diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,324723 @@ +{ + "best_metric": 0.40609586353161314, + "best_model_checkpoint": "checkpoints/misunderstood-oath-216/checkpoint-30000", + "epoch": 3.4742327735958307, + "eval_steps": 1000, + "global_step": 54000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0, + "learning_rate": 0.0, + "loss": 5.3501, + "step": 1 + }, + { + "epoch": 0.0, + "learning_rate": 2.0000000000000002e-07, + "loss": 5.3956, + "step": 2 + }, + { + "epoch": 0.0, + "learning_rate": 4.0000000000000003e-07, + "loss": 5.3784, + "step": 3 + }, + { + "epoch": 0.0, + "learning_rate": 6.000000000000001e-07, + "loss": 5.2865, + "step": 4 + }, + { + "epoch": 0.0, + "learning_rate": 8.000000000000001e-07, + "loss": 5.4317, + "step": 5 + }, + { + "epoch": 0.0, + "learning_rate": 1.0000000000000002e-06, + "loss": 5.1846, + "step": 6 + }, + { + "epoch": 0.0, + "learning_rate": 1.2000000000000002e-06, + "loss": 5.1307, + "step": 7 + }, + { + "epoch": 0.0, + "learning_rate": 1.4000000000000001e-06, + "loss": 5.1022, + "step": 8 + }, + { + "epoch": 0.0, + "learning_rate": 1.4000000000000001e-06, + "loss": 5.0478, + "step": 9 + }, + { + "epoch": 0.0, + "learning_rate": 1.6000000000000001e-06, + "loss": 5.1205, + "step": 10 + }, + { + "epoch": 0.0, + "learning_rate": 1.8000000000000001e-06, + "loss": 5.0833, + "step": 11 + }, + { + "epoch": 0.0, + "learning_rate": 2.0000000000000003e-06, + "loss": 5.0956, + "step": 12 + }, + { + "epoch": 0.0, + "learning_rate": 2.2e-06, + "loss": 4.6399, + "step": 13 + }, + { + "epoch": 0.0, + "learning_rate": 2.4000000000000003e-06, + "loss": 4.6217, + "step": 14 + }, + { + "epoch": 0.0, + "learning_rate": 2.6e-06, + "loss": 4.7764, + "step": 15 + }, + { + "epoch": 0.0, + "learning_rate": 2.8000000000000003e-06, + "loss": 4.6695, + "step": 16 + }, + { + "epoch": 0.0, + "learning_rate": 3e-06, + "loss": 4.4532, + "step": 17 + }, + { + "epoch": 0.0, + "learning_rate": 3.2000000000000003e-06, + "loss": 4.3873, + "step": 18 + }, + { + "epoch": 0.0, + "learning_rate": 3.4000000000000005e-06, + "loss": 4.1236, + "step": 19 + }, + { + "epoch": 0.0, + "learning_rate": 3.6000000000000003e-06, + "loss": 4.1087, + "step": 20 + }, + { + "epoch": 0.0, + "learning_rate": 3.8000000000000005e-06, + "loss": 3.9695, + "step": 21 + }, + { + "epoch": 0.0, + "learning_rate": 4.000000000000001e-06, + "loss": 3.8101, + "step": 22 + }, + { + "epoch": 0.0, + "learning_rate": 4.2000000000000004e-06, + "loss": 3.77, + "step": 23 + }, + { + "epoch": 0.0, + "learning_rate": 4.4e-06, + "loss": 3.6102, + "step": 24 + }, + { + "epoch": 0.0, + "learning_rate": 4.600000000000001e-06, + "loss": 3.4966, + "step": 25 + }, + { + "epoch": 0.0, + "learning_rate": 4.800000000000001e-06, + "loss": 3.314, + "step": 26 + }, + { + "epoch": 0.0, + "learning_rate": 5e-06, + "loss": 3.2284, + "step": 27 + }, + { + "epoch": 0.0, + "learning_rate": 5.2e-06, + "loss": 3.2582, + "step": 28 + }, + { + "epoch": 0.0, + "learning_rate": 5.400000000000001e-06, + "loss": 3.0677, + "step": 29 + }, + { + "epoch": 0.0, + "learning_rate": 5.600000000000001e-06, + "loss": 3.0885, + "step": 30 + }, + { + "epoch": 0.0, + "learning_rate": 5.8e-06, + "loss": 2.9582, + "step": 31 + }, + { + "epoch": 0.0, + "learning_rate": 6e-06, + "loss": 2.7603, + "step": 32 + }, + { + "epoch": 0.0, + "learning_rate": 6.200000000000001e-06, + "loss": 2.7556, + "step": 33 + }, + { + "epoch": 0.0, + "learning_rate": 6.4000000000000006e-06, + "loss": 2.6821, + "step": 34 + }, + { + "epoch": 0.0, + "learning_rate": 6.600000000000001e-06, + "loss": 2.6172, + "step": 35 + }, + { + "epoch": 0.0, + "learning_rate": 6.800000000000001e-06, + "loss": 2.6612, + "step": 36 + }, + { + "epoch": 0.0, + "learning_rate": 7e-06, + "loss": 2.4991, + "step": 37 + }, + { + "epoch": 0.0, + "learning_rate": 7.2000000000000005e-06, + "loss": 2.3305, + "step": 38 + }, + { + "epoch": 0.0, + "learning_rate": 7.4e-06, + "loss": 2.5087, + "step": 39 + }, + { + "epoch": 0.0, + "learning_rate": 7.600000000000001e-06, + "loss": 2.1899, + "step": 40 + }, + { + "epoch": 0.0, + "learning_rate": 7.800000000000002e-06, + "loss": 2.3546, + "step": 41 + }, + { + "epoch": 0.0, + "learning_rate": 8.000000000000001e-06, + "loss": 2.2266, + "step": 42 + }, + { + "epoch": 0.0, + "learning_rate": 8.2e-06, + "loss": 2.1673, + "step": 43 + }, + { + "epoch": 0.0, + "learning_rate": 8.400000000000001e-06, + "loss": 2.2075, + "step": 44 + }, + { + "epoch": 0.0, + "learning_rate": 8.6e-06, + "loss": 2.1283, + "step": 45 + }, + { + "epoch": 0.0, + "learning_rate": 8.8e-06, + "loss": 2.2437, + "step": 46 + }, + { + "epoch": 0.0, + "learning_rate": 9e-06, + "loss": 2.1244, + "step": 47 + }, + { + "epoch": 0.0, + "learning_rate": 9.200000000000002e-06, + "loss": 2.1517, + "step": 48 + }, + { + "epoch": 0.0, + "learning_rate": 9.4e-06, + "loss": 2.1929, + "step": 49 + }, + { + "epoch": 0.0, + "learning_rate": 9.600000000000001e-06, + "loss": 1.9339, + "step": 50 + }, + { + "epoch": 0.0, + "learning_rate": 9.800000000000001e-06, + "loss": 1.9246, + "step": 51 + }, + { + "epoch": 0.0, + "learning_rate": 1e-05, + "loss": 2.0175, + "step": 52 + }, + { + "epoch": 0.0, + "learning_rate": 1.02e-05, + "loss": 1.9385, + "step": 53 + }, + { + "epoch": 0.0, + "learning_rate": 1.04e-05, + "loss": 1.9292, + "step": 54 + }, + { + "epoch": 0.0, + "learning_rate": 1.0600000000000002e-05, + "loss": 1.9037, + "step": 55 + }, + { + "epoch": 0.0, + "learning_rate": 1.0800000000000002e-05, + "loss": 1.9823, + "step": 56 + }, + { + "epoch": 0.0, + "learning_rate": 1.1000000000000001e-05, + "loss": 1.8485, + "step": 57 + }, + { + "epoch": 0.0, + "learning_rate": 1.1200000000000001e-05, + "loss": 1.8657, + "step": 58 + }, + { + "epoch": 0.0, + "learning_rate": 1.14e-05, + "loss": 1.9203, + "step": 59 + }, + { + "epoch": 0.0, + "learning_rate": 1.16e-05, + "loss": 1.873, + "step": 60 + }, + { + "epoch": 0.0, + "learning_rate": 1.18e-05, + "loss": 2.066, + "step": 61 + }, + { + "epoch": 0.0, + "learning_rate": 1.2e-05, + "loss": 1.9169, + "step": 62 + }, + { + "epoch": 0.0, + "learning_rate": 1.22e-05, + "loss": 1.7883, + "step": 63 + }, + { + "epoch": 0.0, + "learning_rate": 1.2400000000000002e-05, + "loss": 1.8737, + "step": 64 + }, + { + "epoch": 0.0, + "learning_rate": 1.2600000000000001e-05, + "loss": 1.8874, + "step": 65 + }, + { + "epoch": 0.0, + "learning_rate": 1.2800000000000001e-05, + "loss": 1.8512, + "step": 66 + }, + { + "epoch": 0.0, + "learning_rate": 1.3000000000000001e-05, + "loss": 1.7456, + "step": 67 + }, + { + "epoch": 0.0, + "learning_rate": 1.3200000000000002e-05, + "loss": 1.7904, + "step": 68 + }, + { + "epoch": 0.0, + "learning_rate": 1.3400000000000002e-05, + "loss": 1.9096, + "step": 69 + }, + { + "epoch": 0.0, + "learning_rate": 1.3600000000000002e-05, + "loss": 1.7047, + "step": 70 + }, + { + "epoch": 0.0, + "learning_rate": 1.38e-05, + "loss": 1.7912, + "step": 71 + }, + { + "epoch": 0.0, + "learning_rate": 1.4e-05, + "loss": 1.8078, + "step": 72 + }, + { + "epoch": 0.0, + "learning_rate": 1.4200000000000001e-05, + "loss": 1.8265, + "step": 73 + }, + { + "epoch": 0.0, + "learning_rate": 1.4400000000000001e-05, + "loss": 1.8332, + "step": 74 + }, + { + "epoch": 0.0, + "learning_rate": 1.46e-05, + "loss": 1.7074, + "step": 75 + }, + { + "epoch": 0.0, + "learning_rate": 1.48e-05, + "loss": 1.8431, + "step": 76 + }, + { + "epoch": 0.0, + "learning_rate": 1.5000000000000002e-05, + "loss": 1.721, + "step": 77 + }, + { + "epoch": 0.01, + "learning_rate": 1.5200000000000002e-05, + "loss": 1.8039, + "step": 78 + }, + { + "epoch": 0.01, + "learning_rate": 1.54e-05, + "loss": 1.8019, + "step": 79 + }, + { + "epoch": 0.01, + "learning_rate": 1.5600000000000003e-05, + "loss": 1.6346, + "step": 80 + }, + { + "epoch": 0.01, + "learning_rate": 1.58e-05, + "loss": 1.6903, + "step": 81 + }, + { + "epoch": 0.01, + "learning_rate": 1.6000000000000003e-05, + "loss": 1.6262, + "step": 82 + }, + { + "epoch": 0.01, + "learning_rate": 1.62e-05, + "loss": 1.7798, + "step": 83 + }, + { + "epoch": 0.01, + "learning_rate": 1.64e-05, + "loss": 1.5539, + "step": 84 + }, + { + "epoch": 0.01, + "learning_rate": 1.66e-05, + "loss": 1.7486, + "step": 85 + }, + { + "epoch": 0.01, + "learning_rate": 1.6800000000000002e-05, + "loss": 1.6887, + "step": 86 + }, + { + "epoch": 0.01, + "learning_rate": 1.7e-05, + "loss": 1.6411, + "step": 87 + }, + { + "epoch": 0.01, + "learning_rate": 1.72e-05, + "loss": 1.5932, + "step": 88 + }, + { + "epoch": 0.01, + "learning_rate": 1.7400000000000003e-05, + "loss": 1.6465, + "step": 89 + }, + { + "epoch": 0.01, + "learning_rate": 1.76e-05, + "loss": 1.5225, + "step": 90 + }, + { + "epoch": 0.01, + "learning_rate": 1.7800000000000002e-05, + "loss": 1.5533, + "step": 91 + }, + { + "epoch": 0.01, + "learning_rate": 1.8e-05, + "loss": 1.6825, + "step": 92 + }, + { + "epoch": 0.01, + "learning_rate": 1.8200000000000002e-05, + "loss": 1.5618, + "step": 93 + }, + { + "epoch": 0.01, + "learning_rate": 1.8400000000000003e-05, + "loss": 1.6228, + "step": 94 + }, + { + "epoch": 0.01, + "learning_rate": 1.86e-05, + "loss": 1.6515, + "step": 95 + }, + { + "epoch": 0.01, + "learning_rate": 1.88e-05, + "loss": 1.6866, + "step": 96 + }, + { + "epoch": 0.01, + "learning_rate": 1.9e-05, + "loss": 1.5618, + "step": 97 + }, + { + "epoch": 0.01, + "learning_rate": 1.9200000000000003e-05, + "loss": 1.5686, + "step": 98 + }, + { + "epoch": 0.01, + "learning_rate": 1.94e-05, + "loss": 1.6962, + "step": 99 + }, + { + "epoch": 0.01, + "learning_rate": 1.9600000000000002e-05, + "loss": 1.6023, + "step": 100 + }, + { + "epoch": 0.01, + "learning_rate": 1.98e-05, + "loss": 1.6147, + "step": 101 + }, + { + "epoch": 0.01, + "learning_rate": 2e-05, + "loss": 1.4402, + "step": 102 + }, + { + "epoch": 0.01, + "learning_rate": 1.999978531097705e-05, + "loss": 1.7844, + "step": 103 + }, + { + "epoch": 0.01, + "learning_rate": 1.99995706219541e-05, + "loss": 1.5232, + "step": 104 + }, + { + "epoch": 0.01, + "learning_rate": 1.9999355932931152e-05, + "loss": 1.4863, + "step": 105 + }, + { + "epoch": 0.01, + "learning_rate": 1.99991412439082e-05, + "loss": 1.5387, + "step": 106 + }, + { + "epoch": 0.01, + "learning_rate": 1.999892655488525e-05, + "loss": 1.4253, + "step": 107 + }, + { + "epoch": 0.01, + "learning_rate": 1.99987118658623e-05, + "loss": 1.5311, + "step": 108 + }, + { + "epoch": 0.01, + "learning_rate": 1.9998497176839348e-05, + "loss": 1.6278, + "step": 109 + }, + { + "epoch": 0.01, + "learning_rate": 1.99982824878164e-05, + "loss": 1.5171, + "step": 110 + }, + { + "epoch": 0.01, + "learning_rate": 1.999806779879345e-05, + "loss": 1.4893, + "step": 111 + }, + { + "epoch": 0.01, + "learning_rate": 1.99978531097705e-05, + "loss": 1.6068, + "step": 112 + }, + { + "epoch": 0.01, + "learning_rate": 1.999763842074755e-05, + "loss": 1.6329, + "step": 113 + }, + { + "epoch": 0.01, + "learning_rate": 1.99974237317246e-05, + "loss": 1.5418, + "step": 114 + }, + { + "epoch": 0.01, + "learning_rate": 1.999720904270165e-05, + "loss": 1.4332, + "step": 115 + }, + { + "epoch": 0.01, + "learning_rate": 1.9996994353678697e-05, + "loss": 1.4685, + "step": 116 + }, + { + "epoch": 0.01, + "learning_rate": 1.9996779664655746e-05, + "loss": 1.4613, + "step": 117 + }, + { + "epoch": 0.01, + "learning_rate": 1.99965649756328e-05, + "loss": 1.477, + "step": 118 + }, + { + "epoch": 0.01, + "learning_rate": 1.9996350286609848e-05, + "loss": 1.6885, + "step": 119 + }, + { + "epoch": 0.01, + "learning_rate": 1.9996135597586897e-05, + "loss": 1.3529, + "step": 120 + }, + { + "epoch": 0.01, + "learning_rate": 1.9995920908563946e-05, + "loss": 1.498, + "step": 121 + }, + { + "epoch": 0.01, + "learning_rate": 1.9995706219540994e-05, + "loss": 1.4229, + "step": 122 + }, + { + "epoch": 0.01, + "learning_rate": 1.9995491530518047e-05, + "loss": 1.5141, + "step": 123 + }, + { + "epoch": 0.01, + "learning_rate": 1.9995276841495096e-05, + "loss": 1.4024, + "step": 124 + }, + { + "epoch": 0.01, + "learning_rate": 1.9995062152472148e-05, + "loss": 1.5359, + "step": 125 + }, + { + "epoch": 0.01, + "learning_rate": 1.9994847463449197e-05, + "loss": 1.3967, + "step": 126 + }, + { + "epoch": 0.01, + "learning_rate": 1.9994632774426246e-05, + "loss": 1.4626, + "step": 127 + }, + { + "epoch": 0.01, + "learning_rate": 1.9994418085403295e-05, + "loss": 1.5078, + "step": 128 + }, + { + "epoch": 0.01, + "learning_rate": 1.9994203396380344e-05, + "loss": 1.4291, + "step": 129 + }, + { + "epoch": 0.01, + "learning_rate": 1.9993988707357393e-05, + "loss": 1.5107, + "step": 130 + }, + { + "epoch": 0.01, + "learning_rate": 1.9993774018334445e-05, + "loss": 1.4166, + "step": 131 + }, + { + "epoch": 0.01, + "learning_rate": 1.9993559329311494e-05, + "loss": 1.5026, + "step": 132 + }, + { + "epoch": 0.01, + "learning_rate": 1.9993344640288543e-05, + "loss": 1.4271, + "step": 133 + }, + { + "epoch": 0.01, + "learning_rate": 1.9993129951265592e-05, + "loss": 1.5964, + "step": 134 + }, + { + "epoch": 0.01, + "learning_rate": 1.999291526224264e-05, + "loss": 1.5907, + "step": 135 + }, + { + "epoch": 0.01, + "learning_rate": 1.9992700573219694e-05, + "loss": 1.429, + "step": 136 + }, + { + "epoch": 0.01, + "learning_rate": 1.9992485884196743e-05, + "loss": 1.4342, + "step": 137 + }, + { + "epoch": 0.01, + "learning_rate": 1.999227119517379e-05, + "loss": 1.5285, + "step": 138 + }, + { + "epoch": 0.01, + "learning_rate": 1.9992056506150844e-05, + "loss": 1.4634, + "step": 139 + }, + { + "epoch": 0.01, + "learning_rate": 1.9991841817127893e-05, + "loss": 1.4943, + "step": 140 + }, + { + "epoch": 0.01, + "learning_rate": 1.9991627128104942e-05, + "loss": 1.4631, + "step": 141 + }, + { + "epoch": 0.01, + "learning_rate": 1.999141243908199e-05, + "loss": 1.6667, + "step": 142 + }, + { + "epoch": 0.01, + "learning_rate": 1.9991197750059043e-05, + "loss": 1.4552, + "step": 143 + }, + { + "epoch": 0.01, + "learning_rate": 1.9990983061036092e-05, + "loss": 1.4297, + "step": 144 + }, + { + "epoch": 0.01, + "learning_rate": 1.999076837201314e-05, + "loss": 1.3702, + "step": 145 + }, + { + "epoch": 0.01, + "learning_rate": 1.999055368299019e-05, + "loss": 1.4936, + "step": 146 + }, + { + "epoch": 0.01, + "learning_rate": 1.999033899396724e-05, + "loss": 1.6323, + "step": 147 + }, + { + "epoch": 0.01, + "learning_rate": 1.9990124304944288e-05, + "loss": 1.469, + "step": 148 + }, + { + "epoch": 0.01, + "learning_rate": 1.998990961592134e-05, + "loss": 1.4835, + "step": 149 + }, + { + "epoch": 0.01, + "learning_rate": 1.998969492689839e-05, + "loss": 1.618, + "step": 150 + }, + { + "epoch": 0.01, + "learning_rate": 1.9989480237875438e-05, + "loss": 1.4658, + "step": 151 + }, + { + "epoch": 0.01, + "learning_rate": 1.9989265548852487e-05, + "loss": 1.4158, + "step": 152 + }, + { + "epoch": 0.01, + "learning_rate": 1.998905085982954e-05, + "loss": 1.4331, + "step": 153 + }, + { + "epoch": 0.01, + "learning_rate": 1.998883617080659e-05, + "loss": 1.3965, + "step": 154 + }, + { + "epoch": 0.01, + "learning_rate": 1.9988621481783637e-05, + "loss": 1.4156, + "step": 155 + }, + { + "epoch": 0.01, + "learning_rate": 1.998840679276069e-05, + "loss": 1.4441, + "step": 156 + }, + { + "epoch": 0.01, + "learning_rate": 1.998819210373774e-05, + "loss": 1.4665, + "step": 157 + }, + { + "epoch": 0.01, + "learning_rate": 1.9987977414714788e-05, + "loss": 1.3424, + "step": 158 + }, + { + "epoch": 0.01, + "learning_rate": 1.9987762725691837e-05, + "loss": 1.4151, + "step": 159 + }, + { + "epoch": 0.01, + "learning_rate": 1.9987548036668886e-05, + "loss": 1.4264, + "step": 160 + }, + { + "epoch": 0.01, + "learning_rate": 1.9987333347645938e-05, + "loss": 1.4506, + "step": 161 + }, + { + "epoch": 0.01, + "learning_rate": 1.9987118658622987e-05, + "loss": 1.5392, + "step": 162 + }, + { + "epoch": 0.01, + "learning_rate": 1.9986903969600036e-05, + "loss": 1.3007, + "step": 163 + }, + { + "epoch": 0.01, + "learning_rate": 1.9986689280577085e-05, + "loss": 1.3759, + "step": 164 + }, + { + "epoch": 0.01, + "learning_rate": 1.9986474591554134e-05, + "loss": 1.5224, + "step": 165 + }, + { + "epoch": 0.01, + "learning_rate": 1.9986259902531183e-05, + "loss": 1.3736, + "step": 166 + }, + { + "epoch": 0.01, + "learning_rate": 1.9986045213508235e-05, + "loss": 1.4525, + "step": 167 + }, + { + "epoch": 0.01, + "learning_rate": 1.9985830524485284e-05, + "loss": 1.4804, + "step": 168 + }, + { + "epoch": 0.01, + "learning_rate": 1.9985615835462336e-05, + "loss": 1.4911, + "step": 169 + }, + { + "epoch": 0.01, + "learning_rate": 1.9985401146439385e-05, + "loss": 1.4254, + "step": 170 + }, + { + "epoch": 0.01, + "learning_rate": 1.9985186457416434e-05, + "loss": 1.4016, + "step": 171 + }, + { + "epoch": 0.01, + "learning_rate": 1.9984971768393483e-05, + "loss": 1.5936, + "step": 172 + }, + { + "epoch": 0.01, + "learning_rate": 1.9984757079370532e-05, + "loss": 1.3892, + "step": 173 + }, + { + "epoch": 0.01, + "learning_rate": 1.9984542390347585e-05, + "loss": 1.3223, + "step": 174 + }, + { + "epoch": 0.01, + "learning_rate": 1.9984327701324634e-05, + "loss": 1.4443, + "step": 175 + }, + { + "epoch": 0.01, + "learning_rate": 1.9984113012301683e-05, + "loss": 1.3711, + "step": 176 + }, + { + "epoch": 0.01, + "learning_rate": 1.998389832327873e-05, + "loss": 1.4154, + "step": 177 + }, + { + "epoch": 0.01, + "learning_rate": 1.998368363425578e-05, + "loss": 1.3584, + "step": 178 + }, + { + "epoch": 0.01, + "learning_rate": 1.998346894523283e-05, + "loss": 1.4978, + "step": 179 + }, + { + "epoch": 0.01, + "learning_rate": 1.9983254256209882e-05, + "loss": 1.4173, + "step": 180 + }, + { + "epoch": 0.01, + "learning_rate": 1.998303956718693e-05, + "loss": 1.4096, + "step": 181 + }, + { + "epoch": 0.01, + "learning_rate": 1.9982824878163983e-05, + "loss": 1.4568, + "step": 182 + }, + { + "epoch": 0.01, + "learning_rate": 1.9982610189141032e-05, + "loss": 1.3597, + "step": 183 + }, + { + "epoch": 0.01, + "learning_rate": 1.998239550011808e-05, + "loss": 1.371, + "step": 184 + }, + { + "epoch": 0.01, + "learning_rate": 1.998218081109513e-05, + "loss": 1.4398, + "step": 185 + }, + { + "epoch": 0.01, + "learning_rate": 1.998196612207218e-05, + "loss": 1.4292, + "step": 186 + }, + { + "epoch": 0.01, + "learning_rate": 1.998175143304923e-05, + "loss": 1.4285, + "step": 187 + }, + { + "epoch": 0.01, + "learning_rate": 1.998153674402628e-05, + "loss": 1.4514, + "step": 188 + }, + { + "epoch": 0.01, + "learning_rate": 1.998132205500333e-05, + "loss": 1.4439, + "step": 189 + }, + { + "epoch": 0.01, + "learning_rate": 1.998110736598038e-05, + "loss": 1.3794, + "step": 190 + }, + { + "epoch": 0.01, + "learning_rate": 1.9980892676957427e-05, + "loss": 1.4907, + "step": 191 + }, + { + "epoch": 0.01, + "learning_rate": 1.998067798793448e-05, + "loss": 1.2773, + "step": 192 + }, + { + "epoch": 0.01, + "learning_rate": 1.998046329891153e-05, + "loss": 1.3374, + "step": 193 + }, + { + "epoch": 0.01, + "learning_rate": 1.9980248609888578e-05, + "loss": 1.4242, + "step": 194 + }, + { + "epoch": 0.01, + "learning_rate": 1.9980033920865627e-05, + "loss": 1.5418, + "step": 195 + }, + { + "epoch": 0.01, + "learning_rate": 1.9979819231842675e-05, + "loss": 1.4763, + "step": 196 + }, + { + "epoch": 0.01, + "learning_rate": 1.9979604542819728e-05, + "loss": 1.3906, + "step": 197 + }, + { + "epoch": 0.01, + "learning_rate": 1.9979389853796777e-05, + "loss": 1.4122, + "step": 198 + }, + { + "epoch": 0.01, + "learning_rate": 1.9979175164773826e-05, + "loss": 1.4258, + "step": 199 + }, + { + "epoch": 0.01, + "learning_rate": 1.9978960475750878e-05, + "loss": 1.3533, + "step": 200 + }, + { + "epoch": 0.01, + "learning_rate": 1.9978745786727927e-05, + "loss": 1.357, + "step": 201 + }, + { + "epoch": 0.01, + "learning_rate": 1.9978531097704976e-05, + "loss": 1.3911, + "step": 202 + }, + { + "epoch": 0.01, + "learning_rate": 1.9978316408682025e-05, + "loss": 1.4728, + "step": 203 + }, + { + "epoch": 0.01, + "learning_rate": 1.9978101719659074e-05, + "loss": 1.3565, + "step": 204 + }, + { + "epoch": 0.01, + "learning_rate": 1.9977887030636126e-05, + "loss": 1.4445, + "step": 205 + }, + { + "epoch": 0.01, + "learning_rate": 1.9977672341613175e-05, + "loss": 1.358, + "step": 206 + }, + { + "epoch": 0.01, + "learning_rate": 1.9977457652590224e-05, + "loss": 1.378, + "step": 207 + }, + { + "epoch": 0.01, + "learning_rate": 1.9977242963567273e-05, + "loss": 1.3419, + "step": 208 + }, + { + "epoch": 0.01, + "learning_rate": 1.9977028274544322e-05, + "loss": 1.3284, + "step": 209 + }, + { + "epoch": 0.01, + "learning_rate": 1.9976813585521375e-05, + "loss": 1.3453, + "step": 210 + }, + { + "epoch": 0.01, + "learning_rate": 1.9976598896498424e-05, + "loss": 1.4048, + "step": 211 + }, + { + "epoch": 0.01, + "learning_rate": 1.9976384207475476e-05, + "loss": 1.3382, + "step": 212 + }, + { + "epoch": 0.01, + "learning_rate": 1.9976169518452525e-05, + "loss": 1.4272, + "step": 213 + }, + { + "epoch": 0.01, + "learning_rate": 1.9975954829429574e-05, + "loss": 1.45, + "step": 214 + }, + { + "epoch": 0.01, + "learning_rate": 1.9975740140406623e-05, + "loss": 1.3386, + "step": 215 + }, + { + "epoch": 0.01, + "learning_rate": 1.9975525451383672e-05, + "loss": 1.3179, + "step": 216 + }, + { + "epoch": 0.01, + "learning_rate": 1.997531076236072e-05, + "loss": 1.3377, + "step": 217 + }, + { + "epoch": 0.01, + "learning_rate": 1.9975096073337773e-05, + "loss": 1.359, + "step": 218 + }, + { + "epoch": 0.01, + "learning_rate": 1.9974881384314822e-05, + "loss": 1.4262, + "step": 219 + }, + { + "epoch": 0.01, + "learning_rate": 1.997466669529187e-05, + "loss": 1.4015, + "step": 220 + }, + { + "epoch": 0.01, + "learning_rate": 1.997445200626892e-05, + "loss": 1.3431, + "step": 221 + }, + { + "epoch": 0.01, + "learning_rate": 1.997423731724597e-05, + "loss": 1.3978, + "step": 222 + }, + { + "epoch": 0.01, + "learning_rate": 1.997402262822302e-05, + "loss": 1.2628, + "step": 223 + }, + { + "epoch": 0.01, + "learning_rate": 1.997380793920007e-05, + "loss": 1.3724, + "step": 224 + }, + { + "epoch": 0.01, + "learning_rate": 1.997359325017712e-05, + "loss": 1.319, + "step": 225 + }, + { + "epoch": 0.01, + "learning_rate": 1.997337856115417e-05, + "loss": 1.3899, + "step": 226 + }, + { + "epoch": 0.01, + "learning_rate": 1.997316387213122e-05, + "loss": 1.3299, + "step": 227 + }, + { + "epoch": 0.01, + "learning_rate": 1.997294918310827e-05, + "loss": 1.4814, + "step": 228 + }, + { + "epoch": 0.01, + "learning_rate": 1.997273449408532e-05, + "loss": 1.2986, + "step": 229 + }, + { + "epoch": 0.01, + "learning_rate": 1.997251980506237e-05, + "loss": 1.4916, + "step": 230 + }, + { + "epoch": 0.01, + "learning_rate": 1.997230511603942e-05, + "loss": 1.3971, + "step": 231 + }, + { + "epoch": 0.01, + "learning_rate": 1.997209042701647e-05, + "loss": 1.4418, + "step": 232 + }, + { + "epoch": 0.01, + "learning_rate": 1.9971875737993518e-05, + "loss": 1.361, + "step": 233 + }, + { + "epoch": 0.02, + "learning_rate": 1.9971661048970567e-05, + "loss": 1.3601, + "step": 234 + }, + { + "epoch": 0.02, + "learning_rate": 1.9971446359947616e-05, + "loss": 1.3434, + "step": 235 + }, + { + "epoch": 0.02, + "learning_rate": 1.9971231670924668e-05, + "loss": 1.4039, + "step": 236 + }, + { + "epoch": 0.02, + "learning_rate": 1.9971016981901717e-05, + "loss": 1.3295, + "step": 237 + }, + { + "epoch": 0.02, + "learning_rate": 1.9970802292878766e-05, + "loss": 1.3892, + "step": 238 + }, + { + "epoch": 0.02, + "learning_rate": 1.9970587603855815e-05, + "loss": 1.3852, + "step": 239 + }, + { + "epoch": 0.02, + "learning_rate": 1.9970372914832867e-05, + "loss": 1.4076, + "step": 240 + }, + { + "epoch": 0.02, + "learning_rate": 1.9970158225809916e-05, + "loss": 1.4299, + "step": 241 + }, + { + "epoch": 0.02, + "learning_rate": 1.9969943536786965e-05, + "loss": 1.4072, + "step": 242 + }, + { + "epoch": 0.02, + "learning_rate": 1.9969728847764018e-05, + "loss": 1.3068, + "step": 243 + }, + { + "epoch": 0.02, + "learning_rate": 1.9969514158741066e-05, + "loss": 1.4004, + "step": 244 + }, + { + "epoch": 0.02, + "learning_rate": 1.9969299469718115e-05, + "loss": 1.341, + "step": 245 + }, + { + "epoch": 0.02, + "learning_rate": 1.9969084780695164e-05, + "loss": 1.3683, + "step": 246 + }, + { + "epoch": 0.02, + "learning_rate": 1.9968870091672213e-05, + "loss": 1.3611, + "step": 247 + }, + { + "epoch": 0.02, + "learning_rate": 1.9968655402649266e-05, + "loss": 1.2452, + "step": 248 + }, + { + "epoch": 0.02, + "learning_rate": 1.9968440713626315e-05, + "loss": 1.3642, + "step": 249 + }, + { + "epoch": 0.02, + "learning_rate": 1.9968226024603364e-05, + "loss": 1.3395, + "step": 250 + }, + { + "epoch": 0.02, + "learning_rate": 1.9968011335580413e-05, + "loss": 1.2755, + "step": 251 + }, + { + "epoch": 0.02, + "learning_rate": 1.996779664655746e-05, + "loss": 1.3234, + "step": 252 + }, + { + "epoch": 0.02, + "learning_rate": 1.996758195753451e-05, + "loss": 1.4413, + "step": 253 + }, + { + "epoch": 0.02, + "learning_rate": 1.9967367268511563e-05, + "loss": 1.3231, + "step": 254 + }, + { + "epoch": 0.02, + "learning_rate": 1.9967152579488612e-05, + "loss": 1.3214, + "step": 255 + }, + { + "epoch": 0.02, + "learning_rate": 1.9966937890465664e-05, + "loss": 1.2221, + "step": 256 + }, + { + "epoch": 0.02, + "learning_rate": 1.9966723201442713e-05, + "loss": 1.3871, + "step": 257 + }, + { + "epoch": 0.02, + "learning_rate": 1.9966508512419762e-05, + "loss": 1.4322, + "step": 258 + }, + { + "epoch": 0.02, + "learning_rate": 1.996629382339681e-05, + "loss": 1.2531, + "step": 259 + }, + { + "epoch": 0.02, + "learning_rate": 1.996607913437386e-05, + "loss": 1.4322, + "step": 260 + }, + { + "epoch": 0.02, + "learning_rate": 1.9965864445350912e-05, + "loss": 1.3311, + "step": 261 + }, + { + "epoch": 0.02, + "learning_rate": 1.996564975632796e-05, + "loss": 1.3599, + "step": 262 + }, + { + "epoch": 0.02, + "learning_rate": 1.996543506730501e-05, + "loss": 1.3773, + "step": 263 + }, + { + "epoch": 0.02, + "learning_rate": 1.996522037828206e-05, + "loss": 1.3567, + "step": 264 + }, + { + "epoch": 0.02, + "learning_rate": 1.9965005689259108e-05, + "loss": 1.348, + "step": 265 + }, + { + "epoch": 0.02, + "learning_rate": 1.9964791000236157e-05, + "loss": 1.3334, + "step": 266 + }, + { + "epoch": 0.02, + "learning_rate": 1.996457631121321e-05, + "loss": 1.4055, + "step": 267 + }, + { + "epoch": 0.02, + "learning_rate": 1.996436162219026e-05, + "loss": 1.425, + "step": 268 + }, + { + "epoch": 0.02, + "learning_rate": 1.996414693316731e-05, + "loss": 1.3252, + "step": 269 + }, + { + "epoch": 0.02, + "learning_rate": 1.996393224414436e-05, + "loss": 1.4151, + "step": 270 + }, + { + "epoch": 0.02, + "learning_rate": 1.996371755512141e-05, + "loss": 1.2307, + "step": 271 + }, + { + "epoch": 0.02, + "learning_rate": 1.9963502866098458e-05, + "loss": 1.3723, + "step": 272 + }, + { + "epoch": 0.02, + "learning_rate": 1.9963288177075507e-05, + "loss": 1.3097, + "step": 273 + }, + { + "epoch": 0.02, + "learning_rate": 1.996307348805256e-05, + "loss": 1.1767, + "step": 274 + }, + { + "epoch": 0.02, + "learning_rate": 1.9962858799029608e-05, + "loss": 1.3731, + "step": 275 + }, + { + "epoch": 0.02, + "learning_rate": 1.9962644110006657e-05, + "loss": 1.3506, + "step": 276 + }, + { + "epoch": 0.02, + "learning_rate": 1.9962429420983706e-05, + "loss": 1.2651, + "step": 277 + }, + { + "epoch": 0.02, + "learning_rate": 1.9962214731960755e-05, + "loss": 1.3173, + "step": 278 + }, + { + "epoch": 0.02, + "learning_rate": 1.9962000042937807e-05, + "loss": 1.2482, + "step": 279 + }, + { + "epoch": 0.02, + "learning_rate": 1.9961785353914856e-05, + "loss": 1.2631, + "step": 280 + }, + { + "epoch": 0.02, + "learning_rate": 1.9961570664891905e-05, + "loss": 1.3574, + "step": 281 + }, + { + "epoch": 0.02, + "learning_rate": 1.9961355975868954e-05, + "loss": 1.3976, + "step": 282 + }, + { + "epoch": 0.02, + "learning_rate": 1.9961141286846003e-05, + "loss": 1.3424, + "step": 283 + }, + { + "epoch": 0.02, + "learning_rate": 1.9960926597823056e-05, + "loss": 1.3342, + "step": 284 + }, + { + "epoch": 0.02, + "learning_rate": 1.9960711908800105e-05, + "loss": 1.29, + "step": 285 + }, + { + "epoch": 0.02, + "learning_rate": 1.9960497219777153e-05, + "loss": 1.384, + "step": 286 + }, + { + "epoch": 0.02, + "learning_rate": 1.9960282530754206e-05, + "loss": 1.3628, + "step": 287 + }, + { + "epoch": 0.02, + "learning_rate": 1.9960067841731255e-05, + "loss": 1.3474, + "step": 288 + }, + { + "epoch": 0.02, + "learning_rate": 1.9959853152708304e-05, + "loss": 1.3927, + "step": 289 + }, + { + "epoch": 0.02, + "learning_rate": 1.9959638463685353e-05, + "loss": 1.3593, + "step": 290 + }, + { + "epoch": 0.02, + "learning_rate": 1.9959423774662402e-05, + "loss": 1.259, + "step": 291 + }, + { + "epoch": 0.02, + "learning_rate": 1.9959209085639454e-05, + "loss": 1.2861, + "step": 292 + }, + { + "epoch": 0.02, + "learning_rate": 1.9958994396616503e-05, + "loss": 1.3795, + "step": 293 + }, + { + "epoch": 0.02, + "learning_rate": 1.9958779707593552e-05, + "loss": 1.4071, + "step": 294 + }, + { + "epoch": 0.02, + "learning_rate": 1.99585650185706e-05, + "loss": 1.2824, + "step": 295 + }, + { + "epoch": 0.02, + "learning_rate": 1.995835032954765e-05, + "loss": 1.1944, + "step": 296 + }, + { + "epoch": 0.02, + "learning_rate": 1.9958135640524702e-05, + "loss": 1.3398, + "step": 297 + }, + { + "epoch": 0.02, + "learning_rate": 1.995792095150175e-05, + "loss": 1.3513, + "step": 298 + }, + { + "epoch": 0.02, + "learning_rate": 1.9957706262478804e-05, + "loss": 1.297, + "step": 299 + }, + { + "epoch": 0.02, + "learning_rate": 1.9957491573455853e-05, + "loss": 1.3472, + "step": 300 + }, + { + "epoch": 0.02, + "learning_rate": 1.99572768844329e-05, + "loss": 1.3623, + "step": 301 + }, + { + "epoch": 0.02, + "learning_rate": 1.995706219540995e-05, + "loss": 1.3669, + "step": 302 + }, + { + "epoch": 0.02, + "learning_rate": 1.9956847506387e-05, + "loss": 1.3961, + "step": 303 + }, + { + "epoch": 0.02, + "learning_rate": 1.995663281736405e-05, + "loss": 1.2239, + "step": 304 + }, + { + "epoch": 0.02, + "learning_rate": 1.99564181283411e-05, + "loss": 1.3374, + "step": 305 + }, + { + "epoch": 0.02, + "learning_rate": 1.995620343931815e-05, + "loss": 1.3295, + "step": 306 + }, + { + "epoch": 0.02, + "learning_rate": 1.99559887502952e-05, + "loss": 1.2517, + "step": 307 + }, + { + "epoch": 0.02, + "learning_rate": 1.9955774061272248e-05, + "loss": 1.2219, + "step": 308 + }, + { + "epoch": 0.02, + "learning_rate": 1.9955559372249297e-05, + "loss": 1.3443, + "step": 309 + }, + { + "epoch": 0.02, + "learning_rate": 1.995534468322635e-05, + "loss": 1.439, + "step": 310 + }, + { + "epoch": 0.02, + "learning_rate": 1.9955129994203398e-05, + "loss": 1.4139, + "step": 311 + }, + { + "epoch": 0.02, + "learning_rate": 1.9954915305180447e-05, + "loss": 1.3497, + "step": 312 + }, + { + "epoch": 0.02, + "learning_rate": 1.99547006161575e-05, + "loss": 1.3312, + "step": 313 + }, + { + "epoch": 0.02, + "learning_rate": 1.9954485927134548e-05, + "loss": 1.2666, + "step": 314 + }, + { + "epoch": 0.02, + "learning_rate": 1.9954271238111597e-05, + "loss": 1.1807, + "step": 315 + }, + { + "epoch": 0.02, + "learning_rate": 1.9954056549088646e-05, + "loss": 1.3662, + "step": 316 + }, + { + "epoch": 0.02, + "learning_rate": 1.99538418600657e-05, + "loss": 1.2285, + "step": 317 + }, + { + "epoch": 0.02, + "learning_rate": 1.9953627171042747e-05, + "loss": 1.2805, + "step": 318 + }, + { + "epoch": 0.02, + "learning_rate": 1.9953412482019796e-05, + "loss": 1.2967, + "step": 319 + }, + { + "epoch": 0.02, + "learning_rate": 1.9953197792996845e-05, + "loss": 1.2994, + "step": 320 + }, + { + "epoch": 0.02, + "learning_rate": 1.9952983103973894e-05, + "loss": 1.374, + "step": 321 + }, + { + "epoch": 0.02, + "learning_rate": 1.9952768414950943e-05, + "loss": 1.4242, + "step": 322 + }, + { + "epoch": 0.02, + "learning_rate": 1.9952553725927996e-05, + "loss": 1.2158, + "step": 323 + }, + { + "epoch": 0.02, + "learning_rate": 1.9952339036905045e-05, + "loss": 1.2927, + "step": 324 + }, + { + "epoch": 0.02, + "learning_rate": 1.9952124347882094e-05, + "loss": 1.3916, + "step": 325 + }, + { + "epoch": 0.02, + "learning_rate": 1.9951909658859143e-05, + "loss": 1.3626, + "step": 326 + }, + { + "epoch": 0.02, + "learning_rate": 1.995169496983619e-05, + "loss": 1.4873, + "step": 327 + }, + { + "epoch": 0.02, + "learning_rate": 1.9951480280813244e-05, + "loss": 1.3687, + "step": 328 + }, + { + "epoch": 0.02, + "learning_rate": 1.9951265591790293e-05, + "loss": 1.372, + "step": 329 + }, + { + "epoch": 0.02, + "learning_rate": 1.9951050902767345e-05, + "loss": 1.304, + "step": 330 + }, + { + "epoch": 0.02, + "learning_rate": 1.9950836213744394e-05, + "loss": 1.3708, + "step": 331 + }, + { + "epoch": 0.02, + "learning_rate": 1.9950621524721443e-05, + "loss": 1.3745, + "step": 332 + }, + { + "epoch": 0.02, + "learning_rate": 1.9950406835698492e-05, + "loss": 1.3497, + "step": 333 + }, + { + "epoch": 0.02, + "learning_rate": 1.995019214667554e-05, + "loss": 1.2521, + "step": 334 + }, + { + "epoch": 0.02, + "learning_rate": 1.9949977457652593e-05, + "loss": 1.2898, + "step": 335 + }, + { + "epoch": 0.02, + "learning_rate": 1.9949762768629642e-05, + "loss": 1.483, + "step": 336 + }, + { + "epoch": 0.02, + "learning_rate": 1.994954807960669e-05, + "loss": 1.2406, + "step": 337 + }, + { + "epoch": 0.02, + "learning_rate": 1.994933339058374e-05, + "loss": 1.4686, + "step": 338 + }, + { + "epoch": 0.02, + "learning_rate": 1.994911870156079e-05, + "loss": 1.4571, + "step": 339 + }, + { + "epoch": 0.02, + "learning_rate": 1.9948904012537838e-05, + "loss": 1.2849, + "step": 340 + }, + { + "epoch": 0.02, + "learning_rate": 1.994868932351489e-05, + "loss": 1.3619, + "step": 341 + }, + { + "epoch": 0.02, + "learning_rate": 1.994847463449194e-05, + "loss": 1.2946, + "step": 342 + }, + { + "epoch": 0.02, + "learning_rate": 1.9948259945468992e-05, + "loss": 1.2735, + "step": 343 + }, + { + "epoch": 0.02, + "learning_rate": 1.994804525644604e-05, + "loss": 1.2837, + "step": 344 + }, + { + "epoch": 0.02, + "learning_rate": 1.994783056742309e-05, + "loss": 1.3459, + "step": 345 + }, + { + "epoch": 0.02, + "learning_rate": 1.994761587840014e-05, + "loss": 1.318, + "step": 346 + }, + { + "epoch": 0.02, + "learning_rate": 1.9947401189377188e-05, + "loss": 1.2803, + "step": 347 + }, + { + "epoch": 0.02, + "learning_rate": 1.994718650035424e-05, + "loss": 1.248, + "step": 348 + }, + { + "epoch": 0.02, + "learning_rate": 1.994697181133129e-05, + "loss": 1.3219, + "step": 349 + }, + { + "epoch": 0.02, + "learning_rate": 1.9946757122308338e-05, + "loss": 1.4075, + "step": 350 + }, + { + "epoch": 0.02, + "learning_rate": 1.9946542433285387e-05, + "loss": 1.3157, + "step": 351 + }, + { + "epoch": 0.02, + "learning_rate": 1.9946327744262436e-05, + "loss": 1.2686, + "step": 352 + }, + { + "epoch": 0.02, + "learning_rate": 1.9946113055239485e-05, + "loss": 1.3433, + "step": 353 + }, + { + "epoch": 0.02, + "learning_rate": 1.9945898366216537e-05, + "loss": 1.252, + "step": 354 + }, + { + "epoch": 0.02, + "learning_rate": 1.9945683677193586e-05, + "loss": 1.2813, + "step": 355 + }, + { + "epoch": 0.02, + "learning_rate": 1.9945468988170635e-05, + "loss": 1.4863, + "step": 356 + }, + { + "epoch": 0.02, + "learning_rate": 1.9945254299147688e-05, + "loss": 1.2543, + "step": 357 + }, + { + "epoch": 0.02, + "learning_rate": 1.9945039610124737e-05, + "loss": 1.2229, + "step": 358 + }, + { + "epoch": 0.02, + "learning_rate": 1.9944824921101786e-05, + "loss": 1.3595, + "step": 359 + }, + { + "epoch": 0.02, + "learning_rate": 1.9944610232078834e-05, + "loss": 1.3063, + "step": 360 + }, + { + "epoch": 0.02, + "learning_rate": 1.9944395543055887e-05, + "loss": 1.3464, + "step": 361 + }, + { + "epoch": 0.02, + "learning_rate": 1.9944180854032936e-05, + "loss": 1.2666, + "step": 362 + }, + { + "epoch": 0.02, + "learning_rate": 1.9943966165009985e-05, + "loss": 1.3662, + "step": 363 + }, + { + "epoch": 0.02, + "learning_rate": 1.9943751475987034e-05, + "loss": 1.3413, + "step": 364 + }, + { + "epoch": 0.02, + "learning_rate": 1.9943536786964083e-05, + "loss": 1.5005, + "step": 365 + }, + { + "epoch": 0.02, + "learning_rate": 1.9943322097941135e-05, + "loss": 1.3393, + "step": 366 + }, + { + "epoch": 0.02, + "learning_rate": 1.9943107408918184e-05, + "loss": 1.2843, + "step": 367 + }, + { + "epoch": 0.02, + "learning_rate": 1.9942892719895233e-05, + "loss": 1.2588, + "step": 368 + }, + { + "epoch": 0.02, + "learning_rate": 1.9942678030872282e-05, + "loss": 1.248, + "step": 369 + }, + { + "epoch": 0.02, + "learning_rate": 1.994246334184933e-05, + "loss": 1.3741, + "step": 370 + }, + { + "epoch": 0.02, + "learning_rate": 1.9942248652826383e-05, + "loss": 1.2721, + "step": 371 + }, + { + "epoch": 0.02, + "learning_rate": 1.9942033963803432e-05, + "loss": 1.2413, + "step": 372 + }, + { + "epoch": 0.02, + "learning_rate": 1.994181927478048e-05, + "loss": 1.2692, + "step": 373 + }, + { + "epoch": 0.02, + "learning_rate": 1.9941604585757534e-05, + "loss": 1.2846, + "step": 374 + }, + { + "epoch": 0.02, + "learning_rate": 1.9941389896734583e-05, + "loss": 1.3197, + "step": 375 + }, + { + "epoch": 0.02, + "learning_rate": 1.994117520771163e-05, + "loss": 1.3236, + "step": 376 + }, + { + "epoch": 0.02, + "learning_rate": 1.994096051868868e-05, + "loss": 1.3173, + "step": 377 + }, + { + "epoch": 0.02, + "learning_rate": 1.994074582966573e-05, + "loss": 1.2561, + "step": 378 + }, + { + "epoch": 0.02, + "learning_rate": 1.9940531140642782e-05, + "loss": 1.3395, + "step": 379 + }, + { + "epoch": 0.02, + "learning_rate": 1.994031645161983e-05, + "loss": 1.2797, + "step": 380 + }, + { + "epoch": 0.02, + "learning_rate": 1.994010176259688e-05, + "loss": 1.3405, + "step": 381 + }, + { + "epoch": 0.02, + "learning_rate": 1.993988707357393e-05, + "loss": 1.3165, + "step": 382 + }, + { + "epoch": 0.02, + "learning_rate": 1.9939672384550978e-05, + "loss": 1.3632, + "step": 383 + }, + { + "epoch": 0.02, + "learning_rate": 1.993945769552803e-05, + "loss": 1.2895, + "step": 384 + }, + { + "epoch": 0.02, + "learning_rate": 1.993924300650508e-05, + "loss": 1.2605, + "step": 385 + }, + { + "epoch": 0.02, + "learning_rate": 1.993902831748213e-05, + "loss": 1.3502, + "step": 386 + }, + { + "epoch": 0.02, + "learning_rate": 1.993881362845918e-05, + "loss": 1.2443, + "step": 387 + }, + { + "epoch": 0.02, + "learning_rate": 1.993859893943623e-05, + "loss": 1.3325, + "step": 388 + }, + { + "epoch": 0.03, + "learning_rate": 1.9938384250413278e-05, + "loss": 1.3504, + "step": 389 + }, + { + "epoch": 0.03, + "learning_rate": 1.9938169561390327e-05, + "loss": 1.2418, + "step": 390 + }, + { + "epoch": 0.03, + "learning_rate": 1.9937954872367376e-05, + "loss": 1.3168, + "step": 391 + }, + { + "epoch": 0.03, + "learning_rate": 1.993774018334443e-05, + "loss": 1.2282, + "step": 392 + }, + { + "epoch": 0.03, + "learning_rate": 1.9937525494321477e-05, + "loss": 1.2458, + "step": 393 + }, + { + "epoch": 0.03, + "learning_rate": 1.9937310805298526e-05, + "loss": 1.3633, + "step": 394 + }, + { + "epoch": 0.03, + "learning_rate": 1.9937096116275575e-05, + "loss": 1.2598, + "step": 395 + }, + { + "epoch": 0.03, + "learning_rate": 1.9936881427252624e-05, + "loss": 1.3485, + "step": 396 + }, + { + "epoch": 0.03, + "learning_rate": 1.9936666738229677e-05, + "loss": 1.3879, + "step": 397 + }, + { + "epoch": 0.03, + "learning_rate": 1.9936452049206726e-05, + "loss": 1.3254, + "step": 398 + }, + { + "epoch": 0.03, + "learning_rate": 1.9936237360183775e-05, + "loss": 1.3001, + "step": 399 + }, + { + "epoch": 0.03, + "learning_rate": 1.9936022671160827e-05, + "loss": 1.3218, + "step": 400 + }, + { + "epoch": 0.03, + "learning_rate": 1.9935807982137876e-05, + "loss": 1.2857, + "step": 401 + }, + { + "epoch": 0.03, + "learning_rate": 1.9935593293114925e-05, + "loss": 1.2542, + "step": 402 + }, + { + "epoch": 0.03, + "learning_rate": 1.9935378604091974e-05, + "loss": 1.2277, + "step": 403 + }, + { + "epoch": 0.03, + "learning_rate": 1.9935163915069026e-05, + "loss": 1.3082, + "step": 404 + }, + { + "epoch": 0.03, + "learning_rate": 1.9934949226046075e-05, + "loss": 1.3334, + "step": 405 + }, + { + "epoch": 0.03, + "learning_rate": 1.9934734537023124e-05, + "loss": 1.3246, + "step": 406 + }, + { + "epoch": 0.03, + "learning_rate": 1.9934519848000173e-05, + "loss": 1.254, + "step": 407 + }, + { + "epoch": 0.03, + "learning_rate": 1.9934305158977222e-05, + "loss": 1.2494, + "step": 408 + }, + { + "epoch": 0.03, + "learning_rate": 1.993409046995427e-05, + "loss": 1.372, + "step": 409 + }, + { + "epoch": 0.03, + "learning_rate": 1.9933875780931323e-05, + "loss": 1.1959, + "step": 410 + }, + { + "epoch": 0.03, + "learning_rate": 1.9933661091908372e-05, + "loss": 1.3262, + "step": 411 + }, + { + "epoch": 0.03, + "learning_rate": 1.993344640288542e-05, + "loss": 1.402, + "step": 412 + }, + { + "epoch": 0.03, + "learning_rate": 1.993323171386247e-05, + "loss": 1.31, + "step": 413 + }, + { + "epoch": 0.03, + "learning_rate": 1.993301702483952e-05, + "loss": 1.2797, + "step": 414 + }, + { + "epoch": 0.03, + "learning_rate": 1.993280233581657e-05, + "loss": 1.3126, + "step": 415 + }, + { + "epoch": 0.03, + "learning_rate": 1.993258764679362e-05, + "loss": 1.1859, + "step": 416 + }, + { + "epoch": 0.03, + "learning_rate": 1.9932372957770673e-05, + "loss": 1.3581, + "step": 417 + }, + { + "epoch": 0.03, + "learning_rate": 1.9932158268747722e-05, + "loss": 1.306, + "step": 418 + }, + { + "epoch": 0.03, + "learning_rate": 1.993194357972477e-05, + "loss": 1.3369, + "step": 419 + }, + { + "epoch": 0.03, + "learning_rate": 1.993172889070182e-05, + "loss": 1.2883, + "step": 420 + }, + { + "epoch": 0.03, + "learning_rate": 1.993151420167887e-05, + "loss": 1.2809, + "step": 421 + }, + { + "epoch": 0.03, + "learning_rate": 1.993129951265592e-05, + "loss": 1.3735, + "step": 422 + }, + { + "epoch": 0.03, + "learning_rate": 1.993108482363297e-05, + "loss": 1.346, + "step": 423 + }, + { + "epoch": 0.03, + "learning_rate": 1.993087013461002e-05, + "loss": 1.3647, + "step": 424 + }, + { + "epoch": 0.03, + "learning_rate": 1.9930655445587068e-05, + "loss": 1.3653, + "step": 425 + }, + { + "epoch": 0.03, + "learning_rate": 1.9930440756564117e-05, + "loss": 1.1652, + "step": 426 + }, + { + "epoch": 0.03, + "learning_rate": 1.9930226067541166e-05, + "loss": 1.284, + "step": 427 + }, + { + "epoch": 0.03, + "learning_rate": 1.993001137851822e-05, + "loss": 1.443, + "step": 428 + }, + { + "epoch": 0.03, + "learning_rate": 1.9929796689495267e-05, + "loss": 1.304, + "step": 429 + }, + { + "epoch": 0.03, + "learning_rate": 1.992958200047232e-05, + "loss": 1.3005, + "step": 430 + }, + { + "epoch": 0.03, + "learning_rate": 1.992936731144937e-05, + "loss": 1.3678, + "step": 431 + }, + { + "epoch": 0.03, + "learning_rate": 1.9929152622426418e-05, + "loss": 1.3621, + "step": 432 + }, + { + "epoch": 0.03, + "learning_rate": 1.9928937933403467e-05, + "loss": 1.2992, + "step": 433 + }, + { + "epoch": 0.03, + "learning_rate": 1.9928723244380515e-05, + "loss": 1.2106, + "step": 434 + }, + { + "epoch": 0.03, + "learning_rate": 1.9928508555357568e-05, + "loss": 1.2285, + "step": 435 + }, + { + "epoch": 0.03, + "learning_rate": 1.9928293866334617e-05, + "loss": 1.3217, + "step": 436 + }, + { + "epoch": 0.03, + "learning_rate": 1.9928079177311666e-05, + "loss": 1.3803, + "step": 437 + }, + { + "epoch": 0.03, + "learning_rate": 1.9927864488288715e-05, + "loss": 1.2312, + "step": 438 + }, + { + "epoch": 0.03, + "learning_rate": 1.9927649799265764e-05, + "loss": 1.3215, + "step": 439 + }, + { + "epoch": 0.03, + "learning_rate": 1.9927435110242813e-05, + "loss": 1.2913, + "step": 440 + }, + { + "epoch": 0.03, + "learning_rate": 1.9927220421219865e-05, + "loss": 1.1649, + "step": 441 + }, + { + "epoch": 0.03, + "learning_rate": 1.9927005732196914e-05, + "loss": 1.1652, + "step": 442 + }, + { + "epoch": 0.03, + "learning_rate": 1.9926791043173963e-05, + "loss": 1.2619, + "step": 443 + }, + { + "epoch": 0.03, + "learning_rate": 1.9926576354151015e-05, + "loss": 1.3073, + "step": 444 + }, + { + "epoch": 0.03, + "learning_rate": 1.9926361665128064e-05, + "loss": 1.3124, + "step": 445 + }, + { + "epoch": 0.03, + "learning_rate": 1.9926146976105113e-05, + "loss": 1.292, + "step": 446 + }, + { + "epoch": 0.03, + "learning_rate": 1.9925932287082162e-05, + "loss": 1.2435, + "step": 447 + }, + { + "epoch": 0.03, + "learning_rate": 1.9925717598059215e-05, + "loss": 1.4361, + "step": 448 + }, + { + "epoch": 0.03, + "learning_rate": 1.9925502909036264e-05, + "loss": 1.2515, + "step": 449 + }, + { + "epoch": 0.03, + "learning_rate": 1.9925288220013312e-05, + "loss": 1.2867, + "step": 450 + }, + { + "epoch": 0.03, + "learning_rate": 1.992507353099036e-05, + "loss": 1.2464, + "step": 451 + }, + { + "epoch": 0.03, + "learning_rate": 1.992485884196741e-05, + "loss": 1.2665, + "step": 452 + }, + { + "epoch": 0.03, + "learning_rate": 1.9924644152944463e-05, + "loss": 1.3577, + "step": 453 + }, + { + "epoch": 0.03, + "learning_rate": 1.9924429463921512e-05, + "loss": 1.4195, + "step": 454 + }, + { + "epoch": 0.03, + "learning_rate": 1.992421477489856e-05, + "loss": 1.3266, + "step": 455 + }, + { + "epoch": 0.03, + "learning_rate": 1.992400008587561e-05, + "loss": 1.3045, + "step": 456 + }, + { + "epoch": 0.03, + "learning_rate": 1.992378539685266e-05, + "loss": 1.3171, + "step": 457 + }, + { + "epoch": 0.03, + "learning_rate": 1.992357070782971e-05, + "loss": 1.2999, + "step": 458 + }, + { + "epoch": 0.03, + "learning_rate": 1.992335601880676e-05, + "loss": 1.2467, + "step": 459 + }, + { + "epoch": 0.03, + "learning_rate": 1.992314132978381e-05, + "loss": 1.4044, + "step": 460 + }, + { + "epoch": 0.03, + "learning_rate": 1.992292664076086e-05, + "loss": 1.2309, + "step": 461 + }, + { + "epoch": 0.03, + "learning_rate": 1.992271195173791e-05, + "loss": 1.2554, + "step": 462 + }, + { + "epoch": 0.03, + "learning_rate": 1.992249726271496e-05, + "loss": 1.3982, + "step": 463 + }, + { + "epoch": 0.03, + "learning_rate": 1.9922282573692008e-05, + "loss": 1.337, + "step": 464 + }, + { + "epoch": 0.03, + "learning_rate": 1.9922067884669057e-05, + "loss": 1.2545, + "step": 465 + }, + { + "epoch": 0.03, + "learning_rate": 1.992185319564611e-05, + "loss": 1.2857, + "step": 466 + }, + { + "epoch": 0.03, + "learning_rate": 1.992163850662316e-05, + "loss": 1.2854, + "step": 467 + }, + { + "epoch": 0.03, + "learning_rate": 1.9921423817600207e-05, + "loss": 1.2823, + "step": 468 + }, + { + "epoch": 0.03, + "learning_rate": 1.9921209128577256e-05, + "loss": 1.2136, + "step": 469 + }, + { + "epoch": 0.03, + "learning_rate": 1.9920994439554305e-05, + "loss": 1.3617, + "step": 470 + }, + { + "epoch": 0.03, + "learning_rate": 1.9920779750531358e-05, + "loss": 1.2488, + "step": 471 + }, + { + "epoch": 0.03, + "learning_rate": 1.9920565061508407e-05, + "loss": 1.1286, + "step": 472 + }, + { + "epoch": 0.03, + "learning_rate": 1.992035037248546e-05, + "loss": 1.3493, + "step": 473 + }, + { + "epoch": 0.03, + "learning_rate": 1.9920135683462508e-05, + "loss": 1.2387, + "step": 474 + }, + { + "epoch": 0.03, + "learning_rate": 1.9919920994439557e-05, + "loss": 1.2913, + "step": 475 + }, + { + "epoch": 0.03, + "learning_rate": 1.9919706305416606e-05, + "loss": 1.2439, + "step": 476 + }, + { + "epoch": 0.03, + "learning_rate": 1.9919491616393655e-05, + "loss": 1.2274, + "step": 477 + }, + { + "epoch": 0.03, + "learning_rate": 1.9919276927370704e-05, + "loss": 1.2566, + "step": 478 + }, + { + "epoch": 0.03, + "learning_rate": 1.9919062238347756e-05, + "loss": 1.3235, + "step": 479 + }, + { + "epoch": 0.03, + "learning_rate": 1.9918847549324805e-05, + "loss": 1.1897, + "step": 480 + }, + { + "epoch": 0.03, + "learning_rate": 1.9918632860301854e-05, + "loss": 1.336, + "step": 481 + }, + { + "epoch": 0.03, + "learning_rate": 1.9918418171278903e-05, + "loss": 1.3097, + "step": 482 + }, + { + "epoch": 0.03, + "learning_rate": 1.9918203482255952e-05, + "loss": 1.2677, + "step": 483 + }, + { + "epoch": 0.03, + "learning_rate": 1.9917988793233004e-05, + "loss": 1.2427, + "step": 484 + }, + { + "epoch": 0.03, + "learning_rate": 1.9917774104210053e-05, + "loss": 1.258, + "step": 485 + }, + { + "epoch": 0.03, + "learning_rate": 1.9917559415187102e-05, + "loss": 1.3399, + "step": 486 + }, + { + "epoch": 0.03, + "learning_rate": 1.9917344726164155e-05, + "loss": 1.3085, + "step": 487 + }, + { + "epoch": 0.03, + "learning_rate": 1.9917130037141204e-05, + "loss": 1.2177, + "step": 488 + }, + { + "epoch": 0.03, + "learning_rate": 1.9916915348118253e-05, + "loss": 1.168, + "step": 489 + }, + { + "epoch": 0.03, + "learning_rate": 1.99167006590953e-05, + "loss": 1.2305, + "step": 490 + }, + { + "epoch": 0.03, + "learning_rate": 1.9916485970072354e-05, + "loss": 1.1987, + "step": 491 + }, + { + "epoch": 0.03, + "learning_rate": 1.9916271281049403e-05, + "loss": 1.3319, + "step": 492 + }, + { + "epoch": 0.03, + "learning_rate": 1.9916056592026452e-05, + "loss": 1.2458, + "step": 493 + }, + { + "epoch": 0.03, + "learning_rate": 1.99158419030035e-05, + "loss": 1.236, + "step": 494 + }, + { + "epoch": 0.03, + "learning_rate": 1.991562721398055e-05, + "loss": 1.2396, + "step": 495 + }, + { + "epoch": 0.03, + "learning_rate": 1.99154125249576e-05, + "loss": 1.2638, + "step": 496 + }, + { + "epoch": 0.03, + "learning_rate": 1.991519783593465e-05, + "loss": 1.3089, + "step": 497 + }, + { + "epoch": 0.03, + "learning_rate": 1.99149831469117e-05, + "loss": 1.2066, + "step": 498 + }, + { + "epoch": 0.03, + "learning_rate": 1.991476845788875e-05, + "loss": 1.2348, + "step": 499 + }, + { + "epoch": 0.03, + "learning_rate": 1.9914553768865798e-05, + "loss": 1.2695, + "step": 500 + }, + { + "epoch": 0.03, + "learning_rate": 1.9914339079842847e-05, + "loss": 1.4123, + "step": 501 + }, + { + "epoch": 0.03, + "learning_rate": 1.99141243908199e-05, + "loss": 1.2968, + "step": 502 + }, + { + "epoch": 0.03, + "learning_rate": 1.9913909701796948e-05, + "loss": 1.2773, + "step": 503 + }, + { + "epoch": 0.03, + "learning_rate": 1.9913695012774e-05, + "loss": 1.3708, + "step": 504 + }, + { + "epoch": 0.03, + "learning_rate": 1.991348032375105e-05, + "loss": 1.2647, + "step": 505 + }, + { + "epoch": 0.03, + "learning_rate": 1.99132656347281e-05, + "loss": 1.3358, + "step": 506 + }, + { + "epoch": 0.03, + "learning_rate": 1.9913050945705148e-05, + "loss": 1.221, + "step": 507 + }, + { + "epoch": 0.03, + "learning_rate": 1.9912836256682196e-05, + "loss": 1.3338, + "step": 508 + }, + { + "epoch": 0.03, + "learning_rate": 1.9912621567659245e-05, + "loss": 1.2286, + "step": 509 + }, + { + "epoch": 0.03, + "learning_rate": 1.9912406878636298e-05, + "loss": 1.2981, + "step": 510 + }, + { + "epoch": 0.03, + "learning_rate": 1.9912192189613347e-05, + "loss": 1.3016, + "step": 511 + }, + { + "epoch": 0.03, + "learning_rate": 1.9911977500590396e-05, + "loss": 1.3132, + "step": 512 + }, + { + "epoch": 0.03, + "learning_rate": 1.9911762811567445e-05, + "loss": 1.294, + "step": 513 + }, + { + "epoch": 0.03, + "learning_rate": 1.9911548122544494e-05, + "loss": 1.2346, + "step": 514 + }, + { + "epoch": 0.03, + "learning_rate": 1.9911333433521546e-05, + "loss": 1.2086, + "step": 515 + }, + { + "epoch": 0.03, + "learning_rate": 1.9911118744498595e-05, + "loss": 1.3237, + "step": 516 + }, + { + "epoch": 0.03, + "learning_rate": 1.9910904055475647e-05, + "loss": 1.3673, + "step": 517 + }, + { + "epoch": 0.03, + "learning_rate": 1.9910689366452696e-05, + "loss": 1.3581, + "step": 518 + }, + { + "epoch": 0.03, + "learning_rate": 1.9910474677429745e-05, + "loss": 1.2874, + "step": 519 + }, + { + "epoch": 0.03, + "learning_rate": 1.9910259988406794e-05, + "loss": 1.2815, + "step": 520 + }, + { + "epoch": 0.03, + "learning_rate": 1.9910045299383843e-05, + "loss": 1.2511, + "step": 521 + }, + { + "epoch": 0.03, + "learning_rate": 1.9909830610360896e-05, + "loss": 1.241, + "step": 522 + }, + { + "epoch": 0.03, + "learning_rate": 1.9909615921337945e-05, + "loss": 1.2973, + "step": 523 + }, + { + "epoch": 0.03, + "learning_rate": 1.9909401232314993e-05, + "loss": 1.3459, + "step": 524 + }, + { + "epoch": 0.03, + "learning_rate": 1.9909186543292042e-05, + "loss": 1.3174, + "step": 525 + }, + { + "epoch": 0.03, + "learning_rate": 1.990897185426909e-05, + "loss": 1.3113, + "step": 526 + }, + { + "epoch": 0.03, + "learning_rate": 1.990875716524614e-05, + "loss": 1.2688, + "step": 527 + }, + { + "epoch": 0.03, + "learning_rate": 1.9908542476223193e-05, + "loss": 1.3121, + "step": 528 + }, + { + "epoch": 0.03, + "learning_rate": 1.9908327787200242e-05, + "loss": 1.2237, + "step": 529 + }, + { + "epoch": 0.03, + "learning_rate": 1.990811309817729e-05, + "loss": 1.3083, + "step": 530 + }, + { + "epoch": 0.03, + "learning_rate": 1.9907898409154343e-05, + "loss": 1.3134, + "step": 531 + }, + { + "epoch": 0.03, + "learning_rate": 1.9907683720131392e-05, + "loss": 1.2179, + "step": 532 + }, + { + "epoch": 0.03, + "learning_rate": 1.990746903110844e-05, + "loss": 1.2093, + "step": 533 + }, + { + "epoch": 0.03, + "learning_rate": 1.990725434208549e-05, + "loss": 1.3244, + "step": 534 + }, + { + "epoch": 0.03, + "learning_rate": 1.9907039653062542e-05, + "loss": 1.3224, + "step": 535 + }, + { + "epoch": 0.03, + "learning_rate": 1.990682496403959e-05, + "loss": 1.3057, + "step": 536 + }, + { + "epoch": 0.03, + "learning_rate": 1.990661027501664e-05, + "loss": 1.2811, + "step": 537 + }, + { + "epoch": 0.03, + "learning_rate": 1.990639558599369e-05, + "loss": 1.2521, + "step": 538 + }, + { + "epoch": 0.03, + "learning_rate": 1.9906180896970738e-05, + "loss": 1.23, + "step": 539 + }, + { + "epoch": 0.03, + "learning_rate": 1.990596620794779e-05, + "loss": 1.2372, + "step": 540 + }, + { + "epoch": 0.03, + "learning_rate": 1.990575151892484e-05, + "loss": 1.2847, + "step": 541 + }, + { + "epoch": 0.03, + "learning_rate": 1.990553682990189e-05, + "loss": 1.3078, + "step": 542 + }, + { + "epoch": 0.03, + "learning_rate": 1.9905322140878937e-05, + "loss": 1.2863, + "step": 543 + }, + { + "epoch": 0.03, + "learning_rate": 1.9905107451855986e-05, + "loss": 1.2597, + "step": 544 + }, + { + "epoch": 0.04, + "learning_rate": 1.990489276283304e-05, + "loss": 1.3159, + "step": 545 + }, + { + "epoch": 0.04, + "learning_rate": 1.9904678073810088e-05, + "loss": 1.2398, + "step": 546 + }, + { + "epoch": 0.04, + "learning_rate": 1.9904463384787137e-05, + "loss": 1.3349, + "step": 547 + }, + { + "epoch": 0.04, + "learning_rate": 1.990424869576419e-05, + "loss": 1.1487, + "step": 548 + }, + { + "epoch": 0.04, + "learning_rate": 1.9904034006741238e-05, + "loss": 1.255, + "step": 549 + }, + { + "epoch": 0.04, + "learning_rate": 1.9903819317718287e-05, + "loss": 1.1379, + "step": 550 + }, + { + "epoch": 0.04, + "learning_rate": 1.9903604628695336e-05, + "loss": 1.2965, + "step": 551 + }, + { + "epoch": 0.04, + "learning_rate": 1.9903389939672385e-05, + "loss": 1.267, + "step": 552 + }, + { + "epoch": 0.04, + "learning_rate": 1.9903175250649437e-05, + "loss": 1.2846, + "step": 553 + }, + { + "epoch": 0.04, + "learning_rate": 1.9902960561626486e-05, + "loss": 1.1973, + "step": 554 + }, + { + "epoch": 0.04, + "learning_rate": 1.9902745872603535e-05, + "loss": 1.325, + "step": 555 + }, + { + "epoch": 0.04, + "learning_rate": 1.9902531183580584e-05, + "loss": 1.3583, + "step": 556 + }, + { + "epoch": 0.04, + "learning_rate": 1.9902316494557633e-05, + "loss": 1.2204, + "step": 557 + }, + { + "epoch": 0.04, + "learning_rate": 1.9902101805534685e-05, + "loss": 1.2149, + "step": 558 + }, + { + "epoch": 0.04, + "learning_rate": 1.9901887116511734e-05, + "loss": 1.322, + "step": 559 + }, + { + "epoch": 0.04, + "learning_rate": 1.9901672427488787e-05, + "loss": 1.3805, + "step": 560 + }, + { + "epoch": 0.04, + "learning_rate": 1.9901457738465836e-05, + "loss": 1.309, + "step": 561 + }, + { + "epoch": 0.04, + "learning_rate": 1.9901243049442885e-05, + "loss": 1.2272, + "step": 562 + }, + { + "epoch": 0.04, + "learning_rate": 1.9901028360419934e-05, + "loss": 1.1782, + "step": 563 + }, + { + "epoch": 0.04, + "learning_rate": 1.9900813671396983e-05, + "loss": 1.3416, + "step": 564 + }, + { + "epoch": 0.04, + "learning_rate": 1.990059898237403e-05, + "loss": 1.3912, + "step": 565 + }, + { + "epoch": 0.04, + "learning_rate": 1.9900384293351084e-05, + "loss": 1.2276, + "step": 566 + }, + { + "epoch": 0.04, + "learning_rate": 1.9900169604328133e-05, + "loss": 1.2221, + "step": 567 + }, + { + "epoch": 0.04, + "learning_rate": 1.9899954915305182e-05, + "loss": 1.2771, + "step": 568 + }, + { + "epoch": 0.04, + "learning_rate": 1.989974022628223e-05, + "loss": 1.1887, + "step": 569 + }, + { + "epoch": 0.04, + "learning_rate": 1.989952553725928e-05, + "loss": 1.3046, + "step": 570 + }, + { + "epoch": 0.04, + "learning_rate": 1.9899310848236332e-05, + "loss": 1.2781, + "step": 571 + }, + { + "epoch": 0.04, + "learning_rate": 1.989909615921338e-05, + "loss": 1.224, + "step": 572 + }, + { + "epoch": 0.04, + "learning_rate": 1.989888147019043e-05, + "loss": 1.2338, + "step": 573 + }, + { + "epoch": 0.04, + "learning_rate": 1.9898666781167482e-05, + "loss": 1.2546, + "step": 574 + }, + { + "epoch": 0.04, + "learning_rate": 1.989845209214453e-05, + "loss": 1.2811, + "step": 575 + }, + { + "epoch": 0.04, + "learning_rate": 1.989823740312158e-05, + "loss": 1.3074, + "step": 576 + }, + { + "epoch": 0.04, + "learning_rate": 1.989802271409863e-05, + "loss": 1.2935, + "step": 577 + }, + { + "epoch": 0.04, + "learning_rate": 1.989780802507568e-05, + "loss": 1.2029, + "step": 578 + }, + { + "epoch": 0.04, + "learning_rate": 1.989759333605273e-05, + "loss": 1.1812, + "step": 579 + }, + { + "epoch": 0.04, + "learning_rate": 1.989737864702978e-05, + "loss": 1.2049, + "step": 580 + }, + { + "epoch": 0.04, + "learning_rate": 1.989716395800683e-05, + "loss": 1.217, + "step": 581 + }, + { + "epoch": 0.04, + "learning_rate": 1.9896949268983877e-05, + "loss": 1.2683, + "step": 582 + }, + { + "epoch": 0.04, + "learning_rate": 1.9896734579960926e-05, + "loss": 1.2877, + "step": 583 + }, + { + "epoch": 0.04, + "learning_rate": 1.989651989093798e-05, + "loss": 1.3049, + "step": 584 + }, + { + "epoch": 0.04, + "learning_rate": 1.9896305201915028e-05, + "loss": 1.2096, + "step": 585 + }, + { + "epoch": 0.04, + "learning_rate": 1.9896090512892077e-05, + "loss": 1.2504, + "step": 586 + }, + { + "epoch": 0.04, + "learning_rate": 1.9895875823869126e-05, + "loss": 1.3437, + "step": 587 + }, + { + "epoch": 0.04, + "learning_rate": 1.9895661134846175e-05, + "loss": 1.3346, + "step": 588 + }, + { + "epoch": 0.04, + "learning_rate": 1.9895446445823227e-05, + "loss": 1.2583, + "step": 589 + }, + { + "epoch": 0.04, + "learning_rate": 1.9895231756800276e-05, + "loss": 1.1931, + "step": 590 + }, + { + "epoch": 0.04, + "learning_rate": 1.989501706777733e-05, + "loss": 1.3522, + "step": 591 + }, + { + "epoch": 0.04, + "learning_rate": 1.9894802378754377e-05, + "loss": 1.3301, + "step": 592 + }, + { + "epoch": 0.04, + "learning_rate": 1.9894587689731426e-05, + "loss": 1.2803, + "step": 593 + }, + { + "epoch": 0.04, + "learning_rate": 1.9894373000708475e-05, + "loss": 1.2959, + "step": 594 + }, + { + "epoch": 0.04, + "learning_rate": 1.9894158311685524e-05, + "loss": 1.2756, + "step": 595 + }, + { + "epoch": 0.04, + "learning_rate": 1.9893943622662573e-05, + "loss": 1.3003, + "step": 596 + }, + { + "epoch": 0.04, + "learning_rate": 1.9893728933639626e-05, + "loss": 1.2908, + "step": 597 + }, + { + "epoch": 0.04, + "learning_rate": 1.9893514244616674e-05, + "loss": 1.3649, + "step": 598 + }, + { + "epoch": 0.04, + "learning_rate": 1.9893299555593723e-05, + "loss": 1.2486, + "step": 599 + }, + { + "epoch": 0.04, + "learning_rate": 1.9893084866570772e-05, + "loss": 1.2809, + "step": 600 + }, + { + "epoch": 0.04, + "learning_rate": 1.989287017754782e-05, + "loss": 1.2703, + "step": 601 + }, + { + "epoch": 0.04, + "learning_rate": 1.9892655488524874e-05, + "loss": 1.266, + "step": 602 + }, + { + "epoch": 0.04, + "learning_rate": 1.9892440799501923e-05, + "loss": 1.1596, + "step": 603 + }, + { + "epoch": 0.04, + "learning_rate": 1.9892226110478975e-05, + "loss": 1.2645, + "step": 604 + }, + { + "epoch": 0.04, + "learning_rate": 1.9892011421456024e-05, + "loss": 1.2699, + "step": 605 + }, + { + "epoch": 0.04, + "learning_rate": 1.9891796732433073e-05, + "loss": 1.2469, + "step": 606 + }, + { + "epoch": 0.04, + "learning_rate": 1.9891582043410122e-05, + "loss": 1.2347, + "step": 607 + }, + { + "epoch": 0.04, + "learning_rate": 1.989136735438717e-05, + "loss": 1.3059, + "step": 608 + }, + { + "epoch": 0.04, + "learning_rate": 1.9891152665364223e-05, + "loss": 1.3392, + "step": 609 + }, + { + "epoch": 0.04, + "learning_rate": 1.9890937976341272e-05, + "loss": 1.2721, + "step": 610 + }, + { + "epoch": 0.04, + "learning_rate": 1.989072328731832e-05, + "loss": 1.2966, + "step": 611 + }, + { + "epoch": 0.04, + "learning_rate": 1.989050859829537e-05, + "loss": 1.3149, + "step": 612 + }, + { + "epoch": 0.04, + "learning_rate": 1.989029390927242e-05, + "loss": 1.3834, + "step": 613 + }, + { + "epoch": 0.04, + "learning_rate": 1.9890079220249468e-05, + "loss": 1.2941, + "step": 614 + }, + { + "epoch": 0.04, + "learning_rate": 1.988986453122652e-05, + "loss": 1.2352, + "step": 615 + }, + { + "epoch": 0.04, + "learning_rate": 1.988964984220357e-05, + "loss": 1.2554, + "step": 616 + }, + { + "epoch": 0.04, + "learning_rate": 1.988943515318062e-05, + "loss": 1.2907, + "step": 617 + }, + { + "epoch": 0.04, + "learning_rate": 1.988922046415767e-05, + "loss": 1.2774, + "step": 618 + }, + { + "epoch": 0.04, + "learning_rate": 1.988900577513472e-05, + "loss": 1.2675, + "step": 619 + }, + { + "epoch": 0.04, + "learning_rate": 1.988879108611177e-05, + "loss": 1.3354, + "step": 620 + }, + { + "epoch": 0.04, + "learning_rate": 1.9888576397088818e-05, + "loss": 1.1955, + "step": 621 + }, + { + "epoch": 0.04, + "learning_rate": 1.988836170806587e-05, + "loss": 1.2176, + "step": 622 + }, + { + "epoch": 0.04, + "learning_rate": 1.988814701904292e-05, + "loss": 1.3886, + "step": 623 + }, + { + "epoch": 0.04, + "learning_rate": 1.9887932330019968e-05, + "loss": 1.2391, + "step": 624 + }, + { + "epoch": 0.04, + "learning_rate": 1.9887717640997017e-05, + "loss": 1.1574, + "step": 625 + }, + { + "epoch": 0.04, + "learning_rate": 1.9887502951974066e-05, + "loss": 1.2905, + "step": 626 + }, + { + "epoch": 0.04, + "learning_rate": 1.9887288262951118e-05, + "loss": 1.2342, + "step": 627 + }, + { + "epoch": 0.04, + "learning_rate": 1.9887073573928167e-05, + "loss": 1.2118, + "step": 628 + }, + { + "epoch": 0.04, + "learning_rate": 1.9886858884905216e-05, + "loss": 1.2165, + "step": 629 + }, + { + "epoch": 0.04, + "learning_rate": 1.9886644195882265e-05, + "loss": 1.294, + "step": 630 + }, + { + "epoch": 0.04, + "learning_rate": 1.9886429506859314e-05, + "loss": 1.2709, + "step": 631 + }, + { + "epoch": 0.04, + "learning_rate": 1.9886214817836366e-05, + "loss": 1.1949, + "step": 632 + }, + { + "epoch": 0.04, + "learning_rate": 1.9886000128813415e-05, + "loss": 1.1611, + "step": 633 + }, + { + "epoch": 0.04, + "learning_rate": 1.9885785439790464e-05, + "loss": 1.2573, + "step": 634 + }, + { + "epoch": 0.04, + "learning_rate": 1.9885570750767517e-05, + "loss": 1.2557, + "step": 635 + }, + { + "epoch": 0.04, + "learning_rate": 1.9885356061744566e-05, + "loss": 1.2299, + "step": 636 + }, + { + "epoch": 0.04, + "learning_rate": 1.9885141372721615e-05, + "loss": 1.28, + "step": 637 + }, + { + "epoch": 0.04, + "learning_rate": 1.9884926683698664e-05, + "loss": 1.1739, + "step": 638 + }, + { + "epoch": 0.04, + "learning_rate": 1.9884711994675713e-05, + "loss": 1.2834, + "step": 639 + }, + { + "epoch": 0.04, + "learning_rate": 1.9884497305652765e-05, + "loss": 1.2211, + "step": 640 + }, + { + "epoch": 0.04, + "learning_rate": 1.9884282616629814e-05, + "loss": 1.2488, + "step": 641 + }, + { + "epoch": 0.04, + "learning_rate": 1.9884067927606863e-05, + "loss": 1.2035, + "step": 642 + }, + { + "epoch": 0.04, + "learning_rate": 1.9883853238583912e-05, + "loss": 1.2922, + "step": 643 + }, + { + "epoch": 0.04, + "learning_rate": 1.988363854956096e-05, + "loss": 1.2862, + "step": 644 + }, + { + "epoch": 0.04, + "learning_rate": 1.9883423860538013e-05, + "loss": 1.2795, + "step": 645 + }, + { + "epoch": 0.04, + "learning_rate": 1.9883209171515062e-05, + "loss": 1.2492, + "step": 646 + }, + { + "epoch": 0.04, + "learning_rate": 1.9882994482492114e-05, + "loss": 1.1684, + "step": 647 + }, + { + "epoch": 0.04, + "learning_rate": 1.9882779793469163e-05, + "loss": 1.2366, + "step": 648 + }, + { + "epoch": 0.04, + "learning_rate": 1.9882565104446212e-05, + "loss": 1.3013, + "step": 649 + }, + { + "epoch": 0.04, + "learning_rate": 1.988235041542326e-05, + "loss": 1.21, + "step": 650 + }, + { + "epoch": 0.04, + "learning_rate": 1.988213572640031e-05, + "loss": 1.2385, + "step": 651 + }, + { + "epoch": 0.04, + "learning_rate": 1.988192103737736e-05, + "loss": 1.217, + "step": 652 + }, + { + "epoch": 0.04, + "learning_rate": 1.988170634835441e-05, + "loss": 1.2503, + "step": 653 + }, + { + "epoch": 0.04, + "learning_rate": 1.988149165933146e-05, + "loss": 1.3369, + "step": 654 + }, + { + "epoch": 0.04, + "learning_rate": 1.988127697030851e-05, + "loss": 1.2615, + "step": 655 + }, + { + "epoch": 0.04, + "learning_rate": 1.988106228128556e-05, + "loss": 1.2744, + "step": 656 + }, + { + "epoch": 0.04, + "learning_rate": 1.9880847592262607e-05, + "loss": 1.3297, + "step": 657 + }, + { + "epoch": 0.04, + "learning_rate": 1.988063290323966e-05, + "loss": 1.1935, + "step": 658 + }, + { + "epoch": 0.04, + "learning_rate": 1.988041821421671e-05, + "loss": 1.2464, + "step": 659 + }, + { + "epoch": 0.04, + "learning_rate": 1.9880203525193758e-05, + "loss": 1.2497, + "step": 660 + }, + { + "epoch": 0.04, + "learning_rate": 1.9879988836170807e-05, + "loss": 1.1858, + "step": 661 + }, + { + "epoch": 0.04, + "learning_rate": 1.987977414714786e-05, + "loss": 1.2577, + "step": 662 + }, + { + "epoch": 0.04, + "learning_rate": 1.9879559458124908e-05, + "loss": 1.219, + "step": 663 + }, + { + "epoch": 0.04, + "learning_rate": 1.9879344769101957e-05, + "loss": 1.3487, + "step": 664 + }, + { + "epoch": 0.04, + "learning_rate": 1.987913008007901e-05, + "loss": 1.2351, + "step": 665 + }, + { + "epoch": 0.04, + "learning_rate": 1.987891539105606e-05, + "loss": 1.2964, + "step": 666 + }, + { + "epoch": 0.04, + "learning_rate": 1.9878700702033107e-05, + "loss": 1.2424, + "step": 667 + }, + { + "epoch": 0.04, + "learning_rate": 1.9878486013010156e-05, + "loss": 1.2953, + "step": 668 + }, + { + "epoch": 0.04, + "learning_rate": 1.9878271323987205e-05, + "loss": 1.2752, + "step": 669 + }, + { + "epoch": 0.04, + "learning_rate": 1.9878056634964254e-05, + "loss": 1.3269, + "step": 670 + }, + { + "epoch": 0.04, + "learning_rate": 1.9877841945941307e-05, + "loss": 1.2411, + "step": 671 + }, + { + "epoch": 0.04, + "learning_rate": 1.9877627256918355e-05, + "loss": 1.2959, + "step": 672 + }, + { + "epoch": 0.04, + "learning_rate": 1.9877412567895404e-05, + "loss": 1.272, + "step": 673 + }, + { + "epoch": 0.04, + "learning_rate": 1.9877197878872453e-05, + "loss": 1.2386, + "step": 674 + }, + { + "epoch": 0.04, + "learning_rate": 1.9876983189849502e-05, + "loss": 1.2849, + "step": 675 + }, + { + "epoch": 0.04, + "learning_rate": 1.9876768500826555e-05, + "loss": 1.2722, + "step": 676 + }, + { + "epoch": 0.04, + "learning_rate": 1.9876553811803604e-05, + "loss": 1.2331, + "step": 677 + }, + { + "epoch": 0.04, + "learning_rate": 1.9876339122780656e-05, + "loss": 1.2954, + "step": 678 + }, + { + "epoch": 0.04, + "learning_rate": 1.9876124433757705e-05, + "loss": 1.2461, + "step": 679 + }, + { + "epoch": 0.04, + "learning_rate": 1.9875909744734754e-05, + "loss": 1.1602, + "step": 680 + }, + { + "epoch": 0.04, + "learning_rate": 1.9875695055711803e-05, + "loss": 1.1634, + "step": 681 + }, + { + "epoch": 0.04, + "learning_rate": 1.9875480366688852e-05, + "loss": 1.3161, + "step": 682 + }, + { + "epoch": 0.04, + "learning_rate": 1.98752656776659e-05, + "loss": 1.2774, + "step": 683 + }, + { + "epoch": 0.04, + "learning_rate": 1.9875050988642953e-05, + "loss": 1.2535, + "step": 684 + }, + { + "epoch": 0.04, + "learning_rate": 1.9874836299620002e-05, + "loss": 1.1964, + "step": 685 + }, + { + "epoch": 0.04, + "learning_rate": 1.987462161059705e-05, + "loss": 1.2631, + "step": 686 + }, + { + "epoch": 0.04, + "learning_rate": 1.98744069215741e-05, + "loss": 1.247, + "step": 687 + }, + { + "epoch": 0.04, + "learning_rate": 1.987419223255115e-05, + "loss": 1.2549, + "step": 688 + }, + { + "epoch": 0.04, + "learning_rate": 1.98739775435282e-05, + "loss": 1.2668, + "step": 689 + }, + { + "epoch": 0.04, + "learning_rate": 1.987376285450525e-05, + "loss": 1.2784, + "step": 690 + }, + { + "epoch": 0.04, + "learning_rate": 1.9873548165482303e-05, + "loss": 1.145, + "step": 691 + }, + { + "epoch": 0.04, + "learning_rate": 1.9873333476459352e-05, + "loss": 1.3134, + "step": 692 + }, + { + "epoch": 0.04, + "learning_rate": 1.98731187874364e-05, + "loss": 1.292, + "step": 693 + }, + { + "epoch": 0.04, + "learning_rate": 1.987290409841345e-05, + "loss": 1.2579, + "step": 694 + }, + { + "epoch": 0.04, + "learning_rate": 1.98726894093905e-05, + "loss": 1.243, + "step": 695 + }, + { + "epoch": 0.04, + "learning_rate": 1.987247472036755e-05, + "loss": 1.373, + "step": 696 + }, + { + "epoch": 0.04, + "learning_rate": 1.98722600313446e-05, + "loss": 1.2982, + "step": 697 + }, + { + "epoch": 0.04, + "learning_rate": 1.987204534232165e-05, + "loss": 1.189, + "step": 698 + }, + { + "epoch": 0.04, + "learning_rate": 1.9871830653298698e-05, + "loss": 1.2608, + "step": 699 + }, + { + "epoch": 0.05, + "learning_rate": 1.9871615964275747e-05, + "loss": 1.2946, + "step": 700 + }, + { + "epoch": 0.05, + "learning_rate": 1.9871401275252796e-05, + "loss": 1.1971, + "step": 701 + }, + { + "epoch": 0.05, + "learning_rate": 1.9871186586229848e-05, + "loss": 1.2302, + "step": 702 + }, + { + "epoch": 0.05, + "learning_rate": 1.9870971897206897e-05, + "loss": 1.2869, + "step": 703 + }, + { + "epoch": 0.05, + "learning_rate": 1.9870757208183946e-05, + "loss": 1.198, + "step": 704 + }, + { + "epoch": 0.05, + "learning_rate": 1.9870542519161e-05, + "loss": 1.2519, + "step": 705 + }, + { + "epoch": 0.05, + "learning_rate": 1.9870327830138047e-05, + "loss": 1.2225, + "step": 706 + }, + { + "epoch": 0.05, + "learning_rate": 1.9870113141115096e-05, + "loss": 1.2381, + "step": 707 + }, + { + "epoch": 0.05, + "learning_rate": 1.9869898452092145e-05, + "loss": 1.192, + "step": 708 + }, + { + "epoch": 0.05, + "learning_rate": 1.9869683763069198e-05, + "loss": 1.2135, + "step": 709 + }, + { + "epoch": 0.05, + "learning_rate": 1.9869469074046247e-05, + "loss": 1.2069, + "step": 710 + }, + { + "epoch": 0.05, + "learning_rate": 1.9869254385023296e-05, + "loss": 1.2292, + "step": 711 + }, + { + "epoch": 0.05, + "learning_rate": 1.9869039696000345e-05, + "loss": 1.2095, + "step": 712 + }, + { + "epoch": 0.05, + "learning_rate": 1.9868825006977394e-05, + "loss": 1.3154, + "step": 713 + }, + { + "epoch": 0.05, + "learning_rate": 1.9868610317954446e-05, + "loss": 1.3304, + "step": 714 + }, + { + "epoch": 0.05, + "learning_rate": 1.9868395628931495e-05, + "loss": 1.3259, + "step": 715 + }, + { + "epoch": 0.05, + "learning_rate": 1.9868180939908544e-05, + "loss": 1.1951, + "step": 716 + }, + { + "epoch": 0.05, + "learning_rate": 1.9867966250885593e-05, + "loss": 1.3179, + "step": 717 + }, + { + "epoch": 0.05, + "learning_rate": 1.9867751561862642e-05, + "loss": 1.2513, + "step": 718 + }, + { + "epoch": 0.05, + "learning_rate": 1.986753687283969e-05, + "loss": 1.2128, + "step": 719 + }, + { + "epoch": 0.05, + "learning_rate": 1.9867322183816743e-05, + "loss": 1.224, + "step": 720 + }, + { + "epoch": 0.05, + "learning_rate": 1.9867107494793792e-05, + "loss": 1.2742, + "step": 721 + }, + { + "epoch": 0.05, + "learning_rate": 1.9866892805770844e-05, + "loss": 1.4352, + "step": 722 + }, + { + "epoch": 0.05, + "learning_rate": 1.9866678116747893e-05, + "loss": 1.2677, + "step": 723 + }, + { + "epoch": 0.05, + "learning_rate": 1.9866463427724942e-05, + "loss": 1.235, + "step": 724 + }, + { + "epoch": 0.05, + "learning_rate": 1.986624873870199e-05, + "loss": 1.3304, + "step": 725 + }, + { + "epoch": 0.05, + "learning_rate": 1.986603404967904e-05, + "loss": 1.136, + "step": 726 + }, + { + "epoch": 0.05, + "learning_rate": 1.9865819360656093e-05, + "loss": 1.2368, + "step": 727 + }, + { + "epoch": 0.05, + "learning_rate": 1.986560467163314e-05, + "loss": 1.1737, + "step": 728 + }, + { + "epoch": 0.05, + "learning_rate": 1.986538998261019e-05, + "loss": 1.2708, + "step": 729 + }, + { + "epoch": 0.05, + "learning_rate": 1.986517529358724e-05, + "loss": 1.2146, + "step": 730 + }, + { + "epoch": 0.05, + "learning_rate": 1.986496060456429e-05, + "loss": 1.3752, + "step": 731 + }, + { + "epoch": 0.05, + "learning_rate": 1.986474591554134e-05, + "loss": 1.2737, + "step": 732 + }, + { + "epoch": 0.05, + "learning_rate": 1.986453122651839e-05, + "loss": 1.3699, + "step": 733 + }, + { + "epoch": 0.05, + "learning_rate": 1.9864316537495442e-05, + "loss": 1.2132, + "step": 734 + }, + { + "epoch": 0.05, + "learning_rate": 1.986410184847249e-05, + "loss": 1.284, + "step": 735 + }, + { + "epoch": 0.05, + "learning_rate": 1.986388715944954e-05, + "loss": 1.2519, + "step": 736 + }, + { + "epoch": 0.05, + "learning_rate": 1.986367247042659e-05, + "loss": 1.2245, + "step": 737 + }, + { + "epoch": 0.05, + "learning_rate": 1.9863457781403638e-05, + "loss": 1.2086, + "step": 738 + }, + { + "epoch": 0.05, + "learning_rate": 1.9863243092380687e-05, + "loss": 1.2033, + "step": 739 + }, + { + "epoch": 0.05, + "learning_rate": 1.986302840335774e-05, + "loss": 1.2369, + "step": 740 + }, + { + "epoch": 0.05, + "learning_rate": 1.9862813714334788e-05, + "loss": 1.2544, + "step": 741 + }, + { + "epoch": 0.05, + "learning_rate": 1.9862599025311837e-05, + "loss": 1.2733, + "step": 742 + }, + { + "epoch": 0.05, + "learning_rate": 1.9862384336288886e-05, + "loss": 1.2762, + "step": 743 + }, + { + "epoch": 0.05, + "learning_rate": 1.9862169647265935e-05, + "loss": 1.1931, + "step": 744 + }, + { + "epoch": 0.05, + "learning_rate": 1.9861954958242988e-05, + "loss": 1.3562, + "step": 745 + }, + { + "epoch": 0.05, + "learning_rate": 1.9861740269220036e-05, + "loss": 1.2675, + "step": 746 + }, + { + "epoch": 0.05, + "learning_rate": 1.9861525580197085e-05, + "loss": 1.2755, + "step": 747 + }, + { + "epoch": 0.05, + "learning_rate": 1.9861310891174134e-05, + "loss": 1.243, + "step": 748 + }, + { + "epoch": 0.05, + "learning_rate": 1.9861096202151187e-05, + "loss": 1.2551, + "step": 749 + }, + { + "epoch": 0.05, + "learning_rate": 1.9860881513128236e-05, + "loss": 1.1789, + "step": 750 + }, + { + "epoch": 0.05, + "learning_rate": 1.9860666824105285e-05, + "loss": 1.1886, + "step": 751 + }, + { + "epoch": 0.05, + "learning_rate": 1.9860452135082337e-05, + "loss": 1.2742, + "step": 752 + }, + { + "epoch": 0.05, + "learning_rate": 1.9860237446059386e-05, + "loss": 1.3408, + "step": 753 + }, + { + "epoch": 0.05, + "learning_rate": 1.9860022757036435e-05, + "loss": 1.3166, + "step": 754 + }, + { + "epoch": 0.05, + "learning_rate": 1.9859808068013484e-05, + "loss": 1.2141, + "step": 755 + }, + { + "epoch": 0.05, + "learning_rate": 1.9859593378990533e-05, + "loss": 1.2234, + "step": 756 + }, + { + "epoch": 0.05, + "learning_rate": 1.9859378689967582e-05, + "loss": 1.1905, + "step": 757 + }, + { + "epoch": 0.05, + "learning_rate": 1.9859164000944634e-05, + "loss": 1.2256, + "step": 758 + }, + { + "epoch": 0.05, + "learning_rate": 1.9858949311921683e-05, + "loss": 1.2494, + "step": 759 + }, + { + "epoch": 0.05, + "learning_rate": 1.9858734622898732e-05, + "loss": 1.213, + "step": 760 + }, + { + "epoch": 0.05, + "learning_rate": 1.985851993387578e-05, + "loss": 1.2562, + "step": 761 + }, + { + "epoch": 0.05, + "learning_rate": 1.985830524485283e-05, + "loss": 1.1794, + "step": 762 + }, + { + "epoch": 0.05, + "learning_rate": 1.9858090555829882e-05, + "loss": 1.2321, + "step": 763 + }, + { + "epoch": 0.05, + "learning_rate": 1.985787586680693e-05, + "loss": 1.1829, + "step": 764 + }, + { + "epoch": 0.05, + "learning_rate": 1.9857661177783984e-05, + "loss": 1.2205, + "step": 765 + }, + { + "epoch": 0.05, + "learning_rate": 1.9857446488761033e-05, + "loss": 1.2624, + "step": 766 + }, + { + "epoch": 0.05, + "learning_rate": 1.9857231799738082e-05, + "loss": 1.3652, + "step": 767 + }, + { + "epoch": 0.05, + "learning_rate": 1.985701711071513e-05, + "loss": 1.2358, + "step": 768 + }, + { + "epoch": 0.05, + "learning_rate": 1.985680242169218e-05, + "loss": 1.1847, + "step": 769 + }, + { + "epoch": 0.05, + "learning_rate": 1.985658773266923e-05, + "loss": 1.2445, + "step": 770 + }, + { + "epoch": 0.05, + "learning_rate": 1.985637304364628e-05, + "loss": 1.2326, + "step": 771 + }, + { + "epoch": 0.05, + "learning_rate": 1.985615835462333e-05, + "loss": 1.2052, + "step": 772 + }, + { + "epoch": 0.05, + "learning_rate": 1.985594366560038e-05, + "loss": 1.2669, + "step": 773 + }, + { + "epoch": 0.05, + "learning_rate": 1.9855728976577428e-05, + "loss": 1.2293, + "step": 774 + }, + { + "epoch": 0.05, + "learning_rate": 1.9855514287554477e-05, + "loss": 1.2939, + "step": 775 + }, + { + "epoch": 0.05, + "learning_rate": 1.985529959853153e-05, + "loss": 1.2687, + "step": 776 + }, + { + "epoch": 0.05, + "learning_rate": 1.9855084909508578e-05, + "loss": 1.2984, + "step": 777 + }, + { + "epoch": 0.05, + "learning_rate": 1.985487022048563e-05, + "loss": 1.2951, + "step": 778 + }, + { + "epoch": 0.05, + "learning_rate": 1.985465553146268e-05, + "loss": 1.2553, + "step": 779 + }, + { + "epoch": 0.05, + "learning_rate": 1.985444084243973e-05, + "loss": 1.2832, + "step": 780 + }, + { + "epoch": 0.05, + "learning_rate": 1.9854226153416777e-05, + "loss": 1.178, + "step": 781 + }, + { + "epoch": 0.05, + "learning_rate": 1.9854011464393826e-05, + "loss": 1.2918, + "step": 782 + }, + { + "epoch": 0.05, + "learning_rate": 1.985379677537088e-05, + "loss": 1.2426, + "step": 783 + }, + { + "epoch": 0.05, + "learning_rate": 1.9853582086347928e-05, + "loss": 1.2666, + "step": 784 + }, + { + "epoch": 0.05, + "learning_rate": 1.9853367397324977e-05, + "loss": 1.3116, + "step": 785 + }, + { + "epoch": 0.05, + "learning_rate": 1.9853152708302026e-05, + "loss": 1.3802, + "step": 786 + }, + { + "epoch": 0.05, + "learning_rate": 1.9852938019279075e-05, + "loss": 1.2914, + "step": 787 + }, + { + "epoch": 0.05, + "learning_rate": 1.9852723330256124e-05, + "loss": 1.2543, + "step": 788 + }, + { + "epoch": 0.05, + "learning_rate": 1.9852508641233176e-05, + "loss": 1.0462, + "step": 789 + }, + { + "epoch": 0.05, + "learning_rate": 1.9852293952210225e-05, + "loss": 1.1978, + "step": 790 + }, + { + "epoch": 0.05, + "learning_rate": 1.9852079263187274e-05, + "loss": 1.214, + "step": 791 + }, + { + "epoch": 0.05, + "learning_rate": 1.9851864574164326e-05, + "loss": 1.2312, + "step": 792 + }, + { + "epoch": 0.05, + "learning_rate": 1.9851649885141375e-05, + "loss": 1.2255, + "step": 793 + }, + { + "epoch": 0.05, + "learning_rate": 1.9851435196118424e-05, + "loss": 1.2165, + "step": 794 + }, + { + "epoch": 0.05, + "learning_rate": 1.9851220507095473e-05, + "loss": 1.3108, + "step": 795 + }, + { + "epoch": 0.05, + "learning_rate": 1.9851005818072525e-05, + "loss": 1.3565, + "step": 796 + }, + { + "epoch": 0.05, + "learning_rate": 1.9850791129049574e-05, + "loss": 1.1768, + "step": 797 + }, + { + "epoch": 0.05, + "learning_rate": 1.9850576440026623e-05, + "loss": 1.1846, + "step": 798 + }, + { + "epoch": 0.05, + "learning_rate": 1.9850361751003672e-05, + "loss": 1.2012, + "step": 799 + }, + { + "epoch": 0.05, + "learning_rate": 1.985014706198072e-05, + "loss": 1.2349, + "step": 800 + }, + { + "epoch": 0.05, + "learning_rate": 1.9849932372957774e-05, + "loss": 1.2705, + "step": 801 + }, + { + "epoch": 0.05, + "learning_rate": 1.9849717683934823e-05, + "loss": 1.197, + "step": 802 + }, + { + "epoch": 0.05, + "learning_rate": 1.984950299491187e-05, + "loss": 1.1257, + "step": 803 + }, + { + "epoch": 0.05, + "learning_rate": 1.984928830588892e-05, + "loss": 1.2828, + "step": 804 + }, + { + "epoch": 0.05, + "learning_rate": 1.984907361686597e-05, + "loss": 1.2945, + "step": 805 + }, + { + "epoch": 0.05, + "learning_rate": 1.984885892784302e-05, + "loss": 1.2299, + "step": 806 + }, + { + "epoch": 0.05, + "learning_rate": 1.984864423882007e-05, + "loss": 1.304, + "step": 807 + }, + { + "epoch": 0.05, + "learning_rate": 1.984842954979712e-05, + "loss": 1.2078, + "step": 808 + }, + { + "epoch": 0.05, + "learning_rate": 1.9848214860774172e-05, + "loss": 1.2694, + "step": 809 + }, + { + "epoch": 0.05, + "learning_rate": 1.984800017175122e-05, + "loss": 1.2079, + "step": 810 + }, + { + "epoch": 0.05, + "learning_rate": 1.984778548272827e-05, + "loss": 1.1501, + "step": 811 + }, + { + "epoch": 0.05, + "learning_rate": 1.984757079370532e-05, + "loss": 1.2813, + "step": 812 + }, + { + "epoch": 0.05, + "learning_rate": 1.9847356104682368e-05, + "loss": 1.225, + "step": 813 + }, + { + "epoch": 0.05, + "learning_rate": 1.984714141565942e-05, + "loss": 1.2717, + "step": 814 + }, + { + "epoch": 0.05, + "learning_rate": 1.984692672663647e-05, + "loss": 1.292, + "step": 815 + }, + { + "epoch": 0.05, + "learning_rate": 1.9846712037613518e-05, + "loss": 1.137, + "step": 816 + }, + { + "epoch": 0.05, + "learning_rate": 1.9846497348590567e-05, + "loss": 1.2554, + "step": 817 + }, + { + "epoch": 0.05, + "learning_rate": 1.9846282659567616e-05, + "loss": 1.2312, + "step": 818 + }, + { + "epoch": 0.05, + "learning_rate": 1.9846067970544665e-05, + "loss": 1.2035, + "step": 819 + }, + { + "epoch": 0.05, + "learning_rate": 1.9845853281521718e-05, + "loss": 1.2621, + "step": 820 + }, + { + "epoch": 0.05, + "learning_rate": 1.984563859249877e-05, + "loss": 1.2195, + "step": 821 + }, + { + "epoch": 0.05, + "learning_rate": 1.984542390347582e-05, + "loss": 1.3479, + "step": 822 + }, + { + "epoch": 0.05, + "learning_rate": 1.9845209214452868e-05, + "loss": 1.2664, + "step": 823 + }, + { + "epoch": 0.05, + "learning_rate": 1.9844994525429917e-05, + "loss": 1.2146, + "step": 824 + }, + { + "epoch": 0.05, + "learning_rate": 1.9844779836406966e-05, + "loss": 1.3113, + "step": 825 + }, + { + "epoch": 0.05, + "learning_rate": 1.9844565147384015e-05, + "loss": 1.1563, + "step": 826 + }, + { + "epoch": 0.05, + "learning_rate": 1.9844350458361067e-05, + "loss": 1.2696, + "step": 827 + }, + { + "epoch": 0.05, + "learning_rate": 1.9844135769338116e-05, + "loss": 1.1412, + "step": 828 + }, + { + "epoch": 0.05, + "learning_rate": 1.9843921080315165e-05, + "loss": 1.2931, + "step": 829 + }, + { + "epoch": 0.05, + "learning_rate": 1.9843706391292214e-05, + "loss": 1.3367, + "step": 830 + }, + { + "epoch": 0.05, + "learning_rate": 1.9843491702269263e-05, + "loss": 1.1792, + "step": 831 + }, + { + "epoch": 0.05, + "learning_rate": 1.9843277013246315e-05, + "loss": 1.2788, + "step": 832 + }, + { + "epoch": 0.05, + "learning_rate": 1.9843062324223364e-05, + "loss": 1.1546, + "step": 833 + }, + { + "epoch": 0.05, + "learning_rate": 1.9842847635200413e-05, + "loss": 1.2027, + "step": 834 + }, + { + "epoch": 0.05, + "learning_rate": 1.9842632946177462e-05, + "loss": 1.3154, + "step": 835 + }, + { + "epoch": 0.05, + "learning_rate": 1.9842418257154514e-05, + "loss": 1.2876, + "step": 836 + }, + { + "epoch": 0.05, + "learning_rate": 1.9842203568131563e-05, + "loss": 1.239, + "step": 837 + }, + { + "epoch": 0.05, + "learning_rate": 1.9841988879108612e-05, + "loss": 1.1291, + "step": 838 + }, + { + "epoch": 0.05, + "learning_rate": 1.984177419008566e-05, + "loss": 1.1577, + "step": 839 + }, + { + "epoch": 0.05, + "learning_rate": 1.9841559501062714e-05, + "loss": 1.2193, + "step": 840 + }, + { + "epoch": 0.05, + "learning_rate": 1.9841344812039763e-05, + "loss": 1.1824, + "step": 841 + }, + { + "epoch": 0.05, + "learning_rate": 1.984113012301681e-05, + "loss": 1.2542, + "step": 842 + }, + { + "epoch": 0.05, + "learning_rate": 1.984091543399386e-05, + "loss": 1.2254, + "step": 843 + }, + { + "epoch": 0.05, + "learning_rate": 1.984070074497091e-05, + "loss": 1.2053, + "step": 844 + }, + { + "epoch": 0.05, + "learning_rate": 1.9840486055947962e-05, + "loss": 1.2222, + "step": 845 + }, + { + "epoch": 0.05, + "learning_rate": 1.984027136692501e-05, + "loss": 1.1923, + "step": 846 + }, + { + "epoch": 0.05, + "learning_rate": 1.984005667790206e-05, + "loss": 1.2419, + "step": 847 + }, + { + "epoch": 0.05, + "learning_rate": 1.983984198887911e-05, + "loss": 1.2697, + "step": 848 + }, + { + "epoch": 0.05, + "learning_rate": 1.9839627299856158e-05, + "loss": 1.2436, + "step": 849 + }, + { + "epoch": 0.05, + "learning_rate": 1.983941261083321e-05, + "loss": 1.2717, + "step": 850 + }, + { + "epoch": 0.05, + "learning_rate": 1.983919792181026e-05, + "loss": 1.2778, + "step": 851 + }, + { + "epoch": 0.05, + "learning_rate": 1.983898323278731e-05, + "loss": 1.1926, + "step": 852 + }, + { + "epoch": 0.05, + "learning_rate": 1.983876854376436e-05, + "loss": 1.2714, + "step": 853 + }, + { + "epoch": 0.05, + "learning_rate": 1.983855385474141e-05, + "loss": 1.3405, + "step": 854 + }, + { + "epoch": 0.06, + "learning_rate": 1.983833916571846e-05, + "loss": 1.1759, + "step": 855 + }, + { + "epoch": 0.06, + "learning_rate": 1.9838124476695507e-05, + "loss": 1.261, + "step": 856 + }, + { + "epoch": 0.06, + "learning_rate": 1.9837909787672556e-05, + "loss": 1.249, + "step": 857 + }, + { + "epoch": 0.06, + "learning_rate": 1.983769509864961e-05, + "loss": 1.1461, + "step": 858 + }, + { + "epoch": 0.06, + "learning_rate": 1.9837480409626658e-05, + "loss": 1.2031, + "step": 859 + }, + { + "epoch": 0.06, + "learning_rate": 1.9837265720603707e-05, + "loss": 1.1922, + "step": 860 + }, + { + "epoch": 0.06, + "learning_rate": 1.9837051031580756e-05, + "loss": 1.2483, + "step": 861 + }, + { + "epoch": 0.06, + "learning_rate": 1.9836836342557805e-05, + "loss": 1.214, + "step": 862 + }, + { + "epoch": 0.06, + "learning_rate": 1.9836621653534857e-05, + "loss": 1.1749, + "step": 863 + }, + { + "epoch": 0.06, + "learning_rate": 1.9836406964511906e-05, + "loss": 1.1229, + "step": 864 + }, + { + "epoch": 0.06, + "learning_rate": 1.9836192275488958e-05, + "loss": 1.2996, + "step": 865 + }, + { + "epoch": 0.06, + "learning_rate": 1.9835977586466007e-05, + "loss": 1.2767, + "step": 866 + }, + { + "epoch": 0.06, + "learning_rate": 1.9835762897443056e-05, + "loss": 1.2078, + "step": 867 + }, + { + "epoch": 0.06, + "learning_rate": 1.9835548208420105e-05, + "loss": 1.2443, + "step": 868 + }, + { + "epoch": 0.06, + "learning_rate": 1.9835333519397154e-05, + "loss": 1.2569, + "step": 869 + }, + { + "epoch": 0.06, + "learning_rate": 1.9835118830374206e-05, + "loss": 1.2777, + "step": 870 + }, + { + "epoch": 0.06, + "learning_rate": 1.9834904141351255e-05, + "loss": 1.2286, + "step": 871 + }, + { + "epoch": 0.06, + "learning_rate": 1.9834689452328304e-05, + "loss": 1.2117, + "step": 872 + }, + { + "epoch": 0.06, + "learning_rate": 1.9834474763305353e-05, + "loss": 1.2414, + "step": 873 + }, + { + "epoch": 0.06, + "learning_rate": 1.9834260074282402e-05, + "loss": 1.2176, + "step": 874 + }, + { + "epoch": 0.06, + "learning_rate": 1.983404538525945e-05, + "loss": 1.1173, + "step": 875 + }, + { + "epoch": 0.06, + "learning_rate": 1.9833830696236504e-05, + "loss": 1.2277, + "step": 876 + }, + { + "epoch": 0.06, + "learning_rate": 1.9833616007213553e-05, + "loss": 1.1961, + "step": 877 + }, + { + "epoch": 0.06, + "learning_rate": 1.98334013181906e-05, + "loss": 1.1931, + "step": 878 + }, + { + "epoch": 0.06, + "learning_rate": 1.9833186629167654e-05, + "loss": 1.2621, + "step": 879 + }, + { + "epoch": 0.06, + "learning_rate": 1.9832971940144703e-05, + "loss": 1.2657, + "step": 880 + }, + { + "epoch": 0.06, + "learning_rate": 1.9832757251121752e-05, + "loss": 1.3039, + "step": 881 + }, + { + "epoch": 0.06, + "learning_rate": 1.98325425620988e-05, + "loss": 1.2288, + "step": 882 + }, + { + "epoch": 0.06, + "learning_rate": 1.9832327873075853e-05, + "loss": 1.2388, + "step": 883 + }, + { + "epoch": 0.06, + "learning_rate": 1.9832113184052902e-05, + "loss": 1.2195, + "step": 884 + }, + { + "epoch": 0.06, + "learning_rate": 1.983189849502995e-05, + "loss": 1.2499, + "step": 885 + }, + { + "epoch": 0.06, + "learning_rate": 1.9831683806007e-05, + "loss": 1.2958, + "step": 886 + }, + { + "epoch": 0.06, + "learning_rate": 1.983146911698405e-05, + "loss": 1.187, + "step": 887 + }, + { + "epoch": 0.06, + "learning_rate": 1.98312544279611e-05, + "loss": 1.2342, + "step": 888 + }, + { + "epoch": 0.06, + "learning_rate": 1.983103973893815e-05, + "loss": 1.2492, + "step": 889 + }, + { + "epoch": 0.06, + "learning_rate": 1.98308250499152e-05, + "loss": 1.202, + "step": 890 + }, + { + "epoch": 0.06, + "learning_rate": 1.9830610360892248e-05, + "loss": 1.1819, + "step": 891 + }, + { + "epoch": 0.06, + "learning_rate": 1.9830395671869297e-05, + "loss": 1.2276, + "step": 892 + }, + { + "epoch": 0.06, + "learning_rate": 1.9830180982846346e-05, + "loss": 1.2047, + "step": 893 + }, + { + "epoch": 0.06, + "learning_rate": 1.98299662938234e-05, + "loss": 1.2098, + "step": 894 + }, + { + "epoch": 0.06, + "learning_rate": 1.9829751604800447e-05, + "loss": 1.2273, + "step": 895 + }, + { + "epoch": 0.06, + "learning_rate": 1.98295369157775e-05, + "loss": 1.2506, + "step": 896 + }, + { + "epoch": 0.06, + "learning_rate": 1.982932222675455e-05, + "loss": 1.2329, + "step": 897 + }, + { + "epoch": 0.06, + "learning_rate": 1.9829107537731598e-05, + "loss": 1.2332, + "step": 898 + }, + { + "epoch": 0.06, + "learning_rate": 1.9828892848708647e-05, + "loss": 1.2035, + "step": 899 + }, + { + "epoch": 0.06, + "learning_rate": 1.9828678159685696e-05, + "loss": 1.2539, + "step": 900 + }, + { + "epoch": 0.06, + "learning_rate": 1.9828463470662748e-05, + "loss": 1.236, + "step": 901 + }, + { + "epoch": 0.06, + "learning_rate": 1.9828248781639797e-05, + "loss": 1.3451, + "step": 902 + }, + { + "epoch": 0.06, + "learning_rate": 1.9828034092616846e-05, + "loss": 1.2005, + "step": 903 + }, + { + "epoch": 0.06, + "learning_rate": 1.9827819403593895e-05, + "loss": 1.3209, + "step": 904 + }, + { + "epoch": 0.06, + "learning_rate": 1.9827604714570944e-05, + "loss": 1.2647, + "step": 905 + }, + { + "epoch": 0.06, + "learning_rate": 1.9827390025547993e-05, + "loss": 1.2059, + "step": 906 + }, + { + "epoch": 0.06, + "learning_rate": 1.9827175336525045e-05, + "loss": 1.3972, + "step": 907 + }, + { + "epoch": 0.06, + "learning_rate": 1.9826960647502098e-05, + "loss": 1.2767, + "step": 908 + }, + { + "epoch": 0.06, + "learning_rate": 1.9826745958479147e-05, + "loss": 1.2407, + "step": 909 + }, + { + "epoch": 0.06, + "learning_rate": 1.9826531269456195e-05, + "loss": 1.2792, + "step": 910 + }, + { + "epoch": 0.06, + "learning_rate": 1.9826316580433244e-05, + "loss": 1.3056, + "step": 911 + }, + { + "epoch": 0.06, + "learning_rate": 1.9826101891410293e-05, + "loss": 1.2036, + "step": 912 + }, + { + "epoch": 0.06, + "learning_rate": 1.9825887202387342e-05, + "loss": 1.181, + "step": 913 + }, + { + "epoch": 0.06, + "learning_rate": 1.9825672513364395e-05, + "loss": 1.176, + "step": 914 + }, + { + "epoch": 0.06, + "learning_rate": 1.9825457824341444e-05, + "loss": 1.3131, + "step": 915 + }, + { + "epoch": 0.06, + "learning_rate": 1.9825243135318493e-05, + "loss": 1.2632, + "step": 916 + }, + { + "epoch": 0.06, + "learning_rate": 1.982502844629554e-05, + "loss": 1.232, + "step": 917 + }, + { + "epoch": 0.06, + "learning_rate": 1.982481375727259e-05, + "loss": 1.1167, + "step": 918 + }, + { + "epoch": 0.06, + "learning_rate": 1.9824599068249643e-05, + "loss": 1.1483, + "step": 919 + }, + { + "epoch": 0.06, + "learning_rate": 1.9824384379226692e-05, + "loss": 1.1485, + "step": 920 + }, + { + "epoch": 0.06, + "learning_rate": 1.982416969020374e-05, + "loss": 1.2264, + "step": 921 + }, + { + "epoch": 0.06, + "learning_rate": 1.982395500118079e-05, + "loss": 1.2298, + "step": 922 + }, + { + "epoch": 0.06, + "learning_rate": 1.9823740312157842e-05, + "loss": 1.171, + "step": 923 + }, + { + "epoch": 0.06, + "learning_rate": 1.982352562313489e-05, + "loss": 1.3125, + "step": 924 + }, + { + "epoch": 0.06, + "learning_rate": 1.982331093411194e-05, + "loss": 1.3058, + "step": 925 + }, + { + "epoch": 0.06, + "learning_rate": 1.982309624508899e-05, + "loss": 1.2694, + "step": 926 + }, + { + "epoch": 0.06, + "learning_rate": 1.982288155606604e-05, + "loss": 1.2323, + "step": 927 + }, + { + "epoch": 0.06, + "learning_rate": 1.982266686704309e-05, + "loss": 1.2136, + "step": 928 + }, + { + "epoch": 0.06, + "learning_rate": 1.982245217802014e-05, + "loss": 1.2163, + "step": 929 + }, + { + "epoch": 0.06, + "learning_rate": 1.982223748899719e-05, + "loss": 1.2395, + "step": 930 + }, + { + "epoch": 0.06, + "learning_rate": 1.9822022799974237e-05, + "loss": 1.1892, + "step": 931 + }, + { + "epoch": 0.06, + "learning_rate": 1.982180811095129e-05, + "loss": 1.3018, + "step": 932 + }, + { + "epoch": 0.06, + "learning_rate": 1.982159342192834e-05, + "loss": 1.2466, + "step": 933 + }, + { + "epoch": 0.06, + "learning_rate": 1.9821378732905388e-05, + "loss": 1.3064, + "step": 934 + }, + { + "epoch": 0.06, + "learning_rate": 1.9821164043882437e-05, + "loss": 1.2026, + "step": 935 + }, + { + "epoch": 0.06, + "learning_rate": 1.9820949354859486e-05, + "loss": 1.2187, + "step": 936 + }, + { + "epoch": 0.06, + "learning_rate": 1.9820734665836538e-05, + "loss": 1.2474, + "step": 937 + }, + { + "epoch": 0.06, + "learning_rate": 1.9820519976813587e-05, + "loss": 1.3399, + "step": 938 + }, + { + "epoch": 0.06, + "learning_rate": 1.982030528779064e-05, + "loss": 1.2461, + "step": 939 + }, + { + "epoch": 0.06, + "learning_rate": 1.9820090598767688e-05, + "loss": 1.2464, + "step": 940 + }, + { + "epoch": 0.06, + "learning_rate": 1.9819875909744737e-05, + "loss": 1.1312, + "step": 941 + }, + { + "epoch": 0.06, + "learning_rate": 1.9819661220721786e-05, + "loss": 1.3039, + "step": 942 + }, + { + "epoch": 0.06, + "learning_rate": 1.9819446531698835e-05, + "loss": 1.1874, + "step": 943 + }, + { + "epoch": 0.06, + "learning_rate": 1.9819231842675884e-05, + "loss": 1.2139, + "step": 944 + }, + { + "epoch": 0.06, + "learning_rate": 1.9819017153652936e-05, + "loss": 1.2547, + "step": 945 + }, + { + "epoch": 0.06, + "learning_rate": 1.9818802464629985e-05, + "loss": 1.2231, + "step": 946 + }, + { + "epoch": 0.06, + "learning_rate": 1.9818587775607034e-05, + "loss": 1.2451, + "step": 947 + }, + { + "epoch": 0.06, + "learning_rate": 1.9818373086584083e-05, + "loss": 1.1766, + "step": 948 + }, + { + "epoch": 0.06, + "learning_rate": 1.9818158397561132e-05, + "loss": 1.1823, + "step": 949 + }, + { + "epoch": 0.06, + "learning_rate": 1.9817943708538185e-05, + "loss": 1.3066, + "step": 950 + }, + { + "epoch": 0.06, + "learning_rate": 1.9817729019515234e-05, + "loss": 1.2017, + "step": 951 + }, + { + "epoch": 0.06, + "learning_rate": 1.9817514330492286e-05, + "loss": 1.3071, + "step": 952 + }, + { + "epoch": 0.06, + "learning_rate": 1.9817299641469335e-05, + "loss": 1.3468, + "step": 953 + }, + { + "epoch": 0.06, + "learning_rate": 1.9817084952446384e-05, + "loss": 1.2294, + "step": 954 + }, + { + "epoch": 0.06, + "learning_rate": 1.9816870263423433e-05, + "loss": 1.2246, + "step": 955 + }, + { + "epoch": 0.06, + "learning_rate": 1.9816655574400482e-05, + "loss": 1.1436, + "step": 956 + }, + { + "epoch": 0.06, + "learning_rate": 1.9816440885377534e-05, + "loss": 1.339, + "step": 957 + }, + { + "epoch": 0.06, + "learning_rate": 1.9816226196354583e-05, + "loss": 1.3121, + "step": 958 + }, + { + "epoch": 0.06, + "learning_rate": 1.9816011507331632e-05, + "loss": 1.2032, + "step": 959 + }, + { + "epoch": 0.06, + "learning_rate": 1.981579681830868e-05, + "loss": 1.2538, + "step": 960 + }, + { + "epoch": 0.06, + "learning_rate": 1.981558212928573e-05, + "loss": 1.1405, + "step": 961 + }, + { + "epoch": 0.06, + "learning_rate": 1.981536744026278e-05, + "loss": 1.1917, + "step": 962 + }, + { + "epoch": 0.06, + "learning_rate": 1.981515275123983e-05, + "loss": 1.2032, + "step": 963 + }, + { + "epoch": 0.06, + "learning_rate": 1.981493806221688e-05, + "loss": 1.1296, + "step": 964 + }, + { + "epoch": 0.06, + "learning_rate": 1.981472337319393e-05, + "loss": 1.1697, + "step": 965 + }, + { + "epoch": 0.06, + "learning_rate": 1.981450868417098e-05, + "loss": 1.2271, + "step": 966 + }, + { + "epoch": 0.06, + "learning_rate": 1.981429399514803e-05, + "loss": 1.1865, + "step": 967 + }, + { + "epoch": 0.06, + "learning_rate": 1.981407930612508e-05, + "loss": 1.2634, + "step": 968 + }, + { + "epoch": 0.06, + "learning_rate": 1.981386461710213e-05, + "loss": 1.1959, + "step": 969 + }, + { + "epoch": 0.06, + "learning_rate": 1.981364992807918e-05, + "loss": 1.1665, + "step": 970 + }, + { + "epoch": 0.06, + "learning_rate": 1.981343523905623e-05, + "loss": 1.2741, + "step": 971 + }, + { + "epoch": 0.06, + "learning_rate": 1.981322055003328e-05, + "loss": 1.2032, + "step": 972 + }, + { + "epoch": 0.06, + "learning_rate": 1.9813005861010328e-05, + "loss": 1.2639, + "step": 973 + }, + { + "epoch": 0.06, + "learning_rate": 1.9812791171987377e-05, + "loss": 1.3577, + "step": 974 + }, + { + "epoch": 0.06, + "learning_rate": 1.981257648296443e-05, + "loss": 1.2286, + "step": 975 + }, + { + "epoch": 0.06, + "learning_rate": 1.9812361793941478e-05, + "loss": 1.3496, + "step": 976 + }, + { + "epoch": 0.06, + "learning_rate": 1.9812147104918527e-05, + "loss": 1.2322, + "step": 977 + }, + { + "epoch": 0.06, + "learning_rate": 1.9811932415895576e-05, + "loss": 1.291, + "step": 978 + }, + { + "epoch": 0.06, + "learning_rate": 1.9811717726872625e-05, + "loss": 1.3375, + "step": 979 + }, + { + "epoch": 0.06, + "learning_rate": 1.9811503037849674e-05, + "loss": 1.1179, + "step": 980 + }, + { + "epoch": 0.06, + "learning_rate": 1.9811288348826726e-05, + "loss": 1.2068, + "step": 981 + }, + { + "epoch": 0.06, + "learning_rate": 1.9811073659803775e-05, + "loss": 1.3097, + "step": 982 + }, + { + "epoch": 0.06, + "learning_rate": 1.9810858970780828e-05, + "loss": 1.2738, + "step": 983 + }, + { + "epoch": 0.06, + "learning_rate": 1.9810644281757876e-05, + "loss": 1.1101, + "step": 984 + }, + { + "epoch": 0.06, + "learning_rate": 1.9810429592734925e-05, + "loss": 1.2541, + "step": 985 + }, + { + "epoch": 0.06, + "learning_rate": 1.9810214903711974e-05, + "loss": 1.2424, + "step": 986 + }, + { + "epoch": 0.06, + "learning_rate": 1.9810000214689023e-05, + "loss": 1.2792, + "step": 987 + }, + { + "epoch": 0.06, + "learning_rate": 1.9809785525666076e-05, + "loss": 1.3012, + "step": 988 + }, + { + "epoch": 0.06, + "learning_rate": 1.9809570836643125e-05, + "loss": 1.3041, + "step": 989 + }, + { + "epoch": 0.06, + "learning_rate": 1.9809356147620174e-05, + "loss": 1.1256, + "step": 990 + }, + { + "epoch": 0.06, + "learning_rate": 1.9809141458597223e-05, + "loss": 1.2683, + "step": 991 + }, + { + "epoch": 0.06, + "learning_rate": 1.980892676957427e-05, + "loss": 1.2169, + "step": 992 + }, + { + "epoch": 0.06, + "learning_rate": 1.980871208055132e-05, + "loss": 1.1461, + "step": 993 + }, + { + "epoch": 0.06, + "learning_rate": 1.9808497391528373e-05, + "loss": 1.2326, + "step": 994 + }, + { + "epoch": 0.06, + "learning_rate": 1.9808282702505425e-05, + "loss": 1.214, + "step": 995 + }, + { + "epoch": 0.06, + "learning_rate": 1.9808068013482474e-05, + "loss": 1.2799, + "step": 996 + }, + { + "epoch": 0.06, + "learning_rate": 1.9807853324459523e-05, + "loss": 1.2095, + "step": 997 + }, + { + "epoch": 0.06, + "learning_rate": 1.9807638635436572e-05, + "loss": 1.2562, + "step": 998 + }, + { + "epoch": 0.06, + "learning_rate": 1.980742394641362e-05, + "loss": 1.1647, + "step": 999 + }, + { + "epoch": 0.06, + "learning_rate": 1.980720925739067e-05, + "loss": 1.2964, + "step": 1000 + }, + { + "epoch": 0.06, + "eval_laion/cider": 0.4061030941176318, + "eval_laion/meteor": 0.36913355480596377, + "eval_laion/pred_num_words": 34.19140625, + "eval_laion/sacrebleu": 13.886638908718858, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.2321223020553589, + "eval_runtime": 92.783, + "eval_samples_per_second": 5.378, + "eval_steps_per_second": 0.345, + "step": 1000 + }, + { + "epoch": 0.06, + "learning_rate": 1.9806994568367722e-05, + "loss": 1.204, + "step": 1001 + }, + { + "epoch": 0.06, + "learning_rate": 1.980677987934477e-05, + "loss": 1.1887, + "step": 1002 + }, + { + "epoch": 0.06, + "learning_rate": 1.980656519032182e-05, + "loss": 1.2225, + "step": 1003 + }, + { + "epoch": 0.06, + "learning_rate": 1.980635050129887e-05, + "loss": 1.1969, + "step": 1004 + }, + { + "epoch": 0.06, + "learning_rate": 1.980613581227592e-05, + "loss": 1.2846, + "step": 1005 + }, + { + "epoch": 0.06, + "learning_rate": 1.980592112325297e-05, + "loss": 1.2798, + "step": 1006 + }, + { + "epoch": 0.06, + "learning_rate": 1.980570643423002e-05, + "loss": 1.2013, + "step": 1007 + }, + { + "epoch": 0.06, + "learning_rate": 1.980549174520707e-05, + "loss": 1.1936, + "step": 1008 + }, + { + "epoch": 0.06, + "learning_rate": 1.9805277056184118e-05, + "loss": 1.1472, + "step": 1009 + }, + { + "epoch": 0.06, + "learning_rate": 1.980506236716117e-05, + "loss": 1.1931, + "step": 1010 + }, + { + "epoch": 0.07, + "learning_rate": 1.980484767813822e-05, + "loss": 1.3104, + "step": 1011 + }, + { + "epoch": 0.07, + "learning_rate": 1.9804632989115268e-05, + "loss": 1.27, + "step": 1012 + }, + { + "epoch": 0.07, + "learning_rate": 1.9804418300092317e-05, + "loss": 1.1835, + "step": 1013 + }, + { + "epoch": 0.07, + "learning_rate": 1.980420361106937e-05, + "loss": 1.1944, + "step": 1014 + }, + { + "epoch": 0.07, + "learning_rate": 1.9803988922046418e-05, + "loss": 1.2053, + "step": 1015 + }, + { + "epoch": 0.07, + "learning_rate": 1.9803774233023467e-05, + "loss": 1.1867, + "step": 1016 + }, + { + "epoch": 0.07, + "learning_rate": 1.9803559544000516e-05, + "loss": 1.213, + "step": 1017 + }, + { + "epoch": 0.07, + "learning_rate": 1.9803344854977565e-05, + "loss": 1.2854, + "step": 1018 + }, + { + "epoch": 0.07, + "learning_rate": 1.9803130165954617e-05, + "loss": 1.2425, + "step": 1019 + }, + { + "epoch": 0.07, + "learning_rate": 1.9802915476931666e-05, + "loss": 1.217, + "step": 1020 + }, + { + "epoch": 0.07, + "learning_rate": 1.9802700787908715e-05, + "loss": 1.2321, + "step": 1021 + }, + { + "epoch": 0.07, + "learning_rate": 1.9802486098885764e-05, + "loss": 1.3939, + "step": 1022 + }, + { + "epoch": 0.07, + "learning_rate": 1.9802271409862813e-05, + "loss": 1.2661, + "step": 1023 + }, + { + "epoch": 0.07, + "learning_rate": 1.9802056720839866e-05, + "loss": 1.2187, + "step": 1024 + }, + { + "epoch": 0.07, + "learning_rate": 1.9801842031816915e-05, + "loss": 1.1911, + "step": 1025 + }, + { + "epoch": 0.07, + "learning_rate": 1.9801627342793967e-05, + "loss": 1.1403, + "step": 1026 + }, + { + "epoch": 0.07, + "learning_rate": 1.9801412653771016e-05, + "loss": 1.2479, + "step": 1027 + }, + { + "epoch": 0.07, + "learning_rate": 1.9801197964748065e-05, + "loss": 1.2887, + "step": 1028 + }, + { + "epoch": 0.07, + "learning_rate": 1.9800983275725114e-05, + "loss": 1.1316, + "step": 1029 + }, + { + "epoch": 0.07, + "learning_rate": 1.9800768586702163e-05, + "loss": 1.2818, + "step": 1030 + }, + { + "epoch": 0.07, + "learning_rate": 1.9800553897679212e-05, + "loss": 1.2306, + "step": 1031 + }, + { + "epoch": 0.07, + "learning_rate": 1.9800339208656264e-05, + "loss": 1.2133, + "step": 1032 + }, + { + "epoch": 0.07, + "learning_rate": 1.9800124519633313e-05, + "loss": 1.2731, + "step": 1033 + }, + { + "epoch": 0.07, + "learning_rate": 1.9799909830610362e-05, + "loss": 1.2836, + "step": 1034 + }, + { + "epoch": 0.07, + "learning_rate": 1.979969514158741e-05, + "loss": 1.2189, + "step": 1035 + }, + { + "epoch": 0.07, + "learning_rate": 1.979948045256446e-05, + "loss": 1.2504, + "step": 1036 + }, + { + "epoch": 0.07, + "learning_rate": 1.9799265763541512e-05, + "loss": 1.1951, + "step": 1037 + }, + { + "epoch": 0.07, + "learning_rate": 1.979905107451856e-05, + "loss": 1.2014, + "step": 1038 + }, + { + "epoch": 0.07, + "learning_rate": 1.9798836385495614e-05, + "loss": 1.2411, + "step": 1039 + }, + { + "epoch": 0.07, + "learning_rate": 1.9798621696472663e-05, + "loss": 1.1093, + "step": 1040 + }, + { + "epoch": 0.07, + "learning_rate": 1.979840700744971e-05, + "loss": 1.3067, + "step": 1041 + }, + { + "epoch": 0.07, + "learning_rate": 1.979819231842676e-05, + "loss": 1.1495, + "step": 1042 + }, + { + "epoch": 0.07, + "learning_rate": 1.979797762940381e-05, + "loss": 1.2376, + "step": 1043 + }, + { + "epoch": 0.07, + "learning_rate": 1.9797762940380862e-05, + "loss": 1.2938, + "step": 1044 + }, + { + "epoch": 0.07, + "learning_rate": 1.979754825135791e-05, + "loss": 1.1086, + "step": 1045 + }, + { + "epoch": 0.07, + "learning_rate": 1.979733356233496e-05, + "loss": 1.2189, + "step": 1046 + }, + { + "epoch": 0.07, + "learning_rate": 1.979711887331201e-05, + "loss": 1.2665, + "step": 1047 + }, + { + "epoch": 0.07, + "learning_rate": 1.9796904184289058e-05, + "loss": 1.2596, + "step": 1048 + }, + { + "epoch": 0.07, + "learning_rate": 1.9796689495266107e-05, + "loss": 1.2218, + "step": 1049 + }, + { + "epoch": 0.07, + "learning_rate": 1.979647480624316e-05, + "loss": 1.1629, + "step": 1050 + }, + { + "epoch": 0.07, + "learning_rate": 1.9796260117220208e-05, + "loss": 1.1553, + "step": 1051 + }, + { + "epoch": 0.07, + "learning_rate": 1.9796045428197257e-05, + "loss": 1.1843, + "step": 1052 + }, + { + "epoch": 0.07, + "learning_rate": 1.9795830739174306e-05, + "loss": 1.2314, + "step": 1053 + }, + { + "epoch": 0.07, + "learning_rate": 1.9795616050151358e-05, + "loss": 1.2139, + "step": 1054 + }, + { + "epoch": 0.07, + "learning_rate": 1.9795401361128407e-05, + "loss": 1.2679, + "step": 1055 + }, + { + "epoch": 0.07, + "learning_rate": 1.9795186672105456e-05, + "loss": 1.3483, + "step": 1056 + }, + { + "epoch": 0.07, + "learning_rate": 1.979497198308251e-05, + "loss": 1.1752, + "step": 1057 + }, + { + "epoch": 0.07, + "learning_rate": 1.9794757294059558e-05, + "loss": 1.2757, + "step": 1058 + }, + { + "epoch": 0.07, + "learning_rate": 1.9794542605036606e-05, + "loss": 1.1268, + "step": 1059 + }, + { + "epoch": 0.07, + "learning_rate": 1.9794327916013655e-05, + "loss": 1.2233, + "step": 1060 + }, + { + "epoch": 0.07, + "learning_rate": 1.9794113226990704e-05, + "loss": 1.2, + "step": 1061 + }, + { + "epoch": 0.07, + "learning_rate": 1.9793898537967757e-05, + "loss": 1.2077, + "step": 1062 + }, + { + "epoch": 0.07, + "learning_rate": 1.9793683848944806e-05, + "loss": 1.1871, + "step": 1063 + }, + { + "epoch": 0.07, + "learning_rate": 1.9793469159921855e-05, + "loss": 1.3104, + "step": 1064 + }, + { + "epoch": 0.07, + "learning_rate": 1.9793254470898904e-05, + "loss": 1.3399, + "step": 1065 + }, + { + "epoch": 0.07, + "learning_rate": 1.9793039781875953e-05, + "loss": 1.3197, + "step": 1066 + }, + { + "epoch": 0.07, + "learning_rate": 1.9792825092853e-05, + "loss": 1.2396, + "step": 1067 + }, + { + "epoch": 0.07, + "learning_rate": 1.9792610403830054e-05, + "loss": 1.2547, + "step": 1068 + }, + { + "epoch": 0.07, + "learning_rate": 1.9792395714807103e-05, + "loss": 1.1605, + "step": 1069 + }, + { + "epoch": 0.07, + "learning_rate": 1.9792181025784155e-05, + "loss": 1.1878, + "step": 1070 + }, + { + "epoch": 0.07, + "learning_rate": 1.9791966336761204e-05, + "loss": 1.2844, + "step": 1071 + }, + { + "epoch": 0.07, + "learning_rate": 1.9791751647738253e-05, + "loss": 1.1787, + "step": 1072 + }, + { + "epoch": 0.07, + "learning_rate": 1.9791536958715302e-05, + "loss": 1.1737, + "step": 1073 + }, + { + "epoch": 0.07, + "learning_rate": 1.979132226969235e-05, + "loss": 1.2003, + "step": 1074 + }, + { + "epoch": 0.07, + "learning_rate": 1.9791107580669403e-05, + "loss": 1.2303, + "step": 1075 + }, + { + "epoch": 0.07, + "learning_rate": 1.9790892891646452e-05, + "loss": 1.2843, + "step": 1076 + }, + { + "epoch": 0.07, + "learning_rate": 1.97906782026235e-05, + "loss": 1.3114, + "step": 1077 + }, + { + "epoch": 0.07, + "learning_rate": 1.979046351360055e-05, + "loss": 1.2545, + "step": 1078 + }, + { + "epoch": 0.07, + "learning_rate": 1.97902488245776e-05, + "loss": 1.3324, + "step": 1079 + }, + { + "epoch": 0.07, + "learning_rate": 1.9790034135554648e-05, + "loss": 1.1651, + "step": 1080 + }, + { + "epoch": 0.07, + "learning_rate": 1.97898194465317e-05, + "loss": 1.2027, + "step": 1081 + }, + { + "epoch": 0.07, + "learning_rate": 1.978960475750875e-05, + "loss": 1.1009, + "step": 1082 + }, + { + "epoch": 0.07, + "learning_rate": 1.9789390068485802e-05, + "loss": 1.192, + "step": 1083 + }, + { + "epoch": 0.07, + "learning_rate": 1.978917537946285e-05, + "loss": 1.1853, + "step": 1084 + }, + { + "epoch": 0.07, + "learning_rate": 1.97889606904399e-05, + "loss": 1.1287, + "step": 1085 + }, + { + "epoch": 0.07, + "learning_rate": 1.978874600141695e-05, + "loss": 1.2518, + "step": 1086 + }, + { + "epoch": 0.07, + "learning_rate": 1.9788531312393998e-05, + "loss": 1.1961, + "step": 1087 + }, + { + "epoch": 0.07, + "learning_rate": 1.978831662337105e-05, + "loss": 1.1202, + "step": 1088 + }, + { + "epoch": 0.07, + "learning_rate": 1.97881019343481e-05, + "loss": 1.2661, + "step": 1089 + }, + { + "epoch": 0.07, + "learning_rate": 1.9787887245325148e-05, + "loss": 1.2168, + "step": 1090 + }, + { + "epoch": 0.07, + "learning_rate": 1.9787672556302197e-05, + "loss": 1.2329, + "step": 1091 + }, + { + "epoch": 0.07, + "learning_rate": 1.9787457867279246e-05, + "loss": 1.2356, + "step": 1092 + }, + { + "epoch": 0.07, + "learning_rate": 1.97872431782563e-05, + "loss": 1.2077, + "step": 1093 + }, + { + "epoch": 0.07, + "learning_rate": 1.9787028489233347e-05, + "loss": 1.1767, + "step": 1094 + }, + { + "epoch": 0.07, + "learning_rate": 1.9786813800210396e-05, + "loss": 1.1114, + "step": 1095 + }, + { + "epoch": 0.07, + "learning_rate": 1.9786599111187445e-05, + "loss": 1.2608, + "step": 1096 + }, + { + "epoch": 0.07, + "learning_rate": 1.9786384422164498e-05, + "loss": 1.2646, + "step": 1097 + }, + { + "epoch": 0.07, + "learning_rate": 1.9786169733141547e-05, + "loss": 1.3002, + "step": 1098 + }, + { + "epoch": 0.07, + "learning_rate": 1.9785955044118596e-05, + "loss": 1.2539, + "step": 1099 + }, + { + "epoch": 0.07, + "learning_rate": 1.9785740355095645e-05, + "loss": 1.2247, + "step": 1100 + }, + { + "epoch": 0.07, + "learning_rate": 1.9785525666072697e-05, + "loss": 1.1447, + "step": 1101 + }, + { + "epoch": 0.07, + "learning_rate": 1.9785310977049746e-05, + "loss": 1.1611, + "step": 1102 + }, + { + "epoch": 0.07, + "learning_rate": 1.9785096288026795e-05, + "loss": 1.2177, + "step": 1103 + }, + { + "epoch": 0.07, + "learning_rate": 1.9784881599003844e-05, + "loss": 1.211, + "step": 1104 + }, + { + "epoch": 0.07, + "learning_rate": 1.9784666909980893e-05, + "loss": 1.1791, + "step": 1105 + }, + { + "epoch": 0.07, + "learning_rate": 1.9784452220957945e-05, + "loss": 1.2313, + "step": 1106 + }, + { + "epoch": 0.07, + "learning_rate": 1.9784237531934994e-05, + "loss": 1.1926, + "step": 1107 + }, + { + "epoch": 0.07, + "learning_rate": 1.9784022842912043e-05, + "loss": 1.165, + "step": 1108 + }, + { + "epoch": 0.07, + "learning_rate": 1.9783808153889092e-05, + "loss": 1.1874, + "step": 1109 + }, + { + "epoch": 0.07, + "learning_rate": 1.978359346486614e-05, + "loss": 1.2368, + "step": 1110 + }, + { + "epoch": 0.07, + "learning_rate": 1.9783378775843193e-05, + "loss": 1.3449, + "step": 1111 + }, + { + "epoch": 0.07, + "learning_rate": 1.9783164086820242e-05, + "loss": 1.1961, + "step": 1112 + }, + { + "epoch": 0.07, + "learning_rate": 1.9782949397797295e-05, + "loss": 1.2201, + "step": 1113 + }, + { + "epoch": 0.07, + "learning_rate": 1.9782734708774344e-05, + "loss": 1.2485, + "step": 1114 + }, + { + "epoch": 0.07, + "learning_rate": 1.9782520019751393e-05, + "loss": 1.1665, + "step": 1115 + }, + { + "epoch": 0.07, + "learning_rate": 1.978230533072844e-05, + "loss": 1.2387, + "step": 1116 + }, + { + "epoch": 0.07, + "learning_rate": 1.978209064170549e-05, + "loss": 1.1594, + "step": 1117 + }, + { + "epoch": 0.07, + "learning_rate": 1.978187595268254e-05, + "loss": 1.3221, + "step": 1118 + }, + { + "epoch": 0.07, + "learning_rate": 1.9781661263659592e-05, + "loss": 1.1965, + "step": 1119 + }, + { + "epoch": 0.07, + "learning_rate": 1.978144657463664e-05, + "loss": 1.2438, + "step": 1120 + }, + { + "epoch": 0.07, + "learning_rate": 1.978123188561369e-05, + "loss": 1.2214, + "step": 1121 + }, + { + "epoch": 0.07, + "learning_rate": 1.978101719659074e-05, + "loss": 1.2125, + "step": 1122 + }, + { + "epoch": 0.07, + "learning_rate": 1.9780802507567788e-05, + "loss": 1.2462, + "step": 1123 + }, + { + "epoch": 0.07, + "learning_rate": 1.978058781854484e-05, + "loss": 1.212, + "step": 1124 + }, + { + "epoch": 0.07, + "learning_rate": 1.978037312952189e-05, + "loss": 1.2541, + "step": 1125 + }, + { + "epoch": 0.07, + "learning_rate": 1.978015844049894e-05, + "loss": 1.2103, + "step": 1126 + }, + { + "epoch": 0.07, + "learning_rate": 1.977994375147599e-05, + "loss": 1.3448, + "step": 1127 + }, + { + "epoch": 0.07, + "learning_rate": 1.977972906245304e-05, + "loss": 1.1536, + "step": 1128 + }, + { + "epoch": 0.07, + "learning_rate": 1.9779514373430088e-05, + "loss": 1.2497, + "step": 1129 + }, + { + "epoch": 0.07, + "learning_rate": 1.9779299684407137e-05, + "loss": 1.4028, + "step": 1130 + }, + { + "epoch": 0.07, + "learning_rate": 1.977908499538419e-05, + "loss": 1.2769, + "step": 1131 + }, + { + "epoch": 0.07, + "learning_rate": 1.977887030636124e-05, + "loss": 1.292, + "step": 1132 + }, + { + "epoch": 0.07, + "learning_rate": 1.9778655617338287e-05, + "loss": 1.21, + "step": 1133 + }, + { + "epoch": 0.07, + "learning_rate": 1.9778440928315336e-05, + "loss": 1.2054, + "step": 1134 + }, + { + "epoch": 0.07, + "learning_rate": 1.9778226239292385e-05, + "loss": 1.1846, + "step": 1135 + }, + { + "epoch": 0.07, + "learning_rate": 1.9778011550269434e-05, + "loss": 1.1271, + "step": 1136 + }, + { + "epoch": 0.07, + "learning_rate": 1.9777796861246487e-05, + "loss": 1.2817, + "step": 1137 + }, + { + "epoch": 0.07, + "learning_rate": 1.9777582172223536e-05, + "loss": 1.1865, + "step": 1138 + }, + { + "epoch": 0.07, + "learning_rate": 1.9777367483200585e-05, + "loss": 1.2144, + "step": 1139 + }, + { + "epoch": 0.07, + "learning_rate": 1.9777152794177634e-05, + "loss": 1.228, + "step": 1140 + }, + { + "epoch": 0.07, + "learning_rate": 1.9776938105154686e-05, + "loss": 1.2701, + "step": 1141 + }, + { + "epoch": 0.07, + "learning_rate": 1.9776723416131735e-05, + "loss": 1.2582, + "step": 1142 + }, + { + "epoch": 0.07, + "learning_rate": 1.9776508727108784e-05, + "loss": 1.1793, + "step": 1143 + }, + { + "epoch": 0.07, + "learning_rate": 1.9776294038085836e-05, + "loss": 1.1463, + "step": 1144 + }, + { + "epoch": 0.07, + "learning_rate": 1.9776079349062885e-05, + "loss": 1.2364, + "step": 1145 + }, + { + "epoch": 0.07, + "learning_rate": 1.9775864660039934e-05, + "loss": 1.1604, + "step": 1146 + }, + { + "epoch": 0.07, + "learning_rate": 1.9775649971016983e-05, + "loss": 1.2539, + "step": 1147 + }, + { + "epoch": 0.07, + "learning_rate": 1.9775435281994032e-05, + "loss": 1.2078, + "step": 1148 + }, + { + "epoch": 0.07, + "learning_rate": 1.977522059297108e-05, + "loss": 1.2282, + "step": 1149 + }, + { + "epoch": 0.07, + "learning_rate": 1.9775005903948133e-05, + "loss": 1.3704, + "step": 1150 + }, + { + "epoch": 0.07, + "learning_rate": 1.9774791214925182e-05, + "loss": 1.1704, + "step": 1151 + }, + { + "epoch": 0.07, + "learning_rate": 1.977457652590223e-05, + "loss": 1.2382, + "step": 1152 + }, + { + "epoch": 0.07, + "learning_rate": 1.977436183687928e-05, + "loss": 1.2324, + "step": 1153 + }, + { + "epoch": 0.07, + "learning_rate": 1.977414714785633e-05, + "loss": 1.2307, + "step": 1154 + }, + { + "epoch": 0.07, + "learning_rate": 1.977393245883338e-05, + "loss": 1.1719, + "step": 1155 + }, + { + "epoch": 0.07, + "learning_rate": 1.977371776981043e-05, + "loss": 1.2163, + "step": 1156 + }, + { + "epoch": 0.07, + "learning_rate": 1.9773503080787483e-05, + "loss": 1.2947, + "step": 1157 + }, + { + "epoch": 0.07, + "learning_rate": 1.9773288391764532e-05, + "loss": 1.2032, + "step": 1158 + }, + { + "epoch": 0.07, + "learning_rate": 1.977307370274158e-05, + "loss": 1.3635, + "step": 1159 + }, + { + "epoch": 0.07, + "learning_rate": 1.977285901371863e-05, + "loss": 1.2142, + "step": 1160 + }, + { + "epoch": 0.07, + "learning_rate": 1.977264432469568e-05, + "loss": 1.2271, + "step": 1161 + }, + { + "epoch": 0.07, + "learning_rate": 1.977242963567273e-05, + "loss": 1.2398, + "step": 1162 + }, + { + "epoch": 0.07, + "learning_rate": 1.977221494664978e-05, + "loss": 1.2289, + "step": 1163 + }, + { + "epoch": 0.07, + "learning_rate": 1.977200025762683e-05, + "loss": 1.2491, + "step": 1164 + }, + { + "epoch": 0.07, + "learning_rate": 1.9771785568603878e-05, + "loss": 1.1847, + "step": 1165 + }, + { + "epoch": 0.08, + "learning_rate": 1.9771570879580927e-05, + "loss": 1.1802, + "step": 1166 + }, + { + "epoch": 0.08, + "learning_rate": 1.9771356190557976e-05, + "loss": 1.1204, + "step": 1167 + }, + { + "epoch": 0.08, + "learning_rate": 1.977114150153503e-05, + "loss": 1.1478, + "step": 1168 + }, + { + "epoch": 0.08, + "learning_rate": 1.9770926812512077e-05, + "loss": 1.2379, + "step": 1169 + }, + { + "epoch": 0.08, + "learning_rate": 1.977071212348913e-05, + "loss": 1.2392, + "step": 1170 + }, + { + "epoch": 0.08, + "learning_rate": 1.977049743446618e-05, + "loss": 1.1873, + "step": 1171 + }, + { + "epoch": 0.08, + "learning_rate": 1.9770282745443228e-05, + "loss": 1.2544, + "step": 1172 + }, + { + "epoch": 0.08, + "learning_rate": 1.9770068056420277e-05, + "loss": 1.3034, + "step": 1173 + }, + { + "epoch": 0.08, + "learning_rate": 1.9769853367397326e-05, + "loss": 1.2022, + "step": 1174 + }, + { + "epoch": 0.08, + "learning_rate": 1.9769638678374378e-05, + "loss": 1.1837, + "step": 1175 + }, + { + "epoch": 0.08, + "learning_rate": 1.9769423989351427e-05, + "loss": 1.1398, + "step": 1176 + }, + { + "epoch": 0.08, + "learning_rate": 1.9769209300328476e-05, + "loss": 1.2349, + "step": 1177 + }, + { + "epoch": 0.08, + "learning_rate": 1.9768994611305525e-05, + "loss": 1.2062, + "step": 1178 + }, + { + "epoch": 0.08, + "learning_rate": 1.9768779922282574e-05, + "loss": 1.2072, + "step": 1179 + }, + { + "epoch": 0.08, + "learning_rate": 1.9768565233259626e-05, + "loss": 1.1927, + "step": 1180 + }, + { + "epoch": 0.08, + "learning_rate": 1.9768350544236675e-05, + "loss": 1.1381, + "step": 1181 + }, + { + "epoch": 0.08, + "learning_rate": 1.9768135855213724e-05, + "loss": 1.1314, + "step": 1182 + }, + { + "epoch": 0.08, + "learning_rate": 1.9767921166190773e-05, + "loss": 1.136, + "step": 1183 + }, + { + "epoch": 0.08, + "learning_rate": 1.9767706477167825e-05, + "loss": 1.1464, + "step": 1184 + }, + { + "epoch": 0.08, + "learning_rate": 1.9767491788144874e-05, + "loss": 1.2487, + "step": 1185 + }, + { + "epoch": 0.08, + "learning_rate": 1.9767277099121923e-05, + "loss": 1.144, + "step": 1186 + }, + { + "epoch": 0.08, + "learning_rate": 1.9767062410098972e-05, + "loss": 1.1304, + "step": 1187 + }, + { + "epoch": 0.08, + "learning_rate": 1.9766847721076025e-05, + "loss": 1.0767, + "step": 1188 + }, + { + "epoch": 0.08, + "learning_rate": 1.9766633032053074e-05, + "loss": 1.1656, + "step": 1189 + }, + { + "epoch": 0.08, + "learning_rate": 1.9766418343030123e-05, + "loss": 1.2545, + "step": 1190 + }, + { + "epoch": 0.08, + "learning_rate": 1.976620365400717e-05, + "loss": 1.1106, + "step": 1191 + }, + { + "epoch": 0.08, + "learning_rate": 1.976598896498422e-05, + "loss": 1.2583, + "step": 1192 + }, + { + "epoch": 0.08, + "learning_rate": 1.9765774275961273e-05, + "loss": 1.1948, + "step": 1193 + }, + { + "epoch": 0.08, + "learning_rate": 1.9765559586938322e-05, + "loss": 1.1933, + "step": 1194 + }, + { + "epoch": 0.08, + "learning_rate": 1.976534489791537e-05, + "loss": 1.149, + "step": 1195 + }, + { + "epoch": 0.08, + "learning_rate": 1.976513020889242e-05, + "loss": 1.292, + "step": 1196 + }, + { + "epoch": 0.08, + "learning_rate": 1.976491551986947e-05, + "loss": 1.2338, + "step": 1197 + }, + { + "epoch": 0.08, + "learning_rate": 1.976470083084652e-05, + "loss": 1.2021, + "step": 1198 + }, + { + "epoch": 0.08, + "learning_rate": 1.976448614182357e-05, + "loss": 1.3188, + "step": 1199 + }, + { + "epoch": 0.08, + "learning_rate": 1.9764271452800622e-05, + "loss": 1.2274, + "step": 1200 + }, + { + "epoch": 0.08, + "learning_rate": 1.976405676377767e-05, + "loss": 1.2473, + "step": 1201 + }, + { + "epoch": 0.08, + "learning_rate": 1.976384207475472e-05, + "loss": 1.2957, + "step": 1202 + }, + { + "epoch": 0.08, + "learning_rate": 1.976362738573177e-05, + "loss": 1.1486, + "step": 1203 + }, + { + "epoch": 0.08, + "learning_rate": 1.9763412696708818e-05, + "loss": 1.2678, + "step": 1204 + }, + { + "epoch": 0.08, + "learning_rate": 1.9763198007685867e-05, + "loss": 1.1919, + "step": 1205 + }, + { + "epoch": 0.08, + "learning_rate": 1.976298331866292e-05, + "loss": 1.1512, + "step": 1206 + }, + { + "epoch": 0.08, + "learning_rate": 1.976276862963997e-05, + "loss": 1.1566, + "step": 1207 + }, + { + "epoch": 0.08, + "learning_rate": 1.9762553940617017e-05, + "loss": 1.1961, + "step": 1208 + }, + { + "epoch": 0.08, + "learning_rate": 1.9762339251594066e-05, + "loss": 1.1597, + "step": 1209 + }, + { + "epoch": 0.08, + "learning_rate": 1.9762124562571115e-05, + "loss": 1.1165, + "step": 1210 + }, + { + "epoch": 0.08, + "learning_rate": 1.9761909873548168e-05, + "loss": 1.1855, + "step": 1211 + }, + { + "epoch": 0.08, + "learning_rate": 1.9761695184525217e-05, + "loss": 1.1335, + "step": 1212 + }, + { + "epoch": 0.08, + "learning_rate": 1.976148049550227e-05, + "loss": 1.236, + "step": 1213 + }, + { + "epoch": 0.08, + "learning_rate": 1.9761265806479318e-05, + "loss": 1.2579, + "step": 1214 + }, + { + "epoch": 0.08, + "learning_rate": 1.9761051117456367e-05, + "loss": 1.2992, + "step": 1215 + }, + { + "epoch": 0.08, + "learning_rate": 1.9760836428433416e-05, + "loss": 1.1069, + "step": 1216 + }, + { + "epoch": 0.08, + "learning_rate": 1.9760621739410465e-05, + "loss": 1.2313, + "step": 1217 + }, + { + "epoch": 0.08, + "learning_rate": 1.9760407050387517e-05, + "loss": 1.1288, + "step": 1218 + }, + { + "epoch": 0.08, + "learning_rate": 1.9760192361364566e-05, + "loss": 1.1644, + "step": 1219 + }, + { + "epoch": 0.08, + "learning_rate": 1.9759977672341615e-05, + "loss": 1.126, + "step": 1220 + }, + { + "epoch": 0.08, + "learning_rate": 1.9759762983318664e-05, + "loss": 1.2019, + "step": 1221 + }, + { + "epoch": 0.08, + "learning_rate": 1.9759548294295713e-05, + "loss": 1.1246, + "step": 1222 + }, + { + "epoch": 0.08, + "learning_rate": 1.9759333605272762e-05, + "loss": 1.1762, + "step": 1223 + }, + { + "epoch": 0.08, + "learning_rate": 1.9759118916249814e-05, + "loss": 1.1885, + "step": 1224 + }, + { + "epoch": 0.08, + "learning_rate": 1.9758904227226863e-05, + "loss": 1.2973, + "step": 1225 + }, + { + "epoch": 0.08, + "learning_rate": 1.9758689538203912e-05, + "loss": 1.2743, + "step": 1226 + }, + { + "epoch": 0.08, + "learning_rate": 1.975847484918096e-05, + "loss": 1.2774, + "step": 1227 + }, + { + "epoch": 0.08, + "learning_rate": 1.9758260160158014e-05, + "loss": 1.249, + "step": 1228 + }, + { + "epoch": 0.08, + "learning_rate": 1.9758045471135063e-05, + "loss": 1.1541, + "step": 1229 + }, + { + "epoch": 0.08, + "learning_rate": 1.975783078211211e-05, + "loss": 1.1494, + "step": 1230 + }, + { + "epoch": 0.08, + "learning_rate": 1.9757616093089164e-05, + "loss": 1.2141, + "step": 1231 + }, + { + "epoch": 0.08, + "learning_rate": 1.9757401404066213e-05, + "loss": 1.2797, + "step": 1232 + }, + { + "epoch": 0.08, + "learning_rate": 1.9757186715043262e-05, + "loss": 1.2827, + "step": 1233 + }, + { + "epoch": 0.08, + "learning_rate": 1.975697202602031e-05, + "loss": 1.1232, + "step": 1234 + }, + { + "epoch": 0.08, + "learning_rate": 1.975675733699736e-05, + "loss": 1.1775, + "step": 1235 + }, + { + "epoch": 0.08, + "learning_rate": 1.975654264797441e-05, + "loss": 1.3074, + "step": 1236 + }, + { + "epoch": 0.08, + "learning_rate": 1.975632795895146e-05, + "loss": 1.1875, + "step": 1237 + }, + { + "epoch": 0.08, + "learning_rate": 1.975611326992851e-05, + "loss": 1.2621, + "step": 1238 + }, + { + "epoch": 0.08, + "learning_rate": 1.975589858090556e-05, + "loss": 1.3221, + "step": 1239 + }, + { + "epoch": 0.08, + "learning_rate": 1.9755683891882608e-05, + "loss": 1.27, + "step": 1240 + }, + { + "epoch": 0.08, + "learning_rate": 1.9755469202859657e-05, + "loss": 1.2445, + "step": 1241 + }, + { + "epoch": 0.08, + "learning_rate": 1.975525451383671e-05, + "loss": 1.3289, + "step": 1242 + }, + { + "epoch": 0.08, + "learning_rate": 1.975503982481376e-05, + "loss": 1.159, + "step": 1243 + }, + { + "epoch": 0.08, + "learning_rate": 1.975482513579081e-05, + "loss": 1.2635, + "step": 1244 + }, + { + "epoch": 0.08, + "learning_rate": 1.975461044676786e-05, + "loss": 1.2034, + "step": 1245 + }, + { + "epoch": 0.08, + "learning_rate": 1.975439575774491e-05, + "loss": 1.1639, + "step": 1246 + }, + { + "epoch": 0.08, + "learning_rate": 1.9754181068721958e-05, + "loss": 1.1595, + "step": 1247 + }, + { + "epoch": 0.08, + "learning_rate": 1.9753966379699007e-05, + "loss": 1.2268, + "step": 1248 + }, + { + "epoch": 0.08, + "learning_rate": 1.975375169067606e-05, + "loss": 1.1376, + "step": 1249 + }, + { + "epoch": 0.08, + "learning_rate": 1.9753537001653108e-05, + "loss": 1.1575, + "step": 1250 + }, + { + "epoch": 0.08, + "learning_rate": 1.9753322312630157e-05, + "loss": 1.1672, + "step": 1251 + }, + { + "epoch": 0.08, + "learning_rate": 1.9753107623607206e-05, + "loss": 1.2666, + "step": 1252 + }, + { + "epoch": 0.08, + "learning_rate": 1.9752892934584255e-05, + "loss": 1.2691, + "step": 1253 + }, + { + "epoch": 0.08, + "learning_rate": 1.9752678245561304e-05, + "loss": 1.1889, + "step": 1254 + }, + { + "epoch": 0.08, + "learning_rate": 1.9752463556538356e-05, + "loss": 1.2606, + "step": 1255 + }, + { + "epoch": 0.08, + "learning_rate": 1.9752248867515405e-05, + "loss": 1.0892, + "step": 1256 + }, + { + "epoch": 0.08, + "learning_rate": 1.9752034178492457e-05, + "loss": 1.277, + "step": 1257 + }, + { + "epoch": 0.08, + "learning_rate": 1.9751819489469506e-05, + "loss": 1.2973, + "step": 1258 + }, + { + "epoch": 0.08, + "learning_rate": 1.9751604800446555e-05, + "loss": 1.2142, + "step": 1259 + }, + { + "epoch": 0.08, + "learning_rate": 1.9751390111423604e-05, + "loss": 1.2649, + "step": 1260 + }, + { + "epoch": 0.08, + "learning_rate": 1.9751175422400653e-05, + "loss": 1.1479, + "step": 1261 + }, + { + "epoch": 0.08, + "learning_rate": 1.9750960733377706e-05, + "loss": 1.3107, + "step": 1262 + }, + { + "epoch": 0.08, + "learning_rate": 1.9750746044354755e-05, + "loss": 1.17, + "step": 1263 + }, + { + "epoch": 0.08, + "learning_rate": 1.9750531355331804e-05, + "loss": 1.1596, + "step": 1264 + }, + { + "epoch": 0.08, + "learning_rate": 1.9750316666308852e-05, + "loss": 1.1811, + "step": 1265 + }, + { + "epoch": 0.08, + "learning_rate": 1.97501019772859e-05, + "loss": 1.2486, + "step": 1266 + }, + { + "epoch": 0.08, + "learning_rate": 1.9749887288262954e-05, + "loss": 1.2067, + "step": 1267 + }, + { + "epoch": 0.08, + "learning_rate": 1.9749672599240003e-05, + "loss": 1.2362, + "step": 1268 + }, + { + "epoch": 0.08, + "learning_rate": 1.9749457910217052e-05, + "loss": 1.1863, + "step": 1269 + }, + { + "epoch": 0.08, + "learning_rate": 1.97492432211941e-05, + "loss": 1.1432, + "step": 1270 + }, + { + "epoch": 0.08, + "learning_rate": 1.9749028532171153e-05, + "loss": 1.176, + "step": 1271 + }, + { + "epoch": 0.08, + "learning_rate": 1.9748813843148202e-05, + "loss": 1.1998, + "step": 1272 + }, + { + "epoch": 0.08, + "learning_rate": 1.974859915412525e-05, + "loss": 1.206, + "step": 1273 + }, + { + "epoch": 0.08, + "learning_rate": 1.97483844651023e-05, + "loss": 1.1575, + "step": 1274 + }, + { + "epoch": 0.08, + "learning_rate": 1.9748169776079352e-05, + "loss": 1.1307, + "step": 1275 + }, + { + "epoch": 0.08, + "learning_rate": 1.97479550870564e-05, + "loss": 1.2497, + "step": 1276 + }, + { + "epoch": 0.08, + "learning_rate": 1.974774039803345e-05, + "loss": 1.1631, + "step": 1277 + }, + { + "epoch": 0.08, + "learning_rate": 1.97475257090105e-05, + "loss": 1.2634, + "step": 1278 + }, + { + "epoch": 0.08, + "learning_rate": 1.9747311019987548e-05, + "loss": 1.1939, + "step": 1279 + }, + { + "epoch": 0.08, + "learning_rate": 1.97470963309646e-05, + "loss": 1.2339, + "step": 1280 + }, + { + "epoch": 0.08, + "learning_rate": 1.974688164194165e-05, + "loss": 1.2467, + "step": 1281 + }, + { + "epoch": 0.08, + "learning_rate": 1.97466669529187e-05, + "loss": 1.3212, + "step": 1282 + }, + { + "epoch": 0.08, + "learning_rate": 1.9746452263895747e-05, + "loss": 1.1184, + "step": 1283 + }, + { + "epoch": 0.08, + "learning_rate": 1.9746237574872796e-05, + "loss": 1.2576, + "step": 1284 + }, + { + "epoch": 0.08, + "learning_rate": 1.974602288584985e-05, + "loss": 1.2682, + "step": 1285 + }, + { + "epoch": 0.08, + "learning_rate": 1.9745808196826898e-05, + "loss": 1.2693, + "step": 1286 + }, + { + "epoch": 0.08, + "learning_rate": 1.974559350780395e-05, + "loss": 1.1337, + "step": 1287 + }, + { + "epoch": 0.08, + "learning_rate": 1.9745378818781e-05, + "loss": 1.2037, + "step": 1288 + }, + { + "epoch": 0.08, + "learning_rate": 1.9745164129758048e-05, + "loss": 1.1392, + "step": 1289 + }, + { + "epoch": 0.08, + "learning_rate": 1.9744949440735097e-05, + "loss": 1.2046, + "step": 1290 + }, + { + "epoch": 0.08, + "learning_rate": 1.9744734751712146e-05, + "loss": 1.2, + "step": 1291 + }, + { + "epoch": 0.08, + "learning_rate": 1.9744520062689195e-05, + "loss": 1.1948, + "step": 1292 + }, + { + "epoch": 0.08, + "learning_rate": 1.9744305373666247e-05, + "loss": 1.1979, + "step": 1293 + }, + { + "epoch": 0.08, + "learning_rate": 1.9744090684643296e-05, + "loss": 1.1551, + "step": 1294 + }, + { + "epoch": 0.08, + "learning_rate": 1.9743875995620345e-05, + "loss": 1.2809, + "step": 1295 + }, + { + "epoch": 0.08, + "learning_rate": 1.9743661306597394e-05, + "loss": 1.1712, + "step": 1296 + }, + { + "epoch": 0.08, + "learning_rate": 1.9743446617574443e-05, + "loss": 1.219, + "step": 1297 + }, + { + "epoch": 0.08, + "learning_rate": 1.9743231928551495e-05, + "loss": 1.1621, + "step": 1298 + }, + { + "epoch": 0.08, + "learning_rate": 1.9743017239528544e-05, + "loss": 1.1905, + "step": 1299 + }, + { + "epoch": 0.08, + "learning_rate": 1.9742802550505597e-05, + "loss": 1.2518, + "step": 1300 + }, + { + "epoch": 0.08, + "learning_rate": 1.9742587861482646e-05, + "loss": 1.2395, + "step": 1301 + }, + { + "epoch": 0.08, + "learning_rate": 1.9742373172459695e-05, + "loss": 1.2054, + "step": 1302 + }, + { + "epoch": 0.08, + "learning_rate": 1.9742158483436744e-05, + "loss": 1.188, + "step": 1303 + }, + { + "epoch": 0.08, + "learning_rate": 1.9741943794413793e-05, + "loss": 1.2064, + "step": 1304 + }, + { + "epoch": 0.08, + "learning_rate": 1.9741729105390845e-05, + "loss": 1.1685, + "step": 1305 + }, + { + "epoch": 0.08, + "learning_rate": 1.9741514416367894e-05, + "loss": 1.2232, + "step": 1306 + }, + { + "epoch": 0.08, + "learning_rate": 1.9741299727344943e-05, + "loss": 1.1919, + "step": 1307 + }, + { + "epoch": 0.08, + "learning_rate": 1.9741085038321992e-05, + "loss": 1.0119, + "step": 1308 + }, + { + "epoch": 0.08, + "learning_rate": 1.974087034929904e-05, + "loss": 1.3104, + "step": 1309 + }, + { + "epoch": 0.08, + "learning_rate": 1.974065566027609e-05, + "loss": 1.2243, + "step": 1310 + }, + { + "epoch": 0.08, + "learning_rate": 1.9740440971253142e-05, + "loss": 1.2106, + "step": 1311 + }, + { + "epoch": 0.08, + "learning_rate": 1.974022628223019e-05, + "loss": 1.1889, + "step": 1312 + }, + { + "epoch": 0.08, + "learning_rate": 1.974001159320724e-05, + "loss": 1.1292, + "step": 1313 + }, + { + "epoch": 0.08, + "learning_rate": 1.973979690418429e-05, + "loss": 1.2357, + "step": 1314 + }, + { + "epoch": 0.08, + "learning_rate": 1.973958221516134e-05, + "loss": 1.1893, + "step": 1315 + }, + { + "epoch": 0.08, + "learning_rate": 1.973936752613839e-05, + "loss": 1.1643, + "step": 1316 + }, + { + "epoch": 0.08, + "learning_rate": 1.973915283711544e-05, + "loss": 1.1938, + "step": 1317 + }, + { + "epoch": 0.08, + "learning_rate": 1.973893814809249e-05, + "loss": 1.1708, + "step": 1318 + }, + { + "epoch": 0.08, + "learning_rate": 1.973872345906954e-05, + "loss": 1.1674, + "step": 1319 + }, + { + "epoch": 0.08, + "learning_rate": 1.973850877004659e-05, + "loss": 1.2253, + "step": 1320 + }, + { + "epoch": 0.08, + "learning_rate": 1.973829408102364e-05, + "loss": 1.2232, + "step": 1321 + }, + { + "epoch": 0.09, + "learning_rate": 1.9738079392000688e-05, + "loss": 1.1083, + "step": 1322 + }, + { + "epoch": 0.09, + "learning_rate": 1.9737864702977736e-05, + "loss": 1.0723, + "step": 1323 + }, + { + "epoch": 0.09, + "learning_rate": 1.973765001395479e-05, + "loss": 1.1838, + "step": 1324 + }, + { + "epoch": 0.09, + "learning_rate": 1.9737435324931838e-05, + "loss": 1.2202, + "step": 1325 + }, + { + "epoch": 0.09, + "learning_rate": 1.9737220635908887e-05, + "loss": 1.2637, + "step": 1326 + }, + { + "epoch": 0.09, + "learning_rate": 1.9737005946885936e-05, + "loss": 1.1868, + "step": 1327 + }, + { + "epoch": 0.09, + "learning_rate": 1.9736791257862985e-05, + "loss": 1.2073, + "step": 1328 + }, + { + "epoch": 0.09, + "learning_rate": 1.9736576568840037e-05, + "loss": 1.3066, + "step": 1329 + }, + { + "epoch": 0.09, + "learning_rate": 1.9736361879817086e-05, + "loss": 1.2953, + "step": 1330 + }, + { + "epoch": 0.09, + "learning_rate": 1.973614719079414e-05, + "loss": 1.107, + "step": 1331 + }, + { + "epoch": 0.09, + "learning_rate": 1.9735932501771187e-05, + "loss": 1.0577, + "step": 1332 + }, + { + "epoch": 0.09, + "learning_rate": 1.9735717812748236e-05, + "loss": 1.261, + "step": 1333 + }, + { + "epoch": 0.09, + "learning_rate": 1.9735503123725285e-05, + "loss": 1.2048, + "step": 1334 + }, + { + "epoch": 0.09, + "learning_rate": 1.9735288434702334e-05, + "loss": 1.2749, + "step": 1335 + }, + { + "epoch": 0.09, + "learning_rate": 1.9735073745679387e-05, + "loss": 1.0983, + "step": 1336 + }, + { + "epoch": 0.09, + "learning_rate": 1.9734859056656436e-05, + "loss": 1.2123, + "step": 1337 + }, + { + "epoch": 0.09, + "learning_rate": 1.9734644367633485e-05, + "loss": 1.1733, + "step": 1338 + }, + { + "epoch": 0.09, + "learning_rate": 1.9734429678610533e-05, + "loss": 1.2322, + "step": 1339 + }, + { + "epoch": 0.09, + "learning_rate": 1.9734214989587582e-05, + "loss": 1.2315, + "step": 1340 + }, + { + "epoch": 0.09, + "learning_rate": 1.973400030056463e-05, + "loss": 1.2058, + "step": 1341 + }, + { + "epoch": 0.09, + "learning_rate": 1.9733785611541684e-05, + "loss": 1.1804, + "step": 1342 + }, + { + "epoch": 0.09, + "learning_rate": 1.9733570922518733e-05, + "loss": 1.1583, + "step": 1343 + }, + { + "epoch": 0.09, + "learning_rate": 1.9733356233495785e-05, + "loss": 1.1911, + "step": 1344 + }, + { + "epoch": 0.09, + "learning_rate": 1.9733141544472834e-05, + "loss": 1.2165, + "step": 1345 + }, + { + "epoch": 0.09, + "learning_rate": 1.9732926855449883e-05, + "loss": 1.2317, + "step": 1346 + }, + { + "epoch": 0.09, + "learning_rate": 1.9732712166426932e-05, + "loss": 1.0999, + "step": 1347 + }, + { + "epoch": 0.09, + "learning_rate": 1.973249747740398e-05, + "loss": 1.2231, + "step": 1348 + }, + { + "epoch": 0.09, + "learning_rate": 1.9732282788381033e-05, + "loss": 1.2449, + "step": 1349 + }, + { + "epoch": 0.09, + "learning_rate": 1.9732068099358082e-05, + "loss": 1.2391, + "step": 1350 + }, + { + "epoch": 0.09, + "learning_rate": 1.973185341033513e-05, + "loss": 1.1664, + "step": 1351 + }, + { + "epoch": 0.09, + "learning_rate": 1.973163872131218e-05, + "loss": 1.1668, + "step": 1352 + }, + { + "epoch": 0.09, + "learning_rate": 1.973142403228923e-05, + "loss": 1.1833, + "step": 1353 + }, + { + "epoch": 0.09, + "learning_rate": 1.973120934326628e-05, + "loss": 1.2554, + "step": 1354 + }, + { + "epoch": 0.09, + "learning_rate": 1.973099465424333e-05, + "loss": 1.2622, + "step": 1355 + }, + { + "epoch": 0.09, + "learning_rate": 1.973077996522038e-05, + "loss": 1.1884, + "step": 1356 + }, + { + "epoch": 0.09, + "learning_rate": 1.973056527619743e-05, + "loss": 1.1964, + "step": 1357 + }, + { + "epoch": 0.09, + "learning_rate": 1.973035058717448e-05, + "loss": 1.1829, + "step": 1358 + }, + { + "epoch": 0.09, + "learning_rate": 1.973013589815153e-05, + "loss": 1.152, + "step": 1359 + }, + { + "epoch": 0.09, + "learning_rate": 1.972992120912858e-05, + "loss": 1.2205, + "step": 1360 + }, + { + "epoch": 0.09, + "learning_rate": 1.9729706520105628e-05, + "loss": 1.2007, + "step": 1361 + }, + { + "epoch": 0.09, + "learning_rate": 1.972949183108268e-05, + "loss": 1.2065, + "step": 1362 + }, + { + "epoch": 0.09, + "learning_rate": 1.972927714205973e-05, + "loss": 1.1821, + "step": 1363 + }, + { + "epoch": 0.09, + "learning_rate": 1.9729062453036778e-05, + "loss": 1.1307, + "step": 1364 + }, + { + "epoch": 0.09, + "learning_rate": 1.9728847764013827e-05, + "loss": 1.2437, + "step": 1365 + }, + { + "epoch": 0.09, + "learning_rate": 1.9728633074990876e-05, + "loss": 1.1746, + "step": 1366 + }, + { + "epoch": 0.09, + "learning_rate": 1.9728418385967928e-05, + "loss": 1.2874, + "step": 1367 + }, + { + "epoch": 0.09, + "learning_rate": 1.9728203696944977e-05, + "loss": 1.2871, + "step": 1368 + }, + { + "epoch": 0.09, + "learning_rate": 1.9727989007922026e-05, + "loss": 1.1235, + "step": 1369 + }, + { + "epoch": 0.09, + "learning_rate": 1.9727774318899075e-05, + "loss": 1.1681, + "step": 1370 + }, + { + "epoch": 0.09, + "learning_rate": 1.9727559629876124e-05, + "loss": 1.1711, + "step": 1371 + }, + { + "epoch": 0.09, + "learning_rate": 1.9727344940853176e-05, + "loss": 1.1661, + "step": 1372 + }, + { + "epoch": 0.09, + "learning_rate": 1.9727130251830225e-05, + "loss": 1.185, + "step": 1373 + }, + { + "epoch": 0.09, + "learning_rate": 1.9726915562807278e-05, + "loss": 1.2199, + "step": 1374 + }, + { + "epoch": 0.09, + "learning_rate": 1.9726700873784327e-05, + "loss": 1.1548, + "step": 1375 + }, + { + "epoch": 0.09, + "learning_rate": 1.9726486184761376e-05, + "loss": 1.264, + "step": 1376 + }, + { + "epoch": 0.09, + "learning_rate": 1.9726271495738425e-05, + "loss": 1.1464, + "step": 1377 + }, + { + "epoch": 0.09, + "learning_rate": 1.9726056806715474e-05, + "loss": 1.186, + "step": 1378 + }, + { + "epoch": 0.09, + "learning_rate": 1.9725842117692523e-05, + "loss": 1.251, + "step": 1379 + }, + { + "epoch": 0.09, + "learning_rate": 1.9725627428669575e-05, + "loss": 1.2668, + "step": 1380 + }, + { + "epoch": 0.09, + "learning_rate": 1.9725412739646624e-05, + "loss": 1.1227, + "step": 1381 + }, + { + "epoch": 0.09, + "learning_rate": 1.9725198050623673e-05, + "loss": 1.199, + "step": 1382 + }, + { + "epoch": 0.09, + "learning_rate": 1.9724983361600722e-05, + "loss": 1.1994, + "step": 1383 + }, + { + "epoch": 0.09, + "learning_rate": 1.972476867257777e-05, + "loss": 1.1378, + "step": 1384 + }, + { + "epoch": 0.09, + "learning_rate": 1.9724553983554823e-05, + "loss": 1.1375, + "step": 1385 + }, + { + "epoch": 0.09, + "learning_rate": 1.9724339294531872e-05, + "loss": 1.2511, + "step": 1386 + }, + { + "epoch": 0.09, + "learning_rate": 1.9724124605508924e-05, + "loss": 1.282, + "step": 1387 + }, + { + "epoch": 0.09, + "learning_rate": 1.9723909916485973e-05, + "loss": 1.0846, + "step": 1388 + }, + { + "epoch": 0.09, + "learning_rate": 1.9723695227463022e-05, + "loss": 1.1917, + "step": 1389 + }, + { + "epoch": 0.09, + "learning_rate": 1.972348053844007e-05, + "loss": 1.1381, + "step": 1390 + }, + { + "epoch": 0.09, + "learning_rate": 1.972326584941712e-05, + "loss": 1.2912, + "step": 1391 + }, + { + "epoch": 0.09, + "learning_rate": 1.9723051160394173e-05, + "loss": 1.139, + "step": 1392 + }, + { + "epoch": 0.09, + "learning_rate": 1.972283647137122e-05, + "loss": 1.2303, + "step": 1393 + }, + { + "epoch": 0.09, + "learning_rate": 1.972262178234827e-05, + "loss": 1.2343, + "step": 1394 + }, + { + "epoch": 0.09, + "learning_rate": 1.972240709332532e-05, + "loss": 1.067, + "step": 1395 + }, + { + "epoch": 0.09, + "learning_rate": 1.972219240430237e-05, + "loss": 1.1329, + "step": 1396 + }, + { + "epoch": 0.09, + "learning_rate": 1.9721977715279418e-05, + "loss": 1.22, + "step": 1397 + }, + { + "epoch": 0.09, + "learning_rate": 1.972176302625647e-05, + "loss": 1.0906, + "step": 1398 + }, + { + "epoch": 0.09, + "learning_rate": 1.972154833723352e-05, + "loss": 1.2044, + "step": 1399 + }, + { + "epoch": 0.09, + "learning_rate": 1.9721333648210568e-05, + "loss": 1.2253, + "step": 1400 + }, + { + "epoch": 0.09, + "learning_rate": 1.9721118959187617e-05, + "loss": 1.2134, + "step": 1401 + }, + { + "epoch": 0.09, + "learning_rate": 1.972090427016467e-05, + "loss": 1.3101, + "step": 1402 + }, + { + "epoch": 0.09, + "learning_rate": 1.9720689581141718e-05, + "loss": 1.3768, + "step": 1403 + }, + { + "epoch": 0.09, + "learning_rate": 1.9720474892118767e-05, + "loss": 1.2963, + "step": 1404 + }, + { + "epoch": 0.09, + "learning_rate": 1.972026020309582e-05, + "loss": 1.145, + "step": 1405 + }, + { + "epoch": 0.09, + "learning_rate": 1.972004551407287e-05, + "loss": 1.1394, + "step": 1406 + }, + { + "epoch": 0.09, + "learning_rate": 1.9719830825049917e-05, + "loss": 1.2007, + "step": 1407 + }, + { + "epoch": 0.09, + "learning_rate": 1.9719616136026966e-05, + "loss": 1.2192, + "step": 1408 + }, + { + "epoch": 0.09, + "learning_rate": 1.9719401447004015e-05, + "loss": 1.2958, + "step": 1409 + }, + { + "epoch": 0.09, + "learning_rate": 1.9719186757981064e-05, + "loss": 1.2127, + "step": 1410 + }, + { + "epoch": 0.09, + "learning_rate": 1.9718972068958117e-05, + "loss": 1.1142, + "step": 1411 + }, + { + "epoch": 0.09, + "learning_rate": 1.9718757379935166e-05, + "loss": 1.1807, + "step": 1412 + }, + { + "epoch": 0.09, + "learning_rate": 1.9718542690912214e-05, + "loss": 1.1355, + "step": 1413 + }, + { + "epoch": 0.09, + "learning_rate": 1.9718328001889263e-05, + "loss": 1.2543, + "step": 1414 + }, + { + "epoch": 0.09, + "learning_rate": 1.9718113312866312e-05, + "loss": 1.321, + "step": 1415 + }, + { + "epoch": 0.09, + "learning_rate": 1.9717898623843365e-05, + "loss": 1.2319, + "step": 1416 + }, + { + "epoch": 0.09, + "learning_rate": 1.9717683934820414e-05, + "loss": 1.1149, + "step": 1417 + }, + { + "epoch": 0.09, + "learning_rate": 1.9717469245797466e-05, + "loss": 1.1211, + "step": 1418 + }, + { + "epoch": 0.09, + "learning_rate": 1.9717254556774515e-05, + "loss": 1.1736, + "step": 1419 + }, + { + "epoch": 0.09, + "learning_rate": 1.9717039867751564e-05, + "loss": 1.2945, + "step": 1420 + }, + { + "epoch": 0.09, + "learning_rate": 1.9716825178728613e-05, + "loss": 1.1629, + "step": 1421 + }, + { + "epoch": 0.09, + "learning_rate": 1.9716610489705662e-05, + "loss": 1.2616, + "step": 1422 + }, + { + "epoch": 0.09, + "learning_rate": 1.9716395800682714e-05, + "loss": 1.1101, + "step": 1423 + }, + { + "epoch": 0.09, + "learning_rate": 1.9716181111659763e-05, + "loss": 1.1622, + "step": 1424 + }, + { + "epoch": 0.09, + "learning_rate": 1.9715966422636812e-05, + "loss": 1.2062, + "step": 1425 + }, + { + "epoch": 0.09, + "learning_rate": 1.971575173361386e-05, + "loss": 1.2204, + "step": 1426 + }, + { + "epoch": 0.09, + "learning_rate": 1.971553704459091e-05, + "loss": 1.24, + "step": 1427 + }, + { + "epoch": 0.09, + "learning_rate": 1.971532235556796e-05, + "loss": 1.2914, + "step": 1428 + }, + { + "epoch": 0.09, + "learning_rate": 1.971510766654501e-05, + "loss": 1.22, + "step": 1429 + }, + { + "epoch": 0.09, + "learning_rate": 1.971489297752206e-05, + "loss": 1.2406, + "step": 1430 + }, + { + "epoch": 0.09, + "learning_rate": 1.9714678288499113e-05, + "loss": 1.1549, + "step": 1431 + }, + { + "epoch": 0.09, + "learning_rate": 1.9714463599476162e-05, + "loss": 1.1518, + "step": 1432 + }, + { + "epoch": 0.09, + "learning_rate": 1.971424891045321e-05, + "loss": 1.1886, + "step": 1433 + }, + { + "epoch": 0.09, + "learning_rate": 1.971403422143026e-05, + "loss": 1.1462, + "step": 1434 + }, + { + "epoch": 0.09, + "learning_rate": 1.971381953240731e-05, + "loss": 1.1321, + "step": 1435 + }, + { + "epoch": 0.09, + "learning_rate": 1.971360484338436e-05, + "loss": 1.3143, + "step": 1436 + }, + { + "epoch": 0.09, + "learning_rate": 1.971339015436141e-05, + "loss": 1.1243, + "step": 1437 + }, + { + "epoch": 0.09, + "learning_rate": 1.971317546533846e-05, + "loss": 1.1144, + "step": 1438 + }, + { + "epoch": 0.09, + "learning_rate": 1.9712960776315508e-05, + "loss": 1.2292, + "step": 1439 + }, + { + "epoch": 0.09, + "learning_rate": 1.9712746087292557e-05, + "loss": 1.2229, + "step": 1440 + }, + { + "epoch": 0.09, + "learning_rate": 1.971253139826961e-05, + "loss": 1.1581, + "step": 1441 + }, + { + "epoch": 0.09, + "learning_rate": 1.9712316709246658e-05, + "loss": 1.2466, + "step": 1442 + }, + { + "epoch": 0.09, + "learning_rate": 1.9712102020223707e-05, + "loss": 1.0996, + "step": 1443 + }, + { + "epoch": 0.09, + "learning_rate": 1.9711887331200756e-05, + "loss": 1.3098, + "step": 1444 + }, + { + "epoch": 0.09, + "learning_rate": 1.9711672642177805e-05, + "loss": 1.2276, + "step": 1445 + }, + { + "epoch": 0.09, + "learning_rate": 1.9711457953154857e-05, + "loss": 1.2774, + "step": 1446 + }, + { + "epoch": 0.09, + "learning_rate": 1.9711243264131906e-05, + "loss": 1.2373, + "step": 1447 + }, + { + "epoch": 0.09, + "learning_rate": 1.9711028575108955e-05, + "loss": 1.1486, + "step": 1448 + }, + { + "epoch": 0.09, + "learning_rate": 1.9710813886086008e-05, + "loss": 1.198, + "step": 1449 + }, + { + "epoch": 0.09, + "learning_rate": 1.9710599197063057e-05, + "loss": 1.164, + "step": 1450 + }, + { + "epoch": 0.09, + "learning_rate": 1.9710384508040106e-05, + "loss": 1.293, + "step": 1451 + }, + { + "epoch": 0.09, + "learning_rate": 1.9710169819017155e-05, + "loss": 1.1515, + "step": 1452 + }, + { + "epoch": 0.09, + "learning_rate": 1.9709955129994204e-05, + "loss": 1.2291, + "step": 1453 + }, + { + "epoch": 0.09, + "learning_rate": 1.9709740440971256e-05, + "loss": 1.2851, + "step": 1454 + }, + { + "epoch": 0.09, + "learning_rate": 1.9709525751948305e-05, + "loss": 1.2151, + "step": 1455 + }, + { + "epoch": 0.09, + "learning_rate": 1.9709311062925354e-05, + "loss": 1.2015, + "step": 1456 + }, + { + "epoch": 0.09, + "learning_rate": 1.9709096373902403e-05, + "loss": 1.212, + "step": 1457 + }, + { + "epoch": 0.09, + "learning_rate": 1.9708881684879452e-05, + "loss": 1.2551, + "step": 1458 + }, + { + "epoch": 0.09, + "learning_rate": 1.97086669958565e-05, + "loss": 1.184, + "step": 1459 + }, + { + "epoch": 0.09, + "learning_rate": 1.9708452306833553e-05, + "loss": 1.0909, + "step": 1460 + }, + { + "epoch": 0.09, + "learning_rate": 1.9708237617810605e-05, + "loss": 1.1201, + "step": 1461 + }, + { + "epoch": 0.09, + "learning_rate": 1.9708022928787654e-05, + "loss": 1.2172, + "step": 1462 + }, + { + "epoch": 0.09, + "learning_rate": 1.9707808239764703e-05, + "loss": 1.1502, + "step": 1463 + }, + { + "epoch": 0.09, + "learning_rate": 1.9707593550741752e-05, + "loss": 1.2024, + "step": 1464 + }, + { + "epoch": 0.09, + "learning_rate": 1.97073788617188e-05, + "loss": 1.1917, + "step": 1465 + }, + { + "epoch": 0.09, + "learning_rate": 1.970716417269585e-05, + "loss": 1.2691, + "step": 1466 + }, + { + "epoch": 0.09, + "learning_rate": 1.9706949483672903e-05, + "loss": 1.2469, + "step": 1467 + }, + { + "epoch": 0.09, + "learning_rate": 1.970673479464995e-05, + "loss": 1.2043, + "step": 1468 + }, + { + "epoch": 0.09, + "learning_rate": 1.9706520105627e-05, + "loss": 1.2135, + "step": 1469 + }, + { + "epoch": 0.09, + "learning_rate": 1.970630541660405e-05, + "loss": 1.2363, + "step": 1470 + }, + { + "epoch": 0.09, + "learning_rate": 1.97060907275811e-05, + "loss": 1.2028, + "step": 1471 + }, + { + "epoch": 0.09, + "learning_rate": 1.970587603855815e-05, + "loss": 1.1474, + "step": 1472 + }, + { + "epoch": 0.09, + "learning_rate": 1.97056613495352e-05, + "loss": 1.1883, + "step": 1473 + }, + { + "epoch": 0.09, + "learning_rate": 1.970544666051225e-05, + "loss": 1.1354, + "step": 1474 + }, + { + "epoch": 0.09, + "learning_rate": 1.97052319714893e-05, + "loss": 1.1563, + "step": 1475 + }, + { + "epoch": 0.09, + "learning_rate": 1.970501728246635e-05, + "loss": 1.3477, + "step": 1476 + }, + { + "epoch": 0.1, + "learning_rate": 1.97048025934434e-05, + "loss": 1.1807, + "step": 1477 + }, + { + "epoch": 0.1, + "learning_rate": 1.9704587904420448e-05, + "loss": 1.2983, + "step": 1478 + }, + { + "epoch": 0.1, + "learning_rate": 1.97043732153975e-05, + "loss": 1.2365, + "step": 1479 + }, + { + "epoch": 0.1, + "learning_rate": 1.970415852637455e-05, + "loss": 1.1271, + "step": 1480 + }, + { + "epoch": 0.1, + "learning_rate": 1.97039438373516e-05, + "loss": 1.1842, + "step": 1481 + }, + { + "epoch": 0.1, + "learning_rate": 1.9703729148328647e-05, + "loss": 1.1854, + "step": 1482 + }, + { + "epoch": 0.1, + "learning_rate": 1.9703514459305696e-05, + "loss": 1.2145, + "step": 1483 + }, + { + "epoch": 0.1, + "learning_rate": 1.9703299770282745e-05, + "loss": 1.2608, + "step": 1484 + }, + { + "epoch": 0.1, + "learning_rate": 1.9703085081259798e-05, + "loss": 1.1622, + "step": 1485 + }, + { + "epoch": 0.1, + "learning_rate": 1.9702870392236847e-05, + "loss": 1.1319, + "step": 1486 + }, + { + "epoch": 0.1, + "learning_rate": 1.9702655703213895e-05, + "loss": 1.123, + "step": 1487 + }, + { + "epoch": 0.1, + "learning_rate": 1.9702441014190944e-05, + "loss": 1.1633, + "step": 1488 + }, + { + "epoch": 0.1, + "learning_rate": 1.9702226325167997e-05, + "loss": 1.2308, + "step": 1489 + }, + { + "epoch": 0.1, + "learning_rate": 1.9702011636145046e-05, + "loss": 1.2147, + "step": 1490 + }, + { + "epoch": 0.1, + "learning_rate": 1.9701796947122095e-05, + "loss": 1.102, + "step": 1491 + }, + { + "epoch": 0.1, + "learning_rate": 1.9701582258099147e-05, + "loss": 1.0962, + "step": 1492 + }, + { + "epoch": 0.1, + "learning_rate": 1.9701367569076196e-05, + "loss": 1.2706, + "step": 1493 + }, + { + "epoch": 0.1, + "learning_rate": 1.9701152880053245e-05, + "loss": 1.2264, + "step": 1494 + }, + { + "epoch": 0.1, + "learning_rate": 1.9700938191030294e-05, + "loss": 1.2008, + "step": 1495 + }, + { + "epoch": 0.1, + "learning_rate": 1.9700723502007343e-05, + "loss": 1.2013, + "step": 1496 + }, + { + "epoch": 0.1, + "learning_rate": 1.9700508812984392e-05, + "loss": 1.1922, + "step": 1497 + }, + { + "epoch": 0.1, + "learning_rate": 1.9700294123961444e-05, + "loss": 1.165, + "step": 1498 + }, + { + "epoch": 0.1, + "learning_rate": 1.9700079434938493e-05, + "loss": 1.1554, + "step": 1499 + }, + { + "epoch": 0.1, + "learning_rate": 1.9699864745915542e-05, + "loss": 1.2208, + "step": 1500 + }, + { + "epoch": 0.1, + "learning_rate": 1.969965005689259e-05, + "loss": 1.1871, + "step": 1501 + }, + { + "epoch": 0.1, + "learning_rate": 1.969943536786964e-05, + "loss": 1.1664, + "step": 1502 + }, + { + "epoch": 0.1, + "learning_rate": 1.9699220678846692e-05, + "loss": 1.192, + "step": 1503 + }, + { + "epoch": 0.1, + "learning_rate": 1.969900598982374e-05, + "loss": 1.1453, + "step": 1504 + }, + { + "epoch": 0.1, + "learning_rate": 1.9698791300800794e-05, + "loss": 1.2984, + "step": 1505 + }, + { + "epoch": 0.1, + "learning_rate": 1.9698576611777843e-05, + "loss": 1.1342, + "step": 1506 + }, + { + "epoch": 0.1, + "learning_rate": 1.9698361922754892e-05, + "loss": 1.2285, + "step": 1507 + }, + { + "epoch": 0.1, + "learning_rate": 1.969814723373194e-05, + "loss": 1.1199, + "step": 1508 + }, + { + "epoch": 0.1, + "learning_rate": 1.969793254470899e-05, + "loss": 1.0254, + "step": 1509 + }, + { + "epoch": 0.1, + "learning_rate": 1.9697717855686042e-05, + "loss": 1.227, + "step": 1510 + }, + { + "epoch": 0.1, + "learning_rate": 1.969750316666309e-05, + "loss": 1.1918, + "step": 1511 + }, + { + "epoch": 0.1, + "learning_rate": 1.969728847764014e-05, + "loss": 1.1219, + "step": 1512 + }, + { + "epoch": 0.1, + "learning_rate": 1.969707378861719e-05, + "loss": 1.1971, + "step": 1513 + }, + { + "epoch": 0.1, + "learning_rate": 1.9696859099594238e-05, + "loss": 1.1665, + "step": 1514 + }, + { + "epoch": 0.1, + "learning_rate": 1.9696644410571287e-05, + "loss": 1.2301, + "step": 1515 + }, + { + "epoch": 0.1, + "learning_rate": 1.969642972154834e-05, + "loss": 1.1714, + "step": 1516 + }, + { + "epoch": 0.1, + "learning_rate": 1.9696215032525388e-05, + "loss": 1.2214, + "step": 1517 + }, + { + "epoch": 0.1, + "learning_rate": 1.969600034350244e-05, + "loss": 1.1636, + "step": 1518 + }, + { + "epoch": 0.1, + "learning_rate": 1.969578565447949e-05, + "loss": 1.2497, + "step": 1519 + }, + { + "epoch": 0.1, + "learning_rate": 1.969557096545654e-05, + "loss": 1.239, + "step": 1520 + }, + { + "epoch": 0.1, + "learning_rate": 1.9695356276433587e-05, + "loss": 1.2311, + "step": 1521 + }, + { + "epoch": 0.1, + "learning_rate": 1.9695141587410636e-05, + "loss": 1.2134, + "step": 1522 + }, + { + "epoch": 0.1, + "learning_rate": 1.969492689838769e-05, + "loss": 1.0687, + "step": 1523 + }, + { + "epoch": 0.1, + "learning_rate": 1.9694712209364738e-05, + "loss": 1.1006, + "step": 1524 + }, + { + "epoch": 0.1, + "learning_rate": 1.9694497520341787e-05, + "loss": 1.1755, + "step": 1525 + }, + { + "epoch": 0.1, + "learning_rate": 1.9694282831318836e-05, + "loss": 1.1113, + "step": 1526 + }, + { + "epoch": 0.1, + "learning_rate": 1.9694068142295885e-05, + "loss": 1.2037, + "step": 1527 + }, + { + "epoch": 0.1, + "learning_rate": 1.9693853453272937e-05, + "loss": 1.1059, + "step": 1528 + }, + { + "epoch": 0.1, + "learning_rate": 1.9693638764249986e-05, + "loss": 1.2674, + "step": 1529 + }, + { + "epoch": 0.1, + "learning_rate": 1.9693424075227035e-05, + "loss": 1.1047, + "step": 1530 + }, + { + "epoch": 0.1, + "learning_rate": 1.9693209386204084e-05, + "loss": 1.2869, + "step": 1531 + }, + { + "epoch": 0.1, + "learning_rate": 1.9692994697181133e-05, + "loss": 1.2106, + "step": 1532 + }, + { + "epoch": 0.1, + "learning_rate": 1.9692780008158185e-05, + "loss": 1.1567, + "step": 1533 + }, + { + "epoch": 0.1, + "learning_rate": 1.9692565319135234e-05, + "loss": 1.3113, + "step": 1534 + }, + { + "epoch": 0.1, + "learning_rate": 1.9692350630112283e-05, + "loss": 1.1662, + "step": 1535 + }, + { + "epoch": 0.1, + "learning_rate": 1.9692135941089335e-05, + "loss": 1.204, + "step": 1536 + }, + { + "epoch": 0.1, + "learning_rate": 1.9691921252066384e-05, + "loss": 1.1462, + "step": 1537 + }, + { + "epoch": 0.1, + "learning_rate": 1.9691706563043433e-05, + "loss": 1.1541, + "step": 1538 + }, + { + "epoch": 0.1, + "learning_rate": 1.9691491874020482e-05, + "loss": 1.2413, + "step": 1539 + }, + { + "epoch": 0.1, + "learning_rate": 1.969127718499753e-05, + "loss": 1.2861, + "step": 1540 + }, + { + "epoch": 0.1, + "learning_rate": 1.9691062495974584e-05, + "loss": 1.1056, + "step": 1541 + }, + { + "epoch": 0.1, + "learning_rate": 1.9690847806951633e-05, + "loss": 1.1588, + "step": 1542 + }, + { + "epoch": 0.1, + "learning_rate": 1.969063311792868e-05, + "loss": 1.1627, + "step": 1543 + }, + { + "epoch": 0.1, + "learning_rate": 1.969041842890573e-05, + "loss": 1.107, + "step": 1544 + }, + { + "epoch": 0.1, + "learning_rate": 1.969020373988278e-05, + "loss": 1.1185, + "step": 1545 + }, + { + "epoch": 0.1, + "learning_rate": 1.968998905085983e-05, + "loss": 1.1177, + "step": 1546 + }, + { + "epoch": 0.1, + "learning_rate": 1.968977436183688e-05, + "loss": 1.2219, + "step": 1547 + }, + { + "epoch": 0.1, + "learning_rate": 1.9689559672813933e-05, + "loss": 1.2039, + "step": 1548 + }, + { + "epoch": 0.1, + "learning_rate": 1.9689344983790982e-05, + "loss": 1.1115, + "step": 1549 + }, + { + "epoch": 0.1, + "learning_rate": 1.968913029476803e-05, + "loss": 1.1733, + "step": 1550 + }, + { + "epoch": 0.1, + "learning_rate": 1.968891560574508e-05, + "loss": 1.1421, + "step": 1551 + }, + { + "epoch": 0.1, + "learning_rate": 1.968870091672213e-05, + "loss": 1.1801, + "step": 1552 + }, + { + "epoch": 0.1, + "learning_rate": 1.9688486227699178e-05, + "loss": 1.1553, + "step": 1553 + }, + { + "epoch": 0.1, + "learning_rate": 1.968827153867623e-05, + "loss": 1.2341, + "step": 1554 + }, + { + "epoch": 0.1, + "learning_rate": 1.968805684965328e-05, + "loss": 1.1416, + "step": 1555 + }, + { + "epoch": 0.1, + "learning_rate": 1.9687842160630328e-05, + "loss": 1.1725, + "step": 1556 + }, + { + "epoch": 0.1, + "learning_rate": 1.9687627471607377e-05, + "loss": 1.1256, + "step": 1557 + }, + { + "epoch": 0.1, + "learning_rate": 1.9687412782584426e-05, + "loss": 1.1395, + "step": 1558 + }, + { + "epoch": 0.1, + "learning_rate": 1.968719809356148e-05, + "loss": 1.126, + "step": 1559 + }, + { + "epoch": 0.1, + "learning_rate": 1.9686983404538528e-05, + "loss": 1.1769, + "step": 1560 + }, + { + "epoch": 0.1, + "learning_rate": 1.9686768715515577e-05, + "loss": 1.1923, + "step": 1561 + }, + { + "epoch": 0.1, + "learning_rate": 1.968655402649263e-05, + "loss": 1.2451, + "step": 1562 + }, + { + "epoch": 0.1, + "learning_rate": 1.9686339337469678e-05, + "loss": 1.181, + "step": 1563 + }, + { + "epoch": 0.1, + "learning_rate": 1.9686124648446727e-05, + "loss": 1.2379, + "step": 1564 + }, + { + "epoch": 0.1, + "learning_rate": 1.9685909959423776e-05, + "loss": 1.1465, + "step": 1565 + }, + { + "epoch": 0.1, + "learning_rate": 1.9685695270400825e-05, + "loss": 1.2188, + "step": 1566 + }, + { + "epoch": 0.1, + "learning_rate": 1.9685480581377877e-05, + "loss": 1.184, + "step": 1567 + }, + { + "epoch": 0.1, + "learning_rate": 1.9685265892354926e-05, + "loss": 1.1711, + "step": 1568 + }, + { + "epoch": 0.1, + "learning_rate": 1.9685051203331975e-05, + "loss": 1.1657, + "step": 1569 + }, + { + "epoch": 0.1, + "learning_rate": 1.9684836514309024e-05, + "loss": 1.1974, + "step": 1570 + }, + { + "epoch": 0.1, + "learning_rate": 1.9684621825286073e-05, + "loss": 1.1713, + "step": 1571 + }, + { + "epoch": 0.1, + "learning_rate": 1.9684407136263125e-05, + "loss": 1.1588, + "step": 1572 + }, + { + "epoch": 0.1, + "learning_rate": 1.9684192447240174e-05, + "loss": 1.2142, + "step": 1573 + }, + { + "epoch": 0.1, + "learning_rate": 1.9683977758217223e-05, + "loss": 1.2675, + "step": 1574 + }, + { + "epoch": 0.1, + "learning_rate": 1.9683763069194272e-05, + "loss": 1.1178, + "step": 1575 + }, + { + "epoch": 0.1, + "learning_rate": 1.9683548380171325e-05, + "loss": 1.2873, + "step": 1576 + }, + { + "epoch": 0.1, + "learning_rate": 1.9683333691148373e-05, + "loss": 1.1946, + "step": 1577 + }, + { + "epoch": 0.1, + "learning_rate": 1.9683119002125422e-05, + "loss": 1.1589, + "step": 1578 + }, + { + "epoch": 0.1, + "learning_rate": 1.9682904313102475e-05, + "loss": 1.2, + "step": 1579 + }, + { + "epoch": 0.1, + "learning_rate": 1.9682689624079524e-05, + "loss": 1.1829, + "step": 1580 + }, + { + "epoch": 0.1, + "learning_rate": 1.9682474935056573e-05, + "loss": 1.2162, + "step": 1581 + }, + { + "epoch": 0.1, + "learning_rate": 1.9682260246033622e-05, + "loss": 1.1611, + "step": 1582 + }, + { + "epoch": 0.1, + "learning_rate": 1.968204555701067e-05, + "loss": 1.2168, + "step": 1583 + }, + { + "epoch": 0.1, + "learning_rate": 1.968183086798772e-05, + "loss": 1.2508, + "step": 1584 + }, + { + "epoch": 0.1, + "learning_rate": 1.9681616178964772e-05, + "loss": 1.2641, + "step": 1585 + }, + { + "epoch": 0.1, + "learning_rate": 1.968140148994182e-05, + "loss": 1.2012, + "step": 1586 + }, + { + "epoch": 0.1, + "learning_rate": 1.968118680091887e-05, + "loss": 1.2272, + "step": 1587 + }, + { + "epoch": 0.1, + "learning_rate": 1.968097211189592e-05, + "loss": 1.2376, + "step": 1588 + }, + { + "epoch": 0.1, + "learning_rate": 1.9680757422872968e-05, + "loss": 1.1132, + "step": 1589 + }, + { + "epoch": 0.1, + "learning_rate": 1.968054273385002e-05, + "loss": 1.1868, + "step": 1590 + }, + { + "epoch": 0.1, + "learning_rate": 1.968032804482707e-05, + "loss": 1.2249, + "step": 1591 + }, + { + "epoch": 0.1, + "learning_rate": 1.968011335580412e-05, + "loss": 1.2426, + "step": 1592 + }, + { + "epoch": 0.1, + "learning_rate": 1.967989866678117e-05, + "loss": 1.1453, + "step": 1593 + }, + { + "epoch": 0.1, + "learning_rate": 1.967968397775822e-05, + "loss": 1.1208, + "step": 1594 + }, + { + "epoch": 0.1, + "learning_rate": 1.967946928873527e-05, + "loss": 1.184, + "step": 1595 + }, + { + "epoch": 0.1, + "learning_rate": 1.9679254599712317e-05, + "loss": 1.2862, + "step": 1596 + }, + { + "epoch": 0.1, + "learning_rate": 1.967903991068937e-05, + "loss": 1.1707, + "step": 1597 + }, + { + "epoch": 0.1, + "learning_rate": 1.967882522166642e-05, + "loss": 1.1738, + "step": 1598 + }, + { + "epoch": 0.1, + "learning_rate": 1.9678610532643468e-05, + "loss": 1.2202, + "step": 1599 + }, + { + "epoch": 0.1, + "learning_rate": 1.9678395843620517e-05, + "loss": 1.1674, + "step": 1600 + }, + { + "epoch": 0.1, + "learning_rate": 1.9678181154597566e-05, + "loss": 1.2001, + "step": 1601 + }, + { + "epoch": 0.1, + "learning_rate": 1.9677966465574615e-05, + "loss": 1.1854, + "step": 1602 + }, + { + "epoch": 0.1, + "learning_rate": 1.9677751776551667e-05, + "loss": 1.2741, + "step": 1603 + }, + { + "epoch": 0.1, + "learning_rate": 1.9677537087528716e-05, + "loss": 1.1904, + "step": 1604 + }, + { + "epoch": 0.1, + "learning_rate": 1.9677322398505768e-05, + "loss": 1.1117, + "step": 1605 + }, + { + "epoch": 0.1, + "learning_rate": 1.9677107709482817e-05, + "loss": 1.1868, + "step": 1606 + }, + { + "epoch": 0.1, + "learning_rate": 1.9676893020459866e-05, + "loss": 1.1395, + "step": 1607 + }, + { + "epoch": 0.1, + "learning_rate": 1.9676678331436915e-05, + "loss": 1.1712, + "step": 1608 + }, + { + "epoch": 0.1, + "learning_rate": 1.9676463642413964e-05, + "loss": 1.1315, + "step": 1609 + }, + { + "epoch": 0.1, + "learning_rate": 1.9676248953391016e-05, + "loss": 1.2108, + "step": 1610 + }, + { + "epoch": 0.1, + "learning_rate": 1.9676034264368065e-05, + "loss": 1.1589, + "step": 1611 + }, + { + "epoch": 0.1, + "learning_rate": 1.9675819575345114e-05, + "loss": 1.2443, + "step": 1612 + }, + { + "epoch": 0.1, + "learning_rate": 1.9675604886322163e-05, + "loss": 1.131, + "step": 1613 + }, + { + "epoch": 0.1, + "learning_rate": 1.9675390197299212e-05, + "loss": 1.1511, + "step": 1614 + }, + { + "epoch": 0.1, + "learning_rate": 1.9675175508276265e-05, + "loss": 1.2049, + "step": 1615 + }, + { + "epoch": 0.1, + "learning_rate": 1.9674960819253314e-05, + "loss": 1.1436, + "step": 1616 + }, + { + "epoch": 0.1, + "learning_rate": 1.9674746130230363e-05, + "loss": 1.1679, + "step": 1617 + }, + { + "epoch": 0.1, + "learning_rate": 1.967453144120741e-05, + "loss": 1.1625, + "step": 1618 + }, + { + "epoch": 0.1, + "learning_rate": 1.967431675218446e-05, + "loss": 1.1978, + "step": 1619 + }, + { + "epoch": 0.1, + "learning_rate": 1.9674102063161513e-05, + "loss": 1.1019, + "step": 1620 + }, + { + "epoch": 0.1, + "learning_rate": 1.9673887374138562e-05, + "loss": 1.1818, + "step": 1621 + }, + { + "epoch": 0.1, + "learning_rate": 1.967367268511561e-05, + "loss": 1.1171, + "step": 1622 + }, + { + "epoch": 0.1, + "learning_rate": 1.9673457996092663e-05, + "loss": 1.2807, + "step": 1623 + }, + { + "epoch": 0.1, + "learning_rate": 1.9673243307069712e-05, + "loss": 1.2119, + "step": 1624 + }, + { + "epoch": 0.1, + "learning_rate": 1.967302861804676e-05, + "loss": 1.2978, + "step": 1625 + }, + { + "epoch": 0.1, + "learning_rate": 1.967281392902381e-05, + "loss": 1.196, + "step": 1626 + }, + { + "epoch": 0.1, + "learning_rate": 1.967259924000086e-05, + "loss": 1.2288, + "step": 1627 + }, + { + "epoch": 0.1, + "learning_rate": 1.967238455097791e-05, + "loss": 1.0742, + "step": 1628 + }, + { + "epoch": 0.1, + "learning_rate": 1.967216986195496e-05, + "loss": 1.1089, + "step": 1629 + }, + { + "epoch": 0.1, + "learning_rate": 1.967195517293201e-05, + "loss": 1.1901, + "step": 1630 + }, + { + "epoch": 0.1, + "learning_rate": 1.9671740483909058e-05, + "loss": 1.2702, + "step": 1631 + }, + { + "epoch": 0.1, + "learning_rate": 1.9671525794886107e-05, + "loss": 1.2166, + "step": 1632 + }, + { + "epoch": 0.11, + "learning_rate": 1.9671311105863156e-05, + "loss": 1.2266, + "step": 1633 + }, + { + "epoch": 0.11, + "learning_rate": 1.967109641684021e-05, + "loss": 1.1652, + "step": 1634 + }, + { + "epoch": 0.11, + "learning_rate": 1.967088172781726e-05, + "loss": 1.1406, + "step": 1635 + }, + { + "epoch": 0.11, + "learning_rate": 1.967066703879431e-05, + "loss": 1.2018, + "step": 1636 + }, + { + "epoch": 0.11, + "learning_rate": 1.967045234977136e-05, + "loss": 1.2053, + "step": 1637 + }, + { + "epoch": 0.11, + "learning_rate": 1.9670237660748408e-05, + "loss": 1.1779, + "step": 1638 + }, + { + "epoch": 0.11, + "learning_rate": 1.9670022971725457e-05, + "loss": 1.2111, + "step": 1639 + }, + { + "epoch": 0.11, + "learning_rate": 1.9669808282702506e-05, + "loss": 1.2019, + "step": 1640 + }, + { + "epoch": 0.11, + "learning_rate": 1.9669593593679558e-05, + "loss": 1.2566, + "step": 1641 + }, + { + "epoch": 0.11, + "learning_rate": 1.9669378904656607e-05, + "loss": 1.1876, + "step": 1642 + }, + { + "epoch": 0.11, + "learning_rate": 1.9669164215633656e-05, + "loss": 1.1763, + "step": 1643 + }, + { + "epoch": 0.11, + "learning_rate": 1.9668949526610705e-05, + "loss": 1.1218, + "step": 1644 + }, + { + "epoch": 0.11, + "learning_rate": 1.9668734837587754e-05, + "loss": 1.0922, + "step": 1645 + }, + { + "epoch": 0.11, + "learning_rate": 1.9668520148564806e-05, + "loss": 1.217, + "step": 1646 + }, + { + "epoch": 0.11, + "learning_rate": 1.9668305459541855e-05, + "loss": 1.1492, + "step": 1647 + }, + { + "epoch": 0.11, + "learning_rate": 1.9668090770518904e-05, + "loss": 1.1844, + "step": 1648 + }, + { + "epoch": 0.11, + "learning_rate": 1.9667876081495957e-05, + "loss": 1.0664, + "step": 1649 + }, + { + "epoch": 0.11, + "learning_rate": 1.9667661392473006e-05, + "loss": 1.1495, + "step": 1650 + }, + { + "epoch": 0.11, + "learning_rate": 1.9667446703450054e-05, + "loss": 1.2036, + "step": 1651 + }, + { + "epoch": 0.11, + "learning_rate": 1.9667232014427103e-05, + "loss": 1.1069, + "step": 1652 + }, + { + "epoch": 0.11, + "learning_rate": 1.9667017325404152e-05, + "loss": 1.1677, + "step": 1653 + }, + { + "epoch": 0.11, + "learning_rate": 1.9666802636381205e-05, + "loss": 1.2228, + "step": 1654 + }, + { + "epoch": 0.11, + "learning_rate": 1.9666587947358254e-05, + "loss": 1.2128, + "step": 1655 + }, + { + "epoch": 0.11, + "learning_rate": 1.9666373258335303e-05, + "loss": 1.1959, + "step": 1656 + }, + { + "epoch": 0.11, + "learning_rate": 1.966615856931235e-05, + "loss": 1.2219, + "step": 1657 + }, + { + "epoch": 0.11, + "learning_rate": 1.96659438802894e-05, + "loss": 1.2725, + "step": 1658 + }, + { + "epoch": 0.11, + "learning_rate": 1.9665729191266453e-05, + "loss": 1.2227, + "step": 1659 + }, + { + "epoch": 0.11, + "learning_rate": 1.9665514502243502e-05, + "loss": 1.2795, + "step": 1660 + }, + { + "epoch": 0.11, + "learning_rate": 1.966529981322055e-05, + "loss": 1.2238, + "step": 1661 + }, + { + "epoch": 0.11, + "learning_rate": 1.96650851241976e-05, + "loss": 1.0548, + "step": 1662 + }, + { + "epoch": 0.11, + "learning_rate": 1.9664870435174652e-05, + "loss": 1.195, + "step": 1663 + }, + { + "epoch": 0.11, + "learning_rate": 1.96646557461517e-05, + "loss": 1.2613, + "step": 1664 + }, + { + "epoch": 0.11, + "learning_rate": 1.966444105712875e-05, + "loss": 1.1009, + "step": 1665 + }, + { + "epoch": 0.11, + "learning_rate": 1.9664226368105803e-05, + "loss": 1.2148, + "step": 1666 + }, + { + "epoch": 0.11, + "learning_rate": 1.966401167908285e-05, + "loss": 1.2395, + "step": 1667 + }, + { + "epoch": 0.11, + "learning_rate": 1.96637969900599e-05, + "loss": 1.148, + "step": 1668 + }, + { + "epoch": 0.11, + "learning_rate": 1.966358230103695e-05, + "loss": 1.2105, + "step": 1669 + }, + { + "epoch": 0.11, + "learning_rate": 1.9663367612014e-05, + "loss": 1.1922, + "step": 1670 + }, + { + "epoch": 0.11, + "learning_rate": 1.9663152922991047e-05, + "loss": 1.2241, + "step": 1671 + }, + { + "epoch": 0.11, + "learning_rate": 1.96629382339681e-05, + "loss": 1.0475, + "step": 1672 + }, + { + "epoch": 0.11, + "learning_rate": 1.966272354494515e-05, + "loss": 1.1562, + "step": 1673 + }, + { + "epoch": 0.11, + "learning_rate": 1.9662508855922198e-05, + "loss": 1.2098, + "step": 1674 + }, + { + "epoch": 0.11, + "learning_rate": 1.9662294166899247e-05, + "loss": 1.1999, + "step": 1675 + }, + { + "epoch": 0.11, + "learning_rate": 1.9662079477876296e-05, + "loss": 1.0968, + "step": 1676 + }, + { + "epoch": 0.11, + "learning_rate": 1.9661864788853348e-05, + "loss": 1.0867, + "step": 1677 + }, + { + "epoch": 0.11, + "learning_rate": 1.9661650099830397e-05, + "loss": 1.1869, + "step": 1678 + }, + { + "epoch": 0.11, + "learning_rate": 1.966143541080745e-05, + "loss": 1.1652, + "step": 1679 + }, + { + "epoch": 0.11, + "learning_rate": 1.9661220721784498e-05, + "loss": 1.21, + "step": 1680 + }, + { + "epoch": 0.11, + "learning_rate": 1.9661006032761547e-05, + "loss": 1.1687, + "step": 1681 + }, + { + "epoch": 0.11, + "learning_rate": 1.9660791343738596e-05, + "loss": 1.113, + "step": 1682 + }, + { + "epoch": 0.11, + "learning_rate": 1.9660576654715645e-05, + "loss": 1.1984, + "step": 1683 + }, + { + "epoch": 0.11, + "learning_rate": 1.9660361965692697e-05, + "loss": 1.1216, + "step": 1684 + }, + { + "epoch": 0.11, + "learning_rate": 1.9660147276669746e-05, + "loss": 1.1985, + "step": 1685 + }, + { + "epoch": 0.11, + "learning_rate": 1.9659932587646795e-05, + "loss": 1.0894, + "step": 1686 + }, + { + "epoch": 0.11, + "learning_rate": 1.9659717898623844e-05, + "loss": 1.2357, + "step": 1687 + }, + { + "epoch": 0.11, + "learning_rate": 1.9659503209600893e-05, + "loss": 1.2029, + "step": 1688 + }, + { + "epoch": 0.11, + "learning_rate": 1.9659288520577942e-05, + "loss": 1.1466, + "step": 1689 + }, + { + "epoch": 0.11, + "learning_rate": 1.9659073831554995e-05, + "loss": 1.1115, + "step": 1690 + }, + { + "epoch": 0.11, + "learning_rate": 1.9658859142532044e-05, + "loss": 1.0983, + "step": 1691 + }, + { + "epoch": 0.11, + "learning_rate": 1.9658644453509096e-05, + "loss": 1.1543, + "step": 1692 + }, + { + "epoch": 0.11, + "learning_rate": 1.9658429764486145e-05, + "loss": 1.2181, + "step": 1693 + }, + { + "epoch": 0.11, + "learning_rate": 1.9658215075463194e-05, + "loss": 1.1163, + "step": 1694 + }, + { + "epoch": 0.11, + "learning_rate": 1.9658000386440243e-05, + "loss": 1.1814, + "step": 1695 + }, + { + "epoch": 0.11, + "learning_rate": 1.9657785697417292e-05, + "loss": 1.2953, + "step": 1696 + }, + { + "epoch": 0.11, + "learning_rate": 1.9657571008394344e-05, + "loss": 1.2185, + "step": 1697 + }, + { + "epoch": 0.11, + "learning_rate": 1.9657356319371393e-05, + "loss": 1.2142, + "step": 1698 + }, + { + "epoch": 0.11, + "learning_rate": 1.9657141630348442e-05, + "loss": 1.2498, + "step": 1699 + }, + { + "epoch": 0.11, + "learning_rate": 1.965692694132549e-05, + "loss": 1.1734, + "step": 1700 + }, + { + "epoch": 0.11, + "learning_rate": 1.965671225230254e-05, + "loss": 1.225, + "step": 1701 + }, + { + "epoch": 0.11, + "learning_rate": 1.9656497563279592e-05, + "loss": 1.1388, + "step": 1702 + }, + { + "epoch": 0.11, + "learning_rate": 1.965628287425664e-05, + "loss": 1.1795, + "step": 1703 + }, + { + "epoch": 0.11, + "learning_rate": 1.965606818523369e-05, + "loss": 1.0904, + "step": 1704 + }, + { + "epoch": 0.11, + "learning_rate": 1.965585349621074e-05, + "loss": 1.2862, + "step": 1705 + }, + { + "epoch": 0.11, + "learning_rate": 1.9655638807187788e-05, + "loss": 1.1476, + "step": 1706 + }, + { + "epoch": 0.11, + "learning_rate": 1.965542411816484e-05, + "loss": 1.1899, + "step": 1707 + }, + { + "epoch": 0.11, + "learning_rate": 1.965520942914189e-05, + "loss": 1.1419, + "step": 1708 + }, + { + "epoch": 0.11, + "learning_rate": 1.965499474011894e-05, + "loss": 1.1489, + "step": 1709 + }, + { + "epoch": 0.11, + "learning_rate": 1.965478005109599e-05, + "loss": 1.2649, + "step": 1710 + }, + { + "epoch": 0.11, + "learning_rate": 1.965456536207304e-05, + "loss": 1.2039, + "step": 1711 + }, + { + "epoch": 0.11, + "learning_rate": 1.965435067305009e-05, + "loss": 1.2004, + "step": 1712 + }, + { + "epoch": 0.11, + "learning_rate": 1.9654135984027138e-05, + "loss": 1.1369, + "step": 1713 + }, + { + "epoch": 0.11, + "learning_rate": 1.9653921295004187e-05, + "loss": 1.2394, + "step": 1714 + }, + { + "epoch": 0.11, + "learning_rate": 1.965370660598124e-05, + "loss": 1.1681, + "step": 1715 + }, + { + "epoch": 0.11, + "learning_rate": 1.9653491916958288e-05, + "loss": 1.2065, + "step": 1716 + }, + { + "epoch": 0.11, + "learning_rate": 1.9653277227935337e-05, + "loss": 1.1011, + "step": 1717 + }, + { + "epoch": 0.11, + "learning_rate": 1.9653062538912386e-05, + "loss": 1.1585, + "step": 1718 + }, + { + "epoch": 0.11, + "learning_rate": 1.9652847849889435e-05, + "loss": 1.1695, + "step": 1719 + }, + { + "epoch": 0.11, + "learning_rate": 1.9652633160866484e-05, + "loss": 1.198, + "step": 1720 + }, + { + "epoch": 0.11, + "learning_rate": 1.9652418471843536e-05, + "loss": 1.2704, + "step": 1721 + }, + { + "epoch": 0.11, + "learning_rate": 1.965220378282059e-05, + "loss": 1.1236, + "step": 1722 + }, + { + "epoch": 0.11, + "learning_rate": 1.9651989093797638e-05, + "loss": 1.1529, + "step": 1723 + }, + { + "epoch": 0.11, + "learning_rate": 1.9651774404774687e-05, + "loss": 1.2662, + "step": 1724 + }, + { + "epoch": 0.11, + "learning_rate": 1.9651559715751735e-05, + "loss": 1.0523, + "step": 1725 + }, + { + "epoch": 0.11, + "learning_rate": 1.9651345026728784e-05, + "loss": 1.1996, + "step": 1726 + }, + { + "epoch": 0.11, + "learning_rate": 1.9651130337705833e-05, + "loss": 1.2042, + "step": 1727 + }, + { + "epoch": 0.11, + "learning_rate": 1.9650915648682886e-05, + "loss": 1.1619, + "step": 1728 + }, + { + "epoch": 0.11, + "learning_rate": 1.9650700959659935e-05, + "loss": 1.2114, + "step": 1729 + }, + { + "epoch": 0.11, + "learning_rate": 1.9650486270636984e-05, + "loss": 1.2633, + "step": 1730 + }, + { + "epoch": 0.11, + "learning_rate": 1.9650271581614033e-05, + "loss": 1.1646, + "step": 1731 + }, + { + "epoch": 0.11, + "learning_rate": 1.965005689259108e-05, + "loss": 1.1329, + "step": 1732 + }, + { + "epoch": 0.11, + "learning_rate": 1.9649842203568134e-05, + "loss": 1.1168, + "step": 1733 + }, + { + "epoch": 0.11, + "learning_rate": 1.9649627514545183e-05, + "loss": 1.1985, + "step": 1734 + }, + { + "epoch": 0.11, + "learning_rate": 1.9649412825522232e-05, + "loss": 1.1756, + "step": 1735 + }, + { + "epoch": 0.11, + "learning_rate": 1.9649198136499284e-05, + "loss": 1.2585, + "step": 1736 + }, + { + "epoch": 0.11, + "learning_rate": 1.9648983447476333e-05, + "loss": 1.2212, + "step": 1737 + }, + { + "epoch": 0.11, + "learning_rate": 1.9648768758453382e-05, + "loss": 1.2159, + "step": 1738 + }, + { + "epoch": 0.11, + "learning_rate": 1.964855406943043e-05, + "loss": 1.1679, + "step": 1739 + }, + { + "epoch": 0.11, + "learning_rate": 1.964833938040748e-05, + "loss": 1.047, + "step": 1740 + }, + { + "epoch": 0.11, + "learning_rate": 1.9648124691384532e-05, + "loss": 1.1874, + "step": 1741 + }, + { + "epoch": 0.11, + "learning_rate": 1.964791000236158e-05, + "loss": 1.1553, + "step": 1742 + }, + { + "epoch": 0.11, + "learning_rate": 1.964769531333863e-05, + "loss": 1.182, + "step": 1743 + }, + { + "epoch": 0.11, + "learning_rate": 1.964748062431568e-05, + "loss": 1.2165, + "step": 1744 + }, + { + "epoch": 0.11, + "learning_rate": 1.964726593529273e-05, + "loss": 1.1541, + "step": 1745 + }, + { + "epoch": 0.11, + "learning_rate": 1.964705124626978e-05, + "loss": 1.1321, + "step": 1746 + }, + { + "epoch": 0.11, + "learning_rate": 1.964683655724683e-05, + "loss": 1.2915, + "step": 1747 + }, + { + "epoch": 0.11, + "learning_rate": 1.964662186822388e-05, + "loss": 1.1721, + "step": 1748 + }, + { + "epoch": 0.11, + "learning_rate": 1.9646407179200928e-05, + "loss": 1.2336, + "step": 1749 + }, + { + "epoch": 0.11, + "learning_rate": 1.964619249017798e-05, + "loss": 1.1998, + "step": 1750 + }, + { + "epoch": 0.11, + "learning_rate": 1.964597780115503e-05, + "loss": 1.1792, + "step": 1751 + }, + { + "epoch": 0.11, + "learning_rate": 1.9645763112132078e-05, + "loss": 1.2025, + "step": 1752 + }, + { + "epoch": 0.11, + "learning_rate": 1.964554842310913e-05, + "loss": 1.2887, + "step": 1753 + }, + { + "epoch": 0.11, + "learning_rate": 1.964533373408618e-05, + "loss": 1.195, + "step": 1754 + }, + { + "epoch": 0.11, + "learning_rate": 1.9645119045063228e-05, + "loss": 1.1869, + "step": 1755 + }, + { + "epoch": 0.11, + "learning_rate": 1.9644904356040277e-05, + "loss": 1.122, + "step": 1756 + }, + { + "epoch": 0.11, + "learning_rate": 1.9644689667017326e-05, + "loss": 1.0944, + "step": 1757 + }, + { + "epoch": 0.11, + "learning_rate": 1.9644474977994375e-05, + "loss": 1.1783, + "step": 1758 + }, + { + "epoch": 0.11, + "learning_rate": 1.9644260288971427e-05, + "loss": 1.1632, + "step": 1759 + }, + { + "epoch": 0.11, + "learning_rate": 1.9644045599948476e-05, + "loss": 1.2229, + "step": 1760 + }, + { + "epoch": 0.11, + "learning_rate": 1.9643830910925525e-05, + "loss": 1.085, + "step": 1761 + }, + { + "epoch": 0.11, + "learning_rate": 1.9643616221902574e-05, + "loss": 1.1432, + "step": 1762 + }, + { + "epoch": 0.11, + "learning_rate": 1.9643401532879623e-05, + "loss": 1.2294, + "step": 1763 + }, + { + "epoch": 0.11, + "learning_rate": 1.9643186843856676e-05, + "loss": 1.1908, + "step": 1764 + }, + { + "epoch": 0.11, + "learning_rate": 1.9642972154833725e-05, + "loss": 1.2722, + "step": 1765 + }, + { + "epoch": 0.11, + "learning_rate": 1.9642757465810777e-05, + "loss": 1.2348, + "step": 1766 + }, + { + "epoch": 0.11, + "learning_rate": 1.9642542776787826e-05, + "loss": 1.2167, + "step": 1767 + }, + { + "epoch": 0.11, + "learning_rate": 1.9642328087764875e-05, + "loss": 1.2298, + "step": 1768 + }, + { + "epoch": 0.11, + "learning_rate": 1.9642113398741924e-05, + "loss": 1.208, + "step": 1769 + }, + { + "epoch": 0.11, + "learning_rate": 1.9641898709718973e-05, + "loss": 1.1575, + "step": 1770 + }, + { + "epoch": 0.11, + "learning_rate": 1.9641684020696025e-05, + "loss": 1.2014, + "step": 1771 + }, + { + "epoch": 0.11, + "learning_rate": 1.9641469331673074e-05, + "loss": 1.196, + "step": 1772 + }, + { + "epoch": 0.11, + "learning_rate": 1.9641254642650123e-05, + "loss": 1.1721, + "step": 1773 + }, + { + "epoch": 0.11, + "learning_rate": 1.9641039953627172e-05, + "loss": 1.2271, + "step": 1774 + }, + { + "epoch": 0.11, + "learning_rate": 1.964082526460422e-05, + "loss": 1.1325, + "step": 1775 + }, + { + "epoch": 0.11, + "learning_rate": 1.964061057558127e-05, + "loss": 1.1325, + "step": 1776 + }, + { + "epoch": 0.11, + "learning_rate": 1.9640395886558322e-05, + "loss": 1.1934, + "step": 1777 + }, + { + "epoch": 0.11, + "learning_rate": 1.964018119753537e-05, + "loss": 1.1, + "step": 1778 + }, + { + "epoch": 0.11, + "learning_rate": 1.963996650851242e-05, + "loss": 1.1754, + "step": 1779 + }, + { + "epoch": 0.11, + "learning_rate": 1.9639751819489473e-05, + "loss": 1.1453, + "step": 1780 + }, + { + "epoch": 0.11, + "learning_rate": 1.963953713046652e-05, + "loss": 1.2008, + "step": 1781 + }, + { + "epoch": 0.11, + "learning_rate": 1.963932244144357e-05, + "loss": 1.194, + "step": 1782 + }, + { + "epoch": 0.11, + "learning_rate": 1.963910775242062e-05, + "loss": 1.2341, + "step": 1783 + }, + { + "epoch": 0.11, + "learning_rate": 1.9638893063397672e-05, + "loss": 1.2057, + "step": 1784 + }, + { + "epoch": 0.11, + "learning_rate": 1.963867837437472e-05, + "loss": 1.2504, + "step": 1785 + }, + { + "epoch": 0.11, + "learning_rate": 1.963846368535177e-05, + "loss": 1.1617, + "step": 1786 + }, + { + "epoch": 0.11, + "learning_rate": 1.963824899632882e-05, + "loss": 1.1896, + "step": 1787 + }, + { + "epoch": 0.12, + "learning_rate": 1.9638034307305868e-05, + "loss": 1.2224, + "step": 1788 + }, + { + "epoch": 0.12, + "learning_rate": 1.963781961828292e-05, + "loss": 1.1911, + "step": 1789 + }, + { + "epoch": 0.12, + "learning_rate": 1.963760492925997e-05, + "loss": 1.2888, + "step": 1790 + }, + { + "epoch": 0.12, + "learning_rate": 1.9637390240237018e-05, + "loss": 1.2003, + "step": 1791 + }, + { + "epoch": 0.12, + "learning_rate": 1.9637175551214067e-05, + "loss": 1.1602, + "step": 1792 + }, + { + "epoch": 0.12, + "learning_rate": 1.9636960862191116e-05, + "loss": 1.1483, + "step": 1793 + }, + { + "epoch": 0.12, + "learning_rate": 1.9636746173168168e-05, + "loss": 1.1008, + "step": 1794 + }, + { + "epoch": 0.12, + "learning_rate": 1.9636531484145217e-05, + "loss": 1.1428, + "step": 1795 + }, + { + "epoch": 0.12, + "learning_rate": 1.9636316795122266e-05, + "loss": 1.1677, + "step": 1796 + }, + { + "epoch": 0.12, + "learning_rate": 1.963610210609932e-05, + "loss": 1.1409, + "step": 1797 + }, + { + "epoch": 0.12, + "learning_rate": 1.9635887417076368e-05, + "loss": 1.1934, + "step": 1798 + }, + { + "epoch": 0.12, + "learning_rate": 1.9635672728053417e-05, + "loss": 1.1714, + "step": 1799 + }, + { + "epoch": 0.12, + "learning_rate": 1.9635458039030465e-05, + "loss": 1.2808, + "step": 1800 + }, + { + "epoch": 0.12, + "learning_rate": 1.9635243350007514e-05, + "loss": 1.1449, + "step": 1801 + }, + { + "epoch": 0.12, + "learning_rate": 1.9635028660984567e-05, + "loss": 1.1724, + "step": 1802 + }, + { + "epoch": 0.12, + "learning_rate": 1.9634813971961616e-05, + "loss": 1.2578, + "step": 1803 + }, + { + "epoch": 0.12, + "learning_rate": 1.9634599282938665e-05, + "loss": 1.1862, + "step": 1804 + }, + { + "epoch": 0.12, + "learning_rate": 1.9634384593915714e-05, + "loss": 1.1225, + "step": 1805 + }, + { + "epoch": 0.12, + "learning_rate": 1.9634169904892763e-05, + "loss": 1.0883, + "step": 1806 + }, + { + "epoch": 0.12, + "learning_rate": 1.963395521586981e-05, + "loss": 1.2513, + "step": 1807 + }, + { + "epoch": 0.12, + "learning_rate": 1.9633740526846864e-05, + "loss": 1.1529, + "step": 1808 + }, + { + "epoch": 0.12, + "learning_rate": 1.9633525837823916e-05, + "loss": 1.2622, + "step": 1809 + }, + { + "epoch": 0.12, + "learning_rate": 1.9633311148800965e-05, + "loss": 1.1227, + "step": 1810 + }, + { + "epoch": 0.12, + "learning_rate": 1.9633096459778014e-05, + "loss": 1.2549, + "step": 1811 + }, + { + "epoch": 0.12, + "learning_rate": 1.9632881770755063e-05, + "loss": 1.123, + "step": 1812 + }, + { + "epoch": 0.12, + "learning_rate": 1.9632667081732112e-05, + "loss": 1.2254, + "step": 1813 + }, + { + "epoch": 0.12, + "learning_rate": 1.963245239270916e-05, + "loss": 1.1721, + "step": 1814 + }, + { + "epoch": 0.12, + "learning_rate": 1.9632237703686213e-05, + "loss": 1.1505, + "step": 1815 + }, + { + "epoch": 0.12, + "learning_rate": 1.9632023014663262e-05, + "loss": 1.2702, + "step": 1816 + }, + { + "epoch": 0.12, + "learning_rate": 1.963180832564031e-05, + "loss": 1.2504, + "step": 1817 + }, + { + "epoch": 0.12, + "learning_rate": 1.963159363661736e-05, + "loss": 1.2421, + "step": 1818 + }, + { + "epoch": 0.12, + "learning_rate": 1.963137894759441e-05, + "loss": 1.2493, + "step": 1819 + }, + { + "epoch": 0.12, + "learning_rate": 1.9631164258571462e-05, + "loss": 1.2666, + "step": 1820 + }, + { + "epoch": 0.12, + "learning_rate": 1.963094956954851e-05, + "loss": 1.1381, + "step": 1821 + }, + { + "epoch": 0.12, + "learning_rate": 1.963073488052556e-05, + "loss": 1.2516, + "step": 1822 + }, + { + "epoch": 0.12, + "learning_rate": 1.9630520191502612e-05, + "loss": 1.1438, + "step": 1823 + }, + { + "epoch": 0.12, + "learning_rate": 1.963030550247966e-05, + "loss": 1.122, + "step": 1824 + }, + { + "epoch": 0.12, + "learning_rate": 1.963009081345671e-05, + "loss": 1.2435, + "step": 1825 + }, + { + "epoch": 0.12, + "learning_rate": 1.962987612443376e-05, + "loss": 1.2308, + "step": 1826 + }, + { + "epoch": 0.12, + "learning_rate": 1.9629661435410808e-05, + "loss": 1.1807, + "step": 1827 + }, + { + "epoch": 0.12, + "learning_rate": 1.962944674638786e-05, + "loss": 1.2083, + "step": 1828 + }, + { + "epoch": 0.12, + "learning_rate": 1.962923205736491e-05, + "loss": 1.1431, + "step": 1829 + }, + { + "epoch": 0.12, + "learning_rate": 1.9629017368341958e-05, + "loss": 1.2331, + "step": 1830 + }, + { + "epoch": 0.12, + "learning_rate": 1.9628802679319007e-05, + "loss": 1.2137, + "step": 1831 + }, + { + "epoch": 0.12, + "learning_rate": 1.9628587990296056e-05, + "loss": 1.2464, + "step": 1832 + }, + { + "epoch": 0.12, + "learning_rate": 1.962837330127311e-05, + "loss": 1.0973, + "step": 1833 + }, + { + "epoch": 0.12, + "learning_rate": 1.9628158612250157e-05, + "loss": 1.269, + "step": 1834 + }, + { + "epoch": 0.12, + "learning_rate": 1.9627943923227206e-05, + "loss": 1.1531, + "step": 1835 + }, + { + "epoch": 0.12, + "learning_rate": 1.9627729234204255e-05, + "loss": 1.2285, + "step": 1836 + }, + { + "epoch": 0.12, + "learning_rate": 1.9627514545181304e-05, + "loss": 1.1744, + "step": 1837 + }, + { + "epoch": 0.12, + "learning_rate": 1.9627299856158357e-05, + "loss": 1.307, + "step": 1838 + }, + { + "epoch": 0.12, + "learning_rate": 1.9627085167135406e-05, + "loss": 1.2217, + "step": 1839 + }, + { + "epoch": 0.12, + "learning_rate": 1.9626870478112458e-05, + "loss": 1.178, + "step": 1840 + }, + { + "epoch": 0.12, + "learning_rate": 1.9626655789089507e-05, + "loss": 1.1488, + "step": 1841 + }, + { + "epoch": 0.12, + "learning_rate": 1.9626441100066556e-05, + "loss": 1.1349, + "step": 1842 + }, + { + "epoch": 0.12, + "learning_rate": 1.9626226411043605e-05, + "loss": 1.2107, + "step": 1843 + }, + { + "epoch": 0.12, + "learning_rate": 1.9626011722020654e-05, + "loss": 1.1959, + "step": 1844 + }, + { + "epoch": 0.12, + "learning_rate": 1.9625797032997703e-05, + "loss": 1.1316, + "step": 1845 + }, + { + "epoch": 0.12, + "learning_rate": 1.9625582343974755e-05, + "loss": 1.2404, + "step": 1846 + }, + { + "epoch": 0.12, + "learning_rate": 1.9625367654951804e-05, + "loss": 1.2606, + "step": 1847 + }, + { + "epoch": 0.12, + "learning_rate": 1.9625152965928853e-05, + "loss": 1.1225, + "step": 1848 + }, + { + "epoch": 0.12, + "learning_rate": 1.9624938276905902e-05, + "loss": 1.1323, + "step": 1849 + }, + { + "epoch": 0.12, + "learning_rate": 1.962472358788295e-05, + "loss": 1.234, + "step": 1850 + }, + { + "epoch": 0.12, + "learning_rate": 1.9624508898860003e-05, + "loss": 1.1395, + "step": 1851 + }, + { + "epoch": 0.12, + "learning_rate": 1.9624294209837052e-05, + "loss": 1.1813, + "step": 1852 + }, + { + "epoch": 0.12, + "learning_rate": 1.9624079520814105e-05, + "loss": 1.1952, + "step": 1853 + }, + { + "epoch": 0.12, + "learning_rate": 1.9623864831791154e-05, + "loss": 1.2076, + "step": 1854 + }, + { + "epoch": 0.12, + "learning_rate": 1.9623650142768203e-05, + "loss": 1.154, + "step": 1855 + }, + { + "epoch": 0.12, + "learning_rate": 1.962343545374525e-05, + "loss": 1.2675, + "step": 1856 + }, + { + "epoch": 0.12, + "learning_rate": 1.96232207647223e-05, + "loss": 1.2219, + "step": 1857 + }, + { + "epoch": 0.12, + "learning_rate": 1.9623006075699353e-05, + "loss": 1.1708, + "step": 1858 + }, + { + "epoch": 0.12, + "learning_rate": 1.9622791386676402e-05, + "loss": 1.1963, + "step": 1859 + }, + { + "epoch": 0.12, + "learning_rate": 1.962257669765345e-05, + "loss": 1.1587, + "step": 1860 + }, + { + "epoch": 0.12, + "learning_rate": 1.96223620086305e-05, + "loss": 1.081, + "step": 1861 + }, + { + "epoch": 0.12, + "learning_rate": 1.962214731960755e-05, + "loss": 1.2054, + "step": 1862 + }, + { + "epoch": 0.12, + "learning_rate": 1.9621932630584598e-05, + "loss": 1.1764, + "step": 1863 + }, + { + "epoch": 0.12, + "learning_rate": 1.962171794156165e-05, + "loss": 1.0662, + "step": 1864 + }, + { + "epoch": 0.12, + "learning_rate": 1.96215032525387e-05, + "loss": 1.1517, + "step": 1865 + }, + { + "epoch": 0.12, + "learning_rate": 1.9621288563515748e-05, + "loss": 1.2106, + "step": 1866 + }, + { + "epoch": 0.12, + "learning_rate": 1.96210738744928e-05, + "loss": 1.1861, + "step": 1867 + }, + { + "epoch": 0.12, + "learning_rate": 1.962085918546985e-05, + "loss": 1.2303, + "step": 1868 + }, + { + "epoch": 0.12, + "learning_rate": 1.9620644496446898e-05, + "loss": 1.1689, + "step": 1869 + }, + { + "epoch": 0.12, + "learning_rate": 1.9620429807423947e-05, + "loss": 1.1922, + "step": 1870 + }, + { + "epoch": 0.12, + "learning_rate": 1.9620215118401e-05, + "loss": 1.202, + "step": 1871 + }, + { + "epoch": 0.12, + "learning_rate": 1.962000042937805e-05, + "loss": 1.1906, + "step": 1872 + }, + { + "epoch": 0.12, + "learning_rate": 1.9619785740355098e-05, + "loss": 1.2197, + "step": 1873 + }, + { + "epoch": 0.12, + "learning_rate": 1.9619571051332146e-05, + "loss": 1.2864, + "step": 1874 + }, + { + "epoch": 0.12, + "learning_rate": 1.9619356362309195e-05, + "loss": 1.1114, + "step": 1875 + }, + { + "epoch": 0.12, + "learning_rate": 1.9619141673286244e-05, + "loss": 1.1602, + "step": 1876 + }, + { + "epoch": 0.12, + "learning_rate": 1.9618926984263297e-05, + "loss": 1.1606, + "step": 1877 + }, + { + "epoch": 0.12, + "learning_rate": 1.9618712295240346e-05, + "loss": 1.1949, + "step": 1878 + }, + { + "epoch": 0.12, + "learning_rate": 1.9618497606217395e-05, + "loss": 1.1816, + "step": 1879 + }, + { + "epoch": 0.12, + "learning_rate": 1.9618282917194444e-05, + "loss": 1.1614, + "step": 1880 + }, + { + "epoch": 0.12, + "learning_rate": 1.9618068228171496e-05, + "loss": 1.0592, + "step": 1881 + }, + { + "epoch": 0.12, + "learning_rate": 1.9617853539148545e-05, + "loss": 1.2413, + "step": 1882 + }, + { + "epoch": 0.12, + "learning_rate": 1.9617638850125594e-05, + "loss": 1.2936, + "step": 1883 + }, + { + "epoch": 0.12, + "learning_rate": 1.9617424161102646e-05, + "loss": 1.278, + "step": 1884 + }, + { + "epoch": 0.12, + "learning_rate": 1.9617209472079695e-05, + "loss": 1.1009, + "step": 1885 + }, + { + "epoch": 0.12, + "learning_rate": 1.9616994783056744e-05, + "loss": 1.0836, + "step": 1886 + }, + { + "epoch": 0.12, + "learning_rate": 1.9616780094033793e-05, + "loss": 1.1075, + "step": 1887 + }, + { + "epoch": 0.12, + "learning_rate": 1.9616565405010842e-05, + "loss": 1.1833, + "step": 1888 + }, + { + "epoch": 0.12, + "learning_rate": 1.9616350715987894e-05, + "loss": 1.1891, + "step": 1889 + }, + { + "epoch": 0.12, + "learning_rate": 1.9616136026964943e-05, + "loss": 1.1593, + "step": 1890 + }, + { + "epoch": 0.12, + "learning_rate": 1.9615921337941992e-05, + "loss": 1.1665, + "step": 1891 + }, + { + "epoch": 0.12, + "learning_rate": 1.961570664891904e-05, + "loss": 1.1503, + "step": 1892 + }, + { + "epoch": 0.12, + "learning_rate": 1.961549195989609e-05, + "loss": 1.1906, + "step": 1893 + }, + { + "epoch": 0.12, + "learning_rate": 1.961527727087314e-05, + "loss": 1.1683, + "step": 1894 + }, + { + "epoch": 0.12, + "learning_rate": 1.961506258185019e-05, + "loss": 1.2403, + "step": 1895 + }, + { + "epoch": 0.12, + "learning_rate": 1.961484789282724e-05, + "loss": 1.1722, + "step": 1896 + }, + { + "epoch": 0.12, + "learning_rate": 1.9614633203804293e-05, + "loss": 1.2419, + "step": 1897 + }, + { + "epoch": 0.12, + "learning_rate": 1.9614418514781342e-05, + "loss": 1.25, + "step": 1898 + }, + { + "epoch": 0.12, + "learning_rate": 1.961420382575839e-05, + "loss": 1.1623, + "step": 1899 + }, + { + "epoch": 0.12, + "learning_rate": 1.961398913673544e-05, + "loss": 1.2329, + "step": 1900 + }, + { + "epoch": 0.12, + "learning_rate": 1.961377444771249e-05, + "loss": 1.3362, + "step": 1901 + }, + { + "epoch": 0.12, + "learning_rate": 1.961355975868954e-05, + "loss": 1.0995, + "step": 1902 + }, + { + "epoch": 0.12, + "learning_rate": 1.961334506966659e-05, + "loss": 1.1186, + "step": 1903 + }, + { + "epoch": 0.12, + "learning_rate": 1.961313038064364e-05, + "loss": 1.1343, + "step": 1904 + }, + { + "epoch": 0.12, + "learning_rate": 1.9612915691620688e-05, + "loss": 1.0885, + "step": 1905 + }, + { + "epoch": 0.12, + "learning_rate": 1.9612701002597737e-05, + "loss": 1.1248, + "step": 1906 + }, + { + "epoch": 0.12, + "learning_rate": 1.961248631357479e-05, + "loss": 1.1709, + "step": 1907 + }, + { + "epoch": 0.12, + "learning_rate": 1.961227162455184e-05, + "loss": 1.1324, + "step": 1908 + }, + { + "epoch": 0.12, + "learning_rate": 1.9612056935528887e-05, + "loss": 1.1868, + "step": 1909 + }, + { + "epoch": 0.12, + "learning_rate": 1.961184224650594e-05, + "loss": 1.1556, + "step": 1910 + }, + { + "epoch": 0.12, + "learning_rate": 1.961162755748299e-05, + "loss": 1.2281, + "step": 1911 + }, + { + "epoch": 0.12, + "learning_rate": 1.9611412868460038e-05, + "loss": 1.1741, + "step": 1912 + }, + { + "epoch": 0.12, + "learning_rate": 1.9611198179437087e-05, + "loss": 1.2089, + "step": 1913 + }, + { + "epoch": 0.12, + "learning_rate": 1.9610983490414136e-05, + "loss": 1.2776, + "step": 1914 + }, + { + "epoch": 0.12, + "learning_rate": 1.9610768801391188e-05, + "loss": 1.1471, + "step": 1915 + }, + { + "epoch": 0.12, + "learning_rate": 1.9610554112368237e-05, + "loss": 1.2479, + "step": 1916 + }, + { + "epoch": 0.12, + "learning_rate": 1.9610339423345286e-05, + "loss": 1.1378, + "step": 1917 + }, + { + "epoch": 0.12, + "learning_rate": 1.9610124734322335e-05, + "loss": 1.1829, + "step": 1918 + }, + { + "epoch": 0.12, + "learning_rate": 1.9609910045299384e-05, + "loss": 1.2632, + "step": 1919 + }, + { + "epoch": 0.12, + "learning_rate": 1.9609695356276436e-05, + "loss": 1.0215, + "step": 1920 + }, + { + "epoch": 0.12, + "learning_rate": 1.9609480667253485e-05, + "loss": 1.1106, + "step": 1921 + }, + { + "epoch": 0.12, + "learning_rate": 1.9609265978230534e-05, + "loss": 1.1263, + "step": 1922 + }, + { + "epoch": 0.12, + "learning_rate": 1.9609051289207583e-05, + "loss": 1.1635, + "step": 1923 + }, + { + "epoch": 0.12, + "learning_rate": 1.9608836600184632e-05, + "loss": 1.1111, + "step": 1924 + }, + { + "epoch": 0.12, + "learning_rate": 1.9608621911161684e-05, + "loss": 1.1631, + "step": 1925 + }, + { + "epoch": 0.12, + "learning_rate": 1.9608407222138733e-05, + "loss": 1.2447, + "step": 1926 + }, + { + "epoch": 0.12, + "learning_rate": 1.9608192533115786e-05, + "loss": 1.2382, + "step": 1927 + }, + { + "epoch": 0.12, + "learning_rate": 1.9607977844092835e-05, + "loss": 1.2433, + "step": 1928 + }, + { + "epoch": 0.12, + "learning_rate": 1.9607763155069884e-05, + "loss": 1.1487, + "step": 1929 + }, + { + "epoch": 0.12, + "learning_rate": 1.9607548466046933e-05, + "loss": 1.1434, + "step": 1930 + }, + { + "epoch": 0.12, + "learning_rate": 1.960733377702398e-05, + "loss": 1.1647, + "step": 1931 + }, + { + "epoch": 0.12, + "learning_rate": 1.960711908800103e-05, + "loss": 1.1087, + "step": 1932 + }, + { + "epoch": 0.12, + "learning_rate": 1.9606904398978083e-05, + "loss": 1.2585, + "step": 1933 + }, + { + "epoch": 0.12, + "learning_rate": 1.9606689709955132e-05, + "loss": 1.2089, + "step": 1934 + }, + { + "epoch": 0.12, + "learning_rate": 1.960647502093218e-05, + "loss": 1.2512, + "step": 1935 + }, + { + "epoch": 0.12, + "learning_rate": 1.960626033190923e-05, + "loss": 1.11, + "step": 1936 + }, + { + "epoch": 0.12, + "learning_rate": 1.960604564288628e-05, + "loss": 1.1166, + "step": 1937 + }, + { + "epoch": 0.12, + "learning_rate": 1.960583095386333e-05, + "loss": 1.0135, + "step": 1938 + }, + { + "epoch": 0.12, + "learning_rate": 1.960561626484038e-05, + "loss": 1.1326, + "step": 1939 + }, + { + "epoch": 0.12, + "learning_rate": 1.9605401575817432e-05, + "loss": 1.0692, + "step": 1940 + }, + { + "epoch": 0.12, + "learning_rate": 1.960518688679448e-05, + "loss": 1.1452, + "step": 1941 + }, + { + "epoch": 0.12, + "learning_rate": 1.960497219777153e-05, + "loss": 1.1826, + "step": 1942 + }, + { + "epoch": 0.13, + "learning_rate": 1.960475750874858e-05, + "loss": 1.1601, + "step": 1943 + }, + { + "epoch": 0.13, + "learning_rate": 1.9604542819725628e-05, + "loss": 1.0688, + "step": 1944 + }, + { + "epoch": 0.13, + "learning_rate": 1.960432813070268e-05, + "loss": 1.2042, + "step": 1945 + }, + { + "epoch": 0.13, + "learning_rate": 1.960411344167973e-05, + "loss": 1.0846, + "step": 1946 + }, + { + "epoch": 0.13, + "learning_rate": 1.960389875265678e-05, + "loss": 1.2672, + "step": 1947 + }, + { + "epoch": 0.13, + "learning_rate": 1.9603684063633827e-05, + "loss": 1.085, + "step": 1948 + }, + { + "epoch": 0.13, + "learning_rate": 1.9603469374610876e-05, + "loss": 1.1662, + "step": 1949 + }, + { + "epoch": 0.13, + "learning_rate": 1.9603254685587925e-05, + "loss": 1.1555, + "step": 1950 + }, + { + "epoch": 0.13, + "learning_rate": 1.9603039996564978e-05, + "loss": 1.1543, + "step": 1951 + }, + { + "epoch": 0.13, + "learning_rate": 1.9602825307542027e-05, + "loss": 1.206, + "step": 1952 + }, + { + "epoch": 0.13, + "learning_rate": 1.9602610618519076e-05, + "loss": 1.2499, + "step": 1953 + }, + { + "epoch": 0.13, + "learning_rate": 1.9602395929496128e-05, + "loss": 1.0919, + "step": 1954 + }, + { + "epoch": 0.13, + "learning_rate": 1.9602181240473177e-05, + "loss": 1.3053, + "step": 1955 + }, + { + "epoch": 0.13, + "learning_rate": 1.9601966551450226e-05, + "loss": 1.1221, + "step": 1956 + }, + { + "epoch": 0.13, + "learning_rate": 1.9601751862427275e-05, + "loss": 1.1907, + "step": 1957 + }, + { + "epoch": 0.13, + "learning_rate": 1.9601537173404327e-05, + "loss": 1.2189, + "step": 1958 + }, + { + "epoch": 0.13, + "learning_rate": 1.9601322484381376e-05, + "loss": 1.1949, + "step": 1959 + }, + { + "epoch": 0.13, + "learning_rate": 1.9601107795358425e-05, + "loss": 1.0796, + "step": 1960 + }, + { + "epoch": 0.13, + "learning_rate": 1.9600893106335474e-05, + "loss": 1.1942, + "step": 1961 + }, + { + "epoch": 0.13, + "learning_rate": 1.9600678417312523e-05, + "loss": 1.2087, + "step": 1962 + }, + { + "epoch": 0.13, + "learning_rate": 1.9600463728289572e-05, + "loss": 1.1964, + "step": 1963 + }, + { + "epoch": 0.13, + "learning_rate": 1.9600249039266624e-05, + "loss": 1.2839, + "step": 1964 + }, + { + "epoch": 0.13, + "learning_rate": 1.9600034350243673e-05, + "loss": 1.2799, + "step": 1965 + }, + { + "epoch": 0.13, + "learning_rate": 1.9599819661220722e-05, + "loss": 1.1056, + "step": 1966 + }, + { + "epoch": 0.13, + "learning_rate": 1.959960497219777e-05, + "loss": 1.0388, + "step": 1967 + }, + { + "epoch": 0.13, + "learning_rate": 1.9599390283174824e-05, + "loss": 1.2056, + "step": 1968 + }, + { + "epoch": 0.13, + "learning_rate": 1.9599175594151873e-05, + "loss": 1.1585, + "step": 1969 + }, + { + "epoch": 0.13, + "learning_rate": 1.959896090512892e-05, + "loss": 1.2226, + "step": 1970 + }, + { + "epoch": 0.13, + "learning_rate": 1.9598746216105974e-05, + "loss": 1.1097, + "step": 1971 + }, + { + "epoch": 0.13, + "learning_rate": 1.9598531527083023e-05, + "loss": 1.2308, + "step": 1972 + }, + { + "epoch": 0.13, + "learning_rate": 1.9598316838060072e-05, + "loss": 1.1439, + "step": 1973 + }, + { + "epoch": 0.13, + "learning_rate": 1.959810214903712e-05, + "loss": 1.1691, + "step": 1974 + }, + { + "epoch": 0.13, + "learning_rate": 1.959788746001417e-05, + "loss": 1.1807, + "step": 1975 + }, + { + "epoch": 0.13, + "learning_rate": 1.9597672770991222e-05, + "loss": 1.1812, + "step": 1976 + }, + { + "epoch": 0.13, + "learning_rate": 1.959745808196827e-05, + "loss": 1.1958, + "step": 1977 + }, + { + "epoch": 0.13, + "learning_rate": 1.959724339294532e-05, + "loss": 1.1617, + "step": 1978 + }, + { + "epoch": 0.13, + "learning_rate": 1.959702870392237e-05, + "loss": 1.218, + "step": 1979 + }, + { + "epoch": 0.13, + "learning_rate": 1.9596814014899418e-05, + "loss": 1.1557, + "step": 1980 + }, + { + "epoch": 0.13, + "learning_rate": 1.9596599325876467e-05, + "loss": 1.1872, + "step": 1981 + }, + { + "epoch": 0.13, + "learning_rate": 1.959638463685352e-05, + "loss": 1.1372, + "step": 1982 + }, + { + "epoch": 0.13, + "learning_rate": 1.959616994783057e-05, + "loss": 1.2495, + "step": 1983 + }, + { + "epoch": 0.13, + "learning_rate": 1.959595525880762e-05, + "loss": 1.0708, + "step": 1984 + }, + { + "epoch": 0.13, + "learning_rate": 1.959574056978467e-05, + "loss": 1.1931, + "step": 1985 + }, + { + "epoch": 0.13, + "learning_rate": 1.959552588076172e-05, + "loss": 1.2306, + "step": 1986 + }, + { + "epoch": 0.13, + "learning_rate": 1.9595311191738768e-05, + "loss": 1.1412, + "step": 1987 + }, + { + "epoch": 0.13, + "learning_rate": 1.9595096502715817e-05, + "loss": 1.1353, + "step": 1988 + }, + { + "epoch": 0.13, + "learning_rate": 1.959488181369287e-05, + "loss": 1.2268, + "step": 1989 + }, + { + "epoch": 0.13, + "learning_rate": 1.9594667124669918e-05, + "loss": 1.1687, + "step": 1990 + }, + { + "epoch": 0.13, + "learning_rate": 1.9594452435646967e-05, + "loss": 1.1301, + "step": 1991 + }, + { + "epoch": 0.13, + "learning_rate": 1.9594237746624016e-05, + "loss": 1.1651, + "step": 1992 + }, + { + "epoch": 0.13, + "learning_rate": 1.9594023057601065e-05, + "loss": 1.2787, + "step": 1993 + }, + { + "epoch": 0.13, + "learning_rate": 1.9593808368578117e-05, + "loss": 1.1954, + "step": 1994 + }, + { + "epoch": 0.13, + "learning_rate": 1.9593593679555166e-05, + "loss": 1.246, + "step": 1995 + }, + { + "epoch": 0.13, + "learning_rate": 1.9593378990532215e-05, + "loss": 1.117, + "step": 1996 + }, + { + "epoch": 0.13, + "learning_rate": 1.9593164301509267e-05, + "loss": 1.1147, + "step": 1997 + }, + { + "epoch": 0.13, + "learning_rate": 1.9592949612486316e-05, + "loss": 1.1293, + "step": 1998 + }, + { + "epoch": 0.13, + "learning_rate": 1.9592734923463365e-05, + "loss": 1.1784, + "step": 1999 + }, + { + "epoch": 0.13, + "learning_rate": 1.9592520234440414e-05, + "loss": 1.2374, + "step": 2000 + }, + { + "epoch": 0.13, + "eval_laion/cider": 0.4587446210259467, + "eval_laion/meteor": 0.3763751930808341, + "eval_laion/pred_num_words": 36.162109375, + "eval_laion/sacrebleu": 14.697159648170405, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.182318925857544, + "eval_runtime": 94.4167, + "eval_samples_per_second": 5.285, + "eval_steps_per_second": 0.339, + "step": 2000 + }, + { + "epoch": 0.13, + "learning_rate": 1.9592305545417463e-05, + "loss": 1.1027, + "step": 2001 + }, + { + "epoch": 0.13, + "learning_rate": 1.9592090856394516e-05, + "loss": 1.1285, + "step": 2002 + }, + { + "epoch": 0.13, + "learning_rate": 1.9591876167371565e-05, + "loss": 1.1424, + "step": 2003 + }, + { + "epoch": 0.13, + "learning_rate": 1.9591661478348614e-05, + "loss": 1.2235, + "step": 2004 + }, + { + "epoch": 0.13, + "learning_rate": 1.9591446789325663e-05, + "loss": 1.2071, + "step": 2005 + }, + { + "epoch": 0.13, + "learning_rate": 1.959123210030271e-05, + "loss": 1.1871, + "step": 2006 + }, + { + "epoch": 0.13, + "learning_rate": 1.9591017411279764e-05, + "loss": 1.1999, + "step": 2007 + }, + { + "epoch": 0.13, + "learning_rate": 1.9590802722256813e-05, + "loss": 1.1409, + "step": 2008 + }, + { + "epoch": 0.13, + "learning_rate": 1.9590588033233862e-05, + "loss": 1.1369, + "step": 2009 + }, + { + "epoch": 0.13, + "learning_rate": 1.959037334421091e-05, + "loss": 1.1044, + "step": 2010 + }, + { + "epoch": 0.13, + "learning_rate": 1.959015865518796e-05, + "loss": 1.0908, + "step": 2011 + }, + { + "epoch": 0.13, + "learning_rate": 1.9589943966165012e-05, + "loss": 1.1592, + "step": 2012 + }, + { + "epoch": 0.13, + "learning_rate": 1.958972927714206e-05, + "loss": 1.1653, + "step": 2013 + }, + { + "epoch": 0.13, + "learning_rate": 1.9589514588119113e-05, + "loss": 1.1746, + "step": 2014 + }, + { + "epoch": 0.13, + "learning_rate": 1.9589299899096162e-05, + "loss": 1.1809, + "step": 2015 + }, + { + "epoch": 0.13, + "learning_rate": 1.958908521007321e-05, + "loss": 1.2119, + "step": 2016 + }, + { + "epoch": 0.13, + "learning_rate": 1.958887052105026e-05, + "loss": 1.201, + "step": 2017 + }, + { + "epoch": 0.13, + "learning_rate": 1.958865583202731e-05, + "loss": 1.1545, + "step": 2018 + }, + { + "epoch": 0.13, + "learning_rate": 1.9588441143004358e-05, + "loss": 1.1395, + "step": 2019 + }, + { + "epoch": 0.13, + "learning_rate": 1.958822645398141e-05, + "loss": 1.1352, + "step": 2020 + }, + { + "epoch": 0.13, + "learning_rate": 1.958801176495846e-05, + "loss": 1.1879, + "step": 2021 + }, + { + "epoch": 0.13, + "learning_rate": 1.958779707593551e-05, + "loss": 1.1654, + "step": 2022 + }, + { + "epoch": 0.13, + "learning_rate": 1.9587582386912557e-05, + "loss": 1.229, + "step": 2023 + }, + { + "epoch": 0.13, + "learning_rate": 1.9587367697889606e-05, + "loss": 1.1898, + "step": 2024 + }, + { + "epoch": 0.13, + "learning_rate": 1.958715300886666e-05, + "loss": 1.1143, + "step": 2025 + }, + { + "epoch": 0.13, + "learning_rate": 1.9586938319843708e-05, + "loss": 1.1284, + "step": 2026 + }, + { + "epoch": 0.13, + "learning_rate": 1.958672363082076e-05, + "loss": 1.1411, + "step": 2027 + }, + { + "epoch": 0.13, + "learning_rate": 1.958650894179781e-05, + "loss": 1.2702, + "step": 2028 + }, + { + "epoch": 0.13, + "learning_rate": 1.9586294252774858e-05, + "loss": 1.1945, + "step": 2029 + }, + { + "epoch": 0.13, + "learning_rate": 1.9586079563751907e-05, + "loss": 1.2084, + "step": 2030 + }, + { + "epoch": 0.13, + "learning_rate": 1.9585864874728956e-05, + "loss": 1.1956, + "step": 2031 + }, + { + "epoch": 0.13, + "learning_rate": 1.9585650185706008e-05, + "loss": 1.0842, + "step": 2032 + }, + { + "epoch": 0.13, + "learning_rate": 1.9585435496683057e-05, + "loss": 1.189, + "step": 2033 + }, + { + "epoch": 0.13, + "learning_rate": 1.9585220807660106e-05, + "loss": 1.0234, + "step": 2034 + }, + { + "epoch": 0.13, + "learning_rate": 1.9585006118637155e-05, + "loss": 1.1724, + "step": 2035 + }, + { + "epoch": 0.13, + "learning_rate": 1.9584791429614204e-05, + "loss": 1.1276, + "step": 2036 + }, + { + "epoch": 0.13, + "learning_rate": 1.9584576740591253e-05, + "loss": 1.1319, + "step": 2037 + }, + { + "epoch": 0.13, + "learning_rate": 1.9584362051568305e-05, + "loss": 1.1962, + "step": 2038 + }, + { + "epoch": 0.13, + "learning_rate": 1.9584147362545354e-05, + "loss": 1.1751, + "step": 2039 + }, + { + "epoch": 0.13, + "learning_rate": 1.9583932673522403e-05, + "loss": 1.0712, + "step": 2040 + }, + { + "epoch": 0.13, + "learning_rate": 1.9583717984499456e-05, + "loss": 1.1658, + "step": 2041 + }, + { + "epoch": 0.13, + "learning_rate": 1.9583503295476505e-05, + "loss": 1.1815, + "step": 2042 + }, + { + "epoch": 0.13, + "learning_rate": 1.9583288606453554e-05, + "loss": 1.1341, + "step": 2043 + }, + { + "epoch": 0.13, + "learning_rate": 1.9583073917430603e-05, + "loss": 1.3205, + "step": 2044 + }, + { + "epoch": 0.13, + "learning_rate": 1.9582859228407655e-05, + "loss": 1.1432, + "step": 2045 + }, + { + "epoch": 0.13, + "learning_rate": 1.9582644539384704e-05, + "loss": 1.196, + "step": 2046 + }, + { + "epoch": 0.13, + "learning_rate": 1.9582429850361753e-05, + "loss": 1.2076, + "step": 2047 + }, + { + "epoch": 0.13, + "learning_rate": 1.9582215161338802e-05, + "loss": 1.1444, + "step": 2048 + }, + { + "epoch": 0.13, + "learning_rate": 1.958200047231585e-05, + "loss": 1.207, + "step": 2049 + }, + { + "epoch": 0.13, + "learning_rate": 1.95817857832929e-05, + "loss": 1.1339, + "step": 2050 + }, + { + "epoch": 0.13, + "learning_rate": 1.9581571094269952e-05, + "loss": 1.1362, + "step": 2051 + }, + { + "epoch": 0.13, + "learning_rate": 1.9581356405247e-05, + "loss": 1.3049, + "step": 2052 + }, + { + "epoch": 0.13, + "learning_rate": 1.958114171622405e-05, + "loss": 1.2371, + "step": 2053 + }, + { + "epoch": 0.13, + "learning_rate": 1.95809270272011e-05, + "loss": 1.1603, + "step": 2054 + }, + { + "epoch": 0.13, + "learning_rate": 1.958071233817815e-05, + "loss": 1.1215, + "step": 2055 + }, + { + "epoch": 0.13, + "learning_rate": 1.95804976491552e-05, + "loss": 1.1907, + "step": 2056 + }, + { + "epoch": 0.13, + "learning_rate": 1.958028296013225e-05, + "loss": 1.1845, + "step": 2057 + }, + { + "epoch": 0.13, + "learning_rate": 1.9580068271109302e-05, + "loss": 1.1884, + "step": 2058 + }, + { + "epoch": 0.13, + "learning_rate": 1.957985358208635e-05, + "loss": 1.181, + "step": 2059 + }, + { + "epoch": 0.13, + "learning_rate": 1.95796388930634e-05, + "loss": 1.2027, + "step": 2060 + }, + { + "epoch": 0.13, + "learning_rate": 1.957942420404045e-05, + "loss": 1.1337, + "step": 2061 + }, + { + "epoch": 0.13, + "learning_rate": 1.9579209515017498e-05, + "loss": 1.0886, + "step": 2062 + }, + { + "epoch": 0.13, + "learning_rate": 1.957899482599455e-05, + "loss": 1.2655, + "step": 2063 + }, + { + "epoch": 0.13, + "learning_rate": 1.95787801369716e-05, + "loss": 1.1735, + "step": 2064 + }, + { + "epoch": 0.13, + "learning_rate": 1.9578565447948648e-05, + "loss": 1.2285, + "step": 2065 + }, + { + "epoch": 0.13, + "learning_rate": 1.9578350758925697e-05, + "loss": 1.2664, + "step": 2066 + }, + { + "epoch": 0.13, + "learning_rate": 1.9578136069902746e-05, + "loss": 1.2686, + "step": 2067 + }, + { + "epoch": 0.13, + "learning_rate": 1.9577921380879795e-05, + "loss": 1.0795, + "step": 2068 + }, + { + "epoch": 0.13, + "learning_rate": 1.9577706691856847e-05, + "loss": 1.1508, + "step": 2069 + }, + { + "epoch": 0.13, + "learning_rate": 1.9577492002833896e-05, + "loss": 1.1437, + "step": 2070 + }, + { + "epoch": 0.13, + "learning_rate": 1.957727731381095e-05, + "loss": 1.2731, + "step": 2071 + }, + { + "epoch": 0.13, + "learning_rate": 1.9577062624787997e-05, + "loss": 1.1374, + "step": 2072 + }, + { + "epoch": 0.13, + "learning_rate": 1.9576847935765046e-05, + "loss": 1.1176, + "step": 2073 + }, + { + "epoch": 0.13, + "learning_rate": 1.9576633246742095e-05, + "loss": 1.116, + "step": 2074 + }, + { + "epoch": 0.13, + "learning_rate": 1.9576418557719144e-05, + "loss": 1.185, + "step": 2075 + }, + { + "epoch": 0.13, + "learning_rate": 1.9576203868696197e-05, + "loss": 1.2109, + "step": 2076 + }, + { + "epoch": 0.13, + "learning_rate": 1.9575989179673246e-05, + "loss": 1.1776, + "step": 2077 + }, + { + "epoch": 0.13, + "learning_rate": 1.9575774490650295e-05, + "loss": 1.0462, + "step": 2078 + }, + { + "epoch": 0.13, + "learning_rate": 1.9575559801627344e-05, + "loss": 1.1298, + "step": 2079 + }, + { + "epoch": 0.13, + "learning_rate": 1.9575345112604392e-05, + "loss": 1.2123, + "step": 2080 + }, + { + "epoch": 0.13, + "learning_rate": 1.9575130423581445e-05, + "loss": 1.1666, + "step": 2081 + }, + { + "epoch": 0.13, + "learning_rate": 1.9574915734558494e-05, + "loss": 1.112, + "step": 2082 + }, + { + "epoch": 0.13, + "learning_rate": 1.9574701045535543e-05, + "loss": 1.3192, + "step": 2083 + }, + { + "epoch": 0.13, + "learning_rate": 1.9574486356512595e-05, + "loss": 1.1665, + "step": 2084 + }, + { + "epoch": 0.13, + "learning_rate": 1.9574271667489644e-05, + "loss": 1.1586, + "step": 2085 + }, + { + "epoch": 0.13, + "learning_rate": 1.9574056978466693e-05, + "loss": 1.1574, + "step": 2086 + }, + { + "epoch": 0.13, + "learning_rate": 1.9573842289443742e-05, + "loss": 1.0861, + "step": 2087 + }, + { + "epoch": 0.13, + "learning_rate": 1.957362760042079e-05, + "loss": 1.1603, + "step": 2088 + }, + { + "epoch": 0.13, + "learning_rate": 1.9573412911397843e-05, + "loss": 1.1088, + "step": 2089 + }, + { + "epoch": 0.13, + "learning_rate": 1.9573198222374892e-05, + "loss": 1.2366, + "step": 2090 + }, + { + "epoch": 0.13, + "learning_rate": 1.957298353335194e-05, + "loss": 1.2196, + "step": 2091 + }, + { + "epoch": 0.13, + "learning_rate": 1.957276884432899e-05, + "loss": 1.1302, + "step": 2092 + }, + { + "epoch": 0.13, + "learning_rate": 1.957255415530604e-05, + "loss": 1.1775, + "step": 2093 + }, + { + "epoch": 0.13, + "learning_rate": 1.957233946628309e-05, + "loss": 1.1473, + "step": 2094 + }, + { + "epoch": 0.13, + "learning_rate": 1.957212477726014e-05, + "loss": 1.1228, + "step": 2095 + }, + { + "epoch": 0.13, + "learning_rate": 1.957191008823719e-05, + "loss": 1.083, + "step": 2096 + }, + { + "epoch": 0.13, + "learning_rate": 1.957169539921424e-05, + "loss": 1.1218, + "step": 2097 + }, + { + "epoch": 0.13, + "learning_rate": 1.9571480710191287e-05, + "loss": 1.0837, + "step": 2098 + }, + { + "epoch": 0.14, + "learning_rate": 1.957126602116834e-05, + "loss": 1.1721, + "step": 2099 + }, + { + "epoch": 0.14, + "learning_rate": 1.957105133214539e-05, + "loss": 1.2627, + "step": 2100 + }, + { + "epoch": 0.14, + "learning_rate": 1.957083664312244e-05, + "loss": 1.247, + "step": 2101 + }, + { + "epoch": 0.14, + "learning_rate": 1.957062195409949e-05, + "loss": 1.1438, + "step": 2102 + }, + { + "epoch": 0.14, + "learning_rate": 1.957040726507654e-05, + "loss": 1.167, + "step": 2103 + }, + { + "epoch": 0.14, + "learning_rate": 1.9570192576053588e-05, + "loss": 1.1524, + "step": 2104 + }, + { + "epoch": 0.14, + "learning_rate": 1.9569977887030637e-05, + "loss": 1.2354, + "step": 2105 + }, + { + "epoch": 0.14, + "learning_rate": 1.9569763198007686e-05, + "loss": 1.1982, + "step": 2106 + }, + { + "epoch": 0.14, + "learning_rate": 1.9569548508984738e-05, + "loss": 1.1253, + "step": 2107 + }, + { + "epoch": 0.14, + "learning_rate": 1.9569333819961787e-05, + "loss": 1.0953, + "step": 2108 + }, + { + "epoch": 0.14, + "learning_rate": 1.9569119130938836e-05, + "loss": 1.2157, + "step": 2109 + }, + { + "epoch": 0.14, + "learning_rate": 1.9568904441915885e-05, + "loss": 1.2277, + "step": 2110 + }, + { + "epoch": 0.14, + "learning_rate": 1.9568689752892934e-05, + "loss": 1.1087, + "step": 2111 + }, + { + "epoch": 0.14, + "learning_rate": 1.9568475063869986e-05, + "loss": 1.2237, + "step": 2112 + }, + { + "epoch": 0.14, + "learning_rate": 1.9568260374847035e-05, + "loss": 1.1658, + "step": 2113 + }, + { + "epoch": 0.14, + "learning_rate": 1.9568045685824088e-05, + "loss": 1.2122, + "step": 2114 + }, + { + "epoch": 0.14, + "learning_rate": 1.9567830996801137e-05, + "loss": 1.226, + "step": 2115 + }, + { + "epoch": 0.14, + "learning_rate": 1.9567616307778186e-05, + "loss": 1.1043, + "step": 2116 + }, + { + "epoch": 0.14, + "learning_rate": 1.9567401618755235e-05, + "loss": 1.1537, + "step": 2117 + }, + { + "epoch": 0.14, + "learning_rate": 1.9567186929732284e-05, + "loss": 1.1585, + "step": 2118 + }, + { + "epoch": 0.14, + "learning_rate": 1.9566972240709336e-05, + "loss": 1.1201, + "step": 2119 + }, + { + "epoch": 0.14, + "learning_rate": 1.9566757551686385e-05, + "loss": 1.1826, + "step": 2120 + }, + { + "epoch": 0.14, + "learning_rate": 1.9566542862663434e-05, + "loss": 1.2621, + "step": 2121 + }, + { + "epoch": 0.14, + "learning_rate": 1.9566328173640483e-05, + "loss": 1.1338, + "step": 2122 + }, + { + "epoch": 0.14, + "learning_rate": 1.9566113484617532e-05, + "loss": 1.1601, + "step": 2123 + }, + { + "epoch": 0.14, + "learning_rate": 1.956589879559458e-05, + "loss": 1.1856, + "step": 2124 + }, + { + "epoch": 0.14, + "learning_rate": 1.9565684106571633e-05, + "loss": 1.1395, + "step": 2125 + }, + { + "epoch": 0.14, + "learning_rate": 1.9565469417548682e-05, + "loss": 1.164, + "step": 2126 + }, + { + "epoch": 0.14, + "learning_rate": 1.956525472852573e-05, + "loss": 1.0823, + "step": 2127 + }, + { + "epoch": 0.14, + "learning_rate": 1.9565040039502783e-05, + "loss": 1.2513, + "step": 2128 + }, + { + "epoch": 0.14, + "learning_rate": 1.9564825350479832e-05, + "loss": 1.2675, + "step": 2129 + }, + { + "epoch": 0.14, + "learning_rate": 1.956461066145688e-05, + "loss": 1.1697, + "step": 2130 + }, + { + "epoch": 0.14, + "learning_rate": 1.956439597243393e-05, + "loss": 1.1773, + "step": 2131 + }, + { + "epoch": 0.14, + "learning_rate": 1.9564181283410983e-05, + "loss": 1.178, + "step": 2132 + }, + { + "epoch": 0.14, + "learning_rate": 1.956396659438803e-05, + "loss": 1.1718, + "step": 2133 + }, + { + "epoch": 0.14, + "learning_rate": 1.956375190536508e-05, + "loss": 1.1422, + "step": 2134 + }, + { + "epoch": 0.14, + "learning_rate": 1.956353721634213e-05, + "loss": 1.1092, + "step": 2135 + }, + { + "epoch": 0.14, + "learning_rate": 1.956332252731918e-05, + "loss": 1.1106, + "step": 2136 + }, + { + "epoch": 0.14, + "learning_rate": 1.9563107838296228e-05, + "loss": 1.1874, + "step": 2137 + }, + { + "epoch": 0.14, + "learning_rate": 1.956289314927328e-05, + "loss": 1.1484, + "step": 2138 + }, + { + "epoch": 0.14, + "learning_rate": 1.956267846025033e-05, + "loss": 1.1712, + "step": 2139 + }, + { + "epoch": 0.14, + "learning_rate": 1.9562463771227378e-05, + "loss": 1.2448, + "step": 2140 + }, + { + "epoch": 0.14, + "learning_rate": 1.9562249082204427e-05, + "loss": 1.1665, + "step": 2141 + }, + { + "epoch": 0.14, + "learning_rate": 1.9562034393181476e-05, + "loss": 1.1169, + "step": 2142 + }, + { + "epoch": 0.14, + "learning_rate": 1.9561819704158528e-05, + "loss": 1.176, + "step": 2143 + }, + { + "epoch": 0.14, + "learning_rate": 1.9561605015135577e-05, + "loss": 1.1987, + "step": 2144 + }, + { + "epoch": 0.14, + "learning_rate": 1.956139032611263e-05, + "loss": 1.0669, + "step": 2145 + }, + { + "epoch": 0.14, + "learning_rate": 1.956117563708968e-05, + "loss": 1.182, + "step": 2146 + }, + { + "epoch": 0.14, + "learning_rate": 1.9560960948066727e-05, + "loss": 1.1963, + "step": 2147 + }, + { + "epoch": 0.14, + "learning_rate": 1.9560746259043776e-05, + "loss": 1.2022, + "step": 2148 + }, + { + "epoch": 0.14, + "learning_rate": 1.9560531570020825e-05, + "loss": 1.1509, + "step": 2149 + }, + { + "epoch": 0.14, + "learning_rate": 1.9560316880997878e-05, + "loss": 1.1173, + "step": 2150 + }, + { + "epoch": 0.14, + "learning_rate": 1.9560102191974927e-05, + "loss": 1.1478, + "step": 2151 + }, + { + "epoch": 0.14, + "learning_rate": 1.9559887502951976e-05, + "loss": 1.0673, + "step": 2152 + }, + { + "epoch": 0.14, + "learning_rate": 1.9559672813929025e-05, + "loss": 1.1692, + "step": 2153 + }, + { + "epoch": 0.14, + "learning_rate": 1.9559458124906073e-05, + "loss": 1.202, + "step": 2154 + }, + { + "epoch": 0.14, + "learning_rate": 1.9559243435883122e-05, + "loss": 1.1834, + "step": 2155 + }, + { + "epoch": 0.14, + "learning_rate": 1.9559028746860175e-05, + "loss": 1.1236, + "step": 2156 + }, + { + "epoch": 0.14, + "learning_rate": 1.9558814057837224e-05, + "loss": 1.2501, + "step": 2157 + }, + { + "epoch": 0.14, + "learning_rate": 1.9558599368814276e-05, + "loss": 1.3183, + "step": 2158 + }, + { + "epoch": 0.14, + "learning_rate": 1.9558384679791325e-05, + "loss": 1.1323, + "step": 2159 + }, + { + "epoch": 0.14, + "learning_rate": 1.9558169990768374e-05, + "loss": 1.13, + "step": 2160 + }, + { + "epoch": 0.14, + "learning_rate": 1.9557955301745423e-05, + "loss": 1.2179, + "step": 2161 + }, + { + "epoch": 0.14, + "learning_rate": 1.9557740612722472e-05, + "loss": 1.185, + "step": 2162 + }, + { + "epoch": 0.14, + "learning_rate": 1.9557525923699524e-05, + "loss": 1.1402, + "step": 2163 + }, + { + "epoch": 0.14, + "learning_rate": 1.9557311234676573e-05, + "loss": 1.1577, + "step": 2164 + }, + { + "epoch": 0.14, + "learning_rate": 1.9557096545653622e-05, + "loss": 1.3174, + "step": 2165 + }, + { + "epoch": 0.14, + "learning_rate": 1.955688185663067e-05, + "loss": 1.2788, + "step": 2166 + }, + { + "epoch": 0.14, + "learning_rate": 1.955666716760772e-05, + "loss": 1.15, + "step": 2167 + }, + { + "epoch": 0.14, + "learning_rate": 1.9556452478584773e-05, + "loss": 1.1448, + "step": 2168 + }, + { + "epoch": 0.14, + "learning_rate": 1.955623778956182e-05, + "loss": 1.1723, + "step": 2169 + }, + { + "epoch": 0.14, + "learning_rate": 1.955602310053887e-05, + "loss": 1.1644, + "step": 2170 + }, + { + "epoch": 0.14, + "learning_rate": 1.955580841151592e-05, + "loss": 1.1146, + "step": 2171 + }, + { + "epoch": 0.14, + "learning_rate": 1.9555593722492972e-05, + "loss": 1.2487, + "step": 2172 + }, + { + "epoch": 0.14, + "learning_rate": 1.955537903347002e-05, + "loss": 1.2159, + "step": 2173 + }, + { + "epoch": 0.14, + "learning_rate": 1.955516434444707e-05, + "loss": 1.1305, + "step": 2174 + }, + { + "epoch": 0.14, + "learning_rate": 1.955494965542412e-05, + "loss": 1.1798, + "step": 2175 + }, + { + "epoch": 0.14, + "learning_rate": 1.955473496640117e-05, + "loss": 1.141, + "step": 2176 + }, + { + "epoch": 0.14, + "learning_rate": 1.955452027737822e-05, + "loss": 1.1239, + "step": 2177 + }, + { + "epoch": 0.14, + "learning_rate": 1.955430558835527e-05, + "loss": 1.0894, + "step": 2178 + }, + { + "epoch": 0.14, + "learning_rate": 1.9554090899332318e-05, + "loss": 1.1335, + "step": 2179 + }, + { + "epoch": 0.14, + "learning_rate": 1.9553876210309367e-05, + "loss": 1.1999, + "step": 2180 + }, + { + "epoch": 0.14, + "learning_rate": 1.955366152128642e-05, + "loss": 1.094, + "step": 2181 + }, + { + "epoch": 0.14, + "learning_rate": 1.9553446832263468e-05, + "loss": 1.0847, + "step": 2182 + }, + { + "epoch": 0.14, + "learning_rate": 1.9553232143240517e-05, + "loss": 1.2194, + "step": 2183 + }, + { + "epoch": 0.14, + "learning_rate": 1.9553017454217566e-05, + "loss": 1.1685, + "step": 2184 + }, + { + "epoch": 0.14, + "learning_rate": 1.9552802765194615e-05, + "loss": 1.1637, + "step": 2185 + }, + { + "epoch": 0.14, + "learning_rate": 1.9552588076171667e-05, + "loss": 1.1362, + "step": 2186 + }, + { + "epoch": 0.14, + "learning_rate": 1.9552373387148716e-05, + "loss": 1.1566, + "step": 2187 + }, + { + "epoch": 0.14, + "learning_rate": 1.955215869812577e-05, + "loss": 1.2416, + "step": 2188 + }, + { + "epoch": 0.14, + "learning_rate": 1.9551944009102818e-05, + "loss": 1.1094, + "step": 2189 + }, + { + "epoch": 0.14, + "learning_rate": 1.9551729320079867e-05, + "loss": 1.1465, + "step": 2190 + }, + { + "epoch": 0.14, + "learning_rate": 1.9551514631056916e-05, + "loss": 1.1136, + "step": 2191 + }, + { + "epoch": 0.14, + "learning_rate": 1.9551299942033965e-05, + "loss": 1.2364, + "step": 2192 + }, + { + "epoch": 0.14, + "learning_rate": 1.9551085253011014e-05, + "loss": 1.0688, + "step": 2193 + }, + { + "epoch": 0.14, + "learning_rate": 1.9550870563988066e-05, + "loss": 1.2476, + "step": 2194 + }, + { + "epoch": 0.14, + "learning_rate": 1.9550655874965115e-05, + "loss": 1.1344, + "step": 2195 + }, + { + "epoch": 0.14, + "learning_rate": 1.9550441185942164e-05, + "loss": 1.2732, + "step": 2196 + }, + { + "epoch": 0.14, + "learning_rate": 1.9550226496919213e-05, + "loss": 1.108, + "step": 2197 + }, + { + "epoch": 0.14, + "learning_rate": 1.9550011807896262e-05, + "loss": 1.1755, + "step": 2198 + }, + { + "epoch": 0.14, + "learning_rate": 1.9549797118873314e-05, + "loss": 1.1546, + "step": 2199 + }, + { + "epoch": 0.14, + "learning_rate": 1.9549582429850363e-05, + "loss": 1.2536, + "step": 2200 + }, + { + "epoch": 0.14, + "learning_rate": 1.9549367740827416e-05, + "loss": 1.2329, + "step": 2201 + }, + { + "epoch": 0.14, + "learning_rate": 1.9549153051804464e-05, + "loss": 1.159, + "step": 2202 + }, + { + "epoch": 0.14, + "learning_rate": 1.9548938362781513e-05, + "loss": 1.1292, + "step": 2203 + }, + { + "epoch": 0.14, + "learning_rate": 1.9548723673758562e-05, + "loss": 1.2212, + "step": 2204 + }, + { + "epoch": 0.14, + "learning_rate": 1.954850898473561e-05, + "loss": 1.1352, + "step": 2205 + }, + { + "epoch": 0.14, + "learning_rate": 1.954829429571266e-05, + "loss": 1.2148, + "step": 2206 + }, + { + "epoch": 0.14, + "learning_rate": 1.9548079606689713e-05, + "loss": 1.1571, + "step": 2207 + }, + { + "epoch": 0.14, + "learning_rate": 1.954786491766676e-05, + "loss": 1.1651, + "step": 2208 + }, + { + "epoch": 0.14, + "learning_rate": 1.954765022864381e-05, + "loss": 1.248, + "step": 2209 + }, + { + "epoch": 0.14, + "learning_rate": 1.954743553962086e-05, + "loss": 1.0464, + "step": 2210 + }, + { + "epoch": 0.14, + "learning_rate": 1.954722085059791e-05, + "loss": 1.1176, + "step": 2211 + }, + { + "epoch": 0.14, + "learning_rate": 1.954700616157496e-05, + "loss": 1.0557, + "step": 2212 + }, + { + "epoch": 0.14, + "learning_rate": 1.954679147255201e-05, + "loss": 1.2156, + "step": 2213 + }, + { + "epoch": 0.14, + "learning_rate": 1.954657678352906e-05, + "loss": 1.2689, + "step": 2214 + }, + { + "epoch": 0.14, + "learning_rate": 1.954636209450611e-05, + "loss": 1.1406, + "step": 2215 + }, + { + "epoch": 0.14, + "learning_rate": 1.954614740548316e-05, + "loss": 1.1278, + "step": 2216 + }, + { + "epoch": 0.14, + "learning_rate": 1.954593271646021e-05, + "loss": 1.2008, + "step": 2217 + }, + { + "epoch": 0.14, + "learning_rate": 1.9545718027437258e-05, + "loss": 1.1314, + "step": 2218 + }, + { + "epoch": 0.14, + "learning_rate": 1.954550333841431e-05, + "loss": 1.0615, + "step": 2219 + }, + { + "epoch": 0.14, + "learning_rate": 1.954528864939136e-05, + "loss": 1.1722, + "step": 2220 + }, + { + "epoch": 0.14, + "learning_rate": 1.954507396036841e-05, + "loss": 1.1139, + "step": 2221 + }, + { + "epoch": 0.14, + "learning_rate": 1.9544859271345457e-05, + "loss": 1.09, + "step": 2222 + }, + { + "epoch": 0.14, + "learning_rate": 1.9544644582322506e-05, + "loss": 1.1526, + "step": 2223 + }, + { + "epoch": 0.14, + "learning_rate": 1.9544429893299555e-05, + "loss": 1.1346, + "step": 2224 + }, + { + "epoch": 0.14, + "learning_rate": 1.9544215204276608e-05, + "loss": 1.1199, + "step": 2225 + }, + { + "epoch": 0.14, + "learning_rate": 1.9544000515253657e-05, + "loss": 1.1288, + "step": 2226 + }, + { + "epoch": 0.14, + "learning_rate": 1.9543785826230706e-05, + "loss": 1.193, + "step": 2227 + }, + { + "epoch": 0.14, + "learning_rate": 1.9543571137207754e-05, + "loss": 1.1733, + "step": 2228 + }, + { + "epoch": 0.14, + "learning_rate": 1.9543356448184803e-05, + "loss": 1.0655, + "step": 2229 + }, + { + "epoch": 0.14, + "learning_rate": 1.9543141759161856e-05, + "loss": 1.1442, + "step": 2230 + }, + { + "epoch": 0.14, + "learning_rate": 1.9542927070138905e-05, + "loss": 1.1131, + "step": 2231 + }, + { + "epoch": 0.14, + "learning_rate": 1.9542712381115957e-05, + "loss": 1.2659, + "step": 2232 + }, + { + "epoch": 0.14, + "learning_rate": 1.9542497692093006e-05, + "loss": 1.1343, + "step": 2233 + }, + { + "epoch": 0.14, + "learning_rate": 1.9542283003070055e-05, + "loss": 1.148, + "step": 2234 + }, + { + "epoch": 0.14, + "learning_rate": 1.9542068314047104e-05, + "loss": 1.174, + "step": 2235 + }, + { + "epoch": 0.14, + "learning_rate": 1.9541853625024153e-05, + "loss": 1.1003, + "step": 2236 + }, + { + "epoch": 0.14, + "learning_rate": 1.9541638936001205e-05, + "loss": 1.1484, + "step": 2237 + }, + { + "epoch": 0.14, + "learning_rate": 1.9541424246978254e-05, + "loss": 1.1463, + "step": 2238 + }, + { + "epoch": 0.14, + "learning_rate": 1.9541209557955303e-05, + "loss": 1.2094, + "step": 2239 + }, + { + "epoch": 0.14, + "learning_rate": 1.9540994868932352e-05, + "loss": 1.1095, + "step": 2240 + }, + { + "epoch": 0.14, + "learning_rate": 1.95407801799094e-05, + "loss": 1.2154, + "step": 2241 + }, + { + "epoch": 0.14, + "learning_rate": 1.954056549088645e-05, + "loss": 1.1306, + "step": 2242 + }, + { + "epoch": 0.14, + "learning_rate": 1.9540350801863503e-05, + "loss": 1.1349, + "step": 2243 + }, + { + "epoch": 0.14, + "learning_rate": 1.954013611284055e-05, + "loss": 1.1697, + "step": 2244 + }, + { + "epoch": 0.14, + "learning_rate": 1.9539921423817604e-05, + "loss": 1.1638, + "step": 2245 + }, + { + "epoch": 0.14, + "learning_rate": 1.9539706734794653e-05, + "loss": 1.1442, + "step": 2246 + }, + { + "epoch": 0.14, + "learning_rate": 1.9539492045771702e-05, + "loss": 1.1585, + "step": 2247 + }, + { + "epoch": 0.14, + "learning_rate": 1.953927735674875e-05, + "loss": 1.1518, + "step": 2248 + }, + { + "epoch": 0.14, + "learning_rate": 1.95390626677258e-05, + "loss": 1.1098, + "step": 2249 + }, + { + "epoch": 0.14, + "learning_rate": 1.9538847978702852e-05, + "loss": 1.2719, + "step": 2250 + }, + { + "epoch": 0.14, + "learning_rate": 1.95386332896799e-05, + "loss": 1.1223, + "step": 2251 + }, + { + "epoch": 0.14, + "learning_rate": 1.953841860065695e-05, + "loss": 1.1948, + "step": 2252 + }, + { + "epoch": 0.14, + "learning_rate": 1.9538203911634e-05, + "loss": 1.1513, + "step": 2253 + }, + { + "epoch": 0.15, + "learning_rate": 1.9537989222611048e-05, + "loss": 1.1797, + "step": 2254 + }, + { + "epoch": 0.15, + "learning_rate": 1.95377745335881e-05, + "loss": 1.1176, + "step": 2255 + }, + { + "epoch": 0.15, + "learning_rate": 1.953755984456515e-05, + "loss": 1.1153, + "step": 2256 + }, + { + "epoch": 0.15, + "learning_rate": 1.9537345155542198e-05, + "loss": 1.1306, + "step": 2257 + }, + { + "epoch": 0.15, + "learning_rate": 1.9537130466519247e-05, + "loss": 1.1234, + "step": 2258 + }, + { + "epoch": 0.15, + "learning_rate": 1.95369157774963e-05, + "loss": 1.1593, + "step": 2259 + }, + { + "epoch": 0.15, + "learning_rate": 1.953670108847335e-05, + "loss": 1.1574, + "step": 2260 + }, + { + "epoch": 0.15, + "learning_rate": 1.9536486399450397e-05, + "loss": 1.0623, + "step": 2261 + }, + { + "epoch": 0.15, + "learning_rate": 1.9536271710427446e-05, + "loss": 1.2352, + "step": 2262 + }, + { + "epoch": 0.15, + "learning_rate": 1.95360570214045e-05, + "loss": 1.024, + "step": 2263 + }, + { + "epoch": 0.15, + "learning_rate": 1.9535842332381548e-05, + "loss": 1.1625, + "step": 2264 + }, + { + "epoch": 0.15, + "learning_rate": 1.9535627643358597e-05, + "loss": 1.1302, + "step": 2265 + }, + { + "epoch": 0.15, + "learning_rate": 1.9535412954335646e-05, + "loss": 1.122, + "step": 2266 + }, + { + "epoch": 0.15, + "learning_rate": 1.9535198265312695e-05, + "loss": 1.1599, + "step": 2267 + }, + { + "epoch": 0.15, + "learning_rate": 1.9534983576289747e-05, + "loss": 1.1699, + "step": 2268 + }, + { + "epoch": 0.15, + "learning_rate": 1.9534768887266796e-05, + "loss": 1.1802, + "step": 2269 + }, + { + "epoch": 0.15, + "learning_rate": 1.9534554198243845e-05, + "loss": 1.1174, + "step": 2270 + }, + { + "epoch": 0.15, + "learning_rate": 1.9534339509220894e-05, + "loss": 1.157, + "step": 2271 + }, + { + "epoch": 0.15, + "learning_rate": 1.9534124820197943e-05, + "loss": 1.1158, + "step": 2272 + }, + { + "epoch": 0.15, + "learning_rate": 1.9533910131174995e-05, + "loss": 1.1698, + "step": 2273 + }, + { + "epoch": 0.15, + "learning_rate": 1.9533695442152044e-05, + "loss": 1.144, + "step": 2274 + }, + { + "epoch": 0.15, + "learning_rate": 1.9533480753129097e-05, + "loss": 1.1628, + "step": 2275 + }, + { + "epoch": 0.15, + "learning_rate": 1.9533266064106145e-05, + "loss": 1.0658, + "step": 2276 + }, + { + "epoch": 0.15, + "learning_rate": 1.9533051375083194e-05, + "loss": 1.1687, + "step": 2277 + }, + { + "epoch": 0.15, + "learning_rate": 1.9532836686060243e-05, + "loss": 1.148, + "step": 2278 + }, + { + "epoch": 0.15, + "learning_rate": 1.9532621997037292e-05, + "loss": 1.1119, + "step": 2279 + }, + { + "epoch": 0.15, + "learning_rate": 1.953240730801434e-05, + "loss": 1.2051, + "step": 2280 + }, + { + "epoch": 0.15, + "learning_rate": 1.9532192618991394e-05, + "loss": 1.1539, + "step": 2281 + }, + { + "epoch": 0.15, + "learning_rate": 1.9531977929968443e-05, + "loss": 1.1146, + "step": 2282 + }, + { + "epoch": 0.15, + "learning_rate": 1.953176324094549e-05, + "loss": 1.1265, + "step": 2283 + }, + { + "epoch": 0.15, + "learning_rate": 1.953154855192254e-05, + "loss": 1.1383, + "step": 2284 + }, + { + "epoch": 0.15, + "learning_rate": 1.953133386289959e-05, + "loss": 1.1444, + "step": 2285 + }, + { + "epoch": 0.15, + "learning_rate": 1.9531119173876642e-05, + "loss": 1.1957, + "step": 2286 + }, + { + "epoch": 0.15, + "learning_rate": 1.953090448485369e-05, + "loss": 1.1497, + "step": 2287 + }, + { + "epoch": 0.15, + "learning_rate": 1.9530689795830743e-05, + "loss": 1.1753, + "step": 2288 + }, + { + "epoch": 0.15, + "learning_rate": 1.9530475106807792e-05, + "loss": 1.1525, + "step": 2289 + }, + { + "epoch": 0.15, + "learning_rate": 1.953026041778484e-05, + "loss": 1.1125, + "step": 2290 + }, + { + "epoch": 0.15, + "learning_rate": 1.953004572876189e-05, + "loss": 1.2096, + "step": 2291 + }, + { + "epoch": 0.15, + "learning_rate": 1.952983103973894e-05, + "loss": 1.139, + "step": 2292 + }, + { + "epoch": 0.15, + "learning_rate": 1.9529616350715988e-05, + "loss": 1.1014, + "step": 2293 + }, + { + "epoch": 0.15, + "learning_rate": 1.952940166169304e-05, + "loss": 1.1106, + "step": 2294 + }, + { + "epoch": 0.15, + "learning_rate": 1.952918697267009e-05, + "loss": 1.163, + "step": 2295 + }, + { + "epoch": 0.15, + "learning_rate": 1.952897228364714e-05, + "loss": 1.1073, + "step": 2296 + }, + { + "epoch": 0.15, + "learning_rate": 1.9528757594624187e-05, + "loss": 1.2478, + "step": 2297 + }, + { + "epoch": 0.15, + "learning_rate": 1.9528542905601236e-05, + "loss": 1.1999, + "step": 2298 + }, + { + "epoch": 0.15, + "learning_rate": 1.952832821657829e-05, + "loss": 1.1821, + "step": 2299 + }, + { + "epoch": 0.15, + "learning_rate": 1.9528113527555338e-05, + "loss": 1.1759, + "step": 2300 + }, + { + "epoch": 0.15, + "learning_rate": 1.9527898838532387e-05, + "loss": 1.1245, + "step": 2301 + }, + { + "epoch": 0.15, + "learning_rate": 1.952768414950944e-05, + "loss": 1.1877, + "step": 2302 + }, + { + "epoch": 0.15, + "learning_rate": 1.9527469460486488e-05, + "loss": 1.1259, + "step": 2303 + }, + { + "epoch": 0.15, + "learning_rate": 1.9527254771463537e-05, + "loss": 1.1948, + "step": 2304 + }, + { + "epoch": 0.15, + "learning_rate": 1.9527040082440586e-05, + "loss": 1.1203, + "step": 2305 + }, + { + "epoch": 0.15, + "learning_rate": 1.9526825393417638e-05, + "loss": 1.2014, + "step": 2306 + }, + { + "epoch": 0.15, + "learning_rate": 1.9526610704394687e-05, + "loss": 1.1765, + "step": 2307 + }, + { + "epoch": 0.15, + "learning_rate": 1.9526396015371736e-05, + "loss": 1.1141, + "step": 2308 + }, + { + "epoch": 0.15, + "learning_rate": 1.9526181326348785e-05, + "loss": 1.1557, + "step": 2309 + }, + { + "epoch": 0.15, + "learning_rate": 1.9525966637325834e-05, + "loss": 1.1212, + "step": 2310 + }, + { + "epoch": 0.15, + "learning_rate": 1.9525751948302883e-05, + "loss": 1.1507, + "step": 2311 + }, + { + "epoch": 0.15, + "learning_rate": 1.9525537259279935e-05, + "loss": 1.2396, + "step": 2312 + }, + { + "epoch": 0.15, + "learning_rate": 1.9525322570256984e-05, + "loss": 1.0723, + "step": 2313 + }, + { + "epoch": 0.15, + "learning_rate": 1.9525107881234033e-05, + "loss": 1.1715, + "step": 2314 + }, + { + "epoch": 0.15, + "learning_rate": 1.9524893192211082e-05, + "loss": 1.0892, + "step": 2315 + }, + { + "epoch": 0.15, + "learning_rate": 1.952467850318813e-05, + "loss": 1.1941, + "step": 2316 + }, + { + "epoch": 0.15, + "learning_rate": 1.9524463814165184e-05, + "loss": 1.156, + "step": 2317 + }, + { + "epoch": 0.15, + "learning_rate": 1.9524249125142232e-05, + "loss": 1.1738, + "step": 2318 + }, + { + "epoch": 0.15, + "learning_rate": 1.9524034436119285e-05, + "loss": 1.1093, + "step": 2319 + }, + { + "epoch": 0.15, + "learning_rate": 1.9523819747096334e-05, + "loss": 1.218, + "step": 2320 + }, + { + "epoch": 0.15, + "learning_rate": 1.9523605058073383e-05, + "loss": 1.0759, + "step": 2321 + }, + { + "epoch": 0.15, + "learning_rate": 1.9523390369050432e-05, + "loss": 1.1963, + "step": 2322 + }, + { + "epoch": 0.15, + "learning_rate": 1.952317568002748e-05, + "loss": 1.223, + "step": 2323 + }, + { + "epoch": 0.15, + "learning_rate": 1.9522960991004533e-05, + "loss": 1.146, + "step": 2324 + }, + { + "epoch": 0.15, + "learning_rate": 1.9522746301981582e-05, + "loss": 1.2152, + "step": 2325 + }, + { + "epoch": 0.15, + "learning_rate": 1.952253161295863e-05, + "loss": 1.1818, + "step": 2326 + }, + { + "epoch": 0.15, + "learning_rate": 1.952231692393568e-05, + "loss": 1.1095, + "step": 2327 + }, + { + "epoch": 0.15, + "learning_rate": 1.952210223491273e-05, + "loss": 1.2126, + "step": 2328 + }, + { + "epoch": 0.15, + "learning_rate": 1.9521887545889778e-05, + "loss": 1.1845, + "step": 2329 + }, + { + "epoch": 0.15, + "learning_rate": 1.952167285686683e-05, + "loss": 1.1936, + "step": 2330 + }, + { + "epoch": 0.15, + "learning_rate": 1.952145816784388e-05, + "loss": 1.0638, + "step": 2331 + }, + { + "epoch": 0.15, + "learning_rate": 1.952124347882093e-05, + "loss": 1.1219, + "step": 2332 + }, + { + "epoch": 0.15, + "learning_rate": 1.952102878979798e-05, + "loss": 1.1547, + "step": 2333 + }, + { + "epoch": 0.15, + "learning_rate": 1.952081410077503e-05, + "loss": 1.1237, + "step": 2334 + }, + { + "epoch": 0.15, + "learning_rate": 1.952059941175208e-05, + "loss": 1.1228, + "step": 2335 + }, + { + "epoch": 0.15, + "learning_rate": 1.9520384722729127e-05, + "loss": 1.1461, + "step": 2336 + }, + { + "epoch": 0.15, + "learning_rate": 1.952017003370618e-05, + "loss": 1.1184, + "step": 2337 + }, + { + "epoch": 0.15, + "learning_rate": 1.951995534468323e-05, + "loss": 1.163, + "step": 2338 + }, + { + "epoch": 0.15, + "learning_rate": 1.9519740655660278e-05, + "loss": 1.1811, + "step": 2339 + }, + { + "epoch": 0.15, + "learning_rate": 1.9519525966637327e-05, + "loss": 1.1087, + "step": 2340 + }, + { + "epoch": 0.15, + "learning_rate": 1.9519311277614376e-05, + "loss": 1.1697, + "step": 2341 + }, + { + "epoch": 0.15, + "learning_rate": 1.9519096588591428e-05, + "loss": 1.1366, + "step": 2342 + }, + { + "epoch": 0.15, + "learning_rate": 1.9518881899568477e-05, + "loss": 1.1785, + "step": 2343 + }, + { + "epoch": 0.15, + "learning_rate": 1.9518667210545526e-05, + "loss": 1.179, + "step": 2344 + }, + { + "epoch": 0.15, + "learning_rate": 1.9518452521522575e-05, + "loss": 1.1844, + "step": 2345 + }, + { + "epoch": 0.15, + "learning_rate": 1.9518237832499627e-05, + "loss": 1.1329, + "step": 2346 + }, + { + "epoch": 0.15, + "learning_rate": 1.9518023143476676e-05, + "loss": 1.1653, + "step": 2347 + }, + { + "epoch": 0.15, + "learning_rate": 1.9517808454453725e-05, + "loss": 1.1351, + "step": 2348 + }, + { + "epoch": 0.15, + "learning_rate": 1.9517593765430774e-05, + "loss": 1.1492, + "step": 2349 + }, + { + "epoch": 0.15, + "learning_rate": 1.9517379076407826e-05, + "loss": 1.1658, + "step": 2350 + }, + { + "epoch": 0.15, + "learning_rate": 1.9517164387384875e-05, + "loss": 1.1633, + "step": 2351 + }, + { + "epoch": 0.15, + "learning_rate": 1.9516949698361924e-05, + "loss": 1.1394, + "step": 2352 + }, + { + "epoch": 0.15, + "learning_rate": 1.9516735009338973e-05, + "loss": 1.0991, + "step": 2353 + }, + { + "epoch": 0.15, + "learning_rate": 1.9516520320316022e-05, + "loss": 1.0806, + "step": 2354 + }, + { + "epoch": 0.15, + "learning_rate": 1.9516305631293075e-05, + "loss": 1.1385, + "step": 2355 + }, + { + "epoch": 0.15, + "learning_rate": 1.9516090942270124e-05, + "loss": 1.3194, + "step": 2356 + }, + { + "epoch": 0.15, + "learning_rate": 1.9515876253247173e-05, + "loss": 1.1135, + "step": 2357 + }, + { + "epoch": 0.15, + "learning_rate": 1.951566156422422e-05, + "loss": 1.1383, + "step": 2358 + }, + { + "epoch": 0.15, + "learning_rate": 1.951544687520127e-05, + "loss": 1.1946, + "step": 2359 + }, + { + "epoch": 0.15, + "learning_rate": 1.9515232186178323e-05, + "loss": 1.2221, + "step": 2360 + }, + { + "epoch": 0.15, + "learning_rate": 1.9515017497155372e-05, + "loss": 1.2094, + "step": 2361 + }, + { + "epoch": 0.15, + "learning_rate": 1.9514802808132424e-05, + "loss": 1.2378, + "step": 2362 + }, + { + "epoch": 0.15, + "learning_rate": 1.9514588119109473e-05, + "loss": 1.1919, + "step": 2363 + }, + { + "epoch": 0.15, + "learning_rate": 1.9514373430086522e-05, + "loss": 1.1185, + "step": 2364 + }, + { + "epoch": 0.15, + "learning_rate": 1.951415874106357e-05, + "loss": 1.1549, + "step": 2365 + }, + { + "epoch": 0.15, + "learning_rate": 1.951394405204062e-05, + "loss": 1.0522, + "step": 2366 + }, + { + "epoch": 0.15, + "learning_rate": 1.951372936301767e-05, + "loss": 1.1523, + "step": 2367 + }, + { + "epoch": 0.15, + "learning_rate": 1.951351467399472e-05, + "loss": 1.1329, + "step": 2368 + }, + { + "epoch": 0.15, + "learning_rate": 1.951329998497177e-05, + "loss": 1.1253, + "step": 2369 + }, + { + "epoch": 0.15, + "learning_rate": 1.951308529594882e-05, + "loss": 1.1466, + "step": 2370 + }, + { + "epoch": 0.15, + "learning_rate": 1.9512870606925868e-05, + "loss": 1.1665, + "step": 2371 + }, + { + "epoch": 0.15, + "learning_rate": 1.9512655917902917e-05, + "loss": 1.1189, + "step": 2372 + }, + { + "epoch": 0.15, + "learning_rate": 1.951244122887997e-05, + "loss": 1.166, + "step": 2373 + }, + { + "epoch": 0.15, + "learning_rate": 1.951222653985702e-05, + "loss": 1.1816, + "step": 2374 + }, + { + "epoch": 0.15, + "learning_rate": 1.951201185083407e-05, + "loss": 1.1043, + "step": 2375 + }, + { + "epoch": 0.15, + "learning_rate": 1.951179716181112e-05, + "loss": 1.2111, + "step": 2376 + }, + { + "epoch": 0.15, + "learning_rate": 1.951158247278817e-05, + "loss": 1.1813, + "step": 2377 + }, + { + "epoch": 0.15, + "learning_rate": 1.9511367783765218e-05, + "loss": 1.0937, + "step": 2378 + }, + { + "epoch": 0.15, + "learning_rate": 1.9511153094742267e-05, + "loss": 1.17, + "step": 2379 + }, + { + "epoch": 0.15, + "learning_rate": 1.9510938405719316e-05, + "loss": 1.2198, + "step": 2380 + }, + { + "epoch": 0.15, + "learning_rate": 1.9510723716696368e-05, + "loss": 1.2138, + "step": 2381 + }, + { + "epoch": 0.15, + "learning_rate": 1.9510509027673417e-05, + "loss": 1.1633, + "step": 2382 + }, + { + "epoch": 0.15, + "learning_rate": 1.9510294338650466e-05, + "loss": 1.0408, + "step": 2383 + }, + { + "epoch": 0.15, + "learning_rate": 1.9510079649627515e-05, + "loss": 1.2289, + "step": 2384 + }, + { + "epoch": 0.15, + "learning_rate": 1.9509864960604564e-05, + "loss": 1.264, + "step": 2385 + }, + { + "epoch": 0.15, + "learning_rate": 1.9509650271581616e-05, + "loss": 1.0286, + "step": 2386 + }, + { + "epoch": 0.15, + "learning_rate": 1.9509435582558665e-05, + "loss": 1.1115, + "step": 2387 + }, + { + "epoch": 0.15, + "learning_rate": 1.9509220893535714e-05, + "loss": 1.1357, + "step": 2388 + }, + { + "epoch": 0.15, + "learning_rate": 1.9509006204512767e-05, + "loss": 1.1671, + "step": 2389 + }, + { + "epoch": 0.15, + "learning_rate": 1.9508791515489816e-05, + "loss": 1.0965, + "step": 2390 + }, + { + "epoch": 0.15, + "learning_rate": 1.9508576826466865e-05, + "loss": 1.0713, + "step": 2391 + }, + { + "epoch": 0.15, + "learning_rate": 1.9508362137443913e-05, + "loss": 1.1426, + "step": 2392 + }, + { + "epoch": 0.15, + "learning_rate": 1.9508147448420966e-05, + "loss": 1.199, + "step": 2393 + }, + { + "epoch": 0.15, + "learning_rate": 1.9507932759398015e-05, + "loss": 1.1342, + "step": 2394 + }, + { + "epoch": 0.15, + "learning_rate": 1.9507718070375064e-05, + "loss": 1.1686, + "step": 2395 + }, + { + "epoch": 0.15, + "learning_rate": 1.9507503381352113e-05, + "loss": 1.0714, + "step": 2396 + }, + { + "epoch": 0.15, + "learning_rate": 1.9507288692329162e-05, + "loss": 1.1917, + "step": 2397 + }, + { + "epoch": 0.15, + "learning_rate": 1.950707400330621e-05, + "loss": 1.1198, + "step": 2398 + }, + { + "epoch": 0.15, + "learning_rate": 1.9506859314283263e-05, + "loss": 1.1876, + "step": 2399 + }, + { + "epoch": 0.15, + "learning_rate": 1.9506644625260312e-05, + "loss": 1.153, + "step": 2400 + }, + { + "epoch": 0.15, + "learning_rate": 1.950642993623736e-05, + "loss": 1.0982, + "step": 2401 + }, + { + "epoch": 0.15, + "learning_rate": 1.950621524721441e-05, + "loss": 1.1597, + "step": 2402 + }, + { + "epoch": 0.15, + "learning_rate": 1.950600055819146e-05, + "loss": 1.139, + "step": 2403 + }, + { + "epoch": 0.15, + "learning_rate": 1.950578586916851e-05, + "loss": 1.1684, + "step": 2404 + }, + { + "epoch": 0.15, + "learning_rate": 1.950557118014556e-05, + "loss": 1.1556, + "step": 2405 + }, + { + "epoch": 0.15, + "learning_rate": 1.9505356491122613e-05, + "loss": 1.18, + "step": 2406 + }, + { + "epoch": 0.15, + "learning_rate": 1.950514180209966e-05, + "loss": 1.1625, + "step": 2407 + }, + { + "epoch": 0.15, + "learning_rate": 1.950492711307671e-05, + "loss": 1.1979, + "step": 2408 + }, + { + "epoch": 0.15, + "learning_rate": 1.950471242405376e-05, + "loss": 1.1322, + "step": 2409 + }, + { + "epoch": 0.16, + "learning_rate": 1.950449773503081e-05, + "loss": 1.1359, + "step": 2410 + }, + { + "epoch": 0.16, + "learning_rate": 1.950428304600786e-05, + "loss": 1.136, + "step": 2411 + }, + { + "epoch": 0.16, + "learning_rate": 1.950406835698491e-05, + "loss": 1.1999, + "step": 2412 + }, + { + "epoch": 0.16, + "learning_rate": 1.950385366796196e-05, + "loss": 1.2135, + "step": 2413 + }, + { + "epoch": 0.16, + "learning_rate": 1.9503638978939008e-05, + "loss": 1.1229, + "step": 2414 + }, + { + "epoch": 0.16, + "learning_rate": 1.9503424289916057e-05, + "loss": 1.1857, + "step": 2415 + }, + { + "epoch": 0.16, + "learning_rate": 1.9503209600893106e-05, + "loss": 1.213, + "step": 2416 + }, + { + "epoch": 0.16, + "learning_rate": 1.9502994911870158e-05, + "loss": 1.1338, + "step": 2417 + }, + { + "epoch": 0.16, + "learning_rate": 1.9502780222847207e-05, + "loss": 1.2088, + "step": 2418 + }, + { + "epoch": 0.16, + "learning_rate": 1.950256553382426e-05, + "loss": 1.1383, + "step": 2419 + }, + { + "epoch": 0.16, + "learning_rate": 1.9502350844801308e-05, + "loss": 1.195, + "step": 2420 + }, + { + "epoch": 0.16, + "learning_rate": 1.9502136155778357e-05, + "loss": 1.1614, + "step": 2421 + }, + { + "epoch": 0.16, + "learning_rate": 1.9501921466755406e-05, + "loss": 1.1788, + "step": 2422 + }, + { + "epoch": 0.16, + "learning_rate": 1.9501706777732455e-05, + "loss": 1.1861, + "step": 2423 + }, + { + "epoch": 0.16, + "learning_rate": 1.9501492088709507e-05, + "loss": 1.0674, + "step": 2424 + }, + { + "epoch": 0.16, + "learning_rate": 1.9501277399686556e-05, + "loss": 1.1556, + "step": 2425 + }, + { + "epoch": 0.16, + "learning_rate": 1.9501062710663605e-05, + "loss": 1.1327, + "step": 2426 + }, + { + "epoch": 0.16, + "learning_rate": 1.9500848021640654e-05, + "loss": 1.0967, + "step": 2427 + }, + { + "epoch": 0.16, + "learning_rate": 1.9500633332617703e-05, + "loss": 1.1689, + "step": 2428 + }, + { + "epoch": 0.16, + "learning_rate": 1.9500418643594756e-05, + "loss": 1.1361, + "step": 2429 + }, + { + "epoch": 0.16, + "learning_rate": 1.9500203954571805e-05, + "loss": 1.2501, + "step": 2430 + }, + { + "epoch": 0.16, + "learning_rate": 1.9499989265548854e-05, + "loss": 1.2356, + "step": 2431 + }, + { + "epoch": 0.16, + "learning_rate": 1.9499774576525903e-05, + "loss": 1.2652, + "step": 2432 + }, + { + "epoch": 0.16, + "learning_rate": 1.9499559887502955e-05, + "loss": 1.0645, + "step": 2433 + }, + { + "epoch": 0.16, + "learning_rate": 1.9499345198480004e-05, + "loss": 1.179, + "step": 2434 + }, + { + "epoch": 0.16, + "learning_rate": 1.9499130509457053e-05, + "loss": 1.189, + "step": 2435 + }, + { + "epoch": 0.16, + "learning_rate": 1.9498915820434102e-05, + "loss": 1.2432, + "step": 2436 + }, + { + "epoch": 0.16, + "learning_rate": 1.9498701131411154e-05, + "loss": 1.1887, + "step": 2437 + }, + { + "epoch": 0.16, + "learning_rate": 1.9498486442388203e-05, + "loss": 1.0971, + "step": 2438 + }, + { + "epoch": 0.16, + "learning_rate": 1.9498271753365252e-05, + "loss": 1.1236, + "step": 2439 + }, + { + "epoch": 0.16, + "learning_rate": 1.94980570643423e-05, + "loss": 1.1332, + "step": 2440 + }, + { + "epoch": 0.16, + "learning_rate": 1.949784237531935e-05, + "loss": 1.1216, + "step": 2441 + }, + { + "epoch": 0.16, + "learning_rate": 1.9497627686296402e-05, + "loss": 1.1244, + "step": 2442 + }, + { + "epoch": 0.16, + "learning_rate": 1.949741299727345e-05, + "loss": 1.1185, + "step": 2443 + }, + { + "epoch": 0.16, + "learning_rate": 1.94971983082505e-05, + "loss": 1.2313, + "step": 2444 + }, + { + "epoch": 0.16, + "learning_rate": 1.949698361922755e-05, + "loss": 1.2277, + "step": 2445 + }, + { + "epoch": 0.16, + "learning_rate": 1.9496768930204598e-05, + "loss": 1.1691, + "step": 2446 + }, + { + "epoch": 0.16, + "learning_rate": 1.949655424118165e-05, + "loss": 1.1585, + "step": 2447 + }, + { + "epoch": 0.16, + "learning_rate": 1.94963395521587e-05, + "loss": 1.1845, + "step": 2448 + }, + { + "epoch": 0.16, + "learning_rate": 1.9496124863135752e-05, + "loss": 1.1576, + "step": 2449 + }, + { + "epoch": 0.16, + "learning_rate": 1.94959101741128e-05, + "loss": 1.2019, + "step": 2450 + }, + { + "epoch": 0.16, + "learning_rate": 1.949569548508985e-05, + "loss": 1.2077, + "step": 2451 + }, + { + "epoch": 0.16, + "learning_rate": 1.94954807960669e-05, + "loss": 1.1419, + "step": 2452 + }, + { + "epoch": 0.16, + "learning_rate": 1.9495266107043948e-05, + "loss": 1.0667, + "step": 2453 + }, + { + "epoch": 0.16, + "learning_rate": 1.9495051418020997e-05, + "loss": 1.1173, + "step": 2454 + }, + { + "epoch": 0.16, + "learning_rate": 1.949483672899805e-05, + "loss": 1.0788, + "step": 2455 + }, + { + "epoch": 0.16, + "learning_rate": 1.9494622039975098e-05, + "loss": 1.1669, + "step": 2456 + }, + { + "epoch": 0.16, + "learning_rate": 1.9494407350952147e-05, + "loss": 1.166, + "step": 2457 + }, + { + "epoch": 0.16, + "learning_rate": 1.9494192661929196e-05, + "loss": 1.1984, + "step": 2458 + }, + { + "epoch": 0.16, + "learning_rate": 1.9493977972906245e-05, + "loss": 1.1046, + "step": 2459 + }, + { + "epoch": 0.16, + "learning_rate": 1.9493763283883297e-05, + "loss": 1.138, + "step": 2460 + }, + { + "epoch": 0.16, + "learning_rate": 1.9493548594860346e-05, + "loss": 1.2091, + "step": 2461 + }, + { + "epoch": 0.16, + "learning_rate": 1.94933339058374e-05, + "loss": 1.1657, + "step": 2462 + }, + { + "epoch": 0.16, + "learning_rate": 1.9493119216814448e-05, + "loss": 1.2607, + "step": 2463 + }, + { + "epoch": 0.16, + "learning_rate": 1.9492904527791497e-05, + "loss": 1.2214, + "step": 2464 + }, + { + "epoch": 0.16, + "learning_rate": 1.9492689838768546e-05, + "loss": 1.0839, + "step": 2465 + }, + { + "epoch": 0.16, + "learning_rate": 1.9492475149745594e-05, + "loss": 1.1703, + "step": 2466 + }, + { + "epoch": 0.16, + "learning_rate": 1.9492260460722643e-05, + "loss": 1.1473, + "step": 2467 + }, + { + "epoch": 0.16, + "learning_rate": 1.9492045771699696e-05, + "loss": 1.2477, + "step": 2468 + }, + { + "epoch": 0.16, + "learning_rate": 1.9491831082676745e-05, + "loss": 1.1258, + "step": 2469 + }, + { + "epoch": 0.16, + "learning_rate": 1.9491616393653794e-05, + "loss": 1.2168, + "step": 2470 + }, + { + "epoch": 0.16, + "learning_rate": 1.9491401704630843e-05, + "loss": 1.1559, + "step": 2471 + }, + { + "epoch": 0.16, + "learning_rate": 1.949118701560789e-05, + "loss": 1.1827, + "step": 2472 + }, + { + "epoch": 0.16, + "learning_rate": 1.9490972326584944e-05, + "loss": 1.2202, + "step": 2473 + }, + { + "epoch": 0.16, + "learning_rate": 1.9490757637561993e-05, + "loss": 1.1487, + "step": 2474 + }, + { + "epoch": 0.16, + "learning_rate": 1.9490542948539042e-05, + "loss": 1.2134, + "step": 2475 + }, + { + "epoch": 0.16, + "learning_rate": 1.9490328259516094e-05, + "loss": 1.1219, + "step": 2476 + }, + { + "epoch": 0.16, + "learning_rate": 1.9490113570493143e-05, + "loss": 1.152, + "step": 2477 + }, + { + "epoch": 0.16, + "learning_rate": 1.9489898881470192e-05, + "loss": 1.1812, + "step": 2478 + }, + { + "epoch": 0.16, + "learning_rate": 1.948968419244724e-05, + "loss": 1.1722, + "step": 2479 + }, + { + "epoch": 0.16, + "learning_rate": 1.9489469503424294e-05, + "loss": 1.0457, + "step": 2480 + }, + { + "epoch": 0.16, + "learning_rate": 1.9489254814401343e-05, + "loss": 1.1716, + "step": 2481 + }, + { + "epoch": 0.16, + "learning_rate": 1.948904012537839e-05, + "loss": 1.2443, + "step": 2482 + }, + { + "epoch": 0.16, + "learning_rate": 1.948882543635544e-05, + "loss": 1.1555, + "step": 2483 + }, + { + "epoch": 0.16, + "learning_rate": 1.948861074733249e-05, + "loss": 1.2042, + "step": 2484 + }, + { + "epoch": 0.16, + "learning_rate": 1.948839605830954e-05, + "loss": 1.2015, + "step": 2485 + }, + { + "epoch": 0.16, + "learning_rate": 1.948818136928659e-05, + "loss": 1.1782, + "step": 2486 + }, + { + "epoch": 0.16, + "learning_rate": 1.948796668026364e-05, + "loss": 1.1403, + "step": 2487 + }, + { + "epoch": 0.16, + "learning_rate": 1.948775199124069e-05, + "loss": 1.2219, + "step": 2488 + }, + { + "epoch": 0.16, + "learning_rate": 1.9487537302217738e-05, + "loss": 1.1779, + "step": 2489 + }, + { + "epoch": 0.16, + "learning_rate": 1.9487322613194787e-05, + "loss": 1.0668, + "step": 2490 + }, + { + "epoch": 0.16, + "learning_rate": 1.948710792417184e-05, + "loss": 1.1533, + "step": 2491 + }, + { + "epoch": 0.16, + "learning_rate": 1.9486893235148888e-05, + "loss": 1.1298, + "step": 2492 + }, + { + "epoch": 0.16, + "learning_rate": 1.948667854612594e-05, + "loss": 1.1862, + "step": 2493 + }, + { + "epoch": 0.16, + "learning_rate": 1.948646385710299e-05, + "loss": 1.1161, + "step": 2494 + }, + { + "epoch": 0.16, + "learning_rate": 1.9486249168080038e-05, + "loss": 1.1855, + "step": 2495 + }, + { + "epoch": 0.16, + "learning_rate": 1.9486034479057087e-05, + "loss": 1.1577, + "step": 2496 + }, + { + "epoch": 0.16, + "learning_rate": 1.9485819790034136e-05, + "loss": 1.1639, + "step": 2497 + }, + { + "epoch": 0.16, + "learning_rate": 1.948560510101119e-05, + "loss": 1.1383, + "step": 2498 + }, + { + "epoch": 0.16, + "learning_rate": 1.9485390411988237e-05, + "loss": 1.0987, + "step": 2499 + }, + { + "epoch": 0.16, + "learning_rate": 1.9485175722965286e-05, + "loss": 1.1551, + "step": 2500 + }, + { + "epoch": 0.16, + "learning_rate": 1.9484961033942335e-05, + "loss": 1.1697, + "step": 2501 + }, + { + "epoch": 0.16, + "learning_rate": 1.9484746344919384e-05, + "loss": 1.2636, + "step": 2502 + }, + { + "epoch": 0.16, + "learning_rate": 1.9484531655896433e-05, + "loss": 1.1633, + "step": 2503 + }, + { + "epoch": 0.16, + "learning_rate": 1.9484316966873486e-05, + "loss": 1.1743, + "step": 2504 + }, + { + "epoch": 0.16, + "learning_rate": 1.9484102277850535e-05, + "loss": 1.1384, + "step": 2505 + }, + { + "epoch": 0.16, + "learning_rate": 1.9483887588827587e-05, + "loss": 1.2291, + "step": 2506 + }, + { + "epoch": 0.16, + "learning_rate": 1.9483672899804636e-05, + "loss": 1.2754, + "step": 2507 + }, + { + "epoch": 0.16, + "learning_rate": 1.9483458210781685e-05, + "loss": 1.0748, + "step": 2508 + }, + { + "epoch": 0.16, + "learning_rate": 1.9483243521758734e-05, + "loss": 1.1332, + "step": 2509 + }, + { + "epoch": 0.16, + "learning_rate": 1.9483028832735783e-05, + "loss": 1.166, + "step": 2510 + }, + { + "epoch": 0.16, + "learning_rate": 1.9482814143712835e-05, + "loss": 1.1102, + "step": 2511 + }, + { + "epoch": 0.16, + "learning_rate": 1.9482599454689884e-05, + "loss": 1.2127, + "step": 2512 + }, + { + "epoch": 0.16, + "learning_rate": 1.9482384765666933e-05, + "loss": 1.1524, + "step": 2513 + }, + { + "epoch": 0.16, + "learning_rate": 1.9482170076643982e-05, + "loss": 1.1124, + "step": 2514 + }, + { + "epoch": 0.16, + "learning_rate": 1.948195538762103e-05, + "loss": 1.1367, + "step": 2515 + }, + { + "epoch": 0.16, + "learning_rate": 1.948174069859808e-05, + "loss": 1.1278, + "step": 2516 + }, + { + "epoch": 0.16, + "learning_rate": 1.9481526009575132e-05, + "loss": 1.2711, + "step": 2517 + }, + { + "epoch": 0.16, + "learning_rate": 1.948131132055218e-05, + "loss": 1.1216, + "step": 2518 + }, + { + "epoch": 0.16, + "learning_rate": 1.948109663152923e-05, + "loss": 1.1825, + "step": 2519 + }, + { + "epoch": 0.16, + "learning_rate": 1.9480881942506283e-05, + "loss": 1.107, + "step": 2520 + }, + { + "epoch": 0.16, + "learning_rate": 1.948066725348333e-05, + "loss": 1.0669, + "step": 2521 + }, + { + "epoch": 0.16, + "learning_rate": 1.948045256446038e-05, + "loss": 1.2528, + "step": 2522 + }, + { + "epoch": 0.16, + "learning_rate": 1.948023787543743e-05, + "loss": 1.2006, + "step": 2523 + }, + { + "epoch": 0.16, + "learning_rate": 1.9480023186414482e-05, + "loss": 1.2279, + "step": 2524 + }, + { + "epoch": 0.16, + "learning_rate": 1.947980849739153e-05, + "loss": 1.1551, + "step": 2525 + }, + { + "epoch": 0.16, + "learning_rate": 1.947959380836858e-05, + "loss": 1.2022, + "step": 2526 + }, + { + "epoch": 0.16, + "learning_rate": 1.947937911934563e-05, + "loss": 1.2189, + "step": 2527 + }, + { + "epoch": 0.16, + "learning_rate": 1.9479164430322678e-05, + "loss": 1.21, + "step": 2528 + }, + { + "epoch": 0.16, + "learning_rate": 1.947894974129973e-05, + "loss": 1.215, + "step": 2529 + }, + { + "epoch": 0.16, + "learning_rate": 1.947873505227678e-05, + "loss": 1.0137, + "step": 2530 + }, + { + "epoch": 0.16, + "learning_rate": 1.9478520363253828e-05, + "loss": 1.1423, + "step": 2531 + }, + { + "epoch": 0.16, + "learning_rate": 1.9478305674230877e-05, + "loss": 1.1563, + "step": 2532 + }, + { + "epoch": 0.16, + "learning_rate": 1.9478090985207926e-05, + "loss": 1.1702, + "step": 2533 + }, + { + "epoch": 0.16, + "learning_rate": 1.9477876296184975e-05, + "loss": 1.1313, + "step": 2534 + }, + { + "epoch": 0.16, + "learning_rate": 1.9477661607162027e-05, + "loss": 1.2146, + "step": 2535 + }, + { + "epoch": 0.16, + "learning_rate": 1.947744691813908e-05, + "loss": 1.1611, + "step": 2536 + }, + { + "epoch": 0.16, + "learning_rate": 1.947723222911613e-05, + "loss": 1.1383, + "step": 2537 + }, + { + "epoch": 0.16, + "learning_rate": 1.9477017540093178e-05, + "loss": 1.072, + "step": 2538 + }, + { + "epoch": 0.16, + "learning_rate": 1.9476802851070227e-05, + "loss": 1.1922, + "step": 2539 + }, + { + "epoch": 0.16, + "learning_rate": 1.9476588162047276e-05, + "loss": 1.1762, + "step": 2540 + }, + { + "epoch": 0.16, + "learning_rate": 1.9476373473024324e-05, + "loss": 1.21, + "step": 2541 + }, + { + "epoch": 0.16, + "learning_rate": 1.9476158784001377e-05, + "loss": 1.1834, + "step": 2542 + }, + { + "epoch": 0.16, + "learning_rate": 1.9475944094978426e-05, + "loss": 1.1246, + "step": 2543 + }, + { + "epoch": 0.16, + "learning_rate": 1.9475729405955475e-05, + "loss": 1.1436, + "step": 2544 + }, + { + "epoch": 0.16, + "learning_rate": 1.9475514716932524e-05, + "loss": 1.1191, + "step": 2545 + }, + { + "epoch": 0.16, + "learning_rate": 1.9475300027909573e-05, + "loss": 1.1461, + "step": 2546 + }, + { + "epoch": 0.16, + "learning_rate": 1.9475085338886625e-05, + "loss": 1.102, + "step": 2547 + }, + { + "epoch": 0.16, + "learning_rate": 1.9474870649863674e-05, + "loss": 1.114, + "step": 2548 + }, + { + "epoch": 0.16, + "learning_rate": 1.9474655960840726e-05, + "loss": 1.1721, + "step": 2549 + }, + { + "epoch": 0.16, + "learning_rate": 1.9474441271817775e-05, + "loss": 1.1281, + "step": 2550 + }, + { + "epoch": 0.16, + "learning_rate": 1.9474226582794824e-05, + "loss": 1.1631, + "step": 2551 + }, + { + "epoch": 0.16, + "learning_rate": 1.9474011893771873e-05, + "loss": 1.1889, + "step": 2552 + }, + { + "epoch": 0.16, + "learning_rate": 1.9473797204748922e-05, + "loss": 1.1473, + "step": 2553 + }, + { + "epoch": 0.16, + "learning_rate": 1.947358251572597e-05, + "loss": 1.2144, + "step": 2554 + }, + { + "epoch": 0.16, + "learning_rate": 1.9473367826703024e-05, + "loss": 1.1197, + "step": 2555 + }, + { + "epoch": 0.16, + "learning_rate": 1.9473153137680072e-05, + "loss": 1.1265, + "step": 2556 + }, + { + "epoch": 0.16, + "learning_rate": 1.947293844865712e-05, + "loss": 1.1418, + "step": 2557 + }, + { + "epoch": 0.16, + "learning_rate": 1.947272375963417e-05, + "loss": 1.2675, + "step": 2558 + }, + { + "epoch": 0.16, + "learning_rate": 1.947250907061122e-05, + "loss": 1.2117, + "step": 2559 + }, + { + "epoch": 0.16, + "learning_rate": 1.9472294381588272e-05, + "loss": 1.1501, + "step": 2560 + }, + { + "epoch": 0.16, + "learning_rate": 1.947207969256532e-05, + "loss": 1.1978, + "step": 2561 + }, + { + "epoch": 0.16, + "learning_rate": 1.947186500354237e-05, + "loss": 1.1862, + "step": 2562 + }, + { + "epoch": 0.16, + "learning_rate": 1.947165031451942e-05, + "loss": 1.1189, + "step": 2563 + }, + { + "epoch": 0.16, + "learning_rate": 1.947143562549647e-05, + "loss": 1.1204, + "step": 2564 + }, + { + "epoch": 0.17, + "learning_rate": 1.947122093647352e-05, + "loss": 1.2266, + "step": 2565 + }, + { + "epoch": 0.17, + "learning_rate": 1.947100624745057e-05, + "loss": 1.0661, + "step": 2566 + }, + { + "epoch": 0.17, + "learning_rate": 1.947079155842762e-05, + "loss": 1.1875, + "step": 2567 + }, + { + "epoch": 0.17, + "learning_rate": 1.947057686940467e-05, + "loss": 1.2036, + "step": 2568 + }, + { + "epoch": 0.17, + "learning_rate": 1.947036218038172e-05, + "loss": 1.1257, + "step": 2569 + }, + { + "epoch": 0.17, + "learning_rate": 1.9470147491358768e-05, + "loss": 1.1823, + "step": 2570 + }, + { + "epoch": 0.17, + "learning_rate": 1.9469932802335817e-05, + "loss": 1.0854, + "step": 2571 + }, + { + "epoch": 0.17, + "learning_rate": 1.9469718113312866e-05, + "loss": 1.242, + "step": 2572 + }, + { + "epoch": 0.17, + "learning_rate": 1.946950342428992e-05, + "loss": 1.1646, + "step": 2573 + }, + { + "epoch": 0.17, + "learning_rate": 1.9469288735266967e-05, + "loss": 1.1591, + "step": 2574 + }, + { + "epoch": 0.17, + "learning_rate": 1.9469074046244016e-05, + "loss": 1.1584, + "step": 2575 + }, + { + "epoch": 0.17, + "learning_rate": 1.9468859357221065e-05, + "loss": 1.1602, + "step": 2576 + }, + { + "epoch": 0.17, + "learning_rate": 1.9468644668198114e-05, + "loss": 1.1509, + "step": 2577 + }, + { + "epoch": 0.17, + "learning_rate": 1.9468429979175167e-05, + "loss": 1.2701, + "step": 2578 + }, + { + "epoch": 0.17, + "learning_rate": 1.9468215290152216e-05, + "loss": 1.1787, + "step": 2579 + }, + { + "epoch": 0.17, + "learning_rate": 1.9468000601129268e-05, + "loss": 1.2342, + "step": 2580 + }, + { + "epoch": 0.17, + "learning_rate": 1.9467785912106317e-05, + "loss": 1.0764, + "step": 2581 + }, + { + "epoch": 0.17, + "learning_rate": 1.9467571223083366e-05, + "loss": 1.1948, + "step": 2582 + }, + { + "epoch": 0.17, + "learning_rate": 1.9467356534060415e-05, + "loss": 1.1166, + "step": 2583 + }, + { + "epoch": 0.17, + "learning_rate": 1.9467141845037464e-05, + "loss": 1.1994, + "step": 2584 + }, + { + "epoch": 0.17, + "learning_rate": 1.9466927156014516e-05, + "loss": 1.228, + "step": 2585 + }, + { + "epoch": 0.17, + "learning_rate": 1.9466712466991565e-05, + "loss": 1.199, + "step": 2586 + }, + { + "epoch": 0.17, + "learning_rate": 1.9466497777968614e-05, + "loss": 1.14, + "step": 2587 + }, + { + "epoch": 0.17, + "learning_rate": 1.9466283088945663e-05, + "loss": 1.1111, + "step": 2588 + }, + { + "epoch": 0.17, + "learning_rate": 1.9466068399922712e-05, + "loss": 1.1454, + "step": 2589 + }, + { + "epoch": 0.17, + "learning_rate": 1.946585371089976e-05, + "loss": 1.1403, + "step": 2590 + }, + { + "epoch": 0.17, + "learning_rate": 1.9465639021876813e-05, + "loss": 1.2539, + "step": 2591 + }, + { + "epoch": 0.17, + "learning_rate": 1.9465424332853862e-05, + "loss": 1.1456, + "step": 2592 + }, + { + "epoch": 0.17, + "learning_rate": 1.9465209643830915e-05, + "loss": 1.1102, + "step": 2593 + }, + { + "epoch": 0.17, + "learning_rate": 1.9464994954807964e-05, + "loss": 1.1915, + "step": 2594 + }, + { + "epoch": 0.17, + "learning_rate": 1.9464780265785013e-05, + "loss": 1.18, + "step": 2595 + }, + { + "epoch": 0.17, + "learning_rate": 1.946456557676206e-05, + "loss": 1.1812, + "step": 2596 + }, + { + "epoch": 0.17, + "learning_rate": 1.946435088773911e-05, + "loss": 1.1068, + "step": 2597 + }, + { + "epoch": 0.17, + "learning_rate": 1.9464136198716163e-05, + "loss": 1.1675, + "step": 2598 + }, + { + "epoch": 0.17, + "learning_rate": 1.9463921509693212e-05, + "loss": 1.05, + "step": 2599 + }, + { + "epoch": 0.17, + "learning_rate": 1.946370682067026e-05, + "loss": 1.1299, + "step": 2600 + }, + { + "epoch": 0.17, + "learning_rate": 1.946349213164731e-05, + "loss": 1.2623, + "step": 2601 + }, + { + "epoch": 0.17, + "learning_rate": 1.946327744262436e-05, + "loss": 1.1352, + "step": 2602 + }, + { + "epoch": 0.17, + "learning_rate": 1.9463062753601408e-05, + "loss": 1.086, + "step": 2603 + }, + { + "epoch": 0.17, + "learning_rate": 1.946284806457846e-05, + "loss": 1.2488, + "step": 2604 + }, + { + "epoch": 0.17, + "learning_rate": 1.946263337555551e-05, + "loss": 1.128, + "step": 2605 + }, + { + "epoch": 0.17, + "learning_rate": 1.9462418686532558e-05, + "loss": 1.1622, + "step": 2606 + }, + { + "epoch": 0.17, + "learning_rate": 1.946220399750961e-05, + "loss": 1.1161, + "step": 2607 + }, + { + "epoch": 0.17, + "learning_rate": 1.946198930848666e-05, + "loss": 1.1418, + "step": 2608 + }, + { + "epoch": 0.17, + "learning_rate": 1.9461774619463708e-05, + "loss": 1.2166, + "step": 2609 + }, + { + "epoch": 0.17, + "learning_rate": 1.9461559930440757e-05, + "loss": 1.116, + "step": 2610 + }, + { + "epoch": 0.17, + "learning_rate": 1.946134524141781e-05, + "loss": 1.1074, + "step": 2611 + }, + { + "epoch": 0.17, + "learning_rate": 1.946113055239486e-05, + "loss": 1.1686, + "step": 2612 + }, + { + "epoch": 0.17, + "learning_rate": 1.9460915863371908e-05, + "loss": 1.1384, + "step": 2613 + }, + { + "epoch": 0.17, + "learning_rate": 1.9460701174348957e-05, + "loss": 1.1033, + "step": 2614 + }, + { + "epoch": 0.17, + "learning_rate": 1.9460486485326005e-05, + "loss": 1.246, + "step": 2615 + }, + { + "epoch": 0.17, + "learning_rate": 1.9460271796303058e-05, + "loss": 1.1299, + "step": 2616 + }, + { + "epoch": 0.17, + "learning_rate": 1.9460057107280107e-05, + "loss": 1.1283, + "step": 2617 + }, + { + "epoch": 0.17, + "learning_rate": 1.9459842418257156e-05, + "loss": 1.1255, + "step": 2618 + }, + { + "epoch": 0.17, + "learning_rate": 1.9459627729234205e-05, + "loss": 1.1494, + "step": 2619 + }, + { + "epoch": 0.17, + "learning_rate": 1.9459413040211254e-05, + "loss": 1.1789, + "step": 2620 + }, + { + "epoch": 0.17, + "learning_rate": 1.9459198351188303e-05, + "loss": 1.2385, + "step": 2621 + }, + { + "epoch": 0.17, + "learning_rate": 1.9458983662165355e-05, + "loss": 1.1742, + "step": 2622 + }, + { + "epoch": 0.17, + "learning_rate": 1.9458768973142404e-05, + "loss": 1.2057, + "step": 2623 + }, + { + "epoch": 0.17, + "learning_rate": 1.9458554284119456e-05, + "loss": 1.083, + "step": 2624 + }, + { + "epoch": 0.17, + "learning_rate": 1.9458339595096505e-05, + "loss": 1.0839, + "step": 2625 + }, + { + "epoch": 0.17, + "learning_rate": 1.9458124906073554e-05, + "loss": 1.1102, + "step": 2626 + }, + { + "epoch": 0.17, + "learning_rate": 1.9457910217050603e-05, + "loss": 1.1487, + "step": 2627 + }, + { + "epoch": 0.17, + "learning_rate": 1.9457695528027652e-05, + "loss": 1.1397, + "step": 2628 + }, + { + "epoch": 0.17, + "learning_rate": 1.9457480839004705e-05, + "loss": 1.0584, + "step": 2629 + }, + { + "epoch": 0.17, + "learning_rate": 1.9457266149981753e-05, + "loss": 1.1347, + "step": 2630 + }, + { + "epoch": 0.17, + "learning_rate": 1.9457051460958802e-05, + "loss": 1.0725, + "step": 2631 + }, + { + "epoch": 0.17, + "learning_rate": 1.945683677193585e-05, + "loss": 1.1249, + "step": 2632 + }, + { + "epoch": 0.17, + "learning_rate": 1.94566220829129e-05, + "loss": 1.0496, + "step": 2633 + }, + { + "epoch": 0.17, + "learning_rate": 1.9456407393889953e-05, + "loss": 1.2004, + "step": 2634 + }, + { + "epoch": 0.17, + "learning_rate": 1.9456192704867002e-05, + "loss": 1.179, + "step": 2635 + }, + { + "epoch": 0.17, + "learning_rate": 1.9455978015844054e-05, + "loss": 1.1845, + "step": 2636 + }, + { + "epoch": 0.17, + "learning_rate": 1.9455763326821103e-05, + "loss": 1.1198, + "step": 2637 + }, + { + "epoch": 0.17, + "learning_rate": 1.9455548637798152e-05, + "loss": 1.1038, + "step": 2638 + }, + { + "epoch": 0.17, + "learning_rate": 1.94553339487752e-05, + "loss": 1.1338, + "step": 2639 + }, + { + "epoch": 0.17, + "learning_rate": 1.945511925975225e-05, + "loss": 1.1514, + "step": 2640 + }, + { + "epoch": 0.17, + "learning_rate": 1.94549045707293e-05, + "loss": 1.1019, + "step": 2641 + }, + { + "epoch": 0.17, + "learning_rate": 1.945468988170635e-05, + "loss": 1.1485, + "step": 2642 + }, + { + "epoch": 0.17, + "learning_rate": 1.94544751926834e-05, + "loss": 1.2102, + "step": 2643 + }, + { + "epoch": 0.17, + "learning_rate": 1.945426050366045e-05, + "loss": 1.1696, + "step": 2644 + }, + { + "epoch": 0.17, + "learning_rate": 1.9454045814637498e-05, + "loss": 1.0962, + "step": 2645 + }, + { + "epoch": 0.17, + "learning_rate": 1.9453831125614547e-05, + "loss": 1.2259, + "step": 2646 + }, + { + "epoch": 0.17, + "learning_rate": 1.94536164365916e-05, + "loss": 1.2271, + "step": 2647 + }, + { + "epoch": 0.17, + "learning_rate": 1.945340174756865e-05, + "loss": 1.1917, + "step": 2648 + }, + { + "epoch": 0.17, + "learning_rate": 1.9453187058545697e-05, + "loss": 1.2309, + "step": 2649 + }, + { + "epoch": 0.17, + "learning_rate": 1.9452972369522746e-05, + "loss": 1.1994, + "step": 2650 + }, + { + "epoch": 0.17, + "learning_rate": 1.94527576804998e-05, + "loss": 1.1888, + "step": 2651 + }, + { + "epoch": 0.17, + "learning_rate": 1.9452542991476848e-05, + "loss": 1.2046, + "step": 2652 + }, + { + "epoch": 0.17, + "learning_rate": 1.9452328302453897e-05, + "loss": 1.1569, + "step": 2653 + }, + { + "epoch": 0.17, + "learning_rate": 1.945211361343095e-05, + "loss": 1.105, + "step": 2654 + }, + { + "epoch": 0.17, + "learning_rate": 1.9451898924407998e-05, + "loss": 1.0528, + "step": 2655 + }, + { + "epoch": 0.17, + "learning_rate": 1.9451684235385047e-05, + "loss": 1.0742, + "step": 2656 + }, + { + "epoch": 0.17, + "learning_rate": 1.9451469546362096e-05, + "loss": 1.1855, + "step": 2657 + }, + { + "epoch": 0.17, + "learning_rate": 1.9451254857339145e-05, + "loss": 1.226, + "step": 2658 + }, + { + "epoch": 0.17, + "learning_rate": 1.9451040168316194e-05, + "loss": 1.1687, + "step": 2659 + }, + { + "epoch": 0.17, + "learning_rate": 1.9450825479293246e-05, + "loss": 1.1485, + "step": 2660 + }, + { + "epoch": 0.17, + "learning_rate": 1.9450610790270295e-05, + "loss": 1.0886, + "step": 2661 + }, + { + "epoch": 0.17, + "learning_rate": 1.9450396101247344e-05, + "loss": 1.1341, + "step": 2662 + }, + { + "epoch": 0.17, + "learning_rate": 1.9450181412224393e-05, + "loss": 1.1848, + "step": 2663 + }, + { + "epoch": 0.17, + "learning_rate": 1.9449966723201442e-05, + "loss": 1.0526, + "step": 2664 + }, + { + "epoch": 0.17, + "learning_rate": 1.9449752034178494e-05, + "loss": 1.1386, + "step": 2665 + }, + { + "epoch": 0.17, + "learning_rate": 1.9449537345155543e-05, + "loss": 1.08, + "step": 2666 + }, + { + "epoch": 0.17, + "learning_rate": 1.9449322656132596e-05, + "loss": 1.1395, + "step": 2667 + }, + { + "epoch": 0.17, + "learning_rate": 1.9449107967109645e-05, + "loss": 1.18, + "step": 2668 + }, + { + "epoch": 0.17, + "learning_rate": 1.9448893278086694e-05, + "loss": 1.231, + "step": 2669 + }, + { + "epoch": 0.17, + "learning_rate": 1.9448678589063743e-05, + "loss": 1.0907, + "step": 2670 + }, + { + "epoch": 0.17, + "learning_rate": 1.944846390004079e-05, + "loss": 1.2103, + "step": 2671 + }, + { + "epoch": 0.17, + "learning_rate": 1.9448249211017844e-05, + "loss": 1.2903, + "step": 2672 + }, + { + "epoch": 0.17, + "learning_rate": 1.9448034521994893e-05, + "loss": 1.1098, + "step": 2673 + }, + { + "epoch": 0.17, + "learning_rate": 1.9447819832971942e-05, + "loss": 1.1088, + "step": 2674 + }, + { + "epoch": 0.17, + "learning_rate": 1.944760514394899e-05, + "loss": 1.3082, + "step": 2675 + }, + { + "epoch": 0.17, + "learning_rate": 1.944739045492604e-05, + "loss": 1.1152, + "step": 2676 + }, + { + "epoch": 0.17, + "learning_rate": 1.944717576590309e-05, + "loss": 1.1373, + "step": 2677 + }, + { + "epoch": 0.17, + "learning_rate": 1.944696107688014e-05, + "loss": 1.1753, + "step": 2678 + }, + { + "epoch": 0.17, + "learning_rate": 1.944674638785719e-05, + "loss": 1.1188, + "step": 2679 + }, + { + "epoch": 0.17, + "learning_rate": 1.9446531698834242e-05, + "loss": 1.1741, + "step": 2680 + }, + { + "epoch": 0.17, + "learning_rate": 1.944631700981129e-05, + "loss": 1.218, + "step": 2681 + }, + { + "epoch": 0.17, + "learning_rate": 1.944610232078834e-05, + "loss": 1.0656, + "step": 2682 + }, + { + "epoch": 0.17, + "learning_rate": 1.944588763176539e-05, + "loss": 1.176, + "step": 2683 + }, + { + "epoch": 0.17, + "learning_rate": 1.9445672942742438e-05, + "loss": 1.0911, + "step": 2684 + }, + { + "epoch": 0.17, + "learning_rate": 1.944545825371949e-05, + "loss": 1.0672, + "step": 2685 + }, + { + "epoch": 0.17, + "learning_rate": 1.944524356469654e-05, + "loss": 1.1793, + "step": 2686 + }, + { + "epoch": 0.17, + "learning_rate": 1.944502887567359e-05, + "loss": 1.1151, + "step": 2687 + }, + { + "epoch": 0.17, + "learning_rate": 1.9444814186650638e-05, + "loss": 1.0774, + "step": 2688 + }, + { + "epoch": 0.17, + "learning_rate": 1.9444599497627686e-05, + "loss": 1.1265, + "step": 2689 + }, + { + "epoch": 0.17, + "learning_rate": 1.9444384808604735e-05, + "loss": 1.1347, + "step": 2690 + }, + { + "epoch": 0.17, + "learning_rate": 1.9444170119581788e-05, + "loss": 1.1219, + "step": 2691 + }, + { + "epoch": 0.17, + "learning_rate": 1.9443955430558837e-05, + "loss": 1.0789, + "step": 2692 + }, + { + "epoch": 0.17, + "learning_rate": 1.9443740741535886e-05, + "loss": 1.1379, + "step": 2693 + }, + { + "epoch": 0.17, + "learning_rate": 1.9443526052512938e-05, + "loss": 1.1191, + "step": 2694 + }, + { + "epoch": 0.17, + "learning_rate": 1.9443311363489987e-05, + "loss": 1.1127, + "step": 2695 + }, + { + "epoch": 0.17, + "learning_rate": 1.9443096674467036e-05, + "loss": 1.0758, + "step": 2696 + }, + { + "epoch": 0.17, + "learning_rate": 1.9442881985444085e-05, + "loss": 1.1843, + "step": 2697 + }, + { + "epoch": 0.17, + "learning_rate": 1.9442667296421137e-05, + "loss": 1.0609, + "step": 2698 + }, + { + "epoch": 0.17, + "learning_rate": 1.9442452607398186e-05, + "loss": 1.084, + "step": 2699 + }, + { + "epoch": 0.17, + "learning_rate": 1.9442237918375235e-05, + "loss": 1.1101, + "step": 2700 + }, + { + "epoch": 0.17, + "learning_rate": 1.9442023229352284e-05, + "loss": 1.1177, + "step": 2701 + }, + { + "epoch": 0.17, + "learning_rate": 1.9441808540329333e-05, + "loss": 1.0929, + "step": 2702 + }, + { + "epoch": 0.17, + "learning_rate": 1.9441593851306386e-05, + "loss": 1.2011, + "step": 2703 + }, + { + "epoch": 0.17, + "learning_rate": 1.9441379162283434e-05, + "loss": 1.0479, + "step": 2704 + }, + { + "epoch": 0.17, + "learning_rate": 1.9441164473260483e-05, + "loss": 1.0834, + "step": 2705 + }, + { + "epoch": 0.17, + "learning_rate": 1.9440949784237532e-05, + "loss": 1.1348, + "step": 2706 + }, + { + "epoch": 0.17, + "learning_rate": 1.944073509521458e-05, + "loss": 1.11, + "step": 2707 + }, + { + "epoch": 0.17, + "learning_rate": 1.944052040619163e-05, + "loss": 1.1703, + "step": 2708 + }, + { + "epoch": 0.17, + "learning_rate": 1.9440305717168683e-05, + "loss": 1.1722, + "step": 2709 + }, + { + "epoch": 0.17, + "learning_rate": 1.944009102814573e-05, + "loss": 1.2069, + "step": 2710 + }, + { + "epoch": 0.17, + "learning_rate": 1.9439876339122784e-05, + "loss": 1.1618, + "step": 2711 + }, + { + "epoch": 0.17, + "learning_rate": 1.9439661650099833e-05, + "loss": 1.1641, + "step": 2712 + }, + { + "epoch": 0.17, + "learning_rate": 1.9439446961076882e-05, + "loss": 1.2324, + "step": 2713 + }, + { + "epoch": 0.17, + "learning_rate": 1.943923227205393e-05, + "loss": 1.1153, + "step": 2714 + }, + { + "epoch": 0.17, + "learning_rate": 1.943901758303098e-05, + "loss": 1.1849, + "step": 2715 + }, + { + "epoch": 0.17, + "learning_rate": 1.9438802894008032e-05, + "loss": 1.1701, + "step": 2716 + }, + { + "epoch": 0.17, + "learning_rate": 1.943858820498508e-05, + "loss": 1.1722, + "step": 2717 + }, + { + "epoch": 0.17, + "learning_rate": 1.943837351596213e-05, + "loss": 1.193, + "step": 2718 + }, + { + "epoch": 0.17, + "learning_rate": 1.943815882693918e-05, + "loss": 1.2758, + "step": 2719 + }, + { + "epoch": 0.17, + "learning_rate": 1.9437944137916228e-05, + "loss": 1.1043, + "step": 2720 + }, + { + "epoch": 0.18, + "learning_rate": 1.943772944889328e-05, + "loss": 1.1839, + "step": 2721 + }, + { + "epoch": 0.18, + "learning_rate": 1.943751475987033e-05, + "loss": 1.1603, + "step": 2722 + }, + { + "epoch": 0.18, + "learning_rate": 1.9437300070847382e-05, + "loss": 1.145, + "step": 2723 + }, + { + "epoch": 0.18, + "learning_rate": 1.943708538182443e-05, + "loss": 1.1491, + "step": 2724 + }, + { + "epoch": 0.18, + "learning_rate": 1.943687069280148e-05, + "loss": 1.1468, + "step": 2725 + }, + { + "epoch": 0.18, + "learning_rate": 1.943665600377853e-05, + "loss": 1.1364, + "step": 2726 + }, + { + "epoch": 0.18, + "learning_rate": 1.9436441314755578e-05, + "loss": 1.2003, + "step": 2727 + }, + { + "epoch": 0.18, + "learning_rate": 1.9436226625732627e-05, + "loss": 1.1836, + "step": 2728 + }, + { + "epoch": 0.18, + "learning_rate": 1.943601193670968e-05, + "loss": 1.2085, + "step": 2729 + }, + { + "epoch": 0.18, + "learning_rate": 1.9435797247686728e-05, + "loss": 1.0945, + "step": 2730 + }, + { + "epoch": 0.18, + "learning_rate": 1.9435582558663777e-05, + "loss": 1.0748, + "step": 2731 + }, + { + "epoch": 0.18, + "learning_rate": 1.9435367869640826e-05, + "loss": 1.2276, + "step": 2732 + }, + { + "epoch": 0.18, + "learning_rate": 1.9435153180617875e-05, + "loss": 1.127, + "step": 2733 + }, + { + "epoch": 0.18, + "learning_rate": 1.9434938491594927e-05, + "loss": 1.1325, + "step": 2734 + }, + { + "epoch": 0.18, + "learning_rate": 1.9434723802571976e-05, + "loss": 1.1699, + "step": 2735 + }, + { + "epoch": 0.18, + "learning_rate": 1.9434509113549025e-05, + "loss": 1.1685, + "step": 2736 + }, + { + "epoch": 0.18, + "learning_rate": 1.9434294424526074e-05, + "loss": 1.2225, + "step": 2737 + }, + { + "epoch": 0.18, + "learning_rate": 1.9434079735503126e-05, + "loss": 1.1106, + "step": 2738 + }, + { + "epoch": 0.18, + "learning_rate": 1.9433865046480175e-05, + "loss": 1.1381, + "step": 2739 + }, + { + "epoch": 0.18, + "learning_rate": 1.9433650357457224e-05, + "loss": 1.1179, + "step": 2740 + }, + { + "epoch": 0.18, + "learning_rate": 1.9433435668434277e-05, + "loss": 1.1657, + "step": 2741 + }, + { + "epoch": 0.18, + "learning_rate": 1.9433220979411326e-05, + "loss": 1.1194, + "step": 2742 + }, + { + "epoch": 0.18, + "learning_rate": 1.9433006290388375e-05, + "loss": 1.0666, + "step": 2743 + }, + { + "epoch": 0.18, + "learning_rate": 1.9432791601365424e-05, + "loss": 1.0376, + "step": 2744 + }, + { + "epoch": 0.18, + "learning_rate": 1.9432576912342473e-05, + "loss": 1.0148, + "step": 2745 + }, + { + "epoch": 0.18, + "learning_rate": 1.943236222331952e-05, + "loss": 1.1174, + "step": 2746 + }, + { + "epoch": 0.18, + "learning_rate": 1.9432147534296574e-05, + "loss": 1.2896, + "step": 2747 + }, + { + "epoch": 0.18, + "learning_rate": 1.9431932845273623e-05, + "loss": 1.0992, + "step": 2748 + }, + { + "epoch": 0.18, + "learning_rate": 1.9431718156250672e-05, + "loss": 1.1508, + "step": 2749 + }, + { + "epoch": 0.18, + "learning_rate": 1.943150346722772e-05, + "loss": 1.1585, + "step": 2750 + }, + { + "epoch": 0.18, + "learning_rate": 1.943128877820477e-05, + "loss": 1.1594, + "step": 2751 + }, + { + "epoch": 0.18, + "learning_rate": 1.9431074089181822e-05, + "loss": 1.1704, + "step": 2752 + }, + { + "epoch": 0.18, + "learning_rate": 1.943085940015887e-05, + "loss": 1.2009, + "step": 2753 + }, + { + "epoch": 0.18, + "learning_rate": 1.9430644711135923e-05, + "loss": 1.1206, + "step": 2754 + }, + { + "epoch": 0.18, + "learning_rate": 1.9430430022112972e-05, + "loss": 1.1078, + "step": 2755 + }, + { + "epoch": 0.18, + "learning_rate": 1.943021533309002e-05, + "loss": 1.1165, + "step": 2756 + }, + { + "epoch": 0.18, + "learning_rate": 1.943000064406707e-05, + "loss": 1.1578, + "step": 2757 + }, + { + "epoch": 0.18, + "learning_rate": 1.942978595504412e-05, + "loss": 1.0566, + "step": 2758 + }, + { + "epoch": 0.18, + "learning_rate": 1.942957126602117e-05, + "loss": 1.1225, + "step": 2759 + }, + { + "epoch": 0.18, + "learning_rate": 1.942935657699822e-05, + "loss": 1.1472, + "step": 2760 + }, + { + "epoch": 0.18, + "learning_rate": 1.942914188797527e-05, + "loss": 1.1309, + "step": 2761 + }, + { + "epoch": 0.18, + "learning_rate": 1.942892719895232e-05, + "loss": 1.1206, + "step": 2762 + }, + { + "epoch": 0.18, + "learning_rate": 1.9428712509929367e-05, + "loss": 1.1202, + "step": 2763 + }, + { + "epoch": 0.18, + "learning_rate": 1.9428497820906416e-05, + "loss": 1.1604, + "step": 2764 + }, + { + "epoch": 0.18, + "learning_rate": 1.942828313188347e-05, + "loss": 1.1445, + "step": 2765 + }, + { + "epoch": 0.18, + "learning_rate": 1.9428068442860518e-05, + "loss": 1.0566, + "step": 2766 + }, + { + "epoch": 0.18, + "learning_rate": 1.942785375383757e-05, + "loss": 1.1984, + "step": 2767 + }, + { + "epoch": 0.18, + "learning_rate": 1.942763906481462e-05, + "loss": 1.0795, + "step": 2768 + }, + { + "epoch": 0.18, + "learning_rate": 1.9427424375791668e-05, + "loss": 1.0741, + "step": 2769 + }, + { + "epoch": 0.18, + "learning_rate": 1.9427209686768717e-05, + "loss": 1.1897, + "step": 2770 + }, + { + "epoch": 0.18, + "learning_rate": 1.9426994997745766e-05, + "loss": 1.1785, + "step": 2771 + }, + { + "epoch": 0.18, + "learning_rate": 1.942678030872282e-05, + "loss": 1.0842, + "step": 2772 + }, + { + "epoch": 0.18, + "learning_rate": 1.9426565619699867e-05, + "loss": 1.1633, + "step": 2773 + }, + { + "epoch": 0.18, + "learning_rate": 1.9426350930676916e-05, + "loss": 1.1484, + "step": 2774 + }, + { + "epoch": 0.18, + "learning_rate": 1.9426136241653965e-05, + "loss": 1.1634, + "step": 2775 + }, + { + "epoch": 0.18, + "learning_rate": 1.9425921552631014e-05, + "loss": 1.1403, + "step": 2776 + }, + { + "epoch": 0.18, + "learning_rate": 1.9425706863608063e-05, + "loss": 1.1717, + "step": 2777 + }, + { + "epoch": 0.18, + "learning_rate": 1.9425492174585116e-05, + "loss": 1.196, + "step": 2778 + }, + { + "epoch": 0.18, + "learning_rate": 1.9425277485562164e-05, + "loss": 1.1081, + "step": 2779 + }, + { + "epoch": 0.18, + "learning_rate": 1.9425062796539213e-05, + "loss": 1.0854, + "step": 2780 + }, + { + "epoch": 0.18, + "learning_rate": 1.9424848107516266e-05, + "loss": 1.0812, + "step": 2781 + }, + { + "epoch": 0.18, + "learning_rate": 1.9424633418493315e-05, + "loss": 1.1301, + "step": 2782 + }, + { + "epoch": 0.18, + "learning_rate": 1.9424418729470364e-05, + "loss": 1.0552, + "step": 2783 + }, + { + "epoch": 0.18, + "learning_rate": 1.9424204040447413e-05, + "loss": 1.1574, + "step": 2784 + }, + { + "epoch": 0.18, + "learning_rate": 1.9423989351424465e-05, + "loss": 1.0813, + "step": 2785 + }, + { + "epoch": 0.18, + "learning_rate": 1.9423774662401514e-05, + "loss": 1.0969, + "step": 2786 + }, + { + "epoch": 0.18, + "learning_rate": 1.9423559973378563e-05, + "loss": 1.1549, + "step": 2787 + }, + { + "epoch": 0.18, + "learning_rate": 1.9423345284355612e-05, + "loss": 1.1769, + "step": 2788 + }, + { + "epoch": 0.18, + "learning_rate": 1.942313059533266e-05, + "loss": 1.0977, + "step": 2789 + }, + { + "epoch": 0.18, + "learning_rate": 1.9422915906309713e-05, + "loss": 1.2118, + "step": 2790 + }, + { + "epoch": 0.18, + "learning_rate": 1.9422701217286762e-05, + "loss": 1.1925, + "step": 2791 + }, + { + "epoch": 0.18, + "learning_rate": 1.942248652826381e-05, + "loss": 1.1554, + "step": 2792 + }, + { + "epoch": 0.18, + "learning_rate": 1.942227183924086e-05, + "loss": 1.1812, + "step": 2793 + }, + { + "epoch": 0.18, + "learning_rate": 1.942205715021791e-05, + "loss": 1.0622, + "step": 2794 + }, + { + "epoch": 0.18, + "learning_rate": 1.9421842461194958e-05, + "loss": 1.2339, + "step": 2795 + }, + { + "epoch": 0.18, + "learning_rate": 1.942162777217201e-05, + "loss": 1.1886, + "step": 2796 + }, + { + "epoch": 0.18, + "learning_rate": 1.942141308314906e-05, + "loss": 1.2169, + "step": 2797 + }, + { + "epoch": 0.18, + "learning_rate": 1.9421198394126112e-05, + "loss": 1.1105, + "step": 2798 + }, + { + "epoch": 0.18, + "learning_rate": 1.942098370510316e-05, + "loss": 1.1455, + "step": 2799 + }, + { + "epoch": 0.18, + "learning_rate": 1.942076901608021e-05, + "loss": 1.1984, + "step": 2800 + }, + { + "epoch": 0.18, + "learning_rate": 1.942055432705726e-05, + "loss": 1.1763, + "step": 2801 + }, + { + "epoch": 0.18, + "learning_rate": 1.9420339638034308e-05, + "loss": 1.1319, + "step": 2802 + }, + { + "epoch": 0.18, + "learning_rate": 1.942012494901136e-05, + "loss": 1.1789, + "step": 2803 + }, + { + "epoch": 0.18, + "learning_rate": 1.941991025998841e-05, + "loss": 1.1167, + "step": 2804 + }, + { + "epoch": 0.18, + "learning_rate": 1.9419695570965458e-05, + "loss": 1.0781, + "step": 2805 + }, + { + "epoch": 0.18, + "learning_rate": 1.9419480881942507e-05, + "loss": 1.0967, + "step": 2806 + }, + { + "epoch": 0.18, + "learning_rate": 1.9419266192919556e-05, + "loss": 1.1001, + "step": 2807 + }, + { + "epoch": 0.18, + "learning_rate": 1.9419051503896608e-05, + "loss": 1.2022, + "step": 2808 + }, + { + "epoch": 0.18, + "learning_rate": 1.9418836814873657e-05, + "loss": 1.0937, + "step": 2809 + }, + { + "epoch": 0.18, + "learning_rate": 1.941862212585071e-05, + "loss": 1.2063, + "step": 2810 + }, + { + "epoch": 0.18, + "learning_rate": 1.941840743682776e-05, + "loss": 1.1425, + "step": 2811 + }, + { + "epoch": 0.18, + "learning_rate": 1.9418192747804807e-05, + "loss": 1.1677, + "step": 2812 + }, + { + "epoch": 0.18, + "learning_rate": 1.9417978058781856e-05, + "loss": 1.1387, + "step": 2813 + }, + { + "epoch": 0.18, + "learning_rate": 1.9417763369758905e-05, + "loss": 1.1449, + "step": 2814 + }, + { + "epoch": 0.18, + "learning_rate": 1.9417548680735954e-05, + "loss": 1.0925, + "step": 2815 + }, + { + "epoch": 0.18, + "learning_rate": 1.9417333991713007e-05, + "loss": 1.1802, + "step": 2816 + }, + { + "epoch": 0.18, + "learning_rate": 1.9417119302690056e-05, + "loss": 1.2295, + "step": 2817 + }, + { + "epoch": 0.18, + "learning_rate": 1.9416904613667105e-05, + "loss": 1.1202, + "step": 2818 + }, + { + "epoch": 0.18, + "learning_rate": 1.9416689924644154e-05, + "loss": 1.1501, + "step": 2819 + }, + { + "epoch": 0.18, + "learning_rate": 1.9416475235621203e-05, + "loss": 1.1161, + "step": 2820 + }, + { + "epoch": 0.18, + "learning_rate": 1.9416260546598255e-05, + "loss": 1.1333, + "step": 2821 + }, + { + "epoch": 0.18, + "learning_rate": 1.9416045857575304e-05, + "loss": 1.1929, + "step": 2822 + }, + { + "epoch": 0.18, + "learning_rate": 1.9415831168552353e-05, + "loss": 1.2066, + "step": 2823 + }, + { + "epoch": 0.18, + "learning_rate": 1.9415616479529402e-05, + "loss": 1.1824, + "step": 2824 + }, + { + "epoch": 0.18, + "learning_rate": 1.9415401790506454e-05, + "loss": 1.0842, + "step": 2825 + }, + { + "epoch": 0.18, + "learning_rate": 1.9415187101483503e-05, + "loss": 1.1445, + "step": 2826 + }, + { + "epoch": 0.18, + "learning_rate": 1.9414972412460552e-05, + "loss": 1.2123, + "step": 2827 + }, + { + "epoch": 0.18, + "learning_rate": 1.9414757723437604e-05, + "loss": 1.136, + "step": 2828 + }, + { + "epoch": 0.18, + "learning_rate": 1.9414543034414653e-05, + "loss": 1.2212, + "step": 2829 + }, + { + "epoch": 0.18, + "learning_rate": 1.9414328345391702e-05, + "loss": 1.0942, + "step": 2830 + }, + { + "epoch": 0.18, + "learning_rate": 1.941411365636875e-05, + "loss": 1.1808, + "step": 2831 + }, + { + "epoch": 0.18, + "learning_rate": 1.94138989673458e-05, + "loss": 1.1293, + "step": 2832 + }, + { + "epoch": 0.18, + "learning_rate": 1.941368427832285e-05, + "loss": 1.1521, + "step": 2833 + }, + { + "epoch": 0.18, + "learning_rate": 1.94134695892999e-05, + "loss": 1.1473, + "step": 2834 + }, + { + "epoch": 0.18, + "learning_rate": 1.941325490027695e-05, + "loss": 1.1719, + "step": 2835 + }, + { + "epoch": 0.18, + "learning_rate": 1.9413040211254e-05, + "loss": 1.1791, + "step": 2836 + }, + { + "epoch": 0.18, + "learning_rate": 1.941282552223105e-05, + "loss": 1.1499, + "step": 2837 + }, + { + "epoch": 0.18, + "learning_rate": 1.9412610833208097e-05, + "loss": 1.1016, + "step": 2838 + }, + { + "epoch": 0.18, + "learning_rate": 1.941239614418515e-05, + "loss": 1.1572, + "step": 2839 + }, + { + "epoch": 0.18, + "learning_rate": 1.94121814551622e-05, + "loss": 1.1623, + "step": 2840 + }, + { + "epoch": 0.18, + "learning_rate": 1.941196676613925e-05, + "loss": 1.1763, + "step": 2841 + }, + { + "epoch": 0.18, + "learning_rate": 1.94117520771163e-05, + "loss": 1.1588, + "step": 2842 + }, + { + "epoch": 0.18, + "learning_rate": 1.941153738809335e-05, + "loss": 1.0589, + "step": 2843 + }, + { + "epoch": 0.18, + "learning_rate": 1.9411322699070398e-05, + "loss": 1.1528, + "step": 2844 + }, + { + "epoch": 0.18, + "learning_rate": 1.9411108010047447e-05, + "loss": 1.1558, + "step": 2845 + }, + { + "epoch": 0.18, + "learning_rate": 1.94108933210245e-05, + "loss": 1.1285, + "step": 2846 + }, + { + "epoch": 0.18, + "learning_rate": 1.9410678632001548e-05, + "loss": 1.0879, + "step": 2847 + }, + { + "epoch": 0.18, + "learning_rate": 1.9410463942978597e-05, + "loss": 1.142, + "step": 2848 + }, + { + "epoch": 0.18, + "learning_rate": 1.9410249253955646e-05, + "loss": 1.0921, + "step": 2849 + }, + { + "epoch": 0.18, + "learning_rate": 1.9410034564932695e-05, + "loss": 1.1074, + "step": 2850 + }, + { + "epoch": 0.18, + "learning_rate": 1.9409819875909744e-05, + "loss": 1.1897, + "step": 2851 + }, + { + "epoch": 0.18, + "learning_rate": 1.9409605186886797e-05, + "loss": 1.2032, + "step": 2852 + }, + { + "epoch": 0.18, + "learning_rate": 1.9409390497863845e-05, + "loss": 1.1829, + "step": 2853 + }, + { + "epoch": 0.18, + "learning_rate": 1.9409175808840898e-05, + "loss": 1.132, + "step": 2854 + }, + { + "epoch": 0.18, + "learning_rate": 1.9408961119817947e-05, + "loss": 1.0845, + "step": 2855 + }, + { + "epoch": 0.18, + "learning_rate": 1.9408746430794996e-05, + "loss": 1.0582, + "step": 2856 + }, + { + "epoch": 0.18, + "learning_rate": 1.9408531741772045e-05, + "loss": 1.0559, + "step": 2857 + }, + { + "epoch": 0.18, + "learning_rate": 1.9408317052749094e-05, + "loss": 1.0905, + "step": 2858 + }, + { + "epoch": 0.18, + "learning_rate": 1.9408102363726146e-05, + "loss": 1.0543, + "step": 2859 + }, + { + "epoch": 0.18, + "learning_rate": 1.9407887674703195e-05, + "loss": 1.0664, + "step": 2860 + }, + { + "epoch": 0.18, + "learning_rate": 1.9407672985680244e-05, + "loss": 1.1351, + "step": 2861 + }, + { + "epoch": 0.18, + "learning_rate": 1.9407458296657293e-05, + "loss": 1.1554, + "step": 2862 + }, + { + "epoch": 0.18, + "learning_rate": 1.9407243607634342e-05, + "loss": 1.1307, + "step": 2863 + }, + { + "epoch": 0.18, + "learning_rate": 1.940702891861139e-05, + "loss": 1.1401, + "step": 2864 + }, + { + "epoch": 0.18, + "learning_rate": 1.9406814229588443e-05, + "loss": 1.2363, + "step": 2865 + }, + { + "epoch": 0.18, + "learning_rate": 1.9406599540565492e-05, + "loss": 1.1774, + "step": 2866 + }, + { + "epoch": 0.18, + "learning_rate": 1.940638485154254e-05, + "loss": 1.1364, + "step": 2867 + }, + { + "epoch": 0.18, + "learning_rate": 1.9406170162519593e-05, + "loss": 1.2614, + "step": 2868 + }, + { + "epoch": 0.18, + "learning_rate": 1.9405955473496642e-05, + "loss": 1.0516, + "step": 2869 + }, + { + "epoch": 0.18, + "learning_rate": 1.940574078447369e-05, + "loss": 1.1212, + "step": 2870 + }, + { + "epoch": 0.18, + "learning_rate": 1.940552609545074e-05, + "loss": 1.1125, + "step": 2871 + }, + { + "epoch": 0.18, + "learning_rate": 1.9405311406427793e-05, + "loss": 1.1635, + "step": 2872 + }, + { + "epoch": 0.18, + "learning_rate": 1.9405096717404842e-05, + "loss": 1.2521, + "step": 2873 + }, + { + "epoch": 0.18, + "learning_rate": 1.940488202838189e-05, + "loss": 1.1552, + "step": 2874 + }, + { + "epoch": 0.18, + "learning_rate": 1.940466733935894e-05, + "loss": 1.0616, + "step": 2875 + }, + { + "epoch": 0.19, + "learning_rate": 1.940445265033599e-05, + "loss": 1.1364, + "step": 2876 + }, + { + "epoch": 0.19, + "learning_rate": 1.940423796131304e-05, + "loss": 1.1552, + "step": 2877 + }, + { + "epoch": 0.19, + "learning_rate": 1.940402327229009e-05, + "loss": 1.1549, + "step": 2878 + }, + { + "epoch": 0.19, + "learning_rate": 1.940380858326714e-05, + "loss": 1.1915, + "step": 2879 + }, + { + "epoch": 0.19, + "learning_rate": 1.9403593894244188e-05, + "loss": 1.1506, + "step": 2880 + }, + { + "epoch": 0.19, + "learning_rate": 1.9403379205221237e-05, + "loss": 1.1753, + "step": 2881 + }, + { + "epoch": 0.19, + "learning_rate": 1.9403164516198286e-05, + "loss": 1.1236, + "step": 2882 + }, + { + "epoch": 0.19, + "learning_rate": 1.9402949827175338e-05, + "loss": 1.159, + "step": 2883 + }, + { + "epoch": 0.19, + "learning_rate": 1.9402735138152387e-05, + "loss": 1.1245, + "step": 2884 + }, + { + "epoch": 0.19, + "learning_rate": 1.940252044912944e-05, + "loss": 1.0309, + "step": 2885 + }, + { + "epoch": 0.19, + "learning_rate": 1.940230576010649e-05, + "loss": 1.1707, + "step": 2886 + }, + { + "epoch": 0.19, + "learning_rate": 1.9402091071083537e-05, + "loss": 1.149, + "step": 2887 + }, + { + "epoch": 0.19, + "learning_rate": 1.9401876382060586e-05, + "loss": 1.1615, + "step": 2888 + }, + { + "epoch": 0.19, + "learning_rate": 1.9401661693037635e-05, + "loss": 1.0781, + "step": 2889 + }, + { + "epoch": 0.19, + "learning_rate": 1.9401447004014688e-05, + "loss": 1.0858, + "step": 2890 + }, + { + "epoch": 0.19, + "learning_rate": 1.9401232314991737e-05, + "loss": 1.1706, + "step": 2891 + }, + { + "epoch": 0.19, + "learning_rate": 1.9401017625968786e-05, + "loss": 1.1856, + "step": 2892 + }, + { + "epoch": 0.19, + "learning_rate": 1.9400802936945835e-05, + "loss": 1.0848, + "step": 2893 + }, + { + "epoch": 0.19, + "learning_rate": 1.9400588247922884e-05, + "loss": 1.1408, + "step": 2894 + }, + { + "epoch": 0.19, + "learning_rate": 1.9400373558899936e-05, + "loss": 1.2688, + "step": 2895 + }, + { + "epoch": 0.19, + "learning_rate": 1.9400158869876985e-05, + "loss": 1.211, + "step": 2896 + }, + { + "epoch": 0.19, + "learning_rate": 1.9399944180854034e-05, + "loss": 1.1092, + "step": 2897 + }, + { + "epoch": 0.19, + "learning_rate": 1.9399729491831086e-05, + "loss": 1.1162, + "step": 2898 + }, + { + "epoch": 0.19, + "learning_rate": 1.9399514802808135e-05, + "loss": 1.1257, + "step": 2899 + }, + { + "epoch": 0.19, + "learning_rate": 1.9399300113785184e-05, + "loss": 1.1161, + "step": 2900 + }, + { + "epoch": 0.19, + "learning_rate": 1.9399085424762233e-05, + "loss": 1.1228, + "step": 2901 + }, + { + "epoch": 0.19, + "learning_rate": 1.9398870735739282e-05, + "loss": 1.1985, + "step": 2902 + }, + { + "epoch": 0.19, + "learning_rate": 1.9398656046716334e-05, + "loss": 1.1459, + "step": 2903 + }, + { + "epoch": 0.19, + "learning_rate": 1.9398441357693383e-05, + "loss": 1.131, + "step": 2904 + }, + { + "epoch": 0.19, + "learning_rate": 1.9398226668670432e-05, + "loss": 1.2138, + "step": 2905 + }, + { + "epoch": 0.19, + "learning_rate": 1.939801197964748e-05, + "loss": 1.2752, + "step": 2906 + }, + { + "epoch": 0.19, + "learning_rate": 1.939779729062453e-05, + "loss": 1.1163, + "step": 2907 + }, + { + "epoch": 0.19, + "learning_rate": 1.9397582601601583e-05, + "loss": 1.1593, + "step": 2908 + }, + { + "epoch": 0.19, + "learning_rate": 1.939736791257863e-05, + "loss": 1.3137, + "step": 2909 + }, + { + "epoch": 0.19, + "learning_rate": 1.939715322355568e-05, + "loss": 1.0889, + "step": 2910 + }, + { + "epoch": 0.19, + "learning_rate": 1.939693853453273e-05, + "loss": 1.0981, + "step": 2911 + }, + { + "epoch": 0.19, + "learning_rate": 1.9396723845509782e-05, + "loss": 1.1503, + "step": 2912 + }, + { + "epoch": 0.19, + "learning_rate": 1.939650915648683e-05, + "loss": 1.1546, + "step": 2913 + }, + { + "epoch": 0.19, + "learning_rate": 1.939629446746388e-05, + "loss": 1.1942, + "step": 2914 + }, + { + "epoch": 0.19, + "learning_rate": 1.9396079778440932e-05, + "loss": 1.1143, + "step": 2915 + }, + { + "epoch": 0.19, + "learning_rate": 1.939586508941798e-05, + "loss": 1.1743, + "step": 2916 + }, + { + "epoch": 0.19, + "learning_rate": 1.939565040039503e-05, + "loss": 1.1964, + "step": 2917 + }, + { + "epoch": 0.19, + "learning_rate": 1.939543571137208e-05, + "loss": 1.105, + "step": 2918 + }, + { + "epoch": 0.19, + "learning_rate": 1.9395221022349128e-05, + "loss": 1.1274, + "step": 2919 + }, + { + "epoch": 0.19, + "learning_rate": 1.9395006333326177e-05, + "loss": 1.1476, + "step": 2920 + }, + { + "epoch": 0.19, + "learning_rate": 1.939479164430323e-05, + "loss": 1.2128, + "step": 2921 + }, + { + "epoch": 0.19, + "learning_rate": 1.9394576955280278e-05, + "loss": 1.0742, + "step": 2922 + }, + { + "epoch": 0.19, + "learning_rate": 1.9394362266257327e-05, + "loss": 1.1582, + "step": 2923 + }, + { + "epoch": 0.19, + "learning_rate": 1.9394147577234376e-05, + "loss": 1.0713, + "step": 2924 + }, + { + "epoch": 0.19, + "learning_rate": 1.9393932888211425e-05, + "loss": 1.1705, + "step": 2925 + }, + { + "epoch": 0.19, + "learning_rate": 1.9393718199188478e-05, + "loss": 1.1764, + "step": 2926 + }, + { + "epoch": 0.19, + "learning_rate": 1.9393503510165526e-05, + "loss": 1.2098, + "step": 2927 + }, + { + "epoch": 0.19, + "learning_rate": 1.939328882114258e-05, + "loss": 1.2221, + "step": 2928 + }, + { + "epoch": 0.19, + "learning_rate": 1.9393074132119628e-05, + "loss": 1.1594, + "step": 2929 + }, + { + "epoch": 0.19, + "learning_rate": 1.9392859443096677e-05, + "loss": 1.1093, + "step": 2930 + }, + { + "epoch": 0.19, + "learning_rate": 1.9392644754073726e-05, + "loss": 1.1272, + "step": 2931 + }, + { + "epoch": 0.19, + "learning_rate": 1.9392430065050775e-05, + "loss": 1.1201, + "step": 2932 + }, + { + "epoch": 0.19, + "learning_rate": 1.9392215376027824e-05, + "loss": 1.1179, + "step": 2933 + }, + { + "epoch": 0.19, + "learning_rate": 1.9392000687004876e-05, + "loss": 1.1498, + "step": 2934 + }, + { + "epoch": 0.19, + "learning_rate": 1.9391785997981925e-05, + "loss": 1.0477, + "step": 2935 + }, + { + "epoch": 0.19, + "learning_rate": 1.9391571308958974e-05, + "loss": 1.0797, + "step": 2936 + }, + { + "epoch": 0.19, + "learning_rate": 1.9391356619936023e-05, + "loss": 1.1033, + "step": 2937 + }, + { + "epoch": 0.19, + "learning_rate": 1.9391141930913072e-05, + "loss": 1.0928, + "step": 2938 + }, + { + "epoch": 0.19, + "learning_rate": 1.9390927241890124e-05, + "loss": 1.0819, + "step": 2939 + }, + { + "epoch": 0.19, + "learning_rate": 1.9390712552867173e-05, + "loss": 1.1537, + "step": 2940 + }, + { + "epoch": 0.19, + "learning_rate": 1.9390497863844226e-05, + "loss": 1.1969, + "step": 2941 + }, + { + "epoch": 0.19, + "learning_rate": 1.9390283174821275e-05, + "loss": 1.1876, + "step": 2942 + }, + { + "epoch": 0.19, + "learning_rate": 1.9390068485798323e-05, + "loss": 1.0858, + "step": 2943 + }, + { + "epoch": 0.19, + "learning_rate": 1.9389853796775372e-05, + "loss": 1.178, + "step": 2944 + }, + { + "epoch": 0.19, + "learning_rate": 1.938963910775242e-05, + "loss": 1.1589, + "step": 2945 + }, + { + "epoch": 0.19, + "learning_rate": 1.9389424418729474e-05, + "loss": 1.1837, + "step": 2946 + }, + { + "epoch": 0.19, + "learning_rate": 1.9389209729706523e-05, + "loss": 1.2026, + "step": 2947 + }, + { + "epoch": 0.19, + "learning_rate": 1.938899504068357e-05, + "loss": 1.1143, + "step": 2948 + }, + { + "epoch": 0.19, + "learning_rate": 1.938878035166062e-05, + "loss": 1.1346, + "step": 2949 + }, + { + "epoch": 0.19, + "learning_rate": 1.938856566263767e-05, + "loss": 1.2017, + "step": 2950 + }, + { + "epoch": 0.19, + "learning_rate": 1.938835097361472e-05, + "loss": 1.2202, + "step": 2951 + }, + { + "epoch": 0.19, + "learning_rate": 1.938813628459177e-05, + "loss": 1.1407, + "step": 2952 + }, + { + "epoch": 0.19, + "learning_rate": 1.938792159556882e-05, + "loss": 1.1094, + "step": 2953 + }, + { + "epoch": 0.19, + "learning_rate": 1.938770690654587e-05, + "loss": 1.1105, + "step": 2954 + }, + { + "epoch": 0.19, + "learning_rate": 1.9387492217522918e-05, + "loss": 1.0976, + "step": 2955 + }, + { + "epoch": 0.19, + "learning_rate": 1.938727752849997e-05, + "loss": 1.1744, + "step": 2956 + }, + { + "epoch": 0.19, + "learning_rate": 1.938706283947702e-05, + "loss": 1.2582, + "step": 2957 + }, + { + "epoch": 0.19, + "learning_rate": 1.9386848150454068e-05, + "loss": 1.1557, + "step": 2958 + }, + { + "epoch": 0.19, + "learning_rate": 1.938663346143112e-05, + "loss": 1.1404, + "step": 2959 + }, + { + "epoch": 0.19, + "learning_rate": 1.938641877240817e-05, + "loss": 1.2441, + "step": 2960 + }, + { + "epoch": 0.19, + "learning_rate": 1.938620408338522e-05, + "loss": 1.1841, + "step": 2961 + }, + { + "epoch": 0.19, + "learning_rate": 1.9385989394362267e-05, + "loss": 1.1915, + "step": 2962 + }, + { + "epoch": 0.19, + "learning_rate": 1.9385774705339316e-05, + "loss": 1.0925, + "step": 2963 + }, + { + "epoch": 0.19, + "learning_rate": 1.938556001631637e-05, + "loss": 1.1712, + "step": 2964 + }, + { + "epoch": 0.19, + "learning_rate": 1.9385345327293418e-05, + "loss": 1.153, + "step": 2965 + }, + { + "epoch": 0.19, + "learning_rate": 1.9385130638270467e-05, + "loss": 1.1411, + "step": 2966 + }, + { + "epoch": 0.19, + "learning_rate": 1.9384915949247516e-05, + "loss": 1.2191, + "step": 2967 + }, + { + "epoch": 0.19, + "learning_rate": 1.9384701260224565e-05, + "loss": 1.0797, + "step": 2968 + }, + { + "epoch": 0.19, + "learning_rate": 1.9384486571201613e-05, + "loss": 1.1045, + "step": 2969 + }, + { + "epoch": 0.19, + "learning_rate": 1.9384271882178666e-05, + "loss": 1.2041, + "step": 2970 + }, + { + "epoch": 0.19, + "learning_rate": 1.9384057193155715e-05, + "loss": 1.2147, + "step": 2971 + }, + { + "epoch": 0.19, + "learning_rate": 1.9383842504132767e-05, + "loss": 1.1437, + "step": 2972 + }, + { + "epoch": 0.19, + "learning_rate": 1.9383627815109816e-05, + "loss": 1.1203, + "step": 2973 + }, + { + "epoch": 0.19, + "learning_rate": 1.9383413126086865e-05, + "loss": 1.0911, + "step": 2974 + }, + { + "epoch": 0.19, + "learning_rate": 1.9383198437063914e-05, + "loss": 1.187, + "step": 2975 + }, + { + "epoch": 0.19, + "learning_rate": 1.9382983748040963e-05, + "loss": 1.1172, + "step": 2976 + }, + { + "epoch": 0.19, + "learning_rate": 1.9382769059018015e-05, + "loss": 1.104, + "step": 2977 + }, + { + "epoch": 0.19, + "learning_rate": 1.9382554369995064e-05, + "loss": 1.2397, + "step": 2978 + }, + { + "epoch": 0.19, + "learning_rate": 1.9382339680972113e-05, + "loss": 1.1005, + "step": 2979 + }, + { + "epoch": 0.19, + "learning_rate": 1.9382124991949162e-05, + "loss": 1.1863, + "step": 2980 + }, + { + "epoch": 0.19, + "learning_rate": 1.938191030292621e-05, + "loss": 1.0998, + "step": 2981 + }, + { + "epoch": 0.19, + "learning_rate": 1.9381695613903264e-05, + "loss": 1.1636, + "step": 2982 + }, + { + "epoch": 0.19, + "learning_rate": 1.9381480924880313e-05, + "loss": 1.1261, + "step": 2983 + }, + { + "epoch": 0.19, + "learning_rate": 1.938126623585736e-05, + "loss": 1.0141, + "step": 2984 + }, + { + "epoch": 0.19, + "learning_rate": 1.9381051546834414e-05, + "loss": 1.1379, + "step": 2985 + }, + { + "epoch": 0.19, + "learning_rate": 1.9380836857811463e-05, + "loss": 1.1823, + "step": 2986 + }, + { + "epoch": 0.19, + "learning_rate": 1.9380622168788512e-05, + "loss": 1.1339, + "step": 2987 + }, + { + "epoch": 0.19, + "learning_rate": 1.938040747976556e-05, + "loss": 1.1611, + "step": 2988 + }, + { + "epoch": 0.19, + "learning_rate": 1.938019279074261e-05, + "loss": 1.132, + "step": 2989 + }, + { + "epoch": 0.19, + "learning_rate": 1.9379978101719662e-05, + "loss": 1.1487, + "step": 2990 + }, + { + "epoch": 0.19, + "learning_rate": 1.937976341269671e-05, + "loss": 1.1784, + "step": 2991 + }, + { + "epoch": 0.19, + "learning_rate": 1.937954872367376e-05, + "loss": 1.188, + "step": 2992 + }, + { + "epoch": 0.19, + "learning_rate": 1.937933403465081e-05, + "loss": 1.1742, + "step": 2993 + }, + { + "epoch": 0.19, + "learning_rate": 1.9379119345627858e-05, + "loss": 1.1309, + "step": 2994 + }, + { + "epoch": 0.19, + "learning_rate": 1.937890465660491e-05, + "loss": 1.2252, + "step": 2995 + }, + { + "epoch": 0.19, + "learning_rate": 1.937868996758196e-05, + "loss": 1.1594, + "step": 2996 + }, + { + "epoch": 0.19, + "learning_rate": 1.9378475278559008e-05, + "loss": 1.0901, + "step": 2997 + }, + { + "epoch": 0.19, + "learning_rate": 1.9378260589536057e-05, + "loss": 1.2, + "step": 2998 + }, + { + "epoch": 0.19, + "learning_rate": 1.937804590051311e-05, + "loss": 1.1366, + "step": 2999 + }, + { + "epoch": 0.19, + "learning_rate": 1.937783121149016e-05, + "loss": 1.2292, + "step": 3000 + }, + { + "epoch": 0.19, + "eval_laion/cider": 0.44376184324315476, + "eval_laion/meteor": 0.37961963501315543, + "eval_laion/pred_num_words": 38.666015625, + "eval_laion/sacrebleu": 15.21276865720541, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.1588678359985352, + "eval_runtime": 97.0914, + "eval_samples_per_second": 5.139, + "eval_steps_per_second": 0.33, + "step": 3000 + }, + { + "epoch": 0.19, + "learning_rate": 1.9377616522467207e-05, + "loss": 1.1289, + "step": 3001 + }, + { + "epoch": 0.19, + "learning_rate": 1.937740183344426e-05, + "loss": 1.1191, + "step": 3002 + }, + { + "epoch": 0.19, + "learning_rate": 1.937718714442131e-05, + "loss": 1.0408, + "step": 3003 + }, + { + "epoch": 0.19, + "learning_rate": 1.9376972455398358e-05, + "loss": 1.1567, + "step": 3004 + }, + { + "epoch": 0.19, + "learning_rate": 1.9376757766375407e-05, + "loss": 1.1647, + "step": 3005 + }, + { + "epoch": 0.19, + "learning_rate": 1.9376543077352456e-05, + "loss": 1.0627, + "step": 3006 + }, + { + "epoch": 0.19, + "learning_rate": 1.9376328388329505e-05, + "loss": 1.2048, + "step": 3007 + }, + { + "epoch": 0.19, + "learning_rate": 1.9376113699306557e-05, + "loss": 1.1828, + "step": 3008 + }, + { + "epoch": 0.19, + "learning_rate": 1.9375899010283606e-05, + "loss": 1.0275, + "step": 3009 + }, + { + "epoch": 0.19, + "learning_rate": 1.9375684321260655e-05, + "loss": 1.1662, + "step": 3010 + }, + { + "epoch": 0.19, + "learning_rate": 1.9375469632237704e-05, + "loss": 1.0095, + "step": 3011 + }, + { + "epoch": 0.19, + "learning_rate": 1.9375254943214753e-05, + "loss": 1.1073, + "step": 3012 + }, + { + "epoch": 0.19, + "learning_rate": 1.9375040254191805e-05, + "loss": 1.1235, + "step": 3013 + }, + { + "epoch": 0.19, + "learning_rate": 1.9374825565168854e-05, + "loss": 1.0553, + "step": 3014 + }, + { + "epoch": 0.19, + "learning_rate": 1.9374610876145907e-05, + "loss": 1.0703, + "step": 3015 + }, + { + "epoch": 0.19, + "learning_rate": 1.9374396187122956e-05, + "loss": 1.1425, + "step": 3016 + }, + { + "epoch": 0.19, + "learning_rate": 1.9374181498100004e-05, + "loss": 1.1411, + "step": 3017 + }, + { + "epoch": 0.19, + "learning_rate": 1.9373966809077053e-05, + "loss": 1.0212, + "step": 3018 + }, + { + "epoch": 0.19, + "learning_rate": 1.9373752120054102e-05, + "loss": 1.1332, + "step": 3019 + }, + { + "epoch": 0.19, + "learning_rate": 1.937353743103115e-05, + "loss": 1.2063, + "step": 3020 + }, + { + "epoch": 0.19, + "learning_rate": 1.9373322742008204e-05, + "loss": 1.2209, + "step": 3021 + }, + { + "epoch": 0.19, + "learning_rate": 1.9373108052985253e-05, + "loss": 1.1876, + "step": 3022 + }, + { + "epoch": 0.19, + "learning_rate": 1.93728933639623e-05, + "loss": 1.3172, + "step": 3023 + }, + { + "epoch": 0.19, + "learning_rate": 1.937267867493935e-05, + "loss": 1.1573, + "step": 3024 + }, + { + "epoch": 0.19, + "learning_rate": 1.93724639859164e-05, + "loss": 1.1954, + "step": 3025 + }, + { + "epoch": 0.19, + "learning_rate": 1.9372249296893452e-05, + "loss": 1.1391, + "step": 3026 + }, + { + "epoch": 0.19, + "learning_rate": 1.93720346078705e-05, + "loss": 1.1097, + "step": 3027 + }, + { + "epoch": 0.19, + "learning_rate": 1.9371819918847553e-05, + "loss": 1.1351, + "step": 3028 + }, + { + "epoch": 0.19, + "learning_rate": 1.9371605229824602e-05, + "loss": 1.1664, + "step": 3029 + }, + { + "epoch": 0.19, + "learning_rate": 1.937139054080165e-05, + "loss": 1.1485, + "step": 3030 + }, + { + "epoch": 0.2, + "learning_rate": 1.93711758517787e-05, + "loss": 1.1777, + "step": 3031 + }, + { + "epoch": 0.2, + "learning_rate": 1.937096116275575e-05, + "loss": 1.1834, + "step": 3032 + }, + { + "epoch": 0.2, + "learning_rate": 1.93707464737328e-05, + "loss": 1.158, + "step": 3033 + }, + { + "epoch": 0.2, + "learning_rate": 1.937053178470985e-05, + "loss": 1.1262, + "step": 3034 + }, + { + "epoch": 0.2, + "learning_rate": 1.93703170956869e-05, + "loss": 1.161, + "step": 3035 + }, + { + "epoch": 0.2, + "learning_rate": 1.937010240666395e-05, + "loss": 1.0593, + "step": 3036 + }, + { + "epoch": 0.2, + "learning_rate": 1.9369887717640997e-05, + "loss": 1.1326, + "step": 3037 + }, + { + "epoch": 0.2, + "learning_rate": 1.9369673028618046e-05, + "loss": 1.1329, + "step": 3038 + }, + { + "epoch": 0.2, + "learning_rate": 1.93694583395951e-05, + "loss": 1.1756, + "step": 3039 + }, + { + "epoch": 0.2, + "learning_rate": 1.9369243650572148e-05, + "loss": 1.1286, + "step": 3040 + }, + { + "epoch": 0.2, + "learning_rate": 1.9369028961549197e-05, + "loss": 1.2003, + "step": 3041 + }, + { + "epoch": 0.2, + "learning_rate": 1.9368814272526246e-05, + "loss": 1.3104, + "step": 3042 + }, + { + "epoch": 0.2, + "learning_rate": 1.9368599583503298e-05, + "loss": 1.1137, + "step": 3043 + }, + { + "epoch": 0.2, + "learning_rate": 1.9368384894480347e-05, + "loss": 1.1292, + "step": 3044 + }, + { + "epoch": 0.2, + "learning_rate": 1.9368170205457396e-05, + "loss": 1.2032, + "step": 3045 + }, + { + "epoch": 0.2, + "learning_rate": 1.9367955516434448e-05, + "loss": 1.1426, + "step": 3046 + }, + { + "epoch": 0.2, + "learning_rate": 1.9367740827411497e-05, + "loss": 1.1641, + "step": 3047 + }, + { + "epoch": 0.2, + "learning_rate": 1.9367526138388546e-05, + "loss": 1.1626, + "step": 3048 + }, + { + "epoch": 0.2, + "learning_rate": 1.9367311449365595e-05, + "loss": 1.0703, + "step": 3049 + }, + { + "epoch": 0.2, + "learning_rate": 1.9367096760342644e-05, + "loss": 1.1828, + "step": 3050 + }, + { + "epoch": 0.2, + "learning_rate": 1.9366882071319696e-05, + "loss": 1.1318, + "step": 3051 + }, + { + "epoch": 0.2, + "learning_rate": 1.9366667382296745e-05, + "loss": 1.1855, + "step": 3052 + }, + { + "epoch": 0.2, + "learning_rate": 1.9366452693273794e-05, + "loss": 1.1707, + "step": 3053 + }, + { + "epoch": 0.2, + "learning_rate": 1.9366238004250843e-05, + "loss": 1.1831, + "step": 3054 + }, + { + "epoch": 0.2, + "learning_rate": 1.9366023315227892e-05, + "loss": 1.0957, + "step": 3055 + }, + { + "epoch": 0.2, + "learning_rate": 1.936580862620494e-05, + "loss": 1.076, + "step": 3056 + }, + { + "epoch": 0.2, + "learning_rate": 1.9365593937181994e-05, + "loss": 1.122, + "step": 3057 + }, + { + "epoch": 0.2, + "learning_rate": 1.9365379248159043e-05, + "loss": 1.1467, + "step": 3058 + }, + { + "epoch": 0.2, + "learning_rate": 1.9365164559136095e-05, + "loss": 1.1121, + "step": 3059 + }, + { + "epoch": 0.2, + "learning_rate": 1.9364949870113144e-05, + "loss": 1.1492, + "step": 3060 + }, + { + "epoch": 0.2, + "learning_rate": 1.9364735181090193e-05, + "loss": 1.1833, + "step": 3061 + }, + { + "epoch": 0.2, + "learning_rate": 1.9364520492067242e-05, + "loss": 1.1808, + "step": 3062 + }, + { + "epoch": 0.2, + "learning_rate": 1.936430580304429e-05, + "loss": 1.1738, + "step": 3063 + }, + { + "epoch": 0.2, + "learning_rate": 1.9364091114021343e-05, + "loss": 1.1664, + "step": 3064 + }, + { + "epoch": 0.2, + "learning_rate": 1.9363876424998392e-05, + "loss": 1.067, + "step": 3065 + }, + { + "epoch": 0.2, + "learning_rate": 1.936366173597544e-05, + "loss": 1.1192, + "step": 3066 + }, + { + "epoch": 0.2, + "learning_rate": 1.936344704695249e-05, + "loss": 1.1338, + "step": 3067 + }, + { + "epoch": 0.2, + "learning_rate": 1.936323235792954e-05, + "loss": 1.1466, + "step": 3068 + }, + { + "epoch": 0.2, + "learning_rate": 1.936301766890659e-05, + "loss": 1.1173, + "step": 3069 + }, + { + "epoch": 0.2, + "learning_rate": 1.936280297988364e-05, + "loss": 1.2281, + "step": 3070 + }, + { + "epoch": 0.2, + "learning_rate": 1.936258829086069e-05, + "loss": 1.1842, + "step": 3071 + }, + { + "epoch": 0.2, + "learning_rate": 1.936237360183774e-05, + "loss": 1.0798, + "step": 3072 + }, + { + "epoch": 0.2, + "learning_rate": 1.936215891281479e-05, + "loss": 1.1123, + "step": 3073 + }, + { + "epoch": 0.2, + "learning_rate": 1.936194422379184e-05, + "loss": 1.1955, + "step": 3074 + }, + { + "epoch": 0.2, + "learning_rate": 1.936172953476889e-05, + "loss": 1.0987, + "step": 3075 + }, + { + "epoch": 0.2, + "learning_rate": 1.9361514845745937e-05, + "loss": 1.1095, + "step": 3076 + }, + { + "epoch": 0.2, + "learning_rate": 1.936130015672299e-05, + "loss": 1.1529, + "step": 3077 + }, + { + "epoch": 0.2, + "learning_rate": 1.936108546770004e-05, + "loss": 1.1004, + "step": 3078 + }, + { + "epoch": 0.2, + "learning_rate": 1.9360870778677088e-05, + "loss": 1.1439, + "step": 3079 + }, + { + "epoch": 0.2, + "learning_rate": 1.9360656089654137e-05, + "loss": 1.1028, + "step": 3080 + }, + { + "epoch": 0.2, + "learning_rate": 1.9360441400631186e-05, + "loss": 1.1101, + "step": 3081 + }, + { + "epoch": 0.2, + "learning_rate": 1.9360226711608238e-05, + "loss": 1.1281, + "step": 3082 + }, + { + "epoch": 0.2, + "learning_rate": 1.9360012022585287e-05, + "loss": 1.1596, + "step": 3083 + }, + { + "epoch": 0.2, + "learning_rate": 1.9359797333562336e-05, + "loss": 1.1713, + "step": 3084 + }, + { + "epoch": 0.2, + "learning_rate": 1.9359582644539385e-05, + "loss": 1.1674, + "step": 3085 + }, + { + "epoch": 0.2, + "learning_rate": 1.9359367955516437e-05, + "loss": 1.2143, + "step": 3086 + }, + { + "epoch": 0.2, + "learning_rate": 1.9359153266493486e-05, + "loss": 1.1236, + "step": 3087 + }, + { + "epoch": 0.2, + "learning_rate": 1.9358938577470535e-05, + "loss": 1.2471, + "step": 3088 + }, + { + "epoch": 0.2, + "learning_rate": 1.9358723888447588e-05, + "loss": 1.1392, + "step": 3089 + }, + { + "epoch": 0.2, + "learning_rate": 1.9358509199424637e-05, + "loss": 1.0981, + "step": 3090 + }, + { + "epoch": 0.2, + "learning_rate": 1.9358294510401685e-05, + "loss": 1.1633, + "step": 3091 + }, + { + "epoch": 0.2, + "learning_rate": 1.9358079821378734e-05, + "loss": 1.07, + "step": 3092 + }, + { + "epoch": 0.2, + "learning_rate": 1.9357865132355783e-05, + "loss": 1.0293, + "step": 3093 + }, + { + "epoch": 0.2, + "learning_rate": 1.9357650443332832e-05, + "loss": 1.116, + "step": 3094 + }, + { + "epoch": 0.2, + "learning_rate": 1.9357435754309885e-05, + "loss": 1.2188, + "step": 3095 + }, + { + "epoch": 0.2, + "learning_rate": 1.9357221065286934e-05, + "loss": 1.1758, + "step": 3096 + }, + { + "epoch": 0.2, + "learning_rate": 1.9357006376263983e-05, + "loss": 1.0037, + "step": 3097 + }, + { + "epoch": 0.2, + "learning_rate": 1.935679168724103e-05, + "loss": 1.147, + "step": 3098 + }, + { + "epoch": 0.2, + "learning_rate": 1.935657699821808e-05, + "loss": 1.1508, + "step": 3099 + }, + { + "epoch": 0.2, + "learning_rate": 1.9356362309195133e-05, + "loss": 1.0817, + "step": 3100 + }, + { + "epoch": 0.2, + "learning_rate": 1.9356147620172182e-05, + "loss": 1.1163, + "step": 3101 + }, + { + "epoch": 0.2, + "learning_rate": 1.9355932931149234e-05, + "loss": 1.0582, + "step": 3102 + }, + { + "epoch": 0.2, + "learning_rate": 1.9355718242126283e-05, + "loss": 1.1144, + "step": 3103 + }, + { + "epoch": 0.2, + "learning_rate": 1.9355503553103332e-05, + "loss": 1.1456, + "step": 3104 + }, + { + "epoch": 0.2, + "learning_rate": 1.935528886408038e-05, + "loss": 1.0313, + "step": 3105 + }, + { + "epoch": 0.2, + "learning_rate": 1.935507417505743e-05, + "loss": 1.2054, + "step": 3106 + }, + { + "epoch": 0.2, + "learning_rate": 1.935485948603448e-05, + "loss": 1.1623, + "step": 3107 + }, + { + "epoch": 0.2, + "learning_rate": 1.935464479701153e-05, + "loss": 1.175, + "step": 3108 + }, + { + "epoch": 0.2, + "learning_rate": 1.935443010798858e-05, + "loss": 1.1849, + "step": 3109 + }, + { + "epoch": 0.2, + "learning_rate": 1.935421541896563e-05, + "loss": 1.0798, + "step": 3110 + }, + { + "epoch": 0.2, + "learning_rate": 1.935400072994268e-05, + "loss": 1.0544, + "step": 3111 + }, + { + "epoch": 0.2, + "learning_rate": 1.9353786040919727e-05, + "loss": 1.0722, + "step": 3112 + }, + { + "epoch": 0.2, + "learning_rate": 1.935357135189678e-05, + "loss": 1.1568, + "step": 3113 + }, + { + "epoch": 0.2, + "learning_rate": 1.935335666287383e-05, + "loss": 1.2096, + "step": 3114 + }, + { + "epoch": 0.2, + "learning_rate": 1.935314197385088e-05, + "loss": 1.154, + "step": 3115 + }, + { + "epoch": 0.2, + "learning_rate": 1.935292728482793e-05, + "loss": 1.1465, + "step": 3116 + }, + { + "epoch": 0.2, + "learning_rate": 1.935271259580498e-05, + "loss": 1.1767, + "step": 3117 + }, + { + "epoch": 0.2, + "learning_rate": 1.9352497906782028e-05, + "loss": 1.1973, + "step": 3118 + }, + { + "epoch": 0.2, + "learning_rate": 1.9352283217759077e-05, + "loss": 1.134, + "step": 3119 + }, + { + "epoch": 0.2, + "learning_rate": 1.935206852873613e-05, + "loss": 1.1622, + "step": 3120 + }, + { + "epoch": 0.2, + "learning_rate": 1.9351853839713178e-05, + "loss": 1.1449, + "step": 3121 + }, + { + "epoch": 0.2, + "learning_rate": 1.9351639150690227e-05, + "loss": 1.1817, + "step": 3122 + }, + { + "epoch": 0.2, + "learning_rate": 1.9351424461667276e-05, + "loss": 1.0948, + "step": 3123 + }, + { + "epoch": 0.2, + "learning_rate": 1.9351209772644325e-05, + "loss": 1.1513, + "step": 3124 + }, + { + "epoch": 0.2, + "learning_rate": 1.9350995083621374e-05, + "loss": 1.2398, + "step": 3125 + }, + { + "epoch": 0.2, + "learning_rate": 1.9350780394598426e-05, + "loss": 1.1129, + "step": 3126 + }, + { + "epoch": 0.2, + "learning_rate": 1.9350565705575475e-05, + "loss": 1.1001, + "step": 3127 + }, + { + "epoch": 0.2, + "learning_rate": 1.9350351016552524e-05, + "loss": 1.0936, + "step": 3128 + }, + { + "epoch": 0.2, + "learning_rate": 1.9350136327529573e-05, + "loss": 1.1177, + "step": 3129 + }, + { + "epoch": 0.2, + "learning_rate": 1.9349921638506626e-05, + "loss": 1.1307, + "step": 3130 + }, + { + "epoch": 0.2, + "learning_rate": 1.9349706949483675e-05, + "loss": 1.1316, + "step": 3131 + }, + { + "epoch": 0.2, + "learning_rate": 1.9349492260460724e-05, + "loss": 1.2712, + "step": 3132 + }, + { + "epoch": 0.2, + "learning_rate": 1.9349277571437776e-05, + "loss": 1.0535, + "step": 3133 + }, + { + "epoch": 0.2, + "learning_rate": 1.9349062882414825e-05, + "loss": 1.173, + "step": 3134 + }, + { + "epoch": 0.2, + "learning_rate": 1.9348848193391874e-05, + "loss": 1.1062, + "step": 3135 + }, + { + "epoch": 0.2, + "learning_rate": 1.9348633504368923e-05, + "loss": 1.0967, + "step": 3136 + }, + { + "epoch": 0.2, + "learning_rate": 1.9348418815345972e-05, + "loss": 1.1024, + "step": 3137 + }, + { + "epoch": 0.2, + "learning_rate": 1.9348204126323024e-05, + "loss": 1.1205, + "step": 3138 + }, + { + "epoch": 0.2, + "learning_rate": 1.9347989437300073e-05, + "loss": 1.1722, + "step": 3139 + }, + { + "epoch": 0.2, + "learning_rate": 1.9347774748277122e-05, + "loss": 1.098, + "step": 3140 + }, + { + "epoch": 0.2, + "learning_rate": 1.934756005925417e-05, + "loss": 1.1729, + "step": 3141 + }, + { + "epoch": 0.2, + "learning_rate": 1.934734537023122e-05, + "loss": 1.2102, + "step": 3142 + }, + { + "epoch": 0.2, + "learning_rate": 1.934713068120827e-05, + "loss": 1.1165, + "step": 3143 + }, + { + "epoch": 0.2, + "learning_rate": 1.934691599218532e-05, + "loss": 1.2036, + "step": 3144 + }, + { + "epoch": 0.2, + "learning_rate": 1.934670130316237e-05, + "loss": 1.0775, + "step": 3145 + }, + { + "epoch": 0.2, + "learning_rate": 1.9346486614139423e-05, + "loss": 1.1743, + "step": 3146 + }, + { + "epoch": 0.2, + "learning_rate": 1.934627192511647e-05, + "loss": 1.0734, + "step": 3147 + }, + { + "epoch": 0.2, + "learning_rate": 1.934605723609352e-05, + "loss": 1.1158, + "step": 3148 + }, + { + "epoch": 0.2, + "learning_rate": 1.934584254707057e-05, + "loss": 1.0792, + "step": 3149 + }, + { + "epoch": 0.2, + "learning_rate": 1.934562785804762e-05, + "loss": 1.1429, + "step": 3150 + }, + { + "epoch": 0.2, + "learning_rate": 1.934541316902467e-05, + "loss": 1.1167, + "step": 3151 + }, + { + "epoch": 0.2, + "learning_rate": 1.934519848000172e-05, + "loss": 1.1078, + "step": 3152 + }, + { + "epoch": 0.2, + "learning_rate": 1.934498379097877e-05, + "loss": 1.099, + "step": 3153 + }, + { + "epoch": 0.2, + "learning_rate": 1.9344769101955818e-05, + "loss": 1.1919, + "step": 3154 + }, + { + "epoch": 0.2, + "learning_rate": 1.9344554412932867e-05, + "loss": 1.2403, + "step": 3155 + }, + { + "epoch": 0.2, + "learning_rate": 1.934433972390992e-05, + "loss": 1.0793, + "step": 3156 + }, + { + "epoch": 0.2, + "learning_rate": 1.9344125034886968e-05, + "loss": 1.0841, + "step": 3157 + }, + { + "epoch": 0.2, + "learning_rate": 1.9343910345864017e-05, + "loss": 1.2085, + "step": 3158 + }, + { + "epoch": 0.2, + "learning_rate": 1.934369565684107e-05, + "loss": 1.1473, + "step": 3159 + }, + { + "epoch": 0.2, + "learning_rate": 1.9343480967818118e-05, + "loss": 1.1606, + "step": 3160 + }, + { + "epoch": 0.2, + "learning_rate": 1.9343266278795167e-05, + "loss": 1.1869, + "step": 3161 + }, + { + "epoch": 0.2, + "learning_rate": 1.9343051589772216e-05, + "loss": 1.1012, + "step": 3162 + }, + { + "epoch": 0.2, + "learning_rate": 1.9342836900749265e-05, + "loss": 1.0709, + "step": 3163 + }, + { + "epoch": 0.2, + "learning_rate": 1.9342622211726318e-05, + "loss": 1.2122, + "step": 3164 + }, + { + "epoch": 0.2, + "learning_rate": 1.9342407522703366e-05, + "loss": 1.1216, + "step": 3165 + }, + { + "epoch": 0.2, + "learning_rate": 1.9342192833680415e-05, + "loss": 1.0784, + "step": 3166 + }, + { + "epoch": 0.2, + "learning_rate": 1.9341978144657464e-05, + "loss": 1.1373, + "step": 3167 + }, + { + "epoch": 0.2, + "learning_rate": 1.9341763455634513e-05, + "loss": 1.1402, + "step": 3168 + }, + { + "epoch": 0.2, + "learning_rate": 1.9341548766611566e-05, + "loss": 1.2176, + "step": 3169 + }, + { + "epoch": 0.2, + "learning_rate": 1.9341334077588615e-05, + "loss": 1.1442, + "step": 3170 + }, + { + "epoch": 0.2, + "learning_rate": 1.9341119388565664e-05, + "loss": 1.091, + "step": 3171 + }, + { + "epoch": 0.2, + "learning_rate": 1.9340904699542713e-05, + "loss": 1.1245, + "step": 3172 + }, + { + "epoch": 0.2, + "learning_rate": 1.9340690010519765e-05, + "loss": 1.1674, + "step": 3173 + }, + { + "epoch": 0.2, + "learning_rate": 1.9340475321496814e-05, + "loss": 1.1939, + "step": 3174 + }, + { + "epoch": 0.2, + "learning_rate": 1.9340260632473863e-05, + "loss": 1.1538, + "step": 3175 + }, + { + "epoch": 0.2, + "learning_rate": 1.9340045943450915e-05, + "loss": 1.0611, + "step": 3176 + }, + { + "epoch": 0.2, + "learning_rate": 1.9339831254427964e-05, + "loss": 1.1854, + "step": 3177 + }, + { + "epoch": 0.2, + "learning_rate": 1.9339616565405013e-05, + "loss": 1.1441, + "step": 3178 + }, + { + "epoch": 0.2, + "learning_rate": 1.9339401876382062e-05, + "loss": 1.1061, + "step": 3179 + }, + { + "epoch": 0.2, + "learning_rate": 1.933918718735911e-05, + "loss": 1.0271, + "step": 3180 + }, + { + "epoch": 0.2, + "learning_rate": 1.933897249833616e-05, + "loss": 1.1566, + "step": 3181 + }, + { + "epoch": 0.2, + "learning_rate": 1.9338757809313212e-05, + "loss": 1.1342, + "step": 3182 + }, + { + "epoch": 0.2, + "learning_rate": 1.933854312029026e-05, + "loss": 1.124, + "step": 3183 + }, + { + "epoch": 0.2, + "learning_rate": 1.933832843126731e-05, + "loss": 1.1487, + "step": 3184 + }, + { + "epoch": 0.2, + "learning_rate": 1.933811374224436e-05, + "loss": 1.1394, + "step": 3185 + }, + { + "epoch": 0.2, + "learning_rate": 1.9337899053221408e-05, + "loss": 1.0639, + "step": 3186 + }, + { + "epoch": 0.21, + "learning_rate": 1.933768436419846e-05, + "loss": 1.1018, + "step": 3187 + }, + { + "epoch": 0.21, + "learning_rate": 1.933746967517551e-05, + "loss": 1.2302, + "step": 3188 + }, + { + "epoch": 0.21, + "learning_rate": 1.9337254986152562e-05, + "loss": 1.1443, + "step": 3189 + }, + { + "epoch": 0.21, + "learning_rate": 1.933704029712961e-05, + "loss": 1.1279, + "step": 3190 + }, + { + "epoch": 0.21, + "learning_rate": 1.933682560810666e-05, + "loss": 1.149, + "step": 3191 + }, + { + "epoch": 0.21, + "learning_rate": 1.933661091908371e-05, + "loss": 1.1431, + "step": 3192 + }, + { + "epoch": 0.21, + "learning_rate": 1.9336396230060758e-05, + "loss": 1.1844, + "step": 3193 + }, + { + "epoch": 0.21, + "learning_rate": 1.9336181541037807e-05, + "loss": 1.1728, + "step": 3194 + }, + { + "epoch": 0.21, + "learning_rate": 1.933596685201486e-05, + "loss": 1.1553, + "step": 3195 + }, + { + "epoch": 0.21, + "learning_rate": 1.9335752162991908e-05, + "loss": 1.1849, + "step": 3196 + }, + { + "epoch": 0.21, + "learning_rate": 1.9335537473968957e-05, + "loss": 1.0824, + "step": 3197 + }, + { + "epoch": 0.21, + "learning_rate": 1.9335322784946006e-05, + "loss": 1.0664, + "step": 3198 + }, + { + "epoch": 0.21, + "learning_rate": 1.9335108095923055e-05, + "loss": 1.1342, + "step": 3199 + }, + { + "epoch": 0.21, + "learning_rate": 1.9334893406900107e-05, + "loss": 1.1239, + "step": 3200 + }, + { + "epoch": 0.21, + "learning_rate": 1.9334678717877156e-05, + "loss": 1.2573, + "step": 3201 + }, + { + "epoch": 0.21, + "learning_rate": 1.933446402885421e-05, + "loss": 1.145, + "step": 3202 + }, + { + "epoch": 0.21, + "learning_rate": 1.9334249339831258e-05, + "loss": 1.0952, + "step": 3203 + }, + { + "epoch": 0.21, + "learning_rate": 1.9334034650808307e-05, + "loss": 1.1601, + "step": 3204 + }, + { + "epoch": 0.21, + "learning_rate": 1.9333819961785356e-05, + "loss": 1.1246, + "step": 3205 + }, + { + "epoch": 0.21, + "learning_rate": 1.9333605272762405e-05, + "loss": 1.1986, + "step": 3206 + }, + { + "epoch": 0.21, + "learning_rate": 1.9333390583739457e-05, + "loss": 1.1489, + "step": 3207 + }, + { + "epoch": 0.21, + "learning_rate": 1.9333175894716506e-05, + "loss": 1.1238, + "step": 3208 + }, + { + "epoch": 0.21, + "learning_rate": 1.9332961205693555e-05, + "loss": 1.0995, + "step": 3209 + }, + { + "epoch": 0.21, + "learning_rate": 1.9332746516670604e-05, + "loss": 1.1897, + "step": 3210 + }, + { + "epoch": 0.21, + "learning_rate": 1.9332531827647653e-05, + "loss": 1.1497, + "step": 3211 + }, + { + "epoch": 0.21, + "learning_rate": 1.9332317138624702e-05, + "loss": 1.2804, + "step": 3212 + }, + { + "epoch": 0.21, + "learning_rate": 1.9332102449601754e-05, + "loss": 1.0805, + "step": 3213 + }, + { + "epoch": 0.21, + "learning_rate": 1.9331887760578803e-05, + "loss": 1.1403, + "step": 3214 + }, + { + "epoch": 0.21, + "learning_rate": 1.9331673071555852e-05, + "loss": 1.1572, + "step": 3215 + }, + { + "epoch": 0.21, + "learning_rate": 1.93314583825329e-05, + "loss": 1.1838, + "step": 3216 + }, + { + "epoch": 0.21, + "learning_rate": 1.9331243693509953e-05, + "loss": 1.2252, + "step": 3217 + }, + { + "epoch": 0.21, + "learning_rate": 1.9331029004487002e-05, + "loss": 1.2155, + "step": 3218 + }, + { + "epoch": 0.21, + "learning_rate": 1.933081431546405e-05, + "loss": 1.1511, + "step": 3219 + }, + { + "epoch": 0.21, + "learning_rate": 1.9330599626441104e-05, + "loss": 1.2091, + "step": 3220 + }, + { + "epoch": 0.21, + "learning_rate": 1.9330384937418153e-05, + "loss": 1.2342, + "step": 3221 + }, + { + "epoch": 0.21, + "learning_rate": 1.93301702483952e-05, + "loss": 1.1896, + "step": 3222 + }, + { + "epoch": 0.21, + "learning_rate": 1.932995555937225e-05, + "loss": 1.0835, + "step": 3223 + }, + { + "epoch": 0.21, + "learning_rate": 1.93297408703493e-05, + "loss": 1.1388, + "step": 3224 + }, + { + "epoch": 0.21, + "learning_rate": 1.9329526181326352e-05, + "loss": 1.1905, + "step": 3225 + }, + { + "epoch": 0.21, + "learning_rate": 1.93293114923034e-05, + "loss": 1.0383, + "step": 3226 + }, + { + "epoch": 0.21, + "learning_rate": 1.932909680328045e-05, + "loss": 1.1223, + "step": 3227 + }, + { + "epoch": 0.21, + "learning_rate": 1.93288821142575e-05, + "loss": 1.1207, + "step": 3228 + }, + { + "epoch": 0.21, + "learning_rate": 1.9328667425234548e-05, + "loss": 1.1229, + "step": 3229 + }, + { + "epoch": 0.21, + "learning_rate": 1.9328452736211597e-05, + "loss": 1.1486, + "step": 3230 + }, + { + "epoch": 0.21, + "learning_rate": 1.932823804718865e-05, + "loss": 1.2172, + "step": 3231 + }, + { + "epoch": 0.21, + "learning_rate": 1.9328023358165698e-05, + "loss": 1.0705, + "step": 3232 + }, + { + "epoch": 0.21, + "learning_rate": 1.932780866914275e-05, + "loss": 1.1981, + "step": 3233 + }, + { + "epoch": 0.21, + "learning_rate": 1.93275939801198e-05, + "loss": 1.1235, + "step": 3234 + }, + { + "epoch": 0.21, + "learning_rate": 1.9327379291096848e-05, + "loss": 1.135, + "step": 3235 + }, + { + "epoch": 0.21, + "learning_rate": 1.9327164602073897e-05, + "loss": 1.0948, + "step": 3236 + }, + { + "epoch": 0.21, + "learning_rate": 1.9326949913050946e-05, + "loss": 0.984, + "step": 3237 + }, + { + "epoch": 0.21, + "learning_rate": 1.9326735224028e-05, + "loss": 1.1947, + "step": 3238 + }, + { + "epoch": 0.21, + "learning_rate": 1.9326520535005047e-05, + "loss": 1.1107, + "step": 3239 + }, + { + "epoch": 0.21, + "learning_rate": 1.9326305845982096e-05, + "loss": 1.1454, + "step": 3240 + }, + { + "epoch": 0.21, + "learning_rate": 1.9326091156959145e-05, + "loss": 1.1478, + "step": 3241 + }, + { + "epoch": 0.21, + "learning_rate": 1.9325876467936194e-05, + "loss": 1.0866, + "step": 3242 + }, + { + "epoch": 0.21, + "learning_rate": 1.9325661778913243e-05, + "loss": 1.1821, + "step": 3243 + }, + { + "epoch": 0.21, + "learning_rate": 1.9325447089890296e-05, + "loss": 1.1627, + "step": 3244 + }, + { + "epoch": 0.21, + "learning_rate": 1.9325232400867345e-05, + "loss": 1.0974, + "step": 3245 + }, + { + "epoch": 0.21, + "learning_rate": 1.9325017711844397e-05, + "loss": 1.1703, + "step": 3246 + }, + { + "epoch": 0.21, + "learning_rate": 1.9324803022821446e-05, + "loss": 1.145, + "step": 3247 + }, + { + "epoch": 0.21, + "learning_rate": 1.9324588333798495e-05, + "loss": 1.0543, + "step": 3248 + }, + { + "epoch": 0.21, + "learning_rate": 1.9324373644775544e-05, + "loss": 1.1306, + "step": 3249 + }, + { + "epoch": 0.21, + "learning_rate": 1.9324158955752593e-05, + "loss": 1.2352, + "step": 3250 + }, + { + "epoch": 0.21, + "learning_rate": 1.9323944266729645e-05, + "loss": 1.2201, + "step": 3251 + }, + { + "epoch": 0.21, + "learning_rate": 1.9323729577706694e-05, + "loss": 1.1472, + "step": 3252 + }, + { + "epoch": 0.21, + "learning_rate": 1.9323514888683743e-05, + "loss": 1.0785, + "step": 3253 + }, + { + "epoch": 0.21, + "learning_rate": 1.9323300199660792e-05, + "loss": 1.2053, + "step": 3254 + }, + { + "epoch": 0.21, + "learning_rate": 1.932308551063784e-05, + "loss": 1.1868, + "step": 3255 + }, + { + "epoch": 0.21, + "learning_rate": 1.9322870821614893e-05, + "loss": 1.0349, + "step": 3256 + }, + { + "epoch": 0.21, + "learning_rate": 1.9322656132591942e-05, + "loss": 1.1369, + "step": 3257 + }, + { + "epoch": 0.21, + "learning_rate": 1.932244144356899e-05, + "loss": 1.1462, + "step": 3258 + }, + { + "epoch": 0.21, + "learning_rate": 1.932222675454604e-05, + "loss": 1.0847, + "step": 3259 + }, + { + "epoch": 0.21, + "learning_rate": 1.932201206552309e-05, + "loss": 1.0275, + "step": 3260 + }, + { + "epoch": 0.21, + "learning_rate": 1.932179737650014e-05, + "loss": 1.0729, + "step": 3261 + }, + { + "epoch": 0.21, + "learning_rate": 1.932158268747719e-05, + "loss": 1.1349, + "step": 3262 + }, + { + "epoch": 0.21, + "learning_rate": 1.932136799845424e-05, + "loss": 1.1589, + "step": 3263 + }, + { + "epoch": 0.21, + "learning_rate": 1.9321153309431292e-05, + "loss": 1.1286, + "step": 3264 + }, + { + "epoch": 0.21, + "learning_rate": 1.932093862040834e-05, + "loss": 1.1812, + "step": 3265 + }, + { + "epoch": 0.21, + "learning_rate": 1.932072393138539e-05, + "loss": 1.2189, + "step": 3266 + }, + { + "epoch": 0.21, + "learning_rate": 1.932050924236244e-05, + "loss": 1.1324, + "step": 3267 + }, + { + "epoch": 0.21, + "learning_rate": 1.9320294553339488e-05, + "loss": 1.1777, + "step": 3268 + }, + { + "epoch": 0.21, + "learning_rate": 1.932007986431654e-05, + "loss": 1.1547, + "step": 3269 + }, + { + "epoch": 0.21, + "learning_rate": 1.931986517529359e-05, + "loss": 1.0794, + "step": 3270 + }, + { + "epoch": 0.21, + "learning_rate": 1.9319650486270638e-05, + "loss": 1.1811, + "step": 3271 + }, + { + "epoch": 0.21, + "learning_rate": 1.9319435797247687e-05, + "loss": 1.1662, + "step": 3272 + }, + { + "epoch": 0.21, + "learning_rate": 1.9319221108224736e-05, + "loss": 0.9902, + "step": 3273 + }, + { + "epoch": 0.21, + "learning_rate": 1.931900641920179e-05, + "loss": 1.1637, + "step": 3274 + }, + { + "epoch": 0.21, + "learning_rate": 1.9318791730178837e-05, + "loss": 1.1597, + "step": 3275 + }, + { + "epoch": 0.21, + "learning_rate": 1.931857704115589e-05, + "loss": 1.082, + "step": 3276 + }, + { + "epoch": 0.21, + "learning_rate": 1.931836235213294e-05, + "loss": 1.1398, + "step": 3277 + }, + { + "epoch": 0.21, + "learning_rate": 1.9318147663109988e-05, + "loss": 1.0834, + "step": 3278 + }, + { + "epoch": 0.21, + "learning_rate": 1.9317932974087037e-05, + "loss": 1.1347, + "step": 3279 + }, + { + "epoch": 0.21, + "learning_rate": 1.9317718285064086e-05, + "loss": 1.1213, + "step": 3280 + }, + { + "epoch": 0.21, + "learning_rate": 1.9317503596041135e-05, + "loss": 1.2955, + "step": 3281 + }, + { + "epoch": 0.21, + "learning_rate": 1.9317288907018187e-05, + "loss": 1.0815, + "step": 3282 + }, + { + "epoch": 0.21, + "learning_rate": 1.9317074217995236e-05, + "loss": 1.1257, + "step": 3283 + }, + { + "epoch": 0.21, + "learning_rate": 1.9316859528972285e-05, + "loss": 1.0895, + "step": 3284 + }, + { + "epoch": 0.21, + "learning_rate": 1.9316644839949334e-05, + "loss": 1.1113, + "step": 3285 + }, + { + "epoch": 0.21, + "learning_rate": 1.9316430150926383e-05, + "loss": 1.124, + "step": 3286 + }, + { + "epoch": 0.21, + "learning_rate": 1.9316215461903435e-05, + "loss": 1.1623, + "step": 3287 + }, + { + "epoch": 0.21, + "learning_rate": 1.9316000772880484e-05, + "loss": 1.1367, + "step": 3288 + }, + { + "epoch": 0.21, + "learning_rate": 1.9315786083857533e-05, + "loss": 1.0763, + "step": 3289 + }, + { + "epoch": 0.21, + "learning_rate": 1.9315571394834585e-05, + "loss": 1.235, + "step": 3290 + }, + { + "epoch": 0.21, + "learning_rate": 1.9315356705811634e-05, + "loss": 1.0413, + "step": 3291 + }, + { + "epoch": 0.21, + "learning_rate": 1.9315142016788683e-05, + "loss": 1.0845, + "step": 3292 + }, + { + "epoch": 0.21, + "learning_rate": 1.9314927327765732e-05, + "loss": 1.0786, + "step": 3293 + }, + { + "epoch": 0.21, + "learning_rate": 1.9314712638742785e-05, + "loss": 1.1016, + "step": 3294 + }, + { + "epoch": 0.21, + "learning_rate": 1.9314497949719834e-05, + "loss": 1.1411, + "step": 3295 + }, + { + "epoch": 0.21, + "learning_rate": 1.9314283260696883e-05, + "loss": 1.1084, + "step": 3296 + }, + { + "epoch": 0.21, + "learning_rate": 1.931406857167393e-05, + "loss": 1.087, + "step": 3297 + }, + { + "epoch": 0.21, + "learning_rate": 1.931385388265098e-05, + "loss": 1.1092, + "step": 3298 + }, + { + "epoch": 0.21, + "learning_rate": 1.931363919362803e-05, + "loss": 1.1853, + "step": 3299 + }, + { + "epoch": 0.21, + "learning_rate": 1.9313424504605082e-05, + "loss": 1.1496, + "step": 3300 + }, + { + "epoch": 0.21, + "learning_rate": 1.931320981558213e-05, + "loss": 1.0902, + "step": 3301 + }, + { + "epoch": 0.21, + "learning_rate": 1.931299512655918e-05, + "loss": 1.1393, + "step": 3302 + }, + { + "epoch": 0.21, + "learning_rate": 1.931278043753623e-05, + "loss": 1.1516, + "step": 3303 + }, + { + "epoch": 0.21, + "learning_rate": 1.931256574851328e-05, + "loss": 1.1136, + "step": 3304 + }, + { + "epoch": 0.21, + "learning_rate": 1.931235105949033e-05, + "loss": 1.1322, + "step": 3305 + }, + { + "epoch": 0.21, + "learning_rate": 1.931213637046738e-05, + "loss": 1.1333, + "step": 3306 + }, + { + "epoch": 0.21, + "learning_rate": 1.931192168144443e-05, + "loss": 1.1548, + "step": 3307 + }, + { + "epoch": 0.21, + "learning_rate": 1.931170699242148e-05, + "loss": 1.1443, + "step": 3308 + }, + { + "epoch": 0.21, + "learning_rate": 1.931149230339853e-05, + "loss": 1.1817, + "step": 3309 + }, + { + "epoch": 0.21, + "learning_rate": 1.9311277614375578e-05, + "loss": 1.0868, + "step": 3310 + }, + { + "epoch": 0.21, + "learning_rate": 1.9311062925352627e-05, + "loss": 1.1163, + "step": 3311 + }, + { + "epoch": 0.21, + "learning_rate": 1.931084823632968e-05, + "loss": 1.2029, + "step": 3312 + }, + { + "epoch": 0.21, + "learning_rate": 1.931063354730673e-05, + "loss": 1.198, + "step": 3313 + }, + { + "epoch": 0.21, + "learning_rate": 1.9310418858283777e-05, + "loss": 1.1768, + "step": 3314 + }, + { + "epoch": 0.21, + "learning_rate": 1.9310204169260826e-05, + "loss": 1.1113, + "step": 3315 + }, + { + "epoch": 0.21, + "learning_rate": 1.9309989480237875e-05, + "loss": 1.3036, + "step": 3316 + }, + { + "epoch": 0.21, + "learning_rate": 1.9309774791214924e-05, + "loss": 1.1511, + "step": 3317 + }, + { + "epoch": 0.21, + "learning_rate": 1.9309560102191977e-05, + "loss": 1.1902, + "step": 3318 + }, + { + "epoch": 0.21, + "learning_rate": 1.9309345413169026e-05, + "loss": 1.1319, + "step": 3319 + }, + { + "epoch": 0.21, + "learning_rate": 1.9309130724146078e-05, + "loss": 1.1322, + "step": 3320 + }, + { + "epoch": 0.21, + "learning_rate": 1.9308916035123127e-05, + "loss": 1.104, + "step": 3321 + }, + { + "epoch": 0.21, + "learning_rate": 1.9308701346100176e-05, + "loss": 1.0099, + "step": 3322 + }, + { + "epoch": 0.21, + "learning_rate": 1.9308486657077225e-05, + "loss": 1.0937, + "step": 3323 + }, + { + "epoch": 0.21, + "learning_rate": 1.9308271968054274e-05, + "loss": 1.1687, + "step": 3324 + }, + { + "epoch": 0.21, + "learning_rate": 1.9308057279031326e-05, + "loss": 1.1695, + "step": 3325 + }, + { + "epoch": 0.21, + "learning_rate": 1.9307842590008375e-05, + "loss": 1.2567, + "step": 3326 + }, + { + "epoch": 0.21, + "learning_rate": 1.9307627900985424e-05, + "loss": 1.2074, + "step": 3327 + }, + { + "epoch": 0.21, + "learning_rate": 1.9307413211962473e-05, + "loss": 1.1069, + "step": 3328 + }, + { + "epoch": 0.21, + "learning_rate": 1.9307198522939522e-05, + "loss": 1.1664, + "step": 3329 + }, + { + "epoch": 0.21, + "learning_rate": 1.930698383391657e-05, + "loss": 1.1713, + "step": 3330 + }, + { + "epoch": 0.21, + "learning_rate": 1.9306769144893623e-05, + "loss": 1.1478, + "step": 3331 + }, + { + "epoch": 0.21, + "learning_rate": 1.9306554455870672e-05, + "loss": 1.1245, + "step": 3332 + }, + { + "epoch": 0.21, + "learning_rate": 1.9306339766847725e-05, + "loss": 1.1059, + "step": 3333 + }, + { + "epoch": 0.21, + "learning_rate": 1.9306125077824774e-05, + "loss": 1.1423, + "step": 3334 + }, + { + "epoch": 0.21, + "learning_rate": 1.9305910388801823e-05, + "loss": 1.2025, + "step": 3335 + }, + { + "epoch": 0.21, + "learning_rate": 1.930569569977887e-05, + "loss": 1.1492, + "step": 3336 + }, + { + "epoch": 0.21, + "learning_rate": 1.930548101075592e-05, + "loss": 1.1524, + "step": 3337 + }, + { + "epoch": 0.21, + "learning_rate": 1.9305266321732973e-05, + "loss": 1.0993, + "step": 3338 + }, + { + "epoch": 0.21, + "learning_rate": 1.9305051632710022e-05, + "loss": 1.254, + "step": 3339 + }, + { + "epoch": 0.21, + "learning_rate": 1.930483694368707e-05, + "loss": 1.1428, + "step": 3340 + }, + { + "epoch": 0.21, + "learning_rate": 1.930462225466412e-05, + "loss": 1.1282, + "step": 3341 + }, + { + "epoch": 0.22, + "learning_rate": 1.930440756564117e-05, + "loss": 1.1764, + "step": 3342 + }, + { + "epoch": 0.22, + "learning_rate": 1.930419287661822e-05, + "loss": 1.1346, + "step": 3343 + }, + { + "epoch": 0.22, + "learning_rate": 1.930397818759527e-05, + "loss": 1.0839, + "step": 3344 + }, + { + "epoch": 0.22, + "learning_rate": 1.930376349857232e-05, + "loss": 1.1168, + "step": 3345 + }, + { + "epoch": 0.22, + "learning_rate": 1.9303548809549368e-05, + "loss": 1.1442, + "step": 3346 + }, + { + "epoch": 0.22, + "learning_rate": 1.9303334120526417e-05, + "loss": 1.1504, + "step": 3347 + }, + { + "epoch": 0.22, + "learning_rate": 1.930311943150347e-05, + "loss": 1.1385, + "step": 3348 + }, + { + "epoch": 0.22, + "learning_rate": 1.930290474248052e-05, + "loss": 1.1921, + "step": 3349 + }, + { + "epoch": 0.22, + "learning_rate": 1.9302690053457567e-05, + "loss": 1.0153, + "step": 3350 + }, + { + "epoch": 0.22, + "learning_rate": 1.930247536443462e-05, + "loss": 1.0325, + "step": 3351 + }, + { + "epoch": 0.22, + "learning_rate": 1.930226067541167e-05, + "loss": 1.0843, + "step": 3352 + }, + { + "epoch": 0.22, + "learning_rate": 1.9302045986388718e-05, + "loss": 1.2156, + "step": 3353 + }, + { + "epoch": 0.22, + "learning_rate": 1.9301831297365767e-05, + "loss": 1.1314, + "step": 3354 + }, + { + "epoch": 0.22, + "learning_rate": 1.9301616608342816e-05, + "loss": 1.1688, + "step": 3355 + }, + { + "epoch": 0.22, + "learning_rate": 1.9301401919319868e-05, + "loss": 1.1031, + "step": 3356 + }, + { + "epoch": 0.22, + "learning_rate": 1.9301187230296917e-05, + "loss": 1.1055, + "step": 3357 + }, + { + "epoch": 0.22, + "learning_rate": 1.9300972541273966e-05, + "loss": 1.1622, + "step": 3358 + }, + { + "epoch": 0.22, + "learning_rate": 1.9300757852251015e-05, + "loss": 1.0689, + "step": 3359 + }, + { + "epoch": 0.22, + "learning_rate": 1.9300543163228064e-05, + "loss": 1.1008, + "step": 3360 + }, + { + "epoch": 0.22, + "learning_rate": 1.9300328474205116e-05, + "loss": 1.1078, + "step": 3361 + }, + { + "epoch": 0.22, + "learning_rate": 1.9300113785182165e-05, + "loss": 1.153, + "step": 3362 + }, + { + "epoch": 0.22, + "learning_rate": 1.9299899096159217e-05, + "loss": 1.1743, + "step": 3363 + }, + { + "epoch": 0.22, + "learning_rate": 1.9299684407136266e-05, + "loss": 1.1461, + "step": 3364 + }, + { + "epoch": 0.22, + "learning_rate": 1.9299469718113315e-05, + "loss": 1.1733, + "step": 3365 + }, + { + "epoch": 0.22, + "learning_rate": 1.9299255029090364e-05, + "loss": 1.1389, + "step": 3366 + }, + { + "epoch": 0.22, + "learning_rate": 1.9299040340067413e-05, + "loss": 1.1758, + "step": 3367 + }, + { + "epoch": 0.22, + "learning_rate": 1.9298825651044462e-05, + "loss": 1.0887, + "step": 3368 + }, + { + "epoch": 0.22, + "learning_rate": 1.9298610962021515e-05, + "loss": 1.1187, + "step": 3369 + }, + { + "epoch": 0.22, + "learning_rate": 1.9298396272998564e-05, + "loss": 1.2207, + "step": 3370 + }, + { + "epoch": 0.22, + "learning_rate": 1.9298181583975612e-05, + "loss": 1.1703, + "step": 3371 + }, + { + "epoch": 0.22, + "learning_rate": 1.929796689495266e-05, + "loss": 1.1721, + "step": 3372 + }, + { + "epoch": 0.22, + "learning_rate": 1.929775220592971e-05, + "loss": 1.1053, + "step": 3373 + }, + { + "epoch": 0.22, + "learning_rate": 1.9297537516906763e-05, + "loss": 1.1627, + "step": 3374 + }, + { + "epoch": 0.22, + "learning_rate": 1.9297322827883812e-05, + "loss": 1.1717, + "step": 3375 + }, + { + "epoch": 0.22, + "learning_rate": 1.929710813886086e-05, + "loss": 1.1231, + "step": 3376 + }, + { + "epoch": 0.22, + "learning_rate": 1.9296893449837913e-05, + "loss": 1.0821, + "step": 3377 + }, + { + "epoch": 0.22, + "learning_rate": 1.9296678760814962e-05, + "loss": 1.1754, + "step": 3378 + }, + { + "epoch": 0.22, + "learning_rate": 1.929646407179201e-05, + "loss": 1.1884, + "step": 3379 + }, + { + "epoch": 0.22, + "learning_rate": 1.929624938276906e-05, + "loss": 1.1161, + "step": 3380 + }, + { + "epoch": 0.22, + "learning_rate": 1.9296034693746112e-05, + "loss": 1.1475, + "step": 3381 + }, + { + "epoch": 0.22, + "learning_rate": 1.929582000472316e-05, + "loss": 1.1451, + "step": 3382 + }, + { + "epoch": 0.22, + "learning_rate": 1.929560531570021e-05, + "loss": 1.1654, + "step": 3383 + }, + { + "epoch": 0.22, + "learning_rate": 1.929539062667726e-05, + "loss": 1.2175, + "step": 3384 + }, + { + "epoch": 0.22, + "learning_rate": 1.9295175937654308e-05, + "loss": 1.0307, + "step": 3385 + }, + { + "epoch": 0.22, + "learning_rate": 1.9294961248631357e-05, + "loss": 1.1252, + "step": 3386 + }, + { + "epoch": 0.22, + "learning_rate": 1.929474655960841e-05, + "loss": 1.1875, + "step": 3387 + }, + { + "epoch": 0.22, + "learning_rate": 1.929453187058546e-05, + "loss": 1.152, + "step": 3388 + }, + { + "epoch": 0.22, + "learning_rate": 1.9294317181562507e-05, + "loss": 1.1457, + "step": 3389 + }, + { + "epoch": 0.22, + "learning_rate": 1.9294102492539556e-05, + "loss": 1.1185, + "step": 3390 + }, + { + "epoch": 0.22, + "learning_rate": 1.929388780351661e-05, + "loss": 1.2007, + "step": 3391 + }, + { + "epoch": 0.22, + "learning_rate": 1.9293673114493658e-05, + "loss": 1.077, + "step": 3392 + }, + { + "epoch": 0.22, + "learning_rate": 1.9293458425470707e-05, + "loss": 1.2241, + "step": 3393 + }, + { + "epoch": 0.22, + "learning_rate": 1.929324373644776e-05, + "loss": 1.1794, + "step": 3394 + }, + { + "epoch": 0.22, + "learning_rate": 1.9293029047424808e-05, + "loss": 1.1583, + "step": 3395 + }, + { + "epoch": 0.22, + "learning_rate": 1.9292814358401857e-05, + "loss": 1.2128, + "step": 3396 + }, + { + "epoch": 0.22, + "learning_rate": 1.9292599669378906e-05, + "loss": 1.1635, + "step": 3397 + }, + { + "epoch": 0.22, + "learning_rate": 1.9292384980355955e-05, + "loss": 1.208, + "step": 3398 + }, + { + "epoch": 0.22, + "learning_rate": 1.9292170291333007e-05, + "loss": 1.1681, + "step": 3399 + }, + { + "epoch": 0.22, + "learning_rate": 1.9291955602310056e-05, + "loss": 1.1513, + "step": 3400 + }, + { + "epoch": 0.22, + "learning_rate": 1.9291740913287105e-05, + "loss": 1.1978, + "step": 3401 + }, + { + "epoch": 0.22, + "learning_rate": 1.9291526224264154e-05, + "loss": 1.0927, + "step": 3402 + }, + { + "epoch": 0.22, + "learning_rate": 1.9291311535241203e-05, + "loss": 1.0446, + "step": 3403 + }, + { + "epoch": 0.22, + "learning_rate": 1.9291096846218252e-05, + "loss": 1.1518, + "step": 3404 + }, + { + "epoch": 0.22, + "learning_rate": 1.9290882157195304e-05, + "loss": 1.1005, + "step": 3405 + }, + { + "epoch": 0.22, + "learning_rate": 1.9290667468172353e-05, + "loss": 1.2343, + "step": 3406 + }, + { + "epoch": 0.22, + "learning_rate": 1.9290452779149406e-05, + "loss": 1.1631, + "step": 3407 + }, + { + "epoch": 0.22, + "learning_rate": 1.9290238090126455e-05, + "loss": 1.1434, + "step": 3408 + }, + { + "epoch": 0.22, + "learning_rate": 1.9290023401103504e-05, + "loss": 1.1283, + "step": 3409 + }, + { + "epoch": 0.22, + "learning_rate": 1.9289808712080553e-05, + "loss": 1.0601, + "step": 3410 + }, + { + "epoch": 0.22, + "learning_rate": 1.92895940230576e-05, + "loss": 1.1587, + "step": 3411 + }, + { + "epoch": 0.22, + "learning_rate": 1.9289379334034654e-05, + "loss": 1.1014, + "step": 3412 + }, + { + "epoch": 0.22, + "learning_rate": 1.9289164645011703e-05, + "loss": 1.1733, + "step": 3413 + }, + { + "epoch": 0.22, + "learning_rate": 1.9288949955988752e-05, + "loss": 1.1922, + "step": 3414 + }, + { + "epoch": 0.22, + "learning_rate": 1.92887352669658e-05, + "loss": 1.1, + "step": 3415 + }, + { + "epoch": 0.22, + "learning_rate": 1.928852057794285e-05, + "loss": 1.1226, + "step": 3416 + }, + { + "epoch": 0.22, + "learning_rate": 1.92883058889199e-05, + "loss": 1.0614, + "step": 3417 + }, + { + "epoch": 0.22, + "learning_rate": 1.928809119989695e-05, + "loss": 1.1327, + "step": 3418 + }, + { + "epoch": 0.22, + "learning_rate": 1.9287876510874e-05, + "loss": 1.1985, + "step": 3419 + }, + { + "epoch": 0.22, + "learning_rate": 1.9287661821851052e-05, + "loss": 1.2019, + "step": 3420 + }, + { + "epoch": 0.22, + "learning_rate": 1.92874471328281e-05, + "loss": 1.2018, + "step": 3421 + }, + { + "epoch": 0.22, + "learning_rate": 1.928723244380515e-05, + "loss": 1.1391, + "step": 3422 + }, + { + "epoch": 0.22, + "learning_rate": 1.92870177547822e-05, + "loss": 1.1807, + "step": 3423 + }, + { + "epoch": 0.22, + "learning_rate": 1.9286803065759248e-05, + "loss": 1.1544, + "step": 3424 + }, + { + "epoch": 0.22, + "learning_rate": 1.92865883767363e-05, + "loss": 1.1002, + "step": 3425 + }, + { + "epoch": 0.22, + "learning_rate": 1.928637368771335e-05, + "loss": 1.0104, + "step": 3426 + }, + { + "epoch": 0.22, + "learning_rate": 1.92861589986904e-05, + "loss": 1.0805, + "step": 3427 + }, + { + "epoch": 0.22, + "learning_rate": 1.9285944309667448e-05, + "loss": 1.1179, + "step": 3428 + }, + { + "epoch": 0.22, + "learning_rate": 1.9285729620644497e-05, + "loss": 1.1618, + "step": 3429 + }, + { + "epoch": 0.22, + "learning_rate": 1.928551493162155e-05, + "loss": 1.0944, + "step": 3430 + }, + { + "epoch": 0.22, + "learning_rate": 1.9285300242598598e-05, + "loss": 1.1147, + "step": 3431 + }, + { + "epoch": 0.22, + "learning_rate": 1.9285085553575647e-05, + "loss": 1.1131, + "step": 3432 + }, + { + "epoch": 0.22, + "learning_rate": 1.9284870864552696e-05, + "loss": 1.1406, + "step": 3433 + }, + { + "epoch": 0.22, + "learning_rate": 1.9284656175529745e-05, + "loss": 1.1889, + "step": 3434 + }, + { + "epoch": 0.22, + "learning_rate": 1.9284441486506797e-05, + "loss": 1.0945, + "step": 3435 + }, + { + "epoch": 0.22, + "learning_rate": 1.9284226797483846e-05, + "loss": 1.1676, + "step": 3436 + }, + { + "epoch": 0.22, + "learning_rate": 1.9284012108460895e-05, + "loss": 1.1899, + "step": 3437 + }, + { + "epoch": 0.22, + "learning_rate": 1.9283797419437947e-05, + "loss": 1.1832, + "step": 3438 + }, + { + "epoch": 0.22, + "learning_rate": 1.9283582730414996e-05, + "loss": 1.1836, + "step": 3439 + }, + { + "epoch": 0.22, + "learning_rate": 1.9283368041392045e-05, + "loss": 1.1677, + "step": 3440 + }, + { + "epoch": 0.22, + "learning_rate": 1.9283153352369094e-05, + "loss": 1.1372, + "step": 3441 + }, + { + "epoch": 0.22, + "learning_rate": 1.9282938663346143e-05, + "loss": 1.0381, + "step": 3442 + }, + { + "epoch": 0.22, + "learning_rate": 1.9282723974323196e-05, + "loss": 1.1342, + "step": 3443 + }, + { + "epoch": 0.22, + "learning_rate": 1.9282509285300245e-05, + "loss": 1.1352, + "step": 3444 + }, + { + "epoch": 0.22, + "learning_rate": 1.9282294596277293e-05, + "loss": 1.1029, + "step": 3445 + }, + { + "epoch": 0.22, + "learning_rate": 1.9282079907254342e-05, + "loss": 1.1373, + "step": 3446 + }, + { + "epoch": 0.22, + "learning_rate": 1.928186521823139e-05, + "loss": 1.2008, + "step": 3447 + }, + { + "epoch": 0.22, + "learning_rate": 1.9281650529208444e-05, + "loss": 1.117, + "step": 3448 + }, + { + "epoch": 0.22, + "learning_rate": 1.9281435840185493e-05, + "loss": 1.1442, + "step": 3449 + }, + { + "epoch": 0.22, + "learning_rate": 1.9281221151162545e-05, + "loss": 1.0231, + "step": 3450 + }, + { + "epoch": 0.22, + "learning_rate": 1.9281006462139594e-05, + "loss": 1.1472, + "step": 3451 + }, + { + "epoch": 0.22, + "learning_rate": 1.9280791773116643e-05, + "loss": 1.1633, + "step": 3452 + }, + { + "epoch": 0.22, + "learning_rate": 1.9280577084093692e-05, + "loss": 1.2747, + "step": 3453 + }, + { + "epoch": 0.22, + "learning_rate": 1.928036239507074e-05, + "loss": 1.1766, + "step": 3454 + }, + { + "epoch": 0.22, + "learning_rate": 1.928014770604779e-05, + "loss": 1.1444, + "step": 3455 + }, + { + "epoch": 0.22, + "learning_rate": 1.9279933017024842e-05, + "loss": 1.1306, + "step": 3456 + }, + { + "epoch": 0.22, + "learning_rate": 1.927971832800189e-05, + "loss": 1.1116, + "step": 3457 + }, + { + "epoch": 0.22, + "learning_rate": 1.927950363897894e-05, + "loss": 1.2026, + "step": 3458 + }, + { + "epoch": 0.22, + "learning_rate": 1.927928894995599e-05, + "loss": 1.1434, + "step": 3459 + }, + { + "epoch": 0.22, + "learning_rate": 1.9279074260933038e-05, + "loss": 1.1752, + "step": 3460 + }, + { + "epoch": 0.22, + "learning_rate": 1.927885957191009e-05, + "loss": 1.2176, + "step": 3461 + }, + { + "epoch": 0.22, + "learning_rate": 1.927864488288714e-05, + "loss": 1.0668, + "step": 3462 + }, + { + "epoch": 0.22, + "learning_rate": 1.927843019386419e-05, + "loss": 1.1565, + "step": 3463 + }, + { + "epoch": 0.22, + "learning_rate": 1.927821550484124e-05, + "loss": 1.1377, + "step": 3464 + }, + { + "epoch": 0.22, + "learning_rate": 1.927800081581829e-05, + "loss": 1.1057, + "step": 3465 + }, + { + "epoch": 0.22, + "learning_rate": 1.927778612679534e-05, + "loss": 1.1072, + "step": 3466 + }, + { + "epoch": 0.22, + "learning_rate": 1.9277571437772388e-05, + "loss": 1.1333, + "step": 3467 + }, + { + "epoch": 0.22, + "learning_rate": 1.927735674874944e-05, + "loss": 1.168, + "step": 3468 + }, + { + "epoch": 0.22, + "learning_rate": 1.927714205972649e-05, + "loss": 1.1187, + "step": 3469 + }, + { + "epoch": 0.22, + "learning_rate": 1.9276927370703538e-05, + "loss": 1.128, + "step": 3470 + }, + { + "epoch": 0.22, + "learning_rate": 1.9276712681680587e-05, + "loss": 1.1604, + "step": 3471 + }, + { + "epoch": 0.22, + "learning_rate": 1.9276497992657636e-05, + "loss": 1.0315, + "step": 3472 + }, + { + "epoch": 0.22, + "learning_rate": 1.9276283303634685e-05, + "loss": 1.2129, + "step": 3473 + }, + { + "epoch": 0.22, + "learning_rate": 1.9276068614611737e-05, + "loss": 1.1118, + "step": 3474 + }, + { + "epoch": 0.22, + "learning_rate": 1.9275853925588786e-05, + "loss": 1.1576, + "step": 3475 + }, + { + "epoch": 0.22, + "learning_rate": 1.9275639236565835e-05, + "loss": 1.1236, + "step": 3476 + }, + { + "epoch": 0.22, + "learning_rate": 1.9275424547542884e-05, + "loss": 1.2262, + "step": 3477 + }, + { + "epoch": 0.22, + "learning_rate": 1.9275209858519936e-05, + "loss": 1.1924, + "step": 3478 + }, + { + "epoch": 0.22, + "learning_rate": 1.9274995169496985e-05, + "loss": 1.2209, + "step": 3479 + }, + { + "epoch": 0.22, + "learning_rate": 1.9274780480474034e-05, + "loss": 1.086, + "step": 3480 + }, + { + "epoch": 0.22, + "learning_rate": 1.9274565791451087e-05, + "loss": 1.1475, + "step": 3481 + }, + { + "epoch": 0.22, + "learning_rate": 1.9274351102428136e-05, + "loss": 1.2604, + "step": 3482 + }, + { + "epoch": 0.22, + "learning_rate": 1.9274136413405185e-05, + "loss": 1.1146, + "step": 3483 + }, + { + "epoch": 0.22, + "learning_rate": 1.9273921724382234e-05, + "loss": 1.0827, + "step": 3484 + }, + { + "epoch": 0.22, + "learning_rate": 1.9273707035359283e-05, + "loss": 1.1078, + "step": 3485 + }, + { + "epoch": 0.22, + "learning_rate": 1.9273492346336335e-05, + "loss": 1.0854, + "step": 3486 + }, + { + "epoch": 0.22, + "learning_rate": 1.9273277657313384e-05, + "loss": 1.2167, + "step": 3487 + }, + { + "epoch": 0.22, + "learning_rate": 1.9273062968290433e-05, + "loss": 1.1315, + "step": 3488 + }, + { + "epoch": 0.22, + "learning_rate": 1.9272848279267482e-05, + "loss": 1.0875, + "step": 3489 + }, + { + "epoch": 0.22, + "learning_rate": 1.927263359024453e-05, + "loss": 1.1005, + "step": 3490 + }, + { + "epoch": 0.22, + "learning_rate": 1.927241890122158e-05, + "loss": 1.186, + "step": 3491 + }, + { + "epoch": 0.22, + "learning_rate": 1.9272204212198632e-05, + "loss": 1.2191, + "step": 3492 + }, + { + "epoch": 0.22, + "learning_rate": 1.927198952317568e-05, + "loss": 1.2162, + "step": 3493 + }, + { + "epoch": 0.22, + "learning_rate": 1.9271774834152733e-05, + "loss": 1.1451, + "step": 3494 + }, + { + "epoch": 0.22, + "learning_rate": 1.9271560145129782e-05, + "loss": 1.1447, + "step": 3495 + }, + { + "epoch": 0.22, + "learning_rate": 1.927134545610683e-05, + "loss": 1.0448, + "step": 3496 + }, + { + "epoch": 0.22, + "learning_rate": 1.927113076708388e-05, + "loss": 1.1604, + "step": 3497 + }, + { + "epoch": 0.23, + "learning_rate": 1.927091607806093e-05, + "loss": 1.0585, + "step": 3498 + }, + { + "epoch": 0.23, + "learning_rate": 1.927070138903798e-05, + "loss": 1.16, + "step": 3499 + }, + { + "epoch": 0.23, + "learning_rate": 1.927048670001503e-05, + "loss": 1.1236, + "step": 3500 + }, + { + "epoch": 0.23, + "learning_rate": 1.927027201099208e-05, + "loss": 1.1185, + "step": 3501 + }, + { + "epoch": 0.23, + "learning_rate": 1.927005732196913e-05, + "loss": 1.1362, + "step": 3502 + }, + { + "epoch": 0.23, + "learning_rate": 1.9269842632946178e-05, + "loss": 1.1657, + "step": 3503 + }, + { + "epoch": 0.23, + "learning_rate": 1.9269627943923226e-05, + "loss": 1.2051, + "step": 3504 + }, + { + "epoch": 0.23, + "learning_rate": 1.926941325490028e-05, + "loss": 1.1852, + "step": 3505 + }, + { + "epoch": 0.23, + "learning_rate": 1.9269198565877328e-05, + "loss": 0.9656, + "step": 3506 + }, + { + "epoch": 0.23, + "learning_rate": 1.926898387685438e-05, + "loss": 1.1965, + "step": 3507 + }, + { + "epoch": 0.23, + "learning_rate": 1.926876918783143e-05, + "loss": 1.0709, + "step": 3508 + }, + { + "epoch": 0.23, + "learning_rate": 1.9268554498808478e-05, + "loss": 1.1417, + "step": 3509 + }, + { + "epoch": 0.23, + "learning_rate": 1.9268339809785527e-05, + "loss": 1.0398, + "step": 3510 + }, + { + "epoch": 0.23, + "learning_rate": 1.9268125120762576e-05, + "loss": 1.2032, + "step": 3511 + }, + { + "epoch": 0.23, + "learning_rate": 1.926791043173963e-05, + "loss": 1.2082, + "step": 3512 + }, + { + "epoch": 0.23, + "learning_rate": 1.9267695742716677e-05, + "loss": 1.1748, + "step": 3513 + }, + { + "epoch": 0.23, + "learning_rate": 1.9267481053693726e-05, + "loss": 1.1307, + "step": 3514 + }, + { + "epoch": 0.23, + "learning_rate": 1.9267266364670775e-05, + "loss": 1.1527, + "step": 3515 + }, + { + "epoch": 0.23, + "learning_rate": 1.9267051675647824e-05, + "loss": 1.123, + "step": 3516 + }, + { + "epoch": 0.23, + "learning_rate": 1.9266836986624877e-05, + "loss": 1.2264, + "step": 3517 + }, + { + "epoch": 0.23, + "learning_rate": 1.9266622297601926e-05, + "loss": 1.1029, + "step": 3518 + }, + { + "epoch": 0.23, + "learning_rate": 1.9266407608578975e-05, + "loss": 1.1455, + "step": 3519 + }, + { + "epoch": 0.23, + "learning_rate": 1.9266192919556023e-05, + "loss": 1.1304, + "step": 3520 + }, + { + "epoch": 0.23, + "learning_rate": 1.9265978230533072e-05, + "loss": 1.1738, + "step": 3521 + }, + { + "epoch": 0.23, + "learning_rate": 1.9265763541510125e-05, + "loss": 1.0805, + "step": 3522 + }, + { + "epoch": 0.23, + "learning_rate": 1.9265548852487174e-05, + "loss": 1.0898, + "step": 3523 + }, + { + "epoch": 0.23, + "learning_rate": 1.9265334163464223e-05, + "loss": 1.0912, + "step": 3524 + }, + { + "epoch": 0.23, + "learning_rate": 1.9265119474441275e-05, + "loss": 1.1131, + "step": 3525 + }, + { + "epoch": 0.23, + "learning_rate": 1.9264904785418324e-05, + "loss": 1.1104, + "step": 3526 + }, + { + "epoch": 0.23, + "learning_rate": 1.9264690096395373e-05, + "loss": 1.1873, + "step": 3527 + }, + { + "epoch": 0.23, + "learning_rate": 1.9264475407372422e-05, + "loss": 1.0925, + "step": 3528 + }, + { + "epoch": 0.23, + "learning_rate": 1.926426071834947e-05, + "loss": 1.3153, + "step": 3529 + }, + { + "epoch": 0.23, + "learning_rate": 1.9264046029326523e-05, + "loss": 1.1881, + "step": 3530 + }, + { + "epoch": 0.23, + "learning_rate": 1.9263831340303572e-05, + "loss": 1.1524, + "step": 3531 + }, + { + "epoch": 0.23, + "learning_rate": 1.926361665128062e-05, + "loss": 1.0295, + "step": 3532 + }, + { + "epoch": 0.23, + "learning_rate": 1.926340196225767e-05, + "loss": 1.183, + "step": 3533 + }, + { + "epoch": 0.23, + "learning_rate": 1.926318727323472e-05, + "loss": 1.0757, + "step": 3534 + }, + { + "epoch": 0.23, + "learning_rate": 1.926297258421177e-05, + "loss": 1.117, + "step": 3535 + }, + { + "epoch": 0.23, + "learning_rate": 1.926275789518882e-05, + "loss": 1.1288, + "step": 3536 + }, + { + "epoch": 0.23, + "learning_rate": 1.9262543206165873e-05, + "loss": 1.1505, + "step": 3537 + }, + { + "epoch": 0.23, + "learning_rate": 1.9262328517142922e-05, + "loss": 1.0993, + "step": 3538 + }, + { + "epoch": 0.23, + "learning_rate": 1.926211382811997e-05, + "loss": 1.0827, + "step": 3539 + }, + { + "epoch": 0.23, + "learning_rate": 1.926189913909702e-05, + "loss": 1.1172, + "step": 3540 + }, + { + "epoch": 0.23, + "learning_rate": 1.926168445007407e-05, + "loss": 1.2081, + "step": 3541 + }, + { + "epoch": 0.23, + "learning_rate": 1.9261469761051118e-05, + "loss": 1.0881, + "step": 3542 + }, + { + "epoch": 0.23, + "learning_rate": 1.926125507202817e-05, + "loss": 1.0836, + "step": 3543 + }, + { + "epoch": 0.23, + "learning_rate": 1.926104038300522e-05, + "loss": 1.1534, + "step": 3544 + }, + { + "epoch": 0.23, + "learning_rate": 1.9260825693982268e-05, + "loss": 1.0929, + "step": 3545 + }, + { + "epoch": 0.23, + "learning_rate": 1.9260611004959317e-05, + "loss": 1.1426, + "step": 3546 + }, + { + "epoch": 0.23, + "learning_rate": 1.9260396315936366e-05, + "loss": 1.1194, + "step": 3547 + }, + { + "epoch": 0.23, + "learning_rate": 1.9260181626913418e-05, + "loss": 1.1524, + "step": 3548 + }, + { + "epoch": 0.23, + "learning_rate": 1.9259966937890467e-05, + "loss": 1.1692, + "step": 3549 + }, + { + "epoch": 0.23, + "learning_rate": 1.9259752248867516e-05, + "loss": 1.0843, + "step": 3550 + }, + { + "epoch": 0.23, + "learning_rate": 1.925953755984457e-05, + "loss": 0.999, + "step": 3551 + }, + { + "epoch": 0.23, + "learning_rate": 1.9259322870821617e-05, + "loss": 1.1307, + "step": 3552 + }, + { + "epoch": 0.23, + "learning_rate": 1.9259108181798666e-05, + "loss": 1.1949, + "step": 3553 + }, + { + "epoch": 0.23, + "learning_rate": 1.9258893492775715e-05, + "loss": 1.0689, + "step": 3554 + }, + { + "epoch": 0.23, + "learning_rate": 1.9258678803752768e-05, + "loss": 1.0772, + "step": 3555 + }, + { + "epoch": 0.23, + "learning_rate": 1.9258464114729817e-05, + "loss": 1.0529, + "step": 3556 + }, + { + "epoch": 0.23, + "learning_rate": 1.9258249425706866e-05, + "loss": 1.1471, + "step": 3557 + }, + { + "epoch": 0.23, + "learning_rate": 1.9258034736683915e-05, + "loss": 1.1855, + "step": 3558 + }, + { + "epoch": 0.23, + "learning_rate": 1.9257820047660964e-05, + "loss": 1.0345, + "step": 3559 + }, + { + "epoch": 0.23, + "learning_rate": 1.9257605358638013e-05, + "loss": 1.0014, + "step": 3560 + }, + { + "epoch": 0.23, + "learning_rate": 1.9257390669615065e-05, + "loss": 1.0689, + "step": 3561 + }, + { + "epoch": 0.23, + "learning_rate": 1.9257175980592114e-05, + "loss": 1.1325, + "step": 3562 + }, + { + "epoch": 0.23, + "learning_rate": 1.9256961291569163e-05, + "loss": 1.1706, + "step": 3563 + }, + { + "epoch": 0.23, + "learning_rate": 1.9256746602546212e-05, + "loss": 1.1406, + "step": 3564 + }, + { + "epoch": 0.23, + "learning_rate": 1.9256531913523264e-05, + "loss": 1.2003, + "step": 3565 + }, + { + "epoch": 0.23, + "learning_rate": 1.9256317224500313e-05, + "loss": 1.1304, + "step": 3566 + }, + { + "epoch": 0.23, + "learning_rate": 1.9256102535477362e-05, + "loss": 1.2106, + "step": 3567 + }, + { + "epoch": 0.23, + "learning_rate": 1.9255887846454414e-05, + "loss": 1.2879, + "step": 3568 + }, + { + "epoch": 0.23, + "learning_rate": 1.9255673157431463e-05, + "loss": 1.0869, + "step": 3569 + }, + { + "epoch": 0.23, + "learning_rate": 1.9255458468408512e-05, + "loss": 1.1532, + "step": 3570 + }, + { + "epoch": 0.23, + "learning_rate": 1.925524377938556e-05, + "loss": 1.0969, + "step": 3571 + }, + { + "epoch": 0.23, + "learning_rate": 1.925502909036261e-05, + "loss": 1.1539, + "step": 3572 + }, + { + "epoch": 0.23, + "learning_rate": 1.925481440133966e-05, + "loss": 1.0499, + "step": 3573 + }, + { + "epoch": 0.23, + "learning_rate": 1.925459971231671e-05, + "loss": 1.1813, + "step": 3574 + }, + { + "epoch": 0.23, + "learning_rate": 1.925438502329376e-05, + "loss": 1.0857, + "step": 3575 + }, + { + "epoch": 0.23, + "learning_rate": 1.925417033427081e-05, + "loss": 1.0553, + "step": 3576 + }, + { + "epoch": 0.23, + "learning_rate": 1.925395564524786e-05, + "loss": 1.1687, + "step": 3577 + }, + { + "epoch": 0.23, + "learning_rate": 1.9253740956224907e-05, + "loss": 1.0661, + "step": 3578 + }, + { + "epoch": 0.23, + "learning_rate": 1.925352626720196e-05, + "loss": 1.1149, + "step": 3579 + }, + { + "epoch": 0.23, + "learning_rate": 1.925331157817901e-05, + "loss": 1.1165, + "step": 3580 + }, + { + "epoch": 0.23, + "learning_rate": 1.925309688915606e-05, + "loss": 1.1577, + "step": 3581 + }, + { + "epoch": 0.23, + "learning_rate": 1.925288220013311e-05, + "loss": 1.1214, + "step": 3582 + }, + { + "epoch": 0.23, + "learning_rate": 1.925266751111016e-05, + "loss": 1.2316, + "step": 3583 + }, + { + "epoch": 0.23, + "learning_rate": 1.9252452822087208e-05, + "loss": 1.1607, + "step": 3584 + }, + { + "epoch": 0.23, + "learning_rate": 1.9252238133064257e-05, + "loss": 1.0783, + "step": 3585 + }, + { + "epoch": 0.23, + "learning_rate": 1.925202344404131e-05, + "loss": 1.16, + "step": 3586 + }, + { + "epoch": 0.23, + "learning_rate": 1.925180875501836e-05, + "loss": 1.0434, + "step": 3587 + }, + { + "epoch": 0.23, + "learning_rate": 1.9251594065995407e-05, + "loss": 1.0901, + "step": 3588 + }, + { + "epoch": 0.23, + "learning_rate": 1.9251379376972456e-05, + "loss": 1.0997, + "step": 3589 + }, + { + "epoch": 0.23, + "learning_rate": 1.9251164687949505e-05, + "loss": 1.0954, + "step": 3590 + }, + { + "epoch": 0.23, + "learning_rate": 1.9250949998926554e-05, + "loss": 1.1091, + "step": 3591 + }, + { + "epoch": 0.23, + "learning_rate": 1.9250735309903607e-05, + "loss": 1.1139, + "step": 3592 + }, + { + "epoch": 0.23, + "learning_rate": 1.9250520620880656e-05, + "loss": 1.1027, + "step": 3593 + }, + { + "epoch": 0.23, + "learning_rate": 1.9250305931857708e-05, + "loss": 1.1138, + "step": 3594 + }, + { + "epoch": 0.23, + "learning_rate": 1.9250091242834757e-05, + "loss": 1.0895, + "step": 3595 + }, + { + "epoch": 0.23, + "learning_rate": 1.9249876553811806e-05, + "loss": 1.088, + "step": 3596 + }, + { + "epoch": 0.23, + "learning_rate": 1.9249661864788855e-05, + "loss": 1.1457, + "step": 3597 + }, + { + "epoch": 0.23, + "learning_rate": 1.9249447175765904e-05, + "loss": 1.1157, + "step": 3598 + }, + { + "epoch": 0.23, + "learning_rate": 1.9249232486742956e-05, + "loss": 1.1887, + "step": 3599 + }, + { + "epoch": 0.23, + "learning_rate": 1.9249017797720005e-05, + "loss": 1.2286, + "step": 3600 + }, + { + "epoch": 0.23, + "learning_rate": 1.9248803108697054e-05, + "loss": 1.118, + "step": 3601 + }, + { + "epoch": 0.23, + "learning_rate": 1.9248588419674103e-05, + "loss": 1.1453, + "step": 3602 + }, + { + "epoch": 0.23, + "learning_rate": 1.9248373730651152e-05, + "loss": 1.2026, + "step": 3603 + }, + { + "epoch": 0.23, + "learning_rate": 1.9248159041628204e-05, + "loss": 1.1813, + "step": 3604 + }, + { + "epoch": 0.23, + "learning_rate": 1.9247944352605253e-05, + "loss": 1.098, + "step": 3605 + }, + { + "epoch": 0.23, + "learning_rate": 1.9247729663582302e-05, + "loss": 1.1632, + "step": 3606 + }, + { + "epoch": 0.23, + "learning_rate": 1.924751497455935e-05, + "loss": 1.2474, + "step": 3607 + }, + { + "epoch": 0.23, + "learning_rate": 1.92473002855364e-05, + "loss": 1.1555, + "step": 3608 + }, + { + "epoch": 0.23, + "learning_rate": 1.9247085596513452e-05, + "loss": 1.1633, + "step": 3609 + }, + { + "epoch": 0.23, + "learning_rate": 1.92468709074905e-05, + "loss": 1.1456, + "step": 3610 + }, + { + "epoch": 0.23, + "learning_rate": 1.924665621846755e-05, + "loss": 1.1225, + "step": 3611 + }, + { + "epoch": 0.23, + "learning_rate": 1.9246441529444603e-05, + "loss": 1.1625, + "step": 3612 + }, + { + "epoch": 0.23, + "learning_rate": 1.9246226840421652e-05, + "loss": 1.0667, + "step": 3613 + }, + { + "epoch": 0.23, + "learning_rate": 1.92460121513987e-05, + "loss": 1.2063, + "step": 3614 + }, + { + "epoch": 0.23, + "learning_rate": 1.924579746237575e-05, + "loss": 1.1362, + "step": 3615 + }, + { + "epoch": 0.23, + "learning_rate": 1.92455827733528e-05, + "loss": 1.1625, + "step": 3616 + }, + { + "epoch": 0.23, + "learning_rate": 1.924536808432985e-05, + "loss": 1.1196, + "step": 3617 + }, + { + "epoch": 0.23, + "learning_rate": 1.92451533953069e-05, + "loss": 1.0874, + "step": 3618 + }, + { + "epoch": 0.23, + "learning_rate": 1.924493870628395e-05, + "loss": 1.1939, + "step": 3619 + }, + { + "epoch": 0.23, + "learning_rate": 1.9244724017260998e-05, + "loss": 1.1904, + "step": 3620 + }, + { + "epoch": 0.23, + "learning_rate": 1.9244509328238047e-05, + "loss": 1.0739, + "step": 3621 + }, + { + "epoch": 0.23, + "learning_rate": 1.92442946392151e-05, + "loss": 1.1156, + "step": 3622 + }, + { + "epoch": 0.23, + "learning_rate": 1.9244079950192148e-05, + "loss": 1.1939, + "step": 3623 + }, + { + "epoch": 0.23, + "learning_rate": 1.92438652611692e-05, + "loss": 1.0351, + "step": 3624 + }, + { + "epoch": 0.23, + "learning_rate": 1.924365057214625e-05, + "loss": 1.2452, + "step": 3625 + }, + { + "epoch": 0.23, + "learning_rate": 1.92434358831233e-05, + "loss": 1.1576, + "step": 3626 + }, + { + "epoch": 0.23, + "learning_rate": 1.9243221194100347e-05, + "loss": 1.1438, + "step": 3627 + }, + { + "epoch": 0.23, + "learning_rate": 1.9243006505077396e-05, + "loss": 1.0828, + "step": 3628 + }, + { + "epoch": 0.23, + "learning_rate": 1.9242791816054445e-05, + "loss": 1.1281, + "step": 3629 + }, + { + "epoch": 0.23, + "learning_rate": 1.9242577127031498e-05, + "loss": 1.0755, + "step": 3630 + }, + { + "epoch": 0.23, + "learning_rate": 1.9242362438008547e-05, + "loss": 1.2308, + "step": 3631 + }, + { + "epoch": 0.23, + "learning_rate": 1.9242147748985596e-05, + "loss": 1.0757, + "step": 3632 + }, + { + "epoch": 0.23, + "learning_rate": 1.9241933059962645e-05, + "loss": 1.1858, + "step": 3633 + }, + { + "epoch": 0.23, + "learning_rate": 1.9241718370939694e-05, + "loss": 1.2437, + "step": 3634 + }, + { + "epoch": 0.23, + "learning_rate": 1.9241503681916746e-05, + "loss": 1.1597, + "step": 3635 + }, + { + "epoch": 0.23, + "learning_rate": 1.9241288992893795e-05, + "loss": 1.0551, + "step": 3636 + }, + { + "epoch": 0.23, + "learning_rate": 1.9241074303870844e-05, + "loss": 1.111, + "step": 3637 + }, + { + "epoch": 0.23, + "learning_rate": 1.9240859614847896e-05, + "loss": 1.0775, + "step": 3638 + }, + { + "epoch": 0.23, + "learning_rate": 1.9240644925824945e-05, + "loss": 1.121, + "step": 3639 + }, + { + "epoch": 0.23, + "learning_rate": 1.9240430236801994e-05, + "loss": 1.1494, + "step": 3640 + }, + { + "epoch": 0.23, + "learning_rate": 1.9240215547779043e-05, + "loss": 1.1344, + "step": 3641 + }, + { + "epoch": 0.23, + "learning_rate": 1.9240000858756095e-05, + "loss": 1.101, + "step": 3642 + }, + { + "epoch": 0.23, + "learning_rate": 1.9239786169733144e-05, + "loss": 1.1576, + "step": 3643 + }, + { + "epoch": 0.23, + "learning_rate": 1.9239571480710193e-05, + "loss": 1.0974, + "step": 3644 + }, + { + "epoch": 0.23, + "learning_rate": 1.9239356791687242e-05, + "loss": 1.1985, + "step": 3645 + }, + { + "epoch": 0.23, + "learning_rate": 1.923914210266429e-05, + "loss": 1.1518, + "step": 3646 + }, + { + "epoch": 0.23, + "learning_rate": 1.923892741364134e-05, + "loss": 1.1052, + "step": 3647 + }, + { + "epoch": 0.23, + "learning_rate": 1.9238712724618393e-05, + "loss": 1.0928, + "step": 3648 + }, + { + "epoch": 0.23, + "learning_rate": 1.923849803559544e-05, + "loss": 1.1761, + "step": 3649 + }, + { + "epoch": 0.23, + "learning_rate": 1.923828334657249e-05, + "loss": 1.1312, + "step": 3650 + }, + { + "epoch": 0.23, + "learning_rate": 1.923806865754954e-05, + "loss": 1.1202, + "step": 3651 + }, + { + "epoch": 0.23, + "learning_rate": 1.923785396852659e-05, + "loss": 1.1405, + "step": 3652 + }, + { + "epoch": 0.24, + "learning_rate": 1.923763927950364e-05, + "loss": 1.1054, + "step": 3653 + }, + { + "epoch": 0.24, + "learning_rate": 1.923742459048069e-05, + "loss": 1.0863, + "step": 3654 + }, + { + "epoch": 0.24, + "learning_rate": 1.9237209901457742e-05, + "loss": 1.1384, + "step": 3655 + }, + { + "epoch": 0.24, + "learning_rate": 1.923699521243479e-05, + "loss": 1.0926, + "step": 3656 + }, + { + "epoch": 0.24, + "learning_rate": 1.923678052341184e-05, + "loss": 1.1722, + "step": 3657 + }, + { + "epoch": 0.24, + "learning_rate": 1.923656583438889e-05, + "loss": 1.1476, + "step": 3658 + }, + { + "epoch": 0.24, + "learning_rate": 1.9236351145365938e-05, + "loss": 1.1377, + "step": 3659 + }, + { + "epoch": 0.24, + "learning_rate": 1.9236136456342987e-05, + "loss": 1.0915, + "step": 3660 + }, + { + "epoch": 0.24, + "learning_rate": 1.923592176732004e-05, + "loss": 1.1257, + "step": 3661 + }, + { + "epoch": 0.24, + "learning_rate": 1.9235707078297088e-05, + "loss": 1.0839, + "step": 3662 + }, + { + "epoch": 0.24, + "learning_rate": 1.9235492389274137e-05, + "loss": 1.158, + "step": 3663 + }, + { + "epoch": 0.24, + "learning_rate": 1.9235277700251186e-05, + "loss": 1.1006, + "step": 3664 + }, + { + "epoch": 0.24, + "learning_rate": 1.9235063011228235e-05, + "loss": 1.1576, + "step": 3665 + }, + { + "epoch": 0.24, + "learning_rate": 1.9234848322205288e-05, + "loss": 1.0649, + "step": 3666 + }, + { + "epoch": 0.24, + "learning_rate": 1.9234633633182337e-05, + "loss": 1.1752, + "step": 3667 + }, + { + "epoch": 0.24, + "learning_rate": 1.923441894415939e-05, + "loss": 1.1581, + "step": 3668 + }, + { + "epoch": 0.24, + "learning_rate": 1.9234204255136438e-05, + "loss": 1.1871, + "step": 3669 + }, + { + "epoch": 0.24, + "learning_rate": 1.9233989566113487e-05, + "loss": 1.0921, + "step": 3670 + }, + { + "epoch": 0.24, + "learning_rate": 1.9233774877090536e-05, + "loss": 1.1776, + "step": 3671 + }, + { + "epoch": 0.24, + "learning_rate": 1.9233560188067585e-05, + "loss": 1.0424, + "step": 3672 + }, + { + "epoch": 0.24, + "learning_rate": 1.9233345499044637e-05, + "loss": 1.097, + "step": 3673 + }, + { + "epoch": 0.24, + "learning_rate": 1.9233130810021686e-05, + "loss": 1.0827, + "step": 3674 + }, + { + "epoch": 0.24, + "learning_rate": 1.9232916120998735e-05, + "loss": 1.038, + "step": 3675 + }, + { + "epoch": 0.24, + "learning_rate": 1.9232701431975784e-05, + "loss": 1.0512, + "step": 3676 + }, + { + "epoch": 0.24, + "learning_rate": 1.9232486742952833e-05, + "loss": 1.1525, + "step": 3677 + }, + { + "epoch": 0.24, + "learning_rate": 1.9232272053929882e-05, + "loss": 1.0554, + "step": 3678 + }, + { + "epoch": 0.24, + "learning_rate": 1.9232057364906934e-05, + "loss": 1.1135, + "step": 3679 + }, + { + "epoch": 0.24, + "learning_rate": 1.9231842675883983e-05, + "loss": 1.1394, + "step": 3680 + }, + { + "epoch": 0.24, + "learning_rate": 1.9231627986861032e-05, + "loss": 1.1732, + "step": 3681 + }, + { + "epoch": 0.24, + "learning_rate": 1.9231413297838085e-05, + "loss": 1.1088, + "step": 3682 + }, + { + "epoch": 0.24, + "learning_rate": 1.9231198608815134e-05, + "loss": 1.1363, + "step": 3683 + }, + { + "epoch": 0.24, + "learning_rate": 1.9230983919792182e-05, + "loss": 1.1205, + "step": 3684 + }, + { + "epoch": 0.24, + "learning_rate": 1.923076923076923e-05, + "loss": 0.9837, + "step": 3685 + }, + { + "epoch": 0.24, + "learning_rate": 1.9230554541746284e-05, + "loss": 1.1164, + "step": 3686 + }, + { + "epoch": 0.24, + "learning_rate": 1.9230339852723333e-05, + "loss": 1.233, + "step": 3687 + }, + { + "epoch": 0.24, + "learning_rate": 1.9230125163700382e-05, + "loss": 1.1722, + "step": 3688 + }, + { + "epoch": 0.24, + "learning_rate": 1.922991047467743e-05, + "loss": 1.0924, + "step": 3689 + }, + { + "epoch": 0.24, + "learning_rate": 1.922969578565448e-05, + "loss": 1.0956, + "step": 3690 + }, + { + "epoch": 0.24, + "learning_rate": 1.9229481096631532e-05, + "loss": 1.1694, + "step": 3691 + }, + { + "epoch": 0.24, + "learning_rate": 1.922926640760858e-05, + "loss": 1.1336, + "step": 3692 + }, + { + "epoch": 0.24, + "learning_rate": 1.922905171858563e-05, + "loss": 1.1706, + "step": 3693 + }, + { + "epoch": 0.24, + "learning_rate": 1.922883702956268e-05, + "loss": 1.1656, + "step": 3694 + }, + { + "epoch": 0.24, + "learning_rate": 1.9228622340539728e-05, + "loss": 1.2346, + "step": 3695 + }, + { + "epoch": 0.24, + "learning_rate": 1.922840765151678e-05, + "loss": 1.0399, + "step": 3696 + }, + { + "epoch": 0.24, + "learning_rate": 1.922819296249383e-05, + "loss": 1.1295, + "step": 3697 + }, + { + "epoch": 0.24, + "learning_rate": 1.9227978273470878e-05, + "loss": 1.1063, + "step": 3698 + }, + { + "epoch": 0.24, + "learning_rate": 1.922776358444793e-05, + "loss": 1.0817, + "step": 3699 + }, + { + "epoch": 0.24, + "learning_rate": 1.922754889542498e-05, + "loss": 1.2031, + "step": 3700 + }, + { + "epoch": 0.24, + "learning_rate": 1.922733420640203e-05, + "loss": 1.1531, + "step": 3701 + }, + { + "epoch": 0.24, + "learning_rate": 1.9227119517379077e-05, + "loss": 1.1678, + "step": 3702 + }, + { + "epoch": 0.24, + "learning_rate": 1.9226904828356126e-05, + "loss": 1.1077, + "step": 3703 + }, + { + "epoch": 0.24, + "learning_rate": 1.922669013933318e-05, + "loss": 1.048, + "step": 3704 + }, + { + "epoch": 0.24, + "learning_rate": 1.9226475450310228e-05, + "loss": 1.1227, + "step": 3705 + }, + { + "epoch": 0.24, + "learning_rate": 1.9226260761287277e-05, + "loss": 1.0917, + "step": 3706 + }, + { + "epoch": 0.24, + "learning_rate": 1.9226046072264326e-05, + "loss": 1.1255, + "step": 3707 + }, + { + "epoch": 0.24, + "learning_rate": 1.9225831383241375e-05, + "loss": 1.1009, + "step": 3708 + }, + { + "epoch": 0.24, + "learning_rate": 1.9225616694218427e-05, + "loss": 1.1245, + "step": 3709 + }, + { + "epoch": 0.24, + "learning_rate": 1.9225402005195476e-05, + "loss": 1.217, + "step": 3710 + }, + { + "epoch": 0.24, + "learning_rate": 1.9225187316172528e-05, + "loss": 1.1115, + "step": 3711 + }, + { + "epoch": 0.24, + "learning_rate": 1.9224972627149577e-05, + "loss": 1.1303, + "step": 3712 + }, + { + "epoch": 0.24, + "learning_rate": 1.9224757938126626e-05, + "loss": 1.1602, + "step": 3713 + }, + { + "epoch": 0.24, + "learning_rate": 1.9224543249103675e-05, + "loss": 1.1728, + "step": 3714 + }, + { + "epoch": 0.24, + "learning_rate": 1.9224328560080724e-05, + "loss": 1.0782, + "step": 3715 + }, + { + "epoch": 0.24, + "learning_rate": 1.9224113871057773e-05, + "loss": 1.1665, + "step": 3716 + }, + { + "epoch": 0.24, + "learning_rate": 1.9223899182034825e-05, + "loss": 1.0802, + "step": 3717 + }, + { + "epoch": 0.24, + "learning_rate": 1.9223684493011874e-05, + "loss": 1.1204, + "step": 3718 + }, + { + "epoch": 0.24, + "learning_rate": 1.9223469803988923e-05, + "loss": 1.21, + "step": 3719 + }, + { + "epoch": 0.24, + "learning_rate": 1.9223255114965972e-05, + "loss": 1.1589, + "step": 3720 + }, + { + "epoch": 0.24, + "learning_rate": 1.922304042594302e-05, + "loss": 1.0958, + "step": 3721 + }, + { + "epoch": 0.24, + "learning_rate": 1.9222825736920074e-05, + "loss": 1.2212, + "step": 3722 + }, + { + "epoch": 0.24, + "learning_rate": 1.9222611047897123e-05, + "loss": 1.0624, + "step": 3723 + }, + { + "epoch": 0.24, + "learning_rate": 1.922239635887417e-05, + "loss": 1.1231, + "step": 3724 + }, + { + "epoch": 0.24, + "learning_rate": 1.9222181669851224e-05, + "loss": 1.0908, + "step": 3725 + }, + { + "epoch": 0.24, + "learning_rate": 1.9221966980828273e-05, + "loss": 1.2636, + "step": 3726 + }, + { + "epoch": 0.24, + "learning_rate": 1.9221752291805322e-05, + "loss": 1.2079, + "step": 3727 + }, + { + "epoch": 0.24, + "learning_rate": 1.922153760278237e-05, + "loss": 1.1476, + "step": 3728 + }, + { + "epoch": 0.24, + "learning_rate": 1.9221322913759423e-05, + "loss": 1.1365, + "step": 3729 + }, + { + "epoch": 0.24, + "learning_rate": 1.9221108224736472e-05, + "loss": 1.1571, + "step": 3730 + }, + { + "epoch": 0.24, + "learning_rate": 1.922089353571352e-05, + "loss": 1.0776, + "step": 3731 + }, + { + "epoch": 0.24, + "learning_rate": 1.922067884669057e-05, + "loss": 1.1236, + "step": 3732 + }, + { + "epoch": 0.24, + "learning_rate": 1.922046415766762e-05, + "loss": 1.1252, + "step": 3733 + }, + { + "epoch": 0.24, + "learning_rate": 1.9220249468644668e-05, + "loss": 1.0967, + "step": 3734 + }, + { + "epoch": 0.24, + "learning_rate": 1.922003477962172e-05, + "loss": 1.206, + "step": 3735 + }, + { + "epoch": 0.24, + "learning_rate": 1.921982009059877e-05, + "loss": 1.0783, + "step": 3736 + }, + { + "epoch": 0.24, + "learning_rate": 1.9219605401575818e-05, + "loss": 1.1654, + "step": 3737 + }, + { + "epoch": 0.24, + "learning_rate": 1.9219390712552867e-05, + "loss": 1.1203, + "step": 3738 + }, + { + "epoch": 0.24, + "learning_rate": 1.9219176023529916e-05, + "loss": 1.17, + "step": 3739 + }, + { + "epoch": 0.24, + "learning_rate": 1.921896133450697e-05, + "loss": 1.1159, + "step": 3740 + }, + { + "epoch": 0.24, + "learning_rate": 1.9218746645484018e-05, + "loss": 1.221, + "step": 3741 + }, + { + "epoch": 0.24, + "learning_rate": 1.921853195646107e-05, + "loss": 1.0719, + "step": 3742 + }, + { + "epoch": 0.24, + "learning_rate": 1.921831726743812e-05, + "loss": 1.101, + "step": 3743 + }, + { + "epoch": 0.24, + "learning_rate": 1.9218102578415168e-05, + "loss": 1.1288, + "step": 3744 + }, + { + "epoch": 0.24, + "learning_rate": 1.9217887889392217e-05, + "loss": 1.0883, + "step": 3745 + }, + { + "epoch": 0.24, + "learning_rate": 1.9217673200369266e-05, + "loss": 1.1281, + "step": 3746 + }, + { + "epoch": 0.24, + "learning_rate": 1.9217458511346315e-05, + "loss": 1.0942, + "step": 3747 + }, + { + "epoch": 0.24, + "learning_rate": 1.9217243822323367e-05, + "loss": 1.1575, + "step": 3748 + }, + { + "epoch": 0.24, + "learning_rate": 1.9217029133300416e-05, + "loss": 1.1393, + "step": 3749 + }, + { + "epoch": 0.24, + "learning_rate": 1.9216814444277465e-05, + "loss": 1.0662, + "step": 3750 + }, + { + "epoch": 0.24, + "learning_rate": 1.9216599755254514e-05, + "loss": 1.0949, + "step": 3751 + }, + { + "epoch": 0.24, + "learning_rate": 1.9216385066231563e-05, + "loss": 1.0532, + "step": 3752 + }, + { + "epoch": 0.24, + "learning_rate": 1.9216170377208615e-05, + "loss": 1.2743, + "step": 3753 + }, + { + "epoch": 0.24, + "learning_rate": 1.9215955688185664e-05, + "loss": 1.1082, + "step": 3754 + }, + { + "epoch": 0.24, + "learning_rate": 1.9215740999162717e-05, + "loss": 1.0942, + "step": 3755 + }, + { + "epoch": 0.24, + "learning_rate": 1.9215526310139766e-05, + "loss": 1.1074, + "step": 3756 + }, + { + "epoch": 0.24, + "learning_rate": 1.9215311621116815e-05, + "loss": 1.1633, + "step": 3757 + }, + { + "epoch": 0.24, + "learning_rate": 1.9215096932093863e-05, + "loss": 1.1018, + "step": 3758 + }, + { + "epoch": 0.24, + "learning_rate": 1.9214882243070912e-05, + "loss": 1.1744, + "step": 3759 + }, + { + "epoch": 0.24, + "learning_rate": 1.9214667554047965e-05, + "loss": 1.0827, + "step": 3760 + }, + { + "epoch": 0.24, + "learning_rate": 1.9214452865025014e-05, + "loss": 1.1226, + "step": 3761 + }, + { + "epoch": 0.24, + "learning_rate": 1.9214238176002063e-05, + "loss": 1.1424, + "step": 3762 + }, + { + "epoch": 0.24, + "learning_rate": 1.921402348697911e-05, + "loss": 1.1561, + "step": 3763 + }, + { + "epoch": 0.24, + "learning_rate": 1.921380879795616e-05, + "loss": 1.0993, + "step": 3764 + }, + { + "epoch": 0.24, + "learning_rate": 1.921359410893321e-05, + "loss": 1.221, + "step": 3765 + }, + { + "epoch": 0.24, + "learning_rate": 1.9213379419910262e-05, + "loss": 1.1069, + "step": 3766 + }, + { + "epoch": 0.24, + "learning_rate": 1.921316473088731e-05, + "loss": 1.1436, + "step": 3767 + }, + { + "epoch": 0.24, + "learning_rate": 1.921295004186436e-05, + "loss": 1.1197, + "step": 3768 + }, + { + "epoch": 0.24, + "learning_rate": 1.9212735352841412e-05, + "loss": 1.1233, + "step": 3769 + }, + { + "epoch": 0.24, + "learning_rate": 1.921252066381846e-05, + "loss": 1.0625, + "step": 3770 + }, + { + "epoch": 0.24, + "learning_rate": 1.921230597479551e-05, + "loss": 1.0705, + "step": 3771 + }, + { + "epoch": 0.24, + "learning_rate": 1.921209128577256e-05, + "loss": 1.1565, + "step": 3772 + }, + { + "epoch": 0.24, + "learning_rate": 1.921187659674961e-05, + "loss": 1.1901, + "step": 3773 + }, + { + "epoch": 0.24, + "learning_rate": 1.921166190772666e-05, + "loss": 1.1861, + "step": 3774 + }, + { + "epoch": 0.24, + "learning_rate": 1.921144721870371e-05, + "loss": 1.0816, + "step": 3775 + }, + { + "epoch": 0.24, + "learning_rate": 1.921123252968076e-05, + "loss": 1.1335, + "step": 3776 + }, + { + "epoch": 0.24, + "learning_rate": 1.9211017840657807e-05, + "loss": 1.1722, + "step": 3777 + }, + { + "epoch": 0.24, + "learning_rate": 1.921080315163486e-05, + "loss": 1.188, + "step": 3778 + }, + { + "epoch": 0.24, + "learning_rate": 1.921058846261191e-05, + "loss": 1.1876, + "step": 3779 + }, + { + "epoch": 0.24, + "learning_rate": 1.9210373773588958e-05, + "loss": 1.1743, + "step": 3780 + }, + { + "epoch": 0.24, + "learning_rate": 1.9210159084566007e-05, + "loss": 1.2217, + "step": 3781 + }, + { + "epoch": 0.24, + "learning_rate": 1.9209944395543056e-05, + "loss": 1.178, + "step": 3782 + }, + { + "epoch": 0.24, + "learning_rate": 1.9209729706520108e-05, + "loss": 1.1456, + "step": 3783 + }, + { + "epoch": 0.24, + "learning_rate": 1.9209515017497157e-05, + "loss": 1.1021, + "step": 3784 + }, + { + "epoch": 0.24, + "learning_rate": 1.9209300328474206e-05, + "loss": 1.1545, + "step": 3785 + }, + { + "epoch": 0.24, + "learning_rate": 1.9209085639451258e-05, + "loss": 1.1861, + "step": 3786 + }, + { + "epoch": 0.24, + "learning_rate": 1.9208870950428307e-05, + "loss": 1.1609, + "step": 3787 + }, + { + "epoch": 0.24, + "learning_rate": 1.9208656261405356e-05, + "loss": 1.1573, + "step": 3788 + }, + { + "epoch": 0.24, + "learning_rate": 1.9208441572382405e-05, + "loss": 1.1726, + "step": 3789 + }, + { + "epoch": 0.24, + "learning_rate": 1.9208226883359454e-05, + "loss": 1.1052, + "step": 3790 + }, + { + "epoch": 0.24, + "learning_rate": 1.9208012194336506e-05, + "loss": 1.0393, + "step": 3791 + }, + { + "epoch": 0.24, + "learning_rate": 1.9207797505313555e-05, + "loss": 1.1131, + "step": 3792 + }, + { + "epoch": 0.24, + "learning_rate": 1.9207582816290604e-05, + "loss": 1.1219, + "step": 3793 + }, + { + "epoch": 0.24, + "learning_rate": 1.9207368127267653e-05, + "loss": 1.1496, + "step": 3794 + }, + { + "epoch": 0.24, + "learning_rate": 1.9207153438244702e-05, + "loss": 1.1809, + "step": 3795 + }, + { + "epoch": 0.24, + "learning_rate": 1.9206938749221755e-05, + "loss": 1.1434, + "step": 3796 + }, + { + "epoch": 0.24, + "learning_rate": 1.9206724060198804e-05, + "loss": 1.0915, + "step": 3797 + }, + { + "epoch": 0.24, + "learning_rate": 1.9206509371175856e-05, + "loss": 1.1838, + "step": 3798 + }, + { + "epoch": 0.24, + "learning_rate": 1.9206294682152905e-05, + "loss": 1.0546, + "step": 3799 + }, + { + "epoch": 0.24, + "learning_rate": 1.9206079993129954e-05, + "loss": 1.1598, + "step": 3800 + }, + { + "epoch": 0.24, + "learning_rate": 1.9205865304107003e-05, + "loss": 1.0957, + "step": 3801 + }, + { + "epoch": 0.24, + "learning_rate": 1.9205650615084052e-05, + "loss": 1.1161, + "step": 3802 + }, + { + "epoch": 0.24, + "learning_rate": 1.92054359260611e-05, + "loss": 1.1242, + "step": 3803 + }, + { + "epoch": 0.24, + "learning_rate": 1.9205221237038153e-05, + "loss": 1.1481, + "step": 3804 + }, + { + "epoch": 0.24, + "learning_rate": 1.9205006548015202e-05, + "loss": 1.1421, + "step": 3805 + }, + { + "epoch": 0.24, + "learning_rate": 1.920479185899225e-05, + "loss": 1.1127, + "step": 3806 + }, + { + "epoch": 0.24, + "learning_rate": 1.92045771699693e-05, + "loss": 1.1822, + "step": 3807 + }, + { + "epoch": 0.24, + "learning_rate": 1.920436248094635e-05, + "loss": 1.0999, + "step": 3808 + }, + { + "epoch": 0.25, + "learning_rate": 1.92041477919234e-05, + "loss": 1.134, + "step": 3809 + }, + { + "epoch": 0.25, + "learning_rate": 1.920393310290045e-05, + "loss": 1.1456, + "step": 3810 + }, + { + "epoch": 0.25, + "learning_rate": 1.92037184138775e-05, + "loss": 1.1539, + "step": 3811 + }, + { + "epoch": 0.25, + "learning_rate": 1.920350372485455e-05, + "loss": 1.1121, + "step": 3812 + }, + { + "epoch": 0.25, + "learning_rate": 1.92032890358316e-05, + "loss": 1.0705, + "step": 3813 + }, + { + "epoch": 0.25, + "learning_rate": 1.920307434680865e-05, + "loss": 1.0769, + "step": 3814 + }, + { + "epoch": 0.25, + "learning_rate": 1.92028596577857e-05, + "loss": 1.0963, + "step": 3815 + }, + { + "epoch": 0.25, + "learning_rate": 1.920264496876275e-05, + "loss": 1.0415, + "step": 3816 + }, + { + "epoch": 0.25, + "learning_rate": 1.92024302797398e-05, + "loss": 1.0152, + "step": 3817 + }, + { + "epoch": 0.25, + "learning_rate": 1.920221559071685e-05, + "loss": 1.1363, + "step": 3818 + }, + { + "epoch": 0.25, + "learning_rate": 1.9202000901693898e-05, + "loss": 1.1914, + "step": 3819 + }, + { + "epoch": 0.25, + "learning_rate": 1.9201786212670947e-05, + "loss": 1.1419, + "step": 3820 + }, + { + "epoch": 0.25, + "learning_rate": 1.9201571523647996e-05, + "loss": 1.2064, + "step": 3821 + }, + { + "epoch": 0.25, + "learning_rate": 1.9201356834625048e-05, + "loss": 1.1501, + "step": 3822 + }, + { + "epoch": 0.25, + "learning_rate": 1.9201142145602097e-05, + "loss": 1.0814, + "step": 3823 + }, + { + "epoch": 0.25, + "learning_rate": 1.9200927456579146e-05, + "loss": 1.0608, + "step": 3824 + }, + { + "epoch": 0.25, + "learning_rate": 1.9200712767556195e-05, + "loss": 1.1651, + "step": 3825 + }, + { + "epoch": 0.25, + "learning_rate": 1.9200498078533244e-05, + "loss": 1.1011, + "step": 3826 + }, + { + "epoch": 0.25, + "learning_rate": 1.9200283389510296e-05, + "loss": 1.0373, + "step": 3827 + }, + { + "epoch": 0.25, + "learning_rate": 1.9200068700487345e-05, + "loss": 1.2064, + "step": 3828 + }, + { + "epoch": 0.25, + "learning_rate": 1.9199854011464398e-05, + "loss": 1.0803, + "step": 3829 + }, + { + "epoch": 0.25, + "learning_rate": 1.9199639322441447e-05, + "loss": 1.2081, + "step": 3830 + }, + { + "epoch": 0.25, + "learning_rate": 1.9199424633418496e-05, + "loss": 1.134, + "step": 3831 + }, + { + "epoch": 0.25, + "learning_rate": 1.9199209944395544e-05, + "loss": 1.0084, + "step": 3832 + }, + { + "epoch": 0.25, + "learning_rate": 1.9198995255372593e-05, + "loss": 1.1674, + "step": 3833 + }, + { + "epoch": 0.25, + "learning_rate": 1.9198780566349642e-05, + "loss": 1.049, + "step": 3834 + }, + { + "epoch": 0.25, + "learning_rate": 1.9198565877326695e-05, + "loss": 1.1651, + "step": 3835 + }, + { + "epoch": 0.25, + "learning_rate": 1.9198351188303744e-05, + "loss": 1.0368, + "step": 3836 + }, + { + "epoch": 0.25, + "learning_rate": 1.9198136499280793e-05, + "loss": 1.1802, + "step": 3837 + }, + { + "epoch": 0.25, + "learning_rate": 1.919792181025784e-05, + "loss": 1.1492, + "step": 3838 + }, + { + "epoch": 0.25, + "learning_rate": 1.919770712123489e-05, + "loss": 1.139, + "step": 3839 + }, + { + "epoch": 0.25, + "learning_rate": 1.9197492432211943e-05, + "loss": 1.1384, + "step": 3840 + }, + { + "epoch": 0.25, + "learning_rate": 1.9197277743188992e-05, + "loss": 1.0212, + "step": 3841 + }, + { + "epoch": 0.25, + "learning_rate": 1.9197063054166044e-05, + "loss": 1.2359, + "step": 3842 + }, + { + "epoch": 0.25, + "learning_rate": 1.9196848365143093e-05, + "loss": 1.2216, + "step": 3843 + }, + { + "epoch": 0.25, + "learning_rate": 1.9196633676120142e-05, + "loss": 1.1687, + "step": 3844 + }, + { + "epoch": 0.25, + "learning_rate": 1.919641898709719e-05, + "loss": 1.1515, + "step": 3845 + }, + { + "epoch": 0.25, + "learning_rate": 1.919620429807424e-05, + "loss": 1.1137, + "step": 3846 + }, + { + "epoch": 0.25, + "learning_rate": 1.9195989609051292e-05, + "loss": 1.1634, + "step": 3847 + }, + { + "epoch": 0.25, + "learning_rate": 1.919577492002834e-05, + "loss": 1.1251, + "step": 3848 + }, + { + "epoch": 0.25, + "learning_rate": 1.919556023100539e-05, + "loss": 1.139, + "step": 3849 + }, + { + "epoch": 0.25, + "learning_rate": 1.919534554198244e-05, + "loss": 1.101, + "step": 3850 + }, + { + "epoch": 0.25, + "learning_rate": 1.919513085295949e-05, + "loss": 1.086, + "step": 3851 + }, + { + "epoch": 0.25, + "learning_rate": 1.9194916163936537e-05, + "loss": 1.2182, + "step": 3852 + }, + { + "epoch": 0.25, + "learning_rate": 1.919470147491359e-05, + "loss": 1.1804, + "step": 3853 + }, + { + "epoch": 0.25, + "learning_rate": 1.919448678589064e-05, + "loss": 1.048, + "step": 3854 + }, + { + "epoch": 0.25, + "learning_rate": 1.9194272096867688e-05, + "loss": 1.1337, + "step": 3855 + }, + { + "epoch": 0.25, + "learning_rate": 1.919405740784474e-05, + "loss": 1.1448, + "step": 3856 + }, + { + "epoch": 0.25, + "learning_rate": 1.919384271882179e-05, + "loss": 1.2157, + "step": 3857 + }, + { + "epoch": 0.25, + "learning_rate": 1.9193628029798838e-05, + "loss": 1.1472, + "step": 3858 + }, + { + "epoch": 0.25, + "learning_rate": 1.9193413340775887e-05, + "loss": 1.1247, + "step": 3859 + }, + { + "epoch": 0.25, + "learning_rate": 1.919319865175294e-05, + "loss": 1.1652, + "step": 3860 + }, + { + "epoch": 0.25, + "learning_rate": 1.9192983962729988e-05, + "loss": 1.2641, + "step": 3861 + }, + { + "epoch": 0.25, + "learning_rate": 1.9192769273707037e-05, + "loss": 1.0305, + "step": 3862 + }, + { + "epoch": 0.25, + "learning_rate": 1.9192554584684086e-05, + "loss": 1.121, + "step": 3863 + }, + { + "epoch": 0.25, + "learning_rate": 1.9192339895661135e-05, + "loss": 1.0727, + "step": 3864 + }, + { + "epoch": 0.25, + "learning_rate": 1.9192125206638187e-05, + "loss": 1.1684, + "step": 3865 + }, + { + "epoch": 0.25, + "learning_rate": 1.9191910517615236e-05, + "loss": 1.2357, + "step": 3866 + }, + { + "epoch": 0.25, + "learning_rate": 1.9191695828592285e-05, + "loss": 1.0713, + "step": 3867 + }, + { + "epoch": 0.25, + "learning_rate": 1.9191481139569334e-05, + "loss": 1.1543, + "step": 3868 + }, + { + "epoch": 0.25, + "learning_rate": 1.9191266450546383e-05, + "loss": 1.0749, + "step": 3869 + }, + { + "epoch": 0.25, + "learning_rate": 1.9191051761523436e-05, + "loss": 1.2176, + "step": 3870 + }, + { + "epoch": 0.25, + "learning_rate": 1.9190837072500485e-05, + "loss": 1.0598, + "step": 3871 + }, + { + "epoch": 0.25, + "learning_rate": 1.9190622383477534e-05, + "loss": 1.0415, + "step": 3872 + }, + { + "epoch": 0.25, + "learning_rate": 1.9190407694454586e-05, + "loss": 1.1422, + "step": 3873 + }, + { + "epoch": 0.25, + "learning_rate": 1.9190193005431635e-05, + "loss": 1.1759, + "step": 3874 + }, + { + "epoch": 0.25, + "learning_rate": 1.9189978316408684e-05, + "loss": 1.0817, + "step": 3875 + }, + { + "epoch": 0.25, + "learning_rate": 1.9189763627385733e-05, + "loss": 1.1389, + "step": 3876 + }, + { + "epoch": 0.25, + "learning_rate": 1.9189548938362782e-05, + "loss": 1.0661, + "step": 3877 + }, + { + "epoch": 0.25, + "learning_rate": 1.9189334249339834e-05, + "loss": 1.1719, + "step": 3878 + }, + { + "epoch": 0.25, + "learning_rate": 1.9189119560316883e-05, + "loss": 1.0707, + "step": 3879 + }, + { + "epoch": 0.25, + "learning_rate": 1.9188904871293932e-05, + "loss": 1.1414, + "step": 3880 + }, + { + "epoch": 0.25, + "learning_rate": 1.918869018227098e-05, + "loss": 1.1059, + "step": 3881 + }, + { + "epoch": 0.25, + "learning_rate": 1.918847549324803e-05, + "loss": 1.1401, + "step": 3882 + }, + { + "epoch": 0.25, + "learning_rate": 1.918826080422508e-05, + "loss": 1.1141, + "step": 3883 + }, + { + "epoch": 0.25, + "learning_rate": 1.918804611520213e-05, + "loss": 1.1623, + "step": 3884 + }, + { + "epoch": 0.25, + "learning_rate": 1.9187831426179184e-05, + "loss": 1.0985, + "step": 3885 + }, + { + "epoch": 0.25, + "learning_rate": 1.9187616737156233e-05, + "loss": 1.1149, + "step": 3886 + }, + { + "epoch": 0.25, + "learning_rate": 1.918740204813328e-05, + "loss": 1.1548, + "step": 3887 + }, + { + "epoch": 0.25, + "learning_rate": 1.918718735911033e-05, + "loss": 1.131, + "step": 3888 + }, + { + "epoch": 0.25, + "learning_rate": 1.918697267008738e-05, + "loss": 1.13, + "step": 3889 + }, + { + "epoch": 0.25, + "learning_rate": 1.918675798106443e-05, + "loss": 1.1764, + "step": 3890 + }, + { + "epoch": 0.25, + "learning_rate": 1.918654329204148e-05, + "loss": 1.0623, + "step": 3891 + }, + { + "epoch": 0.25, + "learning_rate": 1.918632860301853e-05, + "loss": 1.061, + "step": 3892 + }, + { + "epoch": 0.25, + "learning_rate": 1.918611391399558e-05, + "loss": 1.2174, + "step": 3893 + }, + { + "epoch": 0.25, + "learning_rate": 1.9185899224972628e-05, + "loss": 1.1315, + "step": 3894 + }, + { + "epoch": 0.25, + "learning_rate": 1.9185684535949677e-05, + "loss": 1.1642, + "step": 3895 + }, + { + "epoch": 0.25, + "learning_rate": 1.918546984692673e-05, + "loss": 1.1891, + "step": 3896 + }, + { + "epoch": 0.25, + "learning_rate": 1.9185255157903778e-05, + "loss": 1.1714, + "step": 3897 + }, + { + "epoch": 0.25, + "learning_rate": 1.9185040468880827e-05, + "loss": 1.077, + "step": 3898 + }, + { + "epoch": 0.25, + "learning_rate": 1.918482577985788e-05, + "loss": 1.1098, + "step": 3899 + }, + { + "epoch": 0.25, + "learning_rate": 1.9184611090834928e-05, + "loss": 1.1286, + "step": 3900 + }, + { + "epoch": 0.25, + "learning_rate": 1.9184396401811977e-05, + "loss": 1.1593, + "step": 3901 + }, + { + "epoch": 0.25, + "learning_rate": 1.9184181712789026e-05, + "loss": 1.1126, + "step": 3902 + }, + { + "epoch": 0.25, + "learning_rate": 1.918396702376608e-05, + "loss": 1.122, + "step": 3903 + }, + { + "epoch": 0.25, + "learning_rate": 1.9183752334743128e-05, + "loss": 1.0999, + "step": 3904 + }, + { + "epoch": 0.25, + "learning_rate": 1.9183537645720177e-05, + "loss": 1.2386, + "step": 3905 + }, + { + "epoch": 0.25, + "learning_rate": 1.9183322956697225e-05, + "loss": 1.1332, + "step": 3906 + }, + { + "epoch": 0.25, + "learning_rate": 1.9183108267674274e-05, + "loss": 1.121, + "step": 3907 + }, + { + "epoch": 0.25, + "learning_rate": 1.9182893578651323e-05, + "loss": 1.1756, + "step": 3908 + }, + { + "epoch": 0.25, + "learning_rate": 1.9182678889628376e-05, + "loss": 1.1256, + "step": 3909 + }, + { + "epoch": 0.25, + "learning_rate": 1.9182464200605425e-05, + "loss": 1.1048, + "step": 3910 + }, + { + "epoch": 0.25, + "learning_rate": 1.9182249511582474e-05, + "loss": 1.1045, + "step": 3911 + }, + { + "epoch": 0.25, + "learning_rate": 1.9182034822559523e-05, + "loss": 1.1509, + "step": 3912 + }, + { + "epoch": 0.25, + "learning_rate": 1.918182013353657e-05, + "loss": 1.2535, + "step": 3913 + }, + { + "epoch": 0.25, + "learning_rate": 1.9181605444513624e-05, + "loss": 1.049, + "step": 3914 + }, + { + "epoch": 0.25, + "learning_rate": 1.9181390755490673e-05, + "loss": 1.1043, + "step": 3915 + }, + { + "epoch": 0.25, + "learning_rate": 1.9181176066467725e-05, + "loss": 1.1228, + "step": 3916 + }, + { + "epoch": 0.25, + "learning_rate": 1.9180961377444774e-05, + "loss": 1.1184, + "step": 3917 + }, + { + "epoch": 0.25, + "learning_rate": 1.9180746688421823e-05, + "loss": 1.1864, + "step": 3918 + }, + { + "epoch": 0.25, + "learning_rate": 1.9180531999398872e-05, + "loss": 1.1014, + "step": 3919 + }, + { + "epoch": 0.25, + "learning_rate": 1.918031731037592e-05, + "loss": 1.0963, + "step": 3920 + }, + { + "epoch": 0.25, + "learning_rate": 1.918010262135297e-05, + "loss": 1.1807, + "step": 3921 + }, + { + "epoch": 0.25, + "learning_rate": 1.9179887932330022e-05, + "loss": 1.1303, + "step": 3922 + }, + { + "epoch": 0.25, + "learning_rate": 1.917967324330707e-05, + "loss": 1.1594, + "step": 3923 + }, + { + "epoch": 0.25, + "learning_rate": 1.917945855428412e-05, + "loss": 1.0889, + "step": 3924 + }, + { + "epoch": 0.25, + "learning_rate": 1.917924386526117e-05, + "loss": 1.1337, + "step": 3925 + }, + { + "epoch": 0.25, + "learning_rate": 1.917902917623822e-05, + "loss": 1.1501, + "step": 3926 + }, + { + "epoch": 0.25, + "learning_rate": 1.917881448721527e-05, + "loss": 1.12, + "step": 3927 + }, + { + "epoch": 0.25, + "learning_rate": 1.917859979819232e-05, + "loss": 1.0945, + "step": 3928 + }, + { + "epoch": 0.25, + "learning_rate": 1.9178385109169372e-05, + "loss": 1.0971, + "step": 3929 + }, + { + "epoch": 0.25, + "learning_rate": 1.917817042014642e-05, + "loss": 1.1225, + "step": 3930 + }, + { + "epoch": 0.25, + "learning_rate": 1.917795573112347e-05, + "loss": 1.1505, + "step": 3931 + }, + { + "epoch": 0.25, + "learning_rate": 1.917774104210052e-05, + "loss": 1.1168, + "step": 3932 + }, + { + "epoch": 0.25, + "learning_rate": 1.9177526353077568e-05, + "loss": 1.1347, + "step": 3933 + }, + { + "epoch": 0.25, + "learning_rate": 1.917731166405462e-05, + "loss": 1.1975, + "step": 3934 + }, + { + "epoch": 0.25, + "learning_rate": 1.917709697503167e-05, + "loss": 1.205, + "step": 3935 + }, + { + "epoch": 0.25, + "learning_rate": 1.9176882286008718e-05, + "loss": 1.1081, + "step": 3936 + }, + { + "epoch": 0.25, + "learning_rate": 1.9176667596985767e-05, + "loss": 1.1333, + "step": 3937 + }, + { + "epoch": 0.25, + "learning_rate": 1.9176452907962816e-05, + "loss": 1.0898, + "step": 3938 + }, + { + "epoch": 0.25, + "learning_rate": 1.9176238218939865e-05, + "loss": 1.1206, + "step": 3939 + }, + { + "epoch": 0.25, + "learning_rate": 1.9176023529916917e-05, + "loss": 1.1885, + "step": 3940 + }, + { + "epoch": 0.25, + "learning_rate": 1.9175808840893966e-05, + "loss": 1.1272, + "step": 3941 + }, + { + "epoch": 0.25, + "learning_rate": 1.9175594151871015e-05, + "loss": 1.2157, + "step": 3942 + }, + { + "epoch": 0.25, + "learning_rate": 1.9175379462848068e-05, + "loss": 1.1542, + "step": 3943 + }, + { + "epoch": 0.25, + "learning_rate": 1.9175164773825117e-05, + "loss": 1.0818, + "step": 3944 + }, + { + "epoch": 0.25, + "learning_rate": 1.9174950084802166e-05, + "loss": 1.045, + "step": 3945 + }, + { + "epoch": 0.25, + "learning_rate": 1.9174735395779215e-05, + "loss": 1.1172, + "step": 3946 + }, + { + "epoch": 0.25, + "learning_rate": 1.9174520706756267e-05, + "loss": 1.1768, + "step": 3947 + }, + { + "epoch": 0.25, + "learning_rate": 1.9174306017733316e-05, + "loss": 1.1813, + "step": 3948 + }, + { + "epoch": 0.25, + "learning_rate": 1.9174091328710365e-05, + "loss": 1.1168, + "step": 3949 + }, + { + "epoch": 0.25, + "learning_rate": 1.9173876639687414e-05, + "loss": 1.1265, + "step": 3950 + }, + { + "epoch": 0.25, + "learning_rate": 1.9173661950664463e-05, + "loss": 1.0615, + "step": 3951 + }, + { + "epoch": 0.25, + "learning_rate": 1.9173447261641515e-05, + "loss": 1.1669, + "step": 3952 + }, + { + "epoch": 0.25, + "learning_rate": 1.9173232572618564e-05, + "loss": 1.1276, + "step": 3953 + }, + { + "epoch": 0.25, + "learning_rate": 1.9173017883595613e-05, + "loss": 1.1202, + "step": 3954 + }, + { + "epoch": 0.25, + "learning_rate": 1.9172803194572662e-05, + "loss": 1.0957, + "step": 3955 + }, + { + "epoch": 0.25, + "learning_rate": 1.917258850554971e-05, + "loss": 1.1071, + "step": 3956 + }, + { + "epoch": 0.25, + "learning_rate": 1.9172373816526763e-05, + "loss": 1.1708, + "step": 3957 + }, + { + "epoch": 0.25, + "learning_rate": 1.9172159127503812e-05, + "loss": 1.2107, + "step": 3958 + }, + { + "epoch": 0.25, + "learning_rate": 1.917194443848086e-05, + "loss": 1.1352, + "step": 3959 + }, + { + "epoch": 0.25, + "learning_rate": 1.9171729749457914e-05, + "loss": 1.1572, + "step": 3960 + }, + { + "epoch": 0.25, + "learning_rate": 1.9171515060434963e-05, + "loss": 1.1334, + "step": 3961 + }, + { + "epoch": 0.25, + "learning_rate": 1.917130037141201e-05, + "loss": 1.1497, + "step": 3962 + }, + { + "epoch": 0.25, + "learning_rate": 1.917108568238906e-05, + "loss": 1.0962, + "step": 3963 + }, + { + "epoch": 0.26, + "learning_rate": 1.917087099336611e-05, + "loss": 1.1439, + "step": 3964 + }, + { + "epoch": 0.26, + "learning_rate": 1.9170656304343162e-05, + "loss": 1.1063, + "step": 3965 + }, + { + "epoch": 0.26, + "learning_rate": 1.917044161532021e-05, + "loss": 1.1842, + "step": 3966 + }, + { + "epoch": 0.26, + "learning_rate": 1.917022692629726e-05, + "loss": 1.1933, + "step": 3967 + }, + { + "epoch": 0.26, + "learning_rate": 1.917001223727431e-05, + "loss": 1.1343, + "step": 3968 + }, + { + "epoch": 0.26, + "learning_rate": 1.9169797548251358e-05, + "loss": 1.1076, + "step": 3969 + }, + { + "epoch": 0.26, + "learning_rate": 1.9169582859228407e-05, + "loss": 1.1415, + "step": 3970 + }, + { + "epoch": 0.26, + "learning_rate": 1.916936817020546e-05, + "loss": 1.1892, + "step": 3971 + }, + { + "epoch": 0.26, + "learning_rate": 1.916915348118251e-05, + "loss": 1.1451, + "step": 3972 + }, + { + "epoch": 0.26, + "learning_rate": 1.916893879215956e-05, + "loss": 1.1253, + "step": 3973 + }, + { + "epoch": 0.26, + "learning_rate": 1.916872410313661e-05, + "loss": 1.0679, + "step": 3974 + }, + { + "epoch": 0.26, + "learning_rate": 1.9168509414113658e-05, + "loss": 1.0731, + "step": 3975 + }, + { + "epoch": 0.26, + "learning_rate": 1.9168294725090707e-05, + "loss": 1.0825, + "step": 3976 + }, + { + "epoch": 0.26, + "learning_rate": 1.9168080036067756e-05, + "loss": 1.1061, + "step": 3977 + }, + { + "epoch": 0.26, + "learning_rate": 1.916786534704481e-05, + "loss": 1.1166, + "step": 3978 + }, + { + "epoch": 0.26, + "learning_rate": 1.9167650658021858e-05, + "loss": 1.1668, + "step": 3979 + }, + { + "epoch": 0.26, + "learning_rate": 1.9167435968998906e-05, + "loss": 1.1675, + "step": 3980 + }, + { + "epoch": 0.26, + "learning_rate": 1.9167221279975955e-05, + "loss": 1.1511, + "step": 3981 + }, + { + "epoch": 0.26, + "learning_rate": 1.9167006590953004e-05, + "loss": 1.1086, + "step": 3982 + }, + { + "epoch": 0.26, + "learning_rate": 1.9166791901930057e-05, + "loss": 1.235, + "step": 3983 + }, + { + "epoch": 0.26, + "learning_rate": 1.9166577212907106e-05, + "loss": 1.116, + "step": 3984 + }, + { + "epoch": 0.26, + "learning_rate": 1.9166362523884155e-05, + "loss": 1.036, + "step": 3985 + }, + { + "epoch": 0.26, + "learning_rate": 1.9166147834861204e-05, + "loss": 1.0967, + "step": 3986 + }, + { + "epoch": 0.26, + "learning_rate": 1.9165933145838256e-05, + "loss": 1.1643, + "step": 3987 + }, + { + "epoch": 0.26, + "learning_rate": 1.9165718456815305e-05, + "loss": 1.1282, + "step": 3988 + }, + { + "epoch": 0.26, + "learning_rate": 1.9165503767792354e-05, + "loss": 1.1009, + "step": 3989 + }, + { + "epoch": 0.26, + "learning_rate": 1.9165289078769403e-05, + "loss": 1.1018, + "step": 3990 + }, + { + "epoch": 0.26, + "learning_rate": 1.9165074389746455e-05, + "loss": 1.2282, + "step": 3991 + }, + { + "epoch": 0.26, + "learning_rate": 1.9164859700723504e-05, + "loss": 1.2432, + "step": 3992 + }, + { + "epoch": 0.26, + "learning_rate": 1.9164645011700553e-05, + "loss": 1.1585, + "step": 3993 + }, + { + "epoch": 0.26, + "learning_rate": 1.9164430322677602e-05, + "loss": 1.1233, + "step": 3994 + }, + { + "epoch": 0.26, + "learning_rate": 1.916421563365465e-05, + "loss": 1.1031, + "step": 3995 + }, + { + "epoch": 0.26, + "learning_rate": 1.9164000944631703e-05, + "loss": 1.0302, + "step": 3996 + }, + { + "epoch": 0.26, + "learning_rate": 1.9163786255608752e-05, + "loss": 1.1468, + "step": 3997 + }, + { + "epoch": 0.26, + "learning_rate": 1.91635715665858e-05, + "loss": 1.104, + "step": 3998 + }, + { + "epoch": 0.26, + "learning_rate": 1.916335687756285e-05, + "loss": 1.1036, + "step": 3999 + }, + { + "epoch": 0.26, + "learning_rate": 1.91631421885399e-05, + "loss": 1.0471, + "step": 4000 + }, + { + "epoch": 0.26, + "eval_laion/cider": 0.5101882411680806, + "eval_laion/meteor": 0.3843941180901681, + "eval_laion/pred_num_words": 36.53515625, + "eval_laion/sacrebleu": 15.794838453659754, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.1389846801757812, + "eval_runtime": 97.3045, + "eval_samples_per_second": 5.128, + "eval_steps_per_second": 0.329, + "step": 4000 + }, + { + "epoch": 0.26, + "learning_rate": 1.916292749951695e-05, + "loss": 1.1302, + "step": 4001 + }, + { + "epoch": 0.26, + "learning_rate": 1.9162712810494e-05, + "loss": 1.1033, + "step": 4002 + }, + { + "epoch": 0.26, + "learning_rate": 1.9162498121471053e-05, + "loss": 1.1405, + "step": 4003 + }, + { + "epoch": 0.26, + "learning_rate": 1.9162283432448102e-05, + "loss": 1.199, + "step": 4004 + }, + { + "epoch": 0.26, + "learning_rate": 1.916206874342515e-05, + "loss": 1.1655, + "step": 4005 + }, + { + "epoch": 0.26, + "learning_rate": 1.91618540544022e-05, + "loss": 1.1209, + "step": 4006 + }, + { + "epoch": 0.26, + "learning_rate": 1.916163936537925e-05, + "loss": 1.1553, + "step": 4007 + }, + { + "epoch": 0.26, + "learning_rate": 1.9161424676356298e-05, + "loss": 1.1315, + "step": 4008 + }, + { + "epoch": 0.26, + "learning_rate": 1.916120998733335e-05, + "loss": 1.1579, + "step": 4009 + }, + { + "epoch": 0.26, + "learning_rate": 1.91609952983104e-05, + "loss": 1.191, + "step": 4010 + }, + { + "epoch": 0.26, + "learning_rate": 1.9160780609287448e-05, + "loss": 1.0635, + "step": 4011 + }, + { + "epoch": 0.26, + "learning_rate": 1.9160780609287448e-05, + "loss": 1.1711, + "step": 4012 + }, + { + "epoch": 0.26, + "learning_rate": 1.9160565920264497e-05, + "loss": 1.1001, + "step": 4013 + }, + { + "epoch": 0.26, + "learning_rate": 1.9160351231241546e-05, + "loss": 1.1341, + "step": 4014 + }, + { + "epoch": 0.26, + "learning_rate": 1.91601365422186e-05, + "loss": 1.0283, + "step": 4015 + }, + { + "epoch": 0.26, + "learning_rate": 1.9159921853195647e-05, + "loss": 1.0719, + "step": 4016 + }, + { + "epoch": 0.26, + "learning_rate": 1.91597071641727e-05, + "loss": 1.1167, + "step": 4017 + }, + { + "epoch": 0.26, + "learning_rate": 1.915949247514975e-05, + "loss": 1.1066, + "step": 4018 + }, + { + "epoch": 0.26, + "learning_rate": 1.9159277786126798e-05, + "loss": 1.2545, + "step": 4019 + }, + { + "epoch": 0.26, + "learning_rate": 1.9159063097103847e-05, + "loss": 1.1104, + "step": 4020 + }, + { + "epoch": 0.26, + "learning_rate": 1.9158848408080896e-05, + "loss": 1.1187, + "step": 4021 + }, + { + "epoch": 0.26, + "learning_rate": 1.9158633719057948e-05, + "loss": 1.1308, + "step": 4022 + }, + { + "epoch": 0.26, + "learning_rate": 1.9158419030034997e-05, + "loss": 1.1589, + "step": 4023 + }, + { + "epoch": 0.26, + "learning_rate": 1.9158204341012046e-05, + "loss": 1.0988, + "step": 4024 + }, + { + "epoch": 0.26, + "learning_rate": 1.9157989651989095e-05, + "loss": 1.0923, + "step": 4025 + }, + { + "epoch": 0.26, + "learning_rate": 1.9157774962966144e-05, + "loss": 1.0967, + "step": 4026 + }, + { + "epoch": 0.26, + "learning_rate": 1.9157560273943193e-05, + "loss": 1.1937, + "step": 4027 + }, + { + "epoch": 0.26, + "learning_rate": 1.9157345584920245e-05, + "loss": 1.136, + "step": 4028 + }, + { + "epoch": 0.26, + "learning_rate": 1.9157130895897294e-05, + "loss": 1.1616, + "step": 4029 + }, + { + "epoch": 0.26, + "learning_rate": 1.9156916206874343e-05, + "loss": 1.12, + "step": 4030 + }, + { + "epoch": 0.26, + "learning_rate": 1.9156701517851395e-05, + "loss": 1.145, + "step": 4031 + }, + { + "epoch": 0.26, + "learning_rate": 1.9156486828828444e-05, + "loss": 1.1373, + "step": 4032 + }, + { + "epoch": 0.26, + "learning_rate": 1.9156272139805493e-05, + "loss": 1.2022, + "step": 4033 + }, + { + "epoch": 0.26, + "learning_rate": 1.9156057450782542e-05, + "loss": 1.1797, + "step": 4034 + }, + { + "epoch": 0.26, + "learning_rate": 1.9155842761759595e-05, + "loss": 1.1437, + "step": 4035 + }, + { + "epoch": 0.26, + "learning_rate": 1.9155628072736644e-05, + "loss": 1.0856, + "step": 4036 + }, + { + "epoch": 0.26, + "learning_rate": 1.9155413383713693e-05, + "loss": 1.1292, + "step": 4037 + }, + { + "epoch": 0.26, + "learning_rate": 1.915519869469074e-05, + "loss": 1.218, + "step": 4038 + }, + { + "epoch": 0.26, + "learning_rate": 1.915498400566779e-05, + "loss": 1.0659, + "step": 4039 + }, + { + "epoch": 0.26, + "learning_rate": 1.9154769316644843e-05, + "loss": 1.1237, + "step": 4040 + }, + { + "epoch": 0.26, + "learning_rate": 1.9154554627621892e-05, + "loss": 1.1037, + "step": 4041 + }, + { + "epoch": 0.26, + "learning_rate": 1.915433993859894e-05, + "loss": 0.9852, + "step": 4042 + }, + { + "epoch": 0.26, + "learning_rate": 1.915412524957599e-05, + "loss": 1.1047, + "step": 4043 + }, + { + "epoch": 0.26, + "learning_rate": 1.915391056055304e-05, + "loss": 1.1698, + "step": 4044 + }, + { + "epoch": 0.26, + "learning_rate": 1.9153695871530088e-05, + "loss": 1.0946, + "step": 4045 + }, + { + "epoch": 0.26, + "learning_rate": 1.915348118250714e-05, + "loss": 1.1451, + "step": 4046 + }, + { + "epoch": 0.26, + "learning_rate": 1.915326649348419e-05, + "loss": 1.1101, + "step": 4047 + }, + { + "epoch": 0.26, + "learning_rate": 1.915305180446124e-05, + "loss": 1.0314, + "step": 4048 + }, + { + "epoch": 0.26, + "learning_rate": 1.915283711543829e-05, + "loss": 1.101, + "step": 4049 + }, + { + "epoch": 0.26, + "learning_rate": 1.915262242641534e-05, + "loss": 1.2001, + "step": 4050 + }, + { + "epoch": 0.26, + "learning_rate": 1.9152407737392388e-05, + "loss": 1.0622, + "step": 4051 + }, + { + "epoch": 0.26, + "learning_rate": 1.9152193048369437e-05, + "loss": 1.0881, + "step": 4052 + }, + { + "epoch": 0.26, + "learning_rate": 1.915197835934649e-05, + "loss": 1.1119, + "step": 4053 + }, + { + "epoch": 0.26, + "learning_rate": 1.915176367032354e-05, + "loss": 1.2028, + "step": 4054 + }, + { + "epoch": 0.26, + "learning_rate": 1.9151548981300587e-05, + "loss": 1.0983, + "step": 4055 + }, + { + "epoch": 0.26, + "learning_rate": 1.9151334292277636e-05, + "loss": 1.085, + "step": 4056 + }, + { + "epoch": 0.26, + "learning_rate": 1.9151119603254685e-05, + "loss": 1.1868, + "step": 4057 + }, + { + "epoch": 0.26, + "learning_rate": 1.9150904914231734e-05, + "loss": 1.1124, + "step": 4058 + }, + { + "epoch": 0.26, + "learning_rate": 1.9150690225208787e-05, + "loss": 1.0801, + "step": 4059 + }, + { + "epoch": 0.26, + "learning_rate": 1.915047553618584e-05, + "loss": 1.039, + "step": 4060 + }, + { + "epoch": 0.26, + "learning_rate": 1.9150260847162888e-05, + "loss": 1.0896, + "step": 4061 + }, + { + "epoch": 0.26, + "learning_rate": 1.9150046158139937e-05, + "loss": 1.121, + "step": 4062 + }, + { + "epoch": 0.26, + "learning_rate": 1.9149831469116986e-05, + "loss": 1.1263, + "step": 4063 + }, + { + "epoch": 0.26, + "learning_rate": 1.9149616780094035e-05, + "loss": 1.2043, + "step": 4064 + }, + { + "epoch": 0.26, + "learning_rate": 1.9149402091071084e-05, + "loss": 1.2041, + "step": 4065 + }, + { + "epoch": 0.26, + "learning_rate": 1.9149187402048136e-05, + "loss": 1.1264, + "step": 4066 + }, + { + "epoch": 0.26, + "learning_rate": 1.9148972713025185e-05, + "loss": 1.1294, + "step": 4067 + }, + { + "epoch": 0.26, + "learning_rate": 1.9148758024002234e-05, + "loss": 1.1618, + "step": 4068 + }, + { + "epoch": 0.26, + "learning_rate": 1.9148543334979283e-05, + "loss": 1.1093, + "step": 4069 + }, + { + "epoch": 0.26, + "learning_rate": 1.9148328645956332e-05, + "loss": 1.1561, + "step": 4070 + }, + { + "epoch": 0.26, + "learning_rate": 1.9148113956933384e-05, + "loss": 1.2423, + "step": 4071 + }, + { + "epoch": 0.26, + "learning_rate": 1.9147899267910433e-05, + "loss": 1.1216, + "step": 4072 + }, + { + "epoch": 0.26, + "learning_rate": 1.9147684578887482e-05, + "loss": 1.1274, + "step": 4073 + }, + { + "epoch": 0.26, + "learning_rate": 1.914746988986453e-05, + "loss": 1.1068, + "step": 4074 + }, + { + "epoch": 0.26, + "learning_rate": 1.9147255200841584e-05, + "loss": 1.0716, + "step": 4075 + }, + { + "epoch": 0.26, + "learning_rate": 1.9147040511818633e-05, + "loss": 1.1473, + "step": 4076 + }, + { + "epoch": 0.26, + "learning_rate": 1.914682582279568e-05, + "loss": 1.095, + "step": 4077 + }, + { + "epoch": 0.26, + "learning_rate": 1.914661113377273e-05, + "loss": 1.0352, + "step": 4078 + }, + { + "epoch": 0.26, + "learning_rate": 1.9146396444749783e-05, + "loss": 1.1538, + "step": 4079 + }, + { + "epoch": 0.26, + "learning_rate": 1.9146181755726832e-05, + "loss": 1.1466, + "step": 4080 + }, + { + "epoch": 0.26, + "learning_rate": 1.914596706670388e-05, + "loss": 1.0626, + "step": 4081 + }, + { + "epoch": 0.26, + "learning_rate": 1.914575237768093e-05, + "loss": 1.1923, + "step": 4082 + }, + { + "epoch": 0.26, + "learning_rate": 1.914553768865798e-05, + "loss": 1.1399, + "step": 4083 + }, + { + "epoch": 0.26, + "learning_rate": 1.914532299963503e-05, + "loss": 1.0283, + "step": 4084 + }, + { + "epoch": 0.26, + "learning_rate": 1.914510831061208e-05, + "loss": 1.1125, + "step": 4085 + }, + { + "epoch": 0.26, + "learning_rate": 1.914489362158913e-05, + "loss": 1.1183, + "step": 4086 + }, + { + "epoch": 0.26, + "learning_rate": 1.9144678932566178e-05, + "loss": 1.0589, + "step": 4087 + }, + { + "epoch": 0.26, + "learning_rate": 1.9144464243543227e-05, + "loss": 1.129, + "step": 4088 + }, + { + "epoch": 0.26, + "learning_rate": 1.914424955452028e-05, + "loss": 1.1352, + "step": 4089 + }, + { + "epoch": 0.26, + "learning_rate": 1.914403486549733e-05, + "loss": 1.1243, + "step": 4090 + }, + { + "epoch": 0.26, + "learning_rate": 1.914382017647438e-05, + "loss": 1.1223, + "step": 4091 + }, + { + "epoch": 0.26, + "learning_rate": 1.914360548745143e-05, + "loss": 1.157, + "step": 4092 + }, + { + "epoch": 0.26, + "learning_rate": 1.914339079842848e-05, + "loss": 1.092, + "step": 4093 + }, + { + "epoch": 0.26, + "learning_rate": 1.9143176109405528e-05, + "loss": 1.1458, + "step": 4094 + }, + { + "epoch": 0.26, + "learning_rate": 1.9142961420382577e-05, + "loss": 1.112, + "step": 4095 + }, + { + "epoch": 0.26, + "learning_rate": 1.9142746731359626e-05, + "loss": 1.1231, + "step": 4096 + }, + { + "epoch": 0.26, + "learning_rate": 1.9142532042336678e-05, + "loss": 1.0594, + "step": 4097 + }, + { + "epoch": 0.26, + "learning_rate": 1.9142317353313727e-05, + "loss": 1.0782, + "step": 4098 + }, + { + "epoch": 0.26, + "learning_rate": 1.9142102664290776e-05, + "loss": 1.1167, + "step": 4099 + }, + { + "epoch": 0.26, + "learning_rate": 1.9141887975267825e-05, + "loss": 1.0797, + "step": 4100 + }, + { + "epoch": 0.26, + "learning_rate": 1.9141673286244874e-05, + "loss": 1.1462, + "step": 4101 + }, + { + "epoch": 0.26, + "learning_rate": 1.9141458597221926e-05, + "loss": 1.1202, + "step": 4102 + }, + { + "epoch": 0.26, + "learning_rate": 1.9141243908198975e-05, + "loss": 1.1947, + "step": 4103 + }, + { + "epoch": 0.26, + "learning_rate": 1.9141029219176027e-05, + "loss": 1.1597, + "step": 4104 + }, + { + "epoch": 0.26, + "learning_rate": 1.9140814530153076e-05, + "loss": 1.0491, + "step": 4105 + }, + { + "epoch": 0.26, + "learning_rate": 1.9140599841130125e-05, + "loss": 1.1616, + "step": 4106 + }, + { + "epoch": 0.26, + "learning_rate": 1.9140385152107174e-05, + "loss": 1.1362, + "step": 4107 + }, + { + "epoch": 0.26, + "learning_rate": 1.9140170463084223e-05, + "loss": 1.0954, + "step": 4108 + }, + { + "epoch": 0.26, + "learning_rate": 1.9139955774061276e-05, + "loss": 1.0761, + "step": 4109 + }, + { + "epoch": 0.26, + "learning_rate": 1.9139741085038325e-05, + "loss": 1.1765, + "step": 4110 + }, + { + "epoch": 0.26, + "learning_rate": 1.9139526396015374e-05, + "loss": 1.0773, + "step": 4111 + }, + { + "epoch": 0.26, + "learning_rate": 1.9139311706992423e-05, + "loss": 1.0792, + "step": 4112 + }, + { + "epoch": 0.26, + "learning_rate": 1.913909701796947e-05, + "loss": 1.083, + "step": 4113 + }, + { + "epoch": 0.26, + "learning_rate": 1.913888232894652e-05, + "loss": 1.1141, + "step": 4114 + }, + { + "epoch": 0.26, + "learning_rate": 1.9138667639923573e-05, + "loss": 1.0699, + "step": 4115 + }, + { + "epoch": 0.26, + "learning_rate": 1.9138452950900622e-05, + "loss": 1.1337, + "step": 4116 + }, + { + "epoch": 0.26, + "learning_rate": 1.913823826187767e-05, + "loss": 1.1476, + "step": 4117 + }, + { + "epoch": 0.26, + "learning_rate": 1.9138023572854723e-05, + "loss": 1.1154, + "step": 4118 + }, + { + "epoch": 0.27, + "learning_rate": 1.9137808883831772e-05, + "loss": 1.1914, + "step": 4119 + }, + { + "epoch": 0.27, + "learning_rate": 1.913759419480882e-05, + "loss": 1.1082, + "step": 4120 + }, + { + "epoch": 0.27, + "learning_rate": 1.913737950578587e-05, + "loss": 1.2053, + "step": 4121 + }, + { + "epoch": 0.27, + "learning_rate": 1.9137164816762922e-05, + "loss": 1.1464, + "step": 4122 + }, + { + "epoch": 0.27, + "learning_rate": 1.913695012773997e-05, + "loss": 1.1252, + "step": 4123 + }, + { + "epoch": 0.27, + "learning_rate": 1.913673543871702e-05, + "loss": 1.1105, + "step": 4124 + }, + { + "epoch": 0.27, + "learning_rate": 1.913652074969407e-05, + "loss": 1.0666, + "step": 4125 + }, + { + "epoch": 0.27, + "learning_rate": 1.9136306060671118e-05, + "loss": 1.1318, + "step": 4126 + }, + { + "epoch": 0.27, + "learning_rate": 1.913609137164817e-05, + "loss": 1.1197, + "step": 4127 + }, + { + "epoch": 0.27, + "learning_rate": 1.913587668262522e-05, + "loss": 1.0656, + "step": 4128 + }, + { + "epoch": 0.27, + "learning_rate": 1.913566199360227e-05, + "loss": 1.1445, + "step": 4129 + }, + { + "epoch": 0.27, + "learning_rate": 1.9135447304579317e-05, + "loss": 1.079, + "step": 4130 + }, + { + "epoch": 0.27, + "learning_rate": 1.9135232615556366e-05, + "loss": 1.1579, + "step": 4131 + }, + { + "epoch": 0.27, + "learning_rate": 1.9135017926533415e-05, + "loss": 1.1684, + "step": 4132 + }, + { + "epoch": 0.27, + "learning_rate": 1.9134803237510468e-05, + "loss": 1.0778, + "step": 4133 + }, + { + "epoch": 0.27, + "learning_rate": 1.9134588548487517e-05, + "loss": 1.1571, + "step": 4134 + }, + { + "epoch": 0.27, + "learning_rate": 1.913437385946457e-05, + "loss": 1.0661, + "step": 4135 + }, + { + "epoch": 0.27, + "learning_rate": 1.9134159170441618e-05, + "loss": 1.1758, + "step": 4136 + }, + { + "epoch": 0.27, + "learning_rate": 1.9133944481418667e-05, + "loss": 1.1846, + "step": 4137 + }, + { + "epoch": 0.27, + "learning_rate": 1.9133729792395716e-05, + "loss": 1.0818, + "step": 4138 + }, + { + "epoch": 0.27, + "learning_rate": 1.9133515103372765e-05, + "loss": 1.1806, + "step": 4139 + }, + { + "epoch": 0.27, + "learning_rate": 1.9133300414349817e-05, + "loss": 1.1454, + "step": 4140 + }, + { + "epoch": 0.27, + "learning_rate": 1.9133085725326866e-05, + "loss": 1.2229, + "step": 4141 + }, + { + "epoch": 0.27, + "learning_rate": 1.9132871036303915e-05, + "loss": 1.1672, + "step": 4142 + }, + { + "epoch": 0.27, + "learning_rate": 1.9132656347280964e-05, + "loss": 1.1727, + "step": 4143 + }, + { + "epoch": 0.27, + "learning_rate": 1.9132441658258013e-05, + "loss": 1.1624, + "step": 4144 + }, + { + "epoch": 0.27, + "learning_rate": 1.9132226969235062e-05, + "loss": 1.1533, + "step": 4145 + }, + { + "epoch": 0.27, + "learning_rate": 1.9132012280212114e-05, + "loss": 1.1065, + "step": 4146 + }, + { + "epoch": 0.27, + "learning_rate": 1.9131797591189167e-05, + "loss": 1.1222, + "step": 4147 + }, + { + "epoch": 0.27, + "learning_rate": 1.9131582902166216e-05, + "loss": 1.1214, + "step": 4148 + }, + { + "epoch": 0.27, + "learning_rate": 1.9131368213143265e-05, + "loss": 1.1175, + "step": 4149 + }, + { + "epoch": 0.27, + "learning_rate": 1.9131153524120314e-05, + "loss": 1.0419, + "step": 4150 + }, + { + "epoch": 0.27, + "learning_rate": 1.9130938835097363e-05, + "loss": 1.0431, + "step": 4151 + }, + { + "epoch": 0.27, + "learning_rate": 1.913072414607441e-05, + "loss": 1.1325, + "step": 4152 + }, + { + "epoch": 0.27, + "learning_rate": 1.9130509457051464e-05, + "loss": 1.128, + "step": 4153 + }, + { + "epoch": 0.27, + "learning_rate": 1.9130294768028513e-05, + "loss": 1.0729, + "step": 4154 + }, + { + "epoch": 0.27, + "learning_rate": 1.9130080079005562e-05, + "loss": 1.1803, + "step": 4155 + }, + { + "epoch": 0.27, + "learning_rate": 1.912986538998261e-05, + "loss": 1.1389, + "step": 4156 + }, + { + "epoch": 0.27, + "learning_rate": 1.912965070095966e-05, + "loss": 1.1399, + "step": 4157 + }, + { + "epoch": 0.27, + "learning_rate": 1.9129436011936712e-05, + "loss": 1.0883, + "step": 4158 + }, + { + "epoch": 0.27, + "learning_rate": 1.912922132291376e-05, + "loss": 1.1227, + "step": 4159 + }, + { + "epoch": 0.27, + "learning_rate": 1.912900663389081e-05, + "loss": 1.1538, + "step": 4160 + }, + { + "epoch": 0.27, + "learning_rate": 1.912879194486786e-05, + "loss": 1.1072, + "step": 4161 + }, + { + "epoch": 0.27, + "learning_rate": 1.912857725584491e-05, + "loss": 1.1613, + "step": 4162 + }, + { + "epoch": 0.27, + "learning_rate": 1.912836256682196e-05, + "loss": 1.1689, + "step": 4163 + }, + { + "epoch": 0.27, + "learning_rate": 1.912814787779901e-05, + "loss": 1.1032, + "step": 4164 + }, + { + "epoch": 0.27, + "learning_rate": 1.912793318877606e-05, + "loss": 1.1582, + "step": 4165 + }, + { + "epoch": 0.27, + "learning_rate": 1.912771849975311e-05, + "loss": 1.1971, + "step": 4166 + }, + { + "epoch": 0.27, + "learning_rate": 1.912750381073016e-05, + "loss": 1.1248, + "step": 4167 + }, + { + "epoch": 0.27, + "learning_rate": 1.912728912170721e-05, + "loss": 1.0568, + "step": 4168 + }, + { + "epoch": 0.27, + "learning_rate": 1.9127074432684258e-05, + "loss": 1.1326, + "step": 4169 + }, + { + "epoch": 0.27, + "learning_rate": 1.9126859743661307e-05, + "loss": 1.1358, + "step": 4170 + }, + { + "epoch": 0.27, + "learning_rate": 1.912664505463836e-05, + "loss": 1.0439, + "step": 4171 + }, + { + "epoch": 0.27, + "learning_rate": 1.9126430365615408e-05, + "loss": 1.0547, + "step": 4172 + }, + { + "epoch": 0.27, + "learning_rate": 1.9126215676592457e-05, + "loss": 1.1807, + "step": 4173 + }, + { + "epoch": 0.27, + "learning_rate": 1.9126000987569506e-05, + "loss": 1.1676, + "step": 4174 + }, + { + "epoch": 0.27, + "learning_rate": 1.9125786298546555e-05, + "loss": 1.0851, + "step": 4175 + }, + { + "epoch": 0.27, + "learning_rate": 1.9125571609523607e-05, + "loss": 1.1338, + "step": 4176 + }, + { + "epoch": 0.27, + "learning_rate": 1.9125356920500656e-05, + "loss": 1.1595, + "step": 4177 + }, + { + "epoch": 0.27, + "learning_rate": 1.912514223147771e-05, + "loss": 1.1021, + "step": 4178 + }, + { + "epoch": 0.27, + "learning_rate": 1.9124927542454757e-05, + "loss": 1.1451, + "step": 4179 + }, + { + "epoch": 0.27, + "learning_rate": 1.9124712853431806e-05, + "loss": 1.12, + "step": 4180 + }, + { + "epoch": 0.27, + "learning_rate": 1.9124498164408855e-05, + "loss": 1.1003, + "step": 4181 + }, + { + "epoch": 0.27, + "learning_rate": 1.9124283475385904e-05, + "loss": 1.0997, + "step": 4182 + }, + { + "epoch": 0.27, + "learning_rate": 1.9124068786362953e-05, + "loss": 1.1305, + "step": 4183 + }, + { + "epoch": 0.27, + "learning_rate": 1.9123854097340006e-05, + "loss": 1.0514, + "step": 4184 + }, + { + "epoch": 0.27, + "learning_rate": 1.9123639408317055e-05, + "loss": 1.0731, + "step": 4185 + }, + { + "epoch": 0.27, + "learning_rate": 1.9123424719294104e-05, + "loss": 1.166, + "step": 4186 + }, + { + "epoch": 0.27, + "learning_rate": 1.9123210030271152e-05, + "loss": 1.0642, + "step": 4187 + }, + { + "epoch": 0.27, + "learning_rate": 1.91229953412482e-05, + "loss": 1.1387, + "step": 4188 + }, + { + "epoch": 0.27, + "learning_rate": 1.9122780652225254e-05, + "loss": 1.1553, + "step": 4189 + }, + { + "epoch": 0.27, + "learning_rate": 1.9122565963202303e-05, + "loss": 1.1225, + "step": 4190 + }, + { + "epoch": 0.27, + "learning_rate": 1.9122351274179355e-05, + "loss": 1.1487, + "step": 4191 + }, + { + "epoch": 0.27, + "learning_rate": 1.9122136585156404e-05, + "loss": 1.091, + "step": 4192 + }, + { + "epoch": 0.27, + "learning_rate": 1.9121921896133453e-05, + "loss": 1.0241, + "step": 4193 + }, + { + "epoch": 0.27, + "learning_rate": 1.9121707207110502e-05, + "loss": 1.0514, + "step": 4194 + }, + { + "epoch": 0.27, + "learning_rate": 1.912149251808755e-05, + "loss": 1.1378, + "step": 4195 + }, + { + "epoch": 0.27, + "learning_rate": 1.9121277829064603e-05, + "loss": 1.1306, + "step": 4196 + }, + { + "epoch": 0.27, + "learning_rate": 1.9121063140041652e-05, + "loss": 1.0955, + "step": 4197 + }, + { + "epoch": 0.27, + "learning_rate": 1.91208484510187e-05, + "loss": 1.1277, + "step": 4198 + }, + { + "epoch": 0.27, + "learning_rate": 1.912063376199575e-05, + "loss": 1.1993, + "step": 4199 + }, + { + "epoch": 0.27, + "learning_rate": 1.91204190729728e-05, + "loss": 1.0998, + "step": 4200 + }, + { + "epoch": 0.27, + "learning_rate": 1.9120204383949848e-05, + "loss": 1.0498, + "step": 4201 + }, + { + "epoch": 0.27, + "learning_rate": 1.91199896949269e-05, + "loss": 1.1231, + "step": 4202 + }, + { + "epoch": 0.27, + "learning_rate": 1.911977500590395e-05, + "loss": 1.099, + "step": 4203 + }, + { + "epoch": 0.27, + "learning_rate": 1.9119560316881e-05, + "loss": 1.1141, + "step": 4204 + }, + { + "epoch": 0.27, + "learning_rate": 1.911934562785805e-05, + "loss": 1.1059, + "step": 4205 + }, + { + "epoch": 0.27, + "learning_rate": 1.91191309388351e-05, + "loss": 1.1748, + "step": 4206 + }, + { + "epoch": 0.27, + "learning_rate": 1.911891624981215e-05, + "loss": 1.0946, + "step": 4207 + }, + { + "epoch": 0.27, + "learning_rate": 1.9118701560789198e-05, + "loss": 1.169, + "step": 4208 + }, + { + "epoch": 0.27, + "learning_rate": 1.911848687176625e-05, + "loss": 1.1264, + "step": 4209 + }, + { + "epoch": 0.27, + "learning_rate": 1.91182721827433e-05, + "loss": 1.1467, + "step": 4210 + }, + { + "epoch": 0.27, + "learning_rate": 1.9118057493720348e-05, + "loss": 1.2344, + "step": 4211 + }, + { + "epoch": 0.27, + "learning_rate": 1.9117842804697397e-05, + "loss": 1.098, + "step": 4212 + }, + { + "epoch": 0.27, + "learning_rate": 1.9117628115674446e-05, + "loss": 1.1125, + "step": 4213 + }, + { + "epoch": 0.27, + "learning_rate": 1.9117413426651498e-05, + "loss": 1.1806, + "step": 4214 + }, + { + "epoch": 0.27, + "learning_rate": 1.9117198737628547e-05, + "loss": 1.0518, + "step": 4215 + }, + { + "epoch": 0.27, + "learning_rate": 1.9116984048605596e-05, + "loss": 1.1303, + "step": 4216 + }, + { + "epoch": 0.27, + "learning_rate": 1.9116769359582645e-05, + "loss": 1.0839, + "step": 4217 + }, + { + "epoch": 0.27, + "learning_rate": 1.9116554670559694e-05, + "loss": 1.119, + "step": 4218 + }, + { + "epoch": 0.27, + "learning_rate": 1.9116339981536743e-05, + "loss": 1.1447, + "step": 4219 + }, + { + "epoch": 0.27, + "learning_rate": 1.9116125292513795e-05, + "loss": 1.084, + "step": 4220 + }, + { + "epoch": 0.27, + "learning_rate": 1.9115910603490844e-05, + "loss": 1.1459, + "step": 4221 + }, + { + "epoch": 0.27, + "learning_rate": 1.9115695914467897e-05, + "loss": 1.1143, + "step": 4222 + }, + { + "epoch": 0.27, + "learning_rate": 1.9115481225444946e-05, + "loss": 1.0374, + "step": 4223 + }, + { + "epoch": 0.27, + "learning_rate": 1.9115266536421995e-05, + "loss": 1.2082, + "step": 4224 + }, + { + "epoch": 0.27, + "learning_rate": 1.9115051847399044e-05, + "loss": 1.0905, + "step": 4225 + }, + { + "epoch": 0.27, + "learning_rate": 1.9114837158376093e-05, + "loss": 1.2186, + "step": 4226 + }, + { + "epoch": 0.27, + "learning_rate": 1.9114622469353145e-05, + "loss": 1.2091, + "step": 4227 + }, + { + "epoch": 0.27, + "learning_rate": 1.9114407780330194e-05, + "loss": 1.1104, + "step": 4228 + }, + { + "epoch": 0.27, + "learning_rate": 1.9114193091307243e-05, + "loss": 1.0617, + "step": 4229 + }, + { + "epoch": 0.27, + "learning_rate": 1.9113978402284292e-05, + "loss": 1.1382, + "step": 4230 + }, + { + "epoch": 0.27, + "learning_rate": 1.911376371326134e-05, + "loss": 1.1518, + "step": 4231 + }, + { + "epoch": 0.27, + "learning_rate": 1.911354902423839e-05, + "loss": 1.0813, + "step": 4232 + }, + { + "epoch": 0.27, + "learning_rate": 1.9113334335215442e-05, + "loss": 1.0286, + "step": 4233 + }, + { + "epoch": 0.27, + "learning_rate": 1.9113119646192495e-05, + "loss": 1.2163, + "step": 4234 + }, + { + "epoch": 0.27, + "learning_rate": 1.9112904957169543e-05, + "loss": 1.109, + "step": 4235 + }, + { + "epoch": 0.27, + "learning_rate": 1.9112690268146592e-05, + "loss": 1.1828, + "step": 4236 + }, + { + "epoch": 0.27, + "learning_rate": 1.911247557912364e-05, + "loss": 1.1377, + "step": 4237 + }, + { + "epoch": 0.27, + "learning_rate": 1.911226089010069e-05, + "loss": 1.2082, + "step": 4238 + }, + { + "epoch": 0.27, + "learning_rate": 1.911204620107774e-05, + "loss": 1.0828, + "step": 4239 + }, + { + "epoch": 0.27, + "learning_rate": 1.911183151205479e-05, + "loss": 1.0373, + "step": 4240 + }, + { + "epoch": 0.27, + "learning_rate": 1.911161682303184e-05, + "loss": 1.1949, + "step": 4241 + }, + { + "epoch": 0.27, + "learning_rate": 1.911140213400889e-05, + "loss": 1.1585, + "step": 4242 + }, + { + "epoch": 0.27, + "learning_rate": 1.911118744498594e-05, + "loss": 1.1086, + "step": 4243 + }, + { + "epoch": 0.27, + "learning_rate": 1.9110972755962988e-05, + "loss": 1.0092, + "step": 4244 + }, + { + "epoch": 0.27, + "learning_rate": 1.911075806694004e-05, + "loss": 1.1051, + "step": 4245 + }, + { + "epoch": 0.27, + "learning_rate": 1.911054337791709e-05, + "loss": 1.072, + "step": 4246 + }, + { + "epoch": 0.27, + "learning_rate": 1.9110328688894138e-05, + "loss": 1.114, + "step": 4247 + }, + { + "epoch": 0.27, + "learning_rate": 1.9110113999871187e-05, + "loss": 1.0728, + "step": 4248 + }, + { + "epoch": 0.27, + "learning_rate": 1.910989931084824e-05, + "loss": 1.124, + "step": 4249 + }, + { + "epoch": 0.27, + "learning_rate": 1.9109684621825288e-05, + "loss": 1.1267, + "step": 4250 + }, + { + "epoch": 0.27, + "learning_rate": 1.9109469932802337e-05, + "loss": 1.165, + "step": 4251 + }, + { + "epoch": 0.27, + "learning_rate": 1.9109255243779386e-05, + "loss": 1.0752, + "step": 4252 + }, + { + "epoch": 0.27, + "learning_rate": 1.910904055475644e-05, + "loss": 1.1085, + "step": 4253 + }, + { + "epoch": 0.27, + "learning_rate": 1.9108825865733487e-05, + "loss": 1.0994, + "step": 4254 + }, + { + "epoch": 0.27, + "learning_rate": 1.9108611176710536e-05, + "loss": 1.1608, + "step": 4255 + }, + { + "epoch": 0.27, + "learning_rate": 1.9108396487687585e-05, + "loss": 1.2016, + "step": 4256 + }, + { + "epoch": 0.27, + "learning_rate": 1.9108181798664634e-05, + "loss": 1.0625, + "step": 4257 + }, + { + "epoch": 0.27, + "learning_rate": 1.9107967109641687e-05, + "loss": 1.0299, + "step": 4258 + }, + { + "epoch": 0.27, + "learning_rate": 1.9107752420618736e-05, + "loss": 1.1426, + "step": 4259 + }, + { + "epoch": 0.27, + "learning_rate": 1.9107537731595785e-05, + "loss": 1.0372, + "step": 4260 + }, + { + "epoch": 0.27, + "learning_rate": 1.9107323042572834e-05, + "loss": 1.1233, + "step": 4261 + }, + { + "epoch": 0.27, + "learning_rate": 1.9107108353549882e-05, + "loss": 1.1012, + "step": 4262 + }, + { + "epoch": 0.27, + "learning_rate": 1.9106893664526935e-05, + "loss": 1.2365, + "step": 4263 + }, + { + "epoch": 0.27, + "learning_rate": 1.9106678975503984e-05, + "loss": 1.1616, + "step": 4264 + }, + { + "epoch": 0.27, + "learning_rate": 1.9106464286481036e-05, + "loss": 1.2729, + "step": 4265 + }, + { + "epoch": 0.27, + "learning_rate": 1.9106249597458085e-05, + "loss": 1.0863, + "step": 4266 + }, + { + "epoch": 0.27, + "learning_rate": 1.9106034908435134e-05, + "loss": 1.1973, + "step": 4267 + }, + { + "epoch": 0.27, + "learning_rate": 1.9105820219412183e-05, + "loss": 1.1256, + "step": 4268 + }, + { + "epoch": 0.27, + "learning_rate": 1.9105605530389232e-05, + "loss": 1.1334, + "step": 4269 + }, + { + "epoch": 0.27, + "learning_rate": 1.910539084136628e-05, + "loss": 1.2269, + "step": 4270 + }, + { + "epoch": 0.27, + "learning_rate": 1.9105176152343333e-05, + "loss": 1.1226, + "step": 4271 + }, + { + "epoch": 0.27, + "learning_rate": 1.9104961463320382e-05, + "loss": 1.1182, + "step": 4272 + }, + { + "epoch": 0.27, + "learning_rate": 1.910474677429743e-05, + "loss": 1.1015, + "step": 4273 + }, + { + "epoch": 0.27, + "learning_rate": 1.910453208527448e-05, + "loss": 1.2048, + "step": 4274 + }, + { + "epoch": 0.28, + "learning_rate": 1.910431739625153e-05, + "loss": 1.1532, + "step": 4275 + }, + { + "epoch": 0.28, + "learning_rate": 1.910410270722858e-05, + "loss": 1.1014, + "step": 4276 + }, + { + "epoch": 0.28, + "learning_rate": 1.910388801820563e-05, + "loss": 1.207, + "step": 4277 + }, + { + "epoch": 0.28, + "learning_rate": 1.9103673329182683e-05, + "loss": 1.0746, + "step": 4278 + }, + { + "epoch": 0.28, + "learning_rate": 1.9103458640159732e-05, + "loss": 1.1504, + "step": 4279 + }, + { + "epoch": 0.28, + "learning_rate": 1.910324395113678e-05, + "loss": 1.1513, + "step": 4280 + }, + { + "epoch": 0.28, + "learning_rate": 1.910302926211383e-05, + "loss": 1.191, + "step": 4281 + }, + { + "epoch": 0.28, + "learning_rate": 1.910281457309088e-05, + "loss": 1.0918, + "step": 4282 + }, + { + "epoch": 0.28, + "learning_rate": 1.910259988406793e-05, + "loss": 1.1394, + "step": 4283 + }, + { + "epoch": 0.28, + "learning_rate": 1.910238519504498e-05, + "loss": 1.1381, + "step": 4284 + }, + { + "epoch": 0.28, + "learning_rate": 1.910217050602203e-05, + "loss": 1.1743, + "step": 4285 + }, + { + "epoch": 0.28, + "learning_rate": 1.9101955816999078e-05, + "loss": 1.1616, + "step": 4286 + }, + { + "epoch": 0.28, + "learning_rate": 1.9101741127976127e-05, + "loss": 1.1013, + "step": 4287 + }, + { + "epoch": 0.28, + "learning_rate": 1.9101526438953176e-05, + "loss": 1.1189, + "step": 4288 + }, + { + "epoch": 0.28, + "learning_rate": 1.9101311749930228e-05, + "loss": 1.0853, + "step": 4289 + }, + { + "epoch": 0.28, + "learning_rate": 1.9101097060907277e-05, + "loss": 1.1951, + "step": 4290 + }, + { + "epoch": 0.28, + "learning_rate": 1.9100882371884326e-05, + "loss": 1.1015, + "step": 4291 + }, + { + "epoch": 0.28, + "learning_rate": 1.910066768286138e-05, + "loss": 1.2401, + "step": 4292 + }, + { + "epoch": 0.28, + "learning_rate": 1.9100452993838427e-05, + "loss": 1.1217, + "step": 4293 + }, + { + "epoch": 0.28, + "learning_rate": 1.9100238304815476e-05, + "loss": 1.0292, + "step": 4294 + }, + { + "epoch": 0.28, + "learning_rate": 1.9100023615792525e-05, + "loss": 1.1596, + "step": 4295 + }, + { + "epoch": 0.28, + "learning_rate": 1.9099808926769578e-05, + "loss": 1.1358, + "step": 4296 + }, + { + "epoch": 0.28, + "learning_rate": 1.9099594237746627e-05, + "loss": 1.1315, + "step": 4297 + }, + { + "epoch": 0.28, + "learning_rate": 1.9099379548723676e-05, + "loss": 0.991, + "step": 4298 + }, + { + "epoch": 0.28, + "learning_rate": 1.9099164859700725e-05, + "loss": 1.182, + "step": 4299 + }, + { + "epoch": 0.28, + "learning_rate": 1.9098950170677774e-05, + "loss": 1.1432, + "step": 4300 + }, + { + "epoch": 0.28, + "learning_rate": 1.9098735481654823e-05, + "loss": 1.193, + "step": 4301 + }, + { + "epoch": 0.28, + "learning_rate": 1.9098520792631875e-05, + "loss": 1.0968, + "step": 4302 + }, + { + "epoch": 0.28, + "learning_rate": 1.9098306103608924e-05, + "loss": 1.0942, + "step": 4303 + }, + { + "epoch": 0.28, + "learning_rate": 1.9098091414585973e-05, + "loss": 1.1042, + "step": 4304 + }, + { + "epoch": 0.28, + "learning_rate": 1.9097876725563022e-05, + "loss": 1.1483, + "step": 4305 + }, + { + "epoch": 0.28, + "learning_rate": 1.909766203654007e-05, + "loss": 1.1789, + "step": 4306 + }, + { + "epoch": 0.28, + "learning_rate": 1.9097447347517123e-05, + "loss": 1.1151, + "step": 4307 + }, + { + "epoch": 0.28, + "learning_rate": 1.9097232658494172e-05, + "loss": 1.1285, + "step": 4308 + }, + { + "epoch": 0.28, + "learning_rate": 1.9097017969471224e-05, + "loss": 1.1019, + "step": 4309 + }, + { + "epoch": 0.28, + "learning_rate": 1.9096803280448273e-05, + "loss": 1.1071, + "step": 4310 + }, + { + "epoch": 0.28, + "learning_rate": 1.9096588591425322e-05, + "loss": 1.1555, + "step": 4311 + }, + { + "epoch": 0.28, + "learning_rate": 1.909637390240237e-05, + "loss": 1.1575, + "step": 4312 + }, + { + "epoch": 0.28, + "learning_rate": 1.909615921337942e-05, + "loss": 1.1508, + "step": 4313 + }, + { + "epoch": 0.28, + "learning_rate": 1.9095944524356473e-05, + "loss": 1.2459, + "step": 4314 + }, + { + "epoch": 0.28, + "learning_rate": 1.909572983533352e-05, + "loss": 1.1226, + "step": 4315 + }, + { + "epoch": 0.28, + "learning_rate": 1.909551514631057e-05, + "loss": 1.1578, + "step": 4316 + }, + { + "epoch": 0.28, + "learning_rate": 1.909530045728762e-05, + "loss": 1.0513, + "step": 4317 + }, + { + "epoch": 0.28, + "learning_rate": 1.909508576826467e-05, + "loss": 1.0823, + "step": 4318 + }, + { + "epoch": 0.28, + "learning_rate": 1.9094871079241718e-05, + "loss": 1.1267, + "step": 4319 + }, + { + "epoch": 0.28, + "learning_rate": 1.909465639021877e-05, + "loss": 1.0861, + "step": 4320 + }, + { + "epoch": 0.28, + "learning_rate": 1.909444170119582e-05, + "loss": 1.0401, + "step": 4321 + }, + { + "epoch": 0.28, + "learning_rate": 1.909422701217287e-05, + "loss": 1.1007, + "step": 4322 + }, + { + "epoch": 0.28, + "learning_rate": 1.909401232314992e-05, + "loss": 1.1745, + "step": 4323 + }, + { + "epoch": 0.28, + "learning_rate": 1.909379763412697e-05, + "loss": 1.0452, + "step": 4324 + }, + { + "epoch": 0.28, + "learning_rate": 1.9093582945104018e-05, + "loss": 1.1554, + "step": 4325 + }, + { + "epoch": 0.28, + "learning_rate": 1.9093368256081067e-05, + "loss": 1.1565, + "step": 4326 + }, + { + "epoch": 0.28, + "learning_rate": 1.909315356705812e-05, + "loss": 1.0992, + "step": 4327 + }, + { + "epoch": 0.28, + "learning_rate": 1.909293887803517e-05, + "loss": 1.1872, + "step": 4328 + }, + { + "epoch": 0.28, + "learning_rate": 1.9092724189012217e-05, + "loss": 1.1632, + "step": 4329 + }, + { + "epoch": 0.28, + "learning_rate": 1.9092509499989266e-05, + "loss": 1.1174, + "step": 4330 + }, + { + "epoch": 0.28, + "learning_rate": 1.9092294810966315e-05, + "loss": 1.0633, + "step": 4331 + }, + { + "epoch": 0.28, + "learning_rate": 1.9092080121943368e-05, + "loss": 1.0799, + "step": 4332 + }, + { + "epoch": 0.28, + "learning_rate": 1.9091865432920417e-05, + "loss": 1.305, + "step": 4333 + }, + { + "epoch": 0.28, + "learning_rate": 1.9091650743897466e-05, + "loss": 1.1701, + "step": 4334 + }, + { + "epoch": 0.28, + "learning_rate": 1.9091436054874515e-05, + "loss": 1.2279, + "step": 4335 + }, + { + "epoch": 0.28, + "learning_rate": 1.9091221365851567e-05, + "loss": 1.0515, + "step": 4336 + }, + { + "epoch": 0.28, + "learning_rate": 1.9091006676828616e-05, + "loss": 1.1361, + "step": 4337 + }, + { + "epoch": 0.28, + "learning_rate": 1.9090791987805665e-05, + "loss": 1.1691, + "step": 4338 + }, + { + "epoch": 0.28, + "learning_rate": 1.9090577298782714e-05, + "loss": 1.1024, + "step": 4339 + }, + { + "epoch": 0.28, + "learning_rate": 1.9090362609759766e-05, + "loss": 1.2388, + "step": 4340 + }, + { + "epoch": 0.28, + "learning_rate": 1.9090147920736815e-05, + "loss": 1.1512, + "step": 4341 + }, + { + "epoch": 0.28, + "learning_rate": 1.9089933231713864e-05, + "loss": 1.0656, + "step": 4342 + }, + { + "epoch": 0.28, + "learning_rate": 1.9089718542690913e-05, + "loss": 1.084, + "step": 4343 + }, + { + "epoch": 0.28, + "learning_rate": 1.9089503853667962e-05, + "loss": 1.1535, + "step": 4344 + }, + { + "epoch": 0.28, + "learning_rate": 1.9089289164645014e-05, + "loss": 1.2256, + "step": 4345 + }, + { + "epoch": 0.28, + "learning_rate": 1.9089074475622063e-05, + "loss": 1.1707, + "step": 4346 + }, + { + "epoch": 0.28, + "learning_rate": 1.9088859786599112e-05, + "loss": 1.0241, + "step": 4347 + }, + { + "epoch": 0.28, + "learning_rate": 1.908864509757616e-05, + "loss": 1.1396, + "step": 4348 + }, + { + "epoch": 0.28, + "learning_rate": 1.908843040855321e-05, + "loss": 1.1262, + "step": 4349 + }, + { + "epoch": 0.28, + "learning_rate": 1.9088215719530263e-05, + "loss": 1.1167, + "step": 4350 + }, + { + "epoch": 0.28, + "learning_rate": 1.908800103050731e-05, + "loss": 1.1303, + "step": 4351 + }, + { + "epoch": 0.28, + "learning_rate": 1.9087786341484364e-05, + "loss": 1.0957, + "step": 4352 + }, + { + "epoch": 0.28, + "learning_rate": 1.9087571652461413e-05, + "loss": 1.1624, + "step": 4353 + }, + { + "epoch": 0.28, + "learning_rate": 1.9087356963438462e-05, + "loss": 1.1462, + "step": 4354 + }, + { + "epoch": 0.28, + "learning_rate": 1.908714227441551e-05, + "loss": 1.0176, + "step": 4355 + }, + { + "epoch": 0.28, + "learning_rate": 1.908692758539256e-05, + "loss": 1.1852, + "step": 4356 + }, + { + "epoch": 0.28, + "learning_rate": 1.908671289636961e-05, + "loss": 1.0842, + "step": 4357 + }, + { + "epoch": 0.28, + "learning_rate": 1.908649820734666e-05, + "loss": 1.0967, + "step": 4358 + }, + { + "epoch": 0.28, + "learning_rate": 1.908628351832371e-05, + "loss": 1.1223, + "step": 4359 + }, + { + "epoch": 0.28, + "learning_rate": 1.908606882930076e-05, + "loss": 1.1038, + "step": 4360 + }, + { + "epoch": 0.28, + "learning_rate": 1.9085854140277808e-05, + "loss": 1.1462, + "step": 4361 + }, + { + "epoch": 0.28, + "learning_rate": 1.9085639451254857e-05, + "loss": 1.1394, + "step": 4362 + }, + { + "epoch": 0.28, + "learning_rate": 1.908542476223191e-05, + "loss": 1.0354, + "step": 4363 + }, + { + "epoch": 0.28, + "learning_rate": 1.9085210073208958e-05, + "loss": 1.1558, + "step": 4364 + }, + { + "epoch": 0.28, + "learning_rate": 1.908499538418601e-05, + "loss": 1.0468, + "step": 4365 + }, + { + "epoch": 0.28, + "learning_rate": 1.908478069516306e-05, + "loss": 1.1362, + "step": 4366 + }, + { + "epoch": 0.28, + "learning_rate": 1.908456600614011e-05, + "loss": 1.1393, + "step": 4367 + }, + { + "epoch": 0.28, + "learning_rate": 1.9084351317117157e-05, + "loss": 1.1701, + "step": 4368 + }, + { + "epoch": 0.28, + "learning_rate": 1.9084136628094206e-05, + "loss": 1.2427, + "step": 4369 + }, + { + "epoch": 0.28, + "learning_rate": 1.908392193907126e-05, + "loss": 1.0831, + "step": 4370 + }, + { + "epoch": 0.28, + "learning_rate": 1.9083707250048308e-05, + "loss": 1.1324, + "step": 4371 + }, + { + "epoch": 0.28, + "learning_rate": 1.9083492561025357e-05, + "loss": 1.0642, + "step": 4372 + }, + { + "epoch": 0.28, + "learning_rate": 1.9083277872002406e-05, + "loss": 1.1503, + "step": 4373 + }, + { + "epoch": 0.28, + "learning_rate": 1.9083063182979455e-05, + "loss": 1.0854, + "step": 4374 + }, + { + "epoch": 0.28, + "learning_rate": 1.9082848493956504e-05, + "loss": 1.1177, + "step": 4375 + }, + { + "epoch": 0.28, + "learning_rate": 1.9082633804933556e-05, + "loss": 1.061, + "step": 4376 + }, + { + "epoch": 0.28, + "learning_rate": 1.9082419115910605e-05, + "loss": 1.0605, + "step": 4377 + }, + { + "epoch": 0.28, + "learning_rate": 1.9082204426887654e-05, + "loss": 1.064, + "step": 4378 + }, + { + "epoch": 0.28, + "learning_rate": 1.9081989737864703e-05, + "loss": 1.1607, + "step": 4379 + }, + { + "epoch": 0.28, + "learning_rate": 1.9081775048841755e-05, + "loss": 1.0606, + "step": 4380 + }, + { + "epoch": 0.28, + "learning_rate": 1.9081560359818804e-05, + "loss": 1.0928, + "step": 4381 + }, + { + "epoch": 0.28, + "learning_rate": 1.9081345670795853e-05, + "loss": 1.1654, + "step": 4382 + }, + { + "epoch": 0.28, + "learning_rate": 1.9081130981772905e-05, + "loss": 1.0837, + "step": 4383 + }, + { + "epoch": 0.28, + "learning_rate": 1.9080916292749954e-05, + "loss": 1.1279, + "step": 4384 + }, + { + "epoch": 0.28, + "learning_rate": 1.9080701603727003e-05, + "loss": 1.1251, + "step": 4385 + }, + { + "epoch": 0.28, + "learning_rate": 1.9080486914704052e-05, + "loss": 1.1635, + "step": 4386 + }, + { + "epoch": 0.28, + "learning_rate": 1.90802722256811e-05, + "loss": 1.0708, + "step": 4387 + }, + { + "epoch": 0.28, + "learning_rate": 1.908005753665815e-05, + "loss": 1.155, + "step": 4388 + }, + { + "epoch": 0.28, + "learning_rate": 1.9079842847635203e-05, + "loss": 1.1163, + "step": 4389 + }, + { + "epoch": 0.28, + "learning_rate": 1.907962815861225e-05, + "loss": 1.1978, + "step": 4390 + }, + { + "epoch": 0.28, + "learning_rate": 1.90794134695893e-05, + "loss": 1.0624, + "step": 4391 + }, + { + "epoch": 0.28, + "learning_rate": 1.907919878056635e-05, + "loss": 1.1655, + "step": 4392 + }, + { + "epoch": 0.28, + "learning_rate": 1.90789840915434e-05, + "loss": 1.138, + "step": 4393 + }, + { + "epoch": 0.28, + "learning_rate": 1.907876940252045e-05, + "loss": 1.1268, + "step": 4394 + }, + { + "epoch": 0.28, + "learning_rate": 1.90785547134975e-05, + "loss": 1.1075, + "step": 4395 + }, + { + "epoch": 0.28, + "learning_rate": 1.9078340024474552e-05, + "loss": 1.0851, + "step": 4396 + }, + { + "epoch": 0.28, + "learning_rate": 1.90781253354516e-05, + "loss": 1.1304, + "step": 4397 + }, + { + "epoch": 0.28, + "learning_rate": 1.907791064642865e-05, + "loss": 1.1173, + "step": 4398 + }, + { + "epoch": 0.28, + "learning_rate": 1.90776959574057e-05, + "loss": 1.2021, + "step": 4399 + }, + { + "epoch": 0.28, + "learning_rate": 1.9077481268382748e-05, + "loss": 1.083, + "step": 4400 + }, + { + "epoch": 0.28, + "learning_rate": 1.90772665793598e-05, + "loss": 1.1999, + "step": 4401 + }, + { + "epoch": 0.28, + "learning_rate": 1.907705189033685e-05, + "loss": 1.0429, + "step": 4402 + }, + { + "epoch": 0.28, + "learning_rate": 1.90768372013139e-05, + "loss": 1.0672, + "step": 4403 + }, + { + "epoch": 0.28, + "learning_rate": 1.9076622512290947e-05, + "loss": 1.1665, + "step": 4404 + }, + { + "epoch": 0.28, + "learning_rate": 1.9076407823267996e-05, + "loss": 1.0115, + "step": 4405 + }, + { + "epoch": 0.28, + "learning_rate": 1.9076193134245045e-05, + "loss": 1.1208, + "step": 4406 + }, + { + "epoch": 0.28, + "learning_rate": 1.9075978445222098e-05, + "loss": 1.1412, + "step": 4407 + }, + { + "epoch": 0.28, + "learning_rate": 1.9075763756199147e-05, + "loss": 1.0663, + "step": 4408 + }, + { + "epoch": 0.28, + "learning_rate": 1.90755490671762e-05, + "loss": 1.1936, + "step": 4409 + }, + { + "epoch": 0.28, + "learning_rate": 1.9075334378153248e-05, + "loss": 1.1343, + "step": 4410 + }, + { + "epoch": 0.28, + "learning_rate": 1.9075119689130297e-05, + "loss": 1.205, + "step": 4411 + }, + { + "epoch": 0.28, + "learning_rate": 1.9074905000107346e-05, + "loss": 1.1824, + "step": 4412 + }, + { + "epoch": 0.28, + "learning_rate": 1.9074690311084395e-05, + "loss": 1.1425, + "step": 4413 + }, + { + "epoch": 0.28, + "learning_rate": 1.9074475622061447e-05, + "loss": 1.0844, + "step": 4414 + }, + { + "epoch": 0.28, + "learning_rate": 1.9074260933038496e-05, + "loss": 1.1043, + "step": 4415 + }, + { + "epoch": 0.28, + "learning_rate": 1.9074046244015545e-05, + "loss": 1.1157, + "step": 4416 + }, + { + "epoch": 0.28, + "learning_rate": 1.9073831554992594e-05, + "loss": 1.145, + "step": 4417 + }, + { + "epoch": 0.28, + "learning_rate": 1.9073616865969643e-05, + "loss": 1.2547, + "step": 4418 + }, + { + "epoch": 0.28, + "learning_rate": 1.9073402176946695e-05, + "loss": 1.1133, + "step": 4419 + }, + { + "epoch": 0.28, + "learning_rate": 1.9073187487923744e-05, + "loss": 1.1504, + "step": 4420 + }, + { + "epoch": 0.28, + "learning_rate": 1.9072972798900793e-05, + "loss": 1.0857, + "step": 4421 + }, + { + "epoch": 0.28, + "learning_rate": 1.9072758109877842e-05, + "loss": 1.169, + "step": 4422 + }, + { + "epoch": 0.28, + "learning_rate": 1.9072543420854895e-05, + "loss": 1.0799, + "step": 4423 + }, + { + "epoch": 0.28, + "learning_rate": 1.9072328731831944e-05, + "loss": 1.0824, + "step": 4424 + }, + { + "epoch": 0.28, + "learning_rate": 1.9072114042808992e-05, + "loss": 1.1657, + "step": 4425 + }, + { + "epoch": 0.28, + "learning_rate": 1.907189935378604e-05, + "loss": 1.1604, + "step": 4426 + }, + { + "epoch": 0.28, + "learning_rate": 1.9071684664763094e-05, + "loss": 1.0925, + "step": 4427 + }, + { + "epoch": 0.28, + "learning_rate": 1.9071469975740143e-05, + "loss": 1.0802, + "step": 4428 + }, + { + "epoch": 0.28, + "learning_rate": 1.9071255286717192e-05, + "loss": 1.1244, + "step": 4429 + }, + { + "epoch": 0.29, + "learning_rate": 1.907104059769424e-05, + "loss": 1.1354, + "step": 4430 + }, + { + "epoch": 0.29, + "learning_rate": 1.907082590867129e-05, + "loss": 1.0227, + "step": 4431 + }, + { + "epoch": 0.29, + "learning_rate": 1.9070611219648342e-05, + "loss": 1.1119, + "step": 4432 + }, + { + "epoch": 0.29, + "learning_rate": 1.907039653062539e-05, + "loss": 1.1148, + "step": 4433 + }, + { + "epoch": 0.29, + "learning_rate": 1.907018184160244e-05, + "loss": 0.9941, + "step": 4434 + }, + { + "epoch": 0.29, + "learning_rate": 1.906996715257949e-05, + "loss": 1.1325, + "step": 4435 + }, + { + "epoch": 0.29, + "learning_rate": 1.9069752463556538e-05, + "loss": 1.0991, + "step": 4436 + }, + { + "epoch": 0.29, + "learning_rate": 1.906953777453359e-05, + "loss": 1.1013, + "step": 4437 + }, + { + "epoch": 0.29, + "learning_rate": 1.906932308551064e-05, + "loss": 1.163, + "step": 4438 + }, + { + "epoch": 0.29, + "learning_rate": 1.906910839648769e-05, + "loss": 1.2469, + "step": 4439 + }, + { + "epoch": 0.29, + "learning_rate": 1.906889370746474e-05, + "loss": 1.1247, + "step": 4440 + }, + { + "epoch": 0.29, + "learning_rate": 1.906867901844179e-05, + "loss": 1.1152, + "step": 4441 + }, + { + "epoch": 0.29, + "learning_rate": 1.906846432941884e-05, + "loss": 1.0659, + "step": 4442 + }, + { + "epoch": 0.29, + "learning_rate": 1.9068249640395887e-05, + "loss": 1.1611, + "step": 4443 + }, + { + "epoch": 0.29, + "learning_rate": 1.9068034951372936e-05, + "loss": 1.1229, + "step": 4444 + }, + { + "epoch": 0.29, + "learning_rate": 1.906782026234999e-05, + "loss": 1.1463, + "step": 4445 + }, + { + "epoch": 0.29, + "learning_rate": 1.9067605573327038e-05, + "loss": 1.1069, + "step": 4446 + }, + { + "epoch": 0.29, + "learning_rate": 1.9067390884304087e-05, + "loss": 1.0656, + "step": 4447 + }, + { + "epoch": 0.29, + "learning_rate": 1.9067176195281136e-05, + "loss": 1.1062, + "step": 4448 + }, + { + "epoch": 0.29, + "learning_rate": 1.9066961506258185e-05, + "loss": 1.0602, + "step": 4449 + }, + { + "epoch": 0.29, + "learning_rate": 1.9066746817235237e-05, + "loss": 1.1898, + "step": 4450 + }, + { + "epoch": 0.29, + "learning_rate": 1.9066532128212286e-05, + "loss": 1.0927, + "step": 4451 + }, + { + "epoch": 0.29, + "learning_rate": 1.9066317439189338e-05, + "loss": 1.0893, + "step": 4452 + }, + { + "epoch": 0.29, + "learning_rate": 1.9066102750166387e-05, + "loss": 1.0767, + "step": 4453 + }, + { + "epoch": 0.29, + "learning_rate": 1.9065888061143436e-05, + "loss": 1.0902, + "step": 4454 + }, + { + "epoch": 0.29, + "learning_rate": 1.9065673372120485e-05, + "loss": 1.1413, + "step": 4455 + }, + { + "epoch": 0.29, + "learning_rate": 1.9065458683097534e-05, + "loss": 1.0813, + "step": 4456 + }, + { + "epoch": 0.29, + "learning_rate": 1.9065243994074586e-05, + "loss": 1.1245, + "step": 4457 + }, + { + "epoch": 0.29, + "learning_rate": 1.9065029305051635e-05, + "loss": 1.1715, + "step": 4458 + }, + { + "epoch": 0.29, + "learning_rate": 1.9064814616028684e-05, + "loss": 1.0704, + "step": 4459 + }, + { + "epoch": 0.29, + "learning_rate": 1.9064599927005733e-05, + "loss": 1.19, + "step": 4460 + }, + { + "epoch": 0.29, + "learning_rate": 1.9064385237982782e-05, + "loss": 1.1542, + "step": 4461 + }, + { + "epoch": 0.29, + "learning_rate": 1.906417054895983e-05, + "loss": 1.1681, + "step": 4462 + }, + { + "epoch": 0.29, + "learning_rate": 1.9063955859936884e-05, + "loss": 1.0888, + "step": 4463 + }, + { + "epoch": 0.29, + "learning_rate": 1.9063741170913933e-05, + "loss": 1.0972, + "step": 4464 + }, + { + "epoch": 0.29, + "learning_rate": 1.906352648189098e-05, + "loss": 1.095, + "step": 4465 + }, + { + "epoch": 0.29, + "learning_rate": 1.906331179286803e-05, + "loss": 1.1591, + "step": 4466 + }, + { + "epoch": 0.29, + "learning_rate": 1.9063097103845083e-05, + "loss": 1.0873, + "step": 4467 + }, + { + "epoch": 0.29, + "learning_rate": 1.9062882414822132e-05, + "loss": 1.2293, + "step": 4468 + }, + { + "epoch": 0.29, + "learning_rate": 1.906266772579918e-05, + "loss": 1.0611, + "step": 4469 + }, + { + "epoch": 0.29, + "learning_rate": 1.9062453036776233e-05, + "loss": 1.1548, + "step": 4470 + }, + { + "epoch": 0.29, + "learning_rate": 1.9062238347753282e-05, + "loss": 1.0644, + "step": 4471 + }, + { + "epoch": 0.29, + "learning_rate": 1.906202365873033e-05, + "loss": 1.1537, + "step": 4472 + }, + { + "epoch": 0.29, + "learning_rate": 1.906180896970738e-05, + "loss": 1.1213, + "step": 4473 + }, + { + "epoch": 0.29, + "learning_rate": 1.906159428068443e-05, + "loss": 1.0663, + "step": 4474 + }, + { + "epoch": 0.29, + "learning_rate": 1.9061379591661478e-05, + "loss": 1.0143, + "step": 4475 + }, + { + "epoch": 0.29, + "learning_rate": 1.906116490263853e-05, + "loss": 1.0942, + "step": 4476 + }, + { + "epoch": 0.29, + "learning_rate": 1.906095021361558e-05, + "loss": 1.1292, + "step": 4477 + }, + { + "epoch": 0.29, + "learning_rate": 1.9060735524592628e-05, + "loss": 1.0504, + "step": 4478 + }, + { + "epoch": 0.29, + "learning_rate": 1.9060520835569677e-05, + "loss": 1.1335, + "step": 4479 + }, + { + "epoch": 0.29, + "learning_rate": 1.9060306146546726e-05, + "loss": 1.1268, + "step": 4480 + }, + { + "epoch": 0.29, + "learning_rate": 1.906009145752378e-05, + "loss": 1.2243, + "step": 4481 + }, + { + "epoch": 0.29, + "learning_rate": 1.9059876768500828e-05, + "loss": 1.2321, + "step": 4482 + }, + { + "epoch": 0.29, + "learning_rate": 1.905966207947788e-05, + "loss": 1.0847, + "step": 4483 + }, + { + "epoch": 0.29, + "learning_rate": 1.905944739045493e-05, + "loss": 1.1195, + "step": 4484 + }, + { + "epoch": 0.29, + "learning_rate": 1.9059232701431978e-05, + "loss": 1.0722, + "step": 4485 + }, + { + "epoch": 0.29, + "learning_rate": 1.9059018012409027e-05, + "loss": 1.1006, + "step": 4486 + }, + { + "epoch": 0.29, + "learning_rate": 1.9058803323386076e-05, + "loss": 1.0858, + "step": 4487 + }, + { + "epoch": 0.29, + "learning_rate": 1.9058588634363128e-05, + "loss": 1.1509, + "step": 4488 + }, + { + "epoch": 0.29, + "learning_rate": 1.9058373945340177e-05, + "loss": 1.1438, + "step": 4489 + }, + { + "epoch": 0.29, + "learning_rate": 1.9058159256317226e-05, + "loss": 1.1236, + "step": 4490 + }, + { + "epoch": 0.29, + "learning_rate": 1.9057944567294275e-05, + "loss": 1.0259, + "step": 4491 + }, + { + "epoch": 0.29, + "learning_rate": 1.9057729878271324e-05, + "loss": 1.1714, + "step": 4492 + }, + { + "epoch": 0.29, + "learning_rate": 1.9057515189248373e-05, + "loss": 1.036, + "step": 4493 + }, + { + "epoch": 0.29, + "learning_rate": 1.9057300500225425e-05, + "loss": 1.1169, + "step": 4494 + }, + { + "epoch": 0.29, + "learning_rate": 1.9057085811202474e-05, + "loss": 1.122, + "step": 4495 + }, + { + "epoch": 0.29, + "learning_rate": 1.9056871122179527e-05, + "loss": 1.0505, + "step": 4496 + }, + { + "epoch": 0.29, + "learning_rate": 1.9056656433156576e-05, + "loss": 1.1217, + "step": 4497 + }, + { + "epoch": 0.29, + "learning_rate": 1.9056441744133625e-05, + "loss": 1.1442, + "step": 4498 + }, + { + "epoch": 0.29, + "learning_rate": 1.9056227055110674e-05, + "loss": 1.0639, + "step": 4499 + }, + { + "epoch": 0.29, + "learning_rate": 1.9056012366087722e-05, + "loss": 1.1743, + "step": 4500 + }, + { + "epoch": 0.29, + "learning_rate": 1.9055797677064775e-05, + "loss": 1.0557, + "step": 4501 + }, + { + "epoch": 0.29, + "learning_rate": 1.9055582988041824e-05, + "loss": 1.1289, + "step": 4502 + }, + { + "epoch": 0.29, + "learning_rate": 1.9055368299018873e-05, + "loss": 1.0825, + "step": 4503 + }, + { + "epoch": 0.29, + "learning_rate": 1.9055153609995922e-05, + "loss": 1.1079, + "step": 4504 + }, + { + "epoch": 0.29, + "learning_rate": 1.905493892097297e-05, + "loss": 1.1719, + "step": 4505 + }, + { + "epoch": 0.29, + "learning_rate": 1.9054724231950023e-05, + "loss": 1.1598, + "step": 4506 + }, + { + "epoch": 0.29, + "learning_rate": 1.9054509542927072e-05, + "loss": 1.1108, + "step": 4507 + }, + { + "epoch": 0.29, + "learning_rate": 1.905429485390412e-05, + "loss": 1.0569, + "step": 4508 + }, + { + "epoch": 0.29, + "learning_rate": 1.905408016488117e-05, + "loss": 1.1818, + "step": 4509 + }, + { + "epoch": 0.29, + "learning_rate": 1.9053865475858222e-05, + "loss": 1.0654, + "step": 4510 + }, + { + "epoch": 0.29, + "learning_rate": 1.905365078683527e-05, + "loss": 1.1253, + "step": 4511 + }, + { + "epoch": 0.29, + "learning_rate": 1.905343609781232e-05, + "loss": 1.1219, + "step": 4512 + }, + { + "epoch": 0.29, + "learning_rate": 1.905322140878937e-05, + "loss": 1.2099, + "step": 4513 + }, + { + "epoch": 0.29, + "learning_rate": 1.905300671976642e-05, + "loss": 1.1633, + "step": 4514 + }, + { + "epoch": 0.29, + "learning_rate": 1.905279203074347e-05, + "loss": 1.1674, + "step": 4515 + }, + { + "epoch": 0.29, + "learning_rate": 1.905257734172052e-05, + "loss": 1.1047, + "step": 4516 + }, + { + "epoch": 0.29, + "learning_rate": 1.905236265269757e-05, + "loss": 1.1186, + "step": 4517 + }, + { + "epoch": 0.29, + "learning_rate": 1.9052147963674617e-05, + "loss": 1.1959, + "step": 4518 + }, + { + "epoch": 0.29, + "learning_rate": 1.905193327465167e-05, + "loss": 1.1239, + "step": 4519 + }, + { + "epoch": 0.29, + "learning_rate": 1.905171858562872e-05, + "loss": 1.185, + "step": 4520 + }, + { + "epoch": 0.29, + "learning_rate": 1.9051503896605768e-05, + "loss": 1.1592, + "step": 4521 + }, + { + "epoch": 0.29, + "learning_rate": 1.9051289207582817e-05, + "loss": 1.0448, + "step": 4522 + }, + { + "epoch": 0.29, + "learning_rate": 1.9051074518559866e-05, + "loss": 1.1048, + "step": 4523 + }, + { + "epoch": 0.29, + "learning_rate": 1.9050859829536915e-05, + "loss": 1.0755, + "step": 4524 + }, + { + "epoch": 0.29, + "learning_rate": 1.9050645140513967e-05, + "loss": 1.0857, + "step": 4525 + }, + { + "epoch": 0.29, + "learning_rate": 1.905043045149102e-05, + "loss": 1.1526, + "step": 4526 + }, + { + "epoch": 0.29, + "learning_rate": 1.9050215762468068e-05, + "loss": 1.1542, + "step": 4527 + }, + { + "epoch": 0.29, + "learning_rate": 1.9050001073445117e-05, + "loss": 1.1133, + "step": 4528 + }, + { + "epoch": 0.29, + "learning_rate": 1.9049786384422166e-05, + "loss": 1.0908, + "step": 4529 + }, + { + "epoch": 0.29, + "learning_rate": 1.9049571695399215e-05, + "loss": 1.2014, + "step": 4530 + }, + { + "epoch": 0.29, + "learning_rate": 1.9049357006376264e-05, + "loss": 1.1495, + "step": 4531 + }, + { + "epoch": 0.29, + "learning_rate": 1.9049142317353316e-05, + "loss": 1.1035, + "step": 4532 + }, + { + "epoch": 0.29, + "learning_rate": 1.9048927628330365e-05, + "loss": 1.2176, + "step": 4533 + }, + { + "epoch": 0.29, + "learning_rate": 1.9048712939307414e-05, + "loss": 1.1806, + "step": 4534 + }, + { + "epoch": 0.29, + "learning_rate": 1.9048498250284463e-05, + "loss": 1.1005, + "step": 4535 + }, + { + "epoch": 0.29, + "learning_rate": 1.9048283561261512e-05, + "loss": 1.1167, + "step": 4536 + }, + { + "epoch": 0.29, + "learning_rate": 1.9048068872238565e-05, + "loss": 1.0838, + "step": 4537 + }, + { + "epoch": 0.29, + "learning_rate": 1.9047854183215614e-05, + "loss": 1.14, + "step": 4538 + }, + { + "epoch": 0.29, + "learning_rate": 1.9047639494192666e-05, + "loss": 1.1045, + "step": 4539 + }, + { + "epoch": 0.29, + "learning_rate": 1.9047424805169715e-05, + "loss": 1.0856, + "step": 4540 + }, + { + "epoch": 0.29, + "learning_rate": 1.9047210116146764e-05, + "loss": 1.1655, + "step": 4541 + }, + { + "epoch": 0.29, + "learning_rate": 1.9046995427123813e-05, + "loss": 1.156, + "step": 4542 + }, + { + "epoch": 0.29, + "learning_rate": 1.9046780738100862e-05, + "loss": 1.1482, + "step": 4543 + }, + { + "epoch": 0.29, + "learning_rate": 1.9046566049077914e-05, + "loss": 1.2469, + "step": 4544 + }, + { + "epoch": 0.29, + "learning_rate": 1.9046351360054963e-05, + "loss": 1.0828, + "step": 4545 + }, + { + "epoch": 0.29, + "learning_rate": 1.9046136671032012e-05, + "loss": 1.1931, + "step": 4546 + }, + { + "epoch": 0.29, + "learning_rate": 1.904592198200906e-05, + "loss": 1.1078, + "step": 4547 + }, + { + "epoch": 0.29, + "learning_rate": 1.904570729298611e-05, + "loss": 1.1142, + "step": 4548 + }, + { + "epoch": 0.29, + "learning_rate": 1.904549260396316e-05, + "loss": 1.1227, + "step": 4549 + }, + { + "epoch": 0.29, + "learning_rate": 1.904527791494021e-05, + "loss": 1.023, + "step": 4550 + }, + { + "epoch": 0.29, + "learning_rate": 1.904506322591726e-05, + "loss": 1.0766, + "step": 4551 + }, + { + "epoch": 0.29, + "learning_rate": 1.904484853689431e-05, + "loss": 1.1963, + "step": 4552 + }, + { + "epoch": 0.29, + "learning_rate": 1.9044633847871358e-05, + "loss": 1.114, + "step": 4553 + }, + { + "epoch": 0.29, + "learning_rate": 1.904441915884841e-05, + "loss": 1.1252, + "step": 4554 + }, + { + "epoch": 0.29, + "learning_rate": 1.904420446982546e-05, + "loss": 1.0099, + "step": 4555 + }, + { + "epoch": 0.29, + "learning_rate": 1.904398978080251e-05, + "loss": 1.1015, + "step": 4556 + }, + { + "epoch": 0.29, + "learning_rate": 1.904377509177956e-05, + "loss": 1.1136, + "step": 4557 + }, + { + "epoch": 0.29, + "learning_rate": 1.904356040275661e-05, + "loss": 1.0418, + "step": 4558 + }, + { + "epoch": 0.29, + "learning_rate": 1.904334571373366e-05, + "loss": 1.1796, + "step": 4559 + }, + { + "epoch": 0.29, + "learning_rate": 1.9043131024710708e-05, + "loss": 1.2231, + "step": 4560 + }, + { + "epoch": 0.29, + "learning_rate": 1.9042916335687757e-05, + "loss": 1.1375, + "step": 4561 + }, + { + "epoch": 0.29, + "learning_rate": 1.9042701646664806e-05, + "loss": 1.1974, + "step": 4562 + }, + { + "epoch": 0.29, + "learning_rate": 1.9042486957641858e-05, + "loss": 1.1658, + "step": 4563 + }, + { + "epoch": 0.29, + "learning_rate": 1.9042272268618907e-05, + "loss": 1.1447, + "step": 4564 + }, + { + "epoch": 0.29, + "learning_rate": 1.9042057579595956e-05, + "loss": 0.9993, + "step": 4565 + }, + { + "epoch": 0.29, + "learning_rate": 1.9041842890573005e-05, + "loss": 1.0926, + "step": 4566 + }, + { + "epoch": 0.29, + "learning_rate": 1.9041628201550054e-05, + "loss": 1.1334, + "step": 4567 + }, + { + "epoch": 0.29, + "learning_rate": 1.9041413512527106e-05, + "loss": 1.095, + "step": 4568 + }, + { + "epoch": 0.29, + "learning_rate": 1.9041198823504155e-05, + "loss": 1.1806, + "step": 4569 + }, + { + "epoch": 0.29, + "learning_rate": 1.9040984134481208e-05, + "loss": 1.1585, + "step": 4570 + }, + { + "epoch": 0.29, + "learning_rate": 1.9040769445458257e-05, + "loss": 1.1222, + "step": 4571 + }, + { + "epoch": 0.29, + "learning_rate": 1.9040554756435306e-05, + "loss": 1.088, + "step": 4572 + }, + { + "epoch": 0.29, + "learning_rate": 1.9040340067412355e-05, + "loss": 1.0699, + "step": 4573 + }, + { + "epoch": 0.29, + "learning_rate": 1.9040125378389403e-05, + "loss": 1.1106, + "step": 4574 + }, + { + "epoch": 0.29, + "learning_rate": 1.9039910689366456e-05, + "loss": 1.1155, + "step": 4575 + }, + { + "epoch": 0.29, + "learning_rate": 1.9039696000343505e-05, + "loss": 1.1984, + "step": 4576 + }, + { + "epoch": 0.29, + "learning_rate": 1.9039481311320554e-05, + "loss": 1.0949, + "step": 4577 + }, + { + "epoch": 0.29, + "learning_rate": 1.9039266622297603e-05, + "loss": 1.1378, + "step": 4578 + }, + { + "epoch": 0.29, + "learning_rate": 1.903905193327465e-05, + "loss": 1.1701, + "step": 4579 + }, + { + "epoch": 0.29, + "learning_rate": 1.90388372442517e-05, + "loss": 1.1463, + "step": 4580 + }, + { + "epoch": 0.29, + "learning_rate": 1.9038622555228753e-05, + "loss": 1.091, + "step": 4581 + }, + { + "epoch": 0.29, + "learning_rate": 1.9038407866205802e-05, + "loss": 1.1669, + "step": 4582 + }, + { + "epoch": 0.29, + "learning_rate": 1.9038193177182854e-05, + "loss": 1.1168, + "step": 4583 + }, + { + "epoch": 0.29, + "learning_rate": 1.9037978488159903e-05, + "loss": 1.0346, + "step": 4584 + }, + { + "epoch": 0.29, + "learning_rate": 1.9037763799136952e-05, + "loss": 1.1222, + "step": 4585 + }, + { + "epoch": 0.3, + "learning_rate": 1.9037549110114e-05, + "loss": 1.1252, + "step": 4586 + }, + { + "epoch": 0.3, + "learning_rate": 1.903733442109105e-05, + "loss": 1.0856, + "step": 4587 + }, + { + "epoch": 0.3, + "learning_rate": 1.9037119732068103e-05, + "loss": 1.243, + "step": 4588 + }, + { + "epoch": 0.3, + "learning_rate": 1.903690504304515e-05, + "loss": 1.1029, + "step": 4589 + }, + { + "epoch": 0.3, + "learning_rate": 1.90366903540222e-05, + "loss": 1.1151, + "step": 4590 + }, + { + "epoch": 0.3, + "learning_rate": 1.903647566499925e-05, + "loss": 1.1268, + "step": 4591 + }, + { + "epoch": 0.3, + "learning_rate": 1.90362609759763e-05, + "loss": 1.1707, + "step": 4592 + }, + { + "epoch": 0.3, + "learning_rate": 1.903604628695335e-05, + "loss": 1.1473, + "step": 4593 + }, + { + "epoch": 0.3, + "learning_rate": 1.90358315979304e-05, + "loss": 1.1672, + "step": 4594 + }, + { + "epoch": 0.3, + "learning_rate": 1.903561690890745e-05, + "loss": 0.9678, + "step": 4595 + }, + { + "epoch": 0.3, + "learning_rate": 1.9035402219884498e-05, + "loss": 1.0796, + "step": 4596 + }, + { + "epoch": 0.3, + "learning_rate": 1.903518753086155e-05, + "loss": 1.1057, + "step": 4597 + }, + { + "epoch": 0.3, + "learning_rate": 1.90349728418386e-05, + "loss": 1.0411, + "step": 4598 + }, + { + "epoch": 0.3, + "learning_rate": 1.9034758152815648e-05, + "loss": 1.1287, + "step": 4599 + }, + { + "epoch": 0.3, + "learning_rate": 1.9034543463792697e-05, + "loss": 1.1134, + "step": 4600 + }, + { + "epoch": 0.3, + "learning_rate": 1.903432877476975e-05, + "loss": 1.0573, + "step": 4601 + }, + { + "epoch": 0.3, + "learning_rate": 1.9034114085746798e-05, + "loss": 1.1468, + "step": 4602 + }, + { + "epoch": 0.3, + "learning_rate": 1.9033899396723847e-05, + "loss": 1.0138, + "step": 4603 + }, + { + "epoch": 0.3, + "learning_rate": 1.9033684707700896e-05, + "loss": 1.2123, + "step": 4604 + }, + { + "epoch": 0.3, + "learning_rate": 1.9033470018677945e-05, + "loss": 1.1509, + "step": 4605 + }, + { + "epoch": 0.3, + "learning_rate": 1.9033255329654997e-05, + "loss": 1.043, + "step": 4606 + }, + { + "epoch": 0.3, + "learning_rate": 1.9033040640632046e-05, + "loss": 1.1876, + "step": 4607 + }, + { + "epoch": 0.3, + "learning_rate": 1.9032825951609095e-05, + "loss": 1.2077, + "step": 4608 + }, + { + "epoch": 0.3, + "learning_rate": 1.9032611262586144e-05, + "loss": 1.1428, + "step": 4609 + }, + { + "epoch": 0.3, + "learning_rate": 1.9032396573563193e-05, + "loss": 1.1464, + "step": 4610 + }, + { + "epoch": 0.3, + "learning_rate": 1.9032181884540242e-05, + "loss": 1.1561, + "step": 4611 + }, + { + "epoch": 0.3, + "learning_rate": 1.9031967195517295e-05, + "loss": 1.0673, + "step": 4612 + }, + { + "epoch": 0.3, + "learning_rate": 1.9031752506494347e-05, + "loss": 1.162, + "step": 4613 + }, + { + "epoch": 0.3, + "learning_rate": 1.9031537817471396e-05, + "loss": 1.1303, + "step": 4614 + }, + { + "epoch": 0.3, + "learning_rate": 1.9031323128448445e-05, + "loss": 1.1385, + "step": 4615 + }, + { + "epoch": 0.3, + "learning_rate": 1.9031108439425494e-05, + "loss": 1.1772, + "step": 4616 + }, + { + "epoch": 0.3, + "learning_rate": 1.9030893750402543e-05, + "loss": 1.0624, + "step": 4617 + }, + { + "epoch": 0.3, + "learning_rate": 1.9030679061379592e-05, + "loss": 1.176, + "step": 4618 + }, + { + "epoch": 0.3, + "learning_rate": 1.9030464372356644e-05, + "loss": 1.1651, + "step": 4619 + }, + { + "epoch": 0.3, + "learning_rate": 1.9030249683333693e-05, + "loss": 1.1088, + "step": 4620 + }, + { + "epoch": 0.3, + "learning_rate": 1.9030034994310742e-05, + "loss": 1.0997, + "step": 4621 + }, + { + "epoch": 0.3, + "learning_rate": 1.902982030528779e-05, + "loss": 1.1574, + "step": 4622 + }, + { + "epoch": 0.3, + "learning_rate": 1.902960561626484e-05, + "loss": 1.19, + "step": 4623 + }, + { + "epoch": 0.3, + "learning_rate": 1.9029390927241892e-05, + "loss": 1.125, + "step": 4624 + }, + { + "epoch": 0.3, + "learning_rate": 1.902917623821894e-05, + "loss": 1.112, + "step": 4625 + }, + { + "epoch": 0.3, + "learning_rate": 1.9028961549195994e-05, + "loss": 1.1094, + "step": 4626 + }, + { + "epoch": 0.3, + "learning_rate": 1.9028746860173043e-05, + "loss": 1.08, + "step": 4627 + }, + { + "epoch": 0.3, + "learning_rate": 1.902853217115009e-05, + "loss": 1.0683, + "step": 4628 + }, + { + "epoch": 0.3, + "learning_rate": 1.902831748212714e-05, + "loss": 1.0773, + "step": 4629 + }, + { + "epoch": 0.3, + "learning_rate": 1.902810279310419e-05, + "loss": 1.119, + "step": 4630 + }, + { + "epoch": 0.3, + "learning_rate": 1.902788810408124e-05, + "loss": 1.1372, + "step": 4631 + }, + { + "epoch": 0.3, + "learning_rate": 1.902767341505829e-05, + "loss": 1.1444, + "step": 4632 + }, + { + "epoch": 0.3, + "learning_rate": 1.902745872603534e-05, + "loss": 1.0357, + "step": 4633 + }, + { + "epoch": 0.3, + "learning_rate": 1.902724403701239e-05, + "loss": 1.0855, + "step": 4634 + }, + { + "epoch": 0.3, + "learning_rate": 1.9027029347989438e-05, + "loss": 1.0674, + "step": 4635 + }, + { + "epoch": 0.3, + "learning_rate": 1.9026814658966487e-05, + "loss": 1.1621, + "step": 4636 + }, + { + "epoch": 0.3, + "learning_rate": 1.902659996994354e-05, + "loss": 1.118, + "step": 4637 + }, + { + "epoch": 0.3, + "learning_rate": 1.9026385280920588e-05, + "loss": 1.1023, + "step": 4638 + }, + { + "epoch": 0.3, + "learning_rate": 1.9026170591897637e-05, + "loss": 1.2138, + "step": 4639 + }, + { + "epoch": 0.3, + "learning_rate": 1.9025955902874686e-05, + "loss": 1.2174, + "step": 4640 + }, + { + "epoch": 0.3, + "learning_rate": 1.902574121385174e-05, + "loss": 1.1315, + "step": 4641 + }, + { + "epoch": 0.3, + "learning_rate": 1.9025526524828787e-05, + "loss": 1.1709, + "step": 4642 + }, + { + "epoch": 0.3, + "learning_rate": 1.9025311835805836e-05, + "loss": 1.0831, + "step": 4643 + }, + { + "epoch": 0.3, + "learning_rate": 1.902509714678289e-05, + "loss": 1.0394, + "step": 4644 + }, + { + "epoch": 0.3, + "learning_rate": 1.9024882457759938e-05, + "loss": 1.1418, + "step": 4645 + }, + { + "epoch": 0.3, + "learning_rate": 1.9024667768736987e-05, + "loss": 1.0573, + "step": 4646 + }, + { + "epoch": 0.3, + "learning_rate": 1.9024453079714036e-05, + "loss": 1.1298, + "step": 4647 + }, + { + "epoch": 0.3, + "learning_rate": 1.9024238390691084e-05, + "loss": 1.029, + "step": 4648 + }, + { + "epoch": 0.3, + "learning_rate": 1.9024023701668133e-05, + "loss": 1.0371, + "step": 4649 + }, + { + "epoch": 0.3, + "learning_rate": 1.9023809012645186e-05, + "loss": 1.136, + "step": 4650 + }, + { + "epoch": 0.3, + "learning_rate": 1.9023594323622235e-05, + "loss": 1.1341, + "step": 4651 + }, + { + "epoch": 0.3, + "learning_rate": 1.9023379634599284e-05, + "loss": 1.1981, + "step": 4652 + }, + { + "epoch": 0.3, + "learning_rate": 1.9023164945576333e-05, + "loss": 1.1176, + "step": 4653 + }, + { + "epoch": 0.3, + "learning_rate": 1.902295025655338e-05, + "loss": 0.9523, + "step": 4654 + }, + { + "epoch": 0.3, + "learning_rate": 1.9022735567530434e-05, + "loss": 1.1102, + "step": 4655 + }, + { + "epoch": 0.3, + "learning_rate": 1.9022520878507483e-05, + "loss": 1.1598, + "step": 4656 + }, + { + "epoch": 0.3, + "learning_rate": 1.9022306189484535e-05, + "loss": 1.1719, + "step": 4657 + }, + { + "epoch": 0.3, + "learning_rate": 1.9022091500461584e-05, + "loss": 1.1047, + "step": 4658 + }, + { + "epoch": 0.3, + "learning_rate": 1.9021876811438633e-05, + "loss": 1.1291, + "step": 4659 + }, + { + "epoch": 0.3, + "learning_rate": 1.9021662122415682e-05, + "loss": 1.0906, + "step": 4660 + }, + { + "epoch": 0.3, + "learning_rate": 1.902144743339273e-05, + "loss": 1.1151, + "step": 4661 + }, + { + "epoch": 0.3, + "learning_rate": 1.9021232744369784e-05, + "loss": 1.0078, + "step": 4662 + }, + { + "epoch": 0.3, + "learning_rate": 1.9021018055346833e-05, + "loss": 1.0207, + "step": 4663 + }, + { + "epoch": 0.3, + "learning_rate": 1.902080336632388e-05, + "loss": 1.0864, + "step": 4664 + }, + { + "epoch": 0.3, + "learning_rate": 1.902058867730093e-05, + "loss": 1.1086, + "step": 4665 + }, + { + "epoch": 0.3, + "learning_rate": 1.902037398827798e-05, + "loss": 1.0949, + "step": 4666 + }, + { + "epoch": 0.3, + "learning_rate": 1.902015929925503e-05, + "loss": 1.0521, + "step": 4667 + }, + { + "epoch": 0.3, + "learning_rate": 1.901994461023208e-05, + "loss": 1.1411, + "step": 4668 + }, + { + "epoch": 0.3, + "learning_rate": 1.901972992120913e-05, + "loss": 1.1254, + "step": 4669 + }, + { + "epoch": 0.3, + "learning_rate": 1.9019515232186182e-05, + "loss": 1.1699, + "step": 4670 + }, + { + "epoch": 0.3, + "learning_rate": 1.901930054316323e-05, + "loss": 1.0263, + "step": 4671 + }, + { + "epoch": 0.3, + "learning_rate": 1.901908585414028e-05, + "loss": 1.1014, + "step": 4672 + }, + { + "epoch": 0.3, + "learning_rate": 1.901887116511733e-05, + "loss": 1.2298, + "step": 4673 + }, + { + "epoch": 0.3, + "learning_rate": 1.9018656476094378e-05, + "loss": 1.1154, + "step": 4674 + }, + { + "epoch": 0.3, + "learning_rate": 1.901844178707143e-05, + "loss": 1.0763, + "step": 4675 + }, + { + "epoch": 0.3, + "learning_rate": 1.901822709804848e-05, + "loss": 1.0864, + "step": 4676 + }, + { + "epoch": 0.3, + "learning_rate": 1.9018012409025528e-05, + "loss": 1.0854, + "step": 4677 + }, + { + "epoch": 0.3, + "learning_rate": 1.9017797720002577e-05, + "loss": 1.1303, + "step": 4678 + }, + { + "epoch": 0.3, + "learning_rate": 1.9017583030979626e-05, + "loss": 1.0256, + "step": 4679 + }, + { + "epoch": 0.3, + "learning_rate": 1.901736834195668e-05, + "loss": 1.0839, + "step": 4680 + }, + { + "epoch": 0.3, + "learning_rate": 1.9017153652933727e-05, + "loss": 1.2053, + "step": 4681 + }, + { + "epoch": 0.3, + "learning_rate": 1.9016938963910776e-05, + "loss": 1.0406, + "step": 4682 + }, + { + "epoch": 0.3, + "learning_rate": 1.9016724274887825e-05, + "loss": 1.1677, + "step": 4683 + }, + { + "epoch": 0.3, + "learning_rate": 1.9016509585864878e-05, + "loss": 1.1619, + "step": 4684 + }, + { + "epoch": 0.3, + "learning_rate": 1.9016294896841927e-05, + "loss": 1.1429, + "step": 4685 + }, + { + "epoch": 0.3, + "learning_rate": 1.9016080207818976e-05, + "loss": 1.1093, + "step": 4686 + }, + { + "epoch": 0.3, + "learning_rate": 1.9015865518796025e-05, + "loss": 1.0914, + "step": 4687 + }, + { + "epoch": 0.3, + "learning_rate": 1.9015650829773077e-05, + "loss": 1.1283, + "step": 4688 + }, + { + "epoch": 0.3, + "learning_rate": 1.9015436140750126e-05, + "loss": 1.1366, + "step": 4689 + }, + { + "epoch": 0.3, + "learning_rate": 1.9015221451727175e-05, + "loss": 1.1275, + "step": 4690 + }, + { + "epoch": 0.3, + "learning_rate": 1.9015006762704224e-05, + "loss": 1.2194, + "step": 4691 + }, + { + "epoch": 0.3, + "learning_rate": 1.9014792073681273e-05, + "loss": 1.1316, + "step": 4692 + }, + { + "epoch": 0.3, + "learning_rate": 1.9014577384658325e-05, + "loss": 1.1453, + "step": 4693 + }, + { + "epoch": 0.3, + "learning_rate": 1.9014362695635374e-05, + "loss": 1.1106, + "step": 4694 + }, + { + "epoch": 0.3, + "learning_rate": 1.9014148006612423e-05, + "loss": 1.0968, + "step": 4695 + }, + { + "epoch": 0.3, + "learning_rate": 1.9013933317589472e-05, + "loss": 1.1124, + "step": 4696 + }, + { + "epoch": 0.3, + "learning_rate": 1.901371862856652e-05, + "loss": 1.1953, + "step": 4697 + }, + { + "epoch": 0.3, + "learning_rate": 1.901350393954357e-05, + "loss": 1.1208, + "step": 4698 + }, + { + "epoch": 0.3, + "learning_rate": 1.9013289250520622e-05, + "loss": 1.087, + "step": 4699 + }, + { + "epoch": 0.3, + "learning_rate": 1.9013074561497675e-05, + "loss": 1.0986, + "step": 4700 + }, + { + "epoch": 0.3, + "learning_rate": 1.9012859872474724e-05, + "loss": 1.0976, + "step": 4701 + }, + { + "epoch": 0.3, + "learning_rate": 1.9012645183451773e-05, + "loss": 1.0941, + "step": 4702 + }, + { + "epoch": 0.3, + "learning_rate": 1.901243049442882e-05, + "loss": 1.1141, + "step": 4703 + }, + { + "epoch": 0.3, + "learning_rate": 1.901221580540587e-05, + "loss": 1.0859, + "step": 4704 + }, + { + "epoch": 0.3, + "learning_rate": 1.901200111638292e-05, + "loss": 1.1146, + "step": 4705 + }, + { + "epoch": 0.3, + "learning_rate": 1.9011786427359972e-05, + "loss": 1.0201, + "step": 4706 + }, + { + "epoch": 0.3, + "learning_rate": 1.901157173833702e-05, + "loss": 1.1343, + "step": 4707 + }, + { + "epoch": 0.3, + "learning_rate": 1.901135704931407e-05, + "loss": 1.1757, + "step": 4708 + }, + { + "epoch": 0.3, + "learning_rate": 1.901114236029112e-05, + "loss": 1.1081, + "step": 4709 + }, + { + "epoch": 0.3, + "learning_rate": 1.9010927671268168e-05, + "loss": 1.0746, + "step": 4710 + }, + { + "epoch": 0.3, + "learning_rate": 1.901071298224522e-05, + "loss": 1.167, + "step": 4711 + }, + { + "epoch": 0.3, + "learning_rate": 1.901049829322227e-05, + "loss": 1.1406, + "step": 4712 + }, + { + "epoch": 0.3, + "learning_rate": 1.901028360419932e-05, + "loss": 1.0146, + "step": 4713 + }, + { + "epoch": 0.3, + "learning_rate": 1.901006891517637e-05, + "loss": 1.106, + "step": 4714 + }, + { + "epoch": 0.3, + "learning_rate": 1.900985422615342e-05, + "loss": 1.0967, + "step": 4715 + }, + { + "epoch": 0.3, + "learning_rate": 1.9009639537130468e-05, + "loss": 1.1911, + "step": 4716 + }, + { + "epoch": 0.3, + "learning_rate": 1.9009424848107517e-05, + "loss": 1.1298, + "step": 4717 + }, + { + "epoch": 0.3, + "learning_rate": 1.9009210159084566e-05, + "loss": 1.0823, + "step": 4718 + }, + { + "epoch": 0.3, + "learning_rate": 1.900899547006162e-05, + "loss": 1.138, + "step": 4719 + }, + { + "epoch": 0.3, + "learning_rate": 1.9008780781038668e-05, + "loss": 0.9828, + "step": 4720 + }, + { + "epoch": 0.3, + "learning_rate": 1.9008566092015717e-05, + "loss": 1.0567, + "step": 4721 + }, + { + "epoch": 0.3, + "learning_rate": 1.9008351402992765e-05, + "loss": 1.1556, + "step": 4722 + }, + { + "epoch": 0.3, + "learning_rate": 1.9008136713969814e-05, + "loss": 1.1921, + "step": 4723 + }, + { + "epoch": 0.3, + "learning_rate": 1.9007922024946867e-05, + "loss": 1.0995, + "step": 4724 + }, + { + "epoch": 0.3, + "learning_rate": 1.9007707335923916e-05, + "loss": 1.1106, + "step": 4725 + }, + { + "epoch": 0.3, + "learning_rate": 1.9007492646900965e-05, + "loss": 1.1133, + "step": 4726 + }, + { + "epoch": 0.3, + "learning_rate": 1.9007277957878014e-05, + "loss": 1.0831, + "step": 4727 + }, + { + "epoch": 0.3, + "learning_rate": 1.9007063268855066e-05, + "loss": 1.0299, + "step": 4728 + }, + { + "epoch": 0.3, + "learning_rate": 1.9006848579832115e-05, + "loss": 1.1327, + "step": 4729 + }, + { + "epoch": 0.3, + "learning_rate": 1.9006633890809164e-05, + "loss": 1.1461, + "step": 4730 + }, + { + "epoch": 0.3, + "learning_rate": 1.9006419201786216e-05, + "loss": 1.2172, + "step": 4731 + }, + { + "epoch": 0.3, + "learning_rate": 1.9006204512763265e-05, + "loss": 1.1165, + "step": 4732 + }, + { + "epoch": 0.3, + "learning_rate": 1.9005989823740314e-05, + "loss": 1.1555, + "step": 4733 + }, + { + "epoch": 0.3, + "learning_rate": 1.9005775134717363e-05, + "loss": 1.1195, + "step": 4734 + }, + { + "epoch": 0.3, + "learning_rate": 1.9005560445694412e-05, + "loss": 1.0647, + "step": 4735 + }, + { + "epoch": 0.3, + "learning_rate": 1.900534575667146e-05, + "loss": 1.0678, + "step": 4736 + }, + { + "epoch": 0.3, + "learning_rate": 1.9005131067648514e-05, + "loss": 1.1151, + "step": 4737 + }, + { + "epoch": 0.3, + "learning_rate": 1.9004916378625562e-05, + "loss": 1.1175, + "step": 4738 + }, + { + "epoch": 0.3, + "learning_rate": 1.900470168960261e-05, + "loss": 1.1257, + "step": 4739 + }, + { + "epoch": 0.3, + "learning_rate": 1.900448700057966e-05, + "loss": 1.1304, + "step": 4740 + }, + { + "epoch": 0.31, + "learning_rate": 1.900427231155671e-05, + "loss": 1.1202, + "step": 4741 + }, + { + "epoch": 0.31, + "learning_rate": 1.9004057622533762e-05, + "loss": 1.1062, + "step": 4742 + }, + { + "epoch": 0.31, + "learning_rate": 1.900384293351081e-05, + "loss": 1.0777, + "step": 4743 + }, + { + "epoch": 0.31, + "learning_rate": 1.9003628244487863e-05, + "loss": 1.1512, + "step": 4744 + }, + { + "epoch": 0.31, + "learning_rate": 1.9003413555464912e-05, + "loss": 1.161, + "step": 4745 + }, + { + "epoch": 0.31, + "learning_rate": 1.900319886644196e-05, + "loss": 1.1605, + "step": 4746 + }, + { + "epoch": 0.31, + "learning_rate": 1.900298417741901e-05, + "loss": 1.0723, + "step": 4747 + }, + { + "epoch": 0.31, + "learning_rate": 1.900276948839606e-05, + "loss": 1.0615, + "step": 4748 + }, + { + "epoch": 0.31, + "learning_rate": 1.900255479937311e-05, + "loss": 1.157, + "step": 4749 + }, + { + "epoch": 0.31, + "learning_rate": 1.900234011035016e-05, + "loss": 1.1079, + "step": 4750 + }, + { + "epoch": 0.31, + "learning_rate": 1.900212542132721e-05, + "loss": 1.0822, + "step": 4751 + }, + { + "epoch": 0.31, + "learning_rate": 1.9001910732304258e-05, + "loss": 1.0669, + "step": 4752 + }, + { + "epoch": 0.31, + "learning_rate": 1.9001696043281307e-05, + "loss": 1.0575, + "step": 4753 + }, + { + "epoch": 0.31, + "learning_rate": 1.9001481354258356e-05, + "loss": 1.1113, + "step": 4754 + }, + { + "epoch": 0.31, + "learning_rate": 1.900126666523541e-05, + "loss": 1.1665, + "step": 4755 + }, + { + "epoch": 0.31, + "learning_rate": 1.9001051976212457e-05, + "loss": 1.0624, + "step": 4756 + }, + { + "epoch": 0.31, + "learning_rate": 1.900083728718951e-05, + "loss": 1.1401, + "step": 4757 + }, + { + "epoch": 0.31, + "learning_rate": 1.900062259816656e-05, + "loss": 1.0884, + "step": 4758 + }, + { + "epoch": 0.31, + "learning_rate": 1.9000407909143608e-05, + "loss": 1.1359, + "step": 4759 + }, + { + "epoch": 0.31, + "learning_rate": 1.9000193220120657e-05, + "loss": 1.0689, + "step": 4760 + }, + { + "epoch": 0.31, + "learning_rate": 1.8999978531097706e-05, + "loss": 1.002, + "step": 4761 + }, + { + "epoch": 0.31, + "learning_rate": 1.8999763842074758e-05, + "loss": 1.0923, + "step": 4762 + }, + { + "epoch": 0.31, + "learning_rate": 1.8999549153051807e-05, + "loss": 1.0982, + "step": 4763 + }, + { + "epoch": 0.31, + "learning_rate": 1.8999334464028856e-05, + "loss": 1.0909, + "step": 4764 + }, + { + "epoch": 0.31, + "learning_rate": 1.8999119775005905e-05, + "loss": 1.1402, + "step": 4765 + }, + { + "epoch": 0.31, + "learning_rate": 1.8998905085982954e-05, + "loss": 1.1092, + "step": 4766 + }, + { + "epoch": 0.31, + "learning_rate": 1.8998690396960006e-05, + "loss": 1.1206, + "step": 4767 + }, + { + "epoch": 0.31, + "learning_rate": 1.8998475707937055e-05, + "loss": 1.1417, + "step": 4768 + }, + { + "epoch": 0.31, + "learning_rate": 1.8998261018914104e-05, + "loss": 1.1539, + "step": 4769 + }, + { + "epoch": 0.31, + "learning_rate": 1.8998046329891153e-05, + "loss": 1.1098, + "step": 4770 + }, + { + "epoch": 0.31, + "learning_rate": 1.8997831640868202e-05, + "loss": 1.045, + "step": 4771 + }, + { + "epoch": 0.31, + "learning_rate": 1.8997616951845254e-05, + "loss": 1.1187, + "step": 4772 + }, + { + "epoch": 0.31, + "learning_rate": 1.8997402262822303e-05, + "loss": 1.199, + "step": 4773 + }, + { + "epoch": 0.31, + "learning_rate": 1.8997187573799352e-05, + "loss": 1.097, + "step": 4774 + }, + { + "epoch": 0.31, + "learning_rate": 1.8996972884776405e-05, + "loss": 1.0733, + "step": 4775 + }, + { + "epoch": 0.31, + "learning_rate": 1.8996758195753454e-05, + "loss": 1.1295, + "step": 4776 + }, + { + "epoch": 0.31, + "learning_rate": 1.8996543506730503e-05, + "loss": 1.0342, + "step": 4777 + }, + { + "epoch": 0.31, + "learning_rate": 1.899632881770755e-05, + "loss": 1.2294, + "step": 4778 + }, + { + "epoch": 0.31, + "learning_rate": 1.89961141286846e-05, + "loss": 1.1013, + "step": 4779 + }, + { + "epoch": 0.31, + "learning_rate": 1.8995899439661653e-05, + "loss": 1.146, + "step": 4780 + }, + { + "epoch": 0.31, + "learning_rate": 1.8995684750638702e-05, + "loss": 1.0788, + "step": 4781 + }, + { + "epoch": 0.31, + "learning_rate": 1.899547006161575e-05, + "loss": 1.1258, + "step": 4782 + }, + { + "epoch": 0.31, + "learning_rate": 1.89952553725928e-05, + "loss": 1.1384, + "step": 4783 + }, + { + "epoch": 0.31, + "learning_rate": 1.899504068356985e-05, + "loss": 1.1746, + "step": 4784 + }, + { + "epoch": 0.31, + "learning_rate": 1.8994825994546898e-05, + "loss": 1.1231, + "step": 4785 + }, + { + "epoch": 0.31, + "learning_rate": 1.899461130552395e-05, + "loss": 1.078, + "step": 4786 + }, + { + "epoch": 0.31, + "learning_rate": 1.8994396616501002e-05, + "loss": 1.0883, + "step": 4787 + }, + { + "epoch": 0.31, + "learning_rate": 1.899418192747805e-05, + "loss": 1.0831, + "step": 4788 + }, + { + "epoch": 0.31, + "learning_rate": 1.89939672384551e-05, + "loss": 1.1029, + "step": 4789 + }, + { + "epoch": 0.31, + "learning_rate": 1.899375254943215e-05, + "loss": 1.1691, + "step": 4790 + }, + { + "epoch": 0.31, + "learning_rate": 1.8993537860409198e-05, + "loss": 1.1349, + "step": 4791 + }, + { + "epoch": 0.31, + "learning_rate": 1.8993323171386247e-05, + "loss": 1.1032, + "step": 4792 + }, + { + "epoch": 0.31, + "learning_rate": 1.89931084823633e-05, + "loss": 1.0747, + "step": 4793 + }, + { + "epoch": 0.31, + "learning_rate": 1.899289379334035e-05, + "loss": 1.1275, + "step": 4794 + }, + { + "epoch": 0.31, + "learning_rate": 1.8992679104317398e-05, + "loss": 1.0945, + "step": 4795 + }, + { + "epoch": 0.31, + "learning_rate": 1.8992464415294446e-05, + "loss": 1.1385, + "step": 4796 + }, + { + "epoch": 0.31, + "learning_rate": 1.8992249726271495e-05, + "loss": 1.101, + "step": 4797 + }, + { + "epoch": 0.31, + "learning_rate": 1.8992035037248548e-05, + "loss": 1.136, + "step": 4798 + }, + { + "epoch": 0.31, + "learning_rate": 1.8991820348225597e-05, + "loss": 1.1635, + "step": 4799 + }, + { + "epoch": 0.31, + "learning_rate": 1.8991605659202646e-05, + "loss": 1.164, + "step": 4800 + }, + { + "epoch": 0.31, + "learning_rate": 1.8991390970179698e-05, + "loss": 1.236, + "step": 4801 + }, + { + "epoch": 0.31, + "learning_rate": 1.8991176281156747e-05, + "loss": 1.0353, + "step": 4802 + }, + { + "epoch": 0.31, + "learning_rate": 1.8990961592133796e-05, + "loss": 1.1463, + "step": 4803 + }, + { + "epoch": 0.31, + "learning_rate": 1.8990746903110845e-05, + "loss": 1.037, + "step": 4804 + }, + { + "epoch": 0.31, + "learning_rate": 1.8990532214087894e-05, + "loss": 1.1422, + "step": 4805 + }, + { + "epoch": 0.31, + "learning_rate": 1.8990317525064946e-05, + "loss": 1.1269, + "step": 4806 + }, + { + "epoch": 0.31, + "learning_rate": 1.8990102836041995e-05, + "loss": 1.1137, + "step": 4807 + }, + { + "epoch": 0.31, + "learning_rate": 1.8989888147019044e-05, + "loss": 1.1223, + "step": 4808 + }, + { + "epoch": 0.31, + "learning_rate": 1.8989673457996093e-05, + "loss": 1.1406, + "step": 4809 + }, + { + "epoch": 0.31, + "learning_rate": 1.8989458768973142e-05, + "loss": 1.103, + "step": 4810 + }, + { + "epoch": 0.31, + "learning_rate": 1.8989244079950195e-05, + "loss": 1.0905, + "step": 4811 + }, + { + "epoch": 0.31, + "learning_rate": 1.8989029390927243e-05, + "loss": 1.234, + "step": 4812 + }, + { + "epoch": 0.31, + "learning_rate": 1.8988814701904292e-05, + "loss": 1.0943, + "step": 4813 + }, + { + "epoch": 0.31, + "learning_rate": 1.898860001288134e-05, + "loss": 1.0991, + "step": 4814 + }, + { + "epoch": 0.31, + "learning_rate": 1.8988385323858394e-05, + "loss": 1.191, + "step": 4815 + }, + { + "epoch": 0.31, + "learning_rate": 1.8988170634835443e-05, + "loss": 1.1988, + "step": 4816 + }, + { + "epoch": 0.31, + "learning_rate": 1.898795594581249e-05, + "loss": 1.118, + "step": 4817 + }, + { + "epoch": 0.31, + "learning_rate": 1.8987741256789544e-05, + "loss": 1.0483, + "step": 4818 + }, + { + "epoch": 0.31, + "learning_rate": 1.8987526567766593e-05, + "loss": 1.098, + "step": 4819 + }, + { + "epoch": 0.31, + "learning_rate": 1.8987311878743642e-05, + "loss": 1.0819, + "step": 4820 + }, + { + "epoch": 0.31, + "learning_rate": 1.898709718972069e-05, + "loss": 1.0878, + "step": 4821 + }, + { + "epoch": 0.31, + "learning_rate": 1.898688250069774e-05, + "loss": 1.1297, + "step": 4822 + }, + { + "epoch": 0.31, + "learning_rate": 1.898666781167479e-05, + "loss": 1.0433, + "step": 4823 + }, + { + "epoch": 0.31, + "learning_rate": 1.898645312265184e-05, + "loss": 1.1709, + "step": 4824 + }, + { + "epoch": 0.31, + "learning_rate": 1.898623843362889e-05, + "loss": 1.0927, + "step": 4825 + }, + { + "epoch": 0.31, + "learning_rate": 1.898602374460594e-05, + "loss": 1.0917, + "step": 4826 + }, + { + "epoch": 0.31, + "learning_rate": 1.8985809055582988e-05, + "loss": 1.0883, + "step": 4827 + }, + { + "epoch": 0.31, + "learning_rate": 1.8985594366560037e-05, + "loss": 1.0944, + "step": 4828 + }, + { + "epoch": 0.31, + "learning_rate": 1.898537967753709e-05, + "loss": 1.2169, + "step": 4829 + }, + { + "epoch": 0.31, + "learning_rate": 1.898516498851414e-05, + "loss": 1.1713, + "step": 4830 + }, + { + "epoch": 0.31, + "learning_rate": 1.898495029949119e-05, + "loss": 1.1505, + "step": 4831 + }, + { + "epoch": 0.31, + "learning_rate": 1.898473561046824e-05, + "loss": 1.0542, + "step": 4832 + }, + { + "epoch": 0.31, + "learning_rate": 1.898452092144529e-05, + "loss": 1.109, + "step": 4833 + }, + { + "epoch": 0.31, + "learning_rate": 1.8984306232422338e-05, + "loss": 1.1297, + "step": 4834 + }, + { + "epoch": 0.31, + "learning_rate": 1.8984091543399387e-05, + "loss": 1.1122, + "step": 4835 + }, + { + "epoch": 0.31, + "learning_rate": 1.898387685437644e-05, + "loss": 1.1924, + "step": 4836 + }, + { + "epoch": 0.31, + "learning_rate": 1.8983662165353488e-05, + "loss": 1.1191, + "step": 4837 + }, + { + "epoch": 0.31, + "learning_rate": 1.8983447476330537e-05, + "loss": 1.0804, + "step": 4838 + }, + { + "epoch": 0.31, + "learning_rate": 1.8983232787307586e-05, + "loss": 1.1022, + "step": 4839 + }, + { + "epoch": 0.31, + "learning_rate": 1.8983018098284635e-05, + "loss": 1.1246, + "step": 4840 + }, + { + "epoch": 0.31, + "learning_rate": 1.8982803409261684e-05, + "loss": 1.0989, + "step": 4841 + }, + { + "epoch": 0.31, + "learning_rate": 1.8982588720238736e-05, + "loss": 1.0504, + "step": 4842 + }, + { + "epoch": 0.31, + "learning_rate": 1.8982374031215785e-05, + "loss": 1.0819, + "step": 4843 + }, + { + "epoch": 0.31, + "learning_rate": 1.8982159342192837e-05, + "loss": 1.1016, + "step": 4844 + }, + { + "epoch": 0.31, + "learning_rate": 1.8981944653169886e-05, + "loss": 1.2942, + "step": 4845 + }, + { + "epoch": 0.31, + "learning_rate": 1.8981729964146935e-05, + "loss": 1.0831, + "step": 4846 + }, + { + "epoch": 0.31, + "learning_rate": 1.8981515275123984e-05, + "loss": 1.1921, + "step": 4847 + }, + { + "epoch": 0.31, + "learning_rate": 1.8981300586101033e-05, + "loss": 1.2056, + "step": 4848 + }, + { + "epoch": 0.31, + "learning_rate": 1.8981085897078086e-05, + "loss": 1.0694, + "step": 4849 + }, + { + "epoch": 0.31, + "learning_rate": 1.8980871208055135e-05, + "loss": 1.1275, + "step": 4850 + }, + { + "epoch": 0.31, + "learning_rate": 1.8980656519032184e-05, + "loss": 1.1568, + "step": 4851 + }, + { + "epoch": 0.31, + "learning_rate": 1.8980441830009233e-05, + "loss": 1.1119, + "step": 4852 + }, + { + "epoch": 0.31, + "learning_rate": 1.898022714098628e-05, + "loss": 1.052, + "step": 4853 + }, + { + "epoch": 0.31, + "learning_rate": 1.8980012451963334e-05, + "loss": 1.014, + "step": 4854 + }, + { + "epoch": 0.31, + "learning_rate": 1.8979797762940383e-05, + "loss": 1.1356, + "step": 4855 + }, + { + "epoch": 0.31, + "learning_rate": 1.8979583073917432e-05, + "loss": 1.1176, + "step": 4856 + }, + { + "epoch": 0.31, + "learning_rate": 1.897936838489448e-05, + "loss": 1.1399, + "step": 4857 + }, + { + "epoch": 0.31, + "learning_rate": 1.897915369587153e-05, + "loss": 1.1619, + "step": 4858 + }, + { + "epoch": 0.31, + "learning_rate": 1.8978939006848582e-05, + "loss": 1.1639, + "step": 4859 + }, + { + "epoch": 0.31, + "learning_rate": 1.897872431782563e-05, + "loss": 1.0768, + "step": 4860 + }, + { + "epoch": 0.31, + "learning_rate": 1.897850962880268e-05, + "loss": 1.0456, + "step": 4861 + }, + { + "epoch": 0.31, + "learning_rate": 1.8978294939779732e-05, + "loss": 1.1359, + "step": 4862 + }, + { + "epoch": 0.31, + "learning_rate": 1.897808025075678e-05, + "loss": 1.1366, + "step": 4863 + }, + { + "epoch": 0.31, + "learning_rate": 1.897786556173383e-05, + "loss": 1.1256, + "step": 4864 + }, + { + "epoch": 0.31, + "learning_rate": 1.897765087271088e-05, + "loss": 1.1588, + "step": 4865 + }, + { + "epoch": 0.31, + "learning_rate": 1.8977436183687928e-05, + "loss": 1.1147, + "step": 4866 + }, + { + "epoch": 0.31, + "learning_rate": 1.897722149466498e-05, + "loss": 1.0782, + "step": 4867 + }, + { + "epoch": 0.31, + "learning_rate": 1.897700680564203e-05, + "loss": 1.2044, + "step": 4868 + }, + { + "epoch": 0.31, + "learning_rate": 1.897679211661908e-05, + "loss": 1.1442, + "step": 4869 + }, + { + "epoch": 0.31, + "learning_rate": 1.8976577427596127e-05, + "loss": 1.1469, + "step": 4870 + }, + { + "epoch": 0.31, + "learning_rate": 1.8976362738573176e-05, + "loss": 1.1134, + "step": 4871 + }, + { + "epoch": 0.31, + "learning_rate": 1.8976148049550225e-05, + "loss": 1.0758, + "step": 4872 + }, + { + "epoch": 0.31, + "learning_rate": 1.8975933360527278e-05, + "loss": 1.1378, + "step": 4873 + }, + { + "epoch": 0.31, + "learning_rate": 1.897571867150433e-05, + "loss": 1.1546, + "step": 4874 + }, + { + "epoch": 0.31, + "learning_rate": 1.897550398248138e-05, + "loss": 1.0818, + "step": 4875 + }, + { + "epoch": 0.31, + "learning_rate": 1.8975289293458428e-05, + "loss": 1.1562, + "step": 4876 + }, + { + "epoch": 0.31, + "learning_rate": 1.8975074604435477e-05, + "loss": 1.1503, + "step": 4877 + }, + { + "epoch": 0.31, + "learning_rate": 1.8974859915412526e-05, + "loss": 1.1009, + "step": 4878 + }, + { + "epoch": 0.31, + "learning_rate": 1.8974645226389575e-05, + "loss": 1.09, + "step": 4879 + }, + { + "epoch": 0.31, + "learning_rate": 1.8974430537366627e-05, + "loss": 1.1144, + "step": 4880 + }, + { + "epoch": 0.31, + "learning_rate": 1.8974215848343676e-05, + "loss": 1.1225, + "step": 4881 + }, + { + "epoch": 0.31, + "learning_rate": 1.8974001159320725e-05, + "loss": 1.0645, + "step": 4882 + }, + { + "epoch": 0.31, + "learning_rate": 1.8973786470297774e-05, + "loss": 1.1448, + "step": 4883 + }, + { + "epoch": 0.31, + "learning_rate": 1.8973571781274823e-05, + "loss": 1.107, + "step": 4884 + }, + { + "epoch": 0.31, + "learning_rate": 1.8973357092251876e-05, + "loss": 1.141, + "step": 4885 + }, + { + "epoch": 0.31, + "learning_rate": 1.8973142403228924e-05, + "loss": 1.0653, + "step": 4886 + }, + { + "epoch": 0.31, + "learning_rate": 1.8972927714205973e-05, + "loss": 1.1048, + "step": 4887 + }, + { + "epoch": 0.31, + "learning_rate": 1.8972713025183026e-05, + "loss": 1.1037, + "step": 4888 + }, + { + "epoch": 0.31, + "learning_rate": 1.8972498336160075e-05, + "loss": 1.1952, + "step": 4889 + }, + { + "epoch": 0.31, + "learning_rate": 1.8972283647137124e-05, + "loss": 1.1719, + "step": 4890 + }, + { + "epoch": 0.31, + "learning_rate": 1.8972068958114173e-05, + "loss": 1.059, + "step": 4891 + }, + { + "epoch": 0.31, + "learning_rate": 1.897185426909122e-05, + "loss": 1.1321, + "step": 4892 + }, + { + "epoch": 0.31, + "learning_rate": 1.8971639580068274e-05, + "loss": 1.0468, + "step": 4893 + }, + { + "epoch": 0.31, + "learning_rate": 1.8971424891045323e-05, + "loss": 1.1754, + "step": 4894 + }, + { + "epoch": 0.31, + "learning_rate": 1.8971210202022372e-05, + "loss": 1.1275, + "step": 4895 + }, + { + "epoch": 0.31, + "learning_rate": 1.897099551299942e-05, + "loss": 1.0416, + "step": 4896 + }, + { + "epoch": 0.32, + "learning_rate": 1.897078082397647e-05, + "loss": 1.1243, + "step": 4897 + }, + { + "epoch": 0.32, + "learning_rate": 1.8970566134953522e-05, + "loss": 1.1618, + "step": 4898 + }, + { + "epoch": 0.32, + "learning_rate": 1.897035144593057e-05, + "loss": 1.2119, + "step": 4899 + }, + { + "epoch": 0.32, + "learning_rate": 1.897013675690762e-05, + "loss": 1.0907, + "step": 4900 + }, + { + "epoch": 0.32, + "learning_rate": 1.896992206788467e-05, + "loss": 1.1561, + "step": 4901 + }, + { + "epoch": 0.32, + "learning_rate": 1.896970737886172e-05, + "loss": 1.1704, + "step": 4902 + }, + { + "epoch": 0.32, + "learning_rate": 1.896949268983877e-05, + "loss": 1.1065, + "step": 4903 + }, + { + "epoch": 0.32, + "learning_rate": 1.896927800081582e-05, + "loss": 1.0498, + "step": 4904 + }, + { + "epoch": 0.32, + "learning_rate": 1.8969063311792872e-05, + "loss": 1.07, + "step": 4905 + }, + { + "epoch": 0.32, + "learning_rate": 1.896884862276992e-05, + "loss": 1.1372, + "step": 4906 + }, + { + "epoch": 0.32, + "learning_rate": 1.896863393374697e-05, + "loss": 1.0443, + "step": 4907 + }, + { + "epoch": 0.32, + "learning_rate": 1.896841924472402e-05, + "loss": 1.0624, + "step": 4908 + }, + { + "epoch": 0.32, + "learning_rate": 1.8968204555701068e-05, + "loss": 1.1241, + "step": 4909 + }, + { + "epoch": 0.32, + "learning_rate": 1.8967989866678117e-05, + "loss": 1.0276, + "step": 4910 + }, + { + "epoch": 0.32, + "learning_rate": 1.896777517765517e-05, + "loss": 1.0622, + "step": 4911 + }, + { + "epoch": 0.32, + "learning_rate": 1.8967560488632218e-05, + "loss": 1.097, + "step": 4912 + }, + { + "epoch": 0.32, + "learning_rate": 1.8967345799609267e-05, + "loss": 1.0892, + "step": 4913 + }, + { + "epoch": 0.32, + "learning_rate": 1.8967131110586316e-05, + "loss": 1.0705, + "step": 4914 + }, + { + "epoch": 0.32, + "learning_rate": 1.8966916421563365e-05, + "loss": 1.1949, + "step": 4915 + }, + { + "epoch": 0.32, + "learning_rate": 1.8966701732540417e-05, + "loss": 1.1228, + "step": 4916 + }, + { + "epoch": 0.32, + "learning_rate": 1.8966487043517466e-05, + "loss": 1.2217, + "step": 4917 + }, + { + "epoch": 0.32, + "learning_rate": 1.896627235449452e-05, + "loss": 1.0941, + "step": 4918 + }, + { + "epoch": 0.32, + "learning_rate": 1.8966057665471567e-05, + "loss": 1.1571, + "step": 4919 + }, + { + "epoch": 0.32, + "learning_rate": 1.8965842976448616e-05, + "loss": 1.0769, + "step": 4920 + }, + { + "epoch": 0.32, + "learning_rate": 1.8965628287425665e-05, + "loss": 1.0793, + "step": 4921 + }, + { + "epoch": 0.32, + "learning_rate": 1.8965413598402714e-05, + "loss": 1.1189, + "step": 4922 + }, + { + "epoch": 0.32, + "learning_rate": 1.8965198909379767e-05, + "loss": 1.1948, + "step": 4923 + }, + { + "epoch": 0.32, + "learning_rate": 1.8964984220356816e-05, + "loss": 1.0965, + "step": 4924 + }, + { + "epoch": 0.32, + "learning_rate": 1.8964769531333865e-05, + "loss": 1.1066, + "step": 4925 + }, + { + "epoch": 0.32, + "learning_rate": 1.8964554842310914e-05, + "loss": 1.1236, + "step": 4926 + }, + { + "epoch": 0.32, + "learning_rate": 1.8964340153287963e-05, + "loss": 1.0951, + "step": 4927 + }, + { + "epoch": 0.32, + "learning_rate": 1.896412546426501e-05, + "loss": 1.1252, + "step": 4928 + }, + { + "epoch": 0.32, + "learning_rate": 1.8963910775242064e-05, + "loss": 1.1315, + "step": 4929 + }, + { + "epoch": 0.32, + "learning_rate": 1.8963696086219113e-05, + "loss": 1.0882, + "step": 4930 + }, + { + "epoch": 0.32, + "learning_rate": 1.8963481397196165e-05, + "loss": 1.1228, + "step": 4931 + }, + { + "epoch": 0.32, + "learning_rate": 1.8963266708173214e-05, + "loss": 1.1708, + "step": 4932 + }, + { + "epoch": 0.32, + "learning_rate": 1.8963052019150263e-05, + "loss": 1.2178, + "step": 4933 + }, + { + "epoch": 0.32, + "learning_rate": 1.8962837330127312e-05, + "loss": 1.0576, + "step": 4934 + }, + { + "epoch": 0.32, + "learning_rate": 1.896262264110436e-05, + "loss": 1.0671, + "step": 4935 + }, + { + "epoch": 0.32, + "learning_rate": 1.8962407952081413e-05, + "loss": 1.1776, + "step": 4936 + }, + { + "epoch": 0.32, + "learning_rate": 1.8962193263058462e-05, + "loss": 1.0681, + "step": 4937 + }, + { + "epoch": 0.32, + "learning_rate": 1.896197857403551e-05, + "loss": 1.0938, + "step": 4938 + }, + { + "epoch": 0.32, + "learning_rate": 1.896176388501256e-05, + "loss": 1.1238, + "step": 4939 + }, + { + "epoch": 0.32, + "learning_rate": 1.896154919598961e-05, + "loss": 1.1624, + "step": 4940 + }, + { + "epoch": 0.32, + "learning_rate": 1.8961334506966658e-05, + "loss": 1.054, + "step": 4941 + }, + { + "epoch": 0.32, + "learning_rate": 1.896111981794371e-05, + "loss": 1.0767, + "step": 4942 + }, + { + "epoch": 0.32, + "learning_rate": 1.896090512892076e-05, + "loss": 1.0919, + "step": 4943 + }, + { + "epoch": 0.32, + "learning_rate": 1.896069043989781e-05, + "loss": 1.1167, + "step": 4944 + }, + { + "epoch": 0.32, + "learning_rate": 1.8960475750874857e-05, + "loss": 1.0935, + "step": 4945 + }, + { + "epoch": 0.32, + "learning_rate": 1.896026106185191e-05, + "loss": 1.1357, + "step": 4946 + }, + { + "epoch": 0.32, + "learning_rate": 1.896004637282896e-05, + "loss": 1.1374, + "step": 4947 + }, + { + "epoch": 0.32, + "learning_rate": 1.8959831683806008e-05, + "loss": 1.0625, + "step": 4948 + }, + { + "epoch": 0.32, + "learning_rate": 1.895961699478306e-05, + "loss": 1.1008, + "step": 4949 + }, + { + "epoch": 0.32, + "learning_rate": 1.895940230576011e-05, + "loss": 1.2023, + "step": 4950 + }, + { + "epoch": 0.32, + "learning_rate": 1.8959187616737158e-05, + "loss": 1.1161, + "step": 4951 + }, + { + "epoch": 0.32, + "learning_rate": 1.8958972927714207e-05, + "loss": 1.0941, + "step": 4952 + }, + { + "epoch": 0.32, + "learning_rate": 1.8958758238691256e-05, + "loss": 1.1837, + "step": 4953 + }, + { + "epoch": 0.32, + "learning_rate": 1.8958543549668308e-05, + "loss": 1.1207, + "step": 4954 + }, + { + "epoch": 0.32, + "learning_rate": 1.8958328860645357e-05, + "loss": 1.0094, + "step": 4955 + }, + { + "epoch": 0.32, + "learning_rate": 1.8958114171622406e-05, + "loss": 1.1068, + "step": 4956 + }, + { + "epoch": 0.32, + "learning_rate": 1.8957899482599455e-05, + "loss": 1.1388, + "step": 4957 + }, + { + "epoch": 0.32, + "learning_rate": 1.8957684793576504e-05, + "loss": 1.22, + "step": 4958 + }, + { + "epoch": 0.32, + "learning_rate": 1.8957470104553553e-05, + "loss": 1.1127, + "step": 4959 + }, + { + "epoch": 0.32, + "learning_rate": 1.8957255415530605e-05, + "loss": 1.072, + "step": 4960 + }, + { + "epoch": 0.32, + "learning_rate": 1.8957040726507658e-05, + "loss": 1.0466, + "step": 4961 + }, + { + "epoch": 0.32, + "learning_rate": 1.8956826037484707e-05, + "loss": 1.0006, + "step": 4962 + }, + { + "epoch": 0.32, + "learning_rate": 1.8956611348461756e-05, + "loss": 1.0881, + "step": 4963 + }, + { + "epoch": 0.32, + "learning_rate": 1.8956396659438805e-05, + "loss": 1.1231, + "step": 4964 + }, + { + "epoch": 0.32, + "learning_rate": 1.8956181970415854e-05, + "loss": 1.1848, + "step": 4965 + }, + { + "epoch": 0.32, + "learning_rate": 1.8955967281392903e-05, + "loss": 1.0964, + "step": 4966 + }, + { + "epoch": 0.32, + "learning_rate": 1.8955752592369955e-05, + "loss": 1.1459, + "step": 4967 + }, + { + "epoch": 0.32, + "learning_rate": 1.8955537903347004e-05, + "loss": 1.1416, + "step": 4968 + }, + { + "epoch": 0.32, + "learning_rate": 1.8955323214324053e-05, + "loss": 1.1829, + "step": 4969 + }, + { + "epoch": 0.32, + "learning_rate": 1.8955108525301102e-05, + "loss": 1.0934, + "step": 4970 + }, + { + "epoch": 0.32, + "learning_rate": 1.895489383627815e-05, + "loss": 1.2176, + "step": 4971 + }, + { + "epoch": 0.32, + "learning_rate": 1.8954679147255203e-05, + "loss": 1.0658, + "step": 4972 + }, + { + "epoch": 0.32, + "learning_rate": 1.8954464458232252e-05, + "loss": 1.0897, + "step": 4973 + }, + { + "epoch": 0.32, + "learning_rate": 1.89542497692093e-05, + "loss": 1.0756, + "step": 4974 + }, + { + "epoch": 0.32, + "learning_rate": 1.8954035080186354e-05, + "loss": 1.0755, + "step": 4975 + }, + { + "epoch": 0.32, + "learning_rate": 1.8953820391163402e-05, + "loss": 1.1307, + "step": 4976 + }, + { + "epoch": 0.32, + "learning_rate": 1.895360570214045e-05, + "loss": 1.0541, + "step": 4977 + }, + { + "epoch": 0.32, + "learning_rate": 1.89533910131175e-05, + "loss": 1.165, + "step": 4978 + }, + { + "epoch": 0.32, + "learning_rate": 1.895317632409455e-05, + "loss": 1.112, + "step": 4979 + }, + { + "epoch": 0.32, + "learning_rate": 1.8952961635071602e-05, + "loss": 1.0244, + "step": 4980 + }, + { + "epoch": 0.32, + "learning_rate": 1.895274694604865e-05, + "loss": 1.0149, + "step": 4981 + }, + { + "epoch": 0.32, + "learning_rate": 1.89525322570257e-05, + "loss": 1.1532, + "step": 4982 + }, + { + "epoch": 0.32, + "learning_rate": 1.895231756800275e-05, + "loss": 1.1508, + "step": 4983 + }, + { + "epoch": 0.32, + "learning_rate": 1.8952102878979798e-05, + "loss": 1.0236, + "step": 4984 + }, + { + "epoch": 0.32, + "learning_rate": 1.895188818995685e-05, + "loss": 1.1376, + "step": 4985 + }, + { + "epoch": 0.32, + "learning_rate": 1.89516735009339e-05, + "loss": 1.1083, + "step": 4986 + }, + { + "epoch": 0.32, + "learning_rate": 1.8951458811910948e-05, + "loss": 1.0428, + "step": 4987 + }, + { + "epoch": 0.32, + "learning_rate": 1.8951244122887997e-05, + "loss": 1.1336, + "step": 4988 + }, + { + "epoch": 0.32, + "learning_rate": 1.895102943386505e-05, + "loss": 1.0989, + "step": 4989 + }, + { + "epoch": 0.32, + "learning_rate": 1.8950814744842098e-05, + "loss": 1.1546, + "step": 4990 + }, + { + "epoch": 0.32, + "learning_rate": 1.8950600055819147e-05, + "loss": 1.152, + "step": 4991 + }, + { + "epoch": 0.32, + "learning_rate": 1.89503853667962e-05, + "loss": 1.094, + "step": 4992 + }, + { + "epoch": 0.32, + "learning_rate": 1.895017067777325e-05, + "loss": 1.201, + "step": 4993 + }, + { + "epoch": 0.32, + "learning_rate": 1.8949955988750297e-05, + "loss": 1.1816, + "step": 4994 + }, + { + "epoch": 0.32, + "learning_rate": 1.8949741299727346e-05, + "loss": 1.1301, + "step": 4995 + }, + { + "epoch": 0.32, + "learning_rate": 1.8949526610704395e-05, + "loss": 1.087, + "step": 4996 + }, + { + "epoch": 0.32, + "learning_rate": 1.8949311921681444e-05, + "loss": 1.2146, + "step": 4997 + }, + { + "epoch": 0.32, + "learning_rate": 1.8949097232658497e-05, + "loss": 1.1171, + "step": 4998 + }, + { + "epoch": 0.32, + "learning_rate": 1.8948882543635546e-05, + "loss": 1.071, + "step": 4999 + }, + { + "epoch": 0.32, + "learning_rate": 1.8948667854612595e-05, + "loss": 1.1631, + "step": 5000 + }, + { + "epoch": 0.32, + "eval_laion/cider": 0.5111077502372141, + "eval_laion/meteor": 0.3847717148944281, + "eval_laion/pred_num_words": 35.205078125, + "eval_laion/sacrebleu": 15.038419408038733, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.1262047290802002, + "eval_runtime": 95.3978, + "eval_samples_per_second": 5.231, + "eval_steps_per_second": 0.335, + "step": 5000 + }, + { + "epoch": 0.32, + "learning_rate": 1.8948453165589644e-05, + "loss": 1.1497, + "step": 5001 + }, + { + "epoch": 0.32, + "learning_rate": 1.8948238476566693e-05, + "loss": 1.1339, + "step": 5002 + }, + { + "epoch": 0.32, + "learning_rate": 1.8948023787543745e-05, + "loss": 1.1158, + "step": 5003 + }, + { + "epoch": 0.32, + "learning_rate": 1.8947809098520794e-05, + "loss": 1.147, + "step": 5004 + }, + { + "epoch": 0.32, + "learning_rate": 1.8947594409497846e-05, + "loss": 1.1859, + "step": 5005 + }, + { + "epoch": 0.32, + "learning_rate": 1.8947379720474895e-05, + "loss": 1.1754, + "step": 5006 + }, + { + "epoch": 0.32, + "learning_rate": 1.8947165031451944e-05, + "loss": 1.0537, + "step": 5007 + }, + { + "epoch": 0.32, + "learning_rate": 1.8946950342428993e-05, + "loss": 1.0818, + "step": 5008 + }, + { + "epoch": 0.32, + "learning_rate": 1.8946735653406042e-05, + "loss": 1.1102, + "step": 5009 + }, + { + "epoch": 0.32, + "learning_rate": 1.8946520964383094e-05, + "loss": 1.1435, + "step": 5010 + }, + { + "epoch": 0.32, + "learning_rate": 1.8946306275360143e-05, + "loss": 1.0807, + "step": 5011 + }, + { + "epoch": 0.32, + "learning_rate": 1.8946091586337192e-05, + "loss": 1.0789, + "step": 5012 + }, + { + "epoch": 0.32, + "learning_rate": 1.894587689731424e-05, + "loss": 1.1515, + "step": 5013 + }, + { + "epoch": 0.32, + "learning_rate": 1.894566220829129e-05, + "loss": 1.0007, + "step": 5014 + }, + { + "epoch": 0.32, + "learning_rate": 1.894544751926834e-05, + "loss": 1.1696, + "step": 5015 + }, + { + "epoch": 0.32, + "learning_rate": 1.894523283024539e-05, + "loss": 1.2087, + "step": 5016 + }, + { + "epoch": 0.32, + "learning_rate": 1.894501814122244e-05, + "loss": 1.0914, + "step": 5017 + }, + { + "epoch": 0.32, + "learning_rate": 1.8944803452199493e-05, + "loss": 1.1343, + "step": 5018 + }, + { + "epoch": 0.32, + "learning_rate": 1.8944588763176542e-05, + "loss": 1.0818, + "step": 5019 + }, + { + "epoch": 0.32, + "learning_rate": 1.894437407415359e-05, + "loss": 1.0963, + "step": 5020 + }, + { + "epoch": 0.32, + "learning_rate": 1.894415938513064e-05, + "loss": 1.164, + "step": 5021 + }, + { + "epoch": 0.32, + "learning_rate": 1.894394469610769e-05, + "loss": 1.1507, + "step": 5022 + }, + { + "epoch": 0.32, + "learning_rate": 1.894373000708474e-05, + "loss": 1.09, + "step": 5023 + }, + { + "epoch": 0.32, + "learning_rate": 1.894351531806179e-05, + "loss": 1.1109, + "step": 5024 + }, + { + "epoch": 0.32, + "learning_rate": 1.894330062903884e-05, + "loss": 1.1727, + "step": 5025 + }, + { + "epoch": 0.32, + "learning_rate": 1.8943085940015888e-05, + "loss": 1.1269, + "step": 5026 + }, + { + "epoch": 0.32, + "learning_rate": 1.8942871250992937e-05, + "loss": 1.0889, + "step": 5027 + }, + { + "epoch": 0.32, + "learning_rate": 1.8942656561969986e-05, + "loss": 1.1463, + "step": 5028 + }, + { + "epoch": 0.32, + "learning_rate": 1.8942441872947038e-05, + "loss": 1.1378, + "step": 5029 + }, + { + "epoch": 0.32, + "learning_rate": 1.8942227183924087e-05, + "loss": 1.0871, + "step": 5030 + }, + { + "epoch": 0.32, + "learning_rate": 1.8942012494901136e-05, + "loss": 1.1522, + "step": 5031 + }, + { + "epoch": 0.32, + "learning_rate": 1.8941797805878185e-05, + "loss": 1.1008, + "step": 5032 + }, + { + "epoch": 0.32, + "learning_rate": 1.8941583116855238e-05, + "loss": 1.1613, + "step": 5033 + }, + { + "epoch": 0.32, + "learning_rate": 1.8941368427832286e-05, + "loss": 1.1791, + "step": 5034 + }, + { + "epoch": 0.32, + "learning_rate": 1.8941153738809335e-05, + "loss": 1.1234, + "step": 5035 + }, + { + "epoch": 0.32, + "learning_rate": 1.8940939049786388e-05, + "loss": 1.1741, + "step": 5036 + }, + { + "epoch": 0.32, + "learning_rate": 1.8940724360763437e-05, + "loss": 1.1236, + "step": 5037 + }, + { + "epoch": 0.32, + "learning_rate": 1.8940509671740486e-05, + "loss": 1.1024, + "step": 5038 + }, + { + "epoch": 0.32, + "learning_rate": 1.8940294982717535e-05, + "loss": 1.1235, + "step": 5039 + }, + { + "epoch": 0.32, + "learning_rate": 1.8940080293694584e-05, + "loss": 1.1004, + "step": 5040 + }, + { + "epoch": 0.32, + "learning_rate": 1.8939865604671636e-05, + "loss": 1.0454, + "step": 5041 + }, + { + "epoch": 0.32, + "learning_rate": 1.8939650915648685e-05, + "loss": 1.1291, + "step": 5042 + }, + { + "epoch": 0.32, + "learning_rate": 1.8939436226625734e-05, + "loss": 1.1841, + "step": 5043 + }, + { + "epoch": 0.32, + "learning_rate": 1.8939221537602783e-05, + "loss": 1.1811, + "step": 5044 + }, + { + "epoch": 0.32, + "learning_rate": 1.8939006848579832e-05, + "loss": 1.1127, + "step": 5045 + }, + { + "epoch": 0.32, + "learning_rate": 1.893879215955688e-05, + "loss": 1.2592, + "step": 5046 + }, + { + "epoch": 0.32, + "learning_rate": 1.8938577470533933e-05, + "loss": 1.1319, + "step": 5047 + }, + { + "epoch": 0.32, + "learning_rate": 1.8938362781510982e-05, + "loss": 1.0763, + "step": 5048 + }, + { + "epoch": 0.32, + "learning_rate": 1.8938148092488035e-05, + "loss": 1.1502, + "step": 5049 + }, + { + "epoch": 0.32, + "learning_rate": 1.8937933403465083e-05, + "loss": 1.1187, + "step": 5050 + }, + { + "epoch": 0.32, + "learning_rate": 1.8937718714442132e-05, + "loss": 1.1006, + "step": 5051 + }, + { + "epoch": 0.33, + "learning_rate": 1.893750402541918e-05, + "loss": 1.1406, + "step": 5052 + }, + { + "epoch": 0.33, + "learning_rate": 1.893728933639623e-05, + "loss": 1.0772, + "step": 5053 + }, + { + "epoch": 0.33, + "learning_rate": 1.8937074647373283e-05, + "loss": 1.1126, + "step": 5054 + }, + { + "epoch": 0.33, + "learning_rate": 1.893685995835033e-05, + "loss": 1.1231, + "step": 5055 + }, + { + "epoch": 0.33, + "learning_rate": 1.893664526932738e-05, + "loss": 1.0998, + "step": 5056 + }, + { + "epoch": 0.33, + "learning_rate": 1.893643058030443e-05, + "loss": 0.9815, + "step": 5057 + }, + { + "epoch": 0.33, + "learning_rate": 1.893621589128148e-05, + "loss": 1.0924, + "step": 5058 + }, + { + "epoch": 0.33, + "learning_rate": 1.893600120225853e-05, + "loss": 1.1374, + "step": 5059 + }, + { + "epoch": 0.33, + "learning_rate": 1.893578651323558e-05, + "loss": 1.1058, + "step": 5060 + }, + { + "epoch": 0.33, + "learning_rate": 1.893557182421263e-05, + "loss": 1.1879, + "step": 5061 + }, + { + "epoch": 0.33, + "learning_rate": 1.893535713518968e-05, + "loss": 1.0604, + "step": 5062 + }, + { + "epoch": 0.33, + "learning_rate": 1.893514244616673e-05, + "loss": 1.0735, + "step": 5063 + }, + { + "epoch": 0.33, + "learning_rate": 1.893492775714378e-05, + "loss": 1.0917, + "step": 5064 + }, + { + "epoch": 0.33, + "learning_rate": 1.8934713068120828e-05, + "loss": 1.1807, + "step": 5065 + }, + { + "epoch": 0.33, + "learning_rate": 1.8934498379097877e-05, + "loss": 1.0459, + "step": 5066 + }, + { + "epoch": 0.33, + "learning_rate": 1.893428369007493e-05, + "loss": 1.0991, + "step": 5067 + }, + { + "epoch": 0.33, + "learning_rate": 1.893406900105198e-05, + "loss": 1.1105, + "step": 5068 + }, + { + "epoch": 0.33, + "learning_rate": 1.8933854312029027e-05, + "loss": 1.16, + "step": 5069 + }, + { + "epoch": 0.33, + "learning_rate": 1.8933639623006076e-05, + "loss": 1.1207, + "step": 5070 + }, + { + "epoch": 0.33, + "learning_rate": 1.8933424933983125e-05, + "loss": 1.1121, + "step": 5071 + }, + { + "epoch": 0.33, + "learning_rate": 1.8933210244960178e-05, + "loss": 1.1072, + "step": 5072 + }, + { + "epoch": 0.33, + "learning_rate": 1.8932995555937227e-05, + "loss": 1.0811, + "step": 5073 + }, + { + "epoch": 0.33, + "learning_rate": 1.8932780866914276e-05, + "loss": 1.1927, + "step": 5074 + }, + { + "epoch": 0.33, + "learning_rate": 1.8932566177891325e-05, + "loss": 1.1449, + "step": 5075 + }, + { + "epoch": 0.33, + "learning_rate": 1.8932351488868377e-05, + "loss": 1.0569, + "step": 5076 + }, + { + "epoch": 0.33, + "learning_rate": 1.8932136799845426e-05, + "loss": 1.1243, + "step": 5077 + }, + { + "epoch": 0.33, + "learning_rate": 1.8931922110822475e-05, + "loss": 1.0685, + "step": 5078 + }, + { + "epoch": 0.33, + "learning_rate": 1.8931707421799527e-05, + "loss": 1.055, + "step": 5079 + }, + { + "epoch": 0.33, + "learning_rate": 1.8931492732776576e-05, + "loss": 1.1063, + "step": 5080 + }, + { + "epoch": 0.33, + "learning_rate": 1.8931278043753625e-05, + "loss": 1.1358, + "step": 5081 + }, + { + "epoch": 0.33, + "learning_rate": 1.8931063354730674e-05, + "loss": 1.158, + "step": 5082 + }, + { + "epoch": 0.33, + "learning_rate": 1.8930848665707723e-05, + "loss": 1.1619, + "step": 5083 + }, + { + "epoch": 0.33, + "learning_rate": 1.8930633976684772e-05, + "loss": 1.0734, + "step": 5084 + }, + { + "epoch": 0.33, + "learning_rate": 1.8930419287661824e-05, + "loss": 1.0876, + "step": 5085 + }, + { + "epoch": 0.33, + "learning_rate": 1.8930204598638873e-05, + "loss": 0.9988, + "step": 5086 + }, + { + "epoch": 0.33, + "learning_rate": 1.8929989909615922e-05, + "loss": 1.0846, + "step": 5087 + }, + { + "epoch": 0.33, + "learning_rate": 1.892977522059297e-05, + "loss": 1.0978, + "step": 5088 + }, + { + "epoch": 0.33, + "learning_rate": 1.892956053157002e-05, + "loss": 1.0771, + "step": 5089 + }, + { + "epoch": 0.33, + "learning_rate": 1.8929345842547073e-05, + "loss": 1.0591, + "step": 5090 + }, + { + "epoch": 0.33, + "learning_rate": 1.892913115352412e-05, + "loss": 1.1012, + "step": 5091 + }, + { + "epoch": 0.33, + "learning_rate": 1.8928916464501174e-05, + "loss": 1.1095, + "step": 5092 + }, + { + "epoch": 0.33, + "learning_rate": 1.8928701775478223e-05, + "loss": 1.1412, + "step": 5093 + }, + { + "epoch": 0.33, + "learning_rate": 1.8928487086455272e-05, + "loss": 1.1156, + "step": 5094 + }, + { + "epoch": 0.33, + "learning_rate": 1.892827239743232e-05, + "loss": 1.2026, + "step": 5095 + }, + { + "epoch": 0.33, + "learning_rate": 1.892805770840937e-05, + "loss": 1.1404, + "step": 5096 + }, + { + "epoch": 0.33, + "learning_rate": 1.8927843019386422e-05, + "loss": 1.0815, + "step": 5097 + }, + { + "epoch": 0.33, + "learning_rate": 1.892762833036347e-05, + "loss": 1.0992, + "step": 5098 + }, + { + "epoch": 0.33, + "learning_rate": 1.892741364134052e-05, + "loss": 1.0428, + "step": 5099 + }, + { + "epoch": 0.33, + "learning_rate": 1.892719895231757e-05, + "loss": 1.0976, + "step": 5100 + }, + { + "epoch": 0.33, + "learning_rate": 1.8926984263294618e-05, + "loss": 1.125, + "step": 5101 + }, + { + "epoch": 0.33, + "learning_rate": 1.8926769574271667e-05, + "loss": 1.1128, + "step": 5102 + }, + { + "epoch": 0.33, + "learning_rate": 1.892655488524872e-05, + "loss": 1.0904, + "step": 5103 + }, + { + "epoch": 0.33, + "learning_rate": 1.8926340196225768e-05, + "loss": 1.0866, + "step": 5104 + }, + { + "epoch": 0.33, + "learning_rate": 1.8926125507202817e-05, + "loss": 1.214, + "step": 5105 + }, + { + "epoch": 0.33, + "learning_rate": 1.892591081817987e-05, + "loss": 1.2106, + "step": 5106 + }, + { + "epoch": 0.33, + "learning_rate": 1.892569612915692e-05, + "loss": 1.0826, + "step": 5107 + }, + { + "epoch": 0.33, + "learning_rate": 1.8925481440133967e-05, + "loss": 1.1875, + "step": 5108 + }, + { + "epoch": 0.33, + "learning_rate": 1.8925266751111016e-05, + "loss": 1.0702, + "step": 5109 + }, + { + "epoch": 0.33, + "learning_rate": 1.892505206208807e-05, + "loss": 1.1435, + "step": 5110 + }, + { + "epoch": 0.33, + "learning_rate": 1.8924837373065118e-05, + "loss": 1.1471, + "step": 5111 + }, + { + "epoch": 0.33, + "learning_rate": 1.8924622684042167e-05, + "loss": 1.1115, + "step": 5112 + }, + { + "epoch": 0.33, + "learning_rate": 1.8924407995019216e-05, + "loss": 1.1205, + "step": 5113 + }, + { + "epoch": 0.33, + "learning_rate": 1.8924193305996265e-05, + "loss": 1.1247, + "step": 5114 + }, + { + "epoch": 0.33, + "learning_rate": 1.8923978616973314e-05, + "loss": 1.1307, + "step": 5115 + }, + { + "epoch": 0.33, + "learning_rate": 1.8923763927950366e-05, + "loss": 1.1209, + "step": 5116 + }, + { + "epoch": 0.33, + "learning_rate": 1.8923549238927415e-05, + "loss": 1.142, + "step": 5117 + }, + { + "epoch": 0.33, + "learning_rate": 1.8923334549904464e-05, + "loss": 1.0676, + "step": 5118 + }, + { + "epoch": 0.33, + "learning_rate": 1.8923119860881513e-05, + "loss": 1.0876, + "step": 5119 + }, + { + "epoch": 0.33, + "learning_rate": 1.8922905171858565e-05, + "loss": 1.1515, + "step": 5120 + }, + { + "epoch": 0.33, + "learning_rate": 1.8922690482835614e-05, + "loss": 1.0921, + "step": 5121 + }, + { + "epoch": 0.33, + "learning_rate": 1.8922475793812663e-05, + "loss": 1.0747, + "step": 5122 + }, + { + "epoch": 0.33, + "learning_rate": 1.8922261104789716e-05, + "loss": 1.0855, + "step": 5123 + }, + { + "epoch": 0.33, + "learning_rate": 1.8922046415766764e-05, + "loss": 1.06, + "step": 5124 + }, + { + "epoch": 0.33, + "learning_rate": 1.8921831726743813e-05, + "loss": 1.0829, + "step": 5125 + }, + { + "epoch": 0.33, + "learning_rate": 1.8921617037720862e-05, + "loss": 1.0904, + "step": 5126 + }, + { + "epoch": 0.33, + "learning_rate": 1.892140234869791e-05, + "loss": 1.1253, + "step": 5127 + }, + { + "epoch": 0.33, + "learning_rate": 1.8921187659674964e-05, + "loss": 1.0274, + "step": 5128 + }, + { + "epoch": 0.33, + "learning_rate": 1.8920972970652013e-05, + "loss": 1.1224, + "step": 5129 + }, + { + "epoch": 0.33, + "learning_rate": 1.892075828162906e-05, + "loss": 1.1377, + "step": 5130 + }, + { + "epoch": 0.33, + "learning_rate": 1.892054359260611e-05, + "loss": 1.1387, + "step": 5131 + }, + { + "epoch": 0.33, + "learning_rate": 1.892032890358316e-05, + "loss": 1.0395, + "step": 5132 + }, + { + "epoch": 0.33, + "learning_rate": 1.892011421456021e-05, + "loss": 1.1503, + "step": 5133 + }, + { + "epoch": 0.33, + "learning_rate": 1.891989952553726e-05, + "loss": 1.0938, + "step": 5134 + }, + { + "epoch": 0.33, + "learning_rate": 1.891968483651431e-05, + "loss": 1.0792, + "step": 5135 + }, + { + "epoch": 0.33, + "learning_rate": 1.8919470147491362e-05, + "loss": 1.0673, + "step": 5136 + }, + { + "epoch": 0.33, + "learning_rate": 1.891925545846841e-05, + "loss": 1.0632, + "step": 5137 + }, + { + "epoch": 0.33, + "learning_rate": 1.891904076944546e-05, + "loss": 1.112, + "step": 5138 + }, + { + "epoch": 0.33, + "learning_rate": 1.891882608042251e-05, + "loss": 1.1443, + "step": 5139 + }, + { + "epoch": 0.33, + "learning_rate": 1.8918611391399558e-05, + "loss": 1.1236, + "step": 5140 + }, + { + "epoch": 0.33, + "learning_rate": 1.891839670237661e-05, + "loss": 1.0847, + "step": 5141 + }, + { + "epoch": 0.33, + "learning_rate": 1.891818201335366e-05, + "loss": 1.0982, + "step": 5142 + }, + { + "epoch": 0.33, + "learning_rate": 1.891796732433071e-05, + "loss": 1.0761, + "step": 5143 + }, + { + "epoch": 0.33, + "learning_rate": 1.8917752635307757e-05, + "loss": 1.1682, + "step": 5144 + }, + { + "epoch": 0.33, + "learning_rate": 1.8917537946284806e-05, + "loss": 1.015, + "step": 5145 + }, + { + "epoch": 0.33, + "learning_rate": 1.891732325726186e-05, + "loss": 1.1435, + "step": 5146 + }, + { + "epoch": 0.33, + "learning_rate": 1.8917108568238908e-05, + "loss": 1.0633, + "step": 5147 + }, + { + "epoch": 0.33, + "learning_rate": 1.8916893879215957e-05, + "loss": 1.206, + "step": 5148 + }, + { + "epoch": 0.33, + "learning_rate": 1.891667919019301e-05, + "loss": 1.0828, + "step": 5149 + }, + { + "epoch": 0.33, + "learning_rate": 1.8916464501170058e-05, + "loss": 1.1103, + "step": 5150 + }, + { + "epoch": 0.33, + "learning_rate": 1.8916249812147107e-05, + "loss": 1.0804, + "step": 5151 + }, + { + "epoch": 0.33, + "learning_rate": 1.8916035123124156e-05, + "loss": 1.1641, + "step": 5152 + }, + { + "epoch": 0.33, + "learning_rate": 1.8915820434101205e-05, + "loss": 1.0857, + "step": 5153 + }, + { + "epoch": 0.33, + "learning_rate": 1.8915605745078257e-05, + "loss": 1.0209, + "step": 5154 + }, + { + "epoch": 0.33, + "learning_rate": 1.8915391056055306e-05, + "loss": 1.1116, + "step": 5155 + }, + { + "epoch": 0.33, + "learning_rate": 1.8915176367032355e-05, + "loss": 1.0768, + "step": 5156 + }, + { + "epoch": 0.33, + "learning_rate": 1.8914961678009404e-05, + "loss": 1.1338, + "step": 5157 + }, + { + "epoch": 0.33, + "learning_rate": 1.8914746988986453e-05, + "loss": 1.0888, + "step": 5158 + }, + { + "epoch": 0.33, + "learning_rate": 1.8914532299963505e-05, + "loss": 1.0128, + "step": 5159 + }, + { + "epoch": 0.33, + "learning_rate": 1.8914317610940554e-05, + "loss": 1.1177, + "step": 5160 + }, + { + "epoch": 0.33, + "learning_rate": 1.8914102921917603e-05, + "loss": 1.0861, + "step": 5161 + }, + { + "epoch": 0.33, + "learning_rate": 1.8913888232894652e-05, + "loss": 1.1147, + "step": 5162 + }, + { + "epoch": 0.33, + "learning_rate": 1.89136735438717e-05, + "loss": 1.0296, + "step": 5163 + }, + { + "epoch": 0.33, + "learning_rate": 1.8913458854848754e-05, + "loss": 1.0342, + "step": 5164 + }, + { + "epoch": 0.33, + "learning_rate": 1.8913244165825803e-05, + "loss": 1.0956, + "step": 5165 + }, + { + "epoch": 0.33, + "learning_rate": 1.8913029476802855e-05, + "loss": 1.2352, + "step": 5166 + }, + { + "epoch": 0.33, + "learning_rate": 1.8912814787779904e-05, + "loss": 1.14, + "step": 5167 + }, + { + "epoch": 0.33, + "learning_rate": 1.8912600098756953e-05, + "loss": 1.0927, + "step": 5168 + }, + { + "epoch": 0.33, + "learning_rate": 1.8912385409734002e-05, + "loss": 1.1035, + "step": 5169 + }, + { + "epoch": 0.33, + "learning_rate": 1.891217072071105e-05, + "loss": 1.2284, + "step": 5170 + }, + { + "epoch": 0.33, + "learning_rate": 1.89119560316881e-05, + "loss": 1.1084, + "step": 5171 + }, + { + "epoch": 0.33, + "learning_rate": 1.8911741342665152e-05, + "loss": 1.1222, + "step": 5172 + }, + { + "epoch": 0.33, + "learning_rate": 1.89115266536422e-05, + "loss": 1.0826, + "step": 5173 + }, + { + "epoch": 0.33, + "learning_rate": 1.891131196461925e-05, + "loss": 1.1856, + "step": 5174 + }, + { + "epoch": 0.33, + "learning_rate": 1.89110972755963e-05, + "loss": 1.0333, + "step": 5175 + }, + { + "epoch": 0.33, + "learning_rate": 1.8910882586573348e-05, + "loss": 1.1943, + "step": 5176 + }, + { + "epoch": 0.33, + "learning_rate": 1.89106678975504e-05, + "loss": 1.1185, + "step": 5177 + }, + { + "epoch": 0.33, + "learning_rate": 1.891045320852745e-05, + "loss": 1.1723, + "step": 5178 + }, + { + "epoch": 0.33, + "learning_rate": 1.89102385195045e-05, + "loss": 1.1026, + "step": 5179 + }, + { + "epoch": 0.33, + "learning_rate": 1.891002383048155e-05, + "loss": 1.0796, + "step": 5180 + }, + { + "epoch": 0.33, + "learning_rate": 1.89098091414586e-05, + "loss": 1.1184, + "step": 5181 + }, + { + "epoch": 0.33, + "learning_rate": 1.890959445243565e-05, + "loss": 1.0996, + "step": 5182 + }, + { + "epoch": 0.33, + "learning_rate": 1.8909379763412697e-05, + "loss": 1.1733, + "step": 5183 + }, + { + "epoch": 0.33, + "learning_rate": 1.890916507438975e-05, + "loss": 1.0393, + "step": 5184 + }, + { + "epoch": 0.33, + "learning_rate": 1.89089503853668e-05, + "loss": 1.1034, + "step": 5185 + }, + { + "epoch": 0.33, + "learning_rate": 1.8908735696343848e-05, + "loss": 1.0823, + "step": 5186 + }, + { + "epoch": 0.33, + "learning_rate": 1.8908521007320897e-05, + "loss": 1.2296, + "step": 5187 + }, + { + "epoch": 0.33, + "learning_rate": 1.8908306318297946e-05, + "loss": 1.1358, + "step": 5188 + }, + { + "epoch": 0.33, + "learning_rate": 1.8908091629274995e-05, + "loss": 1.1559, + "step": 5189 + }, + { + "epoch": 0.33, + "learning_rate": 1.8907876940252047e-05, + "loss": 1.0803, + "step": 5190 + }, + { + "epoch": 0.33, + "learning_rate": 1.8907662251229096e-05, + "loss": 1.0992, + "step": 5191 + }, + { + "epoch": 0.33, + "learning_rate": 1.8907447562206145e-05, + "loss": 1.179, + "step": 5192 + }, + { + "epoch": 0.33, + "learning_rate": 1.8907232873183197e-05, + "loss": 1.0957, + "step": 5193 + }, + { + "epoch": 0.33, + "learning_rate": 1.8907018184160246e-05, + "loss": 1.1445, + "step": 5194 + }, + { + "epoch": 0.33, + "learning_rate": 1.8906803495137295e-05, + "loss": 1.0991, + "step": 5195 + }, + { + "epoch": 0.33, + "learning_rate": 1.8906588806114344e-05, + "loss": 1.1692, + "step": 5196 + }, + { + "epoch": 0.33, + "learning_rate": 1.8906374117091397e-05, + "loss": 1.1569, + "step": 5197 + }, + { + "epoch": 0.33, + "learning_rate": 1.8906159428068445e-05, + "loss": 1.1431, + "step": 5198 + }, + { + "epoch": 0.33, + "learning_rate": 1.8905944739045494e-05, + "loss": 1.1635, + "step": 5199 + }, + { + "epoch": 0.33, + "learning_rate": 1.8905730050022543e-05, + "loss": 1.1894, + "step": 5200 + }, + { + "epoch": 0.33, + "learning_rate": 1.8905515360999592e-05, + "loss": 1.1564, + "step": 5201 + }, + { + "epoch": 0.33, + "learning_rate": 1.890530067197664e-05, + "loss": 1.1068, + "step": 5202 + }, + { + "epoch": 0.33, + "learning_rate": 1.8905085982953694e-05, + "loss": 1.1569, + "step": 5203 + }, + { + "epoch": 0.33, + "learning_rate": 1.8904871293930743e-05, + "loss": 1.0639, + "step": 5204 + }, + { + "epoch": 0.33, + "learning_rate": 1.890465660490779e-05, + "loss": 1.1308, + "step": 5205 + }, + { + "epoch": 0.33, + "learning_rate": 1.890444191588484e-05, + "loss": 1.0887, + "step": 5206 + }, + { + "epoch": 0.34, + "learning_rate": 1.8904227226861893e-05, + "loss": 1.0809, + "step": 5207 + }, + { + "epoch": 0.34, + "learning_rate": 1.8904012537838942e-05, + "loss": 1.186, + "step": 5208 + }, + { + "epoch": 0.34, + "learning_rate": 1.890379784881599e-05, + "loss": 1.1903, + "step": 5209 + }, + { + "epoch": 0.34, + "learning_rate": 1.8903583159793043e-05, + "loss": 1.1136, + "step": 5210 + }, + { + "epoch": 0.34, + "learning_rate": 1.8903368470770092e-05, + "loss": 1.0807, + "step": 5211 + }, + { + "epoch": 0.34, + "learning_rate": 1.890315378174714e-05, + "loss": 1.0896, + "step": 5212 + }, + { + "epoch": 0.34, + "learning_rate": 1.890293909272419e-05, + "loss": 1.0902, + "step": 5213 + }, + { + "epoch": 0.34, + "learning_rate": 1.890272440370124e-05, + "loss": 1.0528, + "step": 5214 + }, + { + "epoch": 0.34, + "learning_rate": 1.890250971467829e-05, + "loss": 1.1238, + "step": 5215 + }, + { + "epoch": 0.34, + "learning_rate": 1.890229502565534e-05, + "loss": 1.1338, + "step": 5216 + }, + { + "epoch": 0.34, + "learning_rate": 1.890208033663239e-05, + "loss": 1.1266, + "step": 5217 + }, + { + "epoch": 0.34, + "learning_rate": 1.890186564760944e-05, + "loss": 1.1407, + "step": 5218 + }, + { + "epoch": 0.34, + "learning_rate": 1.8901650958586487e-05, + "loss": 1.0695, + "step": 5219 + }, + { + "epoch": 0.34, + "learning_rate": 1.8901436269563536e-05, + "loss": 1.1196, + "step": 5220 + }, + { + "epoch": 0.34, + "learning_rate": 1.890122158054059e-05, + "loss": 1.1435, + "step": 5221 + }, + { + "epoch": 0.34, + "learning_rate": 1.8901006891517638e-05, + "loss": 1.2099, + "step": 5222 + }, + { + "epoch": 0.34, + "learning_rate": 1.890079220249469e-05, + "loss": 1.1532, + "step": 5223 + }, + { + "epoch": 0.34, + "learning_rate": 1.890057751347174e-05, + "loss": 1.0724, + "step": 5224 + }, + { + "epoch": 0.34, + "learning_rate": 1.8900362824448788e-05, + "loss": 1.1328, + "step": 5225 + }, + { + "epoch": 0.34, + "learning_rate": 1.8900148135425837e-05, + "loss": 1.1096, + "step": 5226 + }, + { + "epoch": 0.34, + "learning_rate": 1.8899933446402886e-05, + "loss": 1.2071, + "step": 5227 + }, + { + "epoch": 0.34, + "learning_rate": 1.8899718757379938e-05, + "loss": 1.1017, + "step": 5228 + }, + { + "epoch": 0.34, + "learning_rate": 1.8899504068356987e-05, + "loss": 1.1327, + "step": 5229 + }, + { + "epoch": 0.34, + "learning_rate": 1.8899289379334036e-05, + "loss": 1.0276, + "step": 5230 + }, + { + "epoch": 0.34, + "learning_rate": 1.8899074690311085e-05, + "loss": 1.19, + "step": 5231 + }, + { + "epoch": 0.34, + "learning_rate": 1.8898860001288134e-05, + "loss": 1.139, + "step": 5232 + }, + { + "epoch": 0.34, + "learning_rate": 1.8898645312265186e-05, + "loss": 1.0903, + "step": 5233 + }, + { + "epoch": 0.34, + "learning_rate": 1.8898430623242235e-05, + "loss": 1.0277, + "step": 5234 + }, + { + "epoch": 0.34, + "learning_rate": 1.8898215934219284e-05, + "loss": 1.2088, + "step": 5235 + }, + { + "epoch": 0.34, + "learning_rate": 1.8898001245196337e-05, + "loss": 1.0486, + "step": 5236 + }, + { + "epoch": 0.34, + "learning_rate": 1.8897786556173386e-05, + "loss": 1.1125, + "step": 5237 + }, + { + "epoch": 0.34, + "learning_rate": 1.8897571867150435e-05, + "loss": 1.1215, + "step": 5238 + }, + { + "epoch": 0.34, + "learning_rate": 1.8897357178127484e-05, + "loss": 1.1524, + "step": 5239 + }, + { + "epoch": 0.34, + "learning_rate": 1.8897142489104533e-05, + "loss": 1.0622, + "step": 5240 + }, + { + "epoch": 0.34, + "learning_rate": 1.8896927800081585e-05, + "loss": 1.1168, + "step": 5241 + }, + { + "epoch": 0.34, + "learning_rate": 1.8896713111058634e-05, + "loss": 1.0668, + "step": 5242 + }, + { + "epoch": 0.34, + "learning_rate": 1.8896498422035683e-05, + "loss": 1.1185, + "step": 5243 + }, + { + "epoch": 0.34, + "learning_rate": 1.8896283733012732e-05, + "loss": 1.2358, + "step": 5244 + }, + { + "epoch": 0.34, + "learning_rate": 1.889606904398978e-05, + "loss": 1.1188, + "step": 5245 + }, + { + "epoch": 0.34, + "learning_rate": 1.8895854354966833e-05, + "loss": 1.1453, + "step": 5246 + }, + { + "epoch": 0.34, + "learning_rate": 1.8895639665943882e-05, + "loss": 1.0887, + "step": 5247 + }, + { + "epoch": 0.34, + "learning_rate": 1.889542497692093e-05, + "loss": 1.1458, + "step": 5248 + }, + { + "epoch": 0.34, + "learning_rate": 1.889521028789798e-05, + "loss": 1.0803, + "step": 5249 + }, + { + "epoch": 0.34, + "learning_rate": 1.889499559887503e-05, + "loss": 1.1063, + "step": 5250 + }, + { + "epoch": 0.34, + "learning_rate": 1.889478090985208e-05, + "loss": 1.1065, + "step": 5251 + }, + { + "epoch": 0.34, + "learning_rate": 1.889456622082913e-05, + "loss": 1.1081, + "step": 5252 + }, + { + "epoch": 0.34, + "learning_rate": 1.8894351531806183e-05, + "loss": 1.0376, + "step": 5253 + }, + { + "epoch": 0.34, + "learning_rate": 1.889413684278323e-05, + "loss": 1.1704, + "step": 5254 + }, + { + "epoch": 0.34, + "learning_rate": 1.889392215376028e-05, + "loss": 1.1411, + "step": 5255 + }, + { + "epoch": 0.34, + "learning_rate": 1.889370746473733e-05, + "loss": 1.1694, + "step": 5256 + }, + { + "epoch": 0.34, + "learning_rate": 1.889349277571438e-05, + "loss": 1.1289, + "step": 5257 + }, + { + "epoch": 0.34, + "learning_rate": 1.8893278086691427e-05, + "loss": 1.1635, + "step": 5258 + }, + { + "epoch": 0.34, + "learning_rate": 1.889306339766848e-05, + "loss": 1.141, + "step": 5259 + }, + { + "epoch": 0.34, + "learning_rate": 1.889284870864553e-05, + "loss": 1.0906, + "step": 5260 + }, + { + "epoch": 0.34, + "learning_rate": 1.8892634019622578e-05, + "loss": 1.1025, + "step": 5261 + }, + { + "epoch": 0.34, + "learning_rate": 1.8892419330599627e-05, + "loss": 1.1784, + "step": 5262 + }, + { + "epoch": 0.34, + "learning_rate": 1.8892204641576676e-05, + "loss": 1.166, + "step": 5263 + }, + { + "epoch": 0.34, + "learning_rate": 1.8891989952553728e-05, + "loss": 1.0909, + "step": 5264 + }, + { + "epoch": 0.34, + "learning_rate": 1.8891775263530777e-05, + "loss": 1.0569, + "step": 5265 + }, + { + "epoch": 0.34, + "learning_rate": 1.889156057450783e-05, + "loss": 1.1012, + "step": 5266 + }, + { + "epoch": 0.34, + "learning_rate": 1.8891345885484878e-05, + "loss": 1.1498, + "step": 5267 + }, + { + "epoch": 0.34, + "learning_rate": 1.8891131196461927e-05, + "loss": 1.0612, + "step": 5268 + }, + { + "epoch": 0.34, + "learning_rate": 1.8890916507438976e-05, + "loss": 1.0749, + "step": 5269 + }, + { + "epoch": 0.34, + "learning_rate": 1.8890701818416025e-05, + "loss": 1.0979, + "step": 5270 + }, + { + "epoch": 0.34, + "learning_rate": 1.8890487129393078e-05, + "loss": 0.9892, + "step": 5271 + }, + { + "epoch": 0.34, + "learning_rate": 1.8890272440370126e-05, + "loss": 1.063, + "step": 5272 + }, + { + "epoch": 0.34, + "learning_rate": 1.8890057751347175e-05, + "loss": 1.0584, + "step": 5273 + }, + { + "epoch": 0.34, + "learning_rate": 1.8889843062324224e-05, + "loss": 1.0606, + "step": 5274 + }, + { + "epoch": 0.34, + "learning_rate": 1.8889628373301273e-05, + "loss": 1.0865, + "step": 5275 + }, + { + "epoch": 0.34, + "learning_rate": 1.8889413684278322e-05, + "loss": 1.1492, + "step": 5276 + }, + { + "epoch": 0.34, + "learning_rate": 1.8889198995255375e-05, + "loss": 1.0473, + "step": 5277 + }, + { + "epoch": 0.34, + "learning_rate": 1.8888984306232424e-05, + "loss": 1.0828, + "step": 5278 + }, + { + "epoch": 0.34, + "learning_rate": 1.8888769617209473e-05, + "loss": 1.1823, + "step": 5279 + }, + { + "epoch": 0.34, + "learning_rate": 1.8888554928186525e-05, + "loss": 1.0962, + "step": 5280 + }, + { + "epoch": 0.34, + "learning_rate": 1.8888340239163574e-05, + "loss": 1.1273, + "step": 5281 + }, + { + "epoch": 0.34, + "learning_rate": 1.8888125550140623e-05, + "loss": 1.0992, + "step": 5282 + }, + { + "epoch": 0.34, + "learning_rate": 1.8887910861117672e-05, + "loss": 1.0809, + "step": 5283 + }, + { + "epoch": 0.34, + "learning_rate": 1.8887696172094724e-05, + "loss": 1.1213, + "step": 5284 + }, + { + "epoch": 0.34, + "learning_rate": 1.8887481483071773e-05, + "loss": 1.2016, + "step": 5285 + }, + { + "epoch": 0.34, + "learning_rate": 1.8887266794048822e-05, + "loss": 1.1172, + "step": 5286 + }, + { + "epoch": 0.34, + "learning_rate": 1.888705210502587e-05, + "loss": 1.099, + "step": 5287 + }, + { + "epoch": 0.34, + "learning_rate": 1.888683741600292e-05, + "loss": 1.1552, + "step": 5288 + }, + { + "epoch": 0.34, + "learning_rate": 1.888662272697997e-05, + "loss": 1.1238, + "step": 5289 + }, + { + "epoch": 0.34, + "learning_rate": 1.888640803795702e-05, + "loss": 1.1611, + "step": 5290 + }, + { + "epoch": 0.34, + "learning_rate": 1.888619334893407e-05, + "loss": 1.1133, + "step": 5291 + }, + { + "epoch": 0.34, + "learning_rate": 1.888597865991112e-05, + "loss": 1.0762, + "step": 5292 + }, + { + "epoch": 0.34, + "learning_rate": 1.8885763970888168e-05, + "loss": 1.1341, + "step": 5293 + }, + { + "epoch": 0.34, + "learning_rate": 1.888554928186522e-05, + "loss": 1.0993, + "step": 5294 + }, + { + "epoch": 0.34, + "learning_rate": 1.888533459284227e-05, + "loss": 1.1052, + "step": 5295 + }, + { + "epoch": 0.34, + "learning_rate": 1.888511990381932e-05, + "loss": 1.1156, + "step": 5296 + }, + { + "epoch": 0.34, + "learning_rate": 1.888490521479637e-05, + "loss": 1.1003, + "step": 5297 + }, + { + "epoch": 0.34, + "learning_rate": 1.888469052577342e-05, + "loss": 1.1014, + "step": 5298 + }, + { + "epoch": 0.34, + "learning_rate": 1.888447583675047e-05, + "loss": 1.0576, + "step": 5299 + }, + { + "epoch": 0.34, + "learning_rate": 1.8884261147727518e-05, + "loss": 1.1471, + "step": 5300 + }, + { + "epoch": 0.34, + "learning_rate": 1.8884046458704567e-05, + "loss": 1.1205, + "step": 5301 + }, + { + "epoch": 0.34, + "learning_rate": 1.888383176968162e-05, + "loss": 1.1206, + "step": 5302 + }, + { + "epoch": 0.34, + "learning_rate": 1.8883617080658668e-05, + "loss": 1.1473, + "step": 5303 + }, + { + "epoch": 0.34, + "learning_rate": 1.8883402391635717e-05, + "loss": 1.0581, + "step": 5304 + }, + { + "epoch": 0.34, + "learning_rate": 1.8883187702612766e-05, + "loss": 1.1157, + "step": 5305 + }, + { + "epoch": 0.34, + "learning_rate": 1.8882973013589815e-05, + "loss": 1.1281, + "step": 5306 + }, + { + "epoch": 0.34, + "learning_rate": 1.8882758324566864e-05, + "loss": 1.1384, + "step": 5307 + }, + { + "epoch": 0.34, + "learning_rate": 1.8882543635543916e-05, + "loss": 1.1151, + "step": 5308 + }, + { + "epoch": 0.34, + "learning_rate": 1.8882328946520965e-05, + "loss": 1.1571, + "step": 5309 + }, + { + "epoch": 0.34, + "learning_rate": 1.8882114257498018e-05, + "loss": 1.1022, + "step": 5310 + }, + { + "epoch": 0.34, + "learning_rate": 1.8881899568475067e-05, + "loss": 1.0492, + "step": 5311 + }, + { + "epoch": 0.34, + "learning_rate": 1.8881684879452116e-05, + "loss": 1.0991, + "step": 5312 + }, + { + "epoch": 0.34, + "learning_rate": 1.8881470190429165e-05, + "loss": 1.1969, + "step": 5313 + }, + { + "epoch": 0.34, + "learning_rate": 1.8881255501406214e-05, + "loss": 1.0531, + "step": 5314 + }, + { + "epoch": 0.34, + "learning_rate": 1.8881040812383266e-05, + "loss": 1.0957, + "step": 5315 + }, + { + "epoch": 0.34, + "learning_rate": 1.8880826123360315e-05, + "loss": 1.2534, + "step": 5316 + }, + { + "epoch": 0.34, + "learning_rate": 1.8880611434337364e-05, + "loss": 1.138, + "step": 5317 + }, + { + "epoch": 0.34, + "learning_rate": 1.8880396745314413e-05, + "loss": 1.121, + "step": 5318 + }, + { + "epoch": 0.34, + "learning_rate": 1.8880182056291462e-05, + "loss": 1.1217, + "step": 5319 + }, + { + "epoch": 0.34, + "learning_rate": 1.8879967367268514e-05, + "loss": 1.0405, + "step": 5320 + }, + { + "epoch": 0.34, + "learning_rate": 1.8879752678245563e-05, + "loss": 1.1172, + "step": 5321 + }, + { + "epoch": 0.34, + "learning_rate": 1.8879537989222612e-05, + "loss": 1.1695, + "step": 5322 + }, + { + "epoch": 0.34, + "learning_rate": 1.8879323300199664e-05, + "loss": 1.0262, + "step": 5323 + }, + { + "epoch": 0.34, + "learning_rate": 1.8879108611176713e-05, + "loss": 1.0935, + "step": 5324 + }, + { + "epoch": 0.34, + "learning_rate": 1.8878893922153762e-05, + "loss": 1.1096, + "step": 5325 + }, + { + "epoch": 0.34, + "learning_rate": 1.887867923313081e-05, + "loss": 1.1086, + "step": 5326 + }, + { + "epoch": 0.34, + "learning_rate": 1.887846454410786e-05, + "loss": 1.1358, + "step": 5327 + }, + { + "epoch": 0.34, + "learning_rate": 1.8878249855084913e-05, + "loss": 1.0995, + "step": 5328 + }, + { + "epoch": 0.34, + "learning_rate": 1.887803516606196e-05, + "loss": 1.1068, + "step": 5329 + }, + { + "epoch": 0.34, + "learning_rate": 1.887782047703901e-05, + "loss": 1.15, + "step": 5330 + }, + { + "epoch": 0.34, + "learning_rate": 1.887760578801606e-05, + "loss": 1.1533, + "step": 5331 + }, + { + "epoch": 0.34, + "learning_rate": 1.887739109899311e-05, + "loss": 1.1238, + "step": 5332 + }, + { + "epoch": 0.34, + "learning_rate": 1.887717640997016e-05, + "loss": 1.2381, + "step": 5333 + }, + { + "epoch": 0.34, + "learning_rate": 1.887696172094721e-05, + "loss": 1.0698, + "step": 5334 + }, + { + "epoch": 0.34, + "learning_rate": 1.887674703192426e-05, + "loss": 1.0878, + "step": 5335 + }, + { + "epoch": 0.34, + "learning_rate": 1.8876532342901308e-05, + "loss": 1.0884, + "step": 5336 + }, + { + "epoch": 0.34, + "learning_rate": 1.8876317653878357e-05, + "loss": 1.0817, + "step": 5337 + }, + { + "epoch": 0.34, + "learning_rate": 1.887610296485541e-05, + "loss": 1.1129, + "step": 5338 + }, + { + "epoch": 0.34, + "learning_rate": 1.8875888275832458e-05, + "loss": 1.1165, + "step": 5339 + }, + { + "epoch": 0.34, + "learning_rate": 1.887567358680951e-05, + "loss": 1.063, + "step": 5340 + }, + { + "epoch": 0.34, + "learning_rate": 1.887545889778656e-05, + "loss": 1.0286, + "step": 5341 + }, + { + "epoch": 0.34, + "learning_rate": 1.8875244208763608e-05, + "loss": 1.0811, + "step": 5342 + }, + { + "epoch": 0.34, + "learning_rate": 1.8875029519740657e-05, + "loss": 1.1143, + "step": 5343 + }, + { + "epoch": 0.34, + "learning_rate": 1.8874814830717706e-05, + "loss": 1.0883, + "step": 5344 + }, + { + "epoch": 0.34, + "learning_rate": 1.8874600141694755e-05, + "loss": 1.1388, + "step": 5345 + }, + { + "epoch": 0.34, + "learning_rate": 1.8874385452671807e-05, + "loss": 1.1045, + "step": 5346 + }, + { + "epoch": 0.34, + "learning_rate": 1.8874170763648856e-05, + "loss": 1.1026, + "step": 5347 + }, + { + "epoch": 0.34, + "learning_rate": 1.8873956074625905e-05, + "loss": 1.1288, + "step": 5348 + }, + { + "epoch": 0.34, + "learning_rate": 1.8873741385602954e-05, + "loss": 1.0095, + "step": 5349 + }, + { + "epoch": 0.34, + "learning_rate": 1.8873526696580003e-05, + "loss": 1.1185, + "step": 5350 + }, + { + "epoch": 0.34, + "learning_rate": 1.8873312007557056e-05, + "loss": 1.1557, + "step": 5351 + }, + { + "epoch": 0.34, + "learning_rate": 1.8873097318534105e-05, + "loss": 1.1302, + "step": 5352 + }, + { + "epoch": 0.34, + "learning_rate": 1.8872882629511157e-05, + "loss": 1.1683, + "step": 5353 + }, + { + "epoch": 0.34, + "learning_rate": 1.8872667940488206e-05, + "loss": 1.1477, + "step": 5354 + }, + { + "epoch": 0.34, + "learning_rate": 1.8872453251465255e-05, + "loss": 1.1271, + "step": 5355 + }, + { + "epoch": 0.34, + "learning_rate": 1.8872238562442304e-05, + "loss": 1.121, + "step": 5356 + }, + { + "epoch": 0.34, + "learning_rate": 1.8872023873419353e-05, + "loss": 1.1422, + "step": 5357 + }, + { + "epoch": 0.34, + "learning_rate": 1.8871809184396402e-05, + "loss": 1.0732, + "step": 5358 + }, + { + "epoch": 0.34, + "learning_rate": 1.8871594495373454e-05, + "loss": 1.1025, + "step": 5359 + }, + { + "epoch": 0.34, + "learning_rate": 1.8871379806350503e-05, + "loss": 1.0945, + "step": 5360 + }, + { + "epoch": 0.34, + "learning_rate": 1.8871165117327552e-05, + "loss": 1.1661, + "step": 5361 + }, + { + "epoch": 0.34, + "learning_rate": 1.88709504283046e-05, + "loss": 1.099, + "step": 5362 + }, + { + "epoch": 0.35, + "learning_rate": 1.887073573928165e-05, + "loss": 1.1738, + "step": 5363 + }, + { + "epoch": 0.35, + "learning_rate": 1.8870521050258702e-05, + "loss": 1.0965, + "step": 5364 + }, + { + "epoch": 0.35, + "learning_rate": 1.887030636123575e-05, + "loss": 1.0638, + "step": 5365 + }, + { + "epoch": 0.35, + "learning_rate": 1.88700916722128e-05, + "loss": 1.0094, + "step": 5366 + }, + { + "epoch": 0.35, + "learning_rate": 1.8869876983189853e-05, + "loss": 1.1071, + "step": 5367 + }, + { + "epoch": 0.35, + "learning_rate": 1.88696622941669e-05, + "loss": 1.0974, + "step": 5368 + }, + { + "epoch": 0.35, + "learning_rate": 1.886944760514395e-05, + "loss": 1.2825, + "step": 5369 + }, + { + "epoch": 0.35, + "learning_rate": 1.8869232916121e-05, + "loss": 1.0856, + "step": 5370 + }, + { + "epoch": 0.35, + "learning_rate": 1.8869018227098052e-05, + "loss": 1.1789, + "step": 5371 + }, + { + "epoch": 0.35, + "learning_rate": 1.88688035380751e-05, + "loss": 1.1094, + "step": 5372 + }, + { + "epoch": 0.35, + "learning_rate": 1.886858884905215e-05, + "loss": 1.1705, + "step": 5373 + }, + { + "epoch": 0.35, + "learning_rate": 1.88683741600292e-05, + "loss": 1.1048, + "step": 5374 + }, + { + "epoch": 0.35, + "learning_rate": 1.8868159471006248e-05, + "loss": 1.0763, + "step": 5375 + }, + { + "epoch": 0.35, + "learning_rate": 1.8867944781983297e-05, + "loss": 1.071, + "step": 5376 + }, + { + "epoch": 0.35, + "learning_rate": 1.886773009296035e-05, + "loss": 1.0065, + "step": 5377 + }, + { + "epoch": 0.35, + "learning_rate": 1.8867515403937398e-05, + "loss": 1.0542, + "step": 5378 + }, + { + "epoch": 0.35, + "learning_rate": 1.8867300714914447e-05, + "loss": 1.1081, + "step": 5379 + }, + { + "epoch": 0.35, + "learning_rate": 1.8867086025891496e-05, + "loss": 1.1681, + "step": 5380 + }, + { + "epoch": 0.35, + "learning_rate": 1.886687133686855e-05, + "loss": 1.115, + "step": 5381 + }, + { + "epoch": 0.35, + "learning_rate": 1.8866656647845597e-05, + "loss": 1.1327, + "step": 5382 + }, + { + "epoch": 0.35, + "learning_rate": 1.8866441958822646e-05, + "loss": 1.0453, + "step": 5383 + }, + { + "epoch": 0.35, + "learning_rate": 1.88662272697997e-05, + "loss": 1.023, + "step": 5384 + }, + { + "epoch": 0.35, + "learning_rate": 1.8866012580776748e-05, + "loss": 1.0894, + "step": 5385 + }, + { + "epoch": 0.35, + "learning_rate": 1.8865797891753797e-05, + "loss": 1.1321, + "step": 5386 + }, + { + "epoch": 0.35, + "learning_rate": 1.8865583202730846e-05, + "loss": 1.0891, + "step": 5387 + }, + { + "epoch": 0.35, + "learning_rate": 1.8865368513707895e-05, + "loss": 1.065, + "step": 5388 + }, + { + "epoch": 0.35, + "learning_rate": 1.8865153824684947e-05, + "loss": 1.1234, + "step": 5389 + }, + { + "epoch": 0.35, + "learning_rate": 1.8864939135661996e-05, + "loss": 1.1265, + "step": 5390 + }, + { + "epoch": 0.35, + "learning_rate": 1.8864724446639045e-05, + "loss": 1.0612, + "step": 5391 + }, + { + "epoch": 0.35, + "learning_rate": 1.8864509757616094e-05, + "loss": 1.0391, + "step": 5392 + }, + { + "epoch": 0.35, + "learning_rate": 1.8864295068593143e-05, + "loss": 1.0791, + "step": 5393 + }, + { + "epoch": 0.35, + "learning_rate": 1.886408037957019e-05, + "loss": 1.0332, + "step": 5394 + }, + { + "epoch": 0.35, + "learning_rate": 1.8863865690547244e-05, + "loss": 1.1005, + "step": 5395 + }, + { + "epoch": 0.35, + "learning_rate": 1.8863651001524293e-05, + "loss": 1.1239, + "step": 5396 + }, + { + "epoch": 0.35, + "learning_rate": 1.8863436312501345e-05, + "loss": 1.0685, + "step": 5397 + }, + { + "epoch": 0.35, + "learning_rate": 1.8863221623478394e-05, + "loss": 1.1126, + "step": 5398 + }, + { + "epoch": 0.35, + "learning_rate": 1.8863006934455443e-05, + "loss": 1.1862, + "step": 5399 + }, + { + "epoch": 0.35, + "learning_rate": 1.8862792245432492e-05, + "loss": 1.1432, + "step": 5400 + }, + { + "epoch": 0.35, + "learning_rate": 1.886257755640954e-05, + "loss": 1.1319, + "step": 5401 + }, + { + "epoch": 0.35, + "learning_rate": 1.8862362867386594e-05, + "loss": 1.2087, + "step": 5402 + }, + { + "epoch": 0.35, + "learning_rate": 1.8862148178363643e-05, + "loss": 1.1282, + "step": 5403 + }, + { + "epoch": 0.35, + "learning_rate": 1.886193348934069e-05, + "loss": 1.1003, + "step": 5404 + }, + { + "epoch": 0.35, + "learning_rate": 1.886171880031774e-05, + "loss": 1.0939, + "step": 5405 + }, + { + "epoch": 0.35, + "learning_rate": 1.886150411129479e-05, + "loss": 1.104, + "step": 5406 + }, + { + "epoch": 0.35, + "learning_rate": 1.8861289422271842e-05, + "loss": 1.1394, + "step": 5407 + }, + { + "epoch": 0.35, + "learning_rate": 1.886107473324889e-05, + "loss": 1.1657, + "step": 5408 + }, + { + "epoch": 0.35, + "learning_rate": 1.886086004422594e-05, + "loss": 1.142, + "step": 5409 + }, + { + "epoch": 0.35, + "learning_rate": 1.8860645355202992e-05, + "loss": 1.0998, + "step": 5410 + }, + { + "epoch": 0.35, + "learning_rate": 1.886043066618004e-05, + "loss": 1.1893, + "step": 5411 + }, + { + "epoch": 0.35, + "learning_rate": 1.886021597715709e-05, + "loss": 1.0372, + "step": 5412 + }, + { + "epoch": 0.35, + "learning_rate": 1.886000128813414e-05, + "loss": 1.0828, + "step": 5413 + }, + { + "epoch": 0.35, + "learning_rate": 1.8859786599111188e-05, + "loss": 1.0655, + "step": 5414 + }, + { + "epoch": 0.35, + "learning_rate": 1.885957191008824e-05, + "loss": 1.1484, + "step": 5415 + }, + { + "epoch": 0.35, + "learning_rate": 1.885935722106529e-05, + "loss": 1.1022, + "step": 5416 + }, + { + "epoch": 0.35, + "learning_rate": 1.8859142532042338e-05, + "loss": 1.1495, + "step": 5417 + }, + { + "epoch": 0.35, + "learning_rate": 1.8858927843019387e-05, + "loss": 1.0893, + "step": 5418 + }, + { + "epoch": 0.35, + "learning_rate": 1.8858713153996436e-05, + "loss": 1.1563, + "step": 5419 + }, + { + "epoch": 0.35, + "learning_rate": 1.885849846497349e-05, + "loss": 1.1397, + "step": 5420 + }, + { + "epoch": 0.35, + "learning_rate": 1.8858283775950537e-05, + "loss": 1.0862, + "step": 5421 + }, + { + "epoch": 0.35, + "learning_rate": 1.8858069086927586e-05, + "loss": 1.0837, + "step": 5422 + }, + { + "epoch": 0.35, + "learning_rate": 1.8857854397904635e-05, + "loss": 1.1158, + "step": 5423 + }, + { + "epoch": 0.35, + "learning_rate": 1.8857639708881684e-05, + "loss": 1.1217, + "step": 5424 + }, + { + "epoch": 0.35, + "learning_rate": 1.8857425019858737e-05, + "loss": 1.0658, + "step": 5425 + }, + { + "epoch": 0.35, + "learning_rate": 1.8857210330835786e-05, + "loss": 1.0364, + "step": 5426 + }, + { + "epoch": 0.35, + "learning_rate": 1.8856995641812838e-05, + "loss": 1.0423, + "step": 5427 + }, + { + "epoch": 0.35, + "learning_rate": 1.8856780952789887e-05, + "loss": 1.1614, + "step": 5428 + }, + { + "epoch": 0.35, + "learning_rate": 1.8856566263766936e-05, + "loss": 1.0626, + "step": 5429 + }, + { + "epoch": 0.35, + "learning_rate": 1.8856351574743985e-05, + "loss": 1.1517, + "step": 5430 + }, + { + "epoch": 0.35, + "learning_rate": 1.8856136885721034e-05, + "loss": 1.1655, + "step": 5431 + }, + { + "epoch": 0.35, + "learning_rate": 1.8855922196698083e-05, + "loss": 1.1314, + "step": 5432 + }, + { + "epoch": 0.35, + "learning_rate": 1.8855707507675135e-05, + "loss": 1.0578, + "step": 5433 + }, + { + "epoch": 0.35, + "learning_rate": 1.8855492818652184e-05, + "loss": 1.1644, + "step": 5434 + }, + { + "epoch": 0.35, + "learning_rate": 1.8855278129629233e-05, + "loss": 1.1458, + "step": 5435 + }, + { + "epoch": 0.35, + "learning_rate": 1.8855063440606282e-05, + "loss": 1.1723, + "step": 5436 + }, + { + "epoch": 0.35, + "learning_rate": 1.885484875158333e-05, + "loss": 1.2061, + "step": 5437 + }, + { + "epoch": 0.35, + "learning_rate": 1.8854634062560383e-05, + "loss": 1.0692, + "step": 5438 + }, + { + "epoch": 0.35, + "learning_rate": 1.8854419373537432e-05, + "loss": 1.0996, + "step": 5439 + }, + { + "epoch": 0.35, + "learning_rate": 1.8854204684514485e-05, + "loss": 1.0482, + "step": 5440 + }, + { + "epoch": 0.35, + "learning_rate": 1.8853989995491534e-05, + "loss": 1.0982, + "step": 5441 + }, + { + "epoch": 0.35, + "learning_rate": 1.8853775306468583e-05, + "loss": 0.9893, + "step": 5442 + }, + { + "epoch": 0.35, + "learning_rate": 1.885356061744563e-05, + "loss": 1.1441, + "step": 5443 + }, + { + "epoch": 0.35, + "learning_rate": 1.885334592842268e-05, + "loss": 1.1309, + "step": 5444 + }, + { + "epoch": 0.35, + "learning_rate": 1.885313123939973e-05, + "loss": 1.1159, + "step": 5445 + }, + { + "epoch": 0.35, + "learning_rate": 1.8852916550376782e-05, + "loss": 1.1389, + "step": 5446 + }, + { + "epoch": 0.35, + "learning_rate": 1.885270186135383e-05, + "loss": 1.0464, + "step": 5447 + }, + { + "epoch": 0.35, + "learning_rate": 1.885248717233088e-05, + "loss": 1.0922, + "step": 5448 + }, + { + "epoch": 0.35, + "learning_rate": 1.885227248330793e-05, + "loss": 1.0751, + "step": 5449 + }, + { + "epoch": 0.35, + "learning_rate": 1.8852057794284978e-05, + "loss": 1.1391, + "step": 5450 + }, + { + "epoch": 0.35, + "learning_rate": 1.885184310526203e-05, + "loss": 1.069, + "step": 5451 + }, + { + "epoch": 0.35, + "learning_rate": 1.885162841623908e-05, + "loss": 1.0817, + "step": 5452 + }, + { + "epoch": 0.35, + "learning_rate": 1.8851413727216128e-05, + "loss": 1.0683, + "step": 5453 + }, + { + "epoch": 0.35, + "learning_rate": 1.885119903819318e-05, + "loss": 1.1287, + "step": 5454 + }, + { + "epoch": 0.35, + "learning_rate": 1.885098434917023e-05, + "loss": 1.0707, + "step": 5455 + }, + { + "epoch": 0.35, + "learning_rate": 1.885076966014728e-05, + "loss": 1.1814, + "step": 5456 + }, + { + "epoch": 0.35, + "learning_rate": 1.8850554971124327e-05, + "loss": 1.1136, + "step": 5457 + }, + { + "epoch": 0.35, + "learning_rate": 1.885034028210138e-05, + "loss": 1.0778, + "step": 5458 + }, + { + "epoch": 0.35, + "learning_rate": 1.885012559307843e-05, + "loss": 1.2135, + "step": 5459 + }, + { + "epoch": 0.35, + "learning_rate": 1.8849910904055478e-05, + "loss": 1.124, + "step": 5460 + }, + { + "epoch": 0.35, + "learning_rate": 1.8849696215032527e-05, + "loss": 1.0626, + "step": 5461 + }, + { + "epoch": 0.35, + "learning_rate": 1.8849481526009576e-05, + "loss": 1.1863, + "step": 5462 + }, + { + "epoch": 0.35, + "learning_rate": 1.8849266836986624e-05, + "loss": 1.0869, + "step": 5463 + }, + { + "epoch": 0.35, + "learning_rate": 1.8849052147963677e-05, + "loss": 1.0701, + "step": 5464 + }, + { + "epoch": 0.35, + "learning_rate": 1.8848837458940726e-05, + "loss": 1.0728, + "step": 5465 + }, + { + "epoch": 0.35, + "learning_rate": 1.8848622769917775e-05, + "loss": 1.1704, + "step": 5466 + }, + { + "epoch": 0.35, + "learning_rate": 1.8848408080894824e-05, + "loss": 1.1265, + "step": 5467 + }, + { + "epoch": 0.35, + "learning_rate": 1.8848193391871873e-05, + "loss": 1.0715, + "step": 5468 + }, + { + "epoch": 0.35, + "learning_rate": 1.8847978702848925e-05, + "loss": 1.1327, + "step": 5469 + }, + { + "epoch": 0.35, + "learning_rate": 1.8847764013825974e-05, + "loss": 1.1392, + "step": 5470 + }, + { + "epoch": 0.35, + "learning_rate": 1.8847549324803026e-05, + "loss": 1.1032, + "step": 5471 + }, + { + "epoch": 0.35, + "learning_rate": 1.8847334635780075e-05, + "loss": 1.1769, + "step": 5472 + }, + { + "epoch": 0.35, + "learning_rate": 1.8847119946757124e-05, + "loss": 0.9961, + "step": 5473 + }, + { + "epoch": 0.35, + "learning_rate": 1.8846905257734173e-05, + "loss": 1.0731, + "step": 5474 + }, + { + "epoch": 0.35, + "learning_rate": 1.8846690568711222e-05, + "loss": 1.1318, + "step": 5475 + }, + { + "epoch": 0.35, + "learning_rate": 1.8846475879688275e-05, + "loss": 1.1159, + "step": 5476 + }, + { + "epoch": 0.35, + "learning_rate": 1.8846261190665324e-05, + "loss": 1.0334, + "step": 5477 + }, + { + "epoch": 0.35, + "learning_rate": 1.8846046501642373e-05, + "loss": 1.0953, + "step": 5478 + }, + { + "epoch": 0.35, + "learning_rate": 1.884583181261942e-05, + "loss": 1.1293, + "step": 5479 + }, + { + "epoch": 0.35, + "learning_rate": 1.884561712359647e-05, + "loss": 1.0602, + "step": 5480 + }, + { + "epoch": 0.35, + "learning_rate": 1.884540243457352e-05, + "loss": 1.0612, + "step": 5481 + }, + { + "epoch": 0.35, + "learning_rate": 1.8845187745550572e-05, + "loss": 1.0524, + "step": 5482 + }, + { + "epoch": 0.35, + "learning_rate": 1.884497305652762e-05, + "loss": 1.0577, + "step": 5483 + }, + { + "epoch": 0.35, + "learning_rate": 1.8844758367504673e-05, + "loss": 1.0873, + "step": 5484 + }, + { + "epoch": 0.35, + "learning_rate": 1.8844543678481722e-05, + "loss": 1.0555, + "step": 5485 + }, + { + "epoch": 0.35, + "learning_rate": 1.884432898945877e-05, + "loss": 1.0743, + "step": 5486 + }, + { + "epoch": 0.35, + "learning_rate": 1.884411430043582e-05, + "loss": 1.0556, + "step": 5487 + }, + { + "epoch": 0.35, + "learning_rate": 1.884389961141287e-05, + "loss": 1.0862, + "step": 5488 + }, + { + "epoch": 0.35, + "learning_rate": 1.884368492238992e-05, + "loss": 1.0091, + "step": 5489 + }, + { + "epoch": 0.35, + "learning_rate": 1.884347023336697e-05, + "loss": 1.1333, + "step": 5490 + }, + { + "epoch": 0.35, + "learning_rate": 1.884325554434402e-05, + "loss": 1.2062, + "step": 5491 + }, + { + "epoch": 0.35, + "learning_rate": 1.8843040855321068e-05, + "loss": 1.1076, + "step": 5492 + }, + { + "epoch": 0.35, + "learning_rate": 1.8842826166298117e-05, + "loss": 1.1612, + "step": 5493 + }, + { + "epoch": 0.35, + "learning_rate": 1.884261147727517e-05, + "loss": 1.0539, + "step": 5494 + }, + { + "epoch": 0.35, + "learning_rate": 1.884239678825222e-05, + "loss": 1.0011, + "step": 5495 + }, + { + "epoch": 0.35, + "learning_rate": 1.8842182099229267e-05, + "loss": 1.0677, + "step": 5496 + }, + { + "epoch": 0.35, + "learning_rate": 1.8841967410206316e-05, + "loss": 1.1385, + "step": 5497 + }, + { + "epoch": 0.35, + "learning_rate": 1.884175272118337e-05, + "loss": 1.0211, + "step": 5498 + }, + { + "epoch": 0.35, + "learning_rate": 1.8841538032160418e-05, + "loss": 1.0828, + "step": 5499 + }, + { + "epoch": 0.35, + "learning_rate": 1.8841323343137467e-05, + "loss": 1.0909, + "step": 5500 + }, + { + "epoch": 0.35, + "learning_rate": 1.8841108654114516e-05, + "loss": 1.191, + "step": 5501 + }, + { + "epoch": 0.35, + "learning_rate": 1.8840893965091568e-05, + "loss": 1.0779, + "step": 5502 + }, + { + "epoch": 0.35, + "learning_rate": 1.8840679276068617e-05, + "loss": 1.132, + "step": 5503 + }, + { + "epoch": 0.35, + "learning_rate": 1.8840464587045666e-05, + "loss": 1.0611, + "step": 5504 + }, + { + "epoch": 0.35, + "learning_rate": 1.8840249898022715e-05, + "loss": 1.0584, + "step": 5505 + }, + { + "epoch": 0.35, + "learning_rate": 1.8840035208999764e-05, + "loss": 1.0125, + "step": 5506 + }, + { + "epoch": 0.35, + "learning_rate": 1.8839820519976816e-05, + "loss": 1.0623, + "step": 5507 + }, + { + "epoch": 0.35, + "learning_rate": 1.8839605830953865e-05, + "loss": 1.0447, + "step": 5508 + }, + { + "epoch": 0.35, + "learning_rate": 1.8839391141930914e-05, + "loss": 1.2274, + "step": 5509 + }, + { + "epoch": 0.35, + "learning_rate": 1.8839176452907963e-05, + "loss": 1.0356, + "step": 5510 + }, + { + "epoch": 0.35, + "learning_rate": 1.8838961763885012e-05, + "loss": 1.1297, + "step": 5511 + }, + { + "epoch": 0.35, + "learning_rate": 1.8838747074862064e-05, + "loss": 1.1785, + "step": 5512 + }, + { + "epoch": 0.35, + "learning_rate": 1.8838532385839113e-05, + "loss": 1.0542, + "step": 5513 + }, + { + "epoch": 0.35, + "learning_rate": 1.8838317696816166e-05, + "loss": 1.0484, + "step": 5514 + }, + { + "epoch": 0.35, + "learning_rate": 1.8838103007793215e-05, + "loss": 1.1411, + "step": 5515 + }, + { + "epoch": 0.35, + "learning_rate": 1.8837888318770264e-05, + "loss": 1.0583, + "step": 5516 + }, + { + "epoch": 0.35, + "learning_rate": 1.8837673629747313e-05, + "loss": 1.1717, + "step": 5517 + }, + { + "epoch": 0.36, + "learning_rate": 1.883745894072436e-05, + "loss": 1.0745, + "step": 5518 + }, + { + "epoch": 0.36, + "learning_rate": 1.883724425170141e-05, + "loss": 1.1373, + "step": 5519 + }, + { + "epoch": 0.36, + "learning_rate": 1.8837029562678463e-05, + "loss": 1.0665, + "step": 5520 + }, + { + "epoch": 0.36, + "learning_rate": 1.8836814873655512e-05, + "loss": 1.0879, + "step": 5521 + }, + { + "epoch": 0.36, + "learning_rate": 1.883660018463256e-05, + "loss": 1.1382, + "step": 5522 + }, + { + "epoch": 0.36, + "learning_rate": 1.883638549560961e-05, + "loss": 1.0758, + "step": 5523 + }, + { + "epoch": 0.36, + "learning_rate": 1.883617080658666e-05, + "loss": 1.1462, + "step": 5524 + }, + { + "epoch": 0.36, + "learning_rate": 1.883595611756371e-05, + "loss": 1.0513, + "step": 5525 + }, + { + "epoch": 0.36, + "learning_rate": 1.883574142854076e-05, + "loss": 1.2588, + "step": 5526 + }, + { + "epoch": 0.36, + "learning_rate": 1.8835526739517812e-05, + "loss": 1.0911, + "step": 5527 + }, + { + "epoch": 0.36, + "learning_rate": 1.883531205049486e-05, + "loss": 1.1706, + "step": 5528 + }, + { + "epoch": 0.36, + "learning_rate": 1.883509736147191e-05, + "loss": 1.0875, + "step": 5529 + }, + { + "epoch": 0.36, + "learning_rate": 1.883488267244896e-05, + "loss": 1.1202, + "step": 5530 + }, + { + "epoch": 0.36, + "learning_rate": 1.8834667983426008e-05, + "loss": 1.0597, + "step": 5531 + }, + { + "epoch": 0.36, + "learning_rate": 1.8834453294403057e-05, + "loss": 1.1312, + "step": 5532 + }, + { + "epoch": 0.36, + "learning_rate": 1.883423860538011e-05, + "loss": 1.0586, + "step": 5533 + }, + { + "epoch": 0.36, + "learning_rate": 1.883402391635716e-05, + "loss": 1.1084, + "step": 5534 + }, + { + "epoch": 0.36, + "learning_rate": 1.8833809227334208e-05, + "loss": 0.9735, + "step": 5535 + }, + { + "epoch": 0.36, + "learning_rate": 1.8833594538311257e-05, + "loss": 1.1703, + "step": 5536 + }, + { + "epoch": 0.36, + "learning_rate": 1.8833379849288305e-05, + "loss": 1.1991, + "step": 5537 + }, + { + "epoch": 0.36, + "learning_rate": 1.8833165160265358e-05, + "loss": 1.0906, + "step": 5538 + }, + { + "epoch": 0.36, + "learning_rate": 1.8832950471242407e-05, + "loss": 1.0866, + "step": 5539 + }, + { + "epoch": 0.36, + "learning_rate": 1.8832735782219456e-05, + "loss": 1.0728, + "step": 5540 + }, + { + "epoch": 0.36, + "learning_rate": 1.8832521093196508e-05, + "loss": 1.1101, + "step": 5541 + }, + { + "epoch": 0.36, + "learning_rate": 1.8832306404173557e-05, + "loss": 1.1436, + "step": 5542 + }, + { + "epoch": 0.36, + "learning_rate": 1.8832091715150606e-05, + "loss": 1.0707, + "step": 5543 + }, + { + "epoch": 0.36, + "learning_rate": 1.8831877026127655e-05, + "loss": 1.1396, + "step": 5544 + }, + { + "epoch": 0.36, + "learning_rate": 1.8831662337104707e-05, + "loss": 1.1999, + "step": 5545 + }, + { + "epoch": 0.36, + "learning_rate": 1.8831447648081756e-05, + "loss": 1.0595, + "step": 5546 + }, + { + "epoch": 0.36, + "learning_rate": 1.8831232959058805e-05, + "loss": 1.1542, + "step": 5547 + }, + { + "epoch": 0.36, + "learning_rate": 1.8831018270035854e-05, + "loss": 1.0739, + "step": 5548 + }, + { + "epoch": 0.36, + "learning_rate": 1.8830803581012903e-05, + "loss": 1.0585, + "step": 5549 + }, + { + "epoch": 0.36, + "learning_rate": 1.8830588891989952e-05, + "loss": 1.0391, + "step": 5550 + }, + { + "epoch": 0.36, + "learning_rate": 1.8830374202967005e-05, + "loss": 1.1288, + "step": 5551 + }, + { + "epoch": 0.36, + "learning_rate": 1.8830159513944054e-05, + "loss": 1.0886, + "step": 5552 + }, + { + "epoch": 0.36, + "learning_rate": 1.8829944824921102e-05, + "loss": 1.0785, + "step": 5553 + }, + { + "epoch": 0.36, + "learning_rate": 1.882973013589815e-05, + "loss": 1.1427, + "step": 5554 + }, + { + "epoch": 0.36, + "learning_rate": 1.88295154468752e-05, + "loss": 1.1001, + "step": 5555 + }, + { + "epoch": 0.36, + "learning_rate": 1.8829300757852253e-05, + "loss": 1.1091, + "step": 5556 + }, + { + "epoch": 0.36, + "learning_rate": 1.8829086068829302e-05, + "loss": 1.0287, + "step": 5557 + }, + { + "epoch": 0.36, + "learning_rate": 1.8828871379806354e-05, + "loss": 1.109, + "step": 5558 + }, + { + "epoch": 0.36, + "learning_rate": 1.8828656690783403e-05, + "loss": 1.1175, + "step": 5559 + }, + { + "epoch": 0.36, + "learning_rate": 1.8828442001760452e-05, + "loss": 1.1277, + "step": 5560 + }, + { + "epoch": 0.36, + "learning_rate": 1.88282273127375e-05, + "loss": 1.0821, + "step": 5561 + }, + { + "epoch": 0.36, + "learning_rate": 1.882801262371455e-05, + "loss": 1.0791, + "step": 5562 + }, + { + "epoch": 0.36, + "learning_rate": 1.8827797934691602e-05, + "loss": 1.1558, + "step": 5563 + }, + { + "epoch": 0.36, + "learning_rate": 1.882758324566865e-05, + "loss": 1.0229, + "step": 5564 + }, + { + "epoch": 0.36, + "learning_rate": 1.88273685566457e-05, + "loss": 1.1339, + "step": 5565 + }, + { + "epoch": 0.36, + "learning_rate": 1.882715386762275e-05, + "loss": 1.0027, + "step": 5566 + }, + { + "epoch": 0.36, + "learning_rate": 1.8826939178599798e-05, + "loss": 1.0679, + "step": 5567 + }, + { + "epoch": 0.36, + "learning_rate": 1.8826724489576847e-05, + "loss": 1.0832, + "step": 5568 + }, + { + "epoch": 0.36, + "learning_rate": 1.88265098005539e-05, + "loss": 1.045, + "step": 5569 + }, + { + "epoch": 0.36, + "learning_rate": 1.882629511153095e-05, + "loss": 1.077, + "step": 5570 + }, + { + "epoch": 0.36, + "learning_rate": 1.8826080422508e-05, + "loss": 1.0788, + "step": 5571 + }, + { + "epoch": 0.36, + "learning_rate": 1.882586573348505e-05, + "loss": 1.0887, + "step": 5572 + }, + { + "epoch": 0.36, + "learning_rate": 1.88256510444621e-05, + "loss": 1.1651, + "step": 5573 + }, + { + "epoch": 0.36, + "learning_rate": 1.8825436355439148e-05, + "loss": 1.096, + "step": 5574 + }, + { + "epoch": 0.36, + "learning_rate": 1.8825221666416197e-05, + "loss": 1.0183, + "step": 5575 + }, + { + "epoch": 0.36, + "learning_rate": 1.882500697739325e-05, + "loss": 1.2253, + "step": 5576 + }, + { + "epoch": 0.36, + "learning_rate": 1.8824792288370298e-05, + "loss": 1.1395, + "step": 5577 + }, + { + "epoch": 0.36, + "learning_rate": 1.8824577599347347e-05, + "loss": 1.1442, + "step": 5578 + }, + { + "epoch": 0.36, + "learning_rate": 1.8824362910324396e-05, + "loss": 1.1554, + "step": 5579 + }, + { + "epoch": 0.36, + "learning_rate": 1.8824148221301445e-05, + "loss": 1.1708, + "step": 5580 + }, + { + "epoch": 0.36, + "learning_rate": 1.8823933532278497e-05, + "loss": 1.0986, + "step": 5581 + }, + { + "epoch": 0.36, + "learning_rate": 1.8823718843255546e-05, + "loss": 1.1457, + "step": 5582 + }, + { + "epoch": 0.36, + "learning_rate": 1.8823504154232595e-05, + "loss": 1.0584, + "step": 5583 + }, + { + "epoch": 0.36, + "learning_rate": 1.8823289465209644e-05, + "loss": 1.11, + "step": 5584 + }, + { + "epoch": 0.36, + "learning_rate": 1.8823074776186696e-05, + "loss": 1.0636, + "step": 5585 + }, + { + "epoch": 0.36, + "learning_rate": 1.8822860087163745e-05, + "loss": 1.1616, + "step": 5586 + }, + { + "epoch": 0.36, + "learning_rate": 1.8822645398140794e-05, + "loss": 1.1515, + "step": 5587 + }, + { + "epoch": 0.36, + "learning_rate": 1.8822430709117843e-05, + "loss": 1.0702, + "step": 5588 + }, + { + "epoch": 0.36, + "learning_rate": 1.8822216020094896e-05, + "loss": 1.1128, + "step": 5589 + }, + { + "epoch": 0.36, + "learning_rate": 1.8822001331071945e-05, + "loss": 1.199, + "step": 5590 + }, + { + "epoch": 0.36, + "learning_rate": 1.8821786642048994e-05, + "loss": 1.1617, + "step": 5591 + }, + { + "epoch": 0.36, + "learning_rate": 1.8821571953026043e-05, + "loss": 1.082, + "step": 5592 + }, + { + "epoch": 0.36, + "learning_rate": 1.882135726400309e-05, + "loss": 1.119, + "step": 5593 + }, + { + "epoch": 0.36, + "learning_rate": 1.8821142574980144e-05, + "loss": 1.2205, + "step": 5594 + }, + { + "epoch": 0.36, + "learning_rate": 1.8820927885957193e-05, + "loss": 1.0888, + "step": 5595 + }, + { + "epoch": 0.36, + "learning_rate": 1.8820713196934242e-05, + "loss": 1.0434, + "step": 5596 + }, + { + "epoch": 0.36, + "learning_rate": 1.882049850791129e-05, + "loss": 1.1041, + "step": 5597 + }, + { + "epoch": 0.36, + "learning_rate": 1.882028381888834e-05, + "loss": 1.1295, + "step": 5598 + }, + { + "epoch": 0.36, + "learning_rate": 1.8820069129865392e-05, + "loss": 1.1059, + "step": 5599 + }, + { + "epoch": 0.36, + "learning_rate": 1.881985444084244e-05, + "loss": 1.0542, + "step": 5600 + }, + { + "epoch": 0.36, + "learning_rate": 1.8819639751819493e-05, + "loss": 1.1166, + "step": 5601 + }, + { + "epoch": 0.36, + "learning_rate": 1.8819425062796542e-05, + "loss": 1.0459, + "step": 5602 + }, + { + "epoch": 0.36, + "learning_rate": 1.881921037377359e-05, + "loss": 1.2105, + "step": 5603 + }, + { + "epoch": 0.36, + "learning_rate": 1.881899568475064e-05, + "loss": 1.0745, + "step": 5604 + }, + { + "epoch": 0.36, + "learning_rate": 1.881878099572769e-05, + "loss": 1.0992, + "step": 5605 + }, + { + "epoch": 0.36, + "learning_rate": 1.8818566306704738e-05, + "loss": 1.0801, + "step": 5606 + }, + { + "epoch": 0.36, + "learning_rate": 1.881835161768179e-05, + "loss": 1.0883, + "step": 5607 + }, + { + "epoch": 0.36, + "learning_rate": 1.881813692865884e-05, + "loss": 1.082, + "step": 5608 + }, + { + "epoch": 0.36, + "learning_rate": 1.881792223963589e-05, + "loss": 1.0706, + "step": 5609 + }, + { + "epoch": 0.36, + "learning_rate": 1.8817707550612938e-05, + "loss": 1.0208, + "step": 5610 + }, + { + "epoch": 0.36, + "learning_rate": 1.8817492861589986e-05, + "loss": 1.1288, + "step": 5611 + }, + { + "epoch": 0.36, + "learning_rate": 1.881727817256704e-05, + "loss": 1.0375, + "step": 5612 + }, + { + "epoch": 0.36, + "learning_rate": 1.8817063483544088e-05, + "loss": 1.028, + "step": 5613 + }, + { + "epoch": 0.36, + "learning_rate": 1.881684879452114e-05, + "loss": 1.0698, + "step": 5614 + }, + { + "epoch": 0.36, + "learning_rate": 1.881663410549819e-05, + "loss": 1.0659, + "step": 5615 + }, + { + "epoch": 0.36, + "learning_rate": 1.8816419416475238e-05, + "loss": 1.0833, + "step": 5616 + }, + { + "epoch": 0.36, + "learning_rate": 1.8816204727452287e-05, + "loss": 1.1424, + "step": 5617 + }, + { + "epoch": 0.36, + "learning_rate": 1.8815990038429336e-05, + "loss": 0.9544, + "step": 5618 + }, + { + "epoch": 0.36, + "learning_rate": 1.8815775349406385e-05, + "loss": 1.085, + "step": 5619 + }, + { + "epoch": 0.36, + "learning_rate": 1.8815560660383437e-05, + "loss": 1.1607, + "step": 5620 + }, + { + "epoch": 0.36, + "learning_rate": 1.8815345971360486e-05, + "loss": 1.0793, + "step": 5621 + }, + { + "epoch": 0.36, + "learning_rate": 1.8815131282337535e-05, + "loss": 1.0398, + "step": 5622 + }, + { + "epoch": 0.36, + "learning_rate": 1.8814916593314584e-05, + "loss": 1.0725, + "step": 5623 + }, + { + "epoch": 0.36, + "learning_rate": 1.8814701904291633e-05, + "loss": 1.0719, + "step": 5624 + }, + { + "epoch": 0.36, + "learning_rate": 1.8814487215268686e-05, + "loss": 1.1725, + "step": 5625 + }, + { + "epoch": 0.36, + "learning_rate": 1.8814272526245735e-05, + "loss": 1.1141, + "step": 5626 + }, + { + "epoch": 0.36, + "learning_rate": 1.8814057837222783e-05, + "loss": 1.1227, + "step": 5627 + }, + { + "epoch": 0.36, + "learning_rate": 1.8813843148199836e-05, + "loss": 1.0399, + "step": 5628 + }, + { + "epoch": 0.36, + "learning_rate": 1.8813628459176885e-05, + "loss": 1.1881, + "step": 5629 + }, + { + "epoch": 0.36, + "learning_rate": 1.8813413770153934e-05, + "loss": 1.1931, + "step": 5630 + }, + { + "epoch": 0.36, + "learning_rate": 1.8813199081130983e-05, + "loss": 1.1492, + "step": 5631 + }, + { + "epoch": 0.36, + "learning_rate": 1.8812984392108035e-05, + "loss": 1.0456, + "step": 5632 + }, + { + "epoch": 0.36, + "learning_rate": 1.8812769703085084e-05, + "loss": 1.1388, + "step": 5633 + }, + { + "epoch": 0.36, + "learning_rate": 1.8812555014062133e-05, + "loss": 1.1424, + "step": 5634 + }, + { + "epoch": 0.36, + "learning_rate": 1.8812340325039182e-05, + "loss": 1.0715, + "step": 5635 + }, + { + "epoch": 0.36, + "learning_rate": 1.881212563601623e-05, + "loss": 1.1195, + "step": 5636 + }, + { + "epoch": 0.36, + "learning_rate": 1.881191094699328e-05, + "loss": 1.1602, + "step": 5637 + }, + { + "epoch": 0.36, + "learning_rate": 1.8811696257970332e-05, + "loss": 1.1105, + "step": 5638 + }, + { + "epoch": 0.36, + "learning_rate": 1.881148156894738e-05, + "loss": 1.1723, + "step": 5639 + }, + { + "epoch": 0.36, + "learning_rate": 1.881126687992443e-05, + "loss": 1.0536, + "step": 5640 + }, + { + "epoch": 0.36, + "learning_rate": 1.881105219090148e-05, + "loss": 1.1244, + "step": 5641 + }, + { + "epoch": 0.36, + "learning_rate": 1.8810837501878528e-05, + "loss": 1.0191, + "step": 5642 + }, + { + "epoch": 0.36, + "learning_rate": 1.881062281285558e-05, + "loss": 1.113, + "step": 5643 + }, + { + "epoch": 0.36, + "learning_rate": 1.881040812383263e-05, + "loss": 1.1275, + "step": 5644 + }, + { + "epoch": 0.36, + "learning_rate": 1.8810193434809682e-05, + "loss": 1.0786, + "step": 5645 + }, + { + "epoch": 0.36, + "learning_rate": 1.880997874578673e-05, + "loss": 1.035, + "step": 5646 + }, + { + "epoch": 0.36, + "learning_rate": 1.880976405676378e-05, + "loss": 1.0771, + "step": 5647 + }, + { + "epoch": 0.36, + "learning_rate": 1.880954936774083e-05, + "loss": 1.128, + "step": 5648 + }, + { + "epoch": 0.36, + "learning_rate": 1.8809334678717878e-05, + "loss": 1.1049, + "step": 5649 + }, + { + "epoch": 0.36, + "learning_rate": 1.880911998969493e-05, + "loss": 1.1242, + "step": 5650 + }, + { + "epoch": 0.36, + "learning_rate": 1.880890530067198e-05, + "loss": 1.0636, + "step": 5651 + }, + { + "epoch": 0.36, + "learning_rate": 1.8808690611649028e-05, + "loss": 1.0661, + "step": 5652 + }, + { + "epoch": 0.36, + "learning_rate": 1.8808475922626077e-05, + "loss": 1.1413, + "step": 5653 + }, + { + "epoch": 0.36, + "learning_rate": 1.8808261233603126e-05, + "loss": 1.1864, + "step": 5654 + }, + { + "epoch": 0.36, + "learning_rate": 1.8808046544580175e-05, + "loss": 1.0915, + "step": 5655 + }, + { + "epoch": 0.36, + "learning_rate": 1.8807831855557227e-05, + "loss": 1.1338, + "step": 5656 + }, + { + "epoch": 0.36, + "learning_rate": 1.8807617166534276e-05, + "loss": 1.1205, + "step": 5657 + }, + { + "epoch": 0.36, + "learning_rate": 1.880740247751133e-05, + "loss": 1.1166, + "step": 5658 + }, + { + "epoch": 0.36, + "learning_rate": 1.8807187788488377e-05, + "loss": 1.0338, + "step": 5659 + }, + { + "epoch": 0.36, + "learning_rate": 1.8806973099465426e-05, + "loss": 1.1869, + "step": 5660 + }, + { + "epoch": 0.36, + "learning_rate": 1.8806758410442475e-05, + "loss": 1.1212, + "step": 5661 + }, + { + "epoch": 0.36, + "learning_rate": 1.8806543721419524e-05, + "loss": 1.0451, + "step": 5662 + }, + { + "epoch": 0.36, + "learning_rate": 1.8806329032396577e-05, + "loss": 1.0738, + "step": 5663 + }, + { + "epoch": 0.36, + "learning_rate": 1.8806114343373626e-05, + "loss": 1.0848, + "step": 5664 + }, + { + "epoch": 0.36, + "learning_rate": 1.8805899654350675e-05, + "loss": 1.1561, + "step": 5665 + }, + { + "epoch": 0.36, + "learning_rate": 1.8805684965327724e-05, + "loss": 1.1049, + "step": 5666 + }, + { + "epoch": 0.36, + "learning_rate": 1.8805470276304773e-05, + "loss": 1.1425, + "step": 5667 + }, + { + "epoch": 0.36, + "learning_rate": 1.880525558728182e-05, + "loss": 1.1324, + "step": 5668 + }, + { + "epoch": 0.36, + "learning_rate": 1.8805040898258874e-05, + "loss": 1.0374, + "step": 5669 + }, + { + "epoch": 0.36, + "learning_rate": 1.8804826209235923e-05, + "loss": 1.1091, + "step": 5670 + }, + { + "epoch": 0.36, + "learning_rate": 1.8804611520212972e-05, + "loss": 1.0199, + "step": 5671 + }, + { + "epoch": 0.36, + "learning_rate": 1.8804396831190024e-05, + "loss": 1.0729, + "step": 5672 + }, + { + "epoch": 0.36, + "learning_rate": 1.8804182142167073e-05, + "loss": 1.0574, + "step": 5673 + }, + { + "epoch": 0.37, + "learning_rate": 1.8803967453144122e-05, + "loss": 1.0992, + "step": 5674 + }, + { + "epoch": 0.37, + "learning_rate": 1.880375276412117e-05, + "loss": 1.0963, + "step": 5675 + }, + { + "epoch": 0.37, + "learning_rate": 1.8803538075098223e-05, + "loss": 1.11, + "step": 5676 + }, + { + "epoch": 0.37, + "learning_rate": 1.8803323386075272e-05, + "loss": 1.1233, + "step": 5677 + }, + { + "epoch": 0.37, + "learning_rate": 1.880310869705232e-05, + "loss": 1.0722, + "step": 5678 + }, + { + "epoch": 0.37, + "learning_rate": 1.880289400802937e-05, + "loss": 1.0952, + "step": 5679 + }, + { + "epoch": 0.37, + "learning_rate": 1.880267931900642e-05, + "loss": 1.0618, + "step": 5680 + }, + { + "epoch": 0.37, + "learning_rate": 1.880246462998347e-05, + "loss": 1.1231, + "step": 5681 + }, + { + "epoch": 0.37, + "learning_rate": 1.880224994096052e-05, + "loss": 1.0591, + "step": 5682 + }, + { + "epoch": 0.37, + "learning_rate": 1.880203525193757e-05, + "loss": 1.0715, + "step": 5683 + }, + { + "epoch": 0.37, + "learning_rate": 1.880182056291462e-05, + "loss": 1.1481, + "step": 5684 + }, + { + "epoch": 0.37, + "learning_rate": 1.8801605873891667e-05, + "loss": 1.1464, + "step": 5685 + }, + { + "epoch": 0.37, + "learning_rate": 1.880139118486872e-05, + "loss": 1.087, + "step": 5686 + }, + { + "epoch": 0.37, + "learning_rate": 1.880117649584577e-05, + "loss": 1.1277, + "step": 5687 + }, + { + "epoch": 0.37, + "learning_rate": 1.8800961806822818e-05, + "loss": 1.109, + "step": 5688 + }, + { + "epoch": 0.37, + "learning_rate": 1.880074711779987e-05, + "loss": 1.0798, + "step": 5689 + }, + { + "epoch": 0.37, + "learning_rate": 1.880053242877692e-05, + "loss": 1.1262, + "step": 5690 + }, + { + "epoch": 0.37, + "learning_rate": 1.8800317739753968e-05, + "loss": 1.1036, + "step": 5691 + }, + { + "epoch": 0.37, + "learning_rate": 1.8800103050731017e-05, + "loss": 1.147, + "step": 5692 + }, + { + "epoch": 0.37, + "learning_rate": 1.8799888361708066e-05, + "loss": 0.9771, + "step": 5693 + }, + { + "epoch": 0.37, + "learning_rate": 1.879967367268512e-05, + "loss": 1.0966, + "step": 5694 + }, + { + "epoch": 0.37, + "learning_rate": 1.8799458983662167e-05, + "loss": 1.0496, + "step": 5695 + }, + { + "epoch": 0.37, + "learning_rate": 1.8799244294639216e-05, + "loss": 0.951, + "step": 5696 + }, + { + "epoch": 0.37, + "learning_rate": 1.8799029605616265e-05, + "loss": 1.1013, + "step": 5697 + }, + { + "epoch": 0.37, + "learning_rate": 1.8798814916593314e-05, + "loss": 1.1525, + "step": 5698 + }, + { + "epoch": 0.37, + "learning_rate": 1.8798600227570367e-05, + "loss": 1.172, + "step": 5699 + }, + { + "epoch": 0.37, + "learning_rate": 1.8798385538547416e-05, + "loss": 1.0889, + "step": 5700 + }, + { + "epoch": 0.37, + "learning_rate": 1.8798170849524468e-05, + "loss": 1.0316, + "step": 5701 + }, + { + "epoch": 0.37, + "learning_rate": 1.8797956160501517e-05, + "loss": 1.1679, + "step": 5702 + }, + { + "epoch": 0.37, + "learning_rate": 1.8797741471478566e-05, + "loss": 1.1274, + "step": 5703 + }, + { + "epoch": 0.37, + "learning_rate": 1.8797526782455615e-05, + "loss": 1.1388, + "step": 5704 + }, + { + "epoch": 0.37, + "learning_rate": 1.8797312093432664e-05, + "loss": 1.1388, + "step": 5705 + }, + { + "epoch": 0.37, + "learning_rate": 1.8797097404409713e-05, + "loss": 1.0861, + "step": 5706 + }, + { + "epoch": 0.37, + "learning_rate": 1.8796882715386765e-05, + "loss": 1.1187, + "step": 5707 + }, + { + "epoch": 0.37, + "learning_rate": 1.8796668026363814e-05, + "loss": 1.0592, + "step": 5708 + }, + { + "epoch": 0.37, + "learning_rate": 1.8796453337340863e-05, + "loss": 1.0692, + "step": 5709 + }, + { + "epoch": 0.37, + "learning_rate": 1.8796238648317912e-05, + "loss": 1.1727, + "step": 5710 + }, + { + "epoch": 0.37, + "learning_rate": 1.879602395929496e-05, + "loss": 1.1212, + "step": 5711 + }, + { + "epoch": 0.37, + "learning_rate": 1.8795809270272013e-05, + "loss": 1.047, + "step": 5712 + }, + { + "epoch": 0.37, + "learning_rate": 1.8795594581249062e-05, + "loss": 1.2056, + "step": 5713 + }, + { + "epoch": 0.37, + "learning_rate": 1.879537989222611e-05, + "loss": 1.1166, + "step": 5714 + }, + { + "epoch": 0.37, + "learning_rate": 1.8795165203203164e-05, + "loss": 1.0234, + "step": 5715 + }, + { + "epoch": 0.37, + "learning_rate": 1.8794950514180213e-05, + "loss": 1.0788, + "step": 5716 + }, + { + "epoch": 0.37, + "learning_rate": 1.879473582515726e-05, + "loss": 1.037, + "step": 5717 + }, + { + "epoch": 0.37, + "learning_rate": 1.879452113613431e-05, + "loss": 1.1194, + "step": 5718 + }, + { + "epoch": 0.37, + "learning_rate": 1.8794306447111363e-05, + "loss": 1.0475, + "step": 5719 + }, + { + "epoch": 0.37, + "learning_rate": 1.8794091758088412e-05, + "loss": 1.1331, + "step": 5720 + }, + { + "epoch": 0.37, + "learning_rate": 1.879387706906546e-05, + "loss": 1.0383, + "step": 5721 + }, + { + "epoch": 0.37, + "learning_rate": 1.879366238004251e-05, + "loss": 1.0883, + "step": 5722 + }, + { + "epoch": 0.37, + "learning_rate": 1.879344769101956e-05, + "loss": 1.1551, + "step": 5723 + }, + { + "epoch": 0.37, + "learning_rate": 1.8793233001996608e-05, + "loss": 1.1389, + "step": 5724 + }, + { + "epoch": 0.37, + "learning_rate": 1.879301831297366e-05, + "loss": 1.1007, + "step": 5725 + }, + { + "epoch": 0.37, + "learning_rate": 1.879280362395071e-05, + "loss": 1.086, + "step": 5726 + }, + { + "epoch": 0.37, + "learning_rate": 1.8792588934927758e-05, + "loss": 1.1541, + "step": 5727 + }, + { + "epoch": 0.37, + "learning_rate": 1.8792374245904807e-05, + "loss": 1.1443, + "step": 5728 + }, + { + "epoch": 0.37, + "learning_rate": 1.8792159556881856e-05, + "loss": 1.1103, + "step": 5729 + }, + { + "epoch": 0.37, + "learning_rate": 1.8791944867858908e-05, + "loss": 1.2186, + "step": 5730 + }, + { + "epoch": 0.37, + "learning_rate": 1.8791730178835957e-05, + "loss": 1.1814, + "step": 5731 + }, + { + "epoch": 0.37, + "learning_rate": 1.879151548981301e-05, + "loss": 1.1147, + "step": 5732 + }, + { + "epoch": 0.37, + "learning_rate": 1.879130080079006e-05, + "loss": 1.1818, + "step": 5733 + }, + { + "epoch": 0.37, + "learning_rate": 1.8791086111767107e-05, + "loss": 1.0546, + "step": 5734 + }, + { + "epoch": 0.37, + "learning_rate": 1.8790871422744156e-05, + "loss": 1.0786, + "step": 5735 + }, + { + "epoch": 0.37, + "learning_rate": 1.8790656733721205e-05, + "loss": 1.1183, + "step": 5736 + }, + { + "epoch": 0.37, + "learning_rate": 1.8790442044698258e-05, + "loss": 1.0636, + "step": 5737 + }, + { + "epoch": 0.37, + "learning_rate": 1.8790227355675307e-05, + "loss": 1.1088, + "step": 5738 + }, + { + "epoch": 0.37, + "learning_rate": 1.8790012666652356e-05, + "loss": 1.1597, + "step": 5739 + }, + { + "epoch": 0.37, + "learning_rate": 1.8789797977629405e-05, + "loss": 1.2251, + "step": 5740 + }, + { + "epoch": 0.37, + "learning_rate": 1.8789583288606454e-05, + "loss": 1.1471, + "step": 5741 + }, + { + "epoch": 0.37, + "learning_rate": 1.8789368599583503e-05, + "loss": 1.1032, + "step": 5742 + }, + { + "epoch": 0.37, + "learning_rate": 1.8789153910560555e-05, + "loss": 1.0997, + "step": 5743 + }, + { + "epoch": 0.37, + "learning_rate": 1.8788939221537604e-05, + "loss": 1.1678, + "step": 5744 + }, + { + "epoch": 0.37, + "learning_rate": 1.8788724532514656e-05, + "loss": 1.0849, + "step": 5745 + }, + { + "epoch": 0.37, + "learning_rate": 1.8788509843491705e-05, + "loss": 1.1049, + "step": 5746 + }, + { + "epoch": 0.37, + "learning_rate": 1.8788295154468754e-05, + "loss": 1.1479, + "step": 5747 + }, + { + "epoch": 0.37, + "learning_rate": 1.8788080465445803e-05, + "loss": 1.1338, + "step": 5748 + }, + { + "epoch": 0.37, + "learning_rate": 1.8787865776422852e-05, + "loss": 1.1116, + "step": 5749 + }, + { + "epoch": 0.37, + "learning_rate": 1.8787651087399904e-05, + "loss": 1.1179, + "step": 5750 + }, + { + "epoch": 0.37, + "learning_rate": 1.8787436398376953e-05, + "loss": 1.0119, + "step": 5751 + }, + { + "epoch": 0.37, + "learning_rate": 1.8787221709354002e-05, + "loss": 0.9689, + "step": 5752 + }, + { + "epoch": 0.37, + "learning_rate": 1.878700702033105e-05, + "loss": 1.1718, + "step": 5753 + }, + { + "epoch": 0.37, + "learning_rate": 1.87867923313081e-05, + "loss": 1.136, + "step": 5754 + }, + { + "epoch": 0.37, + "learning_rate": 1.878657764228515e-05, + "loss": 1.0403, + "step": 5755 + }, + { + "epoch": 0.37, + "learning_rate": 1.87863629532622e-05, + "loss": 1.1285, + "step": 5756 + }, + { + "epoch": 0.37, + "learning_rate": 1.878614826423925e-05, + "loss": 1.106, + "step": 5757 + }, + { + "epoch": 0.37, + "learning_rate": 1.87859335752163e-05, + "loss": 1.1381, + "step": 5758 + }, + { + "epoch": 0.37, + "learning_rate": 1.8785718886193352e-05, + "loss": 1.1144, + "step": 5759 + }, + { + "epoch": 0.37, + "learning_rate": 1.87855041971704e-05, + "loss": 1.066, + "step": 5760 + }, + { + "epoch": 0.37, + "learning_rate": 1.878528950814745e-05, + "loss": 1.2302, + "step": 5761 + }, + { + "epoch": 0.37, + "learning_rate": 1.87850748191245e-05, + "loss": 1.1082, + "step": 5762 + }, + { + "epoch": 0.37, + "learning_rate": 1.878486013010155e-05, + "loss": 1.111, + "step": 5763 + }, + { + "epoch": 0.37, + "learning_rate": 1.87846454410786e-05, + "loss": 1.0875, + "step": 5764 + }, + { + "epoch": 0.37, + "learning_rate": 1.878443075205565e-05, + "loss": 1.0544, + "step": 5765 + }, + { + "epoch": 0.37, + "learning_rate": 1.8784216063032698e-05, + "loss": 1.0888, + "step": 5766 + }, + { + "epoch": 0.37, + "learning_rate": 1.8784001374009747e-05, + "loss": 1.0442, + "step": 5767 + }, + { + "epoch": 0.37, + "learning_rate": 1.87837866849868e-05, + "loss": 1.1406, + "step": 5768 + }, + { + "epoch": 0.37, + "learning_rate": 1.8783571995963848e-05, + "loss": 1.0732, + "step": 5769 + }, + { + "epoch": 0.37, + "learning_rate": 1.8783357306940897e-05, + "loss": 1.2046, + "step": 5770 + }, + { + "epoch": 0.37, + "learning_rate": 1.8783142617917946e-05, + "loss": 1.0796, + "step": 5771 + }, + { + "epoch": 0.37, + "learning_rate": 1.8782927928894995e-05, + "loss": 1.1408, + "step": 5772 + }, + { + "epoch": 0.37, + "learning_rate": 1.8782713239872048e-05, + "loss": 1.06, + "step": 5773 + }, + { + "epoch": 0.37, + "learning_rate": 1.8782498550849097e-05, + "loss": 1.0984, + "step": 5774 + }, + { + "epoch": 0.37, + "learning_rate": 1.8782283861826145e-05, + "loss": 1.0692, + "step": 5775 + }, + { + "epoch": 0.37, + "learning_rate": 1.8782069172803198e-05, + "loss": 1.105, + "step": 5776 + }, + { + "epoch": 0.37, + "learning_rate": 1.8781854483780247e-05, + "loss": 0.9768, + "step": 5777 + }, + { + "epoch": 0.37, + "learning_rate": 1.8781639794757296e-05, + "loss": 1.0769, + "step": 5778 + }, + { + "epoch": 0.37, + "learning_rate": 1.8781425105734345e-05, + "loss": 1.1696, + "step": 5779 + }, + { + "epoch": 0.37, + "learning_rate": 1.8781210416711394e-05, + "loss": 1.0332, + "step": 5780 + }, + { + "epoch": 0.37, + "learning_rate": 1.8780995727688446e-05, + "loss": 1.0945, + "step": 5781 + }, + { + "epoch": 0.37, + "learning_rate": 1.8780781038665495e-05, + "loss": 1.1728, + "step": 5782 + }, + { + "epoch": 0.37, + "learning_rate": 1.8780566349642544e-05, + "loss": 1.0321, + "step": 5783 + }, + { + "epoch": 0.37, + "learning_rate": 1.8780351660619593e-05, + "loss": 1.1649, + "step": 5784 + }, + { + "epoch": 0.37, + "learning_rate": 1.8780136971596642e-05, + "loss": 1.1995, + "step": 5785 + }, + { + "epoch": 0.37, + "learning_rate": 1.8779922282573694e-05, + "loss": 1.1841, + "step": 5786 + }, + { + "epoch": 0.37, + "learning_rate": 1.8779707593550743e-05, + "loss": 1.0755, + "step": 5787 + }, + { + "epoch": 0.37, + "learning_rate": 1.8779492904527796e-05, + "loss": 1.1963, + "step": 5788 + }, + { + "epoch": 0.37, + "learning_rate": 1.8779278215504845e-05, + "loss": 1.1983, + "step": 5789 + }, + { + "epoch": 0.37, + "learning_rate": 1.8779063526481894e-05, + "loss": 1.1357, + "step": 5790 + }, + { + "epoch": 0.37, + "learning_rate": 1.8778848837458942e-05, + "loss": 1.122, + "step": 5791 + }, + { + "epoch": 0.37, + "learning_rate": 1.877863414843599e-05, + "loss": 1.061, + "step": 5792 + }, + { + "epoch": 0.37, + "learning_rate": 1.877841945941304e-05, + "loss": 1.1178, + "step": 5793 + }, + { + "epoch": 0.37, + "learning_rate": 1.8778204770390093e-05, + "loss": 1.1013, + "step": 5794 + }, + { + "epoch": 0.37, + "learning_rate": 1.8777990081367142e-05, + "loss": 1.0937, + "step": 5795 + }, + { + "epoch": 0.37, + "learning_rate": 1.877777539234419e-05, + "loss": 1.129, + "step": 5796 + }, + { + "epoch": 0.37, + "learning_rate": 1.877756070332124e-05, + "loss": 1.1339, + "step": 5797 + }, + { + "epoch": 0.37, + "learning_rate": 1.877734601429829e-05, + "loss": 1.0641, + "step": 5798 + }, + { + "epoch": 0.37, + "learning_rate": 1.877713132527534e-05, + "loss": 1.1277, + "step": 5799 + }, + { + "epoch": 0.37, + "learning_rate": 1.877691663625239e-05, + "loss": 1.117, + "step": 5800 + }, + { + "epoch": 0.37, + "learning_rate": 1.877670194722944e-05, + "loss": 1.0341, + "step": 5801 + }, + { + "epoch": 0.37, + "learning_rate": 1.877648725820649e-05, + "loss": 1.0188, + "step": 5802 + }, + { + "epoch": 0.37, + "learning_rate": 1.877627256918354e-05, + "loss": 1.0079, + "step": 5803 + }, + { + "epoch": 0.37, + "learning_rate": 1.877605788016059e-05, + "loss": 1.0694, + "step": 5804 + }, + { + "epoch": 0.37, + "learning_rate": 1.8775843191137638e-05, + "loss": 1.1629, + "step": 5805 + }, + { + "epoch": 0.37, + "learning_rate": 1.877562850211469e-05, + "loss": 1.1512, + "step": 5806 + }, + { + "epoch": 0.37, + "learning_rate": 1.877541381309174e-05, + "loss": 1.161, + "step": 5807 + }, + { + "epoch": 0.37, + "learning_rate": 1.877519912406879e-05, + "loss": 1.1739, + "step": 5808 + }, + { + "epoch": 0.37, + "learning_rate": 1.8774984435045837e-05, + "loss": 1.1387, + "step": 5809 + }, + { + "epoch": 0.37, + "learning_rate": 1.8774769746022886e-05, + "loss": 1.0878, + "step": 5810 + }, + { + "epoch": 0.37, + "learning_rate": 1.8774555056999935e-05, + "loss": 1.0708, + "step": 5811 + }, + { + "epoch": 0.37, + "learning_rate": 1.8774340367976988e-05, + "loss": 1.0889, + "step": 5812 + }, + { + "epoch": 0.37, + "learning_rate": 1.8774125678954037e-05, + "loss": 1.1014, + "step": 5813 + }, + { + "epoch": 0.37, + "learning_rate": 1.8773910989931086e-05, + "loss": 1.0654, + "step": 5814 + }, + { + "epoch": 0.37, + "learning_rate": 1.8773696300908135e-05, + "loss": 1.1667, + "step": 5815 + }, + { + "epoch": 0.37, + "learning_rate": 1.8773481611885184e-05, + "loss": 1.0406, + "step": 5816 + }, + { + "epoch": 0.37, + "learning_rate": 1.8773266922862236e-05, + "loss": 1.1272, + "step": 5817 + }, + { + "epoch": 0.37, + "learning_rate": 1.8773052233839285e-05, + "loss": 1.1303, + "step": 5818 + }, + { + "epoch": 0.37, + "learning_rate": 1.8772837544816337e-05, + "loss": 1.2023, + "step": 5819 + }, + { + "epoch": 0.37, + "learning_rate": 1.8772622855793386e-05, + "loss": 1.1235, + "step": 5820 + }, + { + "epoch": 0.37, + "learning_rate": 1.8772408166770435e-05, + "loss": 1.0626, + "step": 5821 + }, + { + "epoch": 0.37, + "learning_rate": 1.8772193477747484e-05, + "loss": 1.0977, + "step": 5822 + }, + { + "epoch": 0.37, + "learning_rate": 1.8771978788724533e-05, + "loss": 1.0886, + "step": 5823 + }, + { + "epoch": 0.37, + "learning_rate": 1.8771764099701585e-05, + "loss": 1.0978, + "step": 5824 + }, + { + "epoch": 0.37, + "learning_rate": 1.8771549410678634e-05, + "loss": 1.0952, + "step": 5825 + }, + { + "epoch": 0.37, + "learning_rate": 1.8771334721655683e-05, + "loss": 1.1202, + "step": 5826 + }, + { + "epoch": 0.37, + "learning_rate": 1.8771120032632732e-05, + "loss": 1.1307, + "step": 5827 + }, + { + "epoch": 0.37, + "learning_rate": 1.877090534360978e-05, + "loss": 1.1534, + "step": 5828 + }, + { + "epoch": 0.38, + "learning_rate": 1.877069065458683e-05, + "loss": 1.021, + "step": 5829 + }, + { + "epoch": 0.38, + "learning_rate": 1.8770475965563883e-05, + "loss": 1.1197, + "step": 5830 + }, + { + "epoch": 0.38, + "learning_rate": 1.877026127654093e-05, + "loss": 1.0677, + "step": 5831 + }, + { + "epoch": 0.38, + "learning_rate": 1.8770046587517984e-05, + "loss": 1.1635, + "step": 5832 + }, + { + "epoch": 0.38, + "learning_rate": 1.8769831898495033e-05, + "loss": 1.1088, + "step": 5833 + }, + { + "epoch": 0.38, + "learning_rate": 1.8769617209472082e-05, + "loss": 0.9925, + "step": 5834 + }, + { + "epoch": 0.38, + "learning_rate": 1.876940252044913e-05, + "loss": 1.1138, + "step": 5835 + }, + { + "epoch": 0.38, + "learning_rate": 1.876918783142618e-05, + "loss": 1.0973, + "step": 5836 + }, + { + "epoch": 0.38, + "learning_rate": 1.8768973142403232e-05, + "loss": 1.1059, + "step": 5837 + }, + { + "epoch": 0.38, + "learning_rate": 1.876875845338028e-05, + "loss": 1.1067, + "step": 5838 + }, + { + "epoch": 0.38, + "learning_rate": 1.876854376435733e-05, + "loss": 1.1412, + "step": 5839 + }, + { + "epoch": 0.38, + "learning_rate": 1.876832907533438e-05, + "loss": 1.0965, + "step": 5840 + }, + { + "epoch": 0.38, + "learning_rate": 1.8768114386311428e-05, + "loss": 1.0188, + "step": 5841 + }, + { + "epoch": 0.38, + "learning_rate": 1.8767899697288477e-05, + "loss": 1.123, + "step": 5842 + }, + { + "epoch": 0.38, + "learning_rate": 1.876768500826553e-05, + "loss": 1.1353, + "step": 5843 + }, + { + "epoch": 0.38, + "learning_rate": 1.8767470319242578e-05, + "loss": 1.0999, + "step": 5844 + }, + { + "epoch": 0.38, + "learning_rate": 1.8767255630219627e-05, + "loss": 1.1332, + "step": 5845 + }, + { + "epoch": 0.38, + "learning_rate": 1.876704094119668e-05, + "loss": 1.1463, + "step": 5846 + }, + { + "epoch": 0.38, + "learning_rate": 1.876682625217373e-05, + "loss": 1.1651, + "step": 5847 + }, + { + "epoch": 0.38, + "learning_rate": 1.8766611563150778e-05, + "loss": 1.0632, + "step": 5848 + }, + { + "epoch": 0.38, + "learning_rate": 1.8766396874127826e-05, + "loss": 1.1542, + "step": 5849 + }, + { + "epoch": 0.38, + "learning_rate": 1.876618218510488e-05, + "loss": 1.1678, + "step": 5850 + }, + { + "epoch": 0.38, + "learning_rate": 1.8765967496081928e-05, + "loss": 1.0917, + "step": 5851 + }, + { + "epoch": 0.38, + "learning_rate": 1.8765752807058977e-05, + "loss": 1.0788, + "step": 5852 + }, + { + "epoch": 0.38, + "learning_rate": 1.8765538118036026e-05, + "loss": 1.0868, + "step": 5853 + }, + { + "epoch": 0.38, + "learning_rate": 1.8765323429013075e-05, + "loss": 1.0072, + "step": 5854 + }, + { + "epoch": 0.38, + "learning_rate": 1.8765108739990127e-05, + "loss": 1.1547, + "step": 5855 + }, + { + "epoch": 0.38, + "learning_rate": 1.8764894050967176e-05, + "loss": 1.135, + "step": 5856 + }, + { + "epoch": 0.38, + "learning_rate": 1.8764679361944225e-05, + "loss": 1.096, + "step": 5857 + }, + { + "epoch": 0.38, + "learning_rate": 1.8764464672921274e-05, + "loss": 1.0712, + "step": 5858 + }, + { + "epoch": 0.38, + "learning_rate": 1.8764249983898323e-05, + "loss": 1.1398, + "step": 5859 + }, + { + "epoch": 0.38, + "learning_rate": 1.8764035294875372e-05, + "loss": 1.1646, + "step": 5860 + }, + { + "epoch": 0.38, + "learning_rate": 1.8763820605852424e-05, + "loss": 1.1584, + "step": 5861 + }, + { + "epoch": 0.38, + "learning_rate": 1.8763605916829473e-05, + "loss": 1.2331, + "step": 5862 + }, + { + "epoch": 0.38, + "learning_rate": 1.8763391227806526e-05, + "loss": 1.0658, + "step": 5863 + }, + { + "epoch": 0.38, + "learning_rate": 1.8763176538783575e-05, + "loss": 1.1748, + "step": 5864 + }, + { + "epoch": 0.38, + "learning_rate": 1.8762961849760623e-05, + "loss": 1.1242, + "step": 5865 + }, + { + "epoch": 0.38, + "learning_rate": 1.8762747160737672e-05, + "loss": 1.0642, + "step": 5866 + }, + { + "epoch": 0.38, + "learning_rate": 1.876253247171472e-05, + "loss": 1.0399, + "step": 5867 + }, + { + "epoch": 0.38, + "learning_rate": 1.8762317782691774e-05, + "loss": 1.0717, + "step": 5868 + }, + { + "epoch": 0.38, + "learning_rate": 1.8762103093668823e-05, + "loss": 1.091, + "step": 5869 + }, + { + "epoch": 0.38, + "learning_rate": 1.876188840464587e-05, + "loss": 1.0922, + "step": 5870 + }, + { + "epoch": 0.38, + "learning_rate": 1.876167371562292e-05, + "loss": 1.0709, + "step": 5871 + }, + { + "epoch": 0.38, + "learning_rate": 1.876145902659997e-05, + "loss": 1.089, + "step": 5872 + }, + { + "epoch": 0.38, + "learning_rate": 1.8761244337577022e-05, + "loss": 1.1452, + "step": 5873 + }, + { + "epoch": 0.38, + "learning_rate": 1.876102964855407e-05, + "loss": 1.1082, + "step": 5874 + }, + { + "epoch": 0.38, + "learning_rate": 1.8760814959531123e-05, + "loss": 1.0606, + "step": 5875 + }, + { + "epoch": 0.38, + "learning_rate": 1.8760600270508172e-05, + "loss": 1.0544, + "step": 5876 + }, + { + "epoch": 0.38, + "learning_rate": 1.876038558148522e-05, + "loss": 1.077, + "step": 5877 + }, + { + "epoch": 0.38, + "learning_rate": 1.876017089246227e-05, + "loss": 1.1555, + "step": 5878 + }, + { + "epoch": 0.38, + "learning_rate": 1.875995620343932e-05, + "loss": 1.1367, + "step": 5879 + }, + { + "epoch": 0.38, + "learning_rate": 1.8759741514416368e-05, + "loss": 1.0979, + "step": 5880 + }, + { + "epoch": 0.38, + "learning_rate": 1.875952682539342e-05, + "loss": 1.1145, + "step": 5881 + }, + { + "epoch": 0.38, + "learning_rate": 1.875931213637047e-05, + "loss": 1.1297, + "step": 5882 + }, + { + "epoch": 0.38, + "learning_rate": 1.875909744734752e-05, + "loss": 1.1059, + "step": 5883 + }, + { + "epoch": 0.38, + "learning_rate": 1.8758882758324567e-05, + "loss": 1.1624, + "step": 5884 + }, + { + "epoch": 0.38, + "learning_rate": 1.8758668069301616e-05, + "loss": 1.2265, + "step": 5885 + }, + { + "epoch": 0.38, + "learning_rate": 1.875845338027867e-05, + "loss": 0.9749, + "step": 5886 + }, + { + "epoch": 0.38, + "learning_rate": 1.8758238691255718e-05, + "loss": 1.1091, + "step": 5887 + }, + { + "epoch": 0.38, + "learning_rate": 1.8758024002232767e-05, + "loss": 1.0949, + "step": 5888 + }, + { + "epoch": 0.38, + "learning_rate": 1.8757809313209816e-05, + "loss": 1.1167, + "step": 5889 + }, + { + "epoch": 0.38, + "learning_rate": 1.8757594624186868e-05, + "loss": 1.2185, + "step": 5890 + }, + { + "epoch": 0.38, + "learning_rate": 1.8757379935163917e-05, + "loss": 1.0444, + "step": 5891 + }, + { + "epoch": 0.38, + "learning_rate": 1.8757165246140966e-05, + "loss": 1.1128, + "step": 5892 + }, + { + "epoch": 0.38, + "learning_rate": 1.8756950557118018e-05, + "loss": 1.0369, + "step": 5893 + }, + { + "epoch": 0.38, + "learning_rate": 1.8756735868095067e-05, + "loss": 1.0674, + "step": 5894 + }, + { + "epoch": 0.38, + "learning_rate": 1.8756521179072116e-05, + "loss": 1.1179, + "step": 5895 + }, + { + "epoch": 0.38, + "learning_rate": 1.8756306490049165e-05, + "loss": 1.1001, + "step": 5896 + }, + { + "epoch": 0.38, + "learning_rate": 1.8756091801026214e-05, + "loss": 1.1878, + "step": 5897 + }, + { + "epoch": 0.38, + "learning_rate": 1.8755877112003263e-05, + "loss": 1.1286, + "step": 5898 + }, + { + "epoch": 0.38, + "learning_rate": 1.8755662422980315e-05, + "loss": 1.0535, + "step": 5899 + }, + { + "epoch": 0.38, + "learning_rate": 1.8755447733957364e-05, + "loss": 1.1139, + "step": 5900 + }, + { + "epoch": 0.38, + "learning_rate": 1.8755233044934413e-05, + "loss": 1.1256, + "step": 5901 + }, + { + "epoch": 0.38, + "learning_rate": 1.8755018355911462e-05, + "loss": 1.0898, + "step": 5902 + }, + { + "epoch": 0.38, + "learning_rate": 1.875480366688851e-05, + "loss": 0.993, + "step": 5903 + }, + { + "epoch": 0.38, + "learning_rate": 1.8754588977865564e-05, + "loss": 0.9874, + "step": 5904 + }, + { + "epoch": 0.38, + "learning_rate": 1.8754374288842613e-05, + "loss": 1.0459, + "step": 5905 + }, + { + "epoch": 0.38, + "learning_rate": 1.8754159599819665e-05, + "loss": 1.1919, + "step": 5906 + }, + { + "epoch": 0.38, + "learning_rate": 1.8753944910796714e-05, + "loss": 1.135, + "step": 5907 + }, + { + "epoch": 0.38, + "learning_rate": 1.8753730221773763e-05, + "loss": 1.141, + "step": 5908 + }, + { + "epoch": 0.38, + "learning_rate": 1.8753515532750812e-05, + "loss": 1.0481, + "step": 5909 + }, + { + "epoch": 0.38, + "learning_rate": 1.875330084372786e-05, + "loss": 1.0807, + "step": 5910 + }, + { + "epoch": 0.38, + "learning_rate": 1.8753086154704913e-05, + "loss": 1.1651, + "step": 5911 + }, + { + "epoch": 0.38, + "learning_rate": 1.8752871465681962e-05, + "loss": 1.0118, + "step": 5912 + }, + { + "epoch": 0.38, + "learning_rate": 1.875265677665901e-05, + "loss": 1.2092, + "step": 5913 + }, + { + "epoch": 0.38, + "learning_rate": 1.875244208763606e-05, + "loss": 1.0543, + "step": 5914 + }, + { + "epoch": 0.38, + "learning_rate": 1.875222739861311e-05, + "loss": 1.1816, + "step": 5915 + }, + { + "epoch": 0.38, + "learning_rate": 1.8752012709590158e-05, + "loss": 1.1634, + "step": 5916 + }, + { + "epoch": 0.38, + "learning_rate": 1.875179802056721e-05, + "loss": 1.0952, + "step": 5917 + }, + { + "epoch": 0.38, + "learning_rate": 1.875158333154426e-05, + "loss": 1.0474, + "step": 5918 + }, + { + "epoch": 0.38, + "learning_rate": 1.875136864252131e-05, + "loss": 1.1588, + "step": 5919 + }, + { + "epoch": 0.38, + "learning_rate": 1.875115395349836e-05, + "loss": 1.098, + "step": 5920 + }, + { + "epoch": 0.38, + "learning_rate": 1.875093926447541e-05, + "loss": 1.1084, + "step": 5921 + }, + { + "epoch": 0.38, + "learning_rate": 1.875072457545246e-05, + "loss": 1.0803, + "step": 5922 + }, + { + "epoch": 0.38, + "learning_rate": 1.8750509886429507e-05, + "loss": 1.0861, + "step": 5923 + }, + { + "epoch": 0.38, + "learning_rate": 1.875029519740656e-05, + "loss": 1.1702, + "step": 5924 + }, + { + "epoch": 0.38, + "learning_rate": 1.875008050838361e-05, + "loss": 1.0485, + "step": 5925 + }, + { + "epoch": 0.38, + "learning_rate": 1.8749865819360658e-05, + "loss": 1.1889, + "step": 5926 + }, + { + "epoch": 0.38, + "learning_rate": 1.8749651130337707e-05, + "loss": 1.1978, + "step": 5927 + }, + { + "epoch": 0.38, + "learning_rate": 1.8749436441314756e-05, + "loss": 1.0339, + "step": 5928 + }, + { + "epoch": 0.38, + "learning_rate": 1.8749221752291805e-05, + "loss": 1.0333, + "step": 5929 + }, + { + "epoch": 0.38, + "learning_rate": 1.8749007063268857e-05, + "loss": 1.0884, + "step": 5930 + }, + { + "epoch": 0.38, + "learning_rate": 1.8748792374245906e-05, + "loss": 1.0988, + "step": 5931 + }, + { + "epoch": 0.38, + "learning_rate": 1.8748577685222955e-05, + "loss": 1.163, + "step": 5932 + }, + { + "epoch": 0.38, + "learning_rate": 1.8748362996200007e-05, + "loss": 1.1484, + "step": 5933 + }, + { + "epoch": 0.38, + "learning_rate": 1.8748148307177056e-05, + "loss": 1.1452, + "step": 5934 + }, + { + "epoch": 0.38, + "learning_rate": 1.8747933618154105e-05, + "loss": 1.1449, + "step": 5935 + }, + { + "epoch": 0.38, + "learning_rate": 1.8747718929131154e-05, + "loss": 1.0828, + "step": 5936 + }, + { + "epoch": 0.38, + "learning_rate": 1.8747504240108207e-05, + "loss": 1.1546, + "step": 5937 + }, + { + "epoch": 0.38, + "learning_rate": 1.8747289551085256e-05, + "loss": 1.0872, + "step": 5938 + }, + { + "epoch": 0.38, + "learning_rate": 1.8747074862062304e-05, + "loss": 1.0514, + "step": 5939 + }, + { + "epoch": 0.38, + "learning_rate": 1.8746860173039353e-05, + "loss": 1.1603, + "step": 5940 + }, + { + "epoch": 0.38, + "learning_rate": 1.8746645484016402e-05, + "loss": 1.1572, + "step": 5941 + }, + { + "epoch": 0.38, + "learning_rate": 1.8746430794993455e-05, + "loss": 1.0906, + "step": 5942 + }, + { + "epoch": 0.38, + "learning_rate": 1.8746216105970504e-05, + "loss": 1.1004, + "step": 5943 + }, + { + "epoch": 0.38, + "learning_rate": 1.8746001416947553e-05, + "loss": 1.1216, + "step": 5944 + }, + { + "epoch": 0.38, + "learning_rate": 1.87457867279246e-05, + "loss": 1.0458, + "step": 5945 + }, + { + "epoch": 0.38, + "learning_rate": 1.874557203890165e-05, + "loss": 1.0386, + "step": 5946 + }, + { + "epoch": 0.38, + "learning_rate": 1.87453573498787e-05, + "loss": 1.0261, + "step": 5947 + }, + { + "epoch": 0.38, + "learning_rate": 1.8745142660855752e-05, + "loss": 1.0556, + "step": 5948 + }, + { + "epoch": 0.38, + "learning_rate": 1.87449279718328e-05, + "loss": 1.0692, + "step": 5949 + }, + { + "epoch": 0.38, + "learning_rate": 1.8744713282809853e-05, + "loss": 1.1455, + "step": 5950 + }, + { + "epoch": 0.38, + "learning_rate": 1.8744498593786902e-05, + "loss": 1.0179, + "step": 5951 + }, + { + "epoch": 0.38, + "learning_rate": 1.874428390476395e-05, + "loss": 1.1109, + "step": 5952 + }, + { + "epoch": 0.38, + "learning_rate": 1.8744069215741e-05, + "loss": 1.1394, + "step": 5953 + }, + { + "epoch": 0.38, + "learning_rate": 1.874385452671805e-05, + "loss": 1.0801, + "step": 5954 + }, + { + "epoch": 0.38, + "learning_rate": 1.87436398376951e-05, + "loss": 1.06, + "step": 5955 + }, + { + "epoch": 0.38, + "learning_rate": 1.874342514867215e-05, + "loss": 1.0507, + "step": 5956 + }, + { + "epoch": 0.38, + "learning_rate": 1.87432104596492e-05, + "loss": 1.1397, + "step": 5957 + }, + { + "epoch": 0.38, + "learning_rate": 1.874299577062625e-05, + "loss": 1.0137, + "step": 5958 + }, + { + "epoch": 0.38, + "learning_rate": 1.8742781081603297e-05, + "loss": 1.0267, + "step": 5959 + }, + { + "epoch": 0.38, + "learning_rate": 1.874256639258035e-05, + "loss": 1.068, + "step": 5960 + }, + { + "epoch": 0.38, + "learning_rate": 1.87423517035574e-05, + "loss": 1.16, + "step": 5961 + }, + { + "epoch": 0.38, + "learning_rate": 1.874213701453445e-05, + "loss": 1.1746, + "step": 5962 + }, + { + "epoch": 0.38, + "learning_rate": 1.87419223255115e-05, + "loss": 1.038, + "step": 5963 + }, + { + "epoch": 0.38, + "learning_rate": 1.874170763648855e-05, + "loss": 1.1811, + "step": 5964 + }, + { + "epoch": 0.38, + "learning_rate": 1.8741492947465598e-05, + "loss": 1.0789, + "step": 5965 + }, + { + "epoch": 0.38, + "learning_rate": 1.8741278258442647e-05, + "loss": 1.1322, + "step": 5966 + }, + { + "epoch": 0.38, + "learning_rate": 1.8741063569419696e-05, + "loss": 1.1354, + "step": 5967 + }, + { + "epoch": 0.38, + "learning_rate": 1.8740848880396748e-05, + "loss": 1.0925, + "step": 5968 + }, + { + "epoch": 0.38, + "learning_rate": 1.8740634191373797e-05, + "loss": 1.213, + "step": 5969 + }, + { + "epoch": 0.38, + "learning_rate": 1.8740419502350846e-05, + "loss": 1.123, + "step": 5970 + }, + { + "epoch": 0.38, + "learning_rate": 1.8740204813327895e-05, + "loss": 1.1919, + "step": 5971 + }, + { + "epoch": 0.38, + "learning_rate": 1.8739990124304944e-05, + "loss": 1.1269, + "step": 5972 + }, + { + "epoch": 0.38, + "learning_rate": 1.8739775435281996e-05, + "loss": 1.0578, + "step": 5973 + }, + { + "epoch": 0.38, + "learning_rate": 1.8739560746259045e-05, + "loss": 1.0289, + "step": 5974 + }, + { + "epoch": 0.38, + "learning_rate": 1.8739346057236094e-05, + "loss": 0.9759, + "step": 5975 + }, + { + "epoch": 0.38, + "learning_rate": 1.8739131368213143e-05, + "loss": 1.0692, + "step": 5976 + }, + { + "epoch": 0.38, + "learning_rate": 1.8738916679190196e-05, + "loss": 1.1185, + "step": 5977 + }, + { + "epoch": 0.38, + "learning_rate": 1.8738701990167245e-05, + "loss": 1.1829, + "step": 5978 + }, + { + "epoch": 0.38, + "learning_rate": 1.8738487301144294e-05, + "loss": 1.1475, + "step": 5979 + }, + { + "epoch": 0.38, + "learning_rate": 1.8738272612121346e-05, + "loss": 1.0899, + "step": 5980 + }, + { + "epoch": 0.38, + "learning_rate": 1.8738057923098395e-05, + "loss": 1.1436, + "step": 5981 + }, + { + "epoch": 0.38, + "learning_rate": 1.8737843234075444e-05, + "loss": 1.121, + "step": 5982 + }, + { + "epoch": 0.38, + "learning_rate": 1.8737628545052493e-05, + "loss": 1.0841, + "step": 5983 + }, + { + "epoch": 0.38, + "learning_rate": 1.8737413856029542e-05, + "loss": 1.1148, + "step": 5984 + }, + { + "epoch": 0.39, + "learning_rate": 1.873719916700659e-05, + "loss": 1.0801, + "step": 5985 + }, + { + "epoch": 0.39, + "learning_rate": 1.8736984477983643e-05, + "loss": 1.1904, + "step": 5986 + }, + { + "epoch": 0.39, + "learning_rate": 1.8736769788960692e-05, + "loss": 1.0455, + "step": 5987 + }, + { + "epoch": 0.39, + "learning_rate": 1.873655509993774e-05, + "loss": 1.1155, + "step": 5988 + }, + { + "epoch": 0.39, + "learning_rate": 1.873634041091479e-05, + "loss": 1.0146, + "step": 5989 + }, + { + "epoch": 0.39, + "learning_rate": 1.873612572189184e-05, + "loss": 1.0915, + "step": 5990 + }, + { + "epoch": 0.39, + "learning_rate": 1.873591103286889e-05, + "loss": 1.0704, + "step": 5991 + }, + { + "epoch": 0.39, + "learning_rate": 1.873569634384594e-05, + "loss": 1.086, + "step": 5992 + }, + { + "epoch": 0.39, + "learning_rate": 1.8735481654822993e-05, + "loss": 1.1535, + "step": 5993 + }, + { + "epoch": 0.39, + "learning_rate": 1.873526696580004e-05, + "loss": 1.1026, + "step": 5994 + }, + { + "epoch": 0.39, + "learning_rate": 1.873505227677709e-05, + "loss": 1.1297, + "step": 5995 + }, + { + "epoch": 0.39, + "learning_rate": 1.873483758775414e-05, + "loss": 1.1422, + "step": 5996 + }, + { + "epoch": 0.39, + "learning_rate": 1.873462289873119e-05, + "loss": 1.0338, + "step": 5997 + }, + { + "epoch": 0.39, + "learning_rate": 1.8734408209708237e-05, + "loss": 1.0764, + "step": 5998 + }, + { + "epoch": 0.39, + "learning_rate": 1.873419352068529e-05, + "loss": 1.1562, + "step": 5999 + }, + { + "epoch": 0.39, + "learning_rate": 1.873397883166234e-05, + "loss": 1.1183, + "step": 6000 + }, + { + "epoch": 0.39, + "eval_laion/cider": 0.48334081504313, + "eval_laion/meteor": 0.38958639317303834, + "eval_laion/pred_num_words": 35.322265625, + "eval_laion/sacrebleu": 15.970916356582249, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.116838812828064, + "eval_runtime": 100.8793, + "eval_samples_per_second": 4.947, + "eval_steps_per_second": 0.317, + "step": 6000 + }, + { + "epoch": 0.39, + "learning_rate": 1.8733764142639388e-05, + "loss": 1.0709, + "step": 6001 + }, + { + "epoch": 0.39, + "learning_rate": 1.8733549453616437e-05, + "loss": 1.1083, + "step": 6002 + }, + { + "epoch": 0.39, + "learning_rate": 1.8733334764593486e-05, + "loss": 1.0979, + "step": 6003 + }, + { + "epoch": 0.39, + "learning_rate": 1.8733120075570538e-05, + "loss": 1.1479, + "step": 6004 + }, + { + "epoch": 0.39, + "learning_rate": 1.8732905386547587e-05, + "loss": 1.0401, + "step": 6005 + }, + { + "epoch": 0.39, + "learning_rate": 1.873269069752464e-05, + "loss": 1.1028, + "step": 6006 + }, + { + "epoch": 0.39, + "learning_rate": 1.873247600850169e-05, + "loss": 1.1482, + "step": 6007 + }, + { + "epoch": 0.39, + "learning_rate": 1.8732261319478737e-05, + "loss": 1.0905, + "step": 6008 + }, + { + "epoch": 0.39, + "learning_rate": 1.8732046630455786e-05, + "loss": 1.105, + "step": 6009 + }, + { + "epoch": 0.39, + "learning_rate": 1.8731831941432835e-05, + "loss": 1.0141, + "step": 6010 + }, + { + "epoch": 0.39, + "learning_rate": 1.8731617252409888e-05, + "loss": 1.081, + "step": 6011 + }, + { + "epoch": 0.39, + "learning_rate": 1.8731402563386937e-05, + "loss": 1.0765, + "step": 6012 + }, + { + "epoch": 0.39, + "learning_rate": 1.8731187874363985e-05, + "loss": 1.082, + "step": 6013 + }, + { + "epoch": 0.39, + "learning_rate": 1.8730973185341034e-05, + "loss": 1.0553, + "step": 6014 + }, + { + "epoch": 0.39, + "learning_rate": 1.8730758496318083e-05, + "loss": 1.0956, + "step": 6015 + }, + { + "epoch": 0.39, + "learning_rate": 1.8730543807295132e-05, + "loss": 1.1457, + "step": 6016 + }, + { + "epoch": 0.39, + "learning_rate": 1.8730329118272185e-05, + "loss": 1.0389, + "step": 6017 + }, + { + "epoch": 0.39, + "learning_rate": 1.8730329118272185e-05, + "loss": 1.1059, + "step": 6018 + }, + { + "epoch": 0.39, + "learning_rate": 1.8730114429249234e-05, + "loss": 1.1682, + "step": 6019 + }, + { + "epoch": 0.39, + "learning_rate": 1.8729899740226283e-05, + "loss": 1.039, + "step": 6020 + }, + { + "epoch": 0.39, + "learning_rate": 1.8729685051203335e-05, + "loss": 1.1511, + "step": 6021 + }, + { + "epoch": 0.39, + "learning_rate": 1.8729470362180384e-05, + "loss": 1.1958, + "step": 6022 + }, + { + "epoch": 0.39, + "learning_rate": 1.8729255673157433e-05, + "loss": 1.0131, + "step": 6023 + }, + { + "epoch": 0.39, + "learning_rate": 1.8729040984134482e-05, + "loss": 1.0053, + "step": 6024 + }, + { + "epoch": 0.39, + "learning_rate": 1.8728826295111534e-05, + "loss": 1.0868, + "step": 6025 + }, + { + "epoch": 0.39, + "learning_rate": 1.8728611606088583e-05, + "loss": 1.0367, + "step": 6026 + }, + { + "epoch": 0.39, + "learning_rate": 1.8728396917065632e-05, + "loss": 1.0989, + "step": 6027 + }, + { + "epoch": 0.39, + "learning_rate": 1.872818222804268e-05, + "loss": 1.1011, + "step": 6028 + }, + { + "epoch": 0.39, + "learning_rate": 1.872796753901973e-05, + "loss": 0.9781, + "step": 6029 + }, + { + "epoch": 0.39, + "learning_rate": 1.8727752849996782e-05, + "loss": 1.151, + "step": 6030 + }, + { + "epoch": 0.39, + "learning_rate": 1.872753816097383e-05, + "loss": 1.1484, + "step": 6031 + }, + { + "epoch": 0.39, + "learning_rate": 1.872732347195088e-05, + "loss": 1.0356, + "step": 6032 + }, + { + "epoch": 0.39, + "learning_rate": 1.872710878292793e-05, + "loss": 1.0072, + "step": 6033 + }, + { + "epoch": 0.39, + "learning_rate": 1.872689409390498e-05, + "loss": 1.0192, + "step": 6034 + }, + { + "epoch": 0.39, + "learning_rate": 1.8726679404882027e-05, + "loss": 1.2, + "step": 6035 + }, + { + "epoch": 0.39, + "learning_rate": 1.872646471585908e-05, + "loss": 1.0568, + "step": 6036 + }, + { + "epoch": 0.39, + "learning_rate": 1.872625002683613e-05, + "loss": 1.1136, + "step": 6037 + }, + { + "epoch": 0.39, + "learning_rate": 1.872603533781318e-05, + "loss": 1.1844, + "step": 6038 + }, + { + "epoch": 0.39, + "learning_rate": 1.872582064879023e-05, + "loss": 1.1592, + "step": 6039 + }, + { + "epoch": 0.39, + "learning_rate": 1.872560595976728e-05, + "loss": 1.0479, + "step": 6040 + }, + { + "epoch": 0.39, + "learning_rate": 1.8725391270744328e-05, + "loss": 0.9992, + "step": 6041 + }, + { + "epoch": 0.39, + "learning_rate": 1.8725176581721377e-05, + "loss": 1.1058, + "step": 6042 + }, + { + "epoch": 0.39, + "learning_rate": 1.872496189269843e-05, + "loss": 1.1385, + "step": 6043 + }, + { + "epoch": 0.39, + "learning_rate": 1.8724747203675478e-05, + "loss": 1.0265, + "step": 6044 + }, + { + "epoch": 0.39, + "learning_rate": 1.8724532514652527e-05, + "loss": 1.059, + "step": 6045 + }, + { + "epoch": 0.39, + "learning_rate": 1.8724317825629576e-05, + "loss": 1.0477, + "step": 6046 + }, + { + "epoch": 0.39, + "learning_rate": 1.8724103136606625e-05, + "loss": 1.1156, + "step": 6047 + }, + { + "epoch": 0.39, + "learning_rate": 1.8723888447583677e-05, + "loss": 1.1706, + "step": 6048 + }, + { + "epoch": 0.39, + "learning_rate": 1.8723673758560726e-05, + "loss": 1.1429, + "step": 6049 + }, + { + "epoch": 0.39, + "learning_rate": 1.872345906953778e-05, + "loss": 1.1343, + "step": 6050 + }, + { + "epoch": 0.39, + "learning_rate": 1.8723244380514828e-05, + "loss": 1.1506, + "step": 6051 + }, + { + "epoch": 0.39, + "learning_rate": 1.8723029691491877e-05, + "loss": 1.1125, + "step": 6052 + }, + { + "epoch": 0.39, + "learning_rate": 1.8722815002468926e-05, + "loss": 1.039, + "step": 6053 + }, + { + "epoch": 0.39, + "learning_rate": 1.8722600313445975e-05, + "loss": 1.0175, + "step": 6054 + }, + { + "epoch": 0.39, + "learning_rate": 1.8722385624423024e-05, + "loss": 1.1266, + "step": 6055 + }, + { + "epoch": 0.39, + "learning_rate": 1.8722170935400076e-05, + "loss": 1.1246, + "step": 6056 + }, + { + "epoch": 0.39, + "learning_rate": 1.8721956246377125e-05, + "loss": 1.0617, + "step": 6057 + }, + { + "epoch": 0.39, + "learning_rate": 1.8721741557354174e-05, + "loss": 1.1148, + "step": 6058 + }, + { + "epoch": 0.39, + "learning_rate": 1.8721526868331223e-05, + "loss": 1.1307, + "step": 6059 + }, + { + "epoch": 0.39, + "learning_rate": 1.8721312179308272e-05, + "loss": 1.1468, + "step": 6060 + }, + { + "epoch": 0.39, + "learning_rate": 1.8721097490285324e-05, + "loss": 1.0924, + "step": 6061 + }, + { + "epoch": 0.39, + "learning_rate": 1.8720882801262373e-05, + "loss": 1.0893, + "step": 6062 + }, + { + "epoch": 0.39, + "learning_rate": 1.8720668112239422e-05, + "loss": 1.1982, + "step": 6063 + }, + { + "epoch": 0.39, + "learning_rate": 1.872045342321647e-05, + "loss": 1.0877, + "step": 6064 + }, + { + "epoch": 0.39, + "learning_rate": 1.8720238734193523e-05, + "loss": 1.0404, + "step": 6065 + }, + { + "epoch": 0.39, + "learning_rate": 1.8720024045170572e-05, + "loss": 1.1037, + "step": 6066 + }, + { + "epoch": 0.39, + "learning_rate": 1.871980935614762e-05, + "loss": 1.0296, + "step": 6067 + }, + { + "epoch": 0.39, + "learning_rate": 1.8719594667124674e-05, + "loss": 1.0095, + "step": 6068 + }, + { + "epoch": 0.39, + "learning_rate": 1.8719379978101723e-05, + "loss": 1.1589, + "step": 6069 + }, + { + "epoch": 0.39, + "learning_rate": 1.871916528907877e-05, + "loss": 1.1343, + "step": 6070 + }, + { + "epoch": 0.39, + "learning_rate": 1.871895060005582e-05, + "loss": 1.1064, + "step": 6071 + }, + { + "epoch": 0.39, + "learning_rate": 1.871873591103287e-05, + "loss": 1.0663, + "step": 6072 + }, + { + "epoch": 0.39, + "learning_rate": 1.871852122200992e-05, + "loss": 1.0924, + "step": 6073 + }, + { + "epoch": 0.39, + "learning_rate": 1.871830653298697e-05, + "loss": 1.0463, + "step": 6074 + }, + { + "epoch": 0.39, + "learning_rate": 1.871809184396402e-05, + "loss": 1.069, + "step": 6075 + }, + { + "epoch": 0.39, + "learning_rate": 1.871787715494107e-05, + "loss": 1.0937, + "step": 6076 + }, + { + "epoch": 0.39, + "learning_rate": 1.8717662465918118e-05, + "loss": 1.1505, + "step": 6077 + }, + { + "epoch": 0.39, + "learning_rate": 1.8717447776895167e-05, + "loss": 1.0438, + "step": 6078 + }, + { + "epoch": 0.39, + "learning_rate": 1.871723308787222e-05, + "loss": 1.0968, + "step": 6079 + }, + { + "epoch": 0.39, + "learning_rate": 1.8717018398849268e-05, + "loss": 1.0788, + "step": 6080 + }, + { + "epoch": 0.39, + "learning_rate": 1.871680370982632e-05, + "loss": 1.1307, + "step": 6081 + }, + { + "epoch": 0.39, + "learning_rate": 1.871658902080337e-05, + "loss": 1.1478, + "step": 6082 + }, + { + "epoch": 0.39, + "learning_rate": 1.8716374331780418e-05, + "loss": 1.0533, + "step": 6083 + }, + { + "epoch": 0.39, + "learning_rate": 1.8716159642757467e-05, + "loss": 1.0303, + "step": 6084 + }, + { + "epoch": 0.39, + "learning_rate": 1.8715944953734516e-05, + "loss": 1.1162, + "step": 6085 + }, + { + "epoch": 0.39, + "learning_rate": 1.8715730264711565e-05, + "loss": 1.1043, + "step": 6086 + }, + { + "epoch": 0.39, + "learning_rate": 1.8715515575688618e-05, + "loss": 1.0785, + "step": 6087 + }, + { + "epoch": 0.39, + "learning_rate": 1.8715300886665666e-05, + "loss": 1.1102, + "step": 6088 + }, + { + "epoch": 0.39, + "learning_rate": 1.8715086197642715e-05, + "loss": 1.0994, + "step": 6089 + }, + { + "epoch": 0.39, + "learning_rate": 1.8714871508619764e-05, + "loss": 1.1089, + "step": 6090 + }, + { + "epoch": 0.39, + "learning_rate": 1.8714656819596813e-05, + "loss": 1.1484, + "step": 6091 + }, + { + "epoch": 0.39, + "learning_rate": 1.8714442130573866e-05, + "loss": 1.1046, + "step": 6092 + }, + { + "epoch": 0.39, + "learning_rate": 1.8714227441550915e-05, + "loss": 1.0806, + "step": 6093 + }, + { + "epoch": 0.39, + "learning_rate": 1.8714012752527967e-05, + "loss": 1.1127, + "step": 6094 + }, + { + "epoch": 0.39, + "learning_rate": 1.8713798063505016e-05, + "loss": 1.0706, + "step": 6095 + }, + { + "epoch": 0.39, + "learning_rate": 1.8713583374482065e-05, + "loss": 1.1493, + "step": 6096 + }, + { + "epoch": 0.39, + "learning_rate": 1.8713368685459114e-05, + "loss": 1.0435, + "step": 6097 + }, + { + "epoch": 0.39, + "learning_rate": 1.8713153996436163e-05, + "loss": 1.1107, + "step": 6098 + }, + { + "epoch": 0.39, + "learning_rate": 1.8712939307413215e-05, + "loss": 1.0991, + "step": 6099 + }, + { + "epoch": 0.39, + "learning_rate": 1.8712724618390264e-05, + "loss": 1.0784, + "step": 6100 + }, + { + "epoch": 0.39, + "learning_rate": 1.8712509929367313e-05, + "loss": 1.0823, + "step": 6101 + }, + { + "epoch": 0.39, + "learning_rate": 1.8712295240344362e-05, + "loss": 1.037, + "step": 6102 + }, + { + "epoch": 0.39, + "learning_rate": 1.871208055132141e-05, + "loss": 1.1342, + "step": 6103 + }, + { + "epoch": 0.39, + "learning_rate": 1.871186586229846e-05, + "loss": 1.042, + "step": 6104 + }, + { + "epoch": 0.39, + "learning_rate": 1.8711651173275512e-05, + "loss": 1.1244, + "step": 6105 + }, + { + "epoch": 0.39, + "learning_rate": 1.871143648425256e-05, + "loss": 1.0856, + "step": 6106 + }, + { + "epoch": 0.39, + "learning_rate": 1.871122179522961e-05, + "loss": 1.0782, + "step": 6107 + }, + { + "epoch": 0.39, + "learning_rate": 1.8711007106206663e-05, + "loss": 1.1104, + "step": 6108 + }, + { + "epoch": 0.39, + "learning_rate": 1.871079241718371e-05, + "loss": 1.0895, + "step": 6109 + }, + { + "epoch": 0.39, + "learning_rate": 1.871057772816076e-05, + "loss": 1.0708, + "step": 6110 + }, + { + "epoch": 0.39, + "learning_rate": 1.871036303913781e-05, + "loss": 1.0385, + "step": 6111 + }, + { + "epoch": 0.39, + "learning_rate": 1.8710148350114862e-05, + "loss": 1.2052, + "step": 6112 + }, + { + "epoch": 0.39, + "learning_rate": 1.870993366109191e-05, + "loss": 1.1054, + "step": 6113 + }, + { + "epoch": 0.39, + "learning_rate": 1.870971897206896e-05, + "loss": 1.0545, + "step": 6114 + }, + { + "epoch": 0.39, + "learning_rate": 1.870950428304601e-05, + "loss": 1.2041, + "step": 6115 + }, + { + "epoch": 0.39, + "learning_rate": 1.8709289594023058e-05, + "loss": 0.9936, + "step": 6116 + }, + { + "epoch": 0.39, + "learning_rate": 1.870907490500011e-05, + "loss": 1.1145, + "step": 6117 + }, + { + "epoch": 0.39, + "learning_rate": 1.870886021597716e-05, + "loss": 1.0948, + "step": 6118 + }, + { + "epoch": 0.39, + "learning_rate": 1.8708645526954208e-05, + "loss": 1.0453, + "step": 6119 + }, + { + "epoch": 0.39, + "learning_rate": 1.8708430837931257e-05, + "loss": 1.176, + "step": 6120 + }, + { + "epoch": 0.39, + "learning_rate": 1.8708216148908306e-05, + "loss": 1.1159, + "step": 6121 + }, + { + "epoch": 0.39, + "learning_rate": 1.8708001459885355e-05, + "loss": 1.1724, + "step": 6122 + }, + { + "epoch": 0.39, + "learning_rate": 1.8707786770862407e-05, + "loss": 1.11, + "step": 6123 + }, + { + "epoch": 0.39, + "learning_rate": 1.8707572081839456e-05, + "loss": 1.0779, + "step": 6124 + }, + { + "epoch": 0.39, + "learning_rate": 1.870735739281651e-05, + "loss": 1.0779, + "step": 6125 + }, + { + "epoch": 0.39, + "learning_rate": 1.8707142703793558e-05, + "loss": 1.0726, + "step": 6126 + }, + { + "epoch": 0.39, + "learning_rate": 1.8706928014770607e-05, + "loss": 1.0343, + "step": 6127 + }, + { + "epoch": 0.39, + "learning_rate": 1.8706713325747656e-05, + "loss": 1.0882, + "step": 6128 + }, + { + "epoch": 0.39, + "learning_rate": 1.8706498636724705e-05, + "loss": 1.0302, + "step": 6129 + }, + { + "epoch": 0.39, + "learning_rate": 1.8706283947701757e-05, + "loss": 1.1125, + "step": 6130 + }, + { + "epoch": 0.39, + "learning_rate": 1.8706069258678806e-05, + "loss": 1.0993, + "step": 6131 + }, + { + "epoch": 0.39, + "learning_rate": 1.8705854569655855e-05, + "loss": 1.0827, + "step": 6132 + }, + { + "epoch": 0.39, + "learning_rate": 1.8705639880632904e-05, + "loss": 1.0697, + "step": 6133 + }, + { + "epoch": 0.39, + "learning_rate": 1.8705425191609953e-05, + "loss": 1.1265, + "step": 6134 + }, + { + "epoch": 0.39, + "learning_rate": 1.8705210502587005e-05, + "loss": 1.2116, + "step": 6135 + }, + { + "epoch": 0.39, + "learning_rate": 1.8704995813564054e-05, + "loss": 1.0783, + "step": 6136 + }, + { + "epoch": 0.39, + "learning_rate": 1.8704781124541106e-05, + "loss": 1.186, + "step": 6137 + }, + { + "epoch": 0.39, + "learning_rate": 1.8704566435518155e-05, + "loss": 1.0588, + "step": 6138 + }, + { + "epoch": 0.39, + "learning_rate": 1.8704351746495204e-05, + "loss": 1.0498, + "step": 6139 + }, + { + "epoch": 0.4, + "learning_rate": 1.8704137057472253e-05, + "loss": 1.0692, + "step": 6140 + }, + { + "epoch": 0.4, + "learning_rate": 1.8703922368449302e-05, + "loss": 1.1288, + "step": 6141 + }, + { + "epoch": 0.4, + "learning_rate": 1.870370767942635e-05, + "loss": 1.1583, + "step": 6142 + }, + { + "epoch": 0.4, + "learning_rate": 1.8703492990403404e-05, + "loss": 1.0584, + "step": 6143 + }, + { + "epoch": 0.4, + "learning_rate": 1.8703278301380453e-05, + "loss": 1.1381, + "step": 6144 + }, + { + "epoch": 0.4, + "learning_rate": 1.87030636123575e-05, + "loss": 1.127, + "step": 6145 + }, + { + "epoch": 0.4, + "learning_rate": 1.870284892333455e-05, + "loss": 1.0977, + "step": 6146 + }, + { + "epoch": 0.4, + "learning_rate": 1.87026342343116e-05, + "loss": 1.0692, + "step": 6147 + }, + { + "epoch": 0.4, + "learning_rate": 1.8702419545288652e-05, + "loss": 1.1053, + "step": 6148 + }, + { + "epoch": 0.4, + "learning_rate": 1.87022048562657e-05, + "loss": 1.0537, + "step": 6149 + }, + { + "epoch": 0.4, + "learning_rate": 1.870199016724275e-05, + "loss": 1.1169, + "step": 6150 + }, + { + "epoch": 0.4, + "learning_rate": 1.87017754782198e-05, + "loss": 1.0887, + "step": 6151 + }, + { + "epoch": 0.4, + "learning_rate": 1.870156078919685e-05, + "loss": 1.092, + "step": 6152 + }, + { + "epoch": 0.4, + "learning_rate": 1.87013461001739e-05, + "loss": 1.1248, + "step": 6153 + }, + { + "epoch": 0.4, + "learning_rate": 1.870113141115095e-05, + "loss": 1.1422, + "step": 6154 + }, + { + "epoch": 0.4, + "learning_rate": 1.8700916722128e-05, + "loss": 1.1812, + "step": 6155 + }, + { + "epoch": 0.4, + "learning_rate": 1.870070203310505e-05, + "loss": 1.0695, + "step": 6156 + }, + { + "epoch": 0.4, + "learning_rate": 1.87004873440821e-05, + "loss": 1.053, + "step": 6157 + }, + { + "epoch": 0.4, + "learning_rate": 1.8700272655059148e-05, + "loss": 1.0811, + "step": 6158 + }, + { + "epoch": 0.4, + "learning_rate": 1.8700057966036197e-05, + "loss": 1.0954, + "step": 6159 + }, + { + "epoch": 0.4, + "learning_rate": 1.8699843277013246e-05, + "loss": 1.1389, + "step": 6160 + }, + { + "epoch": 0.4, + "learning_rate": 1.86996285879903e-05, + "loss": 1.091, + "step": 6161 + }, + { + "epoch": 0.4, + "learning_rate": 1.8699413898967347e-05, + "loss": 1.1503, + "step": 6162 + }, + { + "epoch": 0.4, + "learning_rate": 1.8699199209944396e-05, + "loss": 1.0847, + "step": 6163 + }, + { + "epoch": 0.4, + "learning_rate": 1.8698984520921445e-05, + "loss": 1.0867, + "step": 6164 + }, + { + "epoch": 0.4, + "learning_rate": 1.8698769831898494e-05, + "loss": 0.9664, + "step": 6165 + }, + { + "epoch": 0.4, + "learning_rate": 1.8698555142875547e-05, + "loss": 1.0875, + "step": 6166 + }, + { + "epoch": 0.4, + "learning_rate": 1.8698340453852596e-05, + "loss": 1.1392, + "step": 6167 + }, + { + "epoch": 0.4, + "learning_rate": 1.8698125764829648e-05, + "loss": 1.0999, + "step": 6168 + }, + { + "epoch": 0.4, + "learning_rate": 1.8697911075806697e-05, + "loss": 1.1771, + "step": 6169 + }, + { + "epoch": 0.4, + "learning_rate": 1.8697696386783746e-05, + "loss": 1.0942, + "step": 6170 + }, + { + "epoch": 0.4, + "learning_rate": 1.8697481697760795e-05, + "loss": 1.144, + "step": 6171 + }, + { + "epoch": 0.4, + "learning_rate": 1.8697267008737844e-05, + "loss": 1.1122, + "step": 6172 + }, + { + "epoch": 0.4, + "learning_rate": 1.8697052319714893e-05, + "loss": 1.1587, + "step": 6173 + }, + { + "epoch": 0.4, + "learning_rate": 1.8696837630691945e-05, + "loss": 1.1077, + "step": 6174 + }, + { + "epoch": 0.4, + "learning_rate": 1.8696622941668994e-05, + "loss": 1.0834, + "step": 6175 + }, + { + "epoch": 0.4, + "learning_rate": 1.8696408252646043e-05, + "loss": 1.0094, + "step": 6176 + }, + { + "epoch": 0.4, + "learning_rate": 1.8696193563623092e-05, + "loss": 1.1075, + "step": 6177 + }, + { + "epoch": 0.4, + "learning_rate": 1.869597887460014e-05, + "loss": 1.0968, + "step": 6178 + }, + { + "epoch": 0.4, + "learning_rate": 1.8695764185577193e-05, + "loss": 1.1199, + "step": 6179 + }, + { + "epoch": 0.4, + "learning_rate": 1.8695549496554242e-05, + "loss": 1.1518, + "step": 6180 + }, + { + "epoch": 0.4, + "learning_rate": 1.8695334807531295e-05, + "loss": 1.172, + "step": 6181 + }, + { + "epoch": 0.4, + "learning_rate": 1.8695120118508344e-05, + "loss": 1.0909, + "step": 6182 + }, + { + "epoch": 0.4, + "learning_rate": 1.8694905429485393e-05, + "loss": 1.071, + "step": 6183 + }, + { + "epoch": 0.4, + "learning_rate": 1.869469074046244e-05, + "loss": 1.1105, + "step": 6184 + }, + { + "epoch": 0.4, + "learning_rate": 1.869447605143949e-05, + "loss": 1.2067, + "step": 6185 + }, + { + "epoch": 0.4, + "learning_rate": 1.8694261362416543e-05, + "loss": 1.0822, + "step": 6186 + }, + { + "epoch": 0.4, + "learning_rate": 1.8694046673393592e-05, + "loss": 1.1048, + "step": 6187 + }, + { + "epoch": 0.4, + "learning_rate": 1.869383198437064e-05, + "loss": 1.0914, + "step": 6188 + }, + { + "epoch": 0.4, + "learning_rate": 1.869361729534769e-05, + "loss": 1.0632, + "step": 6189 + }, + { + "epoch": 0.4, + "learning_rate": 1.869340260632474e-05, + "loss": 1.1183, + "step": 6190 + }, + { + "epoch": 0.4, + "learning_rate": 1.8693187917301788e-05, + "loss": 0.9963, + "step": 6191 + }, + { + "epoch": 0.4, + "learning_rate": 1.869297322827884e-05, + "loss": 1.0725, + "step": 6192 + }, + { + "epoch": 0.4, + "learning_rate": 1.869275853925589e-05, + "loss": 1.0972, + "step": 6193 + }, + { + "epoch": 0.4, + "learning_rate": 1.8692543850232938e-05, + "loss": 1.1027, + "step": 6194 + }, + { + "epoch": 0.4, + "learning_rate": 1.869232916120999e-05, + "loss": 1.0797, + "step": 6195 + }, + { + "epoch": 0.4, + "learning_rate": 1.869211447218704e-05, + "loss": 1.1192, + "step": 6196 + }, + { + "epoch": 0.4, + "learning_rate": 1.869189978316409e-05, + "loss": 1.1026, + "step": 6197 + }, + { + "epoch": 0.4, + "learning_rate": 1.8691685094141137e-05, + "loss": 1.11, + "step": 6198 + }, + { + "epoch": 0.4, + "learning_rate": 1.869147040511819e-05, + "loss": 1.0106, + "step": 6199 + }, + { + "epoch": 0.4, + "learning_rate": 1.869125571609524e-05, + "loss": 1.1195, + "step": 6200 + }, + { + "epoch": 0.4, + "learning_rate": 1.8691041027072288e-05, + "loss": 1.1814, + "step": 6201 + }, + { + "epoch": 0.4, + "learning_rate": 1.8690826338049337e-05, + "loss": 1.1002, + "step": 6202 + }, + { + "epoch": 0.4, + "learning_rate": 1.8690611649026386e-05, + "loss": 1.1658, + "step": 6203 + }, + { + "epoch": 0.4, + "learning_rate": 1.8690396960003438e-05, + "loss": 1.1509, + "step": 6204 + }, + { + "epoch": 0.4, + "learning_rate": 1.8690182270980487e-05, + "loss": 0.991, + "step": 6205 + }, + { + "epoch": 0.4, + "learning_rate": 1.8689967581957536e-05, + "loss": 1.1366, + "step": 6206 + }, + { + "epoch": 0.4, + "learning_rate": 1.8689752892934585e-05, + "loss": 1.0846, + "step": 6207 + }, + { + "epoch": 0.4, + "learning_rate": 1.8689538203911634e-05, + "loss": 1.0642, + "step": 6208 + }, + { + "epoch": 0.4, + "learning_rate": 1.8689323514888683e-05, + "loss": 1.1036, + "step": 6209 + }, + { + "epoch": 0.4, + "learning_rate": 1.8689108825865735e-05, + "loss": 1.1636, + "step": 6210 + }, + { + "epoch": 0.4, + "learning_rate": 1.8688894136842784e-05, + "loss": 1.1171, + "step": 6211 + }, + { + "epoch": 0.4, + "learning_rate": 1.8688679447819836e-05, + "loss": 1.1217, + "step": 6212 + }, + { + "epoch": 0.4, + "learning_rate": 1.8688464758796885e-05, + "loss": 1.1053, + "step": 6213 + }, + { + "epoch": 0.4, + "learning_rate": 1.8688250069773934e-05, + "loss": 1.1379, + "step": 6214 + }, + { + "epoch": 0.4, + "learning_rate": 1.8688035380750983e-05, + "loss": 1.0871, + "step": 6215 + }, + { + "epoch": 0.4, + "learning_rate": 1.8687820691728032e-05, + "loss": 1.0311, + "step": 6216 + }, + { + "epoch": 0.4, + "learning_rate": 1.8687606002705085e-05, + "loss": 1.1213, + "step": 6217 + }, + { + "epoch": 0.4, + "learning_rate": 1.8687391313682134e-05, + "loss": 1.1566, + "step": 6218 + }, + { + "epoch": 0.4, + "learning_rate": 1.8687176624659183e-05, + "loss": 1.1173, + "step": 6219 + }, + { + "epoch": 0.4, + "learning_rate": 1.868696193563623e-05, + "loss": 1.1845, + "step": 6220 + }, + { + "epoch": 0.4, + "learning_rate": 1.868674724661328e-05, + "loss": 1.0908, + "step": 6221 + }, + { + "epoch": 0.4, + "learning_rate": 1.8686532557590333e-05, + "loss": 0.9696, + "step": 6222 + }, + { + "epoch": 0.4, + "learning_rate": 1.8686317868567382e-05, + "loss": 1.1072, + "step": 6223 + }, + { + "epoch": 0.4, + "learning_rate": 1.868610317954443e-05, + "loss": 1.1167, + "step": 6224 + }, + { + "epoch": 0.4, + "learning_rate": 1.8685888490521483e-05, + "loss": 1.1178, + "step": 6225 + }, + { + "epoch": 0.4, + "learning_rate": 1.8685673801498532e-05, + "loss": 1.0417, + "step": 6226 + }, + { + "epoch": 0.4, + "learning_rate": 1.868545911247558e-05, + "loss": 1.1112, + "step": 6227 + }, + { + "epoch": 0.4, + "learning_rate": 1.868524442345263e-05, + "loss": 1.0362, + "step": 6228 + }, + { + "epoch": 0.4, + "learning_rate": 1.868502973442968e-05, + "loss": 0.9964, + "step": 6229 + }, + { + "epoch": 0.4, + "learning_rate": 1.868481504540673e-05, + "loss": 1.1033, + "step": 6230 + }, + { + "epoch": 0.4, + "learning_rate": 1.868460035638378e-05, + "loss": 1.0969, + "step": 6231 + }, + { + "epoch": 0.4, + "learning_rate": 1.868438566736083e-05, + "loss": 1.1335, + "step": 6232 + }, + { + "epoch": 0.4, + "learning_rate": 1.8684170978337878e-05, + "loss": 1.0587, + "step": 6233 + }, + { + "epoch": 0.4, + "learning_rate": 1.8683956289314927e-05, + "loss": 1.1277, + "step": 6234 + }, + { + "epoch": 0.4, + "learning_rate": 1.868374160029198e-05, + "loss": 1.1664, + "step": 6235 + }, + { + "epoch": 0.4, + "learning_rate": 1.868352691126903e-05, + "loss": 1.1072, + "step": 6236 + }, + { + "epoch": 0.4, + "learning_rate": 1.8683312222246077e-05, + "loss": 1.071, + "step": 6237 + }, + { + "epoch": 0.4, + "learning_rate": 1.8683097533223126e-05, + "loss": 1.0952, + "step": 6238 + }, + { + "epoch": 0.4, + "learning_rate": 1.868288284420018e-05, + "loss": 1.1445, + "step": 6239 + }, + { + "epoch": 0.4, + "learning_rate": 1.8682668155177228e-05, + "loss": 1.0668, + "step": 6240 + }, + { + "epoch": 0.4, + "learning_rate": 1.8682453466154277e-05, + "loss": 1.0985, + "step": 6241 + }, + { + "epoch": 0.4, + "learning_rate": 1.868223877713133e-05, + "loss": 1.1366, + "step": 6242 + }, + { + "epoch": 0.4, + "learning_rate": 1.8682024088108378e-05, + "loss": 1.0944, + "step": 6243 + }, + { + "epoch": 0.4, + "learning_rate": 1.8681809399085427e-05, + "loss": 1.1454, + "step": 6244 + }, + { + "epoch": 0.4, + "learning_rate": 1.8681594710062476e-05, + "loss": 1.1018, + "step": 6245 + }, + { + "epoch": 0.4, + "learning_rate": 1.8681380021039525e-05, + "loss": 1.1007, + "step": 6246 + }, + { + "epoch": 0.4, + "learning_rate": 1.8681165332016574e-05, + "loss": 1.1094, + "step": 6247 + }, + { + "epoch": 0.4, + "learning_rate": 1.8680950642993626e-05, + "loss": 1.1108, + "step": 6248 + }, + { + "epoch": 0.4, + "learning_rate": 1.8680735953970675e-05, + "loss": 1.0824, + "step": 6249 + }, + { + "epoch": 0.4, + "learning_rate": 1.8680521264947724e-05, + "loss": 1.0262, + "step": 6250 + }, + { + "epoch": 0.4, + "learning_rate": 1.8680306575924773e-05, + "loss": 1.0778, + "step": 6251 + }, + { + "epoch": 0.4, + "learning_rate": 1.8680091886901822e-05, + "loss": 1.1187, + "step": 6252 + }, + { + "epoch": 0.4, + "learning_rate": 1.8679877197878874e-05, + "loss": 1.0433, + "step": 6253 + }, + { + "epoch": 0.4, + "learning_rate": 1.8679662508855923e-05, + "loss": 1.1082, + "step": 6254 + }, + { + "epoch": 0.4, + "learning_rate": 1.8679447819832976e-05, + "loss": 1.1294, + "step": 6255 + }, + { + "epoch": 0.4, + "learning_rate": 1.8679233130810025e-05, + "loss": 1.1005, + "step": 6256 + }, + { + "epoch": 0.4, + "learning_rate": 1.8679018441787074e-05, + "loss": 1.0604, + "step": 6257 + }, + { + "epoch": 0.4, + "learning_rate": 1.8678803752764123e-05, + "loss": 1.1882, + "step": 6258 + }, + { + "epoch": 0.4, + "learning_rate": 1.867858906374117e-05, + "loss": 1.1047, + "step": 6259 + }, + { + "epoch": 0.4, + "learning_rate": 1.867837437471822e-05, + "loss": 1.1463, + "step": 6260 + }, + { + "epoch": 0.4, + "learning_rate": 1.8678159685695273e-05, + "loss": 1.0535, + "step": 6261 + }, + { + "epoch": 0.4, + "learning_rate": 1.8677944996672322e-05, + "loss": 1.0885, + "step": 6262 + }, + { + "epoch": 0.4, + "learning_rate": 1.867773030764937e-05, + "loss": 1.1403, + "step": 6263 + }, + { + "epoch": 0.4, + "learning_rate": 1.867751561862642e-05, + "loss": 1.1374, + "step": 6264 + }, + { + "epoch": 0.4, + "learning_rate": 1.867730092960347e-05, + "loss": 1.0815, + "step": 6265 + }, + { + "epoch": 0.4, + "learning_rate": 1.867708624058052e-05, + "loss": 1.1155, + "step": 6266 + }, + { + "epoch": 0.4, + "learning_rate": 1.867687155155757e-05, + "loss": 1.0836, + "step": 6267 + }, + { + "epoch": 0.4, + "learning_rate": 1.8676656862534622e-05, + "loss": 1.0906, + "step": 6268 + }, + { + "epoch": 0.4, + "learning_rate": 1.867644217351167e-05, + "loss": 1.0529, + "step": 6269 + }, + { + "epoch": 0.4, + "learning_rate": 1.867622748448872e-05, + "loss": 1.1316, + "step": 6270 + }, + { + "epoch": 0.4, + "learning_rate": 1.867601279546577e-05, + "loss": 1.0584, + "step": 6271 + }, + { + "epoch": 0.4, + "learning_rate": 1.867579810644282e-05, + "loss": 1.0286, + "step": 6272 + }, + { + "epoch": 0.4, + "learning_rate": 1.867558341741987e-05, + "loss": 1.1609, + "step": 6273 + }, + { + "epoch": 0.4, + "learning_rate": 1.867536872839692e-05, + "loss": 1.1896, + "step": 6274 + }, + { + "epoch": 0.4, + "learning_rate": 1.867515403937397e-05, + "loss": 1.0305, + "step": 6275 + }, + { + "epoch": 0.4, + "learning_rate": 1.8674939350351018e-05, + "loss": 1.1484, + "step": 6276 + }, + { + "epoch": 0.4, + "learning_rate": 1.8674724661328067e-05, + "loss": 1.066, + "step": 6277 + }, + { + "epoch": 0.4, + "learning_rate": 1.8674509972305116e-05, + "loss": 1.0453, + "step": 6278 + }, + { + "epoch": 0.4, + "learning_rate": 1.8674295283282168e-05, + "loss": 1.0735, + "step": 6279 + }, + { + "epoch": 0.4, + "learning_rate": 1.8674080594259217e-05, + "loss": 1.0949, + "step": 6280 + }, + { + "epoch": 0.4, + "learning_rate": 1.8673865905236266e-05, + "loss": 1.0912, + "step": 6281 + }, + { + "epoch": 0.4, + "learning_rate": 1.8673651216213315e-05, + "loss": 1.078, + "step": 6282 + }, + { + "epoch": 0.4, + "learning_rate": 1.8673436527190367e-05, + "loss": 1.075, + "step": 6283 + }, + { + "epoch": 0.4, + "learning_rate": 1.8673221838167416e-05, + "loss": 1.0998, + "step": 6284 + }, + { + "epoch": 0.4, + "learning_rate": 1.8673007149144465e-05, + "loss": 1.1269, + "step": 6285 + }, + { + "epoch": 0.4, + "learning_rate": 1.8672792460121517e-05, + "loss": 1.0405, + "step": 6286 + }, + { + "epoch": 0.4, + "learning_rate": 1.8672577771098566e-05, + "loss": 1.081, + "step": 6287 + }, + { + "epoch": 0.4, + "learning_rate": 1.8672363082075615e-05, + "loss": 1.2024, + "step": 6288 + }, + { + "epoch": 0.4, + "learning_rate": 1.8672148393052664e-05, + "loss": 1.118, + "step": 6289 + }, + { + "epoch": 0.4, + "learning_rate": 1.8671933704029713e-05, + "loss": 1.1263, + "step": 6290 + }, + { + "epoch": 0.4, + "learning_rate": 1.8671719015006766e-05, + "loss": 1.1008, + "step": 6291 + }, + { + "epoch": 0.4, + "learning_rate": 1.8671504325983815e-05, + "loss": 1.1379, + "step": 6292 + }, + { + "epoch": 0.4, + "learning_rate": 1.8671289636960864e-05, + "loss": 1.1564, + "step": 6293 + }, + { + "epoch": 0.4, + "learning_rate": 1.8671074947937913e-05, + "loss": 1.0638, + "step": 6294 + }, + { + "epoch": 0.41, + "learning_rate": 1.867086025891496e-05, + "loss": 1.0258, + "step": 6295 + }, + { + "epoch": 0.41, + "learning_rate": 1.867064556989201e-05, + "loss": 1.0877, + "step": 6296 + }, + { + "epoch": 0.41, + "learning_rate": 1.8670430880869063e-05, + "loss": 1.0251, + "step": 6297 + }, + { + "epoch": 0.41, + "learning_rate": 1.8670216191846112e-05, + "loss": 1.2101, + "step": 6298 + }, + { + "epoch": 0.41, + "learning_rate": 1.8670001502823164e-05, + "loss": 1.1194, + "step": 6299 + }, + { + "epoch": 0.41, + "learning_rate": 1.8669786813800213e-05, + "loss": 1.1973, + "step": 6300 + }, + { + "epoch": 0.41, + "learning_rate": 1.8669572124777262e-05, + "loss": 1.0723, + "step": 6301 + }, + { + "epoch": 0.41, + "learning_rate": 1.866935743575431e-05, + "loss": 1.0815, + "step": 6302 + }, + { + "epoch": 0.41, + "learning_rate": 1.866914274673136e-05, + "loss": 1.0766, + "step": 6303 + }, + { + "epoch": 0.41, + "learning_rate": 1.8668928057708412e-05, + "loss": 1.1685, + "step": 6304 + }, + { + "epoch": 0.41, + "learning_rate": 1.866871336868546e-05, + "loss": 1.0606, + "step": 6305 + }, + { + "epoch": 0.41, + "learning_rate": 1.866849867966251e-05, + "loss": 1.0276, + "step": 6306 + }, + { + "epoch": 0.41, + "learning_rate": 1.866828399063956e-05, + "loss": 1.0411, + "step": 6307 + }, + { + "epoch": 0.41, + "learning_rate": 1.8668069301616608e-05, + "loss": 1.1533, + "step": 6308 + }, + { + "epoch": 0.41, + "learning_rate": 1.8667854612593657e-05, + "loss": 1.1393, + "step": 6309 + }, + { + "epoch": 0.41, + "learning_rate": 1.866763992357071e-05, + "loss": 0.9712, + "step": 6310 + }, + { + "epoch": 0.41, + "learning_rate": 1.866742523454776e-05, + "loss": 1.1749, + "step": 6311 + }, + { + "epoch": 0.41, + "learning_rate": 1.866721054552481e-05, + "loss": 1.1084, + "step": 6312 + }, + { + "epoch": 0.41, + "learning_rate": 1.866699585650186e-05, + "loss": 1.0824, + "step": 6313 + }, + { + "epoch": 0.41, + "learning_rate": 1.866678116747891e-05, + "loss": 1.0505, + "step": 6314 + }, + { + "epoch": 0.41, + "learning_rate": 1.8666566478455958e-05, + "loss": 1.1828, + "step": 6315 + }, + { + "epoch": 0.41, + "learning_rate": 1.8666351789433007e-05, + "loss": 1.0798, + "step": 6316 + }, + { + "epoch": 0.41, + "learning_rate": 1.866613710041006e-05, + "loss": 1.0414, + "step": 6317 + }, + { + "epoch": 0.41, + "learning_rate": 1.8665922411387108e-05, + "loss": 1.0785, + "step": 6318 + }, + { + "epoch": 0.41, + "learning_rate": 1.8665707722364157e-05, + "loss": 1.0934, + "step": 6319 + }, + { + "epoch": 0.41, + "learning_rate": 1.8665493033341206e-05, + "loss": 1.0569, + "step": 6320 + }, + { + "epoch": 0.41, + "learning_rate": 1.8665278344318255e-05, + "loss": 1.1759, + "step": 6321 + }, + { + "epoch": 0.41, + "learning_rate": 1.8665063655295307e-05, + "loss": 1.1814, + "step": 6322 + }, + { + "epoch": 0.41, + "learning_rate": 1.8664848966272356e-05, + "loss": 1.0721, + "step": 6323 + }, + { + "epoch": 0.41, + "learning_rate": 1.8664634277249405e-05, + "loss": 1.0972, + "step": 6324 + }, + { + "epoch": 0.41, + "learning_rate": 1.8664419588226454e-05, + "loss": 1.0014, + "step": 6325 + }, + { + "epoch": 0.41, + "learning_rate": 1.8664204899203506e-05, + "loss": 1.1164, + "step": 6326 + }, + { + "epoch": 0.41, + "learning_rate": 1.8663990210180555e-05, + "loss": 1.1556, + "step": 6327 + }, + { + "epoch": 0.41, + "learning_rate": 1.8663775521157604e-05, + "loss": 1.1197, + "step": 6328 + }, + { + "epoch": 0.41, + "learning_rate": 1.8663560832134657e-05, + "loss": 1.1614, + "step": 6329 + }, + { + "epoch": 0.41, + "learning_rate": 1.8663346143111706e-05, + "loss": 1.1834, + "step": 6330 + }, + { + "epoch": 0.41, + "learning_rate": 1.8663131454088755e-05, + "loss": 1.0574, + "step": 6331 + }, + { + "epoch": 0.41, + "learning_rate": 1.8662916765065804e-05, + "loss": 1.1258, + "step": 6332 + }, + { + "epoch": 0.41, + "learning_rate": 1.8662702076042853e-05, + "loss": 1.1402, + "step": 6333 + }, + { + "epoch": 0.41, + "learning_rate": 1.86624873870199e-05, + "loss": 1.1772, + "step": 6334 + }, + { + "epoch": 0.41, + "learning_rate": 1.8662272697996954e-05, + "loss": 1.2307, + "step": 6335 + }, + { + "epoch": 0.41, + "learning_rate": 1.8662058008974003e-05, + "loss": 1.099, + "step": 6336 + }, + { + "epoch": 0.41, + "learning_rate": 1.8661843319951052e-05, + "loss": 1.0292, + "step": 6337 + }, + { + "epoch": 0.41, + "learning_rate": 1.86616286309281e-05, + "loss": 1.0727, + "step": 6338 + }, + { + "epoch": 0.41, + "learning_rate": 1.866141394190515e-05, + "loss": 1.1311, + "step": 6339 + }, + { + "epoch": 0.41, + "learning_rate": 1.8661199252882202e-05, + "loss": 1.1339, + "step": 6340 + }, + { + "epoch": 0.41, + "learning_rate": 1.866098456385925e-05, + "loss": 1.2164, + "step": 6341 + }, + { + "epoch": 0.41, + "learning_rate": 1.8660769874836303e-05, + "loss": 1.0956, + "step": 6342 + }, + { + "epoch": 0.41, + "learning_rate": 1.8660555185813352e-05, + "loss": 1.1133, + "step": 6343 + }, + { + "epoch": 0.41, + "learning_rate": 1.86603404967904e-05, + "loss": 1.1315, + "step": 6344 + }, + { + "epoch": 0.41, + "learning_rate": 1.866012580776745e-05, + "loss": 1.0163, + "step": 6345 + }, + { + "epoch": 0.41, + "learning_rate": 1.86599111187445e-05, + "loss": 1.1217, + "step": 6346 + }, + { + "epoch": 0.41, + "learning_rate": 1.8659696429721548e-05, + "loss": 1.1755, + "step": 6347 + }, + { + "epoch": 0.41, + "learning_rate": 1.86594817406986e-05, + "loss": 1.1159, + "step": 6348 + }, + { + "epoch": 0.41, + "learning_rate": 1.865926705167565e-05, + "loss": 1.099, + "step": 6349 + }, + { + "epoch": 0.41, + "learning_rate": 1.86590523626527e-05, + "loss": 1.0853, + "step": 6350 + }, + { + "epoch": 0.41, + "learning_rate": 1.8658837673629748e-05, + "loss": 1.0756, + "step": 6351 + }, + { + "epoch": 0.41, + "learning_rate": 1.8658622984606797e-05, + "loss": 1.1027, + "step": 6352 + }, + { + "epoch": 0.41, + "learning_rate": 1.865840829558385e-05, + "loss": 1.0822, + "step": 6353 + }, + { + "epoch": 0.41, + "learning_rate": 1.8658193606560898e-05, + "loss": 1.1025, + "step": 6354 + }, + { + "epoch": 0.41, + "learning_rate": 1.865797891753795e-05, + "loss": 1.0497, + "step": 6355 + }, + { + "epoch": 0.41, + "learning_rate": 1.8657764228515e-05, + "loss": 1.1113, + "step": 6356 + }, + { + "epoch": 0.41, + "learning_rate": 1.8657549539492048e-05, + "loss": 1.1415, + "step": 6357 + }, + { + "epoch": 0.41, + "learning_rate": 1.8657334850469097e-05, + "loss": 1.0308, + "step": 6358 + }, + { + "epoch": 0.41, + "learning_rate": 1.8657120161446146e-05, + "loss": 1.0658, + "step": 6359 + }, + { + "epoch": 0.41, + "learning_rate": 1.86569054724232e-05, + "loss": 1.2059, + "step": 6360 + }, + { + "epoch": 0.41, + "learning_rate": 1.8656690783400247e-05, + "loss": 1.1035, + "step": 6361 + }, + { + "epoch": 0.41, + "learning_rate": 1.8656476094377296e-05, + "loss": 1.0472, + "step": 6362 + }, + { + "epoch": 0.41, + "learning_rate": 1.8656261405354345e-05, + "loss": 1.0181, + "step": 6363 + }, + { + "epoch": 0.41, + "learning_rate": 1.8656046716331394e-05, + "loss": 1.0912, + "step": 6364 + }, + { + "epoch": 0.41, + "learning_rate": 1.8655832027308443e-05, + "loss": 1.1006, + "step": 6365 + }, + { + "epoch": 0.41, + "learning_rate": 1.8655617338285496e-05, + "loss": 1.1811, + "step": 6366 + }, + { + "epoch": 0.41, + "learning_rate": 1.8655402649262545e-05, + "loss": 1.0325, + "step": 6367 + }, + { + "epoch": 0.41, + "learning_rate": 1.8655187960239594e-05, + "loss": 1.0983, + "step": 6368 + }, + { + "epoch": 0.41, + "learning_rate": 1.8654973271216642e-05, + "loss": 1.1194, + "step": 6369 + }, + { + "epoch": 0.41, + "learning_rate": 1.8654758582193695e-05, + "loss": 1.0845, + "step": 6370 + }, + { + "epoch": 0.41, + "learning_rate": 1.8654543893170744e-05, + "loss": 1.2296, + "step": 6371 + }, + { + "epoch": 0.41, + "learning_rate": 1.8654329204147793e-05, + "loss": 1.0819, + "step": 6372 + }, + { + "epoch": 0.41, + "learning_rate": 1.8654114515124845e-05, + "loss": 1.1085, + "step": 6373 + }, + { + "epoch": 0.41, + "learning_rate": 1.8653899826101894e-05, + "loss": 1.0493, + "step": 6374 + }, + { + "epoch": 0.41, + "learning_rate": 1.8653685137078943e-05, + "loss": 1.1259, + "step": 6375 + }, + { + "epoch": 0.41, + "learning_rate": 1.8653470448055992e-05, + "loss": 1.0931, + "step": 6376 + }, + { + "epoch": 0.41, + "learning_rate": 1.865325575903304e-05, + "loss": 1.1321, + "step": 6377 + }, + { + "epoch": 0.41, + "learning_rate": 1.8653041070010093e-05, + "loss": 1.0478, + "step": 6378 + }, + { + "epoch": 0.41, + "learning_rate": 1.8652826380987142e-05, + "loss": 1.068, + "step": 6379 + }, + { + "epoch": 0.41, + "learning_rate": 1.865261169196419e-05, + "loss": 1.1248, + "step": 6380 + }, + { + "epoch": 0.41, + "learning_rate": 1.865239700294124e-05, + "loss": 1.057, + "step": 6381 + }, + { + "epoch": 0.41, + "learning_rate": 1.865218231391829e-05, + "loss": 1.1252, + "step": 6382 + }, + { + "epoch": 0.41, + "learning_rate": 1.8651967624895338e-05, + "loss": 1.0841, + "step": 6383 + }, + { + "epoch": 0.41, + "learning_rate": 1.865175293587239e-05, + "loss": 1.0777, + "step": 6384 + }, + { + "epoch": 0.41, + "learning_rate": 1.865153824684944e-05, + "loss": 1.0553, + "step": 6385 + }, + { + "epoch": 0.41, + "learning_rate": 1.8651323557826492e-05, + "loss": 1.1093, + "step": 6386 + }, + { + "epoch": 0.41, + "learning_rate": 1.865110886880354e-05, + "loss": 1.1121, + "step": 6387 + }, + { + "epoch": 0.41, + "learning_rate": 1.865089417978059e-05, + "loss": 1.1138, + "step": 6388 + }, + { + "epoch": 0.41, + "learning_rate": 1.865067949075764e-05, + "loss": 1.0981, + "step": 6389 + }, + { + "epoch": 0.41, + "learning_rate": 1.8650464801734688e-05, + "loss": 1.0225, + "step": 6390 + }, + { + "epoch": 0.41, + "learning_rate": 1.865025011271174e-05, + "loss": 1.1442, + "step": 6391 + }, + { + "epoch": 0.41, + "learning_rate": 1.865003542368879e-05, + "loss": 1.0616, + "step": 6392 + }, + { + "epoch": 0.41, + "learning_rate": 1.8649820734665838e-05, + "loss": 1.0598, + "step": 6393 + }, + { + "epoch": 0.41, + "learning_rate": 1.8649606045642887e-05, + "loss": 1.0573, + "step": 6394 + }, + { + "epoch": 0.41, + "learning_rate": 1.8649391356619936e-05, + "loss": 1.0965, + "step": 6395 + }, + { + "epoch": 0.41, + "learning_rate": 1.8649176667596985e-05, + "loss": 1.1476, + "step": 6396 + }, + { + "epoch": 0.41, + "learning_rate": 1.8648961978574037e-05, + "loss": 1.1333, + "step": 6397 + }, + { + "epoch": 0.41, + "learning_rate": 1.8648747289551086e-05, + "loss": 1.1432, + "step": 6398 + }, + { + "epoch": 0.41, + "learning_rate": 1.864853260052814e-05, + "loss": 1.0989, + "step": 6399 + }, + { + "epoch": 0.41, + "learning_rate": 1.8648317911505187e-05, + "loss": 1.1048, + "step": 6400 + }, + { + "epoch": 0.41, + "learning_rate": 1.8648103222482236e-05, + "loss": 1.0672, + "step": 6401 + }, + { + "epoch": 0.41, + "learning_rate": 1.8647888533459285e-05, + "loss": 1.0961, + "step": 6402 + }, + { + "epoch": 0.41, + "learning_rate": 1.8647673844436334e-05, + "loss": 1.0954, + "step": 6403 + }, + { + "epoch": 0.41, + "learning_rate": 1.8647459155413387e-05, + "loss": 1.1279, + "step": 6404 + }, + { + "epoch": 0.41, + "learning_rate": 1.8647244466390436e-05, + "loss": 1.1387, + "step": 6405 + }, + { + "epoch": 0.41, + "learning_rate": 1.8647029777367485e-05, + "loss": 1.0039, + "step": 6406 + }, + { + "epoch": 0.41, + "learning_rate": 1.8646815088344534e-05, + "loss": 1.1648, + "step": 6407 + }, + { + "epoch": 0.41, + "learning_rate": 1.8646600399321583e-05, + "loss": 1.1269, + "step": 6408 + }, + { + "epoch": 0.41, + "learning_rate": 1.8646385710298635e-05, + "loss": 1.1916, + "step": 6409 + }, + { + "epoch": 0.41, + "learning_rate": 1.8646171021275684e-05, + "loss": 1.1177, + "step": 6410 + }, + { + "epoch": 0.41, + "learning_rate": 1.8645956332252733e-05, + "loss": 1.0571, + "step": 6411 + }, + { + "epoch": 0.41, + "learning_rate": 1.8645741643229782e-05, + "loss": 1.0201, + "step": 6412 + }, + { + "epoch": 0.41, + "learning_rate": 1.8645526954206834e-05, + "loss": 1.1076, + "step": 6413 + }, + { + "epoch": 0.41, + "learning_rate": 1.8645312265183883e-05, + "loss": 1.0853, + "step": 6414 + }, + { + "epoch": 0.41, + "learning_rate": 1.8645097576160932e-05, + "loss": 1.1446, + "step": 6415 + }, + { + "epoch": 0.41, + "learning_rate": 1.864488288713798e-05, + "loss": 1.1276, + "step": 6416 + }, + { + "epoch": 0.41, + "learning_rate": 1.8644668198115033e-05, + "loss": 1.07, + "step": 6417 + }, + { + "epoch": 0.41, + "learning_rate": 1.8644453509092082e-05, + "loss": 1.1112, + "step": 6418 + }, + { + "epoch": 0.41, + "learning_rate": 1.864423882006913e-05, + "loss": 1.0668, + "step": 6419 + }, + { + "epoch": 0.41, + "learning_rate": 1.864402413104618e-05, + "loss": 1.2494, + "step": 6420 + }, + { + "epoch": 0.41, + "learning_rate": 1.864380944202323e-05, + "loss": 1.0996, + "step": 6421 + }, + { + "epoch": 0.41, + "learning_rate": 1.864359475300028e-05, + "loss": 1.1447, + "step": 6422 + }, + { + "epoch": 0.41, + "learning_rate": 1.864338006397733e-05, + "loss": 1.1772, + "step": 6423 + }, + { + "epoch": 0.41, + "learning_rate": 1.864316537495438e-05, + "loss": 1.1008, + "step": 6424 + }, + { + "epoch": 0.41, + "learning_rate": 1.864295068593143e-05, + "loss": 1.0756, + "step": 6425 + }, + { + "epoch": 0.41, + "learning_rate": 1.8642735996908478e-05, + "loss": 1.139, + "step": 6426 + }, + { + "epoch": 0.41, + "learning_rate": 1.864252130788553e-05, + "loss": 1.136, + "step": 6427 + }, + { + "epoch": 0.41, + "learning_rate": 1.864230661886258e-05, + "loss": 1.0862, + "step": 6428 + }, + { + "epoch": 0.41, + "learning_rate": 1.864209192983963e-05, + "loss": 1.129, + "step": 6429 + }, + { + "epoch": 0.41, + "learning_rate": 1.864187724081668e-05, + "loss": 1.0489, + "step": 6430 + }, + { + "epoch": 0.41, + "learning_rate": 1.864166255179373e-05, + "loss": 1.0976, + "step": 6431 + }, + { + "epoch": 0.41, + "learning_rate": 1.8641447862770778e-05, + "loss": 1.0438, + "step": 6432 + }, + { + "epoch": 0.41, + "learning_rate": 1.8641233173747827e-05, + "loss": 1.2227, + "step": 6433 + }, + { + "epoch": 0.41, + "learning_rate": 1.8641018484724876e-05, + "loss": 1.0836, + "step": 6434 + }, + { + "epoch": 0.41, + "learning_rate": 1.864080379570193e-05, + "loss": 1.1095, + "step": 6435 + }, + { + "epoch": 0.41, + "learning_rate": 1.8640589106678977e-05, + "loss": 1.1794, + "step": 6436 + }, + { + "epoch": 0.41, + "learning_rate": 1.8640374417656026e-05, + "loss": 1.0939, + "step": 6437 + }, + { + "epoch": 0.41, + "learning_rate": 1.8640159728633075e-05, + "loss": 1.1952, + "step": 6438 + }, + { + "epoch": 0.41, + "learning_rate": 1.8639945039610124e-05, + "loss": 1.0873, + "step": 6439 + }, + { + "epoch": 0.41, + "learning_rate": 1.8639730350587177e-05, + "loss": 1.1212, + "step": 6440 + }, + { + "epoch": 0.41, + "learning_rate": 1.8639515661564226e-05, + "loss": 1.1214, + "step": 6441 + }, + { + "epoch": 0.41, + "learning_rate": 1.8639300972541278e-05, + "loss": 1.0387, + "step": 6442 + }, + { + "epoch": 0.41, + "learning_rate": 1.8639086283518327e-05, + "loss": 1.0843, + "step": 6443 + }, + { + "epoch": 0.41, + "learning_rate": 1.8638871594495376e-05, + "loss": 1.1085, + "step": 6444 + }, + { + "epoch": 0.41, + "learning_rate": 1.8638656905472425e-05, + "loss": 1.0702, + "step": 6445 + }, + { + "epoch": 0.41, + "learning_rate": 1.8638442216449474e-05, + "loss": 1.1284, + "step": 6446 + }, + { + "epoch": 0.41, + "learning_rate": 1.8638227527426526e-05, + "loss": 1.0731, + "step": 6447 + }, + { + "epoch": 0.41, + "learning_rate": 1.8638012838403575e-05, + "loss": 1.0648, + "step": 6448 + }, + { + "epoch": 0.41, + "learning_rate": 1.8637798149380624e-05, + "loss": 1.128, + "step": 6449 + }, + { + "epoch": 0.41, + "learning_rate": 1.8637583460357673e-05, + "loss": 1.1738, + "step": 6450 + }, + { + "epoch": 0.42, + "learning_rate": 1.8637368771334722e-05, + "loss": 1.1503, + "step": 6451 + }, + { + "epoch": 0.42, + "learning_rate": 1.863715408231177e-05, + "loss": 1.1153, + "step": 6452 + }, + { + "epoch": 0.42, + "learning_rate": 1.8636939393288823e-05, + "loss": 1.0063, + "step": 6453 + }, + { + "epoch": 0.42, + "learning_rate": 1.8636724704265872e-05, + "loss": 1.0384, + "step": 6454 + }, + { + "epoch": 0.42, + "learning_rate": 1.863651001524292e-05, + "loss": 1.2267, + "step": 6455 + }, + { + "epoch": 0.42, + "learning_rate": 1.863629532621997e-05, + "loss": 1.1365, + "step": 6456 + }, + { + "epoch": 0.42, + "learning_rate": 1.8636080637197023e-05, + "loss": 1.1441, + "step": 6457 + }, + { + "epoch": 0.42, + "learning_rate": 1.863586594817407e-05, + "loss": 1.1367, + "step": 6458 + }, + { + "epoch": 0.42, + "learning_rate": 1.863565125915112e-05, + "loss": 1.1483, + "step": 6459 + }, + { + "epoch": 0.42, + "learning_rate": 1.8635436570128173e-05, + "loss": 1.0619, + "step": 6460 + }, + { + "epoch": 0.42, + "learning_rate": 1.8635221881105222e-05, + "loss": 1.1422, + "step": 6461 + }, + { + "epoch": 0.42, + "learning_rate": 1.863500719208227e-05, + "loss": 1.1156, + "step": 6462 + }, + { + "epoch": 0.42, + "learning_rate": 1.863479250305932e-05, + "loss": 1.0953, + "step": 6463 + }, + { + "epoch": 0.42, + "learning_rate": 1.863457781403637e-05, + "loss": 1.0782, + "step": 6464 + }, + { + "epoch": 0.42, + "learning_rate": 1.863436312501342e-05, + "loss": 1.0636, + "step": 6465 + }, + { + "epoch": 0.42, + "learning_rate": 1.863414843599047e-05, + "loss": 1.0294, + "step": 6466 + }, + { + "epoch": 0.42, + "learning_rate": 1.863393374696752e-05, + "loss": 1.0413, + "step": 6467 + }, + { + "epoch": 0.42, + "learning_rate": 1.8633719057944568e-05, + "loss": 1.1321, + "step": 6468 + }, + { + "epoch": 0.42, + "learning_rate": 1.8633504368921617e-05, + "loss": 1.2039, + "step": 6469 + }, + { + "epoch": 0.42, + "learning_rate": 1.8633289679898666e-05, + "loss": 1.1065, + "step": 6470 + }, + { + "epoch": 0.42, + "learning_rate": 1.8633074990875718e-05, + "loss": 1.1008, + "step": 6471 + }, + { + "epoch": 0.42, + "learning_rate": 1.8632860301852767e-05, + "loss": 1.0587, + "step": 6472 + }, + { + "epoch": 0.42, + "learning_rate": 1.863264561282982e-05, + "loss": 1.1162, + "step": 6473 + }, + { + "epoch": 0.42, + "learning_rate": 1.863243092380687e-05, + "loss": 1.0911, + "step": 6474 + }, + { + "epoch": 0.42, + "learning_rate": 1.8632216234783917e-05, + "loss": 1.0885, + "step": 6475 + }, + { + "epoch": 0.42, + "learning_rate": 1.8632001545760966e-05, + "loss": 1.0742, + "step": 6476 + }, + { + "epoch": 0.42, + "learning_rate": 1.8631786856738015e-05, + "loss": 1.109, + "step": 6477 + }, + { + "epoch": 0.42, + "learning_rate": 1.8631572167715068e-05, + "loss": 1.2158, + "step": 6478 + }, + { + "epoch": 0.42, + "learning_rate": 1.8631357478692117e-05, + "loss": 1.1712, + "step": 6479 + }, + { + "epoch": 0.42, + "learning_rate": 1.8631142789669166e-05, + "loss": 1.1595, + "step": 6480 + }, + { + "epoch": 0.42, + "learning_rate": 1.8630928100646215e-05, + "loss": 1.1862, + "step": 6481 + }, + { + "epoch": 0.42, + "learning_rate": 1.8630713411623264e-05, + "loss": 1.0992, + "step": 6482 + }, + { + "epoch": 0.42, + "learning_rate": 1.8630498722600313e-05, + "loss": 1.1143, + "step": 6483 + }, + { + "epoch": 0.42, + "learning_rate": 1.8630284033577365e-05, + "loss": 1.1506, + "step": 6484 + }, + { + "epoch": 0.42, + "learning_rate": 1.8630069344554414e-05, + "loss": 1.2043, + "step": 6485 + }, + { + "epoch": 0.42, + "learning_rate": 1.8629854655531466e-05, + "loss": 1.1752, + "step": 6486 + }, + { + "epoch": 0.42, + "learning_rate": 1.8629639966508515e-05, + "loss": 1.0107, + "step": 6487 + }, + { + "epoch": 0.42, + "learning_rate": 1.8629425277485564e-05, + "loss": 1.1394, + "step": 6488 + }, + { + "epoch": 0.42, + "learning_rate": 1.8629210588462613e-05, + "loss": 1.1595, + "step": 6489 + }, + { + "epoch": 0.42, + "learning_rate": 1.8628995899439662e-05, + "loss": 1.104, + "step": 6490 + }, + { + "epoch": 0.42, + "learning_rate": 1.8628781210416714e-05, + "loss": 1.0047, + "step": 6491 + }, + { + "epoch": 0.42, + "learning_rate": 1.8628566521393763e-05, + "loss": 1.0843, + "step": 6492 + }, + { + "epoch": 0.42, + "learning_rate": 1.8628351832370812e-05, + "loss": 1.1876, + "step": 6493 + }, + { + "epoch": 0.42, + "learning_rate": 1.862813714334786e-05, + "loss": 1.0595, + "step": 6494 + }, + { + "epoch": 0.42, + "learning_rate": 1.862792245432491e-05, + "loss": 1.1763, + "step": 6495 + }, + { + "epoch": 0.42, + "learning_rate": 1.8627707765301963e-05, + "loss": 1.114, + "step": 6496 + }, + { + "epoch": 0.42, + "learning_rate": 1.862749307627901e-05, + "loss": 1.1028, + "step": 6497 + }, + { + "epoch": 0.42, + "learning_rate": 1.862727838725606e-05, + "loss": 1.056, + "step": 6498 + }, + { + "epoch": 0.42, + "learning_rate": 1.862706369823311e-05, + "loss": 1.058, + "step": 6499 + }, + { + "epoch": 0.42, + "learning_rate": 1.8626849009210162e-05, + "loss": 1.1417, + "step": 6500 + }, + { + "epoch": 0.42, + "learning_rate": 1.862663432018721e-05, + "loss": 1.1367, + "step": 6501 + }, + { + "epoch": 0.42, + "learning_rate": 1.862641963116426e-05, + "loss": 1.0686, + "step": 6502 + }, + { + "epoch": 0.42, + "learning_rate": 1.862620494214131e-05, + "loss": 1.113, + "step": 6503 + }, + { + "epoch": 0.42, + "learning_rate": 1.862599025311836e-05, + "loss": 1.0445, + "step": 6504 + }, + { + "epoch": 0.42, + "learning_rate": 1.862577556409541e-05, + "loss": 1.1511, + "step": 6505 + }, + { + "epoch": 0.42, + "learning_rate": 1.862556087507246e-05, + "loss": 1.1215, + "step": 6506 + }, + { + "epoch": 0.42, + "learning_rate": 1.8625346186049508e-05, + "loss": 1.1358, + "step": 6507 + }, + { + "epoch": 0.42, + "learning_rate": 1.8625131497026557e-05, + "loss": 1.1665, + "step": 6508 + }, + { + "epoch": 0.42, + "learning_rate": 1.862491680800361e-05, + "loss": 1.1063, + "step": 6509 + }, + { + "epoch": 0.42, + "learning_rate": 1.862470211898066e-05, + "loss": 1.1059, + "step": 6510 + }, + { + "epoch": 0.42, + "learning_rate": 1.8624487429957707e-05, + "loss": 1.1366, + "step": 6511 + }, + { + "epoch": 0.42, + "learning_rate": 1.8624272740934756e-05, + "loss": 1.0316, + "step": 6512 + }, + { + "epoch": 0.42, + "learning_rate": 1.8624058051911805e-05, + "loss": 1.0846, + "step": 6513 + }, + { + "epoch": 0.42, + "learning_rate": 1.8623843362888858e-05, + "loss": 1.0937, + "step": 6514 + }, + { + "epoch": 0.42, + "learning_rate": 1.8623628673865907e-05, + "loss": 1.0442, + "step": 6515 + }, + { + "epoch": 0.42, + "learning_rate": 1.862341398484296e-05, + "loss": 1.015, + "step": 6516 + }, + { + "epoch": 0.42, + "learning_rate": 1.8623199295820008e-05, + "loss": 1.1402, + "step": 6517 + }, + { + "epoch": 0.42, + "learning_rate": 1.8622984606797057e-05, + "loss": 1.1861, + "step": 6518 + }, + { + "epoch": 0.42, + "learning_rate": 1.8622769917774106e-05, + "loss": 1.1004, + "step": 6519 + }, + { + "epoch": 0.42, + "learning_rate": 1.8622555228751155e-05, + "loss": 1.1616, + "step": 6520 + }, + { + "epoch": 0.42, + "learning_rate": 1.8622340539728204e-05, + "loss": 1.1194, + "step": 6521 + }, + { + "epoch": 0.42, + "learning_rate": 1.8622125850705256e-05, + "loss": 1.099, + "step": 6522 + }, + { + "epoch": 0.42, + "learning_rate": 1.8621911161682305e-05, + "loss": 1.0228, + "step": 6523 + }, + { + "epoch": 0.42, + "learning_rate": 1.8621696472659354e-05, + "loss": 1.1223, + "step": 6524 + }, + { + "epoch": 0.42, + "learning_rate": 1.8621481783636403e-05, + "loss": 1.0734, + "step": 6525 + }, + { + "epoch": 0.42, + "learning_rate": 1.8621267094613452e-05, + "loss": 1.1514, + "step": 6526 + }, + { + "epoch": 0.42, + "learning_rate": 1.8621052405590504e-05, + "loss": 1.1508, + "step": 6527 + }, + { + "epoch": 0.42, + "learning_rate": 1.8620837716567553e-05, + "loss": 1.0253, + "step": 6528 + }, + { + "epoch": 0.42, + "learning_rate": 1.8620623027544606e-05, + "loss": 1.1447, + "step": 6529 + }, + { + "epoch": 0.42, + "learning_rate": 1.8620408338521655e-05, + "loss": 1.0934, + "step": 6530 + }, + { + "epoch": 0.42, + "learning_rate": 1.8620193649498704e-05, + "loss": 1.106, + "step": 6531 + }, + { + "epoch": 0.42, + "learning_rate": 1.8619978960475753e-05, + "loss": 1.087, + "step": 6532 + }, + { + "epoch": 0.42, + "learning_rate": 1.86197642714528e-05, + "loss": 1.0461, + "step": 6533 + }, + { + "epoch": 0.42, + "learning_rate": 1.8619549582429854e-05, + "loss": 1.1268, + "step": 6534 + }, + { + "epoch": 0.42, + "learning_rate": 1.8619334893406903e-05, + "loss": 1.0824, + "step": 6535 + }, + { + "epoch": 0.42, + "learning_rate": 1.8619120204383952e-05, + "loss": 1.0732, + "step": 6536 + }, + { + "epoch": 0.42, + "learning_rate": 1.8618905515361e-05, + "loss": 1.1196, + "step": 6537 + }, + { + "epoch": 0.42, + "learning_rate": 1.861869082633805e-05, + "loss": 1.094, + "step": 6538 + }, + { + "epoch": 0.42, + "learning_rate": 1.86184761373151e-05, + "loss": 1.0833, + "step": 6539 + }, + { + "epoch": 0.42, + "learning_rate": 1.861826144829215e-05, + "loss": 1.0177, + "step": 6540 + }, + { + "epoch": 0.42, + "learning_rate": 1.86180467592692e-05, + "loss": 1.1461, + "step": 6541 + }, + { + "epoch": 0.42, + "learning_rate": 1.861783207024625e-05, + "loss": 1.0277, + "step": 6542 + }, + { + "epoch": 0.42, + "learning_rate": 1.8617617381223298e-05, + "loss": 1.0879, + "step": 6543 + }, + { + "epoch": 0.42, + "learning_rate": 1.861740269220035e-05, + "loss": 1.0884, + "step": 6544 + }, + { + "epoch": 0.42, + "learning_rate": 1.86171880031774e-05, + "loss": 1.0638, + "step": 6545 + }, + { + "epoch": 0.42, + "learning_rate": 1.8616973314154448e-05, + "loss": 1.0945, + "step": 6546 + }, + { + "epoch": 0.42, + "learning_rate": 1.86167586251315e-05, + "loss": 1.0182, + "step": 6547 + }, + { + "epoch": 0.42, + "learning_rate": 1.861654393610855e-05, + "loss": 1.0651, + "step": 6548 + }, + { + "epoch": 0.42, + "learning_rate": 1.86163292470856e-05, + "loss": 0.9853, + "step": 6549 + }, + { + "epoch": 0.42, + "learning_rate": 1.8616114558062647e-05, + "loss": 1.1101, + "step": 6550 + }, + { + "epoch": 0.42, + "learning_rate": 1.8615899869039696e-05, + "loss": 1.1056, + "step": 6551 + }, + { + "epoch": 0.42, + "learning_rate": 1.861568518001675e-05, + "loss": 1.0543, + "step": 6552 + }, + { + "epoch": 0.42, + "learning_rate": 1.8615470490993798e-05, + "loss": 1.1279, + "step": 6553 + }, + { + "epoch": 0.42, + "learning_rate": 1.8615255801970847e-05, + "loss": 1.2299, + "step": 6554 + }, + { + "epoch": 0.42, + "learning_rate": 1.8615041112947896e-05, + "loss": 1.0856, + "step": 6555 + }, + { + "epoch": 0.42, + "learning_rate": 1.8614826423924945e-05, + "loss": 1.1326, + "step": 6556 + }, + { + "epoch": 0.42, + "learning_rate": 1.8614611734901994e-05, + "loss": 1.061, + "step": 6557 + }, + { + "epoch": 0.42, + "learning_rate": 1.8614397045879046e-05, + "loss": 1.1318, + "step": 6558 + }, + { + "epoch": 0.42, + "learning_rate": 1.8614182356856095e-05, + "loss": 1.0299, + "step": 6559 + }, + { + "epoch": 0.42, + "learning_rate": 1.8613967667833147e-05, + "loss": 1.1424, + "step": 6560 + }, + { + "epoch": 0.42, + "learning_rate": 1.8613752978810196e-05, + "loss": 1.105, + "step": 6561 + }, + { + "epoch": 0.42, + "learning_rate": 1.8613538289787245e-05, + "loss": 1.1121, + "step": 6562 + }, + { + "epoch": 0.42, + "learning_rate": 1.8613323600764294e-05, + "loss": 1.1769, + "step": 6563 + }, + { + "epoch": 0.42, + "learning_rate": 1.8613108911741343e-05, + "loss": 1.107, + "step": 6564 + }, + { + "epoch": 0.42, + "learning_rate": 1.8612894222718395e-05, + "loss": 1.1438, + "step": 6565 + }, + { + "epoch": 0.42, + "learning_rate": 1.8612679533695444e-05, + "loss": 1.1, + "step": 6566 + }, + { + "epoch": 0.42, + "learning_rate": 1.8612464844672493e-05, + "loss": 1.0814, + "step": 6567 + }, + { + "epoch": 0.42, + "learning_rate": 1.8612250155649542e-05, + "loss": 1.1305, + "step": 6568 + }, + { + "epoch": 0.42, + "learning_rate": 1.861203546662659e-05, + "loss": 1.1503, + "step": 6569 + }, + { + "epoch": 0.42, + "learning_rate": 1.861182077760364e-05, + "loss": 1.0662, + "step": 6570 + }, + { + "epoch": 0.42, + "learning_rate": 1.8611606088580693e-05, + "loss": 1.0868, + "step": 6571 + }, + { + "epoch": 0.42, + "learning_rate": 1.861139139955774e-05, + "loss": 1.1078, + "step": 6572 + }, + { + "epoch": 0.42, + "learning_rate": 1.8611176710534794e-05, + "loss": 1.1139, + "step": 6573 + }, + { + "epoch": 0.42, + "learning_rate": 1.8610962021511843e-05, + "loss": 1.0977, + "step": 6574 + }, + { + "epoch": 0.42, + "learning_rate": 1.8610747332488892e-05, + "loss": 1.1067, + "step": 6575 + }, + { + "epoch": 0.42, + "learning_rate": 1.861053264346594e-05, + "loss": 1.0674, + "step": 6576 + }, + { + "epoch": 0.42, + "learning_rate": 1.861031795444299e-05, + "loss": 1.0836, + "step": 6577 + }, + { + "epoch": 0.42, + "learning_rate": 1.8610103265420042e-05, + "loss": 1.1434, + "step": 6578 + }, + { + "epoch": 0.42, + "learning_rate": 1.860988857639709e-05, + "loss": 1.0244, + "step": 6579 + }, + { + "epoch": 0.42, + "learning_rate": 1.860967388737414e-05, + "loss": 1.05, + "step": 6580 + }, + { + "epoch": 0.42, + "learning_rate": 1.860945919835119e-05, + "loss": 1.1353, + "step": 6581 + }, + { + "epoch": 0.42, + "learning_rate": 1.8609244509328238e-05, + "loss": 1.0854, + "step": 6582 + }, + { + "epoch": 0.42, + "learning_rate": 1.860902982030529e-05, + "loss": 1.1492, + "step": 6583 + }, + { + "epoch": 0.42, + "learning_rate": 1.860881513128234e-05, + "loss": 1.033, + "step": 6584 + }, + { + "epoch": 0.42, + "learning_rate": 1.860860044225939e-05, + "loss": 1.0841, + "step": 6585 + }, + { + "epoch": 0.42, + "learning_rate": 1.8608385753236437e-05, + "loss": 1.1642, + "step": 6586 + }, + { + "epoch": 0.42, + "learning_rate": 1.8608171064213486e-05, + "loss": 1.0768, + "step": 6587 + }, + { + "epoch": 0.42, + "learning_rate": 1.860795637519054e-05, + "loss": 1.1473, + "step": 6588 + }, + { + "epoch": 0.42, + "learning_rate": 1.8607741686167588e-05, + "loss": 1.123, + "step": 6589 + }, + { + "epoch": 0.42, + "learning_rate": 1.8607526997144637e-05, + "loss": 1.0641, + "step": 6590 + }, + { + "epoch": 0.42, + "learning_rate": 1.860731230812169e-05, + "loss": 1.0686, + "step": 6591 + }, + { + "epoch": 0.42, + "learning_rate": 1.8607097619098738e-05, + "loss": 1.0717, + "step": 6592 + }, + { + "epoch": 0.42, + "learning_rate": 1.8606882930075787e-05, + "loss": 1.0488, + "step": 6593 + }, + { + "epoch": 0.42, + "learning_rate": 1.8606668241052836e-05, + "loss": 1.041, + "step": 6594 + }, + { + "epoch": 0.42, + "learning_rate": 1.8606453552029885e-05, + "loss": 0.9924, + "step": 6595 + }, + { + "epoch": 0.42, + "learning_rate": 1.8606238863006937e-05, + "loss": 1.1261, + "step": 6596 + }, + { + "epoch": 0.42, + "learning_rate": 1.8606024173983986e-05, + "loss": 1.1104, + "step": 6597 + }, + { + "epoch": 0.42, + "learning_rate": 1.8605809484961035e-05, + "loss": 1.1862, + "step": 6598 + }, + { + "epoch": 0.42, + "learning_rate": 1.8605594795938084e-05, + "loss": 1.1124, + "step": 6599 + }, + { + "epoch": 0.42, + "learning_rate": 1.8605380106915133e-05, + "loss": 1.0991, + "step": 6600 + }, + { + "epoch": 0.42, + "learning_rate": 1.8605165417892185e-05, + "loss": 1.1034, + "step": 6601 + }, + { + "epoch": 0.42, + "learning_rate": 1.8604950728869234e-05, + "loss": 0.9997, + "step": 6602 + }, + { + "epoch": 0.42, + "learning_rate": 1.8604736039846287e-05, + "loss": 1.035, + "step": 6603 + }, + { + "epoch": 0.42, + "learning_rate": 1.8604521350823336e-05, + "loss": 1.0816, + "step": 6604 + }, + { + "epoch": 0.42, + "learning_rate": 1.8604306661800385e-05, + "loss": 1.0779, + "step": 6605 + }, + { + "epoch": 0.43, + "learning_rate": 1.8604091972777434e-05, + "loss": 1.1994, + "step": 6606 + }, + { + "epoch": 0.43, + "learning_rate": 1.8603877283754482e-05, + "loss": 1.1398, + "step": 6607 + }, + { + "epoch": 0.43, + "learning_rate": 1.860366259473153e-05, + "loss": 1.1529, + "step": 6608 + }, + { + "epoch": 0.43, + "learning_rate": 1.8603447905708584e-05, + "loss": 1.0519, + "step": 6609 + }, + { + "epoch": 0.43, + "learning_rate": 1.8603233216685633e-05, + "loss": 1.0041, + "step": 6610 + }, + { + "epoch": 0.43, + "learning_rate": 1.8603018527662682e-05, + "loss": 1.1154, + "step": 6611 + }, + { + "epoch": 0.43, + "learning_rate": 1.860280383863973e-05, + "loss": 1.0446, + "step": 6612 + }, + { + "epoch": 0.43, + "learning_rate": 1.860258914961678e-05, + "loss": 1.141, + "step": 6613 + }, + { + "epoch": 0.43, + "learning_rate": 1.8602374460593832e-05, + "loss": 1.1183, + "step": 6614 + }, + { + "epoch": 0.43, + "learning_rate": 1.860215977157088e-05, + "loss": 1.1177, + "step": 6615 + }, + { + "epoch": 0.43, + "learning_rate": 1.860194508254793e-05, + "loss": 1.0487, + "step": 6616 + }, + { + "epoch": 0.43, + "learning_rate": 1.8601730393524982e-05, + "loss": 1.079, + "step": 6617 + }, + { + "epoch": 0.43, + "learning_rate": 1.860151570450203e-05, + "loss": 1.1145, + "step": 6618 + }, + { + "epoch": 0.43, + "learning_rate": 1.860130101547908e-05, + "loss": 1.0643, + "step": 6619 + }, + { + "epoch": 0.43, + "learning_rate": 1.860108632645613e-05, + "loss": 1.0607, + "step": 6620 + }, + { + "epoch": 0.43, + "learning_rate": 1.860087163743318e-05, + "loss": 1.047, + "step": 6621 + }, + { + "epoch": 0.43, + "learning_rate": 1.860065694841023e-05, + "loss": 1.1068, + "step": 6622 + }, + { + "epoch": 0.43, + "learning_rate": 1.860044225938728e-05, + "loss": 1.0877, + "step": 6623 + }, + { + "epoch": 0.43, + "learning_rate": 1.860022757036433e-05, + "loss": 1.0846, + "step": 6624 + }, + { + "epoch": 0.43, + "learning_rate": 1.8600012881341377e-05, + "loss": 1.2471, + "step": 6625 + }, + { + "epoch": 0.43, + "learning_rate": 1.8599798192318426e-05, + "loss": 1.0778, + "step": 6626 + }, + { + "epoch": 0.43, + "learning_rate": 1.859958350329548e-05, + "loss": 1.0726, + "step": 6627 + }, + { + "epoch": 0.43, + "learning_rate": 1.8599368814272528e-05, + "loss": 1.0409, + "step": 6628 + }, + { + "epoch": 0.43, + "learning_rate": 1.8599154125249577e-05, + "loss": 1.1483, + "step": 6629 + }, + { + "epoch": 0.43, + "learning_rate": 1.8598939436226626e-05, + "loss": 1.1098, + "step": 6630 + }, + { + "epoch": 0.43, + "learning_rate": 1.8598724747203678e-05, + "loss": 1.047, + "step": 6631 + }, + { + "epoch": 0.43, + "learning_rate": 1.8598510058180727e-05, + "loss": 1.0512, + "step": 6632 + }, + { + "epoch": 0.43, + "learning_rate": 1.8598295369157776e-05, + "loss": 1.1523, + "step": 6633 + }, + { + "epoch": 0.43, + "learning_rate": 1.8598080680134828e-05, + "loss": 1.0747, + "step": 6634 + }, + { + "epoch": 0.43, + "learning_rate": 1.8597865991111877e-05, + "loss": 1.054, + "step": 6635 + }, + { + "epoch": 0.43, + "learning_rate": 1.8597651302088926e-05, + "loss": 1.1239, + "step": 6636 + }, + { + "epoch": 0.43, + "learning_rate": 1.8597436613065975e-05, + "loss": 1.0698, + "step": 6637 + }, + { + "epoch": 0.43, + "learning_rate": 1.8597221924043024e-05, + "loss": 1.1489, + "step": 6638 + }, + { + "epoch": 0.43, + "learning_rate": 1.8597007235020076e-05, + "loss": 1.2198, + "step": 6639 + }, + { + "epoch": 0.43, + "learning_rate": 1.8596792545997125e-05, + "loss": 1.0654, + "step": 6640 + }, + { + "epoch": 0.43, + "learning_rate": 1.8596577856974174e-05, + "loss": 1.2045, + "step": 6641 + }, + { + "epoch": 0.43, + "learning_rate": 1.8596363167951223e-05, + "loss": 1.0596, + "step": 6642 + }, + { + "epoch": 0.43, + "learning_rate": 1.8596148478928272e-05, + "loss": 1.1241, + "step": 6643 + }, + { + "epoch": 0.43, + "learning_rate": 1.859593378990532e-05, + "loss": 1.1285, + "step": 6644 + }, + { + "epoch": 0.43, + "learning_rate": 1.8595719100882374e-05, + "loss": 1.0592, + "step": 6645 + }, + { + "epoch": 0.43, + "learning_rate": 1.8595504411859423e-05, + "loss": 1.0435, + "step": 6646 + }, + { + "epoch": 0.43, + "learning_rate": 1.8595289722836475e-05, + "loss": 1.1032, + "step": 6647 + }, + { + "epoch": 0.43, + "learning_rate": 1.8595075033813524e-05, + "loss": 1.0635, + "step": 6648 + }, + { + "epoch": 0.43, + "learning_rate": 1.8594860344790573e-05, + "loss": 1.1489, + "step": 6649 + }, + { + "epoch": 0.43, + "learning_rate": 1.8594645655767622e-05, + "loss": 1.1029, + "step": 6650 + }, + { + "epoch": 0.43, + "learning_rate": 1.859443096674467e-05, + "loss": 1.0569, + "step": 6651 + }, + { + "epoch": 0.43, + "learning_rate": 1.8594216277721723e-05, + "loss": 1.1098, + "step": 6652 + }, + { + "epoch": 0.43, + "learning_rate": 1.8594001588698772e-05, + "loss": 1.0746, + "step": 6653 + }, + { + "epoch": 0.43, + "learning_rate": 1.859378689967582e-05, + "loss": 1.1648, + "step": 6654 + }, + { + "epoch": 0.43, + "learning_rate": 1.859357221065287e-05, + "loss": 1.1326, + "step": 6655 + }, + { + "epoch": 0.43, + "learning_rate": 1.859335752162992e-05, + "loss": 1.135, + "step": 6656 + }, + { + "epoch": 0.43, + "learning_rate": 1.8593142832606968e-05, + "loss": 0.9727, + "step": 6657 + }, + { + "epoch": 0.43, + "learning_rate": 1.859292814358402e-05, + "loss": 1.0849, + "step": 6658 + }, + { + "epoch": 0.43, + "learning_rate": 1.859271345456107e-05, + "loss": 1.0255, + "step": 6659 + }, + { + "epoch": 0.43, + "learning_rate": 1.859249876553812e-05, + "loss": 1.0953, + "step": 6660 + }, + { + "epoch": 0.43, + "learning_rate": 1.859228407651517e-05, + "loss": 1.2237, + "step": 6661 + }, + { + "epoch": 0.43, + "learning_rate": 1.859206938749222e-05, + "loss": 1.0799, + "step": 6662 + }, + { + "epoch": 0.43, + "learning_rate": 1.859185469846927e-05, + "loss": 1.1231, + "step": 6663 + }, + { + "epoch": 0.43, + "learning_rate": 1.8591640009446318e-05, + "loss": 1.1272, + "step": 6664 + }, + { + "epoch": 0.43, + "learning_rate": 1.859142532042337e-05, + "loss": 1.1407, + "step": 6665 + }, + { + "epoch": 0.43, + "learning_rate": 1.859121063140042e-05, + "loss": 1.1535, + "step": 6666 + }, + { + "epoch": 0.43, + "learning_rate": 1.8590995942377468e-05, + "loss": 1.1013, + "step": 6667 + }, + { + "epoch": 0.43, + "learning_rate": 1.8590781253354517e-05, + "loss": 1.1281, + "step": 6668 + }, + { + "epoch": 0.43, + "learning_rate": 1.8590566564331566e-05, + "loss": 1.0566, + "step": 6669 + }, + { + "epoch": 0.43, + "learning_rate": 1.8590351875308618e-05, + "loss": 1.1263, + "step": 6670 + }, + { + "epoch": 0.43, + "learning_rate": 1.8590137186285667e-05, + "loss": 1.0964, + "step": 6671 + }, + { + "epoch": 0.43, + "learning_rate": 1.8589922497262716e-05, + "loss": 1.0691, + "step": 6672 + }, + { + "epoch": 0.43, + "learning_rate": 1.8589707808239765e-05, + "loss": 1.1632, + "step": 6673 + }, + { + "epoch": 0.43, + "learning_rate": 1.8589493119216814e-05, + "loss": 1.0713, + "step": 6674 + }, + { + "epoch": 0.43, + "learning_rate": 1.8589278430193866e-05, + "loss": 1.1332, + "step": 6675 + }, + { + "epoch": 0.43, + "learning_rate": 1.8589063741170915e-05, + "loss": 1.2065, + "step": 6676 + }, + { + "epoch": 0.43, + "learning_rate": 1.8588849052147964e-05, + "loss": 1.0099, + "step": 6677 + }, + { + "epoch": 0.43, + "learning_rate": 1.8588634363125017e-05, + "loss": 1.1465, + "step": 6678 + }, + { + "epoch": 0.43, + "learning_rate": 1.8588419674102066e-05, + "loss": 1.1085, + "step": 6679 + }, + { + "epoch": 0.43, + "learning_rate": 1.8588204985079115e-05, + "loss": 1.1539, + "step": 6680 + }, + { + "epoch": 0.43, + "learning_rate": 1.8587990296056163e-05, + "loss": 1.1606, + "step": 6681 + }, + { + "epoch": 0.43, + "learning_rate": 1.8587775607033212e-05, + "loss": 1.0927, + "step": 6682 + }, + { + "epoch": 0.43, + "learning_rate": 1.8587560918010265e-05, + "loss": 1.1248, + "step": 6683 + }, + { + "epoch": 0.43, + "learning_rate": 1.8587346228987314e-05, + "loss": 1.1138, + "step": 6684 + }, + { + "epoch": 0.43, + "learning_rate": 1.8587131539964363e-05, + "loss": 1.1232, + "step": 6685 + }, + { + "epoch": 0.43, + "learning_rate": 1.858691685094141e-05, + "loss": 1.0866, + "step": 6686 + }, + { + "epoch": 0.43, + "learning_rate": 1.858670216191846e-05, + "loss": 1.1206, + "step": 6687 + }, + { + "epoch": 0.43, + "learning_rate": 1.8586487472895513e-05, + "loss": 1.0646, + "step": 6688 + }, + { + "epoch": 0.43, + "learning_rate": 1.8586272783872562e-05, + "loss": 1.084, + "step": 6689 + }, + { + "epoch": 0.43, + "learning_rate": 1.8586058094849614e-05, + "loss": 1.1463, + "step": 6690 + }, + { + "epoch": 0.43, + "learning_rate": 1.8585843405826663e-05, + "loss": 1.0554, + "step": 6691 + }, + { + "epoch": 0.43, + "learning_rate": 1.8585628716803712e-05, + "loss": 1.085, + "step": 6692 + }, + { + "epoch": 0.43, + "learning_rate": 1.858541402778076e-05, + "loss": 1.1033, + "step": 6693 + }, + { + "epoch": 0.43, + "learning_rate": 1.858519933875781e-05, + "loss": 1.0781, + "step": 6694 + }, + { + "epoch": 0.43, + "learning_rate": 1.858498464973486e-05, + "loss": 1.1982, + "step": 6695 + }, + { + "epoch": 0.43, + "learning_rate": 1.858476996071191e-05, + "loss": 1.1152, + "step": 6696 + }, + { + "epoch": 0.43, + "learning_rate": 1.858455527168896e-05, + "loss": 1.0274, + "step": 6697 + }, + { + "epoch": 0.43, + "learning_rate": 1.858434058266601e-05, + "loss": 1.127, + "step": 6698 + }, + { + "epoch": 0.43, + "learning_rate": 1.858412589364306e-05, + "loss": 1.1247, + "step": 6699 + }, + { + "epoch": 0.43, + "learning_rate": 1.8583911204620107e-05, + "loss": 1.1184, + "step": 6700 + }, + { + "epoch": 0.43, + "learning_rate": 1.858369651559716e-05, + "loss": 1.0469, + "step": 6701 + }, + { + "epoch": 0.43, + "learning_rate": 1.858348182657421e-05, + "loss": 1.0856, + "step": 6702 + }, + { + "epoch": 0.43, + "learning_rate": 1.8583267137551258e-05, + "loss": 1.1006, + "step": 6703 + }, + { + "epoch": 0.43, + "learning_rate": 1.858305244852831e-05, + "loss": 1.1347, + "step": 6704 + }, + { + "epoch": 0.43, + "learning_rate": 1.858283775950536e-05, + "loss": 1.0539, + "step": 6705 + }, + { + "epoch": 0.43, + "learning_rate": 1.8582623070482408e-05, + "loss": 1.0803, + "step": 6706 + }, + { + "epoch": 0.43, + "learning_rate": 1.8582408381459457e-05, + "loss": 1.0438, + "step": 6707 + }, + { + "epoch": 0.43, + "learning_rate": 1.858219369243651e-05, + "loss": 1.1234, + "step": 6708 + }, + { + "epoch": 0.43, + "learning_rate": 1.8581979003413558e-05, + "loss": 1.1499, + "step": 6709 + }, + { + "epoch": 0.43, + "learning_rate": 1.8581764314390607e-05, + "loss": 1.1316, + "step": 6710 + }, + { + "epoch": 0.43, + "learning_rate": 1.8581549625367656e-05, + "loss": 1.1279, + "step": 6711 + }, + { + "epoch": 0.43, + "learning_rate": 1.8581334936344705e-05, + "loss": 1.0469, + "step": 6712 + }, + { + "epoch": 0.43, + "learning_rate": 1.8581120247321754e-05, + "loss": 1.1505, + "step": 6713 + }, + { + "epoch": 0.43, + "learning_rate": 1.8580905558298806e-05, + "loss": 1.0429, + "step": 6714 + }, + { + "epoch": 0.43, + "learning_rate": 1.8580690869275855e-05, + "loss": 1.1981, + "step": 6715 + }, + { + "epoch": 0.43, + "learning_rate": 1.8580476180252904e-05, + "loss": 1.08, + "step": 6716 + }, + { + "epoch": 0.43, + "learning_rate": 1.8580261491229953e-05, + "loss": 1.1707, + "step": 6717 + }, + { + "epoch": 0.43, + "learning_rate": 1.8580046802207006e-05, + "loss": 1.0363, + "step": 6718 + }, + { + "epoch": 0.43, + "learning_rate": 1.8579832113184055e-05, + "loss": 1.0296, + "step": 6719 + }, + { + "epoch": 0.43, + "learning_rate": 1.8579617424161104e-05, + "loss": 1.1582, + "step": 6720 + }, + { + "epoch": 0.43, + "learning_rate": 1.8579402735138156e-05, + "loss": 1.1643, + "step": 6721 + }, + { + "epoch": 0.43, + "learning_rate": 1.8579188046115205e-05, + "loss": 1.1319, + "step": 6722 + }, + { + "epoch": 0.43, + "learning_rate": 1.8578973357092254e-05, + "loss": 1.0466, + "step": 6723 + }, + { + "epoch": 0.43, + "learning_rate": 1.8578758668069303e-05, + "loss": 1.0519, + "step": 6724 + }, + { + "epoch": 0.43, + "learning_rate": 1.8578543979046352e-05, + "loss": 1.0343, + "step": 6725 + }, + { + "epoch": 0.43, + "learning_rate": 1.85783292900234e-05, + "loss": 1.0338, + "step": 6726 + }, + { + "epoch": 0.43, + "learning_rate": 1.8578114601000453e-05, + "loss": 1.0956, + "step": 6727 + }, + { + "epoch": 0.43, + "learning_rate": 1.8577899911977502e-05, + "loss": 1.088, + "step": 6728 + }, + { + "epoch": 0.43, + "learning_rate": 1.857768522295455e-05, + "loss": 1.024, + "step": 6729 + }, + { + "epoch": 0.43, + "learning_rate": 1.85774705339316e-05, + "loss": 1.1252, + "step": 6730 + }, + { + "epoch": 0.43, + "learning_rate": 1.857725584490865e-05, + "loss": 0.9847, + "step": 6731 + }, + { + "epoch": 0.43, + "learning_rate": 1.85770411558857e-05, + "loss": 1.1612, + "step": 6732 + }, + { + "epoch": 0.43, + "learning_rate": 1.857682646686275e-05, + "loss": 1.1267, + "step": 6733 + }, + { + "epoch": 0.43, + "learning_rate": 1.8576611777839803e-05, + "loss": 1.1009, + "step": 6734 + }, + { + "epoch": 0.43, + "learning_rate": 1.857639708881685e-05, + "loss": 1.1662, + "step": 6735 + }, + { + "epoch": 0.43, + "learning_rate": 1.85761823997939e-05, + "loss": 1.1026, + "step": 6736 + }, + { + "epoch": 0.43, + "learning_rate": 1.857596771077095e-05, + "loss": 1.0866, + "step": 6737 + }, + { + "epoch": 0.43, + "learning_rate": 1.8575753021748e-05, + "loss": 1.1115, + "step": 6738 + }, + { + "epoch": 0.43, + "learning_rate": 1.857553833272505e-05, + "loss": 1.135, + "step": 6739 + }, + { + "epoch": 0.43, + "learning_rate": 1.85753236437021e-05, + "loss": 1.1286, + "step": 6740 + }, + { + "epoch": 0.43, + "learning_rate": 1.857510895467915e-05, + "loss": 1.1072, + "step": 6741 + }, + { + "epoch": 0.43, + "learning_rate": 1.8574894265656198e-05, + "loss": 1.1415, + "step": 6742 + }, + { + "epoch": 0.43, + "learning_rate": 1.8574679576633247e-05, + "loss": 1.0937, + "step": 6743 + }, + { + "epoch": 0.43, + "learning_rate": 1.8574464887610296e-05, + "loss": 1.1538, + "step": 6744 + }, + { + "epoch": 0.43, + "learning_rate": 1.8574250198587348e-05, + "loss": 1.1371, + "step": 6745 + }, + { + "epoch": 0.43, + "learning_rate": 1.8574035509564397e-05, + "loss": 1.1036, + "step": 6746 + }, + { + "epoch": 0.43, + "learning_rate": 1.857382082054145e-05, + "loss": 1.0967, + "step": 6747 + }, + { + "epoch": 0.43, + "learning_rate": 1.85736061315185e-05, + "loss": 1.1103, + "step": 6748 + }, + { + "epoch": 0.43, + "learning_rate": 1.8573391442495547e-05, + "loss": 1.0724, + "step": 6749 + }, + { + "epoch": 0.43, + "learning_rate": 1.8573176753472596e-05, + "loss": 1.0765, + "step": 6750 + }, + { + "epoch": 0.43, + "learning_rate": 1.8572962064449645e-05, + "loss": 1.1201, + "step": 6751 + }, + { + "epoch": 0.43, + "learning_rate": 1.8572747375426698e-05, + "loss": 1.2085, + "step": 6752 + }, + { + "epoch": 0.43, + "learning_rate": 1.8572532686403747e-05, + "loss": 1.0203, + "step": 6753 + }, + { + "epoch": 0.43, + "learning_rate": 1.8572317997380796e-05, + "loss": 1.0812, + "step": 6754 + }, + { + "epoch": 0.43, + "learning_rate": 1.8572103308357844e-05, + "loss": 1.0932, + "step": 6755 + }, + { + "epoch": 0.43, + "learning_rate": 1.8571888619334893e-05, + "loss": 1.1436, + "step": 6756 + }, + { + "epoch": 0.43, + "learning_rate": 1.8571673930311946e-05, + "loss": 1.1043, + "step": 6757 + }, + { + "epoch": 0.43, + "learning_rate": 1.8571459241288995e-05, + "loss": 1.0858, + "step": 6758 + }, + { + "epoch": 0.43, + "learning_rate": 1.8571244552266044e-05, + "loss": 1.0608, + "step": 6759 + }, + { + "epoch": 0.43, + "learning_rate": 1.8571029863243093e-05, + "loss": 1.0143, + "step": 6760 + }, + { + "epoch": 0.43, + "learning_rate": 1.857081517422014e-05, + "loss": 1.0841, + "step": 6761 + }, + { + "epoch": 0.44, + "learning_rate": 1.8570600485197194e-05, + "loss": 1.0417, + "step": 6762 + }, + { + "epoch": 0.44, + "learning_rate": 1.8570385796174243e-05, + "loss": 1.0966, + "step": 6763 + }, + { + "epoch": 0.44, + "learning_rate": 1.8570171107151292e-05, + "loss": 1.0641, + "step": 6764 + }, + { + "epoch": 0.44, + "learning_rate": 1.8569956418128344e-05, + "loss": 1.1901, + "step": 6765 + }, + { + "epoch": 0.44, + "learning_rate": 1.8569741729105393e-05, + "loss": 1.1315, + "step": 6766 + }, + { + "epoch": 0.44, + "learning_rate": 1.8569527040082442e-05, + "loss": 1.0644, + "step": 6767 + }, + { + "epoch": 0.44, + "learning_rate": 1.856931235105949e-05, + "loss": 1.0704, + "step": 6768 + }, + { + "epoch": 0.44, + "learning_rate": 1.856909766203654e-05, + "loss": 1.1016, + "step": 6769 + }, + { + "epoch": 0.44, + "learning_rate": 1.8568882973013593e-05, + "loss": 1.0867, + "step": 6770 + }, + { + "epoch": 0.44, + "learning_rate": 1.856866828399064e-05, + "loss": 1.1085, + "step": 6771 + }, + { + "epoch": 0.44, + "learning_rate": 1.856845359496769e-05, + "loss": 1.1598, + "step": 6772 + }, + { + "epoch": 0.44, + "learning_rate": 1.856823890594474e-05, + "loss": 1.0519, + "step": 6773 + }, + { + "epoch": 0.44, + "learning_rate": 1.856802421692179e-05, + "loss": 1.0978, + "step": 6774 + }, + { + "epoch": 0.44, + "learning_rate": 1.856780952789884e-05, + "loss": 0.9799, + "step": 6775 + }, + { + "epoch": 0.44, + "learning_rate": 1.856759483887589e-05, + "loss": 1.0543, + "step": 6776 + }, + { + "epoch": 0.44, + "learning_rate": 1.8567380149852942e-05, + "loss": 1.154, + "step": 6777 + }, + { + "epoch": 0.44, + "learning_rate": 1.856716546082999e-05, + "loss": 1.1064, + "step": 6778 + }, + { + "epoch": 0.44, + "learning_rate": 1.856695077180704e-05, + "loss": 1.1092, + "step": 6779 + }, + { + "epoch": 0.44, + "learning_rate": 1.856673608278409e-05, + "loss": 1.1054, + "step": 6780 + }, + { + "epoch": 0.44, + "learning_rate": 1.8566521393761138e-05, + "loss": 1.1374, + "step": 6781 + }, + { + "epoch": 0.44, + "learning_rate": 1.8566306704738187e-05, + "loss": 1.0752, + "step": 6782 + }, + { + "epoch": 0.44, + "learning_rate": 1.856609201571524e-05, + "loss": 1.1248, + "step": 6783 + }, + { + "epoch": 0.44, + "learning_rate": 1.8565877326692288e-05, + "loss": 1.0796, + "step": 6784 + }, + { + "epoch": 0.44, + "learning_rate": 1.8565662637669337e-05, + "loss": 1.119, + "step": 6785 + }, + { + "epoch": 0.44, + "learning_rate": 1.8565447948646386e-05, + "loss": 1.0678, + "step": 6786 + }, + { + "epoch": 0.44, + "learning_rate": 1.8565233259623435e-05, + "loss": 1.1416, + "step": 6787 + }, + { + "epoch": 0.44, + "learning_rate": 1.8565018570600487e-05, + "loss": 1.166, + "step": 6788 + }, + { + "epoch": 0.44, + "learning_rate": 1.8564803881577536e-05, + "loss": 1.1618, + "step": 6789 + }, + { + "epoch": 0.44, + "learning_rate": 1.8564589192554585e-05, + "loss": 0.9866, + "step": 6790 + }, + { + "epoch": 0.44, + "learning_rate": 1.8564374503531638e-05, + "loss": 1.0598, + "step": 6791 + }, + { + "epoch": 0.44, + "learning_rate": 1.8564159814508687e-05, + "loss": 1.0947, + "step": 6792 + }, + { + "epoch": 0.44, + "learning_rate": 1.8563945125485736e-05, + "loss": 1.1271, + "step": 6793 + }, + { + "epoch": 0.44, + "learning_rate": 1.8563730436462785e-05, + "loss": 1.0923, + "step": 6794 + }, + { + "epoch": 0.44, + "learning_rate": 1.8563515747439837e-05, + "loss": 1.1563, + "step": 6795 + }, + { + "epoch": 0.44, + "learning_rate": 1.8563301058416886e-05, + "loss": 1.1341, + "step": 6796 + }, + { + "epoch": 0.44, + "learning_rate": 1.8563086369393935e-05, + "loss": 1.0649, + "step": 6797 + }, + { + "epoch": 0.44, + "learning_rate": 1.8562871680370984e-05, + "loss": 1.0589, + "step": 6798 + }, + { + "epoch": 0.44, + "learning_rate": 1.8562656991348033e-05, + "loss": 1.1018, + "step": 6799 + }, + { + "epoch": 0.44, + "learning_rate": 1.8562442302325082e-05, + "loss": 1.1278, + "step": 6800 + }, + { + "epoch": 0.44, + "learning_rate": 1.8562227613302134e-05, + "loss": 1.1104, + "step": 6801 + }, + { + "epoch": 0.44, + "learning_rate": 1.8562012924279183e-05, + "loss": 1.0446, + "step": 6802 + }, + { + "epoch": 0.44, + "learning_rate": 1.8561798235256232e-05, + "loss": 1.1481, + "step": 6803 + }, + { + "epoch": 0.44, + "learning_rate": 1.856158354623328e-05, + "loss": 1.0386, + "step": 6804 + }, + { + "epoch": 0.44, + "learning_rate": 1.8561368857210333e-05, + "loss": 1.0432, + "step": 6805 + }, + { + "epoch": 0.44, + "learning_rate": 1.8561154168187382e-05, + "loss": 1.132, + "step": 6806 + }, + { + "epoch": 0.44, + "learning_rate": 1.856093947916443e-05, + "loss": 1.0989, + "step": 6807 + }, + { + "epoch": 0.44, + "learning_rate": 1.8560724790141484e-05, + "loss": 1.1073, + "step": 6808 + }, + { + "epoch": 0.44, + "learning_rate": 1.8560510101118533e-05, + "loss": 1.0895, + "step": 6809 + }, + { + "epoch": 0.44, + "learning_rate": 1.856029541209558e-05, + "loss": 1.1582, + "step": 6810 + }, + { + "epoch": 0.44, + "learning_rate": 1.856008072307263e-05, + "loss": 1.1839, + "step": 6811 + }, + { + "epoch": 0.44, + "learning_rate": 1.855986603404968e-05, + "loss": 1.1715, + "step": 6812 + }, + { + "epoch": 0.44, + "learning_rate": 1.855965134502673e-05, + "loss": 1.111, + "step": 6813 + }, + { + "epoch": 0.44, + "learning_rate": 1.855943665600378e-05, + "loss": 1.1221, + "step": 6814 + }, + { + "epoch": 0.44, + "learning_rate": 1.855922196698083e-05, + "loss": 1.1986, + "step": 6815 + }, + { + "epoch": 0.44, + "learning_rate": 1.855900727795788e-05, + "loss": 1.0662, + "step": 6816 + }, + { + "epoch": 0.44, + "learning_rate": 1.8558792588934928e-05, + "loss": 1.0454, + "step": 6817 + }, + { + "epoch": 0.44, + "learning_rate": 1.8558577899911977e-05, + "loss": 1.1496, + "step": 6818 + }, + { + "epoch": 0.44, + "learning_rate": 1.855836321088903e-05, + "loss": 1.0072, + "step": 6819 + }, + { + "epoch": 0.44, + "learning_rate": 1.8558148521866078e-05, + "loss": 1.1401, + "step": 6820 + }, + { + "epoch": 0.44, + "learning_rate": 1.855793383284313e-05, + "loss": 1.1251, + "step": 6821 + }, + { + "epoch": 0.44, + "learning_rate": 1.855771914382018e-05, + "loss": 1.0612, + "step": 6822 + }, + { + "epoch": 0.44, + "learning_rate": 1.855750445479723e-05, + "loss": 1.0708, + "step": 6823 + }, + { + "epoch": 0.44, + "learning_rate": 1.8557289765774277e-05, + "loss": 1.054, + "step": 6824 + }, + { + "epoch": 0.44, + "learning_rate": 1.8557075076751326e-05, + "loss": 1.2213, + "step": 6825 + }, + { + "epoch": 0.44, + "learning_rate": 1.855686038772838e-05, + "loss": 1.108, + "step": 6826 + }, + { + "epoch": 0.44, + "learning_rate": 1.8556645698705428e-05, + "loss": 1.0768, + "step": 6827 + }, + { + "epoch": 0.44, + "learning_rate": 1.8556431009682477e-05, + "loss": 1.1253, + "step": 6828 + }, + { + "epoch": 0.44, + "learning_rate": 1.8556216320659525e-05, + "loss": 1.159, + "step": 6829 + }, + { + "epoch": 0.44, + "learning_rate": 1.8556001631636574e-05, + "loss": 1.0942, + "step": 6830 + }, + { + "epoch": 0.44, + "learning_rate": 1.8555786942613623e-05, + "loss": 1.0518, + "step": 6831 + }, + { + "epoch": 0.44, + "learning_rate": 1.8555572253590676e-05, + "loss": 1.0645, + "step": 6832 + }, + { + "epoch": 0.44, + "learning_rate": 1.8555357564567725e-05, + "loss": 1.0662, + "step": 6833 + }, + { + "epoch": 0.44, + "learning_rate": 1.8555142875544777e-05, + "loss": 1.1008, + "step": 6834 + }, + { + "epoch": 0.44, + "learning_rate": 1.8554928186521826e-05, + "loss": 1.1046, + "step": 6835 + }, + { + "epoch": 0.44, + "learning_rate": 1.8554713497498875e-05, + "loss": 1.2652, + "step": 6836 + }, + { + "epoch": 0.44, + "learning_rate": 1.8554498808475924e-05, + "loss": 1.0457, + "step": 6837 + }, + { + "epoch": 0.44, + "learning_rate": 1.8554284119452973e-05, + "loss": 1.0096, + "step": 6838 + }, + { + "epoch": 0.44, + "learning_rate": 1.8554069430430025e-05, + "loss": 1.133, + "step": 6839 + }, + { + "epoch": 0.44, + "learning_rate": 1.8553854741407074e-05, + "loss": 1.0372, + "step": 6840 + }, + { + "epoch": 0.44, + "learning_rate": 1.8553640052384123e-05, + "loss": 1.1287, + "step": 6841 + }, + { + "epoch": 0.44, + "learning_rate": 1.8553425363361172e-05, + "loss": 1.0706, + "step": 6842 + }, + { + "epoch": 0.44, + "learning_rate": 1.855321067433822e-05, + "loss": 1.1427, + "step": 6843 + }, + { + "epoch": 0.44, + "learning_rate": 1.8552995985315274e-05, + "loss": 1.0302, + "step": 6844 + }, + { + "epoch": 0.44, + "learning_rate": 1.8552781296292322e-05, + "loss": 1.0771, + "step": 6845 + }, + { + "epoch": 0.44, + "learning_rate": 1.855256660726937e-05, + "loss": 1.1971, + "step": 6846 + }, + { + "epoch": 0.44, + "learning_rate": 1.855235191824642e-05, + "loss": 1.0545, + "step": 6847 + }, + { + "epoch": 0.44, + "learning_rate": 1.855213722922347e-05, + "loss": 1.077, + "step": 6848 + }, + { + "epoch": 0.44, + "learning_rate": 1.8551922540200522e-05, + "loss": 1.1455, + "step": 6849 + }, + { + "epoch": 0.44, + "learning_rate": 1.855170785117757e-05, + "loss": 1.161, + "step": 6850 + }, + { + "epoch": 0.44, + "learning_rate": 1.855149316215462e-05, + "loss": 1.0823, + "step": 6851 + }, + { + "epoch": 0.44, + "learning_rate": 1.8551278473131672e-05, + "loss": 1.1241, + "step": 6852 + }, + { + "epoch": 0.44, + "learning_rate": 1.855106378410872e-05, + "loss": 1.0802, + "step": 6853 + }, + { + "epoch": 0.44, + "learning_rate": 1.855084909508577e-05, + "loss": 1.0424, + "step": 6854 + }, + { + "epoch": 0.44, + "learning_rate": 1.855063440606282e-05, + "loss": 1.0756, + "step": 6855 + }, + { + "epoch": 0.44, + "learning_rate": 1.8550419717039868e-05, + "loss": 1.1102, + "step": 6856 + }, + { + "epoch": 0.44, + "learning_rate": 1.855020502801692e-05, + "loss": 1.0706, + "step": 6857 + }, + { + "epoch": 0.44, + "learning_rate": 1.854999033899397e-05, + "loss": 0.9889, + "step": 6858 + }, + { + "epoch": 0.44, + "learning_rate": 1.8549775649971018e-05, + "loss": 1.0447, + "step": 6859 + }, + { + "epoch": 0.44, + "learning_rate": 1.8549560960948067e-05, + "loss": 1.137, + "step": 6860 + }, + { + "epoch": 0.44, + "learning_rate": 1.8549346271925116e-05, + "loss": 1.0603, + "step": 6861 + }, + { + "epoch": 0.44, + "learning_rate": 1.854913158290217e-05, + "loss": 1.1639, + "step": 6862 + }, + { + "epoch": 0.44, + "learning_rate": 1.8548916893879217e-05, + "loss": 1.064, + "step": 6863 + }, + { + "epoch": 0.44, + "learning_rate": 1.854870220485627e-05, + "loss": 1.1967, + "step": 6864 + }, + { + "epoch": 0.44, + "learning_rate": 1.854848751583332e-05, + "loss": 1.0942, + "step": 6865 + }, + { + "epoch": 0.44, + "learning_rate": 1.8548272826810368e-05, + "loss": 1.0776, + "step": 6866 + }, + { + "epoch": 0.44, + "learning_rate": 1.8548058137787417e-05, + "loss": 1.1291, + "step": 6867 + }, + { + "epoch": 0.44, + "learning_rate": 1.8547843448764466e-05, + "loss": 1.1282, + "step": 6868 + }, + { + "epoch": 0.44, + "learning_rate": 1.8547628759741515e-05, + "loss": 1.0641, + "step": 6869 + }, + { + "epoch": 0.44, + "learning_rate": 1.8547414070718567e-05, + "loss": 1.0982, + "step": 6870 + }, + { + "epoch": 0.44, + "learning_rate": 1.8547199381695616e-05, + "loss": 1.1658, + "step": 6871 + }, + { + "epoch": 0.44, + "learning_rate": 1.8546984692672665e-05, + "loss": 1.1008, + "step": 6872 + }, + { + "epoch": 0.44, + "learning_rate": 1.8546770003649714e-05, + "loss": 1.0737, + "step": 6873 + }, + { + "epoch": 0.44, + "learning_rate": 1.8546555314626763e-05, + "loss": 1.175, + "step": 6874 + }, + { + "epoch": 0.44, + "learning_rate": 1.8546340625603815e-05, + "loss": 1.1118, + "step": 6875 + }, + { + "epoch": 0.44, + "learning_rate": 1.8546125936580864e-05, + "loss": 1.0882, + "step": 6876 + }, + { + "epoch": 0.44, + "learning_rate": 1.8545911247557913e-05, + "loss": 1.08, + "step": 6877 + }, + { + "epoch": 0.44, + "learning_rate": 1.8545696558534965e-05, + "loss": 1.0722, + "step": 6878 + }, + { + "epoch": 0.44, + "learning_rate": 1.8545481869512014e-05, + "loss": 1.1063, + "step": 6879 + }, + { + "epoch": 0.44, + "learning_rate": 1.8545267180489063e-05, + "loss": 1.0954, + "step": 6880 + }, + { + "epoch": 0.44, + "learning_rate": 1.8545052491466112e-05, + "loss": 1.0574, + "step": 6881 + }, + { + "epoch": 0.44, + "learning_rate": 1.8544837802443165e-05, + "loss": 1.0753, + "step": 6882 + }, + { + "epoch": 0.44, + "learning_rate": 1.8544623113420214e-05, + "loss": 1.1249, + "step": 6883 + }, + { + "epoch": 0.44, + "learning_rate": 1.8544408424397263e-05, + "loss": 1.1627, + "step": 6884 + }, + { + "epoch": 0.44, + "learning_rate": 1.854419373537431e-05, + "loss": 1.1921, + "step": 6885 + }, + { + "epoch": 0.44, + "learning_rate": 1.854397904635136e-05, + "loss": 1.1034, + "step": 6886 + }, + { + "epoch": 0.44, + "learning_rate": 1.854376435732841e-05, + "loss": 1.1758, + "step": 6887 + }, + { + "epoch": 0.44, + "learning_rate": 1.8543549668305462e-05, + "loss": 1.0965, + "step": 6888 + }, + { + "epoch": 0.44, + "learning_rate": 1.854333497928251e-05, + "loss": 1.1, + "step": 6889 + }, + { + "epoch": 0.44, + "learning_rate": 1.854312029025956e-05, + "loss": 1.0779, + "step": 6890 + }, + { + "epoch": 0.44, + "learning_rate": 1.854290560123661e-05, + "loss": 1.1316, + "step": 6891 + }, + { + "epoch": 0.44, + "learning_rate": 1.854269091221366e-05, + "loss": 1.0469, + "step": 6892 + }, + { + "epoch": 0.44, + "learning_rate": 1.854247622319071e-05, + "loss": 1.0939, + "step": 6893 + }, + { + "epoch": 0.44, + "learning_rate": 1.854226153416776e-05, + "loss": 1.0232, + "step": 6894 + }, + { + "epoch": 0.44, + "learning_rate": 1.854204684514481e-05, + "loss": 1.0914, + "step": 6895 + }, + { + "epoch": 0.44, + "learning_rate": 1.854183215612186e-05, + "loss": 1.1235, + "step": 6896 + }, + { + "epoch": 0.44, + "learning_rate": 1.854161746709891e-05, + "loss": 1.0556, + "step": 6897 + }, + { + "epoch": 0.44, + "learning_rate": 1.8541402778075958e-05, + "loss": 1.1404, + "step": 6898 + }, + { + "epoch": 0.44, + "learning_rate": 1.8541188089053007e-05, + "loss": 1.0729, + "step": 6899 + }, + { + "epoch": 0.44, + "learning_rate": 1.8540973400030056e-05, + "loss": 1.1231, + "step": 6900 + }, + { + "epoch": 0.44, + "learning_rate": 1.854075871100711e-05, + "loss": 1.1098, + "step": 6901 + }, + { + "epoch": 0.44, + "learning_rate": 1.8540544021984158e-05, + "loss": 1.0585, + "step": 6902 + }, + { + "epoch": 0.44, + "learning_rate": 1.8540329332961206e-05, + "loss": 1.1355, + "step": 6903 + }, + { + "epoch": 0.44, + "learning_rate": 1.8540114643938255e-05, + "loss": 1.0336, + "step": 6904 + }, + { + "epoch": 0.44, + "learning_rate": 1.8539899954915304e-05, + "loss": 1.0932, + "step": 6905 + }, + { + "epoch": 0.44, + "learning_rate": 1.8539685265892357e-05, + "loss": 1.123, + "step": 6906 + }, + { + "epoch": 0.44, + "learning_rate": 1.8539470576869406e-05, + "loss": 1.1505, + "step": 6907 + }, + { + "epoch": 0.44, + "learning_rate": 1.8539255887846458e-05, + "loss": 1.1071, + "step": 6908 + }, + { + "epoch": 0.44, + "learning_rate": 1.8539041198823507e-05, + "loss": 1.06, + "step": 6909 + }, + { + "epoch": 0.44, + "learning_rate": 1.8538826509800556e-05, + "loss": 1.0466, + "step": 6910 + }, + { + "epoch": 0.44, + "learning_rate": 1.8538611820777605e-05, + "loss": 1.1219, + "step": 6911 + }, + { + "epoch": 0.44, + "learning_rate": 1.8538397131754654e-05, + "loss": 1.0472, + "step": 6912 + }, + { + "epoch": 0.44, + "learning_rate": 1.8538182442731706e-05, + "loss": 1.0059, + "step": 6913 + }, + { + "epoch": 0.44, + "learning_rate": 1.8537967753708755e-05, + "loss": 1.0174, + "step": 6914 + }, + { + "epoch": 0.44, + "learning_rate": 1.8537753064685804e-05, + "loss": 1.0643, + "step": 6915 + }, + { + "epoch": 0.44, + "learning_rate": 1.8537538375662853e-05, + "loss": 1.0822, + "step": 6916 + }, + { + "epoch": 0.45, + "learning_rate": 1.8537323686639902e-05, + "loss": 1.0547, + "step": 6917 + }, + { + "epoch": 0.45, + "learning_rate": 1.853710899761695e-05, + "loss": 1.1846, + "step": 6918 + }, + { + "epoch": 0.45, + "learning_rate": 1.8536894308594003e-05, + "loss": 1.0959, + "step": 6919 + }, + { + "epoch": 0.45, + "learning_rate": 1.8536679619571052e-05, + "loss": 1.1239, + "step": 6920 + }, + { + "epoch": 0.45, + "learning_rate": 1.8536464930548105e-05, + "loss": 1.0525, + "step": 6921 + }, + { + "epoch": 0.45, + "learning_rate": 1.8536250241525154e-05, + "loss": 1.0885, + "step": 6922 + }, + { + "epoch": 0.45, + "learning_rate": 1.8536035552502203e-05, + "loss": 1.0518, + "step": 6923 + }, + { + "epoch": 0.45, + "learning_rate": 1.8535820863479252e-05, + "loss": 1.094, + "step": 6924 + }, + { + "epoch": 0.45, + "learning_rate": 1.85356061744563e-05, + "loss": 1.1296, + "step": 6925 + }, + { + "epoch": 0.45, + "learning_rate": 1.8535391485433353e-05, + "loss": 1.011, + "step": 6926 + }, + { + "epoch": 0.45, + "learning_rate": 1.8535176796410402e-05, + "loss": 1.0869, + "step": 6927 + }, + { + "epoch": 0.45, + "learning_rate": 1.853496210738745e-05, + "loss": 1.2315, + "step": 6928 + }, + { + "epoch": 0.45, + "learning_rate": 1.85347474183645e-05, + "loss": 1.1277, + "step": 6929 + }, + { + "epoch": 0.45, + "learning_rate": 1.853453272934155e-05, + "loss": 1.118, + "step": 6930 + }, + { + "epoch": 0.45, + "learning_rate": 1.85343180403186e-05, + "loss": 1.0598, + "step": 6931 + }, + { + "epoch": 0.45, + "learning_rate": 1.853410335129565e-05, + "loss": 1.134, + "step": 6932 + }, + { + "epoch": 0.45, + "learning_rate": 1.85338886622727e-05, + "loss": 1.1795, + "step": 6933 + }, + { + "epoch": 0.45, + "learning_rate": 1.8533673973249748e-05, + "loss": 1.0471, + "step": 6934 + }, + { + "epoch": 0.45, + "learning_rate": 1.8533459284226797e-05, + "loss": 1.1523, + "step": 6935 + }, + { + "epoch": 0.45, + "learning_rate": 1.853324459520385e-05, + "loss": 1.1274, + "step": 6936 + }, + { + "epoch": 0.45, + "learning_rate": 1.85330299061809e-05, + "loss": 1.1072, + "step": 6937 + }, + { + "epoch": 0.45, + "learning_rate": 1.8532815217157947e-05, + "loss": 1.1158, + "step": 6938 + }, + { + "epoch": 0.45, + "learning_rate": 1.8532600528135e-05, + "loss": 1.1393, + "step": 6939 + }, + { + "epoch": 0.45, + "learning_rate": 1.853238583911205e-05, + "loss": 1.1047, + "step": 6940 + }, + { + "epoch": 0.45, + "learning_rate": 1.8532171150089098e-05, + "loss": 1.0928, + "step": 6941 + }, + { + "epoch": 0.45, + "learning_rate": 1.8531956461066147e-05, + "loss": 1.1118, + "step": 6942 + }, + { + "epoch": 0.45, + "learning_rate": 1.8531741772043196e-05, + "loss": 1.1015, + "step": 6943 + }, + { + "epoch": 0.45, + "learning_rate": 1.8531527083020248e-05, + "loss": 1.038, + "step": 6944 + }, + { + "epoch": 0.45, + "learning_rate": 1.8531312393997297e-05, + "loss": 0.9865, + "step": 6945 + }, + { + "epoch": 0.45, + "learning_rate": 1.8531097704974346e-05, + "loss": 1.1297, + "step": 6946 + }, + { + "epoch": 0.45, + "learning_rate": 1.8530883015951395e-05, + "loss": 1.0608, + "step": 6947 + }, + { + "epoch": 0.45, + "learning_rate": 1.8530668326928444e-05, + "loss": 1.1436, + "step": 6948 + }, + { + "epoch": 0.45, + "learning_rate": 1.8530453637905493e-05, + "loss": 1.1162, + "step": 6949 + }, + { + "epoch": 0.45, + "learning_rate": 1.8530238948882545e-05, + "loss": 1.0649, + "step": 6950 + }, + { + "epoch": 0.45, + "learning_rate": 1.8530024259859597e-05, + "loss": 1.0508, + "step": 6951 + }, + { + "epoch": 0.45, + "learning_rate": 1.8529809570836646e-05, + "loss": 1.0911, + "step": 6952 + }, + { + "epoch": 0.45, + "learning_rate": 1.8529594881813695e-05, + "loss": 1.1062, + "step": 6953 + }, + { + "epoch": 0.45, + "learning_rate": 1.8529380192790744e-05, + "loss": 1.1264, + "step": 6954 + }, + { + "epoch": 0.45, + "learning_rate": 1.8529165503767793e-05, + "loss": 1.1327, + "step": 6955 + }, + { + "epoch": 0.45, + "learning_rate": 1.8528950814744842e-05, + "loss": 1.0471, + "step": 6956 + }, + { + "epoch": 0.45, + "learning_rate": 1.8528736125721895e-05, + "loss": 1.1318, + "step": 6957 + }, + { + "epoch": 0.45, + "learning_rate": 1.8528521436698944e-05, + "loss": 1.0482, + "step": 6958 + }, + { + "epoch": 0.45, + "learning_rate": 1.8528306747675993e-05, + "loss": 1.1148, + "step": 6959 + }, + { + "epoch": 0.45, + "learning_rate": 1.852809205865304e-05, + "loss": 1.0355, + "step": 6960 + }, + { + "epoch": 0.45, + "learning_rate": 1.852787736963009e-05, + "loss": 1.0918, + "step": 6961 + }, + { + "epoch": 0.45, + "learning_rate": 1.8527662680607143e-05, + "loss": 1.1793, + "step": 6962 + }, + { + "epoch": 0.45, + "learning_rate": 1.8527447991584192e-05, + "loss": 1.099, + "step": 6963 + }, + { + "epoch": 0.45, + "learning_rate": 1.852723330256124e-05, + "loss": 1.0455, + "step": 6964 + }, + { + "epoch": 0.45, + "learning_rate": 1.8527018613538293e-05, + "loss": 1.1359, + "step": 6965 + }, + { + "epoch": 0.45, + "learning_rate": 1.8526803924515342e-05, + "loss": 1.137, + "step": 6966 + }, + { + "epoch": 0.45, + "learning_rate": 1.852658923549239e-05, + "loss": 1.1818, + "step": 6967 + }, + { + "epoch": 0.45, + "learning_rate": 1.852637454646944e-05, + "loss": 1.0564, + "step": 6968 + }, + { + "epoch": 0.45, + "learning_rate": 1.8526159857446492e-05, + "loss": 1.085, + "step": 6969 + }, + { + "epoch": 0.45, + "learning_rate": 1.852594516842354e-05, + "loss": 1.1376, + "step": 6970 + }, + { + "epoch": 0.45, + "learning_rate": 1.852573047940059e-05, + "loss": 1.1401, + "step": 6971 + }, + { + "epoch": 0.45, + "learning_rate": 1.852551579037764e-05, + "loss": 1.1012, + "step": 6972 + }, + { + "epoch": 0.45, + "learning_rate": 1.8525301101354688e-05, + "loss": 1.08, + "step": 6973 + }, + { + "epoch": 0.45, + "learning_rate": 1.8525086412331737e-05, + "loss": 0.9661, + "step": 6974 + }, + { + "epoch": 0.45, + "learning_rate": 1.852487172330879e-05, + "loss": 0.9948, + "step": 6975 + }, + { + "epoch": 0.45, + "learning_rate": 1.852465703428584e-05, + "loss": 1.1064, + "step": 6976 + }, + { + "epoch": 0.45, + "learning_rate": 1.8524442345262887e-05, + "loss": 1.1239, + "step": 6977 + }, + { + "epoch": 0.45, + "learning_rate": 1.8524227656239936e-05, + "loss": 1.1004, + "step": 6978 + }, + { + "epoch": 0.45, + "learning_rate": 1.8524012967216985e-05, + "loss": 1.1646, + "step": 6979 + }, + { + "epoch": 0.45, + "learning_rate": 1.8523798278194038e-05, + "loss": 1.1151, + "step": 6980 + }, + { + "epoch": 0.45, + "learning_rate": 1.8523583589171087e-05, + "loss": 1.126, + "step": 6981 + }, + { + "epoch": 0.45, + "learning_rate": 1.852336890014814e-05, + "loss": 1.0478, + "step": 6982 + }, + { + "epoch": 0.45, + "learning_rate": 1.8523154211125188e-05, + "loss": 1.144, + "step": 6983 + }, + { + "epoch": 0.45, + "learning_rate": 1.8522939522102237e-05, + "loss": 1.1685, + "step": 6984 + }, + { + "epoch": 0.45, + "learning_rate": 1.8522724833079286e-05, + "loss": 1.0387, + "step": 6985 + }, + { + "epoch": 0.45, + "learning_rate": 1.8522510144056335e-05, + "loss": 1.0571, + "step": 6986 + }, + { + "epoch": 0.45, + "learning_rate": 1.8522295455033384e-05, + "loss": 1.0978, + "step": 6987 + }, + { + "epoch": 0.45, + "learning_rate": 1.8522080766010436e-05, + "loss": 1.1485, + "step": 6988 + }, + { + "epoch": 0.45, + "learning_rate": 1.8521866076987485e-05, + "loss": 1.0417, + "step": 6989 + }, + { + "epoch": 0.45, + "learning_rate": 1.8521651387964534e-05, + "loss": 1.0412, + "step": 6990 + }, + { + "epoch": 0.45, + "learning_rate": 1.8521436698941583e-05, + "loss": 1.0477, + "step": 6991 + }, + { + "epoch": 0.45, + "learning_rate": 1.8521222009918632e-05, + "loss": 1.0862, + "step": 6992 + }, + { + "epoch": 0.45, + "learning_rate": 1.8521007320895684e-05, + "loss": 1.101, + "step": 6993 + }, + { + "epoch": 0.45, + "learning_rate": 1.8520792631872733e-05, + "loss": 1.0936, + "step": 6994 + }, + { + "epoch": 0.45, + "learning_rate": 1.8520577942849786e-05, + "loss": 1.0945, + "step": 6995 + }, + { + "epoch": 0.45, + "learning_rate": 1.8520363253826835e-05, + "loss": 1.109, + "step": 6996 + }, + { + "epoch": 0.45, + "learning_rate": 1.8520148564803884e-05, + "loss": 1.0961, + "step": 6997 + }, + { + "epoch": 0.45, + "learning_rate": 1.8519933875780933e-05, + "loss": 1.0396, + "step": 6998 + }, + { + "epoch": 0.45, + "learning_rate": 1.851971918675798e-05, + "loss": 1.1309, + "step": 6999 + }, + { + "epoch": 0.45, + "learning_rate": 1.8519504497735034e-05, + "loss": 1.0654, + "step": 7000 + }, + { + "epoch": 0.45, + "eval_laion/cider": 0.5961587666882281, + "eval_laion/meteor": 0.39994827488723417, + "eval_laion/pred_num_words": 37.69140625, + "eval_laion/sacrebleu": 16.54413707677776, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.1071584224700928, + "eval_runtime": 100.7831, + "eval_samples_per_second": 4.951, + "eval_steps_per_second": 0.318, + "step": 7000 + }, + { + "epoch": 0.45, + "learning_rate": 1.8519289808712083e-05, + "loss": 1.1438, + "step": 7001 + }, + { + "epoch": 0.45, + "learning_rate": 1.8519075119689132e-05, + "loss": 1.1328, + "step": 7002 + }, + { + "epoch": 0.45, + "learning_rate": 1.851886043066618e-05, + "loss": 1.1722, + "step": 7003 + }, + { + "epoch": 0.45, + "learning_rate": 1.851864574164323e-05, + "loss": 1.0088, + "step": 7004 + }, + { + "epoch": 0.45, + "learning_rate": 1.851843105262028e-05, + "loss": 1.1287, + "step": 7005 + }, + { + "epoch": 0.45, + "learning_rate": 1.851821636359733e-05, + "loss": 1.1353, + "step": 7006 + }, + { + "epoch": 0.45, + "learning_rate": 1.851800167457438e-05, + "loss": 1.0873, + "step": 7007 + }, + { + "epoch": 0.45, + "learning_rate": 1.851778698555143e-05, + "loss": 0.9389, + "step": 7008 + }, + { + "epoch": 0.45, + "learning_rate": 1.851757229652848e-05, + "loss": 1.0922, + "step": 7009 + }, + { + "epoch": 0.45, + "learning_rate": 1.851735760750553e-05, + "loss": 1.1625, + "step": 7010 + }, + { + "epoch": 0.45, + "learning_rate": 1.851714291848258e-05, + "loss": 1.0965, + "step": 7011 + }, + { + "epoch": 0.45, + "learning_rate": 1.851692822945963e-05, + "loss": 1.0852, + "step": 7012 + }, + { + "epoch": 0.45, + "learning_rate": 1.851671354043668e-05, + "loss": 1.0959, + "step": 7013 + }, + { + "epoch": 0.45, + "learning_rate": 1.851649885141373e-05, + "loss": 1.1154, + "step": 7014 + }, + { + "epoch": 0.45, + "learning_rate": 1.851628416239078e-05, + "loss": 1.1367, + "step": 7015 + }, + { + "epoch": 0.45, + "learning_rate": 1.8516069473367828e-05, + "loss": 1.0692, + "step": 7016 + }, + { + "epoch": 0.45, + "learning_rate": 1.8515854784344877e-05, + "loss": 1.0659, + "step": 7017 + }, + { + "epoch": 0.45, + "learning_rate": 1.851564009532193e-05, + "loss": 1.0945, + "step": 7018 + }, + { + "epoch": 0.45, + "learning_rate": 1.8515425406298978e-05, + "loss": 1.035, + "step": 7019 + }, + { + "epoch": 0.45, + "learning_rate": 1.8515210717276027e-05, + "loss": 1.1315, + "step": 7020 + }, + { + "epoch": 0.45, + "learning_rate": 1.8514996028253076e-05, + "loss": 1.0248, + "step": 7021 + }, + { + "epoch": 0.45, + "learning_rate": 1.8514781339230125e-05, + "loss": 1.1471, + "step": 7022 + }, + { + "epoch": 0.45, + "learning_rate": 1.8514566650207177e-05, + "loss": 1.0958, + "step": 7023 + }, + { + "epoch": 0.45, + "learning_rate": 1.8514351961184226e-05, + "loss": 1.0478, + "step": 7024 + }, + { + "epoch": 0.45, + "learning_rate": 1.8514137272161275e-05, + "loss": 1.1337, + "step": 7025 + }, + { + "epoch": 0.45, + "learning_rate": 1.8513922583138327e-05, + "loss": 1.0308, + "step": 7026 + }, + { + "epoch": 0.45, + "learning_rate": 1.8513707894115376e-05, + "loss": 1.1954, + "step": 7027 + }, + { + "epoch": 0.45, + "learning_rate": 1.8513493205092425e-05, + "loss": 1.1067, + "step": 7028 + }, + { + "epoch": 0.45, + "learning_rate": 1.8513278516069474e-05, + "loss": 1.0567, + "step": 7029 + }, + { + "epoch": 0.45, + "learning_rate": 1.8513063827046523e-05, + "loss": 1.0377, + "step": 7030 + }, + { + "epoch": 0.45, + "learning_rate": 1.8512849138023576e-05, + "loss": 1.152, + "step": 7031 + }, + { + "epoch": 0.45, + "learning_rate": 1.8512634449000625e-05, + "loss": 1.1657, + "step": 7032 + }, + { + "epoch": 0.45, + "learning_rate": 1.8512419759977674e-05, + "loss": 1.0402, + "step": 7033 + }, + { + "epoch": 0.45, + "learning_rate": 1.8512205070954723e-05, + "loss": 1.0943, + "step": 7034 + }, + { + "epoch": 0.45, + "learning_rate": 1.851199038193177e-05, + "loss": 1.209, + "step": 7035 + }, + { + "epoch": 0.45, + "learning_rate": 1.851177569290882e-05, + "loss": 0.9962, + "step": 7036 + }, + { + "epoch": 0.45, + "learning_rate": 1.8511561003885873e-05, + "loss": 1.0698, + "step": 7037 + }, + { + "epoch": 0.45, + "learning_rate": 1.8511346314862925e-05, + "loss": 1.1267, + "step": 7038 + }, + { + "epoch": 0.45, + "learning_rate": 1.8511131625839974e-05, + "loss": 1.1089, + "step": 7039 + }, + { + "epoch": 0.45, + "learning_rate": 1.8510916936817023e-05, + "loss": 1.1316, + "step": 7040 + }, + { + "epoch": 0.45, + "learning_rate": 1.8510702247794072e-05, + "loss": 1.0001, + "step": 7041 + }, + { + "epoch": 0.45, + "learning_rate": 1.851048755877112e-05, + "loss": 1.1055, + "step": 7042 + }, + { + "epoch": 0.45, + "learning_rate": 1.851027286974817e-05, + "loss": 1.1051, + "step": 7043 + }, + { + "epoch": 0.45, + "learning_rate": 1.8510058180725222e-05, + "loss": 1.1145, + "step": 7044 + }, + { + "epoch": 0.45, + "learning_rate": 1.850984349170227e-05, + "loss": 1.0938, + "step": 7045 + }, + { + "epoch": 0.45, + "learning_rate": 1.850962880267932e-05, + "loss": 1.0245, + "step": 7046 + }, + { + "epoch": 0.45, + "learning_rate": 1.850941411365637e-05, + "loss": 1.0168, + "step": 7047 + }, + { + "epoch": 0.45, + "learning_rate": 1.8509199424633418e-05, + "loss": 1.0474, + "step": 7048 + }, + { + "epoch": 0.45, + "learning_rate": 1.850898473561047e-05, + "loss": 1.0713, + "step": 7049 + }, + { + "epoch": 0.45, + "learning_rate": 1.850877004658752e-05, + "loss": 1.052, + "step": 7050 + }, + { + "epoch": 0.45, + "learning_rate": 1.850855535756457e-05, + "loss": 1.1221, + "step": 7051 + }, + { + "epoch": 0.45, + "learning_rate": 1.850834066854162e-05, + "loss": 1.118, + "step": 7052 + }, + { + "epoch": 0.45, + "learning_rate": 1.850812597951867e-05, + "loss": 1.1433, + "step": 7053 + }, + { + "epoch": 0.45, + "learning_rate": 1.850791129049572e-05, + "loss": 1.1549, + "step": 7054 + }, + { + "epoch": 0.45, + "learning_rate": 1.8507696601472768e-05, + "loss": 1.1437, + "step": 7055 + }, + { + "epoch": 0.45, + "learning_rate": 1.8507481912449817e-05, + "loss": 1.0883, + "step": 7056 + }, + { + "epoch": 0.45, + "learning_rate": 1.850726722342687e-05, + "loss": 1.055, + "step": 7057 + }, + { + "epoch": 0.45, + "learning_rate": 1.8507052534403918e-05, + "loss": 1.2134, + "step": 7058 + }, + { + "epoch": 0.45, + "learning_rate": 1.8506837845380967e-05, + "loss": 1.0623, + "step": 7059 + }, + { + "epoch": 0.45, + "learning_rate": 1.8506623156358016e-05, + "loss": 1.1141, + "step": 7060 + }, + { + "epoch": 0.45, + "learning_rate": 1.8506408467335065e-05, + "loss": 1.1117, + "step": 7061 + }, + { + "epoch": 0.45, + "learning_rate": 1.8506193778312117e-05, + "loss": 1.0659, + "step": 7062 + }, + { + "epoch": 0.45, + "learning_rate": 1.8505979089289166e-05, + "loss": 1.1308, + "step": 7063 + }, + { + "epoch": 0.45, + "learning_rate": 1.8505764400266215e-05, + "loss": 1.0859, + "step": 7064 + }, + { + "epoch": 0.45, + "learning_rate": 1.8505549711243264e-05, + "loss": 1.0064, + "step": 7065 + }, + { + "epoch": 0.45, + "learning_rate": 1.8505335022220313e-05, + "loss": 1.0845, + "step": 7066 + }, + { + "epoch": 0.45, + "learning_rate": 1.8505120333197365e-05, + "loss": 1.1149, + "step": 7067 + }, + { + "epoch": 0.45, + "learning_rate": 1.8504905644174414e-05, + "loss": 0.995, + "step": 7068 + }, + { + "epoch": 0.45, + "learning_rate": 1.8504690955151467e-05, + "loss": 1.1874, + "step": 7069 + }, + { + "epoch": 0.45, + "learning_rate": 1.8504476266128516e-05, + "loss": 1.1355, + "step": 7070 + }, + { + "epoch": 0.45, + "learning_rate": 1.8504261577105565e-05, + "loss": 1.1464, + "step": 7071 + }, + { + "epoch": 0.45, + "learning_rate": 1.8504046888082614e-05, + "loss": 1.0979, + "step": 7072 + }, + { + "epoch": 0.46, + "learning_rate": 1.8503832199059663e-05, + "loss": 1.0894, + "step": 7073 + }, + { + "epoch": 0.46, + "learning_rate": 1.850361751003671e-05, + "loss": 1.0999, + "step": 7074 + }, + { + "epoch": 0.46, + "learning_rate": 1.8503402821013764e-05, + "loss": 1.0999, + "step": 7075 + }, + { + "epoch": 0.46, + "learning_rate": 1.8503188131990813e-05, + "loss": 1.1268, + "step": 7076 + }, + { + "epoch": 0.46, + "learning_rate": 1.8502973442967862e-05, + "loss": 1.1544, + "step": 7077 + }, + { + "epoch": 0.46, + "learning_rate": 1.850275875394491e-05, + "loss": 1.0979, + "step": 7078 + }, + { + "epoch": 0.46, + "learning_rate": 1.850254406492196e-05, + "loss": 1.0598, + "step": 7079 + }, + { + "epoch": 0.46, + "learning_rate": 1.8502329375899012e-05, + "loss": 1.0622, + "step": 7080 + }, + { + "epoch": 0.46, + "learning_rate": 1.850211468687606e-05, + "loss": 1.0936, + "step": 7081 + }, + { + "epoch": 0.46, + "learning_rate": 1.8501899997853114e-05, + "loss": 1.007, + "step": 7082 + }, + { + "epoch": 0.46, + "learning_rate": 1.8501685308830162e-05, + "loss": 1.0309, + "step": 7083 + }, + { + "epoch": 0.46, + "learning_rate": 1.850147061980721e-05, + "loss": 1.0429, + "step": 7084 + }, + { + "epoch": 0.46, + "learning_rate": 1.850125593078426e-05, + "loss": 1.1024, + "step": 7085 + }, + { + "epoch": 0.46, + "learning_rate": 1.850104124176131e-05, + "loss": 1.1573, + "step": 7086 + }, + { + "epoch": 0.46, + "learning_rate": 1.8500826552738362e-05, + "loss": 1.053, + "step": 7087 + }, + { + "epoch": 0.46, + "learning_rate": 1.850061186371541e-05, + "loss": 1.1074, + "step": 7088 + }, + { + "epoch": 0.46, + "learning_rate": 1.850039717469246e-05, + "loss": 1.1135, + "step": 7089 + }, + { + "epoch": 0.46, + "learning_rate": 1.850018248566951e-05, + "loss": 1.0607, + "step": 7090 + }, + { + "epoch": 0.46, + "learning_rate": 1.8499967796646558e-05, + "loss": 1.1251, + "step": 7091 + }, + { + "epoch": 0.46, + "learning_rate": 1.8499753107623607e-05, + "loss": 1.0388, + "step": 7092 + }, + { + "epoch": 0.46, + "learning_rate": 1.849953841860066e-05, + "loss": 1.1073, + "step": 7093 + }, + { + "epoch": 0.46, + "learning_rate": 1.8499323729577708e-05, + "loss": 1.0111, + "step": 7094 + }, + { + "epoch": 0.46, + "learning_rate": 1.8499109040554757e-05, + "loss": 1.0726, + "step": 7095 + }, + { + "epoch": 0.46, + "learning_rate": 1.849889435153181e-05, + "loss": 1.1092, + "step": 7096 + }, + { + "epoch": 0.46, + "learning_rate": 1.8498679662508858e-05, + "loss": 1.1329, + "step": 7097 + }, + { + "epoch": 0.46, + "learning_rate": 1.8498464973485907e-05, + "loss": 1.1376, + "step": 7098 + }, + { + "epoch": 0.46, + "learning_rate": 1.8498250284462956e-05, + "loss": 1.0622, + "step": 7099 + }, + { + "epoch": 0.46, + "learning_rate": 1.849803559544001e-05, + "loss": 1.0725, + "step": 7100 + }, + { + "epoch": 0.46, + "learning_rate": 1.8497820906417057e-05, + "loss": 1.0319, + "step": 7101 + }, + { + "epoch": 0.46, + "learning_rate": 1.8497606217394106e-05, + "loss": 1.1248, + "step": 7102 + }, + { + "epoch": 0.46, + "learning_rate": 1.8497391528371155e-05, + "loss": 1.076, + "step": 7103 + }, + { + "epoch": 0.46, + "learning_rate": 1.8497176839348204e-05, + "loss": 1.161, + "step": 7104 + }, + { + "epoch": 0.46, + "learning_rate": 1.8496962150325257e-05, + "loss": 1.2203, + "step": 7105 + }, + { + "epoch": 0.46, + "learning_rate": 1.8496747461302306e-05, + "loss": 1.1066, + "step": 7106 + }, + { + "epoch": 0.46, + "learning_rate": 1.8496532772279355e-05, + "loss": 1.1182, + "step": 7107 + }, + { + "epoch": 0.46, + "learning_rate": 1.8496318083256404e-05, + "loss": 1.1126, + "step": 7108 + }, + { + "epoch": 0.46, + "learning_rate": 1.8496103394233453e-05, + "loss": 1.088, + "step": 7109 + }, + { + "epoch": 0.46, + "learning_rate": 1.8495888705210505e-05, + "loss": 1.0813, + "step": 7110 + }, + { + "epoch": 0.46, + "learning_rate": 1.8495674016187554e-05, + "loss": 1.1043, + "step": 7111 + }, + { + "epoch": 0.46, + "learning_rate": 1.8495459327164603e-05, + "loss": 1.0917, + "step": 7112 + }, + { + "epoch": 0.46, + "learning_rate": 1.8495244638141655e-05, + "loss": 1.0595, + "step": 7113 + }, + { + "epoch": 0.46, + "learning_rate": 1.8495029949118704e-05, + "loss": 1.0762, + "step": 7114 + }, + { + "epoch": 0.46, + "learning_rate": 1.8494815260095753e-05, + "loss": 1.1559, + "step": 7115 + }, + { + "epoch": 0.46, + "learning_rate": 1.8494600571072802e-05, + "loss": 1.0718, + "step": 7116 + }, + { + "epoch": 0.46, + "learning_rate": 1.849438588204985e-05, + "loss": 1.1752, + "step": 7117 + }, + { + "epoch": 0.46, + "learning_rate": 1.8494171193026903e-05, + "loss": 1.072, + "step": 7118 + }, + { + "epoch": 0.46, + "learning_rate": 1.8493956504003952e-05, + "loss": 1.2047, + "step": 7119 + }, + { + "epoch": 0.46, + "learning_rate": 1.8493741814981e-05, + "loss": 1.0991, + "step": 7120 + }, + { + "epoch": 0.46, + "learning_rate": 1.849352712595805e-05, + "loss": 1.0258, + "step": 7121 + }, + { + "epoch": 0.46, + "learning_rate": 1.84933124369351e-05, + "loss": 1.1111, + "step": 7122 + }, + { + "epoch": 0.46, + "learning_rate": 1.8493097747912148e-05, + "loss": 1.0398, + "step": 7123 + }, + { + "epoch": 0.46, + "learning_rate": 1.84928830588892e-05, + "loss": 1.0782, + "step": 7124 + }, + { + "epoch": 0.46, + "learning_rate": 1.8492668369866253e-05, + "loss": 1.1016, + "step": 7125 + }, + { + "epoch": 0.46, + "learning_rate": 1.8492453680843302e-05, + "loss": 1.1384, + "step": 7126 + }, + { + "epoch": 0.46, + "learning_rate": 1.849223899182035e-05, + "loss": 1.0869, + "step": 7127 + }, + { + "epoch": 0.46, + "learning_rate": 1.84920243027974e-05, + "loss": 1.0682, + "step": 7128 + }, + { + "epoch": 0.46, + "learning_rate": 1.849180961377445e-05, + "loss": 1.0941, + "step": 7129 + }, + { + "epoch": 0.46, + "learning_rate": 1.8491594924751498e-05, + "loss": 1.0588, + "step": 7130 + }, + { + "epoch": 0.46, + "learning_rate": 1.849138023572855e-05, + "loss": 1.0958, + "step": 7131 + }, + { + "epoch": 0.46, + "learning_rate": 1.84911655467056e-05, + "loss": 1.0752, + "step": 7132 + }, + { + "epoch": 0.46, + "learning_rate": 1.8490950857682648e-05, + "loss": 1.0801, + "step": 7133 + }, + { + "epoch": 0.46, + "learning_rate": 1.8490736168659697e-05, + "loss": 1.1189, + "step": 7134 + }, + { + "epoch": 0.46, + "learning_rate": 1.8490521479636746e-05, + "loss": 1.0753, + "step": 7135 + }, + { + "epoch": 0.46, + "learning_rate": 1.8490306790613798e-05, + "loss": 1.1542, + "step": 7136 + }, + { + "epoch": 0.46, + "learning_rate": 1.8490092101590847e-05, + "loss": 1.0081, + "step": 7137 + }, + { + "epoch": 0.46, + "learning_rate": 1.8489877412567896e-05, + "loss": 1.0433, + "step": 7138 + }, + { + "epoch": 0.46, + "learning_rate": 1.848966272354495e-05, + "loss": 0.9868, + "step": 7139 + }, + { + "epoch": 0.46, + "learning_rate": 1.8489448034521998e-05, + "loss": 1.1599, + "step": 7140 + }, + { + "epoch": 0.46, + "learning_rate": 1.8489233345499046e-05, + "loss": 1.1859, + "step": 7141 + }, + { + "epoch": 0.46, + "learning_rate": 1.8489018656476095e-05, + "loss": 1.1362, + "step": 7142 + }, + { + "epoch": 0.46, + "learning_rate": 1.8488803967453144e-05, + "loss": 1.1244, + "step": 7143 + }, + { + "epoch": 0.46, + "learning_rate": 1.8488589278430197e-05, + "loss": 1.0399, + "step": 7144 + }, + { + "epoch": 0.46, + "learning_rate": 1.8488374589407246e-05, + "loss": 1.1742, + "step": 7145 + }, + { + "epoch": 0.46, + "learning_rate": 1.8488159900384295e-05, + "loss": 1.1193, + "step": 7146 + }, + { + "epoch": 0.46, + "learning_rate": 1.8487945211361344e-05, + "loss": 1.1171, + "step": 7147 + }, + { + "epoch": 0.46, + "learning_rate": 1.8487730522338393e-05, + "loss": 1.0833, + "step": 7148 + }, + { + "epoch": 0.46, + "learning_rate": 1.8487515833315445e-05, + "loss": 1.0509, + "step": 7149 + }, + { + "epoch": 0.46, + "learning_rate": 1.8487301144292494e-05, + "loss": 1.1438, + "step": 7150 + }, + { + "epoch": 0.46, + "learning_rate": 1.8487086455269543e-05, + "loss": 1.0774, + "step": 7151 + }, + { + "epoch": 0.46, + "learning_rate": 1.8486871766246592e-05, + "loss": 1.0264, + "step": 7152 + }, + { + "epoch": 0.46, + "learning_rate": 1.848665707722364e-05, + "loss": 1.1517, + "step": 7153 + }, + { + "epoch": 0.46, + "learning_rate": 1.8486442388200693e-05, + "loss": 1.1441, + "step": 7154 + }, + { + "epoch": 0.46, + "learning_rate": 1.8486227699177742e-05, + "loss": 1.1103, + "step": 7155 + }, + { + "epoch": 0.46, + "learning_rate": 1.8486013010154795e-05, + "loss": 1.0312, + "step": 7156 + }, + { + "epoch": 0.46, + "learning_rate": 1.8485798321131843e-05, + "loss": 1.049, + "step": 7157 + }, + { + "epoch": 0.46, + "learning_rate": 1.8485583632108892e-05, + "loss": 1.0836, + "step": 7158 + }, + { + "epoch": 0.46, + "learning_rate": 1.848536894308594e-05, + "loss": 1.119, + "step": 7159 + }, + { + "epoch": 0.46, + "learning_rate": 1.848515425406299e-05, + "loss": 1.0095, + "step": 7160 + }, + { + "epoch": 0.46, + "learning_rate": 1.848493956504004e-05, + "loss": 1.0203, + "step": 7161 + }, + { + "epoch": 0.46, + "learning_rate": 1.8484724876017092e-05, + "loss": 1.0482, + "step": 7162 + }, + { + "epoch": 0.46, + "learning_rate": 1.848451018699414e-05, + "loss": 1.006, + "step": 7163 + }, + { + "epoch": 0.46, + "learning_rate": 1.848429549797119e-05, + "loss": 1.1113, + "step": 7164 + }, + { + "epoch": 0.46, + "learning_rate": 1.848408080894824e-05, + "loss": 1.1003, + "step": 7165 + }, + { + "epoch": 0.46, + "learning_rate": 1.8483866119925288e-05, + "loss": 1.1697, + "step": 7166 + }, + { + "epoch": 0.46, + "learning_rate": 1.848365143090234e-05, + "loss": 1.0634, + "step": 7167 + }, + { + "epoch": 0.46, + "learning_rate": 1.848343674187939e-05, + "loss": 1.0903, + "step": 7168 + }, + { + "epoch": 0.46, + "learning_rate": 1.848322205285644e-05, + "loss": 1.0647, + "step": 7169 + }, + { + "epoch": 0.46, + "learning_rate": 1.848300736383349e-05, + "loss": 1.1215, + "step": 7170 + }, + { + "epoch": 0.46, + "learning_rate": 1.848279267481054e-05, + "loss": 1.1332, + "step": 7171 + }, + { + "epoch": 0.46, + "learning_rate": 1.8482577985787588e-05, + "loss": 1.0525, + "step": 7172 + }, + { + "epoch": 0.46, + "learning_rate": 1.8482363296764637e-05, + "loss": 1.1754, + "step": 7173 + }, + { + "epoch": 0.46, + "learning_rate": 1.848214860774169e-05, + "loss": 1.1026, + "step": 7174 + }, + { + "epoch": 0.46, + "learning_rate": 1.848193391871874e-05, + "loss": 1.0763, + "step": 7175 + }, + { + "epoch": 0.46, + "learning_rate": 1.8481719229695787e-05, + "loss": 1.106, + "step": 7176 + }, + { + "epoch": 0.46, + "learning_rate": 1.8481504540672836e-05, + "loss": 1.0609, + "step": 7177 + }, + { + "epoch": 0.46, + "learning_rate": 1.8481289851649885e-05, + "loss": 1.0495, + "step": 7178 + }, + { + "epoch": 0.46, + "learning_rate": 1.8481075162626934e-05, + "loss": 1.1025, + "step": 7179 + }, + { + "epoch": 0.46, + "learning_rate": 1.8480860473603987e-05, + "loss": 0.988, + "step": 7180 + }, + { + "epoch": 0.46, + "learning_rate": 1.8480645784581036e-05, + "loss": 1.1055, + "step": 7181 + }, + { + "epoch": 0.46, + "learning_rate": 1.8480431095558085e-05, + "loss": 1.1029, + "step": 7182 + }, + { + "epoch": 0.46, + "learning_rate": 1.8480216406535137e-05, + "loss": 1.0236, + "step": 7183 + }, + { + "epoch": 0.46, + "learning_rate": 1.8480001717512186e-05, + "loss": 1.0344, + "step": 7184 + }, + { + "epoch": 0.46, + "learning_rate": 1.8479787028489235e-05, + "loss": 1.0979, + "step": 7185 + }, + { + "epoch": 0.46, + "learning_rate": 1.8479572339466284e-05, + "loss": 1.0466, + "step": 7186 + }, + { + "epoch": 0.46, + "learning_rate": 1.8479357650443336e-05, + "loss": 1.1341, + "step": 7187 + }, + { + "epoch": 0.46, + "learning_rate": 1.8479142961420385e-05, + "loss": 1.0941, + "step": 7188 + }, + { + "epoch": 0.46, + "learning_rate": 1.8478928272397434e-05, + "loss": 1.1197, + "step": 7189 + }, + { + "epoch": 0.46, + "learning_rate": 1.8478713583374483e-05, + "loss": 1.1334, + "step": 7190 + }, + { + "epoch": 0.46, + "learning_rate": 1.8478498894351532e-05, + "loss": 1.0455, + "step": 7191 + }, + { + "epoch": 0.46, + "learning_rate": 1.8478284205328584e-05, + "loss": 1.0431, + "step": 7192 + }, + { + "epoch": 0.46, + "learning_rate": 1.8478069516305633e-05, + "loss": 1.0302, + "step": 7193 + }, + { + "epoch": 0.46, + "learning_rate": 1.8477854827282682e-05, + "loss": 1.1125, + "step": 7194 + }, + { + "epoch": 0.46, + "learning_rate": 1.847764013825973e-05, + "loss": 0.9809, + "step": 7195 + }, + { + "epoch": 0.46, + "learning_rate": 1.847742544923678e-05, + "loss": 1.0551, + "step": 7196 + }, + { + "epoch": 0.46, + "learning_rate": 1.8477210760213833e-05, + "loss": 1.0358, + "step": 7197 + }, + { + "epoch": 0.46, + "learning_rate": 1.847699607119088e-05, + "loss": 1.0726, + "step": 7198 + }, + { + "epoch": 0.46, + "learning_rate": 1.847678138216793e-05, + "loss": 1.0818, + "step": 7199 + }, + { + "epoch": 0.46, + "learning_rate": 1.8476566693144983e-05, + "loss": 1.1208, + "step": 7200 + }, + { + "epoch": 0.46, + "learning_rate": 1.8476352004122032e-05, + "loss": 1.0868, + "step": 7201 + }, + { + "epoch": 0.46, + "learning_rate": 1.847613731509908e-05, + "loss": 1.0534, + "step": 7202 + }, + { + "epoch": 0.46, + "learning_rate": 1.847592262607613e-05, + "loss": 1.0745, + "step": 7203 + }, + { + "epoch": 0.46, + "learning_rate": 1.847570793705318e-05, + "loss": 1.1532, + "step": 7204 + }, + { + "epoch": 0.46, + "learning_rate": 1.847549324803023e-05, + "loss": 1.0714, + "step": 7205 + }, + { + "epoch": 0.46, + "learning_rate": 1.847527855900728e-05, + "loss": 1.093, + "step": 7206 + }, + { + "epoch": 0.46, + "learning_rate": 1.847506386998433e-05, + "loss": 1.0627, + "step": 7207 + }, + { + "epoch": 0.46, + "learning_rate": 1.8474849180961378e-05, + "loss": 1.1093, + "step": 7208 + }, + { + "epoch": 0.46, + "learning_rate": 1.8474634491938427e-05, + "loss": 1.0702, + "step": 7209 + }, + { + "epoch": 0.46, + "learning_rate": 1.8474419802915476e-05, + "loss": 1.0565, + "step": 7210 + }, + { + "epoch": 0.46, + "learning_rate": 1.8474205113892528e-05, + "loss": 1.1156, + "step": 7211 + }, + { + "epoch": 0.46, + "learning_rate": 1.847399042486958e-05, + "loss": 1.0429, + "step": 7212 + }, + { + "epoch": 0.46, + "learning_rate": 1.847377573584663e-05, + "loss": 1.1389, + "step": 7213 + }, + { + "epoch": 0.46, + "learning_rate": 1.847356104682368e-05, + "loss": 1.0453, + "step": 7214 + }, + { + "epoch": 0.46, + "learning_rate": 1.8473346357800727e-05, + "loss": 1.0348, + "step": 7215 + }, + { + "epoch": 0.46, + "learning_rate": 1.8473131668777776e-05, + "loss": 1.1409, + "step": 7216 + }, + { + "epoch": 0.46, + "learning_rate": 1.8472916979754825e-05, + "loss": 1.0514, + "step": 7217 + }, + { + "epoch": 0.46, + "learning_rate": 1.8472702290731878e-05, + "loss": 1.0939, + "step": 7218 + }, + { + "epoch": 0.46, + "learning_rate": 1.8472487601708927e-05, + "loss": 1.1156, + "step": 7219 + }, + { + "epoch": 0.46, + "learning_rate": 1.8472272912685976e-05, + "loss": 1.1463, + "step": 7220 + }, + { + "epoch": 0.46, + "learning_rate": 1.8472058223663025e-05, + "loss": 1.0467, + "step": 7221 + }, + { + "epoch": 0.46, + "learning_rate": 1.8471843534640074e-05, + "loss": 1.0555, + "step": 7222 + }, + { + "epoch": 0.46, + "learning_rate": 1.8471628845617126e-05, + "loss": 1.095, + "step": 7223 + }, + { + "epoch": 0.46, + "learning_rate": 1.8471414156594175e-05, + "loss": 1.1141, + "step": 7224 + }, + { + "epoch": 0.46, + "learning_rate": 1.8471199467571224e-05, + "loss": 1.0842, + "step": 7225 + }, + { + "epoch": 0.46, + "learning_rate": 1.8470984778548276e-05, + "loss": 1.095, + "step": 7226 + }, + { + "epoch": 0.46, + "learning_rate": 1.8470770089525325e-05, + "loss": 1.0667, + "step": 7227 + }, + { + "epoch": 0.47, + "learning_rate": 1.8470555400502374e-05, + "loss": 1.1312, + "step": 7228 + }, + { + "epoch": 0.47, + "learning_rate": 1.8470340711479423e-05, + "loss": 1.0598, + "step": 7229 + }, + { + "epoch": 0.47, + "learning_rate": 1.8470126022456472e-05, + "loss": 1.0759, + "step": 7230 + }, + { + "epoch": 0.47, + "learning_rate": 1.8469911333433524e-05, + "loss": 1.0728, + "step": 7231 + }, + { + "epoch": 0.47, + "learning_rate": 1.8469696644410573e-05, + "loss": 1.1063, + "step": 7232 + }, + { + "epoch": 0.47, + "learning_rate": 1.8469481955387622e-05, + "loss": 1.1693, + "step": 7233 + }, + { + "epoch": 0.47, + "learning_rate": 1.846926726636467e-05, + "loss": 1.1327, + "step": 7234 + }, + { + "epoch": 0.47, + "learning_rate": 1.846905257734172e-05, + "loss": 1.0927, + "step": 7235 + }, + { + "epoch": 0.47, + "learning_rate": 1.8468837888318773e-05, + "loss": 1.0916, + "step": 7236 + }, + { + "epoch": 0.47, + "learning_rate": 1.846862319929582e-05, + "loss": 1.0531, + "step": 7237 + }, + { + "epoch": 0.47, + "learning_rate": 1.846840851027287e-05, + "loss": 1.075, + "step": 7238 + }, + { + "epoch": 0.47, + "learning_rate": 1.846819382124992e-05, + "loss": 1.1308, + "step": 7239 + }, + { + "epoch": 0.47, + "learning_rate": 1.846797913222697e-05, + "loss": 1.1484, + "step": 7240 + }, + { + "epoch": 0.47, + "learning_rate": 1.846776444320402e-05, + "loss": 1.1523, + "step": 7241 + }, + { + "epoch": 0.47, + "learning_rate": 1.846754975418107e-05, + "loss": 1.0666, + "step": 7242 + }, + { + "epoch": 0.47, + "learning_rate": 1.8467335065158122e-05, + "loss": 1.0844, + "step": 7243 + }, + { + "epoch": 0.47, + "learning_rate": 1.846712037613517e-05, + "loss": 1.1993, + "step": 7244 + }, + { + "epoch": 0.47, + "learning_rate": 1.846690568711222e-05, + "loss": 1.1789, + "step": 7245 + }, + { + "epoch": 0.47, + "learning_rate": 1.846669099808927e-05, + "loss": 1.127, + "step": 7246 + }, + { + "epoch": 0.47, + "learning_rate": 1.8466476309066318e-05, + "loss": 1.0977, + "step": 7247 + }, + { + "epoch": 0.47, + "learning_rate": 1.8466261620043367e-05, + "loss": 1.135, + "step": 7248 + }, + { + "epoch": 0.47, + "learning_rate": 1.846604693102042e-05, + "loss": 1.1464, + "step": 7249 + }, + { + "epoch": 0.47, + "learning_rate": 1.846583224199747e-05, + "loss": 1.102, + "step": 7250 + }, + { + "epoch": 0.47, + "learning_rate": 1.8465617552974517e-05, + "loss": 1.0711, + "step": 7251 + }, + { + "epoch": 0.47, + "learning_rate": 1.8465402863951566e-05, + "loss": 1.0855, + "step": 7252 + }, + { + "epoch": 0.47, + "learning_rate": 1.8465188174928615e-05, + "loss": 1.0901, + "step": 7253 + }, + { + "epoch": 0.47, + "learning_rate": 1.8464973485905668e-05, + "loss": 1.0375, + "step": 7254 + }, + { + "epoch": 0.47, + "learning_rate": 1.8464758796882717e-05, + "loss": 1.1391, + "step": 7255 + }, + { + "epoch": 0.47, + "learning_rate": 1.846454410785977e-05, + "loss": 1.1796, + "step": 7256 + }, + { + "epoch": 0.47, + "learning_rate": 1.8464329418836818e-05, + "loss": 1.0312, + "step": 7257 + }, + { + "epoch": 0.47, + "learning_rate": 1.8464114729813867e-05, + "loss": 1.1233, + "step": 7258 + }, + { + "epoch": 0.47, + "learning_rate": 1.8463900040790916e-05, + "loss": 1.0855, + "step": 7259 + }, + { + "epoch": 0.47, + "learning_rate": 1.8463685351767965e-05, + "loss": 1.1412, + "step": 7260 + }, + { + "epoch": 0.47, + "learning_rate": 1.8463470662745017e-05, + "loss": 1.039, + "step": 7261 + }, + { + "epoch": 0.47, + "learning_rate": 1.8463255973722066e-05, + "loss": 1.0082, + "step": 7262 + }, + { + "epoch": 0.47, + "learning_rate": 1.8463041284699115e-05, + "loss": 1.0645, + "step": 7263 + }, + { + "epoch": 0.47, + "learning_rate": 1.8462826595676164e-05, + "loss": 1.1047, + "step": 7264 + }, + { + "epoch": 0.47, + "learning_rate": 1.8462611906653213e-05, + "loss": 1.1192, + "step": 7265 + }, + { + "epoch": 0.47, + "learning_rate": 1.8462397217630262e-05, + "loss": 1.1178, + "step": 7266 + }, + { + "epoch": 0.47, + "learning_rate": 1.8462182528607314e-05, + "loss": 1.112, + "step": 7267 + }, + { + "epoch": 0.47, + "learning_rate": 1.8461967839584363e-05, + "loss": 1.0813, + "step": 7268 + }, + { + "epoch": 0.47, + "learning_rate": 1.8461753150561412e-05, + "loss": 1.089, + "step": 7269 + }, + { + "epoch": 0.47, + "learning_rate": 1.8461538461538465e-05, + "loss": 1.099, + "step": 7270 + }, + { + "epoch": 0.47, + "learning_rate": 1.8461323772515514e-05, + "loss": 1.0804, + "step": 7271 + }, + { + "epoch": 0.47, + "learning_rate": 1.8461109083492563e-05, + "loss": 1.1345, + "step": 7272 + }, + { + "epoch": 0.47, + "learning_rate": 1.846089439446961e-05, + "loss": 1.1135, + "step": 7273 + }, + { + "epoch": 0.47, + "learning_rate": 1.8460679705446664e-05, + "loss": 1.0743, + "step": 7274 + }, + { + "epoch": 0.47, + "learning_rate": 1.8460465016423713e-05, + "loss": 1.2202, + "step": 7275 + }, + { + "epoch": 0.47, + "learning_rate": 1.8460250327400762e-05, + "loss": 1.0395, + "step": 7276 + }, + { + "epoch": 0.47, + "learning_rate": 1.846003563837781e-05, + "loss": 1.2026, + "step": 7277 + }, + { + "epoch": 0.47, + "learning_rate": 1.845982094935486e-05, + "loss": 1.1211, + "step": 7278 + }, + { + "epoch": 0.47, + "learning_rate": 1.8459606260331912e-05, + "loss": 1.1313, + "step": 7279 + }, + { + "epoch": 0.47, + "learning_rate": 1.845939157130896e-05, + "loss": 1.0741, + "step": 7280 + }, + { + "epoch": 0.47, + "learning_rate": 1.845917688228601e-05, + "loss": 1.09, + "step": 7281 + }, + { + "epoch": 0.47, + "learning_rate": 1.845896219326306e-05, + "loss": 1.0747, + "step": 7282 + }, + { + "epoch": 0.47, + "learning_rate": 1.8458747504240108e-05, + "loss": 1.1163, + "step": 7283 + }, + { + "epoch": 0.47, + "learning_rate": 1.845853281521716e-05, + "loss": 1.019, + "step": 7284 + }, + { + "epoch": 0.47, + "learning_rate": 1.845831812619421e-05, + "loss": 1.0824, + "step": 7285 + }, + { + "epoch": 0.47, + "learning_rate": 1.8458103437171258e-05, + "loss": 1.0605, + "step": 7286 + }, + { + "epoch": 0.47, + "learning_rate": 1.845788874814831e-05, + "loss": 1.0759, + "step": 7287 + }, + { + "epoch": 0.47, + "learning_rate": 1.845767405912536e-05, + "loss": 1.0839, + "step": 7288 + }, + { + "epoch": 0.47, + "learning_rate": 1.845745937010241e-05, + "loss": 1.1477, + "step": 7289 + }, + { + "epoch": 0.47, + "learning_rate": 1.8457244681079457e-05, + "loss": 1.108, + "step": 7290 + }, + { + "epoch": 0.47, + "learning_rate": 1.8457029992056506e-05, + "loss": 1.1429, + "step": 7291 + }, + { + "epoch": 0.47, + "learning_rate": 1.845681530303356e-05, + "loss": 1.0762, + "step": 7292 + }, + { + "epoch": 0.47, + "learning_rate": 1.8456600614010608e-05, + "loss": 1.0452, + "step": 7293 + }, + { + "epoch": 0.47, + "learning_rate": 1.8456385924987657e-05, + "loss": 1.0961, + "step": 7294 + }, + { + "epoch": 0.47, + "learning_rate": 1.8456171235964706e-05, + "loss": 1.0805, + "step": 7295 + }, + { + "epoch": 0.47, + "learning_rate": 1.8455956546941755e-05, + "loss": 1.0529, + "step": 7296 + }, + { + "epoch": 0.47, + "learning_rate": 1.8455741857918804e-05, + "loss": 1.0519, + "step": 7297 + }, + { + "epoch": 0.47, + "learning_rate": 1.8455527168895856e-05, + "loss": 1.1258, + "step": 7298 + }, + { + "epoch": 0.47, + "learning_rate": 1.845531247987291e-05, + "loss": 1.1408, + "step": 7299 + }, + { + "epoch": 0.47, + "learning_rate": 1.8455097790849957e-05, + "loss": 1.0449, + "step": 7300 + }, + { + "epoch": 0.47, + "learning_rate": 1.8454883101827006e-05, + "loss": 1.0076, + "step": 7301 + }, + { + "epoch": 0.47, + "learning_rate": 1.8454668412804055e-05, + "loss": 1.0843, + "step": 7302 + }, + { + "epoch": 0.47, + "learning_rate": 1.8454453723781104e-05, + "loss": 0.9551, + "step": 7303 + }, + { + "epoch": 0.47, + "learning_rate": 1.8454239034758153e-05, + "loss": 1.1478, + "step": 7304 + }, + { + "epoch": 0.47, + "learning_rate": 1.8454024345735205e-05, + "loss": 1.0745, + "step": 7305 + }, + { + "epoch": 0.47, + "learning_rate": 1.8453809656712254e-05, + "loss": 0.9802, + "step": 7306 + }, + { + "epoch": 0.47, + "learning_rate": 1.8453594967689303e-05, + "loss": 1.0951, + "step": 7307 + }, + { + "epoch": 0.47, + "learning_rate": 1.8453380278666352e-05, + "loss": 1.1329, + "step": 7308 + }, + { + "epoch": 0.47, + "learning_rate": 1.84531655896434e-05, + "loss": 1.0684, + "step": 7309 + }, + { + "epoch": 0.47, + "learning_rate": 1.8452950900620454e-05, + "loss": 1.1216, + "step": 7310 + }, + { + "epoch": 0.47, + "learning_rate": 1.8452736211597503e-05, + "loss": 1.1302, + "step": 7311 + }, + { + "epoch": 0.47, + "learning_rate": 1.845252152257455e-05, + "loss": 1.1328, + "step": 7312 + }, + { + "epoch": 0.47, + "learning_rate": 1.84523068335516e-05, + "loss": 1.0747, + "step": 7313 + }, + { + "epoch": 0.47, + "learning_rate": 1.8452092144528653e-05, + "loss": 1.0266, + "step": 7314 + }, + { + "epoch": 0.47, + "learning_rate": 1.8451877455505702e-05, + "loss": 1.0788, + "step": 7315 + }, + { + "epoch": 0.47, + "learning_rate": 1.845166276648275e-05, + "loss": 1.0719, + "step": 7316 + }, + { + "epoch": 0.47, + "learning_rate": 1.84514480774598e-05, + "loss": 1.0158, + "step": 7317 + }, + { + "epoch": 0.47, + "learning_rate": 1.8451233388436852e-05, + "loss": 1.1707, + "step": 7318 + }, + { + "epoch": 0.47, + "learning_rate": 1.84510186994139e-05, + "loss": 1.033, + "step": 7319 + }, + { + "epoch": 0.47, + "learning_rate": 1.845080401039095e-05, + "loss": 1.1285, + "step": 7320 + }, + { + "epoch": 0.47, + "learning_rate": 1.8450589321368e-05, + "loss": 1.0805, + "step": 7321 + }, + { + "epoch": 0.47, + "learning_rate": 1.8450374632345048e-05, + "loss": 1.1272, + "step": 7322 + }, + { + "epoch": 0.47, + "learning_rate": 1.84501599433221e-05, + "loss": 1.018, + "step": 7323 + }, + { + "epoch": 0.47, + "learning_rate": 1.844994525429915e-05, + "loss": 1.0493, + "step": 7324 + }, + { + "epoch": 0.47, + "learning_rate": 1.84497305652762e-05, + "loss": 1.1104, + "step": 7325 + }, + { + "epoch": 0.47, + "learning_rate": 1.8449515876253247e-05, + "loss": 1.0303, + "step": 7326 + }, + { + "epoch": 0.47, + "learning_rate": 1.8449301187230296e-05, + "loss": 1.1299, + "step": 7327 + }, + { + "epoch": 0.47, + "learning_rate": 1.844908649820735e-05, + "loss": 1.103, + "step": 7328 + }, + { + "epoch": 0.47, + "learning_rate": 1.8448871809184398e-05, + "loss": 1.0363, + "step": 7329 + }, + { + "epoch": 0.47, + "learning_rate": 1.844865712016145e-05, + "loss": 1.0159, + "step": 7330 + }, + { + "epoch": 0.47, + "learning_rate": 1.84484424311385e-05, + "loss": 1.1115, + "step": 7331 + }, + { + "epoch": 0.47, + "learning_rate": 1.8448227742115548e-05, + "loss": 1.1266, + "step": 7332 + }, + { + "epoch": 0.47, + "learning_rate": 1.8448013053092597e-05, + "loss": 1.1102, + "step": 7333 + }, + { + "epoch": 0.47, + "learning_rate": 1.8447798364069646e-05, + "loss": 1.0426, + "step": 7334 + }, + { + "epoch": 0.47, + "learning_rate": 1.8447583675046695e-05, + "loss": 1.0754, + "step": 7335 + }, + { + "epoch": 0.47, + "learning_rate": 1.8447368986023747e-05, + "loss": 1.0463, + "step": 7336 + }, + { + "epoch": 0.47, + "learning_rate": 1.8447154297000796e-05, + "loss": 1.1038, + "step": 7337 + }, + { + "epoch": 0.47, + "learning_rate": 1.8446939607977845e-05, + "loss": 1.0705, + "step": 7338 + }, + { + "epoch": 0.47, + "learning_rate": 1.8446724918954894e-05, + "loss": 1.1107, + "step": 7339 + }, + { + "epoch": 0.47, + "learning_rate": 1.8446510229931943e-05, + "loss": 1.1314, + "step": 7340 + }, + { + "epoch": 0.47, + "learning_rate": 1.8446295540908995e-05, + "loss": 1.1106, + "step": 7341 + }, + { + "epoch": 0.47, + "learning_rate": 1.8446080851886044e-05, + "loss": 1.189, + "step": 7342 + }, + { + "epoch": 0.47, + "learning_rate": 1.8445866162863097e-05, + "loss": 1.0638, + "step": 7343 + }, + { + "epoch": 0.47, + "learning_rate": 1.8445651473840146e-05, + "loss": 1.1508, + "step": 7344 + }, + { + "epoch": 0.47, + "learning_rate": 1.8445436784817195e-05, + "loss": 1.1524, + "step": 7345 + }, + { + "epoch": 0.47, + "learning_rate": 1.8445222095794244e-05, + "loss": 1.1303, + "step": 7346 + }, + { + "epoch": 0.47, + "learning_rate": 1.8445007406771293e-05, + "loss": 1.0686, + "step": 7347 + }, + { + "epoch": 0.47, + "learning_rate": 1.8444792717748345e-05, + "loss": 1.1433, + "step": 7348 + }, + { + "epoch": 0.47, + "learning_rate": 1.8444578028725394e-05, + "loss": 1.153, + "step": 7349 + }, + { + "epoch": 0.47, + "learning_rate": 1.8444363339702443e-05, + "loss": 1.1005, + "step": 7350 + }, + { + "epoch": 0.47, + "learning_rate": 1.8444148650679492e-05, + "loss": 1.1012, + "step": 7351 + }, + { + "epoch": 0.47, + "learning_rate": 1.844393396165654e-05, + "loss": 1.0775, + "step": 7352 + }, + { + "epoch": 0.47, + "learning_rate": 1.844371927263359e-05, + "loss": 1.1242, + "step": 7353 + }, + { + "epoch": 0.47, + "learning_rate": 1.8443504583610642e-05, + "loss": 1.0707, + "step": 7354 + }, + { + "epoch": 0.47, + "learning_rate": 1.844328989458769e-05, + "loss": 1.0703, + "step": 7355 + }, + { + "epoch": 0.47, + "learning_rate": 1.844307520556474e-05, + "loss": 1.093, + "step": 7356 + }, + { + "epoch": 0.47, + "learning_rate": 1.8442860516541792e-05, + "loss": 1.0376, + "step": 7357 + }, + { + "epoch": 0.47, + "learning_rate": 1.844264582751884e-05, + "loss": 1.0453, + "step": 7358 + }, + { + "epoch": 0.47, + "learning_rate": 1.844243113849589e-05, + "loss": 1.1179, + "step": 7359 + }, + { + "epoch": 0.47, + "learning_rate": 1.844221644947294e-05, + "loss": 1.0926, + "step": 7360 + }, + { + "epoch": 0.47, + "learning_rate": 1.844200176044999e-05, + "loss": 1.0344, + "step": 7361 + }, + { + "epoch": 0.47, + "learning_rate": 1.844178707142704e-05, + "loss": 1.0862, + "step": 7362 + }, + { + "epoch": 0.47, + "learning_rate": 1.844157238240409e-05, + "loss": 1.0479, + "step": 7363 + }, + { + "epoch": 0.47, + "learning_rate": 1.844135769338114e-05, + "loss": 1.0246, + "step": 7364 + }, + { + "epoch": 0.47, + "learning_rate": 1.8441143004358187e-05, + "loss": 1.1135, + "step": 7365 + }, + { + "epoch": 0.47, + "learning_rate": 1.8440928315335236e-05, + "loss": 1.1377, + "step": 7366 + }, + { + "epoch": 0.47, + "learning_rate": 1.844071362631229e-05, + "loss": 1.0916, + "step": 7367 + }, + { + "epoch": 0.47, + "learning_rate": 1.8440498937289338e-05, + "loss": 1.1492, + "step": 7368 + }, + { + "epoch": 0.47, + "learning_rate": 1.8440284248266387e-05, + "loss": 1.2166, + "step": 7369 + }, + { + "epoch": 0.47, + "learning_rate": 1.8440069559243436e-05, + "loss": 1.1934, + "step": 7370 + }, + { + "epoch": 0.47, + "learning_rate": 1.8439854870220485e-05, + "loss": 1.1049, + "step": 7371 + }, + { + "epoch": 0.47, + "learning_rate": 1.8439640181197537e-05, + "loss": 1.0806, + "step": 7372 + }, + { + "epoch": 0.47, + "learning_rate": 1.8439425492174586e-05, + "loss": 1.1539, + "step": 7373 + }, + { + "epoch": 0.47, + "learning_rate": 1.8439210803151638e-05, + "loss": 1.1115, + "step": 7374 + }, + { + "epoch": 0.47, + "learning_rate": 1.8438996114128687e-05, + "loss": 1.1008, + "step": 7375 + }, + { + "epoch": 0.47, + "learning_rate": 1.8438781425105736e-05, + "loss": 1.2011, + "step": 7376 + }, + { + "epoch": 0.47, + "learning_rate": 1.8438566736082785e-05, + "loss": 1.0283, + "step": 7377 + }, + { + "epoch": 0.47, + "learning_rate": 1.8438352047059834e-05, + "loss": 0.9905, + "step": 7378 + }, + { + "epoch": 0.47, + "learning_rate": 1.8438137358036886e-05, + "loss": 1.1192, + "step": 7379 + }, + { + "epoch": 0.47, + "learning_rate": 1.8437922669013935e-05, + "loss": 1.0847, + "step": 7380 + }, + { + "epoch": 0.47, + "learning_rate": 1.8437707979990984e-05, + "loss": 1.0397, + "step": 7381 + }, + { + "epoch": 0.47, + "learning_rate": 1.8437493290968033e-05, + "loss": 1.0762, + "step": 7382 + }, + { + "epoch": 0.48, + "learning_rate": 1.8437278601945082e-05, + "loss": 1.0926, + "step": 7383 + }, + { + "epoch": 0.48, + "learning_rate": 1.843706391292213e-05, + "loss": 1.1594, + "step": 7384 + }, + { + "epoch": 0.48, + "learning_rate": 1.8436849223899184e-05, + "loss": 1.1102, + "step": 7385 + }, + { + "epoch": 0.48, + "learning_rate": 1.8436634534876236e-05, + "loss": 1.139, + "step": 7386 + }, + { + "epoch": 0.48, + "learning_rate": 1.8436419845853285e-05, + "loss": 1.2326, + "step": 7387 + }, + { + "epoch": 0.48, + "learning_rate": 1.8436205156830334e-05, + "loss": 1.0059, + "step": 7388 + }, + { + "epoch": 0.48, + "learning_rate": 1.8435990467807383e-05, + "loss": 1.0985, + "step": 7389 + }, + { + "epoch": 0.48, + "learning_rate": 1.8435775778784432e-05, + "loss": 1.1148, + "step": 7390 + }, + { + "epoch": 0.48, + "learning_rate": 1.843556108976148e-05, + "loss": 1.0138, + "step": 7391 + }, + { + "epoch": 0.48, + "learning_rate": 1.8435346400738533e-05, + "loss": 1.1079, + "step": 7392 + }, + { + "epoch": 0.48, + "learning_rate": 1.8435131711715582e-05, + "loss": 1.0844, + "step": 7393 + }, + { + "epoch": 0.48, + "learning_rate": 1.843491702269263e-05, + "loss": 1.0859, + "step": 7394 + }, + { + "epoch": 0.48, + "learning_rate": 1.843470233366968e-05, + "loss": 1.1553, + "step": 7395 + }, + { + "epoch": 0.48, + "learning_rate": 1.843448764464673e-05, + "loss": 1.0774, + "step": 7396 + }, + { + "epoch": 0.48, + "learning_rate": 1.843427295562378e-05, + "loss": 1.0468, + "step": 7397 + }, + { + "epoch": 0.48, + "learning_rate": 1.843405826660083e-05, + "loss": 1.0895, + "step": 7398 + }, + { + "epoch": 0.48, + "learning_rate": 1.843384357757788e-05, + "loss": 1.1231, + "step": 7399 + }, + { + "epoch": 0.48, + "learning_rate": 1.843362888855493e-05, + "loss": 1.0279, + "step": 7400 + }, + { + "epoch": 0.48, + "learning_rate": 1.843341419953198e-05, + "loss": 1.0648, + "step": 7401 + }, + { + "epoch": 0.48, + "learning_rate": 1.843319951050903e-05, + "loss": 1.0517, + "step": 7402 + }, + { + "epoch": 0.48, + "learning_rate": 1.843298482148608e-05, + "loss": 1.1282, + "step": 7403 + }, + { + "epoch": 0.48, + "learning_rate": 1.8432770132463128e-05, + "loss": 1.1065, + "step": 7404 + }, + { + "epoch": 0.48, + "learning_rate": 1.843255544344018e-05, + "loss": 1.1637, + "step": 7405 + }, + { + "epoch": 0.48, + "learning_rate": 1.843234075441723e-05, + "loss": 1.0732, + "step": 7406 + }, + { + "epoch": 0.48, + "learning_rate": 1.8432126065394278e-05, + "loss": 1.0488, + "step": 7407 + }, + { + "epoch": 0.48, + "learning_rate": 1.8431911376371327e-05, + "loss": 1.1186, + "step": 7408 + }, + { + "epoch": 0.48, + "learning_rate": 1.8431696687348376e-05, + "loss": 1.1899, + "step": 7409 + }, + { + "epoch": 0.48, + "learning_rate": 1.8431481998325428e-05, + "loss": 1.0114, + "step": 7410 + }, + { + "epoch": 0.48, + "learning_rate": 1.8431267309302477e-05, + "loss": 1.0421, + "step": 7411 + }, + { + "epoch": 0.48, + "learning_rate": 1.8431052620279526e-05, + "loss": 1.0632, + "step": 7412 + }, + { + "epoch": 0.48, + "learning_rate": 1.8430837931256575e-05, + "loss": 1.0818, + "step": 7413 + }, + { + "epoch": 0.48, + "learning_rate": 1.8430623242233624e-05, + "loss": 1.1613, + "step": 7414 + }, + { + "epoch": 0.48, + "learning_rate": 1.8430408553210676e-05, + "loss": 1.1081, + "step": 7415 + }, + { + "epoch": 0.48, + "learning_rate": 1.8430193864187725e-05, + "loss": 1.1338, + "step": 7416 + }, + { + "epoch": 0.48, + "learning_rate": 1.8429979175164778e-05, + "loss": 0.9952, + "step": 7417 + }, + { + "epoch": 0.48, + "learning_rate": 1.8429764486141827e-05, + "loss": 0.9675, + "step": 7418 + }, + { + "epoch": 0.48, + "learning_rate": 1.8429549797118876e-05, + "loss": 1.0787, + "step": 7419 + }, + { + "epoch": 0.48, + "learning_rate": 1.8429335108095925e-05, + "loss": 1.0453, + "step": 7420 + }, + { + "epoch": 0.48, + "learning_rate": 1.8429120419072974e-05, + "loss": 1.112, + "step": 7421 + }, + { + "epoch": 0.48, + "learning_rate": 1.8428905730050022e-05, + "loss": 1.1207, + "step": 7422 + }, + { + "epoch": 0.48, + "learning_rate": 1.8428691041027075e-05, + "loss": 1.079, + "step": 7423 + }, + { + "epoch": 0.48, + "learning_rate": 1.8428476352004124e-05, + "loss": 1.1112, + "step": 7424 + }, + { + "epoch": 0.48, + "learning_rate": 1.8428261662981173e-05, + "loss": 1.0747, + "step": 7425 + }, + { + "epoch": 0.48, + "learning_rate": 1.8428046973958222e-05, + "loss": 1.1337, + "step": 7426 + }, + { + "epoch": 0.48, + "learning_rate": 1.842783228493527e-05, + "loss": 1.0687, + "step": 7427 + }, + { + "epoch": 0.48, + "learning_rate": 1.8427617595912323e-05, + "loss": 1.0614, + "step": 7428 + }, + { + "epoch": 0.48, + "learning_rate": 1.8427402906889372e-05, + "loss": 1.0084, + "step": 7429 + }, + { + "epoch": 0.48, + "learning_rate": 1.8427188217866424e-05, + "loss": 1.0029, + "step": 7430 + }, + { + "epoch": 0.48, + "learning_rate": 1.8426973528843473e-05, + "loss": 1.0827, + "step": 7431 + }, + { + "epoch": 0.48, + "learning_rate": 1.8426758839820522e-05, + "loss": 1.0543, + "step": 7432 + }, + { + "epoch": 0.48, + "learning_rate": 1.842654415079757e-05, + "loss": 1.2042, + "step": 7433 + }, + { + "epoch": 0.48, + "learning_rate": 1.842632946177462e-05, + "loss": 1.0834, + "step": 7434 + }, + { + "epoch": 0.48, + "learning_rate": 1.8426114772751673e-05, + "loss": 1.056, + "step": 7435 + }, + { + "epoch": 0.48, + "learning_rate": 1.842590008372872e-05, + "loss": 1.0945, + "step": 7436 + }, + { + "epoch": 0.48, + "learning_rate": 1.842568539470577e-05, + "loss": 1.055, + "step": 7437 + }, + { + "epoch": 0.48, + "learning_rate": 1.842547070568282e-05, + "loss": 1.0803, + "step": 7438 + }, + { + "epoch": 0.48, + "learning_rate": 1.842525601665987e-05, + "loss": 1.065, + "step": 7439 + }, + { + "epoch": 0.48, + "learning_rate": 1.8425041327636917e-05, + "loss": 1.0657, + "step": 7440 + }, + { + "epoch": 0.48, + "learning_rate": 1.842482663861397e-05, + "loss": 1.1056, + "step": 7441 + }, + { + "epoch": 0.48, + "learning_rate": 1.842461194959102e-05, + "loss": 1.1319, + "step": 7442 + }, + { + "epoch": 0.48, + "learning_rate": 1.8424397260568068e-05, + "loss": 1.1788, + "step": 7443 + }, + { + "epoch": 0.48, + "learning_rate": 1.842418257154512e-05, + "loss": 1.1464, + "step": 7444 + }, + { + "epoch": 0.48, + "learning_rate": 1.842396788252217e-05, + "loss": 1.1829, + "step": 7445 + }, + { + "epoch": 0.48, + "learning_rate": 1.8423753193499218e-05, + "loss": 1.0711, + "step": 7446 + }, + { + "epoch": 0.48, + "learning_rate": 1.8423538504476267e-05, + "loss": 1.1409, + "step": 7447 + }, + { + "epoch": 0.48, + "learning_rate": 1.842332381545332e-05, + "loss": 1.121, + "step": 7448 + }, + { + "epoch": 0.48, + "learning_rate": 1.8423109126430368e-05, + "loss": 1.1313, + "step": 7449 + }, + { + "epoch": 0.48, + "learning_rate": 1.8422894437407417e-05, + "loss": 1.0716, + "step": 7450 + }, + { + "epoch": 0.48, + "learning_rate": 1.8422679748384466e-05, + "loss": 1.1557, + "step": 7451 + }, + { + "epoch": 0.48, + "learning_rate": 1.8422465059361515e-05, + "loss": 1.0424, + "step": 7452 + }, + { + "epoch": 0.48, + "learning_rate": 1.8422250370338564e-05, + "loss": 1.0844, + "step": 7453 + }, + { + "epoch": 0.48, + "learning_rate": 1.8422035681315616e-05, + "loss": 1.059, + "step": 7454 + }, + { + "epoch": 0.48, + "learning_rate": 1.8421820992292665e-05, + "loss": 1.1194, + "step": 7455 + }, + { + "epoch": 0.48, + "learning_rate": 1.8421606303269714e-05, + "loss": 1.1269, + "step": 7456 + }, + { + "epoch": 0.48, + "learning_rate": 1.8421391614246763e-05, + "loss": 1.0028, + "step": 7457 + }, + { + "epoch": 0.48, + "learning_rate": 1.8421176925223812e-05, + "loss": 1.0713, + "step": 7458 + }, + { + "epoch": 0.48, + "learning_rate": 1.8420962236200865e-05, + "loss": 1.0565, + "step": 7459 + }, + { + "epoch": 0.48, + "learning_rate": 1.8420747547177914e-05, + "loss": 1.009, + "step": 7460 + }, + { + "epoch": 0.48, + "learning_rate": 1.8420532858154966e-05, + "loss": 1.0256, + "step": 7461 + }, + { + "epoch": 0.48, + "learning_rate": 1.8420318169132015e-05, + "loss": 1.2004, + "step": 7462 + }, + { + "epoch": 0.48, + "learning_rate": 1.8420103480109064e-05, + "loss": 1.2176, + "step": 7463 + }, + { + "epoch": 0.48, + "learning_rate": 1.8419888791086113e-05, + "loss": 1.0643, + "step": 7464 + }, + { + "epoch": 0.48, + "learning_rate": 1.8419674102063162e-05, + "loss": 1.1449, + "step": 7465 + }, + { + "epoch": 0.48, + "learning_rate": 1.8419459413040214e-05, + "loss": 1.0374, + "step": 7466 + }, + { + "epoch": 0.48, + "learning_rate": 1.8419244724017263e-05, + "loss": 1.0928, + "step": 7467 + }, + { + "epoch": 0.48, + "learning_rate": 1.8419030034994312e-05, + "loss": 1.022, + "step": 7468 + }, + { + "epoch": 0.48, + "learning_rate": 1.841881534597136e-05, + "loss": 1.0688, + "step": 7469 + }, + { + "epoch": 0.48, + "learning_rate": 1.841860065694841e-05, + "loss": 1.0945, + "step": 7470 + }, + { + "epoch": 0.48, + "learning_rate": 1.841838596792546e-05, + "loss": 1.0707, + "step": 7471 + }, + { + "epoch": 0.48, + "learning_rate": 1.841817127890251e-05, + "loss": 1.1038, + "step": 7472 + }, + { + "epoch": 0.48, + "learning_rate": 1.841795658987956e-05, + "loss": 1.1166, + "step": 7473 + }, + { + "epoch": 0.48, + "learning_rate": 1.8417741900856613e-05, + "loss": 1.1369, + "step": 7474 + }, + { + "epoch": 0.48, + "learning_rate": 1.841752721183366e-05, + "loss": 0.9796, + "step": 7475 + }, + { + "epoch": 0.48, + "learning_rate": 1.841731252281071e-05, + "loss": 1.121, + "step": 7476 + }, + { + "epoch": 0.48, + "learning_rate": 1.841709783378776e-05, + "loss": 1.0841, + "step": 7477 + }, + { + "epoch": 0.48, + "learning_rate": 1.841688314476481e-05, + "loss": 1.041, + "step": 7478 + }, + { + "epoch": 0.48, + "learning_rate": 1.841666845574186e-05, + "loss": 1.0502, + "step": 7479 + }, + { + "epoch": 0.48, + "learning_rate": 1.841645376671891e-05, + "loss": 0.9982, + "step": 7480 + }, + { + "epoch": 0.48, + "learning_rate": 1.841623907769596e-05, + "loss": 1.1289, + "step": 7481 + }, + { + "epoch": 0.48, + "learning_rate": 1.8416024388673008e-05, + "loss": 0.9663, + "step": 7482 + }, + { + "epoch": 0.48, + "learning_rate": 1.8415809699650057e-05, + "loss": 1.0506, + "step": 7483 + }, + { + "epoch": 0.48, + "learning_rate": 1.841559501062711e-05, + "loss": 1.0584, + "step": 7484 + }, + { + "epoch": 0.48, + "learning_rate": 1.8415380321604158e-05, + "loss": 1.1786, + "step": 7485 + }, + { + "epoch": 0.48, + "learning_rate": 1.8415165632581207e-05, + "loss": 1.1095, + "step": 7486 + }, + { + "epoch": 0.48, + "learning_rate": 1.8414950943558256e-05, + "loss": 1.0956, + "step": 7487 + }, + { + "epoch": 0.48, + "learning_rate": 1.841473625453531e-05, + "loss": 1.0323, + "step": 7488 + }, + { + "epoch": 0.48, + "learning_rate": 1.8414521565512357e-05, + "loss": 1.0896, + "step": 7489 + }, + { + "epoch": 0.48, + "learning_rate": 1.8414306876489406e-05, + "loss": 1.0523, + "step": 7490 + }, + { + "epoch": 0.48, + "learning_rate": 1.8414092187466455e-05, + "loss": 1.0359, + "step": 7491 + }, + { + "epoch": 0.48, + "learning_rate": 1.8413877498443508e-05, + "loss": 1.1985, + "step": 7492 + }, + { + "epoch": 0.48, + "learning_rate": 1.8413662809420557e-05, + "loss": 1.1406, + "step": 7493 + }, + { + "epoch": 0.48, + "learning_rate": 1.8413448120397606e-05, + "loss": 1.0748, + "step": 7494 + }, + { + "epoch": 0.48, + "learning_rate": 1.8413233431374655e-05, + "loss": 0.9416, + "step": 7495 + }, + { + "epoch": 0.48, + "learning_rate": 1.8413018742351703e-05, + "loss": 1.0518, + "step": 7496 + }, + { + "epoch": 0.48, + "learning_rate": 1.8412804053328756e-05, + "loss": 1.191, + "step": 7497 + }, + { + "epoch": 0.48, + "learning_rate": 1.8412589364305805e-05, + "loss": 1.0926, + "step": 7498 + }, + { + "epoch": 0.48, + "learning_rate": 1.8412374675282854e-05, + "loss": 1.0955, + "step": 7499 + }, + { + "epoch": 0.48, + "learning_rate": 1.8412159986259903e-05, + "loss": 1.0187, + "step": 7500 + }, + { + "epoch": 0.48, + "learning_rate": 1.8411945297236952e-05, + "loss": 1.1066, + "step": 7501 + }, + { + "epoch": 0.48, + "learning_rate": 1.8411730608214004e-05, + "loss": 1.1287, + "step": 7502 + }, + { + "epoch": 0.48, + "learning_rate": 1.8411515919191053e-05, + "loss": 1.0739, + "step": 7503 + }, + { + "epoch": 0.48, + "learning_rate": 1.8411301230168105e-05, + "loss": 1.0328, + "step": 7504 + }, + { + "epoch": 0.48, + "learning_rate": 1.8411086541145154e-05, + "loss": 1.1727, + "step": 7505 + }, + { + "epoch": 0.48, + "learning_rate": 1.8410871852122203e-05, + "loss": 1.0682, + "step": 7506 + }, + { + "epoch": 0.48, + "learning_rate": 1.8410657163099252e-05, + "loss": 1.1117, + "step": 7507 + }, + { + "epoch": 0.48, + "learning_rate": 1.84104424740763e-05, + "loss": 1.0659, + "step": 7508 + }, + { + "epoch": 0.48, + "learning_rate": 1.841022778505335e-05, + "loss": 1.0539, + "step": 7509 + }, + { + "epoch": 0.48, + "learning_rate": 1.8410013096030403e-05, + "loss": 1.032, + "step": 7510 + }, + { + "epoch": 0.48, + "learning_rate": 1.840979840700745e-05, + "loss": 1.0853, + "step": 7511 + }, + { + "epoch": 0.48, + "learning_rate": 1.84095837179845e-05, + "loss": 1.0586, + "step": 7512 + }, + { + "epoch": 0.48, + "learning_rate": 1.840936902896155e-05, + "loss": 1.1443, + "step": 7513 + }, + { + "epoch": 0.48, + "learning_rate": 1.84091543399386e-05, + "loss": 1.1176, + "step": 7514 + }, + { + "epoch": 0.48, + "learning_rate": 1.840893965091565e-05, + "loss": 1.1274, + "step": 7515 + }, + { + "epoch": 0.48, + "learning_rate": 1.84087249618927e-05, + "loss": 1.0902, + "step": 7516 + }, + { + "epoch": 0.48, + "learning_rate": 1.8408510272869752e-05, + "loss": 1.1779, + "step": 7517 + }, + { + "epoch": 0.48, + "learning_rate": 1.84082955838468e-05, + "loss": 1.0937, + "step": 7518 + }, + { + "epoch": 0.48, + "learning_rate": 1.840808089482385e-05, + "loss": 1.0733, + "step": 7519 + }, + { + "epoch": 0.48, + "learning_rate": 1.84078662058009e-05, + "loss": 1.0135, + "step": 7520 + }, + { + "epoch": 0.48, + "learning_rate": 1.8407651516777948e-05, + "loss": 1.1373, + "step": 7521 + }, + { + "epoch": 0.48, + "learning_rate": 1.8407436827755e-05, + "loss": 1.0053, + "step": 7522 + }, + { + "epoch": 0.48, + "learning_rate": 1.840722213873205e-05, + "loss": 1.1146, + "step": 7523 + }, + { + "epoch": 0.48, + "learning_rate": 1.8407007449709098e-05, + "loss": 1.0602, + "step": 7524 + }, + { + "epoch": 0.48, + "learning_rate": 1.8406792760686147e-05, + "loss": 1.0803, + "step": 7525 + }, + { + "epoch": 0.48, + "learning_rate": 1.8406578071663196e-05, + "loss": 1.142, + "step": 7526 + }, + { + "epoch": 0.48, + "learning_rate": 1.8406363382640245e-05, + "loss": 1.0316, + "step": 7527 + }, + { + "epoch": 0.48, + "learning_rate": 1.8406148693617297e-05, + "loss": 0.9831, + "step": 7528 + }, + { + "epoch": 0.48, + "learning_rate": 1.8405934004594346e-05, + "loss": 1.2131, + "step": 7529 + }, + { + "epoch": 0.48, + "learning_rate": 1.8405719315571395e-05, + "loss": 1.1046, + "step": 7530 + }, + { + "epoch": 0.48, + "learning_rate": 1.8405504626548448e-05, + "loss": 1.0996, + "step": 7531 + }, + { + "epoch": 0.48, + "learning_rate": 1.8405289937525497e-05, + "loss": 1.1003, + "step": 7532 + }, + { + "epoch": 0.48, + "learning_rate": 1.8405075248502546e-05, + "loss": 1.1029, + "step": 7533 + }, + { + "epoch": 0.48, + "learning_rate": 1.8404860559479595e-05, + "loss": 1.0864, + "step": 7534 + }, + { + "epoch": 0.48, + "learning_rate": 1.8404645870456647e-05, + "loss": 1.0066, + "step": 7535 + }, + { + "epoch": 0.48, + "learning_rate": 1.8404431181433696e-05, + "loss": 1.0811, + "step": 7536 + }, + { + "epoch": 0.48, + "learning_rate": 1.8404216492410745e-05, + "loss": 1.0939, + "step": 7537 + }, + { + "epoch": 0.48, + "learning_rate": 1.8404001803387794e-05, + "loss": 1.0818, + "step": 7538 + }, + { + "epoch": 0.49, + "learning_rate": 1.8403787114364843e-05, + "loss": 1.1206, + "step": 7539 + }, + { + "epoch": 0.49, + "learning_rate": 1.8403572425341892e-05, + "loss": 1.1297, + "step": 7540 + }, + { + "epoch": 0.49, + "learning_rate": 1.8403357736318944e-05, + "loss": 1.0869, + "step": 7541 + }, + { + "epoch": 0.49, + "learning_rate": 1.8403143047295993e-05, + "loss": 1.0705, + "step": 7542 + }, + { + "epoch": 0.49, + "learning_rate": 1.8402928358273042e-05, + "loss": 1.0542, + "step": 7543 + }, + { + "epoch": 0.49, + "learning_rate": 1.840271366925009e-05, + "loss": 1.0286, + "step": 7544 + }, + { + "epoch": 0.49, + "learning_rate": 1.840249898022714e-05, + "loss": 1.0683, + "step": 7545 + }, + { + "epoch": 0.49, + "learning_rate": 1.8402284291204192e-05, + "loss": 1.059, + "step": 7546 + }, + { + "epoch": 0.49, + "learning_rate": 1.840206960218124e-05, + "loss": 1.1201, + "step": 7547 + }, + { + "epoch": 0.49, + "learning_rate": 1.8401854913158294e-05, + "loss": 1.1157, + "step": 7548 + }, + { + "epoch": 0.49, + "learning_rate": 1.8401640224135343e-05, + "loss": 1.1022, + "step": 7549 + }, + { + "epoch": 0.49, + "learning_rate": 1.840142553511239e-05, + "loss": 1.1253, + "step": 7550 + }, + { + "epoch": 0.49, + "learning_rate": 1.840121084608944e-05, + "loss": 1.1711, + "step": 7551 + }, + { + "epoch": 0.49, + "learning_rate": 1.840099615706649e-05, + "loss": 1.0788, + "step": 7552 + }, + { + "epoch": 0.49, + "learning_rate": 1.8400781468043542e-05, + "loss": 1.0754, + "step": 7553 + }, + { + "epoch": 0.49, + "learning_rate": 1.840056677902059e-05, + "loss": 1.0668, + "step": 7554 + }, + { + "epoch": 0.49, + "learning_rate": 1.840035208999764e-05, + "loss": 1.108, + "step": 7555 + }, + { + "epoch": 0.49, + "learning_rate": 1.840013740097469e-05, + "loss": 1.1421, + "step": 7556 + }, + { + "epoch": 0.49, + "learning_rate": 1.8399922711951738e-05, + "loss": 1.0875, + "step": 7557 + }, + { + "epoch": 0.49, + "learning_rate": 1.8399708022928787e-05, + "loss": 1.032, + "step": 7558 + }, + { + "epoch": 0.49, + "learning_rate": 1.839949333390584e-05, + "loss": 1.1291, + "step": 7559 + }, + { + "epoch": 0.49, + "learning_rate": 1.8399278644882888e-05, + "loss": 1.1273, + "step": 7560 + }, + { + "epoch": 0.49, + "learning_rate": 1.839906395585994e-05, + "loss": 1.0177, + "step": 7561 + }, + { + "epoch": 0.49, + "learning_rate": 1.839884926683699e-05, + "loss": 1.0656, + "step": 7562 + }, + { + "epoch": 0.49, + "learning_rate": 1.839863457781404e-05, + "loss": 0.9782, + "step": 7563 + }, + { + "epoch": 0.49, + "learning_rate": 1.8398419888791087e-05, + "loss": 1.1352, + "step": 7564 + }, + { + "epoch": 0.49, + "learning_rate": 1.8398205199768136e-05, + "loss": 1.127, + "step": 7565 + }, + { + "epoch": 0.49, + "learning_rate": 1.839799051074519e-05, + "loss": 1.0627, + "step": 7566 + }, + { + "epoch": 0.49, + "learning_rate": 1.8397775821722238e-05, + "loss": 1.0357, + "step": 7567 + }, + { + "epoch": 0.49, + "learning_rate": 1.8397561132699287e-05, + "loss": 1.0966, + "step": 7568 + }, + { + "epoch": 0.49, + "learning_rate": 1.8397346443676336e-05, + "loss": 1.1573, + "step": 7569 + }, + { + "epoch": 0.49, + "learning_rate": 1.8397131754653384e-05, + "loss": 1.0927, + "step": 7570 + }, + { + "epoch": 0.49, + "learning_rate": 1.8396917065630437e-05, + "loss": 1.1162, + "step": 7571 + }, + { + "epoch": 0.49, + "learning_rate": 1.8396702376607486e-05, + "loss": 1.0993, + "step": 7572 + }, + { + "epoch": 0.49, + "learning_rate": 1.8396487687584535e-05, + "loss": 1.1245, + "step": 7573 + }, + { + "epoch": 0.49, + "learning_rate": 1.8396272998561584e-05, + "loss": 0.9934, + "step": 7574 + }, + { + "epoch": 0.49, + "learning_rate": 1.8396058309538636e-05, + "loss": 1.0563, + "step": 7575 + }, + { + "epoch": 0.49, + "learning_rate": 1.8395843620515685e-05, + "loss": 1.0719, + "step": 7576 + }, + { + "epoch": 0.49, + "learning_rate": 1.8395628931492734e-05, + "loss": 1.069, + "step": 7577 + }, + { + "epoch": 0.49, + "learning_rate": 1.8395414242469783e-05, + "loss": 1.0846, + "step": 7578 + }, + { + "epoch": 0.49, + "learning_rate": 1.8395199553446835e-05, + "loss": 1.1338, + "step": 7579 + }, + { + "epoch": 0.49, + "learning_rate": 1.8394984864423884e-05, + "loss": 1.081, + "step": 7580 + }, + { + "epoch": 0.49, + "learning_rate": 1.8394770175400933e-05, + "loss": 1.0288, + "step": 7581 + }, + { + "epoch": 0.49, + "learning_rate": 1.8394555486377982e-05, + "loss": 1.1235, + "step": 7582 + }, + { + "epoch": 0.49, + "learning_rate": 1.839434079735503e-05, + "loss": 1.0841, + "step": 7583 + }, + { + "epoch": 0.49, + "learning_rate": 1.8394126108332084e-05, + "loss": 1.0973, + "step": 7584 + }, + { + "epoch": 0.49, + "learning_rate": 1.8393911419309133e-05, + "loss": 0.9496, + "step": 7585 + }, + { + "epoch": 0.49, + "learning_rate": 1.839369673028618e-05, + "loss": 1.1864, + "step": 7586 + }, + { + "epoch": 0.49, + "learning_rate": 1.839348204126323e-05, + "loss": 1.0261, + "step": 7587 + }, + { + "epoch": 0.49, + "learning_rate": 1.839326735224028e-05, + "loss": 1.0032, + "step": 7588 + }, + { + "epoch": 0.49, + "learning_rate": 1.8393052663217332e-05, + "loss": 1.1676, + "step": 7589 + }, + { + "epoch": 0.49, + "learning_rate": 1.839283797419438e-05, + "loss": 1.0831, + "step": 7590 + }, + { + "epoch": 0.49, + "learning_rate": 1.8392623285171433e-05, + "loss": 1.1554, + "step": 7591 + }, + { + "epoch": 0.49, + "learning_rate": 1.8392408596148482e-05, + "loss": 1.1675, + "step": 7592 + }, + { + "epoch": 0.49, + "learning_rate": 1.839219390712553e-05, + "loss": 1.0372, + "step": 7593 + }, + { + "epoch": 0.49, + "learning_rate": 1.839197921810258e-05, + "loss": 1.1639, + "step": 7594 + }, + { + "epoch": 0.49, + "learning_rate": 1.839176452907963e-05, + "loss": 1.1624, + "step": 7595 + }, + { + "epoch": 0.49, + "learning_rate": 1.8391549840056678e-05, + "loss": 1.1448, + "step": 7596 + }, + { + "epoch": 0.49, + "learning_rate": 1.839133515103373e-05, + "loss": 1.1166, + "step": 7597 + }, + { + "epoch": 0.49, + "learning_rate": 1.839112046201078e-05, + "loss": 1.1389, + "step": 7598 + }, + { + "epoch": 0.49, + "learning_rate": 1.8390905772987828e-05, + "loss": 1.0639, + "step": 7599 + }, + { + "epoch": 0.49, + "learning_rate": 1.8390691083964877e-05, + "loss": 1.1488, + "step": 7600 + }, + { + "epoch": 0.49, + "learning_rate": 1.8390476394941926e-05, + "loss": 1.1176, + "step": 7601 + }, + { + "epoch": 0.49, + "learning_rate": 1.839026170591898e-05, + "loss": 1.1533, + "step": 7602 + }, + { + "epoch": 0.49, + "learning_rate": 1.8390047016896027e-05, + "loss": 1.0302, + "step": 7603 + }, + { + "epoch": 0.49, + "learning_rate": 1.838983232787308e-05, + "loss": 1.1526, + "step": 7604 + }, + { + "epoch": 0.49, + "learning_rate": 1.838961763885013e-05, + "loss": 1.0034, + "step": 7605 + }, + { + "epoch": 0.49, + "learning_rate": 1.8389402949827178e-05, + "loss": 1.1932, + "step": 7606 + }, + { + "epoch": 0.49, + "learning_rate": 1.8389188260804227e-05, + "loss": 1.1866, + "step": 7607 + }, + { + "epoch": 0.49, + "learning_rate": 1.8388973571781276e-05, + "loss": 1.0894, + "step": 7608 + }, + { + "epoch": 0.49, + "learning_rate": 1.8388758882758328e-05, + "loss": 1.1281, + "step": 7609 + }, + { + "epoch": 0.49, + "learning_rate": 1.8388544193735377e-05, + "loss": 1.1185, + "step": 7610 + }, + { + "epoch": 0.49, + "learning_rate": 1.8388329504712426e-05, + "loss": 1.0474, + "step": 7611 + }, + { + "epoch": 0.49, + "learning_rate": 1.8388114815689475e-05, + "loss": 1.1038, + "step": 7612 + }, + { + "epoch": 0.49, + "learning_rate": 1.8387900126666524e-05, + "loss": 1.0685, + "step": 7613 + }, + { + "epoch": 0.49, + "learning_rate": 1.8387685437643573e-05, + "loss": 1.011, + "step": 7614 + }, + { + "epoch": 0.49, + "learning_rate": 1.8387470748620625e-05, + "loss": 1.1534, + "step": 7615 + }, + { + "epoch": 0.49, + "learning_rate": 1.8387256059597674e-05, + "loss": 1.0621, + "step": 7616 + }, + { + "epoch": 0.49, + "learning_rate": 1.8387041370574723e-05, + "loss": 1.0527, + "step": 7617 + }, + { + "epoch": 0.49, + "learning_rate": 1.8386826681551775e-05, + "loss": 1.1324, + "step": 7618 + }, + { + "epoch": 0.49, + "learning_rate": 1.8386611992528824e-05, + "loss": 1.0514, + "step": 7619 + }, + { + "epoch": 0.49, + "learning_rate": 1.8386397303505873e-05, + "loss": 1.0771, + "step": 7620 + }, + { + "epoch": 0.49, + "learning_rate": 1.8386182614482922e-05, + "loss": 1.1337, + "step": 7621 + }, + { + "epoch": 0.49, + "learning_rate": 1.8385967925459975e-05, + "loss": 1.0585, + "step": 7622 + }, + { + "epoch": 0.49, + "learning_rate": 1.8385753236437024e-05, + "loss": 1.1078, + "step": 7623 + }, + { + "epoch": 0.49, + "learning_rate": 1.8385538547414073e-05, + "loss": 1.1238, + "step": 7624 + }, + { + "epoch": 0.49, + "learning_rate": 1.838532385839112e-05, + "loss": 1.0049, + "step": 7625 + }, + { + "epoch": 0.49, + "learning_rate": 1.838510916936817e-05, + "loss": 1.0602, + "step": 7626 + }, + { + "epoch": 0.49, + "learning_rate": 1.838489448034522e-05, + "loss": 1.0306, + "step": 7627 + }, + { + "epoch": 0.49, + "learning_rate": 1.8384679791322272e-05, + "loss": 1.0554, + "step": 7628 + }, + { + "epoch": 0.49, + "learning_rate": 1.838446510229932e-05, + "loss": 0.9928, + "step": 7629 + }, + { + "epoch": 0.49, + "learning_rate": 1.838425041327637e-05, + "loss": 1.0496, + "step": 7630 + }, + { + "epoch": 0.49, + "learning_rate": 1.838403572425342e-05, + "loss": 1.0376, + "step": 7631 + }, + { + "epoch": 0.49, + "learning_rate": 1.8383821035230468e-05, + "loss": 1.0643, + "step": 7632 + }, + { + "epoch": 0.49, + "learning_rate": 1.838360634620752e-05, + "loss": 1.0833, + "step": 7633 + }, + { + "epoch": 0.49, + "learning_rate": 1.838339165718457e-05, + "loss": 1.1532, + "step": 7634 + }, + { + "epoch": 0.49, + "learning_rate": 1.838317696816162e-05, + "loss": 1.043, + "step": 7635 + }, + { + "epoch": 0.49, + "learning_rate": 1.838296227913867e-05, + "loss": 1.1305, + "step": 7636 + }, + { + "epoch": 0.49, + "learning_rate": 1.838274759011572e-05, + "loss": 1.0833, + "step": 7637 + }, + { + "epoch": 0.49, + "learning_rate": 1.838253290109277e-05, + "loss": 0.9908, + "step": 7638 + }, + { + "epoch": 0.49, + "learning_rate": 1.8382318212069817e-05, + "loss": 1.133, + "step": 7639 + }, + { + "epoch": 0.49, + "learning_rate": 1.838210352304687e-05, + "loss": 1.1356, + "step": 7640 + }, + { + "epoch": 0.49, + "learning_rate": 1.838188883402392e-05, + "loss": 1.0949, + "step": 7641 + }, + { + "epoch": 0.49, + "learning_rate": 1.8381674145000968e-05, + "loss": 1.086, + "step": 7642 + }, + { + "epoch": 0.49, + "learning_rate": 1.8381459455978017e-05, + "loss": 1.1064, + "step": 7643 + }, + { + "epoch": 0.49, + "learning_rate": 1.8381244766955065e-05, + "loss": 0.9938, + "step": 7644 + }, + { + "epoch": 0.49, + "learning_rate": 1.8381030077932114e-05, + "loss": 1.0186, + "step": 7645 + }, + { + "epoch": 0.49, + "learning_rate": 1.8380815388909167e-05, + "loss": 1.0534, + "step": 7646 + }, + { + "epoch": 0.49, + "learning_rate": 1.8380600699886216e-05, + "loss": 1.1025, + "step": 7647 + }, + { + "epoch": 0.49, + "learning_rate": 1.8380386010863268e-05, + "loss": 1.0459, + "step": 7648 + }, + { + "epoch": 0.49, + "learning_rate": 1.8380171321840317e-05, + "loss": 1.0465, + "step": 7649 + }, + { + "epoch": 0.49, + "learning_rate": 1.8379956632817366e-05, + "loss": 1.0793, + "step": 7650 + }, + { + "epoch": 0.49, + "learning_rate": 1.8379741943794415e-05, + "loss": 1.0567, + "step": 7651 + }, + { + "epoch": 0.49, + "learning_rate": 1.8379527254771464e-05, + "loss": 1.0869, + "step": 7652 + }, + { + "epoch": 0.49, + "learning_rate": 1.8379312565748516e-05, + "loss": 1.169, + "step": 7653 + }, + { + "epoch": 0.49, + "learning_rate": 1.8379097876725565e-05, + "loss": 1.0914, + "step": 7654 + }, + { + "epoch": 0.49, + "learning_rate": 1.8378883187702614e-05, + "loss": 1.0511, + "step": 7655 + }, + { + "epoch": 0.49, + "learning_rate": 1.8378668498679663e-05, + "loss": 1.1649, + "step": 7656 + }, + { + "epoch": 0.49, + "learning_rate": 1.8378453809656712e-05, + "loss": 1.1897, + "step": 7657 + }, + { + "epoch": 0.49, + "learning_rate": 1.8378239120633765e-05, + "loss": 1.0477, + "step": 7658 + }, + { + "epoch": 0.49, + "learning_rate": 1.8378024431610814e-05, + "loss": 1.0871, + "step": 7659 + }, + { + "epoch": 0.49, + "learning_rate": 1.8377809742587862e-05, + "loss": 1.0894, + "step": 7660 + }, + { + "epoch": 0.49, + "learning_rate": 1.837759505356491e-05, + "loss": 1.1591, + "step": 7661 + }, + { + "epoch": 0.49, + "learning_rate": 1.8377380364541964e-05, + "loss": 1.0133, + "step": 7662 + }, + { + "epoch": 0.49, + "learning_rate": 1.8377165675519013e-05, + "loss": 1.1614, + "step": 7663 + }, + { + "epoch": 0.49, + "learning_rate": 1.8376950986496062e-05, + "loss": 1.174, + "step": 7664 + }, + { + "epoch": 0.49, + "learning_rate": 1.837673629747311e-05, + "loss": 1.0902, + "step": 7665 + }, + { + "epoch": 0.49, + "learning_rate": 1.8376521608450163e-05, + "loss": 1.0547, + "step": 7666 + }, + { + "epoch": 0.49, + "learning_rate": 1.8376306919427212e-05, + "loss": 1.1123, + "step": 7667 + }, + { + "epoch": 0.49, + "learning_rate": 1.837609223040426e-05, + "loss": 1.1195, + "step": 7668 + }, + { + "epoch": 0.49, + "learning_rate": 1.837587754138131e-05, + "loss": 1.0446, + "step": 7669 + }, + { + "epoch": 0.49, + "learning_rate": 1.837566285235836e-05, + "loss": 1.0823, + "step": 7670 + }, + { + "epoch": 0.49, + "learning_rate": 1.837544816333541e-05, + "loss": 1.1368, + "step": 7671 + }, + { + "epoch": 0.49, + "learning_rate": 1.837523347431246e-05, + "loss": 1.0513, + "step": 7672 + }, + { + "epoch": 0.49, + "learning_rate": 1.837501878528951e-05, + "loss": 1.1581, + "step": 7673 + }, + { + "epoch": 0.49, + "learning_rate": 1.8374804096266558e-05, + "loss": 1.1791, + "step": 7674 + }, + { + "epoch": 0.49, + "learning_rate": 1.8374589407243607e-05, + "loss": 1.1467, + "step": 7675 + }, + { + "epoch": 0.49, + "learning_rate": 1.837437471822066e-05, + "loss": 1.0998, + "step": 7676 + }, + { + "epoch": 0.49, + "learning_rate": 1.837416002919771e-05, + "loss": 1.0496, + "step": 7677 + }, + { + "epoch": 0.49, + "learning_rate": 1.837394534017476e-05, + "loss": 1.0941, + "step": 7678 + }, + { + "epoch": 0.49, + "learning_rate": 1.837373065115181e-05, + "loss": 1.0725, + "step": 7679 + }, + { + "epoch": 0.49, + "learning_rate": 1.837351596212886e-05, + "loss": 1.1428, + "step": 7680 + }, + { + "epoch": 0.49, + "learning_rate": 1.8373301273105908e-05, + "loss": 1.0621, + "step": 7681 + }, + { + "epoch": 0.49, + "learning_rate": 1.8373086584082957e-05, + "loss": 1.0633, + "step": 7682 + }, + { + "epoch": 0.49, + "learning_rate": 1.8372871895060006e-05, + "loss": 1.0498, + "step": 7683 + }, + { + "epoch": 0.49, + "learning_rate": 1.8372657206037058e-05, + "loss": 1.0917, + "step": 7684 + }, + { + "epoch": 0.49, + "learning_rate": 1.8372442517014107e-05, + "loss": 1.0146, + "step": 7685 + }, + { + "epoch": 0.49, + "learning_rate": 1.8372227827991156e-05, + "loss": 1.209, + "step": 7686 + }, + { + "epoch": 0.49, + "learning_rate": 1.8372013138968205e-05, + "loss": 1.0764, + "step": 7687 + }, + { + "epoch": 0.49, + "learning_rate": 1.8371798449945254e-05, + "loss": 1.0745, + "step": 7688 + }, + { + "epoch": 0.49, + "learning_rate": 1.8371583760922306e-05, + "loss": 1.1772, + "step": 7689 + }, + { + "epoch": 0.49, + "learning_rate": 1.8371369071899355e-05, + "loss": 1.1023, + "step": 7690 + }, + { + "epoch": 0.49, + "learning_rate": 1.8371154382876407e-05, + "loss": 1.1037, + "step": 7691 + }, + { + "epoch": 0.49, + "learning_rate": 1.8370939693853456e-05, + "loss": 1.0788, + "step": 7692 + }, + { + "epoch": 0.49, + "learning_rate": 1.8370725004830505e-05, + "loss": 1.1206, + "step": 7693 + }, + { + "epoch": 0.5, + "learning_rate": 1.8370510315807554e-05, + "loss": 1.1119, + "step": 7694 + }, + { + "epoch": 0.5, + "learning_rate": 1.8370295626784603e-05, + "loss": 1.1141, + "step": 7695 + }, + { + "epoch": 0.5, + "learning_rate": 1.8370080937761656e-05, + "loss": 1.1459, + "step": 7696 + }, + { + "epoch": 0.5, + "learning_rate": 1.8369866248738705e-05, + "loss": 1.0489, + "step": 7697 + }, + { + "epoch": 0.5, + "learning_rate": 1.8369651559715754e-05, + "loss": 1.1155, + "step": 7698 + }, + { + "epoch": 0.5, + "learning_rate": 1.8369436870692803e-05, + "loss": 1.0376, + "step": 7699 + }, + { + "epoch": 0.5, + "learning_rate": 1.836922218166985e-05, + "loss": 1.0989, + "step": 7700 + }, + { + "epoch": 0.5, + "learning_rate": 1.83690074926469e-05, + "loss": 1.0886, + "step": 7701 + }, + { + "epoch": 0.5, + "learning_rate": 1.8368792803623953e-05, + "loss": 1.1106, + "step": 7702 + }, + { + "epoch": 0.5, + "learning_rate": 1.8368578114601002e-05, + "loss": 1.1586, + "step": 7703 + }, + { + "epoch": 0.5, + "learning_rate": 1.836836342557805e-05, + "loss": 1.0968, + "step": 7704 + }, + { + "epoch": 0.5, + "learning_rate": 1.83681487365551e-05, + "loss": 1.0838, + "step": 7705 + }, + { + "epoch": 0.5, + "learning_rate": 1.8367934047532152e-05, + "loss": 1.1244, + "step": 7706 + }, + { + "epoch": 0.5, + "learning_rate": 1.83677193585092e-05, + "loss": 1.0978, + "step": 7707 + }, + { + "epoch": 0.5, + "learning_rate": 1.836750466948625e-05, + "loss": 1.1472, + "step": 7708 + }, + { + "epoch": 0.5, + "learning_rate": 1.8367289980463302e-05, + "loss": 1.0145, + "step": 7709 + }, + { + "epoch": 0.5, + "learning_rate": 1.836707529144035e-05, + "loss": 1.1652, + "step": 7710 + }, + { + "epoch": 0.5, + "learning_rate": 1.83668606024174e-05, + "loss": 0.9888, + "step": 7711 + }, + { + "epoch": 0.5, + "learning_rate": 1.836664591339445e-05, + "loss": 1.0174, + "step": 7712 + }, + { + "epoch": 0.5, + "learning_rate": 1.8366431224371498e-05, + "loss": 1.099, + "step": 7713 + }, + { + "epoch": 0.5, + "learning_rate": 1.8366216535348547e-05, + "loss": 1.0966, + "step": 7714 + }, + { + "epoch": 0.5, + "learning_rate": 1.83660018463256e-05, + "loss": 1.1156, + "step": 7715 + }, + { + "epoch": 0.5, + "learning_rate": 1.836578715730265e-05, + "loss": 1.0982, + "step": 7716 + }, + { + "epoch": 0.5, + "learning_rate": 1.8365572468279698e-05, + "loss": 1.0545, + "step": 7717 + }, + { + "epoch": 0.5, + "learning_rate": 1.8365357779256746e-05, + "loss": 1.106, + "step": 7718 + }, + { + "epoch": 0.5, + "learning_rate": 1.8365143090233795e-05, + "loss": 1.0893, + "step": 7719 + }, + { + "epoch": 0.5, + "learning_rate": 1.8364928401210848e-05, + "loss": 1.0146, + "step": 7720 + }, + { + "epoch": 0.5, + "learning_rate": 1.8364713712187897e-05, + "loss": 1.1339, + "step": 7721 + }, + { + "epoch": 0.5, + "learning_rate": 1.836449902316495e-05, + "loss": 1.0682, + "step": 7722 + }, + { + "epoch": 0.5, + "learning_rate": 1.8364284334141998e-05, + "loss": 1.1193, + "step": 7723 + }, + { + "epoch": 0.5, + "learning_rate": 1.8364069645119047e-05, + "loss": 1.0684, + "step": 7724 + }, + { + "epoch": 0.5, + "learning_rate": 1.8363854956096096e-05, + "loss": 1.2112, + "step": 7725 + }, + { + "epoch": 0.5, + "learning_rate": 1.8363640267073145e-05, + "loss": 1.1138, + "step": 7726 + }, + { + "epoch": 0.5, + "learning_rate": 1.8363425578050197e-05, + "loss": 1.0896, + "step": 7727 + }, + { + "epoch": 0.5, + "learning_rate": 1.8363210889027246e-05, + "loss": 1.058, + "step": 7728 + }, + { + "epoch": 0.5, + "learning_rate": 1.8362996200004295e-05, + "loss": 1.1491, + "step": 7729 + }, + { + "epoch": 0.5, + "learning_rate": 1.8362781510981344e-05, + "loss": 1.0254, + "step": 7730 + }, + { + "epoch": 0.5, + "learning_rate": 1.8362566821958393e-05, + "loss": 1.0906, + "step": 7731 + }, + { + "epoch": 0.5, + "learning_rate": 1.8362352132935442e-05, + "loss": 1.0392, + "step": 7732 + }, + { + "epoch": 0.5, + "learning_rate": 1.8362137443912495e-05, + "loss": 1.0766, + "step": 7733 + }, + { + "epoch": 0.5, + "learning_rate": 1.8361922754889543e-05, + "loss": 1.0264, + "step": 7734 + }, + { + "epoch": 0.5, + "learning_rate": 1.8361708065866596e-05, + "loss": 1.1818, + "step": 7735 + }, + { + "epoch": 0.5, + "learning_rate": 1.8361493376843645e-05, + "loss": 1.1331, + "step": 7736 + }, + { + "epoch": 0.5, + "learning_rate": 1.8361278687820694e-05, + "loss": 1.1478, + "step": 7737 + }, + { + "epoch": 0.5, + "learning_rate": 1.8361063998797743e-05, + "loss": 1.0986, + "step": 7738 + }, + { + "epoch": 0.5, + "learning_rate": 1.8360849309774792e-05, + "loss": 1.044, + "step": 7739 + }, + { + "epoch": 0.5, + "learning_rate": 1.8360634620751844e-05, + "loss": 1.0471, + "step": 7740 + }, + { + "epoch": 0.5, + "learning_rate": 1.8360419931728893e-05, + "loss": 1.1584, + "step": 7741 + }, + { + "epoch": 0.5, + "learning_rate": 1.8360205242705942e-05, + "loss": 1.0272, + "step": 7742 + }, + { + "epoch": 0.5, + "learning_rate": 1.835999055368299e-05, + "loss": 1.0617, + "step": 7743 + }, + { + "epoch": 0.5, + "learning_rate": 1.835977586466004e-05, + "loss": 1.1766, + "step": 7744 + }, + { + "epoch": 0.5, + "learning_rate": 1.8359561175637092e-05, + "loss": 1.1076, + "step": 7745 + }, + { + "epoch": 0.5, + "learning_rate": 1.835934648661414e-05, + "loss": 1.0786, + "step": 7746 + }, + { + "epoch": 0.5, + "learning_rate": 1.835913179759119e-05, + "loss": 0.9801, + "step": 7747 + }, + { + "epoch": 0.5, + "learning_rate": 1.835891710856824e-05, + "loss": 1.1206, + "step": 7748 + }, + { + "epoch": 0.5, + "learning_rate": 1.835870241954529e-05, + "loss": 1.2167, + "step": 7749 + }, + { + "epoch": 0.5, + "learning_rate": 1.835848773052234e-05, + "loss": 1.0404, + "step": 7750 + }, + { + "epoch": 0.5, + "learning_rate": 1.835827304149939e-05, + "loss": 1.0504, + "step": 7751 + }, + { + "epoch": 0.5, + "learning_rate": 1.835805835247644e-05, + "loss": 1.2506, + "step": 7752 + }, + { + "epoch": 0.5, + "learning_rate": 1.835784366345349e-05, + "loss": 1.1409, + "step": 7753 + }, + { + "epoch": 0.5, + "learning_rate": 1.835762897443054e-05, + "loss": 1.0868, + "step": 7754 + }, + { + "epoch": 0.5, + "learning_rate": 1.835741428540759e-05, + "loss": 1.0837, + "step": 7755 + }, + { + "epoch": 0.5, + "learning_rate": 1.8357199596384638e-05, + "loss": 1.1579, + "step": 7756 + }, + { + "epoch": 0.5, + "learning_rate": 1.8356984907361687e-05, + "loss": 1.1233, + "step": 7757 + }, + { + "epoch": 0.5, + "learning_rate": 1.835677021833874e-05, + "loss": 1.1478, + "step": 7758 + }, + { + "epoch": 0.5, + "learning_rate": 1.8356555529315788e-05, + "loss": 1.0881, + "step": 7759 + }, + { + "epoch": 0.5, + "learning_rate": 1.8356340840292837e-05, + "loss": 1.1422, + "step": 7760 + }, + { + "epoch": 0.5, + "learning_rate": 1.8356126151269886e-05, + "loss": 1.079, + "step": 7761 + }, + { + "epoch": 0.5, + "learning_rate": 1.8355911462246935e-05, + "loss": 1.0763, + "step": 7762 + }, + { + "epoch": 0.5, + "learning_rate": 1.8355696773223984e-05, + "loss": 1.1203, + "step": 7763 + }, + { + "epoch": 0.5, + "learning_rate": 1.8355482084201036e-05, + "loss": 1.0792, + "step": 7764 + }, + { + "epoch": 0.5, + "learning_rate": 1.835526739517809e-05, + "loss": 1.0902, + "step": 7765 + }, + { + "epoch": 0.5, + "learning_rate": 1.8355052706155137e-05, + "loss": 1.179, + "step": 7766 + }, + { + "epoch": 0.5, + "learning_rate": 1.8354838017132186e-05, + "loss": 1.0704, + "step": 7767 + }, + { + "epoch": 0.5, + "learning_rate": 1.8354623328109235e-05, + "loss": 1.0881, + "step": 7768 + }, + { + "epoch": 0.5, + "learning_rate": 1.8354408639086284e-05, + "loss": 1.089, + "step": 7769 + }, + { + "epoch": 0.5, + "learning_rate": 1.8354193950063333e-05, + "loss": 1.021, + "step": 7770 + }, + { + "epoch": 0.5, + "learning_rate": 1.8353979261040386e-05, + "loss": 1.2033, + "step": 7771 + }, + { + "epoch": 0.5, + "learning_rate": 1.8353764572017435e-05, + "loss": 1.0926, + "step": 7772 + }, + { + "epoch": 0.5, + "learning_rate": 1.8353549882994484e-05, + "loss": 1.1075, + "step": 7773 + }, + { + "epoch": 0.5, + "learning_rate": 1.8353335193971533e-05, + "loss": 1.1751, + "step": 7774 + }, + { + "epoch": 0.5, + "learning_rate": 1.835312050494858e-05, + "loss": 1.0214, + "step": 7775 + }, + { + "epoch": 0.5, + "learning_rate": 1.8352905815925634e-05, + "loss": 1.0842, + "step": 7776 + }, + { + "epoch": 0.5, + "learning_rate": 1.8352691126902683e-05, + "loss": 1.1028, + "step": 7777 + }, + { + "epoch": 0.5, + "learning_rate": 1.8352476437879735e-05, + "loss": 1.0345, + "step": 7778 + }, + { + "epoch": 0.5, + "learning_rate": 1.8352261748856784e-05, + "loss": 1.1167, + "step": 7779 + }, + { + "epoch": 0.5, + "learning_rate": 1.8352047059833833e-05, + "loss": 1.0369, + "step": 7780 + }, + { + "epoch": 0.5, + "learning_rate": 1.8351832370810882e-05, + "loss": 1.0479, + "step": 7781 + }, + { + "epoch": 0.5, + "learning_rate": 1.835161768178793e-05, + "loss": 1.0857, + "step": 7782 + }, + { + "epoch": 0.5, + "learning_rate": 1.835140299276498e-05, + "loss": 1.0943, + "step": 7783 + }, + { + "epoch": 0.5, + "learning_rate": 1.8351188303742032e-05, + "loss": 1.0246, + "step": 7784 + }, + { + "epoch": 0.5, + "learning_rate": 1.835097361471908e-05, + "loss": 1.0477, + "step": 7785 + }, + { + "epoch": 0.5, + "learning_rate": 1.835075892569613e-05, + "loss": 1.1532, + "step": 7786 + }, + { + "epoch": 0.5, + "learning_rate": 1.835054423667318e-05, + "loss": 1.0619, + "step": 7787 + }, + { + "epoch": 0.5, + "learning_rate": 1.8350329547650228e-05, + "loss": 1.1139, + "step": 7788 + }, + { + "epoch": 0.5, + "learning_rate": 1.835011485862728e-05, + "loss": 1.1164, + "step": 7789 + }, + { + "epoch": 0.5, + "learning_rate": 1.834990016960433e-05, + "loss": 1.1427, + "step": 7790 + }, + { + "epoch": 0.5, + "learning_rate": 1.834968548058138e-05, + "loss": 1.121, + "step": 7791 + }, + { + "epoch": 0.5, + "learning_rate": 1.8349470791558427e-05, + "loss": 1.0525, + "step": 7792 + }, + { + "epoch": 0.5, + "learning_rate": 1.834925610253548e-05, + "loss": 1.1269, + "step": 7793 + }, + { + "epoch": 0.5, + "learning_rate": 1.834904141351253e-05, + "loss": 1.1172, + "step": 7794 + }, + { + "epoch": 0.5, + "learning_rate": 1.8348826724489578e-05, + "loss": 1.0662, + "step": 7795 + }, + { + "epoch": 0.5, + "learning_rate": 1.834861203546663e-05, + "loss": 1.2538, + "step": 7796 + }, + { + "epoch": 0.5, + "learning_rate": 1.834839734644368e-05, + "loss": 1.0521, + "step": 7797 + }, + { + "epoch": 0.5, + "learning_rate": 1.8348182657420728e-05, + "loss": 1.1588, + "step": 7798 + }, + { + "epoch": 0.5, + "learning_rate": 1.8347967968397777e-05, + "loss": 1.0946, + "step": 7799 + }, + { + "epoch": 0.5, + "learning_rate": 1.8347753279374826e-05, + "loss": 1.0183, + "step": 7800 + }, + { + "epoch": 0.5, + "learning_rate": 1.8347538590351875e-05, + "loss": 1.1325, + "step": 7801 + }, + { + "epoch": 0.5, + "learning_rate": 1.8347323901328927e-05, + "loss": 1.1001, + "step": 7802 + }, + { + "epoch": 0.5, + "learning_rate": 1.8347109212305976e-05, + "loss": 1.0936, + "step": 7803 + }, + { + "epoch": 0.5, + "learning_rate": 1.8346894523283025e-05, + "loss": 1.0462, + "step": 7804 + }, + { + "epoch": 0.5, + "learning_rate": 1.8346679834260074e-05, + "loss": 1.0403, + "step": 7805 + }, + { + "epoch": 0.5, + "learning_rate": 1.8346465145237123e-05, + "loss": 1.1005, + "step": 7806 + }, + { + "epoch": 0.5, + "learning_rate": 1.8346250456214176e-05, + "loss": 1.0777, + "step": 7807 + }, + { + "epoch": 0.5, + "learning_rate": 1.8346035767191224e-05, + "loss": 1.0509, + "step": 7808 + }, + { + "epoch": 0.5, + "learning_rate": 1.8345821078168277e-05, + "loss": 1.0659, + "step": 7809 + }, + { + "epoch": 0.5, + "learning_rate": 1.8345606389145326e-05, + "loss": 1.0888, + "step": 7810 + }, + { + "epoch": 0.5, + "learning_rate": 1.8345391700122375e-05, + "loss": 1.0718, + "step": 7811 + }, + { + "epoch": 0.5, + "learning_rate": 1.8345177011099424e-05, + "loss": 1.0103, + "step": 7812 + }, + { + "epoch": 0.5, + "learning_rate": 1.8344962322076473e-05, + "loss": 1.1039, + "step": 7813 + }, + { + "epoch": 0.5, + "learning_rate": 1.8344747633053525e-05, + "loss": 1.0999, + "step": 7814 + }, + { + "epoch": 0.5, + "learning_rate": 1.8344532944030574e-05, + "loss": 1.0961, + "step": 7815 + }, + { + "epoch": 0.5, + "learning_rate": 1.8344318255007623e-05, + "loss": 1.0683, + "step": 7816 + }, + { + "epoch": 0.5, + "learning_rate": 1.8344103565984672e-05, + "loss": 1.2129, + "step": 7817 + }, + { + "epoch": 0.5, + "learning_rate": 1.834388887696172e-05, + "loss": 1.1022, + "step": 7818 + }, + { + "epoch": 0.5, + "learning_rate": 1.834367418793877e-05, + "loss": 1.1115, + "step": 7819 + }, + { + "epoch": 0.5, + "learning_rate": 1.8343459498915822e-05, + "loss": 1.0976, + "step": 7820 + }, + { + "epoch": 0.5, + "learning_rate": 1.834324480989287e-05, + "loss": 1.0208, + "step": 7821 + }, + { + "epoch": 0.5, + "learning_rate": 1.8343030120869924e-05, + "loss": 1.0129, + "step": 7822 + }, + { + "epoch": 0.5, + "learning_rate": 1.8342815431846973e-05, + "loss": 1.1385, + "step": 7823 + }, + { + "epoch": 0.5, + "learning_rate": 1.834260074282402e-05, + "loss": 1.0927, + "step": 7824 + }, + { + "epoch": 0.5, + "learning_rate": 1.834238605380107e-05, + "loss": 1.1107, + "step": 7825 + }, + { + "epoch": 0.5, + "learning_rate": 1.834217136477812e-05, + "loss": 1.0915, + "step": 7826 + }, + { + "epoch": 0.5, + "learning_rate": 1.8341956675755172e-05, + "loss": 1.0807, + "step": 7827 + }, + { + "epoch": 0.5, + "learning_rate": 1.834174198673222e-05, + "loss": 1.0606, + "step": 7828 + }, + { + "epoch": 0.5, + "learning_rate": 1.834152729770927e-05, + "loss": 1.0672, + "step": 7829 + }, + { + "epoch": 0.5, + "learning_rate": 1.834131260868632e-05, + "loss": 1.0494, + "step": 7830 + }, + { + "epoch": 0.5, + "learning_rate": 1.8341097919663368e-05, + "loss": 1.0664, + "step": 7831 + }, + { + "epoch": 0.5, + "learning_rate": 1.834088323064042e-05, + "loss": 1.0888, + "step": 7832 + }, + { + "epoch": 0.5, + "learning_rate": 1.834066854161747e-05, + "loss": 1.1401, + "step": 7833 + }, + { + "epoch": 0.5, + "learning_rate": 1.8340453852594518e-05, + "loss": 1.0234, + "step": 7834 + }, + { + "epoch": 0.5, + "learning_rate": 1.8340239163571567e-05, + "loss": 1.0968, + "step": 7835 + }, + { + "epoch": 0.5, + "learning_rate": 1.834002447454862e-05, + "loss": 1.0526, + "step": 7836 + }, + { + "epoch": 0.5, + "learning_rate": 1.8339809785525668e-05, + "loss": 1.1345, + "step": 7837 + }, + { + "epoch": 0.5, + "learning_rate": 1.8339595096502717e-05, + "loss": 0.995, + "step": 7838 + }, + { + "epoch": 0.5, + "learning_rate": 1.8339380407479766e-05, + "loss": 1.1148, + "step": 7839 + }, + { + "epoch": 0.5, + "learning_rate": 1.833916571845682e-05, + "loss": 1.0602, + "step": 7840 + }, + { + "epoch": 0.5, + "learning_rate": 1.8338951029433867e-05, + "loss": 1.0506, + "step": 7841 + }, + { + "epoch": 0.5, + "learning_rate": 1.8338736340410916e-05, + "loss": 1.0995, + "step": 7842 + }, + { + "epoch": 0.5, + "learning_rate": 1.8338521651387965e-05, + "loss": 1.077, + "step": 7843 + }, + { + "epoch": 0.5, + "learning_rate": 1.8338306962365014e-05, + "loss": 1.084, + "step": 7844 + }, + { + "epoch": 0.5, + "learning_rate": 1.8338092273342067e-05, + "loss": 1.1587, + "step": 7845 + }, + { + "epoch": 0.5, + "learning_rate": 1.8337877584319116e-05, + "loss": 1.118, + "step": 7846 + }, + { + "epoch": 0.5, + "learning_rate": 1.8337662895296165e-05, + "loss": 1.0973, + "step": 7847 + }, + { + "epoch": 0.5, + "learning_rate": 1.8337448206273214e-05, + "loss": 1.1255, + "step": 7848 + }, + { + "epoch": 0.5, + "learning_rate": 1.8337233517250263e-05, + "loss": 1.0677, + "step": 7849 + }, + { + "epoch": 0.51, + "learning_rate": 1.833701882822731e-05, + "loss": 1.1551, + "step": 7850 + }, + { + "epoch": 0.51, + "learning_rate": 1.8336804139204364e-05, + "loss": 1.0639, + "step": 7851 + }, + { + "epoch": 0.51, + "learning_rate": 1.8336589450181416e-05, + "loss": 1.0495, + "step": 7852 + }, + { + "epoch": 0.51, + "learning_rate": 1.8336374761158465e-05, + "loss": 1.1475, + "step": 7853 + }, + { + "epoch": 0.51, + "learning_rate": 1.8336160072135514e-05, + "loss": 1.1223, + "step": 7854 + }, + { + "epoch": 0.51, + "learning_rate": 1.8335945383112563e-05, + "loss": 1.0598, + "step": 7855 + }, + { + "epoch": 0.51, + "learning_rate": 1.8335730694089612e-05, + "loss": 1.0319, + "step": 7856 + }, + { + "epoch": 0.51, + "learning_rate": 1.833551600506666e-05, + "loss": 1.0913, + "step": 7857 + }, + { + "epoch": 0.51, + "learning_rate": 1.8335301316043713e-05, + "loss": 1.1736, + "step": 7858 + }, + { + "epoch": 0.51, + "learning_rate": 1.8335086627020762e-05, + "loss": 1.015, + "step": 7859 + }, + { + "epoch": 0.51, + "learning_rate": 1.833487193799781e-05, + "loss": 1.1295, + "step": 7860 + }, + { + "epoch": 0.51, + "learning_rate": 1.833465724897486e-05, + "loss": 1.1, + "step": 7861 + }, + { + "epoch": 0.51, + "learning_rate": 1.833444255995191e-05, + "loss": 1.0696, + "step": 7862 + }, + { + "epoch": 0.51, + "learning_rate": 1.833422787092896e-05, + "loss": 1.0281, + "step": 7863 + }, + { + "epoch": 0.51, + "learning_rate": 1.833401318190601e-05, + "loss": 1.1287, + "step": 7864 + }, + { + "epoch": 0.51, + "learning_rate": 1.8333798492883063e-05, + "loss": 1.1036, + "step": 7865 + }, + { + "epoch": 0.51, + "learning_rate": 1.8333583803860112e-05, + "loss": 1.1808, + "step": 7866 + }, + { + "epoch": 0.51, + "learning_rate": 1.833336911483716e-05, + "loss": 1.1188, + "step": 7867 + }, + { + "epoch": 0.51, + "learning_rate": 1.833315442581421e-05, + "loss": 1.1073, + "step": 7868 + }, + { + "epoch": 0.51, + "learning_rate": 1.833293973679126e-05, + "loss": 1.0733, + "step": 7869 + }, + { + "epoch": 0.51, + "learning_rate": 1.8332725047768308e-05, + "loss": 1.1234, + "step": 7870 + }, + { + "epoch": 0.51, + "learning_rate": 1.833251035874536e-05, + "loss": 1.0255, + "step": 7871 + }, + { + "epoch": 0.51, + "learning_rate": 1.833229566972241e-05, + "loss": 1.0703, + "step": 7872 + }, + { + "epoch": 0.51, + "learning_rate": 1.8332080980699458e-05, + "loss": 1.1372, + "step": 7873 + }, + { + "epoch": 0.51, + "learning_rate": 1.8331866291676507e-05, + "loss": 1.0719, + "step": 7874 + }, + { + "epoch": 0.51, + "learning_rate": 1.8331651602653556e-05, + "loss": 1.0973, + "step": 7875 + }, + { + "epoch": 0.51, + "learning_rate": 1.833143691363061e-05, + "loss": 1.1327, + "step": 7876 + }, + { + "epoch": 0.51, + "learning_rate": 1.8331222224607657e-05, + "loss": 1.0943, + "step": 7877 + }, + { + "epoch": 0.51, + "learning_rate": 1.8331007535584706e-05, + "loss": 1.0804, + "step": 7878 + }, + { + "epoch": 0.51, + "learning_rate": 1.8330792846561755e-05, + "loss": 1.0865, + "step": 7879 + }, + { + "epoch": 0.51, + "learning_rate": 1.8330578157538808e-05, + "loss": 1.0773, + "step": 7880 + }, + { + "epoch": 0.51, + "learning_rate": 1.8330363468515857e-05, + "loss": 1.0147, + "step": 7881 + }, + { + "epoch": 0.51, + "learning_rate": 1.8330148779492905e-05, + "loss": 1.0442, + "step": 7882 + }, + { + "epoch": 0.51, + "learning_rate": 1.8329934090469958e-05, + "loss": 1.1314, + "step": 7883 + }, + { + "epoch": 0.51, + "learning_rate": 1.8329719401447007e-05, + "loss": 1.0411, + "step": 7884 + }, + { + "epoch": 0.51, + "learning_rate": 1.8329504712424056e-05, + "loss": 1.1481, + "step": 7885 + }, + { + "epoch": 0.51, + "learning_rate": 1.8329290023401105e-05, + "loss": 1.0456, + "step": 7886 + }, + { + "epoch": 0.51, + "learning_rate": 1.8329075334378154e-05, + "loss": 1.0336, + "step": 7887 + }, + { + "epoch": 0.51, + "learning_rate": 1.8328860645355203e-05, + "loss": 1.0952, + "step": 7888 + }, + { + "epoch": 0.51, + "learning_rate": 1.8328645956332255e-05, + "loss": 1.0292, + "step": 7889 + }, + { + "epoch": 0.51, + "learning_rate": 1.8328431267309304e-05, + "loss": 1.0323, + "step": 7890 + }, + { + "epoch": 0.51, + "learning_rate": 1.8328216578286353e-05, + "loss": 1.0762, + "step": 7891 + }, + { + "epoch": 0.51, + "learning_rate": 1.8328001889263402e-05, + "loss": 1.1018, + "step": 7892 + }, + { + "epoch": 0.51, + "learning_rate": 1.832778720024045e-05, + "loss": 1.1475, + "step": 7893 + }, + { + "epoch": 0.51, + "learning_rate": 1.8327572511217503e-05, + "loss": 1.0731, + "step": 7894 + }, + { + "epoch": 0.51, + "learning_rate": 1.8327357822194552e-05, + "loss": 1.1142, + "step": 7895 + }, + { + "epoch": 0.51, + "learning_rate": 1.8327143133171605e-05, + "loss": 1.0817, + "step": 7896 + }, + { + "epoch": 0.51, + "learning_rate": 1.8326928444148654e-05, + "loss": 0.9961, + "step": 7897 + }, + { + "epoch": 0.51, + "learning_rate": 1.8326713755125702e-05, + "loss": 1.0939, + "step": 7898 + }, + { + "epoch": 0.51, + "learning_rate": 1.832649906610275e-05, + "loss": 0.9703, + "step": 7899 + }, + { + "epoch": 0.51, + "learning_rate": 1.83262843770798e-05, + "loss": 1.1728, + "step": 7900 + }, + { + "epoch": 0.51, + "learning_rate": 1.8326069688056853e-05, + "loss": 1.1302, + "step": 7901 + }, + { + "epoch": 0.51, + "learning_rate": 1.8325854999033902e-05, + "loss": 1.0437, + "step": 7902 + }, + { + "epoch": 0.51, + "learning_rate": 1.832564031001095e-05, + "loss": 1.1035, + "step": 7903 + }, + { + "epoch": 0.51, + "learning_rate": 1.8325425620988e-05, + "loss": 1.0879, + "step": 7904 + }, + { + "epoch": 0.51, + "learning_rate": 1.832521093196505e-05, + "loss": 1.0275, + "step": 7905 + }, + { + "epoch": 0.51, + "learning_rate": 1.8324996242942098e-05, + "loss": 1.0616, + "step": 7906 + }, + { + "epoch": 0.51, + "learning_rate": 1.832478155391915e-05, + "loss": 1.091, + "step": 7907 + }, + { + "epoch": 0.51, + "learning_rate": 1.83245668648962e-05, + "loss": 1.0752, + "step": 7908 + }, + { + "epoch": 0.51, + "learning_rate": 1.832435217587325e-05, + "loss": 1.0476, + "step": 7909 + }, + { + "epoch": 0.51, + "learning_rate": 1.83241374868503e-05, + "loss": 1.0623, + "step": 7910 + }, + { + "epoch": 0.51, + "learning_rate": 1.832392279782735e-05, + "loss": 1.1686, + "step": 7911 + }, + { + "epoch": 0.51, + "learning_rate": 1.8323708108804398e-05, + "loss": 1.0363, + "step": 7912 + }, + { + "epoch": 0.51, + "learning_rate": 1.8323493419781447e-05, + "loss": 1.0638, + "step": 7913 + }, + { + "epoch": 0.51, + "learning_rate": 1.83232787307585e-05, + "loss": 1.1062, + "step": 7914 + }, + { + "epoch": 0.51, + "learning_rate": 1.832306404173555e-05, + "loss": 1.0542, + "step": 7915 + }, + { + "epoch": 0.51, + "learning_rate": 1.8322849352712597e-05, + "loss": 1.0808, + "step": 7916 + }, + { + "epoch": 0.51, + "learning_rate": 1.8322634663689646e-05, + "loss": 1.1183, + "step": 7917 + }, + { + "epoch": 0.51, + "learning_rate": 1.8322419974666695e-05, + "loss": 1.1339, + "step": 7918 + }, + { + "epoch": 0.51, + "learning_rate": 1.8322205285643748e-05, + "loss": 1.0989, + "step": 7919 + }, + { + "epoch": 0.51, + "learning_rate": 1.8321990596620797e-05, + "loss": 1.0888, + "step": 7920 + }, + { + "epoch": 0.51, + "learning_rate": 1.8321775907597846e-05, + "loss": 1.0845, + "step": 7921 + }, + { + "epoch": 0.51, + "learning_rate": 1.8321561218574895e-05, + "loss": 1.1319, + "step": 7922 + }, + { + "epoch": 0.51, + "learning_rate": 1.8321346529551947e-05, + "loss": 1.1041, + "step": 7923 + }, + { + "epoch": 0.51, + "learning_rate": 1.8321131840528996e-05, + "loss": 1.088, + "step": 7924 + }, + { + "epoch": 0.51, + "learning_rate": 1.8320917151506045e-05, + "loss": 1.0702, + "step": 7925 + }, + { + "epoch": 0.51, + "learning_rate": 1.8320702462483094e-05, + "loss": 1.0676, + "step": 7926 + }, + { + "epoch": 0.51, + "learning_rate": 1.8320487773460146e-05, + "loss": 1.0049, + "step": 7927 + }, + { + "epoch": 0.51, + "learning_rate": 1.8320273084437195e-05, + "loss": 1.1108, + "step": 7928 + }, + { + "epoch": 0.51, + "learning_rate": 1.8320058395414244e-05, + "loss": 1.1167, + "step": 7929 + }, + { + "epoch": 0.51, + "learning_rate": 1.8319843706391293e-05, + "loss": 1.1286, + "step": 7930 + }, + { + "epoch": 0.51, + "learning_rate": 1.8319629017368342e-05, + "loss": 1.0762, + "step": 7931 + }, + { + "epoch": 0.51, + "learning_rate": 1.8319414328345394e-05, + "loss": 1.0674, + "step": 7932 + }, + { + "epoch": 0.51, + "learning_rate": 1.8319199639322443e-05, + "loss": 1.1975, + "step": 7933 + }, + { + "epoch": 0.51, + "learning_rate": 1.8318984950299492e-05, + "loss": 1.0318, + "step": 7934 + }, + { + "epoch": 0.51, + "learning_rate": 1.831877026127654e-05, + "loss": 1.0979, + "step": 7935 + }, + { + "epoch": 0.51, + "learning_rate": 1.831855557225359e-05, + "loss": 1.0122, + "step": 7936 + }, + { + "epoch": 0.51, + "learning_rate": 1.831834088323064e-05, + "loss": 1.0418, + "step": 7937 + }, + { + "epoch": 0.51, + "learning_rate": 1.831812619420769e-05, + "loss": 1.1518, + "step": 7938 + }, + { + "epoch": 0.51, + "learning_rate": 1.8317911505184744e-05, + "loss": 1.123, + "step": 7939 + }, + { + "epoch": 0.51, + "learning_rate": 1.8317696816161793e-05, + "loss": 1.0787, + "step": 7940 + }, + { + "epoch": 0.51, + "learning_rate": 1.8317482127138842e-05, + "loss": 1.0783, + "step": 7941 + }, + { + "epoch": 0.51, + "learning_rate": 1.831726743811589e-05, + "loss": 1.0961, + "step": 7942 + }, + { + "epoch": 0.51, + "learning_rate": 1.831705274909294e-05, + "loss": 1.1437, + "step": 7943 + }, + { + "epoch": 0.51, + "learning_rate": 1.831683806006999e-05, + "loss": 1.0189, + "step": 7944 + }, + { + "epoch": 0.51, + "learning_rate": 1.831662337104704e-05, + "loss": 1.1168, + "step": 7945 + }, + { + "epoch": 0.51, + "learning_rate": 1.831640868202409e-05, + "loss": 1.0108, + "step": 7946 + }, + { + "epoch": 0.51, + "learning_rate": 1.831619399300114e-05, + "loss": 1.0711, + "step": 7947 + }, + { + "epoch": 0.51, + "learning_rate": 1.8315979303978188e-05, + "loss": 1.1435, + "step": 7948 + }, + { + "epoch": 0.51, + "learning_rate": 1.8315764614955237e-05, + "loss": 1.128, + "step": 7949 + }, + { + "epoch": 0.51, + "learning_rate": 1.831554992593229e-05, + "loss": 1.0264, + "step": 7950 + }, + { + "epoch": 0.51, + "learning_rate": 1.8315335236909338e-05, + "loss": 1.1576, + "step": 7951 + }, + { + "epoch": 0.51, + "learning_rate": 1.831512054788639e-05, + "loss": 1.04, + "step": 7952 + }, + { + "epoch": 0.51, + "learning_rate": 1.831490585886344e-05, + "loss": 1.1119, + "step": 7953 + }, + { + "epoch": 0.51, + "learning_rate": 1.831469116984049e-05, + "loss": 1.0283, + "step": 7954 + }, + { + "epoch": 0.51, + "learning_rate": 1.8314476480817538e-05, + "loss": 1.025, + "step": 7955 + }, + { + "epoch": 0.51, + "learning_rate": 1.8314261791794586e-05, + "loss": 1.1047, + "step": 7956 + }, + { + "epoch": 0.51, + "learning_rate": 1.8314047102771635e-05, + "loss": 1.1808, + "step": 7957 + }, + { + "epoch": 0.51, + "learning_rate": 1.8313832413748688e-05, + "loss": 1.1089, + "step": 7958 + }, + { + "epoch": 0.51, + "learning_rate": 1.8313617724725737e-05, + "loss": 1.1522, + "step": 7959 + }, + { + "epoch": 0.51, + "learning_rate": 1.8313403035702786e-05, + "loss": 1.1139, + "step": 7960 + }, + { + "epoch": 0.51, + "learning_rate": 1.8313188346679835e-05, + "loss": 1.0638, + "step": 7961 + }, + { + "epoch": 0.51, + "learning_rate": 1.8312973657656884e-05, + "loss": 1.0488, + "step": 7962 + }, + { + "epoch": 0.51, + "learning_rate": 1.8312758968633936e-05, + "loss": 1.1374, + "step": 7963 + }, + { + "epoch": 0.51, + "learning_rate": 1.8312544279610985e-05, + "loss": 1.0946, + "step": 7964 + }, + { + "epoch": 0.51, + "learning_rate": 1.8312329590588034e-05, + "loss": 1.1456, + "step": 7965 + }, + { + "epoch": 0.51, + "learning_rate": 1.8312114901565083e-05, + "loss": 1.1151, + "step": 7966 + }, + { + "epoch": 0.51, + "learning_rate": 1.8311900212542135e-05, + "loss": 1.0666, + "step": 7967 + }, + { + "epoch": 0.51, + "learning_rate": 1.8311685523519184e-05, + "loss": 1.054, + "step": 7968 + }, + { + "epoch": 0.51, + "learning_rate": 1.8311470834496233e-05, + "loss": 1.0886, + "step": 7969 + }, + { + "epoch": 0.51, + "learning_rate": 1.8311256145473286e-05, + "loss": 1.0135, + "step": 7970 + }, + { + "epoch": 0.51, + "learning_rate": 1.8311041456450335e-05, + "loss": 1.0963, + "step": 7971 + }, + { + "epoch": 0.51, + "learning_rate": 1.8310826767427383e-05, + "loss": 1.0593, + "step": 7972 + }, + { + "epoch": 0.51, + "learning_rate": 1.8310612078404432e-05, + "loss": 1.0582, + "step": 7973 + }, + { + "epoch": 0.51, + "learning_rate": 1.831039738938148e-05, + "loss": 1.0686, + "step": 7974 + }, + { + "epoch": 0.51, + "learning_rate": 1.831018270035853e-05, + "loss": 1.0663, + "step": 7975 + }, + { + "epoch": 0.51, + "learning_rate": 1.8309968011335583e-05, + "loss": 1.122, + "step": 7976 + }, + { + "epoch": 0.51, + "learning_rate": 1.8309753322312632e-05, + "loss": 1.1616, + "step": 7977 + }, + { + "epoch": 0.51, + "learning_rate": 1.830953863328968e-05, + "loss": 1.127, + "step": 7978 + }, + { + "epoch": 0.51, + "learning_rate": 1.830932394426673e-05, + "loss": 1.079, + "step": 7979 + }, + { + "epoch": 0.51, + "learning_rate": 1.830910925524378e-05, + "loss": 1.0555, + "step": 7980 + }, + { + "epoch": 0.51, + "learning_rate": 1.830889456622083e-05, + "loss": 1.1457, + "step": 7981 + }, + { + "epoch": 0.51, + "learning_rate": 1.830867987719788e-05, + "loss": 1.1355, + "step": 7982 + }, + { + "epoch": 0.51, + "learning_rate": 1.8308465188174932e-05, + "loss": 1.0397, + "step": 7983 + }, + { + "epoch": 0.51, + "learning_rate": 1.830825049915198e-05, + "loss": 1.1248, + "step": 7984 + }, + { + "epoch": 0.51, + "learning_rate": 1.830803581012903e-05, + "loss": 1.0625, + "step": 7985 + }, + { + "epoch": 0.51, + "learning_rate": 1.830782112110608e-05, + "loss": 1.0728, + "step": 7986 + }, + { + "epoch": 0.51, + "learning_rate": 1.8307606432083128e-05, + "loss": 1.1688, + "step": 7987 + }, + { + "epoch": 0.51, + "learning_rate": 1.830739174306018e-05, + "loss": 1.0585, + "step": 7988 + }, + { + "epoch": 0.51, + "learning_rate": 1.830717705403723e-05, + "loss": 1.0881, + "step": 7989 + }, + { + "epoch": 0.51, + "learning_rate": 1.830696236501428e-05, + "loss": 1.0283, + "step": 7990 + }, + { + "epoch": 0.51, + "learning_rate": 1.8306747675991327e-05, + "loss": 1.2156, + "step": 7991 + }, + { + "epoch": 0.51, + "learning_rate": 1.8306532986968376e-05, + "loss": 1.1062, + "step": 7992 + }, + { + "epoch": 0.51, + "learning_rate": 1.8306318297945425e-05, + "loss": 1.1977, + "step": 7993 + }, + { + "epoch": 0.51, + "learning_rate": 1.8306103608922478e-05, + "loss": 1.0513, + "step": 7994 + }, + { + "epoch": 0.51, + "learning_rate": 1.8305888919899527e-05, + "loss": 1.0399, + "step": 7995 + }, + { + "epoch": 0.51, + "learning_rate": 1.830567423087658e-05, + "loss": 1.077, + "step": 7996 + }, + { + "epoch": 0.51, + "learning_rate": 1.8305459541853628e-05, + "loss": 1.1325, + "step": 7997 + }, + { + "epoch": 0.51, + "learning_rate": 1.8305244852830677e-05, + "loss": 1.1634, + "step": 7998 + }, + { + "epoch": 0.51, + "learning_rate": 1.8305030163807726e-05, + "loss": 1.0376, + "step": 7999 + }, + { + "epoch": 0.51, + "learning_rate": 1.8304815474784775e-05, + "loss": 1.025, + "step": 8000 + }, + { + "epoch": 0.51, + "eval_laion/cider": 0.48436194187301174, + "eval_laion/meteor": 0.38180602976137945, + "eval_laion/pred_num_words": 36.21875, + "eval_laion/sacrebleu": 15.14206874920634, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.098731517791748, + "eval_runtime": 97.5301, + "eval_samples_per_second": 5.116, + "eval_steps_per_second": 0.328, + "step": 8000 + }, + { + "epoch": 0.51, + "learning_rate": 1.8304600785761827e-05, + "loss": 1.0667, + "step": 8001 + }, + { + "epoch": 0.51, + "learning_rate": 1.8304386096738876e-05, + "loss": 1.0353, + "step": 8002 + }, + { + "epoch": 0.51, + "learning_rate": 1.8304171407715925e-05, + "loss": 1.1437, + "step": 8003 + }, + { + "epoch": 0.51, + "learning_rate": 1.8303956718692974e-05, + "loss": 1.0171, + "step": 8004 + }, + { + "epoch": 0.52, + "learning_rate": 1.8303742029670023e-05, + "loss": 1.0612, + "step": 8005 + }, + { + "epoch": 0.52, + "learning_rate": 1.8303527340647075e-05, + "loss": 1.1065, + "step": 8006 + }, + { + "epoch": 0.52, + "learning_rate": 1.8303312651624124e-05, + "loss": 1.089, + "step": 8007 + }, + { + "epoch": 0.52, + "learning_rate": 1.8303097962601173e-05, + "loss": 1.0661, + "step": 8008 + }, + { + "epoch": 0.52, + "learning_rate": 1.8302883273578222e-05, + "loss": 1.0614, + "step": 8009 + }, + { + "epoch": 0.52, + "learning_rate": 1.8302668584555275e-05, + "loss": 1.0549, + "step": 8010 + }, + { + "epoch": 0.52, + "learning_rate": 1.8302453895532324e-05, + "loss": 1.0712, + "step": 8011 + }, + { + "epoch": 0.52, + "learning_rate": 1.8302239206509373e-05, + "loss": 1.0601, + "step": 8012 + }, + { + "epoch": 0.52, + "learning_rate": 1.830202451748642e-05, + "loss": 1.1753, + "step": 8013 + }, + { + "epoch": 0.52, + "learning_rate": 1.8301809828463474e-05, + "loss": 1.1065, + "step": 8014 + }, + { + "epoch": 0.52, + "learning_rate": 1.8301595139440523e-05, + "loss": 1.0881, + "step": 8015 + }, + { + "epoch": 0.52, + "learning_rate": 1.8301380450417572e-05, + "loss": 0.9727, + "step": 8016 + }, + { + "epoch": 0.52, + "learning_rate": 1.830116576139462e-05, + "loss": 1.1011, + "step": 8017 + }, + { + "epoch": 0.52, + "learning_rate": 1.830095107237167e-05, + "loss": 1.0768, + "step": 8018 + }, + { + "epoch": 0.52, + "learning_rate": 1.8300736383348722e-05, + "loss": 1.0877, + "step": 8019 + }, + { + "epoch": 0.52, + "learning_rate": 1.830052169432577e-05, + "loss": 1.0133, + "step": 8020 + }, + { + "epoch": 0.52, + "learning_rate": 1.830030700530282e-05, + "loss": 1.0701, + "step": 8021 + }, + { + "epoch": 0.52, + "learning_rate": 1.830009231627987e-05, + "loss": 1.0836, + "step": 8022 + }, + { + "epoch": 0.52, + "learning_rate": 1.8299877627256918e-05, + "loss": 1.0784, + "step": 8023 + }, + { + "epoch": 0.52, + "learning_rate": 1.8299662938233967e-05, + "loss": 1.0469, + "step": 8024 + }, + { + "epoch": 0.52, + "learning_rate": 1.829944824921102e-05, + "loss": 1.125, + "step": 8025 + }, + { + "epoch": 0.52, + "learning_rate": 1.829923356018807e-05, + "loss": 1.1091, + "step": 8026 + }, + { + "epoch": 0.52, + "learning_rate": 1.829901887116512e-05, + "loss": 0.9893, + "step": 8027 + }, + { + "epoch": 0.52, + "learning_rate": 1.829880418214217e-05, + "loss": 1.0506, + "step": 8028 + }, + { + "epoch": 0.52, + "learning_rate": 1.829858949311922e-05, + "loss": 1.0986, + "step": 8029 + }, + { + "epoch": 0.52, + "learning_rate": 1.8298374804096267e-05, + "loss": 0.9835, + "step": 8030 + }, + { + "epoch": 0.52, + "learning_rate": 1.8298160115073316e-05, + "loss": 1.0974, + "step": 8031 + }, + { + "epoch": 0.52, + "learning_rate": 1.829794542605037e-05, + "loss": 1.036, + "step": 8032 + }, + { + "epoch": 0.52, + "learning_rate": 1.8297730737027418e-05, + "loss": 1.063, + "step": 8033 + }, + { + "epoch": 0.52, + "learning_rate": 1.8297516048004467e-05, + "loss": 1.0596, + "step": 8034 + }, + { + "epoch": 0.52, + "learning_rate": 1.8297301358981516e-05, + "loss": 0.9899, + "step": 8035 + }, + { + "epoch": 0.52, + "learning_rate": 1.8297086669958565e-05, + "loss": 1.0395, + "step": 8036 + }, + { + "epoch": 0.52, + "learning_rate": 1.8296871980935617e-05, + "loss": 1.0997, + "step": 8037 + }, + { + "epoch": 0.52, + "learning_rate": 1.8296657291912666e-05, + "loss": 1.0678, + "step": 8038 + }, + { + "epoch": 0.52, + "learning_rate": 1.829644260288972e-05, + "loss": 1.1101, + "step": 8039 + }, + { + "epoch": 0.52, + "learning_rate": 1.8296227913866767e-05, + "loss": 1.099, + "step": 8040 + }, + { + "epoch": 0.52, + "learning_rate": 1.8296227913866767e-05, + "loss": 1.0898, + "step": 8041 + }, + { + "epoch": 0.52, + "learning_rate": 1.8296013224843816e-05, + "loss": 1.1505, + "step": 8042 + }, + { + "epoch": 0.52, + "learning_rate": 1.8295798535820865e-05, + "loss": 1.0668, + "step": 8043 + }, + { + "epoch": 0.52, + "learning_rate": 1.8295583846797914e-05, + "loss": 0.9702, + "step": 8044 + }, + { + "epoch": 0.52, + "learning_rate": 1.8295369157774963e-05, + "loss": 1.0245, + "step": 8045 + }, + { + "epoch": 0.52, + "learning_rate": 1.8295154468752016e-05, + "loss": 1.0176, + "step": 8046 + }, + { + "epoch": 0.52, + "learning_rate": 1.8294939779729064e-05, + "loss": 1.1694, + "step": 8047 + }, + { + "epoch": 0.52, + "learning_rate": 1.8294725090706113e-05, + "loss": 1.105, + "step": 8048 + }, + { + "epoch": 0.52, + "learning_rate": 1.8294510401683162e-05, + "loss": 0.9998, + "step": 8049 + }, + { + "epoch": 0.52, + "learning_rate": 1.829429571266021e-05, + "loss": 1.0675, + "step": 8050 + }, + { + "epoch": 0.52, + "learning_rate": 1.8294081023637264e-05, + "loss": 1.1379, + "step": 8051 + }, + { + "epoch": 0.52, + "learning_rate": 1.8293866334614313e-05, + "loss": 1.1549, + "step": 8052 + }, + { + "epoch": 0.52, + "learning_rate": 1.829365164559136e-05, + "loss": 1.034, + "step": 8053 + }, + { + "epoch": 0.52, + "learning_rate": 1.829343695656841e-05, + "loss": 1.036, + "step": 8054 + }, + { + "epoch": 0.52, + "learning_rate": 1.8293222267545463e-05, + "loss": 1.1473, + "step": 8055 + }, + { + "epoch": 0.52, + "learning_rate": 1.8293007578522512e-05, + "loss": 1.117, + "step": 8056 + }, + { + "epoch": 0.52, + "learning_rate": 1.829279288949956e-05, + "loss": 1.0316, + "step": 8057 + }, + { + "epoch": 0.52, + "learning_rate": 1.8292578200476613e-05, + "loss": 1.1121, + "step": 8058 + }, + { + "epoch": 0.52, + "learning_rate": 1.8292363511453662e-05, + "loss": 1.1404, + "step": 8059 + }, + { + "epoch": 0.52, + "learning_rate": 1.829214882243071e-05, + "loss": 1.0579, + "step": 8060 + }, + { + "epoch": 0.52, + "learning_rate": 1.829193413340776e-05, + "loss": 1.172, + "step": 8061 + }, + { + "epoch": 0.52, + "learning_rate": 1.829171944438481e-05, + "loss": 1.1006, + "step": 8062 + }, + { + "epoch": 0.52, + "learning_rate": 1.8291504755361858e-05, + "loss": 1.1101, + "step": 8063 + }, + { + "epoch": 0.52, + "learning_rate": 1.829129006633891e-05, + "loss": 1.0446, + "step": 8064 + }, + { + "epoch": 0.52, + "learning_rate": 1.829107537731596e-05, + "loss": 1.0743, + "step": 8065 + }, + { + "epoch": 0.52, + "learning_rate": 1.829086068829301e-05, + "loss": 1.0811, + "step": 8066 + }, + { + "epoch": 0.52, + "learning_rate": 1.8290645999270057e-05, + "loss": 1.0718, + "step": 8067 + }, + { + "epoch": 0.52, + "learning_rate": 1.8290431310247106e-05, + "loss": 1.1018, + "step": 8068 + }, + { + "epoch": 0.52, + "learning_rate": 1.829021662122416e-05, + "loss": 1.0546, + "step": 8069 + }, + { + "epoch": 0.52, + "learning_rate": 1.8290001932201208e-05, + "loss": 1.0163, + "step": 8070 + }, + { + "epoch": 0.52, + "learning_rate": 1.828978724317826e-05, + "loss": 1.1906, + "step": 8071 + }, + { + "epoch": 0.52, + "learning_rate": 1.828957255415531e-05, + "loss": 1.1072, + "step": 8072 + }, + { + "epoch": 0.52, + "learning_rate": 1.8289357865132358e-05, + "loss": 1.0785, + "step": 8073 + }, + { + "epoch": 0.52, + "learning_rate": 1.8289143176109407e-05, + "loss": 1.1534, + "step": 8074 + }, + { + "epoch": 0.52, + "learning_rate": 1.8288928487086456e-05, + "loss": 1.0229, + "step": 8075 + }, + { + "epoch": 0.52, + "learning_rate": 1.8288713798063508e-05, + "loss": 1.0494, + "step": 8076 + }, + { + "epoch": 0.52, + "learning_rate": 1.8288499109040557e-05, + "loss": 1.0794, + "step": 8077 + }, + { + "epoch": 0.52, + "learning_rate": 1.8288284420017606e-05, + "loss": 1.0659, + "step": 8078 + }, + { + "epoch": 0.52, + "learning_rate": 1.8288069730994655e-05, + "loss": 1.157, + "step": 8079 + }, + { + "epoch": 0.52, + "learning_rate": 1.8287855041971704e-05, + "loss": 1.0973, + "step": 8080 + }, + { + "epoch": 0.52, + "learning_rate": 1.8287640352948753e-05, + "loss": 1.0919, + "step": 8081 + }, + { + "epoch": 0.52, + "learning_rate": 1.8287425663925805e-05, + "loss": 1.1291, + "step": 8082 + }, + { + "epoch": 0.52, + "learning_rate": 1.8287210974902854e-05, + "loss": 1.106, + "step": 8083 + }, + { + "epoch": 0.52, + "learning_rate": 1.8286996285879907e-05, + "loss": 1.0507, + "step": 8084 + }, + { + "epoch": 0.52, + "learning_rate": 1.8286781596856956e-05, + "loss": 0.9903, + "step": 8085 + }, + { + "epoch": 0.52, + "learning_rate": 1.8286566907834005e-05, + "loss": 1.1128, + "step": 8086 + }, + { + "epoch": 0.52, + "learning_rate": 1.8286352218811054e-05, + "loss": 1.0885, + "step": 8087 + }, + { + "epoch": 0.52, + "learning_rate": 1.8286137529788103e-05, + "loss": 1.107, + "step": 8088 + }, + { + "epoch": 0.52, + "learning_rate": 1.8285922840765155e-05, + "loss": 1.0569, + "step": 8089 + }, + { + "epoch": 0.52, + "learning_rate": 1.8285708151742204e-05, + "loss": 1.0698, + "step": 8090 + }, + { + "epoch": 0.52, + "learning_rate": 1.8285493462719253e-05, + "loss": 1.0638, + "step": 8091 + }, + { + "epoch": 0.52, + "learning_rate": 1.8285278773696302e-05, + "loss": 1.0836, + "step": 8092 + }, + { + "epoch": 0.52, + "learning_rate": 1.828506408467335e-05, + "loss": 1.0922, + "step": 8093 + }, + { + "epoch": 0.52, + "learning_rate": 1.82848493956504e-05, + "loss": 1.0614, + "step": 8094 + }, + { + "epoch": 0.52, + "learning_rate": 1.8284634706627452e-05, + "loss": 1.0367, + "step": 8095 + }, + { + "epoch": 0.52, + "learning_rate": 1.82844200176045e-05, + "loss": 1.1503, + "step": 8096 + }, + { + "epoch": 0.52, + "learning_rate": 1.828420532858155e-05, + "loss": 1.0949, + "step": 8097 + }, + { + "epoch": 0.52, + "learning_rate": 1.82839906395586e-05, + "loss": 1.0578, + "step": 8098 + }, + { + "epoch": 0.52, + "learning_rate": 1.828377595053565e-05, + "loss": 1.1246, + "step": 8099 + }, + { + "epoch": 0.52, + "learning_rate": 1.82835612615127e-05, + "loss": 1.0203, + "step": 8100 + }, + { + "epoch": 0.52, + "learning_rate": 1.828334657248975e-05, + "loss": 1.0423, + "step": 8101 + }, + { + "epoch": 0.52, + "learning_rate": 1.82831318834668e-05, + "loss": 1.0409, + "step": 8102 + }, + { + "epoch": 0.52, + "learning_rate": 1.828291719444385e-05, + "loss": 1.1407, + "step": 8103 + }, + { + "epoch": 0.52, + "learning_rate": 1.82827025054209e-05, + "loss": 1.1429, + "step": 8104 + }, + { + "epoch": 0.52, + "learning_rate": 1.828248781639795e-05, + "loss": 1.0366, + "step": 8105 + }, + { + "epoch": 0.52, + "learning_rate": 1.8282273127374997e-05, + "loss": 1.0255, + "step": 8106 + }, + { + "epoch": 0.52, + "learning_rate": 1.828205843835205e-05, + "loss": 1.0886, + "step": 8107 + }, + { + "epoch": 0.52, + "learning_rate": 1.82818437493291e-05, + "loss": 1.0264, + "step": 8108 + }, + { + "epoch": 0.52, + "learning_rate": 1.8281629060306148e-05, + "loss": 1.0731, + "step": 8109 + }, + { + "epoch": 0.52, + "learning_rate": 1.8281414371283197e-05, + "loss": 1.0868, + "step": 8110 + }, + { + "epoch": 0.52, + "learning_rate": 1.8281199682260246e-05, + "loss": 1.2463, + "step": 8111 + }, + { + "epoch": 0.52, + "learning_rate": 1.8280984993237295e-05, + "loss": 1.153, + "step": 8112 + }, + { + "epoch": 0.52, + "learning_rate": 1.8280770304214347e-05, + "loss": 1.0927, + "step": 8113 + }, + { + "epoch": 0.52, + "learning_rate": 1.8280555615191396e-05, + "loss": 1.0837, + "step": 8114 + }, + { + "epoch": 0.52, + "learning_rate": 1.828034092616845e-05, + "loss": 1.0975, + "step": 8115 + }, + { + "epoch": 0.52, + "learning_rate": 1.8280126237145497e-05, + "loss": 1.1028, + "step": 8116 + }, + { + "epoch": 0.52, + "learning_rate": 1.8279911548122546e-05, + "loss": 1.0904, + "step": 8117 + }, + { + "epoch": 0.52, + "learning_rate": 1.8279696859099595e-05, + "loss": 1.131, + "step": 8118 + }, + { + "epoch": 0.52, + "learning_rate": 1.8279482170076644e-05, + "loss": 1.0674, + "step": 8119 + }, + { + "epoch": 0.52, + "learning_rate": 1.8279267481053697e-05, + "loss": 1.0507, + "step": 8120 + }, + { + "epoch": 0.52, + "learning_rate": 1.8279052792030745e-05, + "loss": 1.0655, + "step": 8121 + }, + { + "epoch": 0.52, + "learning_rate": 1.8278838103007794e-05, + "loss": 1.1079, + "step": 8122 + }, + { + "epoch": 0.52, + "learning_rate": 1.8278623413984843e-05, + "loss": 1.1073, + "step": 8123 + }, + { + "epoch": 0.52, + "learning_rate": 1.8278408724961892e-05, + "loss": 1.1309, + "step": 8124 + }, + { + "epoch": 0.52, + "learning_rate": 1.8278194035938945e-05, + "loss": 1.0941, + "step": 8125 + }, + { + "epoch": 0.52, + "learning_rate": 1.8277979346915994e-05, + "loss": 1.0954, + "step": 8126 + }, + { + "epoch": 0.52, + "learning_rate": 1.8277764657893043e-05, + "loss": 1.0224, + "step": 8127 + }, + { + "epoch": 0.52, + "learning_rate": 1.8277549968870095e-05, + "loss": 1.0162, + "step": 8128 + }, + { + "epoch": 0.52, + "learning_rate": 1.8277335279847144e-05, + "loss": 1.1002, + "step": 8129 + }, + { + "epoch": 0.52, + "learning_rate": 1.8277120590824193e-05, + "loss": 1.1034, + "step": 8130 + }, + { + "epoch": 0.52, + "learning_rate": 1.8276905901801242e-05, + "loss": 1.0318, + "step": 8131 + }, + { + "epoch": 0.52, + "learning_rate": 1.827669121277829e-05, + "loss": 1.078, + "step": 8132 + }, + { + "epoch": 0.52, + "learning_rate": 1.8276476523755343e-05, + "loss": 1.0838, + "step": 8133 + }, + { + "epoch": 0.52, + "learning_rate": 1.8276261834732392e-05, + "loss": 1.1434, + "step": 8134 + }, + { + "epoch": 0.52, + "learning_rate": 1.827604714570944e-05, + "loss": 1.0807, + "step": 8135 + }, + { + "epoch": 0.52, + "learning_rate": 1.827583245668649e-05, + "loss": 1.0733, + "step": 8136 + }, + { + "epoch": 0.52, + "learning_rate": 1.827561776766354e-05, + "loss": 1.08, + "step": 8137 + }, + { + "epoch": 0.52, + "learning_rate": 1.827540307864059e-05, + "loss": 1.0479, + "step": 8138 + }, + { + "epoch": 0.52, + "learning_rate": 1.827518838961764e-05, + "loss": 1.1431, + "step": 8139 + }, + { + "epoch": 0.52, + "learning_rate": 1.827497370059469e-05, + "loss": 1.0651, + "step": 8140 + }, + { + "epoch": 0.52, + "learning_rate": 1.827475901157174e-05, + "loss": 1.1234, + "step": 8141 + }, + { + "epoch": 0.52, + "learning_rate": 1.827454432254879e-05, + "loss": 1.0927, + "step": 8142 + }, + { + "epoch": 0.52, + "learning_rate": 1.827432963352584e-05, + "loss": 1.1291, + "step": 8143 + }, + { + "epoch": 0.52, + "learning_rate": 1.827411494450289e-05, + "loss": 1.0331, + "step": 8144 + }, + { + "epoch": 0.52, + "learning_rate": 1.827390025547994e-05, + "loss": 1.0441, + "step": 8145 + }, + { + "epoch": 0.52, + "learning_rate": 1.827368556645699e-05, + "loss": 0.9945, + "step": 8146 + }, + { + "epoch": 0.52, + "learning_rate": 1.827347087743404e-05, + "loss": 1.0828, + "step": 8147 + }, + { + "epoch": 0.52, + "learning_rate": 1.8273256188411088e-05, + "loss": 1.1223, + "step": 8148 + }, + { + "epoch": 0.52, + "learning_rate": 1.8273041499388137e-05, + "loss": 1.1323, + "step": 8149 + }, + { + "epoch": 0.52, + "learning_rate": 1.8272826810365186e-05, + "loss": 1.0664, + "step": 8150 + }, + { + "epoch": 0.52, + "learning_rate": 1.8272612121342238e-05, + "loss": 1.0888, + "step": 8151 + }, + { + "epoch": 0.52, + "learning_rate": 1.8272397432319287e-05, + "loss": 1.0927, + "step": 8152 + }, + { + "epoch": 0.52, + "learning_rate": 1.8272182743296336e-05, + "loss": 1.1309, + "step": 8153 + }, + { + "epoch": 0.52, + "learning_rate": 1.8271968054273385e-05, + "loss": 0.9996, + "step": 8154 + }, + { + "epoch": 0.52, + "learning_rate": 1.8271753365250434e-05, + "loss": 1.0739, + "step": 8155 + }, + { + "epoch": 0.52, + "learning_rate": 1.8271538676227486e-05, + "loss": 1.1302, + "step": 8156 + }, + { + "epoch": 0.52, + "learning_rate": 1.8271323987204535e-05, + "loss": 1.0923, + "step": 8157 + }, + { + "epoch": 0.52, + "learning_rate": 1.8271109298181588e-05, + "loss": 1.0742, + "step": 8158 + }, + { + "epoch": 0.52, + "learning_rate": 1.8270894609158637e-05, + "loss": 1.0981, + "step": 8159 + }, + { + "epoch": 0.52, + "learning_rate": 1.8270679920135686e-05, + "loss": 1.1537, + "step": 8160 + }, + { + "epoch": 0.53, + "learning_rate": 1.8270465231112735e-05, + "loss": 1.0676, + "step": 8161 + }, + { + "epoch": 0.53, + "learning_rate": 1.8270250542089784e-05, + "loss": 1.1043, + "step": 8162 + }, + { + "epoch": 0.53, + "learning_rate": 1.8270035853066836e-05, + "loss": 1.0435, + "step": 8163 + }, + { + "epoch": 0.53, + "learning_rate": 1.8269821164043885e-05, + "loss": 1.1427, + "step": 8164 + }, + { + "epoch": 0.53, + "learning_rate": 1.8269606475020934e-05, + "loss": 1.1137, + "step": 8165 + }, + { + "epoch": 0.53, + "learning_rate": 1.8269391785997983e-05, + "loss": 1.1583, + "step": 8166 + }, + { + "epoch": 0.53, + "learning_rate": 1.8269177096975032e-05, + "loss": 1.1812, + "step": 8167 + }, + { + "epoch": 0.53, + "learning_rate": 1.826896240795208e-05, + "loss": 1.0823, + "step": 8168 + }, + { + "epoch": 0.53, + "learning_rate": 1.8268747718929133e-05, + "loss": 1.1305, + "step": 8169 + }, + { + "epoch": 0.53, + "learning_rate": 1.8268533029906182e-05, + "loss": 1.096, + "step": 8170 + }, + { + "epoch": 0.53, + "learning_rate": 1.8268318340883234e-05, + "loss": 1.0772, + "step": 8171 + }, + { + "epoch": 0.53, + "learning_rate": 1.8268103651860283e-05, + "loss": 1.1406, + "step": 8172 + }, + { + "epoch": 0.53, + "learning_rate": 1.8267888962837332e-05, + "loss": 1.0235, + "step": 8173 + }, + { + "epoch": 0.53, + "learning_rate": 1.826767427381438e-05, + "loss": 1.1012, + "step": 8174 + }, + { + "epoch": 0.53, + "learning_rate": 1.826745958479143e-05, + "loss": 1.0874, + "step": 8175 + }, + { + "epoch": 0.53, + "learning_rate": 1.8267244895768483e-05, + "loss": 1.0771, + "step": 8176 + }, + { + "epoch": 0.53, + "learning_rate": 1.826703020674553e-05, + "loss": 1.1187, + "step": 8177 + }, + { + "epoch": 0.53, + "learning_rate": 1.826681551772258e-05, + "loss": 1.1181, + "step": 8178 + }, + { + "epoch": 0.53, + "learning_rate": 1.826660082869963e-05, + "loss": 1.1321, + "step": 8179 + }, + { + "epoch": 0.53, + "learning_rate": 1.826638613967668e-05, + "loss": 1.1168, + "step": 8180 + }, + { + "epoch": 0.53, + "learning_rate": 1.8266171450653727e-05, + "loss": 1.0463, + "step": 8181 + }, + { + "epoch": 0.53, + "learning_rate": 1.826595676163078e-05, + "loss": 1.0416, + "step": 8182 + }, + { + "epoch": 0.53, + "learning_rate": 1.826574207260783e-05, + "loss": 1.102, + "step": 8183 + }, + { + "epoch": 0.53, + "learning_rate": 1.8265527383584878e-05, + "loss": 1.0736, + "step": 8184 + }, + { + "epoch": 0.53, + "learning_rate": 1.8265312694561927e-05, + "loss": 1.0493, + "step": 8185 + }, + { + "epoch": 0.53, + "learning_rate": 1.826509800553898e-05, + "loss": 1.064, + "step": 8186 + }, + { + "epoch": 0.53, + "learning_rate": 1.8264883316516028e-05, + "loss": 1.1294, + "step": 8187 + }, + { + "epoch": 0.53, + "learning_rate": 1.8264668627493077e-05, + "loss": 1.1239, + "step": 8188 + }, + { + "epoch": 0.53, + "learning_rate": 1.826445393847013e-05, + "loss": 0.9723, + "step": 8189 + }, + { + "epoch": 0.53, + "learning_rate": 1.8264239249447178e-05, + "loss": 1.0753, + "step": 8190 + }, + { + "epoch": 0.53, + "learning_rate": 1.8264024560424227e-05, + "loss": 1.1173, + "step": 8191 + }, + { + "epoch": 0.53, + "learning_rate": 1.8263809871401276e-05, + "loss": 1.1891, + "step": 8192 + }, + { + "epoch": 0.53, + "learning_rate": 1.8263595182378325e-05, + "loss": 1.0281, + "step": 8193 + }, + { + "epoch": 0.53, + "learning_rate": 1.8263380493355378e-05, + "loss": 1.065, + "step": 8194 + }, + { + "epoch": 0.53, + "learning_rate": 1.8263165804332426e-05, + "loss": 1.0444, + "step": 8195 + }, + { + "epoch": 0.53, + "learning_rate": 1.8262951115309475e-05, + "loss": 1.0223, + "step": 8196 + }, + { + "epoch": 0.53, + "learning_rate": 1.8262736426286524e-05, + "loss": 1.1386, + "step": 8197 + }, + { + "epoch": 0.53, + "learning_rate": 1.8262521737263573e-05, + "loss": 1.0742, + "step": 8198 + }, + { + "epoch": 0.53, + "learning_rate": 1.8262307048240622e-05, + "loss": 1.1126, + "step": 8199 + }, + { + "epoch": 0.53, + "learning_rate": 1.8262092359217675e-05, + "loss": 1.1073, + "step": 8200 + }, + { + "epoch": 0.53, + "learning_rate": 1.8261877670194724e-05, + "loss": 1.1378, + "step": 8201 + }, + { + "epoch": 0.53, + "learning_rate": 1.8261662981171776e-05, + "loss": 1.0283, + "step": 8202 + }, + { + "epoch": 0.53, + "learning_rate": 1.8261448292148825e-05, + "loss": 1.1157, + "step": 8203 + }, + { + "epoch": 0.53, + "learning_rate": 1.8261233603125874e-05, + "loss": 1.0393, + "step": 8204 + }, + { + "epoch": 0.53, + "learning_rate": 1.8261018914102923e-05, + "loss": 1.1031, + "step": 8205 + }, + { + "epoch": 0.53, + "learning_rate": 1.8260804225079972e-05, + "loss": 1.0948, + "step": 8206 + }, + { + "epoch": 0.53, + "learning_rate": 1.8260589536057024e-05, + "loss": 1.0029, + "step": 8207 + }, + { + "epoch": 0.53, + "learning_rate": 1.8260374847034073e-05, + "loss": 1.0943, + "step": 8208 + }, + { + "epoch": 0.53, + "learning_rate": 1.8260160158011122e-05, + "loss": 1.1599, + "step": 8209 + }, + { + "epoch": 0.53, + "learning_rate": 1.825994546898817e-05, + "loss": 1.1075, + "step": 8210 + }, + { + "epoch": 0.53, + "learning_rate": 1.825973077996522e-05, + "loss": 1.0768, + "step": 8211 + }, + { + "epoch": 0.53, + "learning_rate": 1.8259516090942272e-05, + "loss": 1.1317, + "step": 8212 + }, + { + "epoch": 0.53, + "learning_rate": 1.825930140191932e-05, + "loss": 1.0844, + "step": 8213 + }, + { + "epoch": 0.53, + "learning_rate": 1.825908671289637e-05, + "loss": 1.087, + "step": 8214 + }, + { + "epoch": 0.53, + "learning_rate": 1.8258872023873423e-05, + "loss": 1.1009, + "step": 8215 + }, + { + "epoch": 0.53, + "learning_rate": 1.8258657334850472e-05, + "loss": 1.0307, + "step": 8216 + }, + { + "epoch": 0.53, + "learning_rate": 1.825844264582752e-05, + "loss": 1.1551, + "step": 8217 + }, + { + "epoch": 0.53, + "learning_rate": 1.825822795680457e-05, + "loss": 1.0747, + "step": 8218 + }, + { + "epoch": 0.53, + "learning_rate": 1.825801326778162e-05, + "loss": 1.0571, + "step": 8219 + }, + { + "epoch": 0.53, + "learning_rate": 1.825779857875867e-05, + "loss": 1.0274, + "step": 8220 + }, + { + "epoch": 0.53, + "learning_rate": 1.825758388973572e-05, + "loss": 1.083, + "step": 8221 + }, + { + "epoch": 0.53, + "learning_rate": 1.825736920071277e-05, + "loss": 1.1053, + "step": 8222 + }, + { + "epoch": 0.53, + "learning_rate": 1.8257154511689818e-05, + "loss": 1.0685, + "step": 8223 + }, + { + "epoch": 0.53, + "learning_rate": 1.8256939822666867e-05, + "loss": 1.0771, + "step": 8224 + }, + { + "epoch": 0.53, + "learning_rate": 1.825672513364392e-05, + "loss": 1.0701, + "step": 8225 + }, + { + "epoch": 0.53, + "learning_rate": 1.8256510444620968e-05, + "loss": 1.0124, + "step": 8226 + }, + { + "epoch": 0.53, + "learning_rate": 1.8256295755598017e-05, + "loss": 1.0765, + "step": 8227 + }, + { + "epoch": 0.53, + "learning_rate": 1.8256081066575066e-05, + "loss": 1.0771, + "step": 8228 + }, + { + "epoch": 0.53, + "learning_rate": 1.825586637755212e-05, + "loss": 1.0389, + "step": 8229 + }, + { + "epoch": 0.53, + "learning_rate": 1.8255651688529167e-05, + "loss": 1.0942, + "step": 8230 + }, + { + "epoch": 0.53, + "learning_rate": 1.8255436999506216e-05, + "loss": 1.0835, + "step": 8231 + }, + { + "epoch": 0.53, + "learning_rate": 1.825522231048327e-05, + "loss": 1.1289, + "step": 8232 + }, + { + "epoch": 0.53, + "learning_rate": 1.8255007621460318e-05, + "loss": 1.0657, + "step": 8233 + }, + { + "epoch": 0.53, + "learning_rate": 1.8254792932437367e-05, + "loss": 1.1168, + "step": 8234 + }, + { + "epoch": 0.53, + "learning_rate": 1.8254578243414416e-05, + "loss": 1.1326, + "step": 8235 + }, + { + "epoch": 0.53, + "learning_rate": 1.8254363554391465e-05, + "loss": 1.0333, + "step": 8236 + }, + { + "epoch": 0.53, + "learning_rate": 1.8254148865368514e-05, + "loss": 1.0363, + "step": 8237 + }, + { + "epoch": 0.53, + "learning_rate": 1.8253934176345566e-05, + "loss": 1.0607, + "step": 8238 + }, + { + "epoch": 0.53, + "learning_rate": 1.8253719487322615e-05, + "loss": 1.1055, + "step": 8239 + }, + { + "epoch": 0.53, + "learning_rate": 1.8253504798299664e-05, + "loss": 1.0474, + "step": 8240 + }, + { + "epoch": 0.53, + "learning_rate": 1.8253290109276713e-05, + "loss": 0.974, + "step": 8241 + }, + { + "epoch": 0.53, + "learning_rate": 1.8253075420253762e-05, + "loss": 1.1319, + "step": 8242 + }, + { + "epoch": 0.53, + "learning_rate": 1.8252860731230814e-05, + "loss": 1.1181, + "step": 8243 + }, + { + "epoch": 0.53, + "learning_rate": 1.8252646042207863e-05, + "loss": 1.0954, + "step": 8244 + }, + { + "epoch": 0.53, + "learning_rate": 1.8252431353184915e-05, + "loss": 1.0821, + "step": 8245 + }, + { + "epoch": 0.53, + "learning_rate": 1.8252216664161964e-05, + "loss": 1.0731, + "step": 8246 + }, + { + "epoch": 0.53, + "learning_rate": 1.8252001975139013e-05, + "loss": 1.0041, + "step": 8247 + }, + { + "epoch": 0.53, + "learning_rate": 1.8251787286116062e-05, + "loss": 1.0669, + "step": 8248 + }, + { + "epoch": 0.53, + "learning_rate": 1.825157259709311e-05, + "loss": 1.0428, + "step": 8249 + }, + { + "epoch": 0.53, + "learning_rate": 1.8251357908070164e-05, + "loss": 1.1349, + "step": 8250 + }, + { + "epoch": 0.53, + "learning_rate": 1.8251143219047213e-05, + "loss": 1.09, + "step": 8251 + }, + { + "epoch": 0.53, + "learning_rate": 1.825092853002426e-05, + "loss": 1.1057, + "step": 8252 + }, + { + "epoch": 0.53, + "learning_rate": 1.825071384100131e-05, + "loss": 1.1332, + "step": 8253 + }, + { + "epoch": 0.53, + "learning_rate": 1.825049915197836e-05, + "loss": 1.1586, + "step": 8254 + }, + { + "epoch": 0.53, + "learning_rate": 1.825028446295541e-05, + "loss": 1.0668, + "step": 8255 + }, + { + "epoch": 0.53, + "learning_rate": 1.825006977393246e-05, + "loss": 1.0991, + "step": 8256 + }, + { + "epoch": 0.53, + "learning_rate": 1.824985508490951e-05, + "loss": 1.1077, + "step": 8257 + }, + { + "epoch": 0.53, + "learning_rate": 1.8249640395886562e-05, + "loss": 1.0538, + "step": 8258 + }, + { + "epoch": 0.53, + "learning_rate": 1.824942570686361e-05, + "loss": 1.0958, + "step": 8259 + }, + { + "epoch": 0.53, + "learning_rate": 1.824921101784066e-05, + "loss": 1.162, + "step": 8260 + }, + { + "epoch": 0.53, + "learning_rate": 1.824899632881771e-05, + "loss": 0.9873, + "step": 8261 + }, + { + "epoch": 0.53, + "learning_rate": 1.8248781639794758e-05, + "loss": 1.0539, + "step": 8262 + }, + { + "epoch": 0.53, + "learning_rate": 1.824856695077181e-05, + "loss": 1.0695, + "step": 8263 + }, + { + "epoch": 0.53, + "learning_rate": 1.824835226174886e-05, + "loss": 1.0419, + "step": 8264 + }, + { + "epoch": 0.53, + "learning_rate": 1.8248137572725908e-05, + "loss": 1.0682, + "step": 8265 + }, + { + "epoch": 0.53, + "learning_rate": 1.8247922883702957e-05, + "loss": 1.1059, + "step": 8266 + }, + { + "epoch": 0.53, + "learning_rate": 1.8247708194680006e-05, + "loss": 1.0929, + "step": 8267 + }, + { + "epoch": 0.53, + "learning_rate": 1.8247493505657055e-05, + "loss": 1.0792, + "step": 8268 + }, + { + "epoch": 0.53, + "learning_rate": 1.8247278816634107e-05, + "loss": 1.0586, + "step": 8269 + }, + { + "epoch": 0.53, + "learning_rate": 1.8247064127611156e-05, + "loss": 1.1317, + "step": 8270 + }, + { + "epoch": 0.53, + "learning_rate": 1.8246849438588205e-05, + "loss": 1.1444, + "step": 8271 + }, + { + "epoch": 0.53, + "learning_rate": 1.8246634749565254e-05, + "loss": 1.0813, + "step": 8272 + }, + { + "epoch": 0.53, + "learning_rate": 1.8246420060542307e-05, + "loss": 1.1118, + "step": 8273 + }, + { + "epoch": 0.53, + "learning_rate": 1.8246205371519356e-05, + "loss": 1.0622, + "step": 8274 + }, + { + "epoch": 0.53, + "learning_rate": 1.8245990682496405e-05, + "loss": 1.0044, + "step": 8275 + }, + { + "epoch": 0.53, + "learning_rate": 1.8245775993473457e-05, + "loss": 1.0742, + "step": 8276 + }, + { + "epoch": 0.53, + "learning_rate": 1.8245561304450506e-05, + "loss": 1.121, + "step": 8277 + }, + { + "epoch": 0.53, + "learning_rate": 1.8245346615427555e-05, + "loss": 1.059, + "step": 8278 + }, + { + "epoch": 0.53, + "learning_rate": 1.8245131926404604e-05, + "loss": 0.9765, + "step": 8279 + }, + { + "epoch": 0.53, + "learning_rate": 1.8244917237381653e-05, + "loss": 0.9981, + "step": 8280 + }, + { + "epoch": 0.53, + "learning_rate": 1.8244702548358705e-05, + "loss": 1.1183, + "step": 8281 + }, + { + "epoch": 0.53, + "learning_rate": 1.8244487859335754e-05, + "loss": 1.0465, + "step": 8282 + }, + { + "epoch": 0.53, + "learning_rate": 1.8244273170312803e-05, + "loss": 1.0752, + "step": 8283 + }, + { + "epoch": 0.53, + "learning_rate": 1.8244058481289852e-05, + "loss": 1.098, + "step": 8284 + }, + { + "epoch": 0.53, + "learning_rate": 1.82438437922669e-05, + "loss": 1.0475, + "step": 8285 + }, + { + "epoch": 0.53, + "learning_rate": 1.824362910324395e-05, + "loss": 1.1452, + "step": 8286 + }, + { + "epoch": 0.53, + "learning_rate": 1.8243414414221002e-05, + "loss": 1.0753, + "step": 8287 + }, + { + "epoch": 0.53, + "learning_rate": 1.824319972519805e-05, + "loss": 1.0723, + "step": 8288 + }, + { + "epoch": 0.53, + "learning_rate": 1.8242985036175104e-05, + "loss": 1.0369, + "step": 8289 + }, + { + "epoch": 0.53, + "learning_rate": 1.8242770347152153e-05, + "loss": 1.0428, + "step": 8290 + }, + { + "epoch": 0.53, + "learning_rate": 1.82425556581292e-05, + "loss": 1.0669, + "step": 8291 + }, + { + "epoch": 0.53, + "learning_rate": 1.824234096910625e-05, + "loss": 0.9928, + "step": 8292 + }, + { + "epoch": 0.53, + "learning_rate": 1.82421262800833e-05, + "loss": 1.0336, + "step": 8293 + }, + { + "epoch": 0.53, + "learning_rate": 1.8241911591060352e-05, + "loss": 1.104, + "step": 8294 + }, + { + "epoch": 0.53, + "learning_rate": 1.82416969020374e-05, + "loss": 1.0237, + "step": 8295 + }, + { + "epoch": 0.53, + "learning_rate": 1.824148221301445e-05, + "loss": 1.0991, + "step": 8296 + }, + { + "epoch": 0.53, + "learning_rate": 1.82412675239915e-05, + "loss": 1.0446, + "step": 8297 + }, + { + "epoch": 0.53, + "learning_rate": 1.8241052834968548e-05, + "loss": 1.0893, + "step": 8298 + }, + { + "epoch": 0.53, + "learning_rate": 1.82408381459456e-05, + "loss": 1.0642, + "step": 8299 + }, + { + "epoch": 0.53, + "learning_rate": 1.824062345692265e-05, + "loss": 1.0953, + "step": 8300 + }, + { + "epoch": 0.53, + "learning_rate": 1.8240408767899698e-05, + "loss": 1.0872, + "step": 8301 + }, + { + "epoch": 0.53, + "learning_rate": 1.824019407887675e-05, + "loss": 0.9992, + "step": 8302 + }, + { + "epoch": 0.53, + "learning_rate": 1.82399793898538e-05, + "loss": 1.0547, + "step": 8303 + }, + { + "epoch": 0.53, + "learning_rate": 1.823976470083085e-05, + "loss": 1.0282, + "step": 8304 + }, + { + "epoch": 0.53, + "learning_rate": 1.8239550011807897e-05, + "loss": 1.1349, + "step": 8305 + }, + { + "epoch": 0.53, + "learning_rate": 1.8239335322784946e-05, + "loss": 1.05, + "step": 8306 + }, + { + "epoch": 0.53, + "learning_rate": 1.8239120633762e-05, + "loss": 1.1113, + "step": 8307 + }, + { + "epoch": 0.53, + "learning_rate": 1.8238905944739048e-05, + "loss": 1.058, + "step": 8308 + }, + { + "epoch": 0.53, + "learning_rate": 1.8238691255716097e-05, + "loss": 1.0113, + "step": 8309 + }, + { + "epoch": 0.53, + "learning_rate": 1.8238476566693146e-05, + "loss": 1.0055, + "step": 8310 + }, + { + "epoch": 0.53, + "learning_rate": 1.8238261877670195e-05, + "loss": 1.1078, + "step": 8311 + }, + { + "epoch": 0.53, + "learning_rate": 1.8238047188647247e-05, + "loss": 1.0737, + "step": 8312 + }, + { + "epoch": 0.53, + "learning_rate": 1.8237832499624296e-05, + "loss": 1.1618, + "step": 8313 + }, + { + "epoch": 0.53, + "learning_rate": 1.8237617810601345e-05, + "loss": 1.0593, + "step": 8314 + }, + { + "epoch": 0.53, + "learning_rate": 1.8237403121578394e-05, + "loss": 1.0413, + "step": 8315 + }, + { + "epoch": 0.54, + "learning_rate": 1.8237188432555446e-05, + "loss": 1.0921, + "step": 8316 + }, + { + "epoch": 0.54, + "learning_rate": 1.8236973743532495e-05, + "loss": 1.0573, + "step": 8317 + }, + { + "epoch": 0.54, + "learning_rate": 1.8236759054509544e-05, + "loss": 1.065, + "step": 8318 + }, + { + "epoch": 0.54, + "learning_rate": 1.8236544365486596e-05, + "loss": 1.1986, + "step": 8319 + }, + { + "epoch": 0.54, + "learning_rate": 1.8236329676463645e-05, + "loss": 1.1385, + "step": 8320 + }, + { + "epoch": 0.54, + "learning_rate": 1.8236114987440694e-05, + "loss": 1.0907, + "step": 8321 + }, + { + "epoch": 0.54, + "learning_rate": 1.8235900298417743e-05, + "loss": 1.1461, + "step": 8322 + }, + { + "epoch": 0.54, + "learning_rate": 1.8235685609394792e-05, + "loss": 1.0828, + "step": 8323 + }, + { + "epoch": 0.54, + "learning_rate": 1.823547092037184e-05, + "loss": 1.064, + "step": 8324 + }, + { + "epoch": 0.54, + "learning_rate": 1.8235256231348894e-05, + "loss": 1.135, + "step": 8325 + }, + { + "epoch": 0.54, + "learning_rate": 1.8235041542325943e-05, + "loss": 1.0885, + "step": 8326 + }, + { + "epoch": 0.54, + "learning_rate": 1.823482685330299e-05, + "loss": 1.0939, + "step": 8327 + }, + { + "epoch": 0.54, + "learning_rate": 1.823461216428004e-05, + "loss": 1.0265, + "step": 8328 + }, + { + "epoch": 0.54, + "learning_rate": 1.823439747525709e-05, + "loss": 1.0857, + "step": 8329 + }, + { + "epoch": 0.54, + "learning_rate": 1.8234182786234142e-05, + "loss": 1.0874, + "step": 8330 + }, + { + "epoch": 0.54, + "learning_rate": 1.823396809721119e-05, + "loss": 1.1332, + "step": 8331 + }, + { + "epoch": 0.54, + "learning_rate": 1.8233753408188243e-05, + "loss": 1.0839, + "step": 8332 + }, + { + "epoch": 0.54, + "learning_rate": 1.8233538719165292e-05, + "loss": 1.0365, + "step": 8333 + }, + { + "epoch": 0.54, + "learning_rate": 1.823332403014234e-05, + "loss": 1.1407, + "step": 8334 + }, + { + "epoch": 0.54, + "learning_rate": 1.823310934111939e-05, + "loss": 1.1302, + "step": 8335 + }, + { + "epoch": 0.54, + "learning_rate": 1.823289465209644e-05, + "loss": 1.0686, + "step": 8336 + }, + { + "epoch": 0.54, + "learning_rate": 1.823267996307349e-05, + "loss": 1.0782, + "step": 8337 + }, + { + "epoch": 0.54, + "learning_rate": 1.823246527405054e-05, + "loss": 1.1352, + "step": 8338 + }, + { + "epoch": 0.54, + "learning_rate": 1.823225058502759e-05, + "loss": 1.1093, + "step": 8339 + }, + { + "epoch": 0.54, + "learning_rate": 1.8232035896004638e-05, + "loss": 1.057, + "step": 8340 + }, + { + "epoch": 0.54, + "learning_rate": 1.8231821206981687e-05, + "loss": 1.1032, + "step": 8341 + }, + { + "epoch": 0.54, + "learning_rate": 1.8231606517958736e-05, + "loss": 1.0774, + "step": 8342 + }, + { + "epoch": 0.54, + "learning_rate": 1.823139182893579e-05, + "loss": 1.0312, + "step": 8343 + }, + { + "epoch": 0.54, + "learning_rate": 1.8231177139912837e-05, + "loss": 1.0695, + "step": 8344 + }, + { + "epoch": 0.54, + "learning_rate": 1.823096245088989e-05, + "loss": 1.1042, + "step": 8345 + }, + { + "epoch": 0.54, + "learning_rate": 1.823074776186694e-05, + "loss": 1.0663, + "step": 8346 + }, + { + "epoch": 0.54, + "learning_rate": 1.8230533072843988e-05, + "loss": 1.1408, + "step": 8347 + }, + { + "epoch": 0.54, + "learning_rate": 1.8230318383821037e-05, + "loss": 1.0295, + "step": 8348 + }, + { + "epoch": 0.54, + "learning_rate": 1.8230103694798086e-05, + "loss": 1.0784, + "step": 8349 + }, + { + "epoch": 0.54, + "learning_rate": 1.8229889005775138e-05, + "loss": 0.9889, + "step": 8350 + }, + { + "epoch": 0.54, + "learning_rate": 1.8229674316752187e-05, + "loss": 1.1285, + "step": 8351 + }, + { + "epoch": 0.54, + "learning_rate": 1.8229459627729236e-05, + "loss": 1.1206, + "step": 8352 + }, + { + "epoch": 0.54, + "learning_rate": 1.8229244938706285e-05, + "loss": 1.0579, + "step": 8353 + }, + { + "epoch": 0.54, + "learning_rate": 1.8229030249683334e-05, + "loss": 1.0275, + "step": 8354 + }, + { + "epoch": 0.54, + "learning_rate": 1.8228815560660383e-05, + "loss": 1.1206, + "step": 8355 + }, + { + "epoch": 0.54, + "learning_rate": 1.8228600871637435e-05, + "loss": 1.0617, + "step": 8356 + }, + { + "epoch": 0.54, + "learning_rate": 1.8228386182614484e-05, + "loss": 1.1172, + "step": 8357 + }, + { + "epoch": 0.54, + "learning_rate": 1.8228171493591533e-05, + "loss": 1.1211, + "step": 8358 + }, + { + "epoch": 0.54, + "learning_rate": 1.8227956804568582e-05, + "loss": 1.0808, + "step": 8359 + }, + { + "epoch": 0.54, + "learning_rate": 1.8227742115545634e-05, + "loss": 1.0629, + "step": 8360 + }, + { + "epoch": 0.54, + "learning_rate": 1.8227527426522683e-05, + "loss": 1.0902, + "step": 8361 + }, + { + "epoch": 0.54, + "learning_rate": 1.8227312737499732e-05, + "loss": 0.9904, + "step": 8362 + }, + { + "epoch": 0.54, + "learning_rate": 1.8227098048476785e-05, + "loss": 1.1316, + "step": 8363 + }, + { + "epoch": 0.54, + "learning_rate": 1.8226883359453834e-05, + "loss": 1.1052, + "step": 8364 + }, + { + "epoch": 0.54, + "learning_rate": 1.8226668670430883e-05, + "loss": 1.0407, + "step": 8365 + }, + { + "epoch": 0.54, + "learning_rate": 1.822645398140793e-05, + "loss": 1.0993, + "step": 8366 + }, + { + "epoch": 0.54, + "learning_rate": 1.822623929238498e-05, + "loss": 1.0833, + "step": 8367 + }, + { + "epoch": 0.54, + "learning_rate": 1.8226024603362033e-05, + "loss": 1.0791, + "step": 8368 + }, + { + "epoch": 0.54, + "learning_rate": 1.8225809914339082e-05, + "loss": 1.1238, + "step": 8369 + }, + { + "epoch": 0.54, + "learning_rate": 1.822559522531613e-05, + "loss": 1.1409, + "step": 8370 + }, + { + "epoch": 0.54, + "learning_rate": 1.822538053629318e-05, + "loss": 1.0368, + "step": 8371 + }, + { + "epoch": 0.54, + "learning_rate": 1.822516584727023e-05, + "loss": 1.0741, + "step": 8372 + }, + { + "epoch": 0.54, + "learning_rate": 1.8224951158247278e-05, + "loss": 1.0254, + "step": 8373 + }, + { + "epoch": 0.54, + "learning_rate": 1.822473646922433e-05, + "loss": 1.066, + "step": 8374 + }, + { + "epoch": 0.54, + "learning_rate": 1.822452178020138e-05, + "loss": 0.9837, + "step": 8375 + }, + { + "epoch": 0.54, + "learning_rate": 1.822430709117843e-05, + "loss": 1.0574, + "step": 8376 + }, + { + "epoch": 0.54, + "learning_rate": 1.822409240215548e-05, + "loss": 1.1057, + "step": 8377 + }, + { + "epoch": 0.54, + "learning_rate": 1.822387771313253e-05, + "loss": 0.9901, + "step": 8378 + }, + { + "epoch": 0.54, + "learning_rate": 1.822366302410958e-05, + "loss": 1.1151, + "step": 8379 + }, + { + "epoch": 0.54, + "learning_rate": 1.8223448335086627e-05, + "loss": 1.0234, + "step": 8380 + }, + { + "epoch": 0.54, + "learning_rate": 1.822323364606368e-05, + "loss": 1.1214, + "step": 8381 + }, + { + "epoch": 0.54, + "learning_rate": 1.822301895704073e-05, + "loss": 1.1232, + "step": 8382 + }, + { + "epoch": 0.54, + "learning_rate": 1.8222804268017778e-05, + "loss": 1.0895, + "step": 8383 + }, + { + "epoch": 0.54, + "learning_rate": 1.8222589578994827e-05, + "loss": 1.0364, + "step": 8384 + }, + { + "epoch": 0.54, + "learning_rate": 1.8222374889971876e-05, + "loss": 1.1005, + "step": 8385 + }, + { + "epoch": 0.54, + "learning_rate": 1.8222160200948928e-05, + "loss": 1.0783, + "step": 8386 + }, + { + "epoch": 0.54, + "learning_rate": 1.8221945511925977e-05, + "loss": 1.1233, + "step": 8387 + }, + { + "epoch": 0.54, + "learning_rate": 1.8221730822903026e-05, + "loss": 1.1329, + "step": 8388 + }, + { + "epoch": 0.54, + "learning_rate": 1.8221516133880078e-05, + "loss": 0.993, + "step": 8389 + }, + { + "epoch": 0.54, + "learning_rate": 1.8221301444857127e-05, + "loss": 1.0515, + "step": 8390 + }, + { + "epoch": 0.54, + "learning_rate": 1.8221086755834176e-05, + "loss": 1.057, + "step": 8391 + }, + { + "epoch": 0.54, + "learning_rate": 1.8220872066811225e-05, + "loss": 1.0465, + "step": 8392 + }, + { + "epoch": 0.54, + "learning_rate": 1.8220657377788274e-05, + "loss": 1.0742, + "step": 8393 + }, + { + "epoch": 0.54, + "learning_rate": 1.8220442688765326e-05, + "loss": 0.9773, + "step": 8394 + }, + { + "epoch": 0.54, + "learning_rate": 1.8220227999742375e-05, + "loss": 1.1233, + "step": 8395 + }, + { + "epoch": 0.54, + "learning_rate": 1.8220013310719424e-05, + "loss": 1.1162, + "step": 8396 + }, + { + "epoch": 0.54, + "learning_rate": 1.8219798621696473e-05, + "loss": 1.1347, + "step": 8397 + }, + { + "epoch": 0.54, + "learning_rate": 1.8219583932673522e-05, + "loss": 1.0479, + "step": 8398 + }, + { + "epoch": 0.54, + "learning_rate": 1.8219369243650575e-05, + "loss": 1.0719, + "step": 8399 + }, + { + "epoch": 0.54, + "learning_rate": 1.8219154554627624e-05, + "loss": 1.0979, + "step": 8400 + }, + { + "epoch": 0.54, + "learning_rate": 1.8218939865604673e-05, + "loss": 1.1261, + "step": 8401 + }, + { + "epoch": 0.54, + "learning_rate": 1.821872517658172e-05, + "loss": 1.1502, + "step": 8402 + }, + { + "epoch": 0.54, + "learning_rate": 1.8218510487558774e-05, + "loss": 1.0814, + "step": 8403 + }, + { + "epoch": 0.54, + "learning_rate": 1.8218295798535823e-05, + "loss": 1.2006, + "step": 8404 + }, + { + "epoch": 0.54, + "learning_rate": 1.8218081109512872e-05, + "loss": 1.1173, + "step": 8405 + }, + { + "epoch": 0.54, + "learning_rate": 1.8217866420489924e-05, + "loss": 1.057, + "step": 8406 + }, + { + "epoch": 0.54, + "learning_rate": 1.8217651731466973e-05, + "loss": 1.0662, + "step": 8407 + }, + { + "epoch": 0.54, + "learning_rate": 1.8217437042444022e-05, + "loss": 1.0402, + "step": 8408 + }, + { + "epoch": 0.54, + "learning_rate": 1.821722235342107e-05, + "loss": 1.0858, + "step": 8409 + }, + { + "epoch": 0.54, + "learning_rate": 1.821700766439812e-05, + "loss": 1.0819, + "step": 8410 + }, + { + "epoch": 0.54, + "learning_rate": 1.821679297537517e-05, + "loss": 1.0719, + "step": 8411 + }, + { + "epoch": 0.54, + "learning_rate": 1.821657828635222e-05, + "loss": 1.0993, + "step": 8412 + }, + { + "epoch": 0.54, + "learning_rate": 1.821636359732927e-05, + "loss": 1.0418, + "step": 8413 + }, + { + "epoch": 0.54, + "learning_rate": 1.821614890830632e-05, + "loss": 1.1267, + "step": 8414 + }, + { + "epoch": 0.54, + "learning_rate": 1.8215934219283368e-05, + "loss": 1.1094, + "step": 8415 + }, + { + "epoch": 0.54, + "learning_rate": 1.8215719530260417e-05, + "loss": 1.1025, + "step": 8416 + }, + { + "epoch": 0.54, + "learning_rate": 1.821550484123747e-05, + "loss": 1.0742, + "step": 8417 + }, + { + "epoch": 0.54, + "learning_rate": 1.821529015221452e-05, + "loss": 1.1459, + "step": 8418 + }, + { + "epoch": 0.54, + "learning_rate": 1.821507546319157e-05, + "loss": 1.1113, + "step": 8419 + }, + { + "epoch": 0.54, + "learning_rate": 1.821486077416862e-05, + "loss": 1.0647, + "step": 8420 + }, + { + "epoch": 0.54, + "learning_rate": 1.821464608514567e-05, + "loss": 1.0215, + "step": 8421 + }, + { + "epoch": 0.54, + "learning_rate": 1.8214431396122718e-05, + "loss": 1.1119, + "step": 8422 + }, + { + "epoch": 0.54, + "learning_rate": 1.8214216707099767e-05, + "loss": 1.058, + "step": 8423 + }, + { + "epoch": 0.54, + "learning_rate": 1.8214002018076816e-05, + "loss": 1.0605, + "step": 8424 + }, + { + "epoch": 0.54, + "learning_rate": 1.8213787329053868e-05, + "loss": 0.9818, + "step": 8425 + }, + { + "epoch": 0.54, + "learning_rate": 1.8213572640030917e-05, + "loss": 1.0806, + "step": 8426 + }, + { + "epoch": 0.54, + "learning_rate": 1.8213357951007966e-05, + "loss": 1.0355, + "step": 8427 + }, + { + "epoch": 0.54, + "learning_rate": 1.8213143261985015e-05, + "loss": 1.0788, + "step": 8428 + }, + { + "epoch": 0.54, + "learning_rate": 1.8212928572962064e-05, + "loss": 1.028, + "step": 8429 + }, + { + "epoch": 0.54, + "learning_rate": 1.8212713883939116e-05, + "loss": 1.0758, + "step": 8430 + }, + { + "epoch": 0.54, + "learning_rate": 1.8212499194916165e-05, + "loss": 1.0652, + "step": 8431 + }, + { + "epoch": 0.54, + "learning_rate": 1.8212284505893214e-05, + "loss": 1.0755, + "step": 8432 + }, + { + "epoch": 0.54, + "learning_rate": 1.8212069816870266e-05, + "loss": 1.0231, + "step": 8433 + }, + { + "epoch": 0.54, + "learning_rate": 1.8211855127847315e-05, + "loss": 1.0261, + "step": 8434 + }, + { + "epoch": 0.54, + "learning_rate": 1.8211640438824364e-05, + "loss": 1.0472, + "step": 8435 + }, + { + "epoch": 0.54, + "learning_rate": 1.8211425749801413e-05, + "loss": 1.1724, + "step": 8436 + }, + { + "epoch": 0.54, + "learning_rate": 1.8211211060778466e-05, + "loss": 1.0892, + "step": 8437 + }, + { + "epoch": 0.54, + "learning_rate": 1.8210996371755515e-05, + "loss": 1.1086, + "step": 8438 + }, + { + "epoch": 0.54, + "learning_rate": 1.8210781682732564e-05, + "loss": 1.0511, + "step": 8439 + }, + { + "epoch": 0.54, + "learning_rate": 1.8210566993709613e-05, + "loss": 0.9428, + "step": 8440 + }, + { + "epoch": 0.54, + "learning_rate": 1.821035230468666e-05, + "loss": 1.115, + "step": 8441 + }, + { + "epoch": 0.54, + "learning_rate": 1.821013761566371e-05, + "loss": 1.0368, + "step": 8442 + }, + { + "epoch": 0.54, + "learning_rate": 1.8209922926640763e-05, + "loss": 1.1179, + "step": 8443 + }, + { + "epoch": 0.54, + "learning_rate": 1.8209708237617812e-05, + "loss": 1.0634, + "step": 8444 + }, + { + "epoch": 0.54, + "learning_rate": 1.820949354859486e-05, + "loss": 1.077, + "step": 8445 + }, + { + "epoch": 0.54, + "learning_rate": 1.820927885957191e-05, + "loss": 1.0888, + "step": 8446 + }, + { + "epoch": 0.54, + "learning_rate": 1.8209064170548962e-05, + "loss": 1.1137, + "step": 8447 + }, + { + "epoch": 0.54, + "learning_rate": 1.820884948152601e-05, + "loss": 1.1243, + "step": 8448 + }, + { + "epoch": 0.54, + "learning_rate": 1.820863479250306e-05, + "loss": 1.0779, + "step": 8449 + }, + { + "epoch": 0.54, + "learning_rate": 1.8208420103480112e-05, + "loss": 1.0474, + "step": 8450 + }, + { + "epoch": 0.54, + "learning_rate": 1.820820541445716e-05, + "loss": 1.0827, + "step": 8451 + }, + { + "epoch": 0.54, + "learning_rate": 1.820799072543421e-05, + "loss": 1.1353, + "step": 8452 + }, + { + "epoch": 0.54, + "learning_rate": 1.820777603641126e-05, + "loss": 1.0739, + "step": 8453 + }, + { + "epoch": 0.54, + "learning_rate": 1.820756134738831e-05, + "loss": 1.1408, + "step": 8454 + }, + { + "epoch": 0.54, + "learning_rate": 1.820734665836536e-05, + "loss": 1.0782, + "step": 8455 + }, + { + "epoch": 0.54, + "learning_rate": 1.820713196934241e-05, + "loss": 1.1066, + "step": 8456 + }, + { + "epoch": 0.54, + "learning_rate": 1.820691728031946e-05, + "loss": 1.0596, + "step": 8457 + }, + { + "epoch": 0.54, + "learning_rate": 1.8206702591296508e-05, + "loss": 1.1264, + "step": 8458 + }, + { + "epoch": 0.54, + "learning_rate": 1.8206487902273557e-05, + "loss": 1.1095, + "step": 8459 + }, + { + "epoch": 0.54, + "learning_rate": 1.8206273213250605e-05, + "loss": 1.0931, + "step": 8460 + }, + { + "epoch": 0.54, + "learning_rate": 1.8206058524227658e-05, + "loss": 1.1337, + "step": 8461 + }, + { + "epoch": 0.54, + "learning_rate": 1.8205843835204707e-05, + "loss": 1.1339, + "step": 8462 + }, + { + "epoch": 0.54, + "learning_rate": 1.820562914618176e-05, + "loss": 1.0377, + "step": 8463 + }, + { + "epoch": 0.54, + "learning_rate": 1.8205414457158808e-05, + "loss": 1.1528, + "step": 8464 + }, + { + "epoch": 0.54, + "learning_rate": 1.8205199768135857e-05, + "loss": 0.9894, + "step": 8465 + }, + { + "epoch": 0.54, + "learning_rate": 1.8204985079112906e-05, + "loss": 1.0319, + "step": 8466 + }, + { + "epoch": 0.54, + "learning_rate": 1.8204770390089955e-05, + "loss": 1.0572, + "step": 8467 + }, + { + "epoch": 0.54, + "learning_rate": 1.8204555701067007e-05, + "loss": 1.0348, + "step": 8468 + }, + { + "epoch": 0.54, + "learning_rate": 1.8204341012044056e-05, + "loss": 1.0845, + "step": 8469 + }, + { + "epoch": 0.54, + "learning_rate": 1.8204126323021105e-05, + "loss": 1.1557, + "step": 8470 + }, + { + "epoch": 0.55, + "learning_rate": 1.8203911633998154e-05, + "loss": 1.103, + "step": 8471 + }, + { + "epoch": 0.55, + "learning_rate": 1.8203696944975203e-05, + "loss": 1.0756, + "step": 8472 + }, + { + "epoch": 0.55, + "learning_rate": 1.8203482255952256e-05, + "loss": 1.1442, + "step": 8473 + }, + { + "epoch": 0.55, + "learning_rate": 1.8203267566929305e-05, + "loss": 1.1048, + "step": 8474 + }, + { + "epoch": 0.55, + "learning_rate": 1.8203052877906354e-05, + "loss": 1.0909, + "step": 8475 + }, + { + "epoch": 0.55, + "learning_rate": 1.8202838188883406e-05, + "loss": 1.0643, + "step": 8476 + }, + { + "epoch": 0.55, + "learning_rate": 1.8202623499860455e-05, + "loss": 1.1064, + "step": 8477 + }, + { + "epoch": 0.55, + "learning_rate": 1.8202408810837504e-05, + "loss": 1.0226, + "step": 8478 + }, + { + "epoch": 0.55, + "learning_rate": 1.8202194121814553e-05, + "loss": 1.077, + "step": 8479 + }, + { + "epoch": 0.55, + "learning_rate": 1.8201979432791602e-05, + "loss": 1.059, + "step": 8480 + }, + { + "epoch": 0.55, + "learning_rate": 1.8201764743768654e-05, + "loss": 1.078, + "step": 8481 + }, + { + "epoch": 0.55, + "learning_rate": 1.8201550054745703e-05, + "loss": 1.1862, + "step": 8482 + }, + { + "epoch": 0.55, + "learning_rate": 1.8201335365722752e-05, + "loss": 1.0778, + "step": 8483 + }, + { + "epoch": 0.55, + "learning_rate": 1.82011206766998e-05, + "loss": 0.9821, + "step": 8484 + }, + { + "epoch": 0.55, + "learning_rate": 1.820090598767685e-05, + "loss": 1.0985, + "step": 8485 + }, + { + "epoch": 0.55, + "learning_rate": 1.8200691298653902e-05, + "loss": 1.0027, + "step": 8486 + }, + { + "epoch": 0.55, + "learning_rate": 1.820047660963095e-05, + "loss": 1.0437, + "step": 8487 + }, + { + "epoch": 0.55, + "learning_rate": 1.8200261920608e-05, + "loss": 1.0597, + "step": 8488 + }, + { + "epoch": 0.55, + "learning_rate": 1.820004723158505e-05, + "loss": 1.0899, + "step": 8489 + }, + { + "epoch": 0.55, + "learning_rate": 1.8199832542562098e-05, + "loss": 1.0685, + "step": 8490 + }, + { + "epoch": 0.55, + "learning_rate": 1.819961785353915e-05, + "loss": 1.1049, + "step": 8491 + }, + { + "epoch": 0.55, + "learning_rate": 1.81994031645162e-05, + "loss": 1.1328, + "step": 8492 + }, + { + "epoch": 0.55, + "learning_rate": 1.8199188475493252e-05, + "loss": 1.0031, + "step": 8493 + }, + { + "epoch": 0.55, + "learning_rate": 1.81989737864703e-05, + "loss": 1.1732, + "step": 8494 + }, + { + "epoch": 0.55, + "learning_rate": 1.819875909744735e-05, + "loss": 0.9984, + "step": 8495 + }, + { + "epoch": 0.55, + "learning_rate": 1.81985444084244e-05, + "loss": 1.1366, + "step": 8496 + }, + { + "epoch": 0.55, + "learning_rate": 1.8198329719401448e-05, + "loss": 1.1919, + "step": 8497 + }, + { + "epoch": 0.55, + "learning_rate": 1.8198115030378497e-05, + "loss": 1.1152, + "step": 8498 + }, + { + "epoch": 0.55, + "learning_rate": 1.819790034135555e-05, + "loss": 1.1216, + "step": 8499 + }, + { + "epoch": 0.55, + "learning_rate": 1.8197685652332598e-05, + "loss": 1.0967, + "step": 8500 + }, + { + "epoch": 0.55, + "learning_rate": 1.8197470963309647e-05, + "loss": 1.1067, + "step": 8501 + }, + { + "epoch": 0.55, + "learning_rate": 1.8197256274286696e-05, + "loss": 1.0805, + "step": 8502 + }, + { + "epoch": 0.55, + "learning_rate": 1.8197041585263745e-05, + "loss": 1.123, + "step": 8503 + }, + { + "epoch": 0.55, + "learning_rate": 1.8196826896240797e-05, + "loss": 0.9595, + "step": 8504 + }, + { + "epoch": 0.55, + "learning_rate": 1.8196612207217846e-05, + "loss": 1.226, + "step": 8505 + }, + { + "epoch": 0.55, + "learning_rate": 1.81963975181949e-05, + "loss": 1.0864, + "step": 8506 + }, + { + "epoch": 0.55, + "learning_rate": 1.8196182829171948e-05, + "loss": 1.0569, + "step": 8507 + }, + { + "epoch": 0.55, + "learning_rate": 1.8195968140148996e-05, + "loss": 1.1766, + "step": 8508 + }, + { + "epoch": 0.55, + "learning_rate": 1.8195753451126045e-05, + "loss": 1.0919, + "step": 8509 + }, + { + "epoch": 0.55, + "learning_rate": 1.8195538762103094e-05, + "loss": 1.1028, + "step": 8510 + }, + { + "epoch": 0.55, + "learning_rate": 1.8195324073080143e-05, + "loss": 1.0937, + "step": 8511 + }, + { + "epoch": 0.55, + "learning_rate": 1.8195109384057196e-05, + "loss": 0.9874, + "step": 8512 + }, + { + "epoch": 0.55, + "learning_rate": 1.8194894695034245e-05, + "loss": 1.1288, + "step": 8513 + }, + { + "epoch": 0.55, + "learning_rate": 1.8194680006011294e-05, + "loss": 1.1393, + "step": 8514 + }, + { + "epoch": 0.55, + "learning_rate": 1.8194465316988343e-05, + "loss": 1.0495, + "step": 8515 + }, + { + "epoch": 0.55, + "learning_rate": 1.819425062796539e-05, + "loss": 1.0727, + "step": 8516 + }, + { + "epoch": 0.55, + "learning_rate": 1.8194035938942444e-05, + "loss": 1.0271, + "step": 8517 + }, + { + "epoch": 0.55, + "learning_rate": 1.8193821249919493e-05, + "loss": 1.172, + "step": 8518 + }, + { + "epoch": 0.55, + "learning_rate": 1.8193606560896542e-05, + "loss": 1.167, + "step": 8519 + }, + { + "epoch": 0.55, + "learning_rate": 1.8193391871873594e-05, + "loss": 1.0077, + "step": 8520 + }, + { + "epoch": 0.55, + "learning_rate": 1.8193177182850643e-05, + "loss": 1.0647, + "step": 8521 + }, + { + "epoch": 0.55, + "learning_rate": 1.8192962493827692e-05, + "loss": 1.0702, + "step": 8522 + }, + { + "epoch": 0.55, + "learning_rate": 1.819274780480474e-05, + "loss": 1.0498, + "step": 8523 + }, + { + "epoch": 0.55, + "learning_rate": 1.8192533115781793e-05, + "loss": 1.0939, + "step": 8524 + }, + { + "epoch": 0.55, + "learning_rate": 1.8192318426758842e-05, + "loss": 1.0856, + "step": 8525 + }, + { + "epoch": 0.55, + "learning_rate": 1.819210373773589e-05, + "loss": 1.1095, + "step": 8526 + }, + { + "epoch": 0.55, + "learning_rate": 1.819188904871294e-05, + "loss": 1.0989, + "step": 8527 + }, + { + "epoch": 0.55, + "learning_rate": 1.819167435968999e-05, + "loss": 1.1245, + "step": 8528 + }, + { + "epoch": 0.55, + "learning_rate": 1.8191459670667038e-05, + "loss": 1.1502, + "step": 8529 + }, + { + "epoch": 0.55, + "learning_rate": 1.819124498164409e-05, + "loss": 1.0768, + "step": 8530 + }, + { + "epoch": 0.55, + "learning_rate": 1.819103029262114e-05, + "loss": 1.0594, + "step": 8531 + }, + { + "epoch": 0.55, + "learning_rate": 1.819081560359819e-05, + "loss": 1.1196, + "step": 8532 + }, + { + "epoch": 0.55, + "learning_rate": 1.8190600914575238e-05, + "loss": 1.0662, + "step": 8533 + }, + { + "epoch": 0.55, + "learning_rate": 1.819038622555229e-05, + "loss": 1.0113, + "step": 8534 + }, + { + "epoch": 0.55, + "learning_rate": 1.819017153652934e-05, + "loss": 1.0711, + "step": 8535 + }, + { + "epoch": 0.55, + "learning_rate": 1.8189956847506388e-05, + "loss": 1.0259, + "step": 8536 + }, + { + "epoch": 0.55, + "learning_rate": 1.818974215848344e-05, + "loss": 1.0921, + "step": 8537 + }, + { + "epoch": 0.55, + "learning_rate": 1.818952746946049e-05, + "loss": 1.0906, + "step": 8538 + }, + { + "epoch": 0.55, + "learning_rate": 1.8189312780437538e-05, + "loss": 1.0686, + "step": 8539 + }, + { + "epoch": 0.55, + "learning_rate": 1.8189098091414587e-05, + "loss": 1.1385, + "step": 8540 + }, + { + "epoch": 0.55, + "learning_rate": 1.8188883402391636e-05, + "loss": 1.1489, + "step": 8541 + }, + { + "epoch": 0.55, + "learning_rate": 1.818866871336869e-05, + "loss": 1.1377, + "step": 8542 + }, + { + "epoch": 0.55, + "learning_rate": 1.8188454024345737e-05, + "loss": 1.0348, + "step": 8543 + }, + { + "epoch": 0.55, + "learning_rate": 1.8188239335322786e-05, + "loss": 1.1298, + "step": 8544 + }, + { + "epoch": 0.55, + "learning_rate": 1.8188024646299835e-05, + "loss": 1.1307, + "step": 8545 + }, + { + "epoch": 0.55, + "learning_rate": 1.8187809957276884e-05, + "loss": 1.0961, + "step": 8546 + }, + { + "epoch": 0.55, + "learning_rate": 1.8187595268253933e-05, + "loss": 1.0878, + "step": 8547 + }, + { + "epoch": 0.55, + "learning_rate": 1.8187380579230986e-05, + "loss": 1.0702, + "step": 8548 + }, + { + "epoch": 0.55, + "learning_rate": 1.8187165890208035e-05, + "loss": 1.1695, + "step": 8549 + }, + { + "epoch": 0.55, + "learning_rate": 1.8186951201185087e-05, + "loss": 0.9735, + "step": 8550 + }, + { + "epoch": 0.55, + "learning_rate": 1.8186736512162136e-05, + "loss": 1.0886, + "step": 8551 + }, + { + "epoch": 0.55, + "learning_rate": 1.8186521823139185e-05, + "loss": 1.0184, + "step": 8552 + }, + { + "epoch": 0.55, + "learning_rate": 1.8186307134116234e-05, + "loss": 1.0664, + "step": 8553 + }, + { + "epoch": 0.55, + "learning_rate": 1.8186092445093283e-05, + "loss": 1.1285, + "step": 8554 + }, + { + "epoch": 0.55, + "learning_rate": 1.8185877756070335e-05, + "loss": 1.1138, + "step": 8555 + }, + { + "epoch": 0.55, + "learning_rate": 1.8185663067047384e-05, + "loss": 1.0657, + "step": 8556 + }, + { + "epoch": 0.55, + "learning_rate": 1.8185448378024433e-05, + "loss": 1.0271, + "step": 8557 + }, + { + "epoch": 0.55, + "learning_rate": 1.8185233689001482e-05, + "loss": 1.0367, + "step": 8558 + }, + { + "epoch": 0.55, + "learning_rate": 1.818501899997853e-05, + "loss": 1.0791, + "step": 8559 + }, + { + "epoch": 0.55, + "learning_rate": 1.8184804310955583e-05, + "loss": 1.0701, + "step": 8560 + }, + { + "epoch": 0.55, + "learning_rate": 1.8184589621932632e-05, + "loss": 1.1456, + "step": 8561 + }, + { + "epoch": 0.55, + "learning_rate": 1.818437493290968e-05, + "loss": 1.0497, + "step": 8562 + }, + { + "epoch": 0.55, + "learning_rate": 1.8184160243886734e-05, + "loss": 1.1011, + "step": 8563 + }, + { + "epoch": 0.55, + "learning_rate": 1.8183945554863783e-05, + "loss": 1.108, + "step": 8564 + }, + { + "epoch": 0.55, + "learning_rate": 1.818373086584083e-05, + "loss": 1.0654, + "step": 8565 + }, + { + "epoch": 0.55, + "learning_rate": 1.818351617681788e-05, + "loss": 1.1437, + "step": 8566 + }, + { + "epoch": 0.55, + "learning_rate": 1.818330148779493e-05, + "loss": 1.0861, + "step": 8567 + }, + { + "epoch": 0.55, + "learning_rate": 1.8183086798771982e-05, + "loss": 1.0812, + "step": 8568 + }, + { + "epoch": 0.55, + "learning_rate": 1.818287210974903e-05, + "loss": 1.0794, + "step": 8569 + }, + { + "epoch": 0.55, + "learning_rate": 1.818265742072608e-05, + "loss": 1.1337, + "step": 8570 + }, + { + "epoch": 0.55, + "learning_rate": 1.818244273170313e-05, + "loss": 1.067, + "step": 8571 + }, + { + "epoch": 0.55, + "learning_rate": 1.8182228042680178e-05, + "loss": 1.0067, + "step": 8572 + }, + { + "epoch": 0.55, + "learning_rate": 1.818201335365723e-05, + "loss": 1.0429, + "step": 8573 + }, + { + "epoch": 0.55, + "learning_rate": 1.818179866463428e-05, + "loss": 1.0246, + "step": 8574 + }, + { + "epoch": 0.55, + "learning_rate": 1.8181583975611328e-05, + "loss": 1.1041, + "step": 8575 + }, + { + "epoch": 0.55, + "learning_rate": 1.8181369286588377e-05, + "loss": 1.0105, + "step": 8576 + }, + { + "epoch": 0.55, + "learning_rate": 1.8181154597565426e-05, + "loss": 1.1085, + "step": 8577 + }, + { + "epoch": 0.55, + "learning_rate": 1.8180939908542478e-05, + "loss": 1.1823, + "step": 8578 + }, + { + "epoch": 0.55, + "learning_rate": 1.8180725219519527e-05, + "loss": 1.0132, + "step": 8579 + }, + { + "epoch": 0.55, + "learning_rate": 1.818051053049658e-05, + "loss": 1.1319, + "step": 8580 + }, + { + "epoch": 0.55, + "learning_rate": 1.818029584147363e-05, + "loss": 1.0745, + "step": 8581 + }, + { + "epoch": 0.55, + "learning_rate": 1.8180081152450677e-05, + "loss": 1.113, + "step": 8582 + }, + { + "epoch": 0.55, + "learning_rate": 1.8179866463427726e-05, + "loss": 0.9913, + "step": 8583 + }, + { + "epoch": 0.55, + "learning_rate": 1.8179651774404775e-05, + "loss": 1.1522, + "step": 8584 + }, + { + "epoch": 0.55, + "learning_rate": 1.8179437085381824e-05, + "loss": 1.1807, + "step": 8585 + }, + { + "epoch": 0.55, + "learning_rate": 1.8179222396358877e-05, + "loss": 1.032, + "step": 8586 + }, + { + "epoch": 0.55, + "learning_rate": 1.8179007707335926e-05, + "loss": 1.0926, + "step": 8587 + }, + { + "epoch": 0.55, + "learning_rate": 1.8178793018312975e-05, + "loss": 1.027, + "step": 8588 + }, + { + "epoch": 0.55, + "learning_rate": 1.8178578329290024e-05, + "loss": 1.0196, + "step": 8589 + }, + { + "epoch": 0.55, + "learning_rate": 1.8178363640267073e-05, + "loss": 1.1346, + "step": 8590 + }, + { + "epoch": 0.55, + "learning_rate": 1.8178148951244125e-05, + "loss": 1.1478, + "step": 8591 + }, + { + "epoch": 0.55, + "learning_rate": 1.8177934262221174e-05, + "loss": 1.1595, + "step": 8592 + }, + { + "epoch": 0.55, + "learning_rate": 1.8177719573198226e-05, + "loss": 1.1392, + "step": 8593 + }, + { + "epoch": 0.55, + "learning_rate": 1.8177504884175275e-05, + "loss": 1.0187, + "step": 8594 + }, + { + "epoch": 0.55, + "learning_rate": 1.8177290195152324e-05, + "loss": 1.1029, + "step": 8595 + }, + { + "epoch": 0.55, + "learning_rate": 1.8177075506129373e-05, + "loss": 1.0415, + "step": 8596 + }, + { + "epoch": 0.55, + "learning_rate": 1.8176860817106422e-05, + "loss": 1.0597, + "step": 8597 + }, + { + "epoch": 0.55, + "learning_rate": 1.817664612808347e-05, + "loss": 1.1271, + "step": 8598 + }, + { + "epoch": 0.55, + "learning_rate": 1.8176431439060523e-05, + "loss": 1.089, + "step": 8599 + }, + { + "epoch": 0.55, + "learning_rate": 1.8176216750037572e-05, + "loss": 1.0441, + "step": 8600 + }, + { + "epoch": 0.55, + "learning_rate": 1.817600206101462e-05, + "loss": 1.0744, + "step": 8601 + }, + { + "epoch": 0.55, + "learning_rate": 1.817578737199167e-05, + "loss": 1.0724, + "step": 8602 + }, + { + "epoch": 0.55, + "learning_rate": 1.817557268296872e-05, + "loss": 1.0731, + "step": 8603 + }, + { + "epoch": 0.55, + "learning_rate": 1.817535799394577e-05, + "loss": 1.041, + "step": 8604 + }, + { + "epoch": 0.55, + "learning_rate": 1.817514330492282e-05, + "loss": 1.0496, + "step": 8605 + }, + { + "epoch": 0.55, + "learning_rate": 1.817492861589987e-05, + "loss": 1.0518, + "step": 8606 + }, + { + "epoch": 0.55, + "learning_rate": 1.8174713926876922e-05, + "loss": 1.1084, + "step": 8607 + }, + { + "epoch": 0.55, + "learning_rate": 1.817449923785397e-05, + "loss": 1.1741, + "step": 8608 + }, + { + "epoch": 0.55, + "learning_rate": 1.817428454883102e-05, + "loss": 1.1259, + "step": 8609 + }, + { + "epoch": 0.55, + "learning_rate": 1.817406985980807e-05, + "loss": 1.075, + "step": 8610 + }, + { + "epoch": 0.55, + "learning_rate": 1.817385517078512e-05, + "loss": 1.1049, + "step": 8611 + }, + { + "epoch": 0.55, + "learning_rate": 1.817364048176217e-05, + "loss": 1.0826, + "step": 8612 + }, + { + "epoch": 0.55, + "learning_rate": 1.817342579273922e-05, + "loss": 1.0728, + "step": 8613 + }, + { + "epoch": 0.55, + "learning_rate": 1.8173211103716268e-05, + "loss": 1.1082, + "step": 8614 + }, + { + "epoch": 0.55, + "learning_rate": 1.8172996414693317e-05, + "loss": 1.0743, + "step": 8615 + }, + { + "epoch": 0.55, + "learning_rate": 1.8172781725670366e-05, + "loss": 1.0824, + "step": 8616 + }, + { + "epoch": 0.55, + "learning_rate": 1.817256703664742e-05, + "loss": 1.1413, + "step": 8617 + }, + { + "epoch": 0.55, + "learning_rate": 1.8172352347624467e-05, + "loss": 1.1199, + "step": 8618 + }, + { + "epoch": 0.55, + "learning_rate": 1.8172137658601516e-05, + "loss": 1.1664, + "step": 8619 + }, + { + "epoch": 0.55, + "learning_rate": 1.8171922969578565e-05, + "loss": 1.0544, + "step": 8620 + }, + { + "epoch": 0.55, + "learning_rate": 1.8171708280555618e-05, + "loss": 1.0989, + "step": 8621 + }, + { + "epoch": 0.55, + "learning_rate": 1.8171493591532667e-05, + "loss": 1.2033, + "step": 8622 + }, + { + "epoch": 0.55, + "learning_rate": 1.8171278902509716e-05, + "loss": 1.0736, + "step": 8623 + }, + { + "epoch": 0.55, + "learning_rate": 1.8171064213486768e-05, + "loss": 1.0543, + "step": 8624 + }, + { + "epoch": 0.55, + "learning_rate": 1.8170849524463817e-05, + "loss": 1.0372, + "step": 8625 + }, + { + "epoch": 0.55, + "learning_rate": 1.8170634835440866e-05, + "loss": 1.1085, + "step": 8626 + }, + { + "epoch": 0.56, + "learning_rate": 1.8170420146417915e-05, + "loss": 1.1027, + "step": 8627 + }, + { + "epoch": 0.56, + "learning_rate": 1.8170205457394964e-05, + "loss": 1.1352, + "step": 8628 + }, + { + "epoch": 0.56, + "learning_rate": 1.8169990768372016e-05, + "loss": 1.0602, + "step": 8629 + }, + { + "epoch": 0.56, + "learning_rate": 1.8169776079349065e-05, + "loss": 1.0935, + "step": 8630 + }, + { + "epoch": 0.56, + "learning_rate": 1.8169561390326114e-05, + "loss": 1.0789, + "step": 8631 + }, + { + "epoch": 0.56, + "learning_rate": 1.8169346701303163e-05, + "loss": 0.9554, + "step": 8632 + }, + { + "epoch": 0.56, + "learning_rate": 1.8169132012280212e-05, + "loss": 1.0275, + "step": 8633 + }, + { + "epoch": 0.56, + "learning_rate": 1.816891732325726e-05, + "loss": 1.1111, + "step": 8634 + }, + { + "epoch": 0.56, + "learning_rate": 1.8168702634234313e-05, + "loss": 1.0756, + "step": 8635 + }, + { + "epoch": 0.56, + "learning_rate": 1.8168487945211362e-05, + "loss": 1.1418, + "step": 8636 + }, + { + "epoch": 0.56, + "learning_rate": 1.8168273256188415e-05, + "loss": 1.0828, + "step": 8637 + }, + { + "epoch": 0.56, + "learning_rate": 1.8168058567165464e-05, + "loss": 1.0736, + "step": 8638 + }, + { + "epoch": 0.56, + "learning_rate": 1.8167843878142513e-05, + "loss": 1.1542, + "step": 8639 + }, + { + "epoch": 0.56, + "learning_rate": 1.816762918911956e-05, + "loss": 1.169, + "step": 8640 + }, + { + "epoch": 0.56, + "learning_rate": 1.816741450009661e-05, + "loss": 1.0183, + "step": 8641 + }, + { + "epoch": 0.56, + "learning_rate": 1.8167199811073663e-05, + "loss": 1.0361, + "step": 8642 + }, + { + "epoch": 0.56, + "learning_rate": 1.8166985122050712e-05, + "loss": 1.1147, + "step": 8643 + }, + { + "epoch": 0.56, + "learning_rate": 1.816677043302776e-05, + "loss": 1.1498, + "step": 8644 + }, + { + "epoch": 0.56, + "learning_rate": 1.816655574400481e-05, + "loss": 1.1034, + "step": 8645 + }, + { + "epoch": 0.56, + "learning_rate": 1.816634105498186e-05, + "loss": 1.0446, + "step": 8646 + }, + { + "epoch": 0.56, + "learning_rate": 1.816612636595891e-05, + "loss": 1.0511, + "step": 8647 + }, + { + "epoch": 0.56, + "learning_rate": 1.816591167693596e-05, + "loss": 1.0744, + "step": 8648 + }, + { + "epoch": 0.56, + "learning_rate": 1.816569698791301e-05, + "loss": 1.1109, + "step": 8649 + }, + { + "epoch": 0.56, + "learning_rate": 1.816548229889006e-05, + "loss": 1.1202, + "step": 8650 + }, + { + "epoch": 0.56, + "learning_rate": 1.816526760986711e-05, + "loss": 1.1049, + "step": 8651 + }, + { + "epoch": 0.56, + "learning_rate": 1.816505292084416e-05, + "loss": 1.1829, + "step": 8652 + }, + { + "epoch": 0.56, + "learning_rate": 1.8164838231821208e-05, + "loss": 1.1183, + "step": 8653 + }, + { + "epoch": 0.56, + "learning_rate": 1.8164623542798257e-05, + "loss": 1.0491, + "step": 8654 + }, + { + "epoch": 0.56, + "learning_rate": 1.816440885377531e-05, + "loss": 1.0335, + "step": 8655 + }, + { + "epoch": 0.56, + "learning_rate": 1.816419416475236e-05, + "loss": 1.1587, + "step": 8656 + }, + { + "epoch": 0.56, + "learning_rate": 1.8163979475729407e-05, + "loss": 1.0843, + "step": 8657 + }, + { + "epoch": 0.56, + "learning_rate": 1.8163764786706456e-05, + "loss": 1.0152, + "step": 8658 + }, + { + "epoch": 0.56, + "learning_rate": 1.8163550097683505e-05, + "loss": 1.0975, + "step": 8659 + }, + { + "epoch": 0.56, + "learning_rate": 1.8163335408660558e-05, + "loss": 1.1035, + "step": 8660 + }, + { + "epoch": 0.56, + "learning_rate": 1.8163120719637607e-05, + "loss": 1.0443, + "step": 8661 + }, + { + "epoch": 0.56, + "learning_rate": 1.8162906030614656e-05, + "loss": 1.0816, + "step": 8662 + }, + { + "epoch": 0.56, + "learning_rate": 1.8162691341591705e-05, + "loss": 1.1154, + "step": 8663 + }, + { + "epoch": 0.56, + "learning_rate": 1.8162476652568754e-05, + "loss": 1.0134, + "step": 8664 + }, + { + "epoch": 0.56, + "learning_rate": 1.8162261963545806e-05, + "loss": 1.0901, + "step": 8665 + }, + { + "epoch": 0.56, + "learning_rate": 1.8162047274522855e-05, + "loss": 1.0616, + "step": 8666 + }, + { + "epoch": 0.56, + "learning_rate": 1.8161832585499907e-05, + "loss": 1.0995, + "step": 8667 + }, + { + "epoch": 0.56, + "learning_rate": 1.8161617896476956e-05, + "loss": 1.0388, + "step": 8668 + }, + { + "epoch": 0.56, + "learning_rate": 1.8161403207454005e-05, + "loss": 1.1693, + "step": 8669 + }, + { + "epoch": 0.56, + "learning_rate": 1.8161188518431054e-05, + "loss": 1.1382, + "step": 8670 + }, + { + "epoch": 0.56, + "learning_rate": 1.8160973829408103e-05, + "loss": 1.0509, + "step": 8671 + }, + { + "epoch": 0.56, + "learning_rate": 1.8160759140385152e-05, + "loss": 1.0419, + "step": 8672 + }, + { + "epoch": 0.56, + "learning_rate": 1.8160544451362204e-05, + "loss": 1.032, + "step": 8673 + }, + { + "epoch": 0.56, + "learning_rate": 1.8160329762339253e-05, + "loss": 1.046, + "step": 8674 + }, + { + "epoch": 0.56, + "learning_rate": 1.8160115073316302e-05, + "loss": 1.1254, + "step": 8675 + }, + { + "epoch": 0.56, + "learning_rate": 1.815990038429335e-05, + "loss": 0.9958, + "step": 8676 + }, + { + "epoch": 0.56, + "learning_rate": 1.81596856952704e-05, + "loss": 0.9646, + "step": 8677 + }, + { + "epoch": 0.56, + "learning_rate": 1.8159471006247453e-05, + "loss": 1.0879, + "step": 8678 + }, + { + "epoch": 0.56, + "learning_rate": 1.81592563172245e-05, + "loss": 1.1243, + "step": 8679 + }, + { + "epoch": 0.56, + "learning_rate": 1.8159041628201554e-05, + "loss": 1.1765, + "step": 8680 + }, + { + "epoch": 0.56, + "learning_rate": 1.8158826939178603e-05, + "loss": 1.02, + "step": 8681 + }, + { + "epoch": 0.56, + "learning_rate": 1.8158612250155652e-05, + "loss": 1.1099, + "step": 8682 + }, + { + "epoch": 0.56, + "learning_rate": 1.81583975611327e-05, + "loss": 1.0482, + "step": 8683 + }, + { + "epoch": 0.56, + "learning_rate": 1.815818287210975e-05, + "loss": 1.0194, + "step": 8684 + }, + { + "epoch": 0.56, + "learning_rate": 1.81579681830868e-05, + "loss": 1.1209, + "step": 8685 + }, + { + "epoch": 0.56, + "learning_rate": 1.815775349406385e-05, + "loss": 1.0718, + "step": 8686 + }, + { + "epoch": 0.56, + "learning_rate": 1.81575388050409e-05, + "loss": 1.2089, + "step": 8687 + }, + { + "epoch": 0.56, + "learning_rate": 1.815732411601795e-05, + "loss": 1.083, + "step": 8688 + }, + { + "epoch": 0.56, + "learning_rate": 1.8157109426994998e-05, + "loss": 1.1962, + "step": 8689 + }, + { + "epoch": 0.56, + "learning_rate": 1.8156894737972047e-05, + "loss": 1.067, + "step": 8690 + }, + { + "epoch": 0.56, + "learning_rate": 1.81566800489491e-05, + "loss": 1.014, + "step": 8691 + }, + { + "epoch": 0.56, + "learning_rate": 1.815646535992615e-05, + "loss": 1.1193, + "step": 8692 + }, + { + "epoch": 0.56, + "learning_rate": 1.8156250670903197e-05, + "loss": 1.0665, + "step": 8693 + }, + { + "epoch": 0.56, + "learning_rate": 1.815603598188025e-05, + "loss": 1.1078, + "step": 8694 + }, + { + "epoch": 0.56, + "learning_rate": 1.81558212928573e-05, + "loss": 1.0524, + "step": 8695 + }, + { + "epoch": 0.56, + "learning_rate": 1.8155606603834348e-05, + "loss": 1.1154, + "step": 8696 + }, + { + "epoch": 0.56, + "learning_rate": 1.8155391914811397e-05, + "loss": 1.0358, + "step": 8697 + }, + { + "epoch": 0.56, + "learning_rate": 1.815517722578845e-05, + "loss": 1.0695, + "step": 8698 + }, + { + "epoch": 0.56, + "learning_rate": 1.8154962536765498e-05, + "loss": 1.0392, + "step": 8699 + }, + { + "epoch": 0.56, + "learning_rate": 1.8154747847742547e-05, + "loss": 1.1617, + "step": 8700 + }, + { + "epoch": 0.56, + "learning_rate": 1.8154533158719596e-05, + "loss": 1.0962, + "step": 8701 + }, + { + "epoch": 0.56, + "learning_rate": 1.8154318469696645e-05, + "loss": 1.0838, + "step": 8702 + }, + { + "epoch": 0.56, + "learning_rate": 1.8154103780673694e-05, + "loss": 1.0657, + "step": 8703 + }, + { + "epoch": 0.56, + "learning_rate": 1.8153889091650746e-05, + "loss": 1.1168, + "step": 8704 + }, + { + "epoch": 0.56, + "learning_rate": 1.8153674402627795e-05, + "loss": 1.081, + "step": 8705 + }, + { + "epoch": 0.56, + "learning_rate": 1.8153459713604844e-05, + "loss": 1.1427, + "step": 8706 + }, + { + "epoch": 0.56, + "learning_rate": 1.8153245024581893e-05, + "loss": 1.0946, + "step": 8707 + }, + { + "epoch": 0.56, + "learning_rate": 1.8153030335558945e-05, + "loss": 1.2182, + "step": 8708 + }, + { + "epoch": 0.56, + "learning_rate": 1.8152815646535994e-05, + "loss": 1.0573, + "step": 8709 + }, + { + "epoch": 0.56, + "learning_rate": 1.8152600957513043e-05, + "loss": 1.1035, + "step": 8710 + }, + { + "epoch": 0.56, + "learning_rate": 1.8152386268490096e-05, + "loss": 0.9854, + "step": 8711 + }, + { + "epoch": 0.56, + "learning_rate": 1.8152171579467145e-05, + "loss": 1.084, + "step": 8712 + }, + { + "epoch": 0.56, + "learning_rate": 1.8151956890444194e-05, + "loss": 1.0061, + "step": 8713 + }, + { + "epoch": 0.56, + "learning_rate": 1.8151742201421242e-05, + "loss": 1.1584, + "step": 8714 + }, + { + "epoch": 0.56, + "learning_rate": 1.815152751239829e-05, + "loss": 1.0017, + "step": 8715 + }, + { + "epoch": 0.56, + "learning_rate": 1.8151312823375344e-05, + "loss": 1.0773, + "step": 8716 + }, + { + "epoch": 0.56, + "learning_rate": 1.8151098134352393e-05, + "loss": 1.0898, + "step": 8717 + }, + { + "epoch": 0.56, + "learning_rate": 1.8150883445329442e-05, + "loss": 1.1897, + "step": 8718 + }, + { + "epoch": 0.56, + "learning_rate": 1.815066875630649e-05, + "loss": 1.0589, + "step": 8719 + }, + { + "epoch": 0.56, + "learning_rate": 1.815045406728354e-05, + "loss": 1.0484, + "step": 8720 + }, + { + "epoch": 0.56, + "learning_rate": 1.815023937826059e-05, + "loss": 1.157, + "step": 8721 + }, + { + "epoch": 0.56, + "learning_rate": 1.815002468923764e-05, + "loss": 1.2178, + "step": 8722 + }, + { + "epoch": 0.56, + "learning_rate": 1.814981000021469e-05, + "loss": 1.1023, + "step": 8723 + }, + { + "epoch": 0.56, + "learning_rate": 1.8149595311191742e-05, + "loss": 1.0322, + "step": 8724 + }, + { + "epoch": 0.56, + "learning_rate": 1.814938062216879e-05, + "loss": 1.0138, + "step": 8725 + }, + { + "epoch": 0.56, + "learning_rate": 1.814916593314584e-05, + "loss": 1.1343, + "step": 8726 + }, + { + "epoch": 0.56, + "learning_rate": 1.814895124412289e-05, + "loss": 1.0106, + "step": 8727 + }, + { + "epoch": 0.56, + "learning_rate": 1.8148736555099938e-05, + "loss": 1.0768, + "step": 8728 + }, + { + "epoch": 0.56, + "learning_rate": 1.814852186607699e-05, + "loss": 1.1206, + "step": 8729 + }, + { + "epoch": 0.56, + "learning_rate": 1.814830717705404e-05, + "loss": 1.0993, + "step": 8730 + }, + { + "epoch": 0.56, + "learning_rate": 1.814809248803109e-05, + "loss": 1.0795, + "step": 8731 + }, + { + "epoch": 0.56, + "learning_rate": 1.8147877799008137e-05, + "loss": 1.1079, + "step": 8732 + }, + { + "epoch": 0.56, + "learning_rate": 1.8147663109985186e-05, + "loss": 1.0567, + "step": 8733 + }, + { + "epoch": 0.56, + "learning_rate": 1.8147448420962235e-05, + "loss": 1.2028, + "step": 8734 + }, + { + "epoch": 0.56, + "learning_rate": 1.8147233731939288e-05, + "loss": 1.086, + "step": 8735 + }, + { + "epoch": 0.56, + "learning_rate": 1.8147019042916337e-05, + "loss": 1.1697, + "step": 8736 + }, + { + "epoch": 0.56, + "learning_rate": 1.814680435389339e-05, + "loss": 1.1067, + "step": 8737 + }, + { + "epoch": 0.56, + "learning_rate": 1.8146589664870438e-05, + "loss": 1.1552, + "step": 8738 + }, + { + "epoch": 0.56, + "learning_rate": 1.8146374975847487e-05, + "loss": 1.0874, + "step": 8739 + }, + { + "epoch": 0.56, + "learning_rate": 1.8146160286824536e-05, + "loss": 1.0964, + "step": 8740 + }, + { + "epoch": 0.56, + "learning_rate": 1.8145945597801585e-05, + "loss": 1.1366, + "step": 8741 + }, + { + "epoch": 0.56, + "learning_rate": 1.8145730908778637e-05, + "loss": 1.1011, + "step": 8742 + }, + { + "epoch": 0.56, + "learning_rate": 1.8145516219755686e-05, + "loss": 1.0728, + "step": 8743 + }, + { + "epoch": 0.56, + "learning_rate": 1.8145301530732735e-05, + "loss": 0.9829, + "step": 8744 + }, + { + "epoch": 0.56, + "learning_rate": 1.8145086841709784e-05, + "loss": 1.0129, + "step": 8745 + }, + { + "epoch": 0.56, + "learning_rate": 1.8144872152686833e-05, + "loss": 1.113, + "step": 8746 + }, + { + "epoch": 0.56, + "learning_rate": 1.8144657463663885e-05, + "loss": 1.1487, + "step": 8747 + }, + { + "epoch": 0.56, + "learning_rate": 1.8144442774640934e-05, + "loss": 1.0474, + "step": 8748 + }, + { + "epoch": 0.56, + "learning_rate": 1.8144228085617983e-05, + "loss": 1.062, + "step": 8749 + }, + { + "epoch": 0.56, + "learning_rate": 1.8144013396595032e-05, + "loss": 1.1037, + "step": 8750 + }, + { + "epoch": 0.56, + "learning_rate": 1.814379870757208e-05, + "loss": 1.0802, + "step": 8751 + }, + { + "epoch": 0.56, + "learning_rate": 1.8143584018549134e-05, + "loss": 1.1417, + "step": 8752 + }, + { + "epoch": 0.56, + "learning_rate": 1.8143369329526183e-05, + "loss": 1.062, + "step": 8753 + }, + { + "epoch": 0.56, + "learning_rate": 1.8143154640503235e-05, + "loss": 1.1583, + "step": 8754 + }, + { + "epoch": 0.56, + "learning_rate": 1.8142939951480284e-05, + "loss": 1.1017, + "step": 8755 + }, + { + "epoch": 0.56, + "learning_rate": 1.8142725262457333e-05, + "loss": 1.0871, + "step": 8756 + }, + { + "epoch": 0.56, + "learning_rate": 1.8142510573434382e-05, + "loss": 1.1401, + "step": 8757 + }, + { + "epoch": 0.56, + "learning_rate": 1.814229588441143e-05, + "loss": 1.0957, + "step": 8758 + }, + { + "epoch": 0.56, + "learning_rate": 1.814208119538848e-05, + "loss": 1.021, + "step": 8759 + }, + { + "epoch": 0.56, + "learning_rate": 1.8141866506365532e-05, + "loss": 1.0305, + "step": 8760 + }, + { + "epoch": 0.56, + "learning_rate": 1.814165181734258e-05, + "loss": 1.0919, + "step": 8761 + }, + { + "epoch": 0.56, + "learning_rate": 1.814143712831963e-05, + "loss": 1.1632, + "step": 8762 + }, + { + "epoch": 0.56, + "learning_rate": 1.814122243929668e-05, + "loss": 1.0233, + "step": 8763 + }, + { + "epoch": 0.56, + "learning_rate": 1.8141007750273728e-05, + "loss": 1.1512, + "step": 8764 + }, + { + "epoch": 0.56, + "learning_rate": 1.814079306125078e-05, + "loss": 1.0322, + "step": 8765 + }, + { + "epoch": 0.56, + "learning_rate": 1.814057837222783e-05, + "loss": 1.021, + "step": 8766 + }, + { + "epoch": 0.56, + "learning_rate": 1.814036368320488e-05, + "loss": 1.1224, + "step": 8767 + }, + { + "epoch": 0.56, + "learning_rate": 1.814014899418193e-05, + "loss": 1.0736, + "step": 8768 + }, + { + "epoch": 0.56, + "learning_rate": 1.813993430515898e-05, + "loss": 1.1462, + "step": 8769 + }, + { + "epoch": 0.56, + "learning_rate": 1.813971961613603e-05, + "loss": 1.0855, + "step": 8770 + }, + { + "epoch": 0.56, + "learning_rate": 1.8139504927113078e-05, + "loss": 1.0008, + "step": 8771 + }, + { + "epoch": 0.56, + "learning_rate": 1.8139290238090126e-05, + "loss": 1.0224, + "step": 8772 + }, + { + "epoch": 0.56, + "learning_rate": 1.813907554906718e-05, + "loss": 1.0758, + "step": 8773 + }, + { + "epoch": 0.56, + "learning_rate": 1.8138860860044228e-05, + "loss": 0.9914, + "step": 8774 + }, + { + "epoch": 0.56, + "learning_rate": 1.8138646171021277e-05, + "loss": 1.1188, + "step": 8775 + }, + { + "epoch": 0.56, + "learning_rate": 1.8138431481998326e-05, + "loss": 1.1095, + "step": 8776 + }, + { + "epoch": 0.56, + "learning_rate": 1.8138216792975375e-05, + "loss": 1.0521, + "step": 8777 + }, + { + "epoch": 0.56, + "learning_rate": 1.8138002103952427e-05, + "loss": 1.0778, + "step": 8778 + }, + { + "epoch": 0.56, + "learning_rate": 1.8137787414929476e-05, + "loss": 1.1301, + "step": 8779 + }, + { + "epoch": 0.56, + "learning_rate": 1.8137572725906525e-05, + "loss": 1.094, + "step": 8780 + }, + { + "epoch": 0.56, + "learning_rate": 1.8137358036883577e-05, + "loss": 1.1704, + "step": 8781 + }, + { + "epoch": 0.57, + "learning_rate": 1.8137143347860626e-05, + "loss": 1.0574, + "step": 8782 + }, + { + "epoch": 0.57, + "learning_rate": 1.8136928658837675e-05, + "loss": 0.9956, + "step": 8783 + }, + { + "epoch": 0.57, + "learning_rate": 1.8136713969814724e-05, + "loss": 1.0096, + "step": 8784 + }, + { + "epoch": 0.57, + "learning_rate": 1.8136499280791777e-05, + "loss": 1.1302, + "step": 8785 + }, + { + "epoch": 0.57, + "learning_rate": 1.8136284591768826e-05, + "loss": 1.1304, + "step": 8786 + }, + { + "epoch": 0.57, + "learning_rate": 1.8136069902745875e-05, + "loss": 1.0764, + "step": 8787 + }, + { + "epoch": 0.57, + "learning_rate": 1.8135855213722923e-05, + "loss": 1.0984, + "step": 8788 + }, + { + "epoch": 0.57, + "learning_rate": 1.8135640524699972e-05, + "loss": 1.0695, + "step": 8789 + }, + { + "epoch": 0.57, + "learning_rate": 1.813542583567702e-05, + "loss": 1.0828, + "step": 8790 + }, + { + "epoch": 0.57, + "learning_rate": 1.8135211146654074e-05, + "loss": 1.0141, + "step": 8791 + }, + { + "epoch": 0.57, + "learning_rate": 1.8134996457631123e-05, + "loss": 1.1515, + "step": 8792 + }, + { + "epoch": 0.57, + "learning_rate": 1.8134781768608172e-05, + "loss": 1.1199, + "step": 8793 + }, + { + "epoch": 0.57, + "learning_rate": 1.813456707958522e-05, + "loss": 1.0752, + "step": 8794 + }, + { + "epoch": 0.57, + "learning_rate": 1.813435239056227e-05, + "loss": 1.0414, + "step": 8795 + }, + { + "epoch": 0.57, + "learning_rate": 1.8134137701539322e-05, + "loss": 1.0805, + "step": 8796 + }, + { + "epoch": 0.57, + "learning_rate": 1.813392301251637e-05, + "loss": 0.969, + "step": 8797 + }, + { + "epoch": 0.57, + "learning_rate": 1.8133708323493423e-05, + "loss": 1.0339, + "step": 8798 + }, + { + "epoch": 0.57, + "learning_rate": 1.8133493634470472e-05, + "loss": 1.0277, + "step": 8799 + }, + { + "epoch": 0.57, + "learning_rate": 1.813327894544752e-05, + "loss": 1.0119, + "step": 8800 + }, + { + "epoch": 0.57, + "learning_rate": 1.813306425642457e-05, + "loss": 1.0916, + "step": 8801 + }, + { + "epoch": 0.57, + "learning_rate": 1.813284956740162e-05, + "loss": 1.0701, + "step": 8802 + }, + { + "epoch": 0.57, + "learning_rate": 1.813263487837867e-05, + "loss": 1.0886, + "step": 8803 + }, + { + "epoch": 0.57, + "learning_rate": 1.813242018935572e-05, + "loss": 0.9969, + "step": 8804 + }, + { + "epoch": 0.57, + "learning_rate": 1.813220550033277e-05, + "loss": 1.1232, + "step": 8805 + }, + { + "epoch": 0.57, + "learning_rate": 1.813199081130982e-05, + "loss": 1.1054, + "step": 8806 + }, + { + "epoch": 0.57, + "learning_rate": 1.8131776122286867e-05, + "loss": 1.1396, + "step": 8807 + }, + { + "epoch": 0.57, + "learning_rate": 1.8131561433263916e-05, + "loss": 1.0229, + "step": 8808 + }, + { + "epoch": 0.57, + "learning_rate": 1.813134674424097e-05, + "loss": 1.0577, + "step": 8809 + }, + { + "epoch": 0.57, + "learning_rate": 1.8131132055218018e-05, + "loss": 1.1393, + "step": 8810 + }, + { + "epoch": 0.57, + "learning_rate": 1.813091736619507e-05, + "loss": 1.0556, + "step": 8811 + }, + { + "epoch": 0.57, + "learning_rate": 1.813070267717212e-05, + "loss": 1.0885, + "step": 8812 + }, + { + "epoch": 0.57, + "learning_rate": 1.8130487988149168e-05, + "loss": 1.1149, + "step": 8813 + }, + { + "epoch": 0.57, + "learning_rate": 1.8130273299126217e-05, + "loss": 1.0143, + "step": 8814 + }, + { + "epoch": 0.57, + "learning_rate": 1.8130058610103266e-05, + "loss": 0.994, + "step": 8815 + }, + { + "epoch": 0.57, + "learning_rate": 1.8129843921080318e-05, + "loss": 1.1534, + "step": 8816 + }, + { + "epoch": 0.57, + "learning_rate": 1.8129629232057367e-05, + "loss": 1.092, + "step": 8817 + }, + { + "epoch": 0.57, + "learning_rate": 1.8129414543034416e-05, + "loss": 1.11, + "step": 8818 + }, + { + "epoch": 0.57, + "learning_rate": 1.8129199854011465e-05, + "loss": 1.0792, + "step": 8819 + }, + { + "epoch": 0.57, + "learning_rate": 1.8128985164988514e-05, + "loss": 1.0406, + "step": 8820 + }, + { + "epoch": 0.57, + "learning_rate": 1.8128770475965563e-05, + "loss": 1.0314, + "step": 8821 + }, + { + "epoch": 0.57, + "learning_rate": 1.8128555786942615e-05, + "loss": 1.1269, + "step": 8822 + }, + { + "epoch": 0.57, + "learning_rate": 1.8128341097919664e-05, + "loss": 1.0469, + "step": 8823 + }, + { + "epoch": 0.57, + "learning_rate": 1.8128126408896713e-05, + "loss": 1.1183, + "step": 8824 + }, + { + "epoch": 0.57, + "learning_rate": 1.8127911719873766e-05, + "loss": 1.0767, + "step": 8825 + }, + { + "epoch": 0.57, + "learning_rate": 1.8127697030850815e-05, + "loss": 1.0796, + "step": 8826 + }, + { + "epoch": 0.57, + "learning_rate": 1.8127482341827864e-05, + "loss": 0.9855, + "step": 8827 + }, + { + "epoch": 0.57, + "learning_rate": 1.8127267652804913e-05, + "loss": 1.0807, + "step": 8828 + }, + { + "epoch": 0.57, + "learning_rate": 1.8127052963781965e-05, + "loss": 1.0907, + "step": 8829 + }, + { + "epoch": 0.57, + "learning_rate": 1.8126838274759014e-05, + "loss": 1.0734, + "step": 8830 + }, + { + "epoch": 0.57, + "learning_rate": 1.8126623585736063e-05, + "loss": 1.0762, + "step": 8831 + }, + { + "epoch": 0.57, + "learning_rate": 1.8126408896713112e-05, + "loss": 1.0857, + "step": 8832 + }, + { + "epoch": 0.57, + "learning_rate": 1.812619420769016e-05, + "loss": 1.0484, + "step": 8833 + }, + { + "epoch": 0.57, + "learning_rate": 1.8125979518667213e-05, + "loss": 1.1048, + "step": 8834 + }, + { + "epoch": 0.57, + "learning_rate": 1.8125764829644262e-05, + "loss": 1.0905, + "step": 8835 + }, + { + "epoch": 0.57, + "learning_rate": 1.812555014062131e-05, + "loss": 1.0473, + "step": 8836 + }, + { + "epoch": 0.57, + "learning_rate": 1.812533545159836e-05, + "loss": 1.052, + "step": 8837 + }, + { + "epoch": 0.57, + "learning_rate": 1.812512076257541e-05, + "loss": 1.1002, + "step": 8838 + }, + { + "epoch": 0.57, + "learning_rate": 1.812490607355246e-05, + "loss": 1.0572, + "step": 8839 + }, + { + "epoch": 0.57, + "learning_rate": 1.812469138452951e-05, + "loss": 1.0813, + "step": 8840 + }, + { + "epoch": 0.57, + "learning_rate": 1.812447669550656e-05, + "loss": 0.9966, + "step": 8841 + }, + { + "epoch": 0.57, + "learning_rate": 1.812426200648361e-05, + "loss": 1.2097, + "step": 8842 + }, + { + "epoch": 0.57, + "learning_rate": 1.812404731746066e-05, + "loss": 1.14, + "step": 8843 + }, + { + "epoch": 0.57, + "learning_rate": 1.812383262843771e-05, + "loss": 1.1594, + "step": 8844 + }, + { + "epoch": 0.57, + "learning_rate": 1.812361793941476e-05, + "loss": 1.0754, + "step": 8845 + }, + { + "epoch": 0.57, + "learning_rate": 1.8123403250391808e-05, + "loss": 1.0814, + "step": 8846 + }, + { + "epoch": 0.57, + "learning_rate": 1.812318856136886e-05, + "loss": 1.0481, + "step": 8847 + }, + { + "epoch": 0.57, + "learning_rate": 1.812297387234591e-05, + "loss": 1.0418, + "step": 8848 + }, + { + "epoch": 0.57, + "learning_rate": 1.8122759183322958e-05, + "loss": 1.043, + "step": 8849 + }, + { + "epoch": 0.57, + "learning_rate": 1.8122544494300007e-05, + "loss": 1.1211, + "step": 8850 + }, + { + "epoch": 0.57, + "learning_rate": 1.8122329805277056e-05, + "loss": 1.1124, + "step": 8851 + }, + { + "epoch": 0.57, + "learning_rate": 1.8122115116254108e-05, + "loss": 1.0972, + "step": 8852 + }, + { + "epoch": 0.57, + "learning_rate": 1.8121900427231157e-05, + "loss": 1.0592, + "step": 8853 + }, + { + "epoch": 0.57, + "learning_rate": 1.812168573820821e-05, + "loss": 1.1307, + "step": 8854 + }, + { + "epoch": 0.57, + "learning_rate": 1.812147104918526e-05, + "loss": 1.0388, + "step": 8855 + }, + { + "epoch": 0.57, + "learning_rate": 1.8121256360162307e-05, + "loss": 0.9704, + "step": 8856 + }, + { + "epoch": 0.57, + "learning_rate": 1.8121041671139356e-05, + "loss": 1.0633, + "step": 8857 + }, + { + "epoch": 0.57, + "learning_rate": 1.8120826982116405e-05, + "loss": 1.0974, + "step": 8858 + }, + { + "epoch": 0.57, + "learning_rate": 1.8120612293093454e-05, + "loss": 1.137, + "step": 8859 + }, + { + "epoch": 0.57, + "learning_rate": 1.8120397604070507e-05, + "loss": 1.0329, + "step": 8860 + }, + { + "epoch": 0.57, + "learning_rate": 1.8120182915047556e-05, + "loss": 1.0998, + "step": 8861 + }, + { + "epoch": 0.57, + "learning_rate": 1.8119968226024604e-05, + "loss": 1.0538, + "step": 8862 + }, + { + "epoch": 0.57, + "learning_rate": 1.8119753537001653e-05, + "loss": 1.0645, + "step": 8863 + }, + { + "epoch": 0.57, + "learning_rate": 1.8119538847978702e-05, + "loss": 1.1516, + "step": 8864 + }, + { + "epoch": 0.57, + "learning_rate": 1.8119324158955755e-05, + "loss": 1.1342, + "step": 8865 + }, + { + "epoch": 0.57, + "learning_rate": 1.8119109469932804e-05, + "loss": 1.0643, + "step": 8866 + }, + { + "epoch": 0.57, + "learning_rate": 1.8118894780909853e-05, + "loss": 1.0013, + "step": 8867 + }, + { + "epoch": 0.57, + "learning_rate": 1.8118680091886905e-05, + "loss": 1.0428, + "step": 8868 + }, + { + "epoch": 0.57, + "learning_rate": 1.8118465402863954e-05, + "loss": 1.056, + "step": 8869 + }, + { + "epoch": 0.57, + "learning_rate": 1.8118250713841003e-05, + "loss": 0.9788, + "step": 8870 + }, + { + "epoch": 0.57, + "learning_rate": 1.8118036024818052e-05, + "loss": 1.1695, + "step": 8871 + }, + { + "epoch": 0.57, + "learning_rate": 1.8117821335795104e-05, + "loss": 1.1305, + "step": 8872 + }, + { + "epoch": 0.57, + "learning_rate": 1.8117606646772153e-05, + "loss": 1.0938, + "step": 8873 + }, + { + "epoch": 0.57, + "learning_rate": 1.8117391957749202e-05, + "loss": 1.025, + "step": 8874 + }, + { + "epoch": 0.57, + "learning_rate": 1.811717726872625e-05, + "loss": 1.1611, + "step": 8875 + }, + { + "epoch": 0.57, + "learning_rate": 1.81169625797033e-05, + "loss": 1.0891, + "step": 8876 + }, + { + "epoch": 0.57, + "learning_rate": 1.811674789068035e-05, + "loss": 1.0752, + "step": 8877 + }, + { + "epoch": 0.57, + "learning_rate": 1.81165332016574e-05, + "loss": 1.135, + "step": 8878 + }, + { + "epoch": 0.57, + "learning_rate": 1.811631851263445e-05, + "loss": 1.0614, + "step": 8879 + }, + { + "epoch": 0.57, + "learning_rate": 1.81161038236115e-05, + "loss": 1.059, + "step": 8880 + }, + { + "epoch": 0.57, + "learning_rate": 1.811588913458855e-05, + "loss": 1.2251, + "step": 8881 + }, + { + "epoch": 0.57, + "learning_rate": 1.8115674445565597e-05, + "loss": 1.0944, + "step": 8882 + }, + { + "epoch": 0.57, + "learning_rate": 1.811545975654265e-05, + "loss": 1.1087, + "step": 8883 + }, + { + "epoch": 0.57, + "learning_rate": 1.81152450675197e-05, + "loss": 1.038, + "step": 8884 + }, + { + "epoch": 0.57, + "learning_rate": 1.811503037849675e-05, + "loss": 1.1205, + "step": 8885 + }, + { + "epoch": 0.57, + "learning_rate": 1.81148156894738e-05, + "loss": 1.1076, + "step": 8886 + }, + { + "epoch": 0.57, + "learning_rate": 1.811460100045085e-05, + "loss": 1.0103, + "step": 8887 + }, + { + "epoch": 0.57, + "learning_rate": 1.8114386311427898e-05, + "loss": 1.0724, + "step": 8888 + }, + { + "epoch": 0.57, + "learning_rate": 1.8114171622404947e-05, + "loss": 1.1312, + "step": 8889 + }, + { + "epoch": 0.57, + "learning_rate": 1.8113956933382e-05, + "loss": 1.0475, + "step": 8890 + }, + { + "epoch": 0.57, + "learning_rate": 1.8113742244359048e-05, + "loss": 1.0327, + "step": 8891 + }, + { + "epoch": 0.57, + "learning_rate": 1.8113527555336097e-05, + "loss": 1.0366, + "step": 8892 + }, + { + "epoch": 0.57, + "learning_rate": 1.8113312866313146e-05, + "loss": 1.1177, + "step": 8893 + }, + { + "epoch": 0.57, + "learning_rate": 1.8113098177290195e-05, + "loss": 1.1169, + "step": 8894 + }, + { + "epoch": 0.57, + "learning_rate": 1.8112883488267244e-05, + "loss": 1.0563, + "step": 8895 + }, + { + "epoch": 0.57, + "learning_rate": 1.8112668799244296e-05, + "loss": 1.0366, + "step": 8896 + }, + { + "epoch": 0.57, + "learning_rate": 1.8112454110221345e-05, + "loss": 1.0889, + "step": 8897 + }, + { + "epoch": 0.57, + "learning_rate": 1.8112239421198398e-05, + "loss": 1.055, + "step": 8898 + }, + { + "epoch": 0.57, + "learning_rate": 1.8112024732175447e-05, + "loss": 1.0055, + "step": 8899 + }, + { + "epoch": 0.57, + "learning_rate": 1.8111810043152496e-05, + "loss": 1.1368, + "step": 8900 + }, + { + "epoch": 0.57, + "learning_rate": 1.8111595354129545e-05, + "loss": 1.1733, + "step": 8901 + }, + { + "epoch": 0.57, + "learning_rate": 1.8111380665106594e-05, + "loss": 1.0671, + "step": 8902 + }, + { + "epoch": 0.57, + "learning_rate": 1.8111165976083646e-05, + "loss": 1.0968, + "step": 8903 + }, + { + "epoch": 0.57, + "learning_rate": 1.8110951287060695e-05, + "loss": 1.0897, + "step": 8904 + }, + { + "epoch": 0.57, + "learning_rate": 1.8110736598037744e-05, + "loss": 1.1686, + "step": 8905 + }, + { + "epoch": 0.57, + "learning_rate": 1.8110521909014793e-05, + "loss": 1.112, + "step": 8906 + }, + { + "epoch": 0.57, + "learning_rate": 1.8110307219991842e-05, + "loss": 1.0357, + "step": 8907 + }, + { + "epoch": 0.57, + "learning_rate": 1.811009253096889e-05, + "loss": 1.15, + "step": 8908 + }, + { + "epoch": 0.57, + "learning_rate": 1.8109877841945943e-05, + "loss": 1.0186, + "step": 8909 + }, + { + "epoch": 0.57, + "learning_rate": 1.8109663152922992e-05, + "loss": 1.0979, + "step": 8910 + }, + { + "epoch": 0.57, + "learning_rate": 1.810944846390004e-05, + "loss": 0.9946, + "step": 8911 + }, + { + "epoch": 0.57, + "learning_rate": 1.8109233774877093e-05, + "loss": 0.9948, + "step": 8912 + }, + { + "epoch": 0.57, + "learning_rate": 1.8109019085854142e-05, + "loss": 1.0136, + "step": 8913 + }, + { + "epoch": 0.57, + "learning_rate": 1.810880439683119e-05, + "loss": 1.0221, + "step": 8914 + }, + { + "epoch": 0.57, + "learning_rate": 1.810858970780824e-05, + "loss": 1.0783, + "step": 8915 + }, + { + "epoch": 0.57, + "learning_rate": 1.8108375018785293e-05, + "loss": 1.156, + "step": 8916 + }, + { + "epoch": 0.57, + "learning_rate": 1.810816032976234e-05, + "loss": 1.1126, + "step": 8917 + }, + { + "epoch": 0.57, + "learning_rate": 1.810794564073939e-05, + "loss": 1.1142, + "step": 8918 + }, + { + "epoch": 0.57, + "learning_rate": 1.810773095171644e-05, + "loss": 1.0487, + "step": 8919 + }, + { + "epoch": 0.57, + "learning_rate": 1.810751626269349e-05, + "loss": 1.1163, + "step": 8920 + }, + { + "epoch": 0.57, + "learning_rate": 1.810730157367054e-05, + "loss": 1.0754, + "step": 8921 + }, + { + "epoch": 0.57, + "learning_rate": 1.810708688464759e-05, + "loss": 0.9858, + "step": 8922 + }, + { + "epoch": 0.57, + "learning_rate": 1.810687219562464e-05, + "loss": 1.061, + "step": 8923 + }, + { + "epoch": 0.57, + "learning_rate": 1.8106657506601688e-05, + "loss": 1.0396, + "step": 8924 + }, + { + "epoch": 0.57, + "learning_rate": 1.8106442817578737e-05, + "loss": 1.0114, + "step": 8925 + }, + { + "epoch": 0.57, + "learning_rate": 1.810622812855579e-05, + "loss": 1.0624, + "step": 8926 + }, + { + "epoch": 0.57, + "learning_rate": 1.8106013439532838e-05, + "loss": 1.0906, + "step": 8927 + }, + { + "epoch": 0.57, + "learning_rate": 1.8105798750509887e-05, + "loss": 1.15, + "step": 8928 + }, + { + "epoch": 0.57, + "learning_rate": 1.810558406148694e-05, + "loss": 1.0957, + "step": 8929 + }, + { + "epoch": 0.57, + "learning_rate": 1.810536937246399e-05, + "loss": 1.0986, + "step": 8930 + }, + { + "epoch": 0.57, + "learning_rate": 1.8105154683441037e-05, + "loss": 1.149, + "step": 8931 + }, + { + "epoch": 0.57, + "learning_rate": 1.8104939994418086e-05, + "loss": 1.0815, + "step": 8932 + }, + { + "epoch": 0.57, + "learning_rate": 1.8104725305395135e-05, + "loss": 1.0392, + "step": 8933 + }, + { + "epoch": 0.57, + "learning_rate": 1.8104510616372188e-05, + "loss": 1.1174, + "step": 8934 + }, + { + "epoch": 0.57, + "learning_rate": 1.8104295927349237e-05, + "loss": 1.0218, + "step": 8935 + }, + { + "epoch": 0.57, + "learning_rate": 1.8104081238326285e-05, + "loss": 1.1225, + "step": 8936 + }, + { + "epoch": 0.57, + "learning_rate": 1.8103866549303334e-05, + "loss": 1.07, + "step": 8937 + }, + { + "epoch": 0.58, + "learning_rate": 1.8103651860280383e-05, + "loss": 1.1138, + "step": 8938 + }, + { + "epoch": 0.58, + "learning_rate": 1.8103437171257436e-05, + "loss": 1.1054, + "step": 8939 + }, + { + "epoch": 0.58, + "learning_rate": 1.8103222482234485e-05, + "loss": 1.1668, + "step": 8940 + }, + { + "epoch": 0.58, + "learning_rate": 1.8103007793211537e-05, + "loss": 1.0989, + "step": 8941 + }, + { + "epoch": 0.58, + "learning_rate": 1.8102793104188586e-05, + "loss": 1.1255, + "step": 8942 + }, + { + "epoch": 0.58, + "learning_rate": 1.8102578415165635e-05, + "loss": 1.0648, + "step": 8943 + }, + { + "epoch": 0.58, + "learning_rate": 1.8102363726142684e-05, + "loss": 1.1814, + "step": 8944 + }, + { + "epoch": 0.58, + "learning_rate": 1.8102149037119733e-05, + "loss": 1.0473, + "step": 8945 + }, + { + "epoch": 0.58, + "learning_rate": 1.8101934348096782e-05, + "loss": 1.0559, + "step": 8946 + }, + { + "epoch": 0.58, + "learning_rate": 1.8101719659073834e-05, + "loss": 1.0294, + "step": 8947 + }, + { + "epoch": 0.58, + "learning_rate": 1.8101504970050883e-05, + "loss": 1.0121, + "step": 8948 + }, + { + "epoch": 0.58, + "learning_rate": 1.8101290281027932e-05, + "loss": 1.04, + "step": 8949 + }, + { + "epoch": 0.58, + "learning_rate": 1.810107559200498e-05, + "loss": 1.0244, + "step": 8950 + }, + { + "epoch": 0.58, + "learning_rate": 1.810086090298203e-05, + "loss": 1.1082, + "step": 8951 + }, + { + "epoch": 0.58, + "learning_rate": 1.8100646213959082e-05, + "loss": 1.0827, + "step": 8952 + }, + { + "epoch": 0.58, + "learning_rate": 1.810043152493613e-05, + "loss": 1.0965, + "step": 8953 + }, + { + "epoch": 0.58, + "learning_rate": 1.810021683591318e-05, + "loss": 1.1606, + "step": 8954 + }, + { + "epoch": 0.58, + "learning_rate": 1.8100002146890233e-05, + "loss": 1.1117, + "step": 8955 + }, + { + "epoch": 0.58, + "learning_rate": 1.8099787457867282e-05, + "loss": 1.0719, + "step": 8956 + }, + { + "epoch": 0.58, + "learning_rate": 1.809957276884433e-05, + "loss": 1.0479, + "step": 8957 + }, + { + "epoch": 0.58, + "learning_rate": 1.809935807982138e-05, + "loss": 1.1105, + "step": 8958 + }, + { + "epoch": 0.58, + "learning_rate": 1.8099143390798432e-05, + "loss": 1.054, + "step": 8959 + }, + { + "epoch": 0.58, + "learning_rate": 1.809892870177548e-05, + "loss": 1.1379, + "step": 8960 + }, + { + "epoch": 0.58, + "learning_rate": 1.809871401275253e-05, + "loss": 1.11, + "step": 8961 + }, + { + "epoch": 0.58, + "learning_rate": 1.809849932372958e-05, + "loss": 1.126, + "step": 8962 + }, + { + "epoch": 0.58, + "learning_rate": 1.8098284634706628e-05, + "loss": 1.0591, + "step": 8963 + }, + { + "epoch": 0.58, + "learning_rate": 1.8098069945683677e-05, + "loss": 1.0692, + "step": 8964 + }, + { + "epoch": 0.58, + "learning_rate": 1.809785525666073e-05, + "loss": 1.1775, + "step": 8965 + }, + { + "epoch": 0.58, + "learning_rate": 1.8097640567637778e-05, + "loss": 1.0116, + "step": 8966 + }, + { + "epoch": 0.58, + "learning_rate": 1.8097425878614827e-05, + "loss": 1.0708, + "step": 8967 + }, + { + "epoch": 0.58, + "learning_rate": 1.8097211189591876e-05, + "loss": 1.1223, + "step": 8968 + }, + { + "epoch": 0.58, + "learning_rate": 1.8096996500568925e-05, + "loss": 1.103, + "step": 8969 + }, + { + "epoch": 0.58, + "learning_rate": 1.8096781811545977e-05, + "loss": 1.0997, + "step": 8970 + }, + { + "epoch": 0.58, + "learning_rate": 1.8096567122523026e-05, + "loss": 1.0023, + "step": 8971 + }, + { + "epoch": 0.58, + "learning_rate": 1.809635243350008e-05, + "loss": 1.0249, + "step": 8972 + }, + { + "epoch": 0.58, + "learning_rate": 1.8096137744477128e-05, + "loss": 1.0886, + "step": 8973 + }, + { + "epoch": 0.58, + "learning_rate": 1.8095923055454177e-05, + "loss": 1.136, + "step": 8974 + }, + { + "epoch": 0.58, + "learning_rate": 1.8095708366431226e-05, + "loss": 1.0638, + "step": 8975 + }, + { + "epoch": 0.58, + "learning_rate": 1.8095493677408275e-05, + "loss": 1.0464, + "step": 8976 + }, + { + "epoch": 0.58, + "learning_rate": 1.8095278988385327e-05, + "loss": 1.063, + "step": 8977 + }, + { + "epoch": 0.58, + "learning_rate": 1.8095064299362376e-05, + "loss": 1.0398, + "step": 8978 + }, + { + "epoch": 0.58, + "learning_rate": 1.8094849610339425e-05, + "loss": 1.0777, + "step": 8979 + }, + { + "epoch": 0.58, + "learning_rate": 1.8094634921316474e-05, + "loss": 1.0672, + "step": 8980 + }, + { + "epoch": 0.58, + "learning_rate": 1.8094420232293523e-05, + "loss": 1.058, + "step": 8981 + }, + { + "epoch": 0.58, + "learning_rate": 1.8094205543270572e-05, + "loss": 1.0745, + "step": 8982 + }, + { + "epoch": 0.58, + "learning_rate": 1.8093990854247624e-05, + "loss": 1.0026, + "step": 8983 + }, + { + "epoch": 0.58, + "learning_rate": 1.8093776165224673e-05, + "loss": 1.16, + "step": 8984 + }, + { + "epoch": 0.58, + "learning_rate": 1.8093561476201725e-05, + "loss": 1.1128, + "step": 8985 + }, + { + "epoch": 0.58, + "learning_rate": 1.8093346787178774e-05, + "loss": 1.0402, + "step": 8986 + }, + { + "epoch": 0.58, + "learning_rate": 1.8093132098155823e-05, + "loss": 1.0219, + "step": 8987 + }, + { + "epoch": 0.58, + "learning_rate": 1.8092917409132872e-05, + "loss": 1.11, + "step": 8988 + }, + { + "epoch": 0.58, + "learning_rate": 1.809270272010992e-05, + "loss": 1.1073, + "step": 8989 + }, + { + "epoch": 0.58, + "learning_rate": 1.8092488031086974e-05, + "loss": 1.109, + "step": 8990 + }, + { + "epoch": 0.58, + "learning_rate": 1.8092273342064023e-05, + "loss": 1.0826, + "step": 8991 + }, + { + "epoch": 0.58, + "learning_rate": 1.809205865304107e-05, + "loss": 1.1733, + "step": 8992 + }, + { + "epoch": 0.58, + "learning_rate": 1.809184396401812e-05, + "loss": 1.0204, + "step": 8993 + }, + { + "epoch": 0.58, + "learning_rate": 1.809162927499517e-05, + "loss": 1.0787, + "step": 8994 + }, + { + "epoch": 0.58, + "learning_rate": 1.809141458597222e-05, + "loss": 1.0482, + "step": 8995 + }, + { + "epoch": 0.58, + "learning_rate": 1.809119989694927e-05, + "loss": 1.0918, + "step": 8996 + }, + { + "epoch": 0.58, + "learning_rate": 1.809098520792632e-05, + "loss": 1.0638, + "step": 8997 + }, + { + "epoch": 0.58, + "learning_rate": 1.809077051890337e-05, + "loss": 1.131, + "step": 8998 + }, + { + "epoch": 0.58, + "learning_rate": 1.809055582988042e-05, + "loss": 1.086, + "step": 8999 + }, + { + "epoch": 0.58, + "learning_rate": 1.809034114085747e-05, + "loss": 1.1318, + "step": 9000 + }, + { + "epoch": 0.58, + "eval_laion/cider": 0.5028093212260166, + "eval_laion/meteor": 0.38587967342969565, + "eval_laion/pred_num_words": 35.810546875, + "eval_laion/sacrebleu": 15.264527164319304, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0959758758544922, + "eval_runtime": 103.2448, + "eval_samples_per_second": 4.833, + "eval_steps_per_second": 0.31, + "step": 9000 + }, + { + "epoch": 0.58, + "learning_rate": 1.809012645183452e-05, + "loss": 1.0412, + "step": 9001 + }, + { + "epoch": 0.58, + "learning_rate": 1.8089911762811568e-05, + "loss": 1.1204, + "step": 9002 + }, + { + "epoch": 0.58, + "learning_rate": 1.808969707378862e-05, + "loss": 1.1334, + "step": 9003 + }, + { + "epoch": 0.58, + "learning_rate": 1.808948238476567e-05, + "loss": 1.1291, + "step": 9004 + }, + { + "epoch": 0.58, + "learning_rate": 1.8089267695742718e-05, + "loss": 1.0999, + "step": 9005 + }, + { + "epoch": 0.58, + "learning_rate": 1.8089053006719767e-05, + "loss": 1.0748, + "step": 9006 + }, + { + "epoch": 0.58, + "learning_rate": 1.8088838317696816e-05, + "loss": 1.0661, + "step": 9007 + }, + { + "epoch": 0.58, + "learning_rate": 1.808862362867387e-05, + "loss": 1.0445, + "step": 9008 + }, + { + "epoch": 0.58, + "learning_rate": 1.8088408939650918e-05, + "loss": 1.144, + "step": 9009 + }, + { + "epoch": 0.58, + "learning_rate": 1.8088194250627966e-05, + "loss": 1.1166, + "step": 9010 + }, + { + "epoch": 0.58, + "learning_rate": 1.8087979561605015e-05, + "loss": 1.0413, + "step": 9011 + }, + { + "epoch": 0.58, + "learning_rate": 1.8087764872582064e-05, + "loss": 1.0127, + "step": 9012 + }, + { + "epoch": 0.58, + "learning_rate": 1.8087550183559117e-05, + "loss": 1.063, + "step": 9013 + }, + { + "epoch": 0.58, + "learning_rate": 1.8087335494536166e-05, + "loss": 1.0995, + "step": 9014 + }, + { + "epoch": 0.58, + "learning_rate": 1.8087120805513215e-05, + "loss": 1.0025, + "step": 9015 + }, + { + "epoch": 0.58, + "learning_rate": 1.8086906116490267e-05, + "loss": 1.1316, + "step": 9016 + }, + { + "epoch": 0.58, + "learning_rate": 1.8086691427467316e-05, + "loss": 1.1506, + "step": 9017 + }, + { + "epoch": 0.58, + "learning_rate": 1.8086476738444365e-05, + "loss": 1.0157, + "step": 9018 + }, + { + "epoch": 0.58, + "learning_rate": 1.8086262049421414e-05, + "loss": 1.0659, + "step": 9019 + }, + { + "epoch": 0.58, + "learning_rate": 1.8086047360398463e-05, + "loss": 1.0928, + "step": 9020 + }, + { + "epoch": 0.58, + "learning_rate": 1.8085832671375515e-05, + "loss": 1.158, + "step": 9021 + }, + { + "epoch": 0.58, + "learning_rate": 1.8085617982352564e-05, + "loss": 0.9964, + "step": 9022 + }, + { + "epoch": 0.58, + "learning_rate": 1.8085403293329613e-05, + "loss": 1.1142, + "step": 9023 + }, + { + "epoch": 0.58, + "learning_rate": 1.8085188604306662e-05, + "loss": 1.0435, + "step": 9024 + }, + { + "epoch": 0.58, + "learning_rate": 1.808497391528371e-05, + "loss": 1.1058, + "step": 9025 + }, + { + "epoch": 0.58, + "learning_rate": 1.8084759226260763e-05, + "loss": 1.05, + "step": 9026 + }, + { + "epoch": 0.58, + "learning_rate": 1.8084544537237812e-05, + "loss": 1.0987, + "step": 9027 + }, + { + "epoch": 0.58, + "learning_rate": 1.8084329848214865e-05, + "loss": 1.1289, + "step": 9028 + }, + { + "epoch": 0.58, + "learning_rate": 1.8084115159191914e-05, + "loss": 0.9928, + "step": 9029 + }, + { + "epoch": 0.58, + "learning_rate": 1.8083900470168963e-05, + "loss": 1.1118, + "step": 9030 + }, + { + "epoch": 0.58, + "learning_rate": 1.8083685781146012e-05, + "loss": 1.0082, + "step": 9031 + }, + { + "epoch": 0.58, + "learning_rate": 1.808347109212306e-05, + "loss": 1.0486, + "step": 9032 + }, + { + "epoch": 0.58, + "learning_rate": 1.808325640310011e-05, + "loss": 1.1082, + "step": 9033 + }, + { + "epoch": 0.58, + "learning_rate": 1.8083041714077162e-05, + "loss": 1.1515, + "step": 9034 + }, + { + "epoch": 0.58, + "learning_rate": 1.808282702505421e-05, + "loss": 1.091, + "step": 9035 + }, + { + "epoch": 0.58, + "learning_rate": 1.808261233603126e-05, + "loss": 1.0791, + "step": 9036 + }, + { + "epoch": 0.58, + "learning_rate": 1.808239764700831e-05, + "loss": 1.0732, + "step": 9037 + }, + { + "epoch": 0.58, + "learning_rate": 1.8082182957985358e-05, + "loss": 1.0522, + "step": 9038 + }, + { + "epoch": 0.58, + "learning_rate": 1.808196826896241e-05, + "loss": 1.0993, + "step": 9039 + }, + { + "epoch": 0.58, + "learning_rate": 1.808175357993946e-05, + "loss": 1.121, + "step": 9040 + }, + { + "epoch": 0.58, + "learning_rate": 1.8081538890916508e-05, + "loss": 1.0587, + "step": 9041 + }, + { + "epoch": 0.58, + "learning_rate": 1.808132420189356e-05, + "loss": 1.0461, + "step": 9042 + }, + { + "epoch": 0.58, + "learning_rate": 1.808110951287061e-05, + "loss": 1.0441, + "step": 9043 + }, + { + "epoch": 0.58, + "learning_rate": 1.808089482384766e-05, + "loss": 1.0749, + "step": 9044 + }, + { + "epoch": 0.58, + "learning_rate": 1.8080680134824707e-05, + "loss": 1.0792, + "step": 9045 + }, + { + "epoch": 0.58, + "learning_rate": 1.808046544580176e-05, + "loss": 1.103, + "step": 9046 + }, + { + "epoch": 0.58, + "learning_rate": 1.808025075677881e-05, + "loss": 1.0642, + "step": 9047 + }, + { + "epoch": 0.58, + "learning_rate": 1.8080036067755858e-05, + "loss": 1.0547, + "step": 9048 + }, + { + "epoch": 0.58, + "learning_rate": 1.8079821378732907e-05, + "loss": 1.0904, + "step": 9049 + }, + { + "epoch": 0.58, + "learning_rate": 1.8079606689709956e-05, + "loss": 1.0942, + "step": 9050 + }, + { + "epoch": 0.58, + "learning_rate": 1.8079392000687005e-05, + "loss": 1.0712, + "step": 9051 + }, + { + "epoch": 0.58, + "learning_rate": 1.8079177311664057e-05, + "loss": 1.0646, + "step": 9052 + }, + { + "epoch": 0.58, + "learning_rate": 1.8078962622641106e-05, + "loss": 1.0832, + "step": 9053 + }, + { + "epoch": 0.58, + "learning_rate": 1.8078747933618155e-05, + "loss": 1.0964, + "step": 9054 + }, + { + "epoch": 0.58, + "learning_rate": 1.8078533244595204e-05, + "loss": 1.0925, + "step": 9055 + }, + { + "epoch": 0.58, + "learning_rate": 1.8078318555572253e-05, + "loss": 1.1009, + "step": 9056 + }, + { + "epoch": 0.58, + "learning_rate": 1.8078103866549305e-05, + "loss": 1.0512, + "step": 9057 + }, + { + "epoch": 0.58, + "learning_rate": 1.8077889177526354e-05, + "loss": 1.1699, + "step": 9058 + }, + { + "epoch": 0.58, + "learning_rate": 1.8077674488503406e-05, + "loss": 1.0907, + "step": 9059 + }, + { + "epoch": 0.58, + "learning_rate": 1.8077459799480455e-05, + "loss": 1.061, + "step": 9060 + }, + { + "epoch": 0.58, + "learning_rate": 1.8077245110457504e-05, + "loss": 0.9938, + "step": 9061 + }, + { + "epoch": 0.58, + "learning_rate": 1.8077030421434553e-05, + "loss": 1.0999, + "step": 9062 + }, + { + "epoch": 0.58, + "learning_rate": 1.8076815732411602e-05, + "loss": 1.0607, + "step": 9063 + }, + { + "epoch": 0.58, + "learning_rate": 1.8076601043388655e-05, + "loss": 1.0685, + "step": 9064 + }, + { + "epoch": 0.58, + "learning_rate": 1.8076386354365704e-05, + "loss": 1.0866, + "step": 9065 + }, + { + "epoch": 0.58, + "learning_rate": 1.8076171665342753e-05, + "loss": 1.058, + "step": 9066 + }, + { + "epoch": 0.58, + "learning_rate": 1.80759569763198e-05, + "loss": 1.1283, + "step": 9067 + }, + { + "epoch": 0.58, + "learning_rate": 1.807574228729685e-05, + "loss": 1.198, + "step": 9068 + }, + { + "epoch": 0.58, + "learning_rate": 1.80755275982739e-05, + "loss": 1.1696, + "step": 9069 + }, + { + "epoch": 0.58, + "learning_rate": 1.8075312909250952e-05, + "loss": 1.0135, + "step": 9070 + }, + { + "epoch": 0.58, + "learning_rate": 1.8075098220228e-05, + "loss": 1.1461, + "step": 9071 + }, + { + "epoch": 0.58, + "learning_rate": 1.8074883531205053e-05, + "loss": 1.035, + "step": 9072 + }, + { + "epoch": 0.58, + "learning_rate": 1.8074668842182102e-05, + "loss": 1.0241, + "step": 9073 + }, + { + "epoch": 0.58, + "learning_rate": 1.807445415315915e-05, + "loss": 0.989, + "step": 9074 + }, + { + "epoch": 0.58, + "learning_rate": 1.80742394641362e-05, + "loss": 1.1051, + "step": 9075 + }, + { + "epoch": 0.58, + "learning_rate": 1.807402477511325e-05, + "loss": 1.1415, + "step": 9076 + }, + { + "epoch": 0.58, + "learning_rate": 1.80738100860903e-05, + "loss": 1.0515, + "step": 9077 + }, + { + "epoch": 0.58, + "learning_rate": 1.807359539706735e-05, + "loss": 1.0481, + "step": 9078 + }, + { + "epoch": 0.58, + "learning_rate": 1.80733807080444e-05, + "loss": 1.0656, + "step": 9079 + }, + { + "epoch": 0.58, + "learning_rate": 1.8073166019021448e-05, + "loss": 1.0658, + "step": 9080 + }, + { + "epoch": 0.58, + "learning_rate": 1.8072951329998497e-05, + "loss": 0.9594, + "step": 9081 + }, + { + "epoch": 0.58, + "learning_rate": 1.8072736640975546e-05, + "loss": 1.04, + "step": 9082 + }, + { + "epoch": 0.58, + "learning_rate": 1.80725219519526e-05, + "loss": 1.0511, + "step": 9083 + }, + { + "epoch": 0.58, + "learning_rate": 1.8072307262929648e-05, + "loss": 1.087, + "step": 9084 + }, + { + "epoch": 0.58, + "learning_rate": 1.8072092573906696e-05, + "loss": 1.1659, + "step": 9085 + }, + { + "epoch": 0.58, + "learning_rate": 1.807187788488375e-05, + "loss": 1.0542, + "step": 9086 + }, + { + "epoch": 0.58, + "learning_rate": 1.8071663195860798e-05, + "loss": 1.1059, + "step": 9087 + }, + { + "epoch": 0.58, + "learning_rate": 1.8071448506837847e-05, + "loss": 0.9405, + "step": 9088 + }, + { + "epoch": 0.58, + "learning_rate": 1.8071233817814896e-05, + "loss": 1.0192, + "step": 9089 + }, + { + "epoch": 0.58, + "learning_rate": 1.8071019128791948e-05, + "loss": 1.1412, + "step": 9090 + }, + { + "epoch": 0.58, + "learning_rate": 1.8070804439768997e-05, + "loss": 1.006, + "step": 9091 + }, + { + "epoch": 0.58, + "learning_rate": 1.8070589750746046e-05, + "loss": 1.1107, + "step": 9092 + }, + { + "epoch": 0.59, + "learning_rate": 1.8070375061723095e-05, + "loss": 1.1185, + "step": 9093 + }, + { + "epoch": 0.59, + "learning_rate": 1.8070160372700144e-05, + "loss": 1.0211, + "step": 9094 + }, + { + "epoch": 0.59, + "learning_rate": 1.8069945683677196e-05, + "loss": 1.1813, + "step": 9095 + }, + { + "epoch": 0.59, + "learning_rate": 1.8069730994654245e-05, + "loss": 1.0571, + "step": 9096 + }, + { + "epoch": 0.59, + "learning_rate": 1.8069516305631294e-05, + "loss": 1.0492, + "step": 9097 + }, + { + "epoch": 0.59, + "learning_rate": 1.8069301616608343e-05, + "loss": 1.137, + "step": 9098 + }, + { + "epoch": 0.59, + "learning_rate": 1.8069086927585392e-05, + "loss": 1.0475, + "step": 9099 + }, + { + "epoch": 0.59, + "learning_rate": 1.8068872238562444e-05, + "loss": 1.0854, + "step": 9100 + }, + { + "epoch": 0.59, + "learning_rate": 1.8068657549539493e-05, + "loss": 1.0721, + "step": 9101 + }, + { + "epoch": 0.59, + "learning_rate": 1.8068442860516542e-05, + "loss": 1.1059, + "step": 9102 + }, + { + "epoch": 0.59, + "learning_rate": 1.8068228171493595e-05, + "loss": 1.1177, + "step": 9103 + }, + { + "epoch": 0.59, + "learning_rate": 1.8068013482470644e-05, + "loss": 1.1264, + "step": 9104 + }, + { + "epoch": 0.59, + "learning_rate": 1.8067798793447693e-05, + "loss": 1.0423, + "step": 9105 + }, + { + "epoch": 0.59, + "learning_rate": 1.806758410442474e-05, + "loss": 0.9967, + "step": 9106 + }, + { + "epoch": 0.59, + "learning_rate": 1.806736941540179e-05, + "loss": 1.1383, + "step": 9107 + }, + { + "epoch": 0.59, + "learning_rate": 1.8067154726378843e-05, + "loss": 1.1568, + "step": 9108 + }, + { + "epoch": 0.59, + "learning_rate": 1.8066940037355892e-05, + "loss": 1.0557, + "step": 9109 + }, + { + "epoch": 0.59, + "learning_rate": 1.806672534833294e-05, + "loss": 1.0611, + "step": 9110 + }, + { + "epoch": 0.59, + "learning_rate": 1.806651065930999e-05, + "loss": 1.0381, + "step": 9111 + }, + { + "epoch": 0.59, + "learning_rate": 1.806629597028704e-05, + "loss": 1.0741, + "step": 9112 + }, + { + "epoch": 0.59, + "learning_rate": 1.806608128126409e-05, + "loss": 1.0421, + "step": 9113 + }, + { + "epoch": 0.59, + "learning_rate": 1.806586659224114e-05, + "loss": 1.1117, + "step": 9114 + }, + { + "epoch": 0.59, + "learning_rate": 1.8065651903218193e-05, + "loss": 1.0906, + "step": 9115 + }, + { + "epoch": 0.59, + "learning_rate": 1.806543721419524e-05, + "loss": 1.1639, + "step": 9116 + }, + { + "epoch": 0.59, + "learning_rate": 1.806522252517229e-05, + "loss": 1.0429, + "step": 9117 + }, + { + "epoch": 0.59, + "learning_rate": 1.806500783614934e-05, + "loss": 1.0839, + "step": 9118 + }, + { + "epoch": 0.59, + "learning_rate": 1.806479314712639e-05, + "loss": 1.037, + "step": 9119 + }, + { + "epoch": 0.59, + "learning_rate": 1.8064578458103437e-05, + "loss": 1.0671, + "step": 9120 + }, + { + "epoch": 0.59, + "learning_rate": 1.806436376908049e-05, + "loss": 1.0611, + "step": 9121 + }, + { + "epoch": 0.59, + "learning_rate": 1.806414908005754e-05, + "loss": 1.1281, + "step": 9122 + }, + { + "epoch": 0.59, + "learning_rate": 1.8063934391034588e-05, + "loss": 1.0036, + "step": 9123 + }, + { + "epoch": 0.59, + "learning_rate": 1.8063719702011637e-05, + "loss": 1.0676, + "step": 9124 + }, + { + "epoch": 0.59, + "learning_rate": 1.8063505012988686e-05, + "loss": 1.1734, + "step": 9125 + }, + { + "epoch": 0.59, + "learning_rate": 1.8063290323965738e-05, + "loss": 0.9978, + "step": 9126 + }, + { + "epoch": 0.59, + "learning_rate": 1.8063075634942787e-05, + "loss": 1.0218, + "step": 9127 + }, + { + "epoch": 0.59, + "learning_rate": 1.8062860945919836e-05, + "loss": 0.9893, + "step": 9128 + }, + { + "epoch": 0.59, + "learning_rate": 1.8062646256896888e-05, + "loss": 1.0695, + "step": 9129 + }, + { + "epoch": 0.59, + "learning_rate": 1.8062431567873937e-05, + "loss": 1.0411, + "step": 9130 + }, + { + "epoch": 0.59, + "learning_rate": 1.8062216878850986e-05, + "loss": 1.1328, + "step": 9131 + }, + { + "epoch": 0.59, + "learning_rate": 1.8062002189828035e-05, + "loss": 1.1954, + "step": 9132 + }, + { + "epoch": 0.59, + "learning_rate": 1.8061787500805087e-05, + "loss": 1.0612, + "step": 9133 + }, + { + "epoch": 0.59, + "learning_rate": 1.8061572811782136e-05, + "loss": 1.1083, + "step": 9134 + }, + { + "epoch": 0.59, + "learning_rate": 1.8061358122759185e-05, + "loss": 1.0454, + "step": 9135 + }, + { + "epoch": 0.59, + "learning_rate": 1.8061143433736234e-05, + "loss": 1.0248, + "step": 9136 + }, + { + "epoch": 0.59, + "learning_rate": 1.8060928744713283e-05, + "loss": 1.1366, + "step": 9137 + }, + { + "epoch": 0.59, + "learning_rate": 1.8060714055690332e-05, + "loss": 1.1331, + "step": 9138 + }, + { + "epoch": 0.59, + "learning_rate": 1.8060499366667385e-05, + "loss": 1.0376, + "step": 9139 + }, + { + "epoch": 0.59, + "learning_rate": 1.8060284677644434e-05, + "loss": 1.0327, + "step": 9140 + }, + { + "epoch": 0.59, + "learning_rate": 1.8060069988621483e-05, + "loss": 1.1456, + "step": 9141 + }, + { + "epoch": 0.59, + "learning_rate": 1.805985529959853e-05, + "loss": 1.1203, + "step": 9142 + }, + { + "epoch": 0.59, + "learning_rate": 1.805964061057558e-05, + "loss": 1.0277, + "step": 9143 + }, + { + "epoch": 0.59, + "learning_rate": 1.8059425921552633e-05, + "loss": 1.1054, + "step": 9144 + }, + { + "epoch": 0.59, + "learning_rate": 1.8059211232529682e-05, + "loss": 1.1491, + "step": 9145 + }, + { + "epoch": 0.59, + "learning_rate": 1.8058996543506734e-05, + "loss": 1.0683, + "step": 9146 + }, + { + "epoch": 0.59, + "learning_rate": 1.8058781854483783e-05, + "loss": 1.0437, + "step": 9147 + }, + { + "epoch": 0.59, + "learning_rate": 1.8058567165460832e-05, + "loss": 1.1281, + "step": 9148 + }, + { + "epoch": 0.59, + "learning_rate": 1.805835247643788e-05, + "loss": 1.0097, + "step": 9149 + }, + { + "epoch": 0.59, + "learning_rate": 1.805813778741493e-05, + "loss": 1.0937, + "step": 9150 + }, + { + "epoch": 0.59, + "learning_rate": 1.805792309839198e-05, + "loss": 1.0386, + "step": 9151 + }, + { + "epoch": 0.59, + "learning_rate": 1.805770840936903e-05, + "loss": 1.0145, + "step": 9152 + }, + { + "epoch": 0.59, + "learning_rate": 1.805749372034608e-05, + "loss": 1.1012, + "step": 9153 + }, + { + "epoch": 0.59, + "learning_rate": 1.805727903132313e-05, + "loss": 1.1144, + "step": 9154 + }, + { + "epoch": 0.59, + "learning_rate": 1.8057064342300178e-05, + "loss": 1.0865, + "step": 9155 + }, + { + "epoch": 0.59, + "learning_rate": 1.8056849653277227e-05, + "loss": 0.9962, + "step": 9156 + }, + { + "epoch": 0.59, + "learning_rate": 1.805663496425428e-05, + "loss": 1.1411, + "step": 9157 + }, + { + "epoch": 0.59, + "learning_rate": 1.805642027523133e-05, + "loss": 1.0645, + "step": 9158 + }, + { + "epoch": 0.59, + "learning_rate": 1.805620558620838e-05, + "loss": 1.045, + "step": 9159 + }, + { + "epoch": 0.59, + "learning_rate": 1.805599089718543e-05, + "loss": 1.0495, + "step": 9160 + }, + { + "epoch": 0.59, + "learning_rate": 1.805577620816248e-05, + "loss": 1.0951, + "step": 9161 + }, + { + "epoch": 0.59, + "learning_rate": 1.8055561519139528e-05, + "loss": 1.0282, + "step": 9162 + }, + { + "epoch": 0.59, + "learning_rate": 1.8055346830116577e-05, + "loss": 1.0376, + "step": 9163 + }, + { + "epoch": 0.59, + "learning_rate": 1.805513214109363e-05, + "loss": 1.1133, + "step": 9164 + }, + { + "epoch": 0.59, + "learning_rate": 1.8054917452070678e-05, + "loss": 1.1165, + "step": 9165 + }, + { + "epoch": 0.59, + "learning_rate": 1.8054702763047727e-05, + "loss": 1.0475, + "step": 9166 + }, + { + "epoch": 0.59, + "learning_rate": 1.8054488074024776e-05, + "loss": 1.2285, + "step": 9167 + }, + { + "epoch": 0.59, + "learning_rate": 1.8054273385001825e-05, + "loss": 1.0534, + "step": 9168 + }, + { + "epoch": 0.59, + "learning_rate": 1.8054058695978874e-05, + "loss": 1.146, + "step": 9169 + }, + { + "epoch": 0.59, + "learning_rate": 1.8053844006955926e-05, + "loss": 1.1288, + "step": 9170 + }, + { + "epoch": 0.59, + "learning_rate": 1.8053629317932975e-05, + "loss": 1.0872, + "step": 9171 + }, + { + "epoch": 0.59, + "learning_rate": 1.8053414628910024e-05, + "loss": 1.0816, + "step": 9172 + }, + { + "epoch": 0.59, + "learning_rate": 1.8053199939887077e-05, + "loss": 1.0189, + "step": 9173 + }, + { + "epoch": 0.59, + "learning_rate": 1.8052985250864125e-05, + "loss": 1.1192, + "step": 9174 + }, + { + "epoch": 0.59, + "learning_rate": 1.8052770561841174e-05, + "loss": 1.0223, + "step": 9175 + }, + { + "epoch": 0.59, + "learning_rate": 1.8052555872818223e-05, + "loss": 1.0898, + "step": 9176 + }, + { + "epoch": 0.59, + "learning_rate": 1.8052341183795276e-05, + "loss": 1.0785, + "step": 9177 + }, + { + "epoch": 0.59, + "learning_rate": 1.8052126494772325e-05, + "loss": 1.1387, + "step": 9178 + }, + { + "epoch": 0.59, + "learning_rate": 1.8051911805749374e-05, + "loss": 1.1087, + "step": 9179 + }, + { + "epoch": 0.59, + "learning_rate": 1.8051697116726423e-05, + "loss": 0.9878, + "step": 9180 + }, + { + "epoch": 0.59, + "learning_rate": 1.805148242770347e-05, + "loss": 1.0452, + "step": 9181 + }, + { + "epoch": 0.59, + "learning_rate": 1.8051267738680524e-05, + "loss": 1.0875, + "step": 9182 + }, + { + "epoch": 0.59, + "learning_rate": 1.8051053049657573e-05, + "loss": 1.0868, + "step": 9183 + }, + { + "epoch": 0.59, + "learning_rate": 1.8050838360634622e-05, + "loss": 1.0688, + "step": 9184 + }, + { + "epoch": 0.59, + "learning_rate": 1.805062367161167e-05, + "loss": 1.0669, + "step": 9185 + }, + { + "epoch": 0.59, + "learning_rate": 1.805040898258872e-05, + "loss": 1.0761, + "step": 9186 + }, + { + "epoch": 0.59, + "learning_rate": 1.805019429356577e-05, + "loss": 1.1094, + "step": 9187 + }, + { + "epoch": 0.59, + "learning_rate": 1.804997960454282e-05, + "loss": 1.0992, + "step": 9188 + }, + { + "epoch": 0.59, + "learning_rate": 1.804976491551987e-05, + "loss": 0.9778, + "step": 9189 + }, + { + "epoch": 0.59, + "learning_rate": 1.8049550226496922e-05, + "loss": 1.0601, + "step": 9190 + }, + { + "epoch": 0.59, + "learning_rate": 1.804933553747397e-05, + "loss": 1.0919, + "step": 9191 + }, + { + "epoch": 0.59, + "learning_rate": 1.804912084845102e-05, + "loss": 1.0613, + "step": 9192 + }, + { + "epoch": 0.59, + "learning_rate": 1.804890615942807e-05, + "loss": 1.0748, + "step": 9193 + }, + { + "epoch": 0.59, + "learning_rate": 1.804869147040512e-05, + "loss": 0.9962, + "step": 9194 + }, + { + "epoch": 0.59, + "learning_rate": 1.804847678138217e-05, + "loss": 1.0707, + "step": 9195 + }, + { + "epoch": 0.59, + "learning_rate": 1.804826209235922e-05, + "loss": 1.0861, + "step": 9196 + }, + { + "epoch": 0.59, + "learning_rate": 1.804804740333627e-05, + "loss": 1.1483, + "step": 9197 + }, + { + "epoch": 0.59, + "learning_rate": 1.8047832714313318e-05, + "loss": 1.1053, + "step": 9198 + }, + { + "epoch": 0.59, + "learning_rate": 1.8047618025290367e-05, + "loss": 1.0356, + "step": 9199 + }, + { + "epoch": 0.59, + "learning_rate": 1.804740333626742e-05, + "loss": 1.0947, + "step": 9200 + }, + { + "epoch": 0.59, + "learning_rate": 1.8047188647244468e-05, + "loss": 1.0988, + "step": 9201 + }, + { + "epoch": 0.59, + "learning_rate": 1.804697395822152e-05, + "loss": 1.1571, + "step": 9202 + }, + { + "epoch": 0.59, + "learning_rate": 1.804675926919857e-05, + "loss": 1.0593, + "step": 9203 + }, + { + "epoch": 0.59, + "learning_rate": 1.8046544580175618e-05, + "loss": 1.0384, + "step": 9204 + }, + { + "epoch": 0.59, + "learning_rate": 1.8046329891152667e-05, + "loss": 1.1281, + "step": 9205 + }, + { + "epoch": 0.59, + "learning_rate": 1.8046115202129716e-05, + "loss": 1.0315, + "step": 9206 + }, + { + "epoch": 0.59, + "learning_rate": 1.8045900513106765e-05, + "loss": 1.0062, + "step": 9207 + }, + { + "epoch": 0.59, + "learning_rate": 1.8045685824083817e-05, + "loss": 1.0084, + "step": 9208 + }, + { + "epoch": 0.59, + "learning_rate": 1.8045471135060866e-05, + "loss": 1.0698, + "step": 9209 + }, + { + "epoch": 0.59, + "learning_rate": 1.8045256446037915e-05, + "loss": 1.105, + "step": 9210 + }, + { + "epoch": 0.59, + "learning_rate": 1.8045041757014964e-05, + "loss": 1.0694, + "step": 9211 + }, + { + "epoch": 0.59, + "learning_rate": 1.8044827067992013e-05, + "loss": 1.0827, + "step": 9212 + }, + { + "epoch": 0.59, + "learning_rate": 1.8044612378969066e-05, + "loss": 1.1039, + "step": 9213 + }, + { + "epoch": 0.59, + "learning_rate": 1.8044397689946115e-05, + "loss": 1.0897, + "step": 9214 + }, + { + "epoch": 0.59, + "learning_rate": 1.8044183000923164e-05, + "loss": 1.0314, + "step": 9215 + }, + { + "epoch": 0.59, + "learning_rate": 1.8043968311900213e-05, + "loss": 0.9958, + "step": 9216 + }, + { + "epoch": 0.59, + "learning_rate": 1.8043753622877265e-05, + "loss": 1.1026, + "step": 9217 + }, + { + "epoch": 0.59, + "learning_rate": 1.8043538933854314e-05, + "loss": 1.1294, + "step": 9218 + }, + { + "epoch": 0.59, + "learning_rate": 1.8043324244831363e-05, + "loss": 1.1919, + "step": 9219 + }, + { + "epoch": 0.59, + "learning_rate": 1.8043109555808415e-05, + "loss": 1.1116, + "step": 9220 + }, + { + "epoch": 0.59, + "learning_rate": 1.8042894866785464e-05, + "loss": 1.0742, + "step": 9221 + }, + { + "epoch": 0.59, + "learning_rate": 1.8042680177762513e-05, + "loss": 1.0975, + "step": 9222 + }, + { + "epoch": 0.59, + "learning_rate": 1.8042465488739562e-05, + "loss": 1.2368, + "step": 9223 + }, + { + "epoch": 0.59, + "learning_rate": 1.804225079971661e-05, + "loss": 1.0724, + "step": 9224 + }, + { + "epoch": 0.59, + "learning_rate": 1.804203611069366e-05, + "loss": 1.1204, + "step": 9225 + }, + { + "epoch": 0.59, + "learning_rate": 1.8041821421670712e-05, + "loss": 1.101, + "step": 9226 + }, + { + "epoch": 0.59, + "learning_rate": 1.804160673264776e-05, + "loss": 1.0757, + "step": 9227 + }, + { + "epoch": 0.59, + "learning_rate": 1.804139204362481e-05, + "loss": 1.0682, + "step": 9228 + }, + { + "epoch": 0.59, + "learning_rate": 1.804117735460186e-05, + "loss": 1.0845, + "step": 9229 + }, + { + "epoch": 0.59, + "learning_rate": 1.8040962665578908e-05, + "loss": 1.1141, + "step": 9230 + }, + { + "epoch": 0.59, + "learning_rate": 1.804074797655596e-05, + "loss": 1.0438, + "step": 9231 + }, + { + "epoch": 0.59, + "learning_rate": 1.804053328753301e-05, + "loss": 0.9997, + "step": 9232 + }, + { + "epoch": 0.59, + "learning_rate": 1.8040318598510062e-05, + "loss": 1.0556, + "step": 9233 + }, + { + "epoch": 0.59, + "learning_rate": 1.804010390948711e-05, + "loss": 1.0981, + "step": 9234 + }, + { + "epoch": 0.59, + "learning_rate": 1.803988922046416e-05, + "loss": 1.2983, + "step": 9235 + }, + { + "epoch": 0.59, + "learning_rate": 1.803967453144121e-05, + "loss": 1.1485, + "step": 9236 + }, + { + "epoch": 0.59, + "learning_rate": 1.8039459842418258e-05, + "loss": 1.0903, + "step": 9237 + }, + { + "epoch": 0.59, + "learning_rate": 1.8039245153395307e-05, + "loss": 1.1099, + "step": 9238 + }, + { + "epoch": 0.59, + "learning_rate": 1.803903046437236e-05, + "loss": 1.1255, + "step": 9239 + }, + { + "epoch": 0.59, + "learning_rate": 1.8038815775349408e-05, + "loss": 1.0878, + "step": 9240 + }, + { + "epoch": 0.59, + "learning_rate": 1.8038601086326457e-05, + "loss": 1.0398, + "step": 9241 + }, + { + "epoch": 0.59, + "learning_rate": 1.8038386397303506e-05, + "loss": 1.1672, + "step": 9242 + }, + { + "epoch": 0.59, + "learning_rate": 1.8038171708280555e-05, + "loss": 1.0172, + "step": 9243 + }, + { + "epoch": 0.59, + "learning_rate": 1.8037957019257607e-05, + "loss": 1.0695, + "step": 9244 + }, + { + "epoch": 0.59, + "learning_rate": 1.8037742330234656e-05, + "loss": 1.053, + "step": 9245 + }, + { + "epoch": 0.59, + "learning_rate": 1.803752764121171e-05, + "loss": 1.0324, + "step": 9246 + }, + { + "epoch": 0.59, + "learning_rate": 1.8037312952188758e-05, + "loss": 1.1137, + "step": 9247 + }, + { + "epoch": 0.59, + "learning_rate": 1.8037098263165807e-05, + "loss": 1.1275, + "step": 9248 + }, + { + "epoch": 0.6, + "learning_rate": 1.8036883574142855e-05, + "loss": 1.1475, + "step": 9249 + }, + { + "epoch": 0.6, + "learning_rate": 1.8036668885119904e-05, + "loss": 1.0622, + "step": 9250 + }, + { + "epoch": 0.6, + "learning_rate": 1.8036454196096957e-05, + "loss": 1.0509, + "step": 9251 + }, + { + "epoch": 0.6, + "learning_rate": 1.8036239507074006e-05, + "loss": 1.0874, + "step": 9252 + }, + { + "epoch": 0.6, + "learning_rate": 1.8036024818051055e-05, + "loss": 1.0421, + "step": 9253 + }, + { + "epoch": 0.6, + "learning_rate": 1.8035810129028104e-05, + "loss": 1.13, + "step": 9254 + }, + { + "epoch": 0.6, + "learning_rate": 1.8035595440005153e-05, + "loss": 1.0745, + "step": 9255 + }, + { + "epoch": 0.6, + "learning_rate": 1.80353807509822e-05, + "loss": 1.0537, + "step": 9256 + }, + { + "epoch": 0.6, + "learning_rate": 1.8035166061959254e-05, + "loss": 1.0461, + "step": 9257 + }, + { + "epoch": 0.6, + "learning_rate": 1.8034951372936303e-05, + "loss": 1.0736, + "step": 9258 + }, + { + "epoch": 0.6, + "learning_rate": 1.8034736683913352e-05, + "loss": 0.9995, + "step": 9259 + }, + { + "epoch": 0.6, + "learning_rate": 1.8034521994890404e-05, + "loss": 1.123, + "step": 9260 + }, + { + "epoch": 0.6, + "learning_rate": 1.8034307305867453e-05, + "loss": 1.1138, + "step": 9261 + }, + { + "epoch": 0.6, + "learning_rate": 1.8034092616844502e-05, + "loss": 1.1194, + "step": 9262 + }, + { + "epoch": 0.6, + "learning_rate": 1.803387792782155e-05, + "loss": 1.0327, + "step": 9263 + }, + { + "epoch": 0.6, + "learning_rate": 1.8033663238798603e-05, + "loss": 1.1272, + "step": 9264 + }, + { + "epoch": 0.6, + "learning_rate": 1.8033448549775652e-05, + "loss": 1.1812, + "step": 9265 + }, + { + "epoch": 0.6, + "learning_rate": 1.80332338607527e-05, + "loss": 1.0069, + "step": 9266 + }, + { + "epoch": 0.6, + "learning_rate": 1.803301917172975e-05, + "loss": 1.1378, + "step": 9267 + }, + { + "epoch": 0.6, + "learning_rate": 1.80328044827068e-05, + "loss": 1.041, + "step": 9268 + }, + { + "epoch": 0.6, + "learning_rate": 1.8032589793683852e-05, + "loss": 1.1216, + "step": 9269 + }, + { + "epoch": 0.6, + "learning_rate": 1.80323751046609e-05, + "loss": 1.1101, + "step": 9270 + }, + { + "epoch": 0.6, + "learning_rate": 1.803216041563795e-05, + "loss": 1.1374, + "step": 9271 + }, + { + "epoch": 0.6, + "learning_rate": 1.8031945726615e-05, + "loss": 1.0826, + "step": 9272 + }, + { + "epoch": 0.6, + "learning_rate": 1.8031731037592048e-05, + "loss": 1.0878, + "step": 9273 + }, + { + "epoch": 0.6, + "learning_rate": 1.8031516348569097e-05, + "loss": 1.0639, + "step": 9274 + }, + { + "epoch": 0.6, + "learning_rate": 1.803130165954615e-05, + "loss": 1.0194, + "step": 9275 + }, + { + "epoch": 0.6, + "learning_rate": 1.8031086970523198e-05, + "loss": 1.147, + "step": 9276 + }, + { + "epoch": 0.6, + "learning_rate": 1.803087228150025e-05, + "loss": 1.1341, + "step": 9277 + }, + { + "epoch": 0.6, + "learning_rate": 1.80306575924773e-05, + "loss": 1.0326, + "step": 9278 + }, + { + "epoch": 0.6, + "learning_rate": 1.8030442903454348e-05, + "loss": 1.0786, + "step": 9279 + }, + { + "epoch": 0.6, + "learning_rate": 1.8030228214431397e-05, + "loss": 1.0484, + "step": 9280 + }, + { + "epoch": 0.6, + "learning_rate": 1.8030013525408446e-05, + "loss": 1.1439, + "step": 9281 + }, + { + "epoch": 0.6, + "learning_rate": 1.80297988363855e-05, + "loss": 0.9719, + "step": 9282 + }, + { + "epoch": 0.6, + "learning_rate": 1.8029584147362547e-05, + "loss": 1.0856, + "step": 9283 + }, + { + "epoch": 0.6, + "learning_rate": 1.8029369458339596e-05, + "loss": 1.1337, + "step": 9284 + }, + { + "epoch": 0.6, + "learning_rate": 1.8029154769316645e-05, + "loss": 1.0335, + "step": 9285 + }, + { + "epoch": 0.6, + "learning_rate": 1.8028940080293694e-05, + "loss": 1.1185, + "step": 9286 + }, + { + "epoch": 0.6, + "learning_rate": 1.8028725391270747e-05, + "loss": 1.0849, + "step": 9287 + }, + { + "epoch": 0.6, + "learning_rate": 1.8028510702247796e-05, + "loss": 1.0482, + "step": 9288 + }, + { + "epoch": 0.6, + "learning_rate": 1.8028296013224848e-05, + "loss": 0.9462, + "step": 9289 + }, + { + "epoch": 0.6, + "learning_rate": 1.8028081324201897e-05, + "loss": 1.092, + "step": 9290 + }, + { + "epoch": 0.6, + "learning_rate": 1.8027866635178946e-05, + "loss": 1.1081, + "step": 9291 + }, + { + "epoch": 0.6, + "learning_rate": 1.8027651946155995e-05, + "loss": 1.1134, + "step": 9292 + }, + { + "epoch": 0.6, + "learning_rate": 1.8027437257133044e-05, + "loss": 1.1037, + "step": 9293 + }, + { + "epoch": 0.6, + "learning_rate": 1.8027222568110093e-05, + "loss": 1.1312, + "step": 9294 + }, + { + "epoch": 0.6, + "learning_rate": 1.8027007879087145e-05, + "loss": 1.0466, + "step": 9295 + }, + { + "epoch": 0.6, + "learning_rate": 1.8026793190064194e-05, + "loss": 1.044, + "step": 9296 + }, + { + "epoch": 0.6, + "learning_rate": 1.8026578501041243e-05, + "loss": 1.0969, + "step": 9297 + }, + { + "epoch": 0.6, + "learning_rate": 1.8026363812018292e-05, + "loss": 1.0536, + "step": 9298 + }, + { + "epoch": 0.6, + "learning_rate": 1.802614912299534e-05, + "loss": 1.0421, + "step": 9299 + }, + { + "epoch": 0.6, + "learning_rate": 1.8025934433972393e-05, + "loss": 1.1407, + "step": 9300 + }, + { + "epoch": 0.6, + "learning_rate": 1.8025719744949442e-05, + "loss": 1.0755, + "step": 9301 + }, + { + "epoch": 0.6, + "learning_rate": 1.802550505592649e-05, + "loss": 1.1716, + "step": 9302 + }, + { + "epoch": 0.6, + "learning_rate": 1.802529036690354e-05, + "loss": 0.9916, + "step": 9303 + }, + { + "epoch": 0.6, + "learning_rate": 1.8025075677880593e-05, + "loss": 1.1098, + "step": 9304 + }, + { + "epoch": 0.6, + "learning_rate": 1.802486098885764e-05, + "loss": 1.0851, + "step": 9305 + }, + { + "epoch": 0.6, + "learning_rate": 1.802464629983469e-05, + "loss": 1.0792, + "step": 9306 + }, + { + "epoch": 0.6, + "learning_rate": 1.8024431610811743e-05, + "loss": 1.1023, + "step": 9307 + }, + { + "epoch": 0.6, + "learning_rate": 1.8024216921788792e-05, + "loss": 1.1273, + "step": 9308 + }, + { + "epoch": 0.6, + "learning_rate": 1.802400223276584e-05, + "loss": 1.0821, + "step": 9309 + }, + { + "epoch": 0.6, + "learning_rate": 1.802378754374289e-05, + "loss": 1.0398, + "step": 9310 + }, + { + "epoch": 0.6, + "learning_rate": 1.802357285471994e-05, + "loss": 1.0163, + "step": 9311 + }, + { + "epoch": 0.6, + "learning_rate": 1.8023358165696988e-05, + "loss": 1.0565, + "step": 9312 + }, + { + "epoch": 0.6, + "learning_rate": 1.802314347667404e-05, + "loss": 1.1286, + "step": 9313 + }, + { + "epoch": 0.6, + "learning_rate": 1.802292878765109e-05, + "loss": 1.1078, + "step": 9314 + }, + { + "epoch": 0.6, + "learning_rate": 1.8022714098628138e-05, + "loss": 1.0847, + "step": 9315 + }, + { + "epoch": 0.6, + "learning_rate": 1.8022499409605187e-05, + "loss": 1.0869, + "step": 9316 + }, + { + "epoch": 0.6, + "learning_rate": 1.8022284720582236e-05, + "loss": 1.0598, + "step": 9317 + }, + { + "epoch": 0.6, + "learning_rate": 1.8022070031559288e-05, + "loss": 1.1274, + "step": 9318 + }, + { + "epoch": 0.6, + "learning_rate": 1.8021855342536337e-05, + "loss": 1.0609, + "step": 9319 + }, + { + "epoch": 0.6, + "learning_rate": 1.802164065351339e-05, + "loss": 1.0685, + "step": 9320 + }, + { + "epoch": 0.6, + "learning_rate": 1.802142596449044e-05, + "loss": 1.0262, + "step": 9321 + }, + { + "epoch": 0.6, + "learning_rate": 1.8021211275467488e-05, + "loss": 1.0257, + "step": 9322 + }, + { + "epoch": 0.6, + "learning_rate": 1.8020996586444536e-05, + "loss": 1.0789, + "step": 9323 + }, + { + "epoch": 0.6, + "learning_rate": 1.8020781897421585e-05, + "loss": 1.1437, + "step": 9324 + }, + { + "epoch": 0.6, + "learning_rate": 1.8020567208398634e-05, + "loss": 1.1156, + "step": 9325 + }, + { + "epoch": 0.6, + "learning_rate": 1.8020352519375687e-05, + "loss": 1.0915, + "step": 9326 + }, + { + "epoch": 0.6, + "learning_rate": 1.8020137830352736e-05, + "loss": 1.123, + "step": 9327 + }, + { + "epoch": 0.6, + "learning_rate": 1.8019923141329785e-05, + "loss": 1.0153, + "step": 9328 + }, + { + "epoch": 0.6, + "learning_rate": 1.8019708452306834e-05, + "loss": 1.0813, + "step": 9329 + }, + { + "epoch": 0.6, + "learning_rate": 1.8019493763283883e-05, + "loss": 1.144, + "step": 9330 + }, + { + "epoch": 0.6, + "learning_rate": 1.8019279074260935e-05, + "loss": 1.0969, + "step": 9331 + }, + { + "epoch": 0.6, + "learning_rate": 1.8019064385237984e-05, + "loss": 1.0829, + "step": 9332 + }, + { + "epoch": 0.6, + "learning_rate": 1.8018849696215036e-05, + "loss": 1.1014, + "step": 9333 + }, + { + "epoch": 0.6, + "learning_rate": 1.8018635007192085e-05, + "loss": 1.0536, + "step": 9334 + }, + { + "epoch": 0.6, + "learning_rate": 1.8018420318169134e-05, + "loss": 1.0705, + "step": 9335 + }, + { + "epoch": 0.6, + "learning_rate": 1.8018205629146183e-05, + "loss": 1.0024, + "step": 9336 + }, + { + "epoch": 0.6, + "learning_rate": 1.8017990940123232e-05, + "loss": 1.0708, + "step": 9337 + }, + { + "epoch": 0.6, + "learning_rate": 1.8017776251100284e-05, + "loss": 1.0245, + "step": 9338 + }, + { + "epoch": 0.6, + "learning_rate": 1.8017561562077333e-05, + "loss": 1.0939, + "step": 9339 + }, + { + "epoch": 0.6, + "learning_rate": 1.8017346873054382e-05, + "loss": 1.1339, + "step": 9340 + }, + { + "epoch": 0.6, + "learning_rate": 1.801713218403143e-05, + "loss": 1.0249, + "step": 9341 + }, + { + "epoch": 0.6, + "learning_rate": 1.801691749500848e-05, + "loss": 1.0808, + "step": 9342 + }, + { + "epoch": 0.6, + "learning_rate": 1.801670280598553e-05, + "loss": 1.0931, + "step": 9343 + }, + { + "epoch": 0.6, + "learning_rate": 1.801648811696258e-05, + "loss": 1.1278, + "step": 9344 + }, + { + "epoch": 0.6, + "learning_rate": 1.801627342793963e-05, + "loss": 1.0867, + "step": 9345 + }, + { + "epoch": 0.6, + "learning_rate": 1.801605873891668e-05, + "loss": 1.1182, + "step": 9346 + }, + { + "epoch": 0.6, + "learning_rate": 1.8015844049893732e-05, + "loss": 1.0273, + "step": 9347 + }, + { + "epoch": 0.6, + "learning_rate": 1.801562936087078e-05, + "loss": 1.0575, + "step": 9348 + }, + { + "epoch": 0.6, + "learning_rate": 1.801541467184783e-05, + "loss": 0.9835, + "step": 9349 + }, + { + "epoch": 0.6, + "learning_rate": 1.801519998282488e-05, + "loss": 1.0798, + "step": 9350 + }, + { + "epoch": 0.6, + "learning_rate": 1.801498529380193e-05, + "loss": 1.1214, + "step": 9351 + }, + { + "epoch": 0.6, + "learning_rate": 1.801477060477898e-05, + "loss": 1.1355, + "step": 9352 + }, + { + "epoch": 0.6, + "learning_rate": 1.801455591575603e-05, + "loss": 1.1218, + "step": 9353 + }, + { + "epoch": 0.6, + "learning_rate": 1.8014341226733078e-05, + "loss": 1.0915, + "step": 9354 + }, + { + "epoch": 0.6, + "learning_rate": 1.8014126537710127e-05, + "loss": 1.0475, + "step": 9355 + }, + { + "epoch": 0.6, + "learning_rate": 1.801391184868718e-05, + "loss": 1.0462, + "step": 9356 + }, + { + "epoch": 0.6, + "learning_rate": 1.801369715966423e-05, + "loss": 1.0789, + "step": 9357 + }, + { + "epoch": 0.6, + "learning_rate": 1.8013482470641277e-05, + "loss": 1.029, + "step": 9358 + }, + { + "epoch": 0.6, + "learning_rate": 1.8013267781618326e-05, + "loss": 1.1662, + "step": 9359 + }, + { + "epoch": 0.6, + "learning_rate": 1.8013053092595375e-05, + "loss": 1.0807, + "step": 9360 + }, + { + "epoch": 0.6, + "learning_rate": 1.8012838403572424e-05, + "loss": 1.1454, + "step": 9361 + }, + { + "epoch": 0.6, + "learning_rate": 1.8012623714549477e-05, + "loss": 1.1009, + "step": 9362 + }, + { + "epoch": 0.6, + "learning_rate": 1.8012409025526526e-05, + "loss": 1.1154, + "step": 9363 + }, + { + "epoch": 0.6, + "learning_rate": 1.8012194336503578e-05, + "loss": 1.104, + "step": 9364 + }, + { + "epoch": 0.6, + "learning_rate": 1.8011979647480627e-05, + "loss": 1.1816, + "step": 9365 + }, + { + "epoch": 0.6, + "learning_rate": 1.8011764958457676e-05, + "loss": 0.9953, + "step": 9366 + }, + { + "epoch": 0.6, + "learning_rate": 1.8011550269434725e-05, + "loss": 1.1246, + "step": 9367 + }, + { + "epoch": 0.6, + "learning_rate": 1.8011335580411774e-05, + "loss": 1.0768, + "step": 9368 + }, + { + "epoch": 0.6, + "learning_rate": 1.8011120891388826e-05, + "loss": 0.9981, + "step": 9369 + }, + { + "epoch": 0.6, + "learning_rate": 1.8010906202365875e-05, + "loss": 1.052, + "step": 9370 + }, + { + "epoch": 0.6, + "learning_rate": 1.8010691513342924e-05, + "loss": 1.0826, + "step": 9371 + }, + { + "epoch": 0.6, + "learning_rate": 1.8010476824319973e-05, + "loss": 1.0156, + "step": 9372 + }, + { + "epoch": 0.6, + "learning_rate": 1.8010262135297022e-05, + "loss": 1.0251, + "step": 9373 + }, + { + "epoch": 0.6, + "learning_rate": 1.801004744627407e-05, + "loss": 1.0799, + "step": 9374 + }, + { + "epoch": 0.6, + "learning_rate": 1.8009832757251123e-05, + "loss": 1.0804, + "step": 9375 + }, + { + "epoch": 0.6, + "learning_rate": 1.8009618068228176e-05, + "loss": 1.0359, + "step": 9376 + }, + { + "epoch": 0.6, + "learning_rate": 1.8009403379205225e-05, + "loss": 1.0885, + "step": 9377 + }, + { + "epoch": 0.6, + "learning_rate": 1.8009188690182274e-05, + "loss": 1.0492, + "step": 9378 + }, + { + "epoch": 0.6, + "learning_rate": 1.8008974001159323e-05, + "loss": 1.0763, + "step": 9379 + }, + { + "epoch": 0.6, + "learning_rate": 1.800875931213637e-05, + "loss": 1.0079, + "step": 9380 + }, + { + "epoch": 0.6, + "learning_rate": 1.800854462311342e-05, + "loss": 1.0534, + "step": 9381 + }, + { + "epoch": 0.6, + "learning_rate": 1.8008329934090473e-05, + "loss": 1.1816, + "step": 9382 + }, + { + "epoch": 0.6, + "learning_rate": 1.8008115245067522e-05, + "loss": 1.0497, + "step": 9383 + }, + { + "epoch": 0.6, + "learning_rate": 1.800790055604457e-05, + "loss": 1.1015, + "step": 9384 + }, + { + "epoch": 0.6, + "learning_rate": 1.800768586702162e-05, + "loss": 1.0305, + "step": 9385 + }, + { + "epoch": 0.6, + "learning_rate": 1.800747117799867e-05, + "loss": 1.1262, + "step": 9386 + }, + { + "epoch": 0.6, + "learning_rate": 1.800725648897572e-05, + "loss": 1.0358, + "step": 9387 + }, + { + "epoch": 0.6, + "learning_rate": 1.800704179995277e-05, + "loss": 1.075, + "step": 9388 + }, + { + "epoch": 0.6, + "learning_rate": 1.800682711092982e-05, + "loss": 1.1041, + "step": 9389 + }, + { + "epoch": 0.6, + "learning_rate": 1.8006612421906868e-05, + "loss": 1.1032, + "step": 9390 + }, + { + "epoch": 0.6, + "learning_rate": 1.800639773288392e-05, + "loss": 1.0617, + "step": 9391 + }, + { + "epoch": 0.6, + "learning_rate": 1.800618304386097e-05, + "loss": 1.0475, + "step": 9392 + }, + { + "epoch": 0.6, + "learning_rate": 1.8005968354838018e-05, + "loss": 1.1301, + "step": 9393 + }, + { + "epoch": 0.6, + "learning_rate": 1.800575366581507e-05, + "loss": 1.048, + "step": 9394 + }, + { + "epoch": 0.6, + "learning_rate": 1.800553897679212e-05, + "loss": 1.132, + "step": 9395 + }, + { + "epoch": 0.6, + "learning_rate": 1.800532428776917e-05, + "loss": 1.1389, + "step": 9396 + }, + { + "epoch": 0.6, + "learning_rate": 1.8005109598746217e-05, + "loss": 1.0682, + "step": 9397 + }, + { + "epoch": 0.6, + "learning_rate": 1.8004894909723266e-05, + "loss": 1.1061, + "step": 9398 + }, + { + "epoch": 0.6, + "learning_rate": 1.8004680220700315e-05, + "loss": 1.026, + "step": 9399 + }, + { + "epoch": 0.6, + "learning_rate": 1.8004465531677368e-05, + "loss": 1.0881, + "step": 9400 + }, + { + "epoch": 0.6, + "learning_rate": 1.8004250842654417e-05, + "loss": 1.119, + "step": 9401 + }, + { + "epoch": 0.6, + "learning_rate": 1.8004036153631466e-05, + "loss": 1.0506, + "step": 9402 + }, + { + "epoch": 0.6, + "learning_rate": 1.8003821464608515e-05, + "loss": 1.0984, + "step": 9403 + }, + { + "epoch": 0.61, + "learning_rate": 1.8003606775585564e-05, + "loss": 1.1061, + "step": 9404 + }, + { + "epoch": 0.61, + "learning_rate": 1.8003392086562616e-05, + "loss": 1.1915, + "step": 9405 + }, + { + "epoch": 0.61, + "learning_rate": 1.8003177397539665e-05, + "loss": 1.0774, + "step": 9406 + }, + { + "epoch": 0.61, + "learning_rate": 1.8002962708516717e-05, + "loss": 1.0893, + "step": 9407 + }, + { + "epoch": 0.61, + "learning_rate": 1.8002748019493766e-05, + "loss": 1.0793, + "step": 9408 + }, + { + "epoch": 0.61, + "learning_rate": 1.8002533330470815e-05, + "loss": 1.1111, + "step": 9409 + }, + { + "epoch": 0.61, + "learning_rate": 1.8002318641447864e-05, + "loss": 1.0849, + "step": 9410 + }, + { + "epoch": 0.61, + "learning_rate": 1.8002103952424913e-05, + "loss": 1.1098, + "step": 9411 + }, + { + "epoch": 0.61, + "learning_rate": 1.8001889263401962e-05, + "loss": 1.1876, + "step": 9412 + }, + { + "epoch": 0.61, + "learning_rate": 1.8001674574379014e-05, + "loss": 1.0156, + "step": 9413 + }, + { + "epoch": 0.61, + "learning_rate": 1.8001459885356063e-05, + "loss": 1.0365, + "step": 9414 + }, + { + "epoch": 0.61, + "learning_rate": 1.8001245196333112e-05, + "loss": 0.9735, + "step": 9415 + }, + { + "epoch": 0.61, + "learning_rate": 1.800103050731016e-05, + "loss": 1.0845, + "step": 9416 + }, + { + "epoch": 0.61, + "learning_rate": 1.800081581828721e-05, + "loss": 1.097, + "step": 9417 + }, + { + "epoch": 0.61, + "learning_rate": 1.8000601129264263e-05, + "loss": 1.087, + "step": 9418 + }, + { + "epoch": 0.61, + "learning_rate": 1.800038644024131e-05, + "loss": 1.0964, + "step": 9419 + }, + { + "epoch": 0.61, + "learning_rate": 1.8000171751218364e-05, + "loss": 1.0509, + "step": 9420 + }, + { + "epoch": 0.61, + "learning_rate": 1.7999957062195413e-05, + "loss": 1.108, + "step": 9421 + }, + { + "epoch": 0.61, + "learning_rate": 1.7999742373172462e-05, + "loss": 0.9957, + "step": 9422 + }, + { + "epoch": 0.61, + "learning_rate": 1.799952768414951e-05, + "loss": 1.0426, + "step": 9423 + }, + { + "epoch": 0.61, + "learning_rate": 1.799931299512656e-05, + "loss": 1.139, + "step": 9424 + }, + { + "epoch": 0.61, + "learning_rate": 1.7999098306103612e-05, + "loss": 1.1072, + "step": 9425 + }, + { + "epoch": 0.61, + "learning_rate": 1.799888361708066e-05, + "loss": 1.0548, + "step": 9426 + }, + { + "epoch": 0.61, + "learning_rate": 1.799866892805771e-05, + "loss": 1.0073, + "step": 9427 + }, + { + "epoch": 0.61, + "learning_rate": 1.799845423903476e-05, + "loss": 1.0804, + "step": 9428 + }, + { + "epoch": 0.61, + "learning_rate": 1.7998239550011808e-05, + "loss": 1.0514, + "step": 9429 + }, + { + "epoch": 0.61, + "learning_rate": 1.7998024860988857e-05, + "loss": 1.1087, + "step": 9430 + }, + { + "epoch": 0.61, + "learning_rate": 1.799781017196591e-05, + "loss": 1.1009, + "step": 9431 + }, + { + "epoch": 0.61, + "learning_rate": 1.799759548294296e-05, + "loss": 0.9958, + "step": 9432 + }, + { + "epoch": 0.61, + "learning_rate": 1.7997380793920007e-05, + "loss": 1.1085, + "step": 9433 + }, + { + "epoch": 0.61, + "learning_rate": 1.799716610489706e-05, + "loss": 1.0761, + "step": 9434 + }, + { + "epoch": 0.61, + "learning_rate": 1.799695141587411e-05, + "loss": 1.126, + "step": 9435 + }, + { + "epoch": 0.61, + "learning_rate": 1.7996736726851158e-05, + "loss": 1.0664, + "step": 9436 + }, + { + "epoch": 0.61, + "learning_rate": 1.7996522037828207e-05, + "loss": 1.0371, + "step": 9437 + }, + { + "epoch": 0.61, + "learning_rate": 1.799630734880526e-05, + "loss": 1.1221, + "step": 9438 + }, + { + "epoch": 0.61, + "learning_rate": 1.7996092659782308e-05, + "loss": 1.0049, + "step": 9439 + }, + { + "epoch": 0.61, + "learning_rate": 1.7995877970759357e-05, + "loss": 1.1178, + "step": 9440 + }, + { + "epoch": 0.61, + "learning_rate": 1.7995663281736406e-05, + "loss": 1.0942, + "step": 9441 + }, + { + "epoch": 0.61, + "learning_rate": 1.7995448592713455e-05, + "loss": 1.0277, + "step": 9442 + }, + { + "epoch": 0.61, + "learning_rate": 1.7995233903690507e-05, + "loss": 1.1524, + "step": 9443 + }, + { + "epoch": 0.61, + "learning_rate": 1.7995019214667556e-05, + "loss": 1.0845, + "step": 9444 + }, + { + "epoch": 0.61, + "learning_rate": 1.7994804525644605e-05, + "loss": 1.047, + "step": 9445 + }, + { + "epoch": 0.61, + "learning_rate": 1.7994589836621654e-05, + "loss": 1.0828, + "step": 9446 + }, + { + "epoch": 0.61, + "learning_rate": 1.7994375147598703e-05, + "loss": 1.0947, + "step": 9447 + }, + { + "epoch": 0.61, + "learning_rate": 1.7994160458575752e-05, + "loss": 1.0752, + "step": 9448 + }, + { + "epoch": 0.61, + "learning_rate": 1.7993945769552804e-05, + "loss": 1.0424, + "step": 9449 + }, + { + "epoch": 0.61, + "learning_rate": 1.7993731080529853e-05, + "loss": 1.0104, + "step": 9450 + }, + { + "epoch": 0.61, + "learning_rate": 1.7993516391506906e-05, + "loss": 1.0944, + "step": 9451 + }, + { + "epoch": 0.61, + "learning_rate": 1.7993301702483955e-05, + "loss": 1.0736, + "step": 9452 + }, + { + "epoch": 0.61, + "learning_rate": 1.7993087013461004e-05, + "loss": 1.0579, + "step": 9453 + }, + { + "epoch": 0.61, + "learning_rate": 1.7992872324438053e-05, + "loss": 1.0297, + "step": 9454 + }, + { + "epoch": 0.61, + "learning_rate": 1.79926576354151e-05, + "loss": 1.1011, + "step": 9455 + }, + { + "epoch": 0.61, + "learning_rate": 1.7992442946392154e-05, + "loss": 1.0669, + "step": 9456 + }, + { + "epoch": 0.61, + "learning_rate": 1.7992228257369203e-05, + "loss": 1.0877, + "step": 9457 + }, + { + "epoch": 0.61, + "learning_rate": 1.7992013568346252e-05, + "loss": 1.0481, + "step": 9458 + }, + { + "epoch": 0.61, + "learning_rate": 1.79917988793233e-05, + "loss": 1.0647, + "step": 9459 + }, + { + "epoch": 0.61, + "learning_rate": 1.799158419030035e-05, + "loss": 1.0637, + "step": 9460 + }, + { + "epoch": 0.61, + "learning_rate": 1.79913695012774e-05, + "loss": 1.1257, + "step": 9461 + }, + { + "epoch": 0.61, + "learning_rate": 1.799115481225445e-05, + "loss": 1.0846, + "step": 9462 + }, + { + "epoch": 0.61, + "learning_rate": 1.7990940123231503e-05, + "loss": 1.0409, + "step": 9463 + }, + { + "epoch": 0.61, + "learning_rate": 1.7990725434208552e-05, + "loss": 1.0315, + "step": 9464 + }, + { + "epoch": 0.61, + "learning_rate": 1.79905107451856e-05, + "loss": 0.9507, + "step": 9465 + }, + { + "epoch": 0.61, + "learning_rate": 1.799029605616265e-05, + "loss": 1.1544, + "step": 9466 + }, + { + "epoch": 0.61, + "learning_rate": 1.79900813671397e-05, + "loss": 1.1319, + "step": 9467 + }, + { + "epoch": 0.61, + "learning_rate": 1.7989866678116748e-05, + "loss": 1.0029, + "step": 9468 + }, + { + "epoch": 0.61, + "learning_rate": 1.79896519890938e-05, + "loss": 1.1443, + "step": 9469 + }, + { + "epoch": 0.61, + "learning_rate": 1.798943730007085e-05, + "loss": 1.1373, + "step": 9470 + }, + { + "epoch": 0.61, + "learning_rate": 1.79892226110479e-05, + "loss": 1.1101, + "step": 9471 + }, + { + "epoch": 0.61, + "learning_rate": 1.7989007922024947e-05, + "loss": 0.9992, + "step": 9472 + }, + { + "epoch": 0.61, + "learning_rate": 1.7988793233001996e-05, + "loss": 1.0283, + "step": 9473 + }, + { + "epoch": 0.61, + "learning_rate": 1.798857854397905e-05, + "loss": 1.0244, + "step": 9474 + }, + { + "epoch": 0.61, + "learning_rate": 1.7988363854956098e-05, + "loss": 1.003, + "step": 9475 + }, + { + "epoch": 0.61, + "learning_rate": 1.7988149165933147e-05, + "loss": 1.0421, + "step": 9476 + }, + { + "epoch": 0.61, + "learning_rate": 1.7987934476910196e-05, + "loss": 0.984, + "step": 9477 + }, + { + "epoch": 0.61, + "learning_rate": 1.7987719787887248e-05, + "loss": 1.0857, + "step": 9478 + }, + { + "epoch": 0.61, + "learning_rate": 1.7987505098864297e-05, + "loss": 1.0235, + "step": 9479 + }, + { + "epoch": 0.61, + "learning_rate": 1.7987290409841346e-05, + "loss": 1.08, + "step": 9480 + }, + { + "epoch": 0.61, + "learning_rate": 1.7987075720818395e-05, + "loss": 1.083, + "step": 9481 + }, + { + "epoch": 0.61, + "learning_rate": 1.7986861031795447e-05, + "loss": 1.0344, + "step": 9482 + }, + { + "epoch": 0.61, + "learning_rate": 1.7986646342772496e-05, + "loss": 0.9948, + "step": 9483 + }, + { + "epoch": 0.61, + "learning_rate": 1.7986431653749545e-05, + "loss": 1.0354, + "step": 9484 + }, + { + "epoch": 0.61, + "learning_rate": 1.7986216964726594e-05, + "loss": 1.093, + "step": 9485 + }, + { + "epoch": 0.61, + "learning_rate": 1.7986002275703643e-05, + "loss": 1.0028, + "step": 9486 + }, + { + "epoch": 0.61, + "learning_rate": 1.7985787586680695e-05, + "loss": 1.0551, + "step": 9487 + }, + { + "epoch": 0.61, + "learning_rate": 1.7985572897657744e-05, + "loss": 1.1326, + "step": 9488 + }, + { + "epoch": 0.61, + "learning_rate": 1.7985358208634793e-05, + "loss": 1.0694, + "step": 9489 + }, + { + "epoch": 0.61, + "learning_rate": 1.7985143519611842e-05, + "loss": 1.0242, + "step": 9490 + }, + { + "epoch": 0.61, + "learning_rate": 1.798492883058889e-05, + "loss": 1.0971, + "step": 9491 + }, + { + "epoch": 0.61, + "learning_rate": 1.7984714141565944e-05, + "loss": 1.0521, + "step": 9492 + }, + { + "epoch": 0.61, + "learning_rate": 1.7984499452542993e-05, + "loss": 1.0511, + "step": 9493 + }, + { + "epoch": 0.61, + "learning_rate": 1.7984284763520045e-05, + "loss": 1.0485, + "step": 9494 + }, + { + "epoch": 0.61, + "learning_rate": 1.7984070074497094e-05, + "loss": 1.0243, + "step": 9495 + }, + { + "epoch": 0.61, + "learning_rate": 1.7983855385474143e-05, + "loss": 1.0226, + "step": 9496 + }, + { + "epoch": 0.61, + "learning_rate": 1.7983640696451192e-05, + "loss": 1.1924, + "step": 9497 + }, + { + "epoch": 0.61, + "learning_rate": 1.798342600742824e-05, + "loss": 1.1679, + "step": 9498 + }, + { + "epoch": 0.61, + "learning_rate": 1.798321131840529e-05, + "loss": 1.021, + "step": 9499 + }, + { + "epoch": 0.61, + "learning_rate": 1.7982996629382342e-05, + "loss": 1.0918, + "step": 9500 + }, + { + "epoch": 0.61, + "learning_rate": 1.798278194035939e-05, + "loss": 1.0587, + "step": 9501 + }, + { + "epoch": 0.61, + "learning_rate": 1.798256725133644e-05, + "loss": 1.1048, + "step": 9502 + }, + { + "epoch": 0.61, + "learning_rate": 1.798235256231349e-05, + "loss": 1.0893, + "step": 9503 + }, + { + "epoch": 0.61, + "learning_rate": 1.7982137873290538e-05, + "loss": 1.0498, + "step": 9504 + }, + { + "epoch": 0.61, + "learning_rate": 1.798192318426759e-05, + "loss": 0.9638, + "step": 9505 + }, + { + "epoch": 0.61, + "learning_rate": 1.798170849524464e-05, + "loss": 1.0466, + "step": 9506 + }, + { + "epoch": 0.61, + "learning_rate": 1.7981493806221692e-05, + "loss": 1.0413, + "step": 9507 + }, + { + "epoch": 0.61, + "learning_rate": 1.798127911719874e-05, + "loss": 1.0542, + "step": 9508 + }, + { + "epoch": 0.61, + "learning_rate": 1.798106442817579e-05, + "loss": 1.0799, + "step": 9509 + }, + { + "epoch": 0.61, + "learning_rate": 1.798084973915284e-05, + "loss": 1.0543, + "step": 9510 + }, + { + "epoch": 0.61, + "learning_rate": 1.7980635050129888e-05, + "loss": 1.0828, + "step": 9511 + }, + { + "epoch": 0.61, + "learning_rate": 1.798042036110694e-05, + "loss": 1.0408, + "step": 9512 + }, + { + "epoch": 0.61, + "learning_rate": 1.798020567208399e-05, + "loss": 1.0568, + "step": 9513 + }, + { + "epoch": 0.61, + "learning_rate": 1.7979990983061038e-05, + "loss": 1.1599, + "step": 9514 + }, + { + "epoch": 0.61, + "learning_rate": 1.7979776294038087e-05, + "loss": 1.132, + "step": 9515 + }, + { + "epoch": 0.61, + "learning_rate": 1.7979561605015136e-05, + "loss": 1.0579, + "step": 9516 + }, + { + "epoch": 0.61, + "learning_rate": 1.7979346915992185e-05, + "loss": 1.0125, + "step": 9517 + }, + { + "epoch": 0.61, + "learning_rate": 1.7979132226969237e-05, + "loss": 1.0375, + "step": 9518 + }, + { + "epoch": 0.61, + "learning_rate": 1.7978917537946286e-05, + "loss": 1.0296, + "step": 9519 + }, + { + "epoch": 0.61, + "learning_rate": 1.7978702848923335e-05, + "loss": 1.0806, + "step": 9520 + }, + { + "epoch": 0.61, + "learning_rate": 1.7978488159900387e-05, + "loss": 1.1595, + "step": 9521 + }, + { + "epoch": 0.61, + "learning_rate": 1.7978273470877436e-05, + "loss": 1.0719, + "step": 9522 + }, + { + "epoch": 0.61, + "learning_rate": 1.7978058781854485e-05, + "loss": 1.0535, + "step": 9523 + }, + { + "epoch": 0.61, + "learning_rate": 1.7977844092831534e-05, + "loss": 1.0558, + "step": 9524 + }, + { + "epoch": 0.61, + "learning_rate": 1.7977629403808587e-05, + "loss": 1.075, + "step": 9525 + }, + { + "epoch": 0.61, + "learning_rate": 1.7977414714785636e-05, + "loss": 1.1011, + "step": 9526 + }, + { + "epoch": 0.61, + "learning_rate": 1.7977200025762685e-05, + "loss": 1.1183, + "step": 9527 + }, + { + "epoch": 0.61, + "learning_rate": 1.7976985336739734e-05, + "loss": 1.0251, + "step": 9528 + }, + { + "epoch": 0.61, + "learning_rate": 1.7976770647716782e-05, + "loss": 0.9797, + "step": 9529 + }, + { + "epoch": 0.61, + "learning_rate": 1.7976555958693835e-05, + "loss": 1.1217, + "step": 9530 + }, + { + "epoch": 0.61, + "learning_rate": 1.7976341269670884e-05, + "loss": 1.0778, + "step": 9531 + }, + { + "epoch": 0.61, + "learning_rate": 1.7976126580647933e-05, + "loss": 1.0807, + "step": 9532 + }, + { + "epoch": 0.61, + "learning_rate": 1.7975911891624982e-05, + "loss": 1.0826, + "step": 9533 + }, + { + "epoch": 0.61, + "learning_rate": 1.797569720260203e-05, + "loss": 1.1329, + "step": 9534 + }, + { + "epoch": 0.61, + "learning_rate": 1.797548251357908e-05, + "loss": 1.0623, + "step": 9535 + }, + { + "epoch": 0.61, + "learning_rate": 1.7975267824556132e-05, + "loss": 1.0428, + "step": 9536 + }, + { + "epoch": 0.61, + "learning_rate": 1.797505313553318e-05, + "loss": 1.1611, + "step": 9537 + }, + { + "epoch": 0.61, + "learning_rate": 1.7974838446510233e-05, + "loss": 1.13, + "step": 9538 + }, + { + "epoch": 0.61, + "learning_rate": 1.7974623757487282e-05, + "loss": 0.9777, + "step": 9539 + }, + { + "epoch": 0.61, + "learning_rate": 1.797440906846433e-05, + "loss": 1.1887, + "step": 9540 + }, + { + "epoch": 0.61, + "learning_rate": 1.797419437944138e-05, + "loss": 1.0383, + "step": 9541 + }, + { + "epoch": 0.61, + "learning_rate": 1.797397969041843e-05, + "loss": 1.0601, + "step": 9542 + }, + { + "epoch": 0.61, + "learning_rate": 1.797376500139548e-05, + "loss": 1.1193, + "step": 9543 + }, + { + "epoch": 0.61, + "learning_rate": 1.797355031237253e-05, + "loss": 1.1067, + "step": 9544 + }, + { + "epoch": 0.61, + "learning_rate": 1.797333562334958e-05, + "loss": 1.0913, + "step": 9545 + }, + { + "epoch": 0.61, + "learning_rate": 1.797312093432663e-05, + "loss": 1.1007, + "step": 9546 + }, + { + "epoch": 0.61, + "learning_rate": 1.7972906245303677e-05, + "loss": 1.1096, + "step": 9547 + }, + { + "epoch": 0.61, + "learning_rate": 1.7972691556280726e-05, + "loss": 1.054, + "step": 9548 + }, + { + "epoch": 0.61, + "learning_rate": 1.797247686725778e-05, + "loss": 1.0427, + "step": 9549 + }, + { + "epoch": 0.61, + "learning_rate": 1.7972262178234828e-05, + "loss": 1.0492, + "step": 9550 + }, + { + "epoch": 0.61, + "learning_rate": 1.797204748921188e-05, + "loss": 1.0851, + "step": 9551 + }, + { + "epoch": 0.61, + "learning_rate": 1.797183280018893e-05, + "loss": 0.9673, + "step": 9552 + }, + { + "epoch": 0.61, + "learning_rate": 1.7971618111165978e-05, + "loss": 1.0693, + "step": 9553 + }, + { + "epoch": 0.61, + "learning_rate": 1.7971403422143027e-05, + "loss": 1.0931, + "step": 9554 + }, + { + "epoch": 0.61, + "learning_rate": 1.7971188733120076e-05, + "loss": 1.0977, + "step": 9555 + }, + { + "epoch": 0.61, + "learning_rate": 1.7970974044097128e-05, + "loss": 1.1606, + "step": 9556 + }, + { + "epoch": 0.61, + "learning_rate": 1.7970759355074177e-05, + "loss": 1.159, + "step": 9557 + }, + { + "epoch": 0.61, + "learning_rate": 1.7970544666051226e-05, + "loss": 1.1275, + "step": 9558 + }, + { + "epoch": 0.62, + "learning_rate": 1.7970329977028275e-05, + "loss": 1.1611, + "step": 9559 + }, + { + "epoch": 0.62, + "learning_rate": 1.7970115288005324e-05, + "loss": 1.0673, + "step": 9560 + }, + { + "epoch": 0.62, + "learning_rate": 1.7969900598982376e-05, + "loss": 1.2469, + "step": 9561 + }, + { + "epoch": 0.62, + "learning_rate": 1.7969685909959425e-05, + "loss": 1.0243, + "step": 9562 + }, + { + "epoch": 0.62, + "learning_rate": 1.7969471220936474e-05, + "loss": 1.0803, + "step": 9563 + }, + { + "epoch": 0.62, + "learning_rate": 1.7969256531913523e-05, + "loss": 1.1635, + "step": 9564 + }, + { + "epoch": 0.62, + "learning_rate": 1.7969041842890576e-05, + "loss": 1.0655, + "step": 9565 + }, + { + "epoch": 0.62, + "learning_rate": 1.7968827153867625e-05, + "loss": 1.1129, + "step": 9566 + }, + { + "epoch": 0.62, + "learning_rate": 1.7968612464844674e-05, + "loss": 1.0778, + "step": 9567 + }, + { + "epoch": 0.62, + "learning_rate": 1.7968397775821723e-05, + "loss": 1.0569, + "step": 9568 + }, + { + "epoch": 0.62, + "learning_rate": 1.7968183086798775e-05, + "loss": 1.166, + "step": 9569 + }, + { + "epoch": 0.62, + "learning_rate": 1.7967968397775824e-05, + "loss": 1.0068, + "step": 9570 + }, + { + "epoch": 0.62, + "learning_rate": 1.7967753708752873e-05, + "loss": 1.118, + "step": 9571 + }, + { + "epoch": 0.62, + "learning_rate": 1.7967539019729922e-05, + "loss": 1.0236, + "step": 9572 + }, + { + "epoch": 0.62, + "learning_rate": 1.796732433070697e-05, + "loss": 1.0755, + "step": 9573 + }, + { + "epoch": 0.62, + "learning_rate": 1.7967109641684023e-05, + "loss": 1.1107, + "step": 9574 + }, + { + "epoch": 0.62, + "learning_rate": 1.7966894952661072e-05, + "loss": 1.095, + "step": 9575 + }, + { + "epoch": 0.62, + "learning_rate": 1.796668026363812e-05, + "loss": 1.0227, + "step": 9576 + }, + { + "epoch": 0.62, + "learning_rate": 1.796646557461517e-05, + "loss": 1.0453, + "step": 9577 + }, + { + "epoch": 0.62, + "learning_rate": 1.796625088559222e-05, + "loss": 1.0511, + "step": 9578 + }, + { + "epoch": 0.62, + "learning_rate": 1.796603619656927e-05, + "loss": 1.0971, + "step": 9579 + }, + { + "epoch": 0.62, + "learning_rate": 1.796582150754632e-05, + "loss": 1.1237, + "step": 9580 + }, + { + "epoch": 0.62, + "learning_rate": 1.7965606818523373e-05, + "loss": 1.1637, + "step": 9581 + }, + { + "epoch": 0.62, + "learning_rate": 1.796539212950042e-05, + "loss": 1.0331, + "step": 9582 + }, + { + "epoch": 0.62, + "learning_rate": 1.796517744047747e-05, + "loss": 1.1355, + "step": 9583 + }, + { + "epoch": 0.62, + "learning_rate": 1.796496275145452e-05, + "loss": 1.1082, + "step": 9584 + }, + { + "epoch": 0.62, + "learning_rate": 1.796474806243157e-05, + "loss": 1.0108, + "step": 9585 + }, + { + "epoch": 0.62, + "learning_rate": 1.7964533373408618e-05, + "loss": 1.1082, + "step": 9586 + }, + { + "epoch": 0.62, + "learning_rate": 1.796431868438567e-05, + "loss": 1.0412, + "step": 9587 + }, + { + "epoch": 0.62, + "learning_rate": 1.796410399536272e-05, + "loss": 1.0697, + "step": 9588 + }, + { + "epoch": 0.62, + "learning_rate": 1.7963889306339768e-05, + "loss": 1.0794, + "step": 9589 + }, + { + "epoch": 0.62, + "learning_rate": 1.7963674617316817e-05, + "loss": 1.0497, + "step": 9590 + }, + { + "epoch": 0.62, + "learning_rate": 1.7963459928293866e-05, + "loss": 1.1515, + "step": 9591 + }, + { + "epoch": 0.62, + "learning_rate": 1.7963245239270918e-05, + "loss": 1.1063, + "step": 9592 + }, + { + "epoch": 0.62, + "learning_rate": 1.7963030550247967e-05, + "loss": 1.0413, + "step": 9593 + }, + { + "epoch": 0.62, + "learning_rate": 1.796281586122502e-05, + "loss": 1.0992, + "step": 9594 + }, + { + "epoch": 0.62, + "learning_rate": 1.796260117220207e-05, + "loss": 1.0711, + "step": 9595 + }, + { + "epoch": 0.62, + "learning_rate": 1.7962386483179117e-05, + "loss": 1.1247, + "step": 9596 + }, + { + "epoch": 0.62, + "learning_rate": 1.7962171794156166e-05, + "loss": 1.0964, + "step": 9597 + }, + { + "epoch": 0.62, + "learning_rate": 1.7961957105133215e-05, + "loss": 1.1145, + "step": 9598 + }, + { + "epoch": 0.62, + "learning_rate": 1.7961742416110268e-05, + "loss": 1.0521, + "step": 9599 + }, + { + "epoch": 0.62, + "learning_rate": 1.7961527727087317e-05, + "loss": 1.0353, + "step": 9600 + }, + { + "epoch": 0.62, + "learning_rate": 1.7961313038064366e-05, + "loss": 1.0746, + "step": 9601 + }, + { + "epoch": 0.62, + "learning_rate": 1.7961098349041415e-05, + "loss": 1.0975, + "step": 9602 + }, + { + "epoch": 0.62, + "learning_rate": 1.7960883660018463e-05, + "loss": 1.1777, + "step": 9603 + }, + { + "epoch": 0.62, + "learning_rate": 1.7960668970995512e-05, + "loss": 1.1341, + "step": 9604 + }, + { + "epoch": 0.62, + "learning_rate": 1.7960454281972565e-05, + "loss": 1.0348, + "step": 9605 + }, + { + "epoch": 0.62, + "learning_rate": 1.7960239592949614e-05, + "loss": 1.1344, + "step": 9606 + }, + { + "epoch": 0.62, + "learning_rate": 1.7960024903926663e-05, + "loss": 1.197, + "step": 9607 + }, + { + "epoch": 0.62, + "learning_rate": 1.7959810214903712e-05, + "loss": 1.0801, + "step": 9608 + }, + { + "epoch": 0.62, + "learning_rate": 1.7959595525880764e-05, + "loss": 1.024, + "step": 9609 + }, + { + "epoch": 0.62, + "learning_rate": 1.7959380836857813e-05, + "loss": 0.9866, + "step": 9610 + }, + { + "epoch": 0.62, + "learning_rate": 1.7959166147834862e-05, + "loss": 1.024, + "step": 9611 + }, + { + "epoch": 0.62, + "learning_rate": 1.7958951458811914e-05, + "loss": 1.1076, + "step": 9612 + }, + { + "epoch": 0.62, + "learning_rate": 1.7958736769788963e-05, + "loss": 1.1086, + "step": 9613 + }, + { + "epoch": 0.62, + "learning_rate": 1.7958522080766012e-05, + "loss": 1.1163, + "step": 9614 + }, + { + "epoch": 0.62, + "learning_rate": 1.795830739174306e-05, + "loss": 0.9857, + "step": 9615 + }, + { + "epoch": 0.62, + "learning_rate": 1.795809270272011e-05, + "loss": 1.0918, + "step": 9616 + }, + { + "epoch": 0.62, + "learning_rate": 1.7957878013697163e-05, + "loss": 1.0381, + "step": 9617 + }, + { + "epoch": 0.62, + "learning_rate": 1.795766332467421e-05, + "loss": 1.0048, + "step": 9618 + }, + { + "epoch": 0.62, + "learning_rate": 1.795744863565126e-05, + "loss": 1.1097, + "step": 9619 + }, + { + "epoch": 0.62, + "learning_rate": 1.795723394662831e-05, + "loss": 1.1051, + "step": 9620 + }, + { + "epoch": 0.62, + "learning_rate": 1.795701925760536e-05, + "loss": 1.0692, + "step": 9621 + }, + { + "epoch": 0.62, + "learning_rate": 1.7956804568582407e-05, + "loss": 1.0879, + "step": 9622 + }, + { + "epoch": 0.62, + "learning_rate": 1.795658987955946e-05, + "loss": 0.9895, + "step": 9623 + }, + { + "epoch": 0.62, + "learning_rate": 1.795637519053651e-05, + "loss": 1.1204, + "step": 9624 + }, + { + "epoch": 0.62, + "learning_rate": 1.795616050151356e-05, + "loss": 1.0619, + "step": 9625 + }, + { + "epoch": 0.62, + "learning_rate": 1.795594581249061e-05, + "loss": 1.1585, + "step": 9626 + }, + { + "epoch": 0.62, + "learning_rate": 1.795573112346766e-05, + "loss": 1.1319, + "step": 9627 + }, + { + "epoch": 0.62, + "learning_rate": 1.7955516434444708e-05, + "loss": 1.1348, + "step": 9628 + }, + { + "epoch": 0.62, + "learning_rate": 1.7955301745421757e-05, + "loss": 1.0815, + "step": 9629 + }, + { + "epoch": 0.62, + "learning_rate": 1.795508705639881e-05, + "loss": 1.0068, + "step": 9630 + }, + { + "epoch": 0.62, + "learning_rate": 1.7954872367375858e-05, + "loss": 1.1003, + "step": 9631 + }, + { + "epoch": 0.62, + "learning_rate": 1.7954657678352907e-05, + "loss": 1.1391, + "step": 9632 + }, + { + "epoch": 0.62, + "learning_rate": 1.7954442989329956e-05, + "loss": 1.1186, + "step": 9633 + }, + { + "epoch": 0.62, + "learning_rate": 1.7954228300307005e-05, + "loss": 1.0666, + "step": 9634 + }, + { + "epoch": 0.62, + "learning_rate": 1.7954013611284054e-05, + "loss": 1.0936, + "step": 9635 + }, + { + "epoch": 0.62, + "learning_rate": 1.7953798922261106e-05, + "loss": 1.0642, + "step": 9636 + }, + { + "epoch": 0.62, + "learning_rate": 1.7953584233238155e-05, + "loss": 1.1999, + "step": 9637 + }, + { + "epoch": 0.62, + "learning_rate": 1.7953369544215208e-05, + "loss": 1.0895, + "step": 9638 + }, + { + "epoch": 0.62, + "learning_rate": 1.7953154855192257e-05, + "loss": 1.1067, + "step": 9639 + }, + { + "epoch": 0.62, + "learning_rate": 1.7952940166169306e-05, + "loss": 1.1297, + "step": 9640 + }, + { + "epoch": 0.62, + "learning_rate": 1.7952725477146355e-05, + "loss": 0.9948, + "step": 9641 + }, + { + "epoch": 0.62, + "learning_rate": 1.7952510788123404e-05, + "loss": 1.0852, + "step": 9642 + }, + { + "epoch": 0.62, + "learning_rate": 1.7952296099100456e-05, + "loss": 1.1455, + "step": 9643 + }, + { + "epoch": 0.62, + "learning_rate": 1.7952081410077505e-05, + "loss": 1.0678, + "step": 9644 + }, + { + "epoch": 0.62, + "learning_rate": 1.7951866721054554e-05, + "loss": 1.0838, + "step": 9645 + }, + { + "epoch": 0.62, + "learning_rate": 1.7951652032031603e-05, + "loss": 1.0502, + "step": 9646 + }, + { + "epoch": 0.62, + "learning_rate": 1.7951437343008652e-05, + "loss": 1.0189, + "step": 9647 + }, + { + "epoch": 0.62, + "learning_rate": 1.7951222653985704e-05, + "loss": 1.106, + "step": 9648 + }, + { + "epoch": 0.62, + "learning_rate": 1.7951007964962753e-05, + "loss": 1.0998, + "step": 9649 + }, + { + "epoch": 0.62, + "learning_rate": 1.7950793275939802e-05, + "loss": 1.0651, + "step": 9650 + }, + { + "epoch": 0.62, + "learning_rate": 1.795057858691685e-05, + "loss": 1.0063, + "step": 9651 + }, + { + "epoch": 0.62, + "learning_rate": 1.7950363897893903e-05, + "loss": 1.0716, + "step": 9652 + }, + { + "epoch": 0.62, + "learning_rate": 1.7950149208870952e-05, + "loss": 1.02, + "step": 9653 + }, + { + "epoch": 0.62, + "learning_rate": 1.7949934519848e-05, + "loss": 1.0948, + "step": 9654 + }, + { + "epoch": 0.62, + "learning_rate": 1.794971983082505e-05, + "loss": 1.0988, + "step": 9655 + }, + { + "epoch": 0.62, + "learning_rate": 1.7949505141802103e-05, + "loss": 1.0741, + "step": 9656 + }, + { + "epoch": 0.62, + "learning_rate": 1.794929045277915e-05, + "loss": 1.0952, + "step": 9657 + }, + { + "epoch": 0.62, + "learning_rate": 1.79490757637562e-05, + "loss": 1.0896, + "step": 9658 + }, + { + "epoch": 0.62, + "learning_rate": 1.794886107473325e-05, + "loss": 1.0195, + "step": 9659 + }, + { + "epoch": 0.62, + "learning_rate": 1.79486463857103e-05, + "loss": 1.0728, + "step": 9660 + }, + { + "epoch": 0.62, + "learning_rate": 1.794843169668735e-05, + "loss": 1.0483, + "step": 9661 + }, + { + "epoch": 0.62, + "learning_rate": 1.79482170076644e-05, + "loss": 1.1019, + "step": 9662 + }, + { + "epoch": 0.62, + "learning_rate": 1.794800231864145e-05, + "loss": 1.1161, + "step": 9663 + }, + { + "epoch": 0.62, + "learning_rate": 1.7947787629618498e-05, + "loss": 1.0083, + "step": 9664 + }, + { + "epoch": 0.62, + "learning_rate": 1.7947572940595547e-05, + "loss": 1.077, + "step": 9665 + }, + { + "epoch": 0.62, + "learning_rate": 1.79473582515726e-05, + "loss": 1.0949, + "step": 9666 + }, + { + "epoch": 0.62, + "learning_rate": 1.7947143562549648e-05, + "loss": 1.0124, + "step": 9667 + }, + { + "epoch": 0.62, + "learning_rate": 1.79469288735267e-05, + "loss": 1.1093, + "step": 9668 + }, + { + "epoch": 0.62, + "learning_rate": 1.794671418450375e-05, + "loss": 1.0588, + "step": 9669 + }, + { + "epoch": 0.62, + "learning_rate": 1.79464994954808e-05, + "loss": 1.1282, + "step": 9670 + }, + { + "epoch": 0.62, + "learning_rate": 1.7946284806457847e-05, + "loss": 1.0489, + "step": 9671 + }, + { + "epoch": 0.62, + "learning_rate": 1.7946070117434896e-05, + "loss": 1.1098, + "step": 9672 + }, + { + "epoch": 0.62, + "learning_rate": 1.7945855428411945e-05, + "loss": 1.0446, + "step": 9673 + }, + { + "epoch": 0.62, + "learning_rate": 1.7945640739388998e-05, + "loss": 1.0045, + "step": 9674 + }, + { + "epoch": 0.62, + "learning_rate": 1.7945426050366047e-05, + "loss": 1.1253, + "step": 9675 + }, + { + "epoch": 0.62, + "learning_rate": 1.7945211361343096e-05, + "loss": 0.9948, + "step": 9676 + }, + { + "epoch": 0.62, + "learning_rate": 1.7944996672320144e-05, + "loss": 1.1134, + "step": 9677 + }, + { + "epoch": 0.62, + "learning_rate": 1.7944781983297193e-05, + "loss": 1.0383, + "step": 9678 + }, + { + "epoch": 0.62, + "learning_rate": 1.7944567294274246e-05, + "loss": 1.1466, + "step": 9679 + }, + { + "epoch": 0.62, + "learning_rate": 1.7944352605251295e-05, + "loss": 1.1197, + "step": 9680 + }, + { + "epoch": 0.62, + "learning_rate": 1.7944137916228347e-05, + "loss": 1.0785, + "step": 9681 + }, + { + "epoch": 0.62, + "learning_rate": 1.7943923227205396e-05, + "loss": 1.075, + "step": 9682 + }, + { + "epoch": 0.62, + "learning_rate": 1.7943708538182445e-05, + "loss": 1.0765, + "step": 9683 + }, + { + "epoch": 0.62, + "learning_rate": 1.7943493849159494e-05, + "loss": 1.1335, + "step": 9684 + }, + { + "epoch": 0.62, + "learning_rate": 1.7943279160136543e-05, + "loss": 1.1535, + "step": 9685 + }, + { + "epoch": 0.62, + "learning_rate": 1.7943064471113595e-05, + "loss": 1.0575, + "step": 9686 + }, + { + "epoch": 0.62, + "learning_rate": 1.7942849782090644e-05, + "loss": 1.0618, + "step": 9687 + }, + { + "epoch": 0.62, + "learning_rate": 1.7942635093067693e-05, + "loss": 1.0651, + "step": 9688 + }, + { + "epoch": 0.62, + "learning_rate": 1.7942420404044742e-05, + "loss": 1.0772, + "step": 9689 + }, + { + "epoch": 0.62, + "learning_rate": 1.794220571502179e-05, + "loss": 1.1447, + "step": 9690 + }, + { + "epoch": 0.62, + "learning_rate": 1.794199102599884e-05, + "loss": 1.1684, + "step": 9691 + }, + { + "epoch": 0.62, + "learning_rate": 1.7941776336975893e-05, + "loss": 1.0661, + "step": 9692 + }, + { + "epoch": 0.62, + "learning_rate": 1.794156164795294e-05, + "loss": 1.1097, + "step": 9693 + }, + { + "epoch": 0.62, + "learning_rate": 1.794134695892999e-05, + "loss": 1.044, + "step": 9694 + }, + { + "epoch": 0.62, + "learning_rate": 1.794113226990704e-05, + "loss": 1.1283, + "step": 9695 + }, + { + "epoch": 0.62, + "learning_rate": 1.7940917580884092e-05, + "loss": 1.0537, + "step": 9696 + }, + { + "epoch": 0.62, + "learning_rate": 1.794070289186114e-05, + "loss": 0.973, + "step": 9697 + }, + { + "epoch": 0.62, + "learning_rate": 1.794048820283819e-05, + "loss": 1.0104, + "step": 9698 + }, + { + "epoch": 0.62, + "learning_rate": 1.7940273513815242e-05, + "loss": 1.054, + "step": 9699 + }, + { + "epoch": 0.62, + "learning_rate": 1.794005882479229e-05, + "loss": 1.1112, + "step": 9700 + }, + { + "epoch": 0.62, + "learning_rate": 1.793984413576934e-05, + "loss": 1.1083, + "step": 9701 + }, + { + "epoch": 0.62, + "learning_rate": 1.793962944674639e-05, + "loss": 1.0418, + "step": 9702 + }, + { + "epoch": 0.62, + "learning_rate": 1.7939414757723438e-05, + "loss": 1.1002, + "step": 9703 + }, + { + "epoch": 0.62, + "learning_rate": 1.793920006870049e-05, + "loss": 1.0648, + "step": 9704 + }, + { + "epoch": 0.62, + "learning_rate": 1.793898537967754e-05, + "loss": 1.0615, + "step": 9705 + }, + { + "epoch": 0.62, + "learning_rate": 1.7938770690654588e-05, + "loss": 1.1405, + "step": 9706 + }, + { + "epoch": 0.62, + "learning_rate": 1.7938556001631637e-05, + "loss": 1.1191, + "step": 9707 + }, + { + "epoch": 0.62, + "learning_rate": 1.7938341312608686e-05, + "loss": 1.0022, + "step": 9708 + }, + { + "epoch": 0.62, + "learning_rate": 1.7938126623585735e-05, + "loss": 1.0611, + "step": 9709 + }, + { + "epoch": 0.62, + "learning_rate": 1.7937911934562787e-05, + "loss": 1.1023, + "step": 9710 + }, + { + "epoch": 0.62, + "learning_rate": 1.7937697245539836e-05, + "loss": 1.0326, + "step": 9711 + }, + { + "epoch": 0.62, + "learning_rate": 1.793748255651689e-05, + "loss": 1.0075, + "step": 9712 + }, + { + "epoch": 0.62, + "learning_rate": 1.7937267867493938e-05, + "loss": 1.085, + "step": 9713 + }, + { + "epoch": 0.62, + "learning_rate": 1.7937053178470987e-05, + "loss": 1.1196, + "step": 9714 + }, + { + "epoch": 0.63, + "learning_rate": 1.7936838489448036e-05, + "loss": 1.1129, + "step": 9715 + }, + { + "epoch": 0.63, + "learning_rate": 1.7936623800425085e-05, + "loss": 1.0767, + "step": 9716 + }, + { + "epoch": 0.63, + "learning_rate": 1.7936409111402137e-05, + "loss": 1.1183, + "step": 9717 + }, + { + "epoch": 0.63, + "learning_rate": 1.7936194422379186e-05, + "loss": 1.1128, + "step": 9718 + }, + { + "epoch": 0.63, + "learning_rate": 1.7935979733356235e-05, + "loss": 1.1248, + "step": 9719 + }, + { + "epoch": 0.63, + "learning_rate": 1.7935765044333284e-05, + "loss": 1.0741, + "step": 9720 + }, + { + "epoch": 0.63, + "learning_rate": 1.7935550355310333e-05, + "loss": 1.0901, + "step": 9721 + }, + { + "epoch": 0.63, + "learning_rate": 1.7935335666287382e-05, + "loss": 1.0641, + "step": 9722 + }, + { + "epoch": 0.63, + "learning_rate": 1.7935120977264434e-05, + "loss": 1.1171, + "step": 9723 + }, + { + "epoch": 0.63, + "learning_rate": 1.7934906288241483e-05, + "loss": 1.05, + "step": 9724 + }, + { + "epoch": 0.63, + "learning_rate": 1.7934691599218535e-05, + "loss": 1.114, + "step": 9725 + }, + { + "epoch": 0.63, + "learning_rate": 1.7934476910195584e-05, + "loss": 1.1215, + "step": 9726 + }, + { + "epoch": 0.63, + "learning_rate": 1.7934262221172633e-05, + "loss": 1.0568, + "step": 9727 + }, + { + "epoch": 0.63, + "learning_rate": 1.7934047532149682e-05, + "loss": 0.9999, + "step": 9728 + }, + { + "epoch": 0.63, + "learning_rate": 1.793383284312673e-05, + "loss": 1.1287, + "step": 9729 + }, + { + "epoch": 0.63, + "learning_rate": 1.7933618154103784e-05, + "loss": 1.068, + "step": 9730 + }, + { + "epoch": 0.63, + "learning_rate": 1.7933403465080833e-05, + "loss": 1.0648, + "step": 9731 + }, + { + "epoch": 0.63, + "learning_rate": 1.793318877605788e-05, + "loss": 1.0952, + "step": 9732 + }, + { + "epoch": 0.63, + "learning_rate": 1.793297408703493e-05, + "loss": 1.118, + "step": 9733 + }, + { + "epoch": 0.63, + "learning_rate": 1.793275939801198e-05, + "loss": 1.0524, + "step": 9734 + }, + { + "epoch": 0.63, + "learning_rate": 1.7932544708989032e-05, + "loss": 1.1115, + "step": 9735 + }, + { + "epoch": 0.63, + "learning_rate": 1.793233001996608e-05, + "loss": 1.0001, + "step": 9736 + }, + { + "epoch": 0.63, + "learning_rate": 1.793211533094313e-05, + "loss": 1.0728, + "step": 9737 + }, + { + "epoch": 0.63, + "learning_rate": 1.793190064192018e-05, + "loss": 1.0427, + "step": 9738 + }, + { + "epoch": 0.63, + "learning_rate": 1.793168595289723e-05, + "loss": 1.1531, + "step": 9739 + }, + { + "epoch": 0.63, + "learning_rate": 1.793147126387428e-05, + "loss": 1.0473, + "step": 9740 + }, + { + "epoch": 0.63, + "learning_rate": 1.793125657485133e-05, + "loss": 1.1538, + "step": 9741 + }, + { + "epoch": 0.63, + "learning_rate": 1.7931041885828378e-05, + "loss": 1.0445, + "step": 9742 + }, + { + "epoch": 0.63, + "learning_rate": 1.793082719680543e-05, + "loss": 1.02, + "step": 9743 + }, + { + "epoch": 0.63, + "learning_rate": 1.793061250778248e-05, + "loss": 1.0463, + "step": 9744 + }, + { + "epoch": 0.63, + "learning_rate": 1.793039781875953e-05, + "loss": 1.0613, + "step": 9745 + }, + { + "epoch": 0.63, + "learning_rate": 1.7930183129736577e-05, + "loss": 1.1236, + "step": 9746 + }, + { + "epoch": 0.63, + "learning_rate": 1.7929968440713626e-05, + "loss": 1.1781, + "step": 9747 + }, + { + "epoch": 0.63, + "learning_rate": 1.792975375169068e-05, + "loss": 1.0916, + "step": 9748 + }, + { + "epoch": 0.63, + "learning_rate": 1.7929539062667728e-05, + "loss": 1.1158, + "step": 9749 + }, + { + "epoch": 0.63, + "learning_rate": 1.7929324373644777e-05, + "loss": 1.0143, + "step": 9750 + }, + { + "epoch": 0.63, + "learning_rate": 1.7929109684621825e-05, + "loss": 1.0772, + "step": 9751 + }, + { + "epoch": 0.63, + "learning_rate": 1.7928894995598874e-05, + "loss": 1.0588, + "step": 9752 + }, + { + "epoch": 0.63, + "learning_rate": 1.7928680306575927e-05, + "loss": 1.0413, + "step": 9753 + }, + { + "epoch": 0.63, + "learning_rate": 1.7928465617552976e-05, + "loss": 1.0841, + "step": 9754 + }, + { + "epoch": 0.63, + "learning_rate": 1.7928250928530028e-05, + "loss": 1.0277, + "step": 9755 + }, + { + "epoch": 0.63, + "learning_rate": 1.7928036239507077e-05, + "loss": 1.0736, + "step": 9756 + }, + { + "epoch": 0.63, + "learning_rate": 1.7927821550484126e-05, + "loss": 1.0894, + "step": 9757 + }, + { + "epoch": 0.63, + "learning_rate": 1.7927606861461175e-05, + "loss": 1.0917, + "step": 9758 + }, + { + "epoch": 0.63, + "learning_rate": 1.7927392172438224e-05, + "loss": 1.0498, + "step": 9759 + }, + { + "epoch": 0.63, + "learning_rate": 1.7927177483415273e-05, + "loss": 1.0764, + "step": 9760 + }, + { + "epoch": 0.63, + "learning_rate": 1.7926962794392325e-05, + "loss": 0.9146, + "step": 9761 + }, + { + "epoch": 0.63, + "learning_rate": 1.7926748105369374e-05, + "loss": 1.0427, + "step": 9762 + }, + { + "epoch": 0.63, + "learning_rate": 1.7926533416346423e-05, + "loss": 1.0021, + "step": 9763 + }, + { + "epoch": 0.63, + "learning_rate": 1.7926318727323472e-05, + "loss": 1.0292, + "step": 9764 + }, + { + "epoch": 0.63, + "learning_rate": 1.792610403830052e-05, + "loss": 1.0836, + "step": 9765 + }, + { + "epoch": 0.63, + "learning_rate": 1.7925889349277574e-05, + "loss": 1.0919, + "step": 9766 + }, + { + "epoch": 0.63, + "learning_rate": 1.7925674660254622e-05, + "loss": 1.1675, + "step": 9767 + }, + { + "epoch": 0.63, + "learning_rate": 1.7925459971231675e-05, + "loss": 1.1057, + "step": 9768 + }, + { + "epoch": 0.63, + "learning_rate": 1.7925245282208724e-05, + "loss": 1.1041, + "step": 9769 + }, + { + "epoch": 0.63, + "learning_rate": 1.7925030593185773e-05, + "loss": 1.1042, + "step": 9770 + }, + { + "epoch": 0.63, + "learning_rate": 1.7924815904162822e-05, + "loss": 1.0906, + "step": 9771 + }, + { + "epoch": 0.63, + "learning_rate": 1.792460121513987e-05, + "loss": 1.0071, + "step": 9772 + }, + { + "epoch": 0.63, + "learning_rate": 1.7924386526116923e-05, + "loss": 1.0762, + "step": 9773 + }, + { + "epoch": 0.63, + "learning_rate": 1.7924171837093972e-05, + "loss": 1.0718, + "step": 9774 + }, + { + "epoch": 0.63, + "learning_rate": 1.792395714807102e-05, + "loss": 1.1134, + "step": 9775 + }, + { + "epoch": 0.63, + "learning_rate": 1.792374245904807e-05, + "loss": 1.0494, + "step": 9776 + }, + { + "epoch": 0.63, + "learning_rate": 1.792352777002512e-05, + "loss": 1.1099, + "step": 9777 + }, + { + "epoch": 0.63, + "learning_rate": 1.7923313081002168e-05, + "loss": 1.0978, + "step": 9778 + }, + { + "epoch": 0.63, + "learning_rate": 1.792309839197922e-05, + "loss": 1.0792, + "step": 9779 + }, + { + "epoch": 0.63, + "learning_rate": 1.792288370295627e-05, + "loss": 1.1023, + "step": 9780 + }, + { + "epoch": 0.63, + "learning_rate": 1.7922669013933318e-05, + "loss": 1.0827, + "step": 9781 + }, + { + "epoch": 0.63, + "learning_rate": 1.7922454324910367e-05, + "loss": 1.1785, + "step": 9782 + }, + { + "epoch": 0.63, + "learning_rate": 1.792223963588742e-05, + "loss": 1.1292, + "step": 9783 + }, + { + "epoch": 0.63, + "learning_rate": 1.792202494686447e-05, + "loss": 1.0095, + "step": 9784 + }, + { + "epoch": 0.63, + "learning_rate": 1.7921810257841517e-05, + "loss": 1.0151, + "step": 9785 + }, + { + "epoch": 0.63, + "learning_rate": 1.792159556881857e-05, + "loss": 1.0998, + "step": 9786 + }, + { + "epoch": 0.63, + "learning_rate": 1.792138087979562e-05, + "loss": 0.9612, + "step": 9787 + }, + { + "epoch": 0.63, + "learning_rate": 1.7921166190772668e-05, + "loss": 1.1048, + "step": 9788 + }, + { + "epoch": 0.63, + "learning_rate": 1.7920951501749717e-05, + "loss": 1.0057, + "step": 9789 + }, + { + "epoch": 0.63, + "learning_rate": 1.7920736812726766e-05, + "loss": 1.0699, + "step": 9790 + }, + { + "epoch": 0.63, + "learning_rate": 1.7920522123703815e-05, + "loss": 1.0671, + "step": 9791 + }, + { + "epoch": 0.63, + "learning_rate": 1.7920307434680867e-05, + "loss": 1.0654, + "step": 9792 + }, + { + "epoch": 0.63, + "learning_rate": 1.7920092745657916e-05, + "loss": 1.1094, + "step": 9793 + }, + { + "epoch": 0.63, + "learning_rate": 1.7919878056634965e-05, + "loss": 1.0001, + "step": 9794 + }, + { + "epoch": 0.63, + "learning_rate": 1.7919663367612014e-05, + "loss": 0.9997, + "step": 9795 + }, + { + "epoch": 0.63, + "learning_rate": 1.7919448678589063e-05, + "loss": 1.0888, + "step": 9796 + }, + { + "epoch": 0.63, + "learning_rate": 1.7919233989566115e-05, + "loss": 1.134, + "step": 9797 + }, + { + "epoch": 0.63, + "learning_rate": 1.7919019300543164e-05, + "loss": 1.1282, + "step": 9798 + }, + { + "epoch": 0.63, + "learning_rate": 1.7918804611520216e-05, + "loss": 1.1582, + "step": 9799 + }, + { + "epoch": 0.63, + "learning_rate": 1.7918589922497265e-05, + "loss": 1.1096, + "step": 9800 + }, + { + "epoch": 0.63, + "learning_rate": 1.7918375233474314e-05, + "loss": 1.2449, + "step": 9801 + }, + { + "epoch": 0.63, + "learning_rate": 1.7918160544451363e-05, + "loss": 1.0853, + "step": 9802 + }, + { + "epoch": 0.63, + "learning_rate": 1.7917945855428412e-05, + "loss": 1.1056, + "step": 9803 + }, + { + "epoch": 0.63, + "learning_rate": 1.7917731166405465e-05, + "loss": 1.1531, + "step": 9804 + }, + { + "epoch": 0.63, + "learning_rate": 1.7917516477382514e-05, + "loss": 1.0128, + "step": 9805 + }, + { + "epoch": 0.63, + "learning_rate": 1.7917301788359563e-05, + "loss": 1.1626, + "step": 9806 + }, + { + "epoch": 0.63, + "learning_rate": 1.791708709933661e-05, + "loss": 1.0249, + "step": 9807 + }, + { + "epoch": 0.63, + "learning_rate": 1.791687241031366e-05, + "loss": 1.0458, + "step": 9808 + }, + { + "epoch": 0.63, + "learning_rate": 1.791665772129071e-05, + "loss": 1.1092, + "step": 9809 + }, + { + "epoch": 0.63, + "learning_rate": 1.7916443032267762e-05, + "loss": 1.0526, + "step": 9810 + }, + { + "epoch": 0.63, + "learning_rate": 1.791622834324481e-05, + "loss": 1.0212, + "step": 9811 + }, + { + "epoch": 0.63, + "learning_rate": 1.7916013654221863e-05, + "loss": 0.9984, + "step": 9812 + }, + { + "epoch": 0.63, + "learning_rate": 1.7915798965198912e-05, + "loss": 1.123, + "step": 9813 + }, + { + "epoch": 0.63, + "learning_rate": 1.791558427617596e-05, + "loss": 1.0116, + "step": 9814 + }, + { + "epoch": 0.63, + "learning_rate": 1.791536958715301e-05, + "loss": 1.1259, + "step": 9815 + }, + { + "epoch": 0.63, + "learning_rate": 1.791515489813006e-05, + "loss": 1.0212, + "step": 9816 + }, + { + "epoch": 0.63, + "learning_rate": 1.791494020910711e-05, + "loss": 1.0076, + "step": 9817 + }, + { + "epoch": 0.63, + "learning_rate": 1.791472552008416e-05, + "loss": 1.0504, + "step": 9818 + }, + { + "epoch": 0.63, + "learning_rate": 1.791451083106121e-05, + "loss": 0.9625, + "step": 9819 + }, + { + "epoch": 0.63, + "learning_rate": 1.7914296142038258e-05, + "loss": 1.0446, + "step": 9820 + }, + { + "epoch": 0.63, + "learning_rate": 1.7914081453015307e-05, + "loss": 1.056, + "step": 9821 + }, + { + "epoch": 0.63, + "learning_rate": 1.791386676399236e-05, + "loss": 1.1859, + "step": 9822 + }, + { + "epoch": 0.63, + "learning_rate": 1.791365207496941e-05, + "loss": 1.0781, + "step": 9823 + }, + { + "epoch": 0.63, + "learning_rate": 1.7913437385946458e-05, + "loss": 1.05, + "step": 9824 + }, + { + "epoch": 0.63, + "learning_rate": 1.7913222696923507e-05, + "loss": 1.0141, + "step": 9825 + }, + { + "epoch": 0.63, + "learning_rate": 1.791300800790056e-05, + "loss": 1.0943, + "step": 9826 + }, + { + "epoch": 0.63, + "learning_rate": 1.7912793318877608e-05, + "loss": 1.1473, + "step": 9827 + }, + { + "epoch": 0.63, + "learning_rate": 1.7912578629854657e-05, + "loss": 1.0165, + "step": 9828 + }, + { + "epoch": 0.63, + "learning_rate": 1.7912363940831706e-05, + "loss": 1.1494, + "step": 9829 + }, + { + "epoch": 0.63, + "learning_rate": 1.7912149251808758e-05, + "loss": 1.0313, + "step": 9830 + }, + { + "epoch": 0.63, + "learning_rate": 1.7911934562785807e-05, + "loss": 1.0201, + "step": 9831 + }, + { + "epoch": 0.63, + "learning_rate": 1.7911719873762856e-05, + "loss": 1.0448, + "step": 9832 + }, + { + "epoch": 0.63, + "learning_rate": 1.7911505184739905e-05, + "loss": 1.0562, + "step": 9833 + }, + { + "epoch": 0.63, + "learning_rate": 1.7911290495716954e-05, + "loss": 1.0776, + "step": 9834 + }, + { + "epoch": 0.63, + "learning_rate": 1.7911075806694006e-05, + "loss": 1.0573, + "step": 9835 + }, + { + "epoch": 0.63, + "learning_rate": 1.7910861117671055e-05, + "loss": 1.1387, + "step": 9836 + }, + { + "epoch": 0.63, + "learning_rate": 1.7910646428648104e-05, + "loss": 1.1109, + "step": 9837 + }, + { + "epoch": 0.63, + "learning_rate": 1.7910431739625153e-05, + "loss": 1.0622, + "step": 9838 + }, + { + "epoch": 0.63, + "learning_rate": 1.7910217050602202e-05, + "loss": 1.0163, + "step": 9839 + }, + { + "epoch": 0.63, + "learning_rate": 1.7910002361579255e-05, + "loss": 1.0414, + "step": 9840 + }, + { + "epoch": 0.63, + "learning_rate": 1.7909787672556303e-05, + "loss": 0.992, + "step": 9841 + }, + { + "epoch": 0.63, + "learning_rate": 1.7909572983533356e-05, + "loss": 1.0256, + "step": 9842 + }, + { + "epoch": 0.63, + "learning_rate": 1.7909358294510405e-05, + "loss": 1.0618, + "step": 9843 + }, + { + "epoch": 0.63, + "learning_rate": 1.7909143605487454e-05, + "loss": 1.0528, + "step": 9844 + }, + { + "epoch": 0.63, + "learning_rate": 1.7908928916464503e-05, + "loss": 1.0542, + "step": 9845 + }, + { + "epoch": 0.63, + "learning_rate": 1.7908714227441552e-05, + "loss": 1.042, + "step": 9846 + }, + { + "epoch": 0.63, + "learning_rate": 1.79084995384186e-05, + "loss": 1.1023, + "step": 9847 + }, + { + "epoch": 0.63, + "learning_rate": 1.7908284849395653e-05, + "loss": 0.9787, + "step": 9848 + }, + { + "epoch": 0.63, + "learning_rate": 1.7908070160372702e-05, + "loss": 1.1048, + "step": 9849 + }, + { + "epoch": 0.63, + "learning_rate": 1.790785547134975e-05, + "loss": 1.0649, + "step": 9850 + }, + { + "epoch": 0.63, + "learning_rate": 1.79076407823268e-05, + "loss": 1.0562, + "step": 9851 + }, + { + "epoch": 0.63, + "learning_rate": 1.790742609330385e-05, + "loss": 1.0638, + "step": 9852 + }, + { + "epoch": 0.63, + "learning_rate": 1.79072114042809e-05, + "loss": 1.0734, + "step": 9853 + }, + { + "epoch": 0.63, + "learning_rate": 1.790699671525795e-05, + "loss": 1.0564, + "step": 9854 + }, + { + "epoch": 0.63, + "learning_rate": 1.7906782026235003e-05, + "loss": 1.09, + "step": 9855 + }, + { + "epoch": 0.63, + "learning_rate": 1.790656733721205e-05, + "loss": 1.0301, + "step": 9856 + }, + { + "epoch": 0.63, + "learning_rate": 1.79063526481891e-05, + "loss": 1.1573, + "step": 9857 + }, + { + "epoch": 0.63, + "learning_rate": 1.790613795916615e-05, + "loss": 1.0569, + "step": 9858 + }, + { + "epoch": 0.63, + "learning_rate": 1.79059232701432e-05, + "loss": 1.1262, + "step": 9859 + }, + { + "epoch": 0.63, + "learning_rate": 1.790570858112025e-05, + "loss": 1.1404, + "step": 9860 + }, + { + "epoch": 0.63, + "learning_rate": 1.79054938920973e-05, + "loss": 0.9931, + "step": 9861 + }, + { + "epoch": 0.63, + "learning_rate": 1.790527920307435e-05, + "loss": 1.0996, + "step": 9862 + }, + { + "epoch": 0.63, + "learning_rate": 1.7905064514051398e-05, + "loss": 1.1592, + "step": 9863 + }, + { + "epoch": 0.63, + "learning_rate": 1.7904849825028447e-05, + "loss": 1.0549, + "step": 9864 + }, + { + "epoch": 0.63, + "learning_rate": 1.7904635136005496e-05, + "loss": 1.1306, + "step": 9865 + }, + { + "epoch": 0.63, + "learning_rate": 1.7904420446982548e-05, + "loss": 1.1392, + "step": 9866 + }, + { + "epoch": 0.63, + "learning_rate": 1.7904205757959597e-05, + "loss": 1.1343, + "step": 9867 + }, + { + "epoch": 0.63, + "learning_rate": 1.7903991068936646e-05, + "loss": 1.1332, + "step": 9868 + }, + { + "epoch": 0.63, + "learning_rate": 1.7903776379913695e-05, + "loss": 1.1346, + "step": 9869 + }, + { + "epoch": 0.64, + "learning_rate": 1.7903561690890747e-05, + "loss": 1.0277, + "step": 9870 + }, + { + "epoch": 0.64, + "learning_rate": 1.7903347001867796e-05, + "loss": 1.0944, + "step": 9871 + }, + { + "epoch": 0.64, + "learning_rate": 1.7903132312844845e-05, + "loss": 1.0546, + "step": 9872 + }, + { + "epoch": 0.64, + "learning_rate": 1.7902917623821897e-05, + "loss": 1.0485, + "step": 9873 + }, + { + "epoch": 0.64, + "learning_rate": 1.7902702934798946e-05, + "loss": 1.0664, + "step": 9874 + }, + { + "epoch": 0.64, + "learning_rate": 1.7902488245775995e-05, + "loss": 1.0126, + "step": 9875 + }, + { + "epoch": 0.64, + "learning_rate": 1.7902273556753044e-05, + "loss": 1.0745, + "step": 9876 + }, + { + "epoch": 0.64, + "learning_rate": 1.7902058867730093e-05, + "loss": 1.1046, + "step": 9877 + }, + { + "epoch": 0.64, + "learning_rate": 1.7901844178707142e-05, + "loss": 1.0697, + "step": 9878 + }, + { + "epoch": 0.64, + "learning_rate": 1.7901629489684195e-05, + "loss": 1.1134, + "step": 9879 + }, + { + "epoch": 0.64, + "learning_rate": 1.7901414800661244e-05, + "loss": 1.0726, + "step": 9880 + }, + { + "epoch": 0.64, + "learning_rate": 1.7901200111638293e-05, + "loss": 1.0718, + "step": 9881 + }, + { + "epoch": 0.64, + "learning_rate": 1.790098542261534e-05, + "loss": 1.0822, + "step": 9882 + }, + { + "epoch": 0.64, + "learning_rate": 1.790077073359239e-05, + "loss": 1.0794, + "step": 9883 + }, + { + "epoch": 0.64, + "learning_rate": 1.7900556044569443e-05, + "loss": 1.0595, + "step": 9884 + }, + { + "epoch": 0.64, + "learning_rate": 1.7900341355546492e-05, + "loss": 1.1485, + "step": 9885 + }, + { + "epoch": 0.64, + "learning_rate": 1.7900126666523544e-05, + "loss": 1.0024, + "step": 9886 + }, + { + "epoch": 0.64, + "learning_rate": 1.7899911977500593e-05, + "loss": 1.0489, + "step": 9887 + }, + { + "epoch": 0.64, + "learning_rate": 1.7899697288477642e-05, + "loss": 1.0308, + "step": 9888 + }, + { + "epoch": 0.64, + "learning_rate": 1.789948259945469e-05, + "loss": 1.0452, + "step": 9889 + }, + { + "epoch": 0.64, + "learning_rate": 1.789926791043174e-05, + "loss": 1.1354, + "step": 9890 + }, + { + "epoch": 0.64, + "learning_rate": 1.7899053221408792e-05, + "loss": 1.1336, + "step": 9891 + }, + { + "epoch": 0.64, + "learning_rate": 1.789883853238584e-05, + "loss": 1.1357, + "step": 9892 + }, + { + "epoch": 0.64, + "learning_rate": 1.789862384336289e-05, + "loss": 1.1234, + "step": 9893 + }, + { + "epoch": 0.64, + "learning_rate": 1.789840915433994e-05, + "loss": 1.1075, + "step": 9894 + }, + { + "epoch": 0.64, + "learning_rate": 1.7898194465316988e-05, + "loss": 1.1265, + "step": 9895 + }, + { + "epoch": 0.64, + "learning_rate": 1.7897979776294037e-05, + "loss": 1.0686, + "step": 9896 + }, + { + "epoch": 0.64, + "learning_rate": 1.789776508727109e-05, + "loss": 1.137, + "step": 9897 + }, + { + "epoch": 0.64, + "learning_rate": 1.789755039824814e-05, + "loss": 1.1286, + "step": 9898 + }, + { + "epoch": 0.64, + "learning_rate": 1.789733570922519e-05, + "loss": 1.0949, + "step": 9899 + }, + { + "epoch": 0.64, + "learning_rate": 1.789712102020224e-05, + "loss": 1.1134, + "step": 9900 + }, + { + "epoch": 0.64, + "learning_rate": 1.789690633117929e-05, + "loss": 1.0538, + "step": 9901 + }, + { + "epoch": 0.64, + "learning_rate": 1.7896691642156338e-05, + "loss": 0.999, + "step": 9902 + }, + { + "epoch": 0.64, + "learning_rate": 1.7896476953133387e-05, + "loss": 1.0568, + "step": 9903 + }, + { + "epoch": 0.64, + "learning_rate": 1.789626226411044e-05, + "loss": 1.1226, + "step": 9904 + }, + { + "epoch": 0.64, + "learning_rate": 1.7896047575087488e-05, + "loss": 1.0928, + "step": 9905 + }, + { + "epoch": 0.64, + "learning_rate": 1.7895832886064537e-05, + "loss": 1.0518, + "step": 9906 + }, + { + "epoch": 0.64, + "learning_rate": 1.7895618197041586e-05, + "loss": 1.1406, + "step": 9907 + }, + { + "epoch": 0.64, + "learning_rate": 1.7895403508018635e-05, + "loss": 1.0914, + "step": 9908 + }, + { + "epoch": 0.64, + "learning_rate": 1.7895188818995687e-05, + "loss": 1.0526, + "step": 9909 + }, + { + "epoch": 0.64, + "learning_rate": 1.7894974129972736e-05, + "loss": 1.0629, + "step": 9910 + }, + { + "epoch": 0.64, + "learning_rate": 1.7894759440949785e-05, + "loss": 1.0874, + "step": 9911 + }, + { + "epoch": 0.64, + "learning_rate": 1.7894544751926834e-05, + "loss": 1.1097, + "step": 9912 + }, + { + "epoch": 0.64, + "learning_rate": 1.7894330062903883e-05, + "loss": 1.1176, + "step": 9913 + }, + { + "epoch": 0.64, + "learning_rate": 1.7894115373880936e-05, + "loss": 1.1276, + "step": 9914 + }, + { + "epoch": 0.64, + "learning_rate": 1.7893900684857984e-05, + "loss": 1.0878, + "step": 9915 + }, + { + "epoch": 0.64, + "learning_rate": 1.7893685995835033e-05, + "loss": 0.9907, + "step": 9916 + }, + { + "epoch": 0.64, + "learning_rate": 1.7893471306812086e-05, + "loss": 1.0882, + "step": 9917 + }, + { + "epoch": 0.64, + "learning_rate": 1.7893256617789135e-05, + "loss": 1.0386, + "step": 9918 + }, + { + "epoch": 0.64, + "learning_rate": 1.7893041928766184e-05, + "loss": 1.077, + "step": 9919 + }, + { + "epoch": 0.64, + "learning_rate": 1.7892827239743233e-05, + "loss": 0.9934, + "step": 9920 + }, + { + "epoch": 0.64, + "learning_rate": 1.789261255072028e-05, + "loss": 1.1483, + "step": 9921 + }, + { + "epoch": 0.64, + "learning_rate": 1.7892397861697334e-05, + "loss": 1.0003, + "step": 9922 + }, + { + "epoch": 0.64, + "learning_rate": 1.7892183172674383e-05, + "loss": 1.0479, + "step": 9923 + }, + { + "epoch": 0.64, + "learning_rate": 1.7891968483651432e-05, + "loss": 1.0177, + "step": 9924 + }, + { + "epoch": 0.64, + "learning_rate": 1.789175379462848e-05, + "loss": 1.1464, + "step": 9925 + }, + { + "epoch": 0.64, + "learning_rate": 1.789153910560553e-05, + "loss": 1.1337, + "step": 9926 + }, + { + "epoch": 0.64, + "learning_rate": 1.7891324416582582e-05, + "loss": 1.0509, + "step": 9927 + }, + { + "epoch": 0.64, + "learning_rate": 1.789110972755963e-05, + "loss": 0.9409, + "step": 9928 + }, + { + "epoch": 0.64, + "learning_rate": 1.7890895038536684e-05, + "loss": 1.1348, + "step": 9929 + }, + { + "epoch": 0.64, + "learning_rate": 1.7890680349513733e-05, + "loss": 1.0319, + "step": 9930 + }, + { + "epoch": 0.64, + "learning_rate": 1.789046566049078e-05, + "loss": 1.0309, + "step": 9931 + }, + { + "epoch": 0.64, + "learning_rate": 1.789025097146783e-05, + "loss": 1.1096, + "step": 9932 + }, + { + "epoch": 0.64, + "learning_rate": 1.789003628244488e-05, + "loss": 1.1837, + "step": 9933 + }, + { + "epoch": 0.64, + "learning_rate": 1.788982159342193e-05, + "loss": 1.0158, + "step": 9934 + }, + { + "epoch": 0.64, + "learning_rate": 1.788960690439898e-05, + "loss": 1.1148, + "step": 9935 + }, + { + "epoch": 0.64, + "learning_rate": 1.788939221537603e-05, + "loss": 1.0914, + "step": 9936 + }, + { + "epoch": 0.64, + "learning_rate": 1.788917752635308e-05, + "loss": 1.0134, + "step": 9937 + }, + { + "epoch": 0.64, + "learning_rate": 1.7888962837330128e-05, + "loss": 1.1936, + "step": 9938 + }, + { + "epoch": 0.64, + "learning_rate": 1.7888748148307177e-05, + "loss": 1.1183, + "step": 9939 + }, + { + "epoch": 0.64, + "learning_rate": 1.788853345928423e-05, + "loss": 1.053, + "step": 9940 + }, + { + "epoch": 0.64, + "learning_rate": 1.7888318770261278e-05, + "loss": 1.0534, + "step": 9941 + }, + { + "epoch": 0.64, + "learning_rate": 1.7888104081238327e-05, + "loss": 1.0684, + "step": 9942 + }, + { + "epoch": 0.64, + "learning_rate": 1.788788939221538e-05, + "loss": 1.0735, + "step": 9943 + }, + { + "epoch": 0.64, + "learning_rate": 1.7887674703192428e-05, + "loss": 1.0341, + "step": 9944 + }, + { + "epoch": 0.64, + "learning_rate": 1.7887460014169477e-05, + "loss": 1.0027, + "step": 9945 + }, + { + "epoch": 0.64, + "learning_rate": 1.7887245325146526e-05, + "loss": 1.0183, + "step": 9946 + }, + { + "epoch": 0.64, + "learning_rate": 1.788703063612358e-05, + "loss": 1.0335, + "step": 9947 + }, + { + "epoch": 0.64, + "learning_rate": 1.7886815947100627e-05, + "loss": 1.0361, + "step": 9948 + }, + { + "epoch": 0.64, + "learning_rate": 1.7886601258077676e-05, + "loss": 1.1169, + "step": 9949 + }, + { + "epoch": 0.64, + "learning_rate": 1.7886386569054725e-05, + "loss": 1.0534, + "step": 9950 + }, + { + "epoch": 0.64, + "learning_rate": 1.7886171880031774e-05, + "loss": 1.0445, + "step": 9951 + }, + { + "epoch": 0.64, + "learning_rate": 1.7885957191008823e-05, + "loss": 1.0232, + "step": 9952 + }, + { + "epoch": 0.64, + "learning_rate": 1.7885742501985876e-05, + "loss": 1.0879, + "step": 9953 + }, + { + "epoch": 0.64, + "learning_rate": 1.7885527812962925e-05, + "loss": 1.0626, + "step": 9954 + }, + { + "epoch": 0.64, + "learning_rate": 1.7885313123939974e-05, + "loss": 0.9699, + "step": 9955 + }, + { + "epoch": 0.64, + "learning_rate": 1.7885098434917023e-05, + "loss": 1.076, + "step": 9956 + }, + { + "epoch": 0.64, + "learning_rate": 1.7884883745894075e-05, + "loss": 1.0436, + "step": 9957 + }, + { + "epoch": 0.64, + "learning_rate": 1.7884669056871124e-05, + "loss": 1.0904, + "step": 9958 + }, + { + "epoch": 0.64, + "learning_rate": 1.7884454367848173e-05, + "loss": 1.1061, + "step": 9959 + }, + { + "epoch": 0.64, + "learning_rate": 1.7884239678825225e-05, + "loss": 1.0553, + "step": 9960 + }, + { + "epoch": 0.64, + "learning_rate": 1.7884024989802274e-05, + "loss": 1.0458, + "step": 9961 + }, + { + "epoch": 0.64, + "learning_rate": 1.7883810300779323e-05, + "loss": 1.0491, + "step": 9962 + }, + { + "epoch": 0.64, + "learning_rate": 1.7883595611756372e-05, + "loss": 1.0851, + "step": 9963 + }, + { + "epoch": 0.64, + "learning_rate": 1.788338092273342e-05, + "loss": 1.1504, + "step": 9964 + }, + { + "epoch": 0.64, + "learning_rate": 1.788316623371047e-05, + "loss": 1.1091, + "step": 9965 + }, + { + "epoch": 0.64, + "learning_rate": 1.7882951544687522e-05, + "loss": 1.0973, + "step": 9966 + }, + { + "epoch": 0.64, + "learning_rate": 1.788273685566457e-05, + "loss": 0.9944, + "step": 9967 + }, + { + "epoch": 0.64, + "learning_rate": 1.788252216664162e-05, + "loss": 0.9927, + "step": 9968 + }, + { + "epoch": 0.64, + "learning_rate": 1.788230747761867e-05, + "loss": 1.0545, + "step": 9969 + }, + { + "epoch": 0.64, + "learning_rate": 1.7882092788595718e-05, + "loss": 1.0896, + "step": 9970 + }, + { + "epoch": 0.64, + "learning_rate": 1.788187809957277e-05, + "loss": 1.0796, + "step": 9971 + }, + { + "epoch": 0.64, + "learning_rate": 1.788166341054982e-05, + "loss": 1.0156, + "step": 9972 + }, + { + "epoch": 0.64, + "learning_rate": 1.7881448721526872e-05, + "loss": 0.9778, + "step": 9973 + }, + { + "epoch": 0.64, + "learning_rate": 1.788123403250392e-05, + "loss": 1.1249, + "step": 9974 + }, + { + "epoch": 0.64, + "learning_rate": 1.788101934348097e-05, + "loss": 1.1702, + "step": 9975 + }, + { + "epoch": 0.64, + "learning_rate": 1.788080465445802e-05, + "loss": 1.0955, + "step": 9976 + }, + { + "epoch": 0.64, + "learning_rate": 1.7880589965435068e-05, + "loss": 1.1826, + "step": 9977 + }, + { + "epoch": 0.64, + "learning_rate": 1.788037527641212e-05, + "loss": 1.0541, + "step": 9978 + }, + { + "epoch": 0.64, + "learning_rate": 1.788016058738917e-05, + "loss": 1.0728, + "step": 9979 + }, + { + "epoch": 0.64, + "learning_rate": 1.7879945898366218e-05, + "loss": 1.1153, + "step": 9980 + }, + { + "epoch": 0.64, + "learning_rate": 1.7879731209343267e-05, + "loss": 1.0079, + "step": 9981 + }, + { + "epoch": 0.64, + "learning_rate": 1.7879516520320316e-05, + "loss": 1.0403, + "step": 9982 + }, + { + "epoch": 0.64, + "learning_rate": 1.7879301831297365e-05, + "loss": 1.1166, + "step": 9983 + }, + { + "epoch": 0.64, + "learning_rate": 1.7879087142274417e-05, + "loss": 1.0414, + "step": 9984 + }, + { + "epoch": 0.64, + "learning_rate": 1.7878872453251466e-05, + "loss": 1.1257, + "step": 9985 + }, + { + "epoch": 0.64, + "learning_rate": 1.787865776422852e-05, + "loss": 1.1303, + "step": 9986 + }, + { + "epoch": 0.64, + "learning_rate": 1.7878443075205568e-05, + "loss": 1.1503, + "step": 9987 + }, + { + "epoch": 0.64, + "learning_rate": 1.7878228386182617e-05, + "loss": 0.9903, + "step": 9988 + }, + { + "epoch": 0.64, + "learning_rate": 1.7878013697159665e-05, + "loss": 1.0884, + "step": 9989 + }, + { + "epoch": 0.64, + "learning_rate": 1.7877799008136714e-05, + "loss": 1.1447, + "step": 9990 + }, + { + "epoch": 0.64, + "learning_rate": 1.7877584319113767e-05, + "loss": 1.0537, + "step": 9991 + }, + { + "epoch": 0.64, + "learning_rate": 1.7877369630090816e-05, + "loss": 1.0287, + "step": 9992 + }, + { + "epoch": 0.64, + "learning_rate": 1.7877154941067865e-05, + "loss": 1.1558, + "step": 9993 + }, + { + "epoch": 0.64, + "learning_rate": 1.7876940252044914e-05, + "loss": 1.1092, + "step": 9994 + }, + { + "epoch": 0.64, + "learning_rate": 1.7876725563021963e-05, + "loss": 1.1314, + "step": 9995 + }, + { + "epoch": 0.64, + "learning_rate": 1.7876510873999015e-05, + "loss": 1.061, + "step": 9996 + }, + { + "epoch": 0.64, + "learning_rate": 1.7876296184976064e-05, + "loss": 1.1312, + "step": 9997 + }, + { + "epoch": 0.64, + "learning_rate": 1.7876081495953113e-05, + "loss": 1.0627, + "step": 9998 + }, + { + "epoch": 0.64, + "learning_rate": 1.7875866806930162e-05, + "loss": 1.0828, + "step": 9999 + }, + { + "epoch": 0.64, + "learning_rate": 1.787565211790721e-05, + "loss": 1.0779, + "step": 10000 + }, + { + "epoch": 0.64, + "eval_laion/cider": 0.5115066314370493, + "eval_laion/meteor": 0.3836147300648682, + "eval_laion/pred_num_words": 36.388671875, + "eval_laion/sacrebleu": 15.428113020426395, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0860545635223389, + "eval_runtime": 106.0733, + "eval_samples_per_second": 4.704, + "eval_steps_per_second": 0.302, + "step": 10000 + }, + { + "epoch": 0.64, + "learning_rate": 1.7875437428884263e-05, + "loss": 1.0745, + "step": 10001 + }, + { + "epoch": 0.64, + "learning_rate": 1.7875222739861312e-05, + "loss": 1.113, + "step": 10002 + }, + { + "epoch": 0.64, + "learning_rate": 1.787500805083836e-05, + "loss": 1.137, + "step": 10003 + }, + { + "epoch": 0.64, + "learning_rate": 1.7874793361815414e-05, + "loss": 1.0771, + "step": 10004 + }, + { + "epoch": 0.64, + "learning_rate": 1.7874578672792462e-05, + "loss": 1.0255, + "step": 10005 + }, + { + "epoch": 0.64, + "learning_rate": 1.787436398376951e-05, + "loss": 1.0697, + "step": 10006 + }, + { + "epoch": 0.64, + "learning_rate": 1.787414929474656e-05, + "loss": 1.0836, + "step": 10007 + }, + { + "epoch": 0.64, + "learning_rate": 1.787393460572361e-05, + "loss": 1.0952, + "step": 10008 + }, + { + "epoch": 0.64, + "learning_rate": 1.7873719916700662e-05, + "loss": 1.0377, + "step": 10009 + }, + { + "epoch": 0.64, + "learning_rate": 1.787350522767771e-05, + "loss": 1.0772, + "step": 10010 + }, + { + "epoch": 0.64, + "learning_rate": 1.787329053865476e-05, + "loss": 1.0959, + "step": 10011 + }, + { + "epoch": 0.64, + "learning_rate": 1.787307584963181e-05, + "loss": 1.1237, + "step": 10012 + }, + { + "epoch": 0.64, + "learning_rate": 1.7872861160608858e-05, + "loss": 1.1018, + "step": 10013 + }, + { + "epoch": 0.64, + "learning_rate": 1.787264647158591e-05, + "loss": 1.0727, + "step": 10014 + }, + { + "epoch": 0.64, + "learning_rate": 1.787243178256296e-05, + "loss": 1.0267, + "step": 10015 + }, + { + "epoch": 0.64, + "learning_rate": 1.787221709354001e-05, + "loss": 1.0955, + "step": 10016 + }, + { + "epoch": 0.64, + "learning_rate": 1.787200240451706e-05, + "loss": 1.1004, + "step": 10017 + }, + { + "epoch": 0.64, + "learning_rate": 1.787178771549411e-05, + "loss": 1.0586, + "step": 10018 + }, + { + "epoch": 0.64, + "learning_rate": 1.7871573026471158e-05, + "loss": 1.1624, + "step": 10019 + }, + { + "epoch": 0.64, + "learning_rate": 1.7871358337448207e-05, + "loss": 1.0018, + "step": 10020 + }, + { + "epoch": 0.64, + "learning_rate": 1.7871143648425256e-05, + "loss": 1.0793, + "step": 10021 + }, + { + "epoch": 0.64, + "learning_rate": 1.787092895940231e-05, + "loss": 1.015, + "step": 10022 + }, + { + "epoch": 0.64, + "learning_rate": 1.7870714270379357e-05, + "loss": 1.0073, + "step": 10023 + }, + { + "epoch": 0.64, + "learning_rate": 1.7870499581356406e-05, + "loss": 1.0121, + "step": 10024 + }, + { + "epoch": 0.64, + "learning_rate": 1.7870284892333455e-05, + "loss": 1.0651, + "step": 10025 + }, + { + "epoch": 0.65, + "learning_rate": 1.7870070203310504e-05, + "loss": 1.1073, + "step": 10026 + }, + { + "epoch": 0.65, + "learning_rate": 1.7869855514287557e-05, + "loss": 1.0189, + "step": 10027 + }, + { + "epoch": 0.65, + "learning_rate": 1.7869640825264606e-05, + "loss": 1.1343, + "step": 10028 + }, + { + "epoch": 0.65, + "learning_rate": 1.7869426136241655e-05, + "loss": 1.1025, + "step": 10029 + }, + { + "epoch": 0.65, + "learning_rate": 1.7869211447218707e-05, + "loss": 1.1005, + "step": 10030 + }, + { + "epoch": 0.65, + "learning_rate": 1.7868996758195756e-05, + "loss": 1.0478, + "step": 10031 + }, + { + "epoch": 0.65, + "learning_rate": 1.7868782069172805e-05, + "loss": 1.0632, + "step": 10032 + }, + { + "epoch": 0.65, + "learning_rate": 1.7868567380149854e-05, + "loss": 1.0394, + "step": 10033 + }, + { + "epoch": 0.65, + "learning_rate": 1.7868352691126906e-05, + "loss": 0.9819, + "step": 10034 + }, + { + "epoch": 0.65, + "learning_rate": 1.7868138002103955e-05, + "loss": 1.0452, + "step": 10035 + }, + { + "epoch": 0.65, + "learning_rate": 1.7867923313081004e-05, + "loss": 1.0245, + "step": 10036 + }, + { + "epoch": 0.65, + "learning_rate": 1.7867708624058053e-05, + "loss": 1.1233, + "step": 10037 + }, + { + "epoch": 0.65, + "learning_rate": 1.7867493935035102e-05, + "loss": 0.9964, + "step": 10038 + }, + { + "epoch": 0.65, + "learning_rate": 1.786727924601215e-05, + "loss": 1.1095, + "step": 10039 + }, + { + "epoch": 0.65, + "learning_rate": 1.7867064556989203e-05, + "loss": 1.0931, + "step": 10040 + }, + { + "epoch": 0.65, + "learning_rate": 1.7866849867966252e-05, + "loss": 1.0422, + "step": 10041 + }, + { + "epoch": 0.65, + "learning_rate": 1.78666351789433e-05, + "loss": 1.113, + "step": 10042 + }, + { + "epoch": 0.65, + "learning_rate": 1.786642048992035e-05, + "loss": 1.1025, + "step": 10043 + }, + { + "epoch": 0.65, + "learning_rate": 1.7866205800897403e-05, + "loss": 1.1496, + "step": 10044 + }, + { + "epoch": 0.65, + "learning_rate": 1.786599111187445e-05, + "loss": 1.0783, + "step": 10045 + }, + { + "epoch": 0.65, + "learning_rate": 1.78657764228515e-05, + "loss": 1.1194, + "step": 10046 + }, + { + "epoch": 0.65, + "learning_rate": 1.7865561733828553e-05, + "loss": 1.0658, + "step": 10047 + }, + { + "epoch": 0.65, + "learning_rate": 1.7865347044805602e-05, + "loss": 1.1339, + "step": 10048 + }, + { + "epoch": 0.65, + "learning_rate": 1.786513235578265e-05, + "loss": 1.0591, + "step": 10049 + }, + { + "epoch": 0.65, + "learning_rate": 1.78649176667597e-05, + "loss": 0.9439, + "step": 10050 + }, + { + "epoch": 0.65, + "learning_rate": 1.786470297773675e-05, + "loss": 1.1086, + "step": 10051 + }, + { + "epoch": 0.65, + "learning_rate": 1.7864488288713798e-05, + "loss": 1.0538, + "step": 10052 + }, + { + "epoch": 0.65, + "learning_rate": 1.786427359969085e-05, + "loss": 1.1065, + "step": 10053 + }, + { + "epoch": 0.65, + "learning_rate": 1.78640589106679e-05, + "loss": 1.0101, + "step": 10054 + }, + { + "epoch": 0.65, + "learning_rate": 1.7863844221644948e-05, + "loss": 1.0704, + "step": 10055 + }, + { + "epoch": 0.65, + "learning_rate": 1.7863629532621997e-05, + "loss": 1.0587, + "step": 10056 + }, + { + "epoch": 0.65, + "learning_rate": 1.7863414843599046e-05, + "loss": 0.9698, + "step": 10057 + }, + { + "epoch": 0.65, + "learning_rate": 1.7863200154576098e-05, + "loss": 1.006, + "step": 10058 + }, + { + "epoch": 0.65, + "learning_rate": 1.7862985465553147e-05, + "loss": 1.0555, + "step": 10059 + }, + { + "epoch": 0.65, + "learning_rate": 1.78627707765302e-05, + "loss": 1.0177, + "step": 10060 + }, + { + "epoch": 0.65, + "learning_rate": 1.786255608750725e-05, + "loss": 1.049, + "step": 10061 + }, + { + "epoch": 0.65, + "learning_rate": 1.7862341398484298e-05, + "loss": 1.1133, + "step": 10062 + }, + { + "epoch": 0.65, + "learning_rate": 1.7862126709461347e-05, + "loss": 1.0239, + "step": 10063 + }, + { + "epoch": 0.65, + "learning_rate": 1.7861912020438395e-05, + "loss": 1.0361, + "step": 10064 + }, + { + "epoch": 0.65, + "learning_rate": 1.7861697331415448e-05, + "loss": 1.1136, + "step": 10065 + }, + { + "epoch": 0.65, + "learning_rate": 1.7861482642392497e-05, + "loss": 1.0341, + "step": 10066 + }, + { + "epoch": 0.65, + "learning_rate": 1.7861267953369546e-05, + "loss": 1.0514, + "step": 10067 + }, + { + "epoch": 0.65, + "learning_rate": 1.7861053264346595e-05, + "loss": 1.0946, + "step": 10068 + }, + { + "epoch": 0.65, + "learning_rate": 1.7860838575323644e-05, + "loss": 1.0889, + "step": 10069 + }, + { + "epoch": 0.65, + "learning_rate": 1.7860623886300693e-05, + "loss": 1.0415, + "step": 10070 + }, + { + "epoch": 0.65, + "learning_rate": 1.7860409197277745e-05, + "loss": 1.0508, + "step": 10071 + }, + { + "epoch": 0.65, + "learning_rate": 1.7860194508254794e-05, + "loss": 1.0407, + "step": 10072 + }, + { + "epoch": 0.65, + "learning_rate": 1.7859979819231846e-05, + "loss": 1.1606, + "step": 10073 + }, + { + "epoch": 0.65, + "learning_rate": 1.7859979819231846e-05, + "loss": 1.061, + "step": 10074 + }, + { + "epoch": 0.65, + "learning_rate": 1.7859765130208895e-05, + "loss": 1.1637, + "step": 10075 + }, + { + "epoch": 0.65, + "learning_rate": 1.7859550441185944e-05, + "loss": 1.0061, + "step": 10076 + }, + { + "epoch": 0.65, + "learning_rate": 1.7859335752162993e-05, + "loss": 1.1365, + "step": 10077 + }, + { + "epoch": 0.65, + "learning_rate": 1.7859121063140042e-05, + "loss": 1.0588, + "step": 10078 + }, + { + "epoch": 0.65, + "learning_rate": 1.7858906374117095e-05, + "loss": 0.98, + "step": 10079 + }, + { + "epoch": 0.65, + "learning_rate": 1.7858691685094143e-05, + "loss": 1.0397, + "step": 10080 + }, + { + "epoch": 0.65, + "learning_rate": 1.7858476996071192e-05, + "loss": 1.0954, + "step": 10081 + }, + { + "epoch": 0.65, + "learning_rate": 1.785826230704824e-05, + "loss": 1.0621, + "step": 10082 + }, + { + "epoch": 0.65, + "learning_rate": 1.785804761802529e-05, + "loss": 0.9739, + "step": 10083 + }, + { + "epoch": 0.65, + "learning_rate": 1.7857832929002343e-05, + "loss": 1.0958, + "step": 10084 + }, + { + "epoch": 0.65, + "learning_rate": 1.7857618239979392e-05, + "loss": 1.0678, + "step": 10085 + }, + { + "epoch": 0.65, + "learning_rate": 1.785740355095644e-05, + "loss": 1.1401, + "step": 10086 + }, + { + "epoch": 0.65, + "learning_rate": 1.785718886193349e-05, + "loss": 1.1245, + "step": 10087 + }, + { + "epoch": 0.65, + "learning_rate": 1.785697417291054e-05, + "loss": 1.048, + "step": 10088 + }, + { + "epoch": 0.65, + "learning_rate": 1.785675948388759e-05, + "loss": 1.0644, + "step": 10089 + }, + { + "epoch": 0.65, + "learning_rate": 1.785654479486464e-05, + "loss": 1.1023, + "step": 10090 + }, + { + "epoch": 0.65, + "learning_rate": 1.785633010584169e-05, + "loss": 1.0298, + "step": 10091 + }, + { + "epoch": 0.65, + "learning_rate": 1.785611541681874e-05, + "loss": 1.0449, + "step": 10092 + }, + { + "epoch": 0.65, + "learning_rate": 1.785590072779579e-05, + "loss": 1.1299, + "step": 10093 + }, + { + "epoch": 0.65, + "learning_rate": 1.785568603877284e-05, + "loss": 1.0457, + "step": 10094 + }, + { + "epoch": 0.65, + "learning_rate": 1.7855471349749888e-05, + "loss": 1.1075, + "step": 10095 + }, + { + "epoch": 0.65, + "learning_rate": 1.7855256660726937e-05, + "loss": 1.0089, + "step": 10096 + }, + { + "epoch": 0.65, + "learning_rate": 1.785504197170399e-05, + "loss": 1.2047, + "step": 10097 + }, + { + "epoch": 0.65, + "learning_rate": 1.785482728268104e-05, + "loss": 1.0775, + "step": 10098 + }, + { + "epoch": 0.65, + "learning_rate": 1.7854612593658087e-05, + "loss": 1.0797, + "step": 10099 + }, + { + "epoch": 0.65, + "learning_rate": 1.7854397904635136e-05, + "loss": 0.9808, + "step": 10100 + }, + { + "epoch": 0.65, + "learning_rate": 1.7854183215612185e-05, + "loss": 1.1009, + "step": 10101 + }, + { + "epoch": 0.65, + "learning_rate": 1.7853968526589234e-05, + "loss": 1.088, + "step": 10102 + }, + { + "epoch": 0.65, + "learning_rate": 1.7853753837566287e-05, + "loss": 1.1067, + "step": 10103 + }, + { + "epoch": 0.65, + "learning_rate": 1.785353914854334e-05, + "loss": 1.0607, + "step": 10104 + }, + { + "epoch": 0.65, + "learning_rate": 1.7853324459520388e-05, + "loss": 1.0962, + "step": 10105 + }, + { + "epoch": 0.65, + "learning_rate": 1.7853109770497437e-05, + "loss": 1.1697, + "step": 10106 + }, + { + "epoch": 0.65, + "learning_rate": 1.7852895081474486e-05, + "loss": 1.0438, + "step": 10107 + }, + { + "epoch": 0.65, + "learning_rate": 1.7852680392451535e-05, + "loss": 1.0789, + "step": 10108 + }, + { + "epoch": 0.65, + "learning_rate": 1.7852465703428584e-05, + "loss": 0.9249, + "step": 10109 + }, + { + "epoch": 0.65, + "learning_rate": 1.7852251014405636e-05, + "loss": 1.0447, + "step": 10110 + }, + { + "epoch": 0.65, + "learning_rate": 1.7852036325382685e-05, + "loss": 1.0224, + "step": 10111 + }, + { + "epoch": 0.65, + "learning_rate": 1.7851821636359734e-05, + "loss": 1.0619, + "step": 10112 + }, + { + "epoch": 0.65, + "learning_rate": 1.7851606947336783e-05, + "loss": 1.0853, + "step": 10113 + }, + { + "epoch": 0.65, + "learning_rate": 1.7851392258313832e-05, + "loss": 1.1336, + "step": 10114 + }, + { + "epoch": 0.65, + "learning_rate": 1.7851177569290884e-05, + "loss": 1.0321, + "step": 10115 + }, + { + "epoch": 0.65, + "learning_rate": 1.7850962880267933e-05, + "loss": 1.0626, + "step": 10116 + }, + { + "epoch": 0.65, + "learning_rate": 1.7850748191244982e-05, + "loss": 1.0635, + "step": 10117 + }, + { + "epoch": 0.65, + "learning_rate": 1.7850533502222035e-05, + "loss": 1.1578, + "step": 10118 + }, + { + "epoch": 0.65, + "learning_rate": 1.7850318813199084e-05, + "loss": 1.1016, + "step": 10119 + }, + { + "epoch": 0.65, + "learning_rate": 1.7850104124176133e-05, + "loss": 1.0745, + "step": 10120 + }, + { + "epoch": 0.65, + "learning_rate": 1.784988943515318e-05, + "loss": 1.187, + "step": 10121 + }, + { + "epoch": 0.65, + "learning_rate": 1.7849674746130234e-05, + "loss": 1.131, + "step": 10122 + }, + { + "epoch": 0.65, + "learning_rate": 1.7849460057107283e-05, + "loss": 1.1024, + "step": 10123 + }, + { + "epoch": 0.65, + "learning_rate": 1.7849245368084332e-05, + "loss": 1.1002, + "step": 10124 + }, + { + "epoch": 0.65, + "learning_rate": 1.784903067906138e-05, + "loss": 0.9864, + "step": 10125 + }, + { + "epoch": 0.65, + "learning_rate": 1.784881599003843e-05, + "loss": 1.0912, + "step": 10126 + }, + { + "epoch": 0.65, + "learning_rate": 1.784860130101548e-05, + "loss": 1.1644, + "step": 10127 + }, + { + "epoch": 0.65, + "learning_rate": 1.784838661199253e-05, + "loss": 1.0208, + "step": 10128 + }, + { + "epoch": 0.65, + "learning_rate": 1.784817192296958e-05, + "loss": 1.1165, + "step": 10129 + }, + { + "epoch": 0.65, + "learning_rate": 1.784795723394663e-05, + "loss": 1.0091, + "step": 10130 + }, + { + "epoch": 0.65, + "learning_rate": 1.7847742544923678e-05, + "loss": 0.9963, + "step": 10131 + }, + { + "epoch": 0.65, + "learning_rate": 1.784752785590073e-05, + "loss": 1.0954, + "step": 10132 + }, + { + "epoch": 0.65, + "learning_rate": 1.784731316687778e-05, + "loss": 1.0789, + "step": 10133 + }, + { + "epoch": 0.65, + "learning_rate": 1.7847098477854828e-05, + "loss": 1.0099, + "step": 10134 + }, + { + "epoch": 0.65, + "learning_rate": 1.784688378883188e-05, + "loss": 1.0161, + "step": 10135 + }, + { + "epoch": 0.65, + "learning_rate": 1.784666909980893e-05, + "loss": 1.0891, + "step": 10136 + }, + { + "epoch": 0.65, + "learning_rate": 1.784645441078598e-05, + "loss": 1.0494, + "step": 10137 + }, + { + "epoch": 0.65, + "learning_rate": 1.7846239721763028e-05, + "loss": 1.1164, + "step": 10138 + }, + { + "epoch": 0.65, + "learning_rate": 1.7846025032740076e-05, + "loss": 0.9925, + "step": 10139 + }, + { + "epoch": 0.65, + "learning_rate": 1.7845810343717125e-05, + "loss": 1.0948, + "step": 10140 + }, + { + "epoch": 0.65, + "learning_rate": 1.7845595654694178e-05, + "loss": 1.034, + "step": 10141 + }, + { + "epoch": 0.65, + "learning_rate": 1.7845380965671227e-05, + "loss": 1.0832, + "step": 10142 + }, + { + "epoch": 0.65, + "learning_rate": 1.7845166276648276e-05, + "loss": 1.0238, + "step": 10143 + }, + { + "epoch": 0.65, + "learning_rate": 1.7844951587625325e-05, + "loss": 0.9974, + "step": 10144 + }, + { + "epoch": 0.65, + "learning_rate": 1.7844736898602374e-05, + "loss": 0.9767, + "step": 10145 + }, + { + "epoch": 0.65, + "learning_rate": 1.7844522209579426e-05, + "loss": 1.0606, + "step": 10146 + }, + { + "epoch": 0.65, + "learning_rate": 1.7844307520556475e-05, + "loss": 1.1606, + "step": 10147 + }, + { + "epoch": 0.65, + "learning_rate": 1.7844092831533527e-05, + "loss": 1.0682, + "step": 10148 + }, + { + "epoch": 0.65, + "learning_rate": 1.7843878142510576e-05, + "loss": 1.0289, + "step": 10149 + }, + { + "epoch": 0.65, + "learning_rate": 1.7843663453487625e-05, + "loss": 1.1175, + "step": 10150 + }, + { + "epoch": 0.65, + "learning_rate": 1.7843448764464674e-05, + "loss": 0.9985, + "step": 10151 + }, + { + "epoch": 0.65, + "learning_rate": 1.7843234075441723e-05, + "loss": 1.0545, + "step": 10152 + }, + { + "epoch": 0.65, + "learning_rate": 1.7843019386418776e-05, + "loss": 1.1041, + "step": 10153 + }, + { + "epoch": 0.65, + "learning_rate": 1.7842804697395824e-05, + "loss": 1.1256, + "step": 10154 + }, + { + "epoch": 0.65, + "learning_rate": 1.7842590008372873e-05, + "loss": 1.1112, + "step": 10155 + }, + { + "epoch": 0.65, + "learning_rate": 1.7842375319349922e-05, + "loss": 1.0485, + "step": 10156 + }, + { + "epoch": 0.65, + "learning_rate": 1.784216063032697e-05, + "loss": 1.1265, + "step": 10157 + }, + { + "epoch": 0.65, + "learning_rate": 1.784194594130402e-05, + "loss": 1.0399, + "step": 10158 + }, + { + "epoch": 0.65, + "learning_rate": 1.7841731252281073e-05, + "loss": 1.1469, + "step": 10159 + }, + { + "epoch": 0.65, + "learning_rate": 1.784151656325812e-05, + "loss": 1.1646, + "step": 10160 + }, + { + "epoch": 0.65, + "learning_rate": 1.7841301874235174e-05, + "loss": 1.1309, + "step": 10161 + }, + { + "epoch": 0.65, + "learning_rate": 1.7841087185212223e-05, + "loss": 1.0679, + "step": 10162 + }, + { + "epoch": 0.65, + "learning_rate": 1.7840872496189272e-05, + "loss": 1.0289, + "step": 10163 + }, + { + "epoch": 0.65, + "learning_rate": 1.784065780716632e-05, + "loss": 1.0297, + "step": 10164 + }, + { + "epoch": 0.65, + "learning_rate": 1.784044311814337e-05, + "loss": 1.0221, + "step": 10165 + }, + { + "epoch": 0.65, + "learning_rate": 1.7840228429120422e-05, + "loss": 1.1164, + "step": 10166 + }, + { + "epoch": 0.65, + "learning_rate": 1.784001374009747e-05, + "loss": 0.9979, + "step": 10167 + }, + { + "epoch": 0.65, + "learning_rate": 1.783979905107452e-05, + "loss": 1.0536, + "step": 10168 + }, + { + "epoch": 0.65, + "learning_rate": 1.783958436205157e-05, + "loss": 1.0858, + "step": 10169 + }, + { + "epoch": 0.65, + "learning_rate": 1.7839369673028618e-05, + "loss": 1.0393, + "step": 10170 + }, + { + "epoch": 0.65, + "learning_rate": 1.783915498400567e-05, + "loss": 1.0107, + "step": 10171 + }, + { + "epoch": 0.65, + "learning_rate": 1.783894029498272e-05, + "loss": 1.0705, + "step": 10172 + }, + { + "epoch": 0.65, + "learning_rate": 1.783872560595977e-05, + "loss": 1.025, + "step": 10173 + }, + { + "epoch": 0.65, + "learning_rate": 1.7838510916936817e-05, + "loss": 1.057, + "step": 10174 + }, + { + "epoch": 0.65, + "learning_rate": 1.7838296227913866e-05, + "loss": 1.1414, + "step": 10175 + }, + { + "epoch": 0.65, + "learning_rate": 1.783808153889092e-05, + "loss": 1.0977, + "step": 10176 + }, + { + "epoch": 0.65, + "learning_rate": 1.7837866849867968e-05, + "loss": 1.1156, + "step": 10177 + }, + { + "epoch": 0.65, + "learning_rate": 1.7837652160845017e-05, + "loss": 1.0067, + "step": 10178 + }, + { + "epoch": 0.65, + "learning_rate": 1.783743747182207e-05, + "loss": 1.0205, + "step": 10179 + }, + { + "epoch": 0.65, + "learning_rate": 1.7837222782799118e-05, + "loss": 1.048, + "step": 10180 + }, + { + "epoch": 0.66, + "learning_rate": 1.7837008093776167e-05, + "loss": 1.0991, + "step": 10181 + }, + { + "epoch": 0.66, + "learning_rate": 1.7836793404753216e-05, + "loss": 1.1473, + "step": 10182 + }, + { + "epoch": 0.66, + "learning_rate": 1.7836578715730265e-05, + "loss": 1.041, + "step": 10183 + }, + { + "epoch": 0.66, + "learning_rate": 1.7836364026707317e-05, + "loss": 1.0738, + "step": 10184 + }, + { + "epoch": 0.66, + "learning_rate": 1.7836149337684366e-05, + "loss": 1.0543, + "step": 10185 + }, + { + "epoch": 0.66, + "learning_rate": 1.7835934648661415e-05, + "loss": 1.029, + "step": 10186 + }, + { + "epoch": 0.66, + "learning_rate": 1.7835719959638464e-05, + "loss": 1.1072, + "step": 10187 + }, + { + "epoch": 0.66, + "learning_rate": 1.7835505270615513e-05, + "loss": 1.0556, + "step": 10188 + }, + { + "epoch": 0.66, + "learning_rate": 1.7835290581592562e-05, + "loss": 1.0896, + "step": 10189 + }, + { + "epoch": 0.66, + "learning_rate": 1.7835075892569614e-05, + "loss": 1.1033, + "step": 10190 + }, + { + "epoch": 0.66, + "learning_rate": 1.7834861203546667e-05, + "loss": 1.03, + "step": 10191 + }, + { + "epoch": 0.66, + "learning_rate": 1.7834646514523716e-05, + "loss": 1.0726, + "step": 10192 + }, + { + "epoch": 0.66, + "learning_rate": 1.7834431825500765e-05, + "loss": 1.0835, + "step": 10193 + }, + { + "epoch": 0.66, + "learning_rate": 1.7834217136477814e-05, + "loss": 1.0337, + "step": 10194 + }, + { + "epoch": 0.66, + "learning_rate": 1.7834002447454863e-05, + "loss": 1.0958, + "step": 10195 + }, + { + "epoch": 0.66, + "learning_rate": 1.783378775843191e-05, + "loss": 1.1074, + "step": 10196 + }, + { + "epoch": 0.66, + "learning_rate": 1.7833573069408964e-05, + "loss": 1.0405, + "step": 10197 + }, + { + "epoch": 0.66, + "learning_rate": 1.7833358380386013e-05, + "loss": 1.0566, + "step": 10198 + }, + { + "epoch": 0.66, + "learning_rate": 1.7833143691363062e-05, + "loss": 1.0363, + "step": 10199 + }, + { + "epoch": 0.66, + "learning_rate": 1.783292900234011e-05, + "loss": 1.0348, + "step": 10200 + }, + { + "epoch": 0.66, + "learning_rate": 1.783271431331716e-05, + "loss": 1.0482, + "step": 10201 + }, + { + "epoch": 0.66, + "learning_rate": 1.7832499624294212e-05, + "loss": 0.9293, + "step": 10202 + }, + { + "epoch": 0.66, + "learning_rate": 1.783228493527126e-05, + "loss": 1.14, + "step": 10203 + }, + { + "epoch": 0.66, + "learning_rate": 1.783207024624831e-05, + "loss": 1.0201, + "step": 10204 + }, + { + "epoch": 0.66, + "learning_rate": 1.7831855557225362e-05, + "loss": 1.132, + "step": 10205 + }, + { + "epoch": 0.66, + "learning_rate": 1.783164086820241e-05, + "loss": 1.1071, + "step": 10206 + }, + { + "epoch": 0.66, + "learning_rate": 1.783142617917946e-05, + "loss": 1.0733, + "step": 10207 + }, + { + "epoch": 0.66, + "learning_rate": 1.783121149015651e-05, + "loss": 0.9969, + "step": 10208 + }, + { + "epoch": 0.66, + "learning_rate": 1.7830996801133558e-05, + "loss": 0.9652, + "step": 10209 + }, + { + "epoch": 0.66, + "learning_rate": 1.783078211211061e-05, + "loss": 0.9999, + "step": 10210 + }, + { + "epoch": 0.66, + "learning_rate": 1.783056742308766e-05, + "loss": 1.0975, + "step": 10211 + }, + { + "epoch": 0.66, + "learning_rate": 1.783035273406471e-05, + "loss": 1.1403, + "step": 10212 + }, + { + "epoch": 0.66, + "learning_rate": 1.7830138045041757e-05, + "loss": 1.0199, + "step": 10213 + }, + { + "epoch": 0.66, + "learning_rate": 1.7829923356018806e-05, + "loss": 1.1365, + "step": 10214 + }, + { + "epoch": 0.66, + "learning_rate": 1.782970866699586e-05, + "loss": 1.0695, + "step": 10215 + }, + { + "epoch": 0.66, + "learning_rate": 1.7829493977972908e-05, + "loss": 1.0304, + "step": 10216 + }, + { + "epoch": 0.66, + "learning_rate": 1.7829279288949957e-05, + "loss": 0.9964, + "step": 10217 + }, + { + "epoch": 0.66, + "learning_rate": 1.7829064599927006e-05, + "loss": 1.0343, + "step": 10218 + }, + { + "epoch": 0.66, + "learning_rate": 1.7828849910904058e-05, + "loss": 1.0179, + "step": 10219 + }, + { + "epoch": 0.66, + "learning_rate": 1.7828635221881107e-05, + "loss": 1.0875, + "step": 10220 + }, + { + "epoch": 0.66, + "learning_rate": 1.7828420532858156e-05, + "loss": 1.0554, + "step": 10221 + }, + { + "epoch": 0.66, + "learning_rate": 1.782820584383521e-05, + "loss": 1.0598, + "step": 10222 + }, + { + "epoch": 0.66, + "learning_rate": 1.7827991154812257e-05, + "loss": 1.0167, + "step": 10223 + }, + { + "epoch": 0.66, + "learning_rate": 1.7827776465789306e-05, + "loss": 1.1302, + "step": 10224 + }, + { + "epoch": 0.66, + "learning_rate": 1.7827561776766355e-05, + "loss": 1.1529, + "step": 10225 + }, + { + "epoch": 0.66, + "learning_rate": 1.7827347087743404e-05, + "loss": 1.0598, + "step": 10226 + }, + { + "epoch": 0.66, + "learning_rate": 1.7827132398720453e-05, + "loss": 1.0658, + "step": 10227 + }, + { + "epoch": 0.66, + "learning_rate": 1.7826917709697506e-05, + "loss": 1.141, + "step": 10228 + }, + { + "epoch": 0.66, + "learning_rate": 1.7826703020674554e-05, + "loss": 1.0997, + "step": 10229 + }, + { + "epoch": 0.66, + "learning_rate": 1.7826488331651603e-05, + "loss": 1.1039, + "step": 10230 + }, + { + "epoch": 0.66, + "learning_rate": 1.7826273642628652e-05, + "loss": 1.0327, + "step": 10231 + }, + { + "epoch": 0.66, + "learning_rate": 1.78260589536057e-05, + "loss": 1.0793, + "step": 10232 + }, + { + "epoch": 0.66, + "learning_rate": 1.7825844264582754e-05, + "loss": 1.0657, + "step": 10233 + }, + { + "epoch": 0.66, + "learning_rate": 1.7825629575559803e-05, + "loss": 1.1114, + "step": 10234 + }, + { + "epoch": 0.66, + "learning_rate": 1.7825414886536855e-05, + "loss": 1.0558, + "step": 10235 + }, + { + "epoch": 0.66, + "learning_rate": 1.7825200197513904e-05, + "loss": 1.1227, + "step": 10236 + }, + { + "epoch": 0.66, + "learning_rate": 1.7824985508490953e-05, + "loss": 1.033, + "step": 10237 + }, + { + "epoch": 0.66, + "learning_rate": 1.7824770819468002e-05, + "loss": 1.1082, + "step": 10238 + }, + { + "epoch": 0.66, + "learning_rate": 1.782455613044505e-05, + "loss": 1.0568, + "step": 10239 + }, + { + "epoch": 0.66, + "learning_rate": 1.7824341441422103e-05, + "loss": 1.1007, + "step": 10240 + }, + { + "epoch": 0.66, + "learning_rate": 1.7824126752399152e-05, + "loss": 1.0597, + "step": 10241 + }, + { + "epoch": 0.66, + "learning_rate": 1.78239120633762e-05, + "loss": 1.116, + "step": 10242 + }, + { + "epoch": 0.66, + "learning_rate": 1.782369737435325e-05, + "loss": 0.9946, + "step": 10243 + }, + { + "epoch": 0.66, + "learning_rate": 1.78234826853303e-05, + "loss": 1.0201, + "step": 10244 + }, + { + "epoch": 0.66, + "learning_rate": 1.7823267996307348e-05, + "loss": 1.1672, + "step": 10245 + }, + { + "epoch": 0.66, + "learning_rate": 1.78230533072844e-05, + "loss": 0.9706, + "step": 10246 + }, + { + "epoch": 0.66, + "learning_rate": 1.782283861826145e-05, + "loss": 1.0211, + "step": 10247 + }, + { + "epoch": 0.66, + "learning_rate": 1.7822623929238502e-05, + "loss": 1.0055, + "step": 10248 + }, + { + "epoch": 0.66, + "learning_rate": 1.782240924021555e-05, + "loss": 1.0682, + "step": 10249 + }, + { + "epoch": 0.66, + "learning_rate": 1.78221945511926e-05, + "loss": 1.1591, + "step": 10250 + }, + { + "epoch": 0.66, + "learning_rate": 1.782197986216965e-05, + "loss": 1.0542, + "step": 10251 + }, + { + "epoch": 0.66, + "learning_rate": 1.7821765173146698e-05, + "loss": 1.1035, + "step": 10252 + }, + { + "epoch": 0.66, + "learning_rate": 1.782155048412375e-05, + "loss": 1.0979, + "step": 10253 + }, + { + "epoch": 0.66, + "learning_rate": 1.78213357951008e-05, + "loss": 0.9757, + "step": 10254 + }, + { + "epoch": 0.66, + "learning_rate": 1.7821121106077848e-05, + "loss": 1.1129, + "step": 10255 + }, + { + "epoch": 0.66, + "learning_rate": 1.7820906417054897e-05, + "loss": 1.101, + "step": 10256 + }, + { + "epoch": 0.66, + "learning_rate": 1.7820691728031946e-05, + "loss": 1.1366, + "step": 10257 + }, + { + "epoch": 0.66, + "learning_rate": 1.7820477039008998e-05, + "loss": 1.1002, + "step": 10258 + }, + { + "epoch": 0.66, + "learning_rate": 1.7820262349986047e-05, + "loss": 1.1033, + "step": 10259 + }, + { + "epoch": 0.66, + "learning_rate": 1.7820047660963096e-05, + "loss": 1.0131, + "step": 10260 + }, + { + "epoch": 0.66, + "learning_rate": 1.7819832971940145e-05, + "loss": 1.0944, + "step": 10261 + }, + { + "epoch": 0.66, + "learning_rate": 1.7819618282917194e-05, + "loss": 1.1331, + "step": 10262 + }, + { + "epoch": 0.66, + "learning_rate": 1.7819403593894246e-05, + "loss": 1.0164, + "step": 10263 + }, + { + "epoch": 0.66, + "learning_rate": 1.7819188904871295e-05, + "loss": 0.9895, + "step": 10264 + }, + { + "epoch": 0.66, + "learning_rate": 1.7818974215848344e-05, + "loss": 1.0398, + "step": 10265 + }, + { + "epoch": 0.66, + "learning_rate": 1.7818759526825397e-05, + "loss": 1.0318, + "step": 10266 + }, + { + "epoch": 0.66, + "learning_rate": 1.7818544837802446e-05, + "loss": 1.1246, + "step": 10267 + }, + { + "epoch": 0.66, + "learning_rate": 1.7818330148779495e-05, + "loss": 1.0212, + "step": 10268 + }, + { + "epoch": 0.66, + "learning_rate": 1.7818115459756544e-05, + "loss": 1.1489, + "step": 10269 + }, + { + "epoch": 0.66, + "learning_rate": 1.7817900770733593e-05, + "loss": 1.0601, + "step": 10270 + }, + { + "epoch": 0.66, + "learning_rate": 1.7817686081710645e-05, + "loss": 1.0715, + "step": 10271 + }, + { + "epoch": 0.66, + "learning_rate": 1.7817471392687694e-05, + "loss": 1.0275, + "step": 10272 + }, + { + "epoch": 0.66, + "learning_rate": 1.7817256703664743e-05, + "loss": 1.0848, + "step": 10273 + }, + { + "epoch": 0.66, + "learning_rate": 1.7817042014641792e-05, + "loss": 1.1707, + "step": 10274 + }, + { + "epoch": 0.66, + "learning_rate": 1.781682732561884e-05, + "loss": 1.0836, + "step": 10275 + }, + { + "epoch": 0.66, + "learning_rate": 1.781661263659589e-05, + "loss": 1.108, + "step": 10276 + }, + { + "epoch": 0.66, + "learning_rate": 1.7816397947572942e-05, + "loss": 1.0133, + "step": 10277 + }, + { + "epoch": 0.66, + "learning_rate": 1.7816183258549994e-05, + "loss": 1.1099, + "step": 10278 + }, + { + "epoch": 0.66, + "learning_rate": 1.7815968569527043e-05, + "loss": 1.0878, + "step": 10279 + }, + { + "epoch": 0.66, + "learning_rate": 1.7815753880504092e-05, + "loss": 1.0666, + "step": 10280 + }, + { + "epoch": 0.66, + "learning_rate": 1.781553919148114e-05, + "loss": 1.0442, + "step": 10281 + }, + { + "epoch": 0.66, + "learning_rate": 1.781532450245819e-05, + "loss": 1.1506, + "step": 10282 + }, + { + "epoch": 0.66, + "learning_rate": 1.781510981343524e-05, + "loss": 1.1403, + "step": 10283 + }, + { + "epoch": 0.66, + "learning_rate": 1.781489512441229e-05, + "loss": 1.0636, + "step": 10284 + }, + { + "epoch": 0.66, + "learning_rate": 1.781468043538934e-05, + "loss": 1.044, + "step": 10285 + }, + { + "epoch": 0.66, + "learning_rate": 1.781446574636639e-05, + "loss": 1.0987, + "step": 10286 + }, + { + "epoch": 0.66, + "learning_rate": 1.781425105734344e-05, + "loss": 1.0905, + "step": 10287 + }, + { + "epoch": 0.66, + "learning_rate": 1.7814036368320487e-05, + "loss": 1.1315, + "step": 10288 + }, + { + "epoch": 0.66, + "learning_rate": 1.781382167929754e-05, + "loss": 1.0949, + "step": 10289 + }, + { + "epoch": 0.66, + "learning_rate": 1.781360699027459e-05, + "loss": 1.1036, + "step": 10290 + }, + { + "epoch": 0.66, + "learning_rate": 1.7813392301251638e-05, + "loss": 1.1467, + "step": 10291 + }, + { + "epoch": 0.66, + "learning_rate": 1.781317761222869e-05, + "loss": 1.0721, + "step": 10292 + }, + { + "epoch": 0.66, + "learning_rate": 1.781296292320574e-05, + "loss": 0.9962, + "step": 10293 + }, + { + "epoch": 0.66, + "learning_rate": 1.7812748234182788e-05, + "loss": 1.0464, + "step": 10294 + }, + { + "epoch": 0.66, + "learning_rate": 1.7812533545159837e-05, + "loss": 1.1496, + "step": 10295 + }, + { + "epoch": 0.66, + "learning_rate": 1.7812318856136886e-05, + "loss": 1.0562, + "step": 10296 + }, + { + "epoch": 0.66, + "learning_rate": 1.7812104167113938e-05, + "loss": 1.1056, + "step": 10297 + }, + { + "epoch": 0.66, + "learning_rate": 1.7811889478090987e-05, + "loss": 1.053, + "step": 10298 + }, + { + "epoch": 0.66, + "learning_rate": 1.7811674789068036e-05, + "loss": 1.0393, + "step": 10299 + }, + { + "epoch": 0.66, + "learning_rate": 1.7811460100045085e-05, + "loss": 1.0299, + "step": 10300 + }, + { + "epoch": 0.66, + "learning_rate": 1.7811245411022134e-05, + "loss": 1.0786, + "step": 10301 + }, + { + "epoch": 0.66, + "learning_rate": 1.7811030721999187e-05, + "loss": 1.1287, + "step": 10302 + }, + { + "epoch": 0.66, + "learning_rate": 1.7810816032976235e-05, + "loss": 1.1386, + "step": 10303 + }, + { + "epoch": 0.66, + "learning_rate": 1.7810601343953284e-05, + "loss": 1.0843, + "step": 10304 + }, + { + "epoch": 0.66, + "learning_rate": 1.7810386654930333e-05, + "loss": 1.0017, + "step": 10305 + }, + { + "epoch": 0.66, + "learning_rate": 1.7810171965907382e-05, + "loss": 1.0465, + "step": 10306 + }, + { + "epoch": 0.66, + "learning_rate": 1.7809957276884435e-05, + "loss": 1.1151, + "step": 10307 + }, + { + "epoch": 0.66, + "learning_rate": 1.7809742587861484e-05, + "loss": 1.0068, + "step": 10308 + }, + { + "epoch": 0.66, + "learning_rate": 1.7809527898838536e-05, + "loss": 1.082, + "step": 10309 + }, + { + "epoch": 0.66, + "learning_rate": 1.7809313209815585e-05, + "loss": 1.1219, + "step": 10310 + }, + { + "epoch": 0.66, + "learning_rate": 1.7809098520792634e-05, + "loss": 1.0809, + "step": 10311 + }, + { + "epoch": 0.66, + "learning_rate": 1.7808883831769683e-05, + "loss": 1.0906, + "step": 10312 + }, + { + "epoch": 0.66, + "learning_rate": 1.7808669142746732e-05, + "loss": 1.0817, + "step": 10313 + }, + { + "epoch": 0.66, + "learning_rate": 1.780845445372378e-05, + "loss": 1.0689, + "step": 10314 + }, + { + "epoch": 0.66, + "learning_rate": 1.7808239764700833e-05, + "loss": 1.1023, + "step": 10315 + }, + { + "epoch": 0.66, + "learning_rate": 1.7808025075677882e-05, + "loss": 1.1339, + "step": 10316 + }, + { + "epoch": 0.66, + "learning_rate": 1.780781038665493e-05, + "loss": 1.2016, + "step": 10317 + }, + { + "epoch": 0.66, + "learning_rate": 1.780759569763198e-05, + "loss": 1.0273, + "step": 10318 + }, + { + "epoch": 0.66, + "learning_rate": 1.780738100860903e-05, + "loss": 1.1514, + "step": 10319 + }, + { + "epoch": 0.66, + "learning_rate": 1.780716631958608e-05, + "loss": 1.1086, + "step": 10320 + }, + { + "epoch": 0.66, + "learning_rate": 1.780695163056313e-05, + "loss": 1.1115, + "step": 10321 + }, + { + "epoch": 0.66, + "learning_rate": 1.7806736941540183e-05, + "loss": 1.1051, + "step": 10322 + }, + { + "epoch": 0.66, + "learning_rate": 1.7806522252517232e-05, + "loss": 1.1384, + "step": 10323 + }, + { + "epoch": 0.66, + "learning_rate": 1.780630756349428e-05, + "loss": 1.0154, + "step": 10324 + }, + { + "epoch": 0.66, + "learning_rate": 1.780609287447133e-05, + "loss": 1.0162, + "step": 10325 + }, + { + "epoch": 0.66, + "learning_rate": 1.780587818544838e-05, + "loss": 1.0792, + "step": 10326 + }, + { + "epoch": 0.66, + "learning_rate": 1.780566349642543e-05, + "loss": 1.12, + "step": 10327 + }, + { + "epoch": 0.66, + "learning_rate": 1.780544880740248e-05, + "loss": 1.0366, + "step": 10328 + }, + { + "epoch": 0.66, + "learning_rate": 1.780523411837953e-05, + "loss": 1.0406, + "step": 10329 + }, + { + "epoch": 0.66, + "learning_rate": 1.7805019429356578e-05, + "loss": 1.0707, + "step": 10330 + }, + { + "epoch": 0.66, + "learning_rate": 1.7804804740333627e-05, + "loss": 1.1096, + "step": 10331 + }, + { + "epoch": 0.66, + "learning_rate": 1.7804590051310676e-05, + "loss": 0.9967, + "step": 10332 + }, + { + "epoch": 0.66, + "learning_rate": 1.7804375362287728e-05, + "loss": 1.0456, + "step": 10333 + }, + { + "epoch": 0.66, + "learning_rate": 1.7804160673264777e-05, + "loss": 1.0307, + "step": 10334 + }, + { + "epoch": 0.66, + "learning_rate": 1.7803945984241826e-05, + "loss": 1.0467, + "step": 10335 + }, + { + "epoch": 0.66, + "learning_rate": 1.780373129521888e-05, + "loss": 1.173, + "step": 10336 + }, + { + "epoch": 0.67, + "learning_rate": 1.7803516606195927e-05, + "loss": 1.0999, + "step": 10337 + }, + { + "epoch": 0.67, + "learning_rate": 1.7803301917172976e-05, + "loss": 1.0582, + "step": 10338 + }, + { + "epoch": 0.67, + "learning_rate": 1.7803087228150025e-05, + "loss": 1.0788, + "step": 10339 + }, + { + "epoch": 0.67, + "learning_rate": 1.7802872539127078e-05, + "loss": 1.0849, + "step": 10340 + }, + { + "epoch": 0.67, + "learning_rate": 1.7802657850104127e-05, + "loss": 0.9831, + "step": 10341 + }, + { + "epoch": 0.67, + "learning_rate": 1.7802443161081176e-05, + "loss": 1.1437, + "step": 10342 + }, + { + "epoch": 0.67, + "learning_rate": 1.7802228472058225e-05, + "loss": 1.1578, + "step": 10343 + }, + { + "epoch": 0.67, + "learning_rate": 1.7802013783035274e-05, + "loss": 1.0932, + "step": 10344 + }, + { + "epoch": 0.67, + "learning_rate": 1.7801799094012326e-05, + "loss": 1.2066, + "step": 10345 + }, + { + "epoch": 0.67, + "learning_rate": 1.7801584404989375e-05, + "loss": 1.0342, + "step": 10346 + }, + { + "epoch": 0.67, + "learning_rate": 1.7801369715966424e-05, + "loss": 1.1225, + "step": 10347 + }, + { + "epoch": 0.67, + "learning_rate": 1.7801155026943473e-05, + "loss": 1.031, + "step": 10348 + }, + { + "epoch": 0.67, + "learning_rate": 1.7800940337920522e-05, + "loss": 1.0011, + "step": 10349 + }, + { + "epoch": 0.67, + "learning_rate": 1.7800725648897574e-05, + "loss": 1.0584, + "step": 10350 + }, + { + "epoch": 0.67, + "learning_rate": 1.7800510959874623e-05, + "loss": 1.0949, + "step": 10351 + }, + { + "epoch": 0.67, + "learning_rate": 1.7800296270851672e-05, + "loss": 1.0841, + "step": 10352 + }, + { + "epoch": 0.67, + "learning_rate": 1.7800081581828724e-05, + "loss": 1.0157, + "step": 10353 + }, + { + "epoch": 0.67, + "learning_rate": 1.7799866892805773e-05, + "loss": 1.0281, + "step": 10354 + }, + { + "epoch": 0.67, + "learning_rate": 1.7799652203782822e-05, + "loss": 1.1017, + "step": 10355 + }, + { + "epoch": 0.67, + "learning_rate": 1.779943751475987e-05, + "loss": 1.0399, + "step": 10356 + }, + { + "epoch": 0.67, + "learning_rate": 1.779922282573692e-05, + "loss": 1.1046, + "step": 10357 + }, + { + "epoch": 0.67, + "learning_rate": 1.7799008136713973e-05, + "loss": 1.0539, + "step": 10358 + }, + { + "epoch": 0.67, + "learning_rate": 1.779879344769102e-05, + "loss": 1.0669, + "step": 10359 + }, + { + "epoch": 0.67, + "learning_rate": 1.779857875866807e-05, + "loss": 1.0161, + "step": 10360 + }, + { + "epoch": 0.67, + "learning_rate": 1.779836406964512e-05, + "loss": 1.1125, + "step": 10361 + }, + { + "epoch": 0.67, + "learning_rate": 1.779814938062217e-05, + "loss": 1.116, + "step": 10362 + }, + { + "epoch": 0.67, + "learning_rate": 1.7797934691599217e-05, + "loss": 1.0817, + "step": 10363 + }, + { + "epoch": 0.67, + "learning_rate": 1.779772000257627e-05, + "loss": 0.9378, + "step": 10364 + }, + { + "epoch": 0.67, + "learning_rate": 1.7797505313553322e-05, + "loss": 1.0642, + "step": 10365 + }, + { + "epoch": 0.67, + "learning_rate": 1.779729062453037e-05, + "loss": 1.024, + "step": 10366 + }, + { + "epoch": 0.67, + "learning_rate": 1.779707593550742e-05, + "loss": 1.0715, + "step": 10367 + }, + { + "epoch": 0.67, + "learning_rate": 1.779686124648447e-05, + "loss": 0.9713, + "step": 10368 + }, + { + "epoch": 0.67, + "learning_rate": 1.7796646557461518e-05, + "loss": 1.0192, + "step": 10369 + }, + { + "epoch": 0.67, + "learning_rate": 1.7796431868438567e-05, + "loss": 1.0501, + "step": 10370 + }, + { + "epoch": 0.67, + "learning_rate": 1.779621717941562e-05, + "loss": 0.9866, + "step": 10371 + }, + { + "epoch": 0.67, + "learning_rate": 1.7796002490392668e-05, + "loss": 1.0416, + "step": 10372 + }, + { + "epoch": 0.67, + "learning_rate": 1.7795787801369717e-05, + "loss": 1.0698, + "step": 10373 + }, + { + "epoch": 0.67, + "learning_rate": 1.7795573112346766e-05, + "loss": 1.0756, + "step": 10374 + }, + { + "epoch": 0.67, + "learning_rate": 1.7795358423323815e-05, + "loss": 1.0473, + "step": 10375 + }, + { + "epoch": 0.67, + "learning_rate": 1.7795143734300868e-05, + "loss": 1.0961, + "step": 10376 + }, + { + "epoch": 0.67, + "learning_rate": 1.7794929045277916e-05, + "loss": 1.1001, + "step": 10377 + }, + { + "epoch": 0.67, + "learning_rate": 1.7794714356254965e-05, + "loss": 1.0944, + "step": 10378 + }, + { + "epoch": 0.67, + "learning_rate": 1.7794499667232018e-05, + "loss": 1.1209, + "step": 10379 + }, + { + "epoch": 0.67, + "learning_rate": 1.7794284978209067e-05, + "loss": 1.122, + "step": 10380 + }, + { + "epoch": 0.67, + "learning_rate": 1.7794070289186116e-05, + "loss": 0.963, + "step": 10381 + }, + { + "epoch": 0.67, + "learning_rate": 1.7793855600163165e-05, + "loss": 1.0227, + "step": 10382 + }, + { + "epoch": 0.67, + "learning_rate": 1.7793640911140214e-05, + "loss": 1.0871, + "step": 10383 + }, + { + "epoch": 0.67, + "learning_rate": 1.7793426222117266e-05, + "loss": 1.0789, + "step": 10384 + }, + { + "epoch": 0.67, + "learning_rate": 1.7793211533094315e-05, + "loss": 1.0413, + "step": 10385 + }, + { + "epoch": 0.67, + "learning_rate": 1.7792996844071364e-05, + "loss": 1.0647, + "step": 10386 + }, + { + "epoch": 0.67, + "learning_rate": 1.7792782155048413e-05, + "loss": 0.9816, + "step": 10387 + }, + { + "epoch": 0.67, + "learning_rate": 1.7792567466025462e-05, + "loss": 1.0678, + "step": 10388 + }, + { + "epoch": 0.67, + "learning_rate": 1.7792352777002514e-05, + "loss": 1.1925, + "step": 10389 + }, + { + "epoch": 0.67, + "learning_rate": 1.7792138087979563e-05, + "loss": 1.115, + "step": 10390 + }, + { + "epoch": 0.67, + "learning_rate": 1.7791923398956612e-05, + "loss": 1.0854, + "step": 10391 + }, + { + "epoch": 0.67, + "learning_rate": 1.779170870993366e-05, + "loss": 1.0642, + "step": 10392 + }, + { + "epoch": 0.67, + "learning_rate": 1.779149402091071e-05, + "loss": 1.0348, + "step": 10393 + }, + { + "epoch": 0.67, + "learning_rate": 1.7791279331887762e-05, + "loss": 0.9701, + "step": 10394 + }, + { + "epoch": 0.67, + "learning_rate": 1.779106464286481e-05, + "loss": 1.0704, + "step": 10395 + }, + { + "epoch": 0.67, + "learning_rate": 1.7790849953841864e-05, + "loss": 1.0616, + "step": 10396 + }, + { + "epoch": 0.67, + "learning_rate": 1.7790635264818913e-05, + "loss": 1.0613, + "step": 10397 + }, + { + "epoch": 0.67, + "learning_rate": 1.779042057579596e-05, + "loss": 1.0208, + "step": 10398 + }, + { + "epoch": 0.67, + "learning_rate": 1.779020588677301e-05, + "loss": 1.0572, + "step": 10399 + }, + { + "epoch": 0.67, + "learning_rate": 1.778999119775006e-05, + "loss": 1.1098, + "step": 10400 + }, + { + "epoch": 0.67, + "learning_rate": 1.778977650872711e-05, + "loss": 1.0472, + "step": 10401 + }, + { + "epoch": 0.67, + "learning_rate": 1.778956181970416e-05, + "loss": 1.0768, + "step": 10402 + }, + { + "epoch": 0.67, + "learning_rate": 1.778934713068121e-05, + "loss": 0.986, + "step": 10403 + }, + { + "epoch": 0.67, + "learning_rate": 1.778913244165826e-05, + "loss": 1.112, + "step": 10404 + }, + { + "epoch": 0.67, + "learning_rate": 1.7788917752635308e-05, + "loss": 1.0801, + "step": 10405 + }, + { + "epoch": 0.67, + "learning_rate": 1.7788703063612357e-05, + "loss": 1.1057, + "step": 10406 + }, + { + "epoch": 0.67, + "learning_rate": 1.778848837458941e-05, + "loss": 1.1621, + "step": 10407 + }, + { + "epoch": 0.67, + "learning_rate": 1.7788273685566458e-05, + "loss": 1.0456, + "step": 10408 + }, + { + "epoch": 0.67, + "learning_rate": 1.778805899654351e-05, + "loss": 1.0799, + "step": 10409 + }, + { + "epoch": 0.67, + "learning_rate": 1.778784430752056e-05, + "loss": 0.9779, + "step": 10410 + }, + { + "epoch": 0.67, + "learning_rate": 1.778762961849761e-05, + "loss": 1.0973, + "step": 10411 + }, + { + "epoch": 0.67, + "learning_rate": 1.7787414929474657e-05, + "loss": 1.0471, + "step": 10412 + }, + { + "epoch": 0.67, + "learning_rate": 1.7787200240451706e-05, + "loss": 1.0546, + "step": 10413 + }, + { + "epoch": 0.67, + "learning_rate": 1.778698555142876e-05, + "loss": 1.0244, + "step": 10414 + }, + { + "epoch": 0.67, + "learning_rate": 1.7786770862405808e-05, + "loss": 1.065, + "step": 10415 + }, + { + "epoch": 0.67, + "learning_rate": 1.7786556173382857e-05, + "loss": 1.0605, + "step": 10416 + }, + { + "epoch": 0.67, + "learning_rate": 1.7786341484359906e-05, + "loss": 0.9851, + "step": 10417 + }, + { + "epoch": 0.67, + "learning_rate": 1.7786126795336955e-05, + "loss": 1.0665, + "step": 10418 + }, + { + "epoch": 0.67, + "learning_rate": 1.7785912106314003e-05, + "loss": 1.0286, + "step": 10419 + }, + { + "epoch": 0.67, + "learning_rate": 1.7785697417291056e-05, + "loss": 1.0568, + "step": 10420 + }, + { + "epoch": 0.67, + "learning_rate": 1.7785482728268105e-05, + "loss": 1.0641, + "step": 10421 + }, + { + "epoch": 0.67, + "learning_rate": 1.7785268039245154e-05, + "loss": 1.0495, + "step": 10422 + }, + { + "epoch": 0.67, + "learning_rate": 1.7785053350222206e-05, + "loss": 1.0288, + "step": 10423 + }, + { + "epoch": 0.67, + "learning_rate": 1.7784838661199255e-05, + "loss": 1.058, + "step": 10424 + }, + { + "epoch": 0.67, + "learning_rate": 1.7784623972176304e-05, + "loss": 1.072, + "step": 10425 + }, + { + "epoch": 0.67, + "learning_rate": 1.7784409283153353e-05, + "loss": 1.0787, + "step": 10426 + }, + { + "epoch": 0.67, + "learning_rate": 1.7784194594130405e-05, + "loss": 1.103, + "step": 10427 + }, + { + "epoch": 0.67, + "learning_rate": 1.7783979905107454e-05, + "loss": 1.1082, + "step": 10428 + }, + { + "epoch": 0.67, + "learning_rate": 1.7783765216084503e-05, + "loss": 0.9909, + "step": 10429 + }, + { + "epoch": 0.67, + "learning_rate": 1.7783550527061552e-05, + "loss": 1.146, + "step": 10430 + }, + { + "epoch": 0.67, + "learning_rate": 1.77833358380386e-05, + "loss": 1.0695, + "step": 10431 + }, + { + "epoch": 0.67, + "learning_rate": 1.7783121149015654e-05, + "loss": 1.1563, + "step": 10432 + }, + { + "epoch": 0.67, + "learning_rate": 1.7782906459992703e-05, + "loss": 1.0066, + "step": 10433 + }, + { + "epoch": 0.67, + "learning_rate": 1.778269177096975e-05, + "loss": 1.1851, + "step": 10434 + }, + { + "epoch": 0.67, + "learning_rate": 1.77824770819468e-05, + "loss": 1.1045, + "step": 10435 + }, + { + "epoch": 0.67, + "learning_rate": 1.778226239292385e-05, + "loss": 1.0411, + "step": 10436 + }, + { + "epoch": 0.67, + "learning_rate": 1.7782047703900902e-05, + "loss": 1.1119, + "step": 10437 + }, + { + "epoch": 0.67, + "learning_rate": 1.778183301487795e-05, + "loss": 1.0931, + "step": 10438 + }, + { + "epoch": 0.67, + "learning_rate": 1.7781618325855e-05, + "loss": 1.0676, + "step": 10439 + }, + { + "epoch": 0.67, + "learning_rate": 1.7781403636832052e-05, + "loss": 1.0698, + "step": 10440 + }, + { + "epoch": 0.67, + "learning_rate": 1.77811889478091e-05, + "loss": 1.0256, + "step": 10441 + }, + { + "epoch": 0.67, + "learning_rate": 1.778097425878615e-05, + "loss": 1.1025, + "step": 10442 + }, + { + "epoch": 0.67, + "learning_rate": 1.77807595697632e-05, + "loss": 1.0941, + "step": 10443 + }, + { + "epoch": 0.67, + "learning_rate": 1.7780544880740248e-05, + "loss": 1.0748, + "step": 10444 + }, + { + "epoch": 0.67, + "learning_rate": 1.77803301917173e-05, + "loss": 1.0628, + "step": 10445 + }, + { + "epoch": 0.67, + "learning_rate": 1.778011550269435e-05, + "loss": 1.077, + "step": 10446 + }, + { + "epoch": 0.67, + "learning_rate": 1.7779900813671398e-05, + "loss": 1.0445, + "step": 10447 + }, + { + "epoch": 0.67, + "learning_rate": 1.7779686124648447e-05, + "loss": 1.1254, + "step": 10448 + }, + { + "epoch": 0.67, + "learning_rate": 1.7779471435625496e-05, + "loss": 1.1515, + "step": 10449 + }, + { + "epoch": 0.67, + "learning_rate": 1.7779256746602545e-05, + "loss": 1.0429, + "step": 10450 + }, + { + "epoch": 0.67, + "learning_rate": 1.7779042057579597e-05, + "loss": 1.0438, + "step": 10451 + }, + { + "epoch": 0.67, + "learning_rate": 1.777882736855665e-05, + "loss": 1.0916, + "step": 10452 + }, + { + "epoch": 0.67, + "learning_rate": 1.77786126795337e-05, + "loss": 1.0285, + "step": 10453 + }, + { + "epoch": 0.67, + "learning_rate": 1.7778397990510748e-05, + "loss": 1.2037, + "step": 10454 + }, + { + "epoch": 0.67, + "learning_rate": 1.7778183301487797e-05, + "loss": 1.0737, + "step": 10455 + }, + { + "epoch": 0.67, + "learning_rate": 1.7777968612464846e-05, + "loss": 1.0328, + "step": 10456 + }, + { + "epoch": 0.67, + "learning_rate": 1.7777753923441895e-05, + "loss": 1.1103, + "step": 10457 + }, + { + "epoch": 0.67, + "learning_rate": 1.7777539234418947e-05, + "loss": 1.1129, + "step": 10458 + }, + { + "epoch": 0.67, + "learning_rate": 1.7777324545395996e-05, + "loss": 1.1172, + "step": 10459 + }, + { + "epoch": 0.67, + "learning_rate": 1.7777109856373045e-05, + "loss": 1.0538, + "step": 10460 + }, + { + "epoch": 0.67, + "learning_rate": 1.7776895167350094e-05, + "loss": 1.0606, + "step": 10461 + }, + { + "epoch": 0.67, + "learning_rate": 1.7776680478327143e-05, + "loss": 1.0703, + "step": 10462 + }, + { + "epoch": 0.67, + "learning_rate": 1.7776465789304195e-05, + "loss": 1.047, + "step": 10463 + }, + { + "epoch": 0.67, + "learning_rate": 1.7776251100281244e-05, + "loss": 1.0455, + "step": 10464 + }, + { + "epoch": 0.67, + "learning_rate": 1.7776036411258293e-05, + "loss": 1.1146, + "step": 10465 + }, + { + "epoch": 0.67, + "learning_rate": 1.7775821722235346e-05, + "loss": 1.1119, + "step": 10466 + }, + { + "epoch": 0.67, + "learning_rate": 1.7775607033212394e-05, + "loss": 1.0587, + "step": 10467 + }, + { + "epoch": 0.67, + "learning_rate": 1.7775392344189443e-05, + "loss": 1.0677, + "step": 10468 + }, + { + "epoch": 0.67, + "learning_rate": 1.7775177655166492e-05, + "loss": 1.0447, + "step": 10469 + }, + { + "epoch": 0.67, + "learning_rate": 1.777496296614354e-05, + "loss": 1.0167, + "step": 10470 + }, + { + "epoch": 0.67, + "learning_rate": 1.7774748277120594e-05, + "loss": 1.0459, + "step": 10471 + }, + { + "epoch": 0.67, + "learning_rate": 1.7774533588097643e-05, + "loss": 1.1478, + "step": 10472 + }, + { + "epoch": 0.67, + "learning_rate": 1.777431889907469e-05, + "loss": 1.095, + "step": 10473 + }, + { + "epoch": 0.67, + "learning_rate": 1.777410421005174e-05, + "loss": 1.107, + "step": 10474 + }, + { + "epoch": 0.67, + "learning_rate": 1.777388952102879e-05, + "loss": 1.0505, + "step": 10475 + }, + { + "epoch": 0.67, + "learning_rate": 1.7773674832005842e-05, + "loss": 1.0723, + "step": 10476 + }, + { + "epoch": 0.67, + "learning_rate": 1.777346014298289e-05, + "loss": 1.1145, + "step": 10477 + }, + { + "epoch": 0.67, + "learning_rate": 1.777324545395994e-05, + "loss": 1.0809, + "step": 10478 + }, + { + "epoch": 0.67, + "learning_rate": 1.777303076493699e-05, + "loss": 1.046, + "step": 10479 + }, + { + "epoch": 0.67, + "learning_rate": 1.7772816075914038e-05, + "loss": 1.078, + "step": 10480 + }, + { + "epoch": 0.67, + "learning_rate": 1.777260138689109e-05, + "loss": 1.0961, + "step": 10481 + }, + { + "epoch": 0.67, + "learning_rate": 1.777238669786814e-05, + "loss": 1.0054, + "step": 10482 + }, + { + "epoch": 0.67, + "learning_rate": 1.777217200884519e-05, + "loss": 1.1311, + "step": 10483 + }, + { + "epoch": 0.67, + "learning_rate": 1.777195731982224e-05, + "loss": 1.0045, + "step": 10484 + }, + { + "epoch": 0.67, + "learning_rate": 1.777174263079929e-05, + "loss": 1.0181, + "step": 10485 + }, + { + "epoch": 0.67, + "learning_rate": 1.777152794177634e-05, + "loss": 1.1151, + "step": 10486 + }, + { + "epoch": 0.67, + "learning_rate": 1.7771313252753387e-05, + "loss": 1.0442, + "step": 10487 + }, + { + "epoch": 0.67, + "learning_rate": 1.7771098563730436e-05, + "loss": 1.028, + "step": 10488 + }, + { + "epoch": 0.67, + "learning_rate": 1.777088387470749e-05, + "loss": 1.1175, + "step": 10489 + }, + { + "epoch": 0.67, + "learning_rate": 1.7770669185684538e-05, + "loss": 1.0614, + "step": 10490 + }, + { + "epoch": 0.67, + "learning_rate": 1.7770454496661587e-05, + "loss": 1.0569, + "step": 10491 + }, + { + "epoch": 0.68, + "learning_rate": 1.7770239807638636e-05, + "loss": 1.0501, + "step": 10492 + }, + { + "epoch": 0.68, + "learning_rate": 1.7770025118615684e-05, + "loss": 1.0953, + "step": 10493 + }, + { + "epoch": 0.68, + "learning_rate": 1.7769810429592737e-05, + "loss": 1.1285, + "step": 10494 + }, + { + "epoch": 0.68, + "learning_rate": 1.7769595740569786e-05, + "loss": 1.0901, + "step": 10495 + }, + { + "epoch": 0.68, + "learning_rate": 1.7769381051546838e-05, + "loss": 1.1318, + "step": 10496 + }, + { + "epoch": 0.68, + "learning_rate": 1.7769166362523887e-05, + "loss": 1.0241, + "step": 10497 + }, + { + "epoch": 0.68, + "learning_rate": 1.7768951673500936e-05, + "loss": 1.0449, + "step": 10498 + }, + { + "epoch": 0.68, + "learning_rate": 1.7768736984477985e-05, + "loss": 1.0511, + "step": 10499 + }, + { + "epoch": 0.68, + "learning_rate": 1.7768522295455034e-05, + "loss": 1.0177, + "step": 10500 + }, + { + "epoch": 0.68, + "learning_rate": 1.7768307606432086e-05, + "loss": 0.9575, + "step": 10501 + }, + { + "epoch": 0.68, + "learning_rate": 1.7768092917409135e-05, + "loss": 1.0697, + "step": 10502 + }, + { + "epoch": 0.68, + "learning_rate": 1.7767878228386184e-05, + "loss": 1.0899, + "step": 10503 + }, + { + "epoch": 0.68, + "learning_rate": 1.7767663539363233e-05, + "loss": 1.0821, + "step": 10504 + }, + { + "epoch": 0.68, + "learning_rate": 1.7767448850340282e-05, + "loss": 1.1389, + "step": 10505 + }, + { + "epoch": 0.68, + "learning_rate": 1.776723416131733e-05, + "loss": 1.0569, + "step": 10506 + }, + { + "epoch": 0.68, + "learning_rate": 1.7767019472294384e-05, + "loss": 1.05, + "step": 10507 + }, + { + "epoch": 0.68, + "learning_rate": 1.7766804783271433e-05, + "loss": 1.0234, + "step": 10508 + }, + { + "epoch": 0.68, + "learning_rate": 1.776659009424848e-05, + "loss": 1.0063, + "step": 10509 + }, + { + "epoch": 0.68, + "learning_rate": 1.7766375405225534e-05, + "loss": 1.0655, + "step": 10510 + }, + { + "epoch": 0.68, + "learning_rate": 1.7766160716202583e-05, + "loss": 1.1237, + "step": 10511 + }, + { + "epoch": 0.68, + "learning_rate": 1.7765946027179632e-05, + "loss": 1.0635, + "step": 10512 + }, + { + "epoch": 0.68, + "learning_rate": 1.776573133815668e-05, + "loss": 0.9888, + "step": 10513 + }, + { + "epoch": 0.68, + "learning_rate": 1.7765516649133733e-05, + "loss": 1.0456, + "step": 10514 + }, + { + "epoch": 0.68, + "learning_rate": 1.7765301960110782e-05, + "loss": 0.9892, + "step": 10515 + }, + { + "epoch": 0.68, + "learning_rate": 1.776508727108783e-05, + "loss": 1.0711, + "step": 10516 + }, + { + "epoch": 0.68, + "learning_rate": 1.776487258206488e-05, + "loss": 1.161, + "step": 10517 + }, + { + "epoch": 0.68, + "learning_rate": 1.776465789304193e-05, + "loss": 1.0584, + "step": 10518 + }, + { + "epoch": 0.68, + "learning_rate": 1.7764443204018978e-05, + "loss": 1.0828, + "step": 10519 + }, + { + "epoch": 0.68, + "learning_rate": 1.776422851499603e-05, + "loss": 1.0554, + "step": 10520 + }, + { + "epoch": 0.68, + "learning_rate": 1.776401382597308e-05, + "loss": 1.0888, + "step": 10521 + }, + { + "epoch": 0.68, + "learning_rate": 1.7763799136950128e-05, + "loss": 1.0664, + "step": 10522 + }, + { + "epoch": 0.68, + "learning_rate": 1.7763584447927177e-05, + "loss": 1.0128, + "step": 10523 + }, + { + "epoch": 0.68, + "learning_rate": 1.776336975890423e-05, + "loss": 1.0568, + "step": 10524 + }, + { + "epoch": 0.68, + "learning_rate": 1.776315506988128e-05, + "loss": 0.9554, + "step": 10525 + }, + { + "epoch": 0.68, + "learning_rate": 1.7762940380858327e-05, + "loss": 1.0818, + "step": 10526 + }, + { + "epoch": 0.68, + "learning_rate": 1.776272569183538e-05, + "loss": 1.1386, + "step": 10527 + }, + { + "epoch": 0.68, + "learning_rate": 1.776251100281243e-05, + "loss": 1.0504, + "step": 10528 + }, + { + "epoch": 0.68, + "learning_rate": 1.7762296313789478e-05, + "loss": 0.9949, + "step": 10529 + }, + { + "epoch": 0.68, + "learning_rate": 1.7762081624766527e-05, + "loss": 1.0987, + "step": 10530 + }, + { + "epoch": 0.68, + "learning_rate": 1.7761866935743576e-05, + "loss": 1.0514, + "step": 10531 + }, + { + "epoch": 0.68, + "learning_rate": 1.7761652246720628e-05, + "loss": 1.0716, + "step": 10532 + }, + { + "epoch": 0.68, + "learning_rate": 1.7761437557697677e-05, + "loss": 1.0364, + "step": 10533 + }, + { + "epoch": 0.68, + "learning_rate": 1.7761222868674726e-05, + "loss": 1.0495, + "step": 10534 + }, + { + "epoch": 0.68, + "learning_rate": 1.7761008179651775e-05, + "loss": 1.1384, + "step": 10535 + }, + { + "epoch": 0.68, + "learning_rate": 1.7760793490628824e-05, + "loss": 1.0064, + "step": 10536 + }, + { + "epoch": 0.68, + "learning_rate": 1.7760578801605873e-05, + "loss": 1.0648, + "step": 10537 + }, + { + "epoch": 0.68, + "learning_rate": 1.7760364112582925e-05, + "loss": 1.0807, + "step": 10538 + }, + { + "epoch": 0.68, + "learning_rate": 1.7760149423559974e-05, + "loss": 1.0454, + "step": 10539 + }, + { + "epoch": 0.68, + "learning_rate": 1.7759934734537027e-05, + "loss": 1.026, + "step": 10540 + }, + { + "epoch": 0.68, + "learning_rate": 1.7759720045514075e-05, + "loss": 1.1175, + "step": 10541 + }, + { + "epoch": 0.68, + "learning_rate": 1.7759505356491124e-05, + "loss": 0.9727, + "step": 10542 + }, + { + "epoch": 0.68, + "learning_rate": 1.7759290667468173e-05, + "loss": 1.0693, + "step": 10543 + }, + { + "epoch": 0.68, + "learning_rate": 1.7759075978445222e-05, + "loss": 1.1203, + "step": 10544 + }, + { + "epoch": 0.68, + "learning_rate": 1.7758861289422275e-05, + "loss": 1.0598, + "step": 10545 + }, + { + "epoch": 0.68, + "learning_rate": 1.7758646600399324e-05, + "loss": 1.1347, + "step": 10546 + }, + { + "epoch": 0.68, + "learning_rate": 1.7758431911376373e-05, + "loss": 1.1617, + "step": 10547 + }, + { + "epoch": 0.68, + "learning_rate": 1.775821722235342e-05, + "loss": 1.0258, + "step": 10548 + }, + { + "epoch": 0.68, + "learning_rate": 1.775800253333047e-05, + "loss": 1.0955, + "step": 10549 + }, + { + "epoch": 0.68, + "learning_rate": 1.7757787844307523e-05, + "loss": 1.0302, + "step": 10550 + }, + { + "epoch": 0.68, + "learning_rate": 1.7757573155284572e-05, + "loss": 1.1362, + "step": 10551 + }, + { + "epoch": 0.68, + "learning_rate": 1.775735846626162e-05, + "loss": 1.0723, + "step": 10552 + }, + { + "epoch": 0.68, + "learning_rate": 1.7757143777238673e-05, + "loss": 1.0702, + "step": 10553 + }, + { + "epoch": 0.68, + "learning_rate": 1.7756929088215722e-05, + "loss": 0.9722, + "step": 10554 + }, + { + "epoch": 0.68, + "learning_rate": 1.775671439919277e-05, + "loss": 1.0776, + "step": 10555 + }, + { + "epoch": 0.68, + "learning_rate": 1.775649971016982e-05, + "loss": 1.1133, + "step": 10556 + }, + { + "epoch": 0.68, + "learning_rate": 1.775628502114687e-05, + "loss": 1.109, + "step": 10557 + }, + { + "epoch": 0.68, + "learning_rate": 1.775607033212392e-05, + "loss": 1.1012, + "step": 10558 + }, + { + "epoch": 0.68, + "learning_rate": 1.775585564310097e-05, + "loss": 1.0329, + "step": 10559 + }, + { + "epoch": 0.68, + "learning_rate": 1.775564095407802e-05, + "loss": 1.121, + "step": 10560 + }, + { + "epoch": 0.68, + "learning_rate": 1.775542626505507e-05, + "loss": 1.0166, + "step": 10561 + }, + { + "epoch": 0.68, + "learning_rate": 1.7755211576032117e-05, + "loss": 1.0443, + "step": 10562 + }, + { + "epoch": 0.68, + "learning_rate": 1.775499688700917e-05, + "loss": 1.06, + "step": 10563 + }, + { + "epoch": 0.68, + "learning_rate": 1.775478219798622e-05, + "loss": 1.0791, + "step": 10564 + }, + { + "epoch": 0.68, + "learning_rate": 1.7754567508963268e-05, + "loss": 1.1097, + "step": 10565 + }, + { + "epoch": 0.68, + "learning_rate": 1.7754352819940317e-05, + "loss": 1.0626, + "step": 10566 + }, + { + "epoch": 0.68, + "learning_rate": 1.7754138130917366e-05, + "loss": 1.099, + "step": 10567 + }, + { + "epoch": 0.68, + "learning_rate": 1.7753923441894418e-05, + "loss": 1.0712, + "step": 10568 + }, + { + "epoch": 0.68, + "learning_rate": 1.7753708752871467e-05, + "loss": 1.0162, + "step": 10569 + }, + { + "epoch": 0.68, + "learning_rate": 1.775349406384852e-05, + "loss": 1.1086, + "step": 10570 + }, + { + "epoch": 0.68, + "learning_rate": 1.7753279374825568e-05, + "loss": 1.1323, + "step": 10571 + }, + { + "epoch": 0.68, + "learning_rate": 1.7753064685802617e-05, + "loss": 1.0705, + "step": 10572 + }, + { + "epoch": 0.68, + "learning_rate": 1.7752849996779666e-05, + "loss": 1.1159, + "step": 10573 + }, + { + "epoch": 0.68, + "learning_rate": 1.7752635307756715e-05, + "loss": 1.122, + "step": 10574 + }, + { + "epoch": 0.68, + "learning_rate": 1.7752420618733764e-05, + "loss": 0.9804, + "step": 10575 + }, + { + "epoch": 0.68, + "learning_rate": 1.7752205929710816e-05, + "loss": 1.0096, + "step": 10576 + }, + { + "epoch": 0.68, + "learning_rate": 1.7751991240687865e-05, + "loss": 1.0835, + "step": 10577 + }, + { + "epoch": 0.68, + "learning_rate": 1.7751776551664914e-05, + "loss": 1.0651, + "step": 10578 + }, + { + "epoch": 0.68, + "learning_rate": 1.7751561862641963e-05, + "loss": 1.0488, + "step": 10579 + }, + { + "epoch": 0.68, + "learning_rate": 1.7751347173619012e-05, + "loss": 1.0498, + "step": 10580 + }, + { + "epoch": 0.68, + "learning_rate": 1.7751132484596065e-05, + "loss": 1.1131, + "step": 10581 + }, + { + "epoch": 0.68, + "learning_rate": 1.7750917795573114e-05, + "loss": 1.1295, + "step": 10582 + }, + { + "epoch": 0.68, + "learning_rate": 1.7750703106550166e-05, + "loss": 1.0804, + "step": 10583 + }, + { + "epoch": 0.68, + "learning_rate": 1.7750488417527215e-05, + "loss": 1.0547, + "step": 10584 + }, + { + "epoch": 0.68, + "learning_rate": 1.7750273728504264e-05, + "loss": 1.0868, + "step": 10585 + }, + { + "epoch": 0.68, + "learning_rate": 1.7750059039481313e-05, + "loss": 1.1223, + "step": 10586 + }, + { + "epoch": 0.68, + "learning_rate": 1.7749844350458362e-05, + "loss": 1.0181, + "step": 10587 + }, + { + "epoch": 0.68, + "learning_rate": 1.7749629661435414e-05, + "loss": 1.02, + "step": 10588 + }, + { + "epoch": 0.68, + "learning_rate": 1.7749414972412463e-05, + "loss": 1.064, + "step": 10589 + }, + { + "epoch": 0.68, + "learning_rate": 1.7749200283389512e-05, + "loss": 1.1004, + "step": 10590 + }, + { + "epoch": 0.68, + "learning_rate": 1.774898559436656e-05, + "loss": 1.0456, + "step": 10591 + }, + { + "epoch": 0.68, + "learning_rate": 1.774877090534361e-05, + "loss": 1.1636, + "step": 10592 + }, + { + "epoch": 0.68, + "learning_rate": 1.774855621632066e-05, + "loss": 1.0087, + "step": 10593 + }, + { + "epoch": 0.68, + "learning_rate": 1.774834152729771e-05, + "loss": 1.1094, + "step": 10594 + }, + { + "epoch": 0.68, + "learning_rate": 1.774812683827476e-05, + "loss": 1.106, + "step": 10595 + }, + { + "epoch": 0.68, + "learning_rate": 1.774791214925181e-05, + "loss": 1.0054, + "step": 10596 + }, + { + "epoch": 0.68, + "learning_rate": 1.774769746022886e-05, + "loss": 1.0999, + "step": 10597 + }, + { + "epoch": 0.68, + "learning_rate": 1.774748277120591e-05, + "loss": 1.0614, + "step": 10598 + }, + { + "epoch": 0.68, + "learning_rate": 1.774726808218296e-05, + "loss": 1.1396, + "step": 10599 + }, + { + "epoch": 0.68, + "learning_rate": 1.774705339316001e-05, + "loss": 1.0787, + "step": 10600 + }, + { + "epoch": 0.68, + "learning_rate": 1.774683870413706e-05, + "loss": 1.1308, + "step": 10601 + }, + { + "epoch": 0.68, + "learning_rate": 1.774662401511411e-05, + "loss": 1.1795, + "step": 10602 + }, + { + "epoch": 0.68, + "learning_rate": 1.774640932609116e-05, + "loss": 1.037, + "step": 10603 + }, + { + "epoch": 0.68, + "learning_rate": 1.7746194637068208e-05, + "loss": 1.0483, + "step": 10604 + }, + { + "epoch": 0.68, + "learning_rate": 1.7745979948045257e-05, + "loss": 1.0927, + "step": 10605 + }, + { + "epoch": 0.68, + "learning_rate": 1.7745765259022306e-05, + "loss": 1.1191, + "step": 10606 + }, + { + "epoch": 0.68, + "learning_rate": 1.7745550569999358e-05, + "loss": 1.1469, + "step": 10607 + }, + { + "epoch": 0.68, + "learning_rate": 1.7745335880976407e-05, + "loss": 1.0378, + "step": 10608 + }, + { + "epoch": 0.68, + "learning_rate": 1.7745121191953456e-05, + "loss": 1.0767, + "step": 10609 + }, + { + "epoch": 0.68, + "learning_rate": 1.7744906502930505e-05, + "loss": 1.1373, + "step": 10610 + }, + { + "epoch": 0.68, + "learning_rate": 1.7744691813907557e-05, + "loss": 1.0315, + "step": 10611 + }, + { + "epoch": 0.68, + "learning_rate": 1.7744477124884606e-05, + "loss": 1.1828, + "step": 10612 + }, + { + "epoch": 0.68, + "learning_rate": 1.7744262435861655e-05, + "loss": 1.102, + "step": 10613 + }, + { + "epoch": 0.68, + "learning_rate": 1.7744047746838708e-05, + "loss": 1.1031, + "step": 10614 + }, + { + "epoch": 0.68, + "learning_rate": 1.7743833057815756e-05, + "loss": 1.1155, + "step": 10615 + }, + { + "epoch": 0.68, + "learning_rate": 1.7743618368792805e-05, + "loss": 1.052, + "step": 10616 + }, + { + "epoch": 0.68, + "learning_rate": 1.7743403679769854e-05, + "loss": 1.0782, + "step": 10617 + }, + { + "epoch": 0.68, + "learning_rate": 1.7743188990746903e-05, + "loss": 1.1332, + "step": 10618 + }, + { + "epoch": 0.68, + "learning_rate": 1.7742974301723956e-05, + "loss": 1.1117, + "step": 10619 + }, + { + "epoch": 0.68, + "learning_rate": 1.7742759612701005e-05, + "loss": 1.0312, + "step": 10620 + }, + { + "epoch": 0.68, + "learning_rate": 1.7742544923678054e-05, + "loss": 1.0536, + "step": 10621 + }, + { + "epoch": 0.68, + "learning_rate": 1.7742330234655103e-05, + "loss": 0.9649, + "step": 10622 + }, + { + "epoch": 0.68, + "learning_rate": 1.774211554563215e-05, + "loss": 1.1258, + "step": 10623 + }, + { + "epoch": 0.68, + "learning_rate": 1.77419008566092e-05, + "loss": 1.1683, + "step": 10624 + }, + { + "epoch": 0.68, + "learning_rate": 1.7741686167586253e-05, + "loss": 1.086, + "step": 10625 + }, + { + "epoch": 0.68, + "learning_rate": 1.7741471478563302e-05, + "loss": 1.0112, + "step": 10626 + }, + { + "epoch": 0.68, + "learning_rate": 1.7741256789540354e-05, + "loss": 1.073, + "step": 10627 + }, + { + "epoch": 0.68, + "learning_rate": 1.7741042100517403e-05, + "loss": 1.1106, + "step": 10628 + }, + { + "epoch": 0.68, + "learning_rate": 1.7740827411494452e-05, + "loss": 1.1388, + "step": 10629 + }, + { + "epoch": 0.68, + "learning_rate": 1.77406127224715e-05, + "loss": 1.0797, + "step": 10630 + }, + { + "epoch": 0.68, + "learning_rate": 1.774039803344855e-05, + "loss": 1.0498, + "step": 10631 + }, + { + "epoch": 0.68, + "learning_rate": 1.7740183344425602e-05, + "loss": 1.0816, + "step": 10632 + }, + { + "epoch": 0.68, + "learning_rate": 1.773996865540265e-05, + "loss": 1.0323, + "step": 10633 + }, + { + "epoch": 0.68, + "learning_rate": 1.77397539663797e-05, + "loss": 1.0123, + "step": 10634 + }, + { + "epoch": 0.68, + "learning_rate": 1.773953927735675e-05, + "loss": 1.1201, + "step": 10635 + }, + { + "epoch": 0.68, + "learning_rate": 1.7739324588333798e-05, + "loss": 1.0441, + "step": 10636 + }, + { + "epoch": 0.68, + "learning_rate": 1.773910989931085e-05, + "loss": 1.1226, + "step": 10637 + }, + { + "epoch": 0.68, + "learning_rate": 1.77388952102879e-05, + "loss": 1.0302, + "step": 10638 + }, + { + "epoch": 0.68, + "learning_rate": 1.773868052126495e-05, + "loss": 0.9898, + "step": 10639 + }, + { + "epoch": 0.68, + "learning_rate": 1.7738465832242e-05, + "loss": 1.0501, + "step": 10640 + }, + { + "epoch": 0.68, + "learning_rate": 1.773825114321905e-05, + "loss": 1.1013, + "step": 10641 + }, + { + "epoch": 0.68, + "learning_rate": 1.77380364541961e-05, + "loss": 1.1084, + "step": 10642 + }, + { + "epoch": 0.68, + "learning_rate": 1.7737821765173148e-05, + "loss": 1.1222, + "step": 10643 + }, + { + "epoch": 0.68, + "learning_rate": 1.7737607076150197e-05, + "loss": 1.0442, + "step": 10644 + }, + { + "epoch": 0.68, + "learning_rate": 1.773739238712725e-05, + "loss": 1.0987, + "step": 10645 + }, + { + "epoch": 0.68, + "learning_rate": 1.7737177698104298e-05, + "loss": 0.9766, + "step": 10646 + }, + { + "epoch": 0.69, + "learning_rate": 1.7736963009081347e-05, + "loss": 1.0441, + "step": 10647 + }, + { + "epoch": 0.69, + "learning_rate": 1.7736748320058396e-05, + "loss": 1.0092, + "step": 10648 + }, + { + "epoch": 0.69, + "learning_rate": 1.7736533631035445e-05, + "loss": 1.0724, + "step": 10649 + }, + { + "epoch": 0.69, + "learning_rate": 1.7736318942012497e-05, + "loss": 1.0185, + "step": 10650 + }, + { + "epoch": 0.69, + "learning_rate": 1.7736104252989546e-05, + "loss": 1.0758, + "step": 10651 + }, + { + "epoch": 0.69, + "learning_rate": 1.7735889563966595e-05, + "loss": 1.1199, + "step": 10652 + }, + { + "epoch": 0.69, + "learning_rate": 1.7735674874943644e-05, + "loss": 1.0844, + "step": 10653 + }, + { + "epoch": 0.69, + "learning_rate": 1.7735460185920693e-05, + "loss": 1.0838, + "step": 10654 + }, + { + "epoch": 0.69, + "learning_rate": 1.7735245496897746e-05, + "loss": 1.0936, + "step": 10655 + }, + { + "epoch": 0.69, + "learning_rate": 1.7735030807874795e-05, + "loss": 1.0009, + "step": 10656 + }, + { + "epoch": 0.69, + "learning_rate": 1.7734816118851847e-05, + "loss": 1.1395, + "step": 10657 + }, + { + "epoch": 0.69, + "learning_rate": 1.7734601429828896e-05, + "loss": 1.1281, + "step": 10658 + }, + { + "epoch": 0.69, + "learning_rate": 1.7734386740805945e-05, + "loss": 1.0549, + "step": 10659 + }, + { + "epoch": 0.69, + "learning_rate": 1.7734172051782994e-05, + "loss": 1.022, + "step": 10660 + }, + { + "epoch": 0.69, + "learning_rate": 1.7733957362760043e-05, + "loss": 1.0652, + "step": 10661 + }, + { + "epoch": 0.69, + "learning_rate": 1.7733742673737092e-05, + "loss": 1.0509, + "step": 10662 + }, + { + "epoch": 0.69, + "learning_rate": 1.7733527984714144e-05, + "loss": 1.1081, + "step": 10663 + }, + { + "epoch": 0.69, + "learning_rate": 1.7733313295691193e-05, + "loss": 1.0499, + "step": 10664 + }, + { + "epoch": 0.69, + "learning_rate": 1.7733098606668242e-05, + "loss": 1.0412, + "step": 10665 + }, + { + "epoch": 0.69, + "learning_rate": 1.773288391764529e-05, + "loss": 1.0546, + "step": 10666 + }, + { + "epoch": 0.69, + "learning_rate": 1.773266922862234e-05, + "loss": 0.9964, + "step": 10667 + }, + { + "epoch": 0.69, + "learning_rate": 1.7732454539599392e-05, + "loss": 0.9853, + "step": 10668 + }, + { + "epoch": 0.69, + "learning_rate": 1.773223985057644e-05, + "loss": 1.1085, + "step": 10669 + }, + { + "epoch": 0.69, + "learning_rate": 1.7732025161553494e-05, + "loss": 0.9622, + "step": 10670 + }, + { + "epoch": 0.69, + "learning_rate": 1.7731810472530543e-05, + "loss": 1.0584, + "step": 10671 + }, + { + "epoch": 0.69, + "learning_rate": 1.773159578350759e-05, + "loss": 1.063, + "step": 10672 + }, + { + "epoch": 0.69, + "learning_rate": 1.773138109448464e-05, + "loss": 1.0392, + "step": 10673 + }, + { + "epoch": 0.69, + "learning_rate": 1.773116640546169e-05, + "loss": 1.0388, + "step": 10674 + }, + { + "epoch": 0.69, + "learning_rate": 1.7730951716438742e-05, + "loss": 1.0783, + "step": 10675 + }, + { + "epoch": 0.69, + "learning_rate": 1.773073702741579e-05, + "loss": 1.117, + "step": 10676 + }, + { + "epoch": 0.69, + "learning_rate": 1.773052233839284e-05, + "loss": 1.1158, + "step": 10677 + }, + { + "epoch": 0.69, + "learning_rate": 1.773030764936989e-05, + "loss": 0.9959, + "step": 10678 + }, + { + "epoch": 0.69, + "learning_rate": 1.7730092960346938e-05, + "loss": 1.1067, + "step": 10679 + }, + { + "epoch": 0.69, + "learning_rate": 1.7729878271323987e-05, + "loss": 1.1028, + "step": 10680 + }, + { + "epoch": 0.69, + "learning_rate": 1.772966358230104e-05, + "loss": 1.1475, + "step": 10681 + }, + { + "epoch": 0.69, + "learning_rate": 1.7729448893278088e-05, + "loss": 1.0635, + "step": 10682 + }, + { + "epoch": 0.69, + "learning_rate": 1.7729234204255137e-05, + "loss": 1.0979, + "step": 10683 + }, + { + "epoch": 0.69, + "learning_rate": 1.772901951523219e-05, + "loss": 1.0785, + "step": 10684 + }, + { + "epoch": 0.69, + "learning_rate": 1.7728804826209238e-05, + "loss": 1.0611, + "step": 10685 + }, + { + "epoch": 0.69, + "learning_rate": 1.7728590137186287e-05, + "loss": 1.0794, + "step": 10686 + }, + { + "epoch": 0.69, + "learning_rate": 1.7728375448163336e-05, + "loss": 1.0813, + "step": 10687 + }, + { + "epoch": 0.69, + "learning_rate": 1.772816075914039e-05, + "loss": 1.101, + "step": 10688 + }, + { + "epoch": 0.69, + "learning_rate": 1.7727946070117437e-05, + "loss": 1.0415, + "step": 10689 + }, + { + "epoch": 0.69, + "learning_rate": 1.7727731381094486e-05, + "loss": 1.1064, + "step": 10690 + }, + { + "epoch": 0.69, + "learning_rate": 1.7727516692071535e-05, + "loss": 1.1201, + "step": 10691 + }, + { + "epoch": 0.69, + "learning_rate": 1.7727302003048584e-05, + "loss": 1.0427, + "step": 10692 + }, + { + "epoch": 0.69, + "learning_rate": 1.7727087314025633e-05, + "loss": 1.0972, + "step": 10693 + }, + { + "epoch": 0.69, + "learning_rate": 1.7726872625002686e-05, + "loss": 1.1455, + "step": 10694 + }, + { + "epoch": 0.69, + "learning_rate": 1.7726657935979735e-05, + "loss": 0.9739, + "step": 10695 + }, + { + "epoch": 0.69, + "learning_rate": 1.7726443246956784e-05, + "loss": 1.112, + "step": 10696 + }, + { + "epoch": 0.69, + "learning_rate": 1.7726228557933833e-05, + "loss": 1.0267, + "step": 10697 + }, + { + "epoch": 0.69, + "learning_rate": 1.772601386891088e-05, + "loss": 1.154, + "step": 10698 + }, + { + "epoch": 0.69, + "learning_rate": 1.7725799179887934e-05, + "loss": 1.0045, + "step": 10699 + }, + { + "epoch": 0.69, + "learning_rate": 1.7725584490864983e-05, + "loss": 1.089, + "step": 10700 + }, + { + "epoch": 0.69, + "learning_rate": 1.7725369801842035e-05, + "loss": 1.063, + "step": 10701 + }, + { + "epoch": 0.69, + "learning_rate": 1.7725155112819084e-05, + "loss": 1.1237, + "step": 10702 + }, + { + "epoch": 0.69, + "learning_rate": 1.7724940423796133e-05, + "loss": 1.0444, + "step": 10703 + }, + { + "epoch": 0.69, + "learning_rate": 1.7724725734773182e-05, + "loss": 1.0439, + "step": 10704 + }, + { + "epoch": 0.69, + "learning_rate": 1.772451104575023e-05, + "loss": 1.1023, + "step": 10705 + }, + { + "epoch": 0.69, + "learning_rate": 1.7724296356727283e-05, + "loss": 0.9358, + "step": 10706 + }, + { + "epoch": 0.69, + "learning_rate": 1.7724081667704332e-05, + "loss": 1.0349, + "step": 10707 + }, + { + "epoch": 0.69, + "learning_rate": 1.772386697868138e-05, + "loss": 1.0303, + "step": 10708 + }, + { + "epoch": 0.69, + "learning_rate": 1.772365228965843e-05, + "loss": 1.0504, + "step": 10709 + }, + { + "epoch": 0.69, + "learning_rate": 1.772343760063548e-05, + "loss": 1.0763, + "step": 10710 + }, + { + "epoch": 0.69, + "learning_rate": 1.7723222911612528e-05, + "loss": 1.0892, + "step": 10711 + }, + { + "epoch": 0.69, + "learning_rate": 1.772300822258958e-05, + "loss": 1.0369, + "step": 10712 + }, + { + "epoch": 0.69, + "learning_rate": 1.772279353356663e-05, + "loss": 1.138, + "step": 10713 + }, + { + "epoch": 0.69, + "learning_rate": 1.7722578844543682e-05, + "loss": 1.0095, + "step": 10714 + }, + { + "epoch": 0.69, + "learning_rate": 1.772236415552073e-05, + "loss": 1.0298, + "step": 10715 + }, + { + "epoch": 0.69, + "learning_rate": 1.772214946649778e-05, + "loss": 1.0505, + "step": 10716 + }, + { + "epoch": 0.69, + "learning_rate": 1.772193477747483e-05, + "loss": 1.0431, + "step": 10717 + }, + { + "epoch": 0.69, + "learning_rate": 1.7721720088451878e-05, + "loss": 1.0673, + "step": 10718 + }, + { + "epoch": 0.69, + "learning_rate": 1.772150539942893e-05, + "loss": 1.0569, + "step": 10719 + }, + { + "epoch": 0.69, + "learning_rate": 1.772129071040598e-05, + "loss": 1.0618, + "step": 10720 + }, + { + "epoch": 0.69, + "learning_rate": 1.7721076021383028e-05, + "loss": 1.0928, + "step": 10721 + }, + { + "epoch": 0.69, + "learning_rate": 1.7720861332360077e-05, + "loss": 1.0914, + "step": 10722 + }, + { + "epoch": 0.69, + "learning_rate": 1.7720646643337126e-05, + "loss": 1.0035, + "step": 10723 + }, + { + "epoch": 0.69, + "learning_rate": 1.772043195431418e-05, + "loss": 1.0453, + "step": 10724 + }, + { + "epoch": 0.69, + "learning_rate": 1.7720217265291227e-05, + "loss": 1.0546, + "step": 10725 + }, + { + "epoch": 0.69, + "learning_rate": 1.7720002576268276e-05, + "loss": 1.0709, + "step": 10726 + }, + { + "epoch": 0.69, + "learning_rate": 1.7719787887245325e-05, + "loss": 1.1059, + "step": 10727 + }, + { + "epoch": 0.69, + "learning_rate": 1.7719573198222378e-05, + "loss": 1.0397, + "step": 10728 + }, + { + "epoch": 0.69, + "learning_rate": 1.7719358509199427e-05, + "loss": 1.0866, + "step": 10729 + }, + { + "epoch": 0.69, + "learning_rate": 1.7719143820176476e-05, + "loss": 1.1474, + "step": 10730 + }, + { + "epoch": 0.69, + "learning_rate": 1.7718929131153524e-05, + "loss": 1.102, + "step": 10731 + }, + { + "epoch": 0.69, + "learning_rate": 1.7718714442130577e-05, + "loss": 1.0897, + "step": 10732 + }, + { + "epoch": 0.69, + "learning_rate": 1.7718499753107626e-05, + "loss": 1.0595, + "step": 10733 + }, + { + "epoch": 0.69, + "learning_rate": 1.7718285064084675e-05, + "loss": 1.0244, + "step": 10734 + }, + { + "epoch": 0.69, + "learning_rate": 1.7718070375061724e-05, + "loss": 1.1235, + "step": 10735 + }, + { + "epoch": 0.69, + "learning_rate": 1.7717855686038773e-05, + "loss": 1.0702, + "step": 10736 + }, + { + "epoch": 0.69, + "learning_rate": 1.7717640997015825e-05, + "loss": 1.1484, + "step": 10737 + }, + { + "epoch": 0.69, + "learning_rate": 1.7717426307992874e-05, + "loss": 1.074, + "step": 10738 + }, + { + "epoch": 0.69, + "learning_rate": 1.7717211618969923e-05, + "loss": 1.0901, + "step": 10739 + }, + { + "epoch": 0.69, + "learning_rate": 1.7716996929946972e-05, + "loss": 1.0131, + "step": 10740 + }, + { + "epoch": 0.69, + "learning_rate": 1.771678224092402e-05, + "loss": 1.056, + "step": 10741 + }, + { + "epoch": 0.69, + "learning_rate": 1.7716567551901073e-05, + "loss": 1.052, + "step": 10742 + }, + { + "epoch": 0.69, + "learning_rate": 1.7716352862878122e-05, + "loss": 1.0496, + "step": 10743 + }, + { + "epoch": 0.69, + "learning_rate": 1.7716138173855175e-05, + "loss": 1.0813, + "step": 10744 + }, + { + "epoch": 0.69, + "learning_rate": 1.7715923484832224e-05, + "loss": 1.0468, + "step": 10745 + }, + { + "epoch": 0.69, + "learning_rate": 1.7715708795809273e-05, + "loss": 1.0983, + "step": 10746 + }, + { + "epoch": 0.69, + "learning_rate": 1.771549410678632e-05, + "loss": 1.0155, + "step": 10747 + }, + { + "epoch": 0.69, + "learning_rate": 1.771527941776337e-05, + "loss": 0.979, + "step": 10748 + }, + { + "epoch": 0.69, + "learning_rate": 1.771506472874042e-05, + "loss": 1.0443, + "step": 10749 + }, + { + "epoch": 0.69, + "learning_rate": 1.7714850039717472e-05, + "loss": 1.0333, + "step": 10750 + }, + { + "epoch": 0.69, + "learning_rate": 1.771463535069452e-05, + "loss": 1.0929, + "step": 10751 + }, + { + "epoch": 0.69, + "learning_rate": 1.771442066167157e-05, + "loss": 1.081, + "step": 10752 + }, + { + "epoch": 0.69, + "learning_rate": 1.771420597264862e-05, + "loss": 1.0317, + "step": 10753 + }, + { + "epoch": 0.69, + "learning_rate": 1.7713991283625668e-05, + "loss": 1.1085, + "step": 10754 + }, + { + "epoch": 0.69, + "learning_rate": 1.771377659460272e-05, + "loss": 1.0993, + "step": 10755 + }, + { + "epoch": 0.69, + "learning_rate": 1.771356190557977e-05, + "loss": 1.0679, + "step": 10756 + }, + { + "epoch": 0.69, + "learning_rate": 1.771334721655682e-05, + "loss": 1.0521, + "step": 10757 + }, + { + "epoch": 0.69, + "learning_rate": 1.771313252753387e-05, + "loss": 1.0744, + "step": 10758 + }, + { + "epoch": 0.69, + "learning_rate": 1.771291783851092e-05, + "loss": 1.0466, + "step": 10759 + }, + { + "epoch": 0.69, + "learning_rate": 1.7712703149487968e-05, + "loss": 1.0433, + "step": 10760 + }, + { + "epoch": 0.69, + "learning_rate": 1.7712488460465017e-05, + "loss": 1.1497, + "step": 10761 + }, + { + "epoch": 0.69, + "learning_rate": 1.771227377144207e-05, + "loss": 1.1422, + "step": 10762 + }, + { + "epoch": 0.69, + "learning_rate": 1.771205908241912e-05, + "loss": 1.0132, + "step": 10763 + }, + { + "epoch": 0.69, + "learning_rate": 1.7711844393396167e-05, + "loss": 1.1251, + "step": 10764 + }, + { + "epoch": 0.69, + "learning_rate": 1.7711629704373216e-05, + "loss": 1.0646, + "step": 10765 + }, + { + "epoch": 0.69, + "learning_rate": 1.7711415015350265e-05, + "loss": 1.1423, + "step": 10766 + }, + { + "epoch": 0.69, + "learning_rate": 1.7711200326327314e-05, + "loss": 1.0025, + "step": 10767 + }, + { + "epoch": 0.69, + "learning_rate": 1.7710985637304367e-05, + "loss": 1.1017, + "step": 10768 + }, + { + "epoch": 0.69, + "learning_rate": 1.7710770948281416e-05, + "loss": 1.079, + "step": 10769 + }, + { + "epoch": 0.69, + "learning_rate": 1.7710556259258465e-05, + "loss": 1.0307, + "step": 10770 + }, + { + "epoch": 0.69, + "learning_rate": 1.7710341570235517e-05, + "loss": 1.0154, + "step": 10771 + }, + { + "epoch": 0.69, + "learning_rate": 1.7710126881212566e-05, + "loss": 1.1478, + "step": 10772 + }, + { + "epoch": 0.69, + "learning_rate": 1.7709912192189615e-05, + "loss": 0.9801, + "step": 10773 + }, + { + "epoch": 0.69, + "learning_rate": 1.7709697503166664e-05, + "loss": 1.0337, + "step": 10774 + }, + { + "epoch": 0.69, + "learning_rate": 1.7709482814143716e-05, + "loss": 1.1115, + "step": 10775 + }, + { + "epoch": 0.69, + "learning_rate": 1.7709268125120765e-05, + "loss": 1.0411, + "step": 10776 + }, + { + "epoch": 0.69, + "learning_rate": 1.7709053436097814e-05, + "loss": 0.9975, + "step": 10777 + }, + { + "epoch": 0.69, + "learning_rate": 1.7708838747074863e-05, + "loss": 1.0938, + "step": 10778 + }, + { + "epoch": 0.69, + "learning_rate": 1.7708624058051912e-05, + "loss": 1.096, + "step": 10779 + }, + { + "epoch": 0.69, + "learning_rate": 1.770840936902896e-05, + "loss": 1.0436, + "step": 10780 + }, + { + "epoch": 0.69, + "learning_rate": 1.7708194680006013e-05, + "loss": 1.0976, + "step": 10781 + }, + { + "epoch": 0.69, + "learning_rate": 1.7707979990983062e-05, + "loss": 1.0309, + "step": 10782 + }, + { + "epoch": 0.69, + "learning_rate": 1.770776530196011e-05, + "loss": 1.009, + "step": 10783 + }, + { + "epoch": 0.69, + "learning_rate": 1.770755061293716e-05, + "loss": 1.1055, + "step": 10784 + }, + { + "epoch": 0.69, + "learning_rate": 1.770733592391421e-05, + "loss": 1.0577, + "step": 10785 + }, + { + "epoch": 0.69, + "learning_rate": 1.770712123489126e-05, + "loss": 1.0058, + "step": 10786 + }, + { + "epoch": 0.69, + "learning_rate": 1.770690654586831e-05, + "loss": 1.0863, + "step": 10787 + }, + { + "epoch": 0.69, + "learning_rate": 1.7706691856845363e-05, + "loss": 1.0586, + "step": 10788 + }, + { + "epoch": 0.69, + "learning_rate": 1.7706477167822412e-05, + "loss": 1.0973, + "step": 10789 + }, + { + "epoch": 0.69, + "learning_rate": 1.770626247879946e-05, + "loss": 1.0785, + "step": 10790 + }, + { + "epoch": 0.69, + "learning_rate": 1.770604778977651e-05, + "loss": 1.1535, + "step": 10791 + }, + { + "epoch": 0.69, + "learning_rate": 1.770583310075356e-05, + "loss": 1.0651, + "step": 10792 + }, + { + "epoch": 0.69, + "learning_rate": 1.770561841173061e-05, + "loss": 1.0855, + "step": 10793 + }, + { + "epoch": 0.69, + "learning_rate": 1.770540372270766e-05, + "loss": 1.0829, + "step": 10794 + }, + { + "epoch": 0.69, + "learning_rate": 1.770518903368471e-05, + "loss": 0.9908, + "step": 10795 + }, + { + "epoch": 0.69, + "learning_rate": 1.7704974344661758e-05, + "loss": 1.0663, + "step": 10796 + }, + { + "epoch": 0.69, + "learning_rate": 1.7704759655638807e-05, + "loss": 1.038, + "step": 10797 + }, + { + "epoch": 0.69, + "learning_rate": 1.7704544966615856e-05, + "loss": 1.1502, + "step": 10798 + }, + { + "epoch": 0.69, + "learning_rate": 1.770433027759291e-05, + "loss": 1.02, + "step": 10799 + }, + { + "epoch": 0.69, + "learning_rate": 1.7704115588569957e-05, + "loss": 1.0971, + "step": 10800 + }, + { + "epoch": 0.69, + "learning_rate": 1.770390089954701e-05, + "loss": 1.0098, + "step": 10801 + }, + { + "epoch": 0.69, + "learning_rate": 1.770368621052406e-05, + "loss": 0.9358, + "step": 10802 + }, + { + "epoch": 0.7, + "learning_rate": 1.7703471521501108e-05, + "loss": 1.0568, + "step": 10803 + }, + { + "epoch": 0.7, + "learning_rate": 1.7703256832478157e-05, + "loss": 1.0538, + "step": 10804 + }, + { + "epoch": 0.7, + "learning_rate": 1.7703042143455206e-05, + "loss": 1.0904, + "step": 10805 + }, + { + "epoch": 0.7, + "learning_rate": 1.7702827454432258e-05, + "loss": 1.0645, + "step": 10806 + }, + { + "epoch": 0.7, + "learning_rate": 1.7702612765409307e-05, + "loss": 1.0307, + "step": 10807 + }, + { + "epoch": 0.7, + "learning_rate": 1.7702398076386356e-05, + "loss": 1.0295, + "step": 10808 + }, + { + "epoch": 0.7, + "learning_rate": 1.7702183387363405e-05, + "loss": 1.048, + "step": 10809 + }, + { + "epoch": 0.7, + "learning_rate": 1.7701968698340454e-05, + "loss": 1.1078, + "step": 10810 + }, + { + "epoch": 0.7, + "learning_rate": 1.7701754009317506e-05, + "loss": 1.0679, + "step": 10811 + }, + { + "epoch": 0.7, + "learning_rate": 1.7701539320294555e-05, + "loss": 1.0747, + "step": 10812 + }, + { + "epoch": 0.7, + "learning_rate": 1.7701324631271604e-05, + "loss": 1.1304, + "step": 10813 + }, + { + "epoch": 0.7, + "learning_rate": 1.7701109942248653e-05, + "loss": 1.0506, + "step": 10814 + }, + { + "epoch": 0.7, + "learning_rate": 1.7700895253225705e-05, + "loss": 1.045, + "step": 10815 + }, + { + "epoch": 0.7, + "learning_rate": 1.7700680564202754e-05, + "loss": 1.021, + "step": 10816 + }, + { + "epoch": 0.7, + "learning_rate": 1.7700465875179803e-05, + "loss": 1.1176, + "step": 10817 + }, + { + "epoch": 0.7, + "learning_rate": 1.7700251186156852e-05, + "loss": 1.1198, + "step": 10818 + }, + { + "epoch": 0.7, + "learning_rate": 1.7700036497133905e-05, + "loss": 1.1508, + "step": 10819 + }, + { + "epoch": 0.7, + "learning_rate": 1.7699821808110954e-05, + "loss": 1.0982, + "step": 10820 + }, + { + "epoch": 0.7, + "learning_rate": 1.7699607119088002e-05, + "loss": 1.1218, + "step": 10821 + }, + { + "epoch": 0.7, + "learning_rate": 1.769939243006505e-05, + "loss": 1.0942, + "step": 10822 + }, + { + "epoch": 0.7, + "learning_rate": 1.76991777410421e-05, + "loss": 1.0142, + "step": 10823 + }, + { + "epoch": 0.7, + "learning_rate": 1.7698963052019153e-05, + "loss": 1.0476, + "step": 10824 + }, + { + "epoch": 0.7, + "learning_rate": 1.7698748362996202e-05, + "loss": 1.0943, + "step": 10825 + }, + { + "epoch": 0.7, + "learning_rate": 1.769853367397325e-05, + "loss": 1.1169, + "step": 10826 + }, + { + "epoch": 0.7, + "learning_rate": 1.76983189849503e-05, + "loss": 1.0161, + "step": 10827 + }, + { + "epoch": 0.7, + "learning_rate": 1.769810429592735e-05, + "loss": 1.1024, + "step": 10828 + }, + { + "epoch": 0.7, + "learning_rate": 1.76978896069044e-05, + "loss": 1.0553, + "step": 10829 + }, + { + "epoch": 0.7, + "learning_rate": 1.769767491788145e-05, + "loss": 0.9966, + "step": 10830 + }, + { + "epoch": 0.7, + "learning_rate": 1.7697460228858502e-05, + "loss": 1.1667, + "step": 10831 + }, + { + "epoch": 0.7, + "learning_rate": 1.769724553983555e-05, + "loss": 1.0881, + "step": 10832 + }, + { + "epoch": 0.7, + "learning_rate": 1.76970308508126e-05, + "loss": 1.0426, + "step": 10833 + }, + { + "epoch": 0.7, + "learning_rate": 1.769681616178965e-05, + "loss": 1.0153, + "step": 10834 + }, + { + "epoch": 0.7, + "learning_rate": 1.7696601472766698e-05, + "loss": 1.0685, + "step": 10835 + }, + { + "epoch": 0.7, + "learning_rate": 1.7696386783743747e-05, + "loss": 1.1695, + "step": 10836 + }, + { + "epoch": 0.7, + "learning_rate": 1.76961720947208e-05, + "loss": 1.0969, + "step": 10837 + }, + { + "epoch": 0.7, + "learning_rate": 1.769595740569785e-05, + "loss": 1.0291, + "step": 10838 + }, + { + "epoch": 0.7, + "learning_rate": 1.7695742716674897e-05, + "loss": 1.0796, + "step": 10839 + }, + { + "epoch": 0.7, + "learning_rate": 1.7695528027651946e-05, + "loss": 1.1305, + "step": 10840 + }, + { + "epoch": 0.7, + "learning_rate": 1.7695313338628995e-05, + "loss": 0.9948, + "step": 10841 + }, + { + "epoch": 0.7, + "learning_rate": 1.7695098649606048e-05, + "loss": 1.0979, + "step": 10842 + }, + { + "epoch": 0.7, + "learning_rate": 1.7694883960583097e-05, + "loss": 1.0928, + "step": 10843 + }, + { + "epoch": 0.7, + "learning_rate": 1.769466927156015e-05, + "loss": 1.1429, + "step": 10844 + }, + { + "epoch": 0.7, + "learning_rate": 1.7694454582537198e-05, + "loss": 1.1284, + "step": 10845 + }, + { + "epoch": 0.7, + "learning_rate": 1.7694239893514247e-05, + "loss": 1.0376, + "step": 10846 + }, + { + "epoch": 0.7, + "learning_rate": 1.7694025204491296e-05, + "loss": 1.0482, + "step": 10847 + }, + { + "epoch": 0.7, + "learning_rate": 1.7693810515468345e-05, + "loss": 1.0221, + "step": 10848 + }, + { + "epoch": 0.7, + "learning_rate": 1.7693595826445394e-05, + "loss": 1.0599, + "step": 10849 + }, + { + "epoch": 0.7, + "learning_rate": 1.7693381137422446e-05, + "loss": 1.078, + "step": 10850 + }, + { + "epoch": 0.7, + "learning_rate": 1.7693166448399495e-05, + "loss": 1.1172, + "step": 10851 + }, + { + "epoch": 0.7, + "learning_rate": 1.7692951759376544e-05, + "loss": 1.0999, + "step": 10852 + }, + { + "epoch": 0.7, + "learning_rate": 1.7692737070353593e-05, + "loss": 0.9931, + "step": 10853 + }, + { + "epoch": 0.7, + "learning_rate": 1.7692522381330642e-05, + "loss": 1.1187, + "step": 10854 + }, + { + "epoch": 0.7, + "learning_rate": 1.7692307692307694e-05, + "loss": 1.059, + "step": 10855 + }, + { + "epoch": 0.7, + "learning_rate": 1.7692093003284743e-05, + "loss": 1.0818, + "step": 10856 + }, + { + "epoch": 0.7, + "learning_rate": 1.7691878314261792e-05, + "loss": 1.0933, + "step": 10857 + }, + { + "epoch": 0.7, + "learning_rate": 1.7691663625238845e-05, + "loss": 1.1134, + "step": 10858 + }, + { + "epoch": 0.7, + "learning_rate": 1.7691448936215894e-05, + "loss": 1.0477, + "step": 10859 + }, + { + "epoch": 0.7, + "learning_rate": 1.7691234247192943e-05, + "loss": 1.116, + "step": 10860 + }, + { + "epoch": 0.7, + "learning_rate": 1.769101955816999e-05, + "loss": 1.0664, + "step": 10861 + }, + { + "epoch": 0.7, + "learning_rate": 1.7690804869147044e-05, + "loss": 1.0873, + "step": 10862 + }, + { + "epoch": 0.7, + "learning_rate": 1.7690590180124093e-05, + "loss": 1.0551, + "step": 10863 + }, + { + "epoch": 0.7, + "learning_rate": 1.7690375491101142e-05, + "loss": 1.0376, + "step": 10864 + }, + { + "epoch": 0.7, + "learning_rate": 1.769016080207819e-05, + "loss": 1.0854, + "step": 10865 + }, + { + "epoch": 0.7, + "learning_rate": 1.768994611305524e-05, + "loss": 1.0928, + "step": 10866 + }, + { + "epoch": 0.7, + "learning_rate": 1.768973142403229e-05, + "loss": 0.993, + "step": 10867 + }, + { + "epoch": 0.7, + "learning_rate": 1.768951673500934e-05, + "loss": 1.0786, + "step": 10868 + }, + { + "epoch": 0.7, + "learning_rate": 1.768930204598639e-05, + "loss": 1.0616, + "step": 10869 + }, + { + "epoch": 0.7, + "learning_rate": 1.768908735696344e-05, + "loss": 1.0225, + "step": 10870 + }, + { + "epoch": 0.7, + "learning_rate": 1.7688872667940488e-05, + "loss": 1.0731, + "step": 10871 + }, + { + "epoch": 0.7, + "learning_rate": 1.7688657978917537e-05, + "loss": 1.1441, + "step": 10872 + }, + { + "epoch": 0.7, + "learning_rate": 1.768844328989459e-05, + "loss": 1.0095, + "step": 10873 + }, + { + "epoch": 0.7, + "learning_rate": 1.7688228600871638e-05, + "loss": 1.1306, + "step": 10874 + }, + { + "epoch": 0.7, + "learning_rate": 1.768801391184869e-05, + "loss": 1.0088, + "step": 10875 + }, + { + "epoch": 0.7, + "learning_rate": 1.768779922282574e-05, + "loss": 1.033, + "step": 10876 + }, + { + "epoch": 0.7, + "learning_rate": 1.768758453380279e-05, + "loss": 1.1046, + "step": 10877 + }, + { + "epoch": 0.7, + "learning_rate": 1.7687369844779838e-05, + "loss": 1.1034, + "step": 10878 + }, + { + "epoch": 0.7, + "learning_rate": 1.7687155155756887e-05, + "loss": 1.0387, + "step": 10879 + }, + { + "epoch": 0.7, + "learning_rate": 1.768694046673394e-05, + "loss": 1.0549, + "step": 10880 + }, + { + "epoch": 0.7, + "learning_rate": 1.7686725777710988e-05, + "loss": 1.0032, + "step": 10881 + }, + { + "epoch": 0.7, + "learning_rate": 1.7686511088688037e-05, + "loss": 1.1156, + "step": 10882 + }, + { + "epoch": 0.7, + "learning_rate": 1.7686296399665086e-05, + "loss": 1.0447, + "step": 10883 + }, + { + "epoch": 0.7, + "learning_rate": 1.7686081710642135e-05, + "loss": 1.0804, + "step": 10884 + }, + { + "epoch": 0.7, + "learning_rate": 1.7685867021619184e-05, + "loss": 1.0115, + "step": 10885 + }, + { + "epoch": 0.7, + "learning_rate": 1.7685652332596236e-05, + "loss": 1.0639, + "step": 10886 + }, + { + "epoch": 0.7, + "learning_rate": 1.7685437643573285e-05, + "loss": 1.0532, + "step": 10887 + }, + { + "epoch": 0.7, + "learning_rate": 1.7685222954550337e-05, + "loss": 1.0384, + "step": 10888 + }, + { + "epoch": 0.7, + "learning_rate": 1.7685008265527386e-05, + "loss": 1.0365, + "step": 10889 + }, + { + "epoch": 0.7, + "learning_rate": 1.7684793576504435e-05, + "loss": 1.0452, + "step": 10890 + }, + { + "epoch": 0.7, + "learning_rate": 1.7684578887481484e-05, + "loss": 1.0793, + "step": 10891 + }, + { + "epoch": 0.7, + "learning_rate": 1.7684364198458533e-05, + "loss": 1.1348, + "step": 10892 + }, + { + "epoch": 0.7, + "learning_rate": 1.7684149509435586e-05, + "loss": 1.1208, + "step": 10893 + }, + { + "epoch": 0.7, + "learning_rate": 1.7683934820412635e-05, + "loss": 1.0316, + "step": 10894 + }, + { + "epoch": 0.7, + "learning_rate": 1.7683720131389683e-05, + "loss": 1.0845, + "step": 10895 + }, + { + "epoch": 0.7, + "learning_rate": 1.7683505442366732e-05, + "loss": 1.0002, + "step": 10896 + }, + { + "epoch": 0.7, + "learning_rate": 1.768329075334378e-05, + "loss": 1.037, + "step": 10897 + }, + { + "epoch": 0.7, + "learning_rate": 1.7683076064320834e-05, + "loss": 1.0503, + "step": 10898 + }, + { + "epoch": 0.7, + "learning_rate": 1.7682861375297883e-05, + "loss": 1.0517, + "step": 10899 + }, + { + "epoch": 0.7, + "learning_rate": 1.7682646686274932e-05, + "loss": 1.0544, + "step": 10900 + }, + { + "epoch": 0.7, + "learning_rate": 1.768243199725198e-05, + "loss": 1.0729, + "step": 10901 + }, + { + "epoch": 0.7, + "learning_rate": 1.7682217308229033e-05, + "loss": 1.1123, + "step": 10902 + }, + { + "epoch": 0.7, + "learning_rate": 1.7682002619206082e-05, + "loss": 1.0422, + "step": 10903 + }, + { + "epoch": 0.7, + "learning_rate": 1.768178793018313e-05, + "loss": 1.0107, + "step": 10904 + }, + { + "epoch": 0.7, + "learning_rate": 1.768157324116018e-05, + "loss": 1.1477, + "step": 10905 + }, + { + "epoch": 0.7, + "learning_rate": 1.7681358552137232e-05, + "loss": 1.0593, + "step": 10906 + }, + { + "epoch": 0.7, + "learning_rate": 1.768114386311428e-05, + "loss": 1.1049, + "step": 10907 + }, + { + "epoch": 0.7, + "learning_rate": 1.768092917409133e-05, + "loss": 1.1191, + "step": 10908 + }, + { + "epoch": 0.7, + "learning_rate": 1.768071448506838e-05, + "loss": 1.0296, + "step": 10909 + }, + { + "epoch": 0.7, + "learning_rate": 1.7680499796045428e-05, + "loss": 1.003, + "step": 10910 + }, + { + "epoch": 0.7, + "learning_rate": 1.768028510702248e-05, + "loss": 1.0796, + "step": 10911 + }, + { + "epoch": 0.7, + "learning_rate": 1.768007041799953e-05, + "loss": 1.0674, + "step": 10912 + }, + { + "epoch": 0.7, + "learning_rate": 1.767985572897658e-05, + "loss": 1.0859, + "step": 10913 + }, + { + "epoch": 0.7, + "learning_rate": 1.7679641039953627e-05, + "loss": 1.0996, + "step": 10914 + }, + { + "epoch": 0.7, + "learning_rate": 1.7679426350930676e-05, + "loss": 1.1068, + "step": 10915 + }, + { + "epoch": 0.7, + "learning_rate": 1.767921166190773e-05, + "loss": 1.1748, + "step": 10916 + }, + { + "epoch": 0.7, + "learning_rate": 1.7678996972884778e-05, + "loss": 1.0408, + "step": 10917 + }, + { + "epoch": 0.7, + "learning_rate": 1.767878228386183e-05, + "loss": 1.0111, + "step": 10918 + }, + { + "epoch": 0.7, + "learning_rate": 1.767856759483888e-05, + "loss": 1.086, + "step": 10919 + }, + { + "epoch": 0.7, + "learning_rate": 1.7678352905815928e-05, + "loss": 0.9939, + "step": 10920 + }, + { + "epoch": 0.7, + "learning_rate": 1.7678138216792977e-05, + "loss": 1.0612, + "step": 10921 + }, + { + "epoch": 0.7, + "learning_rate": 1.7677923527770026e-05, + "loss": 0.9998, + "step": 10922 + }, + { + "epoch": 0.7, + "learning_rate": 1.7677708838747075e-05, + "loss": 1.0082, + "step": 10923 + }, + { + "epoch": 0.7, + "learning_rate": 1.7677494149724127e-05, + "loss": 1.1225, + "step": 10924 + }, + { + "epoch": 0.7, + "learning_rate": 1.7677279460701176e-05, + "loss": 1.0253, + "step": 10925 + }, + { + "epoch": 0.7, + "learning_rate": 1.7677064771678225e-05, + "loss": 1.0357, + "step": 10926 + }, + { + "epoch": 0.7, + "learning_rate": 1.7676850082655274e-05, + "loss": 1.0659, + "step": 10927 + }, + { + "epoch": 0.7, + "learning_rate": 1.7676635393632323e-05, + "loss": 1.0572, + "step": 10928 + }, + { + "epoch": 0.7, + "learning_rate": 1.7676420704609375e-05, + "loss": 1.0341, + "step": 10929 + }, + { + "epoch": 0.7, + "learning_rate": 1.7676206015586424e-05, + "loss": 1.0743, + "step": 10930 + }, + { + "epoch": 0.7, + "learning_rate": 1.7675991326563477e-05, + "loss": 1.0094, + "step": 10931 + }, + { + "epoch": 0.7, + "learning_rate": 1.7675776637540526e-05, + "loss": 1.1177, + "step": 10932 + }, + { + "epoch": 0.7, + "learning_rate": 1.7675561948517575e-05, + "loss": 1.0191, + "step": 10933 + }, + { + "epoch": 0.7, + "learning_rate": 1.7675347259494624e-05, + "loss": 0.9602, + "step": 10934 + }, + { + "epoch": 0.7, + "learning_rate": 1.7675132570471673e-05, + "loss": 1.094, + "step": 10935 + }, + { + "epoch": 0.7, + "learning_rate": 1.767491788144872e-05, + "loss": 1.0155, + "step": 10936 + }, + { + "epoch": 0.7, + "learning_rate": 1.7674703192425774e-05, + "loss": 1.081, + "step": 10937 + }, + { + "epoch": 0.7, + "learning_rate": 1.7674488503402823e-05, + "loss": 1.075, + "step": 10938 + }, + { + "epoch": 0.7, + "learning_rate": 1.7674273814379872e-05, + "loss": 0.9845, + "step": 10939 + }, + { + "epoch": 0.7, + "learning_rate": 1.767405912535692e-05, + "loss": 1.0694, + "step": 10940 + }, + { + "epoch": 0.7, + "learning_rate": 1.767384443633397e-05, + "loss": 1.0582, + "step": 10941 + }, + { + "epoch": 0.7, + "learning_rate": 1.7673629747311022e-05, + "loss": 1.12, + "step": 10942 + }, + { + "epoch": 0.7, + "learning_rate": 1.767341505828807e-05, + "loss": 1.1147, + "step": 10943 + }, + { + "epoch": 0.7, + "learning_rate": 1.767320036926512e-05, + "loss": 1.0389, + "step": 10944 + }, + { + "epoch": 0.7, + "learning_rate": 1.7672985680242172e-05, + "loss": 1.0211, + "step": 10945 + }, + { + "epoch": 0.7, + "learning_rate": 1.767277099121922e-05, + "loss": 1.0641, + "step": 10946 + }, + { + "epoch": 0.7, + "learning_rate": 1.767255630219627e-05, + "loss": 1.0638, + "step": 10947 + }, + { + "epoch": 0.7, + "learning_rate": 1.767234161317332e-05, + "loss": 1.0916, + "step": 10948 + }, + { + "epoch": 0.7, + "learning_rate": 1.767212692415037e-05, + "loss": 1.1411, + "step": 10949 + }, + { + "epoch": 0.7, + "learning_rate": 1.767191223512742e-05, + "loss": 1.0151, + "step": 10950 + }, + { + "epoch": 0.7, + "learning_rate": 1.767169754610447e-05, + "loss": 1.07, + "step": 10951 + }, + { + "epoch": 0.7, + "learning_rate": 1.767148285708152e-05, + "loss": 1.1033, + "step": 10952 + }, + { + "epoch": 0.7, + "learning_rate": 1.7671268168058568e-05, + "loss": 1.0762, + "step": 10953 + }, + { + "epoch": 0.7, + "learning_rate": 1.7671053479035616e-05, + "loss": 1.0754, + "step": 10954 + }, + { + "epoch": 0.7, + "learning_rate": 1.767083879001267e-05, + "loss": 1.0508, + "step": 10955 + }, + { + "epoch": 0.7, + "learning_rate": 1.7670624100989718e-05, + "loss": 1.0894, + "step": 10956 + }, + { + "epoch": 0.7, + "learning_rate": 1.7670409411966767e-05, + "loss": 1.03, + "step": 10957 + }, + { + "epoch": 0.71, + "learning_rate": 1.7670194722943816e-05, + "loss": 1.0258, + "step": 10958 + }, + { + "epoch": 0.71, + "learning_rate": 1.7669980033920865e-05, + "loss": 1.0878, + "step": 10959 + }, + { + "epoch": 0.71, + "learning_rate": 1.7669765344897917e-05, + "loss": 1.0974, + "step": 10960 + }, + { + "epoch": 0.71, + "learning_rate": 1.7669550655874966e-05, + "loss": 1.0465, + "step": 10961 + }, + { + "epoch": 0.71, + "learning_rate": 1.766933596685202e-05, + "loss": 1.1116, + "step": 10962 + }, + { + "epoch": 0.71, + "learning_rate": 1.7669121277829067e-05, + "loss": 1.092, + "step": 10963 + }, + { + "epoch": 0.71, + "learning_rate": 1.7668906588806116e-05, + "loss": 1.0285, + "step": 10964 + }, + { + "epoch": 0.71, + "learning_rate": 1.7668691899783165e-05, + "loss": 1.0964, + "step": 10965 + }, + { + "epoch": 0.71, + "learning_rate": 1.7668477210760214e-05, + "loss": 0.9801, + "step": 10966 + }, + { + "epoch": 0.71, + "learning_rate": 1.7668262521737267e-05, + "loss": 1.0166, + "step": 10967 + }, + { + "epoch": 0.71, + "learning_rate": 1.7668047832714316e-05, + "loss": 1.0353, + "step": 10968 + }, + { + "epoch": 0.71, + "learning_rate": 1.7667833143691365e-05, + "loss": 1.0041, + "step": 10969 + }, + { + "epoch": 0.71, + "learning_rate": 1.7667618454668413e-05, + "loss": 1.0364, + "step": 10970 + }, + { + "epoch": 0.71, + "learning_rate": 1.7667403765645462e-05, + "loss": 1.1285, + "step": 10971 + }, + { + "epoch": 0.71, + "learning_rate": 1.766718907662251e-05, + "loss": 1.0885, + "step": 10972 + }, + { + "epoch": 0.71, + "learning_rate": 1.7666974387599564e-05, + "loss": 1.095, + "step": 10973 + }, + { + "epoch": 0.71, + "learning_rate": 1.7666759698576613e-05, + "loss": 1.0389, + "step": 10974 + }, + { + "epoch": 0.71, + "learning_rate": 1.7666545009553665e-05, + "loss": 1.0961, + "step": 10975 + }, + { + "epoch": 0.71, + "learning_rate": 1.7666330320530714e-05, + "loss": 1.1339, + "step": 10976 + }, + { + "epoch": 0.71, + "learning_rate": 1.7666115631507763e-05, + "loss": 1.1192, + "step": 10977 + }, + { + "epoch": 0.71, + "learning_rate": 1.7665900942484812e-05, + "loss": 1.1045, + "step": 10978 + }, + { + "epoch": 0.71, + "learning_rate": 1.766568625346186e-05, + "loss": 1.1344, + "step": 10979 + }, + { + "epoch": 0.71, + "learning_rate": 1.7665471564438913e-05, + "loss": 1.0882, + "step": 10980 + }, + { + "epoch": 0.71, + "learning_rate": 1.7665256875415962e-05, + "loss": 1.0448, + "step": 10981 + }, + { + "epoch": 0.71, + "learning_rate": 1.766504218639301e-05, + "loss": 1.0983, + "step": 10982 + }, + { + "epoch": 0.71, + "learning_rate": 1.766482749737006e-05, + "loss": 1.1149, + "step": 10983 + }, + { + "epoch": 0.71, + "learning_rate": 1.766461280834711e-05, + "loss": 1.0871, + "step": 10984 + }, + { + "epoch": 0.71, + "learning_rate": 1.766439811932416e-05, + "loss": 1.0691, + "step": 10985 + }, + { + "epoch": 0.71, + "learning_rate": 1.766418343030121e-05, + "loss": 1.0481, + "step": 10986 + }, + { + "epoch": 0.71, + "learning_rate": 1.766396874127826e-05, + "loss": 1.0134, + "step": 10987 + }, + { + "epoch": 0.71, + "learning_rate": 1.766375405225531e-05, + "loss": 1.0644, + "step": 10988 + }, + { + "epoch": 0.71, + "learning_rate": 1.766353936323236e-05, + "loss": 1.118, + "step": 10989 + }, + { + "epoch": 0.71, + "learning_rate": 1.766332467420941e-05, + "loss": 1.0435, + "step": 10990 + }, + { + "epoch": 0.71, + "learning_rate": 1.766310998518646e-05, + "loss": 1.0507, + "step": 10991 + }, + { + "epoch": 0.71, + "learning_rate": 1.7662895296163508e-05, + "loss": 1.1815, + "step": 10992 + }, + { + "epoch": 0.71, + "learning_rate": 1.766268060714056e-05, + "loss": 1.056, + "step": 10993 + }, + { + "epoch": 0.71, + "learning_rate": 1.766246591811761e-05, + "loss": 1.0913, + "step": 10994 + }, + { + "epoch": 0.71, + "learning_rate": 1.7662251229094658e-05, + "loss": 1.0472, + "step": 10995 + }, + { + "epoch": 0.71, + "learning_rate": 1.7662036540071707e-05, + "loss": 1.0488, + "step": 10996 + }, + { + "epoch": 0.71, + "learning_rate": 1.7661821851048756e-05, + "loss": 1.0738, + "step": 10997 + }, + { + "epoch": 0.71, + "learning_rate": 1.7661607162025808e-05, + "loss": 1.0846, + "step": 10998 + }, + { + "epoch": 0.71, + "learning_rate": 1.7661392473002857e-05, + "loss": 1.0907, + "step": 10999 + }, + { + "epoch": 0.71, + "learning_rate": 1.7661177783979906e-05, + "loss": 1.0443, + "step": 11000 + }, + { + "epoch": 0.71, + "eval_laion/cider": 0.531899064318802, + "eval_laion/meteor": 0.3838324053135861, + "eval_laion/pred_num_words": 36.255859375, + "eval_laion/sacrebleu": 15.145693695909271, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.082856297492981, + "eval_runtime": 104.6754, + "eval_samples_per_second": 4.767, + "eval_steps_per_second": 0.306, + "step": 11000 + }, + { + "epoch": 0.71, + "learning_rate": 1.7660963094956955e-05, + "loss": 1.1197, + "step": 11001 + }, + { + "epoch": 0.71, + "learning_rate": 1.7660748405934004e-05, + "loss": 1.0788, + "step": 11002 + }, + { + "epoch": 0.71, + "learning_rate": 1.7660533716911056e-05, + "loss": 1.0366, + "step": 11003 + }, + { + "epoch": 0.71, + "learning_rate": 1.7660319027888105e-05, + "loss": 1.0588, + "step": 11004 + }, + { + "epoch": 0.71, + "learning_rate": 1.7660104338865158e-05, + "loss": 1.0177, + "step": 11005 + }, + { + "epoch": 0.71, + "learning_rate": 1.7659889649842207e-05, + "loss": 1.1248, + "step": 11006 + }, + { + "epoch": 0.71, + "learning_rate": 1.7659674960819256e-05, + "loss": 1.0749, + "step": 11007 + }, + { + "epoch": 0.71, + "learning_rate": 1.7659460271796305e-05, + "loss": 1.0293, + "step": 11008 + }, + { + "epoch": 0.71, + "learning_rate": 1.7659245582773354e-05, + "loss": 1.1086, + "step": 11009 + }, + { + "epoch": 0.71, + "learning_rate": 1.7659030893750403e-05, + "loss": 1.0489, + "step": 11010 + }, + { + "epoch": 0.71, + "learning_rate": 1.7658816204727455e-05, + "loss": 1.0991, + "step": 11011 + }, + { + "epoch": 0.71, + "learning_rate": 1.7658601515704504e-05, + "loss": 0.9703, + "step": 11012 + }, + { + "epoch": 0.71, + "learning_rate": 1.7658386826681553e-05, + "loss": 0.9876, + "step": 11013 + }, + { + "epoch": 0.71, + "learning_rate": 1.7658172137658602e-05, + "loss": 1.1259, + "step": 11014 + }, + { + "epoch": 0.71, + "learning_rate": 1.765795744863565e-05, + "loss": 1.0643, + "step": 11015 + }, + { + "epoch": 0.71, + "learning_rate": 1.7657742759612703e-05, + "loss": 1.085, + "step": 11016 + }, + { + "epoch": 0.71, + "learning_rate": 1.7657528070589752e-05, + "loss": 1.0516, + "step": 11017 + }, + { + "epoch": 0.71, + "learning_rate": 1.7657313381566804e-05, + "loss": 1.0491, + "step": 11018 + }, + { + "epoch": 0.71, + "learning_rate": 1.7657098692543853e-05, + "loss": 1.0126, + "step": 11019 + }, + { + "epoch": 0.71, + "learning_rate": 1.7656884003520902e-05, + "loss": 1.0279, + "step": 11020 + }, + { + "epoch": 0.71, + "learning_rate": 1.765666931449795e-05, + "loss": 1.04, + "step": 11021 + }, + { + "epoch": 0.71, + "learning_rate": 1.7656454625475e-05, + "loss": 1.031, + "step": 11022 + }, + { + "epoch": 0.71, + "learning_rate": 1.765623993645205e-05, + "loss": 1.1037, + "step": 11023 + }, + { + "epoch": 0.71, + "learning_rate": 1.76560252474291e-05, + "loss": 1.0793, + "step": 11024 + }, + { + "epoch": 0.71, + "learning_rate": 1.765581055840615e-05, + "loss": 1.104, + "step": 11025 + }, + { + "epoch": 0.71, + "learning_rate": 1.76555958693832e-05, + "loss": 1.0124, + "step": 11026 + }, + { + "epoch": 0.71, + "learning_rate": 1.765538118036025e-05, + "loss": 1.0733, + "step": 11027 + }, + { + "epoch": 0.71, + "learning_rate": 1.7655166491337297e-05, + "loss": 1.1194, + "step": 11028 + }, + { + "epoch": 0.71, + "learning_rate": 1.765495180231435e-05, + "loss": 1.095, + "step": 11029 + }, + { + "epoch": 0.71, + "learning_rate": 1.76547371132914e-05, + "loss": 1.1269, + "step": 11030 + }, + { + "epoch": 0.71, + "learning_rate": 1.7654522424268448e-05, + "loss": 1.0178, + "step": 11031 + }, + { + "epoch": 0.71, + "learning_rate": 1.7654307735245497e-05, + "loss": 1.0917, + "step": 11032 + }, + { + "epoch": 0.71, + "learning_rate": 1.765409304622255e-05, + "loss": 1.1029, + "step": 11033 + }, + { + "epoch": 0.71, + "learning_rate": 1.7653878357199598e-05, + "loss": 1.0726, + "step": 11034 + }, + { + "epoch": 0.71, + "learning_rate": 1.7653663668176647e-05, + "loss": 1.0645, + "step": 11035 + }, + { + "epoch": 0.71, + "learning_rate": 1.76534489791537e-05, + "loss": 1.1121, + "step": 11036 + }, + { + "epoch": 0.71, + "learning_rate": 1.765323429013075e-05, + "loss": 1.0637, + "step": 11037 + }, + { + "epoch": 0.71, + "learning_rate": 1.7653019601107797e-05, + "loss": 1.0662, + "step": 11038 + }, + { + "epoch": 0.71, + "learning_rate": 1.7652804912084846e-05, + "loss": 1.145, + "step": 11039 + }, + { + "epoch": 0.71, + "learning_rate": 1.7652590223061895e-05, + "loss": 1.1582, + "step": 11040 + }, + { + "epoch": 0.71, + "learning_rate": 1.7652375534038944e-05, + "loss": 1.0866, + "step": 11041 + }, + { + "epoch": 0.71, + "learning_rate": 1.7652160845015997e-05, + "loss": 1.027, + "step": 11042 + }, + { + "epoch": 0.71, + "learning_rate": 1.7651946155993046e-05, + "loss": 1.0724, + "step": 11043 + }, + { + "epoch": 0.71, + "learning_rate": 1.7651731466970094e-05, + "loss": 1.065, + "step": 11044 + }, + { + "epoch": 0.71, + "learning_rate": 1.7651516777947143e-05, + "loss": 1.0993, + "step": 11045 + }, + { + "epoch": 0.71, + "learning_rate": 1.7651302088924192e-05, + "loss": 1.1439, + "step": 11046 + }, + { + "epoch": 0.71, + "learning_rate": 1.7651087399901245e-05, + "loss": 0.9609, + "step": 11047 + }, + { + "epoch": 0.71, + "learning_rate": 1.7650872710878294e-05, + "loss": 0.9567, + "step": 11048 + }, + { + "epoch": 0.71, + "learning_rate": 1.7650658021855346e-05, + "loss": 1.0624, + "step": 11049 + }, + { + "epoch": 0.71, + "learning_rate": 1.7650443332832395e-05, + "loss": 1.0831, + "step": 11050 + }, + { + "epoch": 0.71, + "learning_rate": 1.7650228643809444e-05, + "loss": 0.971, + "step": 11051 + }, + { + "epoch": 0.71, + "learning_rate": 1.7650013954786493e-05, + "loss": 0.999, + "step": 11052 + }, + { + "epoch": 0.71, + "learning_rate": 1.7649799265763542e-05, + "loss": 1.1, + "step": 11053 + }, + { + "epoch": 0.71, + "learning_rate": 1.7649584576740594e-05, + "loss": 1.0872, + "step": 11054 + }, + { + "epoch": 0.71, + "learning_rate": 1.7649369887717643e-05, + "loss": 1.0272, + "step": 11055 + }, + { + "epoch": 0.71, + "learning_rate": 1.7649155198694692e-05, + "loss": 1.0764, + "step": 11056 + }, + { + "epoch": 0.71, + "learning_rate": 1.764894050967174e-05, + "loss": 1.0903, + "step": 11057 + }, + { + "epoch": 0.71, + "learning_rate": 1.764872582064879e-05, + "loss": 1.1248, + "step": 11058 + }, + { + "epoch": 0.71, + "learning_rate": 1.764851113162584e-05, + "loss": 1.1172, + "step": 11059 + }, + { + "epoch": 0.71, + "learning_rate": 1.764829644260289e-05, + "loss": 1.0597, + "step": 11060 + }, + { + "epoch": 0.71, + "learning_rate": 1.764808175357994e-05, + "loss": 1.0914, + "step": 11061 + }, + { + "epoch": 0.71, + "learning_rate": 1.7647867064556993e-05, + "loss": 1.0862, + "step": 11062 + }, + { + "epoch": 0.71, + "learning_rate": 1.7647652375534042e-05, + "loss": 1.047, + "step": 11063 + }, + { + "epoch": 0.71, + "learning_rate": 1.764743768651109e-05, + "loss": 1.0387, + "step": 11064 + }, + { + "epoch": 0.71, + "learning_rate": 1.764722299748814e-05, + "loss": 1.0717, + "step": 11065 + }, + { + "epoch": 0.71, + "learning_rate": 1.764700830846519e-05, + "loss": 1.1737, + "step": 11066 + }, + { + "epoch": 0.71, + "learning_rate": 1.764679361944224e-05, + "loss": 1.0679, + "step": 11067 + }, + { + "epoch": 0.71, + "learning_rate": 1.764657893041929e-05, + "loss": 1.0622, + "step": 11068 + }, + { + "epoch": 0.71, + "learning_rate": 1.764636424139634e-05, + "loss": 1.009, + "step": 11069 + }, + { + "epoch": 0.71, + "learning_rate": 1.7646149552373388e-05, + "loss": 1.1027, + "step": 11070 + }, + { + "epoch": 0.71, + "learning_rate": 1.7645934863350437e-05, + "loss": 1.0838, + "step": 11071 + }, + { + "epoch": 0.71, + "learning_rate": 1.764572017432749e-05, + "loss": 1.0316, + "step": 11072 + }, + { + "epoch": 0.71, + "learning_rate": 1.7645505485304538e-05, + "loss": 1.1132, + "step": 11073 + }, + { + "epoch": 0.71, + "learning_rate": 1.7645290796281587e-05, + "loss": 1.0131, + "step": 11074 + }, + { + "epoch": 0.71, + "learning_rate": 1.7645076107258636e-05, + "loss": 1.0673, + "step": 11075 + }, + { + "epoch": 0.71, + "learning_rate": 1.764486141823569e-05, + "loss": 1.1071, + "step": 11076 + }, + { + "epoch": 0.71, + "learning_rate": 1.7644646729212737e-05, + "loss": 1.1123, + "step": 11077 + }, + { + "epoch": 0.71, + "learning_rate": 1.7644432040189786e-05, + "loss": 1.034, + "step": 11078 + }, + { + "epoch": 0.71, + "learning_rate": 1.7644217351166835e-05, + "loss": 1.0122, + "step": 11079 + }, + { + "epoch": 0.71, + "learning_rate": 1.7644002662143888e-05, + "loss": 1.0789, + "step": 11080 + }, + { + "epoch": 0.71, + "learning_rate": 1.7643787973120937e-05, + "loss": 1.1113, + "step": 11081 + }, + { + "epoch": 0.71, + "learning_rate": 1.7643573284097986e-05, + "loss": 0.9775, + "step": 11082 + }, + { + "epoch": 0.71, + "learning_rate": 1.7643358595075035e-05, + "loss": 1.0431, + "step": 11083 + }, + { + "epoch": 0.71, + "learning_rate": 1.7643143906052084e-05, + "loss": 1.0906, + "step": 11084 + }, + { + "epoch": 0.71, + "learning_rate": 1.7642929217029136e-05, + "loss": 1.0453, + "step": 11085 + }, + { + "epoch": 0.71, + "learning_rate": 1.7642714528006185e-05, + "loss": 1.1192, + "step": 11086 + }, + { + "epoch": 0.71, + "learning_rate": 1.7642499838983234e-05, + "loss": 1.0417, + "step": 11087 + }, + { + "epoch": 0.71, + "learning_rate": 1.7642285149960283e-05, + "loss": 1.0409, + "step": 11088 + }, + { + "epoch": 0.71, + "learning_rate": 1.7642070460937332e-05, + "loss": 1.0828, + "step": 11089 + }, + { + "epoch": 0.71, + "learning_rate": 1.764185577191438e-05, + "loss": 1.0543, + "step": 11090 + }, + { + "epoch": 0.71, + "learning_rate": 1.7641641082891433e-05, + "loss": 1.1478, + "step": 11091 + }, + { + "epoch": 0.71, + "learning_rate": 1.7641426393868485e-05, + "loss": 1.0937, + "step": 11092 + }, + { + "epoch": 0.71, + "learning_rate": 1.7641211704845534e-05, + "loss": 1.0899, + "step": 11093 + }, + { + "epoch": 0.71, + "learning_rate": 1.7640997015822583e-05, + "loss": 1.0529, + "step": 11094 + }, + { + "epoch": 0.71, + "learning_rate": 1.7640782326799632e-05, + "loss": 1.1242, + "step": 11095 + }, + { + "epoch": 0.71, + "learning_rate": 1.764056763777668e-05, + "loss": 1.0386, + "step": 11096 + }, + { + "epoch": 0.71, + "learning_rate": 1.764035294875373e-05, + "loss": 1.0382, + "step": 11097 + }, + { + "epoch": 0.71, + "learning_rate": 1.7640138259730783e-05, + "loss": 1.1305, + "step": 11098 + }, + { + "epoch": 0.71, + "learning_rate": 1.763992357070783e-05, + "loss": 1.0019, + "step": 11099 + }, + { + "epoch": 0.71, + "learning_rate": 1.763970888168488e-05, + "loss": 1.0122, + "step": 11100 + }, + { + "epoch": 0.71, + "learning_rate": 1.763949419266193e-05, + "loss": 1.1312, + "step": 11101 + }, + { + "epoch": 0.71, + "learning_rate": 1.763927950363898e-05, + "loss": 1.0803, + "step": 11102 + }, + { + "epoch": 0.71, + "learning_rate": 1.763906481461603e-05, + "loss": 1.06, + "step": 11103 + }, + { + "epoch": 0.71, + "learning_rate": 1.763885012559308e-05, + "loss": 1.0222, + "step": 11104 + }, + { + "epoch": 0.71, + "learning_rate": 1.7638635436570132e-05, + "loss": 1.0219, + "step": 11105 + }, + { + "epoch": 0.71, + "learning_rate": 1.763842074754718e-05, + "loss": 1.125, + "step": 11106 + }, + { + "epoch": 0.71, + "learning_rate": 1.763820605852423e-05, + "loss": 1.0776, + "step": 11107 + }, + { + "epoch": 0.71, + "learning_rate": 1.763799136950128e-05, + "loss": 1.1404, + "step": 11108 + }, + { + "epoch": 0.71, + "learning_rate": 1.7637776680478328e-05, + "loss": 1.0567, + "step": 11109 + }, + { + "epoch": 0.71, + "learning_rate": 1.7637561991455377e-05, + "loss": 1.0907, + "step": 11110 + }, + { + "epoch": 0.71, + "learning_rate": 1.763734730243243e-05, + "loss": 1.1043, + "step": 11111 + }, + { + "epoch": 0.71, + "learning_rate": 1.7637132613409478e-05, + "loss": 1.0665, + "step": 11112 + }, + { + "epoch": 0.71, + "learning_rate": 1.7636917924386527e-05, + "loss": 1.0038, + "step": 11113 + }, + { + "epoch": 0.72, + "learning_rate": 1.7636703235363576e-05, + "loss": 1.0683, + "step": 11114 + }, + { + "epoch": 0.72, + "learning_rate": 1.7636488546340625e-05, + "loss": 1.0824, + "step": 11115 + }, + { + "epoch": 0.72, + "learning_rate": 1.7636273857317678e-05, + "loss": 1.0798, + "step": 11116 + }, + { + "epoch": 0.72, + "learning_rate": 1.7636059168294727e-05, + "loss": 1.1074, + "step": 11117 + }, + { + "epoch": 0.72, + "learning_rate": 1.7635844479271775e-05, + "loss": 1.1209, + "step": 11118 + }, + { + "epoch": 0.72, + "learning_rate": 1.7635629790248824e-05, + "loss": 1.0411, + "step": 11119 + }, + { + "epoch": 0.72, + "learning_rate": 1.7635415101225877e-05, + "loss": 1.1035, + "step": 11120 + }, + { + "epoch": 0.72, + "learning_rate": 1.7635200412202926e-05, + "loss": 1.1167, + "step": 11121 + }, + { + "epoch": 0.72, + "learning_rate": 1.7634985723179975e-05, + "loss": 1.1253, + "step": 11122 + }, + { + "epoch": 0.72, + "learning_rate": 1.7634771034157027e-05, + "loss": 1.0517, + "step": 11123 + }, + { + "epoch": 0.72, + "learning_rate": 1.7634556345134076e-05, + "loss": 1.0047, + "step": 11124 + }, + { + "epoch": 0.72, + "learning_rate": 1.7634341656111125e-05, + "loss": 1.1172, + "step": 11125 + }, + { + "epoch": 0.72, + "learning_rate": 1.7634126967088174e-05, + "loss": 1.1008, + "step": 11126 + }, + { + "epoch": 0.72, + "learning_rate": 1.7633912278065223e-05, + "loss": 0.9943, + "step": 11127 + }, + { + "epoch": 0.72, + "learning_rate": 1.7633697589042272e-05, + "loss": 1.0108, + "step": 11128 + }, + { + "epoch": 0.72, + "learning_rate": 1.7633482900019324e-05, + "loss": 1.0114, + "step": 11129 + }, + { + "epoch": 0.72, + "learning_rate": 1.7633268210996373e-05, + "loss": 1.0421, + "step": 11130 + }, + { + "epoch": 0.72, + "learning_rate": 1.7633053521973422e-05, + "loss": 1.0239, + "step": 11131 + }, + { + "epoch": 0.72, + "learning_rate": 1.763283883295047e-05, + "loss": 1.0, + "step": 11132 + }, + { + "epoch": 0.72, + "learning_rate": 1.763262414392752e-05, + "loss": 1.1248, + "step": 11133 + }, + { + "epoch": 0.72, + "learning_rate": 1.7632409454904572e-05, + "loss": 1.0732, + "step": 11134 + }, + { + "epoch": 0.72, + "learning_rate": 1.763219476588162e-05, + "loss": 1.0615, + "step": 11135 + }, + { + "epoch": 0.72, + "learning_rate": 1.7631980076858674e-05, + "loss": 1.0119, + "step": 11136 + }, + { + "epoch": 0.72, + "learning_rate": 1.7631765387835723e-05, + "loss": 1.1684, + "step": 11137 + }, + { + "epoch": 0.72, + "learning_rate": 1.7631550698812772e-05, + "loss": 0.9711, + "step": 11138 + }, + { + "epoch": 0.72, + "learning_rate": 1.763133600978982e-05, + "loss": 1.129, + "step": 11139 + }, + { + "epoch": 0.72, + "learning_rate": 1.763112132076687e-05, + "loss": 1.0659, + "step": 11140 + }, + { + "epoch": 0.72, + "learning_rate": 1.7630906631743922e-05, + "loss": 1.058, + "step": 11141 + }, + { + "epoch": 0.72, + "learning_rate": 1.763069194272097e-05, + "loss": 1.1426, + "step": 11142 + }, + { + "epoch": 0.72, + "learning_rate": 1.763047725369802e-05, + "loss": 1.1075, + "step": 11143 + }, + { + "epoch": 0.72, + "learning_rate": 1.763026256467507e-05, + "loss": 1.0583, + "step": 11144 + }, + { + "epoch": 0.72, + "learning_rate": 1.7630047875652118e-05, + "loss": 1.061, + "step": 11145 + }, + { + "epoch": 0.72, + "learning_rate": 1.7629833186629167e-05, + "loss": 0.9933, + "step": 11146 + }, + { + "epoch": 0.72, + "learning_rate": 1.762961849760622e-05, + "loss": 1.1147, + "step": 11147 + }, + { + "epoch": 0.72, + "learning_rate": 1.7629403808583268e-05, + "loss": 1.0606, + "step": 11148 + }, + { + "epoch": 0.72, + "learning_rate": 1.762918911956032e-05, + "loss": 1.0201, + "step": 11149 + }, + { + "epoch": 0.72, + "learning_rate": 1.762897443053737e-05, + "loss": 1.0546, + "step": 11150 + }, + { + "epoch": 0.72, + "learning_rate": 1.762875974151442e-05, + "loss": 1.0551, + "step": 11151 + }, + { + "epoch": 0.72, + "learning_rate": 1.7628545052491467e-05, + "loss": 1.0465, + "step": 11152 + }, + { + "epoch": 0.72, + "learning_rate": 1.7628330363468516e-05, + "loss": 1.0089, + "step": 11153 + }, + { + "epoch": 0.72, + "learning_rate": 1.762811567444557e-05, + "loss": 1.0043, + "step": 11154 + }, + { + "epoch": 0.72, + "learning_rate": 1.7627900985422618e-05, + "loss": 1.059, + "step": 11155 + }, + { + "epoch": 0.72, + "learning_rate": 1.7627686296399667e-05, + "loss": 1.092, + "step": 11156 + }, + { + "epoch": 0.72, + "learning_rate": 1.7627471607376716e-05, + "loss": 1.1107, + "step": 11157 + }, + { + "epoch": 0.72, + "learning_rate": 1.7627256918353765e-05, + "loss": 1.0781, + "step": 11158 + }, + { + "epoch": 0.72, + "learning_rate": 1.7627042229330814e-05, + "loss": 1.0754, + "step": 11159 + }, + { + "epoch": 0.72, + "learning_rate": 1.7626827540307866e-05, + "loss": 1.0327, + "step": 11160 + }, + { + "epoch": 0.72, + "learning_rate": 1.7626612851284915e-05, + "loss": 1.0936, + "step": 11161 + }, + { + "epoch": 0.72, + "learning_rate": 1.7626398162261964e-05, + "loss": 1.0779, + "step": 11162 + }, + { + "epoch": 0.72, + "learning_rate": 1.7626183473239016e-05, + "loss": 1.0949, + "step": 11163 + }, + { + "epoch": 0.72, + "learning_rate": 1.7625968784216065e-05, + "loss": 1.1323, + "step": 11164 + }, + { + "epoch": 0.72, + "learning_rate": 1.7625754095193114e-05, + "loss": 1.0888, + "step": 11165 + }, + { + "epoch": 0.72, + "learning_rate": 1.7625539406170163e-05, + "loss": 1.1013, + "step": 11166 + }, + { + "epoch": 0.72, + "learning_rate": 1.7625324717147215e-05, + "loss": 1.0968, + "step": 11167 + }, + { + "epoch": 0.72, + "learning_rate": 1.7625110028124264e-05, + "loss": 1.1119, + "step": 11168 + }, + { + "epoch": 0.72, + "learning_rate": 1.7624895339101313e-05, + "loss": 1.0712, + "step": 11169 + }, + { + "epoch": 0.72, + "learning_rate": 1.7624680650078362e-05, + "loss": 1.1235, + "step": 11170 + }, + { + "epoch": 0.72, + "learning_rate": 1.762446596105541e-05, + "loss": 1.1005, + "step": 11171 + }, + { + "epoch": 0.72, + "learning_rate": 1.7624251272032464e-05, + "loss": 1.1375, + "step": 11172 + }, + { + "epoch": 0.72, + "learning_rate": 1.7624036583009513e-05, + "loss": 1.0399, + "step": 11173 + }, + { + "epoch": 0.72, + "learning_rate": 1.762382189398656e-05, + "loss": 1.111, + "step": 11174 + }, + { + "epoch": 0.72, + "learning_rate": 1.762360720496361e-05, + "loss": 1.1096, + "step": 11175 + }, + { + "epoch": 0.72, + "learning_rate": 1.762339251594066e-05, + "loss": 1.1145, + "step": 11176 + }, + { + "epoch": 0.72, + "learning_rate": 1.762317782691771e-05, + "loss": 1.0792, + "step": 11177 + }, + { + "epoch": 0.72, + "learning_rate": 1.762296313789476e-05, + "loss": 1.0778, + "step": 11178 + }, + { + "epoch": 0.72, + "learning_rate": 1.7622748448871813e-05, + "loss": 1.0079, + "step": 11179 + }, + { + "epoch": 0.72, + "learning_rate": 1.7622533759848862e-05, + "loss": 1.1083, + "step": 11180 + }, + { + "epoch": 0.72, + "learning_rate": 1.762231907082591e-05, + "loss": 1.1431, + "step": 11181 + }, + { + "epoch": 0.72, + "learning_rate": 1.762210438180296e-05, + "loss": 1.0981, + "step": 11182 + }, + { + "epoch": 0.72, + "learning_rate": 1.762188969278001e-05, + "loss": 1.1421, + "step": 11183 + }, + { + "epoch": 0.72, + "learning_rate": 1.7621675003757058e-05, + "loss": 1.0502, + "step": 11184 + }, + { + "epoch": 0.72, + "learning_rate": 1.762146031473411e-05, + "loss": 1.062, + "step": 11185 + }, + { + "epoch": 0.72, + "learning_rate": 1.762124562571116e-05, + "loss": 1.1225, + "step": 11186 + }, + { + "epoch": 0.72, + "learning_rate": 1.7621030936688208e-05, + "loss": 1.0703, + "step": 11187 + }, + { + "epoch": 0.72, + "learning_rate": 1.7620816247665257e-05, + "loss": 0.9925, + "step": 11188 + }, + { + "epoch": 0.72, + "learning_rate": 1.7620601558642306e-05, + "loss": 0.9958, + "step": 11189 + }, + { + "epoch": 0.72, + "learning_rate": 1.762038686961936e-05, + "loss": 1.0542, + "step": 11190 + }, + { + "epoch": 0.72, + "learning_rate": 1.7620172180596408e-05, + "loss": 1.0686, + "step": 11191 + }, + { + "epoch": 0.72, + "learning_rate": 1.761995749157346e-05, + "loss": 1.0488, + "step": 11192 + }, + { + "epoch": 0.72, + "learning_rate": 1.761974280255051e-05, + "loss": 1.0403, + "step": 11193 + }, + { + "epoch": 0.72, + "learning_rate": 1.7619528113527558e-05, + "loss": 1.0252, + "step": 11194 + }, + { + "epoch": 0.72, + "learning_rate": 1.7619313424504607e-05, + "loss": 1.1047, + "step": 11195 + }, + { + "epoch": 0.72, + "learning_rate": 1.7619098735481656e-05, + "loss": 0.9934, + "step": 11196 + }, + { + "epoch": 0.72, + "learning_rate": 1.7618884046458705e-05, + "loss": 1.0297, + "step": 11197 + }, + { + "epoch": 0.72, + "learning_rate": 1.7618669357435757e-05, + "loss": 1.025, + "step": 11198 + }, + { + "epoch": 0.72, + "learning_rate": 1.7618454668412806e-05, + "loss": 0.997, + "step": 11199 + }, + { + "epoch": 0.72, + "learning_rate": 1.7618239979389855e-05, + "loss": 1.1613, + "step": 11200 + }, + { + "epoch": 0.72, + "learning_rate": 1.7618025290366904e-05, + "loss": 1.0887, + "step": 11201 + }, + { + "epoch": 0.72, + "learning_rate": 1.7617810601343953e-05, + "loss": 1.0505, + "step": 11202 + }, + { + "epoch": 0.72, + "learning_rate": 1.7617595912321005e-05, + "loss": 1.0861, + "step": 11203 + }, + { + "epoch": 0.72, + "learning_rate": 1.7617381223298054e-05, + "loss": 1.1205, + "step": 11204 + }, + { + "epoch": 0.72, + "learning_rate": 1.7617166534275103e-05, + "loss": 0.9479, + "step": 11205 + }, + { + "epoch": 0.72, + "learning_rate": 1.7616951845252152e-05, + "loss": 0.996, + "step": 11206 + }, + { + "epoch": 0.72, + "learning_rate": 1.7616737156229205e-05, + "loss": 1.033, + "step": 11207 + }, + { + "epoch": 0.72, + "learning_rate": 1.7616522467206253e-05, + "loss": 1.0418, + "step": 11208 + }, + { + "epoch": 0.72, + "learning_rate": 1.7616307778183302e-05, + "loss": 1.1568, + "step": 11209 + }, + { + "epoch": 0.72, + "learning_rate": 1.7616093089160355e-05, + "loss": 1.1055, + "step": 11210 + }, + { + "epoch": 0.72, + "learning_rate": 1.7615878400137404e-05, + "loss": 1.0537, + "step": 11211 + }, + { + "epoch": 0.72, + "learning_rate": 1.7615663711114453e-05, + "loss": 1.1437, + "step": 11212 + }, + { + "epoch": 0.72, + "learning_rate": 1.76154490220915e-05, + "loss": 1.0111, + "step": 11213 + }, + { + "epoch": 0.72, + "learning_rate": 1.761523433306855e-05, + "loss": 1.026, + "step": 11214 + }, + { + "epoch": 0.72, + "learning_rate": 1.76150196440456e-05, + "loss": 0.9871, + "step": 11215 + }, + { + "epoch": 0.72, + "learning_rate": 1.7614804955022652e-05, + "loss": 1.1203, + "step": 11216 + }, + { + "epoch": 0.72, + "learning_rate": 1.76145902659997e-05, + "loss": 1.0271, + "step": 11217 + }, + { + "epoch": 0.72, + "learning_rate": 1.761437557697675e-05, + "loss": 1.0541, + "step": 11218 + }, + { + "epoch": 0.72, + "learning_rate": 1.76141608879538e-05, + "loss": 1.0239, + "step": 11219 + }, + { + "epoch": 0.72, + "learning_rate": 1.7613946198930848e-05, + "loss": 1.1505, + "step": 11220 + }, + { + "epoch": 0.72, + "learning_rate": 1.76137315099079e-05, + "loss": 1.0987, + "step": 11221 + }, + { + "epoch": 0.72, + "learning_rate": 1.761351682088495e-05, + "loss": 1.0443, + "step": 11222 + }, + { + "epoch": 0.72, + "learning_rate": 1.7613302131862e-05, + "loss": 1.0219, + "step": 11223 + }, + { + "epoch": 0.72, + "learning_rate": 1.761308744283905e-05, + "loss": 1.0928, + "step": 11224 + }, + { + "epoch": 0.72, + "learning_rate": 1.76128727538161e-05, + "loss": 1.1663, + "step": 11225 + }, + { + "epoch": 0.72, + "learning_rate": 1.761265806479315e-05, + "loss": 1.0265, + "step": 11226 + }, + { + "epoch": 0.72, + "learning_rate": 1.7612443375770197e-05, + "loss": 1.0968, + "step": 11227 + }, + { + "epoch": 0.72, + "learning_rate": 1.761222868674725e-05, + "loss": 1.1009, + "step": 11228 + }, + { + "epoch": 0.72, + "learning_rate": 1.76120139977243e-05, + "loss": 1.1246, + "step": 11229 + }, + { + "epoch": 0.72, + "learning_rate": 1.7611799308701348e-05, + "loss": 0.9984, + "step": 11230 + }, + { + "epoch": 0.72, + "learning_rate": 1.7611584619678397e-05, + "loss": 1.0329, + "step": 11231 + }, + { + "epoch": 0.72, + "learning_rate": 1.7611369930655446e-05, + "loss": 1.0006, + "step": 11232 + }, + { + "epoch": 0.72, + "learning_rate": 1.7611155241632495e-05, + "loss": 1.074, + "step": 11233 + }, + { + "epoch": 0.72, + "learning_rate": 1.7610940552609547e-05, + "loss": 1.0457, + "step": 11234 + }, + { + "epoch": 0.72, + "learning_rate": 1.7610725863586596e-05, + "loss": 1.0512, + "step": 11235 + }, + { + "epoch": 0.72, + "learning_rate": 1.7610511174563648e-05, + "loss": 1.0163, + "step": 11236 + }, + { + "epoch": 0.72, + "learning_rate": 1.7610296485540697e-05, + "loss": 1.0545, + "step": 11237 + }, + { + "epoch": 0.72, + "learning_rate": 1.7610081796517746e-05, + "loss": 1.0176, + "step": 11238 + }, + { + "epoch": 0.72, + "learning_rate": 1.7609867107494795e-05, + "loss": 1.0192, + "step": 11239 + }, + { + "epoch": 0.72, + "learning_rate": 1.7609652418471844e-05, + "loss": 1.0812, + "step": 11240 + }, + { + "epoch": 0.72, + "learning_rate": 1.7609437729448896e-05, + "loss": 1.1425, + "step": 11241 + }, + { + "epoch": 0.72, + "learning_rate": 1.7609223040425945e-05, + "loss": 1.0659, + "step": 11242 + }, + { + "epoch": 0.72, + "learning_rate": 1.7609008351402994e-05, + "loss": 1.2169, + "step": 11243 + }, + { + "epoch": 0.72, + "learning_rate": 1.7608793662380043e-05, + "loss": 1.0817, + "step": 11244 + }, + { + "epoch": 0.72, + "learning_rate": 1.7608578973357092e-05, + "loss": 1.1392, + "step": 11245 + }, + { + "epoch": 0.72, + "learning_rate": 1.760836428433414e-05, + "loss": 1.0636, + "step": 11246 + }, + { + "epoch": 0.72, + "learning_rate": 1.7608149595311194e-05, + "loss": 1.1253, + "step": 11247 + }, + { + "epoch": 0.72, + "learning_rate": 1.7607934906288243e-05, + "loss": 1.0271, + "step": 11248 + }, + { + "epoch": 0.72, + "learning_rate": 1.760772021726529e-05, + "loss": 1.0422, + "step": 11249 + }, + { + "epoch": 0.72, + "learning_rate": 1.7607505528242344e-05, + "loss": 1.097, + "step": 11250 + }, + { + "epoch": 0.72, + "learning_rate": 1.7607290839219393e-05, + "loss": 1.0103, + "step": 11251 + }, + { + "epoch": 0.72, + "learning_rate": 1.7607076150196442e-05, + "loss": 1.1105, + "step": 11252 + }, + { + "epoch": 0.72, + "learning_rate": 1.760686146117349e-05, + "loss": 0.9675, + "step": 11253 + }, + { + "epoch": 0.72, + "learning_rate": 1.7606646772150543e-05, + "loss": 1.0124, + "step": 11254 + }, + { + "epoch": 0.72, + "learning_rate": 1.7606432083127592e-05, + "loss": 1.0664, + "step": 11255 + }, + { + "epoch": 0.72, + "learning_rate": 1.760621739410464e-05, + "loss": 1.0574, + "step": 11256 + }, + { + "epoch": 0.72, + "learning_rate": 1.760600270508169e-05, + "loss": 0.9911, + "step": 11257 + }, + { + "epoch": 0.72, + "learning_rate": 1.760578801605874e-05, + "loss": 1.054, + "step": 11258 + }, + { + "epoch": 0.72, + "learning_rate": 1.760557332703579e-05, + "loss": 1.0173, + "step": 11259 + }, + { + "epoch": 0.72, + "learning_rate": 1.760535863801284e-05, + "loss": 1.0085, + "step": 11260 + }, + { + "epoch": 0.72, + "learning_rate": 1.760514394898989e-05, + "loss": 1.0378, + "step": 11261 + }, + { + "epoch": 0.72, + "learning_rate": 1.7604929259966938e-05, + "loss": 1.07, + "step": 11262 + }, + { + "epoch": 0.72, + "learning_rate": 1.7604714570943987e-05, + "loss": 1.0782, + "step": 11263 + }, + { + "epoch": 0.72, + "learning_rate": 1.7604499881921036e-05, + "loss": 1.0124, + "step": 11264 + }, + { + "epoch": 0.72, + "learning_rate": 1.760428519289809e-05, + "loss": 1.0381, + "step": 11265 + }, + { + "epoch": 0.72, + "learning_rate": 1.7604070503875137e-05, + "loss": 1.091, + "step": 11266 + }, + { + "epoch": 0.72, + "learning_rate": 1.760385581485219e-05, + "loss": 1.0542, + "step": 11267 + }, + { + "epoch": 0.72, + "learning_rate": 1.760364112582924e-05, + "loss": 1.1377, + "step": 11268 + }, + { + "epoch": 0.73, + "learning_rate": 1.7603426436806288e-05, + "loss": 1.1191, + "step": 11269 + }, + { + "epoch": 0.73, + "learning_rate": 1.7603211747783337e-05, + "loss": 1.0153, + "step": 11270 + }, + { + "epoch": 0.73, + "learning_rate": 1.7602997058760386e-05, + "loss": 1.0578, + "step": 11271 + }, + { + "epoch": 0.73, + "learning_rate": 1.7602782369737438e-05, + "loss": 1.0689, + "step": 11272 + }, + { + "epoch": 0.73, + "learning_rate": 1.7602567680714487e-05, + "loss": 1.0368, + "step": 11273 + }, + { + "epoch": 0.73, + "learning_rate": 1.7602352991691536e-05, + "loss": 1.0401, + "step": 11274 + }, + { + "epoch": 0.73, + "learning_rate": 1.7602138302668585e-05, + "loss": 1.1004, + "step": 11275 + }, + { + "epoch": 0.73, + "learning_rate": 1.7601923613645634e-05, + "loss": 1.1013, + "step": 11276 + }, + { + "epoch": 0.73, + "learning_rate": 1.7601708924622686e-05, + "loss": 1.1208, + "step": 11277 + }, + { + "epoch": 0.73, + "learning_rate": 1.7601494235599735e-05, + "loss": 1.0344, + "step": 11278 + }, + { + "epoch": 0.73, + "learning_rate": 1.7601279546576788e-05, + "loss": 1.052, + "step": 11279 + }, + { + "epoch": 0.73, + "learning_rate": 1.7601064857553837e-05, + "loss": 1.0349, + "step": 11280 + }, + { + "epoch": 0.73, + "learning_rate": 1.7600850168530886e-05, + "loss": 1.1935, + "step": 11281 + }, + { + "epoch": 0.73, + "learning_rate": 1.7600635479507934e-05, + "loss": 1.0468, + "step": 11282 + }, + { + "epoch": 0.73, + "learning_rate": 1.7600420790484983e-05, + "loss": 1.0273, + "step": 11283 + }, + { + "epoch": 0.73, + "learning_rate": 1.7600206101462032e-05, + "loss": 1.1544, + "step": 11284 + }, + { + "epoch": 0.73, + "learning_rate": 1.7599991412439085e-05, + "loss": 1.1039, + "step": 11285 + }, + { + "epoch": 0.73, + "learning_rate": 1.7599776723416134e-05, + "loss": 1.0901, + "step": 11286 + }, + { + "epoch": 0.73, + "learning_rate": 1.7599562034393183e-05, + "loss": 1.0328, + "step": 11287 + }, + { + "epoch": 0.73, + "learning_rate": 1.759934734537023e-05, + "loss": 1.1084, + "step": 11288 + }, + { + "epoch": 0.73, + "learning_rate": 1.759913265634728e-05, + "loss": 1.1026, + "step": 11289 + }, + { + "epoch": 0.73, + "learning_rate": 1.7598917967324333e-05, + "loss": 1.0243, + "step": 11290 + }, + { + "epoch": 0.73, + "learning_rate": 1.7598703278301382e-05, + "loss": 1.0486, + "step": 11291 + }, + { + "epoch": 0.73, + "learning_rate": 1.759848858927843e-05, + "loss": 1.0754, + "step": 11292 + }, + { + "epoch": 0.73, + "learning_rate": 1.759827390025548e-05, + "loss": 1.0553, + "step": 11293 + }, + { + "epoch": 0.73, + "learning_rate": 1.7598059211232532e-05, + "loss": 1.0947, + "step": 11294 + }, + { + "epoch": 0.73, + "learning_rate": 1.759784452220958e-05, + "loss": 1.0685, + "step": 11295 + }, + { + "epoch": 0.73, + "learning_rate": 1.759762983318663e-05, + "loss": 0.9861, + "step": 11296 + }, + { + "epoch": 0.73, + "learning_rate": 1.7597415144163682e-05, + "loss": 1.1002, + "step": 11297 + }, + { + "epoch": 0.73, + "learning_rate": 1.759720045514073e-05, + "loss": 1.031, + "step": 11298 + }, + { + "epoch": 0.73, + "learning_rate": 1.759698576611778e-05, + "loss": 1.0183, + "step": 11299 + }, + { + "epoch": 0.73, + "learning_rate": 1.759677107709483e-05, + "loss": 1.0817, + "step": 11300 + }, + { + "epoch": 0.73, + "learning_rate": 1.759655638807188e-05, + "loss": 1.0678, + "step": 11301 + }, + { + "epoch": 0.73, + "learning_rate": 1.7596341699048927e-05, + "loss": 1.0454, + "step": 11302 + }, + { + "epoch": 0.73, + "learning_rate": 1.759612701002598e-05, + "loss": 1.0045, + "step": 11303 + }, + { + "epoch": 0.73, + "learning_rate": 1.759591232100303e-05, + "loss": 0.9988, + "step": 11304 + }, + { + "epoch": 0.73, + "learning_rate": 1.7595697631980078e-05, + "loss": 1.1929, + "step": 11305 + }, + { + "epoch": 0.73, + "learning_rate": 1.7595482942957127e-05, + "loss": 1.0567, + "step": 11306 + }, + { + "epoch": 0.73, + "learning_rate": 1.7595268253934176e-05, + "loss": 1.0206, + "step": 11307 + }, + { + "epoch": 0.73, + "learning_rate": 1.7595053564911228e-05, + "loss": 0.9598, + "step": 11308 + }, + { + "epoch": 0.73, + "learning_rate": 1.7594838875888277e-05, + "loss": 1.0812, + "step": 11309 + }, + { + "epoch": 0.73, + "learning_rate": 1.759462418686533e-05, + "loss": 1.0803, + "step": 11310 + }, + { + "epoch": 0.73, + "learning_rate": 1.7594409497842378e-05, + "loss": 1.0244, + "step": 11311 + }, + { + "epoch": 0.73, + "learning_rate": 1.7594194808819427e-05, + "loss": 1.1514, + "step": 11312 + }, + { + "epoch": 0.73, + "learning_rate": 1.7593980119796476e-05, + "loss": 0.9643, + "step": 11313 + }, + { + "epoch": 0.73, + "learning_rate": 1.7593765430773525e-05, + "loss": 0.9923, + "step": 11314 + }, + { + "epoch": 0.73, + "learning_rate": 1.7593550741750577e-05, + "loss": 1.0193, + "step": 11315 + }, + { + "epoch": 0.73, + "learning_rate": 1.7593336052727626e-05, + "loss": 1.107, + "step": 11316 + }, + { + "epoch": 0.73, + "learning_rate": 1.7593121363704675e-05, + "loss": 1.0108, + "step": 11317 + }, + { + "epoch": 0.73, + "learning_rate": 1.7592906674681724e-05, + "loss": 1.0732, + "step": 11318 + }, + { + "epoch": 0.73, + "learning_rate": 1.7592691985658773e-05, + "loss": 1.0815, + "step": 11319 + }, + { + "epoch": 0.73, + "learning_rate": 1.7592477296635822e-05, + "loss": 1.0579, + "step": 11320 + }, + { + "epoch": 0.73, + "learning_rate": 1.7592262607612875e-05, + "loss": 1.1173, + "step": 11321 + }, + { + "epoch": 0.73, + "learning_rate": 1.7592047918589924e-05, + "loss": 1.0215, + "step": 11322 + }, + { + "epoch": 0.73, + "learning_rate": 1.7591833229566976e-05, + "loss": 1.0153, + "step": 11323 + }, + { + "epoch": 0.73, + "learning_rate": 1.7591618540544025e-05, + "loss": 1.109, + "step": 11324 + }, + { + "epoch": 0.73, + "learning_rate": 1.7591403851521074e-05, + "loss": 1.0966, + "step": 11325 + }, + { + "epoch": 0.73, + "learning_rate": 1.7591189162498123e-05, + "loss": 1.1211, + "step": 11326 + }, + { + "epoch": 0.73, + "learning_rate": 1.7590974473475172e-05, + "loss": 1.0326, + "step": 11327 + }, + { + "epoch": 0.73, + "learning_rate": 1.7590759784452224e-05, + "loss": 1.0828, + "step": 11328 + }, + { + "epoch": 0.73, + "learning_rate": 1.7590545095429273e-05, + "loss": 1.0924, + "step": 11329 + }, + { + "epoch": 0.73, + "learning_rate": 1.7590330406406322e-05, + "loss": 1.1117, + "step": 11330 + }, + { + "epoch": 0.73, + "learning_rate": 1.759011571738337e-05, + "loss": 0.9256, + "step": 11331 + }, + { + "epoch": 0.73, + "learning_rate": 1.758990102836042e-05, + "loss": 1.0468, + "step": 11332 + }, + { + "epoch": 0.73, + "learning_rate": 1.758968633933747e-05, + "loss": 0.9708, + "step": 11333 + }, + { + "epoch": 0.73, + "learning_rate": 1.758947165031452e-05, + "loss": 1.0219, + "step": 11334 + }, + { + "epoch": 0.73, + "learning_rate": 1.758925696129157e-05, + "loss": 1.0763, + "step": 11335 + }, + { + "epoch": 0.73, + "learning_rate": 1.758904227226862e-05, + "loss": 0.9974, + "step": 11336 + }, + { + "epoch": 0.73, + "learning_rate": 1.758882758324567e-05, + "loss": 1.0181, + "step": 11337 + }, + { + "epoch": 0.73, + "learning_rate": 1.758861289422272e-05, + "loss": 1.0635, + "step": 11338 + }, + { + "epoch": 0.73, + "learning_rate": 1.758839820519977e-05, + "loss": 1.0175, + "step": 11339 + }, + { + "epoch": 0.73, + "learning_rate": 1.758818351617682e-05, + "loss": 1.0972, + "step": 11340 + }, + { + "epoch": 0.73, + "learning_rate": 1.758796882715387e-05, + "loss": 1.0791, + "step": 11341 + }, + { + "epoch": 0.73, + "learning_rate": 1.758775413813092e-05, + "loss": 0.9878, + "step": 11342 + }, + { + "epoch": 0.73, + "learning_rate": 1.758753944910797e-05, + "loss": 1.0074, + "step": 11343 + }, + { + "epoch": 0.73, + "learning_rate": 1.7587324760085018e-05, + "loss": 1.0031, + "step": 11344 + }, + { + "epoch": 0.73, + "learning_rate": 1.7587110071062067e-05, + "loss": 1.1064, + "step": 11345 + }, + { + "epoch": 0.73, + "learning_rate": 1.758689538203912e-05, + "loss": 1.017, + "step": 11346 + }, + { + "epoch": 0.73, + "learning_rate": 1.7586680693016168e-05, + "loss": 1.0227, + "step": 11347 + }, + { + "epoch": 0.73, + "learning_rate": 1.7586466003993217e-05, + "loss": 1.0658, + "step": 11348 + }, + { + "epoch": 0.73, + "learning_rate": 1.7586251314970266e-05, + "loss": 1.0063, + "step": 11349 + }, + { + "epoch": 0.73, + "learning_rate": 1.7586036625947315e-05, + "loss": 1.131, + "step": 11350 + }, + { + "epoch": 0.73, + "learning_rate": 1.7585821936924364e-05, + "loss": 1.0407, + "step": 11351 + }, + { + "epoch": 0.73, + "learning_rate": 1.7585607247901416e-05, + "loss": 1.0012, + "step": 11352 + }, + { + "epoch": 0.73, + "learning_rate": 1.7585392558878465e-05, + "loss": 1.0793, + "step": 11353 + }, + { + "epoch": 0.73, + "learning_rate": 1.7585177869855518e-05, + "loss": 1.0739, + "step": 11354 + }, + { + "epoch": 0.73, + "learning_rate": 1.7584963180832567e-05, + "loss": 1.0815, + "step": 11355 + }, + { + "epoch": 0.73, + "learning_rate": 1.7584748491809615e-05, + "loss": 1.0671, + "step": 11356 + }, + { + "epoch": 0.73, + "learning_rate": 1.7584533802786664e-05, + "loss": 1.0074, + "step": 11357 + }, + { + "epoch": 0.73, + "learning_rate": 1.7584319113763713e-05, + "loss": 1.0914, + "step": 11358 + }, + { + "epoch": 0.73, + "learning_rate": 1.7584104424740766e-05, + "loss": 1.124, + "step": 11359 + }, + { + "epoch": 0.73, + "learning_rate": 1.7583889735717815e-05, + "loss": 1.0318, + "step": 11360 + }, + { + "epoch": 0.73, + "learning_rate": 1.7583675046694864e-05, + "loss": 0.9782, + "step": 11361 + }, + { + "epoch": 0.73, + "learning_rate": 1.7583460357671913e-05, + "loss": 1.0156, + "step": 11362 + }, + { + "epoch": 0.73, + "learning_rate": 1.758324566864896e-05, + "loss": 1.1009, + "step": 11363 + }, + { + "epoch": 0.73, + "learning_rate": 1.7583030979626014e-05, + "loss": 1.111, + "step": 11364 + }, + { + "epoch": 0.73, + "learning_rate": 1.7582816290603063e-05, + "loss": 1.0254, + "step": 11365 + }, + { + "epoch": 0.73, + "learning_rate": 1.7582601601580115e-05, + "loss": 1.0707, + "step": 11366 + }, + { + "epoch": 0.73, + "learning_rate": 1.7582386912557164e-05, + "loss": 1.0477, + "step": 11367 + }, + { + "epoch": 0.73, + "learning_rate": 1.7582172223534213e-05, + "loss": 1.09, + "step": 11368 + }, + { + "epoch": 0.73, + "learning_rate": 1.7581957534511262e-05, + "loss": 1.0348, + "step": 11369 + }, + { + "epoch": 0.73, + "learning_rate": 1.758174284548831e-05, + "loss": 1.0626, + "step": 11370 + }, + { + "epoch": 0.73, + "learning_rate": 1.758152815646536e-05, + "loss": 1.0165, + "step": 11371 + }, + { + "epoch": 0.73, + "learning_rate": 1.7581313467442412e-05, + "loss": 1.0263, + "step": 11372 + }, + { + "epoch": 0.73, + "learning_rate": 1.758109877841946e-05, + "loss": 1.1307, + "step": 11373 + }, + { + "epoch": 0.73, + "learning_rate": 1.758088408939651e-05, + "loss": 1.0578, + "step": 11374 + }, + { + "epoch": 0.73, + "learning_rate": 1.758066940037356e-05, + "loss": 1.0819, + "step": 11375 + }, + { + "epoch": 0.73, + "learning_rate": 1.758045471135061e-05, + "loss": 1.027, + "step": 11376 + }, + { + "epoch": 0.73, + "learning_rate": 1.758024002232766e-05, + "loss": 1.0497, + "step": 11377 + }, + { + "epoch": 0.73, + "learning_rate": 1.758002533330471e-05, + "loss": 1.1322, + "step": 11378 + }, + { + "epoch": 0.73, + "learning_rate": 1.757981064428176e-05, + "loss": 1.0261, + "step": 11379 + }, + { + "epoch": 0.73, + "learning_rate": 1.7579595955258808e-05, + "loss": 1.0619, + "step": 11380 + }, + { + "epoch": 0.73, + "learning_rate": 1.757938126623586e-05, + "loss": 1.0329, + "step": 11381 + }, + { + "epoch": 0.73, + "learning_rate": 1.757916657721291e-05, + "loss": 1.0522, + "step": 11382 + }, + { + "epoch": 0.73, + "learning_rate": 1.7578951888189958e-05, + "loss": 1.0649, + "step": 11383 + }, + { + "epoch": 0.73, + "learning_rate": 1.757873719916701e-05, + "loss": 1.0777, + "step": 11384 + }, + { + "epoch": 0.73, + "learning_rate": 1.757852251014406e-05, + "loss": 1.0799, + "step": 11385 + }, + { + "epoch": 0.73, + "learning_rate": 1.7578307821121108e-05, + "loss": 1.0432, + "step": 11386 + }, + { + "epoch": 0.73, + "learning_rate": 1.7578093132098157e-05, + "loss": 1.0253, + "step": 11387 + }, + { + "epoch": 0.73, + "learning_rate": 1.7577878443075206e-05, + "loss": 1.1383, + "step": 11388 + }, + { + "epoch": 0.73, + "learning_rate": 1.7577663754052255e-05, + "loss": 1.0057, + "step": 11389 + }, + { + "epoch": 0.73, + "learning_rate": 1.7577449065029307e-05, + "loss": 1.0787, + "step": 11390 + }, + { + "epoch": 0.73, + "learning_rate": 1.7577234376006356e-05, + "loss": 0.9886, + "step": 11391 + }, + { + "epoch": 0.73, + "learning_rate": 1.7577019686983405e-05, + "loss": 1.1077, + "step": 11392 + }, + { + "epoch": 0.73, + "learning_rate": 1.7576804997960454e-05, + "loss": 1.1374, + "step": 11393 + }, + { + "epoch": 0.73, + "learning_rate": 1.7576590308937503e-05, + "loss": 1.0516, + "step": 11394 + }, + { + "epoch": 0.73, + "learning_rate": 1.7576375619914556e-05, + "loss": 1.128, + "step": 11395 + }, + { + "epoch": 0.73, + "learning_rate": 1.7576160930891605e-05, + "loss": 1.1143, + "step": 11396 + }, + { + "epoch": 0.73, + "learning_rate": 1.7575946241868657e-05, + "loss": 1.1485, + "step": 11397 + }, + { + "epoch": 0.73, + "learning_rate": 1.7575731552845706e-05, + "loss": 1.0786, + "step": 11398 + }, + { + "epoch": 0.73, + "learning_rate": 1.7575516863822755e-05, + "loss": 1.0175, + "step": 11399 + }, + { + "epoch": 0.73, + "learning_rate": 1.7575302174799804e-05, + "loss": 1.0557, + "step": 11400 + }, + { + "epoch": 0.73, + "learning_rate": 1.7575087485776853e-05, + "loss": 1.072, + "step": 11401 + }, + { + "epoch": 0.73, + "learning_rate": 1.7574872796753905e-05, + "loss": 1.0519, + "step": 11402 + }, + { + "epoch": 0.73, + "learning_rate": 1.7574658107730954e-05, + "loss": 1.037, + "step": 11403 + }, + { + "epoch": 0.73, + "learning_rate": 1.7574443418708003e-05, + "loss": 1.1432, + "step": 11404 + }, + { + "epoch": 0.73, + "learning_rate": 1.7574228729685052e-05, + "loss": 1.0567, + "step": 11405 + }, + { + "epoch": 0.73, + "learning_rate": 1.75740140406621e-05, + "loss": 1.0756, + "step": 11406 + }, + { + "epoch": 0.73, + "learning_rate": 1.757379935163915e-05, + "loss": 1.037, + "step": 11407 + }, + { + "epoch": 0.73, + "learning_rate": 1.7573584662616202e-05, + "loss": 1.1501, + "step": 11408 + }, + { + "epoch": 0.73, + "learning_rate": 1.757336997359325e-05, + "loss": 1.0674, + "step": 11409 + }, + { + "epoch": 0.73, + "learning_rate": 1.7573155284570304e-05, + "loss": 1.0889, + "step": 11410 + }, + { + "epoch": 0.73, + "learning_rate": 1.7572940595547353e-05, + "loss": 1.0271, + "step": 11411 + }, + { + "epoch": 0.73, + "learning_rate": 1.75727259065244e-05, + "loss": 0.9809, + "step": 11412 + }, + { + "epoch": 0.73, + "learning_rate": 1.757251121750145e-05, + "loss": 1.0678, + "step": 11413 + }, + { + "epoch": 0.73, + "learning_rate": 1.75722965284785e-05, + "loss": 1.0661, + "step": 11414 + }, + { + "epoch": 0.73, + "learning_rate": 1.7572081839455552e-05, + "loss": 0.9831, + "step": 11415 + }, + { + "epoch": 0.73, + "learning_rate": 1.75718671504326e-05, + "loss": 1.0499, + "step": 11416 + }, + { + "epoch": 0.73, + "learning_rate": 1.757165246140965e-05, + "loss": 1.0017, + "step": 11417 + }, + { + "epoch": 0.73, + "learning_rate": 1.75714377723867e-05, + "loss": 0.996, + "step": 11418 + }, + { + "epoch": 0.73, + "learning_rate": 1.7571223083363748e-05, + "loss": 1.0609, + "step": 11419 + }, + { + "epoch": 0.73, + "learning_rate": 1.7571008394340797e-05, + "loss": 1.0803, + "step": 11420 + }, + { + "epoch": 0.73, + "learning_rate": 1.757079370531785e-05, + "loss": 1.0392, + "step": 11421 + }, + { + "epoch": 0.73, + "learning_rate": 1.7570579016294898e-05, + "loss": 1.0318, + "step": 11422 + }, + { + "epoch": 0.73, + "learning_rate": 1.7570364327271947e-05, + "loss": 1.0626, + "step": 11423 + }, + { + "epoch": 0.73, + "learning_rate": 1.7570149638248996e-05, + "loss": 1.0671, + "step": 11424 + }, + { + "epoch": 0.74, + "learning_rate": 1.7569934949226048e-05, + "loss": 0.9846, + "step": 11425 + }, + { + "epoch": 0.74, + "learning_rate": 1.7569720260203097e-05, + "loss": 1.0857, + "step": 11426 + }, + { + "epoch": 0.74, + "learning_rate": 1.7569505571180146e-05, + "loss": 1.0889, + "step": 11427 + }, + { + "epoch": 0.74, + "learning_rate": 1.75692908821572e-05, + "loss": 1.1011, + "step": 11428 + }, + { + "epoch": 0.74, + "learning_rate": 1.7569076193134248e-05, + "loss": 1.0227, + "step": 11429 + }, + { + "epoch": 0.74, + "learning_rate": 1.7568861504111296e-05, + "loss": 1.1177, + "step": 11430 + }, + { + "epoch": 0.74, + "learning_rate": 1.7568646815088345e-05, + "loss": 0.9883, + "step": 11431 + }, + { + "epoch": 0.74, + "learning_rate": 1.7568432126065394e-05, + "loss": 1.0291, + "step": 11432 + }, + { + "epoch": 0.74, + "learning_rate": 1.7568217437042447e-05, + "loss": 1.076, + "step": 11433 + }, + { + "epoch": 0.74, + "learning_rate": 1.7568002748019496e-05, + "loss": 0.9882, + "step": 11434 + }, + { + "epoch": 0.74, + "learning_rate": 1.7567788058996545e-05, + "loss": 1.1542, + "step": 11435 + }, + { + "epoch": 0.74, + "learning_rate": 1.7567573369973594e-05, + "loss": 1.0606, + "step": 11436 + }, + { + "epoch": 0.74, + "learning_rate": 1.7567358680950643e-05, + "loss": 1.0917, + "step": 11437 + }, + { + "epoch": 0.74, + "learning_rate": 1.756714399192769e-05, + "loss": 1.0504, + "step": 11438 + }, + { + "epoch": 0.74, + "learning_rate": 1.7566929302904744e-05, + "loss": 1.0961, + "step": 11439 + }, + { + "epoch": 0.74, + "learning_rate": 1.7566714613881793e-05, + "loss": 1.0434, + "step": 11440 + }, + { + "epoch": 0.74, + "learning_rate": 1.7566499924858845e-05, + "loss": 1.0464, + "step": 11441 + }, + { + "epoch": 0.74, + "learning_rate": 1.7566285235835894e-05, + "loss": 1.0884, + "step": 11442 + }, + { + "epoch": 0.74, + "learning_rate": 1.7566070546812943e-05, + "loss": 1.0497, + "step": 11443 + }, + { + "epoch": 0.74, + "learning_rate": 1.7565855857789992e-05, + "loss": 1.0974, + "step": 11444 + }, + { + "epoch": 0.74, + "learning_rate": 1.756564116876704e-05, + "loss": 1.1584, + "step": 11445 + }, + { + "epoch": 0.74, + "learning_rate": 1.7565426479744093e-05, + "loss": 1.1144, + "step": 11446 + }, + { + "epoch": 0.74, + "learning_rate": 1.7565211790721142e-05, + "loss": 1.0663, + "step": 11447 + }, + { + "epoch": 0.74, + "learning_rate": 1.756499710169819e-05, + "loss": 1.0518, + "step": 11448 + }, + { + "epoch": 0.74, + "learning_rate": 1.756478241267524e-05, + "loss": 1.0579, + "step": 11449 + }, + { + "epoch": 0.74, + "learning_rate": 1.756456772365229e-05, + "loss": 1.0761, + "step": 11450 + }, + { + "epoch": 0.74, + "learning_rate": 1.756435303462934e-05, + "loss": 1.033, + "step": 11451 + }, + { + "epoch": 0.74, + "learning_rate": 1.756413834560639e-05, + "loss": 1.0388, + "step": 11452 + }, + { + "epoch": 0.74, + "learning_rate": 1.756392365658344e-05, + "loss": 1.1172, + "step": 11453 + }, + { + "epoch": 0.74, + "learning_rate": 1.7563708967560492e-05, + "loss": 1.1575, + "step": 11454 + }, + { + "epoch": 0.74, + "learning_rate": 1.756349427853754e-05, + "loss": 1.0618, + "step": 11455 + }, + { + "epoch": 0.74, + "learning_rate": 1.756327958951459e-05, + "loss": 1.1085, + "step": 11456 + }, + { + "epoch": 0.74, + "learning_rate": 1.756306490049164e-05, + "loss": 1.1041, + "step": 11457 + }, + { + "epoch": 0.74, + "learning_rate": 1.7562850211468688e-05, + "loss": 1.0474, + "step": 11458 + }, + { + "epoch": 0.74, + "learning_rate": 1.756263552244574e-05, + "loss": 1.0401, + "step": 11459 + }, + { + "epoch": 0.74, + "learning_rate": 1.756242083342279e-05, + "loss": 0.9946, + "step": 11460 + }, + { + "epoch": 0.74, + "learning_rate": 1.7562206144399838e-05, + "loss": 1.1252, + "step": 11461 + }, + { + "epoch": 0.74, + "learning_rate": 1.7561991455376887e-05, + "loss": 1.0581, + "step": 11462 + }, + { + "epoch": 0.74, + "learning_rate": 1.7561776766353936e-05, + "loss": 1.046, + "step": 11463 + }, + { + "epoch": 0.74, + "learning_rate": 1.756156207733099e-05, + "loss": 1.0912, + "step": 11464 + }, + { + "epoch": 0.74, + "learning_rate": 1.7561347388308037e-05, + "loss": 1.0666, + "step": 11465 + }, + { + "epoch": 0.74, + "learning_rate": 1.7561132699285086e-05, + "loss": 1.1365, + "step": 11466 + }, + { + "epoch": 0.74, + "learning_rate": 1.7560918010262135e-05, + "loss": 1.1001, + "step": 11467 + }, + { + "epoch": 0.74, + "learning_rate": 1.7560703321239188e-05, + "loss": 1.0072, + "step": 11468 + }, + { + "epoch": 0.74, + "learning_rate": 1.7560488632216237e-05, + "loss": 1.0608, + "step": 11469 + }, + { + "epoch": 0.74, + "learning_rate": 1.7560273943193286e-05, + "loss": 1.1565, + "step": 11470 + }, + { + "epoch": 0.74, + "learning_rate": 1.7560059254170338e-05, + "loss": 1.052, + "step": 11471 + }, + { + "epoch": 0.74, + "learning_rate": 1.7559844565147387e-05, + "loss": 0.9685, + "step": 11472 + }, + { + "epoch": 0.74, + "learning_rate": 1.7559629876124436e-05, + "loss": 1.1632, + "step": 11473 + }, + { + "epoch": 0.74, + "learning_rate": 1.7559415187101485e-05, + "loss": 1.1487, + "step": 11474 + }, + { + "epoch": 0.74, + "learning_rate": 1.7559200498078534e-05, + "loss": 1.0851, + "step": 11475 + }, + { + "epoch": 0.74, + "learning_rate": 1.7558985809055583e-05, + "loss": 1.0853, + "step": 11476 + }, + { + "epoch": 0.74, + "learning_rate": 1.7558771120032635e-05, + "loss": 1.0332, + "step": 11477 + }, + { + "epoch": 0.74, + "learning_rate": 1.7558556431009684e-05, + "loss": 1.1005, + "step": 11478 + }, + { + "epoch": 0.74, + "learning_rate": 1.7558341741986733e-05, + "loss": 1.0391, + "step": 11479 + }, + { + "epoch": 0.74, + "learning_rate": 1.7558127052963782e-05, + "loss": 1.0502, + "step": 11480 + }, + { + "epoch": 0.74, + "learning_rate": 1.755791236394083e-05, + "loss": 1.1575, + "step": 11481 + }, + { + "epoch": 0.74, + "learning_rate": 1.7557697674917883e-05, + "loss": 1.0677, + "step": 11482 + }, + { + "epoch": 0.74, + "learning_rate": 1.7557482985894932e-05, + "loss": 1.036, + "step": 11483 + }, + { + "epoch": 0.74, + "learning_rate": 1.7557268296871985e-05, + "loss": 1.0664, + "step": 11484 + }, + { + "epoch": 0.74, + "learning_rate": 1.7557053607849034e-05, + "loss": 1.0625, + "step": 11485 + }, + { + "epoch": 0.74, + "learning_rate": 1.7556838918826083e-05, + "loss": 1.0777, + "step": 11486 + }, + { + "epoch": 0.74, + "learning_rate": 1.755662422980313e-05, + "loss": 1.0061, + "step": 11487 + }, + { + "epoch": 0.74, + "learning_rate": 1.755640954078018e-05, + "loss": 1.0181, + "step": 11488 + }, + { + "epoch": 0.74, + "learning_rate": 1.7556194851757233e-05, + "loss": 1.0606, + "step": 11489 + }, + { + "epoch": 0.74, + "learning_rate": 1.7555980162734282e-05, + "loss": 1.0858, + "step": 11490 + }, + { + "epoch": 0.74, + "learning_rate": 1.755576547371133e-05, + "loss": 1.1239, + "step": 11491 + }, + { + "epoch": 0.74, + "learning_rate": 1.755555078468838e-05, + "loss": 1.0071, + "step": 11492 + }, + { + "epoch": 0.74, + "learning_rate": 1.755533609566543e-05, + "loss": 1.0701, + "step": 11493 + }, + { + "epoch": 0.74, + "learning_rate": 1.7555121406642478e-05, + "loss": 1.0176, + "step": 11494 + }, + { + "epoch": 0.74, + "learning_rate": 1.755490671761953e-05, + "loss": 1.0548, + "step": 11495 + }, + { + "epoch": 0.74, + "learning_rate": 1.755469202859658e-05, + "loss": 1.1059, + "step": 11496 + }, + { + "epoch": 0.74, + "learning_rate": 1.755447733957363e-05, + "loss": 1.1662, + "step": 11497 + }, + { + "epoch": 0.74, + "learning_rate": 1.755426265055068e-05, + "loss": 1.1553, + "step": 11498 + }, + { + "epoch": 0.74, + "learning_rate": 1.755404796152773e-05, + "loss": 1.0979, + "step": 11499 + }, + { + "epoch": 0.74, + "learning_rate": 1.7553833272504778e-05, + "loss": 1.0234, + "step": 11500 + }, + { + "epoch": 0.74, + "learning_rate": 1.7553618583481827e-05, + "loss": 1.0266, + "step": 11501 + }, + { + "epoch": 0.74, + "learning_rate": 1.755340389445888e-05, + "loss": 0.9763, + "step": 11502 + }, + { + "epoch": 0.74, + "learning_rate": 1.755318920543593e-05, + "loss": 1.1298, + "step": 11503 + }, + { + "epoch": 0.74, + "learning_rate": 1.7552974516412977e-05, + "loss": 1.0399, + "step": 11504 + }, + { + "epoch": 0.74, + "learning_rate": 1.7552759827390026e-05, + "loss": 1.0179, + "step": 11505 + }, + { + "epoch": 0.74, + "learning_rate": 1.7552545138367075e-05, + "loss": 1.0054, + "step": 11506 + }, + { + "epoch": 0.74, + "learning_rate": 1.7552330449344124e-05, + "loss": 1.0575, + "step": 11507 + }, + { + "epoch": 0.74, + "learning_rate": 1.7552115760321177e-05, + "loss": 1.0526, + "step": 11508 + }, + { + "epoch": 0.74, + "learning_rate": 1.7551901071298226e-05, + "loss": 1.0654, + "step": 11509 + }, + { + "epoch": 0.74, + "learning_rate": 1.7551686382275275e-05, + "loss": 1.0703, + "step": 11510 + }, + { + "epoch": 0.74, + "learning_rate": 1.7551471693252324e-05, + "loss": 1.1639, + "step": 11511 + }, + { + "epoch": 0.74, + "learning_rate": 1.7551257004229376e-05, + "loss": 1.1293, + "step": 11512 + }, + { + "epoch": 0.74, + "learning_rate": 1.7551042315206425e-05, + "loss": 1.0808, + "step": 11513 + }, + { + "epoch": 0.74, + "learning_rate": 1.7550827626183474e-05, + "loss": 1.0249, + "step": 11514 + }, + { + "epoch": 0.74, + "learning_rate": 1.7550612937160526e-05, + "loss": 1.0873, + "step": 11515 + }, + { + "epoch": 0.74, + "learning_rate": 1.7550398248137575e-05, + "loss": 1.0597, + "step": 11516 + }, + { + "epoch": 0.74, + "learning_rate": 1.7550183559114624e-05, + "loss": 1.0357, + "step": 11517 + }, + { + "epoch": 0.74, + "learning_rate": 1.7549968870091673e-05, + "loss": 1.1107, + "step": 11518 + }, + { + "epoch": 0.74, + "learning_rate": 1.7549754181068722e-05, + "loss": 1.1331, + "step": 11519 + }, + { + "epoch": 0.74, + "learning_rate": 1.7549539492045774e-05, + "loss": 1.0384, + "step": 11520 + }, + { + "epoch": 0.74, + "learning_rate": 1.7549324803022823e-05, + "loss": 1.0367, + "step": 11521 + }, + { + "epoch": 0.74, + "learning_rate": 1.7549110113999872e-05, + "loss": 1.0216, + "step": 11522 + }, + { + "epoch": 0.74, + "learning_rate": 1.754889542497692e-05, + "loss": 1.0412, + "step": 11523 + }, + { + "epoch": 0.74, + "learning_rate": 1.754868073595397e-05, + "loss": 1.1365, + "step": 11524 + }, + { + "epoch": 0.74, + "learning_rate": 1.754846604693102e-05, + "loss": 0.9723, + "step": 11525 + }, + { + "epoch": 0.74, + "learning_rate": 1.754825135790807e-05, + "loss": 1.0255, + "step": 11526 + }, + { + "epoch": 0.74, + "learning_rate": 1.754803666888512e-05, + "loss": 1.0814, + "step": 11527 + }, + { + "epoch": 0.74, + "learning_rate": 1.7547821979862173e-05, + "loss": 1.0964, + "step": 11528 + }, + { + "epoch": 0.74, + "learning_rate": 1.7547607290839222e-05, + "loss": 1.0535, + "step": 11529 + }, + { + "epoch": 0.74, + "learning_rate": 1.754739260181627e-05, + "loss": 1.0659, + "step": 11530 + }, + { + "epoch": 0.74, + "learning_rate": 1.754717791279332e-05, + "loss": 1.1686, + "step": 11531 + }, + { + "epoch": 0.74, + "learning_rate": 1.754696322377037e-05, + "loss": 1.0695, + "step": 11532 + }, + { + "epoch": 0.74, + "learning_rate": 1.754674853474742e-05, + "loss": 1.0974, + "step": 11533 + }, + { + "epoch": 0.74, + "learning_rate": 1.754653384572447e-05, + "loss": 1.031, + "step": 11534 + }, + { + "epoch": 0.74, + "learning_rate": 1.754631915670152e-05, + "loss": 1.0417, + "step": 11535 + }, + { + "epoch": 0.74, + "learning_rate": 1.7546104467678568e-05, + "loss": 1.0609, + "step": 11536 + }, + { + "epoch": 0.74, + "learning_rate": 1.7545889778655617e-05, + "loss": 1.0409, + "step": 11537 + }, + { + "epoch": 0.74, + "learning_rate": 1.754567508963267e-05, + "loss": 1.0819, + "step": 11538 + }, + { + "epoch": 0.74, + "learning_rate": 1.754546040060972e-05, + "loss": 1.0749, + "step": 11539 + }, + { + "epoch": 0.74, + "learning_rate": 1.7545245711586767e-05, + "loss": 1.0089, + "step": 11540 + }, + { + "epoch": 0.74, + "learning_rate": 1.754503102256382e-05, + "loss": 1.0858, + "step": 11541 + }, + { + "epoch": 0.74, + "learning_rate": 1.754481633354087e-05, + "loss": 0.9622, + "step": 11542 + }, + { + "epoch": 0.74, + "learning_rate": 1.7544601644517918e-05, + "loss": 1.1527, + "step": 11543 + }, + { + "epoch": 0.74, + "learning_rate": 1.7544386955494967e-05, + "loss": 1.1114, + "step": 11544 + }, + { + "epoch": 0.74, + "learning_rate": 1.7544172266472016e-05, + "loss": 1.032, + "step": 11545 + }, + { + "epoch": 0.74, + "learning_rate": 1.7543957577449068e-05, + "loss": 1.1546, + "step": 11546 + }, + { + "epoch": 0.74, + "learning_rate": 1.7543742888426117e-05, + "loss": 1.1241, + "step": 11547 + }, + { + "epoch": 0.74, + "learning_rate": 1.7543528199403166e-05, + "loss": 0.9682, + "step": 11548 + }, + { + "epoch": 0.74, + "learning_rate": 1.7543313510380215e-05, + "loss": 1.0688, + "step": 11549 + }, + { + "epoch": 0.74, + "learning_rate": 1.7543098821357264e-05, + "loss": 1.076, + "step": 11550 + }, + { + "epoch": 0.74, + "learning_rate": 1.7542884132334316e-05, + "loss": 1.0633, + "step": 11551 + }, + { + "epoch": 0.74, + "learning_rate": 1.7542669443311365e-05, + "loss": 1.0051, + "step": 11552 + }, + { + "epoch": 0.74, + "learning_rate": 1.7542454754288414e-05, + "loss": 1.0802, + "step": 11553 + }, + { + "epoch": 0.74, + "learning_rate": 1.7542240065265463e-05, + "loss": 1.1324, + "step": 11554 + }, + { + "epoch": 0.74, + "learning_rate": 1.7542025376242515e-05, + "loss": 1.1256, + "step": 11555 + }, + { + "epoch": 0.74, + "learning_rate": 1.7541810687219564e-05, + "loss": 1.0814, + "step": 11556 + }, + { + "epoch": 0.74, + "learning_rate": 1.7541595998196613e-05, + "loss": 1.0309, + "step": 11557 + }, + { + "epoch": 0.74, + "learning_rate": 1.7541381309173666e-05, + "loss": 1.0454, + "step": 11558 + }, + { + "epoch": 0.74, + "learning_rate": 1.7541166620150715e-05, + "loss": 1.1011, + "step": 11559 + }, + { + "epoch": 0.74, + "learning_rate": 1.7540951931127764e-05, + "loss": 0.9389, + "step": 11560 + }, + { + "epoch": 0.74, + "learning_rate": 1.7540737242104813e-05, + "loss": 1.0473, + "step": 11561 + }, + { + "epoch": 0.74, + "learning_rate": 1.754052255308186e-05, + "loss": 1.0485, + "step": 11562 + }, + { + "epoch": 0.74, + "learning_rate": 1.754030786405891e-05, + "loss": 1.039, + "step": 11563 + }, + { + "epoch": 0.74, + "learning_rate": 1.7540093175035963e-05, + "loss": 1.0257, + "step": 11564 + }, + { + "epoch": 0.74, + "learning_rate": 1.7539878486013012e-05, + "loss": 1.0235, + "step": 11565 + }, + { + "epoch": 0.74, + "learning_rate": 1.753966379699006e-05, + "loss": 1.1049, + "step": 11566 + }, + { + "epoch": 0.74, + "learning_rate": 1.753944910796711e-05, + "loss": 1.1067, + "step": 11567 + }, + { + "epoch": 0.74, + "learning_rate": 1.753923441894416e-05, + "loss": 1.0481, + "step": 11568 + }, + { + "epoch": 0.74, + "learning_rate": 1.753901972992121e-05, + "loss": 1.0227, + "step": 11569 + }, + { + "epoch": 0.74, + "learning_rate": 1.753880504089826e-05, + "loss": 1.0876, + "step": 11570 + }, + { + "epoch": 0.74, + "learning_rate": 1.7538590351875312e-05, + "loss": 0.9983, + "step": 11571 + }, + { + "epoch": 0.74, + "learning_rate": 1.753837566285236e-05, + "loss": 1.1273, + "step": 11572 + }, + { + "epoch": 0.74, + "learning_rate": 1.753816097382941e-05, + "loss": 1.1181, + "step": 11573 + }, + { + "epoch": 0.74, + "learning_rate": 1.753794628480646e-05, + "loss": 1.0277, + "step": 11574 + }, + { + "epoch": 0.74, + "learning_rate": 1.7537731595783508e-05, + "loss": 1.0999, + "step": 11575 + }, + { + "epoch": 0.74, + "learning_rate": 1.7537516906760557e-05, + "loss": 1.1332, + "step": 11576 + }, + { + "epoch": 0.74, + "learning_rate": 1.753730221773761e-05, + "loss": 1.0455, + "step": 11577 + }, + { + "epoch": 0.74, + "learning_rate": 1.753708752871466e-05, + "loss": 1.0216, + "step": 11578 + }, + { + "epoch": 0.74, + "learning_rate": 1.7536872839691707e-05, + "loss": 0.9759, + "step": 11579 + }, + { + "epoch": 0.75, + "learning_rate": 1.7536658150668756e-05, + "loss": 1.0889, + "step": 11580 + }, + { + "epoch": 0.75, + "learning_rate": 1.7536443461645805e-05, + "loss": 0.9453, + "step": 11581 + }, + { + "epoch": 0.75, + "learning_rate": 1.7536228772622858e-05, + "loss": 1.0242, + "step": 11582 + }, + { + "epoch": 0.75, + "learning_rate": 1.7536014083599907e-05, + "loss": 1.0412, + "step": 11583 + }, + { + "epoch": 0.75, + "learning_rate": 1.753579939457696e-05, + "loss": 1.0855, + "step": 11584 + }, + { + "epoch": 0.75, + "learning_rate": 1.7535584705554008e-05, + "loss": 1.0622, + "step": 11585 + }, + { + "epoch": 0.75, + "learning_rate": 1.7535370016531057e-05, + "loss": 1.0432, + "step": 11586 + }, + { + "epoch": 0.75, + "learning_rate": 1.7535155327508106e-05, + "loss": 1.0361, + "step": 11587 + }, + { + "epoch": 0.75, + "learning_rate": 1.7534940638485155e-05, + "loss": 1.0417, + "step": 11588 + }, + { + "epoch": 0.75, + "learning_rate": 1.7534725949462207e-05, + "loss": 1.0923, + "step": 11589 + }, + { + "epoch": 0.75, + "learning_rate": 1.7534511260439256e-05, + "loss": 1.1697, + "step": 11590 + }, + { + "epoch": 0.75, + "learning_rate": 1.7534511260439256e-05, + "loss": 1.0373, + "step": 11591 + }, + { + "epoch": 0.75, + "learning_rate": 1.7534296571416305e-05, + "loss": 0.9309, + "step": 11592 + }, + { + "epoch": 0.75, + "learning_rate": 1.7534081882393354e-05, + "loss": 1.0508, + "step": 11593 + }, + { + "epoch": 0.75, + "learning_rate": 1.7533867193370403e-05, + "loss": 1.1055, + "step": 11594 + }, + { + "epoch": 0.75, + "learning_rate": 1.7533652504347452e-05, + "loss": 1.017, + "step": 11595 + }, + { + "epoch": 0.75, + "learning_rate": 1.7533437815324504e-05, + "loss": 1.1593, + "step": 11596 + }, + { + "epoch": 0.75, + "learning_rate": 1.7533223126301553e-05, + "loss": 1.0981, + "step": 11597 + }, + { + "epoch": 0.75, + "learning_rate": 1.7533008437278602e-05, + "loss": 1.0236, + "step": 11598 + }, + { + "epoch": 0.75, + "learning_rate": 1.753279374825565e-05, + "loss": 1.0721, + "step": 11599 + }, + { + "epoch": 0.75, + "learning_rate": 1.7532579059232704e-05, + "loss": 1.1281, + "step": 11600 + }, + { + "epoch": 0.75, + "learning_rate": 1.7532364370209753e-05, + "loss": 1.0788, + "step": 11601 + }, + { + "epoch": 0.75, + "learning_rate": 1.75321496811868e-05, + "loss": 0.9923, + "step": 11602 + }, + { + "epoch": 0.75, + "learning_rate": 1.7531934992163854e-05, + "loss": 1.0099, + "step": 11603 + }, + { + "epoch": 0.75, + "learning_rate": 1.7531720303140903e-05, + "loss": 1.1537, + "step": 11604 + }, + { + "epoch": 0.75, + "learning_rate": 1.7531505614117952e-05, + "loss": 1.0808, + "step": 11605 + }, + { + "epoch": 0.75, + "learning_rate": 1.7531290925095e-05, + "loss": 1.0859, + "step": 11606 + }, + { + "epoch": 0.75, + "learning_rate": 1.753107623607205e-05, + "loss": 1.1734, + "step": 11607 + }, + { + "epoch": 0.75, + "learning_rate": 1.7530861547049102e-05, + "loss": 1.2417, + "step": 11608 + }, + { + "epoch": 0.75, + "learning_rate": 1.753064685802615e-05, + "loss": 1.1875, + "step": 11609 + }, + { + "epoch": 0.75, + "learning_rate": 1.75304321690032e-05, + "loss": 1.3327, + "step": 11610 + }, + { + "epoch": 0.75, + "learning_rate": 1.753021747998025e-05, + "loss": 1.2297, + "step": 11611 + }, + { + "epoch": 0.75, + "learning_rate": 1.7530002790957298e-05, + "loss": 1.2025, + "step": 11612 + }, + { + "epoch": 0.75, + "learning_rate": 1.7529788101934347e-05, + "loss": 1.119, + "step": 11613 + }, + { + "epoch": 0.75, + "learning_rate": 1.75295734129114e-05, + "loss": 1.2298, + "step": 11614 + }, + { + "epoch": 0.75, + "learning_rate": 1.752935872388845e-05, + "loss": 1.1815, + "step": 11615 + }, + { + "epoch": 0.75, + "learning_rate": 1.75291440348655e-05, + "loss": 1.0911, + "step": 11616 + }, + { + "epoch": 0.75, + "learning_rate": 1.752892934584255e-05, + "loss": 1.1525, + "step": 11617 + }, + { + "epoch": 0.75, + "learning_rate": 1.752892934584255e-05, + "loss": 1.2187, + "step": 11618 + }, + { + "epoch": 0.75, + "learning_rate": 1.75287146568196e-05, + "loss": 1.1358, + "step": 11619 + }, + { + "epoch": 0.75, + "learning_rate": 1.7528499967796648e-05, + "loss": 1.1785, + "step": 11620 + }, + { + "epoch": 0.75, + "learning_rate": 1.7528285278773697e-05, + "loss": 1.0984, + "step": 11621 + }, + { + "epoch": 0.75, + "learning_rate": 1.752807058975075e-05, + "loss": 1.1589, + "step": 11622 + }, + { + "epoch": 0.75, + "learning_rate": 1.7527855900727798e-05, + "loss": 1.0249, + "step": 11623 + }, + { + "epoch": 0.75, + "learning_rate": 1.7527641211704847e-05, + "loss": 1.0892, + "step": 11624 + }, + { + "epoch": 0.75, + "learning_rate": 1.7527426522681896e-05, + "loss": 1.0529, + "step": 11625 + }, + { + "epoch": 0.75, + "learning_rate": 1.7527211833658945e-05, + "loss": 1.1571, + "step": 11626 + }, + { + "epoch": 0.75, + "learning_rate": 1.7526997144635997e-05, + "loss": 1.1022, + "step": 11627 + }, + { + "epoch": 0.75, + "learning_rate": 1.7526782455613046e-05, + "loss": 1.1348, + "step": 11628 + }, + { + "epoch": 0.75, + "learning_rate": 1.7526567766590095e-05, + "loss": 1.098, + "step": 11629 + }, + { + "epoch": 0.75, + "learning_rate": 1.7526353077567147e-05, + "loss": 1.078, + "step": 11630 + }, + { + "epoch": 0.75, + "learning_rate": 1.7526138388544196e-05, + "loss": 1.1825, + "step": 11631 + }, + { + "epoch": 0.75, + "learning_rate": 1.7525923699521245e-05, + "loss": 1.0965, + "step": 11632 + }, + { + "epoch": 0.75, + "learning_rate": 1.7525923699521245e-05, + "loss": 1.1369, + "step": 11633 + }, + { + "epoch": 0.75, + "learning_rate": 1.7525709010498294e-05, + "loss": 1.1628, + "step": 11634 + }, + { + "epoch": 0.75, + "learning_rate": 1.7525494321475343e-05, + "loss": 1.111, + "step": 11635 + }, + { + "epoch": 0.75, + "learning_rate": 1.7525279632452396e-05, + "loss": 1.1009, + "step": 11636 + }, + { + "epoch": 0.75, + "learning_rate": 1.7525064943429445e-05, + "loss": 1.2373, + "step": 11637 + }, + { + "epoch": 0.75, + "learning_rate": 1.7524850254406494e-05, + "loss": 1.1955, + "step": 11638 + }, + { + "epoch": 0.75, + "learning_rate": 1.7524635565383542e-05, + "loss": 1.197, + "step": 11639 + }, + { + "epoch": 0.75, + "learning_rate": 1.752442087636059e-05, + "loss": 1.2346, + "step": 11640 + }, + { + "epoch": 0.75, + "learning_rate": 1.7524206187337644e-05, + "loss": 1.2104, + "step": 11641 + }, + { + "epoch": 0.75, + "learning_rate": 1.7523991498314693e-05, + "loss": 1.349, + "step": 11642 + }, + { + "epoch": 0.75, + "learning_rate": 1.7523776809291742e-05, + "loss": 1.135, + "step": 11643 + }, + { + "epoch": 0.75, + "learning_rate": 1.752356212026879e-05, + "loss": 1.2906, + "step": 11644 + }, + { + "epoch": 0.75, + "learning_rate": 1.7523347431245843e-05, + "loss": 1.2378, + "step": 11645 + }, + { + "epoch": 0.75, + "learning_rate": 1.7523132742222892e-05, + "loss": 1.4414, + "step": 11646 + }, + { + "epoch": 0.75, + "learning_rate": 1.752291805319994e-05, + "loss": 1.2779, + "step": 11647 + }, + { + "epoch": 0.75, + "learning_rate": 1.7522703364176993e-05, + "loss": 1.1727, + "step": 11648 + }, + { + "epoch": 0.75, + "learning_rate": 1.7522488675154042e-05, + "loss": 1.3666, + "step": 11649 + }, + { + "epoch": 0.75, + "learning_rate": 1.752227398613109e-05, + "loss": 1.2437, + "step": 11650 + }, + { + "epoch": 0.75, + "learning_rate": 1.752205929710814e-05, + "loss": 1.0544, + "step": 11651 + }, + { + "epoch": 0.75, + "learning_rate": 1.752184460808519e-05, + "loss": 1.0924, + "step": 11652 + }, + { + "epoch": 0.75, + "learning_rate": 1.7521629919062238e-05, + "loss": 1.0594, + "step": 11653 + }, + { + "epoch": 0.75, + "learning_rate": 1.752141523003929e-05, + "loss": 1.1596, + "step": 11654 + }, + { + "epoch": 0.75, + "learning_rate": 1.752120054101634e-05, + "loss": 1.0016, + "step": 11655 + }, + { + "epoch": 0.75, + "learning_rate": 1.752098585199339e-05, + "loss": 1.0684, + "step": 11656 + }, + { + "epoch": 0.75, + "learning_rate": 1.7520771162970437e-05, + "loss": 1.0865, + "step": 11657 + }, + { + "epoch": 0.75, + "learning_rate": 1.7520556473947486e-05, + "loss": 1.0182, + "step": 11658 + }, + { + "epoch": 0.75, + "learning_rate": 1.752034178492454e-05, + "loss": 1.0901, + "step": 11659 + }, + { + "epoch": 0.75, + "learning_rate": 1.7520127095901588e-05, + "loss": 1.0659, + "step": 11660 + }, + { + "epoch": 0.75, + "learning_rate": 1.751991240687864e-05, + "loss": 1.0962, + "step": 11661 + }, + { + "epoch": 0.75, + "learning_rate": 1.751969771785569e-05, + "loss": 1.0287, + "step": 11662 + }, + { + "epoch": 0.75, + "learning_rate": 1.7519483028832738e-05, + "loss": 1.1313, + "step": 11663 + }, + { + "epoch": 0.75, + "learning_rate": 1.7519268339809787e-05, + "loss": 1.0688, + "step": 11664 + }, + { + "epoch": 0.75, + "learning_rate": 1.7519053650786836e-05, + "loss": 0.9543, + "step": 11665 + }, + { + "epoch": 0.75, + "learning_rate": 1.7518838961763885e-05, + "loss": 1.0892, + "step": 11666 + }, + { + "epoch": 0.75, + "learning_rate": 1.7518624272740937e-05, + "loss": 1.1121, + "step": 11667 + }, + { + "epoch": 0.75, + "learning_rate": 1.7518409583717986e-05, + "loss": 1.0944, + "step": 11668 + }, + { + "epoch": 0.75, + "learning_rate": 1.7518194894695035e-05, + "loss": 1.0147, + "step": 11669 + }, + { + "epoch": 0.75, + "learning_rate": 1.7517980205672084e-05, + "loss": 1.0593, + "step": 11670 + }, + { + "epoch": 0.75, + "learning_rate": 1.7517765516649133e-05, + "loss": 1.0916, + "step": 11671 + }, + { + "epoch": 0.75, + "learning_rate": 1.7517550827626185e-05, + "loss": 1.0716, + "step": 11672 + }, + { + "epoch": 0.75, + "learning_rate": 1.7517336138603234e-05, + "loss": 1.1482, + "step": 11673 + }, + { + "epoch": 0.75, + "learning_rate": 1.7517121449580287e-05, + "loss": 1.0138, + "step": 11674 + }, + { + "epoch": 0.75, + "learning_rate": 1.7516906760557336e-05, + "loss": 1.037, + "step": 11675 + }, + { + "epoch": 0.75, + "learning_rate": 1.7516692071534385e-05, + "loss": 1.1765, + "step": 11676 + }, + { + "epoch": 0.75, + "learning_rate": 1.7516477382511434e-05, + "loss": 1.0385, + "step": 11677 + }, + { + "epoch": 0.75, + "learning_rate": 1.7516262693488483e-05, + "loss": 1.0137, + "step": 11678 + }, + { + "epoch": 0.75, + "learning_rate": 1.7516048004465535e-05, + "loss": 1.058, + "step": 11679 + }, + { + "epoch": 0.75, + "learning_rate": 1.7515833315442584e-05, + "loss": 1.0658, + "step": 11680 + }, + { + "epoch": 0.75, + "learning_rate": 1.7515618626419633e-05, + "loss": 1.0579, + "step": 11681 + }, + { + "epoch": 0.75, + "learning_rate": 1.7515403937396682e-05, + "loss": 0.9767, + "step": 11682 + }, + { + "epoch": 0.75, + "learning_rate": 1.751518924837373e-05, + "loss": 1.096, + "step": 11683 + }, + { + "epoch": 0.75, + "learning_rate": 1.751497455935078e-05, + "loss": 1.1797, + "step": 11684 + }, + { + "epoch": 0.75, + "learning_rate": 1.7514759870327832e-05, + "loss": 1.0183, + "step": 11685 + }, + { + "epoch": 0.75, + "learning_rate": 1.751454518130488e-05, + "loss": 1.0891, + "step": 11686 + }, + { + "epoch": 0.75, + "learning_rate": 1.751433049228193e-05, + "loss": 1.0531, + "step": 11687 + }, + { + "epoch": 0.75, + "learning_rate": 1.751411580325898e-05, + "loss": 1.1051, + "step": 11688 + }, + { + "epoch": 0.75, + "learning_rate": 1.751390111423603e-05, + "loss": 1.0436, + "step": 11689 + }, + { + "epoch": 0.75, + "learning_rate": 1.751368642521308e-05, + "loss": 1.0211, + "step": 11690 + }, + { + "epoch": 0.75, + "learning_rate": 1.751347173619013e-05, + "loss": 0.9892, + "step": 11691 + }, + { + "epoch": 0.75, + "learning_rate": 1.751325704716718e-05, + "loss": 1.1417, + "step": 11692 + }, + { + "epoch": 0.75, + "learning_rate": 1.751304235814423e-05, + "loss": 1.0631, + "step": 11693 + }, + { + "epoch": 0.75, + "learning_rate": 1.751282766912128e-05, + "loss": 1.0696, + "step": 11694 + }, + { + "epoch": 0.75, + "learning_rate": 1.751261298009833e-05, + "loss": 1.021, + "step": 11695 + }, + { + "epoch": 0.75, + "learning_rate": 1.7512398291075378e-05, + "loss": 1.0206, + "step": 11696 + }, + { + "epoch": 0.75, + "learning_rate": 1.751218360205243e-05, + "loss": 1.1225, + "step": 11697 + }, + { + "epoch": 0.75, + "learning_rate": 1.751196891302948e-05, + "loss": 1.0477, + "step": 11698 + }, + { + "epoch": 0.75, + "learning_rate": 1.7511754224006528e-05, + "loss": 1.0393, + "step": 11699 + }, + { + "epoch": 0.75, + "learning_rate": 1.7511539534983577e-05, + "loss": 1.0716, + "step": 11700 + }, + { + "epoch": 0.75, + "learning_rate": 1.7511324845960626e-05, + "loss": 1.0636, + "step": 11701 + }, + { + "epoch": 0.75, + "learning_rate": 1.7511110156937675e-05, + "loss": 1.0719, + "step": 11702 + }, + { + "epoch": 0.75, + "learning_rate": 1.7510895467914727e-05, + "loss": 1.0256, + "step": 11703 + }, + { + "epoch": 0.75, + "learning_rate": 1.7510680778891776e-05, + "loss": 1.0486, + "step": 11704 + }, + { + "epoch": 0.75, + "learning_rate": 1.751046608986883e-05, + "loss": 1.0465, + "step": 11705 + }, + { + "epoch": 0.75, + "learning_rate": 1.7510251400845877e-05, + "loss": 0.9805, + "step": 11706 + }, + { + "epoch": 0.75, + "learning_rate": 1.7510036711822926e-05, + "loss": 1.0004, + "step": 11707 + }, + { + "epoch": 0.75, + "learning_rate": 1.7509822022799975e-05, + "loss": 1.0585, + "step": 11708 + }, + { + "epoch": 0.75, + "learning_rate": 1.7509607333777024e-05, + "loss": 1.0335, + "step": 11709 + }, + { + "epoch": 0.75, + "learning_rate": 1.7509392644754077e-05, + "loss": 1.1194, + "step": 11710 + }, + { + "epoch": 0.75, + "learning_rate": 1.7509177955731126e-05, + "loss": 1.1164, + "step": 11711 + }, + { + "epoch": 0.75, + "learning_rate": 1.7508963266708175e-05, + "loss": 1.1425, + "step": 11712 + }, + { + "epoch": 0.75, + "learning_rate": 1.7508748577685223e-05, + "loss": 1.0379, + "step": 11713 + }, + { + "epoch": 0.75, + "learning_rate": 1.7508533888662272e-05, + "loss": 0.9971, + "step": 11714 + }, + { + "epoch": 0.75, + "learning_rate": 1.7508319199639325e-05, + "loss": 0.9696, + "step": 11715 + }, + { + "epoch": 0.75, + "learning_rate": 1.7508104510616374e-05, + "loss": 1.1326, + "step": 11716 + }, + { + "epoch": 0.75, + "learning_rate": 1.7507889821593423e-05, + "loss": 1.0826, + "step": 11717 + }, + { + "epoch": 0.75, + "learning_rate": 1.7507675132570475e-05, + "loss": 1.1145, + "step": 11718 + }, + { + "epoch": 0.75, + "learning_rate": 1.7507460443547524e-05, + "loss": 1.1147, + "step": 11719 + }, + { + "epoch": 0.75, + "learning_rate": 1.7507245754524573e-05, + "loss": 1.1092, + "step": 11720 + }, + { + "epoch": 0.75, + "learning_rate": 1.7507031065501622e-05, + "loss": 1.0039, + "step": 11721 + }, + { + "epoch": 0.75, + "learning_rate": 1.750681637647867e-05, + "loss": 1.1039, + "step": 11722 + }, + { + "epoch": 0.75, + "learning_rate": 1.7506601687455723e-05, + "loss": 1.0454, + "step": 11723 + }, + { + "epoch": 0.75, + "learning_rate": 1.7506386998432772e-05, + "loss": 1.012, + "step": 11724 + }, + { + "epoch": 0.75, + "learning_rate": 1.750617230940982e-05, + "loss": 0.9821, + "step": 11725 + }, + { + "epoch": 0.75, + "learning_rate": 1.750595762038687e-05, + "loss": 1.0008, + "step": 11726 + }, + { + "epoch": 0.75, + "learning_rate": 1.750574293136392e-05, + "loss": 1.0261, + "step": 11727 + }, + { + "epoch": 0.75, + "learning_rate": 1.750552824234097e-05, + "loss": 1.0447, + "step": 11728 + }, + { + "epoch": 0.75, + "learning_rate": 1.750531355331802e-05, + "loss": 1.0508, + "step": 11729 + }, + { + "epoch": 0.75, + "learning_rate": 1.750509886429507e-05, + "loss": 1.131, + "step": 11730 + }, + { + "epoch": 0.75, + "learning_rate": 1.750488417527212e-05, + "loss": 1.0074, + "step": 11731 + }, + { + "epoch": 0.75, + "learning_rate": 1.750466948624917e-05, + "loss": 1.092, + "step": 11732 + }, + { + "epoch": 0.75, + "learning_rate": 1.750445479722622e-05, + "loss": 1.0265, + "step": 11733 + }, + { + "epoch": 0.75, + "learning_rate": 1.750424010820327e-05, + "loss": 1.0859, + "step": 11734 + }, + { + "epoch": 0.76, + "learning_rate": 1.750402541918032e-05, + "loss": 1.0601, + "step": 11735 + }, + { + "epoch": 0.76, + "learning_rate": 1.750381073015737e-05, + "loss": 1.2011, + "step": 11736 + }, + { + "epoch": 0.76, + "learning_rate": 1.750359604113442e-05, + "loss": 1.0536, + "step": 11737 + }, + { + "epoch": 0.76, + "learning_rate": 1.7503381352111468e-05, + "loss": 1.0382, + "step": 11738 + }, + { + "epoch": 0.76, + "learning_rate": 1.7503166663088517e-05, + "loss": 1.0663, + "step": 11739 + }, + { + "epoch": 0.76, + "learning_rate": 1.7502951974065566e-05, + "loss": 1.0092, + "step": 11740 + }, + { + "epoch": 0.76, + "learning_rate": 1.7502737285042618e-05, + "loss": 1.0559, + "step": 11741 + }, + { + "epoch": 0.76, + "learning_rate": 1.7502522596019667e-05, + "loss": 0.9704, + "step": 11742 + }, + { + "epoch": 0.76, + "learning_rate": 1.7502307906996716e-05, + "loss": 1.105, + "step": 11743 + }, + { + "epoch": 0.76, + "learning_rate": 1.7502093217973765e-05, + "loss": 1.1072, + "step": 11744 + }, + { + "epoch": 0.76, + "learning_rate": 1.7501878528950814e-05, + "loss": 1.0502, + "step": 11745 + }, + { + "epoch": 0.76, + "learning_rate": 1.7501663839927866e-05, + "loss": 1.1555, + "step": 11746 + }, + { + "epoch": 0.76, + "learning_rate": 1.7501449150904915e-05, + "loss": 1.0124, + "step": 11747 + }, + { + "epoch": 0.76, + "learning_rate": 1.7501234461881968e-05, + "loss": 1.1007, + "step": 11748 + }, + { + "epoch": 0.76, + "learning_rate": 1.7501019772859017e-05, + "loss": 1.1097, + "step": 11749 + }, + { + "epoch": 0.76, + "learning_rate": 1.7500805083836066e-05, + "loss": 1.0086, + "step": 11750 + }, + { + "epoch": 0.76, + "learning_rate": 1.7500590394813115e-05, + "loss": 1.0864, + "step": 11751 + }, + { + "epoch": 0.76, + "learning_rate": 1.7500375705790164e-05, + "loss": 1.1717, + "step": 11752 + }, + { + "epoch": 0.76, + "learning_rate": 1.7500161016767213e-05, + "loss": 1.0614, + "step": 11753 + }, + { + "epoch": 0.76, + "learning_rate": 1.7499946327744265e-05, + "loss": 1.0652, + "step": 11754 + }, + { + "epoch": 0.76, + "learning_rate": 1.7499731638721314e-05, + "loss": 1.0942, + "step": 11755 + }, + { + "epoch": 0.76, + "learning_rate": 1.7499516949698363e-05, + "loss": 1.0353, + "step": 11756 + }, + { + "epoch": 0.76, + "learning_rate": 1.7499302260675412e-05, + "loss": 1.048, + "step": 11757 + }, + { + "epoch": 0.76, + "learning_rate": 1.749908757165246e-05, + "loss": 1.0066, + "step": 11758 + }, + { + "epoch": 0.76, + "learning_rate": 1.7498872882629513e-05, + "loss": 1.0776, + "step": 11759 + }, + { + "epoch": 0.76, + "learning_rate": 1.7498658193606562e-05, + "loss": 1.0467, + "step": 11760 + }, + { + "epoch": 0.76, + "learning_rate": 1.749844350458361e-05, + "loss": 1.0516, + "step": 11761 + }, + { + "epoch": 0.76, + "learning_rate": 1.7498228815560663e-05, + "loss": 1.0083, + "step": 11762 + }, + { + "epoch": 0.76, + "learning_rate": 1.7498014126537712e-05, + "loss": 1.0273, + "step": 11763 + }, + { + "epoch": 0.76, + "learning_rate": 1.749779943751476e-05, + "loss": 1.0675, + "step": 11764 + }, + { + "epoch": 0.76, + "learning_rate": 1.749758474849181e-05, + "loss": 1.0659, + "step": 11765 + }, + { + "epoch": 0.76, + "learning_rate": 1.7497370059468863e-05, + "loss": 1.035, + "step": 11766 + }, + { + "epoch": 0.76, + "learning_rate": 1.749715537044591e-05, + "loss": 1.0714, + "step": 11767 + }, + { + "epoch": 0.76, + "learning_rate": 1.749694068142296e-05, + "loss": 1.1882, + "step": 11768 + }, + { + "epoch": 0.76, + "learning_rate": 1.749672599240001e-05, + "loss": 1.0143, + "step": 11769 + }, + { + "epoch": 0.76, + "learning_rate": 1.749651130337706e-05, + "loss": 1.0373, + "step": 11770 + }, + { + "epoch": 0.76, + "learning_rate": 1.7496296614354108e-05, + "loss": 1.2159, + "step": 11771 + }, + { + "epoch": 0.76, + "learning_rate": 1.749608192533116e-05, + "loss": 1.0668, + "step": 11772 + }, + { + "epoch": 0.76, + "learning_rate": 1.749586723630821e-05, + "loss": 1.1179, + "step": 11773 + }, + { + "epoch": 0.76, + "learning_rate": 1.7495652547285258e-05, + "loss": 1.1236, + "step": 11774 + }, + { + "epoch": 0.76, + "learning_rate": 1.7495437858262307e-05, + "loss": 1.0949, + "step": 11775 + }, + { + "epoch": 0.76, + "learning_rate": 1.749522316923936e-05, + "loss": 1.0946, + "step": 11776 + }, + { + "epoch": 0.76, + "learning_rate": 1.7495008480216408e-05, + "loss": 1.0578, + "step": 11777 + }, + { + "epoch": 0.76, + "learning_rate": 1.7494793791193457e-05, + "loss": 1.0372, + "step": 11778 + }, + { + "epoch": 0.76, + "learning_rate": 1.749457910217051e-05, + "loss": 1.0541, + "step": 11779 + }, + { + "epoch": 0.76, + "learning_rate": 1.749436441314756e-05, + "loss": 1.1217, + "step": 11780 + }, + { + "epoch": 0.76, + "learning_rate": 1.7494149724124607e-05, + "loss": 1.1292, + "step": 11781 + }, + { + "epoch": 0.76, + "learning_rate": 1.7493935035101656e-05, + "loss": 1.0279, + "step": 11782 + }, + { + "epoch": 0.76, + "learning_rate": 1.7493720346078705e-05, + "loss": 1.0625, + "step": 11783 + }, + { + "epoch": 0.76, + "learning_rate": 1.7493505657055758e-05, + "loss": 1.0303, + "step": 11784 + }, + { + "epoch": 0.76, + "learning_rate": 1.7493290968032807e-05, + "loss": 1.0861, + "step": 11785 + }, + { + "epoch": 0.76, + "learning_rate": 1.7493076279009856e-05, + "loss": 1.0871, + "step": 11786 + }, + { + "epoch": 0.76, + "learning_rate": 1.7492861589986905e-05, + "loss": 1.0857, + "step": 11787 + }, + { + "epoch": 0.76, + "learning_rate": 1.7492646900963953e-05, + "loss": 1.173, + "step": 11788 + }, + { + "epoch": 0.76, + "learning_rate": 1.7492432211941002e-05, + "loss": 1.0063, + "step": 11789 + }, + { + "epoch": 0.76, + "learning_rate": 1.7492217522918055e-05, + "loss": 1.1129, + "step": 11790 + }, + { + "epoch": 0.76, + "learning_rate": 1.7492002833895104e-05, + "loss": 1.0537, + "step": 11791 + }, + { + "epoch": 0.76, + "learning_rate": 1.7491788144872156e-05, + "loss": 1.1019, + "step": 11792 + }, + { + "epoch": 0.76, + "learning_rate": 1.7491573455849205e-05, + "loss": 1.0803, + "step": 11793 + }, + { + "epoch": 0.76, + "learning_rate": 1.7491358766826254e-05, + "loss": 1.0344, + "step": 11794 + }, + { + "epoch": 0.76, + "learning_rate": 1.7491144077803303e-05, + "loss": 0.9881, + "step": 11795 + }, + { + "epoch": 0.76, + "learning_rate": 1.7490929388780352e-05, + "loss": 1.0681, + "step": 11796 + }, + { + "epoch": 0.76, + "learning_rate": 1.7490714699757404e-05, + "loss": 1.0607, + "step": 11797 + }, + { + "epoch": 0.76, + "learning_rate": 1.7490500010734453e-05, + "loss": 1.0921, + "step": 11798 + }, + { + "epoch": 0.76, + "learning_rate": 1.7490285321711502e-05, + "loss": 1.0614, + "step": 11799 + }, + { + "epoch": 0.76, + "learning_rate": 1.749007063268855e-05, + "loss": 1.0481, + "step": 11800 + }, + { + "epoch": 0.76, + "learning_rate": 1.74898559436656e-05, + "loss": 1.0481, + "step": 11801 + }, + { + "epoch": 0.76, + "learning_rate": 1.748964125464265e-05, + "loss": 1.0954, + "step": 11802 + }, + { + "epoch": 0.76, + "learning_rate": 1.74894265656197e-05, + "loss": 1.002, + "step": 11803 + }, + { + "epoch": 0.76, + "learning_rate": 1.748921187659675e-05, + "loss": 1.0527, + "step": 11804 + }, + { + "epoch": 0.76, + "learning_rate": 1.7488997187573803e-05, + "loss": 1.1104, + "step": 11805 + }, + { + "epoch": 0.76, + "learning_rate": 1.7488782498550852e-05, + "loss": 1.0797, + "step": 11806 + }, + { + "epoch": 0.76, + "learning_rate": 1.74885678095279e-05, + "loss": 1.0199, + "step": 11807 + }, + { + "epoch": 0.76, + "learning_rate": 1.748835312050495e-05, + "loss": 1.0248, + "step": 11808 + }, + { + "epoch": 0.76, + "learning_rate": 1.7488138431482e-05, + "loss": 1.0356, + "step": 11809 + }, + { + "epoch": 0.76, + "learning_rate": 1.748792374245905e-05, + "loss": 0.9795, + "step": 11810 + }, + { + "epoch": 0.76, + "learning_rate": 1.74877090534361e-05, + "loss": 1.0853, + "step": 11811 + }, + { + "epoch": 0.76, + "learning_rate": 1.748749436441315e-05, + "loss": 1.052, + "step": 11812 + }, + { + "epoch": 0.76, + "learning_rate": 1.7487279675390198e-05, + "loss": 1.0039, + "step": 11813 + }, + { + "epoch": 0.76, + "learning_rate": 1.7487064986367247e-05, + "loss": 1.0855, + "step": 11814 + }, + { + "epoch": 0.76, + "learning_rate": 1.74868502973443e-05, + "loss": 1.1281, + "step": 11815 + }, + { + "epoch": 0.76, + "learning_rate": 1.7486635608321348e-05, + "loss": 1.0925, + "step": 11816 + }, + { + "epoch": 0.76, + "learning_rate": 1.7486420919298397e-05, + "loss": 1.0434, + "step": 11817 + }, + { + "epoch": 0.76, + "learning_rate": 1.7486206230275446e-05, + "loss": 1.0046, + "step": 11818 + }, + { + "epoch": 0.76, + "learning_rate": 1.7485991541252495e-05, + "loss": 1.0905, + "step": 11819 + }, + { + "epoch": 0.76, + "learning_rate": 1.7485776852229547e-05, + "loss": 0.9903, + "step": 11820 + }, + { + "epoch": 0.76, + "learning_rate": 1.7485562163206596e-05, + "loss": 1.0865, + "step": 11821 + }, + { + "epoch": 0.76, + "learning_rate": 1.748534747418365e-05, + "loss": 1.0476, + "step": 11822 + }, + { + "epoch": 0.76, + "learning_rate": 1.7485132785160698e-05, + "loss": 1.1104, + "step": 11823 + }, + { + "epoch": 0.76, + "learning_rate": 1.7484918096137747e-05, + "loss": 1.0873, + "step": 11824 + }, + { + "epoch": 0.76, + "learning_rate": 1.7484703407114796e-05, + "loss": 0.9244, + "step": 11825 + }, + { + "epoch": 0.76, + "learning_rate": 1.7484488718091845e-05, + "loss": 1.1207, + "step": 11826 + }, + { + "epoch": 0.76, + "learning_rate": 1.7484274029068894e-05, + "loss": 1.0974, + "step": 11827 + }, + { + "epoch": 0.76, + "learning_rate": 1.7484059340045946e-05, + "loss": 1.0299, + "step": 11828 + }, + { + "epoch": 0.76, + "learning_rate": 1.7483844651022995e-05, + "loss": 1.0812, + "step": 11829 + }, + { + "epoch": 0.76, + "learning_rate": 1.7483629962000044e-05, + "loss": 1.1629, + "step": 11830 + }, + { + "epoch": 0.76, + "learning_rate": 1.7483415272977093e-05, + "loss": 1.0079, + "step": 11831 + }, + { + "epoch": 0.76, + "learning_rate": 1.7483200583954142e-05, + "loss": 1.0631, + "step": 11832 + }, + { + "epoch": 0.76, + "learning_rate": 1.7482985894931194e-05, + "loss": 1.0324, + "step": 11833 + }, + { + "epoch": 0.76, + "learning_rate": 1.7482771205908243e-05, + "loss": 1.0971, + "step": 11834 + }, + { + "epoch": 0.76, + "learning_rate": 1.7482556516885295e-05, + "loss": 1.0992, + "step": 11835 + }, + { + "epoch": 0.76, + "learning_rate": 1.7482341827862344e-05, + "loss": 1.0106, + "step": 11836 + }, + { + "epoch": 0.76, + "learning_rate": 1.7482127138839393e-05, + "loss": 0.9539, + "step": 11837 + }, + { + "epoch": 0.76, + "learning_rate": 1.7481912449816442e-05, + "loss": 1.0193, + "step": 11838 + }, + { + "epoch": 0.76, + "learning_rate": 1.748169776079349e-05, + "loss": 1.1066, + "step": 11839 + }, + { + "epoch": 0.76, + "learning_rate": 1.748148307177054e-05, + "loss": 1.1314, + "step": 11840 + }, + { + "epoch": 0.76, + "learning_rate": 1.7481268382747593e-05, + "loss": 1.0531, + "step": 11841 + }, + { + "epoch": 0.76, + "learning_rate": 1.748105369372464e-05, + "loss": 1.152, + "step": 11842 + }, + { + "epoch": 0.76, + "learning_rate": 1.748083900470169e-05, + "loss": 1.0999, + "step": 11843 + }, + { + "epoch": 0.76, + "learning_rate": 1.748062431567874e-05, + "loss": 1.059, + "step": 11844 + }, + { + "epoch": 0.76, + "learning_rate": 1.748040962665579e-05, + "loss": 1.0485, + "step": 11845 + }, + { + "epoch": 0.76, + "learning_rate": 1.748019493763284e-05, + "loss": 1.0201, + "step": 11846 + }, + { + "epoch": 0.76, + "learning_rate": 1.747998024860989e-05, + "loss": 1.0553, + "step": 11847 + }, + { + "epoch": 0.76, + "learning_rate": 1.747976555958694e-05, + "loss": 1.051, + "step": 11848 + }, + { + "epoch": 0.76, + "learning_rate": 1.747955087056399e-05, + "loss": 1.141, + "step": 11849 + }, + { + "epoch": 0.76, + "learning_rate": 1.747933618154104e-05, + "loss": 1.0345, + "step": 11850 + }, + { + "epoch": 0.76, + "learning_rate": 1.747912149251809e-05, + "loss": 1.0003, + "step": 11851 + }, + { + "epoch": 0.76, + "learning_rate": 1.7478906803495138e-05, + "loss": 1.0067, + "step": 11852 + }, + { + "epoch": 0.76, + "learning_rate": 1.747869211447219e-05, + "loss": 1.021, + "step": 11853 + }, + { + "epoch": 0.76, + "learning_rate": 1.747847742544924e-05, + "loss": 1.0588, + "step": 11854 + }, + { + "epoch": 0.76, + "learning_rate": 1.747826273642629e-05, + "loss": 1.079, + "step": 11855 + }, + { + "epoch": 0.76, + "learning_rate": 1.7478048047403337e-05, + "loss": 1.0509, + "step": 11856 + }, + { + "epoch": 0.76, + "learning_rate": 1.7477833358380386e-05, + "loss": 1.1229, + "step": 11857 + }, + { + "epoch": 0.76, + "learning_rate": 1.7477618669357435e-05, + "loss": 1.0169, + "step": 11858 + }, + { + "epoch": 0.76, + "learning_rate": 1.7477403980334488e-05, + "loss": 1.0298, + "step": 11859 + }, + { + "epoch": 0.76, + "learning_rate": 1.7477189291311537e-05, + "loss": 1.0749, + "step": 11860 + }, + { + "epoch": 0.76, + "learning_rate": 1.7476974602288586e-05, + "loss": 1.148, + "step": 11861 + }, + { + "epoch": 0.76, + "learning_rate": 1.7476759913265634e-05, + "loss": 1.0617, + "step": 11862 + }, + { + "epoch": 0.76, + "learning_rate": 1.7476545224242687e-05, + "loss": 1.0721, + "step": 11863 + }, + { + "epoch": 0.76, + "learning_rate": 1.7476330535219736e-05, + "loss": 1.1476, + "step": 11864 + }, + { + "epoch": 0.76, + "learning_rate": 1.7476115846196785e-05, + "loss": 1.0335, + "step": 11865 + }, + { + "epoch": 0.76, + "learning_rate": 1.7475901157173837e-05, + "loss": 1.1008, + "step": 11866 + }, + { + "epoch": 0.76, + "learning_rate": 1.7475686468150886e-05, + "loss": 1.1075, + "step": 11867 + }, + { + "epoch": 0.76, + "learning_rate": 1.7475471779127935e-05, + "loss": 1.126, + "step": 11868 + }, + { + "epoch": 0.76, + "learning_rate": 1.7475257090104984e-05, + "loss": 1.0479, + "step": 11869 + }, + { + "epoch": 0.76, + "learning_rate": 1.7475042401082033e-05, + "loss": 1.0854, + "step": 11870 + }, + { + "epoch": 0.76, + "learning_rate": 1.7474827712059085e-05, + "loss": 1.0082, + "step": 11871 + }, + { + "epoch": 0.76, + "learning_rate": 1.7474613023036134e-05, + "loss": 1.0011, + "step": 11872 + }, + { + "epoch": 0.76, + "learning_rate": 1.7474398334013183e-05, + "loss": 1.0838, + "step": 11873 + }, + { + "epoch": 0.76, + "learning_rate": 1.7474183644990232e-05, + "loss": 1.013, + "step": 11874 + }, + { + "epoch": 0.76, + "learning_rate": 1.747396895596728e-05, + "loss": 0.9701, + "step": 11875 + }, + { + "epoch": 0.76, + "learning_rate": 1.747375426694433e-05, + "loss": 1.0534, + "step": 11876 + }, + { + "epoch": 0.76, + "learning_rate": 1.7473539577921382e-05, + "loss": 1.0442, + "step": 11877 + }, + { + "epoch": 0.76, + "learning_rate": 1.747332488889843e-05, + "loss": 1.0943, + "step": 11878 + }, + { + "epoch": 0.76, + "learning_rate": 1.7473110199875484e-05, + "loss": 1.0532, + "step": 11879 + }, + { + "epoch": 0.76, + "learning_rate": 1.7472895510852533e-05, + "loss": 1.0415, + "step": 11880 + }, + { + "epoch": 0.76, + "learning_rate": 1.7472680821829582e-05, + "loss": 1.105, + "step": 11881 + }, + { + "epoch": 0.76, + "learning_rate": 1.747246613280663e-05, + "loss": 1.0664, + "step": 11882 + }, + { + "epoch": 0.76, + "learning_rate": 1.747225144378368e-05, + "loss": 1.0979, + "step": 11883 + }, + { + "epoch": 0.76, + "learning_rate": 1.7472036754760732e-05, + "loss": 1.0755, + "step": 11884 + }, + { + "epoch": 0.76, + "learning_rate": 1.747182206573778e-05, + "loss": 1.0341, + "step": 11885 + }, + { + "epoch": 0.76, + "learning_rate": 1.747160737671483e-05, + "loss": 1.0611, + "step": 11886 + }, + { + "epoch": 0.76, + "learning_rate": 1.747139268769188e-05, + "loss": 1.0167, + "step": 11887 + }, + { + "epoch": 0.76, + "learning_rate": 1.7471177998668928e-05, + "loss": 1.1291, + "step": 11888 + }, + { + "epoch": 0.76, + "learning_rate": 1.7470963309645977e-05, + "loss": 0.9737, + "step": 11889 + }, + { + "epoch": 0.76, + "learning_rate": 1.747074862062303e-05, + "loss": 1.0539, + "step": 11890 + }, + { + "epoch": 0.77, + "learning_rate": 1.7470533931600078e-05, + "loss": 1.0209, + "step": 11891 + }, + { + "epoch": 0.77, + "learning_rate": 1.747031924257713e-05, + "loss": 1.029, + "step": 11892 + }, + { + "epoch": 0.77, + "learning_rate": 1.747010455355418e-05, + "loss": 1.0715, + "step": 11893 + }, + { + "epoch": 0.77, + "learning_rate": 1.746988986453123e-05, + "loss": 1.0502, + "step": 11894 + }, + { + "epoch": 0.77, + "learning_rate": 1.7469675175508277e-05, + "loss": 1.03, + "step": 11895 + }, + { + "epoch": 0.77, + "learning_rate": 1.7469460486485326e-05, + "loss": 1.0894, + "step": 11896 + }, + { + "epoch": 0.77, + "learning_rate": 1.746924579746238e-05, + "loss": 1.068, + "step": 11897 + }, + { + "epoch": 0.77, + "learning_rate": 1.7469031108439428e-05, + "loss": 1.0218, + "step": 11898 + }, + { + "epoch": 0.77, + "learning_rate": 1.7468816419416477e-05, + "loss": 1.0291, + "step": 11899 + }, + { + "epoch": 0.77, + "learning_rate": 1.7468601730393526e-05, + "loss": 1.1132, + "step": 11900 + }, + { + "epoch": 0.77, + "learning_rate": 1.7468387041370575e-05, + "loss": 1.0628, + "step": 11901 + }, + { + "epoch": 0.77, + "learning_rate": 1.7468172352347627e-05, + "loss": 1.0153, + "step": 11902 + }, + { + "epoch": 0.77, + "learning_rate": 1.7467957663324676e-05, + "loss": 1.0704, + "step": 11903 + }, + { + "epoch": 0.77, + "learning_rate": 1.7467742974301725e-05, + "loss": 1.013, + "step": 11904 + }, + { + "epoch": 0.77, + "learning_rate": 1.7467528285278774e-05, + "loss": 1.0248, + "step": 11905 + }, + { + "epoch": 0.77, + "learning_rate": 1.7467313596255823e-05, + "loss": 1.0711, + "step": 11906 + }, + { + "epoch": 0.77, + "learning_rate": 1.7467098907232875e-05, + "loss": 1.0991, + "step": 11907 + }, + { + "epoch": 0.77, + "learning_rate": 1.7466884218209924e-05, + "loss": 1.0488, + "step": 11908 + }, + { + "epoch": 0.77, + "learning_rate": 1.7466669529186973e-05, + "loss": 1.0672, + "step": 11909 + }, + { + "epoch": 0.77, + "learning_rate": 1.7466454840164025e-05, + "loss": 1.0747, + "step": 11910 + }, + { + "epoch": 0.77, + "learning_rate": 1.7466240151141074e-05, + "loss": 1.0742, + "step": 11911 + }, + { + "epoch": 0.77, + "learning_rate": 1.7466025462118123e-05, + "loss": 1.1429, + "step": 11912 + }, + { + "epoch": 0.77, + "learning_rate": 1.7465810773095172e-05, + "loss": 1.039, + "step": 11913 + }, + { + "epoch": 0.77, + "learning_rate": 1.746559608407222e-05, + "loss": 1.0929, + "step": 11914 + }, + { + "epoch": 0.77, + "learning_rate": 1.7465381395049274e-05, + "loss": 1.0675, + "step": 11915 + }, + { + "epoch": 0.77, + "learning_rate": 1.7465166706026323e-05, + "loss": 1.0694, + "step": 11916 + }, + { + "epoch": 0.77, + "learning_rate": 1.746495201700337e-05, + "loss": 1.122, + "step": 11917 + }, + { + "epoch": 0.77, + "learning_rate": 1.746473732798042e-05, + "loss": 1.076, + "step": 11918 + }, + { + "epoch": 0.77, + "learning_rate": 1.746452263895747e-05, + "loss": 1.0581, + "step": 11919 + }, + { + "epoch": 0.77, + "learning_rate": 1.7464307949934522e-05, + "loss": 1.0049, + "step": 11920 + }, + { + "epoch": 0.77, + "learning_rate": 1.746409326091157e-05, + "loss": 1.02, + "step": 11921 + }, + { + "epoch": 0.77, + "learning_rate": 1.7463878571888623e-05, + "loss": 1.0408, + "step": 11922 + }, + { + "epoch": 0.77, + "learning_rate": 1.7463663882865672e-05, + "loss": 1.0246, + "step": 11923 + }, + { + "epoch": 0.77, + "learning_rate": 1.746344919384272e-05, + "loss": 1.0541, + "step": 11924 + }, + { + "epoch": 0.77, + "learning_rate": 1.746323450481977e-05, + "loss": 1.0844, + "step": 11925 + }, + { + "epoch": 0.77, + "learning_rate": 1.746301981579682e-05, + "loss": 1.0557, + "step": 11926 + }, + { + "epoch": 0.77, + "learning_rate": 1.7462805126773868e-05, + "loss": 1.1334, + "step": 11927 + }, + { + "epoch": 0.77, + "learning_rate": 1.746259043775092e-05, + "loss": 1.0371, + "step": 11928 + }, + { + "epoch": 0.77, + "learning_rate": 1.746237574872797e-05, + "loss": 1.1202, + "step": 11929 + }, + { + "epoch": 0.77, + "learning_rate": 1.7462161059705018e-05, + "loss": 0.9924, + "step": 11930 + }, + { + "epoch": 0.77, + "learning_rate": 1.7461946370682067e-05, + "loss": 1.0313, + "step": 11931 + }, + { + "epoch": 0.77, + "learning_rate": 1.7461731681659116e-05, + "loss": 1.0686, + "step": 11932 + }, + { + "epoch": 0.77, + "learning_rate": 1.746151699263617e-05, + "loss": 1.0537, + "step": 11933 + }, + { + "epoch": 0.77, + "learning_rate": 1.7461302303613218e-05, + "loss": 1.0612, + "step": 11934 + }, + { + "epoch": 0.77, + "learning_rate": 1.7461087614590267e-05, + "loss": 1.0304, + "step": 11935 + }, + { + "epoch": 0.77, + "learning_rate": 1.746087292556732e-05, + "loss": 1.0614, + "step": 11936 + }, + { + "epoch": 0.77, + "learning_rate": 1.7460658236544368e-05, + "loss": 1.0796, + "step": 11937 + }, + { + "epoch": 0.77, + "learning_rate": 1.7460443547521417e-05, + "loss": 1.1509, + "step": 11938 + }, + { + "epoch": 0.77, + "learning_rate": 1.7460228858498466e-05, + "loss": 1.0917, + "step": 11939 + }, + { + "epoch": 0.77, + "learning_rate": 1.7460014169475518e-05, + "loss": 1.1056, + "step": 11940 + }, + { + "epoch": 0.77, + "learning_rate": 1.7459799480452567e-05, + "loss": 1.0378, + "step": 11941 + }, + { + "epoch": 0.77, + "learning_rate": 1.7459584791429616e-05, + "loss": 1.0948, + "step": 11942 + }, + { + "epoch": 0.77, + "learning_rate": 1.7459370102406665e-05, + "loss": 1.0327, + "step": 11943 + }, + { + "epoch": 0.77, + "learning_rate": 1.7459155413383714e-05, + "loss": 1.0592, + "step": 11944 + }, + { + "epoch": 0.77, + "learning_rate": 1.7458940724360763e-05, + "loss": 1.1033, + "step": 11945 + }, + { + "epoch": 0.77, + "learning_rate": 1.7458726035337815e-05, + "loss": 1.0779, + "step": 11946 + }, + { + "epoch": 0.77, + "learning_rate": 1.7458511346314864e-05, + "loss": 1.1743, + "step": 11947 + }, + { + "epoch": 0.77, + "learning_rate": 1.7458296657291913e-05, + "loss": 0.9908, + "step": 11948 + }, + { + "epoch": 0.77, + "learning_rate": 1.7458081968268962e-05, + "loss": 1.0987, + "step": 11949 + }, + { + "epoch": 0.77, + "learning_rate": 1.7457867279246015e-05, + "loss": 1.0455, + "step": 11950 + }, + { + "epoch": 0.77, + "learning_rate": 1.7457652590223064e-05, + "loss": 1.0395, + "step": 11951 + }, + { + "epoch": 0.77, + "learning_rate": 1.7457437901200112e-05, + "loss": 1.078, + "step": 11952 + }, + { + "epoch": 0.77, + "learning_rate": 1.7457223212177165e-05, + "loss": 1.0885, + "step": 11953 + }, + { + "epoch": 0.77, + "learning_rate": 1.7457008523154214e-05, + "loss": 1.0501, + "step": 11954 + }, + { + "epoch": 0.77, + "learning_rate": 1.7456793834131263e-05, + "loss": 1.0837, + "step": 11955 + }, + { + "epoch": 0.77, + "learning_rate": 1.7456579145108312e-05, + "loss": 1.043, + "step": 11956 + }, + { + "epoch": 0.77, + "learning_rate": 1.745636445608536e-05, + "loss": 1.0792, + "step": 11957 + }, + { + "epoch": 0.77, + "learning_rate": 1.7456149767062413e-05, + "loss": 1.098, + "step": 11958 + }, + { + "epoch": 0.77, + "learning_rate": 1.7455935078039462e-05, + "loss": 1.0631, + "step": 11959 + }, + { + "epoch": 0.77, + "learning_rate": 1.745572038901651e-05, + "loss": 0.9835, + "step": 11960 + }, + { + "epoch": 0.77, + "learning_rate": 1.745550569999356e-05, + "loss": 1.0687, + "step": 11961 + }, + { + "epoch": 0.77, + "learning_rate": 1.745529101097061e-05, + "loss": 1.0727, + "step": 11962 + }, + { + "epoch": 0.77, + "learning_rate": 1.7455076321947658e-05, + "loss": 1.1637, + "step": 11963 + }, + { + "epoch": 0.77, + "learning_rate": 1.745486163292471e-05, + "loss": 1.0453, + "step": 11964 + }, + { + "epoch": 0.77, + "learning_rate": 1.745464694390176e-05, + "loss": 1.0907, + "step": 11965 + }, + { + "epoch": 0.77, + "learning_rate": 1.745443225487881e-05, + "loss": 1.0369, + "step": 11966 + }, + { + "epoch": 0.77, + "learning_rate": 1.745421756585586e-05, + "loss": 0.9794, + "step": 11967 + }, + { + "epoch": 0.77, + "learning_rate": 1.745400287683291e-05, + "loss": 0.9928, + "step": 11968 + }, + { + "epoch": 0.77, + "learning_rate": 1.745378818780996e-05, + "loss": 1.0812, + "step": 11969 + }, + { + "epoch": 0.77, + "learning_rate": 1.7453573498787007e-05, + "loss": 1.0583, + "step": 11970 + }, + { + "epoch": 0.77, + "learning_rate": 1.745335880976406e-05, + "loss": 1.0844, + "step": 11971 + }, + { + "epoch": 0.77, + "learning_rate": 1.745314412074111e-05, + "loss": 1.0957, + "step": 11972 + }, + { + "epoch": 0.77, + "learning_rate": 1.7452929431718158e-05, + "loss": 1.0873, + "step": 11973 + }, + { + "epoch": 0.77, + "learning_rate": 1.7452714742695207e-05, + "loss": 0.9529, + "step": 11974 + }, + { + "epoch": 0.77, + "learning_rate": 1.7452500053672256e-05, + "loss": 1.0523, + "step": 11975 + }, + { + "epoch": 0.77, + "learning_rate": 1.7452285364649305e-05, + "loss": 1.0685, + "step": 11976 + }, + { + "epoch": 0.77, + "learning_rate": 1.7452070675626357e-05, + "loss": 1.0456, + "step": 11977 + }, + { + "epoch": 0.77, + "learning_rate": 1.7451855986603406e-05, + "loss": 1.0541, + "step": 11978 + }, + { + "epoch": 0.77, + "learning_rate": 1.7451641297580458e-05, + "loss": 1.1169, + "step": 11979 + }, + { + "epoch": 0.77, + "learning_rate": 1.7451426608557507e-05, + "loss": 1.1312, + "step": 11980 + }, + { + "epoch": 0.77, + "learning_rate": 1.7451211919534556e-05, + "loss": 1.1365, + "step": 11981 + }, + { + "epoch": 0.77, + "learning_rate": 1.7450997230511605e-05, + "loss": 1.0232, + "step": 11982 + }, + { + "epoch": 0.77, + "learning_rate": 1.7450782541488654e-05, + "loss": 1.059, + "step": 11983 + }, + { + "epoch": 0.77, + "learning_rate": 1.7450567852465706e-05, + "loss": 1.0593, + "step": 11984 + }, + { + "epoch": 0.77, + "learning_rate": 1.7450353163442755e-05, + "loss": 1.1302, + "step": 11985 + }, + { + "epoch": 0.77, + "learning_rate": 1.7450138474419804e-05, + "loss": 1.1506, + "step": 11986 + }, + { + "epoch": 0.77, + "learning_rate": 1.7449923785396853e-05, + "loss": 1.0027, + "step": 11987 + }, + { + "epoch": 0.77, + "learning_rate": 1.7449709096373902e-05, + "loss": 1.1276, + "step": 11988 + }, + { + "epoch": 0.77, + "learning_rate": 1.7449494407350955e-05, + "loss": 0.993, + "step": 11989 + }, + { + "epoch": 0.77, + "learning_rate": 1.7449279718328004e-05, + "loss": 1.0149, + "step": 11990 + }, + { + "epoch": 0.77, + "learning_rate": 1.7449065029305053e-05, + "loss": 1.1568, + "step": 11991 + }, + { + "epoch": 0.77, + "learning_rate": 1.74488503402821e-05, + "loss": 1.0503, + "step": 11992 + }, + { + "epoch": 0.77, + "learning_rate": 1.744863565125915e-05, + "loss": 1.1576, + "step": 11993 + }, + { + "epoch": 0.77, + "learning_rate": 1.7448420962236203e-05, + "loss": 1.1167, + "step": 11994 + }, + { + "epoch": 0.77, + "learning_rate": 1.7448206273213252e-05, + "loss": 1.1122, + "step": 11995 + }, + { + "epoch": 0.77, + "learning_rate": 1.74479915841903e-05, + "loss": 1.106, + "step": 11996 + }, + { + "epoch": 0.77, + "learning_rate": 1.7447776895167353e-05, + "loss": 1.026, + "step": 11997 + }, + { + "epoch": 0.77, + "learning_rate": 1.7447562206144402e-05, + "loss": 1.0892, + "step": 11998 + }, + { + "epoch": 0.77, + "learning_rate": 1.744734751712145e-05, + "loss": 1.0758, + "step": 11999 + }, + { + "epoch": 0.77, + "learning_rate": 1.74471328280985e-05, + "loss": 1.0326, + "step": 12000 + }, + { + "epoch": 0.77, + "eval_laion/cider": 0.4767249397727165, + "eval_laion/meteor": 0.3807762001141635, + "eval_laion/pred_num_words": 36.751953125, + "eval_laion/sacrebleu": 15.200864937806573, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.076338529586792, + "eval_runtime": 98.1191, + "eval_samples_per_second": 5.086, + "eval_steps_per_second": 0.326, + "step": 12000 + }, + { + "epoch": 0.77, + "learning_rate": 1.744691813907555e-05, + "loss": 1.0146, + "step": 12001 + }, + { + "epoch": 0.77, + "learning_rate": 1.74467034500526e-05, + "loss": 1.0318, + "step": 12002 + }, + { + "epoch": 0.77, + "learning_rate": 1.744648876102965e-05, + "loss": 1.0528, + "step": 12003 + }, + { + "epoch": 0.77, + "learning_rate": 1.74462740720067e-05, + "loss": 1.072, + "step": 12004 + }, + { + "epoch": 0.77, + "learning_rate": 1.7446059382983748e-05, + "loss": 0.9915, + "step": 12005 + }, + { + "epoch": 0.77, + "learning_rate": 1.7445844693960797e-05, + "loss": 1.0201, + "step": 12006 + }, + { + "epoch": 0.77, + "learning_rate": 1.744563000493785e-05, + "loss": 0.9847, + "step": 12007 + }, + { + "epoch": 0.77, + "learning_rate": 1.74454153159149e-05, + "loss": 1.1086, + "step": 12008 + }, + { + "epoch": 0.77, + "learning_rate": 1.744520062689195e-05, + "loss": 1.0045, + "step": 12009 + }, + { + "epoch": 0.77, + "learning_rate": 1.7444985937869e-05, + "loss": 1.0011, + "step": 12010 + }, + { + "epoch": 0.77, + "learning_rate": 1.744477124884605e-05, + "loss": 1.0855, + "step": 12011 + }, + { + "epoch": 0.77, + "learning_rate": 1.7444556559823098e-05, + "loss": 0.9896, + "step": 12012 + }, + { + "epoch": 0.77, + "learning_rate": 1.7444341870800147e-05, + "loss": 1.0403, + "step": 12013 + }, + { + "epoch": 0.77, + "learning_rate": 1.7444127181777196e-05, + "loss": 1.0896, + "step": 12014 + }, + { + "epoch": 0.77, + "learning_rate": 1.7443912492754248e-05, + "loss": 1.0617, + "step": 12015 + }, + { + "epoch": 0.77, + "learning_rate": 1.7443697803731297e-05, + "loss": 1.0922, + "step": 12016 + }, + { + "epoch": 0.77, + "learning_rate": 1.7443483114708346e-05, + "loss": 1.0439, + "step": 12017 + }, + { + "epoch": 0.77, + "learning_rate": 1.7443268425685395e-05, + "loss": 1.1148, + "step": 12018 + }, + { + "epoch": 0.77, + "learning_rate": 1.7443053736662444e-05, + "loss": 1.0703, + "step": 12019 + }, + { + "epoch": 0.77, + "learning_rate": 1.7442839047639496e-05, + "loss": 1.0302, + "step": 12020 + }, + { + "epoch": 0.77, + "learning_rate": 1.7442624358616545e-05, + "loss": 1.1004, + "step": 12021 + }, + { + "epoch": 0.77, + "learning_rate": 1.7442409669593594e-05, + "loss": 0.9349, + "step": 12022 + }, + { + "epoch": 0.77, + "learning_rate": 1.7442194980570647e-05, + "loss": 1.1273, + "step": 12023 + }, + { + "epoch": 0.77, + "learning_rate": 1.7441980291547696e-05, + "loss": 1.0496, + "step": 12024 + }, + { + "epoch": 0.77, + "learning_rate": 1.7441765602524745e-05, + "loss": 1.1115, + "step": 12025 + }, + { + "epoch": 0.77, + "learning_rate": 1.7441550913501793e-05, + "loss": 1.0834, + "step": 12026 + }, + { + "epoch": 0.77, + "learning_rate": 1.7441336224478846e-05, + "loss": 1.0376, + "step": 12027 + }, + { + "epoch": 0.77, + "learning_rate": 1.7441121535455895e-05, + "loss": 1.1606, + "step": 12028 + }, + { + "epoch": 0.77, + "learning_rate": 1.7440906846432944e-05, + "loss": 1.0477, + "step": 12029 + }, + { + "epoch": 0.77, + "learning_rate": 1.7440692157409993e-05, + "loss": 1.0797, + "step": 12030 + }, + { + "epoch": 0.77, + "learning_rate": 1.744047746838704e-05, + "loss": 1.0153, + "step": 12031 + }, + { + "epoch": 0.77, + "learning_rate": 1.744026277936409e-05, + "loss": 1.0157, + "step": 12032 + }, + { + "epoch": 0.77, + "learning_rate": 1.7440048090341143e-05, + "loss": 1.0856, + "step": 12033 + }, + { + "epoch": 0.77, + "learning_rate": 1.7439833401318192e-05, + "loss": 1.0754, + "step": 12034 + }, + { + "epoch": 0.77, + "learning_rate": 1.743961871229524e-05, + "loss": 1.12, + "step": 12035 + }, + { + "epoch": 0.77, + "learning_rate": 1.743940402327229e-05, + "loss": 1.0922, + "step": 12036 + }, + { + "epoch": 0.77, + "learning_rate": 1.7439189334249342e-05, + "loss": 1.1285, + "step": 12037 + }, + { + "epoch": 0.77, + "learning_rate": 1.743897464522639e-05, + "loss": 1.0542, + "step": 12038 + }, + { + "epoch": 0.77, + "learning_rate": 1.743875995620344e-05, + "loss": 1.0739, + "step": 12039 + }, + { + "epoch": 0.77, + "learning_rate": 1.7438545267180493e-05, + "loss": 1.0812, + "step": 12040 + }, + { + "epoch": 0.77, + "learning_rate": 1.743833057815754e-05, + "loss": 1.0341, + "step": 12041 + }, + { + "epoch": 0.77, + "learning_rate": 1.743811588913459e-05, + "loss": 1.0359, + "step": 12042 + }, + { + "epoch": 0.77, + "learning_rate": 1.743790120011164e-05, + "loss": 1.1083, + "step": 12043 + }, + { + "epoch": 0.77, + "learning_rate": 1.743768651108869e-05, + "loss": 1.0994, + "step": 12044 + }, + { + "epoch": 0.77, + "learning_rate": 1.743747182206574e-05, + "loss": 1.0655, + "step": 12045 + }, + { + "epoch": 0.78, + "learning_rate": 1.743725713304279e-05, + "loss": 1.0741, + "step": 12046 + }, + { + "epoch": 0.78, + "learning_rate": 1.743704244401984e-05, + "loss": 1.1069, + "step": 12047 + }, + { + "epoch": 0.78, + "learning_rate": 1.7436827754996888e-05, + "loss": 1.1492, + "step": 12048 + }, + { + "epoch": 0.78, + "learning_rate": 1.7436613065973937e-05, + "loss": 1.0317, + "step": 12049 + }, + { + "epoch": 0.78, + "learning_rate": 1.7436398376950986e-05, + "loss": 1.0556, + "step": 12050 + }, + { + "epoch": 0.78, + "learning_rate": 1.7436183687928038e-05, + "loss": 1.0574, + "step": 12051 + }, + { + "epoch": 0.78, + "learning_rate": 1.7435968998905087e-05, + "loss": 1.1362, + "step": 12052 + }, + { + "epoch": 0.78, + "learning_rate": 1.743575430988214e-05, + "loss": 1.0736, + "step": 12053 + }, + { + "epoch": 0.78, + "learning_rate": 1.7435539620859188e-05, + "loss": 1.1406, + "step": 12054 + }, + { + "epoch": 0.78, + "learning_rate": 1.7435324931836237e-05, + "loss": 1.0311, + "step": 12055 + }, + { + "epoch": 0.78, + "learning_rate": 1.7435110242813286e-05, + "loss": 1.0279, + "step": 12056 + }, + { + "epoch": 0.78, + "learning_rate": 1.7434895553790335e-05, + "loss": 1.1048, + "step": 12057 + }, + { + "epoch": 0.78, + "learning_rate": 1.7434680864767387e-05, + "loss": 1.0403, + "step": 12058 + }, + { + "epoch": 0.78, + "learning_rate": 1.7434466175744436e-05, + "loss": 1.0086, + "step": 12059 + }, + { + "epoch": 0.78, + "learning_rate": 1.7434251486721485e-05, + "loss": 1.0668, + "step": 12060 + }, + { + "epoch": 0.78, + "learning_rate": 1.7434036797698534e-05, + "loss": 1.0699, + "step": 12061 + }, + { + "epoch": 0.78, + "learning_rate": 1.7433822108675583e-05, + "loss": 1.0687, + "step": 12062 + }, + { + "epoch": 0.78, + "learning_rate": 1.7433607419652632e-05, + "loss": 1.0503, + "step": 12063 + }, + { + "epoch": 0.78, + "learning_rate": 1.7433392730629685e-05, + "loss": 1.0196, + "step": 12064 + }, + { + "epoch": 0.78, + "learning_rate": 1.7433178041606734e-05, + "loss": 1.1027, + "step": 12065 + }, + { + "epoch": 0.78, + "learning_rate": 1.7432963352583786e-05, + "loss": 1.0517, + "step": 12066 + }, + { + "epoch": 0.78, + "learning_rate": 1.7432748663560835e-05, + "loss": 1.0002, + "step": 12067 + }, + { + "epoch": 0.78, + "learning_rate": 1.7432533974537884e-05, + "loss": 1.0227, + "step": 12068 + }, + { + "epoch": 0.78, + "learning_rate": 1.7432319285514933e-05, + "loss": 1.0651, + "step": 12069 + }, + { + "epoch": 0.78, + "learning_rate": 1.7432104596491982e-05, + "loss": 1.0349, + "step": 12070 + }, + { + "epoch": 0.78, + "learning_rate": 1.7431889907469034e-05, + "loss": 1.0102, + "step": 12071 + }, + { + "epoch": 0.78, + "learning_rate": 1.7431675218446083e-05, + "loss": 1.1032, + "step": 12072 + }, + { + "epoch": 0.78, + "learning_rate": 1.7431460529423132e-05, + "loss": 1.1086, + "step": 12073 + }, + { + "epoch": 0.78, + "learning_rate": 1.743124584040018e-05, + "loss": 1.0163, + "step": 12074 + }, + { + "epoch": 0.78, + "learning_rate": 1.743103115137723e-05, + "loss": 1.0307, + "step": 12075 + }, + { + "epoch": 0.78, + "learning_rate": 1.7430816462354282e-05, + "loss": 1.0833, + "step": 12076 + }, + { + "epoch": 0.78, + "learning_rate": 1.743060177333133e-05, + "loss": 0.9943, + "step": 12077 + }, + { + "epoch": 0.78, + "learning_rate": 1.743038708430838e-05, + "loss": 1.0865, + "step": 12078 + }, + { + "epoch": 0.78, + "learning_rate": 1.743017239528543e-05, + "loss": 1.0498, + "step": 12079 + }, + { + "epoch": 0.78, + "learning_rate": 1.7429957706262478e-05, + "loss": 1.012, + "step": 12080 + }, + { + "epoch": 0.78, + "learning_rate": 1.742974301723953e-05, + "loss": 1.0151, + "step": 12081 + }, + { + "epoch": 0.78, + "learning_rate": 1.742952832821658e-05, + "loss": 1.0348, + "step": 12082 + }, + { + "epoch": 0.78, + "learning_rate": 1.742931363919363e-05, + "loss": 1.1301, + "step": 12083 + }, + { + "epoch": 0.78, + "learning_rate": 1.742909895017068e-05, + "loss": 1.0451, + "step": 12084 + }, + { + "epoch": 0.78, + "learning_rate": 1.742888426114773e-05, + "loss": 1.0863, + "step": 12085 + }, + { + "epoch": 0.78, + "learning_rate": 1.742866957212478e-05, + "loss": 1.0216, + "step": 12086 + }, + { + "epoch": 0.78, + "learning_rate": 1.7428454883101828e-05, + "loss": 1.0908, + "step": 12087 + }, + { + "epoch": 0.78, + "learning_rate": 1.7428240194078877e-05, + "loss": 1.0634, + "step": 12088 + }, + { + "epoch": 0.78, + "learning_rate": 1.742802550505593e-05, + "loss": 1.0908, + "step": 12089 + }, + { + "epoch": 0.78, + "learning_rate": 1.7427810816032978e-05, + "loss": 1.1064, + "step": 12090 + }, + { + "epoch": 0.78, + "learning_rate": 1.7427596127010027e-05, + "loss": 1.0943, + "step": 12091 + }, + { + "epoch": 0.78, + "learning_rate": 1.7427381437987076e-05, + "loss": 1.0907, + "step": 12092 + }, + { + "epoch": 0.78, + "learning_rate": 1.7427166748964125e-05, + "loss": 1.1119, + "step": 12093 + }, + { + "epoch": 0.78, + "learning_rate": 1.7426952059941177e-05, + "loss": 1.0479, + "step": 12094 + }, + { + "epoch": 0.78, + "learning_rate": 1.7426737370918226e-05, + "loss": 1.0644, + "step": 12095 + }, + { + "epoch": 0.78, + "learning_rate": 1.742652268189528e-05, + "loss": 1.0482, + "step": 12096 + }, + { + "epoch": 0.78, + "learning_rate": 1.7426307992872328e-05, + "loss": 1.0531, + "step": 12097 + }, + { + "epoch": 0.78, + "learning_rate": 1.7426093303849377e-05, + "loss": 1.0522, + "step": 12098 + }, + { + "epoch": 0.78, + "learning_rate": 1.7425878614826426e-05, + "loss": 1.0846, + "step": 12099 + }, + { + "epoch": 0.78, + "learning_rate": 1.7425663925803474e-05, + "loss": 1.0798, + "step": 12100 + }, + { + "epoch": 0.78, + "learning_rate": 1.7425449236780523e-05, + "loss": 1.0286, + "step": 12101 + }, + { + "epoch": 0.78, + "learning_rate": 1.7425234547757576e-05, + "loss": 1.0916, + "step": 12102 + }, + { + "epoch": 0.78, + "learning_rate": 1.7425019858734625e-05, + "loss": 0.9921, + "step": 12103 + }, + { + "epoch": 0.78, + "learning_rate": 1.7424805169711674e-05, + "loss": 1.0189, + "step": 12104 + }, + { + "epoch": 0.78, + "learning_rate": 1.7424590480688723e-05, + "loss": 1.0647, + "step": 12105 + }, + { + "epoch": 0.78, + "learning_rate": 1.742437579166577e-05, + "loss": 1.0089, + "step": 12106 + }, + { + "epoch": 0.78, + "learning_rate": 1.7424161102642824e-05, + "loss": 1.2142, + "step": 12107 + }, + { + "epoch": 0.78, + "learning_rate": 1.7423946413619873e-05, + "loss": 0.9587, + "step": 12108 + }, + { + "epoch": 0.78, + "learning_rate": 1.7423731724596922e-05, + "loss": 0.9584, + "step": 12109 + }, + { + "epoch": 0.78, + "learning_rate": 1.7423517035573974e-05, + "loss": 1.1503, + "step": 12110 + }, + { + "epoch": 0.78, + "learning_rate": 1.7423302346551023e-05, + "loss": 1.0395, + "step": 12111 + }, + { + "epoch": 0.78, + "learning_rate": 1.7423087657528072e-05, + "loss": 1.0983, + "step": 12112 + }, + { + "epoch": 0.78, + "learning_rate": 1.742287296850512e-05, + "loss": 1.0566, + "step": 12113 + }, + { + "epoch": 0.78, + "learning_rate": 1.7422658279482174e-05, + "loss": 1.1471, + "step": 12114 + }, + { + "epoch": 0.78, + "learning_rate": 1.7422443590459222e-05, + "loss": 0.9758, + "step": 12115 + }, + { + "epoch": 0.78, + "learning_rate": 1.742222890143627e-05, + "loss": 1.1093, + "step": 12116 + }, + { + "epoch": 0.78, + "learning_rate": 1.742201421241332e-05, + "loss": 1.1511, + "step": 12117 + }, + { + "epoch": 0.78, + "learning_rate": 1.742179952339037e-05, + "loss": 0.9793, + "step": 12118 + }, + { + "epoch": 0.78, + "learning_rate": 1.742158483436742e-05, + "loss": 1.0674, + "step": 12119 + }, + { + "epoch": 0.78, + "learning_rate": 1.742137014534447e-05, + "loss": 1.106, + "step": 12120 + }, + { + "epoch": 0.78, + "learning_rate": 1.742115545632152e-05, + "loss": 1.0443, + "step": 12121 + }, + { + "epoch": 0.78, + "learning_rate": 1.742094076729857e-05, + "loss": 1.0438, + "step": 12122 + }, + { + "epoch": 0.78, + "learning_rate": 1.7420726078275618e-05, + "loss": 1.0163, + "step": 12123 + }, + { + "epoch": 0.78, + "learning_rate": 1.7420511389252667e-05, + "loss": 1.0492, + "step": 12124 + }, + { + "epoch": 0.78, + "learning_rate": 1.742029670022972e-05, + "loss": 1.0332, + "step": 12125 + }, + { + "epoch": 0.78, + "learning_rate": 1.7420082011206768e-05, + "loss": 1.0469, + "step": 12126 + }, + { + "epoch": 0.78, + "learning_rate": 1.741986732218382e-05, + "loss": 1.0549, + "step": 12127 + }, + { + "epoch": 0.78, + "learning_rate": 1.741965263316087e-05, + "loss": 1.0424, + "step": 12128 + }, + { + "epoch": 0.78, + "learning_rate": 1.7419437944137918e-05, + "loss": 0.9894, + "step": 12129 + }, + { + "epoch": 0.78, + "learning_rate": 1.7419223255114967e-05, + "loss": 1.1038, + "step": 12130 + }, + { + "epoch": 0.78, + "learning_rate": 1.7419008566092016e-05, + "loss": 1.121, + "step": 12131 + }, + { + "epoch": 0.78, + "learning_rate": 1.741879387706907e-05, + "loss": 1.0052, + "step": 12132 + }, + { + "epoch": 0.78, + "learning_rate": 1.7418579188046117e-05, + "loss": 1.0932, + "step": 12133 + }, + { + "epoch": 0.78, + "learning_rate": 1.7418364499023166e-05, + "loss": 1.045, + "step": 12134 + }, + { + "epoch": 0.78, + "learning_rate": 1.7418149810000215e-05, + "loss": 1.0733, + "step": 12135 + }, + { + "epoch": 0.78, + "learning_rate": 1.7417935120977264e-05, + "loss": 1.0178, + "step": 12136 + }, + { + "epoch": 0.78, + "learning_rate": 1.7417720431954313e-05, + "loss": 1.0663, + "step": 12137 + }, + { + "epoch": 0.78, + "learning_rate": 1.7417505742931366e-05, + "loss": 0.9988, + "step": 12138 + }, + { + "epoch": 0.78, + "learning_rate": 1.7417291053908415e-05, + "loss": 1.1499, + "step": 12139 + }, + { + "epoch": 0.78, + "learning_rate": 1.7417076364885467e-05, + "loss": 1.0662, + "step": 12140 + }, + { + "epoch": 0.78, + "learning_rate": 1.7416861675862516e-05, + "loss": 0.9751, + "step": 12141 + }, + { + "epoch": 0.78, + "learning_rate": 1.7416646986839565e-05, + "loss": 1.1754, + "step": 12142 + }, + { + "epoch": 0.78, + "learning_rate": 1.7416432297816614e-05, + "loss": 1.0208, + "step": 12143 + }, + { + "epoch": 0.78, + "learning_rate": 1.7416217608793663e-05, + "loss": 0.9976, + "step": 12144 + }, + { + "epoch": 0.78, + "learning_rate": 1.7416002919770715e-05, + "loss": 1.0315, + "step": 12145 + }, + { + "epoch": 0.78, + "learning_rate": 1.7415788230747764e-05, + "loss": 1.1254, + "step": 12146 + }, + { + "epoch": 0.78, + "learning_rate": 1.7415573541724813e-05, + "loss": 0.9982, + "step": 12147 + }, + { + "epoch": 0.78, + "learning_rate": 1.7415358852701862e-05, + "loss": 1.0345, + "step": 12148 + }, + { + "epoch": 0.78, + "learning_rate": 1.741514416367891e-05, + "loss": 1.1069, + "step": 12149 + }, + { + "epoch": 0.78, + "learning_rate": 1.741492947465596e-05, + "loss": 1.0848, + "step": 12150 + }, + { + "epoch": 0.78, + "learning_rate": 1.7414714785633012e-05, + "loss": 0.9905, + "step": 12151 + }, + { + "epoch": 0.78, + "learning_rate": 1.741450009661006e-05, + "loss": 0.9817, + "step": 12152 + }, + { + "epoch": 0.78, + "learning_rate": 1.741428540758711e-05, + "loss": 1.126, + "step": 12153 + }, + { + "epoch": 0.78, + "learning_rate": 1.7414070718564163e-05, + "loss": 1.0953, + "step": 12154 + }, + { + "epoch": 0.78, + "learning_rate": 1.741385602954121e-05, + "loss": 1.0641, + "step": 12155 + }, + { + "epoch": 0.78, + "learning_rate": 1.741364134051826e-05, + "loss": 1.0843, + "step": 12156 + }, + { + "epoch": 0.78, + "learning_rate": 1.741342665149531e-05, + "loss": 1.0914, + "step": 12157 + }, + { + "epoch": 0.78, + "learning_rate": 1.7413211962472362e-05, + "loss": 1.0736, + "step": 12158 + }, + { + "epoch": 0.78, + "learning_rate": 1.741299727344941e-05, + "loss": 0.9727, + "step": 12159 + }, + { + "epoch": 0.78, + "learning_rate": 1.741278258442646e-05, + "loss": 1.0906, + "step": 12160 + }, + { + "epoch": 0.78, + "learning_rate": 1.741256789540351e-05, + "loss": 1.065, + "step": 12161 + }, + { + "epoch": 0.78, + "learning_rate": 1.7412353206380558e-05, + "loss": 1.0487, + "step": 12162 + }, + { + "epoch": 0.78, + "learning_rate": 1.741213851735761e-05, + "loss": 1.0442, + "step": 12163 + }, + { + "epoch": 0.78, + "learning_rate": 1.741192382833466e-05, + "loss": 0.9968, + "step": 12164 + }, + { + "epoch": 0.78, + "learning_rate": 1.7411709139311708e-05, + "loss": 0.9726, + "step": 12165 + }, + { + "epoch": 0.78, + "learning_rate": 1.7411494450288757e-05, + "loss": 1.0678, + "step": 12166 + }, + { + "epoch": 0.78, + "learning_rate": 1.7411279761265806e-05, + "loss": 1.0899, + "step": 12167 + }, + { + "epoch": 0.78, + "learning_rate": 1.7411065072242858e-05, + "loss": 1.0299, + "step": 12168 + }, + { + "epoch": 0.78, + "learning_rate": 1.7410850383219907e-05, + "loss": 1.0628, + "step": 12169 + }, + { + "epoch": 0.78, + "learning_rate": 1.7410635694196956e-05, + "loss": 1.1165, + "step": 12170 + }, + { + "epoch": 0.78, + "learning_rate": 1.741042100517401e-05, + "loss": 1.0963, + "step": 12171 + }, + { + "epoch": 0.78, + "learning_rate": 1.7410206316151058e-05, + "loss": 1.0471, + "step": 12172 + }, + { + "epoch": 0.78, + "learning_rate": 1.7409991627128107e-05, + "loss": 1.0647, + "step": 12173 + }, + { + "epoch": 0.78, + "learning_rate": 1.7409776938105155e-05, + "loss": 1.0746, + "step": 12174 + }, + { + "epoch": 0.78, + "learning_rate": 1.7409562249082204e-05, + "loss": 1.0665, + "step": 12175 + }, + { + "epoch": 0.78, + "learning_rate": 1.7409347560059257e-05, + "loss": 1.0976, + "step": 12176 + }, + { + "epoch": 0.78, + "learning_rate": 1.7409132871036306e-05, + "loss": 1.0825, + "step": 12177 + }, + { + "epoch": 0.78, + "learning_rate": 1.7408918182013355e-05, + "loss": 1.0844, + "step": 12178 + }, + { + "epoch": 0.78, + "learning_rate": 1.7408703492990404e-05, + "loss": 1.1516, + "step": 12179 + }, + { + "epoch": 0.78, + "learning_rate": 1.7408488803967453e-05, + "loss": 0.8887, + "step": 12180 + }, + { + "epoch": 0.78, + "learning_rate": 1.7408274114944505e-05, + "loss": 1.1086, + "step": 12181 + }, + { + "epoch": 0.78, + "learning_rate": 1.7408059425921554e-05, + "loss": 1.123, + "step": 12182 + }, + { + "epoch": 0.78, + "learning_rate": 1.7407844736898606e-05, + "loss": 1.0093, + "step": 12183 + }, + { + "epoch": 0.78, + "learning_rate": 1.7407630047875655e-05, + "loss": 1.0893, + "step": 12184 + }, + { + "epoch": 0.78, + "learning_rate": 1.7407415358852704e-05, + "loss": 1.0132, + "step": 12185 + }, + { + "epoch": 0.78, + "learning_rate": 1.7407200669829753e-05, + "loss": 1.092, + "step": 12186 + }, + { + "epoch": 0.78, + "learning_rate": 1.7406985980806802e-05, + "loss": 1.0391, + "step": 12187 + }, + { + "epoch": 0.78, + "learning_rate": 1.740677129178385e-05, + "loss": 1.0569, + "step": 12188 + }, + { + "epoch": 0.78, + "learning_rate": 1.7406556602760904e-05, + "loss": 1.0739, + "step": 12189 + }, + { + "epoch": 0.78, + "learning_rate": 1.7406341913737952e-05, + "loss": 1.0447, + "step": 12190 + }, + { + "epoch": 0.78, + "learning_rate": 1.7406127224715e-05, + "loss": 1.0382, + "step": 12191 + }, + { + "epoch": 0.78, + "learning_rate": 1.740591253569205e-05, + "loss": 1.0145, + "step": 12192 + }, + { + "epoch": 0.78, + "learning_rate": 1.74056978466691e-05, + "loss": 1.0502, + "step": 12193 + }, + { + "epoch": 0.78, + "learning_rate": 1.7405483157646152e-05, + "loss": 1.0965, + "step": 12194 + }, + { + "epoch": 0.78, + "learning_rate": 1.74052684686232e-05, + "loss": 1.1062, + "step": 12195 + }, + { + "epoch": 0.78, + "learning_rate": 1.740505377960025e-05, + "loss": 1.0242, + "step": 12196 + }, + { + "epoch": 0.78, + "learning_rate": 1.7404839090577302e-05, + "loss": 1.1332, + "step": 12197 + }, + { + "epoch": 0.78, + "learning_rate": 1.740462440155435e-05, + "loss": 1.0174, + "step": 12198 + }, + { + "epoch": 0.78, + "learning_rate": 1.74044097125314e-05, + "loss": 1.0592, + "step": 12199 + }, + { + "epoch": 0.78, + "learning_rate": 1.740419502350845e-05, + "loss": 1.0459, + "step": 12200 + }, + { + "epoch": 0.78, + "learning_rate": 1.74039803344855e-05, + "loss": 1.0514, + "step": 12201 + }, + { + "epoch": 0.79, + "learning_rate": 1.740376564546255e-05, + "loss": 1.0972, + "step": 12202 + }, + { + "epoch": 0.79, + "learning_rate": 1.74035509564396e-05, + "loss": 1.0251, + "step": 12203 + }, + { + "epoch": 0.79, + "learning_rate": 1.7403336267416648e-05, + "loss": 1.0297, + "step": 12204 + }, + { + "epoch": 0.79, + "learning_rate": 1.7403121578393697e-05, + "loss": 1.133, + "step": 12205 + }, + { + "epoch": 0.79, + "learning_rate": 1.7402906889370746e-05, + "loss": 0.9758, + "step": 12206 + }, + { + "epoch": 0.79, + "learning_rate": 1.74026922003478e-05, + "loss": 1.0341, + "step": 12207 + }, + { + "epoch": 0.79, + "learning_rate": 1.7402477511324847e-05, + "loss": 0.9879, + "step": 12208 + }, + { + "epoch": 0.79, + "learning_rate": 1.7402262822301896e-05, + "loss": 1.1424, + "step": 12209 + }, + { + "epoch": 0.79, + "learning_rate": 1.7402048133278945e-05, + "loss": 1.083, + "step": 12210 + }, + { + "epoch": 0.79, + "learning_rate": 1.7401833444255994e-05, + "loss": 1.1157, + "step": 12211 + }, + { + "epoch": 0.79, + "learning_rate": 1.7401618755233047e-05, + "loss": 1.0595, + "step": 12212 + }, + { + "epoch": 0.79, + "learning_rate": 1.7401404066210096e-05, + "loss": 1.1092, + "step": 12213 + }, + { + "epoch": 0.79, + "learning_rate": 1.7401189377187148e-05, + "loss": 0.9974, + "step": 12214 + }, + { + "epoch": 0.79, + "learning_rate": 1.7400974688164197e-05, + "loss": 1.0424, + "step": 12215 + }, + { + "epoch": 0.79, + "learning_rate": 1.7400759999141246e-05, + "loss": 1.1274, + "step": 12216 + }, + { + "epoch": 0.79, + "learning_rate": 1.7400545310118295e-05, + "loss": 1.0697, + "step": 12217 + }, + { + "epoch": 0.79, + "learning_rate": 1.7400330621095344e-05, + "loss": 1.1289, + "step": 12218 + }, + { + "epoch": 0.79, + "learning_rate": 1.7400115932072393e-05, + "loss": 1.1433, + "step": 12219 + }, + { + "epoch": 0.79, + "learning_rate": 1.7399901243049445e-05, + "loss": 1.1096, + "step": 12220 + }, + { + "epoch": 0.79, + "learning_rate": 1.7399686554026494e-05, + "loss": 1.1142, + "step": 12221 + }, + { + "epoch": 0.79, + "learning_rate": 1.7399471865003543e-05, + "loss": 1.0857, + "step": 12222 + }, + { + "epoch": 0.79, + "learning_rate": 1.7399257175980592e-05, + "loss": 1.096, + "step": 12223 + }, + { + "epoch": 0.79, + "learning_rate": 1.739904248695764e-05, + "loss": 1.0623, + "step": 12224 + }, + { + "epoch": 0.79, + "learning_rate": 1.7398827797934693e-05, + "loss": 1.0398, + "step": 12225 + }, + { + "epoch": 0.79, + "learning_rate": 1.7398613108911742e-05, + "loss": 1.0727, + "step": 12226 + }, + { + "epoch": 0.79, + "learning_rate": 1.7398398419888795e-05, + "loss": 1.0586, + "step": 12227 + }, + { + "epoch": 0.79, + "learning_rate": 1.7398183730865844e-05, + "loss": 1.0673, + "step": 12228 + }, + { + "epoch": 0.79, + "learning_rate": 1.7397969041842893e-05, + "loss": 1.1004, + "step": 12229 + }, + { + "epoch": 0.79, + "learning_rate": 1.739775435281994e-05, + "loss": 1.0769, + "step": 12230 + }, + { + "epoch": 0.79, + "learning_rate": 1.739753966379699e-05, + "loss": 1.0665, + "step": 12231 + }, + { + "epoch": 0.79, + "learning_rate": 1.7397324974774043e-05, + "loss": 1.0739, + "step": 12232 + }, + { + "epoch": 0.79, + "learning_rate": 1.7397110285751092e-05, + "loss": 1.0518, + "step": 12233 + }, + { + "epoch": 0.79, + "learning_rate": 1.739689559672814e-05, + "loss": 1.0378, + "step": 12234 + }, + { + "epoch": 0.79, + "learning_rate": 1.739668090770519e-05, + "loss": 1.0628, + "step": 12235 + }, + { + "epoch": 0.79, + "learning_rate": 1.739646621868224e-05, + "loss": 1.0911, + "step": 12236 + }, + { + "epoch": 0.79, + "learning_rate": 1.7396251529659288e-05, + "loss": 1.1447, + "step": 12237 + }, + { + "epoch": 0.79, + "learning_rate": 1.739603684063634e-05, + "loss": 1.1218, + "step": 12238 + }, + { + "epoch": 0.79, + "learning_rate": 1.739582215161339e-05, + "loss": 1.0308, + "step": 12239 + }, + { + "epoch": 0.79, + "learning_rate": 1.7395607462590438e-05, + "loss": 1.0787, + "step": 12240 + }, + { + "epoch": 0.79, + "learning_rate": 1.739539277356749e-05, + "loss": 1.0288, + "step": 12241 + }, + { + "epoch": 0.79, + "learning_rate": 1.739517808454454e-05, + "loss": 1.1004, + "step": 12242 + }, + { + "epoch": 0.79, + "learning_rate": 1.7394963395521588e-05, + "loss": 1.0614, + "step": 12243 + }, + { + "epoch": 0.79, + "learning_rate": 1.7394748706498637e-05, + "loss": 1.0619, + "step": 12244 + }, + { + "epoch": 0.79, + "learning_rate": 1.739453401747569e-05, + "loss": 1.0463, + "step": 12245 + }, + { + "epoch": 0.79, + "learning_rate": 1.739431932845274e-05, + "loss": 1.0522, + "step": 12246 + }, + { + "epoch": 0.79, + "learning_rate": 1.7394104639429788e-05, + "loss": 1.0538, + "step": 12247 + }, + { + "epoch": 0.79, + "learning_rate": 1.7393889950406836e-05, + "loss": 0.9983, + "step": 12248 + }, + { + "epoch": 0.79, + "learning_rate": 1.7393675261383885e-05, + "loss": 1.077, + "step": 12249 + }, + { + "epoch": 0.79, + "learning_rate": 1.7393460572360938e-05, + "loss": 1.1136, + "step": 12250 + }, + { + "epoch": 0.79, + "learning_rate": 1.7393245883337987e-05, + "loss": 1.0584, + "step": 12251 + }, + { + "epoch": 0.79, + "learning_rate": 1.7393031194315036e-05, + "loss": 1.0668, + "step": 12252 + }, + { + "epoch": 0.79, + "learning_rate": 1.7392816505292085e-05, + "loss": 1.0403, + "step": 12253 + }, + { + "epoch": 0.79, + "learning_rate": 1.7392601816269134e-05, + "loss": 1.0707, + "step": 12254 + }, + { + "epoch": 0.79, + "learning_rate": 1.7392387127246186e-05, + "loss": 1.0273, + "step": 12255 + }, + { + "epoch": 0.79, + "learning_rate": 1.7392172438223235e-05, + "loss": 1.0031, + "step": 12256 + }, + { + "epoch": 0.79, + "learning_rate": 1.7391957749200284e-05, + "loss": 1.0076, + "step": 12257 + }, + { + "epoch": 0.79, + "learning_rate": 1.7391743060177336e-05, + "loss": 1.0437, + "step": 12258 + }, + { + "epoch": 0.79, + "learning_rate": 1.7391528371154385e-05, + "loss": 1.0382, + "step": 12259 + }, + { + "epoch": 0.79, + "learning_rate": 1.7391313682131434e-05, + "loss": 1.0912, + "step": 12260 + }, + { + "epoch": 0.79, + "learning_rate": 1.7391098993108483e-05, + "loss": 0.9978, + "step": 12261 + }, + { + "epoch": 0.79, + "learning_rate": 1.7390884304085532e-05, + "loss": 0.9867, + "step": 12262 + }, + { + "epoch": 0.79, + "learning_rate": 1.7390669615062585e-05, + "loss": 1.065, + "step": 12263 + }, + { + "epoch": 0.79, + "learning_rate": 1.7390454926039633e-05, + "loss": 1.1369, + "step": 12264 + }, + { + "epoch": 0.79, + "learning_rate": 1.7390240237016682e-05, + "loss": 1.0532, + "step": 12265 + }, + { + "epoch": 0.79, + "learning_rate": 1.739002554799373e-05, + "loss": 1.0329, + "step": 12266 + }, + { + "epoch": 0.79, + "learning_rate": 1.738981085897078e-05, + "loss": 1.1119, + "step": 12267 + }, + { + "epoch": 0.79, + "learning_rate": 1.7389596169947833e-05, + "loss": 1.0195, + "step": 12268 + }, + { + "epoch": 0.79, + "learning_rate": 1.738938148092488e-05, + "loss": 1.0219, + "step": 12269 + }, + { + "epoch": 0.79, + "learning_rate": 1.7389166791901934e-05, + "loss": 1.0151, + "step": 12270 + }, + { + "epoch": 0.79, + "learning_rate": 1.7388952102878983e-05, + "loss": 1.1367, + "step": 12271 + }, + { + "epoch": 0.79, + "learning_rate": 1.7388737413856032e-05, + "loss": 1.0669, + "step": 12272 + }, + { + "epoch": 0.79, + "learning_rate": 1.738852272483308e-05, + "loss": 0.9489, + "step": 12273 + }, + { + "epoch": 0.79, + "learning_rate": 1.738830803581013e-05, + "loss": 1.0451, + "step": 12274 + }, + { + "epoch": 0.79, + "learning_rate": 1.738809334678718e-05, + "loss": 1.0421, + "step": 12275 + }, + { + "epoch": 0.79, + "learning_rate": 1.738787865776423e-05, + "loss": 1.1332, + "step": 12276 + }, + { + "epoch": 0.79, + "learning_rate": 1.738766396874128e-05, + "loss": 0.9776, + "step": 12277 + }, + { + "epoch": 0.79, + "learning_rate": 1.738744927971833e-05, + "loss": 1.0203, + "step": 12278 + }, + { + "epoch": 0.79, + "learning_rate": 1.7387234590695378e-05, + "loss": 1.1375, + "step": 12279 + }, + { + "epoch": 0.79, + "learning_rate": 1.7387019901672427e-05, + "loss": 1.0186, + "step": 12280 + }, + { + "epoch": 0.79, + "learning_rate": 1.738680521264948e-05, + "loss": 1.0467, + "step": 12281 + }, + { + "epoch": 0.79, + "learning_rate": 1.738659052362653e-05, + "loss": 1.1552, + "step": 12282 + }, + { + "epoch": 0.79, + "learning_rate": 1.7386375834603577e-05, + "loss": 1.0983, + "step": 12283 + }, + { + "epoch": 0.79, + "learning_rate": 1.738616114558063e-05, + "loss": 1.0696, + "step": 12284 + }, + { + "epoch": 0.79, + "learning_rate": 1.738594645655768e-05, + "loss": 1.1519, + "step": 12285 + }, + { + "epoch": 0.79, + "learning_rate": 1.7385731767534728e-05, + "loss": 1.0118, + "step": 12286 + }, + { + "epoch": 0.79, + "learning_rate": 1.7385517078511777e-05, + "loss": 1.0846, + "step": 12287 + }, + { + "epoch": 0.79, + "learning_rate": 1.738530238948883e-05, + "loss": 1.1291, + "step": 12288 + }, + { + "epoch": 0.79, + "learning_rate": 1.7385087700465878e-05, + "loss": 0.9973, + "step": 12289 + }, + { + "epoch": 0.79, + "learning_rate": 1.7384873011442927e-05, + "loss": 1.11, + "step": 12290 + }, + { + "epoch": 0.79, + "learning_rate": 1.7384658322419976e-05, + "loss": 1.0797, + "step": 12291 + }, + { + "epoch": 0.79, + "learning_rate": 1.7384443633397025e-05, + "loss": 0.9764, + "step": 12292 + }, + { + "epoch": 0.79, + "learning_rate": 1.7384228944374074e-05, + "loss": 1.0304, + "step": 12293 + }, + { + "epoch": 0.79, + "learning_rate": 1.7384014255351126e-05, + "loss": 1.0267, + "step": 12294 + }, + { + "epoch": 0.79, + "learning_rate": 1.7383799566328175e-05, + "loss": 1.1185, + "step": 12295 + }, + { + "epoch": 0.79, + "learning_rate": 1.7383584877305224e-05, + "loss": 1.0215, + "step": 12296 + }, + { + "epoch": 0.79, + "learning_rate": 1.7383370188282273e-05, + "loss": 1.0868, + "step": 12297 + }, + { + "epoch": 0.79, + "learning_rate": 1.7383155499259322e-05, + "loss": 1.0972, + "step": 12298 + }, + { + "epoch": 0.79, + "learning_rate": 1.7382940810236374e-05, + "loss": 0.9863, + "step": 12299 + }, + { + "epoch": 0.79, + "learning_rate": 1.7382726121213423e-05, + "loss": 0.9975, + "step": 12300 + }, + { + "epoch": 0.79, + "learning_rate": 1.7382511432190476e-05, + "loss": 1.0317, + "step": 12301 + }, + { + "epoch": 0.79, + "learning_rate": 1.7382296743167525e-05, + "loss": 1.0012, + "step": 12302 + }, + { + "epoch": 0.79, + "learning_rate": 1.7382082054144574e-05, + "loss": 1.1097, + "step": 12303 + }, + { + "epoch": 0.79, + "learning_rate": 1.7381867365121623e-05, + "loss": 1.0084, + "step": 12304 + }, + { + "epoch": 0.79, + "learning_rate": 1.738165267609867e-05, + "loss": 1.0597, + "step": 12305 + }, + { + "epoch": 0.79, + "learning_rate": 1.738143798707572e-05, + "loss": 1.0837, + "step": 12306 + }, + { + "epoch": 0.79, + "learning_rate": 1.7381223298052773e-05, + "loss": 1.0722, + "step": 12307 + }, + { + "epoch": 0.79, + "learning_rate": 1.7381008609029822e-05, + "loss": 1.0845, + "step": 12308 + }, + { + "epoch": 0.79, + "learning_rate": 1.738079392000687e-05, + "loss": 1.1063, + "step": 12309 + }, + { + "epoch": 0.79, + "learning_rate": 1.738057923098392e-05, + "loss": 1.022, + "step": 12310 + }, + { + "epoch": 0.79, + "learning_rate": 1.738036454196097e-05, + "loss": 1.0556, + "step": 12311 + }, + { + "epoch": 0.79, + "learning_rate": 1.738014985293802e-05, + "loss": 1.0181, + "step": 12312 + }, + { + "epoch": 0.79, + "learning_rate": 1.737993516391507e-05, + "loss": 1.0115, + "step": 12313 + }, + { + "epoch": 0.79, + "learning_rate": 1.7379720474892122e-05, + "loss": 1.0464, + "step": 12314 + }, + { + "epoch": 0.79, + "learning_rate": 1.737950578586917e-05, + "loss": 1.0554, + "step": 12315 + }, + { + "epoch": 0.79, + "learning_rate": 1.737929109684622e-05, + "loss": 1.0863, + "step": 12316 + }, + { + "epoch": 0.79, + "learning_rate": 1.737907640782327e-05, + "loss": 1.0513, + "step": 12317 + }, + { + "epoch": 0.79, + "learning_rate": 1.7378861718800318e-05, + "loss": 1.0328, + "step": 12318 + }, + { + "epoch": 0.79, + "learning_rate": 1.737864702977737e-05, + "loss": 1.1087, + "step": 12319 + }, + { + "epoch": 0.79, + "learning_rate": 1.737843234075442e-05, + "loss": 1.0179, + "step": 12320 + }, + { + "epoch": 0.79, + "learning_rate": 1.737821765173147e-05, + "loss": 1.0805, + "step": 12321 + }, + { + "epoch": 0.79, + "learning_rate": 1.7378002962708517e-05, + "loss": 1.0421, + "step": 12322 + }, + { + "epoch": 0.79, + "learning_rate": 1.7377788273685566e-05, + "loss": 1.0532, + "step": 12323 + }, + { + "epoch": 0.79, + "learning_rate": 1.7377573584662615e-05, + "loss": 0.9348, + "step": 12324 + }, + { + "epoch": 0.79, + "learning_rate": 1.7377358895639668e-05, + "loss": 1.0788, + "step": 12325 + }, + { + "epoch": 0.79, + "learning_rate": 1.7377144206616717e-05, + "loss": 1.0634, + "step": 12326 + }, + { + "epoch": 0.79, + "learning_rate": 1.7376929517593766e-05, + "loss": 1.102, + "step": 12327 + }, + { + "epoch": 0.79, + "learning_rate": 1.7376714828570818e-05, + "loss": 1.0567, + "step": 12328 + }, + { + "epoch": 0.79, + "learning_rate": 1.7376500139547867e-05, + "loss": 1.1134, + "step": 12329 + }, + { + "epoch": 0.79, + "learning_rate": 1.7376285450524916e-05, + "loss": 0.9992, + "step": 12330 + }, + { + "epoch": 0.79, + "learning_rate": 1.7376070761501965e-05, + "loss": 1.0953, + "step": 12331 + }, + { + "epoch": 0.79, + "learning_rate": 1.7375856072479017e-05, + "loss": 1.046, + "step": 12332 + }, + { + "epoch": 0.79, + "learning_rate": 1.7375641383456066e-05, + "loss": 1.0192, + "step": 12333 + }, + { + "epoch": 0.79, + "learning_rate": 1.7375426694433115e-05, + "loss": 1.0703, + "step": 12334 + }, + { + "epoch": 0.79, + "learning_rate": 1.7375212005410164e-05, + "loss": 1.0489, + "step": 12335 + }, + { + "epoch": 0.79, + "learning_rate": 1.7374997316387213e-05, + "loss": 1.1592, + "step": 12336 + }, + { + "epoch": 0.79, + "learning_rate": 1.7374782627364266e-05, + "loss": 1.063, + "step": 12337 + }, + { + "epoch": 0.79, + "learning_rate": 1.7374567938341314e-05, + "loss": 1.0552, + "step": 12338 + }, + { + "epoch": 0.79, + "learning_rate": 1.7374353249318363e-05, + "loss": 1.0123, + "step": 12339 + }, + { + "epoch": 0.79, + "learning_rate": 1.7374138560295412e-05, + "loss": 1.0494, + "step": 12340 + }, + { + "epoch": 0.79, + "learning_rate": 1.737392387127246e-05, + "loss": 1.0626, + "step": 12341 + }, + { + "epoch": 0.79, + "learning_rate": 1.7373709182249514e-05, + "loss": 1.0457, + "step": 12342 + }, + { + "epoch": 0.79, + "learning_rate": 1.7373494493226563e-05, + "loss": 1.0878, + "step": 12343 + }, + { + "epoch": 0.79, + "learning_rate": 1.737327980420361e-05, + "loss": 1.1309, + "step": 12344 + }, + { + "epoch": 0.79, + "learning_rate": 1.7373065115180664e-05, + "loss": 1.1238, + "step": 12345 + }, + { + "epoch": 0.79, + "learning_rate": 1.7372850426157713e-05, + "loss": 0.9976, + "step": 12346 + }, + { + "epoch": 0.79, + "learning_rate": 1.7372635737134762e-05, + "loss": 1.1202, + "step": 12347 + }, + { + "epoch": 0.79, + "learning_rate": 1.737242104811181e-05, + "loss": 1.0859, + "step": 12348 + }, + { + "epoch": 0.79, + "learning_rate": 1.737220635908886e-05, + "loss": 1.1218, + "step": 12349 + }, + { + "epoch": 0.79, + "learning_rate": 1.7371991670065912e-05, + "loss": 1.1458, + "step": 12350 + }, + { + "epoch": 0.79, + "learning_rate": 1.737177698104296e-05, + "loss": 1.1076, + "step": 12351 + }, + { + "epoch": 0.79, + "learning_rate": 1.737156229202001e-05, + "loss": 1.0816, + "step": 12352 + }, + { + "epoch": 0.79, + "learning_rate": 1.737134760299706e-05, + "loss": 1.0582, + "step": 12353 + }, + { + "epoch": 0.79, + "learning_rate": 1.7371132913974108e-05, + "loss": 1.0569, + "step": 12354 + }, + { + "epoch": 0.79, + "learning_rate": 1.737091822495116e-05, + "loss": 1.096, + "step": 12355 + }, + { + "epoch": 0.79, + "learning_rate": 1.737070353592821e-05, + "loss": 1.1393, + "step": 12356 + }, + { + "epoch": 0.8, + "learning_rate": 1.7370488846905262e-05, + "loss": 1.0829, + "step": 12357 + }, + { + "epoch": 0.8, + "learning_rate": 1.737027415788231e-05, + "loss": 1.0116, + "step": 12358 + }, + { + "epoch": 0.8, + "learning_rate": 1.737005946885936e-05, + "loss": 1.0391, + "step": 12359 + }, + { + "epoch": 0.8, + "learning_rate": 1.736984477983641e-05, + "loss": 1.085, + "step": 12360 + }, + { + "epoch": 0.8, + "learning_rate": 1.7369630090813458e-05, + "loss": 1.092, + "step": 12361 + }, + { + "epoch": 0.8, + "learning_rate": 1.7369415401790507e-05, + "loss": 1.1001, + "step": 12362 + }, + { + "epoch": 0.8, + "learning_rate": 1.736920071276756e-05, + "loss": 1.1484, + "step": 12363 + }, + { + "epoch": 0.8, + "learning_rate": 1.7368986023744608e-05, + "loss": 1.0859, + "step": 12364 + }, + { + "epoch": 0.8, + "learning_rate": 1.7368771334721657e-05, + "loss": 1.1396, + "step": 12365 + }, + { + "epoch": 0.8, + "learning_rate": 1.7368556645698706e-05, + "loss": 1.0813, + "step": 12366 + }, + { + "epoch": 0.8, + "learning_rate": 1.7368341956675755e-05, + "loss": 1.0681, + "step": 12367 + }, + { + "epoch": 0.8, + "learning_rate": 1.7368127267652807e-05, + "loss": 1.0474, + "step": 12368 + }, + { + "epoch": 0.8, + "learning_rate": 1.7367912578629856e-05, + "loss": 1.007, + "step": 12369 + }, + { + "epoch": 0.8, + "learning_rate": 1.7367697889606905e-05, + "loss": 0.9971, + "step": 12370 + }, + { + "epoch": 0.8, + "learning_rate": 1.7367483200583957e-05, + "loss": 1.065, + "step": 12371 + }, + { + "epoch": 0.8, + "learning_rate": 1.7367268511561006e-05, + "loss": 1.0789, + "step": 12372 + }, + { + "epoch": 0.8, + "learning_rate": 1.7367053822538055e-05, + "loss": 1.0792, + "step": 12373 + }, + { + "epoch": 0.8, + "learning_rate": 1.7366839133515104e-05, + "loss": 1.0551, + "step": 12374 + }, + { + "epoch": 0.8, + "learning_rate": 1.7366624444492157e-05, + "loss": 1.0868, + "step": 12375 + }, + { + "epoch": 0.8, + "learning_rate": 1.7366409755469206e-05, + "loss": 1.0413, + "step": 12376 + }, + { + "epoch": 0.8, + "learning_rate": 1.7366195066446255e-05, + "loss": 1.1346, + "step": 12377 + }, + { + "epoch": 0.8, + "learning_rate": 1.7365980377423304e-05, + "loss": 1.0363, + "step": 12378 + }, + { + "epoch": 0.8, + "learning_rate": 1.7365765688400353e-05, + "loss": 1.0946, + "step": 12379 + }, + { + "epoch": 0.8, + "learning_rate": 1.73655509993774e-05, + "loss": 1.0514, + "step": 12380 + }, + { + "epoch": 0.8, + "learning_rate": 1.7365336310354454e-05, + "loss": 1.1107, + "step": 12381 + }, + { + "epoch": 0.8, + "learning_rate": 1.7365121621331503e-05, + "loss": 1.1056, + "step": 12382 + }, + { + "epoch": 0.8, + "learning_rate": 1.7364906932308552e-05, + "loss": 0.9918, + "step": 12383 + }, + { + "epoch": 0.8, + "learning_rate": 1.73646922432856e-05, + "loss": 1.0812, + "step": 12384 + }, + { + "epoch": 0.8, + "learning_rate": 1.736447755426265e-05, + "loss": 0.9915, + "step": 12385 + }, + { + "epoch": 0.8, + "learning_rate": 1.7364262865239702e-05, + "loss": 1.0452, + "step": 12386 + }, + { + "epoch": 0.8, + "learning_rate": 1.736404817621675e-05, + "loss": 0.9838, + "step": 12387 + }, + { + "epoch": 0.8, + "learning_rate": 1.7363833487193803e-05, + "loss": 1.0488, + "step": 12388 + }, + { + "epoch": 0.8, + "learning_rate": 1.7363618798170852e-05, + "loss": 1.0502, + "step": 12389 + }, + { + "epoch": 0.8, + "learning_rate": 1.73634041091479e-05, + "loss": 1.0153, + "step": 12390 + }, + { + "epoch": 0.8, + "learning_rate": 1.736318942012495e-05, + "loss": 1.0292, + "step": 12391 + }, + { + "epoch": 0.8, + "learning_rate": 1.7362974731102e-05, + "loss": 1.0712, + "step": 12392 + }, + { + "epoch": 0.8, + "learning_rate": 1.7362760042079048e-05, + "loss": 1.0735, + "step": 12393 + }, + { + "epoch": 0.8, + "learning_rate": 1.73625453530561e-05, + "loss": 1.0518, + "step": 12394 + }, + { + "epoch": 0.8, + "learning_rate": 1.736233066403315e-05, + "loss": 1.1029, + "step": 12395 + }, + { + "epoch": 0.8, + "learning_rate": 1.73621159750102e-05, + "loss": 1.1189, + "step": 12396 + }, + { + "epoch": 0.8, + "learning_rate": 1.7361901285987247e-05, + "loss": 1.0259, + "step": 12397 + }, + { + "epoch": 0.8, + "learning_rate": 1.7361686596964296e-05, + "loss": 1.0807, + "step": 12398 + }, + { + "epoch": 0.8, + "learning_rate": 1.736147190794135e-05, + "loss": 1.0642, + "step": 12399 + }, + { + "epoch": 0.8, + "learning_rate": 1.7361257218918398e-05, + "loss": 1.0546, + "step": 12400 + }, + { + "epoch": 0.8, + "learning_rate": 1.736104252989545e-05, + "loss": 1.0033, + "step": 12401 + }, + { + "epoch": 0.8, + "learning_rate": 1.73608278408725e-05, + "loss": 1.0208, + "step": 12402 + }, + { + "epoch": 0.8, + "learning_rate": 1.7360613151849548e-05, + "loss": 1.0324, + "step": 12403 + }, + { + "epoch": 0.8, + "learning_rate": 1.7360398462826597e-05, + "loss": 1.0237, + "step": 12404 + }, + { + "epoch": 0.8, + "learning_rate": 1.7360183773803646e-05, + "loss": 1.0902, + "step": 12405 + }, + { + "epoch": 0.8, + "learning_rate": 1.7359969084780698e-05, + "loss": 1.0631, + "step": 12406 + }, + { + "epoch": 0.8, + "learning_rate": 1.7359754395757747e-05, + "loss": 1.0526, + "step": 12407 + }, + { + "epoch": 0.8, + "learning_rate": 1.7359539706734796e-05, + "loss": 1.0196, + "step": 12408 + }, + { + "epoch": 0.8, + "learning_rate": 1.7359325017711845e-05, + "loss": 1.0666, + "step": 12409 + }, + { + "epoch": 0.8, + "learning_rate": 1.7359110328688894e-05, + "loss": 1.0304, + "step": 12410 + }, + { + "epoch": 0.8, + "learning_rate": 1.7358895639665943e-05, + "loss": 1.1086, + "step": 12411 + }, + { + "epoch": 0.8, + "learning_rate": 1.7358680950642995e-05, + "loss": 0.99, + "step": 12412 + }, + { + "epoch": 0.8, + "learning_rate": 1.7358466261620044e-05, + "loss": 1.1163, + "step": 12413 + }, + { + "epoch": 0.8, + "learning_rate": 1.7358251572597093e-05, + "loss": 1.0469, + "step": 12414 + }, + { + "epoch": 0.8, + "learning_rate": 1.7358036883574146e-05, + "loss": 1.091, + "step": 12415 + }, + { + "epoch": 0.8, + "learning_rate": 1.7357822194551195e-05, + "loss": 1.0055, + "step": 12416 + }, + { + "epoch": 0.8, + "learning_rate": 1.7357607505528244e-05, + "loss": 1.1035, + "step": 12417 + }, + { + "epoch": 0.8, + "learning_rate": 1.7357392816505293e-05, + "loss": 1.0196, + "step": 12418 + }, + { + "epoch": 0.8, + "learning_rate": 1.7357178127482345e-05, + "loss": 1.1131, + "step": 12419 + }, + { + "epoch": 0.8, + "learning_rate": 1.7356963438459394e-05, + "loss": 1.0704, + "step": 12420 + }, + { + "epoch": 0.8, + "learning_rate": 1.7356748749436443e-05, + "loss": 1.0074, + "step": 12421 + }, + { + "epoch": 0.8, + "learning_rate": 1.7356534060413492e-05, + "loss": 1.0184, + "step": 12422 + }, + { + "epoch": 0.8, + "learning_rate": 1.735631937139054e-05, + "loss": 1.0795, + "step": 12423 + }, + { + "epoch": 0.8, + "learning_rate": 1.7356104682367593e-05, + "loss": 1.0271, + "step": 12424 + }, + { + "epoch": 0.8, + "learning_rate": 1.7355889993344642e-05, + "loss": 1.1739, + "step": 12425 + }, + { + "epoch": 0.8, + "learning_rate": 1.735567530432169e-05, + "loss": 1.0722, + "step": 12426 + }, + { + "epoch": 0.8, + "learning_rate": 1.735546061529874e-05, + "loss": 0.9448, + "step": 12427 + }, + { + "epoch": 0.8, + "learning_rate": 1.735524592627579e-05, + "loss": 1.0171, + "step": 12428 + }, + { + "epoch": 0.8, + "learning_rate": 1.735503123725284e-05, + "loss": 1.0936, + "step": 12429 + }, + { + "epoch": 0.8, + "learning_rate": 1.735481654822989e-05, + "loss": 1.1042, + "step": 12430 + }, + { + "epoch": 0.8, + "learning_rate": 1.735460185920694e-05, + "loss": 1.0577, + "step": 12431 + }, + { + "epoch": 0.8, + "learning_rate": 1.7354387170183992e-05, + "loss": 1.0683, + "step": 12432 + }, + { + "epoch": 0.8, + "learning_rate": 1.735417248116104e-05, + "loss": 1.0561, + "step": 12433 + }, + { + "epoch": 0.8, + "learning_rate": 1.735395779213809e-05, + "loss": 1.0367, + "step": 12434 + }, + { + "epoch": 0.8, + "learning_rate": 1.735374310311514e-05, + "loss": 1.1609, + "step": 12435 + }, + { + "epoch": 0.8, + "learning_rate": 1.7353528414092188e-05, + "loss": 1.086, + "step": 12436 + }, + { + "epoch": 0.8, + "learning_rate": 1.735331372506924e-05, + "loss": 1.0748, + "step": 12437 + }, + { + "epoch": 0.8, + "learning_rate": 1.735309903604629e-05, + "loss": 1.1319, + "step": 12438 + }, + { + "epoch": 0.8, + "learning_rate": 1.7352884347023338e-05, + "loss": 1.069, + "step": 12439 + }, + { + "epoch": 0.8, + "learning_rate": 1.7352669658000387e-05, + "loss": 1.1047, + "step": 12440 + }, + { + "epoch": 0.8, + "learning_rate": 1.7352454968977436e-05, + "loss": 1.067, + "step": 12441 + }, + { + "epoch": 0.8, + "learning_rate": 1.7352240279954488e-05, + "loss": 1.0809, + "step": 12442 + }, + { + "epoch": 0.8, + "learning_rate": 1.7352025590931537e-05, + "loss": 1.0327, + "step": 12443 + }, + { + "epoch": 0.8, + "learning_rate": 1.735181090190859e-05, + "loss": 1.0598, + "step": 12444 + }, + { + "epoch": 0.8, + "learning_rate": 1.735159621288564e-05, + "loss": 1.0126, + "step": 12445 + }, + { + "epoch": 0.8, + "learning_rate": 1.7351381523862687e-05, + "loss": 1.0393, + "step": 12446 + }, + { + "epoch": 0.8, + "learning_rate": 1.7351166834839736e-05, + "loss": 0.9984, + "step": 12447 + }, + { + "epoch": 0.8, + "learning_rate": 1.7350952145816785e-05, + "loss": 1.0612, + "step": 12448 + }, + { + "epoch": 0.8, + "learning_rate": 1.7350737456793834e-05, + "loss": 1.0061, + "step": 12449 + }, + { + "epoch": 0.8, + "learning_rate": 1.7350522767770887e-05, + "loss": 1.0561, + "step": 12450 + }, + { + "epoch": 0.8, + "learning_rate": 1.7350308078747936e-05, + "loss": 1.0794, + "step": 12451 + }, + { + "epoch": 0.8, + "learning_rate": 1.7350093389724985e-05, + "loss": 1.014, + "step": 12452 + }, + { + "epoch": 0.8, + "learning_rate": 1.7349878700702034e-05, + "loss": 1.0033, + "step": 12453 + }, + { + "epoch": 0.8, + "learning_rate": 1.7349664011679082e-05, + "loss": 1.0358, + "step": 12454 + }, + { + "epoch": 0.8, + "learning_rate": 1.7349449322656135e-05, + "loss": 1.1032, + "step": 12455 + }, + { + "epoch": 0.8, + "learning_rate": 1.7349234633633184e-05, + "loss": 1.0102, + "step": 12456 + }, + { + "epoch": 0.8, + "learning_rate": 1.7349019944610233e-05, + "loss": 1.0851, + "step": 12457 + }, + { + "epoch": 0.8, + "learning_rate": 1.7348805255587285e-05, + "loss": 1.128, + "step": 12458 + }, + { + "epoch": 0.8, + "learning_rate": 1.7348590566564334e-05, + "loss": 0.9525, + "step": 12459 + }, + { + "epoch": 0.8, + "learning_rate": 1.7348375877541383e-05, + "loss": 1.1, + "step": 12460 + }, + { + "epoch": 0.8, + "learning_rate": 1.7348161188518432e-05, + "loss": 1.1109, + "step": 12461 + }, + { + "epoch": 0.8, + "learning_rate": 1.7347946499495484e-05, + "loss": 1.0397, + "step": 12462 + }, + { + "epoch": 0.8, + "learning_rate": 1.7347731810472533e-05, + "loss": 1.0164, + "step": 12463 + }, + { + "epoch": 0.8, + "learning_rate": 1.7347517121449582e-05, + "loss": 1.0899, + "step": 12464 + }, + { + "epoch": 0.8, + "learning_rate": 1.734730243242663e-05, + "loss": 0.9647, + "step": 12465 + }, + { + "epoch": 0.8, + "learning_rate": 1.734708774340368e-05, + "loss": 1.0487, + "step": 12466 + }, + { + "epoch": 0.8, + "learning_rate": 1.734687305438073e-05, + "loss": 1.1426, + "step": 12467 + }, + { + "epoch": 0.8, + "learning_rate": 1.734665836535778e-05, + "loss": 1.1155, + "step": 12468 + }, + { + "epoch": 0.8, + "learning_rate": 1.734644367633483e-05, + "loss": 1.1015, + "step": 12469 + }, + { + "epoch": 0.8, + "learning_rate": 1.734622898731188e-05, + "loss": 1.1003, + "step": 12470 + }, + { + "epoch": 0.8, + "learning_rate": 1.734601429828893e-05, + "loss": 1.0281, + "step": 12471 + }, + { + "epoch": 0.8, + "learning_rate": 1.7345799609265977e-05, + "loss": 0.991, + "step": 12472 + }, + { + "epoch": 0.8, + "learning_rate": 1.734558492024303e-05, + "loss": 1.0704, + "step": 12473 + }, + { + "epoch": 0.8, + "learning_rate": 1.734537023122008e-05, + "loss": 1.1231, + "step": 12474 + }, + { + "epoch": 0.8, + "learning_rate": 1.734515554219713e-05, + "loss": 1.065, + "step": 12475 + }, + { + "epoch": 0.8, + "learning_rate": 1.734494085317418e-05, + "loss": 1.0979, + "step": 12476 + }, + { + "epoch": 0.8, + "learning_rate": 1.734472616415123e-05, + "loss": 1.0843, + "step": 12477 + }, + { + "epoch": 0.8, + "learning_rate": 1.7344511475128278e-05, + "loss": 1.0909, + "step": 12478 + }, + { + "epoch": 0.8, + "learning_rate": 1.7344296786105327e-05, + "loss": 1.1274, + "step": 12479 + }, + { + "epoch": 0.8, + "learning_rate": 1.7344082097082376e-05, + "loss": 1.0867, + "step": 12480 + }, + { + "epoch": 0.8, + "learning_rate": 1.7343867408059428e-05, + "loss": 1.0801, + "step": 12481 + }, + { + "epoch": 0.8, + "learning_rate": 1.7343652719036477e-05, + "loss": 1.0737, + "step": 12482 + }, + { + "epoch": 0.8, + "learning_rate": 1.7343438030013526e-05, + "loss": 1.0247, + "step": 12483 + }, + { + "epoch": 0.8, + "learning_rate": 1.7343223340990575e-05, + "loss": 0.9822, + "step": 12484 + }, + { + "epoch": 0.8, + "learning_rate": 1.7343008651967624e-05, + "loss": 0.9961, + "step": 12485 + }, + { + "epoch": 0.8, + "learning_rate": 1.7342793962944676e-05, + "loss": 1.0144, + "step": 12486 + }, + { + "epoch": 0.8, + "learning_rate": 1.7342579273921725e-05, + "loss": 1.0321, + "step": 12487 + }, + { + "epoch": 0.8, + "learning_rate": 1.7342364584898778e-05, + "loss": 1.1384, + "step": 12488 + }, + { + "epoch": 0.8, + "learning_rate": 1.7342149895875827e-05, + "loss": 1.0508, + "step": 12489 + }, + { + "epoch": 0.8, + "learning_rate": 1.7341935206852876e-05, + "loss": 1.1957, + "step": 12490 + }, + { + "epoch": 0.8, + "learning_rate": 1.7341720517829925e-05, + "loss": 0.9932, + "step": 12491 + }, + { + "epoch": 0.8, + "learning_rate": 1.7341505828806974e-05, + "loss": 1.0854, + "step": 12492 + }, + { + "epoch": 0.8, + "learning_rate": 1.7341291139784026e-05, + "loss": 1.0733, + "step": 12493 + }, + { + "epoch": 0.8, + "learning_rate": 1.7341076450761075e-05, + "loss": 1.062, + "step": 12494 + }, + { + "epoch": 0.8, + "learning_rate": 1.7340861761738124e-05, + "loss": 1.0276, + "step": 12495 + }, + { + "epoch": 0.8, + "learning_rate": 1.7340647072715173e-05, + "loss": 1.1217, + "step": 12496 + }, + { + "epoch": 0.8, + "learning_rate": 1.7340432383692222e-05, + "loss": 1.0602, + "step": 12497 + }, + { + "epoch": 0.8, + "learning_rate": 1.734021769466927e-05, + "loss": 1.0777, + "step": 12498 + }, + { + "epoch": 0.8, + "learning_rate": 1.7340003005646323e-05, + "loss": 1.0446, + "step": 12499 + }, + { + "epoch": 0.8, + "learning_rate": 1.7339788316623372e-05, + "loss": 1.1007, + "step": 12500 + }, + { + "epoch": 0.8, + "learning_rate": 1.733957362760042e-05, + "loss": 1.0669, + "step": 12501 + }, + { + "epoch": 0.8, + "learning_rate": 1.7339358938577473e-05, + "loss": 1.0595, + "step": 12502 + }, + { + "epoch": 0.8, + "learning_rate": 1.7339144249554522e-05, + "loss": 1.048, + "step": 12503 + }, + { + "epoch": 0.8, + "learning_rate": 1.733892956053157e-05, + "loss": 1.1075, + "step": 12504 + }, + { + "epoch": 0.8, + "learning_rate": 1.733871487150862e-05, + "loss": 1.0225, + "step": 12505 + }, + { + "epoch": 0.8, + "learning_rate": 1.7338500182485673e-05, + "loss": 1.0803, + "step": 12506 + }, + { + "epoch": 0.8, + "learning_rate": 1.733828549346272e-05, + "loss": 1.1274, + "step": 12507 + }, + { + "epoch": 0.8, + "learning_rate": 1.733807080443977e-05, + "loss": 1.1212, + "step": 12508 + }, + { + "epoch": 0.8, + "learning_rate": 1.733785611541682e-05, + "loss": 1.102, + "step": 12509 + }, + { + "epoch": 0.8, + "learning_rate": 1.733764142639387e-05, + "loss": 1.142, + "step": 12510 + }, + { + "epoch": 0.8, + "learning_rate": 1.733742673737092e-05, + "loss": 1.047, + "step": 12511 + }, + { + "epoch": 0.8, + "learning_rate": 1.733721204834797e-05, + "loss": 1.0551, + "step": 12512 + }, + { + "epoch": 0.81, + "learning_rate": 1.733699735932502e-05, + "loss": 1.1015, + "step": 12513 + }, + { + "epoch": 0.81, + "learning_rate": 1.7336782670302068e-05, + "loss": 1.1899, + "step": 12514 + }, + { + "epoch": 0.81, + "learning_rate": 1.7336567981279117e-05, + "loss": 1.102, + "step": 12515 + }, + { + "epoch": 0.81, + "learning_rate": 1.7336353292256166e-05, + "loss": 1.0584, + "step": 12516 + }, + { + "epoch": 0.81, + "learning_rate": 1.7336138603233218e-05, + "loss": 0.9732, + "step": 12517 + }, + { + "epoch": 0.81, + "learning_rate": 1.7335923914210267e-05, + "loss": 1.0634, + "step": 12518 + }, + { + "epoch": 0.81, + "learning_rate": 1.733570922518732e-05, + "loss": 0.9956, + "step": 12519 + }, + { + "epoch": 0.81, + "learning_rate": 1.733549453616437e-05, + "loss": 0.978, + "step": 12520 + }, + { + "epoch": 0.81, + "learning_rate": 1.7335279847141417e-05, + "loss": 1.0677, + "step": 12521 + }, + { + "epoch": 0.81, + "learning_rate": 1.7335065158118466e-05, + "loss": 1.1637, + "step": 12522 + }, + { + "epoch": 0.81, + "learning_rate": 1.7334850469095515e-05, + "loss": 1.0575, + "step": 12523 + }, + { + "epoch": 0.81, + "learning_rate": 1.7334635780072568e-05, + "loss": 1.0652, + "step": 12524 + }, + { + "epoch": 0.81, + "learning_rate": 1.7334421091049617e-05, + "loss": 1.1112, + "step": 12525 + }, + { + "epoch": 0.81, + "learning_rate": 1.7334206402026666e-05, + "loss": 0.9963, + "step": 12526 + }, + { + "epoch": 0.81, + "learning_rate": 1.7333991713003715e-05, + "loss": 1.1397, + "step": 12527 + }, + { + "epoch": 0.81, + "learning_rate": 1.7333777023980764e-05, + "loss": 1.0673, + "step": 12528 + }, + { + "epoch": 0.81, + "learning_rate": 1.7333562334957812e-05, + "loss": 1.0687, + "step": 12529 + }, + { + "epoch": 0.81, + "learning_rate": 1.7333347645934865e-05, + "loss": 1.1227, + "step": 12530 + }, + { + "epoch": 0.81, + "learning_rate": 1.7333132956911917e-05, + "loss": 0.9963, + "step": 12531 + }, + { + "epoch": 0.81, + "learning_rate": 1.7332918267888966e-05, + "loss": 1.1176, + "step": 12532 + }, + { + "epoch": 0.81, + "learning_rate": 1.7332703578866015e-05, + "loss": 1.0692, + "step": 12533 + }, + { + "epoch": 0.81, + "learning_rate": 1.7332488889843064e-05, + "loss": 1.1292, + "step": 12534 + }, + { + "epoch": 0.81, + "learning_rate": 1.7332274200820113e-05, + "loss": 1.1197, + "step": 12535 + }, + { + "epoch": 0.81, + "learning_rate": 1.7332059511797162e-05, + "loss": 1.1707, + "step": 12536 + }, + { + "epoch": 0.81, + "learning_rate": 1.7331844822774214e-05, + "loss": 1.0741, + "step": 12537 + }, + { + "epoch": 0.81, + "learning_rate": 1.7331630133751263e-05, + "loss": 1.0973, + "step": 12538 + }, + { + "epoch": 0.81, + "learning_rate": 1.7331415444728312e-05, + "loss": 1.0766, + "step": 12539 + }, + { + "epoch": 0.81, + "learning_rate": 1.733120075570536e-05, + "loss": 0.9931, + "step": 12540 + }, + { + "epoch": 0.81, + "learning_rate": 1.733098606668241e-05, + "loss": 0.9685, + "step": 12541 + }, + { + "epoch": 0.81, + "learning_rate": 1.7330771377659463e-05, + "loss": 1.1063, + "step": 12542 + }, + { + "epoch": 0.81, + "learning_rate": 1.733055668863651e-05, + "loss": 1.0188, + "step": 12543 + }, + { + "epoch": 0.81, + "learning_rate": 1.733034199961356e-05, + "loss": 1.3295, + "step": 12544 + }, + { + "epoch": 0.81, + "learning_rate": 1.733012731059061e-05, + "loss": 0.9741, + "step": 12545 + }, + { + "epoch": 0.81, + "learning_rate": 1.7329912621567662e-05, + "loss": 1.0624, + "step": 12546 + }, + { + "epoch": 0.81, + "learning_rate": 1.732969793254471e-05, + "loss": 1.0166, + "step": 12547 + }, + { + "epoch": 0.81, + "learning_rate": 1.732948324352176e-05, + "loss": 0.9654, + "step": 12548 + }, + { + "epoch": 0.81, + "learning_rate": 1.7329268554498812e-05, + "loss": 1.0216, + "step": 12549 + }, + { + "epoch": 0.81, + "learning_rate": 1.732905386547586e-05, + "loss": 1.0506, + "step": 12550 + }, + { + "epoch": 0.81, + "learning_rate": 1.732883917645291e-05, + "loss": 1.12, + "step": 12551 + }, + { + "epoch": 0.81, + "learning_rate": 1.732862448742996e-05, + "loss": 1.1024, + "step": 12552 + }, + { + "epoch": 0.81, + "learning_rate": 1.7328409798407008e-05, + "loss": 1.0165, + "step": 12553 + }, + { + "epoch": 0.81, + "learning_rate": 1.7328195109384057e-05, + "loss": 1.0962, + "step": 12554 + }, + { + "epoch": 0.81, + "learning_rate": 1.732798042036111e-05, + "loss": 1.1788, + "step": 12555 + }, + { + "epoch": 0.81, + "learning_rate": 1.7327765731338158e-05, + "loss": 1.0182, + "step": 12556 + }, + { + "epoch": 0.81, + "learning_rate": 1.7327551042315207e-05, + "loss": 1.0465, + "step": 12557 + }, + { + "epoch": 0.81, + "learning_rate": 1.7327336353292256e-05, + "loss": 1.109, + "step": 12558 + }, + { + "epoch": 0.81, + "learning_rate": 1.7327121664269305e-05, + "loss": 1.0474, + "step": 12559 + }, + { + "epoch": 0.81, + "learning_rate": 1.7326906975246357e-05, + "loss": 1.0763, + "step": 12560 + }, + { + "epoch": 0.81, + "learning_rate": 1.7326692286223406e-05, + "loss": 1.0626, + "step": 12561 + }, + { + "epoch": 0.81, + "learning_rate": 1.732647759720046e-05, + "loss": 1.0815, + "step": 12562 + }, + { + "epoch": 0.81, + "learning_rate": 1.7326262908177508e-05, + "loss": 1.0115, + "step": 12563 + }, + { + "epoch": 0.81, + "learning_rate": 1.7326048219154557e-05, + "loss": 1.1024, + "step": 12564 + }, + { + "epoch": 0.81, + "learning_rate": 1.7325833530131606e-05, + "loss": 1.0552, + "step": 12565 + }, + { + "epoch": 0.81, + "learning_rate": 1.7325618841108655e-05, + "loss": 1.1435, + "step": 12566 + }, + { + "epoch": 0.81, + "learning_rate": 1.7325404152085704e-05, + "loss": 1.0138, + "step": 12567 + }, + { + "epoch": 0.81, + "learning_rate": 1.7325189463062756e-05, + "loss": 1.001, + "step": 12568 + }, + { + "epoch": 0.81, + "learning_rate": 1.7324974774039805e-05, + "loss": 1.041, + "step": 12569 + }, + { + "epoch": 0.81, + "learning_rate": 1.7324760085016854e-05, + "loss": 1.0635, + "step": 12570 + }, + { + "epoch": 0.81, + "learning_rate": 1.7324545395993903e-05, + "loss": 0.9749, + "step": 12571 + }, + { + "epoch": 0.81, + "learning_rate": 1.7324330706970952e-05, + "loss": 1.0615, + "step": 12572 + }, + { + "epoch": 0.81, + "learning_rate": 1.7324116017948004e-05, + "loss": 1.0418, + "step": 12573 + }, + { + "epoch": 0.81, + "learning_rate": 1.7323901328925053e-05, + "loss": 1.1202, + "step": 12574 + }, + { + "epoch": 0.81, + "learning_rate": 1.7323686639902106e-05, + "loss": 1.089, + "step": 12575 + }, + { + "epoch": 0.81, + "learning_rate": 1.7323471950879154e-05, + "loss": 1.0764, + "step": 12576 + }, + { + "epoch": 0.81, + "learning_rate": 1.7323257261856203e-05, + "loss": 1.0844, + "step": 12577 + }, + { + "epoch": 0.81, + "learning_rate": 1.7323042572833252e-05, + "loss": 1.0674, + "step": 12578 + }, + { + "epoch": 0.81, + "learning_rate": 1.73228278838103e-05, + "loss": 1.101, + "step": 12579 + }, + { + "epoch": 0.81, + "learning_rate": 1.7322613194787354e-05, + "loss": 1.0009, + "step": 12580 + }, + { + "epoch": 0.81, + "learning_rate": 1.7322398505764403e-05, + "loss": 1.0518, + "step": 12581 + }, + { + "epoch": 0.81, + "learning_rate": 1.732218381674145e-05, + "loss": 1.2034, + "step": 12582 + }, + { + "epoch": 0.81, + "learning_rate": 1.73219691277185e-05, + "loss": 1.0443, + "step": 12583 + }, + { + "epoch": 0.81, + "learning_rate": 1.732175443869555e-05, + "loss": 1.0371, + "step": 12584 + }, + { + "epoch": 0.81, + "learning_rate": 1.73215397496726e-05, + "loss": 1.0548, + "step": 12585 + }, + { + "epoch": 0.81, + "learning_rate": 1.732132506064965e-05, + "loss": 1.1301, + "step": 12586 + }, + { + "epoch": 0.81, + "learning_rate": 1.73211103716267e-05, + "loss": 1.0722, + "step": 12587 + }, + { + "epoch": 0.81, + "learning_rate": 1.732089568260375e-05, + "loss": 1.0468, + "step": 12588 + }, + { + "epoch": 0.81, + "learning_rate": 1.73206809935808e-05, + "loss": 1.041, + "step": 12589 + }, + { + "epoch": 0.81, + "learning_rate": 1.732046630455785e-05, + "loss": 1.0901, + "step": 12590 + }, + { + "epoch": 0.81, + "learning_rate": 1.73202516155349e-05, + "loss": 0.9495, + "step": 12591 + }, + { + "epoch": 0.81, + "learning_rate": 1.7320036926511948e-05, + "loss": 1.1246, + "step": 12592 + }, + { + "epoch": 0.81, + "learning_rate": 1.7319822237489e-05, + "loss": 0.9648, + "step": 12593 + }, + { + "epoch": 0.81, + "learning_rate": 1.731960754846605e-05, + "loss": 1.074, + "step": 12594 + }, + { + "epoch": 0.81, + "learning_rate": 1.73193928594431e-05, + "loss": 0.9927, + "step": 12595 + }, + { + "epoch": 0.81, + "learning_rate": 1.7319178170420147e-05, + "loss": 1.0087, + "step": 12596 + }, + { + "epoch": 0.81, + "learning_rate": 1.7318963481397196e-05, + "loss": 1.079, + "step": 12597 + }, + { + "epoch": 0.81, + "learning_rate": 1.731874879237425e-05, + "loss": 1.0862, + "step": 12598 + }, + { + "epoch": 0.81, + "learning_rate": 1.7318534103351298e-05, + "loss": 1.0309, + "step": 12599 + }, + { + "epoch": 0.81, + "learning_rate": 1.7318319414328347e-05, + "loss": 1.1244, + "step": 12600 + }, + { + "epoch": 0.81, + "learning_rate": 1.7318104725305396e-05, + "loss": 0.9718, + "step": 12601 + }, + { + "epoch": 0.81, + "learning_rate": 1.7317890036282445e-05, + "loss": 1.0146, + "step": 12602 + }, + { + "epoch": 0.81, + "learning_rate": 1.7317675347259493e-05, + "loss": 0.9479, + "step": 12603 + }, + { + "epoch": 0.81, + "learning_rate": 1.7317460658236546e-05, + "loss": 1.1179, + "step": 12604 + }, + { + "epoch": 0.81, + "learning_rate": 1.7317245969213595e-05, + "loss": 1.1163, + "step": 12605 + }, + { + "epoch": 0.81, + "learning_rate": 1.7317031280190647e-05, + "loss": 1.185, + "step": 12606 + }, + { + "epoch": 0.81, + "learning_rate": 1.7316816591167696e-05, + "loss": 1.0986, + "step": 12607 + }, + { + "epoch": 0.81, + "learning_rate": 1.7316601902144745e-05, + "loss": 1.0778, + "step": 12608 + }, + { + "epoch": 0.81, + "learning_rate": 1.7316387213121794e-05, + "loss": 0.9101, + "step": 12609 + }, + { + "epoch": 0.81, + "learning_rate": 1.7316172524098843e-05, + "loss": 1.1289, + "step": 12610 + }, + { + "epoch": 0.81, + "learning_rate": 1.7315957835075895e-05, + "loss": 1.1358, + "step": 12611 + }, + { + "epoch": 0.81, + "learning_rate": 1.7315743146052944e-05, + "loss": 1.0077, + "step": 12612 + }, + { + "epoch": 0.81, + "learning_rate": 1.7315528457029993e-05, + "loss": 1.0887, + "step": 12613 + }, + { + "epoch": 0.81, + "learning_rate": 1.7315313768007042e-05, + "loss": 1.0759, + "step": 12614 + }, + { + "epoch": 0.81, + "learning_rate": 1.731509907898409e-05, + "loss": 1.0248, + "step": 12615 + }, + { + "epoch": 0.81, + "learning_rate": 1.731488438996114e-05, + "loss": 1.1389, + "step": 12616 + }, + { + "epoch": 0.81, + "learning_rate": 1.7314669700938193e-05, + "loss": 1.0836, + "step": 12617 + }, + { + "epoch": 0.81, + "learning_rate": 1.7314455011915245e-05, + "loss": 1.1601, + "step": 12618 + }, + { + "epoch": 0.81, + "learning_rate": 1.7314240322892294e-05, + "loss": 1.0508, + "step": 12619 + }, + { + "epoch": 0.81, + "learning_rate": 1.7314025633869343e-05, + "loss": 1.0651, + "step": 12620 + }, + { + "epoch": 0.81, + "learning_rate": 1.7313810944846392e-05, + "loss": 1.0532, + "step": 12621 + }, + { + "epoch": 0.81, + "learning_rate": 1.731359625582344e-05, + "loss": 1.0704, + "step": 12622 + }, + { + "epoch": 0.81, + "learning_rate": 1.731338156680049e-05, + "loss": 1.0866, + "step": 12623 + }, + { + "epoch": 0.81, + "learning_rate": 1.7313166877777542e-05, + "loss": 1.0997, + "step": 12624 + }, + { + "epoch": 0.81, + "learning_rate": 1.731295218875459e-05, + "loss": 0.9954, + "step": 12625 + }, + { + "epoch": 0.81, + "learning_rate": 1.731273749973164e-05, + "loss": 1.1484, + "step": 12626 + }, + { + "epoch": 0.81, + "learning_rate": 1.731252281070869e-05, + "loss": 1.0531, + "step": 12627 + }, + { + "epoch": 0.81, + "learning_rate": 1.7312308121685738e-05, + "loss": 1.0419, + "step": 12628 + }, + { + "epoch": 0.81, + "learning_rate": 1.731209343266279e-05, + "loss": 1.0174, + "step": 12629 + }, + { + "epoch": 0.81, + "learning_rate": 1.731187874363984e-05, + "loss": 1.0431, + "step": 12630 + }, + { + "epoch": 0.81, + "learning_rate": 1.7311664054616888e-05, + "loss": 1.0237, + "step": 12631 + }, + { + "epoch": 0.81, + "learning_rate": 1.7311449365593937e-05, + "loss": 1.0363, + "step": 12632 + }, + { + "epoch": 0.81, + "learning_rate": 1.731123467657099e-05, + "loss": 1.0946, + "step": 12633 + }, + { + "epoch": 0.81, + "learning_rate": 1.731101998754804e-05, + "loss": 1.1029, + "step": 12634 + }, + { + "epoch": 0.81, + "learning_rate": 1.7310805298525087e-05, + "loss": 0.9726, + "step": 12635 + }, + { + "epoch": 0.81, + "learning_rate": 1.7310590609502136e-05, + "loss": 1.0677, + "step": 12636 + }, + { + "epoch": 0.81, + "learning_rate": 1.731037592047919e-05, + "loss": 1.0091, + "step": 12637 + }, + { + "epoch": 0.81, + "learning_rate": 1.7310161231456238e-05, + "loss": 1.0259, + "step": 12638 + }, + { + "epoch": 0.81, + "learning_rate": 1.7309946542433287e-05, + "loss": 1.046, + "step": 12639 + }, + { + "epoch": 0.81, + "learning_rate": 1.7309731853410336e-05, + "loss": 1.0478, + "step": 12640 + }, + { + "epoch": 0.81, + "learning_rate": 1.7309517164387385e-05, + "loss": 1.0891, + "step": 12641 + }, + { + "epoch": 0.81, + "learning_rate": 1.7309302475364437e-05, + "loss": 1.027, + "step": 12642 + }, + { + "epoch": 0.81, + "learning_rate": 1.7309087786341486e-05, + "loss": 1.0052, + "step": 12643 + }, + { + "epoch": 0.81, + "learning_rate": 1.7308873097318535e-05, + "loss": 1.0379, + "step": 12644 + }, + { + "epoch": 0.81, + "learning_rate": 1.7308658408295584e-05, + "loss": 1.125, + "step": 12645 + }, + { + "epoch": 0.81, + "learning_rate": 1.7308443719272633e-05, + "loss": 1.0759, + "step": 12646 + }, + { + "epoch": 0.81, + "learning_rate": 1.7308229030249685e-05, + "loss": 1.0117, + "step": 12647 + }, + { + "epoch": 0.81, + "learning_rate": 1.7308014341226734e-05, + "loss": 1.0362, + "step": 12648 + }, + { + "epoch": 0.81, + "learning_rate": 1.7307799652203787e-05, + "loss": 1.0474, + "step": 12649 + }, + { + "epoch": 0.81, + "learning_rate": 1.7307584963180835e-05, + "loss": 1.0918, + "step": 12650 + }, + { + "epoch": 0.81, + "learning_rate": 1.7307370274157884e-05, + "loss": 1.0729, + "step": 12651 + }, + { + "epoch": 0.81, + "learning_rate": 1.7307155585134933e-05, + "loss": 1.0136, + "step": 12652 + }, + { + "epoch": 0.81, + "learning_rate": 1.7306940896111982e-05, + "loss": 1.089, + "step": 12653 + }, + { + "epoch": 0.81, + "learning_rate": 1.730672620708903e-05, + "loss": 0.9884, + "step": 12654 + }, + { + "epoch": 0.81, + "learning_rate": 1.7306511518066084e-05, + "loss": 1.0191, + "step": 12655 + }, + { + "epoch": 0.81, + "learning_rate": 1.7306296829043133e-05, + "loss": 1.1113, + "step": 12656 + }, + { + "epoch": 0.81, + "learning_rate": 1.730608214002018e-05, + "loss": 1.0877, + "step": 12657 + }, + { + "epoch": 0.81, + "learning_rate": 1.730586745099723e-05, + "loss": 1.0022, + "step": 12658 + }, + { + "epoch": 0.81, + "learning_rate": 1.730565276197428e-05, + "loss": 1.0402, + "step": 12659 + }, + { + "epoch": 0.81, + "learning_rate": 1.7305438072951332e-05, + "loss": 1.0869, + "step": 12660 + }, + { + "epoch": 0.81, + "learning_rate": 1.730522338392838e-05, + "loss": 1.0802, + "step": 12661 + }, + { + "epoch": 0.81, + "learning_rate": 1.7305008694905433e-05, + "loss": 1.0577, + "step": 12662 + }, + { + "epoch": 0.81, + "learning_rate": 1.7304794005882482e-05, + "loss": 1.0533, + "step": 12663 + }, + { + "epoch": 0.81, + "learning_rate": 1.730457931685953e-05, + "loss": 1.082, + "step": 12664 + }, + { + "epoch": 0.81, + "learning_rate": 1.730436462783658e-05, + "loss": 1.0552, + "step": 12665 + }, + { + "epoch": 0.81, + "learning_rate": 1.730414993881363e-05, + "loss": 1.0197, + "step": 12666 + }, + { + "epoch": 0.81, + "learning_rate": 1.730393524979068e-05, + "loss": 1.0306, + "step": 12667 + }, + { + "epoch": 0.82, + "learning_rate": 1.730372056076773e-05, + "loss": 1.0902, + "step": 12668 + }, + { + "epoch": 0.82, + "learning_rate": 1.730350587174478e-05, + "loss": 1.062, + "step": 12669 + }, + { + "epoch": 0.82, + "learning_rate": 1.730329118272183e-05, + "loss": 1.0308, + "step": 12670 + }, + { + "epoch": 0.82, + "learning_rate": 1.7303076493698877e-05, + "loss": 1.0915, + "step": 12671 + }, + { + "epoch": 0.82, + "learning_rate": 1.7302861804675926e-05, + "loss": 1.0417, + "step": 12672 + }, + { + "epoch": 0.82, + "learning_rate": 1.730264711565298e-05, + "loss": 1.0305, + "step": 12673 + }, + { + "epoch": 0.82, + "learning_rate": 1.7302432426630028e-05, + "loss": 1.1035, + "step": 12674 + }, + { + "epoch": 0.82, + "learning_rate": 1.7302217737607077e-05, + "loss": 1.0632, + "step": 12675 + }, + { + "epoch": 0.82, + "learning_rate": 1.730200304858413e-05, + "loss": 1.0031, + "step": 12676 + }, + { + "epoch": 0.82, + "learning_rate": 1.7301788359561178e-05, + "loss": 1.0428, + "step": 12677 + }, + { + "epoch": 0.82, + "learning_rate": 1.7301573670538227e-05, + "loss": 1.0013, + "step": 12678 + }, + { + "epoch": 0.82, + "learning_rate": 1.7301358981515276e-05, + "loss": 1.0104, + "step": 12679 + }, + { + "epoch": 0.82, + "learning_rate": 1.7301144292492328e-05, + "loss": 1.0436, + "step": 12680 + }, + { + "epoch": 0.82, + "learning_rate": 1.7300929603469377e-05, + "loss": 1.0887, + "step": 12681 + }, + { + "epoch": 0.82, + "learning_rate": 1.7300714914446426e-05, + "loss": 1.083, + "step": 12682 + }, + { + "epoch": 0.82, + "learning_rate": 1.7300500225423475e-05, + "loss": 1.0167, + "step": 12683 + }, + { + "epoch": 0.82, + "learning_rate": 1.7300285536400524e-05, + "loss": 1.0702, + "step": 12684 + }, + { + "epoch": 0.82, + "learning_rate": 1.7300070847377576e-05, + "loss": 1.1852, + "step": 12685 + }, + { + "epoch": 0.82, + "learning_rate": 1.7299856158354625e-05, + "loss": 1.0409, + "step": 12686 + }, + { + "epoch": 0.82, + "learning_rate": 1.7299641469331674e-05, + "loss": 1.0537, + "step": 12687 + }, + { + "epoch": 0.82, + "learning_rate": 1.7299426780308723e-05, + "loss": 1.1154, + "step": 12688 + }, + { + "epoch": 0.82, + "learning_rate": 1.7299212091285772e-05, + "loss": 0.9931, + "step": 12689 + }, + { + "epoch": 0.82, + "learning_rate": 1.729899740226282e-05, + "loss": 1.0521, + "step": 12690 + }, + { + "epoch": 0.82, + "learning_rate": 1.7298782713239874e-05, + "loss": 1.0093, + "step": 12691 + }, + { + "epoch": 0.82, + "learning_rate": 1.7298568024216923e-05, + "loss": 1.0397, + "step": 12692 + }, + { + "epoch": 0.82, + "learning_rate": 1.7298353335193975e-05, + "loss": 1.0542, + "step": 12693 + }, + { + "epoch": 0.82, + "learning_rate": 1.7298138646171024e-05, + "loss": 1.0252, + "step": 12694 + }, + { + "epoch": 0.82, + "learning_rate": 1.7297923957148073e-05, + "loss": 1.0804, + "step": 12695 + }, + { + "epoch": 0.82, + "learning_rate": 1.7297709268125122e-05, + "loss": 1.1113, + "step": 12696 + }, + { + "epoch": 0.82, + "learning_rate": 1.729749457910217e-05, + "loss": 1.0765, + "step": 12697 + }, + { + "epoch": 0.82, + "learning_rate": 1.7297279890079223e-05, + "loss": 1.1044, + "step": 12698 + }, + { + "epoch": 0.82, + "learning_rate": 1.7297065201056272e-05, + "loss": 1.0912, + "step": 12699 + }, + { + "epoch": 0.82, + "learning_rate": 1.729685051203332e-05, + "loss": 1.0087, + "step": 12700 + }, + { + "epoch": 0.82, + "learning_rate": 1.729663582301037e-05, + "loss": 1.1452, + "step": 12701 + }, + { + "epoch": 0.82, + "learning_rate": 1.729642113398742e-05, + "loss": 1.1041, + "step": 12702 + }, + { + "epoch": 0.82, + "learning_rate": 1.7296206444964468e-05, + "loss": 1.0174, + "step": 12703 + }, + { + "epoch": 0.82, + "learning_rate": 1.729599175594152e-05, + "loss": 1.0908, + "step": 12704 + }, + { + "epoch": 0.82, + "learning_rate": 1.7295777066918573e-05, + "loss": 0.9544, + "step": 12705 + }, + { + "epoch": 0.82, + "learning_rate": 1.729556237789562e-05, + "loss": 1.0503, + "step": 12706 + }, + { + "epoch": 0.82, + "learning_rate": 1.729534768887267e-05, + "loss": 1.1095, + "step": 12707 + }, + { + "epoch": 0.82, + "learning_rate": 1.729513299984972e-05, + "loss": 1.0945, + "step": 12708 + }, + { + "epoch": 0.82, + "learning_rate": 1.729491831082677e-05, + "loss": 1.1195, + "step": 12709 + }, + { + "epoch": 0.82, + "learning_rate": 1.7294703621803817e-05, + "loss": 1.0351, + "step": 12710 + }, + { + "epoch": 0.82, + "learning_rate": 1.729448893278087e-05, + "loss": 1.003, + "step": 12711 + }, + { + "epoch": 0.82, + "learning_rate": 1.729427424375792e-05, + "loss": 1.0112, + "step": 12712 + }, + { + "epoch": 0.82, + "learning_rate": 1.7294059554734968e-05, + "loss": 1.0092, + "step": 12713 + }, + { + "epoch": 0.82, + "learning_rate": 1.7293844865712017e-05, + "loss": 1.0018, + "step": 12714 + }, + { + "epoch": 0.82, + "learning_rate": 1.7293630176689066e-05, + "loss": 1.0735, + "step": 12715 + }, + { + "epoch": 0.82, + "learning_rate": 1.7293415487666118e-05, + "loss": 1.1218, + "step": 12716 + }, + { + "epoch": 0.82, + "learning_rate": 1.7293200798643167e-05, + "loss": 1.0972, + "step": 12717 + }, + { + "epoch": 0.82, + "learning_rate": 1.7292986109620216e-05, + "loss": 1.0606, + "step": 12718 + }, + { + "epoch": 0.82, + "learning_rate": 1.7292771420597265e-05, + "loss": 0.957, + "step": 12719 + }, + { + "epoch": 0.82, + "learning_rate": 1.7292556731574317e-05, + "loss": 1.0035, + "step": 12720 + }, + { + "epoch": 0.82, + "learning_rate": 1.7292342042551366e-05, + "loss": 1.0783, + "step": 12721 + }, + { + "epoch": 0.82, + "learning_rate": 1.7292127353528415e-05, + "loss": 0.9893, + "step": 12722 + }, + { + "epoch": 0.82, + "learning_rate": 1.7291912664505464e-05, + "loss": 0.9778, + "step": 12723 + }, + { + "epoch": 0.82, + "learning_rate": 1.7291697975482516e-05, + "loss": 0.9583, + "step": 12724 + }, + { + "epoch": 0.82, + "learning_rate": 1.7291483286459565e-05, + "loss": 1.0844, + "step": 12725 + }, + { + "epoch": 0.82, + "learning_rate": 1.7291268597436614e-05, + "loss": 1.0452, + "step": 12726 + }, + { + "epoch": 0.82, + "learning_rate": 1.7291053908413663e-05, + "loss": 1.0548, + "step": 12727 + }, + { + "epoch": 0.82, + "learning_rate": 1.7290839219390712e-05, + "loss": 1.0641, + "step": 12728 + }, + { + "epoch": 0.82, + "learning_rate": 1.7290624530367765e-05, + "loss": 1.0237, + "step": 12729 + }, + { + "epoch": 0.82, + "learning_rate": 1.7290409841344814e-05, + "loss": 1.0688, + "step": 12730 + }, + { + "epoch": 0.82, + "learning_rate": 1.7290195152321863e-05, + "loss": 1.0556, + "step": 12731 + }, + { + "epoch": 0.82, + "learning_rate": 1.728998046329891e-05, + "loss": 1.0713, + "step": 12732 + }, + { + "epoch": 0.82, + "learning_rate": 1.728976577427596e-05, + "loss": 1.0612, + "step": 12733 + }, + { + "epoch": 0.82, + "learning_rate": 1.7289551085253013e-05, + "loss": 1.0432, + "step": 12734 + }, + { + "epoch": 0.82, + "learning_rate": 1.7289336396230062e-05, + "loss": 1.0487, + "step": 12735 + }, + { + "epoch": 0.82, + "learning_rate": 1.7289121707207114e-05, + "loss": 0.9715, + "step": 12736 + }, + { + "epoch": 0.82, + "learning_rate": 1.7288907018184163e-05, + "loss": 1.204, + "step": 12737 + }, + { + "epoch": 0.82, + "learning_rate": 1.7288692329161212e-05, + "loss": 1.1132, + "step": 12738 + }, + { + "epoch": 0.82, + "learning_rate": 1.728847764013826e-05, + "loss": 1.0129, + "step": 12739 + }, + { + "epoch": 0.82, + "learning_rate": 1.728826295111531e-05, + "loss": 1.144, + "step": 12740 + }, + { + "epoch": 0.82, + "learning_rate": 1.728804826209236e-05, + "loss": 1.1233, + "step": 12741 + }, + { + "epoch": 0.82, + "learning_rate": 1.728783357306941e-05, + "loss": 1.1115, + "step": 12742 + }, + { + "epoch": 0.82, + "learning_rate": 1.728761888404646e-05, + "loss": 1.0617, + "step": 12743 + }, + { + "epoch": 0.82, + "learning_rate": 1.728740419502351e-05, + "loss": 1.0053, + "step": 12744 + }, + { + "epoch": 0.82, + "learning_rate": 1.7287189506000558e-05, + "loss": 1.1118, + "step": 12745 + }, + { + "epoch": 0.82, + "learning_rate": 1.7286974816977607e-05, + "loss": 1.1125, + "step": 12746 + }, + { + "epoch": 0.82, + "learning_rate": 1.728676012795466e-05, + "loss": 1.1322, + "step": 12747 + }, + { + "epoch": 0.82, + "learning_rate": 1.728654543893171e-05, + "loss": 1.1028, + "step": 12748 + }, + { + "epoch": 0.82, + "learning_rate": 1.728633074990876e-05, + "loss": 1.0102, + "step": 12749 + }, + { + "epoch": 0.82, + "learning_rate": 1.728611606088581e-05, + "loss": 1.1439, + "step": 12750 + }, + { + "epoch": 0.82, + "learning_rate": 1.728590137186286e-05, + "loss": 0.9869, + "step": 12751 + }, + { + "epoch": 0.82, + "learning_rate": 1.7285686682839908e-05, + "loss": 1.0448, + "step": 12752 + }, + { + "epoch": 0.82, + "learning_rate": 1.7285471993816957e-05, + "loss": 1.0495, + "step": 12753 + }, + { + "epoch": 0.82, + "learning_rate": 1.728525730479401e-05, + "loss": 1.0106, + "step": 12754 + }, + { + "epoch": 0.82, + "learning_rate": 1.7285042615771058e-05, + "loss": 1.086, + "step": 12755 + }, + { + "epoch": 0.82, + "learning_rate": 1.7284827926748107e-05, + "loss": 1.0292, + "step": 12756 + }, + { + "epoch": 0.82, + "learning_rate": 1.7284613237725156e-05, + "loss": 1.1152, + "step": 12757 + }, + { + "epoch": 0.82, + "learning_rate": 1.7284398548702205e-05, + "loss": 1.0546, + "step": 12758 + }, + { + "epoch": 0.82, + "learning_rate": 1.7284183859679254e-05, + "loss": 1.0373, + "step": 12759 + }, + { + "epoch": 0.82, + "learning_rate": 1.7283969170656306e-05, + "loss": 1.1294, + "step": 12760 + }, + { + "epoch": 0.82, + "learning_rate": 1.7283754481633355e-05, + "loss": 1.0758, + "step": 12761 + }, + { + "epoch": 0.82, + "learning_rate": 1.7283539792610404e-05, + "loss": 1.0442, + "step": 12762 + }, + { + "epoch": 0.82, + "learning_rate": 1.7283325103587457e-05, + "loss": 1.07, + "step": 12763 + }, + { + "epoch": 0.82, + "learning_rate": 1.7283110414564506e-05, + "loss": 1.0221, + "step": 12764 + }, + { + "epoch": 0.82, + "learning_rate": 1.7282895725541555e-05, + "loss": 1.0631, + "step": 12765 + }, + { + "epoch": 0.82, + "learning_rate": 1.7282681036518604e-05, + "loss": 1.0327, + "step": 12766 + }, + { + "epoch": 0.82, + "learning_rate": 1.7282466347495656e-05, + "loss": 1.0337, + "step": 12767 + }, + { + "epoch": 0.82, + "learning_rate": 1.7282251658472705e-05, + "loss": 1.0062, + "step": 12768 + }, + { + "epoch": 0.82, + "learning_rate": 1.7282036969449754e-05, + "loss": 1.048, + "step": 12769 + }, + { + "epoch": 0.82, + "learning_rate": 1.7281822280426803e-05, + "loss": 0.9494, + "step": 12770 + }, + { + "epoch": 0.82, + "learning_rate": 1.7281607591403852e-05, + "loss": 1.0616, + "step": 12771 + }, + { + "epoch": 0.82, + "learning_rate": 1.7281392902380904e-05, + "loss": 1.0982, + "step": 12772 + }, + { + "epoch": 0.82, + "learning_rate": 1.7281178213357953e-05, + "loss": 1.1396, + "step": 12773 + }, + { + "epoch": 0.82, + "learning_rate": 1.7280963524335002e-05, + "loss": 0.9913, + "step": 12774 + }, + { + "epoch": 0.82, + "learning_rate": 1.728074883531205e-05, + "loss": 1.11, + "step": 12775 + }, + { + "epoch": 0.82, + "learning_rate": 1.72805341462891e-05, + "loss": 1.1211, + "step": 12776 + }, + { + "epoch": 0.82, + "learning_rate": 1.728031945726615e-05, + "loss": 1.0559, + "step": 12777 + }, + { + "epoch": 0.82, + "learning_rate": 1.72801047682432e-05, + "loss": 1.0236, + "step": 12778 + }, + { + "epoch": 0.82, + "learning_rate": 1.727989007922025e-05, + "loss": 1.0529, + "step": 12779 + }, + { + "epoch": 0.82, + "learning_rate": 1.7279675390197303e-05, + "loss": 1.0955, + "step": 12780 + }, + { + "epoch": 0.82, + "learning_rate": 1.727946070117435e-05, + "loss": 1.0624, + "step": 12781 + }, + { + "epoch": 0.82, + "learning_rate": 1.72792460121514e-05, + "loss": 1.0735, + "step": 12782 + }, + { + "epoch": 0.82, + "learning_rate": 1.727903132312845e-05, + "loss": 1.0579, + "step": 12783 + }, + { + "epoch": 0.82, + "learning_rate": 1.72788166341055e-05, + "loss": 1.0905, + "step": 12784 + }, + { + "epoch": 0.82, + "learning_rate": 1.727860194508255e-05, + "loss": 1.0775, + "step": 12785 + }, + { + "epoch": 0.82, + "learning_rate": 1.72783872560596e-05, + "loss": 1.0928, + "step": 12786 + }, + { + "epoch": 0.82, + "learning_rate": 1.727817256703665e-05, + "loss": 1.0256, + "step": 12787 + }, + { + "epoch": 0.82, + "learning_rate": 1.7277957878013698e-05, + "loss": 1.177, + "step": 12788 + }, + { + "epoch": 0.82, + "learning_rate": 1.7277743188990747e-05, + "loss": 0.9958, + "step": 12789 + }, + { + "epoch": 0.82, + "learning_rate": 1.7277528499967796e-05, + "loss": 1.0598, + "step": 12790 + }, + { + "epoch": 0.82, + "learning_rate": 1.7277313810944848e-05, + "loss": 1.0525, + "step": 12791 + }, + { + "epoch": 0.82, + "learning_rate": 1.72770991219219e-05, + "loss": 1.0177, + "step": 12792 + }, + { + "epoch": 0.82, + "learning_rate": 1.727688443289895e-05, + "loss": 1.0199, + "step": 12793 + }, + { + "epoch": 0.82, + "learning_rate": 1.7276669743875998e-05, + "loss": 1.132, + "step": 12794 + }, + { + "epoch": 0.82, + "learning_rate": 1.7276455054853047e-05, + "loss": 1.0715, + "step": 12795 + }, + { + "epoch": 0.82, + "learning_rate": 1.7276240365830096e-05, + "loss": 1.0, + "step": 12796 + }, + { + "epoch": 0.82, + "learning_rate": 1.7276025676807145e-05, + "loss": 1.1185, + "step": 12797 + }, + { + "epoch": 0.82, + "learning_rate": 1.7275810987784197e-05, + "loss": 1.0572, + "step": 12798 + }, + { + "epoch": 0.82, + "learning_rate": 1.7275596298761246e-05, + "loss": 1.0109, + "step": 12799 + }, + { + "epoch": 0.82, + "learning_rate": 1.7275381609738295e-05, + "loss": 1.0057, + "step": 12800 + }, + { + "epoch": 0.82, + "learning_rate": 1.7275166920715344e-05, + "loss": 1.1096, + "step": 12801 + }, + { + "epoch": 0.82, + "learning_rate": 1.7274952231692393e-05, + "loss": 1.0254, + "step": 12802 + }, + { + "epoch": 0.82, + "learning_rate": 1.7274737542669446e-05, + "loss": 1.0089, + "step": 12803 + }, + { + "epoch": 0.82, + "learning_rate": 1.7274522853646495e-05, + "loss": 1.0922, + "step": 12804 + }, + { + "epoch": 0.82, + "learning_rate": 1.7274308164623544e-05, + "loss": 1.02, + "step": 12805 + }, + { + "epoch": 0.82, + "learning_rate": 1.7274093475600593e-05, + "loss": 1.0737, + "step": 12806 + }, + { + "epoch": 0.82, + "learning_rate": 1.7273878786577645e-05, + "loss": 1.074, + "step": 12807 + }, + { + "epoch": 0.82, + "learning_rate": 1.7273664097554694e-05, + "loss": 1.0276, + "step": 12808 + }, + { + "epoch": 0.82, + "learning_rate": 1.7273449408531743e-05, + "loss": 1.0439, + "step": 12809 + }, + { + "epoch": 0.82, + "learning_rate": 1.7273234719508792e-05, + "loss": 1.0849, + "step": 12810 + }, + { + "epoch": 0.82, + "learning_rate": 1.7273020030485844e-05, + "loss": 1.052, + "step": 12811 + }, + { + "epoch": 0.82, + "learning_rate": 1.7272805341462893e-05, + "loss": 1.1008, + "step": 12812 + }, + { + "epoch": 0.82, + "learning_rate": 1.7272590652439942e-05, + "loss": 1.0619, + "step": 12813 + }, + { + "epoch": 0.82, + "learning_rate": 1.727237596341699e-05, + "loss": 1.0964, + "step": 12814 + }, + { + "epoch": 0.82, + "learning_rate": 1.727216127439404e-05, + "loss": 0.9683, + "step": 12815 + }, + { + "epoch": 0.82, + "learning_rate": 1.7271946585371092e-05, + "loss": 1.0351, + "step": 12816 + }, + { + "epoch": 0.82, + "learning_rate": 1.727173189634814e-05, + "loss": 1.0894, + "step": 12817 + }, + { + "epoch": 0.82, + "learning_rate": 1.727151720732519e-05, + "loss": 1.0349, + "step": 12818 + }, + { + "epoch": 0.82, + "learning_rate": 1.727130251830224e-05, + "loss": 1.047, + "step": 12819 + }, + { + "epoch": 0.82, + "learning_rate": 1.7271087829279288e-05, + "loss": 1.1024, + "step": 12820 + }, + { + "epoch": 0.82, + "learning_rate": 1.727087314025634e-05, + "loss": 0.968, + "step": 12821 + }, + { + "epoch": 0.82, + "learning_rate": 1.727065845123339e-05, + "loss": 1.0564, + "step": 12822 + }, + { + "epoch": 0.83, + "learning_rate": 1.7270443762210442e-05, + "loss": 1.0423, + "step": 12823 + }, + { + "epoch": 0.83, + "learning_rate": 1.727022907318749e-05, + "loss": 1.1305, + "step": 12824 + }, + { + "epoch": 0.83, + "learning_rate": 1.727001438416454e-05, + "loss": 1.0527, + "step": 12825 + }, + { + "epoch": 0.83, + "learning_rate": 1.726979969514159e-05, + "loss": 1.2076, + "step": 12826 + }, + { + "epoch": 0.83, + "learning_rate": 1.7269585006118638e-05, + "loss": 1.0662, + "step": 12827 + }, + { + "epoch": 0.83, + "learning_rate": 1.7269370317095687e-05, + "loss": 1.1147, + "step": 12828 + }, + { + "epoch": 0.83, + "learning_rate": 1.726915562807274e-05, + "loss": 1.0364, + "step": 12829 + }, + { + "epoch": 0.83, + "learning_rate": 1.7268940939049788e-05, + "loss": 1.0768, + "step": 12830 + }, + { + "epoch": 0.83, + "learning_rate": 1.7268726250026837e-05, + "loss": 1.0704, + "step": 12831 + }, + { + "epoch": 0.83, + "learning_rate": 1.7268511561003886e-05, + "loss": 1.1275, + "step": 12832 + }, + { + "epoch": 0.83, + "learning_rate": 1.7268296871980935e-05, + "loss": 1.0332, + "step": 12833 + }, + { + "epoch": 0.83, + "learning_rate": 1.7268082182957987e-05, + "loss": 1.0764, + "step": 12834 + }, + { + "epoch": 0.83, + "learning_rate": 1.7267867493935036e-05, + "loss": 1.062, + "step": 12835 + }, + { + "epoch": 0.83, + "learning_rate": 1.726765280491209e-05, + "loss": 1.1064, + "step": 12836 + }, + { + "epoch": 0.83, + "learning_rate": 1.7267438115889138e-05, + "loss": 1.0761, + "step": 12837 + }, + { + "epoch": 0.83, + "learning_rate": 1.7267223426866187e-05, + "loss": 1.0596, + "step": 12838 + }, + { + "epoch": 0.83, + "learning_rate": 1.7267008737843236e-05, + "loss": 0.9894, + "step": 12839 + }, + { + "epoch": 0.83, + "learning_rate": 1.7266794048820285e-05, + "loss": 1.1192, + "step": 12840 + }, + { + "epoch": 0.83, + "learning_rate": 1.7266579359797337e-05, + "loss": 1.0817, + "step": 12841 + }, + { + "epoch": 0.83, + "learning_rate": 1.7266364670774386e-05, + "loss": 1.0566, + "step": 12842 + }, + { + "epoch": 0.83, + "learning_rate": 1.7266149981751435e-05, + "loss": 1.0271, + "step": 12843 + }, + { + "epoch": 0.83, + "learning_rate": 1.7265935292728484e-05, + "loss": 1.0007, + "step": 12844 + }, + { + "epoch": 0.83, + "learning_rate": 1.7265720603705533e-05, + "loss": 1.0506, + "step": 12845 + }, + { + "epoch": 0.83, + "learning_rate": 1.726550591468258e-05, + "loss": 1.0723, + "step": 12846 + }, + { + "epoch": 0.83, + "learning_rate": 1.7265291225659634e-05, + "loss": 1.0521, + "step": 12847 + }, + { + "epoch": 0.83, + "learning_rate": 1.7265076536636683e-05, + "loss": 1.0627, + "step": 12848 + }, + { + "epoch": 0.83, + "learning_rate": 1.7264861847613732e-05, + "loss": 1.1193, + "step": 12849 + }, + { + "epoch": 0.83, + "learning_rate": 1.7264647158590784e-05, + "loss": 0.9827, + "step": 12850 + }, + { + "epoch": 0.83, + "learning_rate": 1.7264432469567833e-05, + "loss": 1.0942, + "step": 12851 + }, + { + "epoch": 0.83, + "learning_rate": 1.7264217780544882e-05, + "loss": 1.0899, + "step": 12852 + }, + { + "epoch": 0.83, + "learning_rate": 1.726400309152193e-05, + "loss": 1.0901, + "step": 12853 + }, + { + "epoch": 0.83, + "learning_rate": 1.7263788402498984e-05, + "loss": 1.0413, + "step": 12854 + }, + { + "epoch": 0.83, + "learning_rate": 1.7263573713476033e-05, + "loss": 1.0464, + "step": 12855 + }, + { + "epoch": 0.83, + "learning_rate": 1.726335902445308e-05, + "loss": 1.1737, + "step": 12856 + }, + { + "epoch": 0.83, + "learning_rate": 1.726314433543013e-05, + "loss": 1.0776, + "step": 12857 + }, + { + "epoch": 0.83, + "learning_rate": 1.726292964640718e-05, + "loss": 1.0975, + "step": 12858 + }, + { + "epoch": 0.83, + "learning_rate": 1.726271495738423e-05, + "loss": 0.9119, + "step": 12859 + }, + { + "epoch": 0.83, + "learning_rate": 1.726250026836128e-05, + "loss": 1.0888, + "step": 12860 + }, + { + "epoch": 0.83, + "learning_rate": 1.726228557933833e-05, + "loss": 1.1036, + "step": 12861 + }, + { + "epoch": 0.83, + "learning_rate": 1.726207089031538e-05, + "loss": 0.9941, + "step": 12862 + }, + { + "epoch": 0.83, + "learning_rate": 1.7261856201292428e-05, + "loss": 1.1202, + "step": 12863 + }, + { + "epoch": 0.83, + "learning_rate": 1.7261641512269477e-05, + "loss": 1.0665, + "step": 12864 + }, + { + "epoch": 0.83, + "learning_rate": 1.726142682324653e-05, + "loss": 1.0375, + "step": 12865 + }, + { + "epoch": 0.83, + "learning_rate": 1.7261212134223578e-05, + "loss": 1.0302, + "step": 12866 + }, + { + "epoch": 0.83, + "learning_rate": 1.726099744520063e-05, + "loss": 1.0114, + "step": 12867 + }, + { + "epoch": 0.83, + "learning_rate": 1.726078275617768e-05, + "loss": 1.1013, + "step": 12868 + }, + { + "epoch": 0.83, + "learning_rate": 1.7260568067154728e-05, + "loss": 1.1192, + "step": 12869 + }, + { + "epoch": 0.83, + "learning_rate": 1.7260353378131777e-05, + "loss": 1.1022, + "step": 12870 + }, + { + "epoch": 0.83, + "learning_rate": 1.7260138689108826e-05, + "loss": 1.0613, + "step": 12871 + }, + { + "epoch": 0.83, + "learning_rate": 1.725992400008588e-05, + "loss": 1.0442, + "step": 12872 + }, + { + "epoch": 0.83, + "learning_rate": 1.7259709311062927e-05, + "loss": 1.0478, + "step": 12873 + }, + { + "epoch": 0.83, + "learning_rate": 1.7259494622039976e-05, + "loss": 1.0137, + "step": 12874 + }, + { + "epoch": 0.83, + "learning_rate": 1.7259279933017025e-05, + "loss": 1.0648, + "step": 12875 + }, + { + "epoch": 0.83, + "learning_rate": 1.7259065243994074e-05, + "loss": 1.0022, + "step": 12876 + }, + { + "epoch": 0.83, + "learning_rate": 1.7258850554971123e-05, + "loss": 0.9555, + "step": 12877 + }, + { + "epoch": 0.83, + "learning_rate": 1.7258635865948176e-05, + "loss": 1.0841, + "step": 12878 + }, + { + "epoch": 0.83, + "learning_rate": 1.7258421176925225e-05, + "loss": 1.0529, + "step": 12879 + }, + { + "epoch": 0.83, + "learning_rate": 1.7258206487902277e-05, + "loss": 1.0664, + "step": 12880 + }, + { + "epoch": 0.83, + "learning_rate": 1.7257991798879326e-05, + "loss": 1.0296, + "step": 12881 + }, + { + "epoch": 0.83, + "learning_rate": 1.7257777109856375e-05, + "loss": 1.0705, + "step": 12882 + }, + { + "epoch": 0.83, + "learning_rate": 1.7257562420833424e-05, + "loss": 1.1175, + "step": 12883 + }, + { + "epoch": 0.83, + "learning_rate": 1.7257347731810473e-05, + "loss": 1.058, + "step": 12884 + }, + { + "epoch": 0.83, + "learning_rate": 1.7257133042787525e-05, + "loss": 1.0135, + "step": 12885 + }, + { + "epoch": 0.83, + "learning_rate": 1.7256918353764574e-05, + "loss": 1.1276, + "step": 12886 + }, + { + "epoch": 0.83, + "learning_rate": 1.7256703664741623e-05, + "loss": 1.0773, + "step": 12887 + }, + { + "epoch": 0.83, + "learning_rate": 1.7256488975718672e-05, + "loss": 1.083, + "step": 12888 + }, + { + "epoch": 0.83, + "learning_rate": 1.725627428669572e-05, + "loss": 1.0381, + "step": 12889 + }, + { + "epoch": 0.83, + "learning_rate": 1.7256059597672773e-05, + "loss": 1.0535, + "step": 12890 + }, + { + "epoch": 0.83, + "learning_rate": 1.7255844908649822e-05, + "loss": 0.992, + "step": 12891 + }, + { + "epoch": 0.83, + "learning_rate": 1.725563021962687e-05, + "loss": 1.0169, + "step": 12892 + }, + { + "epoch": 0.83, + "learning_rate": 1.725541553060392e-05, + "loss": 1.1177, + "step": 12893 + }, + { + "epoch": 0.83, + "learning_rate": 1.7255200841580973e-05, + "loss": 1.0223, + "step": 12894 + }, + { + "epoch": 0.83, + "learning_rate": 1.725498615255802e-05, + "loss": 1.0474, + "step": 12895 + }, + { + "epoch": 0.83, + "learning_rate": 1.725477146353507e-05, + "loss": 1.1262, + "step": 12896 + }, + { + "epoch": 0.83, + "learning_rate": 1.725455677451212e-05, + "loss": 0.9726, + "step": 12897 + }, + { + "epoch": 0.83, + "learning_rate": 1.7254342085489172e-05, + "loss": 1.1051, + "step": 12898 + }, + { + "epoch": 0.83, + "learning_rate": 1.725412739646622e-05, + "loss": 1.0139, + "step": 12899 + }, + { + "epoch": 0.83, + "learning_rate": 1.725391270744327e-05, + "loss": 1.0556, + "step": 12900 + }, + { + "epoch": 0.83, + "learning_rate": 1.725369801842032e-05, + "loss": 1.1142, + "step": 12901 + }, + { + "epoch": 0.83, + "learning_rate": 1.7253483329397368e-05, + "loss": 1.0576, + "step": 12902 + }, + { + "epoch": 0.83, + "learning_rate": 1.725326864037442e-05, + "loss": 1.0488, + "step": 12903 + }, + { + "epoch": 0.83, + "learning_rate": 1.725305395135147e-05, + "loss": 1.1667, + "step": 12904 + }, + { + "epoch": 0.83, + "learning_rate": 1.7252839262328518e-05, + "loss": 1.0804, + "step": 12905 + }, + { + "epoch": 0.83, + "learning_rate": 1.7252624573305567e-05, + "loss": 1.0439, + "step": 12906 + }, + { + "epoch": 0.83, + "learning_rate": 1.7252409884282616e-05, + "loss": 1.0708, + "step": 12907 + }, + { + "epoch": 0.83, + "learning_rate": 1.725219519525967e-05, + "loss": 0.9984, + "step": 12908 + }, + { + "epoch": 0.83, + "learning_rate": 1.7251980506236717e-05, + "loss": 1.0338, + "step": 12909 + }, + { + "epoch": 0.83, + "learning_rate": 1.725176581721377e-05, + "loss": 1.0624, + "step": 12910 + }, + { + "epoch": 0.83, + "learning_rate": 1.725155112819082e-05, + "loss": 1.0103, + "step": 12911 + }, + { + "epoch": 0.83, + "learning_rate": 1.7251336439167868e-05, + "loss": 1.012, + "step": 12912 + }, + { + "epoch": 0.83, + "learning_rate": 1.7251121750144917e-05, + "loss": 1.1071, + "step": 12913 + }, + { + "epoch": 0.83, + "learning_rate": 1.7250907061121966e-05, + "loss": 1.0556, + "step": 12914 + }, + { + "epoch": 0.83, + "learning_rate": 1.7250692372099014e-05, + "loss": 1.0945, + "step": 12915 + }, + { + "epoch": 0.83, + "learning_rate": 1.7250477683076067e-05, + "loss": 1.1081, + "step": 12916 + }, + { + "epoch": 0.83, + "learning_rate": 1.7250262994053116e-05, + "loss": 1.0493, + "step": 12917 + }, + { + "epoch": 0.83, + "learning_rate": 1.7250048305030165e-05, + "loss": 1.0123, + "step": 12918 + }, + { + "epoch": 0.83, + "learning_rate": 1.7249833616007214e-05, + "loss": 1.1122, + "step": 12919 + }, + { + "epoch": 0.83, + "learning_rate": 1.7249618926984263e-05, + "loss": 1.0544, + "step": 12920 + }, + { + "epoch": 0.83, + "learning_rate": 1.7249404237961315e-05, + "loss": 1.0527, + "step": 12921 + }, + { + "epoch": 0.83, + "learning_rate": 1.7249189548938364e-05, + "loss": 1.0999, + "step": 12922 + }, + { + "epoch": 0.83, + "learning_rate": 1.7248974859915416e-05, + "loss": 1.1198, + "step": 12923 + }, + { + "epoch": 0.83, + "learning_rate": 1.7248760170892465e-05, + "loss": 1.0382, + "step": 12924 + }, + { + "epoch": 0.83, + "learning_rate": 1.7248545481869514e-05, + "loss": 0.9902, + "step": 12925 + }, + { + "epoch": 0.83, + "learning_rate": 1.7248330792846563e-05, + "loss": 1.0691, + "step": 12926 + }, + { + "epoch": 0.83, + "learning_rate": 1.7248116103823612e-05, + "loss": 1.1264, + "step": 12927 + }, + { + "epoch": 0.83, + "learning_rate": 1.7247901414800665e-05, + "loss": 1.0454, + "step": 12928 + }, + { + "epoch": 0.83, + "learning_rate": 1.7247686725777714e-05, + "loss": 1.0854, + "step": 12929 + }, + { + "epoch": 0.83, + "learning_rate": 1.7247472036754763e-05, + "loss": 1.0565, + "step": 12930 + }, + { + "epoch": 0.83, + "learning_rate": 1.724725734773181e-05, + "loss": 1.0146, + "step": 12931 + }, + { + "epoch": 0.83, + "learning_rate": 1.724704265870886e-05, + "loss": 1.0387, + "step": 12932 + }, + { + "epoch": 0.83, + "learning_rate": 1.724682796968591e-05, + "loss": 1.0553, + "step": 12933 + }, + { + "epoch": 0.83, + "learning_rate": 1.7246613280662962e-05, + "loss": 1.0615, + "step": 12934 + }, + { + "epoch": 0.83, + "learning_rate": 1.724639859164001e-05, + "loss": 0.9958, + "step": 12935 + }, + { + "epoch": 0.83, + "learning_rate": 1.724618390261706e-05, + "loss": 1.0707, + "step": 12936 + }, + { + "epoch": 0.83, + "learning_rate": 1.724596921359411e-05, + "loss": 1.0512, + "step": 12937 + }, + { + "epoch": 0.83, + "learning_rate": 1.724575452457116e-05, + "loss": 1.0017, + "step": 12938 + }, + { + "epoch": 0.83, + "learning_rate": 1.724553983554821e-05, + "loss": 1.0768, + "step": 12939 + }, + { + "epoch": 0.83, + "learning_rate": 1.724532514652526e-05, + "loss": 1.0822, + "step": 12940 + }, + { + "epoch": 0.83, + "learning_rate": 1.724511045750231e-05, + "loss": 0.9876, + "step": 12941 + }, + { + "epoch": 0.83, + "learning_rate": 1.724489576847936e-05, + "loss": 1.0816, + "step": 12942 + }, + { + "epoch": 0.83, + "learning_rate": 1.724468107945641e-05, + "loss": 1.0697, + "step": 12943 + }, + { + "epoch": 0.83, + "learning_rate": 1.7244466390433458e-05, + "loss": 1.078, + "step": 12944 + }, + { + "epoch": 0.83, + "learning_rate": 1.7244251701410507e-05, + "loss": 1.1073, + "step": 12945 + }, + { + "epoch": 0.83, + "learning_rate": 1.7244037012387556e-05, + "loss": 1.0084, + "step": 12946 + }, + { + "epoch": 0.83, + "learning_rate": 1.724382232336461e-05, + "loss": 1.0135, + "step": 12947 + }, + { + "epoch": 0.83, + "learning_rate": 1.7243607634341657e-05, + "loss": 1.0505, + "step": 12948 + }, + { + "epoch": 0.83, + "learning_rate": 1.7243392945318706e-05, + "loss": 1.1127, + "step": 12949 + }, + { + "epoch": 0.83, + "learning_rate": 1.7243178256295755e-05, + "loss": 1.0174, + "step": 12950 + }, + { + "epoch": 0.83, + "learning_rate": 1.7242963567272804e-05, + "loss": 1.0874, + "step": 12951 + }, + { + "epoch": 0.83, + "learning_rate": 1.7242748878249857e-05, + "loss": 1.0228, + "step": 12952 + }, + { + "epoch": 0.83, + "learning_rate": 1.7242534189226906e-05, + "loss": 1.0613, + "step": 12953 + }, + { + "epoch": 0.83, + "learning_rate": 1.7242319500203958e-05, + "loss": 1.0439, + "step": 12954 + }, + { + "epoch": 0.83, + "learning_rate": 1.7242104811181007e-05, + "loss": 1.0607, + "step": 12955 + }, + { + "epoch": 0.83, + "learning_rate": 1.7241890122158056e-05, + "loss": 1.0309, + "step": 12956 + }, + { + "epoch": 0.83, + "learning_rate": 1.7241675433135105e-05, + "loss": 1.0195, + "step": 12957 + }, + { + "epoch": 0.83, + "learning_rate": 1.7241460744112154e-05, + "loss": 1.0164, + "step": 12958 + }, + { + "epoch": 0.83, + "learning_rate": 1.7241246055089206e-05, + "loss": 1.0852, + "step": 12959 + }, + { + "epoch": 0.83, + "learning_rate": 1.7241031366066255e-05, + "loss": 1.0922, + "step": 12960 + }, + { + "epoch": 0.83, + "learning_rate": 1.7240816677043304e-05, + "loss": 1.1245, + "step": 12961 + }, + { + "epoch": 0.83, + "learning_rate": 1.7240601988020353e-05, + "loss": 1.0359, + "step": 12962 + }, + { + "epoch": 0.83, + "learning_rate": 1.7240387298997402e-05, + "loss": 1.0458, + "step": 12963 + }, + { + "epoch": 0.83, + "learning_rate": 1.724017260997445e-05, + "loss": 1.1392, + "step": 12964 + }, + { + "epoch": 0.83, + "learning_rate": 1.7239957920951503e-05, + "loss": 1.0785, + "step": 12965 + }, + { + "epoch": 0.83, + "learning_rate": 1.7239743231928552e-05, + "loss": 1.0749, + "step": 12966 + }, + { + "epoch": 0.83, + "learning_rate": 1.7239528542905605e-05, + "loss": 1.0983, + "step": 12967 + }, + { + "epoch": 0.83, + "learning_rate": 1.7239313853882654e-05, + "loss": 1.045, + "step": 12968 + }, + { + "epoch": 0.83, + "learning_rate": 1.7239099164859703e-05, + "loss": 1.0329, + "step": 12969 + }, + { + "epoch": 0.83, + "learning_rate": 1.723888447583675e-05, + "loss": 1.072, + "step": 12970 + }, + { + "epoch": 0.83, + "learning_rate": 1.72386697868138e-05, + "loss": 0.97, + "step": 12971 + }, + { + "epoch": 0.83, + "learning_rate": 1.7238455097790853e-05, + "loss": 1.0366, + "step": 12972 + }, + { + "epoch": 0.83, + "learning_rate": 1.7238240408767902e-05, + "loss": 1.0689, + "step": 12973 + }, + { + "epoch": 0.83, + "learning_rate": 1.723802571974495e-05, + "loss": 1.1038, + "step": 12974 + }, + { + "epoch": 0.83, + "learning_rate": 1.7237811030722e-05, + "loss": 1.0274, + "step": 12975 + }, + { + "epoch": 0.83, + "learning_rate": 1.723759634169905e-05, + "loss": 1.0313, + "step": 12976 + }, + { + "epoch": 0.83, + "learning_rate": 1.72373816526761e-05, + "loss": 1.0022, + "step": 12977 + }, + { + "epoch": 0.83, + "learning_rate": 1.723716696365315e-05, + "loss": 1.1142, + "step": 12978 + }, + { + "epoch": 0.84, + "learning_rate": 1.72369522746302e-05, + "loss": 1.1184, + "step": 12979 + }, + { + "epoch": 0.84, + "learning_rate": 1.7236737585607248e-05, + "loss": 1.0689, + "step": 12980 + }, + { + "epoch": 0.84, + "learning_rate": 1.72365228965843e-05, + "loss": 1.0936, + "step": 12981 + }, + { + "epoch": 0.84, + "learning_rate": 1.723630820756135e-05, + "loss": 1.0837, + "step": 12982 + }, + { + "epoch": 0.84, + "learning_rate": 1.7236093518538398e-05, + "loss": 1.0262, + "step": 12983 + }, + { + "epoch": 0.84, + "learning_rate": 1.7235878829515447e-05, + "loss": 0.9602, + "step": 12984 + }, + { + "epoch": 0.84, + "learning_rate": 1.72356641404925e-05, + "loss": 1.1167, + "step": 12985 + }, + { + "epoch": 0.84, + "learning_rate": 1.723544945146955e-05, + "loss": 1.0453, + "step": 12986 + }, + { + "epoch": 0.84, + "learning_rate": 1.7235234762446598e-05, + "loss": 1.0913, + "step": 12987 + }, + { + "epoch": 0.84, + "learning_rate": 1.7235020073423647e-05, + "loss": 0.9906, + "step": 12988 + }, + { + "epoch": 0.84, + "learning_rate": 1.7234805384400695e-05, + "loss": 1.027, + "step": 12989 + }, + { + "epoch": 0.84, + "learning_rate": 1.7234590695377748e-05, + "loss": 1.0165, + "step": 12990 + }, + { + "epoch": 0.84, + "learning_rate": 1.7234376006354797e-05, + "loss": 1.137, + "step": 12991 + }, + { + "epoch": 0.84, + "learning_rate": 1.7234161317331846e-05, + "loss": 1.0192, + "step": 12992 + }, + { + "epoch": 0.84, + "learning_rate": 1.7233946628308895e-05, + "loss": 1.065, + "step": 12993 + }, + { + "epoch": 0.84, + "learning_rate": 1.7233731939285944e-05, + "loss": 1.0018, + "step": 12994 + }, + { + "epoch": 0.84, + "learning_rate": 1.7233517250262996e-05, + "loss": 1.0748, + "step": 12995 + }, + { + "epoch": 0.84, + "learning_rate": 1.7233302561240045e-05, + "loss": 1.0655, + "step": 12996 + }, + { + "epoch": 0.84, + "learning_rate": 1.7233087872217097e-05, + "loss": 1.0181, + "step": 12997 + }, + { + "epoch": 0.84, + "learning_rate": 1.7232873183194146e-05, + "loss": 1.0748, + "step": 12998 + }, + { + "epoch": 0.84, + "learning_rate": 1.7232658494171195e-05, + "loss": 1.0871, + "step": 12999 + }, + { + "epoch": 0.84, + "learning_rate": 1.7232443805148244e-05, + "loss": 1.0243, + "step": 13000 + }, + { + "epoch": 0.84, + "eval_laion/cider": 0.5493278763592271, + "eval_laion/meteor": 0.39185675940918585, + "eval_laion/pred_num_words": 37.41015625, + "eval_laion/sacrebleu": 16.08882057817228, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.074589729309082, + "eval_runtime": 96.5004, + "eval_samples_per_second": 5.171, + "eval_steps_per_second": 0.332, + "step": 13000 + }, + { + "epoch": 0.84, + "learning_rate": 1.7232229116125293e-05, + "loss": 1.0975, + "step": 13001 + }, + { + "epoch": 0.84, + "learning_rate": 1.7232014427102342e-05, + "loss": 1.0092, + "step": 13002 + }, + { + "epoch": 0.84, + "learning_rate": 1.7231799738079395e-05, + "loss": 1.0477, + "step": 13003 + }, + { + "epoch": 0.84, + "learning_rate": 1.7231585049056444e-05, + "loss": 1.036, + "step": 13004 + }, + { + "epoch": 0.84, + "learning_rate": 1.7231370360033492e-05, + "loss": 1.0775, + "step": 13005 + }, + { + "epoch": 0.84, + "learning_rate": 1.723115567101054e-05, + "loss": 1.0032, + "step": 13006 + }, + { + "epoch": 0.84, + "learning_rate": 1.723094098198759e-05, + "loss": 1.093, + "step": 13007 + }, + { + "epoch": 0.84, + "learning_rate": 1.7230726292964643e-05, + "loss": 1.0784, + "step": 13008 + }, + { + "epoch": 0.84, + "learning_rate": 1.7230511603941692e-05, + "loss": 0.9518, + "step": 13009 + }, + { + "epoch": 0.84, + "learning_rate": 1.7230296914918744e-05, + "loss": 1.0153, + "step": 13010 + }, + { + "epoch": 0.84, + "learning_rate": 1.7230082225895793e-05, + "loss": 1.0714, + "step": 13011 + }, + { + "epoch": 0.84, + "learning_rate": 1.7229867536872842e-05, + "loss": 1.0713, + "step": 13012 + }, + { + "epoch": 0.84, + "learning_rate": 1.722965284784989e-05, + "loss": 1.1186, + "step": 13013 + }, + { + "epoch": 0.84, + "learning_rate": 1.722943815882694e-05, + "loss": 1.1098, + "step": 13014 + }, + { + "epoch": 0.84, + "learning_rate": 1.7229223469803992e-05, + "loss": 1.07, + "step": 13015 + }, + { + "epoch": 0.84, + "learning_rate": 1.722900878078104e-05, + "loss": 1.0601, + "step": 13016 + }, + { + "epoch": 0.84, + "learning_rate": 1.722879409175809e-05, + "loss": 1.0329, + "step": 13017 + }, + { + "epoch": 0.84, + "learning_rate": 1.722857940273514e-05, + "loss": 1.0886, + "step": 13018 + }, + { + "epoch": 0.84, + "learning_rate": 1.7228364713712188e-05, + "loss": 1.023, + "step": 13019 + }, + { + "epoch": 0.84, + "learning_rate": 1.7228150024689237e-05, + "loss": 1.0808, + "step": 13020 + }, + { + "epoch": 0.84, + "learning_rate": 1.722793533566629e-05, + "loss": 1.1022, + "step": 13021 + }, + { + "epoch": 0.84, + "learning_rate": 1.722772064664334e-05, + "loss": 1.1011, + "step": 13022 + }, + { + "epoch": 0.84, + "learning_rate": 1.7227505957620387e-05, + "loss": 1.0785, + "step": 13023 + }, + { + "epoch": 0.84, + "learning_rate": 1.7227291268597436e-05, + "loss": 0.9538, + "step": 13024 + }, + { + "epoch": 0.84, + "learning_rate": 1.722707657957449e-05, + "loss": 1.0016, + "step": 13025 + }, + { + "epoch": 0.84, + "learning_rate": 1.7226861890551538e-05, + "loss": 1.0603, + "step": 13026 + }, + { + "epoch": 0.84, + "learning_rate": 1.7226647201528587e-05, + "loss": 1.1077, + "step": 13027 + }, + { + "epoch": 0.84, + "learning_rate": 1.722643251250564e-05, + "loss": 1.0637, + "step": 13028 + }, + { + "epoch": 0.84, + "learning_rate": 1.7226217823482688e-05, + "loss": 1.0289, + "step": 13029 + }, + { + "epoch": 0.84, + "learning_rate": 1.7226003134459737e-05, + "loss": 1.0467, + "step": 13030 + }, + { + "epoch": 0.84, + "learning_rate": 1.7225788445436786e-05, + "loss": 1.0946, + "step": 13031 + }, + { + "epoch": 0.84, + "learning_rate": 1.7225573756413835e-05, + "loss": 1.0326, + "step": 13032 + }, + { + "epoch": 0.84, + "learning_rate": 1.7225359067390884e-05, + "loss": 1.0721, + "step": 13033 + }, + { + "epoch": 0.84, + "learning_rate": 1.7225144378367936e-05, + "loss": 1.1411, + "step": 13034 + }, + { + "epoch": 0.84, + "learning_rate": 1.7224929689344985e-05, + "loss": 1.0576, + "step": 13035 + }, + { + "epoch": 0.84, + "learning_rate": 1.7224715000322034e-05, + "loss": 1.0265, + "step": 13036 + }, + { + "epoch": 0.84, + "learning_rate": 1.7224500311299083e-05, + "loss": 1.0544, + "step": 13037 + }, + { + "epoch": 0.84, + "learning_rate": 1.7224285622276132e-05, + "loss": 1.0125, + "step": 13038 + }, + { + "epoch": 0.84, + "learning_rate": 1.7224070933253184e-05, + "loss": 0.9958, + "step": 13039 + }, + { + "epoch": 0.84, + "learning_rate": 1.7223856244230233e-05, + "loss": 1.0334, + "step": 13040 + }, + { + "epoch": 0.84, + "learning_rate": 1.7223641555207286e-05, + "loss": 1.0387, + "step": 13041 + }, + { + "epoch": 0.84, + "learning_rate": 1.7223426866184335e-05, + "loss": 1.1458, + "step": 13042 + }, + { + "epoch": 0.84, + "learning_rate": 1.7223212177161384e-05, + "loss": 1.0106, + "step": 13043 + }, + { + "epoch": 0.84, + "learning_rate": 1.7222997488138433e-05, + "loss": 0.9068, + "step": 13044 + }, + { + "epoch": 0.84, + "learning_rate": 1.722278279911548e-05, + "loss": 1.1573, + "step": 13045 + }, + { + "epoch": 0.84, + "learning_rate": 1.7222568110092534e-05, + "loss": 1.0313, + "step": 13046 + }, + { + "epoch": 0.84, + "learning_rate": 1.7222353421069583e-05, + "loss": 1.0861, + "step": 13047 + }, + { + "epoch": 0.84, + "learning_rate": 1.7222138732046632e-05, + "loss": 1.0362, + "step": 13048 + }, + { + "epoch": 0.84, + "learning_rate": 1.722192404302368e-05, + "loss": 1.0579, + "step": 13049 + }, + { + "epoch": 0.84, + "learning_rate": 1.722170935400073e-05, + "loss": 1.0361, + "step": 13050 + }, + { + "epoch": 0.84, + "learning_rate": 1.722149466497778e-05, + "loss": 1.0334, + "step": 13051 + }, + { + "epoch": 0.84, + "learning_rate": 1.722127997595483e-05, + "loss": 1.0502, + "step": 13052 + }, + { + "epoch": 0.84, + "learning_rate": 1.722106528693188e-05, + "loss": 1.0843, + "step": 13053 + }, + { + "epoch": 0.84, + "learning_rate": 1.7220850597908932e-05, + "loss": 1.0459, + "step": 13054 + }, + { + "epoch": 0.84, + "learning_rate": 1.722063590888598e-05, + "loss": 1.0121, + "step": 13055 + }, + { + "epoch": 0.84, + "learning_rate": 1.722042121986303e-05, + "loss": 1.0822, + "step": 13056 + }, + { + "epoch": 0.84, + "learning_rate": 1.722020653084008e-05, + "loss": 1.0948, + "step": 13057 + }, + { + "epoch": 0.84, + "learning_rate": 1.7219991841817128e-05, + "loss": 1.1084, + "step": 13058 + }, + { + "epoch": 0.84, + "learning_rate": 1.721977715279418e-05, + "loss": 1.0149, + "step": 13059 + }, + { + "epoch": 0.84, + "learning_rate": 1.721956246377123e-05, + "loss": 1.1319, + "step": 13060 + }, + { + "epoch": 0.84, + "learning_rate": 1.721934777474828e-05, + "loss": 1.0395, + "step": 13061 + }, + { + "epoch": 0.84, + "learning_rate": 1.7219133085725328e-05, + "loss": 0.971, + "step": 13062 + }, + { + "epoch": 0.84, + "learning_rate": 1.7218918396702376e-05, + "loss": 0.9604, + "step": 13063 + }, + { + "epoch": 0.84, + "learning_rate": 1.721870370767943e-05, + "loss": 1.106, + "step": 13064 + }, + { + "epoch": 0.84, + "learning_rate": 1.7218489018656478e-05, + "loss": 1.0864, + "step": 13065 + }, + { + "epoch": 0.84, + "learning_rate": 1.7218274329633527e-05, + "loss": 1.0981, + "step": 13066 + }, + { + "epoch": 0.84, + "learning_rate": 1.7218059640610576e-05, + "loss": 1.0994, + "step": 13067 + }, + { + "epoch": 0.84, + "learning_rate": 1.7217844951587628e-05, + "loss": 1.0655, + "step": 13068 + }, + { + "epoch": 0.84, + "learning_rate": 1.7217630262564677e-05, + "loss": 1.0973, + "step": 13069 + }, + { + "epoch": 0.84, + "learning_rate": 1.7217415573541726e-05, + "loss": 1.042, + "step": 13070 + }, + { + "epoch": 0.84, + "learning_rate": 1.7217200884518775e-05, + "loss": 1.0237, + "step": 13071 + }, + { + "epoch": 0.84, + "learning_rate": 1.7216986195495827e-05, + "loss": 1.0925, + "step": 13072 + }, + { + "epoch": 0.84, + "learning_rate": 1.7216771506472876e-05, + "loss": 1.0084, + "step": 13073 + }, + { + "epoch": 0.84, + "learning_rate": 1.7216556817449925e-05, + "loss": 1.1105, + "step": 13074 + }, + { + "epoch": 0.84, + "learning_rate": 1.7216342128426974e-05, + "loss": 1.0489, + "step": 13075 + }, + { + "epoch": 0.84, + "learning_rate": 1.7216127439404023e-05, + "loss": 0.9937, + "step": 13076 + }, + { + "epoch": 0.84, + "learning_rate": 1.7215912750381076e-05, + "loss": 1.083, + "step": 13077 + }, + { + "epoch": 0.84, + "learning_rate": 1.7215698061358125e-05, + "loss": 1.0138, + "step": 13078 + }, + { + "epoch": 0.84, + "learning_rate": 1.7215483372335173e-05, + "loss": 1.0594, + "step": 13079 + }, + { + "epoch": 0.84, + "learning_rate": 1.7215268683312222e-05, + "loss": 1.123, + "step": 13080 + }, + { + "epoch": 0.84, + "learning_rate": 1.721505399428927e-05, + "loss": 1.0396, + "step": 13081 + }, + { + "epoch": 0.84, + "learning_rate": 1.7214839305266324e-05, + "loss": 1.1064, + "step": 13082 + }, + { + "epoch": 0.84, + "learning_rate": 1.7214624616243373e-05, + "loss": 1.051, + "step": 13083 + }, + { + "epoch": 0.84, + "learning_rate": 1.7214409927220425e-05, + "loss": 1.0775, + "step": 13084 + }, + { + "epoch": 0.84, + "learning_rate": 1.7214195238197474e-05, + "loss": 1.1277, + "step": 13085 + }, + { + "epoch": 0.84, + "learning_rate": 1.7213980549174523e-05, + "loss": 1.0212, + "step": 13086 + }, + { + "epoch": 0.84, + "learning_rate": 1.7213765860151572e-05, + "loss": 1.0331, + "step": 13087 + }, + { + "epoch": 0.84, + "learning_rate": 1.721355117112862e-05, + "loss": 0.9674, + "step": 13088 + }, + { + "epoch": 0.84, + "learning_rate": 1.721333648210567e-05, + "loss": 1.0235, + "step": 13089 + }, + { + "epoch": 0.84, + "learning_rate": 1.7213121793082722e-05, + "loss": 0.9376, + "step": 13090 + }, + { + "epoch": 0.84, + "learning_rate": 1.721290710405977e-05, + "loss": 1.0216, + "step": 13091 + }, + { + "epoch": 0.84, + "learning_rate": 1.721269241503682e-05, + "loss": 1.0562, + "step": 13092 + }, + { + "epoch": 0.84, + "learning_rate": 1.721247772601387e-05, + "loss": 0.9765, + "step": 13093 + }, + { + "epoch": 0.84, + "learning_rate": 1.7212263036990918e-05, + "loss": 0.9917, + "step": 13094 + }, + { + "epoch": 0.84, + "learning_rate": 1.721204834796797e-05, + "loss": 1.0991, + "step": 13095 + }, + { + "epoch": 0.84, + "learning_rate": 1.721183365894502e-05, + "loss": 1.1248, + "step": 13096 + }, + { + "epoch": 0.84, + "learning_rate": 1.7211618969922072e-05, + "loss": 1.0755, + "step": 13097 + }, + { + "epoch": 0.84, + "learning_rate": 1.721140428089912e-05, + "loss": 1.0355, + "step": 13098 + }, + { + "epoch": 0.84, + "learning_rate": 1.721118959187617e-05, + "loss": 0.9973, + "step": 13099 + }, + { + "epoch": 0.84, + "learning_rate": 1.721097490285322e-05, + "loss": 1.1495, + "step": 13100 + }, + { + "epoch": 0.84, + "learning_rate": 1.7210760213830268e-05, + "loss": 1.0802, + "step": 13101 + }, + { + "epoch": 0.84, + "learning_rate": 1.721054552480732e-05, + "loss": 1.1554, + "step": 13102 + }, + { + "epoch": 0.84, + "learning_rate": 1.721033083578437e-05, + "loss": 0.9942, + "step": 13103 + }, + { + "epoch": 0.84, + "learning_rate": 1.7210116146761418e-05, + "loss": 1.1667, + "step": 13104 + }, + { + "epoch": 0.84, + "learning_rate": 1.7209901457738467e-05, + "loss": 1.0816, + "step": 13105 + }, + { + "epoch": 0.84, + "learning_rate": 1.7209686768715516e-05, + "loss": 1.0603, + "step": 13106 + }, + { + "epoch": 0.84, + "learning_rate": 1.7209472079692565e-05, + "loss": 1.0487, + "step": 13107 + }, + { + "epoch": 0.84, + "learning_rate": 1.7209257390669617e-05, + "loss": 1.0095, + "step": 13108 + }, + { + "epoch": 0.84, + "learning_rate": 1.7209042701646666e-05, + "loss": 1.1763, + "step": 13109 + }, + { + "epoch": 0.84, + "learning_rate": 1.7208828012623715e-05, + "loss": 1.015, + "step": 13110 + }, + { + "epoch": 0.84, + "learning_rate": 1.7208613323600764e-05, + "loss": 1.0492, + "step": 13111 + }, + { + "epoch": 0.84, + "learning_rate": 1.7208398634577816e-05, + "loss": 1.0208, + "step": 13112 + }, + { + "epoch": 0.84, + "learning_rate": 1.7208183945554865e-05, + "loss": 1.0364, + "step": 13113 + }, + { + "epoch": 0.84, + "learning_rate": 1.7207969256531914e-05, + "loss": 1.1418, + "step": 13114 + }, + { + "epoch": 0.84, + "learning_rate": 1.7207754567508967e-05, + "loss": 1.022, + "step": 13115 + }, + { + "epoch": 0.84, + "learning_rate": 1.7207539878486016e-05, + "loss": 1.0454, + "step": 13116 + }, + { + "epoch": 0.84, + "learning_rate": 1.7207325189463065e-05, + "loss": 1.0653, + "step": 13117 + }, + { + "epoch": 0.84, + "learning_rate": 1.7207110500440114e-05, + "loss": 1.0946, + "step": 13118 + }, + { + "epoch": 0.84, + "learning_rate": 1.7206895811417163e-05, + "loss": 1.0266, + "step": 13119 + }, + { + "epoch": 0.84, + "learning_rate": 1.720668112239421e-05, + "loss": 1.1578, + "step": 13120 + }, + { + "epoch": 0.84, + "learning_rate": 1.7206466433371264e-05, + "loss": 1.0387, + "step": 13121 + }, + { + "epoch": 0.84, + "learning_rate": 1.7206251744348313e-05, + "loss": 1.0082, + "step": 13122 + }, + { + "epoch": 0.84, + "learning_rate": 1.7206037055325362e-05, + "loss": 1.0686, + "step": 13123 + }, + { + "epoch": 0.84, + "learning_rate": 1.720582236630241e-05, + "loss": 0.9871, + "step": 13124 + }, + { + "epoch": 0.84, + "learning_rate": 1.720560767727946e-05, + "loss": 1.0283, + "step": 13125 + }, + { + "epoch": 0.84, + "learning_rate": 1.7205392988256512e-05, + "loss": 1.1237, + "step": 13126 + }, + { + "epoch": 0.84, + "learning_rate": 1.720517829923356e-05, + "loss": 1.0025, + "step": 13127 + }, + { + "epoch": 0.84, + "learning_rate": 1.7204963610210613e-05, + "loss": 1.0942, + "step": 13128 + }, + { + "epoch": 0.84, + "learning_rate": 1.7204748921187662e-05, + "loss": 1.0246, + "step": 13129 + }, + { + "epoch": 0.84, + "learning_rate": 1.720453423216471e-05, + "loss": 0.9741, + "step": 13130 + }, + { + "epoch": 0.84, + "learning_rate": 1.720431954314176e-05, + "loss": 1.0391, + "step": 13131 + }, + { + "epoch": 0.84, + "learning_rate": 1.720410485411881e-05, + "loss": 1.0984, + "step": 13132 + }, + { + "epoch": 0.84, + "learning_rate": 1.720389016509586e-05, + "loss": 1.0952, + "step": 13133 + }, + { + "epoch": 0.85, + "learning_rate": 1.720367547607291e-05, + "loss": 1.0619, + "step": 13134 + }, + { + "epoch": 0.85, + "learning_rate": 1.720346078704996e-05, + "loss": 1.0369, + "step": 13135 + }, + { + "epoch": 0.85, + "learning_rate": 1.720324609802701e-05, + "loss": 1.0611, + "step": 13136 + }, + { + "epoch": 0.85, + "learning_rate": 1.7203031409004057e-05, + "loss": 1.0729, + "step": 13137 + }, + { + "epoch": 0.85, + "learning_rate": 1.7202816719981106e-05, + "loss": 1.0469, + "step": 13138 + }, + { + "epoch": 0.85, + "learning_rate": 1.720260203095816e-05, + "loss": 0.9785, + "step": 13139 + }, + { + "epoch": 0.85, + "learning_rate": 1.7202387341935208e-05, + "loss": 1.0133, + "step": 13140 + }, + { + "epoch": 0.85, + "learning_rate": 1.720217265291226e-05, + "loss": 1.0737, + "step": 13141 + }, + { + "epoch": 0.85, + "learning_rate": 1.720195796388931e-05, + "loss": 1.1014, + "step": 13142 + }, + { + "epoch": 0.85, + "learning_rate": 1.7201743274866358e-05, + "loss": 1.0967, + "step": 13143 + }, + { + "epoch": 0.85, + "learning_rate": 1.7201528585843407e-05, + "loss": 1.0547, + "step": 13144 + }, + { + "epoch": 0.85, + "learning_rate": 1.7201313896820456e-05, + "loss": 1.065, + "step": 13145 + }, + { + "epoch": 0.85, + "learning_rate": 1.720109920779751e-05, + "loss": 1.013, + "step": 13146 + }, + { + "epoch": 0.85, + "learning_rate": 1.7200884518774557e-05, + "loss": 1.0362, + "step": 13147 + }, + { + "epoch": 0.85, + "learning_rate": 1.7200669829751606e-05, + "loss": 1.0108, + "step": 13148 + }, + { + "epoch": 0.85, + "learning_rate": 1.7200455140728655e-05, + "loss": 0.993, + "step": 13149 + }, + { + "epoch": 0.85, + "learning_rate": 1.7200240451705704e-05, + "loss": 1.1565, + "step": 13150 + }, + { + "epoch": 0.85, + "learning_rate": 1.7200025762682757e-05, + "loss": 1.0196, + "step": 13151 + }, + { + "epoch": 0.85, + "learning_rate": 1.7199811073659806e-05, + "loss": 1.0234, + "step": 13152 + }, + { + "epoch": 0.85, + "learning_rate": 1.7199596384636854e-05, + "loss": 1.1065, + "step": 13153 + }, + { + "epoch": 0.85, + "learning_rate": 1.7199381695613903e-05, + "loss": 1.0975, + "step": 13154 + }, + { + "epoch": 0.85, + "learning_rate": 1.7199167006590956e-05, + "loss": 1.0425, + "step": 13155 + }, + { + "epoch": 0.85, + "learning_rate": 1.7198952317568005e-05, + "loss": 1.1211, + "step": 13156 + }, + { + "epoch": 0.85, + "learning_rate": 1.7198737628545054e-05, + "loss": 1.0492, + "step": 13157 + }, + { + "epoch": 0.85, + "learning_rate": 1.7198522939522103e-05, + "loss": 1.0376, + "step": 13158 + }, + { + "epoch": 0.85, + "learning_rate": 1.7198308250499155e-05, + "loss": 0.983, + "step": 13159 + }, + { + "epoch": 0.85, + "learning_rate": 1.7198093561476204e-05, + "loss": 0.9987, + "step": 13160 + }, + { + "epoch": 0.85, + "learning_rate": 1.7197878872453253e-05, + "loss": 1.065, + "step": 13161 + }, + { + "epoch": 0.85, + "learning_rate": 1.7197664183430302e-05, + "loss": 1.0257, + "step": 13162 + }, + { + "epoch": 0.85, + "learning_rate": 1.719744949440735e-05, + "loss": 1.0328, + "step": 13163 + }, + { + "epoch": 0.85, + "learning_rate": 1.7197234805384403e-05, + "loss": 0.9711, + "step": 13164 + }, + { + "epoch": 0.85, + "learning_rate": 1.7197020116361452e-05, + "loss": 1.0679, + "step": 13165 + }, + { + "epoch": 0.85, + "learning_rate": 1.71968054273385e-05, + "loss": 1.0338, + "step": 13166 + }, + { + "epoch": 0.85, + "learning_rate": 1.719659073831555e-05, + "loss": 1.0401, + "step": 13167 + }, + { + "epoch": 0.85, + "learning_rate": 1.71963760492926e-05, + "loss": 1.0004, + "step": 13168 + }, + { + "epoch": 0.85, + "learning_rate": 1.7196161360269648e-05, + "loss": 1.1379, + "step": 13169 + }, + { + "epoch": 0.85, + "learning_rate": 1.71959466712467e-05, + "loss": 1.0912, + "step": 13170 + }, + { + "epoch": 0.85, + "learning_rate": 1.7195731982223753e-05, + "loss": 1.0722, + "step": 13171 + }, + { + "epoch": 0.85, + "learning_rate": 1.7195517293200802e-05, + "loss": 1.0004, + "step": 13172 + }, + { + "epoch": 0.85, + "learning_rate": 1.719530260417785e-05, + "loss": 1.0347, + "step": 13173 + }, + { + "epoch": 0.85, + "learning_rate": 1.71950879151549e-05, + "loss": 1.0125, + "step": 13174 + }, + { + "epoch": 0.85, + "learning_rate": 1.719487322613195e-05, + "loss": 1.0228, + "step": 13175 + }, + { + "epoch": 0.85, + "learning_rate": 1.7194658537108998e-05, + "loss": 1.1128, + "step": 13176 + }, + { + "epoch": 0.85, + "learning_rate": 1.719444384808605e-05, + "loss": 1.0021, + "step": 13177 + }, + { + "epoch": 0.85, + "learning_rate": 1.71942291590631e-05, + "loss": 1.0088, + "step": 13178 + }, + { + "epoch": 0.85, + "learning_rate": 1.7194014470040148e-05, + "loss": 1.0302, + "step": 13179 + }, + { + "epoch": 0.85, + "learning_rate": 1.7193799781017197e-05, + "loss": 1.0365, + "step": 13180 + }, + { + "epoch": 0.85, + "learning_rate": 1.7193585091994246e-05, + "loss": 1.0686, + "step": 13181 + }, + { + "epoch": 0.85, + "learning_rate": 1.7193370402971298e-05, + "loss": 1.0723, + "step": 13182 + }, + { + "epoch": 0.85, + "learning_rate": 1.7193155713948347e-05, + "loss": 1.0684, + "step": 13183 + }, + { + "epoch": 0.85, + "learning_rate": 1.71929410249254e-05, + "loss": 1.0811, + "step": 13184 + }, + { + "epoch": 0.85, + "learning_rate": 1.719272633590245e-05, + "loss": 1.0466, + "step": 13185 + }, + { + "epoch": 0.85, + "learning_rate": 1.7192511646879497e-05, + "loss": 1.0248, + "step": 13186 + }, + { + "epoch": 0.85, + "learning_rate": 1.7192296957856546e-05, + "loss": 1.0699, + "step": 13187 + }, + { + "epoch": 0.85, + "learning_rate": 1.7192082268833595e-05, + "loss": 1.022, + "step": 13188 + }, + { + "epoch": 0.85, + "learning_rate": 1.7191867579810648e-05, + "loss": 1.0669, + "step": 13189 + }, + { + "epoch": 0.85, + "learning_rate": 1.7191652890787697e-05, + "loss": 1.0313, + "step": 13190 + }, + { + "epoch": 0.85, + "learning_rate": 1.7191438201764746e-05, + "loss": 1.195, + "step": 13191 + }, + { + "epoch": 0.85, + "learning_rate": 1.7191223512741795e-05, + "loss": 0.9555, + "step": 13192 + }, + { + "epoch": 0.85, + "learning_rate": 1.7191008823718844e-05, + "loss": 1.0273, + "step": 13193 + }, + { + "epoch": 0.85, + "learning_rate": 1.7190794134695893e-05, + "loss": 1.0418, + "step": 13194 + }, + { + "epoch": 0.85, + "learning_rate": 1.7190579445672945e-05, + "loss": 1.1015, + "step": 13195 + }, + { + "epoch": 0.85, + "learning_rate": 1.7190364756649994e-05, + "loss": 1.0584, + "step": 13196 + }, + { + "epoch": 0.85, + "learning_rate": 1.7190150067627043e-05, + "loss": 0.9672, + "step": 13197 + }, + { + "epoch": 0.85, + "learning_rate": 1.7189935378604092e-05, + "loss": 1.0508, + "step": 13198 + }, + { + "epoch": 0.85, + "learning_rate": 1.7189720689581144e-05, + "loss": 1.0365, + "step": 13199 + }, + { + "epoch": 0.85, + "learning_rate": 1.7189506000558193e-05, + "loss": 1.1681, + "step": 13200 + }, + { + "epoch": 0.85, + "learning_rate": 1.7189291311535242e-05, + "loss": 1.0905, + "step": 13201 + }, + { + "epoch": 0.85, + "learning_rate": 1.7189076622512294e-05, + "loss": 1.0342, + "step": 13202 + }, + { + "epoch": 0.85, + "learning_rate": 1.7188861933489343e-05, + "loss": 1.0783, + "step": 13203 + }, + { + "epoch": 0.85, + "learning_rate": 1.7188647244466392e-05, + "loss": 1.0381, + "step": 13204 + }, + { + "epoch": 0.85, + "learning_rate": 1.718843255544344e-05, + "loss": 1.0449, + "step": 13205 + }, + { + "epoch": 0.85, + "learning_rate": 1.718821786642049e-05, + "loss": 1.1162, + "step": 13206 + }, + { + "epoch": 0.85, + "learning_rate": 1.718800317739754e-05, + "loss": 1.1022, + "step": 13207 + }, + { + "epoch": 0.85, + "learning_rate": 1.718778848837459e-05, + "loss": 1.0694, + "step": 13208 + }, + { + "epoch": 0.85, + "learning_rate": 1.718757379935164e-05, + "loss": 1.0596, + "step": 13209 + }, + { + "epoch": 0.85, + "learning_rate": 1.718735911032869e-05, + "loss": 1.0935, + "step": 13210 + }, + { + "epoch": 0.85, + "learning_rate": 1.718714442130574e-05, + "loss": 1.0937, + "step": 13211 + }, + { + "epoch": 0.85, + "learning_rate": 1.7186929732282787e-05, + "loss": 1.085, + "step": 13212 + }, + { + "epoch": 0.85, + "learning_rate": 1.718671504325984e-05, + "loss": 1.0868, + "step": 13213 + }, + { + "epoch": 0.85, + "learning_rate": 1.718650035423689e-05, + "loss": 1.0688, + "step": 13214 + }, + { + "epoch": 0.85, + "learning_rate": 1.718628566521394e-05, + "loss": 1.0005, + "step": 13215 + }, + { + "epoch": 0.85, + "learning_rate": 1.718607097619099e-05, + "loss": 1.0764, + "step": 13216 + }, + { + "epoch": 0.85, + "learning_rate": 1.718585628716804e-05, + "loss": 1.0512, + "step": 13217 + }, + { + "epoch": 0.85, + "learning_rate": 1.7185641598145088e-05, + "loss": 1.0302, + "step": 13218 + }, + { + "epoch": 0.85, + "learning_rate": 1.7185426909122137e-05, + "loss": 1.0225, + "step": 13219 + }, + { + "epoch": 0.85, + "learning_rate": 1.718521222009919e-05, + "loss": 1.0, + "step": 13220 + }, + { + "epoch": 0.85, + "learning_rate": 1.7184997531076238e-05, + "loss": 1.0792, + "step": 13221 + }, + { + "epoch": 0.85, + "learning_rate": 1.7184782842053287e-05, + "loss": 1.059, + "step": 13222 + }, + { + "epoch": 0.85, + "learning_rate": 1.7184568153030336e-05, + "loss": 1.0453, + "step": 13223 + }, + { + "epoch": 0.85, + "learning_rate": 1.7184353464007385e-05, + "loss": 1.1246, + "step": 13224 + }, + { + "epoch": 0.85, + "learning_rate": 1.7184138774984434e-05, + "loss": 1.0474, + "step": 13225 + }, + { + "epoch": 0.85, + "learning_rate": 1.7183924085961487e-05, + "loss": 1.0022, + "step": 13226 + }, + { + "epoch": 0.85, + "learning_rate": 1.7183709396938535e-05, + "loss": 1.0682, + "step": 13227 + }, + { + "epoch": 0.85, + "learning_rate": 1.7183494707915588e-05, + "loss": 1.1334, + "step": 13228 + }, + { + "epoch": 0.85, + "learning_rate": 1.7183280018892637e-05, + "loss": 1.0475, + "step": 13229 + }, + { + "epoch": 0.85, + "learning_rate": 1.7183065329869686e-05, + "loss": 1.0604, + "step": 13230 + }, + { + "epoch": 0.85, + "learning_rate": 1.7182850640846735e-05, + "loss": 1.0268, + "step": 13231 + }, + { + "epoch": 0.85, + "learning_rate": 1.7182635951823784e-05, + "loss": 1.0893, + "step": 13232 + }, + { + "epoch": 0.85, + "learning_rate": 1.7182421262800836e-05, + "loss": 1.054, + "step": 13233 + }, + { + "epoch": 0.85, + "learning_rate": 1.7182206573777885e-05, + "loss": 1.1049, + "step": 13234 + }, + { + "epoch": 0.85, + "learning_rate": 1.7181991884754934e-05, + "loss": 0.9891, + "step": 13235 + }, + { + "epoch": 0.85, + "learning_rate": 1.7181777195731983e-05, + "loss": 1.0679, + "step": 13236 + }, + { + "epoch": 0.85, + "learning_rate": 1.7181562506709032e-05, + "loss": 0.9664, + "step": 13237 + }, + { + "epoch": 0.85, + "learning_rate": 1.7181347817686084e-05, + "loss": 1.0846, + "step": 13238 + }, + { + "epoch": 0.85, + "learning_rate": 1.7181133128663133e-05, + "loss": 1.0423, + "step": 13239 + }, + { + "epoch": 0.85, + "learning_rate": 1.7180918439640182e-05, + "loss": 1.0857, + "step": 13240 + }, + { + "epoch": 0.85, + "learning_rate": 1.718070375061723e-05, + "loss": 1.0595, + "step": 13241 + }, + { + "epoch": 0.85, + "learning_rate": 1.718048906159428e-05, + "loss": 0.9824, + "step": 13242 + }, + { + "epoch": 0.85, + "learning_rate": 1.7180274372571332e-05, + "loss": 1.0598, + "step": 13243 + }, + { + "epoch": 0.85, + "learning_rate": 1.718005968354838e-05, + "loss": 1.088, + "step": 13244 + }, + { + "epoch": 0.85, + "learning_rate": 1.717984499452543e-05, + "loss": 1.1494, + "step": 13245 + }, + { + "epoch": 0.85, + "learning_rate": 1.7179630305502483e-05, + "loss": 1.0595, + "step": 13246 + }, + { + "epoch": 0.85, + "learning_rate": 1.7179415616479532e-05, + "loss": 1.0597, + "step": 13247 + }, + { + "epoch": 0.85, + "learning_rate": 1.717920092745658e-05, + "loss": 1.0493, + "step": 13248 + }, + { + "epoch": 0.85, + "learning_rate": 1.717898623843363e-05, + "loss": 1.029, + "step": 13249 + }, + { + "epoch": 0.85, + "learning_rate": 1.717877154941068e-05, + "loss": 1.1222, + "step": 13250 + }, + { + "epoch": 0.85, + "learning_rate": 1.717855686038773e-05, + "loss": 1.0698, + "step": 13251 + }, + { + "epoch": 0.85, + "learning_rate": 1.717834217136478e-05, + "loss": 1.0625, + "step": 13252 + }, + { + "epoch": 0.85, + "learning_rate": 1.717812748234183e-05, + "loss": 1.1305, + "step": 13253 + }, + { + "epoch": 0.85, + "learning_rate": 1.7177912793318878e-05, + "loss": 1.0162, + "step": 13254 + }, + { + "epoch": 0.85, + "learning_rate": 1.7177698104295927e-05, + "loss": 1.1206, + "step": 13255 + }, + { + "epoch": 0.85, + "learning_rate": 1.7177483415272976e-05, + "loss": 1.0611, + "step": 13256 + }, + { + "epoch": 0.85, + "learning_rate": 1.7177268726250028e-05, + "loss": 1.1605, + "step": 13257 + }, + { + "epoch": 0.85, + "learning_rate": 1.717705403722708e-05, + "loss": 1.0074, + "step": 13258 + }, + { + "epoch": 0.85, + "learning_rate": 1.717683934820413e-05, + "loss": 1.0999, + "step": 13259 + }, + { + "epoch": 0.85, + "learning_rate": 1.717662465918118e-05, + "loss": 0.9815, + "step": 13260 + }, + { + "epoch": 0.85, + "learning_rate": 1.7176409970158227e-05, + "loss": 1.0148, + "step": 13261 + }, + { + "epoch": 0.85, + "learning_rate": 1.7176195281135276e-05, + "loss": 1.0696, + "step": 13262 + }, + { + "epoch": 0.85, + "learning_rate": 1.7175980592112325e-05, + "loss": 1.0274, + "step": 13263 + }, + { + "epoch": 0.85, + "learning_rate": 1.7175765903089378e-05, + "loss": 1.0069, + "step": 13264 + }, + { + "epoch": 0.85, + "learning_rate": 1.7175551214066427e-05, + "loss": 1.1253, + "step": 13265 + }, + { + "epoch": 0.85, + "learning_rate": 1.7175336525043476e-05, + "loss": 1.1874, + "step": 13266 + }, + { + "epoch": 0.85, + "learning_rate": 1.7175121836020525e-05, + "loss": 1.0835, + "step": 13267 + }, + { + "epoch": 0.85, + "learning_rate": 1.7174907146997574e-05, + "loss": 1.108, + "step": 13268 + }, + { + "epoch": 0.85, + "learning_rate": 1.7174692457974626e-05, + "loss": 0.9453, + "step": 13269 + }, + { + "epoch": 0.85, + "learning_rate": 1.7174477768951675e-05, + "loss": 1.0629, + "step": 13270 + }, + { + "epoch": 0.85, + "learning_rate": 1.7174263079928724e-05, + "loss": 1.098, + "step": 13271 + }, + { + "epoch": 0.85, + "learning_rate": 1.7174048390905776e-05, + "loss": 0.9938, + "step": 13272 + }, + { + "epoch": 0.85, + "learning_rate": 1.7173833701882825e-05, + "loss": 1.099, + "step": 13273 + }, + { + "epoch": 0.85, + "learning_rate": 1.7173619012859874e-05, + "loss": 1.0572, + "step": 13274 + }, + { + "epoch": 0.85, + "learning_rate": 1.7173404323836923e-05, + "loss": 1.2378, + "step": 13275 + }, + { + "epoch": 0.85, + "learning_rate": 1.7173189634813972e-05, + "loss": 1.0575, + "step": 13276 + }, + { + "epoch": 0.85, + "learning_rate": 1.7172974945791024e-05, + "loss": 1.0197, + "step": 13277 + }, + { + "epoch": 0.85, + "learning_rate": 1.7172760256768073e-05, + "loss": 1.0877, + "step": 13278 + }, + { + "epoch": 0.85, + "learning_rate": 1.7172545567745122e-05, + "loss": 1.0894, + "step": 13279 + }, + { + "epoch": 0.85, + "learning_rate": 1.717233087872217e-05, + "loss": 1.1377, + "step": 13280 + }, + { + "epoch": 0.85, + "learning_rate": 1.717211618969922e-05, + "loss": 1.0027, + "step": 13281 + }, + { + "epoch": 0.85, + "learning_rate": 1.7171901500676273e-05, + "loss": 1.0797, + "step": 13282 + }, + { + "epoch": 0.85, + "learning_rate": 1.717168681165332e-05, + "loss": 1.0574, + "step": 13283 + }, + { + "epoch": 0.85, + "learning_rate": 1.717147212263037e-05, + "loss": 1.0471, + "step": 13284 + }, + { + "epoch": 0.85, + "learning_rate": 1.717125743360742e-05, + "loss": 1.0906, + "step": 13285 + }, + { + "epoch": 0.85, + "learning_rate": 1.7171042744584472e-05, + "loss": 1.1186, + "step": 13286 + }, + { + "epoch": 0.85, + "learning_rate": 1.717082805556152e-05, + "loss": 1.0485, + "step": 13287 + }, + { + "epoch": 0.85, + "learning_rate": 1.717061336653857e-05, + "loss": 1.0373, + "step": 13288 + }, + { + "epoch": 0.85, + "learning_rate": 1.7170398677515622e-05, + "loss": 1.1605, + "step": 13289 + }, + { + "epoch": 0.86, + "learning_rate": 1.717018398849267e-05, + "loss": 1.0485, + "step": 13290 + }, + { + "epoch": 0.86, + "learning_rate": 1.716996929946972e-05, + "loss": 1.0325, + "step": 13291 + }, + { + "epoch": 0.86, + "learning_rate": 1.716975461044677e-05, + "loss": 1.0275, + "step": 13292 + }, + { + "epoch": 0.86, + "learning_rate": 1.7169539921423818e-05, + "loss": 1.0594, + "step": 13293 + }, + { + "epoch": 0.86, + "learning_rate": 1.7169325232400867e-05, + "loss": 1.1022, + "step": 13294 + }, + { + "epoch": 0.86, + "learning_rate": 1.716911054337792e-05, + "loss": 1.0852, + "step": 13295 + }, + { + "epoch": 0.86, + "learning_rate": 1.7168895854354968e-05, + "loss": 1.0528, + "step": 13296 + }, + { + "epoch": 0.86, + "learning_rate": 1.7168681165332017e-05, + "loss": 1.0645, + "step": 13297 + }, + { + "epoch": 0.86, + "learning_rate": 1.7168466476309066e-05, + "loss": 1.0462, + "step": 13298 + }, + { + "epoch": 0.86, + "learning_rate": 1.7168251787286115e-05, + "loss": 1.0933, + "step": 13299 + }, + { + "epoch": 0.86, + "learning_rate": 1.7168037098263168e-05, + "loss": 1.0508, + "step": 13300 + }, + { + "epoch": 0.86, + "learning_rate": 1.7167822409240216e-05, + "loss": 1.0568, + "step": 13301 + }, + { + "epoch": 0.86, + "learning_rate": 1.716760772021727e-05, + "loss": 1.0373, + "step": 13302 + }, + { + "epoch": 0.86, + "learning_rate": 1.7167393031194318e-05, + "loss": 1.1663, + "step": 13303 + }, + { + "epoch": 0.86, + "learning_rate": 1.7167178342171367e-05, + "loss": 1.0394, + "step": 13304 + }, + { + "epoch": 0.86, + "learning_rate": 1.7166963653148416e-05, + "loss": 1.071, + "step": 13305 + }, + { + "epoch": 0.86, + "learning_rate": 1.7166748964125465e-05, + "loss": 1.0524, + "step": 13306 + }, + { + "epoch": 0.86, + "learning_rate": 1.7166534275102517e-05, + "loss": 1.0532, + "step": 13307 + }, + { + "epoch": 0.86, + "learning_rate": 1.7166319586079566e-05, + "loss": 1.1129, + "step": 13308 + }, + { + "epoch": 0.86, + "learning_rate": 1.7166104897056615e-05, + "loss": 1.0385, + "step": 13309 + }, + { + "epoch": 0.86, + "learning_rate": 1.7165890208033664e-05, + "loss": 1.0516, + "step": 13310 + }, + { + "epoch": 0.86, + "learning_rate": 1.7165675519010713e-05, + "loss": 1.0061, + "step": 13311 + }, + { + "epoch": 0.86, + "learning_rate": 1.7165460829987762e-05, + "loss": 1.1074, + "step": 13312 + }, + { + "epoch": 0.86, + "learning_rate": 1.7165246140964814e-05, + "loss": 1.1049, + "step": 13313 + }, + { + "epoch": 0.86, + "learning_rate": 1.7165031451941863e-05, + "loss": 1.0402, + "step": 13314 + }, + { + "epoch": 0.86, + "learning_rate": 1.7164816762918916e-05, + "loss": 1.0696, + "step": 13315 + }, + { + "epoch": 0.86, + "learning_rate": 1.7164602073895965e-05, + "loss": 0.9451, + "step": 13316 + }, + { + "epoch": 0.86, + "learning_rate": 1.7164387384873013e-05, + "loss": 1.066, + "step": 13317 + }, + { + "epoch": 0.86, + "learning_rate": 1.7164172695850062e-05, + "loss": 1.0275, + "step": 13318 + }, + { + "epoch": 0.86, + "learning_rate": 1.716395800682711e-05, + "loss": 1.1188, + "step": 13319 + }, + { + "epoch": 0.86, + "learning_rate": 1.7163743317804164e-05, + "loss": 1.0948, + "step": 13320 + }, + { + "epoch": 0.86, + "learning_rate": 1.7163528628781213e-05, + "loss": 1.0519, + "step": 13321 + }, + { + "epoch": 0.86, + "learning_rate": 1.716331393975826e-05, + "loss": 1.0804, + "step": 13322 + }, + { + "epoch": 0.86, + "learning_rate": 1.716309925073531e-05, + "loss": 1.1553, + "step": 13323 + }, + { + "epoch": 0.86, + "learning_rate": 1.716288456171236e-05, + "loss": 1.021, + "step": 13324 + }, + { + "epoch": 0.86, + "learning_rate": 1.7162669872689412e-05, + "loss": 1.1655, + "step": 13325 + }, + { + "epoch": 0.86, + "learning_rate": 1.716245518366646e-05, + "loss": 0.994, + "step": 13326 + }, + { + "epoch": 0.86, + "learning_rate": 1.716224049464351e-05, + "loss": 1.0919, + "step": 13327 + }, + { + "epoch": 0.86, + "learning_rate": 1.716202580562056e-05, + "loss": 1.0869, + "step": 13328 + }, + { + "epoch": 0.86, + "learning_rate": 1.7161811116597608e-05, + "loss": 1.0722, + "step": 13329 + }, + { + "epoch": 0.86, + "learning_rate": 1.716159642757466e-05, + "loss": 1.0259, + "step": 13330 + }, + { + "epoch": 0.86, + "learning_rate": 1.716138173855171e-05, + "loss": 1.0804, + "step": 13331 + }, + { + "epoch": 0.86, + "learning_rate": 1.7161167049528758e-05, + "loss": 0.9716, + "step": 13332 + }, + { + "epoch": 0.86, + "learning_rate": 1.716095236050581e-05, + "loss": 1.1797, + "step": 13333 + }, + { + "epoch": 0.86, + "learning_rate": 1.716073767148286e-05, + "loss": 1.0552, + "step": 13334 + }, + { + "epoch": 0.86, + "learning_rate": 1.716052298245991e-05, + "loss": 1.1023, + "step": 13335 + }, + { + "epoch": 0.86, + "learning_rate": 1.7160308293436957e-05, + "loss": 1.0831, + "step": 13336 + }, + { + "epoch": 0.86, + "learning_rate": 1.7160093604414006e-05, + "loss": 1.0868, + "step": 13337 + }, + { + "epoch": 0.86, + "learning_rate": 1.715987891539106e-05, + "loss": 1.0959, + "step": 13338 + }, + { + "epoch": 0.86, + "learning_rate": 1.7159664226368108e-05, + "loss": 1.0053, + "step": 13339 + }, + { + "epoch": 0.86, + "learning_rate": 1.7159449537345157e-05, + "loss": 1.0483, + "step": 13340 + }, + { + "epoch": 0.86, + "learning_rate": 1.7159234848322206e-05, + "loss": 1.1072, + "step": 13341 + }, + { + "epoch": 0.86, + "learning_rate": 1.7159020159299255e-05, + "loss": 1.1207, + "step": 13342 + }, + { + "epoch": 0.86, + "learning_rate": 1.7158805470276304e-05, + "loss": 1.0453, + "step": 13343 + }, + { + "epoch": 0.86, + "learning_rate": 1.7158590781253356e-05, + "loss": 1.0624, + "step": 13344 + }, + { + "epoch": 0.86, + "learning_rate": 1.7158376092230408e-05, + "loss": 1.118, + "step": 13345 + }, + { + "epoch": 0.86, + "learning_rate": 1.7158161403207457e-05, + "loss": 1.0099, + "step": 13346 + }, + { + "epoch": 0.86, + "learning_rate": 1.7157946714184506e-05, + "loss": 1.0392, + "step": 13347 + }, + { + "epoch": 0.86, + "learning_rate": 1.7157732025161555e-05, + "loss": 1.1028, + "step": 13348 + }, + { + "epoch": 0.86, + "learning_rate": 1.7157517336138604e-05, + "loss": 1.1171, + "step": 13349 + }, + { + "epoch": 0.86, + "learning_rate": 1.7157302647115653e-05, + "loss": 1.0268, + "step": 13350 + }, + { + "epoch": 0.86, + "learning_rate": 1.7157087958092705e-05, + "loss": 1.025, + "step": 13351 + }, + { + "epoch": 0.86, + "learning_rate": 1.7156873269069754e-05, + "loss": 1.0448, + "step": 13352 + }, + { + "epoch": 0.86, + "learning_rate": 1.7156658580046803e-05, + "loss": 1.0308, + "step": 13353 + }, + { + "epoch": 0.86, + "learning_rate": 1.7156443891023852e-05, + "loss": 1.1057, + "step": 13354 + }, + { + "epoch": 0.86, + "learning_rate": 1.71562292020009e-05, + "loss": 1.1091, + "step": 13355 + }, + { + "epoch": 0.86, + "learning_rate": 1.7156014512977954e-05, + "loss": 1.0376, + "step": 13356 + }, + { + "epoch": 0.86, + "learning_rate": 1.7155799823955003e-05, + "loss": 1.0154, + "step": 13357 + }, + { + "epoch": 0.86, + "learning_rate": 1.715558513493205e-05, + "loss": 1.016, + "step": 13358 + }, + { + "epoch": 0.86, + "learning_rate": 1.7155370445909104e-05, + "loss": 1.0356, + "step": 13359 + }, + { + "epoch": 0.86, + "learning_rate": 1.7155155756886153e-05, + "loss": 1.0708, + "step": 13360 + }, + { + "epoch": 0.86, + "learning_rate": 1.7154941067863202e-05, + "loss": 1.065, + "step": 13361 + }, + { + "epoch": 0.86, + "learning_rate": 1.715472637884025e-05, + "loss": 1.0382, + "step": 13362 + }, + { + "epoch": 0.86, + "learning_rate": 1.71545116898173e-05, + "loss": 1.1125, + "step": 13363 + }, + { + "epoch": 0.86, + "learning_rate": 1.7154297000794352e-05, + "loss": 1.0935, + "step": 13364 + }, + { + "epoch": 0.86, + "learning_rate": 1.71540823117714e-05, + "loss": 1.1286, + "step": 13365 + }, + { + "epoch": 0.86, + "learning_rate": 1.715386762274845e-05, + "loss": 1.0022, + "step": 13366 + }, + { + "epoch": 0.86, + "learning_rate": 1.71536529337255e-05, + "loss": 1.086, + "step": 13367 + }, + { + "epoch": 0.86, + "learning_rate": 1.7153438244702548e-05, + "loss": 1.0722, + "step": 13368 + }, + { + "epoch": 0.86, + "learning_rate": 1.71532235556796e-05, + "loss": 1.0738, + "step": 13369 + }, + { + "epoch": 0.86, + "learning_rate": 1.715300886665665e-05, + "loss": 1.138, + "step": 13370 + }, + { + "epoch": 0.86, + "learning_rate": 1.7152794177633698e-05, + "loss": 1.0115, + "step": 13371 + }, + { + "epoch": 0.86, + "learning_rate": 1.7152579488610747e-05, + "loss": 1.0665, + "step": 13372 + }, + { + "epoch": 0.86, + "learning_rate": 1.71523647995878e-05, + "loss": 1.1163, + "step": 13373 + }, + { + "epoch": 0.86, + "learning_rate": 1.715215011056485e-05, + "loss": 1.1343, + "step": 13374 + }, + { + "epoch": 0.86, + "learning_rate": 1.7151935421541897e-05, + "loss": 1.0032, + "step": 13375 + }, + { + "epoch": 0.86, + "learning_rate": 1.715172073251895e-05, + "loss": 1.1579, + "step": 13376 + }, + { + "epoch": 0.86, + "learning_rate": 1.7151506043496e-05, + "loss": 1.0036, + "step": 13377 + }, + { + "epoch": 0.86, + "learning_rate": 1.7151291354473048e-05, + "loss": 1.095, + "step": 13378 + }, + { + "epoch": 0.86, + "learning_rate": 1.7151076665450097e-05, + "loss": 1.149, + "step": 13379 + }, + { + "epoch": 0.86, + "learning_rate": 1.7150861976427146e-05, + "loss": 1.0305, + "step": 13380 + }, + { + "epoch": 0.86, + "learning_rate": 1.7150647287404195e-05, + "loss": 1.0549, + "step": 13381 + }, + { + "epoch": 0.86, + "learning_rate": 1.7150432598381247e-05, + "loss": 1.1179, + "step": 13382 + }, + { + "epoch": 0.86, + "learning_rate": 1.7150217909358296e-05, + "loss": 1.0684, + "step": 13383 + }, + { + "epoch": 0.86, + "learning_rate": 1.7150003220335345e-05, + "loss": 1.0245, + "step": 13384 + }, + { + "epoch": 0.86, + "learning_rate": 1.7149788531312394e-05, + "loss": 1.0291, + "step": 13385 + }, + { + "epoch": 0.86, + "learning_rate": 1.7149573842289443e-05, + "loss": 1.0694, + "step": 13386 + }, + { + "epoch": 0.86, + "learning_rate": 1.7149359153266495e-05, + "loss": 1.032, + "step": 13387 + }, + { + "epoch": 0.86, + "learning_rate": 1.7149144464243544e-05, + "loss": 0.9677, + "step": 13388 + }, + { + "epoch": 0.86, + "learning_rate": 1.7148929775220597e-05, + "loss": 1.0446, + "step": 13389 + }, + { + "epoch": 0.86, + "learning_rate": 1.7148715086197646e-05, + "loss": 0.9393, + "step": 13390 + }, + { + "epoch": 0.86, + "learning_rate": 1.7148500397174694e-05, + "loss": 1.0405, + "step": 13391 + }, + { + "epoch": 0.86, + "learning_rate": 1.7148285708151743e-05, + "loss": 1.1212, + "step": 13392 + }, + { + "epoch": 0.86, + "learning_rate": 1.7148071019128792e-05, + "loss": 1.0893, + "step": 13393 + }, + { + "epoch": 0.86, + "learning_rate": 1.7147856330105845e-05, + "loss": 1.0979, + "step": 13394 + }, + { + "epoch": 0.86, + "learning_rate": 1.7147641641082894e-05, + "loss": 1.0569, + "step": 13395 + }, + { + "epoch": 0.86, + "learning_rate": 1.7147426952059943e-05, + "loss": 1.1409, + "step": 13396 + }, + { + "epoch": 0.86, + "learning_rate": 1.714721226303699e-05, + "loss": 1.0235, + "step": 13397 + }, + { + "epoch": 0.86, + "learning_rate": 1.714699757401404e-05, + "loss": 1.0389, + "step": 13398 + }, + { + "epoch": 0.86, + "learning_rate": 1.714678288499109e-05, + "loss": 1.0249, + "step": 13399 + }, + { + "epoch": 0.86, + "learning_rate": 1.7146568195968142e-05, + "loss": 1.1324, + "step": 13400 + }, + { + "epoch": 0.86, + "learning_rate": 1.714635350694519e-05, + "loss": 1.0084, + "step": 13401 + }, + { + "epoch": 0.86, + "learning_rate": 1.7146138817922243e-05, + "loss": 1.0459, + "step": 13402 + }, + { + "epoch": 0.86, + "learning_rate": 1.7145924128899292e-05, + "loss": 0.9722, + "step": 13403 + }, + { + "epoch": 0.86, + "learning_rate": 1.714570943987634e-05, + "loss": 1.0659, + "step": 13404 + }, + { + "epoch": 0.86, + "learning_rate": 1.714549475085339e-05, + "loss": 1.0063, + "step": 13405 + }, + { + "epoch": 0.86, + "learning_rate": 1.714528006183044e-05, + "loss": 1.0975, + "step": 13406 + }, + { + "epoch": 0.86, + "learning_rate": 1.714506537280749e-05, + "loss": 1.0308, + "step": 13407 + }, + { + "epoch": 0.86, + "learning_rate": 1.714485068378454e-05, + "loss": 1.0582, + "step": 13408 + }, + { + "epoch": 0.86, + "learning_rate": 1.714463599476159e-05, + "loss": 1.0996, + "step": 13409 + }, + { + "epoch": 0.86, + "learning_rate": 1.714442130573864e-05, + "loss": 1.0377, + "step": 13410 + }, + { + "epoch": 0.86, + "learning_rate": 1.7144206616715687e-05, + "loss": 1.0537, + "step": 13411 + }, + { + "epoch": 0.86, + "learning_rate": 1.714399192769274e-05, + "loss": 1.0174, + "step": 13412 + }, + { + "epoch": 0.86, + "learning_rate": 1.714377723866979e-05, + "loss": 1.0062, + "step": 13413 + }, + { + "epoch": 0.86, + "learning_rate": 1.7143562549646838e-05, + "loss": 1.1274, + "step": 13414 + }, + { + "epoch": 0.86, + "learning_rate": 1.7143347860623887e-05, + "loss": 1.0131, + "step": 13415 + }, + { + "epoch": 0.86, + "learning_rate": 1.7143133171600936e-05, + "loss": 1.0266, + "step": 13416 + }, + { + "epoch": 0.86, + "learning_rate": 1.7142918482577988e-05, + "loss": 1.0243, + "step": 13417 + }, + { + "epoch": 0.86, + "learning_rate": 1.7142703793555037e-05, + "loss": 1.0263, + "step": 13418 + }, + { + "epoch": 0.86, + "learning_rate": 1.7142489104532086e-05, + "loss": 1.0262, + "step": 13419 + }, + { + "epoch": 0.86, + "learning_rate": 1.7142274415509138e-05, + "loss": 1.1583, + "step": 13420 + }, + { + "epoch": 0.86, + "learning_rate": 1.7142059726486187e-05, + "loss": 1.0206, + "step": 13421 + }, + { + "epoch": 0.86, + "learning_rate": 1.7141845037463236e-05, + "loss": 1.0121, + "step": 13422 + }, + { + "epoch": 0.86, + "learning_rate": 1.7141630348440285e-05, + "loss": 1.0859, + "step": 13423 + }, + { + "epoch": 0.86, + "learning_rate": 1.7141415659417334e-05, + "loss": 1.129, + "step": 13424 + }, + { + "epoch": 0.86, + "learning_rate": 1.7141200970394386e-05, + "loss": 0.9582, + "step": 13425 + }, + { + "epoch": 0.86, + "learning_rate": 1.7140986281371435e-05, + "loss": 1.0972, + "step": 13426 + }, + { + "epoch": 0.86, + "learning_rate": 1.7140771592348484e-05, + "loss": 0.9976, + "step": 13427 + }, + { + "epoch": 0.86, + "learning_rate": 1.7140556903325533e-05, + "loss": 1.0888, + "step": 13428 + }, + { + "epoch": 0.86, + "learning_rate": 1.7140342214302582e-05, + "loss": 1.0438, + "step": 13429 + }, + { + "epoch": 0.86, + "learning_rate": 1.714012752527963e-05, + "loss": 1.104, + "step": 13430 + }, + { + "epoch": 0.86, + "learning_rate": 1.7139912836256684e-05, + "loss": 1.0844, + "step": 13431 + }, + { + "epoch": 0.86, + "learning_rate": 1.7139698147233736e-05, + "loss": 1.0715, + "step": 13432 + }, + { + "epoch": 0.86, + "learning_rate": 1.7139483458210785e-05, + "loss": 1.1249, + "step": 13433 + }, + { + "epoch": 0.86, + "learning_rate": 1.7139268769187834e-05, + "loss": 1.0916, + "step": 13434 + }, + { + "epoch": 0.86, + "learning_rate": 1.7139054080164883e-05, + "loss": 1.0619, + "step": 13435 + }, + { + "epoch": 0.86, + "learning_rate": 1.7138839391141932e-05, + "loss": 1.0378, + "step": 13436 + }, + { + "epoch": 0.86, + "learning_rate": 1.713862470211898e-05, + "loss": 1.0222, + "step": 13437 + }, + { + "epoch": 0.86, + "learning_rate": 1.7138410013096033e-05, + "loss": 1.0737, + "step": 13438 + }, + { + "epoch": 0.86, + "learning_rate": 1.7138195324073082e-05, + "loss": 1.0473, + "step": 13439 + }, + { + "epoch": 0.86, + "learning_rate": 1.713798063505013e-05, + "loss": 1.0194, + "step": 13440 + }, + { + "epoch": 0.86, + "learning_rate": 1.713776594602718e-05, + "loss": 1.0291, + "step": 13441 + }, + { + "epoch": 0.86, + "learning_rate": 1.713755125700423e-05, + "loss": 1.0397, + "step": 13442 + }, + { + "epoch": 0.86, + "learning_rate": 1.713733656798128e-05, + "loss": 1.1737, + "step": 13443 + }, + { + "epoch": 0.86, + "learning_rate": 1.713712187895833e-05, + "loss": 1.0462, + "step": 13444 + }, + { + "epoch": 0.87, + "learning_rate": 1.713690718993538e-05, + "loss": 1.076, + "step": 13445 + }, + { + "epoch": 0.87, + "learning_rate": 1.713669250091243e-05, + "loss": 0.9936, + "step": 13446 + }, + { + "epoch": 0.87, + "learning_rate": 1.713647781188948e-05, + "loss": 1.0847, + "step": 13447 + }, + { + "epoch": 0.87, + "learning_rate": 1.713626312286653e-05, + "loss": 1.0885, + "step": 13448 + }, + { + "epoch": 0.87, + "learning_rate": 1.713604843384358e-05, + "loss": 1.1212, + "step": 13449 + }, + { + "epoch": 0.87, + "learning_rate": 1.7135833744820627e-05, + "loss": 1.0545, + "step": 13450 + }, + { + "epoch": 0.87, + "learning_rate": 1.713561905579768e-05, + "loss": 1.1302, + "step": 13451 + }, + { + "epoch": 0.87, + "learning_rate": 1.713540436677473e-05, + "loss": 1.0704, + "step": 13452 + }, + { + "epoch": 0.87, + "learning_rate": 1.7135189677751778e-05, + "loss": 1.0843, + "step": 13453 + }, + { + "epoch": 0.87, + "learning_rate": 1.7134974988728827e-05, + "loss": 1.0835, + "step": 13454 + }, + { + "epoch": 0.87, + "learning_rate": 1.7134760299705876e-05, + "loss": 1.0393, + "step": 13455 + }, + { + "epoch": 0.87, + "learning_rate": 1.7134545610682928e-05, + "loss": 0.9991, + "step": 13456 + }, + { + "epoch": 0.87, + "learning_rate": 1.7134330921659977e-05, + "loss": 1.1264, + "step": 13457 + }, + { + "epoch": 0.87, + "learning_rate": 1.7134116232637026e-05, + "loss": 1.0454, + "step": 13458 + }, + { + "epoch": 0.87, + "learning_rate": 1.7133901543614075e-05, + "loss": 1.0447, + "step": 13459 + }, + { + "epoch": 0.87, + "learning_rate": 1.7133686854591127e-05, + "loss": 1.0237, + "step": 13460 + }, + { + "epoch": 0.87, + "learning_rate": 1.7133472165568176e-05, + "loss": 1.0982, + "step": 13461 + }, + { + "epoch": 0.87, + "learning_rate": 1.7133257476545225e-05, + "loss": 1.0976, + "step": 13462 + }, + { + "epoch": 0.87, + "learning_rate": 1.7133042787522278e-05, + "loss": 1.0038, + "step": 13463 + }, + { + "epoch": 0.87, + "learning_rate": 1.7132828098499327e-05, + "loss": 1.0419, + "step": 13464 + }, + { + "epoch": 0.87, + "learning_rate": 1.7132613409476375e-05, + "loss": 1.0815, + "step": 13465 + }, + { + "epoch": 0.87, + "learning_rate": 1.7132398720453424e-05, + "loss": 1.0676, + "step": 13466 + }, + { + "epoch": 0.87, + "learning_rate": 1.7132184031430473e-05, + "loss": 1.0285, + "step": 13467 + }, + { + "epoch": 0.87, + "learning_rate": 1.7131969342407522e-05, + "loss": 1.0987, + "step": 13468 + }, + { + "epoch": 0.87, + "learning_rate": 1.7131754653384575e-05, + "loss": 1.0334, + "step": 13469 + }, + { + "epoch": 0.87, + "learning_rate": 1.7131539964361624e-05, + "loss": 1.0328, + "step": 13470 + }, + { + "epoch": 0.87, + "learning_rate": 1.7131325275338673e-05, + "loss": 1.0366, + "step": 13471 + }, + { + "epoch": 0.87, + "learning_rate": 1.713111058631572e-05, + "loss": 1.0618, + "step": 13472 + }, + { + "epoch": 0.87, + "learning_rate": 1.713089589729277e-05, + "loss": 1.0464, + "step": 13473 + }, + { + "epoch": 0.87, + "learning_rate": 1.7130681208269823e-05, + "loss": 1.0759, + "step": 13474 + }, + { + "epoch": 0.87, + "learning_rate": 1.7130466519246872e-05, + "loss": 0.9448, + "step": 13475 + }, + { + "epoch": 0.87, + "learning_rate": 1.7130251830223924e-05, + "loss": 0.9472, + "step": 13476 + }, + { + "epoch": 0.87, + "learning_rate": 1.7130037141200973e-05, + "loss": 1.03, + "step": 13477 + }, + { + "epoch": 0.87, + "learning_rate": 1.7129822452178022e-05, + "loss": 1.1393, + "step": 13478 + }, + { + "epoch": 0.87, + "learning_rate": 1.712960776315507e-05, + "loss": 1.0754, + "step": 13479 + }, + { + "epoch": 0.87, + "learning_rate": 1.712939307413212e-05, + "loss": 1.0652, + "step": 13480 + }, + { + "epoch": 0.87, + "learning_rate": 1.7129178385109172e-05, + "loss": 1.1058, + "step": 13481 + }, + { + "epoch": 0.87, + "learning_rate": 1.712896369608622e-05, + "loss": 1.1335, + "step": 13482 + }, + { + "epoch": 0.87, + "learning_rate": 1.712874900706327e-05, + "loss": 1.0182, + "step": 13483 + }, + { + "epoch": 0.87, + "learning_rate": 1.712853431804032e-05, + "loss": 1.004, + "step": 13484 + }, + { + "epoch": 0.87, + "learning_rate": 1.712831962901737e-05, + "loss": 1.0427, + "step": 13485 + }, + { + "epoch": 0.87, + "learning_rate": 1.7128104939994417e-05, + "loss": 1.0541, + "step": 13486 + }, + { + "epoch": 0.87, + "learning_rate": 1.712789025097147e-05, + "loss": 1.1007, + "step": 13487 + }, + { + "epoch": 0.87, + "learning_rate": 1.712767556194852e-05, + "loss": 1.1135, + "step": 13488 + }, + { + "epoch": 0.87, + "learning_rate": 1.712746087292557e-05, + "loss": 0.9652, + "step": 13489 + }, + { + "epoch": 0.87, + "learning_rate": 1.712724618390262e-05, + "loss": 0.9747, + "step": 13490 + }, + { + "epoch": 0.87, + "learning_rate": 1.712703149487967e-05, + "loss": 1.0728, + "step": 13491 + }, + { + "epoch": 0.87, + "learning_rate": 1.7126816805856718e-05, + "loss": 1.0527, + "step": 13492 + }, + { + "epoch": 0.87, + "learning_rate": 1.7126602116833767e-05, + "loss": 1.0533, + "step": 13493 + }, + { + "epoch": 0.87, + "learning_rate": 1.712638742781082e-05, + "loss": 1.0578, + "step": 13494 + }, + { + "epoch": 0.87, + "learning_rate": 1.7126172738787868e-05, + "loss": 1.0904, + "step": 13495 + }, + { + "epoch": 0.87, + "learning_rate": 1.7125958049764917e-05, + "loss": 1.1144, + "step": 13496 + }, + { + "epoch": 0.87, + "learning_rate": 1.7125743360741966e-05, + "loss": 1.0334, + "step": 13497 + }, + { + "epoch": 0.87, + "learning_rate": 1.7125528671719015e-05, + "loss": 1.0918, + "step": 13498 + }, + { + "epoch": 0.87, + "learning_rate": 1.7125313982696067e-05, + "loss": 1.0571, + "step": 13499 + }, + { + "epoch": 0.87, + "learning_rate": 1.7125099293673116e-05, + "loss": 1.0573, + "step": 13500 + }, + { + "epoch": 0.87, + "learning_rate": 1.7124884604650165e-05, + "loss": 1.0738, + "step": 13501 + }, + { + "epoch": 0.87, + "learning_rate": 1.7124669915627214e-05, + "loss": 1.0354, + "step": 13502 + }, + { + "epoch": 0.87, + "learning_rate": 1.7124455226604263e-05, + "loss": 0.9935, + "step": 13503 + }, + { + "epoch": 0.87, + "learning_rate": 1.7124240537581316e-05, + "loss": 1.1254, + "step": 13504 + }, + { + "epoch": 0.87, + "learning_rate": 1.7124025848558365e-05, + "loss": 1.0526, + "step": 13505 + }, + { + "epoch": 0.87, + "learning_rate": 1.7123811159535414e-05, + "loss": 1.0587, + "step": 13506 + }, + { + "epoch": 0.87, + "learning_rate": 1.7123596470512466e-05, + "loss": 1.0653, + "step": 13507 + }, + { + "epoch": 0.87, + "learning_rate": 1.7123381781489515e-05, + "loss": 1.0528, + "step": 13508 + }, + { + "epoch": 0.87, + "learning_rate": 1.7123167092466564e-05, + "loss": 1.0424, + "step": 13509 + }, + { + "epoch": 0.87, + "learning_rate": 1.7122952403443613e-05, + "loss": 1.05, + "step": 13510 + }, + { + "epoch": 0.87, + "learning_rate": 1.7122737714420662e-05, + "loss": 1.0839, + "step": 13511 + }, + { + "epoch": 0.87, + "learning_rate": 1.7122523025397714e-05, + "loss": 1.0418, + "step": 13512 + }, + { + "epoch": 0.87, + "learning_rate": 1.7122308336374763e-05, + "loss": 1.0267, + "step": 13513 + }, + { + "epoch": 0.87, + "learning_rate": 1.7122093647351812e-05, + "loss": 1.0895, + "step": 13514 + }, + { + "epoch": 0.87, + "learning_rate": 1.712187895832886e-05, + "loss": 1.0381, + "step": 13515 + }, + { + "epoch": 0.87, + "learning_rate": 1.712166426930591e-05, + "loss": 1.1304, + "step": 13516 + }, + { + "epoch": 0.87, + "learning_rate": 1.712144958028296e-05, + "loss": 1.1088, + "step": 13517 + }, + { + "epoch": 0.87, + "learning_rate": 1.712123489126001e-05, + "loss": 1.0971, + "step": 13518 + }, + { + "epoch": 0.87, + "learning_rate": 1.7121020202237064e-05, + "loss": 1.0399, + "step": 13519 + }, + { + "epoch": 0.87, + "learning_rate": 1.7120805513214113e-05, + "loss": 1.0759, + "step": 13520 + }, + { + "epoch": 0.87, + "learning_rate": 1.712059082419116e-05, + "loss": 1.0935, + "step": 13521 + }, + { + "epoch": 0.87, + "learning_rate": 1.712037613516821e-05, + "loss": 1.0631, + "step": 13522 + }, + { + "epoch": 0.87, + "learning_rate": 1.712016144614526e-05, + "loss": 0.9715, + "step": 13523 + }, + { + "epoch": 0.87, + "learning_rate": 1.711994675712231e-05, + "loss": 1.088, + "step": 13524 + }, + { + "epoch": 0.87, + "learning_rate": 1.711973206809936e-05, + "loss": 1.0367, + "step": 13525 + }, + { + "epoch": 0.87, + "learning_rate": 1.711951737907641e-05, + "loss": 1.1212, + "step": 13526 + }, + { + "epoch": 0.87, + "learning_rate": 1.711930269005346e-05, + "loss": 1.0568, + "step": 13527 + }, + { + "epoch": 0.87, + "learning_rate": 1.7119088001030508e-05, + "loss": 1.187, + "step": 13528 + }, + { + "epoch": 0.87, + "learning_rate": 1.7118873312007557e-05, + "loss": 1.0082, + "step": 13529 + }, + { + "epoch": 0.87, + "learning_rate": 1.711865862298461e-05, + "loss": 1.103, + "step": 13530 + }, + { + "epoch": 0.87, + "learning_rate": 1.7118443933961658e-05, + "loss": 1.0882, + "step": 13531 + }, + { + "epoch": 0.87, + "learning_rate": 1.7118229244938707e-05, + "loss": 1.0842, + "step": 13532 + }, + { + "epoch": 0.87, + "learning_rate": 1.711801455591576e-05, + "loss": 1.0835, + "step": 13533 + }, + { + "epoch": 0.87, + "learning_rate": 1.7117799866892808e-05, + "loss": 1.059, + "step": 13534 + }, + { + "epoch": 0.87, + "learning_rate": 1.7117585177869857e-05, + "loss": 1.1035, + "step": 13535 + }, + { + "epoch": 0.87, + "learning_rate": 1.7117370488846906e-05, + "loss": 1.1098, + "step": 13536 + }, + { + "epoch": 0.87, + "learning_rate": 1.7117155799823955e-05, + "loss": 1.0198, + "step": 13537 + }, + { + "epoch": 0.87, + "learning_rate": 1.7116941110801008e-05, + "loss": 1.0567, + "step": 13538 + }, + { + "epoch": 0.87, + "learning_rate": 1.7116726421778056e-05, + "loss": 0.9896, + "step": 13539 + }, + { + "epoch": 0.87, + "learning_rate": 1.7116511732755105e-05, + "loss": 1.0452, + "step": 13540 + }, + { + "epoch": 0.87, + "learning_rate": 1.7116297043732154e-05, + "loss": 1.0543, + "step": 13541 + }, + { + "epoch": 0.87, + "learning_rate": 1.7116082354709203e-05, + "loss": 0.9942, + "step": 13542 + }, + { + "epoch": 0.87, + "learning_rate": 1.7115867665686256e-05, + "loss": 0.9948, + "step": 13543 + }, + { + "epoch": 0.87, + "learning_rate": 1.7115652976663305e-05, + "loss": 1.0968, + "step": 13544 + }, + { + "epoch": 0.87, + "learning_rate": 1.7115438287640354e-05, + "loss": 1.0784, + "step": 13545 + }, + { + "epoch": 0.87, + "learning_rate": 1.7115223598617403e-05, + "loss": 1.0963, + "step": 13546 + }, + { + "epoch": 0.87, + "learning_rate": 1.7115008909594455e-05, + "loss": 1.017, + "step": 13547 + }, + { + "epoch": 0.87, + "learning_rate": 1.7114794220571504e-05, + "loss": 1.0349, + "step": 13548 + }, + { + "epoch": 0.87, + "learning_rate": 1.7114579531548553e-05, + "loss": 1.06, + "step": 13549 + }, + { + "epoch": 0.87, + "learning_rate": 1.7114364842525605e-05, + "loss": 1.0692, + "step": 13550 + }, + { + "epoch": 0.87, + "learning_rate": 1.7114150153502654e-05, + "loss": 1.0684, + "step": 13551 + }, + { + "epoch": 0.87, + "learning_rate": 1.7113935464479703e-05, + "loss": 1.0871, + "step": 13552 + }, + { + "epoch": 0.87, + "learning_rate": 1.7113720775456752e-05, + "loss": 1.0677, + "step": 13553 + }, + { + "epoch": 0.87, + "learning_rate": 1.71135060864338e-05, + "loss": 0.9597, + "step": 13554 + }, + { + "epoch": 0.87, + "learning_rate": 1.711329139741085e-05, + "loss": 1.0179, + "step": 13555 + }, + { + "epoch": 0.87, + "learning_rate": 1.7113076708387902e-05, + "loss": 0.9935, + "step": 13556 + }, + { + "epoch": 0.87, + "learning_rate": 1.711286201936495e-05, + "loss": 1.0349, + "step": 13557 + }, + { + "epoch": 0.87, + "learning_rate": 1.7112647330342e-05, + "loss": 1.0582, + "step": 13558 + }, + { + "epoch": 0.87, + "learning_rate": 1.711243264131905e-05, + "loss": 1.1152, + "step": 13559 + }, + { + "epoch": 0.87, + "learning_rate": 1.7112217952296098e-05, + "loss": 1.0952, + "step": 13560 + }, + { + "epoch": 0.87, + "learning_rate": 1.711200326327315e-05, + "loss": 1.0782, + "step": 13561 + }, + { + "epoch": 0.87, + "learning_rate": 1.71117885742502e-05, + "loss": 1.03, + "step": 13562 + }, + { + "epoch": 0.87, + "learning_rate": 1.7111573885227252e-05, + "loss": 0.9805, + "step": 13563 + }, + { + "epoch": 0.87, + "learning_rate": 1.71113591962043e-05, + "loss": 1.1362, + "step": 13564 + }, + { + "epoch": 0.87, + "learning_rate": 1.711114450718135e-05, + "loss": 1.0387, + "step": 13565 + }, + { + "epoch": 0.87, + "learning_rate": 1.71109298181584e-05, + "loss": 1.0501, + "step": 13566 + }, + { + "epoch": 0.87, + "learning_rate": 1.7110715129135448e-05, + "loss": 1.097, + "step": 13567 + }, + { + "epoch": 0.87, + "learning_rate": 1.71105004401125e-05, + "loss": 1.0853, + "step": 13568 + }, + { + "epoch": 0.87, + "learning_rate": 1.711028575108955e-05, + "loss": 1.0905, + "step": 13569 + }, + { + "epoch": 0.87, + "learning_rate": 1.7110071062066598e-05, + "loss": 1.0808, + "step": 13570 + }, + { + "epoch": 0.87, + "learning_rate": 1.7109856373043647e-05, + "loss": 0.935, + "step": 13571 + }, + { + "epoch": 0.87, + "learning_rate": 1.7109641684020696e-05, + "loss": 0.9751, + "step": 13572 + }, + { + "epoch": 0.87, + "learning_rate": 1.7109426994997745e-05, + "loss": 1.0785, + "step": 13573 + }, + { + "epoch": 0.87, + "learning_rate": 1.7109212305974797e-05, + "loss": 1.0476, + "step": 13574 + }, + { + "epoch": 0.87, + "learning_rate": 1.7108997616951846e-05, + "loss": 1.0551, + "step": 13575 + }, + { + "epoch": 0.87, + "learning_rate": 1.71087829279289e-05, + "loss": 1.0213, + "step": 13576 + }, + { + "epoch": 0.87, + "learning_rate": 1.7108568238905948e-05, + "loss": 1.059, + "step": 13577 + }, + { + "epoch": 0.87, + "learning_rate": 1.7108353549882997e-05, + "loss": 1.0646, + "step": 13578 + }, + { + "epoch": 0.87, + "learning_rate": 1.7108138860860046e-05, + "loss": 0.9794, + "step": 13579 + }, + { + "epoch": 0.87, + "learning_rate": 1.7107924171837095e-05, + "loss": 1.1101, + "step": 13580 + }, + { + "epoch": 0.87, + "learning_rate": 1.7107709482814147e-05, + "loss": 1.0932, + "step": 13581 + }, + { + "epoch": 0.87, + "learning_rate": 1.7107494793791196e-05, + "loss": 1.0755, + "step": 13582 + }, + { + "epoch": 0.87, + "learning_rate": 1.7107280104768245e-05, + "loss": 1.0376, + "step": 13583 + }, + { + "epoch": 0.87, + "learning_rate": 1.7107065415745294e-05, + "loss": 1.0754, + "step": 13584 + }, + { + "epoch": 0.87, + "learning_rate": 1.7106850726722343e-05, + "loss": 1.0517, + "step": 13585 + }, + { + "epoch": 0.87, + "learning_rate": 1.7106636037699392e-05, + "loss": 1.111, + "step": 13586 + }, + { + "epoch": 0.87, + "learning_rate": 1.7106421348676444e-05, + "loss": 1.087, + "step": 13587 + }, + { + "epoch": 0.87, + "learning_rate": 1.7106206659653493e-05, + "loss": 1.0357, + "step": 13588 + }, + { + "epoch": 0.87, + "learning_rate": 1.7105991970630542e-05, + "loss": 1.0617, + "step": 13589 + }, + { + "epoch": 0.87, + "learning_rate": 1.710577728160759e-05, + "loss": 1.0522, + "step": 13590 + }, + { + "epoch": 0.87, + "learning_rate": 1.7105562592584643e-05, + "loss": 1.1177, + "step": 13591 + }, + { + "epoch": 0.87, + "learning_rate": 1.7105347903561692e-05, + "loss": 1.0672, + "step": 13592 + }, + { + "epoch": 0.87, + "learning_rate": 1.710513321453874e-05, + "loss": 1.0685, + "step": 13593 + }, + { + "epoch": 0.87, + "learning_rate": 1.7104918525515794e-05, + "loss": 1.1311, + "step": 13594 + }, + { + "epoch": 0.87, + "learning_rate": 1.7104703836492843e-05, + "loss": 1.0957, + "step": 13595 + }, + { + "epoch": 0.87, + "learning_rate": 1.710448914746989e-05, + "loss": 1.1632, + "step": 13596 + }, + { + "epoch": 0.87, + "learning_rate": 1.710427445844694e-05, + "loss": 0.9916, + "step": 13597 + }, + { + "epoch": 0.87, + "learning_rate": 1.710405976942399e-05, + "loss": 1.1711, + "step": 13598 + }, + { + "epoch": 0.87, + "learning_rate": 1.7103845080401042e-05, + "loss": 1.0327, + "step": 13599 + }, + { + "epoch": 0.87, + "learning_rate": 1.710363039137809e-05, + "loss": 1.0299, + "step": 13600 + }, + { + "epoch": 0.88, + "learning_rate": 1.710341570235514e-05, + "loss": 1.062, + "step": 13601 + }, + { + "epoch": 0.88, + "learning_rate": 1.710320101333219e-05, + "loss": 1.0088, + "step": 13602 + }, + { + "epoch": 0.88, + "learning_rate": 1.7102986324309238e-05, + "loss": 1.0157, + "step": 13603 + }, + { + "epoch": 0.88, + "learning_rate": 1.7102771635286287e-05, + "loss": 1.0867, + "step": 13604 + }, + { + "epoch": 0.88, + "learning_rate": 1.710255694626334e-05, + "loss": 1.0475, + "step": 13605 + }, + { + "epoch": 0.88, + "learning_rate": 1.710234225724039e-05, + "loss": 1.0334, + "step": 13606 + }, + { + "epoch": 0.88, + "learning_rate": 1.710212756821744e-05, + "loss": 1.0004, + "step": 13607 + }, + { + "epoch": 0.88, + "learning_rate": 1.710191287919449e-05, + "loss": 1.104, + "step": 13608 + }, + { + "epoch": 0.88, + "learning_rate": 1.7101698190171538e-05, + "loss": 1.0071, + "step": 13609 + }, + { + "epoch": 0.88, + "learning_rate": 1.7101483501148587e-05, + "loss": 0.9987, + "step": 13610 + }, + { + "epoch": 0.88, + "learning_rate": 1.7101268812125636e-05, + "loss": 1.0814, + "step": 13611 + }, + { + "epoch": 0.88, + "learning_rate": 1.710105412310269e-05, + "loss": 1.049, + "step": 13612 + }, + { + "epoch": 0.88, + "learning_rate": 1.7100839434079737e-05, + "loss": 1.0335, + "step": 13613 + }, + { + "epoch": 0.88, + "learning_rate": 1.7100624745056786e-05, + "loss": 1.051, + "step": 13614 + }, + { + "epoch": 0.88, + "learning_rate": 1.7100410056033835e-05, + "loss": 1.0944, + "step": 13615 + }, + { + "epoch": 0.88, + "learning_rate": 1.7100195367010884e-05, + "loss": 1.0401, + "step": 13616 + }, + { + "epoch": 0.88, + "learning_rate": 1.7099980677987937e-05, + "loss": 1.0885, + "step": 13617 + }, + { + "epoch": 0.88, + "learning_rate": 1.7099765988964986e-05, + "loss": 1.0155, + "step": 13618 + }, + { + "epoch": 0.88, + "learning_rate": 1.7099551299942035e-05, + "loss": 1.0821, + "step": 13619 + }, + { + "epoch": 0.88, + "learning_rate": 1.7099336610919087e-05, + "loss": 1.0421, + "step": 13620 + }, + { + "epoch": 0.88, + "learning_rate": 1.7099121921896136e-05, + "loss": 1.0276, + "step": 13621 + }, + { + "epoch": 0.88, + "learning_rate": 1.7098907232873185e-05, + "loss": 1.1051, + "step": 13622 + }, + { + "epoch": 0.88, + "learning_rate": 1.7098692543850234e-05, + "loss": 1.0347, + "step": 13623 + }, + { + "epoch": 0.88, + "learning_rate": 1.7098477854827283e-05, + "loss": 1.0514, + "step": 13624 + }, + { + "epoch": 0.88, + "learning_rate": 1.7098263165804335e-05, + "loss": 1.1408, + "step": 13625 + }, + { + "epoch": 0.88, + "learning_rate": 1.7098048476781384e-05, + "loss": 1.0226, + "step": 13626 + }, + { + "epoch": 0.88, + "learning_rate": 1.7097833787758433e-05, + "loss": 1.0327, + "step": 13627 + }, + { + "epoch": 0.88, + "learning_rate": 1.7097619098735482e-05, + "loss": 1.095, + "step": 13628 + }, + { + "epoch": 0.88, + "learning_rate": 1.709740440971253e-05, + "loss": 1.0052, + "step": 13629 + }, + { + "epoch": 0.88, + "learning_rate": 1.7097189720689583e-05, + "loss": 1.0376, + "step": 13630 + }, + { + "epoch": 0.88, + "learning_rate": 1.7096975031666632e-05, + "loss": 1.0679, + "step": 13631 + }, + { + "epoch": 0.88, + "learning_rate": 1.709676034264368e-05, + "loss": 1.0678, + "step": 13632 + }, + { + "epoch": 0.88, + "learning_rate": 1.709654565362073e-05, + "loss": 1.019, + "step": 13633 + }, + { + "epoch": 0.88, + "learning_rate": 1.709633096459778e-05, + "loss": 1.0678, + "step": 13634 + }, + { + "epoch": 0.88, + "learning_rate": 1.709611627557483e-05, + "loss": 1.1343, + "step": 13635 + }, + { + "epoch": 0.88, + "learning_rate": 1.709590158655188e-05, + "loss": 1.0904, + "step": 13636 + }, + { + "epoch": 0.88, + "learning_rate": 1.7095686897528933e-05, + "loss": 1.0774, + "step": 13637 + }, + { + "epoch": 0.88, + "learning_rate": 1.7095472208505982e-05, + "loss": 1.0052, + "step": 13638 + }, + { + "epoch": 0.88, + "learning_rate": 1.709525751948303e-05, + "loss": 1.107, + "step": 13639 + }, + { + "epoch": 0.88, + "learning_rate": 1.709504283046008e-05, + "loss": 1.1585, + "step": 13640 + }, + { + "epoch": 0.88, + "learning_rate": 1.709482814143713e-05, + "loss": 1.0414, + "step": 13641 + }, + { + "epoch": 0.88, + "learning_rate": 1.7094613452414178e-05, + "loss": 1.0269, + "step": 13642 + }, + { + "epoch": 0.88, + "learning_rate": 1.709439876339123e-05, + "loss": 1.0927, + "step": 13643 + }, + { + "epoch": 0.88, + "learning_rate": 1.709418407436828e-05, + "loss": 1.0903, + "step": 13644 + }, + { + "epoch": 0.88, + "learning_rate": 1.7093969385345328e-05, + "loss": 1.0579, + "step": 13645 + }, + { + "epoch": 0.88, + "learning_rate": 1.7093754696322377e-05, + "loss": 1.0657, + "step": 13646 + }, + { + "epoch": 0.88, + "learning_rate": 1.7093540007299426e-05, + "loss": 1.0584, + "step": 13647 + }, + { + "epoch": 0.88, + "learning_rate": 1.709332531827648e-05, + "loss": 1.0277, + "step": 13648 + }, + { + "epoch": 0.88, + "learning_rate": 1.7093110629253527e-05, + "loss": 0.9736, + "step": 13649 + }, + { + "epoch": 0.88, + "learning_rate": 1.709289594023058e-05, + "loss": 1.0246, + "step": 13650 + }, + { + "epoch": 0.88, + "learning_rate": 1.709268125120763e-05, + "loss": 1.0441, + "step": 13651 + }, + { + "epoch": 0.88, + "learning_rate": 1.7092466562184678e-05, + "loss": 1.0785, + "step": 13652 + }, + { + "epoch": 0.88, + "learning_rate": 1.7092251873161727e-05, + "loss": 1.1465, + "step": 13653 + }, + { + "epoch": 0.88, + "learning_rate": 1.7092037184138776e-05, + "loss": 1.0243, + "step": 13654 + }, + { + "epoch": 0.88, + "learning_rate": 1.7091822495115828e-05, + "loss": 1.0313, + "step": 13655 + }, + { + "epoch": 0.88, + "learning_rate": 1.7091607806092877e-05, + "loss": 1.1645, + "step": 13656 + }, + { + "epoch": 0.88, + "learning_rate": 1.7091393117069926e-05, + "loss": 1.0007, + "step": 13657 + }, + { + "epoch": 0.88, + "learning_rate": 1.7091178428046975e-05, + "loss": 1.0809, + "step": 13658 + }, + { + "epoch": 0.88, + "learning_rate": 1.7090963739024024e-05, + "loss": 1.0467, + "step": 13659 + }, + { + "epoch": 0.88, + "learning_rate": 1.7090749050001073e-05, + "loss": 1.072, + "step": 13660 + }, + { + "epoch": 0.88, + "learning_rate": 1.7090534360978125e-05, + "loss": 1.0865, + "step": 13661 + }, + { + "epoch": 0.88, + "learning_rate": 1.7090319671955174e-05, + "loss": 1.124, + "step": 13662 + }, + { + "epoch": 0.88, + "learning_rate": 1.7090104982932223e-05, + "loss": 1.048, + "step": 13663 + }, + { + "epoch": 0.88, + "learning_rate": 1.7089890293909275e-05, + "loss": 1.0448, + "step": 13664 + }, + { + "epoch": 0.88, + "learning_rate": 1.7089675604886324e-05, + "loss": 1.0938, + "step": 13665 + }, + { + "epoch": 0.88, + "learning_rate": 1.7089460915863373e-05, + "loss": 1.0101, + "step": 13666 + }, + { + "epoch": 0.88, + "learning_rate": 1.7089246226840422e-05, + "loss": 1.0754, + "step": 13667 + }, + { + "epoch": 0.88, + "learning_rate": 1.7089031537817475e-05, + "loss": 1.051, + "step": 13668 + }, + { + "epoch": 0.88, + "learning_rate": 1.7088816848794524e-05, + "loss": 1.0755, + "step": 13669 + }, + { + "epoch": 0.88, + "learning_rate": 1.7088602159771573e-05, + "loss": 1.035, + "step": 13670 + }, + { + "epoch": 0.88, + "learning_rate": 1.708838747074862e-05, + "loss": 1.0237, + "step": 13671 + }, + { + "epoch": 0.88, + "learning_rate": 1.708817278172567e-05, + "loss": 1.044, + "step": 13672 + }, + { + "epoch": 0.88, + "learning_rate": 1.708795809270272e-05, + "loss": 1.0479, + "step": 13673 + }, + { + "epoch": 0.88, + "learning_rate": 1.7087743403679772e-05, + "loss": 1.1102, + "step": 13674 + }, + { + "epoch": 0.88, + "learning_rate": 1.708752871465682e-05, + "loss": 1.0534, + "step": 13675 + }, + { + "epoch": 0.88, + "learning_rate": 1.708731402563387e-05, + "loss": 1.1177, + "step": 13676 + }, + { + "epoch": 0.88, + "learning_rate": 1.708709933661092e-05, + "loss": 1.0746, + "step": 13677 + }, + { + "epoch": 0.88, + "learning_rate": 1.708688464758797e-05, + "loss": 1.0929, + "step": 13678 + }, + { + "epoch": 0.88, + "learning_rate": 1.708666995856502e-05, + "loss": 1.0741, + "step": 13679 + }, + { + "epoch": 0.88, + "learning_rate": 1.708645526954207e-05, + "loss": 1.0957, + "step": 13680 + }, + { + "epoch": 0.88, + "learning_rate": 1.708624058051912e-05, + "loss": 1.0236, + "step": 13681 + }, + { + "epoch": 0.88, + "learning_rate": 1.708602589149617e-05, + "loss": 1.0891, + "step": 13682 + }, + { + "epoch": 0.88, + "learning_rate": 1.708581120247322e-05, + "loss": 1.0407, + "step": 13683 + }, + { + "epoch": 0.88, + "learning_rate": 1.7085596513450268e-05, + "loss": 1.0938, + "step": 13684 + }, + { + "epoch": 0.88, + "learning_rate": 1.7085381824427317e-05, + "loss": 1.1273, + "step": 13685 + }, + { + "epoch": 0.88, + "learning_rate": 1.708516713540437e-05, + "loss": 0.9869, + "step": 13686 + }, + { + "epoch": 0.88, + "learning_rate": 1.708495244638142e-05, + "loss": 1.0707, + "step": 13687 + }, + { + "epoch": 0.88, + "learning_rate": 1.7084737757358467e-05, + "loss": 1.0545, + "step": 13688 + }, + { + "epoch": 0.88, + "learning_rate": 1.7084523068335516e-05, + "loss": 1.0988, + "step": 13689 + }, + { + "epoch": 0.88, + "learning_rate": 1.7084308379312565e-05, + "loss": 1.0991, + "step": 13690 + }, + { + "epoch": 0.88, + "learning_rate": 1.7084093690289614e-05, + "loss": 1.0597, + "step": 13691 + }, + { + "epoch": 0.88, + "learning_rate": 1.7083879001266667e-05, + "loss": 0.8848, + "step": 13692 + }, + { + "epoch": 0.88, + "learning_rate": 1.7083664312243716e-05, + "loss": 1.0483, + "step": 13693 + }, + { + "epoch": 0.88, + "learning_rate": 1.7083449623220768e-05, + "loss": 0.9939, + "step": 13694 + }, + { + "epoch": 0.88, + "learning_rate": 1.7083234934197817e-05, + "loss": 1.0267, + "step": 13695 + }, + { + "epoch": 0.88, + "learning_rate": 1.7083020245174866e-05, + "loss": 1.0233, + "step": 13696 + }, + { + "epoch": 0.88, + "learning_rate": 1.7082805556151915e-05, + "loss": 1.1455, + "step": 13697 + }, + { + "epoch": 0.88, + "learning_rate": 1.7082590867128964e-05, + "loss": 1.0107, + "step": 13698 + }, + { + "epoch": 0.88, + "learning_rate": 1.7082376178106016e-05, + "loss": 1.0629, + "step": 13699 + }, + { + "epoch": 0.88, + "learning_rate": 1.7082161489083065e-05, + "loss": 1.0094, + "step": 13700 + }, + { + "epoch": 0.88, + "learning_rate": 1.7081946800060114e-05, + "loss": 1.0837, + "step": 13701 + }, + { + "epoch": 0.88, + "learning_rate": 1.7081732111037163e-05, + "loss": 1.0387, + "step": 13702 + }, + { + "epoch": 0.88, + "learning_rate": 1.7081517422014212e-05, + "loss": 1.0546, + "step": 13703 + }, + { + "epoch": 0.88, + "learning_rate": 1.7081302732991264e-05, + "loss": 1.028, + "step": 13704 + }, + { + "epoch": 0.88, + "learning_rate": 1.7081088043968313e-05, + "loss": 1.0874, + "step": 13705 + }, + { + "epoch": 0.88, + "learning_rate": 1.7080873354945362e-05, + "loss": 1.0751, + "step": 13706 + }, + { + "epoch": 0.88, + "learning_rate": 1.7080658665922415e-05, + "loss": 1.0615, + "step": 13707 + }, + { + "epoch": 0.88, + "learning_rate": 1.7080443976899464e-05, + "loss": 1.1268, + "step": 13708 + }, + { + "epoch": 0.88, + "learning_rate": 1.7080229287876513e-05, + "loss": 0.9672, + "step": 13709 + }, + { + "epoch": 0.88, + "learning_rate": 1.708001459885356e-05, + "loss": 1.065, + "step": 13710 + }, + { + "epoch": 0.88, + "learning_rate": 1.707979990983061e-05, + "loss": 1.0029, + "step": 13711 + }, + { + "epoch": 0.88, + "learning_rate": 1.7079585220807663e-05, + "loss": 1.0144, + "step": 13712 + }, + { + "epoch": 0.88, + "learning_rate": 1.7079370531784712e-05, + "loss": 0.9959, + "step": 13713 + }, + { + "epoch": 0.88, + "learning_rate": 1.707915584276176e-05, + "loss": 1.0224, + "step": 13714 + }, + { + "epoch": 0.88, + "learning_rate": 1.707894115373881e-05, + "loss": 1.109, + "step": 13715 + }, + { + "epoch": 0.88, + "learning_rate": 1.707872646471586e-05, + "loss": 1.1136, + "step": 13716 + }, + { + "epoch": 0.88, + "learning_rate": 1.707851177569291e-05, + "loss": 1.0734, + "step": 13717 + }, + { + "epoch": 0.88, + "learning_rate": 1.707829708666996e-05, + "loss": 1.1313, + "step": 13718 + }, + { + "epoch": 0.88, + "learning_rate": 1.707808239764701e-05, + "loss": 0.9876, + "step": 13719 + }, + { + "epoch": 0.88, + "learning_rate": 1.7077867708624058e-05, + "loss": 1.0574, + "step": 13720 + }, + { + "epoch": 0.88, + "learning_rate": 1.7077653019601107e-05, + "loss": 1.0793, + "step": 13721 + }, + { + "epoch": 0.88, + "learning_rate": 1.707743833057816e-05, + "loss": 1.0451, + "step": 13722 + }, + { + "epoch": 0.88, + "learning_rate": 1.707722364155521e-05, + "loss": 1.0708, + "step": 13723 + }, + { + "epoch": 0.88, + "learning_rate": 1.707700895253226e-05, + "loss": 1.0184, + "step": 13724 + }, + { + "epoch": 0.88, + "learning_rate": 1.707679426350931e-05, + "loss": 1.0574, + "step": 13725 + }, + { + "epoch": 0.88, + "learning_rate": 1.707657957448636e-05, + "loss": 1.1212, + "step": 13726 + }, + { + "epoch": 0.88, + "learning_rate": 1.7076364885463408e-05, + "loss": 1.0225, + "step": 13727 + }, + { + "epoch": 0.88, + "learning_rate": 1.7076150196440457e-05, + "loss": 1.1299, + "step": 13728 + }, + { + "epoch": 0.88, + "learning_rate": 1.7075935507417506e-05, + "loss": 1.0655, + "step": 13729 + }, + { + "epoch": 0.88, + "learning_rate": 1.7075720818394558e-05, + "loss": 0.9864, + "step": 13730 + }, + { + "epoch": 0.88, + "learning_rate": 1.7075506129371607e-05, + "loss": 1.1457, + "step": 13731 + }, + { + "epoch": 0.88, + "learning_rate": 1.7075291440348656e-05, + "loss": 1.0886, + "step": 13732 + }, + { + "epoch": 0.88, + "learning_rate": 1.7075076751325705e-05, + "loss": 1.0714, + "step": 13733 + }, + { + "epoch": 0.88, + "learning_rate": 1.7074862062302754e-05, + "loss": 1.1043, + "step": 13734 + }, + { + "epoch": 0.88, + "learning_rate": 1.7074647373279806e-05, + "loss": 0.9921, + "step": 13735 + }, + { + "epoch": 0.88, + "learning_rate": 1.7074432684256855e-05, + "loss": 1.0528, + "step": 13736 + }, + { + "epoch": 0.88, + "learning_rate": 1.7074217995233907e-05, + "loss": 1.132, + "step": 13737 + }, + { + "epoch": 0.88, + "learning_rate": 1.7074003306210956e-05, + "loss": 1.0334, + "step": 13738 + }, + { + "epoch": 0.88, + "learning_rate": 1.7073788617188005e-05, + "loss": 1.0714, + "step": 13739 + }, + { + "epoch": 0.88, + "learning_rate": 1.7073573928165054e-05, + "loss": 1.0482, + "step": 13740 + }, + { + "epoch": 0.88, + "learning_rate": 1.7073359239142103e-05, + "loss": 1.0576, + "step": 13741 + }, + { + "epoch": 0.88, + "learning_rate": 1.7073144550119156e-05, + "loss": 1.1036, + "step": 13742 + }, + { + "epoch": 0.88, + "learning_rate": 1.7072929861096205e-05, + "loss": 1.08, + "step": 13743 + }, + { + "epoch": 0.88, + "learning_rate": 1.7072715172073254e-05, + "loss": 0.9994, + "step": 13744 + }, + { + "epoch": 0.88, + "learning_rate": 1.7072500483050303e-05, + "loss": 0.9729, + "step": 13745 + }, + { + "epoch": 0.88, + "learning_rate": 1.707228579402735e-05, + "loss": 1.034, + "step": 13746 + }, + { + "epoch": 0.88, + "learning_rate": 1.70720711050044e-05, + "loss": 1.1227, + "step": 13747 + }, + { + "epoch": 0.88, + "learning_rate": 1.7071856415981453e-05, + "loss": 1.0369, + "step": 13748 + }, + { + "epoch": 0.88, + "learning_rate": 1.7071641726958502e-05, + "loss": 1.1003, + "step": 13749 + }, + { + "epoch": 0.88, + "learning_rate": 1.707142703793555e-05, + "loss": 1.0992, + "step": 13750 + }, + { + "epoch": 0.88, + "learning_rate": 1.7071212348912603e-05, + "loss": 1.1039, + "step": 13751 + }, + { + "epoch": 0.88, + "learning_rate": 1.7070997659889652e-05, + "loss": 1.0863, + "step": 13752 + }, + { + "epoch": 0.88, + "learning_rate": 1.70707829708667e-05, + "loss": 1.0772, + "step": 13753 + }, + { + "epoch": 0.88, + "learning_rate": 1.707056828184375e-05, + "loss": 1.0051, + "step": 13754 + }, + { + "epoch": 0.88, + "learning_rate": 1.7070353592820802e-05, + "loss": 1.1333, + "step": 13755 + }, + { + "epoch": 0.89, + "learning_rate": 1.707013890379785e-05, + "loss": 1.0655, + "step": 13756 + }, + { + "epoch": 0.89, + "learning_rate": 1.70699242147749e-05, + "loss": 1.0819, + "step": 13757 + }, + { + "epoch": 0.89, + "learning_rate": 1.706970952575195e-05, + "loss": 1.1201, + "step": 13758 + }, + { + "epoch": 0.89, + "learning_rate": 1.7069494836728998e-05, + "loss": 1.0857, + "step": 13759 + }, + { + "epoch": 0.89, + "learning_rate": 1.7069280147706047e-05, + "loss": 1.0499, + "step": 13760 + }, + { + "epoch": 0.89, + "learning_rate": 1.70690654586831e-05, + "loss": 1.0155, + "step": 13761 + }, + { + "epoch": 0.89, + "learning_rate": 1.706885076966015e-05, + "loss": 1.0716, + "step": 13762 + }, + { + "epoch": 0.89, + "learning_rate": 1.7068636080637197e-05, + "loss": 1.0643, + "step": 13763 + }, + { + "epoch": 0.89, + "learning_rate": 1.7068421391614246e-05, + "loss": 1.0501, + "step": 13764 + }, + { + "epoch": 0.89, + "learning_rate": 1.70682067025913e-05, + "loss": 1.0338, + "step": 13765 + }, + { + "epoch": 0.89, + "learning_rate": 1.7067992013568348e-05, + "loss": 1.0307, + "step": 13766 + }, + { + "epoch": 0.89, + "learning_rate": 1.7067777324545397e-05, + "loss": 1.0032, + "step": 13767 + }, + { + "epoch": 0.89, + "learning_rate": 1.706756263552245e-05, + "loss": 0.9889, + "step": 13768 + }, + { + "epoch": 0.89, + "learning_rate": 1.7067347946499498e-05, + "loss": 1.0579, + "step": 13769 + }, + { + "epoch": 0.89, + "learning_rate": 1.7067133257476547e-05, + "loss": 1.0582, + "step": 13770 + }, + { + "epoch": 0.89, + "learning_rate": 1.7066918568453596e-05, + "loss": 1.0455, + "step": 13771 + }, + { + "epoch": 0.89, + "learning_rate": 1.7066703879430645e-05, + "loss": 1.1719, + "step": 13772 + }, + { + "epoch": 0.89, + "learning_rate": 1.7066489190407697e-05, + "loss": 1.0651, + "step": 13773 + }, + { + "epoch": 0.89, + "learning_rate": 1.7066274501384746e-05, + "loss": 1.1072, + "step": 13774 + }, + { + "epoch": 0.89, + "learning_rate": 1.7066059812361795e-05, + "loss": 1.0128, + "step": 13775 + }, + { + "epoch": 0.89, + "learning_rate": 1.7065845123338844e-05, + "loss": 1.067, + "step": 13776 + }, + { + "epoch": 0.89, + "learning_rate": 1.7065630434315893e-05, + "loss": 1.0719, + "step": 13777 + }, + { + "epoch": 0.89, + "learning_rate": 1.7065415745292942e-05, + "loss": 1.0692, + "step": 13778 + }, + { + "epoch": 0.89, + "learning_rate": 1.7065201056269994e-05, + "loss": 1.1273, + "step": 13779 + }, + { + "epoch": 0.89, + "learning_rate": 1.7064986367247043e-05, + "loss": 1.0655, + "step": 13780 + }, + { + "epoch": 0.89, + "learning_rate": 1.7064771678224096e-05, + "loss": 1.1614, + "step": 13781 + }, + { + "epoch": 0.89, + "learning_rate": 1.7064556989201145e-05, + "loss": 0.9801, + "step": 13782 + }, + { + "epoch": 0.89, + "learning_rate": 1.7064342300178194e-05, + "loss": 1.1005, + "step": 13783 + }, + { + "epoch": 0.89, + "learning_rate": 1.7064127611155243e-05, + "loss": 1.0406, + "step": 13784 + }, + { + "epoch": 0.89, + "learning_rate": 1.706391292213229e-05, + "loss": 1.0227, + "step": 13785 + }, + { + "epoch": 0.89, + "learning_rate": 1.7063698233109344e-05, + "loss": 1.0055, + "step": 13786 + }, + { + "epoch": 0.89, + "learning_rate": 1.7063483544086393e-05, + "loss": 1.0182, + "step": 13787 + }, + { + "epoch": 0.89, + "learning_rate": 1.7063268855063442e-05, + "loss": 0.963, + "step": 13788 + }, + { + "epoch": 0.89, + "learning_rate": 1.706305416604049e-05, + "loss": 0.9782, + "step": 13789 + }, + { + "epoch": 0.89, + "learning_rate": 1.706283947701754e-05, + "loss": 1.0503, + "step": 13790 + }, + { + "epoch": 0.89, + "learning_rate": 1.7062624787994592e-05, + "loss": 0.9733, + "step": 13791 + }, + { + "epoch": 0.89, + "learning_rate": 1.706241009897164e-05, + "loss": 1.0565, + "step": 13792 + }, + { + "epoch": 0.89, + "learning_rate": 1.706219540994869e-05, + "loss": 1.0915, + "step": 13793 + }, + { + "epoch": 0.89, + "learning_rate": 1.7061980720925742e-05, + "loss": 1.0785, + "step": 13794 + }, + { + "epoch": 0.89, + "learning_rate": 1.706176603190279e-05, + "loss": 1.0432, + "step": 13795 + }, + { + "epoch": 0.89, + "learning_rate": 1.706155134287984e-05, + "loss": 1.031, + "step": 13796 + }, + { + "epoch": 0.89, + "learning_rate": 1.706133665385689e-05, + "loss": 1.0506, + "step": 13797 + }, + { + "epoch": 0.89, + "learning_rate": 1.7061121964833938e-05, + "loss": 1.0509, + "step": 13798 + }, + { + "epoch": 0.89, + "learning_rate": 1.706090727581099e-05, + "loss": 1.0694, + "step": 13799 + }, + { + "epoch": 0.89, + "learning_rate": 1.706069258678804e-05, + "loss": 1.113, + "step": 13800 + }, + { + "epoch": 0.89, + "learning_rate": 1.706047789776509e-05, + "loss": 0.9851, + "step": 13801 + }, + { + "epoch": 0.89, + "learning_rate": 1.7060263208742138e-05, + "loss": 1.1033, + "step": 13802 + }, + { + "epoch": 0.89, + "learning_rate": 1.7060048519719187e-05, + "loss": 1.067, + "step": 13803 + }, + { + "epoch": 0.89, + "learning_rate": 1.705983383069624e-05, + "loss": 1.1517, + "step": 13804 + }, + { + "epoch": 0.89, + "learning_rate": 1.7059619141673288e-05, + "loss": 1.0634, + "step": 13805 + }, + { + "epoch": 0.89, + "learning_rate": 1.7059404452650337e-05, + "loss": 1.0291, + "step": 13806 + }, + { + "epoch": 0.89, + "learning_rate": 1.7059189763627386e-05, + "loss": 1.0254, + "step": 13807 + }, + { + "epoch": 0.89, + "learning_rate": 1.7058975074604435e-05, + "loss": 1.1381, + "step": 13808 + }, + { + "epoch": 0.89, + "learning_rate": 1.7058760385581487e-05, + "loss": 1.1083, + "step": 13809 + }, + { + "epoch": 0.89, + "learning_rate": 1.7058545696558536e-05, + "loss": 1.1076, + "step": 13810 + }, + { + "epoch": 0.89, + "learning_rate": 1.705833100753559e-05, + "loss": 0.9693, + "step": 13811 + }, + { + "epoch": 0.89, + "learning_rate": 1.7058116318512637e-05, + "loss": 1.097, + "step": 13812 + }, + { + "epoch": 0.89, + "learning_rate": 1.7057901629489686e-05, + "loss": 1.068, + "step": 13813 + }, + { + "epoch": 0.89, + "learning_rate": 1.7057686940466735e-05, + "loss": 1.1468, + "step": 13814 + }, + { + "epoch": 0.89, + "learning_rate": 1.7057472251443784e-05, + "loss": 0.9864, + "step": 13815 + }, + { + "epoch": 0.89, + "learning_rate": 1.7057257562420833e-05, + "loss": 1.0807, + "step": 13816 + }, + { + "epoch": 0.89, + "learning_rate": 1.7057042873397886e-05, + "loss": 1.061, + "step": 13817 + }, + { + "epoch": 0.89, + "learning_rate": 1.7056828184374935e-05, + "loss": 1.0847, + "step": 13818 + }, + { + "epoch": 0.89, + "learning_rate": 1.7056613495351984e-05, + "loss": 1.1619, + "step": 13819 + }, + { + "epoch": 0.89, + "learning_rate": 1.7056398806329032e-05, + "loss": 1.1038, + "step": 13820 + }, + { + "epoch": 0.89, + "learning_rate": 1.705618411730608e-05, + "loss": 1.0329, + "step": 13821 + }, + { + "epoch": 0.89, + "learning_rate": 1.7055969428283134e-05, + "loss": 1.064, + "step": 13822 + }, + { + "epoch": 0.89, + "learning_rate": 1.7055754739260183e-05, + "loss": 1.0739, + "step": 13823 + }, + { + "epoch": 0.89, + "learning_rate": 1.7055540050237235e-05, + "loss": 0.9664, + "step": 13824 + }, + { + "epoch": 0.89, + "learning_rate": 1.7055325361214284e-05, + "loss": 1.0202, + "step": 13825 + }, + { + "epoch": 0.89, + "learning_rate": 1.7055110672191333e-05, + "loss": 1.053, + "step": 13826 + }, + { + "epoch": 0.89, + "learning_rate": 1.7054895983168382e-05, + "loss": 1.0265, + "step": 13827 + }, + { + "epoch": 0.89, + "learning_rate": 1.705468129414543e-05, + "loss": 1.0159, + "step": 13828 + }, + { + "epoch": 0.89, + "learning_rate": 1.7054466605122483e-05, + "loss": 1.1748, + "step": 13829 + }, + { + "epoch": 0.89, + "learning_rate": 1.7054251916099532e-05, + "loss": 0.9989, + "step": 13830 + }, + { + "epoch": 0.89, + "learning_rate": 1.705403722707658e-05, + "loss": 1.1025, + "step": 13831 + }, + { + "epoch": 0.89, + "learning_rate": 1.705382253805363e-05, + "loss": 1.0797, + "step": 13832 + }, + { + "epoch": 0.89, + "learning_rate": 1.705360784903068e-05, + "loss": 1.0445, + "step": 13833 + }, + { + "epoch": 0.89, + "learning_rate": 1.7053393160007728e-05, + "loss": 1.0864, + "step": 13834 + }, + { + "epoch": 0.89, + "learning_rate": 1.705317847098478e-05, + "loss": 0.9966, + "step": 13835 + }, + { + "epoch": 0.89, + "learning_rate": 1.705296378196183e-05, + "loss": 1.1319, + "step": 13836 + }, + { + "epoch": 0.89, + "learning_rate": 1.705274909293888e-05, + "loss": 1.0348, + "step": 13837 + }, + { + "epoch": 0.89, + "learning_rate": 1.705253440391593e-05, + "loss": 1.0616, + "step": 13838 + }, + { + "epoch": 0.89, + "learning_rate": 1.705231971489298e-05, + "loss": 1.0701, + "step": 13839 + }, + { + "epoch": 0.89, + "learning_rate": 1.705210502587003e-05, + "loss": 1.0753, + "step": 13840 + }, + { + "epoch": 0.89, + "learning_rate": 1.7051890336847078e-05, + "loss": 1.1398, + "step": 13841 + }, + { + "epoch": 0.89, + "learning_rate": 1.705167564782413e-05, + "loss": 1.0937, + "step": 13842 + }, + { + "epoch": 0.89, + "learning_rate": 1.705146095880118e-05, + "loss": 1.1248, + "step": 13843 + }, + { + "epoch": 0.89, + "learning_rate": 1.7051246269778228e-05, + "loss": 1.003, + "step": 13844 + }, + { + "epoch": 0.89, + "learning_rate": 1.7051031580755277e-05, + "loss": 0.9958, + "step": 13845 + }, + { + "epoch": 0.89, + "learning_rate": 1.7050816891732326e-05, + "loss": 1.0595, + "step": 13846 + }, + { + "epoch": 0.89, + "learning_rate": 1.7050602202709375e-05, + "loss": 1.0842, + "step": 13847 + }, + { + "epoch": 0.89, + "learning_rate": 1.7050387513686427e-05, + "loss": 1.0346, + "step": 13848 + }, + { + "epoch": 0.89, + "learning_rate": 1.7050172824663476e-05, + "loss": 1.0335, + "step": 13849 + }, + { + "epoch": 0.89, + "learning_rate": 1.7049958135640525e-05, + "loss": 1.0396, + "step": 13850 + }, + { + "epoch": 0.89, + "learning_rate": 1.7049743446617574e-05, + "loss": 1.106, + "step": 13851 + }, + { + "epoch": 0.89, + "learning_rate": 1.7049528757594626e-05, + "loss": 1.0029, + "step": 13852 + }, + { + "epoch": 0.89, + "learning_rate": 1.7049314068571675e-05, + "loss": 1.0395, + "step": 13853 + }, + { + "epoch": 0.89, + "learning_rate": 1.7049099379548724e-05, + "loss": 1.0894, + "step": 13854 + }, + { + "epoch": 0.89, + "learning_rate": 1.7048884690525777e-05, + "loss": 1.065, + "step": 13855 + }, + { + "epoch": 0.89, + "learning_rate": 1.7048670001502826e-05, + "loss": 1.0534, + "step": 13856 + }, + { + "epoch": 0.89, + "learning_rate": 1.7048455312479875e-05, + "loss": 1.1139, + "step": 13857 + }, + { + "epoch": 0.89, + "learning_rate": 1.7048240623456924e-05, + "loss": 1.0539, + "step": 13858 + }, + { + "epoch": 0.89, + "learning_rate": 1.7048025934433973e-05, + "loss": 1.0262, + "step": 13859 + }, + { + "epoch": 0.89, + "learning_rate": 1.7047811245411025e-05, + "loss": 1.0546, + "step": 13860 + }, + { + "epoch": 0.89, + "learning_rate": 1.7047596556388074e-05, + "loss": 1.0334, + "step": 13861 + }, + { + "epoch": 0.89, + "learning_rate": 1.7047381867365123e-05, + "loss": 0.9852, + "step": 13862 + }, + { + "epoch": 0.89, + "learning_rate": 1.7047167178342172e-05, + "loss": 1.0182, + "step": 13863 + }, + { + "epoch": 0.89, + "learning_rate": 1.704695248931922e-05, + "loss": 1.0927, + "step": 13864 + }, + { + "epoch": 0.89, + "learning_rate": 1.704673780029627e-05, + "loss": 1.027, + "step": 13865 + }, + { + "epoch": 0.89, + "learning_rate": 1.7046523111273322e-05, + "loss": 1.0626, + "step": 13866 + }, + { + "epoch": 0.89, + "learning_rate": 1.704630842225037e-05, + "loss": 1.0505, + "step": 13867 + }, + { + "epoch": 0.89, + "learning_rate": 1.7046093733227423e-05, + "loss": 1.0277, + "step": 13868 + }, + { + "epoch": 0.89, + "learning_rate": 1.7045879044204472e-05, + "loss": 0.9993, + "step": 13869 + }, + { + "epoch": 0.89, + "learning_rate": 1.704566435518152e-05, + "loss": 1.0736, + "step": 13870 + }, + { + "epoch": 0.89, + "learning_rate": 1.704544966615857e-05, + "loss": 0.9479, + "step": 13871 + }, + { + "epoch": 0.89, + "learning_rate": 1.704523497713562e-05, + "loss": 1.0717, + "step": 13872 + }, + { + "epoch": 0.89, + "learning_rate": 1.704502028811267e-05, + "loss": 1.026, + "step": 13873 + }, + { + "epoch": 0.89, + "learning_rate": 1.704480559908972e-05, + "loss": 1.0014, + "step": 13874 + }, + { + "epoch": 0.89, + "learning_rate": 1.704459091006677e-05, + "loss": 1.0579, + "step": 13875 + }, + { + "epoch": 0.89, + "learning_rate": 1.704437622104382e-05, + "loss": 1.0464, + "step": 13876 + }, + { + "epoch": 0.89, + "learning_rate": 1.7044161532020868e-05, + "loss": 1.1289, + "step": 13877 + }, + { + "epoch": 0.89, + "learning_rate": 1.704394684299792e-05, + "loss": 1.0309, + "step": 13878 + }, + { + "epoch": 0.89, + "learning_rate": 1.704373215397497e-05, + "loss": 1.01, + "step": 13879 + }, + { + "epoch": 0.89, + "learning_rate": 1.7043517464952018e-05, + "loss": 1.0278, + "step": 13880 + }, + { + "epoch": 0.89, + "learning_rate": 1.704330277592907e-05, + "loss": 1.0443, + "step": 13881 + }, + { + "epoch": 0.89, + "learning_rate": 1.704308808690612e-05, + "loss": 1.0502, + "step": 13882 + }, + { + "epoch": 0.89, + "learning_rate": 1.7042873397883168e-05, + "loss": 1.0364, + "step": 13883 + }, + { + "epoch": 0.89, + "learning_rate": 1.7042658708860217e-05, + "loss": 1.0884, + "step": 13884 + }, + { + "epoch": 0.89, + "learning_rate": 1.7042444019837266e-05, + "loss": 1.0522, + "step": 13885 + }, + { + "epoch": 0.89, + "learning_rate": 1.704222933081432e-05, + "loss": 1.0175, + "step": 13886 + }, + { + "epoch": 0.89, + "learning_rate": 1.7042014641791367e-05, + "loss": 1.0539, + "step": 13887 + }, + { + "epoch": 0.89, + "learning_rate": 1.7041799952768416e-05, + "loss": 1.0317, + "step": 13888 + }, + { + "epoch": 0.89, + "learning_rate": 1.7041585263745465e-05, + "loss": 1.0906, + "step": 13889 + }, + { + "epoch": 0.89, + "learning_rate": 1.7041370574722514e-05, + "loss": 1.1201, + "step": 13890 + }, + { + "epoch": 0.89, + "learning_rate": 1.7041155885699567e-05, + "loss": 1.0652, + "step": 13891 + }, + { + "epoch": 0.89, + "learning_rate": 1.7040941196676616e-05, + "loss": 1.0807, + "step": 13892 + }, + { + "epoch": 0.89, + "learning_rate": 1.7040726507653665e-05, + "loss": 1.0914, + "step": 13893 + }, + { + "epoch": 0.89, + "learning_rate": 1.7040511818630713e-05, + "loss": 1.0407, + "step": 13894 + }, + { + "epoch": 0.89, + "learning_rate": 1.7040297129607762e-05, + "loss": 0.981, + "step": 13895 + }, + { + "epoch": 0.89, + "learning_rate": 1.7040082440584815e-05, + "loss": 1.051, + "step": 13896 + }, + { + "epoch": 0.89, + "learning_rate": 1.7039867751561864e-05, + "loss": 1.1668, + "step": 13897 + }, + { + "epoch": 0.89, + "learning_rate": 1.7039653062538916e-05, + "loss": 1.083, + "step": 13898 + }, + { + "epoch": 0.89, + "learning_rate": 1.7039438373515965e-05, + "loss": 1.0738, + "step": 13899 + }, + { + "epoch": 0.89, + "learning_rate": 1.7039223684493014e-05, + "loss": 1.0622, + "step": 13900 + }, + { + "epoch": 0.89, + "learning_rate": 1.7039008995470063e-05, + "loss": 1.0891, + "step": 13901 + }, + { + "epoch": 0.89, + "learning_rate": 1.7038794306447112e-05, + "loss": 1.0429, + "step": 13902 + }, + { + "epoch": 0.89, + "learning_rate": 1.703857961742416e-05, + "loss": 1.054, + "step": 13903 + }, + { + "epoch": 0.89, + "learning_rate": 1.7038364928401213e-05, + "loss": 1.0293, + "step": 13904 + }, + { + "epoch": 0.89, + "learning_rate": 1.7038150239378262e-05, + "loss": 0.9842, + "step": 13905 + }, + { + "epoch": 0.89, + "learning_rate": 1.703793555035531e-05, + "loss": 1.0855, + "step": 13906 + }, + { + "epoch": 0.89, + "learning_rate": 1.703772086133236e-05, + "loss": 1.0782, + "step": 13907 + }, + { + "epoch": 0.89, + "learning_rate": 1.703750617230941e-05, + "loss": 1.0667, + "step": 13908 + }, + { + "epoch": 0.89, + "learning_rate": 1.703729148328646e-05, + "loss": 0.9492, + "step": 13909 + }, + { + "epoch": 0.89, + "learning_rate": 1.703707679426351e-05, + "loss": 1.0456, + "step": 13910 + }, + { + "epoch": 0.9, + "learning_rate": 1.7036862105240563e-05, + "loss": 1.0535, + "step": 13911 + }, + { + "epoch": 0.9, + "learning_rate": 1.7036647416217612e-05, + "loss": 1.0422, + "step": 13912 + }, + { + "epoch": 0.9, + "learning_rate": 1.703643272719466e-05, + "loss": 1.0035, + "step": 13913 + }, + { + "epoch": 0.9, + "learning_rate": 1.703621803817171e-05, + "loss": 1.0655, + "step": 13914 + }, + { + "epoch": 0.9, + "learning_rate": 1.703600334914876e-05, + "loss": 1.0769, + "step": 13915 + }, + { + "epoch": 0.9, + "learning_rate": 1.703578866012581e-05, + "loss": 1.061, + "step": 13916 + }, + { + "epoch": 0.9, + "learning_rate": 1.703557397110286e-05, + "loss": 1.084, + "step": 13917 + }, + { + "epoch": 0.9, + "learning_rate": 1.703535928207991e-05, + "loss": 1.0691, + "step": 13918 + }, + { + "epoch": 0.9, + "learning_rate": 1.7035144593056958e-05, + "loss": 1.0327, + "step": 13919 + }, + { + "epoch": 0.9, + "learning_rate": 1.7034929904034007e-05, + "loss": 0.9623, + "step": 13920 + }, + { + "epoch": 0.9, + "learning_rate": 1.7034715215011056e-05, + "loss": 1.0576, + "step": 13921 + }, + { + "epoch": 0.9, + "learning_rate": 1.7034500525988108e-05, + "loss": 1.0574, + "step": 13922 + }, + { + "epoch": 0.9, + "learning_rate": 1.7034285836965157e-05, + "loss": 1.0805, + "step": 13923 + }, + { + "epoch": 0.9, + "learning_rate": 1.7034071147942206e-05, + "loss": 1.0027, + "step": 13924 + }, + { + "epoch": 0.9, + "learning_rate": 1.703385645891926e-05, + "loss": 1.0313, + "step": 13925 + }, + { + "epoch": 0.9, + "learning_rate": 1.7033641769896307e-05, + "loss": 1.1195, + "step": 13926 + }, + { + "epoch": 0.9, + "learning_rate": 1.7033427080873356e-05, + "loss": 1.1007, + "step": 13927 + }, + { + "epoch": 0.9, + "learning_rate": 1.7033212391850405e-05, + "loss": 1.1108, + "step": 13928 + }, + { + "epoch": 0.9, + "learning_rate": 1.7032997702827458e-05, + "loss": 1.0699, + "step": 13929 + }, + { + "epoch": 0.9, + "learning_rate": 1.7032783013804507e-05, + "loss": 1.1228, + "step": 13930 + }, + { + "epoch": 0.9, + "learning_rate": 1.7032568324781556e-05, + "loss": 1.0752, + "step": 13931 + }, + { + "epoch": 0.9, + "learning_rate": 1.7032353635758605e-05, + "loss": 1.1296, + "step": 13932 + }, + { + "epoch": 0.9, + "learning_rate": 1.7032138946735654e-05, + "loss": 1.0219, + "step": 13933 + }, + { + "epoch": 0.9, + "learning_rate": 1.7031924257712703e-05, + "loss": 1.0914, + "step": 13934 + }, + { + "epoch": 0.9, + "learning_rate": 1.7031709568689755e-05, + "loss": 1.0508, + "step": 13935 + }, + { + "epoch": 0.9, + "learning_rate": 1.7031494879666804e-05, + "loss": 1.0314, + "step": 13936 + }, + { + "epoch": 0.9, + "learning_rate": 1.7031280190643853e-05, + "loss": 1.0634, + "step": 13937 + }, + { + "epoch": 0.9, + "learning_rate": 1.7031065501620902e-05, + "loss": 1.0951, + "step": 13938 + }, + { + "epoch": 0.9, + "learning_rate": 1.7030850812597954e-05, + "loss": 1.002, + "step": 13939 + }, + { + "epoch": 0.9, + "learning_rate": 1.7030636123575003e-05, + "loss": 1.0914, + "step": 13940 + }, + { + "epoch": 0.9, + "learning_rate": 1.7030421434552052e-05, + "loss": 1.0902, + "step": 13941 + }, + { + "epoch": 0.9, + "learning_rate": 1.7030206745529104e-05, + "loss": 1.0739, + "step": 13942 + }, + { + "epoch": 0.9, + "learning_rate": 1.7029992056506153e-05, + "loss": 1.0589, + "step": 13943 + }, + { + "epoch": 0.9, + "learning_rate": 1.7029777367483202e-05, + "loss": 0.9852, + "step": 13944 + }, + { + "epoch": 0.9, + "learning_rate": 1.702956267846025e-05, + "loss": 1.0288, + "step": 13945 + }, + { + "epoch": 0.9, + "learning_rate": 1.70293479894373e-05, + "loss": 1.0565, + "step": 13946 + }, + { + "epoch": 0.9, + "learning_rate": 1.7029133300414353e-05, + "loss": 1.0459, + "step": 13947 + }, + { + "epoch": 0.9, + "learning_rate": 1.70289186113914e-05, + "loss": 1.0899, + "step": 13948 + }, + { + "epoch": 0.9, + "learning_rate": 1.702870392236845e-05, + "loss": 1.0428, + "step": 13949 + }, + { + "epoch": 0.9, + "learning_rate": 1.70284892333455e-05, + "loss": 1.0112, + "step": 13950 + }, + { + "epoch": 0.9, + "learning_rate": 1.702827454432255e-05, + "loss": 1.0733, + "step": 13951 + }, + { + "epoch": 0.9, + "learning_rate": 1.7028059855299597e-05, + "loss": 1.0588, + "step": 13952 + }, + { + "epoch": 0.9, + "learning_rate": 1.702784516627665e-05, + "loss": 1.1385, + "step": 13953 + }, + { + "epoch": 0.9, + "learning_rate": 1.70276304772537e-05, + "loss": 1.1547, + "step": 13954 + }, + { + "epoch": 0.9, + "learning_rate": 1.702741578823075e-05, + "loss": 0.9848, + "step": 13955 + }, + { + "epoch": 0.9, + "learning_rate": 1.70272010992078e-05, + "loss": 1.1154, + "step": 13956 + }, + { + "epoch": 0.9, + "learning_rate": 1.702698641018485e-05, + "loss": 1.0632, + "step": 13957 + }, + { + "epoch": 0.9, + "learning_rate": 1.7026771721161898e-05, + "loss": 1.0787, + "step": 13958 + }, + { + "epoch": 0.9, + "learning_rate": 1.7026557032138947e-05, + "loss": 1.0978, + "step": 13959 + }, + { + "epoch": 0.9, + "learning_rate": 1.7026342343116e-05, + "loss": 1.0926, + "step": 13960 + }, + { + "epoch": 0.9, + "learning_rate": 1.702612765409305e-05, + "loss": 1.0276, + "step": 13961 + }, + { + "epoch": 0.9, + "learning_rate": 1.7025912965070097e-05, + "loss": 1.0751, + "step": 13962 + }, + { + "epoch": 0.9, + "learning_rate": 1.7025698276047146e-05, + "loss": 1.0057, + "step": 13963 + }, + { + "epoch": 0.9, + "learning_rate": 1.7025483587024195e-05, + "loss": 1.0562, + "step": 13964 + }, + { + "epoch": 0.9, + "learning_rate": 1.7025268898001248e-05, + "loss": 1.1517, + "step": 13965 + }, + { + "epoch": 0.9, + "learning_rate": 1.7025054208978297e-05, + "loss": 1.0792, + "step": 13966 + }, + { + "epoch": 0.9, + "learning_rate": 1.7024839519955346e-05, + "loss": 1.0411, + "step": 13967 + }, + { + "epoch": 0.9, + "learning_rate": 1.7024624830932398e-05, + "loss": 1.0008, + "step": 13968 + }, + { + "epoch": 0.9, + "learning_rate": 1.7024410141909447e-05, + "loss": 0.9245, + "step": 13969 + }, + { + "epoch": 0.9, + "learning_rate": 1.7024195452886496e-05, + "loss": 1.0118, + "step": 13970 + }, + { + "epoch": 0.9, + "learning_rate": 1.7023980763863545e-05, + "loss": 1.0327, + "step": 13971 + }, + { + "epoch": 0.9, + "learning_rate": 1.7023766074840594e-05, + "loss": 0.9957, + "step": 13972 + }, + { + "epoch": 0.9, + "learning_rate": 1.7023551385817646e-05, + "loss": 1.0768, + "step": 13973 + }, + { + "epoch": 0.9, + "learning_rate": 1.7023336696794695e-05, + "loss": 1.0626, + "step": 13974 + }, + { + "epoch": 0.9, + "learning_rate": 1.7023122007771744e-05, + "loss": 0.9983, + "step": 13975 + }, + { + "epoch": 0.9, + "learning_rate": 1.7022907318748793e-05, + "loss": 1.0432, + "step": 13976 + }, + { + "epoch": 0.9, + "learning_rate": 1.7022692629725842e-05, + "loss": 1.0228, + "step": 13977 + }, + { + "epoch": 0.9, + "learning_rate": 1.7022477940702894e-05, + "loss": 1.0826, + "step": 13978 + }, + { + "epoch": 0.9, + "learning_rate": 1.7022263251679943e-05, + "loss": 1.0358, + "step": 13979 + }, + { + "epoch": 0.9, + "learning_rate": 1.7022048562656992e-05, + "loss": 1.0598, + "step": 13980 + }, + { + "epoch": 0.9, + "learning_rate": 1.702183387363404e-05, + "loss": 1.0196, + "step": 13981 + }, + { + "epoch": 0.9, + "learning_rate": 1.702161918461109e-05, + "loss": 1.0718, + "step": 13982 + }, + { + "epoch": 0.9, + "learning_rate": 1.7021404495588143e-05, + "loss": 1.1156, + "step": 13983 + }, + { + "epoch": 0.9, + "learning_rate": 1.702118980656519e-05, + "loss": 1.0774, + "step": 13984 + }, + { + "epoch": 0.9, + "learning_rate": 1.7020975117542244e-05, + "loss": 1.1093, + "step": 13985 + }, + { + "epoch": 0.9, + "learning_rate": 1.7020760428519293e-05, + "loss": 1.0597, + "step": 13986 + }, + { + "epoch": 0.9, + "learning_rate": 1.7020545739496342e-05, + "loss": 1.0191, + "step": 13987 + }, + { + "epoch": 0.9, + "learning_rate": 1.702033105047339e-05, + "loss": 0.9897, + "step": 13988 + }, + { + "epoch": 0.9, + "learning_rate": 1.702011636145044e-05, + "loss": 1.0374, + "step": 13989 + }, + { + "epoch": 0.9, + "learning_rate": 1.701990167242749e-05, + "loss": 0.9578, + "step": 13990 + }, + { + "epoch": 0.9, + "learning_rate": 1.701968698340454e-05, + "loss": 1.0715, + "step": 13991 + }, + { + "epoch": 0.9, + "learning_rate": 1.701947229438159e-05, + "loss": 1.0382, + "step": 13992 + }, + { + "epoch": 0.9, + "learning_rate": 1.701925760535864e-05, + "loss": 1.1252, + "step": 13993 + }, + { + "epoch": 0.9, + "learning_rate": 1.7019042916335688e-05, + "loss": 1.1005, + "step": 13994 + }, + { + "epoch": 0.9, + "learning_rate": 1.7018828227312737e-05, + "loss": 1.0093, + "step": 13995 + }, + { + "epoch": 0.9, + "learning_rate": 1.701861353828979e-05, + "loss": 1.084, + "step": 13996 + }, + { + "epoch": 0.9, + "learning_rate": 1.7018398849266838e-05, + "loss": 1.0553, + "step": 13997 + }, + { + "epoch": 0.9, + "learning_rate": 1.701818416024389e-05, + "loss": 1.1366, + "step": 13998 + }, + { + "epoch": 0.9, + "learning_rate": 1.701796947122094e-05, + "loss": 1.0221, + "step": 13999 + }, + { + "epoch": 0.9, + "learning_rate": 1.701775478219799e-05, + "loss": 1.1431, + "step": 14000 + }, + { + "epoch": 0.9, + "eval_laion/cider": 0.4772414414511665, + "eval_laion/meteor": 0.37395394573075413, + "eval_laion/pred_num_words": 36.90234375, + "eval_laion/sacrebleu": 15.273434638635713, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0725994110107422, + "eval_runtime": 108.2253, + "eval_samples_per_second": 4.611, + "eval_steps_per_second": 0.296, + "step": 14000 + }, + { + "epoch": 0.9, + "learning_rate": 1.7017540093175037e-05, + "loss": 1.079, + "step": 14001 + }, + { + "epoch": 0.9, + "learning_rate": 1.7017325404152086e-05, + "loss": 1.0699, + "step": 14002 + }, + { + "epoch": 0.9, + "learning_rate": 1.7017110715129135e-05, + "loss": 1.0903, + "step": 14003 + }, + { + "epoch": 0.9, + "learning_rate": 1.7016896026106188e-05, + "loss": 1.1186, + "step": 14004 + }, + { + "epoch": 0.9, + "learning_rate": 1.7016681337083237e-05, + "loss": 1.1099, + "step": 14005 + }, + { + "epoch": 0.9, + "learning_rate": 1.7016466648060286e-05, + "loss": 1.0234, + "step": 14006 + }, + { + "epoch": 0.9, + "learning_rate": 1.7016251959037335e-05, + "loss": 1.1135, + "step": 14007 + }, + { + "epoch": 0.9, + "learning_rate": 1.7016037270014384e-05, + "loss": 1.0211, + "step": 14008 + }, + { + "epoch": 0.9, + "learning_rate": 1.7015822580991436e-05, + "loss": 0.9972, + "step": 14009 + }, + { + "epoch": 0.9, + "learning_rate": 1.7015607891968485e-05, + "loss": 1.043, + "step": 14010 + }, + { + "epoch": 0.9, + "learning_rate": 1.7015393202945534e-05, + "loss": 1.1342, + "step": 14011 + }, + { + "epoch": 0.9, + "learning_rate": 1.7015178513922586e-05, + "loss": 1.126, + "step": 14012 + }, + { + "epoch": 0.9, + "learning_rate": 1.7014963824899635e-05, + "loss": 1.0878, + "step": 14013 + }, + { + "epoch": 0.9, + "learning_rate": 1.7014749135876684e-05, + "loss": 1.0132, + "step": 14014 + }, + { + "epoch": 0.9, + "learning_rate": 1.7014534446853733e-05, + "loss": 1.0477, + "step": 14015 + }, + { + "epoch": 0.9, + "learning_rate": 1.7014319757830785e-05, + "loss": 1.0683, + "step": 14016 + }, + { + "epoch": 0.9, + "learning_rate": 1.7014105068807834e-05, + "loss": 0.9777, + "step": 14017 + }, + { + "epoch": 0.9, + "learning_rate": 1.7013890379784883e-05, + "loss": 1.0973, + "step": 14018 + }, + { + "epoch": 0.9, + "learning_rate": 1.7013675690761932e-05, + "loss": 1.1057, + "step": 14019 + }, + { + "epoch": 0.9, + "learning_rate": 1.701346100173898e-05, + "loss": 1.0508, + "step": 14020 + }, + { + "epoch": 0.9, + "learning_rate": 1.701324631271603e-05, + "loss": 1.0918, + "step": 14021 + }, + { + "epoch": 0.9, + "learning_rate": 1.7013031623693083e-05, + "loss": 0.9936, + "step": 14022 + }, + { + "epoch": 0.9, + "learning_rate": 1.701281693467013e-05, + "loss": 0.9743, + "step": 14023 + }, + { + "epoch": 0.9, + "learning_rate": 1.701260224564718e-05, + "loss": 1.0412, + "step": 14024 + }, + { + "epoch": 0.9, + "learning_rate": 1.701238755662423e-05, + "loss": 0.9844, + "step": 14025 + }, + { + "epoch": 0.9, + "learning_rate": 1.701217286760128e-05, + "loss": 0.9995, + "step": 14026 + }, + { + "epoch": 0.9, + "learning_rate": 1.701195817857833e-05, + "loss": 1.0873, + "step": 14027 + }, + { + "epoch": 0.9, + "learning_rate": 1.701174348955538e-05, + "loss": 1.0601, + "step": 14028 + }, + { + "epoch": 0.9, + "learning_rate": 1.7011528800532432e-05, + "loss": 1.1333, + "step": 14029 + }, + { + "epoch": 0.9, + "learning_rate": 1.701131411150948e-05, + "loss": 1.1117, + "step": 14030 + }, + { + "epoch": 0.9, + "learning_rate": 1.701109942248653e-05, + "loss": 1.0809, + "step": 14031 + }, + { + "epoch": 0.9, + "learning_rate": 1.701088473346358e-05, + "loss": 1.0337, + "step": 14032 + }, + { + "epoch": 0.9, + "learning_rate": 1.7010670044440628e-05, + "loss": 1.035, + "step": 14033 + }, + { + "epoch": 0.9, + "learning_rate": 1.701045535541768e-05, + "loss": 0.9834, + "step": 14034 + }, + { + "epoch": 0.9, + "learning_rate": 1.701024066639473e-05, + "loss": 1.0718, + "step": 14035 + }, + { + "epoch": 0.9, + "learning_rate": 1.7010025977371778e-05, + "loss": 1.1642, + "step": 14036 + }, + { + "epoch": 0.9, + "learning_rate": 1.7009811288348827e-05, + "loss": 1.0291, + "step": 14037 + }, + { + "epoch": 0.9, + "learning_rate": 1.7009596599325876e-05, + "loss": 1.074, + "step": 14038 + }, + { + "epoch": 0.9, + "learning_rate": 1.7009381910302925e-05, + "loss": 0.9681, + "step": 14039 + }, + { + "epoch": 0.9, + "learning_rate": 1.7009167221279978e-05, + "loss": 1.0617, + "step": 14040 + }, + { + "epoch": 0.9, + "learning_rate": 1.7008952532257027e-05, + "loss": 1.0898, + "step": 14041 + }, + { + "epoch": 0.9, + "learning_rate": 1.700873784323408e-05, + "loss": 0.9988, + "step": 14042 + }, + { + "epoch": 0.9, + "learning_rate": 1.7008523154211128e-05, + "loss": 1.083, + "step": 14043 + }, + { + "epoch": 0.9, + "learning_rate": 1.7008308465188177e-05, + "loss": 1.0064, + "step": 14044 + }, + { + "epoch": 0.9, + "learning_rate": 1.7008093776165226e-05, + "loss": 1.1066, + "step": 14045 + }, + { + "epoch": 0.9, + "learning_rate": 1.7007879087142275e-05, + "loss": 0.9756, + "step": 14046 + }, + { + "epoch": 0.9, + "learning_rate": 1.7007664398119327e-05, + "loss": 1.0269, + "step": 14047 + }, + { + "epoch": 0.9, + "learning_rate": 1.7007449709096376e-05, + "loss": 1.0817, + "step": 14048 + }, + { + "epoch": 0.9, + "learning_rate": 1.7007235020073425e-05, + "loss": 1.1268, + "step": 14049 + }, + { + "epoch": 0.9, + "learning_rate": 1.7007020331050474e-05, + "loss": 1.0355, + "step": 14050 + }, + { + "epoch": 0.9, + "learning_rate": 1.7006805642027523e-05, + "loss": 0.9811, + "step": 14051 + }, + { + "epoch": 0.9, + "learning_rate": 1.7006590953004575e-05, + "loss": 1.0017, + "step": 14052 + }, + { + "epoch": 0.9, + "learning_rate": 1.7006376263981624e-05, + "loss": 1.0303, + "step": 14053 + }, + { + "epoch": 0.9, + "learning_rate": 1.7006161574958673e-05, + "loss": 1.0644, + "step": 14054 + }, + { + "epoch": 0.9, + "learning_rate": 1.7005946885935722e-05, + "loss": 1.0047, + "step": 14055 + }, + { + "epoch": 0.9, + "learning_rate": 1.7005732196912775e-05, + "loss": 0.9662, + "step": 14056 + }, + { + "epoch": 0.9, + "learning_rate": 1.7005517507889824e-05, + "loss": 1.1682, + "step": 14057 + }, + { + "epoch": 0.9, + "learning_rate": 1.7005302818866872e-05, + "loss": 1.0222, + "step": 14058 + }, + { + "epoch": 0.9, + "learning_rate": 1.700508812984392e-05, + "loss": 1.0272, + "step": 14059 + }, + { + "epoch": 0.9, + "learning_rate": 1.7004873440820974e-05, + "loss": 1.0626, + "step": 14060 + }, + { + "epoch": 0.9, + "learning_rate": 1.7004658751798023e-05, + "loss": 1.0146, + "step": 14061 + }, + { + "epoch": 0.9, + "learning_rate": 1.7004444062775072e-05, + "loss": 1.061, + "step": 14062 + }, + { + "epoch": 0.9, + "learning_rate": 1.700422937375212e-05, + "loss": 1.011, + "step": 14063 + }, + { + "epoch": 0.9, + "learning_rate": 1.700401468472917e-05, + "loss": 1.0685, + "step": 14064 + }, + { + "epoch": 0.9, + "learning_rate": 1.7003799995706222e-05, + "loss": 1.0788, + "step": 14065 + }, + { + "epoch": 0.9, + "learning_rate": 1.700358530668327e-05, + "loss": 0.9995, + "step": 14066 + }, + { + "epoch": 0.91, + "learning_rate": 1.700337061766032e-05, + "loss": 1.0865, + "step": 14067 + }, + { + "epoch": 0.91, + "learning_rate": 1.700315592863737e-05, + "loss": 1.0971, + "step": 14068 + }, + { + "epoch": 0.91, + "learning_rate": 1.7002941239614418e-05, + "loss": 1.0609, + "step": 14069 + }, + { + "epoch": 0.91, + "learning_rate": 1.700272655059147e-05, + "loss": 1.088, + "step": 14070 + }, + { + "epoch": 0.91, + "learning_rate": 1.700251186156852e-05, + "loss": 0.988, + "step": 14071 + }, + { + "epoch": 0.91, + "learning_rate": 1.700229717254557e-05, + "loss": 1.1512, + "step": 14072 + }, + { + "epoch": 0.91, + "learning_rate": 1.700208248352262e-05, + "loss": 1.0563, + "step": 14073 + }, + { + "epoch": 0.91, + "learning_rate": 1.700186779449967e-05, + "loss": 1.1901, + "step": 14074 + }, + { + "epoch": 0.91, + "learning_rate": 1.700165310547672e-05, + "loss": 1.0774, + "step": 14075 + }, + { + "epoch": 0.91, + "learning_rate": 1.7001438416453767e-05, + "loss": 1.0744, + "step": 14076 + }, + { + "epoch": 0.91, + "learning_rate": 1.7001223727430816e-05, + "loss": 1.0261, + "step": 14077 + }, + { + "epoch": 0.91, + "learning_rate": 1.700100903840787e-05, + "loss": 1.0521, + "step": 14078 + }, + { + "epoch": 0.91, + "learning_rate": 1.7000794349384918e-05, + "loss": 1.0244, + "step": 14079 + }, + { + "epoch": 0.91, + "learning_rate": 1.7000579660361967e-05, + "loss": 1.1057, + "step": 14080 + }, + { + "epoch": 0.91, + "learning_rate": 1.7000364971339016e-05, + "loss": 0.9812, + "step": 14081 + }, + { + "epoch": 0.91, + "learning_rate": 1.7000150282316065e-05, + "loss": 1.0153, + "step": 14082 + }, + { + "epoch": 0.91, + "learning_rate": 1.6999935593293117e-05, + "loss": 1.0519, + "step": 14083 + }, + { + "epoch": 0.91, + "learning_rate": 1.6999720904270166e-05, + "loss": 1.0148, + "step": 14084 + }, + { + "epoch": 0.91, + "learning_rate": 1.6999506215247218e-05, + "loss": 1.1206, + "step": 14085 + }, + { + "epoch": 0.91, + "learning_rate": 1.6999291526224267e-05, + "loss": 1.146, + "step": 14086 + }, + { + "epoch": 0.91, + "learning_rate": 1.6999076837201316e-05, + "loss": 0.9924, + "step": 14087 + }, + { + "epoch": 0.91, + "learning_rate": 1.6998862148178365e-05, + "loss": 1.1652, + "step": 14088 + }, + { + "epoch": 0.91, + "learning_rate": 1.6998647459155414e-05, + "loss": 1.083, + "step": 14089 + }, + { + "epoch": 0.91, + "learning_rate": 1.6998432770132463e-05, + "loss": 1.0392, + "step": 14090 + }, + { + "epoch": 0.91, + "learning_rate": 1.6998218081109515e-05, + "loss": 0.9776, + "step": 14091 + }, + { + "epoch": 0.91, + "learning_rate": 1.6998003392086564e-05, + "loss": 1.0249, + "step": 14092 + }, + { + "epoch": 0.91, + "learning_rate": 1.6997788703063613e-05, + "loss": 1.1091, + "step": 14093 + }, + { + "epoch": 0.91, + "learning_rate": 1.6997574014040662e-05, + "loss": 1.0033, + "step": 14094 + }, + { + "epoch": 0.91, + "learning_rate": 1.699735932501771e-05, + "loss": 1.0589, + "step": 14095 + }, + { + "epoch": 0.91, + "learning_rate": 1.6997144635994764e-05, + "loss": 1.083, + "step": 14096 + }, + { + "epoch": 0.91, + "learning_rate": 1.6996929946971813e-05, + "loss": 1.0924, + "step": 14097 + }, + { + "epoch": 0.91, + "learning_rate": 1.699671525794886e-05, + "loss": 1.1447, + "step": 14098 + }, + { + "epoch": 0.91, + "learning_rate": 1.6996500568925914e-05, + "loss": 1.0728, + "step": 14099 + }, + { + "epoch": 0.91, + "learning_rate": 1.6996285879902963e-05, + "loss": 0.9985, + "step": 14100 + }, + { + "epoch": 0.91, + "learning_rate": 1.6996071190880012e-05, + "loss": 1.1606, + "step": 14101 + }, + { + "epoch": 0.91, + "learning_rate": 1.699585650185706e-05, + "loss": 1.1087, + "step": 14102 + }, + { + "epoch": 0.91, + "learning_rate": 1.6995641812834113e-05, + "loss": 1.1186, + "step": 14103 + }, + { + "epoch": 0.91, + "learning_rate": 1.6995427123811162e-05, + "loss": 1.0251, + "step": 14104 + }, + { + "epoch": 0.91, + "learning_rate": 1.699521243478821e-05, + "loss": 1.0372, + "step": 14105 + }, + { + "epoch": 0.91, + "learning_rate": 1.699499774576526e-05, + "loss": 0.9679, + "step": 14106 + }, + { + "epoch": 0.91, + "learning_rate": 1.699478305674231e-05, + "loss": 0.9689, + "step": 14107 + }, + { + "epoch": 0.91, + "learning_rate": 1.6994568367719358e-05, + "loss": 1.004, + "step": 14108 + }, + { + "epoch": 0.91, + "learning_rate": 1.699435367869641e-05, + "loss": 1.0896, + "step": 14109 + }, + { + "epoch": 0.91, + "learning_rate": 1.699413898967346e-05, + "loss": 1.0387, + "step": 14110 + }, + { + "epoch": 0.91, + "learning_rate": 1.6993924300650508e-05, + "loss": 1.0488, + "step": 14111 + }, + { + "epoch": 0.91, + "learning_rate": 1.6993709611627557e-05, + "loss": 1.0972, + "step": 14112 + }, + { + "epoch": 0.91, + "learning_rate": 1.6993494922604606e-05, + "loss": 1.0683, + "step": 14113 + }, + { + "epoch": 0.91, + "learning_rate": 1.699328023358166e-05, + "loss": 1.078, + "step": 14114 + }, + { + "epoch": 0.91, + "learning_rate": 1.6993065544558708e-05, + "loss": 1.039, + "step": 14115 + }, + { + "epoch": 0.91, + "learning_rate": 1.699285085553576e-05, + "loss": 1.072, + "step": 14116 + }, + { + "epoch": 0.91, + "learning_rate": 1.699263616651281e-05, + "loss": 1.0093, + "step": 14117 + }, + { + "epoch": 0.91, + "learning_rate": 1.6992421477489858e-05, + "loss": 1.1359, + "step": 14118 + }, + { + "epoch": 0.91, + "learning_rate": 1.6992206788466907e-05, + "loss": 1.0438, + "step": 14119 + }, + { + "epoch": 0.91, + "learning_rate": 1.6991992099443956e-05, + "loss": 1.0171, + "step": 14120 + }, + { + "epoch": 0.91, + "learning_rate": 1.6991777410421008e-05, + "loss": 1.022, + "step": 14121 + }, + { + "epoch": 0.91, + "learning_rate": 1.6991562721398057e-05, + "loss": 1.0913, + "step": 14122 + }, + { + "epoch": 0.91, + "learning_rate": 1.6991348032375106e-05, + "loss": 1.0015, + "step": 14123 + }, + { + "epoch": 0.91, + "learning_rate": 1.6991133343352155e-05, + "loss": 1.0491, + "step": 14124 + }, + { + "epoch": 0.91, + "learning_rate": 1.6990918654329204e-05, + "loss": 1.0239, + "step": 14125 + }, + { + "epoch": 0.91, + "learning_rate": 1.6990703965306253e-05, + "loss": 0.997, + "step": 14126 + }, + { + "epoch": 0.91, + "learning_rate": 1.6990489276283305e-05, + "loss": 1.0391, + "step": 14127 + }, + { + "epoch": 0.91, + "learning_rate": 1.6990274587260354e-05, + "loss": 1.0812, + "step": 14128 + }, + { + "epoch": 0.91, + "learning_rate": 1.6990059898237407e-05, + "loss": 1.0308, + "step": 14129 + }, + { + "epoch": 0.91, + "learning_rate": 1.6989845209214456e-05, + "loss": 1.0135, + "step": 14130 + }, + { + "epoch": 0.91, + "learning_rate": 1.6989630520191505e-05, + "loss": 0.949, + "step": 14131 + }, + { + "epoch": 0.91, + "learning_rate": 1.6989415831168553e-05, + "loss": 1.0288, + "step": 14132 + }, + { + "epoch": 0.91, + "learning_rate": 1.6989201142145602e-05, + "loss": 1.0534, + "step": 14133 + }, + { + "epoch": 0.91, + "learning_rate": 1.6988986453122655e-05, + "loss": 1.1001, + "step": 14134 + }, + { + "epoch": 0.91, + "learning_rate": 1.6988771764099704e-05, + "loss": 1.1529, + "step": 14135 + }, + { + "epoch": 0.91, + "learning_rate": 1.6988557075076753e-05, + "loss": 1.0097, + "step": 14136 + }, + { + "epoch": 0.91, + "learning_rate": 1.69883423860538e-05, + "loss": 1.0885, + "step": 14137 + }, + { + "epoch": 0.91, + "learning_rate": 1.698812769703085e-05, + "loss": 1.0806, + "step": 14138 + }, + { + "epoch": 0.91, + "learning_rate": 1.6987913008007903e-05, + "loss": 1.0444, + "step": 14139 + }, + { + "epoch": 0.91, + "learning_rate": 1.6987698318984952e-05, + "loss": 0.9495, + "step": 14140 + }, + { + "epoch": 0.91, + "learning_rate": 1.6987483629962e-05, + "loss": 1.0913, + "step": 14141 + }, + { + "epoch": 0.91, + "learning_rate": 1.698726894093905e-05, + "loss": 1.0351, + "step": 14142 + }, + { + "epoch": 0.91, + "learning_rate": 1.6987054251916102e-05, + "loss": 1.0524, + "step": 14143 + }, + { + "epoch": 0.91, + "learning_rate": 1.698683956289315e-05, + "loss": 1.0632, + "step": 14144 + }, + { + "epoch": 0.91, + "learning_rate": 1.69866248738702e-05, + "loss": 1.0675, + "step": 14145 + }, + { + "epoch": 0.91, + "learning_rate": 1.698641018484725e-05, + "loss": 1.091, + "step": 14146 + }, + { + "epoch": 0.91, + "learning_rate": 1.69861954958243e-05, + "loss": 1.0419, + "step": 14147 + }, + { + "epoch": 0.91, + "learning_rate": 1.698598080680135e-05, + "loss": 1.0366, + "step": 14148 + }, + { + "epoch": 0.91, + "learning_rate": 1.69857661177784e-05, + "loss": 1.0445, + "step": 14149 + }, + { + "epoch": 0.91, + "learning_rate": 1.698555142875545e-05, + "loss": 1.0282, + "step": 14150 + }, + { + "epoch": 0.91, + "learning_rate": 1.6985336739732497e-05, + "loss": 1.176, + "step": 14151 + }, + { + "epoch": 0.91, + "learning_rate": 1.698512205070955e-05, + "loss": 1.0756, + "step": 14152 + }, + { + "epoch": 0.91, + "learning_rate": 1.69849073616866e-05, + "loss": 1.0383, + "step": 14153 + }, + { + "epoch": 0.91, + "learning_rate": 1.6984692672663648e-05, + "loss": 1.0633, + "step": 14154 + }, + { + "epoch": 0.91, + "learning_rate": 1.6984477983640697e-05, + "loss": 1.0233, + "step": 14155 + }, + { + "epoch": 0.91, + "learning_rate": 1.6984263294617746e-05, + "loss": 0.9657, + "step": 14156 + }, + { + "epoch": 0.91, + "learning_rate": 1.6984048605594798e-05, + "loss": 1.0755, + "step": 14157 + }, + { + "epoch": 0.91, + "learning_rate": 1.6983833916571847e-05, + "loss": 1.1207, + "step": 14158 + }, + { + "epoch": 0.91, + "learning_rate": 1.69836192275489e-05, + "loss": 1.1103, + "step": 14159 + }, + { + "epoch": 0.91, + "learning_rate": 1.6983404538525948e-05, + "loss": 1.0884, + "step": 14160 + }, + { + "epoch": 0.91, + "learning_rate": 1.6983189849502997e-05, + "loss": 1.1102, + "step": 14161 + }, + { + "epoch": 0.91, + "learning_rate": 1.6982975160480046e-05, + "loss": 1.0596, + "step": 14162 + }, + { + "epoch": 0.91, + "learning_rate": 1.6982760471457095e-05, + "loss": 1.0628, + "step": 14163 + }, + { + "epoch": 0.91, + "learning_rate": 1.6982545782434144e-05, + "loss": 1.0809, + "step": 14164 + }, + { + "epoch": 0.91, + "learning_rate": 1.6982331093411196e-05, + "loss": 1.0178, + "step": 14165 + }, + { + "epoch": 0.91, + "learning_rate": 1.6982116404388245e-05, + "loss": 1.0541, + "step": 14166 + }, + { + "epoch": 0.91, + "learning_rate": 1.6981901715365294e-05, + "loss": 1.077, + "step": 14167 + }, + { + "epoch": 0.91, + "learning_rate": 1.6981687026342343e-05, + "loss": 1.0351, + "step": 14168 + }, + { + "epoch": 0.91, + "learning_rate": 1.6981472337319392e-05, + "loss": 1.0241, + "step": 14169 + }, + { + "epoch": 0.91, + "learning_rate": 1.6981257648296445e-05, + "loss": 0.9924, + "step": 14170 + }, + { + "epoch": 0.91, + "learning_rate": 1.6981042959273494e-05, + "loss": 1.0373, + "step": 14171 + }, + { + "epoch": 0.91, + "learning_rate": 1.6980828270250546e-05, + "loss": 1.0751, + "step": 14172 + }, + { + "epoch": 0.91, + "learning_rate": 1.6980613581227595e-05, + "loss": 0.9898, + "step": 14173 + }, + { + "epoch": 0.91, + "learning_rate": 1.6980398892204644e-05, + "loss": 1.2109, + "step": 14174 + }, + { + "epoch": 0.91, + "learning_rate": 1.6980184203181693e-05, + "loss": 1.0742, + "step": 14175 + }, + { + "epoch": 0.91, + "learning_rate": 1.6979969514158742e-05, + "loss": 1.124, + "step": 14176 + }, + { + "epoch": 0.91, + "learning_rate": 1.697975482513579e-05, + "loss": 1.0306, + "step": 14177 + }, + { + "epoch": 0.91, + "learning_rate": 1.6979540136112843e-05, + "loss": 1.0498, + "step": 14178 + }, + { + "epoch": 0.91, + "learning_rate": 1.6979325447089892e-05, + "loss": 1.0073, + "step": 14179 + }, + { + "epoch": 0.91, + "learning_rate": 1.697911075806694e-05, + "loss": 1.0875, + "step": 14180 + }, + { + "epoch": 0.91, + "learning_rate": 1.697889606904399e-05, + "loss": 1.0847, + "step": 14181 + }, + { + "epoch": 0.91, + "learning_rate": 1.697868138002104e-05, + "loss": 1.0824, + "step": 14182 + }, + { + "epoch": 0.91, + "learning_rate": 1.697846669099809e-05, + "loss": 0.9832, + "step": 14183 + }, + { + "epoch": 0.91, + "learning_rate": 1.697825200197514e-05, + "loss": 0.9898, + "step": 14184 + }, + { + "epoch": 0.91, + "learning_rate": 1.697803731295219e-05, + "loss": 1.0832, + "step": 14185 + }, + { + "epoch": 0.91, + "learning_rate": 1.697782262392924e-05, + "loss": 0.955, + "step": 14186 + }, + { + "epoch": 0.91, + "learning_rate": 1.697760793490629e-05, + "loss": 1.0581, + "step": 14187 + }, + { + "epoch": 0.91, + "learning_rate": 1.697739324588334e-05, + "loss": 1.1322, + "step": 14188 + }, + { + "epoch": 0.91, + "learning_rate": 1.697717855686039e-05, + "loss": 1.0584, + "step": 14189 + }, + { + "epoch": 0.91, + "learning_rate": 1.697696386783744e-05, + "loss": 1.1021, + "step": 14190 + }, + { + "epoch": 0.91, + "learning_rate": 1.697674917881449e-05, + "loss": 1.0706, + "step": 14191 + }, + { + "epoch": 0.91, + "learning_rate": 1.697653448979154e-05, + "loss": 1.0561, + "step": 14192 + }, + { + "epoch": 0.91, + "learning_rate": 1.6976319800768588e-05, + "loss": 0.9671, + "step": 14193 + }, + { + "epoch": 0.91, + "learning_rate": 1.6976105111745637e-05, + "loss": 1.0218, + "step": 14194 + }, + { + "epoch": 0.91, + "learning_rate": 1.6975890422722686e-05, + "loss": 1.1055, + "step": 14195 + }, + { + "epoch": 0.91, + "learning_rate": 1.6975675733699738e-05, + "loss": 1.1121, + "step": 14196 + }, + { + "epoch": 0.91, + "learning_rate": 1.6975461044676787e-05, + "loss": 1.1163, + "step": 14197 + }, + { + "epoch": 0.91, + "learning_rate": 1.6975246355653836e-05, + "loss": 1.0808, + "step": 14198 + }, + { + "epoch": 0.91, + "learning_rate": 1.6975031666630885e-05, + "loss": 1.0115, + "step": 14199 + }, + { + "epoch": 0.91, + "learning_rate": 1.6974816977607934e-05, + "loss": 1.0551, + "step": 14200 + }, + { + "epoch": 0.91, + "learning_rate": 1.6974602288584986e-05, + "loss": 1.0643, + "step": 14201 + }, + { + "epoch": 0.91, + "learning_rate": 1.6974387599562035e-05, + "loss": 1.1502, + "step": 14202 + }, + { + "epoch": 0.91, + "learning_rate": 1.6974172910539088e-05, + "loss": 1.0654, + "step": 14203 + }, + { + "epoch": 0.91, + "learning_rate": 1.6973958221516137e-05, + "loss": 1.0026, + "step": 14204 + }, + { + "epoch": 0.91, + "learning_rate": 1.6973743532493186e-05, + "loss": 1.0237, + "step": 14205 + }, + { + "epoch": 0.91, + "learning_rate": 1.6973528843470234e-05, + "loss": 1.139, + "step": 14206 + }, + { + "epoch": 0.91, + "learning_rate": 1.6973314154447283e-05, + "loss": 1.0847, + "step": 14207 + }, + { + "epoch": 0.91, + "learning_rate": 1.6973099465424336e-05, + "loss": 1.0185, + "step": 14208 + }, + { + "epoch": 0.91, + "learning_rate": 1.6972884776401385e-05, + "loss": 1.0619, + "step": 14209 + }, + { + "epoch": 0.91, + "learning_rate": 1.6972670087378434e-05, + "loss": 0.9991, + "step": 14210 + }, + { + "epoch": 0.91, + "learning_rate": 1.6972455398355483e-05, + "loss": 1.1281, + "step": 14211 + }, + { + "epoch": 0.91, + "learning_rate": 1.697224070933253e-05, + "loss": 1.0967, + "step": 14212 + }, + { + "epoch": 0.91, + "learning_rate": 1.697202602030958e-05, + "loss": 1.0127, + "step": 14213 + }, + { + "epoch": 0.91, + "learning_rate": 1.6971811331286633e-05, + "loss": 1.128, + "step": 14214 + }, + { + "epoch": 0.91, + "learning_rate": 1.6971596642263682e-05, + "loss": 0.9925, + "step": 14215 + }, + { + "epoch": 0.91, + "learning_rate": 1.6971381953240734e-05, + "loss": 1.0176, + "step": 14216 + }, + { + "epoch": 0.91, + "learning_rate": 1.6971167264217783e-05, + "loss": 1.0212, + "step": 14217 + }, + { + "epoch": 0.91, + "learning_rate": 1.6970952575194832e-05, + "loss": 1.0046, + "step": 14218 + }, + { + "epoch": 0.91, + "learning_rate": 1.697073788617188e-05, + "loss": 0.9903, + "step": 14219 + }, + { + "epoch": 0.91, + "learning_rate": 1.697052319714893e-05, + "loss": 1.0369, + "step": 14220 + }, + { + "epoch": 0.91, + "learning_rate": 1.6970308508125983e-05, + "loss": 0.9916, + "step": 14221 + }, + { + "epoch": 0.92, + "learning_rate": 1.697009381910303e-05, + "loss": 1.1395, + "step": 14222 + }, + { + "epoch": 0.92, + "learning_rate": 1.696987913008008e-05, + "loss": 1.1019, + "step": 14223 + }, + { + "epoch": 0.92, + "learning_rate": 1.696966444105713e-05, + "loss": 1.0244, + "step": 14224 + }, + { + "epoch": 0.92, + "learning_rate": 1.696944975203418e-05, + "loss": 1.1253, + "step": 14225 + }, + { + "epoch": 0.92, + "learning_rate": 1.6969235063011227e-05, + "loss": 1.0547, + "step": 14226 + }, + { + "epoch": 0.92, + "learning_rate": 1.696902037398828e-05, + "loss": 1.0811, + "step": 14227 + }, + { + "epoch": 0.92, + "learning_rate": 1.696880568496533e-05, + "loss": 1.0106, + "step": 14228 + }, + { + "epoch": 0.92, + "learning_rate": 1.6968590995942378e-05, + "loss": 1.0018, + "step": 14229 + }, + { + "epoch": 0.92, + "learning_rate": 1.696837630691943e-05, + "loss": 1.0036, + "step": 14230 + }, + { + "epoch": 0.92, + "learning_rate": 1.696816161789648e-05, + "loss": 1.0421, + "step": 14231 + }, + { + "epoch": 0.92, + "learning_rate": 1.6967946928873528e-05, + "loss": 0.9989, + "step": 14232 + }, + { + "epoch": 0.92, + "learning_rate": 1.6967732239850577e-05, + "loss": 1.0985, + "step": 14233 + }, + { + "epoch": 0.92, + "learning_rate": 1.696751755082763e-05, + "loss": 1.0096, + "step": 14234 + }, + { + "epoch": 0.92, + "learning_rate": 1.6967302861804678e-05, + "loss": 1.1005, + "step": 14235 + }, + { + "epoch": 0.92, + "learning_rate": 1.6967088172781727e-05, + "loss": 0.9699, + "step": 14236 + }, + { + "epoch": 0.92, + "learning_rate": 1.6966873483758776e-05, + "loss": 1.0757, + "step": 14237 + }, + { + "epoch": 0.92, + "learning_rate": 1.6966658794735825e-05, + "loss": 1.054, + "step": 14238 + }, + { + "epoch": 0.92, + "learning_rate": 1.6966444105712877e-05, + "loss": 1.0148, + "step": 14239 + }, + { + "epoch": 0.92, + "learning_rate": 1.6966229416689926e-05, + "loss": 1.1078, + "step": 14240 + }, + { + "epoch": 0.92, + "learning_rate": 1.6966014727666975e-05, + "loss": 1.0221, + "step": 14241 + }, + { + "epoch": 0.92, + "learning_rate": 1.6965800038644024e-05, + "loss": 1.0843, + "step": 14242 + }, + { + "epoch": 0.92, + "learning_rate": 1.6965585349621073e-05, + "loss": 0.9868, + "step": 14243 + }, + { + "epoch": 0.92, + "learning_rate": 1.6965370660598126e-05, + "loss": 0.9895, + "step": 14244 + }, + { + "epoch": 0.92, + "learning_rate": 1.6965155971575175e-05, + "loss": 1.0718, + "step": 14245 + }, + { + "epoch": 0.92, + "learning_rate": 1.6964941282552227e-05, + "loss": 1.0617, + "step": 14246 + }, + { + "epoch": 0.92, + "learning_rate": 1.6964726593529276e-05, + "loss": 1.1005, + "step": 14247 + }, + { + "epoch": 0.92, + "learning_rate": 1.6964511904506325e-05, + "loss": 1.0368, + "step": 14248 + }, + { + "epoch": 0.92, + "learning_rate": 1.6964297215483374e-05, + "loss": 1.0573, + "step": 14249 + }, + { + "epoch": 0.92, + "learning_rate": 1.6964082526460423e-05, + "loss": 1.0771, + "step": 14250 + }, + { + "epoch": 0.92, + "learning_rate": 1.6963867837437472e-05, + "loss": 0.9854, + "step": 14251 + }, + { + "epoch": 0.92, + "learning_rate": 1.6963653148414524e-05, + "loss": 1.0919, + "step": 14252 + }, + { + "epoch": 0.92, + "learning_rate": 1.6963438459391573e-05, + "loss": 1.0484, + "step": 14253 + }, + { + "epoch": 0.92, + "learning_rate": 1.6963223770368622e-05, + "loss": 1.041, + "step": 14254 + }, + { + "epoch": 0.92, + "learning_rate": 1.696300908134567e-05, + "loss": 1.0417, + "step": 14255 + }, + { + "epoch": 0.92, + "learning_rate": 1.696279439232272e-05, + "loss": 1.0485, + "step": 14256 + }, + { + "epoch": 0.92, + "learning_rate": 1.6962579703299772e-05, + "loss": 1.1078, + "step": 14257 + }, + { + "epoch": 0.92, + "learning_rate": 1.696236501427682e-05, + "loss": 1.1326, + "step": 14258 + }, + { + "epoch": 0.92, + "learning_rate": 1.6962150325253874e-05, + "loss": 1.0531, + "step": 14259 + }, + { + "epoch": 0.92, + "learning_rate": 1.6961935636230923e-05, + "loss": 1.0487, + "step": 14260 + }, + { + "epoch": 0.92, + "learning_rate": 1.696172094720797e-05, + "loss": 1.0646, + "step": 14261 + }, + { + "epoch": 0.92, + "learning_rate": 1.696150625818502e-05, + "loss": 1.0735, + "step": 14262 + }, + { + "epoch": 0.92, + "learning_rate": 1.696129156916207e-05, + "loss": 1.0325, + "step": 14263 + }, + { + "epoch": 0.92, + "learning_rate": 1.696107688013912e-05, + "loss": 1.1099, + "step": 14264 + }, + { + "epoch": 0.92, + "learning_rate": 1.696086219111617e-05, + "loss": 1.0521, + "step": 14265 + }, + { + "epoch": 0.92, + "learning_rate": 1.696064750209322e-05, + "loss": 1.0293, + "step": 14266 + }, + { + "epoch": 0.92, + "learning_rate": 1.696043281307027e-05, + "loss": 1.1315, + "step": 14267 + }, + { + "epoch": 0.92, + "learning_rate": 1.6960218124047318e-05, + "loss": 1.0305, + "step": 14268 + }, + { + "epoch": 0.92, + "learning_rate": 1.6960003435024367e-05, + "loss": 1.0701, + "step": 14269 + }, + { + "epoch": 0.92, + "learning_rate": 1.695978874600142e-05, + "loss": 1.0453, + "step": 14270 + }, + { + "epoch": 0.92, + "learning_rate": 1.6959574056978468e-05, + "loss": 1.0882, + "step": 14271 + }, + { + "epoch": 0.92, + "learning_rate": 1.6959359367955517e-05, + "loss": 1.0403, + "step": 14272 + }, + { + "epoch": 0.92, + "learning_rate": 1.695914467893257e-05, + "loss": 1.0526, + "step": 14273 + }, + { + "epoch": 0.92, + "learning_rate": 1.695892998990962e-05, + "loss": 1.1409, + "step": 14274 + }, + { + "epoch": 0.92, + "learning_rate": 1.6958715300886667e-05, + "loss": 1.0228, + "step": 14275 + }, + { + "epoch": 0.92, + "learning_rate": 1.6958500611863716e-05, + "loss": 1.0588, + "step": 14276 + }, + { + "epoch": 0.92, + "learning_rate": 1.695828592284077e-05, + "loss": 1.0997, + "step": 14277 + }, + { + "epoch": 0.92, + "learning_rate": 1.6958071233817818e-05, + "loss": 1.0156, + "step": 14278 + }, + { + "epoch": 0.92, + "learning_rate": 1.6957856544794867e-05, + "loss": 1.0102, + "step": 14279 + }, + { + "epoch": 0.92, + "learning_rate": 1.6957641855771915e-05, + "loss": 1.0946, + "step": 14280 + }, + { + "epoch": 0.92, + "learning_rate": 1.6957427166748964e-05, + "loss": 1.0095, + "step": 14281 + }, + { + "epoch": 0.92, + "learning_rate": 1.6957212477726013e-05, + "loss": 1.023, + "step": 14282 + }, + { + "epoch": 0.92, + "learning_rate": 1.6956997788703066e-05, + "loss": 1.0742, + "step": 14283 + }, + { + "epoch": 0.92, + "learning_rate": 1.6956783099680115e-05, + "loss": 1.0681, + "step": 14284 + }, + { + "epoch": 0.92, + "learning_rate": 1.6956568410657164e-05, + "loss": 1.094, + "step": 14285 + }, + { + "epoch": 0.92, + "learning_rate": 1.6956353721634213e-05, + "loss": 1.0373, + "step": 14286 + }, + { + "epoch": 0.92, + "learning_rate": 1.695613903261126e-05, + "loss": 1.0191, + "step": 14287 + }, + { + "epoch": 0.92, + "learning_rate": 1.6955924343588314e-05, + "loss": 1.0241, + "step": 14288 + }, + { + "epoch": 0.92, + "learning_rate": 1.6955709654565363e-05, + "loss": 1.0354, + "step": 14289 + }, + { + "epoch": 0.92, + "learning_rate": 1.6955494965542415e-05, + "loss": 1.0401, + "step": 14290 + }, + { + "epoch": 0.92, + "learning_rate": 1.6955280276519464e-05, + "loss": 0.9491, + "step": 14291 + }, + { + "epoch": 0.92, + "learning_rate": 1.6955065587496513e-05, + "loss": 1.0427, + "step": 14292 + }, + { + "epoch": 0.92, + "learning_rate": 1.6954850898473562e-05, + "loss": 1.0757, + "step": 14293 + }, + { + "epoch": 0.92, + "learning_rate": 1.695463620945061e-05, + "loss": 1.0934, + "step": 14294 + }, + { + "epoch": 0.92, + "learning_rate": 1.6954421520427664e-05, + "loss": 1.0384, + "step": 14295 + }, + { + "epoch": 0.92, + "learning_rate": 1.6954206831404712e-05, + "loss": 1.0957, + "step": 14296 + }, + { + "epoch": 0.92, + "learning_rate": 1.695399214238176e-05, + "loss": 1.0697, + "step": 14297 + }, + { + "epoch": 0.92, + "learning_rate": 1.695377745335881e-05, + "loss": 0.9811, + "step": 14298 + }, + { + "epoch": 0.92, + "learning_rate": 1.695356276433586e-05, + "loss": 1.0581, + "step": 14299 + }, + { + "epoch": 0.92, + "learning_rate": 1.695334807531291e-05, + "loss": 1.1551, + "step": 14300 + }, + { + "epoch": 0.92, + "learning_rate": 1.695313338628996e-05, + "loss": 0.9689, + "step": 14301 + }, + { + "epoch": 0.92, + "learning_rate": 1.695291869726701e-05, + "loss": 1.0611, + "step": 14302 + }, + { + "epoch": 0.92, + "learning_rate": 1.6952704008244062e-05, + "loss": 1.0544, + "step": 14303 + }, + { + "epoch": 0.92, + "learning_rate": 1.695248931922111e-05, + "loss": 1.0356, + "step": 14304 + }, + { + "epoch": 0.92, + "learning_rate": 1.695227463019816e-05, + "loss": 0.9726, + "step": 14305 + }, + { + "epoch": 0.92, + "learning_rate": 1.695205994117521e-05, + "loss": 1.082, + "step": 14306 + }, + { + "epoch": 0.92, + "learning_rate": 1.6951845252152258e-05, + "loss": 0.9585, + "step": 14307 + }, + { + "epoch": 0.92, + "learning_rate": 1.695163056312931e-05, + "loss": 1.1251, + "step": 14308 + }, + { + "epoch": 0.92, + "learning_rate": 1.695141587410636e-05, + "loss": 1.0522, + "step": 14309 + }, + { + "epoch": 0.92, + "learning_rate": 1.6951201185083408e-05, + "loss": 0.9583, + "step": 14310 + }, + { + "epoch": 0.92, + "learning_rate": 1.6950986496060457e-05, + "loss": 1.028, + "step": 14311 + }, + { + "epoch": 0.92, + "learning_rate": 1.6950771807037506e-05, + "loss": 1.0114, + "step": 14312 + }, + { + "epoch": 0.92, + "learning_rate": 1.6950557118014555e-05, + "loss": 1.1124, + "step": 14313 + }, + { + "epoch": 0.92, + "learning_rate": 1.6950342428991607e-05, + "loss": 1.0919, + "step": 14314 + }, + { + "epoch": 0.92, + "learning_rate": 1.6950127739968656e-05, + "loss": 1.0049, + "step": 14315 + }, + { + "epoch": 0.92, + "learning_rate": 1.6949913050945705e-05, + "loss": 1.0531, + "step": 14316 + }, + { + "epoch": 0.92, + "learning_rate": 1.6949698361922758e-05, + "loss": 1.061, + "step": 14317 + }, + { + "epoch": 0.92, + "learning_rate": 1.6949483672899807e-05, + "loss": 1.0716, + "step": 14318 + }, + { + "epoch": 0.92, + "learning_rate": 1.6949268983876856e-05, + "loss": 1.0307, + "step": 14319 + }, + { + "epoch": 0.92, + "learning_rate": 1.6949054294853905e-05, + "loss": 1.0152, + "step": 14320 + }, + { + "epoch": 0.92, + "learning_rate": 1.6948839605830957e-05, + "loss": 1.0447, + "step": 14321 + }, + { + "epoch": 0.92, + "learning_rate": 1.6948624916808006e-05, + "loss": 1.01, + "step": 14322 + }, + { + "epoch": 0.92, + "learning_rate": 1.6948410227785055e-05, + "loss": 1.186, + "step": 14323 + }, + { + "epoch": 0.92, + "learning_rate": 1.6948195538762104e-05, + "loss": 0.9917, + "step": 14324 + }, + { + "epoch": 0.92, + "learning_rate": 1.6947980849739153e-05, + "loss": 0.9921, + "step": 14325 + }, + { + "epoch": 0.92, + "learning_rate": 1.6947766160716205e-05, + "loss": 1.0414, + "step": 14326 + }, + { + "epoch": 0.92, + "learning_rate": 1.6947551471693254e-05, + "loss": 1.0921, + "step": 14327 + }, + { + "epoch": 0.92, + "learning_rate": 1.6947336782670303e-05, + "loss": 1.0169, + "step": 14328 + }, + { + "epoch": 0.92, + "learning_rate": 1.6947122093647352e-05, + "loss": 1.0361, + "step": 14329 + }, + { + "epoch": 0.92, + "learning_rate": 1.69469074046244e-05, + "loss": 1.0767, + "step": 14330 + }, + { + "epoch": 0.92, + "learning_rate": 1.6946692715601453e-05, + "loss": 1.1213, + "step": 14331 + }, + { + "epoch": 0.92, + "learning_rate": 1.6946478026578502e-05, + "loss": 1.0046, + "step": 14332 + }, + { + "epoch": 0.92, + "learning_rate": 1.694626333755555e-05, + "loss": 1.1163, + "step": 14333 + }, + { + "epoch": 0.92, + "learning_rate": 1.6946048648532604e-05, + "loss": 1.0347, + "step": 14334 + }, + { + "epoch": 0.92, + "learning_rate": 1.6945833959509653e-05, + "loss": 1.04, + "step": 14335 + }, + { + "epoch": 0.92, + "learning_rate": 1.69456192704867e-05, + "loss": 1.047, + "step": 14336 + }, + { + "epoch": 0.92, + "learning_rate": 1.694540458146375e-05, + "loss": 1.0755, + "step": 14337 + }, + { + "epoch": 0.92, + "learning_rate": 1.69451898924408e-05, + "loss": 1.0431, + "step": 14338 + }, + { + "epoch": 0.92, + "learning_rate": 1.6944975203417852e-05, + "loss": 1.1376, + "step": 14339 + }, + { + "epoch": 0.92, + "learning_rate": 1.69447605143949e-05, + "loss": 1.0556, + "step": 14340 + }, + { + "epoch": 0.92, + "learning_rate": 1.694454582537195e-05, + "loss": 1.1401, + "step": 14341 + }, + { + "epoch": 0.92, + "learning_rate": 1.6944331136349e-05, + "loss": 1.1146, + "step": 14342 + }, + { + "epoch": 0.92, + "learning_rate": 1.6944116447326048e-05, + "loss": 1.0616, + "step": 14343 + }, + { + "epoch": 0.92, + "learning_rate": 1.69439017583031e-05, + "loss": 1.0465, + "step": 14344 + }, + { + "epoch": 0.92, + "learning_rate": 1.694368706928015e-05, + "loss": 1.0217, + "step": 14345 + }, + { + "epoch": 0.92, + "learning_rate": 1.69434723802572e-05, + "loss": 1.093, + "step": 14346 + }, + { + "epoch": 0.92, + "learning_rate": 1.694325769123425e-05, + "loss": 1.1193, + "step": 14347 + }, + { + "epoch": 0.92, + "learning_rate": 1.69430430022113e-05, + "loss": 1.0565, + "step": 14348 + }, + { + "epoch": 0.92, + "learning_rate": 1.6942828313188348e-05, + "loss": 1.0603, + "step": 14349 + }, + { + "epoch": 0.92, + "learning_rate": 1.6942613624165397e-05, + "loss": 1.1238, + "step": 14350 + }, + { + "epoch": 0.92, + "learning_rate": 1.6942398935142446e-05, + "loss": 1.0032, + "step": 14351 + }, + { + "epoch": 0.92, + "learning_rate": 1.69421842461195e-05, + "loss": 1.0092, + "step": 14352 + }, + { + "epoch": 0.92, + "learning_rate": 1.6941969557096548e-05, + "loss": 1.0363, + "step": 14353 + }, + { + "epoch": 0.92, + "learning_rate": 1.6941754868073596e-05, + "loss": 0.9602, + "step": 14354 + }, + { + "epoch": 0.92, + "learning_rate": 1.6941540179050645e-05, + "loss": 1.0395, + "step": 14355 + }, + { + "epoch": 0.92, + "learning_rate": 1.6941325490027694e-05, + "loss": 1.1121, + "step": 14356 + }, + { + "epoch": 0.92, + "learning_rate": 1.6941110801004747e-05, + "loss": 1.0809, + "step": 14357 + }, + { + "epoch": 0.92, + "learning_rate": 1.6940896111981796e-05, + "loss": 1.0393, + "step": 14358 + }, + { + "epoch": 0.92, + "learning_rate": 1.6940681422958845e-05, + "loss": 1.0523, + "step": 14359 + }, + { + "epoch": 0.92, + "learning_rate": 1.6940466733935894e-05, + "loss": 1.0545, + "step": 14360 + }, + { + "epoch": 0.92, + "learning_rate": 1.6940252044912946e-05, + "loss": 1.0668, + "step": 14361 + }, + { + "epoch": 0.92, + "learning_rate": 1.6940037355889995e-05, + "loss": 1.0291, + "step": 14362 + }, + { + "epoch": 0.92, + "learning_rate": 1.6939822666867044e-05, + "loss": 1.0542, + "step": 14363 + }, + { + "epoch": 0.92, + "learning_rate": 1.6939607977844096e-05, + "loss": 1.0651, + "step": 14364 + }, + { + "epoch": 0.92, + "learning_rate": 1.6939393288821145e-05, + "loss": 1.0226, + "step": 14365 + }, + { + "epoch": 0.92, + "learning_rate": 1.6939178599798194e-05, + "loss": 1.1358, + "step": 14366 + }, + { + "epoch": 0.92, + "learning_rate": 1.6938963910775243e-05, + "loss": 1.024, + "step": 14367 + }, + { + "epoch": 0.92, + "learning_rate": 1.6938749221752292e-05, + "loss": 0.9985, + "step": 14368 + }, + { + "epoch": 0.92, + "learning_rate": 1.693853453272934e-05, + "loss": 0.975, + "step": 14369 + }, + { + "epoch": 0.92, + "learning_rate": 1.6938319843706393e-05, + "loss": 1.0519, + "step": 14370 + }, + { + "epoch": 0.92, + "learning_rate": 1.6938105154683442e-05, + "loss": 1.0267, + "step": 14371 + }, + { + "epoch": 0.92, + "learning_rate": 1.693789046566049e-05, + "loss": 0.9934, + "step": 14372 + }, + { + "epoch": 0.92, + "learning_rate": 1.693767577663754e-05, + "loss": 0.9418, + "step": 14373 + }, + { + "epoch": 0.92, + "learning_rate": 1.693746108761459e-05, + "loss": 1.1074, + "step": 14374 + }, + { + "epoch": 0.92, + "learning_rate": 1.693724639859164e-05, + "loss": 1.0303, + "step": 14375 + }, + { + "epoch": 0.92, + "learning_rate": 1.693703170956869e-05, + "loss": 1.0831, + "step": 14376 + }, + { + "epoch": 0.92, + "learning_rate": 1.6936817020545743e-05, + "loss": 1.0692, + "step": 14377 + }, + { + "epoch": 0.93, + "learning_rate": 1.6936602331522792e-05, + "loss": 1.0687, + "step": 14378 + }, + { + "epoch": 0.93, + "learning_rate": 1.693638764249984e-05, + "loss": 1.1117, + "step": 14379 + }, + { + "epoch": 0.93, + "learning_rate": 1.693617295347689e-05, + "loss": 0.9999, + "step": 14380 + }, + { + "epoch": 0.93, + "learning_rate": 1.693595826445394e-05, + "loss": 1.0335, + "step": 14381 + }, + { + "epoch": 0.93, + "learning_rate": 1.693574357543099e-05, + "loss": 1.0893, + "step": 14382 + }, + { + "epoch": 0.93, + "learning_rate": 1.693552888640804e-05, + "loss": 1.0622, + "step": 14383 + }, + { + "epoch": 0.93, + "learning_rate": 1.693531419738509e-05, + "loss": 1.0517, + "step": 14384 + }, + { + "epoch": 0.93, + "learning_rate": 1.6935099508362138e-05, + "loss": 1.0997, + "step": 14385 + }, + { + "epoch": 0.93, + "learning_rate": 1.6934884819339187e-05, + "loss": 1.0568, + "step": 14386 + }, + { + "epoch": 0.93, + "learning_rate": 1.6934670130316236e-05, + "loss": 1.1006, + "step": 14387 + }, + { + "epoch": 0.93, + "learning_rate": 1.693445544129329e-05, + "loss": 1.088, + "step": 14388 + }, + { + "epoch": 0.93, + "learning_rate": 1.6934240752270337e-05, + "loss": 1.0264, + "step": 14389 + }, + { + "epoch": 0.93, + "learning_rate": 1.693402606324739e-05, + "loss": 1.1327, + "step": 14390 + }, + { + "epoch": 0.93, + "learning_rate": 1.693381137422444e-05, + "loss": 1.002, + "step": 14391 + }, + { + "epoch": 0.93, + "learning_rate": 1.6933596685201488e-05, + "loss": 1.0862, + "step": 14392 + }, + { + "epoch": 0.93, + "learning_rate": 1.6933381996178537e-05, + "loss": 1.067, + "step": 14393 + }, + { + "epoch": 0.93, + "learning_rate": 1.6933167307155586e-05, + "loss": 1.0455, + "step": 14394 + }, + { + "epoch": 0.93, + "learning_rate": 1.6932952618132638e-05, + "loss": 1.05, + "step": 14395 + }, + { + "epoch": 0.93, + "learning_rate": 1.6932737929109687e-05, + "loss": 1.137, + "step": 14396 + }, + { + "epoch": 0.93, + "learning_rate": 1.6932523240086736e-05, + "loss": 1.1151, + "step": 14397 + }, + { + "epoch": 0.93, + "learning_rate": 1.6932308551063785e-05, + "loss": 1.0337, + "step": 14398 + }, + { + "epoch": 0.93, + "learning_rate": 1.6932093862040834e-05, + "loss": 0.9958, + "step": 14399 + }, + { + "epoch": 0.93, + "learning_rate": 1.6931879173017883e-05, + "loss": 1.1157, + "step": 14400 + }, + { + "epoch": 0.93, + "learning_rate": 1.6931664483994935e-05, + "loss": 1.0745, + "step": 14401 + }, + { + "epoch": 0.93, + "learning_rate": 1.6931449794971984e-05, + "loss": 1.0869, + "step": 14402 + }, + { + "epoch": 0.93, + "learning_rate": 1.6931235105949033e-05, + "loss": 1.0552, + "step": 14403 + }, + { + "epoch": 0.93, + "learning_rate": 1.6931020416926085e-05, + "loss": 1.101, + "step": 14404 + }, + { + "epoch": 0.93, + "learning_rate": 1.6930805727903134e-05, + "loss": 1.0887, + "step": 14405 + }, + { + "epoch": 0.93, + "learning_rate": 1.6930591038880183e-05, + "loss": 1.0219, + "step": 14406 + }, + { + "epoch": 0.93, + "learning_rate": 1.6930376349857232e-05, + "loss": 1.0158, + "step": 14407 + }, + { + "epoch": 0.93, + "learning_rate": 1.6930161660834285e-05, + "loss": 1.0554, + "step": 14408 + }, + { + "epoch": 0.93, + "learning_rate": 1.6929946971811334e-05, + "loss": 1.0587, + "step": 14409 + }, + { + "epoch": 0.93, + "learning_rate": 1.6929732282788383e-05, + "loss": 0.9995, + "step": 14410 + }, + { + "epoch": 0.93, + "learning_rate": 1.692951759376543e-05, + "loss": 1.0779, + "step": 14411 + }, + { + "epoch": 0.93, + "learning_rate": 1.692930290474248e-05, + "loss": 1.0806, + "step": 14412 + }, + { + "epoch": 0.93, + "learning_rate": 1.6929088215719533e-05, + "loss": 0.9703, + "step": 14413 + }, + { + "epoch": 0.93, + "learning_rate": 1.6928873526696582e-05, + "loss": 1.1541, + "step": 14414 + }, + { + "epoch": 0.93, + "learning_rate": 1.692865883767363e-05, + "loss": 1.076, + "step": 14415 + }, + { + "epoch": 0.93, + "learning_rate": 1.692844414865068e-05, + "loss": 1.0775, + "step": 14416 + }, + { + "epoch": 0.93, + "learning_rate": 1.692822945962773e-05, + "loss": 1.1058, + "step": 14417 + }, + { + "epoch": 0.93, + "learning_rate": 1.6928014770604778e-05, + "loss": 1.0951, + "step": 14418 + }, + { + "epoch": 0.93, + "learning_rate": 1.692780008158183e-05, + "loss": 1.1928, + "step": 14419 + }, + { + "epoch": 0.93, + "learning_rate": 1.692758539255888e-05, + "loss": 1.1312, + "step": 14420 + }, + { + "epoch": 0.93, + "learning_rate": 1.692737070353593e-05, + "loss": 1.0008, + "step": 14421 + }, + { + "epoch": 0.93, + "learning_rate": 1.692715601451298e-05, + "loss": 0.9762, + "step": 14422 + }, + { + "epoch": 0.93, + "learning_rate": 1.692694132549003e-05, + "loss": 1.1825, + "step": 14423 + }, + { + "epoch": 0.93, + "learning_rate": 1.6926726636467078e-05, + "loss": 0.9912, + "step": 14424 + }, + { + "epoch": 0.93, + "learning_rate": 1.6926511947444127e-05, + "loss": 1.1357, + "step": 14425 + }, + { + "epoch": 0.93, + "learning_rate": 1.692629725842118e-05, + "loss": 1.0792, + "step": 14426 + }, + { + "epoch": 0.93, + "learning_rate": 1.692608256939823e-05, + "loss": 1.1004, + "step": 14427 + }, + { + "epoch": 0.93, + "learning_rate": 1.6925867880375277e-05, + "loss": 1.0446, + "step": 14428 + }, + { + "epoch": 0.93, + "learning_rate": 1.6925653191352326e-05, + "loss": 1.1266, + "step": 14429 + }, + { + "epoch": 0.93, + "learning_rate": 1.6925438502329375e-05, + "loss": 1.0426, + "step": 14430 + }, + { + "epoch": 0.93, + "learning_rate": 1.6925223813306428e-05, + "loss": 1.1757, + "step": 14431 + }, + { + "epoch": 0.93, + "learning_rate": 1.6925009124283477e-05, + "loss": 1.0355, + "step": 14432 + }, + { + "epoch": 0.93, + "learning_rate": 1.692479443526053e-05, + "loss": 1.0647, + "step": 14433 + }, + { + "epoch": 0.93, + "learning_rate": 1.6924579746237578e-05, + "loss": 1.1237, + "step": 14434 + }, + { + "epoch": 0.93, + "learning_rate": 1.6924365057214627e-05, + "loss": 1.0988, + "step": 14435 + }, + { + "epoch": 0.93, + "learning_rate": 1.6924150368191676e-05, + "loss": 1.0221, + "step": 14436 + }, + { + "epoch": 0.93, + "learning_rate": 1.6923935679168725e-05, + "loss": 1.0526, + "step": 14437 + }, + { + "epoch": 0.93, + "learning_rate": 1.6923720990145774e-05, + "loss": 1.1018, + "step": 14438 + }, + { + "epoch": 0.93, + "learning_rate": 1.6923506301122826e-05, + "loss": 1.0555, + "step": 14439 + }, + { + "epoch": 0.93, + "learning_rate": 1.6923291612099875e-05, + "loss": 1.0255, + "step": 14440 + }, + { + "epoch": 0.93, + "learning_rate": 1.6923076923076924e-05, + "loss": 1.0854, + "step": 14441 + }, + { + "epoch": 0.93, + "learning_rate": 1.6922862234053973e-05, + "loss": 0.9992, + "step": 14442 + }, + { + "epoch": 0.93, + "learning_rate": 1.6922647545031022e-05, + "loss": 0.9958, + "step": 14443 + }, + { + "epoch": 0.93, + "learning_rate": 1.6922432856008074e-05, + "loss": 1.055, + "step": 14444 + }, + { + "epoch": 0.93, + "learning_rate": 1.6922218166985123e-05, + "loss": 1.0136, + "step": 14445 + }, + { + "epoch": 0.93, + "learning_rate": 1.6922003477962172e-05, + "loss": 1.2379, + "step": 14446 + }, + { + "epoch": 0.93, + "learning_rate": 1.692178878893922e-05, + "loss": 1.078, + "step": 14447 + }, + { + "epoch": 0.93, + "learning_rate": 1.6921574099916274e-05, + "loss": 1.0431, + "step": 14448 + }, + { + "epoch": 0.93, + "learning_rate": 1.6921359410893323e-05, + "loss": 1.1196, + "step": 14449 + }, + { + "epoch": 0.93, + "learning_rate": 1.692114472187037e-05, + "loss": 1.1019, + "step": 14450 + }, + { + "epoch": 0.93, + "learning_rate": 1.6920930032847424e-05, + "loss": 1.0283, + "step": 14451 + }, + { + "epoch": 0.93, + "learning_rate": 1.6920715343824473e-05, + "loss": 1.0551, + "step": 14452 + }, + { + "epoch": 0.93, + "learning_rate": 1.6920500654801522e-05, + "loss": 1.0575, + "step": 14453 + }, + { + "epoch": 0.93, + "learning_rate": 1.692028596577857e-05, + "loss": 0.9616, + "step": 14454 + }, + { + "epoch": 0.93, + "learning_rate": 1.692007127675562e-05, + "loss": 1.0158, + "step": 14455 + }, + { + "epoch": 0.93, + "learning_rate": 1.691985658773267e-05, + "loss": 1.0852, + "step": 14456 + }, + { + "epoch": 0.93, + "learning_rate": 1.691964189870972e-05, + "loss": 1.069, + "step": 14457 + }, + { + "epoch": 0.93, + "learning_rate": 1.691942720968677e-05, + "loss": 1.091, + "step": 14458 + }, + { + "epoch": 0.93, + "learning_rate": 1.691921252066382e-05, + "loss": 1.0717, + "step": 14459 + }, + { + "epoch": 0.93, + "learning_rate": 1.6918997831640868e-05, + "loss": 1.0589, + "step": 14460 + }, + { + "epoch": 0.93, + "learning_rate": 1.6918783142617917e-05, + "loss": 0.9846, + "step": 14461 + }, + { + "epoch": 0.93, + "learning_rate": 1.691856845359497e-05, + "loss": 1.0928, + "step": 14462 + }, + { + "epoch": 0.93, + "learning_rate": 1.691835376457202e-05, + "loss": 1.0878, + "step": 14463 + }, + { + "epoch": 0.93, + "learning_rate": 1.691813907554907e-05, + "loss": 1.035, + "step": 14464 + }, + { + "epoch": 0.93, + "learning_rate": 1.691792438652612e-05, + "loss": 1.0907, + "step": 14465 + }, + { + "epoch": 0.93, + "learning_rate": 1.691770969750317e-05, + "loss": 1.0387, + "step": 14466 + }, + { + "epoch": 0.93, + "learning_rate": 1.6917495008480218e-05, + "loss": 1.0104, + "step": 14467 + }, + { + "epoch": 0.93, + "learning_rate": 1.6917280319457267e-05, + "loss": 1.085, + "step": 14468 + }, + { + "epoch": 0.93, + "learning_rate": 1.691706563043432e-05, + "loss": 0.9835, + "step": 14469 + }, + { + "epoch": 0.93, + "learning_rate": 1.6916850941411368e-05, + "loss": 1.019, + "step": 14470 + }, + { + "epoch": 0.93, + "learning_rate": 1.6916636252388417e-05, + "loss": 1.0763, + "step": 14471 + }, + { + "epoch": 0.93, + "learning_rate": 1.6916421563365466e-05, + "loss": 1.0478, + "step": 14472 + }, + { + "epoch": 0.93, + "learning_rate": 1.6916206874342515e-05, + "loss": 1.0311, + "step": 14473 + }, + { + "epoch": 0.93, + "learning_rate": 1.6915992185319564e-05, + "loss": 1.0859, + "step": 14474 + }, + { + "epoch": 0.93, + "learning_rate": 1.6915777496296616e-05, + "loss": 0.989, + "step": 14475 + }, + { + "epoch": 0.93, + "learning_rate": 1.6915562807273665e-05, + "loss": 1.0465, + "step": 14476 + }, + { + "epoch": 0.93, + "learning_rate": 1.6915348118250717e-05, + "loss": 1.0583, + "step": 14477 + }, + { + "epoch": 0.93, + "learning_rate": 1.6915133429227766e-05, + "loss": 1.0661, + "step": 14478 + }, + { + "epoch": 0.93, + "learning_rate": 1.6914918740204815e-05, + "loss": 1.0376, + "step": 14479 + }, + { + "epoch": 0.93, + "learning_rate": 1.6914704051181864e-05, + "loss": 1.1311, + "step": 14480 + }, + { + "epoch": 0.93, + "learning_rate": 1.6914489362158913e-05, + "loss": 0.9681, + "step": 14481 + }, + { + "epoch": 0.93, + "learning_rate": 1.6914274673135966e-05, + "loss": 1.0264, + "step": 14482 + }, + { + "epoch": 0.93, + "learning_rate": 1.6914059984113015e-05, + "loss": 1.0454, + "step": 14483 + }, + { + "epoch": 0.93, + "learning_rate": 1.6913845295090064e-05, + "loss": 1.0219, + "step": 14484 + }, + { + "epoch": 0.93, + "learning_rate": 1.6913630606067113e-05, + "loss": 1.0675, + "step": 14485 + }, + { + "epoch": 0.93, + "learning_rate": 1.691341591704416e-05, + "loss": 0.9274, + "step": 14486 + }, + { + "epoch": 0.93, + "learning_rate": 1.691320122802121e-05, + "loss": 1.1956, + "step": 14487 + }, + { + "epoch": 0.93, + "learning_rate": 1.6912986538998263e-05, + "loss": 1.0718, + "step": 14488 + }, + { + "epoch": 0.93, + "learning_rate": 1.6912771849975312e-05, + "loss": 1.0674, + "step": 14489 + }, + { + "epoch": 0.93, + "learning_rate": 1.691255716095236e-05, + "loss": 1.036, + "step": 14490 + }, + { + "epoch": 0.93, + "learning_rate": 1.6912342471929413e-05, + "loss": 1.0285, + "step": 14491 + }, + { + "epoch": 0.93, + "learning_rate": 1.6912127782906462e-05, + "loss": 1.0614, + "step": 14492 + }, + { + "epoch": 0.93, + "learning_rate": 1.691191309388351e-05, + "loss": 1.021, + "step": 14493 + }, + { + "epoch": 0.93, + "learning_rate": 1.691169840486056e-05, + "loss": 1.0305, + "step": 14494 + }, + { + "epoch": 0.93, + "learning_rate": 1.6911483715837612e-05, + "loss": 1.0644, + "step": 14495 + }, + { + "epoch": 0.93, + "learning_rate": 1.691126902681466e-05, + "loss": 1.0528, + "step": 14496 + }, + { + "epoch": 0.93, + "learning_rate": 1.691105433779171e-05, + "loss": 1.0805, + "step": 14497 + }, + { + "epoch": 0.93, + "learning_rate": 1.691083964876876e-05, + "loss": 1.0693, + "step": 14498 + }, + { + "epoch": 0.93, + "learning_rate": 1.6910624959745808e-05, + "loss": 1.083, + "step": 14499 + }, + { + "epoch": 0.93, + "learning_rate": 1.691041027072286e-05, + "loss": 1.1729, + "step": 14500 + }, + { + "epoch": 0.93, + "learning_rate": 1.691019558169991e-05, + "loss": 0.9648, + "step": 14501 + }, + { + "epoch": 0.93, + "learning_rate": 1.690998089267696e-05, + "loss": 1.0512, + "step": 14502 + }, + { + "epoch": 0.93, + "learning_rate": 1.6909766203654007e-05, + "loss": 1.1035, + "step": 14503 + }, + { + "epoch": 0.93, + "learning_rate": 1.6909551514631056e-05, + "loss": 0.999, + "step": 14504 + }, + { + "epoch": 0.93, + "learning_rate": 1.6909336825608105e-05, + "loss": 1.0763, + "step": 14505 + }, + { + "epoch": 0.93, + "learning_rate": 1.6909122136585158e-05, + "loss": 1.1076, + "step": 14506 + }, + { + "epoch": 0.93, + "learning_rate": 1.6908907447562207e-05, + "loss": 1.0079, + "step": 14507 + }, + { + "epoch": 0.93, + "learning_rate": 1.690869275853926e-05, + "loss": 1.1233, + "step": 14508 + }, + { + "epoch": 0.93, + "learning_rate": 1.6908478069516308e-05, + "loss": 0.9684, + "step": 14509 + }, + { + "epoch": 0.93, + "learning_rate": 1.6908263380493357e-05, + "loss": 1.1038, + "step": 14510 + }, + { + "epoch": 0.93, + "learning_rate": 1.6908048691470406e-05, + "loss": 1.0994, + "step": 14511 + }, + { + "epoch": 0.93, + "learning_rate": 1.6907834002447455e-05, + "loss": 1.0471, + "step": 14512 + }, + { + "epoch": 0.93, + "learning_rate": 1.6907619313424507e-05, + "loss": 1.0895, + "step": 14513 + }, + { + "epoch": 0.93, + "learning_rate": 1.6907404624401556e-05, + "loss": 1.0614, + "step": 14514 + }, + { + "epoch": 0.93, + "learning_rate": 1.6907189935378605e-05, + "loss": 0.9956, + "step": 14515 + }, + { + "epoch": 0.93, + "learning_rate": 1.6906975246355654e-05, + "loss": 1.013, + "step": 14516 + }, + { + "epoch": 0.93, + "learning_rate": 1.6906760557332703e-05, + "loss": 1.0196, + "step": 14517 + }, + { + "epoch": 0.93, + "learning_rate": 1.6906545868309755e-05, + "loss": 1.0833, + "step": 14518 + }, + { + "epoch": 0.93, + "learning_rate": 1.6906331179286804e-05, + "loss": 1.0309, + "step": 14519 + }, + { + "epoch": 0.93, + "learning_rate": 1.6906116490263857e-05, + "loss": 1.1286, + "step": 14520 + }, + { + "epoch": 0.93, + "learning_rate": 1.6905901801240906e-05, + "loss": 1.0846, + "step": 14521 + }, + { + "epoch": 0.93, + "learning_rate": 1.6905687112217955e-05, + "loss": 1.0536, + "step": 14522 + }, + { + "epoch": 0.93, + "learning_rate": 1.6905472423195004e-05, + "loss": 1.0894, + "step": 14523 + }, + { + "epoch": 0.93, + "learning_rate": 1.6905257734172053e-05, + "loss": 1.072, + "step": 14524 + }, + { + "epoch": 0.93, + "learning_rate": 1.69050430451491e-05, + "loss": 0.981, + "step": 14525 + }, + { + "epoch": 0.93, + "learning_rate": 1.6904828356126154e-05, + "loss": 0.9832, + "step": 14526 + }, + { + "epoch": 0.93, + "learning_rate": 1.6904613667103203e-05, + "loss": 1.0748, + "step": 14527 + }, + { + "epoch": 0.93, + "learning_rate": 1.6904398978080252e-05, + "loss": 1.1106, + "step": 14528 + }, + { + "epoch": 0.93, + "learning_rate": 1.69041842890573e-05, + "loss": 1.0746, + "step": 14529 + }, + { + "epoch": 0.93, + "learning_rate": 1.690396960003435e-05, + "loss": 1.0448, + "step": 14530 + }, + { + "epoch": 0.93, + "learning_rate": 1.6903754911011402e-05, + "loss": 1.0097, + "step": 14531 + }, + { + "epoch": 0.93, + "learning_rate": 1.690354022198845e-05, + "loss": 1.0411, + "step": 14532 + }, + { + "epoch": 0.94, + "learning_rate": 1.69033255329655e-05, + "loss": 1.0608, + "step": 14533 + }, + { + "epoch": 0.94, + "learning_rate": 1.690311084394255e-05, + "loss": 1.0295, + "step": 14534 + }, + { + "epoch": 0.94, + "learning_rate": 1.69028961549196e-05, + "loss": 1.0442, + "step": 14535 + }, + { + "epoch": 0.94, + "learning_rate": 1.690268146589665e-05, + "loss": 1.0024, + "step": 14536 + }, + { + "epoch": 0.94, + "learning_rate": 1.69024667768737e-05, + "loss": 1.0338, + "step": 14537 + }, + { + "epoch": 0.94, + "learning_rate": 1.6902252087850752e-05, + "loss": 1.0241, + "step": 14538 + }, + { + "epoch": 0.94, + "learning_rate": 1.69020373988278e-05, + "loss": 1.0337, + "step": 14539 + }, + { + "epoch": 0.94, + "learning_rate": 1.690182270980485e-05, + "loss": 1.0548, + "step": 14540 + }, + { + "epoch": 0.94, + "learning_rate": 1.69016080207819e-05, + "loss": 1.0524, + "step": 14541 + }, + { + "epoch": 0.94, + "learning_rate": 1.6901393331758948e-05, + "loss": 1.1084, + "step": 14542 + }, + { + "epoch": 0.94, + "learning_rate": 1.6901178642735997e-05, + "loss": 1.0615, + "step": 14543 + }, + { + "epoch": 0.94, + "learning_rate": 1.690096395371305e-05, + "loss": 1.0079, + "step": 14544 + }, + { + "epoch": 0.94, + "learning_rate": 1.6900749264690098e-05, + "loss": 0.9383, + "step": 14545 + }, + { + "epoch": 0.94, + "learning_rate": 1.6900534575667147e-05, + "loss": 1.0204, + "step": 14546 + }, + { + "epoch": 0.94, + "learning_rate": 1.6900319886644196e-05, + "loss": 1.0589, + "step": 14547 + }, + { + "epoch": 0.94, + "learning_rate": 1.6900105197621245e-05, + "loss": 1.0573, + "step": 14548 + }, + { + "epoch": 0.94, + "learning_rate": 1.6899890508598297e-05, + "loss": 1.0656, + "step": 14549 + }, + { + "epoch": 0.94, + "learning_rate": 1.6899675819575346e-05, + "loss": 1.1194, + "step": 14550 + }, + { + "epoch": 0.94, + "learning_rate": 1.68994611305524e-05, + "loss": 1.0355, + "step": 14551 + }, + { + "epoch": 0.94, + "learning_rate": 1.6899246441529447e-05, + "loss": 1.0433, + "step": 14552 + }, + { + "epoch": 0.94, + "learning_rate": 1.6899031752506496e-05, + "loss": 1.028, + "step": 14553 + }, + { + "epoch": 0.94, + "learning_rate": 1.6898817063483545e-05, + "loss": 1.1114, + "step": 14554 + }, + { + "epoch": 0.94, + "learning_rate": 1.6898602374460594e-05, + "loss": 1.0346, + "step": 14555 + }, + { + "epoch": 0.94, + "learning_rate": 1.6898387685437647e-05, + "loss": 1.085, + "step": 14556 + }, + { + "epoch": 0.94, + "learning_rate": 1.6898172996414696e-05, + "loss": 1.0022, + "step": 14557 + }, + { + "epoch": 0.94, + "learning_rate": 1.6897958307391745e-05, + "loss": 1.1132, + "step": 14558 + }, + { + "epoch": 0.94, + "learning_rate": 1.6897743618368794e-05, + "loss": 1.0141, + "step": 14559 + }, + { + "epoch": 0.94, + "learning_rate": 1.6897528929345843e-05, + "loss": 1.0407, + "step": 14560 + }, + { + "epoch": 0.94, + "learning_rate": 1.689731424032289e-05, + "loss": 1.0264, + "step": 14561 + }, + { + "epoch": 0.94, + "learning_rate": 1.6897099551299944e-05, + "loss": 1.0382, + "step": 14562 + }, + { + "epoch": 0.94, + "learning_rate": 1.6896884862276993e-05, + "loss": 1.1047, + "step": 14563 + }, + { + "epoch": 0.94, + "learning_rate": 1.6896670173254045e-05, + "loss": 1.0841, + "step": 14564 + }, + { + "epoch": 0.94, + "learning_rate": 1.6896455484231094e-05, + "loss": 1.0638, + "step": 14565 + }, + { + "epoch": 0.94, + "learning_rate": 1.6896240795208143e-05, + "loss": 1.0355, + "step": 14566 + }, + { + "epoch": 0.94, + "learning_rate": 1.6896026106185192e-05, + "loss": 0.9976, + "step": 14567 + }, + { + "epoch": 0.94, + "learning_rate": 1.689581141716224e-05, + "loss": 1.0163, + "step": 14568 + }, + { + "epoch": 0.94, + "learning_rate": 1.6895596728139293e-05, + "loss": 1.0639, + "step": 14569 + }, + { + "epoch": 0.94, + "learning_rate": 1.6895382039116342e-05, + "loss": 1.0173, + "step": 14570 + }, + { + "epoch": 0.94, + "learning_rate": 1.689516735009339e-05, + "loss": 1.0556, + "step": 14571 + }, + { + "epoch": 0.94, + "learning_rate": 1.689495266107044e-05, + "loss": 1.1199, + "step": 14572 + }, + { + "epoch": 0.94, + "learning_rate": 1.689473797204749e-05, + "loss": 0.9807, + "step": 14573 + }, + { + "epoch": 0.94, + "learning_rate": 1.6894523283024538e-05, + "loss": 1.1018, + "step": 14574 + }, + { + "epoch": 0.94, + "learning_rate": 1.689430859400159e-05, + "loss": 1.0451, + "step": 14575 + }, + { + "epoch": 0.94, + "learning_rate": 1.689409390497864e-05, + "loss": 1.0362, + "step": 14576 + }, + { + "epoch": 0.94, + "learning_rate": 1.689387921595569e-05, + "loss": 0.9984, + "step": 14577 + }, + { + "epoch": 0.94, + "learning_rate": 1.689366452693274e-05, + "loss": 1.0375, + "step": 14578 + }, + { + "epoch": 0.94, + "learning_rate": 1.689344983790979e-05, + "loss": 1.0449, + "step": 14579 + }, + { + "epoch": 0.94, + "learning_rate": 1.689323514888684e-05, + "loss": 0.9609, + "step": 14580 + }, + { + "epoch": 0.94, + "learning_rate": 1.6893020459863888e-05, + "loss": 1.0173, + "step": 14581 + }, + { + "epoch": 0.94, + "learning_rate": 1.689280577084094e-05, + "loss": 1.0574, + "step": 14582 + }, + { + "epoch": 0.94, + "learning_rate": 1.689259108181799e-05, + "loss": 1.047, + "step": 14583 + }, + { + "epoch": 0.94, + "learning_rate": 1.6892376392795038e-05, + "loss": 1.0634, + "step": 14584 + }, + { + "epoch": 0.94, + "learning_rate": 1.6892161703772087e-05, + "loss": 1.0165, + "step": 14585 + }, + { + "epoch": 0.94, + "learning_rate": 1.6891947014749136e-05, + "loss": 1.0463, + "step": 14586 + }, + { + "epoch": 0.94, + "learning_rate": 1.6891732325726188e-05, + "loss": 1.0709, + "step": 14587 + }, + { + "epoch": 0.94, + "learning_rate": 1.6891517636703237e-05, + "loss": 1.0027, + "step": 14588 + }, + { + "epoch": 0.94, + "learning_rate": 1.6891302947680286e-05, + "loss": 1.1096, + "step": 14589 + }, + { + "epoch": 0.94, + "learning_rate": 1.6891088258657335e-05, + "loss": 1.1243, + "step": 14590 + }, + { + "epoch": 0.94, + "learning_rate": 1.6890873569634384e-05, + "loss": 1.0875, + "step": 14591 + }, + { + "epoch": 0.94, + "learning_rate": 1.6890658880611433e-05, + "loss": 1.1025, + "step": 14592 + }, + { + "epoch": 0.94, + "learning_rate": 1.6890444191588485e-05, + "loss": 1.1296, + "step": 14593 + }, + { + "epoch": 0.94, + "learning_rate": 1.6890229502565534e-05, + "loss": 1.1, + "step": 14594 + }, + { + "epoch": 0.94, + "learning_rate": 1.6890014813542587e-05, + "loss": 1.0844, + "step": 14595 + }, + { + "epoch": 0.94, + "learning_rate": 1.6889800124519636e-05, + "loss": 1.0392, + "step": 14596 + }, + { + "epoch": 0.94, + "learning_rate": 1.6889585435496685e-05, + "loss": 1.0973, + "step": 14597 + }, + { + "epoch": 0.94, + "learning_rate": 1.6889370746473734e-05, + "loss": 1.0971, + "step": 14598 + }, + { + "epoch": 0.94, + "learning_rate": 1.6889156057450783e-05, + "loss": 0.978, + "step": 14599 + }, + { + "epoch": 0.94, + "learning_rate": 1.6888941368427835e-05, + "loss": 0.9767, + "step": 14600 + }, + { + "epoch": 0.94, + "learning_rate": 1.6888726679404884e-05, + "loss": 1.0092, + "step": 14601 + }, + { + "epoch": 0.94, + "learning_rate": 1.6888511990381933e-05, + "loss": 1.0032, + "step": 14602 + }, + { + "epoch": 0.94, + "learning_rate": 1.6888297301358982e-05, + "loss": 1.0409, + "step": 14603 + }, + { + "epoch": 0.94, + "learning_rate": 1.688808261233603e-05, + "loss": 0.9802, + "step": 14604 + }, + { + "epoch": 0.94, + "learning_rate": 1.6887867923313083e-05, + "loss": 1.0489, + "step": 14605 + }, + { + "epoch": 0.94, + "learning_rate": 1.6887653234290132e-05, + "loss": 1.1045, + "step": 14606 + }, + { + "epoch": 0.94, + "learning_rate": 1.6887438545267185e-05, + "loss": 1.0799, + "step": 14607 + }, + { + "epoch": 0.94, + "learning_rate": 1.6887223856244233e-05, + "loss": 1.0514, + "step": 14608 + }, + { + "epoch": 0.94, + "learning_rate": 1.6887009167221282e-05, + "loss": 1.0404, + "step": 14609 + }, + { + "epoch": 0.94, + "learning_rate": 1.688679447819833e-05, + "loss": 1.08, + "step": 14610 + }, + { + "epoch": 0.94, + "learning_rate": 1.688657978917538e-05, + "loss": 1.0245, + "step": 14611 + }, + { + "epoch": 0.94, + "learning_rate": 1.688636510015243e-05, + "loss": 1.1166, + "step": 14612 + }, + { + "epoch": 0.94, + "learning_rate": 1.6886150411129482e-05, + "loss": 1.0165, + "step": 14613 + }, + { + "epoch": 0.94, + "learning_rate": 1.688593572210653e-05, + "loss": 1.0157, + "step": 14614 + }, + { + "epoch": 0.94, + "learning_rate": 1.688572103308358e-05, + "loss": 1.0058, + "step": 14615 + }, + { + "epoch": 0.94, + "learning_rate": 1.688550634406063e-05, + "loss": 1.1745, + "step": 14616 + }, + { + "epoch": 0.94, + "learning_rate": 1.6885291655037678e-05, + "loss": 1.1354, + "step": 14617 + }, + { + "epoch": 0.94, + "learning_rate": 1.688507696601473e-05, + "loss": 1.0396, + "step": 14618 + }, + { + "epoch": 0.94, + "learning_rate": 1.688486227699178e-05, + "loss": 1.0499, + "step": 14619 + }, + { + "epoch": 0.94, + "learning_rate": 1.6884647587968828e-05, + "loss": 1.1371, + "step": 14620 + }, + { + "epoch": 0.94, + "learning_rate": 1.6884432898945877e-05, + "loss": 1.093, + "step": 14621 + }, + { + "epoch": 0.94, + "learning_rate": 1.688421820992293e-05, + "loss": 1.0288, + "step": 14622 + }, + { + "epoch": 0.94, + "learning_rate": 1.6884003520899978e-05, + "loss": 1.09, + "step": 14623 + }, + { + "epoch": 0.94, + "learning_rate": 1.6883788831877027e-05, + "loss": 1.0743, + "step": 14624 + }, + { + "epoch": 0.94, + "learning_rate": 1.688357414285408e-05, + "loss": 0.9878, + "step": 14625 + }, + { + "epoch": 0.94, + "learning_rate": 1.688335945383113e-05, + "loss": 1.0011, + "step": 14626 + }, + { + "epoch": 0.94, + "learning_rate": 1.6883144764808177e-05, + "loss": 1.0216, + "step": 14627 + }, + { + "epoch": 0.94, + "learning_rate": 1.6882930075785226e-05, + "loss": 0.9984, + "step": 14628 + }, + { + "epoch": 0.94, + "learning_rate": 1.6882715386762275e-05, + "loss": 0.9861, + "step": 14629 + }, + { + "epoch": 0.94, + "learning_rate": 1.6882500697739324e-05, + "loss": 1.1305, + "step": 14630 + }, + { + "epoch": 0.94, + "learning_rate": 1.6882286008716377e-05, + "loss": 0.9738, + "step": 14631 + }, + { + "epoch": 0.94, + "learning_rate": 1.6882071319693426e-05, + "loss": 1.0905, + "step": 14632 + }, + { + "epoch": 0.94, + "learning_rate": 1.6881856630670475e-05, + "loss": 1.0112, + "step": 14633 + }, + { + "epoch": 0.94, + "learning_rate": 1.6881641941647524e-05, + "loss": 1.0475, + "step": 14634 + }, + { + "epoch": 0.94, + "learning_rate": 1.6881427252624572e-05, + "loss": 0.9994, + "step": 14635 + }, + { + "epoch": 0.94, + "learning_rate": 1.6881212563601625e-05, + "loss": 1.0145, + "step": 14636 + }, + { + "epoch": 0.94, + "learning_rate": 1.6880997874578674e-05, + "loss": 1.0718, + "step": 14637 + }, + { + "epoch": 0.94, + "learning_rate": 1.6880783185555726e-05, + "loss": 1.0964, + "step": 14638 + }, + { + "epoch": 0.94, + "learning_rate": 1.6880568496532775e-05, + "loss": 1.0561, + "step": 14639 + }, + { + "epoch": 0.94, + "learning_rate": 1.6880353807509824e-05, + "loss": 1.0813, + "step": 14640 + }, + { + "epoch": 0.94, + "learning_rate": 1.6880139118486873e-05, + "loss": 1.154, + "step": 14641 + }, + { + "epoch": 0.94, + "learning_rate": 1.6879924429463922e-05, + "loss": 0.9763, + "step": 14642 + }, + { + "epoch": 0.94, + "learning_rate": 1.687970974044097e-05, + "loss": 1.0022, + "step": 14643 + }, + { + "epoch": 0.94, + "learning_rate": 1.6879495051418023e-05, + "loss": 1.1121, + "step": 14644 + }, + { + "epoch": 0.94, + "learning_rate": 1.6879280362395072e-05, + "loss": 1.0, + "step": 14645 + }, + { + "epoch": 0.94, + "learning_rate": 1.687906567337212e-05, + "loss": 1.0259, + "step": 14646 + }, + { + "epoch": 0.94, + "learning_rate": 1.687885098434917e-05, + "loss": 1.0011, + "step": 14647 + }, + { + "epoch": 0.94, + "learning_rate": 1.687863629532622e-05, + "loss": 1.0679, + "step": 14648 + }, + { + "epoch": 0.94, + "learning_rate": 1.687842160630327e-05, + "loss": 1.0242, + "step": 14649 + }, + { + "epoch": 0.94, + "learning_rate": 1.687820691728032e-05, + "loss": 0.9816, + "step": 14650 + }, + { + "epoch": 0.94, + "learning_rate": 1.6877992228257373e-05, + "loss": 1.0966, + "step": 14651 + }, + { + "epoch": 0.94, + "learning_rate": 1.6877777539234422e-05, + "loss": 1.0232, + "step": 14652 + }, + { + "epoch": 0.94, + "learning_rate": 1.687756285021147e-05, + "loss": 1.0978, + "step": 14653 + }, + { + "epoch": 0.94, + "learning_rate": 1.687734816118852e-05, + "loss": 1.0653, + "step": 14654 + }, + { + "epoch": 0.94, + "learning_rate": 1.687713347216557e-05, + "loss": 1.1353, + "step": 14655 + }, + { + "epoch": 0.94, + "learning_rate": 1.687691878314262e-05, + "loss": 1.0028, + "step": 14656 + }, + { + "epoch": 0.94, + "learning_rate": 1.687670409411967e-05, + "loss": 0.9895, + "step": 14657 + }, + { + "epoch": 0.94, + "learning_rate": 1.687648940509672e-05, + "loss": 1.0881, + "step": 14658 + }, + { + "epoch": 0.94, + "learning_rate": 1.6876274716073768e-05, + "loss": 1.0534, + "step": 14659 + }, + { + "epoch": 0.94, + "learning_rate": 1.6876060027050817e-05, + "loss": 1.0515, + "step": 14660 + }, + { + "epoch": 0.94, + "learning_rate": 1.6875845338027866e-05, + "loss": 1.0175, + "step": 14661 + }, + { + "epoch": 0.94, + "learning_rate": 1.6875630649004918e-05, + "loss": 1.0374, + "step": 14662 + }, + { + "epoch": 0.94, + "learning_rate": 1.6875415959981967e-05, + "loss": 1.0216, + "step": 14663 + }, + { + "epoch": 0.94, + "learning_rate": 1.6875201270959016e-05, + "loss": 1.0264, + "step": 14664 + }, + { + "epoch": 0.94, + "learning_rate": 1.687498658193607e-05, + "loss": 1.1055, + "step": 14665 + }, + { + "epoch": 0.94, + "learning_rate": 1.6874771892913117e-05, + "loss": 1.1723, + "step": 14666 + }, + { + "epoch": 0.94, + "learning_rate": 1.6874557203890166e-05, + "loss": 1.0549, + "step": 14667 + }, + { + "epoch": 0.94, + "learning_rate": 1.6874342514867215e-05, + "loss": 1.1875, + "step": 14668 + }, + { + "epoch": 0.94, + "learning_rate": 1.6874127825844268e-05, + "loss": 1.1305, + "step": 14669 + }, + { + "epoch": 0.94, + "learning_rate": 1.6873913136821317e-05, + "loss": 1.0388, + "step": 14670 + }, + { + "epoch": 0.94, + "learning_rate": 1.6873698447798366e-05, + "loss": 0.9978, + "step": 14671 + }, + { + "epoch": 0.94, + "learning_rate": 1.6873483758775415e-05, + "loss": 1.0756, + "step": 14672 + }, + { + "epoch": 0.94, + "learning_rate": 1.6873269069752464e-05, + "loss": 1.046, + "step": 14673 + }, + { + "epoch": 0.94, + "learning_rate": 1.6873054380729516e-05, + "loss": 1.035, + "step": 14674 + }, + { + "epoch": 0.94, + "learning_rate": 1.6872839691706565e-05, + "loss": 1.0957, + "step": 14675 + }, + { + "epoch": 0.94, + "learning_rate": 1.6872625002683614e-05, + "loss": 0.9776, + "step": 14676 + }, + { + "epoch": 0.94, + "learning_rate": 1.6872410313660663e-05, + "loss": 1.0242, + "step": 14677 + }, + { + "epoch": 0.94, + "learning_rate": 1.6872195624637712e-05, + "loss": 1.0848, + "step": 14678 + }, + { + "epoch": 0.94, + "learning_rate": 1.687198093561476e-05, + "loss": 1.0046, + "step": 14679 + }, + { + "epoch": 0.94, + "learning_rate": 1.6871766246591813e-05, + "loss": 1.0261, + "step": 14680 + }, + { + "epoch": 0.94, + "learning_rate": 1.6871551557568862e-05, + "loss": 0.9707, + "step": 14681 + }, + { + "epoch": 0.94, + "learning_rate": 1.6871336868545914e-05, + "loss": 1.0452, + "step": 14682 + }, + { + "epoch": 0.94, + "learning_rate": 1.6871122179522963e-05, + "loss": 1.0035, + "step": 14683 + }, + { + "epoch": 0.94, + "learning_rate": 1.6870907490500012e-05, + "loss": 0.9867, + "step": 14684 + }, + { + "epoch": 0.94, + "learning_rate": 1.687069280147706e-05, + "loss": 1.0704, + "step": 14685 + }, + { + "epoch": 0.94, + "learning_rate": 1.687047811245411e-05, + "loss": 1.0828, + "step": 14686 + }, + { + "epoch": 0.94, + "learning_rate": 1.6870263423431163e-05, + "loss": 1.0663, + "step": 14687 + }, + { + "epoch": 0.94, + "learning_rate": 1.687004873440821e-05, + "loss": 1.1119, + "step": 14688 + }, + { + "epoch": 0.95, + "learning_rate": 1.686983404538526e-05, + "loss": 1.0221, + "step": 14689 + }, + { + "epoch": 0.95, + "learning_rate": 1.686961935636231e-05, + "loss": 1.0443, + "step": 14690 + }, + { + "epoch": 0.95, + "learning_rate": 1.686940466733936e-05, + "loss": 1.0113, + "step": 14691 + }, + { + "epoch": 0.95, + "learning_rate": 1.686918997831641e-05, + "loss": 1.1189, + "step": 14692 + }, + { + "epoch": 0.95, + "learning_rate": 1.686897528929346e-05, + "loss": 1.1121, + "step": 14693 + }, + { + "epoch": 0.95, + "learning_rate": 1.6868760600270512e-05, + "loss": 1.1332, + "step": 14694 + }, + { + "epoch": 0.95, + "learning_rate": 1.686854591124756e-05, + "loss": 1.0473, + "step": 14695 + }, + { + "epoch": 0.95, + "learning_rate": 1.686833122222461e-05, + "loss": 1.0551, + "step": 14696 + }, + { + "epoch": 0.95, + "learning_rate": 1.686811653320166e-05, + "loss": 1.0934, + "step": 14697 + }, + { + "epoch": 0.95, + "learning_rate": 1.6867901844178708e-05, + "loss": 1.0137, + "step": 14698 + }, + { + "epoch": 0.95, + "learning_rate": 1.6867687155155757e-05, + "loss": 1.1167, + "step": 14699 + }, + { + "epoch": 0.95, + "learning_rate": 1.686747246613281e-05, + "loss": 1.0628, + "step": 14700 + }, + { + "epoch": 0.95, + "learning_rate": 1.686725777710986e-05, + "loss": 1.0851, + "step": 14701 + }, + { + "epoch": 0.95, + "learning_rate": 1.6867043088086907e-05, + "loss": 0.993, + "step": 14702 + }, + { + "epoch": 0.95, + "learning_rate": 1.6866828399063956e-05, + "loss": 1.1041, + "step": 14703 + }, + { + "epoch": 0.95, + "learning_rate": 1.6866613710041005e-05, + "loss": 1.0726, + "step": 14704 + }, + { + "epoch": 0.95, + "learning_rate": 1.6866399021018058e-05, + "loss": 1.1099, + "step": 14705 + }, + { + "epoch": 0.95, + "learning_rate": 1.6866184331995107e-05, + "loss": 1.1208, + "step": 14706 + }, + { + "epoch": 0.95, + "learning_rate": 1.6865969642972156e-05, + "loss": 1.0519, + "step": 14707 + }, + { + "epoch": 0.95, + "learning_rate": 1.6865754953949205e-05, + "loss": 1.1139, + "step": 14708 + }, + { + "epoch": 0.95, + "learning_rate": 1.6865540264926257e-05, + "loss": 1.0673, + "step": 14709 + }, + { + "epoch": 0.95, + "learning_rate": 1.6865325575903306e-05, + "loss": 1.0004, + "step": 14710 + }, + { + "epoch": 0.95, + "learning_rate": 1.6865110886880355e-05, + "loss": 1.0966, + "step": 14711 + }, + { + "epoch": 0.95, + "learning_rate": 1.6864896197857407e-05, + "loss": 1.0841, + "step": 14712 + }, + { + "epoch": 0.95, + "learning_rate": 1.6864681508834456e-05, + "loss": 1.0612, + "step": 14713 + }, + { + "epoch": 0.95, + "learning_rate": 1.6864466819811505e-05, + "loss": 1.0667, + "step": 14714 + }, + { + "epoch": 0.95, + "learning_rate": 1.6864252130788554e-05, + "loss": 1.0239, + "step": 14715 + }, + { + "epoch": 0.95, + "learning_rate": 1.6864037441765603e-05, + "loss": 1.0038, + "step": 14716 + }, + { + "epoch": 0.95, + "learning_rate": 1.6863822752742652e-05, + "loss": 1.0594, + "step": 14717 + }, + { + "epoch": 0.95, + "learning_rate": 1.6863608063719704e-05, + "loss": 1.0127, + "step": 14718 + }, + { + "epoch": 0.95, + "learning_rate": 1.6863393374696753e-05, + "loss": 1.0412, + "step": 14719 + }, + { + "epoch": 0.95, + "learning_rate": 1.6863178685673802e-05, + "loss": 1.0625, + "step": 14720 + }, + { + "epoch": 0.95, + "learning_rate": 1.686296399665085e-05, + "loss": 0.9947, + "step": 14721 + }, + { + "epoch": 0.95, + "learning_rate": 1.68627493076279e-05, + "loss": 1.0341, + "step": 14722 + }, + { + "epoch": 0.95, + "learning_rate": 1.6862534618604953e-05, + "loss": 1.0358, + "step": 14723 + }, + { + "epoch": 0.95, + "learning_rate": 1.6862319929582e-05, + "loss": 1.0308, + "step": 14724 + }, + { + "epoch": 0.95, + "learning_rate": 1.6862105240559054e-05, + "loss": 1.0371, + "step": 14725 + }, + { + "epoch": 0.95, + "learning_rate": 1.6861890551536103e-05, + "loss": 1.1008, + "step": 14726 + }, + { + "epoch": 0.95, + "learning_rate": 1.6861675862513152e-05, + "loss": 1.0291, + "step": 14727 + }, + { + "epoch": 0.95, + "learning_rate": 1.68614611734902e-05, + "loss": 1.1421, + "step": 14728 + }, + { + "epoch": 0.95, + "learning_rate": 1.686124648446725e-05, + "loss": 1.0868, + "step": 14729 + }, + { + "epoch": 0.95, + "learning_rate": 1.68610317954443e-05, + "loss": 1.0925, + "step": 14730 + }, + { + "epoch": 0.95, + "learning_rate": 1.686081710642135e-05, + "loss": 1.1849, + "step": 14731 + }, + { + "epoch": 0.95, + "learning_rate": 1.68606024173984e-05, + "loss": 1.1467, + "step": 14732 + }, + { + "epoch": 0.95, + "learning_rate": 1.686038772837545e-05, + "loss": 1.0342, + "step": 14733 + }, + { + "epoch": 0.95, + "learning_rate": 1.6860173039352498e-05, + "loss": 1.0877, + "step": 14734 + }, + { + "epoch": 0.95, + "learning_rate": 1.6859958350329547e-05, + "loss": 1.0656, + "step": 14735 + }, + { + "epoch": 0.95, + "learning_rate": 1.68597436613066e-05, + "loss": 0.9942, + "step": 14736 + }, + { + "epoch": 0.95, + "learning_rate": 1.6859528972283648e-05, + "loss": 1.0747, + "step": 14737 + }, + { + "epoch": 0.95, + "learning_rate": 1.68593142832607e-05, + "loss": 1.0051, + "step": 14738 + }, + { + "epoch": 0.95, + "learning_rate": 1.685909959423775e-05, + "loss": 0.9377, + "step": 14739 + }, + { + "epoch": 0.95, + "learning_rate": 1.68588849052148e-05, + "loss": 1.0856, + "step": 14740 + }, + { + "epoch": 0.95, + "learning_rate": 1.6858670216191847e-05, + "loss": 1.0627, + "step": 14741 + }, + { + "epoch": 0.95, + "learning_rate": 1.6858455527168896e-05, + "loss": 1.0948, + "step": 14742 + }, + { + "epoch": 0.95, + "learning_rate": 1.685824083814595e-05, + "loss": 1.0569, + "step": 14743 + }, + { + "epoch": 0.95, + "learning_rate": 1.6858026149122998e-05, + "loss": 1.0644, + "step": 14744 + }, + { + "epoch": 0.95, + "learning_rate": 1.6857811460100047e-05, + "loss": 1.1687, + "step": 14745 + }, + { + "epoch": 0.95, + "learning_rate": 1.6857596771077096e-05, + "loss": 1.0428, + "step": 14746 + }, + { + "epoch": 0.95, + "learning_rate": 1.6857382082054145e-05, + "loss": 0.9865, + "step": 14747 + }, + { + "epoch": 0.95, + "learning_rate": 1.6857167393031194e-05, + "loss": 1.0854, + "step": 14748 + }, + { + "epoch": 0.95, + "learning_rate": 1.6856952704008246e-05, + "loss": 1.0693, + "step": 14749 + }, + { + "epoch": 0.95, + "learning_rate": 1.6856738014985295e-05, + "loss": 1.0134, + "step": 14750 + }, + { + "epoch": 0.95, + "learning_rate": 1.6856523325962344e-05, + "loss": 1.0219, + "step": 14751 + }, + { + "epoch": 0.95, + "learning_rate": 1.6856308636939393e-05, + "loss": 0.984, + "step": 14752 + }, + { + "epoch": 0.95, + "learning_rate": 1.6856093947916445e-05, + "loss": 1.1026, + "step": 14753 + }, + { + "epoch": 0.95, + "learning_rate": 1.6855879258893494e-05, + "loss": 1.0702, + "step": 14754 + }, + { + "epoch": 0.95, + "learning_rate": 1.6855664569870543e-05, + "loss": 1.0164, + "step": 14755 + }, + { + "epoch": 0.95, + "learning_rate": 1.6855449880847595e-05, + "loss": 1.0916, + "step": 14756 + }, + { + "epoch": 0.95, + "learning_rate": 1.6855235191824644e-05, + "loss": 1.0566, + "step": 14757 + }, + { + "epoch": 0.95, + "learning_rate": 1.6855020502801693e-05, + "loss": 1.0316, + "step": 14758 + }, + { + "epoch": 0.95, + "learning_rate": 1.6854805813778742e-05, + "loss": 1.1425, + "step": 14759 + }, + { + "epoch": 0.95, + "learning_rate": 1.685459112475579e-05, + "loss": 1.1045, + "step": 14760 + }, + { + "epoch": 0.95, + "learning_rate": 1.6854376435732844e-05, + "loss": 1.0759, + "step": 14761 + }, + { + "epoch": 0.95, + "learning_rate": 1.6854161746709893e-05, + "loss": 1.0163, + "step": 14762 + }, + { + "epoch": 0.95, + "learning_rate": 1.685394705768694e-05, + "loss": 1.1507, + "step": 14763 + }, + { + "epoch": 0.95, + "learning_rate": 1.685373236866399e-05, + "loss": 1.0575, + "step": 14764 + }, + { + "epoch": 0.95, + "learning_rate": 1.685351767964104e-05, + "loss": 1.0838, + "step": 14765 + }, + { + "epoch": 0.95, + "learning_rate": 1.685330299061809e-05, + "loss": 1.0796, + "step": 14766 + }, + { + "epoch": 0.95, + "learning_rate": 1.685308830159514e-05, + "loss": 1.0848, + "step": 14767 + }, + { + "epoch": 0.95, + "learning_rate": 1.685287361257219e-05, + "loss": 1.0925, + "step": 14768 + }, + { + "epoch": 0.95, + "learning_rate": 1.6852658923549242e-05, + "loss": 1.0354, + "step": 14769 + }, + { + "epoch": 0.95, + "learning_rate": 1.685244423452629e-05, + "loss": 1.0375, + "step": 14770 + }, + { + "epoch": 0.95, + "learning_rate": 1.685222954550334e-05, + "loss": 1.0981, + "step": 14771 + }, + { + "epoch": 0.95, + "learning_rate": 1.685201485648039e-05, + "loss": 0.9738, + "step": 14772 + }, + { + "epoch": 0.95, + "learning_rate": 1.6851800167457438e-05, + "loss": 1.0503, + "step": 14773 + }, + { + "epoch": 0.95, + "learning_rate": 1.685158547843449e-05, + "loss": 1.0486, + "step": 14774 + }, + { + "epoch": 0.95, + "learning_rate": 1.685137078941154e-05, + "loss": 1.087, + "step": 14775 + }, + { + "epoch": 0.95, + "learning_rate": 1.685115610038859e-05, + "loss": 1.033, + "step": 14776 + }, + { + "epoch": 0.95, + "learning_rate": 1.6850941411365637e-05, + "loss": 1.0881, + "step": 14777 + }, + { + "epoch": 0.95, + "learning_rate": 1.6850726722342686e-05, + "loss": 0.9941, + "step": 14778 + }, + { + "epoch": 0.95, + "learning_rate": 1.685051203331974e-05, + "loss": 1.066, + "step": 14779 + }, + { + "epoch": 0.95, + "learning_rate": 1.6850297344296788e-05, + "loss": 1.0406, + "step": 14780 + }, + { + "epoch": 0.95, + "learning_rate": 1.6850082655273837e-05, + "loss": 1.0194, + "step": 14781 + }, + { + "epoch": 0.95, + "learning_rate": 1.684986796625089e-05, + "loss": 1.0346, + "step": 14782 + }, + { + "epoch": 0.95, + "learning_rate": 1.6849653277227938e-05, + "loss": 1.1374, + "step": 14783 + }, + { + "epoch": 0.95, + "learning_rate": 1.6849438588204987e-05, + "loss": 1.0689, + "step": 14784 + }, + { + "epoch": 0.95, + "learning_rate": 1.6849223899182036e-05, + "loss": 1.1222, + "step": 14785 + }, + { + "epoch": 0.95, + "learning_rate": 1.6849009210159085e-05, + "loss": 1.0722, + "step": 14786 + }, + { + "epoch": 0.95, + "learning_rate": 1.6848794521136137e-05, + "loss": 1.1024, + "step": 14787 + }, + { + "epoch": 0.95, + "learning_rate": 1.6848579832113186e-05, + "loss": 1.1027, + "step": 14788 + }, + { + "epoch": 0.95, + "learning_rate": 1.6848365143090235e-05, + "loss": 1.0095, + "step": 14789 + }, + { + "epoch": 0.95, + "learning_rate": 1.6848150454067284e-05, + "loss": 1.0737, + "step": 14790 + }, + { + "epoch": 0.95, + "learning_rate": 1.6847935765044333e-05, + "loss": 1.0172, + "step": 14791 + }, + { + "epoch": 0.95, + "learning_rate": 1.6847721076021385e-05, + "loss": 0.9513, + "step": 14792 + }, + { + "epoch": 0.95, + "learning_rate": 1.6847506386998434e-05, + "loss": 1.0017, + "step": 14793 + }, + { + "epoch": 0.95, + "learning_rate": 1.6847291697975483e-05, + "loss": 1.0973, + "step": 14794 + }, + { + "epoch": 0.95, + "learning_rate": 1.6847077008952532e-05, + "loss": 1.0827, + "step": 14795 + }, + { + "epoch": 0.95, + "learning_rate": 1.6846862319929585e-05, + "loss": 0.9987, + "step": 14796 + }, + { + "epoch": 0.95, + "learning_rate": 1.6846647630906634e-05, + "loss": 1.0802, + "step": 14797 + }, + { + "epoch": 0.95, + "learning_rate": 1.6846432941883683e-05, + "loss": 1.0302, + "step": 14798 + }, + { + "epoch": 0.95, + "learning_rate": 1.6846218252860735e-05, + "loss": 1.0801, + "step": 14799 + }, + { + "epoch": 0.95, + "learning_rate": 1.6846003563837784e-05, + "loss": 1.0922, + "step": 14800 + }, + { + "epoch": 0.95, + "learning_rate": 1.6845788874814833e-05, + "loss": 1.0272, + "step": 14801 + }, + { + "epoch": 0.95, + "learning_rate": 1.6845574185791882e-05, + "loss": 1.121, + "step": 14802 + }, + { + "epoch": 0.95, + "learning_rate": 1.684535949676893e-05, + "loss": 1.0779, + "step": 14803 + }, + { + "epoch": 0.95, + "learning_rate": 1.684514480774598e-05, + "loss": 1.0916, + "step": 14804 + }, + { + "epoch": 0.95, + "learning_rate": 1.6844930118723032e-05, + "loss": 0.984, + "step": 14805 + }, + { + "epoch": 0.95, + "learning_rate": 1.684471542970008e-05, + "loss": 1.0245, + "step": 14806 + }, + { + "epoch": 0.95, + "learning_rate": 1.684450074067713e-05, + "loss": 1.1131, + "step": 14807 + }, + { + "epoch": 0.95, + "learning_rate": 1.684428605165418e-05, + "loss": 1.0091, + "step": 14808 + }, + { + "epoch": 0.95, + "learning_rate": 1.6844071362631228e-05, + "loss": 1.1092, + "step": 14809 + }, + { + "epoch": 0.95, + "learning_rate": 1.684385667360828e-05, + "loss": 1.1332, + "step": 14810 + }, + { + "epoch": 0.95, + "learning_rate": 1.684364198458533e-05, + "loss": 1.0575, + "step": 14811 + }, + { + "epoch": 0.95, + "learning_rate": 1.684342729556238e-05, + "loss": 1.0043, + "step": 14812 + }, + { + "epoch": 0.95, + "learning_rate": 1.684321260653943e-05, + "loss": 1.0521, + "step": 14813 + }, + { + "epoch": 0.95, + "learning_rate": 1.684299791751648e-05, + "loss": 1.0774, + "step": 14814 + }, + { + "epoch": 0.95, + "learning_rate": 1.684278322849353e-05, + "loss": 1.0974, + "step": 14815 + }, + { + "epoch": 0.95, + "learning_rate": 1.6842568539470577e-05, + "loss": 0.984, + "step": 14816 + }, + { + "epoch": 0.95, + "learning_rate": 1.6842353850447626e-05, + "loss": 1.0281, + "step": 14817 + }, + { + "epoch": 0.95, + "learning_rate": 1.684213916142468e-05, + "loss": 1.0833, + "step": 14818 + }, + { + "epoch": 0.95, + "learning_rate": 1.6841924472401728e-05, + "loss": 1.0503, + "step": 14819 + }, + { + "epoch": 0.95, + "learning_rate": 1.6841709783378777e-05, + "loss": 1.051, + "step": 14820 + }, + { + "epoch": 0.95, + "learning_rate": 1.6841495094355826e-05, + "loss": 1.0008, + "step": 14821 + }, + { + "epoch": 0.95, + "learning_rate": 1.6841280405332875e-05, + "loss": 1.1193, + "step": 14822 + }, + { + "epoch": 0.95, + "learning_rate": 1.6841065716309927e-05, + "loss": 1.0882, + "step": 14823 + }, + { + "epoch": 0.95, + "learning_rate": 1.6840851027286976e-05, + "loss": 1.0534, + "step": 14824 + }, + { + "epoch": 0.95, + "learning_rate": 1.6840636338264028e-05, + "loss": 1.0583, + "step": 14825 + }, + { + "epoch": 0.95, + "learning_rate": 1.6840421649241077e-05, + "loss": 1.1105, + "step": 14826 + }, + { + "epoch": 0.95, + "learning_rate": 1.6840206960218126e-05, + "loss": 1.01, + "step": 14827 + }, + { + "epoch": 0.95, + "learning_rate": 1.6839992271195175e-05, + "loss": 1.0036, + "step": 14828 + }, + { + "epoch": 0.95, + "learning_rate": 1.6839777582172224e-05, + "loss": 1.0875, + "step": 14829 + }, + { + "epoch": 0.95, + "learning_rate": 1.6839562893149276e-05, + "loss": 0.9729, + "step": 14830 + }, + { + "epoch": 0.95, + "learning_rate": 1.6839348204126325e-05, + "loss": 0.9707, + "step": 14831 + }, + { + "epoch": 0.95, + "learning_rate": 1.6839133515103374e-05, + "loss": 1.0399, + "step": 14832 + }, + { + "epoch": 0.95, + "learning_rate": 1.6838918826080423e-05, + "loss": 1.0928, + "step": 14833 + }, + { + "epoch": 0.95, + "learning_rate": 1.6838704137057472e-05, + "loss": 1.0971, + "step": 14834 + }, + { + "epoch": 0.95, + "learning_rate": 1.683848944803452e-05, + "loss": 1.0755, + "step": 14835 + }, + { + "epoch": 0.95, + "learning_rate": 1.6838274759011574e-05, + "loss": 1.0769, + "step": 14836 + }, + { + "epoch": 0.95, + "learning_rate": 1.6838060069988623e-05, + "loss": 1.0881, + "step": 14837 + }, + { + "epoch": 0.95, + "learning_rate": 1.683784538096567e-05, + "loss": 1.124, + "step": 14838 + }, + { + "epoch": 0.95, + "learning_rate": 1.683763069194272e-05, + "loss": 1.0136, + "step": 14839 + }, + { + "epoch": 0.95, + "learning_rate": 1.6837416002919773e-05, + "loss": 1.0522, + "step": 14840 + }, + { + "epoch": 0.95, + "learning_rate": 1.6837201313896822e-05, + "loss": 1.1264, + "step": 14841 + }, + { + "epoch": 0.95, + "learning_rate": 1.683698662487387e-05, + "loss": 0.9833, + "step": 14842 + }, + { + "epoch": 0.95, + "learning_rate": 1.6836771935850923e-05, + "loss": 1.0549, + "step": 14843 + }, + { + "epoch": 0.96, + "learning_rate": 1.6836557246827972e-05, + "loss": 1.0319, + "step": 14844 + }, + { + "epoch": 0.96, + "learning_rate": 1.683634255780502e-05, + "loss": 1.1125, + "step": 14845 + }, + { + "epoch": 0.96, + "learning_rate": 1.683612786878207e-05, + "loss": 1.0135, + "step": 14846 + }, + { + "epoch": 0.96, + "learning_rate": 1.683591317975912e-05, + "loss": 1.1454, + "step": 14847 + }, + { + "epoch": 0.96, + "learning_rate": 1.683569849073617e-05, + "loss": 1.104, + "step": 14848 + }, + { + "epoch": 0.96, + "learning_rate": 1.683548380171322e-05, + "loss": 0.9776, + "step": 14849 + }, + { + "epoch": 0.96, + "learning_rate": 1.683526911269027e-05, + "loss": 1.0183, + "step": 14850 + }, + { + "epoch": 0.96, + "learning_rate": 1.683505442366732e-05, + "loss": 1.1096, + "step": 14851 + }, + { + "epoch": 0.96, + "learning_rate": 1.6834839734644367e-05, + "loss": 1.0861, + "step": 14852 + }, + { + "epoch": 0.96, + "learning_rate": 1.6834625045621416e-05, + "loss": 0.9827, + "step": 14853 + }, + { + "epoch": 0.96, + "learning_rate": 1.683441035659847e-05, + "loss": 1.0518, + "step": 14854 + }, + { + "epoch": 0.96, + "learning_rate": 1.6834195667575518e-05, + "loss": 1.0233, + "step": 14855 + }, + { + "epoch": 0.96, + "learning_rate": 1.683398097855257e-05, + "loss": 1.0106, + "step": 14856 + }, + { + "epoch": 0.96, + "learning_rate": 1.683376628952962e-05, + "loss": 1.1167, + "step": 14857 + }, + { + "epoch": 0.96, + "learning_rate": 1.6833551600506668e-05, + "loss": 1.0229, + "step": 14858 + }, + { + "epoch": 0.96, + "learning_rate": 1.6833336911483717e-05, + "loss": 0.997, + "step": 14859 + }, + { + "epoch": 0.96, + "learning_rate": 1.6833122222460766e-05, + "loss": 1.1067, + "step": 14860 + }, + { + "epoch": 0.96, + "learning_rate": 1.6832907533437818e-05, + "loss": 1.0352, + "step": 14861 + }, + { + "epoch": 0.96, + "learning_rate": 1.6832692844414867e-05, + "loss": 1.0305, + "step": 14862 + }, + { + "epoch": 0.96, + "learning_rate": 1.6832478155391916e-05, + "loss": 1.0337, + "step": 14863 + }, + { + "epoch": 0.96, + "learning_rate": 1.6832263466368965e-05, + "loss": 1.0373, + "step": 14864 + }, + { + "epoch": 0.96, + "learning_rate": 1.6832048777346014e-05, + "loss": 1.1177, + "step": 14865 + }, + { + "epoch": 0.96, + "learning_rate": 1.6831834088323066e-05, + "loss": 1.0387, + "step": 14866 + }, + { + "epoch": 0.96, + "learning_rate": 1.6831619399300115e-05, + "loss": 1.0304, + "step": 14867 + }, + { + "epoch": 0.96, + "learning_rate": 1.6831404710277164e-05, + "loss": 1.0431, + "step": 14868 + }, + { + "epoch": 0.96, + "learning_rate": 1.6831190021254217e-05, + "loss": 1.0871, + "step": 14869 + }, + { + "epoch": 0.96, + "learning_rate": 1.6830975332231266e-05, + "loss": 0.9526, + "step": 14870 + }, + { + "epoch": 0.96, + "learning_rate": 1.6830760643208315e-05, + "loss": 1.0566, + "step": 14871 + }, + { + "epoch": 0.96, + "learning_rate": 1.6830545954185364e-05, + "loss": 1.0949, + "step": 14872 + }, + { + "epoch": 0.96, + "learning_rate": 1.6830331265162412e-05, + "loss": 0.996, + "step": 14873 + }, + { + "epoch": 0.96, + "learning_rate": 1.6830116576139465e-05, + "loss": 1.0829, + "step": 14874 + }, + { + "epoch": 0.96, + "learning_rate": 1.6829901887116514e-05, + "loss": 1.0563, + "step": 14875 + }, + { + "epoch": 0.96, + "learning_rate": 1.6829687198093563e-05, + "loss": 1.0015, + "step": 14876 + }, + { + "epoch": 0.96, + "learning_rate": 1.6829472509070612e-05, + "loss": 1.0585, + "step": 14877 + }, + { + "epoch": 0.96, + "learning_rate": 1.682925782004766e-05, + "loss": 1.0337, + "step": 14878 + }, + { + "epoch": 0.96, + "learning_rate": 1.6829043131024713e-05, + "loss": 1.0343, + "step": 14879 + }, + { + "epoch": 0.96, + "learning_rate": 1.6828828442001762e-05, + "loss": 0.9995, + "step": 14880 + }, + { + "epoch": 0.96, + "learning_rate": 1.682861375297881e-05, + "loss": 1.1257, + "step": 14881 + }, + { + "epoch": 0.96, + "learning_rate": 1.682839906395586e-05, + "loss": 1.0127, + "step": 14882 + }, + { + "epoch": 0.96, + "learning_rate": 1.6828184374932912e-05, + "loss": 1.0022, + "step": 14883 + }, + { + "epoch": 0.96, + "learning_rate": 1.682796968590996e-05, + "loss": 1.0573, + "step": 14884 + }, + { + "epoch": 0.96, + "learning_rate": 1.682775499688701e-05, + "loss": 1.0045, + "step": 14885 + }, + { + "epoch": 0.96, + "learning_rate": 1.6827540307864063e-05, + "loss": 1.0781, + "step": 14886 + }, + { + "epoch": 0.96, + "learning_rate": 1.682732561884111e-05, + "loss": 1.1069, + "step": 14887 + }, + { + "epoch": 0.96, + "learning_rate": 1.682711092981816e-05, + "loss": 1.0098, + "step": 14888 + }, + { + "epoch": 0.96, + "learning_rate": 1.682689624079521e-05, + "loss": 1.0139, + "step": 14889 + }, + { + "epoch": 0.96, + "learning_rate": 1.682668155177226e-05, + "loss": 0.9585, + "step": 14890 + }, + { + "epoch": 0.96, + "learning_rate": 1.6826466862749307e-05, + "loss": 1.0742, + "step": 14891 + }, + { + "epoch": 0.96, + "learning_rate": 1.682625217372636e-05, + "loss": 1.1498, + "step": 14892 + }, + { + "epoch": 0.96, + "learning_rate": 1.682603748470341e-05, + "loss": 1.0415, + "step": 14893 + }, + { + "epoch": 0.96, + "learning_rate": 1.6825822795680458e-05, + "loss": 1.0521, + "step": 14894 + }, + { + "epoch": 0.96, + "learning_rate": 1.6825608106657507e-05, + "loss": 0.9797, + "step": 14895 + }, + { + "epoch": 0.96, + "learning_rate": 1.6825393417634556e-05, + "loss": 0.9992, + "step": 14896 + }, + { + "epoch": 0.96, + "learning_rate": 1.6825178728611608e-05, + "loss": 1.0781, + "step": 14897 + }, + { + "epoch": 0.96, + "learning_rate": 1.6824964039588657e-05, + "loss": 1.0571, + "step": 14898 + }, + { + "epoch": 0.96, + "learning_rate": 1.682474935056571e-05, + "loss": 1.0271, + "step": 14899 + }, + { + "epoch": 0.96, + "learning_rate": 1.6824534661542758e-05, + "loss": 1.0851, + "step": 14900 + }, + { + "epoch": 0.96, + "learning_rate": 1.6824319972519807e-05, + "loss": 1.1257, + "step": 14901 + }, + { + "epoch": 0.96, + "learning_rate": 1.6824105283496856e-05, + "loss": 0.9774, + "step": 14902 + }, + { + "epoch": 0.96, + "learning_rate": 1.6823890594473905e-05, + "loss": 1.0562, + "step": 14903 + }, + { + "epoch": 0.96, + "learning_rate": 1.6823675905450954e-05, + "loss": 1.0581, + "step": 14904 + }, + { + "epoch": 0.96, + "learning_rate": 1.6823461216428006e-05, + "loss": 1.0552, + "step": 14905 + }, + { + "epoch": 0.96, + "learning_rate": 1.6823246527405055e-05, + "loss": 1.1066, + "step": 14906 + }, + { + "epoch": 0.96, + "learning_rate": 1.6823031838382104e-05, + "loss": 1.0686, + "step": 14907 + }, + { + "epoch": 0.96, + "learning_rate": 1.6822817149359153e-05, + "loss": 1.0309, + "step": 14908 + }, + { + "epoch": 0.96, + "learning_rate": 1.6822602460336202e-05, + "loss": 1.0523, + "step": 14909 + }, + { + "epoch": 0.96, + "learning_rate": 1.6822387771313255e-05, + "loss": 1.1114, + "step": 14910 + }, + { + "epoch": 0.96, + "learning_rate": 1.6822173082290304e-05, + "loss": 1.0862, + "step": 14911 + }, + { + "epoch": 0.96, + "learning_rate": 1.6821958393267356e-05, + "loss": 1.0181, + "step": 14912 + }, + { + "epoch": 0.96, + "learning_rate": 1.6821743704244405e-05, + "loss": 1.0021, + "step": 14913 + }, + { + "epoch": 0.96, + "learning_rate": 1.6821529015221454e-05, + "loss": 1.1211, + "step": 14914 + }, + { + "epoch": 0.96, + "learning_rate": 1.6821314326198503e-05, + "loss": 1.1121, + "step": 14915 + }, + { + "epoch": 0.96, + "learning_rate": 1.6821099637175552e-05, + "loss": 1.0822, + "step": 14916 + }, + { + "epoch": 0.96, + "learning_rate": 1.6820884948152604e-05, + "loss": 1.0558, + "step": 14917 + }, + { + "epoch": 0.96, + "learning_rate": 1.6820670259129653e-05, + "loss": 1.1749, + "step": 14918 + }, + { + "epoch": 0.96, + "learning_rate": 1.6820455570106702e-05, + "loss": 1.099, + "step": 14919 + }, + { + "epoch": 0.96, + "learning_rate": 1.682024088108375e-05, + "loss": 1.0685, + "step": 14920 + }, + { + "epoch": 0.96, + "learning_rate": 1.68200261920608e-05, + "loss": 1.0492, + "step": 14921 + }, + { + "epoch": 0.96, + "learning_rate": 1.681981150303785e-05, + "loss": 1.0274, + "step": 14922 + }, + { + "epoch": 0.96, + "learning_rate": 1.68195968140149e-05, + "loss": 1.0086, + "step": 14923 + }, + { + "epoch": 0.96, + "learning_rate": 1.681938212499195e-05, + "loss": 1.0775, + "step": 14924 + }, + { + "epoch": 0.96, + "learning_rate": 1.6819167435969e-05, + "loss": 1.0809, + "step": 14925 + }, + { + "epoch": 0.96, + "learning_rate": 1.6818952746946048e-05, + "loss": 0.9902, + "step": 14926 + }, + { + "epoch": 0.96, + "learning_rate": 1.68187380579231e-05, + "loss": 1.0584, + "step": 14927 + }, + { + "epoch": 0.96, + "learning_rate": 1.681852336890015e-05, + "loss": 1.0224, + "step": 14928 + }, + { + "epoch": 0.96, + "learning_rate": 1.68183086798772e-05, + "loss": 1.097, + "step": 14929 + }, + { + "epoch": 0.96, + "learning_rate": 1.681809399085425e-05, + "loss": 1.0626, + "step": 14930 + }, + { + "epoch": 0.96, + "learning_rate": 1.68178793018313e-05, + "loss": 1.0566, + "step": 14931 + }, + { + "epoch": 0.96, + "learning_rate": 1.681766461280835e-05, + "loss": 1.0556, + "step": 14932 + }, + { + "epoch": 0.96, + "learning_rate": 1.6817449923785398e-05, + "loss": 1.0617, + "step": 14933 + }, + { + "epoch": 0.96, + "learning_rate": 1.6817235234762447e-05, + "loss": 1.092, + "step": 14934 + }, + { + "epoch": 0.96, + "learning_rate": 1.68170205457395e-05, + "loss": 1.0943, + "step": 14935 + }, + { + "epoch": 0.96, + "learning_rate": 1.6816805856716548e-05, + "loss": 1.0616, + "step": 14936 + }, + { + "epoch": 0.96, + "learning_rate": 1.6816591167693597e-05, + "loss": 1.0656, + "step": 14937 + }, + { + "epoch": 0.96, + "learning_rate": 1.6816376478670646e-05, + "loss": 1.024, + "step": 14938 + }, + { + "epoch": 0.96, + "learning_rate": 1.6816161789647695e-05, + "loss": 1.1066, + "step": 14939 + }, + { + "epoch": 0.96, + "learning_rate": 1.6815947100624744e-05, + "loss": 1.021, + "step": 14940 + }, + { + "epoch": 0.96, + "learning_rate": 1.6815732411601796e-05, + "loss": 0.9919, + "step": 14941 + }, + { + "epoch": 0.96, + "learning_rate": 1.6815517722578845e-05, + "loss": 1.0618, + "step": 14942 + }, + { + "epoch": 0.96, + "learning_rate": 1.6815303033555898e-05, + "loss": 1.1362, + "step": 14943 + }, + { + "epoch": 0.96, + "learning_rate": 1.6815088344532947e-05, + "loss": 1.0634, + "step": 14944 + }, + { + "epoch": 0.96, + "learning_rate": 1.6814873655509996e-05, + "loss": 1.0627, + "step": 14945 + }, + { + "epoch": 0.96, + "learning_rate": 1.6814658966487045e-05, + "loss": 1.0454, + "step": 14946 + }, + { + "epoch": 0.96, + "learning_rate": 1.6814444277464093e-05, + "loss": 1.0003, + "step": 14947 + }, + { + "epoch": 0.96, + "learning_rate": 1.6814229588441146e-05, + "loss": 1.1172, + "step": 14948 + }, + { + "epoch": 0.96, + "learning_rate": 1.6814014899418195e-05, + "loss": 1.0823, + "step": 14949 + }, + { + "epoch": 0.96, + "learning_rate": 1.6813800210395244e-05, + "loss": 1.0307, + "step": 14950 + }, + { + "epoch": 0.96, + "learning_rate": 1.6813585521372293e-05, + "loss": 1.0084, + "step": 14951 + }, + { + "epoch": 0.96, + "learning_rate": 1.681337083234934e-05, + "loss": 1.1501, + "step": 14952 + }, + { + "epoch": 0.96, + "learning_rate": 1.681315614332639e-05, + "loss": 1.0304, + "step": 14953 + }, + { + "epoch": 0.96, + "learning_rate": 1.6812941454303443e-05, + "loss": 1.0252, + "step": 14954 + }, + { + "epoch": 0.96, + "learning_rate": 1.6812726765280492e-05, + "loss": 0.9847, + "step": 14955 + }, + { + "epoch": 0.96, + "learning_rate": 1.6812512076257544e-05, + "loss": 1.0236, + "step": 14956 + }, + { + "epoch": 0.96, + "learning_rate": 1.6812297387234593e-05, + "loss": 1.064, + "step": 14957 + }, + { + "epoch": 0.96, + "learning_rate": 1.6812082698211642e-05, + "loss": 1.0959, + "step": 14958 + }, + { + "epoch": 0.96, + "learning_rate": 1.681186800918869e-05, + "loss": 1.0513, + "step": 14959 + }, + { + "epoch": 0.96, + "learning_rate": 1.681165332016574e-05, + "loss": 1.0651, + "step": 14960 + }, + { + "epoch": 0.96, + "learning_rate": 1.6811438631142793e-05, + "loss": 1.158, + "step": 14961 + }, + { + "epoch": 0.96, + "learning_rate": 1.681122394211984e-05, + "loss": 1.0615, + "step": 14962 + }, + { + "epoch": 0.96, + "learning_rate": 1.681100925309689e-05, + "loss": 1.049, + "step": 14963 + }, + { + "epoch": 0.96, + "learning_rate": 1.681079456407394e-05, + "loss": 0.9994, + "step": 14964 + }, + { + "epoch": 0.96, + "learning_rate": 1.681057987505099e-05, + "loss": 0.994, + "step": 14965 + }, + { + "epoch": 0.96, + "learning_rate": 1.681036518602804e-05, + "loss": 0.972, + "step": 14966 + }, + { + "epoch": 0.96, + "learning_rate": 1.681015049700509e-05, + "loss": 1.032, + "step": 14967 + }, + { + "epoch": 0.96, + "learning_rate": 1.680993580798214e-05, + "loss": 1.033, + "step": 14968 + }, + { + "epoch": 0.96, + "learning_rate": 1.6809721118959188e-05, + "loss": 1.0629, + "step": 14969 + }, + { + "epoch": 0.96, + "learning_rate": 1.680950642993624e-05, + "loss": 1.0094, + "step": 14970 + }, + { + "epoch": 0.96, + "learning_rate": 1.680929174091329e-05, + "loss": 1.0582, + "step": 14971 + }, + { + "epoch": 0.96, + "learning_rate": 1.6809077051890338e-05, + "loss": 0.9723, + "step": 14972 + }, + { + "epoch": 0.96, + "learning_rate": 1.680886236286739e-05, + "loss": 1.0723, + "step": 14973 + }, + { + "epoch": 0.96, + "learning_rate": 1.680864767384444e-05, + "loss": 1.1214, + "step": 14974 + }, + { + "epoch": 0.96, + "learning_rate": 1.6808432984821488e-05, + "loss": 1.117, + "step": 14975 + }, + { + "epoch": 0.96, + "learning_rate": 1.6808218295798537e-05, + "loss": 1.0681, + "step": 14976 + }, + { + "epoch": 0.96, + "learning_rate": 1.6808003606775586e-05, + "loss": 1.0942, + "step": 14977 + }, + { + "epoch": 0.96, + "learning_rate": 1.6807788917752635e-05, + "loss": 0.989, + "step": 14978 + }, + { + "epoch": 0.96, + "learning_rate": 1.6807574228729687e-05, + "loss": 1.0285, + "step": 14979 + }, + { + "epoch": 0.96, + "learning_rate": 1.6807359539706736e-05, + "loss": 1.0348, + "step": 14980 + }, + { + "epoch": 0.96, + "learning_rate": 1.6807144850683785e-05, + "loss": 1.0113, + "step": 14981 + }, + { + "epoch": 0.96, + "learning_rate": 1.6806930161660834e-05, + "loss": 1.1046, + "step": 14982 + }, + { + "epoch": 0.96, + "learning_rate": 1.6806715472637883e-05, + "loss": 1.0384, + "step": 14983 + }, + { + "epoch": 0.96, + "learning_rate": 1.6806500783614936e-05, + "loss": 1.0769, + "step": 14984 + }, + { + "epoch": 0.96, + "learning_rate": 1.6806286094591985e-05, + "loss": 1.0509, + "step": 14985 + }, + { + "epoch": 0.96, + "learning_rate": 1.6806071405569037e-05, + "loss": 1.1743, + "step": 14986 + }, + { + "epoch": 0.96, + "learning_rate": 1.6805856716546086e-05, + "loss": 1.0428, + "step": 14987 + }, + { + "epoch": 0.96, + "learning_rate": 1.6805642027523135e-05, + "loss": 1.1735, + "step": 14988 + }, + { + "epoch": 0.96, + "learning_rate": 1.6805427338500184e-05, + "loss": 1.036, + "step": 14989 + }, + { + "epoch": 0.96, + "learning_rate": 1.6805212649477233e-05, + "loss": 1.0243, + "step": 14990 + }, + { + "epoch": 0.96, + "learning_rate": 1.6804997960454282e-05, + "loss": 1.0647, + "step": 14991 + }, + { + "epoch": 0.96, + "learning_rate": 1.6804783271431334e-05, + "loss": 1.0158, + "step": 14992 + }, + { + "epoch": 0.96, + "learning_rate": 1.6804568582408383e-05, + "loss": 1.0651, + "step": 14993 + }, + { + "epoch": 0.96, + "learning_rate": 1.6804353893385432e-05, + "loss": 0.9975, + "step": 14994 + }, + { + "epoch": 0.96, + "learning_rate": 1.680413920436248e-05, + "loss": 1.1232, + "step": 14995 + }, + { + "epoch": 0.96, + "learning_rate": 1.680392451533953e-05, + "loss": 1.0456, + "step": 14996 + }, + { + "epoch": 0.96, + "learning_rate": 1.6803709826316582e-05, + "loss": 1.1505, + "step": 14997 + }, + { + "epoch": 0.96, + "learning_rate": 1.680349513729363e-05, + "loss": 1.0087, + "step": 14998 + }, + { + "epoch": 0.97, + "learning_rate": 1.6803280448270684e-05, + "loss": 1.0063, + "step": 14999 + }, + { + "epoch": 0.97, + "learning_rate": 1.6803065759247733e-05, + "loss": 1.018, + "step": 15000 + }, + { + "epoch": 0.97, + "eval_laion/cider": 0.6067123037618357, + "eval_laion/meteor": 0.39342206160742876, + "eval_laion/pred_num_words": 36.93359375, + "eval_laion/sacrebleu": 16.096921489437996, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0672409534454346, + "eval_runtime": 104.8106, + "eval_samples_per_second": 4.761, + "eval_steps_per_second": 0.305, + "step": 15000 + }, + { + "epoch": 0.97, + "learning_rate": 1.680285107022478e-05, + "loss": 1.188, + "step": 15001 + }, + { + "epoch": 0.97, + "learning_rate": 1.680263638120183e-05, + "loss": 1.1105, + "step": 15002 + }, + { + "epoch": 0.97, + "learning_rate": 1.680242169217888e-05, + "loss": 1.0193, + "step": 15003 + }, + { + "epoch": 0.97, + "learning_rate": 1.6802207003155932e-05, + "loss": 1.0275, + "step": 15004 + }, + { + "epoch": 0.97, + "learning_rate": 1.680199231413298e-05, + "loss": 1.0573, + "step": 15005 + }, + { + "epoch": 0.97, + "learning_rate": 1.680177762511003e-05, + "loss": 1.0992, + "step": 15006 + }, + { + "epoch": 0.97, + "learning_rate": 1.680156293608708e-05, + "loss": 1.0912, + "step": 15007 + }, + { + "epoch": 0.97, + "learning_rate": 1.6801348247064128e-05, + "loss": 1.0384, + "step": 15008 + }, + { + "epoch": 0.97, + "learning_rate": 1.6801133558041177e-05, + "loss": 1.0031, + "step": 15009 + }, + { + "epoch": 0.97, + "learning_rate": 1.680091886901823e-05, + "loss": 1.1272, + "step": 15010 + }, + { + "epoch": 0.97, + "learning_rate": 1.6800704179995278e-05, + "loss": 1.0707, + "step": 15011 + }, + { + "epoch": 0.97, + "learning_rate": 1.6800489490972327e-05, + "loss": 0.959, + "step": 15012 + }, + { + "epoch": 0.97, + "learning_rate": 1.6800274801949376e-05, + "loss": 1.0182, + "step": 15013 + }, + { + "epoch": 0.97, + "learning_rate": 1.680006011292643e-05, + "loss": 1.1108, + "step": 15014 + }, + { + "epoch": 0.97, + "learning_rate": 1.6799845423903477e-05, + "loss": 1.0072, + "step": 15015 + }, + { + "epoch": 0.97, + "learning_rate": 1.6799630734880526e-05, + "loss": 0.9924, + "step": 15016 + }, + { + "epoch": 0.97, + "learning_rate": 1.679941604585758e-05, + "loss": 1.1183, + "step": 15017 + }, + { + "epoch": 0.97, + "learning_rate": 1.6799201356834628e-05, + "loss": 1.0757, + "step": 15018 + }, + { + "epoch": 0.97, + "learning_rate": 1.6798986667811677e-05, + "loss": 1.1591, + "step": 15019 + }, + { + "epoch": 0.97, + "learning_rate": 1.6798771978788726e-05, + "loss": 1.0692, + "step": 15020 + }, + { + "epoch": 0.97, + "learning_rate": 1.6798557289765774e-05, + "loss": 0.9594, + "step": 15021 + }, + { + "epoch": 0.97, + "learning_rate": 1.6798342600742827e-05, + "loss": 1.0006, + "step": 15022 + }, + { + "epoch": 0.97, + "learning_rate": 1.6798127911719876e-05, + "loss": 1.0982, + "step": 15023 + }, + { + "epoch": 0.97, + "learning_rate": 1.6797913222696925e-05, + "loss": 0.9854, + "step": 15024 + }, + { + "epoch": 0.97, + "learning_rate": 1.6797698533673974e-05, + "loss": 1.0106, + "step": 15025 + }, + { + "epoch": 0.97, + "learning_rate": 1.6797483844651023e-05, + "loss": 1.0667, + "step": 15026 + }, + { + "epoch": 0.97, + "learning_rate": 1.679726915562807e-05, + "loss": 1.1032, + "step": 15027 + }, + { + "epoch": 0.97, + "learning_rate": 1.6797054466605124e-05, + "loss": 1.1531, + "step": 15028 + }, + { + "epoch": 0.97, + "learning_rate": 1.6796839777582173e-05, + "loss": 1.1298, + "step": 15029 + }, + { + "epoch": 0.97, + "learning_rate": 1.6796625088559225e-05, + "loss": 1.1416, + "step": 15030 + }, + { + "epoch": 0.97, + "learning_rate": 1.6796410399536274e-05, + "loss": 1.0657, + "step": 15031 + }, + { + "epoch": 0.97, + "learning_rate": 1.6796195710513323e-05, + "loss": 1.1292, + "step": 15032 + }, + { + "epoch": 0.97, + "learning_rate": 1.6795981021490372e-05, + "loss": 1.1758, + "step": 15033 + }, + { + "epoch": 0.97, + "learning_rate": 1.679576633246742e-05, + "loss": 1.0032, + "step": 15034 + }, + { + "epoch": 0.97, + "learning_rate": 1.6795551643444474e-05, + "loss": 1.0467, + "step": 15035 + }, + { + "epoch": 0.97, + "learning_rate": 1.6795336954421523e-05, + "loss": 1.0105, + "step": 15036 + }, + { + "epoch": 0.97, + "learning_rate": 1.679512226539857e-05, + "loss": 1.0617, + "step": 15037 + }, + { + "epoch": 0.97, + "learning_rate": 1.679490757637562e-05, + "loss": 1.0744, + "step": 15038 + }, + { + "epoch": 0.97, + "learning_rate": 1.679469288735267e-05, + "loss": 1.0838, + "step": 15039 + }, + { + "epoch": 0.97, + "learning_rate": 1.679447819832972e-05, + "loss": 1.0182, + "step": 15040 + }, + { + "epoch": 0.97, + "learning_rate": 1.679426350930677e-05, + "loss": 1.0941, + "step": 15041 + }, + { + "epoch": 0.97, + "learning_rate": 1.679404882028382e-05, + "loss": 1.116, + "step": 15042 + }, + { + "epoch": 0.97, + "learning_rate": 1.6793834131260872e-05, + "loss": 1.1037, + "step": 15043 + }, + { + "epoch": 0.97, + "learning_rate": 1.679361944223792e-05, + "loss": 1.0062, + "step": 15044 + }, + { + "epoch": 0.97, + "learning_rate": 1.679340475321497e-05, + "loss": 1.1327, + "step": 15045 + }, + { + "epoch": 0.97, + "learning_rate": 1.679319006419202e-05, + "loss": 1.0235, + "step": 15046 + }, + { + "epoch": 0.97, + "learning_rate": 1.6792975375169068e-05, + "loss": 1.0761, + "step": 15047 + }, + { + "epoch": 0.97, + "learning_rate": 1.679276068614612e-05, + "loss": 1.0607, + "step": 15048 + }, + { + "epoch": 0.97, + "learning_rate": 1.679254599712317e-05, + "loss": 1.1105, + "step": 15049 + }, + { + "epoch": 0.97, + "learning_rate": 1.6792331308100218e-05, + "loss": 1.0999, + "step": 15050 + }, + { + "epoch": 0.97, + "learning_rate": 1.6792116619077267e-05, + "loss": 1.048, + "step": 15051 + }, + { + "epoch": 0.97, + "learning_rate": 1.6791901930054316e-05, + "loss": 1.1129, + "step": 15052 + }, + { + "epoch": 0.97, + "learning_rate": 1.679168724103137e-05, + "loss": 1.0579, + "step": 15053 + }, + { + "epoch": 0.97, + "learning_rate": 1.6791472552008417e-05, + "loss": 1.0387, + "step": 15054 + }, + { + "epoch": 0.97, + "learning_rate": 1.6791257862985466e-05, + "loss": 1.1217, + "step": 15055 + }, + { + "epoch": 0.97, + "learning_rate": 1.6791043173962515e-05, + "loss": 1.0856, + "step": 15056 + }, + { + "epoch": 0.97, + "learning_rate": 1.6790828484939568e-05, + "loss": 1.0108, + "step": 15057 + }, + { + "epoch": 0.97, + "learning_rate": 1.6790613795916617e-05, + "loss": 1.0761, + "step": 15058 + }, + { + "epoch": 0.97, + "learning_rate": 1.6790399106893666e-05, + "loss": 1.1111, + "step": 15059 + }, + { + "epoch": 0.97, + "learning_rate": 1.6790184417870715e-05, + "loss": 0.9724, + "step": 15060 + }, + { + "epoch": 0.97, + "learning_rate": 1.6789969728847767e-05, + "loss": 1.0093, + "step": 15061 + }, + { + "epoch": 0.97, + "learning_rate": 1.6789755039824816e-05, + "loss": 1.1208, + "step": 15062 + }, + { + "epoch": 0.97, + "learning_rate": 1.6789540350801865e-05, + "loss": 1.0842, + "step": 15063 + }, + { + "epoch": 0.97, + "learning_rate": 1.6789325661778914e-05, + "loss": 1.0947, + "step": 15064 + }, + { + "epoch": 0.97, + "learning_rate": 1.6789110972755963e-05, + "loss": 1.0432, + "step": 15065 + }, + { + "epoch": 0.97, + "learning_rate": 1.6788896283733015e-05, + "loss": 1.0297, + "step": 15066 + }, + { + "epoch": 0.97, + "learning_rate": 1.6788681594710064e-05, + "loss": 0.9927, + "step": 15067 + }, + { + "epoch": 0.97, + "learning_rate": 1.6788466905687113e-05, + "loss": 1.0503, + "step": 15068 + }, + { + "epoch": 0.97, + "learning_rate": 1.6788252216664162e-05, + "loss": 1.1837, + "step": 15069 + }, + { + "epoch": 0.97, + "learning_rate": 1.678803752764121e-05, + "loss": 1.0978, + "step": 15070 + }, + { + "epoch": 0.97, + "learning_rate": 1.6787822838618263e-05, + "loss": 1.0557, + "step": 15071 + }, + { + "epoch": 0.97, + "learning_rate": 1.6787608149595312e-05, + "loss": 1.0694, + "step": 15072 + }, + { + "epoch": 0.97, + "learning_rate": 1.6787393460572365e-05, + "loss": 1.0997, + "step": 15073 + }, + { + "epoch": 0.97, + "learning_rate": 1.6787178771549414e-05, + "loss": 1.1026, + "step": 15074 + }, + { + "epoch": 0.97, + "learning_rate": 1.6786964082526463e-05, + "loss": 1.127, + "step": 15075 + }, + { + "epoch": 0.97, + "learning_rate": 1.678674939350351e-05, + "loss": 0.9615, + "step": 15076 + }, + { + "epoch": 0.97, + "learning_rate": 1.678653470448056e-05, + "loss": 1.1278, + "step": 15077 + }, + { + "epoch": 0.97, + "learning_rate": 1.678632001545761e-05, + "loss": 0.9921, + "step": 15078 + }, + { + "epoch": 0.97, + "learning_rate": 1.6786105326434662e-05, + "loss": 1.0016, + "step": 15079 + }, + { + "epoch": 0.97, + "learning_rate": 1.678589063741171e-05, + "loss": 1.0483, + "step": 15080 + }, + { + "epoch": 0.97, + "learning_rate": 1.678567594838876e-05, + "loss": 1.0867, + "step": 15081 + }, + { + "epoch": 0.97, + "learning_rate": 1.678546125936581e-05, + "loss": 1.0287, + "step": 15082 + }, + { + "epoch": 0.97, + "learning_rate": 1.6785246570342858e-05, + "loss": 0.9929, + "step": 15083 + }, + { + "epoch": 0.97, + "learning_rate": 1.678503188131991e-05, + "loss": 0.9805, + "step": 15084 + }, + { + "epoch": 0.97, + "learning_rate": 1.678481719229696e-05, + "loss": 0.9676, + "step": 15085 + }, + { + "epoch": 0.97, + "learning_rate": 1.6784602503274008e-05, + "loss": 1.0763, + "step": 15086 + }, + { + "epoch": 0.97, + "learning_rate": 1.678438781425106e-05, + "loss": 1.1481, + "step": 15087 + }, + { + "epoch": 0.97, + "learning_rate": 1.678417312522811e-05, + "loss": 1.0574, + "step": 15088 + }, + { + "epoch": 0.97, + "learning_rate": 1.678395843620516e-05, + "loss": 1.0478, + "step": 15089 + }, + { + "epoch": 0.97, + "learning_rate": 1.6783743747182207e-05, + "loss": 1.0229, + "step": 15090 + }, + { + "epoch": 0.97, + "learning_rate": 1.678352905815926e-05, + "loss": 1.0128, + "step": 15091 + }, + { + "epoch": 0.97, + "learning_rate": 1.678331436913631e-05, + "loss": 1.0672, + "step": 15092 + }, + { + "epoch": 0.97, + "learning_rate": 1.6783099680113358e-05, + "loss": 1.1408, + "step": 15093 + }, + { + "epoch": 0.97, + "learning_rate": 1.6782884991090407e-05, + "loss": 1.0444, + "step": 15094 + }, + { + "epoch": 0.97, + "learning_rate": 1.6782670302067455e-05, + "loss": 0.9871, + "step": 15095 + }, + { + "epoch": 0.97, + "learning_rate": 1.6782455613044504e-05, + "loss": 1.1096, + "step": 15096 + }, + { + "epoch": 0.97, + "learning_rate": 1.6782240924021557e-05, + "loss": 1.053, + "step": 15097 + }, + { + "epoch": 0.97, + "learning_rate": 1.6782026234998606e-05, + "loss": 1.0179, + "step": 15098 + }, + { + "epoch": 0.97, + "learning_rate": 1.6781811545975655e-05, + "loss": 1.0868, + "step": 15099 + }, + { + "epoch": 0.97, + "learning_rate": 1.6781596856952704e-05, + "loss": 1.0337, + "step": 15100 + }, + { + "epoch": 0.97, + "learning_rate": 1.6781382167929756e-05, + "loss": 1.0535, + "step": 15101 + }, + { + "epoch": 0.97, + "learning_rate": 1.6781167478906805e-05, + "loss": 1.0597, + "step": 15102 + }, + { + "epoch": 0.97, + "learning_rate": 1.6780952789883854e-05, + "loss": 1.0056, + "step": 15103 + }, + { + "epoch": 0.97, + "learning_rate": 1.6780738100860906e-05, + "loss": 1.0507, + "step": 15104 + }, + { + "epoch": 0.97, + "learning_rate": 1.6780523411837955e-05, + "loss": 1.0847, + "step": 15105 + }, + { + "epoch": 0.97, + "learning_rate": 1.6780308722815004e-05, + "loss": 1.0092, + "step": 15106 + }, + { + "epoch": 0.97, + "learning_rate": 1.6780094033792053e-05, + "loss": 1.1222, + "step": 15107 + }, + { + "epoch": 0.97, + "learning_rate": 1.6779879344769102e-05, + "loss": 0.9804, + "step": 15108 + }, + { + "epoch": 0.97, + "learning_rate": 1.6779664655746155e-05, + "loss": 0.9912, + "step": 15109 + }, + { + "epoch": 0.97, + "learning_rate": 1.6779449966723204e-05, + "loss": 0.973, + "step": 15110 + }, + { + "epoch": 0.97, + "learning_rate": 1.6779235277700252e-05, + "loss": 1.0687, + "step": 15111 + }, + { + "epoch": 0.97, + "learning_rate": 1.67790205886773e-05, + "loss": 1.0846, + "step": 15112 + }, + { + "epoch": 0.97, + "learning_rate": 1.677880589965435e-05, + "loss": 0.9859, + "step": 15113 + }, + { + "epoch": 0.97, + "learning_rate": 1.67785912106314e-05, + "loss": 1.0735, + "step": 15114 + }, + { + "epoch": 0.97, + "learning_rate": 1.6778376521608452e-05, + "loss": 1.016, + "step": 15115 + }, + { + "epoch": 0.97, + "learning_rate": 1.67781618325855e-05, + "loss": 1.1193, + "step": 15116 + }, + { + "epoch": 0.97, + "learning_rate": 1.6777947143562553e-05, + "loss": 1.0786, + "step": 15117 + }, + { + "epoch": 0.97, + "learning_rate": 1.6777732454539602e-05, + "loss": 1.0033, + "step": 15118 + }, + { + "epoch": 0.97, + "learning_rate": 1.677751776551665e-05, + "loss": 1.0569, + "step": 15119 + }, + { + "epoch": 0.97, + "learning_rate": 1.67773030764937e-05, + "loss": 1.0508, + "step": 15120 + }, + { + "epoch": 0.97, + "learning_rate": 1.677708838747075e-05, + "loss": 0.9695, + "step": 15121 + }, + { + "epoch": 0.97, + "learning_rate": 1.67768736984478e-05, + "loss": 1.105, + "step": 15122 + }, + { + "epoch": 0.97, + "learning_rate": 1.677665900942485e-05, + "loss": 1.0462, + "step": 15123 + }, + { + "epoch": 0.97, + "learning_rate": 1.67764443204019e-05, + "loss": 1.1098, + "step": 15124 + }, + { + "epoch": 0.97, + "learning_rate": 1.6776229631378948e-05, + "loss": 1.0487, + "step": 15125 + }, + { + "epoch": 0.97, + "learning_rate": 1.6776014942355997e-05, + "loss": 1.0796, + "step": 15126 + }, + { + "epoch": 0.97, + "learning_rate": 1.6775800253333046e-05, + "loss": 1.0575, + "step": 15127 + }, + { + "epoch": 0.97, + "learning_rate": 1.67755855643101e-05, + "loss": 1.0549, + "step": 15128 + }, + { + "epoch": 0.97, + "learning_rate": 1.6775370875287147e-05, + "loss": 1.064, + "step": 15129 + }, + { + "epoch": 0.97, + "learning_rate": 1.67751561862642e-05, + "loss": 1.0386, + "step": 15130 + }, + { + "epoch": 0.97, + "learning_rate": 1.677494149724125e-05, + "loss": 1.018, + "step": 15131 + }, + { + "epoch": 0.97, + "learning_rate": 1.6774726808218298e-05, + "loss": 1.0814, + "step": 15132 + }, + { + "epoch": 0.97, + "learning_rate": 1.6774512119195347e-05, + "loss": 1.0928, + "step": 15133 + }, + { + "epoch": 0.97, + "learning_rate": 1.6774297430172396e-05, + "loss": 1.0426, + "step": 15134 + }, + { + "epoch": 0.97, + "learning_rate": 1.6774082741149448e-05, + "loss": 1.0849, + "step": 15135 + }, + { + "epoch": 0.97, + "learning_rate": 1.6773868052126497e-05, + "loss": 1.0723, + "step": 15136 + }, + { + "epoch": 0.97, + "learning_rate": 1.6773653363103546e-05, + "loss": 0.9981, + "step": 15137 + }, + { + "epoch": 0.97, + "learning_rate": 1.6773438674080595e-05, + "loss": 1.0571, + "step": 15138 + }, + { + "epoch": 0.97, + "learning_rate": 1.6773223985057644e-05, + "loss": 1.1251, + "step": 15139 + }, + { + "epoch": 0.97, + "learning_rate": 1.6773009296034696e-05, + "loss": 1.1537, + "step": 15140 + }, + { + "epoch": 0.97, + "learning_rate": 1.6772794607011745e-05, + "loss": 1.0305, + "step": 15141 + }, + { + "epoch": 0.97, + "learning_rate": 1.6772579917988794e-05, + "loss": 1.0568, + "step": 15142 + }, + { + "epoch": 0.97, + "learning_rate": 1.6772365228965843e-05, + "loss": 1.0482, + "step": 15143 + }, + { + "epoch": 0.97, + "learning_rate": 1.6772150539942892e-05, + "loss": 1.0062, + "step": 15144 + }, + { + "epoch": 0.97, + "learning_rate": 1.6771935850919944e-05, + "loss": 1.0977, + "step": 15145 + }, + { + "epoch": 0.97, + "learning_rate": 1.6771721161896993e-05, + "loss": 1.0876, + "step": 15146 + }, + { + "epoch": 0.97, + "learning_rate": 1.6771506472874042e-05, + "loss": 1.0048, + "step": 15147 + }, + { + "epoch": 0.97, + "learning_rate": 1.6771291783851095e-05, + "loss": 1.0062, + "step": 15148 + }, + { + "epoch": 0.97, + "learning_rate": 1.6771077094828144e-05, + "loss": 1.0798, + "step": 15149 + }, + { + "epoch": 0.97, + "learning_rate": 1.6770862405805193e-05, + "loss": 1.0498, + "step": 15150 + }, + { + "epoch": 0.97, + "learning_rate": 1.677064771678224e-05, + "loss": 0.9856, + "step": 15151 + }, + { + "epoch": 0.97, + "learning_rate": 1.677043302775929e-05, + "loss": 1.0533, + "step": 15152 + }, + { + "epoch": 0.97, + "learning_rate": 1.6770218338736343e-05, + "loss": 1.0863, + "step": 15153 + }, + { + "epoch": 0.97, + "learning_rate": 1.6770003649713392e-05, + "loss": 0.9981, + "step": 15154 + }, + { + "epoch": 0.98, + "learning_rate": 1.676978896069044e-05, + "loss": 1.0737, + "step": 15155 + }, + { + "epoch": 0.98, + "learning_rate": 1.676957427166749e-05, + "loss": 1.0408, + "step": 15156 + }, + { + "epoch": 0.98, + "learning_rate": 1.676935958264454e-05, + "loss": 1.0769, + "step": 15157 + }, + { + "epoch": 0.98, + "learning_rate": 1.676914489362159e-05, + "loss": 1.108, + "step": 15158 + }, + { + "epoch": 0.98, + "learning_rate": 1.676893020459864e-05, + "loss": 0.9783, + "step": 15159 + }, + { + "epoch": 0.98, + "learning_rate": 1.6768715515575692e-05, + "loss": 1.056, + "step": 15160 + }, + { + "epoch": 0.98, + "learning_rate": 1.676850082655274e-05, + "loss": 1.0314, + "step": 15161 + }, + { + "epoch": 0.98, + "learning_rate": 1.676828613752979e-05, + "loss": 1.0342, + "step": 15162 + }, + { + "epoch": 0.98, + "learning_rate": 1.676807144850684e-05, + "loss": 1.0291, + "step": 15163 + }, + { + "epoch": 0.98, + "learning_rate": 1.6767856759483888e-05, + "loss": 1.0865, + "step": 15164 + }, + { + "epoch": 0.98, + "learning_rate": 1.6767642070460937e-05, + "loss": 1.0354, + "step": 15165 + }, + { + "epoch": 0.98, + "learning_rate": 1.676742738143799e-05, + "loss": 1.1002, + "step": 15166 + }, + { + "epoch": 0.98, + "learning_rate": 1.676721269241504e-05, + "loss": 1.1115, + "step": 15167 + }, + { + "epoch": 0.98, + "learning_rate": 1.6766998003392088e-05, + "loss": 1.0474, + "step": 15168 + }, + { + "epoch": 0.98, + "learning_rate": 1.6766783314369136e-05, + "loss": 0.989, + "step": 15169 + }, + { + "epoch": 0.98, + "learning_rate": 1.6766568625346185e-05, + "loss": 0.9475, + "step": 15170 + }, + { + "epoch": 0.98, + "learning_rate": 1.6766353936323238e-05, + "loss": 1.0898, + "step": 15171 + }, + { + "epoch": 0.98, + "learning_rate": 1.6766139247300287e-05, + "loss": 1.1264, + "step": 15172 + }, + { + "epoch": 0.98, + "learning_rate": 1.6765924558277336e-05, + "loss": 1.0794, + "step": 15173 + }, + { + "epoch": 0.98, + "learning_rate": 1.6765709869254388e-05, + "loss": 1.0857, + "step": 15174 + }, + { + "epoch": 0.98, + "learning_rate": 1.6765495180231437e-05, + "loss": 1.0411, + "step": 15175 + }, + { + "epoch": 0.98, + "learning_rate": 1.6765280491208486e-05, + "loss": 0.9912, + "step": 15176 + }, + { + "epoch": 0.98, + "learning_rate": 1.6765065802185535e-05, + "loss": 1.1423, + "step": 15177 + }, + { + "epoch": 0.98, + "learning_rate": 1.6764851113162587e-05, + "loss": 1.0272, + "step": 15178 + }, + { + "epoch": 0.98, + "learning_rate": 1.6764636424139636e-05, + "loss": 1.0807, + "step": 15179 + }, + { + "epoch": 0.98, + "learning_rate": 1.6764421735116685e-05, + "loss": 1.0605, + "step": 15180 + }, + { + "epoch": 0.98, + "learning_rate": 1.6764207046093734e-05, + "loss": 1.0118, + "step": 15181 + }, + { + "epoch": 0.98, + "learning_rate": 1.6763992357070783e-05, + "loss": 1.0042, + "step": 15182 + }, + { + "epoch": 0.98, + "learning_rate": 1.6763777668047832e-05, + "loss": 1.0553, + "step": 15183 + }, + { + "epoch": 0.98, + "learning_rate": 1.6763562979024885e-05, + "loss": 1.144, + "step": 15184 + }, + { + "epoch": 0.98, + "learning_rate": 1.6763348290001933e-05, + "loss": 1.1161, + "step": 15185 + }, + { + "epoch": 0.98, + "learning_rate": 1.6763133600978982e-05, + "loss": 1.0198, + "step": 15186 + }, + { + "epoch": 0.98, + "learning_rate": 1.676291891195603e-05, + "loss": 1.0122, + "step": 15187 + }, + { + "epoch": 0.98, + "learning_rate": 1.6762704222933084e-05, + "loss": 1.0833, + "step": 15188 + }, + { + "epoch": 0.98, + "learning_rate": 1.6762489533910133e-05, + "loss": 1.1318, + "step": 15189 + }, + { + "epoch": 0.98, + "learning_rate": 1.6762274844887182e-05, + "loss": 1.0115, + "step": 15190 + }, + { + "epoch": 0.98, + "learning_rate": 1.6762060155864234e-05, + "loss": 1.0999, + "step": 15191 + }, + { + "epoch": 0.98, + "learning_rate": 1.6761845466841283e-05, + "loss": 1.0482, + "step": 15192 + }, + { + "epoch": 0.98, + "learning_rate": 1.6761630777818332e-05, + "loss": 1.088, + "step": 15193 + }, + { + "epoch": 0.98, + "learning_rate": 1.676141608879538e-05, + "loss": 1.1078, + "step": 15194 + }, + { + "epoch": 0.98, + "learning_rate": 1.676120139977243e-05, + "loss": 1.1025, + "step": 15195 + }, + { + "epoch": 0.98, + "learning_rate": 1.6760986710749482e-05, + "loss": 1.0614, + "step": 15196 + }, + { + "epoch": 0.98, + "learning_rate": 1.676077202172653e-05, + "loss": 1.1562, + "step": 15197 + }, + { + "epoch": 0.98, + "learning_rate": 1.676055733270358e-05, + "loss": 1.0855, + "step": 15198 + }, + { + "epoch": 0.98, + "learning_rate": 1.676034264368063e-05, + "loss": 0.9782, + "step": 15199 + }, + { + "epoch": 0.98, + "learning_rate": 1.6760127954657678e-05, + "loss": 1.0748, + "step": 15200 + }, + { + "epoch": 0.98, + "learning_rate": 1.6759913265634727e-05, + "loss": 1.0732, + "step": 15201 + }, + { + "epoch": 0.98, + "learning_rate": 1.675969857661178e-05, + "loss": 1.1264, + "step": 15202 + }, + { + "epoch": 0.98, + "learning_rate": 1.675948388758883e-05, + "loss": 1.0052, + "step": 15203 + }, + { + "epoch": 0.98, + "learning_rate": 1.675926919856588e-05, + "loss": 1.0815, + "step": 15204 + }, + { + "epoch": 0.98, + "learning_rate": 1.675905450954293e-05, + "loss": 1.1118, + "step": 15205 + }, + { + "epoch": 0.98, + "learning_rate": 1.675883982051998e-05, + "loss": 1.0183, + "step": 15206 + }, + { + "epoch": 0.98, + "learning_rate": 1.6758625131497028e-05, + "loss": 0.9992, + "step": 15207 + }, + { + "epoch": 0.98, + "learning_rate": 1.6758410442474077e-05, + "loss": 1.0576, + "step": 15208 + }, + { + "epoch": 0.98, + "learning_rate": 1.675819575345113e-05, + "loss": 1.0042, + "step": 15209 + }, + { + "epoch": 0.98, + "learning_rate": 1.6757981064428178e-05, + "loss": 1.1331, + "step": 15210 + }, + { + "epoch": 0.98, + "learning_rate": 1.6757766375405227e-05, + "loss": 1.0747, + "step": 15211 + }, + { + "epoch": 0.98, + "learning_rate": 1.6757551686382276e-05, + "loss": 1.0477, + "step": 15212 + }, + { + "epoch": 0.98, + "learning_rate": 1.6757336997359325e-05, + "loss": 1.0961, + "step": 15213 + }, + { + "epoch": 0.98, + "learning_rate": 1.6757122308336374e-05, + "loss": 1.0939, + "step": 15214 + }, + { + "epoch": 0.98, + "learning_rate": 1.6756907619313426e-05, + "loss": 1.0228, + "step": 15215 + }, + { + "epoch": 0.98, + "learning_rate": 1.6756692930290475e-05, + "loss": 0.9222, + "step": 15216 + }, + { + "epoch": 0.98, + "learning_rate": 1.6756478241267527e-05, + "loss": 1.0868, + "step": 15217 + }, + { + "epoch": 0.98, + "learning_rate": 1.6756263552244576e-05, + "loss": 0.981, + "step": 15218 + }, + { + "epoch": 0.98, + "learning_rate": 1.6756048863221625e-05, + "loss": 1.0903, + "step": 15219 + }, + { + "epoch": 0.98, + "learning_rate": 1.6755834174198674e-05, + "loss": 1.0, + "step": 15220 + }, + { + "epoch": 0.98, + "learning_rate": 1.6755619485175723e-05, + "loss": 1.0794, + "step": 15221 + }, + { + "epoch": 0.98, + "learning_rate": 1.6755404796152776e-05, + "loss": 1.0118, + "step": 15222 + }, + { + "epoch": 0.98, + "learning_rate": 1.6755190107129825e-05, + "loss": 1.0778, + "step": 15223 + }, + { + "epoch": 0.98, + "learning_rate": 1.6754975418106874e-05, + "loss": 1.1086, + "step": 15224 + }, + { + "epoch": 0.98, + "learning_rate": 1.6754760729083923e-05, + "loss": 1.0598, + "step": 15225 + }, + { + "epoch": 0.98, + "learning_rate": 1.675454604006097e-05, + "loss": 1.039, + "step": 15226 + }, + { + "epoch": 0.98, + "learning_rate": 1.6754331351038024e-05, + "loss": 1.001, + "step": 15227 + }, + { + "epoch": 0.98, + "learning_rate": 1.6754116662015073e-05, + "loss": 1.0926, + "step": 15228 + }, + { + "epoch": 0.98, + "learning_rate": 1.6753901972992122e-05, + "loss": 1.1184, + "step": 15229 + }, + { + "epoch": 0.98, + "learning_rate": 1.675368728396917e-05, + "loss": 1.0818, + "step": 15230 + }, + { + "epoch": 0.98, + "learning_rate": 1.675347259494622e-05, + "loss": 1.1106, + "step": 15231 + }, + { + "epoch": 0.98, + "learning_rate": 1.6753257905923272e-05, + "loss": 1.0679, + "step": 15232 + }, + { + "epoch": 0.98, + "learning_rate": 1.675304321690032e-05, + "loss": 1.0954, + "step": 15233 + }, + { + "epoch": 0.98, + "learning_rate": 1.675282852787737e-05, + "loss": 0.9667, + "step": 15234 + }, + { + "epoch": 0.98, + "learning_rate": 1.6752613838854422e-05, + "loss": 1.1306, + "step": 15235 + }, + { + "epoch": 0.98, + "learning_rate": 1.675239914983147e-05, + "loss": 1.11, + "step": 15236 + }, + { + "epoch": 0.98, + "learning_rate": 1.675218446080852e-05, + "loss": 1.0234, + "step": 15237 + }, + { + "epoch": 0.98, + "learning_rate": 1.675196977178557e-05, + "loss": 0.9912, + "step": 15238 + }, + { + "epoch": 0.98, + "learning_rate": 1.6751755082762618e-05, + "loss": 0.9807, + "step": 15239 + }, + { + "epoch": 0.98, + "learning_rate": 1.675154039373967e-05, + "loss": 1.0066, + "step": 15240 + }, + { + "epoch": 0.98, + "learning_rate": 1.675132570471672e-05, + "loss": 1.0713, + "step": 15241 + }, + { + "epoch": 0.98, + "learning_rate": 1.675111101569377e-05, + "loss": 1.0546, + "step": 15242 + }, + { + "epoch": 0.98, + "learning_rate": 1.6750896326670817e-05, + "loss": 1.0504, + "step": 15243 + }, + { + "epoch": 0.98, + "learning_rate": 1.6750681637647866e-05, + "loss": 1.0287, + "step": 15244 + }, + { + "epoch": 0.98, + "learning_rate": 1.675046694862492e-05, + "loss": 0.9849, + "step": 15245 + }, + { + "epoch": 0.98, + "learning_rate": 1.6750252259601968e-05, + "loss": 1.0378, + "step": 15246 + }, + { + "epoch": 0.98, + "learning_rate": 1.675003757057902e-05, + "loss": 1.0383, + "step": 15247 + }, + { + "epoch": 0.98, + "learning_rate": 1.674982288155607e-05, + "loss": 1.1175, + "step": 15248 + }, + { + "epoch": 0.98, + "learning_rate": 1.6749608192533118e-05, + "loss": 0.9885, + "step": 15249 + }, + { + "epoch": 0.98, + "learning_rate": 1.6749393503510167e-05, + "loss": 1.0526, + "step": 15250 + }, + { + "epoch": 0.98, + "learning_rate": 1.6749178814487216e-05, + "loss": 0.9997, + "step": 15251 + }, + { + "epoch": 0.98, + "learning_rate": 1.6748964125464265e-05, + "loss": 1.1584, + "step": 15252 + }, + { + "epoch": 0.98, + "learning_rate": 1.6748749436441317e-05, + "loss": 1.1057, + "step": 15253 + }, + { + "epoch": 0.98, + "learning_rate": 1.6748534747418366e-05, + "loss": 0.9586, + "step": 15254 + }, + { + "epoch": 0.98, + "learning_rate": 1.6748320058395415e-05, + "loss": 1.0695, + "step": 15255 + }, + { + "epoch": 0.98, + "learning_rate": 1.6748105369372464e-05, + "loss": 1.081, + "step": 15256 + }, + { + "epoch": 0.98, + "learning_rate": 1.6747890680349513e-05, + "loss": 1.0775, + "step": 15257 + }, + { + "epoch": 0.98, + "learning_rate": 1.6747675991326566e-05, + "loss": 1.0788, + "step": 15258 + }, + { + "epoch": 0.98, + "learning_rate": 1.6747461302303614e-05, + "loss": 1.0269, + "step": 15259 + }, + { + "epoch": 0.98, + "learning_rate": 1.6747246613280663e-05, + "loss": 1.0371, + "step": 15260 + }, + { + "epoch": 0.98, + "learning_rate": 1.6747031924257716e-05, + "loss": 0.9695, + "step": 15261 + }, + { + "epoch": 0.98, + "learning_rate": 1.6746817235234765e-05, + "loss": 0.9842, + "step": 15262 + }, + { + "epoch": 0.98, + "learning_rate": 1.6746602546211814e-05, + "loss": 1.064, + "step": 15263 + }, + { + "epoch": 0.98, + "learning_rate": 1.6746387857188863e-05, + "loss": 0.9899, + "step": 15264 + }, + { + "epoch": 0.98, + "learning_rate": 1.6746173168165915e-05, + "loss": 1.0098, + "step": 15265 + }, + { + "epoch": 0.98, + "learning_rate": 1.6745958479142964e-05, + "loss": 1.0335, + "step": 15266 + }, + { + "epoch": 0.98, + "learning_rate": 1.6745743790120013e-05, + "loss": 1.2023, + "step": 15267 + }, + { + "epoch": 0.98, + "learning_rate": 1.6745529101097062e-05, + "loss": 1.0102, + "step": 15268 + }, + { + "epoch": 0.98, + "learning_rate": 1.674531441207411e-05, + "loss": 1.0324, + "step": 15269 + }, + { + "epoch": 0.98, + "learning_rate": 1.674509972305116e-05, + "loss": 1.0499, + "step": 15270 + }, + { + "epoch": 0.98, + "learning_rate": 1.6744885034028212e-05, + "loss": 1.0281, + "step": 15271 + }, + { + "epoch": 0.98, + "learning_rate": 1.674467034500526e-05, + "loss": 1.0478, + "step": 15272 + }, + { + "epoch": 0.98, + "learning_rate": 1.674445565598231e-05, + "loss": 0.8993, + "step": 15273 + }, + { + "epoch": 0.98, + "learning_rate": 1.674424096695936e-05, + "loss": 1.1118, + "step": 15274 + }, + { + "epoch": 0.98, + "learning_rate": 1.674402627793641e-05, + "loss": 1.0597, + "step": 15275 + }, + { + "epoch": 0.98, + "learning_rate": 1.674381158891346e-05, + "loss": 1.0657, + "step": 15276 + }, + { + "epoch": 0.98, + "learning_rate": 1.674359689989051e-05, + "loss": 1.0749, + "step": 15277 + }, + { + "epoch": 0.98, + "learning_rate": 1.6743382210867562e-05, + "loss": 1.071, + "step": 15278 + }, + { + "epoch": 0.98, + "learning_rate": 1.674316752184461e-05, + "loss": 1.0069, + "step": 15279 + }, + { + "epoch": 0.98, + "learning_rate": 1.674295283282166e-05, + "loss": 0.9646, + "step": 15280 + }, + { + "epoch": 0.98, + "learning_rate": 1.674273814379871e-05, + "loss": 0.9896, + "step": 15281 + }, + { + "epoch": 0.98, + "learning_rate": 1.6742523454775758e-05, + "loss": 1.006, + "step": 15282 + }, + { + "epoch": 0.98, + "learning_rate": 1.6742308765752807e-05, + "loss": 1.0515, + "step": 15283 + }, + { + "epoch": 0.98, + "learning_rate": 1.674209407672986e-05, + "loss": 1.048, + "step": 15284 + }, + { + "epoch": 0.98, + "learning_rate": 1.6741879387706908e-05, + "loss": 1.0794, + "step": 15285 + }, + { + "epoch": 0.98, + "learning_rate": 1.6741664698683957e-05, + "loss": 1.0138, + "step": 15286 + }, + { + "epoch": 0.98, + "learning_rate": 1.6741450009661006e-05, + "loss": 1.0425, + "step": 15287 + }, + { + "epoch": 0.98, + "learning_rate": 1.6741235320638055e-05, + "loss": 1.004, + "step": 15288 + }, + { + "epoch": 0.98, + "learning_rate": 1.6741020631615107e-05, + "loss": 1.1398, + "step": 15289 + }, + { + "epoch": 0.98, + "learning_rate": 1.6740805942592156e-05, + "loss": 1.0828, + "step": 15290 + }, + { + "epoch": 0.98, + "learning_rate": 1.674059125356921e-05, + "loss": 1.0423, + "step": 15291 + }, + { + "epoch": 0.98, + "learning_rate": 1.6740376564546257e-05, + "loss": 1.0913, + "step": 15292 + }, + { + "epoch": 0.98, + "learning_rate": 1.6740161875523306e-05, + "loss": 0.9452, + "step": 15293 + }, + { + "epoch": 0.98, + "learning_rate": 1.6739947186500355e-05, + "loss": 1.0415, + "step": 15294 + }, + { + "epoch": 0.98, + "learning_rate": 1.6739732497477404e-05, + "loss": 1.0834, + "step": 15295 + }, + { + "epoch": 0.98, + "learning_rate": 1.6739517808454457e-05, + "loss": 1.1061, + "step": 15296 + }, + { + "epoch": 0.98, + "learning_rate": 1.6739303119431506e-05, + "loss": 1.0478, + "step": 15297 + }, + { + "epoch": 0.98, + "learning_rate": 1.6739088430408555e-05, + "loss": 1.0563, + "step": 15298 + }, + { + "epoch": 0.98, + "learning_rate": 1.6738873741385604e-05, + "loss": 1.1206, + "step": 15299 + }, + { + "epoch": 0.98, + "learning_rate": 1.6738659052362653e-05, + "loss": 1.0861, + "step": 15300 + }, + { + "epoch": 0.98, + "learning_rate": 1.67384443633397e-05, + "loss": 1.0545, + "step": 15301 + }, + { + "epoch": 0.98, + "learning_rate": 1.6738229674316754e-05, + "loss": 1.0803, + "step": 15302 + }, + { + "epoch": 0.98, + "learning_rate": 1.6738014985293803e-05, + "loss": 1.0527, + "step": 15303 + }, + { + "epoch": 0.98, + "learning_rate": 1.6737800296270855e-05, + "loss": 1.0886, + "step": 15304 + }, + { + "epoch": 0.98, + "learning_rate": 1.6737585607247904e-05, + "loss": 1.153, + "step": 15305 + }, + { + "epoch": 0.98, + "learning_rate": 1.6737370918224953e-05, + "loss": 1.1271, + "step": 15306 + }, + { + "epoch": 0.98, + "learning_rate": 1.6737156229202002e-05, + "loss": 0.9971, + "step": 15307 + }, + { + "epoch": 0.98, + "learning_rate": 1.673694154017905e-05, + "loss": 0.9866, + "step": 15308 + }, + { + "epoch": 0.98, + "learning_rate": 1.6736726851156103e-05, + "loss": 1.0566, + "step": 15309 + }, + { + "epoch": 0.99, + "learning_rate": 1.6736512162133152e-05, + "loss": 1.0942, + "step": 15310 + }, + { + "epoch": 0.99, + "learning_rate": 1.67362974731102e-05, + "loss": 1.0342, + "step": 15311 + }, + { + "epoch": 0.99, + "learning_rate": 1.673608278408725e-05, + "loss": 1.1077, + "step": 15312 + }, + { + "epoch": 0.99, + "learning_rate": 1.67358680950643e-05, + "loss": 1.0144, + "step": 15313 + }, + { + "epoch": 0.99, + "learning_rate": 1.673565340604135e-05, + "loss": 1.0815, + "step": 15314 + }, + { + "epoch": 0.99, + "learning_rate": 1.67354387170184e-05, + "loss": 1.0345, + "step": 15315 + }, + { + "epoch": 0.99, + "learning_rate": 1.673522402799545e-05, + "loss": 1.0566, + "step": 15316 + }, + { + "epoch": 0.99, + "learning_rate": 1.67350093389725e-05, + "loss": 1.0379, + "step": 15317 + }, + { + "epoch": 0.99, + "learning_rate": 1.6734794649949547e-05, + "loss": 1.1218, + "step": 15318 + }, + { + "epoch": 0.99, + "learning_rate": 1.67345799609266e-05, + "loss": 1.0323, + "step": 15319 + }, + { + "epoch": 0.99, + "learning_rate": 1.673436527190365e-05, + "loss": 1.033, + "step": 15320 + }, + { + "epoch": 0.99, + "learning_rate": 1.6734150582880698e-05, + "loss": 1.1373, + "step": 15321 + }, + { + "epoch": 0.99, + "learning_rate": 1.673393589385775e-05, + "loss": 1.1788, + "step": 15322 + }, + { + "epoch": 0.99, + "learning_rate": 1.67337212048348e-05, + "loss": 1.0564, + "step": 15323 + }, + { + "epoch": 0.99, + "learning_rate": 1.6733506515811848e-05, + "loss": 1.0915, + "step": 15324 + }, + { + "epoch": 0.99, + "learning_rate": 1.6733291826788897e-05, + "loss": 1.0298, + "step": 15325 + }, + { + "epoch": 0.99, + "learning_rate": 1.6733077137765946e-05, + "loss": 0.9785, + "step": 15326 + }, + { + "epoch": 0.99, + "learning_rate": 1.6732862448743e-05, + "loss": 1.142, + "step": 15327 + }, + { + "epoch": 0.99, + "learning_rate": 1.6732647759720047e-05, + "loss": 1.052, + "step": 15328 + }, + { + "epoch": 0.99, + "learning_rate": 1.6732433070697096e-05, + "loss": 1.0249, + "step": 15329 + }, + { + "epoch": 0.99, + "learning_rate": 1.6732218381674145e-05, + "loss": 0.9926, + "step": 15330 + }, + { + "epoch": 0.99, + "learning_rate": 1.6732003692651194e-05, + "loss": 1.0018, + "step": 15331 + }, + { + "epoch": 0.99, + "learning_rate": 1.6731789003628247e-05, + "loss": 1.0383, + "step": 15332 + }, + { + "epoch": 0.99, + "learning_rate": 1.6731574314605295e-05, + "loss": 1.0498, + "step": 15333 + }, + { + "epoch": 0.99, + "learning_rate": 1.6731359625582348e-05, + "loss": 1.1101, + "step": 15334 + }, + { + "epoch": 0.99, + "learning_rate": 1.6731144936559397e-05, + "loss": 1.0808, + "step": 15335 + }, + { + "epoch": 0.99, + "learning_rate": 1.6730930247536446e-05, + "loss": 1.0763, + "step": 15336 + }, + { + "epoch": 0.99, + "learning_rate": 1.6730715558513495e-05, + "loss": 1.1429, + "step": 15337 + }, + { + "epoch": 0.99, + "learning_rate": 1.6730500869490544e-05, + "loss": 1.0216, + "step": 15338 + }, + { + "epoch": 0.99, + "learning_rate": 1.6730286180467593e-05, + "loss": 1.1744, + "step": 15339 + }, + { + "epoch": 0.99, + "learning_rate": 1.6730071491444645e-05, + "loss": 0.993, + "step": 15340 + }, + { + "epoch": 0.99, + "learning_rate": 1.6729856802421694e-05, + "loss": 1.0903, + "step": 15341 + }, + { + "epoch": 0.99, + "learning_rate": 1.6729642113398743e-05, + "loss": 1.0501, + "step": 15342 + }, + { + "epoch": 0.99, + "learning_rate": 1.6729427424375792e-05, + "loss": 1.0457, + "step": 15343 + }, + { + "epoch": 0.99, + "learning_rate": 1.672921273535284e-05, + "loss": 1.0913, + "step": 15344 + }, + { + "epoch": 0.99, + "learning_rate": 1.6728998046329893e-05, + "loss": 1.0979, + "step": 15345 + }, + { + "epoch": 0.99, + "learning_rate": 1.6728783357306942e-05, + "loss": 1.0676, + "step": 15346 + }, + { + "epoch": 0.99, + "learning_rate": 1.672856866828399e-05, + "loss": 1.0442, + "step": 15347 + }, + { + "epoch": 0.99, + "learning_rate": 1.6728353979261044e-05, + "loss": 1.0347, + "step": 15348 + }, + { + "epoch": 0.99, + "learning_rate": 1.6728139290238092e-05, + "loss": 1.0595, + "step": 15349 + }, + { + "epoch": 0.99, + "learning_rate": 1.672792460121514e-05, + "loss": 1.0576, + "step": 15350 + }, + { + "epoch": 0.99, + "learning_rate": 1.672770991219219e-05, + "loss": 1.035, + "step": 15351 + }, + { + "epoch": 0.99, + "learning_rate": 1.6727495223169243e-05, + "loss": 0.9982, + "step": 15352 + }, + { + "epoch": 0.99, + "learning_rate": 1.6727280534146292e-05, + "loss": 1.0413, + "step": 15353 + }, + { + "epoch": 0.99, + "learning_rate": 1.672706584512334e-05, + "loss": 1.0705, + "step": 15354 + }, + { + "epoch": 0.99, + "learning_rate": 1.672685115610039e-05, + "loss": 1.0827, + "step": 15355 + }, + { + "epoch": 0.99, + "learning_rate": 1.672663646707744e-05, + "loss": 0.9822, + "step": 15356 + }, + { + "epoch": 0.99, + "learning_rate": 1.6726421778054488e-05, + "loss": 1.0462, + "step": 15357 + }, + { + "epoch": 0.99, + "learning_rate": 1.672620708903154e-05, + "loss": 1.0709, + "step": 15358 + }, + { + "epoch": 0.99, + "learning_rate": 1.672599240000859e-05, + "loss": 1.0429, + "step": 15359 + }, + { + "epoch": 0.99, + "learning_rate": 1.6725777710985638e-05, + "loss": 1.0164, + "step": 15360 + }, + { + "epoch": 0.99, + "learning_rate": 1.6725563021962687e-05, + "loss": 1.0182, + "step": 15361 + }, + { + "epoch": 0.99, + "learning_rate": 1.672534833293974e-05, + "loss": 1.141, + "step": 15362 + }, + { + "epoch": 0.99, + "learning_rate": 1.6725133643916788e-05, + "loss": 1.057, + "step": 15363 + }, + { + "epoch": 0.99, + "learning_rate": 1.6724918954893837e-05, + "loss": 1.0576, + "step": 15364 + }, + { + "epoch": 0.99, + "learning_rate": 1.672470426587089e-05, + "loss": 1.037, + "step": 15365 + }, + { + "epoch": 0.99, + "learning_rate": 1.672448957684794e-05, + "loss": 1.1397, + "step": 15366 + }, + { + "epoch": 0.99, + "learning_rate": 1.6724274887824987e-05, + "loss": 1.0406, + "step": 15367 + }, + { + "epoch": 0.99, + "learning_rate": 1.6724060198802036e-05, + "loss": 1.0934, + "step": 15368 + }, + { + "epoch": 0.99, + "learning_rate": 1.6723845509779085e-05, + "loss": 1.0484, + "step": 15369 + }, + { + "epoch": 0.99, + "learning_rate": 1.6723630820756134e-05, + "loss": 0.9919, + "step": 15370 + }, + { + "epoch": 0.99, + "learning_rate": 1.6723416131733187e-05, + "loss": 1.1144, + "step": 15371 + }, + { + "epoch": 0.99, + "learning_rate": 1.6723201442710236e-05, + "loss": 1.0889, + "step": 15372 + }, + { + "epoch": 0.99, + "learning_rate": 1.6722986753687285e-05, + "loss": 1.0303, + "step": 15373 + }, + { + "epoch": 0.99, + "learning_rate": 1.6722772064664334e-05, + "loss": 1.0817, + "step": 15374 + }, + { + "epoch": 0.99, + "learning_rate": 1.6722557375641383e-05, + "loss": 1.0289, + "step": 15375 + }, + { + "epoch": 0.99, + "learning_rate": 1.6722342686618435e-05, + "loss": 1.0229, + "step": 15376 + }, + { + "epoch": 0.99, + "learning_rate": 1.6722127997595484e-05, + "loss": 1.0172, + "step": 15377 + }, + { + "epoch": 0.99, + "learning_rate": 1.6721913308572536e-05, + "loss": 1.0874, + "step": 15378 + }, + { + "epoch": 0.99, + "learning_rate": 1.6721698619549585e-05, + "loss": 1.026, + "step": 15379 + }, + { + "epoch": 0.99, + "learning_rate": 1.6721483930526634e-05, + "loss": 1.0792, + "step": 15380 + }, + { + "epoch": 0.99, + "learning_rate": 1.6721269241503683e-05, + "loss": 1.0187, + "step": 15381 + }, + { + "epoch": 0.99, + "learning_rate": 1.6721054552480732e-05, + "loss": 1.0743, + "step": 15382 + }, + { + "epoch": 0.99, + "learning_rate": 1.6720839863457784e-05, + "loss": 0.9796, + "step": 15383 + }, + { + "epoch": 0.99, + "learning_rate": 1.6720625174434833e-05, + "loss": 1.0514, + "step": 15384 + }, + { + "epoch": 0.99, + "learning_rate": 1.6720410485411882e-05, + "loss": 1.0192, + "step": 15385 + }, + { + "epoch": 0.99, + "learning_rate": 1.672019579638893e-05, + "loss": 1.0288, + "step": 15386 + }, + { + "epoch": 0.99, + "learning_rate": 1.671998110736598e-05, + "loss": 1.0576, + "step": 15387 + }, + { + "epoch": 0.99, + "learning_rate": 1.671976641834303e-05, + "loss": 1.0998, + "step": 15388 + }, + { + "epoch": 0.99, + "learning_rate": 1.671955172932008e-05, + "loss": 1.066, + "step": 15389 + }, + { + "epoch": 0.99, + "learning_rate": 1.671933704029713e-05, + "loss": 1.0167, + "step": 15390 + }, + { + "epoch": 0.99, + "learning_rate": 1.6719122351274183e-05, + "loss": 1.0277, + "step": 15391 + }, + { + "epoch": 0.99, + "learning_rate": 1.6718907662251232e-05, + "loss": 1.0473, + "step": 15392 + }, + { + "epoch": 0.99, + "learning_rate": 1.671869297322828e-05, + "loss": 0.9798, + "step": 15393 + }, + { + "epoch": 0.99, + "learning_rate": 1.671847828420533e-05, + "loss": 1.0975, + "step": 15394 + }, + { + "epoch": 0.99, + "learning_rate": 1.671826359518238e-05, + "loss": 1.0663, + "step": 15395 + }, + { + "epoch": 0.99, + "learning_rate": 1.671804890615943e-05, + "loss": 1.0051, + "step": 15396 + }, + { + "epoch": 0.99, + "learning_rate": 1.671783421713648e-05, + "loss": 1.0082, + "step": 15397 + }, + { + "epoch": 0.99, + "learning_rate": 1.671761952811353e-05, + "loss": 1.0731, + "step": 15398 + }, + { + "epoch": 0.99, + "learning_rate": 1.6717404839090578e-05, + "loss": 0.9991, + "step": 15399 + }, + { + "epoch": 0.99, + "learning_rate": 1.6717190150067627e-05, + "loss": 1.0585, + "step": 15400 + }, + { + "epoch": 0.99, + "learning_rate": 1.671697546104468e-05, + "loss": 1.1333, + "step": 15401 + }, + { + "epoch": 0.99, + "learning_rate": 1.6716760772021728e-05, + "loss": 1.0746, + "step": 15402 + }, + { + "epoch": 0.99, + "learning_rate": 1.6716546082998777e-05, + "loss": 1.1373, + "step": 15403 + }, + { + "epoch": 0.99, + "learning_rate": 1.6716331393975826e-05, + "loss": 1.1075, + "step": 15404 + }, + { + "epoch": 0.99, + "learning_rate": 1.6716116704952875e-05, + "loss": 1.0619, + "step": 15405 + }, + { + "epoch": 0.99, + "learning_rate": 1.6715902015929928e-05, + "loss": 1.0551, + "step": 15406 + }, + { + "epoch": 0.99, + "learning_rate": 1.6715687326906976e-05, + "loss": 1.0727, + "step": 15407 + }, + { + "epoch": 0.99, + "learning_rate": 1.6715472637884025e-05, + "loss": 1.0326, + "step": 15408 + }, + { + "epoch": 0.99, + "learning_rate": 1.6715257948861078e-05, + "loss": 0.9165, + "step": 15409 + }, + { + "epoch": 0.99, + "learning_rate": 1.6715043259838127e-05, + "loss": 1.0503, + "step": 15410 + }, + { + "epoch": 0.99, + "learning_rate": 1.6714828570815176e-05, + "loss": 1.0365, + "step": 15411 + }, + { + "epoch": 0.99, + "learning_rate": 1.6714613881792225e-05, + "loss": 1.0447, + "step": 15412 + }, + { + "epoch": 0.99, + "learning_rate": 1.6714399192769274e-05, + "loss": 1.0132, + "step": 15413 + }, + { + "epoch": 0.99, + "learning_rate": 1.6714184503746326e-05, + "loss": 0.9893, + "step": 15414 + }, + { + "epoch": 0.99, + "learning_rate": 1.6713969814723375e-05, + "loss": 1.0334, + "step": 15415 + }, + { + "epoch": 0.99, + "learning_rate": 1.6713755125700424e-05, + "loss": 1.1117, + "step": 15416 + }, + { + "epoch": 0.99, + "learning_rate": 1.6713540436677473e-05, + "loss": 1.0377, + "step": 15417 + }, + { + "epoch": 0.99, + "learning_rate": 1.6713325747654522e-05, + "loss": 1.0489, + "step": 15418 + }, + { + "epoch": 0.99, + "learning_rate": 1.6713111058631574e-05, + "loss": 1.0046, + "step": 15419 + }, + { + "epoch": 0.99, + "learning_rate": 1.6712896369608623e-05, + "loss": 1.1317, + "step": 15420 + }, + { + "epoch": 0.99, + "learning_rate": 1.6712681680585676e-05, + "loss": 1.0154, + "step": 15421 + }, + { + "epoch": 0.99, + "learning_rate": 1.6712466991562725e-05, + "loss": 0.9706, + "step": 15422 + }, + { + "epoch": 0.99, + "learning_rate": 1.6712252302539773e-05, + "loss": 1.0685, + "step": 15423 + }, + { + "epoch": 0.99, + "learning_rate": 1.6712037613516822e-05, + "loss": 1.1198, + "step": 15424 + }, + { + "epoch": 0.99, + "learning_rate": 1.671182292449387e-05, + "loss": 1.1492, + "step": 15425 + }, + { + "epoch": 0.99, + "learning_rate": 1.671160823547092e-05, + "loss": 1.0803, + "step": 15426 + }, + { + "epoch": 0.99, + "learning_rate": 1.6711393546447973e-05, + "loss": 0.9718, + "step": 15427 + }, + { + "epoch": 0.99, + "learning_rate": 1.6711178857425022e-05, + "loss": 1.1681, + "step": 15428 + }, + { + "epoch": 0.99, + "learning_rate": 1.671096416840207e-05, + "loss": 0.9984, + "step": 15429 + }, + { + "epoch": 0.99, + "learning_rate": 1.671074947937912e-05, + "loss": 1.0125, + "step": 15430 + }, + { + "epoch": 0.99, + "learning_rate": 1.671053479035617e-05, + "loss": 1.0527, + "step": 15431 + }, + { + "epoch": 0.99, + "learning_rate": 1.671032010133322e-05, + "loss": 1.0172, + "step": 15432 + }, + { + "epoch": 0.99, + "learning_rate": 1.671010541231027e-05, + "loss": 0.9521, + "step": 15433 + }, + { + "epoch": 0.99, + "learning_rate": 1.670989072328732e-05, + "loss": 1.0415, + "step": 15434 + }, + { + "epoch": 0.99, + "learning_rate": 1.670967603426437e-05, + "loss": 1.156, + "step": 15435 + }, + { + "epoch": 0.99, + "learning_rate": 1.670946134524142e-05, + "loss": 1.0381, + "step": 15436 + }, + { + "epoch": 0.99, + "learning_rate": 1.670924665621847e-05, + "loss": 1.1405, + "step": 15437 + }, + { + "epoch": 0.99, + "learning_rate": 1.6709031967195518e-05, + "loss": 1.0332, + "step": 15438 + }, + { + "epoch": 0.99, + "learning_rate": 1.670881727817257e-05, + "loss": 1.0219, + "step": 15439 + }, + { + "epoch": 0.99, + "learning_rate": 1.670860258914962e-05, + "loss": 1.0723, + "step": 15440 + }, + { + "epoch": 0.99, + "learning_rate": 1.670838790012667e-05, + "loss": 0.9868, + "step": 15441 + }, + { + "epoch": 0.99, + "learning_rate": 1.6708173211103717e-05, + "loss": 1.0295, + "step": 15442 + }, + { + "epoch": 0.99, + "learning_rate": 1.6707958522080766e-05, + "loss": 0.9909, + "step": 15443 + }, + { + "epoch": 0.99, + "learning_rate": 1.6707743833057815e-05, + "loss": 1.0484, + "step": 15444 + }, + { + "epoch": 0.99, + "learning_rate": 1.6707529144034868e-05, + "loss": 0.9704, + "step": 15445 + }, + { + "epoch": 0.99, + "learning_rate": 1.6707314455011917e-05, + "loss": 1.091, + "step": 15446 + }, + { + "epoch": 0.99, + "learning_rate": 1.6707099765988966e-05, + "loss": 1.0399, + "step": 15447 + }, + { + "epoch": 0.99, + "learning_rate": 1.6706885076966015e-05, + "loss": 1.082, + "step": 15448 + }, + { + "epoch": 0.99, + "learning_rate": 1.6706670387943067e-05, + "loss": 1.0964, + "step": 15449 + }, + { + "epoch": 0.99, + "learning_rate": 1.6706455698920116e-05, + "loss": 1.0755, + "step": 15450 + }, + { + "epoch": 0.99, + "learning_rate": 1.6706241009897165e-05, + "loss": 1.058, + "step": 15451 + }, + { + "epoch": 0.99, + "learning_rate": 1.6706026320874217e-05, + "loss": 1.008, + "step": 15452 + }, + { + "epoch": 0.99, + "learning_rate": 1.6705811631851266e-05, + "loss": 0.9955, + "step": 15453 + }, + { + "epoch": 0.99, + "learning_rate": 1.6705596942828315e-05, + "loss": 1.1167, + "step": 15454 + }, + { + "epoch": 0.99, + "learning_rate": 1.6705382253805364e-05, + "loss": 1.0852, + "step": 15455 + }, + { + "epoch": 0.99, + "learning_rate": 1.6705167564782413e-05, + "loss": 1.0575, + "step": 15456 + }, + { + "epoch": 0.99, + "learning_rate": 1.6704952875759462e-05, + "loss": 1.0484, + "step": 15457 + }, + { + "epoch": 0.99, + "learning_rate": 1.6704738186736514e-05, + "loss": 1.0987, + "step": 15458 + }, + { + "epoch": 0.99, + "learning_rate": 1.6704523497713563e-05, + "loss": 0.9853, + "step": 15459 + }, + { + "epoch": 0.99, + "learning_rate": 1.6704308808690612e-05, + "loss": 1.0446, + "step": 15460 + }, + { + "epoch": 0.99, + "learning_rate": 1.670409411966766e-05, + "loss": 1.0217, + "step": 15461 + }, + { + "epoch": 0.99, + "learning_rate": 1.670387943064471e-05, + "loss": 1.0751, + "step": 15462 + }, + { + "epoch": 0.99, + "learning_rate": 1.6703664741621763e-05, + "loss": 1.0335, + "step": 15463 + }, + { + "epoch": 0.99, + "learning_rate": 1.670345005259881e-05, + "loss": 1.0079, + "step": 15464 + }, + { + "epoch": 0.99, + "learning_rate": 1.6703235363575864e-05, + "loss": 1.0113, + "step": 15465 + }, + { + "epoch": 1.0, + "learning_rate": 1.6703020674552913e-05, + "loss": 1.0907, + "step": 15466 + }, + { + "epoch": 1.0, + "learning_rate": 1.6702805985529962e-05, + "loss": 0.9725, + "step": 15467 + }, + { + "epoch": 1.0, + "learning_rate": 1.670259129650701e-05, + "loss": 1.0208, + "step": 15468 + }, + { + "epoch": 1.0, + "learning_rate": 1.670237660748406e-05, + "loss": 1.0957, + "step": 15469 + }, + { + "epoch": 1.0, + "learning_rate": 1.6702161918461112e-05, + "loss": 1.0124, + "step": 15470 + }, + { + "epoch": 1.0, + "learning_rate": 1.670194722943816e-05, + "loss": 0.9741, + "step": 15471 + }, + { + "epoch": 1.0, + "learning_rate": 1.670173254041521e-05, + "loss": 0.9221, + "step": 15472 + }, + { + "epoch": 1.0, + "learning_rate": 1.670151785139226e-05, + "loss": 1.1126, + "step": 15473 + }, + { + "epoch": 1.0, + "learning_rate": 1.6701303162369308e-05, + "loss": 1.0741, + "step": 15474 + }, + { + "epoch": 1.0, + "learning_rate": 1.6701088473346357e-05, + "loss": 1.1197, + "step": 15475 + }, + { + "epoch": 1.0, + "learning_rate": 1.670087378432341e-05, + "loss": 1.0971, + "step": 15476 + }, + { + "epoch": 1.0, + "learning_rate": 1.6700659095300458e-05, + "loss": 1.0013, + "step": 15477 + }, + { + "epoch": 1.0, + "learning_rate": 1.6700444406277507e-05, + "loss": 1.0704, + "step": 15478 + }, + { + "epoch": 1.0, + "learning_rate": 1.670022971725456e-05, + "loss": 1.1498, + "step": 15479 + }, + { + "epoch": 1.0, + "learning_rate": 1.670001502823161e-05, + "loss": 1.0393, + "step": 15480 + }, + { + "epoch": 1.0, + "learning_rate": 1.6699800339208657e-05, + "loss": 1.0426, + "step": 15481 + }, + { + "epoch": 1.0, + "learning_rate": 1.6699585650185706e-05, + "loss": 1.0107, + "step": 15482 + }, + { + "epoch": 1.0, + "learning_rate": 1.669937096116276e-05, + "loss": 0.9905, + "step": 15483 + }, + { + "epoch": 1.0, + "learning_rate": 1.6699156272139808e-05, + "loss": 0.9935, + "step": 15484 + }, + { + "epoch": 1.0, + "learning_rate": 1.6698941583116857e-05, + "loss": 1.0506, + "step": 15485 + }, + { + "epoch": 1.0, + "learning_rate": 1.6698726894093906e-05, + "loss": 1.0619, + "step": 15486 + }, + { + "epoch": 1.0, + "learning_rate": 1.6698512205070955e-05, + "loss": 1.0934, + "step": 15487 + }, + { + "epoch": 1.0, + "learning_rate": 1.6698297516048007e-05, + "loss": 1.0734, + "step": 15488 + }, + { + "epoch": 1.0, + "learning_rate": 1.6698082827025056e-05, + "loss": 1.0636, + "step": 15489 + }, + { + "epoch": 1.0, + "learning_rate": 1.6697868138002105e-05, + "loss": 1.0172, + "step": 15490 + }, + { + "epoch": 1.0, + "learning_rate": 1.6697653448979154e-05, + "loss": 1.0099, + "step": 15491 + }, + { + "epoch": 1.0, + "learning_rate": 1.6697438759956203e-05, + "loss": 1.0359, + "step": 15492 + }, + { + "epoch": 1.0, + "learning_rate": 1.6697224070933255e-05, + "loss": 1.0174, + "step": 15493 + }, + { + "epoch": 1.0, + "learning_rate": 1.6697009381910304e-05, + "loss": 1.1376, + "step": 15494 + }, + { + "epoch": 1.0, + "learning_rate": 1.6696794692887353e-05, + "loss": 1.1175, + "step": 15495 + }, + { + "epoch": 1.0, + "learning_rate": 1.6696580003864406e-05, + "loss": 1.142, + "step": 15496 + }, + { + "epoch": 1.0, + "learning_rate": 1.6696365314841454e-05, + "loss": 1.1308, + "step": 15497 + }, + { + "epoch": 1.0, + "learning_rate": 1.6696150625818503e-05, + "loss": 1.082, + "step": 15498 + }, + { + "epoch": 1.0, + "learning_rate": 1.6695935936795552e-05, + "loss": 1.0509, + "step": 15499 + }, + { + "epoch": 1.0, + "learning_rate": 1.66957212477726e-05, + "loss": 1.0173, + "step": 15500 + }, + { + "epoch": 1.0, + "learning_rate": 1.6695506558749654e-05, + "loss": 1.0874, + "step": 15501 + }, + { + "epoch": 1.0, + "learning_rate": 1.6695291869726703e-05, + "loss": 1.0712, + "step": 15502 + }, + { + "epoch": 1.0, + "learning_rate": 1.669507718070375e-05, + "loss": 1.1116, + "step": 15503 + }, + { + "epoch": 1.0, + "learning_rate": 1.66948624916808e-05, + "loss": 1.0445, + "step": 15504 + }, + { + "epoch": 1.0, + "learning_rate": 1.669464780265785e-05, + "loss": 1.0834, + "step": 15505 + }, + { + "epoch": 1.0, + "learning_rate": 1.6694433113634902e-05, + "loss": 1.0052, + "step": 15506 + }, + { + "epoch": 1.0, + "learning_rate": 1.669421842461195e-05, + "loss": 1.0568, + "step": 15507 + }, + { + "epoch": 1.0, + "learning_rate": 1.6694003735589003e-05, + "loss": 1.0023, + "step": 15508 + }, + { + "epoch": 1.0, + "learning_rate": 1.6693789046566052e-05, + "loss": 1.0598, + "step": 15509 + }, + { + "epoch": 1.0, + "learning_rate": 1.66935743575431e-05, + "loss": 1.0218, + "step": 15510 + }, + { + "epoch": 1.0, + "learning_rate": 1.669335966852015e-05, + "loss": 1.0728, + "step": 15511 + }, + { + "epoch": 1.0, + "learning_rate": 1.66931449794972e-05, + "loss": 1.156, + "step": 15512 + }, + { + "epoch": 1.0, + "learning_rate": 1.6692930290474248e-05, + "loss": 1.0473, + "step": 15513 + }, + { + "epoch": 1.0, + "learning_rate": 1.66927156014513e-05, + "loss": 0.9463, + "step": 15514 + }, + { + "epoch": 1.0, + "learning_rate": 1.669250091242835e-05, + "loss": 1.0301, + "step": 15515 + }, + { + "epoch": 1.0, + "learning_rate": 1.66922862234054e-05, + "loss": 1.0042, + "step": 15516 + }, + { + "epoch": 1.0, + "learning_rate": 1.6692071534382447e-05, + "loss": 1.0445, + "step": 15517 + }, + { + "epoch": 1.0, + "learning_rate": 1.6691856845359496e-05, + "loss": 1.0591, + "step": 15518 + }, + { + "epoch": 1.0, + "learning_rate": 1.669164215633655e-05, + "loss": 1.0314, + "step": 15519 + }, + { + "epoch": 1.0, + "learning_rate": 1.6691427467313598e-05, + "loss": 0.9694, + "step": 15520 + }, + { + "epoch": 1.0, + "learning_rate": 1.6691212778290647e-05, + "loss": 1.0219, + "step": 15521 + }, + { + "epoch": 1.0, + "learning_rate": 1.66909980892677e-05, + "loss": 1.0046, + "step": 15522 + }, + { + "epoch": 1.0, + "learning_rate": 1.6690783400244748e-05, + "loss": 1.1236, + "step": 15523 + }, + { + "epoch": 1.0, + "learning_rate": 1.6690568711221797e-05, + "loss": 1.1457, + "step": 15524 + }, + { + "epoch": 1.0, + "learning_rate": 1.6690354022198846e-05, + "loss": 1.068, + "step": 15525 + }, + { + "epoch": 1.0, + "learning_rate": 1.6690139333175898e-05, + "loss": 1.1308, + "step": 15526 + }, + { + "epoch": 1.0, + "learning_rate": 1.6689924644152947e-05, + "loss": 1.0325, + "step": 15527 + }, + { + "epoch": 1.0, + "learning_rate": 1.6689709955129996e-05, + "loss": 1.1406, + "step": 15528 + }, + { + "epoch": 1.0, + "learning_rate": 1.6689495266107045e-05, + "loss": 1.025, + "step": 15529 + }, + { + "epoch": 1.0, + "learning_rate": 1.6689280577084094e-05, + "loss": 1.0676, + "step": 15530 + }, + { + "epoch": 1.0, + "learning_rate": 1.6689065888061143e-05, + "loss": 1.1087, + "step": 15531 + }, + { + "epoch": 1.0, + "learning_rate": 1.6688851199038195e-05, + "loss": 1.0521, + "step": 15532 + }, + { + "epoch": 1.0, + "learning_rate": 1.6688636510015244e-05, + "loss": 1.0309, + "step": 15533 + }, + { + "epoch": 1.0, + "learning_rate": 1.6688421820992293e-05, + "loss": 1.0437, + "step": 15534 + }, + { + "epoch": 1.0, + "learning_rate": 1.6688207131969342e-05, + "loss": 1.048, + "step": 15535 + }, + { + "epoch": 1.0, + "learning_rate": 1.668799244294639e-05, + "loss": 1.0442, + "step": 15536 + }, + { + "epoch": 1.0, + "learning_rate": 1.6687777753923444e-05, + "loss": 1.048, + "step": 15537 + }, + { + "epoch": 1.0, + "learning_rate": 1.6687563064900493e-05, + "loss": 1.0636, + "step": 15538 + }, + { + "epoch": 1.0, + "learning_rate": 1.6687348375877545e-05, + "loss": 0.9711, + "step": 15539 + }, + { + "epoch": 1.0, + "learning_rate": 1.6687133686854594e-05, + "loss": 1.0509, + "step": 15540 + }, + { + "epoch": 1.0, + "learning_rate": 1.6686918997831643e-05, + "loss": 1.1026, + "step": 15541 + }, + { + "epoch": 1.0, + "learning_rate": 1.6686704308808692e-05, + "loss": 1.1124, + "step": 15542 + }, + { + "epoch": 1.0, + "learning_rate": 1.668648961978574e-05, + "loss": 1.0862, + "step": 15543 + }, + { + "epoch": 1.0, + "learning_rate": 1.668627493076279e-05, + "loss": 0.9725, + "step": 15544 + }, + { + "epoch": 1.0, + "learning_rate": 1.6686060241739842e-05, + "loss": 0.92, + "step": 15545 + }, + { + "epoch": 1.0, + "learning_rate": 1.668584555271689e-05, + "loss": 0.974, + "step": 15546 + }, + { + "epoch": 1.0, + "learning_rate": 1.668563086369394e-05, + "loss": 1.0335, + "step": 15547 + }, + { + "epoch": 1.0, + "learning_rate": 1.668541617467099e-05, + "loss": 1.0097, + "step": 15548 + }, + { + "epoch": 1.0, + "learning_rate": 1.6685201485648038e-05, + "loss": 0.95, + "step": 15549 + }, + { + "epoch": 1.0, + "learning_rate": 1.668498679662509e-05, + "loss": 0.9523, + "step": 15550 + }, + { + "epoch": 1.0, + "learning_rate": 1.668477210760214e-05, + "loss": 0.9199, + "step": 15551 + }, + { + "epoch": 1.0, + "learning_rate": 1.668455741857919e-05, + "loss": 0.982, + "step": 15552 + }, + { + "epoch": 1.0, + "learning_rate": 1.668434272955624e-05, + "loss": 0.8921, + "step": 15553 + }, + { + "epoch": 1.0, + "learning_rate": 1.668412804053329e-05, + "loss": 1.0541, + "step": 15554 + }, + { + "epoch": 1.0, + "learning_rate": 1.668391335151034e-05, + "loss": 0.9881, + "step": 15555 + }, + { + "epoch": 1.0, + "learning_rate": 1.6683698662487387e-05, + "loss": 0.9898, + "step": 15556 + }, + { + "epoch": 1.0, + "learning_rate": 1.668348397346444e-05, + "loss": 1.0138, + "step": 15557 + }, + { + "epoch": 1.0, + "learning_rate": 1.668326928444149e-05, + "loss": 1.0039, + "step": 15558 + }, + { + "epoch": 1.0, + "learning_rate": 1.6683054595418538e-05, + "loss": 0.927, + "step": 15559 + }, + { + "epoch": 1.0, + "learning_rate": 1.6682839906395587e-05, + "loss": 0.952, + "step": 15560 + }, + { + "epoch": 1.0, + "learning_rate": 1.6682625217372636e-05, + "loss": 0.9632, + "step": 15561 + }, + { + "epoch": 1.0, + "learning_rate": 1.6682410528349685e-05, + "loss": 0.9883, + "step": 15562 + }, + { + "epoch": 1.0, + "learning_rate": 1.6682195839326737e-05, + "loss": 1.0712, + "step": 15563 + }, + { + "epoch": 1.0, + "learning_rate": 1.6681981150303786e-05, + "loss": 1.01, + "step": 15564 + }, + { + "epoch": 1.0, + "learning_rate": 1.6681766461280835e-05, + "loss": 0.9649, + "step": 15565 + }, + { + "epoch": 1.0, + "learning_rate": 1.6681551772257887e-05, + "loss": 0.9544, + "step": 15566 + }, + { + "epoch": 1.0, + "learning_rate": 1.6681337083234936e-05, + "loss": 0.9686, + "step": 15567 + }, + { + "epoch": 1.0, + "learning_rate": 1.6681122394211985e-05, + "loss": 0.9656, + "step": 15568 + }, + { + "epoch": 1.0, + "learning_rate": 1.6680907705189034e-05, + "loss": 0.9147, + "step": 15569 + }, + { + "epoch": 1.0, + "learning_rate": 1.6680693016166087e-05, + "loss": 0.9793, + "step": 15570 + }, + { + "epoch": 1.0, + "learning_rate": 1.6680478327143135e-05, + "loss": 0.9821, + "step": 15571 + }, + { + "epoch": 1.0, + "learning_rate": 1.6680263638120184e-05, + "loss": 0.9461, + "step": 15572 + }, + { + "epoch": 1.0, + "learning_rate": 1.6680048949097233e-05, + "loss": 0.966, + "step": 15573 + }, + { + "epoch": 1.0, + "learning_rate": 1.6679834260074282e-05, + "loss": 1.0056, + "step": 15574 + }, + { + "epoch": 1.0, + "learning_rate": 1.6679619571051335e-05, + "loss": 0.9795, + "step": 15575 + }, + { + "epoch": 1.0, + "learning_rate": 1.6679404882028384e-05, + "loss": 0.9807, + "step": 15576 + }, + { + "epoch": 1.0, + "learning_rate": 1.6679190193005433e-05, + "loss": 1.0282, + "step": 15577 + }, + { + "epoch": 1.0, + "learning_rate": 1.667897550398248e-05, + "loss": 1.0246, + "step": 15578 + }, + { + "epoch": 1.0, + "learning_rate": 1.667876081495953e-05, + "loss": 0.9493, + "step": 15579 + }, + { + "epoch": 1.0, + "learning_rate": 1.6678546125936583e-05, + "loss": 0.9482, + "step": 15580 + }, + { + "epoch": 1.0, + "learning_rate": 1.6678331436913632e-05, + "loss": 1.1129, + "step": 15581 + }, + { + "epoch": 1.0, + "learning_rate": 1.667811674789068e-05, + "loss": 0.9134, + "step": 15582 + }, + { + "epoch": 1.0, + "learning_rate": 1.6677902058867733e-05, + "loss": 1.0109, + "step": 15583 + }, + { + "epoch": 1.0, + "learning_rate": 1.6677687369844782e-05, + "loss": 0.9772, + "step": 15584 + }, + { + "epoch": 1.0, + "learning_rate": 1.667747268082183e-05, + "loss": 0.9398, + "step": 15585 + }, + { + "epoch": 1.0, + "learning_rate": 1.667725799179888e-05, + "loss": 0.8499, + "step": 15586 + }, + { + "epoch": 1.0, + "learning_rate": 1.667704330277593e-05, + "loss": 1.0638, + "step": 15587 + }, + { + "epoch": 1.0, + "learning_rate": 1.667682861375298e-05, + "loss": 1.0154, + "step": 15588 + }, + { + "epoch": 1.0, + "learning_rate": 1.667661392473003e-05, + "loss": 0.9614, + "step": 15589 + }, + { + "epoch": 1.0, + "learning_rate": 1.667639923570708e-05, + "loss": 0.9383, + "step": 15590 + }, + { + "epoch": 1.0, + "learning_rate": 1.667618454668413e-05, + "loss": 0.9399, + "step": 15591 + }, + { + "epoch": 1.0, + "learning_rate": 1.6675969857661177e-05, + "loss": 0.9318, + "step": 15592 + }, + { + "epoch": 1.0, + "learning_rate": 1.6675755168638226e-05, + "loss": 0.9917, + "step": 15593 + }, + { + "epoch": 1.0, + "learning_rate": 1.667554047961528e-05, + "loss": 1.0141, + "step": 15594 + }, + { + "epoch": 1.0, + "learning_rate": 1.667532579059233e-05, + "loss": 0.9645, + "step": 15595 + }, + { + "epoch": 1.0, + "learning_rate": 1.667511110156938e-05, + "loss": 1.0242, + "step": 15596 + }, + { + "epoch": 1.0, + "learning_rate": 1.667489641254643e-05, + "loss": 0.9725, + "step": 15597 + }, + { + "epoch": 1.0, + "learning_rate": 1.6674681723523478e-05, + "loss": 1.0383, + "step": 15598 + }, + { + "epoch": 1.0, + "learning_rate": 1.6674467034500527e-05, + "loss": 1.0031, + "step": 15599 + }, + { + "epoch": 1.0, + "learning_rate": 1.6674252345477576e-05, + "loss": 1.0599, + "step": 15600 + }, + { + "epoch": 1.0, + "learning_rate": 1.6674037656454628e-05, + "loss": 1.0229, + "step": 15601 + }, + { + "epoch": 1.0, + "learning_rate": 1.6673822967431677e-05, + "loss": 0.9721, + "step": 15602 + }, + { + "epoch": 1.0, + "learning_rate": 1.6673608278408726e-05, + "loss": 0.9406, + "step": 15603 + }, + { + "epoch": 1.0, + "learning_rate": 1.6673393589385775e-05, + "loss": 0.9571, + "step": 15604 + }, + { + "epoch": 1.0, + "learning_rate": 1.6673178900362824e-05, + "loss": 0.9844, + "step": 15605 + }, + { + "epoch": 1.0, + "learning_rate": 1.6672964211339876e-05, + "loss": 1.0213, + "step": 15606 + }, + { + "epoch": 1.0, + "learning_rate": 1.6672749522316925e-05, + "loss": 0.9334, + "step": 15607 + }, + { + "epoch": 1.0, + "learning_rate": 1.6672534833293974e-05, + "loss": 0.9115, + "step": 15608 + }, + { + "epoch": 1.0, + "learning_rate": 1.6672320144271027e-05, + "loss": 0.9637, + "step": 15609 + }, + { + "epoch": 1.0, + "learning_rate": 1.6672105455248076e-05, + "loss": 0.9454, + "step": 15610 + }, + { + "epoch": 1.0, + "learning_rate": 1.6671890766225125e-05, + "loss": 0.9168, + "step": 15611 + }, + { + "epoch": 1.0, + "learning_rate": 1.6671676077202174e-05, + "loss": 1.0135, + "step": 15612 + }, + { + "epoch": 1.0, + "learning_rate": 1.6671461388179226e-05, + "loss": 0.9333, + "step": 15613 + }, + { + "epoch": 1.0, + "learning_rate": 1.6671246699156275e-05, + "loss": 1.0399, + "step": 15614 + }, + { + "epoch": 1.0, + "learning_rate": 1.6671032010133324e-05, + "loss": 1.0056, + "step": 15615 + }, + { + "epoch": 1.0, + "learning_rate": 1.6670817321110373e-05, + "loss": 0.9463, + "step": 15616 + }, + { + "epoch": 1.0, + "learning_rate": 1.6670602632087422e-05, + "loss": 1.0041, + "step": 15617 + }, + { + "epoch": 1.0, + "learning_rate": 1.667038794306447e-05, + "loss": 0.9451, + "step": 15618 + }, + { + "epoch": 1.0, + "learning_rate": 1.6670173254041523e-05, + "loss": 0.9463, + "step": 15619 + }, + { + "epoch": 1.0, + "learning_rate": 1.6669958565018572e-05, + "loss": 0.9408, + "step": 15620 + }, + { + "epoch": 1.01, + "learning_rate": 1.666974387599562e-05, + "loss": 0.9861, + "step": 15621 + }, + { + "epoch": 1.01, + "learning_rate": 1.666952918697267e-05, + "loss": 0.9659, + "step": 15622 + }, + { + "epoch": 1.01, + "learning_rate": 1.666931449794972e-05, + "loss": 0.996, + "step": 15623 + }, + { + "epoch": 1.01, + "learning_rate": 1.666909980892677e-05, + "loss": 0.9511, + "step": 15624 + }, + { + "epoch": 1.01, + "learning_rate": 1.666888511990382e-05, + "loss": 0.9615, + "step": 15625 + }, + { + "epoch": 1.01, + "learning_rate": 1.6668670430880873e-05, + "loss": 1.0307, + "step": 15626 + }, + { + "epoch": 1.01, + "learning_rate": 1.666845574185792e-05, + "loss": 0.9337, + "step": 15627 + }, + { + "epoch": 1.01, + "learning_rate": 1.666824105283497e-05, + "loss": 0.9406, + "step": 15628 + }, + { + "epoch": 1.01, + "learning_rate": 1.666802636381202e-05, + "loss": 0.9998, + "step": 15629 + }, + { + "epoch": 1.01, + "learning_rate": 1.666781167478907e-05, + "loss": 0.9646, + "step": 15630 + }, + { + "epoch": 1.01, + "learning_rate": 1.6667596985766117e-05, + "loss": 0.9579, + "step": 15631 + }, + { + "epoch": 1.01, + "learning_rate": 1.666738229674317e-05, + "loss": 0.9698, + "step": 15632 + }, + { + "epoch": 1.01, + "learning_rate": 1.666716760772022e-05, + "loss": 1.0858, + "step": 15633 + }, + { + "epoch": 1.01, + "learning_rate": 1.6666952918697268e-05, + "loss": 0.9853, + "step": 15634 + }, + { + "epoch": 1.01, + "learning_rate": 1.6666738229674317e-05, + "loss": 0.9292, + "step": 15635 + }, + { + "epoch": 1.01, + "learning_rate": 1.6666523540651366e-05, + "loss": 1.0495, + "step": 15636 + }, + { + "epoch": 1.01, + "learning_rate": 1.6666308851628418e-05, + "loss": 0.9442, + "step": 15637 + }, + { + "epoch": 1.01, + "learning_rate": 1.6666094162605467e-05, + "loss": 0.9326, + "step": 15638 + }, + { + "epoch": 1.01, + "learning_rate": 1.666587947358252e-05, + "loss": 0.9899, + "step": 15639 + }, + { + "epoch": 1.01, + "learning_rate": 1.6665664784559568e-05, + "loss": 0.9006, + "step": 15640 + }, + { + "epoch": 1.01, + "learning_rate": 1.6665450095536617e-05, + "loss": 1.0203, + "step": 15641 + }, + { + "epoch": 1.01, + "learning_rate": 1.6665235406513666e-05, + "loss": 0.9366, + "step": 15642 + }, + { + "epoch": 1.01, + "learning_rate": 1.6665020717490715e-05, + "loss": 0.9331, + "step": 15643 + }, + { + "epoch": 1.01, + "learning_rate": 1.6664806028467768e-05, + "loss": 1.0241, + "step": 15644 + }, + { + "epoch": 1.01, + "learning_rate": 1.6664591339444816e-05, + "loss": 0.96, + "step": 15645 + }, + { + "epoch": 1.01, + "learning_rate": 1.6664376650421865e-05, + "loss": 1.0359, + "step": 15646 + }, + { + "epoch": 1.01, + "learning_rate": 1.6664161961398914e-05, + "loss": 0.8838, + "step": 15647 + }, + { + "epoch": 1.01, + "learning_rate": 1.6663947272375963e-05, + "loss": 0.936, + "step": 15648 + }, + { + "epoch": 1.01, + "learning_rate": 1.6663732583353012e-05, + "loss": 0.9735, + "step": 15649 + }, + { + "epoch": 1.01, + "learning_rate": 1.6663517894330065e-05, + "loss": 1.0035, + "step": 15650 + }, + { + "epoch": 1.01, + "learning_rate": 1.6663303205307114e-05, + "loss": 0.9585, + "step": 15651 + }, + { + "epoch": 1.01, + "learning_rate": 1.6663088516284163e-05, + "loss": 0.9699, + "step": 15652 + }, + { + "epoch": 1.01, + "learning_rate": 1.6662873827261215e-05, + "loss": 1.0632, + "step": 15653 + }, + { + "epoch": 1.01, + "learning_rate": 1.6662659138238264e-05, + "loss": 0.9132, + "step": 15654 + }, + { + "epoch": 1.01, + "learning_rate": 1.6662444449215313e-05, + "loss": 0.916, + "step": 15655 + }, + { + "epoch": 1.01, + "learning_rate": 1.6662229760192362e-05, + "loss": 0.9888, + "step": 15656 + }, + { + "epoch": 1.01, + "learning_rate": 1.6662015071169414e-05, + "loss": 1.005, + "step": 15657 + }, + { + "epoch": 1.01, + "learning_rate": 1.6661800382146463e-05, + "loss": 1.0332, + "step": 15658 + }, + { + "epoch": 1.01, + "learning_rate": 1.6661585693123512e-05, + "loss": 0.9614, + "step": 15659 + }, + { + "epoch": 1.01, + "learning_rate": 1.666137100410056e-05, + "loss": 0.8833, + "step": 15660 + }, + { + "epoch": 1.01, + "learning_rate": 1.666115631507761e-05, + "loss": 0.9739, + "step": 15661 + }, + { + "epoch": 1.01, + "learning_rate": 1.6660941626054662e-05, + "loss": 0.9246, + "step": 15662 + }, + { + "epoch": 1.01, + "learning_rate": 1.666072693703171e-05, + "loss": 0.9541, + "step": 15663 + }, + { + "epoch": 1.01, + "learning_rate": 1.666051224800876e-05, + "loss": 0.9701, + "step": 15664 + }, + { + "epoch": 1.01, + "learning_rate": 1.666029755898581e-05, + "loss": 0.8573, + "step": 15665 + }, + { + "epoch": 1.01, + "learning_rate": 1.666008286996286e-05, + "loss": 0.9834, + "step": 15666 + }, + { + "epoch": 1.01, + "learning_rate": 1.665986818093991e-05, + "loss": 0.9777, + "step": 15667 + }, + { + "epoch": 1.01, + "learning_rate": 1.665965349191696e-05, + "loss": 0.9699, + "step": 15668 + }, + { + "epoch": 1.01, + "learning_rate": 1.665943880289401e-05, + "loss": 0.9427, + "step": 15669 + }, + { + "epoch": 1.01, + "learning_rate": 1.665922411387106e-05, + "loss": 1.0195, + "step": 15670 + }, + { + "epoch": 1.01, + "learning_rate": 1.665900942484811e-05, + "loss": 0.9672, + "step": 15671 + }, + { + "epoch": 1.01, + "learning_rate": 1.665879473582516e-05, + "loss": 1.0185, + "step": 15672 + }, + { + "epoch": 1.01, + "learning_rate": 1.6658580046802208e-05, + "loss": 1.0107, + "step": 15673 + }, + { + "epoch": 1.01, + "learning_rate": 1.6658365357779257e-05, + "loss": 1.0798, + "step": 15674 + }, + { + "epoch": 1.01, + "learning_rate": 1.665815066875631e-05, + "loss": 1.0109, + "step": 15675 + }, + { + "epoch": 1.01, + "learning_rate": 1.6657935979733358e-05, + "loss": 0.9425, + "step": 15676 + }, + { + "epoch": 1.01, + "learning_rate": 1.6657721290710407e-05, + "loss": 0.8828, + "step": 15677 + }, + { + "epoch": 1.01, + "learning_rate": 1.6657506601687456e-05, + "loss": 0.9478, + "step": 15678 + }, + { + "epoch": 1.01, + "learning_rate": 1.6657291912664505e-05, + "loss": 1.0474, + "step": 15679 + }, + { + "epoch": 1.01, + "learning_rate": 1.6657077223641554e-05, + "loss": 0.9844, + "step": 15680 + }, + { + "epoch": 1.01, + "learning_rate": 1.6656862534618606e-05, + "loss": 0.922, + "step": 15681 + }, + { + "epoch": 1.01, + "learning_rate": 1.665664784559566e-05, + "loss": 1.0282, + "step": 15682 + }, + { + "epoch": 1.01, + "learning_rate": 1.6656433156572708e-05, + "loss": 0.9662, + "step": 15683 + }, + { + "epoch": 1.01, + "learning_rate": 1.6656218467549757e-05, + "loss": 0.964, + "step": 15684 + }, + { + "epoch": 1.01, + "learning_rate": 1.6656003778526806e-05, + "loss": 0.9953, + "step": 15685 + }, + { + "epoch": 1.01, + "learning_rate": 1.6655789089503855e-05, + "loss": 0.9084, + "step": 15686 + }, + { + "epoch": 1.01, + "learning_rate": 1.6655574400480904e-05, + "loss": 0.9996, + "step": 15687 + }, + { + "epoch": 1.01, + "learning_rate": 1.6655359711457956e-05, + "loss": 0.9811, + "step": 15688 + }, + { + "epoch": 1.01, + "learning_rate": 1.6655145022435005e-05, + "loss": 1.0094, + "step": 15689 + }, + { + "epoch": 1.01, + "learning_rate": 1.6654930333412054e-05, + "loss": 0.9924, + "step": 15690 + }, + { + "epoch": 1.01, + "learning_rate": 1.6654715644389103e-05, + "loss": 1.0205, + "step": 15691 + }, + { + "epoch": 1.01, + "learning_rate": 1.6654500955366152e-05, + "loss": 0.9806, + "step": 15692 + }, + { + "epoch": 1.01, + "learning_rate": 1.6654286266343204e-05, + "loss": 1.024, + "step": 15693 + }, + { + "epoch": 1.01, + "learning_rate": 1.6654071577320253e-05, + "loss": 0.9995, + "step": 15694 + }, + { + "epoch": 1.01, + "learning_rate": 1.6653856888297302e-05, + "loss": 0.9675, + "step": 15695 + }, + { + "epoch": 1.01, + "learning_rate": 1.6653642199274354e-05, + "loss": 0.936, + "step": 15696 + }, + { + "epoch": 1.01, + "learning_rate": 1.6653427510251403e-05, + "loss": 1.0182, + "step": 15697 + }, + { + "epoch": 1.01, + "learning_rate": 1.6653212821228452e-05, + "loss": 0.9573, + "step": 15698 + }, + { + "epoch": 1.01, + "learning_rate": 1.66529981322055e-05, + "loss": 0.9526, + "step": 15699 + }, + { + "epoch": 1.01, + "learning_rate": 1.665278344318255e-05, + "loss": 0.9808, + "step": 15700 + }, + { + "epoch": 1.01, + "learning_rate": 1.6652568754159603e-05, + "loss": 0.9352, + "step": 15701 + }, + { + "epoch": 1.01, + "learning_rate": 1.665235406513665e-05, + "loss": 1.0005, + "step": 15702 + }, + { + "epoch": 1.01, + "learning_rate": 1.66521393761137e-05, + "loss": 1.0289, + "step": 15703 + }, + { + "epoch": 1.01, + "learning_rate": 1.665192468709075e-05, + "loss": 1.0365, + "step": 15704 + }, + { + "epoch": 1.01, + "learning_rate": 1.66517099980678e-05, + "loss": 0.9695, + "step": 15705 + }, + { + "epoch": 1.01, + "learning_rate": 1.665149530904485e-05, + "loss": 0.9857, + "step": 15706 + }, + { + "epoch": 1.01, + "learning_rate": 1.66512806200219e-05, + "loss": 0.9762, + "step": 15707 + }, + { + "epoch": 1.01, + "learning_rate": 1.665106593099895e-05, + "loss": 0.9286, + "step": 15708 + }, + { + "epoch": 1.01, + "learning_rate": 1.6650851241975998e-05, + "loss": 0.9497, + "step": 15709 + }, + { + "epoch": 1.01, + "learning_rate": 1.6650636552953047e-05, + "loss": 0.999, + "step": 15710 + }, + { + "epoch": 1.01, + "learning_rate": 1.66504218639301e-05, + "loss": 0.9763, + "step": 15711 + }, + { + "epoch": 1.01, + "learning_rate": 1.6650207174907148e-05, + "loss": 0.95, + "step": 15712 + }, + { + "epoch": 1.01, + "learning_rate": 1.66499924858842e-05, + "loss": 1.028, + "step": 15713 + }, + { + "epoch": 1.01, + "learning_rate": 1.664977779686125e-05, + "loss": 0.9192, + "step": 15714 + }, + { + "epoch": 1.01, + "learning_rate": 1.6649563107838298e-05, + "loss": 1.0027, + "step": 15715 + }, + { + "epoch": 1.01, + "learning_rate": 1.6649348418815347e-05, + "loss": 1.0066, + "step": 15716 + }, + { + "epoch": 1.01, + "learning_rate": 1.6649133729792396e-05, + "loss": 0.9709, + "step": 15717 + }, + { + "epoch": 1.01, + "learning_rate": 1.6648919040769445e-05, + "loss": 1.0336, + "step": 15718 + }, + { + "epoch": 1.01, + "learning_rate": 1.6648704351746497e-05, + "loss": 0.9428, + "step": 15719 + }, + { + "epoch": 1.01, + "learning_rate": 1.6648489662723546e-05, + "loss": 0.9682, + "step": 15720 + }, + { + "epoch": 1.01, + "learning_rate": 1.6648274973700595e-05, + "loss": 0.9734, + "step": 15721 + }, + { + "epoch": 1.01, + "learning_rate": 1.6648060284677644e-05, + "loss": 0.9383, + "step": 15722 + }, + { + "epoch": 1.01, + "learning_rate": 1.6647845595654693e-05, + "loss": 1.0932, + "step": 15723 + }, + { + "epoch": 1.01, + "learning_rate": 1.6647630906631746e-05, + "loss": 0.9611, + "step": 15724 + }, + { + "epoch": 1.01, + "learning_rate": 1.6647416217608795e-05, + "loss": 0.9684, + "step": 15725 + }, + { + "epoch": 1.01, + "learning_rate": 1.6647201528585847e-05, + "loss": 0.955, + "step": 15726 + }, + { + "epoch": 1.01, + "learning_rate": 1.6646986839562896e-05, + "loss": 0.9771, + "step": 15727 + }, + { + "epoch": 1.01, + "learning_rate": 1.6646772150539945e-05, + "loss": 1.0018, + "step": 15728 + }, + { + "epoch": 1.01, + "learning_rate": 1.6646557461516994e-05, + "loss": 0.9921, + "step": 15729 + }, + { + "epoch": 1.01, + "learning_rate": 1.6646342772494043e-05, + "loss": 0.99, + "step": 15730 + }, + { + "epoch": 1.01, + "learning_rate": 1.6646128083471095e-05, + "loss": 1.0121, + "step": 15731 + }, + { + "epoch": 1.01, + "learning_rate": 1.6645913394448144e-05, + "loss": 1.0156, + "step": 15732 + }, + { + "epoch": 1.01, + "learning_rate": 1.6645698705425193e-05, + "loss": 0.9447, + "step": 15733 + }, + { + "epoch": 1.01, + "learning_rate": 1.6645484016402242e-05, + "loss": 0.9167, + "step": 15734 + }, + { + "epoch": 1.01, + "learning_rate": 1.664526932737929e-05, + "loss": 1.0539, + "step": 15735 + }, + { + "epoch": 1.01, + "learning_rate": 1.664505463835634e-05, + "loss": 1.0836, + "step": 15736 + }, + { + "epoch": 1.01, + "learning_rate": 1.6644839949333392e-05, + "loss": 0.9908, + "step": 15737 + }, + { + "epoch": 1.01, + "learning_rate": 1.664462526031044e-05, + "loss": 0.9251, + "step": 15738 + }, + { + "epoch": 1.01, + "learning_rate": 1.664441057128749e-05, + "loss": 1.0099, + "step": 15739 + }, + { + "epoch": 1.01, + "learning_rate": 1.6644195882264543e-05, + "loss": 0.9509, + "step": 15740 + }, + { + "epoch": 1.01, + "learning_rate": 1.664398119324159e-05, + "loss": 0.954, + "step": 15741 + }, + { + "epoch": 1.01, + "learning_rate": 1.664376650421864e-05, + "loss": 1.0315, + "step": 15742 + }, + { + "epoch": 1.01, + "learning_rate": 1.664355181519569e-05, + "loss": 0.8963, + "step": 15743 + }, + { + "epoch": 1.01, + "learning_rate": 1.6643337126172742e-05, + "loss": 0.9312, + "step": 15744 + }, + { + "epoch": 1.01, + "learning_rate": 1.664312243714979e-05, + "loss": 0.9279, + "step": 15745 + }, + { + "epoch": 1.01, + "learning_rate": 1.664290774812684e-05, + "loss": 1.0341, + "step": 15746 + }, + { + "epoch": 1.01, + "learning_rate": 1.664269305910389e-05, + "loss": 0.9652, + "step": 15747 + }, + { + "epoch": 1.01, + "learning_rate": 1.6642478370080938e-05, + "loss": 1.0215, + "step": 15748 + }, + { + "epoch": 1.01, + "learning_rate": 1.664226368105799e-05, + "loss": 0.9703, + "step": 15749 + }, + { + "epoch": 1.01, + "learning_rate": 1.664204899203504e-05, + "loss": 0.9444, + "step": 15750 + }, + { + "epoch": 1.01, + "learning_rate": 1.6641834303012088e-05, + "loss": 0.9382, + "step": 15751 + }, + { + "epoch": 1.01, + "learning_rate": 1.6641619613989137e-05, + "loss": 0.9668, + "step": 15752 + }, + { + "epoch": 1.01, + "learning_rate": 1.6641404924966186e-05, + "loss": 0.9031, + "step": 15753 + }, + { + "epoch": 1.01, + "learning_rate": 1.664119023594324e-05, + "loss": 0.9528, + "step": 15754 + }, + { + "epoch": 1.01, + "learning_rate": 1.6640975546920287e-05, + "loss": 0.9224, + "step": 15755 + }, + { + "epoch": 1.01, + "learning_rate": 1.6640760857897336e-05, + "loss": 0.9858, + "step": 15756 + }, + { + "epoch": 1.01, + "learning_rate": 1.664054616887439e-05, + "loss": 1.0024, + "step": 15757 + }, + { + "epoch": 1.01, + "learning_rate": 1.6640331479851438e-05, + "loss": 0.9014, + "step": 15758 + }, + { + "epoch": 1.01, + "learning_rate": 1.6640116790828487e-05, + "loss": 0.962, + "step": 15759 + }, + { + "epoch": 1.01, + "learning_rate": 1.6639902101805536e-05, + "loss": 0.9843, + "step": 15760 + }, + { + "epoch": 1.01, + "learning_rate": 1.6639687412782585e-05, + "loss": 1.0243, + "step": 15761 + }, + { + "epoch": 1.01, + "learning_rate": 1.6639472723759637e-05, + "loss": 0.982, + "step": 15762 + }, + { + "epoch": 1.01, + "learning_rate": 1.6639258034736686e-05, + "loss": 0.9939, + "step": 15763 + }, + { + "epoch": 1.01, + "learning_rate": 1.6639043345713735e-05, + "loss": 0.9584, + "step": 15764 + }, + { + "epoch": 1.01, + "learning_rate": 1.6638828656690784e-05, + "loss": 1.0088, + "step": 15765 + }, + { + "epoch": 1.01, + "learning_rate": 1.6638613967667833e-05, + "loss": 0.9604, + "step": 15766 + }, + { + "epoch": 1.01, + "learning_rate": 1.6638399278644882e-05, + "loss": 0.974, + "step": 15767 + }, + { + "epoch": 1.01, + "learning_rate": 1.6638184589621934e-05, + "loss": 0.9516, + "step": 15768 + }, + { + "epoch": 1.01, + "learning_rate": 1.6637969900598986e-05, + "loss": 0.9237, + "step": 15769 + }, + { + "epoch": 1.01, + "learning_rate": 1.6637755211576035e-05, + "loss": 1.0258, + "step": 15770 + }, + { + "epoch": 1.01, + "learning_rate": 1.6637540522553084e-05, + "loss": 0.9423, + "step": 15771 + }, + { + "epoch": 1.01, + "learning_rate": 1.6637325833530133e-05, + "loss": 1.1101, + "step": 15772 + }, + { + "epoch": 1.01, + "learning_rate": 1.6637111144507182e-05, + "loss": 0.9916, + "step": 15773 + }, + { + "epoch": 1.01, + "learning_rate": 1.663689645548423e-05, + "loss": 0.9778, + "step": 15774 + }, + { + "epoch": 1.01, + "learning_rate": 1.6636681766461284e-05, + "loss": 0.9626, + "step": 15775 + }, + { + "epoch": 1.01, + "learning_rate": 1.6636467077438333e-05, + "loss": 0.9593, + "step": 15776 + }, + { + "epoch": 1.02, + "learning_rate": 1.663625238841538e-05, + "loss": 0.9674, + "step": 15777 + }, + { + "epoch": 1.02, + "learning_rate": 1.663603769939243e-05, + "loss": 1.0183, + "step": 15778 + }, + { + "epoch": 1.02, + "learning_rate": 1.663582301036948e-05, + "loss": 0.9502, + "step": 15779 + }, + { + "epoch": 1.02, + "learning_rate": 1.6635608321346532e-05, + "loss": 1.019, + "step": 15780 + }, + { + "epoch": 1.02, + "learning_rate": 1.663539363232358e-05, + "loss": 0.9606, + "step": 15781 + }, + { + "epoch": 1.02, + "learning_rate": 1.663517894330063e-05, + "loss": 0.9997, + "step": 15782 + }, + { + "epoch": 1.02, + "learning_rate": 1.6634964254277682e-05, + "loss": 0.9547, + "step": 15783 + }, + { + "epoch": 1.02, + "learning_rate": 1.663474956525473e-05, + "loss": 0.9307, + "step": 15784 + }, + { + "epoch": 1.02, + "learning_rate": 1.663453487623178e-05, + "loss": 1.0405, + "step": 15785 + }, + { + "epoch": 1.02, + "learning_rate": 1.663432018720883e-05, + "loss": 0.9818, + "step": 15786 + }, + { + "epoch": 1.02, + "learning_rate": 1.6634105498185878e-05, + "loss": 0.9059, + "step": 15787 + }, + { + "epoch": 1.02, + "learning_rate": 1.663389080916293e-05, + "loss": 0.9468, + "step": 15788 + }, + { + "epoch": 1.02, + "learning_rate": 1.663367612013998e-05, + "loss": 1.0344, + "step": 15789 + }, + { + "epoch": 1.02, + "learning_rate": 1.6633461431117028e-05, + "loss": 0.9389, + "step": 15790 + }, + { + "epoch": 1.02, + "learning_rate": 1.6633246742094077e-05, + "loss": 0.936, + "step": 15791 + }, + { + "epoch": 1.02, + "learning_rate": 1.6633032053071126e-05, + "loss": 1.0106, + "step": 15792 + }, + { + "epoch": 1.02, + "learning_rate": 1.663281736404818e-05, + "loss": 0.9561, + "step": 15793 + }, + { + "epoch": 1.02, + "learning_rate": 1.6632602675025227e-05, + "loss": 0.9674, + "step": 15794 + }, + { + "epoch": 1.02, + "learning_rate": 1.6632387986002276e-05, + "loss": 0.9124, + "step": 15795 + }, + { + "epoch": 1.02, + "learning_rate": 1.6632173296979325e-05, + "loss": 1.006, + "step": 15796 + }, + { + "epoch": 1.02, + "learning_rate": 1.6631958607956374e-05, + "loss": 0.988, + "step": 15797 + }, + { + "epoch": 1.02, + "learning_rate": 1.6631743918933427e-05, + "loss": 0.9727, + "step": 15798 + }, + { + "epoch": 1.02, + "learning_rate": 1.6631529229910476e-05, + "loss": 0.9762, + "step": 15799 + }, + { + "epoch": 1.02, + "learning_rate": 1.6631314540887528e-05, + "loss": 0.9538, + "step": 15800 + }, + { + "epoch": 1.02, + "learning_rate": 1.6631099851864577e-05, + "loss": 0.9801, + "step": 15801 + }, + { + "epoch": 1.02, + "learning_rate": 1.6630885162841626e-05, + "loss": 1.0103, + "step": 15802 + }, + { + "epoch": 1.02, + "learning_rate": 1.6630670473818675e-05, + "loss": 0.9368, + "step": 15803 + }, + { + "epoch": 1.02, + "learning_rate": 1.6630455784795724e-05, + "loss": 0.9181, + "step": 15804 + }, + { + "epoch": 1.02, + "learning_rate": 1.6630241095772773e-05, + "loss": 1.0244, + "step": 15805 + }, + { + "epoch": 1.02, + "learning_rate": 1.6630026406749825e-05, + "loss": 0.9454, + "step": 15806 + }, + { + "epoch": 1.02, + "learning_rate": 1.6629811717726874e-05, + "loss": 0.9338, + "step": 15807 + }, + { + "epoch": 1.02, + "learning_rate": 1.6629597028703923e-05, + "loss": 0.9334, + "step": 15808 + }, + { + "epoch": 1.02, + "learning_rate": 1.6629382339680972e-05, + "loss": 1.0097, + "step": 15809 + }, + { + "epoch": 1.02, + "learning_rate": 1.662916765065802e-05, + "loss": 0.9853, + "step": 15810 + }, + { + "epoch": 1.02, + "learning_rate": 1.6628952961635073e-05, + "loss": 0.9741, + "step": 15811 + }, + { + "epoch": 1.02, + "learning_rate": 1.6628738272612122e-05, + "loss": 0.988, + "step": 15812 + }, + { + "epoch": 1.02, + "learning_rate": 1.6628523583589175e-05, + "loss": 0.9409, + "step": 15813 + }, + { + "epoch": 1.02, + "learning_rate": 1.6628308894566224e-05, + "loss": 0.9527, + "step": 15814 + }, + { + "epoch": 1.02, + "learning_rate": 1.6628094205543273e-05, + "loss": 1.0083, + "step": 15815 + }, + { + "epoch": 1.02, + "learning_rate": 1.662787951652032e-05, + "loss": 0.9154, + "step": 15816 + }, + { + "epoch": 1.02, + "learning_rate": 1.662766482749737e-05, + "loss": 0.9842, + "step": 15817 + }, + { + "epoch": 1.02, + "learning_rate": 1.6627450138474423e-05, + "loss": 0.9883, + "step": 15818 + }, + { + "epoch": 1.02, + "learning_rate": 1.6627235449451472e-05, + "loss": 1.003, + "step": 15819 + }, + { + "epoch": 1.02, + "learning_rate": 1.662702076042852e-05, + "loss": 0.9688, + "step": 15820 + }, + { + "epoch": 1.02, + "learning_rate": 1.662680607140557e-05, + "loss": 1.0329, + "step": 15821 + }, + { + "epoch": 1.02, + "learning_rate": 1.662659138238262e-05, + "loss": 0.899, + "step": 15822 + }, + { + "epoch": 1.02, + "learning_rate": 1.6626376693359668e-05, + "loss": 0.9087, + "step": 15823 + }, + { + "epoch": 1.02, + "learning_rate": 1.662616200433672e-05, + "loss": 1.0233, + "step": 15824 + }, + { + "epoch": 1.02, + "learning_rate": 1.662594731531377e-05, + "loss": 0.9292, + "step": 15825 + }, + { + "epoch": 1.02, + "learning_rate": 1.6625732626290818e-05, + "loss": 0.9996, + "step": 15826 + }, + { + "epoch": 1.02, + "learning_rate": 1.662551793726787e-05, + "loss": 0.9428, + "step": 15827 + }, + { + "epoch": 1.02, + "learning_rate": 1.662530324824492e-05, + "loss": 0.9438, + "step": 15828 + }, + { + "epoch": 1.02, + "learning_rate": 1.662508855922197e-05, + "loss": 0.9798, + "step": 15829 + }, + { + "epoch": 1.02, + "learning_rate": 1.6624873870199017e-05, + "loss": 0.9462, + "step": 15830 + }, + { + "epoch": 1.02, + "learning_rate": 1.662465918117607e-05, + "loss": 1.0114, + "step": 15831 + }, + { + "epoch": 1.02, + "learning_rate": 1.662444449215312e-05, + "loss": 0.9867, + "step": 15832 + }, + { + "epoch": 1.02, + "learning_rate": 1.6624229803130168e-05, + "loss": 0.9733, + "step": 15833 + }, + { + "epoch": 1.02, + "learning_rate": 1.6624015114107217e-05, + "loss": 0.9653, + "step": 15834 + }, + { + "epoch": 1.02, + "learning_rate": 1.6623800425084266e-05, + "loss": 0.9376, + "step": 15835 + }, + { + "epoch": 1.02, + "learning_rate": 1.6623585736061318e-05, + "loss": 0.9066, + "step": 15836 + }, + { + "epoch": 1.02, + "learning_rate": 1.6623371047038367e-05, + "loss": 0.9607, + "step": 15837 + }, + { + "epoch": 1.02, + "learning_rate": 1.6623156358015416e-05, + "loss": 1.0643, + "step": 15838 + }, + { + "epoch": 1.02, + "learning_rate": 1.6622941668992465e-05, + "loss": 0.9877, + "step": 15839 + }, + { + "epoch": 1.02, + "learning_rate": 1.6622726979969514e-05, + "loss": 0.8632, + "step": 15840 + }, + { + "epoch": 1.02, + "learning_rate": 1.6622512290946563e-05, + "loss": 0.9623, + "step": 15841 + }, + { + "epoch": 1.02, + "learning_rate": 1.6622297601923615e-05, + "loss": 0.9216, + "step": 15842 + }, + { + "epoch": 1.02, + "learning_rate": 1.6622082912900664e-05, + "loss": 1.0066, + "step": 15843 + }, + { + "epoch": 1.02, + "learning_rate": 1.6621868223877716e-05, + "loss": 1.0407, + "step": 15844 + }, + { + "epoch": 1.02, + "learning_rate": 1.6621653534854765e-05, + "loss": 1.0128, + "step": 15845 + }, + { + "epoch": 1.02, + "learning_rate": 1.6621438845831814e-05, + "loss": 0.9719, + "step": 15846 + }, + { + "epoch": 1.02, + "learning_rate": 1.6621224156808863e-05, + "loss": 0.932, + "step": 15847 + }, + { + "epoch": 1.02, + "learning_rate": 1.6621009467785912e-05, + "loss": 0.9081, + "step": 15848 + }, + { + "epoch": 1.02, + "learning_rate": 1.6620794778762965e-05, + "loss": 1.0335, + "step": 15849 + }, + { + "epoch": 1.02, + "learning_rate": 1.6620580089740014e-05, + "loss": 0.9499, + "step": 15850 + }, + { + "epoch": 1.02, + "learning_rate": 1.6620365400717063e-05, + "loss": 0.9704, + "step": 15851 + }, + { + "epoch": 1.02, + "learning_rate": 1.662015071169411e-05, + "loss": 0.9341, + "step": 15852 + }, + { + "epoch": 1.02, + "learning_rate": 1.661993602267116e-05, + "loss": 0.9267, + "step": 15853 + }, + { + "epoch": 1.02, + "learning_rate": 1.661972133364821e-05, + "loss": 0.9353, + "step": 15854 + }, + { + "epoch": 1.02, + "learning_rate": 1.6619506644625262e-05, + "loss": 1.0683, + "step": 15855 + }, + { + "epoch": 1.02, + "learning_rate": 1.6619291955602314e-05, + "loss": 0.9703, + "step": 15856 + }, + { + "epoch": 1.02, + "learning_rate": 1.6619077266579363e-05, + "loss": 1.004, + "step": 15857 + }, + { + "epoch": 1.02, + "learning_rate": 1.6618862577556412e-05, + "loss": 0.9687, + "step": 15858 + }, + { + "epoch": 1.02, + "learning_rate": 1.661864788853346e-05, + "loss": 0.9303, + "step": 15859 + }, + { + "epoch": 1.02, + "learning_rate": 1.661843319951051e-05, + "loss": 0.9239, + "step": 15860 + }, + { + "epoch": 1.02, + "learning_rate": 1.661821851048756e-05, + "loss": 0.9183, + "step": 15861 + }, + { + "epoch": 1.02, + "learning_rate": 1.661800382146461e-05, + "loss": 0.9275, + "step": 15862 + }, + { + "epoch": 1.02, + "learning_rate": 1.661778913244166e-05, + "loss": 1.0069, + "step": 15863 + }, + { + "epoch": 1.02, + "learning_rate": 1.661757444341871e-05, + "loss": 1.0409, + "step": 15864 + }, + { + "epoch": 1.02, + "learning_rate": 1.6617359754395758e-05, + "loss": 0.9713, + "step": 15865 + }, + { + "epoch": 1.02, + "learning_rate": 1.6617145065372807e-05, + "loss": 0.9436, + "step": 15866 + }, + { + "epoch": 1.02, + "learning_rate": 1.661693037634986e-05, + "loss": 0.9823, + "step": 15867 + }, + { + "epoch": 1.02, + "learning_rate": 1.661671568732691e-05, + "loss": 0.9233, + "step": 15868 + }, + { + "epoch": 1.02, + "learning_rate": 1.6616500998303957e-05, + "loss": 1.0089, + "step": 15869 + }, + { + "epoch": 1.02, + "learning_rate": 1.6616286309281006e-05, + "loss": 0.9987, + "step": 15870 + }, + { + "epoch": 1.02, + "learning_rate": 1.661607162025806e-05, + "loss": 0.9232, + "step": 15871 + }, + { + "epoch": 1.02, + "learning_rate": 1.6615856931235108e-05, + "loss": 1.0245, + "step": 15872 + }, + { + "epoch": 1.02, + "learning_rate": 1.6615642242212157e-05, + "loss": 0.9302, + "step": 15873 + }, + { + "epoch": 1.02, + "learning_rate": 1.6615427553189206e-05, + "loss": 0.97, + "step": 15874 + }, + { + "epoch": 1.02, + "learning_rate": 1.6615212864166258e-05, + "loss": 0.9135, + "step": 15875 + }, + { + "epoch": 1.02, + "learning_rate": 1.6614998175143307e-05, + "loss": 0.9209, + "step": 15876 + }, + { + "epoch": 1.02, + "learning_rate": 1.6614783486120356e-05, + "loss": 0.9075, + "step": 15877 + }, + { + "epoch": 1.02, + "learning_rate": 1.6614568797097405e-05, + "loss": 1.0208, + "step": 15878 + }, + { + "epoch": 1.02, + "learning_rate": 1.6614354108074454e-05, + "loss": 0.9539, + "step": 15879 + }, + { + "epoch": 1.02, + "learning_rate": 1.6614139419051506e-05, + "loss": 1.0317, + "step": 15880 + }, + { + "epoch": 1.02, + "learning_rate": 1.6613924730028555e-05, + "loss": 0.9237, + "step": 15881 + }, + { + "epoch": 1.02, + "learning_rate": 1.6613710041005604e-05, + "loss": 0.9643, + "step": 15882 + }, + { + "epoch": 1.02, + "learning_rate": 1.6613495351982653e-05, + "loss": 0.95, + "step": 15883 + }, + { + "epoch": 1.02, + "learning_rate": 1.6613280662959702e-05, + "loss": 0.9882, + "step": 15884 + }, + { + "epoch": 1.02, + "learning_rate": 1.6613065973936754e-05, + "loss": 0.9932, + "step": 15885 + }, + { + "epoch": 1.02, + "learning_rate": 1.6612851284913803e-05, + "loss": 0.9477, + "step": 15886 + }, + { + "epoch": 1.02, + "learning_rate": 1.6612636595890856e-05, + "loss": 0.9844, + "step": 15887 + }, + { + "epoch": 1.02, + "learning_rate": 1.6612421906867905e-05, + "loss": 0.9908, + "step": 15888 + }, + { + "epoch": 1.02, + "learning_rate": 1.6612207217844954e-05, + "loss": 0.9686, + "step": 15889 + }, + { + "epoch": 1.02, + "learning_rate": 1.6611992528822003e-05, + "loss": 0.9792, + "step": 15890 + }, + { + "epoch": 1.02, + "learning_rate": 1.661177783979905e-05, + "loss": 0.9644, + "step": 15891 + }, + { + "epoch": 1.02, + "learning_rate": 1.66115631507761e-05, + "loss": 0.9518, + "step": 15892 + }, + { + "epoch": 1.02, + "learning_rate": 1.6611348461753153e-05, + "loss": 0.9773, + "step": 15893 + }, + { + "epoch": 1.02, + "learning_rate": 1.6611133772730202e-05, + "loss": 0.9231, + "step": 15894 + }, + { + "epoch": 1.02, + "learning_rate": 1.661091908370725e-05, + "loss": 0.9093, + "step": 15895 + }, + { + "epoch": 1.02, + "learning_rate": 1.66107043946843e-05, + "loss": 0.9551, + "step": 15896 + }, + { + "epoch": 1.02, + "learning_rate": 1.661048970566135e-05, + "loss": 0.983, + "step": 15897 + }, + { + "epoch": 1.02, + "learning_rate": 1.66102750166384e-05, + "loss": 0.981, + "step": 15898 + }, + { + "epoch": 1.02, + "learning_rate": 1.661006032761545e-05, + "loss": 1.0095, + "step": 15899 + }, + { + "epoch": 1.02, + "learning_rate": 1.6609845638592502e-05, + "loss": 0.9324, + "step": 15900 + }, + { + "epoch": 1.02, + "learning_rate": 1.660963094956955e-05, + "loss": 1.0361, + "step": 15901 + }, + { + "epoch": 1.02, + "learning_rate": 1.66094162605466e-05, + "loss": 1.0672, + "step": 15902 + }, + { + "epoch": 1.02, + "learning_rate": 1.660920157152365e-05, + "loss": 0.919, + "step": 15903 + }, + { + "epoch": 1.02, + "learning_rate": 1.66089868825007e-05, + "loss": 0.8507, + "step": 15904 + }, + { + "epoch": 1.02, + "learning_rate": 1.660877219347775e-05, + "loss": 0.903, + "step": 15905 + }, + { + "epoch": 1.02, + "learning_rate": 1.66085575044548e-05, + "loss": 0.9924, + "step": 15906 + }, + { + "epoch": 1.02, + "learning_rate": 1.660834281543185e-05, + "loss": 0.9953, + "step": 15907 + }, + { + "epoch": 1.02, + "learning_rate": 1.6608128126408898e-05, + "loss": 1.0027, + "step": 15908 + }, + { + "epoch": 1.02, + "learning_rate": 1.6607913437385947e-05, + "loss": 1.0046, + "step": 15909 + }, + { + "epoch": 1.02, + "learning_rate": 1.6607698748362995e-05, + "loss": 0.9561, + "step": 15910 + }, + { + "epoch": 1.02, + "learning_rate": 1.6607484059340048e-05, + "loss": 0.9735, + "step": 15911 + }, + { + "epoch": 1.02, + "learning_rate": 1.6607269370317097e-05, + "loss": 0.9689, + "step": 15912 + }, + { + "epoch": 1.02, + "learning_rate": 1.6607054681294146e-05, + "loss": 0.98, + "step": 15913 + }, + { + "epoch": 1.02, + "learning_rate": 1.6606839992271198e-05, + "loss": 0.8999, + "step": 15914 + }, + { + "epoch": 1.02, + "learning_rate": 1.6606625303248247e-05, + "loss": 0.9642, + "step": 15915 + }, + { + "epoch": 1.02, + "learning_rate": 1.6606410614225296e-05, + "loss": 1.0643, + "step": 15916 + }, + { + "epoch": 1.02, + "learning_rate": 1.6606195925202345e-05, + "loss": 0.9594, + "step": 15917 + }, + { + "epoch": 1.02, + "learning_rate": 1.6605981236179397e-05, + "loss": 1.0838, + "step": 15918 + }, + { + "epoch": 1.02, + "learning_rate": 1.6605766547156446e-05, + "loss": 0.9178, + "step": 15919 + }, + { + "epoch": 1.02, + "learning_rate": 1.6605551858133495e-05, + "loss": 0.9637, + "step": 15920 + }, + { + "epoch": 1.02, + "learning_rate": 1.6605337169110544e-05, + "loss": 1.0445, + "step": 15921 + }, + { + "epoch": 1.02, + "learning_rate": 1.6605122480087593e-05, + "loss": 1.0253, + "step": 15922 + }, + { + "epoch": 1.02, + "learning_rate": 1.6604907791064646e-05, + "loss": 0.9729, + "step": 15923 + }, + { + "epoch": 1.02, + "learning_rate": 1.6604693102041695e-05, + "loss": 1.0405, + "step": 15924 + }, + { + "epoch": 1.02, + "learning_rate": 1.6604478413018744e-05, + "loss": 0.9634, + "step": 15925 + }, + { + "epoch": 1.02, + "learning_rate": 1.6604263723995792e-05, + "loss": 1.0312, + "step": 15926 + }, + { + "epoch": 1.02, + "learning_rate": 1.660404903497284e-05, + "loss": 0.9839, + "step": 15927 + }, + { + "epoch": 1.02, + "learning_rate": 1.660383434594989e-05, + "loss": 0.8671, + "step": 15928 + }, + { + "epoch": 1.02, + "learning_rate": 1.6603619656926943e-05, + "loss": 1.0239, + "step": 15929 + }, + { + "epoch": 1.02, + "learning_rate": 1.6603404967903992e-05, + "loss": 0.9996, + "step": 15930 + }, + { + "epoch": 1.02, + "learning_rate": 1.6603190278881044e-05, + "loss": 0.9634, + "step": 15931 + }, + { + "epoch": 1.03, + "learning_rate": 1.6602975589858093e-05, + "loss": 1.0294, + "step": 15932 + }, + { + "epoch": 1.03, + "learning_rate": 1.6602760900835142e-05, + "loss": 1.0106, + "step": 15933 + }, + { + "epoch": 1.03, + "learning_rate": 1.660254621181219e-05, + "loss": 0.9778, + "step": 15934 + }, + { + "epoch": 1.03, + "learning_rate": 1.660233152278924e-05, + "loss": 1.0819, + "step": 15935 + }, + { + "epoch": 1.03, + "learning_rate": 1.6602116833766292e-05, + "loss": 0.9254, + "step": 15936 + }, + { + "epoch": 1.03, + "learning_rate": 1.660190214474334e-05, + "loss": 0.9463, + "step": 15937 + }, + { + "epoch": 1.03, + "learning_rate": 1.660168745572039e-05, + "loss": 0.9753, + "step": 15938 + }, + { + "epoch": 1.03, + "learning_rate": 1.660147276669744e-05, + "loss": 0.9027, + "step": 15939 + }, + { + "epoch": 1.03, + "learning_rate": 1.6601258077674488e-05, + "loss": 1.0079, + "step": 15940 + }, + { + "epoch": 1.03, + "learning_rate": 1.6601043388651537e-05, + "loss": 0.9838, + "step": 15941 + }, + { + "epoch": 1.03, + "learning_rate": 1.660082869962859e-05, + "loss": 0.9491, + "step": 15942 + }, + { + "epoch": 1.03, + "learning_rate": 1.6600614010605642e-05, + "loss": 0.9675, + "step": 15943 + }, + { + "epoch": 1.03, + "learning_rate": 1.660039932158269e-05, + "loss": 0.9651, + "step": 15944 + }, + { + "epoch": 1.03, + "learning_rate": 1.660018463255974e-05, + "loss": 0.9588, + "step": 15945 + }, + { + "epoch": 1.03, + "learning_rate": 1.659996994353679e-05, + "loss": 0.9687, + "step": 15946 + }, + { + "epoch": 1.03, + "learning_rate": 1.6599755254513838e-05, + "loss": 1.0452, + "step": 15947 + }, + { + "epoch": 1.03, + "learning_rate": 1.6599540565490887e-05, + "loss": 0.9661, + "step": 15948 + }, + { + "epoch": 1.03, + "learning_rate": 1.659932587646794e-05, + "loss": 1.0391, + "step": 15949 + }, + { + "epoch": 1.03, + "learning_rate": 1.6599111187444988e-05, + "loss": 0.9809, + "step": 15950 + }, + { + "epoch": 1.03, + "learning_rate": 1.6598896498422037e-05, + "loss": 0.9362, + "step": 15951 + }, + { + "epoch": 1.03, + "learning_rate": 1.6598681809399086e-05, + "loss": 0.9888, + "step": 15952 + }, + { + "epoch": 1.03, + "learning_rate": 1.6598467120376135e-05, + "loss": 0.9666, + "step": 15953 + }, + { + "epoch": 1.03, + "learning_rate": 1.6598252431353187e-05, + "loss": 1.0343, + "step": 15954 + }, + { + "epoch": 1.03, + "learning_rate": 1.6598037742330236e-05, + "loss": 1.0003, + "step": 15955 + }, + { + "epoch": 1.03, + "learning_rate": 1.6597823053307285e-05, + "loss": 0.9129, + "step": 15956 + }, + { + "epoch": 1.03, + "learning_rate": 1.6597608364284334e-05, + "loss": 0.9076, + "step": 15957 + }, + { + "epoch": 1.03, + "learning_rate": 1.6597393675261386e-05, + "loss": 0.9728, + "step": 15958 + }, + { + "epoch": 1.03, + "learning_rate": 1.6597178986238435e-05, + "loss": 1.0078, + "step": 15959 + }, + { + "epoch": 1.03, + "learning_rate": 1.6596964297215484e-05, + "loss": 1.0235, + "step": 15960 + }, + { + "epoch": 1.03, + "learning_rate": 1.6596749608192533e-05, + "loss": 0.9126, + "step": 15961 + }, + { + "epoch": 1.03, + "learning_rate": 1.6596534919169586e-05, + "loss": 1.0064, + "step": 15962 + }, + { + "epoch": 1.03, + "learning_rate": 1.6596320230146635e-05, + "loss": 0.9777, + "step": 15963 + }, + { + "epoch": 1.03, + "learning_rate": 1.6596105541123684e-05, + "loss": 0.9742, + "step": 15964 + }, + { + "epoch": 1.03, + "learning_rate": 1.6595890852100733e-05, + "loss": 1.0065, + "step": 15965 + }, + { + "epoch": 1.03, + "learning_rate": 1.659567616307778e-05, + "loss": 0.9692, + "step": 15966 + }, + { + "epoch": 1.03, + "learning_rate": 1.6595461474054834e-05, + "loss": 0.9952, + "step": 15967 + }, + { + "epoch": 1.03, + "learning_rate": 1.6595246785031883e-05, + "loss": 0.9641, + "step": 15968 + }, + { + "epoch": 1.03, + "learning_rate": 1.6595032096008932e-05, + "loss": 0.9205, + "step": 15969 + }, + { + "epoch": 1.03, + "learning_rate": 1.659481740698598e-05, + "loss": 1.0278, + "step": 15970 + }, + { + "epoch": 1.03, + "learning_rate": 1.659460271796303e-05, + "loss": 0.9767, + "step": 15971 + }, + { + "epoch": 1.03, + "learning_rate": 1.6594388028940082e-05, + "loss": 0.9442, + "step": 15972 + }, + { + "epoch": 1.03, + "learning_rate": 1.659417333991713e-05, + "loss": 0.9312, + "step": 15973 + }, + { + "epoch": 1.03, + "learning_rate": 1.6593958650894183e-05, + "loss": 1.0026, + "step": 15974 + }, + { + "epoch": 1.03, + "learning_rate": 1.6593743961871232e-05, + "loss": 1.0111, + "step": 15975 + }, + { + "epoch": 1.03, + "learning_rate": 1.659352927284828e-05, + "loss": 0.9012, + "step": 15976 + }, + { + "epoch": 1.03, + "learning_rate": 1.659331458382533e-05, + "loss": 1.0724, + "step": 15977 + }, + { + "epoch": 1.03, + "learning_rate": 1.659309989480238e-05, + "loss": 0.9523, + "step": 15978 + }, + { + "epoch": 1.03, + "learning_rate": 1.6592885205779428e-05, + "loss": 0.9729, + "step": 15979 + }, + { + "epoch": 1.03, + "learning_rate": 1.659267051675648e-05, + "loss": 1.0186, + "step": 15980 + }, + { + "epoch": 1.03, + "learning_rate": 1.659245582773353e-05, + "loss": 0.9791, + "step": 15981 + }, + { + "epoch": 1.03, + "learning_rate": 1.659224113871058e-05, + "loss": 0.9534, + "step": 15982 + }, + { + "epoch": 1.03, + "learning_rate": 1.6592026449687628e-05, + "loss": 0.9518, + "step": 15983 + }, + { + "epoch": 1.03, + "learning_rate": 1.6591811760664676e-05, + "loss": 0.9038, + "step": 15984 + }, + { + "epoch": 1.03, + "learning_rate": 1.659159707164173e-05, + "loss": 1.0152, + "step": 15985 + }, + { + "epoch": 1.03, + "learning_rate": 1.6591382382618778e-05, + "loss": 0.9151, + "step": 15986 + }, + { + "epoch": 1.03, + "learning_rate": 1.659116769359583e-05, + "loss": 1.0247, + "step": 15987 + }, + { + "epoch": 1.03, + "learning_rate": 1.659095300457288e-05, + "loss": 1.0036, + "step": 15988 + }, + { + "epoch": 1.03, + "learning_rate": 1.6590738315549928e-05, + "loss": 0.9521, + "step": 15989 + }, + { + "epoch": 1.03, + "learning_rate": 1.6590523626526977e-05, + "loss": 1.0447, + "step": 15990 + }, + { + "epoch": 1.03, + "learning_rate": 1.6590308937504026e-05, + "loss": 1.0081, + "step": 15991 + }, + { + "epoch": 1.03, + "learning_rate": 1.659009424848108e-05, + "loss": 1.0161, + "step": 15992 + }, + { + "epoch": 1.03, + "learning_rate": 1.6589879559458127e-05, + "loss": 0.9233, + "step": 15993 + }, + { + "epoch": 1.03, + "learning_rate": 1.6589664870435176e-05, + "loss": 0.9756, + "step": 15994 + }, + { + "epoch": 1.03, + "learning_rate": 1.6589450181412225e-05, + "loss": 0.9625, + "step": 15995 + }, + { + "epoch": 1.03, + "learning_rate": 1.6589235492389274e-05, + "loss": 0.9686, + "step": 15996 + }, + { + "epoch": 1.03, + "learning_rate": 1.6589020803366323e-05, + "loss": 0.9598, + "step": 15997 + }, + { + "epoch": 1.03, + "learning_rate": 1.6588806114343376e-05, + "loss": 0.9628, + "step": 15998 + }, + { + "epoch": 1.03, + "learning_rate": 1.6588591425320425e-05, + "loss": 0.9671, + "step": 15999 + }, + { + "epoch": 1.03, + "learning_rate": 1.6588376736297473e-05, + "loss": 0.9654, + "step": 16000 + }, + { + "epoch": 1.03, + "eval_laion/cider": 0.5356252433344917, + "eval_laion/meteor": 0.3974098913169987, + "eval_laion/pred_num_words": 37.876953125, + "eval_laion/sacrebleu": 15.685272536812162, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.064367651939392, + "eval_runtime": 98.0263, + "eval_samples_per_second": 5.09, + "eval_steps_per_second": 0.326, + "step": 16000 + }, + { + "epoch": 1.03, + "learning_rate": 1.6588162047274526e-05, + "loss": 0.9371, + "step": 16001 + }, + { + "epoch": 1.03, + "learning_rate": 1.6587947358251575e-05, + "loss": 0.9698, + "step": 16002 + }, + { + "epoch": 1.03, + "learning_rate": 1.6587732669228624e-05, + "loss": 0.9163, + "step": 16003 + }, + { + "epoch": 1.03, + "learning_rate": 1.6587517980205673e-05, + "loss": 0.9715, + "step": 16004 + }, + { + "epoch": 1.03, + "learning_rate": 1.6587303291182725e-05, + "loss": 1.0283, + "step": 16005 + }, + { + "epoch": 1.03, + "learning_rate": 1.6587088602159774e-05, + "loss": 1.0147, + "step": 16006 + }, + { + "epoch": 1.03, + "learning_rate": 1.6586873913136823e-05, + "loss": 0.995, + "step": 16007 + }, + { + "epoch": 1.03, + "learning_rate": 1.6586659224113872e-05, + "loss": 0.9554, + "step": 16008 + }, + { + "epoch": 1.03, + "learning_rate": 1.658644453509092e-05, + "loss": 0.8964, + "step": 16009 + }, + { + "epoch": 1.03, + "learning_rate": 1.658622984606797e-05, + "loss": 0.9834, + "step": 16010 + }, + { + "epoch": 1.03, + "learning_rate": 1.6586015157045022e-05, + "loss": 1.0153, + "step": 16011 + }, + { + "epoch": 1.03, + "learning_rate": 1.658580046802207e-05, + "loss": 0.9956, + "step": 16012 + }, + { + "epoch": 1.03, + "learning_rate": 1.658558577899912e-05, + "loss": 0.9738, + "step": 16013 + }, + { + "epoch": 1.03, + "learning_rate": 1.658537108997617e-05, + "loss": 1.0098, + "step": 16014 + }, + { + "epoch": 1.03, + "learning_rate": 1.6585156400953218e-05, + "loss": 0.9254, + "step": 16015 + }, + { + "epoch": 1.03, + "learning_rate": 1.658494171193027e-05, + "loss": 0.9607, + "step": 16016 + }, + { + "epoch": 1.03, + "learning_rate": 1.658472702290732e-05, + "loss": 1.0397, + "step": 16017 + }, + { + "epoch": 1.03, + "learning_rate": 1.6584512333884372e-05, + "loss": 0.9526, + "step": 16018 + }, + { + "epoch": 1.03, + "learning_rate": 1.658429764486142e-05, + "loss": 0.9563, + "step": 16019 + }, + { + "epoch": 1.03, + "learning_rate": 1.658408295583847e-05, + "loss": 0.8799, + "step": 16020 + }, + { + "epoch": 1.03, + "learning_rate": 1.658386826681552e-05, + "loss": 0.8916, + "step": 16021 + }, + { + "epoch": 1.03, + "learning_rate": 1.6583653577792568e-05, + "loss": 0.9761, + "step": 16022 + }, + { + "epoch": 1.03, + "learning_rate": 1.658343888876962e-05, + "loss": 0.9436, + "step": 16023 + }, + { + "epoch": 1.03, + "learning_rate": 1.658322419974667e-05, + "loss": 0.96, + "step": 16024 + }, + { + "epoch": 1.03, + "learning_rate": 1.6583009510723718e-05, + "loss": 0.9626, + "step": 16025 + }, + { + "epoch": 1.03, + "learning_rate": 1.6582794821700767e-05, + "loss": 0.9635, + "step": 16026 + }, + { + "epoch": 1.03, + "learning_rate": 1.6582580132677816e-05, + "loss": 0.9215, + "step": 16027 + }, + { + "epoch": 1.03, + "learning_rate": 1.6582365443654865e-05, + "loss": 0.8432, + "step": 16028 + }, + { + "epoch": 1.03, + "learning_rate": 1.6582150754631917e-05, + "loss": 0.8951, + "step": 16029 + }, + { + "epoch": 1.03, + "learning_rate": 1.658193606560897e-05, + "loss": 0.9792, + "step": 16030 + }, + { + "epoch": 1.03, + "learning_rate": 1.658172137658602e-05, + "loss": 1.0534, + "step": 16031 + }, + { + "epoch": 1.03, + "learning_rate": 1.6581506687563067e-05, + "loss": 0.9647, + "step": 16032 + }, + { + "epoch": 1.03, + "learning_rate": 1.6581291998540116e-05, + "loss": 0.9112, + "step": 16033 + }, + { + "epoch": 1.03, + "learning_rate": 1.6581077309517165e-05, + "loss": 0.9454, + "step": 16034 + }, + { + "epoch": 1.03, + "learning_rate": 1.6580862620494214e-05, + "loss": 0.9673, + "step": 16035 + }, + { + "epoch": 1.03, + "learning_rate": 1.6580647931471267e-05, + "loss": 1.0177, + "step": 16036 + }, + { + "epoch": 1.03, + "learning_rate": 1.6580433242448316e-05, + "loss": 0.9733, + "step": 16037 + }, + { + "epoch": 1.03, + "learning_rate": 1.6580218553425365e-05, + "loss": 0.9996, + "step": 16038 + }, + { + "epoch": 1.03, + "learning_rate": 1.6580003864402414e-05, + "loss": 0.87, + "step": 16039 + }, + { + "epoch": 1.03, + "learning_rate": 1.6579789175379463e-05, + "loss": 0.9513, + "step": 16040 + }, + { + "epoch": 1.03, + "learning_rate": 1.6579574486356515e-05, + "loss": 0.9424, + "step": 16041 + }, + { + "epoch": 1.03, + "learning_rate": 1.6579359797333564e-05, + "loss": 1.0065, + "step": 16042 + }, + { + "epoch": 1.03, + "learning_rate": 1.6579145108310613e-05, + "loss": 0.8902, + "step": 16043 + }, + { + "epoch": 1.03, + "learning_rate": 1.6578930419287662e-05, + "loss": 0.9781, + "step": 16044 + }, + { + "epoch": 1.03, + "learning_rate": 1.6578715730264714e-05, + "loss": 0.9566, + "step": 16045 + }, + { + "epoch": 1.03, + "learning_rate": 1.6578501041241763e-05, + "loss": 0.9553, + "step": 16046 + }, + { + "epoch": 1.03, + "learning_rate": 1.6578286352218812e-05, + "loss": 1.0204, + "step": 16047 + }, + { + "epoch": 1.03, + "learning_rate": 1.657807166319586e-05, + "loss": 0.9883, + "step": 16048 + }, + { + "epoch": 1.03, + "learning_rate": 1.6577856974172913e-05, + "loss": 1.0378, + "step": 16049 + }, + { + "epoch": 1.03, + "learning_rate": 1.6577642285149962e-05, + "loss": 1.1019, + "step": 16050 + }, + { + "epoch": 1.03, + "learning_rate": 1.657742759612701e-05, + "loss": 1.0236, + "step": 16051 + }, + { + "epoch": 1.03, + "learning_rate": 1.657721290710406e-05, + "loss": 0.9512, + "step": 16052 + }, + { + "epoch": 1.03, + "learning_rate": 1.657699821808111e-05, + "loss": 1.0464, + "step": 16053 + }, + { + "epoch": 1.03, + "learning_rate": 1.657678352905816e-05, + "loss": 0.944, + "step": 16054 + }, + { + "epoch": 1.03, + "learning_rate": 1.657656884003521e-05, + "loss": 0.8579, + "step": 16055 + }, + { + "epoch": 1.03, + "learning_rate": 1.657635415101226e-05, + "loss": 0.9316, + "step": 16056 + }, + { + "epoch": 1.03, + "learning_rate": 1.657613946198931e-05, + "loss": 1.034, + "step": 16057 + }, + { + "epoch": 1.03, + "learning_rate": 1.6575924772966357e-05, + "loss": 0.9611, + "step": 16058 + }, + { + "epoch": 1.03, + "learning_rate": 1.657571008394341e-05, + "loss": 0.9784, + "step": 16059 + }, + { + "epoch": 1.03, + "learning_rate": 1.657549539492046e-05, + "loss": 1.0064, + "step": 16060 + }, + { + "epoch": 1.03, + "learning_rate": 1.657528070589751e-05, + "loss": 0.985, + "step": 16061 + }, + { + "epoch": 1.03, + "learning_rate": 1.657506601687456e-05, + "loss": 0.9736, + "step": 16062 + }, + { + "epoch": 1.03, + "learning_rate": 1.657485132785161e-05, + "loss": 0.9503, + "step": 16063 + }, + { + "epoch": 1.03, + "learning_rate": 1.6574636638828658e-05, + "loss": 0.9194, + "step": 16064 + }, + { + "epoch": 1.03, + "learning_rate": 1.6574421949805707e-05, + "loss": 0.9408, + "step": 16065 + }, + { + "epoch": 1.03, + "learning_rate": 1.6574207260782756e-05, + "loss": 1.007, + "step": 16066 + }, + { + "epoch": 1.03, + "learning_rate": 1.657399257175981e-05, + "loss": 0.9799, + "step": 16067 + }, + { + "epoch": 1.03, + "learning_rate": 1.6573777882736857e-05, + "loss": 1.0267, + "step": 16068 + }, + { + "epoch": 1.03, + "learning_rate": 1.6573563193713906e-05, + "loss": 0.9607, + "step": 16069 + }, + { + "epoch": 1.03, + "learning_rate": 1.6573348504690955e-05, + "loss": 1.0455, + "step": 16070 + }, + { + "epoch": 1.03, + "learning_rate": 1.6573133815668004e-05, + "loss": 1.0714, + "step": 16071 + }, + { + "epoch": 1.03, + "learning_rate": 1.6572919126645057e-05, + "loss": 0.9604, + "step": 16072 + }, + { + "epoch": 1.03, + "learning_rate": 1.6572704437622106e-05, + "loss": 0.9933, + "step": 16073 + }, + { + "epoch": 1.03, + "learning_rate": 1.6572489748599158e-05, + "loss": 1.0441, + "step": 16074 + }, + { + "epoch": 1.03, + "learning_rate": 1.6572275059576207e-05, + "loss": 1.0621, + "step": 16075 + }, + { + "epoch": 1.03, + "learning_rate": 1.6572060370553256e-05, + "loss": 0.9803, + "step": 16076 + }, + { + "epoch": 1.03, + "learning_rate": 1.6571845681530305e-05, + "loss": 0.9574, + "step": 16077 + }, + { + "epoch": 1.03, + "learning_rate": 1.6571630992507354e-05, + "loss": 0.9609, + "step": 16078 + }, + { + "epoch": 1.03, + "learning_rate": 1.6571416303484406e-05, + "loss": 0.8975, + "step": 16079 + }, + { + "epoch": 1.03, + "learning_rate": 1.6571201614461455e-05, + "loss": 0.9069, + "step": 16080 + }, + { + "epoch": 1.03, + "learning_rate": 1.6570986925438504e-05, + "loss": 0.9529, + "step": 16081 + }, + { + "epoch": 1.03, + "learning_rate": 1.6570772236415553e-05, + "loss": 0.9303, + "step": 16082 + }, + { + "epoch": 1.03, + "learning_rate": 1.6570557547392602e-05, + "loss": 0.9867, + "step": 16083 + }, + { + "epoch": 1.03, + "learning_rate": 1.657034285836965e-05, + "loss": 1.0673, + "step": 16084 + }, + { + "epoch": 1.03, + "learning_rate": 1.6570128169346703e-05, + "loss": 1.1185, + "step": 16085 + }, + { + "epoch": 1.03, + "learning_rate": 1.6569913480323752e-05, + "loss": 0.9692, + "step": 16086 + }, + { + "epoch": 1.03, + "learning_rate": 1.65696987913008e-05, + "loss": 1.0183, + "step": 16087 + }, + { + "epoch": 1.04, + "learning_rate": 1.6569484102277854e-05, + "loss": 0.9804, + "step": 16088 + }, + { + "epoch": 1.04, + "learning_rate": 1.6569269413254903e-05, + "loss": 0.977, + "step": 16089 + }, + { + "epoch": 1.04, + "learning_rate": 1.656905472423195e-05, + "loss": 1.0279, + "step": 16090 + }, + { + "epoch": 1.04, + "learning_rate": 1.6568840035209e-05, + "loss": 1.0016, + "step": 16091 + }, + { + "epoch": 1.04, + "learning_rate": 1.6568625346186053e-05, + "loss": 1.0669, + "step": 16092 + }, + { + "epoch": 1.04, + "learning_rate": 1.6568410657163102e-05, + "loss": 0.9929, + "step": 16093 + }, + { + "epoch": 1.04, + "learning_rate": 1.656819596814015e-05, + "loss": 1.0296, + "step": 16094 + }, + { + "epoch": 1.04, + "learning_rate": 1.65679812791172e-05, + "loss": 0.9956, + "step": 16095 + }, + { + "epoch": 1.04, + "learning_rate": 1.656776659009425e-05, + "loss": 0.9408, + "step": 16096 + }, + { + "epoch": 1.04, + "learning_rate": 1.6567551901071298e-05, + "loss": 1.0381, + "step": 16097 + }, + { + "epoch": 1.04, + "learning_rate": 1.656733721204835e-05, + "loss": 0.94, + "step": 16098 + }, + { + "epoch": 1.04, + "learning_rate": 1.65671225230254e-05, + "loss": 1.0063, + "step": 16099 + }, + { + "epoch": 1.04, + "learning_rate": 1.6566907834002448e-05, + "loss": 0.9661, + "step": 16100 + }, + { + "epoch": 1.04, + "learning_rate": 1.6566693144979497e-05, + "loss": 1.03, + "step": 16101 + }, + { + "epoch": 1.04, + "learning_rate": 1.6566478455956546e-05, + "loss": 0.9226, + "step": 16102 + }, + { + "epoch": 1.04, + "learning_rate": 1.6566263766933598e-05, + "loss": 0.9893, + "step": 16103 + }, + { + "epoch": 1.04, + "learning_rate": 1.6566049077910647e-05, + "loss": 0.8837, + "step": 16104 + }, + { + "epoch": 1.04, + "learning_rate": 1.65658343888877e-05, + "loss": 0.9958, + "step": 16105 + }, + { + "epoch": 1.04, + "learning_rate": 1.656561969986475e-05, + "loss": 0.9787, + "step": 16106 + }, + { + "epoch": 1.04, + "learning_rate": 1.6565405010841797e-05, + "loss": 1.0786, + "step": 16107 + }, + { + "epoch": 1.04, + "learning_rate": 1.6565190321818846e-05, + "loss": 1.0223, + "step": 16108 + }, + { + "epoch": 1.04, + "learning_rate": 1.6564975632795895e-05, + "loss": 0.9601, + "step": 16109 + }, + { + "epoch": 1.04, + "learning_rate": 1.6564760943772948e-05, + "loss": 1.0161, + "step": 16110 + }, + { + "epoch": 1.04, + "learning_rate": 1.6564546254749997e-05, + "loss": 0.9156, + "step": 16111 + }, + { + "epoch": 1.04, + "learning_rate": 1.6564331565727046e-05, + "loss": 0.9662, + "step": 16112 + }, + { + "epoch": 1.04, + "learning_rate": 1.6564116876704095e-05, + "loss": 0.9294, + "step": 16113 + }, + { + "epoch": 1.04, + "learning_rate": 1.6563902187681144e-05, + "loss": 1.0333, + "step": 16114 + }, + { + "epoch": 1.04, + "learning_rate": 1.6563687498658193e-05, + "loss": 0.9681, + "step": 16115 + }, + { + "epoch": 1.04, + "learning_rate": 1.6563472809635245e-05, + "loss": 1.014, + "step": 16116 + }, + { + "epoch": 1.04, + "learning_rate": 1.6563258120612294e-05, + "loss": 0.9427, + "step": 16117 + }, + { + "epoch": 1.04, + "learning_rate": 1.6563043431589346e-05, + "loss": 1.039, + "step": 16118 + }, + { + "epoch": 1.04, + "learning_rate": 1.6562828742566395e-05, + "loss": 0.918, + "step": 16119 + }, + { + "epoch": 1.04, + "learning_rate": 1.6562614053543444e-05, + "loss": 0.9596, + "step": 16120 + }, + { + "epoch": 1.04, + "learning_rate": 1.6562399364520493e-05, + "loss": 1.1433, + "step": 16121 + }, + { + "epoch": 1.04, + "learning_rate": 1.6562184675497542e-05, + "loss": 0.9625, + "step": 16122 + }, + { + "epoch": 1.04, + "learning_rate": 1.6561969986474594e-05, + "loss": 0.9639, + "step": 16123 + }, + { + "epoch": 1.04, + "learning_rate": 1.6561755297451643e-05, + "loss": 1.0061, + "step": 16124 + }, + { + "epoch": 1.04, + "learning_rate": 1.6561540608428692e-05, + "loss": 0.9599, + "step": 16125 + }, + { + "epoch": 1.04, + "learning_rate": 1.656132591940574e-05, + "loss": 1.0206, + "step": 16126 + }, + { + "epoch": 1.04, + "learning_rate": 1.656111123038279e-05, + "loss": 1.0047, + "step": 16127 + }, + { + "epoch": 1.04, + "learning_rate": 1.6560896541359843e-05, + "loss": 0.9126, + "step": 16128 + }, + { + "epoch": 1.04, + "learning_rate": 1.656068185233689e-05, + "loss": 0.9722, + "step": 16129 + }, + { + "epoch": 1.04, + "learning_rate": 1.656046716331394e-05, + "loss": 1.0064, + "step": 16130 + }, + { + "epoch": 1.04, + "learning_rate": 1.656025247429099e-05, + "loss": 1.0056, + "step": 16131 + }, + { + "epoch": 1.04, + "learning_rate": 1.6560037785268042e-05, + "loss": 0.8574, + "step": 16132 + }, + { + "epoch": 1.04, + "learning_rate": 1.655982309624509e-05, + "loss": 0.9589, + "step": 16133 + }, + { + "epoch": 1.04, + "learning_rate": 1.655960840722214e-05, + "loss": 0.9678, + "step": 16134 + }, + { + "epoch": 1.04, + "learning_rate": 1.655939371819919e-05, + "loss": 0.9663, + "step": 16135 + }, + { + "epoch": 1.04, + "learning_rate": 1.655917902917624e-05, + "loss": 1.0123, + "step": 16136 + }, + { + "epoch": 1.04, + "learning_rate": 1.655896434015329e-05, + "loss": 0.9271, + "step": 16137 + }, + { + "epoch": 1.04, + "learning_rate": 1.655874965113034e-05, + "loss": 0.9773, + "step": 16138 + }, + { + "epoch": 1.04, + "learning_rate": 1.6558534962107388e-05, + "loss": 0.8796, + "step": 16139 + }, + { + "epoch": 1.04, + "learning_rate": 1.6558320273084437e-05, + "loss": 1.0159, + "step": 16140 + }, + { + "epoch": 1.04, + "learning_rate": 1.655810558406149e-05, + "loss": 0.9398, + "step": 16141 + }, + { + "epoch": 1.04, + "learning_rate": 1.655789089503854e-05, + "loss": 0.9309, + "step": 16142 + }, + { + "epoch": 1.04, + "learning_rate": 1.6557676206015587e-05, + "loss": 0.9645, + "step": 16143 + }, + { + "epoch": 1.04, + "learning_rate": 1.6557461516992636e-05, + "loss": 0.9537, + "step": 16144 + }, + { + "epoch": 1.04, + "learning_rate": 1.6557246827969685e-05, + "loss": 0.9866, + "step": 16145 + }, + { + "epoch": 1.04, + "learning_rate": 1.6557032138946738e-05, + "loss": 0.9328, + "step": 16146 + }, + { + "epoch": 1.04, + "learning_rate": 1.6556817449923787e-05, + "loss": 0.969, + "step": 16147 + }, + { + "epoch": 1.04, + "learning_rate": 1.655660276090084e-05, + "loss": 0.9456, + "step": 16148 + }, + { + "epoch": 1.04, + "learning_rate": 1.6556388071877888e-05, + "loss": 0.9129, + "step": 16149 + }, + { + "epoch": 1.04, + "learning_rate": 1.6556173382854937e-05, + "loss": 1.0429, + "step": 16150 + }, + { + "epoch": 1.04, + "learning_rate": 1.6555958693831986e-05, + "loss": 0.9613, + "step": 16151 + }, + { + "epoch": 1.04, + "learning_rate": 1.6555744004809035e-05, + "loss": 1.0064, + "step": 16152 + }, + { + "epoch": 1.04, + "learning_rate": 1.6555529315786084e-05, + "loss": 0.983, + "step": 16153 + }, + { + "epoch": 1.04, + "learning_rate": 1.6555314626763136e-05, + "loss": 0.9552, + "step": 16154 + }, + { + "epoch": 1.04, + "learning_rate": 1.6555099937740185e-05, + "loss": 0.9787, + "step": 16155 + }, + { + "epoch": 1.04, + "learning_rate": 1.6554885248717234e-05, + "loss": 1.0172, + "step": 16156 + }, + { + "epoch": 1.04, + "learning_rate": 1.6554670559694283e-05, + "loss": 0.9181, + "step": 16157 + }, + { + "epoch": 1.04, + "learning_rate": 1.6554455870671332e-05, + "loss": 1.03, + "step": 16158 + }, + { + "epoch": 1.04, + "learning_rate": 1.6554241181648384e-05, + "loss": 1.049, + "step": 16159 + }, + { + "epoch": 1.04, + "learning_rate": 1.6554026492625433e-05, + "loss": 1.0105, + "step": 16160 + }, + { + "epoch": 1.04, + "learning_rate": 1.6553811803602486e-05, + "loss": 0.9633, + "step": 16161 + }, + { + "epoch": 1.04, + "learning_rate": 1.6553597114579535e-05, + "loss": 0.9549, + "step": 16162 + }, + { + "epoch": 1.04, + "learning_rate": 1.6553382425556584e-05, + "loss": 0.9702, + "step": 16163 + }, + { + "epoch": 1.04, + "learning_rate": 1.6553167736533632e-05, + "loss": 0.9717, + "step": 16164 + }, + { + "epoch": 1.04, + "learning_rate": 1.655295304751068e-05, + "loss": 1.0031, + "step": 16165 + }, + { + "epoch": 1.04, + "learning_rate": 1.6552738358487734e-05, + "loss": 0.9929, + "step": 16166 + }, + { + "epoch": 1.04, + "learning_rate": 1.6552523669464783e-05, + "loss": 0.9735, + "step": 16167 + }, + { + "epoch": 1.04, + "learning_rate": 1.6552308980441832e-05, + "loss": 0.9501, + "step": 16168 + }, + { + "epoch": 1.04, + "learning_rate": 1.655209429141888e-05, + "loss": 0.9904, + "step": 16169 + }, + { + "epoch": 1.04, + "learning_rate": 1.655187960239593e-05, + "loss": 1.0269, + "step": 16170 + }, + { + "epoch": 1.04, + "learning_rate": 1.655166491337298e-05, + "loss": 0.9853, + "step": 16171 + }, + { + "epoch": 1.04, + "learning_rate": 1.655145022435003e-05, + "loss": 1.0278, + "step": 16172 + }, + { + "epoch": 1.04, + "learning_rate": 1.655123553532708e-05, + "loss": 1.0361, + "step": 16173 + }, + { + "epoch": 1.04, + "learning_rate": 1.655102084630413e-05, + "loss": 0.9594, + "step": 16174 + }, + { + "epoch": 1.04, + "learning_rate": 1.655080615728118e-05, + "loss": 1.032, + "step": 16175 + }, + { + "epoch": 1.04, + "learning_rate": 1.655059146825823e-05, + "loss": 0.9905, + "step": 16176 + }, + { + "epoch": 1.04, + "learning_rate": 1.655037677923528e-05, + "loss": 0.9616, + "step": 16177 + }, + { + "epoch": 1.04, + "learning_rate": 1.6550162090212328e-05, + "loss": 1.0063, + "step": 16178 + }, + { + "epoch": 1.04, + "learning_rate": 1.654994740118938e-05, + "loss": 0.9464, + "step": 16179 + }, + { + "epoch": 1.04, + "learning_rate": 1.654973271216643e-05, + "loss": 1.0293, + "step": 16180 + }, + { + "epoch": 1.04, + "learning_rate": 1.654951802314348e-05, + "loss": 0.9715, + "step": 16181 + }, + { + "epoch": 1.04, + "learning_rate": 1.6549303334120527e-05, + "loss": 0.9646, + "step": 16182 + }, + { + "epoch": 1.04, + "learning_rate": 1.6549088645097576e-05, + "loss": 0.9289, + "step": 16183 + }, + { + "epoch": 1.04, + "learning_rate": 1.6548873956074625e-05, + "loss": 0.9967, + "step": 16184 + }, + { + "epoch": 1.04, + "learning_rate": 1.6548659267051678e-05, + "loss": 0.997, + "step": 16185 + }, + { + "epoch": 1.04, + "learning_rate": 1.6548444578028727e-05, + "loss": 0.967, + "step": 16186 + }, + { + "epoch": 1.04, + "learning_rate": 1.6548229889005776e-05, + "loss": 1.0066, + "step": 16187 + }, + { + "epoch": 1.04, + "learning_rate": 1.6548015199982825e-05, + "loss": 0.9568, + "step": 16188 + }, + { + "epoch": 1.04, + "learning_rate": 1.6547800510959874e-05, + "loss": 1.0299, + "step": 16189 + }, + { + "epoch": 1.04, + "learning_rate": 1.6547585821936926e-05, + "loss": 0.9931, + "step": 16190 + }, + { + "epoch": 1.04, + "learning_rate": 1.6547371132913975e-05, + "loss": 0.951, + "step": 16191 + }, + { + "epoch": 1.04, + "learning_rate": 1.6547156443891027e-05, + "loss": 0.9794, + "step": 16192 + }, + { + "epoch": 1.04, + "learning_rate": 1.6546941754868076e-05, + "loss": 0.9736, + "step": 16193 + }, + { + "epoch": 1.04, + "learning_rate": 1.6546727065845125e-05, + "loss": 1.0205, + "step": 16194 + }, + { + "epoch": 1.04, + "learning_rate": 1.6546512376822174e-05, + "loss": 0.9449, + "step": 16195 + }, + { + "epoch": 1.04, + "learning_rate": 1.6546297687799223e-05, + "loss": 0.9413, + "step": 16196 + }, + { + "epoch": 1.04, + "learning_rate": 1.6546082998776275e-05, + "loss": 0.9879, + "step": 16197 + }, + { + "epoch": 1.04, + "learning_rate": 1.6545868309753324e-05, + "loss": 0.9885, + "step": 16198 + }, + { + "epoch": 1.04, + "learning_rate": 1.6545653620730373e-05, + "loss": 0.9579, + "step": 16199 + }, + { + "epoch": 1.04, + "learning_rate": 1.6545438931707422e-05, + "loss": 0.9616, + "step": 16200 + }, + { + "epoch": 1.04, + "learning_rate": 1.654522424268447e-05, + "loss": 0.9202, + "step": 16201 + }, + { + "epoch": 1.04, + "learning_rate": 1.654500955366152e-05, + "loss": 0.9689, + "step": 16202 + }, + { + "epoch": 1.04, + "learning_rate": 1.6544794864638573e-05, + "loss": 1.0281, + "step": 16203 + }, + { + "epoch": 1.04, + "learning_rate": 1.654458017561562e-05, + "loss": 0.9187, + "step": 16204 + }, + { + "epoch": 1.04, + "learning_rate": 1.6544365486592674e-05, + "loss": 0.9665, + "step": 16205 + }, + { + "epoch": 1.04, + "learning_rate": 1.6544150797569723e-05, + "loss": 0.9602, + "step": 16206 + }, + { + "epoch": 1.04, + "learning_rate": 1.6543936108546772e-05, + "loss": 0.9575, + "step": 16207 + }, + { + "epoch": 1.04, + "learning_rate": 1.654372141952382e-05, + "loss": 1.0335, + "step": 16208 + }, + { + "epoch": 1.04, + "learning_rate": 1.654350673050087e-05, + "loss": 1.0479, + "step": 16209 + }, + { + "epoch": 1.04, + "learning_rate": 1.6543292041477922e-05, + "loss": 0.9948, + "step": 16210 + }, + { + "epoch": 1.04, + "learning_rate": 1.654307735245497e-05, + "loss": 1.0086, + "step": 16211 + }, + { + "epoch": 1.04, + "learning_rate": 1.654286266343202e-05, + "loss": 0.95, + "step": 16212 + }, + { + "epoch": 1.04, + "learning_rate": 1.654264797440907e-05, + "loss": 0.9845, + "step": 16213 + }, + { + "epoch": 1.04, + "learning_rate": 1.6542433285386118e-05, + "loss": 0.9571, + "step": 16214 + }, + { + "epoch": 1.04, + "learning_rate": 1.654221859636317e-05, + "loss": 0.9984, + "step": 16215 + }, + { + "epoch": 1.04, + "learning_rate": 1.654200390734022e-05, + "loss": 0.986, + "step": 16216 + }, + { + "epoch": 1.04, + "learning_rate": 1.6541789218317268e-05, + "loss": 0.9396, + "step": 16217 + }, + { + "epoch": 1.04, + "learning_rate": 1.6541574529294317e-05, + "loss": 0.9817, + "step": 16218 + }, + { + "epoch": 1.04, + "learning_rate": 1.654135984027137e-05, + "loss": 0.946, + "step": 16219 + }, + { + "epoch": 1.04, + "learning_rate": 1.654114515124842e-05, + "loss": 0.9688, + "step": 16220 + }, + { + "epoch": 1.04, + "learning_rate": 1.6540930462225468e-05, + "loss": 0.8839, + "step": 16221 + }, + { + "epoch": 1.04, + "learning_rate": 1.6540715773202516e-05, + "loss": 0.962, + "step": 16222 + }, + { + "epoch": 1.04, + "learning_rate": 1.654050108417957e-05, + "loss": 0.9995, + "step": 16223 + }, + { + "epoch": 1.04, + "learning_rate": 1.6540286395156618e-05, + "loss": 0.9493, + "step": 16224 + }, + { + "epoch": 1.04, + "learning_rate": 1.6540071706133667e-05, + "loss": 0.9629, + "step": 16225 + }, + { + "epoch": 1.04, + "learning_rate": 1.6539857017110716e-05, + "loss": 0.9758, + "step": 16226 + }, + { + "epoch": 1.04, + "learning_rate": 1.6539642328087765e-05, + "loss": 1.0037, + "step": 16227 + }, + { + "epoch": 1.04, + "learning_rate": 1.6539427639064817e-05, + "loss": 1.0145, + "step": 16228 + }, + { + "epoch": 1.04, + "learning_rate": 1.6539212950041866e-05, + "loss": 0.981, + "step": 16229 + }, + { + "epoch": 1.04, + "learning_rate": 1.6538998261018915e-05, + "loss": 1.0118, + "step": 16230 + }, + { + "epoch": 1.04, + "learning_rate": 1.6538783571995964e-05, + "loss": 1.0032, + "step": 16231 + }, + { + "epoch": 1.04, + "learning_rate": 1.6538568882973013e-05, + "loss": 1.0396, + "step": 16232 + }, + { + "epoch": 1.04, + "learning_rate": 1.6538354193950065e-05, + "loss": 0.9937, + "step": 16233 + }, + { + "epoch": 1.04, + "learning_rate": 1.6538139504927114e-05, + "loss": 1.0045, + "step": 16234 + }, + { + "epoch": 1.04, + "learning_rate": 1.6537924815904167e-05, + "loss": 1.0251, + "step": 16235 + }, + { + "epoch": 1.04, + "learning_rate": 1.6537710126881216e-05, + "loss": 1.0025, + "step": 16236 + }, + { + "epoch": 1.04, + "learning_rate": 1.6537495437858265e-05, + "loss": 0.993, + "step": 16237 + }, + { + "epoch": 1.04, + "learning_rate": 1.6537280748835313e-05, + "loss": 0.9967, + "step": 16238 + }, + { + "epoch": 1.04, + "learning_rate": 1.6537066059812362e-05, + "loss": 0.9977, + "step": 16239 + }, + { + "epoch": 1.04, + "learning_rate": 1.653685137078941e-05, + "loss": 0.9306, + "step": 16240 + }, + { + "epoch": 1.04, + "learning_rate": 1.6536636681766464e-05, + "loss": 1.0529, + "step": 16241 + }, + { + "epoch": 1.04, + "learning_rate": 1.6536421992743513e-05, + "loss": 1.0004, + "step": 16242 + }, + { + "epoch": 1.05, + "learning_rate": 1.6536207303720562e-05, + "loss": 1.0596, + "step": 16243 + }, + { + "epoch": 1.05, + "learning_rate": 1.653599261469761e-05, + "loss": 1.0005, + "step": 16244 + }, + { + "epoch": 1.05, + "learning_rate": 1.653577792567466e-05, + "loss": 0.974, + "step": 16245 + }, + { + "epoch": 1.05, + "learning_rate": 1.6535563236651712e-05, + "loss": 0.9959, + "step": 16246 + }, + { + "epoch": 1.05, + "learning_rate": 1.653534854762876e-05, + "loss": 0.9664, + "step": 16247 + }, + { + "epoch": 1.05, + "learning_rate": 1.6535133858605813e-05, + "loss": 0.9637, + "step": 16248 + }, + { + "epoch": 1.05, + "learning_rate": 1.6534919169582862e-05, + "loss": 1.0285, + "step": 16249 + }, + { + "epoch": 1.05, + "learning_rate": 1.653470448055991e-05, + "loss": 0.9599, + "step": 16250 + }, + { + "epoch": 1.05, + "learning_rate": 1.653448979153696e-05, + "loss": 0.9937, + "step": 16251 + }, + { + "epoch": 1.05, + "learning_rate": 1.653427510251401e-05, + "loss": 0.9564, + "step": 16252 + }, + { + "epoch": 1.05, + "learning_rate": 1.653406041349106e-05, + "loss": 0.9777, + "step": 16253 + }, + { + "epoch": 1.05, + "learning_rate": 1.653384572446811e-05, + "loss": 0.885, + "step": 16254 + }, + { + "epoch": 1.05, + "learning_rate": 1.653363103544516e-05, + "loss": 1.0084, + "step": 16255 + }, + { + "epoch": 1.05, + "learning_rate": 1.653341634642221e-05, + "loss": 0.9757, + "step": 16256 + }, + { + "epoch": 1.05, + "learning_rate": 1.6533201657399257e-05, + "loss": 1.0776, + "step": 16257 + }, + { + "epoch": 1.05, + "learning_rate": 1.6532986968376306e-05, + "loss": 1.0182, + "step": 16258 + }, + { + "epoch": 1.05, + "learning_rate": 1.653277227935336e-05, + "loss": 0.9813, + "step": 16259 + }, + { + "epoch": 1.05, + "learning_rate": 1.6532557590330408e-05, + "loss": 0.9994, + "step": 16260 + }, + { + "epoch": 1.05, + "learning_rate": 1.6532342901307457e-05, + "loss": 1.0109, + "step": 16261 + }, + { + "epoch": 1.05, + "learning_rate": 1.6532128212284506e-05, + "loss": 0.9979, + "step": 16262 + }, + { + "epoch": 1.05, + "learning_rate": 1.6531913523261558e-05, + "loss": 0.8916, + "step": 16263 + }, + { + "epoch": 1.05, + "learning_rate": 1.6531698834238607e-05, + "loss": 0.9783, + "step": 16264 + }, + { + "epoch": 1.05, + "learning_rate": 1.6531484145215656e-05, + "loss": 1.0243, + "step": 16265 + }, + { + "epoch": 1.05, + "learning_rate": 1.6531269456192708e-05, + "loss": 1.0788, + "step": 16266 + }, + { + "epoch": 1.05, + "learning_rate": 1.6531054767169757e-05, + "loss": 0.9986, + "step": 16267 + }, + { + "epoch": 1.05, + "learning_rate": 1.6530840078146806e-05, + "loss": 1.0515, + "step": 16268 + }, + { + "epoch": 1.05, + "learning_rate": 1.6530625389123855e-05, + "loss": 0.9534, + "step": 16269 + }, + { + "epoch": 1.05, + "learning_rate": 1.6530410700100904e-05, + "loss": 0.989, + "step": 16270 + }, + { + "epoch": 1.05, + "learning_rate": 1.6530196011077953e-05, + "loss": 0.9798, + "step": 16271 + }, + { + "epoch": 1.05, + "learning_rate": 1.6529981322055005e-05, + "loss": 1.0088, + "step": 16272 + }, + { + "epoch": 1.05, + "learning_rate": 1.6529766633032054e-05, + "loss": 0.9299, + "step": 16273 + }, + { + "epoch": 1.05, + "learning_rate": 1.6529551944009103e-05, + "loss": 0.9881, + "step": 16274 + }, + { + "epoch": 1.05, + "learning_rate": 1.6529337254986152e-05, + "loss": 0.9017, + "step": 16275 + }, + { + "epoch": 1.05, + "learning_rate": 1.65291225659632e-05, + "loss": 1.0542, + "step": 16276 + }, + { + "epoch": 1.05, + "learning_rate": 1.6528907876940254e-05, + "loss": 0.9313, + "step": 16277 + }, + { + "epoch": 1.05, + "learning_rate": 1.6528693187917303e-05, + "loss": 0.9377, + "step": 16278 + }, + { + "epoch": 1.05, + "learning_rate": 1.6528478498894355e-05, + "loss": 0.9533, + "step": 16279 + }, + { + "epoch": 1.05, + "learning_rate": 1.6528263809871404e-05, + "loss": 0.9652, + "step": 16280 + }, + { + "epoch": 1.05, + "learning_rate": 1.6528049120848453e-05, + "loss": 0.9986, + "step": 16281 + }, + { + "epoch": 1.05, + "learning_rate": 1.6527834431825502e-05, + "loss": 0.9566, + "step": 16282 + }, + { + "epoch": 1.05, + "learning_rate": 1.652761974280255e-05, + "loss": 0.9857, + "step": 16283 + }, + { + "epoch": 1.05, + "learning_rate": 1.6527405053779603e-05, + "loss": 1.019, + "step": 16284 + }, + { + "epoch": 1.05, + "learning_rate": 1.6527190364756652e-05, + "loss": 0.985, + "step": 16285 + }, + { + "epoch": 1.05, + "learning_rate": 1.65269756757337e-05, + "loss": 0.927, + "step": 16286 + }, + { + "epoch": 1.05, + "learning_rate": 1.652676098671075e-05, + "loss": 0.9562, + "step": 16287 + }, + { + "epoch": 1.05, + "learning_rate": 1.65265462976878e-05, + "loss": 0.9646, + "step": 16288 + }, + { + "epoch": 1.05, + "learning_rate": 1.6526331608664848e-05, + "loss": 0.9392, + "step": 16289 + }, + { + "epoch": 1.05, + "learning_rate": 1.65261169196419e-05, + "loss": 0.9049, + "step": 16290 + }, + { + "epoch": 1.05, + "learning_rate": 1.652590223061895e-05, + "loss": 1.0203, + "step": 16291 + }, + { + "epoch": 1.05, + "learning_rate": 1.6525687541596e-05, + "loss": 0.9633, + "step": 16292 + }, + { + "epoch": 1.05, + "learning_rate": 1.652547285257305e-05, + "loss": 0.9158, + "step": 16293 + }, + { + "epoch": 1.05, + "learning_rate": 1.65252581635501e-05, + "loss": 0.9772, + "step": 16294 + }, + { + "epoch": 1.05, + "learning_rate": 1.652504347452715e-05, + "loss": 0.9566, + "step": 16295 + }, + { + "epoch": 1.05, + "learning_rate": 1.6524828785504197e-05, + "loss": 0.9678, + "step": 16296 + }, + { + "epoch": 1.05, + "learning_rate": 1.652461409648125e-05, + "loss": 0.9856, + "step": 16297 + }, + { + "epoch": 1.05, + "learning_rate": 1.65243994074583e-05, + "loss": 1.0078, + "step": 16298 + }, + { + "epoch": 1.05, + "learning_rate": 1.6524184718435348e-05, + "loss": 0.9898, + "step": 16299 + }, + { + "epoch": 1.05, + "learning_rate": 1.6523970029412397e-05, + "loss": 0.9255, + "step": 16300 + }, + { + "epoch": 1.05, + "learning_rate": 1.6523755340389446e-05, + "loss": 1.0072, + "step": 16301 + }, + { + "epoch": 1.05, + "learning_rate": 1.6523540651366498e-05, + "loss": 0.9366, + "step": 16302 + }, + { + "epoch": 1.05, + "learning_rate": 1.6523325962343547e-05, + "loss": 1.0338, + "step": 16303 + }, + { + "epoch": 1.05, + "learning_rate": 1.6523111273320596e-05, + "loss": 0.9296, + "step": 16304 + }, + { + "epoch": 1.05, + "learning_rate": 1.6522896584297645e-05, + "loss": 1.0033, + "step": 16305 + }, + { + "epoch": 1.05, + "learning_rate": 1.6522681895274697e-05, + "loss": 0.9484, + "step": 16306 + }, + { + "epoch": 1.05, + "learning_rate": 1.6522467206251746e-05, + "loss": 0.9861, + "step": 16307 + }, + { + "epoch": 1.05, + "learning_rate": 1.6522252517228795e-05, + "loss": 0.8964, + "step": 16308 + }, + { + "epoch": 1.05, + "learning_rate": 1.6522037828205844e-05, + "loss": 0.8916, + "step": 16309 + }, + { + "epoch": 1.05, + "learning_rate": 1.6521823139182897e-05, + "loss": 1.0268, + "step": 16310 + }, + { + "epoch": 1.05, + "learning_rate": 1.6521608450159946e-05, + "loss": 1.0127, + "step": 16311 + }, + { + "epoch": 1.05, + "learning_rate": 1.6521393761136994e-05, + "loss": 0.9676, + "step": 16312 + }, + { + "epoch": 1.05, + "learning_rate": 1.6521179072114043e-05, + "loss": 0.996, + "step": 16313 + }, + { + "epoch": 1.05, + "learning_rate": 1.6520964383091092e-05, + "loss": 0.9663, + "step": 16314 + }, + { + "epoch": 1.05, + "learning_rate": 1.6520749694068145e-05, + "loss": 0.9662, + "step": 16315 + }, + { + "epoch": 1.05, + "learning_rate": 1.6520535005045194e-05, + "loss": 0.9406, + "step": 16316 + }, + { + "epoch": 1.05, + "learning_rate": 1.6520320316022243e-05, + "loss": 0.8966, + "step": 16317 + }, + { + "epoch": 1.05, + "learning_rate": 1.652010562699929e-05, + "loss": 0.9902, + "step": 16318 + }, + { + "epoch": 1.05, + "learning_rate": 1.651989093797634e-05, + "loss": 1.0125, + "step": 16319 + }, + { + "epoch": 1.05, + "learning_rate": 1.651967624895339e-05, + "loss": 1.0163, + "step": 16320 + }, + { + "epoch": 1.05, + "learning_rate": 1.6519461559930442e-05, + "loss": 0.9023, + "step": 16321 + }, + { + "epoch": 1.05, + "learning_rate": 1.6519246870907494e-05, + "loss": 1.0305, + "step": 16322 + }, + { + "epoch": 1.05, + "learning_rate": 1.6519032181884543e-05, + "loss": 0.9109, + "step": 16323 + }, + { + "epoch": 1.05, + "learning_rate": 1.6518817492861592e-05, + "loss": 0.9808, + "step": 16324 + }, + { + "epoch": 1.05, + "learning_rate": 1.651860280383864e-05, + "loss": 1.0418, + "step": 16325 + }, + { + "epoch": 1.05, + "learning_rate": 1.651838811481569e-05, + "loss": 0.9481, + "step": 16326 + }, + { + "epoch": 1.05, + "learning_rate": 1.651817342579274e-05, + "loss": 0.9492, + "step": 16327 + }, + { + "epoch": 1.05, + "learning_rate": 1.651795873676979e-05, + "loss": 0.9725, + "step": 16328 + }, + { + "epoch": 1.05, + "learning_rate": 1.651774404774684e-05, + "loss": 0.961, + "step": 16329 + }, + { + "epoch": 1.05, + "learning_rate": 1.651752935872389e-05, + "loss": 0.947, + "step": 16330 + }, + { + "epoch": 1.05, + "learning_rate": 1.651731466970094e-05, + "loss": 1.011, + "step": 16331 + }, + { + "epoch": 1.05, + "learning_rate": 1.6517099980677987e-05, + "loss": 0.9287, + "step": 16332 + }, + { + "epoch": 1.05, + "learning_rate": 1.651688529165504e-05, + "loss": 0.9826, + "step": 16333 + }, + { + "epoch": 1.05, + "learning_rate": 1.651667060263209e-05, + "loss": 0.9936, + "step": 16334 + }, + { + "epoch": 1.05, + "learning_rate": 1.651645591360914e-05, + "loss": 0.9197, + "step": 16335 + }, + { + "epoch": 1.05, + "learning_rate": 1.651624122458619e-05, + "loss": 0.9847, + "step": 16336 + }, + { + "epoch": 1.05, + "learning_rate": 1.651602653556324e-05, + "loss": 0.9968, + "step": 16337 + }, + { + "epoch": 1.05, + "learning_rate": 1.6515811846540288e-05, + "loss": 0.9922, + "step": 16338 + }, + { + "epoch": 1.05, + "learning_rate": 1.6515597157517337e-05, + "loss": 0.9205, + "step": 16339 + }, + { + "epoch": 1.05, + "learning_rate": 1.651538246849439e-05, + "loss": 0.9455, + "step": 16340 + }, + { + "epoch": 1.05, + "learning_rate": 1.6515167779471438e-05, + "loss": 0.9376, + "step": 16341 + }, + { + "epoch": 1.05, + "learning_rate": 1.6514953090448487e-05, + "loss": 1.0059, + "step": 16342 + }, + { + "epoch": 1.05, + "learning_rate": 1.6514738401425536e-05, + "loss": 1.0081, + "step": 16343 + }, + { + "epoch": 1.05, + "learning_rate": 1.6514523712402585e-05, + "loss": 1.0238, + "step": 16344 + }, + { + "epoch": 1.05, + "learning_rate": 1.6514309023379634e-05, + "loss": 0.9132, + "step": 16345 + }, + { + "epoch": 1.05, + "learning_rate": 1.6514094334356686e-05, + "loss": 0.912, + "step": 16346 + }, + { + "epoch": 1.05, + "learning_rate": 1.6513879645333735e-05, + "loss": 0.9649, + "step": 16347 + }, + { + "epoch": 1.05, + "learning_rate": 1.6513664956310784e-05, + "loss": 0.9721, + "step": 16348 + }, + { + "epoch": 1.05, + "learning_rate": 1.6513450267287833e-05, + "loss": 0.9776, + "step": 16349 + }, + { + "epoch": 1.05, + "learning_rate": 1.6513235578264886e-05, + "loss": 1.0824, + "step": 16350 + }, + { + "epoch": 1.05, + "learning_rate": 1.6513020889241935e-05, + "loss": 1.0175, + "step": 16351 + }, + { + "epoch": 1.05, + "learning_rate": 1.6512806200218984e-05, + "loss": 1.061, + "step": 16352 + }, + { + "epoch": 1.05, + "learning_rate": 1.6512591511196036e-05, + "loss": 1.0638, + "step": 16353 + }, + { + "epoch": 1.05, + "learning_rate": 1.6512376822173085e-05, + "loss": 0.9443, + "step": 16354 + }, + { + "epoch": 1.05, + "learning_rate": 1.6512162133150134e-05, + "loss": 0.9741, + "step": 16355 + }, + { + "epoch": 1.05, + "learning_rate": 1.6511947444127183e-05, + "loss": 0.9158, + "step": 16356 + }, + { + "epoch": 1.05, + "learning_rate": 1.6511732755104232e-05, + "loss": 0.9379, + "step": 16357 + }, + { + "epoch": 1.05, + "learning_rate": 1.651151806608128e-05, + "loss": 0.9186, + "step": 16358 + }, + { + "epoch": 1.05, + "learning_rate": 1.6511303377058333e-05, + "loss": 1.0496, + "step": 16359 + }, + { + "epoch": 1.05, + "learning_rate": 1.6511088688035382e-05, + "loss": 1.0741, + "step": 16360 + }, + { + "epoch": 1.05, + "learning_rate": 1.651087399901243e-05, + "loss": 0.9713, + "step": 16361 + }, + { + "epoch": 1.05, + "learning_rate": 1.651065930998948e-05, + "loss": 1.0199, + "step": 16362 + }, + { + "epoch": 1.05, + "learning_rate": 1.651044462096653e-05, + "loss": 1.0333, + "step": 16363 + }, + { + "epoch": 1.05, + "learning_rate": 1.651022993194358e-05, + "loss": 0.9356, + "step": 16364 + }, + { + "epoch": 1.05, + "learning_rate": 1.651001524292063e-05, + "loss": 1.0475, + "step": 16365 + }, + { + "epoch": 1.05, + "learning_rate": 1.6509800553897683e-05, + "loss": 0.9994, + "step": 16366 + }, + { + "epoch": 1.05, + "learning_rate": 1.650958586487473e-05, + "loss": 0.9838, + "step": 16367 + }, + { + "epoch": 1.05, + "learning_rate": 1.650937117585178e-05, + "loss": 0.9986, + "step": 16368 + }, + { + "epoch": 1.05, + "learning_rate": 1.650915648682883e-05, + "loss": 0.9368, + "step": 16369 + }, + { + "epoch": 1.05, + "learning_rate": 1.650894179780588e-05, + "loss": 1.1065, + "step": 16370 + }, + { + "epoch": 1.05, + "learning_rate": 1.650872710878293e-05, + "loss": 0.9134, + "step": 16371 + }, + { + "epoch": 1.05, + "learning_rate": 1.650851241975998e-05, + "loss": 0.9538, + "step": 16372 + }, + { + "epoch": 1.05, + "learning_rate": 1.650829773073703e-05, + "loss": 0.9715, + "step": 16373 + }, + { + "epoch": 1.05, + "learning_rate": 1.6508083041714078e-05, + "loss": 0.9543, + "step": 16374 + }, + { + "epoch": 1.05, + "learning_rate": 1.6507868352691127e-05, + "loss": 0.9849, + "step": 16375 + }, + { + "epoch": 1.05, + "learning_rate": 1.6507653663668176e-05, + "loss": 1.0154, + "step": 16376 + }, + { + "epoch": 1.05, + "learning_rate": 1.6507438974645228e-05, + "loss": 0.9493, + "step": 16377 + }, + { + "epoch": 1.05, + "learning_rate": 1.6507224285622277e-05, + "loss": 0.9936, + "step": 16378 + }, + { + "epoch": 1.05, + "learning_rate": 1.650700959659933e-05, + "loss": 1.0028, + "step": 16379 + }, + { + "epoch": 1.05, + "learning_rate": 1.650679490757638e-05, + "loss": 0.9654, + "step": 16380 + }, + { + "epoch": 1.05, + "learning_rate": 1.6506580218553427e-05, + "loss": 0.976, + "step": 16381 + }, + { + "epoch": 1.05, + "learning_rate": 1.6506365529530476e-05, + "loss": 1.0055, + "step": 16382 + }, + { + "epoch": 1.05, + "learning_rate": 1.6506150840507525e-05, + "loss": 0.9973, + "step": 16383 + }, + { + "epoch": 1.05, + "learning_rate": 1.6505936151484578e-05, + "loss": 0.9882, + "step": 16384 + }, + { + "epoch": 1.05, + "learning_rate": 1.6505721462461627e-05, + "loss": 1.014, + "step": 16385 + }, + { + "epoch": 1.05, + "learning_rate": 1.6505506773438675e-05, + "loss": 1.0206, + "step": 16386 + }, + { + "epoch": 1.05, + "learning_rate": 1.6505292084415724e-05, + "loss": 0.8984, + "step": 16387 + }, + { + "epoch": 1.05, + "learning_rate": 1.6505077395392773e-05, + "loss": 1.0062, + "step": 16388 + }, + { + "epoch": 1.05, + "learning_rate": 1.6504862706369826e-05, + "loss": 1.0265, + "step": 16389 + }, + { + "epoch": 1.05, + "learning_rate": 1.6504648017346875e-05, + "loss": 0.9732, + "step": 16390 + }, + { + "epoch": 1.05, + "learning_rate": 1.6504433328323924e-05, + "loss": 1.0031, + "step": 16391 + }, + { + "epoch": 1.05, + "learning_rate": 1.6504218639300973e-05, + "loss": 1.025, + "step": 16392 + }, + { + "epoch": 1.05, + "learning_rate": 1.6504003950278025e-05, + "loss": 1.0544, + "step": 16393 + }, + { + "epoch": 1.05, + "learning_rate": 1.6503789261255074e-05, + "loss": 0.9037, + "step": 16394 + }, + { + "epoch": 1.05, + "learning_rate": 1.6503574572232123e-05, + "loss": 0.9285, + "step": 16395 + }, + { + "epoch": 1.05, + "learning_rate": 1.6503574572232123e-05, + "loss": 0.9174, + "step": 16396 + }, + { + "epoch": 1.05, + "learning_rate": 1.6503359883209172e-05, + "loss": 1.0571, + "step": 16397 + }, + { + "epoch": 1.06, + "learning_rate": 1.6503145194186224e-05, + "loss": 0.9785, + "step": 16398 + }, + { + "epoch": 1.06, + "learning_rate": 1.6502930505163273e-05, + "loss": 0.93, + "step": 16399 + }, + { + "epoch": 1.06, + "learning_rate": 1.6502715816140322e-05, + "loss": 0.9432, + "step": 16400 + }, + { + "epoch": 1.06, + "learning_rate": 1.650250112711737e-05, + "loss": 0.949, + "step": 16401 + }, + { + "epoch": 1.06, + "learning_rate": 1.650228643809442e-05, + "loss": 0.9732, + "step": 16402 + }, + { + "epoch": 1.06, + "learning_rate": 1.6502071749071472e-05, + "loss": 1.0054, + "step": 16403 + }, + { + "epoch": 1.06, + "learning_rate": 1.650185706004852e-05, + "loss": 0.9385, + "step": 16404 + }, + { + "epoch": 1.06, + "learning_rate": 1.650164237102557e-05, + "loss": 0.9442, + "step": 16405 + }, + { + "epoch": 1.06, + "learning_rate": 1.650142768200262e-05, + "loss": 0.9636, + "step": 16406 + }, + { + "epoch": 1.06, + "learning_rate": 1.650121299297967e-05, + "loss": 0.9332, + "step": 16407 + }, + { + "epoch": 1.06, + "learning_rate": 1.6500998303956717e-05, + "loss": 0.9785, + "step": 16408 + }, + { + "epoch": 1.06, + "learning_rate": 1.650078361493377e-05, + "loss": 0.9678, + "step": 16409 + }, + { + "epoch": 1.06, + "learning_rate": 1.6500568925910822e-05, + "loss": 1.0171, + "step": 16410 + }, + { + "epoch": 1.06, + "learning_rate": 1.650035423688787e-05, + "loss": 1.078, + "step": 16411 + }, + { + "epoch": 1.06, + "learning_rate": 1.650013954786492e-05, + "loss": 0.9309, + "step": 16412 + }, + { + "epoch": 1.06, + "learning_rate": 1.649992485884197e-05, + "loss": 0.9256, + "step": 16413 + }, + { + "epoch": 1.06, + "learning_rate": 1.6499710169819018e-05, + "loss": 1.0169, + "step": 16414 + }, + { + "epoch": 1.06, + "learning_rate": 1.6499495480796067e-05, + "loss": 1.0514, + "step": 16415 + }, + { + "epoch": 1.06, + "learning_rate": 1.649928079177312e-05, + "loss": 0.9987, + "step": 16416 + }, + { + "epoch": 1.06, + "learning_rate": 1.6499066102750168e-05, + "loss": 0.9203, + "step": 16417 + }, + { + "epoch": 1.06, + "learning_rate": 1.6498851413727217e-05, + "loss": 0.993, + "step": 16418 + }, + { + "epoch": 1.06, + "learning_rate": 1.6498636724704266e-05, + "loss": 0.9696, + "step": 16419 + }, + { + "epoch": 1.06, + "learning_rate": 1.6498422035681315e-05, + "loss": 1.0045, + "step": 16420 + }, + { + "epoch": 1.06, + "learning_rate": 1.6498207346658367e-05, + "loss": 1.011, + "step": 16421 + }, + { + "epoch": 1.06, + "learning_rate": 1.6497992657635416e-05, + "loss": 0.9437, + "step": 16422 + }, + { + "epoch": 1.06, + "learning_rate": 1.649777796861247e-05, + "loss": 0.9705, + "step": 16423 + }, + { + "epoch": 1.06, + "learning_rate": 1.6497563279589518e-05, + "loss": 0.8802, + "step": 16424 + }, + { + "epoch": 1.06, + "learning_rate": 1.6497348590566567e-05, + "loss": 0.9321, + "step": 16425 + }, + { + "epoch": 1.06, + "learning_rate": 1.6497133901543616e-05, + "loss": 0.9229, + "step": 16426 + }, + { + "epoch": 1.06, + "learning_rate": 1.6496919212520665e-05, + "loss": 1.0023, + "step": 16427 + }, + { + "epoch": 1.06, + "learning_rate": 1.6496704523497714e-05, + "loss": 0.9573, + "step": 16428 + }, + { + "epoch": 1.06, + "learning_rate": 1.6496489834474766e-05, + "loss": 0.982, + "step": 16429 + }, + { + "epoch": 1.06, + "learning_rate": 1.6496275145451815e-05, + "loss": 0.9919, + "step": 16430 + }, + { + "epoch": 1.06, + "learning_rate": 1.6496060456428864e-05, + "loss": 0.9864, + "step": 16431 + }, + { + "epoch": 1.06, + "learning_rate": 1.6495845767405913e-05, + "loss": 0.9995, + "step": 16432 + }, + { + "epoch": 1.06, + "learning_rate": 1.6495631078382962e-05, + "loss": 1.0618, + "step": 16433 + }, + { + "epoch": 1.06, + "learning_rate": 1.6495416389360014e-05, + "loss": 0.9553, + "step": 16434 + }, + { + "epoch": 1.06, + "learning_rate": 1.6495201700337063e-05, + "loss": 0.9792, + "step": 16435 + }, + { + "epoch": 1.06, + "learning_rate": 1.6494987011314112e-05, + "loss": 0.9685, + "step": 16436 + }, + { + "epoch": 1.06, + "learning_rate": 1.649477232229116e-05, + "loss": 0.9086, + "step": 16437 + }, + { + "epoch": 1.06, + "learning_rate": 1.6494557633268213e-05, + "loss": 0.9912, + "step": 16438 + }, + { + "epoch": 1.06, + "learning_rate": 1.6494342944245262e-05, + "loss": 0.9753, + "step": 16439 + }, + { + "epoch": 1.06, + "learning_rate": 1.649412825522231e-05, + "loss": 1.0453, + "step": 16440 + }, + { + "epoch": 1.06, + "learning_rate": 1.6493913566199364e-05, + "loss": 0.9875, + "step": 16441 + }, + { + "epoch": 1.06, + "learning_rate": 1.6493698877176413e-05, + "loss": 1.0093, + "step": 16442 + }, + { + "epoch": 1.06, + "learning_rate": 1.649348418815346e-05, + "loss": 0.9241, + "step": 16443 + }, + { + "epoch": 1.06, + "learning_rate": 1.649326949913051e-05, + "loss": 0.9657, + "step": 16444 + }, + { + "epoch": 1.06, + "learning_rate": 1.649305481010756e-05, + "loss": 1.1, + "step": 16445 + }, + { + "epoch": 1.06, + "learning_rate": 1.649284012108461e-05, + "loss": 0.9602, + "step": 16446 + }, + { + "epoch": 1.06, + "learning_rate": 1.649262543206166e-05, + "loss": 0.9978, + "step": 16447 + }, + { + "epoch": 1.06, + "learning_rate": 1.649241074303871e-05, + "loss": 0.9208, + "step": 16448 + }, + { + "epoch": 1.06, + "learning_rate": 1.649219605401576e-05, + "loss": 0.966, + "step": 16449 + }, + { + "epoch": 1.06, + "learning_rate": 1.6491981364992808e-05, + "loss": 0.9306, + "step": 16450 + }, + { + "epoch": 1.06, + "learning_rate": 1.6491766675969857e-05, + "loss": 0.9354, + "step": 16451 + }, + { + "epoch": 1.06, + "learning_rate": 1.649155198694691e-05, + "loss": 0.9122, + "step": 16452 + }, + { + "epoch": 1.06, + "learning_rate": 1.6491337297923958e-05, + "loss": 0.956, + "step": 16453 + }, + { + "epoch": 1.06, + "learning_rate": 1.649112260890101e-05, + "loss": 0.9715, + "step": 16454 + }, + { + "epoch": 1.06, + "learning_rate": 1.649090791987806e-05, + "loss": 0.8924, + "step": 16455 + }, + { + "epoch": 1.06, + "learning_rate": 1.6490693230855108e-05, + "loss": 0.9305, + "step": 16456 + }, + { + "epoch": 1.06, + "learning_rate": 1.6490478541832157e-05, + "loss": 0.9647, + "step": 16457 + }, + { + "epoch": 1.06, + "learning_rate": 1.6490263852809206e-05, + "loss": 0.9631, + "step": 16458 + }, + { + "epoch": 1.06, + "learning_rate": 1.649004916378626e-05, + "loss": 0.9615, + "step": 16459 + }, + { + "epoch": 1.06, + "learning_rate": 1.6489834474763308e-05, + "loss": 1.0026, + "step": 16460 + }, + { + "epoch": 1.06, + "learning_rate": 1.6489619785740356e-05, + "loss": 1.0224, + "step": 16461 + }, + { + "epoch": 1.06, + "learning_rate": 1.6489405096717405e-05, + "loss": 0.9278, + "step": 16462 + }, + { + "epoch": 1.06, + "learning_rate": 1.6489190407694454e-05, + "loss": 0.9469, + "step": 16463 + }, + { + "epoch": 1.06, + "learning_rate": 1.6488975718671503e-05, + "loss": 0.9001, + "step": 16464 + }, + { + "epoch": 1.06, + "learning_rate": 1.6488761029648556e-05, + "loss": 0.9714, + "step": 16465 + }, + { + "epoch": 1.06, + "learning_rate": 1.6488546340625605e-05, + "loss": 0.9821, + "step": 16466 + }, + { + "epoch": 1.06, + "learning_rate": 1.6488331651602657e-05, + "loss": 0.982, + "step": 16467 + }, + { + "epoch": 1.06, + "learning_rate": 1.6488116962579706e-05, + "loss": 0.9967, + "step": 16468 + }, + { + "epoch": 1.06, + "learning_rate": 1.6487902273556755e-05, + "loss": 0.9564, + "step": 16469 + }, + { + "epoch": 1.06, + "learning_rate": 1.6487687584533804e-05, + "loss": 0.9442, + "step": 16470 + }, + { + "epoch": 1.06, + "learning_rate": 1.6487472895510853e-05, + "loss": 0.9424, + "step": 16471 + }, + { + "epoch": 1.06, + "learning_rate": 1.6487258206487905e-05, + "loss": 1.0173, + "step": 16472 + }, + { + "epoch": 1.06, + "learning_rate": 1.6487043517464954e-05, + "loss": 0.9413, + "step": 16473 + }, + { + "epoch": 1.06, + "learning_rate": 1.6486828828442003e-05, + "loss": 0.889, + "step": 16474 + }, + { + "epoch": 1.06, + "learning_rate": 1.6486614139419052e-05, + "loss": 0.915, + "step": 16475 + }, + { + "epoch": 1.06, + "learning_rate": 1.64863994503961e-05, + "loss": 1.0303, + "step": 16476 + }, + { + "epoch": 1.06, + "learning_rate": 1.6486184761373153e-05, + "loss": 0.9521, + "step": 16477 + }, + { + "epoch": 1.06, + "learning_rate": 1.6485970072350202e-05, + "loss": 1.0543, + "step": 16478 + }, + { + "epoch": 1.06, + "learning_rate": 1.648575538332725e-05, + "loss": 0.967, + "step": 16479 + }, + { + "epoch": 1.06, + "learning_rate": 1.64855406943043e-05, + "loss": 1.0047, + "step": 16480 + }, + { + "epoch": 1.06, + "learning_rate": 1.6485326005281353e-05, + "loss": 0.9281, + "step": 16481 + }, + { + "epoch": 1.06, + "learning_rate": 1.6485111316258402e-05, + "loss": 0.9436, + "step": 16482 + }, + { + "epoch": 1.06, + "learning_rate": 1.648489662723545e-05, + "loss": 0.9101, + "step": 16483 + }, + { + "epoch": 1.06, + "learning_rate": 1.64846819382125e-05, + "loss": 1.0764, + "step": 16484 + }, + { + "epoch": 1.06, + "learning_rate": 1.6484467249189552e-05, + "loss": 0.9845, + "step": 16485 + }, + { + "epoch": 1.06, + "learning_rate": 1.64842525601666e-05, + "loss": 0.9904, + "step": 16486 + }, + { + "epoch": 1.06, + "learning_rate": 1.648403787114365e-05, + "loss": 1.0062, + "step": 16487 + }, + { + "epoch": 1.06, + "learning_rate": 1.64838231821207e-05, + "loss": 0.9564, + "step": 16488 + }, + { + "epoch": 1.06, + "learning_rate": 1.6483608493097748e-05, + "loss": 0.9701, + "step": 16489 + }, + { + "epoch": 1.06, + "learning_rate": 1.64833938040748e-05, + "loss": 0.98, + "step": 16490 + }, + { + "epoch": 1.06, + "learning_rate": 1.648317911505185e-05, + "loss": 0.9976, + "step": 16491 + }, + { + "epoch": 1.06, + "learning_rate": 1.6482964426028898e-05, + "loss": 1.0259, + "step": 16492 + }, + { + "epoch": 1.06, + "learning_rate": 1.6482749737005947e-05, + "loss": 1.0342, + "step": 16493 + }, + { + "epoch": 1.06, + "learning_rate": 1.6482535047982996e-05, + "loss": 0.9645, + "step": 16494 + }, + { + "epoch": 1.06, + "learning_rate": 1.6482320358960045e-05, + "loss": 0.9856, + "step": 16495 + }, + { + "epoch": 1.06, + "learning_rate": 1.6482105669937097e-05, + "loss": 0.949, + "step": 16496 + }, + { + "epoch": 1.06, + "learning_rate": 1.648189098091415e-05, + "loss": 1.0308, + "step": 16497 + }, + { + "epoch": 1.06, + "learning_rate": 1.64816762918912e-05, + "loss": 0.934, + "step": 16498 + }, + { + "epoch": 1.06, + "learning_rate": 1.6481461602868248e-05, + "loss": 0.9615, + "step": 16499 + }, + { + "epoch": 1.06, + "learning_rate": 1.6481246913845297e-05, + "loss": 0.9448, + "step": 16500 + }, + { + "epoch": 1.06, + "learning_rate": 1.6481032224822346e-05, + "loss": 1.0142, + "step": 16501 + }, + { + "epoch": 1.06, + "learning_rate": 1.6480817535799395e-05, + "loss": 1.0018, + "step": 16502 + }, + { + "epoch": 1.06, + "learning_rate": 1.6480602846776447e-05, + "loss": 0.9367, + "step": 16503 + }, + { + "epoch": 1.06, + "learning_rate": 1.6480388157753496e-05, + "loss": 1.0221, + "step": 16504 + }, + { + "epoch": 1.06, + "learning_rate": 1.6480173468730545e-05, + "loss": 0.9628, + "step": 16505 + }, + { + "epoch": 1.06, + "learning_rate": 1.6479958779707594e-05, + "loss": 1.0455, + "step": 16506 + }, + { + "epoch": 1.06, + "learning_rate": 1.6479744090684643e-05, + "loss": 0.9973, + "step": 16507 + }, + { + "epoch": 1.06, + "learning_rate": 1.6479529401661695e-05, + "loss": 0.9593, + "step": 16508 + }, + { + "epoch": 1.06, + "learning_rate": 1.6479314712638744e-05, + "loss": 1.0329, + "step": 16509 + }, + { + "epoch": 1.06, + "learning_rate": 1.6479100023615796e-05, + "loss": 0.9304, + "step": 16510 + }, + { + "epoch": 1.06, + "learning_rate": 1.6478885334592845e-05, + "loss": 0.9179, + "step": 16511 + }, + { + "epoch": 1.06, + "learning_rate": 1.6478670645569894e-05, + "loss": 1.0007, + "step": 16512 + }, + { + "epoch": 1.06, + "learning_rate": 1.6478455956546943e-05, + "loss": 1.0064, + "step": 16513 + }, + { + "epoch": 1.06, + "learning_rate": 1.6478241267523992e-05, + "loss": 0.9398, + "step": 16514 + }, + { + "epoch": 1.06, + "learning_rate": 1.647802657850104e-05, + "loss": 1.0041, + "step": 16515 + }, + { + "epoch": 1.06, + "learning_rate": 1.6477811889478094e-05, + "loss": 0.9504, + "step": 16516 + }, + { + "epoch": 1.06, + "learning_rate": 1.6477597200455143e-05, + "loss": 0.9227, + "step": 16517 + }, + { + "epoch": 1.06, + "learning_rate": 1.647738251143219e-05, + "loss": 0.9688, + "step": 16518 + }, + { + "epoch": 1.06, + "learning_rate": 1.647716782240924e-05, + "loss": 0.9845, + "step": 16519 + }, + { + "epoch": 1.06, + "learning_rate": 1.647695313338629e-05, + "loss": 0.9444, + "step": 16520 + }, + { + "epoch": 1.06, + "learning_rate": 1.6476738444363342e-05, + "loss": 0.9446, + "step": 16521 + }, + { + "epoch": 1.06, + "learning_rate": 1.647652375534039e-05, + "loss": 0.9942, + "step": 16522 + }, + { + "epoch": 1.06, + "learning_rate": 1.647630906631744e-05, + "loss": 0.9463, + "step": 16523 + }, + { + "epoch": 1.06, + "learning_rate": 1.647609437729449e-05, + "loss": 1.0009, + "step": 16524 + }, + { + "epoch": 1.06, + "learning_rate": 1.647587968827154e-05, + "loss": 0.9853, + "step": 16525 + }, + { + "epoch": 1.06, + "learning_rate": 1.647566499924859e-05, + "loss": 0.8274, + "step": 16526 + }, + { + "epoch": 1.06, + "learning_rate": 1.647545031022564e-05, + "loss": 1.0019, + "step": 16527 + }, + { + "epoch": 1.06, + "learning_rate": 1.647523562120269e-05, + "loss": 0.975, + "step": 16528 + }, + { + "epoch": 1.06, + "learning_rate": 1.647502093217974e-05, + "loss": 0.9115, + "step": 16529 + }, + { + "epoch": 1.06, + "learning_rate": 1.647480624315679e-05, + "loss": 0.9568, + "step": 16530 + }, + { + "epoch": 1.06, + "learning_rate": 1.6474591554133838e-05, + "loss": 0.9915, + "step": 16531 + }, + { + "epoch": 1.06, + "learning_rate": 1.6474376865110887e-05, + "loss": 1.03, + "step": 16532 + }, + { + "epoch": 1.06, + "learning_rate": 1.6474162176087936e-05, + "loss": 1.0367, + "step": 16533 + }, + { + "epoch": 1.06, + "learning_rate": 1.647394748706499e-05, + "loss": 1.0325, + "step": 16534 + }, + { + "epoch": 1.06, + "learning_rate": 1.6473732798042038e-05, + "loss": 0.9051, + "step": 16535 + }, + { + "epoch": 1.06, + "learning_rate": 1.6473518109019086e-05, + "loss": 1.0082, + "step": 16536 + }, + { + "epoch": 1.06, + "learning_rate": 1.6473303419996135e-05, + "loss": 0.9454, + "step": 16537 + }, + { + "epoch": 1.06, + "learning_rate": 1.6473088730973184e-05, + "loss": 0.8908, + "step": 16538 + }, + { + "epoch": 1.06, + "learning_rate": 1.6472874041950237e-05, + "loss": 0.9654, + "step": 16539 + }, + { + "epoch": 1.06, + "learning_rate": 1.6472659352927286e-05, + "loss": 0.9391, + "step": 16540 + }, + { + "epoch": 1.06, + "learning_rate": 1.6472444663904338e-05, + "loss": 0.869, + "step": 16541 + }, + { + "epoch": 1.06, + "learning_rate": 1.6472229974881387e-05, + "loss": 1.0142, + "step": 16542 + }, + { + "epoch": 1.06, + "learning_rate": 1.6472015285858436e-05, + "loss": 0.9474, + "step": 16543 + }, + { + "epoch": 1.06, + "learning_rate": 1.6471800596835485e-05, + "loss": 1.0393, + "step": 16544 + }, + { + "epoch": 1.06, + "learning_rate": 1.6471585907812534e-05, + "loss": 1.0315, + "step": 16545 + }, + { + "epoch": 1.06, + "learning_rate": 1.6471371218789586e-05, + "loss": 0.9814, + "step": 16546 + }, + { + "epoch": 1.06, + "learning_rate": 1.6471156529766635e-05, + "loss": 1.0213, + "step": 16547 + }, + { + "epoch": 1.06, + "learning_rate": 1.6470941840743684e-05, + "loss": 0.975, + "step": 16548 + }, + { + "epoch": 1.06, + "learning_rate": 1.6470727151720733e-05, + "loss": 0.9788, + "step": 16549 + }, + { + "epoch": 1.06, + "learning_rate": 1.6470512462697782e-05, + "loss": 0.9733, + "step": 16550 + }, + { + "epoch": 1.06, + "learning_rate": 1.647029777367483e-05, + "loss": 0.9723, + "step": 16551 + }, + { + "epoch": 1.06, + "learning_rate": 1.6470083084651883e-05, + "loss": 0.9251, + "step": 16552 + }, + { + "epoch": 1.06, + "learning_rate": 1.6469868395628932e-05, + "loss": 1.0782, + "step": 16553 + }, + { + "epoch": 1.07, + "learning_rate": 1.6469653706605985e-05, + "loss": 0.9531, + "step": 16554 + }, + { + "epoch": 1.07, + "learning_rate": 1.6469439017583034e-05, + "loss": 1.0597, + "step": 16555 + }, + { + "epoch": 1.07, + "learning_rate": 1.6469224328560083e-05, + "loss": 0.9872, + "step": 16556 + }, + { + "epoch": 1.07, + "learning_rate": 1.646900963953713e-05, + "loss": 1.0292, + "step": 16557 + }, + { + "epoch": 1.07, + "learning_rate": 1.646879495051418e-05, + "loss": 0.9626, + "step": 16558 + }, + { + "epoch": 1.07, + "learning_rate": 1.6468580261491233e-05, + "loss": 1.005, + "step": 16559 + }, + { + "epoch": 1.07, + "learning_rate": 1.6468365572468282e-05, + "loss": 0.9448, + "step": 16560 + }, + { + "epoch": 1.07, + "learning_rate": 1.646815088344533e-05, + "loss": 0.9298, + "step": 16561 + }, + { + "epoch": 1.07, + "learning_rate": 1.646793619442238e-05, + "loss": 0.9297, + "step": 16562 + }, + { + "epoch": 1.07, + "learning_rate": 1.646772150539943e-05, + "loss": 0.9766, + "step": 16563 + }, + { + "epoch": 1.07, + "learning_rate": 1.646750681637648e-05, + "loss": 0.9146, + "step": 16564 + }, + { + "epoch": 1.07, + "learning_rate": 1.646729212735353e-05, + "loss": 0.9327, + "step": 16565 + }, + { + "epoch": 1.07, + "learning_rate": 1.646707743833058e-05, + "loss": 0.9713, + "step": 16566 + }, + { + "epoch": 1.07, + "learning_rate": 1.6466862749307628e-05, + "loss": 0.9736, + "step": 16567 + }, + { + "epoch": 1.07, + "learning_rate": 1.6466648060284677e-05, + "loss": 0.906, + "step": 16568 + }, + { + "epoch": 1.07, + "learning_rate": 1.646643337126173e-05, + "loss": 0.9837, + "step": 16569 + }, + { + "epoch": 1.07, + "learning_rate": 1.646621868223878e-05, + "loss": 1.0027, + "step": 16570 + }, + { + "epoch": 1.07, + "learning_rate": 1.6466003993215827e-05, + "loss": 0.8958, + "step": 16571 + }, + { + "epoch": 1.07, + "learning_rate": 1.646578930419288e-05, + "loss": 1.0154, + "step": 16572 + }, + { + "epoch": 1.07, + "learning_rate": 1.646557461516993e-05, + "loss": 1.0072, + "step": 16573 + }, + { + "epoch": 1.07, + "learning_rate": 1.6465359926146978e-05, + "loss": 0.9828, + "step": 16574 + }, + { + "epoch": 1.07, + "learning_rate": 1.6465145237124027e-05, + "loss": 0.9536, + "step": 16575 + }, + { + "epoch": 1.07, + "learning_rate": 1.6464930548101076e-05, + "loss": 0.9584, + "step": 16576 + }, + { + "epoch": 1.07, + "learning_rate": 1.6464715859078128e-05, + "loss": 0.9375, + "step": 16577 + }, + { + "epoch": 1.07, + "learning_rate": 1.6464501170055177e-05, + "loss": 1.0079, + "step": 16578 + }, + { + "epoch": 1.07, + "learning_rate": 1.6464286481032226e-05, + "loss": 0.9805, + "step": 16579 + }, + { + "epoch": 1.07, + "learning_rate": 1.6464071792009275e-05, + "loss": 0.9712, + "step": 16580 + }, + { + "epoch": 1.07, + "learning_rate": 1.6463857102986324e-05, + "loss": 0.9548, + "step": 16581 + }, + { + "epoch": 1.07, + "learning_rate": 1.6463642413963373e-05, + "loss": 0.8906, + "step": 16582 + }, + { + "epoch": 1.07, + "learning_rate": 1.6463427724940425e-05, + "loss": 1.0443, + "step": 16583 + }, + { + "epoch": 1.07, + "learning_rate": 1.6463213035917477e-05, + "loss": 1.008, + "step": 16584 + }, + { + "epoch": 1.07, + "learning_rate": 1.6462998346894526e-05, + "loss": 1.0086, + "step": 16585 + }, + { + "epoch": 1.07, + "learning_rate": 1.6462783657871575e-05, + "loss": 0.986, + "step": 16586 + }, + { + "epoch": 1.07, + "learning_rate": 1.6462568968848624e-05, + "loss": 0.968, + "step": 16587 + }, + { + "epoch": 1.07, + "learning_rate": 1.6462354279825673e-05, + "loss": 0.9703, + "step": 16588 + }, + { + "epoch": 1.07, + "learning_rate": 1.6462139590802722e-05, + "loss": 0.9685, + "step": 16589 + }, + { + "epoch": 1.07, + "learning_rate": 1.6461924901779775e-05, + "loss": 1.0577, + "step": 16590 + }, + { + "epoch": 1.07, + "learning_rate": 1.6461710212756824e-05, + "loss": 1.0665, + "step": 16591 + }, + { + "epoch": 1.07, + "learning_rate": 1.6461495523733873e-05, + "loss": 1.0005, + "step": 16592 + }, + { + "epoch": 1.07, + "learning_rate": 1.646128083471092e-05, + "loss": 1.0298, + "step": 16593 + }, + { + "epoch": 1.07, + "learning_rate": 1.646106614568797e-05, + "loss": 1.0047, + "step": 16594 + }, + { + "epoch": 1.07, + "learning_rate": 1.6460851456665023e-05, + "loss": 0.9602, + "step": 16595 + }, + { + "epoch": 1.07, + "learning_rate": 1.6460636767642072e-05, + "loss": 0.9589, + "step": 16596 + }, + { + "epoch": 1.07, + "learning_rate": 1.646042207861912e-05, + "loss": 1.0338, + "step": 16597 + }, + { + "epoch": 1.07, + "learning_rate": 1.6460207389596173e-05, + "loss": 1.046, + "step": 16598 + }, + { + "epoch": 1.07, + "learning_rate": 1.6459992700573222e-05, + "loss": 0.9987, + "step": 16599 + }, + { + "epoch": 1.07, + "learning_rate": 1.645977801155027e-05, + "loss": 0.9877, + "step": 16600 + }, + { + "epoch": 1.07, + "learning_rate": 1.645956332252732e-05, + "loss": 0.9662, + "step": 16601 + }, + { + "epoch": 1.07, + "learning_rate": 1.645934863350437e-05, + "loss": 0.9588, + "step": 16602 + }, + { + "epoch": 1.07, + "learning_rate": 1.645913394448142e-05, + "loss": 0.8991, + "step": 16603 + }, + { + "epoch": 1.07, + "learning_rate": 1.645891925545847e-05, + "loss": 0.9641, + "step": 16604 + }, + { + "epoch": 1.07, + "learning_rate": 1.645870456643552e-05, + "loss": 0.9421, + "step": 16605 + }, + { + "epoch": 1.07, + "learning_rate": 1.6458489877412568e-05, + "loss": 0.9523, + "step": 16606 + }, + { + "epoch": 1.07, + "learning_rate": 1.6458275188389617e-05, + "loss": 0.9427, + "step": 16607 + }, + { + "epoch": 1.07, + "learning_rate": 1.645806049936667e-05, + "loss": 0.9628, + "step": 16608 + }, + { + "epoch": 1.07, + "learning_rate": 1.645784581034372e-05, + "loss": 0.9404, + "step": 16609 + }, + { + "epoch": 1.07, + "learning_rate": 1.6457631121320767e-05, + "loss": 0.991, + "step": 16610 + }, + { + "epoch": 1.07, + "learning_rate": 1.6457416432297816e-05, + "loss": 0.9219, + "step": 16611 + }, + { + "epoch": 1.07, + "learning_rate": 1.645720174327487e-05, + "loss": 0.9806, + "step": 16612 + }, + { + "epoch": 1.07, + "learning_rate": 1.6456987054251918e-05, + "loss": 1.0415, + "step": 16613 + }, + { + "epoch": 1.07, + "learning_rate": 1.6456772365228967e-05, + "loss": 0.9926, + "step": 16614 + }, + { + "epoch": 1.07, + "learning_rate": 1.645655767620602e-05, + "loss": 1.0069, + "step": 16615 + }, + { + "epoch": 1.07, + "learning_rate": 1.6456342987183068e-05, + "loss": 0.915, + "step": 16616 + }, + { + "epoch": 1.07, + "learning_rate": 1.6456128298160117e-05, + "loss": 0.9458, + "step": 16617 + }, + { + "epoch": 1.07, + "learning_rate": 1.6455913609137166e-05, + "loss": 1.0105, + "step": 16618 + }, + { + "epoch": 1.07, + "learning_rate": 1.6455698920114215e-05, + "loss": 0.9467, + "step": 16619 + }, + { + "epoch": 1.07, + "learning_rate": 1.6455484231091264e-05, + "loss": 0.9515, + "step": 16620 + }, + { + "epoch": 1.07, + "learning_rate": 1.6455269542068316e-05, + "loss": 0.8981, + "step": 16621 + }, + { + "epoch": 1.07, + "learning_rate": 1.6455054853045365e-05, + "loss": 0.968, + "step": 16622 + }, + { + "epoch": 1.07, + "learning_rate": 1.6454840164022414e-05, + "loss": 0.9692, + "step": 16623 + }, + { + "epoch": 1.07, + "learning_rate": 1.6454625474999463e-05, + "loss": 0.9076, + "step": 16624 + }, + { + "epoch": 1.07, + "learning_rate": 1.6454410785976512e-05, + "loss": 1.0451, + "step": 16625 + }, + { + "epoch": 1.07, + "learning_rate": 1.6454196096953564e-05, + "loss": 0.9795, + "step": 16626 + }, + { + "epoch": 1.07, + "learning_rate": 1.6453981407930613e-05, + "loss": 0.8897, + "step": 16627 + }, + { + "epoch": 1.07, + "learning_rate": 1.6453766718907666e-05, + "loss": 0.9384, + "step": 16628 + }, + { + "epoch": 1.07, + "learning_rate": 1.6453552029884715e-05, + "loss": 0.9877, + "step": 16629 + }, + { + "epoch": 1.07, + "learning_rate": 1.6453337340861764e-05, + "loss": 0.9467, + "step": 16630 + }, + { + "epoch": 1.07, + "learning_rate": 1.6453122651838813e-05, + "loss": 0.9883, + "step": 16631 + }, + { + "epoch": 1.07, + "learning_rate": 1.645290796281586e-05, + "loss": 0.8929, + "step": 16632 + }, + { + "epoch": 1.07, + "learning_rate": 1.6452693273792914e-05, + "loss": 1.0266, + "step": 16633 + }, + { + "epoch": 1.07, + "learning_rate": 1.6452478584769963e-05, + "loss": 0.9519, + "step": 16634 + }, + { + "epoch": 1.07, + "learning_rate": 1.6452263895747012e-05, + "loss": 1.0131, + "step": 16635 + }, + { + "epoch": 1.07, + "learning_rate": 1.645204920672406e-05, + "loss": 0.9706, + "step": 16636 + }, + { + "epoch": 1.07, + "learning_rate": 1.645183451770111e-05, + "loss": 1.0099, + "step": 16637 + }, + { + "epoch": 1.07, + "learning_rate": 1.645161982867816e-05, + "loss": 0.9569, + "step": 16638 + }, + { + "epoch": 1.07, + "learning_rate": 1.645140513965521e-05, + "loss": 0.9611, + "step": 16639 + }, + { + "epoch": 1.07, + "learning_rate": 1.645119045063226e-05, + "loss": 1.0709, + "step": 16640 + }, + { + "epoch": 1.07, + "learning_rate": 1.6450975761609312e-05, + "loss": 0.9811, + "step": 16641 + }, + { + "epoch": 1.07, + "learning_rate": 1.645076107258636e-05, + "loss": 0.9644, + "step": 16642 + }, + { + "epoch": 1.07, + "learning_rate": 1.645054638356341e-05, + "loss": 0.9412, + "step": 16643 + }, + { + "epoch": 1.07, + "learning_rate": 1.645033169454046e-05, + "loss": 1.0164, + "step": 16644 + }, + { + "epoch": 1.07, + "learning_rate": 1.645011700551751e-05, + "loss": 0.9318, + "step": 16645 + }, + { + "epoch": 1.07, + "learning_rate": 1.644990231649456e-05, + "loss": 0.9003, + "step": 16646 + }, + { + "epoch": 1.07, + "learning_rate": 1.644968762747161e-05, + "loss": 0.94, + "step": 16647 + }, + { + "epoch": 1.07, + "learning_rate": 1.644947293844866e-05, + "loss": 0.9535, + "step": 16648 + }, + { + "epoch": 1.07, + "learning_rate": 1.6449258249425708e-05, + "loss": 0.9401, + "step": 16649 + }, + { + "epoch": 1.07, + "learning_rate": 1.6449043560402757e-05, + "loss": 0.913, + "step": 16650 + }, + { + "epoch": 1.07, + "learning_rate": 1.6448828871379806e-05, + "loss": 1.0182, + "step": 16651 + }, + { + "epoch": 1.07, + "learning_rate": 1.6448614182356858e-05, + "loss": 0.9827, + "step": 16652 + }, + { + "epoch": 1.07, + "learning_rate": 1.6448399493333907e-05, + "loss": 0.9358, + "step": 16653 + }, + { + "epoch": 1.07, + "learning_rate": 1.6448184804310956e-05, + "loss": 1.002, + "step": 16654 + }, + { + "epoch": 1.07, + "learning_rate": 1.6447970115288005e-05, + "loss": 0.9949, + "step": 16655 + }, + { + "epoch": 1.07, + "learning_rate": 1.6447755426265057e-05, + "loss": 0.9667, + "step": 16656 + }, + { + "epoch": 1.07, + "learning_rate": 1.6447540737242106e-05, + "loss": 0.9659, + "step": 16657 + }, + { + "epoch": 1.07, + "learning_rate": 1.6447326048219155e-05, + "loss": 0.9164, + "step": 16658 + }, + { + "epoch": 1.07, + "learning_rate": 1.6447111359196207e-05, + "loss": 1.01, + "step": 16659 + }, + { + "epoch": 1.07, + "learning_rate": 1.6446896670173256e-05, + "loss": 1.0458, + "step": 16660 + }, + { + "epoch": 1.07, + "learning_rate": 1.6446681981150305e-05, + "loss": 0.9319, + "step": 16661 + }, + { + "epoch": 1.07, + "learning_rate": 1.6446467292127354e-05, + "loss": 0.9627, + "step": 16662 + }, + { + "epoch": 1.07, + "learning_rate": 1.6446252603104403e-05, + "loss": 0.9987, + "step": 16663 + }, + { + "epoch": 1.07, + "learning_rate": 1.6446037914081456e-05, + "loss": 0.9232, + "step": 16664 + }, + { + "epoch": 1.07, + "learning_rate": 1.6445823225058505e-05, + "loss": 1.0066, + "step": 16665 + }, + { + "epoch": 1.07, + "learning_rate": 1.6445608536035554e-05, + "loss": 0.9833, + "step": 16666 + }, + { + "epoch": 1.07, + "learning_rate": 1.6445393847012603e-05, + "loss": 0.9555, + "step": 16667 + }, + { + "epoch": 1.07, + "learning_rate": 1.644517915798965e-05, + "loss": 1.0417, + "step": 16668 + }, + { + "epoch": 1.07, + "learning_rate": 1.64449644689667e-05, + "loss": 1.0447, + "step": 16669 + }, + { + "epoch": 1.07, + "learning_rate": 1.6444749779943753e-05, + "loss": 0.9682, + "step": 16670 + }, + { + "epoch": 1.07, + "learning_rate": 1.6444535090920805e-05, + "loss": 0.9927, + "step": 16671 + }, + { + "epoch": 1.07, + "learning_rate": 1.6444320401897854e-05, + "loss": 0.9938, + "step": 16672 + }, + { + "epoch": 1.07, + "learning_rate": 1.6444105712874903e-05, + "loss": 0.9785, + "step": 16673 + }, + { + "epoch": 1.07, + "learning_rate": 1.6443891023851952e-05, + "loss": 0.9581, + "step": 16674 + }, + { + "epoch": 1.07, + "learning_rate": 1.6443676334829e-05, + "loss": 1.02, + "step": 16675 + }, + { + "epoch": 1.07, + "learning_rate": 1.644346164580605e-05, + "loss": 1.0512, + "step": 16676 + }, + { + "epoch": 1.07, + "learning_rate": 1.6443246956783102e-05, + "loss": 1.0394, + "step": 16677 + }, + { + "epoch": 1.07, + "learning_rate": 1.644303226776015e-05, + "loss": 0.8784, + "step": 16678 + }, + { + "epoch": 1.07, + "learning_rate": 1.64428175787372e-05, + "loss": 0.9868, + "step": 16679 + }, + { + "epoch": 1.07, + "learning_rate": 1.644260288971425e-05, + "loss": 0.9448, + "step": 16680 + }, + { + "epoch": 1.07, + "learning_rate": 1.6442388200691298e-05, + "loss": 0.989, + "step": 16681 + }, + { + "epoch": 1.07, + "learning_rate": 1.644217351166835e-05, + "loss": 0.9445, + "step": 16682 + }, + { + "epoch": 1.07, + "learning_rate": 1.64419588226454e-05, + "loss": 1.0086, + "step": 16683 + }, + { + "epoch": 1.07, + "learning_rate": 1.644174413362245e-05, + "loss": 0.9817, + "step": 16684 + }, + { + "epoch": 1.07, + "learning_rate": 1.64415294445995e-05, + "loss": 0.9226, + "step": 16685 + }, + { + "epoch": 1.07, + "learning_rate": 1.644131475557655e-05, + "loss": 0.9757, + "step": 16686 + }, + { + "epoch": 1.07, + "learning_rate": 1.64411000665536e-05, + "loss": 0.9798, + "step": 16687 + }, + { + "epoch": 1.07, + "learning_rate": 1.6440885377530648e-05, + "loss": 0.9187, + "step": 16688 + }, + { + "epoch": 1.07, + "learning_rate": 1.6440670688507697e-05, + "loss": 0.9157, + "step": 16689 + }, + { + "epoch": 1.07, + "learning_rate": 1.644045599948475e-05, + "loss": 0.9657, + "step": 16690 + }, + { + "epoch": 1.07, + "learning_rate": 1.6440241310461798e-05, + "loss": 0.9547, + "step": 16691 + }, + { + "epoch": 1.07, + "learning_rate": 1.6440026621438847e-05, + "loss": 1.0002, + "step": 16692 + }, + { + "epoch": 1.07, + "learning_rate": 1.6439811932415896e-05, + "loss": 0.9749, + "step": 16693 + }, + { + "epoch": 1.07, + "learning_rate": 1.6439597243392945e-05, + "loss": 0.9562, + "step": 16694 + }, + { + "epoch": 1.07, + "learning_rate": 1.6439382554369997e-05, + "loss": 0.9793, + "step": 16695 + }, + { + "epoch": 1.07, + "learning_rate": 1.6439167865347046e-05, + "loss": 1.0136, + "step": 16696 + }, + { + "epoch": 1.07, + "learning_rate": 1.6438953176324095e-05, + "loss": 0.9558, + "step": 16697 + }, + { + "epoch": 1.07, + "learning_rate": 1.6438738487301144e-05, + "loss": 1.0012, + "step": 16698 + }, + { + "epoch": 1.07, + "learning_rate": 1.6438523798278196e-05, + "loss": 0.9811, + "step": 16699 + }, + { + "epoch": 1.07, + "learning_rate": 1.6438309109255245e-05, + "loss": 0.9628, + "step": 16700 + }, + { + "epoch": 1.07, + "learning_rate": 1.6438094420232294e-05, + "loss": 1.0575, + "step": 16701 + }, + { + "epoch": 1.07, + "learning_rate": 1.6437879731209347e-05, + "loss": 0.9025, + "step": 16702 + }, + { + "epoch": 1.07, + "learning_rate": 1.6437665042186396e-05, + "loss": 0.9438, + "step": 16703 + }, + { + "epoch": 1.07, + "learning_rate": 1.6437450353163445e-05, + "loss": 0.9544, + "step": 16704 + }, + { + "epoch": 1.07, + "learning_rate": 1.6437235664140494e-05, + "loss": 0.965, + "step": 16705 + }, + { + "epoch": 1.07, + "learning_rate": 1.6437020975117543e-05, + "loss": 0.9836, + "step": 16706 + }, + { + "epoch": 1.07, + "learning_rate": 1.643680628609459e-05, + "loss": 0.934, + "step": 16707 + }, + { + "epoch": 1.07, + "learning_rate": 1.6436591597071644e-05, + "loss": 0.999, + "step": 16708 + }, + { + "epoch": 1.08, + "learning_rate": 1.6436376908048693e-05, + "loss": 1.0099, + "step": 16709 + }, + { + "epoch": 1.08, + "learning_rate": 1.6436162219025742e-05, + "loss": 0.9898, + "step": 16710 + }, + { + "epoch": 1.08, + "learning_rate": 1.643594753000279e-05, + "loss": 0.9895, + "step": 16711 + }, + { + "epoch": 1.08, + "learning_rate": 1.643573284097984e-05, + "loss": 0.9634, + "step": 16712 + }, + { + "epoch": 1.08, + "learning_rate": 1.6435518151956892e-05, + "loss": 1.008, + "step": 16713 + }, + { + "epoch": 1.08, + "learning_rate": 1.643530346293394e-05, + "loss": 1.0161, + "step": 16714 + }, + { + "epoch": 1.08, + "learning_rate": 1.6435088773910993e-05, + "loss": 1.0011, + "step": 16715 + }, + { + "epoch": 1.08, + "learning_rate": 1.6434874084888042e-05, + "loss": 1.0019, + "step": 16716 + }, + { + "epoch": 1.08, + "learning_rate": 1.643465939586509e-05, + "loss": 1.0011, + "step": 16717 + }, + { + "epoch": 1.08, + "learning_rate": 1.643444470684214e-05, + "loss": 0.9367, + "step": 16718 + }, + { + "epoch": 1.08, + "learning_rate": 1.643423001781919e-05, + "loss": 0.9502, + "step": 16719 + }, + { + "epoch": 1.08, + "learning_rate": 1.6434015328796242e-05, + "loss": 0.9471, + "step": 16720 + }, + { + "epoch": 1.08, + "learning_rate": 1.643380063977329e-05, + "loss": 0.9889, + "step": 16721 + }, + { + "epoch": 1.08, + "learning_rate": 1.643358595075034e-05, + "loss": 0.9228, + "step": 16722 + }, + { + "epoch": 1.08, + "learning_rate": 1.643337126172739e-05, + "loss": 0.9941, + "step": 16723 + }, + { + "epoch": 1.08, + "learning_rate": 1.6433156572704438e-05, + "loss": 0.9924, + "step": 16724 + }, + { + "epoch": 1.08, + "learning_rate": 1.6432941883681487e-05, + "loss": 0.9766, + "step": 16725 + }, + { + "epoch": 1.08, + "learning_rate": 1.643272719465854e-05, + "loss": 0.9666, + "step": 16726 + }, + { + "epoch": 1.08, + "learning_rate": 1.6432512505635588e-05, + "loss": 0.9549, + "step": 16727 + }, + { + "epoch": 1.08, + "learning_rate": 1.643229781661264e-05, + "loss": 0.9955, + "step": 16728 + }, + { + "epoch": 1.08, + "learning_rate": 1.643208312758969e-05, + "loss": 1.0045, + "step": 16729 + }, + { + "epoch": 1.08, + "learning_rate": 1.6431868438566738e-05, + "loss": 0.9643, + "step": 16730 + }, + { + "epoch": 1.08, + "learning_rate": 1.6431653749543787e-05, + "loss": 0.9823, + "step": 16731 + }, + { + "epoch": 1.08, + "learning_rate": 1.6431439060520836e-05, + "loss": 0.9169, + "step": 16732 + }, + { + "epoch": 1.08, + "learning_rate": 1.643122437149789e-05, + "loss": 0.9521, + "step": 16733 + }, + { + "epoch": 1.08, + "learning_rate": 1.6431009682474937e-05, + "loss": 0.9383, + "step": 16734 + }, + { + "epoch": 1.08, + "learning_rate": 1.6430794993451986e-05, + "loss": 0.9926, + "step": 16735 + }, + { + "epoch": 1.08, + "learning_rate": 1.6430580304429035e-05, + "loss": 1.0153, + "step": 16736 + }, + { + "epoch": 1.08, + "learning_rate": 1.6430365615406084e-05, + "loss": 0.9352, + "step": 16737 + }, + { + "epoch": 1.08, + "learning_rate": 1.6430150926383133e-05, + "loss": 0.927, + "step": 16738 + }, + { + "epoch": 1.08, + "learning_rate": 1.6429936237360186e-05, + "loss": 1.0066, + "step": 16739 + }, + { + "epoch": 1.08, + "learning_rate": 1.6429721548337235e-05, + "loss": 0.8954, + "step": 16740 + }, + { + "epoch": 1.08, + "learning_rate": 1.6429506859314284e-05, + "loss": 1.0798, + "step": 16741 + }, + { + "epoch": 1.08, + "learning_rate": 1.6429292170291332e-05, + "loss": 1.0001, + "step": 16742 + }, + { + "epoch": 1.08, + "learning_rate": 1.6429077481268385e-05, + "loss": 0.9381, + "step": 16743 + }, + { + "epoch": 1.08, + "learning_rate": 1.6428862792245434e-05, + "loss": 0.9145, + "step": 16744 + }, + { + "epoch": 1.08, + "learning_rate": 1.6428648103222483e-05, + "loss": 0.9619, + "step": 16745 + }, + { + "epoch": 1.08, + "learning_rate": 1.6428433414199535e-05, + "loss": 0.9302, + "step": 16746 + }, + { + "epoch": 1.08, + "learning_rate": 1.6428218725176584e-05, + "loss": 0.9608, + "step": 16747 + }, + { + "epoch": 1.08, + "learning_rate": 1.6428004036153633e-05, + "loss": 0.9973, + "step": 16748 + }, + { + "epoch": 1.08, + "learning_rate": 1.6427789347130682e-05, + "loss": 1.0016, + "step": 16749 + }, + { + "epoch": 1.08, + "learning_rate": 1.642757465810773e-05, + "loss": 0.9814, + "step": 16750 + }, + { + "epoch": 1.08, + "learning_rate": 1.6427359969084783e-05, + "loss": 0.9489, + "step": 16751 + }, + { + "epoch": 1.08, + "learning_rate": 1.6427145280061832e-05, + "loss": 0.9714, + "step": 16752 + }, + { + "epoch": 1.08, + "learning_rate": 1.642693059103888e-05, + "loss": 0.9652, + "step": 16753 + }, + { + "epoch": 1.08, + "learning_rate": 1.642671590201593e-05, + "loss": 0.9369, + "step": 16754 + }, + { + "epoch": 1.08, + "learning_rate": 1.642650121299298e-05, + "loss": 1.1071, + "step": 16755 + }, + { + "epoch": 1.08, + "learning_rate": 1.6426286523970028e-05, + "loss": 0.99, + "step": 16756 + }, + { + "epoch": 1.08, + "learning_rate": 1.642607183494708e-05, + "loss": 0.9469, + "step": 16757 + }, + { + "epoch": 1.08, + "learning_rate": 1.642585714592413e-05, + "loss": 0.9693, + "step": 16758 + }, + { + "epoch": 1.08, + "learning_rate": 1.6425642456901182e-05, + "loss": 1.0326, + "step": 16759 + }, + { + "epoch": 1.08, + "learning_rate": 1.642542776787823e-05, + "loss": 0.9954, + "step": 16760 + }, + { + "epoch": 1.08, + "learning_rate": 1.642521307885528e-05, + "loss": 1.0266, + "step": 16761 + }, + { + "epoch": 1.08, + "learning_rate": 1.642499838983233e-05, + "loss": 0.9457, + "step": 16762 + }, + { + "epoch": 1.08, + "learning_rate": 1.6424783700809378e-05, + "loss": 0.9904, + "step": 16763 + }, + { + "epoch": 1.08, + "learning_rate": 1.642456901178643e-05, + "loss": 1.011, + "step": 16764 + }, + { + "epoch": 1.08, + "learning_rate": 1.642435432276348e-05, + "loss": 1.0925, + "step": 16765 + }, + { + "epoch": 1.08, + "learning_rate": 1.6424139633740528e-05, + "loss": 0.9236, + "step": 16766 + }, + { + "epoch": 1.08, + "learning_rate": 1.6423924944717577e-05, + "loss": 0.8665, + "step": 16767 + }, + { + "epoch": 1.08, + "learning_rate": 1.6423710255694626e-05, + "loss": 0.924, + "step": 16768 + }, + { + "epoch": 1.08, + "learning_rate": 1.6423495566671678e-05, + "loss": 1.0225, + "step": 16769 + }, + { + "epoch": 1.08, + "learning_rate": 1.6423280877648727e-05, + "loss": 0.9695, + "step": 16770 + }, + { + "epoch": 1.08, + "learning_rate": 1.6423066188625776e-05, + "loss": 1.0128, + "step": 16771 + }, + { + "epoch": 1.08, + "learning_rate": 1.642285149960283e-05, + "loss": 0.9808, + "step": 16772 + }, + { + "epoch": 1.08, + "learning_rate": 1.6422636810579878e-05, + "loss": 0.9599, + "step": 16773 + }, + { + "epoch": 1.08, + "learning_rate": 1.6422422121556926e-05, + "loss": 1.0357, + "step": 16774 + }, + { + "epoch": 1.08, + "learning_rate": 1.6422207432533975e-05, + "loss": 0.9781, + "step": 16775 + }, + { + "epoch": 1.08, + "learning_rate": 1.6421992743511024e-05, + "loss": 1.0278, + "step": 16776 + }, + { + "epoch": 1.08, + "learning_rate": 1.6421778054488077e-05, + "loss": 0.9198, + "step": 16777 + }, + { + "epoch": 1.08, + "learning_rate": 1.6421563365465126e-05, + "loss": 0.9345, + "step": 16778 + }, + { + "epoch": 1.08, + "learning_rate": 1.6421348676442175e-05, + "loss": 0.9565, + "step": 16779 + }, + { + "epoch": 1.08, + "learning_rate": 1.6421133987419224e-05, + "loss": 0.9186, + "step": 16780 + }, + { + "epoch": 1.08, + "learning_rate": 1.6420919298396273e-05, + "loss": 0.8647, + "step": 16781 + }, + { + "epoch": 1.08, + "learning_rate": 1.6420704609373325e-05, + "loss": 0.9334, + "step": 16782 + }, + { + "epoch": 1.08, + "learning_rate": 1.6420489920350374e-05, + "loss": 1.0058, + "step": 16783 + }, + { + "epoch": 1.08, + "learning_rate": 1.6420275231327423e-05, + "loss": 0.9576, + "step": 16784 + }, + { + "epoch": 1.08, + "learning_rate": 1.6420060542304472e-05, + "loss": 0.9348, + "step": 16785 + }, + { + "epoch": 1.08, + "learning_rate": 1.6419845853281524e-05, + "loss": 0.9498, + "step": 16786 + }, + { + "epoch": 1.08, + "learning_rate": 1.6419631164258573e-05, + "loss": 0.9987, + "step": 16787 + }, + { + "epoch": 1.08, + "learning_rate": 1.6419416475235622e-05, + "loss": 0.9291, + "step": 16788 + }, + { + "epoch": 1.08, + "learning_rate": 1.6419201786212674e-05, + "loss": 0.9462, + "step": 16789 + }, + { + "epoch": 1.08, + "learning_rate": 1.6418987097189723e-05, + "loss": 1.0488, + "step": 16790 + }, + { + "epoch": 1.08, + "learning_rate": 1.6418772408166772e-05, + "loss": 1.0115, + "step": 16791 + }, + { + "epoch": 1.08, + "learning_rate": 1.641855771914382e-05, + "loss": 0.9265, + "step": 16792 + }, + { + "epoch": 1.08, + "learning_rate": 1.641834303012087e-05, + "loss": 1.0045, + "step": 16793 + }, + { + "epoch": 1.08, + "learning_rate": 1.641812834109792e-05, + "loss": 1.0009, + "step": 16794 + }, + { + "epoch": 1.08, + "learning_rate": 1.641791365207497e-05, + "loss": 1.028, + "step": 16795 + }, + { + "epoch": 1.08, + "learning_rate": 1.641769896305202e-05, + "loss": 1.0653, + "step": 16796 + }, + { + "epoch": 1.08, + "learning_rate": 1.641748427402907e-05, + "loss": 0.9002, + "step": 16797 + }, + { + "epoch": 1.08, + "learning_rate": 1.641726958500612e-05, + "loss": 1.0169, + "step": 16798 + }, + { + "epoch": 1.08, + "learning_rate": 1.6417054895983168e-05, + "loss": 0.9546, + "step": 16799 + }, + { + "epoch": 1.08, + "learning_rate": 1.641684020696022e-05, + "loss": 1.003, + "step": 16800 + }, + { + "epoch": 1.08, + "learning_rate": 1.641662551793727e-05, + "loss": 0.9743, + "step": 16801 + }, + { + "epoch": 1.08, + "learning_rate": 1.641641082891432e-05, + "loss": 0.8985, + "step": 16802 + }, + { + "epoch": 1.08, + "learning_rate": 1.641619613989137e-05, + "loss": 0.9444, + "step": 16803 + }, + { + "epoch": 1.08, + "learning_rate": 1.641598145086842e-05, + "loss": 0.9229, + "step": 16804 + }, + { + "epoch": 1.08, + "learning_rate": 1.6415766761845468e-05, + "loss": 0.9103, + "step": 16805 + }, + { + "epoch": 1.08, + "learning_rate": 1.6415552072822517e-05, + "loss": 0.9515, + "step": 16806 + }, + { + "epoch": 1.08, + "learning_rate": 1.641533738379957e-05, + "loss": 1.0009, + "step": 16807 + }, + { + "epoch": 1.08, + "learning_rate": 1.641512269477662e-05, + "loss": 0.9855, + "step": 16808 + }, + { + "epoch": 1.08, + "learning_rate": 1.6414908005753667e-05, + "loss": 1.0623, + "step": 16809 + }, + { + "epoch": 1.08, + "learning_rate": 1.6414693316730716e-05, + "loss": 0.9905, + "step": 16810 + }, + { + "epoch": 1.08, + "learning_rate": 1.6414478627707765e-05, + "loss": 1.0026, + "step": 16811 + }, + { + "epoch": 1.08, + "learning_rate": 1.6414263938684814e-05, + "loss": 1.0115, + "step": 16812 + }, + { + "epoch": 1.08, + "learning_rate": 1.6414049249661867e-05, + "loss": 0.9329, + "step": 16813 + }, + { + "epoch": 1.08, + "learning_rate": 1.6413834560638916e-05, + "loss": 1.0016, + "step": 16814 + }, + { + "epoch": 1.08, + "learning_rate": 1.6413619871615968e-05, + "loss": 1.045, + "step": 16815 + }, + { + "epoch": 1.08, + "learning_rate": 1.6413405182593017e-05, + "loss": 0.9878, + "step": 16816 + }, + { + "epoch": 1.08, + "learning_rate": 1.6413190493570066e-05, + "loss": 0.9542, + "step": 16817 + }, + { + "epoch": 1.08, + "learning_rate": 1.6412975804547115e-05, + "loss": 1.011, + "step": 16818 + }, + { + "epoch": 1.08, + "learning_rate": 1.6412761115524164e-05, + "loss": 1.0193, + "step": 16819 + }, + { + "epoch": 1.08, + "learning_rate": 1.6412546426501216e-05, + "loss": 0.9195, + "step": 16820 + }, + { + "epoch": 1.08, + "learning_rate": 1.6412331737478265e-05, + "loss": 0.9398, + "step": 16821 + }, + { + "epoch": 1.08, + "learning_rate": 1.6412117048455314e-05, + "loss": 0.9573, + "step": 16822 + }, + { + "epoch": 1.08, + "learning_rate": 1.6411902359432363e-05, + "loss": 1.0741, + "step": 16823 + }, + { + "epoch": 1.08, + "learning_rate": 1.6411687670409412e-05, + "loss": 0.9488, + "step": 16824 + }, + { + "epoch": 1.08, + "learning_rate": 1.641147298138646e-05, + "loss": 1.0153, + "step": 16825 + }, + { + "epoch": 1.08, + "learning_rate": 1.6411258292363513e-05, + "loss": 0.9693, + "step": 16826 + }, + { + "epoch": 1.08, + "learning_rate": 1.6411043603340562e-05, + "loss": 0.9172, + "step": 16827 + }, + { + "epoch": 1.08, + "learning_rate": 1.641082891431761e-05, + "loss": 0.9494, + "step": 16828 + }, + { + "epoch": 1.08, + "learning_rate": 1.641061422529466e-05, + "loss": 0.9511, + "step": 16829 + }, + { + "epoch": 1.08, + "learning_rate": 1.6410399536271713e-05, + "loss": 1.0062, + "step": 16830 + }, + { + "epoch": 1.08, + "learning_rate": 1.641018484724876e-05, + "loss": 1.0121, + "step": 16831 + }, + { + "epoch": 1.08, + "learning_rate": 1.640997015822581e-05, + "loss": 0.9378, + "step": 16832 + }, + { + "epoch": 1.08, + "learning_rate": 1.6409755469202863e-05, + "loss": 0.9696, + "step": 16833 + }, + { + "epoch": 1.08, + "learning_rate": 1.6409540780179912e-05, + "loss": 0.9739, + "step": 16834 + }, + { + "epoch": 1.08, + "learning_rate": 1.640932609115696e-05, + "loss": 0.9585, + "step": 16835 + }, + { + "epoch": 1.08, + "learning_rate": 1.640911140213401e-05, + "loss": 1.0682, + "step": 16836 + }, + { + "epoch": 1.08, + "learning_rate": 1.640889671311106e-05, + "loss": 0.9479, + "step": 16837 + }, + { + "epoch": 1.08, + "learning_rate": 1.640868202408811e-05, + "loss": 1.0338, + "step": 16838 + }, + { + "epoch": 1.08, + "learning_rate": 1.640846733506516e-05, + "loss": 0.9752, + "step": 16839 + }, + { + "epoch": 1.08, + "learning_rate": 1.640825264604221e-05, + "loss": 0.9414, + "step": 16840 + }, + { + "epoch": 1.08, + "learning_rate": 1.6408037957019258e-05, + "loss": 0.9613, + "step": 16841 + }, + { + "epoch": 1.08, + "learning_rate": 1.6407823267996307e-05, + "loss": 0.9568, + "step": 16842 + }, + { + "epoch": 1.08, + "learning_rate": 1.6407608578973356e-05, + "loss": 1.0358, + "step": 16843 + }, + { + "epoch": 1.08, + "learning_rate": 1.6407393889950408e-05, + "loss": 1.0325, + "step": 16844 + }, + { + "epoch": 1.08, + "learning_rate": 1.6407179200927457e-05, + "loss": 0.9733, + "step": 16845 + }, + { + "epoch": 1.08, + "learning_rate": 1.640696451190451e-05, + "loss": 0.905, + "step": 16846 + }, + { + "epoch": 1.08, + "learning_rate": 1.640674982288156e-05, + "loss": 0.8976, + "step": 16847 + }, + { + "epoch": 1.08, + "learning_rate": 1.6406535133858607e-05, + "loss": 0.8908, + "step": 16848 + }, + { + "epoch": 1.08, + "learning_rate": 1.6406320444835656e-05, + "loss": 1.0189, + "step": 16849 + }, + { + "epoch": 1.08, + "learning_rate": 1.6406105755812705e-05, + "loss": 0.951, + "step": 16850 + }, + { + "epoch": 1.08, + "learning_rate": 1.6405891066789758e-05, + "loss": 0.9732, + "step": 16851 + }, + { + "epoch": 1.08, + "learning_rate": 1.6405676377766807e-05, + "loss": 0.9771, + "step": 16852 + }, + { + "epoch": 1.08, + "learning_rate": 1.6405461688743856e-05, + "loss": 1.0174, + "step": 16853 + }, + { + "epoch": 1.08, + "learning_rate": 1.6405246999720905e-05, + "loss": 0.9736, + "step": 16854 + }, + { + "epoch": 1.08, + "learning_rate": 1.6405032310697954e-05, + "loss": 1.0116, + "step": 16855 + }, + { + "epoch": 1.08, + "learning_rate": 1.6404817621675006e-05, + "loss": 0.9621, + "step": 16856 + }, + { + "epoch": 1.08, + "learning_rate": 1.6404602932652055e-05, + "loss": 0.9443, + "step": 16857 + }, + { + "epoch": 1.08, + "learning_rate": 1.6404388243629104e-05, + "loss": 0.9385, + "step": 16858 + }, + { + "epoch": 1.08, + "learning_rate": 1.6404173554606156e-05, + "loss": 0.9582, + "step": 16859 + }, + { + "epoch": 1.08, + "learning_rate": 1.6403958865583205e-05, + "loss": 1.09, + "step": 16860 + }, + { + "epoch": 1.08, + "learning_rate": 1.6403744176560254e-05, + "loss": 0.9605, + "step": 16861 + }, + { + "epoch": 1.08, + "learning_rate": 1.6403529487537303e-05, + "loss": 0.9832, + "step": 16862 + }, + { + "epoch": 1.08, + "learning_rate": 1.6403314798514352e-05, + "loss": 0.9634, + "step": 16863 + }, + { + "epoch": 1.08, + "learning_rate": 1.6403100109491404e-05, + "loss": 0.9916, + "step": 16864 + }, + { + "epoch": 1.09, + "learning_rate": 1.6402885420468453e-05, + "loss": 0.9653, + "step": 16865 + }, + { + "epoch": 1.09, + "learning_rate": 1.6402670731445502e-05, + "loss": 0.9521, + "step": 16866 + }, + { + "epoch": 1.09, + "learning_rate": 1.640245604242255e-05, + "loss": 0.9809, + "step": 16867 + }, + { + "epoch": 1.09, + "learning_rate": 1.64022413533996e-05, + "loss": 0.9853, + "step": 16868 + }, + { + "epoch": 1.09, + "learning_rate": 1.6402026664376653e-05, + "loss": 0.9915, + "step": 16869 + }, + { + "epoch": 1.09, + "learning_rate": 1.64018119753537e-05, + "loss": 0.9717, + "step": 16870 + }, + { + "epoch": 1.09, + "learning_rate": 1.640159728633075e-05, + "loss": 1.0015, + "step": 16871 + }, + { + "epoch": 1.09, + "learning_rate": 1.64013825973078e-05, + "loss": 0.9737, + "step": 16872 + }, + { + "epoch": 1.09, + "learning_rate": 1.6401167908284852e-05, + "loss": 0.8897, + "step": 16873 + }, + { + "epoch": 1.09, + "learning_rate": 1.64009532192619e-05, + "loss": 0.919, + "step": 16874 + }, + { + "epoch": 1.09, + "learning_rate": 1.640073853023895e-05, + "loss": 1.0273, + "step": 16875 + }, + { + "epoch": 1.09, + "learning_rate": 1.6400523841216002e-05, + "loss": 1.0718, + "step": 16876 + }, + { + "epoch": 1.09, + "learning_rate": 1.640030915219305e-05, + "loss": 0.9289, + "step": 16877 + }, + { + "epoch": 1.09, + "learning_rate": 1.64000944631701e-05, + "loss": 0.9159, + "step": 16878 + }, + { + "epoch": 1.09, + "learning_rate": 1.639987977414715e-05, + "loss": 1.0228, + "step": 16879 + }, + { + "epoch": 1.09, + "learning_rate": 1.6399665085124198e-05, + "loss": 1.0388, + "step": 16880 + }, + { + "epoch": 1.09, + "learning_rate": 1.6399450396101247e-05, + "loss": 0.9105, + "step": 16881 + }, + { + "epoch": 1.09, + "learning_rate": 1.63992357070783e-05, + "loss": 0.9777, + "step": 16882 + }, + { + "epoch": 1.09, + "learning_rate": 1.639902101805535e-05, + "loss": 1.0322, + "step": 16883 + }, + { + "epoch": 1.09, + "learning_rate": 1.6398806329032397e-05, + "loss": 0.9919, + "step": 16884 + }, + { + "epoch": 1.09, + "learning_rate": 1.6398591640009446e-05, + "loss": 0.9687, + "step": 16885 + }, + { + "epoch": 1.09, + "learning_rate": 1.6398376950986495e-05, + "loss": 1.0116, + "step": 16886 + }, + { + "epoch": 1.09, + "learning_rate": 1.6398162261963548e-05, + "loss": 1.0362, + "step": 16887 + }, + { + "epoch": 1.09, + "learning_rate": 1.6397947572940597e-05, + "loss": 0.9421, + "step": 16888 + }, + { + "epoch": 1.09, + "learning_rate": 1.639773288391765e-05, + "loss": 0.9054, + "step": 16889 + }, + { + "epoch": 1.09, + "learning_rate": 1.6397518194894698e-05, + "loss": 1.0835, + "step": 16890 + }, + { + "epoch": 1.09, + "learning_rate": 1.6397303505871747e-05, + "loss": 0.9535, + "step": 16891 + }, + { + "epoch": 1.09, + "learning_rate": 1.6397088816848796e-05, + "loss": 1.041, + "step": 16892 + }, + { + "epoch": 1.09, + "learning_rate": 1.6396874127825845e-05, + "loss": 0.9752, + "step": 16893 + }, + { + "epoch": 1.09, + "learning_rate": 1.6396659438802897e-05, + "loss": 0.9812, + "step": 16894 + }, + { + "epoch": 1.09, + "learning_rate": 1.6396444749779946e-05, + "loss": 0.948, + "step": 16895 + }, + { + "epoch": 1.09, + "learning_rate": 1.6396230060756995e-05, + "loss": 0.9629, + "step": 16896 + }, + { + "epoch": 1.09, + "learning_rate": 1.6396015371734044e-05, + "loss": 0.9816, + "step": 16897 + }, + { + "epoch": 1.09, + "learning_rate": 1.6395800682711093e-05, + "loss": 0.9916, + "step": 16898 + }, + { + "epoch": 1.09, + "learning_rate": 1.6395585993688142e-05, + "loss": 1.0429, + "step": 16899 + }, + { + "epoch": 1.09, + "learning_rate": 1.6395371304665194e-05, + "loss": 0.9907, + "step": 16900 + }, + { + "epoch": 1.09, + "learning_rate": 1.6395156615642243e-05, + "loss": 0.9628, + "step": 16901 + }, + { + "epoch": 1.09, + "learning_rate": 1.6394941926619296e-05, + "loss": 0.9969, + "step": 16902 + }, + { + "epoch": 1.09, + "learning_rate": 1.6394727237596345e-05, + "loss": 1.0004, + "step": 16903 + }, + { + "epoch": 1.09, + "learning_rate": 1.6394512548573394e-05, + "loss": 1.0279, + "step": 16904 + }, + { + "epoch": 1.09, + "learning_rate": 1.6394297859550443e-05, + "loss": 0.9583, + "step": 16905 + }, + { + "epoch": 1.09, + "learning_rate": 1.639408317052749e-05, + "loss": 0.9596, + "step": 16906 + }, + { + "epoch": 1.09, + "learning_rate": 1.6393868481504544e-05, + "loss": 1.0433, + "step": 16907 + }, + { + "epoch": 1.09, + "learning_rate": 1.6393653792481593e-05, + "loss": 0.9181, + "step": 16908 + }, + { + "epoch": 1.09, + "learning_rate": 1.6393439103458642e-05, + "loss": 0.9045, + "step": 16909 + }, + { + "epoch": 1.09, + "learning_rate": 1.639322441443569e-05, + "loss": 1.0714, + "step": 16910 + }, + { + "epoch": 1.09, + "learning_rate": 1.639300972541274e-05, + "loss": 0.9387, + "step": 16911 + }, + { + "epoch": 1.09, + "learning_rate": 1.639279503638979e-05, + "loss": 1.0419, + "step": 16912 + }, + { + "epoch": 1.09, + "learning_rate": 1.639258034736684e-05, + "loss": 0.9595, + "step": 16913 + }, + { + "epoch": 1.09, + "learning_rate": 1.639236565834389e-05, + "loss": 0.9776, + "step": 16914 + }, + { + "epoch": 1.09, + "learning_rate": 1.639215096932094e-05, + "loss": 0.9681, + "step": 16915 + }, + { + "epoch": 1.09, + "learning_rate": 1.6391936280297988e-05, + "loss": 0.9641, + "step": 16916 + }, + { + "epoch": 1.09, + "learning_rate": 1.639172159127504e-05, + "loss": 0.9824, + "step": 16917 + }, + { + "epoch": 1.09, + "learning_rate": 1.639150690225209e-05, + "loss": 0.8947, + "step": 16918 + }, + { + "epoch": 1.09, + "learning_rate": 1.6391292213229138e-05, + "loss": 0.9112, + "step": 16919 + }, + { + "epoch": 1.09, + "learning_rate": 1.639107752420619e-05, + "loss": 0.9203, + "step": 16920 + }, + { + "epoch": 1.09, + "learning_rate": 1.639086283518324e-05, + "loss": 0.9835, + "step": 16921 + }, + { + "epoch": 1.09, + "learning_rate": 1.639064814616029e-05, + "loss": 0.9966, + "step": 16922 + }, + { + "epoch": 1.09, + "learning_rate": 1.6390433457137337e-05, + "loss": 1.0604, + "step": 16923 + }, + { + "epoch": 1.09, + "learning_rate": 1.6390218768114386e-05, + "loss": 0.9816, + "step": 16924 + }, + { + "epoch": 1.09, + "learning_rate": 1.639000407909144e-05, + "loss": 0.9371, + "step": 16925 + }, + { + "epoch": 1.09, + "learning_rate": 1.6389789390068488e-05, + "loss": 0.9757, + "step": 16926 + }, + { + "epoch": 1.09, + "learning_rate": 1.6389574701045537e-05, + "loss": 1.0502, + "step": 16927 + }, + { + "epoch": 1.09, + "learning_rate": 1.6389360012022586e-05, + "loss": 0.8945, + "step": 16928 + }, + { + "epoch": 1.09, + "learning_rate": 1.6389145322999635e-05, + "loss": 0.9403, + "step": 16929 + }, + { + "epoch": 1.09, + "learning_rate": 1.6388930633976684e-05, + "loss": 0.9591, + "step": 16930 + }, + { + "epoch": 1.09, + "learning_rate": 1.6388715944953736e-05, + "loss": 0.9861, + "step": 16931 + }, + { + "epoch": 1.09, + "learning_rate": 1.6388501255930785e-05, + "loss": 0.9509, + "step": 16932 + }, + { + "epoch": 1.09, + "learning_rate": 1.6388286566907837e-05, + "loss": 0.9901, + "step": 16933 + }, + { + "epoch": 1.09, + "learning_rate": 1.6388071877884886e-05, + "loss": 0.8895, + "step": 16934 + }, + { + "epoch": 1.09, + "learning_rate": 1.6387857188861935e-05, + "loss": 0.9485, + "step": 16935 + }, + { + "epoch": 1.09, + "learning_rate": 1.6387642499838984e-05, + "loss": 1.0741, + "step": 16936 + }, + { + "epoch": 1.09, + "learning_rate": 1.6387427810816033e-05, + "loss": 1.0295, + "step": 16937 + }, + { + "epoch": 1.09, + "learning_rate": 1.6387213121793085e-05, + "loss": 0.9706, + "step": 16938 + }, + { + "epoch": 1.09, + "learning_rate": 1.6386998432770134e-05, + "loss": 0.9757, + "step": 16939 + }, + { + "epoch": 1.09, + "learning_rate": 1.6386783743747183e-05, + "loss": 1.0348, + "step": 16940 + }, + { + "epoch": 1.09, + "learning_rate": 1.6386569054724232e-05, + "loss": 0.9957, + "step": 16941 + }, + { + "epoch": 1.09, + "learning_rate": 1.638635436570128e-05, + "loss": 1.0065, + "step": 16942 + }, + { + "epoch": 1.09, + "learning_rate": 1.6386139676678334e-05, + "loss": 1.0279, + "step": 16943 + }, + { + "epoch": 1.09, + "learning_rate": 1.6385924987655383e-05, + "loss": 1.0365, + "step": 16944 + }, + { + "epoch": 1.09, + "learning_rate": 1.638571029863243e-05, + "loss": 1.0303, + "step": 16945 + }, + { + "epoch": 1.09, + "learning_rate": 1.6385495609609484e-05, + "loss": 0.9978, + "step": 16946 + }, + { + "epoch": 1.09, + "learning_rate": 1.6385280920586533e-05, + "loss": 0.9964, + "step": 16947 + }, + { + "epoch": 1.09, + "learning_rate": 1.6385066231563582e-05, + "loss": 0.8479, + "step": 16948 + }, + { + "epoch": 1.09, + "learning_rate": 1.638485154254063e-05, + "loss": 0.832, + "step": 16949 + }, + { + "epoch": 1.09, + "learning_rate": 1.638463685351768e-05, + "loss": 1.056, + "step": 16950 + }, + { + "epoch": 1.09, + "learning_rate": 1.6384422164494732e-05, + "loss": 0.921, + "step": 16951 + }, + { + "epoch": 1.09, + "learning_rate": 1.638420747547178e-05, + "loss": 0.937, + "step": 16952 + }, + { + "epoch": 1.09, + "learning_rate": 1.638399278644883e-05, + "loss": 0.9529, + "step": 16953 + }, + { + "epoch": 1.09, + "learning_rate": 1.638377809742588e-05, + "loss": 0.9507, + "step": 16954 + }, + { + "epoch": 1.09, + "learning_rate": 1.6383563408402928e-05, + "loss": 1.0054, + "step": 16955 + }, + { + "epoch": 1.09, + "learning_rate": 1.638334871937998e-05, + "loss": 1.0139, + "step": 16956 + }, + { + "epoch": 1.09, + "learning_rate": 1.638313403035703e-05, + "loss": 0.9275, + "step": 16957 + }, + { + "epoch": 1.09, + "learning_rate": 1.638291934133408e-05, + "loss": 0.9217, + "step": 16958 + }, + { + "epoch": 1.09, + "learning_rate": 1.6382704652311127e-05, + "loss": 0.9583, + "step": 16959 + }, + { + "epoch": 1.09, + "learning_rate": 1.6382489963288176e-05, + "loss": 1.0488, + "step": 16960 + }, + { + "epoch": 1.09, + "learning_rate": 1.638227527426523e-05, + "loss": 1.0476, + "step": 16961 + }, + { + "epoch": 1.09, + "learning_rate": 1.6382060585242278e-05, + "loss": 1.0184, + "step": 16962 + }, + { + "epoch": 1.09, + "learning_rate": 1.638184589621933e-05, + "loss": 0.9809, + "step": 16963 + }, + { + "epoch": 1.09, + "learning_rate": 1.638163120719638e-05, + "loss": 1.0487, + "step": 16964 + }, + { + "epoch": 1.09, + "learning_rate": 1.6381416518173428e-05, + "loss": 0.934, + "step": 16965 + }, + { + "epoch": 1.09, + "learning_rate": 1.6381201829150477e-05, + "loss": 0.9122, + "step": 16966 + }, + { + "epoch": 1.09, + "learning_rate": 1.6380987140127526e-05, + "loss": 0.9776, + "step": 16967 + }, + { + "epoch": 1.09, + "learning_rate": 1.6380772451104575e-05, + "loss": 1.0119, + "step": 16968 + }, + { + "epoch": 1.09, + "learning_rate": 1.6380557762081627e-05, + "loss": 1.0061, + "step": 16969 + }, + { + "epoch": 1.09, + "learning_rate": 1.6380343073058676e-05, + "loss": 0.9589, + "step": 16970 + }, + { + "epoch": 1.09, + "learning_rate": 1.6380128384035725e-05, + "loss": 0.9445, + "step": 16971 + }, + { + "epoch": 1.09, + "learning_rate": 1.6379913695012774e-05, + "loss": 0.9981, + "step": 16972 + }, + { + "epoch": 1.09, + "learning_rate": 1.6379699005989823e-05, + "loss": 0.9398, + "step": 16973 + }, + { + "epoch": 1.09, + "learning_rate": 1.6379484316966875e-05, + "loss": 0.9896, + "step": 16974 + }, + { + "epoch": 1.09, + "learning_rate": 1.6379269627943924e-05, + "loss": 0.9929, + "step": 16975 + }, + { + "epoch": 1.09, + "learning_rate": 1.6379054938920977e-05, + "loss": 0.9675, + "step": 16976 + }, + { + "epoch": 1.09, + "learning_rate": 1.6378840249898026e-05, + "loss": 0.9195, + "step": 16977 + }, + { + "epoch": 1.09, + "learning_rate": 1.6378625560875075e-05, + "loss": 1.0063, + "step": 16978 + }, + { + "epoch": 1.09, + "learning_rate": 1.6378410871852124e-05, + "loss": 0.9952, + "step": 16979 + }, + { + "epoch": 1.09, + "learning_rate": 1.6378196182829172e-05, + "loss": 1.0398, + "step": 16980 + }, + { + "epoch": 1.09, + "learning_rate": 1.6377981493806225e-05, + "loss": 0.9364, + "step": 16981 + }, + { + "epoch": 1.09, + "learning_rate": 1.6377766804783274e-05, + "loss": 0.9721, + "step": 16982 + }, + { + "epoch": 1.09, + "learning_rate": 1.6377552115760323e-05, + "loss": 0.9697, + "step": 16983 + }, + { + "epoch": 1.09, + "learning_rate": 1.6377337426737372e-05, + "loss": 0.9865, + "step": 16984 + }, + { + "epoch": 1.09, + "learning_rate": 1.637712273771442e-05, + "loss": 0.978, + "step": 16985 + }, + { + "epoch": 1.09, + "learning_rate": 1.637690804869147e-05, + "loss": 0.948, + "step": 16986 + }, + { + "epoch": 1.09, + "learning_rate": 1.6376693359668522e-05, + "loss": 0.9623, + "step": 16987 + }, + { + "epoch": 1.09, + "learning_rate": 1.637647867064557e-05, + "loss": 0.9693, + "step": 16988 + }, + { + "epoch": 1.09, + "learning_rate": 1.637626398162262e-05, + "loss": 1.0322, + "step": 16989 + }, + { + "epoch": 1.09, + "learning_rate": 1.6376049292599672e-05, + "loss": 0.9701, + "step": 16990 + }, + { + "epoch": 1.09, + "learning_rate": 1.637583460357672e-05, + "loss": 0.9401, + "step": 16991 + }, + { + "epoch": 1.09, + "learning_rate": 1.637561991455377e-05, + "loss": 1.0143, + "step": 16992 + }, + { + "epoch": 1.09, + "learning_rate": 1.637540522553082e-05, + "loss": 0.927, + "step": 16993 + }, + { + "epoch": 1.09, + "learning_rate": 1.637519053650787e-05, + "loss": 0.9502, + "step": 16994 + }, + { + "epoch": 1.09, + "learning_rate": 1.637497584748492e-05, + "loss": 1.0659, + "step": 16995 + }, + { + "epoch": 1.09, + "learning_rate": 1.637476115846197e-05, + "loss": 0.9455, + "step": 16996 + }, + { + "epoch": 1.09, + "learning_rate": 1.637454646943902e-05, + "loss": 1.0202, + "step": 16997 + }, + { + "epoch": 1.09, + "learning_rate": 1.6374331780416067e-05, + "loss": 0.9791, + "step": 16998 + }, + { + "epoch": 1.09, + "learning_rate": 1.6374117091393116e-05, + "loss": 0.9511, + "step": 16999 + }, + { + "epoch": 1.09, + "learning_rate": 1.637390240237017e-05, + "loss": 0.9972, + "step": 17000 + }, + { + "epoch": 1.09, + "eval_laion/cider": 0.5685231075302235, + "eval_laion/meteor": 0.39124563477360685, + "eval_laion/pred_num_words": 38.56640625, + "eval_laion/sacrebleu": 15.72776834567261, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.066876769065857, + "eval_runtime": 101.542, + "eval_samples_per_second": 4.914, + "eval_steps_per_second": 0.315, + "step": 17000 + }, + { + "epoch": 1.09, + "learning_rate": 1.6373687713347218e-05, + "loss": 0.9583, + "step": 17001 + }, + { + "epoch": 1.09, + "learning_rate": 1.6373473024324267e-05, + "loss": 0.9927, + "step": 17002 + }, + { + "epoch": 1.09, + "learning_rate": 1.6373258335301316e-05, + "loss": 0.9976, + "step": 17003 + }, + { + "epoch": 1.09, + "learning_rate": 1.6373043646278368e-05, + "loss": 0.946, + "step": 17004 + }, + { + "epoch": 1.09, + "learning_rate": 1.6372828957255417e-05, + "loss": 0.9644, + "step": 17005 + }, + { + "epoch": 1.09, + "learning_rate": 1.6372614268232466e-05, + "loss": 0.9664, + "step": 17006 + }, + { + "epoch": 1.09, + "learning_rate": 1.6372399579209518e-05, + "loss": 0.8918, + "step": 17007 + }, + { + "epoch": 1.09, + "learning_rate": 1.6372184890186567e-05, + "loss": 0.9913, + "step": 17008 + }, + { + "epoch": 1.09, + "learning_rate": 1.6371970201163616e-05, + "loss": 0.9692, + "step": 17009 + }, + { + "epoch": 1.09, + "learning_rate": 1.6371755512140665e-05, + "loss": 1.0513, + "step": 17010 + }, + { + "epoch": 1.09, + "learning_rate": 1.6371540823117714e-05, + "loss": 0.9901, + "step": 17011 + }, + { + "epoch": 1.09, + "learning_rate": 1.6371326134094766e-05, + "loss": 1.0099, + "step": 17012 + }, + { + "epoch": 1.09, + "learning_rate": 1.6371111445071815e-05, + "loss": 0.9622, + "step": 17013 + }, + { + "epoch": 1.09, + "learning_rate": 1.6370896756048864e-05, + "loss": 0.9432, + "step": 17014 + }, + { + "epoch": 1.09, + "learning_rate": 1.6370682067025913e-05, + "loss": 0.9575, + "step": 17015 + }, + { + "epoch": 1.09, + "learning_rate": 1.6370467378002962e-05, + "loss": 1.0038, + "step": 17016 + }, + { + "epoch": 1.09, + "learning_rate": 1.637025268898001e-05, + "loss": 0.9833, + "step": 17017 + }, + { + "epoch": 1.09, + "learning_rate": 1.6370037999957064e-05, + "loss": 0.9181, + "step": 17018 + }, + { + "epoch": 1.09, + "learning_rate": 1.6369823310934113e-05, + "loss": 0.986, + "step": 17019 + }, + { + "epoch": 1.1, + "learning_rate": 1.6369608621911165e-05, + "loss": 0.9022, + "step": 17020 + }, + { + "epoch": 1.1, + "learning_rate": 1.6369393932888214e-05, + "loss": 0.9751, + "step": 17021 + }, + { + "epoch": 1.1, + "learning_rate": 1.6369179243865263e-05, + "loss": 0.9979, + "step": 17022 + }, + { + "epoch": 1.1, + "learning_rate": 1.6368964554842312e-05, + "loss": 0.9657, + "step": 17023 + }, + { + "epoch": 1.1, + "learning_rate": 1.636874986581936e-05, + "loss": 1.0242, + "step": 17024 + }, + { + "epoch": 1.1, + "learning_rate": 1.6368535176796413e-05, + "loss": 0.9931, + "step": 17025 + }, + { + "epoch": 1.1, + "learning_rate": 1.6368320487773462e-05, + "loss": 1.0242, + "step": 17026 + }, + { + "epoch": 1.1, + "learning_rate": 1.636810579875051e-05, + "loss": 0.9316, + "step": 17027 + }, + { + "epoch": 1.1, + "learning_rate": 1.636789110972756e-05, + "loss": 0.9915, + "step": 17028 + }, + { + "epoch": 1.1, + "learning_rate": 1.636767642070461e-05, + "loss": 1.0276, + "step": 17029 + }, + { + "epoch": 1.1, + "learning_rate": 1.636746173168166e-05, + "loss": 1.0104, + "step": 17030 + }, + { + "epoch": 1.1, + "learning_rate": 1.636724704265871e-05, + "loss": 1.0075, + "step": 17031 + }, + { + "epoch": 1.1, + "learning_rate": 1.636703235363576e-05, + "loss": 0.9968, + "step": 17032 + }, + { + "epoch": 1.1, + "learning_rate": 1.636681766461281e-05, + "loss": 0.9349, + "step": 17033 + }, + { + "epoch": 1.1, + "learning_rate": 1.636660297558986e-05, + "loss": 0.9771, + "step": 17034 + }, + { + "epoch": 1.1, + "learning_rate": 1.636638828656691e-05, + "loss": 0.9771, + "step": 17035 + }, + { + "epoch": 1.1, + "learning_rate": 1.636617359754396e-05, + "loss": 0.9638, + "step": 17036 + }, + { + "epoch": 1.1, + "learning_rate": 1.6365958908521008e-05, + "loss": 0.9559, + "step": 17037 + }, + { + "epoch": 1.1, + "learning_rate": 1.636574421949806e-05, + "loss": 1.0214, + "step": 17038 + }, + { + "epoch": 1.1, + "learning_rate": 1.636552953047511e-05, + "loss": 0.9884, + "step": 17039 + }, + { + "epoch": 1.1, + "learning_rate": 1.6365314841452158e-05, + "loss": 0.9783, + "step": 17040 + }, + { + "epoch": 1.1, + "learning_rate": 1.6365100152429207e-05, + "loss": 1.0365, + "step": 17041 + }, + { + "epoch": 1.1, + "learning_rate": 1.6364885463406256e-05, + "loss": 0.9796, + "step": 17042 + }, + { + "epoch": 1.1, + "learning_rate": 1.6364670774383308e-05, + "loss": 0.9474, + "step": 17043 + }, + { + "epoch": 1.1, + "learning_rate": 1.6364456085360357e-05, + "loss": 0.873, + "step": 17044 + }, + { + "epoch": 1.1, + "learning_rate": 1.6364241396337406e-05, + "loss": 0.962, + "step": 17045 + }, + { + "epoch": 1.1, + "learning_rate": 1.6364026707314455e-05, + "loss": 0.893, + "step": 17046 + }, + { + "epoch": 1.1, + "learning_rate": 1.6363812018291504e-05, + "loss": 0.9119, + "step": 17047 + }, + { + "epoch": 1.1, + "learning_rate": 1.6363597329268556e-05, + "loss": 0.9531, + "step": 17048 + }, + { + "epoch": 1.1, + "learning_rate": 1.6363382640245605e-05, + "loss": 0.8968, + "step": 17049 + }, + { + "epoch": 1.1, + "learning_rate": 1.6363167951222658e-05, + "loss": 0.8964, + "step": 17050 + }, + { + "epoch": 1.1, + "learning_rate": 1.6362953262199707e-05, + "loss": 0.933, + "step": 17051 + }, + { + "epoch": 1.1, + "learning_rate": 1.6362738573176756e-05, + "loss": 0.9848, + "step": 17052 + }, + { + "epoch": 1.1, + "learning_rate": 1.6362523884153805e-05, + "loss": 1.0425, + "step": 17053 + }, + { + "epoch": 1.1, + "learning_rate": 1.6362309195130853e-05, + "loss": 1.0192, + "step": 17054 + }, + { + "epoch": 1.1, + "learning_rate": 1.6362094506107902e-05, + "loss": 0.9961, + "step": 17055 + }, + { + "epoch": 1.1, + "learning_rate": 1.6361879817084955e-05, + "loss": 0.9132, + "step": 17056 + }, + { + "epoch": 1.1, + "learning_rate": 1.6361665128062004e-05, + "loss": 0.9801, + "step": 17057 + }, + { + "epoch": 1.1, + "learning_rate": 1.6361450439039053e-05, + "loss": 0.99, + "step": 17058 + }, + { + "epoch": 1.1, + "learning_rate": 1.6361235750016102e-05, + "loss": 0.994, + "step": 17059 + }, + { + "epoch": 1.1, + "learning_rate": 1.636102106099315e-05, + "loss": 1.0164, + "step": 17060 + }, + { + "epoch": 1.1, + "learning_rate": 1.6360806371970203e-05, + "loss": 1.0102, + "step": 17061 + }, + { + "epoch": 1.1, + "learning_rate": 1.6360591682947252e-05, + "loss": 0.9746, + "step": 17062 + }, + { + "epoch": 1.1, + "learning_rate": 1.6360376993924304e-05, + "loss": 0.9566, + "step": 17063 + }, + { + "epoch": 1.1, + "learning_rate": 1.6360162304901353e-05, + "loss": 1.0101, + "step": 17064 + }, + { + "epoch": 1.1, + "learning_rate": 1.6359947615878402e-05, + "loss": 1.0129, + "step": 17065 + }, + { + "epoch": 1.1, + "learning_rate": 1.635973292685545e-05, + "loss": 0.9581, + "step": 17066 + }, + { + "epoch": 1.1, + "learning_rate": 1.63595182378325e-05, + "loss": 0.9365, + "step": 17067 + }, + { + "epoch": 1.1, + "learning_rate": 1.635930354880955e-05, + "loss": 1.0002, + "step": 17068 + }, + { + "epoch": 1.1, + "learning_rate": 1.63590888597866e-05, + "loss": 0.9919, + "step": 17069 + }, + { + "epoch": 1.1, + "learning_rate": 1.635887417076365e-05, + "loss": 0.9869, + "step": 17070 + }, + { + "epoch": 1.1, + "learning_rate": 1.63586594817407e-05, + "loss": 1.0005, + "step": 17071 + }, + { + "epoch": 1.1, + "learning_rate": 1.635844479271775e-05, + "loss": 1.027, + "step": 17072 + }, + { + "epoch": 1.1, + "learning_rate": 1.6358230103694797e-05, + "loss": 1.0396, + "step": 17073 + }, + { + "epoch": 1.1, + "learning_rate": 1.635801541467185e-05, + "loss": 0.9822, + "step": 17074 + }, + { + "epoch": 1.1, + "learning_rate": 1.63578007256489e-05, + "loss": 0.9544, + "step": 17075 + }, + { + "epoch": 1.1, + "learning_rate": 1.6357586036625948e-05, + "loss": 0.9825, + "step": 17076 + }, + { + "epoch": 1.1, + "learning_rate": 1.6357371347603e-05, + "loss": 1.0167, + "step": 17077 + }, + { + "epoch": 1.1, + "learning_rate": 1.635715665858005e-05, + "loss": 0.9529, + "step": 17078 + }, + { + "epoch": 1.1, + "learning_rate": 1.6356941969557098e-05, + "loss": 1.0561, + "step": 17079 + }, + { + "epoch": 1.1, + "learning_rate": 1.6356727280534147e-05, + "loss": 0.9545, + "step": 17080 + }, + { + "epoch": 1.1, + "learning_rate": 1.63565125915112e-05, + "loss": 0.9863, + "step": 17081 + }, + { + "epoch": 1.1, + "learning_rate": 1.6356297902488248e-05, + "loss": 0.9543, + "step": 17082 + }, + { + "epoch": 1.1, + "learning_rate": 1.6356083213465297e-05, + "loss": 1.0241, + "step": 17083 + }, + { + "epoch": 1.1, + "learning_rate": 1.6355868524442346e-05, + "loss": 1.053, + "step": 17084 + }, + { + "epoch": 1.1, + "learning_rate": 1.6355653835419395e-05, + "loss": 0.998, + "step": 17085 + }, + { + "epoch": 1.1, + "learning_rate": 1.6355439146396444e-05, + "loss": 0.9013, + "step": 17086 + }, + { + "epoch": 1.1, + "learning_rate": 1.6355224457373496e-05, + "loss": 0.9573, + "step": 17087 + }, + { + "epoch": 1.1, + "learning_rate": 1.6355009768350545e-05, + "loss": 1.0906, + "step": 17088 + }, + { + "epoch": 1.1, + "learning_rate": 1.6354795079327594e-05, + "loss": 0.9557, + "step": 17089 + }, + { + "epoch": 1.1, + "learning_rate": 1.6354580390304643e-05, + "loss": 0.9572, + "step": 17090 + }, + { + "epoch": 1.1, + "learning_rate": 1.6354365701281696e-05, + "loss": 1.0066, + "step": 17091 + }, + { + "epoch": 1.1, + "learning_rate": 1.6354151012258745e-05, + "loss": 0.9421, + "step": 17092 + }, + { + "epoch": 1.1, + "learning_rate": 1.6353936323235794e-05, + "loss": 1.0043, + "step": 17093 + }, + { + "epoch": 1.1, + "learning_rate": 1.6353721634212846e-05, + "loss": 1.0043, + "step": 17094 + }, + { + "epoch": 1.1, + "learning_rate": 1.6353506945189895e-05, + "loss": 0.9486, + "step": 17095 + }, + { + "epoch": 1.1, + "learning_rate": 1.6353292256166944e-05, + "loss": 0.9085, + "step": 17096 + }, + { + "epoch": 1.1, + "learning_rate": 1.6353077567143993e-05, + "loss": 0.9118, + "step": 17097 + }, + { + "epoch": 1.1, + "learning_rate": 1.6352862878121042e-05, + "loss": 1.007, + "step": 17098 + }, + { + "epoch": 1.1, + "learning_rate": 1.6352648189098094e-05, + "loss": 0.9386, + "step": 17099 + }, + { + "epoch": 1.1, + "learning_rate": 1.6352433500075143e-05, + "loss": 0.9951, + "step": 17100 + }, + { + "epoch": 1.1, + "learning_rate": 1.6352218811052192e-05, + "loss": 0.9425, + "step": 17101 + }, + { + "epoch": 1.1, + "learning_rate": 1.635200412202924e-05, + "loss": 1.0248, + "step": 17102 + }, + { + "epoch": 1.1, + "learning_rate": 1.635178943300629e-05, + "loss": 0.9515, + "step": 17103 + }, + { + "epoch": 1.1, + "learning_rate": 1.635157474398334e-05, + "loss": 1.0055, + "step": 17104 + }, + { + "epoch": 1.1, + "learning_rate": 1.635136005496039e-05, + "loss": 1.0182, + "step": 17105 + }, + { + "epoch": 1.1, + "learning_rate": 1.635114536593744e-05, + "loss": 1.038, + "step": 17106 + }, + { + "epoch": 1.1, + "learning_rate": 1.6350930676914493e-05, + "loss": 1.0002, + "step": 17107 + }, + { + "epoch": 1.1, + "learning_rate": 1.635071598789154e-05, + "loss": 0.9699, + "step": 17108 + }, + { + "epoch": 1.1, + "learning_rate": 1.635050129886859e-05, + "loss": 0.953, + "step": 17109 + }, + { + "epoch": 1.1, + "learning_rate": 1.635028660984564e-05, + "loss": 1.0285, + "step": 17110 + }, + { + "epoch": 1.1, + "learning_rate": 1.635007192082269e-05, + "loss": 0.9728, + "step": 17111 + }, + { + "epoch": 1.1, + "learning_rate": 1.634985723179974e-05, + "loss": 0.9657, + "step": 17112 + }, + { + "epoch": 1.1, + "learning_rate": 1.634964254277679e-05, + "loss": 0.9537, + "step": 17113 + }, + { + "epoch": 1.1, + "learning_rate": 1.634942785375384e-05, + "loss": 1.0519, + "step": 17114 + }, + { + "epoch": 1.1, + "learning_rate": 1.6349213164730888e-05, + "loss": 1.1006, + "step": 17115 + }, + { + "epoch": 1.1, + "learning_rate": 1.6348998475707937e-05, + "loss": 1.0004, + "step": 17116 + }, + { + "epoch": 1.1, + "learning_rate": 1.634878378668499e-05, + "loss": 0.8967, + "step": 17117 + }, + { + "epoch": 1.1, + "learning_rate": 1.6348569097662038e-05, + "loss": 0.8848, + "step": 17118 + }, + { + "epoch": 1.1, + "learning_rate": 1.6348354408639087e-05, + "loss": 0.9568, + "step": 17119 + }, + { + "epoch": 1.1, + "learning_rate": 1.634813971961614e-05, + "loss": 1.017, + "step": 17120 + }, + { + "epoch": 1.1, + "learning_rate": 1.634792503059319e-05, + "loss": 0.9439, + "step": 17121 + }, + { + "epoch": 1.1, + "learning_rate": 1.6347710341570237e-05, + "loss": 0.8754, + "step": 17122 + }, + { + "epoch": 1.1, + "learning_rate": 1.6347495652547286e-05, + "loss": 0.9464, + "step": 17123 + }, + { + "epoch": 1.1, + "learning_rate": 1.6347280963524335e-05, + "loss": 0.9272, + "step": 17124 + }, + { + "epoch": 1.1, + "learning_rate": 1.6347066274501388e-05, + "loss": 0.9357, + "step": 17125 + }, + { + "epoch": 1.1, + "learning_rate": 1.6346851585478437e-05, + "loss": 0.9596, + "step": 17126 + }, + { + "epoch": 1.1, + "learning_rate": 1.6346636896455486e-05, + "loss": 0.9504, + "step": 17127 + }, + { + "epoch": 1.1, + "learning_rate": 1.6346422207432534e-05, + "loss": 1.0027, + "step": 17128 + }, + { + "epoch": 1.1, + "learning_rate": 1.6346207518409583e-05, + "loss": 0.9773, + "step": 17129 + }, + { + "epoch": 1.1, + "learning_rate": 1.6345992829386636e-05, + "loss": 0.9335, + "step": 17130 + }, + { + "epoch": 1.1, + "learning_rate": 1.6345778140363685e-05, + "loss": 0.9663, + "step": 17131 + }, + { + "epoch": 1.1, + "learning_rate": 1.6345563451340734e-05, + "loss": 0.8956, + "step": 17132 + }, + { + "epoch": 1.1, + "learning_rate": 1.6345348762317783e-05, + "loss": 1.0278, + "step": 17133 + }, + { + "epoch": 1.1, + "learning_rate": 1.634513407329483e-05, + "loss": 0.9895, + "step": 17134 + }, + { + "epoch": 1.1, + "learning_rate": 1.6344919384271884e-05, + "loss": 1.0073, + "step": 17135 + }, + { + "epoch": 1.1, + "learning_rate": 1.6344704695248933e-05, + "loss": 0.9786, + "step": 17136 + }, + { + "epoch": 1.1, + "learning_rate": 1.6344490006225985e-05, + "loss": 1.0017, + "step": 17137 + }, + { + "epoch": 1.1, + "learning_rate": 1.6344275317203034e-05, + "loss": 0.9986, + "step": 17138 + }, + { + "epoch": 1.1, + "learning_rate": 1.6344060628180083e-05, + "loss": 1.0137, + "step": 17139 + }, + { + "epoch": 1.1, + "learning_rate": 1.6343845939157132e-05, + "loss": 0.9697, + "step": 17140 + }, + { + "epoch": 1.1, + "learning_rate": 1.634363125013418e-05, + "loss": 1.0086, + "step": 17141 + }, + { + "epoch": 1.1, + "learning_rate": 1.634341656111123e-05, + "loss": 0.9601, + "step": 17142 + }, + { + "epoch": 1.1, + "learning_rate": 1.6343201872088283e-05, + "loss": 0.9345, + "step": 17143 + }, + { + "epoch": 1.1, + "learning_rate": 1.634298718306533e-05, + "loss": 0.9502, + "step": 17144 + }, + { + "epoch": 1.1, + "learning_rate": 1.634277249404238e-05, + "loss": 0.998, + "step": 17145 + }, + { + "epoch": 1.1, + "learning_rate": 1.634255780501943e-05, + "loss": 0.9573, + "step": 17146 + }, + { + "epoch": 1.1, + "learning_rate": 1.634234311599648e-05, + "loss": 0.9391, + "step": 17147 + }, + { + "epoch": 1.1, + "learning_rate": 1.634212842697353e-05, + "loss": 0.9843, + "step": 17148 + }, + { + "epoch": 1.1, + "learning_rate": 1.634191373795058e-05, + "loss": 0.9939, + "step": 17149 + }, + { + "epoch": 1.1, + "learning_rate": 1.6341699048927632e-05, + "loss": 0.9744, + "step": 17150 + }, + { + "epoch": 1.1, + "learning_rate": 1.634148435990468e-05, + "loss": 0.8975, + "step": 17151 + }, + { + "epoch": 1.1, + "learning_rate": 1.634126967088173e-05, + "loss": 0.9043, + "step": 17152 + }, + { + "epoch": 1.1, + "learning_rate": 1.634105498185878e-05, + "loss": 1.0225, + "step": 17153 + }, + { + "epoch": 1.1, + "learning_rate": 1.6340840292835828e-05, + "loss": 0.9549, + "step": 17154 + }, + { + "epoch": 1.1, + "learning_rate": 1.6340625603812877e-05, + "loss": 0.9888, + "step": 17155 + }, + { + "epoch": 1.1, + "learning_rate": 1.634041091478993e-05, + "loss": 1.0406, + "step": 17156 + }, + { + "epoch": 1.1, + "learning_rate": 1.6340196225766978e-05, + "loss": 0.9693, + "step": 17157 + }, + { + "epoch": 1.1, + "learning_rate": 1.6339981536744027e-05, + "loss": 1.0538, + "step": 17158 + }, + { + "epoch": 1.1, + "learning_rate": 1.6339766847721076e-05, + "loss": 0.9856, + "step": 17159 + }, + { + "epoch": 1.1, + "learning_rate": 1.6339552158698125e-05, + "loss": 0.9267, + "step": 17160 + }, + { + "epoch": 1.1, + "learning_rate": 1.6339337469675177e-05, + "loss": 1.026, + "step": 17161 + }, + { + "epoch": 1.1, + "learning_rate": 1.6339122780652226e-05, + "loss": 0.9125, + "step": 17162 + }, + { + "epoch": 1.1, + "learning_rate": 1.6338908091629275e-05, + "loss": 0.9511, + "step": 17163 + }, + { + "epoch": 1.1, + "learning_rate": 1.6338693402606328e-05, + "loss": 0.9576, + "step": 17164 + }, + { + "epoch": 1.1, + "learning_rate": 1.6338478713583377e-05, + "loss": 0.9751, + "step": 17165 + }, + { + "epoch": 1.1, + "learning_rate": 1.6338264024560426e-05, + "loss": 0.9543, + "step": 17166 + }, + { + "epoch": 1.1, + "learning_rate": 1.6338049335537475e-05, + "loss": 0.9766, + "step": 17167 + }, + { + "epoch": 1.1, + "learning_rate": 1.6337834646514527e-05, + "loss": 1.0139, + "step": 17168 + }, + { + "epoch": 1.1, + "learning_rate": 1.6337619957491576e-05, + "loss": 0.9598, + "step": 17169 + }, + { + "epoch": 1.1, + "learning_rate": 1.6337405268468625e-05, + "loss": 0.9383, + "step": 17170 + }, + { + "epoch": 1.1, + "learning_rate": 1.6337190579445674e-05, + "loss": 1.0569, + "step": 17171 + }, + { + "epoch": 1.1, + "learning_rate": 1.6336975890422723e-05, + "loss": 0.9871, + "step": 17172 + }, + { + "epoch": 1.1, + "learning_rate": 1.6336761201399772e-05, + "loss": 1.0339, + "step": 17173 + }, + { + "epoch": 1.1, + "learning_rate": 1.6336546512376824e-05, + "loss": 0.9749, + "step": 17174 + }, + { + "epoch": 1.1, + "learning_rate": 1.6336331823353873e-05, + "loss": 1.0497, + "step": 17175 + }, + { + "epoch": 1.11, + "learning_rate": 1.6336117134330922e-05, + "loss": 0.9615, + "step": 17176 + }, + { + "epoch": 1.11, + "learning_rate": 1.633590244530797e-05, + "loss": 0.9757, + "step": 17177 + }, + { + "epoch": 1.11, + "learning_rate": 1.6335687756285023e-05, + "loss": 1.0392, + "step": 17178 + }, + { + "epoch": 1.11, + "learning_rate": 1.6335473067262072e-05, + "loss": 0.9208, + "step": 17179 + }, + { + "epoch": 1.11, + "learning_rate": 1.633525837823912e-05, + "loss": 1.0421, + "step": 17180 + }, + { + "epoch": 1.11, + "learning_rate": 1.6335043689216174e-05, + "loss": 0.9754, + "step": 17181 + }, + { + "epoch": 1.11, + "learning_rate": 1.6334829000193223e-05, + "loss": 0.9847, + "step": 17182 + }, + { + "epoch": 1.11, + "learning_rate": 1.633461431117027e-05, + "loss": 0.876, + "step": 17183 + }, + { + "epoch": 1.11, + "learning_rate": 1.633439962214732e-05, + "loss": 0.9822, + "step": 17184 + }, + { + "epoch": 1.11, + "learning_rate": 1.633418493312437e-05, + "loss": 0.9585, + "step": 17185 + }, + { + "epoch": 1.11, + "learning_rate": 1.6333970244101422e-05, + "loss": 1.0385, + "step": 17186 + }, + { + "epoch": 1.11, + "learning_rate": 1.633375555507847e-05, + "loss": 0.9976, + "step": 17187 + }, + { + "epoch": 1.11, + "learning_rate": 1.633354086605552e-05, + "loss": 0.9359, + "step": 17188 + }, + { + "epoch": 1.11, + "learning_rate": 1.633332617703257e-05, + "loss": 0.9987, + "step": 17189 + }, + { + "epoch": 1.11, + "learning_rate": 1.6333111488009618e-05, + "loss": 0.9526, + "step": 17190 + }, + { + "epoch": 1.11, + "learning_rate": 1.6332896798986667e-05, + "loss": 0.9938, + "step": 17191 + }, + { + "epoch": 1.11, + "learning_rate": 1.633268210996372e-05, + "loss": 0.9901, + "step": 17192 + }, + { + "epoch": 1.11, + "learning_rate": 1.6332467420940768e-05, + "loss": 0.957, + "step": 17193 + }, + { + "epoch": 1.11, + "learning_rate": 1.633225273191782e-05, + "loss": 1.0506, + "step": 17194 + }, + { + "epoch": 1.11, + "learning_rate": 1.633203804289487e-05, + "loss": 1.0719, + "step": 17195 + }, + { + "epoch": 1.11, + "learning_rate": 1.633182335387192e-05, + "loss": 1.0152, + "step": 17196 + }, + { + "epoch": 1.11, + "learning_rate": 1.6331608664848967e-05, + "loss": 1.0015, + "step": 17197 + }, + { + "epoch": 1.11, + "learning_rate": 1.6331393975826016e-05, + "loss": 0.9527, + "step": 17198 + }, + { + "epoch": 1.11, + "learning_rate": 1.633117928680307e-05, + "loss": 0.9267, + "step": 17199 + }, + { + "epoch": 1.11, + "learning_rate": 1.6330964597780118e-05, + "loss": 0.9806, + "step": 17200 + }, + { + "epoch": 1.11, + "learning_rate": 1.6330749908757167e-05, + "loss": 1.0311, + "step": 17201 + }, + { + "epoch": 1.11, + "learning_rate": 1.6330535219734215e-05, + "loss": 0.95, + "step": 17202 + }, + { + "epoch": 1.11, + "learning_rate": 1.6330320530711264e-05, + "loss": 0.975, + "step": 17203 + }, + { + "epoch": 1.11, + "learning_rate": 1.6330105841688317e-05, + "loss": 0.9394, + "step": 17204 + }, + { + "epoch": 1.11, + "learning_rate": 1.6329891152665366e-05, + "loss": 0.9758, + "step": 17205 + }, + { + "epoch": 1.11, + "learning_rate": 1.6329676463642415e-05, + "loss": 0.9058, + "step": 17206 + }, + { + "epoch": 1.11, + "learning_rate": 1.6329461774619467e-05, + "loss": 0.9379, + "step": 17207 + }, + { + "epoch": 1.11, + "learning_rate": 1.6329247085596516e-05, + "loss": 0.9363, + "step": 17208 + }, + { + "epoch": 1.11, + "learning_rate": 1.6329032396573565e-05, + "loss": 1.0396, + "step": 17209 + }, + { + "epoch": 1.11, + "learning_rate": 1.6328817707550614e-05, + "loss": 0.9112, + "step": 17210 + }, + { + "epoch": 1.11, + "learning_rate": 1.6328603018527663e-05, + "loss": 0.934, + "step": 17211 + }, + { + "epoch": 1.11, + "learning_rate": 1.6328388329504715e-05, + "loss": 1.0244, + "step": 17212 + }, + { + "epoch": 1.11, + "learning_rate": 1.6328173640481764e-05, + "loss": 0.995, + "step": 17213 + }, + { + "epoch": 1.11, + "learning_rate": 1.6327958951458813e-05, + "loss": 0.9678, + "step": 17214 + }, + { + "epoch": 1.11, + "learning_rate": 1.6327744262435862e-05, + "loss": 0.9251, + "step": 17215 + }, + { + "epoch": 1.11, + "learning_rate": 1.632752957341291e-05, + "loss": 0.9726, + "step": 17216 + }, + { + "epoch": 1.11, + "learning_rate": 1.6327314884389964e-05, + "loss": 0.9394, + "step": 17217 + }, + { + "epoch": 1.11, + "learning_rate": 1.6327100195367012e-05, + "loss": 1.0594, + "step": 17218 + }, + { + "epoch": 1.11, + "learning_rate": 1.632688550634406e-05, + "loss": 1.0192, + "step": 17219 + }, + { + "epoch": 1.11, + "learning_rate": 1.632667081732111e-05, + "loss": 0.9384, + "step": 17220 + }, + { + "epoch": 1.11, + "learning_rate": 1.632645612829816e-05, + "loss": 0.9403, + "step": 17221 + }, + { + "epoch": 1.11, + "learning_rate": 1.6326241439275212e-05, + "loss": 0.9385, + "step": 17222 + }, + { + "epoch": 1.11, + "learning_rate": 1.632602675025226e-05, + "loss": 1.012, + "step": 17223 + }, + { + "epoch": 1.11, + "learning_rate": 1.6325812061229313e-05, + "loss": 1.0033, + "step": 17224 + }, + { + "epoch": 1.11, + "learning_rate": 1.6325597372206362e-05, + "loss": 0.9877, + "step": 17225 + }, + { + "epoch": 1.11, + "learning_rate": 1.632538268318341e-05, + "loss": 0.9999, + "step": 17226 + }, + { + "epoch": 1.11, + "learning_rate": 1.632516799416046e-05, + "loss": 0.9746, + "step": 17227 + }, + { + "epoch": 1.11, + "learning_rate": 1.632495330513751e-05, + "loss": 0.9634, + "step": 17228 + }, + { + "epoch": 1.11, + "learning_rate": 1.6324738616114558e-05, + "loss": 0.9912, + "step": 17229 + }, + { + "epoch": 1.11, + "learning_rate": 1.632452392709161e-05, + "loss": 1.0074, + "step": 17230 + }, + { + "epoch": 1.11, + "learning_rate": 1.632430923806866e-05, + "loss": 0.9594, + "step": 17231 + }, + { + "epoch": 1.11, + "learning_rate": 1.6324094549045708e-05, + "loss": 0.957, + "step": 17232 + }, + { + "epoch": 1.11, + "learning_rate": 1.6323879860022757e-05, + "loss": 1.0212, + "step": 17233 + }, + { + "epoch": 1.11, + "learning_rate": 1.6323665170999806e-05, + "loss": 1.011, + "step": 17234 + }, + { + "epoch": 1.11, + "learning_rate": 1.632345048197686e-05, + "loss": 1.0311, + "step": 17235 + }, + { + "epoch": 1.11, + "learning_rate": 1.6323235792953907e-05, + "loss": 0.9064, + "step": 17236 + }, + { + "epoch": 1.11, + "learning_rate": 1.632302110393096e-05, + "loss": 0.9369, + "step": 17237 + }, + { + "epoch": 1.11, + "learning_rate": 1.632280641490801e-05, + "loss": 0.9729, + "step": 17238 + }, + { + "epoch": 1.11, + "learning_rate": 1.6322591725885058e-05, + "loss": 1.0244, + "step": 17239 + }, + { + "epoch": 1.11, + "learning_rate": 1.6322377036862107e-05, + "loss": 1.0041, + "step": 17240 + }, + { + "epoch": 1.11, + "learning_rate": 1.6322162347839156e-05, + "loss": 0.9326, + "step": 17241 + }, + { + "epoch": 1.11, + "learning_rate": 1.6321947658816205e-05, + "loss": 0.9918, + "step": 17242 + }, + { + "epoch": 1.11, + "learning_rate": 1.6321732969793257e-05, + "loss": 1.0452, + "step": 17243 + }, + { + "epoch": 1.11, + "learning_rate": 1.6321518280770306e-05, + "loss": 1.0, + "step": 17244 + }, + { + "epoch": 1.11, + "learning_rate": 1.6321303591747355e-05, + "loss": 0.9676, + "step": 17245 + }, + { + "epoch": 1.11, + "learning_rate": 1.6321088902724404e-05, + "loss": 1.0133, + "step": 17246 + }, + { + "epoch": 1.11, + "learning_rate": 1.6320874213701453e-05, + "loss": 0.9508, + "step": 17247 + }, + { + "epoch": 1.11, + "learning_rate": 1.6320659524678505e-05, + "loss": 1.0096, + "step": 17248 + }, + { + "epoch": 1.11, + "learning_rate": 1.6320444835655554e-05, + "loss": 0.9885, + "step": 17249 + }, + { + "epoch": 1.11, + "learning_rate": 1.6320230146632603e-05, + "loss": 0.9164, + "step": 17250 + }, + { + "epoch": 1.11, + "learning_rate": 1.6320015457609655e-05, + "loss": 0.9839, + "step": 17251 + }, + { + "epoch": 1.11, + "learning_rate": 1.6319800768586704e-05, + "loss": 1.0334, + "step": 17252 + }, + { + "epoch": 1.11, + "learning_rate": 1.6319586079563753e-05, + "loss": 0.9683, + "step": 17253 + }, + { + "epoch": 1.11, + "learning_rate": 1.6319371390540802e-05, + "loss": 0.9877, + "step": 17254 + }, + { + "epoch": 1.11, + "learning_rate": 1.6319156701517855e-05, + "loss": 0.9114, + "step": 17255 + }, + { + "epoch": 1.11, + "learning_rate": 1.6318942012494904e-05, + "loss": 0.941, + "step": 17256 + }, + { + "epoch": 1.11, + "learning_rate": 1.6318727323471953e-05, + "loss": 0.9651, + "step": 17257 + }, + { + "epoch": 1.11, + "learning_rate": 1.6318512634449e-05, + "loss": 0.9296, + "step": 17258 + }, + { + "epoch": 1.11, + "learning_rate": 1.631829794542605e-05, + "loss": 1.0196, + "step": 17259 + }, + { + "epoch": 1.11, + "learning_rate": 1.63180832564031e-05, + "loss": 0.9905, + "step": 17260 + }, + { + "epoch": 1.11, + "learning_rate": 1.6317868567380152e-05, + "loss": 0.9487, + "step": 17261 + }, + { + "epoch": 1.11, + "learning_rate": 1.63176538783572e-05, + "loss": 0.9344, + "step": 17262 + }, + { + "epoch": 1.11, + "learning_rate": 1.631743918933425e-05, + "loss": 1.0277, + "step": 17263 + }, + { + "epoch": 1.11, + "learning_rate": 1.63172245003113e-05, + "loss": 0.9713, + "step": 17264 + }, + { + "epoch": 1.11, + "learning_rate": 1.631700981128835e-05, + "loss": 1.0205, + "step": 17265 + }, + { + "epoch": 1.11, + "learning_rate": 1.63167951222654e-05, + "loss": 0.9451, + "step": 17266 + }, + { + "epoch": 1.11, + "learning_rate": 1.631658043324245e-05, + "loss": 0.9989, + "step": 17267 + }, + { + "epoch": 1.11, + "learning_rate": 1.63163657442195e-05, + "loss": 0.9678, + "step": 17268 + }, + { + "epoch": 1.11, + "learning_rate": 1.631615105519655e-05, + "loss": 1.001, + "step": 17269 + }, + { + "epoch": 1.11, + "learning_rate": 1.63159363661736e-05, + "loss": 0.9747, + "step": 17270 + }, + { + "epoch": 1.11, + "learning_rate": 1.6315721677150648e-05, + "loss": 1.0177, + "step": 17271 + }, + { + "epoch": 1.11, + "learning_rate": 1.6315506988127697e-05, + "loss": 1.0099, + "step": 17272 + }, + { + "epoch": 1.11, + "learning_rate": 1.631529229910475e-05, + "loss": 0.9089, + "step": 17273 + }, + { + "epoch": 1.11, + "learning_rate": 1.63150776100818e-05, + "loss": 1.0314, + "step": 17274 + }, + { + "epoch": 1.11, + "learning_rate": 1.6314862921058848e-05, + "loss": 0.9759, + "step": 17275 + }, + { + "epoch": 1.11, + "learning_rate": 1.6314648232035896e-05, + "loss": 1.0104, + "step": 17276 + }, + { + "epoch": 1.11, + "learning_rate": 1.6314433543012945e-05, + "loss": 1.0481, + "step": 17277 + }, + { + "epoch": 1.11, + "learning_rate": 1.6314218853989994e-05, + "loss": 1.0109, + "step": 17278 + }, + { + "epoch": 1.11, + "learning_rate": 1.6314004164967047e-05, + "loss": 0.9627, + "step": 17279 + }, + { + "epoch": 1.11, + "learning_rate": 1.6313789475944096e-05, + "loss": 0.9455, + "step": 17280 + }, + { + "epoch": 1.11, + "learning_rate": 1.6313574786921148e-05, + "loss": 0.9732, + "step": 17281 + }, + { + "epoch": 1.11, + "learning_rate": 1.6313360097898197e-05, + "loss": 0.9693, + "step": 17282 + }, + { + "epoch": 1.11, + "learning_rate": 1.6313145408875246e-05, + "loss": 0.9508, + "step": 17283 + }, + { + "epoch": 1.11, + "learning_rate": 1.6312930719852295e-05, + "loss": 1.0143, + "step": 17284 + }, + { + "epoch": 1.11, + "learning_rate": 1.6312716030829344e-05, + "loss": 0.8464, + "step": 17285 + }, + { + "epoch": 1.11, + "learning_rate": 1.6312501341806396e-05, + "loss": 1.0247, + "step": 17286 + }, + { + "epoch": 1.11, + "learning_rate": 1.6312286652783445e-05, + "loss": 1.004, + "step": 17287 + }, + { + "epoch": 1.11, + "learning_rate": 1.6312071963760494e-05, + "loss": 0.9101, + "step": 17288 + }, + { + "epoch": 1.11, + "learning_rate": 1.6311857274737543e-05, + "loss": 1.0161, + "step": 17289 + }, + { + "epoch": 1.11, + "learning_rate": 1.6311642585714592e-05, + "loss": 1.0131, + "step": 17290 + }, + { + "epoch": 1.11, + "learning_rate": 1.6311427896691645e-05, + "loss": 0.958, + "step": 17291 + }, + { + "epoch": 1.11, + "learning_rate": 1.6311213207668693e-05, + "loss": 0.9768, + "step": 17292 + }, + { + "epoch": 1.11, + "learning_rate": 1.6310998518645742e-05, + "loss": 1.0002, + "step": 17293 + }, + { + "epoch": 1.11, + "learning_rate": 1.6310783829622795e-05, + "loss": 0.9031, + "step": 17294 + }, + { + "epoch": 1.11, + "learning_rate": 1.6310569140599844e-05, + "loss": 0.9563, + "step": 17295 + }, + { + "epoch": 1.11, + "learning_rate": 1.6310354451576893e-05, + "loss": 0.9663, + "step": 17296 + }, + { + "epoch": 1.11, + "learning_rate": 1.6310139762553942e-05, + "loss": 0.983, + "step": 17297 + }, + { + "epoch": 1.11, + "learning_rate": 1.630992507353099e-05, + "loss": 0.9649, + "step": 17298 + }, + { + "epoch": 1.11, + "learning_rate": 1.6309710384508043e-05, + "loss": 0.9835, + "step": 17299 + }, + { + "epoch": 1.11, + "learning_rate": 1.6309495695485092e-05, + "loss": 1.0171, + "step": 17300 + }, + { + "epoch": 1.11, + "learning_rate": 1.630928100646214e-05, + "loss": 0.9326, + "step": 17301 + }, + { + "epoch": 1.11, + "learning_rate": 1.630906631743919e-05, + "loss": 0.9487, + "step": 17302 + }, + { + "epoch": 1.11, + "learning_rate": 1.630885162841624e-05, + "loss": 0.9384, + "step": 17303 + }, + { + "epoch": 1.11, + "learning_rate": 1.630863693939329e-05, + "loss": 0.9576, + "step": 17304 + }, + { + "epoch": 1.11, + "learning_rate": 1.630842225037034e-05, + "loss": 1.0443, + "step": 17305 + }, + { + "epoch": 1.11, + "learning_rate": 1.630820756134739e-05, + "loss": 1.0046, + "step": 17306 + }, + { + "epoch": 1.11, + "learning_rate": 1.6307992872324438e-05, + "loss": 1.0078, + "step": 17307 + }, + { + "epoch": 1.11, + "learning_rate": 1.6307778183301487e-05, + "loss": 1.0108, + "step": 17308 + }, + { + "epoch": 1.11, + "learning_rate": 1.630756349427854e-05, + "loss": 0.9849, + "step": 17309 + }, + { + "epoch": 1.11, + "learning_rate": 1.630734880525559e-05, + "loss": 0.987, + "step": 17310 + }, + { + "epoch": 1.11, + "learning_rate": 1.630713411623264e-05, + "loss": 1.0311, + "step": 17311 + }, + { + "epoch": 1.11, + "learning_rate": 1.630691942720969e-05, + "loss": 0.9198, + "step": 17312 + }, + { + "epoch": 1.11, + "learning_rate": 1.630670473818674e-05, + "loss": 1.0343, + "step": 17313 + }, + { + "epoch": 1.11, + "learning_rate": 1.6306490049163788e-05, + "loss": 0.9988, + "step": 17314 + }, + { + "epoch": 1.11, + "learning_rate": 1.6306275360140837e-05, + "loss": 0.9625, + "step": 17315 + }, + { + "epoch": 1.11, + "learning_rate": 1.6306060671117886e-05, + "loss": 0.962, + "step": 17316 + }, + { + "epoch": 1.11, + "learning_rate": 1.6305845982094938e-05, + "loss": 0.9238, + "step": 17317 + }, + { + "epoch": 1.11, + "learning_rate": 1.6305631293071987e-05, + "loss": 1.0168, + "step": 17318 + }, + { + "epoch": 1.11, + "learning_rate": 1.6305416604049036e-05, + "loss": 0.9549, + "step": 17319 + }, + { + "epoch": 1.11, + "learning_rate": 1.6305201915026085e-05, + "loss": 0.9819, + "step": 17320 + }, + { + "epoch": 1.11, + "learning_rate": 1.6304987226003134e-05, + "loss": 0.9926, + "step": 17321 + }, + { + "epoch": 1.11, + "learning_rate": 1.6304772536980186e-05, + "loss": 0.9427, + "step": 17322 + }, + { + "epoch": 1.11, + "learning_rate": 1.6304557847957235e-05, + "loss": 0.947, + "step": 17323 + }, + { + "epoch": 1.11, + "learning_rate": 1.6304343158934287e-05, + "loss": 1.0688, + "step": 17324 + }, + { + "epoch": 1.11, + "learning_rate": 1.6304128469911336e-05, + "loss": 0.9692, + "step": 17325 + }, + { + "epoch": 1.11, + "learning_rate": 1.6303913780888385e-05, + "loss": 1.0597, + "step": 17326 + }, + { + "epoch": 1.11, + "learning_rate": 1.6303699091865434e-05, + "loss": 1.012, + "step": 17327 + }, + { + "epoch": 1.11, + "learning_rate": 1.6303484402842483e-05, + "loss": 0.9486, + "step": 17328 + }, + { + "epoch": 1.11, + "learning_rate": 1.6303269713819532e-05, + "loss": 1.0137, + "step": 17329 + }, + { + "epoch": 1.11, + "learning_rate": 1.6303055024796585e-05, + "loss": 1.0152, + "step": 17330 + }, + { + "epoch": 1.12, + "learning_rate": 1.6302840335773634e-05, + "loss": 0.9285, + "step": 17331 + }, + { + "epoch": 1.12, + "learning_rate": 1.6302625646750683e-05, + "loss": 1.0903, + "step": 17332 + }, + { + "epoch": 1.12, + "learning_rate": 1.630241095772773e-05, + "loss": 0.9987, + "step": 17333 + }, + { + "epoch": 1.12, + "learning_rate": 1.630219626870478e-05, + "loss": 1.0131, + "step": 17334 + }, + { + "epoch": 1.12, + "learning_rate": 1.6301981579681833e-05, + "loss": 0.9243, + "step": 17335 + }, + { + "epoch": 1.12, + "learning_rate": 1.6301766890658882e-05, + "loss": 0.9879, + "step": 17336 + }, + { + "epoch": 1.12, + "learning_rate": 1.630155220163593e-05, + "loss": 0.9837, + "step": 17337 + }, + { + "epoch": 1.12, + "learning_rate": 1.6301337512612983e-05, + "loss": 1.0095, + "step": 17338 + }, + { + "epoch": 1.12, + "learning_rate": 1.6301122823590032e-05, + "loss": 0.977, + "step": 17339 + }, + { + "epoch": 1.12, + "learning_rate": 1.630090813456708e-05, + "loss": 0.8835, + "step": 17340 + }, + { + "epoch": 1.12, + "learning_rate": 1.630069344554413e-05, + "loss": 1.0114, + "step": 17341 + }, + { + "epoch": 1.12, + "learning_rate": 1.6300478756521182e-05, + "loss": 0.9142, + "step": 17342 + }, + { + "epoch": 1.12, + "learning_rate": 1.630026406749823e-05, + "loss": 0.9312, + "step": 17343 + }, + { + "epoch": 1.12, + "learning_rate": 1.630004937847528e-05, + "loss": 0.9836, + "step": 17344 + }, + { + "epoch": 1.12, + "learning_rate": 1.629983468945233e-05, + "loss": 1.0186, + "step": 17345 + }, + { + "epoch": 1.12, + "learning_rate": 1.6299620000429378e-05, + "loss": 0.8993, + "step": 17346 + }, + { + "epoch": 1.12, + "learning_rate": 1.6299405311406427e-05, + "loss": 0.9502, + "step": 17347 + }, + { + "epoch": 1.12, + "learning_rate": 1.629919062238348e-05, + "loss": 1.0017, + "step": 17348 + }, + { + "epoch": 1.12, + "learning_rate": 1.629897593336053e-05, + "loss": 0.9545, + "step": 17349 + }, + { + "epoch": 1.12, + "learning_rate": 1.6298761244337578e-05, + "loss": 0.9484, + "step": 17350 + }, + { + "epoch": 1.12, + "learning_rate": 1.6298546555314626e-05, + "loss": 0.9558, + "step": 17351 + }, + { + "epoch": 1.12, + "learning_rate": 1.6298331866291675e-05, + "loss": 0.9574, + "step": 17352 + }, + { + "epoch": 1.12, + "learning_rate": 1.6298117177268728e-05, + "loss": 0.9211, + "step": 17353 + }, + { + "epoch": 1.12, + "learning_rate": 1.6297902488245777e-05, + "loss": 0.9322, + "step": 17354 + }, + { + "epoch": 1.12, + "learning_rate": 1.629768779922283e-05, + "loss": 0.9894, + "step": 17355 + }, + { + "epoch": 1.12, + "learning_rate": 1.6297473110199878e-05, + "loss": 0.9521, + "step": 17356 + }, + { + "epoch": 1.12, + "learning_rate": 1.6297258421176927e-05, + "loss": 1.0275, + "step": 17357 + }, + { + "epoch": 1.12, + "learning_rate": 1.6297043732153976e-05, + "loss": 0.991, + "step": 17358 + }, + { + "epoch": 1.12, + "learning_rate": 1.6296829043131025e-05, + "loss": 0.9342, + "step": 17359 + }, + { + "epoch": 1.12, + "learning_rate": 1.6296614354108077e-05, + "loss": 1.011, + "step": 17360 + }, + { + "epoch": 1.12, + "learning_rate": 1.6296399665085126e-05, + "loss": 1.0094, + "step": 17361 + }, + { + "epoch": 1.12, + "learning_rate": 1.6296184976062175e-05, + "loss": 1.0248, + "step": 17362 + }, + { + "epoch": 1.12, + "learning_rate": 1.6295970287039224e-05, + "loss": 0.963, + "step": 17363 + }, + { + "epoch": 1.12, + "learning_rate": 1.6295755598016273e-05, + "loss": 0.9762, + "step": 17364 + }, + { + "epoch": 1.12, + "learning_rate": 1.6295540908993322e-05, + "loss": 1.0438, + "step": 17365 + }, + { + "epoch": 1.12, + "learning_rate": 1.6295326219970374e-05, + "loss": 0.9463, + "step": 17366 + }, + { + "epoch": 1.12, + "learning_rate": 1.6295111530947423e-05, + "loss": 0.9659, + "step": 17367 + }, + { + "epoch": 1.12, + "learning_rate": 1.6294896841924476e-05, + "loss": 0.9662, + "step": 17368 + }, + { + "epoch": 1.12, + "learning_rate": 1.6294682152901525e-05, + "loss": 1.0578, + "step": 17369 + }, + { + "epoch": 1.12, + "learning_rate": 1.6294467463878574e-05, + "loss": 1.0069, + "step": 17370 + }, + { + "epoch": 1.12, + "learning_rate": 1.6294252774855623e-05, + "loss": 0.9375, + "step": 17371 + }, + { + "epoch": 1.12, + "learning_rate": 1.629403808583267e-05, + "loss": 0.9915, + "step": 17372 + }, + { + "epoch": 1.12, + "learning_rate": 1.6293823396809724e-05, + "loss": 0.8872, + "step": 17373 + }, + { + "epoch": 1.12, + "learning_rate": 1.6293608707786773e-05, + "loss": 0.9688, + "step": 17374 + }, + { + "epoch": 1.12, + "learning_rate": 1.6293394018763822e-05, + "loss": 0.9368, + "step": 17375 + }, + { + "epoch": 1.12, + "learning_rate": 1.629317932974087e-05, + "loss": 0.9599, + "step": 17376 + }, + { + "epoch": 1.12, + "learning_rate": 1.629296464071792e-05, + "loss": 0.8944, + "step": 17377 + }, + { + "epoch": 1.12, + "learning_rate": 1.629274995169497e-05, + "loss": 0.9669, + "step": 17378 + }, + { + "epoch": 1.12, + "learning_rate": 1.629253526267202e-05, + "loss": 0.9228, + "step": 17379 + }, + { + "epoch": 1.12, + "learning_rate": 1.629232057364907e-05, + "loss": 0.9653, + "step": 17380 + }, + { + "epoch": 1.12, + "learning_rate": 1.629210588462612e-05, + "loss": 0.9047, + "step": 17381 + }, + { + "epoch": 1.12, + "learning_rate": 1.629189119560317e-05, + "loss": 1.0092, + "step": 17382 + }, + { + "epoch": 1.12, + "learning_rate": 1.629167650658022e-05, + "loss": 0.9958, + "step": 17383 + }, + { + "epoch": 1.12, + "learning_rate": 1.629146181755727e-05, + "loss": 0.9778, + "step": 17384 + }, + { + "epoch": 1.12, + "learning_rate": 1.629124712853432e-05, + "loss": 0.911, + "step": 17385 + }, + { + "epoch": 1.12, + "learning_rate": 1.629103243951137e-05, + "loss": 0.9503, + "step": 17386 + }, + { + "epoch": 1.12, + "learning_rate": 1.629081775048842e-05, + "loss": 1.0286, + "step": 17387 + }, + { + "epoch": 1.12, + "learning_rate": 1.629060306146547e-05, + "loss": 0.9807, + "step": 17388 + }, + { + "epoch": 1.12, + "learning_rate": 1.6290388372442518e-05, + "loss": 1.0185, + "step": 17389 + }, + { + "epoch": 1.12, + "learning_rate": 1.6290173683419567e-05, + "loss": 1.032, + "step": 17390 + }, + { + "epoch": 1.12, + "learning_rate": 1.628995899439662e-05, + "loss": 0.9378, + "step": 17391 + }, + { + "epoch": 1.12, + "learning_rate": 1.6289744305373668e-05, + "loss": 0.9896, + "step": 17392 + }, + { + "epoch": 1.12, + "learning_rate": 1.6289529616350717e-05, + "loss": 1.0074, + "step": 17393 + }, + { + "epoch": 1.12, + "learning_rate": 1.6289314927327766e-05, + "loss": 1.0174, + "step": 17394 + }, + { + "epoch": 1.12, + "learning_rate": 1.6289100238304815e-05, + "loss": 1.0309, + "step": 17395 + }, + { + "epoch": 1.12, + "learning_rate": 1.6288885549281867e-05, + "loss": 0.8967, + "step": 17396 + }, + { + "epoch": 1.12, + "learning_rate": 1.6288670860258916e-05, + "loss": 0.9586, + "step": 17397 + }, + { + "epoch": 1.12, + "learning_rate": 1.628845617123597e-05, + "loss": 1.0364, + "step": 17398 + }, + { + "epoch": 1.12, + "learning_rate": 1.6288241482213017e-05, + "loss": 0.9964, + "step": 17399 + }, + { + "epoch": 1.12, + "learning_rate": 1.6288026793190066e-05, + "loss": 0.952, + "step": 17400 + }, + { + "epoch": 1.12, + "learning_rate": 1.6287812104167115e-05, + "loss": 0.9609, + "step": 17401 + }, + { + "epoch": 1.12, + "learning_rate": 1.6287597415144164e-05, + "loss": 0.9342, + "step": 17402 + }, + { + "epoch": 1.12, + "learning_rate": 1.6287382726121213e-05, + "loss": 1.031, + "step": 17403 + }, + { + "epoch": 1.12, + "learning_rate": 1.6287168037098266e-05, + "loss": 0.9778, + "step": 17404 + }, + { + "epoch": 1.12, + "learning_rate": 1.6286953348075315e-05, + "loss": 0.909, + "step": 17405 + }, + { + "epoch": 1.12, + "learning_rate": 1.6286738659052364e-05, + "loss": 0.9402, + "step": 17406 + }, + { + "epoch": 1.12, + "learning_rate": 1.6286523970029413e-05, + "loss": 0.9429, + "step": 17407 + }, + { + "epoch": 1.12, + "learning_rate": 1.628630928100646e-05, + "loss": 1.0155, + "step": 17408 + }, + { + "epoch": 1.12, + "learning_rate": 1.6286094591983514e-05, + "loss": 0.9443, + "step": 17409 + }, + { + "epoch": 1.12, + "learning_rate": 1.6285879902960563e-05, + "loss": 1.0353, + "step": 17410 + }, + { + "epoch": 1.12, + "learning_rate": 1.6285665213937615e-05, + "loss": 0.9169, + "step": 17411 + }, + { + "epoch": 1.12, + "learning_rate": 1.6285450524914664e-05, + "loss": 0.8959, + "step": 17412 + }, + { + "epoch": 1.12, + "learning_rate": 1.6285235835891713e-05, + "loss": 1.0063, + "step": 17413 + }, + { + "epoch": 1.12, + "learning_rate": 1.6285021146868762e-05, + "loss": 0.9795, + "step": 17414 + }, + { + "epoch": 1.12, + "learning_rate": 1.628480645784581e-05, + "loss": 0.9518, + "step": 17415 + }, + { + "epoch": 1.12, + "learning_rate": 1.628459176882286e-05, + "loss": 1.0123, + "step": 17416 + }, + { + "epoch": 1.12, + "learning_rate": 1.6284377079799912e-05, + "loss": 1.004, + "step": 17417 + }, + { + "epoch": 1.12, + "learning_rate": 1.628416239077696e-05, + "loss": 0.9717, + "step": 17418 + }, + { + "epoch": 1.12, + "learning_rate": 1.628394770175401e-05, + "loss": 0.9908, + "step": 17419 + }, + { + "epoch": 1.12, + "learning_rate": 1.628373301273106e-05, + "loss": 0.9251, + "step": 17420 + }, + { + "epoch": 1.12, + "learning_rate": 1.6283518323708108e-05, + "loss": 0.9392, + "step": 17421 + }, + { + "epoch": 1.12, + "learning_rate": 1.628330363468516e-05, + "loss": 1.0337, + "step": 17422 + }, + { + "epoch": 1.12, + "learning_rate": 1.628308894566221e-05, + "loss": 1.0146, + "step": 17423 + }, + { + "epoch": 1.12, + "learning_rate": 1.628287425663926e-05, + "loss": 0.949, + "step": 17424 + }, + { + "epoch": 1.12, + "learning_rate": 1.628265956761631e-05, + "loss": 1.0091, + "step": 17425 + }, + { + "epoch": 1.12, + "learning_rate": 1.628244487859336e-05, + "loss": 0.9751, + "step": 17426 + }, + { + "epoch": 1.12, + "learning_rate": 1.628223018957041e-05, + "loss": 1.0045, + "step": 17427 + }, + { + "epoch": 1.12, + "learning_rate": 1.6282015500547458e-05, + "loss": 0.9789, + "step": 17428 + }, + { + "epoch": 1.12, + "learning_rate": 1.628180081152451e-05, + "loss": 0.9664, + "step": 17429 + }, + { + "epoch": 1.12, + "learning_rate": 1.628158612250156e-05, + "loss": 0.9465, + "step": 17430 + }, + { + "epoch": 1.12, + "learning_rate": 1.6281371433478608e-05, + "loss": 0.9693, + "step": 17431 + }, + { + "epoch": 1.12, + "learning_rate": 1.6281156744455657e-05, + "loss": 0.9669, + "step": 17432 + }, + { + "epoch": 1.12, + "learning_rate": 1.6280942055432706e-05, + "loss": 0.9537, + "step": 17433 + }, + { + "epoch": 1.12, + "learning_rate": 1.6280727366409755e-05, + "loss": 1.0313, + "step": 17434 + }, + { + "epoch": 1.12, + "learning_rate": 1.6280512677386807e-05, + "loss": 0.9591, + "step": 17435 + }, + { + "epoch": 1.12, + "learning_rate": 1.6280297988363856e-05, + "loss": 0.9785, + "step": 17436 + }, + { + "epoch": 1.12, + "learning_rate": 1.6280083299340905e-05, + "loss": 0.9672, + "step": 17437 + }, + { + "epoch": 1.12, + "learning_rate": 1.6279868610317954e-05, + "loss": 0.9899, + "step": 17438 + }, + { + "epoch": 1.12, + "learning_rate": 1.6279653921295003e-05, + "loss": 0.9793, + "step": 17439 + }, + { + "epoch": 1.12, + "learning_rate": 1.6279439232272055e-05, + "loss": 1.0162, + "step": 17440 + }, + { + "epoch": 1.12, + "learning_rate": 1.6279224543249104e-05, + "loss": 1.0385, + "step": 17441 + }, + { + "epoch": 1.12, + "learning_rate": 1.6279009854226157e-05, + "loss": 0.9058, + "step": 17442 + }, + { + "epoch": 1.12, + "learning_rate": 1.6278795165203206e-05, + "loss": 1.0531, + "step": 17443 + }, + { + "epoch": 1.12, + "learning_rate": 1.6278580476180255e-05, + "loss": 0.9567, + "step": 17444 + }, + { + "epoch": 1.12, + "learning_rate": 1.6278365787157304e-05, + "loss": 0.932, + "step": 17445 + }, + { + "epoch": 1.12, + "learning_rate": 1.6278151098134353e-05, + "loss": 0.9859, + "step": 17446 + }, + { + "epoch": 1.12, + "learning_rate": 1.6277936409111405e-05, + "loss": 0.9641, + "step": 17447 + }, + { + "epoch": 1.12, + "learning_rate": 1.6277721720088454e-05, + "loss": 0.9965, + "step": 17448 + }, + { + "epoch": 1.12, + "learning_rate": 1.6277507031065503e-05, + "loss": 0.9632, + "step": 17449 + }, + { + "epoch": 1.12, + "learning_rate": 1.6277292342042552e-05, + "loss": 1.0192, + "step": 17450 + }, + { + "epoch": 1.12, + "learning_rate": 1.62770776530196e-05, + "loss": 0.9728, + "step": 17451 + }, + { + "epoch": 1.12, + "learning_rate": 1.627686296399665e-05, + "loss": 0.9161, + "step": 17452 + }, + { + "epoch": 1.12, + "learning_rate": 1.6276648274973702e-05, + "loss": 1.0323, + "step": 17453 + }, + { + "epoch": 1.12, + "learning_rate": 1.627643358595075e-05, + "loss": 0.9223, + "step": 17454 + }, + { + "epoch": 1.12, + "learning_rate": 1.6276218896927804e-05, + "loss": 1.0493, + "step": 17455 + }, + { + "epoch": 1.12, + "learning_rate": 1.6276004207904852e-05, + "loss": 0.9822, + "step": 17456 + }, + { + "epoch": 1.12, + "learning_rate": 1.62757895188819e-05, + "loss": 0.9594, + "step": 17457 + }, + { + "epoch": 1.12, + "learning_rate": 1.627557482985895e-05, + "loss": 1.0556, + "step": 17458 + }, + { + "epoch": 1.12, + "learning_rate": 1.6275360140836e-05, + "loss": 1.055, + "step": 17459 + }, + { + "epoch": 1.12, + "learning_rate": 1.6275145451813052e-05, + "loss": 0.9901, + "step": 17460 + }, + { + "epoch": 1.12, + "learning_rate": 1.62749307627901e-05, + "loss": 0.8791, + "step": 17461 + }, + { + "epoch": 1.12, + "learning_rate": 1.627471607376715e-05, + "loss": 0.974, + "step": 17462 + }, + { + "epoch": 1.12, + "learning_rate": 1.62745013847442e-05, + "loss": 0.9134, + "step": 17463 + }, + { + "epoch": 1.12, + "learning_rate": 1.6274286695721248e-05, + "loss": 1.055, + "step": 17464 + }, + { + "epoch": 1.12, + "learning_rate": 1.6274072006698297e-05, + "loss": 0.9402, + "step": 17465 + }, + { + "epoch": 1.12, + "learning_rate": 1.627385731767535e-05, + "loss": 1.0025, + "step": 17466 + }, + { + "epoch": 1.12, + "learning_rate": 1.6273642628652398e-05, + "loss": 0.9655, + "step": 17467 + }, + { + "epoch": 1.12, + "learning_rate": 1.6273427939629447e-05, + "loss": 0.9716, + "step": 17468 + }, + { + "epoch": 1.12, + "learning_rate": 1.62732132506065e-05, + "loss": 0.9915, + "step": 17469 + }, + { + "epoch": 1.12, + "learning_rate": 1.6272998561583548e-05, + "loss": 0.9782, + "step": 17470 + }, + { + "epoch": 1.12, + "learning_rate": 1.6272783872560597e-05, + "loss": 0.9826, + "step": 17471 + }, + { + "epoch": 1.12, + "learning_rate": 1.6272569183537646e-05, + "loss": 1.0631, + "step": 17472 + }, + { + "epoch": 1.12, + "learning_rate": 1.62723544945147e-05, + "loss": 0.984, + "step": 17473 + }, + { + "epoch": 1.12, + "learning_rate": 1.6272139805491747e-05, + "loss": 0.9603, + "step": 17474 + }, + { + "epoch": 1.12, + "learning_rate": 1.6271925116468796e-05, + "loss": 0.939, + "step": 17475 + }, + { + "epoch": 1.12, + "learning_rate": 1.6271710427445845e-05, + "loss": 0.945, + "step": 17476 + }, + { + "epoch": 1.12, + "learning_rate": 1.6271495738422894e-05, + "loss": 1.0229, + "step": 17477 + }, + { + "epoch": 1.12, + "learning_rate": 1.6271281049399947e-05, + "loss": 0.9919, + "step": 17478 + }, + { + "epoch": 1.12, + "learning_rate": 1.6271066360376996e-05, + "loss": 0.992, + "step": 17479 + }, + { + "epoch": 1.12, + "learning_rate": 1.6270851671354045e-05, + "loss": 0.9642, + "step": 17480 + }, + { + "epoch": 1.12, + "learning_rate": 1.6270636982331094e-05, + "loss": 0.9992, + "step": 17481 + }, + { + "epoch": 1.12, + "learning_rate": 1.6270422293308143e-05, + "loss": 0.8762, + "step": 17482 + }, + { + "epoch": 1.12, + "learning_rate": 1.6270207604285195e-05, + "loss": 0.9651, + "step": 17483 + }, + { + "epoch": 1.12, + "learning_rate": 1.6269992915262244e-05, + "loss": 1.016, + "step": 17484 + }, + { + "epoch": 1.12, + "learning_rate": 1.6269778226239293e-05, + "loss": 1.0495, + "step": 17485 + }, + { + "epoch": 1.13, + "learning_rate": 1.6269563537216345e-05, + "loss": 1.0374, + "step": 17486 + }, + { + "epoch": 1.13, + "learning_rate": 1.6269348848193394e-05, + "loss": 1.0119, + "step": 17487 + }, + { + "epoch": 1.13, + "learning_rate": 1.6269134159170443e-05, + "loss": 0.9149, + "step": 17488 + }, + { + "epoch": 1.13, + "learning_rate": 1.6268919470147492e-05, + "loss": 0.9904, + "step": 17489 + }, + { + "epoch": 1.13, + "learning_rate": 1.626870478112454e-05, + "loss": 0.9856, + "step": 17490 + }, + { + "epoch": 1.13, + "learning_rate": 1.6268490092101593e-05, + "loss": 0.9886, + "step": 17491 + }, + { + "epoch": 1.13, + "learning_rate": 1.6268275403078642e-05, + "loss": 0.9491, + "step": 17492 + }, + { + "epoch": 1.13, + "learning_rate": 1.626806071405569e-05, + "loss": 1.0332, + "step": 17493 + }, + { + "epoch": 1.13, + "learning_rate": 1.626784602503274e-05, + "loss": 1.0161, + "step": 17494 + }, + { + "epoch": 1.13, + "learning_rate": 1.626763133600979e-05, + "loss": 0.994, + "step": 17495 + }, + { + "epoch": 1.13, + "learning_rate": 1.626741664698684e-05, + "loss": 0.9107, + "step": 17496 + }, + { + "epoch": 1.13, + "learning_rate": 1.626720195796389e-05, + "loss": 1.0165, + "step": 17497 + }, + { + "epoch": 1.13, + "learning_rate": 1.6266987268940943e-05, + "loss": 1.072, + "step": 17498 + }, + { + "epoch": 1.13, + "learning_rate": 1.6266772579917992e-05, + "loss": 0.9965, + "step": 17499 + }, + { + "epoch": 1.13, + "learning_rate": 1.626655789089504e-05, + "loss": 1.0518, + "step": 17500 + }, + { + "epoch": 1.13, + "learning_rate": 1.626634320187209e-05, + "loss": 1.0117, + "step": 17501 + }, + { + "epoch": 1.13, + "learning_rate": 1.626612851284914e-05, + "loss": 0.9111, + "step": 17502 + }, + { + "epoch": 1.13, + "learning_rate": 1.6265913823826188e-05, + "loss": 1.0047, + "step": 17503 + }, + { + "epoch": 1.13, + "learning_rate": 1.626569913480324e-05, + "loss": 0.9806, + "step": 17504 + }, + { + "epoch": 1.13, + "learning_rate": 1.626548444578029e-05, + "loss": 0.9901, + "step": 17505 + }, + { + "epoch": 1.13, + "learning_rate": 1.6265269756757338e-05, + "loss": 1.0817, + "step": 17506 + }, + { + "epoch": 1.13, + "learning_rate": 1.6265055067734387e-05, + "loss": 0.973, + "step": 17507 + }, + { + "epoch": 1.13, + "learning_rate": 1.6264840378711436e-05, + "loss": 0.9792, + "step": 17508 + }, + { + "epoch": 1.13, + "learning_rate": 1.6264625689688488e-05, + "loss": 1.0119, + "step": 17509 + }, + { + "epoch": 1.13, + "learning_rate": 1.6264411000665537e-05, + "loss": 1.009, + "step": 17510 + }, + { + "epoch": 1.13, + "learning_rate": 1.6264196311642586e-05, + "loss": 0.9309, + "step": 17511 + }, + { + "epoch": 1.13, + "learning_rate": 1.626398162261964e-05, + "loss": 1.0585, + "step": 17512 + }, + { + "epoch": 1.13, + "learning_rate": 1.6263766933596688e-05, + "loss": 0.9733, + "step": 17513 + }, + { + "epoch": 1.13, + "learning_rate": 1.6263552244573737e-05, + "loss": 1.0527, + "step": 17514 + }, + { + "epoch": 1.13, + "learning_rate": 1.6263337555550785e-05, + "loss": 0.9445, + "step": 17515 + }, + { + "epoch": 1.13, + "learning_rate": 1.6263122866527838e-05, + "loss": 0.9995, + "step": 17516 + }, + { + "epoch": 1.13, + "learning_rate": 1.6262908177504887e-05, + "loss": 0.9334, + "step": 17517 + }, + { + "epoch": 1.13, + "learning_rate": 1.6262693488481936e-05, + "loss": 0.9351, + "step": 17518 + }, + { + "epoch": 1.13, + "learning_rate": 1.6262478799458985e-05, + "loss": 1.0309, + "step": 17519 + }, + { + "epoch": 1.13, + "learning_rate": 1.6262264110436034e-05, + "loss": 0.9924, + "step": 17520 + }, + { + "epoch": 1.13, + "learning_rate": 1.6262049421413083e-05, + "loss": 0.9128, + "step": 17521 + }, + { + "epoch": 1.13, + "learning_rate": 1.6261834732390135e-05, + "loss": 0.9505, + "step": 17522 + }, + { + "epoch": 1.13, + "learning_rate": 1.6261620043367184e-05, + "loss": 0.9583, + "step": 17523 + }, + { + "epoch": 1.13, + "learning_rate": 1.6261405354344233e-05, + "loss": 1.0281, + "step": 17524 + }, + { + "epoch": 1.13, + "learning_rate": 1.6261190665321282e-05, + "loss": 1.0534, + "step": 17525 + }, + { + "epoch": 1.13, + "learning_rate": 1.626097597629833e-05, + "loss": 0.9456, + "step": 17526 + }, + { + "epoch": 1.13, + "learning_rate": 1.6260761287275383e-05, + "loss": 1.0465, + "step": 17527 + }, + { + "epoch": 1.13, + "learning_rate": 1.6260546598252432e-05, + "loss": 0.9581, + "step": 17528 + }, + { + "epoch": 1.13, + "learning_rate": 1.6260331909229485e-05, + "loss": 0.9812, + "step": 17529 + }, + { + "epoch": 1.13, + "learning_rate": 1.6260117220206533e-05, + "loss": 0.9744, + "step": 17530 + }, + { + "epoch": 1.13, + "learning_rate": 1.6259902531183582e-05, + "loss": 0.9683, + "step": 17531 + }, + { + "epoch": 1.13, + "learning_rate": 1.625968784216063e-05, + "loss": 1.0057, + "step": 17532 + }, + { + "epoch": 1.13, + "learning_rate": 1.625947315313768e-05, + "loss": 1.012, + "step": 17533 + }, + { + "epoch": 1.13, + "learning_rate": 1.6259258464114733e-05, + "loss": 0.9889, + "step": 17534 + }, + { + "epoch": 1.13, + "learning_rate": 1.6259043775091782e-05, + "loss": 1.0004, + "step": 17535 + }, + { + "epoch": 1.13, + "learning_rate": 1.625882908606883e-05, + "loss": 1.0445, + "step": 17536 + }, + { + "epoch": 1.13, + "learning_rate": 1.625861439704588e-05, + "loss": 0.9734, + "step": 17537 + }, + { + "epoch": 1.13, + "learning_rate": 1.625839970802293e-05, + "loss": 0.9769, + "step": 17538 + }, + { + "epoch": 1.13, + "learning_rate": 1.6258185018999978e-05, + "loss": 0.8798, + "step": 17539 + }, + { + "epoch": 1.13, + "learning_rate": 1.625797032997703e-05, + "loss": 1.0416, + "step": 17540 + }, + { + "epoch": 1.13, + "learning_rate": 1.625775564095408e-05, + "loss": 0.9936, + "step": 17541 + }, + { + "epoch": 1.13, + "learning_rate": 1.625754095193113e-05, + "loss": 0.9493, + "step": 17542 + }, + { + "epoch": 1.13, + "learning_rate": 1.625732626290818e-05, + "loss": 0.9592, + "step": 17543 + }, + { + "epoch": 1.13, + "learning_rate": 1.625711157388523e-05, + "loss": 0.9647, + "step": 17544 + }, + { + "epoch": 1.13, + "learning_rate": 1.6256896884862278e-05, + "loss": 0.9249, + "step": 17545 + }, + { + "epoch": 1.13, + "learning_rate": 1.6256682195839327e-05, + "loss": 0.9418, + "step": 17546 + }, + { + "epoch": 1.13, + "learning_rate": 1.625646750681638e-05, + "loss": 0.9665, + "step": 17547 + }, + { + "epoch": 1.13, + "learning_rate": 1.625625281779343e-05, + "loss": 0.9745, + "step": 17548 + }, + { + "epoch": 1.13, + "learning_rate": 1.6256038128770477e-05, + "loss": 0.9987, + "step": 17549 + }, + { + "epoch": 1.13, + "learning_rate": 1.6255823439747526e-05, + "loss": 0.9495, + "step": 17550 + }, + { + "epoch": 1.13, + "learning_rate": 1.6255608750724575e-05, + "loss": 0.9011, + "step": 17551 + }, + { + "epoch": 1.13, + "learning_rate": 1.6255394061701624e-05, + "loss": 1.0269, + "step": 17552 + }, + { + "epoch": 1.13, + "learning_rate": 1.6255179372678677e-05, + "loss": 0.9894, + "step": 17553 + }, + { + "epoch": 1.13, + "learning_rate": 1.6254964683655726e-05, + "loss": 0.9937, + "step": 17554 + }, + { + "epoch": 1.13, + "learning_rate": 1.6254749994632775e-05, + "loss": 0.9594, + "step": 17555 + }, + { + "epoch": 1.13, + "learning_rate": 1.6254535305609827e-05, + "loss": 1.0295, + "step": 17556 + }, + { + "epoch": 1.13, + "learning_rate": 1.6254320616586876e-05, + "loss": 0.9269, + "step": 17557 + }, + { + "epoch": 1.13, + "learning_rate": 1.6254105927563925e-05, + "loss": 1.0403, + "step": 17558 + }, + { + "epoch": 1.13, + "learning_rate": 1.6253891238540974e-05, + "loss": 0.9779, + "step": 17559 + }, + { + "epoch": 1.13, + "learning_rate": 1.6253676549518026e-05, + "loss": 0.9435, + "step": 17560 + }, + { + "epoch": 1.13, + "learning_rate": 1.6253461860495075e-05, + "loss": 0.9133, + "step": 17561 + }, + { + "epoch": 1.13, + "learning_rate": 1.6253247171472124e-05, + "loss": 0.9722, + "step": 17562 + }, + { + "epoch": 1.13, + "learning_rate": 1.6253032482449173e-05, + "loss": 0.9574, + "step": 17563 + }, + { + "epoch": 1.13, + "learning_rate": 1.6252817793426222e-05, + "loss": 0.9956, + "step": 17564 + }, + { + "epoch": 1.13, + "learning_rate": 1.6252603104403274e-05, + "loss": 1.0065, + "step": 17565 + }, + { + "epoch": 1.13, + "learning_rate": 1.6252388415380323e-05, + "loss": 1.0012, + "step": 17566 + }, + { + "epoch": 1.13, + "learning_rate": 1.6252173726357372e-05, + "loss": 1.016, + "step": 17567 + }, + { + "epoch": 1.13, + "learning_rate": 1.625195903733442e-05, + "loss": 0.9952, + "step": 17568 + }, + { + "epoch": 1.13, + "learning_rate": 1.625174434831147e-05, + "loss": 0.9533, + "step": 17569 + }, + { + "epoch": 1.13, + "learning_rate": 1.6251529659288523e-05, + "loss": 0.9805, + "step": 17570 + }, + { + "epoch": 1.13, + "learning_rate": 1.625131497026557e-05, + "loss": 1.0136, + "step": 17571 + }, + { + "epoch": 1.13, + "learning_rate": 1.625110028124262e-05, + "loss": 1.0443, + "step": 17572 + }, + { + "epoch": 1.13, + "learning_rate": 1.6250885592219673e-05, + "loss": 0.9503, + "step": 17573 + }, + { + "epoch": 1.13, + "learning_rate": 1.6250670903196722e-05, + "loss": 0.9663, + "step": 17574 + }, + { + "epoch": 1.13, + "learning_rate": 1.625045621417377e-05, + "loss": 0.9638, + "step": 17575 + }, + { + "epoch": 1.13, + "learning_rate": 1.625024152515082e-05, + "loss": 0.9783, + "step": 17576 + }, + { + "epoch": 1.13, + "learning_rate": 1.625002683612787e-05, + "loss": 1.0671, + "step": 17577 + }, + { + "epoch": 1.13, + "learning_rate": 1.624981214710492e-05, + "loss": 0.9696, + "step": 17578 + }, + { + "epoch": 1.13, + "learning_rate": 1.624959745808197e-05, + "loss": 1.013, + "step": 17579 + }, + { + "epoch": 1.13, + "learning_rate": 1.624938276905902e-05, + "loss": 0.8887, + "step": 17580 + }, + { + "epoch": 1.13, + "learning_rate": 1.6249168080036068e-05, + "loss": 0.941, + "step": 17581 + }, + { + "epoch": 1.13, + "learning_rate": 1.6248953391013117e-05, + "loss": 0.912, + "step": 17582 + }, + { + "epoch": 1.13, + "learning_rate": 1.624873870199017e-05, + "loss": 0.9081, + "step": 17583 + }, + { + "epoch": 1.13, + "learning_rate": 1.6248524012967218e-05, + "loss": 0.9393, + "step": 17584 + }, + { + "epoch": 1.13, + "learning_rate": 1.624830932394427e-05, + "loss": 1.0034, + "step": 17585 + }, + { + "epoch": 1.13, + "learning_rate": 1.624809463492132e-05, + "loss": 0.9172, + "step": 17586 + }, + { + "epoch": 1.13, + "learning_rate": 1.624787994589837e-05, + "loss": 1.0386, + "step": 17587 + }, + { + "epoch": 1.13, + "learning_rate": 1.6247665256875418e-05, + "loss": 0.9878, + "step": 17588 + }, + { + "epoch": 1.13, + "learning_rate": 1.6247450567852466e-05, + "loss": 0.9794, + "step": 17589 + }, + { + "epoch": 1.13, + "learning_rate": 1.6247235878829515e-05, + "loss": 0.9845, + "step": 17590 + }, + { + "epoch": 1.13, + "learning_rate": 1.6247021189806568e-05, + "loss": 0.9631, + "step": 17591 + }, + { + "epoch": 1.13, + "learning_rate": 1.6246806500783617e-05, + "loss": 0.9319, + "step": 17592 + }, + { + "epoch": 1.13, + "learning_rate": 1.6246591811760666e-05, + "loss": 1.0684, + "step": 17593 + }, + { + "epoch": 1.13, + "learning_rate": 1.6246377122737715e-05, + "loss": 0.9954, + "step": 17594 + }, + { + "epoch": 1.13, + "learning_rate": 1.6246162433714764e-05, + "loss": 0.9731, + "step": 17595 + }, + { + "epoch": 1.13, + "learning_rate": 1.6245947744691816e-05, + "loss": 0.988, + "step": 17596 + }, + { + "epoch": 1.13, + "learning_rate": 1.6245733055668865e-05, + "loss": 0.9312, + "step": 17597 + }, + { + "epoch": 1.13, + "learning_rate": 1.6245518366645914e-05, + "loss": 1.0144, + "step": 17598 + }, + { + "epoch": 1.13, + "learning_rate": 1.6245303677622966e-05, + "loss": 0.9534, + "step": 17599 + }, + { + "epoch": 1.13, + "learning_rate": 1.6245088988600015e-05, + "loss": 0.9675, + "step": 17600 + }, + { + "epoch": 1.13, + "learning_rate": 1.6244874299577064e-05, + "loss": 0.9274, + "step": 17601 + }, + { + "epoch": 1.13, + "learning_rate": 1.6244659610554113e-05, + "loss": 0.9205, + "step": 17602 + }, + { + "epoch": 1.13, + "learning_rate": 1.6244444921531166e-05, + "loss": 0.9941, + "step": 17603 + }, + { + "epoch": 1.13, + "learning_rate": 1.6244230232508214e-05, + "loss": 1.0621, + "step": 17604 + }, + { + "epoch": 1.13, + "learning_rate": 1.6244015543485263e-05, + "loss": 1.021, + "step": 17605 + }, + { + "epoch": 1.13, + "learning_rate": 1.6243800854462312e-05, + "loss": 0.9608, + "step": 17606 + }, + { + "epoch": 1.13, + "learning_rate": 1.624358616543936e-05, + "loss": 0.9161, + "step": 17607 + }, + { + "epoch": 1.13, + "learning_rate": 1.624337147641641e-05, + "loss": 0.9431, + "step": 17608 + }, + { + "epoch": 1.13, + "learning_rate": 1.6243156787393463e-05, + "loss": 0.9316, + "step": 17609 + }, + { + "epoch": 1.13, + "learning_rate": 1.624294209837051e-05, + "loss": 0.9929, + "step": 17610 + }, + { + "epoch": 1.13, + "learning_rate": 1.624272740934756e-05, + "loss": 1.0077, + "step": 17611 + }, + { + "epoch": 1.13, + "learning_rate": 1.624251272032461e-05, + "loss": 0.919, + "step": 17612 + }, + { + "epoch": 1.13, + "learning_rate": 1.624229803130166e-05, + "loss": 0.9221, + "step": 17613 + }, + { + "epoch": 1.13, + "learning_rate": 1.624208334227871e-05, + "loss": 0.944, + "step": 17614 + }, + { + "epoch": 1.13, + "learning_rate": 1.624186865325576e-05, + "loss": 0.9103, + "step": 17615 + }, + { + "epoch": 1.13, + "learning_rate": 1.6241653964232812e-05, + "loss": 0.9801, + "step": 17616 + }, + { + "epoch": 1.13, + "learning_rate": 1.624143927520986e-05, + "loss": 0.979, + "step": 17617 + }, + { + "epoch": 1.13, + "learning_rate": 1.624122458618691e-05, + "loss": 0.9787, + "step": 17618 + }, + { + "epoch": 1.13, + "learning_rate": 1.624100989716396e-05, + "loss": 0.9758, + "step": 17619 + }, + { + "epoch": 1.13, + "learning_rate": 1.6240795208141008e-05, + "loss": 0.9563, + "step": 17620 + }, + { + "epoch": 1.13, + "learning_rate": 1.624058051911806e-05, + "loss": 0.9545, + "step": 17621 + }, + { + "epoch": 1.13, + "learning_rate": 1.624036583009511e-05, + "loss": 0.9681, + "step": 17622 + }, + { + "epoch": 1.13, + "learning_rate": 1.624015114107216e-05, + "loss": 1.0439, + "step": 17623 + }, + { + "epoch": 1.13, + "learning_rate": 1.6239936452049207e-05, + "loss": 1.0069, + "step": 17624 + }, + { + "epoch": 1.13, + "learning_rate": 1.6239721763026256e-05, + "loss": 0.9717, + "step": 17625 + }, + { + "epoch": 1.13, + "learning_rate": 1.6239507074003305e-05, + "loss": 0.9825, + "step": 17626 + }, + { + "epoch": 1.13, + "learning_rate": 1.6239292384980358e-05, + "loss": 0.9737, + "step": 17627 + }, + { + "epoch": 1.13, + "learning_rate": 1.6239077695957407e-05, + "loss": 0.9473, + "step": 17628 + }, + { + "epoch": 1.13, + "learning_rate": 1.623886300693446e-05, + "loss": 0.8757, + "step": 17629 + }, + { + "epoch": 1.13, + "learning_rate": 1.6238648317911508e-05, + "loss": 0.9167, + "step": 17630 + }, + { + "epoch": 1.13, + "learning_rate": 1.6238433628888557e-05, + "loss": 0.9712, + "step": 17631 + }, + { + "epoch": 1.13, + "learning_rate": 1.6238218939865606e-05, + "loss": 0.944, + "step": 17632 + }, + { + "epoch": 1.13, + "learning_rate": 1.6238004250842655e-05, + "loss": 0.989, + "step": 17633 + }, + { + "epoch": 1.13, + "learning_rate": 1.6237789561819707e-05, + "loss": 1.0281, + "step": 17634 + }, + { + "epoch": 1.13, + "learning_rate": 1.6237574872796756e-05, + "loss": 0.9903, + "step": 17635 + }, + { + "epoch": 1.13, + "learning_rate": 1.6237360183773805e-05, + "loss": 1.0139, + "step": 17636 + }, + { + "epoch": 1.13, + "learning_rate": 1.6237145494750854e-05, + "loss": 0.9806, + "step": 17637 + }, + { + "epoch": 1.13, + "learning_rate": 1.6236930805727903e-05, + "loss": 1.018, + "step": 17638 + }, + { + "epoch": 1.13, + "learning_rate": 1.6236716116704952e-05, + "loss": 0.9314, + "step": 17639 + }, + { + "epoch": 1.13, + "learning_rate": 1.6236501427682004e-05, + "loss": 0.9986, + "step": 17640 + }, + { + "epoch": 1.13, + "learning_rate": 1.6236286738659053e-05, + "loss": 1.107, + "step": 17641 + }, + { + "epoch": 1.14, + "learning_rate": 1.6236072049636102e-05, + "loss": 0.9613, + "step": 17642 + }, + { + "epoch": 1.14, + "learning_rate": 1.6235857360613155e-05, + "loss": 0.9093, + "step": 17643 + }, + { + "epoch": 1.14, + "learning_rate": 1.6235642671590204e-05, + "loss": 1.0346, + "step": 17644 + }, + { + "epoch": 1.14, + "learning_rate": 1.6235427982567253e-05, + "loss": 0.9567, + "step": 17645 + }, + { + "epoch": 1.14, + "learning_rate": 1.62352132935443e-05, + "loss": 1.0045, + "step": 17646 + }, + { + "epoch": 1.14, + "learning_rate": 1.6234998604521354e-05, + "loss": 1.0438, + "step": 17647 + }, + { + "epoch": 1.14, + "learning_rate": 1.6234783915498403e-05, + "loss": 0.8479, + "step": 17648 + }, + { + "epoch": 1.14, + "learning_rate": 1.6234569226475452e-05, + "loss": 1.006, + "step": 17649 + }, + { + "epoch": 1.14, + "learning_rate": 1.62343545374525e-05, + "loss": 0.9882, + "step": 17650 + }, + { + "epoch": 1.14, + "learning_rate": 1.623413984842955e-05, + "loss": 1.06, + "step": 17651 + }, + { + "epoch": 1.14, + "learning_rate": 1.6233925159406602e-05, + "loss": 0.9738, + "step": 17652 + }, + { + "epoch": 1.14, + "learning_rate": 1.623371047038365e-05, + "loss": 1.0217, + "step": 17653 + }, + { + "epoch": 1.14, + "learning_rate": 1.62334957813607e-05, + "loss": 0.9138, + "step": 17654 + }, + { + "epoch": 1.14, + "learning_rate": 1.623328109233775e-05, + "loss": 0.9282, + "step": 17655 + }, + { + "epoch": 1.14, + "learning_rate": 1.6233066403314798e-05, + "loss": 0.9431, + "step": 17656 + }, + { + "epoch": 1.14, + "learning_rate": 1.623285171429185e-05, + "loss": 0.9551, + "step": 17657 + }, + { + "epoch": 1.14, + "learning_rate": 1.62326370252689e-05, + "loss": 0.978, + "step": 17658 + }, + { + "epoch": 1.14, + "learning_rate": 1.6232422336245948e-05, + "loss": 0.9824, + "step": 17659 + }, + { + "epoch": 1.14, + "learning_rate": 1.6232207647223e-05, + "loss": 1.0929, + "step": 17660 + }, + { + "epoch": 1.14, + "learning_rate": 1.623199295820005e-05, + "loss": 0.944, + "step": 17661 + }, + { + "epoch": 1.14, + "learning_rate": 1.62317782691771e-05, + "loss": 0.9862, + "step": 17662 + }, + { + "epoch": 1.14, + "learning_rate": 1.6231563580154147e-05, + "loss": 1.0085, + "step": 17663 + }, + { + "epoch": 1.14, + "learning_rate": 1.6231348891131196e-05, + "loss": 0.9978, + "step": 17664 + }, + { + "epoch": 1.14, + "learning_rate": 1.623113420210825e-05, + "loss": 0.9245, + "step": 17665 + }, + { + "epoch": 1.14, + "learning_rate": 1.6230919513085298e-05, + "loss": 0.9981, + "step": 17666 + }, + { + "epoch": 1.14, + "learning_rate": 1.6230704824062347e-05, + "loss": 0.9994, + "step": 17667 + }, + { + "epoch": 1.14, + "learning_rate": 1.6230490135039396e-05, + "loss": 0.8591, + "step": 17668 + }, + { + "epoch": 1.14, + "learning_rate": 1.6230275446016445e-05, + "loss": 0.9395, + "step": 17669 + }, + { + "epoch": 1.14, + "learning_rate": 1.6230060756993497e-05, + "loss": 0.9825, + "step": 17670 + }, + { + "epoch": 1.14, + "learning_rate": 1.6229846067970546e-05, + "loss": 0.9449, + "step": 17671 + }, + { + "epoch": 1.14, + "learning_rate": 1.62296313789476e-05, + "loss": 1.0717, + "step": 17672 + }, + { + "epoch": 1.14, + "learning_rate": 1.6229416689924647e-05, + "loss": 0.9782, + "step": 17673 + }, + { + "epoch": 1.14, + "learning_rate": 1.6229202000901696e-05, + "loss": 1.0016, + "step": 17674 + }, + { + "epoch": 1.14, + "learning_rate": 1.6228987311878745e-05, + "loss": 1.0335, + "step": 17675 + }, + { + "epoch": 1.14, + "learning_rate": 1.6228772622855794e-05, + "loss": 0.9879, + "step": 17676 + }, + { + "epoch": 1.14, + "learning_rate": 1.6228557933832843e-05, + "loss": 0.9354, + "step": 17677 + }, + { + "epoch": 1.14, + "learning_rate": 1.6228343244809895e-05, + "loss": 0.9224, + "step": 17678 + }, + { + "epoch": 1.14, + "learning_rate": 1.6228128555786944e-05, + "loss": 0.9681, + "step": 17679 + }, + { + "epoch": 1.14, + "learning_rate": 1.6227913866763993e-05, + "loss": 0.9935, + "step": 17680 + }, + { + "epoch": 1.14, + "learning_rate": 1.6227699177741042e-05, + "loss": 0.9938, + "step": 17681 + }, + { + "epoch": 1.14, + "learning_rate": 1.622748448871809e-05, + "loss": 0.9661, + "step": 17682 + }, + { + "epoch": 1.14, + "learning_rate": 1.6227269799695144e-05, + "loss": 1.0133, + "step": 17683 + }, + { + "epoch": 1.14, + "learning_rate": 1.6227055110672193e-05, + "loss": 0.9545, + "step": 17684 + }, + { + "epoch": 1.14, + "learning_rate": 1.622684042164924e-05, + "loss": 0.9576, + "step": 17685 + }, + { + "epoch": 1.14, + "learning_rate": 1.622662573262629e-05, + "loss": 0.9638, + "step": 17686 + }, + { + "epoch": 1.14, + "learning_rate": 1.6226411043603343e-05, + "loss": 0.978, + "step": 17687 + }, + { + "epoch": 1.14, + "learning_rate": 1.6226196354580392e-05, + "loss": 1.0271, + "step": 17688 + }, + { + "epoch": 1.14, + "learning_rate": 1.622598166555744e-05, + "loss": 0.967, + "step": 17689 + }, + { + "epoch": 1.14, + "learning_rate": 1.6225766976534493e-05, + "loss": 0.9781, + "step": 17690 + }, + { + "epoch": 1.14, + "learning_rate": 1.6225552287511542e-05, + "loss": 1.0057, + "step": 17691 + }, + { + "epoch": 1.14, + "learning_rate": 1.622533759848859e-05, + "loss": 0.9934, + "step": 17692 + }, + { + "epoch": 1.14, + "learning_rate": 1.622512290946564e-05, + "loss": 1.026, + "step": 17693 + }, + { + "epoch": 1.14, + "learning_rate": 1.622490822044269e-05, + "loss": 1.0837, + "step": 17694 + }, + { + "epoch": 1.14, + "learning_rate": 1.6224693531419738e-05, + "loss": 1.0147, + "step": 17695 + }, + { + "epoch": 1.14, + "learning_rate": 1.622447884239679e-05, + "loss": 1.0182, + "step": 17696 + }, + { + "epoch": 1.14, + "learning_rate": 1.622426415337384e-05, + "loss": 1.0158, + "step": 17697 + }, + { + "epoch": 1.14, + "learning_rate": 1.622404946435089e-05, + "loss": 0.9335, + "step": 17698 + }, + { + "epoch": 1.14, + "learning_rate": 1.6223834775327937e-05, + "loss": 1.0476, + "step": 17699 + }, + { + "epoch": 1.14, + "learning_rate": 1.6223620086304986e-05, + "loss": 0.9625, + "step": 17700 + }, + { + "epoch": 1.14, + "learning_rate": 1.622340539728204e-05, + "loss": 0.9788, + "step": 17701 + }, + { + "epoch": 1.14, + "learning_rate": 1.6223190708259088e-05, + "loss": 0.9855, + "step": 17702 + }, + { + "epoch": 1.14, + "learning_rate": 1.622297601923614e-05, + "loss": 0.9553, + "step": 17703 + }, + { + "epoch": 1.14, + "learning_rate": 1.622276133021319e-05, + "loss": 0.9381, + "step": 17704 + }, + { + "epoch": 1.14, + "learning_rate": 1.6222546641190238e-05, + "loss": 0.9822, + "step": 17705 + }, + { + "epoch": 1.14, + "learning_rate": 1.6222331952167287e-05, + "loss": 1.0112, + "step": 17706 + }, + { + "epoch": 1.14, + "learning_rate": 1.6222117263144336e-05, + "loss": 1.0467, + "step": 17707 + }, + { + "epoch": 1.14, + "learning_rate": 1.6221902574121385e-05, + "loss": 0.9552, + "step": 17708 + }, + { + "epoch": 1.14, + "learning_rate": 1.6221687885098437e-05, + "loss": 0.9494, + "step": 17709 + }, + { + "epoch": 1.14, + "learning_rate": 1.6221473196075486e-05, + "loss": 0.9386, + "step": 17710 + }, + { + "epoch": 1.14, + "learning_rate": 1.6221258507052535e-05, + "loss": 0.942, + "step": 17711 + }, + { + "epoch": 1.14, + "learning_rate": 1.6221043818029584e-05, + "loss": 0.9339, + "step": 17712 + }, + { + "epoch": 1.14, + "learning_rate": 1.6220829129006633e-05, + "loss": 1.0113, + "step": 17713 + }, + { + "epoch": 1.14, + "learning_rate": 1.6220614439983685e-05, + "loss": 1.03, + "step": 17714 + }, + { + "epoch": 1.14, + "learning_rate": 1.6220399750960734e-05, + "loss": 1.0079, + "step": 17715 + }, + { + "epoch": 1.14, + "learning_rate": 1.6220185061937787e-05, + "loss": 0.9666, + "step": 17716 + }, + { + "epoch": 1.14, + "learning_rate": 1.6219970372914836e-05, + "loss": 0.9463, + "step": 17717 + }, + { + "epoch": 1.14, + "learning_rate": 1.6219755683891885e-05, + "loss": 0.9946, + "step": 17718 + }, + { + "epoch": 1.14, + "learning_rate": 1.6219540994868934e-05, + "loss": 0.9894, + "step": 17719 + }, + { + "epoch": 1.14, + "learning_rate": 1.6219326305845983e-05, + "loss": 1.0073, + "step": 17720 + }, + { + "epoch": 1.14, + "learning_rate": 1.6219111616823035e-05, + "loss": 0.8966, + "step": 17721 + }, + { + "epoch": 1.14, + "learning_rate": 1.6218896927800084e-05, + "loss": 1.069, + "step": 17722 + }, + { + "epoch": 1.14, + "learning_rate": 1.6218682238777133e-05, + "loss": 0.9814, + "step": 17723 + }, + { + "epoch": 1.14, + "learning_rate": 1.6218467549754182e-05, + "loss": 0.9861, + "step": 17724 + }, + { + "epoch": 1.14, + "learning_rate": 1.621825286073123e-05, + "loss": 0.9859, + "step": 17725 + }, + { + "epoch": 1.14, + "learning_rate": 1.621803817170828e-05, + "loss": 0.9911, + "step": 17726 + }, + { + "epoch": 1.14, + "learning_rate": 1.6217823482685332e-05, + "loss": 0.9388, + "step": 17727 + }, + { + "epoch": 1.14, + "learning_rate": 1.621760879366238e-05, + "loss": 0.9935, + "step": 17728 + }, + { + "epoch": 1.14, + "learning_rate": 1.621739410463943e-05, + "loss": 0.9167, + "step": 17729 + }, + { + "epoch": 1.14, + "learning_rate": 1.6217179415616482e-05, + "loss": 1.0157, + "step": 17730 + }, + { + "epoch": 1.14, + "learning_rate": 1.621696472659353e-05, + "loss": 0.9584, + "step": 17731 + }, + { + "epoch": 1.14, + "learning_rate": 1.621675003757058e-05, + "loss": 0.9341, + "step": 17732 + }, + { + "epoch": 1.14, + "learning_rate": 1.621653534854763e-05, + "loss": 0.927, + "step": 17733 + }, + { + "epoch": 1.14, + "learning_rate": 1.621632065952468e-05, + "loss": 0.9597, + "step": 17734 + }, + { + "epoch": 1.14, + "learning_rate": 1.621610597050173e-05, + "loss": 0.9902, + "step": 17735 + }, + { + "epoch": 1.14, + "learning_rate": 1.621589128147878e-05, + "loss": 0.9601, + "step": 17736 + }, + { + "epoch": 1.14, + "learning_rate": 1.621567659245583e-05, + "loss": 1.0161, + "step": 17737 + }, + { + "epoch": 1.14, + "learning_rate": 1.6215461903432877e-05, + "loss": 0.9944, + "step": 17738 + }, + { + "epoch": 1.14, + "learning_rate": 1.621524721440993e-05, + "loss": 0.9566, + "step": 17739 + }, + { + "epoch": 1.14, + "learning_rate": 1.621503252538698e-05, + "loss": 0.982, + "step": 17740 + }, + { + "epoch": 1.14, + "learning_rate": 1.6214817836364028e-05, + "loss": 0.95, + "step": 17741 + }, + { + "epoch": 1.14, + "learning_rate": 1.6214603147341077e-05, + "loss": 1.0533, + "step": 17742 + }, + { + "epoch": 1.14, + "learning_rate": 1.6214388458318126e-05, + "loss": 0.9094, + "step": 17743 + }, + { + "epoch": 1.14, + "learning_rate": 1.6214173769295175e-05, + "loss": 0.9657, + "step": 17744 + }, + { + "epoch": 1.14, + "learning_rate": 1.6213959080272227e-05, + "loss": 1.0186, + "step": 17745 + }, + { + "epoch": 1.14, + "learning_rate": 1.6213744391249276e-05, + "loss": 0.8702, + "step": 17746 + }, + { + "epoch": 1.14, + "learning_rate": 1.6213529702226328e-05, + "loss": 0.9514, + "step": 17747 + }, + { + "epoch": 1.14, + "learning_rate": 1.6213315013203377e-05, + "loss": 0.9771, + "step": 17748 + }, + { + "epoch": 1.14, + "learning_rate": 1.6213100324180426e-05, + "loss": 0.9897, + "step": 17749 + }, + { + "epoch": 1.14, + "learning_rate": 1.6212885635157475e-05, + "loss": 0.9118, + "step": 17750 + }, + { + "epoch": 1.14, + "learning_rate": 1.6212670946134524e-05, + "loss": 0.9377, + "step": 17751 + }, + { + "epoch": 1.14, + "learning_rate": 1.6212456257111577e-05, + "loss": 0.9187, + "step": 17752 + }, + { + "epoch": 1.14, + "learning_rate": 1.6212241568088625e-05, + "loss": 0.9642, + "step": 17753 + }, + { + "epoch": 1.14, + "learning_rate": 1.6212026879065674e-05, + "loss": 0.9899, + "step": 17754 + }, + { + "epoch": 1.14, + "learning_rate": 1.6211812190042723e-05, + "loss": 0.96, + "step": 17755 + }, + { + "epoch": 1.14, + "learning_rate": 1.6211597501019772e-05, + "loss": 0.9117, + "step": 17756 + }, + { + "epoch": 1.14, + "learning_rate": 1.6211382811996825e-05, + "loss": 0.9052, + "step": 17757 + }, + { + "epoch": 1.14, + "learning_rate": 1.6211168122973874e-05, + "loss": 1.0506, + "step": 17758 + }, + { + "epoch": 1.14, + "learning_rate": 1.6210953433950926e-05, + "loss": 0.9683, + "step": 17759 + }, + { + "epoch": 1.14, + "learning_rate": 1.6210738744927975e-05, + "loss": 0.9578, + "step": 17760 + }, + { + "epoch": 1.14, + "learning_rate": 1.6210524055905024e-05, + "loss": 0.981, + "step": 17761 + }, + { + "epoch": 1.14, + "learning_rate": 1.6210309366882073e-05, + "loss": 0.9974, + "step": 17762 + }, + { + "epoch": 1.14, + "learning_rate": 1.6210094677859122e-05, + "loss": 0.9799, + "step": 17763 + }, + { + "epoch": 1.14, + "learning_rate": 1.620987998883617e-05, + "loss": 0.9433, + "step": 17764 + }, + { + "epoch": 1.14, + "learning_rate": 1.6209665299813223e-05, + "loss": 0.9964, + "step": 17765 + }, + { + "epoch": 1.14, + "learning_rate": 1.6209450610790272e-05, + "loss": 0.9635, + "step": 17766 + }, + { + "epoch": 1.14, + "learning_rate": 1.620923592176732e-05, + "loss": 0.9705, + "step": 17767 + }, + { + "epoch": 1.14, + "learning_rate": 1.620902123274437e-05, + "loss": 0.969, + "step": 17768 + }, + { + "epoch": 1.14, + "learning_rate": 1.620880654372142e-05, + "loss": 1.0307, + "step": 17769 + }, + { + "epoch": 1.14, + "learning_rate": 1.620859185469847e-05, + "loss": 1.0068, + "step": 17770 + }, + { + "epoch": 1.14, + "learning_rate": 1.620837716567552e-05, + "loss": 0.8836, + "step": 17771 + }, + { + "epoch": 1.14, + "learning_rate": 1.620816247665257e-05, + "loss": 1.0291, + "step": 17772 + }, + { + "epoch": 1.14, + "learning_rate": 1.620794778762962e-05, + "loss": 1.0364, + "step": 17773 + }, + { + "epoch": 1.14, + "learning_rate": 1.620773309860667e-05, + "loss": 0.9564, + "step": 17774 + }, + { + "epoch": 1.14, + "learning_rate": 1.620751840958372e-05, + "loss": 1.0418, + "step": 17775 + }, + { + "epoch": 1.14, + "learning_rate": 1.620730372056077e-05, + "loss": 0.9934, + "step": 17776 + }, + { + "epoch": 1.14, + "learning_rate": 1.620708903153782e-05, + "loss": 0.9692, + "step": 17777 + }, + { + "epoch": 1.14, + "learning_rate": 1.620687434251487e-05, + "loss": 0.9747, + "step": 17778 + }, + { + "epoch": 1.14, + "learning_rate": 1.620665965349192e-05, + "loss": 1.0237, + "step": 17779 + }, + { + "epoch": 1.14, + "learning_rate": 1.6206444964468968e-05, + "loss": 1.1238, + "step": 17780 + }, + { + "epoch": 1.14, + "learning_rate": 1.6206230275446017e-05, + "loss": 1.0077, + "step": 17781 + }, + { + "epoch": 1.14, + "learning_rate": 1.6206015586423066e-05, + "loss": 1.0537, + "step": 17782 + }, + { + "epoch": 1.14, + "learning_rate": 1.6205800897400118e-05, + "loss": 1.0811, + "step": 17783 + }, + { + "epoch": 1.14, + "learning_rate": 1.6205586208377167e-05, + "loss": 1.0117, + "step": 17784 + }, + { + "epoch": 1.14, + "learning_rate": 1.6205371519354216e-05, + "loss": 0.9975, + "step": 17785 + }, + { + "epoch": 1.14, + "learning_rate": 1.6205156830331265e-05, + "loss": 0.9918, + "step": 17786 + }, + { + "epoch": 1.14, + "learning_rate": 1.6204942141308314e-05, + "loss": 1.0145, + "step": 17787 + }, + { + "epoch": 1.14, + "learning_rate": 1.6204727452285366e-05, + "loss": 1.0584, + "step": 17788 + }, + { + "epoch": 1.14, + "learning_rate": 1.6204512763262415e-05, + "loss": 1.0602, + "step": 17789 + }, + { + "epoch": 1.14, + "learning_rate": 1.6204298074239468e-05, + "loss": 0.9657, + "step": 17790 + }, + { + "epoch": 1.14, + "learning_rate": 1.6204083385216517e-05, + "loss": 0.9509, + "step": 17791 + }, + { + "epoch": 1.14, + "learning_rate": 1.6203868696193566e-05, + "loss": 0.992, + "step": 17792 + }, + { + "epoch": 1.14, + "learning_rate": 1.6203654007170615e-05, + "loss": 0.9899, + "step": 17793 + }, + { + "epoch": 1.14, + "learning_rate": 1.6203439318147664e-05, + "loss": 1.0045, + "step": 17794 + }, + { + "epoch": 1.14, + "learning_rate": 1.6203224629124712e-05, + "loss": 1.0573, + "step": 17795 + }, + { + "epoch": 1.14, + "learning_rate": 1.6203009940101765e-05, + "loss": 0.9993, + "step": 17796 + }, + { + "epoch": 1.15, + "learning_rate": 1.6202795251078814e-05, + "loss": 0.9295, + "step": 17797 + }, + { + "epoch": 1.15, + "learning_rate": 1.6202580562055863e-05, + "loss": 0.9906, + "step": 17798 + }, + { + "epoch": 1.15, + "learning_rate": 1.6202365873032912e-05, + "loss": 0.9583, + "step": 17799 + }, + { + "epoch": 1.15, + "learning_rate": 1.620215118400996e-05, + "loss": 0.9141, + "step": 17800 + }, + { + "epoch": 1.15, + "learning_rate": 1.6201936494987013e-05, + "loss": 0.9467, + "step": 17801 + }, + { + "epoch": 1.15, + "learning_rate": 1.6201721805964062e-05, + "loss": 0.9709, + "step": 17802 + }, + { + "epoch": 1.15, + "learning_rate": 1.6201507116941114e-05, + "loss": 0.9598, + "step": 17803 + }, + { + "epoch": 1.15, + "learning_rate": 1.6201292427918163e-05, + "loss": 0.9642, + "step": 17804 + }, + { + "epoch": 1.15, + "learning_rate": 1.6201077738895212e-05, + "loss": 0.96, + "step": 17805 + }, + { + "epoch": 1.15, + "learning_rate": 1.620086304987226e-05, + "loss": 0.9559, + "step": 17806 + }, + { + "epoch": 1.15, + "learning_rate": 1.620064836084931e-05, + "loss": 0.9398, + "step": 17807 + }, + { + "epoch": 1.15, + "learning_rate": 1.6200433671826363e-05, + "loss": 1.0313, + "step": 17808 + }, + { + "epoch": 1.15, + "learning_rate": 1.620021898280341e-05, + "loss": 1.0249, + "step": 17809 + }, + { + "epoch": 1.15, + "learning_rate": 1.620000429378046e-05, + "loss": 0.9547, + "step": 17810 + }, + { + "epoch": 1.15, + "learning_rate": 1.619978960475751e-05, + "loss": 0.9688, + "step": 17811 + }, + { + "epoch": 1.15, + "learning_rate": 1.619957491573456e-05, + "loss": 0.951, + "step": 17812 + }, + { + "epoch": 1.15, + "learning_rate": 1.6199360226711607e-05, + "loss": 1.0581, + "step": 17813 + }, + { + "epoch": 1.15, + "learning_rate": 1.619914553768866e-05, + "loss": 1.0022, + "step": 17814 + }, + { + "epoch": 1.15, + "learning_rate": 1.619893084866571e-05, + "loss": 0.9195, + "step": 17815 + }, + { + "epoch": 1.15, + "learning_rate": 1.6198716159642758e-05, + "loss": 0.9165, + "step": 17816 + }, + { + "epoch": 1.15, + "learning_rate": 1.619850147061981e-05, + "loss": 0.9893, + "step": 17817 + }, + { + "epoch": 1.15, + "learning_rate": 1.619828678159686e-05, + "loss": 0.9823, + "step": 17818 + }, + { + "epoch": 1.15, + "learning_rate": 1.6198072092573908e-05, + "loss": 0.9653, + "step": 17819 + }, + { + "epoch": 1.15, + "learning_rate": 1.6197857403550957e-05, + "loss": 0.9803, + "step": 17820 + }, + { + "epoch": 1.15, + "learning_rate": 1.619764271452801e-05, + "loss": 1.0159, + "step": 17821 + }, + { + "epoch": 1.15, + "learning_rate": 1.6197428025505058e-05, + "loss": 0.9853, + "step": 17822 + }, + { + "epoch": 1.15, + "learning_rate": 1.6197213336482107e-05, + "loss": 0.9816, + "step": 17823 + }, + { + "epoch": 1.15, + "learning_rate": 1.6196998647459156e-05, + "loss": 0.9484, + "step": 17824 + }, + { + "epoch": 1.15, + "learning_rate": 1.6196783958436205e-05, + "loss": 1.0056, + "step": 17825 + }, + { + "epoch": 1.15, + "learning_rate": 1.6196569269413258e-05, + "loss": 1.0236, + "step": 17826 + }, + { + "epoch": 1.15, + "learning_rate": 1.6196354580390306e-05, + "loss": 0.9777, + "step": 17827 + }, + { + "epoch": 1.15, + "learning_rate": 1.6196139891367355e-05, + "loss": 0.9616, + "step": 17828 + }, + { + "epoch": 1.15, + "learning_rate": 1.6195925202344404e-05, + "loss": 0.9569, + "step": 17829 + }, + { + "epoch": 1.15, + "learning_rate": 1.6195710513321453e-05, + "loss": 0.9727, + "step": 17830 + }, + { + "epoch": 1.15, + "learning_rate": 1.6195495824298502e-05, + "loss": 1.101, + "step": 17831 + }, + { + "epoch": 1.15, + "learning_rate": 1.6195281135275555e-05, + "loss": 0.9309, + "step": 17832 + }, + { + "epoch": 1.15, + "learning_rate": 1.6195066446252604e-05, + "loss": 1.0673, + "step": 17833 + }, + { + "epoch": 1.15, + "learning_rate": 1.6194851757229656e-05, + "loss": 0.9859, + "step": 17834 + }, + { + "epoch": 1.15, + "learning_rate": 1.6194637068206705e-05, + "loss": 1.0225, + "step": 17835 + }, + { + "epoch": 1.15, + "learning_rate": 1.6194422379183754e-05, + "loss": 0.9618, + "step": 17836 + }, + { + "epoch": 1.15, + "learning_rate": 1.6194207690160803e-05, + "loss": 0.96, + "step": 17837 + }, + { + "epoch": 1.15, + "learning_rate": 1.6193993001137852e-05, + "loss": 0.9478, + "step": 17838 + }, + { + "epoch": 1.15, + "learning_rate": 1.6193778312114904e-05, + "loss": 1.0267, + "step": 17839 + }, + { + "epoch": 1.15, + "learning_rate": 1.6193563623091953e-05, + "loss": 0.9727, + "step": 17840 + }, + { + "epoch": 1.15, + "learning_rate": 1.6193348934069002e-05, + "loss": 0.9266, + "step": 17841 + }, + { + "epoch": 1.15, + "learning_rate": 1.619313424504605e-05, + "loss": 0.9932, + "step": 17842 + }, + { + "epoch": 1.15, + "learning_rate": 1.61929195560231e-05, + "loss": 0.9895, + "step": 17843 + }, + { + "epoch": 1.15, + "learning_rate": 1.6192704867000152e-05, + "loss": 0.9986, + "step": 17844 + }, + { + "epoch": 1.15, + "learning_rate": 1.61924901779772e-05, + "loss": 1.0047, + "step": 17845 + }, + { + "epoch": 1.15, + "learning_rate": 1.6192275488954254e-05, + "loss": 1.0285, + "step": 17846 + }, + { + "epoch": 1.15, + "learning_rate": 1.6192060799931303e-05, + "loss": 1.054, + "step": 17847 + }, + { + "epoch": 1.15, + "learning_rate": 1.619184611090835e-05, + "loss": 0.9891, + "step": 17848 + }, + { + "epoch": 1.15, + "learning_rate": 1.61916314218854e-05, + "loss": 0.9989, + "step": 17849 + }, + { + "epoch": 1.15, + "learning_rate": 1.619141673286245e-05, + "loss": 0.9192, + "step": 17850 + }, + { + "epoch": 1.15, + "learning_rate": 1.61912020438395e-05, + "loss": 0.9891, + "step": 17851 + }, + { + "epoch": 1.15, + "learning_rate": 1.619098735481655e-05, + "loss": 0.9214, + "step": 17852 + }, + { + "epoch": 1.15, + "learning_rate": 1.61907726657936e-05, + "loss": 0.9915, + "step": 17853 + }, + { + "epoch": 1.15, + "learning_rate": 1.619055797677065e-05, + "loss": 0.9929, + "step": 17854 + }, + { + "epoch": 1.15, + "learning_rate": 1.6190343287747698e-05, + "loss": 0.929, + "step": 17855 + }, + { + "epoch": 1.15, + "learning_rate": 1.6190128598724747e-05, + "loss": 0.9997, + "step": 17856 + }, + { + "epoch": 1.15, + "learning_rate": 1.61899139097018e-05, + "loss": 0.9846, + "step": 17857 + }, + { + "epoch": 1.15, + "learning_rate": 1.6189699220678848e-05, + "loss": 0.9461, + "step": 17858 + }, + { + "epoch": 1.15, + "learning_rate": 1.6189484531655897e-05, + "loss": 1.0092, + "step": 17859 + }, + { + "epoch": 1.15, + "learning_rate": 1.6189269842632946e-05, + "loss": 1.0201, + "step": 17860 + }, + { + "epoch": 1.15, + "learning_rate": 1.618905515361e-05, + "loss": 0.9812, + "step": 17861 + }, + { + "epoch": 1.15, + "learning_rate": 1.6188840464587047e-05, + "loss": 0.8967, + "step": 17862 + }, + { + "epoch": 1.15, + "learning_rate": 1.6188625775564096e-05, + "loss": 0.9229, + "step": 17863 + }, + { + "epoch": 1.15, + "learning_rate": 1.618841108654115e-05, + "loss": 1.0372, + "step": 17864 + }, + { + "epoch": 1.15, + "learning_rate": 1.6188196397518198e-05, + "loss": 0.9337, + "step": 17865 + }, + { + "epoch": 1.15, + "learning_rate": 1.6187981708495247e-05, + "loss": 1.007, + "step": 17866 + }, + { + "epoch": 1.15, + "learning_rate": 1.6187767019472296e-05, + "loss": 1.0602, + "step": 17867 + }, + { + "epoch": 1.15, + "learning_rate": 1.6187552330449345e-05, + "loss": 1.0438, + "step": 17868 + }, + { + "epoch": 1.15, + "learning_rate": 1.6187337641426393e-05, + "loss": 0.9796, + "step": 17869 + }, + { + "epoch": 1.15, + "learning_rate": 1.6187122952403446e-05, + "loss": 1.0492, + "step": 17870 + }, + { + "epoch": 1.15, + "learning_rate": 1.6186908263380495e-05, + "loss": 0.9985, + "step": 17871 + }, + { + "epoch": 1.15, + "learning_rate": 1.6186693574357544e-05, + "loss": 1.0909, + "step": 17872 + }, + { + "epoch": 1.15, + "learning_rate": 1.6186478885334593e-05, + "loss": 1.0868, + "step": 17873 + }, + { + "epoch": 1.15, + "learning_rate": 1.6186264196311642e-05, + "loss": 0.9649, + "step": 17874 + }, + { + "epoch": 1.15, + "learning_rate": 1.6186049507288694e-05, + "loss": 1.0212, + "step": 17875 + }, + { + "epoch": 1.15, + "learning_rate": 1.6185834818265743e-05, + "loss": 0.8817, + "step": 17876 + }, + { + "epoch": 1.15, + "learning_rate": 1.6185620129242795e-05, + "loss": 0.9743, + "step": 17877 + }, + { + "epoch": 1.15, + "learning_rate": 1.6185405440219844e-05, + "loss": 0.9522, + "step": 17878 + }, + { + "epoch": 1.15, + "learning_rate": 1.6185190751196893e-05, + "loss": 0.9843, + "step": 17879 + }, + { + "epoch": 1.15, + "learning_rate": 1.6184976062173942e-05, + "loss": 1.0084, + "step": 17880 + }, + { + "epoch": 1.15, + "learning_rate": 1.618476137315099e-05, + "loss": 1.003, + "step": 17881 + }, + { + "epoch": 1.15, + "learning_rate": 1.618454668412804e-05, + "loss": 0.99, + "step": 17882 + }, + { + "epoch": 1.15, + "learning_rate": 1.6184331995105093e-05, + "loss": 0.9116, + "step": 17883 + }, + { + "epoch": 1.15, + "learning_rate": 1.618411730608214e-05, + "loss": 1.0887, + "step": 17884 + }, + { + "epoch": 1.15, + "learning_rate": 1.618390261705919e-05, + "loss": 0.9978, + "step": 17885 + }, + { + "epoch": 1.15, + "learning_rate": 1.618368792803624e-05, + "loss": 0.9596, + "step": 17886 + }, + { + "epoch": 1.15, + "learning_rate": 1.618347323901329e-05, + "loss": 0.9857, + "step": 17887 + }, + { + "epoch": 1.15, + "learning_rate": 1.618325854999034e-05, + "loss": 0.9609, + "step": 17888 + }, + { + "epoch": 1.15, + "learning_rate": 1.618304386096739e-05, + "loss": 1.0008, + "step": 17889 + }, + { + "epoch": 1.15, + "learning_rate": 1.6182829171944442e-05, + "loss": 0.9864, + "step": 17890 + }, + { + "epoch": 1.15, + "learning_rate": 1.618261448292149e-05, + "loss": 0.9689, + "step": 17891 + }, + { + "epoch": 1.15, + "learning_rate": 1.618239979389854e-05, + "loss": 0.9555, + "step": 17892 + }, + { + "epoch": 1.15, + "learning_rate": 1.618218510487559e-05, + "loss": 0.9545, + "step": 17893 + }, + { + "epoch": 1.15, + "learning_rate": 1.6181970415852638e-05, + "loss": 0.9762, + "step": 17894 + }, + { + "epoch": 1.15, + "learning_rate": 1.618175572682969e-05, + "loss": 0.9653, + "step": 17895 + }, + { + "epoch": 1.15, + "learning_rate": 1.618154103780674e-05, + "loss": 0.9052, + "step": 17896 + }, + { + "epoch": 1.15, + "learning_rate": 1.6181326348783788e-05, + "loss": 0.9915, + "step": 17897 + }, + { + "epoch": 1.15, + "learning_rate": 1.6181111659760837e-05, + "loss": 0.9736, + "step": 17898 + }, + { + "epoch": 1.15, + "learning_rate": 1.6180896970737886e-05, + "loss": 0.9285, + "step": 17899 + }, + { + "epoch": 1.15, + "learning_rate": 1.6180682281714935e-05, + "loss": 0.9358, + "step": 17900 + }, + { + "epoch": 1.15, + "learning_rate": 1.6180467592691987e-05, + "loss": 0.9849, + "step": 17901 + }, + { + "epoch": 1.15, + "learning_rate": 1.6180252903669036e-05, + "loss": 0.9733, + "step": 17902 + }, + { + "epoch": 1.15, + "learning_rate": 1.6180038214646085e-05, + "loss": 0.9952, + "step": 17903 + }, + { + "epoch": 1.15, + "learning_rate": 1.6179823525623138e-05, + "loss": 0.9763, + "step": 17904 + }, + { + "epoch": 1.15, + "learning_rate": 1.6179608836600187e-05, + "loss": 1.0479, + "step": 17905 + }, + { + "epoch": 1.15, + "learning_rate": 1.6179394147577236e-05, + "loss": 0.9387, + "step": 17906 + }, + { + "epoch": 1.15, + "learning_rate": 1.6179179458554285e-05, + "loss": 0.9846, + "step": 17907 + }, + { + "epoch": 1.15, + "learning_rate": 1.6178964769531337e-05, + "loss": 1.0005, + "step": 17908 + }, + { + "epoch": 1.15, + "learning_rate": 1.6178750080508386e-05, + "loss": 1.0316, + "step": 17909 + }, + { + "epoch": 1.15, + "learning_rate": 1.6178535391485435e-05, + "loss": 1.0021, + "step": 17910 + }, + { + "epoch": 1.15, + "learning_rate": 1.6178320702462484e-05, + "loss": 0.963, + "step": 17911 + }, + { + "epoch": 1.15, + "learning_rate": 1.6178106013439533e-05, + "loss": 0.9933, + "step": 17912 + }, + { + "epoch": 1.15, + "learning_rate": 1.6177891324416585e-05, + "loss": 0.9871, + "step": 17913 + }, + { + "epoch": 1.15, + "learning_rate": 1.6177676635393634e-05, + "loss": 0.9462, + "step": 17914 + }, + { + "epoch": 1.15, + "learning_rate": 1.6177461946370683e-05, + "loss": 0.9747, + "step": 17915 + }, + { + "epoch": 1.15, + "learning_rate": 1.6177247257347732e-05, + "loss": 0.9529, + "step": 17916 + }, + { + "epoch": 1.15, + "learning_rate": 1.617703256832478e-05, + "loss": 0.9398, + "step": 17917 + }, + { + "epoch": 1.15, + "learning_rate": 1.617681787930183e-05, + "loss": 0.9982, + "step": 17918 + }, + { + "epoch": 1.15, + "learning_rate": 1.6176603190278882e-05, + "loss": 0.9526, + "step": 17919 + }, + { + "epoch": 1.15, + "learning_rate": 1.617638850125593e-05, + "loss": 0.9801, + "step": 17920 + }, + { + "epoch": 1.15, + "learning_rate": 1.6176173812232984e-05, + "loss": 1.0298, + "step": 17921 + }, + { + "epoch": 1.15, + "learning_rate": 1.6175959123210033e-05, + "loss": 1.0573, + "step": 17922 + }, + { + "epoch": 1.15, + "learning_rate": 1.617574443418708e-05, + "loss": 0.9548, + "step": 17923 + }, + { + "epoch": 1.15, + "learning_rate": 1.617552974516413e-05, + "loss": 0.9985, + "step": 17924 + }, + { + "epoch": 1.15, + "learning_rate": 1.617531505614118e-05, + "loss": 0.9556, + "step": 17925 + }, + { + "epoch": 1.15, + "learning_rate": 1.6175100367118232e-05, + "loss": 0.9331, + "step": 17926 + }, + { + "epoch": 1.15, + "learning_rate": 1.617488567809528e-05, + "loss": 0.9361, + "step": 17927 + }, + { + "epoch": 1.15, + "learning_rate": 1.617467098907233e-05, + "loss": 1.0126, + "step": 17928 + }, + { + "epoch": 1.15, + "learning_rate": 1.617445630004938e-05, + "loss": 0.935, + "step": 17929 + }, + { + "epoch": 1.15, + "learning_rate": 1.6174241611026428e-05, + "loss": 0.9669, + "step": 17930 + }, + { + "epoch": 1.15, + "learning_rate": 1.617402692200348e-05, + "loss": 0.9442, + "step": 17931 + }, + { + "epoch": 1.15, + "learning_rate": 1.617381223298053e-05, + "loss": 1.0044, + "step": 17932 + }, + { + "epoch": 1.15, + "learning_rate": 1.617359754395758e-05, + "loss": 1.0534, + "step": 17933 + }, + { + "epoch": 1.15, + "learning_rate": 1.617338285493463e-05, + "loss": 0.9507, + "step": 17934 + }, + { + "epoch": 1.15, + "learning_rate": 1.617316816591168e-05, + "loss": 1.0154, + "step": 17935 + }, + { + "epoch": 1.15, + "learning_rate": 1.617295347688873e-05, + "loss": 1.0193, + "step": 17936 + }, + { + "epoch": 1.15, + "learning_rate": 1.6172738787865777e-05, + "loss": 1.0111, + "step": 17937 + }, + { + "epoch": 1.15, + "learning_rate": 1.6172524098842826e-05, + "loss": 0.9623, + "step": 17938 + }, + { + "epoch": 1.15, + "learning_rate": 1.617230940981988e-05, + "loss": 1.0346, + "step": 17939 + }, + { + "epoch": 1.15, + "learning_rate": 1.6172094720796928e-05, + "loss": 0.9571, + "step": 17940 + }, + { + "epoch": 1.15, + "learning_rate": 1.6171880031773977e-05, + "loss": 0.9613, + "step": 17941 + }, + { + "epoch": 1.15, + "learning_rate": 1.6171665342751026e-05, + "loss": 0.9729, + "step": 17942 + }, + { + "epoch": 1.15, + "learning_rate": 1.6171450653728074e-05, + "loss": 0.9396, + "step": 17943 + }, + { + "epoch": 1.15, + "learning_rate": 1.6171235964705127e-05, + "loss": 0.9079, + "step": 17944 + }, + { + "epoch": 1.15, + "learning_rate": 1.6171021275682176e-05, + "loss": 0.9617, + "step": 17945 + }, + { + "epoch": 1.15, + "learning_rate": 1.6170806586659225e-05, + "loss": 1.0094, + "step": 17946 + }, + { + "epoch": 1.15, + "learning_rate": 1.6170591897636274e-05, + "loss": 0.9809, + "step": 17947 + }, + { + "epoch": 1.15, + "learning_rate": 1.6170377208613326e-05, + "loss": 1.0352, + "step": 17948 + }, + { + "epoch": 1.15, + "learning_rate": 1.6170162519590375e-05, + "loss": 0.8911, + "step": 17949 + }, + { + "epoch": 1.15, + "learning_rate": 1.6169947830567424e-05, + "loss": 0.9619, + "step": 17950 + }, + { + "epoch": 1.15, + "learning_rate": 1.6169733141544476e-05, + "loss": 0.9729, + "step": 17951 + }, + { + "epoch": 1.15, + "learning_rate": 1.6169518452521525e-05, + "loss": 0.9897, + "step": 17952 + }, + { + "epoch": 1.16, + "learning_rate": 1.6169303763498574e-05, + "loss": 0.9549, + "step": 17953 + }, + { + "epoch": 1.16, + "learning_rate": 1.6169089074475623e-05, + "loss": 1.0057, + "step": 17954 + }, + { + "epoch": 1.16, + "learning_rate": 1.6168874385452672e-05, + "loss": 0.836, + "step": 17955 + }, + { + "epoch": 1.16, + "learning_rate": 1.616865969642972e-05, + "loss": 0.9617, + "step": 17956 + }, + { + "epoch": 1.16, + "learning_rate": 1.6168445007406774e-05, + "loss": 0.9411, + "step": 17957 + }, + { + "epoch": 1.16, + "learning_rate": 1.6168230318383823e-05, + "loss": 0.924, + "step": 17958 + }, + { + "epoch": 1.16, + "learning_rate": 1.616801562936087e-05, + "loss": 1.0473, + "step": 17959 + }, + { + "epoch": 1.16, + "learning_rate": 1.616780094033792e-05, + "loss": 0.9759, + "step": 17960 + }, + { + "epoch": 1.16, + "learning_rate": 1.616758625131497e-05, + "loss": 0.8853, + "step": 17961 + }, + { + "epoch": 1.16, + "learning_rate": 1.6167371562292022e-05, + "loss": 0.9636, + "step": 17962 + }, + { + "epoch": 1.16, + "learning_rate": 1.616715687326907e-05, + "loss": 1.0178, + "step": 17963 + }, + { + "epoch": 1.16, + "learning_rate": 1.6166942184246123e-05, + "loss": 1.0583, + "step": 17964 + }, + { + "epoch": 1.16, + "learning_rate": 1.6166727495223172e-05, + "loss": 1.0285, + "step": 17965 + }, + { + "epoch": 1.16, + "learning_rate": 1.616651280620022e-05, + "loss": 0.916, + "step": 17966 + }, + { + "epoch": 1.16, + "learning_rate": 1.616629811717727e-05, + "loss": 0.9032, + "step": 17967 + }, + { + "epoch": 1.16, + "learning_rate": 1.616608342815432e-05, + "loss": 0.9421, + "step": 17968 + }, + { + "epoch": 1.16, + "learning_rate": 1.6165868739131368e-05, + "loss": 1.0058, + "step": 17969 + }, + { + "epoch": 1.16, + "learning_rate": 1.616565405010842e-05, + "loss": 0.999, + "step": 17970 + }, + { + "epoch": 1.16, + "learning_rate": 1.616543936108547e-05, + "loss": 0.975, + "step": 17971 + }, + { + "epoch": 1.16, + "learning_rate": 1.6165224672062518e-05, + "loss": 1.0495, + "step": 17972 + }, + { + "epoch": 1.16, + "learning_rate": 1.6165009983039567e-05, + "loss": 1.0386, + "step": 17973 + }, + { + "epoch": 1.16, + "learning_rate": 1.6164795294016616e-05, + "loss": 0.9984, + "step": 17974 + }, + { + "epoch": 1.16, + "learning_rate": 1.616458060499367e-05, + "loss": 1.045, + "step": 17975 + }, + { + "epoch": 1.16, + "learning_rate": 1.6164365915970717e-05, + "loss": 0.933, + "step": 17976 + }, + { + "epoch": 1.16, + "learning_rate": 1.616415122694777e-05, + "loss": 1.003, + "step": 17977 + }, + { + "epoch": 1.16, + "learning_rate": 1.616393653792482e-05, + "loss": 0.9424, + "step": 17978 + }, + { + "epoch": 1.16, + "learning_rate": 1.6163721848901868e-05, + "loss": 0.9403, + "step": 17979 + }, + { + "epoch": 1.16, + "learning_rate": 1.6163507159878917e-05, + "loss": 0.9268, + "step": 17980 + }, + { + "epoch": 1.16, + "learning_rate": 1.6163292470855966e-05, + "loss": 0.9852, + "step": 17981 + }, + { + "epoch": 1.16, + "learning_rate": 1.6163077781833018e-05, + "loss": 0.9373, + "step": 17982 + }, + { + "epoch": 1.16, + "learning_rate": 1.6162863092810067e-05, + "loss": 1.0436, + "step": 17983 + }, + { + "epoch": 1.16, + "learning_rate": 1.6162648403787116e-05, + "loss": 0.9429, + "step": 17984 + }, + { + "epoch": 1.16, + "learning_rate": 1.6162433714764165e-05, + "loss": 0.9685, + "step": 17985 + }, + { + "epoch": 1.16, + "learning_rate": 1.6162219025741214e-05, + "loss": 0.9557, + "step": 17986 + }, + { + "epoch": 1.16, + "learning_rate": 1.6162004336718263e-05, + "loss": 0.8983, + "step": 17987 + }, + { + "epoch": 1.16, + "learning_rate": 1.6161789647695315e-05, + "loss": 0.9811, + "step": 17988 + }, + { + "epoch": 1.16, + "learning_rate": 1.6161574958672364e-05, + "loss": 0.925, + "step": 17989 + }, + { + "epoch": 1.16, + "learning_rate": 1.6161360269649413e-05, + "loss": 0.9831, + "step": 17990 + }, + { + "epoch": 1.16, + "learning_rate": 1.6161145580626465e-05, + "loss": 0.9925, + "step": 17991 + }, + { + "epoch": 1.16, + "learning_rate": 1.6160930891603514e-05, + "loss": 0.991, + "step": 17992 + }, + { + "epoch": 1.16, + "learning_rate": 1.6160716202580563e-05, + "loss": 0.947, + "step": 17993 + }, + { + "epoch": 1.16, + "learning_rate": 1.6160501513557612e-05, + "loss": 1.0342, + "step": 17994 + }, + { + "epoch": 1.16, + "learning_rate": 1.6160286824534665e-05, + "loss": 0.9747, + "step": 17995 + }, + { + "epoch": 1.16, + "learning_rate": 1.6160072135511714e-05, + "loss": 0.9229, + "step": 17996 + }, + { + "epoch": 1.16, + "learning_rate": 1.6159857446488763e-05, + "loss": 0.9204, + "step": 17997 + }, + { + "epoch": 1.16, + "learning_rate": 1.615964275746581e-05, + "loss": 1.0419, + "step": 17998 + }, + { + "epoch": 1.16, + "learning_rate": 1.615942806844286e-05, + "loss": 0.9608, + "step": 17999 + }, + { + "epoch": 1.16, + "learning_rate": 1.6159213379419913e-05, + "loss": 0.9455, + "step": 18000 + }, + { + "epoch": 1.16, + "eval_laion/cider": 0.5534081179798607, + "eval_laion/meteor": 0.39544480057397735, + "eval_laion/pred_num_words": 36.931640625, + "eval_laion/sacrebleu": 16.358764252048072, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.062328815460205, + "eval_runtime": 95.9043, + "eval_samples_per_second": 5.203, + "eval_steps_per_second": 0.334, + "step": 18000 + }, + { + "epoch": 1.16, + "learning_rate": 1.6158998690396962e-05, + "loss": 0.9527, + "step": 18001 + }, + { + "epoch": 1.16, + "learning_rate": 1.615878400137401e-05, + "loss": 0.9779, + "step": 18002 + }, + { + "epoch": 1.16, + "learning_rate": 1.615856931235106e-05, + "loss": 1.0639, + "step": 18003 + }, + { + "epoch": 1.16, + "learning_rate": 1.615835462332811e-05, + "loss": 0.9997, + "step": 18004 + }, + { + "epoch": 1.16, + "learning_rate": 1.6158139934305158e-05, + "loss": 0.9232, + "step": 18005 + }, + { + "epoch": 1.16, + "learning_rate": 1.615792524528221e-05, + "loss": 1.0191, + "step": 18006 + }, + { + "epoch": 1.16, + "learning_rate": 1.615771055625926e-05, + "loss": 0.9825, + "step": 18007 + }, + { + "epoch": 1.16, + "learning_rate": 1.615749586723631e-05, + "loss": 1.0268, + "step": 18008 + }, + { + "epoch": 1.16, + "learning_rate": 1.615728117821336e-05, + "loss": 0.9644, + "step": 18009 + }, + { + "epoch": 1.16, + "learning_rate": 1.615706648919041e-05, + "loss": 0.9889, + "step": 18010 + }, + { + "epoch": 1.16, + "learning_rate": 1.615685180016746e-05, + "loss": 1.0312, + "step": 18011 + }, + { + "epoch": 1.16, + "learning_rate": 1.6156637111144507e-05, + "loss": 0.9838, + "step": 18012 + }, + { + "epoch": 1.16, + "learning_rate": 1.615642242212156e-05, + "loss": 1.049, + "step": 18013 + }, + { + "epoch": 1.16, + "learning_rate": 1.615620773309861e-05, + "loss": 0.9354, + "step": 18014 + }, + { + "epoch": 1.16, + "learning_rate": 1.6155993044075658e-05, + "loss": 1.0258, + "step": 18015 + }, + { + "epoch": 1.16, + "learning_rate": 1.6155778355052707e-05, + "loss": 0.9117, + "step": 18016 + }, + { + "epoch": 1.16, + "learning_rate": 1.6155563666029755e-05, + "loss": 1.0744, + "step": 18017 + }, + { + "epoch": 1.16, + "learning_rate": 1.6155348977006804e-05, + "loss": 1.0299, + "step": 18018 + }, + { + "epoch": 1.16, + "learning_rate": 1.6155134287983857e-05, + "loss": 0.9791, + "step": 18019 + }, + { + "epoch": 1.16, + "learning_rate": 1.615491959896091e-05, + "loss": 1.0196, + "step": 18020 + }, + { + "epoch": 1.16, + "learning_rate": 1.6154704909937958e-05, + "loss": 0.9717, + "step": 18021 + }, + { + "epoch": 1.16, + "learning_rate": 1.6154490220915007e-05, + "loss": 0.9774, + "step": 18022 + }, + { + "epoch": 1.16, + "learning_rate": 1.6154275531892056e-05, + "loss": 0.9495, + "step": 18023 + }, + { + "epoch": 1.16, + "learning_rate": 1.6154060842869105e-05, + "loss": 1.0498, + "step": 18024 + }, + { + "epoch": 1.16, + "learning_rate": 1.6153846153846154e-05, + "loss": 1.0201, + "step": 18025 + }, + { + "epoch": 1.16, + "learning_rate": 1.6153631464823206e-05, + "loss": 1.0162, + "step": 18026 + }, + { + "epoch": 1.16, + "learning_rate": 1.6153416775800255e-05, + "loss": 1.0409, + "step": 18027 + }, + { + "epoch": 1.16, + "learning_rate": 1.6153202086777304e-05, + "loss": 0.9779, + "step": 18028 + }, + { + "epoch": 1.16, + "learning_rate": 1.6152987397754353e-05, + "loss": 0.9944, + "step": 18029 + }, + { + "epoch": 1.16, + "learning_rate": 1.6152772708731402e-05, + "loss": 1.031, + "step": 18030 + }, + { + "epoch": 1.16, + "learning_rate": 1.6152558019708455e-05, + "loss": 0.9747, + "step": 18031 + }, + { + "epoch": 1.16, + "learning_rate": 1.6152343330685504e-05, + "loss": 1.0646, + "step": 18032 + }, + { + "epoch": 1.16, + "learning_rate": 1.6152128641662552e-05, + "loss": 0.9542, + "step": 18033 + }, + { + "epoch": 1.16, + "learning_rate": 1.61519139526396e-05, + "loss": 0.8614, + "step": 18034 + }, + { + "epoch": 1.16, + "learning_rate": 1.6151699263616654e-05, + "loss": 1.0328, + "step": 18035 + }, + { + "epoch": 1.16, + "learning_rate": 1.6151484574593703e-05, + "loss": 0.9119, + "step": 18036 + }, + { + "epoch": 1.16, + "learning_rate": 1.6151269885570752e-05, + "loss": 0.9459, + "step": 18037 + }, + { + "epoch": 1.16, + "learning_rate": 1.6151055196547804e-05, + "loss": 0.9972, + "step": 18038 + }, + { + "epoch": 1.16, + "learning_rate": 1.6150840507524853e-05, + "loss": 1.0252, + "step": 18039 + }, + { + "epoch": 1.16, + "learning_rate": 1.6150625818501902e-05, + "loss": 0.9372, + "step": 18040 + }, + { + "epoch": 1.16, + "learning_rate": 1.615041112947895e-05, + "loss": 1.0077, + "step": 18041 + }, + { + "epoch": 1.16, + "learning_rate": 1.6150196440456e-05, + "loss": 0.9575, + "step": 18042 + }, + { + "epoch": 1.16, + "learning_rate": 1.614998175143305e-05, + "loss": 1.0127, + "step": 18043 + }, + { + "epoch": 1.16, + "learning_rate": 1.61497670624101e-05, + "loss": 0.9542, + "step": 18044 + }, + { + "epoch": 1.16, + "learning_rate": 1.614955237338715e-05, + "loss": 1.0424, + "step": 18045 + }, + { + "epoch": 1.16, + "learning_rate": 1.61493376843642e-05, + "loss": 0.9344, + "step": 18046 + }, + { + "epoch": 1.16, + "learning_rate": 1.6149122995341248e-05, + "loss": 0.9692, + "step": 18047 + }, + { + "epoch": 1.16, + "learning_rate": 1.6148908306318297e-05, + "loss": 1.0113, + "step": 18048 + }, + { + "epoch": 1.16, + "learning_rate": 1.614869361729535e-05, + "loss": 0.9532, + "step": 18049 + }, + { + "epoch": 1.16, + "learning_rate": 1.61484789282724e-05, + "loss": 0.9436, + "step": 18050 + }, + { + "epoch": 1.16, + "learning_rate": 1.614826423924945e-05, + "loss": 0.9798, + "step": 18051 + }, + { + "epoch": 1.16, + "learning_rate": 1.61480495502265e-05, + "loss": 0.9928, + "step": 18052 + }, + { + "epoch": 1.16, + "learning_rate": 1.614783486120355e-05, + "loss": 0.9449, + "step": 18053 + }, + { + "epoch": 1.16, + "learning_rate": 1.6147620172180598e-05, + "loss": 1.0576, + "step": 18054 + }, + { + "epoch": 1.16, + "learning_rate": 1.6147405483157647e-05, + "loss": 1.0213, + "step": 18055 + }, + { + "epoch": 1.16, + "learning_rate": 1.6147190794134696e-05, + "loss": 0.9457, + "step": 18056 + }, + { + "epoch": 1.16, + "learning_rate": 1.6146976105111748e-05, + "loss": 1.0646, + "step": 18057 + }, + { + "epoch": 1.16, + "learning_rate": 1.6146761416088797e-05, + "loss": 1.032, + "step": 18058 + }, + { + "epoch": 1.16, + "learning_rate": 1.6146546727065846e-05, + "loss": 0.952, + "step": 18059 + }, + { + "epoch": 1.16, + "learning_rate": 1.6146332038042895e-05, + "loss": 1.0255, + "step": 18060 + }, + { + "epoch": 1.16, + "learning_rate": 1.6146117349019944e-05, + "loss": 0.9469, + "step": 18061 + }, + { + "epoch": 1.16, + "learning_rate": 1.6145902659996996e-05, + "loss": 0.9627, + "step": 18062 + }, + { + "epoch": 1.16, + "learning_rate": 1.6145687970974045e-05, + "loss": 0.9697, + "step": 18063 + }, + { + "epoch": 1.16, + "learning_rate": 1.6145473281951098e-05, + "loss": 1.0148, + "step": 18064 + }, + { + "epoch": 1.16, + "learning_rate": 1.6145258592928146e-05, + "loss": 0.9803, + "step": 18065 + }, + { + "epoch": 1.16, + "learning_rate": 1.6145043903905195e-05, + "loss": 0.9745, + "step": 18066 + }, + { + "epoch": 1.16, + "learning_rate": 1.6144829214882244e-05, + "loss": 0.9798, + "step": 18067 + }, + { + "epoch": 1.16, + "learning_rate": 1.6144614525859293e-05, + "loss": 0.9235, + "step": 18068 + }, + { + "epoch": 1.16, + "learning_rate": 1.6144399836836346e-05, + "loss": 0.9612, + "step": 18069 + }, + { + "epoch": 1.16, + "learning_rate": 1.6144185147813395e-05, + "loss": 0.9898, + "step": 18070 + }, + { + "epoch": 1.16, + "learning_rate": 1.6143970458790444e-05, + "loss": 1.0152, + "step": 18071 + }, + { + "epoch": 1.16, + "learning_rate": 1.6143755769767493e-05, + "loss": 1.0409, + "step": 18072 + }, + { + "epoch": 1.16, + "learning_rate": 1.614354108074454e-05, + "loss": 1.0118, + "step": 18073 + }, + { + "epoch": 1.16, + "learning_rate": 1.614332639172159e-05, + "loss": 0.9872, + "step": 18074 + }, + { + "epoch": 1.16, + "learning_rate": 1.6143111702698643e-05, + "loss": 1.0271, + "step": 18075 + }, + { + "epoch": 1.16, + "learning_rate": 1.6142897013675692e-05, + "loss": 0.9933, + "step": 18076 + }, + { + "epoch": 1.16, + "learning_rate": 1.614268232465274e-05, + "loss": 0.9696, + "step": 18077 + }, + { + "epoch": 1.16, + "learning_rate": 1.614246763562979e-05, + "loss": 1.0179, + "step": 18078 + }, + { + "epoch": 1.16, + "learning_rate": 1.6142252946606842e-05, + "loss": 0.9883, + "step": 18079 + }, + { + "epoch": 1.16, + "learning_rate": 1.614203825758389e-05, + "loss": 0.9993, + "step": 18080 + }, + { + "epoch": 1.16, + "learning_rate": 1.614182356856094e-05, + "loss": 0.9972, + "step": 18081 + }, + { + "epoch": 1.16, + "learning_rate": 1.6141608879537992e-05, + "loss": 0.9291, + "step": 18082 + }, + { + "epoch": 1.16, + "learning_rate": 1.614139419051504e-05, + "loss": 1.0104, + "step": 18083 + }, + { + "epoch": 1.16, + "learning_rate": 1.614117950149209e-05, + "loss": 0.9531, + "step": 18084 + }, + { + "epoch": 1.16, + "learning_rate": 1.614096481246914e-05, + "loss": 0.9606, + "step": 18085 + }, + { + "epoch": 1.16, + "learning_rate": 1.6140750123446188e-05, + "loss": 0.9864, + "step": 18086 + }, + { + "epoch": 1.16, + "learning_rate": 1.614053543442324e-05, + "loss": 0.9963, + "step": 18087 + }, + { + "epoch": 1.16, + "learning_rate": 1.614032074540029e-05, + "loss": 0.9866, + "step": 18088 + }, + { + "epoch": 1.16, + "learning_rate": 1.614010605637734e-05, + "loss": 0.9265, + "step": 18089 + }, + { + "epoch": 1.16, + "learning_rate": 1.6139891367354388e-05, + "loss": 1.0177, + "step": 18090 + }, + { + "epoch": 1.16, + "learning_rate": 1.6139676678331437e-05, + "loss": 0.9146, + "step": 18091 + }, + { + "epoch": 1.16, + "learning_rate": 1.6139461989308485e-05, + "loss": 1.0071, + "step": 18092 + }, + { + "epoch": 1.16, + "learning_rate": 1.6139247300285538e-05, + "loss": 0.9747, + "step": 18093 + }, + { + "epoch": 1.16, + "learning_rate": 1.6139032611262587e-05, + "loss": 0.9372, + "step": 18094 + }, + { + "epoch": 1.16, + "learning_rate": 1.613881792223964e-05, + "loss": 1.0765, + "step": 18095 + }, + { + "epoch": 1.16, + "learning_rate": 1.6138603233216688e-05, + "loss": 0.96, + "step": 18096 + }, + { + "epoch": 1.16, + "learning_rate": 1.6138388544193737e-05, + "loss": 0.9871, + "step": 18097 + }, + { + "epoch": 1.16, + "learning_rate": 1.6138173855170786e-05, + "loss": 0.9429, + "step": 18098 + }, + { + "epoch": 1.16, + "learning_rate": 1.6137959166147835e-05, + "loss": 1.0347, + "step": 18099 + }, + { + "epoch": 1.16, + "learning_rate": 1.6137744477124887e-05, + "loss": 0.8951, + "step": 18100 + }, + { + "epoch": 1.16, + "learning_rate": 1.6137529788101936e-05, + "loss": 0.958, + "step": 18101 + }, + { + "epoch": 1.16, + "learning_rate": 1.6137315099078985e-05, + "loss": 0.9524, + "step": 18102 + }, + { + "epoch": 1.16, + "learning_rate": 1.6137100410056034e-05, + "loss": 1.011, + "step": 18103 + }, + { + "epoch": 1.16, + "learning_rate": 1.6136885721033083e-05, + "loss": 1.0074, + "step": 18104 + }, + { + "epoch": 1.16, + "learning_rate": 1.6136671032010132e-05, + "loss": 1.0224, + "step": 18105 + }, + { + "epoch": 1.16, + "learning_rate": 1.6136456342987185e-05, + "loss": 0.9588, + "step": 18106 + }, + { + "epoch": 1.16, + "learning_rate": 1.6136241653964233e-05, + "loss": 1.0061, + "step": 18107 + }, + { + "epoch": 1.17, + "learning_rate": 1.6136026964941286e-05, + "loss": 1.0172, + "step": 18108 + }, + { + "epoch": 1.17, + "learning_rate": 1.6135812275918335e-05, + "loss": 0.9911, + "step": 18109 + }, + { + "epoch": 1.17, + "learning_rate": 1.6135597586895384e-05, + "loss": 0.9725, + "step": 18110 + }, + { + "epoch": 1.17, + "learning_rate": 1.6135382897872433e-05, + "loss": 1.0448, + "step": 18111 + }, + { + "epoch": 1.17, + "learning_rate": 1.6135168208849482e-05, + "loss": 1.0645, + "step": 18112 + }, + { + "epoch": 1.17, + "learning_rate": 1.6134953519826534e-05, + "loss": 0.9938, + "step": 18113 + }, + { + "epoch": 1.17, + "learning_rate": 1.6134738830803583e-05, + "loss": 0.9263, + "step": 18114 + }, + { + "epoch": 1.17, + "learning_rate": 1.6134524141780632e-05, + "loss": 0.9622, + "step": 18115 + }, + { + "epoch": 1.17, + "learning_rate": 1.613430945275768e-05, + "loss": 0.9824, + "step": 18116 + }, + { + "epoch": 1.17, + "learning_rate": 1.613409476373473e-05, + "loss": 1.0392, + "step": 18117 + }, + { + "epoch": 1.17, + "learning_rate": 1.6133880074711782e-05, + "loss": 0.9563, + "step": 18118 + }, + { + "epoch": 1.17, + "learning_rate": 1.613366538568883e-05, + "loss": 1.0394, + "step": 18119 + }, + { + "epoch": 1.17, + "learning_rate": 1.613345069666588e-05, + "loss": 0.8851, + "step": 18120 + }, + { + "epoch": 1.17, + "learning_rate": 1.613323600764293e-05, + "loss": 0.9786, + "step": 18121 + }, + { + "epoch": 1.17, + "learning_rate": 1.613302131861998e-05, + "loss": 0.9793, + "step": 18122 + }, + { + "epoch": 1.17, + "learning_rate": 1.613280662959703e-05, + "loss": 0.9709, + "step": 18123 + }, + { + "epoch": 1.17, + "learning_rate": 1.613259194057408e-05, + "loss": 0.9764, + "step": 18124 + }, + { + "epoch": 1.17, + "learning_rate": 1.613237725155113e-05, + "loss": 0.9331, + "step": 18125 + }, + { + "epoch": 1.17, + "learning_rate": 1.613216256252818e-05, + "loss": 1.0263, + "step": 18126 + }, + { + "epoch": 1.17, + "learning_rate": 1.613194787350523e-05, + "loss": 0.9718, + "step": 18127 + }, + { + "epoch": 1.17, + "learning_rate": 1.613173318448228e-05, + "loss": 1.0103, + "step": 18128 + }, + { + "epoch": 1.17, + "learning_rate": 1.6131518495459328e-05, + "loss": 1.018, + "step": 18129 + }, + { + "epoch": 1.17, + "learning_rate": 1.6131303806436377e-05, + "loss": 0.9744, + "step": 18130 + }, + { + "epoch": 1.17, + "learning_rate": 1.613108911741343e-05, + "loss": 0.9855, + "step": 18131 + }, + { + "epoch": 1.17, + "learning_rate": 1.6130874428390478e-05, + "loss": 0.9842, + "step": 18132 + }, + { + "epoch": 1.17, + "learning_rate": 1.6130659739367527e-05, + "loss": 0.9367, + "step": 18133 + }, + { + "epoch": 1.17, + "learning_rate": 1.6130445050344576e-05, + "loss": 0.9462, + "step": 18134 + }, + { + "epoch": 1.17, + "learning_rate": 1.6130230361321625e-05, + "loss": 0.9358, + "step": 18135 + }, + { + "epoch": 1.17, + "learning_rate": 1.6130015672298677e-05, + "loss": 0.9831, + "step": 18136 + }, + { + "epoch": 1.17, + "learning_rate": 1.6129800983275726e-05, + "loss": 0.957, + "step": 18137 + }, + { + "epoch": 1.17, + "learning_rate": 1.612958629425278e-05, + "loss": 1.0177, + "step": 18138 + }, + { + "epoch": 1.17, + "learning_rate": 1.6129371605229827e-05, + "loss": 0.951, + "step": 18139 + }, + { + "epoch": 1.17, + "learning_rate": 1.6129156916206876e-05, + "loss": 1.0399, + "step": 18140 + }, + { + "epoch": 1.17, + "learning_rate": 1.6128942227183925e-05, + "loss": 0.9967, + "step": 18141 + }, + { + "epoch": 1.17, + "learning_rate": 1.6128727538160974e-05, + "loss": 0.9441, + "step": 18142 + }, + { + "epoch": 1.17, + "learning_rate": 1.6128512849138023e-05, + "loss": 0.8952, + "step": 18143 + }, + { + "epoch": 1.17, + "learning_rate": 1.6128298160115076e-05, + "loss": 0.9699, + "step": 18144 + }, + { + "epoch": 1.17, + "learning_rate": 1.6128083471092125e-05, + "loss": 0.9418, + "step": 18145 + }, + { + "epoch": 1.17, + "learning_rate": 1.6127868782069174e-05, + "loss": 0.9978, + "step": 18146 + }, + { + "epoch": 1.17, + "learning_rate": 1.6127654093046223e-05, + "loss": 0.9666, + "step": 18147 + }, + { + "epoch": 1.17, + "learning_rate": 1.612743940402327e-05, + "loss": 0.9572, + "step": 18148 + }, + { + "epoch": 1.17, + "learning_rate": 1.6127224715000324e-05, + "loss": 0.981, + "step": 18149 + }, + { + "epoch": 1.17, + "learning_rate": 1.6127010025977373e-05, + "loss": 1.0247, + "step": 18150 + }, + { + "epoch": 1.17, + "learning_rate": 1.6126795336954425e-05, + "loss": 1.0438, + "step": 18151 + }, + { + "epoch": 1.17, + "learning_rate": 1.6126580647931474e-05, + "loss": 1.013, + "step": 18152 + }, + { + "epoch": 1.17, + "learning_rate": 1.6126365958908523e-05, + "loss": 1.0656, + "step": 18153 + }, + { + "epoch": 1.17, + "learning_rate": 1.6126151269885572e-05, + "loss": 1.0301, + "step": 18154 + }, + { + "epoch": 1.17, + "learning_rate": 1.612593658086262e-05, + "loss": 0.9625, + "step": 18155 + }, + { + "epoch": 1.17, + "learning_rate": 1.6125721891839673e-05, + "loss": 1.0155, + "step": 18156 + }, + { + "epoch": 1.17, + "learning_rate": 1.6125507202816722e-05, + "loss": 1.0936, + "step": 18157 + }, + { + "epoch": 1.17, + "learning_rate": 1.612529251379377e-05, + "loss": 0.931, + "step": 18158 + }, + { + "epoch": 1.17, + "learning_rate": 1.612507782477082e-05, + "loss": 0.9684, + "step": 18159 + }, + { + "epoch": 1.17, + "learning_rate": 1.612486313574787e-05, + "loss": 0.9363, + "step": 18160 + }, + { + "epoch": 1.17, + "learning_rate": 1.6124648446724918e-05, + "loss": 0.9436, + "step": 18161 + }, + { + "epoch": 1.17, + "learning_rate": 1.612443375770197e-05, + "loss": 0.9297, + "step": 18162 + }, + { + "epoch": 1.17, + "learning_rate": 1.612421906867902e-05, + "loss": 0.9378, + "step": 18163 + }, + { + "epoch": 1.17, + "learning_rate": 1.612400437965607e-05, + "loss": 1.0295, + "step": 18164 + }, + { + "epoch": 1.17, + "learning_rate": 1.6123789690633118e-05, + "loss": 1.0543, + "step": 18165 + }, + { + "epoch": 1.17, + "learning_rate": 1.612357500161017e-05, + "loss": 1.0337, + "step": 18166 + }, + { + "epoch": 1.17, + "learning_rate": 1.612336031258722e-05, + "loss": 0.9916, + "step": 18167 + }, + { + "epoch": 1.17, + "learning_rate": 1.6123145623564268e-05, + "loss": 1.0221, + "step": 18168 + }, + { + "epoch": 1.17, + "learning_rate": 1.612293093454132e-05, + "loss": 1.0166, + "step": 18169 + }, + { + "epoch": 1.17, + "learning_rate": 1.612271624551837e-05, + "loss": 0.9747, + "step": 18170 + }, + { + "epoch": 1.17, + "learning_rate": 1.6122501556495418e-05, + "loss": 1.0154, + "step": 18171 + }, + { + "epoch": 1.17, + "learning_rate": 1.6122286867472467e-05, + "loss": 0.9563, + "step": 18172 + }, + { + "epoch": 1.17, + "learning_rate": 1.6122072178449516e-05, + "loss": 0.8942, + "step": 18173 + }, + { + "epoch": 1.17, + "learning_rate": 1.612185748942657e-05, + "loss": 0.9251, + "step": 18174 + }, + { + "epoch": 1.17, + "learning_rate": 1.6121642800403617e-05, + "loss": 1.0249, + "step": 18175 + }, + { + "epoch": 1.17, + "learning_rate": 1.6121428111380666e-05, + "loss": 0.975, + "step": 18176 + }, + { + "epoch": 1.17, + "learning_rate": 1.6121213422357715e-05, + "loss": 0.9193, + "step": 18177 + }, + { + "epoch": 1.17, + "learning_rate": 1.6120998733334764e-05, + "loss": 0.8589, + "step": 18178 + }, + { + "epoch": 1.17, + "learning_rate": 1.6120784044311813e-05, + "loss": 1.0564, + "step": 18179 + }, + { + "epoch": 1.17, + "learning_rate": 1.6120569355288866e-05, + "loss": 1.09, + "step": 18180 + }, + { + "epoch": 1.17, + "learning_rate": 1.6120354666265914e-05, + "loss": 0.9808, + "step": 18181 + }, + { + "epoch": 1.17, + "learning_rate": 1.6120139977242967e-05, + "loss": 0.955, + "step": 18182 + }, + { + "epoch": 1.17, + "learning_rate": 1.6119925288220016e-05, + "loss": 0.9753, + "step": 18183 + }, + { + "epoch": 1.17, + "learning_rate": 1.6119710599197065e-05, + "loss": 1.0229, + "step": 18184 + }, + { + "epoch": 1.17, + "learning_rate": 1.6119495910174114e-05, + "loss": 0.9383, + "step": 18185 + }, + { + "epoch": 1.17, + "learning_rate": 1.6119281221151163e-05, + "loss": 0.8761, + "step": 18186 + }, + { + "epoch": 1.17, + "learning_rate": 1.6119066532128215e-05, + "loss": 0.9916, + "step": 18187 + }, + { + "epoch": 1.17, + "learning_rate": 1.6118851843105264e-05, + "loss": 0.9895, + "step": 18188 + }, + { + "epoch": 1.17, + "learning_rate": 1.6118637154082313e-05, + "loss": 0.9504, + "step": 18189 + }, + { + "epoch": 1.17, + "learning_rate": 1.6118422465059362e-05, + "loss": 1.0099, + "step": 18190 + }, + { + "epoch": 1.17, + "learning_rate": 1.611820777603641e-05, + "loss": 0.9871, + "step": 18191 + }, + { + "epoch": 1.17, + "learning_rate": 1.611799308701346e-05, + "loss": 1.0012, + "step": 18192 + }, + { + "epoch": 1.17, + "learning_rate": 1.6117778397990512e-05, + "loss": 1.0382, + "step": 18193 + }, + { + "epoch": 1.17, + "learning_rate": 1.611756370896756e-05, + "loss": 0.9737, + "step": 18194 + }, + { + "epoch": 1.17, + "learning_rate": 1.6117349019944614e-05, + "loss": 0.9955, + "step": 18195 + }, + { + "epoch": 1.17, + "learning_rate": 1.6117134330921663e-05, + "loss": 0.9117, + "step": 18196 + }, + { + "epoch": 1.17, + "learning_rate": 1.611691964189871e-05, + "loss": 0.9987, + "step": 18197 + }, + { + "epoch": 1.17, + "learning_rate": 1.611670495287576e-05, + "loss": 1.0259, + "step": 18198 + }, + { + "epoch": 1.17, + "learning_rate": 1.611649026385281e-05, + "loss": 0.8798, + "step": 18199 + }, + { + "epoch": 1.17, + "learning_rate": 1.6116275574829862e-05, + "loss": 0.9852, + "step": 18200 + }, + { + "epoch": 1.17, + "learning_rate": 1.611606088580691e-05, + "loss": 1.0261, + "step": 18201 + }, + { + "epoch": 1.17, + "learning_rate": 1.611584619678396e-05, + "loss": 0.928, + "step": 18202 + }, + { + "epoch": 1.17, + "learning_rate": 1.611563150776101e-05, + "loss": 0.9482, + "step": 18203 + }, + { + "epoch": 1.17, + "learning_rate": 1.6115416818738058e-05, + "loss": 1.1519, + "step": 18204 + }, + { + "epoch": 1.17, + "learning_rate": 1.611520212971511e-05, + "loss": 0.9627, + "step": 18205 + }, + { + "epoch": 1.17, + "learning_rate": 1.611498744069216e-05, + "loss": 1.0393, + "step": 18206 + }, + { + "epoch": 1.17, + "learning_rate": 1.6114772751669208e-05, + "loss": 0.9183, + "step": 18207 + }, + { + "epoch": 1.17, + "learning_rate": 1.6114558062646257e-05, + "loss": 1.1003, + "step": 18208 + }, + { + "epoch": 1.17, + "learning_rate": 1.611434337362331e-05, + "loss": 0.9632, + "step": 18209 + }, + { + "epoch": 1.17, + "learning_rate": 1.6114128684600358e-05, + "loss": 0.9035, + "step": 18210 + }, + { + "epoch": 1.17, + "learning_rate": 1.6113913995577407e-05, + "loss": 0.9395, + "step": 18211 + }, + { + "epoch": 1.17, + "learning_rate": 1.6113699306554456e-05, + "loss": 0.9483, + "step": 18212 + }, + { + "epoch": 1.17, + "learning_rate": 1.611348461753151e-05, + "loss": 0.9939, + "step": 18213 + }, + { + "epoch": 1.17, + "learning_rate": 1.6113269928508557e-05, + "loss": 0.9777, + "step": 18214 + }, + { + "epoch": 1.17, + "learning_rate": 1.6113055239485606e-05, + "loss": 0.9455, + "step": 18215 + }, + { + "epoch": 1.17, + "learning_rate": 1.6112840550462655e-05, + "loss": 0.9373, + "step": 18216 + }, + { + "epoch": 1.17, + "learning_rate": 1.6112625861439704e-05, + "loss": 0.9921, + "step": 18217 + }, + { + "epoch": 1.17, + "learning_rate": 1.6112411172416757e-05, + "loss": 0.9205, + "step": 18218 + }, + { + "epoch": 1.17, + "learning_rate": 1.6112196483393806e-05, + "loss": 0.9813, + "step": 18219 + }, + { + "epoch": 1.17, + "learning_rate": 1.6111981794370855e-05, + "loss": 0.9618, + "step": 18220 + }, + { + "epoch": 1.17, + "learning_rate": 1.6111767105347904e-05, + "loss": 0.9064, + "step": 18221 + }, + { + "epoch": 1.17, + "learning_rate": 1.6111552416324953e-05, + "loss": 0.9141, + "step": 18222 + }, + { + "epoch": 1.17, + "learning_rate": 1.6111337727302005e-05, + "loss": 0.9681, + "step": 18223 + }, + { + "epoch": 1.17, + "learning_rate": 1.6111123038279054e-05, + "loss": 0.9849, + "step": 18224 + }, + { + "epoch": 1.17, + "learning_rate": 1.6110908349256106e-05, + "loss": 0.9833, + "step": 18225 + }, + { + "epoch": 1.17, + "learning_rate": 1.6110693660233155e-05, + "loss": 0.9611, + "step": 18226 + }, + { + "epoch": 1.17, + "learning_rate": 1.6110478971210204e-05, + "loss": 0.926, + "step": 18227 + }, + { + "epoch": 1.17, + "learning_rate": 1.6110264282187253e-05, + "loss": 0.9391, + "step": 18228 + }, + { + "epoch": 1.17, + "learning_rate": 1.6110049593164302e-05, + "loss": 0.9193, + "step": 18229 + }, + { + "epoch": 1.17, + "learning_rate": 1.610983490414135e-05, + "loss": 1.0103, + "step": 18230 + }, + { + "epoch": 1.17, + "learning_rate": 1.6109620215118403e-05, + "loss": 0.9641, + "step": 18231 + }, + { + "epoch": 1.17, + "learning_rate": 1.6109405526095452e-05, + "loss": 0.9643, + "step": 18232 + }, + { + "epoch": 1.17, + "learning_rate": 1.61091908370725e-05, + "loss": 0.9664, + "step": 18233 + }, + { + "epoch": 1.17, + "learning_rate": 1.610897614804955e-05, + "loss": 1.0365, + "step": 18234 + }, + { + "epoch": 1.17, + "learning_rate": 1.61087614590266e-05, + "loss": 1.0005, + "step": 18235 + }, + { + "epoch": 1.17, + "learning_rate": 1.610854677000365e-05, + "loss": 0.9352, + "step": 18236 + }, + { + "epoch": 1.17, + "learning_rate": 1.61083320809807e-05, + "loss": 0.9623, + "step": 18237 + }, + { + "epoch": 1.17, + "learning_rate": 1.6108117391957753e-05, + "loss": 0.9246, + "step": 18238 + }, + { + "epoch": 1.17, + "learning_rate": 1.6107902702934802e-05, + "loss": 0.9812, + "step": 18239 + }, + { + "epoch": 1.17, + "learning_rate": 1.610768801391185e-05, + "loss": 0.9436, + "step": 18240 + }, + { + "epoch": 1.17, + "learning_rate": 1.61074733248889e-05, + "loss": 1.0199, + "step": 18241 + }, + { + "epoch": 1.17, + "learning_rate": 1.610725863586595e-05, + "loss": 1.0317, + "step": 18242 + }, + { + "epoch": 1.17, + "learning_rate": 1.6107043946843e-05, + "loss": 1.007, + "step": 18243 + }, + { + "epoch": 1.17, + "learning_rate": 1.610682925782005e-05, + "loss": 0.952, + "step": 18244 + }, + { + "epoch": 1.17, + "learning_rate": 1.61066145687971e-05, + "loss": 1.0222, + "step": 18245 + }, + { + "epoch": 1.17, + "learning_rate": 1.6106399879774148e-05, + "loss": 0.9856, + "step": 18246 + }, + { + "epoch": 1.17, + "learning_rate": 1.6106185190751197e-05, + "loss": 0.9323, + "step": 18247 + }, + { + "epoch": 1.17, + "learning_rate": 1.6105970501728246e-05, + "loss": 1.063, + "step": 18248 + }, + { + "epoch": 1.17, + "learning_rate": 1.61057558127053e-05, + "loss": 0.9821, + "step": 18249 + }, + { + "epoch": 1.17, + "learning_rate": 1.6105541123682347e-05, + "loss": 0.9541, + "step": 18250 + }, + { + "epoch": 1.17, + "learning_rate": 1.6105326434659396e-05, + "loss": 0.9716, + "step": 18251 + }, + { + "epoch": 1.17, + "learning_rate": 1.6105111745636445e-05, + "loss": 1.0226, + "step": 18252 + }, + { + "epoch": 1.17, + "learning_rate": 1.6104897056613498e-05, + "loss": 1.0654, + "step": 18253 + }, + { + "epoch": 1.17, + "learning_rate": 1.6104682367590547e-05, + "loss": 0.981, + "step": 18254 + }, + { + "epoch": 1.17, + "learning_rate": 1.6104467678567596e-05, + "loss": 1.0346, + "step": 18255 + }, + { + "epoch": 1.17, + "learning_rate": 1.6104252989544648e-05, + "loss": 1.0129, + "step": 18256 + }, + { + "epoch": 1.17, + "learning_rate": 1.6104038300521697e-05, + "loss": 1.0215, + "step": 18257 + }, + { + "epoch": 1.17, + "learning_rate": 1.6103823611498746e-05, + "loss": 1.0081, + "step": 18258 + }, + { + "epoch": 1.17, + "learning_rate": 1.6103608922475795e-05, + "loss": 0.9835, + "step": 18259 + }, + { + "epoch": 1.17, + "learning_rate": 1.6103394233452844e-05, + "loss": 0.8854, + "step": 18260 + }, + { + "epoch": 1.17, + "learning_rate": 1.6103179544429896e-05, + "loss": 0.9555, + "step": 18261 + }, + { + "epoch": 1.17, + "learning_rate": 1.6102964855406945e-05, + "loss": 1.0485, + "step": 18262 + }, + { + "epoch": 1.17, + "learning_rate": 1.6102750166383994e-05, + "loss": 1.0344, + "step": 18263 + }, + { + "epoch": 1.18, + "learning_rate": 1.6102535477361043e-05, + "loss": 0.9974, + "step": 18264 + }, + { + "epoch": 1.18, + "learning_rate": 1.6102320788338092e-05, + "loss": 1.0528, + "step": 18265 + }, + { + "epoch": 1.18, + "learning_rate": 1.610210609931514e-05, + "loss": 0.8909, + "step": 18266 + }, + { + "epoch": 1.18, + "learning_rate": 1.6101891410292193e-05, + "loss": 0.9677, + "step": 18267 + }, + { + "epoch": 1.18, + "learning_rate": 1.6101676721269242e-05, + "loss": 0.8912, + "step": 18268 + }, + { + "epoch": 1.18, + "learning_rate": 1.6101462032246295e-05, + "loss": 1.0061, + "step": 18269 + }, + { + "epoch": 1.18, + "learning_rate": 1.6101247343223344e-05, + "loss": 1.0464, + "step": 18270 + }, + { + "epoch": 1.18, + "learning_rate": 1.6101032654200392e-05, + "loss": 1.073, + "step": 18271 + }, + { + "epoch": 1.18, + "learning_rate": 1.610081796517744e-05, + "loss": 0.9591, + "step": 18272 + }, + { + "epoch": 1.18, + "learning_rate": 1.610060327615449e-05, + "loss": 0.9672, + "step": 18273 + }, + { + "epoch": 1.18, + "learning_rate": 1.6100388587131543e-05, + "loss": 1.0276, + "step": 18274 + }, + { + "epoch": 1.18, + "learning_rate": 1.6100173898108592e-05, + "loss": 0.9757, + "step": 18275 + }, + { + "epoch": 1.18, + "learning_rate": 1.609995920908564e-05, + "loss": 0.9971, + "step": 18276 + }, + { + "epoch": 1.18, + "learning_rate": 1.609974452006269e-05, + "loss": 1.0153, + "step": 18277 + }, + { + "epoch": 1.18, + "learning_rate": 1.609952983103974e-05, + "loss": 1.0252, + "step": 18278 + }, + { + "epoch": 1.18, + "learning_rate": 1.6099315142016788e-05, + "loss": 1.0337, + "step": 18279 + }, + { + "epoch": 1.18, + "learning_rate": 1.609910045299384e-05, + "loss": 0.946, + "step": 18280 + }, + { + "epoch": 1.18, + "learning_rate": 1.609888576397089e-05, + "loss": 0.9469, + "step": 18281 + }, + { + "epoch": 1.18, + "learning_rate": 1.609867107494794e-05, + "loss": 0.9176, + "step": 18282 + }, + { + "epoch": 1.18, + "learning_rate": 1.609845638592499e-05, + "loss": 1.0526, + "step": 18283 + }, + { + "epoch": 1.18, + "learning_rate": 1.609824169690204e-05, + "loss": 0.9446, + "step": 18284 + }, + { + "epoch": 1.18, + "learning_rate": 1.6098027007879088e-05, + "loss": 1.0452, + "step": 18285 + }, + { + "epoch": 1.18, + "learning_rate": 1.6097812318856137e-05, + "loss": 0.9209, + "step": 18286 + }, + { + "epoch": 1.18, + "learning_rate": 1.609759762983319e-05, + "loss": 0.993, + "step": 18287 + }, + { + "epoch": 1.18, + "learning_rate": 1.609738294081024e-05, + "loss": 0.945, + "step": 18288 + }, + { + "epoch": 1.18, + "learning_rate": 1.6097168251787287e-05, + "loss": 0.9581, + "step": 18289 + }, + { + "epoch": 1.18, + "learning_rate": 1.6096953562764336e-05, + "loss": 1.0335, + "step": 18290 + }, + { + "epoch": 1.18, + "learning_rate": 1.6096738873741385e-05, + "loss": 0.9666, + "step": 18291 + }, + { + "epoch": 1.18, + "learning_rate": 1.6096524184718438e-05, + "loss": 1.0483, + "step": 18292 + }, + { + "epoch": 1.18, + "learning_rate": 1.6096309495695487e-05, + "loss": 0.9397, + "step": 18293 + }, + { + "epoch": 1.18, + "learning_rate": 1.6096094806672536e-05, + "loss": 0.946, + "step": 18294 + }, + { + "epoch": 1.18, + "learning_rate": 1.6095880117649585e-05, + "loss": 1.028, + "step": 18295 + }, + { + "epoch": 1.18, + "learning_rate": 1.6095665428626637e-05, + "loss": 0.9914, + "step": 18296 + }, + { + "epoch": 1.18, + "learning_rate": 1.6095450739603686e-05, + "loss": 0.9562, + "step": 18297 + }, + { + "epoch": 1.18, + "learning_rate": 1.6095236050580735e-05, + "loss": 0.9805, + "step": 18298 + }, + { + "epoch": 1.18, + "learning_rate": 1.6095021361557784e-05, + "loss": 0.9309, + "step": 18299 + }, + { + "epoch": 1.18, + "learning_rate": 1.6094806672534836e-05, + "loss": 0.9666, + "step": 18300 + }, + { + "epoch": 1.18, + "learning_rate": 1.6094591983511885e-05, + "loss": 0.974, + "step": 18301 + }, + { + "epoch": 1.18, + "learning_rate": 1.6094377294488934e-05, + "loss": 0.9872, + "step": 18302 + }, + { + "epoch": 1.18, + "learning_rate": 1.6094162605465983e-05, + "loss": 0.9557, + "step": 18303 + }, + { + "epoch": 1.18, + "learning_rate": 1.6093947916443032e-05, + "loss": 0.9721, + "step": 18304 + }, + { + "epoch": 1.18, + "learning_rate": 1.6093733227420084e-05, + "loss": 0.9566, + "step": 18305 + }, + { + "epoch": 1.18, + "learning_rate": 1.6093518538397133e-05, + "loss": 1.0348, + "step": 18306 + }, + { + "epoch": 1.18, + "learning_rate": 1.6093303849374182e-05, + "loss": 0.9147, + "step": 18307 + }, + { + "epoch": 1.18, + "learning_rate": 1.609308916035123e-05, + "loss": 1.004, + "step": 18308 + }, + { + "epoch": 1.18, + "learning_rate": 1.609287447132828e-05, + "loss": 0.9582, + "step": 18309 + }, + { + "epoch": 1.18, + "learning_rate": 1.6092659782305333e-05, + "loss": 0.9272, + "step": 18310 + }, + { + "epoch": 1.18, + "learning_rate": 1.609244509328238e-05, + "loss": 1.0361, + "step": 18311 + }, + { + "epoch": 1.18, + "learning_rate": 1.6092230404259434e-05, + "loss": 1.0068, + "step": 18312 + }, + { + "epoch": 1.18, + "learning_rate": 1.6092015715236483e-05, + "loss": 0.91, + "step": 18313 + }, + { + "epoch": 1.18, + "learning_rate": 1.6091801026213532e-05, + "loss": 0.9841, + "step": 18314 + }, + { + "epoch": 1.18, + "learning_rate": 1.609158633719058e-05, + "loss": 1.0197, + "step": 18315 + }, + { + "epoch": 1.18, + "learning_rate": 1.609137164816763e-05, + "loss": 1.002, + "step": 18316 + }, + { + "epoch": 1.18, + "learning_rate": 1.609115695914468e-05, + "loss": 1.0554, + "step": 18317 + }, + { + "epoch": 1.18, + "learning_rate": 1.609094227012173e-05, + "loss": 1.0505, + "step": 18318 + }, + { + "epoch": 1.18, + "learning_rate": 1.609072758109878e-05, + "loss": 0.9772, + "step": 18319 + }, + { + "epoch": 1.18, + "learning_rate": 1.609051289207583e-05, + "loss": 0.9774, + "step": 18320 + }, + { + "epoch": 1.18, + "learning_rate": 1.6090298203052878e-05, + "loss": 0.9671, + "step": 18321 + }, + { + "epoch": 1.18, + "learning_rate": 1.6090083514029927e-05, + "loss": 0.9333, + "step": 18322 + }, + { + "epoch": 1.18, + "learning_rate": 1.608986882500698e-05, + "loss": 1.0195, + "step": 18323 + }, + { + "epoch": 1.18, + "learning_rate": 1.6089654135984028e-05, + "loss": 0.9664, + "step": 18324 + }, + { + "epoch": 1.18, + "learning_rate": 1.608943944696108e-05, + "loss": 0.9986, + "step": 18325 + }, + { + "epoch": 1.18, + "learning_rate": 1.608922475793813e-05, + "loss": 0.9708, + "step": 18326 + }, + { + "epoch": 1.18, + "learning_rate": 1.608901006891518e-05, + "loss": 0.947, + "step": 18327 + }, + { + "epoch": 1.18, + "learning_rate": 1.6088795379892228e-05, + "loss": 0.9955, + "step": 18328 + }, + { + "epoch": 1.18, + "learning_rate": 1.6088580690869277e-05, + "loss": 1.0195, + "step": 18329 + }, + { + "epoch": 1.18, + "learning_rate": 1.608836600184633e-05, + "loss": 0.9741, + "step": 18330 + }, + { + "epoch": 1.18, + "learning_rate": 1.6088151312823378e-05, + "loss": 0.9073, + "step": 18331 + }, + { + "epoch": 1.18, + "learning_rate": 1.6087936623800427e-05, + "loss": 0.9925, + "step": 18332 + }, + { + "epoch": 1.18, + "learning_rate": 1.6087721934777476e-05, + "loss": 0.9426, + "step": 18333 + }, + { + "epoch": 1.18, + "learning_rate": 1.6087507245754525e-05, + "loss": 0.9807, + "step": 18334 + }, + { + "epoch": 1.18, + "learning_rate": 1.6087292556731574e-05, + "loss": 0.9678, + "step": 18335 + }, + { + "epoch": 1.18, + "learning_rate": 1.6087077867708626e-05, + "loss": 1.0453, + "step": 18336 + }, + { + "epoch": 1.18, + "learning_rate": 1.6086863178685675e-05, + "loss": 0.9613, + "step": 18337 + }, + { + "epoch": 1.18, + "learning_rate": 1.6086648489662724e-05, + "loss": 0.9687, + "step": 18338 + }, + { + "epoch": 1.18, + "learning_rate": 1.6086433800639773e-05, + "loss": 0.9337, + "step": 18339 + }, + { + "epoch": 1.18, + "learning_rate": 1.6086219111616825e-05, + "loss": 0.998, + "step": 18340 + }, + { + "epoch": 1.18, + "learning_rate": 1.6086004422593874e-05, + "loss": 0.9585, + "step": 18341 + }, + { + "epoch": 1.18, + "learning_rate": 1.6085789733570923e-05, + "loss": 1.0558, + "step": 18342 + }, + { + "epoch": 1.18, + "learning_rate": 1.6085575044547976e-05, + "loss": 1.012, + "step": 18343 + }, + { + "epoch": 1.18, + "learning_rate": 1.6085360355525025e-05, + "loss": 0.9441, + "step": 18344 + }, + { + "epoch": 1.18, + "learning_rate": 1.6085145666502073e-05, + "loss": 0.9989, + "step": 18345 + }, + { + "epoch": 1.18, + "learning_rate": 1.6084930977479122e-05, + "loss": 0.9885, + "step": 18346 + }, + { + "epoch": 1.18, + "learning_rate": 1.608471628845617e-05, + "loss": 1.0329, + "step": 18347 + }, + { + "epoch": 1.18, + "learning_rate": 1.6084501599433224e-05, + "loss": 1.0218, + "step": 18348 + }, + { + "epoch": 1.18, + "learning_rate": 1.6084286910410273e-05, + "loss": 0.8711, + "step": 18349 + }, + { + "epoch": 1.18, + "learning_rate": 1.6084072221387322e-05, + "loss": 0.9008, + "step": 18350 + }, + { + "epoch": 1.18, + "learning_rate": 1.608385753236437e-05, + "loss": 0.9294, + "step": 18351 + }, + { + "epoch": 1.18, + "learning_rate": 1.608364284334142e-05, + "loss": 1.0402, + "step": 18352 + }, + { + "epoch": 1.18, + "learning_rate": 1.608342815431847e-05, + "loss": 0.9684, + "step": 18353 + }, + { + "epoch": 1.18, + "learning_rate": 1.608321346529552e-05, + "loss": 0.9544, + "step": 18354 + }, + { + "epoch": 1.18, + "learning_rate": 1.608299877627257e-05, + "loss": 0.9653, + "step": 18355 + }, + { + "epoch": 1.18, + "learning_rate": 1.6082784087249622e-05, + "loss": 0.9768, + "step": 18356 + }, + { + "epoch": 1.18, + "learning_rate": 1.608256939822667e-05, + "loss": 0.9684, + "step": 18357 + }, + { + "epoch": 1.18, + "learning_rate": 1.608235470920372e-05, + "loss": 0.9426, + "step": 18358 + }, + { + "epoch": 1.18, + "learning_rate": 1.608214002018077e-05, + "loss": 0.9308, + "step": 18359 + }, + { + "epoch": 1.18, + "learning_rate": 1.6081925331157818e-05, + "loss": 0.9715, + "step": 18360 + }, + { + "epoch": 1.18, + "learning_rate": 1.608171064213487e-05, + "loss": 0.9954, + "step": 18361 + }, + { + "epoch": 1.18, + "learning_rate": 1.608149595311192e-05, + "loss": 0.9364, + "step": 18362 + }, + { + "epoch": 1.18, + "learning_rate": 1.608128126408897e-05, + "loss": 0.9923, + "step": 18363 + }, + { + "epoch": 1.18, + "learning_rate": 1.6081066575066017e-05, + "loss": 0.976, + "step": 18364 + }, + { + "epoch": 1.18, + "learning_rate": 1.6080851886043066e-05, + "loss": 0.9135, + "step": 18365 + }, + { + "epoch": 1.18, + "learning_rate": 1.6080637197020115e-05, + "loss": 0.9433, + "step": 18366 + }, + { + "epoch": 1.18, + "learning_rate": 1.6080422507997168e-05, + "loss": 1.0067, + "step": 18367 + }, + { + "epoch": 1.18, + "learning_rate": 1.6080207818974217e-05, + "loss": 0.973, + "step": 18368 + }, + { + "epoch": 1.18, + "learning_rate": 1.607999312995127e-05, + "loss": 0.9765, + "step": 18369 + }, + { + "epoch": 1.18, + "learning_rate": 1.6079778440928318e-05, + "loss": 0.9788, + "step": 18370 + }, + { + "epoch": 1.18, + "learning_rate": 1.6079563751905367e-05, + "loss": 0.964, + "step": 18371 + }, + { + "epoch": 1.18, + "learning_rate": 1.6079349062882416e-05, + "loss": 0.9396, + "step": 18372 + }, + { + "epoch": 1.18, + "learning_rate": 1.6079134373859465e-05, + "loss": 0.9858, + "step": 18373 + }, + { + "epoch": 1.18, + "learning_rate": 1.6078919684836517e-05, + "loss": 0.9041, + "step": 18374 + }, + { + "epoch": 1.18, + "learning_rate": 1.6078704995813566e-05, + "loss": 0.9773, + "step": 18375 + }, + { + "epoch": 1.18, + "learning_rate": 1.6078490306790615e-05, + "loss": 0.9369, + "step": 18376 + }, + { + "epoch": 1.18, + "learning_rate": 1.6078275617767664e-05, + "loss": 0.9737, + "step": 18377 + }, + { + "epoch": 1.18, + "learning_rate": 1.6078060928744713e-05, + "loss": 0.926, + "step": 18378 + }, + { + "epoch": 1.18, + "learning_rate": 1.6077846239721765e-05, + "loss": 0.9513, + "step": 18379 + }, + { + "epoch": 1.18, + "learning_rate": 1.6077631550698814e-05, + "loss": 0.9474, + "step": 18380 + }, + { + "epoch": 1.18, + "learning_rate": 1.6077416861675863e-05, + "loss": 0.9942, + "step": 18381 + }, + { + "epoch": 1.18, + "learning_rate": 1.6077202172652912e-05, + "loss": 1.0241, + "step": 18382 + }, + { + "epoch": 1.18, + "learning_rate": 1.6076987483629965e-05, + "loss": 0.9942, + "step": 18383 + }, + { + "epoch": 1.18, + "learning_rate": 1.6076772794607014e-05, + "loss": 1.0194, + "step": 18384 + }, + { + "epoch": 1.18, + "learning_rate": 1.6076558105584063e-05, + "loss": 0.9012, + "step": 18385 + }, + { + "epoch": 1.18, + "learning_rate": 1.607634341656111e-05, + "loss": 0.9845, + "step": 18386 + }, + { + "epoch": 1.18, + "learning_rate": 1.6076128727538164e-05, + "loss": 0.9255, + "step": 18387 + }, + { + "epoch": 1.18, + "learning_rate": 1.6075914038515213e-05, + "loss": 0.9261, + "step": 18388 + }, + { + "epoch": 1.18, + "learning_rate": 1.6075699349492262e-05, + "loss": 0.981, + "step": 18389 + }, + { + "epoch": 1.18, + "learning_rate": 1.607548466046931e-05, + "loss": 0.9566, + "step": 18390 + }, + { + "epoch": 1.18, + "learning_rate": 1.607526997144636e-05, + "loss": 0.9451, + "step": 18391 + }, + { + "epoch": 1.18, + "learning_rate": 1.6075055282423412e-05, + "loss": 0.9982, + "step": 18392 + }, + { + "epoch": 1.18, + "learning_rate": 1.607484059340046e-05, + "loss": 0.953, + "step": 18393 + }, + { + "epoch": 1.18, + "learning_rate": 1.607462590437751e-05, + "loss": 1.0004, + "step": 18394 + }, + { + "epoch": 1.18, + "learning_rate": 1.607441121535456e-05, + "loss": 0.9922, + "step": 18395 + }, + { + "epoch": 1.18, + "learning_rate": 1.6074196526331608e-05, + "loss": 1.0, + "step": 18396 + }, + { + "epoch": 1.18, + "learning_rate": 1.607398183730866e-05, + "loss": 1.0325, + "step": 18397 + }, + { + "epoch": 1.18, + "learning_rate": 1.607376714828571e-05, + "loss": 1.0443, + "step": 18398 + }, + { + "epoch": 1.18, + "learning_rate": 1.607355245926276e-05, + "loss": 0.936, + "step": 18399 + }, + { + "epoch": 1.18, + "learning_rate": 1.607333777023981e-05, + "loss": 0.9432, + "step": 18400 + }, + { + "epoch": 1.18, + "learning_rate": 1.607312308121686e-05, + "loss": 0.9335, + "step": 18401 + }, + { + "epoch": 1.18, + "learning_rate": 1.607290839219391e-05, + "loss": 0.9252, + "step": 18402 + }, + { + "epoch": 1.18, + "learning_rate": 1.6072693703170958e-05, + "loss": 1.0041, + "step": 18403 + }, + { + "epoch": 1.18, + "learning_rate": 1.6072479014148006e-05, + "loss": 1.027, + "step": 18404 + }, + { + "epoch": 1.18, + "learning_rate": 1.607226432512506e-05, + "loss": 1.024, + "step": 18405 + }, + { + "epoch": 1.18, + "learning_rate": 1.6072049636102108e-05, + "loss": 0.9852, + "step": 18406 + }, + { + "epoch": 1.18, + "learning_rate": 1.6071834947079157e-05, + "loss": 0.9526, + "step": 18407 + }, + { + "epoch": 1.18, + "learning_rate": 1.6071620258056206e-05, + "loss": 0.9587, + "step": 18408 + }, + { + "epoch": 1.18, + "learning_rate": 1.6071405569033255e-05, + "loss": 0.9618, + "step": 18409 + }, + { + "epoch": 1.18, + "learning_rate": 1.6071190880010307e-05, + "loss": 1.0187, + "step": 18410 + }, + { + "epoch": 1.18, + "learning_rate": 1.6070976190987356e-05, + "loss": 1.0411, + "step": 18411 + }, + { + "epoch": 1.18, + "learning_rate": 1.6070761501964405e-05, + "loss": 0.9987, + "step": 18412 + }, + { + "epoch": 1.18, + "learning_rate": 1.6070546812941457e-05, + "loss": 1.0042, + "step": 18413 + }, + { + "epoch": 1.18, + "learning_rate": 1.6070332123918506e-05, + "loss": 0.9447, + "step": 18414 + }, + { + "epoch": 1.18, + "learning_rate": 1.6070117434895555e-05, + "loss": 1.0205, + "step": 18415 + }, + { + "epoch": 1.18, + "learning_rate": 1.6069902745872604e-05, + "loss": 1.0398, + "step": 18416 + }, + { + "epoch": 1.18, + "learning_rate": 1.6069688056849657e-05, + "loss": 1.0068, + "step": 18417 + }, + { + "epoch": 1.18, + "learning_rate": 1.6069473367826706e-05, + "loss": 1.0195, + "step": 18418 + }, + { + "epoch": 1.19, + "learning_rate": 1.6069258678803754e-05, + "loss": 1.0142, + "step": 18419 + }, + { + "epoch": 1.19, + "learning_rate": 1.6069043989780803e-05, + "loss": 0.9557, + "step": 18420 + }, + { + "epoch": 1.19, + "learning_rate": 1.6068829300757852e-05, + "loss": 0.9125, + "step": 18421 + }, + { + "epoch": 1.19, + "learning_rate": 1.60686146117349e-05, + "loss": 0.99, + "step": 18422 + }, + { + "epoch": 1.19, + "learning_rate": 1.6068399922711954e-05, + "loss": 0.9685, + "step": 18423 + }, + { + "epoch": 1.19, + "learning_rate": 1.6068185233689003e-05, + "loss": 0.958, + "step": 18424 + }, + { + "epoch": 1.19, + "learning_rate": 1.606797054466605e-05, + "loss": 1.0074, + "step": 18425 + }, + { + "epoch": 1.19, + "learning_rate": 1.60677558556431e-05, + "loss": 0.9552, + "step": 18426 + }, + { + "epoch": 1.19, + "learning_rate": 1.6067541166620153e-05, + "loss": 1.0524, + "step": 18427 + }, + { + "epoch": 1.19, + "learning_rate": 1.6067326477597202e-05, + "loss": 0.974, + "step": 18428 + }, + { + "epoch": 1.19, + "learning_rate": 1.606711178857425e-05, + "loss": 1.0295, + "step": 18429 + }, + { + "epoch": 1.19, + "learning_rate": 1.6066897099551303e-05, + "loss": 1.0875, + "step": 18430 + }, + { + "epoch": 1.19, + "learning_rate": 1.6066682410528352e-05, + "loss": 0.9888, + "step": 18431 + }, + { + "epoch": 1.19, + "learning_rate": 1.60664677215054e-05, + "loss": 1.0194, + "step": 18432 + }, + { + "epoch": 1.19, + "learning_rate": 1.606625303248245e-05, + "loss": 0.9142, + "step": 18433 + }, + { + "epoch": 1.19, + "learning_rate": 1.60660383434595e-05, + "loss": 1.0214, + "step": 18434 + }, + { + "epoch": 1.19, + "learning_rate": 1.6065823654436548e-05, + "loss": 0.9601, + "step": 18435 + }, + { + "epoch": 1.19, + "learning_rate": 1.60656089654136e-05, + "loss": 1.0099, + "step": 18436 + }, + { + "epoch": 1.19, + "learning_rate": 1.606539427639065e-05, + "loss": 0.8862, + "step": 18437 + }, + { + "epoch": 1.19, + "learning_rate": 1.60651795873677e-05, + "loss": 1.0157, + "step": 18438 + }, + { + "epoch": 1.19, + "learning_rate": 1.6064964898344747e-05, + "loss": 0.9693, + "step": 18439 + }, + { + "epoch": 1.19, + "learning_rate": 1.6064750209321796e-05, + "loss": 0.9559, + "step": 18440 + }, + { + "epoch": 1.19, + "learning_rate": 1.606453552029885e-05, + "loss": 0.9636, + "step": 18441 + }, + { + "epoch": 1.19, + "learning_rate": 1.6064320831275898e-05, + "loss": 0.9699, + "step": 18442 + }, + { + "epoch": 1.19, + "learning_rate": 1.606410614225295e-05, + "loss": 1.0336, + "step": 18443 + }, + { + "epoch": 1.19, + "learning_rate": 1.606389145323e-05, + "loss": 0.9814, + "step": 18444 + }, + { + "epoch": 1.19, + "learning_rate": 1.6063676764207048e-05, + "loss": 0.9352, + "step": 18445 + }, + { + "epoch": 1.19, + "learning_rate": 1.6063462075184097e-05, + "loss": 0.9717, + "step": 18446 + }, + { + "epoch": 1.19, + "learning_rate": 1.6063247386161146e-05, + "loss": 0.9805, + "step": 18447 + }, + { + "epoch": 1.19, + "learning_rate": 1.6063032697138198e-05, + "loss": 1.01, + "step": 18448 + }, + { + "epoch": 1.19, + "learning_rate": 1.6062818008115247e-05, + "loss": 0.9669, + "step": 18449 + }, + { + "epoch": 1.19, + "learning_rate": 1.6062603319092296e-05, + "loss": 1.0323, + "step": 18450 + }, + { + "epoch": 1.19, + "learning_rate": 1.6062388630069345e-05, + "loss": 0.9409, + "step": 18451 + }, + { + "epoch": 1.19, + "learning_rate": 1.6062173941046394e-05, + "loss": 0.9785, + "step": 18452 + }, + { + "epoch": 1.19, + "learning_rate": 1.6061959252023443e-05, + "loss": 0.9184, + "step": 18453 + }, + { + "epoch": 1.19, + "learning_rate": 1.6061744563000495e-05, + "loss": 0.9356, + "step": 18454 + }, + { + "epoch": 1.19, + "learning_rate": 1.6061529873977544e-05, + "loss": 1.048, + "step": 18455 + }, + { + "epoch": 1.19, + "learning_rate": 1.6061315184954597e-05, + "loss": 1.0742, + "step": 18456 + }, + { + "epoch": 1.19, + "learning_rate": 1.6061100495931646e-05, + "loss": 0.9429, + "step": 18457 + }, + { + "epoch": 1.19, + "learning_rate": 1.6060885806908695e-05, + "loss": 1.0412, + "step": 18458 + }, + { + "epoch": 1.19, + "learning_rate": 1.6060671117885744e-05, + "loss": 0.9875, + "step": 18459 + }, + { + "epoch": 1.19, + "learning_rate": 1.6060456428862793e-05, + "loss": 0.9736, + "step": 18460 + }, + { + "epoch": 1.19, + "learning_rate": 1.6060241739839845e-05, + "loss": 1.0221, + "step": 18461 + }, + { + "epoch": 1.19, + "learning_rate": 1.6060027050816894e-05, + "loss": 0.9495, + "step": 18462 + }, + { + "epoch": 1.19, + "learning_rate": 1.6059812361793943e-05, + "loss": 0.9647, + "step": 18463 + }, + { + "epoch": 1.19, + "learning_rate": 1.6059597672770992e-05, + "loss": 1.1248, + "step": 18464 + }, + { + "epoch": 1.19, + "learning_rate": 1.605938298374804e-05, + "loss": 1.0067, + "step": 18465 + }, + { + "epoch": 1.19, + "learning_rate": 1.6059168294725093e-05, + "loss": 1.0253, + "step": 18466 + }, + { + "epoch": 1.19, + "learning_rate": 1.6058953605702142e-05, + "loss": 0.9582, + "step": 18467 + }, + { + "epoch": 1.19, + "learning_rate": 1.605873891667919e-05, + "loss": 0.9365, + "step": 18468 + }, + { + "epoch": 1.19, + "learning_rate": 1.605852422765624e-05, + "loss": 0.9795, + "step": 18469 + }, + { + "epoch": 1.19, + "learning_rate": 1.605830953863329e-05, + "loss": 0.9747, + "step": 18470 + }, + { + "epoch": 1.19, + "learning_rate": 1.605809484961034e-05, + "loss": 0.9746, + "step": 18471 + }, + { + "epoch": 1.19, + "learning_rate": 1.605788016058739e-05, + "loss": 0.9516, + "step": 18472 + }, + { + "epoch": 1.19, + "learning_rate": 1.605766547156444e-05, + "loss": 0.9846, + "step": 18473 + }, + { + "epoch": 1.19, + "learning_rate": 1.605745078254149e-05, + "loss": 1.0245, + "step": 18474 + }, + { + "epoch": 1.19, + "learning_rate": 1.605723609351854e-05, + "loss": 0.9952, + "step": 18475 + }, + { + "epoch": 1.19, + "learning_rate": 1.605702140449559e-05, + "loss": 1.0171, + "step": 18476 + }, + { + "epoch": 1.19, + "learning_rate": 1.605680671547264e-05, + "loss": 1.0445, + "step": 18477 + }, + { + "epoch": 1.19, + "learning_rate": 1.6056592026449687e-05, + "loss": 0.9871, + "step": 18478 + }, + { + "epoch": 1.19, + "learning_rate": 1.605637733742674e-05, + "loss": 1.0171, + "step": 18479 + }, + { + "epoch": 1.19, + "learning_rate": 1.605616264840379e-05, + "loss": 0.9607, + "step": 18480 + }, + { + "epoch": 1.19, + "learning_rate": 1.6055947959380838e-05, + "loss": 0.989, + "step": 18481 + }, + { + "epoch": 1.19, + "learning_rate": 1.6055733270357887e-05, + "loss": 1.0043, + "step": 18482 + }, + { + "epoch": 1.19, + "learning_rate": 1.6055518581334936e-05, + "loss": 0.9474, + "step": 18483 + }, + { + "epoch": 1.19, + "learning_rate": 1.6055303892311988e-05, + "loss": 0.972, + "step": 18484 + }, + { + "epoch": 1.19, + "learning_rate": 1.6055089203289037e-05, + "loss": 0.9946, + "step": 18485 + }, + { + "epoch": 1.19, + "learning_rate": 1.605487451426609e-05, + "loss": 0.9705, + "step": 18486 + }, + { + "epoch": 1.19, + "learning_rate": 1.605465982524314e-05, + "loss": 0.9515, + "step": 18487 + }, + { + "epoch": 1.19, + "learning_rate": 1.6054445136220187e-05, + "loss": 1.0262, + "step": 18488 + }, + { + "epoch": 1.19, + "learning_rate": 1.6054230447197236e-05, + "loss": 0.9938, + "step": 18489 + }, + { + "epoch": 1.19, + "learning_rate": 1.6054015758174285e-05, + "loss": 0.9787, + "step": 18490 + }, + { + "epoch": 1.19, + "learning_rate": 1.6053801069151334e-05, + "loss": 0.9764, + "step": 18491 + }, + { + "epoch": 1.19, + "learning_rate": 1.6053586380128387e-05, + "loss": 1.0031, + "step": 18492 + }, + { + "epoch": 1.19, + "learning_rate": 1.6053371691105436e-05, + "loss": 0.9317, + "step": 18493 + }, + { + "epoch": 1.19, + "learning_rate": 1.6053157002082484e-05, + "loss": 0.9976, + "step": 18494 + }, + { + "epoch": 1.19, + "learning_rate": 1.6052942313059533e-05, + "loss": 1.0211, + "step": 18495 + }, + { + "epoch": 1.19, + "learning_rate": 1.6052727624036582e-05, + "loss": 1.001, + "step": 18496 + }, + { + "epoch": 1.19, + "learning_rate": 1.6052512935013635e-05, + "loss": 0.9672, + "step": 18497 + }, + { + "epoch": 1.19, + "learning_rate": 1.6052298245990684e-05, + "loss": 0.9338, + "step": 18498 + }, + { + "epoch": 1.19, + "learning_rate": 1.6052083556967733e-05, + "loss": 0.9912, + "step": 18499 + }, + { + "epoch": 1.19, + "learning_rate": 1.6051868867944785e-05, + "loss": 0.9674, + "step": 18500 + }, + { + "epoch": 1.19, + "learning_rate": 1.6051654178921834e-05, + "loss": 0.9349, + "step": 18501 + }, + { + "epoch": 1.19, + "learning_rate": 1.6051439489898883e-05, + "loss": 0.9585, + "step": 18502 + }, + { + "epoch": 1.19, + "learning_rate": 1.6051224800875932e-05, + "loss": 1.051, + "step": 18503 + }, + { + "epoch": 1.19, + "learning_rate": 1.6051010111852984e-05, + "loss": 1.0485, + "step": 18504 + }, + { + "epoch": 1.19, + "learning_rate": 1.6050795422830033e-05, + "loss": 1.0148, + "step": 18505 + }, + { + "epoch": 1.19, + "learning_rate": 1.6050580733807082e-05, + "loss": 0.994, + "step": 18506 + }, + { + "epoch": 1.19, + "learning_rate": 1.605036604478413e-05, + "loss": 1.0031, + "step": 18507 + }, + { + "epoch": 1.19, + "learning_rate": 1.605015135576118e-05, + "loss": 0.986, + "step": 18508 + }, + { + "epoch": 1.19, + "learning_rate": 1.604993666673823e-05, + "loss": 0.946, + "step": 18509 + }, + { + "epoch": 1.19, + "learning_rate": 1.604972197771528e-05, + "loss": 0.997, + "step": 18510 + }, + { + "epoch": 1.19, + "learning_rate": 1.604950728869233e-05, + "loss": 0.9118, + "step": 18511 + }, + { + "epoch": 1.19, + "learning_rate": 1.604929259966938e-05, + "loss": 1.0783, + "step": 18512 + }, + { + "epoch": 1.19, + "learning_rate": 1.604907791064643e-05, + "loss": 0.9497, + "step": 18513 + }, + { + "epoch": 1.19, + "learning_rate": 1.604886322162348e-05, + "loss": 1.0202, + "step": 18514 + }, + { + "epoch": 1.19, + "learning_rate": 1.604864853260053e-05, + "loss": 0.9561, + "step": 18515 + }, + { + "epoch": 1.19, + "learning_rate": 1.604843384357758e-05, + "loss": 0.9253, + "step": 18516 + }, + { + "epoch": 1.19, + "learning_rate": 1.604821915455463e-05, + "loss": 0.9808, + "step": 18517 + }, + { + "epoch": 1.19, + "learning_rate": 1.604800446553168e-05, + "loss": 0.9599, + "step": 18518 + }, + { + "epoch": 1.19, + "learning_rate": 1.604778977650873e-05, + "loss": 0.9319, + "step": 18519 + }, + { + "epoch": 1.19, + "learning_rate": 1.6047575087485778e-05, + "loss": 0.9862, + "step": 18520 + }, + { + "epoch": 1.19, + "learning_rate": 1.6047360398462827e-05, + "loss": 1.0274, + "step": 18521 + }, + { + "epoch": 1.19, + "learning_rate": 1.6047145709439876e-05, + "loss": 1.0162, + "step": 18522 + }, + { + "epoch": 1.19, + "learning_rate": 1.6046931020416928e-05, + "loss": 0.9438, + "step": 18523 + }, + { + "epoch": 1.19, + "learning_rate": 1.6046716331393977e-05, + "loss": 0.9985, + "step": 18524 + }, + { + "epoch": 1.19, + "learning_rate": 1.6046501642371026e-05, + "loss": 0.9386, + "step": 18525 + }, + { + "epoch": 1.19, + "learning_rate": 1.6046286953348075e-05, + "loss": 0.9382, + "step": 18526 + }, + { + "epoch": 1.19, + "learning_rate": 1.6046072264325124e-05, + "loss": 0.9626, + "step": 18527 + }, + { + "epoch": 1.19, + "learning_rate": 1.6045857575302176e-05, + "loss": 0.9129, + "step": 18528 + }, + { + "epoch": 1.19, + "learning_rate": 1.6045642886279225e-05, + "loss": 1.0279, + "step": 18529 + }, + { + "epoch": 1.19, + "learning_rate": 1.6045428197256278e-05, + "loss": 0.9493, + "step": 18530 + }, + { + "epoch": 1.19, + "learning_rate": 1.6045213508233327e-05, + "loss": 0.9244, + "step": 18531 + }, + { + "epoch": 1.19, + "learning_rate": 1.6044998819210376e-05, + "loss": 0.9852, + "step": 18532 + }, + { + "epoch": 1.19, + "learning_rate": 1.6044784130187425e-05, + "loss": 0.9501, + "step": 18533 + }, + { + "epoch": 1.19, + "learning_rate": 1.6044569441164474e-05, + "loss": 0.919, + "step": 18534 + }, + { + "epoch": 1.19, + "learning_rate": 1.6044354752141526e-05, + "loss": 0.9341, + "step": 18535 + }, + { + "epoch": 1.19, + "learning_rate": 1.6044140063118575e-05, + "loss": 0.9357, + "step": 18536 + }, + { + "epoch": 1.19, + "learning_rate": 1.6043925374095624e-05, + "loss": 0.9106, + "step": 18537 + }, + { + "epoch": 1.19, + "learning_rate": 1.6043710685072673e-05, + "loss": 1.0427, + "step": 18538 + }, + { + "epoch": 1.19, + "learning_rate": 1.6043495996049722e-05, + "loss": 0.9518, + "step": 18539 + }, + { + "epoch": 1.19, + "learning_rate": 1.604328130702677e-05, + "loss": 1.0382, + "step": 18540 + }, + { + "epoch": 1.19, + "learning_rate": 1.6043066618003823e-05, + "loss": 0.9889, + "step": 18541 + }, + { + "epoch": 1.19, + "learning_rate": 1.6042851928980872e-05, + "loss": 0.9315, + "step": 18542 + }, + { + "epoch": 1.19, + "learning_rate": 1.6042637239957924e-05, + "loss": 0.9604, + "step": 18543 + }, + { + "epoch": 1.19, + "learning_rate": 1.6042422550934973e-05, + "loss": 0.9613, + "step": 18544 + }, + { + "epoch": 1.19, + "learning_rate": 1.6042207861912022e-05, + "loss": 0.9487, + "step": 18545 + }, + { + "epoch": 1.19, + "learning_rate": 1.604199317288907e-05, + "loss": 0.9651, + "step": 18546 + }, + { + "epoch": 1.19, + "learning_rate": 1.604177848386612e-05, + "loss": 0.9069, + "step": 18547 + }, + { + "epoch": 1.19, + "learning_rate": 1.6041563794843173e-05, + "loss": 1.0138, + "step": 18548 + }, + { + "epoch": 1.19, + "learning_rate": 1.604134910582022e-05, + "loss": 0.9822, + "step": 18549 + }, + { + "epoch": 1.19, + "learning_rate": 1.604113441679727e-05, + "loss": 1.0149, + "step": 18550 + }, + { + "epoch": 1.19, + "learning_rate": 1.604091972777432e-05, + "loss": 0.9839, + "step": 18551 + }, + { + "epoch": 1.19, + "learning_rate": 1.604070503875137e-05, + "loss": 0.9561, + "step": 18552 + }, + { + "epoch": 1.19, + "learning_rate": 1.604049034972842e-05, + "loss": 0.9922, + "step": 18553 + }, + { + "epoch": 1.19, + "learning_rate": 1.604027566070547e-05, + "loss": 0.9766, + "step": 18554 + }, + { + "epoch": 1.19, + "learning_rate": 1.604006097168252e-05, + "loss": 0.9553, + "step": 18555 + }, + { + "epoch": 1.19, + "learning_rate": 1.6039846282659568e-05, + "loss": 1.0559, + "step": 18556 + }, + { + "epoch": 1.19, + "learning_rate": 1.6039631593636617e-05, + "loss": 0.9376, + "step": 18557 + }, + { + "epoch": 1.19, + "learning_rate": 1.603941690461367e-05, + "loss": 0.9709, + "step": 18558 + }, + { + "epoch": 1.19, + "learning_rate": 1.6039202215590718e-05, + "loss": 0.9942, + "step": 18559 + }, + { + "epoch": 1.19, + "learning_rate": 1.6038987526567767e-05, + "loss": 0.9607, + "step": 18560 + }, + { + "epoch": 1.19, + "learning_rate": 1.603877283754482e-05, + "loss": 0.9538, + "step": 18561 + }, + { + "epoch": 1.19, + "learning_rate": 1.6038558148521868e-05, + "loss": 0.9742, + "step": 18562 + }, + { + "epoch": 1.19, + "learning_rate": 1.6038343459498917e-05, + "loss": 0.979, + "step": 18563 + }, + { + "epoch": 1.19, + "learning_rate": 1.6038128770475966e-05, + "loss": 0.8729, + "step": 18564 + }, + { + "epoch": 1.19, + "learning_rate": 1.6037914081453015e-05, + "loss": 0.8731, + "step": 18565 + }, + { + "epoch": 1.19, + "learning_rate": 1.6037699392430068e-05, + "loss": 1.026, + "step": 18566 + }, + { + "epoch": 1.19, + "learning_rate": 1.6037484703407117e-05, + "loss": 0.9308, + "step": 18567 + }, + { + "epoch": 1.19, + "learning_rate": 1.6037270014384165e-05, + "loss": 1.0739, + "step": 18568 + }, + { + "epoch": 1.19, + "learning_rate": 1.6037055325361214e-05, + "loss": 0.9797, + "step": 18569 + }, + { + "epoch": 1.19, + "learning_rate": 1.6036840636338263e-05, + "loss": 1.0652, + "step": 18570 + }, + { + "epoch": 1.19, + "learning_rate": 1.6036625947315316e-05, + "loss": 1.0108, + "step": 18571 + }, + { + "epoch": 1.19, + "learning_rate": 1.6036411258292365e-05, + "loss": 0.9333, + "step": 18572 + }, + { + "epoch": 1.19, + "learning_rate": 1.6036196569269417e-05, + "loss": 1.0062, + "step": 18573 + }, + { + "epoch": 1.2, + "learning_rate": 1.6035981880246466e-05, + "loss": 1.0266, + "step": 18574 + }, + { + "epoch": 1.2, + "learning_rate": 1.6035767191223515e-05, + "loss": 1.026, + "step": 18575 + }, + { + "epoch": 1.2, + "learning_rate": 1.6035552502200564e-05, + "loss": 0.962, + "step": 18576 + }, + { + "epoch": 1.2, + "learning_rate": 1.6035337813177613e-05, + "loss": 0.8856, + "step": 18577 + }, + { + "epoch": 1.2, + "learning_rate": 1.6035123124154662e-05, + "loss": 0.9459, + "step": 18578 + }, + { + "epoch": 1.2, + "learning_rate": 1.6034908435131714e-05, + "loss": 0.9562, + "step": 18579 + }, + { + "epoch": 1.2, + "learning_rate": 1.6034693746108763e-05, + "loss": 0.9356, + "step": 18580 + }, + { + "epoch": 1.2, + "learning_rate": 1.6034479057085812e-05, + "loss": 0.9711, + "step": 18581 + }, + { + "epoch": 1.2, + "learning_rate": 1.603426436806286e-05, + "loss": 0.9034, + "step": 18582 + }, + { + "epoch": 1.2, + "learning_rate": 1.603404967903991e-05, + "loss": 1.0444, + "step": 18583 + }, + { + "epoch": 1.2, + "learning_rate": 1.6033834990016962e-05, + "loss": 0.9398, + "step": 18584 + }, + { + "epoch": 1.2, + "learning_rate": 1.603362030099401e-05, + "loss": 0.9493, + "step": 18585 + }, + { + "epoch": 1.2, + "learning_rate": 1.603340561197106e-05, + "loss": 0.9603, + "step": 18586 + }, + { + "epoch": 1.2, + "learning_rate": 1.6033190922948113e-05, + "loss": 0.9964, + "step": 18587 + }, + { + "epoch": 1.2, + "learning_rate": 1.6032976233925162e-05, + "loss": 1.0255, + "step": 18588 + }, + { + "epoch": 1.2, + "learning_rate": 1.603276154490221e-05, + "loss": 0.9716, + "step": 18589 + }, + { + "epoch": 1.2, + "learning_rate": 1.603254685587926e-05, + "loss": 0.9469, + "step": 18590 + }, + { + "epoch": 1.2, + "learning_rate": 1.6032332166856312e-05, + "loss": 0.9731, + "step": 18591 + }, + { + "epoch": 1.2, + "learning_rate": 1.603211747783336e-05, + "loss": 1.0157, + "step": 18592 + }, + { + "epoch": 1.2, + "learning_rate": 1.603190278881041e-05, + "loss": 1.0036, + "step": 18593 + }, + { + "epoch": 1.2, + "learning_rate": 1.603168809978746e-05, + "loss": 1.0107, + "step": 18594 + }, + { + "epoch": 1.2, + "learning_rate": 1.6031473410764508e-05, + "loss": 0.9339, + "step": 18595 + }, + { + "epoch": 1.2, + "learning_rate": 1.6031258721741557e-05, + "loss": 0.9348, + "step": 18596 + }, + { + "epoch": 1.2, + "learning_rate": 1.603104403271861e-05, + "loss": 1.0153, + "step": 18597 + }, + { + "epoch": 1.2, + "learning_rate": 1.6030829343695658e-05, + "loss": 0.9789, + "step": 18598 + }, + { + "epoch": 1.2, + "learning_rate": 1.6030614654672707e-05, + "loss": 1.0382, + "step": 18599 + }, + { + "epoch": 1.2, + "learning_rate": 1.6030399965649756e-05, + "loss": 0.956, + "step": 18600 + }, + { + "epoch": 1.2, + "learning_rate": 1.603018527662681e-05, + "loss": 1.0146, + "step": 18601 + }, + { + "epoch": 1.2, + "learning_rate": 1.6029970587603857e-05, + "loss": 0.9691, + "step": 18602 + }, + { + "epoch": 1.2, + "learning_rate": 1.6029755898580906e-05, + "loss": 0.9411, + "step": 18603 + }, + { + "epoch": 1.2, + "learning_rate": 1.602954120955796e-05, + "loss": 0.9713, + "step": 18604 + }, + { + "epoch": 1.2, + "learning_rate": 1.6029326520535008e-05, + "loss": 0.9558, + "step": 18605 + }, + { + "epoch": 1.2, + "learning_rate": 1.6029111831512057e-05, + "loss": 0.9399, + "step": 18606 + }, + { + "epoch": 1.2, + "learning_rate": 1.6028897142489106e-05, + "loss": 0.9906, + "step": 18607 + }, + { + "epoch": 1.2, + "learning_rate": 1.6028682453466155e-05, + "loss": 1.022, + "step": 18608 + }, + { + "epoch": 1.2, + "learning_rate": 1.6028467764443204e-05, + "loss": 1.0112, + "step": 18609 + }, + { + "epoch": 1.2, + "learning_rate": 1.6028253075420256e-05, + "loss": 1.0302, + "step": 18610 + }, + { + "epoch": 1.2, + "learning_rate": 1.6028038386397305e-05, + "loss": 0.9582, + "step": 18611 + }, + { + "epoch": 1.2, + "learning_rate": 1.6027823697374354e-05, + "loss": 0.9483, + "step": 18612 + }, + { + "epoch": 1.2, + "learning_rate": 1.6027609008351403e-05, + "loss": 0.9759, + "step": 18613 + }, + { + "epoch": 1.2, + "learning_rate": 1.6027394319328452e-05, + "loss": 0.9452, + "step": 18614 + }, + { + "epoch": 1.2, + "learning_rate": 1.6027179630305504e-05, + "loss": 0.9035, + "step": 18615 + }, + { + "epoch": 1.2, + "learning_rate": 1.6026964941282553e-05, + "loss": 0.9897, + "step": 18616 + }, + { + "epoch": 1.2, + "learning_rate": 1.6026750252259605e-05, + "loss": 0.9856, + "step": 18617 + }, + { + "epoch": 1.2, + "learning_rate": 1.6026535563236654e-05, + "loss": 0.9837, + "step": 18618 + }, + { + "epoch": 1.2, + "learning_rate": 1.6026320874213703e-05, + "loss": 1.0181, + "step": 18619 + }, + { + "epoch": 1.2, + "learning_rate": 1.6026106185190752e-05, + "loss": 0.9854, + "step": 18620 + }, + { + "epoch": 1.2, + "learning_rate": 1.60258914961678e-05, + "loss": 0.9691, + "step": 18621 + }, + { + "epoch": 1.2, + "learning_rate": 1.6025676807144854e-05, + "loss": 0.9599, + "step": 18622 + }, + { + "epoch": 1.2, + "learning_rate": 1.6025462118121903e-05, + "loss": 0.9688, + "step": 18623 + }, + { + "epoch": 1.2, + "learning_rate": 1.602524742909895e-05, + "loss": 1.0166, + "step": 18624 + }, + { + "epoch": 1.2, + "learning_rate": 1.6025032740076e-05, + "loss": 1.0851, + "step": 18625 + }, + { + "epoch": 1.2, + "learning_rate": 1.602481805105305e-05, + "loss": 0.9914, + "step": 18626 + }, + { + "epoch": 1.2, + "learning_rate": 1.60246033620301e-05, + "loss": 0.9834, + "step": 18627 + }, + { + "epoch": 1.2, + "learning_rate": 1.602438867300715e-05, + "loss": 0.9983, + "step": 18628 + }, + { + "epoch": 1.2, + "learning_rate": 1.60241739839842e-05, + "loss": 0.9295, + "step": 18629 + }, + { + "epoch": 1.2, + "learning_rate": 1.6023959294961252e-05, + "loss": 1.0158, + "step": 18630 + }, + { + "epoch": 1.2, + "learning_rate": 1.60237446059383e-05, + "loss": 0.9913, + "step": 18631 + }, + { + "epoch": 1.2, + "learning_rate": 1.602352991691535e-05, + "loss": 0.9834, + "step": 18632 + }, + { + "epoch": 1.2, + "learning_rate": 1.60233152278924e-05, + "loss": 0.9516, + "step": 18633 + }, + { + "epoch": 1.2, + "learning_rate": 1.6023100538869448e-05, + "loss": 1.0314, + "step": 18634 + }, + { + "epoch": 1.2, + "learning_rate": 1.60228858498465e-05, + "loss": 1.0174, + "step": 18635 + }, + { + "epoch": 1.2, + "learning_rate": 1.602267116082355e-05, + "loss": 0.9929, + "step": 18636 + }, + { + "epoch": 1.2, + "learning_rate": 1.6022456471800598e-05, + "loss": 1.0289, + "step": 18637 + }, + { + "epoch": 1.2, + "learning_rate": 1.6022241782777647e-05, + "loss": 0.9648, + "step": 18638 + }, + { + "epoch": 1.2, + "learning_rate": 1.6022027093754696e-05, + "loss": 1.0006, + "step": 18639 + }, + { + "epoch": 1.2, + "learning_rate": 1.602181240473175e-05, + "loss": 0.9778, + "step": 18640 + }, + { + "epoch": 1.2, + "learning_rate": 1.6021597715708798e-05, + "loss": 0.9944, + "step": 18641 + }, + { + "epoch": 1.2, + "learning_rate": 1.6021383026685846e-05, + "loss": 0.9962, + "step": 18642 + }, + { + "epoch": 1.2, + "learning_rate": 1.6021168337662895e-05, + "loss": 0.963, + "step": 18643 + }, + { + "epoch": 1.2, + "learning_rate": 1.6020953648639944e-05, + "loss": 1.066, + "step": 18644 + }, + { + "epoch": 1.2, + "learning_rate": 1.6020738959616997e-05, + "loss": 0.9906, + "step": 18645 + }, + { + "epoch": 1.2, + "learning_rate": 1.6020524270594046e-05, + "loss": 1.0135, + "step": 18646 + }, + { + "epoch": 1.2, + "learning_rate": 1.6020309581571095e-05, + "loss": 0.9714, + "step": 18647 + }, + { + "epoch": 1.2, + "learning_rate": 1.6020094892548147e-05, + "loss": 0.9881, + "step": 18648 + }, + { + "epoch": 1.2, + "learning_rate": 1.6019880203525196e-05, + "loss": 1.0201, + "step": 18649 + }, + { + "epoch": 1.2, + "learning_rate": 1.6019665514502245e-05, + "loss": 0.9507, + "step": 18650 + }, + { + "epoch": 1.2, + "learning_rate": 1.6019450825479294e-05, + "loss": 0.9775, + "step": 18651 + }, + { + "epoch": 1.2, + "learning_rate": 1.6019236136456343e-05, + "loss": 1.0086, + "step": 18652 + }, + { + "epoch": 1.2, + "learning_rate": 1.6019021447433395e-05, + "loss": 1.0225, + "step": 18653 + }, + { + "epoch": 1.2, + "learning_rate": 1.6018806758410444e-05, + "loss": 0.9812, + "step": 18654 + }, + { + "epoch": 1.2, + "learning_rate": 1.6018592069387493e-05, + "loss": 0.9674, + "step": 18655 + }, + { + "epoch": 1.2, + "learning_rate": 1.6018377380364542e-05, + "loss": 1.0283, + "step": 18656 + }, + { + "epoch": 1.2, + "learning_rate": 1.601816269134159e-05, + "loss": 0.9291, + "step": 18657 + }, + { + "epoch": 1.2, + "learning_rate": 1.6017948002318643e-05, + "loss": 0.9777, + "step": 18658 + }, + { + "epoch": 1.2, + "learning_rate": 1.6017733313295692e-05, + "loss": 0.9945, + "step": 18659 + }, + { + "epoch": 1.2, + "learning_rate": 1.6017518624272745e-05, + "loss": 1.0706, + "step": 18660 + }, + { + "epoch": 1.2, + "learning_rate": 1.6017303935249794e-05, + "loss": 1.0137, + "step": 18661 + }, + { + "epoch": 1.2, + "learning_rate": 1.6017089246226843e-05, + "loss": 0.9472, + "step": 18662 + }, + { + "epoch": 1.2, + "learning_rate": 1.601687455720389e-05, + "loss": 1.023, + "step": 18663 + }, + { + "epoch": 1.2, + "learning_rate": 1.601665986818094e-05, + "loss": 0.9711, + "step": 18664 + }, + { + "epoch": 1.2, + "learning_rate": 1.601644517915799e-05, + "loss": 0.9585, + "step": 18665 + }, + { + "epoch": 1.2, + "learning_rate": 1.6016230490135042e-05, + "loss": 0.9171, + "step": 18666 + }, + { + "epoch": 1.2, + "learning_rate": 1.601601580111209e-05, + "loss": 0.9772, + "step": 18667 + }, + { + "epoch": 1.2, + "learning_rate": 1.601580111208914e-05, + "loss": 0.9846, + "step": 18668 + }, + { + "epoch": 1.2, + "learning_rate": 1.601558642306619e-05, + "loss": 0.997, + "step": 18669 + }, + { + "epoch": 1.2, + "learning_rate": 1.6015371734043238e-05, + "loss": 0.9465, + "step": 18670 + }, + { + "epoch": 1.2, + "learning_rate": 1.601515704502029e-05, + "loss": 1.0817, + "step": 18671 + }, + { + "epoch": 1.2, + "learning_rate": 1.601494235599734e-05, + "loss": 1.0477, + "step": 18672 + }, + { + "epoch": 1.2, + "learning_rate": 1.6014727666974388e-05, + "loss": 0.9485, + "step": 18673 + }, + { + "epoch": 1.2, + "learning_rate": 1.601451297795144e-05, + "loss": 0.9279, + "step": 18674 + }, + { + "epoch": 1.2, + "learning_rate": 1.601429828892849e-05, + "loss": 0.9907, + "step": 18675 + }, + { + "epoch": 1.2, + "learning_rate": 1.601408359990554e-05, + "loss": 0.9606, + "step": 18676 + }, + { + "epoch": 1.2, + "learning_rate": 1.6013868910882587e-05, + "loss": 0.997, + "step": 18677 + }, + { + "epoch": 1.2, + "learning_rate": 1.601365422185964e-05, + "loss": 0.9977, + "step": 18678 + }, + { + "epoch": 1.2, + "learning_rate": 1.601343953283669e-05, + "loss": 0.9509, + "step": 18679 + }, + { + "epoch": 1.2, + "learning_rate": 1.6013224843813738e-05, + "loss": 0.986, + "step": 18680 + }, + { + "epoch": 1.2, + "learning_rate": 1.6013010154790787e-05, + "loss": 1.0084, + "step": 18681 + }, + { + "epoch": 1.2, + "learning_rate": 1.6012795465767836e-05, + "loss": 0.9962, + "step": 18682 + }, + { + "epoch": 1.2, + "learning_rate": 1.6012580776744885e-05, + "loss": 1.0325, + "step": 18683 + }, + { + "epoch": 1.2, + "learning_rate": 1.6012366087721937e-05, + "loss": 0.9786, + "step": 18684 + }, + { + "epoch": 1.2, + "learning_rate": 1.6012151398698986e-05, + "loss": 0.9731, + "step": 18685 + }, + { + "epoch": 1.2, + "learning_rate": 1.6011936709676035e-05, + "loss": 1.051, + "step": 18686 + }, + { + "epoch": 1.2, + "learning_rate": 1.6011722020653084e-05, + "loss": 0.933, + "step": 18687 + }, + { + "epoch": 1.2, + "learning_rate": 1.6011507331630136e-05, + "loss": 1.0323, + "step": 18688 + }, + { + "epoch": 1.2, + "learning_rate": 1.6011292642607185e-05, + "loss": 0.9996, + "step": 18689 + }, + { + "epoch": 1.2, + "learning_rate": 1.6011077953584234e-05, + "loss": 0.9333, + "step": 18690 + }, + { + "epoch": 1.2, + "learning_rate": 1.6010863264561286e-05, + "loss": 1.0342, + "step": 18691 + }, + { + "epoch": 1.2, + "learning_rate": 1.6010648575538335e-05, + "loss": 1.0396, + "step": 18692 + }, + { + "epoch": 1.2, + "learning_rate": 1.6010433886515384e-05, + "loss": 0.9566, + "step": 18693 + }, + { + "epoch": 1.2, + "learning_rate": 1.6010219197492433e-05, + "loss": 0.9542, + "step": 18694 + }, + { + "epoch": 1.2, + "learning_rate": 1.6010004508469482e-05, + "loss": 0.9811, + "step": 18695 + }, + { + "epoch": 1.2, + "learning_rate": 1.600978981944653e-05, + "loss": 0.9498, + "step": 18696 + }, + { + "epoch": 1.2, + "learning_rate": 1.6009575130423584e-05, + "loss": 0.9044, + "step": 18697 + }, + { + "epoch": 1.2, + "learning_rate": 1.6009360441400633e-05, + "loss": 0.9739, + "step": 18698 + }, + { + "epoch": 1.2, + "learning_rate": 1.600914575237768e-05, + "loss": 1.0499, + "step": 18699 + }, + { + "epoch": 1.2, + "learning_rate": 1.600893106335473e-05, + "loss": 0.9996, + "step": 18700 + }, + { + "epoch": 1.2, + "learning_rate": 1.600871637433178e-05, + "loss": 1.0349, + "step": 18701 + }, + { + "epoch": 1.2, + "learning_rate": 1.6008501685308832e-05, + "loss": 1.0191, + "step": 18702 + }, + { + "epoch": 1.2, + "learning_rate": 1.600828699628588e-05, + "loss": 0.9632, + "step": 18703 + }, + { + "epoch": 1.2, + "learning_rate": 1.6008072307262933e-05, + "loss": 1.069, + "step": 18704 + }, + { + "epoch": 1.2, + "learning_rate": 1.6007857618239982e-05, + "loss": 0.9643, + "step": 18705 + }, + { + "epoch": 1.2, + "learning_rate": 1.600764292921703e-05, + "loss": 1.058, + "step": 18706 + }, + { + "epoch": 1.2, + "learning_rate": 1.600742824019408e-05, + "loss": 1.0122, + "step": 18707 + }, + { + "epoch": 1.2, + "learning_rate": 1.600721355117113e-05, + "loss": 0.9206, + "step": 18708 + }, + { + "epoch": 1.2, + "learning_rate": 1.600699886214818e-05, + "loss": 1.0492, + "step": 18709 + }, + { + "epoch": 1.2, + "learning_rate": 1.600678417312523e-05, + "loss": 1.0609, + "step": 18710 + }, + { + "epoch": 1.2, + "learning_rate": 1.600656948410228e-05, + "loss": 0.9572, + "step": 18711 + }, + { + "epoch": 1.2, + "learning_rate": 1.6006354795079328e-05, + "loss": 1.0486, + "step": 18712 + }, + { + "epoch": 1.2, + "learning_rate": 1.6006140106056377e-05, + "loss": 0.9895, + "step": 18713 + }, + { + "epoch": 1.2, + "learning_rate": 1.6005925417033426e-05, + "loss": 0.9487, + "step": 18714 + }, + { + "epoch": 1.2, + "learning_rate": 1.600571072801048e-05, + "loss": 1.0529, + "step": 18715 + }, + { + "epoch": 1.2, + "learning_rate": 1.6005496038987527e-05, + "loss": 1.0093, + "step": 18716 + }, + { + "epoch": 1.2, + "learning_rate": 1.600528134996458e-05, + "loss": 0.9703, + "step": 18717 + }, + { + "epoch": 1.2, + "learning_rate": 1.600506666094163e-05, + "loss": 1.0062, + "step": 18718 + }, + { + "epoch": 1.2, + "learning_rate": 1.6004851971918678e-05, + "loss": 0.9913, + "step": 18719 + }, + { + "epoch": 1.2, + "learning_rate": 1.6004637282895727e-05, + "loss": 1.0046, + "step": 18720 + }, + { + "epoch": 1.2, + "learning_rate": 1.6004422593872776e-05, + "loss": 1.0003, + "step": 18721 + }, + { + "epoch": 1.2, + "learning_rate": 1.6004207904849828e-05, + "loss": 0.914, + "step": 18722 + }, + { + "epoch": 1.2, + "learning_rate": 1.6003993215826877e-05, + "loss": 0.9341, + "step": 18723 + }, + { + "epoch": 1.2, + "learning_rate": 1.6003778526803926e-05, + "loss": 0.9206, + "step": 18724 + }, + { + "epoch": 1.2, + "learning_rate": 1.6003563837780975e-05, + "loss": 1.0346, + "step": 18725 + }, + { + "epoch": 1.2, + "learning_rate": 1.6003349148758024e-05, + "loss": 0.9775, + "step": 18726 + }, + { + "epoch": 1.2, + "learning_rate": 1.6003134459735076e-05, + "loss": 1.0339, + "step": 18727 + }, + { + "epoch": 1.2, + "learning_rate": 1.6002919770712125e-05, + "loss": 1.0704, + "step": 18728 + }, + { + "epoch": 1.2, + "learning_rate": 1.6002705081689174e-05, + "loss": 1.0357, + "step": 18729 + }, + { + "epoch": 1.21, + "learning_rate": 1.6002490392666223e-05, + "loss": 1.0362, + "step": 18730 + }, + { + "epoch": 1.21, + "learning_rate": 1.6002275703643272e-05, + "loss": 0.9655, + "step": 18731 + }, + { + "epoch": 1.21, + "learning_rate": 1.6002061014620324e-05, + "loss": 1.0294, + "step": 18732 + }, + { + "epoch": 1.21, + "learning_rate": 1.6001846325597373e-05, + "loss": 0.9613, + "step": 18733 + }, + { + "epoch": 1.21, + "learning_rate": 1.6001631636574422e-05, + "loss": 1.0044, + "step": 18734 + }, + { + "epoch": 1.21, + "learning_rate": 1.6001416947551475e-05, + "loss": 0.9254, + "step": 18735 + }, + { + "epoch": 1.21, + "learning_rate": 1.6001202258528524e-05, + "loss": 0.9764, + "step": 18736 + }, + { + "epoch": 1.21, + "learning_rate": 1.6000987569505573e-05, + "loss": 0.9227, + "step": 18737 + }, + { + "epoch": 1.21, + "learning_rate": 1.600077288048262e-05, + "loss": 1.1012, + "step": 18738 + }, + { + "epoch": 1.21, + "learning_rate": 1.600055819145967e-05, + "loss": 0.9789, + "step": 18739 + }, + { + "epoch": 1.21, + "learning_rate": 1.6000343502436723e-05, + "loss": 1.0283, + "step": 18740 + }, + { + "epoch": 1.21, + "learning_rate": 1.6000128813413772e-05, + "loss": 0.9554, + "step": 18741 + }, + { + "epoch": 1.21, + "learning_rate": 1.599991412439082e-05, + "loss": 0.9525, + "step": 18742 + }, + { + "epoch": 1.21, + "learning_rate": 1.599969943536787e-05, + "loss": 0.9534, + "step": 18743 + }, + { + "epoch": 1.21, + "learning_rate": 1.599948474634492e-05, + "loss": 0.9202, + "step": 18744 + }, + { + "epoch": 1.21, + "learning_rate": 1.5999270057321968e-05, + "loss": 0.9905, + "step": 18745 + }, + { + "epoch": 1.21, + "learning_rate": 1.599905536829902e-05, + "loss": 0.9648, + "step": 18746 + }, + { + "epoch": 1.21, + "learning_rate": 1.5998840679276072e-05, + "loss": 0.9846, + "step": 18747 + }, + { + "epoch": 1.21, + "learning_rate": 1.599862599025312e-05, + "loss": 0.9519, + "step": 18748 + }, + { + "epoch": 1.21, + "learning_rate": 1.599841130123017e-05, + "loss": 0.9568, + "step": 18749 + }, + { + "epoch": 1.21, + "learning_rate": 1.599819661220722e-05, + "loss": 0.9875, + "step": 18750 + }, + { + "epoch": 1.21, + "learning_rate": 1.599798192318427e-05, + "loss": 0.9797, + "step": 18751 + }, + { + "epoch": 1.21, + "learning_rate": 1.5997767234161317e-05, + "loss": 0.9514, + "step": 18752 + }, + { + "epoch": 1.21, + "learning_rate": 1.599755254513837e-05, + "loss": 0.9799, + "step": 18753 + }, + { + "epoch": 1.21, + "learning_rate": 1.599733785611542e-05, + "loss": 1.0035, + "step": 18754 + }, + { + "epoch": 1.21, + "learning_rate": 1.5997123167092468e-05, + "loss": 1.0021, + "step": 18755 + }, + { + "epoch": 1.21, + "learning_rate": 1.5996908478069517e-05, + "loss": 0.9947, + "step": 18756 + }, + { + "epoch": 1.21, + "learning_rate": 1.5996693789046566e-05, + "loss": 0.9812, + "step": 18757 + }, + { + "epoch": 1.21, + "learning_rate": 1.5996479100023618e-05, + "loss": 0.9784, + "step": 18758 + }, + { + "epoch": 1.21, + "learning_rate": 1.5996264411000667e-05, + "loss": 0.9842, + "step": 18759 + }, + { + "epoch": 1.21, + "learning_rate": 1.5996049721977716e-05, + "loss": 0.9774, + "step": 18760 + }, + { + "epoch": 1.21, + "learning_rate": 1.5995835032954768e-05, + "loss": 1.009, + "step": 18761 + }, + { + "epoch": 1.21, + "learning_rate": 1.5995620343931817e-05, + "loss": 0.9923, + "step": 18762 + }, + { + "epoch": 1.21, + "learning_rate": 1.5995405654908866e-05, + "loss": 1.0058, + "step": 18763 + }, + { + "epoch": 1.21, + "learning_rate": 1.5995190965885915e-05, + "loss": 0.9969, + "step": 18764 + }, + { + "epoch": 1.21, + "learning_rate": 1.5994976276862967e-05, + "loss": 0.9746, + "step": 18765 + }, + { + "epoch": 1.21, + "learning_rate": 1.5994761587840016e-05, + "loss": 0.9345, + "step": 18766 + }, + { + "epoch": 1.21, + "learning_rate": 1.5994546898817065e-05, + "loss": 0.9566, + "step": 18767 + }, + { + "epoch": 1.21, + "learning_rate": 1.5994332209794114e-05, + "loss": 0.8749, + "step": 18768 + }, + { + "epoch": 1.21, + "learning_rate": 1.5994117520771163e-05, + "loss": 1.0354, + "step": 18769 + }, + { + "epoch": 1.21, + "learning_rate": 1.5993902831748212e-05, + "loss": 0.9537, + "step": 18770 + }, + { + "epoch": 1.21, + "learning_rate": 1.5993688142725265e-05, + "loss": 0.9686, + "step": 18771 + }, + { + "epoch": 1.21, + "learning_rate": 1.5993473453702314e-05, + "loss": 0.932, + "step": 18772 + }, + { + "epoch": 1.21, + "learning_rate": 1.5993258764679363e-05, + "loss": 1.1006, + "step": 18773 + }, + { + "epoch": 1.21, + "learning_rate": 1.599304407565641e-05, + "loss": 0.993, + "step": 18774 + }, + { + "epoch": 1.21, + "learning_rate": 1.5992829386633464e-05, + "loss": 0.9211, + "step": 18775 + }, + { + "epoch": 1.21, + "learning_rate": 1.5992614697610513e-05, + "loss": 0.9893, + "step": 18776 + }, + { + "epoch": 1.21, + "learning_rate": 1.5992400008587562e-05, + "loss": 0.9644, + "step": 18777 + }, + { + "epoch": 1.21, + "learning_rate": 1.5992185319564614e-05, + "loss": 0.921, + "step": 18778 + }, + { + "epoch": 1.21, + "learning_rate": 1.5991970630541663e-05, + "loss": 0.9571, + "step": 18779 + }, + { + "epoch": 1.21, + "learning_rate": 1.5991755941518712e-05, + "loss": 0.9891, + "step": 18780 + }, + { + "epoch": 1.21, + "learning_rate": 1.599154125249576e-05, + "loss": 0.9991, + "step": 18781 + }, + { + "epoch": 1.21, + "learning_rate": 1.599132656347281e-05, + "loss": 0.9739, + "step": 18782 + }, + { + "epoch": 1.21, + "learning_rate": 1.599111187444986e-05, + "loss": 1.0512, + "step": 18783 + }, + { + "epoch": 1.21, + "learning_rate": 1.599089718542691e-05, + "loss": 0.9625, + "step": 18784 + }, + { + "epoch": 1.21, + "learning_rate": 1.599068249640396e-05, + "loss": 0.9974, + "step": 18785 + }, + { + "epoch": 1.21, + "learning_rate": 1.599046780738101e-05, + "loss": 0.987, + "step": 18786 + }, + { + "epoch": 1.21, + "learning_rate": 1.5990253118358058e-05, + "loss": 0.9888, + "step": 18787 + }, + { + "epoch": 1.21, + "learning_rate": 1.5990038429335107e-05, + "loss": 0.9514, + "step": 18788 + }, + { + "epoch": 1.21, + "learning_rate": 1.598982374031216e-05, + "loss": 1.0165, + "step": 18789 + }, + { + "epoch": 1.21, + "learning_rate": 1.598960905128921e-05, + "loss": 0.9394, + "step": 18790 + }, + { + "epoch": 1.21, + "learning_rate": 1.598939436226626e-05, + "loss": 0.959, + "step": 18791 + }, + { + "epoch": 1.21, + "learning_rate": 1.598917967324331e-05, + "loss": 0.9669, + "step": 18792 + }, + { + "epoch": 1.21, + "learning_rate": 1.598896498422036e-05, + "loss": 1.0223, + "step": 18793 + }, + { + "epoch": 1.21, + "learning_rate": 1.5988750295197408e-05, + "loss": 1.0309, + "step": 18794 + }, + { + "epoch": 1.21, + "learning_rate": 1.5988535606174457e-05, + "loss": 1.0032, + "step": 18795 + }, + { + "epoch": 1.21, + "learning_rate": 1.598832091715151e-05, + "loss": 1.044, + "step": 18796 + }, + { + "epoch": 1.21, + "learning_rate": 1.5988106228128558e-05, + "loss": 0.9998, + "step": 18797 + }, + { + "epoch": 1.21, + "learning_rate": 1.5987891539105607e-05, + "loss": 0.909, + "step": 18798 + }, + { + "epoch": 1.21, + "learning_rate": 1.5987676850082656e-05, + "loss": 1.0051, + "step": 18799 + }, + { + "epoch": 1.21, + "learning_rate": 1.5987462161059705e-05, + "loss": 0.996, + "step": 18800 + }, + { + "epoch": 1.21, + "learning_rate": 1.5987247472036754e-05, + "loss": 0.9993, + "step": 18801 + }, + { + "epoch": 1.21, + "learning_rate": 1.5987032783013806e-05, + "loss": 0.9533, + "step": 18802 + }, + { + "epoch": 1.21, + "learning_rate": 1.5986818093990855e-05, + "loss": 0.9602, + "step": 18803 + }, + { + "epoch": 1.21, + "learning_rate": 1.5986603404967904e-05, + "loss": 0.9299, + "step": 18804 + }, + { + "epoch": 1.21, + "learning_rate": 1.5986388715944957e-05, + "loss": 1.0164, + "step": 18805 + }, + { + "epoch": 1.21, + "learning_rate": 1.5986174026922005e-05, + "loss": 0.9041, + "step": 18806 + }, + { + "epoch": 1.21, + "learning_rate": 1.5985959337899054e-05, + "loss": 0.9768, + "step": 18807 + }, + { + "epoch": 1.21, + "learning_rate": 1.5985744648876103e-05, + "loss": 0.9887, + "step": 18808 + }, + { + "epoch": 1.21, + "learning_rate": 1.5985529959853156e-05, + "loss": 0.9736, + "step": 18809 + }, + { + "epoch": 1.21, + "learning_rate": 1.5985315270830205e-05, + "loss": 1.011, + "step": 18810 + }, + { + "epoch": 1.21, + "learning_rate": 1.5985100581807254e-05, + "loss": 0.97, + "step": 18811 + }, + { + "epoch": 1.21, + "learning_rate": 1.5984885892784303e-05, + "loss": 0.9499, + "step": 18812 + }, + { + "epoch": 1.21, + "learning_rate": 1.598467120376135e-05, + "loss": 0.9873, + "step": 18813 + }, + { + "epoch": 1.21, + "learning_rate": 1.5984456514738404e-05, + "loss": 1.0939, + "step": 18814 + }, + { + "epoch": 1.21, + "learning_rate": 1.5984241825715453e-05, + "loss": 0.9982, + "step": 18815 + }, + { + "epoch": 1.21, + "learning_rate": 1.5984027136692502e-05, + "loss": 1.0001, + "step": 18816 + }, + { + "epoch": 1.21, + "learning_rate": 1.598381244766955e-05, + "loss": 0.9368, + "step": 18817 + }, + { + "epoch": 1.21, + "learning_rate": 1.59835977586466e-05, + "loss": 1.0302, + "step": 18818 + }, + { + "epoch": 1.21, + "learning_rate": 1.5983383069623652e-05, + "loss": 1.0358, + "step": 18819 + }, + { + "epoch": 1.21, + "learning_rate": 1.59831683806007e-05, + "loss": 0.9756, + "step": 18820 + }, + { + "epoch": 1.21, + "learning_rate": 1.598295369157775e-05, + "loss": 1.093, + "step": 18821 + }, + { + "epoch": 1.21, + "learning_rate": 1.5982739002554802e-05, + "loss": 1.0125, + "step": 18822 + }, + { + "epoch": 1.21, + "learning_rate": 1.598252431353185e-05, + "loss": 0.9469, + "step": 18823 + }, + { + "epoch": 1.21, + "learning_rate": 1.59823096245089e-05, + "loss": 0.912, + "step": 18824 + }, + { + "epoch": 1.21, + "learning_rate": 1.598209493548595e-05, + "loss": 0.9294, + "step": 18825 + }, + { + "epoch": 1.21, + "learning_rate": 1.5981880246463e-05, + "loss": 0.968, + "step": 18826 + }, + { + "epoch": 1.21, + "learning_rate": 1.598166555744005e-05, + "loss": 0.9813, + "step": 18827 + }, + { + "epoch": 1.21, + "learning_rate": 1.59814508684171e-05, + "loss": 1.0142, + "step": 18828 + }, + { + "epoch": 1.21, + "learning_rate": 1.598123617939415e-05, + "loss": 0.9901, + "step": 18829 + }, + { + "epoch": 1.21, + "learning_rate": 1.5981021490371198e-05, + "loss": 0.9152, + "step": 18830 + }, + { + "epoch": 1.21, + "learning_rate": 1.5980806801348247e-05, + "loss": 0.9896, + "step": 18831 + }, + { + "epoch": 1.21, + "learning_rate": 1.5980592112325296e-05, + "loss": 1.0298, + "step": 18832 + }, + { + "epoch": 1.21, + "learning_rate": 1.5980377423302348e-05, + "loss": 1.012, + "step": 18833 + }, + { + "epoch": 1.21, + "learning_rate": 1.59801627342794e-05, + "loss": 0.9562, + "step": 18834 + }, + { + "epoch": 1.21, + "learning_rate": 1.597994804525645e-05, + "loss": 0.9722, + "step": 18835 + }, + { + "epoch": 1.21, + "learning_rate": 1.5979733356233498e-05, + "loss": 0.9965, + "step": 18836 + }, + { + "epoch": 1.21, + "learning_rate": 1.5979518667210547e-05, + "loss": 1.0296, + "step": 18837 + }, + { + "epoch": 1.21, + "learning_rate": 1.5979303978187596e-05, + "loss": 0.9096, + "step": 18838 + }, + { + "epoch": 1.21, + "learning_rate": 1.5979089289164645e-05, + "loss": 0.9335, + "step": 18839 + }, + { + "epoch": 1.21, + "learning_rate": 1.5978874600141697e-05, + "loss": 1.0594, + "step": 18840 + }, + { + "epoch": 1.21, + "learning_rate": 1.5978659911118746e-05, + "loss": 1.0882, + "step": 18841 + }, + { + "epoch": 1.21, + "learning_rate": 1.5978445222095795e-05, + "loss": 0.9358, + "step": 18842 + }, + { + "epoch": 1.21, + "learning_rate": 1.5978230533072844e-05, + "loss": 0.9748, + "step": 18843 + }, + { + "epoch": 1.21, + "learning_rate": 1.5978015844049893e-05, + "loss": 0.9404, + "step": 18844 + }, + { + "epoch": 1.21, + "learning_rate": 1.5977801155026946e-05, + "loss": 1.0024, + "step": 18845 + }, + { + "epoch": 1.21, + "learning_rate": 1.5977586466003995e-05, + "loss": 0.9614, + "step": 18846 + }, + { + "epoch": 1.21, + "learning_rate": 1.5977371776981044e-05, + "loss": 1.0556, + "step": 18847 + }, + { + "epoch": 1.21, + "learning_rate": 1.5977157087958096e-05, + "loss": 0.9288, + "step": 18848 + }, + { + "epoch": 1.21, + "learning_rate": 1.5976942398935145e-05, + "loss": 0.917, + "step": 18849 + }, + { + "epoch": 1.21, + "learning_rate": 1.5976727709912194e-05, + "loss": 0.9636, + "step": 18850 + }, + { + "epoch": 1.21, + "learning_rate": 1.5976513020889243e-05, + "loss": 0.993, + "step": 18851 + }, + { + "epoch": 1.21, + "learning_rate": 1.5976298331866292e-05, + "loss": 0.9334, + "step": 18852 + }, + { + "epoch": 1.21, + "learning_rate": 1.5976083642843344e-05, + "loss": 0.9149, + "step": 18853 + }, + { + "epoch": 1.21, + "learning_rate": 1.5975868953820393e-05, + "loss": 1.0144, + "step": 18854 + }, + { + "epoch": 1.21, + "learning_rate": 1.5975654264797442e-05, + "loss": 0.9526, + "step": 18855 + }, + { + "epoch": 1.21, + "learning_rate": 1.597543957577449e-05, + "loss": 0.9219, + "step": 18856 + }, + { + "epoch": 1.21, + "learning_rate": 1.597522488675154e-05, + "loss": 1.0451, + "step": 18857 + }, + { + "epoch": 1.21, + "learning_rate": 1.5975010197728592e-05, + "loss": 1.0181, + "step": 18858 + }, + { + "epoch": 1.21, + "learning_rate": 1.597479550870564e-05, + "loss": 1.0986, + "step": 18859 + }, + { + "epoch": 1.21, + "learning_rate": 1.597458081968269e-05, + "loss": 0.9816, + "step": 18860 + }, + { + "epoch": 1.21, + "learning_rate": 1.597436613065974e-05, + "loss": 0.944, + "step": 18861 + }, + { + "epoch": 1.21, + "learning_rate": 1.5974151441636788e-05, + "loss": 0.9367, + "step": 18862 + }, + { + "epoch": 1.21, + "learning_rate": 1.597393675261384e-05, + "loss": 0.9582, + "step": 18863 + }, + { + "epoch": 1.21, + "learning_rate": 1.597372206359089e-05, + "loss": 0.9974, + "step": 18864 + }, + { + "epoch": 1.21, + "learning_rate": 1.5973507374567942e-05, + "loss": 0.8928, + "step": 18865 + }, + { + "epoch": 1.21, + "learning_rate": 1.597329268554499e-05, + "loss": 0.9656, + "step": 18866 + }, + { + "epoch": 1.21, + "learning_rate": 1.597307799652204e-05, + "loss": 1.0031, + "step": 18867 + }, + { + "epoch": 1.21, + "learning_rate": 1.597286330749909e-05, + "loss": 1.0506, + "step": 18868 + }, + { + "epoch": 1.21, + "learning_rate": 1.5972648618476138e-05, + "loss": 1.0169, + "step": 18869 + }, + { + "epoch": 1.21, + "learning_rate": 1.5972433929453187e-05, + "loss": 0.9942, + "step": 18870 + }, + { + "epoch": 1.21, + "learning_rate": 1.597221924043024e-05, + "loss": 0.9377, + "step": 18871 + }, + { + "epoch": 1.21, + "learning_rate": 1.5972004551407288e-05, + "loss": 1.0072, + "step": 18872 + }, + { + "epoch": 1.21, + "learning_rate": 1.5971789862384337e-05, + "loss": 1.0003, + "step": 18873 + }, + { + "epoch": 1.21, + "learning_rate": 1.5971575173361386e-05, + "loss": 0.9734, + "step": 18874 + }, + { + "epoch": 1.21, + "learning_rate": 1.5971360484338435e-05, + "loss": 1.0191, + "step": 18875 + }, + { + "epoch": 1.21, + "learning_rate": 1.5971145795315487e-05, + "loss": 1.0535, + "step": 18876 + }, + { + "epoch": 1.21, + "learning_rate": 1.5970931106292536e-05, + "loss": 0.9354, + "step": 18877 + }, + { + "epoch": 1.21, + "learning_rate": 1.597071641726959e-05, + "loss": 1.0002, + "step": 18878 + }, + { + "epoch": 1.21, + "learning_rate": 1.5970501728246638e-05, + "loss": 0.9541, + "step": 18879 + }, + { + "epoch": 1.21, + "learning_rate": 1.5970287039223686e-05, + "loss": 0.9899, + "step": 18880 + }, + { + "epoch": 1.21, + "learning_rate": 1.5970072350200735e-05, + "loss": 0.9535, + "step": 18881 + }, + { + "epoch": 1.21, + "learning_rate": 1.5969857661177784e-05, + "loss": 0.9966, + "step": 18882 + }, + { + "epoch": 1.21, + "learning_rate": 1.5969642972154837e-05, + "loss": 0.9402, + "step": 18883 + }, + { + "epoch": 1.21, + "learning_rate": 1.5969428283131886e-05, + "loss": 1.0167, + "step": 18884 + }, + { + "epoch": 1.22, + "learning_rate": 1.5969213594108935e-05, + "loss": 0.9673, + "step": 18885 + }, + { + "epoch": 1.22, + "learning_rate": 1.5968998905085984e-05, + "loss": 1.0268, + "step": 18886 + }, + { + "epoch": 1.22, + "learning_rate": 1.5968784216063033e-05, + "loss": 0.9807, + "step": 18887 + }, + { + "epoch": 1.22, + "learning_rate": 1.596856952704008e-05, + "loss": 0.9659, + "step": 18888 + }, + { + "epoch": 1.22, + "learning_rate": 1.5968354838017134e-05, + "loss": 0.8528, + "step": 18889 + }, + { + "epoch": 1.22, + "learning_rate": 1.5968140148994183e-05, + "loss": 0.946, + "step": 18890 + }, + { + "epoch": 1.22, + "learning_rate": 1.5967925459971232e-05, + "loss": 0.9755, + "step": 18891 + }, + { + "epoch": 1.22, + "learning_rate": 1.5967710770948284e-05, + "loss": 0.9478, + "step": 18892 + }, + { + "epoch": 1.22, + "learning_rate": 1.5967496081925333e-05, + "loss": 0.9407, + "step": 18893 + }, + { + "epoch": 1.22, + "learning_rate": 1.5967281392902382e-05, + "loss": 0.9409, + "step": 18894 + }, + { + "epoch": 1.22, + "learning_rate": 1.596706670387943e-05, + "loss": 1.02, + "step": 18895 + }, + { + "epoch": 1.22, + "learning_rate": 1.5966852014856483e-05, + "loss": 0.9992, + "step": 18896 + }, + { + "epoch": 1.22, + "learning_rate": 1.5966637325833532e-05, + "loss": 1.0067, + "step": 18897 + }, + { + "epoch": 1.22, + "learning_rate": 1.596642263681058e-05, + "loss": 0.9783, + "step": 18898 + }, + { + "epoch": 1.22, + "learning_rate": 1.596620794778763e-05, + "loss": 1.0027, + "step": 18899 + }, + { + "epoch": 1.22, + "learning_rate": 1.596599325876468e-05, + "loss": 0.9635, + "step": 18900 + }, + { + "epoch": 1.22, + "learning_rate": 1.596577856974173e-05, + "loss": 0.9374, + "step": 18901 + }, + { + "epoch": 1.22, + "learning_rate": 1.596556388071878e-05, + "loss": 1.0213, + "step": 18902 + }, + { + "epoch": 1.22, + "learning_rate": 1.596534919169583e-05, + "loss": 1.0159, + "step": 18903 + }, + { + "epoch": 1.22, + "learning_rate": 1.596513450267288e-05, + "loss": 0.8686, + "step": 18904 + }, + { + "epoch": 1.22, + "learning_rate": 1.5964919813649928e-05, + "loss": 0.9427, + "step": 18905 + }, + { + "epoch": 1.22, + "learning_rate": 1.596470512462698e-05, + "loss": 0.9924, + "step": 18906 + }, + { + "epoch": 1.22, + "learning_rate": 1.596449043560403e-05, + "loss": 0.9121, + "step": 18907 + }, + { + "epoch": 1.22, + "learning_rate": 1.5964275746581078e-05, + "loss": 0.9155, + "step": 18908 + }, + { + "epoch": 1.22, + "learning_rate": 1.596406105755813e-05, + "loss": 0.9054, + "step": 18909 + }, + { + "epoch": 1.22, + "learning_rate": 1.596384636853518e-05, + "loss": 0.9851, + "step": 18910 + }, + { + "epoch": 1.22, + "learning_rate": 1.5963631679512228e-05, + "loss": 1.0334, + "step": 18911 + }, + { + "epoch": 1.22, + "learning_rate": 1.5963416990489277e-05, + "loss": 0.9575, + "step": 18912 + }, + { + "epoch": 1.22, + "learning_rate": 1.5963202301466326e-05, + "loss": 1.0295, + "step": 18913 + }, + { + "epoch": 1.22, + "learning_rate": 1.596298761244338e-05, + "loss": 0.9583, + "step": 18914 + }, + { + "epoch": 1.22, + "learning_rate": 1.5962772923420427e-05, + "loss": 0.892, + "step": 18915 + }, + { + "epoch": 1.22, + "learning_rate": 1.5962558234397476e-05, + "loss": 0.9391, + "step": 18916 + }, + { + "epoch": 1.22, + "learning_rate": 1.5962343545374525e-05, + "loss": 1.0086, + "step": 18917 + }, + { + "epoch": 1.22, + "learning_rate": 1.5962128856351574e-05, + "loss": 1.0013, + "step": 18918 + }, + { + "epoch": 1.22, + "learning_rate": 1.5961914167328623e-05, + "loss": 0.9834, + "step": 18919 + }, + { + "epoch": 1.22, + "learning_rate": 1.5961699478305676e-05, + "loss": 1.0032, + "step": 18920 + }, + { + "epoch": 1.22, + "learning_rate": 1.5961484789282728e-05, + "loss": 0.9334, + "step": 18921 + }, + { + "epoch": 1.22, + "learning_rate": 1.5961270100259777e-05, + "loss": 0.9455, + "step": 18922 + }, + { + "epoch": 1.22, + "learning_rate": 1.5961055411236826e-05, + "loss": 0.9896, + "step": 18923 + }, + { + "epoch": 1.22, + "learning_rate": 1.5960840722213875e-05, + "loss": 0.9644, + "step": 18924 + }, + { + "epoch": 1.22, + "learning_rate": 1.5960626033190924e-05, + "loss": 1.0267, + "step": 18925 + }, + { + "epoch": 1.22, + "learning_rate": 1.5960411344167973e-05, + "loss": 1.0185, + "step": 18926 + }, + { + "epoch": 1.22, + "learning_rate": 1.5960196655145025e-05, + "loss": 0.9453, + "step": 18927 + }, + { + "epoch": 1.22, + "learning_rate": 1.5959981966122074e-05, + "loss": 1.036, + "step": 18928 + }, + { + "epoch": 1.22, + "learning_rate": 1.5959767277099123e-05, + "loss": 1.0148, + "step": 18929 + }, + { + "epoch": 1.22, + "learning_rate": 1.5959552588076172e-05, + "loss": 0.9508, + "step": 18930 + }, + { + "epoch": 1.22, + "learning_rate": 1.595933789905322e-05, + "loss": 1.0357, + "step": 18931 + }, + { + "epoch": 1.22, + "learning_rate": 1.5959123210030273e-05, + "loss": 0.9629, + "step": 18932 + }, + { + "epoch": 1.22, + "learning_rate": 1.5958908521007322e-05, + "loss": 1.0206, + "step": 18933 + }, + { + "epoch": 1.22, + "learning_rate": 1.595869383198437e-05, + "loss": 1.0033, + "step": 18934 + }, + { + "epoch": 1.22, + "learning_rate": 1.5958479142961424e-05, + "loss": 0.9146, + "step": 18935 + }, + { + "epoch": 1.22, + "learning_rate": 1.5958264453938473e-05, + "loss": 0.9208, + "step": 18936 + }, + { + "epoch": 1.22, + "learning_rate": 1.595804976491552e-05, + "loss": 1.0233, + "step": 18937 + }, + { + "epoch": 1.22, + "learning_rate": 1.595783507589257e-05, + "loss": 1.0575, + "step": 18938 + }, + { + "epoch": 1.22, + "learning_rate": 1.595762038686962e-05, + "loss": 0.9425, + "step": 18939 + }, + { + "epoch": 1.22, + "learning_rate": 1.5957405697846672e-05, + "loss": 1.0332, + "step": 18940 + }, + { + "epoch": 1.22, + "learning_rate": 1.595719100882372e-05, + "loss": 0.9931, + "step": 18941 + }, + { + "epoch": 1.22, + "learning_rate": 1.595697631980077e-05, + "loss": 0.9478, + "step": 18942 + }, + { + "epoch": 1.22, + "learning_rate": 1.595676163077782e-05, + "loss": 0.95, + "step": 18943 + }, + { + "epoch": 1.22, + "learning_rate": 1.5956546941754868e-05, + "loss": 0.9674, + "step": 18944 + }, + { + "epoch": 1.22, + "learning_rate": 1.595633225273192e-05, + "loss": 1.0252, + "step": 18945 + }, + { + "epoch": 1.22, + "learning_rate": 1.595611756370897e-05, + "loss": 0.978, + "step": 18946 + }, + { + "epoch": 1.22, + "learning_rate": 1.5955902874686018e-05, + "loss": 0.9515, + "step": 18947 + }, + { + "epoch": 1.22, + "learning_rate": 1.5955688185663067e-05, + "loss": 0.921, + "step": 18948 + }, + { + "epoch": 1.22, + "learning_rate": 1.5955473496640116e-05, + "loss": 0.9133, + "step": 18949 + }, + { + "epoch": 1.22, + "learning_rate": 1.5955258807617168e-05, + "loss": 0.9298, + "step": 18950 + }, + { + "epoch": 1.22, + "learning_rate": 1.5955044118594217e-05, + "loss": 0.957, + "step": 18951 + }, + { + "epoch": 1.22, + "learning_rate": 1.595482942957127e-05, + "loss": 0.9384, + "step": 18952 + }, + { + "epoch": 1.22, + "learning_rate": 1.595461474054832e-05, + "loss": 1.0002, + "step": 18953 + }, + { + "epoch": 1.22, + "learning_rate": 1.5954400051525367e-05, + "loss": 1.0324, + "step": 18954 + }, + { + "epoch": 1.22, + "learning_rate": 1.5954185362502416e-05, + "loss": 1.001, + "step": 18955 + }, + { + "epoch": 1.22, + "learning_rate": 1.5953970673479465e-05, + "loss": 0.9889, + "step": 18956 + }, + { + "epoch": 1.22, + "learning_rate": 1.5953755984456514e-05, + "loss": 0.9739, + "step": 18957 + }, + { + "epoch": 1.22, + "learning_rate": 1.5953541295433567e-05, + "loss": 0.9199, + "step": 18958 + }, + { + "epoch": 1.22, + "learning_rate": 1.5953326606410616e-05, + "loss": 1.0062, + "step": 18959 + }, + { + "epoch": 1.22, + "learning_rate": 1.5953111917387665e-05, + "loss": 0.9764, + "step": 18960 + }, + { + "epoch": 1.22, + "learning_rate": 1.5952897228364714e-05, + "loss": 1.0113, + "step": 18961 + }, + { + "epoch": 1.22, + "learning_rate": 1.5952682539341763e-05, + "loss": 1.0038, + "step": 18962 + }, + { + "epoch": 1.22, + "learning_rate": 1.5952467850318815e-05, + "loss": 0.9205, + "step": 18963 + }, + { + "epoch": 1.22, + "learning_rate": 1.5952253161295864e-05, + "loss": 0.9574, + "step": 18964 + }, + { + "epoch": 1.22, + "learning_rate": 1.5952038472272916e-05, + "loss": 0.9867, + "step": 18965 + }, + { + "epoch": 1.22, + "learning_rate": 1.5951823783249965e-05, + "loss": 1.0454, + "step": 18966 + }, + { + "epoch": 1.22, + "learning_rate": 1.5951609094227014e-05, + "loss": 1.015, + "step": 18967 + }, + { + "epoch": 1.22, + "learning_rate": 1.5951394405204063e-05, + "loss": 0.9054, + "step": 18968 + }, + { + "epoch": 1.22, + "learning_rate": 1.5951179716181112e-05, + "loss": 0.9413, + "step": 18969 + }, + { + "epoch": 1.22, + "learning_rate": 1.5950965027158164e-05, + "loss": 0.9732, + "step": 18970 + }, + { + "epoch": 1.22, + "learning_rate": 1.5950750338135213e-05, + "loss": 0.9258, + "step": 18971 + }, + { + "epoch": 1.22, + "learning_rate": 1.5950535649112262e-05, + "loss": 0.97, + "step": 18972 + }, + { + "epoch": 1.22, + "learning_rate": 1.595032096008931e-05, + "loss": 0.994, + "step": 18973 + }, + { + "epoch": 1.22, + "learning_rate": 1.595010627106636e-05, + "loss": 1.077, + "step": 18974 + }, + { + "epoch": 1.22, + "learning_rate": 1.594989158204341e-05, + "loss": 0.991, + "step": 18975 + }, + { + "epoch": 1.22, + "learning_rate": 1.594967689302046e-05, + "loss": 0.8934, + "step": 18976 + }, + { + "epoch": 1.22, + "learning_rate": 1.594946220399751e-05, + "loss": 1.0031, + "step": 18977 + }, + { + "epoch": 1.22, + "learning_rate": 1.594924751497456e-05, + "loss": 0.9222, + "step": 18978 + }, + { + "epoch": 1.22, + "learning_rate": 1.5949032825951612e-05, + "loss": 1.0184, + "step": 18979 + }, + { + "epoch": 1.22, + "learning_rate": 1.594881813692866e-05, + "loss": 1.0201, + "step": 18980 + }, + { + "epoch": 1.22, + "learning_rate": 1.594860344790571e-05, + "loss": 0.9966, + "step": 18981 + }, + { + "epoch": 1.22, + "learning_rate": 1.594838875888276e-05, + "loss": 0.9517, + "step": 18982 + }, + { + "epoch": 1.22, + "learning_rate": 1.594817406985981e-05, + "loss": 0.9374, + "step": 18983 + }, + { + "epoch": 1.22, + "learning_rate": 1.594795938083686e-05, + "loss": 0.9883, + "step": 18984 + }, + { + "epoch": 1.22, + "learning_rate": 1.594774469181391e-05, + "loss": 1.017, + "step": 18985 + }, + { + "epoch": 1.22, + "learning_rate": 1.5947530002790958e-05, + "loss": 0.899, + "step": 18986 + }, + { + "epoch": 1.22, + "learning_rate": 1.5947315313768007e-05, + "loss": 0.8944, + "step": 18987 + }, + { + "epoch": 1.22, + "learning_rate": 1.594710062474506e-05, + "loss": 0.9995, + "step": 18988 + }, + { + "epoch": 1.22, + "learning_rate": 1.594688593572211e-05, + "loss": 1.015, + "step": 18989 + }, + { + "epoch": 1.22, + "learning_rate": 1.5946671246699157e-05, + "loss": 0.9388, + "step": 18990 + }, + { + "epoch": 1.22, + "learning_rate": 1.5946456557676206e-05, + "loss": 1.0056, + "step": 18991 + }, + { + "epoch": 1.22, + "learning_rate": 1.5946241868653255e-05, + "loss": 0.92, + "step": 18992 + }, + { + "epoch": 1.22, + "learning_rate": 1.5946027179630308e-05, + "loss": 1.0118, + "step": 18993 + }, + { + "epoch": 1.22, + "learning_rate": 1.5945812490607357e-05, + "loss": 0.8955, + "step": 18994 + }, + { + "epoch": 1.22, + "learning_rate": 1.5945597801584406e-05, + "loss": 0.9845, + "step": 18995 + }, + { + "epoch": 1.22, + "learning_rate": 1.5945383112561458e-05, + "loss": 0.9825, + "step": 18996 + }, + { + "epoch": 1.22, + "learning_rate": 1.5945168423538507e-05, + "loss": 0.9855, + "step": 18997 + }, + { + "epoch": 1.22, + "learning_rate": 1.5944953734515556e-05, + "loss": 1.0129, + "step": 18998 + }, + { + "epoch": 1.22, + "learning_rate": 1.5944739045492605e-05, + "loss": 1.0012, + "step": 18999 + }, + { + "epoch": 1.22, + "learning_rate": 1.5944524356469654e-05, + "loss": 0.9814, + "step": 19000 + }, + { + "epoch": 1.22, + "eval_laion/cider": 0.5221260684127664, + "eval_laion/meteor": 0.3897373963294886, + "eval_laion/pred_num_words": 36.41796875, + "eval_laion/sacrebleu": 15.96862083770962, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0591351985931396, + "eval_runtime": 95.7618, + "eval_samples_per_second": 5.211, + "eval_steps_per_second": 0.334, + "step": 19000 + }, + { + "epoch": 1.22, + "learning_rate": 1.5944309667446706e-05, + "loss": 0.974, + "step": 19001 + }, + { + "epoch": 1.22, + "learning_rate": 1.5944094978423755e-05, + "loss": 0.9967, + "step": 19002 + }, + { + "epoch": 1.22, + "learning_rate": 1.5943880289400804e-05, + "loss": 0.9987, + "step": 19003 + }, + { + "epoch": 1.22, + "learning_rate": 1.5943665600377853e-05, + "loss": 0.9983, + "step": 19004 + }, + { + "epoch": 1.22, + "learning_rate": 1.5943450911354902e-05, + "loss": 0.9704, + "step": 19005 + }, + { + "epoch": 1.22, + "learning_rate": 1.594323622233195e-05, + "loss": 1.0119, + "step": 19006 + }, + { + "epoch": 1.22, + "learning_rate": 1.5943021533309003e-05, + "loss": 0.8941, + "step": 19007 + }, + { + "epoch": 1.22, + "learning_rate": 1.5942806844286056e-05, + "loss": 0.923, + "step": 19008 + }, + { + "epoch": 1.22, + "learning_rate": 1.5942592155263105e-05, + "loss": 0.9711, + "step": 19009 + }, + { + "epoch": 1.22, + "learning_rate": 1.5942377466240154e-05, + "loss": 0.968, + "step": 19010 + }, + { + "epoch": 1.22, + "learning_rate": 1.5942162777217203e-05, + "loss": 1.0305, + "step": 19011 + }, + { + "epoch": 1.22, + "learning_rate": 1.594194808819425e-05, + "loss": 1.0167, + "step": 19012 + }, + { + "epoch": 1.22, + "learning_rate": 1.59417333991713e-05, + "loss": 0.9999, + "step": 19013 + }, + { + "epoch": 1.22, + "learning_rate": 1.5941518710148353e-05, + "loss": 0.9702, + "step": 19014 + }, + { + "epoch": 1.22, + "learning_rate": 1.5941304021125402e-05, + "loss": 0.9593, + "step": 19015 + }, + { + "epoch": 1.22, + "learning_rate": 1.594108933210245e-05, + "loss": 0.982, + "step": 19016 + }, + { + "epoch": 1.22, + "learning_rate": 1.59408746430795e-05, + "loss": 1.0175, + "step": 19017 + }, + { + "epoch": 1.22, + "learning_rate": 1.594065995405655e-05, + "loss": 0.9797, + "step": 19018 + }, + { + "epoch": 1.22, + "learning_rate": 1.59404452650336e-05, + "loss": 1.0142, + "step": 19019 + }, + { + "epoch": 1.22, + "learning_rate": 1.594023057601065e-05, + "loss": 1.0024, + "step": 19020 + }, + { + "epoch": 1.22, + "learning_rate": 1.59400158869877e-05, + "loss": 0.9857, + "step": 19021 + }, + { + "epoch": 1.22, + "learning_rate": 1.593980119796475e-05, + "loss": 1.0734, + "step": 19022 + }, + { + "epoch": 1.22, + "learning_rate": 1.59395865089418e-05, + "loss": 0.9877, + "step": 19023 + }, + { + "epoch": 1.22, + "learning_rate": 1.593937181991885e-05, + "loss": 0.9524, + "step": 19024 + }, + { + "epoch": 1.22, + "learning_rate": 1.5939157130895898e-05, + "loss": 0.9674, + "step": 19025 + }, + { + "epoch": 1.22, + "learning_rate": 1.5938942441872947e-05, + "loss": 1.1061, + "step": 19026 + }, + { + "epoch": 1.22, + "learning_rate": 1.593872775285e-05, + "loss": 1.0431, + "step": 19027 + }, + { + "epoch": 1.22, + "learning_rate": 1.593851306382705e-05, + "loss": 0.9631, + "step": 19028 + }, + { + "epoch": 1.22, + "learning_rate": 1.5938298374804097e-05, + "loss": 0.9789, + "step": 19029 + }, + { + "epoch": 1.22, + "learning_rate": 1.5938083685781146e-05, + "loss": 1.0306, + "step": 19030 + }, + { + "epoch": 1.22, + "learning_rate": 1.5937868996758195e-05, + "loss": 1.0121, + "step": 19031 + }, + { + "epoch": 1.22, + "learning_rate": 1.5937654307735248e-05, + "loss": 1.0033, + "step": 19032 + }, + { + "epoch": 1.22, + "learning_rate": 1.5937439618712297e-05, + "loss": 0.951, + "step": 19033 + }, + { + "epoch": 1.22, + "learning_rate": 1.5937224929689346e-05, + "loss": 1.0102, + "step": 19034 + }, + { + "epoch": 1.22, + "learning_rate": 1.5937010240666395e-05, + "loss": 0.9639, + "step": 19035 + }, + { + "epoch": 1.22, + "learning_rate": 1.5936795551643444e-05, + "loss": 0.9552, + "step": 19036 + }, + { + "epoch": 1.22, + "learning_rate": 1.5936580862620496e-05, + "loss": 0.9751, + "step": 19037 + }, + { + "epoch": 1.22, + "learning_rate": 1.5936366173597545e-05, + "loss": 0.9445, + "step": 19038 + }, + { + "epoch": 1.22, + "learning_rate": 1.5936151484574597e-05, + "loss": 0.9078, + "step": 19039 + }, + { + "epoch": 1.22, + "learning_rate": 1.5935936795551646e-05, + "loss": 1.0127, + "step": 19040 + }, + { + "epoch": 1.23, + "learning_rate": 1.5935722106528695e-05, + "loss": 0.9277, + "step": 19041 + }, + { + "epoch": 1.23, + "learning_rate": 1.5935507417505744e-05, + "loss": 0.9077, + "step": 19042 + }, + { + "epoch": 1.23, + "learning_rate": 1.5935292728482793e-05, + "loss": 0.9929, + "step": 19043 + }, + { + "epoch": 1.23, + "learning_rate": 1.5935078039459842e-05, + "loss": 0.9188, + "step": 19044 + }, + { + "epoch": 1.23, + "learning_rate": 1.5934863350436894e-05, + "loss": 0.983, + "step": 19045 + }, + { + "epoch": 1.23, + "learning_rate": 1.5934648661413943e-05, + "loss": 0.9854, + "step": 19046 + }, + { + "epoch": 1.23, + "learning_rate": 1.5934433972390992e-05, + "loss": 0.9833, + "step": 19047 + }, + { + "epoch": 1.23, + "learning_rate": 1.593421928336804e-05, + "loss": 0.887, + "step": 19048 + }, + { + "epoch": 1.23, + "learning_rate": 1.593400459434509e-05, + "loss": 0.9998, + "step": 19049 + }, + { + "epoch": 1.23, + "learning_rate": 1.5933789905322143e-05, + "loss": 0.9435, + "step": 19050 + }, + { + "epoch": 1.23, + "learning_rate": 1.593357521629919e-05, + "loss": 0.8617, + "step": 19051 + }, + { + "epoch": 1.23, + "learning_rate": 1.5933360527276244e-05, + "loss": 1.0312, + "step": 19052 + }, + { + "epoch": 1.23, + "learning_rate": 1.5933145838253293e-05, + "loss": 0.9835, + "step": 19053 + }, + { + "epoch": 1.23, + "learning_rate": 1.5932931149230342e-05, + "loss": 0.9621, + "step": 19054 + }, + { + "epoch": 1.23, + "learning_rate": 1.593271646020739e-05, + "loss": 0.9128, + "step": 19055 + }, + { + "epoch": 1.23, + "learning_rate": 1.593250177118444e-05, + "loss": 1.0748, + "step": 19056 + }, + { + "epoch": 1.23, + "learning_rate": 1.5932287082161492e-05, + "loss": 0.8667, + "step": 19057 + }, + { + "epoch": 1.23, + "learning_rate": 1.593207239313854e-05, + "loss": 0.9589, + "step": 19058 + }, + { + "epoch": 1.23, + "learning_rate": 1.593185770411559e-05, + "loss": 0.9821, + "step": 19059 + }, + { + "epoch": 1.23, + "learning_rate": 1.593164301509264e-05, + "loss": 1.0961, + "step": 19060 + }, + { + "epoch": 1.23, + "learning_rate": 1.5931428326069688e-05, + "loss": 1.0295, + "step": 19061 + }, + { + "epoch": 1.23, + "learning_rate": 1.5931213637046737e-05, + "loss": 0.9703, + "step": 19062 + }, + { + "epoch": 1.23, + "learning_rate": 1.593099894802379e-05, + "loss": 0.9982, + "step": 19063 + }, + { + "epoch": 1.23, + "learning_rate": 1.593078425900084e-05, + "loss": 0.985, + "step": 19064 + }, + { + "epoch": 1.23, + "learning_rate": 1.5930569569977887e-05, + "loss": 0.9995, + "step": 19065 + }, + { + "epoch": 1.23, + "learning_rate": 1.593035488095494e-05, + "loss": 0.9838, + "step": 19066 + }, + { + "epoch": 1.23, + "learning_rate": 1.593014019193199e-05, + "loss": 0.9065, + "step": 19067 + }, + { + "epoch": 1.23, + "learning_rate": 1.5929925502909038e-05, + "loss": 0.9874, + "step": 19068 + }, + { + "epoch": 1.23, + "learning_rate": 1.5929710813886087e-05, + "loss": 0.994, + "step": 19069 + }, + { + "epoch": 1.23, + "learning_rate": 1.592949612486314e-05, + "loss": 1.0381, + "step": 19070 + }, + { + "epoch": 1.23, + "learning_rate": 1.5929281435840188e-05, + "loss": 0.9804, + "step": 19071 + }, + { + "epoch": 1.23, + "learning_rate": 1.5929066746817237e-05, + "loss": 0.9655, + "step": 19072 + }, + { + "epoch": 1.23, + "learning_rate": 1.5928852057794286e-05, + "loss": 0.9486, + "step": 19073 + }, + { + "epoch": 1.23, + "learning_rate": 1.5928637368771335e-05, + "loss": 0.9748, + "step": 19074 + }, + { + "epoch": 1.23, + "learning_rate": 1.5928422679748384e-05, + "loss": 1.0031, + "step": 19075 + }, + { + "epoch": 1.23, + "learning_rate": 1.5928207990725436e-05, + "loss": 0.9485, + "step": 19076 + }, + { + "epoch": 1.23, + "learning_rate": 1.5927993301702485e-05, + "loss": 1.0063, + "step": 19077 + }, + { + "epoch": 1.23, + "learning_rate": 1.5927778612679534e-05, + "loss": 0.9641, + "step": 19078 + }, + { + "epoch": 1.23, + "learning_rate": 1.5927563923656583e-05, + "loss": 0.9986, + "step": 19079 + }, + { + "epoch": 1.23, + "learning_rate": 1.5927349234633635e-05, + "loss": 0.9224, + "step": 19080 + }, + { + "epoch": 1.23, + "learning_rate": 1.5927134545610684e-05, + "loss": 0.9557, + "step": 19081 + }, + { + "epoch": 1.23, + "learning_rate": 1.5926919856587733e-05, + "loss": 1.0164, + "step": 19082 + }, + { + "epoch": 1.23, + "learning_rate": 1.5926705167564786e-05, + "loss": 1.0288, + "step": 19083 + }, + { + "epoch": 1.23, + "learning_rate": 1.5926490478541835e-05, + "loss": 1.0113, + "step": 19084 + }, + { + "epoch": 1.23, + "learning_rate": 1.5926275789518884e-05, + "loss": 0.9502, + "step": 19085 + }, + { + "epoch": 1.23, + "learning_rate": 1.5926061100495932e-05, + "loss": 0.9286, + "step": 19086 + }, + { + "epoch": 1.23, + "learning_rate": 1.592584641147298e-05, + "loss": 0.9518, + "step": 19087 + }, + { + "epoch": 1.23, + "learning_rate": 1.5925631722450034e-05, + "loss": 0.9045, + "step": 19088 + }, + { + "epoch": 1.23, + "learning_rate": 1.5925417033427083e-05, + "loss": 0.9728, + "step": 19089 + }, + { + "epoch": 1.23, + "learning_rate": 1.5925202344404132e-05, + "loss": 1.0008, + "step": 19090 + }, + { + "epoch": 1.23, + "learning_rate": 1.592498765538118e-05, + "loss": 0.9596, + "step": 19091 + }, + { + "epoch": 1.23, + "learning_rate": 1.592477296635823e-05, + "loss": 0.9673, + "step": 19092 + }, + { + "epoch": 1.23, + "learning_rate": 1.592455827733528e-05, + "loss": 0.8888, + "step": 19093 + }, + { + "epoch": 1.23, + "learning_rate": 1.592434358831233e-05, + "loss": 0.967, + "step": 19094 + }, + { + "epoch": 1.23, + "learning_rate": 1.5924128899289383e-05, + "loss": 0.947, + "step": 19095 + }, + { + "epoch": 1.23, + "learning_rate": 1.5923914210266432e-05, + "loss": 0.995, + "step": 19096 + }, + { + "epoch": 1.23, + "learning_rate": 1.592369952124348e-05, + "loss": 1.0109, + "step": 19097 + }, + { + "epoch": 1.23, + "learning_rate": 1.592348483222053e-05, + "loss": 0.95, + "step": 19098 + }, + { + "epoch": 1.23, + "learning_rate": 1.592327014319758e-05, + "loss": 1.0284, + "step": 19099 + }, + { + "epoch": 1.23, + "learning_rate": 1.5923055454174628e-05, + "loss": 0.9515, + "step": 19100 + }, + { + "epoch": 1.23, + "learning_rate": 1.592284076515168e-05, + "loss": 0.9402, + "step": 19101 + }, + { + "epoch": 1.23, + "learning_rate": 1.592262607612873e-05, + "loss": 0.9761, + "step": 19102 + }, + { + "epoch": 1.23, + "learning_rate": 1.592241138710578e-05, + "loss": 0.9116, + "step": 19103 + }, + { + "epoch": 1.23, + "learning_rate": 1.5922196698082827e-05, + "loss": 1.0073, + "step": 19104 + }, + { + "epoch": 1.23, + "learning_rate": 1.5921982009059876e-05, + "loss": 0.9566, + "step": 19105 + }, + { + "epoch": 1.23, + "learning_rate": 1.592176732003693e-05, + "loss": 0.9423, + "step": 19106 + }, + { + "epoch": 1.23, + "learning_rate": 1.5921552631013978e-05, + "loss": 0.9914, + "step": 19107 + }, + { + "epoch": 1.23, + "learning_rate": 1.5921337941991027e-05, + "loss": 1.0209, + "step": 19108 + }, + { + "epoch": 1.23, + "learning_rate": 1.592112325296808e-05, + "loss": 1.0353, + "step": 19109 + }, + { + "epoch": 1.23, + "learning_rate": 1.5920908563945128e-05, + "loss": 0.9287, + "step": 19110 + }, + { + "epoch": 1.23, + "learning_rate": 1.5920693874922177e-05, + "loss": 1.0475, + "step": 19111 + }, + { + "epoch": 1.23, + "learning_rate": 1.5920479185899226e-05, + "loss": 0.9936, + "step": 19112 + }, + { + "epoch": 1.23, + "learning_rate": 1.5920264496876275e-05, + "loss": 0.9417, + "step": 19113 + }, + { + "epoch": 1.23, + "learning_rate": 1.5920049807853327e-05, + "loss": 1.0293, + "step": 19114 + }, + { + "epoch": 1.23, + "learning_rate": 1.5919835118830376e-05, + "loss": 0.9819, + "step": 19115 + }, + { + "epoch": 1.23, + "learning_rate": 1.5919620429807425e-05, + "loss": 0.8595, + "step": 19116 + }, + { + "epoch": 1.23, + "learning_rate": 1.5919405740784474e-05, + "loss": 1.0075, + "step": 19117 + }, + { + "epoch": 1.23, + "learning_rate": 1.5919191051761523e-05, + "loss": 0.9262, + "step": 19118 + }, + { + "epoch": 1.23, + "learning_rate": 1.5918976362738575e-05, + "loss": 0.9506, + "step": 19119 + }, + { + "epoch": 1.23, + "learning_rate": 1.5918761673715624e-05, + "loss": 1.0031, + "step": 19120 + }, + { + "epoch": 1.23, + "learning_rate": 1.5918546984692673e-05, + "loss": 0.9761, + "step": 19121 + }, + { + "epoch": 1.23, + "learning_rate": 1.5918332295669722e-05, + "loss": 0.9414, + "step": 19122 + }, + { + "epoch": 1.23, + "learning_rate": 1.591811760664677e-05, + "loss": 0.9569, + "step": 19123 + }, + { + "epoch": 1.23, + "learning_rate": 1.5917902917623824e-05, + "loss": 0.9419, + "step": 19124 + }, + { + "epoch": 1.23, + "learning_rate": 1.5917688228600873e-05, + "loss": 1.01, + "step": 19125 + }, + { + "epoch": 1.23, + "learning_rate": 1.5917473539577925e-05, + "loss": 0.8785, + "step": 19126 + }, + { + "epoch": 1.23, + "learning_rate": 1.5917258850554974e-05, + "loss": 0.9001, + "step": 19127 + }, + { + "epoch": 1.23, + "learning_rate": 1.5917044161532023e-05, + "loss": 0.9449, + "step": 19128 + }, + { + "epoch": 1.23, + "learning_rate": 1.5916829472509072e-05, + "loss": 0.9637, + "step": 19129 + }, + { + "epoch": 1.23, + "learning_rate": 1.591661478348612e-05, + "loss": 0.9573, + "step": 19130 + }, + { + "epoch": 1.23, + "learning_rate": 1.591640009446317e-05, + "loss": 0.9615, + "step": 19131 + }, + { + "epoch": 1.23, + "learning_rate": 1.5916185405440222e-05, + "loss": 0.9974, + "step": 19132 + }, + { + "epoch": 1.23, + "learning_rate": 1.591597071641727e-05, + "loss": 1.0702, + "step": 19133 + }, + { + "epoch": 1.23, + "learning_rate": 1.591575602739432e-05, + "loss": 1.026, + "step": 19134 + }, + { + "epoch": 1.23, + "learning_rate": 1.591554133837137e-05, + "loss": 0.8889, + "step": 19135 + }, + { + "epoch": 1.23, + "learning_rate": 1.5915326649348418e-05, + "loss": 0.9544, + "step": 19136 + }, + { + "epoch": 1.23, + "learning_rate": 1.591511196032547e-05, + "loss": 0.9813, + "step": 19137 + }, + { + "epoch": 1.23, + "learning_rate": 1.591489727130252e-05, + "loss": 0.9539, + "step": 19138 + }, + { + "epoch": 1.23, + "learning_rate": 1.591468258227957e-05, + "loss": 0.918, + "step": 19139 + }, + { + "epoch": 1.23, + "learning_rate": 1.591446789325662e-05, + "loss": 0.9901, + "step": 19140 + }, + { + "epoch": 1.23, + "learning_rate": 1.591425320423367e-05, + "loss": 0.8998, + "step": 19141 + }, + { + "epoch": 1.23, + "learning_rate": 1.591403851521072e-05, + "loss": 0.9548, + "step": 19142 + }, + { + "epoch": 1.23, + "learning_rate": 1.5913823826187768e-05, + "loss": 0.9685, + "step": 19143 + }, + { + "epoch": 1.23, + "learning_rate": 1.591360913716482e-05, + "loss": 0.9997, + "step": 19144 + }, + { + "epoch": 1.23, + "learning_rate": 1.591339444814187e-05, + "loss": 0.9904, + "step": 19145 + }, + { + "epoch": 1.23, + "learning_rate": 1.5913179759118918e-05, + "loss": 0.9799, + "step": 19146 + }, + { + "epoch": 1.23, + "learning_rate": 1.5912965070095967e-05, + "loss": 0.992, + "step": 19147 + }, + { + "epoch": 1.23, + "learning_rate": 1.5912750381073016e-05, + "loss": 0.9569, + "step": 19148 + }, + { + "epoch": 1.23, + "learning_rate": 1.5912535692050065e-05, + "loss": 1.0143, + "step": 19149 + }, + { + "epoch": 1.23, + "learning_rate": 1.5912321003027117e-05, + "loss": 0.9175, + "step": 19150 + }, + { + "epoch": 1.23, + "learning_rate": 1.5912106314004166e-05, + "loss": 0.9789, + "step": 19151 + }, + { + "epoch": 1.23, + "learning_rate": 1.5911891624981215e-05, + "loss": 0.9952, + "step": 19152 + }, + { + "epoch": 1.23, + "learning_rate": 1.5911676935958267e-05, + "loss": 1.0185, + "step": 19153 + }, + { + "epoch": 1.23, + "learning_rate": 1.5911462246935316e-05, + "loss": 1.0324, + "step": 19154 + }, + { + "epoch": 1.23, + "learning_rate": 1.5911247557912365e-05, + "loss": 1.1185, + "step": 19155 + }, + { + "epoch": 1.23, + "learning_rate": 1.5911032868889414e-05, + "loss": 0.9891, + "step": 19156 + }, + { + "epoch": 1.23, + "learning_rate": 1.5910818179866467e-05, + "loss": 0.9807, + "step": 19157 + }, + { + "epoch": 1.23, + "learning_rate": 1.5910603490843516e-05, + "loss": 0.9376, + "step": 19158 + }, + { + "epoch": 1.23, + "learning_rate": 1.5910388801820565e-05, + "loss": 0.978, + "step": 19159 + }, + { + "epoch": 1.23, + "learning_rate": 1.5910174112797613e-05, + "loss": 0.9616, + "step": 19160 + }, + { + "epoch": 1.23, + "learning_rate": 1.5909959423774662e-05, + "loss": 0.9581, + "step": 19161 + }, + { + "epoch": 1.23, + "learning_rate": 1.590974473475171e-05, + "loss": 0.9617, + "step": 19162 + }, + { + "epoch": 1.23, + "learning_rate": 1.5909530045728764e-05, + "loss": 0.9976, + "step": 19163 + }, + { + "epoch": 1.23, + "learning_rate": 1.5909315356705813e-05, + "loss": 1.0019, + "step": 19164 + }, + { + "epoch": 1.23, + "learning_rate": 1.5909100667682862e-05, + "loss": 0.9713, + "step": 19165 + }, + { + "epoch": 1.23, + "learning_rate": 1.590888597865991e-05, + "loss": 0.9358, + "step": 19166 + }, + { + "epoch": 1.23, + "learning_rate": 1.590867128963696e-05, + "loss": 0.9856, + "step": 19167 + }, + { + "epoch": 1.23, + "learning_rate": 1.5908456600614012e-05, + "loss": 1.0481, + "step": 19168 + }, + { + "epoch": 1.23, + "learning_rate": 1.590824191159106e-05, + "loss": 1.0113, + "step": 19169 + }, + { + "epoch": 1.23, + "learning_rate": 1.5908027222568113e-05, + "loss": 0.9402, + "step": 19170 + }, + { + "epoch": 1.23, + "learning_rate": 1.5907812533545162e-05, + "loss": 0.9685, + "step": 19171 + }, + { + "epoch": 1.23, + "learning_rate": 1.590759784452221e-05, + "loss": 0.9794, + "step": 19172 + }, + { + "epoch": 1.23, + "learning_rate": 1.590738315549926e-05, + "loss": 1.0256, + "step": 19173 + }, + { + "epoch": 1.23, + "learning_rate": 1.590716846647631e-05, + "loss": 1.0082, + "step": 19174 + }, + { + "epoch": 1.23, + "learning_rate": 1.590695377745336e-05, + "loss": 0.982, + "step": 19175 + }, + { + "epoch": 1.23, + "learning_rate": 1.590673908843041e-05, + "loss": 0.918, + "step": 19176 + }, + { + "epoch": 1.23, + "learning_rate": 1.590652439940746e-05, + "loss": 1.0006, + "step": 19177 + }, + { + "epoch": 1.23, + "learning_rate": 1.590630971038451e-05, + "loss": 0.9404, + "step": 19178 + }, + { + "epoch": 1.23, + "learning_rate": 1.5906095021361557e-05, + "loss": 0.988, + "step": 19179 + }, + { + "epoch": 1.23, + "learning_rate": 1.5905880332338606e-05, + "loss": 0.9782, + "step": 19180 + }, + { + "epoch": 1.23, + "learning_rate": 1.590566564331566e-05, + "loss": 0.9568, + "step": 19181 + }, + { + "epoch": 1.23, + "learning_rate": 1.5905450954292708e-05, + "loss": 1.0527, + "step": 19182 + }, + { + "epoch": 1.23, + "learning_rate": 1.590523626526976e-05, + "loss": 1.034, + "step": 19183 + }, + { + "epoch": 1.23, + "learning_rate": 1.590502157624681e-05, + "loss": 0.9926, + "step": 19184 + }, + { + "epoch": 1.23, + "learning_rate": 1.5904806887223858e-05, + "loss": 1.0279, + "step": 19185 + }, + { + "epoch": 1.23, + "learning_rate": 1.5904592198200907e-05, + "loss": 1.0163, + "step": 19186 + }, + { + "epoch": 1.23, + "learning_rate": 1.5904377509177956e-05, + "loss": 0.9803, + "step": 19187 + }, + { + "epoch": 1.23, + "learning_rate": 1.5904162820155008e-05, + "loss": 0.9496, + "step": 19188 + }, + { + "epoch": 1.23, + "learning_rate": 1.5903948131132057e-05, + "loss": 0.9143, + "step": 19189 + }, + { + "epoch": 1.23, + "learning_rate": 1.5903733442109106e-05, + "loss": 1.0093, + "step": 19190 + }, + { + "epoch": 1.23, + "learning_rate": 1.5903518753086155e-05, + "loss": 1.0531, + "step": 19191 + }, + { + "epoch": 1.23, + "learning_rate": 1.5903304064063204e-05, + "loss": 0.9766, + "step": 19192 + }, + { + "epoch": 1.23, + "learning_rate": 1.5903089375040256e-05, + "loss": 0.928, + "step": 19193 + }, + { + "epoch": 1.23, + "learning_rate": 1.5902874686017305e-05, + "loss": 0.9743, + "step": 19194 + }, + { + "epoch": 1.23, + "learning_rate": 1.5902659996994354e-05, + "loss": 0.9811, + "step": 19195 + }, + { + "epoch": 1.24, + "learning_rate": 1.5902445307971403e-05, + "loss": 0.984, + "step": 19196 + }, + { + "epoch": 1.24, + "learning_rate": 1.5902230618948456e-05, + "loss": 0.9862, + "step": 19197 + }, + { + "epoch": 1.24, + "learning_rate": 1.5902015929925505e-05, + "loss": 0.9985, + "step": 19198 + }, + { + "epoch": 1.24, + "learning_rate": 1.5901801240902554e-05, + "loss": 1.0371, + "step": 19199 + }, + { + "epoch": 1.24, + "learning_rate": 1.5901586551879603e-05, + "loss": 0.9088, + "step": 19200 + }, + { + "epoch": 1.24, + "learning_rate": 1.5901371862856655e-05, + "loss": 0.9569, + "step": 19201 + }, + { + "epoch": 1.24, + "learning_rate": 1.5901157173833704e-05, + "loss": 0.9511, + "step": 19202 + }, + { + "epoch": 1.24, + "learning_rate": 1.5900942484810753e-05, + "loss": 0.9237, + "step": 19203 + }, + { + "epoch": 1.24, + "learning_rate": 1.5900727795787802e-05, + "loss": 0.972, + "step": 19204 + }, + { + "epoch": 1.24, + "learning_rate": 1.590051310676485e-05, + "loss": 0.949, + "step": 19205 + }, + { + "epoch": 1.24, + "learning_rate": 1.5900298417741903e-05, + "loss": 1.0257, + "step": 19206 + }, + { + "epoch": 1.24, + "learning_rate": 1.5900083728718952e-05, + "loss": 0.8687, + "step": 19207 + }, + { + "epoch": 1.24, + "learning_rate": 1.5899869039696e-05, + "loss": 0.9984, + "step": 19208 + }, + { + "epoch": 1.24, + "learning_rate": 1.589965435067305e-05, + "loss": 0.9966, + "step": 19209 + }, + { + "epoch": 1.24, + "learning_rate": 1.58994396616501e-05, + "loss": 1.0059, + "step": 19210 + }, + { + "epoch": 1.24, + "learning_rate": 1.589922497262715e-05, + "loss": 0.999, + "step": 19211 + }, + { + "epoch": 1.24, + "learning_rate": 1.58990102836042e-05, + "loss": 0.9702, + "step": 19212 + }, + { + "epoch": 1.24, + "learning_rate": 1.5898795594581253e-05, + "loss": 0.9035, + "step": 19213 + }, + { + "epoch": 1.24, + "learning_rate": 1.58985809055583e-05, + "loss": 1.0067, + "step": 19214 + }, + { + "epoch": 1.24, + "learning_rate": 1.589836621653535e-05, + "loss": 1.0911, + "step": 19215 + }, + { + "epoch": 1.24, + "learning_rate": 1.58981515275124e-05, + "loss": 0.9663, + "step": 19216 + }, + { + "epoch": 1.24, + "learning_rate": 1.589793683848945e-05, + "loss": 1.0405, + "step": 19217 + }, + { + "epoch": 1.24, + "learning_rate": 1.5897722149466498e-05, + "loss": 0.9772, + "step": 19218 + }, + { + "epoch": 1.24, + "learning_rate": 1.589750746044355e-05, + "loss": 1.0069, + "step": 19219 + }, + { + "epoch": 1.24, + "learning_rate": 1.58972927714206e-05, + "loss": 1.0399, + "step": 19220 + }, + { + "epoch": 1.24, + "learning_rate": 1.5897078082397648e-05, + "loss": 1.0071, + "step": 19221 + }, + { + "epoch": 1.24, + "learning_rate": 1.5896863393374697e-05, + "loss": 1.0395, + "step": 19222 + }, + { + "epoch": 1.24, + "learning_rate": 1.5896648704351746e-05, + "loss": 0.9432, + "step": 19223 + }, + { + "epoch": 1.24, + "learning_rate": 1.5896434015328798e-05, + "loss": 1.016, + "step": 19224 + }, + { + "epoch": 1.24, + "learning_rate": 1.5896219326305847e-05, + "loss": 0.9855, + "step": 19225 + }, + { + "epoch": 1.24, + "learning_rate": 1.58960046372829e-05, + "loss": 0.906, + "step": 19226 + }, + { + "epoch": 1.24, + "learning_rate": 1.589578994825995e-05, + "loss": 0.9105, + "step": 19227 + }, + { + "epoch": 1.24, + "learning_rate": 1.5895575259236997e-05, + "loss": 0.9939, + "step": 19228 + }, + { + "epoch": 1.24, + "learning_rate": 1.5895360570214046e-05, + "loss": 0.9937, + "step": 19229 + }, + { + "epoch": 1.24, + "learning_rate": 1.5895145881191095e-05, + "loss": 0.9522, + "step": 19230 + }, + { + "epoch": 1.24, + "learning_rate": 1.5894931192168148e-05, + "loss": 0.8969, + "step": 19231 + }, + { + "epoch": 1.24, + "learning_rate": 1.5894716503145197e-05, + "loss": 0.939, + "step": 19232 + }, + { + "epoch": 1.24, + "learning_rate": 1.5894501814122246e-05, + "loss": 0.9625, + "step": 19233 + }, + { + "epoch": 1.24, + "learning_rate": 1.5894287125099295e-05, + "loss": 0.9795, + "step": 19234 + }, + { + "epoch": 1.24, + "learning_rate": 1.5894072436076343e-05, + "loss": 0.9705, + "step": 19235 + }, + { + "epoch": 1.24, + "learning_rate": 1.5893857747053392e-05, + "loss": 0.9522, + "step": 19236 + }, + { + "epoch": 1.24, + "learning_rate": 1.5893643058030445e-05, + "loss": 1.0016, + "step": 19237 + }, + { + "epoch": 1.24, + "learning_rate": 1.5893428369007494e-05, + "loss": 1.0445, + "step": 19238 + }, + { + "epoch": 1.24, + "learning_rate": 1.5893213679984543e-05, + "loss": 0.9687, + "step": 19239 + }, + { + "epoch": 1.24, + "learning_rate": 1.5892998990961595e-05, + "loss": 0.9534, + "step": 19240 + }, + { + "epoch": 1.24, + "learning_rate": 1.5892784301938644e-05, + "loss": 0.9146, + "step": 19241 + }, + { + "epoch": 1.24, + "learning_rate": 1.5892569612915693e-05, + "loss": 0.9497, + "step": 19242 + }, + { + "epoch": 1.24, + "learning_rate": 1.5892354923892742e-05, + "loss": 0.9594, + "step": 19243 + }, + { + "epoch": 1.24, + "learning_rate": 1.5892140234869794e-05, + "loss": 0.9187, + "step": 19244 + }, + { + "epoch": 1.24, + "learning_rate": 1.5891925545846843e-05, + "loss": 0.9601, + "step": 19245 + }, + { + "epoch": 1.24, + "learning_rate": 1.5891710856823892e-05, + "loss": 1.0165, + "step": 19246 + }, + { + "epoch": 1.24, + "learning_rate": 1.589149616780094e-05, + "loss": 0.9534, + "step": 19247 + }, + { + "epoch": 1.24, + "learning_rate": 1.589128147877799e-05, + "loss": 0.9989, + "step": 19248 + }, + { + "epoch": 1.24, + "learning_rate": 1.589106678975504e-05, + "loss": 0.9962, + "step": 19249 + }, + { + "epoch": 1.24, + "learning_rate": 1.589085210073209e-05, + "loss": 0.9578, + "step": 19250 + }, + { + "epoch": 1.24, + "learning_rate": 1.589063741170914e-05, + "loss": 0.9381, + "step": 19251 + }, + { + "epoch": 1.24, + "learning_rate": 1.589042272268619e-05, + "loss": 1.0216, + "step": 19252 + }, + { + "epoch": 1.24, + "learning_rate": 1.589020803366324e-05, + "loss": 0.9909, + "step": 19253 + }, + { + "epoch": 1.24, + "learning_rate": 1.5889993344640287e-05, + "loss": 0.926, + "step": 19254 + }, + { + "epoch": 1.24, + "learning_rate": 1.588977865561734e-05, + "loss": 0.9852, + "step": 19255 + }, + { + "epoch": 1.24, + "learning_rate": 1.588956396659439e-05, + "loss": 0.9747, + "step": 19256 + }, + { + "epoch": 1.24, + "learning_rate": 1.588934927757144e-05, + "loss": 0.9585, + "step": 19257 + }, + { + "epoch": 1.24, + "learning_rate": 1.588913458854849e-05, + "loss": 0.9589, + "step": 19258 + }, + { + "epoch": 1.24, + "learning_rate": 1.588891989952554e-05, + "loss": 0.9903, + "step": 19259 + }, + { + "epoch": 1.24, + "learning_rate": 1.5888705210502588e-05, + "loss": 0.9605, + "step": 19260 + }, + { + "epoch": 1.24, + "learning_rate": 1.5888490521479637e-05, + "loss": 0.9889, + "step": 19261 + }, + { + "epoch": 1.24, + "learning_rate": 1.588827583245669e-05, + "loss": 0.9712, + "step": 19262 + }, + { + "epoch": 1.24, + "learning_rate": 1.5888061143433738e-05, + "loss": 0.9731, + "step": 19263 + }, + { + "epoch": 1.24, + "learning_rate": 1.5887846454410787e-05, + "loss": 1.0235, + "step": 19264 + }, + { + "epoch": 1.24, + "learning_rate": 1.5887631765387836e-05, + "loss": 1.0305, + "step": 19265 + }, + { + "epoch": 1.24, + "learning_rate": 1.5887417076364885e-05, + "loss": 1.0722, + "step": 19266 + }, + { + "epoch": 1.24, + "learning_rate": 1.5887202387341934e-05, + "loss": 0.9614, + "step": 19267 + }, + { + "epoch": 1.24, + "learning_rate": 1.5886987698318986e-05, + "loss": 0.946, + "step": 19268 + }, + { + "epoch": 1.24, + "learning_rate": 1.5886773009296035e-05, + "loss": 0.9747, + "step": 19269 + }, + { + "epoch": 1.24, + "learning_rate": 1.5886558320273088e-05, + "loss": 0.8935, + "step": 19270 + }, + { + "epoch": 1.24, + "learning_rate": 1.5886343631250137e-05, + "loss": 0.9821, + "step": 19271 + }, + { + "epoch": 1.24, + "learning_rate": 1.5886128942227186e-05, + "loss": 0.9962, + "step": 19272 + }, + { + "epoch": 1.24, + "learning_rate": 1.5885914253204235e-05, + "loss": 0.9682, + "step": 19273 + }, + { + "epoch": 1.24, + "learning_rate": 1.5885699564181284e-05, + "loss": 1.0019, + "step": 19274 + }, + { + "epoch": 1.24, + "learning_rate": 1.5885484875158336e-05, + "loss": 1.001, + "step": 19275 + }, + { + "epoch": 1.24, + "learning_rate": 1.5885270186135385e-05, + "loss": 0.9454, + "step": 19276 + }, + { + "epoch": 1.24, + "learning_rate": 1.5885055497112434e-05, + "loss": 1.0264, + "step": 19277 + }, + { + "epoch": 1.24, + "learning_rate": 1.5884840808089483e-05, + "loss": 0.9004, + "step": 19278 + }, + { + "epoch": 1.24, + "learning_rate": 1.5884626119066532e-05, + "loss": 0.9672, + "step": 19279 + }, + { + "epoch": 1.24, + "learning_rate": 1.5884411430043584e-05, + "loss": 1.0441, + "step": 19280 + }, + { + "epoch": 1.24, + "learning_rate": 1.5884196741020633e-05, + "loss": 1.0505, + "step": 19281 + }, + { + "epoch": 1.24, + "learning_rate": 1.5883982051997682e-05, + "loss": 1.0043, + "step": 19282 + }, + { + "epoch": 1.24, + "learning_rate": 1.588376736297473e-05, + "loss": 0.9636, + "step": 19283 + }, + { + "epoch": 1.24, + "learning_rate": 1.5883552673951783e-05, + "loss": 0.9801, + "step": 19284 + }, + { + "epoch": 1.24, + "learning_rate": 1.5883337984928832e-05, + "loss": 0.8889, + "step": 19285 + }, + { + "epoch": 1.24, + "learning_rate": 1.588312329590588e-05, + "loss": 0.9736, + "step": 19286 + }, + { + "epoch": 1.24, + "learning_rate": 1.588290860688293e-05, + "loss": 0.9133, + "step": 19287 + }, + { + "epoch": 1.24, + "learning_rate": 1.5882693917859983e-05, + "loss": 0.9515, + "step": 19288 + }, + { + "epoch": 1.24, + "learning_rate": 1.588247922883703e-05, + "loss": 1.0223, + "step": 19289 + }, + { + "epoch": 1.24, + "learning_rate": 1.588226453981408e-05, + "loss": 0.9651, + "step": 19290 + }, + { + "epoch": 1.24, + "learning_rate": 1.588204985079113e-05, + "loss": 0.906, + "step": 19291 + }, + { + "epoch": 1.24, + "learning_rate": 1.588183516176818e-05, + "loss": 0.98, + "step": 19292 + }, + { + "epoch": 1.24, + "learning_rate": 1.588162047274523e-05, + "loss": 0.9629, + "step": 19293 + }, + { + "epoch": 1.24, + "learning_rate": 1.588140578372228e-05, + "loss": 1.0123, + "step": 19294 + }, + { + "epoch": 1.24, + "learning_rate": 1.588119109469933e-05, + "loss": 0.9872, + "step": 19295 + }, + { + "epoch": 1.24, + "learning_rate": 1.5880976405676378e-05, + "loss": 0.9614, + "step": 19296 + }, + { + "epoch": 1.24, + "learning_rate": 1.5880761716653427e-05, + "loss": 0.9342, + "step": 19297 + }, + { + "epoch": 1.24, + "learning_rate": 1.588054702763048e-05, + "loss": 0.9493, + "step": 19298 + }, + { + "epoch": 1.24, + "learning_rate": 1.5880332338607528e-05, + "loss": 0.9711, + "step": 19299 + }, + { + "epoch": 1.24, + "learning_rate": 1.588011764958458e-05, + "loss": 0.9803, + "step": 19300 + }, + { + "epoch": 1.24, + "learning_rate": 1.587990296056163e-05, + "loss": 0.9274, + "step": 19301 + }, + { + "epoch": 1.24, + "learning_rate": 1.587968827153868e-05, + "loss": 1.0374, + "step": 19302 + }, + { + "epoch": 1.24, + "learning_rate": 1.5879473582515727e-05, + "loss": 0.9958, + "step": 19303 + }, + { + "epoch": 1.24, + "learning_rate": 1.5879258893492776e-05, + "loss": 0.9666, + "step": 19304 + }, + { + "epoch": 1.24, + "learning_rate": 1.5879044204469825e-05, + "loss": 1.0042, + "step": 19305 + }, + { + "epoch": 1.24, + "learning_rate": 1.5878829515446878e-05, + "loss": 0.9093, + "step": 19306 + }, + { + "epoch": 1.24, + "learning_rate": 1.5878614826423927e-05, + "loss": 0.9612, + "step": 19307 + }, + { + "epoch": 1.24, + "learning_rate": 1.5878400137400976e-05, + "loss": 1.0721, + "step": 19308 + }, + { + "epoch": 1.24, + "learning_rate": 1.5878185448378024e-05, + "loss": 0.961, + "step": 19309 + }, + { + "epoch": 1.24, + "learning_rate": 1.5877970759355073e-05, + "loss": 1.0354, + "step": 19310 + }, + { + "epoch": 1.24, + "learning_rate": 1.5877756070332126e-05, + "loss": 0.8921, + "step": 19311 + }, + { + "epoch": 1.24, + "learning_rate": 1.5877541381309175e-05, + "loss": 1.0208, + "step": 19312 + }, + { + "epoch": 1.24, + "learning_rate": 1.5877326692286227e-05, + "loss": 0.9483, + "step": 19313 + }, + { + "epoch": 1.24, + "learning_rate": 1.5877112003263276e-05, + "loss": 0.985, + "step": 19314 + }, + { + "epoch": 1.24, + "learning_rate": 1.5876897314240325e-05, + "loss": 1.0013, + "step": 19315 + }, + { + "epoch": 1.24, + "learning_rate": 1.5876682625217374e-05, + "loss": 0.979, + "step": 19316 + }, + { + "epoch": 1.24, + "learning_rate": 1.5876467936194423e-05, + "loss": 0.9776, + "step": 19317 + }, + { + "epoch": 1.24, + "learning_rate": 1.5876253247171475e-05, + "loss": 0.9652, + "step": 19318 + }, + { + "epoch": 1.24, + "learning_rate": 1.5876038558148524e-05, + "loss": 1.0191, + "step": 19319 + }, + { + "epoch": 1.24, + "learning_rate": 1.5875823869125573e-05, + "loss": 1.0193, + "step": 19320 + }, + { + "epoch": 1.24, + "learning_rate": 1.5875609180102622e-05, + "loss": 0.9857, + "step": 19321 + }, + { + "epoch": 1.24, + "learning_rate": 1.587539449107967e-05, + "loss": 0.9236, + "step": 19322 + }, + { + "epoch": 1.24, + "learning_rate": 1.587517980205672e-05, + "loss": 1.0068, + "step": 19323 + }, + { + "epoch": 1.24, + "learning_rate": 1.5874965113033772e-05, + "loss": 1.0232, + "step": 19324 + }, + { + "epoch": 1.24, + "learning_rate": 1.587475042401082e-05, + "loss": 1.0227, + "step": 19325 + }, + { + "epoch": 1.24, + "learning_rate": 1.587453573498787e-05, + "loss": 0.9505, + "step": 19326 + }, + { + "epoch": 1.24, + "learning_rate": 1.5874321045964923e-05, + "loss": 1.0053, + "step": 19327 + }, + { + "epoch": 1.24, + "learning_rate": 1.5874106356941972e-05, + "loss": 0.9907, + "step": 19328 + }, + { + "epoch": 1.24, + "learning_rate": 1.587389166791902e-05, + "loss": 1.0183, + "step": 19329 + }, + { + "epoch": 1.24, + "learning_rate": 1.587367697889607e-05, + "loss": 1.0411, + "step": 19330 + }, + { + "epoch": 1.24, + "learning_rate": 1.5873462289873122e-05, + "loss": 0.9633, + "step": 19331 + }, + { + "epoch": 1.24, + "learning_rate": 1.587324760085017e-05, + "loss": 0.9923, + "step": 19332 + }, + { + "epoch": 1.24, + "learning_rate": 1.587303291182722e-05, + "loss": 1.0075, + "step": 19333 + }, + { + "epoch": 1.24, + "learning_rate": 1.587281822280427e-05, + "loss": 0.9377, + "step": 19334 + }, + { + "epoch": 1.24, + "learning_rate": 1.5872603533781318e-05, + "loss": 0.9523, + "step": 19335 + }, + { + "epoch": 1.24, + "learning_rate": 1.5872388844758367e-05, + "loss": 0.9431, + "step": 19336 + }, + { + "epoch": 1.24, + "learning_rate": 1.587217415573542e-05, + "loss": 0.9542, + "step": 19337 + }, + { + "epoch": 1.24, + "learning_rate": 1.5871959466712468e-05, + "loss": 0.9538, + "step": 19338 + }, + { + "epoch": 1.24, + "learning_rate": 1.5871744777689517e-05, + "loss": 0.9457, + "step": 19339 + }, + { + "epoch": 1.24, + "learning_rate": 1.5871530088666566e-05, + "loss": 0.9759, + "step": 19340 + }, + { + "epoch": 1.24, + "learning_rate": 1.5871315399643615e-05, + "loss": 0.9652, + "step": 19341 + }, + { + "epoch": 1.24, + "learning_rate": 1.5871100710620667e-05, + "loss": 0.9386, + "step": 19342 + }, + { + "epoch": 1.24, + "learning_rate": 1.5870886021597716e-05, + "loss": 0.9488, + "step": 19343 + }, + { + "epoch": 1.24, + "learning_rate": 1.587067133257477e-05, + "loss": 1.0228, + "step": 19344 + }, + { + "epoch": 1.24, + "learning_rate": 1.5870456643551818e-05, + "loss": 0.9259, + "step": 19345 + }, + { + "epoch": 1.24, + "learning_rate": 1.5870241954528867e-05, + "loss": 1.0115, + "step": 19346 + }, + { + "epoch": 1.24, + "learning_rate": 1.5870027265505916e-05, + "loss": 0.9605, + "step": 19347 + }, + { + "epoch": 1.24, + "learning_rate": 1.5869812576482965e-05, + "loss": 1.0609, + "step": 19348 + }, + { + "epoch": 1.24, + "learning_rate": 1.5869597887460017e-05, + "loss": 0.9932, + "step": 19349 + }, + { + "epoch": 1.24, + "learning_rate": 1.5869383198437066e-05, + "loss": 0.9346, + "step": 19350 + }, + { + "epoch": 1.24, + "learning_rate": 1.5869168509414115e-05, + "loss": 0.9865, + "step": 19351 + }, + { + "epoch": 1.25, + "learning_rate": 1.5868953820391164e-05, + "loss": 0.9795, + "step": 19352 + }, + { + "epoch": 1.25, + "learning_rate": 1.5868739131368213e-05, + "loss": 1.141, + "step": 19353 + }, + { + "epoch": 1.25, + "learning_rate": 1.5868524442345262e-05, + "loss": 0.9622, + "step": 19354 + }, + { + "epoch": 1.25, + "learning_rate": 1.5868309753322314e-05, + "loss": 1.0294, + "step": 19355 + }, + { + "epoch": 1.25, + "learning_rate": 1.5868095064299363e-05, + "loss": 0.9843, + "step": 19356 + }, + { + "epoch": 1.25, + "learning_rate": 1.5867880375276415e-05, + "loss": 1.0136, + "step": 19357 + }, + { + "epoch": 1.25, + "learning_rate": 1.5867665686253464e-05, + "loss": 0.972, + "step": 19358 + }, + { + "epoch": 1.25, + "learning_rate": 1.5867450997230513e-05, + "loss": 1.0212, + "step": 19359 + }, + { + "epoch": 1.25, + "learning_rate": 1.5867236308207562e-05, + "loss": 1.021, + "step": 19360 + }, + { + "epoch": 1.25, + "learning_rate": 1.586702161918461e-05, + "loss": 1.0105, + "step": 19361 + }, + { + "epoch": 1.25, + "learning_rate": 1.5866806930161664e-05, + "loss": 0.9396, + "step": 19362 + }, + { + "epoch": 1.25, + "learning_rate": 1.5866592241138713e-05, + "loss": 1.0253, + "step": 19363 + }, + { + "epoch": 1.25, + "learning_rate": 1.586637755211576e-05, + "loss": 0.9023, + "step": 19364 + }, + { + "epoch": 1.25, + "learning_rate": 1.586616286309281e-05, + "loss": 1.002, + "step": 19365 + }, + { + "epoch": 1.25, + "learning_rate": 1.586594817406986e-05, + "loss": 0.9126, + "step": 19366 + }, + { + "epoch": 1.25, + "learning_rate": 1.5865733485046912e-05, + "loss": 1.0002, + "step": 19367 + }, + { + "epoch": 1.25, + "learning_rate": 1.586551879602396e-05, + "loss": 1.0099, + "step": 19368 + }, + { + "epoch": 1.25, + "learning_rate": 1.586530410700101e-05, + "loss": 0.9823, + "step": 19369 + }, + { + "epoch": 1.25, + "learning_rate": 1.586508941797806e-05, + "loss": 0.9989, + "step": 19370 + }, + { + "epoch": 1.25, + "learning_rate": 1.586487472895511e-05, + "loss": 0.9388, + "step": 19371 + }, + { + "epoch": 1.25, + "learning_rate": 1.586466003993216e-05, + "loss": 0.9849, + "step": 19372 + }, + { + "epoch": 1.25, + "learning_rate": 1.586444535090921e-05, + "loss": 1.0104, + "step": 19373 + }, + { + "epoch": 1.25, + "learning_rate": 1.5864230661886258e-05, + "loss": 0.9742, + "step": 19374 + }, + { + "epoch": 1.25, + "learning_rate": 1.586401597286331e-05, + "loss": 1.0115, + "step": 19375 + }, + { + "epoch": 1.25, + "learning_rate": 1.586380128384036e-05, + "loss": 0.9323, + "step": 19376 + }, + { + "epoch": 1.25, + "learning_rate": 1.5863586594817408e-05, + "loss": 0.955, + "step": 19377 + }, + { + "epoch": 1.25, + "learning_rate": 1.5863371905794457e-05, + "loss": 1.0154, + "step": 19378 + }, + { + "epoch": 1.25, + "learning_rate": 1.5863157216771506e-05, + "loss": 0.9899, + "step": 19379 + }, + { + "epoch": 1.25, + "learning_rate": 1.586294252774856e-05, + "loss": 1.0187, + "step": 19380 + }, + { + "epoch": 1.25, + "learning_rate": 1.5862727838725608e-05, + "loss": 0.9958, + "step": 19381 + }, + { + "epoch": 1.25, + "learning_rate": 1.5862513149702657e-05, + "loss": 0.9842, + "step": 19382 + }, + { + "epoch": 1.25, + "learning_rate": 1.5862298460679705e-05, + "loss": 1.0017, + "step": 19383 + }, + { + "epoch": 1.25, + "learning_rate": 1.5862083771656754e-05, + "loss": 1.1224, + "step": 19384 + }, + { + "epoch": 1.25, + "learning_rate": 1.5861869082633807e-05, + "loss": 0.9801, + "step": 19385 + }, + { + "epoch": 1.25, + "learning_rate": 1.5861654393610856e-05, + "loss": 1.0059, + "step": 19386 + }, + { + "epoch": 1.25, + "learning_rate": 1.5861439704587908e-05, + "loss": 0.9382, + "step": 19387 + }, + { + "epoch": 1.25, + "learning_rate": 1.5861225015564957e-05, + "loss": 0.9752, + "step": 19388 + }, + { + "epoch": 1.25, + "learning_rate": 1.5861010326542006e-05, + "loss": 0.946, + "step": 19389 + }, + { + "epoch": 1.25, + "learning_rate": 1.5860795637519055e-05, + "loss": 0.9735, + "step": 19390 + }, + { + "epoch": 1.25, + "learning_rate": 1.5860580948496104e-05, + "loss": 1.0513, + "step": 19391 + }, + { + "epoch": 1.25, + "learning_rate": 1.5860366259473153e-05, + "loss": 0.9539, + "step": 19392 + }, + { + "epoch": 1.25, + "learning_rate": 1.5860151570450205e-05, + "loss": 1.0543, + "step": 19393 + }, + { + "epoch": 1.25, + "learning_rate": 1.5859936881427254e-05, + "loss": 0.9845, + "step": 19394 + }, + { + "epoch": 1.25, + "learning_rate": 1.5859722192404303e-05, + "loss": 0.9685, + "step": 19395 + }, + { + "epoch": 1.25, + "learning_rate": 1.5859507503381352e-05, + "loss": 0.9382, + "step": 19396 + }, + { + "epoch": 1.25, + "learning_rate": 1.58592928143584e-05, + "loss": 0.9138, + "step": 19397 + }, + { + "epoch": 1.25, + "learning_rate": 1.5859078125335454e-05, + "loss": 1.0107, + "step": 19398 + }, + { + "epoch": 1.25, + "learning_rate": 1.5858863436312502e-05, + "loss": 0.9723, + "step": 19399 + }, + { + "epoch": 1.25, + "learning_rate": 1.5858648747289555e-05, + "loss": 0.9719, + "step": 19400 + }, + { + "epoch": 1.25, + "learning_rate": 1.5858434058266604e-05, + "loss": 0.9745, + "step": 19401 + }, + { + "epoch": 1.25, + "learning_rate": 1.5858219369243653e-05, + "loss": 0.9662, + "step": 19402 + }, + { + "epoch": 1.25, + "learning_rate": 1.5858004680220702e-05, + "loss": 0.976, + "step": 19403 + }, + { + "epoch": 1.25, + "learning_rate": 1.585778999119775e-05, + "loss": 1.0009, + "step": 19404 + }, + { + "epoch": 1.25, + "learning_rate": 1.5857575302174803e-05, + "loss": 1.0504, + "step": 19405 + }, + { + "epoch": 1.25, + "learning_rate": 1.5857360613151852e-05, + "loss": 0.9502, + "step": 19406 + }, + { + "epoch": 1.25, + "learning_rate": 1.58571459241289e-05, + "loss": 0.9565, + "step": 19407 + }, + { + "epoch": 1.25, + "learning_rate": 1.585693123510595e-05, + "loss": 0.9847, + "step": 19408 + }, + { + "epoch": 1.25, + "learning_rate": 1.5856716546083e-05, + "loss": 0.9863, + "step": 19409 + }, + { + "epoch": 1.25, + "learning_rate": 1.5856501857060048e-05, + "loss": 0.9573, + "step": 19410 + }, + { + "epoch": 1.25, + "learning_rate": 1.58562871680371e-05, + "loss": 1.0128, + "step": 19411 + }, + { + "epoch": 1.25, + "learning_rate": 1.585607247901415e-05, + "loss": 0.9617, + "step": 19412 + }, + { + "epoch": 1.25, + "learning_rate": 1.5855857789991198e-05, + "loss": 1.0145, + "step": 19413 + }, + { + "epoch": 1.25, + "learning_rate": 1.585564310096825e-05, + "loss": 1.0629, + "step": 19414 + }, + { + "epoch": 1.25, + "learning_rate": 1.58554284119453e-05, + "loss": 0.9444, + "step": 19415 + }, + { + "epoch": 1.25, + "learning_rate": 1.585521372292235e-05, + "loss": 0.9913, + "step": 19416 + }, + { + "epoch": 1.25, + "learning_rate": 1.5854999033899397e-05, + "loss": 0.9972, + "step": 19417 + }, + { + "epoch": 1.25, + "learning_rate": 1.585478434487645e-05, + "loss": 0.9558, + "step": 19418 + }, + { + "epoch": 1.25, + "learning_rate": 1.58545696558535e-05, + "loss": 0.9485, + "step": 19419 + }, + { + "epoch": 1.25, + "learning_rate": 1.5854354966830548e-05, + "loss": 0.9697, + "step": 19420 + }, + { + "epoch": 1.25, + "learning_rate": 1.5854140277807597e-05, + "loss": 0.9953, + "step": 19421 + }, + { + "epoch": 1.25, + "learning_rate": 1.5853925588784646e-05, + "loss": 0.9093, + "step": 19422 + }, + { + "epoch": 1.25, + "learning_rate": 1.5853710899761695e-05, + "loss": 0.9768, + "step": 19423 + }, + { + "epoch": 1.25, + "learning_rate": 1.5853496210738747e-05, + "loss": 1.0163, + "step": 19424 + }, + { + "epoch": 1.25, + "learning_rate": 1.5853281521715796e-05, + "loss": 1.016, + "step": 19425 + }, + { + "epoch": 1.25, + "learning_rate": 1.5853066832692845e-05, + "loss": 1.017, + "step": 19426 + }, + { + "epoch": 1.25, + "learning_rate": 1.5852852143669894e-05, + "loss": 0.987, + "step": 19427 + }, + { + "epoch": 1.25, + "learning_rate": 1.5852637454646943e-05, + "loss": 0.8915, + "step": 19428 + }, + { + "epoch": 1.25, + "learning_rate": 1.5852422765623995e-05, + "loss": 0.9562, + "step": 19429 + }, + { + "epoch": 1.25, + "learning_rate": 1.5852208076601044e-05, + "loss": 0.9725, + "step": 19430 + }, + { + "epoch": 1.25, + "learning_rate": 1.5851993387578096e-05, + "loss": 0.9875, + "step": 19431 + }, + { + "epoch": 1.25, + "learning_rate": 1.5851778698555145e-05, + "loss": 0.9676, + "step": 19432 + }, + { + "epoch": 1.25, + "learning_rate": 1.5851564009532194e-05, + "loss": 0.9247, + "step": 19433 + }, + { + "epoch": 1.25, + "learning_rate": 1.5851349320509243e-05, + "loss": 0.9688, + "step": 19434 + }, + { + "epoch": 1.25, + "learning_rate": 1.5851134631486292e-05, + "loss": 0.996, + "step": 19435 + }, + { + "epoch": 1.25, + "learning_rate": 1.5850919942463345e-05, + "loss": 1.0169, + "step": 19436 + }, + { + "epoch": 1.25, + "learning_rate": 1.5850705253440394e-05, + "loss": 0.9997, + "step": 19437 + }, + { + "epoch": 1.25, + "learning_rate": 1.5850490564417443e-05, + "loss": 1.052, + "step": 19438 + }, + { + "epoch": 1.25, + "learning_rate": 1.585027587539449e-05, + "loss": 1.0028, + "step": 19439 + }, + { + "epoch": 1.25, + "learning_rate": 1.585006118637154e-05, + "loss": 1.0148, + "step": 19440 + }, + { + "epoch": 1.25, + "learning_rate": 1.584984649734859e-05, + "loss": 1.0195, + "step": 19441 + }, + { + "epoch": 1.25, + "learning_rate": 1.5849631808325642e-05, + "loss": 1.0068, + "step": 19442 + }, + { + "epoch": 1.25, + "learning_rate": 1.584941711930269e-05, + "loss": 0.908, + "step": 19443 + }, + { + "epoch": 1.25, + "learning_rate": 1.5849202430279743e-05, + "loss": 0.9642, + "step": 19444 + }, + { + "epoch": 1.25, + "learning_rate": 1.5848987741256792e-05, + "loss": 1.003, + "step": 19445 + }, + { + "epoch": 1.25, + "learning_rate": 1.584877305223384e-05, + "loss": 0.9542, + "step": 19446 + }, + { + "epoch": 1.25, + "learning_rate": 1.584855836321089e-05, + "loss": 0.9012, + "step": 19447 + }, + { + "epoch": 1.25, + "learning_rate": 1.584834367418794e-05, + "loss": 1.0079, + "step": 19448 + }, + { + "epoch": 1.25, + "learning_rate": 1.584812898516499e-05, + "loss": 0.9803, + "step": 19449 + }, + { + "epoch": 1.25, + "learning_rate": 1.584791429614204e-05, + "loss": 1.0238, + "step": 19450 + }, + { + "epoch": 1.25, + "learning_rate": 1.584769960711909e-05, + "loss": 1.0116, + "step": 19451 + }, + { + "epoch": 1.25, + "learning_rate": 1.5847484918096138e-05, + "loss": 0.9631, + "step": 19452 + }, + { + "epoch": 1.25, + "learning_rate": 1.5847270229073187e-05, + "loss": 0.9624, + "step": 19453 + }, + { + "epoch": 1.25, + "learning_rate": 1.584705554005024e-05, + "loss": 0.9816, + "step": 19454 + }, + { + "epoch": 1.25, + "learning_rate": 1.584684085102729e-05, + "loss": 1.0114, + "step": 19455 + }, + { + "epoch": 1.25, + "learning_rate": 1.5846626162004338e-05, + "loss": 0.9252, + "step": 19456 + }, + { + "epoch": 1.25, + "learning_rate": 1.5846411472981386e-05, + "loss": 1.0455, + "step": 19457 + }, + { + "epoch": 1.25, + "learning_rate": 1.584619678395844e-05, + "loss": 0.9405, + "step": 19458 + }, + { + "epoch": 1.25, + "learning_rate": 1.5845982094935488e-05, + "loss": 0.9956, + "step": 19459 + }, + { + "epoch": 1.25, + "learning_rate": 1.5845767405912537e-05, + "loss": 0.9504, + "step": 19460 + }, + { + "epoch": 1.25, + "learning_rate": 1.5845552716889586e-05, + "loss": 0.96, + "step": 19461 + }, + { + "epoch": 1.25, + "learning_rate": 1.5845338027866638e-05, + "loss": 1.0422, + "step": 19462 + }, + { + "epoch": 1.25, + "learning_rate": 1.5845123338843687e-05, + "loss": 0.9724, + "step": 19463 + }, + { + "epoch": 1.25, + "learning_rate": 1.5844908649820736e-05, + "loss": 0.9496, + "step": 19464 + }, + { + "epoch": 1.25, + "learning_rate": 1.5844693960797785e-05, + "loss": 0.9621, + "step": 19465 + }, + { + "epoch": 1.25, + "learning_rate": 1.5844479271774834e-05, + "loss": 0.9672, + "step": 19466 + }, + { + "epoch": 1.25, + "learning_rate": 1.5844264582751886e-05, + "loss": 1.0067, + "step": 19467 + }, + { + "epoch": 1.25, + "learning_rate": 1.5844049893728935e-05, + "loss": 0.9261, + "step": 19468 + }, + { + "epoch": 1.25, + "learning_rate": 1.5843835204705984e-05, + "loss": 0.933, + "step": 19469 + }, + { + "epoch": 1.25, + "learning_rate": 1.5843620515683033e-05, + "loss": 0.9615, + "step": 19470 + }, + { + "epoch": 1.25, + "learning_rate": 1.5843405826660082e-05, + "loss": 0.9078, + "step": 19471 + }, + { + "epoch": 1.25, + "learning_rate": 1.5843191137637135e-05, + "loss": 0.9534, + "step": 19472 + }, + { + "epoch": 1.25, + "learning_rate": 1.5842976448614183e-05, + "loss": 0.9884, + "step": 19473 + }, + { + "epoch": 1.25, + "learning_rate": 1.5842761759591236e-05, + "loss": 1.0198, + "step": 19474 + }, + { + "epoch": 1.25, + "learning_rate": 1.5842547070568285e-05, + "loss": 0.9818, + "step": 19475 + }, + { + "epoch": 1.25, + "learning_rate": 1.5842332381545334e-05, + "loss": 0.8728, + "step": 19476 + }, + { + "epoch": 1.25, + "learning_rate": 1.5842117692522383e-05, + "loss": 1.0088, + "step": 19477 + }, + { + "epoch": 1.25, + "learning_rate": 1.584190300349943e-05, + "loss": 0.9087, + "step": 19478 + }, + { + "epoch": 1.25, + "learning_rate": 1.584168831447648e-05, + "loss": 0.9724, + "step": 19479 + }, + { + "epoch": 1.25, + "learning_rate": 1.5841473625453533e-05, + "loss": 0.9472, + "step": 19480 + }, + { + "epoch": 1.25, + "learning_rate": 1.5841258936430582e-05, + "loss": 0.9857, + "step": 19481 + }, + { + "epoch": 1.25, + "learning_rate": 1.584104424740763e-05, + "loss": 1.01, + "step": 19482 + }, + { + "epoch": 1.25, + "learning_rate": 1.584082955838468e-05, + "loss": 0.9893, + "step": 19483 + }, + { + "epoch": 1.25, + "learning_rate": 1.584061486936173e-05, + "loss": 0.9666, + "step": 19484 + }, + { + "epoch": 1.25, + "learning_rate": 1.584040018033878e-05, + "loss": 0.9608, + "step": 19485 + }, + { + "epoch": 1.25, + "learning_rate": 1.584018549131583e-05, + "loss": 1.0364, + "step": 19486 + }, + { + "epoch": 1.25, + "learning_rate": 1.5839970802292883e-05, + "loss": 1.0438, + "step": 19487 + }, + { + "epoch": 1.25, + "learning_rate": 1.583975611326993e-05, + "loss": 0.9976, + "step": 19488 + }, + { + "epoch": 1.25, + "learning_rate": 1.583954142424698e-05, + "loss": 1.0227, + "step": 19489 + }, + { + "epoch": 1.25, + "learning_rate": 1.583932673522403e-05, + "loss": 0.945, + "step": 19490 + }, + { + "epoch": 1.25, + "learning_rate": 1.583911204620108e-05, + "loss": 0.969, + "step": 19491 + }, + { + "epoch": 1.25, + "learning_rate": 1.5838897357178127e-05, + "loss": 1.0496, + "step": 19492 + }, + { + "epoch": 1.25, + "learning_rate": 1.583868266815518e-05, + "loss": 0.9305, + "step": 19493 + }, + { + "epoch": 1.25, + "learning_rate": 1.583846797913223e-05, + "loss": 1.0278, + "step": 19494 + }, + { + "epoch": 1.25, + "learning_rate": 1.5838253290109278e-05, + "loss": 0.9597, + "step": 19495 + }, + { + "epoch": 1.25, + "learning_rate": 1.5838038601086327e-05, + "loss": 0.9797, + "step": 19496 + }, + { + "epoch": 1.25, + "learning_rate": 1.5837823912063376e-05, + "loss": 1.0139, + "step": 19497 + }, + { + "epoch": 1.25, + "learning_rate": 1.5837609223040428e-05, + "loss": 0.91, + "step": 19498 + }, + { + "epoch": 1.25, + "learning_rate": 1.5837394534017477e-05, + "loss": 1.0009, + "step": 19499 + }, + { + "epoch": 1.25, + "learning_rate": 1.5837179844994526e-05, + "loss": 0.9942, + "step": 19500 + }, + { + "epoch": 1.25, + "learning_rate": 1.5836965155971578e-05, + "loss": 0.928, + "step": 19501 + }, + { + "epoch": 1.25, + "learning_rate": 1.5836750466948627e-05, + "loss": 1.0189, + "step": 19502 + }, + { + "epoch": 1.25, + "learning_rate": 1.5836535777925676e-05, + "loss": 1.0166, + "step": 19503 + }, + { + "epoch": 1.25, + "learning_rate": 1.5836321088902725e-05, + "loss": 0.9418, + "step": 19504 + }, + { + "epoch": 1.25, + "learning_rate": 1.5836106399879777e-05, + "loss": 0.9756, + "step": 19505 + }, + { + "epoch": 1.25, + "learning_rate": 1.5835891710856826e-05, + "loss": 0.9253, + "step": 19506 + }, + { + "epoch": 1.26, + "learning_rate": 1.5835677021833875e-05, + "loss": 0.9784, + "step": 19507 + }, + { + "epoch": 1.26, + "learning_rate": 1.5835462332810924e-05, + "loss": 0.9316, + "step": 19508 + }, + { + "epoch": 1.26, + "learning_rate": 1.5835247643787973e-05, + "loss": 0.9656, + "step": 19509 + }, + { + "epoch": 1.26, + "learning_rate": 1.5835032954765022e-05, + "loss": 1.0507, + "step": 19510 + }, + { + "epoch": 1.26, + "learning_rate": 1.5834818265742075e-05, + "loss": 0.9512, + "step": 19511 + }, + { + "epoch": 1.26, + "learning_rate": 1.5834603576719124e-05, + "loss": 1.0211, + "step": 19512 + }, + { + "epoch": 1.26, + "learning_rate": 1.5834388887696173e-05, + "loss": 0.9524, + "step": 19513 + }, + { + "epoch": 1.26, + "learning_rate": 1.583417419867322e-05, + "loss": 0.9651, + "step": 19514 + }, + { + "epoch": 1.26, + "learning_rate": 1.583395950965027e-05, + "loss": 0.9414, + "step": 19515 + }, + { + "epoch": 1.26, + "learning_rate": 1.5833744820627323e-05, + "loss": 0.9157, + "step": 19516 + }, + { + "epoch": 1.26, + "learning_rate": 1.5833530131604372e-05, + "loss": 0.9399, + "step": 19517 + }, + { + "epoch": 1.26, + "learning_rate": 1.5833315442581424e-05, + "loss": 0.9562, + "step": 19518 + }, + { + "epoch": 1.26, + "learning_rate": 1.5833100753558473e-05, + "loss": 0.9886, + "step": 19519 + }, + { + "epoch": 1.26, + "learning_rate": 1.5832886064535522e-05, + "loss": 0.9694, + "step": 19520 + }, + { + "epoch": 1.26, + "learning_rate": 1.583267137551257e-05, + "loss": 0.9939, + "step": 19521 + }, + { + "epoch": 1.26, + "learning_rate": 1.583245668648962e-05, + "loss": 0.9323, + "step": 19522 + }, + { + "epoch": 1.26, + "learning_rate": 1.5832241997466672e-05, + "loss": 1.0209, + "step": 19523 + }, + { + "epoch": 1.26, + "learning_rate": 1.583202730844372e-05, + "loss": 0.898, + "step": 19524 + }, + { + "epoch": 1.26, + "learning_rate": 1.583181261942077e-05, + "loss": 0.9681, + "step": 19525 + }, + { + "epoch": 1.26, + "learning_rate": 1.583159793039782e-05, + "loss": 0.9895, + "step": 19526 + }, + { + "epoch": 1.26, + "learning_rate": 1.5831383241374868e-05, + "loss": 1.0273, + "step": 19527 + }, + { + "epoch": 1.26, + "learning_rate": 1.5831168552351917e-05, + "loss": 1.0062, + "step": 19528 + }, + { + "epoch": 1.26, + "learning_rate": 1.583095386332897e-05, + "loss": 1.0019, + "step": 19529 + }, + { + "epoch": 1.26, + "learning_rate": 1.583073917430602e-05, + "loss": 0.9247, + "step": 19530 + }, + { + "epoch": 1.26, + "learning_rate": 1.583052448528307e-05, + "loss": 0.9923, + "step": 19531 + }, + { + "epoch": 1.26, + "learning_rate": 1.583030979626012e-05, + "loss": 0.9765, + "step": 19532 + }, + { + "epoch": 1.26, + "learning_rate": 1.583009510723717e-05, + "loss": 0.9408, + "step": 19533 + }, + { + "epoch": 1.26, + "learning_rate": 1.5829880418214218e-05, + "loss": 0.9534, + "step": 19534 + }, + { + "epoch": 1.26, + "learning_rate": 1.5829665729191267e-05, + "loss": 0.9657, + "step": 19535 + }, + { + "epoch": 1.26, + "learning_rate": 1.582945104016832e-05, + "loss": 0.9675, + "step": 19536 + }, + { + "epoch": 1.26, + "learning_rate": 1.5829236351145368e-05, + "loss": 0.9502, + "step": 19537 + }, + { + "epoch": 1.26, + "learning_rate": 1.5829021662122417e-05, + "loss": 1.0129, + "step": 19538 + }, + { + "epoch": 1.26, + "learning_rate": 1.5828806973099466e-05, + "loss": 0.9229, + "step": 19539 + }, + { + "epoch": 1.26, + "learning_rate": 1.5828592284076515e-05, + "loss": 1.0109, + "step": 19540 + }, + { + "epoch": 1.26, + "learning_rate": 1.5828377595053567e-05, + "loss": 0.9889, + "step": 19541 + }, + { + "epoch": 1.26, + "learning_rate": 1.5828162906030616e-05, + "loss": 0.9004, + "step": 19542 + }, + { + "epoch": 1.26, + "learning_rate": 1.5827948217007665e-05, + "loss": 0.9311, + "step": 19543 + }, + { + "epoch": 1.26, + "learning_rate": 1.5827733527984714e-05, + "loss": 0.9842, + "step": 19544 + }, + { + "epoch": 1.26, + "learning_rate": 1.5827518838961767e-05, + "loss": 0.9835, + "step": 19545 + }, + { + "epoch": 1.26, + "learning_rate": 1.5827304149938816e-05, + "loss": 1.0531, + "step": 19546 + }, + { + "epoch": 1.26, + "learning_rate": 1.5827089460915864e-05, + "loss": 0.9333, + "step": 19547 + }, + { + "epoch": 1.26, + "learning_rate": 1.5826874771892913e-05, + "loss": 1.0309, + "step": 19548 + }, + { + "epoch": 1.26, + "learning_rate": 1.5826660082869966e-05, + "loss": 0.9652, + "step": 19549 + }, + { + "epoch": 1.26, + "learning_rate": 1.5826445393847015e-05, + "loss": 1.0035, + "step": 19550 + }, + { + "epoch": 1.26, + "learning_rate": 1.5826230704824064e-05, + "loss": 0.9797, + "step": 19551 + }, + { + "epoch": 1.26, + "learning_rate": 1.5826016015801113e-05, + "loss": 0.9981, + "step": 19552 + }, + { + "epoch": 1.26, + "learning_rate": 1.582580132677816e-05, + "loss": 0.9426, + "step": 19553 + }, + { + "epoch": 1.26, + "learning_rate": 1.5825586637755214e-05, + "loss": 1.0063, + "step": 19554 + }, + { + "epoch": 1.26, + "learning_rate": 1.5825371948732263e-05, + "loss": 0.992, + "step": 19555 + }, + { + "epoch": 1.26, + "learning_rate": 1.5825157259709312e-05, + "loss": 0.9219, + "step": 19556 + }, + { + "epoch": 1.26, + "learning_rate": 1.582494257068636e-05, + "loss": 0.9498, + "step": 19557 + }, + { + "epoch": 1.26, + "learning_rate": 1.582472788166341e-05, + "loss": 0.9453, + "step": 19558 + }, + { + "epoch": 1.26, + "learning_rate": 1.582451319264046e-05, + "loss": 0.9941, + "step": 19559 + }, + { + "epoch": 1.26, + "learning_rate": 1.582429850361751e-05, + "loss": 0.9312, + "step": 19560 + }, + { + "epoch": 1.26, + "learning_rate": 1.5824083814594564e-05, + "loss": 0.9616, + "step": 19561 + }, + { + "epoch": 1.26, + "learning_rate": 1.5823869125571612e-05, + "loss": 1.0622, + "step": 19562 + }, + { + "epoch": 1.26, + "learning_rate": 1.582365443654866e-05, + "loss": 1.0389, + "step": 19563 + }, + { + "epoch": 1.26, + "learning_rate": 1.582343974752571e-05, + "loss": 0.9474, + "step": 19564 + }, + { + "epoch": 1.26, + "learning_rate": 1.582322505850276e-05, + "loss": 0.9972, + "step": 19565 + }, + { + "epoch": 1.26, + "learning_rate": 1.582301036947981e-05, + "loss": 1.0686, + "step": 19566 + }, + { + "epoch": 1.26, + "learning_rate": 1.582279568045686e-05, + "loss": 1.0025, + "step": 19567 + }, + { + "epoch": 1.26, + "learning_rate": 1.582258099143391e-05, + "loss": 0.9817, + "step": 19568 + }, + { + "epoch": 1.26, + "learning_rate": 1.582236630241096e-05, + "loss": 1.0209, + "step": 19569 + }, + { + "epoch": 1.26, + "learning_rate": 1.5822151613388008e-05, + "loss": 1.0312, + "step": 19570 + }, + { + "epoch": 1.26, + "learning_rate": 1.5821936924365057e-05, + "loss": 0.9544, + "step": 19571 + }, + { + "epoch": 1.26, + "learning_rate": 1.582172223534211e-05, + "loss": 1.0639, + "step": 19572 + }, + { + "epoch": 1.26, + "learning_rate": 1.5821507546319158e-05, + "loss": 0.9823, + "step": 19573 + }, + { + "epoch": 1.26, + "learning_rate": 1.582129285729621e-05, + "loss": 0.8938, + "step": 19574 + }, + { + "epoch": 1.26, + "learning_rate": 1.582107816827326e-05, + "loss": 1.0318, + "step": 19575 + }, + { + "epoch": 1.26, + "learning_rate": 1.5820863479250308e-05, + "loss": 0.9687, + "step": 19576 + }, + { + "epoch": 1.26, + "learning_rate": 1.5820648790227357e-05, + "loss": 0.9622, + "step": 19577 + }, + { + "epoch": 1.26, + "learning_rate": 1.5820434101204406e-05, + "loss": 1.0135, + "step": 19578 + }, + { + "epoch": 1.26, + "learning_rate": 1.5820219412181455e-05, + "loss": 0.9895, + "step": 19579 + }, + { + "epoch": 1.26, + "learning_rate": 1.5820004723158507e-05, + "loss": 0.9487, + "step": 19580 + }, + { + "epoch": 1.26, + "learning_rate": 1.5819790034135556e-05, + "loss": 1.0069, + "step": 19581 + }, + { + "epoch": 1.26, + "learning_rate": 1.5819575345112605e-05, + "loss": 0.9487, + "step": 19582 + }, + { + "epoch": 1.26, + "learning_rate": 1.5819360656089654e-05, + "loss": 1.0039, + "step": 19583 + }, + { + "epoch": 1.26, + "learning_rate": 1.5819145967066703e-05, + "loss": 0.9759, + "step": 19584 + }, + { + "epoch": 1.26, + "learning_rate": 1.5818931278043756e-05, + "loss": 0.9841, + "step": 19585 + }, + { + "epoch": 1.26, + "learning_rate": 1.5818716589020805e-05, + "loss": 0.9878, + "step": 19586 + }, + { + "epoch": 1.26, + "learning_rate": 1.5818501899997854e-05, + "loss": 0.9614, + "step": 19587 + }, + { + "epoch": 1.26, + "learning_rate": 1.5818287210974903e-05, + "loss": 0.9224, + "step": 19588 + }, + { + "epoch": 1.26, + "learning_rate": 1.5818072521951955e-05, + "loss": 0.9902, + "step": 19589 + }, + { + "epoch": 1.26, + "learning_rate": 1.5817857832929004e-05, + "loss": 0.9715, + "step": 19590 + }, + { + "epoch": 1.26, + "learning_rate": 1.5817643143906053e-05, + "loss": 0.9643, + "step": 19591 + }, + { + "epoch": 1.26, + "learning_rate": 1.5817428454883105e-05, + "loss": 0.924, + "step": 19592 + }, + { + "epoch": 1.26, + "learning_rate": 1.5817213765860154e-05, + "loss": 1.0238, + "step": 19593 + }, + { + "epoch": 1.26, + "learning_rate": 1.5816999076837203e-05, + "loss": 0.9619, + "step": 19594 + }, + { + "epoch": 1.26, + "learning_rate": 1.5816784387814252e-05, + "loss": 1.0352, + "step": 19595 + }, + { + "epoch": 1.26, + "learning_rate": 1.58165696987913e-05, + "loss": 0.9341, + "step": 19596 + }, + { + "epoch": 1.26, + "learning_rate": 1.581635500976835e-05, + "loss": 0.9903, + "step": 19597 + }, + { + "epoch": 1.26, + "learning_rate": 1.5816140320745402e-05, + "loss": 0.9842, + "step": 19598 + }, + { + "epoch": 1.26, + "learning_rate": 1.581592563172245e-05, + "loss": 1.0623, + "step": 19599 + }, + { + "epoch": 1.26, + "learning_rate": 1.58157109426995e-05, + "loss": 0.9484, + "step": 19600 + }, + { + "epoch": 1.26, + "learning_rate": 1.581549625367655e-05, + "loss": 0.9832, + "step": 19601 + }, + { + "epoch": 1.26, + "learning_rate": 1.5815281564653598e-05, + "loss": 0.955, + "step": 19602 + }, + { + "epoch": 1.26, + "learning_rate": 1.581506687563065e-05, + "loss": 0.9629, + "step": 19603 + }, + { + "epoch": 1.26, + "learning_rate": 1.58148521866077e-05, + "loss": 1.0206, + "step": 19604 + }, + { + "epoch": 1.26, + "learning_rate": 1.5814637497584752e-05, + "loss": 0.935, + "step": 19605 + }, + { + "epoch": 1.26, + "learning_rate": 1.58144228085618e-05, + "loss": 0.9537, + "step": 19606 + }, + { + "epoch": 1.26, + "learning_rate": 1.581420811953885e-05, + "loss": 0.9934, + "step": 19607 + }, + { + "epoch": 1.26, + "learning_rate": 1.58139934305159e-05, + "loss": 0.9334, + "step": 19608 + }, + { + "epoch": 1.26, + "learning_rate": 1.5813778741492948e-05, + "loss": 0.9501, + "step": 19609 + }, + { + "epoch": 1.26, + "learning_rate": 1.581356405247e-05, + "loss": 1.008, + "step": 19610 + }, + { + "epoch": 1.26, + "learning_rate": 1.581334936344705e-05, + "loss": 0.9602, + "step": 19611 + }, + { + "epoch": 1.26, + "learning_rate": 1.5813134674424098e-05, + "loss": 0.903, + "step": 19612 + }, + { + "epoch": 1.26, + "learning_rate": 1.5812919985401147e-05, + "loss": 0.9676, + "step": 19613 + }, + { + "epoch": 1.26, + "learning_rate": 1.5812705296378196e-05, + "loss": 1.0365, + "step": 19614 + }, + { + "epoch": 1.26, + "learning_rate": 1.5812490607355245e-05, + "loss": 0.9921, + "step": 19615 + }, + { + "epoch": 1.26, + "learning_rate": 1.5812275918332297e-05, + "loss": 0.9729, + "step": 19616 + }, + { + "epoch": 1.26, + "learning_rate": 1.5812061229309346e-05, + "loss": 0.9589, + "step": 19617 + }, + { + "epoch": 1.26, + "learning_rate": 1.58118465402864e-05, + "loss": 0.9279, + "step": 19618 + }, + { + "epoch": 1.26, + "learning_rate": 1.5811631851263448e-05, + "loss": 1.0544, + "step": 19619 + }, + { + "epoch": 1.26, + "learning_rate": 1.5811417162240497e-05, + "loss": 1.0652, + "step": 19620 + }, + { + "epoch": 1.26, + "learning_rate": 1.5811202473217545e-05, + "loss": 0.9378, + "step": 19621 + }, + { + "epoch": 1.26, + "learning_rate": 1.5810987784194594e-05, + "loss": 0.9092, + "step": 19622 + }, + { + "epoch": 1.26, + "learning_rate": 1.5810773095171647e-05, + "loss": 0.9558, + "step": 19623 + }, + { + "epoch": 1.26, + "learning_rate": 1.5810558406148696e-05, + "loss": 0.9893, + "step": 19624 + }, + { + "epoch": 1.26, + "learning_rate": 1.5810343717125745e-05, + "loss": 0.9914, + "step": 19625 + }, + { + "epoch": 1.26, + "learning_rate": 1.5810129028102794e-05, + "loss": 1.0105, + "step": 19626 + }, + { + "epoch": 1.26, + "learning_rate": 1.5809914339079843e-05, + "loss": 0.9819, + "step": 19627 + }, + { + "epoch": 1.26, + "learning_rate": 1.5809699650056895e-05, + "loss": 0.9858, + "step": 19628 + }, + { + "epoch": 1.26, + "learning_rate": 1.5809484961033944e-05, + "loss": 1.0182, + "step": 19629 + }, + { + "epoch": 1.26, + "learning_rate": 1.5809270272010993e-05, + "loss": 1.0493, + "step": 19630 + }, + { + "epoch": 1.26, + "learning_rate": 1.5809055582988042e-05, + "loss": 0.9114, + "step": 19631 + }, + { + "epoch": 1.26, + "learning_rate": 1.5808840893965094e-05, + "loss": 0.9941, + "step": 19632 + }, + { + "epoch": 1.26, + "learning_rate": 1.5808626204942143e-05, + "loss": 0.8961, + "step": 19633 + }, + { + "epoch": 1.26, + "learning_rate": 1.5808411515919192e-05, + "loss": 0.9242, + "step": 19634 + }, + { + "epoch": 1.26, + "learning_rate": 1.580819682689624e-05, + "loss": 0.915, + "step": 19635 + }, + { + "epoch": 1.26, + "learning_rate": 1.5807982137873294e-05, + "loss": 1.0385, + "step": 19636 + }, + { + "epoch": 1.26, + "learning_rate": 1.5807767448850342e-05, + "loss": 0.9821, + "step": 19637 + }, + { + "epoch": 1.26, + "learning_rate": 1.580755275982739e-05, + "loss": 1.0139, + "step": 19638 + }, + { + "epoch": 1.26, + "learning_rate": 1.580733807080444e-05, + "loss": 1.012, + "step": 19639 + }, + { + "epoch": 1.26, + "learning_rate": 1.580712338178149e-05, + "loss": 1.0204, + "step": 19640 + }, + { + "epoch": 1.26, + "learning_rate": 1.5806908692758542e-05, + "loss": 1.0197, + "step": 19641 + }, + { + "epoch": 1.26, + "learning_rate": 1.580669400373559e-05, + "loss": 1.0469, + "step": 19642 + }, + { + "epoch": 1.26, + "learning_rate": 1.580647931471264e-05, + "loss": 0.9604, + "step": 19643 + }, + { + "epoch": 1.26, + "learning_rate": 1.580626462568969e-05, + "loss": 1.0251, + "step": 19644 + }, + { + "epoch": 1.26, + "learning_rate": 1.5806049936666738e-05, + "loss": 1.029, + "step": 19645 + }, + { + "epoch": 1.26, + "learning_rate": 1.5805835247643787e-05, + "loss": 0.9558, + "step": 19646 + }, + { + "epoch": 1.26, + "learning_rate": 1.580562055862084e-05, + "loss": 0.9225, + "step": 19647 + }, + { + "epoch": 1.26, + "learning_rate": 1.580540586959789e-05, + "loss": 0.9777, + "step": 19648 + }, + { + "epoch": 1.26, + "learning_rate": 1.580519118057494e-05, + "loss": 0.943, + "step": 19649 + }, + { + "epoch": 1.26, + "learning_rate": 1.580497649155199e-05, + "loss": 0.9817, + "step": 19650 + }, + { + "epoch": 1.26, + "learning_rate": 1.5804761802529038e-05, + "loss": 0.9584, + "step": 19651 + }, + { + "epoch": 1.26, + "learning_rate": 1.5804547113506087e-05, + "loss": 0.9183, + "step": 19652 + }, + { + "epoch": 1.26, + "learning_rate": 1.5804332424483136e-05, + "loss": 0.9636, + "step": 19653 + }, + { + "epoch": 1.26, + "learning_rate": 1.580411773546019e-05, + "loss": 0.9615, + "step": 19654 + }, + { + "epoch": 1.26, + "learning_rate": 1.5803903046437237e-05, + "loss": 0.9034, + "step": 19655 + }, + { + "epoch": 1.26, + "learning_rate": 1.5803688357414286e-05, + "loss": 1.0437, + "step": 19656 + }, + { + "epoch": 1.26, + "learning_rate": 1.5803473668391335e-05, + "loss": 1.0014, + "step": 19657 + }, + { + "epoch": 1.26, + "learning_rate": 1.5803258979368384e-05, + "loss": 0.9684, + "step": 19658 + }, + { + "epoch": 1.26, + "learning_rate": 1.5803044290345437e-05, + "loss": 0.8747, + "step": 19659 + }, + { + "epoch": 1.26, + "learning_rate": 1.5802829601322486e-05, + "loss": 1.0462, + "step": 19660 + }, + { + "epoch": 1.26, + "learning_rate": 1.5802614912299538e-05, + "loss": 0.98, + "step": 19661 + }, + { + "epoch": 1.27, + "learning_rate": 1.5802400223276587e-05, + "loss": 0.9612, + "step": 19662 + }, + { + "epoch": 1.27, + "learning_rate": 1.5802185534253636e-05, + "loss": 0.9086, + "step": 19663 + }, + { + "epoch": 1.27, + "learning_rate": 1.5801970845230685e-05, + "loss": 0.9455, + "step": 19664 + }, + { + "epoch": 1.27, + "learning_rate": 1.5801756156207734e-05, + "loss": 0.9406, + "step": 19665 + }, + { + "epoch": 1.27, + "learning_rate": 1.5801541467184783e-05, + "loss": 0.9169, + "step": 19666 + }, + { + "epoch": 1.27, + "learning_rate": 1.5801326778161835e-05, + "loss": 0.97, + "step": 19667 + }, + { + "epoch": 1.27, + "learning_rate": 1.5801112089138884e-05, + "loss": 0.9187, + "step": 19668 + }, + { + "epoch": 1.27, + "learning_rate": 1.5800897400115933e-05, + "loss": 0.9634, + "step": 19669 + }, + { + "epoch": 1.27, + "learning_rate": 1.5800682711092982e-05, + "loss": 0.9163, + "step": 19670 + }, + { + "epoch": 1.27, + "learning_rate": 1.580046802207003e-05, + "loss": 1.0362, + "step": 19671 + }, + { + "epoch": 1.27, + "learning_rate": 1.5800253333047083e-05, + "loss": 1.0114, + "step": 19672 + }, + { + "epoch": 1.27, + "learning_rate": 1.5800038644024132e-05, + "loss": 1.0464, + "step": 19673 + }, + { + "epoch": 1.27, + "learning_rate": 1.579982395500118e-05, + "loss": 0.9505, + "step": 19674 + }, + { + "epoch": 1.27, + "learning_rate": 1.579960926597823e-05, + "loss": 0.9622, + "step": 19675 + }, + { + "epoch": 1.27, + "learning_rate": 1.5799394576955283e-05, + "loss": 1.0028, + "step": 19676 + }, + { + "epoch": 1.27, + "learning_rate": 1.579917988793233e-05, + "loss": 0.9546, + "step": 19677 + }, + { + "epoch": 1.27, + "learning_rate": 1.579896519890938e-05, + "loss": 0.9777, + "step": 19678 + }, + { + "epoch": 1.27, + "learning_rate": 1.5798750509886433e-05, + "loss": 0.9591, + "step": 19679 + }, + { + "epoch": 1.27, + "learning_rate": 1.5798535820863482e-05, + "loss": 0.9089, + "step": 19680 + }, + { + "epoch": 1.27, + "learning_rate": 1.579832113184053e-05, + "loss": 1.0209, + "step": 19681 + }, + { + "epoch": 1.27, + "learning_rate": 1.579810644281758e-05, + "loss": 0.926, + "step": 19682 + }, + { + "epoch": 1.27, + "learning_rate": 1.579789175379463e-05, + "loss": 0.9523, + "step": 19683 + }, + { + "epoch": 1.27, + "learning_rate": 1.5797677064771678e-05, + "loss": 1.017, + "step": 19684 + }, + { + "epoch": 1.27, + "learning_rate": 1.579746237574873e-05, + "loss": 0.9975, + "step": 19685 + }, + { + "epoch": 1.27, + "learning_rate": 1.579724768672578e-05, + "loss": 1.0115, + "step": 19686 + }, + { + "epoch": 1.27, + "learning_rate": 1.5797032997702828e-05, + "loss": 1.0049, + "step": 19687 + }, + { + "epoch": 1.27, + "learning_rate": 1.5796818308679877e-05, + "loss": 1.0426, + "step": 19688 + }, + { + "epoch": 1.27, + "learning_rate": 1.5796603619656926e-05, + "loss": 1.0008, + "step": 19689 + }, + { + "epoch": 1.27, + "learning_rate": 1.5796388930633978e-05, + "loss": 0.9938, + "step": 19690 + }, + { + "epoch": 1.27, + "learning_rate": 1.5796174241611027e-05, + "loss": 0.9715, + "step": 19691 + }, + { + "epoch": 1.27, + "learning_rate": 1.579595955258808e-05, + "loss": 0.94, + "step": 19692 + }, + { + "epoch": 1.27, + "learning_rate": 1.579574486356513e-05, + "loss": 0.9164, + "step": 19693 + }, + { + "epoch": 1.27, + "learning_rate": 1.5795530174542178e-05, + "loss": 0.9396, + "step": 19694 + }, + { + "epoch": 1.27, + "learning_rate": 1.5795315485519226e-05, + "loss": 0.9852, + "step": 19695 + }, + { + "epoch": 1.27, + "learning_rate": 1.5795100796496275e-05, + "loss": 1.003, + "step": 19696 + }, + { + "epoch": 1.27, + "learning_rate": 1.5794886107473328e-05, + "loss": 1.0135, + "step": 19697 + }, + { + "epoch": 1.27, + "learning_rate": 1.5794671418450377e-05, + "loss": 0.9784, + "step": 19698 + }, + { + "epoch": 1.27, + "learning_rate": 1.5794456729427426e-05, + "loss": 1.0154, + "step": 19699 + }, + { + "epoch": 1.27, + "learning_rate": 1.5794242040404475e-05, + "loss": 0.9384, + "step": 19700 + }, + { + "epoch": 1.27, + "learning_rate": 1.5794027351381524e-05, + "loss": 1.0166, + "step": 19701 + }, + { + "epoch": 1.27, + "learning_rate": 1.5793812662358573e-05, + "loss": 0.8923, + "step": 19702 + }, + { + "epoch": 1.27, + "learning_rate": 1.5793597973335625e-05, + "loss": 0.9941, + "step": 19703 + }, + { + "epoch": 1.27, + "learning_rate": 1.5793383284312674e-05, + "loss": 0.9791, + "step": 19704 + }, + { + "epoch": 1.27, + "learning_rate": 1.5793168595289726e-05, + "loss": 0.9765, + "step": 19705 + }, + { + "epoch": 1.27, + "learning_rate": 1.5792953906266775e-05, + "loss": 0.9692, + "step": 19706 + }, + { + "epoch": 1.27, + "learning_rate": 1.5792739217243824e-05, + "loss": 1.0305, + "step": 19707 + }, + { + "epoch": 1.27, + "learning_rate": 1.5792524528220873e-05, + "loss": 0.9556, + "step": 19708 + }, + { + "epoch": 1.27, + "learning_rate": 1.5792309839197922e-05, + "loss": 0.9227, + "step": 19709 + }, + { + "epoch": 1.27, + "learning_rate": 1.5792095150174975e-05, + "loss": 1.0151, + "step": 19710 + }, + { + "epoch": 1.27, + "learning_rate": 1.5791880461152023e-05, + "loss": 0.9705, + "step": 19711 + }, + { + "epoch": 1.27, + "learning_rate": 1.5791665772129072e-05, + "loss": 0.9245, + "step": 19712 + }, + { + "epoch": 1.27, + "learning_rate": 1.579145108310612e-05, + "loss": 1.0431, + "step": 19713 + }, + { + "epoch": 1.27, + "learning_rate": 1.579123639408317e-05, + "loss": 0.9101, + "step": 19714 + }, + { + "epoch": 1.27, + "learning_rate": 1.5791021705060223e-05, + "loss": 0.9928, + "step": 19715 + }, + { + "epoch": 1.27, + "learning_rate": 1.579080701603727e-05, + "loss": 0.9909, + "step": 19716 + }, + { + "epoch": 1.27, + "learning_rate": 1.579059232701432e-05, + "loss": 0.9614, + "step": 19717 + }, + { + "epoch": 1.27, + "learning_rate": 1.579037763799137e-05, + "loss": 1.0071, + "step": 19718 + }, + { + "epoch": 1.27, + "learning_rate": 1.5790162948968422e-05, + "loss": 0.954, + "step": 19719 + }, + { + "epoch": 1.27, + "learning_rate": 1.578994825994547e-05, + "loss": 0.9789, + "step": 19720 + }, + { + "epoch": 1.27, + "learning_rate": 1.578973357092252e-05, + "loss": 0.9226, + "step": 19721 + }, + { + "epoch": 1.27, + "learning_rate": 1.578951888189957e-05, + "loss": 0.91, + "step": 19722 + }, + { + "epoch": 1.27, + "learning_rate": 1.578930419287662e-05, + "loss": 0.9473, + "step": 19723 + }, + { + "epoch": 1.27, + "learning_rate": 1.578908950385367e-05, + "loss": 0.9815, + "step": 19724 + }, + { + "epoch": 1.27, + "learning_rate": 1.578887481483072e-05, + "loss": 1.0316, + "step": 19725 + }, + { + "epoch": 1.27, + "learning_rate": 1.5788660125807768e-05, + "loss": 1.0429, + "step": 19726 + }, + { + "epoch": 1.27, + "learning_rate": 1.5788445436784817e-05, + "loss": 0.9168, + "step": 19727 + }, + { + "epoch": 1.27, + "learning_rate": 1.578823074776187e-05, + "loss": 0.9193, + "step": 19728 + }, + { + "epoch": 1.27, + "learning_rate": 1.578801605873892e-05, + "loss": 0.93, + "step": 19729 + }, + { + "epoch": 1.27, + "learning_rate": 1.5787801369715967e-05, + "loss": 1.0064, + "step": 19730 + }, + { + "epoch": 1.27, + "learning_rate": 1.5787586680693016e-05, + "loss": 0.9474, + "step": 19731 + }, + { + "epoch": 1.27, + "learning_rate": 1.5787371991670065e-05, + "loss": 0.9669, + "step": 19732 + }, + { + "epoch": 1.27, + "learning_rate": 1.5787157302647114e-05, + "loss": 0.9312, + "step": 19733 + }, + { + "epoch": 1.27, + "learning_rate": 1.5786942613624167e-05, + "loss": 0.9037, + "step": 19734 + }, + { + "epoch": 1.27, + "learning_rate": 1.578672792460122e-05, + "loss": 0.905, + "step": 19735 + }, + { + "epoch": 1.27, + "learning_rate": 1.5786513235578268e-05, + "loss": 0.9823, + "step": 19736 + }, + { + "epoch": 1.27, + "learning_rate": 1.5786298546555317e-05, + "loss": 0.959, + "step": 19737 + }, + { + "epoch": 1.27, + "learning_rate": 1.5786083857532366e-05, + "loss": 0.9822, + "step": 19738 + }, + { + "epoch": 1.27, + "learning_rate": 1.5785869168509415e-05, + "loss": 0.9954, + "step": 19739 + }, + { + "epoch": 1.27, + "learning_rate": 1.5785654479486464e-05, + "loss": 0.9784, + "step": 19740 + }, + { + "epoch": 1.27, + "learning_rate": 1.5785439790463516e-05, + "loss": 0.8957, + "step": 19741 + }, + { + "epoch": 1.27, + "learning_rate": 1.5785225101440565e-05, + "loss": 1.0419, + "step": 19742 + }, + { + "epoch": 1.27, + "learning_rate": 1.5785010412417614e-05, + "loss": 0.9709, + "step": 19743 + }, + { + "epoch": 1.27, + "learning_rate": 1.5784795723394663e-05, + "loss": 1.0433, + "step": 19744 + }, + { + "epoch": 1.27, + "learning_rate": 1.5784581034371712e-05, + "loss": 0.9465, + "step": 19745 + }, + { + "epoch": 1.27, + "learning_rate": 1.5784366345348764e-05, + "loss": 0.9196, + "step": 19746 + }, + { + "epoch": 1.27, + "learning_rate": 1.5784151656325813e-05, + "loss": 0.894, + "step": 19747 + }, + { + "epoch": 1.27, + "learning_rate": 1.5783936967302866e-05, + "loss": 0.9548, + "step": 19748 + }, + { + "epoch": 1.27, + "learning_rate": 1.5783722278279915e-05, + "loss": 0.8474, + "step": 19749 + }, + { + "epoch": 1.27, + "learning_rate": 1.5783507589256964e-05, + "loss": 1.0543, + "step": 19750 + }, + { + "epoch": 1.27, + "learning_rate": 1.5783292900234013e-05, + "loss": 1.0069, + "step": 19751 + }, + { + "epoch": 1.27, + "learning_rate": 1.578307821121106e-05, + "loss": 0.9325, + "step": 19752 + }, + { + "epoch": 1.27, + "learning_rate": 1.578286352218811e-05, + "loss": 1.0226, + "step": 19753 + }, + { + "epoch": 1.27, + "learning_rate": 1.5782648833165163e-05, + "loss": 0.957, + "step": 19754 + }, + { + "epoch": 1.27, + "learning_rate": 1.5782434144142212e-05, + "loss": 0.9911, + "step": 19755 + }, + { + "epoch": 1.27, + "learning_rate": 1.578221945511926e-05, + "loss": 0.9606, + "step": 19756 + }, + { + "epoch": 1.27, + "learning_rate": 1.578200476609631e-05, + "loss": 0.9296, + "step": 19757 + }, + { + "epoch": 1.27, + "learning_rate": 1.578179007707336e-05, + "loss": 0.9794, + "step": 19758 + }, + { + "epoch": 1.27, + "learning_rate": 1.578157538805041e-05, + "loss": 0.9351, + "step": 19759 + }, + { + "epoch": 1.27, + "learning_rate": 1.578136069902746e-05, + "loss": 0.9553, + "step": 19760 + }, + { + "epoch": 1.27, + "learning_rate": 1.578114601000451e-05, + "loss": 1.0007, + "step": 19761 + }, + { + "epoch": 1.27, + "learning_rate": 1.5780931320981558e-05, + "loss": 1.0108, + "step": 19762 + }, + { + "epoch": 1.27, + "learning_rate": 1.578071663195861e-05, + "loss": 0.9843, + "step": 19763 + }, + { + "epoch": 1.27, + "learning_rate": 1.578050194293566e-05, + "loss": 1.0215, + "step": 19764 + }, + { + "epoch": 1.27, + "learning_rate": 1.5780287253912708e-05, + "loss": 0.9861, + "step": 19765 + }, + { + "epoch": 1.27, + "learning_rate": 1.578007256488976e-05, + "loss": 1.0335, + "step": 19766 + }, + { + "epoch": 1.27, + "learning_rate": 1.577985787586681e-05, + "loss": 1.0298, + "step": 19767 + }, + { + "epoch": 1.27, + "learning_rate": 1.577964318684386e-05, + "loss": 0.8717, + "step": 19768 + }, + { + "epoch": 1.27, + "learning_rate": 1.5779428497820907e-05, + "loss": 0.9465, + "step": 19769 + }, + { + "epoch": 1.27, + "learning_rate": 1.5779213808797956e-05, + "loss": 1.0624, + "step": 19770 + }, + { + "epoch": 1.27, + "learning_rate": 1.5778999119775005e-05, + "loss": 0.9429, + "step": 19771 + }, + { + "epoch": 1.27, + "learning_rate": 1.5778784430752058e-05, + "loss": 0.9277, + "step": 19772 + }, + { + "epoch": 1.27, + "learning_rate": 1.5778569741729107e-05, + "loss": 0.9632, + "step": 19773 + }, + { + "epoch": 1.27, + "learning_rate": 1.5778355052706156e-05, + "loss": 0.9613, + "step": 19774 + }, + { + "epoch": 1.27, + "learning_rate": 1.5778140363683205e-05, + "loss": 0.8813, + "step": 19775 + }, + { + "epoch": 1.27, + "learning_rate": 1.5777925674660254e-05, + "loss": 1.0061, + "step": 19776 + }, + { + "epoch": 1.27, + "learning_rate": 1.5777710985637306e-05, + "loss": 0.9712, + "step": 19777 + }, + { + "epoch": 1.27, + "learning_rate": 1.5777496296614355e-05, + "loss": 0.9537, + "step": 19778 + }, + { + "epoch": 1.27, + "learning_rate": 1.5777281607591407e-05, + "loss": 0.9764, + "step": 19779 + }, + { + "epoch": 1.27, + "learning_rate": 1.5777066918568456e-05, + "loss": 0.9838, + "step": 19780 + }, + { + "epoch": 1.27, + "learning_rate": 1.5776852229545505e-05, + "loss": 1.0275, + "step": 19781 + }, + { + "epoch": 1.27, + "learning_rate": 1.5776637540522554e-05, + "loss": 0.9723, + "step": 19782 + }, + { + "epoch": 1.27, + "learning_rate": 1.5776422851499603e-05, + "loss": 0.9396, + "step": 19783 + }, + { + "epoch": 1.27, + "learning_rate": 1.5776208162476656e-05, + "loss": 0.8919, + "step": 19784 + }, + { + "epoch": 1.27, + "learning_rate": 1.5775993473453704e-05, + "loss": 0.9946, + "step": 19785 + }, + { + "epoch": 1.27, + "learning_rate": 1.5775778784430753e-05, + "loss": 1.019, + "step": 19786 + }, + { + "epoch": 1.27, + "learning_rate": 1.5775564095407802e-05, + "loss": 1.0331, + "step": 19787 + }, + { + "epoch": 1.27, + "learning_rate": 1.577534940638485e-05, + "loss": 0.9614, + "step": 19788 + }, + { + "epoch": 1.27, + "learning_rate": 1.57751347173619e-05, + "loss": 1.0626, + "step": 19789 + }, + { + "epoch": 1.27, + "learning_rate": 1.5774920028338953e-05, + "loss": 0.917, + "step": 19790 + }, + { + "epoch": 1.27, + "learning_rate": 1.5774705339316e-05, + "loss": 0.9608, + "step": 19791 + }, + { + "epoch": 1.27, + "learning_rate": 1.5774490650293054e-05, + "loss": 1.0488, + "step": 19792 + }, + { + "epoch": 1.27, + "learning_rate": 1.5774275961270103e-05, + "loss": 0.9653, + "step": 19793 + }, + { + "epoch": 1.27, + "learning_rate": 1.5774061272247152e-05, + "loss": 0.927, + "step": 19794 + }, + { + "epoch": 1.27, + "learning_rate": 1.57738465832242e-05, + "loss": 0.9407, + "step": 19795 + }, + { + "epoch": 1.27, + "learning_rate": 1.577363189420125e-05, + "loss": 1.0402, + "step": 19796 + }, + { + "epoch": 1.27, + "learning_rate": 1.5773417205178302e-05, + "loss": 1.0836, + "step": 19797 + }, + { + "epoch": 1.27, + "learning_rate": 1.577320251615535e-05, + "loss": 1.0049, + "step": 19798 + }, + { + "epoch": 1.27, + "learning_rate": 1.57729878271324e-05, + "loss": 0.985, + "step": 19799 + }, + { + "epoch": 1.27, + "learning_rate": 1.577277313810945e-05, + "loss": 1.0535, + "step": 19800 + }, + { + "epoch": 1.27, + "learning_rate": 1.5772558449086498e-05, + "loss": 1.0031, + "step": 19801 + }, + { + "epoch": 1.27, + "learning_rate": 1.5772343760063547e-05, + "loss": 0.9926, + "step": 19802 + }, + { + "epoch": 1.27, + "learning_rate": 1.57721290710406e-05, + "loss": 0.9393, + "step": 19803 + }, + { + "epoch": 1.27, + "learning_rate": 1.577191438201765e-05, + "loss": 0.9201, + "step": 19804 + }, + { + "epoch": 1.27, + "learning_rate": 1.5771699692994697e-05, + "loss": 0.9293, + "step": 19805 + }, + { + "epoch": 1.27, + "learning_rate": 1.577148500397175e-05, + "loss": 0.926, + "step": 19806 + }, + { + "epoch": 1.27, + "learning_rate": 1.57712703149488e-05, + "loss": 1.03, + "step": 19807 + }, + { + "epoch": 1.27, + "learning_rate": 1.5771055625925848e-05, + "loss": 0.9888, + "step": 19808 + }, + { + "epoch": 1.27, + "learning_rate": 1.5770840936902897e-05, + "loss": 0.9607, + "step": 19809 + }, + { + "epoch": 1.27, + "learning_rate": 1.577062624787995e-05, + "loss": 1.0414, + "step": 19810 + }, + { + "epoch": 1.27, + "learning_rate": 1.5770411558856998e-05, + "loss": 0.9307, + "step": 19811 + }, + { + "epoch": 1.27, + "learning_rate": 1.5770196869834047e-05, + "loss": 0.9358, + "step": 19812 + }, + { + "epoch": 1.27, + "learning_rate": 1.5769982180811096e-05, + "loss": 1.0218, + "step": 19813 + }, + { + "epoch": 1.27, + "learning_rate": 1.5769767491788145e-05, + "loss": 0.9656, + "step": 19814 + }, + { + "epoch": 1.27, + "learning_rate": 1.5769552802765197e-05, + "loss": 1.079, + "step": 19815 + }, + { + "epoch": 1.27, + "learning_rate": 1.5769338113742246e-05, + "loss": 1.0484, + "step": 19816 + }, + { + "epoch": 1.27, + "learning_rate": 1.5769123424719295e-05, + "loss": 0.9882, + "step": 19817 + }, + { + "epoch": 1.28, + "learning_rate": 1.5768908735696344e-05, + "loss": 0.9666, + "step": 19818 + }, + { + "epoch": 1.28, + "learning_rate": 1.5768694046673393e-05, + "loss": 1.0209, + "step": 19819 + }, + { + "epoch": 1.28, + "learning_rate": 1.5768479357650442e-05, + "loss": 0.9488, + "step": 19820 + }, + { + "epoch": 1.28, + "learning_rate": 1.5768264668627494e-05, + "loss": 0.9521, + "step": 19821 + }, + { + "epoch": 1.28, + "learning_rate": 1.5768049979604547e-05, + "loss": 1.0114, + "step": 19822 + }, + { + "epoch": 1.28, + "learning_rate": 1.5767835290581596e-05, + "loss": 0.9758, + "step": 19823 + }, + { + "epoch": 1.28, + "learning_rate": 1.5767620601558645e-05, + "loss": 1.0242, + "step": 19824 + }, + { + "epoch": 1.28, + "learning_rate": 1.5767405912535694e-05, + "loss": 1.0401, + "step": 19825 + }, + { + "epoch": 1.28, + "learning_rate": 1.5767191223512743e-05, + "loss": 0.9915, + "step": 19826 + }, + { + "epoch": 1.28, + "learning_rate": 1.576697653448979e-05, + "loss": 1.008, + "step": 19827 + }, + { + "epoch": 1.28, + "learning_rate": 1.5766761845466844e-05, + "loss": 0.9363, + "step": 19828 + }, + { + "epoch": 1.28, + "learning_rate": 1.5766547156443893e-05, + "loss": 0.9849, + "step": 19829 + }, + { + "epoch": 1.28, + "learning_rate": 1.5766332467420942e-05, + "loss": 0.9838, + "step": 19830 + }, + { + "epoch": 1.28, + "learning_rate": 1.576611777839799e-05, + "loss": 0.9801, + "step": 19831 + }, + { + "epoch": 1.28, + "learning_rate": 1.576590308937504e-05, + "loss": 0.9959, + "step": 19832 + }, + { + "epoch": 1.28, + "learning_rate": 1.5765688400352092e-05, + "loss": 0.9785, + "step": 19833 + }, + { + "epoch": 1.28, + "learning_rate": 1.576547371132914e-05, + "loss": 0.9766, + "step": 19834 + }, + { + "epoch": 1.28, + "learning_rate": 1.5765259022306193e-05, + "loss": 1.0248, + "step": 19835 + }, + { + "epoch": 1.28, + "learning_rate": 1.5765044333283242e-05, + "loss": 0.9713, + "step": 19836 + }, + { + "epoch": 1.28, + "learning_rate": 1.576482964426029e-05, + "loss": 1.0205, + "step": 19837 + }, + { + "epoch": 1.28, + "learning_rate": 1.576461495523734e-05, + "loss": 0.9312, + "step": 19838 + }, + { + "epoch": 1.28, + "learning_rate": 1.576440026621439e-05, + "loss": 0.9641, + "step": 19839 + }, + { + "epoch": 1.28, + "learning_rate": 1.5764185577191438e-05, + "loss": 0.9807, + "step": 19840 + }, + { + "epoch": 1.28, + "learning_rate": 1.576397088816849e-05, + "loss": 0.9303, + "step": 19841 + }, + { + "epoch": 1.28, + "learning_rate": 1.576375619914554e-05, + "loss": 0.9653, + "step": 19842 + }, + { + "epoch": 1.28, + "learning_rate": 1.576354151012259e-05, + "loss": 1.0017, + "step": 19843 + }, + { + "epoch": 1.28, + "learning_rate": 1.5763326821099637e-05, + "loss": 1.0167, + "step": 19844 + }, + { + "epoch": 1.28, + "learning_rate": 1.5763112132076686e-05, + "loss": 0.9512, + "step": 19845 + }, + { + "epoch": 1.28, + "learning_rate": 1.576289744305374e-05, + "loss": 0.9838, + "step": 19846 + }, + { + "epoch": 1.28, + "learning_rate": 1.5762682754030788e-05, + "loss": 0.9848, + "step": 19847 + }, + { + "epoch": 1.28, + "learning_rate": 1.5762468065007837e-05, + "loss": 0.9518, + "step": 19848 + }, + { + "epoch": 1.28, + "learning_rate": 1.5762253375984886e-05, + "loss": 1.0147, + "step": 19849 + }, + { + "epoch": 1.28, + "learning_rate": 1.5762038686961938e-05, + "loss": 0.9419, + "step": 19850 + }, + { + "epoch": 1.28, + "learning_rate": 1.5761823997938987e-05, + "loss": 1.0367, + "step": 19851 + }, + { + "epoch": 1.28, + "learning_rate": 1.5761609308916036e-05, + "loss": 0.9995, + "step": 19852 + }, + { + "epoch": 1.28, + "learning_rate": 1.5761394619893088e-05, + "loss": 0.9664, + "step": 19853 + }, + { + "epoch": 1.28, + "learning_rate": 1.5761179930870137e-05, + "loss": 0.9994, + "step": 19854 + }, + { + "epoch": 1.28, + "learning_rate": 1.5760965241847186e-05, + "loss": 1.0317, + "step": 19855 + }, + { + "epoch": 1.28, + "learning_rate": 1.5760750552824235e-05, + "loss": 0.9843, + "step": 19856 + }, + { + "epoch": 1.28, + "learning_rate": 1.5760535863801284e-05, + "loss": 0.9681, + "step": 19857 + }, + { + "epoch": 1.28, + "learning_rate": 1.5760321174778333e-05, + "loss": 0.9943, + "step": 19858 + }, + { + "epoch": 1.28, + "learning_rate": 1.5760106485755385e-05, + "loss": 0.9715, + "step": 19859 + }, + { + "epoch": 1.28, + "learning_rate": 1.5759891796732434e-05, + "loss": 0.997, + "step": 19860 + }, + { + "epoch": 1.28, + "learning_rate": 1.5759677107709483e-05, + "loss": 1.0092, + "step": 19861 + }, + { + "epoch": 1.28, + "learning_rate": 1.5759462418686532e-05, + "loss": 0.9524, + "step": 19862 + }, + { + "epoch": 1.28, + "learning_rate": 1.575924772966358e-05, + "loss": 0.9942, + "step": 19863 + }, + { + "epoch": 1.28, + "learning_rate": 1.5759033040640634e-05, + "loss": 0.8684, + "step": 19864 + }, + { + "epoch": 1.28, + "learning_rate": 1.5758818351617683e-05, + "loss": 1.0231, + "step": 19865 + }, + { + "epoch": 1.28, + "learning_rate": 1.5758603662594735e-05, + "loss": 0.8998, + "step": 19866 + }, + { + "epoch": 1.28, + "learning_rate": 1.5758388973571784e-05, + "loss": 0.9744, + "step": 19867 + }, + { + "epoch": 1.28, + "learning_rate": 1.5758174284548833e-05, + "loss": 1.0112, + "step": 19868 + }, + { + "epoch": 1.28, + "learning_rate": 1.5757959595525882e-05, + "loss": 1.0269, + "step": 19869 + }, + { + "epoch": 1.28, + "learning_rate": 1.575774490650293e-05, + "loss": 0.9558, + "step": 19870 + }, + { + "epoch": 1.28, + "learning_rate": 1.5757530217479983e-05, + "loss": 1.0098, + "step": 19871 + }, + { + "epoch": 1.28, + "learning_rate": 1.5757315528457032e-05, + "loss": 0.9704, + "step": 19872 + }, + { + "epoch": 1.28, + "learning_rate": 1.575710083943408e-05, + "loss": 1.0282, + "step": 19873 + }, + { + "epoch": 1.28, + "learning_rate": 1.575688615041113e-05, + "loss": 0.9184, + "step": 19874 + }, + { + "epoch": 1.28, + "learning_rate": 1.575667146138818e-05, + "loss": 0.9775, + "step": 19875 + }, + { + "epoch": 1.28, + "learning_rate": 1.5756456772365228e-05, + "loss": 0.9645, + "step": 19876 + }, + { + "epoch": 1.28, + "learning_rate": 1.575624208334228e-05, + "loss": 0.9953, + "step": 19877 + }, + { + "epoch": 1.28, + "learning_rate": 1.575602739431933e-05, + "loss": 0.9147, + "step": 19878 + }, + { + "epoch": 1.28, + "learning_rate": 1.5755812705296382e-05, + "loss": 0.9608, + "step": 19879 + }, + { + "epoch": 1.28, + "learning_rate": 1.575559801627343e-05, + "loss": 0.9115, + "step": 19880 + }, + { + "epoch": 1.28, + "learning_rate": 1.575538332725048e-05, + "loss": 0.8861, + "step": 19881 + }, + { + "epoch": 1.28, + "learning_rate": 1.575516863822753e-05, + "loss": 1.0303, + "step": 19882 + }, + { + "epoch": 1.28, + "learning_rate": 1.5754953949204578e-05, + "loss": 0.8944, + "step": 19883 + }, + { + "epoch": 1.28, + "learning_rate": 1.575473926018163e-05, + "loss": 0.942, + "step": 19884 + }, + { + "epoch": 1.28, + "learning_rate": 1.575452457115868e-05, + "loss": 0.9862, + "step": 19885 + }, + { + "epoch": 1.28, + "learning_rate": 1.5754309882135728e-05, + "loss": 0.957, + "step": 19886 + }, + { + "epoch": 1.28, + "learning_rate": 1.5754095193112777e-05, + "loss": 0.9577, + "step": 19887 + }, + { + "epoch": 1.28, + "learning_rate": 1.5753880504089826e-05, + "loss": 0.9787, + "step": 19888 + }, + { + "epoch": 1.28, + "learning_rate": 1.5753665815066875e-05, + "loss": 1.0806, + "step": 19889 + }, + { + "epoch": 1.28, + "learning_rate": 1.5753451126043927e-05, + "loss": 0.9966, + "step": 19890 + }, + { + "epoch": 1.28, + "learning_rate": 1.5753236437020976e-05, + "loss": 0.9933, + "step": 19891 + }, + { + "epoch": 1.28, + "learning_rate": 1.5753021747998025e-05, + "loss": 0.947, + "step": 19892 + }, + { + "epoch": 1.28, + "learning_rate": 1.5752807058975074e-05, + "loss": 1.0435, + "step": 19893 + }, + { + "epoch": 1.28, + "learning_rate": 1.5752592369952126e-05, + "loss": 1.0661, + "step": 19894 + }, + { + "epoch": 1.28, + "learning_rate": 1.5752377680929175e-05, + "loss": 0.8903, + "step": 19895 + }, + { + "epoch": 1.28, + "learning_rate": 1.5752162991906224e-05, + "loss": 0.9374, + "step": 19896 + }, + { + "epoch": 1.28, + "learning_rate": 1.5751948302883277e-05, + "loss": 1.0291, + "step": 19897 + }, + { + "epoch": 1.28, + "learning_rate": 1.5751733613860326e-05, + "loss": 0.9739, + "step": 19898 + }, + { + "epoch": 1.28, + "learning_rate": 1.5751518924837375e-05, + "loss": 0.9413, + "step": 19899 + }, + { + "epoch": 1.28, + "learning_rate": 1.5751304235814424e-05, + "loss": 1.0467, + "step": 19900 + }, + { + "epoch": 1.28, + "learning_rate": 1.5751089546791472e-05, + "loss": 0.9057, + "step": 19901 + }, + { + "epoch": 1.28, + "learning_rate": 1.5750874857768525e-05, + "loss": 1.0443, + "step": 19902 + }, + { + "epoch": 1.28, + "learning_rate": 1.5750660168745574e-05, + "loss": 0.9722, + "step": 19903 + }, + { + "epoch": 1.28, + "learning_rate": 1.5750445479722623e-05, + "loss": 0.9522, + "step": 19904 + }, + { + "epoch": 1.28, + "learning_rate": 1.5750230790699672e-05, + "loss": 0.9878, + "step": 19905 + }, + { + "epoch": 1.28, + "learning_rate": 1.575001610167672e-05, + "loss": 0.9874, + "step": 19906 + }, + { + "epoch": 1.28, + "learning_rate": 1.574980141265377e-05, + "loss": 0.9201, + "step": 19907 + }, + { + "epoch": 1.28, + "learning_rate": 1.5749586723630822e-05, + "loss": 0.8882, + "step": 19908 + }, + { + "epoch": 1.28, + "learning_rate": 1.574937203460787e-05, + "loss": 1.02, + "step": 19909 + }, + { + "epoch": 1.28, + "learning_rate": 1.5749157345584923e-05, + "loss": 0.9135, + "step": 19910 + }, + { + "epoch": 1.28, + "learning_rate": 1.5748942656561972e-05, + "loss": 0.9422, + "step": 19911 + }, + { + "epoch": 1.28, + "learning_rate": 1.574872796753902e-05, + "loss": 0.9769, + "step": 19912 + }, + { + "epoch": 1.28, + "learning_rate": 1.574851327851607e-05, + "loss": 0.9751, + "step": 19913 + }, + { + "epoch": 1.28, + "learning_rate": 1.574829858949312e-05, + "loss": 0.9817, + "step": 19914 + }, + { + "epoch": 1.28, + "learning_rate": 1.574808390047017e-05, + "loss": 1.0416, + "step": 19915 + }, + { + "epoch": 1.28, + "learning_rate": 1.574786921144722e-05, + "loss": 0.99, + "step": 19916 + }, + { + "epoch": 1.28, + "learning_rate": 1.574765452242427e-05, + "loss": 1.0381, + "step": 19917 + }, + { + "epoch": 1.28, + "learning_rate": 1.574743983340132e-05, + "loss": 0.9588, + "step": 19918 + }, + { + "epoch": 1.28, + "learning_rate": 1.5747225144378367e-05, + "loss": 0.8857, + "step": 19919 + }, + { + "epoch": 1.28, + "learning_rate": 1.574701045535542e-05, + "loss": 0.994, + "step": 19920 + }, + { + "epoch": 1.28, + "learning_rate": 1.574679576633247e-05, + "loss": 0.9636, + "step": 19921 + }, + { + "epoch": 1.28, + "learning_rate": 1.5746581077309518e-05, + "loss": 0.9258, + "step": 19922 + }, + { + "epoch": 1.28, + "learning_rate": 1.574636638828657e-05, + "loss": 1.0238, + "step": 19923 + }, + { + "epoch": 1.28, + "learning_rate": 1.574615169926362e-05, + "loss": 1.0018, + "step": 19924 + }, + { + "epoch": 1.28, + "learning_rate": 1.5745937010240668e-05, + "loss": 0.9804, + "step": 19925 + }, + { + "epoch": 1.28, + "learning_rate": 1.5745722321217717e-05, + "loss": 1.0434, + "step": 19926 + }, + { + "epoch": 1.28, + "learning_rate": 1.5745507632194766e-05, + "loss": 0.9862, + "step": 19927 + }, + { + "epoch": 1.28, + "learning_rate": 1.5745292943171818e-05, + "loss": 0.9855, + "step": 19928 + }, + { + "epoch": 1.28, + "learning_rate": 1.5745078254148867e-05, + "loss": 0.9803, + "step": 19929 + }, + { + "epoch": 1.28, + "learning_rate": 1.5744863565125916e-05, + "loss": 0.9435, + "step": 19930 + }, + { + "epoch": 1.28, + "learning_rate": 1.5744648876102965e-05, + "loss": 1.0306, + "step": 19931 + }, + { + "epoch": 1.28, + "learning_rate": 1.5744434187080014e-05, + "loss": 0.8992, + "step": 19932 + }, + { + "epoch": 1.28, + "learning_rate": 1.5744219498057066e-05, + "loss": 0.9641, + "step": 19933 + }, + { + "epoch": 1.28, + "learning_rate": 1.5744004809034115e-05, + "loss": 0.9927, + "step": 19934 + }, + { + "epoch": 1.28, + "learning_rate": 1.5743790120011164e-05, + "loss": 0.9665, + "step": 19935 + }, + { + "epoch": 1.28, + "learning_rate": 1.5743575430988213e-05, + "loss": 0.9631, + "step": 19936 + }, + { + "epoch": 1.28, + "learning_rate": 1.5743360741965266e-05, + "loss": 0.9778, + "step": 19937 + }, + { + "epoch": 1.28, + "learning_rate": 1.5743146052942315e-05, + "loss": 1.0018, + "step": 19938 + }, + { + "epoch": 1.28, + "learning_rate": 1.5742931363919364e-05, + "loss": 0.9186, + "step": 19939 + }, + { + "epoch": 1.28, + "learning_rate": 1.5742716674896416e-05, + "loss": 0.9718, + "step": 19940 + }, + { + "epoch": 1.28, + "learning_rate": 1.5742501985873465e-05, + "loss": 0.8603, + "step": 19941 + }, + { + "epoch": 1.28, + "learning_rate": 1.5742287296850514e-05, + "loss": 1.0474, + "step": 19942 + }, + { + "epoch": 1.28, + "learning_rate": 1.5742072607827563e-05, + "loss": 0.9447, + "step": 19943 + }, + { + "epoch": 1.28, + "learning_rate": 1.5741857918804612e-05, + "loss": 0.9464, + "step": 19944 + }, + { + "epoch": 1.28, + "learning_rate": 1.574164322978166e-05, + "loss": 1.0234, + "step": 19945 + }, + { + "epoch": 1.28, + "learning_rate": 1.5741428540758713e-05, + "loss": 0.9976, + "step": 19946 + }, + { + "epoch": 1.28, + "learning_rate": 1.5741213851735762e-05, + "loss": 1.0152, + "step": 19947 + }, + { + "epoch": 1.28, + "learning_rate": 1.574099916271281e-05, + "loss": 0.992, + "step": 19948 + }, + { + "epoch": 1.28, + "learning_rate": 1.574078447368986e-05, + "loss": 0.9796, + "step": 19949 + }, + { + "epoch": 1.28, + "learning_rate": 1.574056978466691e-05, + "loss": 0.9755, + "step": 19950 + }, + { + "epoch": 1.28, + "learning_rate": 1.574035509564396e-05, + "loss": 1.0334, + "step": 19951 + }, + { + "epoch": 1.28, + "learning_rate": 1.574014040662101e-05, + "loss": 0.9424, + "step": 19952 + }, + { + "epoch": 1.28, + "learning_rate": 1.5739925717598063e-05, + "loss": 0.977, + "step": 19953 + }, + { + "epoch": 1.28, + "learning_rate": 1.573971102857511e-05, + "loss": 0.984, + "step": 19954 + }, + { + "epoch": 1.28, + "learning_rate": 1.573949633955216e-05, + "loss": 0.9613, + "step": 19955 + }, + { + "epoch": 1.28, + "learning_rate": 1.573928165052921e-05, + "loss": 0.911, + "step": 19956 + }, + { + "epoch": 1.28, + "learning_rate": 1.573906696150626e-05, + "loss": 0.912, + "step": 19957 + }, + { + "epoch": 1.28, + "learning_rate": 1.573885227248331e-05, + "loss": 0.9451, + "step": 19958 + }, + { + "epoch": 1.28, + "learning_rate": 1.573863758346036e-05, + "loss": 1.0363, + "step": 19959 + }, + { + "epoch": 1.28, + "learning_rate": 1.573842289443741e-05, + "loss": 0.9599, + "step": 19960 + }, + { + "epoch": 1.28, + "learning_rate": 1.5738208205414458e-05, + "loss": 1.0304, + "step": 19961 + }, + { + "epoch": 1.28, + "learning_rate": 1.5737993516391507e-05, + "loss": 1.04, + "step": 19962 + }, + { + "epoch": 1.28, + "learning_rate": 1.5737778827368556e-05, + "loss": 0.9899, + "step": 19963 + }, + { + "epoch": 1.28, + "learning_rate": 1.5737564138345608e-05, + "loss": 0.9278, + "step": 19964 + }, + { + "epoch": 1.28, + "learning_rate": 1.5737349449322657e-05, + "loss": 1.037, + "step": 19965 + }, + { + "epoch": 1.28, + "learning_rate": 1.573713476029971e-05, + "loss": 1.0873, + "step": 19966 + }, + { + "epoch": 1.28, + "learning_rate": 1.573692007127676e-05, + "loss": 1.0094, + "step": 19967 + }, + { + "epoch": 1.28, + "learning_rate": 1.5736705382253807e-05, + "loss": 0.9368, + "step": 19968 + }, + { + "epoch": 1.28, + "learning_rate": 1.5736490693230856e-05, + "loss": 0.9708, + "step": 19969 + }, + { + "epoch": 1.28, + "learning_rate": 1.5736276004207905e-05, + "loss": 0.9433, + "step": 19970 + }, + { + "epoch": 1.28, + "learning_rate": 1.5736061315184958e-05, + "loss": 0.9849, + "step": 19971 + }, + { + "epoch": 1.28, + "learning_rate": 1.5735846626162007e-05, + "loss": 0.9636, + "step": 19972 + }, + { + "epoch": 1.29, + "learning_rate": 1.5735631937139056e-05, + "loss": 0.9855, + "step": 19973 + }, + { + "epoch": 1.29, + "learning_rate": 1.5735417248116105e-05, + "loss": 0.9683, + "step": 19974 + }, + { + "epoch": 1.29, + "learning_rate": 1.5735202559093154e-05, + "loss": 0.9474, + "step": 19975 + }, + { + "epoch": 1.29, + "learning_rate": 1.5734987870070202e-05, + "loss": 0.9928, + "step": 19976 + }, + { + "epoch": 1.29, + "learning_rate": 1.5734773181047255e-05, + "loss": 1.0543, + "step": 19977 + }, + { + "epoch": 1.29, + "learning_rate": 1.5734558492024304e-05, + "loss": 0.931, + "step": 19978 + }, + { + "epoch": 1.29, + "learning_rate": 1.5734343803001353e-05, + "loss": 0.9409, + "step": 19979 + }, + { + "epoch": 1.29, + "learning_rate": 1.5734129113978402e-05, + "loss": 0.9655, + "step": 19980 + }, + { + "epoch": 1.29, + "learning_rate": 1.5733914424955454e-05, + "loss": 0.8792, + "step": 19981 + }, + { + "epoch": 1.29, + "learning_rate": 1.5733699735932503e-05, + "loss": 0.9522, + "step": 19982 + }, + { + "epoch": 1.29, + "learning_rate": 1.5733485046909552e-05, + "loss": 0.9059, + "step": 19983 + }, + { + "epoch": 1.29, + "learning_rate": 1.5733270357886604e-05, + "loss": 0.9968, + "step": 19984 + }, + { + "epoch": 1.29, + "learning_rate": 1.5733055668863653e-05, + "loss": 0.9947, + "step": 19985 + }, + { + "epoch": 1.29, + "learning_rate": 1.5732840979840702e-05, + "loss": 1.0216, + "step": 19986 + }, + { + "epoch": 1.29, + "learning_rate": 1.573262629081775e-05, + "loss": 0.9801, + "step": 19987 + }, + { + "epoch": 1.29, + "learning_rate": 1.57324116017948e-05, + "loss": 0.966, + "step": 19988 + }, + { + "epoch": 1.29, + "learning_rate": 1.5732196912771853e-05, + "loss": 0.9276, + "step": 19989 + }, + { + "epoch": 1.29, + "learning_rate": 1.57319822237489e-05, + "loss": 0.9976, + "step": 19990 + }, + { + "epoch": 1.29, + "learning_rate": 1.573176753472595e-05, + "loss": 0.9968, + "step": 19991 + }, + { + "epoch": 1.29, + "learning_rate": 1.5731552845703e-05, + "loss": 0.9863, + "step": 19992 + }, + { + "epoch": 1.29, + "learning_rate": 1.573133815668005e-05, + "loss": 0.8841, + "step": 19993 + }, + { + "epoch": 1.29, + "learning_rate": 1.5731123467657097e-05, + "loss": 0.9574, + "step": 19994 + }, + { + "epoch": 1.29, + "learning_rate": 1.573090877863415e-05, + "loss": 0.9246, + "step": 19995 + }, + { + "epoch": 1.29, + "learning_rate": 1.57306940896112e-05, + "loss": 1.0158, + "step": 19996 + }, + { + "epoch": 1.29, + "learning_rate": 1.573047940058825e-05, + "loss": 1.011, + "step": 19997 + }, + { + "epoch": 1.29, + "learning_rate": 1.57302647115653e-05, + "loss": 0.9881, + "step": 19998 + }, + { + "epoch": 1.29, + "learning_rate": 1.573005002254235e-05, + "loss": 0.9647, + "step": 19999 + }, + { + "epoch": 1.29, + "learning_rate": 1.5729835333519398e-05, + "loss": 0.8863, + "step": 20000 + }, + { + "epoch": 1.29, + "eval_laion/cider": 0.535763540768209, + "eval_laion/meteor": 0.3814674784824261, + "eval_laion/pred_num_words": 35.955078125, + "eval_laion/sacrebleu": 14.767416764700334, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0554171800613403, + "eval_runtime": 100.2867, + "eval_samples_per_second": 4.976, + "eval_steps_per_second": 0.319, + "step": 20000 + }, + { + "epoch": 1.29, + "learning_rate": 1.5729620644496447e-05, + "loss": 1.0113, + "step": 20001 + }, + { + "epoch": 1.29, + "learning_rate": 1.57294059554735e-05, + "loss": 1.006, + "step": 20002 + }, + { + "epoch": 1.29, + "learning_rate": 1.5729191266450548e-05, + "loss": 1.039, + "step": 20003 + }, + { + "epoch": 1.29, + "learning_rate": 1.5728976577427597e-05, + "loss": 0.9654, + "step": 20004 + }, + { + "epoch": 1.29, + "learning_rate": 1.5728761888404646e-05, + "loss": 0.9375, + "step": 20005 + }, + { + "epoch": 1.29, + "learning_rate": 1.5728547199381695e-05, + "loss": 1.0116, + "step": 20006 + }, + { + "epoch": 1.29, + "learning_rate": 1.5728332510358747e-05, + "loss": 0.9098, + "step": 20007 + }, + { + "epoch": 1.29, + "learning_rate": 1.5728117821335796e-05, + "loss": 0.9285, + "step": 20008 + }, + { + "epoch": 1.29, + "learning_rate": 1.5727903132312845e-05, + "loss": 0.9616, + "step": 20009 + }, + { + "epoch": 1.29, + "learning_rate": 1.5727688443289898e-05, + "loss": 1.0172, + "step": 20010 + }, + { + "epoch": 1.29, + "learning_rate": 1.5727473754266947e-05, + "loss": 0.9761, + "step": 20011 + }, + { + "epoch": 1.29, + "learning_rate": 1.5727259065243996e-05, + "loss": 0.9796, + "step": 20012 + }, + { + "epoch": 1.29, + "learning_rate": 1.5727044376221045e-05, + "loss": 0.8845, + "step": 20013 + }, + { + "epoch": 1.29, + "learning_rate": 1.5726829687198094e-05, + "loss": 0.9134, + "step": 20014 + }, + { + "epoch": 1.29, + "learning_rate": 1.5726614998175146e-05, + "loss": 1.0034, + "step": 20015 + }, + { + "epoch": 1.29, + "learning_rate": 1.5726400309152195e-05, + "loss": 0.9597, + "step": 20016 + }, + { + "epoch": 1.29, + "learning_rate": 1.5726185620129244e-05, + "loss": 0.9804, + "step": 20017 + }, + { + "epoch": 1.29, + "learning_rate": 1.5725970931106293e-05, + "loss": 1.0451, + "step": 20018 + }, + { + "epoch": 1.29, + "learning_rate": 1.5725756242083342e-05, + "loss": 1.067, + "step": 20019 + }, + { + "epoch": 1.29, + "learning_rate": 1.5725541553060394e-05, + "loss": 0.9902, + "step": 20020 + }, + { + "epoch": 1.29, + "learning_rate": 1.5725326864037443e-05, + "loss": 0.9992, + "step": 20021 + }, + { + "epoch": 1.29, + "learning_rate": 1.5725112175014492e-05, + "loss": 0.9978, + "step": 20022 + }, + { + "epoch": 1.29, + "learning_rate": 1.572489748599154e-05, + "loss": 1.0173, + "step": 20023 + }, + { + "epoch": 1.29, + "learning_rate": 1.5724682796968593e-05, + "loss": 0.9007, + "step": 20024 + }, + { + "epoch": 1.29, + "learning_rate": 1.5724468107945642e-05, + "loss": 0.9855, + "step": 20025 + }, + { + "epoch": 1.29, + "learning_rate": 1.572425341892269e-05, + "loss": 0.9803, + "step": 20026 + }, + { + "epoch": 1.29, + "learning_rate": 1.5724038729899744e-05, + "loss": 0.9599, + "step": 20027 + }, + { + "epoch": 1.29, + "learning_rate": 1.5723824040876793e-05, + "loss": 0.8998, + "step": 20028 + }, + { + "epoch": 1.29, + "learning_rate": 1.572360935185384e-05, + "loss": 0.9607, + "step": 20029 + }, + { + "epoch": 1.29, + "learning_rate": 1.572339466283089e-05, + "loss": 0.9341, + "step": 20030 + }, + { + "epoch": 1.29, + "learning_rate": 1.572317997380794e-05, + "loss": 0.9835, + "step": 20031 + }, + { + "epoch": 1.29, + "learning_rate": 1.572296528478499e-05, + "loss": 0.9694, + "step": 20032 + }, + { + "epoch": 1.29, + "learning_rate": 1.572275059576204e-05, + "loss": 0.8677, + "step": 20033 + }, + { + "epoch": 1.29, + "learning_rate": 1.572253590673909e-05, + "loss": 0.9677, + "step": 20034 + }, + { + "epoch": 1.29, + "learning_rate": 1.572232121771614e-05, + "loss": 1.04, + "step": 20035 + }, + { + "epoch": 1.29, + "learning_rate": 1.5722106528693188e-05, + "loss": 0.9478, + "step": 20036 + }, + { + "epoch": 1.29, + "learning_rate": 1.5721891839670237e-05, + "loss": 0.9409, + "step": 20037 + }, + { + "epoch": 1.29, + "learning_rate": 1.572167715064729e-05, + "loss": 0.9081, + "step": 20038 + }, + { + "epoch": 1.29, + "learning_rate": 1.5721462461624338e-05, + "loss": 0.9554, + "step": 20039 + }, + { + "epoch": 1.29, + "learning_rate": 1.572124777260139e-05, + "loss": 1.0232, + "step": 20040 + }, + { + "epoch": 1.29, + "learning_rate": 1.572103308357844e-05, + "loss": 0.9499, + "step": 20041 + }, + { + "epoch": 1.29, + "learning_rate": 1.572081839455549e-05, + "loss": 0.9387, + "step": 20042 + }, + { + "epoch": 1.29, + "learning_rate": 1.5720603705532537e-05, + "loss": 0.9614, + "step": 20043 + }, + { + "epoch": 1.29, + "learning_rate": 1.5720389016509586e-05, + "loss": 0.8585, + "step": 20044 + }, + { + "epoch": 1.29, + "learning_rate": 1.572017432748664e-05, + "loss": 1.1118, + "step": 20045 + }, + { + "epoch": 1.29, + "learning_rate": 1.5719959638463688e-05, + "loss": 1.02, + "step": 20046 + }, + { + "epoch": 1.29, + "learning_rate": 1.5719744949440737e-05, + "loss": 0.9544, + "step": 20047 + }, + { + "epoch": 1.29, + "learning_rate": 1.5719530260417786e-05, + "loss": 0.9122, + "step": 20048 + }, + { + "epoch": 1.29, + "learning_rate": 1.5719315571394835e-05, + "loss": 1.0268, + "step": 20049 + }, + { + "epoch": 1.29, + "learning_rate": 1.5719100882371883e-05, + "loss": 0.9553, + "step": 20050 + }, + { + "epoch": 1.29, + "learning_rate": 1.5718886193348936e-05, + "loss": 0.9553, + "step": 20051 + }, + { + "epoch": 1.29, + "learning_rate": 1.5718671504325985e-05, + "loss": 1.0371, + "step": 20052 + }, + { + "epoch": 1.29, + "learning_rate": 1.5718456815303037e-05, + "loss": 1.0477, + "step": 20053 + }, + { + "epoch": 1.29, + "learning_rate": 1.5718242126280086e-05, + "loss": 1.0218, + "step": 20054 + }, + { + "epoch": 1.29, + "learning_rate": 1.5718027437257135e-05, + "loss": 1.0072, + "step": 20055 + }, + { + "epoch": 1.29, + "learning_rate": 1.5717812748234184e-05, + "loss": 1.0355, + "step": 20056 + }, + { + "epoch": 1.29, + "learning_rate": 1.5717598059211233e-05, + "loss": 0.9444, + "step": 20057 + }, + { + "epoch": 1.29, + "learning_rate": 1.5717383370188285e-05, + "loss": 0.9065, + "step": 20058 + }, + { + "epoch": 1.29, + "learning_rate": 1.5717168681165334e-05, + "loss": 1.0268, + "step": 20059 + }, + { + "epoch": 1.29, + "learning_rate": 1.5716953992142383e-05, + "loss": 0.9144, + "step": 20060 + }, + { + "epoch": 1.29, + "learning_rate": 1.5716739303119432e-05, + "loss": 0.985, + "step": 20061 + }, + { + "epoch": 1.29, + "learning_rate": 1.571652461409648e-05, + "loss": 1.0277, + "step": 20062 + }, + { + "epoch": 1.29, + "learning_rate": 1.571630992507353e-05, + "loss": 0.9441, + "step": 20063 + }, + { + "epoch": 1.29, + "learning_rate": 1.5716095236050583e-05, + "loss": 0.9462, + "step": 20064 + }, + { + "epoch": 1.29, + "learning_rate": 1.571588054702763e-05, + "loss": 0.9643, + "step": 20065 + }, + { + "epoch": 1.29, + "learning_rate": 1.571566585800468e-05, + "loss": 0.9699, + "step": 20066 + }, + { + "epoch": 1.29, + "learning_rate": 1.571545116898173e-05, + "loss": 0.9783, + "step": 20067 + }, + { + "epoch": 1.29, + "learning_rate": 1.5715236479958782e-05, + "loss": 1.0213, + "step": 20068 + }, + { + "epoch": 1.29, + "learning_rate": 1.571502179093583e-05, + "loss": 0.9699, + "step": 20069 + }, + { + "epoch": 1.29, + "learning_rate": 1.571480710191288e-05, + "loss": 0.9406, + "step": 20070 + }, + { + "epoch": 1.29, + "learning_rate": 1.5714592412889932e-05, + "loss": 1.0596, + "step": 20071 + }, + { + "epoch": 1.29, + "learning_rate": 1.571437772386698e-05, + "loss": 1.0128, + "step": 20072 + }, + { + "epoch": 1.29, + "learning_rate": 1.571416303484403e-05, + "loss": 1.0323, + "step": 20073 + }, + { + "epoch": 1.29, + "learning_rate": 1.571394834582108e-05, + "loss": 0.9902, + "step": 20074 + }, + { + "epoch": 1.29, + "learning_rate": 1.5713733656798128e-05, + "loss": 0.8918, + "step": 20075 + }, + { + "epoch": 1.29, + "learning_rate": 1.571351896777518e-05, + "loss": 1.0391, + "step": 20076 + }, + { + "epoch": 1.29, + "learning_rate": 1.571330427875223e-05, + "loss": 0.9581, + "step": 20077 + }, + { + "epoch": 1.29, + "learning_rate": 1.5713089589729278e-05, + "loss": 0.8727, + "step": 20078 + }, + { + "epoch": 1.29, + "learning_rate": 1.5712874900706327e-05, + "loss": 0.9478, + "step": 20079 + }, + { + "epoch": 1.29, + "learning_rate": 1.5712660211683376e-05, + "loss": 1.011, + "step": 20080 + }, + { + "epoch": 1.29, + "learning_rate": 1.5712445522660425e-05, + "loss": 1.0005, + "step": 20081 + }, + { + "epoch": 1.29, + "learning_rate": 1.5712230833637477e-05, + "loss": 0.9619, + "step": 20082 + }, + { + "epoch": 1.29, + "learning_rate": 1.5712016144614526e-05, + "loss": 0.9732, + "step": 20083 + }, + { + "epoch": 1.29, + "learning_rate": 1.571180145559158e-05, + "loss": 0.9726, + "step": 20084 + }, + { + "epoch": 1.29, + "learning_rate": 1.5711586766568628e-05, + "loss": 0.9683, + "step": 20085 + }, + { + "epoch": 1.29, + "learning_rate": 1.5711372077545677e-05, + "loss": 0.9762, + "step": 20086 + }, + { + "epoch": 1.29, + "learning_rate": 1.5711157388522726e-05, + "loss": 0.8941, + "step": 20087 + }, + { + "epoch": 1.29, + "learning_rate": 1.5710942699499775e-05, + "loss": 1.0244, + "step": 20088 + }, + { + "epoch": 1.29, + "learning_rate": 1.5710728010476827e-05, + "loss": 1.0155, + "step": 20089 + }, + { + "epoch": 1.29, + "learning_rate": 1.5710513321453876e-05, + "loss": 0.9746, + "step": 20090 + }, + { + "epoch": 1.29, + "learning_rate": 1.5710298632430925e-05, + "loss": 0.8995, + "step": 20091 + }, + { + "epoch": 1.29, + "learning_rate": 1.5710083943407974e-05, + "loss": 0.9994, + "step": 20092 + }, + { + "epoch": 1.29, + "learning_rate": 1.5709869254385023e-05, + "loss": 0.9145, + "step": 20093 + }, + { + "epoch": 1.29, + "learning_rate": 1.5709654565362075e-05, + "loss": 0.9857, + "step": 20094 + }, + { + "epoch": 1.29, + "learning_rate": 1.5709439876339124e-05, + "loss": 0.953, + "step": 20095 + }, + { + "epoch": 1.29, + "learning_rate": 1.5709225187316173e-05, + "loss": 0.9997, + "step": 20096 + }, + { + "epoch": 1.29, + "learning_rate": 1.5709010498293225e-05, + "loss": 0.8988, + "step": 20097 + }, + { + "epoch": 1.29, + "learning_rate": 1.5708795809270274e-05, + "loss": 1.0258, + "step": 20098 + }, + { + "epoch": 1.29, + "learning_rate": 1.5708581120247323e-05, + "loss": 1.0762, + "step": 20099 + }, + { + "epoch": 1.29, + "learning_rate": 1.5708366431224372e-05, + "loss": 1.0411, + "step": 20100 + }, + { + "epoch": 1.29, + "learning_rate": 1.570815174220142e-05, + "loss": 0.9515, + "step": 20101 + }, + { + "epoch": 1.29, + "learning_rate": 1.5707937053178474e-05, + "loss": 0.8542, + "step": 20102 + }, + { + "epoch": 1.29, + "learning_rate": 1.5707722364155523e-05, + "loss": 0.9424, + "step": 20103 + }, + { + "epoch": 1.29, + "learning_rate": 1.570750767513257e-05, + "loss": 1.0652, + "step": 20104 + }, + { + "epoch": 1.29, + "learning_rate": 1.570729298610962e-05, + "loss": 0.9955, + "step": 20105 + }, + { + "epoch": 1.29, + "learning_rate": 1.570707829708667e-05, + "loss": 1.0242, + "step": 20106 + }, + { + "epoch": 1.29, + "learning_rate": 1.5706863608063722e-05, + "loss": 0.91, + "step": 20107 + }, + { + "epoch": 1.29, + "learning_rate": 1.570664891904077e-05, + "loss": 1.0085, + "step": 20108 + }, + { + "epoch": 1.29, + "learning_rate": 1.570643423001782e-05, + "loss": 1.0216, + "step": 20109 + }, + { + "epoch": 1.29, + "learning_rate": 1.570621954099487e-05, + "loss": 0.9712, + "step": 20110 + }, + { + "epoch": 1.29, + "learning_rate": 1.570600485197192e-05, + "loss": 0.9995, + "step": 20111 + }, + { + "epoch": 1.29, + "learning_rate": 1.570579016294897e-05, + "loss": 1.0192, + "step": 20112 + }, + { + "epoch": 1.29, + "learning_rate": 1.570557547392602e-05, + "loss": 0.9816, + "step": 20113 + }, + { + "epoch": 1.29, + "learning_rate": 1.570536078490307e-05, + "loss": 0.9799, + "step": 20114 + }, + { + "epoch": 1.29, + "learning_rate": 1.570514609588012e-05, + "loss": 0.9316, + "step": 20115 + }, + { + "epoch": 1.29, + "learning_rate": 1.570493140685717e-05, + "loss": 1.0293, + "step": 20116 + }, + { + "epoch": 1.29, + "learning_rate": 1.570471671783422e-05, + "loss": 0.9259, + "step": 20117 + }, + { + "epoch": 1.29, + "learning_rate": 1.5704502028811267e-05, + "loss": 1.0043, + "step": 20118 + }, + { + "epoch": 1.29, + "learning_rate": 1.5704287339788316e-05, + "loss": 0.9542, + "step": 20119 + }, + { + "epoch": 1.29, + "learning_rate": 1.570407265076537e-05, + "loss": 0.9814, + "step": 20120 + }, + { + "epoch": 1.29, + "learning_rate": 1.5703857961742418e-05, + "loss": 0.9435, + "step": 20121 + }, + { + "epoch": 1.29, + "learning_rate": 1.5703643272719467e-05, + "loss": 1.0145, + "step": 20122 + }, + { + "epoch": 1.29, + "learning_rate": 1.5703428583696516e-05, + "loss": 0.9979, + "step": 20123 + }, + { + "epoch": 1.29, + "learning_rate": 1.5703213894673564e-05, + "loss": 0.9256, + "step": 20124 + }, + { + "epoch": 1.29, + "learning_rate": 1.5702999205650617e-05, + "loss": 1.0317, + "step": 20125 + }, + { + "epoch": 1.29, + "learning_rate": 1.5702784516627666e-05, + "loss": 0.951, + "step": 20126 + }, + { + "epoch": 1.29, + "learning_rate": 1.5702569827604718e-05, + "loss": 0.946, + "step": 20127 + }, + { + "epoch": 1.29, + "learning_rate": 1.5702355138581767e-05, + "loss": 1.0089, + "step": 20128 + }, + { + "epoch": 1.3, + "learning_rate": 1.5702140449558816e-05, + "loss": 0.976, + "step": 20129 + }, + { + "epoch": 1.3, + "learning_rate": 1.5701925760535865e-05, + "loss": 0.9767, + "step": 20130 + }, + { + "epoch": 1.3, + "learning_rate": 1.5701711071512914e-05, + "loss": 0.9575, + "step": 20131 + }, + { + "epoch": 1.3, + "learning_rate": 1.5701496382489963e-05, + "loss": 0.9908, + "step": 20132 + }, + { + "epoch": 1.3, + "learning_rate": 1.5701281693467015e-05, + "loss": 0.9865, + "step": 20133 + }, + { + "epoch": 1.3, + "learning_rate": 1.5701067004444064e-05, + "loss": 1.0434, + "step": 20134 + }, + { + "epoch": 1.3, + "learning_rate": 1.5700852315421113e-05, + "loss": 0.9245, + "step": 20135 + }, + { + "epoch": 1.3, + "learning_rate": 1.5700637626398162e-05, + "loss": 1.0177, + "step": 20136 + }, + { + "epoch": 1.3, + "learning_rate": 1.570042293737521e-05, + "loss": 1.0278, + "step": 20137 + }, + { + "epoch": 1.3, + "learning_rate": 1.5700208248352264e-05, + "loss": 0.9793, + "step": 20138 + }, + { + "epoch": 1.3, + "learning_rate": 1.5699993559329312e-05, + "loss": 0.9403, + "step": 20139 + }, + { + "epoch": 1.3, + "learning_rate": 1.5699778870306365e-05, + "loss": 1.0535, + "step": 20140 + }, + { + "epoch": 1.3, + "learning_rate": 1.5699564181283414e-05, + "loss": 0.9953, + "step": 20141 + }, + { + "epoch": 1.3, + "learning_rate": 1.5699349492260463e-05, + "loss": 0.9339, + "step": 20142 + }, + { + "epoch": 1.3, + "learning_rate": 1.5699134803237512e-05, + "loss": 0.9453, + "step": 20143 + }, + { + "epoch": 1.3, + "learning_rate": 1.569892011421456e-05, + "loss": 0.9822, + "step": 20144 + }, + { + "epoch": 1.3, + "learning_rate": 1.5698705425191613e-05, + "loss": 1.0053, + "step": 20145 + }, + { + "epoch": 1.3, + "learning_rate": 1.5698490736168662e-05, + "loss": 0.9712, + "step": 20146 + }, + { + "epoch": 1.3, + "learning_rate": 1.569827604714571e-05, + "loss": 0.9381, + "step": 20147 + }, + { + "epoch": 1.3, + "learning_rate": 1.569806135812276e-05, + "loss": 0.9916, + "step": 20148 + }, + { + "epoch": 1.3, + "learning_rate": 1.569784666909981e-05, + "loss": 0.9684, + "step": 20149 + }, + { + "epoch": 1.3, + "learning_rate": 1.5697631980076858e-05, + "loss": 1.0159, + "step": 20150 + }, + { + "epoch": 1.3, + "learning_rate": 1.569741729105391e-05, + "loss": 0.9695, + "step": 20151 + }, + { + "epoch": 1.3, + "learning_rate": 1.569720260203096e-05, + "loss": 0.8893, + "step": 20152 + }, + { + "epoch": 1.3, + "learning_rate": 1.5696987913008008e-05, + "loss": 0.9848, + "step": 20153 + }, + { + "epoch": 1.3, + "learning_rate": 1.5696773223985057e-05, + "loss": 0.9821, + "step": 20154 + }, + { + "epoch": 1.3, + "learning_rate": 1.569655853496211e-05, + "loss": 0.8924, + "step": 20155 + }, + { + "epoch": 1.3, + "learning_rate": 1.569634384593916e-05, + "loss": 0.9758, + "step": 20156 + }, + { + "epoch": 1.3, + "learning_rate": 1.5696129156916207e-05, + "loss": 1.0422, + "step": 20157 + }, + { + "epoch": 1.3, + "learning_rate": 1.569591446789326e-05, + "loss": 0.9815, + "step": 20158 + }, + { + "epoch": 1.3, + "learning_rate": 1.569569977887031e-05, + "loss": 0.9515, + "step": 20159 + }, + { + "epoch": 1.3, + "learning_rate": 1.5695485089847358e-05, + "loss": 0.9224, + "step": 20160 + }, + { + "epoch": 1.3, + "learning_rate": 1.5695270400824407e-05, + "loss": 1.0328, + "step": 20161 + }, + { + "epoch": 1.3, + "learning_rate": 1.5695055711801456e-05, + "loss": 0.9583, + "step": 20162 + }, + { + "epoch": 1.3, + "learning_rate": 1.5694841022778508e-05, + "loss": 1.0861, + "step": 20163 + }, + { + "epoch": 1.3, + "learning_rate": 1.5694626333755557e-05, + "loss": 1.051, + "step": 20164 + }, + { + "epoch": 1.3, + "learning_rate": 1.5694411644732606e-05, + "loss": 0.9785, + "step": 20165 + }, + { + "epoch": 1.3, + "learning_rate": 1.5694196955709655e-05, + "loss": 1.0195, + "step": 20166 + }, + { + "epoch": 1.3, + "learning_rate": 1.5693982266686704e-05, + "loss": 0.9672, + "step": 20167 + }, + { + "epoch": 1.3, + "learning_rate": 1.5693767577663753e-05, + "loss": 0.9258, + "step": 20168 + }, + { + "epoch": 1.3, + "learning_rate": 1.5693552888640805e-05, + "loss": 1.0322, + "step": 20169 + }, + { + "epoch": 1.3, + "learning_rate": 1.5693338199617854e-05, + "loss": 0.9769, + "step": 20170 + }, + { + "epoch": 1.3, + "learning_rate": 1.5693123510594906e-05, + "loss": 0.9537, + "step": 20171 + }, + { + "epoch": 1.3, + "learning_rate": 1.5692908821571955e-05, + "loss": 0.9991, + "step": 20172 + }, + { + "epoch": 1.3, + "learning_rate": 1.5692694132549004e-05, + "loss": 0.9966, + "step": 20173 + }, + { + "epoch": 1.3, + "learning_rate": 1.5692479443526053e-05, + "loss": 1.0134, + "step": 20174 + }, + { + "epoch": 1.3, + "learning_rate": 1.5692264754503102e-05, + "loss": 0.9893, + "step": 20175 + }, + { + "epoch": 1.3, + "learning_rate": 1.5692050065480155e-05, + "loss": 0.9437, + "step": 20176 + }, + { + "epoch": 1.3, + "learning_rate": 1.5691835376457204e-05, + "loss": 0.9546, + "step": 20177 + }, + { + "epoch": 1.3, + "learning_rate": 1.5691620687434253e-05, + "loss": 0.9103, + "step": 20178 + }, + { + "epoch": 1.3, + "learning_rate": 1.56914059984113e-05, + "loss": 0.97, + "step": 20179 + }, + { + "epoch": 1.3, + "learning_rate": 1.569119130938835e-05, + "loss": 0.9328, + "step": 20180 + }, + { + "epoch": 1.3, + "learning_rate": 1.5690976620365403e-05, + "loss": 0.9365, + "step": 20181 + }, + { + "epoch": 1.3, + "learning_rate": 1.5690761931342452e-05, + "loss": 1.0391, + "step": 20182 + }, + { + "epoch": 1.3, + "learning_rate": 1.56905472423195e-05, + "loss": 0.9536, + "step": 20183 + }, + { + "epoch": 1.3, + "learning_rate": 1.5690332553296553e-05, + "loss": 0.9557, + "step": 20184 + }, + { + "epoch": 1.3, + "learning_rate": 1.5690117864273602e-05, + "loss": 0.9936, + "step": 20185 + }, + { + "epoch": 1.3, + "learning_rate": 1.568990317525065e-05, + "loss": 0.9144, + "step": 20186 + }, + { + "epoch": 1.3, + "learning_rate": 1.56896884862277e-05, + "loss": 0.9583, + "step": 20187 + }, + { + "epoch": 1.3, + "learning_rate": 1.568947379720475e-05, + "loss": 1.052, + "step": 20188 + }, + { + "epoch": 1.3, + "learning_rate": 1.56892591081818e-05, + "loss": 0.9182, + "step": 20189 + }, + { + "epoch": 1.3, + "learning_rate": 1.568904441915885e-05, + "loss": 0.9504, + "step": 20190 + }, + { + "epoch": 1.3, + "learning_rate": 1.56888297301359e-05, + "loss": 1.0441, + "step": 20191 + }, + { + "epoch": 1.3, + "learning_rate": 1.5688615041112948e-05, + "loss": 0.9585, + "step": 20192 + }, + { + "epoch": 1.3, + "learning_rate": 1.5688400352089997e-05, + "loss": 1.0055, + "step": 20193 + }, + { + "epoch": 1.3, + "learning_rate": 1.568818566306705e-05, + "loss": 0.9816, + "step": 20194 + }, + { + "epoch": 1.3, + "learning_rate": 1.56879709740441e-05, + "loss": 0.948, + "step": 20195 + }, + { + "epoch": 1.3, + "learning_rate": 1.5687756285021148e-05, + "loss": 0.9959, + "step": 20196 + }, + { + "epoch": 1.3, + "learning_rate": 1.5687541595998197e-05, + "loss": 0.9921, + "step": 20197 + }, + { + "epoch": 1.3, + "learning_rate": 1.568732690697525e-05, + "loss": 0.9372, + "step": 20198 + }, + { + "epoch": 1.3, + "learning_rate": 1.5687112217952298e-05, + "loss": 1.0595, + "step": 20199 + }, + { + "epoch": 1.3, + "learning_rate": 1.5686897528929347e-05, + "loss": 0.9845, + "step": 20200 + }, + { + "epoch": 1.3, + "learning_rate": 1.56866828399064e-05, + "loss": 1.0312, + "step": 20201 + }, + { + "epoch": 1.3, + "learning_rate": 1.5686468150883448e-05, + "loss": 0.9853, + "step": 20202 + }, + { + "epoch": 1.3, + "learning_rate": 1.5686253461860497e-05, + "loss": 0.972, + "step": 20203 + }, + { + "epoch": 1.3, + "learning_rate": 1.5686038772837546e-05, + "loss": 1.0013, + "step": 20204 + }, + { + "epoch": 1.3, + "learning_rate": 1.5685824083814595e-05, + "loss": 0.9364, + "step": 20205 + }, + { + "epoch": 1.3, + "learning_rate": 1.5685609394791644e-05, + "loss": 1.0092, + "step": 20206 + }, + { + "epoch": 1.3, + "learning_rate": 1.5685394705768696e-05, + "loss": 0.9064, + "step": 20207 + }, + { + "epoch": 1.3, + "learning_rate": 1.5685180016745745e-05, + "loss": 1.0349, + "step": 20208 + }, + { + "epoch": 1.3, + "learning_rate": 1.5684965327722794e-05, + "loss": 0.9994, + "step": 20209 + }, + { + "epoch": 1.3, + "learning_rate": 1.5684750638699843e-05, + "loss": 1.0072, + "step": 20210 + }, + { + "epoch": 1.3, + "learning_rate": 1.5684535949676892e-05, + "loss": 0.9191, + "step": 20211 + }, + { + "epoch": 1.3, + "learning_rate": 1.5684321260653945e-05, + "loss": 0.9484, + "step": 20212 + }, + { + "epoch": 1.3, + "learning_rate": 1.5684106571630994e-05, + "loss": 0.9683, + "step": 20213 + }, + { + "epoch": 1.3, + "learning_rate": 1.5683891882608046e-05, + "loss": 0.9785, + "step": 20214 + }, + { + "epoch": 1.3, + "learning_rate": 1.5683677193585095e-05, + "loss": 1.0317, + "step": 20215 + }, + { + "epoch": 1.3, + "learning_rate": 1.5683462504562144e-05, + "loss": 0.9828, + "step": 20216 + }, + { + "epoch": 1.3, + "learning_rate": 1.5683247815539193e-05, + "loss": 0.97, + "step": 20217 + }, + { + "epoch": 1.3, + "learning_rate": 1.5683033126516242e-05, + "loss": 1.0325, + "step": 20218 + }, + { + "epoch": 1.3, + "learning_rate": 1.568281843749329e-05, + "loss": 0.9592, + "step": 20219 + }, + { + "epoch": 1.3, + "learning_rate": 1.5682603748470343e-05, + "loss": 0.9855, + "step": 20220 + }, + { + "epoch": 1.3, + "learning_rate": 1.5682389059447392e-05, + "loss": 0.9303, + "step": 20221 + }, + { + "epoch": 1.3, + "learning_rate": 1.568217437042444e-05, + "loss": 0.9536, + "step": 20222 + }, + { + "epoch": 1.3, + "learning_rate": 1.568195968140149e-05, + "loss": 1.016, + "step": 20223 + }, + { + "epoch": 1.3, + "learning_rate": 1.568174499237854e-05, + "loss": 0.9652, + "step": 20224 + }, + { + "epoch": 1.3, + "learning_rate": 1.568153030335559e-05, + "loss": 1.0178, + "step": 20225 + }, + { + "epoch": 1.3, + "learning_rate": 1.568131561433264e-05, + "loss": 0.9948, + "step": 20226 + }, + { + "epoch": 1.3, + "learning_rate": 1.5681100925309693e-05, + "loss": 0.9679, + "step": 20227 + }, + { + "epoch": 1.3, + "learning_rate": 1.568088623628674e-05, + "loss": 1.0363, + "step": 20228 + }, + { + "epoch": 1.3, + "learning_rate": 1.568067154726379e-05, + "loss": 1.0349, + "step": 20229 + }, + { + "epoch": 1.3, + "learning_rate": 1.568045685824084e-05, + "loss": 0.9912, + "step": 20230 + }, + { + "epoch": 1.3, + "learning_rate": 1.568024216921789e-05, + "loss": 0.9634, + "step": 20231 + }, + { + "epoch": 1.3, + "learning_rate": 1.568002748019494e-05, + "loss": 1.0586, + "step": 20232 + }, + { + "epoch": 1.3, + "learning_rate": 1.567981279117199e-05, + "loss": 1.0218, + "step": 20233 + }, + { + "epoch": 1.3, + "learning_rate": 1.567959810214904e-05, + "loss": 0.953, + "step": 20234 + }, + { + "epoch": 1.3, + "learning_rate": 1.5679383413126088e-05, + "loss": 0.9354, + "step": 20235 + }, + { + "epoch": 1.3, + "learning_rate": 1.5679168724103137e-05, + "loss": 0.9491, + "step": 20236 + }, + { + "epoch": 1.3, + "learning_rate": 1.5678954035080186e-05, + "loss": 0.9866, + "step": 20237 + }, + { + "epoch": 1.3, + "learning_rate": 1.5678739346057238e-05, + "loss": 1.0211, + "step": 20238 + }, + { + "epoch": 1.3, + "learning_rate": 1.5678524657034287e-05, + "loss": 1.0169, + "step": 20239 + }, + { + "epoch": 1.3, + "learning_rate": 1.5678309968011336e-05, + "loss": 1.0363, + "step": 20240 + }, + { + "epoch": 1.3, + "learning_rate": 1.5678095278988385e-05, + "loss": 1.0195, + "step": 20241 + }, + { + "epoch": 1.3, + "learning_rate": 1.5677880589965437e-05, + "loss": 0.94, + "step": 20242 + }, + { + "epoch": 1.3, + "learning_rate": 1.5677665900942486e-05, + "loss": 0.9605, + "step": 20243 + }, + { + "epoch": 1.3, + "learning_rate": 1.5677451211919535e-05, + "loss": 0.9957, + "step": 20244 + }, + { + "epoch": 1.3, + "learning_rate": 1.5677236522896587e-05, + "loss": 1.0091, + "step": 20245 + }, + { + "epoch": 1.3, + "learning_rate": 1.5677021833873636e-05, + "loss": 0.9885, + "step": 20246 + }, + { + "epoch": 1.3, + "learning_rate": 1.5676807144850685e-05, + "loss": 1.0453, + "step": 20247 + }, + { + "epoch": 1.3, + "learning_rate": 1.5676592455827734e-05, + "loss": 0.943, + "step": 20248 + }, + { + "epoch": 1.3, + "learning_rate": 1.5676377766804783e-05, + "loss": 0.9745, + "step": 20249 + }, + { + "epoch": 1.3, + "learning_rate": 1.5676163077781836e-05, + "loss": 1.0064, + "step": 20250 + }, + { + "epoch": 1.3, + "learning_rate": 1.5675948388758885e-05, + "loss": 0.9982, + "step": 20251 + }, + { + "epoch": 1.3, + "learning_rate": 1.5675733699735934e-05, + "loss": 0.9492, + "step": 20252 + }, + { + "epoch": 1.3, + "learning_rate": 1.5675519010712983e-05, + "loss": 0.9929, + "step": 20253 + }, + { + "epoch": 1.3, + "learning_rate": 1.567530432169003e-05, + "loss": 1.0149, + "step": 20254 + }, + { + "epoch": 1.3, + "learning_rate": 1.567508963266708e-05, + "loss": 0.9936, + "step": 20255 + }, + { + "epoch": 1.3, + "learning_rate": 1.5674874943644133e-05, + "loss": 0.9454, + "step": 20256 + }, + { + "epoch": 1.3, + "learning_rate": 1.5674660254621182e-05, + "loss": 1.0321, + "step": 20257 + }, + { + "epoch": 1.3, + "learning_rate": 1.5674445565598234e-05, + "loss": 0.9809, + "step": 20258 + }, + { + "epoch": 1.3, + "learning_rate": 1.5674230876575283e-05, + "loss": 0.9654, + "step": 20259 + }, + { + "epoch": 1.3, + "learning_rate": 1.5674016187552332e-05, + "loss": 1.0021, + "step": 20260 + }, + { + "epoch": 1.3, + "learning_rate": 1.567380149852938e-05, + "loss": 1.0491, + "step": 20261 + }, + { + "epoch": 1.3, + "learning_rate": 1.567358680950643e-05, + "loss": 0.9278, + "step": 20262 + }, + { + "epoch": 1.3, + "learning_rate": 1.5673372120483482e-05, + "loss": 0.9377, + "step": 20263 + }, + { + "epoch": 1.3, + "learning_rate": 1.567315743146053e-05, + "loss": 0.9928, + "step": 20264 + }, + { + "epoch": 1.3, + "learning_rate": 1.567294274243758e-05, + "loss": 0.9752, + "step": 20265 + }, + { + "epoch": 1.3, + "learning_rate": 1.567272805341463e-05, + "loss": 1.0051, + "step": 20266 + }, + { + "epoch": 1.3, + "learning_rate": 1.5672513364391678e-05, + "loss": 0.9431, + "step": 20267 + }, + { + "epoch": 1.3, + "learning_rate": 1.567229867536873e-05, + "loss": 0.9927, + "step": 20268 + }, + { + "epoch": 1.3, + "learning_rate": 1.567208398634578e-05, + "loss": 1.0457, + "step": 20269 + }, + { + "epoch": 1.3, + "learning_rate": 1.567186929732283e-05, + "loss": 0.9463, + "step": 20270 + }, + { + "epoch": 1.3, + "learning_rate": 1.567165460829988e-05, + "loss": 0.9706, + "step": 20271 + }, + { + "epoch": 1.3, + "learning_rate": 1.567143991927693e-05, + "loss": 1.0147, + "step": 20272 + }, + { + "epoch": 1.3, + "learning_rate": 1.567122523025398e-05, + "loss": 1.0273, + "step": 20273 + }, + { + "epoch": 1.3, + "learning_rate": 1.5671010541231028e-05, + "loss": 1.0252, + "step": 20274 + }, + { + "epoch": 1.3, + "learning_rate": 1.5670795852208077e-05, + "loss": 0.9271, + "step": 20275 + }, + { + "epoch": 1.3, + "learning_rate": 1.567058116318513e-05, + "loss": 0.9797, + "step": 20276 + }, + { + "epoch": 1.3, + "learning_rate": 1.5670366474162178e-05, + "loss": 0.9283, + "step": 20277 + }, + { + "epoch": 1.3, + "learning_rate": 1.5670151785139227e-05, + "loss": 1.0057, + "step": 20278 + }, + { + "epoch": 1.3, + "learning_rate": 1.5669937096116276e-05, + "loss": 0.9418, + "step": 20279 + }, + { + "epoch": 1.3, + "learning_rate": 1.5669722407093325e-05, + "loss": 1.0119, + "step": 20280 + }, + { + "epoch": 1.3, + "learning_rate": 1.5669507718070377e-05, + "loss": 0.9344, + "step": 20281 + }, + { + "epoch": 1.3, + "learning_rate": 1.5669293029047426e-05, + "loss": 1.0775, + "step": 20282 + }, + { + "epoch": 1.3, + "learning_rate": 1.5669078340024475e-05, + "loss": 0.9986, + "step": 20283 + }, + { + "epoch": 1.31, + "learning_rate": 1.5668863651001524e-05, + "loss": 0.9597, + "step": 20284 + }, + { + "epoch": 1.31, + "learning_rate": 1.5668648961978573e-05, + "loss": 0.9713, + "step": 20285 + }, + { + "epoch": 1.31, + "learning_rate": 1.5668434272955626e-05, + "loss": 1.0526, + "step": 20286 + }, + { + "epoch": 1.31, + "learning_rate": 1.5668219583932675e-05, + "loss": 1.0235, + "step": 20287 + }, + { + "epoch": 1.31, + "learning_rate": 1.5668004894909727e-05, + "loss": 0.9326, + "step": 20288 + }, + { + "epoch": 1.31, + "learning_rate": 1.5667790205886776e-05, + "loss": 1.0582, + "step": 20289 + }, + { + "epoch": 1.31, + "learning_rate": 1.5667575516863825e-05, + "loss": 0.9998, + "step": 20290 + }, + { + "epoch": 1.31, + "learning_rate": 1.5667360827840874e-05, + "loss": 1.0374, + "step": 20291 + }, + { + "epoch": 1.31, + "learning_rate": 1.5667146138817923e-05, + "loss": 1.044, + "step": 20292 + }, + { + "epoch": 1.31, + "learning_rate": 1.566693144979497e-05, + "loss": 1.0129, + "step": 20293 + }, + { + "epoch": 1.31, + "learning_rate": 1.5666716760772024e-05, + "loss": 0.9993, + "step": 20294 + }, + { + "epoch": 1.31, + "learning_rate": 1.5666502071749073e-05, + "loss": 0.9177, + "step": 20295 + }, + { + "epoch": 1.31, + "learning_rate": 1.5666287382726122e-05, + "loss": 0.987, + "step": 20296 + }, + { + "epoch": 1.31, + "learning_rate": 1.566607269370317e-05, + "loss": 0.9483, + "step": 20297 + }, + { + "epoch": 1.31, + "learning_rate": 1.566585800468022e-05, + "loss": 1.0018, + "step": 20298 + }, + { + "epoch": 1.31, + "learning_rate": 1.5665643315657272e-05, + "loss": 0.9597, + "step": 20299 + }, + { + "epoch": 1.31, + "learning_rate": 1.566542862663432e-05, + "loss": 0.9983, + "step": 20300 + }, + { + "epoch": 1.31, + "learning_rate": 1.5665213937611374e-05, + "loss": 0.8899, + "step": 20301 + }, + { + "epoch": 1.31, + "learning_rate": 1.5664999248588423e-05, + "loss": 0.8974, + "step": 20302 + }, + { + "epoch": 1.31, + "learning_rate": 1.566478455956547e-05, + "loss": 0.9599, + "step": 20303 + }, + { + "epoch": 1.31, + "learning_rate": 1.566456987054252e-05, + "loss": 0.9137, + "step": 20304 + }, + { + "epoch": 1.31, + "learning_rate": 1.566435518151957e-05, + "loss": 0.9353, + "step": 20305 + }, + { + "epoch": 1.31, + "learning_rate": 1.566414049249662e-05, + "loss": 0.9647, + "step": 20306 + }, + { + "epoch": 1.31, + "learning_rate": 1.566392580347367e-05, + "loss": 1.0403, + "step": 20307 + }, + { + "epoch": 1.31, + "learning_rate": 1.566371111445072e-05, + "loss": 1.0631, + "step": 20308 + }, + { + "epoch": 1.31, + "learning_rate": 1.566349642542777e-05, + "loss": 1.0227, + "step": 20309 + }, + { + "epoch": 1.31, + "learning_rate": 1.5663281736404818e-05, + "loss": 0.9595, + "step": 20310 + }, + { + "epoch": 1.31, + "learning_rate": 1.5663067047381867e-05, + "loss": 1.024, + "step": 20311 + }, + { + "epoch": 1.31, + "learning_rate": 1.566285235835892e-05, + "loss": 0.9855, + "step": 20312 + }, + { + "epoch": 1.31, + "learning_rate": 1.5662637669335968e-05, + "loss": 0.9911, + "step": 20313 + }, + { + "epoch": 1.31, + "learning_rate": 1.5662422980313017e-05, + "loss": 1.0032, + "step": 20314 + }, + { + "epoch": 1.31, + "learning_rate": 1.566220829129007e-05, + "loss": 1.019, + "step": 20315 + }, + { + "epoch": 1.31, + "learning_rate": 1.5661993602267118e-05, + "loss": 0.9108, + "step": 20316 + }, + { + "epoch": 1.31, + "learning_rate": 1.5661778913244167e-05, + "loss": 1.0419, + "step": 20317 + }, + { + "epoch": 1.31, + "learning_rate": 1.5661564224221216e-05, + "loss": 1.0596, + "step": 20318 + }, + { + "epoch": 1.31, + "learning_rate": 1.566134953519827e-05, + "loss": 1.0234, + "step": 20319 + }, + { + "epoch": 1.31, + "learning_rate": 1.5661134846175317e-05, + "loss": 0.9734, + "step": 20320 + }, + { + "epoch": 1.31, + "learning_rate": 1.5660920157152366e-05, + "loss": 0.9594, + "step": 20321 + }, + { + "epoch": 1.31, + "learning_rate": 1.5660705468129415e-05, + "loss": 1.0372, + "step": 20322 + }, + { + "epoch": 1.31, + "learning_rate": 1.5660490779106464e-05, + "loss": 1.0265, + "step": 20323 + }, + { + "epoch": 1.31, + "learning_rate": 1.5660276090083513e-05, + "loss": 0.9972, + "step": 20324 + }, + { + "epoch": 1.31, + "learning_rate": 1.5660061401060566e-05, + "loss": 0.9685, + "step": 20325 + }, + { + "epoch": 1.31, + "learning_rate": 1.5659846712037615e-05, + "loss": 0.9985, + "step": 20326 + }, + { + "epoch": 1.31, + "learning_rate": 1.5659632023014664e-05, + "loss": 1.0427, + "step": 20327 + }, + { + "epoch": 1.31, + "learning_rate": 1.5659417333991713e-05, + "loss": 0.9563, + "step": 20328 + }, + { + "epoch": 1.31, + "learning_rate": 1.5659202644968765e-05, + "loss": 0.9993, + "step": 20329 + }, + { + "epoch": 1.31, + "learning_rate": 1.5658987955945814e-05, + "loss": 0.9759, + "step": 20330 + }, + { + "epoch": 1.31, + "learning_rate": 1.5658773266922863e-05, + "loss": 0.9784, + "step": 20331 + }, + { + "epoch": 1.31, + "learning_rate": 1.5658558577899915e-05, + "loss": 0.9452, + "step": 20332 + }, + { + "epoch": 1.31, + "learning_rate": 1.5658343888876964e-05, + "loss": 0.9184, + "step": 20333 + }, + { + "epoch": 1.31, + "learning_rate": 1.5658129199854013e-05, + "loss": 1.0103, + "step": 20334 + }, + { + "epoch": 1.31, + "learning_rate": 1.5657914510831062e-05, + "loss": 0.9395, + "step": 20335 + }, + { + "epoch": 1.31, + "learning_rate": 1.565769982180811e-05, + "loss": 1.0508, + "step": 20336 + }, + { + "epoch": 1.31, + "learning_rate": 1.5657485132785163e-05, + "loss": 0.9964, + "step": 20337 + }, + { + "epoch": 1.31, + "learning_rate": 1.5657270443762212e-05, + "loss": 0.9504, + "step": 20338 + }, + { + "epoch": 1.31, + "learning_rate": 1.565705575473926e-05, + "loss": 0.9956, + "step": 20339 + }, + { + "epoch": 1.31, + "learning_rate": 1.565684106571631e-05, + "loss": 0.9747, + "step": 20340 + }, + { + "epoch": 1.31, + "learning_rate": 1.565662637669336e-05, + "loss": 0.9882, + "step": 20341 + }, + { + "epoch": 1.31, + "learning_rate": 1.5656411687670408e-05, + "loss": 1.0378, + "step": 20342 + }, + { + "epoch": 1.31, + "learning_rate": 1.565619699864746e-05, + "loss": 0.9899, + "step": 20343 + }, + { + "epoch": 1.31, + "learning_rate": 1.565598230962451e-05, + "loss": 0.9612, + "step": 20344 + }, + { + "epoch": 1.31, + "learning_rate": 1.5655767620601562e-05, + "loss": 0.911, + "step": 20345 + }, + { + "epoch": 1.31, + "learning_rate": 1.565555293157861e-05, + "loss": 1.0005, + "step": 20346 + }, + { + "epoch": 1.31, + "learning_rate": 1.565533824255566e-05, + "loss": 1.0869, + "step": 20347 + }, + { + "epoch": 1.31, + "learning_rate": 1.565512355353271e-05, + "loss": 0.9513, + "step": 20348 + }, + { + "epoch": 1.31, + "learning_rate": 1.5654908864509758e-05, + "loss": 0.9753, + "step": 20349 + }, + { + "epoch": 1.31, + "learning_rate": 1.565469417548681e-05, + "loss": 1.0158, + "step": 20350 + }, + { + "epoch": 1.31, + "learning_rate": 1.565447948646386e-05, + "loss": 0.9639, + "step": 20351 + }, + { + "epoch": 1.31, + "learning_rate": 1.5654264797440908e-05, + "loss": 0.9821, + "step": 20352 + }, + { + "epoch": 1.31, + "learning_rate": 1.5654050108417957e-05, + "loss": 0.9314, + "step": 20353 + }, + { + "epoch": 1.31, + "learning_rate": 1.5653835419395006e-05, + "loss": 0.9375, + "step": 20354 + }, + { + "epoch": 1.31, + "learning_rate": 1.565362073037206e-05, + "loss": 0.9162, + "step": 20355 + }, + { + "epoch": 1.31, + "learning_rate": 1.5653406041349107e-05, + "loss": 0.9764, + "step": 20356 + }, + { + "epoch": 1.31, + "learning_rate": 1.5653191352326156e-05, + "loss": 0.9821, + "step": 20357 + }, + { + "epoch": 1.31, + "learning_rate": 1.565297666330321e-05, + "loss": 0.9412, + "step": 20358 + }, + { + "epoch": 1.31, + "learning_rate": 1.5652761974280258e-05, + "loss": 0.9632, + "step": 20359 + }, + { + "epoch": 1.31, + "learning_rate": 1.5652547285257307e-05, + "loss": 0.9123, + "step": 20360 + }, + { + "epoch": 1.31, + "learning_rate": 1.5652332596234356e-05, + "loss": 1.0062, + "step": 20361 + }, + { + "epoch": 1.31, + "learning_rate": 1.5652117907211404e-05, + "loss": 1.0065, + "step": 20362 + }, + { + "epoch": 1.31, + "learning_rate": 1.5651903218188457e-05, + "loss": 1.0033, + "step": 20363 + }, + { + "epoch": 1.31, + "learning_rate": 1.5651688529165506e-05, + "loss": 0.9544, + "step": 20364 + }, + { + "epoch": 1.31, + "learning_rate": 1.5651473840142555e-05, + "loss": 0.9311, + "step": 20365 + }, + { + "epoch": 1.31, + "learning_rate": 1.5651259151119604e-05, + "loss": 1.0351, + "step": 20366 + }, + { + "epoch": 1.31, + "learning_rate": 1.5651044462096653e-05, + "loss": 0.953, + "step": 20367 + }, + { + "epoch": 1.31, + "learning_rate": 1.5650829773073705e-05, + "loss": 0.9354, + "step": 20368 + }, + { + "epoch": 1.31, + "learning_rate": 1.5650615084050754e-05, + "loss": 0.9851, + "step": 20369 + }, + { + "epoch": 1.31, + "learning_rate": 1.5650400395027803e-05, + "loss": 1.0102, + "step": 20370 + }, + { + "epoch": 1.31, + "learning_rate": 1.5650185706004852e-05, + "loss": 1.0176, + "step": 20371 + }, + { + "epoch": 1.31, + "learning_rate": 1.56499710169819e-05, + "loss": 1.0925, + "step": 20372 + }, + { + "epoch": 1.31, + "learning_rate": 1.5649756327958953e-05, + "loss": 0.92, + "step": 20373 + }, + { + "epoch": 1.31, + "learning_rate": 1.5649541638936002e-05, + "loss": 0.9811, + "step": 20374 + }, + { + "epoch": 1.31, + "learning_rate": 1.5649326949913055e-05, + "loss": 0.9488, + "step": 20375 + }, + { + "epoch": 1.31, + "learning_rate": 1.5649112260890104e-05, + "loss": 0.9533, + "step": 20376 + }, + { + "epoch": 1.31, + "learning_rate": 1.5648897571867153e-05, + "loss": 1.0078, + "step": 20377 + }, + { + "epoch": 1.31, + "learning_rate": 1.56486828828442e-05, + "loss": 0.9698, + "step": 20378 + }, + { + "epoch": 1.31, + "learning_rate": 1.564846819382125e-05, + "loss": 0.9769, + "step": 20379 + }, + { + "epoch": 1.31, + "learning_rate": 1.56482535047983e-05, + "loss": 1.0232, + "step": 20380 + }, + { + "epoch": 1.31, + "learning_rate": 1.5648038815775352e-05, + "loss": 0.9652, + "step": 20381 + }, + { + "epoch": 1.31, + "learning_rate": 1.56478241267524e-05, + "loss": 0.8865, + "step": 20382 + }, + { + "epoch": 1.31, + "learning_rate": 1.564760943772945e-05, + "loss": 0.9606, + "step": 20383 + }, + { + "epoch": 1.31, + "learning_rate": 1.56473947487065e-05, + "loss": 0.9807, + "step": 20384 + }, + { + "epoch": 1.31, + "learning_rate": 1.5647180059683548e-05, + "loss": 0.9517, + "step": 20385 + }, + { + "epoch": 1.31, + "learning_rate": 1.56469653706606e-05, + "loss": 0.9546, + "step": 20386 + }, + { + "epoch": 1.31, + "learning_rate": 1.564675068163765e-05, + "loss": 0.9606, + "step": 20387 + }, + { + "epoch": 1.31, + "learning_rate": 1.56465359926147e-05, + "loss": 0.9822, + "step": 20388 + }, + { + "epoch": 1.31, + "learning_rate": 1.564632130359175e-05, + "loss": 0.9819, + "step": 20389 + }, + { + "epoch": 1.31, + "learning_rate": 1.56461066145688e-05, + "loss": 0.9626, + "step": 20390 + }, + { + "epoch": 1.31, + "learning_rate": 1.5645891925545848e-05, + "loss": 0.9523, + "step": 20391 + }, + { + "epoch": 1.31, + "learning_rate": 1.5645677236522897e-05, + "loss": 1.1181, + "step": 20392 + }, + { + "epoch": 1.31, + "learning_rate": 1.5645462547499946e-05, + "loss": 1.0326, + "step": 20393 + }, + { + "epoch": 1.31, + "learning_rate": 1.5645247858477e-05, + "loss": 0.8951, + "step": 20394 + }, + { + "epoch": 1.31, + "learning_rate": 1.5645033169454047e-05, + "loss": 0.9494, + "step": 20395 + }, + { + "epoch": 1.31, + "learning_rate": 1.5644818480431096e-05, + "loss": 0.9623, + "step": 20396 + }, + { + "epoch": 1.31, + "learning_rate": 1.5644603791408145e-05, + "loss": 0.9719, + "step": 20397 + }, + { + "epoch": 1.31, + "learning_rate": 1.5644389102385194e-05, + "loss": 0.9661, + "step": 20398 + }, + { + "epoch": 1.31, + "learning_rate": 1.5644174413362247e-05, + "loss": 1.043, + "step": 20399 + }, + { + "epoch": 1.31, + "learning_rate": 1.5643959724339296e-05, + "loss": 0.9545, + "step": 20400 + }, + { + "epoch": 1.31, + "learning_rate": 1.5643745035316345e-05, + "loss": 0.9896, + "step": 20401 + }, + { + "epoch": 1.31, + "learning_rate": 1.5643530346293397e-05, + "loss": 0.9861, + "step": 20402 + }, + { + "epoch": 1.31, + "learning_rate": 1.5643315657270446e-05, + "loss": 1.0239, + "step": 20403 + }, + { + "epoch": 1.31, + "learning_rate": 1.5643100968247495e-05, + "loss": 0.9812, + "step": 20404 + }, + { + "epoch": 1.31, + "learning_rate": 1.5642886279224544e-05, + "loss": 0.9336, + "step": 20405 + }, + { + "epoch": 1.31, + "learning_rate": 1.5642671590201596e-05, + "loss": 1.0838, + "step": 20406 + }, + { + "epoch": 1.31, + "learning_rate": 1.5642456901178645e-05, + "loss": 0.9885, + "step": 20407 + }, + { + "epoch": 1.31, + "learning_rate": 1.5642242212155694e-05, + "loss": 0.9636, + "step": 20408 + }, + { + "epoch": 1.31, + "learning_rate": 1.5642027523132743e-05, + "loss": 0.9149, + "step": 20409 + }, + { + "epoch": 1.31, + "learning_rate": 1.5641812834109792e-05, + "loss": 1.0017, + "step": 20410 + }, + { + "epoch": 1.31, + "learning_rate": 1.564159814508684e-05, + "loss": 0.9823, + "step": 20411 + }, + { + "epoch": 1.31, + "learning_rate": 1.5641383456063893e-05, + "loss": 0.985, + "step": 20412 + }, + { + "epoch": 1.31, + "learning_rate": 1.5641168767040942e-05, + "loss": 1.0383, + "step": 20413 + }, + { + "epoch": 1.31, + "learning_rate": 1.564095407801799e-05, + "loss": 1.0419, + "step": 20414 + }, + { + "epoch": 1.31, + "learning_rate": 1.564073938899504e-05, + "loss": 0.8928, + "step": 20415 + }, + { + "epoch": 1.31, + "learning_rate": 1.5640524699972093e-05, + "loss": 0.978, + "step": 20416 + }, + { + "epoch": 1.31, + "learning_rate": 1.564031001094914e-05, + "loss": 0.986, + "step": 20417 + }, + { + "epoch": 1.31, + "learning_rate": 1.564009532192619e-05, + "loss": 1.0437, + "step": 20418 + }, + { + "epoch": 1.31, + "learning_rate": 1.5639880632903243e-05, + "loss": 0.9608, + "step": 20419 + }, + { + "epoch": 1.31, + "learning_rate": 1.5639665943880292e-05, + "loss": 1.0138, + "step": 20420 + }, + { + "epoch": 1.31, + "learning_rate": 1.563945125485734e-05, + "loss": 0.8886, + "step": 20421 + }, + { + "epoch": 1.31, + "learning_rate": 1.563923656583439e-05, + "loss": 0.9834, + "step": 20422 + }, + { + "epoch": 1.31, + "learning_rate": 1.563902187681144e-05, + "loss": 0.9669, + "step": 20423 + }, + { + "epoch": 1.31, + "learning_rate": 1.563880718778849e-05, + "loss": 0.9074, + "step": 20424 + }, + { + "epoch": 1.31, + "learning_rate": 1.563859249876554e-05, + "loss": 0.9862, + "step": 20425 + }, + { + "epoch": 1.31, + "learning_rate": 1.563837780974259e-05, + "loss": 0.9798, + "step": 20426 + }, + { + "epoch": 1.31, + "learning_rate": 1.5638163120719638e-05, + "loss": 1.0377, + "step": 20427 + }, + { + "epoch": 1.31, + "learning_rate": 1.5637948431696687e-05, + "loss": 0.9462, + "step": 20428 + }, + { + "epoch": 1.31, + "learning_rate": 1.5637733742673736e-05, + "loss": 1.0129, + "step": 20429 + }, + { + "epoch": 1.31, + "learning_rate": 1.5637519053650788e-05, + "loss": 0.9951, + "step": 20430 + }, + { + "epoch": 1.31, + "learning_rate": 1.5637304364627837e-05, + "loss": 0.9442, + "step": 20431 + }, + { + "epoch": 1.31, + "learning_rate": 1.563708967560489e-05, + "loss": 0.9747, + "step": 20432 + }, + { + "epoch": 1.31, + "learning_rate": 1.563687498658194e-05, + "loss": 0.9604, + "step": 20433 + }, + { + "epoch": 1.31, + "learning_rate": 1.5636660297558988e-05, + "loss": 1.0123, + "step": 20434 + }, + { + "epoch": 1.31, + "learning_rate": 1.5636445608536037e-05, + "loss": 0.9309, + "step": 20435 + }, + { + "epoch": 1.31, + "learning_rate": 1.5636230919513085e-05, + "loss": 0.9978, + "step": 20436 + }, + { + "epoch": 1.31, + "learning_rate": 1.5636016230490138e-05, + "loss": 0.9683, + "step": 20437 + }, + { + "epoch": 1.31, + "learning_rate": 1.5635801541467187e-05, + "loss": 1.0061, + "step": 20438 + }, + { + "epoch": 1.31, + "learning_rate": 1.5635586852444236e-05, + "loss": 0.9064, + "step": 20439 + }, + { + "epoch": 1.32, + "learning_rate": 1.5635372163421285e-05, + "loss": 0.988, + "step": 20440 + }, + { + "epoch": 1.32, + "learning_rate": 1.5635157474398334e-05, + "loss": 1.044, + "step": 20441 + }, + { + "epoch": 1.32, + "learning_rate": 1.5634942785375383e-05, + "loss": 1.0547, + "step": 20442 + }, + { + "epoch": 1.32, + "learning_rate": 1.5634728096352435e-05, + "loss": 1.015, + "step": 20443 + }, + { + "epoch": 1.32, + "learning_rate": 1.5634513407329484e-05, + "loss": 0.9461, + "step": 20444 + }, + { + "epoch": 1.32, + "learning_rate": 1.5634298718306536e-05, + "loss": 0.9661, + "step": 20445 + }, + { + "epoch": 1.32, + "learning_rate": 1.5634084029283585e-05, + "loss": 1.0355, + "step": 20446 + }, + { + "epoch": 1.32, + "learning_rate": 1.5633869340260634e-05, + "loss": 0.9576, + "step": 20447 + }, + { + "epoch": 1.32, + "learning_rate": 1.5633654651237683e-05, + "loss": 1.0125, + "step": 20448 + }, + { + "epoch": 1.32, + "learning_rate": 1.5633439962214732e-05, + "loss": 0.9616, + "step": 20449 + }, + { + "epoch": 1.32, + "learning_rate": 1.5633225273191785e-05, + "loss": 0.9682, + "step": 20450 + }, + { + "epoch": 1.32, + "learning_rate": 1.5633010584168834e-05, + "loss": 0.9221, + "step": 20451 + }, + { + "epoch": 1.32, + "learning_rate": 1.5632795895145882e-05, + "loss": 0.975, + "step": 20452 + }, + { + "epoch": 1.32, + "learning_rate": 1.563258120612293e-05, + "loss": 0.9337, + "step": 20453 + }, + { + "epoch": 1.32, + "learning_rate": 1.563236651709998e-05, + "loss": 0.9966, + "step": 20454 + }, + { + "epoch": 1.32, + "learning_rate": 1.5632151828077033e-05, + "loss": 1.0927, + "step": 20455 + }, + { + "epoch": 1.32, + "learning_rate": 1.5631937139054082e-05, + "loss": 0.9659, + "step": 20456 + }, + { + "epoch": 1.32, + "learning_rate": 1.563172245003113e-05, + "loss": 0.9472, + "step": 20457 + }, + { + "epoch": 1.32, + "learning_rate": 1.563150776100818e-05, + "loss": 0.9537, + "step": 20458 + }, + { + "epoch": 1.32, + "learning_rate": 1.563129307198523e-05, + "loss": 0.9569, + "step": 20459 + }, + { + "epoch": 1.32, + "learning_rate": 1.563107838296228e-05, + "loss": 1.0532, + "step": 20460 + }, + { + "epoch": 1.32, + "learning_rate": 1.563086369393933e-05, + "loss": 0.969, + "step": 20461 + }, + { + "epoch": 1.32, + "learning_rate": 1.5630649004916382e-05, + "loss": 0.9944, + "step": 20462 + }, + { + "epoch": 1.32, + "learning_rate": 1.563043431589343e-05, + "loss": 0.9507, + "step": 20463 + }, + { + "epoch": 1.32, + "learning_rate": 1.563021962687048e-05, + "loss": 0.9322, + "step": 20464 + }, + { + "epoch": 1.32, + "learning_rate": 1.563000493784753e-05, + "loss": 0.9587, + "step": 20465 + }, + { + "epoch": 1.32, + "learning_rate": 1.5629790248824578e-05, + "loss": 0.8887, + "step": 20466 + }, + { + "epoch": 1.32, + "learning_rate": 1.5629575559801627e-05, + "loss": 0.9694, + "step": 20467 + }, + { + "epoch": 1.32, + "learning_rate": 1.562936087077868e-05, + "loss": 1.064, + "step": 20468 + }, + { + "epoch": 1.32, + "learning_rate": 1.562914618175573e-05, + "loss": 1.0145, + "step": 20469 + }, + { + "epoch": 1.32, + "learning_rate": 1.5628931492732777e-05, + "loss": 1.0087, + "step": 20470 + }, + { + "epoch": 1.32, + "learning_rate": 1.5628716803709826e-05, + "loss": 0.9508, + "step": 20471 + }, + { + "epoch": 1.32, + "learning_rate": 1.5628502114686875e-05, + "loss": 1.047, + "step": 20472 + }, + { + "epoch": 1.32, + "learning_rate": 1.5628287425663928e-05, + "loss": 0.9202, + "step": 20473 + }, + { + "epoch": 1.32, + "learning_rate": 1.5628072736640977e-05, + "loss": 1.0532, + "step": 20474 + }, + { + "epoch": 1.32, + "learning_rate": 1.562785804761803e-05, + "loss": 0.9978, + "step": 20475 + }, + { + "epoch": 1.32, + "learning_rate": 1.5627643358595078e-05, + "loss": 1.0328, + "step": 20476 + }, + { + "epoch": 1.32, + "learning_rate": 1.5627428669572127e-05, + "loss": 1.0132, + "step": 20477 + }, + { + "epoch": 1.32, + "learning_rate": 1.5627213980549176e-05, + "loss": 0.953, + "step": 20478 + }, + { + "epoch": 1.32, + "learning_rate": 1.5626999291526225e-05, + "loss": 0.9363, + "step": 20479 + }, + { + "epoch": 1.32, + "learning_rate": 1.5626784602503274e-05, + "loss": 1.032, + "step": 20480 + }, + { + "epoch": 1.32, + "learning_rate": 1.5626569913480326e-05, + "loss": 0.9429, + "step": 20481 + }, + { + "epoch": 1.32, + "learning_rate": 1.5626355224457375e-05, + "loss": 0.979, + "step": 20482 + }, + { + "epoch": 1.32, + "learning_rate": 1.5626140535434424e-05, + "loss": 0.9425, + "step": 20483 + }, + { + "epoch": 1.32, + "learning_rate": 1.5625925846411473e-05, + "loss": 0.9706, + "step": 20484 + }, + { + "epoch": 1.32, + "learning_rate": 1.5625711157388522e-05, + "loss": 0.926, + "step": 20485 + }, + { + "epoch": 1.32, + "learning_rate": 1.5625496468365574e-05, + "loss": 0.9247, + "step": 20486 + }, + { + "epoch": 1.32, + "learning_rate": 1.5625281779342623e-05, + "loss": 0.9364, + "step": 20487 + }, + { + "epoch": 1.32, + "learning_rate": 1.5625067090319672e-05, + "loss": 0.96, + "step": 20488 + }, + { + "epoch": 1.32, + "learning_rate": 1.5624852401296725e-05, + "loss": 0.9738, + "step": 20489 + }, + { + "epoch": 1.32, + "learning_rate": 1.5624637712273774e-05, + "loss": 1.0146, + "step": 20490 + }, + { + "epoch": 1.32, + "learning_rate": 1.5624423023250823e-05, + "loss": 0.9422, + "step": 20491 + }, + { + "epoch": 1.32, + "learning_rate": 1.562420833422787e-05, + "loss": 0.9803, + "step": 20492 + }, + { + "epoch": 1.32, + "learning_rate": 1.5623993645204924e-05, + "loss": 1.0289, + "step": 20493 + }, + { + "epoch": 1.32, + "learning_rate": 1.5623778956181973e-05, + "loss": 0.9481, + "step": 20494 + }, + { + "epoch": 1.32, + "learning_rate": 1.5623564267159022e-05, + "loss": 1.0542, + "step": 20495 + }, + { + "epoch": 1.32, + "learning_rate": 1.562334957813607e-05, + "loss": 0.9656, + "step": 20496 + }, + { + "epoch": 1.32, + "learning_rate": 1.562313488911312e-05, + "loss": 0.9888, + "step": 20497 + }, + { + "epoch": 1.32, + "learning_rate": 1.562292020009017e-05, + "loss": 1.0945, + "step": 20498 + }, + { + "epoch": 1.32, + "learning_rate": 1.562270551106722e-05, + "loss": 0.9412, + "step": 20499 + }, + { + "epoch": 1.32, + "learning_rate": 1.562249082204427e-05, + "loss": 1.0226, + "step": 20500 + }, + { + "epoch": 1.32, + "learning_rate": 1.562227613302132e-05, + "loss": 0.9838, + "step": 20501 + }, + { + "epoch": 1.32, + "learning_rate": 1.5622061443998368e-05, + "loss": 0.9247, + "step": 20502 + }, + { + "epoch": 1.32, + "learning_rate": 1.562184675497542e-05, + "loss": 0.9497, + "step": 20503 + }, + { + "epoch": 1.32, + "learning_rate": 1.562163206595247e-05, + "loss": 0.9925, + "step": 20504 + }, + { + "epoch": 1.32, + "learning_rate": 1.5621417376929518e-05, + "loss": 0.9566, + "step": 20505 + }, + { + "epoch": 1.32, + "learning_rate": 1.562120268790657e-05, + "loss": 0.9187, + "step": 20506 + }, + { + "epoch": 1.32, + "learning_rate": 1.562098799888362e-05, + "loss": 1.0068, + "step": 20507 + }, + { + "epoch": 1.32, + "learning_rate": 1.562077330986067e-05, + "loss": 0.9371, + "step": 20508 + }, + { + "epoch": 1.32, + "learning_rate": 1.5620558620837718e-05, + "loss": 1.0225, + "step": 20509 + }, + { + "epoch": 1.32, + "learning_rate": 1.5620343931814766e-05, + "loss": 1.0215, + "step": 20510 + }, + { + "epoch": 1.32, + "learning_rate": 1.562012924279182e-05, + "loss": 0.9669, + "step": 20511 + }, + { + "epoch": 1.32, + "learning_rate": 1.5619914553768868e-05, + "loss": 0.9996, + "step": 20512 + }, + { + "epoch": 1.32, + "learning_rate": 1.5619699864745917e-05, + "loss": 1.0399, + "step": 20513 + }, + { + "epoch": 1.32, + "learning_rate": 1.5619485175722966e-05, + "loss": 0.9956, + "step": 20514 + }, + { + "epoch": 1.32, + "learning_rate": 1.5619270486700015e-05, + "loss": 0.9336, + "step": 20515 + }, + { + "epoch": 1.32, + "learning_rate": 1.5619055797677064e-05, + "loss": 0.9405, + "step": 20516 + }, + { + "epoch": 1.32, + "learning_rate": 1.5618841108654116e-05, + "loss": 0.98, + "step": 20517 + }, + { + "epoch": 1.32, + "learning_rate": 1.5618626419631165e-05, + "loss": 0.9914, + "step": 20518 + }, + { + "epoch": 1.32, + "learning_rate": 1.5618411730608217e-05, + "loss": 0.9734, + "step": 20519 + }, + { + "epoch": 1.32, + "learning_rate": 1.5618197041585266e-05, + "loss": 0.9815, + "step": 20520 + }, + { + "epoch": 1.32, + "learning_rate": 1.5617982352562315e-05, + "loss": 0.9767, + "step": 20521 + }, + { + "epoch": 1.32, + "learning_rate": 1.5617767663539364e-05, + "loss": 1.0024, + "step": 20522 + }, + { + "epoch": 1.32, + "learning_rate": 1.5617552974516413e-05, + "loss": 0.9838, + "step": 20523 + }, + { + "epoch": 1.32, + "learning_rate": 1.5617338285493466e-05, + "loss": 0.9191, + "step": 20524 + }, + { + "epoch": 1.32, + "learning_rate": 1.5617123596470515e-05, + "loss": 1.0037, + "step": 20525 + }, + { + "epoch": 1.32, + "learning_rate": 1.5616908907447563e-05, + "loss": 0.8986, + "step": 20526 + }, + { + "epoch": 1.32, + "learning_rate": 1.5616694218424612e-05, + "loss": 0.989, + "step": 20527 + }, + { + "epoch": 1.32, + "learning_rate": 1.561647952940166e-05, + "loss": 0.9671, + "step": 20528 + }, + { + "epoch": 1.32, + "learning_rate": 1.561626484037871e-05, + "loss": 0.971, + "step": 20529 + }, + { + "epoch": 1.32, + "learning_rate": 1.5616050151355763e-05, + "loss": 0.8817, + "step": 20530 + }, + { + "epoch": 1.32, + "learning_rate": 1.561583546233281e-05, + "loss": 0.934, + "step": 20531 + }, + { + "epoch": 1.32, + "learning_rate": 1.5615620773309864e-05, + "loss": 0.9721, + "step": 20532 + }, + { + "epoch": 1.32, + "learning_rate": 1.5615406084286913e-05, + "loss": 0.9852, + "step": 20533 + }, + { + "epoch": 1.32, + "learning_rate": 1.5615191395263962e-05, + "loss": 0.9881, + "step": 20534 + }, + { + "epoch": 1.32, + "learning_rate": 1.561497670624101e-05, + "loss": 0.9889, + "step": 20535 + }, + { + "epoch": 1.32, + "learning_rate": 1.561476201721806e-05, + "loss": 0.9386, + "step": 20536 + }, + { + "epoch": 1.32, + "learning_rate": 1.5614547328195112e-05, + "loss": 0.984, + "step": 20537 + }, + { + "epoch": 1.32, + "learning_rate": 1.561433263917216e-05, + "loss": 0.9274, + "step": 20538 + }, + { + "epoch": 1.32, + "learning_rate": 1.561411795014921e-05, + "loss": 0.9745, + "step": 20539 + }, + { + "epoch": 1.32, + "learning_rate": 1.561390326112626e-05, + "loss": 0.9756, + "step": 20540 + }, + { + "epoch": 1.32, + "learning_rate": 1.5613688572103308e-05, + "loss": 0.9397, + "step": 20541 + }, + { + "epoch": 1.32, + "learning_rate": 1.561347388308036e-05, + "loss": 1.0257, + "step": 20542 + }, + { + "epoch": 1.32, + "learning_rate": 1.561325919405741e-05, + "loss": 0.9803, + "step": 20543 + }, + { + "epoch": 1.32, + "learning_rate": 1.561304450503446e-05, + "loss": 0.9443, + "step": 20544 + }, + { + "epoch": 1.32, + "learning_rate": 1.5612829816011507e-05, + "loss": 0.901, + "step": 20545 + }, + { + "epoch": 1.32, + "learning_rate": 1.5612615126988556e-05, + "loss": 0.9185, + "step": 20546 + }, + { + "epoch": 1.32, + "learning_rate": 1.561240043796561e-05, + "loss": 0.9729, + "step": 20547 + }, + { + "epoch": 1.32, + "learning_rate": 1.5612185748942658e-05, + "loss": 1.0443, + "step": 20548 + }, + { + "epoch": 1.32, + "learning_rate": 1.5611971059919707e-05, + "loss": 0.8504, + "step": 20549 + }, + { + "epoch": 1.32, + "learning_rate": 1.561175637089676e-05, + "loss": 0.9922, + "step": 20550 + }, + { + "epoch": 1.32, + "learning_rate": 1.5611541681873808e-05, + "loss": 0.9039, + "step": 20551 + }, + { + "epoch": 1.32, + "learning_rate": 1.5611326992850857e-05, + "loss": 1.0389, + "step": 20552 + }, + { + "epoch": 1.32, + "learning_rate": 1.5611112303827906e-05, + "loss": 0.9649, + "step": 20553 + }, + { + "epoch": 1.32, + "learning_rate": 1.5610897614804955e-05, + "loss": 0.9681, + "step": 20554 + }, + { + "epoch": 1.32, + "learning_rate": 1.5610682925782007e-05, + "loss": 0.9046, + "step": 20555 + }, + { + "epoch": 1.32, + "learning_rate": 1.5610468236759056e-05, + "loss": 0.9819, + "step": 20556 + }, + { + "epoch": 1.32, + "learning_rate": 1.5610253547736105e-05, + "loss": 1.0378, + "step": 20557 + }, + { + "epoch": 1.32, + "learning_rate": 1.5610038858713154e-05, + "loss": 1.0041, + "step": 20558 + }, + { + "epoch": 1.32, + "learning_rate": 1.5609824169690203e-05, + "loss": 0.9364, + "step": 20559 + }, + { + "epoch": 1.32, + "learning_rate": 1.5609609480667255e-05, + "loss": 0.9088, + "step": 20560 + }, + { + "epoch": 1.32, + "learning_rate": 1.5609394791644304e-05, + "loss": 1.0193, + "step": 20561 + }, + { + "epoch": 1.32, + "learning_rate": 1.5609180102621357e-05, + "loss": 1.0236, + "step": 20562 + }, + { + "epoch": 1.32, + "learning_rate": 1.5608965413598406e-05, + "loss": 0.9833, + "step": 20563 + }, + { + "epoch": 1.32, + "learning_rate": 1.5608750724575455e-05, + "loss": 1.0007, + "step": 20564 + }, + { + "epoch": 1.32, + "learning_rate": 1.5608536035552504e-05, + "loss": 1.0294, + "step": 20565 + }, + { + "epoch": 1.32, + "learning_rate": 1.5608321346529553e-05, + "loss": 1.0053, + "step": 20566 + }, + { + "epoch": 1.32, + "learning_rate": 1.56081066575066e-05, + "loss": 0.9628, + "step": 20567 + }, + { + "epoch": 1.32, + "learning_rate": 1.5607891968483654e-05, + "loss": 0.992, + "step": 20568 + }, + { + "epoch": 1.32, + "learning_rate": 1.5607677279460703e-05, + "loss": 0.9332, + "step": 20569 + }, + { + "epoch": 1.32, + "learning_rate": 1.5607462590437752e-05, + "loss": 1.012, + "step": 20570 + }, + { + "epoch": 1.32, + "learning_rate": 1.56072479014148e-05, + "loss": 0.995, + "step": 20571 + }, + { + "epoch": 1.32, + "learning_rate": 1.560703321239185e-05, + "loss": 1.0018, + "step": 20572 + }, + { + "epoch": 1.32, + "learning_rate": 1.5606818523368902e-05, + "loss": 0.983, + "step": 20573 + }, + { + "epoch": 1.32, + "learning_rate": 1.560660383434595e-05, + "loss": 1.0333, + "step": 20574 + }, + { + "epoch": 1.32, + "learning_rate": 1.5606389145323e-05, + "loss": 0.9698, + "step": 20575 + }, + { + "epoch": 1.32, + "learning_rate": 1.5606174456300052e-05, + "loss": 0.9882, + "step": 20576 + }, + { + "epoch": 1.32, + "learning_rate": 1.56059597672771e-05, + "loss": 1.0052, + "step": 20577 + }, + { + "epoch": 1.32, + "learning_rate": 1.560574507825415e-05, + "loss": 1.0163, + "step": 20578 + }, + { + "epoch": 1.32, + "learning_rate": 1.56055303892312e-05, + "loss": 0.9432, + "step": 20579 + }, + { + "epoch": 1.32, + "learning_rate": 1.560531570020825e-05, + "loss": 0.9882, + "step": 20580 + }, + { + "epoch": 1.32, + "learning_rate": 1.56051010111853e-05, + "loss": 0.9342, + "step": 20581 + }, + { + "epoch": 1.32, + "learning_rate": 1.560488632216235e-05, + "loss": 0.9713, + "step": 20582 + }, + { + "epoch": 1.32, + "learning_rate": 1.56046716331394e-05, + "loss": 0.9578, + "step": 20583 + }, + { + "epoch": 1.32, + "learning_rate": 1.5604456944116447e-05, + "loss": 0.9827, + "step": 20584 + }, + { + "epoch": 1.32, + "learning_rate": 1.5604242255093496e-05, + "loss": 1.0234, + "step": 20585 + }, + { + "epoch": 1.32, + "learning_rate": 1.560402756607055e-05, + "loss": 0.956, + "step": 20586 + }, + { + "epoch": 1.32, + "learning_rate": 1.5603812877047598e-05, + "loss": 1.0151, + "step": 20587 + }, + { + "epoch": 1.32, + "learning_rate": 1.5603598188024647e-05, + "loss": 0.9681, + "step": 20588 + }, + { + "epoch": 1.32, + "learning_rate": 1.5603383499001696e-05, + "loss": 1.0245, + "step": 20589 + }, + { + "epoch": 1.32, + "learning_rate": 1.5603168809978748e-05, + "loss": 0.9715, + "step": 20590 + }, + { + "epoch": 1.32, + "learning_rate": 1.5602954120955797e-05, + "loss": 0.9464, + "step": 20591 + }, + { + "epoch": 1.32, + "learning_rate": 1.5602739431932846e-05, + "loss": 0.941, + "step": 20592 + }, + { + "epoch": 1.32, + "learning_rate": 1.56025247429099e-05, + "loss": 0.9942, + "step": 20593 + }, + { + "epoch": 1.32, + "learning_rate": 1.5602310053886947e-05, + "loss": 0.9277, + "step": 20594 + }, + { + "epoch": 1.33, + "learning_rate": 1.5602095364863996e-05, + "loss": 0.983, + "step": 20595 + }, + { + "epoch": 1.33, + "learning_rate": 1.5601880675841045e-05, + "loss": 1.0348, + "step": 20596 + }, + { + "epoch": 1.33, + "learning_rate": 1.5601665986818094e-05, + "loss": 1.0126, + "step": 20597 + }, + { + "epoch": 1.33, + "learning_rate": 1.5601451297795147e-05, + "loss": 0.9571, + "step": 20598 + }, + { + "epoch": 1.33, + "learning_rate": 1.5601236608772196e-05, + "loss": 0.9525, + "step": 20599 + }, + { + "epoch": 1.33, + "learning_rate": 1.5601021919749244e-05, + "loss": 0.9889, + "step": 20600 + }, + { + "epoch": 1.33, + "learning_rate": 1.5600807230726293e-05, + "loss": 0.9897, + "step": 20601 + }, + { + "epoch": 1.33, + "learning_rate": 1.5600592541703342e-05, + "loss": 0.9909, + "step": 20602 + }, + { + "epoch": 1.33, + "learning_rate": 1.560037785268039e-05, + "loss": 0.9619, + "step": 20603 + }, + { + "epoch": 1.33, + "learning_rate": 1.5600163163657444e-05, + "loss": 0.9457, + "step": 20604 + }, + { + "epoch": 1.33, + "learning_rate": 1.5599948474634493e-05, + "loss": 1.0443, + "step": 20605 + }, + { + "epoch": 1.33, + "learning_rate": 1.5599733785611545e-05, + "loss": 0.9669, + "step": 20606 + }, + { + "epoch": 1.33, + "learning_rate": 1.5599519096588594e-05, + "loss": 0.9608, + "step": 20607 + }, + { + "epoch": 1.33, + "learning_rate": 1.5599304407565643e-05, + "loss": 1.0376, + "step": 20608 + }, + { + "epoch": 1.33, + "learning_rate": 1.5599089718542692e-05, + "loss": 1.0417, + "step": 20609 + }, + { + "epoch": 1.33, + "learning_rate": 1.559887502951974e-05, + "loss": 0.9765, + "step": 20610 + }, + { + "epoch": 1.33, + "learning_rate": 1.5598660340496793e-05, + "loss": 0.9171, + "step": 20611 + }, + { + "epoch": 1.33, + "learning_rate": 1.5598445651473842e-05, + "loss": 0.9963, + "step": 20612 + }, + { + "epoch": 1.33, + "learning_rate": 1.559823096245089e-05, + "loss": 0.9454, + "step": 20613 + }, + { + "epoch": 1.33, + "learning_rate": 1.559801627342794e-05, + "loss": 1.0157, + "step": 20614 + }, + { + "epoch": 1.33, + "learning_rate": 1.559780158440499e-05, + "loss": 0.9233, + "step": 20615 + }, + { + "epoch": 1.33, + "learning_rate": 1.5597586895382038e-05, + "loss": 0.9515, + "step": 20616 + }, + { + "epoch": 1.33, + "learning_rate": 1.559737220635909e-05, + "loss": 1.0336, + "step": 20617 + }, + { + "epoch": 1.33, + "learning_rate": 1.559715751733614e-05, + "loss": 1.0047, + "step": 20618 + }, + { + "epoch": 1.33, + "learning_rate": 1.5596942828313192e-05, + "loss": 0.9412, + "step": 20619 + }, + { + "epoch": 1.33, + "learning_rate": 1.559672813929024e-05, + "loss": 0.917, + "step": 20620 + }, + { + "epoch": 1.33, + "learning_rate": 1.559651345026729e-05, + "loss": 0.9327, + "step": 20621 + }, + { + "epoch": 1.33, + "learning_rate": 1.559629876124434e-05, + "loss": 0.9919, + "step": 20622 + }, + { + "epoch": 1.33, + "learning_rate": 1.5596084072221388e-05, + "loss": 0.9455, + "step": 20623 + }, + { + "epoch": 1.33, + "learning_rate": 1.559586938319844e-05, + "loss": 0.9494, + "step": 20624 + }, + { + "epoch": 1.33, + "learning_rate": 1.559565469417549e-05, + "loss": 0.9692, + "step": 20625 + }, + { + "epoch": 1.33, + "learning_rate": 1.5595440005152538e-05, + "loss": 0.9664, + "step": 20626 + }, + { + "epoch": 1.33, + "learning_rate": 1.5595225316129587e-05, + "loss": 0.9967, + "step": 20627 + }, + { + "epoch": 1.33, + "learning_rate": 1.5595010627106636e-05, + "loss": 0.9842, + "step": 20628 + }, + { + "epoch": 1.33, + "learning_rate": 1.5594795938083688e-05, + "loss": 1.0145, + "step": 20629 + }, + { + "epoch": 1.33, + "learning_rate": 1.5594581249060737e-05, + "loss": 1.0453, + "step": 20630 + }, + { + "epoch": 1.33, + "learning_rate": 1.5594366560037786e-05, + "loss": 0.9353, + "step": 20631 + }, + { + "epoch": 1.33, + "learning_rate": 1.5594151871014835e-05, + "loss": 1.0495, + "step": 20632 + }, + { + "epoch": 1.33, + "learning_rate": 1.5593937181991884e-05, + "loss": 1.0398, + "step": 20633 + }, + { + "epoch": 1.33, + "learning_rate": 1.5593722492968936e-05, + "loss": 0.8902, + "step": 20634 + }, + { + "epoch": 1.33, + "learning_rate": 1.5593507803945985e-05, + "loss": 0.9559, + "step": 20635 + }, + { + "epoch": 1.33, + "learning_rate": 1.5593293114923034e-05, + "loss": 1.0195, + "step": 20636 + }, + { + "epoch": 1.33, + "learning_rate": 1.5593078425900087e-05, + "loss": 0.9995, + "step": 20637 + }, + { + "epoch": 1.33, + "learning_rate": 1.5592863736877136e-05, + "loss": 0.9427, + "step": 20638 + }, + { + "epoch": 1.33, + "learning_rate": 1.5592649047854185e-05, + "loss": 0.9263, + "step": 20639 + }, + { + "epoch": 1.33, + "learning_rate": 1.5592434358831234e-05, + "loss": 0.9658, + "step": 20640 + }, + { + "epoch": 1.33, + "learning_rate": 1.5592219669808283e-05, + "loss": 0.9272, + "step": 20641 + }, + { + "epoch": 1.33, + "learning_rate": 1.5592004980785335e-05, + "loss": 1.0276, + "step": 20642 + }, + { + "epoch": 1.33, + "learning_rate": 1.5591790291762384e-05, + "loss": 0.9455, + "step": 20643 + }, + { + "epoch": 1.33, + "learning_rate": 1.5591575602739433e-05, + "loss": 1.0721, + "step": 20644 + }, + { + "epoch": 1.33, + "learning_rate": 1.5591360913716482e-05, + "loss": 1.0401, + "step": 20645 + }, + { + "epoch": 1.33, + "learning_rate": 1.559114622469353e-05, + "loss": 0.9754, + "step": 20646 + }, + { + "epoch": 1.33, + "learning_rate": 1.5590931535670583e-05, + "loss": 1.0424, + "step": 20647 + }, + { + "epoch": 1.33, + "learning_rate": 1.5590716846647632e-05, + "loss": 0.9806, + "step": 20648 + }, + { + "epoch": 1.33, + "learning_rate": 1.5590502157624684e-05, + "loss": 0.9456, + "step": 20649 + }, + { + "epoch": 1.33, + "learning_rate": 1.5590287468601733e-05, + "loss": 0.9901, + "step": 20650 + }, + { + "epoch": 1.33, + "learning_rate": 1.5590072779578782e-05, + "loss": 0.9367, + "step": 20651 + }, + { + "epoch": 1.33, + "learning_rate": 1.558985809055583e-05, + "loss": 0.9858, + "step": 20652 + }, + { + "epoch": 1.33, + "learning_rate": 1.558964340153288e-05, + "loss": 0.9374, + "step": 20653 + }, + { + "epoch": 1.33, + "learning_rate": 1.558942871250993e-05, + "loss": 0.9812, + "step": 20654 + }, + { + "epoch": 1.33, + "learning_rate": 1.558921402348698e-05, + "loss": 0.9799, + "step": 20655 + }, + { + "epoch": 1.33, + "learning_rate": 1.558899933446403e-05, + "loss": 0.9921, + "step": 20656 + }, + { + "epoch": 1.33, + "learning_rate": 1.558878464544108e-05, + "loss": 0.9318, + "step": 20657 + }, + { + "epoch": 1.33, + "learning_rate": 1.558856995641813e-05, + "loss": 0.9085, + "step": 20658 + }, + { + "epoch": 1.33, + "learning_rate": 1.5588355267395177e-05, + "loss": 0.9476, + "step": 20659 + }, + { + "epoch": 1.33, + "learning_rate": 1.558814057837223e-05, + "loss": 0.979, + "step": 20660 + }, + { + "epoch": 1.33, + "learning_rate": 1.558792588934928e-05, + "loss": 0.9818, + "step": 20661 + }, + { + "epoch": 1.33, + "learning_rate": 1.5587711200326328e-05, + "loss": 0.9655, + "step": 20662 + }, + { + "epoch": 1.33, + "learning_rate": 1.558749651130338e-05, + "loss": 0.9938, + "step": 20663 + }, + { + "epoch": 1.33, + "learning_rate": 1.558728182228043e-05, + "loss": 0.94, + "step": 20664 + }, + { + "epoch": 1.33, + "learning_rate": 1.5587067133257478e-05, + "loss": 1.0509, + "step": 20665 + }, + { + "epoch": 1.33, + "learning_rate": 1.5586852444234527e-05, + "loss": 0.872, + "step": 20666 + }, + { + "epoch": 1.33, + "learning_rate": 1.558663775521158e-05, + "loss": 0.9623, + "step": 20667 + }, + { + "epoch": 1.33, + "learning_rate": 1.5586423066188628e-05, + "loss": 0.8723, + "step": 20668 + }, + { + "epoch": 1.33, + "learning_rate": 1.5586208377165677e-05, + "loss": 0.9276, + "step": 20669 + }, + { + "epoch": 1.33, + "learning_rate": 1.5585993688142726e-05, + "loss": 0.8791, + "step": 20670 + }, + { + "epoch": 1.33, + "learning_rate": 1.5585778999119775e-05, + "loss": 0.9253, + "step": 20671 + }, + { + "epoch": 1.33, + "learning_rate": 1.5585564310096824e-05, + "loss": 0.9674, + "step": 20672 + }, + { + "epoch": 1.33, + "learning_rate": 1.5585349621073877e-05, + "loss": 1.0733, + "step": 20673 + }, + { + "epoch": 1.33, + "learning_rate": 1.5585134932050925e-05, + "loss": 1.0162, + "step": 20674 + }, + { + "epoch": 1.33, + "learning_rate": 1.5584920243027974e-05, + "loss": 1.0531, + "step": 20675 + }, + { + "epoch": 1.33, + "learning_rate": 1.5584705554005023e-05, + "loss": 1.0167, + "step": 20676 + }, + { + "epoch": 1.33, + "learning_rate": 1.5584490864982072e-05, + "loss": 1.0055, + "step": 20677 + }, + { + "epoch": 1.33, + "learning_rate": 1.5584276175959125e-05, + "loss": 1.0554, + "step": 20678 + }, + { + "epoch": 1.33, + "learning_rate": 1.5584061486936174e-05, + "loss": 1.0091, + "step": 20679 + }, + { + "epoch": 1.33, + "learning_rate": 1.5583846797913226e-05, + "loss": 0.9568, + "step": 20680 + }, + { + "epoch": 1.33, + "learning_rate": 1.5583632108890275e-05, + "loss": 0.9211, + "step": 20681 + }, + { + "epoch": 1.33, + "learning_rate": 1.5583417419867324e-05, + "loss": 0.9788, + "step": 20682 + }, + { + "epoch": 1.33, + "learning_rate": 1.5583202730844373e-05, + "loss": 0.9474, + "step": 20683 + }, + { + "epoch": 1.33, + "learning_rate": 1.5582988041821422e-05, + "loss": 0.9597, + "step": 20684 + }, + { + "epoch": 1.33, + "learning_rate": 1.5582773352798474e-05, + "loss": 0.9599, + "step": 20685 + }, + { + "epoch": 1.33, + "learning_rate": 1.5582558663775523e-05, + "loss": 0.9805, + "step": 20686 + }, + { + "epoch": 1.33, + "learning_rate": 1.5582343974752572e-05, + "loss": 1.0184, + "step": 20687 + }, + { + "epoch": 1.33, + "learning_rate": 1.558212928572962e-05, + "loss": 0.9733, + "step": 20688 + }, + { + "epoch": 1.33, + "learning_rate": 1.558191459670667e-05, + "loss": 0.9781, + "step": 20689 + }, + { + "epoch": 1.33, + "learning_rate": 1.558169990768372e-05, + "loss": 1.0018, + "step": 20690 + }, + { + "epoch": 1.33, + "learning_rate": 1.558148521866077e-05, + "loss": 0.964, + "step": 20691 + }, + { + "epoch": 1.33, + "learning_rate": 1.558127052963782e-05, + "loss": 1.0252, + "step": 20692 + }, + { + "epoch": 1.33, + "learning_rate": 1.5581055840614873e-05, + "loss": 0.9546, + "step": 20693 + }, + { + "epoch": 1.33, + "learning_rate": 1.5580841151591922e-05, + "loss": 0.9869, + "step": 20694 + }, + { + "epoch": 1.33, + "learning_rate": 1.558062646256897e-05, + "loss": 0.9399, + "step": 20695 + }, + { + "epoch": 1.33, + "learning_rate": 1.558041177354602e-05, + "loss": 0.9682, + "step": 20696 + }, + { + "epoch": 1.33, + "learning_rate": 1.558019708452307e-05, + "loss": 1.0202, + "step": 20697 + }, + { + "epoch": 1.33, + "learning_rate": 1.557998239550012e-05, + "loss": 0.9251, + "step": 20698 + }, + { + "epoch": 1.33, + "learning_rate": 1.557976770647717e-05, + "loss": 1.0003, + "step": 20699 + }, + { + "epoch": 1.33, + "learning_rate": 1.557955301745422e-05, + "loss": 0.9126, + "step": 20700 + }, + { + "epoch": 1.33, + "learning_rate": 1.5579338328431268e-05, + "loss": 0.9708, + "step": 20701 + }, + { + "epoch": 1.33, + "learning_rate": 1.5579123639408317e-05, + "loss": 1.0552, + "step": 20702 + }, + { + "epoch": 1.33, + "learning_rate": 1.5578908950385366e-05, + "loss": 0.9548, + "step": 20703 + }, + { + "epoch": 1.33, + "learning_rate": 1.5578694261362418e-05, + "loss": 0.9346, + "step": 20704 + }, + { + "epoch": 1.33, + "learning_rate": 1.5578479572339467e-05, + "loss": 0.9351, + "step": 20705 + }, + { + "epoch": 1.33, + "learning_rate": 1.5578264883316516e-05, + "loss": 0.984, + "step": 20706 + }, + { + "epoch": 1.33, + "learning_rate": 1.557805019429357e-05, + "loss": 0.964, + "step": 20707 + }, + { + "epoch": 1.33, + "learning_rate": 1.5577835505270617e-05, + "loss": 0.9175, + "step": 20708 + }, + { + "epoch": 1.33, + "learning_rate": 1.5577620816247666e-05, + "loss": 0.9969, + "step": 20709 + }, + { + "epoch": 1.33, + "learning_rate": 1.5577406127224715e-05, + "loss": 0.958, + "step": 20710 + }, + { + "epoch": 1.33, + "learning_rate": 1.5577191438201768e-05, + "loss": 0.9668, + "step": 20711 + }, + { + "epoch": 1.33, + "learning_rate": 1.5576976749178817e-05, + "loss": 0.96, + "step": 20712 + }, + { + "epoch": 1.33, + "learning_rate": 1.5576762060155866e-05, + "loss": 1.0164, + "step": 20713 + }, + { + "epoch": 1.33, + "learning_rate": 1.5576547371132915e-05, + "loss": 0.9496, + "step": 20714 + }, + { + "epoch": 1.33, + "learning_rate": 1.5576332682109964e-05, + "loss": 0.9709, + "step": 20715 + }, + { + "epoch": 1.33, + "learning_rate": 1.5576117993087016e-05, + "loss": 0.8927, + "step": 20716 + }, + { + "epoch": 1.33, + "learning_rate": 1.5575903304064065e-05, + "loss": 0.9479, + "step": 20717 + }, + { + "epoch": 1.33, + "learning_rate": 1.5575688615041114e-05, + "loss": 0.9528, + "step": 20718 + }, + { + "epoch": 1.33, + "learning_rate": 1.5575473926018163e-05, + "loss": 0.9725, + "step": 20719 + }, + { + "epoch": 1.33, + "learning_rate": 1.5575259236995212e-05, + "loss": 1.0466, + "step": 20720 + }, + { + "epoch": 1.33, + "learning_rate": 1.5575044547972264e-05, + "loss": 1.0381, + "step": 20721 + }, + { + "epoch": 1.33, + "learning_rate": 1.5574829858949313e-05, + "loss": 1.0734, + "step": 20722 + }, + { + "epoch": 1.33, + "learning_rate": 1.5574615169926362e-05, + "loss": 1.0549, + "step": 20723 + }, + { + "epoch": 1.33, + "learning_rate": 1.5574400480903414e-05, + "loss": 0.9782, + "step": 20724 + }, + { + "epoch": 1.33, + "learning_rate": 1.5574185791880463e-05, + "loss": 0.9364, + "step": 20725 + }, + { + "epoch": 1.33, + "learning_rate": 1.5573971102857512e-05, + "loss": 0.9644, + "step": 20726 + }, + { + "epoch": 1.33, + "learning_rate": 1.557375641383456e-05, + "loss": 0.9988, + "step": 20727 + }, + { + "epoch": 1.33, + "learning_rate": 1.557354172481161e-05, + "loss": 0.9924, + "step": 20728 + }, + { + "epoch": 1.33, + "learning_rate": 1.5573327035788663e-05, + "loss": 0.9024, + "step": 20729 + }, + { + "epoch": 1.33, + "learning_rate": 1.557311234676571e-05, + "loss": 0.9818, + "step": 20730 + }, + { + "epoch": 1.33, + "learning_rate": 1.557289765774276e-05, + "loss": 1.0271, + "step": 20731 + }, + { + "epoch": 1.33, + "learning_rate": 1.557268296871981e-05, + "loss": 1.0919, + "step": 20732 + }, + { + "epoch": 1.33, + "learning_rate": 1.557246827969686e-05, + "loss": 1.049, + "step": 20733 + }, + { + "epoch": 1.33, + "learning_rate": 1.557225359067391e-05, + "loss": 0.9794, + "step": 20734 + }, + { + "epoch": 1.33, + "learning_rate": 1.557203890165096e-05, + "loss": 0.9706, + "step": 20735 + }, + { + "epoch": 1.33, + "learning_rate": 1.5571824212628012e-05, + "loss": 1.0399, + "step": 20736 + }, + { + "epoch": 1.33, + "learning_rate": 1.557160952360506e-05, + "loss": 0.9072, + "step": 20737 + }, + { + "epoch": 1.33, + "learning_rate": 1.557139483458211e-05, + "loss": 0.9816, + "step": 20738 + }, + { + "epoch": 1.33, + "learning_rate": 1.557118014555916e-05, + "loss": 1.0257, + "step": 20739 + }, + { + "epoch": 1.33, + "learning_rate": 1.5570965456536208e-05, + "loss": 1.0048, + "step": 20740 + }, + { + "epoch": 1.33, + "learning_rate": 1.5570750767513257e-05, + "loss": 0.9881, + "step": 20741 + }, + { + "epoch": 1.33, + "learning_rate": 1.557053607849031e-05, + "loss": 0.9769, + "step": 20742 + }, + { + "epoch": 1.33, + "learning_rate": 1.5570321389467358e-05, + "loss": 1.046, + "step": 20743 + }, + { + "epoch": 1.33, + "learning_rate": 1.5570106700444407e-05, + "loss": 0.9766, + "step": 20744 + }, + { + "epoch": 1.33, + "learning_rate": 1.5569892011421456e-05, + "loss": 0.9302, + "step": 20745 + }, + { + "epoch": 1.33, + "learning_rate": 1.5569677322398505e-05, + "loss": 1.0155, + "step": 20746 + }, + { + "epoch": 1.33, + "learning_rate": 1.5569462633375558e-05, + "loss": 0.9381, + "step": 20747 + }, + { + "epoch": 1.33, + "learning_rate": 1.5569247944352606e-05, + "loss": 0.9643, + "step": 20748 + }, + { + "epoch": 1.33, + "learning_rate": 1.5569033255329655e-05, + "loss": 0.9824, + "step": 20749 + }, + { + "epoch": 1.34, + "learning_rate": 1.5568818566306708e-05, + "loss": 0.9689, + "step": 20750 + }, + { + "epoch": 1.34, + "learning_rate": 1.5568603877283757e-05, + "loss": 0.9956, + "step": 20751 + }, + { + "epoch": 1.34, + "learning_rate": 1.5568389188260806e-05, + "loss": 0.9305, + "step": 20752 + }, + { + "epoch": 1.34, + "learning_rate": 1.5568174499237855e-05, + "loss": 1.0257, + "step": 20753 + }, + { + "epoch": 1.34, + "learning_rate": 1.5567959810214907e-05, + "loss": 1.0027, + "step": 20754 + }, + { + "epoch": 1.34, + "learning_rate": 1.5567745121191956e-05, + "loss": 1.0693, + "step": 20755 + }, + { + "epoch": 1.34, + "learning_rate": 1.5567530432169005e-05, + "loss": 1.0484, + "step": 20756 + }, + { + "epoch": 1.34, + "learning_rate": 1.5567315743146054e-05, + "loss": 0.9548, + "step": 20757 + }, + { + "epoch": 1.34, + "learning_rate": 1.5567101054123103e-05, + "loss": 0.9778, + "step": 20758 + }, + { + "epoch": 1.34, + "learning_rate": 1.5566886365100152e-05, + "loss": 0.9765, + "step": 20759 + }, + { + "epoch": 1.34, + "learning_rate": 1.5566671676077204e-05, + "loss": 1.0158, + "step": 20760 + }, + { + "epoch": 1.34, + "learning_rate": 1.5566456987054253e-05, + "loss": 0.9769, + "step": 20761 + }, + { + "epoch": 1.34, + "learning_rate": 1.5566242298031302e-05, + "loss": 0.9468, + "step": 20762 + }, + { + "epoch": 1.34, + "learning_rate": 1.556602760900835e-05, + "loss": 1.0095, + "step": 20763 + }, + { + "epoch": 1.34, + "learning_rate": 1.55658129199854e-05, + "loss": 1.0375, + "step": 20764 + }, + { + "epoch": 1.34, + "learning_rate": 1.5565598230962452e-05, + "loss": 0.9953, + "step": 20765 + }, + { + "epoch": 1.34, + "learning_rate": 1.55653835419395e-05, + "loss": 0.915, + "step": 20766 + }, + { + "epoch": 1.34, + "learning_rate": 1.5565168852916554e-05, + "loss": 0.9601, + "step": 20767 + }, + { + "epoch": 1.34, + "learning_rate": 1.5564954163893603e-05, + "loss": 0.9811, + "step": 20768 + }, + { + "epoch": 1.34, + "learning_rate": 1.556473947487065e-05, + "loss": 0.8933, + "step": 20769 + }, + { + "epoch": 1.34, + "learning_rate": 1.55645247858477e-05, + "loss": 0.9626, + "step": 20770 + }, + { + "epoch": 1.34, + "learning_rate": 1.556431009682475e-05, + "loss": 0.9797, + "step": 20771 + }, + { + "epoch": 1.34, + "learning_rate": 1.5564095407801802e-05, + "loss": 0.9419, + "step": 20772 + }, + { + "epoch": 1.34, + "learning_rate": 1.556388071877885e-05, + "loss": 0.8785, + "step": 20773 + }, + { + "epoch": 1.34, + "learning_rate": 1.55636660297559e-05, + "loss": 0.9275, + "step": 20774 + }, + { + "epoch": 1.34, + "learning_rate": 1.556345134073295e-05, + "loss": 0.9281, + "step": 20775 + }, + { + "epoch": 1.34, + "learning_rate": 1.5563236651709998e-05, + "loss": 0.9902, + "step": 20776 + }, + { + "epoch": 1.34, + "learning_rate": 1.5563021962687047e-05, + "loss": 0.9949, + "step": 20777 + }, + { + "epoch": 1.34, + "learning_rate": 1.55628072736641e-05, + "loss": 0.9465, + "step": 20778 + }, + { + "epoch": 1.34, + "learning_rate": 1.5562592584641148e-05, + "loss": 0.9444, + "step": 20779 + }, + { + "epoch": 1.34, + "learning_rate": 1.55623778956182e-05, + "loss": 0.9778, + "step": 20780 + }, + { + "epoch": 1.34, + "learning_rate": 1.556216320659525e-05, + "loss": 1.0003, + "step": 20781 + }, + { + "epoch": 1.34, + "learning_rate": 1.55619485175723e-05, + "loss": 0.9354, + "step": 20782 + }, + { + "epoch": 1.34, + "learning_rate": 1.5561733828549347e-05, + "loss": 0.9408, + "step": 20783 + }, + { + "epoch": 1.34, + "learning_rate": 1.5561519139526396e-05, + "loss": 0.9802, + "step": 20784 + }, + { + "epoch": 1.34, + "learning_rate": 1.556130445050345e-05, + "loss": 0.9504, + "step": 20785 + }, + { + "epoch": 1.34, + "learning_rate": 1.5561089761480498e-05, + "loss": 0.9787, + "step": 20786 + }, + { + "epoch": 1.34, + "learning_rate": 1.5560875072457547e-05, + "loss": 0.9968, + "step": 20787 + }, + { + "epoch": 1.34, + "learning_rate": 1.5560660383434596e-05, + "loss": 0.9922, + "step": 20788 + }, + { + "epoch": 1.34, + "learning_rate": 1.5560445694411645e-05, + "loss": 0.97, + "step": 20789 + }, + { + "epoch": 1.34, + "learning_rate": 1.5560231005388694e-05, + "loss": 1.0058, + "step": 20790 + }, + { + "epoch": 1.34, + "learning_rate": 1.5560016316365746e-05, + "loss": 1.0651, + "step": 20791 + }, + { + "epoch": 1.34, + "learning_rate": 1.5559801627342795e-05, + "loss": 1.0027, + "step": 20792 + }, + { + "epoch": 1.34, + "learning_rate": 1.5559586938319844e-05, + "loss": 1.0164, + "step": 20793 + }, + { + "epoch": 1.34, + "learning_rate": 1.5559372249296896e-05, + "loss": 0.9967, + "step": 20794 + }, + { + "epoch": 1.34, + "learning_rate": 1.5559157560273945e-05, + "loss": 0.9932, + "step": 20795 + }, + { + "epoch": 1.34, + "learning_rate": 1.5558942871250994e-05, + "loss": 0.9833, + "step": 20796 + }, + { + "epoch": 1.34, + "learning_rate": 1.5558728182228043e-05, + "loss": 0.9228, + "step": 20797 + }, + { + "epoch": 1.34, + "learning_rate": 1.5558513493205095e-05, + "loss": 0.9899, + "step": 20798 + }, + { + "epoch": 1.34, + "learning_rate": 1.5558298804182144e-05, + "loss": 1.0142, + "step": 20799 + }, + { + "epoch": 1.34, + "learning_rate": 1.5558084115159193e-05, + "loss": 1.0232, + "step": 20800 + }, + { + "epoch": 1.34, + "learning_rate": 1.5557869426136242e-05, + "loss": 0.9768, + "step": 20801 + }, + { + "epoch": 1.34, + "learning_rate": 1.555765473711329e-05, + "loss": 0.9448, + "step": 20802 + }, + { + "epoch": 1.34, + "learning_rate": 1.5557440048090344e-05, + "loss": 0.9641, + "step": 20803 + }, + { + "epoch": 1.34, + "learning_rate": 1.5557225359067393e-05, + "loss": 1.0553, + "step": 20804 + }, + { + "epoch": 1.34, + "learning_rate": 1.555701067004444e-05, + "loss": 1.0113, + "step": 20805 + }, + { + "epoch": 1.34, + "learning_rate": 1.555679598102149e-05, + "loss": 0.9001, + "step": 20806 + }, + { + "epoch": 1.34, + "learning_rate": 1.555658129199854e-05, + "loss": 0.9693, + "step": 20807 + }, + { + "epoch": 1.34, + "learning_rate": 1.5556366602975592e-05, + "loss": 0.9762, + "step": 20808 + }, + { + "epoch": 1.34, + "learning_rate": 1.555615191395264e-05, + "loss": 1.0102, + "step": 20809 + }, + { + "epoch": 1.34, + "learning_rate": 1.555593722492969e-05, + "loss": 0.9766, + "step": 20810 + }, + { + "epoch": 1.34, + "learning_rate": 1.5555722535906742e-05, + "loss": 0.9024, + "step": 20811 + }, + { + "epoch": 1.34, + "learning_rate": 1.555550784688379e-05, + "loss": 1.0006, + "step": 20812 + }, + { + "epoch": 1.34, + "learning_rate": 1.555529315786084e-05, + "loss": 0.9224, + "step": 20813 + }, + { + "epoch": 1.34, + "learning_rate": 1.555507846883789e-05, + "loss": 0.9932, + "step": 20814 + }, + { + "epoch": 1.34, + "learning_rate": 1.5554863779814938e-05, + "loss": 0.9331, + "step": 20815 + }, + { + "epoch": 1.34, + "learning_rate": 1.555464909079199e-05, + "loss": 0.9631, + "step": 20816 + }, + { + "epoch": 1.34, + "learning_rate": 1.555443440176904e-05, + "loss": 1.0065, + "step": 20817 + }, + { + "epoch": 1.34, + "learning_rate": 1.5554219712746088e-05, + "loss": 0.8936, + "step": 20818 + }, + { + "epoch": 1.34, + "learning_rate": 1.5554005023723137e-05, + "loss": 0.9513, + "step": 20819 + }, + { + "epoch": 1.34, + "learning_rate": 1.5553790334700186e-05, + "loss": 0.9057, + "step": 20820 + }, + { + "epoch": 1.34, + "learning_rate": 1.555357564567724e-05, + "loss": 0.9399, + "step": 20821 + }, + { + "epoch": 1.34, + "learning_rate": 1.5553360956654287e-05, + "loss": 1.0101, + "step": 20822 + }, + { + "epoch": 1.34, + "learning_rate": 1.555314626763134e-05, + "loss": 0.9673, + "step": 20823 + }, + { + "epoch": 1.34, + "learning_rate": 1.555293157860839e-05, + "loss": 1.0178, + "step": 20824 + }, + { + "epoch": 1.34, + "learning_rate": 1.5552716889585438e-05, + "loss": 0.9908, + "step": 20825 + }, + { + "epoch": 1.34, + "learning_rate": 1.5552502200562487e-05, + "loss": 1.002, + "step": 20826 + }, + { + "epoch": 1.34, + "learning_rate": 1.5552287511539536e-05, + "loss": 1.0329, + "step": 20827 + }, + { + "epoch": 1.34, + "learning_rate": 1.5552072822516585e-05, + "loss": 0.9574, + "step": 20828 + }, + { + "epoch": 1.34, + "learning_rate": 1.5551858133493637e-05, + "loss": 1.0367, + "step": 20829 + }, + { + "epoch": 1.34, + "learning_rate": 1.5551643444470686e-05, + "loss": 0.9438, + "step": 20830 + }, + { + "epoch": 1.34, + "learning_rate": 1.5551428755447735e-05, + "loss": 1.0009, + "step": 20831 + }, + { + "epoch": 1.34, + "learning_rate": 1.5551214066424784e-05, + "loss": 1.0678, + "step": 20832 + }, + { + "epoch": 1.34, + "learning_rate": 1.5550999377401833e-05, + "loss": 0.9845, + "step": 20833 + }, + { + "epoch": 1.34, + "learning_rate": 1.5550784688378885e-05, + "loss": 0.9914, + "step": 20834 + }, + { + "epoch": 1.34, + "learning_rate": 1.5550569999355934e-05, + "loss": 0.9515, + "step": 20835 + }, + { + "epoch": 1.34, + "learning_rate": 1.5550355310332983e-05, + "loss": 1.0199, + "step": 20836 + }, + { + "epoch": 1.34, + "learning_rate": 1.5550140621310036e-05, + "loss": 0.9699, + "step": 20837 + }, + { + "epoch": 1.34, + "learning_rate": 1.5549925932287084e-05, + "loss": 0.952, + "step": 20838 + }, + { + "epoch": 1.34, + "learning_rate": 1.5549711243264133e-05, + "loss": 1.0296, + "step": 20839 + }, + { + "epoch": 1.34, + "learning_rate": 1.5549496554241182e-05, + "loss": 0.9815, + "step": 20840 + }, + { + "epoch": 1.34, + "learning_rate": 1.5549281865218235e-05, + "loss": 0.9353, + "step": 20841 + }, + { + "epoch": 1.34, + "learning_rate": 1.5549067176195284e-05, + "loss": 0.9824, + "step": 20842 + }, + { + "epoch": 1.34, + "learning_rate": 1.5548852487172333e-05, + "loss": 0.9313, + "step": 20843 + }, + { + "epoch": 1.34, + "learning_rate": 1.554863779814938e-05, + "loss": 1.0154, + "step": 20844 + }, + { + "epoch": 1.34, + "learning_rate": 1.554842310912643e-05, + "loss": 1.0702, + "step": 20845 + }, + { + "epoch": 1.34, + "learning_rate": 1.554820842010348e-05, + "loss": 1.0445, + "step": 20846 + }, + { + "epoch": 1.34, + "learning_rate": 1.5547993731080532e-05, + "loss": 0.8964, + "step": 20847 + }, + { + "epoch": 1.34, + "learning_rate": 1.554777904205758e-05, + "loss": 0.967, + "step": 20848 + }, + { + "epoch": 1.34, + "learning_rate": 1.554756435303463e-05, + "loss": 1.0115, + "step": 20849 + }, + { + "epoch": 1.34, + "learning_rate": 1.554734966401168e-05, + "loss": 0.9672, + "step": 20850 + }, + { + "epoch": 1.34, + "learning_rate": 1.5547134974988728e-05, + "loss": 0.9855, + "step": 20851 + }, + { + "epoch": 1.34, + "learning_rate": 1.554692028596578e-05, + "loss": 1.0274, + "step": 20852 + }, + { + "epoch": 1.34, + "learning_rate": 1.554670559694283e-05, + "loss": 0.9899, + "step": 20853 + }, + { + "epoch": 1.34, + "learning_rate": 1.554649090791988e-05, + "loss": 0.945, + "step": 20854 + }, + { + "epoch": 1.34, + "learning_rate": 1.554627621889693e-05, + "loss": 0.9562, + "step": 20855 + }, + { + "epoch": 1.34, + "learning_rate": 1.554606152987398e-05, + "loss": 1.0655, + "step": 20856 + }, + { + "epoch": 1.34, + "learning_rate": 1.554584684085103e-05, + "loss": 0.9533, + "step": 20857 + }, + { + "epoch": 1.34, + "learning_rate": 1.5545632151828077e-05, + "loss": 0.9592, + "step": 20858 + }, + { + "epoch": 1.34, + "learning_rate": 1.5545417462805126e-05, + "loss": 1.057, + "step": 20859 + }, + { + "epoch": 1.34, + "learning_rate": 1.554520277378218e-05, + "loss": 0.9749, + "step": 20860 + }, + { + "epoch": 1.34, + "learning_rate": 1.5544988084759228e-05, + "loss": 1.0069, + "step": 20861 + }, + { + "epoch": 1.34, + "learning_rate": 1.5544773395736277e-05, + "loss": 0.9536, + "step": 20862 + }, + { + "epoch": 1.34, + "learning_rate": 1.5544558706713326e-05, + "loss": 1.1385, + "step": 20863 + }, + { + "epoch": 1.34, + "learning_rate": 1.5544344017690375e-05, + "loss": 1.0222, + "step": 20864 + }, + { + "epoch": 1.34, + "learning_rate": 1.5544129328667427e-05, + "loss": 0.9668, + "step": 20865 + }, + { + "epoch": 1.34, + "learning_rate": 1.5543914639644476e-05, + "loss": 0.9904, + "step": 20866 + }, + { + "epoch": 1.34, + "learning_rate": 1.5543699950621528e-05, + "loss": 0.8872, + "step": 20867 + }, + { + "epoch": 1.34, + "learning_rate": 1.5543485261598577e-05, + "loss": 0.909, + "step": 20868 + }, + { + "epoch": 1.34, + "learning_rate": 1.5543270572575626e-05, + "loss": 0.9907, + "step": 20869 + }, + { + "epoch": 1.34, + "learning_rate": 1.5543055883552675e-05, + "loss": 1.0352, + "step": 20870 + }, + { + "epoch": 1.34, + "learning_rate": 1.5542841194529724e-05, + "loss": 1.0028, + "step": 20871 + }, + { + "epoch": 1.34, + "learning_rate": 1.5542626505506776e-05, + "loss": 1.0084, + "step": 20872 + }, + { + "epoch": 1.34, + "learning_rate": 1.5542411816483825e-05, + "loss": 0.8927, + "step": 20873 + }, + { + "epoch": 1.34, + "learning_rate": 1.5542197127460874e-05, + "loss": 0.9892, + "step": 20874 + }, + { + "epoch": 1.34, + "learning_rate": 1.5541982438437923e-05, + "loss": 0.9019, + "step": 20875 + }, + { + "epoch": 1.34, + "learning_rate": 1.5541767749414972e-05, + "loss": 1.001, + "step": 20876 + }, + { + "epoch": 1.34, + "learning_rate": 1.554155306039202e-05, + "loss": 0.9575, + "step": 20877 + }, + { + "epoch": 1.34, + "learning_rate": 1.5541338371369074e-05, + "loss": 0.9153, + "step": 20878 + }, + { + "epoch": 1.34, + "learning_rate": 1.5541123682346123e-05, + "loss": 0.8902, + "step": 20879 + }, + { + "epoch": 1.34, + "learning_rate": 1.554090899332317e-05, + "loss": 1.0068, + "step": 20880 + }, + { + "epoch": 1.34, + "learning_rate": 1.5540694304300224e-05, + "loss": 0.9719, + "step": 20881 + }, + { + "epoch": 1.34, + "learning_rate": 1.5540479615277273e-05, + "loss": 1.0098, + "step": 20882 + }, + { + "epoch": 1.34, + "learning_rate": 1.5540264926254322e-05, + "loss": 0.9515, + "step": 20883 + }, + { + "epoch": 1.34, + "learning_rate": 1.554005023723137e-05, + "loss": 0.9977, + "step": 20884 + }, + { + "epoch": 1.34, + "learning_rate": 1.5539835548208423e-05, + "loss": 0.9178, + "step": 20885 + }, + { + "epoch": 1.34, + "learning_rate": 1.5539620859185472e-05, + "loss": 0.9849, + "step": 20886 + }, + { + "epoch": 1.34, + "learning_rate": 1.553940617016252e-05, + "loss": 1.0532, + "step": 20887 + }, + { + "epoch": 1.34, + "learning_rate": 1.553919148113957e-05, + "loss": 1.0222, + "step": 20888 + }, + { + "epoch": 1.34, + "learning_rate": 1.553897679211662e-05, + "loss": 0.974, + "step": 20889 + }, + { + "epoch": 1.34, + "learning_rate": 1.553876210309367e-05, + "loss": 0.9477, + "step": 20890 + }, + { + "epoch": 1.34, + "learning_rate": 1.553854741407072e-05, + "loss": 1.0595, + "step": 20891 + }, + { + "epoch": 1.34, + "learning_rate": 1.553833272504777e-05, + "loss": 0.9196, + "step": 20892 + }, + { + "epoch": 1.34, + "learning_rate": 1.5538118036024818e-05, + "loss": 0.9807, + "step": 20893 + }, + { + "epoch": 1.34, + "learning_rate": 1.5537903347001867e-05, + "loss": 0.9797, + "step": 20894 + }, + { + "epoch": 1.34, + "learning_rate": 1.553768865797892e-05, + "loss": 0.9894, + "step": 20895 + }, + { + "epoch": 1.34, + "learning_rate": 1.553747396895597e-05, + "loss": 1.0219, + "step": 20896 + }, + { + "epoch": 1.34, + "learning_rate": 1.5537259279933017e-05, + "loss": 1.0046, + "step": 20897 + }, + { + "epoch": 1.34, + "learning_rate": 1.553704459091007e-05, + "loss": 1.0184, + "step": 20898 + }, + { + "epoch": 1.34, + "learning_rate": 1.553682990188712e-05, + "loss": 1.0259, + "step": 20899 + }, + { + "epoch": 1.34, + "learning_rate": 1.5536615212864168e-05, + "loss": 0.9892, + "step": 20900 + }, + { + "epoch": 1.34, + "learning_rate": 1.5536400523841217e-05, + "loss": 1.0217, + "step": 20901 + }, + { + "epoch": 1.34, + "learning_rate": 1.5536185834818266e-05, + "loss": 0.8876, + "step": 20902 + }, + { + "epoch": 1.34, + "learning_rate": 1.5535971145795318e-05, + "loss": 0.966, + "step": 20903 + }, + { + "epoch": 1.34, + "learning_rate": 1.5535756456772367e-05, + "loss": 1.0222, + "step": 20904 + }, + { + "epoch": 1.34, + "learning_rate": 1.5535541767749416e-05, + "loss": 1.0163, + "step": 20905 + }, + { + "epoch": 1.35, + "learning_rate": 1.5535327078726465e-05, + "loss": 1.0436, + "step": 20906 + }, + { + "epoch": 1.35, + "learning_rate": 1.5535112389703514e-05, + "loss": 1.0207, + "step": 20907 + }, + { + "epoch": 1.35, + "learning_rate": 1.5534897700680566e-05, + "loss": 0.9532, + "step": 20908 + }, + { + "epoch": 1.35, + "learning_rate": 1.5534683011657615e-05, + "loss": 0.9652, + "step": 20909 + }, + { + "epoch": 1.35, + "learning_rate": 1.5534468322634668e-05, + "loss": 1.0079, + "step": 20910 + }, + { + "epoch": 1.35, + "learning_rate": 1.5534253633611717e-05, + "loss": 0.9439, + "step": 20911 + }, + { + "epoch": 1.35, + "learning_rate": 1.5534038944588765e-05, + "loss": 0.9337, + "step": 20912 + }, + { + "epoch": 1.35, + "learning_rate": 1.5533824255565814e-05, + "loss": 0.9547, + "step": 20913 + }, + { + "epoch": 1.35, + "learning_rate": 1.5533609566542863e-05, + "loss": 0.9998, + "step": 20914 + }, + { + "epoch": 1.35, + "learning_rate": 1.5533394877519912e-05, + "loss": 1.0007, + "step": 20915 + }, + { + "epoch": 1.35, + "learning_rate": 1.5533180188496965e-05, + "loss": 0.9908, + "step": 20916 + }, + { + "epoch": 1.35, + "learning_rate": 1.5532965499474014e-05, + "loss": 0.8746, + "step": 20917 + }, + { + "epoch": 1.35, + "learning_rate": 1.5532750810451063e-05, + "loss": 0.9321, + "step": 20918 + }, + { + "epoch": 1.35, + "learning_rate": 1.553253612142811e-05, + "loss": 1.0164, + "step": 20919 + }, + { + "epoch": 1.35, + "learning_rate": 1.553232143240516e-05, + "loss": 1.0338, + "step": 20920 + }, + { + "epoch": 1.35, + "learning_rate": 1.5532106743382213e-05, + "loss": 0.9836, + "step": 20921 + }, + { + "epoch": 1.35, + "learning_rate": 1.5531892054359262e-05, + "loss": 0.9339, + "step": 20922 + }, + { + "epoch": 1.35, + "learning_rate": 1.553167736533631e-05, + "loss": 0.9941, + "step": 20923 + }, + { + "epoch": 1.35, + "learning_rate": 1.5531462676313363e-05, + "loss": 0.98, + "step": 20924 + }, + { + "epoch": 1.35, + "learning_rate": 1.5531247987290412e-05, + "loss": 0.9248, + "step": 20925 + }, + { + "epoch": 1.35, + "learning_rate": 1.553103329826746e-05, + "loss": 1.0727, + "step": 20926 + }, + { + "epoch": 1.35, + "learning_rate": 1.553081860924451e-05, + "loss": 1.016, + "step": 20927 + }, + { + "epoch": 1.35, + "learning_rate": 1.5530603920221562e-05, + "loss": 1.0093, + "step": 20928 + }, + { + "epoch": 1.35, + "learning_rate": 1.553038923119861e-05, + "loss": 0.922, + "step": 20929 + }, + { + "epoch": 1.35, + "learning_rate": 1.553017454217566e-05, + "loss": 0.9414, + "step": 20930 + }, + { + "epoch": 1.35, + "learning_rate": 1.552995985315271e-05, + "loss": 0.919, + "step": 20931 + }, + { + "epoch": 1.35, + "learning_rate": 1.552974516412976e-05, + "loss": 0.945, + "step": 20932 + }, + { + "epoch": 1.35, + "learning_rate": 1.5529530475106807e-05, + "loss": 1.0397, + "step": 20933 + }, + { + "epoch": 1.35, + "learning_rate": 1.552931578608386e-05, + "loss": 0.9772, + "step": 20934 + }, + { + "epoch": 1.35, + "learning_rate": 1.552910109706091e-05, + "loss": 0.9335, + "step": 20935 + }, + { + "epoch": 1.35, + "learning_rate": 1.5528886408037958e-05, + "loss": 0.9854, + "step": 20936 + }, + { + "epoch": 1.35, + "learning_rate": 1.5528671719015007e-05, + "loss": 0.9386, + "step": 20937 + }, + { + "epoch": 1.35, + "learning_rate": 1.5528457029992056e-05, + "loss": 0.9776, + "step": 20938 + }, + { + "epoch": 1.35, + "learning_rate": 1.5528242340969108e-05, + "loss": 0.9697, + "step": 20939 + }, + { + "epoch": 1.35, + "learning_rate": 1.5528027651946157e-05, + "loss": 0.9837, + "step": 20940 + }, + { + "epoch": 1.35, + "learning_rate": 1.552781296292321e-05, + "loss": 0.9272, + "step": 20941 + }, + { + "epoch": 1.35, + "learning_rate": 1.5527598273900258e-05, + "loss": 1.0178, + "step": 20942 + }, + { + "epoch": 1.35, + "learning_rate": 1.5527383584877307e-05, + "loss": 0.9604, + "step": 20943 + }, + { + "epoch": 1.35, + "learning_rate": 1.5527168895854356e-05, + "loss": 0.9472, + "step": 20944 + }, + { + "epoch": 1.35, + "learning_rate": 1.5526954206831405e-05, + "loss": 0.9581, + "step": 20945 + }, + { + "epoch": 1.35, + "learning_rate": 1.5526739517808454e-05, + "loss": 1.0, + "step": 20946 + }, + { + "epoch": 1.35, + "learning_rate": 1.5526524828785506e-05, + "loss": 1.0082, + "step": 20947 + }, + { + "epoch": 1.35, + "learning_rate": 1.5526310139762555e-05, + "loss": 0.9254, + "step": 20948 + }, + { + "epoch": 1.35, + "learning_rate": 1.5526095450739604e-05, + "loss": 0.9599, + "step": 20949 + }, + { + "epoch": 1.35, + "learning_rate": 1.5525880761716653e-05, + "loss": 1.0081, + "step": 20950 + }, + { + "epoch": 1.35, + "learning_rate": 1.5525666072693702e-05, + "loss": 0.9492, + "step": 20951 + }, + { + "epoch": 1.35, + "learning_rate": 1.5525451383670755e-05, + "loss": 0.9319, + "step": 20952 + }, + { + "epoch": 1.35, + "learning_rate": 1.5525236694647804e-05, + "loss": 0.9964, + "step": 20953 + }, + { + "epoch": 1.35, + "learning_rate": 1.5525022005624856e-05, + "loss": 0.9503, + "step": 20954 + }, + { + "epoch": 1.35, + "learning_rate": 1.5524807316601905e-05, + "loss": 0.95, + "step": 20955 + }, + { + "epoch": 1.35, + "learning_rate": 1.5524592627578954e-05, + "loss": 0.9938, + "step": 20956 + }, + { + "epoch": 1.35, + "learning_rate": 1.5524377938556003e-05, + "loss": 1.016, + "step": 20957 + }, + { + "epoch": 1.35, + "learning_rate": 1.5524163249533052e-05, + "loss": 1.0228, + "step": 20958 + }, + { + "epoch": 1.35, + "learning_rate": 1.5523948560510104e-05, + "loss": 0.9902, + "step": 20959 + }, + { + "epoch": 1.35, + "learning_rate": 1.5523733871487153e-05, + "loss": 0.9549, + "step": 20960 + }, + { + "epoch": 1.35, + "learning_rate": 1.5523519182464202e-05, + "loss": 0.9893, + "step": 20961 + }, + { + "epoch": 1.35, + "learning_rate": 1.552330449344125e-05, + "loss": 1.0178, + "step": 20962 + }, + { + "epoch": 1.35, + "learning_rate": 1.55230898044183e-05, + "loss": 1.062, + "step": 20963 + }, + { + "epoch": 1.35, + "learning_rate": 1.552287511539535e-05, + "loss": 1.0376, + "step": 20964 + }, + { + "epoch": 1.35, + "learning_rate": 1.55226604263724e-05, + "loss": 1.0368, + "step": 20965 + }, + { + "epoch": 1.35, + "learning_rate": 1.552244573734945e-05, + "loss": 1.0035, + "step": 20966 + }, + { + "epoch": 1.35, + "learning_rate": 1.55222310483265e-05, + "loss": 0.9559, + "step": 20967 + }, + { + "epoch": 1.35, + "learning_rate": 1.552201635930355e-05, + "loss": 0.9826, + "step": 20968 + }, + { + "epoch": 1.35, + "learning_rate": 1.55218016702806e-05, + "loss": 1.0364, + "step": 20969 + }, + { + "epoch": 1.35, + "learning_rate": 1.552158698125765e-05, + "loss": 0.9602, + "step": 20970 + }, + { + "epoch": 1.35, + "learning_rate": 1.55213722922347e-05, + "loss": 0.9472, + "step": 20971 + }, + { + "epoch": 1.35, + "learning_rate": 1.552115760321175e-05, + "loss": 0.9758, + "step": 20972 + }, + { + "epoch": 1.35, + "learning_rate": 1.55209429141888e-05, + "loss": 0.8946, + "step": 20973 + }, + { + "epoch": 1.35, + "learning_rate": 1.552072822516585e-05, + "loss": 0.8464, + "step": 20974 + }, + { + "epoch": 1.35, + "learning_rate": 1.5520513536142898e-05, + "loss": 0.9779, + "step": 20975 + }, + { + "epoch": 1.35, + "learning_rate": 1.5520298847119947e-05, + "loss": 1.0251, + "step": 20976 + }, + { + "epoch": 1.35, + "learning_rate": 1.5520084158097e-05, + "loss": 0.9406, + "step": 20977 + }, + { + "epoch": 1.35, + "learning_rate": 1.5519869469074048e-05, + "loss": 0.9871, + "step": 20978 + }, + { + "epoch": 1.35, + "learning_rate": 1.5519654780051097e-05, + "loss": 0.9455, + "step": 20979 + }, + { + "epoch": 1.35, + "learning_rate": 1.5519440091028146e-05, + "loss": 1.0143, + "step": 20980 + }, + { + "epoch": 1.35, + "learning_rate": 1.5519225402005195e-05, + "loss": 0.9681, + "step": 20981 + }, + { + "epoch": 1.35, + "learning_rate": 1.5519010712982247e-05, + "loss": 1.0211, + "step": 20982 + }, + { + "epoch": 1.35, + "learning_rate": 1.5518796023959296e-05, + "loss": 0.9696, + "step": 20983 + }, + { + "epoch": 1.35, + "learning_rate": 1.5518581334936345e-05, + "loss": 0.9938, + "step": 20984 + }, + { + "epoch": 1.35, + "learning_rate": 1.5518366645913398e-05, + "loss": 1.0314, + "step": 20985 + }, + { + "epoch": 1.35, + "learning_rate": 1.5518151956890446e-05, + "loss": 0.9226, + "step": 20986 + }, + { + "epoch": 1.35, + "learning_rate": 1.5517937267867495e-05, + "loss": 0.9726, + "step": 20987 + }, + { + "epoch": 1.35, + "learning_rate": 1.5517722578844544e-05, + "loss": 0.9922, + "step": 20988 + }, + { + "epoch": 1.35, + "learning_rate": 1.5517507889821593e-05, + "loss": 0.9338, + "step": 20989 + }, + { + "epoch": 1.35, + "learning_rate": 1.5517293200798646e-05, + "loss": 0.9466, + "step": 20990 + }, + { + "epoch": 1.35, + "learning_rate": 1.5517078511775695e-05, + "loss": 1.0508, + "step": 20991 + }, + { + "epoch": 1.35, + "learning_rate": 1.5516863822752744e-05, + "loss": 0.9434, + "step": 20992 + }, + { + "epoch": 1.35, + "learning_rate": 1.5516649133729793e-05, + "loss": 0.9608, + "step": 20993 + }, + { + "epoch": 1.35, + "learning_rate": 1.551643444470684e-05, + "loss": 0.9924, + "step": 20994 + }, + { + "epoch": 1.35, + "learning_rate": 1.5516219755683894e-05, + "loss": 0.9215, + "step": 20995 + }, + { + "epoch": 1.35, + "learning_rate": 1.5516005066660943e-05, + "loss": 1.0068, + "step": 20996 + }, + { + "epoch": 1.35, + "learning_rate": 1.5515790377637995e-05, + "loss": 0.9881, + "step": 20997 + }, + { + "epoch": 1.35, + "learning_rate": 1.5515575688615044e-05, + "loss": 1.0187, + "step": 20998 + }, + { + "epoch": 1.35, + "learning_rate": 1.5515360999592093e-05, + "loss": 1.0016, + "step": 20999 + }, + { + "epoch": 1.35, + "learning_rate": 1.5515146310569142e-05, + "loss": 1.0869, + "step": 21000 + }, + { + "epoch": 1.35, + "eval_laion/cider": 0.5076302383541398, + "eval_laion/meteor": 0.38824260937117516, + "eval_laion/pred_num_words": 36.822265625, + "eval_laion/sacrebleu": 15.780459170131405, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0543861389160156, + "eval_runtime": 98.1825, + "eval_samples_per_second": 5.082, + "eval_steps_per_second": 0.326, + "step": 21000 + }, + { + "epoch": 1.35, + "learning_rate": 1.551493162154619e-05, + "loss": 0.9261, + "step": 21001 + }, + { + "epoch": 1.35, + "learning_rate": 1.551471693252324e-05, + "loss": 0.9645, + "step": 21002 + }, + { + "epoch": 1.35, + "learning_rate": 1.5514502243500292e-05, + "loss": 0.9938, + "step": 21003 + }, + { + "epoch": 1.35, + "learning_rate": 1.551428755447734e-05, + "loss": 0.9777, + "step": 21004 + }, + { + "epoch": 1.35, + "learning_rate": 1.551407286545439e-05, + "loss": 0.9316, + "step": 21005 + }, + { + "epoch": 1.35, + "learning_rate": 1.551385817643144e-05, + "loss": 0.9566, + "step": 21006 + }, + { + "epoch": 1.35, + "learning_rate": 1.5513643487408488e-05, + "loss": 0.9406, + "step": 21007 + }, + { + "epoch": 1.35, + "learning_rate": 1.551342879838554e-05, + "loss": 1.0524, + "step": 21008 + }, + { + "epoch": 1.35, + "learning_rate": 1.551321410936259e-05, + "loss": 1.0818, + "step": 21009 + }, + { + "epoch": 1.35, + "learning_rate": 1.551299942033964e-05, + "loss": 1.0685, + "step": 21010 + }, + { + "epoch": 1.35, + "learning_rate": 1.5512784731316688e-05, + "loss": 0.9366, + "step": 21011 + }, + { + "epoch": 1.35, + "learning_rate": 1.551257004229374e-05, + "loss": 0.9822, + "step": 21012 + }, + { + "epoch": 1.35, + "learning_rate": 1.551235535327079e-05, + "loss": 0.9765, + "step": 21013 + }, + { + "epoch": 1.35, + "learning_rate": 1.5512140664247838e-05, + "loss": 1.0123, + "step": 21014 + }, + { + "epoch": 1.35, + "learning_rate": 1.551192597522489e-05, + "loss": 0.9857, + "step": 21015 + }, + { + "epoch": 1.35, + "learning_rate": 1.551171128620194e-05, + "loss": 0.985, + "step": 21016 + }, + { + "epoch": 1.35, + "learning_rate": 1.5511496597178988e-05, + "loss": 0.9471, + "step": 21017 + }, + { + "epoch": 1.35, + "learning_rate": 1.5511281908156037e-05, + "loss": 0.9228, + "step": 21018 + }, + { + "epoch": 1.35, + "learning_rate": 1.5511067219133086e-05, + "loss": 0.9846, + "step": 21019 + }, + { + "epoch": 1.35, + "learning_rate": 1.5510852530110135e-05, + "loss": 0.9913, + "step": 21020 + }, + { + "epoch": 1.35, + "learning_rate": 1.5510637841087187e-05, + "loss": 1.0643, + "step": 21021 + }, + { + "epoch": 1.35, + "learning_rate": 1.5510423152064236e-05, + "loss": 1.0315, + "step": 21022 + }, + { + "epoch": 1.35, + "learning_rate": 1.5510208463041285e-05, + "loss": 0.9918, + "step": 21023 + }, + { + "epoch": 1.35, + "learning_rate": 1.5509993774018334e-05, + "loss": 0.921, + "step": 21024 + }, + { + "epoch": 1.35, + "learning_rate": 1.5509779084995383e-05, + "loss": 0.9326, + "step": 21025 + }, + { + "epoch": 1.35, + "learning_rate": 1.5509564395972436e-05, + "loss": 0.9564, + "step": 21026 + }, + { + "epoch": 1.35, + "learning_rate": 1.5509349706949485e-05, + "loss": 1.037, + "step": 21027 + }, + { + "epoch": 1.35, + "learning_rate": 1.5509135017926537e-05, + "loss": 0.9389, + "step": 21028 + }, + { + "epoch": 1.35, + "learning_rate": 1.5508920328903586e-05, + "loss": 0.9861, + "step": 21029 + }, + { + "epoch": 1.35, + "learning_rate": 1.5508705639880635e-05, + "loss": 0.9234, + "step": 21030 + }, + { + "epoch": 1.35, + "learning_rate": 1.5508490950857684e-05, + "loss": 1.0281, + "step": 21031 + }, + { + "epoch": 1.35, + "learning_rate": 1.5508276261834733e-05, + "loss": 1.0387, + "step": 21032 + }, + { + "epoch": 1.35, + "learning_rate": 1.5508061572811782e-05, + "loss": 1.0175, + "step": 21033 + }, + { + "epoch": 1.35, + "learning_rate": 1.5507846883788834e-05, + "loss": 1.0427, + "step": 21034 + }, + { + "epoch": 1.35, + "learning_rate": 1.5507632194765883e-05, + "loss": 1.0041, + "step": 21035 + }, + { + "epoch": 1.35, + "learning_rate": 1.5507417505742932e-05, + "loss": 0.9755, + "step": 21036 + }, + { + "epoch": 1.35, + "learning_rate": 1.550720281671998e-05, + "loss": 0.9176, + "step": 21037 + }, + { + "epoch": 1.35, + "learning_rate": 1.550698812769703e-05, + "loss": 0.9301, + "step": 21038 + }, + { + "epoch": 1.35, + "learning_rate": 1.5506773438674082e-05, + "loss": 0.9746, + "step": 21039 + }, + { + "epoch": 1.35, + "learning_rate": 1.550655874965113e-05, + "loss": 0.9188, + "step": 21040 + }, + { + "epoch": 1.35, + "learning_rate": 1.5506344060628184e-05, + "loss": 1.0071, + "step": 21041 + }, + { + "epoch": 1.35, + "learning_rate": 1.5506129371605233e-05, + "loss": 1.0751, + "step": 21042 + }, + { + "epoch": 1.35, + "learning_rate": 1.550591468258228e-05, + "loss": 0.983, + "step": 21043 + }, + { + "epoch": 1.35, + "learning_rate": 1.550569999355933e-05, + "loss": 0.9175, + "step": 21044 + }, + { + "epoch": 1.35, + "learning_rate": 1.550548530453638e-05, + "loss": 1.0678, + "step": 21045 + }, + { + "epoch": 1.35, + "learning_rate": 1.5505270615513432e-05, + "loss": 0.9738, + "step": 21046 + }, + { + "epoch": 1.35, + "learning_rate": 1.550505592649048e-05, + "loss": 0.9255, + "step": 21047 + }, + { + "epoch": 1.35, + "learning_rate": 1.550484123746753e-05, + "loss": 0.9385, + "step": 21048 + }, + { + "epoch": 1.35, + "learning_rate": 1.550462654844458e-05, + "loss": 0.9521, + "step": 21049 + }, + { + "epoch": 1.35, + "learning_rate": 1.5504411859421628e-05, + "loss": 1.0344, + "step": 21050 + }, + { + "epoch": 1.35, + "learning_rate": 1.5504197170398677e-05, + "loss": 0.9203, + "step": 21051 + }, + { + "epoch": 1.35, + "learning_rate": 1.550398248137573e-05, + "loss": 0.9261, + "step": 21052 + }, + { + "epoch": 1.35, + "learning_rate": 1.5503767792352778e-05, + "loss": 0.9774, + "step": 21053 + }, + { + "epoch": 1.35, + "learning_rate": 1.5503553103329827e-05, + "loss": 0.9529, + "step": 21054 + }, + { + "epoch": 1.35, + "learning_rate": 1.550333841430688e-05, + "loss": 0.9374, + "step": 21055 + }, + { + "epoch": 1.35, + "learning_rate": 1.5503123725283928e-05, + "loss": 1.0331, + "step": 21056 + }, + { + "epoch": 1.35, + "learning_rate": 1.5502909036260977e-05, + "loss": 0.8826, + "step": 21057 + }, + { + "epoch": 1.35, + "learning_rate": 1.5502694347238026e-05, + "loss": 0.9857, + "step": 21058 + }, + { + "epoch": 1.35, + "learning_rate": 1.550247965821508e-05, + "loss": 0.9566, + "step": 21059 + }, + { + "epoch": 1.35, + "learning_rate": 1.5502264969192127e-05, + "loss": 0.9826, + "step": 21060 + }, + { + "epoch": 1.36, + "learning_rate": 1.5502050280169176e-05, + "loss": 0.9898, + "step": 21061 + }, + { + "epoch": 1.36, + "learning_rate": 1.5501835591146225e-05, + "loss": 0.9644, + "step": 21062 + }, + { + "epoch": 1.36, + "learning_rate": 1.5501620902123274e-05, + "loss": 0.9901, + "step": 21063 + }, + { + "epoch": 1.36, + "learning_rate": 1.5501406213100327e-05, + "loss": 0.9535, + "step": 21064 + }, + { + "epoch": 1.36, + "learning_rate": 1.5501191524077376e-05, + "loss": 0.9898, + "step": 21065 + }, + { + "epoch": 1.36, + "learning_rate": 1.5500976835054425e-05, + "loss": 1.0231, + "step": 21066 + }, + { + "epoch": 1.36, + "learning_rate": 1.5500762146031474e-05, + "loss": 0.9285, + "step": 21067 + }, + { + "epoch": 1.36, + "learning_rate": 1.5500547457008523e-05, + "loss": 1.0445, + "step": 21068 + }, + { + "epoch": 1.36, + "learning_rate": 1.550033276798557e-05, + "loss": 0.9166, + "step": 21069 + }, + { + "epoch": 1.36, + "learning_rate": 1.5500118078962624e-05, + "loss": 0.9785, + "step": 21070 + }, + { + "epoch": 1.36, + "learning_rate": 1.5499903389939673e-05, + "loss": 0.9591, + "step": 21071 + }, + { + "epoch": 1.36, + "learning_rate": 1.5499688700916725e-05, + "loss": 0.9233, + "step": 21072 + }, + { + "epoch": 1.36, + "learning_rate": 1.5499474011893774e-05, + "loss": 0.942, + "step": 21073 + }, + { + "epoch": 1.36, + "learning_rate": 1.5499259322870823e-05, + "loss": 0.9868, + "step": 21074 + }, + { + "epoch": 1.36, + "learning_rate": 1.5499044633847872e-05, + "loss": 0.9948, + "step": 21075 + }, + { + "epoch": 1.36, + "learning_rate": 1.549882994482492e-05, + "loss": 0.9493, + "step": 21076 + }, + { + "epoch": 1.36, + "learning_rate": 1.5498615255801973e-05, + "loss": 0.9999, + "step": 21077 + }, + { + "epoch": 1.36, + "learning_rate": 1.5498400566779022e-05, + "loss": 0.9408, + "step": 21078 + }, + { + "epoch": 1.36, + "learning_rate": 1.549818587775607e-05, + "loss": 0.916, + "step": 21079 + }, + { + "epoch": 1.36, + "learning_rate": 1.549797118873312e-05, + "loss": 0.9109, + "step": 21080 + }, + { + "epoch": 1.36, + "learning_rate": 1.549775649971017e-05, + "loss": 0.9699, + "step": 21081 + }, + { + "epoch": 1.36, + "learning_rate": 1.549754181068722e-05, + "loss": 1.0277, + "step": 21082 + }, + { + "epoch": 1.36, + "learning_rate": 1.549732712166427e-05, + "loss": 0.9426, + "step": 21083 + }, + { + "epoch": 1.36, + "learning_rate": 1.5497112432641323e-05, + "loss": 0.9949, + "step": 21084 + }, + { + "epoch": 1.36, + "learning_rate": 1.5496897743618372e-05, + "loss": 0.9474, + "step": 21085 + }, + { + "epoch": 1.36, + "learning_rate": 1.549668305459542e-05, + "loss": 0.9658, + "step": 21086 + }, + { + "epoch": 1.36, + "learning_rate": 1.549646836557247e-05, + "loss": 1.0276, + "step": 21087 + }, + { + "epoch": 1.36, + "learning_rate": 1.549625367654952e-05, + "loss": 0.8453, + "step": 21088 + }, + { + "epoch": 1.36, + "learning_rate": 1.5496038987526568e-05, + "loss": 0.9966, + "step": 21089 + }, + { + "epoch": 1.36, + "learning_rate": 1.549582429850362e-05, + "loss": 0.9723, + "step": 21090 + }, + { + "epoch": 1.36, + "learning_rate": 1.549560960948067e-05, + "loss": 1.0423, + "step": 21091 + }, + { + "epoch": 1.36, + "learning_rate": 1.5495394920457718e-05, + "loss": 1.0028, + "step": 21092 + }, + { + "epoch": 1.36, + "learning_rate": 1.5495180231434767e-05, + "loss": 0.9259, + "step": 21093 + }, + { + "epoch": 1.36, + "learning_rate": 1.5494965542411816e-05, + "loss": 0.8974, + "step": 21094 + }, + { + "epoch": 1.36, + "learning_rate": 1.549475085338887e-05, + "loss": 1.0092, + "step": 21095 + }, + { + "epoch": 1.36, + "learning_rate": 1.5494536164365917e-05, + "loss": 0.985, + "step": 21096 + }, + { + "epoch": 1.36, + "learning_rate": 1.5494321475342966e-05, + "loss": 1.0364, + "step": 21097 + }, + { + "epoch": 1.36, + "learning_rate": 1.5494106786320015e-05, + "loss": 0.949, + "step": 21098 + }, + { + "epoch": 1.36, + "learning_rate": 1.5493892097297068e-05, + "loss": 0.9702, + "step": 21099 + }, + { + "epoch": 1.36, + "learning_rate": 1.5493677408274117e-05, + "loss": 0.8859, + "step": 21100 + }, + { + "epoch": 1.36, + "learning_rate": 1.5493462719251166e-05, + "loss": 0.9551, + "step": 21101 + }, + { + "epoch": 1.36, + "learning_rate": 1.5493248030228218e-05, + "loss": 1.0484, + "step": 21102 + }, + { + "epoch": 1.36, + "learning_rate": 1.5493033341205267e-05, + "loss": 0.9171, + "step": 21103 + }, + { + "epoch": 1.36, + "learning_rate": 1.5492818652182316e-05, + "loss": 0.9364, + "step": 21104 + }, + { + "epoch": 1.36, + "learning_rate": 1.5492603963159365e-05, + "loss": 0.9649, + "step": 21105 + }, + { + "epoch": 1.36, + "learning_rate": 1.5492389274136414e-05, + "loss": 1.0087, + "step": 21106 + }, + { + "epoch": 1.36, + "learning_rate": 1.5492174585113463e-05, + "loss": 0.9985, + "step": 21107 + }, + { + "epoch": 1.36, + "learning_rate": 1.5491959896090515e-05, + "loss": 0.9558, + "step": 21108 + }, + { + "epoch": 1.36, + "learning_rate": 1.5491745207067564e-05, + "loss": 0.9575, + "step": 21109 + }, + { + "epoch": 1.36, + "learning_rate": 1.5491530518044613e-05, + "loss": 0.8676, + "step": 21110 + }, + { + "epoch": 1.36, + "learning_rate": 1.5491315829021662e-05, + "loss": 0.9146, + "step": 21111 + }, + { + "epoch": 1.36, + "learning_rate": 1.549110113999871e-05, + "loss": 0.9976, + "step": 21112 + }, + { + "epoch": 1.36, + "learning_rate": 1.5490886450975763e-05, + "loss": 0.9971, + "step": 21113 + }, + { + "epoch": 1.36, + "learning_rate": 1.5490671761952812e-05, + "loss": 1.0573, + "step": 21114 + }, + { + "epoch": 1.36, + "learning_rate": 1.5490457072929865e-05, + "loss": 0.9803, + "step": 21115 + }, + { + "epoch": 1.36, + "learning_rate": 1.5490242383906914e-05, + "loss": 0.9853, + "step": 21116 + }, + { + "epoch": 1.36, + "learning_rate": 1.5490027694883963e-05, + "loss": 0.9968, + "step": 21117 + }, + { + "epoch": 1.36, + "learning_rate": 1.548981300586101e-05, + "loss": 0.9085, + "step": 21118 + }, + { + "epoch": 1.36, + "learning_rate": 1.548959831683806e-05, + "loss": 0.9814, + "step": 21119 + }, + { + "epoch": 1.36, + "learning_rate": 1.548938362781511e-05, + "loss": 1.0077, + "step": 21120 + }, + { + "epoch": 1.36, + "learning_rate": 1.5489168938792162e-05, + "loss": 0.8803, + "step": 21121 + }, + { + "epoch": 1.36, + "learning_rate": 1.548895424976921e-05, + "loss": 1.0056, + "step": 21122 + }, + { + "epoch": 1.36, + "learning_rate": 1.548873956074626e-05, + "loss": 0.9442, + "step": 21123 + }, + { + "epoch": 1.36, + "learning_rate": 1.548852487172331e-05, + "loss": 0.956, + "step": 21124 + }, + { + "epoch": 1.36, + "learning_rate": 1.5488310182700358e-05, + "loss": 0.9699, + "step": 21125 + }, + { + "epoch": 1.36, + "learning_rate": 1.548809549367741e-05, + "loss": 1.0079, + "step": 21126 + }, + { + "epoch": 1.36, + "learning_rate": 1.548788080465446e-05, + "loss": 1.0059, + "step": 21127 + }, + { + "epoch": 1.36, + "learning_rate": 1.548766611563151e-05, + "loss": 1.0092, + "step": 21128 + }, + { + "epoch": 1.36, + "learning_rate": 1.548745142660856e-05, + "loss": 0.9648, + "step": 21129 + }, + { + "epoch": 1.36, + "learning_rate": 1.548723673758561e-05, + "loss": 1.0585, + "step": 21130 + }, + { + "epoch": 1.36, + "learning_rate": 1.5487022048562658e-05, + "loss": 0.976, + "step": 21131 + }, + { + "epoch": 1.36, + "learning_rate": 1.5486807359539707e-05, + "loss": 0.9725, + "step": 21132 + }, + { + "epoch": 1.36, + "learning_rate": 1.548659267051676e-05, + "loss": 0.9536, + "step": 21133 + }, + { + "epoch": 1.36, + "learning_rate": 1.548637798149381e-05, + "loss": 1.023, + "step": 21134 + }, + { + "epoch": 1.36, + "learning_rate": 1.5486163292470857e-05, + "loss": 0.9846, + "step": 21135 + }, + { + "epoch": 1.36, + "learning_rate": 1.5485948603447906e-05, + "loss": 1.0115, + "step": 21136 + }, + { + "epoch": 1.36, + "learning_rate": 1.5485733914424955e-05, + "loss": 0.9485, + "step": 21137 + }, + { + "epoch": 1.36, + "learning_rate": 1.5485519225402004e-05, + "loss": 0.9832, + "step": 21138 + }, + { + "epoch": 1.36, + "learning_rate": 1.5485304536379057e-05, + "loss": 0.9994, + "step": 21139 + }, + { + "epoch": 1.36, + "learning_rate": 1.5485089847356106e-05, + "loss": 1.0018, + "step": 21140 + }, + { + "epoch": 1.36, + "learning_rate": 1.5484875158333155e-05, + "loss": 0.9773, + "step": 21141 + }, + { + "epoch": 1.36, + "learning_rate": 1.5484660469310207e-05, + "loss": 1.05, + "step": 21142 + }, + { + "epoch": 1.36, + "learning_rate": 1.5484445780287256e-05, + "loss": 0.9795, + "step": 21143 + }, + { + "epoch": 1.36, + "learning_rate": 1.5484231091264305e-05, + "loss": 1.0045, + "step": 21144 + }, + { + "epoch": 1.36, + "learning_rate": 1.5484016402241354e-05, + "loss": 1.0221, + "step": 21145 + }, + { + "epoch": 1.36, + "learning_rate": 1.5483801713218406e-05, + "loss": 1.0316, + "step": 21146 + }, + { + "epoch": 1.36, + "learning_rate": 1.5483587024195455e-05, + "loss": 0.9965, + "step": 21147 + }, + { + "epoch": 1.36, + "learning_rate": 1.5483372335172504e-05, + "loss": 1.0399, + "step": 21148 + }, + { + "epoch": 1.36, + "learning_rate": 1.5483157646149553e-05, + "loss": 0.9679, + "step": 21149 + }, + { + "epoch": 1.36, + "learning_rate": 1.5482942957126602e-05, + "loss": 0.9595, + "step": 21150 + }, + { + "epoch": 1.36, + "learning_rate": 1.5482728268103654e-05, + "loss": 0.9601, + "step": 21151 + }, + { + "epoch": 1.36, + "learning_rate": 1.5482513579080703e-05, + "loss": 0.9901, + "step": 21152 + }, + { + "epoch": 1.36, + "learning_rate": 1.5482298890057752e-05, + "loss": 0.9996, + "step": 21153 + }, + { + "epoch": 1.36, + "learning_rate": 1.54820842010348e-05, + "loss": 1.028, + "step": 21154 + }, + { + "epoch": 1.36, + "learning_rate": 1.548186951201185e-05, + "loss": 0.9504, + "step": 21155 + }, + { + "epoch": 1.36, + "learning_rate": 1.54816548229889e-05, + "loss": 0.9276, + "step": 21156 + }, + { + "epoch": 1.36, + "learning_rate": 1.548144013396595e-05, + "loss": 1.0051, + "step": 21157 + }, + { + "epoch": 1.36, + "learning_rate": 1.5481225444943e-05, + "loss": 0.9355, + "step": 21158 + }, + { + "epoch": 1.36, + "learning_rate": 1.5481010755920053e-05, + "loss": 0.9595, + "step": 21159 + }, + { + "epoch": 1.36, + "learning_rate": 1.5480796066897102e-05, + "loss": 0.9489, + "step": 21160 + }, + { + "epoch": 1.36, + "learning_rate": 1.548058137787415e-05, + "loss": 0.9707, + "step": 21161 + }, + { + "epoch": 1.36, + "learning_rate": 1.54803666888512e-05, + "loss": 0.9297, + "step": 21162 + }, + { + "epoch": 1.36, + "learning_rate": 1.548015199982825e-05, + "loss": 0.9529, + "step": 21163 + }, + { + "epoch": 1.36, + "learning_rate": 1.54799373108053e-05, + "loss": 1.0464, + "step": 21164 + }, + { + "epoch": 1.36, + "learning_rate": 1.547972262178235e-05, + "loss": 1.0145, + "step": 21165 + }, + { + "epoch": 1.36, + "learning_rate": 1.54795079327594e-05, + "loss": 0.9646, + "step": 21166 + }, + { + "epoch": 1.36, + "learning_rate": 1.5479293243736448e-05, + "loss": 0.9813, + "step": 21167 + }, + { + "epoch": 1.36, + "learning_rate": 1.5479078554713497e-05, + "loss": 1.1057, + "step": 21168 + }, + { + "epoch": 1.36, + "learning_rate": 1.5478863865690546e-05, + "loss": 1.0641, + "step": 21169 + }, + { + "epoch": 1.36, + "learning_rate": 1.54786491766676e-05, + "loss": 0.9891, + "step": 21170 + }, + { + "epoch": 1.36, + "learning_rate": 1.547843448764465e-05, + "loss": 1.0122, + "step": 21171 + }, + { + "epoch": 1.36, + "learning_rate": 1.54782197986217e-05, + "loss": 1.0008, + "step": 21172 + }, + { + "epoch": 1.36, + "learning_rate": 1.547800510959875e-05, + "loss": 1.0263, + "step": 21173 + }, + { + "epoch": 1.36, + "learning_rate": 1.5477790420575798e-05, + "loss": 0.9855, + "step": 21174 + }, + { + "epoch": 1.36, + "learning_rate": 1.5477575731552847e-05, + "loss": 0.9835, + "step": 21175 + }, + { + "epoch": 1.36, + "learning_rate": 1.5477361042529896e-05, + "loss": 0.9367, + "step": 21176 + }, + { + "epoch": 1.36, + "learning_rate": 1.5477146353506948e-05, + "loss": 0.91, + "step": 21177 + }, + { + "epoch": 1.36, + "learning_rate": 1.5476931664483997e-05, + "loss": 0.9217, + "step": 21178 + }, + { + "epoch": 1.36, + "learning_rate": 1.5476716975461046e-05, + "loss": 0.9965, + "step": 21179 + }, + { + "epoch": 1.36, + "learning_rate": 1.5476502286438095e-05, + "loss": 0.9869, + "step": 21180 + }, + { + "epoch": 1.36, + "learning_rate": 1.5476287597415144e-05, + "loss": 0.8913, + "step": 21181 + }, + { + "epoch": 1.36, + "learning_rate": 1.5476072908392196e-05, + "loss": 0.971, + "step": 21182 + }, + { + "epoch": 1.36, + "learning_rate": 1.5475858219369245e-05, + "loss": 0.9792, + "step": 21183 + }, + { + "epoch": 1.36, + "learning_rate": 1.5475643530346294e-05, + "loss": 1.0057, + "step": 21184 + }, + { + "epoch": 1.36, + "learning_rate": 1.5475428841323343e-05, + "loss": 1.0761, + "step": 21185 + }, + { + "epoch": 1.36, + "learning_rate": 1.5475214152300395e-05, + "loss": 0.939, + "step": 21186 + }, + { + "epoch": 1.36, + "learning_rate": 1.5474999463277444e-05, + "loss": 0.9936, + "step": 21187 + }, + { + "epoch": 1.36, + "learning_rate": 1.5474784774254493e-05, + "loss": 1.0113, + "step": 21188 + }, + { + "epoch": 1.36, + "learning_rate": 1.5474570085231546e-05, + "loss": 0.9854, + "step": 21189 + }, + { + "epoch": 1.36, + "learning_rate": 1.5474355396208595e-05, + "loss": 1.0641, + "step": 21190 + }, + { + "epoch": 1.36, + "learning_rate": 1.5474140707185644e-05, + "loss": 0.9895, + "step": 21191 + }, + { + "epoch": 1.36, + "learning_rate": 1.5473926018162693e-05, + "loss": 0.9813, + "step": 21192 + }, + { + "epoch": 1.36, + "learning_rate": 1.547371132913974e-05, + "loss": 1.0193, + "step": 21193 + }, + { + "epoch": 1.36, + "learning_rate": 1.547349664011679e-05, + "loss": 0.995, + "step": 21194 + }, + { + "epoch": 1.36, + "learning_rate": 1.5473281951093843e-05, + "loss": 0.988, + "step": 21195 + }, + { + "epoch": 1.36, + "learning_rate": 1.5473067262070892e-05, + "loss": 1.0016, + "step": 21196 + }, + { + "epoch": 1.36, + "learning_rate": 1.547285257304794e-05, + "loss": 1.0799, + "step": 21197 + }, + { + "epoch": 1.36, + "learning_rate": 1.547263788402499e-05, + "loss": 1.0323, + "step": 21198 + }, + { + "epoch": 1.36, + "learning_rate": 1.547242319500204e-05, + "loss": 0.9522, + "step": 21199 + }, + { + "epoch": 1.36, + "learning_rate": 1.547220850597909e-05, + "loss": 0.9541, + "step": 21200 + }, + { + "epoch": 1.36, + "learning_rate": 1.547199381695614e-05, + "loss": 0.9561, + "step": 21201 + }, + { + "epoch": 1.36, + "learning_rate": 1.5471779127933192e-05, + "loss": 0.9716, + "step": 21202 + }, + { + "epoch": 1.36, + "learning_rate": 1.547156443891024e-05, + "loss": 0.9466, + "step": 21203 + }, + { + "epoch": 1.36, + "learning_rate": 1.547134974988729e-05, + "loss": 0.9475, + "step": 21204 + }, + { + "epoch": 1.36, + "learning_rate": 1.547113506086434e-05, + "loss": 0.966, + "step": 21205 + }, + { + "epoch": 1.36, + "learning_rate": 1.5470920371841388e-05, + "loss": 0.9769, + "step": 21206 + }, + { + "epoch": 1.36, + "learning_rate": 1.5470705682818437e-05, + "loss": 0.9983, + "step": 21207 + }, + { + "epoch": 1.36, + "learning_rate": 1.547049099379549e-05, + "loss": 0.9185, + "step": 21208 + }, + { + "epoch": 1.36, + "learning_rate": 1.547027630477254e-05, + "loss": 1.0497, + "step": 21209 + }, + { + "epoch": 1.36, + "learning_rate": 1.5470061615749587e-05, + "loss": 0.978, + "step": 21210 + }, + { + "epoch": 1.36, + "learning_rate": 1.5469846926726636e-05, + "loss": 0.9769, + "step": 21211 + }, + { + "epoch": 1.36, + "learning_rate": 1.5469632237703685e-05, + "loss": 0.9187, + "step": 21212 + }, + { + "epoch": 1.36, + "learning_rate": 1.5469417548680738e-05, + "loss": 0.9546, + "step": 21213 + }, + { + "epoch": 1.36, + "learning_rate": 1.5469202859657787e-05, + "loss": 0.9758, + "step": 21214 + }, + { + "epoch": 1.36, + "learning_rate": 1.546898817063484e-05, + "loss": 0.9555, + "step": 21215 + }, + { + "epoch": 1.36, + "learning_rate": 1.5468773481611888e-05, + "loss": 0.9894, + "step": 21216 + }, + { + "epoch": 1.37, + "learning_rate": 1.5468558792588937e-05, + "loss": 0.9517, + "step": 21217 + }, + { + "epoch": 1.37, + "learning_rate": 1.5468344103565986e-05, + "loss": 0.952, + "step": 21218 + }, + { + "epoch": 1.37, + "learning_rate": 1.5468129414543035e-05, + "loss": 0.9545, + "step": 21219 + }, + { + "epoch": 1.37, + "learning_rate": 1.5467914725520087e-05, + "loss": 0.9812, + "step": 21220 + }, + { + "epoch": 1.37, + "learning_rate": 1.5467700036497136e-05, + "loss": 0.9307, + "step": 21221 + }, + { + "epoch": 1.37, + "learning_rate": 1.5467485347474185e-05, + "loss": 0.9478, + "step": 21222 + }, + { + "epoch": 1.37, + "learning_rate": 1.5467270658451234e-05, + "loss": 0.9585, + "step": 21223 + }, + { + "epoch": 1.37, + "learning_rate": 1.5467055969428283e-05, + "loss": 0.9773, + "step": 21224 + }, + { + "epoch": 1.37, + "learning_rate": 1.5466841280405332e-05, + "loss": 0.922, + "step": 21225 + }, + { + "epoch": 1.37, + "learning_rate": 1.5466626591382384e-05, + "loss": 0.9282, + "step": 21226 + }, + { + "epoch": 1.37, + "learning_rate": 1.5466411902359433e-05, + "loss": 0.9664, + "step": 21227 + }, + { + "epoch": 1.37, + "learning_rate": 1.5466197213336482e-05, + "loss": 1.097, + "step": 21228 + }, + { + "epoch": 1.37, + "learning_rate": 1.5465982524313535e-05, + "loss": 0.9843, + "step": 21229 + }, + { + "epoch": 1.37, + "learning_rate": 1.5465767835290584e-05, + "loss": 0.9997, + "step": 21230 + }, + { + "epoch": 1.37, + "learning_rate": 1.5465553146267633e-05, + "loss": 1.0528, + "step": 21231 + }, + { + "epoch": 1.37, + "learning_rate": 1.546533845724468e-05, + "loss": 0.8918, + "step": 21232 + }, + { + "epoch": 1.37, + "learning_rate": 1.5465123768221734e-05, + "loss": 0.9466, + "step": 21233 + }, + { + "epoch": 1.37, + "learning_rate": 1.5464909079198783e-05, + "loss": 0.9774, + "step": 21234 + }, + { + "epoch": 1.37, + "learning_rate": 1.5464694390175832e-05, + "loss": 1.0352, + "step": 21235 + }, + { + "epoch": 1.37, + "learning_rate": 1.546447970115288e-05, + "loss": 0.9131, + "step": 21236 + }, + { + "epoch": 1.37, + "learning_rate": 1.546426501212993e-05, + "loss": 0.9519, + "step": 21237 + }, + { + "epoch": 1.37, + "learning_rate": 1.5464050323106982e-05, + "loss": 0.9914, + "step": 21238 + }, + { + "epoch": 1.37, + "learning_rate": 1.546383563408403e-05, + "loss": 1.0316, + "step": 21239 + }, + { + "epoch": 1.37, + "learning_rate": 1.546362094506108e-05, + "loss": 0.9291, + "step": 21240 + }, + { + "epoch": 1.37, + "learning_rate": 1.546340625603813e-05, + "loss": 0.951, + "step": 21241 + }, + { + "epoch": 1.37, + "learning_rate": 1.5463191567015178e-05, + "loss": 0.9299, + "step": 21242 + }, + { + "epoch": 1.37, + "learning_rate": 1.5462976877992227e-05, + "loss": 1.0198, + "step": 21243 + }, + { + "epoch": 1.37, + "learning_rate": 1.546276218896928e-05, + "loss": 1.0357, + "step": 21244 + }, + { + "epoch": 1.37, + "learning_rate": 1.5462547499946328e-05, + "loss": 1.0063, + "step": 21245 + }, + { + "epoch": 1.37, + "learning_rate": 1.546233281092338e-05, + "loss": 0.9994, + "step": 21246 + }, + { + "epoch": 1.37, + "learning_rate": 1.546211812190043e-05, + "loss": 0.9587, + "step": 21247 + }, + { + "epoch": 1.37, + "learning_rate": 1.546190343287748e-05, + "loss": 1.0141, + "step": 21248 + }, + { + "epoch": 1.37, + "learning_rate": 1.5461688743854528e-05, + "loss": 0.9156, + "step": 21249 + }, + { + "epoch": 1.37, + "learning_rate": 1.5461474054831577e-05, + "loss": 1.0108, + "step": 21250 + }, + { + "epoch": 1.37, + "learning_rate": 1.546125936580863e-05, + "loss": 1.0332, + "step": 21251 + }, + { + "epoch": 1.37, + "learning_rate": 1.5461044676785678e-05, + "loss": 0.9663, + "step": 21252 + }, + { + "epoch": 1.37, + "learning_rate": 1.5460829987762727e-05, + "loss": 0.9946, + "step": 21253 + }, + { + "epoch": 1.37, + "learning_rate": 1.5460615298739776e-05, + "loss": 0.9795, + "step": 21254 + }, + { + "epoch": 1.37, + "learning_rate": 1.5460400609716825e-05, + "loss": 0.9495, + "step": 21255 + }, + { + "epoch": 1.37, + "learning_rate": 1.5460185920693874e-05, + "loss": 0.9689, + "step": 21256 + }, + { + "epoch": 1.37, + "learning_rate": 1.5459971231670926e-05, + "loss": 1.0289, + "step": 21257 + }, + { + "epoch": 1.37, + "learning_rate": 1.545975654264798e-05, + "loss": 0.9388, + "step": 21258 + }, + { + "epoch": 1.37, + "learning_rate": 1.5459541853625027e-05, + "loss": 0.9403, + "step": 21259 + }, + { + "epoch": 1.37, + "learning_rate": 1.5459327164602076e-05, + "loss": 0.8957, + "step": 21260 + }, + { + "epoch": 1.37, + "learning_rate": 1.5459112475579125e-05, + "loss": 0.97, + "step": 21261 + }, + { + "epoch": 1.37, + "learning_rate": 1.5458897786556174e-05, + "loss": 0.9523, + "step": 21262 + }, + { + "epoch": 1.37, + "learning_rate": 1.5458683097533223e-05, + "loss": 0.9261, + "step": 21263 + }, + { + "epoch": 1.37, + "learning_rate": 1.5458468408510276e-05, + "loss": 0.9801, + "step": 21264 + }, + { + "epoch": 1.37, + "learning_rate": 1.5458253719487325e-05, + "loss": 0.98, + "step": 21265 + }, + { + "epoch": 1.37, + "learning_rate": 1.5458039030464374e-05, + "loss": 1.0236, + "step": 21266 + }, + { + "epoch": 1.37, + "learning_rate": 1.5457824341441422e-05, + "loss": 0.9651, + "step": 21267 + }, + { + "epoch": 1.37, + "learning_rate": 1.545760965241847e-05, + "loss": 0.9568, + "step": 21268 + }, + { + "epoch": 1.37, + "learning_rate": 1.5457394963395524e-05, + "loss": 1.0198, + "step": 21269 + }, + { + "epoch": 1.37, + "learning_rate": 1.5457180274372573e-05, + "loss": 1.0302, + "step": 21270 + }, + { + "epoch": 1.37, + "learning_rate": 1.5456965585349622e-05, + "loss": 1.0014, + "step": 21271 + }, + { + "epoch": 1.37, + "learning_rate": 1.545675089632667e-05, + "loss": 0.9319, + "step": 21272 + }, + { + "epoch": 1.37, + "learning_rate": 1.5456536207303723e-05, + "loss": 0.9388, + "step": 21273 + }, + { + "epoch": 1.37, + "learning_rate": 1.5456321518280772e-05, + "loss": 0.9375, + "step": 21274 + }, + { + "epoch": 1.37, + "learning_rate": 1.545610682925782e-05, + "loss": 0.978, + "step": 21275 + }, + { + "epoch": 1.37, + "learning_rate": 1.545589214023487e-05, + "loss": 0.952, + "step": 21276 + }, + { + "epoch": 1.37, + "learning_rate": 1.5455677451211922e-05, + "loss": 1.0104, + "step": 21277 + }, + { + "epoch": 1.37, + "learning_rate": 1.545546276218897e-05, + "loss": 0.9191, + "step": 21278 + }, + { + "epoch": 1.37, + "learning_rate": 1.545524807316602e-05, + "loss": 0.9439, + "step": 21279 + }, + { + "epoch": 1.37, + "learning_rate": 1.545503338414307e-05, + "loss": 1.0278, + "step": 21280 + }, + { + "epoch": 1.37, + "learning_rate": 1.5454818695120118e-05, + "loss": 0.9697, + "step": 21281 + }, + { + "epoch": 1.37, + "learning_rate": 1.545460400609717e-05, + "loss": 0.9693, + "step": 21282 + }, + { + "epoch": 1.37, + "learning_rate": 1.545438931707422e-05, + "loss": 1.044, + "step": 21283 + }, + { + "epoch": 1.37, + "learning_rate": 1.545417462805127e-05, + "loss": 0.8975, + "step": 21284 + }, + { + "epoch": 1.37, + "learning_rate": 1.5453959939028317e-05, + "loss": 0.9992, + "step": 21285 + }, + { + "epoch": 1.37, + "learning_rate": 1.5453745250005366e-05, + "loss": 0.9525, + "step": 21286 + }, + { + "epoch": 1.37, + "learning_rate": 1.545353056098242e-05, + "loss": 1.0361, + "step": 21287 + }, + { + "epoch": 1.37, + "learning_rate": 1.5453315871959468e-05, + "loss": 1.0736, + "step": 21288 + }, + { + "epoch": 1.37, + "learning_rate": 1.545310118293652e-05, + "loss": 1.0025, + "step": 21289 + }, + { + "epoch": 1.37, + "learning_rate": 1.545288649391357e-05, + "loss": 0.975, + "step": 21290 + }, + { + "epoch": 1.37, + "learning_rate": 1.5452671804890618e-05, + "loss": 0.9758, + "step": 21291 + }, + { + "epoch": 1.37, + "learning_rate": 1.5452457115867667e-05, + "loss": 1.0126, + "step": 21292 + }, + { + "epoch": 1.37, + "learning_rate": 1.5452242426844716e-05, + "loss": 0.9432, + "step": 21293 + }, + { + "epoch": 1.37, + "learning_rate": 1.5452027737821765e-05, + "loss": 0.9481, + "step": 21294 + }, + { + "epoch": 1.37, + "learning_rate": 1.5451813048798817e-05, + "loss": 0.9282, + "step": 21295 + }, + { + "epoch": 1.37, + "learning_rate": 1.5451598359775866e-05, + "loss": 0.9847, + "step": 21296 + }, + { + "epoch": 1.37, + "learning_rate": 1.5451383670752915e-05, + "loss": 1.0064, + "step": 21297 + }, + { + "epoch": 1.37, + "learning_rate": 1.5451168981729964e-05, + "loss": 0.9988, + "step": 21298 + }, + { + "epoch": 1.37, + "learning_rate": 1.5450954292707013e-05, + "loss": 0.9252, + "step": 21299 + }, + { + "epoch": 1.37, + "learning_rate": 1.5450739603684065e-05, + "loss": 0.9963, + "step": 21300 + }, + { + "epoch": 1.37, + "learning_rate": 1.5450524914661114e-05, + "loss": 1.0195, + "step": 21301 + }, + { + "epoch": 1.37, + "learning_rate": 1.5450310225638167e-05, + "loss": 0.9273, + "step": 21302 + }, + { + "epoch": 1.37, + "learning_rate": 1.5450095536615216e-05, + "loss": 0.8981, + "step": 21303 + }, + { + "epoch": 1.37, + "learning_rate": 1.5449880847592265e-05, + "loss": 1.0057, + "step": 21304 + }, + { + "epoch": 1.37, + "learning_rate": 1.5449666158569314e-05, + "loss": 0.9625, + "step": 21305 + }, + { + "epoch": 1.37, + "learning_rate": 1.5449451469546363e-05, + "loss": 0.997, + "step": 21306 + }, + { + "epoch": 1.37, + "learning_rate": 1.5449236780523415e-05, + "loss": 0.9764, + "step": 21307 + }, + { + "epoch": 1.37, + "learning_rate": 1.5449022091500464e-05, + "loss": 1.083, + "step": 21308 + }, + { + "epoch": 1.37, + "learning_rate": 1.5448807402477513e-05, + "loss": 0.9229, + "step": 21309 + }, + { + "epoch": 1.37, + "learning_rate": 1.5448592713454562e-05, + "loss": 0.9757, + "step": 21310 + }, + { + "epoch": 1.37, + "learning_rate": 1.544837802443161e-05, + "loss": 0.9783, + "step": 21311 + }, + { + "epoch": 1.37, + "learning_rate": 1.544816333540866e-05, + "loss": 0.9454, + "step": 21312 + }, + { + "epoch": 1.37, + "learning_rate": 1.5447948646385712e-05, + "loss": 0.9014, + "step": 21313 + }, + { + "epoch": 1.37, + "learning_rate": 1.544773395736276e-05, + "loss": 1.0328, + "step": 21314 + }, + { + "epoch": 1.37, + "learning_rate": 1.544751926833981e-05, + "loss": 1.0455, + "step": 21315 + }, + { + "epoch": 1.37, + "learning_rate": 1.5447304579316862e-05, + "loss": 0.976, + "step": 21316 + }, + { + "epoch": 1.37, + "learning_rate": 1.544708989029391e-05, + "loss": 1.0036, + "step": 21317 + }, + { + "epoch": 1.37, + "learning_rate": 1.544687520127096e-05, + "loss": 1.0467, + "step": 21318 + }, + { + "epoch": 1.37, + "learning_rate": 1.544666051224801e-05, + "loss": 0.9676, + "step": 21319 + }, + { + "epoch": 1.37, + "learning_rate": 1.544644582322506e-05, + "loss": 1.0846, + "step": 21320 + }, + { + "epoch": 1.37, + "learning_rate": 1.544623113420211e-05, + "loss": 0.9748, + "step": 21321 + }, + { + "epoch": 1.37, + "learning_rate": 1.544601644517916e-05, + "loss": 1.0065, + "step": 21322 + }, + { + "epoch": 1.37, + "learning_rate": 1.544580175615621e-05, + "loss": 0.9499, + "step": 21323 + }, + { + "epoch": 1.37, + "learning_rate": 1.5445587067133258e-05, + "loss": 1.0509, + "step": 21324 + }, + { + "epoch": 1.37, + "learning_rate": 1.544537237811031e-05, + "loss": 0.9203, + "step": 21325 + }, + { + "epoch": 1.37, + "learning_rate": 1.544515768908736e-05, + "loss": 0.9596, + "step": 21326 + }, + { + "epoch": 1.37, + "learning_rate": 1.5444943000064408e-05, + "loss": 0.9753, + "step": 21327 + }, + { + "epoch": 1.37, + "learning_rate": 1.5444728311041457e-05, + "loss": 0.9244, + "step": 21328 + }, + { + "epoch": 1.37, + "learning_rate": 1.5444513622018506e-05, + "loss": 1.0023, + "step": 21329 + }, + { + "epoch": 1.37, + "learning_rate": 1.5444298932995555e-05, + "loss": 0.9207, + "step": 21330 + }, + { + "epoch": 1.37, + "learning_rate": 1.5444084243972607e-05, + "loss": 0.9124, + "step": 21331 + }, + { + "epoch": 1.37, + "learning_rate": 1.5443869554949656e-05, + "loss": 1.0297, + "step": 21332 + }, + { + "epoch": 1.37, + "learning_rate": 1.544365486592671e-05, + "loss": 0.982, + "step": 21333 + }, + { + "epoch": 1.37, + "learning_rate": 1.5443440176903757e-05, + "loss": 1.0042, + "step": 21334 + }, + { + "epoch": 1.37, + "learning_rate": 1.5443225487880806e-05, + "loss": 0.9615, + "step": 21335 + }, + { + "epoch": 1.37, + "learning_rate": 1.5443010798857855e-05, + "loss": 0.9766, + "step": 21336 + }, + { + "epoch": 1.37, + "learning_rate": 1.5442796109834904e-05, + "loss": 0.9608, + "step": 21337 + }, + { + "epoch": 1.37, + "learning_rate": 1.5442581420811957e-05, + "loss": 0.9611, + "step": 21338 + }, + { + "epoch": 1.37, + "learning_rate": 1.5442366731789006e-05, + "loss": 0.9492, + "step": 21339 + }, + { + "epoch": 1.37, + "learning_rate": 1.5442152042766055e-05, + "loss": 1.0065, + "step": 21340 + }, + { + "epoch": 1.37, + "learning_rate": 1.5441937353743103e-05, + "loss": 1.013, + "step": 21341 + }, + { + "epoch": 1.37, + "learning_rate": 1.5441722664720152e-05, + "loss": 0.9315, + "step": 21342 + }, + { + "epoch": 1.37, + "learning_rate": 1.54415079756972e-05, + "loss": 0.9263, + "step": 21343 + }, + { + "epoch": 1.37, + "learning_rate": 1.5441293286674254e-05, + "loss": 0.9967, + "step": 21344 + }, + { + "epoch": 1.37, + "learning_rate": 1.5441078597651306e-05, + "loss": 0.9436, + "step": 21345 + }, + { + "epoch": 1.37, + "learning_rate": 1.5440863908628355e-05, + "loss": 0.9567, + "step": 21346 + }, + { + "epoch": 1.37, + "learning_rate": 1.5440649219605404e-05, + "loss": 0.9453, + "step": 21347 + }, + { + "epoch": 1.37, + "learning_rate": 1.5440434530582453e-05, + "loss": 0.8983, + "step": 21348 + }, + { + "epoch": 1.37, + "learning_rate": 1.5440219841559502e-05, + "loss": 0.9435, + "step": 21349 + }, + { + "epoch": 1.37, + "learning_rate": 1.544000515253655e-05, + "loss": 0.9992, + "step": 21350 + }, + { + "epoch": 1.37, + "learning_rate": 1.5439790463513603e-05, + "loss": 0.9534, + "step": 21351 + }, + { + "epoch": 1.37, + "learning_rate": 1.5439575774490652e-05, + "loss": 0.9649, + "step": 21352 + }, + { + "epoch": 1.37, + "learning_rate": 1.54393610854677e-05, + "loss": 0.9802, + "step": 21353 + }, + { + "epoch": 1.37, + "learning_rate": 1.543914639644475e-05, + "loss": 0.9176, + "step": 21354 + }, + { + "epoch": 1.37, + "learning_rate": 1.54389317074218e-05, + "loss": 1.0119, + "step": 21355 + }, + { + "epoch": 1.37, + "learning_rate": 1.543871701839885e-05, + "loss": 0.9148, + "step": 21356 + }, + { + "epoch": 1.37, + "learning_rate": 1.54385023293759e-05, + "loss": 0.9117, + "step": 21357 + }, + { + "epoch": 1.37, + "learning_rate": 1.543828764035295e-05, + "loss": 1.0048, + "step": 21358 + }, + { + "epoch": 1.37, + "learning_rate": 1.543807295133e-05, + "loss": 1.0128, + "step": 21359 + }, + { + "epoch": 1.37, + "learning_rate": 1.543785826230705e-05, + "loss": 1.065, + "step": 21360 + }, + { + "epoch": 1.37, + "learning_rate": 1.54376435732841e-05, + "loss": 1.019, + "step": 21361 + }, + { + "epoch": 1.37, + "learning_rate": 1.543742888426115e-05, + "loss": 0.9717, + "step": 21362 + }, + { + "epoch": 1.37, + "learning_rate": 1.5437214195238198e-05, + "loss": 0.9972, + "step": 21363 + }, + { + "epoch": 1.37, + "learning_rate": 1.543699950621525e-05, + "loss": 0.9126, + "step": 21364 + }, + { + "epoch": 1.37, + "learning_rate": 1.54367848171923e-05, + "loss": 1.0339, + "step": 21365 + }, + { + "epoch": 1.37, + "learning_rate": 1.5436570128169348e-05, + "loss": 0.9716, + "step": 21366 + }, + { + "epoch": 1.37, + "learning_rate": 1.5436355439146397e-05, + "loss": 0.9092, + "step": 21367 + }, + { + "epoch": 1.37, + "learning_rate": 1.5436140750123446e-05, + "loss": 0.978, + "step": 21368 + }, + { + "epoch": 1.37, + "learning_rate": 1.5435926061100498e-05, + "loss": 1.0232, + "step": 21369 + }, + { + "epoch": 1.37, + "learning_rate": 1.5435711372077547e-05, + "loss": 1.0255, + "step": 21370 + }, + { + "epoch": 1.37, + "learning_rate": 1.5435496683054596e-05, + "loss": 1.0156, + "step": 21371 + }, + { + "epoch": 1.38, + "learning_rate": 1.5435281994031645e-05, + "loss": 0.9777, + "step": 21372 + }, + { + "epoch": 1.38, + "learning_rate": 1.5435067305008694e-05, + "loss": 0.8857, + "step": 21373 + }, + { + "epoch": 1.38, + "learning_rate": 1.5434852615985746e-05, + "loss": 0.8854, + "step": 21374 + }, + { + "epoch": 1.38, + "learning_rate": 1.5434637926962795e-05, + "loss": 0.9816, + "step": 21375 + }, + { + "epoch": 1.38, + "learning_rate": 1.5434423237939848e-05, + "loss": 0.9631, + "step": 21376 + }, + { + "epoch": 1.38, + "learning_rate": 1.5434208548916897e-05, + "loss": 0.9405, + "step": 21377 + }, + { + "epoch": 1.38, + "learning_rate": 1.5433993859893946e-05, + "loss": 0.9118, + "step": 21378 + }, + { + "epoch": 1.38, + "learning_rate": 1.5433779170870995e-05, + "loss": 0.9655, + "step": 21379 + }, + { + "epoch": 1.38, + "learning_rate": 1.5433564481848044e-05, + "loss": 0.9895, + "step": 21380 + }, + { + "epoch": 1.38, + "learning_rate": 1.5433349792825093e-05, + "loss": 0.9557, + "step": 21381 + }, + { + "epoch": 1.38, + "learning_rate": 1.5433135103802145e-05, + "loss": 0.9342, + "step": 21382 + }, + { + "epoch": 1.38, + "learning_rate": 1.5432920414779194e-05, + "loss": 0.9997, + "step": 21383 + }, + { + "epoch": 1.38, + "learning_rate": 1.5432705725756243e-05, + "loss": 0.9861, + "step": 21384 + }, + { + "epoch": 1.38, + "learning_rate": 1.5432491036733292e-05, + "loss": 1.0019, + "step": 21385 + }, + { + "epoch": 1.38, + "learning_rate": 1.543227634771034e-05, + "loss": 0.9583, + "step": 21386 + }, + { + "epoch": 1.38, + "learning_rate": 1.5432061658687393e-05, + "loss": 0.9711, + "step": 21387 + }, + { + "epoch": 1.38, + "learning_rate": 1.5431846969664442e-05, + "loss": 0.9665, + "step": 21388 + }, + { + "epoch": 1.38, + "learning_rate": 1.5431632280641494e-05, + "loss": 0.9609, + "step": 21389 + }, + { + "epoch": 1.38, + "learning_rate": 1.5431417591618543e-05, + "loss": 1.0218, + "step": 21390 + }, + { + "epoch": 1.38, + "learning_rate": 1.5431202902595592e-05, + "loss": 1.005, + "step": 21391 + }, + { + "epoch": 1.38, + "learning_rate": 1.543098821357264e-05, + "loss": 0.9965, + "step": 21392 + }, + { + "epoch": 1.38, + "learning_rate": 1.543077352454969e-05, + "loss": 0.9813, + "step": 21393 + }, + { + "epoch": 1.38, + "learning_rate": 1.5430558835526743e-05, + "loss": 0.9219, + "step": 21394 + }, + { + "epoch": 1.38, + "learning_rate": 1.543034414650379e-05, + "loss": 0.9418, + "step": 21395 + }, + { + "epoch": 1.38, + "learning_rate": 1.543012945748084e-05, + "loss": 0.9376, + "step": 21396 + }, + { + "epoch": 1.38, + "learning_rate": 1.542991476845789e-05, + "loss": 0.9228, + "step": 21397 + }, + { + "epoch": 1.38, + "learning_rate": 1.542970007943494e-05, + "loss": 0.8757, + "step": 21398 + }, + { + "epoch": 1.38, + "learning_rate": 1.5429485390411987e-05, + "loss": 0.9215, + "step": 21399 + }, + { + "epoch": 1.38, + "learning_rate": 1.542927070138904e-05, + "loss": 0.9321, + "step": 21400 + }, + { + "epoch": 1.38, + "learning_rate": 1.542905601236609e-05, + "loss": 0.9766, + "step": 21401 + }, + { + "epoch": 1.38, + "learning_rate": 1.5428841323343138e-05, + "loss": 0.9827, + "step": 21402 + }, + { + "epoch": 1.38, + "learning_rate": 1.5428626634320187e-05, + "loss": 0.9005, + "step": 21403 + }, + { + "epoch": 1.38, + "learning_rate": 1.542841194529724e-05, + "loss": 0.8947, + "step": 21404 + }, + { + "epoch": 1.38, + "learning_rate": 1.5428197256274288e-05, + "loss": 1.0365, + "step": 21405 + }, + { + "epoch": 1.38, + "learning_rate": 1.5427982567251337e-05, + "loss": 0.9513, + "step": 21406 + }, + { + "epoch": 1.38, + "learning_rate": 1.542776787822839e-05, + "loss": 0.9439, + "step": 21407 + }, + { + "epoch": 1.38, + "learning_rate": 1.542755318920544e-05, + "loss": 0.9835, + "step": 21408 + }, + { + "epoch": 1.38, + "learning_rate": 1.5427338500182487e-05, + "loss": 0.9631, + "step": 21409 + }, + { + "epoch": 1.38, + "learning_rate": 1.5427123811159536e-05, + "loss": 0.9831, + "step": 21410 + }, + { + "epoch": 1.38, + "learning_rate": 1.5426909122136585e-05, + "loss": 1.0339, + "step": 21411 + }, + { + "epoch": 1.38, + "learning_rate": 1.5426694433113638e-05, + "loss": 0.9401, + "step": 21412 + }, + { + "epoch": 1.38, + "learning_rate": 1.5426479744090687e-05, + "loss": 0.8925, + "step": 21413 + }, + { + "epoch": 1.38, + "learning_rate": 1.5426265055067736e-05, + "loss": 0.9417, + "step": 21414 + }, + { + "epoch": 1.38, + "learning_rate": 1.5426050366044784e-05, + "loss": 1.0239, + "step": 21415 + }, + { + "epoch": 1.38, + "learning_rate": 1.5425835677021833e-05, + "loss": 1.0122, + "step": 21416 + }, + { + "epoch": 1.38, + "learning_rate": 1.5425620987998882e-05, + "loss": 0.9737, + "step": 21417 + }, + { + "epoch": 1.38, + "learning_rate": 1.5425406298975935e-05, + "loss": 0.9957, + "step": 21418 + }, + { + "epoch": 1.38, + "learning_rate": 1.5425191609952984e-05, + "loss": 0.9822, + "step": 21419 + }, + { + "epoch": 1.38, + "learning_rate": 1.5424976920930036e-05, + "loss": 1.0609, + "step": 21420 + }, + { + "epoch": 1.38, + "learning_rate": 1.5424762231907085e-05, + "loss": 0.9219, + "step": 21421 + }, + { + "epoch": 1.38, + "learning_rate": 1.5424547542884134e-05, + "loss": 0.9742, + "step": 21422 + }, + { + "epoch": 1.38, + "learning_rate": 1.5424332853861183e-05, + "loss": 1.0208, + "step": 21423 + }, + { + "epoch": 1.38, + "learning_rate": 1.5424118164838232e-05, + "loss": 0.9729, + "step": 21424 + }, + { + "epoch": 1.38, + "learning_rate": 1.5423903475815284e-05, + "loss": 0.9158, + "step": 21425 + }, + { + "epoch": 1.38, + "learning_rate": 1.5423688786792333e-05, + "loss": 1.0156, + "step": 21426 + }, + { + "epoch": 1.38, + "learning_rate": 1.5423474097769382e-05, + "loss": 0.9832, + "step": 21427 + }, + { + "epoch": 1.38, + "learning_rate": 1.542325940874643e-05, + "loss": 0.9737, + "step": 21428 + }, + { + "epoch": 1.38, + "learning_rate": 1.542304471972348e-05, + "loss": 1.0028, + "step": 21429 + }, + { + "epoch": 1.38, + "learning_rate": 1.542283003070053e-05, + "loss": 0.9267, + "step": 21430 + }, + { + "epoch": 1.38, + "learning_rate": 1.542261534167758e-05, + "loss": 0.9736, + "step": 21431 + }, + { + "epoch": 1.38, + "learning_rate": 1.542240065265463e-05, + "loss": 1.0294, + "step": 21432 + }, + { + "epoch": 1.38, + "learning_rate": 1.5422185963631683e-05, + "loss": 1.0011, + "step": 21433 + }, + { + "epoch": 1.38, + "learning_rate": 1.5421971274608732e-05, + "loss": 0.9661, + "step": 21434 + }, + { + "epoch": 1.38, + "learning_rate": 1.542175658558578e-05, + "loss": 0.8888, + "step": 21435 + }, + { + "epoch": 1.38, + "learning_rate": 1.542154189656283e-05, + "loss": 0.9811, + "step": 21436 + }, + { + "epoch": 1.38, + "learning_rate": 1.542132720753988e-05, + "loss": 0.9576, + "step": 21437 + }, + { + "epoch": 1.38, + "learning_rate": 1.542111251851693e-05, + "loss": 0.9303, + "step": 21438 + }, + { + "epoch": 1.38, + "learning_rate": 1.542089782949398e-05, + "loss": 0.9547, + "step": 21439 + }, + { + "epoch": 1.38, + "learning_rate": 1.542068314047103e-05, + "loss": 1.0098, + "step": 21440 + }, + { + "epoch": 1.38, + "learning_rate": 1.5420468451448078e-05, + "loss": 1.0574, + "step": 21441 + }, + { + "epoch": 1.38, + "learning_rate": 1.5420253762425127e-05, + "loss": 0.9764, + "step": 21442 + }, + { + "epoch": 1.38, + "learning_rate": 1.542003907340218e-05, + "loss": 1.024, + "step": 21443 + }, + { + "epoch": 1.38, + "learning_rate": 1.5419824384379228e-05, + "loss": 0.9178, + "step": 21444 + }, + { + "epoch": 1.38, + "learning_rate": 1.5419609695356277e-05, + "loss": 0.9574, + "step": 21445 + }, + { + "epoch": 1.38, + "learning_rate": 1.5419395006333326e-05, + "loss": 0.9686, + "step": 21446 + }, + { + "epoch": 1.38, + "learning_rate": 1.541918031731038e-05, + "loss": 0.9454, + "step": 21447 + }, + { + "epoch": 1.38, + "learning_rate": 1.5418965628287427e-05, + "loss": 0.9488, + "step": 21448 + }, + { + "epoch": 1.38, + "learning_rate": 1.5418750939264476e-05, + "loss": 0.9694, + "step": 21449 + }, + { + "epoch": 1.38, + "learning_rate": 1.5418536250241525e-05, + "loss": 0.9962, + "step": 21450 + }, + { + "epoch": 1.38, + "learning_rate": 1.5418321561218578e-05, + "loss": 1.023, + "step": 21451 + }, + { + "epoch": 1.38, + "learning_rate": 1.5418106872195627e-05, + "loss": 0.9865, + "step": 21452 + }, + { + "epoch": 1.38, + "learning_rate": 1.5417892183172676e-05, + "loss": 1.0554, + "step": 21453 + }, + { + "epoch": 1.38, + "learning_rate": 1.5417677494149725e-05, + "loss": 0.9975, + "step": 21454 + }, + { + "epoch": 1.38, + "learning_rate": 1.5417462805126774e-05, + "loss": 1.0018, + "step": 21455 + }, + { + "epoch": 1.38, + "learning_rate": 1.5417248116103826e-05, + "loss": 0.9367, + "step": 21456 + }, + { + "epoch": 1.38, + "learning_rate": 1.5417033427080875e-05, + "loss": 0.9019, + "step": 21457 + }, + { + "epoch": 1.38, + "learning_rate": 1.5416818738057924e-05, + "loss": 1.0359, + "step": 21458 + }, + { + "epoch": 1.38, + "learning_rate": 1.5416604049034973e-05, + "loss": 1.007, + "step": 21459 + }, + { + "epoch": 1.38, + "learning_rate": 1.5416389360012022e-05, + "loss": 0.9941, + "step": 21460 + }, + { + "epoch": 1.38, + "learning_rate": 1.5416174670989074e-05, + "loss": 1.0196, + "step": 21461 + }, + { + "epoch": 1.38, + "learning_rate": 1.5415959981966123e-05, + "loss": 0.9468, + "step": 21462 + }, + { + "epoch": 1.38, + "learning_rate": 1.5415745292943175e-05, + "loss": 0.9201, + "step": 21463 + }, + { + "epoch": 1.38, + "learning_rate": 1.5415530603920224e-05, + "loss": 1.014, + "step": 21464 + }, + { + "epoch": 1.38, + "learning_rate": 1.5415315914897273e-05, + "loss": 0.9928, + "step": 21465 + }, + { + "epoch": 1.38, + "learning_rate": 1.5415101225874322e-05, + "loss": 0.9591, + "step": 21466 + }, + { + "epoch": 1.38, + "learning_rate": 1.541488653685137e-05, + "loss": 0.9569, + "step": 21467 + }, + { + "epoch": 1.38, + "learning_rate": 1.541467184782842e-05, + "loss": 0.9921, + "step": 21468 + }, + { + "epoch": 1.38, + "learning_rate": 1.5414457158805473e-05, + "loss": 0.9627, + "step": 21469 + }, + { + "epoch": 1.38, + "learning_rate": 1.541424246978252e-05, + "loss": 0.9698, + "step": 21470 + }, + { + "epoch": 1.38, + "learning_rate": 1.541402778075957e-05, + "loss": 0.9677, + "step": 21471 + }, + { + "epoch": 1.38, + "learning_rate": 1.541381309173662e-05, + "loss": 0.9067, + "step": 21472 + }, + { + "epoch": 1.38, + "learning_rate": 1.541359840271367e-05, + "loss": 0.9169, + "step": 21473 + }, + { + "epoch": 1.38, + "learning_rate": 1.541338371369072e-05, + "loss": 0.9713, + "step": 21474 + }, + { + "epoch": 1.38, + "learning_rate": 1.541316902466777e-05, + "loss": 0.9898, + "step": 21475 + }, + { + "epoch": 1.38, + "learning_rate": 1.5412954335644822e-05, + "loss": 1.0012, + "step": 21476 + }, + { + "epoch": 1.38, + "learning_rate": 1.541273964662187e-05, + "loss": 0.9108, + "step": 21477 + }, + { + "epoch": 1.38, + "learning_rate": 1.541252495759892e-05, + "loss": 1.0011, + "step": 21478 + }, + { + "epoch": 1.38, + "learning_rate": 1.541231026857597e-05, + "loss": 0.9058, + "step": 21479 + }, + { + "epoch": 1.38, + "learning_rate": 1.5412095579553018e-05, + "loss": 0.9851, + "step": 21480 + }, + { + "epoch": 1.38, + "learning_rate": 1.541188089053007e-05, + "loss": 0.9071, + "step": 21481 + }, + { + "epoch": 1.38, + "learning_rate": 1.541166620150712e-05, + "loss": 0.9909, + "step": 21482 + }, + { + "epoch": 1.38, + "learning_rate": 1.5411451512484168e-05, + "loss": 0.9647, + "step": 21483 + }, + { + "epoch": 1.38, + "learning_rate": 1.5411236823461217e-05, + "loss": 1.048, + "step": 21484 + }, + { + "epoch": 1.38, + "learning_rate": 1.5411022134438266e-05, + "loss": 0.9259, + "step": 21485 + }, + { + "epoch": 1.38, + "learning_rate": 1.5410807445415315e-05, + "loss": 0.9841, + "step": 21486 + }, + { + "epoch": 1.38, + "learning_rate": 1.5410592756392368e-05, + "loss": 0.9354, + "step": 21487 + }, + { + "epoch": 1.38, + "learning_rate": 1.5410378067369417e-05, + "loss": 1.0262, + "step": 21488 + }, + { + "epoch": 1.38, + "learning_rate": 1.5410163378346465e-05, + "loss": 1.0512, + "step": 21489 + }, + { + "epoch": 1.38, + "learning_rate": 1.5409948689323514e-05, + "loss": 0.9598, + "step": 21490 + }, + { + "epoch": 1.38, + "learning_rate": 1.5409734000300567e-05, + "loss": 1.0534, + "step": 21491 + }, + { + "epoch": 1.38, + "learning_rate": 1.5409519311277616e-05, + "loss": 0.9694, + "step": 21492 + }, + { + "epoch": 1.38, + "learning_rate": 1.5409304622254665e-05, + "loss": 0.9806, + "step": 21493 + }, + { + "epoch": 1.38, + "learning_rate": 1.5409089933231717e-05, + "loss": 0.9906, + "step": 21494 + }, + { + "epoch": 1.38, + "learning_rate": 1.5408875244208766e-05, + "loss": 0.9975, + "step": 21495 + }, + { + "epoch": 1.38, + "learning_rate": 1.5408660555185815e-05, + "loss": 0.905, + "step": 21496 + }, + { + "epoch": 1.38, + "learning_rate": 1.5408445866162864e-05, + "loss": 0.9783, + "step": 21497 + }, + { + "epoch": 1.38, + "learning_rate": 1.5408231177139913e-05, + "loss": 0.9507, + "step": 21498 + }, + { + "epoch": 1.38, + "learning_rate": 1.5408016488116962e-05, + "loss": 0.9985, + "step": 21499 + }, + { + "epoch": 1.38, + "learning_rate": 1.5407801799094014e-05, + "loss": 0.9621, + "step": 21500 + }, + { + "epoch": 1.38, + "learning_rate": 1.5407587110071063e-05, + "loss": 0.9186, + "step": 21501 + }, + { + "epoch": 1.38, + "learning_rate": 1.5407372421048112e-05, + "loss": 1.0521, + "step": 21502 + }, + { + "epoch": 1.38, + "learning_rate": 1.540715773202516e-05, + "loss": 0.8633, + "step": 21503 + }, + { + "epoch": 1.38, + "learning_rate": 1.540694304300221e-05, + "loss": 0.9422, + "step": 21504 + }, + { + "epoch": 1.38, + "learning_rate": 1.5406728353979262e-05, + "loss": 0.8967, + "step": 21505 + }, + { + "epoch": 1.38, + "learning_rate": 1.540651366495631e-05, + "loss": 0.9915, + "step": 21506 + }, + { + "epoch": 1.38, + "learning_rate": 1.5406298975933364e-05, + "loss": 0.9418, + "step": 21507 + }, + { + "epoch": 1.38, + "learning_rate": 1.5406084286910413e-05, + "loss": 0.9949, + "step": 21508 + }, + { + "epoch": 1.38, + "learning_rate": 1.5405869597887462e-05, + "loss": 0.9295, + "step": 21509 + }, + { + "epoch": 1.38, + "learning_rate": 1.540565490886451e-05, + "loss": 0.9637, + "step": 21510 + }, + { + "epoch": 1.38, + "learning_rate": 1.540544021984156e-05, + "loss": 1.0456, + "step": 21511 + }, + { + "epoch": 1.38, + "learning_rate": 1.5405225530818612e-05, + "loss": 0.9311, + "step": 21512 + }, + { + "epoch": 1.38, + "learning_rate": 1.540501084179566e-05, + "loss": 0.9648, + "step": 21513 + }, + { + "epoch": 1.38, + "learning_rate": 1.540479615277271e-05, + "loss": 0.9967, + "step": 21514 + }, + { + "epoch": 1.38, + "learning_rate": 1.540458146374976e-05, + "loss": 0.9286, + "step": 21515 + }, + { + "epoch": 1.38, + "learning_rate": 1.5404366774726808e-05, + "loss": 0.9766, + "step": 21516 + }, + { + "epoch": 1.38, + "learning_rate": 1.5404152085703857e-05, + "loss": 0.9599, + "step": 21517 + }, + { + "epoch": 1.38, + "learning_rate": 1.540393739668091e-05, + "loss": 1.0173, + "step": 21518 + }, + { + "epoch": 1.38, + "learning_rate": 1.5403722707657958e-05, + "loss": 0.9867, + "step": 21519 + }, + { + "epoch": 1.38, + "learning_rate": 1.540350801863501e-05, + "loss": 0.9861, + "step": 21520 + }, + { + "epoch": 1.38, + "learning_rate": 1.540329332961206e-05, + "loss": 1.017, + "step": 21521 + }, + { + "epoch": 1.38, + "learning_rate": 1.540307864058911e-05, + "loss": 1.0051, + "step": 21522 + }, + { + "epoch": 1.38, + "learning_rate": 1.5402863951566157e-05, + "loss": 0.9787, + "step": 21523 + }, + { + "epoch": 1.38, + "learning_rate": 1.5402649262543206e-05, + "loss": 0.9212, + "step": 21524 + }, + { + "epoch": 1.38, + "learning_rate": 1.540243457352026e-05, + "loss": 0.964, + "step": 21525 + }, + { + "epoch": 1.38, + "learning_rate": 1.5402219884497308e-05, + "loss": 0.9096, + "step": 21526 + }, + { + "epoch": 1.38, + "learning_rate": 1.5402005195474357e-05, + "loss": 0.9691, + "step": 21527 + }, + { + "epoch": 1.39, + "learning_rate": 1.5401790506451406e-05, + "loss": 1.0677, + "step": 21528 + }, + { + "epoch": 1.39, + "learning_rate": 1.5401575817428455e-05, + "loss": 0.999, + "step": 21529 + }, + { + "epoch": 1.39, + "learning_rate": 1.5401361128405507e-05, + "loss": 0.955, + "step": 21530 + }, + { + "epoch": 1.39, + "learning_rate": 1.5401146439382556e-05, + "loss": 0.9621, + "step": 21531 + }, + { + "epoch": 1.39, + "learning_rate": 1.5400931750359605e-05, + "loss": 0.9839, + "step": 21532 + }, + { + "epoch": 1.39, + "learning_rate": 1.5400717061336654e-05, + "loss": 0.9366, + "step": 21533 + }, + { + "epoch": 1.39, + "learning_rate": 1.5400502372313706e-05, + "loss": 0.9787, + "step": 21534 + }, + { + "epoch": 1.39, + "learning_rate": 1.5400287683290755e-05, + "loss": 1.0296, + "step": 21535 + }, + { + "epoch": 1.39, + "learning_rate": 1.5400072994267804e-05, + "loss": 0.99, + "step": 21536 + }, + { + "epoch": 1.39, + "learning_rate": 1.5399858305244853e-05, + "loss": 0.9481, + "step": 21537 + }, + { + "epoch": 1.39, + "learning_rate": 1.5399643616221905e-05, + "loss": 1.0356, + "step": 21538 + }, + { + "epoch": 1.39, + "learning_rate": 1.5399428927198954e-05, + "loss": 0.9463, + "step": 21539 + }, + { + "epoch": 1.39, + "learning_rate": 1.5399214238176003e-05, + "loss": 0.8792, + "step": 21540 + }, + { + "epoch": 1.39, + "learning_rate": 1.5398999549153052e-05, + "loss": 0.9986, + "step": 21541 + }, + { + "epoch": 1.39, + "learning_rate": 1.53987848601301e-05, + "loss": 0.9126, + "step": 21542 + }, + { + "epoch": 1.39, + "learning_rate": 1.5398570171107154e-05, + "loss": 0.9966, + "step": 21543 + }, + { + "epoch": 1.39, + "learning_rate": 1.5398355482084203e-05, + "loss": 1.0009, + "step": 21544 + }, + { + "epoch": 1.39, + "learning_rate": 1.539814079306125e-05, + "loss": 0.9178, + "step": 21545 + }, + { + "epoch": 1.39, + "learning_rate": 1.53979261040383e-05, + "loss": 0.9352, + "step": 21546 + }, + { + "epoch": 1.39, + "learning_rate": 1.539771141501535e-05, + "loss": 1.0502, + "step": 21547 + }, + { + "epoch": 1.39, + "learning_rate": 1.5397496725992402e-05, + "loss": 0.9713, + "step": 21548 + }, + { + "epoch": 1.39, + "learning_rate": 1.539728203696945e-05, + "loss": 0.9123, + "step": 21549 + }, + { + "epoch": 1.39, + "learning_rate": 1.5397067347946503e-05, + "loss": 0.9197, + "step": 21550 + }, + { + "epoch": 1.39, + "learning_rate": 1.5396852658923552e-05, + "loss": 0.9677, + "step": 21551 + }, + { + "epoch": 1.39, + "learning_rate": 1.53966379699006e-05, + "loss": 1.0385, + "step": 21552 + }, + { + "epoch": 1.39, + "learning_rate": 1.539642328087765e-05, + "loss": 0.9657, + "step": 21553 + }, + { + "epoch": 1.39, + "learning_rate": 1.53962085918547e-05, + "loss": 0.9349, + "step": 21554 + }, + { + "epoch": 1.39, + "learning_rate": 1.5395993902831748e-05, + "loss": 0.9664, + "step": 21555 + }, + { + "epoch": 1.39, + "learning_rate": 1.53957792138088e-05, + "loss": 0.9265, + "step": 21556 + }, + { + "epoch": 1.39, + "learning_rate": 1.539556452478585e-05, + "loss": 0.9611, + "step": 21557 + }, + { + "epoch": 1.39, + "learning_rate": 1.5395349835762898e-05, + "loss": 0.9633, + "step": 21558 + }, + { + "epoch": 1.39, + "learning_rate": 1.5395135146739947e-05, + "loss": 1.0637, + "step": 21559 + }, + { + "epoch": 1.39, + "learning_rate": 1.5394920457716996e-05, + "loss": 0.9901, + "step": 21560 + }, + { + "epoch": 1.39, + "learning_rate": 1.539470576869405e-05, + "loss": 0.9735, + "step": 21561 + }, + { + "epoch": 1.39, + "learning_rate": 1.5394491079671098e-05, + "loss": 0.9682, + "step": 21562 + }, + { + "epoch": 1.39, + "learning_rate": 1.539427639064815e-05, + "loss": 0.9315, + "step": 21563 + }, + { + "epoch": 1.39, + "learning_rate": 1.53940617016252e-05, + "loss": 0.986, + "step": 21564 + }, + { + "epoch": 1.39, + "learning_rate": 1.5393847012602248e-05, + "loss": 0.9421, + "step": 21565 + }, + { + "epoch": 1.39, + "learning_rate": 1.5393632323579297e-05, + "loss": 1.0204, + "step": 21566 + }, + { + "epoch": 1.39, + "learning_rate": 1.5393417634556346e-05, + "loss": 0.9679, + "step": 21567 + }, + { + "epoch": 1.39, + "learning_rate": 1.5393202945533398e-05, + "loss": 0.9868, + "step": 21568 + }, + { + "epoch": 1.39, + "learning_rate": 1.5392988256510447e-05, + "loss": 0.957, + "step": 21569 + }, + { + "epoch": 1.39, + "learning_rate": 1.5392773567487496e-05, + "loss": 0.9919, + "step": 21570 + }, + { + "epoch": 1.39, + "learning_rate": 1.5392558878464545e-05, + "loss": 0.9599, + "step": 21571 + }, + { + "epoch": 1.39, + "learning_rate": 1.5392344189441594e-05, + "loss": 1.0365, + "step": 21572 + }, + { + "epoch": 1.39, + "learning_rate": 1.5392129500418643e-05, + "loss": 0.9259, + "step": 21573 + }, + { + "epoch": 1.39, + "learning_rate": 1.5391914811395695e-05, + "loss": 1.0012, + "step": 21574 + }, + { + "epoch": 1.39, + "learning_rate": 1.5391700122372744e-05, + "loss": 0.9576, + "step": 21575 + }, + { + "epoch": 1.39, + "learning_rate": 1.5391485433349793e-05, + "loss": 0.9805, + "step": 21576 + }, + { + "epoch": 1.39, + "learning_rate": 1.5391270744326842e-05, + "loss": 1.0026, + "step": 21577 + }, + { + "epoch": 1.39, + "learning_rate": 1.5391056055303895e-05, + "loss": 0.9703, + "step": 21578 + }, + { + "epoch": 1.39, + "learning_rate": 1.5390841366280943e-05, + "loss": 0.9728, + "step": 21579 + }, + { + "epoch": 1.39, + "learning_rate": 1.5390626677257992e-05, + "loss": 0.9221, + "step": 21580 + }, + { + "epoch": 1.39, + "learning_rate": 1.5390411988235045e-05, + "loss": 1.0094, + "step": 21581 + }, + { + "epoch": 1.39, + "learning_rate": 1.5390197299212094e-05, + "loss": 1.041, + "step": 21582 + }, + { + "epoch": 1.39, + "learning_rate": 1.5389982610189143e-05, + "loss": 0.9562, + "step": 21583 + }, + { + "epoch": 1.39, + "learning_rate": 1.538976792116619e-05, + "loss": 0.9987, + "step": 21584 + }, + { + "epoch": 1.39, + "learning_rate": 1.538955323214324e-05, + "loss": 0.9068, + "step": 21585 + }, + { + "epoch": 1.39, + "learning_rate": 1.538933854312029e-05, + "loss": 0.9894, + "step": 21586 + }, + { + "epoch": 1.39, + "learning_rate": 1.5389123854097342e-05, + "loss": 0.9478, + "step": 21587 + }, + { + "epoch": 1.39, + "learning_rate": 1.538890916507439e-05, + "loss": 0.9425, + "step": 21588 + }, + { + "epoch": 1.39, + "learning_rate": 1.538869447605144e-05, + "loss": 1.0244, + "step": 21589 + }, + { + "epoch": 1.39, + "learning_rate": 1.538847978702849e-05, + "loss": 1.0197, + "step": 21590 + }, + { + "epoch": 1.39, + "learning_rate": 1.5388265098005538e-05, + "loss": 1.0392, + "step": 21591 + }, + { + "epoch": 1.39, + "learning_rate": 1.538805040898259e-05, + "loss": 0.9627, + "step": 21592 + }, + { + "epoch": 1.39, + "learning_rate": 1.538783571995964e-05, + "loss": 0.9229, + "step": 21593 + }, + { + "epoch": 1.39, + "learning_rate": 1.538762103093669e-05, + "loss": 0.9953, + "step": 21594 + }, + { + "epoch": 1.39, + "learning_rate": 1.538740634191374e-05, + "loss": 1.0116, + "step": 21595 + }, + { + "epoch": 1.39, + "learning_rate": 1.538719165289079e-05, + "loss": 0.9147, + "step": 21596 + }, + { + "epoch": 1.39, + "learning_rate": 1.538697696386784e-05, + "loss": 0.9215, + "step": 21597 + }, + { + "epoch": 1.39, + "learning_rate": 1.5386762274844887e-05, + "loss": 0.948, + "step": 21598 + }, + { + "epoch": 1.39, + "learning_rate": 1.538654758582194e-05, + "loss": 1.0111, + "step": 21599 + }, + { + "epoch": 1.39, + "learning_rate": 1.538633289679899e-05, + "loss": 0.8996, + "step": 21600 + }, + { + "epoch": 1.39, + "learning_rate": 1.5386118207776038e-05, + "loss": 0.9018, + "step": 21601 + }, + { + "epoch": 1.39, + "learning_rate": 1.5385903518753087e-05, + "loss": 0.9487, + "step": 21602 + }, + { + "epoch": 1.39, + "learning_rate": 1.5385688829730136e-05, + "loss": 0.9424, + "step": 21603 + }, + { + "epoch": 1.39, + "learning_rate": 1.5385474140707185e-05, + "loss": 0.9664, + "step": 21604 + }, + { + "epoch": 1.39, + "learning_rate": 1.5385259451684237e-05, + "loss": 0.9248, + "step": 21605 + }, + { + "epoch": 1.39, + "learning_rate": 1.5385044762661286e-05, + "loss": 1.0095, + "step": 21606 + }, + { + "epoch": 1.39, + "learning_rate": 1.5384830073638338e-05, + "loss": 0.9781, + "step": 21607 + }, + { + "epoch": 1.39, + "learning_rate": 1.5384615384615387e-05, + "loss": 1.0196, + "step": 21608 + }, + { + "epoch": 1.39, + "learning_rate": 1.5384400695592436e-05, + "loss": 1.0211, + "step": 21609 + }, + { + "epoch": 1.39, + "learning_rate": 1.5384186006569485e-05, + "loss": 1.0341, + "step": 21610 + }, + { + "epoch": 1.39, + "learning_rate": 1.5383971317546534e-05, + "loss": 0.9726, + "step": 21611 + }, + { + "epoch": 1.39, + "learning_rate": 1.5383756628523586e-05, + "loss": 0.9638, + "step": 21612 + }, + { + "epoch": 1.39, + "learning_rate": 1.5383541939500635e-05, + "loss": 0.9198, + "step": 21613 + }, + { + "epoch": 1.39, + "learning_rate": 1.5383327250477684e-05, + "loss": 0.9145, + "step": 21614 + }, + { + "epoch": 1.39, + "learning_rate": 1.5383112561454733e-05, + "loss": 1.1433, + "step": 21615 + }, + { + "epoch": 1.39, + "learning_rate": 1.5382897872431782e-05, + "loss": 0.9633, + "step": 21616 + }, + { + "epoch": 1.39, + "learning_rate": 1.5382683183408835e-05, + "loss": 0.9151, + "step": 21617 + }, + { + "epoch": 1.39, + "learning_rate": 1.5382468494385884e-05, + "loss": 0.9669, + "step": 21618 + }, + { + "epoch": 1.39, + "learning_rate": 1.5382253805362933e-05, + "loss": 1.0125, + "step": 21619 + }, + { + "epoch": 1.39, + "learning_rate": 1.538203911633998e-05, + "loss": 0.9768, + "step": 21620 + }, + { + "epoch": 1.39, + "learning_rate": 1.5381824427317034e-05, + "loss": 0.9667, + "step": 21621 + }, + { + "epoch": 1.39, + "learning_rate": 1.5381609738294083e-05, + "loss": 0.9968, + "step": 21622 + }, + { + "epoch": 1.39, + "learning_rate": 1.5381395049271132e-05, + "loss": 0.8679, + "step": 21623 + }, + { + "epoch": 1.39, + "learning_rate": 1.538118036024818e-05, + "loss": 0.8979, + "step": 21624 + }, + { + "epoch": 1.39, + "learning_rate": 1.5380965671225233e-05, + "loss": 1.0005, + "step": 21625 + }, + { + "epoch": 1.39, + "learning_rate": 1.5380750982202282e-05, + "loss": 0.9375, + "step": 21626 + }, + { + "epoch": 1.39, + "learning_rate": 1.538053629317933e-05, + "loss": 1.0014, + "step": 21627 + }, + { + "epoch": 1.39, + "learning_rate": 1.538032160415638e-05, + "loss": 0.9593, + "step": 21628 + }, + { + "epoch": 1.39, + "learning_rate": 1.538010691513343e-05, + "loss": 0.9419, + "step": 21629 + }, + { + "epoch": 1.39, + "learning_rate": 1.537989222611048e-05, + "loss": 0.9643, + "step": 21630 + }, + { + "epoch": 1.39, + "learning_rate": 1.537967753708753e-05, + "loss": 1.0058, + "step": 21631 + }, + { + "epoch": 1.39, + "learning_rate": 1.537946284806458e-05, + "loss": 1.0989, + "step": 21632 + }, + { + "epoch": 1.39, + "learning_rate": 1.5379248159041628e-05, + "loss": 1.0347, + "step": 21633 + }, + { + "epoch": 1.39, + "learning_rate": 1.5379033470018677e-05, + "loss": 0.9642, + "step": 21634 + }, + { + "epoch": 1.39, + "learning_rate": 1.537881878099573e-05, + "loss": 0.9733, + "step": 21635 + }, + { + "epoch": 1.39, + "learning_rate": 1.537860409197278e-05, + "loss": 0.9258, + "step": 21636 + }, + { + "epoch": 1.39, + "learning_rate": 1.537838940294983e-05, + "loss": 0.9437, + "step": 21637 + }, + { + "epoch": 1.39, + "learning_rate": 1.537817471392688e-05, + "loss": 0.9711, + "step": 21638 + }, + { + "epoch": 1.39, + "learning_rate": 1.537796002490393e-05, + "loss": 0.9794, + "step": 21639 + }, + { + "epoch": 1.39, + "learning_rate": 1.5377745335880978e-05, + "loss": 1.0178, + "step": 21640 + }, + { + "epoch": 1.39, + "learning_rate": 1.5377530646858027e-05, + "loss": 0.9591, + "step": 21641 + }, + { + "epoch": 1.39, + "learning_rate": 1.5377315957835076e-05, + "loss": 1.0397, + "step": 21642 + }, + { + "epoch": 1.39, + "learning_rate": 1.5377101268812128e-05, + "loss": 0.9846, + "step": 21643 + }, + { + "epoch": 1.39, + "learning_rate": 1.5376886579789177e-05, + "loss": 0.9971, + "step": 21644 + }, + { + "epoch": 1.39, + "learning_rate": 1.5376671890766226e-05, + "loss": 0.9447, + "step": 21645 + }, + { + "epoch": 1.39, + "learning_rate": 1.5376457201743275e-05, + "loss": 1.0112, + "step": 21646 + }, + { + "epoch": 1.39, + "learning_rate": 1.5376242512720324e-05, + "loss": 1.0536, + "step": 21647 + }, + { + "epoch": 1.39, + "learning_rate": 1.5376027823697376e-05, + "loss": 0.964, + "step": 21648 + }, + { + "epoch": 1.39, + "learning_rate": 1.5375813134674425e-05, + "loss": 0.9577, + "step": 21649 + }, + { + "epoch": 1.39, + "learning_rate": 1.5375598445651478e-05, + "loss": 1.0158, + "step": 21650 + }, + { + "epoch": 1.39, + "learning_rate": 1.5375383756628527e-05, + "loss": 0.9422, + "step": 21651 + }, + { + "epoch": 1.39, + "learning_rate": 1.5375169067605576e-05, + "loss": 0.9622, + "step": 21652 + }, + { + "epoch": 1.39, + "learning_rate": 1.5374954378582624e-05, + "loss": 0.9908, + "step": 21653 + }, + { + "epoch": 1.39, + "learning_rate": 1.5374739689559673e-05, + "loss": 1.0018, + "step": 21654 + }, + { + "epoch": 1.39, + "learning_rate": 1.5374525000536726e-05, + "loss": 1.0036, + "step": 21655 + }, + { + "epoch": 1.39, + "learning_rate": 1.5374310311513775e-05, + "loss": 0.9895, + "step": 21656 + }, + { + "epoch": 1.39, + "learning_rate": 1.5374095622490824e-05, + "loss": 1.0189, + "step": 21657 + }, + { + "epoch": 1.39, + "learning_rate": 1.5373880933467873e-05, + "loss": 0.9481, + "step": 21658 + }, + { + "epoch": 1.39, + "learning_rate": 1.537366624444492e-05, + "loss": 0.9696, + "step": 21659 + }, + { + "epoch": 1.39, + "learning_rate": 1.537345155542197e-05, + "loss": 0.9954, + "step": 21660 + }, + { + "epoch": 1.39, + "learning_rate": 1.5373236866399023e-05, + "loss": 0.9547, + "step": 21661 + }, + { + "epoch": 1.39, + "learning_rate": 1.5373022177376072e-05, + "loss": 0.9968, + "step": 21662 + }, + { + "epoch": 1.39, + "learning_rate": 1.537280748835312e-05, + "loss": 0.9824, + "step": 21663 + }, + { + "epoch": 1.39, + "learning_rate": 1.537259279933017e-05, + "loss": 0.9729, + "step": 21664 + }, + { + "epoch": 1.39, + "learning_rate": 1.5372378110307222e-05, + "loss": 0.9791, + "step": 21665 + }, + { + "epoch": 1.39, + "learning_rate": 1.537216342128427e-05, + "loss": 0.9675, + "step": 21666 + }, + { + "epoch": 1.39, + "learning_rate": 1.537194873226132e-05, + "loss": 1.0008, + "step": 21667 + }, + { + "epoch": 1.39, + "learning_rate": 1.5371734043238373e-05, + "loss": 0.9587, + "step": 21668 + }, + { + "epoch": 1.39, + "learning_rate": 1.537151935421542e-05, + "loss": 0.9822, + "step": 21669 + }, + { + "epoch": 1.39, + "learning_rate": 1.537130466519247e-05, + "loss": 0.8788, + "step": 21670 + }, + { + "epoch": 1.39, + "learning_rate": 1.537108997616952e-05, + "loss": 0.9902, + "step": 21671 + }, + { + "epoch": 1.39, + "learning_rate": 1.537087528714657e-05, + "loss": 1.0314, + "step": 21672 + }, + { + "epoch": 1.39, + "learning_rate": 1.5370660598123617e-05, + "loss": 0.9586, + "step": 21673 + }, + { + "epoch": 1.39, + "learning_rate": 1.537044590910067e-05, + "loss": 1.0001, + "step": 21674 + }, + { + "epoch": 1.39, + "learning_rate": 1.537023122007772e-05, + "loss": 1.0389, + "step": 21675 + }, + { + "epoch": 1.39, + "learning_rate": 1.5370016531054768e-05, + "loss": 0.9703, + "step": 21676 + }, + { + "epoch": 1.39, + "learning_rate": 1.5369801842031817e-05, + "loss": 0.9912, + "step": 21677 + }, + { + "epoch": 1.39, + "learning_rate": 1.5369587153008866e-05, + "loss": 0.9352, + "step": 21678 + }, + { + "epoch": 1.39, + "learning_rate": 1.5369372463985918e-05, + "loss": 0.8945, + "step": 21679 + }, + { + "epoch": 1.39, + "learning_rate": 1.5369157774962967e-05, + "loss": 0.9745, + "step": 21680 + }, + { + "epoch": 1.39, + "learning_rate": 1.536894308594002e-05, + "loss": 1.0109, + "step": 21681 + }, + { + "epoch": 1.39, + "learning_rate": 1.5368728396917068e-05, + "loss": 0.9778, + "step": 21682 + }, + { + "epoch": 1.4, + "learning_rate": 1.5368513707894117e-05, + "loss": 0.9185, + "step": 21683 + }, + { + "epoch": 1.4, + "learning_rate": 1.5368299018871166e-05, + "loss": 0.879, + "step": 21684 + }, + { + "epoch": 1.4, + "learning_rate": 1.5368084329848215e-05, + "loss": 1.0339, + "step": 21685 + }, + { + "epoch": 1.4, + "learning_rate": 1.5367869640825267e-05, + "loss": 0.9611, + "step": 21686 + }, + { + "epoch": 1.4, + "learning_rate": 1.5367654951802316e-05, + "loss": 0.9474, + "step": 21687 + }, + { + "epoch": 1.4, + "learning_rate": 1.5367440262779365e-05, + "loss": 0.92, + "step": 21688 + }, + { + "epoch": 1.4, + "learning_rate": 1.5367225573756414e-05, + "loss": 0.9546, + "step": 21689 + }, + { + "epoch": 1.4, + "learning_rate": 1.5367010884733463e-05, + "loss": 1.0481, + "step": 21690 + }, + { + "epoch": 1.4, + "learning_rate": 1.5366796195710512e-05, + "loss": 0.9311, + "step": 21691 + }, + { + "epoch": 1.4, + "learning_rate": 1.5366581506687565e-05, + "loss": 0.9912, + "step": 21692 + }, + { + "epoch": 1.4, + "learning_rate": 1.5366366817664614e-05, + "loss": 0.9582, + "step": 21693 + }, + { + "epoch": 1.4, + "learning_rate": 1.5366152128641666e-05, + "loss": 1.1109, + "step": 21694 + }, + { + "epoch": 1.4, + "learning_rate": 1.5365937439618715e-05, + "loss": 0.9334, + "step": 21695 + }, + { + "epoch": 1.4, + "learning_rate": 1.5365722750595764e-05, + "loss": 0.9801, + "step": 21696 + }, + { + "epoch": 1.4, + "learning_rate": 1.5365508061572813e-05, + "loss": 1.0095, + "step": 21697 + }, + { + "epoch": 1.4, + "learning_rate": 1.5365293372549862e-05, + "loss": 0.9383, + "step": 21698 + }, + { + "epoch": 1.4, + "learning_rate": 1.5365078683526914e-05, + "loss": 0.9625, + "step": 21699 + }, + { + "epoch": 1.4, + "learning_rate": 1.5364863994503963e-05, + "loss": 0.9475, + "step": 21700 + }, + { + "epoch": 1.4, + "learning_rate": 1.5364649305481012e-05, + "loss": 1.0159, + "step": 21701 + }, + { + "epoch": 1.4, + "learning_rate": 1.536443461645806e-05, + "loss": 0.9422, + "step": 21702 + }, + { + "epoch": 1.4, + "learning_rate": 1.536421992743511e-05, + "loss": 0.9333, + "step": 21703 + }, + { + "epoch": 1.4, + "learning_rate": 1.5364005238412162e-05, + "loss": 1.0395, + "step": 21704 + }, + { + "epoch": 1.4, + "learning_rate": 1.536379054938921e-05, + "loss": 0.9794, + "step": 21705 + }, + { + "epoch": 1.4, + "learning_rate": 1.536357586036626e-05, + "loss": 0.9407, + "step": 21706 + }, + { + "epoch": 1.4, + "learning_rate": 1.536336117134331e-05, + "loss": 0.957, + "step": 21707 + }, + { + "epoch": 1.4, + "learning_rate": 1.536314648232036e-05, + "loss": 0.8547, + "step": 21708 + }, + { + "epoch": 1.4, + "learning_rate": 1.536293179329741e-05, + "loss": 1.0656, + "step": 21709 + }, + { + "epoch": 1.4, + "learning_rate": 1.536271710427446e-05, + "loss": 0.9708, + "step": 21710 + }, + { + "epoch": 1.4, + "learning_rate": 1.536250241525151e-05, + "loss": 1.0243, + "step": 21711 + }, + { + "epoch": 1.4, + "learning_rate": 1.536228772622856e-05, + "loss": 0.9617, + "step": 21712 + }, + { + "epoch": 1.4, + "learning_rate": 1.536207303720561e-05, + "loss": 0.9836, + "step": 21713 + }, + { + "epoch": 1.4, + "learning_rate": 1.536185834818266e-05, + "loss": 0.9933, + "step": 21714 + }, + { + "epoch": 1.4, + "learning_rate": 1.5361643659159708e-05, + "loss": 0.9742, + "step": 21715 + }, + { + "epoch": 1.4, + "learning_rate": 1.5361428970136757e-05, + "loss": 1.0531, + "step": 21716 + }, + { + "epoch": 1.4, + "learning_rate": 1.536121428111381e-05, + "loss": 0.9004, + "step": 21717 + }, + { + "epoch": 1.4, + "learning_rate": 1.5360999592090858e-05, + "loss": 1.0847, + "step": 21718 + }, + { + "epoch": 1.4, + "learning_rate": 1.5360784903067907e-05, + "loss": 0.9877, + "step": 21719 + }, + { + "epoch": 1.4, + "learning_rate": 1.5360570214044956e-05, + "loss": 0.9906, + "step": 21720 + }, + { + "epoch": 1.4, + "learning_rate": 1.5360355525022005e-05, + "loss": 0.9462, + "step": 21721 + }, + { + "epoch": 1.4, + "learning_rate": 1.5360140835999057e-05, + "loss": 1.0164, + "step": 21722 + }, + { + "epoch": 1.4, + "learning_rate": 1.5359926146976106e-05, + "loss": 0.9571, + "step": 21723 + }, + { + "epoch": 1.4, + "learning_rate": 1.535971145795316e-05, + "loss": 0.9781, + "step": 21724 + }, + { + "epoch": 1.4, + "learning_rate": 1.5359496768930208e-05, + "loss": 0.9624, + "step": 21725 + }, + { + "epoch": 1.4, + "learning_rate": 1.5359282079907257e-05, + "loss": 0.9549, + "step": 21726 + }, + { + "epoch": 1.4, + "learning_rate": 1.5359067390884305e-05, + "loss": 1.0023, + "step": 21727 + }, + { + "epoch": 1.4, + "learning_rate": 1.5358852701861354e-05, + "loss": 0.9184, + "step": 21728 + }, + { + "epoch": 1.4, + "learning_rate": 1.5358638012838403e-05, + "loss": 1.035, + "step": 21729 + }, + { + "epoch": 1.4, + "learning_rate": 1.5358423323815456e-05, + "loss": 0.9023, + "step": 21730 + }, + { + "epoch": 1.4, + "learning_rate": 1.5358208634792505e-05, + "loss": 0.9324, + "step": 21731 + }, + { + "epoch": 1.4, + "learning_rate": 1.5357993945769554e-05, + "loss": 1.0324, + "step": 21732 + }, + { + "epoch": 1.4, + "learning_rate": 1.5357779256746603e-05, + "loss": 0.9672, + "step": 21733 + }, + { + "epoch": 1.4, + "learning_rate": 1.535756456772365e-05, + "loss": 1.0162, + "step": 21734 + }, + { + "epoch": 1.4, + "learning_rate": 1.5357349878700704e-05, + "loss": 0.9727, + "step": 21735 + }, + { + "epoch": 1.4, + "learning_rate": 1.5357135189677753e-05, + "loss": 0.976, + "step": 21736 + }, + { + "epoch": 1.4, + "learning_rate": 1.5356920500654802e-05, + "loss": 1.0144, + "step": 21737 + }, + { + "epoch": 1.4, + "learning_rate": 1.5356705811631854e-05, + "loss": 0.9126, + "step": 21738 + }, + { + "epoch": 1.4, + "learning_rate": 1.5356491122608903e-05, + "loss": 1.0447, + "step": 21739 + }, + { + "epoch": 1.4, + "learning_rate": 1.5356276433585952e-05, + "loss": 0.9802, + "step": 21740 + }, + { + "epoch": 1.4, + "learning_rate": 1.5356061744563e-05, + "loss": 0.9584, + "step": 21741 + }, + { + "epoch": 1.4, + "learning_rate": 1.5355847055540054e-05, + "loss": 0.9664, + "step": 21742 + }, + { + "epoch": 1.4, + "learning_rate": 1.5355632366517102e-05, + "loss": 0.9686, + "step": 21743 + }, + { + "epoch": 1.4, + "learning_rate": 1.535541767749415e-05, + "loss": 1.0087, + "step": 21744 + }, + { + "epoch": 1.4, + "learning_rate": 1.53552029884712e-05, + "loss": 0.9718, + "step": 21745 + }, + { + "epoch": 1.4, + "learning_rate": 1.535498829944825e-05, + "loss": 0.9723, + "step": 21746 + }, + { + "epoch": 1.4, + "learning_rate": 1.53547736104253e-05, + "loss": 1.0261, + "step": 21747 + }, + { + "epoch": 1.4, + "learning_rate": 1.535455892140235e-05, + "loss": 0.9834, + "step": 21748 + }, + { + "epoch": 1.4, + "learning_rate": 1.53543442323794e-05, + "loss": 0.9721, + "step": 21749 + }, + { + "epoch": 1.4, + "learning_rate": 1.535412954335645e-05, + "loss": 0.9889, + "step": 21750 + }, + { + "epoch": 1.4, + "learning_rate": 1.5353914854333498e-05, + "loss": 1.0612, + "step": 21751 + }, + { + "epoch": 1.4, + "learning_rate": 1.535370016531055e-05, + "loss": 0.9399, + "step": 21752 + }, + { + "epoch": 1.4, + "learning_rate": 1.53534854762876e-05, + "loss": 0.92, + "step": 21753 + }, + { + "epoch": 1.4, + "learning_rate": 1.5353270787264648e-05, + "loss": 1.0019, + "step": 21754 + }, + { + "epoch": 1.4, + "learning_rate": 1.53530560982417e-05, + "loss": 0.9149, + "step": 21755 + }, + { + "epoch": 1.4, + "learning_rate": 1.535284140921875e-05, + "loss": 1.005, + "step": 21756 + }, + { + "epoch": 1.4, + "learning_rate": 1.5352626720195798e-05, + "loss": 0.9763, + "step": 21757 + }, + { + "epoch": 1.4, + "learning_rate": 1.5352412031172847e-05, + "loss": 0.9713, + "step": 21758 + }, + { + "epoch": 1.4, + "learning_rate": 1.5352197342149896e-05, + "loss": 0.9913, + "step": 21759 + }, + { + "epoch": 1.4, + "learning_rate": 1.5351982653126945e-05, + "loss": 0.957, + "step": 21760 + }, + { + "epoch": 1.4, + "learning_rate": 1.5351767964103997e-05, + "loss": 0.9989, + "step": 21761 + }, + { + "epoch": 1.4, + "learning_rate": 1.5351553275081046e-05, + "loss": 0.9775, + "step": 21762 + }, + { + "epoch": 1.4, + "learning_rate": 1.5351338586058095e-05, + "loss": 1.0223, + "step": 21763 + }, + { + "epoch": 1.4, + "learning_rate": 1.5351123897035144e-05, + "loss": 1.0004, + "step": 21764 + }, + { + "epoch": 1.4, + "learning_rate": 1.5350909208012193e-05, + "loss": 1.036, + "step": 21765 + }, + { + "epoch": 1.4, + "learning_rate": 1.5350694518989246e-05, + "loss": 1.0404, + "step": 21766 + }, + { + "epoch": 1.4, + "learning_rate": 1.5350479829966295e-05, + "loss": 0.9592, + "step": 21767 + }, + { + "epoch": 1.4, + "learning_rate": 1.5350265140943347e-05, + "loss": 0.9738, + "step": 21768 + }, + { + "epoch": 1.4, + "learning_rate": 1.5350050451920396e-05, + "loss": 0.9258, + "step": 21769 + }, + { + "epoch": 1.4, + "learning_rate": 1.5349835762897445e-05, + "loss": 0.953, + "step": 21770 + }, + { + "epoch": 1.4, + "learning_rate": 1.5349621073874494e-05, + "loss": 0.9393, + "step": 21771 + }, + { + "epoch": 1.4, + "learning_rate": 1.5349406384851543e-05, + "loss": 0.9834, + "step": 21772 + }, + { + "epoch": 1.4, + "learning_rate": 1.5349191695828595e-05, + "loss": 0.9608, + "step": 21773 + }, + { + "epoch": 1.4, + "learning_rate": 1.5348977006805644e-05, + "loss": 1.0163, + "step": 21774 + }, + { + "epoch": 1.4, + "learning_rate": 1.5348762317782693e-05, + "loss": 0.982, + "step": 21775 + }, + { + "epoch": 1.4, + "learning_rate": 1.5348547628759742e-05, + "loss": 1.028, + "step": 21776 + }, + { + "epoch": 1.4, + "learning_rate": 1.534833293973679e-05, + "loss": 0.9855, + "step": 21777 + }, + { + "epoch": 1.4, + "learning_rate": 1.534811825071384e-05, + "loss": 0.9883, + "step": 21778 + }, + { + "epoch": 1.4, + "learning_rate": 1.5347903561690892e-05, + "loss": 1.0254, + "step": 21779 + }, + { + "epoch": 1.4, + "learning_rate": 1.534768887266794e-05, + "loss": 1.0163, + "step": 21780 + }, + { + "epoch": 1.4, + "learning_rate": 1.5347474183644994e-05, + "loss": 0.9106, + "step": 21781 + }, + { + "epoch": 1.4, + "learning_rate": 1.5347259494622043e-05, + "loss": 0.9633, + "step": 21782 + }, + { + "epoch": 1.4, + "learning_rate": 1.534704480559909e-05, + "loss": 0.9754, + "step": 21783 + }, + { + "epoch": 1.4, + "learning_rate": 1.534683011657614e-05, + "loss": 0.894, + "step": 21784 + }, + { + "epoch": 1.4, + "learning_rate": 1.534661542755319e-05, + "loss": 0.9628, + "step": 21785 + }, + { + "epoch": 1.4, + "learning_rate": 1.5346400738530242e-05, + "loss": 1.0036, + "step": 21786 + }, + { + "epoch": 1.4, + "learning_rate": 1.534618604950729e-05, + "loss": 0.9023, + "step": 21787 + }, + { + "epoch": 1.4, + "learning_rate": 1.534597136048434e-05, + "loss": 0.9882, + "step": 21788 + }, + { + "epoch": 1.4, + "learning_rate": 1.534575667146139e-05, + "loss": 0.9616, + "step": 21789 + }, + { + "epoch": 1.4, + "learning_rate": 1.5345541982438438e-05, + "loss": 0.901, + "step": 21790 + }, + { + "epoch": 1.4, + "learning_rate": 1.534532729341549e-05, + "loss": 0.9175, + "step": 21791 + }, + { + "epoch": 1.4, + "learning_rate": 1.534511260439254e-05, + "loss": 0.9903, + "step": 21792 + }, + { + "epoch": 1.4, + "learning_rate": 1.5344897915369588e-05, + "loss": 1.0085, + "step": 21793 + }, + { + "epoch": 1.4, + "learning_rate": 1.5344683226346637e-05, + "loss": 1.0358, + "step": 21794 + }, + { + "epoch": 1.4, + "learning_rate": 1.5344468537323686e-05, + "loss": 0.9823, + "step": 21795 + }, + { + "epoch": 1.4, + "learning_rate": 1.5344253848300738e-05, + "loss": 1.0475, + "step": 21796 + }, + { + "epoch": 1.4, + "learning_rate": 1.5344039159277787e-05, + "loss": 1.0215, + "step": 21797 + }, + { + "epoch": 1.4, + "learning_rate": 1.5343824470254836e-05, + "loss": 0.9319, + "step": 21798 + }, + { + "epoch": 1.4, + "learning_rate": 1.534360978123189e-05, + "loss": 0.9966, + "step": 21799 + }, + { + "epoch": 1.4, + "learning_rate": 1.5343395092208938e-05, + "loss": 0.9774, + "step": 21800 + }, + { + "epoch": 1.4, + "learning_rate": 1.5343180403185986e-05, + "loss": 0.9621, + "step": 21801 + }, + { + "epoch": 1.4, + "learning_rate": 1.5342965714163035e-05, + "loss": 0.9279, + "step": 21802 + }, + { + "epoch": 1.4, + "learning_rate": 1.5342751025140084e-05, + "loss": 1.0385, + "step": 21803 + }, + { + "epoch": 1.4, + "learning_rate": 1.5342536336117137e-05, + "loss": 0.9695, + "step": 21804 + }, + { + "epoch": 1.4, + "learning_rate": 1.5342321647094186e-05, + "loss": 1.0223, + "step": 21805 + }, + { + "epoch": 1.4, + "learning_rate": 1.5342106958071235e-05, + "loss": 1.0386, + "step": 21806 + }, + { + "epoch": 1.4, + "learning_rate": 1.5341892269048284e-05, + "loss": 0.928, + "step": 21807 + }, + { + "epoch": 1.4, + "learning_rate": 1.5341677580025333e-05, + "loss": 1.0003, + "step": 21808 + }, + { + "epoch": 1.4, + "learning_rate": 1.534146289100238e-05, + "loss": 0.9866, + "step": 21809 + }, + { + "epoch": 1.4, + "learning_rate": 1.5341248201979434e-05, + "loss": 1.0262, + "step": 21810 + }, + { + "epoch": 1.4, + "learning_rate": 1.5341033512956486e-05, + "loss": 0.9042, + "step": 21811 + }, + { + "epoch": 1.4, + "learning_rate": 1.5340818823933535e-05, + "loss": 0.9949, + "step": 21812 + }, + { + "epoch": 1.4, + "learning_rate": 1.5340604134910584e-05, + "loss": 0.9414, + "step": 21813 + }, + { + "epoch": 1.4, + "learning_rate": 1.5340389445887633e-05, + "loss": 0.9991, + "step": 21814 + }, + { + "epoch": 1.4, + "learning_rate": 1.5340174756864682e-05, + "loss": 1.0166, + "step": 21815 + }, + { + "epoch": 1.4, + "learning_rate": 1.533996006784173e-05, + "loss": 1.0611, + "step": 21816 + }, + { + "epoch": 1.4, + "learning_rate": 1.5339745378818783e-05, + "loss": 0.8931, + "step": 21817 + }, + { + "epoch": 1.4, + "learning_rate": 1.5339530689795832e-05, + "loss": 0.9447, + "step": 21818 + }, + { + "epoch": 1.4, + "learning_rate": 1.533931600077288e-05, + "loss": 1.0558, + "step": 21819 + }, + { + "epoch": 1.4, + "learning_rate": 1.533910131174993e-05, + "loss": 0.9864, + "step": 21820 + }, + { + "epoch": 1.4, + "learning_rate": 1.533888662272698e-05, + "loss": 0.9455, + "step": 21821 + }, + { + "epoch": 1.4, + "learning_rate": 1.533867193370403e-05, + "loss": 0.9596, + "step": 21822 + }, + { + "epoch": 1.4, + "learning_rate": 1.533845724468108e-05, + "loss": 0.9798, + "step": 21823 + }, + { + "epoch": 1.4, + "learning_rate": 1.533824255565813e-05, + "loss": 1.0113, + "step": 21824 + }, + { + "epoch": 1.4, + "learning_rate": 1.5338027866635182e-05, + "loss": 1.008, + "step": 21825 + }, + { + "epoch": 1.4, + "learning_rate": 1.533781317761223e-05, + "loss": 1.0015, + "step": 21826 + }, + { + "epoch": 1.4, + "learning_rate": 1.533759848858928e-05, + "loss": 0.9859, + "step": 21827 + }, + { + "epoch": 1.4, + "learning_rate": 1.533738379956633e-05, + "loss": 0.9907, + "step": 21828 + }, + { + "epoch": 1.4, + "learning_rate": 1.533716911054338e-05, + "loss": 1.0575, + "step": 21829 + }, + { + "epoch": 1.4, + "learning_rate": 1.533695442152043e-05, + "loss": 0.9615, + "step": 21830 + }, + { + "epoch": 1.4, + "learning_rate": 1.533673973249748e-05, + "loss": 0.9327, + "step": 21831 + }, + { + "epoch": 1.4, + "learning_rate": 1.5336525043474528e-05, + "loss": 1.1799, + "step": 21832 + }, + { + "epoch": 1.4, + "learning_rate": 1.5336310354451577e-05, + "loss": 0.9478, + "step": 21833 + }, + { + "epoch": 1.4, + "learning_rate": 1.5336095665428626e-05, + "loss": 0.9735, + "step": 21834 + }, + { + "epoch": 1.4, + "learning_rate": 1.533588097640568e-05, + "loss": 0.9537, + "step": 21835 + }, + { + "epoch": 1.4, + "learning_rate": 1.5335666287382727e-05, + "loss": 0.9073, + "step": 21836 + }, + { + "epoch": 1.4, + "learning_rate": 1.5335451598359776e-05, + "loss": 0.9995, + "step": 21837 + }, + { + "epoch": 1.41, + "learning_rate": 1.5335236909336825e-05, + "loss": 1.0249, + "step": 21838 + }, + { + "epoch": 1.41, + "learning_rate": 1.5335022220313878e-05, + "loss": 0.9257, + "step": 21839 + }, + { + "epoch": 1.41, + "learning_rate": 1.5334807531290927e-05, + "loss": 0.9826, + "step": 21840 + }, + { + "epoch": 1.41, + "learning_rate": 1.5334592842267976e-05, + "loss": 0.9671, + "step": 21841 + }, + { + "epoch": 1.41, + "learning_rate": 1.5334378153245028e-05, + "loss": 1.0092, + "step": 21842 + }, + { + "epoch": 1.41, + "learning_rate": 1.5334163464222077e-05, + "loss": 0.8544, + "step": 21843 + }, + { + "epoch": 1.41, + "learning_rate": 1.5333948775199126e-05, + "loss": 0.9084, + "step": 21844 + }, + { + "epoch": 1.41, + "learning_rate": 1.5333734086176175e-05, + "loss": 0.9843, + "step": 21845 + }, + { + "epoch": 1.41, + "learning_rate": 1.5333519397153224e-05, + "loss": 0.9951, + "step": 21846 + }, + { + "epoch": 1.41, + "learning_rate": 1.5333304708130273e-05, + "loss": 1.0248, + "step": 21847 + }, + { + "epoch": 1.41, + "learning_rate": 1.5333090019107325e-05, + "loss": 0.9218, + "step": 21848 + }, + { + "epoch": 1.41, + "learning_rate": 1.5332875330084374e-05, + "loss": 0.9123, + "step": 21849 + }, + { + "epoch": 1.41, + "learning_rate": 1.5332660641061423e-05, + "loss": 0.9001, + "step": 21850 + }, + { + "epoch": 1.41, + "learning_rate": 1.5332445952038472e-05, + "loss": 0.9907, + "step": 21851 + }, + { + "epoch": 1.41, + "learning_rate": 1.533223126301552e-05, + "loss": 0.9688, + "step": 21852 + }, + { + "epoch": 1.41, + "learning_rate": 1.5332016573992573e-05, + "loss": 0.9077, + "step": 21853 + }, + { + "epoch": 1.41, + "learning_rate": 1.5331801884969622e-05, + "loss": 1.0092, + "step": 21854 + }, + { + "epoch": 1.41, + "learning_rate": 1.5331587195946675e-05, + "loss": 0.9458, + "step": 21855 + }, + { + "epoch": 1.41, + "learning_rate": 1.5331372506923724e-05, + "loss": 1.0442, + "step": 21856 + }, + { + "epoch": 1.41, + "learning_rate": 1.5331157817900773e-05, + "loss": 0.9552, + "step": 21857 + }, + { + "epoch": 1.41, + "learning_rate": 1.533094312887782e-05, + "loss": 0.9618, + "step": 21858 + }, + { + "epoch": 1.41, + "learning_rate": 1.533072843985487e-05, + "loss": 0.8855, + "step": 21859 + }, + { + "epoch": 1.41, + "learning_rate": 1.5330513750831923e-05, + "loss": 1.0301, + "step": 21860 + }, + { + "epoch": 1.41, + "learning_rate": 1.5330299061808972e-05, + "loss": 1.0204, + "step": 21861 + }, + { + "epoch": 1.41, + "learning_rate": 1.533008437278602e-05, + "loss": 0.9897, + "step": 21862 + }, + { + "epoch": 1.41, + "learning_rate": 1.532986968376307e-05, + "loss": 0.9707, + "step": 21863 + }, + { + "epoch": 1.41, + "learning_rate": 1.532965499474012e-05, + "loss": 0.9448, + "step": 21864 + }, + { + "epoch": 1.41, + "learning_rate": 1.5329440305717168e-05, + "loss": 0.9339, + "step": 21865 + }, + { + "epoch": 1.41, + "learning_rate": 1.532922561669422e-05, + "loss": 1.0635, + "step": 21866 + }, + { + "epoch": 1.41, + "learning_rate": 1.532901092767127e-05, + "loss": 0.9617, + "step": 21867 + }, + { + "epoch": 1.41, + "learning_rate": 1.532879623864832e-05, + "loss": 0.9657, + "step": 21868 + }, + { + "epoch": 1.41, + "learning_rate": 1.532858154962537e-05, + "loss": 1.0187, + "step": 21869 + }, + { + "epoch": 1.41, + "learning_rate": 1.532836686060242e-05, + "loss": 0.9129, + "step": 21870 + }, + { + "epoch": 1.41, + "learning_rate": 1.5328152171579468e-05, + "loss": 1.0162, + "step": 21871 + }, + { + "epoch": 1.41, + "learning_rate": 1.5327937482556517e-05, + "loss": 0.9256, + "step": 21872 + }, + { + "epoch": 1.41, + "learning_rate": 1.532772279353357e-05, + "loss": 0.9847, + "step": 21873 + }, + { + "epoch": 1.41, + "learning_rate": 1.532750810451062e-05, + "loss": 0.9916, + "step": 21874 + }, + { + "epoch": 1.41, + "learning_rate": 1.5327293415487667e-05, + "loss": 1.0025, + "step": 21875 + }, + { + "epoch": 1.41, + "learning_rate": 1.5327078726464716e-05, + "loss": 0.9795, + "step": 21876 + }, + { + "epoch": 1.41, + "learning_rate": 1.5326864037441765e-05, + "loss": 0.9994, + "step": 21877 + }, + { + "epoch": 1.41, + "learning_rate": 1.5326649348418818e-05, + "loss": 0.878, + "step": 21878 + }, + { + "epoch": 1.41, + "learning_rate": 1.5326434659395867e-05, + "loss": 1.0201, + "step": 21879 + }, + { + "epoch": 1.41, + "learning_rate": 1.5326219970372916e-05, + "loss": 1.0222, + "step": 21880 + }, + { + "epoch": 1.41, + "learning_rate": 1.5326005281349965e-05, + "loss": 0.9041, + "step": 21881 + }, + { + "epoch": 1.41, + "learning_rate": 1.5325790592327014e-05, + "loss": 0.975, + "step": 21882 + }, + { + "epoch": 1.41, + "learning_rate": 1.5325575903304066e-05, + "loss": 1.0165, + "step": 21883 + }, + { + "epoch": 1.41, + "learning_rate": 1.5325361214281115e-05, + "loss": 0.9862, + "step": 21884 + }, + { + "epoch": 1.41, + "learning_rate": 1.5325146525258164e-05, + "loss": 1.0054, + "step": 21885 + }, + { + "epoch": 1.41, + "learning_rate": 1.5324931836235216e-05, + "loss": 0.9656, + "step": 21886 + }, + { + "epoch": 1.41, + "learning_rate": 1.5324717147212265e-05, + "loss": 1.0157, + "step": 21887 + }, + { + "epoch": 1.41, + "learning_rate": 1.5324502458189314e-05, + "loss": 1.0105, + "step": 21888 + }, + { + "epoch": 1.41, + "learning_rate": 1.5324287769166363e-05, + "loss": 0.966, + "step": 21889 + }, + { + "epoch": 1.41, + "learning_rate": 1.5324073080143412e-05, + "loss": 1.0367, + "step": 21890 + }, + { + "epoch": 1.41, + "learning_rate": 1.5323858391120464e-05, + "loss": 0.9574, + "step": 21891 + }, + { + "epoch": 1.41, + "learning_rate": 1.5323643702097513e-05, + "loss": 0.9374, + "step": 21892 + }, + { + "epoch": 1.41, + "learning_rate": 1.5323429013074562e-05, + "loss": 1.0267, + "step": 21893 + }, + { + "epoch": 1.41, + "learning_rate": 1.532321432405161e-05, + "loss": 1.0289, + "step": 21894 + }, + { + "epoch": 1.41, + "learning_rate": 1.532299963502866e-05, + "loss": 0.9682, + "step": 21895 + }, + { + "epoch": 1.41, + "learning_rate": 1.532278494600571e-05, + "loss": 1.0136, + "step": 21896 + }, + { + "epoch": 1.41, + "learning_rate": 1.532257025698276e-05, + "loss": 1.0482, + "step": 21897 + }, + { + "epoch": 1.41, + "learning_rate": 1.5322355567959814e-05, + "loss": 0.915, + "step": 21898 + }, + { + "epoch": 1.41, + "learning_rate": 1.5322140878936863e-05, + "loss": 1.0004, + "step": 21899 + }, + { + "epoch": 1.41, + "learning_rate": 1.5321926189913912e-05, + "loss": 0.9249, + "step": 21900 + }, + { + "epoch": 1.41, + "learning_rate": 1.532171150089096e-05, + "loss": 1.0125, + "step": 21901 + }, + { + "epoch": 1.41, + "learning_rate": 1.532149681186801e-05, + "loss": 0.9728, + "step": 21902 + }, + { + "epoch": 1.41, + "learning_rate": 1.532128212284506e-05, + "loss": 0.8999, + "step": 21903 + }, + { + "epoch": 1.41, + "learning_rate": 1.532106743382211e-05, + "loss": 0.964, + "step": 21904 + }, + { + "epoch": 1.41, + "learning_rate": 1.532085274479916e-05, + "loss": 0.9981, + "step": 21905 + }, + { + "epoch": 1.41, + "learning_rate": 1.532063805577621e-05, + "loss": 0.9963, + "step": 21906 + }, + { + "epoch": 1.41, + "learning_rate": 1.5320423366753258e-05, + "loss": 0.9621, + "step": 21907 + }, + { + "epoch": 1.41, + "learning_rate": 1.5320208677730307e-05, + "loss": 0.9625, + "step": 21908 + }, + { + "epoch": 1.41, + "learning_rate": 1.531999398870736e-05, + "loss": 0.9571, + "step": 21909 + }, + { + "epoch": 1.41, + "learning_rate": 1.531977929968441e-05, + "loss": 0.9633, + "step": 21910 + }, + { + "epoch": 1.41, + "learning_rate": 1.5319564610661457e-05, + "loss": 0.9499, + "step": 21911 + }, + { + "epoch": 1.41, + "learning_rate": 1.531934992163851e-05, + "loss": 1.033, + "step": 21912 + }, + { + "epoch": 1.41, + "learning_rate": 1.531913523261556e-05, + "loss": 1.0001, + "step": 21913 + }, + { + "epoch": 1.41, + "learning_rate": 1.5318920543592608e-05, + "loss": 1.0086, + "step": 21914 + }, + { + "epoch": 1.41, + "learning_rate": 1.5318705854569657e-05, + "loss": 0.9918, + "step": 21915 + }, + { + "epoch": 1.41, + "learning_rate": 1.5318491165546706e-05, + "loss": 0.997, + "step": 21916 + }, + { + "epoch": 1.41, + "learning_rate": 1.5318276476523758e-05, + "loss": 1.0273, + "step": 21917 + }, + { + "epoch": 1.41, + "learning_rate": 1.5318061787500807e-05, + "loss": 0.9012, + "step": 21918 + }, + { + "epoch": 1.41, + "learning_rate": 1.5317847098477856e-05, + "loss": 0.9273, + "step": 21919 + }, + { + "epoch": 1.41, + "learning_rate": 1.5317632409454905e-05, + "loss": 0.992, + "step": 21920 + }, + { + "epoch": 1.41, + "learning_rate": 1.5317417720431954e-05, + "loss": 0.9467, + "step": 21921 + }, + { + "epoch": 1.41, + "learning_rate": 1.5317203031409006e-05, + "loss": 1.0437, + "step": 21922 + }, + { + "epoch": 1.41, + "learning_rate": 1.5316988342386055e-05, + "loss": 0.952, + "step": 21923 + }, + { + "epoch": 1.41, + "learning_rate": 1.5316773653363104e-05, + "loss": 0.9611, + "step": 21924 + }, + { + "epoch": 1.41, + "learning_rate": 1.5316558964340153e-05, + "loss": 1.0373, + "step": 21925 + }, + { + "epoch": 1.41, + "learning_rate": 1.5316344275317205e-05, + "loss": 1.0004, + "step": 21926 + }, + { + "epoch": 1.41, + "learning_rate": 1.5316129586294254e-05, + "loss": 0.9737, + "step": 21927 + }, + { + "epoch": 1.41, + "learning_rate": 1.5315914897271303e-05, + "loss": 1.0265, + "step": 21928 + }, + { + "epoch": 1.41, + "learning_rate": 1.5315700208248356e-05, + "loss": 0.9774, + "step": 21929 + }, + { + "epoch": 1.41, + "learning_rate": 1.5315485519225405e-05, + "loss": 0.8974, + "step": 21930 + }, + { + "epoch": 1.41, + "learning_rate": 1.5315270830202454e-05, + "loss": 0.9476, + "step": 21931 + }, + { + "epoch": 1.41, + "learning_rate": 1.5315056141179503e-05, + "loss": 1.011, + "step": 21932 + }, + { + "epoch": 1.41, + "learning_rate": 1.531484145215655e-05, + "loss": 0.9961, + "step": 21933 + }, + { + "epoch": 1.41, + "learning_rate": 1.53146267631336e-05, + "loss": 0.9645, + "step": 21934 + }, + { + "epoch": 1.41, + "learning_rate": 1.5314412074110653e-05, + "loss": 1.0082, + "step": 21935 + }, + { + "epoch": 1.41, + "learning_rate": 1.5314197385087702e-05, + "loss": 1.0319, + "step": 21936 + }, + { + "epoch": 1.41, + "learning_rate": 1.531398269606475e-05, + "loss": 0.9881, + "step": 21937 + }, + { + "epoch": 1.41, + "learning_rate": 1.53137680070418e-05, + "loss": 0.9248, + "step": 21938 + }, + { + "epoch": 1.41, + "learning_rate": 1.531355331801885e-05, + "loss": 0.9605, + "step": 21939 + }, + { + "epoch": 1.41, + "learning_rate": 1.53133386289959e-05, + "loss": 0.9521, + "step": 21940 + }, + { + "epoch": 1.41, + "learning_rate": 1.531312393997295e-05, + "loss": 1.0117, + "step": 21941 + }, + { + "epoch": 1.41, + "learning_rate": 1.5312909250950002e-05, + "loss": 1.0216, + "step": 21942 + }, + { + "epoch": 1.41, + "learning_rate": 1.531269456192705e-05, + "loss": 0.9985, + "step": 21943 + }, + { + "epoch": 1.41, + "learning_rate": 1.53124798729041e-05, + "loss": 0.9085, + "step": 21944 + }, + { + "epoch": 1.41, + "learning_rate": 1.531226518388115e-05, + "loss": 0.9599, + "step": 21945 + }, + { + "epoch": 1.41, + "learning_rate": 1.5312050494858198e-05, + "loss": 0.9624, + "step": 21946 + }, + { + "epoch": 1.41, + "learning_rate": 1.531183580583525e-05, + "loss": 1.0569, + "step": 21947 + }, + { + "epoch": 1.41, + "learning_rate": 1.53116211168123e-05, + "loss": 0.9609, + "step": 21948 + }, + { + "epoch": 1.41, + "learning_rate": 1.531140642778935e-05, + "loss": 0.9757, + "step": 21949 + }, + { + "epoch": 1.41, + "learning_rate": 1.5311191738766397e-05, + "loss": 0.9775, + "step": 21950 + }, + { + "epoch": 1.41, + "learning_rate": 1.5310977049743446e-05, + "loss": 0.9405, + "step": 21951 + }, + { + "epoch": 1.41, + "learning_rate": 1.5310762360720495e-05, + "loss": 0.9453, + "step": 21952 + }, + { + "epoch": 1.41, + "learning_rate": 1.5310547671697548e-05, + "loss": 0.9472, + "step": 21953 + }, + { + "epoch": 1.41, + "learning_rate": 1.5310332982674597e-05, + "loss": 0.9228, + "step": 21954 + }, + { + "epoch": 1.41, + "learning_rate": 1.531011829365165e-05, + "loss": 1.0244, + "step": 21955 + }, + { + "epoch": 1.41, + "learning_rate": 1.5309903604628698e-05, + "loss": 0.9404, + "step": 21956 + }, + { + "epoch": 1.41, + "learning_rate": 1.5309688915605747e-05, + "loss": 0.9619, + "step": 21957 + }, + { + "epoch": 1.41, + "learning_rate": 1.5309474226582796e-05, + "loss": 0.9146, + "step": 21958 + }, + { + "epoch": 1.41, + "learning_rate": 1.5309259537559845e-05, + "loss": 0.9837, + "step": 21959 + }, + { + "epoch": 1.41, + "learning_rate": 1.5309044848536897e-05, + "loss": 1.0191, + "step": 21960 + }, + { + "epoch": 1.41, + "learning_rate": 1.5308830159513946e-05, + "loss": 1.0146, + "step": 21961 + }, + { + "epoch": 1.41, + "learning_rate": 1.5308615470490995e-05, + "loss": 0.9702, + "step": 21962 + }, + { + "epoch": 1.41, + "learning_rate": 1.5308400781468044e-05, + "loss": 0.9224, + "step": 21963 + }, + { + "epoch": 1.41, + "learning_rate": 1.5308186092445093e-05, + "loss": 1.0817, + "step": 21964 + }, + { + "epoch": 1.41, + "learning_rate": 1.5307971403422145e-05, + "loss": 0.9479, + "step": 21965 + }, + { + "epoch": 1.41, + "learning_rate": 1.5307756714399194e-05, + "loss": 0.9452, + "step": 21966 + }, + { + "epoch": 1.41, + "learning_rate": 1.5307542025376243e-05, + "loss": 0.9609, + "step": 21967 + }, + { + "epoch": 1.41, + "learning_rate": 1.5307327336353292e-05, + "loss": 1.0128, + "step": 21968 + }, + { + "epoch": 1.41, + "learning_rate": 1.530711264733034e-05, + "loss": 0.9816, + "step": 21969 + }, + { + "epoch": 1.41, + "learning_rate": 1.5306897958307394e-05, + "loss": 0.926, + "step": 21970 + }, + { + "epoch": 1.41, + "learning_rate": 1.5306683269284443e-05, + "loss": 0.9516, + "step": 21971 + }, + { + "epoch": 1.41, + "learning_rate": 1.530646858026149e-05, + "loss": 0.942, + "step": 21972 + }, + { + "epoch": 1.41, + "learning_rate": 1.5306253891238544e-05, + "loss": 0.9805, + "step": 21973 + }, + { + "epoch": 1.41, + "learning_rate": 1.5306039202215593e-05, + "loss": 0.98, + "step": 21974 + }, + { + "epoch": 1.41, + "learning_rate": 1.5305824513192642e-05, + "loss": 1.0079, + "step": 21975 + }, + { + "epoch": 1.41, + "learning_rate": 1.530560982416969e-05, + "loss": 0.9813, + "step": 21976 + }, + { + "epoch": 1.41, + "learning_rate": 1.530539513514674e-05, + "loss": 0.9869, + "step": 21977 + }, + { + "epoch": 1.41, + "learning_rate": 1.5305180446123792e-05, + "loss": 0.9423, + "step": 21978 + }, + { + "epoch": 1.41, + "learning_rate": 1.530496575710084e-05, + "loss": 0.952, + "step": 21979 + }, + { + "epoch": 1.41, + "learning_rate": 1.530475106807789e-05, + "loss": 0.9573, + "step": 21980 + }, + { + "epoch": 1.41, + "learning_rate": 1.530453637905494e-05, + "loss": 0.9712, + "step": 21981 + }, + { + "epoch": 1.41, + "learning_rate": 1.5304321690031988e-05, + "loss": 0.9844, + "step": 21982 + }, + { + "epoch": 1.41, + "learning_rate": 1.5304107001009037e-05, + "loss": 1.0415, + "step": 21983 + }, + { + "epoch": 1.41, + "learning_rate": 1.530389231198609e-05, + "loss": 0.9969, + "step": 21984 + }, + { + "epoch": 1.41, + "learning_rate": 1.5303677622963142e-05, + "loss": 0.9733, + "step": 21985 + }, + { + "epoch": 1.41, + "learning_rate": 1.530346293394019e-05, + "loss": 0.9186, + "step": 21986 + }, + { + "epoch": 1.41, + "learning_rate": 1.530324824491724e-05, + "loss": 0.9851, + "step": 21987 + }, + { + "epoch": 1.41, + "learning_rate": 1.530303355589429e-05, + "loss": 0.9808, + "step": 21988 + }, + { + "epoch": 1.41, + "learning_rate": 1.5302818866871338e-05, + "loss": 0.8871, + "step": 21989 + }, + { + "epoch": 1.41, + "learning_rate": 1.5302604177848387e-05, + "loss": 0.9172, + "step": 21990 + }, + { + "epoch": 1.41, + "learning_rate": 1.530238948882544e-05, + "loss": 0.9333, + "step": 21991 + }, + { + "epoch": 1.41, + "learning_rate": 1.5302174799802488e-05, + "loss": 0.9508, + "step": 21992 + }, + { + "epoch": 1.41, + "learning_rate": 1.5301960110779537e-05, + "loss": 1.0486, + "step": 21993 + }, + { + "epoch": 1.42, + "learning_rate": 1.5301745421756586e-05, + "loss": 0.9911, + "step": 21994 + }, + { + "epoch": 1.42, + "learning_rate": 1.5301530732733635e-05, + "loss": 0.9444, + "step": 21995 + }, + { + "epoch": 1.42, + "learning_rate": 1.5301316043710687e-05, + "loss": 1.0713, + "step": 21996 + }, + { + "epoch": 1.42, + "learning_rate": 1.5301101354687736e-05, + "loss": 0.9938, + "step": 21997 + }, + { + "epoch": 1.42, + "learning_rate": 1.5300886665664785e-05, + "loss": 0.9098, + "step": 21998 + }, + { + "epoch": 1.42, + "learning_rate": 1.5300671976641837e-05, + "loss": 0.9813, + "step": 21999 + }, + { + "epoch": 1.42, + "learning_rate": 1.5300457287618886e-05, + "loss": 0.9193, + "step": 22000 + }, + { + "epoch": 1.42, + "eval_laion/cider": 0.5909224865620593, + "eval_laion/meteor": 0.4028081781008874, + "eval_laion/pred_num_words": 37.56640625, + "eval_laion/sacrebleu": 16.77001991790098, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0519216060638428, + "eval_runtime": 99.9337, + "eval_samples_per_second": 4.993, + "eval_steps_per_second": 0.32, + "step": 22000 + }, + { + "epoch": 1.42, + "learning_rate": 1.5300242598595935e-05, + "loss": 0.9094, + "step": 22001 + }, + { + "epoch": 1.42, + "learning_rate": 1.5300027909572984e-05, + "loss": 1.0222, + "step": 22002 + }, + { + "epoch": 1.42, + "learning_rate": 1.5299813220550033e-05, + "loss": 0.9859, + "step": 22003 + }, + { + "epoch": 1.42, + "learning_rate": 1.5299598531527086e-05, + "loss": 0.9843, + "step": 22004 + }, + { + "epoch": 1.42, + "learning_rate": 1.5299383842504135e-05, + "loss": 0.9832, + "step": 22005 + }, + { + "epoch": 1.42, + "learning_rate": 1.5299169153481184e-05, + "loss": 1.0088, + "step": 22006 + }, + { + "epoch": 1.42, + "learning_rate": 1.5298954464458233e-05, + "loss": 1.0526, + "step": 22007 + }, + { + "epoch": 1.42, + "learning_rate": 1.529873977543528e-05, + "loss": 0.9658, + "step": 22008 + }, + { + "epoch": 1.42, + "learning_rate": 1.5298525086412334e-05, + "loss": 0.948, + "step": 22009 + }, + { + "epoch": 1.42, + "learning_rate": 1.5298310397389383e-05, + "loss": 0.9501, + "step": 22010 + }, + { + "epoch": 1.42, + "learning_rate": 1.5298095708366432e-05, + "loss": 0.8893, + "step": 22011 + }, + { + "epoch": 1.42, + "learning_rate": 1.529788101934348e-05, + "loss": 0.918, + "step": 22012 + }, + { + "epoch": 1.42, + "learning_rate": 1.5297666330320533e-05, + "loss": 0.9809, + "step": 22013 + }, + { + "epoch": 1.42, + "learning_rate": 1.5297451641297582e-05, + "loss": 1.0454, + "step": 22014 + }, + { + "epoch": 1.42, + "learning_rate": 1.529723695227463e-05, + "loss": 0.9338, + "step": 22015 + }, + { + "epoch": 1.42, + "learning_rate": 1.5297022263251683e-05, + "loss": 0.9586, + "step": 22016 + }, + { + "epoch": 1.42, + "learning_rate": 1.5296807574228732e-05, + "loss": 1.0544, + "step": 22017 + }, + { + "epoch": 1.42, + "learning_rate": 1.529659288520578e-05, + "loss": 0.9425, + "step": 22018 + }, + { + "epoch": 1.42, + "learning_rate": 1.529637819618283e-05, + "loss": 1.0172, + "step": 22019 + }, + { + "epoch": 1.42, + "learning_rate": 1.529616350715988e-05, + "loss": 0.9087, + "step": 22020 + }, + { + "epoch": 1.42, + "learning_rate": 1.5295948818136928e-05, + "loss": 1.0237, + "step": 22021 + }, + { + "epoch": 1.42, + "learning_rate": 1.529573412911398e-05, + "loss": 0.9182, + "step": 22022 + }, + { + "epoch": 1.42, + "learning_rate": 1.529551944009103e-05, + "loss": 0.9985, + "step": 22023 + }, + { + "epoch": 1.42, + "learning_rate": 1.529530475106808e-05, + "loss": 1.0563, + "step": 22024 + }, + { + "epoch": 1.42, + "learning_rate": 1.5295090062045127e-05, + "loss": 0.9822, + "step": 22025 + }, + { + "epoch": 1.42, + "learning_rate": 1.5294875373022176e-05, + "loss": 1.0117, + "step": 22026 + }, + { + "epoch": 1.42, + "learning_rate": 1.529466068399923e-05, + "loss": 0.9285, + "step": 22027 + }, + { + "epoch": 1.42, + "learning_rate": 1.5294445994976278e-05, + "loss": 0.9782, + "step": 22028 + }, + { + "epoch": 1.42, + "learning_rate": 1.529423130595333e-05, + "loss": 1.0224, + "step": 22029 + }, + { + "epoch": 1.42, + "learning_rate": 1.529401661693038e-05, + "loss": 0.9778, + "step": 22030 + }, + { + "epoch": 1.42, + "learning_rate": 1.5293801927907428e-05, + "loss": 0.9764, + "step": 22031 + }, + { + "epoch": 1.42, + "learning_rate": 1.5293587238884477e-05, + "loss": 0.9267, + "step": 22032 + }, + { + "epoch": 1.42, + "learning_rate": 1.5293372549861526e-05, + "loss": 0.913, + "step": 22033 + }, + { + "epoch": 1.42, + "learning_rate": 1.5293157860838578e-05, + "loss": 0.9768, + "step": 22034 + }, + { + "epoch": 1.42, + "learning_rate": 1.5292943171815627e-05, + "loss": 0.9565, + "step": 22035 + }, + { + "epoch": 1.42, + "learning_rate": 1.5292728482792676e-05, + "loss": 0.8786, + "step": 22036 + }, + { + "epoch": 1.42, + "learning_rate": 1.5292513793769725e-05, + "loss": 1.0206, + "step": 22037 + }, + { + "epoch": 1.42, + "learning_rate": 1.5292299104746774e-05, + "loss": 0.961, + "step": 22038 + }, + { + "epoch": 1.42, + "learning_rate": 1.5292084415723823e-05, + "loss": 0.9747, + "step": 22039 + }, + { + "epoch": 1.42, + "learning_rate": 1.5291869726700875e-05, + "loss": 0.9782, + "step": 22040 + }, + { + "epoch": 1.42, + "learning_rate": 1.5291655037677924e-05, + "loss": 0.9842, + "step": 22041 + }, + { + "epoch": 1.42, + "learning_rate": 1.5291440348654977e-05, + "loss": 1.0296, + "step": 22042 + }, + { + "epoch": 1.42, + "learning_rate": 1.5291225659632026e-05, + "loss": 1.0061, + "step": 22043 + }, + { + "epoch": 1.42, + "learning_rate": 1.5291010970609075e-05, + "loss": 0.983, + "step": 22044 + }, + { + "epoch": 1.42, + "learning_rate": 1.5290796281586124e-05, + "loss": 0.9484, + "step": 22045 + }, + { + "epoch": 1.42, + "learning_rate": 1.5290581592563173e-05, + "loss": 0.9816, + "step": 22046 + }, + { + "epoch": 1.42, + "learning_rate": 1.5290366903540225e-05, + "loss": 0.9336, + "step": 22047 + }, + { + "epoch": 1.42, + "learning_rate": 1.5290152214517274e-05, + "loss": 1.0974, + "step": 22048 + }, + { + "epoch": 1.42, + "learning_rate": 1.5289937525494323e-05, + "loss": 1.0415, + "step": 22049 + }, + { + "epoch": 1.42, + "learning_rate": 1.5289722836471372e-05, + "loss": 0.9693, + "step": 22050 + }, + { + "epoch": 1.42, + "learning_rate": 1.528950814744842e-05, + "loss": 0.9802, + "step": 22051 + }, + { + "epoch": 1.42, + "learning_rate": 1.5289293458425473e-05, + "loss": 0.943, + "step": 22052 + }, + { + "epoch": 1.42, + "learning_rate": 1.5289078769402522e-05, + "loss": 1.0165, + "step": 22053 + }, + { + "epoch": 1.42, + "learning_rate": 1.528886408037957e-05, + "loss": 0.9902, + "step": 22054 + }, + { + "epoch": 1.42, + "learning_rate": 1.528864939135662e-05, + "loss": 0.9259, + "step": 22055 + }, + { + "epoch": 1.42, + "learning_rate": 1.528843470233367e-05, + "loss": 1.0022, + "step": 22056 + }, + { + "epoch": 1.42, + "learning_rate": 1.528822001331072e-05, + "loss": 0.9458, + "step": 22057 + }, + { + "epoch": 1.42, + "learning_rate": 1.528800532428777e-05, + "loss": 0.9502, + "step": 22058 + }, + { + "epoch": 1.42, + "learning_rate": 1.528779063526482e-05, + "loss": 0.9447, + "step": 22059 + }, + { + "epoch": 1.42, + "learning_rate": 1.528757594624187e-05, + "loss": 1.0163, + "step": 22060 + }, + { + "epoch": 1.42, + "learning_rate": 1.528736125721892e-05, + "loss": 1.0388, + "step": 22061 + }, + { + "epoch": 1.42, + "learning_rate": 1.528714656819597e-05, + "loss": 0.9569, + "step": 22062 + }, + { + "epoch": 1.42, + "learning_rate": 1.528693187917302e-05, + "loss": 0.9478, + "step": 22063 + }, + { + "epoch": 1.42, + "learning_rate": 1.5286717190150068e-05, + "loss": 0.9911, + "step": 22064 + }, + { + "epoch": 1.42, + "learning_rate": 1.528650250112712e-05, + "loss": 1.0394, + "step": 22065 + }, + { + "epoch": 1.42, + "learning_rate": 1.528628781210417e-05, + "loss": 0.8575, + "step": 22066 + }, + { + "epoch": 1.42, + "learning_rate": 1.5286073123081218e-05, + "loss": 0.9612, + "step": 22067 + }, + { + "epoch": 1.42, + "learning_rate": 1.5285858434058267e-05, + "loss": 0.9758, + "step": 22068 + }, + { + "epoch": 1.42, + "learning_rate": 1.5285643745035316e-05, + "loss": 1.0508, + "step": 22069 + }, + { + "epoch": 1.42, + "learning_rate": 1.5285429056012365e-05, + "loss": 0.9409, + "step": 22070 + }, + { + "epoch": 1.42, + "learning_rate": 1.5285214366989417e-05, + "loss": 1.0107, + "step": 22071 + }, + { + "epoch": 1.42, + "learning_rate": 1.528499967796647e-05, + "loss": 0.9827, + "step": 22072 + }, + { + "epoch": 1.42, + "learning_rate": 1.528478498894352e-05, + "loss": 0.9677, + "step": 22073 + }, + { + "epoch": 1.42, + "learning_rate": 1.5284570299920567e-05, + "loss": 0.9917, + "step": 22074 + }, + { + "epoch": 1.42, + "learning_rate": 1.5284355610897616e-05, + "loss": 0.9545, + "step": 22075 + }, + { + "epoch": 1.42, + "learning_rate": 1.5284140921874665e-05, + "loss": 1.0651, + "step": 22076 + }, + { + "epoch": 1.42, + "learning_rate": 1.5283926232851714e-05, + "loss": 0.976, + "step": 22077 + }, + { + "epoch": 1.42, + "learning_rate": 1.5283711543828767e-05, + "loss": 1.01, + "step": 22078 + }, + { + "epoch": 1.42, + "learning_rate": 1.5283496854805816e-05, + "loss": 0.9688, + "step": 22079 + }, + { + "epoch": 1.42, + "learning_rate": 1.5283282165782865e-05, + "loss": 0.9469, + "step": 22080 + }, + { + "epoch": 1.42, + "learning_rate": 1.5283067476759914e-05, + "loss": 1.0347, + "step": 22081 + }, + { + "epoch": 1.42, + "learning_rate": 1.5282852787736962e-05, + "loss": 1.018, + "step": 22082 + }, + { + "epoch": 1.42, + "learning_rate": 1.5282638098714015e-05, + "loss": 0.9795, + "step": 22083 + }, + { + "epoch": 1.42, + "learning_rate": 1.5282423409691064e-05, + "loss": 0.9311, + "step": 22084 + }, + { + "epoch": 1.42, + "learning_rate": 1.5282208720668113e-05, + "loss": 0.9843, + "step": 22085 + }, + { + "epoch": 1.42, + "learning_rate": 1.5281994031645165e-05, + "loss": 0.8957, + "step": 22086 + }, + { + "epoch": 1.42, + "learning_rate": 1.5281779342622214e-05, + "loss": 0.9797, + "step": 22087 + }, + { + "epoch": 1.42, + "learning_rate": 1.5281564653599263e-05, + "loss": 1.0177, + "step": 22088 + }, + { + "epoch": 1.42, + "learning_rate": 1.5281349964576312e-05, + "loss": 0.9797, + "step": 22089 + }, + { + "epoch": 1.42, + "learning_rate": 1.528113527555336e-05, + "loss": 0.9995, + "step": 22090 + }, + { + "epoch": 1.42, + "learning_rate": 1.5280920586530413e-05, + "loss": 0.9282, + "step": 22091 + }, + { + "epoch": 1.42, + "learning_rate": 1.5280705897507462e-05, + "loss": 0.8649, + "step": 22092 + }, + { + "epoch": 1.42, + "learning_rate": 1.528049120848451e-05, + "loss": 1.0463, + "step": 22093 + }, + { + "epoch": 1.42, + "learning_rate": 1.528027651946156e-05, + "loss": 0.996, + "step": 22094 + }, + { + "epoch": 1.42, + "learning_rate": 1.528006183043861e-05, + "loss": 1.0164, + "step": 22095 + }, + { + "epoch": 1.42, + "learning_rate": 1.527984714141566e-05, + "loss": 0.9084, + "step": 22096 + }, + { + "epoch": 1.42, + "learning_rate": 1.527963245239271e-05, + "loss": 1.0222, + "step": 22097 + }, + { + "epoch": 1.42, + "learning_rate": 1.527941776336976e-05, + "loss": 0.9927, + "step": 22098 + }, + { + "epoch": 1.42, + "learning_rate": 1.527920307434681e-05, + "loss": 0.9423, + "step": 22099 + }, + { + "epoch": 1.42, + "learning_rate": 1.527898838532386e-05, + "loss": 1.0098, + "step": 22100 + }, + { + "epoch": 1.42, + "learning_rate": 1.527877369630091e-05, + "loss": 0.9989, + "step": 22101 + }, + { + "epoch": 1.42, + "learning_rate": 1.527855900727796e-05, + "loss": 0.9642, + "step": 22102 + }, + { + "epoch": 1.42, + "learning_rate": 1.527834431825501e-05, + "loss": 1.0461, + "step": 22103 + }, + { + "epoch": 1.42, + "learning_rate": 1.527812962923206e-05, + "loss": 0.8815, + "step": 22104 + }, + { + "epoch": 1.42, + "learning_rate": 1.527791494020911e-05, + "loss": 0.9455, + "step": 22105 + }, + { + "epoch": 1.42, + "learning_rate": 1.5277700251186158e-05, + "loss": 0.9619, + "step": 22106 + }, + { + "epoch": 1.42, + "learning_rate": 1.5277485562163207e-05, + "loss": 0.9084, + "step": 22107 + }, + { + "epoch": 1.42, + "learning_rate": 1.5277270873140256e-05, + "loss": 1.0023, + "step": 22108 + }, + { + "epoch": 1.42, + "learning_rate": 1.5277056184117308e-05, + "loss": 0.9675, + "step": 22109 + }, + { + "epoch": 1.42, + "learning_rate": 1.5276841495094357e-05, + "loss": 0.9139, + "step": 22110 + }, + { + "epoch": 1.42, + "learning_rate": 1.5276626806071406e-05, + "loss": 0.9411, + "step": 22111 + }, + { + "epoch": 1.42, + "learning_rate": 1.5276412117048455e-05, + "loss": 0.9186, + "step": 22112 + }, + { + "epoch": 1.42, + "learning_rate": 1.5276197428025504e-05, + "loss": 0.9493, + "step": 22113 + }, + { + "epoch": 1.42, + "learning_rate": 1.5275982739002556e-05, + "loss": 1.0114, + "step": 22114 + }, + { + "epoch": 1.42, + "learning_rate": 1.5275768049979605e-05, + "loss": 0.9394, + "step": 22115 + }, + { + "epoch": 1.42, + "learning_rate": 1.5275553360956658e-05, + "loss": 0.9446, + "step": 22116 + }, + { + "epoch": 1.42, + "learning_rate": 1.5275338671933707e-05, + "loss": 0.9249, + "step": 22117 + }, + { + "epoch": 1.42, + "learning_rate": 1.5275123982910756e-05, + "loss": 1.0058, + "step": 22118 + }, + { + "epoch": 1.42, + "learning_rate": 1.5274909293887805e-05, + "loss": 1.0538, + "step": 22119 + }, + { + "epoch": 1.42, + "learning_rate": 1.5274694604864854e-05, + "loss": 1.0038, + "step": 22120 + }, + { + "epoch": 1.42, + "learning_rate": 1.5274479915841906e-05, + "loss": 0.9897, + "step": 22121 + }, + { + "epoch": 1.42, + "learning_rate": 1.5274265226818955e-05, + "loss": 0.9731, + "step": 22122 + }, + { + "epoch": 1.42, + "learning_rate": 1.5274050537796004e-05, + "loss": 1.0693, + "step": 22123 + }, + { + "epoch": 1.42, + "learning_rate": 1.5273835848773053e-05, + "loss": 0.8608, + "step": 22124 + }, + { + "epoch": 1.42, + "learning_rate": 1.5273621159750102e-05, + "loss": 0.9036, + "step": 22125 + }, + { + "epoch": 1.42, + "learning_rate": 1.527340647072715e-05, + "loss": 0.9606, + "step": 22126 + }, + { + "epoch": 1.42, + "learning_rate": 1.5273191781704203e-05, + "loss": 0.9449, + "step": 22127 + }, + { + "epoch": 1.42, + "learning_rate": 1.5272977092681252e-05, + "loss": 0.9526, + "step": 22128 + }, + { + "epoch": 1.42, + "learning_rate": 1.52727624036583e-05, + "loss": 0.9398, + "step": 22129 + }, + { + "epoch": 1.42, + "learning_rate": 1.5272547714635353e-05, + "loss": 1.0376, + "step": 22130 + }, + { + "epoch": 1.42, + "learning_rate": 1.5272333025612402e-05, + "loss": 0.9629, + "step": 22131 + }, + { + "epoch": 1.42, + "learning_rate": 1.527211833658945e-05, + "loss": 0.9412, + "step": 22132 + }, + { + "epoch": 1.42, + "learning_rate": 1.52719036475665e-05, + "loss": 0.978, + "step": 22133 + }, + { + "epoch": 1.42, + "learning_rate": 1.5271688958543553e-05, + "loss": 0.996, + "step": 22134 + }, + { + "epoch": 1.42, + "learning_rate": 1.52714742695206e-05, + "loss": 1.0028, + "step": 22135 + }, + { + "epoch": 1.42, + "learning_rate": 1.527125958049765e-05, + "loss": 1.0278, + "step": 22136 + }, + { + "epoch": 1.42, + "learning_rate": 1.52710448914747e-05, + "loss": 1.0546, + "step": 22137 + }, + { + "epoch": 1.42, + "learning_rate": 1.527083020245175e-05, + "loss": 0.9217, + "step": 22138 + }, + { + "epoch": 1.42, + "learning_rate": 1.52706155134288e-05, + "loss": 0.9304, + "step": 22139 + }, + { + "epoch": 1.42, + "learning_rate": 1.527040082440585e-05, + "loss": 0.9998, + "step": 22140 + }, + { + "epoch": 1.42, + "learning_rate": 1.52701861353829e-05, + "loss": 0.9497, + "step": 22141 + }, + { + "epoch": 1.42, + "learning_rate": 1.5269971446359948e-05, + "loss": 1.004, + "step": 22142 + }, + { + "epoch": 1.42, + "learning_rate": 1.5269756757336997e-05, + "loss": 0.9327, + "step": 22143 + }, + { + "epoch": 1.42, + "learning_rate": 1.526954206831405e-05, + "loss": 0.9056, + "step": 22144 + }, + { + "epoch": 1.42, + "learning_rate": 1.5269327379291098e-05, + "loss": 0.9076, + "step": 22145 + }, + { + "epoch": 1.42, + "learning_rate": 1.5269112690268147e-05, + "loss": 0.9715, + "step": 22146 + }, + { + "epoch": 1.42, + "learning_rate": 1.52688980012452e-05, + "loss": 1.0018, + "step": 22147 + }, + { + "epoch": 1.42, + "learning_rate": 1.526868331222225e-05, + "loss": 0.938, + "step": 22148 + }, + { + "epoch": 1.43, + "learning_rate": 1.5268468623199297e-05, + "loss": 0.9768, + "step": 22149 + }, + { + "epoch": 1.43, + "learning_rate": 1.5268253934176346e-05, + "loss": 1.0328, + "step": 22150 + }, + { + "epoch": 1.43, + "learning_rate": 1.5268039245153395e-05, + "loss": 0.9195, + "step": 22151 + }, + { + "epoch": 1.43, + "learning_rate": 1.5267824556130448e-05, + "loss": 0.9806, + "step": 22152 + }, + { + "epoch": 1.43, + "learning_rate": 1.5267609867107497e-05, + "loss": 1.0078, + "step": 22153 + }, + { + "epoch": 1.43, + "learning_rate": 1.5267395178084546e-05, + "loss": 1.0062, + "step": 22154 + }, + { + "epoch": 1.43, + "learning_rate": 1.5267180489061595e-05, + "loss": 0.9334, + "step": 22155 + }, + { + "epoch": 1.43, + "learning_rate": 1.5266965800038643e-05, + "loss": 0.9398, + "step": 22156 + }, + { + "epoch": 1.43, + "learning_rate": 1.5266751111015692e-05, + "loss": 0.9575, + "step": 22157 + }, + { + "epoch": 1.43, + "learning_rate": 1.5266536421992745e-05, + "loss": 0.9493, + "step": 22158 + }, + { + "epoch": 1.43, + "learning_rate": 1.5266321732969797e-05, + "loss": 0.9322, + "step": 22159 + }, + { + "epoch": 1.43, + "learning_rate": 1.5266107043946846e-05, + "loss": 0.9604, + "step": 22160 + }, + { + "epoch": 1.43, + "learning_rate": 1.5265892354923895e-05, + "loss": 0.9533, + "step": 22161 + }, + { + "epoch": 1.43, + "learning_rate": 1.5265677665900944e-05, + "loss": 0.8819, + "step": 22162 + }, + { + "epoch": 1.43, + "learning_rate": 1.5265462976877993e-05, + "loss": 0.8981, + "step": 22163 + }, + { + "epoch": 1.43, + "learning_rate": 1.5265248287855042e-05, + "loss": 0.957, + "step": 22164 + }, + { + "epoch": 1.43, + "learning_rate": 1.5265033598832094e-05, + "loss": 0.9332, + "step": 22165 + }, + { + "epoch": 1.43, + "learning_rate": 1.5264818909809143e-05, + "loss": 1.0403, + "step": 22166 + }, + { + "epoch": 1.43, + "learning_rate": 1.5264604220786192e-05, + "loss": 0.9569, + "step": 22167 + }, + { + "epoch": 1.43, + "learning_rate": 1.526438953176324e-05, + "loss": 0.9592, + "step": 22168 + }, + { + "epoch": 1.43, + "learning_rate": 1.526417484274029e-05, + "loss": 0.95, + "step": 22169 + }, + { + "epoch": 1.43, + "learning_rate": 1.5263960153717343e-05, + "loss": 0.8685, + "step": 22170 + }, + { + "epoch": 1.43, + "learning_rate": 1.526374546469439e-05, + "loss": 0.9547, + "step": 22171 + }, + { + "epoch": 1.43, + "learning_rate": 1.526353077567144e-05, + "loss": 0.9801, + "step": 22172 + }, + { + "epoch": 1.43, + "learning_rate": 1.5263316086648493e-05, + "loss": 0.996, + "step": 22173 + }, + { + "epoch": 1.43, + "learning_rate": 1.5263101397625542e-05, + "loss": 0.9694, + "step": 22174 + }, + { + "epoch": 1.43, + "learning_rate": 1.526288670860259e-05, + "loss": 0.9518, + "step": 22175 + }, + { + "epoch": 1.43, + "learning_rate": 1.526267201957964e-05, + "loss": 0.9914, + "step": 22176 + }, + { + "epoch": 1.43, + "learning_rate": 1.526245733055669e-05, + "loss": 1.0241, + "step": 22177 + }, + { + "epoch": 1.43, + "learning_rate": 1.526224264153374e-05, + "loss": 0.9613, + "step": 22178 + }, + { + "epoch": 1.43, + "learning_rate": 1.526202795251079e-05, + "loss": 0.9911, + "step": 22179 + }, + { + "epoch": 1.43, + "learning_rate": 1.526181326348784e-05, + "loss": 0.9161, + "step": 22180 + }, + { + "epoch": 1.43, + "learning_rate": 1.5261598574464888e-05, + "loss": 1.0182, + "step": 22181 + }, + { + "epoch": 1.43, + "learning_rate": 1.5261383885441937e-05, + "loss": 0.9922, + "step": 22182 + }, + { + "epoch": 1.43, + "learning_rate": 1.526116919641899e-05, + "loss": 1.0195, + "step": 22183 + }, + { + "epoch": 1.43, + "learning_rate": 1.5260954507396038e-05, + "loss": 1.0132, + "step": 22184 + }, + { + "epoch": 1.43, + "learning_rate": 1.5260739818373087e-05, + "loss": 1.0301, + "step": 22185 + }, + { + "epoch": 1.43, + "learning_rate": 1.5260525129350136e-05, + "loss": 0.9318, + "step": 22186 + }, + { + "epoch": 1.43, + "learning_rate": 1.5260310440327185e-05, + "loss": 1.009, + "step": 22187 + }, + { + "epoch": 1.43, + "learning_rate": 1.5260095751304237e-05, + "loss": 0.9789, + "step": 22188 + }, + { + "epoch": 1.43, + "learning_rate": 1.5259881062281286e-05, + "loss": 0.9754, + "step": 22189 + }, + { + "epoch": 1.43, + "learning_rate": 1.525966637325834e-05, + "loss": 0.9748, + "step": 22190 + }, + { + "epoch": 1.43, + "learning_rate": 1.5259451684235388e-05, + "loss": 0.9748, + "step": 22191 + }, + { + "epoch": 1.43, + "learning_rate": 1.5259236995212437e-05, + "loss": 1.023, + "step": 22192 + }, + { + "epoch": 1.43, + "learning_rate": 1.5259022306189486e-05, + "loss": 0.9402, + "step": 22193 + }, + { + "epoch": 1.43, + "learning_rate": 1.5258807617166535e-05, + "loss": 0.9697, + "step": 22194 + }, + { + "epoch": 1.43, + "learning_rate": 1.5258592928143585e-05, + "loss": 0.9503, + "step": 22195 + }, + { + "epoch": 1.43, + "learning_rate": 1.5258378239120634e-05, + "loss": 0.9247, + "step": 22196 + }, + { + "epoch": 1.43, + "learning_rate": 1.5258163550097685e-05, + "loss": 0.9144, + "step": 22197 + }, + { + "epoch": 1.43, + "learning_rate": 1.5257948861074734e-05, + "loss": 0.9911, + "step": 22198 + }, + { + "epoch": 1.43, + "learning_rate": 1.5257734172051783e-05, + "loss": 0.9905, + "step": 22199 + }, + { + "epoch": 1.43, + "learning_rate": 1.5257519483028834e-05, + "loss": 0.9703, + "step": 22200 + }, + { + "epoch": 1.43, + "learning_rate": 1.5257304794005882e-05, + "loss": 1.0172, + "step": 22201 + }, + { + "epoch": 1.43, + "learning_rate": 1.5257090104982935e-05, + "loss": 0.9881, + "step": 22202 + }, + { + "epoch": 1.43, + "learning_rate": 1.5256875415959984e-05, + "loss": 0.9497, + "step": 22203 + }, + { + "epoch": 1.43, + "learning_rate": 1.5256660726937034e-05, + "loss": 0.9736, + "step": 22204 + }, + { + "epoch": 1.43, + "learning_rate": 1.5256446037914083e-05, + "loss": 0.9636, + "step": 22205 + }, + { + "epoch": 1.43, + "learning_rate": 1.5256231348891132e-05, + "loss": 1.0307, + "step": 22206 + }, + { + "epoch": 1.43, + "learning_rate": 1.5256016659868183e-05, + "loss": 0.9472, + "step": 22207 + }, + { + "epoch": 1.43, + "learning_rate": 1.5255801970845232e-05, + "loss": 0.9556, + "step": 22208 + }, + { + "epoch": 1.43, + "learning_rate": 1.5255587281822281e-05, + "loss": 0.9536, + "step": 22209 + }, + { + "epoch": 1.43, + "learning_rate": 1.5255372592799332e-05, + "loss": 0.9502, + "step": 22210 + }, + { + "epoch": 1.43, + "learning_rate": 1.525515790377638e-05, + "loss": 0.9068, + "step": 22211 + }, + { + "epoch": 1.43, + "learning_rate": 1.5254943214753431e-05, + "loss": 0.9919, + "step": 22212 + }, + { + "epoch": 1.43, + "learning_rate": 1.525472852573048e-05, + "loss": 0.9471, + "step": 22213 + }, + { + "epoch": 1.43, + "learning_rate": 1.525451383670753e-05, + "loss": 0.9397, + "step": 22214 + }, + { + "epoch": 1.43, + "learning_rate": 1.525429914768458e-05, + "loss": 0.9424, + "step": 22215 + }, + { + "epoch": 1.43, + "learning_rate": 1.5254084458661629e-05, + "loss": 0.9258, + "step": 22216 + }, + { + "epoch": 1.43, + "learning_rate": 1.5253869769638681e-05, + "loss": 0.9723, + "step": 22217 + }, + { + "epoch": 1.43, + "learning_rate": 1.525365508061573e-05, + "loss": 0.9386, + "step": 22218 + }, + { + "epoch": 1.43, + "learning_rate": 1.5253440391592779e-05, + "loss": 0.9417, + "step": 22219 + }, + { + "epoch": 1.43, + "learning_rate": 1.525322570256983e-05, + "loss": 0.9672, + "step": 22220 + }, + { + "epoch": 1.43, + "learning_rate": 1.5253011013546879e-05, + "loss": 0.955, + "step": 22221 + }, + { + "epoch": 1.43, + "learning_rate": 1.525279632452393e-05, + "loss": 0.9895, + "step": 22222 + }, + { + "epoch": 1.43, + "learning_rate": 1.5252581635500978e-05, + "loss": 1.0409, + "step": 22223 + }, + { + "epoch": 1.43, + "learning_rate": 1.5252366946478027e-05, + "loss": 0.952, + "step": 22224 + }, + { + "epoch": 1.43, + "learning_rate": 1.5252152257455078e-05, + "loss": 0.9454, + "step": 22225 + }, + { + "epoch": 1.43, + "learning_rate": 1.5251937568432127e-05, + "loss": 0.9513, + "step": 22226 + }, + { + "epoch": 1.43, + "learning_rate": 1.5251722879409176e-05, + "loss": 1.0206, + "step": 22227 + }, + { + "epoch": 1.43, + "learning_rate": 1.5251508190386227e-05, + "loss": 1.0164, + "step": 22228 + }, + { + "epoch": 1.43, + "learning_rate": 1.5251293501363276e-05, + "loss": 0.9831, + "step": 22229 + }, + { + "epoch": 1.43, + "learning_rate": 1.5251078812340324e-05, + "loss": 0.9554, + "step": 22230 + }, + { + "epoch": 1.43, + "learning_rate": 1.5250864123317377e-05, + "loss": 0.9569, + "step": 22231 + }, + { + "epoch": 1.43, + "learning_rate": 1.5250649434294427e-05, + "loss": 1.0223, + "step": 22232 + }, + { + "epoch": 1.43, + "learning_rate": 1.5250434745271476e-05, + "loss": 0.9575, + "step": 22233 + }, + { + "epoch": 1.43, + "learning_rate": 1.5250220056248525e-05, + "loss": 0.9925, + "step": 22234 + }, + { + "epoch": 1.43, + "learning_rate": 1.5250005367225576e-05, + "loss": 1.0571, + "step": 22235 + }, + { + "epoch": 1.43, + "learning_rate": 1.5249790678202625e-05, + "loss": 0.9744, + "step": 22236 + }, + { + "epoch": 1.43, + "learning_rate": 1.5249575989179674e-05, + "loss": 0.9752, + "step": 22237 + }, + { + "epoch": 1.43, + "learning_rate": 1.5249361300156725e-05, + "loss": 0.893, + "step": 22238 + }, + { + "epoch": 1.43, + "learning_rate": 1.5249146611133774e-05, + "loss": 1.0727, + "step": 22239 + }, + { + "epoch": 1.43, + "learning_rate": 1.5248931922110823e-05, + "loss": 1.0239, + "step": 22240 + }, + { + "epoch": 1.43, + "learning_rate": 1.5248717233087873e-05, + "loss": 0.9308, + "step": 22241 + }, + { + "epoch": 1.43, + "learning_rate": 1.5248502544064922e-05, + "loss": 1.0433, + "step": 22242 + }, + { + "epoch": 1.43, + "learning_rate": 1.5248287855041973e-05, + "loss": 1.0095, + "step": 22243 + }, + { + "epoch": 1.43, + "learning_rate": 1.5248073166019022e-05, + "loss": 1.0117, + "step": 22244 + }, + { + "epoch": 1.43, + "learning_rate": 1.524785847699607e-05, + "loss": 1.0607, + "step": 22245 + }, + { + "epoch": 1.43, + "learning_rate": 1.5247643787973123e-05, + "loss": 0.9935, + "step": 22246 + }, + { + "epoch": 1.43, + "learning_rate": 1.5247429098950172e-05, + "loss": 1.0697, + "step": 22247 + }, + { + "epoch": 1.43, + "learning_rate": 1.5247214409927223e-05, + "loss": 0.984, + "step": 22248 + }, + { + "epoch": 1.43, + "learning_rate": 1.5246999720904272e-05, + "loss": 0.9661, + "step": 22249 + }, + { + "epoch": 1.43, + "learning_rate": 1.524678503188132e-05, + "loss": 0.9613, + "step": 22250 + }, + { + "epoch": 1.43, + "learning_rate": 1.5246570342858371e-05, + "loss": 1.0569, + "step": 22251 + }, + { + "epoch": 1.43, + "learning_rate": 1.524635565383542e-05, + "loss": 0.9902, + "step": 22252 + }, + { + "epoch": 1.43, + "learning_rate": 1.5246140964812471e-05, + "loss": 1.07, + "step": 22253 + }, + { + "epoch": 1.43, + "learning_rate": 1.524592627578952e-05, + "loss": 0.9701, + "step": 22254 + }, + { + "epoch": 1.43, + "learning_rate": 1.5245711586766569e-05, + "loss": 0.9028, + "step": 22255 + }, + { + "epoch": 1.43, + "learning_rate": 1.524549689774362e-05, + "loss": 1.0108, + "step": 22256 + }, + { + "epoch": 1.43, + "learning_rate": 1.5245282208720669e-05, + "loss": 0.9041, + "step": 22257 + }, + { + "epoch": 1.43, + "learning_rate": 1.5245067519697718e-05, + "loss": 0.9782, + "step": 22258 + }, + { + "epoch": 1.43, + "learning_rate": 1.5244852830674768e-05, + "loss": 0.9554, + "step": 22259 + }, + { + "epoch": 1.43, + "learning_rate": 1.5244638141651819e-05, + "loss": 1.0285, + "step": 22260 + }, + { + "epoch": 1.43, + "learning_rate": 1.524442345262887e-05, + "loss": 0.9518, + "step": 22261 + }, + { + "epoch": 1.43, + "learning_rate": 1.5244208763605918e-05, + "loss": 0.9706, + "step": 22262 + }, + { + "epoch": 1.43, + "learning_rate": 1.5243994074582969e-05, + "loss": 0.9665, + "step": 22263 + }, + { + "epoch": 1.43, + "learning_rate": 1.5243779385560018e-05, + "loss": 0.9736, + "step": 22264 + }, + { + "epoch": 1.43, + "learning_rate": 1.5243564696537067e-05, + "loss": 1.0838, + "step": 22265 + }, + { + "epoch": 1.43, + "learning_rate": 1.5243350007514118e-05, + "loss": 0.9505, + "step": 22266 + }, + { + "epoch": 1.43, + "learning_rate": 1.5243135318491167e-05, + "loss": 0.9631, + "step": 22267 + }, + { + "epoch": 1.43, + "learning_rate": 1.5242920629468216e-05, + "loss": 1.0333, + "step": 22268 + }, + { + "epoch": 1.43, + "learning_rate": 1.5242705940445266e-05, + "loss": 1.0028, + "step": 22269 + }, + { + "epoch": 1.43, + "learning_rate": 1.5242491251422315e-05, + "loss": 0.9668, + "step": 22270 + }, + { + "epoch": 1.43, + "learning_rate": 1.5242276562399366e-05, + "loss": 1.029, + "step": 22271 + }, + { + "epoch": 1.43, + "learning_rate": 1.5242061873376415e-05, + "loss": 1.0035, + "step": 22272 + }, + { + "epoch": 1.43, + "learning_rate": 1.5241847184353464e-05, + "loss": 0.8744, + "step": 22273 + }, + { + "epoch": 1.43, + "learning_rate": 1.5241632495330515e-05, + "loss": 0.9266, + "step": 22274 + }, + { + "epoch": 1.43, + "learning_rate": 1.5241417806307565e-05, + "loss": 0.9466, + "step": 22275 + }, + { + "epoch": 1.43, + "learning_rate": 1.5241203117284616e-05, + "loss": 0.9524, + "step": 22276 + }, + { + "epoch": 1.43, + "learning_rate": 1.5240988428261665e-05, + "loss": 0.9333, + "step": 22277 + }, + { + "epoch": 1.43, + "learning_rate": 1.5240773739238714e-05, + "loss": 1.0194, + "step": 22278 + }, + { + "epoch": 1.43, + "learning_rate": 1.5240559050215764e-05, + "loss": 0.9106, + "step": 22279 + }, + { + "epoch": 1.43, + "learning_rate": 1.5240344361192813e-05, + "loss": 0.931, + "step": 22280 + }, + { + "epoch": 1.43, + "learning_rate": 1.5240129672169864e-05, + "loss": 0.9612, + "step": 22281 + }, + { + "epoch": 1.43, + "learning_rate": 1.5239914983146913e-05, + "loss": 1.0435, + "step": 22282 + }, + { + "epoch": 1.43, + "learning_rate": 1.5239700294123962e-05, + "loss": 1.0255, + "step": 22283 + }, + { + "epoch": 1.43, + "learning_rate": 1.5239485605101013e-05, + "loss": 0.8828, + "step": 22284 + }, + { + "epoch": 1.43, + "learning_rate": 1.5239270916078062e-05, + "loss": 0.9687, + "step": 22285 + }, + { + "epoch": 1.43, + "learning_rate": 1.523905622705511e-05, + "loss": 0.9804, + "step": 22286 + }, + { + "epoch": 1.43, + "learning_rate": 1.5238841538032161e-05, + "loss": 0.9845, + "step": 22287 + }, + { + "epoch": 1.43, + "learning_rate": 1.523862684900921e-05, + "loss": 0.9895, + "step": 22288 + }, + { + "epoch": 1.43, + "learning_rate": 1.5238412159986263e-05, + "loss": 1.0156, + "step": 22289 + }, + { + "epoch": 1.43, + "learning_rate": 1.5238197470963312e-05, + "loss": 0.9789, + "step": 22290 + }, + { + "epoch": 1.43, + "learning_rate": 1.5237982781940362e-05, + "loss": 0.9474, + "step": 22291 + }, + { + "epoch": 1.43, + "learning_rate": 1.5237768092917411e-05, + "loss": 1.0126, + "step": 22292 + }, + { + "epoch": 1.43, + "learning_rate": 1.523755340389446e-05, + "loss": 0.9935, + "step": 22293 + }, + { + "epoch": 1.43, + "learning_rate": 1.523733871487151e-05, + "loss": 0.9404, + "step": 22294 + }, + { + "epoch": 1.43, + "learning_rate": 1.523712402584856e-05, + "loss": 0.9576, + "step": 22295 + }, + { + "epoch": 1.43, + "learning_rate": 1.5236909336825609e-05, + "loss": 1.0161, + "step": 22296 + }, + { + "epoch": 1.43, + "learning_rate": 1.523669464780266e-05, + "loss": 0.9564, + "step": 22297 + }, + { + "epoch": 1.43, + "learning_rate": 1.5236479958779708e-05, + "loss": 0.9418, + "step": 22298 + }, + { + "epoch": 1.43, + "learning_rate": 1.5236265269756759e-05, + "loss": 0.882, + "step": 22299 + }, + { + "epoch": 1.43, + "learning_rate": 1.5236050580733808e-05, + "loss": 1.0063, + "step": 22300 + }, + { + "epoch": 1.43, + "learning_rate": 1.5235835891710857e-05, + "loss": 0.9716, + "step": 22301 + }, + { + "epoch": 1.43, + "learning_rate": 1.5235621202687908e-05, + "loss": 0.9417, + "step": 22302 + }, + { + "epoch": 1.43, + "learning_rate": 1.5235406513664957e-05, + "loss": 0.9715, + "step": 22303 + }, + { + "epoch": 1.43, + "learning_rate": 1.5235191824642009e-05, + "loss": 0.9648, + "step": 22304 + }, + { + "epoch": 1.44, + "learning_rate": 1.5234977135619058e-05, + "loss": 0.8953, + "step": 22305 + }, + { + "epoch": 1.44, + "learning_rate": 1.5234762446596107e-05, + "loss": 0.9866, + "step": 22306 + }, + { + "epoch": 1.44, + "learning_rate": 1.5234547757573157e-05, + "loss": 0.9402, + "step": 22307 + }, + { + "epoch": 1.44, + "learning_rate": 1.5234333068550206e-05, + "loss": 1.0084, + "step": 22308 + }, + { + "epoch": 1.44, + "learning_rate": 1.5234118379527257e-05, + "loss": 1.0032, + "step": 22309 + }, + { + "epoch": 1.44, + "learning_rate": 1.5233903690504306e-05, + "loss": 0.987, + "step": 22310 + }, + { + "epoch": 1.44, + "learning_rate": 1.5233689001481355e-05, + "loss": 1.0481, + "step": 22311 + }, + { + "epoch": 1.44, + "learning_rate": 1.5233474312458406e-05, + "loss": 0.9544, + "step": 22312 + }, + { + "epoch": 1.44, + "learning_rate": 1.5233259623435455e-05, + "loss": 1.0219, + "step": 22313 + }, + { + "epoch": 1.44, + "learning_rate": 1.5233044934412504e-05, + "loss": 0.9562, + "step": 22314 + }, + { + "epoch": 1.44, + "learning_rate": 1.5232830245389554e-05, + "loss": 1.0005, + "step": 22315 + }, + { + "epoch": 1.44, + "learning_rate": 1.5232615556366603e-05, + "loss": 0.956, + "step": 22316 + }, + { + "epoch": 1.44, + "learning_rate": 1.5232400867343652e-05, + "loss": 0.9538, + "step": 22317 + }, + { + "epoch": 1.44, + "learning_rate": 1.5232186178320705e-05, + "loss": 0.974, + "step": 22318 + }, + { + "epoch": 1.44, + "learning_rate": 1.5231971489297755e-05, + "loss": 0.9889, + "step": 22319 + }, + { + "epoch": 1.44, + "learning_rate": 1.5231756800274804e-05, + "loss": 1.0622, + "step": 22320 + }, + { + "epoch": 1.44, + "learning_rate": 1.5231542111251853e-05, + "loss": 0.964, + "step": 22321 + }, + { + "epoch": 1.44, + "learning_rate": 1.5231327422228904e-05, + "loss": 1.0097, + "step": 22322 + }, + { + "epoch": 1.44, + "learning_rate": 1.5231112733205953e-05, + "loss": 0.9682, + "step": 22323 + }, + { + "epoch": 1.44, + "learning_rate": 1.5230898044183002e-05, + "loss": 1.0025, + "step": 22324 + }, + { + "epoch": 1.44, + "learning_rate": 1.5230683355160052e-05, + "loss": 1.1018, + "step": 22325 + }, + { + "epoch": 1.44, + "learning_rate": 1.5230468666137101e-05, + "loss": 1.0141, + "step": 22326 + }, + { + "epoch": 1.44, + "learning_rate": 1.523025397711415e-05, + "loss": 0.9869, + "step": 22327 + }, + { + "epoch": 1.44, + "learning_rate": 1.5230039288091201e-05, + "loss": 0.9678, + "step": 22328 + }, + { + "epoch": 1.44, + "learning_rate": 1.522982459906825e-05, + "loss": 0.9435, + "step": 22329 + }, + { + "epoch": 1.44, + "learning_rate": 1.52296099100453e-05, + "loss": 0.9834, + "step": 22330 + }, + { + "epoch": 1.44, + "learning_rate": 1.522939522102235e-05, + "loss": 0.9623, + "step": 22331 + }, + { + "epoch": 1.44, + "learning_rate": 1.5229180531999399e-05, + "loss": 1.0345, + "step": 22332 + }, + { + "epoch": 1.44, + "learning_rate": 1.5228965842976451e-05, + "loss": 0.9663, + "step": 22333 + }, + { + "epoch": 1.44, + "learning_rate": 1.52287511539535e-05, + "loss": 0.9555, + "step": 22334 + }, + { + "epoch": 1.44, + "learning_rate": 1.522853646493055e-05, + "loss": 1.0273, + "step": 22335 + }, + { + "epoch": 1.44, + "learning_rate": 1.52283217759076e-05, + "loss": 0.9597, + "step": 22336 + }, + { + "epoch": 1.44, + "learning_rate": 1.5228107086884648e-05, + "loss": 0.9611, + "step": 22337 + }, + { + "epoch": 1.44, + "learning_rate": 1.5227892397861699e-05, + "loss": 1.0137, + "step": 22338 + }, + { + "epoch": 1.44, + "learning_rate": 1.5227677708838748e-05, + "loss": 0.9554, + "step": 22339 + }, + { + "epoch": 1.44, + "learning_rate": 1.5227463019815799e-05, + "loss": 0.9527, + "step": 22340 + }, + { + "epoch": 1.44, + "learning_rate": 1.5227248330792848e-05, + "loss": 0.9604, + "step": 22341 + }, + { + "epoch": 1.44, + "learning_rate": 1.5227033641769897e-05, + "loss": 0.9363, + "step": 22342 + }, + { + "epoch": 1.44, + "learning_rate": 1.5226818952746947e-05, + "loss": 1.0125, + "step": 22343 + }, + { + "epoch": 1.44, + "learning_rate": 1.5226604263723996e-05, + "loss": 1.0185, + "step": 22344 + }, + { + "epoch": 1.44, + "learning_rate": 1.5226389574701045e-05, + "loss": 0.9833, + "step": 22345 + }, + { + "epoch": 1.44, + "learning_rate": 1.5226174885678096e-05, + "loss": 0.8848, + "step": 22346 + }, + { + "epoch": 1.44, + "learning_rate": 1.5225960196655147e-05, + "loss": 1.042, + "step": 22347 + }, + { + "epoch": 1.44, + "learning_rate": 1.5225745507632197e-05, + "loss": 0.9619, + "step": 22348 + }, + { + "epoch": 1.44, + "learning_rate": 1.5225530818609246e-05, + "loss": 0.9909, + "step": 22349 + }, + { + "epoch": 1.44, + "learning_rate": 1.5225316129586297e-05, + "loss": 0.9751, + "step": 22350 + }, + { + "epoch": 1.44, + "learning_rate": 1.5225101440563346e-05, + "loss": 1.0998, + "step": 22351 + }, + { + "epoch": 1.44, + "learning_rate": 1.5224886751540395e-05, + "loss": 0.8981, + "step": 22352 + }, + { + "epoch": 1.44, + "learning_rate": 1.5224672062517445e-05, + "loss": 0.9684, + "step": 22353 + }, + { + "epoch": 1.44, + "learning_rate": 1.5224457373494494e-05, + "loss": 0.9226, + "step": 22354 + }, + { + "epoch": 1.44, + "learning_rate": 1.5224242684471543e-05, + "loss": 0.9926, + "step": 22355 + }, + { + "epoch": 1.44, + "learning_rate": 1.5224027995448594e-05, + "loss": 0.9745, + "step": 22356 + }, + { + "epoch": 1.44, + "learning_rate": 1.5223813306425643e-05, + "loss": 0.9705, + "step": 22357 + }, + { + "epoch": 1.44, + "learning_rate": 1.5223598617402694e-05, + "loss": 0.9058, + "step": 22358 + }, + { + "epoch": 1.44, + "learning_rate": 1.5223383928379743e-05, + "loss": 0.9492, + "step": 22359 + }, + { + "epoch": 1.44, + "learning_rate": 1.5223169239356792e-05, + "loss": 0.9448, + "step": 22360 + }, + { + "epoch": 1.44, + "learning_rate": 1.5222954550333842e-05, + "loss": 0.951, + "step": 22361 + }, + { + "epoch": 1.44, + "learning_rate": 1.5222739861310893e-05, + "loss": 0.9673, + "step": 22362 + }, + { + "epoch": 1.44, + "learning_rate": 1.5222525172287944e-05, + "loss": 0.9903, + "step": 22363 + }, + { + "epoch": 1.44, + "learning_rate": 1.5222310483264993e-05, + "loss": 0.9423, + "step": 22364 + }, + { + "epoch": 1.44, + "learning_rate": 1.5222095794242041e-05, + "loss": 1.0339, + "step": 22365 + }, + { + "epoch": 1.44, + "learning_rate": 1.5221881105219092e-05, + "loss": 0.9942, + "step": 22366 + }, + { + "epoch": 1.44, + "learning_rate": 1.5221666416196141e-05, + "loss": 1.046, + "step": 22367 + }, + { + "epoch": 1.44, + "learning_rate": 1.5221451727173192e-05, + "loss": 0.9515, + "step": 22368 + }, + { + "epoch": 1.44, + "learning_rate": 1.522123703815024e-05, + "loss": 0.9943, + "step": 22369 + }, + { + "epoch": 1.44, + "learning_rate": 1.522102234912729e-05, + "loss": 0.8941, + "step": 22370 + }, + { + "epoch": 1.44, + "learning_rate": 1.522080766010434e-05, + "loss": 0.9676, + "step": 22371 + }, + { + "epoch": 1.44, + "learning_rate": 1.522059297108139e-05, + "loss": 0.9087, + "step": 22372 + }, + { + "epoch": 1.44, + "learning_rate": 1.5220378282058438e-05, + "loss": 1.0021, + "step": 22373 + }, + { + "epoch": 1.44, + "learning_rate": 1.5220163593035489e-05, + "loss": 0.9797, + "step": 22374 + }, + { + "epoch": 1.44, + "learning_rate": 1.5219948904012538e-05, + "loss": 0.9375, + "step": 22375 + }, + { + "epoch": 1.44, + "learning_rate": 1.521973421498959e-05, + "loss": 1.0418, + "step": 22376 + }, + { + "epoch": 1.44, + "learning_rate": 1.521951952596664e-05, + "loss": 0.9448, + "step": 22377 + }, + { + "epoch": 1.44, + "learning_rate": 1.521930483694369e-05, + "loss": 0.9911, + "step": 22378 + }, + { + "epoch": 1.44, + "learning_rate": 1.5219090147920739e-05, + "loss": 0.9401, + "step": 22379 + }, + { + "epoch": 1.44, + "learning_rate": 1.5218875458897788e-05, + "loss": 1.022, + "step": 22380 + }, + { + "epoch": 1.44, + "learning_rate": 1.5218660769874838e-05, + "loss": 0.9581, + "step": 22381 + }, + { + "epoch": 1.44, + "learning_rate": 1.5218446080851887e-05, + "loss": 0.98, + "step": 22382 + }, + { + "epoch": 1.44, + "learning_rate": 1.5218231391828936e-05, + "loss": 0.9182, + "step": 22383 + }, + { + "epoch": 1.44, + "learning_rate": 1.5218016702805987e-05, + "loss": 0.9942, + "step": 22384 + }, + { + "epoch": 1.44, + "learning_rate": 1.5217802013783036e-05, + "loss": 0.9289, + "step": 22385 + }, + { + "epoch": 1.44, + "learning_rate": 1.5217587324760085e-05, + "loss": 1.0132, + "step": 22386 + }, + { + "epoch": 1.44, + "learning_rate": 1.5217372635737136e-05, + "loss": 0.9707, + "step": 22387 + }, + { + "epoch": 1.44, + "learning_rate": 1.5217157946714185e-05, + "loss": 1.0209, + "step": 22388 + }, + { + "epoch": 1.44, + "learning_rate": 1.5216943257691235e-05, + "loss": 1.0006, + "step": 22389 + }, + { + "epoch": 1.44, + "learning_rate": 1.5216728568668284e-05, + "loss": 0.9367, + "step": 22390 + }, + { + "epoch": 1.44, + "learning_rate": 1.5216513879645337e-05, + "loss": 0.9764, + "step": 22391 + }, + { + "epoch": 1.44, + "learning_rate": 1.5216299190622386e-05, + "loss": 0.9781, + "step": 22392 + }, + { + "epoch": 1.44, + "learning_rate": 1.5216084501599435e-05, + "loss": 1.0183, + "step": 22393 + }, + { + "epoch": 1.44, + "learning_rate": 1.5215869812576485e-05, + "loss": 0.9419, + "step": 22394 + }, + { + "epoch": 1.44, + "learning_rate": 1.5215655123553534e-05, + "loss": 0.9501, + "step": 22395 + }, + { + "epoch": 1.44, + "learning_rate": 1.5215440434530583e-05, + "loss": 0.9456, + "step": 22396 + }, + { + "epoch": 1.44, + "learning_rate": 1.5215225745507634e-05, + "loss": 0.9433, + "step": 22397 + }, + { + "epoch": 1.44, + "learning_rate": 1.5215011056484683e-05, + "loss": 0.9536, + "step": 22398 + }, + { + "epoch": 1.44, + "learning_rate": 1.5214796367461733e-05, + "loss": 1.0468, + "step": 22399 + }, + { + "epoch": 1.44, + "learning_rate": 1.5214581678438782e-05, + "loss": 1.0755, + "step": 22400 + }, + { + "epoch": 1.44, + "learning_rate": 1.5214581678438782e-05, + "loss": 0.8839, + "step": 22401 + }, + { + "epoch": 1.44, + "learning_rate": 1.5214366989415831e-05, + "loss": 0.9393, + "step": 22402 + }, + { + "epoch": 1.44, + "learning_rate": 1.5214152300392882e-05, + "loss": 0.9592, + "step": 22403 + }, + { + "epoch": 1.44, + "learning_rate": 1.5213937611369931e-05, + "loss": 0.8882, + "step": 22404 + }, + { + "epoch": 1.44, + "learning_rate": 1.521372292234698e-05, + "loss": 0.9882, + "step": 22405 + }, + { + "epoch": 1.44, + "learning_rate": 1.5213508233324032e-05, + "loss": 0.9656, + "step": 22406 + }, + { + "epoch": 1.44, + "learning_rate": 1.5213293544301083e-05, + "loss": 0.9813, + "step": 22407 + }, + { + "epoch": 1.44, + "learning_rate": 1.5213078855278132e-05, + "loss": 0.9395, + "step": 22408 + }, + { + "epoch": 1.44, + "learning_rate": 1.5212864166255181e-05, + "loss": 1.0328, + "step": 22409 + }, + { + "epoch": 1.44, + "learning_rate": 1.5212649477232232e-05, + "loss": 0.8831, + "step": 22410 + }, + { + "epoch": 1.44, + "learning_rate": 1.521243478820928e-05, + "loss": 1.0358, + "step": 22411 + }, + { + "epoch": 1.44, + "learning_rate": 1.521222009918633e-05, + "loss": 0.9873, + "step": 22412 + }, + { + "epoch": 1.44, + "learning_rate": 1.521200541016338e-05, + "loss": 0.9328, + "step": 22413 + }, + { + "epoch": 1.44, + "learning_rate": 1.5211790721140429e-05, + "loss": 0.9445, + "step": 22414 + }, + { + "epoch": 1.44, + "learning_rate": 1.5211576032117478e-05, + "loss": 0.9498, + "step": 22415 + }, + { + "epoch": 1.44, + "learning_rate": 1.5211361343094529e-05, + "loss": 0.9549, + "step": 22416 + }, + { + "epoch": 1.44, + "learning_rate": 1.5211146654071578e-05, + "loss": 0.9415, + "step": 22417 + }, + { + "epoch": 1.44, + "learning_rate": 1.5210931965048628e-05, + "loss": 1.0121, + "step": 22418 + }, + { + "epoch": 1.44, + "learning_rate": 1.5210717276025677e-05, + "loss": 1.0254, + "step": 22419 + }, + { + "epoch": 1.44, + "learning_rate": 1.5210502587002726e-05, + "loss": 0.9807, + "step": 22420 + }, + { + "epoch": 1.44, + "learning_rate": 1.5210287897979779e-05, + "loss": 1.0019, + "step": 22421 + }, + { + "epoch": 1.44, + "learning_rate": 1.5210073208956828e-05, + "loss": 0.9788, + "step": 22422 + }, + { + "epoch": 1.44, + "learning_rate": 1.5209858519933878e-05, + "loss": 0.9755, + "step": 22423 + }, + { + "epoch": 1.44, + "learning_rate": 1.5209643830910927e-05, + "loss": 0.9344, + "step": 22424 + }, + { + "epoch": 1.44, + "learning_rate": 1.5209429141887976e-05, + "loss": 0.9857, + "step": 22425 + }, + { + "epoch": 1.44, + "learning_rate": 1.5209214452865027e-05, + "loss": 0.9553, + "step": 22426 + }, + { + "epoch": 1.44, + "learning_rate": 1.5208999763842076e-05, + "loss": 0.9493, + "step": 22427 + }, + { + "epoch": 1.44, + "learning_rate": 1.5208785074819126e-05, + "loss": 1.042, + "step": 22428 + }, + { + "epoch": 1.44, + "learning_rate": 1.5208570385796175e-05, + "loss": 0.986, + "step": 22429 + }, + { + "epoch": 1.44, + "learning_rate": 1.5208355696773224e-05, + "loss": 0.9337, + "step": 22430 + }, + { + "epoch": 1.44, + "learning_rate": 1.5208141007750275e-05, + "loss": 0.9924, + "step": 22431 + }, + { + "epoch": 1.44, + "learning_rate": 1.5207926318727324e-05, + "loss": 0.9843, + "step": 22432 + }, + { + "epoch": 1.44, + "learning_rate": 1.5207711629704373e-05, + "loss": 1.058, + "step": 22433 + }, + { + "epoch": 1.44, + "learning_rate": 1.5207496940681424e-05, + "loss": 1.0003, + "step": 22434 + }, + { + "epoch": 1.44, + "learning_rate": 1.5207282251658474e-05, + "loss": 0.9971, + "step": 22435 + }, + { + "epoch": 1.44, + "learning_rate": 1.5207067562635525e-05, + "loss": 1.0636, + "step": 22436 + }, + { + "epoch": 1.44, + "learning_rate": 1.5206852873612574e-05, + "loss": 0.9878, + "step": 22437 + }, + { + "epoch": 1.44, + "learning_rate": 1.5206638184589625e-05, + "loss": 0.9855, + "step": 22438 + }, + { + "epoch": 1.44, + "learning_rate": 1.5206423495566674e-05, + "loss": 0.9583, + "step": 22439 + }, + { + "epoch": 1.44, + "learning_rate": 1.5206208806543722e-05, + "loss": 0.9952, + "step": 22440 + }, + { + "epoch": 1.44, + "learning_rate": 1.5205994117520773e-05, + "loss": 0.9329, + "step": 22441 + }, + { + "epoch": 1.44, + "learning_rate": 1.5205779428497822e-05, + "loss": 0.9888, + "step": 22442 + }, + { + "epoch": 1.44, + "learning_rate": 1.5205564739474871e-05, + "loss": 0.985, + "step": 22443 + }, + { + "epoch": 1.44, + "learning_rate": 1.5205350050451922e-05, + "loss": 0.9562, + "step": 22444 + }, + { + "epoch": 1.44, + "learning_rate": 1.520513536142897e-05, + "loss": 0.875, + "step": 22445 + }, + { + "epoch": 1.44, + "learning_rate": 1.5204920672406021e-05, + "loss": 0.9657, + "step": 22446 + }, + { + "epoch": 1.44, + "learning_rate": 1.520470598338307e-05, + "loss": 0.9275, + "step": 22447 + }, + { + "epoch": 1.44, + "learning_rate": 1.520449129436012e-05, + "loss": 1.0359, + "step": 22448 + }, + { + "epoch": 1.44, + "learning_rate": 1.520427660533717e-05, + "loss": 0.9583, + "step": 22449 + }, + { + "epoch": 1.44, + "learning_rate": 1.520406191631422e-05, + "loss": 0.9509, + "step": 22450 + }, + { + "epoch": 1.44, + "learning_rate": 1.5203847227291271e-05, + "loss": 0.9679, + "step": 22451 + }, + { + "epoch": 1.44, + "learning_rate": 1.520363253826832e-05, + "loss": 0.9791, + "step": 22452 + }, + { + "epoch": 1.44, + "learning_rate": 1.520341784924537e-05, + "loss": 1.0108, + "step": 22453 + }, + { + "epoch": 1.44, + "learning_rate": 1.520320316022242e-05, + "loss": 0.9989, + "step": 22454 + }, + { + "epoch": 1.44, + "learning_rate": 1.5202988471199469e-05, + "loss": 0.9134, + "step": 22455 + }, + { + "epoch": 1.44, + "learning_rate": 1.520277378217652e-05, + "loss": 1.0266, + "step": 22456 + }, + { + "epoch": 1.44, + "learning_rate": 1.5202559093153568e-05, + "loss": 1.0478, + "step": 22457 + }, + { + "epoch": 1.44, + "learning_rate": 1.5202344404130617e-05, + "loss": 1.0134, + "step": 22458 + }, + { + "epoch": 1.44, + "learning_rate": 1.5202129715107668e-05, + "loss": 0.9675, + "step": 22459 + }, + { + "epoch": 1.45, + "learning_rate": 1.5201915026084717e-05, + "loss": 0.9235, + "step": 22460 + }, + { + "epoch": 1.45, + "learning_rate": 1.5201700337061766e-05, + "loss": 0.9807, + "step": 22461 + }, + { + "epoch": 1.45, + "learning_rate": 1.5201485648038817e-05, + "loss": 1.0271, + "step": 22462 + }, + { + "epoch": 1.45, + "learning_rate": 1.5201270959015866e-05, + "loss": 0.9431, + "step": 22463 + }, + { + "epoch": 1.45, + "learning_rate": 1.5201056269992918e-05, + "loss": 0.9895, + "step": 22464 + }, + { + "epoch": 1.45, + "learning_rate": 1.5200841580969967e-05, + "loss": 0.8819, + "step": 22465 + }, + { + "epoch": 1.45, + "learning_rate": 1.5200626891947018e-05, + "loss": 0.9035, + "step": 22466 + }, + { + "epoch": 1.45, + "learning_rate": 1.5200412202924067e-05, + "loss": 0.9284, + "step": 22467 + }, + { + "epoch": 1.45, + "learning_rate": 1.5200197513901116e-05, + "loss": 0.972, + "step": 22468 + }, + { + "epoch": 1.45, + "learning_rate": 1.5199982824878166e-05, + "loss": 0.9778, + "step": 22469 + }, + { + "epoch": 1.45, + "learning_rate": 1.5199768135855215e-05, + "loss": 1.0226, + "step": 22470 + }, + { + "epoch": 1.45, + "learning_rate": 1.5199553446832264e-05, + "loss": 0.9605, + "step": 22471 + }, + { + "epoch": 1.45, + "learning_rate": 1.5199338757809315e-05, + "loss": 0.9272, + "step": 22472 + }, + { + "epoch": 1.45, + "learning_rate": 1.5199124068786364e-05, + "loss": 0.9698, + "step": 22473 + }, + { + "epoch": 1.45, + "learning_rate": 1.5198909379763413e-05, + "loss": 0.9404, + "step": 22474 + }, + { + "epoch": 1.45, + "learning_rate": 1.5198694690740463e-05, + "loss": 0.9838, + "step": 22475 + }, + { + "epoch": 1.45, + "learning_rate": 1.5198480001717512e-05, + "loss": 0.9377, + "step": 22476 + }, + { + "epoch": 1.45, + "learning_rate": 1.5198265312694563e-05, + "loss": 0.9789, + "step": 22477 + }, + { + "epoch": 1.45, + "learning_rate": 1.5198050623671612e-05, + "loss": 0.9948, + "step": 22478 + }, + { + "epoch": 1.45, + "learning_rate": 1.5197835934648664e-05, + "loss": 1.0058, + "step": 22479 + }, + { + "epoch": 1.45, + "learning_rate": 1.5197621245625713e-05, + "loss": 0.9563, + "step": 22480 + }, + { + "epoch": 1.45, + "learning_rate": 1.5197406556602762e-05, + "loss": 0.9653, + "step": 22481 + }, + { + "epoch": 1.45, + "learning_rate": 1.5197191867579813e-05, + "loss": 0.9808, + "step": 22482 + }, + { + "epoch": 1.45, + "learning_rate": 1.5196977178556862e-05, + "loss": 0.9421, + "step": 22483 + }, + { + "epoch": 1.45, + "learning_rate": 1.519676248953391e-05, + "loss": 0.9541, + "step": 22484 + }, + { + "epoch": 1.45, + "learning_rate": 1.5196547800510961e-05, + "loss": 0.9437, + "step": 22485 + }, + { + "epoch": 1.45, + "learning_rate": 1.519633311148801e-05, + "loss": 0.9956, + "step": 22486 + }, + { + "epoch": 1.45, + "learning_rate": 1.5196118422465061e-05, + "loss": 0.9466, + "step": 22487 + }, + { + "epoch": 1.45, + "learning_rate": 1.519590373344211e-05, + "loss": 0.9961, + "step": 22488 + }, + { + "epoch": 1.45, + "learning_rate": 1.5195689044419159e-05, + "loss": 1.0419, + "step": 22489 + }, + { + "epoch": 1.45, + "learning_rate": 1.519547435539621e-05, + "loss": 0.9927, + "step": 22490 + }, + { + "epoch": 1.45, + "learning_rate": 1.5195259666373259e-05, + "loss": 1.0133, + "step": 22491 + }, + { + "epoch": 1.45, + "learning_rate": 1.5195044977350308e-05, + "loss": 0.9249, + "step": 22492 + }, + { + "epoch": 1.45, + "learning_rate": 1.5194830288327358e-05, + "loss": 0.952, + "step": 22493 + }, + { + "epoch": 1.45, + "learning_rate": 1.5194615599304409e-05, + "loss": 0.9699, + "step": 22494 + }, + { + "epoch": 1.45, + "learning_rate": 1.519440091028146e-05, + "loss": 0.8955, + "step": 22495 + }, + { + "epoch": 1.45, + "learning_rate": 1.5194186221258509e-05, + "loss": 0.9074, + "step": 22496 + }, + { + "epoch": 1.45, + "learning_rate": 1.519397153223556e-05, + "loss": 0.9806, + "step": 22497 + }, + { + "epoch": 1.45, + "learning_rate": 1.5193756843212608e-05, + "loss": 0.9683, + "step": 22498 + }, + { + "epoch": 1.45, + "learning_rate": 1.5193542154189657e-05, + "loss": 1.0272, + "step": 22499 + }, + { + "epoch": 1.45, + "learning_rate": 1.5193327465166708e-05, + "loss": 0.8946, + "step": 22500 + }, + { + "epoch": 1.45, + "learning_rate": 1.5193112776143757e-05, + "loss": 0.9481, + "step": 22501 + }, + { + "epoch": 1.45, + "learning_rate": 1.5192898087120806e-05, + "loss": 0.9771, + "step": 22502 + }, + { + "epoch": 1.45, + "learning_rate": 1.5192683398097856e-05, + "loss": 0.9548, + "step": 22503 + }, + { + "epoch": 1.45, + "learning_rate": 1.5192468709074905e-05, + "loss": 0.9721, + "step": 22504 + }, + { + "epoch": 1.45, + "learning_rate": 1.5192254020051956e-05, + "loss": 0.9828, + "step": 22505 + }, + { + "epoch": 1.45, + "learning_rate": 1.5192039331029005e-05, + "loss": 0.9332, + "step": 22506 + }, + { + "epoch": 1.45, + "learning_rate": 1.5191824642006054e-05, + "loss": 0.9833, + "step": 22507 + }, + { + "epoch": 1.45, + "learning_rate": 1.5191609952983106e-05, + "loss": 0.9634, + "step": 22508 + }, + { + "epoch": 1.45, + "learning_rate": 1.5191395263960155e-05, + "loss": 1.0543, + "step": 22509 + }, + { + "epoch": 1.45, + "learning_rate": 1.5191180574937206e-05, + "loss": 0.9026, + "step": 22510 + }, + { + "epoch": 1.45, + "learning_rate": 1.5190965885914255e-05, + "loss": 1.0301, + "step": 22511 + }, + { + "epoch": 1.45, + "learning_rate": 1.5190751196891304e-05, + "loss": 1.0289, + "step": 22512 + }, + { + "epoch": 1.45, + "learning_rate": 1.5190536507868355e-05, + "loss": 0.9952, + "step": 22513 + }, + { + "epoch": 1.45, + "learning_rate": 1.5190321818845403e-05, + "loss": 0.8981, + "step": 22514 + }, + { + "epoch": 1.45, + "learning_rate": 1.5190107129822454e-05, + "loss": 1.0101, + "step": 22515 + }, + { + "epoch": 1.45, + "learning_rate": 1.5189892440799503e-05, + "loss": 0.928, + "step": 22516 + }, + { + "epoch": 1.45, + "learning_rate": 1.5189677751776552e-05, + "loss": 1.0279, + "step": 22517 + }, + { + "epoch": 1.45, + "learning_rate": 1.5189463062753603e-05, + "loss": 0.9669, + "step": 22518 + }, + { + "epoch": 1.45, + "learning_rate": 1.5189248373730652e-05, + "loss": 0.9745, + "step": 22519 + }, + { + "epoch": 1.45, + "learning_rate": 1.51890336847077e-05, + "loss": 1.0054, + "step": 22520 + }, + { + "epoch": 1.45, + "learning_rate": 1.5188818995684751e-05, + "loss": 0.9623, + "step": 22521 + }, + { + "epoch": 1.45, + "learning_rate": 1.51886043066618e-05, + "loss": 1.0037, + "step": 22522 + }, + { + "epoch": 1.45, + "learning_rate": 1.5188389617638853e-05, + "loss": 0.9419, + "step": 22523 + }, + { + "epoch": 1.45, + "learning_rate": 1.5188174928615902e-05, + "loss": 0.9208, + "step": 22524 + }, + { + "epoch": 1.45, + "learning_rate": 1.5187960239592952e-05, + "loss": 0.9762, + "step": 22525 + }, + { + "epoch": 1.45, + "learning_rate": 1.5187745550570001e-05, + "loss": 1.0098, + "step": 22526 + }, + { + "epoch": 1.45, + "learning_rate": 1.518753086154705e-05, + "loss": 0.9561, + "step": 22527 + }, + { + "epoch": 1.45, + "learning_rate": 1.5187316172524101e-05, + "loss": 0.9358, + "step": 22528 + }, + { + "epoch": 1.45, + "learning_rate": 1.518710148350115e-05, + "loss": 0.9619, + "step": 22529 + }, + { + "epoch": 1.45, + "learning_rate": 1.5186886794478199e-05, + "loss": 0.9412, + "step": 22530 + }, + { + "epoch": 1.45, + "learning_rate": 1.518667210545525e-05, + "loss": 0.922, + "step": 22531 + }, + { + "epoch": 1.45, + "learning_rate": 1.5186457416432298e-05, + "loss": 1.0045, + "step": 22532 + }, + { + "epoch": 1.45, + "learning_rate": 1.5186242727409349e-05, + "loss": 1.0449, + "step": 22533 + }, + { + "epoch": 1.45, + "learning_rate": 1.5186028038386398e-05, + "loss": 1.0598, + "step": 22534 + }, + { + "epoch": 1.45, + "learning_rate": 1.5185813349363447e-05, + "loss": 0.9934, + "step": 22535 + }, + { + "epoch": 1.45, + "learning_rate": 1.5185598660340498e-05, + "loss": 0.9408, + "step": 22536 + }, + { + "epoch": 1.45, + "learning_rate": 1.5185383971317548e-05, + "loss": 0.9673, + "step": 22537 + }, + { + "epoch": 1.45, + "learning_rate": 1.5185169282294599e-05, + "loss": 1.0065, + "step": 22538 + }, + { + "epoch": 1.45, + "learning_rate": 1.5184954593271648e-05, + "loss": 1.033, + "step": 22539 + }, + { + "epoch": 1.45, + "learning_rate": 1.5184739904248697e-05, + "loss": 0.9087, + "step": 22540 + }, + { + "epoch": 1.45, + "learning_rate": 1.5184525215225748e-05, + "loss": 1.0436, + "step": 22541 + }, + { + "epoch": 1.45, + "learning_rate": 1.5184310526202797e-05, + "loss": 1.0019, + "step": 22542 + }, + { + "epoch": 1.45, + "learning_rate": 1.5184095837179847e-05, + "loss": 1.0501, + "step": 22543 + }, + { + "epoch": 1.45, + "learning_rate": 1.5183881148156896e-05, + "loss": 0.9332, + "step": 22544 + }, + { + "epoch": 1.45, + "learning_rate": 1.5183666459133945e-05, + "loss": 0.9207, + "step": 22545 + }, + { + "epoch": 1.45, + "learning_rate": 1.5183451770110996e-05, + "loss": 0.9682, + "step": 22546 + }, + { + "epoch": 1.45, + "learning_rate": 1.5183237081088045e-05, + "loss": 0.9809, + "step": 22547 + }, + { + "epoch": 1.45, + "learning_rate": 1.5183022392065094e-05, + "loss": 0.9839, + "step": 22548 + }, + { + "epoch": 1.45, + "learning_rate": 1.5182807703042144e-05, + "loss": 0.955, + "step": 22549 + }, + { + "epoch": 1.45, + "learning_rate": 1.5182593014019193e-05, + "loss": 0.9011, + "step": 22550 + }, + { + "epoch": 1.45, + "learning_rate": 1.5182378324996242e-05, + "loss": 0.9554, + "step": 22551 + }, + { + "epoch": 1.45, + "learning_rate": 1.5182163635973295e-05, + "loss": 1.0262, + "step": 22552 + }, + { + "epoch": 1.45, + "learning_rate": 1.5181948946950345e-05, + "loss": 0.9305, + "step": 22553 + }, + { + "epoch": 1.45, + "learning_rate": 1.5181734257927394e-05, + "loss": 0.9449, + "step": 22554 + }, + { + "epoch": 1.45, + "learning_rate": 1.5181519568904443e-05, + "loss": 0.9389, + "step": 22555 + }, + { + "epoch": 1.45, + "learning_rate": 1.5181304879881494e-05, + "loss": 1.1144, + "step": 22556 + }, + { + "epoch": 1.45, + "learning_rate": 1.5181090190858543e-05, + "loss": 1.0196, + "step": 22557 + }, + { + "epoch": 1.45, + "learning_rate": 1.5180875501835592e-05, + "loss": 0.9118, + "step": 22558 + }, + { + "epoch": 1.45, + "learning_rate": 1.5180660812812642e-05, + "loss": 1.0339, + "step": 22559 + }, + { + "epoch": 1.45, + "learning_rate": 1.5180446123789691e-05, + "loss": 0.9752, + "step": 22560 + }, + { + "epoch": 1.45, + "learning_rate": 1.518023143476674e-05, + "loss": 0.9822, + "step": 22561 + }, + { + "epoch": 1.45, + "learning_rate": 1.5180016745743791e-05, + "loss": 1.0406, + "step": 22562 + }, + { + "epoch": 1.45, + "learning_rate": 1.517980205672084e-05, + "loss": 1.0247, + "step": 22563 + }, + { + "epoch": 1.45, + "learning_rate": 1.517958736769789e-05, + "loss": 0.9991, + "step": 22564 + }, + { + "epoch": 1.45, + "learning_rate": 1.517937267867494e-05, + "loss": 0.9898, + "step": 22565 + }, + { + "epoch": 1.45, + "learning_rate": 1.5179157989651992e-05, + "loss": 0.9688, + "step": 22566 + }, + { + "epoch": 1.45, + "learning_rate": 1.5178943300629041e-05, + "loss": 0.9259, + "step": 22567 + }, + { + "epoch": 1.45, + "learning_rate": 1.517872861160609e-05, + "loss": 1.0105, + "step": 22568 + }, + { + "epoch": 1.45, + "learning_rate": 1.517851392258314e-05, + "loss": 0.8933, + "step": 22569 + }, + { + "epoch": 1.45, + "learning_rate": 1.517829923356019e-05, + "loss": 1.0152, + "step": 22570 + }, + { + "epoch": 1.45, + "learning_rate": 1.5178084544537239e-05, + "loss": 0.966, + "step": 22571 + }, + { + "epoch": 1.45, + "learning_rate": 1.517786985551429e-05, + "loss": 0.9376, + "step": 22572 + }, + { + "epoch": 1.45, + "learning_rate": 1.5177655166491338e-05, + "loss": 1.0401, + "step": 22573 + }, + { + "epoch": 1.45, + "learning_rate": 1.5177440477468389e-05, + "loss": 0.9605, + "step": 22574 + }, + { + "epoch": 1.45, + "learning_rate": 1.5177225788445438e-05, + "loss": 0.9766, + "step": 22575 + }, + { + "epoch": 1.45, + "learning_rate": 1.5177011099422487e-05, + "loss": 0.9179, + "step": 22576 + }, + { + "epoch": 1.45, + "learning_rate": 1.5176796410399537e-05, + "loss": 1.0474, + "step": 22577 + }, + { + "epoch": 1.45, + "learning_rate": 1.5176581721376586e-05, + "loss": 0.9727, + "step": 22578 + }, + { + "epoch": 1.45, + "learning_rate": 1.5176367032353635e-05, + "loss": 0.9594, + "step": 22579 + }, + { + "epoch": 1.45, + "learning_rate": 1.5176152343330686e-05, + "loss": 1.0157, + "step": 22580 + }, + { + "epoch": 1.45, + "learning_rate": 1.5175937654307737e-05, + "loss": 0.9796, + "step": 22581 + }, + { + "epoch": 1.45, + "learning_rate": 1.5175722965284787e-05, + "loss": 1.0137, + "step": 22582 + }, + { + "epoch": 1.45, + "learning_rate": 1.5175508276261836e-05, + "loss": 1.0023, + "step": 22583 + }, + { + "epoch": 1.45, + "learning_rate": 1.5175293587238887e-05, + "loss": 0.892, + "step": 22584 + }, + { + "epoch": 1.45, + "learning_rate": 1.5175078898215936e-05, + "loss": 1.0184, + "step": 22585 + }, + { + "epoch": 1.45, + "learning_rate": 1.5174864209192985e-05, + "loss": 0.9575, + "step": 22586 + }, + { + "epoch": 1.45, + "learning_rate": 1.5174649520170036e-05, + "loss": 0.8961, + "step": 22587 + }, + { + "epoch": 1.45, + "learning_rate": 1.5174434831147084e-05, + "loss": 0.9949, + "step": 22588 + }, + { + "epoch": 1.45, + "learning_rate": 1.5174220142124133e-05, + "loss": 1.0713, + "step": 22589 + }, + { + "epoch": 1.45, + "learning_rate": 1.5174005453101184e-05, + "loss": 0.9716, + "step": 22590 + }, + { + "epoch": 1.45, + "learning_rate": 1.5173790764078233e-05, + "loss": 1.0269, + "step": 22591 + }, + { + "epoch": 1.45, + "learning_rate": 1.5173576075055284e-05, + "loss": 0.952, + "step": 22592 + }, + { + "epoch": 1.45, + "learning_rate": 1.5173361386032333e-05, + "loss": 0.9489, + "step": 22593 + }, + { + "epoch": 1.45, + "learning_rate": 1.5173146697009382e-05, + "loss": 0.967, + "step": 22594 + }, + { + "epoch": 1.45, + "learning_rate": 1.5172932007986434e-05, + "loss": 1.0633, + "step": 22595 + }, + { + "epoch": 1.45, + "learning_rate": 1.5172717318963483e-05, + "loss": 0.9403, + "step": 22596 + }, + { + "epoch": 1.45, + "learning_rate": 1.5172502629940534e-05, + "loss": 1.0455, + "step": 22597 + }, + { + "epoch": 1.45, + "learning_rate": 1.5172287940917583e-05, + "loss": 0.9733, + "step": 22598 + }, + { + "epoch": 1.45, + "learning_rate": 1.5172073251894632e-05, + "loss": 0.9497, + "step": 22599 + }, + { + "epoch": 1.45, + "learning_rate": 1.5171858562871682e-05, + "loss": 0.9381, + "step": 22600 + }, + { + "epoch": 1.45, + "learning_rate": 1.5171643873848731e-05, + "loss": 0.9797, + "step": 22601 + }, + { + "epoch": 1.45, + "learning_rate": 1.5171429184825782e-05, + "loss": 1.0056, + "step": 22602 + }, + { + "epoch": 1.45, + "learning_rate": 1.517121449580283e-05, + "loss": 0.9688, + "step": 22603 + }, + { + "epoch": 1.45, + "learning_rate": 1.517099980677988e-05, + "loss": 1.0137, + "step": 22604 + }, + { + "epoch": 1.45, + "learning_rate": 1.517078511775693e-05, + "loss": 0.9938, + "step": 22605 + }, + { + "epoch": 1.45, + "learning_rate": 1.517057042873398e-05, + "loss": 0.9892, + "step": 22606 + }, + { + "epoch": 1.45, + "learning_rate": 1.5170355739711028e-05, + "loss": 1.0092, + "step": 22607 + }, + { + "epoch": 1.45, + "learning_rate": 1.5170141050688079e-05, + "loss": 0.9427, + "step": 22608 + }, + { + "epoch": 1.45, + "learning_rate": 1.5169926361665128e-05, + "loss": 1.019, + "step": 22609 + }, + { + "epoch": 1.45, + "learning_rate": 1.516971167264218e-05, + "loss": 1.0184, + "step": 22610 + }, + { + "epoch": 1.45, + "learning_rate": 1.516949698361923e-05, + "loss": 0.9729, + "step": 22611 + }, + { + "epoch": 1.45, + "learning_rate": 1.516928229459628e-05, + "loss": 0.9547, + "step": 22612 + }, + { + "epoch": 1.45, + "learning_rate": 1.5169067605573329e-05, + "loss": 0.996, + "step": 22613 + }, + { + "epoch": 1.45, + "learning_rate": 1.5168852916550378e-05, + "loss": 0.9081, + "step": 22614 + }, + { + "epoch": 1.45, + "learning_rate": 1.5168638227527429e-05, + "loss": 0.9895, + "step": 22615 + }, + { + "epoch": 1.46, + "learning_rate": 1.5168423538504478e-05, + "loss": 0.9892, + "step": 22616 + }, + { + "epoch": 1.46, + "learning_rate": 1.5168208849481526e-05, + "loss": 0.9432, + "step": 22617 + }, + { + "epoch": 1.46, + "learning_rate": 1.5167994160458577e-05, + "loss": 1.0215, + "step": 22618 + }, + { + "epoch": 1.46, + "learning_rate": 1.5167779471435626e-05, + "loss": 0.9682, + "step": 22619 + }, + { + "epoch": 1.46, + "learning_rate": 1.5167564782412677e-05, + "loss": 1.0353, + "step": 22620 + }, + { + "epoch": 1.46, + "learning_rate": 1.5167350093389726e-05, + "loss": 0.9629, + "step": 22621 + }, + { + "epoch": 1.46, + "learning_rate": 1.5167135404366775e-05, + "loss": 0.9608, + "step": 22622 + }, + { + "epoch": 1.46, + "learning_rate": 1.5166920715343825e-05, + "loss": 0.9321, + "step": 22623 + }, + { + "epoch": 1.46, + "learning_rate": 1.5166706026320876e-05, + "loss": 1.0748, + "step": 22624 + }, + { + "epoch": 1.46, + "learning_rate": 1.5166491337297927e-05, + "loss": 0.9893, + "step": 22625 + }, + { + "epoch": 1.46, + "learning_rate": 1.5166276648274976e-05, + "loss": 0.9435, + "step": 22626 + }, + { + "epoch": 1.46, + "learning_rate": 1.5166061959252025e-05, + "loss": 0.9946, + "step": 22627 + }, + { + "epoch": 1.46, + "learning_rate": 1.5165847270229075e-05, + "loss": 1.021, + "step": 22628 + }, + { + "epoch": 1.46, + "learning_rate": 1.5165632581206124e-05, + "loss": 0.9427, + "step": 22629 + }, + { + "epoch": 1.46, + "learning_rate": 1.5165417892183175e-05, + "loss": 0.9744, + "step": 22630 + }, + { + "epoch": 1.46, + "learning_rate": 1.5165203203160224e-05, + "loss": 0.9905, + "step": 22631 + }, + { + "epoch": 1.46, + "learning_rate": 1.5164988514137273e-05, + "loss": 0.9757, + "step": 22632 + }, + { + "epoch": 1.46, + "learning_rate": 1.5164773825114323e-05, + "loss": 0.9868, + "step": 22633 + }, + { + "epoch": 1.46, + "learning_rate": 1.5164559136091372e-05, + "loss": 0.9886, + "step": 22634 + }, + { + "epoch": 1.46, + "learning_rate": 1.5164344447068421e-05, + "loss": 1.0397, + "step": 22635 + }, + { + "epoch": 1.46, + "learning_rate": 1.5164129758045472e-05, + "loss": 0.9321, + "step": 22636 + }, + { + "epoch": 1.46, + "learning_rate": 1.5163915069022521e-05, + "loss": 1.025, + "step": 22637 + }, + { + "epoch": 1.46, + "learning_rate": 1.516370037999957e-05, + "loss": 1.044, + "step": 22638 + }, + { + "epoch": 1.46, + "learning_rate": 1.5163485690976622e-05, + "loss": 0.8745, + "step": 22639 + }, + { + "epoch": 1.46, + "learning_rate": 1.5163271001953673e-05, + "loss": 0.9546, + "step": 22640 + }, + { + "epoch": 1.46, + "learning_rate": 1.5163056312930722e-05, + "loss": 0.9664, + "step": 22641 + }, + { + "epoch": 1.46, + "learning_rate": 1.5162841623907771e-05, + "loss": 0.9408, + "step": 22642 + }, + { + "epoch": 1.46, + "learning_rate": 1.5162626934884822e-05, + "loss": 1.0222, + "step": 22643 + }, + { + "epoch": 1.46, + "learning_rate": 1.516241224586187e-05, + "loss": 0.9881, + "step": 22644 + }, + { + "epoch": 1.46, + "learning_rate": 1.516219755683892e-05, + "loss": 0.9094, + "step": 22645 + }, + { + "epoch": 1.46, + "learning_rate": 1.516198286781597e-05, + "loss": 0.9305, + "step": 22646 + }, + { + "epoch": 1.46, + "learning_rate": 1.516176817879302e-05, + "loss": 0.9635, + "step": 22647 + }, + { + "epoch": 1.46, + "learning_rate": 1.5161553489770068e-05, + "loss": 0.9943, + "step": 22648 + }, + { + "epoch": 1.46, + "learning_rate": 1.5161338800747119e-05, + "loss": 0.9839, + "step": 22649 + }, + { + "epoch": 1.46, + "learning_rate": 1.5161124111724168e-05, + "loss": 0.9396, + "step": 22650 + }, + { + "epoch": 1.46, + "learning_rate": 1.5160909422701218e-05, + "loss": 0.972, + "step": 22651 + }, + { + "epoch": 1.46, + "learning_rate": 1.5160694733678267e-05, + "loss": 0.9466, + "step": 22652 + }, + { + "epoch": 1.46, + "learning_rate": 1.516048004465532e-05, + "loss": 0.97, + "step": 22653 + }, + { + "epoch": 1.46, + "learning_rate": 1.5160265355632369e-05, + "loss": 0.9771, + "step": 22654 + }, + { + "epoch": 1.46, + "learning_rate": 1.5160050666609418e-05, + "loss": 0.9398, + "step": 22655 + }, + { + "epoch": 1.46, + "learning_rate": 1.5159835977586468e-05, + "loss": 1.0186, + "step": 22656 + }, + { + "epoch": 1.46, + "learning_rate": 1.5159621288563517e-05, + "loss": 1.0428, + "step": 22657 + }, + { + "epoch": 1.46, + "learning_rate": 1.5159406599540566e-05, + "loss": 0.9442, + "step": 22658 + }, + { + "epoch": 1.46, + "learning_rate": 1.5159191910517617e-05, + "loss": 0.9845, + "step": 22659 + }, + { + "epoch": 1.46, + "learning_rate": 1.5158977221494666e-05, + "loss": 0.9333, + "step": 22660 + }, + { + "epoch": 1.46, + "learning_rate": 1.5158762532471717e-05, + "loss": 0.954, + "step": 22661 + }, + { + "epoch": 1.46, + "learning_rate": 1.5158547843448765e-05, + "loss": 0.9903, + "step": 22662 + }, + { + "epoch": 1.46, + "learning_rate": 1.5158333154425814e-05, + "loss": 0.9245, + "step": 22663 + }, + { + "epoch": 1.46, + "learning_rate": 1.5158118465402865e-05, + "loss": 1.0715, + "step": 22664 + }, + { + "epoch": 1.46, + "learning_rate": 1.5157903776379914e-05, + "loss": 0.9805, + "step": 22665 + }, + { + "epoch": 1.46, + "learning_rate": 1.5157689087356963e-05, + "loss": 0.9895, + "step": 22666 + }, + { + "epoch": 1.46, + "learning_rate": 1.5157474398334014e-05, + "loss": 1.0476, + "step": 22667 + }, + { + "epoch": 1.46, + "learning_rate": 1.5157259709311064e-05, + "loss": 0.9833, + "step": 22668 + }, + { + "epoch": 1.46, + "learning_rate": 1.5157045020288115e-05, + "loss": 0.9049, + "step": 22669 + }, + { + "epoch": 1.46, + "learning_rate": 1.5156830331265164e-05, + "loss": 0.9676, + "step": 22670 + }, + { + "epoch": 1.46, + "learning_rate": 1.5156615642242215e-05, + "loss": 0.9452, + "step": 22671 + }, + { + "epoch": 1.46, + "learning_rate": 1.5156400953219264e-05, + "loss": 0.9376, + "step": 22672 + }, + { + "epoch": 1.46, + "learning_rate": 1.5156186264196313e-05, + "loss": 1.004, + "step": 22673 + }, + { + "epoch": 1.46, + "learning_rate": 1.5155971575173363e-05, + "loss": 0.9771, + "step": 22674 + }, + { + "epoch": 1.46, + "learning_rate": 1.5155756886150412e-05, + "loss": 0.8822, + "step": 22675 + }, + { + "epoch": 1.46, + "learning_rate": 1.5155542197127461e-05, + "loss": 1.058, + "step": 22676 + }, + { + "epoch": 1.46, + "learning_rate": 1.5155327508104512e-05, + "loss": 0.9889, + "step": 22677 + }, + { + "epoch": 1.46, + "learning_rate": 1.515511281908156e-05, + "loss": 1.0727, + "step": 22678 + }, + { + "epoch": 1.46, + "learning_rate": 1.5154898130058611e-05, + "loss": 0.9129, + "step": 22679 + }, + { + "epoch": 1.46, + "learning_rate": 1.515468344103566e-05, + "loss": 0.9134, + "step": 22680 + }, + { + "epoch": 1.46, + "learning_rate": 1.515446875201271e-05, + "loss": 0.9392, + "step": 22681 + }, + { + "epoch": 1.46, + "learning_rate": 1.5154254062989762e-05, + "loss": 1.0336, + "step": 22682 + }, + { + "epoch": 1.46, + "learning_rate": 1.515403937396681e-05, + "loss": 0.9432, + "step": 22683 + }, + { + "epoch": 1.46, + "learning_rate": 1.5153824684943861e-05, + "loss": 0.9554, + "step": 22684 + }, + { + "epoch": 1.46, + "learning_rate": 1.515360999592091e-05, + "loss": 0.9823, + "step": 22685 + }, + { + "epoch": 1.46, + "learning_rate": 1.515339530689796e-05, + "loss": 1.0206, + "step": 22686 + }, + { + "epoch": 1.46, + "learning_rate": 1.515318061787501e-05, + "loss": 0.9986, + "step": 22687 + }, + { + "epoch": 1.46, + "learning_rate": 1.5152965928852059e-05, + "loss": 0.9261, + "step": 22688 + }, + { + "epoch": 1.46, + "learning_rate": 1.515275123982911e-05, + "loss": 0.9506, + "step": 22689 + }, + { + "epoch": 1.46, + "learning_rate": 1.5152536550806159e-05, + "loss": 0.9791, + "step": 22690 + }, + { + "epoch": 1.46, + "learning_rate": 1.5152321861783207e-05, + "loss": 0.9434, + "step": 22691 + }, + { + "epoch": 1.46, + "learning_rate": 1.5152107172760258e-05, + "loss": 0.9227, + "step": 22692 + }, + { + "epoch": 1.46, + "learning_rate": 1.5151892483737307e-05, + "loss": 1.0328, + "step": 22693 + }, + { + "epoch": 1.46, + "learning_rate": 1.5151677794714356e-05, + "loss": 0.9801, + "step": 22694 + }, + { + "epoch": 1.46, + "learning_rate": 1.5151463105691407e-05, + "loss": 0.9251, + "step": 22695 + }, + { + "epoch": 1.46, + "learning_rate": 1.5151248416668456e-05, + "loss": 0.9796, + "step": 22696 + }, + { + "epoch": 1.46, + "learning_rate": 1.5151033727645508e-05, + "loss": 0.9371, + "step": 22697 + }, + { + "epoch": 1.46, + "learning_rate": 1.5150819038622557e-05, + "loss": 0.9133, + "step": 22698 + }, + { + "epoch": 1.46, + "learning_rate": 1.5150604349599608e-05, + "loss": 0.9935, + "step": 22699 + }, + { + "epoch": 1.46, + "learning_rate": 1.5150389660576657e-05, + "loss": 1.0239, + "step": 22700 + }, + { + "epoch": 1.46, + "learning_rate": 1.5150174971553706e-05, + "loss": 0.9376, + "step": 22701 + }, + { + "epoch": 1.46, + "learning_rate": 1.5149960282530756e-05, + "loss": 1.0105, + "step": 22702 + }, + { + "epoch": 1.46, + "learning_rate": 1.5149745593507805e-05, + "loss": 1.0035, + "step": 22703 + }, + { + "epoch": 1.46, + "learning_rate": 1.5149530904484854e-05, + "loss": 0.919, + "step": 22704 + }, + { + "epoch": 1.46, + "learning_rate": 1.5149316215461905e-05, + "loss": 1.0009, + "step": 22705 + }, + { + "epoch": 1.46, + "learning_rate": 1.5149101526438954e-05, + "loss": 0.9946, + "step": 22706 + }, + { + "epoch": 1.46, + "learning_rate": 1.5148886837416003e-05, + "loss": 0.9604, + "step": 22707 + }, + { + "epoch": 1.46, + "learning_rate": 1.5148672148393053e-05, + "loss": 1.0187, + "step": 22708 + }, + { + "epoch": 1.46, + "learning_rate": 1.5148457459370102e-05, + "loss": 1.0348, + "step": 22709 + }, + { + "epoch": 1.46, + "learning_rate": 1.5148242770347153e-05, + "loss": 0.9378, + "step": 22710 + }, + { + "epoch": 1.46, + "learning_rate": 1.5148028081324204e-05, + "loss": 0.9508, + "step": 22711 + }, + { + "epoch": 1.46, + "learning_rate": 1.5147813392301254e-05, + "loss": 0.9598, + "step": 22712 + }, + { + "epoch": 1.46, + "learning_rate": 1.5147598703278303e-05, + "loss": 0.9522, + "step": 22713 + }, + { + "epoch": 1.46, + "learning_rate": 1.5147384014255352e-05, + "loss": 0.9507, + "step": 22714 + }, + { + "epoch": 1.46, + "learning_rate": 1.5147169325232403e-05, + "loss": 1.0038, + "step": 22715 + }, + { + "epoch": 1.46, + "learning_rate": 1.5146954636209452e-05, + "loss": 1.0418, + "step": 22716 + }, + { + "epoch": 1.46, + "learning_rate": 1.5146739947186503e-05, + "loss": 0.9479, + "step": 22717 + }, + { + "epoch": 1.46, + "learning_rate": 1.5146525258163552e-05, + "loss": 0.9557, + "step": 22718 + }, + { + "epoch": 1.46, + "learning_rate": 1.51463105691406e-05, + "loss": 0.9548, + "step": 22719 + }, + { + "epoch": 1.46, + "learning_rate": 1.5146095880117651e-05, + "loss": 0.9069, + "step": 22720 + }, + { + "epoch": 1.46, + "learning_rate": 1.51458811910947e-05, + "loss": 0.9633, + "step": 22721 + }, + { + "epoch": 1.46, + "learning_rate": 1.5145666502071749e-05, + "loss": 0.9294, + "step": 22722 + }, + { + "epoch": 1.46, + "learning_rate": 1.51454518130488e-05, + "loss": 0.9927, + "step": 22723 + }, + { + "epoch": 1.46, + "learning_rate": 1.5145237124025849e-05, + "loss": 0.9764, + "step": 22724 + }, + { + "epoch": 1.46, + "learning_rate": 1.5145022435002898e-05, + "loss": 0.9906, + "step": 22725 + }, + { + "epoch": 1.46, + "learning_rate": 1.514480774597995e-05, + "loss": 0.9502, + "step": 22726 + }, + { + "epoch": 1.46, + "learning_rate": 1.5144593056957e-05, + "loss": 1.0261, + "step": 22727 + }, + { + "epoch": 1.46, + "learning_rate": 1.514437836793405e-05, + "loss": 1.0041, + "step": 22728 + }, + { + "epoch": 1.46, + "learning_rate": 1.5144163678911099e-05, + "loss": 0.9983, + "step": 22729 + }, + { + "epoch": 1.46, + "learning_rate": 1.514394898988815e-05, + "loss": 1.0011, + "step": 22730 + }, + { + "epoch": 1.46, + "learning_rate": 1.5143734300865198e-05, + "loss": 0.9321, + "step": 22731 + }, + { + "epoch": 1.46, + "learning_rate": 1.5143519611842247e-05, + "loss": 0.9421, + "step": 22732 + }, + { + "epoch": 1.46, + "learning_rate": 1.5143304922819298e-05, + "loss": 0.9507, + "step": 22733 + }, + { + "epoch": 1.46, + "learning_rate": 1.5143090233796347e-05, + "loss": 1.015, + "step": 22734 + }, + { + "epoch": 1.46, + "learning_rate": 1.5142875544773396e-05, + "loss": 1.0553, + "step": 22735 + }, + { + "epoch": 1.46, + "learning_rate": 1.5142660855750446e-05, + "loss": 0.9214, + "step": 22736 + }, + { + "epoch": 1.46, + "learning_rate": 1.5142446166727495e-05, + "loss": 1.0496, + "step": 22737 + }, + { + "epoch": 1.46, + "learning_rate": 1.5142231477704546e-05, + "loss": 1.0143, + "step": 22738 + }, + { + "epoch": 1.46, + "learning_rate": 1.5142016788681595e-05, + "loss": 0.931, + "step": 22739 + }, + { + "epoch": 1.46, + "learning_rate": 1.5141802099658647e-05, + "loss": 0.9309, + "step": 22740 + }, + { + "epoch": 1.46, + "learning_rate": 1.5141587410635696e-05, + "loss": 0.9309, + "step": 22741 + }, + { + "epoch": 1.46, + "learning_rate": 1.5141372721612745e-05, + "loss": 1.0472, + "step": 22742 + }, + { + "epoch": 1.46, + "learning_rate": 1.5141158032589796e-05, + "loss": 1.0252, + "step": 22743 + }, + { + "epoch": 1.46, + "learning_rate": 1.5140943343566845e-05, + "loss": 1.0481, + "step": 22744 + }, + { + "epoch": 1.46, + "learning_rate": 1.5140728654543894e-05, + "loss": 1.0158, + "step": 22745 + }, + { + "epoch": 1.46, + "learning_rate": 1.5140513965520945e-05, + "loss": 0.9724, + "step": 22746 + }, + { + "epoch": 1.46, + "learning_rate": 1.5140299276497994e-05, + "loss": 1.008, + "step": 22747 + }, + { + "epoch": 1.46, + "learning_rate": 1.5140084587475044e-05, + "loss": 0.8996, + "step": 22748 + }, + { + "epoch": 1.46, + "learning_rate": 1.5139869898452093e-05, + "loss": 0.9498, + "step": 22749 + }, + { + "epoch": 1.46, + "learning_rate": 1.5139655209429142e-05, + "loss": 0.9479, + "step": 22750 + }, + { + "epoch": 1.46, + "learning_rate": 1.5139440520406193e-05, + "loss": 0.9421, + "step": 22751 + }, + { + "epoch": 1.46, + "learning_rate": 1.5139225831383242e-05, + "loss": 0.9791, + "step": 22752 + }, + { + "epoch": 1.46, + "learning_rate": 1.513901114236029e-05, + "loss": 0.9531, + "step": 22753 + }, + { + "epoch": 1.46, + "learning_rate": 1.5138796453337341e-05, + "loss": 0.9969, + "step": 22754 + }, + { + "epoch": 1.46, + "learning_rate": 1.5138581764314392e-05, + "loss": 1.0576, + "step": 22755 + }, + { + "epoch": 1.46, + "learning_rate": 1.5138367075291443e-05, + "loss": 0.9484, + "step": 22756 + }, + { + "epoch": 1.46, + "learning_rate": 1.5138152386268492e-05, + "loss": 0.9338, + "step": 22757 + }, + { + "epoch": 1.46, + "learning_rate": 1.5137937697245542e-05, + "loss": 1.0245, + "step": 22758 + }, + { + "epoch": 1.46, + "learning_rate": 1.5137723008222591e-05, + "loss": 0.9245, + "step": 22759 + }, + { + "epoch": 1.46, + "learning_rate": 1.513750831919964e-05, + "loss": 0.9668, + "step": 22760 + }, + { + "epoch": 1.46, + "learning_rate": 1.5137293630176691e-05, + "loss": 0.9039, + "step": 22761 + }, + { + "epoch": 1.46, + "learning_rate": 1.513707894115374e-05, + "loss": 0.9044, + "step": 22762 + }, + { + "epoch": 1.46, + "learning_rate": 1.5136864252130789e-05, + "loss": 0.9476, + "step": 22763 + }, + { + "epoch": 1.46, + "learning_rate": 1.513664956310784e-05, + "loss": 0.9793, + "step": 22764 + }, + { + "epoch": 1.46, + "learning_rate": 1.5136434874084888e-05, + "loss": 0.9442, + "step": 22765 + }, + { + "epoch": 1.46, + "learning_rate": 1.513622018506194e-05, + "loss": 0.9913, + "step": 22766 + }, + { + "epoch": 1.46, + "learning_rate": 1.5136005496038988e-05, + "loss": 1.046, + "step": 22767 + }, + { + "epoch": 1.46, + "learning_rate": 1.5135790807016037e-05, + "loss": 0.963, + "step": 22768 + }, + { + "epoch": 1.46, + "learning_rate": 1.513557611799309e-05, + "loss": 1.0412, + "step": 22769 + }, + { + "epoch": 1.46, + "learning_rate": 1.5135361428970138e-05, + "loss": 0.9125, + "step": 22770 + }, + { + "epoch": 1.47, + "learning_rate": 1.5135146739947189e-05, + "loss": 0.9754, + "step": 22771 + }, + { + "epoch": 1.47, + "learning_rate": 1.5134932050924238e-05, + "loss": 1.0256, + "step": 22772 + }, + { + "epoch": 1.47, + "learning_rate": 1.5134717361901287e-05, + "loss": 0.9979, + "step": 22773 + }, + { + "epoch": 1.47, + "learning_rate": 1.5134502672878338e-05, + "loss": 0.9313, + "step": 22774 + }, + { + "epoch": 1.47, + "learning_rate": 1.5134287983855387e-05, + "loss": 0.9606, + "step": 22775 + }, + { + "epoch": 1.47, + "learning_rate": 1.5134073294832437e-05, + "loss": 1.0054, + "step": 22776 + }, + { + "epoch": 1.47, + "learning_rate": 1.5133858605809486e-05, + "loss": 0.9894, + "step": 22777 + }, + { + "epoch": 1.47, + "learning_rate": 1.5133643916786535e-05, + "loss": 0.8675, + "step": 22778 + }, + { + "epoch": 1.47, + "learning_rate": 1.5133429227763586e-05, + "loss": 1.0018, + "step": 22779 + }, + { + "epoch": 1.47, + "learning_rate": 1.5133214538740635e-05, + "loss": 0.9717, + "step": 22780 + }, + { + "epoch": 1.47, + "learning_rate": 1.5132999849717684e-05, + "loss": 0.9345, + "step": 22781 + }, + { + "epoch": 1.47, + "learning_rate": 1.5132785160694734e-05, + "loss": 0.9521, + "step": 22782 + }, + { + "epoch": 1.47, + "learning_rate": 1.5132570471671783e-05, + "loss": 0.8945, + "step": 22783 + }, + { + "epoch": 1.47, + "learning_rate": 1.5132355782648836e-05, + "loss": 0.9692, + "step": 22784 + }, + { + "epoch": 1.47, + "learning_rate": 1.5132141093625885e-05, + "loss": 0.9913, + "step": 22785 + }, + { + "epoch": 1.47, + "learning_rate": 1.5131926404602935e-05, + "loss": 1.0037, + "step": 22786 + }, + { + "epoch": 1.47, + "learning_rate": 1.5131711715579984e-05, + "loss": 1.0024, + "step": 22787 + }, + { + "epoch": 1.47, + "learning_rate": 1.5131497026557033e-05, + "loss": 0.9076, + "step": 22788 + }, + { + "epoch": 1.47, + "learning_rate": 1.5131282337534084e-05, + "loss": 0.924, + "step": 22789 + }, + { + "epoch": 1.47, + "learning_rate": 1.5131067648511133e-05, + "loss": 0.9544, + "step": 22790 + }, + { + "epoch": 1.47, + "learning_rate": 1.5130852959488182e-05, + "loss": 0.9678, + "step": 22791 + }, + { + "epoch": 1.47, + "learning_rate": 1.5130638270465233e-05, + "loss": 1.0618, + "step": 22792 + }, + { + "epoch": 1.47, + "learning_rate": 1.5130423581442282e-05, + "loss": 0.9246, + "step": 22793 + }, + { + "epoch": 1.47, + "learning_rate": 1.513020889241933e-05, + "loss": 1.0269, + "step": 22794 + }, + { + "epoch": 1.47, + "learning_rate": 1.5129994203396381e-05, + "loss": 1.0236, + "step": 22795 + }, + { + "epoch": 1.47, + "learning_rate": 1.512977951437343e-05, + "loss": 1.0159, + "step": 22796 + }, + { + "epoch": 1.47, + "learning_rate": 1.512956482535048e-05, + "loss": 0.8942, + "step": 22797 + }, + { + "epoch": 1.47, + "learning_rate": 1.5129350136327531e-05, + "loss": 0.8748, + "step": 22798 + }, + { + "epoch": 1.47, + "learning_rate": 1.5129135447304582e-05, + "loss": 0.9698, + "step": 22799 + }, + { + "epoch": 1.47, + "learning_rate": 1.5128920758281631e-05, + "loss": 0.9459, + "step": 22800 + }, + { + "epoch": 1.47, + "learning_rate": 1.512870606925868e-05, + "loss": 0.9549, + "step": 22801 + }, + { + "epoch": 1.47, + "learning_rate": 1.512849138023573e-05, + "loss": 1.0407, + "step": 22802 + }, + { + "epoch": 1.47, + "learning_rate": 1.512827669121278e-05, + "loss": 1.0353, + "step": 22803 + }, + { + "epoch": 1.47, + "learning_rate": 1.5128062002189829e-05, + "loss": 0.9514, + "step": 22804 + }, + { + "epoch": 1.47, + "learning_rate": 1.512784731316688e-05, + "loss": 1.0018, + "step": 22805 + }, + { + "epoch": 1.47, + "learning_rate": 1.5127632624143928e-05, + "loss": 0.9616, + "step": 22806 + }, + { + "epoch": 1.47, + "learning_rate": 1.5127417935120979e-05, + "loss": 0.9659, + "step": 22807 + }, + { + "epoch": 1.47, + "learning_rate": 1.5127203246098028e-05, + "loss": 1.0047, + "step": 22808 + }, + { + "epoch": 1.47, + "learning_rate": 1.5126988557075077e-05, + "loss": 0.947, + "step": 22809 + }, + { + "epoch": 1.47, + "learning_rate": 1.5126773868052127e-05, + "loss": 0.9362, + "step": 22810 + }, + { + "epoch": 1.47, + "learning_rate": 1.5126559179029176e-05, + "loss": 0.954, + "step": 22811 + }, + { + "epoch": 1.47, + "learning_rate": 1.5126344490006225e-05, + "loss": 1.0157, + "step": 22812 + }, + { + "epoch": 1.47, + "learning_rate": 1.5126129800983278e-05, + "loss": 1.0043, + "step": 22813 + }, + { + "epoch": 1.47, + "learning_rate": 1.5125915111960327e-05, + "loss": 0.9838, + "step": 22814 + }, + { + "epoch": 1.47, + "learning_rate": 1.5125700422937377e-05, + "loss": 0.9701, + "step": 22815 + }, + { + "epoch": 1.47, + "learning_rate": 1.5125485733914426e-05, + "loss": 0.9992, + "step": 22816 + }, + { + "epoch": 1.47, + "learning_rate": 1.5125271044891477e-05, + "loss": 0.9281, + "step": 22817 + }, + { + "epoch": 1.47, + "learning_rate": 1.5125056355868526e-05, + "loss": 0.9083, + "step": 22818 + }, + { + "epoch": 1.47, + "learning_rate": 1.5124841666845575e-05, + "loss": 0.971, + "step": 22819 + }, + { + "epoch": 1.47, + "learning_rate": 1.5124626977822626e-05, + "loss": 0.9734, + "step": 22820 + }, + { + "epoch": 1.47, + "learning_rate": 1.5124412288799675e-05, + "loss": 0.914, + "step": 22821 + }, + { + "epoch": 1.47, + "learning_rate": 1.5124197599776724e-05, + "loss": 1.0146, + "step": 22822 + }, + { + "epoch": 1.47, + "learning_rate": 1.5123982910753774e-05, + "loss": 0.9462, + "step": 22823 + }, + { + "epoch": 1.47, + "learning_rate": 1.5123768221730823e-05, + "loss": 0.9006, + "step": 22824 + }, + { + "epoch": 1.47, + "learning_rate": 1.5123553532707874e-05, + "loss": 0.9529, + "step": 22825 + }, + { + "epoch": 1.47, + "learning_rate": 1.5123338843684923e-05, + "loss": 1.0482, + "step": 22826 + }, + { + "epoch": 1.47, + "learning_rate": 1.5123124154661975e-05, + "loss": 0.9213, + "step": 22827 + }, + { + "epoch": 1.47, + "learning_rate": 1.5122909465639024e-05, + "loss": 0.9271, + "step": 22828 + }, + { + "epoch": 1.47, + "learning_rate": 1.5122694776616073e-05, + "loss": 0.9668, + "step": 22829 + }, + { + "epoch": 1.47, + "learning_rate": 1.5122480087593124e-05, + "loss": 0.9839, + "step": 22830 + }, + { + "epoch": 1.47, + "learning_rate": 1.5122265398570173e-05, + "loss": 0.9821, + "step": 22831 + }, + { + "epoch": 1.47, + "learning_rate": 1.5122050709547222e-05, + "loss": 0.9119, + "step": 22832 + }, + { + "epoch": 1.47, + "learning_rate": 1.5121836020524272e-05, + "loss": 0.9237, + "step": 22833 + }, + { + "epoch": 1.47, + "learning_rate": 1.5121621331501321e-05, + "loss": 1.0294, + "step": 22834 + }, + { + "epoch": 1.47, + "learning_rate": 1.5121406642478372e-05, + "loss": 1.0623, + "step": 22835 + }, + { + "epoch": 1.47, + "learning_rate": 1.5121191953455421e-05, + "loss": 0.9323, + "step": 22836 + }, + { + "epoch": 1.47, + "learning_rate": 1.512097726443247e-05, + "loss": 1.0217, + "step": 22837 + }, + { + "epoch": 1.47, + "learning_rate": 1.512076257540952e-05, + "loss": 1.0313, + "step": 22838 + }, + { + "epoch": 1.47, + "learning_rate": 1.512054788638657e-05, + "loss": 1.003, + "step": 22839 + }, + { + "epoch": 1.47, + "learning_rate": 1.5120333197363618e-05, + "loss": 1.0293, + "step": 22840 + }, + { + "epoch": 1.47, + "learning_rate": 1.5120118508340669e-05, + "loss": 0.9953, + "step": 22841 + }, + { + "epoch": 1.47, + "learning_rate": 1.511990381931772e-05, + "loss": 0.9702, + "step": 22842 + }, + { + "epoch": 1.47, + "learning_rate": 1.511968913029477e-05, + "loss": 1.0129, + "step": 22843 + }, + { + "epoch": 1.47, + "learning_rate": 1.511947444127182e-05, + "loss": 0.8997, + "step": 22844 + }, + { + "epoch": 1.47, + "learning_rate": 1.511925975224887e-05, + "loss": 0.989, + "step": 22845 + }, + { + "epoch": 1.47, + "learning_rate": 1.5119045063225919e-05, + "loss": 0.9911, + "step": 22846 + }, + { + "epoch": 1.47, + "learning_rate": 1.5118830374202968e-05, + "loss": 0.9917, + "step": 22847 + }, + { + "epoch": 1.47, + "learning_rate": 1.5118615685180019e-05, + "loss": 0.9411, + "step": 22848 + }, + { + "epoch": 1.47, + "learning_rate": 1.5118400996157068e-05, + "loss": 1.0035, + "step": 22849 + }, + { + "epoch": 1.47, + "learning_rate": 1.5118186307134117e-05, + "loss": 0.9841, + "step": 22850 + }, + { + "epoch": 1.47, + "learning_rate": 1.5117971618111167e-05, + "loss": 0.9378, + "step": 22851 + }, + { + "epoch": 1.47, + "learning_rate": 1.5117756929088216e-05, + "loss": 1.0274, + "step": 22852 + }, + { + "epoch": 1.47, + "learning_rate": 1.5117542240065267e-05, + "loss": 1.0305, + "step": 22853 + }, + { + "epoch": 1.47, + "learning_rate": 1.5117327551042316e-05, + "loss": 0.9115, + "step": 22854 + }, + { + "epoch": 1.47, + "learning_rate": 1.5117112862019365e-05, + "loss": 1.009, + "step": 22855 + }, + { + "epoch": 1.47, + "learning_rate": 1.5116898172996415e-05, + "loss": 0.9616, + "step": 22856 + }, + { + "epoch": 1.47, + "learning_rate": 1.5116683483973466e-05, + "loss": 0.9998, + "step": 22857 + }, + { + "epoch": 1.47, + "learning_rate": 1.5116468794950517e-05, + "loss": 0.9226, + "step": 22858 + }, + { + "epoch": 1.47, + "learning_rate": 1.5116254105927566e-05, + "loss": 0.9562, + "step": 22859 + }, + { + "epoch": 1.47, + "learning_rate": 1.5116039416904615e-05, + "loss": 0.8959, + "step": 22860 + }, + { + "epoch": 1.47, + "learning_rate": 1.5115824727881665e-05, + "loss": 1.0295, + "step": 22861 + }, + { + "epoch": 1.47, + "learning_rate": 1.5115610038858714e-05, + "loss": 0.9714, + "step": 22862 + }, + { + "epoch": 1.47, + "learning_rate": 1.5115395349835765e-05, + "loss": 0.9713, + "step": 22863 + }, + { + "epoch": 1.47, + "learning_rate": 1.5115180660812814e-05, + "loss": 0.9258, + "step": 22864 + }, + { + "epoch": 1.47, + "learning_rate": 1.5114965971789863e-05, + "loss": 1.0014, + "step": 22865 + }, + { + "epoch": 1.47, + "learning_rate": 1.5114751282766914e-05, + "loss": 0.9736, + "step": 22866 + }, + { + "epoch": 1.47, + "learning_rate": 1.5114536593743963e-05, + "loss": 0.9207, + "step": 22867 + }, + { + "epoch": 1.47, + "learning_rate": 1.5114321904721012e-05, + "loss": 0.9578, + "step": 22868 + }, + { + "epoch": 1.47, + "learning_rate": 1.5114107215698062e-05, + "loss": 0.944, + "step": 22869 + }, + { + "epoch": 1.47, + "learning_rate": 1.5113892526675111e-05, + "loss": 1.0126, + "step": 22870 + }, + { + "epoch": 1.47, + "learning_rate": 1.5113677837652163e-05, + "loss": 1.0218, + "step": 22871 + }, + { + "epoch": 1.47, + "learning_rate": 1.5113463148629212e-05, + "loss": 0.9289, + "step": 22872 + }, + { + "epoch": 1.47, + "learning_rate": 1.5113248459606263e-05, + "loss": 0.9793, + "step": 22873 + }, + { + "epoch": 1.47, + "learning_rate": 1.5113033770583312e-05, + "loss": 0.9765, + "step": 22874 + }, + { + "epoch": 1.47, + "learning_rate": 1.5112819081560361e-05, + "loss": 0.9146, + "step": 22875 + }, + { + "epoch": 1.47, + "learning_rate": 1.5112604392537412e-05, + "loss": 1.0036, + "step": 22876 + }, + { + "epoch": 1.47, + "learning_rate": 1.511238970351446e-05, + "loss": 0.9575, + "step": 22877 + }, + { + "epoch": 1.47, + "learning_rate": 1.511217501449151e-05, + "loss": 0.9682, + "step": 22878 + }, + { + "epoch": 1.47, + "learning_rate": 1.511196032546856e-05, + "loss": 1.0387, + "step": 22879 + }, + { + "epoch": 1.47, + "learning_rate": 1.511174563644561e-05, + "loss": 0.9422, + "step": 22880 + }, + { + "epoch": 1.47, + "learning_rate": 1.5111530947422658e-05, + "loss": 0.9759, + "step": 22881 + }, + { + "epoch": 1.47, + "learning_rate": 1.5111316258399709e-05, + "loss": 0.9416, + "step": 22882 + }, + { + "epoch": 1.47, + "learning_rate": 1.5111101569376758e-05, + "loss": 1.0092, + "step": 22883 + }, + { + "epoch": 1.47, + "learning_rate": 1.5110886880353808e-05, + "loss": 0.9234, + "step": 22884 + }, + { + "epoch": 1.47, + "learning_rate": 1.5110672191330857e-05, + "loss": 0.993, + "step": 22885 + }, + { + "epoch": 1.47, + "learning_rate": 1.511045750230791e-05, + "loss": 0.9357, + "step": 22886 + }, + { + "epoch": 1.47, + "learning_rate": 1.5110242813284959e-05, + "loss": 0.9785, + "step": 22887 + }, + { + "epoch": 1.47, + "learning_rate": 1.5110028124262008e-05, + "loss": 0.9817, + "step": 22888 + }, + { + "epoch": 1.47, + "learning_rate": 1.5109813435239058e-05, + "loss": 1.0363, + "step": 22889 + }, + { + "epoch": 1.47, + "learning_rate": 1.5109598746216107e-05, + "loss": 0.9553, + "step": 22890 + }, + { + "epoch": 1.47, + "learning_rate": 1.5109384057193156e-05, + "loss": 0.9018, + "step": 22891 + }, + { + "epoch": 1.47, + "learning_rate": 1.5109169368170207e-05, + "loss": 1.1036, + "step": 22892 + }, + { + "epoch": 1.47, + "learning_rate": 1.5108954679147256e-05, + "loss": 1.0209, + "step": 22893 + }, + { + "epoch": 1.47, + "learning_rate": 1.5108739990124307e-05, + "loss": 0.9004, + "step": 22894 + }, + { + "epoch": 1.47, + "learning_rate": 1.5108525301101356e-05, + "loss": 0.9556, + "step": 22895 + }, + { + "epoch": 1.47, + "learning_rate": 1.5108310612078405e-05, + "loss": 1.0342, + "step": 22896 + }, + { + "epoch": 1.47, + "learning_rate": 1.5108095923055455e-05, + "loss": 0.9563, + "step": 22897 + }, + { + "epoch": 1.47, + "learning_rate": 1.5107881234032504e-05, + "loss": 0.9346, + "step": 22898 + }, + { + "epoch": 1.47, + "learning_rate": 1.5107666545009553e-05, + "loss": 0.9913, + "step": 22899 + }, + { + "epoch": 1.47, + "learning_rate": 1.5107451855986605e-05, + "loss": 0.931, + "step": 22900 + }, + { + "epoch": 1.47, + "learning_rate": 1.5107237166963654e-05, + "loss": 0.9971, + "step": 22901 + }, + { + "epoch": 1.47, + "learning_rate": 1.5107022477940705e-05, + "loss": 1.0092, + "step": 22902 + }, + { + "epoch": 1.47, + "learning_rate": 1.5106807788917754e-05, + "loss": 0.9575, + "step": 22903 + }, + { + "epoch": 1.47, + "learning_rate": 1.5106593099894805e-05, + "loss": 0.968, + "step": 22904 + }, + { + "epoch": 1.47, + "learning_rate": 1.5106378410871854e-05, + "loss": 0.9273, + "step": 22905 + }, + { + "epoch": 1.47, + "learning_rate": 1.5106163721848903e-05, + "loss": 0.9257, + "step": 22906 + }, + { + "epoch": 1.47, + "learning_rate": 1.5105949032825953e-05, + "loss": 1.033, + "step": 22907 + }, + { + "epoch": 1.47, + "learning_rate": 1.5105734343803002e-05, + "loss": 0.9898, + "step": 22908 + }, + { + "epoch": 1.47, + "learning_rate": 1.5105519654780051e-05, + "loss": 0.9017, + "step": 22909 + }, + { + "epoch": 1.47, + "learning_rate": 1.5105304965757102e-05, + "loss": 1.0074, + "step": 22910 + }, + { + "epoch": 1.47, + "learning_rate": 1.5105090276734151e-05, + "loss": 1.0142, + "step": 22911 + }, + { + "epoch": 1.47, + "learning_rate": 1.5104875587711202e-05, + "loss": 0.9784, + "step": 22912 + }, + { + "epoch": 1.47, + "learning_rate": 1.510466089868825e-05, + "loss": 0.9249, + "step": 22913 + }, + { + "epoch": 1.47, + "learning_rate": 1.51044462096653e-05, + "loss": 0.9799, + "step": 22914 + }, + { + "epoch": 1.47, + "learning_rate": 1.5104231520642352e-05, + "loss": 0.9521, + "step": 22915 + }, + { + "epoch": 1.47, + "learning_rate": 1.51040168316194e-05, + "loss": 0.9242, + "step": 22916 + }, + { + "epoch": 1.47, + "learning_rate": 1.5103802142596451e-05, + "loss": 0.9161, + "step": 22917 + }, + { + "epoch": 1.47, + "learning_rate": 1.51035874535735e-05, + "loss": 1.0605, + "step": 22918 + }, + { + "epoch": 1.47, + "learning_rate": 1.510337276455055e-05, + "loss": 1.0235, + "step": 22919 + }, + { + "epoch": 1.47, + "learning_rate": 1.51031580755276e-05, + "loss": 0.9067, + "step": 22920 + }, + { + "epoch": 1.47, + "learning_rate": 1.5102943386504649e-05, + "loss": 0.9146, + "step": 22921 + }, + { + "epoch": 1.47, + "learning_rate": 1.51027286974817e-05, + "loss": 1.0302, + "step": 22922 + }, + { + "epoch": 1.47, + "learning_rate": 1.5102514008458749e-05, + "loss": 1.0007, + "step": 22923 + }, + { + "epoch": 1.47, + "learning_rate": 1.5102299319435798e-05, + "loss": 0.9688, + "step": 22924 + }, + { + "epoch": 1.47, + "learning_rate": 1.5102084630412848e-05, + "loss": 0.8904, + "step": 22925 + }, + { + "epoch": 1.48, + "learning_rate": 1.5101869941389897e-05, + "loss": 0.9734, + "step": 22926 + }, + { + "epoch": 1.48, + "learning_rate": 1.5101655252366946e-05, + "loss": 0.9216, + "step": 22927 + }, + { + "epoch": 1.48, + "learning_rate": 1.5101440563343997e-05, + "loss": 0.8991, + "step": 22928 + }, + { + "epoch": 1.48, + "learning_rate": 1.5101225874321047e-05, + "loss": 0.94, + "step": 22929 + }, + { + "epoch": 1.48, + "learning_rate": 1.5101011185298098e-05, + "loss": 1.0236, + "step": 22930 + }, + { + "epoch": 1.48, + "learning_rate": 1.5100796496275147e-05, + "loss": 1.007, + "step": 22931 + }, + { + "epoch": 1.48, + "learning_rate": 1.5100581807252198e-05, + "loss": 0.9478, + "step": 22932 + }, + { + "epoch": 1.48, + "learning_rate": 1.5100367118229247e-05, + "loss": 0.9457, + "step": 22933 + }, + { + "epoch": 1.48, + "learning_rate": 1.5100152429206296e-05, + "loss": 0.9704, + "step": 22934 + }, + { + "epoch": 1.48, + "learning_rate": 1.5099937740183346e-05, + "loss": 1.0768, + "step": 22935 + }, + { + "epoch": 1.48, + "learning_rate": 1.5099723051160395e-05, + "loss": 0.949, + "step": 22936 + }, + { + "epoch": 1.48, + "learning_rate": 1.5099508362137444e-05, + "loss": 0.9548, + "step": 22937 + }, + { + "epoch": 1.48, + "learning_rate": 1.5099293673114495e-05, + "loss": 0.979, + "step": 22938 + }, + { + "epoch": 1.48, + "learning_rate": 1.5099078984091544e-05, + "loss": 0.9823, + "step": 22939 + }, + { + "epoch": 1.48, + "learning_rate": 1.5098864295068595e-05, + "loss": 0.9081, + "step": 22940 + }, + { + "epoch": 1.48, + "learning_rate": 1.5098649606045644e-05, + "loss": 0.9738, + "step": 22941 + }, + { + "epoch": 1.48, + "learning_rate": 1.5098434917022693e-05, + "loss": 0.9544, + "step": 22942 + }, + { + "epoch": 1.48, + "learning_rate": 1.5098220227999743e-05, + "loss": 0.9926, + "step": 22943 + }, + { + "epoch": 1.48, + "learning_rate": 1.5098005538976794e-05, + "loss": 0.9747, + "step": 22944 + }, + { + "epoch": 1.48, + "learning_rate": 1.5097790849953844e-05, + "loss": 0.936, + "step": 22945 + }, + { + "epoch": 1.48, + "learning_rate": 1.5097576160930893e-05, + "loss": 0.9354, + "step": 22946 + }, + { + "epoch": 1.48, + "learning_rate": 1.5097361471907942e-05, + "loss": 0.9433, + "step": 22947 + }, + { + "epoch": 1.48, + "learning_rate": 1.5097146782884993e-05, + "loss": 1.0799, + "step": 22948 + }, + { + "epoch": 1.48, + "learning_rate": 1.5096932093862042e-05, + "loss": 1.0157, + "step": 22949 + }, + { + "epoch": 1.48, + "learning_rate": 1.5096717404839093e-05, + "loss": 0.945, + "step": 22950 + }, + { + "epoch": 1.48, + "learning_rate": 1.5096502715816142e-05, + "loss": 0.9379, + "step": 22951 + }, + { + "epoch": 1.48, + "learning_rate": 1.509628802679319e-05, + "loss": 0.9791, + "step": 22952 + }, + { + "epoch": 1.48, + "learning_rate": 1.5096073337770241e-05, + "loss": 0.9307, + "step": 22953 + }, + { + "epoch": 1.48, + "learning_rate": 1.509585864874729e-05, + "loss": 0.906, + "step": 22954 + }, + { + "epoch": 1.48, + "learning_rate": 1.509564395972434e-05, + "loss": 0.9186, + "step": 22955 + }, + { + "epoch": 1.48, + "learning_rate": 1.509542927070139e-05, + "loss": 0.9748, + "step": 22956 + }, + { + "epoch": 1.48, + "learning_rate": 1.5095214581678439e-05, + "loss": 0.9796, + "step": 22957 + }, + { + "epoch": 1.48, + "learning_rate": 1.5094999892655491e-05, + "loss": 1.0355, + "step": 22958 + }, + { + "epoch": 1.48, + "learning_rate": 1.509478520363254e-05, + "loss": 1.0105, + "step": 22959 + }, + { + "epoch": 1.48, + "learning_rate": 1.509457051460959e-05, + "loss": 0.987, + "step": 22960 + }, + { + "epoch": 1.48, + "learning_rate": 1.509435582558664e-05, + "loss": 0.9296, + "step": 22961 + }, + { + "epoch": 1.48, + "learning_rate": 1.5094141136563689e-05, + "loss": 0.9714, + "step": 22962 + }, + { + "epoch": 1.48, + "learning_rate": 1.509392644754074e-05, + "loss": 0.9627, + "step": 22963 + }, + { + "epoch": 1.48, + "learning_rate": 1.5093711758517788e-05, + "loss": 0.9284, + "step": 22964 + }, + { + "epoch": 1.48, + "learning_rate": 1.5093497069494837e-05, + "loss": 0.9319, + "step": 22965 + }, + { + "epoch": 1.48, + "learning_rate": 1.5093282380471888e-05, + "loss": 0.9877, + "step": 22966 + }, + { + "epoch": 1.48, + "learning_rate": 1.5093067691448937e-05, + "loss": 0.9629, + "step": 22967 + }, + { + "epoch": 1.48, + "learning_rate": 1.5092853002425986e-05, + "loss": 0.9575, + "step": 22968 + }, + { + "epoch": 1.48, + "learning_rate": 1.5092638313403037e-05, + "loss": 1.0004, + "step": 22969 + }, + { + "epoch": 1.48, + "learning_rate": 1.5092423624380086e-05, + "loss": 1.005, + "step": 22970 + }, + { + "epoch": 1.48, + "learning_rate": 1.5092208935357136e-05, + "loss": 1.0211, + "step": 22971 + }, + { + "epoch": 1.48, + "learning_rate": 1.5091994246334185e-05, + "loss": 0.9738, + "step": 22972 + }, + { + "epoch": 1.48, + "learning_rate": 1.5091779557311238e-05, + "loss": 0.9723, + "step": 22973 + }, + { + "epoch": 1.48, + "learning_rate": 1.5091564868288286e-05, + "loss": 0.9194, + "step": 22974 + }, + { + "epoch": 1.48, + "learning_rate": 1.5091350179265335e-05, + "loss": 1.018, + "step": 22975 + }, + { + "epoch": 1.48, + "learning_rate": 1.5091135490242386e-05, + "loss": 0.9766, + "step": 22976 + }, + { + "epoch": 1.48, + "learning_rate": 1.5090920801219435e-05, + "loss": 1.0252, + "step": 22977 + }, + { + "epoch": 1.48, + "learning_rate": 1.5090706112196484e-05, + "loss": 0.9369, + "step": 22978 + }, + { + "epoch": 1.48, + "learning_rate": 1.5090491423173535e-05, + "loss": 1.0851, + "step": 22979 + }, + { + "epoch": 1.48, + "learning_rate": 1.5090276734150584e-05, + "loss": 1.0625, + "step": 22980 + }, + { + "epoch": 1.48, + "learning_rate": 1.5090062045127634e-05, + "loss": 1.1093, + "step": 22981 + }, + { + "epoch": 1.48, + "learning_rate": 1.5089847356104683e-05, + "loss": 1.0286, + "step": 22982 + }, + { + "epoch": 1.48, + "learning_rate": 1.5089632667081732e-05, + "loss": 0.9192, + "step": 22983 + }, + { + "epoch": 1.48, + "learning_rate": 1.5089417978058783e-05, + "loss": 0.9865, + "step": 22984 + }, + { + "epoch": 1.48, + "learning_rate": 1.5089203289035832e-05, + "loss": 0.9884, + "step": 22985 + }, + { + "epoch": 1.48, + "learning_rate": 1.5088988600012881e-05, + "loss": 1.0076, + "step": 22986 + }, + { + "epoch": 1.48, + "learning_rate": 1.5088773910989933e-05, + "loss": 0.9307, + "step": 22987 + }, + { + "epoch": 1.48, + "learning_rate": 1.5088559221966982e-05, + "loss": 1.0355, + "step": 22988 + }, + { + "epoch": 1.48, + "learning_rate": 1.5088344532944033e-05, + "loss": 0.9485, + "step": 22989 + }, + { + "epoch": 1.48, + "learning_rate": 1.5088129843921082e-05, + "loss": 0.9708, + "step": 22990 + }, + { + "epoch": 1.48, + "learning_rate": 1.5087915154898132e-05, + "loss": 1.0089, + "step": 22991 + }, + { + "epoch": 1.48, + "learning_rate": 1.5087700465875181e-05, + "loss": 0.9364, + "step": 22992 + }, + { + "epoch": 1.48, + "learning_rate": 1.508748577685223e-05, + "loss": 1.0039, + "step": 22993 + }, + { + "epoch": 1.48, + "learning_rate": 1.5087271087829281e-05, + "loss": 0.9671, + "step": 22994 + }, + { + "epoch": 1.48, + "learning_rate": 1.508705639880633e-05, + "loss": 0.9921, + "step": 22995 + }, + { + "epoch": 1.48, + "learning_rate": 1.5086841709783379e-05, + "loss": 0.8964, + "step": 22996 + }, + { + "epoch": 1.48, + "learning_rate": 1.508662702076043e-05, + "loss": 1.0038, + "step": 22997 + }, + { + "epoch": 1.48, + "learning_rate": 1.5086412331737479e-05, + "loss": 0.9751, + "step": 22998 + }, + { + "epoch": 1.48, + "learning_rate": 1.508619764271453e-05, + "loss": 1.0067, + "step": 22999 + }, + { + "epoch": 1.48, + "learning_rate": 1.5085982953691578e-05, + "loss": 0.9997, + "step": 23000 + }, + { + "epoch": 1.48, + "eval_laion/cider": 0.48263913201228326, + "eval_laion/meteor": 0.3808464596852842, + "eval_laion/pred_num_words": 35.2890625, + "eval_laion/sacrebleu": 15.142131305309778, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0493484735488892, + "eval_runtime": 95.9675, + "eval_samples_per_second": 5.2, + "eval_steps_per_second": 0.333, + "step": 23000 + }, + { + "epoch": 1.48, + "learning_rate": 1.5085768264668627e-05, + "loss": 1.0895, + "step": 23001 + }, + { + "epoch": 1.48, + "learning_rate": 1.508555357564568e-05, + "loss": 0.9181, + "step": 23002 + }, + { + "epoch": 1.48, + "learning_rate": 1.5085338886622728e-05, + "loss": 0.9535, + "step": 23003 + }, + { + "epoch": 1.48, + "learning_rate": 1.508512419759978e-05, + "loss": 1.0175, + "step": 23004 + }, + { + "epoch": 1.48, + "learning_rate": 1.5084909508576828e-05, + "loss": 0.9695, + "step": 23005 + }, + { + "epoch": 1.48, + "learning_rate": 1.5084694819553877e-05, + "loss": 0.9084, + "step": 23006 + }, + { + "epoch": 1.48, + "learning_rate": 1.5084480130530928e-05, + "loss": 1.0171, + "step": 23007 + }, + { + "epoch": 1.48, + "learning_rate": 1.5084265441507977e-05, + "loss": 1.0047, + "step": 23008 + }, + { + "epoch": 1.48, + "learning_rate": 1.5084050752485027e-05, + "loss": 0.9909, + "step": 23009 + }, + { + "epoch": 1.48, + "learning_rate": 1.5083836063462076e-05, + "loss": 0.9318, + "step": 23010 + }, + { + "epoch": 1.48, + "learning_rate": 1.5083621374439125e-05, + "loss": 0.9506, + "step": 23011 + }, + { + "epoch": 1.48, + "learning_rate": 1.5083406685416176e-05, + "loss": 0.989, + "step": 23012 + }, + { + "epoch": 1.48, + "learning_rate": 1.5083191996393225e-05, + "loss": 1.0033, + "step": 23013 + }, + { + "epoch": 1.48, + "learning_rate": 1.5082977307370274e-05, + "loss": 0.9094, + "step": 23014 + }, + { + "epoch": 1.48, + "learning_rate": 1.5082762618347325e-05, + "loss": 0.9787, + "step": 23015 + }, + { + "epoch": 1.48, + "learning_rate": 1.5082547929324375e-05, + "loss": 0.9543, + "step": 23016 + }, + { + "epoch": 1.48, + "learning_rate": 1.5082333240301426e-05, + "loss": 1.0245, + "step": 23017 + }, + { + "epoch": 1.48, + "learning_rate": 1.5082118551278475e-05, + "loss": 0.9589, + "step": 23018 + }, + { + "epoch": 1.48, + "learning_rate": 1.5081903862255525e-05, + "loss": 0.9703, + "step": 23019 + }, + { + "epoch": 1.48, + "learning_rate": 1.5081689173232574e-05, + "loss": 1.0857, + "step": 23020 + }, + { + "epoch": 1.48, + "learning_rate": 1.5081474484209623e-05, + "loss": 0.9413, + "step": 23021 + }, + { + "epoch": 1.48, + "learning_rate": 1.5081259795186674e-05, + "loss": 0.9598, + "step": 23022 + }, + { + "epoch": 1.48, + "learning_rate": 1.5081045106163723e-05, + "loss": 1.0397, + "step": 23023 + }, + { + "epoch": 1.48, + "learning_rate": 1.5080830417140772e-05, + "loss": 1.0341, + "step": 23024 + }, + { + "epoch": 1.48, + "learning_rate": 1.5080615728117823e-05, + "loss": 1.0511, + "step": 23025 + }, + { + "epoch": 1.48, + "learning_rate": 1.5080401039094872e-05, + "loss": 0.9803, + "step": 23026 + }, + { + "epoch": 1.48, + "learning_rate": 1.508018635007192e-05, + "loss": 0.9694, + "step": 23027 + }, + { + "epoch": 1.48, + "learning_rate": 1.5079971661048971e-05, + "loss": 0.9882, + "step": 23028 + }, + { + "epoch": 1.48, + "learning_rate": 1.507975697202602e-05, + "loss": 0.979, + "step": 23029 + }, + { + "epoch": 1.48, + "learning_rate": 1.5079542283003071e-05, + "loss": 0.945, + "step": 23030 + }, + { + "epoch": 1.48, + "learning_rate": 1.5079327593980122e-05, + "loss": 1.0121, + "step": 23031 + }, + { + "epoch": 1.48, + "learning_rate": 1.5079112904957172e-05, + "loss": 0.9366, + "step": 23032 + }, + { + "epoch": 1.48, + "learning_rate": 1.5078898215934221e-05, + "loss": 0.9894, + "step": 23033 + }, + { + "epoch": 1.48, + "learning_rate": 1.507868352691127e-05, + "loss": 0.8939, + "step": 23034 + }, + { + "epoch": 1.48, + "learning_rate": 1.507846883788832e-05, + "loss": 0.9788, + "step": 23035 + }, + { + "epoch": 1.48, + "learning_rate": 1.507825414886537e-05, + "loss": 0.9891, + "step": 23036 + }, + { + "epoch": 1.48, + "learning_rate": 1.507803945984242e-05, + "loss": 0.9882, + "step": 23037 + }, + { + "epoch": 1.48, + "learning_rate": 1.507782477081947e-05, + "loss": 0.9832, + "step": 23038 + }, + { + "epoch": 1.48, + "learning_rate": 1.5077610081796518e-05, + "loss": 0.952, + "step": 23039 + }, + { + "epoch": 1.48, + "learning_rate": 1.5077395392773569e-05, + "loss": 0.9421, + "step": 23040 + }, + { + "epoch": 1.48, + "learning_rate": 1.5077180703750618e-05, + "loss": 0.988, + "step": 23041 + }, + { + "epoch": 1.48, + "learning_rate": 1.5076966014727667e-05, + "loss": 0.9981, + "step": 23042 + }, + { + "epoch": 1.48, + "learning_rate": 1.5076751325704718e-05, + "loss": 0.9323, + "step": 23043 + }, + { + "epoch": 1.48, + "learning_rate": 1.5076536636681767e-05, + "loss": 0.9594, + "step": 23044 + }, + { + "epoch": 1.48, + "learning_rate": 1.5076321947658819e-05, + "loss": 0.9799, + "step": 23045 + }, + { + "epoch": 1.48, + "learning_rate": 1.5076107258635868e-05, + "loss": 0.9724, + "step": 23046 + }, + { + "epoch": 1.48, + "learning_rate": 1.5075892569612919e-05, + "loss": 0.9785, + "step": 23047 + }, + { + "epoch": 1.48, + "learning_rate": 1.5075677880589967e-05, + "loss": 0.9613, + "step": 23048 + }, + { + "epoch": 1.48, + "learning_rate": 1.5075463191567016e-05, + "loss": 0.9907, + "step": 23049 + }, + { + "epoch": 1.48, + "learning_rate": 1.5075248502544067e-05, + "loss": 0.9849, + "step": 23050 + }, + { + "epoch": 1.48, + "learning_rate": 1.5075033813521116e-05, + "loss": 0.942, + "step": 23051 + }, + { + "epoch": 1.48, + "learning_rate": 1.5074819124498165e-05, + "loss": 1.0245, + "step": 23052 + }, + { + "epoch": 1.48, + "learning_rate": 1.5074604435475216e-05, + "loss": 1.0524, + "step": 23053 + }, + { + "epoch": 1.48, + "learning_rate": 1.5074389746452265e-05, + "loss": 1.0389, + "step": 23054 + }, + { + "epoch": 1.48, + "learning_rate": 1.5074175057429314e-05, + "loss": 0.9862, + "step": 23055 + }, + { + "epoch": 1.48, + "learning_rate": 1.5073960368406364e-05, + "loss": 0.9275, + "step": 23056 + }, + { + "epoch": 1.48, + "learning_rate": 1.5073745679383413e-05, + "loss": 0.9554, + "step": 23057 + }, + { + "epoch": 1.48, + "learning_rate": 1.5073530990360464e-05, + "loss": 0.9571, + "step": 23058 + }, + { + "epoch": 1.48, + "learning_rate": 1.5073316301337513e-05, + "loss": 1.0048, + "step": 23059 + }, + { + "epoch": 1.48, + "learning_rate": 1.5073101612314565e-05, + "loss": 1.0092, + "step": 23060 + }, + { + "epoch": 1.48, + "learning_rate": 1.5072886923291614e-05, + "loss": 0.9557, + "step": 23061 + }, + { + "epoch": 1.48, + "learning_rate": 1.5072672234268663e-05, + "loss": 0.9538, + "step": 23062 + }, + { + "epoch": 1.48, + "learning_rate": 1.5072457545245714e-05, + "loss": 0.9709, + "step": 23063 + }, + { + "epoch": 1.48, + "learning_rate": 1.5072242856222763e-05, + "loss": 1.01, + "step": 23064 + }, + { + "epoch": 1.48, + "learning_rate": 1.5072028167199812e-05, + "loss": 0.931, + "step": 23065 + }, + { + "epoch": 1.48, + "learning_rate": 1.5071813478176862e-05, + "loss": 0.9739, + "step": 23066 + }, + { + "epoch": 1.48, + "learning_rate": 1.5071598789153911e-05, + "loss": 1.0167, + "step": 23067 + }, + { + "epoch": 1.48, + "learning_rate": 1.5071384100130962e-05, + "loss": 0.9917, + "step": 23068 + }, + { + "epoch": 1.48, + "learning_rate": 1.5071169411108011e-05, + "loss": 0.9302, + "step": 23069 + }, + { + "epoch": 1.48, + "learning_rate": 1.507095472208506e-05, + "loss": 0.9547, + "step": 23070 + }, + { + "epoch": 1.48, + "learning_rate": 1.507074003306211e-05, + "loss": 0.9377, + "step": 23071 + }, + { + "epoch": 1.48, + "learning_rate": 1.507052534403916e-05, + "loss": 0.9723, + "step": 23072 + }, + { + "epoch": 1.48, + "learning_rate": 1.5070310655016209e-05, + "loss": 0.9106, + "step": 23073 + }, + { + "epoch": 1.48, + "learning_rate": 1.5070095965993261e-05, + "loss": 0.9136, + "step": 23074 + }, + { + "epoch": 1.48, + "learning_rate": 1.506988127697031e-05, + "loss": 1.0291, + "step": 23075 + }, + { + "epoch": 1.48, + "learning_rate": 1.506966658794736e-05, + "loss": 1.0044, + "step": 23076 + }, + { + "epoch": 1.48, + "learning_rate": 1.506945189892441e-05, + "loss": 0.8915, + "step": 23077 + }, + { + "epoch": 1.48, + "learning_rate": 1.506923720990146e-05, + "loss": 1.0452, + "step": 23078 + }, + { + "epoch": 1.48, + "learning_rate": 1.5069022520878509e-05, + "loss": 0.9709, + "step": 23079 + }, + { + "epoch": 1.48, + "learning_rate": 1.5068807831855558e-05, + "loss": 0.9679, + "step": 23080 + }, + { + "epoch": 1.48, + "learning_rate": 1.5068593142832609e-05, + "loss": 1.0433, + "step": 23081 + }, + { + "epoch": 1.49, + "learning_rate": 1.5068378453809658e-05, + "loss": 0.9133, + "step": 23082 + }, + { + "epoch": 1.49, + "learning_rate": 1.5068163764786707e-05, + "loss": 0.9601, + "step": 23083 + }, + { + "epoch": 1.49, + "learning_rate": 1.5067949075763757e-05, + "loss": 0.9493, + "step": 23084 + }, + { + "epoch": 1.49, + "learning_rate": 1.5067734386740806e-05, + "loss": 0.9448, + "step": 23085 + }, + { + "epoch": 1.49, + "learning_rate": 1.5067519697717857e-05, + "loss": 0.9707, + "step": 23086 + }, + { + "epoch": 1.49, + "learning_rate": 1.5067305008694906e-05, + "loss": 0.9887, + "step": 23087 + }, + { + "epoch": 1.49, + "learning_rate": 1.5067090319671955e-05, + "loss": 0.8787, + "step": 23088 + }, + { + "epoch": 1.49, + "learning_rate": 1.5066875630649007e-05, + "loss": 0.9932, + "step": 23089 + }, + { + "epoch": 1.49, + "learning_rate": 1.5066660941626056e-05, + "loss": 0.9221, + "step": 23090 + }, + { + "epoch": 1.49, + "learning_rate": 1.5066446252603107e-05, + "loss": 0.9624, + "step": 23091 + }, + { + "epoch": 1.49, + "learning_rate": 1.5066231563580156e-05, + "loss": 0.9358, + "step": 23092 + }, + { + "epoch": 1.49, + "learning_rate": 1.5066016874557205e-05, + "loss": 0.9823, + "step": 23093 + }, + { + "epoch": 1.49, + "learning_rate": 1.5065802185534255e-05, + "loss": 0.9888, + "step": 23094 + }, + { + "epoch": 1.49, + "learning_rate": 1.5065587496511304e-05, + "loss": 1.1881, + "step": 23095 + }, + { + "epoch": 1.49, + "learning_rate": 1.5065372807488355e-05, + "loss": 1.1297, + "step": 23096 + }, + { + "epoch": 1.49, + "learning_rate": 1.5065158118465404e-05, + "loss": 1.035, + "step": 23097 + }, + { + "epoch": 1.49, + "learning_rate": 1.5064943429442453e-05, + "loss": 0.9911, + "step": 23098 + }, + { + "epoch": 1.49, + "learning_rate": 1.5064728740419504e-05, + "loss": 1.0188, + "step": 23099 + }, + { + "epoch": 1.49, + "learning_rate": 1.5064514051396553e-05, + "loss": 0.9514, + "step": 23100 + }, + { + "epoch": 1.49, + "learning_rate": 1.5064299362373602e-05, + "loss": 0.9992, + "step": 23101 + }, + { + "epoch": 1.49, + "learning_rate": 1.5064084673350652e-05, + "loss": 0.8969, + "step": 23102 + }, + { + "epoch": 1.49, + "learning_rate": 1.5063869984327703e-05, + "loss": 1.0048, + "step": 23103 + }, + { + "epoch": 1.49, + "learning_rate": 1.5063655295304754e-05, + "loss": 0.9457, + "step": 23104 + }, + { + "epoch": 1.49, + "learning_rate": 1.5063440606281803e-05, + "loss": 0.9494, + "step": 23105 + }, + { + "epoch": 1.49, + "learning_rate": 1.5063225917258853e-05, + "loss": 0.9793, + "step": 23106 + }, + { + "epoch": 1.49, + "learning_rate": 1.5063011228235902e-05, + "loss": 0.9837, + "step": 23107 + }, + { + "epoch": 1.49, + "learning_rate": 1.5062796539212951e-05, + "loss": 1.0381, + "step": 23108 + }, + { + "epoch": 1.49, + "learning_rate": 1.5062581850190002e-05, + "loss": 1.0127, + "step": 23109 + }, + { + "epoch": 1.49, + "learning_rate": 1.506236716116705e-05, + "loss": 0.891, + "step": 23110 + }, + { + "epoch": 1.49, + "learning_rate": 1.50621524721441e-05, + "loss": 0.949, + "step": 23111 + }, + { + "epoch": 1.49, + "learning_rate": 1.506193778312115e-05, + "loss": 0.9864, + "step": 23112 + }, + { + "epoch": 1.49, + "learning_rate": 1.50617230940982e-05, + "loss": 0.9109, + "step": 23113 + }, + { + "epoch": 1.49, + "learning_rate": 1.5061508405075248e-05, + "loss": 0.88, + "step": 23114 + }, + { + "epoch": 1.49, + "learning_rate": 1.5061293716052299e-05, + "loss": 0.9482, + "step": 23115 + }, + { + "epoch": 1.49, + "learning_rate": 1.5061079027029348e-05, + "loss": 0.9619, + "step": 23116 + }, + { + "epoch": 1.49, + "learning_rate": 1.5060864338006399e-05, + "loss": 0.9056, + "step": 23117 + }, + { + "epoch": 1.49, + "learning_rate": 1.506064964898345e-05, + "loss": 0.971, + "step": 23118 + }, + { + "epoch": 1.49, + "learning_rate": 1.50604349599605e-05, + "loss": 1.0344, + "step": 23119 + }, + { + "epoch": 1.49, + "learning_rate": 1.5060220270937549e-05, + "loss": 1.0212, + "step": 23120 + }, + { + "epoch": 1.49, + "learning_rate": 1.5060005581914598e-05, + "loss": 1.0215, + "step": 23121 + }, + { + "epoch": 1.49, + "learning_rate": 1.5059790892891648e-05, + "loss": 0.8673, + "step": 23122 + }, + { + "epoch": 1.49, + "learning_rate": 1.5059576203868697e-05, + "loss": 0.9946, + "step": 23123 + }, + { + "epoch": 1.49, + "learning_rate": 1.5059361514845746e-05, + "loss": 1.0254, + "step": 23124 + }, + { + "epoch": 1.49, + "learning_rate": 1.5059146825822797e-05, + "loss": 0.984, + "step": 23125 + }, + { + "epoch": 1.49, + "learning_rate": 1.5058932136799846e-05, + "loss": 0.9556, + "step": 23126 + }, + { + "epoch": 1.49, + "learning_rate": 1.5058717447776897e-05, + "loss": 0.9384, + "step": 23127 + }, + { + "epoch": 1.49, + "learning_rate": 1.5058502758753946e-05, + "loss": 0.9872, + "step": 23128 + }, + { + "epoch": 1.49, + "learning_rate": 1.5058288069730995e-05, + "loss": 0.9701, + "step": 23129 + }, + { + "epoch": 1.49, + "learning_rate": 1.5058073380708045e-05, + "loss": 0.9001, + "step": 23130 + }, + { + "epoch": 1.49, + "learning_rate": 1.5057858691685094e-05, + "loss": 0.9915, + "step": 23131 + }, + { + "epoch": 1.49, + "learning_rate": 1.5057644002662147e-05, + "loss": 1.0297, + "step": 23132 + }, + { + "epoch": 1.49, + "learning_rate": 1.5057429313639196e-05, + "loss": 0.9829, + "step": 23133 + }, + { + "epoch": 1.49, + "learning_rate": 1.5057214624616245e-05, + "loss": 0.9876, + "step": 23134 + }, + { + "epoch": 1.49, + "learning_rate": 1.5056999935593295e-05, + "loss": 0.961, + "step": 23135 + }, + { + "epoch": 1.49, + "learning_rate": 1.5056785246570344e-05, + "loss": 0.9053, + "step": 23136 + }, + { + "epoch": 1.49, + "learning_rate": 1.5056570557547395e-05, + "loss": 0.9532, + "step": 23137 + }, + { + "epoch": 1.49, + "learning_rate": 1.5056355868524444e-05, + "loss": 1.0074, + "step": 23138 + }, + { + "epoch": 1.49, + "learning_rate": 1.5056141179501493e-05, + "loss": 0.964, + "step": 23139 + }, + { + "epoch": 1.49, + "learning_rate": 1.5055926490478543e-05, + "loss": 1.0012, + "step": 23140 + }, + { + "epoch": 1.49, + "learning_rate": 1.5055711801455592e-05, + "loss": 0.9728, + "step": 23141 + }, + { + "epoch": 1.49, + "learning_rate": 1.5055497112432641e-05, + "loss": 0.9742, + "step": 23142 + }, + { + "epoch": 1.49, + "learning_rate": 1.5055282423409692e-05, + "loss": 1.0106, + "step": 23143 + }, + { + "epoch": 1.49, + "learning_rate": 1.5055067734386741e-05, + "loss": 0.9892, + "step": 23144 + }, + { + "epoch": 1.49, + "learning_rate": 1.5054853045363792e-05, + "loss": 1.0147, + "step": 23145 + }, + { + "epoch": 1.49, + "learning_rate": 1.505463835634084e-05, + "loss": 0.9445, + "step": 23146 + }, + { + "epoch": 1.49, + "learning_rate": 1.5054423667317893e-05, + "loss": 0.9376, + "step": 23147 + }, + { + "epoch": 1.49, + "learning_rate": 1.5054208978294942e-05, + "loss": 1.001, + "step": 23148 + }, + { + "epoch": 1.49, + "learning_rate": 1.5053994289271991e-05, + "loss": 0.9455, + "step": 23149 + }, + { + "epoch": 1.49, + "learning_rate": 1.5053779600249042e-05, + "loss": 1.0578, + "step": 23150 + }, + { + "epoch": 1.49, + "learning_rate": 1.505356491122609e-05, + "loss": 0.9796, + "step": 23151 + }, + { + "epoch": 1.49, + "learning_rate": 1.505335022220314e-05, + "loss": 0.9419, + "step": 23152 + }, + { + "epoch": 1.49, + "learning_rate": 1.505313553318019e-05, + "loss": 1.0075, + "step": 23153 + }, + { + "epoch": 1.49, + "learning_rate": 1.5052920844157239e-05, + "loss": 0.9562, + "step": 23154 + }, + { + "epoch": 1.49, + "learning_rate": 1.505270615513429e-05, + "loss": 0.9881, + "step": 23155 + }, + { + "epoch": 1.49, + "learning_rate": 1.5052491466111339e-05, + "loss": 1.0183, + "step": 23156 + }, + { + "epoch": 1.49, + "learning_rate": 1.5052276777088388e-05, + "loss": 1.0091, + "step": 23157 + }, + { + "epoch": 1.49, + "learning_rate": 1.5052062088065438e-05, + "loss": 0.9534, + "step": 23158 + }, + { + "epoch": 1.49, + "learning_rate": 1.5051847399042487e-05, + "loss": 0.9829, + "step": 23159 + }, + { + "epoch": 1.49, + "learning_rate": 1.5051632710019536e-05, + "loss": 0.96, + "step": 23160 + }, + { + "epoch": 1.49, + "learning_rate": 1.5051418020996589e-05, + "loss": 1.0008, + "step": 23161 + }, + { + "epoch": 1.49, + "learning_rate": 1.5051203331973638e-05, + "loss": 0.9733, + "step": 23162 + }, + { + "epoch": 1.49, + "learning_rate": 1.5050988642950688e-05, + "loss": 0.9696, + "step": 23163 + }, + { + "epoch": 1.49, + "learning_rate": 1.5050773953927737e-05, + "loss": 1.0583, + "step": 23164 + }, + { + "epoch": 1.49, + "learning_rate": 1.5050559264904788e-05, + "loss": 0.8992, + "step": 23165 + }, + { + "epoch": 1.49, + "learning_rate": 1.5050344575881837e-05, + "loss": 0.8676, + "step": 23166 + }, + { + "epoch": 1.49, + "learning_rate": 1.5050129886858886e-05, + "loss": 0.918, + "step": 23167 + }, + { + "epoch": 1.49, + "learning_rate": 1.5049915197835936e-05, + "loss": 0.977, + "step": 23168 + }, + { + "epoch": 1.49, + "learning_rate": 1.5049700508812985e-05, + "loss": 0.9515, + "step": 23169 + }, + { + "epoch": 1.49, + "learning_rate": 1.5049485819790034e-05, + "loss": 0.9901, + "step": 23170 + }, + { + "epoch": 1.49, + "learning_rate": 1.5049271130767085e-05, + "loss": 0.9634, + "step": 23171 + }, + { + "epoch": 1.49, + "learning_rate": 1.5049056441744134e-05, + "loss": 0.947, + "step": 23172 + }, + { + "epoch": 1.49, + "learning_rate": 1.5048841752721185e-05, + "loss": 0.976, + "step": 23173 + }, + { + "epoch": 1.49, + "learning_rate": 1.5048627063698234e-05, + "loss": 0.9142, + "step": 23174 + }, + { + "epoch": 1.49, + "learning_rate": 1.5048412374675283e-05, + "loss": 1.0089, + "step": 23175 + }, + { + "epoch": 1.49, + "learning_rate": 1.5048197685652335e-05, + "loss": 0.932, + "step": 23176 + }, + { + "epoch": 1.49, + "learning_rate": 1.5047982996629384e-05, + "loss": 0.9955, + "step": 23177 + }, + { + "epoch": 1.49, + "learning_rate": 1.5047768307606435e-05, + "loss": 0.9306, + "step": 23178 + }, + { + "epoch": 1.49, + "learning_rate": 1.5047553618583484e-05, + "loss": 0.9219, + "step": 23179 + }, + { + "epoch": 1.49, + "learning_rate": 1.5047338929560533e-05, + "loss": 0.9732, + "step": 23180 + }, + { + "epoch": 1.49, + "learning_rate": 1.5047124240537583e-05, + "loss": 0.9607, + "step": 23181 + }, + { + "epoch": 1.49, + "learning_rate": 1.5046909551514632e-05, + "loss": 1.0648, + "step": 23182 + }, + { + "epoch": 1.49, + "learning_rate": 1.5046694862491683e-05, + "loss": 0.9794, + "step": 23183 + }, + { + "epoch": 1.49, + "learning_rate": 1.5046480173468732e-05, + "loss": 0.9677, + "step": 23184 + }, + { + "epoch": 1.49, + "learning_rate": 1.504626548444578e-05, + "loss": 0.9426, + "step": 23185 + }, + { + "epoch": 1.49, + "learning_rate": 1.5046050795422831e-05, + "loss": 0.9983, + "step": 23186 + }, + { + "epoch": 1.49, + "learning_rate": 1.504583610639988e-05, + "loss": 0.9615, + "step": 23187 + }, + { + "epoch": 1.49, + "learning_rate": 1.504562141737693e-05, + "loss": 0.9832, + "step": 23188 + }, + { + "epoch": 1.49, + "learning_rate": 1.504540672835398e-05, + "loss": 0.9259, + "step": 23189 + }, + { + "epoch": 1.49, + "learning_rate": 1.504519203933103e-05, + "loss": 0.9355, + "step": 23190 + }, + { + "epoch": 1.49, + "learning_rate": 1.5044977350308081e-05, + "loss": 0.9802, + "step": 23191 + }, + { + "epoch": 1.49, + "learning_rate": 1.504476266128513e-05, + "loss": 0.8855, + "step": 23192 + }, + { + "epoch": 1.49, + "learning_rate": 1.5044547972262181e-05, + "loss": 1.0112, + "step": 23193 + }, + { + "epoch": 1.49, + "learning_rate": 1.504433328323923e-05, + "loss": 1.0585, + "step": 23194 + }, + { + "epoch": 1.49, + "learning_rate": 1.5044118594216279e-05, + "loss": 0.9787, + "step": 23195 + }, + { + "epoch": 1.49, + "learning_rate": 1.504390390519333e-05, + "loss": 0.9154, + "step": 23196 + }, + { + "epoch": 1.49, + "learning_rate": 1.5043689216170378e-05, + "loss": 0.9576, + "step": 23197 + }, + { + "epoch": 1.49, + "learning_rate": 1.5043474527147427e-05, + "loss": 0.9954, + "step": 23198 + }, + { + "epoch": 1.49, + "learning_rate": 1.5043259838124478e-05, + "loss": 0.9714, + "step": 23199 + }, + { + "epoch": 1.49, + "learning_rate": 1.5043045149101527e-05, + "loss": 0.9398, + "step": 23200 + }, + { + "epoch": 1.49, + "learning_rate": 1.5042830460078576e-05, + "loss": 0.9743, + "step": 23201 + }, + { + "epoch": 1.49, + "learning_rate": 1.5042615771055627e-05, + "loss": 0.9376, + "step": 23202 + }, + { + "epoch": 1.49, + "learning_rate": 1.5042401082032676e-05, + "loss": 1.0054, + "step": 23203 + }, + { + "epoch": 1.49, + "learning_rate": 1.5042186393009726e-05, + "loss": 1.0079, + "step": 23204 + }, + { + "epoch": 1.49, + "learning_rate": 1.5041971703986777e-05, + "loss": 0.9149, + "step": 23205 + }, + { + "epoch": 1.49, + "learning_rate": 1.5041757014963828e-05, + "loss": 0.9573, + "step": 23206 + }, + { + "epoch": 1.49, + "learning_rate": 1.5041542325940877e-05, + "loss": 1.0187, + "step": 23207 + }, + { + "epoch": 1.49, + "learning_rate": 1.5041327636917926e-05, + "loss": 1.0106, + "step": 23208 + }, + { + "epoch": 1.49, + "learning_rate": 1.5041112947894976e-05, + "loss": 0.9595, + "step": 23209 + }, + { + "epoch": 1.49, + "learning_rate": 1.5040898258872025e-05, + "loss": 0.9893, + "step": 23210 + }, + { + "epoch": 1.49, + "learning_rate": 1.5040683569849074e-05, + "loss": 1.0176, + "step": 23211 + }, + { + "epoch": 1.49, + "learning_rate": 1.5040468880826125e-05, + "loss": 0.9491, + "step": 23212 + }, + { + "epoch": 1.49, + "learning_rate": 1.5040254191803174e-05, + "loss": 0.983, + "step": 23213 + }, + { + "epoch": 1.49, + "learning_rate": 1.5040039502780224e-05, + "loss": 0.9548, + "step": 23214 + }, + { + "epoch": 1.49, + "learning_rate": 1.5039824813757273e-05, + "loss": 0.9274, + "step": 23215 + }, + { + "epoch": 1.49, + "learning_rate": 1.5039610124734322e-05, + "loss": 1.0103, + "step": 23216 + }, + { + "epoch": 1.49, + "learning_rate": 1.5039395435711373e-05, + "loss": 0.9599, + "step": 23217 + }, + { + "epoch": 1.49, + "learning_rate": 1.5039180746688422e-05, + "loss": 0.9325, + "step": 23218 + }, + { + "epoch": 1.49, + "learning_rate": 1.5038966057665471e-05, + "loss": 0.917, + "step": 23219 + }, + { + "epoch": 1.49, + "learning_rate": 1.5038751368642523e-05, + "loss": 1.039, + "step": 23220 + }, + { + "epoch": 1.49, + "learning_rate": 1.5038536679619572e-05, + "loss": 0.9814, + "step": 23221 + }, + { + "epoch": 1.49, + "learning_rate": 1.5038321990596623e-05, + "loss": 0.9119, + "step": 23222 + }, + { + "epoch": 1.49, + "learning_rate": 1.5038107301573672e-05, + "loss": 1.0458, + "step": 23223 + }, + { + "epoch": 1.49, + "learning_rate": 1.5037892612550723e-05, + "loss": 1.0201, + "step": 23224 + }, + { + "epoch": 1.49, + "learning_rate": 1.5037677923527772e-05, + "loss": 0.9971, + "step": 23225 + }, + { + "epoch": 1.49, + "learning_rate": 1.503746323450482e-05, + "loss": 1.0221, + "step": 23226 + }, + { + "epoch": 1.49, + "learning_rate": 1.5037248545481871e-05, + "loss": 0.9907, + "step": 23227 + }, + { + "epoch": 1.49, + "learning_rate": 1.503703385645892e-05, + "loss": 0.9057, + "step": 23228 + }, + { + "epoch": 1.49, + "learning_rate": 1.5036819167435969e-05, + "loss": 1.0314, + "step": 23229 + }, + { + "epoch": 1.49, + "learning_rate": 1.503660447841302e-05, + "loss": 1.0107, + "step": 23230 + }, + { + "epoch": 1.49, + "learning_rate": 1.5036389789390069e-05, + "loss": 0.9773, + "step": 23231 + }, + { + "epoch": 1.49, + "learning_rate": 1.503617510036712e-05, + "loss": 0.9965, + "step": 23232 + }, + { + "epoch": 1.49, + "learning_rate": 1.5035960411344168e-05, + "loss": 0.9318, + "step": 23233 + }, + { + "epoch": 1.49, + "learning_rate": 1.503574572232122e-05, + "loss": 0.9202, + "step": 23234 + }, + { + "epoch": 1.49, + "learning_rate": 1.503553103329827e-05, + "loss": 0.9437, + "step": 23235 + }, + { + "epoch": 1.49, + "learning_rate": 1.5035316344275319e-05, + "loss": 0.9597, + "step": 23236 + }, + { + "epoch": 1.5, + "learning_rate": 1.503510165525237e-05, + "loss": 0.996, + "step": 23237 + }, + { + "epoch": 1.5, + "learning_rate": 1.5034886966229418e-05, + "loss": 0.9775, + "step": 23238 + }, + { + "epoch": 1.5, + "learning_rate": 1.5034672277206467e-05, + "loss": 0.9506, + "step": 23239 + }, + { + "epoch": 1.5, + "learning_rate": 1.5034457588183518e-05, + "loss": 1.0415, + "step": 23240 + }, + { + "epoch": 1.5, + "learning_rate": 1.5034242899160567e-05, + "loss": 1.0125, + "step": 23241 + }, + { + "epoch": 1.5, + "learning_rate": 1.5034028210137617e-05, + "loss": 1.058, + "step": 23242 + }, + { + "epoch": 1.5, + "learning_rate": 1.5033813521114666e-05, + "loss": 1.0529, + "step": 23243 + }, + { + "epoch": 1.5, + "learning_rate": 1.5033598832091715e-05, + "loss": 0.9862, + "step": 23244 + }, + { + "epoch": 1.5, + "learning_rate": 1.5033384143068766e-05, + "loss": 0.9441, + "step": 23245 + }, + { + "epoch": 1.5, + "learning_rate": 1.5033169454045815e-05, + "loss": 0.9898, + "step": 23246 + }, + { + "epoch": 1.5, + "learning_rate": 1.5032954765022864e-05, + "loss": 0.9729, + "step": 23247 + }, + { + "epoch": 1.5, + "learning_rate": 1.5032740075999915e-05, + "loss": 1.035, + "step": 23248 + }, + { + "epoch": 1.5, + "learning_rate": 1.5032525386976965e-05, + "loss": 0.954, + "step": 23249 + }, + { + "epoch": 1.5, + "learning_rate": 1.5032310697954016e-05, + "loss": 0.9551, + "step": 23250 + }, + { + "epoch": 1.5, + "learning_rate": 1.5032096008931065e-05, + "loss": 0.8976, + "step": 23251 + }, + { + "epoch": 1.5, + "learning_rate": 1.5031881319908116e-05, + "loss": 0.9543, + "step": 23252 + }, + { + "epoch": 1.5, + "learning_rate": 1.5031666630885165e-05, + "loss": 0.9974, + "step": 23253 + }, + { + "epoch": 1.5, + "learning_rate": 1.5031451941862214e-05, + "loss": 0.9422, + "step": 23254 + }, + { + "epoch": 1.5, + "learning_rate": 1.5031237252839264e-05, + "loss": 0.9389, + "step": 23255 + }, + { + "epoch": 1.5, + "learning_rate": 1.5031022563816313e-05, + "loss": 1.0737, + "step": 23256 + }, + { + "epoch": 1.5, + "learning_rate": 1.5030807874793362e-05, + "loss": 0.9146, + "step": 23257 + }, + { + "epoch": 1.5, + "learning_rate": 1.5030593185770413e-05, + "loss": 0.989, + "step": 23258 + }, + { + "epoch": 1.5, + "learning_rate": 1.5030378496747462e-05, + "loss": 0.939, + "step": 23259 + }, + { + "epoch": 1.5, + "learning_rate": 1.5030163807724512e-05, + "loss": 0.9668, + "step": 23260 + }, + { + "epoch": 1.5, + "learning_rate": 1.5029949118701561e-05, + "loss": 0.9833, + "step": 23261 + }, + { + "epoch": 1.5, + "learning_rate": 1.502973442967861e-05, + "loss": 1.0037, + "step": 23262 + }, + { + "epoch": 1.5, + "learning_rate": 1.5029519740655663e-05, + "loss": 0.9202, + "step": 23263 + }, + { + "epoch": 1.5, + "learning_rate": 1.5029305051632712e-05, + "loss": 0.9356, + "step": 23264 + }, + { + "epoch": 1.5, + "learning_rate": 1.5029090362609762e-05, + "loss": 0.9876, + "step": 23265 + }, + { + "epoch": 1.5, + "learning_rate": 1.5028875673586811e-05, + "loss": 0.992, + "step": 23266 + }, + { + "epoch": 1.5, + "learning_rate": 1.502866098456386e-05, + "loss": 0.8705, + "step": 23267 + }, + { + "epoch": 1.5, + "learning_rate": 1.5028446295540911e-05, + "loss": 0.918, + "step": 23268 + }, + { + "epoch": 1.5, + "learning_rate": 1.502823160651796e-05, + "loss": 0.9943, + "step": 23269 + }, + { + "epoch": 1.5, + "learning_rate": 1.502801691749501e-05, + "loss": 0.9479, + "step": 23270 + }, + { + "epoch": 1.5, + "learning_rate": 1.502780222847206e-05, + "loss": 0.9445, + "step": 23271 + }, + { + "epoch": 1.5, + "learning_rate": 1.5027587539449108e-05, + "loss": 1.0059, + "step": 23272 + }, + { + "epoch": 1.5, + "learning_rate": 1.5027372850426159e-05, + "loss": 0.9674, + "step": 23273 + }, + { + "epoch": 1.5, + "learning_rate": 1.5027158161403208e-05, + "loss": 0.8892, + "step": 23274 + }, + { + "epoch": 1.5, + "learning_rate": 1.5026943472380257e-05, + "loss": 1.0235, + "step": 23275 + }, + { + "epoch": 1.5, + "learning_rate": 1.5026728783357308e-05, + "loss": 0.9496, + "step": 23276 + }, + { + "epoch": 1.5, + "learning_rate": 1.5026514094334357e-05, + "loss": 0.9535, + "step": 23277 + }, + { + "epoch": 1.5, + "learning_rate": 1.5026299405311409e-05, + "loss": 1.0491, + "step": 23278 + }, + { + "epoch": 1.5, + "learning_rate": 1.5026084716288458e-05, + "loss": 0.9768, + "step": 23279 + }, + { + "epoch": 1.5, + "learning_rate": 1.5025870027265509e-05, + "loss": 0.9535, + "step": 23280 + }, + { + "epoch": 1.5, + "learning_rate": 1.5025655338242558e-05, + "loss": 0.9561, + "step": 23281 + }, + { + "epoch": 1.5, + "learning_rate": 1.5025440649219607e-05, + "loss": 1.0032, + "step": 23282 + }, + { + "epoch": 1.5, + "learning_rate": 1.5025225960196657e-05, + "loss": 1.0423, + "step": 23283 + }, + { + "epoch": 1.5, + "learning_rate": 1.5025011271173706e-05, + "loss": 0.9732, + "step": 23284 + }, + { + "epoch": 1.5, + "learning_rate": 1.5024796582150755e-05, + "loss": 1.0004, + "step": 23285 + }, + { + "epoch": 1.5, + "learning_rate": 1.5024581893127806e-05, + "loss": 0.959, + "step": 23286 + }, + { + "epoch": 1.5, + "learning_rate": 1.5024367204104855e-05, + "loss": 1.0236, + "step": 23287 + }, + { + "epoch": 1.5, + "learning_rate": 1.5024152515081904e-05, + "loss": 0.978, + "step": 23288 + }, + { + "epoch": 1.5, + "learning_rate": 1.5023937826058954e-05, + "loss": 0.9809, + "step": 23289 + }, + { + "epoch": 1.5, + "learning_rate": 1.5023723137036003e-05, + "loss": 1.0188, + "step": 23290 + }, + { + "epoch": 1.5, + "learning_rate": 1.5023508448013054e-05, + "loss": 0.9281, + "step": 23291 + }, + { + "epoch": 1.5, + "learning_rate": 1.5023293758990105e-05, + "loss": 0.9867, + "step": 23292 + }, + { + "epoch": 1.5, + "learning_rate": 1.5023079069967155e-05, + "loss": 0.9145, + "step": 23293 + }, + { + "epoch": 1.5, + "learning_rate": 1.5022864380944204e-05, + "loss": 1.0283, + "step": 23294 + }, + { + "epoch": 1.5, + "learning_rate": 1.5022649691921253e-05, + "loss": 1.0279, + "step": 23295 + }, + { + "epoch": 1.5, + "learning_rate": 1.5022435002898304e-05, + "loss": 0.9904, + "step": 23296 + }, + { + "epoch": 1.5, + "learning_rate": 1.5022220313875353e-05, + "loss": 1.0368, + "step": 23297 + }, + { + "epoch": 1.5, + "learning_rate": 1.5022005624852402e-05, + "loss": 0.9657, + "step": 23298 + }, + { + "epoch": 1.5, + "learning_rate": 1.5021790935829453e-05, + "loss": 1.0204, + "step": 23299 + }, + { + "epoch": 1.5, + "learning_rate": 1.5021576246806501e-05, + "loss": 0.9226, + "step": 23300 + }, + { + "epoch": 1.5, + "learning_rate": 1.5021361557783552e-05, + "loss": 0.9861, + "step": 23301 + }, + { + "epoch": 1.5, + "learning_rate": 1.5021146868760601e-05, + "loss": 0.9362, + "step": 23302 + }, + { + "epoch": 1.5, + "learning_rate": 1.502093217973765e-05, + "loss": 0.9041, + "step": 23303 + }, + { + "epoch": 1.5, + "learning_rate": 1.50207174907147e-05, + "loss": 0.9349, + "step": 23304 + }, + { + "epoch": 1.5, + "learning_rate": 1.502050280169175e-05, + "loss": 0.917, + "step": 23305 + }, + { + "epoch": 1.5, + "learning_rate": 1.5020288112668799e-05, + "loss": 0.9298, + "step": 23306 + }, + { + "epoch": 1.5, + "learning_rate": 1.5020073423645851e-05, + "loss": 0.9922, + "step": 23307 + }, + { + "epoch": 1.5, + "learning_rate": 1.50198587346229e-05, + "loss": 0.9738, + "step": 23308 + }, + { + "epoch": 1.5, + "learning_rate": 1.501964404559995e-05, + "loss": 1.0044, + "step": 23309 + }, + { + "epoch": 1.5, + "learning_rate": 1.5019429356577e-05, + "loss": 0.9404, + "step": 23310 + }, + { + "epoch": 1.5, + "learning_rate": 1.501921466755405e-05, + "loss": 1.0564, + "step": 23311 + }, + { + "epoch": 1.5, + "learning_rate": 1.50189999785311e-05, + "loss": 1.0256, + "step": 23312 + }, + { + "epoch": 1.5, + "learning_rate": 1.5018785289508148e-05, + "loss": 0.9682, + "step": 23313 + }, + { + "epoch": 1.5, + "learning_rate": 1.5018570600485199e-05, + "loss": 1.0365, + "step": 23314 + }, + { + "epoch": 1.5, + "learning_rate": 1.5018355911462248e-05, + "loss": 0.9699, + "step": 23315 + }, + { + "epoch": 1.5, + "learning_rate": 1.5018141222439297e-05, + "loss": 1.0236, + "step": 23316 + }, + { + "epoch": 1.5, + "learning_rate": 1.5017926533416347e-05, + "loss": 0.928, + "step": 23317 + }, + { + "epoch": 1.5, + "learning_rate": 1.5017711844393396e-05, + "loss": 1.0319, + "step": 23318 + }, + { + "epoch": 1.5, + "learning_rate": 1.5017497155370447e-05, + "loss": 1.0256, + "step": 23319 + }, + { + "epoch": 1.5, + "learning_rate": 1.5017282466347496e-05, + "loss": 0.9887, + "step": 23320 + }, + { + "epoch": 1.5, + "learning_rate": 1.5017067777324548e-05, + "loss": 1.0802, + "step": 23321 + }, + { + "epoch": 1.5, + "learning_rate": 1.5016853088301597e-05, + "loss": 0.9738, + "step": 23322 + }, + { + "epoch": 1.5, + "learning_rate": 1.5016638399278646e-05, + "loss": 0.9154, + "step": 23323 + }, + { + "epoch": 1.5, + "learning_rate": 1.5016423710255697e-05, + "loss": 0.972, + "step": 23324 + }, + { + "epoch": 1.5, + "learning_rate": 1.5016209021232746e-05, + "loss": 0.9178, + "step": 23325 + }, + { + "epoch": 1.5, + "learning_rate": 1.5015994332209795e-05, + "loss": 1.005, + "step": 23326 + }, + { + "epoch": 1.5, + "learning_rate": 1.5015779643186846e-05, + "loss": 0.9582, + "step": 23327 + }, + { + "epoch": 1.5, + "learning_rate": 1.5015564954163895e-05, + "loss": 0.9988, + "step": 23328 + }, + { + "epoch": 1.5, + "learning_rate": 1.5015350265140945e-05, + "loss": 0.9807, + "step": 23329 + }, + { + "epoch": 1.5, + "learning_rate": 1.5015135576117994e-05, + "loss": 0.9494, + "step": 23330 + }, + { + "epoch": 1.5, + "learning_rate": 1.5014920887095043e-05, + "loss": 0.9618, + "step": 23331 + }, + { + "epoch": 1.5, + "learning_rate": 1.5014706198072094e-05, + "loss": 0.9574, + "step": 23332 + }, + { + "epoch": 1.5, + "learning_rate": 1.5014491509049143e-05, + "loss": 1.0601, + "step": 23333 + }, + { + "epoch": 1.5, + "learning_rate": 1.5014276820026192e-05, + "loss": 1.0015, + "step": 23334 + }, + { + "epoch": 1.5, + "learning_rate": 1.5014062131003242e-05, + "loss": 0.9226, + "step": 23335 + }, + { + "epoch": 1.5, + "learning_rate": 1.5013847441980293e-05, + "loss": 1.0587, + "step": 23336 + }, + { + "epoch": 1.5, + "learning_rate": 1.5013632752957344e-05, + "loss": 0.96, + "step": 23337 + }, + { + "epoch": 1.5, + "learning_rate": 1.5013418063934393e-05, + "loss": 1.065, + "step": 23338 + }, + { + "epoch": 1.5, + "learning_rate": 1.5013203374911443e-05, + "loss": 0.9814, + "step": 23339 + }, + { + "epoch": 1.5, + "learning_rate": 1.5012988685888492e-05, + "loss": 0.9128, + "step": 23340 + }, + { + "epoch": 1.5, + "learning_rate": 1.5012773996865541e-05, + "loss": 0.9611, + "step": 23341 + }, + { + "epoch": 1.5, + "learning_rate": 1.5012559307842592e-05, + "loss": 1.0235, + "step": 23342 + }, + { + "epoch": 1.5, + "learning_rate": 1.5012344618819641e-05, + "loss": 0.9776, + "step": 23343 + }, + { + "epoch": 1.5, + "learning_rate": 1.501212992979669e-05, + "loss": 0.9697, + "step": 23344 + }, + { + "epoch": 1.5, + "learning_rate": 1.501191524077374e-05, + "loss": 1.0354, + "step": 23345 + }, + { + "epoch": 1.5, + "learning_rate": 1.501170055175079e-05, + "loss": 0.9389, + "step": 23346 + }, + { + "epoch": 1.5, + "learning_rate": 1.501148586272784e-05, + "loss": 1.0293, + "step": 23347 + }, + { + "epoch": 1.5, + "learning_rate": 1.5011271173704889e-05, + "loss": 1.025, + "step": 23348 + }, + { + "epoch": 1.5, + "learning_rate": 1.5011056484681938e-05, + "loss": 1.0027, + "step": 23349 + }, + { + "epoch": 1.5, + "learning_rate": 1.501084179565899e-05, + "loss": 0.9327, + "step": 23350 + }, + { + "epoch": 1.5, + "learning_rate": 1.501062710663604e-05, + "loss": 0.9236, + "step": 23351 + }, + { + "epoch": 1.5, + "learning_rate": 1.501041241761309e-05, + "loss": 0.9902, + "step": 23352 + }, + { + "epoch": 1.5, + "learning_rate": 1.5010197728590139e-05, + "loss": 0.9874, + "step": 23353 + }, + { + "epoch": 1.5, + "learning_rate": 1.5009983039567188e-05, + "loss": 0.99, + "step": 23354 + }, + { + "epoch": 1.5, + "learning_rate": 1.5009768350544239e-05, + "loss": 0.951, + "step": 23355 + }, + { + "epoch": 1.5, + "learning_rate": 1.5009553661521288e-05, + "loss": 1.02, + "step": 23356 + }, + { + "epoch": 1.5, + "learning_rate": 1.5009338972498338e-05, + "loss": 0.9898, + "step": 23357 + }, + { + "epoch": 1.5, + "learning_rate": 1.5009124283475387e-05, + "loss": 0.9594, + "step": 23358 + }, + { + "epoch": 1.5, + "learning_rate": 1.5008909594452436e-05, + "loss": 0.9323, + "step": 23359 + }, + { + "epoch": 1.5, + "learning_rate": 1.5008694905429487e-05, + "loss": 0.9828, + "step": 23360 + }, + { + "epoch": 1.5, + "learning_rate": 1.5008480216406536e-05, + "loss": 0.9242, + "step": 23361 + }, + { + "epoch": 1.5, + "learning_rate": 1.5008265527383585e-05, + "loss": 0.9728, + "step": 23362 + }, + { + "epoch": 1.5, + "learning_rate": 1.5008050838360635e-05, + "loss": 0.9428, + "step": 23363 + }, + { + "epoch": 1.5, + "learning_rate": 1.5007836149337684e-05, + "loss": 0.9906, + "step": 23364 + }, + { + "epoch": 1.5, + "learning_rate": 1.5007621460314737e-05, + "loss": 0.9772, + "step": 23365 + }, + { + "epoch": 1.5, + "learning_rate": 1.5007406771291786e-05, + "loss": 0.9043, + "step": 23366 + }, + { + "epoch": 1.5, + "learning_rate": 1.5007192082268836e-05, + "loss": 0.9948, + "step": 23367 + }, + { + "epoch": 1.5, + "learning_rate": 1.5006977393245885e-05, + "loss": 1.0192, + "step": 23368 + }, + { + "epoch": 1.5, + "learning_rate": 1.5006762704222934e-05, + "loss": 0.9732, + "step": 23369 + }, + { + "epoch": 1.5, + "learning_rate": 1.5006548015199985e-05, + "loss": 0.8774, + "step": 23370 + }, + { + "epoch": 1.5, + "learning_rate": 1.5006333326177034e-05, + "loss": 0.9564, + "step": 23371 + }, + { + "epoch": 1.5, + "learning_rate": 1.5006118637154083e-05, + "loss": 0.9779, + "step": 23372 + }, + { + "epoch": 1.5, + "learning_rate": 1.5005903948131134e-05, + "loss": 0.9797, + "step": 23373 + }, + { + "epoch": 1.5, + "learning_rate": 1.5005689259108182e-05, + "loss": 0.9954, + "step": 23374 + }, + { + "epoch": 1.5, + "learning_rate": 1.5005474570085231e-05, + "loss": 1.096, + "step": 23375 + }, + { + "epoch": 1.5, + "learning_rate": 1.5005259881062282e-05, + "loss": 0.9215, + "step": 23376 + }, + { + "epoch": 1.5, + "learning_rate": 1.5005045192039331e-05, + "loss": 0.9748, + "step": 23377 + }, + { + "epoch": 1.5, + "learning_rate": 1.5004830503016382e-05, + "loss": 1.0379, + "step": 23378 + }, + { + "epoch": 1.5, + "learning_rate": 1.5004615813993432e-05, + "loss": 1.0414, + "step": 23379 + }, + { + "epoch": 1.5, + "learning_rate": 1.5004401124970483e-05, + "loss": 0.9852, + "step": 23380 + }, + { + "epoch": 1.5, + "learning_rate": 1.5004186435947532e-05, + "loss": 1.0079, + "step": 23381 + }, + { + "epoch": 1.5, + "learning_rate": 1.5003971746924581e-05, + "loss": 0.895, + "step": 23382 + }, + { + "epoch": 1.5, + "learning_rate": 1.5003757057901632e-05, + "loss": 0.9503, + "step": 23383 + }, + { + "epoch": 1.5, + "learning_rate": 1.500354236887868e-05, + "loss": 0.9975, + "step": 23384 + }, + { + "epoch": 1.5, + "learning_rate": 1.500332767985573e-05, + "loss": 0.964, + "step": 23385 + }, + { + "epoch": 1.5, + "learning_rate": 1.500311299083278e-05, + "loss": 0.93, + "step": 23386 + }, + { + "epoch": 1.5, + "learning_rate": 1.500289830180983e-05, + "loss": 0.9644, + "step": 23387 + }, + { + "epoch": 1.5, + "learning_rate": 1.500268361278688e-05, + "loss": 0.9553, + "step": 23388 + }, + { + "epoch": 1.5, + "learning_rate": 1.5002468923763929e-05, + "loss": 1.0164, + "step": 23389 + }, + { + "epoch": 1.5, + "learning_rate": 1.5002254234740978e-05, + "loss": 1.0216, + "step": 23390 + }, + { + "epoch": 1.5, + "learning_rate": 1.5002039545718028e-05, + "loss": 0.9774, + "step": 23391 + }, + { + "epoch": 1.5, + "learning_rate": 1.5001824856695077e-05, + "loss": 1.001, + "step": 23392 + }, + { + "epoch": 1.51, + "learning_rate": 1.5001610167672126e-05, + "loss": 1.0522, + "step": 23393 + }, + { + "epoch": 1.51, + "learning_rate": 1.5001395478649179e-05, + "loss": 1.0073, + "step": 23394 + }, + { + "epoch": 1.51, + "learning_rate": 1.5001180789626228e-05, + "loss": 0.9794, + "step": 23395 + }, + { + "epoch": 1.51, + "learning_rate": 1.5000966100603278e-05, + "loss": 0.948, + "step": 23396 + }, + { + "epoch": 1.51, + "learning_rate": 1.5000751411580327e-05, + "loss": 0.9687, + "step": 23397 + }, + { + "epoch": 1.51, + "learning_rate": 1.5000536722557378e-05, + "loss": 0.9621, + "step": 23398 + }, + { + "epoch": 1.51, + "learning_rate": 1.5000322033534427e-05, + "loss": 0.9788, + "step": 23399 + }, + { + "epoch": 1.51, + "learning_rate": 1.5000107344511476e-05, + "loss": 1.0029, + "step": 23400 + }, + { + "epoch": 1.51, + "learning_rate": 1.4999892655488527e-05, + "loss": 1.0052, + "step": 23401 + }, + { + "epoch": 1.51, + "learning_rate": 1.4999677966465576e-05, + "loss": 0.9881, + "step": 23402 + }, + { + "epoch": 1.51, + "learning_rate": 1.4999463277442624e-05, + "loss": 1.0729, + "step": 23403 + }, + { + "epoch": 1.51, + "learning_rate": 1.4999248588419675e-05, + "loss": 0.9909, + "step": 23404 + }, + { + "epoch": 1.51, + "learning_rate": 1.4999033899396724e-05, + "loss": 1.0124, + "step": 23405 + }, + { + "epoch": 1.51, + "learning_rate": 1.4998819210373775e-05, + "loss": 0.8916, + "step": 23406 + }, + { + "epoch": 1.51, + "learning_rate": 1.4998604521350824e-05, + "loss": 0.9304, + "step": 23407 + }, + { + "epoch": 1.51, + "learning_rate": 1.4998389832327876e-05, + "loss": 0.9883, + "step": 23408 + }, + { + "epoch": 1.51, + "learning_rate": 1.4998175143304925e-05, + "loss": 0.9255, + "step": 23409 + }, + { + "epoch": 1.51, + "learning_rate": 1.4997960454281974e-05, + "loss": 0.9902, + "step": 23410 + }, + { + "epoch": 1.51, + "learning_rate": 1.4997745765259025e-05, + "loss": 0.9162, + "step": 23411 + }, + { + "epoch": 1.51, + "learning_rate": 1.4997531076236074e-05, + "loss": 0.9069, + "step": 23412 + }, + { + "epoch": 1.51, + "learning_rate": 1.4997316387213123e-05, + "loss": 0.9295, + "step": 23413 + }, + { + "epoch": 1.51, + "learning_rate": 1.4997101698190173e-05, + "loss": 1.0266, + "step": 23414 + }, + { + "epoch": 1.51, + "learning_rate": 1.4996887009167222e-05, + "loss": 0.9593, + "step": 23415 + }, + { + "epoch": 1.51, + "learning_rate": 1.4996672320144273e-05, + "loss": 0.9715, + "step": 23416 + }, + { + "epoch": 1.51, + "learning_rate": 1.4996457631121322e-05, + "loss": 1.055, + "step": 23417 + }, + { + "epoch": 1.51, + "learning_rate": 1.499624294209837e-05, + "loss": 0.8415, + "step": 23418 + }, + { + "epoch": 1.51, + "learning_rate": 1.4996028253075421e-05, + "loss": 0.9672, + "step": 23419 + }, + { + "epoch": 1.51, + "learning_rate": 1.499581356405247e-05, + "loss": 0.9378, + "step": 23420 + }, + { + "epoch": 1.51, + "learning_rate": 1.499559887502952e-05, + "loss": 0.9524, + "step": 23421 + }, + { + "epoch": 1.51, + "learning_rate": 1.499538418600657e-05, + "loss": 0.9898, + "step": 23422 + }, + { + "epoch": 1.51, + "learning_rate": 1.499516949698362e-05, + "loss": 0.9627, + "step": 23423 + }, + { + "epoch": 1.51, + "learning_rate": 1.4994954807960671e-05, + "loss": 0.9787, + "step": 23424 + }, + { + "epoch": 1.51, + "learning_rate": 1.499474011893772e-05, + "loss": 0.9616, + "step": 23425 + }, + { + "epoch": 1.51, + "learning_rate": 1.4994525429914771e-05, + "loss": 0.9371, + "step": 23426 + }, + { + "epoch": 1.51, + "learning_rate": 1.499431074089182e-05, + "loss": 0.9883, + "step": 23427 + }, + { + "epoch": 1.51, + "learning_rate": 1.4994096051868869e-05, + "loss": 0.9619, + "step": 23428 + }, + { + "epoch": 1.51, + "learning_rate": 1.499388136284592e-05, + "loss": 0.8944, + "step": 23429 + }, + { + "epoch": 1.51, + "learning_rate": 1.4993666673822969e-05, + "loss": 1.0207, + "step": 23430 + }, + { + "epoch": 1.51, + "learning_rate": 1.4993451984800018e-05, + "loss": 0.9079, + "step": 23431 + }, + { + "epoch": 1.51, + "learning_rate": 1.4993237295777068e-05, + "loss": 1.0136, + "step": 23432 + }, + { + "epoch": 1.51, + "learning_rate": 1.4993022606754117e-05, + "loss": 1.0448, + "step": 23433 + }, + { + "epoch": 1.51, + "learning_rate": 1.4992807917731166e-05, + "loss": 1.0233, + "step": 23434 + }, + { + "epoch": 1.51, + "learning_rate": 1.4992593228708217e-05, + "loss": 1.053, + "step": 23435 + }, + { + "epoch": 1.51, + "learning_rate": 1.4992378539685266e-05, + "loss": 0.9921, + "step": 23436 + }, + { + "epoch": 1.51, + "learning_rate": 1.4992163850662318e-05, + "loss": 0.9459, + "step": 23437 + }, + { + "epoch": 1.51, + "learning_rate": 1.4991949161639367e-05, + "loss": 0.9608, + "step": 23438 + }, + { + "epoch": 1.51, + "learning_rate": 1.4991734472616418e-05, + "loss": 0.9779, + "step": 23439 + }, + { + "epoch": 1.51, + "learning_rate": 1.4991519783593467e-05, + "loss": 0.9572, + "step": 23440 + }, + { + "epoch": 1.51, + "learning_rate": 1.4991305094570516e-05, + "loss": 1.0584, + "step": 23441 + }, + { + "epoch": 1.51, + "learning_rate": 1.4991090405547566e-05, + "loss": 0.982, + "step": 23442 + }, + { + "epoch": 1.51, + "learning_rate": 1.4990875716524615e-05, + "loss": 0.9875, + "step": 23443 + }, + { + "epoch": 1.51, + "learning_rate": 1.4990661027501664e-05, + "loss": 0.9192, + "step": 23444 + }, + { + "epoch": 1.51, + "learning_rate": 1.4990446338478715e-05, + "loss": 1.0279, + "step": 23445 + }, + { + "epoch": 1.51, + "learning_rate": 1.4990231649455764e-05, + "loss": 0.972, + "step": 23446 + }, + { + "epoch": 1.51, + "learning_rate": 1.4990016960432815e-05, + "loss": 0.9447, + "step": 23447 + }, + { + "epoch": 1.51, + "learning_rate": 1.4989802271409863e-05, + "loss": 1.0108, + "step": 23448 + }, + { + "epoch": 1.51, + "learning_rate": 1.4989587582386912e-05, + "loss": 0.9661, + "step": 23449 + }, + { + "epoch": 1.51, + "learning_rate": 1.4989372893363963e-05, + "loss": 0.9948, + "step": 23450 + }, + { + "epoch": 1.51, + "learning_rate": 1.4989158204341012e-05, + "loss": 0.9552, + "step": 23451 + }, + { + "epoch": 1.51, + "learning_rate": 1.4988943515318064e-05, + "loss": 0.9782, + "step": 23452 + }, + { + "epoch": 1.51, + "learning_rate": 1.4988728826295113e-05, + "loss": 0.9251, + "step": 23453 + }, + { + "epoch": 1.51, + "learning_rate": 1.4988514137272162e-05, + "loss": 0.9898, + "step": 23454 + }, + { + "epoch": 1.51, + "learning_rate": 1.4988299448249213e-05, + "loss": 1.0115, + "step": 23455 + }, + { + "epoch": 1.51, + "learning_rate": 1.4988084759226262e-05, + "loss": 0.9544, + "step": 23456 + }, + { + "epoch": 1.51, + "learning_rate": 1.4987870070203313e-05, + "loss": 0.9797, + "step": 23457 + }, + { + "epoch": 1.51, + "learning_rate": 1.4987655381180362e-05, + "loss": 0.9281, + "step": 23458 + }, + { + "epoch": 1.51, + "learning_rate": 1.498744069215741e-05, + "loss": 0.8914, + "step": 23459 + }, + { + "epoch": 1.51, + "learning_rate": 1.4987226003134461e-05, + "loss": 1.0208, + "step": 23460 + }, + { + "epoch": 1.51, + "learning_rate": 1.498701131411151e-05, + "loss": 1.0158, + "step": 23461 + }, + { + "epoch": 1.51, + "learning_rate": 1.498679662508856e-05, + "loss": 1.0164, + "step": 23462 + }, + { + "epoch": 1.51, + "learning_rate": 1.498658193606561e-05, + "loss": 0.9441, + "step": 23463 + }, + { + "epoch": 1.51, + "learning_rate": 1.4986367247042659e-05, + "loss": 1.0205, + "step": 23464 + }, + { + "epoch": 1.51, + "learning_rate": 1.498615255801971e-05, + "loss": 0.9193, + "step": 23465 + }, + { + "epoch": 1.51, + "learning_rate": 1.498593786899676e-05, + "loss": 0.9862, + "step": 23466 + }, + { + "epoch": 1.51, + "learning_rate": 1.498572317997381e-05, + "loss": 0.9587, + "step": 23467 + }, + { + "epoch": 1.51, + "learning_rate": 1.498550849095086e-05, + "loss": 0.9627, + "step": 23468 + }, + { + "epoch": 1.51, + "learning_rate": 1.4985293801927909e-05, + "loss": 0.9708, + "step": 23469 + }, + { + "epoch": 1.51, + "learning_rate": 1.498507911290496e-05, + "loss": 0.9364, + "step": 23470 + }, + { + "epoch": 1.51, + "learning_rate": 1.4984864423882008e-05, + "loss": 1.0265, + "step": 23471 + }, + { + "epoch": 1.51, + "learning_rate": 1.4984649734859057e-05, + "loss": 1.0141, + "step": 23472 + }, + { + "epoch": 1.51, + "learning_rate": 1.4984435045836108e-05, + "loss": 0.9929, + "step": 23473 + }, + { + "epoch": 1.51, + "learning_rate": 1.4984220356813157e-05, + "loss": 1.0217, + "step": 23474 + }, + { + "epoch": 1.51, + "learning_rate": 1.4984005667790208e-05, + "loss": 0.9739, + "step": 23475 + }, + { + "epoch": 1.51, + "learning_rate": 1.4983790978767257e-05, + "loss": 0.931, + "step": 23476 + }, + { + "epoch": 1.51, + "learning_rate": 1.4983576289744305e-05, + "loss": 0.9198, + "step": 23477 + }, + { + "epoch": 1.51, + "learning_rate": 1.4983361600721356e-05, + "loss": 1.0755, + "step": 23478 + }, + { + "epoch": 1.51, + "learning_rate": 1.4983146911698405e-05, + "loss": 0.98, + "step": 23479 + }, + { + "epoch": 1.51, + "learning_rate": 1.4982932222675454e-05, + "loss": 0.9134, + "step": 23480 + }, + { + "epoch": 1.51, + "learning_rate": 1.4982717533652506e-05, + "loss": 0.954, + "step": 23481 + }, + { + "epoch": 1.51, + "learning_rate": 1.4982502844629555e-05, + "loss": 0.9969, + "step": 23482 + }, + { + "epoch": 1.51, + "learning_rate": 1.4982288155606606e-05, + "loss": 1.0305, + "step": 23483 + }, + { + "epoch": 1.51, + "learning_rate": 1.4982073466583655e-05, + "loss": 1.0373, + "step": 23484 + }, + { + "epoch": 1.51, + "learning_rate": 1.4981858777560706e-05, + "loss": 0.9564, + "step": 23485 + }, + { + "epoch": 1.51, + "learning_rate": 1.4981644088537755e-05, + "loss": 0.9591, + "step": 23486 + }, + { + "epoch": 1.51, + "learning_rate": 1.4981429399514804e-05, + "loss": 1.0651, + "step": 23487 + }, + { + "epoch": 1.51, + "learning_rate": 1.4981214710491854e-05, + "loss": 1.0025, + "step": 23488 + }, + { + "epoch": 1.51, + "learning_rate": 1.4981000021468903e-05, + "loss": 0.8939, + "step": 23489 + }, + { + "epoch": 1.51, + "learning_rate": 1.4980785332445952e-05, + "loss": 0.9854, + "step": 23490 + }, + { + "epoch": 1.51, + "learning_rate": 1.4980570643423003e-05, + "loss": 0.9444, + "step": 23491 + }, + { + "epoch": 1.51, + "learning_rate": 1.4980355954400052e-05, + "loss": 0.9821, + "step": 23492 + }, + { + "epoch": 1.51, + "learning_rate": 1.4980141265377102e-05, + "loss": 0.9717, + "step": 23493 + }, + { + "epoch": 1.51, + "learning_rate": 1.4979926576354151e-05, + "loss": 1.0463, + "step": 23494 + }, + { + "epoch": 1.51, + "learning_rate": 1.4979711887331204e-05, + "loss": 0.975, + "step": 23495 + }, + { + "epoch": 1.51, + "learning_rate": 1.4979497198308253e-05, + "loss": 0.9315, + "step": 23496 + }, + { + "epoch": 1.51, + "learning_rate": 1.4979282509285302e-05, + "loss": 0.924, + "step": 23497 + }, + { + "epoch": 1.51, + "learning_rate": 1.4979067820262352e-05, + "loss": 0.9924, + "step": 23498 + }, + { + "epoch": 1.51, + "learning_rate": 1.4978853131239401e-05, + "loss": 1.0438, + "step": 23499 + }, + { + "epoch": 1.51, + "learning_rate": 1.497863844221645e-05, + "loss": 0.9988, + "step": 23500 + }, + { + "epoch": 1.51, + "learning_rate": 1.4978423753193501e-05, + "loss": 1.0, + "step": 23501 + }, + { + "epoch": 1.51, + "learning_rate": 1.497820906417055e-05, + "loss": 1.002, + "step": 23502 + }, + { + "epoch": 1.51, + "learning_rate": 1.49779943751476e-05, + "loss": 0.9808, + "step": 23503 + }, + { + "epoch": 1.51, + "learning_rate": 1.497777968612465e-05, + "loss": 1.0063, + "step": 23504 + }, + { + "epoch": 1.51, + "learning_rate": 1.4977564997101699e-05, + "loss": 0.9569, + "step": 23505 + }, + { + "epoch": 1.51, + "learning_rate": 1.497735030807875e-05, + "loss": 1.0034, + "step": 23506 + }, + { + "epoch": 1.51, + "learning_rate": 1.4977135619055798e-05, + "loss": 0.9598, + "step": 23507 + }, + { + "epoch": 1.51, + "learning_rate": 1.4976920930032847e-05, + "loss": 1.0266, + "step": 23508 + }, + { + "epoch": 1.51, + "learning_rate": 1.4976706241009898e-05, + "loss": 0.9633, + "step": 23509 + }, + { + "epoch": 1.51, + "learning_rate": 1.4976491551986948e-05, + "loss": 0.9128, + "step": 23510 + }, + { + "epoch": 1.51, + "learning_rate": 1.4976276862963999e-05, + "loss": 1.0025, + "step": 23511 + }, + { + "epoch": 1.51, + "learning_rate": 1.4976062173941048e-05, + "loss": 0.9923, + "step": 23512 + }, + { + "epoch": 1.51, + "learning_rate": 1.4975847484918099e-05, + "loss": 0.9785, + "step": 23513 + }, + { + "epoch": 1.51, + "learning_rate": 1.4975632795895148e-05, + "loss": 1.0729, + "step": 23514 + }, + { + "epoch": 1.51, + "learning_rate": 1.4975418106872197e-05, + "loss": 0.9406, + "step": 23515 + }, + { + "epoch": 1.51, + "learning_rate": 1.4975203417849247e-05, + "loss": 0.9967, + "step": 23516 + }, + { + "epoch": 1.51, + "learning_rate": 1.4974988728826296e-05, + "loss": 0.9454, + "step": 23517 + }, + { + "epoch": 1.51, + "learning_rate": 1.4974774039803345e-05, + "loss": 1.0229, + "step": 23518 + }, + { + "epoch": 1.51, + "learning_rate": 1.4974559350780396e-05, + "loss": 0.9705, + "step": 23519 + }, + { + "epoch": 1.51, + "learning_rate": 1.4974344661757445e-05, + "loss": 0.9398, + "step": 23520 + }, + { + "epoch": 1.51, + "learning_rate": 1.4974129972734494e-05, + "loss": 0.9777, + "step": 23521 + }, + { + "epoch": 1.51, + "learning_rate": 1.4973915283711544e-05, + "loss": 1.0453, + "step": 23522 + }, + { + "epoch": 1.51, + "learning_rate": 1.4973700594688593e-05, + "loss": 1.052, + "step": 23523 + }, + { + "epoch": 1.51, + "learning_rate": 1.4973485905665646e-05, + "loss": 0.9596, + "step": 23524 + }, + { + "epoch": 1.51, + "learning_rate": 1.4973271216642695e-05, + "loss": 0.9679, + "step": 23525 + }, + { + "epoch": 1.51, + "learning_rate": 1.4973056527619745e-05, + "loss": 1.0053, + "step": 23526 + }, + { + "epoch": 1.51, + "learning_rate": 1.4972841838596794e-05, + "loss": 0.9311, + "step": 23527 + }, + { + "epoch": 1.51, + "learning_rate": 1.4972627149573843e-05, + "loss": 0.8697, + "step": 23528 + }, + { + "epoch": 1.51, + "learning_rate": 1.4972412460550894e-05, + "loss": 1.0084, + "step": 23529 + }, + { + "epoch": 1.51, + "learning_rate": 1.4972197771527943e-05, + "loss": 0.9559, + "step": 23530 + }, + { + "epoch": 1.51, + "learning_rate": 1.4971983082504992e-05, + "loss": 0.9461, + "step": 23531 + }, + { + "epoch": 1.51, + "learning_rate": 1.4971768393482043e-05, + "loss": 0.8932, + "step": 23532 + }, + { + "epoch": 1.51, + "learning_rate": 1.4971553704459092e-05, + "loss": 0.993, + "step": 23533 + }, + { + "epoch": 1.51, + "learning_rate": 1.4971339015436142e-05, + "loss": 1.0237, + "step": 23534 + }, + { + "epoch": 1.51, + "learning_rate": 1.4971124326413191e-05, + "loss": 0.9483, + "step": 23535 + }, + { + "epoch": 1.51, + "learning_rate": 1.497090963739024e-05, + "loss": 0.9505, + "step": 23536 + }, + { + "epoch": 1.51, + "learning_rate": 1.497069494836729e-05, + "loss": 0.9253, + "step": 23537 + }, + { + "epoch": 1.51, + "learning_rate": 1.497048025934434e-05, + "loss": 0.989, + "step": 23538 + }, + { + "epoch": 1.51, + "learning_rate": 1.4970265570321392e-05, + "loss": 0.9707, + "step": 23539 + }, + { + "epoch": 1.51, + "learning_rate": 1.4970050881298441e-05, + "loss": 0.9753, + "step": 23540 + }, + { + "epoch": 1.51, + "learning_rate": 1.496983619227549e-05, + "loss": 0.953, + "step": 23541 + }, + { + "epoch": 1.51, + "learning_rate": 1.496962150325254e-05, + "loss": 1.0183, + "step": 23542 + }, + { + "epoch": 1.51, + "learning_rate": 1.496940681422959e-05, + "loss": 0.9716, + "step": 23543 + }, + { + "epoch": 1.51, + "learning_rate": 1.496919212520664e-05, + "loss": 0.9589, + "step": 23544 + }, + { + "epoch": 1.51, + "learning_rate": 1.496897743618369e-05, + "loss": 1.0389, + "step": 23545 + }, + { + "epoch": 1.51, + "learning_rate": 1.4968762747160738e-05, + "loss": 0.9433, + "step": 23546 + }, + { + "epoch": 1.51, + "learning_rate": 1.4968548058137789e-05, + "loss": 0.959, + "step": 23547 + }, + { + "epoch": 1.52, + "learning_rate": 1.4968333369114838e-05, + "loss": 1.0137, + "step": 23548 + }, + { + "epoch": 1.52, + "learning_rate": 1.4968118680091887e-05, + "loss": 0.9703, + "step": 23549 + }, + { + "epoch": 1.52, + "learning_rate": 1.4967903991068938e-05, + "loss": 0.9167, + "step": 23550 + }, + { + "epoch": 1.52, + "learning_rate": 1.4967689302045986e-05, + "loss": 0.9482, + "step": 23551 + }, + { + "epoch": 1.52, + "learning_rate": 1.4967474613023037e-05, + "loss": 0.9495, + "step": 23552 + }, + { + "epoch": 1.52, + "learning_rate": 1.4967259924000088e-05, + "loss": 0.9694, + "step": 23553 + }, + { + "epoch": 1.52, + "learning_rate": 1.4967045234977138e-05, + "loss": 0.9194, + "step": 23554 + }, + { + "epoch": 1.52, + "learning_rate": 1.4966830545954187e-05, + "loss": 1.0673, + "step": 23555 + }, + { + "epoch": 1.52, + "learning_rate": 1.4966615856931236e-05, + "loss": 1.0368, + "step": 23556 + }, + { + "epoch": 1.52, + "learning_rate": 1.4966401167908287e-05, + "loss": 1.0241, + "step": 23557 + }, + { + "epoch": 1.52, + "learning_rate": 1.4966186478885336e-05, + "loss": 0.9867, + "step": 23558 + }, + { + "epoch": 1.52, + "learning_rate": 1.4965971789862385e-05, + "loss": 1.0205, + "step": 23559 + }, + { + "epoch": 1.52, + "learning_rate": 1.4965757100839436e-05, + "loss": 0.9894, + "step": 23560 + }, + { + "epoch": 1.52, + "learning_rate": 1.4965542411816485e-05, + "loss": 0.9863, + "step": 23561 + }, + { + "epoch": 1.52, + "learning_rate": 1.4965327722793535e-05, + "loss": 0.99, + "step": 23562 + }, + { + "epoch": 1.52, + "learning_rate": 1.4965113033770584e-05, + "loss": 0.9839, + "step": 23563 + }, + { + "epoch": 1.52, + "learning_rate": 1.4964898344747633e-05, + "loss": 0.9483, + "step": 23564 + }, + { + "epoch": 1.52, + "learning_rate": 1.4964683655724684e-05, + "loss": 1.0125, + "step": 23565 + }, + { + "epoch": 1.52, + "learning_rate": 1.4964468966701733e-05, + "loss": 0.9723, + "step": 23566 + }, + { + "epoch": 1.52, + "learning_rate": 1.4964254277678782e-05, + "loss": 0.9297, + "step": 23567 + }, + { + "epoch": 1.52, + "learning_rate": 1.4964039588655834e-05, + "loss": 0.9809, + "step": 23568 + }, + { + "epoch": 1.52, + "learning_rate": 1.4963824899632883e-05, + "loss": 0.953, + "step": 23569 + }, + { + "epoch": 1.52, + "learning_rate": 1.4963610210609934e-05, + "loss": 1.036, + "step": 23570 + }, + { + "epoch": 1.52, + "learning_rate": 1.4963395521586983e-05, + "loss": 1.0859, + "step": 23571 + }, + { + "epoch": 1.52, + "learning_rate": 1.4963180832564033e-05, + "loss": 0.9838, + "step": 23572 + }, + { + "epoch": 1.52, + "learning_rate": 1.4962966143541082e-05, + "loss": 1.0352, + "step": 23573 + }, + { + "epoch": 1.52, + "learning_rate": 1.4962751454518131e-05, + "loss": 0.9675, + "step": 23574 + }, + { + "epoch": 1.52, + "learning_rate": 1.4962536765495182e-05, + "loss": 0.9696, + "step": 23575 + }, + { + "epoch": 1.52, + "learning_rate": 1.4962322076472231e-05, + "loss": 0.9435, + "step": 23576 + }, + { + "epoch": 1.52, + "learning_rate": 1.496210738744928e-05, + "loss": 0.9578, + "step": 23577 + }, + { + "epoch": 1.52, + "learning_rate": 1.496189269842633e-05, + "loss": 0.9398, + "step": 23578 + }, + { + "epoch": 1.52, + "learning_rate": 1.496167800940338e-05, + "loss": 0.9395, + "step": 23579 + }, + { + "epoch": 1.52, + "learning_rate": 1.496146332038043e-05, + "loss": 0.9987, + "step": 23580 + }, + { + "epoch": 1.52, + "learning_rate": 1.496124863135748e-05, + "loss": 0.9909, + "step": 23581 + }, + { + "epoch": 1.52, + "learning_rate": 1.4961033942334532e-05, + "loss": 0.9677, + "step": 23582 + }, + { + "epoch": 1.52, + "learning_rate": 1.496081925331158e-05, + "loss": 0.8936, + "step": 23583 + }, + { + "epoch": 1.52, + "learning_rate": 1.496060456428863e-05, + "loss": 1.0381, + "step": 23584 + }, + { + "epoch": 1.52, + "learning_rate": 1.496038987526568e-05, + "loss": 1.0126, + "step": 23585 + }, + { + "epoch": 1.52, + "learning_rate": 1.4960175186242729e-05, + "loss": 0.9408, + "step": 23586 + }, + { + "epoch": 1.52, + "learning_rate": 1.4959960497219778e-05, + "loss": 0.9061, + "step": 23587 + }, + { + "epoch": 1.52, + "learning_rate": 1.4959745808196829e-05, + "loss": 1.0161, + "step": 23588 + }, + { + "epoch": 1.52, + "learning_rate": 1.4959531119173878e-05, + "loss": 0.8998, + "step": 23589 + }, + { + "epoch": 1.52, + "learning_rate": 1.4959316430150928e-05, + "loss": 1.0952, + "step": 23590 + }, + { + "epoch": 1.52, + "learning_rate": 1.4959101741127977e-05, + "loss": 1.0292, + "step": 23591 + }, + { + "epoch": 1.52, + "learning_rate": 1.4958887052105026e-05, + "loss": 0.9569, + "step": 23592 + }, + { + "epoch": 1.52, + "learning_rate": 1.4958672363082077e-05, + "loss": 0.9113, + "step": 23593 + }, + { + "epoch": 1.52, + "learning_rate": 1.4958457674059126e-05, + "loss": 0.9893, + "step": 23594 + }, + { + "epoch": 1.52, + "learning_rate": 1.4958242985036175e-05, + "loss": 0.9743, + "step": 23595 + }, + { + "epoch": 1.52, + "learning_rate": 1.4958028296013225e-05, + "loss": 0.9352, + "step": 23596 + }, + { + "epoch": 1.52, + "learning_rate": 1.4957813606990276e-05, + "loss": 1.0409, + "step": 23597 + }, + { + "epoch": 1.52, + "learning_rate": 1.4957598917967327e-05, + "loss": 0.9992, + "step": 23598 + }, + { + "epoch": 1.52, + "learning_rate": 1.4957384228944376e-05, + "loss": 0.9615, + "step": 23599 + }, + { + "epoch": 1.52, + "learning_rate": 1.4957169539921426e-05, + "loss": 0.9819, + "step": 23600 + }, + { + "epoch": 1.52, + "learning_rate": 1.4956954850898475e-05, + "loss": 0.973, + "step": 23601 + }, + { + "epoch": 1.52, + "learning_rate": 1.4956740161875524e-05, + "loss": 0.8885, + "step": 23602 + }, + { + "epoch": 1.52, + "learning_rate": 1.4956525472852575e-05, + "loss": 0.9265, + "step": 23603 + }, + { + "epoch": 1.52, + "learning_rate": 1.4956310783829624e-05, + "loss": 1.024, + "step": 23604 + }, + { + "epoch": 1.52, + "learning_rate": 1.4956096094806673e-05, + "loss": 1.0079, + "step": 23605 + }, + { + "epoch": 1.52, + "learning_rate": 1.4955881405783724e-05, + "loss": 0.9949, + "step": 23606 + }, + { + "epoch": 1.52, + "learning_rate": 1.4955666716760773e-05, + "loss": 0.9444, + "step": 23607 + }, + { + "epoch": 1.52, + "learning_rate": 1.4955452027737822e-05, + "loss": 0.9359, + "step": 23608 + }, + { + "epoch": 1.52, + "learning_rate": 1.4955237338714872e-05, + "loss": 0.9994, + "step": 23609 + }, + { + "epoch": 1.52, + "learning_rate": 1.4955022649691921e-05, + "loss": 0.9701, + "step": 23610 + }, + { + "epoch": 1.52, + "learning_rate": 1.4954807960668972e-05, + "loss": 0.9605, + "step": 23611 + }, + { + "epoch": 1.52, + "learning_rate": 1.4954593271646022e-05, + "loss": 0.957, + "step": 23612 + }, + { + "epoch": 1.52, + "learning_rate": 1.4954378582623073e-05, + "loss": 0.9392, + "step": 23613 + }, + { + "epoch": 1.52, + "learning_rate": 1.4954163893600122e-05, + "loss": 1.0831, + "step": 23614 + }, + { + "epoch": 1.52, + "learning_rate": 1.4953949204577171e-05, + "loss": 0.9397, + "step": 23615 + }, + { + "epoch": 1.52, + "learning_rate": 1.4953734515554222e-05, + "loss": 1.0134, + "step": 23616 + }, + { + "epoch": 1.52, + "learning_rate": 1.495351982653127e-05, + "loss": 1.0181, + "step": 23617 + }, + { + "epoch": 1.52, + "learning_rate": 1.495330513750832e-05, + "loss": 0.9289, + "step": 23618 + }, + { + "epoch": 1.52, + "learning_rate": 1.495309044848537e-05, + "loss": 0.9166, + "step": 23619 + }, + { + "epoch": 1.52, + "learning_rate": 1.495287575946242e-05, + "loss": 1.0222, + "step": 23620 + }, + { + "epoch": 1.52, + "learning_rate": 1.495266107043947e-05, + "loss": 1.0212, + "step": 23621 + }, + { + "epoch": 1.52, + "learning_rate": 1.4952446381416519e-05, + "loss": 0.9554, + "step": 23622 + }, + { + "epoch": 1.52, + "learning_rate": 1.4952231692393568e-05, + "loss": 0.9803, + "step": 23623 + }, + { + "epoch": 1.52, + "learning_rate": 1.4952017003370619e-05, + "loss": 0.9821, + "step": 23624 + }, + { + "epoch": 1.52, + "learning_rate": 1.4951802314347667e-05, + "loss": 0.948, + "step": 23625 + }, + { + "epoch": 1.52, + "learning_rate": 1.495158762532472e-05, + "loss": 1.0201, + "step": 23626 + }, + { + "epoch": 1.52, + "learning_rate": 1.4951372936301769e-05, + "loss": 0.9622, + "step": 23627 + }, + { + "epoch": 1.52, + "learning_rate": 1.4951158247278818e-05, + "loss": 0.9047, + "step": 23628 + }, + { + "epoch": 1.52, + "learning_rate": 1.4950943558255868e-05, + "loss": 0.9285, + "step": 23629 + }, + { + "epoch": 1.52, + "learning_rate": 1.4950728869232917e-05, + "loss": 0.9984, + "step": 23630 + }, + { + "epoch": 1.52, + "learning_rate": 1.4950514180209968e-05, + "loss": 0.9971, + "step": 23631 + }, + { + "epoch": 1.52, + "learning_rate": 1.4950299491187017e-05, + "loss": 0.9438, + "step": 23632 + }, + { + "epoch": 1.52, + "learning_rate": 1.4950084802164066e-05, + "loss": 0.9591, + "step": 23633 + }, + { + "epoch": 1.52, + "learning_rate": 1.4949870113141117e-05, + "loss": 0.9875, + "step": 23634 + }, + { + "epoch": 1.52, + "learning_rate": 1.4949655424118166e-05, + "loss": 0.9236, + "step": 23635 + }, + { + "epoch": 1.52, + "learning_rate": 1.4949440735095215e-05, + "loss": 0.973, + "step": 23636 + }, + { + "epoch": 1.52, + "learning_rate": 1.4949226046072265e-05, + "loss": 0.9947, + "step": 23637 + }, + { + "epoch": 1.52, + "learning_rate": 1.4949011357049314e-05, + "loss": 1.0023, + "step": 23638 + }, + { + "epoch": 1.52, + "learning_rate": 1.4948796668026365e-05, + "loss": 1.0558, + "step": 23639 + }, + { + "epoch": 1.52, + "learning_rate": 1.4948581979003414e-05, + "loss": 1.0327, + "step": 23640 + }, + { + "epoch": 1.52, + "learning_rate": 1.4948367289980466e-05, + "loss": 1.0039, + "step": 23641 + }, + { + "epoch": 1.52, + "learning_rate": 1.4948152600957515e-05, + "loss": 0.974, + "step": 23642 + }, + { + "epoch": 1.52, + "learning_rate": 1.4947937911934564e-05, + "loss": 0.976, + "step": 23643 + }, + { + "epoch": 1.52, + "learning_rate": 1.4947723222911615e-05, + "loss": 0.9612, + "step": 23644 + }, + { + "epoch": 1.52, + "learning_rate": 1.4947508533888664e-05, + "loss": 1.0033, + "step": 23645 + }, + { + "epoch": 1.52, + "learning_rate": 1.4947293844865713e-05, + "loss": 1.0234, + "step": 23646 + }, + { + "epoch": 1.52, + "learning_rate": 1.4947079155842763e-05, + "loss": 1.0049, + "step": 23647 + }, + { + "epoch": 1.52, + "learning_rate": 1.4946864466819812e-05, + "loss": 0.9121, + "step": 23648 + }, + { + "epoch": 1.52, + "learning_rate": 1.4946649777796863e-05, + "loss": 0.9945, + "step": 23649 + }, + { + "epoch": 1.52, + "learning_rate": 1.4946435088773912e-05, + "loss": 0.9561, + "step": 23650 + }, + { + "epoch": 1.52, + "learning_rate": 1.4946220399750961e-05, + "loss": 0.9193, + "step": 23651 + }, + { + "epoch": 1.52, + "learning_rate": 1.4946005710728012e-05, + "loss": 1.0038, + "step": 23652 + }, + { + "epoch": 1.52, + "learning_rate": 1.494579102170506e-05, + "loss": 0.9778, + "step": 23653 + }, + { + "epoch": 1.52, + "learning_rate": 1.494557633268211e-05, + "loss": 1.0099, + "step": 23654 + }, + { + "epoch": 1.52, + "learning_rate": 1.4945361643659162e-05, + "loss": 1.0087, + "step": 23655 + }, + { + "epoch": 1.52, + "learning_rate": 1.494514695463621e-05, + "loss": 0.9564, + "step": 23656 + }, + { + "epoch": 1.52, + "learning_rate": 1.4944932265613261e-05, + "loss": 0.9771, + "step": 23657 + }, + { + "epoch": 1.52, + "learning_rate": 1.494471757659031e-05, + "loss": 1.0261, + "step": 23658 + }, + { + "epoch": 1.52, + "learning_rate": 1.4944502887567361e-05, + "loss": 0.9749, + "step": 23659 + }, + { + "epoch": 1.52, + "learning_rate": 1.494428819854441e-05, + "loss": 0.864, + "step": 23660 + }, + { + "epoch": 1.52, + "learning_rate": 1.4944073509521459e-05, + "loss": 0.9769, + "step": 23661 + }, + { + "epoch": 1.52, + "learning_rate": 1.494385882049851e-05, + "loss": 0.9928, + "step": 23662 + }, + { + "epoch": 1.52, + "learning_rate": 1.4943644131475559e-05, + "loss": 0.9935, + "step": 23663 + }, + { + "epoch": 1.52, + "learning_rate": 1.4943429442452608e-05, + "loss": 0.976, + "step": 23664 + }, + { + "epoch": 1.52, + "learning_rate": 1.4943214753429658e-05, + "loss": 0.956, + "step": 23665 + }, + { + "epoch": 1.52, + "learning_rate": 1.4943000064406707e-05, + "loss": 0.9012, + "step": 23666 + }, + { + "epoch": 1.52, + "learning_rate": 1.4942785375383758e-05, + "loss": 1.0041, + "step": 23667 + }, + { + "epoch": 1.52, + "learning_rate": 1.4942570686360807e-05, + "loss": 0.9963, + "step": 23668 + }, + { + "epoch": 1.52, + "learning_rate": 1.4942355997337856e-05, + "loss": 0.9462, + "step": 23669 + }, + { + "epoch": 1.52, + "learning_rate": 1.4942141308314908e-05, + "loss": 0.9939, + "step": 23670 + }, + { + "epoch": 1.52, + "learning_rate": 1.4941926619291957e-05, + "loss": 0.9661, + "step": 23671 + }, + { + "epoch": 1.52, + "learning_rate": 1.4941711930269008e-05, + "loss": 0.9232, + "step": 23672 + }, + { + "epoch": 1.52, + "learning_rate": 1.4941497241246057e-05, + "loss": 1.0048, + "step": 23673 + }, + { + "epoch": 1.52, + "learning_rate": 1.4941282552223106e-05, + "loss": 1.0268, + "step": 23674 + }, + { + "epoch": 1.52, + "learning_rate": 1.4941067863200156e-05, + "loss": 0.9247, + "step": 23675 + }, + { + "epoch": 1.52, + "learning_rate": 1.4940853174177205e-05, + "loss": 0.9993, + "step": 23676 + }, + { + "epoch": 1.52, + "learning_rate": 1.4940638485154256e-05, + "loss": 1.0239, + "step": 23677 + }, + { + "epoch": 1.52, + "learning_rate": 1.4940423796131305e-05, + "loss": 0.8947, + "step": 23678 + }, + { + "epoch": 1.52, + "learning_rate": 1.4940209107108354e-05, + "loss": 0.9738, + "step": 23679 + }, + { + "epoch": 1.52, + "learning_rate": 1.4939994418085405e-05, + "loss": 1.0266, + "step": 23680 + }, + { + "epoch": 1.52, + "learning_rate": 1.4939779729062454e-05, + "loss": 0.9355, + "step": 23681 + }, + { + "epoch": 1.52, + "learning_rate": 1.4939565040039503e-05, + "loss": 0.9644, + "step": 23682 + }, + { + "epoch": 1.52, + "learning_rate": 1.4939350351016553e-05, + "loss": 0.9622, + "step": 23683 + }, + { + "epoch": 1.52, + "learning_rate": 1.4939135661993604e-05, + "loss": 0.9898, + "step": 23684 + }, + { + "epoch": 1.52, + "learning_rate": 1.4938920972970655e-05, + "loss": 1.0058, + "step": 23685 + }, + { + "epoch": 1.52, + "learning_rate": 1.4938706283947703e-05, + "loss": 0.9939, + "step": 23686 + }, + { + "epoch": 1.52, + "learning_rate": 1.4938491594924754e-05, + "loss": 0.9909, + "step": 23687 + }, + { + "epoch": 1.52, + "learning_rate": 1.4938276905901803e-05, + "loss": 0.9694, + "step": 23688 + }, + { + "epoch": 1.52, + "learning_rate": 1.4938062216878852e-05, + "loss": 0.9597, + "step": 23689 + }, + { + "epoch": 1.52, + "learning_rate": 1.4937847527855903e-05, + "loss": 0.906, + "step": 23690 + }, + { + "epoch": 1.52, + "learning_rate": 1.4937632838832952e-05, + "loss": 1.0361, + "step": 23691 + }, + { + "epoch": 1.52, + "learning_rate": 1.493741814981e-05, + "loss": 0.9916, + "step": 23692 + }, + { + "epoch": 1.52, + "learning_rate": 1.4937203460787051e-05, + "loss": 0.988, + "step": 23693 + }, + { + "epoch": 1.52, + "learning_rate": 1.49369887717641e-05, + "loss": 1.0333, + "step": 23694 + }, + { + "epoch": 1.52, + "learning_rate": 1.493677408274115e-05, + "loss": 1.0222, + "step": 23695 + }, + { + "epoch": 1.52, + "learning_rate": 1.49365593937182e-05, + "loss": 0.9773, + "step": 23696 + }, + { + "epoch": 1.52, + "learning_rate": 1.4936344704695249e-05, + "loss": 0.9979, + "step": 23697 + }, + { + "epoch": 1.52, + "learning_rate": 1.49361300156723e-05, + "loss": 0.9365, + "step": 23698 + }, + { + "epoch": 1.52, + "learning_rate": 1.493591532664935e-05, + "loss": 0.9227, + "step": 23699 + }, + { + "epoch": 1.52, + "learning_rate": 1.4935700637626401e-05, + "loss": 0.8946, + "step": 23700 + }, + { + "epoch": 1.52, + "learning_rate": 1.493548594860345e-05, + "loss": 1.0252, + "step": 23701 + }, + { + "epoch": 1.52, + "learning_rate": 1.4935271259580499e-05, + "loss": 1.0236, + "step": 23702 + }, + { + "epoch": 1.52, + "learning_rate": 1.493505657055755e-05, + "loss": 1.0012, + "step": 23703 + }, + { + "epoch": 1.53, + "learning_rate": 1.4934841881534598e-05, + "loss": 0.9728, + "step": 23704 + }, + { + "epoch": 1.53, + "learning_rate": 1.4934627192511647e-05, + "loss": 0.9384, + "step": 23705 + }, + { + "epoch": 1.53, + "learning_rate": 1.4934412503488698e-05, + "loss": 1.0074, + "step": 23706 + }, + { + "epoch": 1.53, + "learning_rate": 1.4934197814465747e-05, + "loss": 0.984, + "step": 23707 + }, + { + "epoch": 1.53, + "learning_rate": 1.4933983125442798e-05, + "loss": 1.0098, + "step": 23708 + }, + { + "epoch": 1.53, + "learning_rate": 1.4933768436419847e-05, + "loss": 0.9064, + "step": 23709 + }, + { + "epoch": 1.53, + "learning_rate": 1.4933553747396896e-05, + "loss": 0.9579, + "step": 23710 + }, + { + "epoch": 1.53, + "learning_rate": 1.4933339058373946e-05, + "loss": 0.9962, + "step": 23711 + }, + { + "epoch": 1.53, + "learning_rate": 1.4933124369350995e-05, + "loss": 0.8903, + "step": 23712 + }, + { + "epoch": 1.53, + "learning_rate": 1.4932909680328048e-05, + "loss": 0.9504, + "step": 23713 + }, + { + "epoch": 1.53, + "learning_rate": 1.4932694991305097e-05, + "loss": 0.9711, + "step": 23714 + }, + { + "epoch": 1.53, + "learning_rate": 1.4932480302282145e-05, + "loss": 0.9811, + "step": 23715 + }, + { + "epoch": 1.53, + "learning_rate": 1.4932265613259196e-05, + "loss": 0.9193, + "step": 23716 + }, + { + "epoch": 1.53, + "learning_rate": 1.4932050924236245e-05, + "loss": 0.9943, + "step": 23717 + }, + { + "epoch": 1.53, + "learning_rate": 1.4931836235213296e-05, + "loss": 0.9906, + "step": 23718 + }, + { + "epoch": 1.53, + "learning_rate": 1.4931621546190345e-05, + "loss": 1.0375, + "step": 23719 + }, + { + "epoch": 1.53, + "learning_rate": 1.4931406857167394e-05, + "loss": 0.9216, + "step": 23720 + }, + { + "epoch": 1.53, + "learning_rate": 1.4931192168144444e-05, + "loss": 0.9097, + "step": 23721 + }, + { + "epoch": 1.53, + "learning_rate": 1.4930977479121493e-05, + "loss": 0.9968, + "step": 23722 + }, + { + "epoch": 1.53, + "learning_rate": 1.4930762790098542e-05, + "loss": 0.9334, + "step": 23723 + }, + { + "epoch": 1.53, + "learning_rate": 1.4930548101075593e-05, + "loss": 0.9335, + "step": 23724 + }, + { + "epoch": 1.53, + "learning_rate": 1.4930333412052642e-05, + "loss": 0.8991, + "step": 23725 + }, + { + "epoch": 1.53, + "learning_rate": 1.4930118723029693e-05, + "loss": 0.9855, + "step": 23726 + }, + { + "epoch": 1.53, + "learning_rate": 1.4929904034006742e-05, + "loss": 1.029, + "step": 23727 + }, + { + "epoch": 1.53, + "learning_rate": 1.4929689344983794e-05, + "loss": 0.9975, + "step": 23728 + }, + { + "epoch": 1.53, + "learning_rate": 1.4929474655960843e-05, + "loss": 0.9988, + "step": 23729 + }, + { + "epoch": 1.53, + "learning_rate": 1.4929259966937892e-05, + "loss": 0.9636, + "step": 23730 + }, + { + "epoch": 1.53, + "learning_rate": 1.4929045277914942e-05, + "loss": 1.0489, + "step": 23731 + }, + { + "epoch": 1.53, + "learning_rate": 1.4928830588891991e-05, + "loss": 1.0264, + "step": 23732 + }, + { + "epoch": 1.53, + "learning_rate": 1.492861589986904e-05, + "loss": 0.9631, + "step": 23733 + }, + { + "epoch": 1.53, + "learning_rate": 1.4928401210846091e-05, + "loss": 1.034, + "step": 23734 + }, + { + "epoch": 1.53, + "learning_rate": 1.492818652182314e-05, + "loss": 0.9602, + "step": 23735 + }, + { + "epoch": 1.53, + "learning_rate": 1.492797183280019e-05, + "loss": 0.938, + "step": 23736 + }, + { + "epoch": 1.53, + "learning_rate": 1.492775714377724e-05, + "loss": 0.9609, + "step": 23737 + }, + { + "epoch": 1.53, + "learning_rate": 1.4927542454754289e-05, + "loss": 0.856, + "step": 23738 + }, + { + "epoch": 1.53, + "learning_rate": 1.492732776573134e-05, + "loss": 0.9875, + "step": 23739 + }, + { + "epoch": 1.53, + "learning_rate": 1.4927113076708388e-05, + "loss": 1.0599, + "step": 23740 + }, + { + "epoch": 1.53, + "learning_rate": 1.4926898387685437e-05, + "loss": 0.9625, + "step": 23741 + }, + { + "epoch": 1.53, + "learning_rate": 1.492668369866249e-05, + "loss": 0.9616, + "step": 23742 + }, + { + "epoch": 1.53, + "learning_rate": 1.4926469009639539e-05, + "loss": 0.9724, + "step": 23743 + }, + { + "epoch": 1.53, + "learning_rate": 1.492625432061659e-05, + "loss": 0.977, + "step": 23744 + }, + { + "epoch": 1.53, + "learning_rate": 1.4926039631593638e-05, + "loss": 0.8967, + "step": 23745 + }, + { + "epoch": 1.53, + "learning_rate": 1.4925824942570689e-05, + "loss": 1.0262, + "step": 23746 + }, + { + "epoch": 1.53, + "learning_rate": 1.4925610253547738e-05, + "loss": 0.9868, + "step": 23747 + }, + { + "epoch": 1.53, + "learning_rate": 1.4925395564524787e-05, + "loss": 0.903, + "step": 23748 + }, + { + "epoch": 1.53, + "learning_rate": 1.4925180875501837e-05, + "loss": 0.9707, + "step": 23749 + }, + { + "epoch": 1.53, + "learning_rate": 1.4924966186478886e-05, + "loss": 0.9721, + "step": 23750 + }, + { + "epoch": 1.53, + "learning_rate": 1.4924751497455935e-05, + "loss": 0.8938, + "step": 23751 + }, + { + "epoch": 1.53, + "learning_rate": 1.4924536808432986e-05, + "loss": 0.9664, + "step": 23752 + }, + { + "epoch": 1.53, + "learning_rate": 1.4924322119410035e-05, + "loss": 0.9776, + "step": 23753 + }, + { + "epoch": 1.53, + "learning_rate": 1.4924107430387084e-05, + "loss": 0.9749, + "step": 23754 + }, + { + "epoch": 1.53, + "learning_rate": 1.4923892741364135e-05, + "loss": 1.0444, + "step": 23755 + }, + { + "epoch": 1.53, + "learning_rate": 1.4923678052341184e-05, + "loss": 1.0294, + "step": 23756 + }, + { + "epoch": 1.53, + "learning_rate": 1.4923463363318236e-05, + "loss": 0.9436, + "step": 23757 + }, + { + "epoch": 1.53, + "learning_rate": 1.4923248674295285e-05, + "loss": 1.0352, + "step": 23758 + }, + { + "epoch": 1.53, + "learning_rate": 1.4923033985272336e-05, + "loss": 0.9933, + "step": 23759 + }, + { + "epoch": 1.53, + "learning_rate": 1.4922819296249384e-05, + "loss": 0.997, + "step": 23760 + }, + { + "epoch": 1.53, + "learning_rate": 1.4922604607226433e-05, + "loss": 0.9478, + "step": 23761 + }, + { + "epoch": 1.53, + "learning_rate": 1.4922389918203484e-05, + "loss": 0.9935, + "step": 23762 + }, + { + "epoch": 1.53, + "learning_rate": 1.4922175229180533e-05, + "loss": 0.9735, + "step": 23763 + }, + { + "epoch": 1.53, + "learning_rate": 1.4921960540157582e-05, + "loss": 0.9089, + "step": 23764 + }, + { + "epoch": 1.53, + "learning_rate": 1.4921745851134633e-05, + "loss": 1.0139, + "step": 23765 + }, + { + "epoch": 1.53, + "learning_rate": 1.4921531162111682e-05, + "loss": 0.971, + "step": 23766 + }, + { + "epoch": 1.53, + "learning_rate": 1.4921316473088732e-05, + "loss": 0.9695, + "step": 23767 + }, + { + "epoch": 1.53, + "learning_rate": 1.4921101784065781e-05, + "loss": 0.9828, + "step": 23768 + }, + { + "epoch": 1.53, + "learning_rate": 1.492088709504283e-05, + "loss": 0.9474, + "step": 23769 + }, + { + "epoch": 1.53, + "learning_rate": 1.4920672406019881e-05, + "loss": 0.9247, + "step": 23770 + }, + { + "epoch": 1.53, + "learning_rate": 1.4920457716996932e-05, + "loss": 0.9203, + "step": 23771 + }, + { + "epoch": 1.53, + "learning_rate": 1.4920243027973982e-05, + "loss": 0.9516, + "step": 23772 + }, + { + "epoch": 1.53, + "learning_rate": 1.4920028338951031e-05, + "loss": 0.9909, + "step": 23773 + }, + { + "epoch": 1.53, + "learning_rate": 1.4919813649928082e-05, + "loss": 0.9364, + "step": 23774 + }, + { + "epoch": 1.53, + "learning_rate": 1.491959896090513e-05, + "loss": 0.9754, + "step": 23775 + }, + { + "epoch": 1.53, + "learning_rate": 1.491938427188218e-05, + "loss": 0.9525, + "step": 23776 + }, + { + "epoch": 1.53, + "learning_rate": 1.491916958285923e-05, + "loss": 0.9524, + "step": 23777 + }, + { + "epoch": 1.53, + "learning_rate": 1.491895489383628e-05, + "loss": 0.9563, + "step": 23778 + }, + { + "epoch": 1.53, + "learning_rate": 1.4918740204813328e-05, + "loss": 0.9387, + "step": 23779 + }, + { + "epoch": 1.53, + "learning_rate": 1.4918525515790379e-05, + "loss": 0.9338, + "step": 23780 + }, + { + "epoch": 1.53, + "learning_rate": 1.4918310826767428e-05, + "loss": 0.9514, + "step": 23781 + }, + { + "epoch": 1.53, + "learning_rate": 1.4918096137744477e-05, + "loss": 1.0278, + "step": 23782 + }, + { + "epoch": 1.53, + "learning_rate": 1.4917881448721528e-05, + "loss": 1.0483, + "step": 23783 + }, + { + "epoch": 1.53, + "learning_rate": 1.4917666759698577e-05, + "loss": 1.0201, + "step": 23784 + }, + { + "epoch": 1.53, + "learning_rate": 1.4917452070675627e-05, + "loss": 1.0432, + "step": 23785 + }, + { + "epoch": 1.53, + "learning_rate": 1.4917237381652678e-05, + "loss": 0.9758, + "step": 23786 + }, + { + "epoch": 1.53, + "learning_rate": 1.4917022692629729e-05, + "loss": 0.9077, + "step": 23787 + }, + { + "epoch": 1.53, + "learning_rate": 1.4916808003606778e-05, + "loss": 0.9148, + "step": 23788 + }, + { + "epoch": 1.53, + "learning_rate": 1.4916593314583826e-05, + "loss": 0.9825, + "step": 23789 + }, + { + "epoch": 1.53, + "learning_rate": 1.4916378625560877e-05, + "loss": 0.9591, + "step": 23790 + }, + { + "epoch": 1.53, + "learning_rate": 1.4916163936537926e-05, + "loss": 1.0191, + "step": 23791 + }, + { + "epoch": 1.53, + "learning_rate": 1.4915949247514975e-05, + "loss": 0.9176, + "step": 23792 + }, + { + "epoch": 1.53, + "learning_rate": 1.4915734558492026e-05, + "loss": 0.887, + "step": 23793 + }, + { + "epoch": 1.53, + "learning_rate": 1.4915519869469075e-05, + "loss": 1.0407, + "step": 23794 + }, + { + "epoch": 1.53, + "learning_rate": 1.4915305180446125e-05, + "loss": 0.9838, + "step": 23795 + }, + { + "epoch": 1.53, + "learning_rate": 1.4915090491423174e-05, + "loss": 1.0093, + "step": 23796 + }, + { + "epoch": 1.53, + "learning_rate": 1.4914875802400223e-05, + "loss": 0.9598, + "step": 23797 + }, + { + "epoch": 1.53, + "learning_rate": 1.4914661113377274e-05, + "loss": 0.9788, + "step": 23798 + }, + { + "epoch": 1.53, + "learning_rate": 1.4914446424354323e-05, + "loss": 0.9869, + "step": 23799 + }, + { + "epoch": 1.53, + "learning_rate": 1.4914231735331375e-05, + "loss": 0.9334, + "step": 23800 + }, + { + "epoch": 1.53, + "learning_rate": 1.4914017046308424e-05, + "loss": 0.9249, + "step": 23801 + }, + { + "epoch": 1.53, + "learning_rate": 1.4913802357285473e-05, + "loss": 0.957, + "step": 23802 + }, + { + "epoch": 1.53, + "learning_rate": 1.4913587668262524e-05, + "loss": 0.9575, + "step": 23803 + }, + { + "epoch": 1.53, + "learning_rate": 1.4913372979239573e-05, + "loss": 1.0392, + "step": 23804 + }, + { + "epoch": 1.53, + "learning_rate": 1.4913158290216623e-05, + "loss": 0.9275, + "step": 23805 + }, + { + "epoch": 1.53, + "learning_rate": 1.4912943601193672e-05, + "loss": 1.028, + "step": 23806 + }, + { + "epoch": 1.53, + "learning_rate": 1.4912728912170721e-05, + "loss": 1.0343, + "step": 23807 + }, + { + "epoch": 1.53, + "learning_rate": 1.4912514223147772e-05, + "loss": 0.9104, + "step": 23808 + }, + { + "epoch": 1.53, + "learning_rate": 1.4912299534124821e-05, + "loss": 0.9927, + "step": 23809 + }, + { + "epoch": 1.53, + "learning_rate": 1.491208484510187e-05, + "loss": 1.0321, + "step": 23810 + }, + { + "epoch": 1.53, + "learning_rate": 1.491187015607892e-05, + "loss": 1.0135, + "step": 23811 + }, + { + "epoch": 1.53, + "learning_rate": 1.491165546705597e-05, + "loss": 0.973, + "step": 23812 + }, + { + "epoch": 1.53, + "learning_rate": 1.491144077803302e-05, + "loss": 0.9998, + "step": 23813 + }, + { + "epoch": 1.53, + "learning_rate": 1.491122608901007e-05, + "loss": 1.0688, + "step": 23814 + }, + { + "epoch": 1.53, + "learning_rate": 1.4911011399987122e-05, + "loss": 0.9835, + "step": 23815 + }, + { + "epoch": 1.53, + "learning_rate": 1.491079671096417e-05, + "loss": 0.9111, + "step": 23816 + }, + { + "epoch": 1.53, + "learning_rate": 1.491058202194122e-05, + "loss": 0.9269, + "step": 23817 + }, + { + "epoch": 1.53, + "learning_rate": 1.491036733291827e-05, + "loss": 1.012, + "step": 23818 + }, + { + "epoch": 1.53, + "learning_rate": 1.491015264389532e-05, + "loss": 0.9343, + "step": 23819 + }, + { + "epoch": 1.53, + "learning_rate": 1.4909937954872368e-05, + "loss": 1.0608, + "step": 23820 + }, + { + "epoch": 1.53, + "learning_rate": 1.4909723265849419e-05, + "loss": 0.9472, + "step": 23821 + }, + { + "epoch": 1.53, + "learning_rate": 1.4909508576826468e-05, + "loss": 0.9938, + "step": 23822 + }, + { + "epoch": 1.53, + "learning_rate": 1.4909293887803518e-05, + "loss": 0.9824, + "step": 23823 + }, + { + "epoch": 1.53, + "learning_rate": 1.4909079198780567e-05, + "loss": 0.9284, + "step": 23824 + }, + { + "epoch": 1.53, + "learning_rate": 1.4908864509757616e-05, + "loss": 0.9641, + "step": 23825 + }, + { + "epoch": 1.53, + "learning_rate": 1.4908649820734667e-05, + "loss": 1.0558, + "step": 23826 + }, + { + "epoch": 1.53, + "learning_rate": 1.4908435131711716e-05, + "loss": 1.0069, + "step": 23827 + }, + { + "epoch": 1.53, + "learning_rate": 1.4908220442688765e-05, + "loss": 0.9996, + "step": 23828 + }, + { + "epoch": 1.53, + "learning_rate": 1.4908005753665817e-05, + "loss": 1.0158, + "step": 23829 + }, + { + "epoch": 1.53, + "learning_rate": 1.4907791064642866e-05, + "loss": 0.9841, + "step": 23830 + }, + { + "epoch": 1.53, + "learning_rate": 1.4907576375619917e-05, + "loss": 0.9834, + "step": 23831 + }, + { + "epoch": 1.53, + "learning_rate": 1.4907361686596966e-05, + "loss": 0.8819, + "step": 23832 + }, + { + "epoch": 1.53, + "learning_rate": 1.4907146997574017e-05, + "loss": 1.0282, + "step": 23833 + }, + { + "epoch": 1.53, + "learning_rate": 1.4906932308551065e-05, + "loss": 0.971, + "step": 23834 + }, + { + "epoch": 1.53, + "learning_rate": 1.4906717619528114e-05, + "loss": 0.9356, + "step": 23835 + }, + { + "epoch": 1.53, + "learning_rate": 1.4906502930505165e-05, + "loss": 0.9484, + "step": 23836 + }, + { + "epoch": 1.53, + "learning_rate": 1.4906288241482214e-05, + "loss": 0.9444, + "step": 23837 + }, + { + "epoch": 1.53, + "learning_rate": 1.4906073552459263e-05, + "loss": 1.0286, + "step": 23838 + }, + { + "epoch": 1.53, + "learning_rate": 1.4905858863436314e-05, + "loss": 1.0355, + "step": 23839 + }, + { + "epoch": 1.53, + "learning_rate": 1.4905644174413363e-05, + "loss": 1.0096, + "step": 23840 + }, + { + "epoch": 1.53, + "learning_rate": 1.4905429485390412e-05, + "loss": 0.9758, + "step": 23841 + }, + { + "epoch": 1.53, + "learning_rate": 1.4905214796367462e-05, + "loss": 1.0689, + "step": 23842 + }, + { + "epoch": 1.53, + "learning_rate": 1.4905000107344511e-05, + "loss": 0.9101, + "step": 23843 + }, + { + "epoch": 1.53, + "learning_rate": 1.4904785418321564e-05, + "loss": 0.9712, + "step": 23844 + }, + { + "epoch": 1.53, + "learning_rate": 1.4904570729298613e-05, + "loss": 1.0111, + "step": 23845 + }, + { + "epoch": 1.53, + "learning_rate": 1.4904356040275663e-05, + "loss": 1.0211, + "step": 23846 + }, + { + "epoch": 1.53, + "learning_rate": 1.4904141351252712e-05, + "loss": 1.0523, + "step": 23847 + }, + { + "epoch": 1.53, + "learning_rate": 1.4903926662229761e-05, + "loss": 0.9712, + "step": 23848 + }, + { + "epoch": 1.53, + "learning_rate": 1.4903711973206812e-05, + "loss": 0.9472, + "step": 23849 + }, + { + "epoch": 1.53, + "learning_rate": 1.490349728418386e-05, + "loss": 0.9366, + "step": 23850 + }, + { + "epoch": 1.53, + "learning_rate": 1.490328259516091e-05, + "loss": 0.9595, + "step": 23851 + }, + { + "epoch": 1.53, + "learning_rate": 1.490306790613796e-05, + "loss": 1.0216, + "step": 23852 + }, + { + "epoch": 1.53, + "learning_rate": 1.490285321711501e-05, + "loss": 1.0197, + "step": 23853 + }, + { + "epoch": 1.53, + "learning_rate": 1.490263852809206e-05, + "loss": 0.9085, + "step": 23854 + }, + { + "epoch": 1.53, + "learning_rate": 1.4902423839069109e-05, + "loss": 0.9177, + "step": 23855 + }, + { + "epoch": 1.53, + "learning_rate": 1.4902209150046158e-05, + "loss": 0.954, + "step": 23856 + }, + { + "epoch": 1.53, + "learning_rate": 1.4901994461023209e-05, + "loss": 1.0363, + "step": 23857 + }, + { + "epoch": 1.53, + "learning_rate": 1.490177977200026e-05, + "loss": 0.9749, + "step": 23858 + }, + { + "epoch": 1.54, + "learning_rate": 1.490156508297731e-05, + "loss": 0.9444, + "step": 23859 + }, + { + "epoch": 1.54, + "learning_rate": 1.4901350393954359e-05, + "loss": 0.9724, + "step": 23860 + }, + { + "epoch": 1.54, + "learning_rate": 1.4901135704931408e-05, + "loss": 1.0185, + "step": 23861 + }, + { + "epoch": 1.54, + "learning_rate": 1.4900921015908459e-05, + "loss": 0.9711, + "step": 23862 + }, + { + "epoch": 1.54, + "learning_rate": 1.4900706326885507e-05, + "loss": 0.9767, + "step": 23863 + }, + { + "epoch": 1.54, + "learning_rate": 1.4900491637862558e-05, + "loss": 1.0033, + "step": 23864 + }, + { + "epoch": 1.54, + "learning_rate": 1.4900276948839607e-05, + "loss": 0.9784, + "step": 23865 + }, + { + "epoch": 1.54, + "learning_rate": 1.4900062259816656e-05, + "loss": 0.9733, + "step": 23866 + }, + { + "epoch": 1.54, + "learning_rate": 1.4899847570793707e-05, + "loss": 0.9701, + "step": 23867 + }, + { + "epoch": 1.54, + "learning_rate": 1.4899632881770756e-05, + "loss": 0.9452, + "step": 23868 + }, + { + "epoch": 1.54, + "learning_rate": 1.4899418192747805e-05, + "loss": 1.0505, + "step": 23869 + }, + { + "epoch": 1.54, + "learning_rate": 1.4899203503724855e-05, + "loss": 0.9822, + "step": 23870 + }, + { + "epoch": 1.54, + "learning_rate": 1.4898988814701904e-05, + "loss": 0.9457, + "step": 23871 + }, + { + "epoch": 1.54, + "learning_rate": 1.4898774125678955e-05, + "loss": 0.9922, + "step": 23872 + }, + { + "epoch": 1.54, + "learning_rate": 1.4898559436656006e-05, + "loss": 0.9835, + "step": 23873 + }, + { + "epoch": 1.54, + "learning_rate": 1.4898344747633056e-05, + "loss": 0.9406, + "step": 23874 + }, + { + "epoch": 1.54, + "learning_rate": 1.4898130058610105e-05, + "loss": 1.0194, + "step": 23875 + }, + { + "epoch": 1.54, + "learning_rate": 1.4897915369587154e-05, + "loss": 0.9538, + "step": 23876 + }, + { + "epoch": 1.54, + "learning_rate": 1.4897700680564205e-05, + "loss": 0.9526, + "step": 23877 + }, + { + "epoch": 1.54, + "learning_rate": 1.4897485991541254e-05, + "loss": 0.9938, + "step": 23878 + }, + { + "epoch": 1.54, + "learning_rate": 1.4897271302518303e-05, + "loss": 0.9656, + "step": 23879 + }, + { + "epoch": 1.54, + "learning_rate": 1.4897056613495353e-05, + "loss": 0.993, + "step": 23880 + }, + { + "epoch": 1.54, + "learning_rate": 1.4896841924472402e-05, + "loss": 1.0248, + "step": 23881 + }, + { + "epoch": 1.54, + "learning_rate": 1.4896627235449453e-05, + "loss": 0.9694, + "step": 23882 + }, + { + "epoch": 1.54, + "learning_rate": 1.4896412546426502e-05, + "loss": 0.9018, + "step": 23883 + }, + { + "epoch": 1.54, + "learning_rate": 1.4896197857403551e-05, + "loss": 0.9714, + "step": 23884 + }, + { + "epoch": 1.54, + "learning_rate": 1.4895983168380602e-05, + "loss": 0.9398, + "step": 23885 + }, + { + "epoch": 1.54, + "learning_rate": 1.489576847935765e-05, + "loss": 0.9861, + "step": 23886 + }, + { + "epoch": 1.54, + "learning_rate": 1.4895553790334703e-05, + "loss": 0.9044, + "step": 23887 + }, + { + "epoch": 1.54, + "learning_rate": 1.4895339101311752e-05, + "loss": 0.9373, + "step": 23888 + }, + { + "epoch": 1.54, + "learning_rate": 1.4895124412288801e-05, + "loss": 0.947, + "step": 23889 + }, + { + "epoch": 1.54, + "learning_rate": 1.4894909723265852e-05, + "loss": 0.9865, + "step": 23890 + }, + { + "epoch": 1.54, + "learning_rate": 1.48946950342429e-05, + "loss": 0.9716, + "step": 23891 + }, + { + "epoch": 1.54, + "learning_rate": 1.4894480345219951e-05, + "loss": 1.0019, + "step": 23892 + }, + { + "epoch": 1.54, + "learning_rate": 1.4894265656197e-05, + "loss": 0.8962, + "step": 23893 + }, + { + "epoch": 1.54, + "learning_rate": 1.4894050967174049e-05, + "loss": 1.0257, + "step": 23894 + }, + { + "epoch": 1.54, + "learning_rate": 1.48938362781511e-05, + "loss": 0.972, + "step": 23895 + }, + { + "epoch": 1.54, + "learning_rate": 1.4893621589128149e-05, + "loss": 0.9792, + "step": 23896 + }, + { + "epoch": 1.54, + "learning_rate": 1.4893406900105198e-05, + "loss": 0.999, + "step": 23897 + }, + { + "epoch": 1.54, + "learning_rate": 1.4893192211082248e-05, + "loss": 0.9445, + "step": 23898 + }, + { + "epoch": 1.54, + "learning_rate": 1.4892977522059297e-05, + "loss": 1.0009, + "step": 23899 + }, + { + "epoch": 1.54, + "learning_rate": 1.4892762833036348e-05, + "loss": 0.9758, + "step": 23900 + }, + { + "epoch": 1.54, + "learning_rate": 1.4892548144013397e-05, + "loss": 0.9239, + "step": 23901 + }, + { + "epoch": 1.54, + "learning_rate": 1.489233345499045e-05, + "loss": 0.9999, + "step": 23902 + }, + { + "epoch": 1.54, + "learning_rate": 1.4892118765967498e-05, + "loss": 1.0075, + "step": 23903 + }, + { + "epoch": 1.54, + "learning_rate": 1.4891904076944547e-05, + "loss": 0.9394, + "step": 23904 + }, + { + "epoch": 1.54, + "learning_rate": 1.4891689387921598e-05, + "loss": 1.0388, + "step": 23905 + }, + { + "epoch": 1.54, + "learning_rate": 1.4891474698898647e-05, + "loss": 0.9399, + "step": 23906 + }, + { + "epoch": 1.54, + "learning_rate": 1.4891260009875696e-05, + "loss": 0.9295, + "step": 23907 + }, + { + "epoch": 1.54, + "learning_rate": 1.4891045320852746e-05, + "loss": 0.8754, + "step": 23908 + }, + { + "epoch": 1.54, + "learning_rate": 1.4890830631829795e-05, + "loss": 0.9898, + "step": 23909 + }, + { + "epoch": 1.54, + "learning_rate": 1.4890615942806846e-05, + "loss": 1.0141, + "step": 23910 + }, + { + "epoch": 1.54, + "learning_rate": 1.4890401253783895e-05, + "loss": 0.8799, + "step": 23911 + }, + { + "epoch": 1.54, + "learning_rate": 1.4890186564760944e-05, + "loss": 0.9633, + "step": 23912 + }, + { + "epoch": 1.54, + "learning_rate": 1.4889971875737995e-05, + "loss": 0.8643, + "step": 23913 + }, + { + "epoch": 1.54, + "learning_rate": 1.4889757186715044e-05, + "loss": 1.0251, + "step": 23914 + }, + { + "epoch": 1.54, + "learning_rate": 1.4889542497692093e-05, + "loss": 1.0339, + "step": 23915 + }, + { + "epoch": 1.54, + "learning_rate": 1.4889327808669145e-05, + "loss": 0.957, + "step": 23916 + }, + { + "epoch": 1.54, + "learning_rate": 1.4889113119646194e-05, + "loss": 1.0028, + "step": 23917 + }, + { + "epoch": 1.54, + "learning_rate": 1.4888898430623245e-05, + "loss": 0.9159, + "step": 23918 + }, + { + "epoch": 1.54, + "learning_rate": 1.4888683741600294e-05, + "loss": 0.9501, + "step": 23919 + }, + { + "epoch": 1.54, + "learning_rate": 1.4888469052577344e-05, + "loss": 0.9339, + "step": 23920 + }, + { + "epoch": 1.54, + "learning_rate": 1.4888254363554393e-05, + "loss": 0.9569, + "step": 23921 + }, + { + "epoch": 1.54, + "learning_rate": 1.4888039674531442e-05, + "loss": 0.9972, + "step": 23922 + }, + { + "epoch": 1.54, + "learning_rate": 1.4887824985508493e-05, + "loss": 1.0081, + "step": 23923 + }, + { + "epoch": 1.54, + "learning_rate": 1.4887610296485542e-05, + "loss": 0.9588, + "step": 23924 + }, + { + "epoch": 1.54, + "learning_rate": 1.488739560746259e-05, + "loss": 0.9746, + "step": 23925 + }, + { + "epoch": 1.54, + "learning_rate": 1.4887180918439641e-05, + "loss": 0.9737, + "step": 23926 + }, + { + "epoch": 1.54, + "learning_rate": 1.488696622941669e-05, + "loss": 0.9742, + "step": 23927 + }, + { + "epoch": 1.54, + "learning_rate": 1.488675154039374e-05, + "loss": 0.8969, + "step": 23928 + }, + { + "epoch": 1.54, + "learning_rate": 1.488653685137079e-05, + "loss": 0.981, + "step": 23929 + }, + { + "epoch": 1.54, + "learning_rate": 1.4886322162347839e-05, + "loss": 0.9617, + "step": 23930 + }, + { + "epoch": 1.54, + "learning_rate": 1.4886107473324891e-05, + "loss": 0.9267, + "step": 23931 + }, + { + "epoch": 1.54, + "learning_rate": 1.488589278430194e-05, + "loss": 0.9739, + "step": 23932 + }, + { + "epoch": 1.54, + "learning_rate": 1.4885678095278991e-05, + "loss": 0.96, + "step": 23933 + }, + { + "epoch": 1.54, + "learning_rate": 1.488546340625604e-05, + "loss": 0.9548, + "step": 23934 + }, + { + "epoch": 1.54, + "learning_rate": 1.4885248717233089e-05, + "loss": 0.9394, + "step": 23935 + }, + { + "epoch": 1.54, + "learning_rate": 1.488503402821014e-05, + "loss": 0.9825, + "step": 23936 + }, + { + "epoch": 1.54, + "learning_rate": 1.4884819339187189e-05, + "loss": 0.9804, + "step": 23937 + }, + { + "epoch": 1.54, + "learning_rate": 1.4884604650164237e-05, + "loss": 1.0318, + "step": 23938 + }, + { + "epoch": 1.54, + "learning_rate": 1.4884389961141288e-05, + "loss": 1.0209, + "step": 23939 + }, + { + "epoch": 1.54, + "learning_rate": 1.4884175272118337e-05, + "loss": 0.9987, + "step": 23940 + }, + { + "epoch": 1.54, + "learning_rate": 1.4883960583095388e-05, + "loss": 1.0539, + "step": 23941 + }, + { + "epoch": 1.54, + "learning_rate": 1.4883745894072437e-05, + "loss": 0.9355, + "step": 23942 + }, + { + "epoch": 1.54, + "learning_rate": 1.4883531205049486e-05, + "loss": 0.9741, + "step": 23943 + }, + { + "epoch": 1.54, + "learning_rate": 1.4883316516026536e-05, + "loss": 0.9609, + "step": 23944 + }, + { + "epoch": 1.54, + "learning_rate": 1.4883101827003587e-05, + "loss": 0.8891, + "step": 23945 + }, + { + "epoch": 1.54, + "learning_rate": 1.4882887137980638e-05, + "loss": 1.0206, + "step": 23946 + }, + { + "epoch": 1.54, + "learning_rate": 1.4882672448957687e-05, + "loss": 0.9986, + "step": 23947 + }, + { + "epoch": 1.54, + "learning_rate": 1.4882457759934736e-05, + "loss": 0.9438, + "step": 23948 + }, + { + "epoch": 1.54, + "learning_rate": 1.4882243070911786e-05, + "loss": 0.9296, + "step": 23949 + }, + { + "epoch": 1.54, + "learning_rate": 1.4882028381888835e-05, + "loss": 0.9153, + "step": 23950 + }, + { + "epoch": 1.54, + "learning_rate": 1.4881813692865886e-05, + "loss": 0.959, + "step": 23951 + }, + { + "epoch": 1.54, + "learning_rate": 1.4881599003842935e-05, + "loss": 0.9671, + "step": 23952 + }, + { + "epoch": 1.54, + "learning_rate": 1.4881384314819984e-05, + "loss": 0.9848, + "step": 23953 + }, + { + "epoch": 1.54, + "learning_rate": 1.4881169625797034e-05, + "loss": 0.9603, + "step": 23954 + }, + { + "epoch": 1.54, + "learning_rate": 1.4880954936774083e-05, + "loss": 0.8728, + "step": 23955 + }, + { + "epoch": 1.54, + "learning_rate": 1.4880740247751132e-05, + "loss": 1.0557, + "step": 23956 + }, + { + "epoch": 1.54, + "learning_rate": 1.4880525558728183e-05, + "loss": 0.9812, + "step": 23957 + }, + { + "epoch": 1.54, + "learning_rate": 1.4880310869705232e-05, + "loss": 0.9493, + "step": 23958 + }, + { + "epoch": 1.54, + "learning_rate": 1.4880096180682283e-05, + "loss": 0.9937, + "step": 23959 + }, + { + "epoch": 1.54, + "learning_rate": 1.4879881491659333e-05, + "loss": 1.0206, + "step": 23960 + }, + { + "epoch": 1.54, + "learning_rate": 1.4879666802636384e-05, + "loss": 0.9223, + "step": 23961 + }, + { + "epoch": 1.54, + "learning_rate": 1.4879452113613433e-05, + "loss": 0.9716, + "step": 23962 + }, + { + "epoch": 1.54, + "learning_rate": 1.4879237424590482e-05, + "loss": 0.935, + "step": 23963 + }, + { + "epoch": 1.54, + "learning_rate": 1.4879022735567533e-05, + "loss": 0.9549, + "step": 23964 + }, + { + "epoch": 1.54, + "learning_rate": 1.4878808046544582e-05, + "loss": 1.0173, + "step": 23965 + }, + { + "epoch": 1.54, + "learning_rate": 1.487859335752163e-05, + "loss": 1.0396, + "step": 23966 + }, + { + "epoch": 1.54, + "learning_rate": 1.4878378668498681e-05, + "loss": 0.936, + "step": 23967 + }, + { + "epoch": 1.54, + "learning_rate": 1.487816397947573e-05, + "loss": 0.9344, + "step": 23968 + }, + { + "epoch": 1.54, + "learning_rate": 1.487794929045278e-05, + "loss": 0.9738, + "step": 23969 + }, + { + "epoch": 1.54, + "learning_rate": 1.487773460142983e-05, + "loss": 1.0633, + "step": 23970 + }, + { + "epoch": 1.54, + "learning_rate": 1.4877519912406879e-05, + "loss": 0.9545, + "step": 23971 + }, + { + "epoch": 1.54, + "learning_rate": 1.487730522338393e-05, + "loss": 0.9901, + "step": 23972 + }, + { + "epoch": 1.54, + "learning_rate": 1.4877090534360978e-05, + "loss": 0.9326, + "step": 23973 + }, + { + "epoch": 1.54, + "learning_rate": 1.4876875845338027e-05, + "loss": 0.9581, + "step": 23974 + }, + { + "epoch": 1.54, + "learning_rate": 1.487666115631508e-05, + "loss": 0.9821, + "step": 23975 + }, + { + "epoch": 1.54, + "learning_rate": 1.4876446467292129e-05, + "loss": 1.0028, + "step": 23976 + }, + { + "epoch": 1.54, + "learning_rate": 1.487623177826918e-05, + "loss": 0.9602, + "step": 23977 + }, + { + "epoch": 1.54, + "learning_rate": 1.4876017089246228e-05, + "loss": 1.0037, + "step": 23978 + }, + { + "epoch": 1.54, + "learning_rate": 1.4875802400223279e-05, + "loss": 0.9543, + "step": 23979 + }, + { + "epoch": 1.54, + "learning_rate": 1.4875587711200328e-05, + "loss": 0.9412, + "step": 23980 + }, + { + "epoch": 1.54, + "learning_rate": 1.4875373022177377e-05, + "loss": 1.026, + "step": 23981 + }, + { + "epoch": 1.54, + "learning_rate": 1.4875158333154427e-05, + "loss": 1.0082, + "step": 23982 + }, + { + "epoch": 1.54, + "learning_rate": 1.4874943644131476e-05, + "loss": 0.9313, + "step": 23983 + }, + { + "epoch": 1.54, + "learning_rate": 1.4874728955108525e-05, + "loss": 0.9877, + "step": 23984 + }, + { + "epoch": 1.54, + "learning_rate": 1.4874514266085576e-05, + "loss": 0.9431, + "step": 23985 + }, + { + "epoch": 1.54, + "learning_rate": 1.4874299577062625e-05, + "loss": 1.0232, + "step": 23986 + }, + { + "epoch": 1.54, + "learning_rate": 1.4874084888039676e-05, + "loss": 0.9531, + "step": 23987 + }, + { + "epoch": 1.54, + "learning_rate": 1.4873870199016725e-05, + "loss": 0.9627, + "step": 23988 + }, + { + "epoch": 1.54, + "learning_rate": 1.4873655509993777e-05, + "loss": 0.9495, + "step": 23989 + }, + { + "epoch": 1.54, + "learning_rate": 1.4873440820970826e-05, + "loss": 1.0864, + "step": 23990 + }, + { + "epoch": 1.54, + "learning_rate": 1.4873226131947875e-05, + "loss": 0.9422, + "step": 23991 + }, + { + "epoch": 1.54, + "learning_rate": 1.4873011442924926e-05, + "loss": 0.9681, + "step": 23992 + }, + { + "epoch": 1.54, + "learning_rate": 1.4872796753901975e-05, + "loss": 1.0293, + "step": 23993 + }, + { + "epoch": 1.54, + "learning_rate": 1.4872582064879024e-05, + "loss": 0.9727, + "step": 23994 + }, + { + "epoch": 1.54, + "learning_rate": 1.4872367375856074e-05, + "loss": 0.9045, + "step": 23995 + }, + { + "epoch": 1.54, + "learning_rate": 1.4872152686833123e-05, + "loss": 0.9314, + "step": 23996 + }, + { + "epoch": 1.54, + "learning_rate": 1.4871937997810174e-05, + "loss": 0.9348, + "step": 23997 + }, + { + "epoch": 1.54, + "learning_rate": 1.4871723308787223e-05, + "loss": 0.9383, + "step": 23998 + }, + { + "epoch": 1.54, + "learning_rate": 1.4871508619764272e-05, + "loss": 0.9769, + "step": 23999 + }, + { + "epoch": 1.54, + "learning_rate": 1.4871293930741322e-05, + "loss": 1.0345, + "step": 24000 + }, + { + "epoch": 1.54, + "eval_laion/cider": 0.5399276055632253, + "eval_laion/meteor": 0.3853825023478007, + "eval_laion/pred_num_words": 35.30859375, + "eval_laion/sacrebleu": 15.450505103042543, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0495561361312866, + "eval_runtime": 94.0441, + "eval_samples_per_second": 5.306, + "eval_steps_per_second": 0.34, + "step": 24000 + }, + { + "epoch": 1.54, + "learning_rate": 1.4871079241718371e-05, + "loss": 0.9156, + "step": 24001 + }, + { + "epoch": 1.54, + "learning_rate": 1.487086455269542e-05, + "loss": 0.9477, + "step": 24002 + }, + { + "epoch": 1.54, + "learning_rate": 1.4870649863672471e-05, + "loss": 1.0153, + "step": 24003 + }, + { + "epoch": 1.54, + "learning_rate": 1.4870435174649522e-05, + "loss": 1.0441, + "step": 24004 + }, + { + "epoch": 1.54, + "learning_rate": 1.4870220485626572e-05, + "loss": 1.0458, + "step": 24005 + }, + { + "epoch": 1.54, + "learning_rate": 1.4870005796603621e-05, + "loss": 0.9639, + "step": 24006 + }, + { + "epoch": 1.54, + "learning_rate": 1.4869791107580672e-05, + "loss": 0.9484, + "step": 24007 + }, + { + "epoch": 1.54, + "learning_rate": 1.4869576418557721e-05, + "loss": 0.9176, + "step": 24008 + }, + { + "epoch": 1.54, + "learning_rate": 1.486936172953477e-05, + "loss": 1.0534, + "step": 24009 + }, + { + "epoch": 1.54, + "learning_rate": 1.486914704051182e-05, + "loss": 0.952, + "step": 24010 + }, + { + "epoch": 1.54, + "learning_rate": 1.486893235148887e-05, + "loss": 0.9878, + "step": 24011 + }, + { + "epoch": 1.54, + "learning_rate": 1.4868717662465918e-05, + "loss": 0.9758, + "step": 24012 + }, + { + "epoch": 1.54, + "learning_rate": 1.4868502973442969e-05, + "loss": 1.028, + "step": 24013 + }, + { + "epoch": 1.55, + "learning_rate": 1.4868288284420018e-05, + "loss": 1.0789, + "step": 24014 + }, + { + "epoch": 1.55, + "learning_rate": 1.4868073595397067e-05, + "loss": 1.0003, + "step": 24015 + }, + { + "epoch": 1.55, + "learning_rate": 1.4867858906374118e-05, + "loss": 0.9315, + "step": 24016 + }, + { + "epoch": 1.55, + "learning_rate": 1.4867644217351167e-05, + "loss": 0.9452, + "step": 24017 + }, + { + "epoch": 1.55, + "learning_rate": 1.4867429528328219e-05, + "loss": 1.0718, + "step": 24018 + }, + { + "epoch": 1.55, + "learning_rate": 1.4867214839305268e-05, + "loss": 1.0502, + "step": 24019 + }, + { + "epoch": 1.55, + "learning_rate": 1.4867000150282319e-05, + "loss": 1.0265, + "step": 24020 + }, + { + "epoch": 1.55, + "learning_rate": 1.4866785461259368e-05, + "loss": 0.9218, + "step": 24021 + }, + { + "epoch": 1.55, + "learning_rate": 1.4866570772236417e-05, + "loss": 0.9555, + "step": 24022 + }, + { + "epoch": 1.55, + "learning_rate": 1.4866356083213467e-05, + "loss": 0.9092, + "step": 24023 + }, + { + "epoch": 1.55, + "learning_rate": 1.4866141394190516e-05, + "loss": 0.9724, + "step": 24024 + }, + { + "epoch": 1.55, + "learning_rate": 1.4865926705167565e-05, + "loss": 0.985, + "step": 24025 + }, + { + "epoch": 1.55, + "learning_rate": 1.4865712016144616e-05, + "loss": 0.953, + "step": 24026 + }, + { + "epoch": 1.55, + "learning_rate": 1.4865497327121665e-05, + "loss": 0.9462, + "step": 24027 + }, + { + "epoch": 1.55, + "learning_rate": 1.4865282638098715e-05, + "loss": 0.9565, + "step": 24028 + }, + { + "epoch": 1.55, + "learning_rate": 1.4865067949075764e-05, + "loss": 0.934, + "step": 24029 + }, + { + "epoch": 1.55, + "learning_rate": 1.4864853260052813e-05, + "loss": 0.9305, + "step": 24030 + }, + { + "epoch": 1.55, + "learning_rate": 1.4864638571029864e-05, + "loss": 1.0079, + "step": 24031 + }, + { + "epoch": 1.55, + "learning_rate": 1.4864423882006913e-05, + "loss": 1.0523, + "step": 24032 + }, + { + "epoch": 1.55, + "learning_rate": 1.4864209192983965e-05, + "loss": 0.9714, + "step": 24033 + }, + { + "epoch": 1.55, + "learning_rate": 1.4863994503961014e-05, + "loss": 0.9899, + "step": 24034 + }, + { + "epoch": 1.55, + "learning_rate": 1.4863779814938063e-05, + "loss": 0.956, + "step": 24035 + }, + { + "epoch": 1.55, + "learning_rate": 1.4863565125915114e-05, + "loss": 1.0264, + "step": 24036 + }, + { + "epoch": 1.55, + "learning_rate": 1.4863350436892163e-05, + "loss": 0.9998, + "step": 24037 + }, + { + "epoch": 1.55, + "learning_rate": 1.4863135747869214e-05, + "loss": 1.0126, + "step": 24038 + }, + { + "epoch": 1.55, + "learning_rate": 1.4862921058846263e-05, + "loss": 0.9346, + "step": 24039 + }, + { + "epoch": 1.55, + "learning_rate": 1.4862706369823312e-05, + "loss": 0.9438, + "step": 24040 + }, + { + "epoch": 1.55, + "learning_rate": 1.4862491680800362e-05, + "loss": 1.0062, + "step": 24041 + }, + { + "epoch": 1.55, + "learning_rate": 1.4862276991777411e-05, + "loss": 0.9799, + "step": 24042 + }, + { + "epoch": 1.55, + "learning_rate": 1.486206230275446e-05, + "loss": 1.0325, + "step": 24043 + }, + { + "epoch": 1.55, + "learning_rate": 1.486184761373151e-05, + "loss": 0.9947, + "step": 24044 + }, + { + "epoch": 1.55, + "learning_rate": 1.486163292470856e-05, + "loss": 0.9048, + "step": 24045 + }, + { + "epoch": 1.55, + "learning_rate": 1.486141823568561e-05, + "loss": 0.9055, + "step": 24046 + }, + { + "epoch": 1.55, + "learning_rate": 1.4861203546662661e-05, + "loss": 1.0059, + "step": 24047 + }, + { + "epoch": 1.55, + "learning_rate": 1.4860988857639712e-05, + "loss": 0.979, + "step": 24048 + }, + { + "epoch": 1.55, + "learning_rate": 1.486077416861676e-05, + "loss": 1.0871, + "step": 24049 + }, + { + "epoch": 1.55, + "learning_rate": 1.486055947959381e-05, + "loss": 0.9478, + "step": 24050 + }, + { + "epoch": 1.55, + "learning_rate": 1.486034479057086e-05, + "loss": 1.0586, + "step": 24051 + }, + { + "epoch": 1.55, + "learning_rate": 1.486013010154791e-05, + "loss": 0.897, + "step": 24052 + }, + { + "epoch": 1.55, + "learning_rate": 1.4859915412524958e-05, + "loss": 0.9789, + "step": 24053 + }, + { + "epoch": 1.55, + "learning_rate": 1.4859700723502009e-05, + "loss": 1.005, + "step": 24054 + }, + { + "epoch": 1.55, + "learning_rate": 1.4859486034479058e-05, + "loss": 0.9458, + "step": 24055 + }, + { + "epoch": 1.55, + "learning_rate": 1.4859271345456109e-05, + "loss": 1.0128, + "step": 24056 + }, + { + "epoch": 1.55, + "learning_rate": 1.4859056656433157e-05, + "loss": 0.9328, + "step": 24057 + }, + { + "epoch": 1.55, + "learning_rate": 1.4858841967410206e-05, + "loss": 0.9404, + "step": 24058 + }, + { + "epoch": 1.55, + "learning_rate": 1.4858627278387257e-05, + "loss": 0.9927, + "step": 24059 + }, + { + "epoch": 1.55, + "learning_rate": 1.4858412589364306e-05, + "loss": 0.969, + "step": 24060 + }, + { + "epoch": 1.55, + "learning_rate": 1.4858197900341355e-05, + "loss": 0.9966, + "step": 24061 + }, + { + "epoch": 1.55, + "learning_rate": 1.4857983211318407e-05, + "loss": 0.9422, + "step": 24062 + }, + { + "epoch": 1.55, + "learning_rate": 1.4857768522295456e-05, + "loss": 0.9285, + "step": 24063 + }, + { + "epoch": 1.55, + "learning_rate": 1.4857553833272507e-05, + "loss": 1.0049, + "step": 24064 + }, + { + "epoch": 1.55, + "learning_rate": 1.4857339144249556e-05, + "loss": 0.9903, + "step": 24065 + }, + { + "epoch": 1.55, + "learning_rate": 1.4857124455226607e-05, + "loss": 0.8906, + "step": 24066 + }, + { + "epoch": 1.55, + "learning_rate": 1.4856909766203656e-05, + "loss": 0.9608, + "step": 24067 + }, + { + "epoch": 1.55, + "learning_rate": 1.4856695077180705e-05, + "loss": 0.9214, + "step": 24068 + }, + { + "epoch": 1.55, + "learning_rate": 1.4856480388157755e-05, + "loss": 0.9604, + "step": 24069 + }, + { + "epoch": 1.55, + "learning_rate": 1.4856265699134804e-05, + "loss": 1.0328, + "step": 24070 + }, + { + "epoch": 1.55, + "learning_rate": 1.4856051010111853e-05, + "loss": 1.0225, + "step": 24071 + }, + { + "epoch": 1.55, + "learning_rate": 1.4855836321088904e-05, + "loss": 0.93, + "step": 24072 + }, + { + "epoch": 1.55, + "learning_rate": 1.4855621632065953e-05, + "loss": 0.9518, + "step": 24073 + }, + { + "epoch": 1.55, + "learning_rate": 1.4855406943043002e-05, + "loss": 0.9798, + "step": 24074 + }, + { + "epoch": 1.55, + "learning_rate": 1.4855192254020052e-05, + "loss": 0.9809, + "step": 24075 + }, + { + "epoch": 1.55, + "learning_rate": 1.4854977564997105e-05, + "loss": 1.0129, + "step": 24076 + }, + { + "epoch": 1.55, + "learning_rate": 1.4854762875974154e-05, + "loss": 1.0145, + "step": 24077 + }, + { + "epoch": 1.55, + "learning_rate": 1.4854548186951203e-05, + "loss": 0.9783, + "step": 24078 + }, + { + "epoch": 1.55, + "learning_rate": 1.4854333497928253e-05, + "loss": 1.0072, + "step": 24079 + }, + { + "epoch": 1.55, + "learning_rate": 1.4854118808905302e-05, + "loss": 0.9907, + "step": 24080 + }, + { + "epoch": 1.55, + "learning_rate": 1.4853904119882351e-05, + "loss": 1.0881, + "step": 24081 + }, + { + "epoch": 1.55, + "learning_rate": 1.4853689430859402e-05, + "loss": 1.0007, + "step": 24082 + }, + { + "epoch": 1.55, + "learning_rate": 1.4853474741836451e-05, + "loss": 0.9473, + "step": 24083 + }, + { + "epoch": 1.55, + "learning_rate": 1.48532600528135e-05, + "loss": 0.9881, + "step": 24084 + }, + { + "epoch": 1.55, + "learning_rate": 1.485304536379055e-05, + "loss": 0.9424, + "step": 24085 + }, + { + "epoch": 1.55, + "learning_rate": 1.48528306747676e-05, + "loss": 0.9195, + "step": 24086 + }, + { + "epoch": 1.55, + "learning_rate": 1.485261598574465e-05, + "loss": 1.0112, + "step": 24087 + }, + { + "epoch": 1.55, + "learning_rate": 1.4852401296721699e-05, + "loss": 0.9447, + "step": 24088 + }, + { + "epoch": 1.55, + "learning_rate": 1.4852186607698748e-05, + "loss": 1.0047, + "step": 24089 + }, + { + "epoch": 1.55, + "learning_rate": 1.4851971918675799e-05, + "loss": 0.9591, + "step": 24090 + }, + { + "epoch": 1.55, + "learning_rate": 1.485175722965285e-05, + "loss": 0.9174, + "step": 24091 + }, + { + "epoch": 1.55, + "learning_rate": 1.48515425406299e-05, + "loss": 0.9248, + "step": 24092 + }, + { + "epoch": 1.55, + "learning_rate": 1.4851327851606949e-05, + "loss": 0.9748, + "step": 24093 + }, + { + "epoch": 1.55, + "learning_rate": 1.4851113162584e-05, + "loss": 0.9757, + "step": 24094 + }, + { + "epoch": 1.55, + "learning_rate": 1.4850898473561049e-05, + "loss": 0.9672, + "step": 24095 + }, + { + "epoch": 1.55, + "learning_rate": 1.4850683784538098e-05, + "loss": 0.9876, + "step": 24096 + }, + { + "epoch": 1.55, + "learning_rate": 1.4850469095515148e-05, + "loss": 0.9778, + "step": 24097 + }, + { + "epoch": 1.55, + "learning_rate": 1.4850254406492197e-05, + "loss": 1.0226, + "step": 24098 + }, + { + "epoch": 1.55, + "learning_rate": 1.4850039717469246e-05, + "loss": 0.9643, + "step": 24099 + }, + { + "epoch": 1.55, + "learning_rate": 1.4849825028446297e-05, + "loss": 0.9271, + "step": 24100 + }, + { + "epoch": 1.55, + "learning_rate": 1.4849610339423346e-05, + "loss": 0.9596, + "step": 24101 + }, + { + "epoch": 1.55, + "learning_rate": 1.4849395650400395e-05, + "loss": 0.9664, + "step": 24102 + }, + { + "epoch": 1.55, + "learning_rate": 1.4849180961377445e-05, + "loss": 1.0088, + "step": 24103 + }, + { + "epoch": 1.55, + "learning_rate": 1.4848966272354494e-05, + "loss": 0.8831, + "step": 24104 + }, + { + "epoch": 1.55, + "learning_rate": 1.4848751583331547e-05, + "loss": 1.0037, + "step": 24105 + }, + { + "epoch": 1.55, + "learning_rate": 1.4848536894308596e-05, + "loss": 0.8655, + "step": 24106 + }, + { + "epoch": 1.55, + "learning_rate": 1.4848322205285646e-05, + "loss": 0.9661, + "step": 24107 + }, + { + "epoch": 1.55, + "learning_rate": 1.4848107516262695e-05, + "loss": 0.9185, + "step": 24108 + }, + { + "epoch": 1.55, + "learning_rate": 1.4847892827239744e-05, + "loss": 0.9869, + "step": 24109 + }, + { + "epoch": 1.55, + "learning_rate": 1.4847678138216795e-05, + "loss": 0.9604, + "step": 24110 + }, + { + "epoch": 1.55, + "learning_rate": 1.4847463449193844e-05, + "loss": 0.9858, + "step": 24111 + }, + { + "epoch": 1.55, + "learning_rate": 1.4847248760170893e-05, + "loss": 0.9293, + "step": 24112 + }, + { + "epoch": 1.55, + "learning_rate": 1.4847034071147944e-05, + "loss": 0.934, + "step": 24113 + }, + { + "epoch": 1.55, + "learning_rate": 1.4846819382124993e-05, + "loss": 1.0251, + "step": 24114 + }, + { + "epoch": 1.55, + "learning_rate": 1.4846604693102043e-05, + "loss": 0.9991, + "step": 24115 + }, + { + "epoch": 1.55, + "learning_rate": 1.4846390004079092e-05, + "loss": 0.9674, + "step": 24116 + }, + { + "epoch": 1.55, + "learning_rate": 1.4846175315056141e-05, + "loss": 0.9575, + "step": 24117 + }, + { + "epoch": 1.55, + "learning_rate": 1.4845960626033192e-05, + "loss": 0.9174, + "step": 24118 + }, + { + "epoch": 1.55, + "learning_rate": 1.484574593701024e-05, + "loss": 1.0129, + "step": 24119 + }, + { + "epoch": 1.55, + "learning_rate": 1.4845531247987293e-05, + "loss": 0.99, + "step": 24120 + }, + { + "epoch": 1.55, + "learning_rate": 1.4845316558964342e-05, + "loss": 0.9208, + "step": 24121 + }, + { + "epoch": 1.55, + "learning_rate": 1.4845101869941391e-05, + "loss": 0.8509, + "step": 24122 + }, + { + "epoch": 1.55, + "learning_rate": 1.4844887180918442e-05, + "loss": 0.9424, + "step": 24123 + }, + { + "epoch": 1.55, + "learning_rate": 1.484467249189549e-05, + "loss": 1.0177, + "step": 24124 + }, + { + "epoch": 1.55, + "learning_rate": 1.4844457802872541e-05, + "loss": 0.9856, + "step": 24125 + }, + { + "epoch": 1.55, + "learning_rate": 1.484424311384959e-05, + "loss": 0.9685, + "step": 24126 + }, + { + "epoch": 1.55, + "learning_rate": 1.484402842482664e-05, + "loss": 0.978, + "step": 24127 + }, + { + "epoch": 1.55, + "learning_rate": 1.484381373580369e-05, + "loss": 0.9508, + "step": 24128 + }, + { + "epoch": 1.55, + "learning_rate": 1.4843599046780739e-05, + "loss": 0.9356, + "step": 24129 + }, + { + "epoch": 1.55, + "learning_rate": 1.4843384357757788e-05, + "loss": 0.9608, + "step": 24130 + }, + { + "epoch": 1.55, + "learning_rate": 1.4843169668734838e-05, + "loss": 0.946, + "step": 24131 + }, + { + "epoch": 1.55, + "learning_rate": 1.4842954979711887e-05, + "loss": 1.0053, + "step": 24132 + }, + { + "epoch": 1.55, + "learning_rate": 1.4842740290688938e-05, + "loss": 0.9627, + "step": 24133 + }, + { + "epoch": 1.55, + "learning_rate": 1.4842525601665989e-05, + "loss": 0.9665, + "step": 24134 + }, + { + "epoch": 1.55, + "learning_rate": 1.484231091264304e-05, + "loss": 0.9884, + "step": 24135 + }, + { + "epoch": 1.55, + "learning_rate": 1.4842096223620088e-05, + "loss": 0.9692, + "step": 24136 + }, + { + "epoch": 1.55, + "learning_rate": 1.4841881534597137e-05, + "loss": 0.9316, + "step": 24137 + }, + { + "epoch": 1.55, + "learning_rate": 1.4841666845574188e-05, + "loss": 0.9386, + "step": 24138 + }, + { + "epoch": 1.55, + "learning_rate": 1.4841452156551237e-05, + "loss": 1.0055, + "step": 24139 + }, + { + "epoch": 1.55, + "learning_rate": 1.4841237467528286e-05, + "loss": 0.95, + "step": 24140 + }, + { + "epoch": 1.55, + "learning_rate": 1.4841022778505337e-05, + "loss": 0.956, + "step": 24141 + }, + { + "epoch": 1.55, + "learning_rate": 1.4840808089482386e-05, + "loss": 0.917, + "step": 24142 + }, + { + "epoch": 1.55, + "learning_rate": 1.4840593400459436e-05, + "loss": 0.951, + "step": 24143 + }, + { + "epoch": 1.55, + "learning_rate": 1.4840378711436485e-05, + "loss": 1.0466, + "step": 24144 + }, + { + "epoch": 1.55, + "learning_rate": 1.4840164022413534e-05, + "loss": 0.9579, + "step": 24145 + }, + { + "epoch": 1.55, + "learning_rate": 1.4839949333390585e-05, + "loss": 0.9867, + "step": 24146 + }, + { + "epoch": 1.55, + "learning_rate": 1.4839734644367634e-05, + "loss": 0.999, + "step": 24147 + }, + { + "epoch": 1.55, + "learning_rate": 1.4839519955344683e-05, + "loss": 0.9847, + "step": 24148 + }, + { + "epoch": 1.55, + "learning_rate": 1.4839305266321735e-05, + "loss": 0.9867, + "step": 24149 + }, + { + "epoch": 1.55, + "learning_rate": 1.4839090577298784e-05, + "loss": 0.9402, + "step": 24150 + }, + { + "epoch": 1.55, + "learning_rate": 1.4838875888275835e-05, + "loss": 1.0116, + "step": 24151 + }, + { + "epoch": 1.55, + "learning_rate": 1.4838661199252884e-05, + "loss": 0.9807, + "step": 24152 + }, + { + "epoch": 1.55, + "learning_rate": 1.4838446510229934e-05, + "loss": 0.9999, + "step": 24153 + }, + { + "epoch": 1.55, + "learning_rate": 1.4838231821206983e-05, + "loss": 0.9582, + "step": 24154 + }, + { + "epoch": 1.55, + "learning_rate": 1.4838017132184032e-05, + "loss": 0.9083, + "step": 24155 + }, + { + "epoch": 1.55, + "learning_rate": 1.4837802443161083e-05, + "loss": 0.9861, + "step": 24156 + }, + { + "epoch": 1.55, + "learning_rate": 1.4837587754138132e-05, + "loss": 0.8615, + "step": 24157 + }, + { + "epoch": 1.55, + "learning_rate": 1.4837373065115181e-05, + "loss": 0.9825, + "step": 24158 + }, + { + "epoch": 1.55, + "learning_rate": 1.4837158376092232e-05, + "loss": 1.0445, + "step": 24159 + }, + { + "epoch": 1.55, + "learning_rate": 1.483694368706928e-05, + "loss": 0.9432, + "step": 24160 + }, + { + "epoch": 1.55, + "learning_rate": 1.483672899804633e-05, + "loss": 0.91, + "step": 24161 + }, + { + "epoch": 1.55, + "learning_rate": 1.483651430902338e-05, + "loss": 0.9896, + "step": 24162 + }, + { + "epoch": 1.55, + "learning_rate": 1.4836299620000432e-05, + "loss": 0.9947, + "step": 24163 + }, + { + "epoch": 1.55, + "learning_rate": 1.4836084930977481e-05, + "loss": 0.9254, + "step": 24164 + }, + { + "epoch": 1.55, + "learning_rate": 1.483587024195453e-05, + "loss": 1.0499, + "step": 24165 + }, + { + "epoch": 1.55, + "learning_rate": 1.4835655552931581e-05, + "loss": 0.9717, + "step": 24166 + }, + { + "epoch": 1.55, + "learning_rate": 1.483544086390863e-05, + "loss": 0.9655, + "step": 24167 + }, + { + "epoch": 1.55, + "learning_rate": 1.4835226174885679e-05, + "loss": 1.0072, + "step": 24168 + }, + { + "epoch": 1.55, + "learning_rate": 1.483501148586273e-05, + "loss": 0.9846, + "step": 24169 + }, + { + "epoch": 1.56, + "learning_rate": 1.4834796796839779e-05, + "loss": 0.9027, + "step": 24170 + }, + { + "epoch": 1.56, + "learning_rate": 1.4834582107816828e-05, + "loss": 0.9642, + "step": 24171 + }, + { + "epoch": 1.56, + "learning_rate": 1.4834367418793878e-05, + "loss": 0.9426, + "step": 24172 + }, + { + "epoch": 1.56, + "learning_rate": 1.4834152729770927e-05, + "loss": 0.9574, + "step": 24173 + }, + { + "epoch": 1.56, + "learning_rate": 1.4833938040747978e-05, + "loss": 1.0182, + "step": 24174 + }, + { + "epoch": 1.56, + "learning_rate": 1.4833723351725027e-05, + "loss": 0.9903, + "step": 24175 + }, + { + "epoch": 1.56, + "learning_rate": 1.4833508662702076e-05, + "loss": 1.0024, + "step": 24176 + }, + { + "epoch": 1.56, + "learning_rate": 1.4833293973679126e-05, + "loss": 1.1119, + "step": 24177 + }, + { + "epoch": 1.56, + "learning_rate": 1.4833079284656177e-05, + "loss": 0.9587, + "step": 24178 + }, + { + "epoch": 1.56, + "learning_rate": 1.4832864595633228e-05, + "loss": 0.9452, + "step": 24179 + }, + { + "epoch": 1.56, + "learning_rate": 1.4832649906610277e-05, + "loss": 0.9845, + "step": 24180 + }, + { + "epoch": 1.56, + "learning_rate": 1.4832435217587326e-05, + "loss": 0.9362, + "step": 24181 + }, + { + "epoch": 1.56, + "learning_rate": 1.4832220528564376e-05, + "loss": 0.9477, + "step": 24182 + }, + { + "epoch": 1.56, + "learning_rate": 1.4832005839541425e-05, + "loss": 0.921, + "step": 24183 + }, + { + "epoch": 1.56, + "learning_rate": 1.4831791150518476e-05, + "loss": 1.0666, + "step": 24184 + }, + { + "epoch": 1.56, + "learning_rate": 1.4831576461495525e-05, + "loss": 0.8879, + "step": 24185 + }, + { + "epoch": 1.56, + "learning_rate": 1.4831361772472574e-05, + "loss": 0.941, + "step": 24186 + }, + { + "epoch": 1.56, + "learning_rate": 1.4831147083449625e-05, + "loss": 0.9554, + "step": 24187 + }, + { + "epoch": 1.56, + "learning_rate": 1.4830932394426674e-05, + "loss": 0.9564, + "step": 24188 + }, + { + "epoch": 1.56, + "learning_rate": 1.4830717705403722e-05, + "loss": 0.954, + "step": 24189 + }, + { + "epoch": 1.56, + "learning_rate": 1.4830503016380773e-05, + "loss": 1.0203, + "step": 24190 + }, + { + "epoch": 1.56, + "learning_rate": 1.4830288327357822e-05, + "loss": 0.9827, + "step": 24191 + }, + { + "epoch": 1.56, + "learning_rate": 1.4830073638334874e-05, + "loss": 0.8527, + "step": 24192 + }, + { + "epoch": 1.56, + "learning_rate": 1.4829858949311923e-05, + "loss": 0.9815, + "step": 24193 + }, + { + "epoch": 1.56, + "learning_rate": 1.4829644260288974e-05, + "loss": 0.957, + "step": 24194 + }, + { + "epoch": 1.56, + "learning_rate": 1.4829429571266023e-05, + "loss": 0.9639, + "step": 24195 + }, + { + "epoch": 1.56, + "learning_rate": 1.4829214882243072e-05, + "loss": 0.9398, + "step": 24196 + }, + { + "epoch": 1.56, + "learning_rate": 1.4829000193220123e-05, + "loss": 0.9823, + "step": 24197 + }, + { + "epoch": 1.56, + "learning_rate": 1.4828785504197172e-05, + "loss": 0.9401, + "step": 24198 + }, + { + "epoch": 1.56, + "learning_rate": 1.482857081517422e-05, + "loss": 1.0207, + "step": 24199 + }, + { + "epoch": 1.56, + "learning_rate": 1.4828356126151271e-05, + "loss": 0.9923, + "step": 24200 + }, + { + "epoch": 1.56, + "learning_rate": 1.482814143712832e-05, + "loss": 0.9534, + "step": 24201 + }, + { + "epoch": 1.56, + "learning_rate": 1.4827926748105371e-05, + "loss": 1.0416, + "step": 24202 + }, + { + "epoch": 1.56, + "learning_rate": 1.482771205908242e-05, + "loss": 1.0541, + "step": 24203 + }, + { + "epoch": 1.56, + "learning_rate": 1.4827497370059469e-05, + "loss": 0.9583, + "step": 24204 + }, + { + "epoch": 1.56, + "learning_rate": 1.482728268103652e-05, + "loss": 0.979, + "step": 24205 + }, + { + "epoch": 1.56, + "learning_rate": 1.4827067992013568e-05, + "loss": 1.0075, + "step": 24206 + }, + { + "epoch": 1.56, + "learning_rate": 1.482685330299062e-05, + "loss": 1.0093, + "step": 24207 + }, + { + "epoch": 1.56, + "learning_rate": 1.482663861396767e-05, + "loss": 0.9854, + "step": 24208 + }, + { + "epoch": 1.56, + "learning_rate": 1.4826423924944719e-05, + "loss": 0.9834, + "step": 24209 + }, + { + "epoch": 1.56, + "learning_rate": 1.482620923592177e-05, + "loss": 1.0306, + "step": 24210 + }, + { + "epoch": 1.56, + "learning_rate": 1.4825994546898818e-05, + "loss": 1.0244, + "step": 24211 + }, + { + "epoch": 1.56, + "learning_rate": 1.4825779857875869e-05, + "loss": 1.0385, + "step": 24212 + }, + { + "epoch": 1.56, + "learning_rate": 1.4825565168852918e-05, + "loss": 0.9365, + "step": 24213 + }, + { + "epoch": 1.56, + "learning_rate": 1.4825350479829967e-05, + "loss": 0.9237, + "step": 24214 + }, + { + "epoch": 1.56, + "learning_rate": 1.4825135790807018e-05, + "loss": 0.9232, + "step": 24215 + }, + { + "epoch": 1.56, + "learning_rate": 1.4824921101784067e-05, + "loss": 0.9712, + "step": 24216 + }, + { + "epoch": 1.56, + "learning_rate": 1.4824706412761116e-05, + "loss": 1.015, + "step": 24217 + }, + { + "epoch": 1.56, + "learning_rate": 1.4824491723738166e-05, + "loss": 1.0041, + "step": 24218 + }, + { + "epoch": 1.56, + "learning_rate": 1.4824277034715215e-05, + "loss": 0.9192, + "step": 24219 + }, + { + "epoch": 1.56, + "learning_rate": 1.4824062345692266e-05, + "loss": 0.9845, + "step": 24220 + }, + { + "epoch": 1.56, + "learning_rate": 1.4823847656669316e-05, + "loss": 0.8582, + "step": 24221 + }, + { + "epoch": 1.56, + "learning_rate": 1.4823632967646367e-05, + "loss": 0.928, + "step": 24222 + }, + { + "epoch": 1.56, + "learning_rate": 1.4823418278623416e-05, + "loss": 0.988, + "step": 24223 + }, + { + "epoch": 1.56, + "learning_rate": 1.4823203589600465e-05, + "loss": 0.9529, + "step": 24224 + }, + { + "epoch": 1.56, + "learning_rate": 1.4822988900577516e-05, + "loss": 0.9474, + "step": 24225 + }, + { + "epoch": 1.56, + "learning_rate": 1.4822774211554565e-05, + "loss": 0.9725, + "step": 24226 + }, + { + "epoch": 1.56, + "learning_rate": 1.4822559522531614e-05, + "loss": 1.0588, + "step": 24227 + }, + { + "epoch": 1.56, + "learning_rate": 1.4822344833508664e-05, + "loss": 0.9359, + "step": 24228 + }, + { + "epoch": 1.56, + "learning_rate": 1.4822130144485713e-05, + "loss": 0.9638, + "step": 24229 + }, + { + "epoch": 1.56, + "learning_rate": 1.4821915455462764e-05, + "loss": 1.0119, + "step": 24230 + }, + { + "epoch": 1.56, + "learning_rate": 1.4821700766439813e-05, + "loss": 1.0305, + "step": 24231 + }, + { + "epoch": 1.56, + "learning_rate": 1.4821486077416862e-05, + "loss": 0.9647, + "step": 24232 + }, + { + "epoch": 1.56, + "learning_rate": 1.4821271388393913e-05, + "loss": 0.9507, + "step": 24233 + }, + { + "epoch": 1.56, + "learning_rate": 1.4821056699370961e-05, + "loss": 1.0396, + "step": 24234 + }, + { + "epoch": 1.56, + "learning_rate": 1.482084201034801e-05, + "loss": 1.0004, + "step": 24235 + }, + { + "epoch": 1.56, + "learning_rate": 1.4820627321325063e-05, + "loss": 0.9047, + "step": 24236 + }, + { + "epoch": 1.56, + "learning_rate": 1.4820412632302112e-05, + "loss": 0.9831, + "step": 24237 + }, + { + "epoch": 1.56, + "learning_rate": 1.4820197943279162e-05, + "loss": 0.9063, + "step": 24238 + }, + { + "epoch": 1.56, + "learning_rate": 1.4819983254256211e-05, + "loss": 0.9523, + "step": 24239 + }, + { + "epoch": 1.56, + "learning_rate": 1.4819768565233262e-05, + "loss": 0.9674, + "step": 24240 + }, + { + "epoch": 1.56, + "learning_rate": 1.4819553876210311e-05, + "loss": 0.9941, + "step": 24241 + }, + { + "epoch": 1.56, + "learning_rate": 1.481933918718736e-05, + "loss": 0.9489, + "step": 24242 + }, + { + "epoch": 1.56, + "learning_rate": 1.481912449816441e-05, + "loss": 0.9904, + "step": 24243 + }, + { + "epoch": 1.56, + "learning_rate": 1.481890980914146e-05, + "loss": 1.014, + "step": 24244 + }, + { + "epoch": 1.56, + "learning_rate": 1.4818695120118509e-05, + "loss": 0.8864, + "step": 24245 + }, + { + "epoch": 1.56, + "learning_rate": 1.481848043109556e-05, + "loss": 1.0009, + "step": 24246 + }, + { + "epoch": 1.56, + "learning_rate": 1.4818265742072608e-05, + "loss": 1.0734, + "step": 24247 + }, + { + "epoch": 1.56, + "learning_rate": 1.4818051053049657e-05, + "loss": 0.9507, + "step": 24248 + }, + { + "epoch": 1.56, + "learning_rate": 1.4817836364026708e-05, + "loss": 1.0818, + "step": 24249 + }, + { + "epoch": 1.56, + "learning_rate": 1.481762167500376e-05, + "loss": 0.9736, + "step": 24250 + }, + { + "epoch": 1.56, + "learning_rate": 1.4817406985980809e-05, + "loss": 1.0034, + "step": 24251 + }, + { + "epoch": 1.56, + "learning_rate": 1.4817192296957858e-05, + "loss": 0.9775, + "step": 24252 + }, + { + "epoch": 1.56, + "learning_rate": 1.4816977607934909e-05, + "loss": 0.997, + "step": 24253 + }, + { + "epoch": 1.56, + "learning_rate": 1.4816762918911958e-05, + "loss": 0.9534, + "step": 24254 + }, + { + "epoch": 1.56, + "learning_rate": 1.4816548229889007e-05, + "loss": 0.9823, + "step": 24255 + }, + { + "epoch": 1.56, + "learning_rate": 1.4816333540866057e-05, + "loss": 1.0206, + "step": 24256 + }, + { + "epoch": 1.56, + "learning_rate": 1.4816118851843106e-05, + "loss": 1.004, + "step": 24257 + }, + { + "epoch": 1.56, + "learning_rate": 1.4815904162820155e-05, + "loss": 1.0359, + "step": 24258 + }, + { + "epoch": 1.56, + "learning_rate": 1.4815689473797206e-05, + "loss": 0.9774, + "step": 24259 + }, + { + "epoch": 1.56, + "learning_rate": 1.4815474784774255e-05, + "loss": 0.9312, + "step": 24260 + }, + { + "epoch": 1.56, + "learning_rate": 1.4815260095751306e-05, + "loss": 0.9835, + "step": 24261 + }, + { + "epoch": 1.56, + "learning_rate": 1.4815045406728355e-05, + "loss": 0.9677, + "step": 24262 + }, + { + "epoch": 1.56, + "learning_rate": 1.4814830717705403e-05, + "loss": 0.9914, + "step": 24263 + }, + { + "epoch": 1.56, + "learning_rate": 1.4814616028682454e-05, + "loss": 0.9122, + "step": 24264 + }, + { + "epoch": 1.56, + "learning_rate": 1.4814401339659505e-05, + "loss": 0.9733, + "step": 24265 + }, + { + "epoch": 1.56, + "learning_rate": 1.4814186650636555e-05, + "loss": 1.0099, + "step": 24266 + }, + { + "epoch": 1.56, + "learning_rate": 1.4813971961613604e-05, + "loss": 0.9831, + "step": 24267 + }, + { + "epoch": 1.56, + "learning_rate": 1.4813757272590653e-05, + "loss": 1.0071, + "step": 24268 + }, + { + "epoch": 1.56, + "learning_rate": 1.4813542583567704e-05, + "loss": 0.9645, + "step": 24269 + }, + { + "epoch": 1.56, + "learning_rate": 1.4813327894544753e-05, + "loss": 0.979, + "step": 24270 + }, + { + "epoch": 1.56, + "learning_rate": 1.4813113205521804e-05, + "loss": 0.9708, + "step": 24271 + }, + { + "epoch": 1.56, + "learning_rate": 1.4812898516498853e-05, + "loss": 0.9586, + "step": 24272 + }, + { + "epoch": 1.56, + "learning_rate": 1.4812683827475902e-05, + "loss": 1.0005, + "step": 24273 + }, + { + "epoch": 1.56, + "learning_rate": 1.4812469138452952e-05, + "loss": 0.9305, + "step": 24274 + }, + { + "epoch": 1.56, + "learning_rate": 1.4812254449430001e-05, + "loss": 0.9773, + "step": 24275 + }, + { + "epoch": 1.56, + "learning_rate": 1.481203976040705e-05, + "loss": 0.9584, + "step": 24276 + }, + { + "epoch": 1.56, + "learning_rate": 1.4811825071384101e-05, + "loss": 0.9913, + "step": 24277 + }, + { + "epoch": 1.56, + "learning_rate": 1.481161038236115e-05, + "loss": 0.9915, + "step": 24278 + }, + { + "epoch": 1.56, + "learning_rate": 1.4811395693338202e-05, + "loss": 1.0246, + "step": 24279 + }, + { + "epoch": 1.56, + "learning_rate": 1.4811181004315251e-05, + "loss": 1.046, + "step": 24280 + }, + { + "epoch": 1.56, + "learning_rate": 1.4810966315292302e-05, + "loss": 0.9332, + "step": 24281 + }, + { + "epoch": 1.56, + "learning_rate": 1.481075162626935e-05, + "loss": 1.0182, + "step": 24282 + }, + { + "epoch": 1.56, + "learning_rate": 1.48105369372464e-05, + "loss": 1.0021, + "step": 24283 + }, + { + "epoch": 1.56, + "learning_rate": 1.481032224822345e-05, + "loss": 0.9616, + "step": 24284 + }, + { + "epoch": 1.56, + "learning_rate": 1.48101075592005e-05, + "loss": 0.9941, + "step": 24285 + }, + { + "epoch": 1.56, + "learning_rate": 1.4809892870177548e-05, + "loss": 0.9119, + "step": 24286 + }, + { + "epoch": 1.56, + "learning_rate": 1.4809678181154599e-05, + "loss": 0.9897, + "step": 24287 + }, + { + "epoch": 1.56, + "learning_rate": 1.4809463492131648e-05, + "loss": 1.0205, + "step": 24288 + }, + { + "epoch": 1.56, + "learning_rate": 1.4809248803108699e-05, + "loss": 1.063, + "step": 24289 + }, + { + "epoch": 1.56, + "learning_rate": 1.4809034114085748e-05, + "loss": 1.0314, + "step": 24290 + }, + { + "epoch": 1.56, + "learning_rate": 1.4808819425062797e-05, + "loss": 0.9759, + "step": 24291 + }, + { + "epoch": 1.56, + "learning_rate": 1.4808604736039847e-05, + "loss": 1.0187, + "step": 24292 + }, + { + "epoch": 1.56, + "learning_rate": 1.4808390047016896e-05, + "loss": 0.9976, + "step": 24293 + }, + { + "epoch": 1.56, + "learning_rate": 1.4808175357993949e-05, + "loss": 0.9627, + "step": 24294 + }, + { + "epoch": 1.56, + "learning_rate": 1.4807960668970997e-05, + "loss": 0.9876, + "step": 24295 + }, + { + "epoch": 1.56, + "learning_rate": 1.4807745979948046e-05, + "loss": 1.0609, + "step": 24296 + }, + { + "epoch": 1.56, + "learning_rate": 1.4807531290925097e-05, + "loss": 1.0123, + "step": 24297 + }, + { + "epoch": 1.56, + "learning_rate": 1.4807316601902146e-05, + "loss": 0.9977, + "step": 24298 + }, + { + "epoch": 1.56, + "learning_rate": 1.4807101912879197e-05, + "loss": 0.9436, + "step": 24299 + }, + { + "epoch": 1.56, + "learning_rate": 1.4806887223856246e-05, + "loss": 0.9677, + "step": 24300 + }, + { + "epoch": 1.56, + "learning_rate": 1.4806672534833295e-05, + "loss": 0.9533, + "step": 24301 + }, + { + "epoch": 1.56, + "learning_rate": 1.4806457845810345e-05, + "loss": 0.9052, + "step": 24302 + }, + { + "epoch": 1.56, + "learning_rate": 1.4806243156787394e-05, + "loss": 1.0303, + "step": 24303 + }, + { + "epoch": 1.56, + "learning_rate": 1.4806028467764443e-05, + "loss": 0.8909, + "step": 24304 + }, + { + "epoch": 1.56, + "learning_rate": 1.4805813778741494e-05, + "loss": 0.9416, + "step": 24305 + }, + { + "epoch": 1.56, + "learning_rate": 1.4805599089718543e-05, + "loss": 0.9941, + "step": 24306 + }, + { + "epoch": 1.56, + "learning_rate": 1.4805384400695594e-05, + "loss": 0.9693, + "step": 24307 + }, + { + "epoch": 1.56, + "learning_rate": 1.4805169711672644e-05, + "loss": 0.9821, + "step": 24308 + }, + { + "epoch": 1.56, + "learning_rate": 1.4804955022649695e-05, + "loss": 0.9791, + "step": 24309 + }, + { + "epoch": 1.56, + "learning_rate": 1.4804740333626744e-05, + "loss": 0.9666, + "step": 24310 + }, + { + "epoch": 1.56, + "learning_rate": 1.4804525644603793e-05, + "loss": 1.0028, + "step": 24311 + }, + { + "epoch": 1.56, + "learning_rate": 1.4804310955580843e-05, + "loss": 0.9433, + "step": 24312 + }, + { + "epoch": 1.56, + "learning_rate": 1.4804096266557892e-05, + "loss": 0.9849, + "step": 24313 + }, + { + "epoch": 1.56, + "learning_rate": 1.4803881577534941e-05, + "loss": 0.9905, + "step": 24314 + }, + { + "epoch": 1.56, + "learning_rate": 1.4803666888511992e-05, + "loss": 0.9602, + "step": 24315 + }, + { + "epoch": 1.56, + "learning_rate": 1.4803452199489041e-05, + "loss": 1.0012, + "step": 24316 + }, + { + "epoch": 1.56, + "learning_rate": 1.4803237510466092e-05, + "loss": 0.8967, + "step": 24317 + }, + { + "epoch": 1.56, + "learning_rate": 1.480302282144314e-05, + "loss": 1.0037, + "step": 24318 + }, + { + "epoch": 1.56, + "learning_rate": 1.480280813242019e-05, + "loss": 0.9472, + "step": 24319 + }, + { + "epoch": 1.56, + "learning_rate": 1.480259344339724e-05, + "loss": 0.9559, + "step": 24320 + }, + { + "epoch": 1.56, + "learning_rate": 1.480237875437429e-05, + "loss": 0.9568, + "step": 24321 + }, + { + "epoch": 1.56, + "learning_rate": 1.4802164065351338e-05, + "loss": 0.9876, + "step": 24322 + }, + { + "epoch": 1.56, + "learning_rate": 1.480194937632839e-05, + "loss": 1.0065, + "step": 24323 + }, + { + "epoch": 1.56, + "learning_rate": 1.480173468730544e-05, + "loss": 0.9966, + "step": 24324 + }, + { + "epoch": 1.57, + "learning_rate": 1.480151999828249e-05, + "loss": 0.9786, + "step": 24325 + }, + { + "epoch": 1.57, + "learning_rate": 1.4801305309259539e-05, + "loss": 0.9814, + "step": 24326 + }, + { + "epoch": 1.57, + "learning_rate": 1.480109062023659e-05, + "loss": 0.9436, + "step": 24327 + }, + { + "epoch": 1.57, + "learning_rate": 1.4800875931213639e-05, + "loss": 0.9797, + "step": 24328 + }, + { + "epoch": 1.57, + "learning_rate": 1.4800661242190688e-05, + "loss": 0.9161, + "step": 24329 + }, + { + "epoch": 1.57, + "learning_rate": 1.4800446553167738e-05, + "loss": 1.0043, + "step": 24330 + }, + { + "epoch": 1.57, + "learning_rate": 1.4800231864144787e-05, + "loss": 1.0175, + "step": 24331 + }, + { + "epoch": 1.57, + "learning_rate": 1.4800017175121836e-05, + "loss": 0.9559, + "step": 24332 + }, + { + "epoch": 1.57, + "learning_rate": 1.4799802486098887e-05, + "loss": 1.0213, + "step": 24333 + }, + { + "epoch": 1.57, + "learning_rate": 1.4799587797075936e-05, + "loss": 0.9373, + "step": 24334 + }, + { + "epoch": 1.57, + "learning_rate": 1.4799373108052985e-05, + "loss": 1.0369, + "step": 24335 + }, + { + "epoch": 1.57, + "learning_rate": 1.4799158419030036e-05, + "loss": 0.9699, + "step": 24336 + }, + { + "epoch": 1.57, + "learning_rate": 1.4798943730007084e-05, + "loss": 0.9659, + "step": 24337 + }, + { + "epoch": 1.57, + "learning_rate": 1.4798729040984137e-05, + "loss": 0.9786, + "step": 24338 + }, + { + "epoch": 1.57, + "learning_rate": 1.4798514351961186e-05, + "loss": 0.9257, + "step": 24339 + }, + { + "epoch": 1.57, + "learning_rate": 1.4798299662938236e-05, + "loss": 0.9528, + "step": 24340 + }, + { + "epoch": 1.57, + "learning_rate": 1.4798084973915285e-05, + "loss": 0.9985, + "step": 24341 + }, + { + "epoch": 1.57, + "learning_rate": 1.4797870284892334e-05, + "loss": 0.9251, + "step": 24342 + }, + { + "epoch": 1.57, + "learning_rate": 1.4797655595869385e-05, + "loss": 1.039, + "step": 24343 + }, + { + "epoch": 1.57, + "learning_rate": 1.4797440906846434e-05, + "loss": 0.9411, + "step": 24344 + }, + { + "epoch": 1.57, + "learning_rate": 1.4797226217823483e-05, + "loss": 1.0097, + "step": 24345 + }, + { + "epoch": 1.57, + "learning_rate": 1.4797011528800534e-05, + "loss": 1.0816, + "step": 24346 + }, + { + "epoch": 1.57, + "learning_rate": 1.4796796839777583e-05, + "loss": 0.9197, + "step": 24347 + }, + { + "epoch": 1.57, + "learning_rate": 1.4796582150754633e-05, + "loss": 0.8968, + "step": 24348 + }, + { + "epoch": 1.57, + "learning_rate": 1.4796367461731682e-05, + "loss": 0.9292, + "step": 24349 + }, + { + "epoch": 1.57, + "learning_rate": 1.4796152772708731e-05, + "loss": 0.9925, + "step": 24350 + }, + { + "epoch": 1.57, + "learning_rate": 1.4795938083685782e-05, + "loss": 0.9627, + "step": 24351 + }, + { + "epoch": 1.57, + "learning_rate": 1.4795723394662833e-05, + "loss": 1.0402, + "step": 24352 + }, + { + "epoch": 1.57, + "learning_rate": 1.4795508705639883e-05, + "loss": 1.048, + "step": 24353 + }, + { + "epoch": 1.57, + "learning_rate": 1.4795294016616932e-05, + "loss": 0.9495, + "step": 24354 + }, + { + "epoch": 1.57, + "learning_rate": 1.4795079327593981e-05, + "loss": 1.0396, + "step": 24355 + }, + { + "epoch": 1.57, + "learning_rate": 1.4794864638571032e-05, + "loss": 0.9734, + "step": 24356 + }, + { + "epoch": 1.57, + "learning_rate": 1.479464994954808e-05, + "loss": 0.9878, + "step": 24357 + }, + { + "epoch": 1.57, + "learning_rate": 1.4794435260525131e-05, + "loss": 0.9404, + "step": 24358 + }, + { + "epoch": 1.57, + "learning_rate": 1.479422057150218e-05, + "loss": 0.9872, + "step": 24359 + }, + { + "epoch": 1.57, + "learning_rate": 1.479400588247923e-05, + "loss": 0.9734, + "step": 24360 + }, + { + "epoch": 1.57, + "learning_rate": 1.479379119345628e-05, + "loss": 1.0118, + "step": 24361 + }, + { + "epoch": 1.57, + "learning_rate": 1.4793576504433329e-05, + "loss": 0.9343, + "step": 24362 + }, + { + "epoch": 1.57, + "learning_rate": 1.4793361815410378e-05, + "loss": 0.9332, + "step": 24363 + }, + { + "epoch": 1.57, + "learning_rate": 1.4793147126387429e-05, + "loss": 0.9411, + "step": 24364 + }, + { + "epoch": 1.57, + "learning_rate": 1.4792932437364478e-05, + "loss": 1.0077, + "step": 24365 + }, + { + "epoch": 1.57, + "learning_rate": 1.4792717748341528e-05, + "loss": 1.0084, + "step": 24366 + }, + { + "epoch": 1.57, + "learning_rate": 1.4792503059318579e-05, + "loss": 0.9767, + "step": 24367 + }, + { + "epoch": 1.57, + "learning_rate": 1.479228837029563e-05, + "loss": 1.041, + "step": 24368 + }, + { + "epoch": 1.57, + "learning_rate": 1.4792073681272678e-05, + "loss": 0.9292, + "step": 24369 + }, + { + "epoch": 1.57, + "learning_rate": 1.4791858992249727e-05, + "loss": 0.9143, + "step": 24370 + }, + { + "epoch": 1.57, + "learning_rate": 1.4791644303226778e-05, + "loss": 0.887, + "step": 24371 + }, + { + "epoch": 1.57, + "learning_rate": 1.4791429614203827e-05, + "loss": 1.0627, + "step": 24372 + }, + { + "epoch": 1.57, + "learning_rate": 1.4791214925180876e-05, + "loss": 0.9478, + "step": 24373 + }, + { + "epoch": 1.57, + "learning_rate": 1.4791000236157927e-05, + "loss": 1.0091, + "step": 24374 + }, + { + "epoch": 1.57, + "learning_rate": 1.4790785547134976e-05, + "loss": 0.9941, + "step": 24375 + }, + { + "epoch": 1.57, + "learning_rate": 1.4790570858112026e-05, + "loss": 0.995, + "step": 24376 + }, + { + "epoch": 1.57, + "learning_rate": 1.4790356169089075e-05, + "loss": 0.9827, + "step": 24377 + }, + { + "epoch": 1.57, + "learning_rate": 1.4790141480066124e-05, + "loss": 0.9884, + "step": 24378 + }, + { + "epoch": 1.57, + "learning_rate": 1.4789926791043175e-05, + "loss": 1.0058, + "step": 24379 + }, + { + "epoch": 1.57, + "learning_rate": 1.4789712102020224e-05, + "loss": 1.0246, + "step": 24380 + }, + { + "epoch": 1.57, + "learning_rate": 1.4789497412997276e-05, + "loss": 0.9652, + "step": 24381 + }, + { + "epoch": 1.57, + "learning_rate": 1.4789282723974325e-05, + "loss": 1.0082, + "step": 24382 + }, + { + "epoch": 1.57, + "learning_rate": 1.4789068034951374e-05, + "loss": 0.9775, + "step": 24383 + }, + { + "epoch": 1.57, + "learning_rate": 1.4788853345928425e-05, + "loss": 0.9583, + "step": 24384 + }, + { + "epoch": 1.57, + "learning_rate": 1.4788638656905474e-05, + "loss": 0.9632, + "step": 24385 + }, + { + "epoch": 1.57, + "learning_rate": 1.4788423967882524e-05, + "loss": 0.9253, + "step": 24386 + }, + { + "epoch": 1.57, + "learning_rate": 1.4788209278859573e-05, + "loss": 0.9561, + "step": 24387 + }, + { + "epoch": 1.57, + "learning_rate": 1.4787994589836622e-05, + "loss": 0.9268, + "step": 24388 + }, + { + "epoch": 1.57, + "learning_rate": 1.4787779900813673e-05, + "loss": 0.9245, + "step": 24389 + }, + { + "epoch": 1.57, + "learning_rate": 1.4787565211790722e-05, + "loss": 1.0174, + "step": 24390 + }, + { + "epoch": 1.57, + "learning_rate": 1.4787350522767771e-05, + "loss": 0.9522, + "step": 24391 + }, + { + "epoch": 1.57, + "learning_rate": 1.4787135833744822e-05, + "loss": 0.9213, + "step": 24392 + }, + { + "epoch": 1.57, + "learning_rate": 1.478692114472187e-05, + "loss": 0.9981, + "step": 24393 + }, + { + "epoch": 1.57, + "learning_rate": 1.478670645569892e-05, + "loss": 0.9467, + "step": 24394 + }, + { + "epoch": 1.57, + "learning_rate": 1.478649176667597e-05, + "loss": 1.0708, + "step": 24395 + }, + { + "epoch": 1.57, + "learning_rate": 1.4786277077653023e-05, + "loss": 0.9147, + "step": 24396 + }, + { + "epoch": 1.57, + "learning_rate": 1.4786062388630072e-05, + "loss": 0.964, + "step": 24397 + }, + { + "epoch": 1.57, + "learning_rate": 1.478584769960712e-05, + "loss": 0.9974, + "step": 24398 + }, + { + "epoch": 1.57, + "learning_rate": 1.4785633010584171e-05, + "loss": 0.9585, + "step": 24399 + }, + { + "epoch": 1.57, + "learning_rate": 1.478541832156122e-05, + "loss": 0.9852, + "step": 24400 + }, + { + "epoch": 1.57, + "learning_rate": 1.4785203632538269e-05, + "loss": 0.9369, + "step": 24401 + }, + { + "epoch": 1.57, + "learning_rate": 1.478498894351532e-05, + "loss": 0.9761, + "step": 24402 + }, + { + "epoch": 1.57, + "learning_rate": 1.4784774254492369e-05, + "loss": 1.0209, + "step": 24403 + }, + { + "epoch": 1.57, + "learning_rate": 1.478455956546942e-05, + "loss": 0.9274, + "step": 24404 + }, + { + "epoch": 1.57, + "learning_rate": 1.4784344876446468e-05, + "loss": 0.946, + "step": 24405 + }, + { + "epoch": 1.57, + "learning_rate": 1.4784130187423517e-05, + "loss": 0.9352, + "step": 24406 + }, + { + "epoch": 1.57, + "learning_rate": 1.4783915498400568e-05, + "loss": 0.9334, + "step": 24407 + }, + { + "epoch": 1.57, + "learning_rate": 1.4783700809377617e-05, + "loss": 1.0103, + "step": 24408 + }, + { + "epoch": 1.57, + "learning_rate": 1.4783486120354666e-05, + "loss": 0.9891, + "step": 24409 + }, + { + "epoch": 1.57, + "learning_rate": 1.4783271431331718e-05, + "loss": 0.9084, + "step": 24410 + }, + { + "epoch": 1.57, + "learning_rate": 1.4783056742308767e-05, + "loss": 0.9508, + "step": 24411 + }, + { + "epoch": 1.57, + "learning_rate": 1.4782842053285818e-05, + "loss": 0.9576, + "step": 24412 + }, + { + "epoch": 1.57, + "learning_rate": 1.4782627364262867e-05, + "loss": 1.1023, + "step": 24413 + }, + { + "epoch": 1.57, + "learning_rate": 1.4782412675239917e-05, + "loss": 0.977, + "step": 24414 + }, + { + "epoch": 1.57, + "learning_rate": 1.4782197986216966e-05, + "loss": 1.0143, + "step": 24415 + }, + { + "epoch": 1.57, + "learning_rate": 1.4781983297194015e-05, + "loss": 0.9331, + "step": 24416 + }, + { + "epoch": 1.57, + "learning_rate": 1.4781768608171066e-05, + "loss": 0.9193, + "step": 24417 + }, + { + "epoch": 1.57, + "learning_rate": 1.4781553919148115e-05, + "loss": 0.9573, + "step": 24418 + }, + { + "epoch": 1.57, + "learning_rate": 1.4781339230125164e-05, + "loss": 1.0009, + "step": 24419 + }, + { + "epoch": 1.57, + "learning_rate": 1.4781124541102215e-05, + "loss": 0.9844, + "step": 24420 + }, + { + "epoch": 1.57, + "learning_rate": 1.4780909852079264e-05, + "loss": 0.9695, + "step": 24421 + }, + { + "epoch": 1.57, + "learning_rate": 1.4780695163056313e-05, + "loss": 0.9679, + "step": 24422 + }, + { + "epoch": 1.57, + "learning_rate": 1.4780480474033363e-05, + "loss": 0.9556, + "step": 24423 + }, + { + "epoch": 1.57, + "learning_rate": 1.4780265785010412e-05, + "loss": 0.9536, + "step": 24424 + }, + { + "epoch": 1.57, + "learning_rate": 1.4780051095987465e-05, + "loss": 0.971, + "step": 24425 + }, + { + "epoch": 1.57, + "learning_rate": 1.4779836406964514e-05, + "loss": 1.0118, + "step": 24426 + }, + { + "epoch": 1.57, + "learning_rate": 1.4779621717941564e-05, + "loss": 0.9456, + "step": 24427 + }, + { + "epoch": 1.57, + "learning_rate": 1.4779407028918613e-05, + "loss": 1.0323, + "step": 24428 + }, + { + "epoch": 1.57, + "learning_rate": 1.4779192339895662e-05, + "loss": 0.93, + "step": 24429 + }, + { + "epoch": 1.57, + "learning_rate": 1.4778977650872713e-05, + "loss": 0.9927, + "step": 24430 + }, + { + "epoch": 1.57, + "learning_rate": 1.4778977650872713e-05, + "loss": 0.9234, + "step": 24431 + }, + { + "epoch": 1.57, + "learning_rate": 1.4778762961849762e-05, + "loss": 0.9404, + "step": 24432 + }, + { + "epoch": 1.57, + "learning_rate": 1.477854827282681e-05, + "loss": 0.971, + "step": 24433 + }, + { + "epoch": 1.57, + "learning_rate": 1.4778333583803861e-05, + "loss": 0.9673, + "step": 24434 + }, + { + "epoch": 1.57, + "learning_rate": 1.477811889478091e-05, + "loss": 0.9605, + "step": 24435 + }, + { + "epoch": 1.57, + "learning_rate": 1.4777904205757961e-05, + "loss": 0.9287, + "step": 24436 + }, + { + "epoch": 1.57, + "learning_rate": 1.477768951673501e-05, + "loss": 0.9784, + "step": 24437 + }, + { + "epoch": 1.57, + "learning_rate": 1.4777474827712059e-05, + "loss": 0.9881, + "step": 24438 + }, + { + "epoch": 1.57, + "learning_rate": 1.477726013868911e-05, + "loss": 0.9412, + "step": 24439 + }, + { + "epoch": 1.57, + "learning_rate": 1.477704544966616e-05, + "loss": 0.9519, + "step": 24440 + }, + { + "epoch": 1.57, + "learning_rate": 1.4776830760643211e-05, + "loss": 1.0263, + "step": 24441 + }, + { + "epoch": 1.57, + "learning_rate": 1.477661607162026e-05, + "loss": 1.0111, + "step": 24442 + }, + { + "epoch": 1.57, + "learning_rate": 1.4776401382597309e-05, + "loss": 0.9499, + "step": 24443 + }, + { + "epoch": 1.57, + "learning_rate": 1.477618669357436e-05, + "loss": 0.9464, + "step": 24444 + }, + { + "epoch": 1.57, + "learning_rate": 1.4775972004551408e-05, + "loss": 1.0153, + "step": 24445 + }, + { + "epoch": 1.57, + "learning_rate": 1.4775757315528459e-05, + "loss": 0.996, + "step": 24446 + }, + { + "epoch": 1.57, + "learning_rate": 1.4775542626505508e-05, + "loss": 0.9836, + "step": 24447 + }, + { + "epoch": 1.57, + "learning_rate": 1.4775327937482557e-05, + "loss": 0.8918, + "step": 24448 + }, + { + "epoch": 1.57, + "learning_rate": 1.4775113248459608e-05, + "loss": 0.9252, + "step": 24449 + }, + { + "epoch": 1.57, + "learning_rate": 1.4774898559436657e-05, + "loss": 1.0265, + "step": 24450 + }, + { + "epoch": 1.57, + "learning_rate": 1.4774683870413706e-05, + "loss": 0.9986, + "step": 24451 + }, + { + "epoch": 1.57, + "learning_rate": 1.4774469181390756e-05, + "loss": 1.0468, + "step": 24452 + }, + { + "epoch": 1.57, + "learning_rate": 1.4774254492367805e-05, + "loss": 0.961, + "step": 24453 + }, + { + "epoch": 1.57, + "learning_rate": 1.4774039803344856e-05, + "loss": 1.0125, + "step": 24454 + }, + { + "epoch": 1.57, + "learning_rate": 1.4773825114321907e-05, + "loss": 0.8737, + "step": 24455 + }, + { + "epoch": 1.57, + "learning_rate": 1.4773610425298957e-05, + "loss": 1.007, + "step": 24456 + }, + { + "epoch": 1.57, + "learning_rate": 1.4773395736276006e-05, + "loss": 0.9267, + "step": 24457 + }, + { + "epoch": 1.57, + "learning_rate": 1.4773181047253055e-05, + "loss": 0.9868, + "step": 24458 + }, + { + "epoch": 1.57, + "learning_rate": 1.4772966358230106e-05, + "loss": 0.9877, + "step": 24459 + }, + { + "epoch": 1.57, + "learning_rate": 1.4772751669207155e-05, + "loss": 0.9362, + "step": 24460 + }, + { + "epoch": 1.57, + "learning_rate": 1.4772536980184204e-05, + "loss": 0.966, + "step": 24461 + }, + { + "epoch": 1.57, + "learning_rate": 1.4772322291161254e-05, + "loss": 1.0315, + "step": 24462 + }, + { + "epoch": 1.57, + "learning_rate": 1.4772107602138303e-05, + "loss": 1.0197, + "step": 24463 + }, + { + "epoch": 1.57, + "learning_rate": 1.4771892913115354e-05, + "loss": 0.9716, + "step": 24464 + }, + { + "epoch": 1.57, + "learning_rate": 1.4771678224092403e-05, + "loss": 0.9865, + "step": 24465 + }, + { + "epoch": 1.57, + "learning_rate": 1.4771463535069452e-05, + "loss": 0.9447, + "step": 24466 + }, + { + "epoch": 1.57, + "learning_rate": 1.4771248846046503e-05, + "loss": 0.9906, + "step": 24467 + }, + { + "epoch": 1.57, + "learning_rate": 1.4771034157023552e-05, + "loss": 1.0084, + "step": 24468 + }, + { + "epoch": 1.57, + "learning_rate": 1.4770819468000604e-05, + "loss": 1.031, + "step": 24469 + }, + { + "epoch": 1.57, + "learning_rate": 1.4770604778977653e-05, + "loss": 1.0323, + "step": 24470 + }, + { + "epoch": 1.57, + "learning_rate": 1.4770390089954702e-05, + "loss": 0.9389, + "step": 24471 + }, + { + "epoch": 1.57, + "learning_rate": 1.4770175400931753e-05, + "loss": 0.8916, + "step": 24472 + }, + { + "epoch": 1.57, + "learning_rate": 1.4769960711908801e-05, + "loss": 0.9107, + "step": 24473 + }, + { + "epoch": 1.57, + "learning_rate": 1.4769746022885852e-05, + "loss": 0.9653, + "step": 24474 + }, + { + "epoch": 1.57, + "learning_rate": 1.4769531333862901e-05, + "loss": 0.9515, + "step": 24475 + }, + { + "epoch": 1.57, + "learning_rate": 1.476931664483995e-05, + "loss": 0.8927, + "step": 24476 + }, + { + "epoch": 1.57, + "learning_rate": 1.4769101955817e-05, + "loss": 0.9671, + "step": 24477 + }, + { + "epoch": 1.57, + "learning_rate": 1.476888726679405e-05, + "loss": 1.0389, + "step": 24478 + }, + { + "epoch": 1.57, + "learning_rate": 1.4768672577771099e-05, + "loss": 0.9947, + "step": 24479 + }, + { + "epoch": 1.57, + "learning_rate": 1.476845788874815e-05, + "loss": 1.0218, + "step": 24480 + }, + { + "epoch": 1.58, + "learning_rate": 1.4768243199725198e-05, + "loss": 0.9706, + "step": 24481 + }, + { + "epoch": 1.58, + "learning_rate": 1.4768028510702247e-05, + "loss": 0.9205, + "step": 24482 + }, + { + "epoch": 1.58, + "learning_rate": 1.4767813821679298e-05, + "loss": 0.9348, + "step": 24483 + }, + { + "epoch": 1.58, + "learning_rate": 1.476759913265635e-05, + "loss": 1.0539, + "step": 24484 + }, + { + "epoch": 1.58, + "learning_rate": 1.47673844436334e-05, + "loss": 0.9433, + "step": 24485 + }, + { + "epoch": 1.58, + "learning_rate": 1.4767169754610448e-05, + "loss": 0.9555, + "step": 24486 + }, + { + "epoch": 1.58, + "learning_rate": 1.4766955065587499e-05, + "loss": 1.0364, + "step": 24487 + }, + { + "epoch": 1.58, + "learning_rate": 1.4766740376564548e-05, + "loss": 0.9564, + "step": 24488 + }, + { + "epoch": 1.58, + "learning_rate": 1.4766525687541597e-05, + "loss": 0.915, + "step": 24489 + }, + { + "epoch": 1.58, + "learning_rate": 1.4766310998518647e-05, + "loss": 0.9691, + "step": 24490 + }, + { + "epoch": 1.58, + "learning_rate": 1.4766096309495696e-05, + "loss": 0.9558, + "step": 24491 + }, + { + "epoch": 1.58, + "learning_rate": 1.4765881620472745e-05, + "loss": 1.0277, + "step": 24492 + }, + { + "epoch": 1.58, + "learning_rate": 1.4765666931449796e-05, + "loss": 0.9275, + "step": 24493 + }, + { + "epoch": 1.58, + "learning_rate": 1.4765452242426845e-05, + "loss": 1.0267, + "step": 24494 + }, + { + "epoch": 1.58, + "learning_rate": 1.4765237553403896e-05, + "loss": 0.9163, + "step": 24495 + }, + { + "epoch": 1.58, + "learning_rate": 1.4765022864380945e-05, + "loss": 0.9983, + "step": 24496 + }, + { + "epoch": 1.58, + "learning_rate": 1.4764808175357994e-05, + "loss": 0.936, + "step": 24497 + }, + { + "epoch": 1.58, + "learning_rate": 1.4764593486335046e-05, + "loss": 0.9661, + "step": 24498 + }, + { + "epoch": 1.58, + "learning_rate": 1.4764378797312095e-05, + "loss": 0.9733, + "step": 24499 + }, + { + "epoch": 1.58, + "learning_rate": 1.4764164108289146e-05, + "loss": 0.9942, + "step": 24500 + }, + { + "epoch": 1.58, + "learning_rate": 1.4763949419266195e-05, + "loss": 0.9675, + "step": 24501 + }, + { + "epoch": 1.58, + "learning_rate": 1.4763734730243243e-05, + "loss": 0.9792, + "step": 24502 + }, + { + "epoch": 1.58, + "learning_rate": 1.4763520041220294e-05, + "loss": 0.9182, + "step": 24503 + }, + { + "epoch": 1.58, + "learning_rate": 1.4763305352197343e-05, + "loss": 0.9783, + "step": 24504 + }, + { + "epoch": 1.58, + "learning_rate": 1.4763090663174394e-05, + "loss": 0.9234, + "step": 24505 + }, + { + "epoch": 1.58, + "learning_rate": 1.4762875974151443e-05, + "loss": 0.9294, + "step": 24506 + }, + { + "epoch": 1.58, + "learning_rate": 1.4762661285128492e-05, + "loss": 1.0175, + "step": 24507 + }, + { + "epoch": 1.58, + "learning_rate": 1.4762446596105542e-05, + "loss": 0.968, + "step": 24508 + }, + { + "epoch": 1.58, + "learning_rate": 1.4762231907082591e-05, + "loss": 0.9658, + "step": 24509 + }, + { + "epoch": 1.58, + "learning_rate": 1.476201721805964e-05, + "loss": 0.9301, + "step": 24510 + }, + { + "epoch": 1.58, + "learning_rate": 1.4761802529036691e-05, + "loss": 0.9501, + "step": 24511 + }, + { + "epoch": 1.58, + "learning_rate": 1.476158784001374e-05, + "loss": 0.9466, + "step": 24512 + }, + { + "epoch": 1.58, + "learning_rate": 1.4761373150990792e-05, + "loss": 1.0253, + "step": 24513 + }, + { + "epoch": 1.58, + "learning_rate": 1.4761158461967841e-05, + "loss": 1.018, + "step": 24514 + }, + { + "epoch": 1.58, + "learning_rate": 1.4760943772944892e-05, + "loss": 0.9446, + "step": 24515 + }, + { + "epoch": 1.58, + "learning_rate": 1.4760729083921941e-05, + "loss": 1.0398, + "step": 24516 + }, + { + "epoch": 1.58, + "learning_rate": 1.476051439489899e-05, + "loss": 0.9601, + "step": 24517 + }, + { + "epoch": 1.58, + "learning_rate": 1.476029970587604e-05, + "loss": 0.964, + "step": 24518 + }, + { + "epoch": 1.58, + "learning_rate": 1.476008501685309e-05, + "loss": 0.906, + "step": 24519 + }, + { + "epoch": 1.58, + "learning_rate": 1.4759870327830138e-05, + "loss": 0.9976, + "step": 24520 + }, + { + "epoch": 1.58, + "learning_rate": 1.4759655638807189e-05, + "loss": 0.9403, + "step": 24521 + }, + { + "epoch": 1.58, + "learning_rate": 1.4759440949784238e-05, + "loss": 0.9715, + "step": 24522 + }, + { + "epoch": 1.58, + "learning_rate": 1.4759226260761289e-05, + "loss": 0.9486, + "step": 24523 + }, + { + "epoch": 1.58, + "learning_rate": 1.4759011571738338e-05, + "loss": 1.0578, + "step": 24524 + }, + { + "epoch": 1.58, + "learning_rate": 1.4758796882715387e-05, + "loss": 0.9779, + "step": 24525 + }, + { + "epoch": 1.58, + "learning_rate": 1.4758582193692437e-05, + "loss": 0.9218, + "step": 24526 + }, + { + "epoch": 1.58, + "learning_rate": 1.4758367504669488e-05, + "loss": 0.9776, + "step": 24527 + }, + { + "epoch": 1.58, + "learning_rate": 1.4758152815646539e-05, + "loss": 0.9674, + "step": 24528 + }, + { + "epoch": 1.58, + "learning_rate": 1.4757938126623588e-05, + "loss": 0.9999, + "step": 24529 + }, + { + "epoch": 1.58, + "learning_rate": 1.4757723437600637e-05, + "loss": 0.9595, + "step": 24530 + }, + { + "epoch": 1.58, + "learning_rate": 1.4757508748577687e-05, + "loss": 0.952, + "step": 24531 + }, + { + "epoch": 1.58, + "learning_rate": 1.4757294059554736e-05, + "loss": 1.0636, + "step": 24532 + }, + { + "epoch": 1.58, + "learning_rate": 1.4757079370531787e-05, + "loss": 0.942, + "step": 24533 + }, + { + "epoch": 1.58, + "learning_rate": 1.4756864681508836e-05, + "loss": 0.9567, + "step": 24534 + }, + { + "epoch": 1.58, + "learning_rate": 1.4756649992485885e-05, + "loss": 0.95, + "step": 24535 + }, + { + "epoch": 1.58, + "learning_rate": 1.4756435303462935e-05, + "loss": 0.9739, + "step": 24536 + }, + { + "epoch": 1.58, + "learning_rate": 1.4756220614439984e-05, + "loss": 0.8963, + "step": 24537 + }, + { + "epoch": 1.58, + "learning_rate": 1.4756005925417033e-05, + "loss": 0.9683, + "step": 24538 + }, + { + "epoch": 1.58, + "learning_rate": 1.4755791236394084e-05, + "loss": 1.0229, + "step": 24539 + }, + { + "epoch": 1.58, + "learning_rate": 1.4755576547371133e-05, + "loss": 0.9946, + "step": 24540 + }, + { + "epoch": 1.58, + "learning_rate": 1.4755361858348184e-05, + "loss": 1.0606, + "step": 24541 + }, + { + "epoch": 1.58, + "learning_rate": 1.4755147169325234e-05, + "loss": 1.0142, + "step": 24542 + }, + { + "epoch": 1.58, + "learning_rate": 1.4754932480302285e-05, + "loss": 1.0613, + "step": 24543 + }, + { + "epoch": 1.58, + "learning_rate": 1.4754717791279334e-05, + "loss": 0.9727, + "step": 24544 + }, + { + "epoch": 1.58, + "learning_rate": 1.4754503102256383e-05, + "loss": 1.0064, + "step": 24545 + }, + { + "epoch": 1.58, + "learning_rate": 1.4754288413233434e-05, + "loss": 0.9883, + "step": 24546 + }, + { + "epoch": 1.58, + "learning_rate": 1.4754073724210482e-05, + "loss": 0.9757, + "step": 24547 + }, + { + "epoch": 1.58, + "learning_rate": 1.4753859035187531e-05, + "loss": 0.9637, + "step": 24548 + }, + { + "epoch": 1.58, + "learning_rate": 1.4753644346164582e-05, + "loss": 0.8759, + "step": 24549 + }, + { + "epoch": 1.58, + "learning_rate": 1.4753429657141631e-05, + "loss": 0.9496, + "step": 24550 + }, + { + "epoch": 1.58, + "learning_rate": 1.4753214968118682e-05, + "loss": 1.0214, + "step": 24551 + }, + { + "epoch": 1.58, + "learning_rate": 1.475300027909573e-05, + "loss": 0.9669, + "step": 24552 + }, + { + "epoch": 1.58, + "learning_rate": 1.475278559007278e-05, + "loss": 1.0182, + "step": 24553 + }, + { + "epoch": 1.58, + "learning_rate": 1.475257090104983e-05, + "loss": 0.934, + "step": 24554 + }, + { + "epoch": 1.58, + "learning_rate": 1.475235621202688e-05, + "loss": 0.9193, + "step": 24555 + }, + { + "epoch": 1.58, + "learning_rate": 1.4752141523003932e-05, + "loss": 1.0225, + "step": 24556 + }, + { + "epoch": 1.58, + "learning_rate": 1.475192683398098e-05, + "loss": 0.9664, + "step": 24557 + }, + { + "epoch": 1.58, + "learning_rate": 1.475171214495803e-05, + "loss": 0.96, + "step": 24558 + }, + { + "epoch": 1.58, + "learning_rate": 1.475149745593508e-05, + "loss": 0.9726, + "step": 24559 + }, + { + "epoch": 1.58, + "learning_rate": 1.475128276691213e-05, + "loss": 0.9719, + "step": 24560 + }, + { + "epoch": 1.58, + "learning_rate": 1.475106807788918e-05, + "loss": 0.9678, + "step": 24561 + }, + { + "epoch": 1.58, + "learning_rate": 1.4750853388866229e-05, + "loss": 0.9392, + "step": 24562 + }, + { + "epoch": 1.58, + "learning_rate": 1.4750638699843278e-05, + "loss": 0.9545, + "step": 24563 + }, + { + "epoch": 1.58, + "learning_rate": 1.4750424010820328e-05, + "loss": 0.9375, + "step": 24564 + }, + { + "epoch": 1.58, + "learning_rate": 1.4750209321797377e-05, + "loss": 0.9543, + "step": 24565 + }, + { + "epoch": 1.58, + "learning_rate": 1.4749994632774426e-05, + "loss": 0.9882, + "step": 24566 + }, + { + "epoch": 1.58, + "learning_rate": 1.4749779943751477e-05, + "loss": 0.9929, + "step": 24567 + }, + { + "epoch": 1.58, + "learning_rate": 1.4749565254728526e-05, + "loss": 0.9521, + "step": 24568 + }, + { + "epoch": 1.58, + "learning_rate": 1.4749350565705575e-05, + "loss": 0.9525, + "step": 24569 + }, + { + "epoch": 1.58, + "learning_rate": 1.4749135876682626e-05, + "loss": 1.0331, + "step": 24570 + }, + { + "epoch": 1.58, + "learning_rate": 1.4748921187659678e-05, + "loss": 1.0086, + "step": 24571 + }, + { + "epoch": 1.58, + "learning_rate": 1.4748706498636727e-05, + "loss": 1.0425, + "step": 24572 + }, + { + "epoch": 1.58, + "learning_rate": 1.4748491809613776e-05, + "loss": 0.9862, + "step": 24573 + }, + { + "epoch": 1.58, + "learning_rate": 1.4748277120590827e-05, + "loss": 0.978, + "step": 24574 + }, + { + "epoch": 1.58, + "learning_rate": 1.4748062431567876e-05, + "loss": 0.99, + "step": 24575 + }, + { + "epoch": 1.58, + "learning_rate": 1.4747847742544924e-05, + "loss": 0.9919, + "step": 24576 + }, + { + "epoch": 1.58, + "learning_rate": 1.4747633053521975e-05, + "loss": 0.9323, + "step": 24577 + }, + { + "epoch": 1.58, + "learning_rate": 1.4747418364499024e-05, + "loss": 0.9687, + "step": 24578 + }, + { + "epoch": 1.58, + "learning_rate": 1.4747203675476073e-05, + "loss": 0.9458, + "step": 24579 + }, + { + "epoch": 1.58, + "learning_rate": 1.4746988986453124e-05, + "loss": 1.039, + "step": 24580 + }, + { + "epoch": 1.58, + "learning_rate": 1.4746774297430173e-05, + "loss": 1.0184, + "step": 24581 + }, + { + "epoch": 1.58, + "learning_rate": 1.4746559608407223e-05, + "loss": 0.9185, + "step": 24582 + }, + { + "epoch": 1.58, + "learning_rate": 1.4746344919384272e-05, + "loss": 0.9989, + "step": 24583 + }, + { + "epoch": 1.58, + "learning_rate": 1.4746130230361321e-05, + "loss": 0.9629, + "step": 24584 + }, + { + "epoch": 1.58, + "learning_rate": 1.4745915541338374e-05, + "loss": 0.9011, + "step": 24585 + }, + { + "epoch": 1.58, + "learning_rate": 1.4745700852315423e-05, + "loss": 1.0366, + "step": 24586 + }, + { + "epoch": 1.58, + "learning_rate": 1.4745486163292473e-05, + "loss": 0.9749, + "step": 24587 + }, + { + "epoch": 1.58, + "learning_rate": 1.4745271474269522e-05, + "loss": 0.9371, + "step": 24588 + }, + { + "epoch": 1.58, + "learning_rate": 1.4745056785246571e-05, + "loss": 0.924, + "step": 24589 + }, + { + "epoch": 1.58, + "learning_rate": 1.4744842096223622e-05, + "loss": 0.9738, + "step": 24590 + }, + { + "epoch": 1.58, + "learning_rate": 1.474462740720067e-05, + "loss": 0.9275, + "step": 24591 + }, + { + "epoch": 1.58, + "learning_rate": 1.4744412718177721e-05, + "loss": 0.9458, + "step": 24592 + }, + { + "epoch": 1.58, + "learning_rate": 1.474419802915477e-05, + "loss": 0.948, + "step": 24593 + }, + { + "epoch": 1.58, + "learning_rate": 1.474398334013182e-05, + "loss": 1.0097, + "step": 24594 + }, + { + "epoch": 1.58, + "learning_rate": 1.474376865110887e-05, + "loss": 0.9842, + "step": 24595 + }, + { + "epoch": 1.58, + "learning_rate": 1.4743553962085919e-05, + "loss": 0.936, + "step": 24596 + }, + { + "epoch": 1.58, + "learning_rate": 1.4743339273062968e-05, + "loss": 1.0405, + "step": 24597 + }, + { + "epoch": 1.58, + "learning_rate": 1.4743124584040019e-05, + "loss": 0.9483, + "step": 24598 + }, + { + "epoch": 1.58, + "learning_rate": 1.4742909895017068e-05, + "loss": 1.009, + "step": 24599 + }, + { + "epoch": 1.58, + "learning_rate": 1.474269520599412e-05, + "loss": 1.0041, + "step": 24600 + }, + { + "epoch": 1.58, + "learning_rate": 1.4742480516971169e-05, + "loss": 0.9355, + "step": 24601 + }, + { + "epoch": 1.58, + "learning_rate": 1.474226582794822e-05, + "loss": 0.9752, + "step": 24602 + }, + { + "epoch": 1.58, + "learning_rate": 1.4742051138925269e-05, + "loss": 0.9792, + "step": 24603 + }, + { + "epoch": 1.58, + "learning_rate": 1.4741836449902318e-05, + "loss": 1.0548, + "step": 24604 + }, + { + "epoch": 1.58, + "learning_rate": 1.4741621760879368e-05, + "loss": 0.9828, + "step": 24605 + }, + { + "epoch": 1.58, + "learning_rate": 1.4741407071856417e-05, + "loss": 1.0119, + "step": 24606 + }, + { + "epoch": 1.58, + "learning_rate": 1.4741192382833466e-05, + "loss": 0.9457, + "step": 24607 + }, + { + "epoch": 1.58, + "learning_rate": 1.4740977693810517e-05, + "loss": 0.9692, + "step": 24608 + }, + { + "epoch": 1.58, + "learning_rate": 1.4740763004787566e-05, + "loss": 0.9607, + "step": 24609 + }, + { + "epoch": 1.58, + "learning_rate": 1.4740548315764616e-05, + "loss": 0.9459, + "step": 24610 + }, + { + "epoch": 1.58, + "learning_rate": 1.4740333626741665e-05, + "loss": 0.9604, + "step": 24611 + }, + { + "epoch": 1.58, + "learning_rate": 1.4740118937718714e-05, + "loss": 0.9198, + "step": 24612 + }, + { + "epoch": 1.58, + "learning_rate": 1.4739904248695765e-05, + "loss": 0.961, + "step": 24613 + }, + { + "epoch": 1.58, + "learning_rate": 1.4739689559672816e-05, + "loss": 0.9988, + "step": 24614 + }, + { + "epoch": 1.58, + "learning_rate": 1.4739474870649866e-05, + "loss": 0.951, + "step": 24615 + }, + { + "epoch": 1.58, + "learning_rate": 1.4739260181626915e-05, + "loss": 0.9366, + "step": 24616 + }, + { + "epoch": 1.58, + "learning_rate": 1.4739045492603964e-05, + "loss": 0.9721, + "step": 24617 + }, + { + "epoch": 1.58, + "learning_rate": 1.4738830803581015e-05, + "loss": 1.0189, + "step": 24618 + }, + { + "epoch": 1.58, + "learning_rate": 1.4738616114558064e-05, + "loss": 0.8683, + "step": 24619 + }, + { + "epoch": 1.58, + "learning_rate": 1.4738401425535115e-05, + "loss": 0.9089, + "step": 24620 + }, + { + "epoch": 1.58, + "learning_rate": 1.4738186736512163e-05, + "loss": 1.0092, + "step": 24621 + }, + { + "epoch": 1.58, + "learning_rate": 1.4737972047489212e-05, + "loss": 1.0047, + "step": 24622 + }, + { + "epoch": 1.58, + "learning_rate": 1.4737757358466263e-05, + "loss": 0.8986, + "step": 24623 + }, + { + "epoch": 1.58, + "learning_rate": 1.4737542669443312e-05, + "loss": 0.9812, + "step": 24624 + }, + { + "epoch": 1.58, + "learning_rate": 1.4737327980420361e-05, + "loss": 0.9898, + "step": 24625 + }, + { + "epoch": 1.58, + "learning_rate": 1.4737113291397412e-05, + "loss": 1.0041, + "step": 24626 + }, + { + "epoch": 1.58, + "learning_rate": 1.473689860237446e-05, + "loss": 0.9245, + "step": 24627 + }, + { + "epoch": 1.58, + "learning_rate": 1.4736683913351511e-05, + "loss": 1.0282, + "step": 24628 + }, + { + "epoch": 1.58, + "learning_rate": 1.4736469224328562e-05, + "loss": 0.9735, + "step": 24629 + }, + { + "epoch": 1.58, + "learning_rate": 1.4736254535305613e-05, + "loss": 1.0085, + "step": 24630 + }, + { + "epoch": 1.58, + "learning_rate": 1.4736039846282662e-05, + "loss": 0.9235, + "step": 24631 + }, + { + "epoch": 1.58, + "learning_rate": 1.473582515725971e-05, + "loss": 0.9647, + "step": 24632 + }, + { + "epoch": 1.58, + "learning_rate": 1.4735610468236761e-05, + "loss": 1.0172, + "step": 24633 + }, + { + "epoch": 1.58, + "learning_rate": 1.473539577921381e-05, + "loss": 0.9428, + "step": 24634 + }, + { + "epoch": 1.58, + "learning_rate": 1.473518109019086e-05, + "loss": 0.9983, + "step": 24635 + }, + { + "epoch": 1.59, + "learning_rate": 1.473496640116791e-05, + "loss": 0.9904, + "step": 24636 + }, + { + "epoch": 1.59, + "learning_rate": 1.4734751712144959e-05, + "loss": 1.0689, + "step": 24637 + }, + { + "epoch": 1.59, + "learning_rate": 1.473453702312201e-05, + "loss": 0.9974, + "step": 24638 + }, + { + "epoch": 1.59, + "learning_rate": 1.4734322334099058e-05, + "loss": 0.9519, + "step": 24639 + }, + { + "epoch": 1.59, + "learning_rate": 1.4734107645076107e-05, + "loss": 0.8849, + "step": 24640 + }, + { + "epoch": 1.59, + "learning_rate": 1.4733892956053158e-05, + "loss": 0.9173, + "step": 24641 + }, + { + "epoch": 1.59, + "learning_rate": 1.4733678267030207e-05, + "loss": 0.9088, + "step": 24642 + }, + { + "epoch": 1.59, + "learning_rate": 1.473346357800726e-05, + "loss": 0.9246, + "step": 24643 + }, + { + "epoch": 1.59, + "learning_rate": 1.4733248888984308e-05, + "loss": 0.9924, + "step": 24644 + }, + { + "epoch": 1.59, + "learning_rate": 1.4733034199961357e-05, + "loss": 0.9505, + "step": 24645 + }, + { + "epoch": 1.59, + "learning_rate": 1.4732819510938408e-05, + "loss": 0.96, + "step": 24646 + }, + { + "epoch": 1.59, + "learning_rate": 1.4732604821915457e-05, + "loss": 0.9828, + "step": 24647 + }, + { + "epoch": 1.59, + "learning_rate": 1.4732390132892508e-05, + "loss": 0.9976, + "step": 24648 + }, + { + "epoch": 1.59, + "learning_rate": 1.4732175443869557e-05, + "loss": 0.9819, + "step": 24649 + }, + { + "epoch": 1.59, + "learning_rate": 1.4731960754846605e-05, + "loss": 0.9836, + "step": 24650 + }, + { + "epoch": 1.59, + "learning_rate": 1.4731746065823656e-05, + "loss": 0.9734, + "step": 24651 + }, + { + "epoch": 1.59, + "learning_rate": 1.4731531376800705e-05, + "loss": 1.0336, + "step": 24652 + }, + { + "epoch": 1.59, + "learning_rate": 1.4731316687777754e-05, + "loss": 1.0305, + "step": 24653 + }, + { + "epoch": 1.59, + "learning_rate": 1.4731101998754805e-05, + "loss": 1.0053, + "step": 24654 + }, + { + "epoch": 1.59, + "learning_rate": 1.4730887309731854e-05, + "loss": 1.0096, + "step": 24655 + }, + { + "epoch": 1.59, + "learning_rate": 1.4730672620708903e-05, + "loss": 0.9665, + "step": 24656 + }, + { + "epoch": 1.59, + "learning_rate": 1.4730457931685953e-05, + "loss": 0.9578, + "step": 24657 + }, + { + "epoch": 1.59, + "learning_rate": 1.4730243242663006e-05, + "loss": 0.9784, + "step": 24658 + }, + { + "epoch": 1.59, + "learning_rate": 1.4730028553640055e-05, + "loss": 0.9907, + "step": 24659 + }, + { + "epoch": 1.59, + "learning_rate": 1.4729813864617104e-05, + "loss": 1.0059, + "step": 24660 + }, + { + "epoch": 1.59, + "learning_rate": 1.4729599175594154e-05, + "loss": 0.9707, + "step": 24661 + }, + { + "epoch": 1.59, + "learning_rate": 1.4729384486571203e-05, + "loss": 0.9941, + "step": 24662 + }, + { + "epoch": 1.59, + "learning_rate": 1.4729169797548252e-05, + "loss": 0.9456, + "step": 24663 + }, + { + "epoch": 1.59, + "learning_rate": 1.4728955108525303e-05, + "loss": 0.9602, + "step": 24664 + }, + { + "epoch": 1.59, + "learning_rate": 1.4728740419502352e-05, + "loss": 1.0135, + "step": 24665 + }, + { + "epoch": 1.59, + "learning_rate": 1.47285257304794e-05, + "loss": 0.9652, + "step": 24666 + }, + { + "epoch": 1.59, + "learning_rate": 1.4728311041456451e-05, + "loss": 0.956, + "step": 24667 + }, + { + "epoch": 1.59, + "learning_rate": 1.47280963524335e-05, + "loss": 1.0279, + "step": 24668 + }, + { + "epoch": 1.59, + "learning_rate": 1.4727881663410551e-05, + "loss": 1.0384, + "step": 24669 + }, + { + "epoch": 1.59, + "learning_rate": 1.47276669743876e-05, + "loss": 0.9777, + "step": 24670 + }, + { + "epoch": 1.59, + "learning_rate": 1.4727452285364649e-05, + "loss": 1.0071, + "step": 24671 + }, + { + "epoch": 1.59, + "learning_rate": 1.4727237596341701e-05, + "loss": 0.9664, + "step": 24672 + }, + { + "epoch": 1.59, + "learning_rate": 1.472702290731875e-05, + "loss": 1.0092, + "step": 24673 + }, + { + "epoch": 1.59, + "learning_rate": 1.4726808218295801e-05, + "loss": 0.9726, + "step": 24674 + }, + { + "epoch": 1.59, + "learning_rate": 1.472659352927285e-05, + "loss": 0.9786, + "step": 24675 + }, + { + "epoch": 1.59, + "learning_rate": 1.4726378840249899e-05, + "loss": 0.9876, + "step": 24676 + }, + { + "epoch": 1.59, + "learning_rate": 1.472616415122695e-05, + "loss": 0.9213, + "step": 24677 + }, + { + "epoch": 1.59, + "learning_rate": 1.4725949462203999e-05, + "loss": 0.9667, + "step": 24678 + }, + { + "epoch": 1.59, + "learning_rate": 1.472573477318105e-05, + "loss": 0.9405, + "step": 24679 + }, + { + "epoch": 1.59, + "learning_rate": 1.4725520084158098e-05, + "loss": 0.9395, + "step": 24680 + }, + { + "epoch": 1.59, + "learning_rate": 1.4725305395135147e-05, + "loss": 0.9473, + "step": 24681 + }, + { + "epoch": 1.59, + "learning_rate": 1.4725090706112198e-05, + "loss": 0.9591, + "step": 24682 + }, + { + "epoch": 1.59, + "learning_rate": 1.4724876017089247e-05, + "loss": 0.9891, + "step": 24683 + }, + { + "epoch": 1.59, + "learning_rate": 1.4724661328066296e-05, + "loss": 0.9509, + "step": 24684 + }, + { + "epoch": 1.59, + "learning_rate": 1.4724446639043346e-05, + "loss": 0.9647, + "step": 24685 + }, + { + "epoch": 1.59, + "learning_rate": 1.4724231950020395e-05, + "loss": 1.1343, + "step": 24686 + }, + { + "epoch": 1.59, + "learning_rate": 1.4724017260997448e-05, + "loss": 1.0272, + "step": 24687 + }, + { + "epoch": 1.59, + "learning_rate": 1.4723802571974497e-05, + "loss": 0.9267, + "step": 24688 + }, + { + "epoch": 1.59, + "learning_rate": 1.4723587882951547e-05, + "loss": 0.9381, + "step": 24689 + }, + { + "epoch": 1.59, + "learning_rate": 1.4723373193928596e-05, + "loss": 0.9359, + "step": 24690 + }, + { + "epoch": 1.59, + "learning_rate": 1.4723158504905645e-05, + "loss": 0.9674, + "step": 24691 + }, + { + "epoch": 1.59, + "learning_rate": 1.4722943815882696e-05, + "loss": 0.9658, + "step": 24692 + }, + { + "epoch": 1.59, + "learning_rate": 1.4722729126859745e-05, + "loss": 0.9383, + "step": 24693 + }, + { + "epoch": 1.59, + "learning_rate": 1.4722514437836794e-05, + "loss": 0.9747, + "step": 24694 + }, + { + "epoch": 1.59, + "learning_rate": 1.4722299748813844e-05, + "loss": 0.949, + "step": 24695 + }, + { + "epoch": 1.59, + "learning_rate": 1.4722085059790893e-05, + "loss": 0.9969, + "step": 24696 + }, + { + "epoch": 1.59, + "learning_rate": 1.4721870370767944e-05, + "loss": 0.9508, + "step": 24697 + }, + { + "epoch": 1.59, + "learning_rate": 1.4721655681744993e-05, + "loss": 0.9489, + "step": 24698 + }, + { + "epoch": 1.59, + "learning_rate": 1.4721440992722042e-05, + "loss": 0.913, + "step": 24699 + }, + { + "epoch": 1.59, + "learning_rate": 1.4721226303699093e-05, + "loss": 0.9401, + "step": 24700 + }, + { + "epoch": 1.59, + "learning_rate": 1.4721011614676142e-05, + "loss": 0.9872, + "step": 24701 + }, + { + "epoch": 1.59, + "learning_rate": 1.4720796925653194e-05, + "loss": 0.985, + "step": 24702 + }, + { + "epoch": 1.59, + "learning_rate": 1.4720582236630243e-05, + "loss": 0.9548, + "step": 24703 + }, + { + "epoch": 1.59, + "learning_rate": 1.4720367547607292e-05, + "loss": 1.0081, + "step": 24704 + }, + { + "epoch": 1.59, + "learning_rate": 1.4720152858584343e-05, + "loss": 1.0223, + "step": 24705 + }, + { + "epoch": 1.59, + "learning_rate": 1.4719938169561392e-05, + "loss": 1.0053, + "step": 24706 + }, + { + "epoch": 1.59, + "learning_rate": 1.4719723480538442e-05, + "loss": 0.96, + "step": 24707 + }, + { + "epoch": 1.59, + "learning_rate": 1.4719508791515491e-05, + "loss": 0.9547, + "step": 24708 + }, + { + "epoch": 1.59, + "learning_rate": 1.471929410249254e-05, + "loss": 1.0418, + "step": 24709 + }, + { + "epoch": 1.59, + "learning_rate": 1.471907941346959e-05, + "loss": 0.9511, + "step": 24710 + }, + { + "epoch": 1.59, + "learning_rate": 1.471886472444664e-05, + "loss": 1.0795, + "step": 24711 + }, + { + "epoch": 1.59, + "learning_rate": 1.4718650035423689e-05, + "loss": 0.9349, + "step": 24712 + }, + { + "epoch": 1.59, + "learning_rate": 1.471843534640074e-05, + "loss": 0.9472, + "step": 24713 + }, + { + "epoch": 1.59, + "learning_rate": 1.4718220657377788e-05, + "loss": 1.0192, + "step": 24714 + }, + { + "epoch": 1.59, + "learning_rate": 1.4718005968354839e-05, + "loss": 0.9642, + "step": 24715 + }, + { + "epoch": 1.59, + "learning_rate": 1.471779127933189e-05, + "loss": 1.0173, + "step": 24716 + }, + { + "epoch": 1.59, + "learning_rate": 1.471757659030894e-05, + "loss": 0.9413, + "step": 24717 + }, + { + "epoch": 1.59, + "learning_rate": 1.471736190128599e-05, + "loss": 0.9541, + "step": 24718 + }, + { + "epoch": 1.59, + "learning_rate": 1.4717147212263038e-05, + "loss": 0.9369, + "step": 24719 + }, + { + "epoch": 1.59, + "learning_rate": 1.4716932523240089e-05, + "loss": 1.0507, + "step": 24720 + }, + { + "epoch": 1.59, + "learning_rate": 1.4716717834217138e-05, + "loss": 0.9928, + "step": 24721 + }, + { + "epoch": 1.59, + "learning_rate": 1.4716503145194187e-05, + "loss": 1.0055, + "step": 24722 + }, + { + "epoch": 1.59, + "learning_rate": 1.4716288456171238e-05, + "loss": 0.9724, + "step": 24723 + }, + { + "epoch": 1.59, + "learning_rate": 1.4716073767148286e-05, + "loss": 0.8966, + "step": 24724 + }, + { + "epoch": 1.59, + "learning_rate": 1.4715859078125337e-05, + "loss": 0.9532, + "step": 24725 + }, + { + "epoch": 1.59, + "learning_rate": 1.4715644389102386e-05, + "loss": 0.9821, + "step": 24726 + }, + { + "epoch": 1.59, + "learning_rate": 1.4715429700079435e-05, + "loss": 0.9881, + "step": 24727 + }, + { + "epoch": 1.59, + "learning_rate": 1.4715215011056486e-05, + "loss": 1.0224, + "step": 24728 + }, + { + "epoch": 1.59, + "learning_rate": 1.4715000322033535e-05, + "loss": 0.931, + "step": 24729 + }, + { + "epoch": 1.59, + "learning_rate": 1.4714785633010584e-05, + "loss": 0.9871, + "step": 24730 + }, + { + "epoch": 1.59, + "learning_rate": 1.4714570943987636e-05, + "loss": 0.974, + "step": 24731 + }, + { + "epoch": 1.59, + "learning_rate": 1.4714356254964685e-05, + "loss": 0.9773, + "step": 24732 + }, + { + "epoch": 1.59, + "learning_rate": 1.4714141565941736e-05, + "loss": 1.0032, + "step": 24733 + }, + { + "epoch": 1.59, + "learning_rate": 1.4713926876918785e-05, + "loss": 0.9925, + "step": 24734 + }, + { + "epoch": 1.59, + "learning_rate": 1.4713712187895835e-05, + "loss": 0.9295, + "step": 24735 + }, + { + "epoch": 1.59, + "learning_rate": 1.4713497498872884e-05, + "loss": 1.0311, + "step": 24736 + }, + { + "epoch": 1.59, + "learning_rate": 1.4713282809849933e-05, + "loss": 0.9588, + "step": 24737 + }, + { + "epoch": 1.59, + "learning_rate": 1.4713068120826984e-05, + "loss": 1.0356, + "step": 24738 + }, + { + "epoch": 1.59, + "learning_rate": 1.4712853431804033e-05, + "loss": 1.0078, + "step": 24739 + }, + { + "epoch": 1.59, + "learning_rate": 1.4712638742781082e-05, + "loss": 1.0701, + "step": 24740 + }, + { + "epoch": 1.59, + "learning_rate": 1.4712424053758132e-05, + "loss": 0.9822, + "step": 24741 + }, + { + "epoch": 1.59, + "learning_rate": 1.4712209364735181e-05, + "loss": 0.9535, + "step": 24742 + }, + { + "epoch": 1.59, + "learning_rate": 1.471199467571223e-05, + "loss": 0.9884, + "step": 24743 + }, + { + "epoch": 1.59, + "learning_rate": 1.4711779986689281e-05, + "loss": 0.9732, + "step": 24744 + }, + { + "epoch": 1.59, + "learning_rate": 1.4711565297666333e-05, + "loss": 0.9666, + "step": 24745 + }, + { + "epoch": 1.59, + "learning_rate": 1.4711350608643382e-05, + "loss": 1.0431, + "step": 24746 + }, + { + "epoch": 1.59, + "learning_rate": 1.4711135919620431e-05, + "loss": 0.9916, + "step": 24747 + }, + { + "epoch": 1.59, + "learning_rate": 1.4710921230597482e-05, + "loss": 0.9441, + "step": 24748 + }, + { + "epoch": 1.59, + "learning_rate": 1.4710706541574531e-05, + "loss": 0.9483, + "step": 24749 + }, + { + "epoch": 1.59, + "learning_rate": 1.471049185255158e-05, + "loss": 1.0125, + "step": 24750 + }, + { + "epoch": 1.59, + "learning_rate": 1.471027716352863e-05, + "loss": 0.9899, + "step": 24751 + }, + { + "epoch": 1.59, + "learning_rate": 1.471006247450568e-05, + "loss": 0.954, + "step": 24752 + }, + { + "epoch": 1.59, + "learning_rate": 1.4709847785482729e-05, + "loss": 1.0383, + "step": 24753 + }, + { + "epoch": 1.59, + "learning_rate": 1.470963309645978e-05, + "loss": 0.9861, + "step": 24754 + }, + { + "epoch": 1.59, + "learning_rate": 1.4709418407436828e-05, + "loss": 0.8878, + "step": 24755 + }, + { + "epoch": 1.59, + "learning_rate": 1.4709203718413879e-05, + "loss": 1.0008, + "step": 24756 + }, + { + "epoch": 1.59, + "learning_rate": 1.4708989029390928e-05, + "loss": 0.991, + "step": 24757 + }, + { + "epoch": 1.59, + "learning_rate": 1.4708774340367977e-05, + "loss": 0.9426, + "step": 24758 + }, + { + "epoch": 1.59, + "learning_rate": 1.4708559651345027e-05, + "loss": 1.0225, + "step": 24759 + }, + { + "epoch": 1.59, + "learning_rate": 1.4708344962322078e-05, + "loss": 0.9443, + "step": 24760 + }, + { + "epoch": 1.59, + "learning_rate": 1.4708130273299129e-05, + "loss": 0.9766, + "step": 24761 + }, + { + "epoch": 1.59, + "learning_rate": 1.4707915584276178e-05, + "loss": 0.9001, + "step": 24762 + }, + { + "epoch": 1.59, + "learning_rate": 1.4707700895253227e-05, + "loss": 1.0778, + "step": 24763 + }, + { + "epoch": 1.59, + "learning_rate": 1.4707486206230277e-05, + "loss": 0.9306, + "step": 24764 + }, + { + "epoch": 1.59, + "learning_rate": 1.4707271517207326e-05, + "loss": 1.0032, + "step": 24765 + }, + { + "epoch": 1.59, + "learning_rate": 1.4707056828184377e-05, + "loss": 0.94, + "step": 24766 + }, + { + "epoch": 1.59, + "learning_rate": 1.4706842139161426e-05, + "loss": 0.9856, + "step": 24767 + }, + { + "epoch": 1.59, + "learning_rate": 1.4706627450138475e-05, + "loss": 0.9244, + "step": 24768 + }, + { + "epoch": 1.59, + "learning_rate": 1.4706412761115525e-05, + "loss": 0.9432, + "step": 24769 + }, + { + "epoch": 1.59, + "learning_rate": 1.4706198072092574e-05, + "loss": 0.9338, + "step": 24770 + }, + { + "epoch": 1.59, + "learning_rate": 1.4705983383069623e-05, + "loss": 0.9455, + "step": 24771 + }, + { + "epoch": 1.59, + "learning_rate": 1.4705768694046674e-05, + "loss": 1.006, + "step": 24772 + }, + { + "epoch": 1.59, + "learning_rate": 1.4705554005023723e-05, + "loss": 0.9525, + "step": 24773 + }, + { + "epoch": 1.59, + "learning_rate": 1.4705339316000775e-05, + "loss": 0.9951, + "step": 24774 + }, + { + "epoch": 1.59, + "learning_rate": 1.4705124626977824e-05, + "loss": 1.0071, + "step": 24775 + }, + { + "epoch": 1.59, + "learning_rate": 1.4704909937954875e-05, + "loss": 0.9907, + "step": 24776 + }, + { + "epoch": 1.59, + "learning_rate": 1.4704695248931924e-05, + "loss": 0.9218, + "step": 24777 + }, + { + "epoch": 1.59, + "learning_rate": 1.4704480559908973e-05, + "loss": 1.0529, + "step": 24778 + }, + { + "epoch": 1.59, + "learning_rate": 1.4704265870886024e-05, + "loss": 0.9668, + "step": 24779 + }, + { + "epoch": 1.59, + "learning_rate": 1.4704051181863073e-05, + "loss": 0.9617, + "step": 24780 + }, + { + "epoch": 1.59, + "learning_rate": 1.4703836492840122e-05, + "loss": 0.9936, + "step": 24781 + }, + { + "epoch": 1.59, + "learning_rate": 1.4703621803817172e-05, + "loss": 0.9639, + "step": 24782 + }, + { + "epoch": 1.59, + "learning_rate": 1.4703407114794221e-05, + "loss": 1.0014, + "step": 24783 + }, + { + "epoch": 1.59, + "learning_rate": 1.4703192425771272e-05, + "loss": 1.0377, + "step": 24784 + }, + { + "epoch": 1.59, + "learning_rate": 1.470297773674832e-05, + "loss": 0.9605, + "step": 24785 + }, + { + "epoch": 1.59, + "learning_rate": 1.470276304772537e-05, + "loss": 1.029, + "step": 24786 + }, + { + "epoch": 1.59, + "learning_rate": 1.470254835870242e-05, + "loss": 1.1261, + "step": 24787 + }, + { + "epoch": 1.59, + "learning_rate": 1.470233366967947e-05, + "loss": 0.9109, + "step": 24788 + }, + { + "epoch": 1.59, + "learning_rate": 1.4702118980656522e-05, + "loss": 0.9766, + "step": 24789 + }, + { + "epoch": 1.59, + "learning_rate": 1.470190429163357e-05, + "loss": 0.9646, + "step": 24790 + }, + { + "epoch": 1.59, + "learning_rate": 1.470168960261062e-05, + "loss": 0.97, + "step": 24791 + }, + { + "epoch": 1.6, + "learning_rate": 1.470147491358767e-05, + "loss": 1.0343, + "step": 24792 + }, + { + "epoch": 1.6, + "learning_rate": 1.470126022456472e-05, + "loss": 0.973, + "step": 24793 + }, + { + "epoch": 1.6, + "learning_rate": 1.470104553554177e-05, + "loss": 0.9281, + "step": 24794 + }, + { + "epoch": 1.6, + "learning_rate": 1.4700830846518819e-05, + "loss": 0.9805, + "step": 24795 + }, + { + "epoch": 1.6, + "learning_rate": 1.4700616157495868e-05, + "loss": 0.9958, + "step": 24796 + }, + { + "epoch": 1.6, + "learning_rate": 1.4700401468472919e-05, + "loss": 0.9784, + "step": 24797 + }, + { + "epoch": 1.6, + "learning_rate": 1.4700186779449968e-05, + "loss": 0.9776, + "step": 24798 + }, + { + "epoch": 1.6, + "learning_rate": 1.4699972090427016e-05, + "loss": 0.9669, + "step": 24799 + }, + { + "epoch": 1.6, + "learning_rate": 1.4699757401404067e-05, + "loss": 1.0122, + "step": 24800 + }, + { + "epoch": 1.6, + "learning_rate": 1.4699542712381116e-05, + "loss": 0.9118, + "step": 24801 + }, + { + "epoch": 1.6, + "learning_rate": 1.4699328023358165e-05, + "loss": 1.0035, + "step": 24802 + }, + { + "epoch": 1.6, + "learning_rate": 1.4699113334335217e-05, + "loss": 0.8566, + "step": 24803 + }, + { + "epoch": 1.6, + "learning_rate": 1.4698898645312268e-05, + "loss": 0.9811, + "step": 24804 + }, + { + "epoch": 1.6, + "learning_rate": 1.4698683956289317e-05, + "loss": 0.996, + "step": 24805 + }, + { + "epoch": 1.6, + "learning_rate": 1.4698469267266366e-05, + "loss": 0.9663, + "step": 24806 + }, + { + "epoch": 1.6, + "learning_rate": 1.4698254578243417e-05, + "loss": 0.9964, + "step": 24807 + }, + { + "epoch": 1.6, + "learning_rate": 1.4698039889220466e-05, + "loss": 0.9759, + "step": 24808 + }, + { + "epoch": 1.6, + "learning_rate": 1.4697825200197515e-05, + "loss": 1.0253, + "step": 24809 + }, + { + "epoch": 1.6, + "learning_rate": 1.4697610511174565e-05, + "loss": 1.0224, + "step": 24810 + }, + { + "epoch": 1.6, + "learning_rate": 1.4697395822151614e-05, + "loss": 0.9454, + "step": 24811 + }, + { + "epoch": 1.6, + "learning_rate": 1.4697181133128663e-05, + "loss": 0.9635, + "step": 24812 + }, + { + "epoch": 1.6, + "learning_rate": 1.4696966444105714e-05, + "loss": 1.0286, + "step": 24813 + }, + { + "epoch": 1.6, + "learning_rate": 1.4696751755082763e-05, + "loss": 0.9581, + "step": 24814 + }, + { + "epoch": 1.6, + "learning_rate": 1.4696537066059813e-05, + "loss": 1.0351, + "step": 24815 + }, + { + "epoch": 1.6, + "learning_rate": 1.4696322377036862e-05, + "loss": 0.9901, + "step": 24816 + }, + { + "epoch": 1.6, + "learning_rate": 1.4696107688013911e-05, + "loss": 0.981, + "step": 24817 + }, + { + "epoch": 1.6, + "learning_rate": 1.4695892998990964e-05, + "loss": 0.958, + "step": 24818 + }, + { + "epoch": 1.6, + "learning_rate": 1.4695678309968013e-05, + "loss": 0.9417, + "step": 24819 + }, + { + "epoch": 1.6, + "learning_rate": 1.4695463620945063e-05, + "loss": 0.9674, + "step": 24820 + }, + { + "epoch": 1.6, + "learning_rate": 1.4695248931922112e-05, + "loss": 0.9924, + "step": 24821 + }, + { + "epoch": 1.6, + "learning_rate": 1.4695034242899161e-05, + "loss": 0.8988, + "step": 24822 + }, + { + "epoch": 1.6, + "learning_rate": 1.4694819553876212e-05, + "loss": 1.0062, + "step": 24823 + }, + { + "epoch": 1.6, + "learning_rate": 1.4694604864853261e-05, + "loss": 1.0122, + "step": 24824 + }, + { + "epoch": 1.6, + "learning_rate": 1.4694390175830312e-05, + "loss": 0.9602, + "step": 24825 + }, + { + "epoch": 1.6, + "learning_rate": 1.469417548680736e-05, + "loss": 0.9857, + "step": 24826 + }, + { + "epoch": 1.6, + "learning_rate": 1.469396079778441e-05, + "loss": 0.9633, + "step": 24827 + }, + { + "epoch": 1.6, + "learning_rate": 1.469374610876146e-05, + "loss": 0.9663, + "step": 24828 + }, + { + "epoch": 1.6, + "learning_rate": 1.4693531419738509e-05, + "loss": 0.847, + "step": 24829 + }, + { + "epoch": 1.6, + "learning_rate": 1.4693316730715558e-05, + "loss": 0.9239, + "step": 24830 + }, + { + "epoch": 1.6, + "learning_rate": 1.4693102041692609e-05, + "loss": 0.9634, + "step": 24831 + }, + { + "epoch": 1.6, + "learning_rate": 1.4692887352669661e-05, + "loss": 0.9964, + "step": 24832 + }, + { + "epoch": 1.6, + "learning_rate": 1.469267266364671e-05, + "loss": 0.9534, + "step": 24833 + }, + { + "epoch": 1.6, + "learning_rate": 1.4692457974623759e-05, + "loss": 1.0199, + "step": 24834 + }, + { + "epoch": 1.6, + "learning_rate": 1.469224328560081e-05, + "loss": 0.9348, + "step": 24835 + }, + { + "epoch": 1.6, + "learning_rate": 1.4692028596577859e-05, + "loss": 0.9397, + "step": 24836 + }, + { + "epoch": 1.6, + "learning_rate": 1.4691813907554908e-05, + "loss": 1.0086, + "step": 24837 + }, + { + "epoch": 1.6, + "learning_rate": 1.4691599218531958e-05, + "loss": 1.0036, + "step": 24838 + }, + { + "epoch": 1.6, + "learning_rate": 1.4691384529509007e-05, + "loss": 1.0061, + "step": 24839 + }, + { + "epoch": 1.6, + "learning_rate": 1.4691169840486056e-05, + "loss": 1.0296, + "step": 24840 + }, + { + "epoch": 1.6, + "learning_rate": 1.4690955151463107e-05, + "loss": 0.911, + "step": 24841 + }, + { + "epoch": 1.6, + "learning_rate": 1.4690740462440156e-05, + "loss": 0.9666, + "step": 24842 + }, + { + "epoch": 1.6, + "learning_rate": 1.4690525773417206e-05, + "loss": 1.0125, + "step": 24843 + }, + { + "epoch": 1.6, + "learning_rate": 1.4690311084394255e-05, + "loss": 0.9904, + "step": 24844 + }, + { + "epoch": 1.6, + "learning_rate": 1.4690096395371304e-05, + "loss": 0.9933, + "step": 24845 + }, + { + "epoch": 1.6, + "learning_rate": 1.4689881706348355e-05, + "loss": 0.9036, + "step": 24846 + }, + { + "epoch": 1.6, + "learning_rate": 1.4689667017325406e-05, + "loss": 0.9749, + "step": 24847 + }, + { + "epoch": 1.6, + "learning_rate": 1.4689452328302456e-05, + "loss": 0.983, + "step": 24848 + }, + { + "epoch": 1.6, + "learning_rate": 1.4689237639279505e-05, + "loss": 0.9737, + "step": 24849 + }, + { + "epoch": 1.6, + "learning_rate": 1.4689022950256554e-05, + "loss": 1.0679, + "step": 24850 + }, + { + "epoch": 1.6, + "learning_rate": 1.4688808261233605e-05, + "loss": 0.9286, + "step": 24851 + }, + { + "epoch": 1.6, + "learning_rate": 1.4688593572210654e-05, + "loss": 0.9693, + "step": 24852 + }, + { + "epoch": 1.6, + "learning_rate": 1.4688378883187705e-05, + "loss": 0.9605, + "step": 24853 + }, + { + "epoch": 1.6, + "learning_rate": 1.4688164194164754e-05, + "loss": 1.0835, + "step": 24854 + }, + { + "epoch": 1.6, + "learning_rate": 1.4687949505141803e-05, + "loss": 0.9664, + "step": 24855 + }, + { + "epoch": 1.6, + "learning_rate": 1.4687734816118853e-05, + "loss": 0.9973, + "step": 24856 + }, + { + "epoch": 1.6, + "learning_rate": 1.4687520127095902e-05, + "loss": 0.9264, + "step": 24857 + }, + { + "epoch": 1.6, + "learning_rate": 1.4687305438072951e-05, + "loss": 0.9474, + "step": 24858 + }, + { + "epoch": 1.6, + "learning_rate": 1.4687090749050002e-05, + "loss": 0.9223, + "step": 24859 + }, + { + "epoch": 1.6, + "learning_rate": 1.468687606002705e-05, + "loss": 0.9555, + "step": 24860 + }, + { + "epoch": 1.6, + "learning_rate": 1.4686661371004103e-05, + "loss": 1.0438, + "step": 24861 + }, + { + "epoch": 1.6, + "learning_rate": 1.4686446681981152e-05, + "loss": 1.0498, + "step": 24862 + }, + { + "epoch": 1.6, + "learning_rate": 1.4686231992958203e-05, + "loss": 0.9868, + "step": 24863 + }, + { + "epoch": 1.6, + "learning_rate": 1.4686017303935252e-05, + "loss": 1.0289, + "step": 24864 + }, + { + "epoch": 1.6, + "learning_rate": 1.46858026149123e-05, + "loss": 0.9344, + "step": 24865 + }, + { + "epoch": 1.6, + "learning_rate": 1.4685587925889351e-05, + "loss": 1.091, + "step": 24866 + }, + { + "epoch": 1.6, + "learning_rate": 1.46853732368664e-05, + "loss": 0.9254, + "step": 24867 + }, + { + "epoch": 1.6, + "learning_rate": 1.468515854784345e-05, + "loss": 0.9332, + "step": 24868 + }, + { + "epoch": 1.6, + "learning_rate": 1.46849438588205e-05, + "loss": 0.9924, + "step": 24869 + }, + { + "epoch": 1.6, + "learning_rate": 1.4684729169797549e-05, + "loss": 0.9522, + "step": 24870 + }, + { + "epoch": 1.6, + "learning_rate": 1.46845144807746e-05, + "loss": 0.9768, + "step": 24871 + }, + { + "epoch": 1.6, + "learning_rate": 1.4684299791751649e-05, + "loss": 1.0615, + "step": 24872 + }, + { + "epoch": 1.6, + "learning_rate": 1.4684085102728697e-05, + "loss": 1.0651, + "step": 24873 + }, + { + "epoch": 1.6, + "learning_rate": 1.4683870413705748e-05, + "loss": 0.9987, + "step": 24874 + }, + { + "epoch": 1.6, + "learning_rate": 1.4683655724682797e-05, + "loss": 0.9867, + "step": 24875 + }, + { + "epoch": 1.6, + "learning_rate": 1.468344103565985e-05, + "loss": 0.9245, + "step": 24876 + }, + { + "epoch": 1.6, + "learning_rate": 1.4683226346636898e-05, + "loss": 0.9599, + "step": 24877 + }, + { + "epoch": 1.6, + "learning_rate": 1.4683011657613947e-05, + "loss": 0.954, + "step": 24878 + }, + { + "epoch": 1.6, + "learning_rate": 1.4682796968590998e-05, + "loss": 0.9821, + "step": 24879 + }, + { + "epoch": 1.6, + "learning_rate": 1.4682582279568047e-05, + "loss": 0.9357, + "step": 24880 + }, + { + "epoch": 1.6, + "learning_rate": 1.4682367590545098e-05, + "loss": 0.999, + "step": 24881 + }, + { + "epoch": 1.6, + "learning_rate": 1.4682152901522147e-05, + "loss": 0.9762, + "step": 24882 + }, + { + "epoch": 1.6, + "learning_rate": 1.4681938212499196e-05, + "loss": 0.9746, + "step": 24883 + }, + { + "epoch": 1.6, + "learning_rate": 1.4681723523476246e-05, + "loss": 1.077, + "step": 24884 + }, + { + "epoch": 1.6, + "learning_rate": 1.4681508834453295e-05, + "loss": 1.0352, + "step": 24885 + }, + { + "epoch": 1.6, + "learning_rate": 1.4681294145430344e-05, + "loss": 0.9722, + "step": 24886 + }, + { + "epoch": 1.6, + "learning_rate": 1.4681079456407395e-05, + "loss": 0.976, + "step": 24887 + }, + { + "epoch": 1.6, + "learning_rate": 1.4680864767384444e-05, + "loss": 0.935, + "step": 24888 + }, + { + "epoch": 1.6, + "learning_rate": 1.4680650078361493e-05, + "loss": 1.0529, + "step": 24889 + }, + { + "epoch": 1.6, + "learning_rate": 1.4680435389338545e-05, + "loss": 0.9996, + "step": 24890 + }, + { + "epoch": 1.6, + "learning_rate": 1.4680220700315596e-05, + "loss": 0.9646, + "step": 24891 + }, + { + "epoch": 1.6, + "learning_rate": 1.4680006011292645e-05, + "loss": 0.9494, + "step": 24892 + }, + { + "epoch": 1.6, + "learning_rate": 1.4679791322269694e-05, + "loss": 0.9059, + "step": 24893 + }, + { + "epoch": 1.6, + "learning_rate": 1.4679576633246744e-05, + "loss": 0.982, + "step": 24894 + }, + { + "epoch": 1.6, + "learning_rate": 1.4679361944223793e-05, + "loss": 1.0271, + "step": 24895 + }, + { + "epoch": 1.6, + "learning_rate": 1.4679147255200842e-05, + "loss": 1.0833, + "step": 24896 + }, + { + "epoch": 1.6, + "learning_rate": 1.4678932566177893e-05, + "loss": 0.9656, + "step": 24897 + }, + { + "epoch": 1.6, + "learning_rate": 1.4678717877154942e-05, + "loss": 0.962, + "step": 24898 + }, + { + "epoch": 1.6, + "learning_rate": 1.4678503188131991e-05, + "loss": 1.053, + "step": 24899 + }, + { + "epoch": 1.6, + "learning_rate": 1.4678288499109042e-05, + "loss": 1.0212, + "step": 24900 + }, + { + "epoch": 1.6, + "learning_rate": 1.467807381008609e-05, + "loss": 0.9947, + "step": 24901 + }, + { + "epoch": 1.6, + "learning_rate": 1.4677859121063141e-05, + "loss": 0.9716, + "step": 24902 + }, + { + "epoch": 1.6, + "learning_rate": 1.467764443204019e-05, + "loss": 0.9956, + "step": 24903 + }, + { + "epoch": 1.6, + "learning_rate": 1.4677429743017239e-05, + "loss": 0.9456, + "step": 24904 + }, + { + "epoch": 1.6, + "learning_rate": 1.4677215053994291e-05, + "loss": 0.9917, + "step": 24905 + }, + { + "epoch": 1.6, + "learning_rate": 1.467700036497134e-05, + "loss": 0.895, + "step": 24906 + }, + { + "epoch": 1.6, + "learning_rate": 1.4676785675948391e-05, + "loss": 0.9515, + "step": 24907 + }, + { + "epoch": 1.6, + "learning_rate": 1.467657098692544e-05, + "loss": 0.976, + "step": 24908 + }, + { + "epoch": 1.6, + "learning_rate": 1.4676356297902489e-05, + "loss": 0.9304, + "step": 24909 + }, + { + "epoch": 1.6, + "learning_rate": 1.467614160887954e-05, + "loss": 0.9038, + "step": 24910 + }, + { + "epoch": 1.6, + "learning_rate": 1.4675926919856589e-05, + "loss": 0.9245, + "step": 24911 + }, + { + "epoch": 1.6, + "learning_rate": 1.467571223083364e-05, + "loss": 0.9715, + "step": 24912 + }, + { + "epoch": 1.6, + "learning_rate": 1.4675497541810688e-05, + "loss": 1.0172, + "step": 24913 + }, + { + "epoch": 1.6, + "learning_rate": 1.4675282852787737e-05, + "loss": 0.9306, + "step": 24914 + }, + { + "epoch": 1.6, + "learning_rate": 1.4675068163764788e-05, + "loss": 0.9471, + "step": 24915 + }, + { + "epoch": 1.6, + "learning_rate": 1.4674853474741837e-05, + "loss": 1.024, + "step": 24916 + }, + { + "epoch": 1.6, + "learning_rate": 1.4674638785718886e-05, + "loss": 0.9919, + "step": 24917 + }, + { + "epoch": 1.6, + "learning_rate": 1.4674424096695936e-05, + "loss": 0.9825, + "step": 24918 + }, + { + "epoch": 1.6, + "learning_rate": 1.4674209407672987e-05, + "loss": 0.9689, + "step": 24919 + }, + { + "epoch": 1.6, + "learning_rate": 1.4673994718650038e-05, + "loss": 0.9543, + "step": 24920 + }, + { + "epoch": 1.6, + "learning_rate": 1.4673780029627087e-05, + "loss": 0.9364, + "step": 24921 + }, + { + "epoch": 1.6, + "learning_rate": 1.4673565340604137e-05, + "loss": 0.9244, + "step": 24922 + }, + { + "epoch": 1.6, + "learning_rate": 1.4673350651581186e-05, + "loss": 0.9945, + "step": 24923 + }, + { + "epoch": 1.6, + "learning_rate": 1.4673135962558235e-05, + "loss": 0.9823, + "step": 24924 + }, + { + "epoch": 1.6, + "learning_rate": 1.4672921273535286e-05, + "loss": 0.9002, + "step": 24925 + }, + { + "epoch": 1.6, + "learning_rate": 1.4672706584512335e-05, + "loss": 0.9389, + "step": 24926 + }, + { + "epoch": 1.6, + "learning_rate": 1.4672491895489384e-05, + "loss": 0.9316, + "step": 24927 + }, + { + "epoch": 1.6, + "learning_rate": 1.4672277206466435e-05, + "loss": 0.9333, + "step": 24928 + }, + { + "epoch": 1.6, + "learning_rate": 1.4672062517443484e-05, + "loss": 0.9117, + "step": 24929 + }, + { + "epoch": 1.6, + "learning_rate": 1.4671847828420534e-05, + "loss": 1.005, + "step": 24930 + }, + { + "epoch": 1.6, + "learning_rate": 1.4671633139397583e-05, + "loss": 1.0043, + "step": 24931 + }, + { + "epoch": 1.6, + "learning_rate": 1.4671418450374632e-05, + "loss": 0.9588, + "step": 24932 + }, + { + "epoch": 1.6, + "learning_rate": 1.4671203761351683e-05, + "loss": 0.9912, + "step": 24933 + }, + { + "epoch": 1.6, + "learning_rate": 1.4670989072328733e-05, + "loss": 1.0092, + "step": 24934 + }, + { + "epoch": 1.6, + "learning_rate": 1.4670774383305784e-05, + "loss": 1.0099, + "step": 24935 + }, + { + "epoch": 1.6, + "learning_rate": 1.4670559694282833e-05, + "loss": 0.987, + "step": 24936 + }, + { + "epoch": 1.6, + "learning_rate": 1.4670345005259882e-05, + "loss": 0.9539, + "step": 24937 + }, + { + "epoch": 1.6, + "learning_rate": 1.4670130316236933e-05, + "loss": 0.9892, + "step": 24938 + }, + { + "epoch": 1.6, + "learning_rate": 1.4669915627213982e-05, + "loss": 0.9896, + "step": 24939 + }, + { + "epoch": 1.6, + "learning_rate": 1.4669700938191032e-05, + "loss": 1.053, + "step": 24940 + }, + { + "epoch": 1.6, + "learning_rate": 1.4669486249168081e-05, + "loss": 0.9556, + "step": 24941 + }, + { + "epoch": 1.6, + "learning_rate": 1.466927156014513e-05, + "loss": 1.0287, + "step": 24942 + }, + { + "epoch": 1.6, + "learning_rate": 1.4669056871122181e-05, + "loss": 0.9095, + "step": 24943 + }, + { + "epoch": 1.6, + "learning_rate": 1.466884218209923e-05, + "loss": 1.0608, + "step": 24944 + }, + { + "epoch": 1.6, + "learning_rate": 1.4668627493076279e-05, + "loss": 1.0429, + "step": 24945 + }, + { + "epoch": 1.6, + "learning_rate": 1.466841280405333e-05, + "loss": 0.9674, + "step": 24946 + }, + { + "epoch": 1.61, + "learning_rate": 1.4668198115030378e-05, + "loss": 1.009, + "step": 24947 + }, + { + "epoch": 1.61, + "learning_rate": 1.466798342600743e-05, + "loss": 0.9826, + "step": 24948 + }, + { + "epoch": 1.61, + "learning_rate": 1.466776873698448e-05, + "loss": 1.0117, + "step": 24949 + }, + { + "epoch": 1.61, + "learning_rate": 1.466755404796153e-05, + "loss": 1.0045, + "step": 24950 + }, + { + "epoch": 1.61, + "learning_rate": 1.466733935893858e-05, + "loss": 1.0241, + "step": 24951 + }, + { + "epoch": 1.61, + "learning_rate": 1.4667124669915628e-05, + "loss": 0.9872, + "step": 24952 + }, + { + "epoch": 1.61, + "learning_rate": 1.4666909980892679e-05, + "loss": 1.0152, + "step": 24953 + }, + { + "epoch": 1.61, + "learning_rate": 1.4666695291869728e-05, + "loss": 1.0087, + "step": 24954 + }, + { + "epoch": 1.61, + "learning_rate": 1.4666480602846777e-05, + "loss": 0.9886, + "step": 24955 + }, + { + "epoch": 1.61, + "learning_rate": 1.4666265913823828e-05, + "loss": 1.0218, + "step": 24956 + }, + { + "epoch": 1.61, + "learning_rate": 1.4666051224800877e-05, + "loss": 1.048, + "step": 24957 + }, + { + "epoch": 1.61, + "learning_rate": 1.4665836535777927e-05, + "loss": 0.925, + "step": 24958 + }, + { + "epoch": 1.61, + "learning_rate": 1.4665621846754976e-05, + "loss": 0.9668, + "step": 24959 + }, + { + "epoch": 1.61, + "learning_rate": 1.4665407157732025e-05, + "loss": 1.0028, + "step": 24960 + }, + { + "epoch": 1.61, + "learning_rate": 1.4665192468709076e-05, + "loss": 0.9731, + "step": 24961 + }, + { + "epoch": 1.61, + "learning_rate": 1.4664977779686125e-05, + "loss": 0.9577, + "step": 24962 + }, + { + "epoch": 1.61, + "learning_rate": 1.4664763090663177e-05, + "loss": 0.9251, + "step": 24963 + }, + { + "epoch": 1.61, + "learning_rate": 1.4664548401640226e-05, + "loss": 0.9877, + "step": 24964 + }, + { + "epoch": 1.61, + "learning_rate": 1.4664333712617275e-05, + "loss": 0.9942, + "step": 24965 + }, + { + "epoch": 1.61, + "learning_rate": 1.4664119023594326e-05, + "loss": 0.98, + "step": 24966 + }, + { + "epoch": 1.61, + "learning_rate": 1.4663904334571375e-05, + "loss": 1.0135, + "step": 24967 + }, + { + "epoch": 1.61, + "learning_rate": 1.4663689645548425e-05, + "loss": 1.0153, + "step": 24968 + }, + { + "epoch": 1.61, + "learning_rate": 1.4663474956525474e-05, + "loss": 0.9388, + "step": 24969 + }, + { + "epoch": 1.61, + "learning_rate": 1.4663260267502523e-05, + "loss": 0.9277, + "step": 24970 + }, + { + "epoch": 1.61, + "learning_rate": 1.4663045578479574e-05, + "loss": 0.9761, + "step": 24971 + }, + { + "epoch": 1.61, + "learning_rate": 1.4662830889456623e-05, + "loss": 0.9583, + "step": 24972 + }, + { + "epoch": 1.61, + "learning_rate": 1.4662616200433672e-05, + "loss": 1.0333, + "step": 24973 + }, + { + "epoch": 1.61, + "learning_rate": 1.4662401511410723e-05, + "loss": 0.9556, + "step": 24974 + }, + { + "epoch": 1.61, + "learning_rate": 1.4662186822387772e-05, + "loss": 0.9439, + "step": 24975 + }, + { + "epoch": 1.61, + "learning_rate": 1.466197213336482e-05, + "loss": 0.9454, + "step": 24976 + }, + { + "epoch": 1.61, + "learning_rate": 1.4661757444341873e-05, + "loss": 0.9459, + "step": 24977 + }, + { + "epoch": 1.61, + "learning_rate": 1.4661542755318923e-05, + "loss": 0.9862, + "step": 24978 + }, + { + "epoch": 1.61, + "learning_rate": 1.4661328066295972e-05, + "loss": 1.0088, + "step": 24979 + }, + { + "epoch": 1.61, + "learning_rate": 1.4661113377273021e-05, + "loss": 0.9385, + "step": 24980 + }, + { + "epoch": 1.61, + "learning_rate": 1.4660898688250072e-05, + "loss": 0.9596, + "step": 24981 + }, + { + "epoch": 1.61, + "learning_rate": 1.4660683999227121e-05, + "loss": 0.9315, + "step": 24982 + }, + { + "epoch": 1.61, + "learning_rate": 1.466046931020417e-05, + "loss": 0.9276, + "step": 24983 + }, + { + "epoch": 1.61, + "learning_rate": 1.466025462118122e-05, + "loss": 0.9007, + "step": 24984 + }, + { + "epoch": 1.61, + "learning_rate": 1.466003993215827e-05, + "loss": 0.9695, + "step": 24985 + }, + { + "epoch": 1.61, + "learning_rate": 1.4659825243135319e-05, + "loss": 0.9531, + "step": 24986 + }, + { + "epoch": 1.61, + "learning_rate": 1.465961055411237e-05, + "loss": 1.0158, + "step": 24987 + }, + { + "epoch": 1.61, + "learning_rate": 1.4659395865089418e-05, + "loss": 1.0052, + "step": 24988 + }, + { + "epoch": 1.61, + "learning_rate": 1.4659181176066469e-05, + "loss": 0.8966, + "step": 24989 + }, + { + "epoch": 1.61, + "learning_rate": 1.4658966487043518e-05, + "loss": 1.0394, + "step": 24990 + }, + { + "epoch": 1.61, + "learning_rate": 1.4658751798020567e-05, + "loss": 1.0089, + "step": 24991 + }, + { + "epoch": 1.61, + "learning_rate": 1.465853710899762e-05, + "loss": 0.9979, + "step": 24992 + }, + { + "epoch": 1.61, + "learning_rate": 1.4658322419974668e-05, + "loss": 0.9468, + "step": 24993 + }, + { + "epoch": 1.61, + "learning_rate": 1.4658107730951719e-05, + "loss": 0.9409, + "step": 24994 + }, + { + "epoch": 1.61, + "learning_rate": 1.4657893041928768e-05, + "loss": 0.9589, + "step": 24995 + }, + { + "epoch": 1.61, + "learning_rate": 1.4657678352905817e-05, + "loss": 0.999, + "step": 24996 + }, + { + "epoch": 1.61, + "learning_rate": 1.4657463663882867e-05, + "loss": 1.0151, + "step": 24997 + }, + { + "epoch": 1.61, + "learning_rate": 1.4657248974859916e-05, + "loss": 0.9882, + "step": 24998 + }, + { + "epoch": 1.61, + "learning_rate": 1.4657034285836967e-05, + "loss": 0.9891, + "step": 24999 + }, + { + "epoch": 1.61, + "learning_rate": 1.4656819596814016e-05, + "loss": 0.9564, + "step": 25000 + }, + { + "epoch": 1.61, + "eval_laion/cider": 0.5620856172397861, + "eval_laion/meteor": 0.3900086753435405, + "eval_laion/pred_num_words": 37.08984375, + "eval_laion/sacrebleu": 15.24858722448092, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0463902950286865, + "eval_runtime": 98.542, + "eval_samples_per_second": 5.064, + "eval_steps_per_second": 0.325, + "step": 25000 + }, + { + "epoch": 1.61, + "learning_rate": 1.4656604907791065e-05, + "loss": 0.9532, + "step": 25001 + }, + { + "epoch": 1.61, + "learning_rate": 1.4656390218768116e-05, + "loss": 0.9875, + "step": 25002 + }, + { + "epoch": 1.61, + "learning_rate": 1.4656175529745165e-05, + "loss": 0.9809, + "step": 25003 + }, + { + "epoch": 1.61, + "learning_rate": 1.4655960840722214e-05, + "loss": 0.9717, + "step": 25004 + }, + { + "epoch": 1.61, + "learning_rate": 1.4655746151699264e-05, + "loss": 0.9851, + "step": 25005 + }, + { + "epoch": 1.61, + "learning_rate": 1.4655531462676315e-05, + "loss": 0.9868, + "step": 25006 + }, + { + "epoch": 1.61, + "learning_rate": 1.4655316773653365e-05, + "loss": 1.0519, + "step": 25007 + }, + { + "epoch": 1.61, + "learning_rate": 1.4655102084630414e-05, + "loss": 1.0767, + "step": 25008 + }, + { + "epoch": 1.61, + "learning_rate": 1.4654887395607465e-05, + "loss": 0.9992, + "step": 25009 + }, + { + "epoch": 1.61, + "learning_rate": 1.4654672706584514e-05, + "loss": 0.9447, + "step": 25010 + }, + { + "epoch": 1.61, + "learning_rate": 1.4654458017561563e-05, + "loss": 0.9195, + "step": 25011 + }, + { + "epoch": 1.61, + "learning_rate": 1.4654243328538614e-05, + "loss": 0.9301, + "step": 25012 + }, + { + "epoch": 1.61, + "learning_rate": 1.4654028639515663e-05, + "loss": 1.0006, + "step": 25013 + }, + { + "epoch": 1.61, + "learning_rate": 1.4653813950492712e-05, + "loss": 0.9476, + "step": 25014 + }, + { + "epoch": 1.61, + "learning_rate": 1.4653599261469762e-05, + "loss": 0.9792, + "step": 25015 + }, + { + "epoch": 1.61, + "learning_rate": 1.4653384572446811e-05, + "loss": 0.9685, + "step": 25016 + }, + { + "epoch": 1.61, + "learning_rate": 1.4653169883423862e-05, + "loss": 1.0412, + "step": 25017 + }, + { + "epoch": 1.61, + "learning_rate": 1.4652955194400911e-05, + "loss": 1.0333, + "step": 25018 + }, + { + "epoch": 1.61, + "learning_rate": 1.465274050537796e-05, + "loss": 0.9519, + "step": 25019 + }, + { + "epoch": 1.61, + "learning_rate": 1.465252581635501e-05, + "loss": 0.9336, + "step": 25020 + }, + { + "epoch": 1.61, + "learning_rate": 1.4652311127332061e-05, + "loss": 1.097, + "step": 25021 + }, + { + "epoch": 1.61, + "learning_rate": 1.4652096438309112e-05, + "loss": 1.0215, + "step": 25022 + }, + { + "epoch": 1.61, + "learning_rate": 1.465188174928616e-05, + "loss": 0.9602, + "step": 25023 + }, + { + "epoch": 1.61, + "learning_rate": 1.465166706026321e-05, + "loss": 0.9799, + "step": 25024 + }, + { + "epoch": 1.61, + "learning_rate": 1.465145237124026e-05, + "loss": 1.0477, + "step": 25025 + }, + { + "epoch": 1.61, + "learning_rate": 1.465123768221731e-05, + "loss": 0.9116, + "step": 25026 + }, + { + "epoch": 1.61, + "learning_rate": 1.465102299319436e-05, + "loss": 0.9807, + "step": 25027 + }, + { + "epoch": 1.61, + "learning_rate": 1.4650808304171409e-05, + "loss": 0.974, + "step": 25028 + }, + { + "epoch": 1.61, + "learning_rate": 1.4650593615148458e-05, + "loss": 0.985, + "step": 25029 + }, + { + "epoch": 1.61, + "learning_rate": 1.4650378926125509e-05, + "loss": 0.9764, + "step": 25030 + }, + { + "epoch": 1.61, + "learning_rate": 1.4650164237102558e-05, + "loss": 1.0059, + "step": 25031 + }, + { + "epoch": 1.61, + "learning_rate": 1.4649949548079607e-05, + "loss": 0.9471, + "step": 25032 + }, + { + "epoch": 1.61, + "learning_rate": 1.4649734859056657e-05, + "loss": 1.0206, + "step": 25033 + }, + { + "epoch": 1.61, + "learning_rate": 1.4649520170033706e-05, + "loss": 0.9363, + "step": 25034 + }, + { + "epoch": 1.61, + "learning_rate": 1.4649305481010759e-05, + "loss": 1.049, + "step": 25035 + }, + { + "epoch": 1.61, + "learning_rate": 1.4649090791987808e-05, + "loss": 0.9416, + "step": 25036 + }, + { + "epoch": 1.61, + "learning_rate": 1.4648876102964858e-05, + "loss": 0.9807, + "step": 25037 + }, + { + "epoch": 1.61, + "learning_rate": 1.4648661413941907e-05, + "loss": 1.006, + "step": 25038 + }, + { + "epoch": 1.61, + "learning_rate": 1.4648446724918956e-05, + "loss": 0.9484, + "step": 25039 + }, + { + "epoch": 1.61, + "learning_rate": 1.4648232035896007e-05, + "loss": 0.9426, + "step": 25040 + }, + { + "epoch": 1.61, + "learning_rate": 1.4648017346873056e-05, + "loss": 0.9489, + "step": 25041 + }, + { + "epoch": 1.61, + "learning_rate": 1.4647802657850105e-05, + "loss": 0.9722, + "step": 25042 + }, + { + "epoch": 1.61, + "learning_rate": 1.4647587968827155e-05, + "loss": 0.9984, + "step": 25043 + }, + { + "epoch": 1.61, + "learning_rate": 1.4647373279804204e-05, + "loss": 0.9836, + "step": 25044 + }, + { + "epoch": 1.61, + "learning_rate": 1.4647158590781255e-05, + "loss": 0.9944, + "step": 25045 + }, + { + "epoch": 1.61, + "learning_rate": 1.4646943901758304e-05, + "loss": 0.9197, + "step": 25046 + }, + { + "epoch": 1.61, + "learning_rate": 1.4646729212735353e-05, + "loss": 0.9068, + "step": 25047 + }, + { + "epoch": 1.61, + "learning_rate": 1.4646514523712404e-05, + "loss": 0.9483, + "step": 25048 + }, + { + "epoch": 1.61, + "learning_rate": 1.4646299834689453e-05, + "loss": 0.9569, + "step": 25049 + }, + { + "epoch": 1.61, + "learning_rate": 1.4646085145666505e-05, + "loss": 1.0328, + "step": 25050 + }, + { + "epoch": 1.61, + "learning_rate": 1.4645870456643554e-05, + "loss": 0.9657, + "step": 25051 + }, + { + "epoch": 1.61, + "learning_rate": 1.4645655767620603e-05, + "loss": 1.0769, + "step": 25052 + }, + { + "epoch": 1.61, + "learning_rate": 1.4645441078597653e-05, + "loss": 0.9896, + "step": 25053 + }, + { + "epoch": 1.61, + "learning_rate": 1.4645226389574702e-05, + "loss": 0.998, + "step": 25054 + }, + { + "epoch": 1.61, + "learning_rate": 1.4645011700551753e-05, + "loss": 0.9791, + "step": 25055 + }, + { + "epoch": 1.61, + "learning_rate": 1.4644797011528802e-05, + "loss": 0.9114, + "step": 25056 + }, + { + "epoch": 1.61, + "learning_rate": 1.4644582322505851e-05, + "loss": 1.0535, + "step": 25057 + }, + { + "epoch": 1.61, + "learning_rate": 1.4644367633482902e-05, + "loss": 0.9173, + "step": 25058 + }, + { + "epoch": 1.61, + "learning_rate": 1.464415294445995e-05, + "loss": 0.9945, + "step": 25059 + }, + { + "epoch": 1.61, + "learning_rate": 1.4643938255437e-05, + "loss": 0.9717, + "step": 25060 + }, + { + "epoch": 1.61, + "learning_rate": 1.464372356641405e-05, + "loss": 0.9383, + "step": 25061 + }, + { + "epoch": 1.61, + "learning_rate": 1.46435088773911e-05, + "loss": 0.9799, + "step": 25062 + }, + { + "epoch": 1.61, + "learning_rate": 1.4643294188368148e-05, + "loss": 0.9325, + "step": 25063 + }, + { + "epoch": 1.61, + "learning_rate": 1.46430794993452e-05, + "loss": 0.9862, + "step": 25064 + }, + { + "epoch": 1.61, + "learning_rate": 1.4642864810322251e-05, + "loss": 1.0396, + "step": 25065 + }, + { + "epoch": 1.61, + "learning_rate": 1.46426501212993e-05, + "loss": 0.8945, + "step": 25066 + }, + { + "epoch": 1.61, + "learning_rate": 1.4642435432276349e-05, + "loss": 0.9766, + "step": 25067 + }, + { + "epoch": 1.61, + "learning_rate": 1.46422207432534e-05, + "loss": 0.9941, + "step": 25068 + }, + { + "epoch": 1.61, + "learning_rate": 1.4642006054230449e-05, + "loss": 1.0214, + "step": 25069 + }, + { + "epoch": 1.61, + "learning_rate": 1.4641791365207498e-05, + "loss": 1.0487, + "step": 25070 + }, + { + "epoch": 1.61, + "learning_rate": 1.4641576676184548e-05, + "loss": 0.933, + "step": 25071 + }, + { + "epoch": 1.61, + "learning_rate": 1.4641361987161597e-05, + "loss": 0.9771, + "step": 25072 + }, + { + "epoch": 1.61, + "learning_rate": 1.4641147298138646e-05, + "loss": 1.0093, + "step": 25073 + }, + { + "epoch": 1.61, + "learning_rate": 1.4640932609115697e-05, + "loss": 1.0058, + "step": 25074 + }, + { + "epoch": 1.61, + "learning_rate": 1.4640717920092746e-05, + "loss": 1.0389, + "step": 25075 + }, + { + "epoch": 1.61, + "learning_rate": 1.4640503231069797e-05, + "loss": 0.8846, + "step": 25076 + }, + { + "epoch": 1.61, + "learning_rate": 1.4640288542046846e-05, + "loss": 0.9398, + "step": 25077 + }, + { + "epoch": 1.61, + "learning_rate": 1.4640073853023895e-05, + "loss": 1.0163, + "step": 25078 + }, + { + "epoch": 1.61, + "learning_rate": 1.4639859164000947e-05, + "loss": 0.9752, + "step": 25079 + }, + { + "epoch": 1.61, + "learning_rate": 1.4639644474977996e-05, + "loss": 1.0181, + "step": 25080 + }, + { + "epoch": 1.61, + "learning_rate": 1.4639429785955047e-05, + "loss": 0.9898, + "step": 25081 + }, + { + "epoch": 1.61, + "learning_rate": 1.4639215096932095e-05, + "loss": 1.0022, + "step": 25082 + }, + { + "epoch": 1.61, + "learning_rate": 1.4639000407909144e-05, + "loss": 0.9646, + "step": 25083 + }, + { + "epoch": 1.61, + "learning_rate": 1.4638785718886195e-05, + "loss": 0.9758, + "step": 25084 + }, + { + "epoch": 1.61, + "learning_rate": 1.4638571029863244e-05, + "loss": 0.976, + "step": 25085 + }, + { + "epoch": 1.61, + "learning_rate": 1.4638356340840295e-05, + "loss": 1.007, + "step": 25086 + }, + { + "epoch": 1.61, + "learning_rate": 1.4638141651817344e-05, + "loss": 0.9637, + "step": 25087 + }, + { + "epoch": 1.61, + "learning_rate": 1.4637926962794393e-05, + "loss": 0.9492, + "step": 25088 + }, + { + "epoch": 1.61, + "learning_rate": 1.4637712273771443e-05, + "loss": 0.9739, + "step": 25089 + }, + { + "epoch": 1.61, + "learning_rate": 1.4637497584748492e-05, + "loss": 0.9782, + "step": 25090 + }, + { + "epoch": 1.61, + "learning_rate": 1.4637282895725541e-05, + "loss": 1.0157, + "step": 25091 + }, + { + "epoch": 1.61, + "learning_rate": 1.4637068206702592e-05, + "loss": 0.9752, + "step": 25092 + }, + { + "epoch": 1.61, + "learning_rate": 1.4636853517679641e-05, + "loss": 0.9647, + "step": 25093 + }, + { + "epoch": 1.61, + "learning_rate": 1.4636638828656693e-05, + "loss": 0.9943, + "step": 25094 + }, + { + "epoch": 1.61, + "learning_rate": 1.4636424139633742e-05, + "loss": 0.9508, + "step": 25095 + }, + { + "epoch": 1.61, + "learning_rate": 1.4636209450610793e-05, + "loss": 0.9605, + "step": 25096 + }, + { + "epoch": 1.61, + "learning_rate": 1.4635994761587842e-05, + "loss": 0.9049, + "step": 25097 + }, + { + "epoch": 1.61, + "learning_rate": 1.463578007256489e-05, + "loss": 0.9379, + "step": 25098 + }, + { + "epoch": 1.61, + "learning_rate": 1.4635565383541941e-05, + "loss": 1.0263, + "step": 25099 + }, + { + "epoch": 1.61, + "learning_rate": 1.463535069451899e-05, + "loss": 0.9597, + "step": 25100 + }, + { + "epoch": 1.61, + "learning_rate": 1.463513600549604e-05, + "loss": 0.9822, + "step": 25101 + }, + { + "epoch": 1.62, + "learning_rate": 1.463492131647309e-05, + "loss": 0.9406, + "step": 25102 + }, + { + "epoch": 1.62, + "learning_rate": 1.4634706627450139e-05, + "loss": 1.0357, + "step": 25103 + }, + { + "epoch": 1.62, + "learning_rate": 1.463449193842719e-05, + "loss": 1.0318, + "step": 25104 + }, + { + "epoch": 1.62, + "learning_rate": 1.4634277249404239e-05, + "loss": 0.9938, + "step": 25105 + }, + { + "epoch": 1.62, + "learning_rate": 1.4634062560381288e-05, + "loss": 0.9847, + "step": 25106 + }, + { + "epoch": 1.62, + "learning_rate": 1.4633847871358338e-05, + "loss": 0.9596, + "step": 25107 + }, + { + "epoch": 1.62, + "learning_rate": 1.4633633182335389e-05, + "loss": 1.052, + "step": 25108 + }, + { + "epoch": 1.62, + "learning_rate": 1.463341849331244e-05, + "loss": 0.9285, + "step": 25109 + }, + { + "epoch": 1.62, + "learning_rate": 1.4633203804289489e-05, + "loss": 0.99, + "step": 25110 + }, + { + "epoch": 1.62, + "learning_rate": 1.4632989115266537e-05, + "loss": 0.936, + "step": 25111 + }, + { + "epoch": 1.62, + "learning_rate": 1.4632774426243588e-05, + "loss": 0.973, + "step": 25112 + }, + { + "epoch": 1.62, + "learning_rate": 1.4632559737220637e-05, + "loss": 0.9924, + "step": 25113 + }, + { + "epoch": 1.62, + "learning_rate": 1.4632345048197688e-05, + "loss": 0.9529, + "step": 25114 + }, + { + "epoch": 1.62, + "learning_rate": 1.4632130359174737e-05, + "loss": 1.0323, + "step": 25115 + }, + { + "epoch": 1.62, + "learning_rate": 1.4631915670151786e-05, + "loss": 0.9479, + "step": 25116 + }, + { + "epoch": 1.62, + "learning_rate": 1.4631700981128836e-05, + "loss": 0.9777, + "step": 25117 + }, + { + "epoch": 1.62, + "learning_rate": 1.4631486292105885e-05, + "loss": 1.0731, + "step": 25118 + }, + { + "epoch": 1.62, + "learning_rate": 1.4631271603082934e-05, + "loss": 1.0134, + "step": 25119 + }, + { + "epoch": 1.62, + "learning_rate": 1.4631056914059985e-05, + "loss": 1.055, + "step": 25120 + }, + { + "epoch": 1.62, + "learning_rate": 1.4630842225037034e-05, + "loss": 1.006, + "step": 25121 + }, + { + "epoch": 1.62, + "learning_rate": 1.4630627536014083e-05, + "loss": 0.951, + "step": 25122 + }, + { + "epoch": 1.62, + "learning_rate": 1.4630412846991135e-05, + "loss": 0.9777, + "step": 25123 + }, + { + "epoch": 1.62, + "learning_rate": 1.4630198157968186e-05, + "loss": 0.9683, + "step": 25124 + }, + { + "epoch": 1.62, + "learning_rate": 1.4629983468945235e-05, + "loss": 0.9439, + "step": 25125 + }, + { + "epoch": 1.62, + "learning_rate": 1.4629768779922284e-05, + "loss": 0.9434, + "step": 25126 + }, + { + "epoch": 1.62, + "learning_rate": 1.4629554090899334e-05, + "loss": 0.9172, + "step": 25127 + }, + { + "epoch": 1.62, + "learning_rate": 1.4629339401876383e-05, + "loss": 0.9853, + "step": 25128 + }, + { + "epoch": 1.62, + "learning_rate": 1.4629124712853432e-05, + "loss": 0.9881, + "step": 25129 + }, + { + "epoch": 1.62, + "learning_rate": 1.4628910023830483e-05, + "loss": 1.0206, + "step": 25130 + }, + { + "epoch": 1.62, + "learning_rate": 1.4628695334807532e-05, + "loss": 1.0324, + "step": 25131 + }, + { + "epoch": 1.62, + "learning_rate": 1.4628480645784581e-05, + "loss": 1.018, + "step": 25132 + }, + { + "epoch": 1.62, + "learning_rate": 1.4628265956761632e-05, + "loss": 1.0251, + "step": 25133 + }, + { + "epoch": 1.62, + "learning_rate": 1.462805126773868e-05, + "loss": 0.9844, + "step": 25134 + }, + { + "epoch": 1.62, + "learning_rate": 1.4627836578715731e-05, + "loss": 0.9439, + "step": 25135 + }, + { + "epoch": 1.62, + "learning_rate": 1.462762188969278e-05, + "loss": 0.9417, + "step": 25136 + }, + { + "epoch": 1.62, + "learning_rate": 1.4627407200669833e-05, + "loss": 0.98, + "step": 25137 + }, + { + "epoch": 1.62, + "learning_rate": 1.4627192511646882e-05, + "loss": 0.9394, + "step": 25138 + }, + { + "epoch": 1.62, + "learning_rate": 1.462697782262393e-05, + "loss": 0.977, + "step": 25139 + }, + { + "epoch": 1.62, + "learning_rate": 1.4626763133600981e-05, + "loss": 1.0234, + "step": 25140 + }, + { + "epoch": 1.62, + "learning_rate": 1.462654844457803e-05, + "loss": 1.014, + "step": 25141 + }, + { + "epoch": 1.62, + "learning_rate": 1.462633375555508e-05, + "loss": 1.0014, + "step": 25142 + }, + { + "epoch": 1.62, + "learning_rate": 1.462611906653213e-05, + "loss": 0.982, + "step": 25143 + }, + { + "epoch": 1.62, + "learning_rate": 1.4625904377509179e-05, + "loss": 0.9986, + "step": 25144 + }, + { + "epoch": 1.62, + "learning_rate": 1.462568968848623e-05, + "loss": 0.9154, + "step": 25145 + }, + { + "epoch": 1.62, + "learning_rate": 1.4625474999463278e-05, + "loss": 1.0034, + "step": 25146 + }, + { + "epoch": 1.62, + "learning_rate": 1.4625260310440327e-05, + "loss": 0.9615, + "step": 25147 + }, + { + "epoch": 1.62, + "learning_rate": 1.4625045621417378e-05, + "loss": 1.0487, + "step": 25148 + }, + { + "epoch": 1.62, + "learning_rate": 1.4624830932394427e-05, + "loss": 0.896, + "step": 25149 + }, + { + "epoch": 1.62, + "learning_rate": 1.4624616243371476e-05, + "loss": 1.0099, + "step": 25150 + }, + { + "epoch": 1.62, + "learning_rate": 1.4624401554348527e-05, + "loss": 1.0206, + "step": 25151 + }, + { + "epoch": 1.62, + "learning_rate": 1.4624186865325579e-05, + "loss": 0.9728, + "step": 25152 + }, + { + "epoch": 1.62, + "learning_rate": 1.4623972176302628e-05, + "loss": 0.9725, + "step": 25153 + }, + { + "epoch": 1.62, + "learning_rate": 1.4623757487279677e-05, + "loss": 0.9633, + "step": 25154 + }, + { + "epoch": 1.62, + "learning_rate": 1.4623542798256728e-05, + "loss": 0.9253, + "step": 25155 + }, + { + "epoch": 1.62, + "learning_rate": 1.4623328109233776e-05, + "loss": 0.9739, + "step": 25156 + }, + { + "epoch": 1.62, + "learning_rate": 1.4623113420210825e-05, + "loss": 0.9629, + "step": 25157 + }, + { + "epoch": 1.62, + "learning_rate": 1.4622898731187876e-05, + "loss": 0.9846, + "step": 25158 + }, + { + "epoch": 1.62, + "learning_rate": 1.4622684042164925e-05, + "loss": 0.9177, + "step": 25159 + }, + { + "epoch": 1.62, + "learning_rate": 1.4622469353141974e-05, + "loss": 0.9114, + "step": 25160 + }, + { + "epoch": 1.62, + "learning_rate": 1.4622254664119025e-05, + "loss": 0.9169, + "step": 25161 + }, + { + "epoch": 1.62, + "learning_rate": 1.4622039975096074e-05, + "loss": 0.9518, + "step": 25162 + }, + { + "epoch": 1.62, + "learning_rate": 1.4621825286073124e-05, + "loss": 0.9006, + "step": 25163 + }, + { + "epoch": 1.62, + "learning_rate": 1.4621610597050173e-05, + "loss": 1.034, + "step": 25164 + }, + { + "epoch": 1.62, + "learning_rate": 1.4621395908027222e-05, + "loss": 1.0026, + "step": 25165 + }, + { + "epoch": 1.62, + "learning_rate": 1.4621181219004275e-05, + "loss": 1.0133, + "step": 25166 + }, + { + "epoch": 1.62, + "learning_rate": 1.4620966529981324e-05, + "loss": 0.983, + "step": 25167 + }, + { + "epoch": 1.62, + "learning_rate": 1.4620751840958374e-05, + "loss": 1.0368, + "step": 25168 + }, + { + "epoch": 1.62, + "learning_rate": 1.4620537151935423e-05, + "loss": 0.9607, + "step": 25169 + }, + { + "epoch": 1.62, + "learning_rate": 1.4620322462912472e-05, + "loss": 0.9567, + "step": 25170 + }, + { + "epoch": 1.62, + "learning_rate": 1.4620107773889523e-05, + "loss": 1.0192, + "step": 25171 + }, + { + "epoch": 1.62, + "learning_rate": 1.4619893084866572e-05, + "loss": 0.9586, + "step": 25172 + }, + { + "epoch": 1.62, + "learning_rate": 1.4619678395843622e-05, + "loss": 0.9265, + "step": 25173 + }, + { + "epoch": 1.62, + "learning_rate": 1.4619463706820671e-05, + "loss": 0.9381, + "step": 25174 + }, + { + "epoch": 1.62, + "learning_rate": 1.461924901779772e-05, + "loss": 0.97, + "step": 25175 + }, + { + "epoch": 1.62, + "learning_rate": 1.4619034328774771e-05, + "loss": 0.9117, + "step": 25176 + }, + { + "epoch": 1.62, + "learning_rate": 1.461881963975182e-05, + "loss": 1.0011, + "step": 25177 + }, + { + "epoch": 1.62, + "learning_rate": 1.4618604950728869e-05, + "loss": 0.9783, + "step": 25178 + }, + { + "epoch": 1.62, + "learning_rate": 1.461839026170592e-05, + "loss": 0.9691, + "step": 25179 + }, + { + "epoch": 1.62, + "learning_rate": 1.4618175572682969e-05, + "loss": 1.0072, + "step": 25180 + }, + { + "epoch": 1.62, + "learning_rate": 1.4617960883660021e-05, + "loss": 0.9638, + "step": 25181 + }, + { + "epoch": 1.62, + "learning_rate": 1.461774619463707e-05, + "loss": 0.8607, + "step": 25182 + }, + { + "epoch": 1.62, + "learning_rate": 1.461753150561412e-05, + "loss": 0.9691, + "step": 25183 + }, + { + "epoch": 1.62, + "learning_rate": 1.461731681659117e-05, + "loss": 0.8804, + "step": 25184 + }, + { + "epoch": 1.62, + "learning_rate": 1.4617102127568218e-05, + "loss": 0.9554, + "step": 25185 + }, + { + "epoch": 1.62, + "learning_rate": 1.4616887438545269e-05, + "loss": 0.9699, + "step": 25186 + }, + { + "epoch": 1.62, + "learning_rate": 1.4616672749522318e-05, + "loss": 0.9116, + "step": 25187 + }, + { + "epoch": 1.62, + "learning_rate": 1.4616458060499367e-05, + "loss": 0.9958, + "step": 25188 + }, + { + "epoch": 1.62, + "learning_rate": 1.4616243371476418e-05, + "loss": 0.908, + "step": 25189 + }, + { + "epoch": 1.62, + "learning_rate": 1.4616028682453467e-05, + "loss": 1.0212, + "step": 25190 + }, + { + "epoch": 1.62, + "learning_rate": 1.4615813993430517e-05, + "loss": 0.9238, + "step": 25191 + }, + { + "epoch": 1.62, + "learning_rate": 1.4615599304407566e-05, + "loss": 0.9617, + "step": 25192 + }, + { + "epoch": 1.62, + "learning_rate": 1.4615384615384615e-05, + "loss": 0.9488, + "step": 25193 + }, + { + "epoch": 1.62, + "learning_rate": 1.4615169926361666e-05, + "loss": 1.0548, + "step": 25194 + }, + { + "epoch": 1.62, + "learning_rate": 1.4614955237338717e-05, + "loss": 1.019, + "step": 25195 + }, + { + "epoch": 1.62, + "learning_rate": 1.4614740548315767e-05, + "loss": 0.9186, + "step": 25196 + }, + { + "epoch": 1.62, + "learning_rate": 1.4614525859292816e-05, + "loss": 0.9932, + "step": 25197 + }, + { + "epoch": 1.62, + "learning_rate": 1.4614311170269865e-05, + "loss": 0.9699, + "step": 25198 + }, + { + "epoch": 1.62, + "learning_rate": 1.4614096481246916e-05, + "loss": 0.9237, + "step": 25199 + }, + { + "epoch": 1.62, + "learning_rate": 1.4613881792223965e-05, + "loss": 1.0311, + "step": 25200 + }, + { + "epoch": 1.62, + "learning_rate": 1.4613667103201015e-05, + "loss": 0.9777, + "step": 25201 + }, + { + "epoch": 1.62, + "learning_rate": 1.4613452414178064e-05, + "loss": 0.9758, + "step": 25202 + }, + { + "epoch": 1.62, + "learning_rate": 1.4613237725155113e-05, + "loss": 0.9096, + "step": 25203 + }, + { + "epoch": 1.62, + "learning_rate": 1.4613023036132164e-05, + "loss": 1.0134, + "step": 25204 + }, + { + "epoch": 1.62, + "learning_rate": 1.4612808347109213e-05, + "loss": 1.0447, + "step": 25205 + }, + { + "epoch": 1.62, + "learning_rate": 1.4612593658086262e-05, + "loss": 1.0215, + "step": 25206 + }, + { + "epoch": 1.62, + "learning_rate": 1.4612378969063313e-05, + "loss": 0.9997, + "step": 25207 + }, + { + "epoch": 1.62, + "learning_rate": 1.4612164280040362e-05, + "loss": 0.9651, + "step": 25208 + }, + { + "epoch": 1.62, + "learning_rate": 1.461194959101741e-05, + "loss": 0.9245, + "step": 25209 + }, + { + "epoch": 1.62, + "learning_rate": 1.4611734901994463e-05, + "loss": 1.0235, + "step": 25210 + }, + { + "epoch": 1.62, + "learning_rate": 1.4611520212971514e-05, + "loss": 0.917, + "step": 25211 + }, + { + "epoch": 1.62, + "learning_rate": 1.4611305523948563e-05, + "loss": 0.8913, + "step": 25212 + }, + { + "epoch": 1.62, + "learning_rate": 1.4611090834925612e-05, + "loss": 0.9641, + "step": 25213 + }, + { + "epoch": 1.62, + "learning_rate": 1.4610876145902662e-05, + "loss": 0.9774, + "step": 25214 + }, + { + "epoch": 1.62, + "learning_rate": 1.4610661456879711e-05, + "loss": 1.0183, + "step": 25215 + }, + { + "epoch": 1.62, + "learning_rate": 1.461044676785676e-05, + "loss": 0.9157, + "step": 25216 + }, + { + "epoch": 1.62, + "learning_rate": 1.461023207883381e-05, + "loss": 1.0264, + "step": 25217 + }, + { + "epoch": 1.62, + "learning_rate": 1.461001738981086e-05, + "loss": 0.972, + "step": 25218 + }, + { + "epoch": 1.62, + "learning_rate": 1.4609802700787909e-05, + "loss": 0.9903, + "step": 25219 + }, + { + "epoch": 1.62, + "learning_rate": 1.460958801176496e-05, + "loss": 0.9328, + "step": 25220 + }, + { + "epoch": 1.62, + "learning_rate": 1.4609373322742008e-05, + "loss": 0.9325, + "step": 25221 + }, + { + "epoch": 1.62, + "learning_rate": 1.4609158633719059e-05, + "loss": 1.004, + "step": 25222 + }, + { + "epoch": 1.62, + "learning_rate": 1.4608943944696108e-05, + "loss": 0.95, + "step": 25223 + }, + { + "epoch": 1.62, + "learning_rate": 1.460872925567316e-05, + "loss": 0.9704, + "step": 25224 + }, + { + "epoch": 1.62, + "learning_rate": 1.460851456665021e-05, + "loss": 0.9179, + "step": 25225 + }, + { + "epoch": 1.62, + "learning_rate": 1.4608299877627258e-05, + "loss": 0.9879, + "step": 25226 + }, + { + "epoch": 1.62, + "learning_rate": 1.4608085188604309e-05, + "loss": 0.9365, + "step": 25227 + }, + { + "epoch": 1.62, + "learning_rate": 1.4607870499581358e-05, + "loss": 0.9573, + "step": 25228 + }, + { + "epoch": 1.62, + "learning_rate": 1.4607655810558407e-05, + "loss": 1.0151, + "step": 25229 + }, + { + "epoch": 1.62, + "learning_rate": 1.4607441121535457e-05, + "loss": 0.9746, + "step": 25230 + }, + { + "epoch": 1.62, + "learning_rate": 1.4607226432512506e-05, + "loss": 0.9965, + "step": 25231 + }, + { + "epoch": 1.62, + "learning_rate": 1.4607011743489557e-05, + "loss": 0.9408, + "step": 25232 + }, + { + "epoch": 1.62, + "learning_rate": 1.4606797054466606e-05, + "loss": 0.9627, + "step": 25233 + }, + { + "epoch": 1.62, + "learning_rate": 1.4606582365443655e-05, + "loss": 0.9732, + "step": 25234 + }, + { + "epoch": 1.62, + "learning_rate": 1.4606367676420706e-05, + "loss": 0.9916, + "step": 25235 + }, + { + "epoch": 1.62, + "learning_rate": 1.4606152987397755e-05, + "loss": 0.9305, + "step": 25236 + }, + { + "epoch": 1.62, + "learning_rate": 1.4605938298374804e-05, + "loss": 0.9972, + "step": 25237 + }, + { + "epoch": 1.62, + "learning_rate": 1.4605723609351854e-05, + "loss": 0.963, + "step": 25238 + }, + { + "epoch": 1.62, + "learning_rate": 1.4605508920328905e-05, + "loss": 0.9452, + "step": 25239 + }, + { + "epoch": 1.62, + "learning_rate": 1.4605294231305956e-05, + "loss": 0.9744, + "step": 25240 + }, + { + "epoch": 1.62, + "learning_rate": 1.4605079542283005e-05, + "loss": 0.8939, + "step": 25241 + }, + { + "epoch": 1.62, + "learning_rate": 1.4604864853260055e-05, + "loss": 0.9385, + "step": 25242 + }, + { + "epoch": 1.62, + "learning_rate": 1.4604650164237104e-05, + "loss": 0.9616, + "step": 25243 + }, + { + "epoch": 1.62, + "learning_rate": 1.4604435475214153e-05, + "loss": 0.9877, + "step": 25244 + }, + { + "epoch": 1.62, + "learning_rate": 1.4604220786191204e-05, + "loss": 0.9786, + "step": 25245 + }, + { + "epoch": 1.62, + "learning_rate": 1.4604006097168253e-05, + "loss": 1.0244, + "step": 25246 + }, + { + "epoch": 1.62, + "learning_rate": 1.4603791408145302e-05, + "loss": 0.9604, + "step": 25247 + }, + { + "epoch": 1.62, + "learning_rate": 1.4603576719122352e-05, + "loss": 1.0198, + "step": 25248 + }, + { + "epoch": 1.62, + "learning_rate": 1.4603362030099401e-05, + "loss": 0.9814, + "step": 25249 + }, + { + "epoch": 1.62, + "learning_rate": 1.4603147341076452e-05, + "loss": 0.9322, + "step": 25250 + }, + { + "epoch": 1.62, + "learning_rate": 1.4602932652053501e-05, + "loss": 0.9637, + "step": 25251 + }, + { + "epoch": 1.62, + "learning_rate": 1.460271796303055e-05, + "loss": 0.946, + "step": 25252 + }, + { + "epoch": 1.62, + "learning_rate": 1.4602503274007602e-05, + "loss": 1.0101, + "step": 25253 + }, + { + "epoch": 1.62, + "learning_rate": 1.4602288584984651e-05, + "loss": 0.8919, + "step": 25254 + }, + { + "epoch": 1.62, + "learning_rate": 1.4602073895961702e-05, + "loss": 1.0842, + "step": 25255 + }, + { + "epoch": 1.62, + "learning_rate": 1.4601859206938751e-05, + "loss": 1.0145, + "step": 25256 + }, + { + "epoch": 1.62, + "learning_rate": 1.46016445179158e-05, + "loss": 0.9555, + "step": 25257 + }, + { + "epoch": 1.63, + "learning_rate": 1.460142982889285e-05, + "loss": 0.9879, + "step": 25258 + }, + { + "epoch": 1.63, + "learning_rate": 1.46012151398699e-05, + "loss": 0.9396, + "step": 25259 + }, + { + "epoch": 1.63, + "learning_rate": 1.460100045084695e-05, + "loss": 0.985, + "step": 25260 + }, + { + "epoch": 1.63, + "learning_rate": 1.4600785761823999e-05, + "loss": 1.004, + "step": 25261 + }, + { + "epoch": 1.63, + "learning_rate": 1.4600571072801048e-05, + "loss": 0.9336, + "step": 25262 + }, + { + "epoch": 1.63, + "learning_rate": 1.4600356383778099e-05, + "loss": 0.9883, + "step": 25263 + }, + { + "epoch": 1.63, + "learning_rate": 1.4600141694755148e-05, + "loss": 1.0093, + "step": 25264 + }, + { + "epoch": 1.63, + "learning_rate": 1.4599927005732197e-05, + "loss": 1.0249, + "step": 25265 + }, + { + "epoch": 1.63, + "learning_rate": 1.4599712316709247e-05, + "loss": 0.9381, + "step": 25266 + }, + { + "epoch": 1.63, + "learning_rate": 1.4599497627686296e-05, + "loss": 0.9703, + "step": 25267 + }, + { + "epoch": 1.63, + "learning_rate": 1.4599282938663349e-05, + "loss": 0.9364, + "step": 25268 + }, + { + "epoch": 1.63, + "learning_rate": 1.4599068249640398e-05, + "loss": 0.9876, + "step": 25269 + }, + { + "epoch": 1.63, + "learning_rate": 1.4598853560617448e-05, + "loss": 0.9183, + "step": 25270 + }, + { + "epoch": 1.63, + "learning_rate": 1.4598638871594497e-05, + "loss": 1.0025, + "step": 25271 + }, + { + "epoch": 1.63, + "learning_rate": 1.4598424182571546e-05, + "loss": 1.0303, + "step": 25272 + }, + { + "epoch": 1.63, + "learning_rate": 1.4598209493548597e-05, + "loss": 0.9345, + "step": 25273 + }, + { + "epoch": 1.63, + "learning_rate": 1.4597994804525646e-05, + "loss": 0.9994, + "step": 25274 + }, + { + "epoch": 1.63, + "learning_rate": 1.4597780115502695e-05, + "loss": 1.0253, + "step": 25275 + }, + { + "epoch": 1.63, + "learning_rate": 1.4597565426479745e-05, + "loss": 0.9738, + "step": 25276 + }, + { + "epoch": 1.63, + "learning_rate": 1.4597350737456794e-05, + "loss": 0.9438, + "step": 25277 + }, + { + "epoch": 1.63, + "learning_rate": 1.4597136048433845e-05, + "loss": 0.9675, + "step": 25278 + }, + { + "epoch": 1.63, + "learning_rate": 1.4596921359410894e-05, + "loss": 0.9004, + "step": 25279 + }, + { + "epoch": 1.63, + "learning_rate": 1.4596706670387943e-05, + "loss": 0.9689, + "step": 25280 + }, + { + "epoch": 1.63, + "learning_rate": 1.4596491981364994e-05, + "loss": 0.9539, + "step": 25281 + }, + { + "epoch": 1.63, + "learning_rate": 1.4596277292342044e-05, + "loss": 1.0191, + "step": 25282 + }, + { + "epoch": 1.63, + "learning_rate": 1.4596062603319095e-05, + "loss": 1.0146, + "step": 25283 + }, + { + "epoch": 1.63, + "learning_rate": 1.4595847914296144e-05, + "loss": 1.0277, + "step": 25284 + }, + { + "epoch": 1.63, + "learning_rate": 1.4595633225273193e-05, + "loss": 0.9452, + "step": 25285 + }, + { + "epoch": 1.63, + "learning_rate": 1.4595418536250244e-05, + "loss": 0.9234, + "step": 25286 + }, + { + "epoch": 1.63, + "learning_rate": 1.4595203847227293e-05, + "loss": 0.9876, + "step": 25287 + }, + { + "epoch": 1.63, + "learning_rate": 1.4594989158204343e-05, + "loss": 1.0211, + "step": 25288 + }, + { + "epoch": 1.63, + "learning_rate": 1.4594774469181392e-05, + "loss": 0.8708, + "step": 25289 + }, + { + "epoch": 1.63, + "learning_rate": 1.4594559780158441e-05, + "loss": 0.9597, + "step": 25290 + }, + { + "epoch": 1.63, + "learning_rate": 1.4594345091135492e-05, + "loss": 0.9614, + "step": 25291 + }, + { + "epoch": 1.63, + "learning_rate": 1.459413040211254e-05, + "loss": 1.0342, + "step": 25292 + }, + { + "epoch": 1.63, + "learning_rate": 1.459391571308959e-05, + "loss": 0.9557, + "step": 25293 + }, + { + "epoch": 1.63, + "learning_rate": 1.459370102406664e-05, + "loss": 0.975, + "step": 25294 + }, + { + "epoch": 1.63, + "learning_rate": 1.459348633504369e-05, + "loss": 1.075, + "step": 25295 + }, + { + "epoch": 1.63, + "learning_rate": 1.4593271646020738e-05, + "loss": 1.0132, + "step": 25296 + }, + { + "epoch": 1.63, + "learning_rate": 1.459305695699779e-05, + "loss": 1.0225, + "step": 25297 + }, + { + "epoch": 1.63, + "learning_rate": 1.4592842267974841e-05, + "loss": 0.9994, + "step": 25298 + }, + { + "epoch": 1.63, + "learning_rate": 1.459262757895189e-05, + "loss": 0.8998, + "step": 25299 + }, + { + "epoch": 1.63, + "learning_rate": 1.459241288992894e-05, + "loss": 0.9453, + "step": 25300 + }, + { + "epoch": 1.63, + "learning_rate": 1.459219820090599e-05, + "loss": 0.9319, + "step": 25301 + }, + { + "epoch": 1.63, + "learning_rate": 1.4591983511883039e-05, + "loss": 1.028, + "step": 25302 + }, + { + "epoch": 1.63, + "learning_rate": 1.4591768822860088e-05, + "loss": 0.9578, + "step": 25303 + }, + { + "epoch": 1.63, + "learning_rate": 1.4591554133837138e-05, + "loss": 0.9505, + "step": 25304 + }, + { + "epoch": 1.63, + "learning_rate": 1.4591339444814187e-05, + "loss": 0.956, + "step": 25305 + }, + { + "epoch": 1.63, + "learning_rate": 1.4591124755791236e-05, + "loss": 0.9713, + "step": 25306 + }, + { + "epoch": 1.63, + "learning_rate": 1.4590910066768287e-05, + "loss": 0.979, + "step": 25307 + }, + { + "epoch": 1.63, + "learning_rate": 1.4590695377745336e-05, + "loss": 1.0179, + "step": 25308 + }, + { + "epoch": 1.63, + "learning_rate": 1.4590480688722387e-05, + "loss": 0.9401, + "step": 25309 + }, + { + "epoch": 1.63, + "learning_rate": 1.4590265999699436e-05, + "loss": 1.0157, + "step": 25310 + }, + { + "epoch": 1.63, + "learning_rate": 1.4590051310676488e-05, + "loss": 0.9691, + "step": 25311 + }, + { + "epoch": 1.63, + "learning_rate": 1.4589836621653537e-05, + "loss": 1.0267, + "step": 25312 + }, + { + "epoch": 1.63, + "learning_rate": 1.4589621932630586e-05, + "loss": 0.9987, + "step": 25313 + }, + { + "epoch": 1.63, + "learning_rate": 1.4589407243607637e-05, + "loss": 1.0221, + "step": 25314 + }, + { + "epoch": 1.63, + "learning_rate": 1.4589192554584686e-05, + "loss": 0.9378, + "step": 25315 + }, + { + "epoch": 1.63, + "learning_rate": 1.4588977865561735e-05, + "loss": 1.045, + "step": 25316 + }, + { + "epoch": 1.63, + "learning_rate": 1.4588763176538785e-05, + "loss": 1.0085, + "step": 25317 + }, + { + "epoch": 1.63, + "learning_rate": 1.4588548487515834e-05, + "loss": 0.9695, + "step": 25318 + }, + { + "epoch": 1.63, + "learning_rate": 1.4588333798492885e-05, + "loss": 1.0128, + "step": 25319 + }, + { + "epoch": 1.63, + "learning_rate": 1.4588119109469934e-05, + "loss": 0.9344, + "step": 25320 + }, + { + "epoch": 1.63, + "learning_rate": 1.4587904420446983e-05, + "loss": 0.9747, + "step": 25321 + }, + { + "epoch": 1.63, + "learning_rate": 1.4587689731424033e-05, + "loss": 0.9754, + "step": 25322 + }, + { + "epoch": 1.63, + "learning_rate": 1.4587475042401082e-05, + "loss": 1.0142, + "step": 25323 + }, + { + "epoch": 1.63, + "learning_rate": 1.4587260353378131e-05, + "loss": 0.9746, + "step": 25324 + }, + { + "epoch": 1.63, + "learning_rate": 1.4587045664355182e-05, + "loss": 0.9598, + "step": 25325 + }, + { + "epoch": 1.63, + "learning_rate": 1.4586830975332233e-05, + "loss": 1.0049, + "step": 25326 + }, + { + "epoch": 1.63, + "learning_rate": 1.4586616286309283e-05, + "loss": 0.9637, + "step": 25327 + }, + { + "epoch": 1.63, + "learning_rate": 1.4586401597286332e-05, + "loss": 1.0588, + "step": 25328 + }, + { + "epoch": 1.63, + "learning_rate": 1.4586186908263383e-05, + "loss": 0.9231, + "step": 25329 + }, + { + "epoch": 1.63, + "learning_rate": 1.4585972219240432e-05, + "loss": 0.9577, + "step": 25330 + }, + { + "epoch": 1.63, + "learning_rate": 1.4585757530217481e-05, + "loss": 0.9421, + "step": 25331 + }, + { + "epoch": 1.63, + "learning_rate": 1.4585542841194532e-05, + "loss": 0.9274, + "step": 25332 + }, + { + "epoch": 1.63, + "learning_rate": 1.458532815217158e-05, + "loss": 0.9096, + "step": 25333 + }, + { + "epoch": 1.63, + "learning_rate": 1.458511346314863e-05, + "loss": 0.9625, + "step": 25334 + }, + { + "epoch": 1.63, + "learning_rate": 1.458489877412568e-05, + "loss": 0.9902, + "step": 25335 + }, + { + "epoch": 1.63, + "learning_rate": 1.4584684085102729e-05, + "loss": 0.9617, + "step": 25336 + }, + { + "epoch": 1.63, + "learning_rate": 1.458446939607978e-05, + "loss": 1.0196, + "step": 25337 + }, + { + "epoch": 1.63, + "learning_rate": 1.4584254707056829e-05, + "loss": 1.0504, + "step": 25338 + }, + { + "epoch": 1.63, + "learning_rate": 1.4584040018033878e-05, + "loss": 0.9709, + "step": 25339 + }, + { + "epoch": 1.63, + "learning_rate": 1.458382532901093e-05, + "loss": 0.9228, + "step": 25340 + }, + { + "epoch": 1.63, + "learning_rate": 1.4583610639987979e-05, + "loss": 1.0059, + "step": 25341 + }, + { + "epoch": 1.63, + "learning_rate": 1.458339595096503e-05, + "loss": 0.894, + "step": 25342 + }, + { + "epoch": 1.63, + "learning_rate": 1.4583181261942079e-05, + "loss": 0.9748, + "step": 25343 + }, + { + "epoch": 1.63, + "learning_rate": 1.4582966572919128e-05, + "loss": 0.923, + "step": 25344 + }, + { + "epoch": 1.63, + "learning_rate": 1.4582751883896178e-05, + "loss": 0.9621, + "step": 25345 + }, + { + "epoch": 1.63, + "learning_rate": 1.4582537194873227e-05, + "loss": 0.9785, + "step": 25346 + }, + { + "epoch": 1.63, + "learning_rate": 1.4582322505850278e-05, + "loss": 1.0268, + "step": 25347 + }, + { + "epoch": 1.63, + "learning_rate": 1.4582107816827327e-05, + "loss": 0.9941, + "step": 25348 + }, + { + "epoch": 1.63, + "learning_rate": 1.4581893127804376e-05, + "loss": 0.9018, + "step": 25349 + }, + { + "epoch": 1.63, + "learning_rate": 1.4581678438781426e-05, + "loss": 1.0249, + "step": 25350 + }, + { + "epoch": 1.63, + "learning_rate": 1.4581463749758475e-05, + "loss": 0.9836, + "step": 25351 + }, + { + "epoch": 1.63, + "learning_rate": 1.4581249060735524e-05, + "loss": 0.923, + "step": 25352 + }, + { + "epoch": 1.63, + "learning_rate": 1.4581034371712575e-05, + "loss": 0.9437, + "step": 25353 + }, + { + "epoch": 1.63, + "learning_rate": 1.4580819682689624e-05, + "loss": 0.9389, + "step": 25354 + }, + { + "epoch": 1.63, + "learning_rate": 1.4580604993666676e-05, + "loss": 1.0212, + "step": 25355 + }, + { + "epoch": 1.63, + "learning_rate": 1.4580390304643725e-05, + "loss": 1.0075, + "step": 25356 + }, + { + "epoch": 1.63, + "learning_rate": 1.4580175615620776e-05, + "loss": 0.9639, + "step": 25357 + }, + { + "epoch": 1.63, + "learning_rate": 1.4579960926597825e-05, + "loss": 0.9257, + "step": 25358 + }, + { + "epoch": 1.63, + "learning_rate": 1.4579746237574874e-05, + "loss": 1.044, + "step": 25359 + }, + { + "epoch": 1.63, + "learning_rate": 1.4579531548551925e-05, + "loss": 0.9792, + "step": 25360 + }, + { + "epoch": 1.63, + "learning_rate": 1.4579316859528974e-05, + "loss": 0.8868, + "step": 25361 + }, + { + "epoch": 1.63, + "learning_rate": 1.4579102170506022e-05, + "loss": 0.9496, + "step": 25362 + }, + { + "epoch": 1.63, + "learning_rate": 1.4578887481483073e-05, + "loss": 0.8741, + "step": 25363 + }, + { + "epoch": 1.63, + "learning_rate": 1.4578672792460122e-05, + "loss": 1.0357, + "step": 25364 + }, + { + "epoch": 1.63, + "learning_rate": 1.4578458103437173e-05, + "loss": 0.9807, + "step": 25365 + }, + { + "epoch": 1.63, + "learning_rate": 1.4578243414414222e-05, + "loss": 0.9752, + "step": 25366 + }, + { + "epoch": 1.63, + "learning_rate": 1.457802872539127e-05, + "loss": 0.9744, + "step": 25367 + }, + { + "epoch": 1.63, + "learning_rate": 1.4577814036368321e-05, + "loss": 0.9494, + "step": 25368 + }, + { + "epoch": 1.63, + "learning_rate": 1.4577599347345372e-05, + "loss": 1.0066, + "step": 25369 + }, + { + "epoch": 1.63, + "learning_rate": 1.4577384658322423e-05, + "loss": 0.9788, + "step": 25370 + }, + { + "epoch": 1.63, + "learning_rate": 1.4577169969299472e-05, + "loss": 0.9825, + "step": 25371 + }, + { + "epoch": 1.63, + "learning_rate": 1.457695528027652e-05, + "loss": 1.0091, + "step": 25372 + }, + { + "epoch": 1.63, + "learning_rate": 1.4576740591253571e-05, + "loss": 0.9804, + "step": 25373 + }, + { + "epoch": 1.63, + "learning_rate": 1.457652590223062e-05, + "loss": 1.0108, + "step": 25374 + }, + { + "epoch": 1.63, + "learning_rate": 1.4576311213207671e-05, + "loss": 0.9491, + "step": 25375 + }, + { + "epoch": 1.63, + "learning_rate": 1.457609652418472e-05, + "loss": 0.9311, + "step": 25376 + }, + { + "epoch": 1.63, + "learning_rate": 1.4575881835161769e-05, + "loss": 0.9374, + "step": 25377 + }, + { + "epoch": 1.63, + "learning_rate": 1.457566714613882e-05, + "loss": 0.8924, + "step": 25378 + }, + { + "epoch": 1.63, + "learning_rate": 1.4575452457115868e-05, + "loss": 0.9569, + "step": 25379 + }, + { + "epoch": 1.63, + "learning_rate": 1.4575237768092917e-05, + "loss": 0.9253, + "step": 25380 + }, + { + "epoch": 1.63, + "learning_rate": 1.4575023079069968e-05, + "loss": 0.9543, + "step": 25381 + }, + { + "epoch": 1.63, + "learning_rate": 1.4574808390047017e-05, + "loss": 0.9726, + "step": 25382 + }, + { + "epoch": 1.63, + "learning_rate": 1.4574593701024066e-05, + "loss": 0.9799, + "step": 25383 + }, + { + "epoch": 1.63, + "learning_rate": 1.4574379012001118e-05, + "loss": 0.9888, + "step": 25384 + }, + { + "epoch": 1.63, + "learning_rate": 1.4574164322978169e-05, + "loss": 0.9228, + "step": 25385 + }, + { + "epoch": 1.63, + "learning_rate": 1.4573949633955218e-05, + "loss": 0.9406, + "step": 25386 + }, + { + "epoch": 1.63, + "learning_rate": 1.4573734944932267e-05, + "loss": 0.9793, + "step": 25387 + }, + { + "epoch": 1.63, + "learning_rate": 1.4573520255909318e-05, + "loss": 0.9863, + "step": 25388 + }, + { + "epoch": 1.63, + "learning_rate": 1.4573305566886367e-05, + "loss": 0.9589, + "step": 25389 + }, + { + "epoch": 1.63, + "learning_rate": 1.4573090877863416e-05, + "loss": 0.9206, + "step": 25390 + }, + { + "epoch": 1.63, + "learning_rate": 1.4572876188840466e-05, + "loss": 0.9976, + "step": 25391 + }, + { + "epoch": 1.63, + "learning_rate": 1.4572661499817515e-05, + "loss": 0.9912, + "step": 25392 + }, + { + "epoch": 1.63, + "learning_rate": 1.4572446810794564e-05, + "loss": 0.9853, + "step": 25393 + }, + { + "epoch": 1.63, + "learning_rate": 1.4572232121771615e-05, + "loss": 0.9618, + "step": 25394 + }, + { + "epoch": 1.63, + "learning_rate": 1.4572017432748664e-05, + "loss": 1.0054, + "step": 25395 + }, + { + "epoch": 1.63, + "learning_rate": 1.4571802743725714e-05, + "loss": 1.0065, + "step": 25396 + }, + { + "epoch": 1.63, + "learning_rate": 1.4571588054702763e-05, + "loss": 0.9648, + "step": 25397 + }, + { + "epoch": 1.63, + "learning_rate": 1.4571373365679816e-05, + "loss": 0.9548, + "step": 25398 + }, + { + "epoch": 1.63, + "learning_rate": 1.4571158676656865e-05, + "loss": 0.9291, + "step": 25399 + }, + { + "epoch": 1.63, + "learning_rate": 1.4570943987633914e-05, + "loss": 1.0058, + "step": 25400 + }, + { + "epoch": 1.63, + "learning_rate": 1.4570729298610964e-05, + "loss": 0.9019, + "step": 25401 + }, + { + "epoch": 1.63, + "learning_rate": 1.4570514609588013e-05, + "loss": 0.9806, + "step": 25402 + }, + { + "epoch": 1.63, + "learning_rate": 1.4570299920565062e-05, + "loss": 0.9367, + "step": 25403 + }, + { + "epoch": 1.63, + "learning_rate": 1.4570085231542113e-05, + "loss": 1.0014, + "step": 25404 + }, + { + "epoch": 1.63, + "learning_rate": 1.4569870542519162e-05, + "loss": 0.949, + "step": 25405 + }, + { + "epoch": 1.63, + "learning_rate": 1.4569655853496213e-05, + "loss": 0.9382, + "step": 25406 + }, + { + "epoch": 1.63, + "learning_rate": 1.4569441164473261e-05, + "loss": 0.9719, + "step": 25407 + }, + { + "epoch": 1.63, + "learning_rate": 1.456922647545031e-05, + "loss": 0.9925, + "step": 25408 + }, + { + "epoch": 1.63, + "learning_rate": 1.4569011786427361e-05, + "loss": 0.9887, + "step": 25409 + }, + { + "epoch": 1.63, + "learning_rate": 1.456879709740441e-05, + "loss": 0.9791, + "step": 25410 + }, + { + "epoch": 1.63, + "learning_rate": 1.4568582408381459e-05, + "loss": 0.9331, + "step": 25411 + }, + { + "epoch": 1.63, + "learning_rate": 1.456836771935851e-05, + "loss": 0.9829, + "step": 25412 + }, + { + "epoch": 1.64, + "learning_rate": 1.456815303033556e-05, + "loss": 0.9993, + "step": 25413 + }, + { + "epoch": 1.64, + "learning_rate": 1.4567938341312611e-05, + "loss": 1.0057, + "step": 25414 + }, + { + "epoch": 1.64, + "learning_rate": 1.456772365228966e-05, + "loss": 0.9043, + "step": 25415 + }, + { + "epoch": 1.64, + "learning_rate": 1.456750896326671e-05, + "loss": 0.9594, + "step": 25416 + }, + { + "epoch": 1.64, + "learning_rate": 1.456729427424376e-05, + "loss": 0.9481, + "step": 25417 + }, + { + "epoch": 1.64, + "learning_rate": 1.4567079585220809e-05, + "loss": 1.0543, + "step": 25418 + }, + { + "epoch": 1.64, + "learning_rate": 1.456686489619786e-05, + "loss": 0.9962, + "step": 25419 + }, + { + "epoch": 1.64, + "learning_rate": 1.4566650207174908e-05, + "loss": 0.9822, + "step": 25420 + }, + { + "epoch": 1.64, + "learning_rate": 1.4566435518151957e-05, + "loss": 0.9424, + "step": 25421 + }, + { + "epoch": 1.64, + "learning_rate": 1.4566220829129008e-05, + "loss": 0.9317, + "step": 25422 + }, + { + "epoch": 1.64, + "learning_rate": 1.4566006140106057e-05, + "loss": 0.9059, + "step": 25423 + }, + { + "epoch": 1.64, + "learning_rate": 1.4565791451083107e-05, + "loss": 0.9244, + "step": 25424 + }, + { + "epoch": 1.64, + "learning_rate": 1.4565576762060156e-05, + "loss": 0.9787, + "step": 25425 + }, + { + "epoch": 1.64, + "learning_rate": 1.4565362073037205e-05, + "loss": 0.9835, + "step": 25426 + }, + { + "epoch": 1.64, + "learning_rate": 1.4565147384014258e-05, + "loss": 1.0093, + "step": 25427 + }, + { + "epoch": 1.64, + "learning_rate": 1.4564932694991307e-05, + "loss": 0.9865, + "step": 25428 + }, + { + "epoch": 1.64, + "learning_rate": 1.4564718005968357e-05, + "loss": 1.0995, + "step": 25429 + }, + { + "epoch": 1.64, + "learning_rate": 1.4564503316945406e-05, + "loss": 1.0541, + "step": 25430 + }, + { + "epoch": 1.64, + "learning_rate": 1.4564288627922455e-05, + "loss": 0.9241, + "step": 25431 + }, + { + "epoch": 1.64, + "learning_rate": 1.4564073938899506e-05, + "loss": 0.9848, + "step": 25432 + }, + { + "epoch": 1.64, + "learning_rate": 1.4563859249876555e-05, + "loss": 1.0616, + "step": 25433 + }, + { + "epoch": 1.64, + "learning_rate": 1.4563644560853606e-05, + "loss": 0.917, + "step": 25434 + }, + { + "epoch": 1.64, + "learning_rate": 1.4563429871830655e-05, + "loss": 1.0077, + "step": 25435 + }, + { + "epoch": 1.64, + "learning_rate": 1.4563215182807703e-05, + "loss": 0.9769, + "step": 25436 + }, + { + "epoch": 1.64, + "learning_rate": 1.4563000493784754e-05, + "loss": 0.9676, + "step": 25437 + }, + { + "epoch": 1.64, + "learning_rate": 1.4562785804761803e-05, + "loss": 0.9294, + "step": 25438 + }, + { + "epoch": 1.64, + "learning_rate": 1.4562571115738852e-05, + "loss": 0.9818, + "step": 25439 + }, + { + "epoch": 1.64, + "learning_rate": 1.4562356426715903e-05, + "loss": 0.927, + "step": 25440 + }, + { + "epoch": 1.64, + "learning_rate": 1.4562141737692952e-05, + "loss": 1.0314, + "step": 25441 + }, + { + "epoch": 1.64, + "learning_rate": 1.4561927048670004e-05, + "loss": 0.9092, + "step": 25442 + }, + { + "epoch": 1.64, + "learning_rate": 1.4561712359647053e-05, + "loss": 0.9507, + "step": 25443 + }, + { + "epoch": 1.64, + "learning_rate": 1.4561497670624104e-05, + "loss": 1.0419, + "step": 25444 + }, + { + "epoch": 1.64, + "learning_rate": 1.4561282981601153e-05, + "loss": 1.0158, + "step": 25445 + }, + { + "epoch": 1.64, + "learning_rate": 1.4561068292578202e-05, + "loss": 0.8981, + "step": 25446 + }, + { + "epoch": 1.64, + "learning_rate": 1.4560853603555252e-05, + "loss": 0.9776, + "step": 25447 + }, + { + "epoch": 1.64, + "learning_rate": 1.4560638914532301e-05, + "loss": 0.957, + "step": 25448 + }, + { + "epoch": 1.64, + "learning_rate": 1.456042422550935e-05, + "loss": 0.9898, + "step": 25449 + }, + { + "epoch": 1.64, + "learning_rate": 1.4560209536486401e-05, + "loss": 0.9253, + "step": 25450 + }, + { + "epoch": 1.64, + "learning_rate": 1.455999484746345e-05, + "loss": 0.9871, + "step": 25451 + }, + { + "epoch": 1.64, + "learning_rate": 1.4559780158440499e-05, + "loss": 0.9273, + "step": 25452 + }, + { + "epoch": 1.64, + "learning_rate": 1.455956546941755e-05, + "loss": 0.9598, + "step": 25453 + }, + { + "epoch": 1.64, + "learning_rate": 1.4559350780394598e-05, + "loss": 0.9864, + "step": 25454 + }, + { + "epoch": 1.64, + "learning_rate": 1.4559136091371649e-05, + "loss": 0.9262, + "step": 25455 + }, + { + "epoch": 1.64, + "learning_rate": 1.4558921402348698e-05, + "loss": 1.046, + "step": 25456 + }, + { + "epoch": 1.64, + "learning_rate": 1.455870671332575e-05, + "loss": 0.929, + "step": 25457 + }, + { + "epoch": 1.64, + "learning_rate": 1.45584920243028e-05, + "loss": 0.9749, + "step": 25458 + }, + { + "epoch": 1.64, + "learning_rate": 1.4558277335279848e-05, + "loss": 0.9509, + "step": 25459 + }, + { + "epoch": 1.64, + "learning_rate": 1.4558062646256899e-05, + "loss": 0.9123, + "step": 25460 + }, + { + "epoch": 1.64, + "learning_rate": 1.4557847957233948e-05, + "loss": 0.9143, + "step": 25461 + }, + { + "epoch": 1.64, + "learning_rate": 1.4557633268210999e-05, + "loss": 0.9856, + "step": 25462 + }, + { + "epoch": 1.64, + "learning_rate": 1.4557418579188048e-05, + "loss": 1.0116, + "step": 25463 + }, + { + "epoch": 1.64, + "learning_rate": 1.4557203890165097e-05, + "loss": 1.0191, + "step": 25464 + }, + { + "epoch": 1.64, + "learning_rate": 1.4556989201142147e-05, + "loss": 1.0029, + "step": 25465 + }, + { + "epoch": 1.64, + "learning_rate": 1.4556774512119196e-05, + "loss": 1.0017, + "step": 25466 + }, + { + "epoch": 1.64, + "learning_rate": 1.4556559823096245e-05, + "loss": 1.0002, + "step": 25467 + }, + { + "epoch": 1.64, + "learning_rate": 1.4556345134073296e-05, + "loss": 0.9894, + "step": 25468 + }, + { + "epoch": 1.64, + "learning_rate": 1.4556130445050345e-05, + "loss": 0.9984, + "step": 25469 + }, + { + "epoch": 1.64, + "learning_rate": 1.4555915756027394e-05, + "loss": 0.9583, + "step": 25470 + }, + { + "epoch": 1.64, + "learning_rate": 1.4555701067004446e-05, + "loss": 0.923, + "step": 25471 + }, + { + "epoch": 1.64, + "learning_rate": 1.4555486377981497e-05, + "loss": 0.9122, + "step": 25472 + }, + { + "epoch": 1.64, + "learning_rate": 1.4555271688958546e-05, + "loss": 1.0272, + "step": 25473 + }, + { + "epoch": 1.64, + "learning_rate": 1.4555056999935595e-05, + "loss": 1.0449, + "step": 25474 + }, + { + "epoch": 1.64, + "learning_rate": 1.4554842310912645e-05, + "loss": 0.9965, + "step": 25475 + }, + { + "epoch": 1.64, + "learning_rate": 1.4554627621889694e-05, + "loss": 1.0286, + "step": 25476 + }, + { + "epoch": 1.64, + "learning_rate": 1.4554412932866743e-05, + "loss": 1.0154, + "step": 25477 + }, + { + "epoch": 1.64, + "learning_rate": 1.4554198243843794e-05, + "loss": 0.946, + "step": 25478 + }, + { + "epoch": 1.64, + "learning_rate": 1.4553983554820843e-05, + "loss": 0.9775, + "step": 25479 + }, + { + "epoch": 1.64, + "learning_rate": 1.4553768865797892e-05, + "loss": 0.9704, + "step": 25480 + }, + { + "epoch": 1.64, + "learning_rate": 1.4553554176774942e-05, + "loss": 1.0397, + "step": 25481 + }, + { + "epoch": 1.64, + "learning_rate": 1.4553339487751991e-05, + "loss": 1.0003, + "step": 25482 + }, + { + "epoch": 1.64, + "learning_rate": 1.4553124798729042e-05, + "loss": 1.0039, + "step": 25483 + }, + { + "epoch": 1.64, + "learning_rate": 1.4552910109706091e-05, + "loss": 0.9751, + "step": 25484 + }, + { + "epoch": 1.64, + "learning_rate": 1.455269542068314e-05, + "loss": 0.9586, + "step": 25485 + }, + { + "epoch": 1.64, + "learning_rate": 1.4552480731660192e-05, + "loss": 0.9874, + "step": 25486 + }, + { + "epoch": 1.64, + "learning_rate": 1.4552266042637241e-05, + "loss": 0.9782, + "step": 25487 + }, + { + "epoch": 1.64, + "learning_rate": 1.4552051353614292e-05, + "loss": 0.9001, + "step": 25488 + }, + { + "epoch": 1.64, + "learning_rate": 1.4551836664591341e-05, + "loss": 0.9554, + "step": 25489 + }, + { + "epoch": 1.64, + "learning_rate": 1.455162197556839e-05, + "loss": 0.9094, + "step": 25490 + }, + { + "epoch": 1.64, + "learning_rate": 1.455140728654544e-05, + "loss": 0.9215, + "step": 25491 + }, + { + "epoch": 1.64, + "learning_rate": 1.455119259752249e-05, + "loss": 1.0063, + "step": 25492 + }, + { + "epoch": 1.64, + "learning_rate": 1.455097790849954e-05, + "loss": 0.9342, + "step": 25493 + }, + { + "epoch": 1.64, + "learning_rate": 1.455076321947659e-05, + "loss": 0.9438, + "step": 25494 + }, + { + "epoch": 1.64, + "learning_rate": 1.4550548530453638e-05, + "loss": 0.9746, + "step": 25495 + }, + { + "epoch": 1.64, + "learning_rate": 1.4550333841430689e-05, + "loss": 0.9952, + "step": 25496 + }, + { + "epoch": 1.64, + "learning_rate": 1.4550119152407738e-05, + "loss": 1.0008, + "step": 25497 + }, + { + "epoch": 1.64, + "learning_rate": 1.4549904463384787e-05, + "loss": 0.9819, + "step": 25498 + }, + { + "epoch": 1.64, + "learning_rate": 1.4549689774361837e-05, + "loss": 0.95, + "step": 25499 + }, + { + "epoch": 1.64, + "learning_rate": 1.4549475085338888e-05, + "loss": 0.9308, + "step": 25500 + }, + { + "epoch": 1.64, + "learning_rate": 1.4549260396315939e-05, + "loss": 0.9311, + "step": 25501 + }, + { + "epoch": 1.64, + "learning_rate": 1.4549045707292988e-05, + "loss": 1.0239, + "step": 25502 + }, + { + "epoch": 1.64, + "learning_rate": 1.4548831018270038e-05, + "loss": 0.9846, + "step": 25503 + }, + { + "epoch": 1.64, + "learning_rate": 1.4548616329247087e-05, + "loss": 0.9889, + "step": 25504 + }, + { + "epoch": 1.64, + "learning_rate": 1.4548401640224136e-05, + "loss": 0.9732, + "step": 25505 + }, + { + "epoch": 1.64, + "learning_rate": 1.4548186951201187e-05, + "loss": 0.9663, + "step": 25506 + }, + { + "epoch": 1.64, + "learning_rate": 1.4547972262178236e-05, + "loss": 0.9801, + "step": 25507 + }, + { + "epoch": 1.64, + "learning_rate": 1.4547757573155285e-05, + "loss": 0.9356, + "step": 25508 + }, + { + "epoch": 1.64, + "learning_rate": 1.4547542884132336e-05, + "loss": 0.9626, + "step": 25509 + }, + { + "epoch": 1.64, + "learning_rate": 1.4547328195109384e-05, + "loss": 0.9853, + "step": 25510 + }, + { + "epoch": 1.64, + "learning_rate": 1.4547113506086435e-05, + "loss": 0.9137, + "step": 25511 + }, + { + "epoch": 1.64, + "learning_rate": 1.4546898817063484e-05, + "loss": 1.0372, + "step": 25512 + }, + { + "epoch": 1.64, + "learning_rate": 1.4546684128040533e-05, + "loss": 1.0153, + "step": 25513 + }, + { + "epoch": 1.64, + "learning_rate": 1.4546469439017584e-05, + "loss": 1.0352, + "step": 25514 + }, + { + "epoch": 1.64, + "learning_rate": 1.4546254749994634e-05, + "loss": 0.8982, + "step": 25515 + }, + { + "epoch": 1.64, + "learning_rate": 1.4546040060971685e-05, + "loss": 0.8966, + "step": 25516 + }, + { + "epoch": 1.64, + "learning_rate": 1.4545825371948734e-05, + "loss": 0.9351, + "step": 25517 + }, + { + "epoch": 1.64, + "learning_rate": 1.4545610682925783e-05, + "loss": 1.0125, + "step": 25518 + }, + { + "epoch": 1.64, + "learning_rate": 1.4545395993902834e-05, + "loss": 1.0035, + "step": 25519 + }, + { + "epoch": 1.64, + "learning_rate": 1.4545181304879883e-05, + "loss": 0.9332, + "step": 25520 + }, + { + "epoch": 1.64, + "learning_rate": 1.4544966615856933e-05, + "loss": 0.8759, + "step": 25521 + }, + { + "epoch": 1.64, + "learning_rate": 1.4544751926833982e-05, + "loss": 0.8608, + "step": 25522 + }, + { + "epoch": 1.64, + "learning_rate": 1.4544537237811031e-05, + "loss": 1.0059, + "step": 25523 + }, + { + "epoch": 1.64, + "learning_rate": 1.4544322548788082e-05, + "loss": 1.0378, + "step": 25524 + }, + { + "epoch": 1.64, + "learning_rate": 1.454410785976513e-05, + "loss": 0.9651, + "step": 25525 + }, + { + "epoch": 1.64, + "learning_rate": 1.454389317074218e-05, + "loss": 0.9553, + "step": 25526 + }, + { + "epoch": 1.64, + "learning_rate": 1.454367848171923e-05, + "loss": 0.9481, + "step": 25527 + }, + { + "epoch": 1.64, + "learning_rate": 1.454346379269628e-05, + "loss": 0.9968, + "step": 25528 + }, + { + "epoch": 1.64, + "learning_rate": 1.4543249103673332e-05, + "loss": 0.9729, + "step": 25529 + }, + { + "epoch": 1.64, + "learning_rate": 1.454303441465038e-05, + "loss": 0.9161, + "step": 25530 + }, + { + "epoch": 1.64, + "learning_rate": 1.4542819725627431e-05, + "loss": 1.0511, + "step": 25531 + }, + { + "epoch": 1.64, + "learning_rate": 1.454260503660448e-05, + "loss": 0.9011, + "step": 25532 + }, + { + "epoch": 1.64, + "learning_rate": 1.454239034758153e-05, + "loss": 0.9319, + "step": 25533 + }, + { + "epoch": 1.64, + "learning_rate": 1.454217565855858e-05, + "loss": 0.9722, + "step": 25534 + }, + { + "epoch": 1.64, + "learning_rate": 1.4541960969535629e-05, + "loss": 0.9924, + "step": 25535 + }, + { + "epoch": 1.64, + "learning_rate": 1.4541746280512678e-05, + "loss": 0.969, + "step": 25536 + }, + { + "epoch": 1.64, + "learning_rate": 1.4541531591489729e-05, + "loss": 1.0564, + "step": 25537 + }, + { + "epoch": 1.64, + "learning_rate": 1.4541316902466778e-05, + "loss": 0.9608, + "step": 25538 + }, + { + "epoch": 1.64, + "learning_rate": 1.4541102213443827e-05, + "loss": 0.9112, + "step": 25539 + }, + { + "epoch": 1.64, + "learning_rate": 1.4540887524420877e-05, + "loss": 0.9819, + "step": 25540 + }, + { + "epoch": 1.64, + "learning_rate": 1.4540672835397926e-05, + "loss": 0.9962, + "step": 25541 + }, + { + "epoch": 1.64, + "learning_rate": 1.4540458146374977e-05, + "loss": 1.0191, + "step": 25542 + }, + { + "epoch": 1.64, + "learning_rate": 1.4540243457352026e-05, + "loss": 0.9736, + "step": 25543 + }, + { + "epoch": 1.64, + "learning_rate": 1.4540028768329078e-05, + "loss": 0.9242, + "step": 25544 + }, + { + "epoch": 1.64, + "learning_rate": 1.4539814079306127e-05, + "loss": 0.9265, + "step": 25545 + }, + { + "epoch": 1.64, + "learning_rate": 1.4539599390283176e-05, + "loss": 1.0055, + "step": 25546 + }, + { + "epoch": 1.64, + "learning_rate": 1.4539384701260227e-05, + "loss": 0.9206, + "step": 25547 + }, + { + "epoch": 1.64, + "learning_rate": 1.4539170012237276e-05, + "loss": 0.973, + "step": 25548 + }, + { + "epoch": 1.64, + "learning_rate": 1.4538955323214325e-05, + "loss": 0.9808, + "step": 25549 + }, + { + "epoch": 1.64, + "learning_rate": 1.4538740634191375e-05, + "loss": 0.9342, + "step": 25550 + }, + { + "epoch": 1.64, + "learning_rate": 1.4538525945168424e-05, + "loss": 1.0168, + "step": 25551 + }, + { + "epoch": 1.64, + "learning_rate": 1.4538311256145475e-05, + "loss": 1.0063, + "step": 25552 + }, + { + "epoch": 1.64, + "learning_rate": 1.4538096567122524e-05, + "loss": 0.9892, + "step": 25553 + }, + { + "epoch": 1.64, + "learning_rate": 1.4537881878099573e-05, + "loss": 0.9202, + "step": 25554 + }, + { + "epoch": 1.64, + "learning_rate": 1.4537667189076623e-05, + "loss": 0.9766, + "step": 25555 + }, + { + "epoch": 1.64, + "learning_rate": 1.4537452500053672e-05, + "loss": 0.9823, + "step": 25556 + }, + { + "epoch": 1.64, + "learning_rate": 1.4537237811030721e-05, + "loss": 1.002, + "step": 25557 + }, + { + "epoch": 1.64, + "learning_rate": 1.4537023122007774e-05, + "loss": 0.892, + "step": 25558 + }, + { + "epoch": 1.64, + "learning_rate": 1.4536808432984823e-05, + "loss": 0.9207, + "step": 25559 + }, + { + "epoch": 1.64, + "learning_rate": 1.4536593743961873e-05, + "loss": 0.9064, + "step": 25560 + }, + { + "epoch": 1.64, + "learning_rate": 1.4536379054938922e-05, + "loss": 1.0115, + "step": 25561 + }, + { + "epoch": 1.64, + "learning_rate": 1.4536164365915973e-05, + "loss": 1.0521, + "step": 25562 + }, + { + "epoch": 1.64, + "learning_rate": 1.4535949676893022e-05, + "loss": 1.0402, + "step": 25563 + }, + { + "epoch": 1.64, + "learning_rate": 1.4535734987870071e-05, + "loss": 1.0217, + "step": 25564 + }, + { + "epoch": 1.64, + "learning_rate": 1.4535520298847122e-05, + "loss": 0.9117, + "step": 25565 + }, + { + "epoch": 1.64, + "learning_rate": 1.453530560982417e-05, + "loss": 1.0426, + "step": 25566 + }, + { + "epoch": 1.64, + "learning_rate": 1.453509092080122e-05, + "loss": 1.0153, + "step": 25567 + }, + { + "epoch": 1.64, + "learning_rate": 1.453487623177827e-05, + "loss": 0.999, + "step": 25568 + }, + { + "epoch": 1.65, + "learning_rate": 1.453466154275532e-05, + "loss": 0.9802, + "step": 25569 + }, + { + "epoch": 1.65, + "learning_rate": 1.453444685373237e-05, + "loss": 0.9121, + "step": 25570 + }, + { + "epoch": 1.65, + "learning_rate": 1.4534232164709419e-05, + "loss": 1.0162, + "step": 25571 + }, + { + "epoch": 1.65, + "learning_rate": 1.4534017475686468e-05, + "loss": 0.9921, + "step": 25572 + }, + { + "epoch": 1.65, + "learning_rate": 1.453380278666352e-05, + "loss": 0.9786, + "step": 25573 + }, + { + "epoch": 1.65, + "learning_rate": 1.4533588097640569e-05, + "loss": 0.9756, + "step": 25574 + }, + { + "epoch": 1.65, + "learning_rate": 1.453337340861762e-05, + "loss": 0.9568, + "step": 25575 + }, + { + "epoch": 1.65, + "learning_rate": 1.4533158719594669e-05, + "loss": 0.9847, + "step": 25576 + }, + { + "epoch": 1.65, + "learning_rate": 1.4532944030571718e-05, + "loss": 0.9629, + "step": 25577 + }, + { + "epoch": 1.65, + "learning_rate": 1.4532729341548768e-05, + "loss": 0.9927, + "step": 25578 + }, + { + "epoch": 1.65, + "learning_rate": 1.4532514652525817e-05, + "loss": 0.9101, + "step": 25579 + }, + { + "epoch": 1.65, + "learning_rate": 1.4532299963502868e-05, + "loss": 0.9824, + "step": 25580 + }, + { + "epoch": 1.65, + "learning_rate": 1.4532085274479917e-05, + "loss": 0.9744, + "step": 25581 + }, + { + "epoch": 1.65, + "learning_rate": 1.4531870585456966e-05, + "loss": 0.9956, + "step": 25582 + }, + { + "epoch": 1.65, + "learning_rate": 1.4531655896434017e-05, + "loss": 1.0192, + "step": 25583 + }, + { + "epoch": 1.65, + "learning_rate": 1.4531441207411065e-05, + "loss": 0.9372, + "step": 25584 + }, + { + "epoch": 1.65, + "learning_rate": 1.4531226518388114e-05, + "loss": 0.9159, + "step": 25585 + }, + { + "epoch": 1.65, + "learning_rate": 1.4531011829365165e-05, + "loss": 0.9502, + "step": 25586 + }, + { + "epoch": 1.65, + "learning_rate": 1.4530797140342216e-05, + "loss": 0.9581, + "step": 25587 + }, + { + "epoch": 1.65, + "learning_rate": 1.4530582451319266e-05, + "loss": 0.9607, + "step": 25588 + }, + { + "epoch": 1.65, + "learning_rate": 1.4530367762296315e-05, + "loss": 0.9592, + "step": 25589 + }, + { + "epoch": 1.65, + "learning_rate": 1.4530153073273366e-05, + "loss": 0.8667, + "step": 25590 + }, + { + "epoch": 1.65, + "learning_rate": 1.4529938384250415e-05, + "loss": 1.0247, + "step": 25591 + }, + { + "epoch": 1.65, + "learning_rate": 1.4529723695227464e-05, + "loss": 0.9884, + "step": 25592 + }, + { + "epoch": 1.65, + "learning_rate": 1.4529509006204515e-05, + "loss": 0.9964, + "step": 25593 + }, + { + "epoch": 1.65, + "learning_rate": 1.4529294317181564e-05, + "loss": 0.9899, + "step": 25594 + }, + { + "epoch": 1.65, + "learning_rate": 1.4529079628158613e-05, + "loss": 0.9719, + "step": 25595 + }, + { + "epoch": 1.65, + "learning_rate": 1.4528864939135663e-05, + "loss": 1.0676, + "step": 25596 + }, + { + "epoch": 1.65, + "learning_rate": 1.4528650250112712e-05, + "loss": 0.9698, + "step": 25597 + }, + { + "epoch": 1.65, + "learning_rate": 1.4528435561089763e-05, + "loss": 0.9295, + "step": 25598 + }, + { + "epoch": 1.65, + "learning_rate": 1.4528220872066812e-05, + "loss": 0.9453, + "step": 25599 + }, + { + "epoch": 1.65, + "learning_rate": 1.452800618304386e-05, + "loss": 1.0384, + "step": 25600 + }, + { + "epoch": 1.65, + "learning_rate": 1.4527791494020911e-05, + "loss": 1.0868, + "step": 25601 + }, + { + "epoch": 1.65, + "learning_rate": 1.4527576804997962e-05, + "loss": 0.9435, + "step": 25602 + }, + { + "epoch": 1.65, + "learning_rate": 1.4527362115975013e-05, + "loss": 0.941, + "step": 25603 + }, + { + "epoch": 1.65, + "learning_rate": 1.4527147426952062e-05, + "loss": 0.9958, + "step": 25604 + }, + { + "epoch": 1.65, + "learning_rate": 1.452693273792911e-05, + "loss": 0.9598, + "step": 25605 + }, + { + "epoch": 1.65, + "learning_rate": 1.4526718048906161e-05, + "loss": 0.9651, + "step": 25606 + }, + { + "epoch": 1.65, + "learning_rate": 1.452650335988321e-05, + "loss": 0.9745, + "step": 25607 + }, + { + "epoch": 1.65, + "learning_rate": 1.4526288670860261e-05, + "loss": 1.0045, + "step": 25608 + }, + { + "epoch": 1.65, + "learning_rate": 1.452607398183731e-05, + "loss": 0.9397, + "step": 25609 + }, + { + "epoch": 1.65, + "learning_rate": 1.4525859292814359e-05, + "loss": 0.9827, + "step": 25610 + }, + { + "epoch": 1.65, + "learning_rate": 1.452564460379141e-05, + "loss": 0.9317, + "step": 25611 + }, + { + "epoch": 1.65, + "learning_rate": 1.4525429914768459e-05, + "loss": 0.9616, + "step": 25612 + }, + { + "epoch": 1.65, + "learning_rate": 1.4525215225745508e-05, + "loss": 0.9761, + "step": 25613 + }, + { + "epoch": 1.65, + "learning_rate": 1.4525000536722558e-05, + "loss": 1.0081, + "step": 25614 + }, + { + "epoch": 1.65, + "learning_rate": 1.4524785847699607e-05, + "loss": 0.9676, + "step": 25615 + }, + { + "epoch": 1.65, + "learning_rate": 1.452457115867666e-05, + "loss": 0.9677, + "step": 25616 + }, + { + "epoch": 1.65, + "learning_rate": 1.4524356469653708e-05, + "loss": 0.9515, + "step": 25617 + }, + { + "epoch": 1.65, + "learning_rate": 1.4524141780630759e-05, + "loss": 1.0362, + "step": 25618 + }, + { + "epoch": 1.65, + "learning_rate": 1.4523927091607808e-05, + "loss": 0.9274, + "step": 25619 + }, + { + "epoch": 1.65, + "learning_rate": 1.4523712402584857e-05, + "loss": 1.034, + "step": 25620 + }, + { + "epoch": 1.65, + "learning_rate": 1.4523497713561908e-05, + "loss": 0.909, + "step": 25621 + }, + { + "epoch": 1.65, + "learning_rate": 1.4523283024538957e-05, + "loss": 0.9463, + "step": 25622 + }, + { + "epoch": 1.65, + "learning_rate": 1.4523068335516006e-05, + "loss": 1.016, + "step": 25623 + }, + { + "epoch": 1.65, + "learning_rate": 1.4522853646493056e-05, + "loss": 0.9625, + "step": 25624 + }, + { + "epoch": 1.65, + "learning_rate": 1.4522638957470105e-05, + "loss": 0.9519, + "step": 25625 + }, + { + "epoch": 1.65, + "learning_rate": 1.4522424268447154e-05, + "loss": 1.0271, + "step": 25626 + }, + { + "epoch": 1.65, + "learning_rate": 1.4522209579424205e-05, + "loss": 0.9372, + "step": 25627 + }, + { + "epoch": 1.65, + "learning_rate": 1.4521994890401254e-05, + "loss": 1.0302, + "step": 25628 + }, + { + "epoch": 1.65, + "learning_rate": 1.4521780201378304e-05, + "loss": 0.9719, + "step": 25629 + }, + { + "epoch": 1.65, + "learning_rate": 1.4521565512355353e-05, + "loss": 0.9711, + "step": 25630 + }, + { + "epoch": 1.65, + "learning_rate": 1.4521350823332406e-05, + "loss": 0.977, + "step": 25631 + }, + { + "epoch": 1.65, + "learning_rate": 1.4521136134309455e-05, + "loss": 0.9528, + "step": 25632 + }, + { + "epoch": 1.65, + "learning_rate": 1.4520921445286504e-05, + "loss": 1.0394, + "step": 25633 + }, + { + "epoch": 1.65, + "learning_rate": 1.4520706756263554e-05, + "loss": 0.9428, + "step": 25634 + }, + { + "epoch": 1.65, + "learning_rate": 1.4520492067240603e-05, + "loss": 1.0091, + "step": 25635 + }, + { + "epoch": 1.65, + "learning_rate": 1.4520277378217652e-05, + "loss": 0.9159, + "step": 25636 + }, + { + "epoch": 1.65, + "learning_rate": 1.4520062689194703e-05, + "loss": 1.0299, + "step": 25637 + }, + { + "epoch": 1.65, + "learning_rate": 1.4519848000171752e-05, + "loss": 0.9211, + "step": 25638 + }, + { + "epoch": 1.65, + "learning_rate": 1.4519633311148803e-05, + "loss": 1.0535, + "step": 25639 + }, + { + "epoch": 1.65, + "learning_rate": 1.4519418622125852e-05, + "loss": 1.0121, + "step": 25640 + }, + { + "epoch": 1.65, + "learning_rate": 1.45192039331029e-05, + "loss": 0.9684, + "step": 25641 + }, + { + "epoch": 1.65, + "learning_rate": 1.4518989244079951e-05, + "loss": 0.9325, + "step": 25642 + }, + { + "epoch": 1.65, + "learning_rate": 1.4518774555057e-05, + "loss": 0.9894, + "step": 25643 + }, + { + "epoch": 1.65, + "learning_rate": 1.4518559866034049e-05, + "loss": 0.9614, + "step": 25644 + }, + { + "epoch": 1.65, + "learning_rate": 1.4518345177011101e-05, + "loss": 0.9422, + "step": 25645 + }, + { + "epoch": 1.65, + "learning_rate": 1.451813048798815e-05, + "loss": 0.9592, + "step": 25646 + }, + { + "epoch": 1.65, + "learning_rate": 1.4517915798965201e-05, + "loss": 0.9573, + "step": 25647 + }, + { + "epoch": 1.65, + "learning_rate": 1.451770110994225e-05, + "loss": 0.9027, + "step": 25648 + }, + { + "epoch": 1.65, + "learning_rate": 1.45174864209193e-05, + "loss": 0.9481, + "step": 25649 + }, + { + "epoch": 1.65, + "learning_rate": 1.451727173189635e-05, + "loss": 1.0375, + "step": 25650 + }, + { + "epoch": 1.65, + "learning_rate": 1.4517057042873399e-05, + "loss": 0.9812, + "step": 25651 + }, + { + "epoch": 1.65, + "learning_rate": 1.451684235385045e-05, + "loss": 0.9489, + "step": 25652 + }, + { + "epoch": 1.65, + "learning_rate": 1.4516627664827498e-05, + "loss": 0.9822, + "step": 25653 + }, + { + "epoch": 1.65, + "learning_rate": 1.4516412975804547e-05, + "loss": 0.978, + "step": 25654 + }, + { + "epoch": 1.65, + "learning_rate": 1.4516198286781598e-05, + "loss": 0.9262, + "step": 25655 + }, + { + "epoch": 1.65, + "learning_rate": 1.4515983597758647e-05, + "loss": 1.0036, + "step": 25656 + }, + { + "epoch": 1.65, + "learning_rate": 1.4515768908735698e-05, + "loss": 0.9723, + "step": 25657 + }, + { + "epoch": 1.65, + "learning_rate": 1.4515554219712747e-05, + "loss": 0.9721, + "step": 25658 + }, + { + "epoch": 1.65, + "learning_rate": 1.4515339530689795e-05, + "loss": 0.9583, + "step": 25659 + }, + { + "epoch": 1.65, + "learning_rate": 1.4515124841666848e-05, + "loss": 1.0023, + "step": 25660 + }, + { + "epoch": 1.65, + "learning_rate": 1.4514910152643897e-05, + "loss": 1.0782, + "step": 25661 + }, + { + "epoch": 1.65, + "learning_rate": 1.4514695463620947e-05, + "loss": 0.8845, + "step": 25662 + }, + { + "epoch": 1.65, + "learning_rate": 1.4514480774597996e-05, + "loss": 0.9801, + "step": 25663 + }, + { + "epoch": 1.65, + "learning_rate": 1.4514266085575045e-05, + "loss": 1.0028, + "step": 25664 + }, + { + "epoch": 1.65, + "learning_rate": 1.4514051396552096e-05, + "loss": 0.9657, + "step": 25665 + }, + { + "epoch": 1.65, + "learning_rate": 1.4513836707529145e-05, + "loss": 0.9905, + "step": 25666 + }, + { + "epoch": 1.65, + "learning_rate": 1.4513622018506196e-05, + "loss": 0.9744, + "step": 25667 + }, + { + "epoch": 1.65, + "learning_rate": 1.4513407329483245e-05, + "loss": 0.9285, + "step": 25668 + }, + { + "epoch": 1.65, + "learning_rate": 1.4513192640460294e-05, + "loss": 0.9519, + "step": 25669 + }, + { + "epoch": 1.65, + "learning_rate": 1.4512977951437344e-05, + "loss": 0.9736, + "step": 25670 + }, + { + "epoch": 1.65, + "learning_rate": 1.4512763262414393e-05, + "loss": 1.0507, + "step": 25671 + }, + { + "epoch": 1.65, + "learning_rate": 1.4512548573391442e-05, + "loss": 1.044, + "step": 25672 + }, + { + "epoch": 1.65, + "learning_rate": 1.4512333884368493e-05, + "loss": 0.9517, + "step": 25673 + }, + { + "epoch": 1.65, + "learning_rate": 1.4512119195345543e-05, + "loss": 1.0075, + "step": 25674 + }, + { + "epoch": 1.65, + "learning_rate": 1.4511904506322594e-05, + "loss": 1.0664, + "step": 25675 + }, + { + "epoch": 1.65, + "learning_rate": 1.4511689817299643e-05, + "loss": 0.9904, + "step": 25676 + }, + { + "epoch": 1.65, + "learning_rate": 1.4511475128276694e-05, + "loss": 0.9902, + "step": 25677 + }, + { + "epoch": 1.65, + "learning_rate": 1.4511260439253743e-05, + "loss": 0.9718, + "step": 25678 + }, + { + "epoch": 1.65, + "learning_rate": 1.4511045750230792e-05, + "loss": 0.9323, + "step": 25679 + }, + { + "epoch": 1.65, + "learning_rate": 1.4510831061207842e-05, + "loss": 0.9411, + "step": 25680 + }, + { + "epoch": 1.65, + "learning_rate": 1.4510616372184891e-05, + "loss": 1.0183, + "step": 25681 + }, + { + "epoch": 1.65, + "learning_rate": 1.451040168316194e-05, + "loss": 0.9144, + "step": 25682 + }, + { + "epoch": 1.65, + "learning_rate": 1.4510186994138991e-05, + "loss": 0.9874, + "step": 25683 + }, + { + "epoch": 1.65, + "learning_rate": 1.450997230511604e-05, + "loss": 0.9873, + "step": 25684 + }, + { + "epoch": 1.65, + "learning_rate": 1.450975761609309e-05, + "loss": 0.9449, + "step": 25685 + }, + { + "epoch": 1.65, + "learning_rate": 1.450954292707014e-05, + "loss": 1.0223, + "step": 25686 + }, + { + "epoch": 1.65, + "learning_rate": 1.4509328238047189e-05, + "loss": 1.0105, + "step": 25687 + }, + { + "epoch": 1.65, + "learning_rate": 1.450911354902424e-05, + "loss": 0.9898, + "step": 25688 + }, + { + "epoch": 1.65, + "learning_rate": 1.450889886000129e-05, + "loss": 1.0188, + "step": 25689 + }, + { + "epoch": 1.65, + "learning_rate": 1.450868417097834e-05, + "loss": 0.947, + "step": 25690 + }, + { + "epoch": 1.65, + "learning_rate": 1.450846948195539e-05, + "loss": 1.0122, + "step": 25691 + }, + { + "epoch": 1.65, + "learning_rate": 1.4508254792932438e-05, + "loss": 0.9615, + "step": 25692 + }, + { + "epoch": 1.65, + "learning_rate": 1.4508040103909489e-05, + "loss": 0.9598, + "step": 25693 + }, + { + "epoch": 1.65, + "learning_rate": 1.4507825414886538e-05, + "loss": 0.9013, + "step": 25694 + }, + { + "epoch": 1.65, + "learning_rate": 1.4507610725863589e-05, + "loss": 0.9756, + "step": 25695 + }, + { + "epoch": 1.65, + "learning_rate": 1.4507396036840638e-05, + "loss": 0.9832, + "step": 25696 + }, + { + "epoch": 1.65, + "learning_rate": 1.4507181347817687e-05, + "loss": 0.8961, + "step": 25697 + }, + { + "epoch": 1.65, + "learning_rate": 1.4506966658794737e-05, + "loss": 1.0525, + "step": 25698 + }, + { + "epoch": 1.65, + "learning_rate": 1.4506751969771786e-05, + "loss": 0.9273, + "step": 25699 + }, + { + "epoch": 1.65, + "learning_rate": 1.4506537280748835e-05, + "loss": 0.978, + "step": 25700 + }, + { + "epoch": 1.65, + "learning_rate": 1.4506322591725886e-05, + "loss": 0.9904, + "step": 25701 + }, + { + "epoch": 1.65, + "learning_rate": 1.4506107902702935e-05, + "loss": 0.9766, + "step": 25702 + }, + { + "epoch": 1.65, + "learning_rate": 1.4505893213679987e-05, + "loss": 0.983, + "step": 25703 + }, + { + "epoch": 1.65, + "learning_rate": 1.4505678524657036e-05, + "loss": 1.0199, + "step": 25704 + }, + { + "epoch": 1.65, + "learning_rate": 1.4505463835634087e-05, + "loss": 1.01, + "step": 25705 + }, + { + "epoch": 1.65, + "learning_rate": 1.4505249146611136e-05, + "loss": 1.0318, + "step": 25706 + }, + { + "epoch": 1.65, + "learning_rate": 1.4505034457588185e-05, + "loss": 1.0528, + "step": 25707 + }, + { + "epoch": 1.65, + "learning_rate": 1.4504819768565235e-05, + "loss": 0.8961, + "step": 25708 + }, + { + "epoch": 1.65, + "learning_rate": 1.4504605079542284e-05, + "loss": 0.9836, + "step": 25709 + }, + { + "epoch": 1.65, + "learning_rate": 1.4504390390519333e-05, + "loss": 0.9286, + "step": 25710 + }, + { + "epoch": 1.65, + "learning_rate": 1.4504175701496384e-05, + "loss": 0.9296, + "step": 25711 + }, + { + "epoch": 1.65, + "learning_rate": 1.4503961012473433e-05, + "loss": 0.9079, + "step": 25712 + }, + { + "epoch": 1.65, + "learning_rate": 1.4503746323450482e-05, + "loss": 0.9953, + "step": 25713 + }, + { + "epoch": 1.65, + "learning_rate": 1.4503531634427533e-05, + "loss": 0.9834, + "step": 25714 + }, + { + "epoch": 1.65, + "learning_rate": 1.4503316945404582e-05, + "loss": 0.9881, + "step": 25715 + }, + { + "epoch": 1.65, + "learning_rate": 1.4503102256381632e-05, + "loss": 0.9646, + "step": 25716 + }, + { + "epoch": 1.65, + "learning_rate": 1.4502887567358681e-05, + "loss": 0.9405, + "step": 25717 + }, + { + "epoch": 1.65, + "learning_rate": 1.4502672878335734e-05, + "loss": 0.9877, + "step": 25718 + }, + { + "epoch": 1.65, + "learning_rate": 1.4502458189312782e-05, + "loss": 0.915, + "step": 25719 + }, + { + "epoch": 1.65, + "learning_rate": 1.4502243500289831e-05, + "loss": 0.9297, + "step": 25720 + }, + { + "epoch": 1.65, + "learning_rate": 1.4502028811266882e-05, + "loss": 0.9436, + "step": 25721 + }, + { + "epoch": 1.65, + "learning_rate": 1.4501814122243931e-05, + "loss": 0.9009, + "step": 25722 + }, + { + "epoch": 1.65, + "learning_rate": 1.450159943322098e-05, + "loss": 1.0168, + "step": 25723 + }, + { + "epoch": 1.66, + "learning_rate": 1.450138474419803e-05, + "loss": 0.9734, + "step": 25724 + }, + { + "epoch": 1.66, + "learning_rate": 1.450117005517508e-05, + "loss": 0.946, + "step": 25725 + }, + { + "epoch": 1.66, + "learning_rate": 1.450095536615213e-05, + "loss": 0.9438, + "step": 25726 + }, + { + "epoch": 1.66, + "learning_rate": 1.450074067712918e-05, + "loss": 0.9785, + "step": 25727 + }, + { + "epoch": 1.66, + "learning_rate": 1.4500525988106228e-05, + "loss": 1.0298, + "step": 25728 + }, + { + "epoch": 1.66, + "learning_rate": 1.4500311299083279e-05, + "loss": 0.9813, + "step": 25729 + }, + { + "epoch": 1.66, + "learning_rate": 1.4500096610060328e-05, + "loss": 0.9432, + "step": 25730 + }, + { + "epoch": 1.66, + "learning_rate": 1.4499881921037377e-05, + "loss": 0.9901, + "step": 25731 + }, + { + "epoch": 1.66, + "learning_rate": 1.449966723201443e-05, + "loss": 0.9743, + "step": 25732 + }, + { + "epoch": 1.66, + "learning_rate": 1.4499452542991478e-05, + "loss": 0.9917, + "step": 25733 + }, + { + "epoch": 1.66, + "learning_rate": 1.4499237853968529e-05, + "loss": 0.9946, + "step": 25734 + }, + { + "epoch": 1.66, + "learning_rate": 1.4499023164945578e-05, + "loss": 1.0022, + "step": 25735 + }, + { + "epoch": 1.66, + "learning_rate": 1.4498808475922628e-05, + "loss": 0.9571, + "step": 25736 + }, + { + "epoch": 1.66, + "learning_rate": 1.4498593786899677e-05, + "loss": 0.9406, + "step": 25737 + }, + { + "epoch": 1.66, + "learning_rate": 1.4498379097876726e-05, + "loss": 0.9346, + "step": 25738 + }, + { + "epoch": 1.66, + "learning_rate": 1.4498164408853777e-05, + "loss": 1.0038, + "step": 25739 + }, + { + "epoch": 1.66, + "learning_rate": 1.4497949719830826e-05, + "loss": 0.8952, + "step": 25740 + }, + { + "epoch": 1.66, + "learning_rate": 1.4497735030807875e-05, + "loss": 0.9943, + "step": 25741 + }, + { + "epoch": 1.66, + "learning_rate": 1.4497520341784926e-05, + "loss": 1.0097, + "step": 25742 + }, + { + "epoch": 1.66, + "learning_rate": 1.4497305652761975e-05, + "loss": 0.9423, + "step": 25743 + }, + { + "epoch": 1.66, + "learning_rate": 1.4497090963739025e-05, + "loss": 1.0417, + "step": 25744 + }, + { + "epoch": 1.66, + "learning_rate": 1.4496876274716074e-05, + "loss": 0.9513, + "step": 25745 + }, + { + "epoch": 1.66, + "learning_rate": 1.4496661585693123e-05, + "loss": 1.0424, + "step": 25746 + }, + { + "epoch": 1.66, + "learning_rate": 1.4496446896670176e-05, + "loss": 0.9941, + "step": 25747 + }, + { + "epoch": 1.66, + "learning_rate": 1.4496232207647224e-05, + "loss": 0.9839, + "step": 25748 + }, + { + "epoch": 1.66, + "learning_rate": 1.4496017518624275e-05, + "loss": 0.8591, + "step": 25749 + }, + { + "epoch": 1.66, + "learning_rate": 1.4495802829601324e-05, + "loss": 1.0268, + "step": 25750 + }, + { + "epoch": 1.66, + "learning_rate": 1.4495588140578373e-05, + "loss": 0.9749, + "step": 25751 + }, + { + "epoch": 1.66, + "learning_rate": 1.4495373451555424e-05, + "loss": 0.9692, + "step": 25752 + }, + { + "epoch": 1.66, + "learning_rate": 1.4495158762532473e-05, + "loss": 0.9751, + "step": 25753 + }, + { + "epoch": 1.66, + "learning_rate": 1.4494944073509523e-05, + "loss": 0.9446, + "step": 25754 + }, + { + "epoch": 1.66, + "learning_rate": 1.4494729384486572e-05, + "loss": 1.005, + "step": 25755 + }, + { + "epoch": 1.66, + "learning_rate": 1.4494514695463621e-05, + "loss": 0.9563, + "step": 25756 + }, + { + "epoch": 1.66, + "learning_rate": 1.4494300006440672e-05, + "loss": 0.9028, + "step": 25757 + }, + { + "epoch": 1.66, + "learning_rate": 1.4494085317417721e-05, + "loss": 0.8746, + "step": 25758 + }, + { + "epoch": 1.66, + "learning_rate": 1.449387062839477e-05, + "loss": 0.9236, + "step": 25759 + }, + { + "epoch": 1.66, + "learning_rate": 1.449365593937182e-05, + "loss": 1.0294, + "step": 25760 + }, + { + "epoch": 1.66, + "learning_rate": 1.4493441250348871e-05, + "loss": 1.0055, + "step": 25761 + }, + { + "epoch": 1.66, + "learning_rate": 1.4493226561325922e-05, + "loss": 0.9112, + "step": 25762 + }, + { + "epoch": 1.66, + "learning_rate": 1.449301187230297e-05, + "loss": 0.9195, + "step": 25763 + }, + { + "epoch": 1.66, + "learning_rate": 1.4492797183280021e-05, + "loss": 0.9202, + "step": 25764 + }, + { + "epoch": 1.66, + "learning_rate": 1.449258249425707e-05, + "loss": 0.9316, + "step": 25765 + }, + { + "epoch": 1.66, + "learning_rate": 1.449236780523412e-05, + "loss": 1.0116, + "step": 25766 + }, + { + "epoch": 1.66, + "learning_rate": 1.449215311621117e-05, + "loss": 1.0226, + "step": 25767 + }, + { + "epoch": 1.66, + "learning_rate": 1.4491938427188219e-05, + "loss": 0.8521, + "step": 25768 + }, + { + "epoch": 1.66, + "learning_rate": 1.4491723738165268e-05, + "loss": 0.9852, + "step": 25769 + }, + { + "epoch": 1.66, + "learning_rate": 1.4491509049142319e-05, + "loss": 0.9039, + "step": 25770 + }, + { + "epoch": 1.66, + "learning_rate": 1.4491294360119368e-05, + "loss": 0.9999, + "step": 25771 + }, + { + "epoch": 1.66, + "learning_rate": 1.4491079671096418e-05, + "loss": 0.9122, + "step": 25772 + }, + { + "epoch": 1.66, + "learning_rate": 1.4490864982073467e-05, + "loss": 0.9195, + "step": 25773 + }, + { + "epoch": 1.66, + "learning_rate": 1.4490650293050516e-05, + "loss": 0.9064, + "step": 25774 + }, + { + "epoch": 1.66, + "learning_rate": 1.4490435604027567e-05, + "loss": 0.9186, + "step": 25775 + }, + { + "epoch": 1.66, + "learning_rate": 1.4490220915004618e-05, + "loss": 0.95, + "step": 25776 + }, + { + "epoch": 1.66, + "learning_rate": 1.4490006225981668e-05, + "loss": 0.9498, + "step": 25777 + }, + { + "epoch": 1.66, + "learning_rate": 1.4489791536958717e-05, + "loss": 0.9473, + "step": 25778 + }, + { + "epoch": 1.66, + "learning_rate": 1.4489576847935766e-05, + "loss": 0.9681, + "step": 25779 + }, + { + "epoch": 1.66, + "learning_rate": 1.4489362158912817e-05, + "loss": 0.9498, + "step": 25780 + }, + { + "epoch": 1.66, + "learning_rate": 1.4489147469889866e-05, + "loss": 1.0009, + "step": 25781 + }, + { + "epoch": 1.66, + "learning_rate": 1.4488932780866916e-05, + "loss": 0.9386, + "step": 25782 + }, + { + "epoch": 1.66, + "learning_rate": 1.4488718091843965e-05, + "loss": 0.9882, + "step": 25783 + }, + { + "epoch": 1.66, + "learning_rate": 1.4488503402821014e-05, + "loss": 0.9968, + "step": 25784 + }, + { + "epoch": 1.66, + "learning_rate": 1.4488288713798065e-05, + "loss": 1.0243, + "step": 25785 + }, + { + "epoch": 1.66, + "learning_rate": 1.4488074024775114e-05, + "loss": 0.9226, + "step": 25786 + }, + { + "epoch": 1.66, + "learning_rate": 1.4487859335752163e-05, + "loss": 0.9218, + "step": 25787 + }, + { + "epoch": 1.66, + "learning_rate": 1.4487644646729214e-05, + "loss": 0.8915, + "step": 25788 + }, + { + "epoch": 1.66, + "learning_rate": 1.4487429957706263e-05, + "loss": 1.019, + "step": 25789 + }, + { + "epoch": 1.66, + "learning_rate": 1.4487215268683315e-05, + "loss": 1.0395, + "step": 25790 + }, + { + "epoch": 1.66, + "learning_rate": 1.4487000579660364e-05, + "loss": 1.0486, + "step": 25791 + }, + { + "epoch": 1.66, + "learning_rate": 1.4486785890637415e-05, + "loss": 0.9364, + "step": 25792 + }, + { + "epoch": 1.66, + "learning_rate": 1.4486571201614463e-05, + "loss": 0.8822, + "step": 25793 + }, + { + "epoch": 1.66, + "learning_rate": 1.4486356512591512e-05, + "loss": 1.0581, + "step": 25794 + }, + { + "epoch": 1.66, + "learning_rate": 1.4486141823568563e-05, + "loss": 0.9675, + "step": 25795 + }, + { + "epoch": 1.66, + "learning_rate": 1.4485927134545612e-05, + "loss": 0.976, + "step": 25796 + }, + { + "epoch": 1.66, + "learning_rate": 1.4485712445522661e-05, + "loss": 1.0085, + "step": 25797 + }, + { + "epoch": 1.66, + "learning_rate": 1.4485497756499712e-05, + "loss": 0.9724, + "step": 25798 + }, + { + "epoch": 1.66, + "learning_rate": 1.448528306747676e-05, + "loss": 0.999, + "step": 25799 + }, + { + "epoch": 1.66, + "learning_rate": 1.448506837845381e-05, + "loss": 1.11, + "step": 25800 + }, + { + "epoch": 1.66, + "learning_rate": 1.448485368943086e-05, + "loss": 0.9269, + "step": 25801 + }, + { + "epoch": 1.66, + "learning_rate": 1.448463900040791e-05, + "loss": 0.9871, + "step": 25802 + }, + { + "epoch": 1.66, + "learning_rate": 1.448442431138496e-05, + "loss": 0.9835, + "step": 25803 + }, + { + "epoch": 1.66, + "learning_rate": 1.4484209622362009e-05, + "loss": 0.8414, + "step": 25804 + }, + { + "epoch": 1.66, + "learning_rate": 1.4483994933339061e-05, + "loss": 1.0196, + "step": 25805 + }, + { + "epoch": 1.66, + "learning_rate": 1.448378024431611e-05, + "loss": 0.9454, + "step": 25806 + }, + { + "epoch": 1.66, + "learning_rate": 1.448356555529316e-05, + "loss": 0.9889, + "step": 25807 + }, + { + "epoch": 1.66, + "learning_rate": 1.448335086627021e-05, + "loss": 1.0193, + "step": 25808 + }, + { + "epoch": 1.66, + "learning_rate": 1.4483136177247259e-05, + "loss": 1.0247, + "step": 25809 + }, + { + "epoch": 1.66, + "learning_rate": 1.4482921488224308e-05, + "loss": 0.896, + "step": 25810 + }, + { + "epoch": 1.66, + "learning_rate": 1.4482706799201358e-05, + "loss": 0.9188, + "step": 25811 + }, + { + "epoch": 1.66, + "learning_rate": 1.4482492110178407e-05, + "loss": 0.9178, + "step": 25812 + }, + { + "epoch": 1.66, + "learning_rate": 1.4482277421155458e-05, + "loss": 0.9145, + "step": 25813 + }, + { + "epoch": 1.66, + "learning_rate": 1.4482062732132507e-05, + "loss": 1.0097, + "step": 25814 + }, + { + "epoch": 1.66, + "learning_rate": 1.4481848043109556e-05, + "loss": 0.9327, + "step": 25815 + }, + { + "epoch": 1.66, + "learning_rate": 1.4481633354086607e-05, + "loss": 0.955, + "step": 25816 + }, + { + "epoch": 1.66, + "learning_rate": 1.4481418665063656e-05, + "loss": 1.0374, + "step": 25817 + }, + { + "epoch": 1.66, + "learning_rate": 1.4481203976040705e-05, + "loss": 0.9882, + "step": 25818 + }, + { + "epoch": 1.66, + "learning_rate": 1.4480989287017755e-05, + "loss": 0.985, + "step": 25819 + }, + { + "epoch": 1.66, + "learning_rate": 1.4480774597994806e-05, + "loss": 0.9404, + "step": 25820 + }, + { + "epoch": 1.66, + "learning_rate": 1.4480559908971857e-05, + "loss": 1.005, + "step": 25821 + }, + { + "epoch": 1.66, + "learning_rate": 1.4480345219948906e-05, + "loss": 0.9359, + "step": 25822 + }, + { + "epoch": 1.66, + "learning_rate": 1.4480130530925956e-05, + "loss": 0.9951, + "step": 25823 + }, + { + "epoch": 1.66, + "learning_rate": 1.4479915841903005e-05, + "loss": 0.9355, + "step": 25824 + }, + { + "epoch": 1.66, + "learning_rate": 1.4479701152880054e-05, + "loss": 0.8954, + "step": 25825 + }, + { + "epoch": 1.66, + "learning_rate": 1.4479486463857105e-05, + "loss": 1.0522, + "step": 25826 + }, + { + "epoch": 1.66, + "learning_rate": 1.4479271774834154e-05, + "loss": 0.9185, + "step": 25827 + }, + { + "epoch": 1.66, + "learning_rate": 1.4479057085811203e-05, + "loss": 0.9437, + "step": 25828 + }, + { + "epoch": 1.66, + "learning_rate": 1.4478842396788253e-05, + "loss": 1.0209, + "step": 25829 + }, + { + "epoch": 1.66, + "learning_rate": 1.4478627707765302e-05, + "loss": 0.9785, + "step": 25830 + }, + { + "epoch": 1.66, + "learning_rate": 1.4478413018742353e-05, + "loss": 0.9473, + "step": 25831 + }, + { + "epoch": 1.66, + "learning_rate": 1.4478198329719402e-05, + "loss": 0.9434, + "step": 25832 + }, + { + "epoch": 1.66, + "learning_rate": 1.4477983640696451e-05, + "loss": 1.0292, + "step": 25833 + }, + { + "epoch": 1.66, + "learning_rate": 1.4477768951673503e-05, + "loss": 0.9619, + "step": 25834 + }, + { + "epoch": 1.66, + "learning_rate": 1.4477554262650552e-05, + "loss": 0.9697, + "step": 25835 + }, + { + "epoch": 1.66, + "learning_rate": 1.4477339573627603e-05, + "loss": 1.012, + "step": 25836 + }, + { + "epoch": 1.66, + "learning_rate": 1.4477124884604652e-05, + "loss": 0.9286, + "step": 25837 + }, + { + "epoch": 1.66, + "learning_rate": 1.44769101955817e-05, + "loss": 0.9792, + "step": 25838 + }, + { + "epoch": 1.66, + "learning_rate": 1.4476695506558751e-05, + "loss": 0.933, + "step": 25839 + }, + { + "epoch": 1.66, + "learning_rate": 1.44764808175358e-05, + "loss": 0.9403, + "step": 25840 + }, + { + "epoch": 1.66, + "learning_rate": 1.4476266128512851e-05, + "loss": 0.9391, + "step": 25841 + }, + { + "epoch": 1.66, + "learning_rate": 1.44760514394899e-05, + "loss": 0.9464, + "step": 25842 + }, + { + "epoch": 1.66, + "learning_rate": 1.4475836750466949e-05, + "loss": 0.976, + "step": 25843 + }, + { + "epoch": 1.66, + "learning_rate": 1.4475622061444e-05, + "loss": 0.9344, + "step": 25844 + }, + { + "epoch": 1.66, + "learning_rate": 1.4475407372421049e-05, + "loss": 0.8898, + "step": 25845 + }, + { + "epoch": 1.66, + "learning_rate": 1.4475192683398098e-05, + "loss": 0.9628, + "step": 25846 + }, + { + "epoch": 1.66, + "learning_rate": 1.4474977994375148e-05, + "loss": 0.9387, + "step": 25847 + }, + { + "epoch": 1.66, + "learning_rate": 1.4474763305352197e-05, + "loss": 0.9239, + "step": 25848 + }, + { + "epoch": 1.66, + "learning_rate": 1.447454861632925e-05, + "loss": 0.9622, + "step": 25849 + }, + { + "epoch": 1.66, + "learning_rate": 1.4474333927306299e-05, + "loss": 0.9432, + "step": 25850 + }, + { + "epoch": 1.66, + "learning_rate": 1.447411923828335e-05, + "loss": 0.9444, + "step": 25851 + }, + { + "epoch": 1.66, + "learning_rate": 1.4473904549260398e-05, + "loss": 0.9186, + "step": 25852 + }, + { + "epoch": 1.66, + "learning_rate": 1.4473689860237447e-05, + "loss": 0.936, + "step": 25853 + }, + { + "epoch": 1.66, + "learning_rate": 1.4473475171214498e-05, + "loss": 0.9466, + "step": 25854 + }, + { + "epoch": 1.66, + "learning_rate": 1.4473260482191547e-05, + "loss": 0.9709, + "step": 25855 + }, + { + "epoch": 1.66, + "learning_rate": 1.4473045793168596e-05, + "loss": 0.9636, + "step": 25856 + }, + { + "epoch": 1.66, + "learning_rate": 1.4472831104145646e-05, + "loss": 0.9563, + "step": 25857 + }, + { + "epoch": 1.66, + "learning_rate": 1.4472616415122695e-05, + "loss": 0.9625, + "step": 25858 + }, + { + "epoch": 1.66, + "learning_rate": 1.4472401726099744e-05, + "loss": 0.9532, + "step": 25859 + }, + { + "epoch": 1.66, + "learning_rate": 1.4472187037076795e-05, + "loss": 1.002, + "step": 25860 + }, + { + "epoch": 1.66, + "learning_rate": 1.4471972348053844e-05, + "loss": 1.0166, + "step": 25861 + }, + { + "epoch": 1.66, + "learning_rate": 1.4471757659030895e-05, + "loss": 0.9799, + "step": 25862 + }, + { + "epoch": 1.66, + "learning_rate": 1.4471542970007945e-05, + "loss": 0.9704, + "step": 25863 + }, + { + "epoch": 1.66, + "learning_rate": 1.4471328280984996e-05, + "loss": 1.0361, + "step": 25864 + }, + { + "epoch": 1.66, + "learning_rate": 1.4471113591962045e-05, + "loss": 0.9545, + "step": 25865 + }, + { + "epoch": 1.66, + "learning_rate": 1.4470898902939094e-05, + "loss": 0.9939, + "step": 25866 + }, + { + "epoch": 1.66, + "learning_rate": 1.4470684213916144e-05, + "loss": 0.9608, + "step": 25867 + }, + { + "epoch": 1.66, + "learning_rate": 1.4470469524893193e-05, + "loss": 0.9804, + "step": 25868 + }, + { + "epoch": 1.66, + "learning_rate": 1.4470254835870242e-05, + "loss": 0.9734, + "step": 25869 + }, + { + "epoch": 1.66, + "learning_rate": 1.4470040146847293e-05, + "loss": 0.9749, + "step": 25870 + }, + { + "epoch": 1.66, + "learning_rate": 1.4469825457824342e-05, + "loss": 1.0154, + "step": 25871 + }, + { + "epoch": 1.66, + "learning_rate": 1.4469610768801393e-05, + "loss": 0.9426, + "step": 25872 + }, + { + "epoch": 1.66, + "learning_rate": 1.4469396079778442e-05, + "loss": 0.9132, + "step": 25873 + }, + { + "epoch": 1.66, + "learning_rate": 1.446918139075549e-05, + "loss": 0.9779, + "step": 25874 + }, + { + "epoch": 1.66, + "learning_rate": 1.4468966701732541e-05, + "loss": 1.0005, + "step": 25875 + }, + { + "epoch": 1.66, + "learning_rate": 1.446875201270959e-05, + "loss": 0.9977, + "step": 25876 + }, + { + "epoch": 1.66, + "learning_rate": 1.446853732368664e-05, + "loss": 0.9447, + "step": 25877 + }, + { + "epoch": 1.66, + "learning_rate": 1.4468322634663692e-05, + "loss": 0.9648, + "step": 25878 + }, + { + "epoch": 1.66, + "learning_rate": 1.446810794564074e-05, + "loss": 1.0243, + "step": 25879 + }, + { + "epoch": 1.67, + "learning_rate": 1.4467893256617791e-05, + "loss": 0.9407, + "step": 25880 + }, + { + "epoch": 1.67, + "learning_rate": 1.446767856759484e-05, + "loss": 0.9404, + "step": 25881 + }, + { + "epoch": 1.67, + "learning_rate": 1.446746387857189e-05, + "loss": 0.9073, + "step": 25882 + }, + { + "epoch": 1.67, + "learning_rate": 1.446724918954894e-05, + "loss": 1.0796, + "step": 25883 + }, + { + "epoch": 1.67, + "learning_rate": 1.4467034500525989e-05, + "loss": 0.975, + "step": 25884 + }, + { + "epoch": 1.67, + "learning_rate": 1.446681981150304e-05, + "loss": 1.0188, + "step": 25885 + }, + { + "epoch": 1.67, + "learning_rate": 1.4466605122480088e-05, + "loss": 0.9654, + "step": 25886 + }, + { + "epoch": 1.67, + "learning_rate": 1.4466390433457137e-05, + "loss": 0.9735, + "step": 25887 + }, + { + "epoch": 1.67, + "learning_rate": 1.4466175744434188e-05, + "loss": 0.9185, + "step": 25888 + }, + { + "epoch": 1.67, + "learning_rate": 1.4465961055411237e-05, + "loss": 0.9405, + "step": 25889 + }, + { + "epoch": 1.67, + "learning_rate": 1.4465746366388288e-05, + "loss": 1.008, + "step": 25890 + }, + { + "epoch": 1.67, + "learning_rate": 1.4465531677365337e-05, + "loss": 0.9824, + "step": 25891 + }, + { + "epoch": 1.67, + "learning_rate": 1.4465316988342389e-05, + "loss": 0.9832, + "step": 25892 + }, + { + "epoch": 1.67, + "learning_rate": 1.4465102299319438e-05, + "loss": 1.0199, + "step": 25893 + }, + { + "epoch": 1.67, + "learning_rate": 1.4464887610296487e-05, + "loss": 0.9913, + "step": 25894 + }, + { + "epoch": 1.67, + "learning_rate": 1.4464672921273538e-05, + "loss": 1.0458, + "step": 25895 + }, + { + "epoch": 1.67, + "learning_rate": 1.4464458232250587e-05, + "loss": 0.9883, + "step": 25896 + }, + { + "epoch": 1.67, + "learning_rate": 1.4464243543227635e-05, + "loss": 0.9682, + "step": 25897 + }, + { + "epoch": 1.67, + "learning_rate": 1.4464028854204686e-05, + "loss": 0.9666, + "step": 25898 + }, + { + "epoch": 1.67, + "learning_rate": 1.4463814165181735e-05, + "loss": 0.9808, + "step": 25899 + }, + { + "epoch": 1.67, + "learning_rate": 1.4463599476158786e-05, + "loss": 0.9071, + "step": 25900 + }, + { + "epoch": 1.67, + "learning_rate": 1.4463384787135835e-05, + "loss": 1.007, + "step": 25901 + }, + { + "epoch": 1.67, + "learning_rate": 1.4463170098112884e-05, + "loss": 1.0175, + "step": 25902 + }, + { + "epoch": 1.67, + "learning_rate": 1.4462955409089934e-05, + "loss": 0.9664, + "step": 25903 + }, + { + "epoch": 1.67, + "learning_rate": 1.4462740720066983e-05, + "loss": 0.9882, + "step": 25904 + }, + { + "epoch": 1.67, + "learning_rate": 1.4462526031044032e-05, + "loss": 0.9627, + "step": 25905 + }, + { + "epoch": 1.67, + "learning_rate": 1.4462311342021083e-05, + "loss": 1.1012, + "step": 25906 + }, + { + "epoch": 1.67, + "learning_rate": 1.4462096652998134e-05, + "loss": 0.9902, + "step": 25907 + }, + { + "epoch": 1.67, + "learning_rate": 1.4461881963975184e-05, + "loss": 0.9727, + "step": 25908 + }, + { + "epoch": 1.67, + "learning_rate": 1.4461667274952233e-05, + "loss": 0.9979, + "step": 25909 + }, + { + "epoch": 1.67, + "learning_rate": 1.4461452585929284e-05, + "loss": 0.9612, + "step": 25910 + }, + { + "epoch": 1.67, + "learning_rate": 1.4461237896906333e-05, + "loss": 0.9222, + "step": 25911 + }, + { + "epoch": 1.67, + "learning_rate": 1.4461023207883382e-05, + "loss": 0.9487, + "step": 25912 + }, + { + "epoch": 1.67, + "learning_rate": 1.4460808518860432e-05, + "loss": 0.9718, + "step": 25913 + }, + { + "epoch": 1.67, + "learning_rate": 1.4460593829837481e-05, + "loss": 0.992, + "step": 25914 + }, + { + "epoch": 1.67, + "learning_rate": 1.446037914081453e-05, + "loss": 0.9921, + "step": 25915 + }, + { + "epoch": 1.67, + "learning_rate": 1.4460164451791581e-05, + "loss": 0.9249, + "step": 25916 + }, + { + "epoch": 1.67, + "learning_rate": 1.445994976276863e-05, + "loss": 0.9445, + "step": 25917 + }, + { + "epoch": 1.67, + "learning_rate": 1.445973507374568e-05, + "loss": 1.0263, + "step": 25918 + }, + { + "epoch": 1.67, + "learning_rate": 1.445952038472273e-05, + "loss": 0.96, + "step": 25919 + }, + { + "epoch": 1.67, + "learning_rate": 1.4459305695699779e-05, + "loss": 0.9801, + "step": 25920 + }, + { + "epoch": 1.67, + "learning_rate": 1.4459091006676831e-05, + "loss": 0.9299, + "step": 25921 + }, + { + "epoch": 1.67, + "learning_rate": 1.445887631765388e-05, + "loss": 0.9257, + "step": 25922 + }, + { + "epoch": 1.67, + "learning_rate": 1.445866162863093e-05, + "loss": 1.0155, + "step": 25923 + }, + { + "epoch": 1.67, + "learning_rate": 1.445844693960798e-05, + "loss": 1.0515, + "step": 25924 + }, + { + "epoch": 1.67, + "learning_rate": 1.4458232250585029e-05, + "loss": 0.936, + "step": 25925 + }, + { + "epoch": 1.67, + "learning_rate": 1.445801756156208e-05, + "loss": 0.9751, + "step": 25926 + }, + { + "epoch": 1.67, + "learning_rate": 1.4457802872539128e-05, + "loss": 0.8965, + "step": 25927 + }, + { + "epoch": 1.67, + "learning_rate": 1.4457588183516179e-05, + "loss": 0.9693, + "step": 25928 + }, + { + "epoch": 1.67, + "learning_rate": 1.4457373494493228e-05, + "loss": 0.9867, + "step": 25929 + }, + { + "epoch": 1.67, + "learning_rate": 1.4457158805470277e-05, + "loss": 1.0295, + "step": 25930 + }, + { + "epoch": 1.67, + "learning_rate": 1.4456944116447327e-05, + "loss": 0.9533, + "step": 25931 + }, + { + "epoch": 1.67, + "learning_rate": 1.4456729427424376e-05, + "loss": 0.9367, + "step": 25932 + }, + { + "epoch": 1.67, + "learning_rate": 1.4456514738401425e-05, + "loss": 0.9082, + "step": 25933 + }, + { + "epoch": 1.67, + "learning_rate": 1.4456300049378476e-05, + "loss": 1.0458, + "step": 25934 + }, + { + "epoch": 1.67, + "learning_rate": 1.4456085360355525e-05, + "loss": 0.9787, + "step": 25935 + }, + { + "epoch": 1.67, + "learning_rate": 1.4455870671332577e-05, + "loss": 0.9836, + "step": 25936 + }, + { + "epoch": 1.67, + "learning_rate": 1.4455655982309626e-05, + "loss": 1.0139, + "step": 25937 + }, + { + "epoch": 1.67, + "learning_rate": 1.4455441293286677e-05, + "loss": 0.9002, + "step": 25938 + }, + { + "epoch": 1.67, + "learning_rate": 1.4455226604263726e-05, + "loss": 1.0306, + "step": 25939 + }, + { + "epoch": 1.67, + "learning_rate": 1.4455011915240775e-05, + "loss": 0.9552, + "step": 25940 + }, + { + "epoch": 1.67, + "learning_rate": 1.4454797226217826e-05, + "loss": 0.974, + "step": 25941 + }, + { + "epoch": 1.67, + "learning_rate": 1.4454582537194874e-05, + "loss": 0.9278, + "step": 25942 + }, + { + "epoch": 1.67, + "learning_rate": 1.4454367848171923e-05, + "loss": 1.0333, + "step": 25943 + }, + { + "epoch": 1.67, + "learning_rate": 1.4454153159148974e-05, + "loss": 0.9621, + "step": 25944 + }, + { + "epoch": 1.67, + "learning_rate": 1.4453938470126023e-05, + "loss": 0.9844, + "step": 25945 + }, + { + "epoch": 1.67, + "learning_rate": 1.4453723781103072e-05, + "loss": 0.9656, + "step": 25946 + }, + { + "epoch": 1.67, + "learning_rate": 1.4453509092080123e-05, + "loss": 1.0362, + "step": 25947 + }, + { + "epoch": 1.67, + "learning_rate": 1.4453294403057172e-05, + "loss": 0.9343, + "step": 25948 + }, + { + "epoch": 1.67, + "learning_rate": 1.4453079714034222e-05, + "loss": 0.9048, + "step": 25949 + }, + { + "epoch": 1.67, + "learning_rate": 1.4452865025011273e-05, + "loss": 0.9642, + "step": 25950 + }, + { + "epoch": 1.67, + "learning_rate": 1.4452650335988324e-05, + "loss": 1.0301, + "step": 25951 + }, + { + "epoch": 1.67, + "learning_rate": 1.4452435646965373e-05, + "loss": 1.0179, + "step": 25952 + }, + { + "epoch": 1.67, + "learning_rate": 1.4452220957942422e-05, + "loss": 1.0001, + "step": 25953 + }, + { + "epoch": 1.67, + "learning_rate": 1.4452006268919472e-05, + "loss": 0.9602, + "step": 25954 + }, + { + "epoch": 1.67, + "learning_rate": 1.4451791579896521e-05, + "loss": 1.0314, + "step": 25955 + }, + { + "epoch": 1.67, + "learning_rate": 1.445157689087357e-05, + "loss": 1.0587, + "step": 25956 + }, + { + "epoch": 1.67, + "learning_rate": 1.445136220185062e-05, + "loss": 0.9597, + "step": 25957 + }, + { + "epoch": 1.67, + "learning_rate": 1.445114751282767e-05, + "loss": 0.9771, + "step": 25958 + }, + { + "epoch": 1.67, + "learning_rate": 1.445093282380472e-05, + "loss": 0.9488, + "step": 25959 + }, + { + "epoch": 1.67, + "learning_rate": 1.445071813478177e-05, + "loss": 0.9472, + "step": 25960 + }, + { + "epoch": 1.67, + "learning_rate": 1.4450503445758818e-05, + "loss": 0.9796, + "step": 25961 + }, + { + "epoch": 1.67, + "learning_rate": 1.4450288756735869e-05, + "loss": 0.9824, + "step": 25962 + }, + { + "epoch": 1.67, + "learning_rate": 1.4450074067712918e-05, + "loss": 0.9544, + "step": 25963 + }, + { + "epoch": 1.67, + "learning_rate": 1.4449859378689967e-05, + "loss": 0.9679, + "step": 25964 + }, + { + "epoch": 1.67, + "learning_rate": 1.444964468966702e-05, + "loss": 1.0239, + "step": 25965 + }, + { + "epoch": 1.67, + "learning_rate": 1.4449430000644068e-05, + "loss": 0.9582, + "step": 25966 + }, + { + "epoch": 1.67, + "learning_rate": 1.4449215311621119e-05, + "loss": 1.0717, + "step": 25967 + }, + { + "epoch": 1.67, + "learning_rate": 1.4449000622598168e-05, + "loss": 1.0435, + "step": 25968 + }, + { + "epoch": 1.67, + "learning_rate": 1.4448785933575219e-05, + "loss": 0.9862, + "step": 25969 + }, + { + "epoch": 1.67, + "learning_rate": 1.4448571244552268e-05, + "loss": 0.974, + "step": 25970 + }, + { + "epoch": 1.67, + "learning_rate": 1.4448356555529316e-05, + "loss": 0.9718, + "step": 25971 + }, + { + "epoch": 1.67, + "learning_rate": 1.4448141866506367e-05, + "loss": 0.9722, + "step": 25972 + }, + { + "epoch": 1.67, + "learning_rate": 1.4447927177483416e-05, + "loss": 0.9654, + "step": 25973 + }, + { + "epoch": 1.67, + "learning_rate": 1.4447712488460465e-05, + "loss": 0.9145, + "step": 25974 + }, + { + "epoch": 1.67, + "learning_rate": 1.4447497799437516e-05, + "loss": 1.0332, + "step": 25975 + }, + { + "epoch": 1.67, + "learning_rate": 1.4447283110414565e-05, + "loss": 0.9712, + "step": 25976 + }, + { + "epoch": 1.67, + "learning_rate": 1.4447068421391615e-05, + "loss": 0.9592, + "step": 25977 + }, + { + "epoch": 1.67, + "learning_rate": 1.4446853732368664e-05, + "loss": 0.9044, + "step": 25978 + }, + { + "epoch": 1.67, + "learning_rate": 1.4446639043345717e-05, + "loss": 0.9624, + "step": 25979 + }, + { + "epoch": 1.67, + "learning_rate": 1.4446424354322766e-05, + "loss": 0.9582, + "step": 25980 + }, + { + "epoch": 1.67, + "learning_rate": 1.4446209665299815e-05, + "loss": 1.0183, + "step": 25981 + }, + { + "epoch": 1.67, + "learning_rate": 1.4445994976276865e-05, + "loss": 1.0255, + "step": 25982 + }, + { + "epoch": 1.67, + "learning_rate": 1.4445780287253914e-05, + "loss": 0.8969, + "step": 25983 + }, + { + "epoch": 1.67, + "learning_rate": 1.4445565598230963e-05, + "loss": 1.0265, + "step": 25984 + }, + { + "epoch": 1.67, + "learning_rate": 1.4445350909208014e-05, + "loss": 0.9344, + "step": 25985 + }, + { + "epoch": 1.67, + "learning_rate": 1.4445136220185063e-05, + "loss": 0.9799, + "step": 25986 + }, + { + "epoch": 1.67, + "learning_rate": 1.4444921531162113e-05, + "loss": 0.9657, + "step": 25987 + }, + { + "epoch": 1.67, + "learning_rate": 1.4444706842139162e-05, + "loss": 0.8776, + "step": 25988 + }, + { + "epoch": 1.67, + "learning_rate": 1.4444492153116211e-05, + "loss": 0.9028, + "step": 25989 + }, + { + "epoch": 1.67, + "learning_rate": 1.4444277464093262e-05, + "loss": 0.9571, + "step": 25990 + }, + { + "epoch": 1.67, + "learning_rate": 1.4444062775070311e-05, + "loss": 0.9794, + "step": 25991 + }, + { + "epoch": 1.67, + "learning_rate": 1.444384808604736e-05, + "loss": 0.9809, + "step": 25992 + }, + { + "epoch": 1.67, + "learning_rate": 1.444363339702441e-05, + "loss": 0.9484, + "step": 25993 + }, + { + "epoch": 1.67, + "learning_rate": 1.4443418708001461e-05, + "loss": 0.9634, + "step": 25994 + }, + { + "epoch": 1.67, + "learning_rate": 1.4443204018978512e-05, + "loss": 0.9546, + "step": 25995 + }, + { + "epoch": 1.67, + "learning_rate": 1.4442989329955561e-05, + "loss": 0.9875, + "step": 25996 + }, + { + "epoch": 1.67, + "learning_rate": 1.4442774640932612e-05, + "loss": 1.0394, + "step": 25997 + }, + { + "epoch": 1.67, + "learning_rate": 1.444255995190966e-05, + "loss": 0.9239, + "step": 25998 + }, + { + "epoch": 1.67, + "learning_rate": 1.444234526288671e-05, + "loss": 0.9535, + "step": 25999 + }, + { + "epoch": 1.67, + "learning_rate": 1.444213057386376e-05, + "loss": 0.9629, + "step": 26000 + }, + { + "epoch": 1.67, + "eval_laion/cider": 0.5354723143995257, + "eval_laion/meteor": 0.39123690836011216, + "eval_laion/pred_num_words": 36.6953125, + "eval_laion/sacrebleu": 15.685677958823284, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.047519326210022, + "eval_runtime": 100.2227, + "eval_samples_per_second": 4.979, + "eval_steps_per_second": 0.319, + "step": 26000 + }, + { + "epoch": 1.67, + "learning_rate": 1.4441915884840809e-05, + "loss": 0.9567, + "step": 26001 + }, + { + "epoch": 1.67, + "learning_rate": 1.4441701195817858e-05, + "loss": 0.9441, + "step": 26002 + }, + { + "epoch": 1.67, + "learning_rate": 1.4441486506794909e-05, + "loss": 1.0007, + "step": 26003 + }, + { + "epoch": 1.67, + "learning_rate": 1.4441271817771958e-05, + "loss": 0.912, + "step": 26004 + }, + { + "epoch": 1.67, + "learning_rate": 1.4441057128749008e-05, + "loss": 1.0095, + "step": 26005 + }, + { + "epoch": 1.67, + "learning_rate": 1.4440842439726057e-05, + "loss": 0.9455, + "step": 26006 + }, + { + "epoch": 1.67, + "learning_rate": 1.4440627750703106e-05, + "loss": 0.9692, + "step": 26007 + }, + { + "epoch": 1.67, + "learning_rate": 1.4440413061680159e-05, + "loss": 0.9442, + "step": 26008 + }, + { + "epoch": 1.67, + "learning_rate": 1.4440198372657208e-05, + "loss": 0.9268, + "step": 26009 + }, + { + "epoch": 1.67, + "learning_rate": 1.4439983683634258e-05, + "loss": 0.957, + "step": 26010 + }, + { + "epoch": 1.67, + "learning_rate": 1.4439768994611307e-05, + "loss": 0.9959, + "step": 26011 + }, + { + "epoch": 1.67, + "learning_rate": 1.4439554305588356e-05, + "loss": 0.8946, + "step": 26012 + }, + { + "epoch": 1.67, + "learning_rate": 1.4439339616565407e-05, + "loss": 1.0957, + "step": 26013 + }, + { + "epoch": 1.67, + "learning_rate": 1.4439124927542456e-05, + "loss": 0.9918, + "step": 26014 + }, + { + "epoch": 1.67, + "learning_rate": 1.4438910238519507e-05, + "loss": 0.9934, + "step": 26015 + }, + { + "epoch": 1.67, + "learning_rate": 1.4438695549496555e-05, + "loss": 0.944, + "step": 26016 + }, + { + "epoch": 1.67, + "learning_rate": 1.4438480860473604e-05, + "loss": 0.9208, + "step": 26017 + }, + { + "epoch": 1.67, + "learning_rate": 1.4438266171450655e-05, + "loss": 1.0289, + "step": 26018 + }, + { + "epoch": 1.67, + "learning_rate": 1.4438051482427704e-05, + "loss": 0.9797, + "step": 26019 + }, + { + "epoch": 1.67, + "learning_rate": 1.4437836793404753e-05, + "loss": 1.0328, + "step": 26020 + }, + { + "epoch": 1.67, + "learning_rate": 1.4437622104381804e-05, + "loss": 0.9591, + "step": 26021 + }, + { + "epoch": 1.67, + "learning_rate": 1.4437407415358853e-05, + "loss": 0.8967, + "step": 26022 + }, + { + "epoch": 1.67, + "learning_rate": 1.4437192726335905e-05, + "loss": 1.0087, + "step": 26023 + }, + { + "epoch": 1.67, + "learning_rate": 1.4436978037312954e-05, + "loss": 1.0297, + "step": 26024 + }, + { + "epoch": 1.67, + "learning_rate": 1.4436763348290005e-05, + "loss": 0.9135, + "step": 26025 + }, + { + "epoch": 1.67, + "learning_rate": 1.4436548659267054e-05, + "loss": 1.0128, + "step": 26026 + }, + { + "epoch": 1.67, + "learning_rate": 1.4436333970244103e-05, + "loss": 0.9641, + "step": 26027 + }, + { + "epoch": 1.67, + "learning_rate": 1.4436119281221153e-05, + "loss": 0.9808, + "step": 26028 + }, + { + "epoch": 1.67, + "learning_rate": 1.4435904592198202e-05, + "loss": 1.0458, + "step": 26029 + }, + { + "epoch": 1.67, + "learning_rate": 1.4435689903175251e-05, + "loss": 0.9806, + "step": 26030 + }, + { + "epoch": 1.67, + "learning_rate": 1.4435475214152302e-05, + "loss": 0.9105, + "step": 26031 + }, + { + "epoch": 1.67, + "learning_rate": 1.443526052512935e-05, + "loss": 1.0084, + "step": 26032 + }, + { + "epoch": 1.67, + "learning_rate": 1.44350458361064e-05, + "loss": 1.0909, + "step": 26033 + }, + { + "epoch": 1.67, + "learning_rate": 1.443483114708345e-05, + "loss": 0.9914, + "step": 26034 + }, + { + "epoch": 1.68, + "learning_rate": 1.44346164580605e-05, + "loss": 0.977, + "step": 26035 + }, + { + "epoch": 1.68, + "learning_rate": 1.443440176903755e-05, + "loss": 0.969, + "step": 26036 + }, + { + "epoch": 1.68, + "learning_rate": 1.44341870800146e-05, + "loss": 0.9786, + "step": 26037 + }, + { + "epoch": 1.68, + "learning_rate": 1.4433972390991651e-05, + "loss": 0.8799, + "step": 26038 + }, + { + "epoch": 1.68, + "learning_rate": 1.44337577019687e-05, + "loss": 0.9341, + "step": 26039 + }, + { + "epoch": 1.68, + "learning_rate": 1.443354301294575e-05, + "loss": 1.0068, + "step": 26040 + }, + { + "epoch": 1.68, + "learning_rate": 1.44333283239228e-05, + "loss": 0.9302, + "step": 26041 + }, + { + "epoch": 1.68, + "learning_rate": 1.4433113634899849e-05, + "loss": 1.0397, + "step": 26042 + }, + { + "epoch": 1.68, + "learning_rate": 1.4432898945876898e-05, + "loss": 0.9518, + "step": 26043 + }, + { + "epoch": 1.68, + "learning_rate": 1.4432684256853949e-05, + "loss": 0.9223, + "step": 26044 + }, + { + "epoch": 1.68, + "learning_rate": 1.4432469567830997e-05, + "loss": 0.9972, + "step": 26045 + }, + { + "epoch": 1.68, + "learning_rate": 1.4432254878808048e-05, + "loss": 0.956, + "step": 26046 + }, + { + "epoch": 1.68, + "learning_rate": 1.4432040189785097e-05, + "loss": 0.9499, + "step": 26047 + }, + { + "epoch": 1.68, + "learning_rate": 1.4431825500762146e-05, + "loss": 1.019, + "step": 26048 + }, + { + "epoch": 1.68, + "learning_rate": 1.4431610811739197e-05, + "loss": 0.9177, + "step": 26049 + }, + { + "epoch": 1.68, + "learning_rate": 1.4431396122716246e-05, + "loss": 1.0339, + "step": 26050 + }, + { + "epoch": 1.68, + "learning_rate": 1.4431181433693295e-05, + "loss": 0.9709, + "step": 26051 + }, + { + "epoch": 1.68, + "learning_rate": 1.4430966744670347e-05, + "loss": 1.0056, + "step": 26052 + }, + { + "epoch": 1.68, + "learning_rate": 1.4430752055647396e-05, + "loss": 0.9801, + "step": 26053 + }, + { + "epoch": 1.68, + "learning_rate": 1.4430537366624447e-05, + "loss": 1.0066, + "step": 26054 + }, + { + "epoch": 1.68, + "learning_rate": 1.4430322677601496e-05, + "loss": 0.961, + "step": 26055 + }, + { + "epoch": 1.68, + "learning_rate": 1.4430107988578546e-05, + "loss": 0.9511, + "step": 26056 + }, + { + "epoch": 1.68, + "learning_rate": 1.4429893299555595e-05, + "loss": 0.998, + "step": 26057 + }, + { + "epoch": 1.68, + "learning_rate": 1.4429678610532644e-05, + "loss": 0.9674, + "step": 26058 + }, + { + "epoch": 1.68, + "learning_rate": 1.4429463921509695e-05, + "loss": 0.8625, + "step": 26059 + }, + { + "epoch": 1.68, + "learning_rate": 1.4429249232486744e-05, + "loss": 1.0893, + "step": 26060 + }, + { + "epoch": 1.68, + "learning_rate": 1.4429034543463793e-05, + "loss": 0.9319, + "step": 26061 + }, + { + "epoch": 1.68, + "learning_rate": 1.4428819854440843e-05, + "loss": 1.0217, + "step": 26062 + }, + { + "epoch": 1.68, + "learning_rate": 1.4428605165417892e-05, + "loss": 1.0401, + "step": 26063 + }, + { + "epoch": 1.68, + "learning_rate": 1.4428390476394943e-05, + "loss": 0.9596, + "step": 26064 + }, + { + "epoch": 1.68, + "learning_rate": 1.4428175787371992e-05, + "loss": 0.9625, + "step": 26065 + }, + { + "epoch": 1.68, + "learning_rate": 1.4427961098349044e-05, + "loss": 0.9876, + "step": 26066 + }, + { + "epoch": 1.68, + "learning_rate": 1.4427746409326093e-05, + "loss": 0.9014, + "step": 26067 + }, + { + "epoch": 1.68, + "learning_rate": 1.4427531720303142e-05, + "loss": 0.9927, + "step": 26068 + }, + { + "epoch": 1.68, + "learning_rate": 1.4427317031280193e-05, + "loss": 0.9809, + "step": 26069 + }, + { + "epoch": 1.68, + "learning_rate": 1.4427102342257242e-05, + "loss": 0.9604, + "step": 26070 + }, + { + "epoch": 1.68, + "learning_rate": 1.4426887653234291e-05, + "loss": 0.9463, + "step": 26071 + }, + { + "epoch": 1.68, + "learning_rate": 1.4426672964211342e-05, + "loss": 1.0038, + "step": 26072 + }, + { + "epoch": 1.68, + "learning_rate": 1.442645827518839e-05, + "loss": 1.0342, + "step": 26073 + }, + { + "epoch": 1.68, + "learning_rate": 1.4426243586165441e-05, + "loss": 1.0083, + "step": 26074 + }, + { + "epoch": 1.68, + "learning_rate": 1.442602889714249e-05, + "loss": 1.0226, + "step": 26075 + }, + { + "epoch": 1.68, + "learning_rate": 1.4425814208119539e-05, + "loss": 0.9061, + "step": 26076 + }, + { + "epoch": 1.68, + "learning_rate": 1.442559951909659e-05, + "loss": 0.9506, + "step": 26077 + }, + { + "epoch": 1.68, + "learning_rate": 1.4425384830073639e-05, + "loss": 0.99, + "step": 26078 + }, + { + "epoch": 1.68, + "learning_rate": 1.4425170141050688e-05, + "loss": 1.0045, + "step": 26079 + }, + { + "epoch": 1.68, + "learning_rate": 1.4424955452027738e-05, + "loss": 1.0757, + "step": 26080 + }, + { + "epoch": 1.68, + "learning_rate": 1.4424740763004789e-05, + "loss": 0.9377, + "step": 26081 + }, + { + "epoch": 1.68, + "learning_rate": 1.442452607398184e-05, + "loss": 0.9272, + "step": 26082 + }, + { + "epoch": 1.68, + "learning_rate": 1.4424311384958889e-05, + "loss": 1.0096, + "step": 26083 + }, + { + "epoch": 1.68, + "learning_rate": 1.442409669593594e-05, + "loss": 0.9624, + "step": 26084 + }, + { + "epoch": 1.68, + "learning_rate": 1.4423882006912988e-05, + "loss": 0.9667, + "step": 26085 + }, + { + "epoch": 1.68, + "learning_rate": 1.4423667317890037e-05, + "loss": 0.9923, + "step": 26086 + }, + { + "epoch": 1.68, + "learning_rate": 1.4423452628867088e-05, + "loss": 0.9999, + "step": 26087 + }, + { + "epoch": 1.68, + "learning_rate": 1.4423237939844137e-05, + "loss": 1.0471, + "step": 26088 + }, + { + "epoch": 1.68, + "learning_rate": 1.4423023250821186e-05, + "loss": 1.0002, + "step": 26089 + }, + { + "epoch": 1.68, + "learning_rate": 1.4422808561798236e-05, + "loss": 0.9658, + "step": 26090 + }, + { + "epoch": 1.68, + "learning_rate": 1.4422593872775285e-05, + "loss": 1.0806, + "step": 26091 + }, + { + "epoch": 1.68, + "learning_rate": 1.4422379183752336e-05, + "loss": 1.0198, + "step": 26092 + }, + { + "epoch": 1.68, + "learning_rate": 1.4422164494729385e-05, + "loss": 0.9463, + "step": 26093 + }, + { + "epoch": 1.68, + "learning_rate": 1.4421949805706434e-05, + "loss": 0.9383, + "step": 26094 + }, + { + "epoch": 1.68, + "learning_rate": 1.4421735116683486e-05, + "loss": 0.9728, + "step": 26095 + }, + { + "epoch": 1.68, + "learning_rate": 1.4421520427660535e-05, + "loss": 1.011, + "step": 26096 + }, + { + "epoch": 1.68, + "learning_rate": 1.4421305738637586e-05, + "loss": 0.9643, + "step": 26097 + }, + { + "epoch": 1.68, + "learning_rate": 1.4421091049614635e-05, + "loss": 1.0044, + "step": 26098 + }, + { + "epoch": 1.68, + "learning_rate": 1.4420876360591684e-05, + "loss": 0.9335, + "step": 26099 + }, + { + "epoch": 1.68, + "learning_rate": 1.4420661671568735e-05, + "loss": 1.0232, + "step": 26100 + }, + { + "epoch": 1.68, + "learning_rate": 1.4420446982545784e-05, + "loss": 0.9284, + "step": 26101 + }, + { + "epoch": 1.68, + "learning_rate": 1.4420232293522834e-05, + "loss": 1.018, + "step": 26102 + }, + { + "epoch": 1.68, + "learning_rate": 1.4420017604499883e-05, + "loss": 0.9075, + "step": 26103 + }, + { + "epoch": 1.68, + "learning_rate": 1.4419802915476932e-05, + "loss": 0.9045, + "step": 26104 + }, + { + "epoch": 1.68, + "learning_rate": 1.4419588226453983e-05, + "loss": 0.9326, + "step": 26105 + }, + { + "epoch": 1.68, + "learning_rate": 1.4419373537431032e-05, + "loss": 0.8844, + "step": 26106 + }, + { + "epoch": 1.68, + "learning_rate": 1.441915884840808e-05, + "loss": 1.0675, + "step": 26107 + }, + { + "epoch": 1.68, + "learning_rate": 1.4418944159385131e-05, + "loss": 0.9757, + "step": 26108 + }, + { + "epoch": 1.68, + "learning_rate": 1.441872947036218e-05, + "loss": 0.9229, + "step": 26109 + }, + { + "epoch": 1.68, + "learning_rate": 1.4418514781339233e-05, + "loss": 1.0072, + "step": 26110 + }, + { + "epoch": 1.68, + "learning_rate": 1.4418300092316282e-05, + "loss": 0.9948, + "step": 26111 + }, + { + "epoch": 1.68, + "learning_rate": 1.4418085403293332e-05, + "loss": 0.9645, + "step": 26112 + }, + { + "epoch": 1.68, + "learning_rate": 1.4417870714270381e-05, + "loss": 0.9981, + "step": 26113 + }, + { + "epoch": 1.68, + "learning_rate": 1.441765602524743e-05, + "loss": 0.998, + "step": 26114 + }, + { + "epoch": 1.68, + "learning_rate": 1.4417441336224481e-05, + "loss": 0.9544, + "step": 26115 + }, + { + "epoch": 1.68, + "learning_rate": 1.441722664720153e-05, + "loss": 0.9773, + "step": 26116 + }, + { + "epoch": 1.68, + "learning_rate": 1.4417011958178579e-05, + "loss": 0.995, + "step": 26117 + }, + { + "epoch": 1.68, + "learning_rate": 1.441679726915563e-05, + "loss": 0.9608, + "step": 26118 + }, + { + "epoch": 1.68, + "learning_rate": 1.4416582580132678e-05, + "loss": 0.9803, + "step": 26119 + }, + { + "epoch": 1.68, + "learning_rate": 1.4416367891109727e-05, + "loss": 1.0055, + "step": 26120 + }, + { + "epoch": 1.68, + "learning_rate": 1.4416153202086778e-05, + "loss": 0.9598, + "step": 26121 + }, + { + "epoch": 1.68, + "learning_rate": 1.4415938513063827e-05, + "loss": 0.9422, + "step": 26122 + }, + { + "epoch": 1.68, + "learning_rate": 1.4415723824040878e-05, + "loss": 0.98, + "step": 26123 + }, + { + "epoch": 1.68, + "learning_rate": 1.4415509135017928e-05, + "loss": 0.9576, + "step": 26124 + }, + { + "epoch": 1.68, + "learning_rate": 1.4415294445994979e-05, + "loss": 1.0274, + "step": 26125 + }, + { + "epoch": 1.68, + "learning_rate": 1.4415079756972028e-05, + "loss": 0.9945, + "step": 26126 + }, + { + "epoch": 1.68, + "learning_rate": 1.4414865067949077e-05, + "loss": 0.9636, + "step": 26127 + }, + { + "epoch": 1.68, + "learning_rate": 1.4414650378926128e-05, + "loss": 0.9647, + "step": 26128 + }, + { + "epoch": 1.68, + "learning_rate": 1.4414435689903177e-05, + "loss": 1.0026, + "step": 26129 + }, + { + "epoch": 1.68, + "learning_rate": 1.4414221000880226e-05, + "loss": 0.9911, + "step": 26130 + }, + { + "epoch": 1.68, + "learning_rate": 1.4414006311857276e-05, + "loss": 1.0291, + "step": 26131 + }, + { + "epoch": 1.68, + "learning_rate": 1.4413791622834325e-05, + "loss": 1.0828, + "step": 26132 + }, + { + "epoch": 1.68, + "learning_rate": 1.4413576933811376e-05, + "loss": 0.9487, + "step": 26133 + }, + { + "epoch": 1.68, + "learning_rate": 1.4413362244788425e-05, + "loss": 0.9529, + "step": 26134 + }, + { + "epoch": 1.68, + "learning_rate": 1.4413147555765474e-05, + "loss": 0.9355, + "step": 26135 + }, + { + "epoch": 1.68, + "learning_rate": 1.4412932866742524e-05, + "loss": 1.0454, + "step": 26136 + }, + { + "epoch": 1.68, + "learning_rate": 1.4412718177719573e-05, + "loss": 0.9503, + "step": 26137 + }, + { + "epoch": 1.68, + "learning_rate": 1.4412503488696622e-05, + "loss": 1.0614, + "step": 26138 + }, + { + "epoch": 1.68, + "learning_rate": 1.4412288799673675e-05, + "loss": 1.0518, + "step": 26139 + }, + { + "epoch": 1.68, + "learning_rate": 1.4412074110650724e-05, + "loss": 0.9427, + "step": 26140 + }, + { + "epoch": 1.68, + "learning_rate": 1.4411859421627774e-05, + "loss": 0.9867, + "step": 26141 + }, + { + "epoch": 1.68, + "learning_rate": 1.4411644732604823e-05, + "loss": 0.9195, + "step": 26142 + }, + { + "epoch": 1.68, + "learning_rate": 1.4411430043581874e-05, + "loss": 0.9767, + "step": 26143 + }, + { + "epoch": 1.68, + "learning_rate": 1.4411215354558923e-05, + "loss": 0.915, + "step": 26144 + }, + { + "epoch": 1.68, + "learning_rate": 1.4411000665535972e-05, + "loss": 1.0703, + "step": 26145 + }, + { + "epoch": 1.68, + "learning_rate": 1.4410785976513023e-05, + "loss": 0.9638, + "step": 26146 + }, + { + "epoch": 1.68, + "learning_rate": 1.4410571287490072e-05, + "loss": 1.0035, + "step": 26147 + }, + { + "epoch": 1.68, + "learning_rate": 1.441035659846712e-05, + "loss": 1.0069, + "step": 26148 + }, + { + "epoch": 1.68, + "learning_rate": 1.4410141909444171e-05, + "loss": 1.0129, + "step": 26149 + }, + { + "epoch": 1.68, + "learning_rate": 1.440992722042122e-05, + "loss": 1.0041, + "step": 26150 + }, + { + "epoch": 1.68, + "learning_rate": 1.440971253139827e-05, + "loss": 0.9207, + "step": 26151 + }, + { + "epoch": 1.68, + "learning_rate": 1.440949784237532e-05, + "loss": 1.0223, + "step": 26152 + }, + { + "epoch": 1.68, + "learning_rate": 1.4409283153352372e-05, + "loss": 1.0337, + "step": 26153 + }, + { + "epoch": 1.68, + "learning_rate": 1.4409068464329421e-05, + "loss": 0.9292, + "step": 26154 + }, + { + "epoch": 1.68, + "learning_rate": 1.440885377530647e-05, + "loss": 1.0644, + "step": 26155 + }, + { + "epoch": 1.68, + "learning_rate": 1.440863908628352e-05, + "loss": 1.0172, + "step": 26156 + }, + { + "epoch": 1.68, + "learning_rate": 1.440842439726057e-05, + "loss": 0.9515, + "step": 26157 + }, + { + "epoch": 1.68, + "learning_rate": 1.4408209708237619e-05, + "loss": 1.0134, + "step": 26158 + }, + { + "epoch": 1.68, + "learning_rate": 1.440799501921467e-05, + "loss": 0.9089, + "step": 26159 + }, + { + "epoch": 1.68, + "learning_rate": 1.4407780330191718e-05, + "loss": 0.9702, + "step": 26160 + }, + { + "epoch": 1.68, + "learning_rate": 1.4407565641168769e-05, + "loss": 0.9895, + "step": 26161 + }, + { + "epoch": 1.68, + "learning_rate": 1.4407350952145818e-05, + "loss": 0.9667, + "step": 26162 + }, + { + "epoch": 1.68, + "learning_rate": 1.4407136263122867e-05, + "loss": 1.0537, + "step": 26163 + }, + { + "epoch": 1.68, + "learning_rate": 1.4406921574099917e-05, + "loss": 0.9673, + "step": 26164 + }, + { + "epoch": 1.68, + "learning_rate": 1.4406706885076966e-05, + "loss": 1.0014, + "step": 26165 + }, + { + "epoch": 1.68, + "learning_rate": 1.4406492196054015e-05, + "loss": 1.002, + "step": 26166 + }, + { + "epoch": 1.68, + "learning_rate": 1.4406277507031066e-05, + "loss": 1.0142, + "step": 26167 + }, + { + "epoch": 1.68, + "learning_rate": 1.4406062818008117e-05, + "loss": 0.9791, + "step": 26168 + }, + { + "epoch": 1.68, + "learning_rate": 1.4405848128985167e-05, + "loss": 0.9292, + "step": 26169 + }, + { + "epoch": 1.68, + "learning_rate": 1.4405633439962216e-05, + "loss": 0.9987, + "step": 26170 + }, + { + "epoch": 1.68, + "learning_rate": 1.4405418750939267e-05, + "loss": 1.026, + "step": 26171 + }, + { + "epoch": 1.68, + "learning_rate": 1.4405204061916316e-05, + "loss": 0.9026, + "step": 26172 + }, + { + "epoch": 1.68, + "learning_rate": 1.4404989372893365e-05, + "loss": 0.9532, + "step": 26173 + }, + { + "epoch": 1.68, + "learning_rate": 1.4404774683870416e-05, + "loss": 0.9681, + "step": 26174 + }, + { + "epoch": 1.68, + "learning_rate": 1.4404559994847465e-05, + "loss": 1.0312, + "step": 26175 + }, + { + "epoch": 1.68, + "learning_rate": 1.4404345305824514e-05, + "loss": 0.93, + "step": 26176 + }, + { + "epoch": 1.68, + "learning_rate": 1.4404130616801564e-05, + "loss": 0.9545, + "step": 26177 + }, + { + "epoch": 1.68, + "learning_rate": 1.4403915927778613e-05, + "loss": 0.9966, + "step": 26178 + }, + { + "epoch": 1.68, + "learning_rate": 1.4403701238755662e-05, + "loss": 1.026, + "step": 26179 + }, + { + "epoch": 1.68, + "learning_rate": 1.4403486549732713e-05, + "loss": 1.0004, + "step": 26180 + }, + { + "epoch": 1.68, + "learning_rate": 1.4403271860709762e-05, + "loss": 0.9769, + "step": 26181 + }, + { + "epoch": 1.68, + "learning_rate": 1.4403057171686812e-05, + "loss": 0.9175, + "step": 26182 + }, + { + "epoch": 1.68, + "learning_rate": 1.4402842482663863e-05, + "loss": 0.9324, + "step": 26183 + }, + { + "epoch": 1.68, + "learning_rate": 1.4402627793640914e-05, + "loss": 1.0325, + "step": 26184 + }, + { + "epoch": 1.68, + "learning_rate": 1.4402413104617963e-05, + "loss": 1.0295, + "step": 26185 + }, + { + "epoch": 1.68, + "learning_rate": 1.4402198415595012e-05, + "loss": 0.9351, + "step": 26186 + }, + { + "epoch": 1.68, + "learning_rate": 1.4401983726572062e-05, + "loss": 0.9929, + "step": 26187 + }, + { + "epoch": 1.68, + "learning_rate": 1.4401769037549111e-05, + "loss": 0.9989, + "step": 26188 + }, + { + "epoch": 1.68, + "learning_rate": 1.440155434852616e-05, + "loss": 0.9775, + "step": 26189 + }, + { + "epoch": 1.69, + "learning_rate": 1.4401339659503211e-05, + "loss": 0.9947, + "step": 26190 + }, + { + "epoch": 1.69, + "learning_rate": 1.440112497048026e-05, + "loss": 1.0009, + "step": 26191 + }, + { + "epoch": 1.69, + "learning_rate": 1.440091028145731e-05, + "loss": 0.9204, + "step": 26192 + }, + { + "epoch": 1.69, + "learning_rate": 1.440069559243436e-05, + "loss": 0.9657, + "step": 26193 + }, + { + "epoch": 1.69, + "learning_rate": 1.4400480903411408e-05, + "loss": 1.0706, + "step": 26194 + }, + { + "epoch": 1.69, + "learning_rate": 1.4400266214388459e-05, + "loss": 0.9761, + "step": 26195 + }, + { + "epoch": 1.69, + "learning_rate": 1.4400051525365508e-05, + "loss": 0.9722, + "step": 26196 + }, + { + "epoch": 1.69, + "learning_rate": 1.439983683634256e-05, + "loss": 0.9488, + "step": 26197 + }, + { + "epoch": 1.69, + "learning_rate": 1.439962214731961e-05, + "loss": 0.8881, + "step": 26198 + }, + { + "epoch": 1.69, + "learning_rate": 1.439940745829666e-05, + "loss": 1.0063, + "step": 26199 + }, + { + "epoch": 1.69, + "learning_rate": 1.4399192769273709e-05, + "loss": 0.8729, + "step": 26200 + }, + { + "epoch": 1.69, + "learning_rate": 1.4398978080250758e-05, + "loss": 0.966, + "step": 26201 + }, + { + "epoch": 1.69, + "learning_rate": 1.4398763391227809e-05, + "loss": 0.9612, + "step": 26202 + }, + { + "epoch": 1.69, + "learning_rate": 1.4398548702204858e-05, + "loss": 0.9462, + "step": 26203 + }, + { + "epoch": 1.69, + "learning_rate": 1.4398334013181907e-05, + "loss": 0.9872, + "step": 26204 + }, + { + "epoch": 1.69, + "learning_rate": 1.4398119324158957e-05, + "loss": 1.049, + "step": 26205 + }, + { + "epoch": 1.69, + "learning_rate": 1.4397904635136006e-05, + "loss": 0.9528, + "step": 26206 + }, + { + "epoch": 1.69, + "learning_rate": 1.4397689946113055e-05, + "loss": 1.079, + "step": 26207 + }, + { + "epoch": 1.69, + "learning_rate": 1.4397475257090106e-05, + "loss": 0.9736, + "step": 26208 + }, + { + "epoch": 1.69, + "learning_rate": 1.4397260568067155e-05, + "loss": 1.0, + "step": 26209 + }, + { + "epoch": 1.69, + "learning_rate": 1.4397045879044205e-05, + "loss": 0.9174, + "step": 26210 + }, + { + "epoch": 1.69, + "learning_rate": 1.4396831190021254e-05, + "loss": 0.9337, + "step": 26211 + }, + { + "epoch": 1.69, + "learning_rate": 1.4396616500998307e-05, + "loss": 0.9552, + "step": 26212 + }, + { + "epoch": 1.69, + "learning_rate": 1.4396401811975356e-05, + "loss": 0.972, + "step": 26213 + }, + { + "epoch": 1.69, + "learning_rate": 1.4396187122952405e-05, + "loss": 0.993, + "step": 26214 + }, + { + "epoch": 1.69, + "learning_rate": 1.4395972433929455e-05, + "loss": 0.986, + "step": 26215 + }, + { + "epoch": 1.69, + "learning_rate": 1.4395757744906504e-05, + "loss": 0.9754, + "step": 26216 + }, + { + "epoch": 1.69, + "learning_rate": 1.4395543055883553e-05, + "loss": 0.9401, + "step": 26217 + }, + { + "epoch": 1.69, + "learning_rate": 1.4395328366860604e-05, + "loss": 0.9825, + "step": 26218 + }, + { + "epoch": 1.69, + "learning_rate": 1.4395113677837653e-05, + "loss": 0.9067, + "step": 26219 + }, + { + "epoch": 1.69, + "learning_rate": 1.4394898988814704e-05, + "loss": 1.0414, + "step": 26220 + }, + { + "epoch": 1.69, + "learning_rate": 1.4394684299791753e-05, + "loss": 0.9176, + "step": 26221 + }, + { + "epoch": 1.69, + "learning_rate": 1.4394469610768801e-05, + "loss": 0.9423, + "step": 26222 + }, + { + "epoch": 1.69, + "learning_rate": 1.4394254921745852e-05, + "loss": 1.0196, + "step": 26223 + }, + { + "epoch": 1.69, + "learning_rate": 1.4394040232722901e-05, + "loss": 0.8725, + "step": 26224 + }, + { + "epoch": 1.69, + "learning_rate": 1.439382554369995e-05, + "loss": 1.038, + "step": 26225 + }, + { + "epoch": 1.69, + "learning_rate": 1.4393610854677002e-05, + "loss": 0.9435, + "step": 26226 + }, + { + "epoch": 1.69, + "learning_rate": 1.4393396165654051e-05, + "loss": 0.9341, + "step": 26227 + }, + { + "epoch": 1.69, + "learning_rate": 1.4393181476631102e-05, + "loss": 1.1302, + "step": 26228 + }, + { + "epoch": 1.69, + "learning_rate": 1.4392966787608151e-05, + "loss": 0.9577, + "step": 26229 + }, + { + "epoch": 1.69, + "learning_rate": 1.4392752098585202e-05, + "loss": 0.9705, + "step": 26230 + }, + { + "epoch": 1.69, + "learning_rate": 1.439253740956225e-05, + "loss": 0.9711, + "step": 26231 + }, + { + "epoch": 1.69, + "learning_rate": 1.43923227205393e-05, + "loss": 1.0115, + "step": 26232 + }, + { + "epoch": 1.69, + "learning_rate": 1.439210803151635e-05, + "loss": 1.0392, + "step": 26233 + }, + { + "epoch": 1.69, + "learning_rate": 1.43918933424934e-05, + "loss": 0.9768, + "step": 26234 + }, + { + "epoch": 1.69, + "learning_rate": 1.4391678653470448e-05, + "loss": 0.9412, + "step": 26235 + }, + { + "epoch": 1.69, + "learning_rate": 1.4391463964447499e-05, + "loss": 0.9757, + "step": 26236 + }, + { + "epoch": 1.69, + "learning_rate": 1.4391249275424548e-05, + "loss": 1.0306, + "step": 26237 + }, + { + "epoch": 1.69, + "learning_rate": 1.4391034586401598e-05, + "loss": 0.9788, + "step": 26238 + }, + { + "epoch": 1.69, + "learning_rate": 1.4390819897378647e-05, + "loss": 0.9484, + "step": 26239 + }, + { + "epoch": 1.69, + "learning_rate": 1.4390605208355696e-05, + "loss": 0.9632, + "step": 26240 + }, + { + "epoch": 1.69, + "learning_rate": 1.4390390519332749e-05, + "loss": 0.9564, + "step": 26241 + }, + { + "epoch": 1.69, + "learning_rate": 1.4390175830309798e-05, + "loss": 0.9394, + "step": 26242 + }, + { + "epoch": 1.69, + "learning_rate": 1.4389961141286848e-05, + "loss": 1.0007, + "step": 26243 + }, + { + "epoch": 1.69, + "learning_rate": 1.4389746452263897e-05, + "loss": 1.0156, + "step": 26244 + }, + { + "epoch": 1.69, + "learning_rate": 1.4389531763240946e-05, + "loss": 1.0042, + "step": 26245 + }, + { + "epoch": 1.69, + "learning_rate": 1.4389317074217997e-05, + "loss": 0.9682, + "step": 26246 + }, + { + "epoch": 1.69, + "learning_rate": 1.4389102385195046e-05, + "loss": 1.0146, + "step": 26247 + }, + { + "epoch": 1.69, + "learning_rate": 1.4388887696172097e-05, + "loss": 1.0155, + "step": 26248 + }, + { + "epoch": 1.69, + "learning_rate": 1.4388673007149146e-05, + "loss": 1.0052, + "step": 26249 + }, + { + "epoch": 1.69, + "learning_rate": 1.4388458318126195e-05, + "loss": 1.0485, + "step": 26250 + }, + { + "epoch": 1.69, + "learning_rate": 1.4388243629103245e-05, + "loss": 1.0082, + "step": 26251 + }, + { + "epoch": 1.69, + "learning_rate": 1.4388028940080294e-05, + "loss": 0.9089, + "step": 26252 + }, + { + "epoch": 1.69, + "learning_rate": 1.4387814251057343e-05, + "loss": 0.8897, + "step": 26253 + }, + { + "epoch": 1.69, + "learning_rate": 1.4387599562034394e-05, + "loss": 0.9213, + "step": 26254 + }, + { + "epoch": 1.69, + "learning_rate": 1.4387384873011444e-05, + "loss": 0.9035, + "step": 26255 + }, + { + "epoch": 1.69, + "learning_rate": 1.4387170183988495e-05, + "loss": 0.9576, + "step": 26256 + }, + { + "epoch": 1.69, + "learning_rate": 1.4386955494965544e-05, + "loss": 1.0006, + "step": 26257 + }, + { + "epoch": 1.69, + "learning_rate": 1.4386740805942595e-05, + "loss": 1.0207, + "step": 26258 + }, + { + "epoch": 1.69, + "learning_rate": 1.4386526116919644e-05, + "loss": 0.9355, + "step": 26259 + }, + { + "epoch": 1.69, + "learning_rate": 1.4386311427896693e-05, + "loss": 0.9692, + "step": 26260 + }, + { + "epoch": 1.69, + "learning_rate": 1.4386096738873743e-05, + "loss": 1.0078, + "step": 26261 + }, + { + "epoch": 1.69, + "learning_rate": 1.4385882049850792e-05, + "loss": 1.0915, + "step": 26262 + }, + { + "epoch": 1.69, + "learning_rate": 1.4385667360827841e-05, + "loss": 0.9679, + "step": 26263 + }, + { + "epoch": 1.69, + "learning_rate": 1.4385452671804892e-05, + "loss": 0.9401, + "step": 26264 + }, + { + "epoch": 1.69, + "learning_rate": 1.4385237982781941e-05, + "loss": 0.9666, + "step": 26265 + }, + { + "epoch": 1.69, + "learning_rate": 1.438502329375899e-05, + "loss": 1.0154, + "step": 26266 + }, + { + "epoch": 1.69, + "learning_rate": 1.438480860473604e-05, + "loss": 1.0221, + "step": 26267 + }, + { + "epoch": 1.69, + "learning_rate": 1.438459391571309e-05, + "loss": 0.9863, + "step": 26268 + }, + { + "epoch": 1.69, + "learning_rate": 1.438437922669014e-05, + "loss": 1.0841, + "step": 26269 + }, + { + "epoch": 1.69, + "learning_rate": 1.438416453766719e-05, + "loss": 0.9516, + "step": 26270 + }, + { + "epoch": 1.69, + "learning_rate": 1.4383949848644241e-05, + "loss": 1.0113, + "step": 26271 + }, + { + "epoch": 1.69, + "learning_rate": 1.438373515962129e-05, + "loss": 0.9687, + "step": 26272 + }, + { + "epoch": 1.69, + "learning_rate": 1.438352047059834e-05, + "loss": 0.9017, + "step": 26273 + }, + { + "epoch": 1.69, + "learning_rate": 1.438330578157539e-05, + "loss": 0.9871, + "step": 26274 + }, + { + "epoch": 1.69, + "learning_rate": 1.4383091092552439e-05, + "loss": 0.9761, + "step": 26275 + }, + { + "epoch": 1.69, + "learning_rate": 1.4382876403529488e-05, + "loss": 0.9677, + "step": 26276 + }, + { + "epoch": 1.69, + "learning_rate": 1.4382661714506539e-05, + "loss": 0.8915, + "step": 26277 + }, + { + "epoch": 1.69, + "learning_rate": 1.4382447025483588e-05, + "loss": 0.9686, + "step": 26278 + }, + { + "epoch": 1.69, + "learning_rate": 1.4382232336460638e-05, + "loss": 0.9432, + "step": 26279 + }, + { + "epoch": 1.69, + "learning_rate": 1.4382017647437687e-05, + "loss": 0.9939, + "step": 26280 + }, + { + "epoch": 1.69, + "learning_rate": 1.4381802958414736e-05, + "loss": 0.9754, + "step": 26281 + }, + { + "epoch": 1.69, + "learning_rate": 1.4381588269391787e-05, + "loss": 0.9782, + "step": 26282 + }, + { + "epoch": 1.69, + "learning_rate": 1.4381373580368836e-05, + "loss": 1.0477, + "step": 26283 + }, + { + "epoch": 1.69, + "learning_rate": 1.4381158891345888e-05, + "loss": 0.9804, + "step": 26284 + }, + { + "epoch": 1.69, + "learning_rate": 1.4380944202322937e-05, + "loss": 1.0149, + "step": 26285 + }, + { + "epoch": 1.69, + "learning_rate": 1.4380729513299986e-05, + "loss": 0.8816, + "step": 26286 + }, + { + "epoch": 1.69, + "learning_rate": 1.4380514824277037e-05, + "loss": 1.0459, + "step": 26287 + }, + { + "epoch": 1.69, + "learning_rate": 1.4380300135254086e-05, + "loss": 0.9864, + "step": 26288 + }, + { + "epoch": 1.69, + "learning_rate": 1.4380085446231136e-05, + "loss": 0.9715, + "step": 26289 + }, + { + "epoch": 1.69, + "learning_rate": 1.4379870757208185e-05, + "loss": 0.9885, + "step": 26290 + }, + { + "epoch": 1.69, + "learning_rate": 1.4379656068185234e-05, + "loss": 0.92, + "step": 26291 + }, + { + "epoch": 1.69, + "learning_rate": 1.4379441379162285e-05, + "loss": 0.9758, + "step": 26292 + }, + { + "epoch": 1.69, + "learning_rate": 1.4379226690139334e-05, + "loss": 0.9618, + "step": 26293 + }, + { + "epoch": 1.69, + "learning_rate": 1.4379012001116383e-05, + "loss": 0.9485, + "step": 26294 + }, + { + "epoch": 1.69, + "learning_rate": 1.4378797312093434e-05, + "loss": 0.8795, + "step": 26295 + }, + { + "epoch": 1.69, + "learning_rate": 1.4378582623070482e-05, + "loss": 0.908, + "step": 26296 + }, + { + "epoch": 1.69, + "learning_rate": 1.4378367934047533e-05, + "loss": 1.0528, + "step": 26297 + }, + { + "epoch": 1.69, + "learning_rate": 1.4378153245024582e-05, + "loss": 0.9631, + "step": 26298 + }, + { + "epoch": 1.69, + "learning_rate": 1.4377938556001634e-05, + "loss": 1.0176, + "step": 26299 + }, + { + "epoch": 1.69, + "learning_rate": 1.4377723866978683e-05, + "loss": 0.9744, + "step": 26300 + }, + { + "epoch": 1.69, + "learning_rate": 1.4377509177955732e-05, + "loss": 0.9469, + "step": 26301 + }, + { + "epoch": 1.69, + "learning_rate": 1.4377294488932783e-05, + "loss": 1.0139, + "step": 26302 + }, + { + "epoch": 1.69, + "learning_rate": 1.4377079799909832e-05, + "loss": 1.0095, + "step": 26303 + }, + { + "epoch": 1.69, + "learning_rate": 1.4376865110886881e-05, + "loss": 0.9573, + "step": 26304 + }, + { + "epoch": 1.69, + "learning_rate": 1.4376650421863932e-05, + "loss": 0.9583, + "step": 26305 + }, + { + "epoch": 1.69, + "learning_rate": 1.437643573284098e-05, + "loss": 0.9895, + "step": 26306 + }, + { + "epoch": 1.69, + "learning_rate": 1.4376221043818031e-05, + "loss": 0.991, + "step": 26307 + }, + { + "epoch": 1.69, + "learning_rate": 1.437600635479508e-05, + "loss": 1.0415, + "step": 26308 + }, + { + "epoch": 1.69, + "learning_rate": 1.437579166577213e-05, + "loss": 0.9843, + "step": 26309 + }, + { + "epoch": 1.69, + "learning_rate": 1.437557697674918e-05, + "loss": 1.003, + "step": 26310 + }, + { + "epoch": 1.69, + "learning_rate": 1.4375362287726229e-05, + "loss": 0.997, + "step": 26311 + }, + { + "epoch": 1.69, + "learning_rate": 1.4375147598703278e-05, + "loss": 0.9513, + "step": 26312 + }, + { + "epoch": 1.69, + "learning_rate": 1.437493290968033e-05, + "loss": 0.9911, + "step": 26313 + }, + { + "epoch": 1.69, + "learning_rate": 1.4374718220657379e-05, + "loss": 0.9871, + "step": 26314 + }, + { + "epoch": 1.69, + "learning_rate": 1.437450353163443e-05, + "loss": 1.0238, + "step": 26315 + }, + { + "epoch": 1.69, + "learning_rate": 1.4374288842611479e-05, + "loss": 0.997, + "step": 26316 + }, + { + "epoch": 1.69, + "learning_rate": 1.437407415358853e-05, + "loss": 1.0236, + "step": 26317 + }, + { + "epoch": 1.69, + "learning_rate": 1.4373859464565578e-05, + "loss": 0.9037, + "step": 26318 + }, + { + "epoch": 1.69, + "learning_rate": 1.4373644775542627e-05, + "loss": 0.9925, + "step": 26319 + }, + { + "epoch": 1.69, + "learning_rate": 1.4373430086519678e-05, + "loss": 1.0269, + "step": 26320 + }, + { + "epoch": 1.69, + "learning_rate": 1.4373215397496727e-05, + "loss": 0.9841, + "step": 26321 + }, + { + "epoch": 1.69, + "learning_rate": 1.4373000708473776e-05, + "loss": 0.9769, + "step": 26322 + }, + { + "epoch": 1.69, + "learning_rate": 1.4372786019450827e-05, + "loss": 0.9665, + "step": 26323 + }, + { + "epoch": 1.69, + "learning_rate": 1.4372571330427876e-05, + "loss": 0.9433, + "step": 26324 + }, + { + "epoch": 1.69, + "learning_rate": 1.4372356641404926e-05, + "loss": 0.8979, + "step": 26325 + }, + { + "epoch": 1.69, + "learning_rate": 1.4372141952381975e-05, + "loss": 0.907, + "step": 26326 + }, + { + "epoch": 1.69, + "learning_rate": 1.4371927263359024e-05, + "loss": 1.042, + "step": 26327 + }, + { + "epoch": 1.69, + "learning_rate": 1.4371712574336076e-05, + "loss": 0.9803, + "step": 26328 + }, + { + "epoch": 1.69, + "learning_rate": 1.4371497885313125e-05, + "loss": 1.0274, + "step": 26329 + }, + { + "epoch": 1.69, + "learning_rate": 1.4371283196290176e-05, + "loss": 0.9941, + "step": 26330 + }, + { + "epoch": 1.69, + "learning_rate": 1.4371068507267225e-05, + "loss": 0.9212, + "step": 26331 + }, + { + "epoch": 1.69, + "learning_rate": 1.4370853818244274e-05, + "loss": 0.9304, + "step": 26332 + }, + { + "epoch": 1.69, + "learning_rate": 1.4370639129221325e-05, + "loss": 0.9078, + "step": 26333 + }, + { + "epoch": 1.69, + "learning_rate": 1.4370424440198374e-05, + "loss": 0.9667, + "step": 26334 + }, + { + "epoch": 1.69, + "learning_rate": 1.4370209751175424e-05, + "loss": 1.0084, + "step": 26335 + }, + { + "epoch": 1.69, + "learning_rate": 1.4369995062152473e-05, + "loss": 0.981, + "step": 26336 + }, + { + "epoch": 1.69, + "learning_rate": 1.4369780373129522e-05, + "loss": 1.0145, + "step": 26337 + }, + { + "epoch": 1.69, + "learning_rate": 1.4369565684106573e-05, + "loss": 1.0077, + "step": 26338 + }, + { + "epoch": 1.69, + "learning_rate": 1.4369350995083622e-05, + "loss": 1.0508, + "step": 26339 + }, + { + "epoch": 1.69, + "learning_rate": 1.436913630606067e-05, + "loss": 0.9737, + "step": 26340 + }, + { + "epoch": 1.69, + "learning_rate": 1.4368921617037721e-05, + "loss": 1.0013, + "step": 26341 + }, + { + "epoch": 1.69, + "learning_rate": 1.4368706928014772e-05, + "loss": 0.9471, + "step": 26342 + }, + { + "epoch": 1.69, + "learning_rate": 1.4368492238991823e-05, + "loss": 1.0266, + "step": 26343 + }, + { + "epoch": 1.69, + "learning_rate": 1.4368277549968872e-05, + "loss": 0.9596, + "step": 26344 + }, + { + "epoch": 1.69, + "learning_rate": 1.4368062860945922e-05, + "loss": 1.0323, + "step": 26345 + }, + { + "epoch": 1.7, + "learning_rate": 1.4367848171922971e-05, + "loss": 0.9611, + "step": 26346 + }, + { + "epoch": 1.7, + "learning_rate": 1.436763348290002e-05, + "loss": 0.9486, + "step": 26347 + }, + { + "epoch": 1.7, + "learning_rate": 1.4367418793877071e-05, + "loss": 0.8926, + "step": 26348 + }, + { + "epoch": 1.7, + "learning_rate": 1.436720410485412e-05, + "loss": 0.9207, + "step": 26349 + }, + { + "epoch": 1.7, + "learning_rate": 1.4366989415831169e-05, + "loss": 0.8619, + "step": 26350 + }, + { + "epoch": 1.7, + "learning_rate": 1.436677472680822e-05, + "loss": 1.0221, + "step": 26351 + }, + { + "epoch": 1.7, + "learning_rate": 1.4366560037785269e-05, + "loss": 0.9787, + "step": 26352 + }, + { + "epoch": 1.7, + "learning_rate": 1.4366345348762318e-05, + "loss": 0.9842, + "step": 26353 + }, + { + "epoch": 1.7, + "learning_rate": 1.4366130659739368e-05, + "loss": 0.9485, + "step": 26354 + }, + { + "epoch": 1.7, + "learning_rate": 1.4365915970716417e-05, + "loss": 1.0054, + "step": 26355 + }, + { + "epoch": 1.7, + "learning_rate": 1.4365701281693468e-05, + "loss": 1.0659, + "step": 26356 + }, + { + "epoch": 1.7, + "learning_rate": 1.4365486592670518e-05, + "loss": 0.9713, + "step": 26357 + }, + { + "epoch": 1.7, + "learning_rate": 1.4365271903647569e-05, + "loss": 1.0279, + "step": 26358 + }, + { + "epoch": 1.7, + "learning_rate": 1.4365057214624618e-05, + "loss": 0.9757, + "step": 26359 + }, + { + "epoch": 1.7, + "learning_rate": 1.4364842525601667e-05, + "loss": 1.0139, + "step": 26360 + }, + { + "epoch": 1.7, + "learning_rate": 1.4364627836578718e-05, + "loss": 0.9446, + "step": 26361 + }, + { + "epoch": 1.7, + "learning_rate": 1.4364413147555767e-05, + "loss": 0.9456, + "step": 26362 + }, + { + "epoch": 1.7, + "learning_rate": 1.4364198458532816e-05, + "loss": 0.953, + "step": 26363 + }, + { + "epoch": 1.7, + "learning_rate": 1.4363983769509866e-05, + "loss": 0.9177, + "step": 26364 + }, + { + "epoch": 1.7, + "learning_rate": 1.4363769080486915e-05, + "loss": 0.9142, + "step": 26365 + }, + { + "epoch": 1.7, + "learning_rate": 1.4363554391463966e-05, + "loss": 0.958, + "step": 26366 + }, + { + "epoch": 1.7, + "learning_rate": 1.4363339702441015e-05, + "loss": 0.9247, + "step": 26367 + }, + { + "epoch": 1.7, + "learning_rate": 1.4363125013418064e-05, + "loss": 0.9844, + "step": 26368 + }, + { + "epoch": 1.7, + "learning_rate": 1.4362910324395115e-05, + "loss": 0.9907, + "step": 26369 + }, + { + "epoch": 1.7, + "learning_rate": 1.4362695635372163e-05, + "loss": 0.9625, + "step": 26370 + }, + { + "epoch": 1.7, + "learning_rate": 1.4362480946349216e-05, + "loss": 0.9824, + "step": 26371 + }, + { + "epoch": 1.7, + "learning_rate": 1.4362266257326265e-05, + "loss": 1.0239, + "step": 26372 + }, + { + "epoch": 1.7, + "learning_rate": 1.4362051568303314e-05, + "loss": 0.943, + "step": 26373 + }, + { + "epoch": 1.7, + "learning_rate": 1.4361836879280364e-05, + "loss": 0.9743, + "step": 26374 + }, + { + "epoch": 1.7, + "learning_rate": 1.4361622190257413e-05, + "loss": 1.0103, + "step": 26375 + }, + { + "epoch": 1.7, + "learning_rate": 1.4361407501234464e-05, + "loss": 0.9957, + "step": 26376 + }, + { + "epoch": 1.7, + "learning_rate": 1.4361192812211513e-05, + "loss": 0.9506, + "step": 26377 + }, + { + "epoch": 1.7, + "learning_rate": 1.4360978123188562e-05, + "loss": 1.052, + "step": 26378 + }, + { + "epoch": 1.7, + "learning_rate": 1.4360763434165613e-05, + "loss": 1.0397, + "step": 26379 + }, + { + "epoch": 1.7, + "learning_rate": 1.4360548745142662e-05, + "loss": 0.9285, + "step": 26380 + }, + { + "epoch": 1.7, + "learning_rate": 1.436033405611971e-05, + "loss": 0.9275, + "step": 26381 + }, + { + "epoch": 1.7, + "learning_rate": 1.4360119367096761e-05, + "loss": 0.9946, + "step": 26382 + }, + { + "epoch": 1.7, + "learning_rate": 1.435990467807381e-05, + "loss": 0.9653, + "step": 26383 + }, + { + "epoch": 1.7, + "learning_rate": 1.4359689989050861e-05, + "loss": 0.9668, + "step": 26384 + }, + { + "epoch": 1.7, + "learning_rate": 1.435947530002791e-05, + "loss": 0.9542, + "step": 26385 + }, + { + "epoch": 1.7, + "learning_rate": 1.4359260611004962e-05, + "loss": 0.93, + "step": 26386 + }, + { + "epoch": 1.7, + "learning_rate": 1.4359045921982011e-05, + "loss": 0.8844, + "step": 26387 + }, + { + "epoch": 1.7, + "learning_rate": 1.435883123295906e-05, + "loss": 0.9708, + "step": 26388 + }, + { + "epoch": 1.7, + "learning_rate": 1.435861654393611e-05, + "loss": 0.9682, + "step": 26389 + }, + { + "epoch": 1.7, + "learning_rate": 1.435840185491316e-05, + "loss": 0.9565, + "step": 26390 + }, + { + "epoch": 1.7, + "learning_rate": 1.4358187165890209e-05, + "loss": 0.9821, + "step": 26391 + }, + { + "epoch": 1.7, + "learning_rate": 1.435797247686726e-05, + "loss": 0.9724, + "step": 26392 + }, + { + "epoch": 1.7, + "learning_rate": 1.4357757787844308e-05, + "loss": 1.0198, + "step": 26393 + }, + { + "epoch": 1.7, + "learning_rate": 1.4357543098821359e-05, + "loss": 0.9165, + "step": 26394 + }, + { + "epoch": 1.7, + "learning_rate": 1.4357328409798408e-05, + "loss": 0.9616, + "step": 26395 + }, + { + "epoch": 1.7, + "learning_rate": 1.4357113720775457e-05, + "loss": 0.9903, + "step": 26396 + }, + { + "epoch": 1.7, + "learning_rate": 1.4356899031752508e-05, + "loss": 0.9392, + "step": 26397 + }, + { + "epoch": 1.7, + "learning_rate": 1.4356684342729557e-05, + "loss": 0.9632, + "step": 26398 + }, + { + "epoch": 1.7, + "learning_rate": 1.4356469653706606e-05, + "loss": 0.8788, + "step": 26399 + }, + { + "epoch": 1.7, + "learning_rate": 1.4356254964683658e-05, + "loss": 0.9408, + "step": 26400 + }, + { + "epoch": 1.7, + "learning_rate": 1.4356040275660707e-05, + "loss": 0.9999, + "step": 26401 + }, + { + "epoch": 1.7, + "learning_rate": 1.4355825586637757e-05, + "loss": 1.0463, + "step": 26402 + }, + { + "epoch": 1.7, + "learning_rate": 1.4355610897614806e-05, + "loss": 0.976, + "step": 26403 + }, + { + "epoch": 1.7, + "learning_rate": 1.4355396208591857e-05, + "loss": 0.9177, + "step": 26404 + }, + { + "epoch": 1.7, + "learning_rate": 1.4355181519568906e-05, + "loss": 0.8957, + "step": 26405 + }, + { + "epoch": 1.7, + "learning_rate": 1.4354966830545955e-05, + "loss": 0.8913, + "step": 26406 + }, + { + "epoch": 1.7, + "learning_rate": 1.4354752141523006e-05, + "loss": 0.9837, + "step": 26407 + }, + { + "epoch": 1.7, + "learning_rate": 1.4354537452500055e-05, + "loss": 0.9705, + "step": 26408 + }, + { + "epoch": 1.7, + "learning_rate": 1.4354322763477104e-05, + "loss": 0.9689, + "step": 26409 + }, + { + "epoch": 1.7, + "learning_rate": 1.4354108074454154e-05, + "loss": 0.959, + "step": 26410 + }, + { + "epoch": 1.7, + "learning_rate": 1.4353893385431203e-05, + "loss": 0.9641, + "step": 26411 + }, + { + "epoch": 1.7, + "learning_rate": 1.4353678696408254e-05, + "loss": 0.8865, + "step": 26412 + }, + { + "epoch": 1.7, + "learning_rate": 1.4353464007385303e-05, + "loss": 0.9745, + "step": 26413 + }, + { + "epoch": 1.7, + "learning_rate": 1.4353249318362352e-05, + "loss": 1.0648, + "step": 26414 + }, + { + "epoch": 1.7, + "learning_rate": 1.4353034629339404e-05, + "loss": 0.9558, + "step": 26415 + }, + { + "epoch": 1.7, + "learning_rate": 1.4352819940316453e-05, + "loss": 0.9421, + "step": 26416 + }, + { + "epoch": 1.7, + "learning_rate": 1.4352605251293504e-05, + "loss": 0.9376, + "step": 26417 + }, + { + "epoch": 1.7, + "learning_rate": 1.4352390562270553e-05, + "loss": 0.9425, + "step": 26418 + }, + { + "epoch": 1.7, + "learning_rate": 1.4352175873247602e-05, + "loss": 0.9446, + "step": 26419 + }, + { + "epoch": 1.7, + "learning_rate": 1.4351961184224652e-05, + "loss": 0.9592, + "step": 26420 + }, + { + "epoch": 1.7, + "learning_rate": 1.4351746495201701e-05, + "loss": 1.0053, + "step": 26421 + }, + { + "epoch": 1.7, + "learning_rate": 1.4351531806178752e-05, + "loss": 1.0403, + "step": 26422 + }, + { + "epoch": 1.7, + "learning_rate": 1.4351317117155801e-05, + "loss": 0.9601, + "step": 26423 + }, + { + "epoch": 1.7, + "learning_rate": 1.435110242813285e-05, + "loss": 0.9226, + "step": 26424 + }, + { + "epoch": 1.7, + "learning_rate": 1.43508877391099e-05, + "loss": 0.9911, + "step": 26425 + }, + { + "epoch": 1.7, + "learning_rate": 1.435067305008695e-05, + "loss": 0.9724, + "step": 26426 + }, + { + "epoch": 1.7, + "learning_rate": 1.4350458361063999e-05, + "loss": 0.9318, + "step": 26427 + }, + { + "epoch": 1.7, + "learning_rate": 1.435024367204105e-05, + "loss": 0.9494, + "step": 26428 + }, + { + "epoch": 1.7, + "learning_rate": 1.43500289830181e-05, + "loss": 0.9265, + "step": 26429 + }, + { + "epoch": 1.7, + "learning_rate": 1.434981429399515e-05, + "loss": 0.9802, + "step": 26430 + }, + { + "epoch": 1.7, + "learning_rate": 1.43495996049722e-05, + "loss": 0.8782, + "step": 26431 + }, + { + "epoch": 1.7, + "learning_rate": 1.434938491594925e-05, + "loss": 0.993, + "step": 26432 + }, + { + "epoch": 1.7, + "learning_rate": 1.4349170226926299e-05, + "loss": 0.9151, + "step": 26433 + }, + { + "epoch": 1.7, + "learning_rate": 1.4348955537903348e-05, + "loss": 1.0232, + "step": 26434 + }, + { + "epoch": 1.7, + "learning_rate": 1.4348740848880399e-05, + "loss": 0.9924, + "step": 26435 + }, + { + "epoch": 1.7, + "learning_rate": 1.4348526159857448e-05, + "loss": 1.0999, + "step": 26436 + }, + { + "epoch": 1.7, + "learning_rate": 1.4348311470834497e-05, + "loss": 1.0211, + "step": 26437 + }, + { + "epoch": 1.7, + "learning_rate": 1.4348096781811547e-05, + "loss": 0.9409, + "step": 26438 + }, + { + "epoch": 1.7, + "learning_rate": 1.4348096781811547e-05, + "loss": 0.9883, + "step": 26439 + }, + { + "epoch": 1.7, + "learning_rate": 1.4347882092788596e-05, + "loss": 0.9887, + "step": 26440 + }, + { + "epoch": 1.7, + "learning_rate": 1.4347667403765645e-05, + "loss": 0.961, + "step": 26441 + }, + { + "epoch": 1.7, + "learning_rate": 1.4347452714742696e-05, + "loss": 1.0285, + "step": 26442 + }, + { + "epoch": 1.7, + "learning_rate": 1.4347238025719745e-05, + "loss": 0.9931, + "step": 26443 + }, + { + "epoch": 1.7, + "learning_rate": 1.4347023336696796e-05, + "loss": 0.9951, + "step": 26444 + }, + { + "epoch": 1.7, + "learning_rate": 1.4346808647673846e-05, + "loss": 0.9237, + "step": 26445 + }, + { + "epoch": 1.7, + "learning_rate": 1.4346593958650897e-05, + "loss": 1.0178, + "step": 26446 + }, + { + "epoch": 1.7, + "learning_rate": 1.4346379269627946e-05, + "loss": 1.0051, + "step": 26447 + }, + { + "epoch": 1.7, + "learning_rate": 1.4346164580604995e-05, + "loss": 0.968, + "step": 26448 + }, + { + "epoch": 1.7, + "learning_rate": 1.4345949891582045e-05, + "loss": 1.0103, + "step": 26449 + }, + { + "epoch": 1.7, + "learning_rate": 1.4345735202559094e-05, + "loss": 0.9648, + "step": 26450 + }, + { + "epoch": 1.7, + "learning_rate": 1.4345520513536143e-05, + "loss": 0.93, + "step": 26451 + }, + { + "epoch": 1.7, + "learning_rate": 1.4345305824513194e-05, + "loss": 0.9534, + "step": 26452 + }, + { + "epoch": 1.7, + "learning_rate": 1.4345091135490243e-05, + "loss": 0.9323, + "step": 26453 + }, + { + "epoch": 1.7, + "learning_rate": 1.4344876446467294e-05, + "loss": 0.8985, + "step": 26454 + }, + { + "epoch": 1.7, + "learning_rate": 1.4344661757444343e-05, + "loss": 1.0005, + "step": 26455 + }, + { + "epoch": 1.7, + "learning_rate": 1.4344447068421392e-05, + "loss": 0.932, + "step": 26456 + }, + { + "epoch": 1.7, + "learning_rate": 1.4344232379398442e-05, + "loss": 0.9414, + "step": 26457 + }, + { + "epoch": 1.7, + "learning_rate": 1.4344017690375491e-05, + "loss": 0.9574, + "step": 26458 + }, + { + "epoch": 1.7, + "learning_rate": 1.4343803001352544e-05, + "loss": 0.8976, + "step": 26459 + }, + { + "epoch": 1.7, + "learning_rate": 1.4343588312329593e-05, + "loss": 1.0124, + "step": 26460 + }, + { + "epoch": 1.7, + "learning_rate": 1.4343373623306641e-05, + "loss": 0.9945, + "step": 26461 + }, + { + "epoch": 1.7, + "learning_rate": 1.4343158934283692e-05, + "loss": 0.9535, + "step": 26462 + }, + { + "epoch": 1.7, + "learning_rate": 1.4342944245260741e-05, + "loss": 0.9021, + "step": 26463 + }, + { + "epoch": 1.7, + "learning_rate": 1.4342729556237792e-05, + "loss": 0.9281, + "step": 26464 + }, + { + "epoch": 1.7, + "learning_rate": 1.434251486721484e-05, + "loss": 0.9645, + "step": 26465 + }, + { + "epoch": 1.7, + "learning_rate": 1.434230017819189e-05, + "loss": 0.9972, + "step": 26466 + }, + { + "epoch": 1.7, + "learning_rate": 1.434208548916894e-05, + "loss": 0.9074, + "step": 26467 + }, + { + "epoch": 1.7, + "learning_rate": 1.434187080014599e-05, + "loss": 0.9014, + "step": 26468 + }, + { + "epoch": 1.7, + "learning_rate": 1.4341656111123038e-05, + "loss": 0.9993, + "step": 26469 + }, + { + "epoch": 1.7, + "learning_rate": 1.4341441422100089e-05, + "loss": 1.012, + "step": 26470 + }, + { + "epoch": 1.7, + "learning_rate": 1.4341226733077138e-05, + "loss": 1.0146, + "step": 26471 + }, + { + "epoch": 1.7, + "learning_rate": 1.4341012044054189e-05, + "loss": 0.9651, + "step": 26472 + }, + { + "epoch": 1.7, + "learning_rate": 1.4340797355031238e-05, + "loss": 1.024, + "step": 26473 + }, + { + "epoch": 1.7, + "learning_rate": 1.434058266600829e-05, + "loss": 0.9412, + "step": 26474 + }, + { + "epoch": 1.7, + "learning_rate": 1.4340367976985339e-05, + "loss": 0.9941, + "step": 26475 + }, + { + "epoch": 1.7, + "learning_rate": 1.4340153287962388e-05, + "loss": 0.9563, + "step": 26476 + }, + { + "epoch": 1.7, + "learning_rate": 1.4339938598939438e-05, + "loss": 0.9423, + "step": 26477 + }, + { + "epoch": 1.7, + "learning_rate": 1.4339723909916487e-05, + "loss": 1.0376, + "step": 26478 + }, + { + "epoch": 1.7, + "learning_rate": 1.4339509220893536e-05, + "loss": 0.8783, + "step": 26479 + }, + { + "epoch": 1.7, + "learning_rate": 1.4339294531870587e-05, + "loss": 0.9772, + "step": 26480 + }, + { + "epoch": 1.7, + "learning_rate": 1.4339079842847636e-05, + "loss": 0.9287, + "step": 26481 + }, + { + "epoch": 1.7, + "learning_rate": 1.4338865153824687e-05, + "loss": 0.9796, + "step": 26482 + }, + { + "epoch": 1.7, + "learning_rate": 1.4338650464801736e-05, + "loss": 1.0284, + "step": 26483 + }, + { + "epoch": 1.7, + "learning_rate": 1.4338435775778785e-05, + "loss": 0.9337, + "step": 26484 + }, + { + "epoch": 1.7, + "learning_rate": 1.4338221086755835e-05, + "loss": 0.9751, + "step": 26485 + }, + { + "epoch": 1.7, + "learning_rate": 1.4338006397732884e-05, + "loss": 0.9604, + "step": 26486 + }, + { + "epoch": 1.7, + "learning_rate": 1.4337791708709933e-05, + "loss": 0.9839, + "step": 26487 + }, + { + "epoch": 1.7, + "learning_rate": 1.4337577019686986e-05, + "loss": 1.0086, + "step": 26488 + }, + { + "epoch": 1.7, + "learning_rate": 1.4337362330664035e-05, + "loss": 0.929, + "step": 26489 + }, + { + "epoch": 1.7, + "learning_rate": 1.4337147641641085e-05, + "loss": 0.9516, + "step": 26490 + }, + { + "epoch": 1.7, + "learning_rate": 1.4336932952618134e-05, + "loss": 0.9472, + "step": 26491 + }, + { + "epoch": 1.7, + "learning_rate": 1.4336718263595185e-05, + "loss": 0.9404, + "step": 26492 + }, + { + "epoch": 1.7, + "learning_rate": 1.4336503574572234e-05, + "loss": 0.9687, + "step": 26493 + }, + { + "epoch": 1.7, + "learning_rate": 1.4336288885549283e-05, + "loss": 1.0004, + "step": 26494 + }, + { + "epoch": 1.7, + "learning_rate": 1.4336074196526333e-05, + "loss": 0.9619, + "step": 26495 + }, + { + "epoch": 1.7, + "learning_rate": 1.4335859507503382e-05, + "loss": 0.942, + "step": 26496 + }, + { + "epoch": 1.7, + "learning_rate": 1.4335644818480431e-05, + "loss": 0.9725, + "step": 26497 + }, + { + "epoch": 1.7, + "learning_rate": 1.4335430129457482e-05, + "loss": 1.0347, + "step": 26498 + }, + { + "epoch": 1.7, + "learning_rate": 1.4335215440434531e-05, + "loss": 0.9624, + "step": 26499 + }, + { + "epoch": 1.7, + "learning_rate": 1.433500075141158e-05, + "loss": 0.9989, + "step": 26500 + }, + { + "epoch": 1.71, + "learning_rate": 1.433478606238863e-05, + "loss": 0.9311, + "step": 26501 + }, + { + "epoch": 1.71, + "learning_rate": 1.433457137336568e-05, + "loss": 0.9727, + "step": 26502 + }, + { + "epoch": 1.71, + "learning_rate": 1.4334356684342732e-05, + "loss": 0.9472, + "step": 26503 + }, + { + "epoch": 1.71, + "learning_rate": 1.4334141995319781e-05, + "loss": 1.0838, + "step": 26504 + }, + { + "epoch": 1.71, + "learning_rate": 1.4333927306296832e-05, + "loss": 0.9938, + "step": 26505 + }, + { + "epoch": 1.71, + "learning_rate": 1.433371261727388e-05, + "loss": 1.0681, + "step": 26506 + }, + { + "epoch": 1.71, + "learning_rate": 1.433349792825093e-05, + "loss": 0.9891, + "step": 26507 + }, + { + "epoch": 1.71, + "learning_rate": 1.433328323922798e-05, + "loss": 0.9371, + "step": 26508 + }, + { + "epoch": 1.71, + "learning_rate": 1.4333068550205029e-05, + "loss": 0.9956, + "step": 26509 + }, + { + "epoch": 1.71, + "learning_rate": 1.4332853861182078e-05, + "loss": 1.0064, + "step": 26510 + }, + { + "epoch": 1.71, + "learning_rate": 1.4332639172159129e-05, + "loss": 0.9975, + "step": 26511 + }, + { + "epoch": 1.71, + "learning_rate": 1.4332424483136178e-05, + "loss": 1.0188, + "step": 26512 + }, + { + "epoch": 1.71, + "learning_rate": 1.4332209794113228e-05, + "loss": 0.8942, + "step": 26513 + }, + { + "epoch": 1.71, + "learning_rate": 1.4331995105090277e-05, + "loss": 1.0121, + "step": 26514 + }, + { + "epoch": 1.71, + "learning_rate": 1.4331780416067326e-05, + "loss": 1.0035, + "step": 26515 + }, + { + "epoch": 1.71, + "learning_rate": 1.4331565727044377e-05, + "loss": 0.9998, + "step": 26516 + }, + { + "epoch": 1.71, + "learning_rate": 1.4331351038021428e-05, + "loss": 0.8442, + "step": 26517 + }, + { + "epoch": 1.71, + "learning_rate": 1.4331136348998478e-05, + "loss": 0.9284, + "step": 26518 + }, + { + "epoch": 1.71, + "learning_rate": 1.4330921659975527e-05, + "loss": 0.9255, + "step": 26519 + }, + { + "epoch": 1.71, + "learning_rate": 1.4330706970952578e-05, + "loss": 0.9823, + "step": 26520 + }, + { + "epoch": 1.71, + "learning_rate": 1.4330492281929627e-05, + "loss": 0.9492, + "step": 26521 + }, + { + "epoch": 1.71, + "learning_rate": 1.4330277592906676e-05, + "loss": 0.9658, + "step": 26522 + }, + { + "epoch": 1.71, + "learning_rate": 1.4330062903883726e-05, + "loss": 0.9524, + "step": 26523 + }, + { + "epoch": 1.71, + "learning_rate": 1.4329848214860775e-05, + "loss": 1.0502, + "step": 26524 + }, + { + "epoch": 1.71, + "learning_rate": 1.4329633525837824e-05, + "loss": 0.9824, + "step": 26525 + }, + { + "epoch": 1.71, + "learning_rate": 1.4329418836814875e-05, + "loss": 1.0015, + "step": 26526 + }, + { + "epoch": 1.71, + "learning_rate": 1.4329204147791924e-05, + "loss": 1.0129, + "step": 26527 + }, + { + "epoch": 1.71, + "learning_rate": 1.4328989458768973e-05, + "loss": 0.9342, + "step": 26528 + }, + { + "epoch": 1.71, + "learning_rate": 1.4328774769746024e-05, + "loss": 0.9571, + "step": 26529 + }, + { + "epoch": 1.71, + "learning_rate": 1.4328560080723073e-05, + "loss": 0.9383, + "step": 26530 + }, + { + "epoch": 1.71, + "learning_rate": 1.4328345391700123e-05, + "loss": 0.9508, + "step": 26531 + }, + { + "epoch": 1.71, + "learning_rate": 1.4328130702677174e-05, + "loss": 0.964, + "step": 26532 + }, + { + "epoch": 1.71, + "learning_rate": 1.4327916013654225e-05, + "loss": 0.9776, + "step": 26533 + }, + { + "epoch": 1.71, + "learning_rate": 1.4327701324631274e-05, + "loss": 1.0494, + "step": 26534 + }, + { + "epoch": 1.71, + "learning_rate": 1.4327486635608322e-05, + "loss": 0.9271, + "step": 26535 + }, + { + "epoch": 1.71, + "learning_rate": 1.4327271946585373e-05, + "loss": 1.0093, + "step": 26536 + }, + { + "epoch": 1.71, + "learning_rate": 1.4327057257562422e-05, + "loss": 1.0185, + "step": 26537 + }, + { + "epoch": 1.71, + "learning_rate": 1.4326842568539471e-05, + "loss": 0.9889, + "step": 26538 + }, + { + "epoch": 1.71, + "learning_rate": 1.4326627879516522e-05, + "loss": 1.0331, + "step": 26539 + }, + { + "epoch": 1.71, + "learning_rate": 1.432641319049357e-05, + "loss": 1.0009, + "step": 26540 + }, + { + "epoch": 1.71, + "learning_rate": 1.4326198501470621e-05, + "loss": 0.9882, + "step": 26541 + }, + { + "epoch": 1.71, + "learning_rate": 1.432598381244767e-05, + "loss": 0.9951, + "step": 26542 + }, + { + "epoch": 1.71, + "learning_rate": 1.432576912342472e-05, + "loss": 1.0477, + "step": 26543 + }, + { + "epoch": 1.71, + "learning_rate": 1.432555443440177e-05, + "loss": 0.9154, + "step": 26544 + }, + { + "epoch": 1.71, + "learning_rate": 1.4325339745378819e-05, + "loss": 0.9718, + "step": 26545 + }, + { + "epoch": 1.71, + "learning_rate": 1.4325125056355868e-05, + "loss": 0.9635, + "step": 26546 + }, + { + "epoch": 1.71, + "learning_rate": 1.432491036733292e-05, + "loss": 0.9831, + "step": 26547 + }, + { + "epoch": 1.71, + "learning_rate": 1.432469567830997e-05, + "loss": 0.9787, + "step": 26548 + }, + { + "epoch": 1.71, + "learning_rate": 1.432448098928702e-05, + "loss": 0.858, + "step": 26549 + }, + { + "epoch": 1.71, + "learning_rate": 1.4324266300264069e-05, + "loss": 0.9043, + "step": 26550 + }, + { + "epoch": 1.71, + "learning_rate": 1.432405161124112e-05, + "loss": 0.959, + "step": 26551 + }, + { + "epoch": 1.71, + "learning_rate": 1.4323836922218168e-05, + "loss": 0.9692, + "step": 26552 + }, + { + "epoch": 1.71, + "learning_rate": 1.4323622233195217e-05, + "loss": 0.9798, + "step": 26553 + }, + { + "epoch": 1.71, + "learning_rate": 1.4323407544172268e-05, + "loss": 1.0064, + "step": 26554 + }, + { + "epoch": 1.71, + "learning_rate": 1.4323192855149317e-05, + "loss": 0.8573, + "step": 26555 + }, + { + "epoch": 1.71, + "learning_rate": 1.4322978166126366e-05, + "loss": 0.9568, + "step": 26556 + }, + { + "epoch": 1.71, + "learning_rate": 1.4322763477103417e-05, + "loss": 0.9707, + "step": 26557 + }, + { + "epoch": 1.71, + "learning_rate": 1.4322548788080466e-05, + "loss": 0.9335, + "step": 26558 + }, + { + "epoch": 1.71, + "learning_rate": 1.4322334099057516e-05, + "loss": 0.9993, + "step": 26559 + }, + { + "epoch": 1.71, + "learning_rate": 1.4322119410034565e-05, + "loss": 0.9446, + "step": 26560 + }, + { + "epoch": 1.71, + "learning_rate": 1.4321904721011618e-05, + "loss": 0.9927, + "step": 26561 + }, + { + "epoch": 1.71, + "learning_rate": 1.4321690031988667e-05, + "loss": 0.9423, + "step": 26562 + }, + { + "epoch": 1.71, + "learning_rate": 1.4321475342965716e-05, + "loss": 0.9373, + "step": 26563 + }, + { + "epoch": 1.71, + "learning_rate": 1.4321260653942766e-05, + "loss": 1.0036, + "step": 26564 + }, + { + "epoch": 1.71, + "learning_rate": 1.4321045964919815e-05, + "loss": 1.0129, + "step": 26565 + }, + { + "epoch": 1.71, + "learning_rate": 1.4320831275896864e-05, + "loss": 0.9589, + "step": 26566 + }, + { + "epoch": 1.71, + "learning_rate": 1.4320616586873915e-05, + "loss": 0.961, + "step": 26567 + }, + { + "epoch": 1.71, + "learning_rate": 1.4320401897850964e-05, + "loss": 1.0083, + "step": 26568 + }, + { + "epoch": 1.71, + "learning_rate": 1.4320187208828014e-05, + "loss": 1.0042, + "step": 26569 + }, + { + "epoch": 1.71, + "learning_rate": 1.4319972519805063e-05, + "loss": 0.9469, + "step": 26570 + }, + { + "epoch": 1.71, + "learning_rate": 1.4319757830782112e-05, + "loss": 1.0572, + "step": 26571 + }, + { + "epoch": 1.71, + "learning_rate": 1.4319543141759163e-05, + "loss": 0.9092, + "step": 26572 + }, + { + "epoch": 1.71, + "learning_rate": 1.4319328452736212e-05, + "loss": 1.0161, + "step": 26573 + }, + { + "epoch": 1.71, + "learning_rate": 1.4319113763713261e-05, + "loss": 0.9277, + "step": 26574 + }, + { + "epoch": 1.71, + "learning_rate": 1.4318899074690312e-05, + "loss": 0.9448, + "step": 26575 + }, + { + "epoch": 1.71, + "learning_rate": 1.4318684385667362e-05, + "loss": 0.9699, + "step": 26576 + }, + { + "epoch": 1.71, + "learning_rate": 1.4318469696644413e-05, + "loss": 0.953, + "step": 26577 + }, + { + "epoch": 1.71, + "learning_rate": 1.4318255007621462e-05, + "loss": 0.9227, + "step": 26578 + }, + { + "epoch": 1.71, + "learning_rate": 1.4318040318598513e-05, + "loss": 0.9484, + "step": 26579 + }, + { + "epoch": 1.71, + "learning_rate": 1.4317825629575561e-05, + "loss": 0.9989, + "step": 26580 + }, + { + "epoch": 1.71, + "learning_rate": 1.431761094055261e-05, + "loss": 1.0454, + "step": 26581 + }, + { + "epoch": 1.71, + "learning_rate": 1.4317396251529661e-05, + "loss": 1.0108, + "step": 26582 + }, + { + "epoch": 1.71, + "learning_rate": 1.431718156250671e-05, + "loss": 0.9384, + "step": 26583 + }, + { + "epoch": 1.71, + "learning_rate": 1.4316966873483759e-05, + "loss": 0.9488, + "step": 26584 + }, + { + "epoch": 1.71, + "learning_rate": 1.431675218446081e-05, + "loss": 0.9905, + "step": 26585 + }, + { + "epoch": 1.71, + "learning_rate": 1.4316537495437859e-05, + "loss": 0.9188, + "step": 26586 + }, + { + "epoch": 1.71, + "learning_rate": 1.4316322806414908e-05, + "loss": 1.0081, + "step": 26587 + }, + { + "epoch": 1.71, + "learning_rate": 1.4316108117391958e-05, + "loss": 0.9681, + "step": 26588 + }, + { + "epoch": 1.71, + "learning_rate": 1.4315893428369007e-05, + "loss": 1.0151, + "step": 26589 + }, + { + "epoch": 1.71, + "learning_rate": 1.431567873934606e-05, + "loss": 1.0077, + "step": 26590 + }, + { + "epoch": 1.71, + "learning_rate": 1.4315464050323109e-05, + "loss": 0.9638, + "step": 26591 + }, + { + "epoch": 1.71, + "learning_rate": 1.431524936130016e-05, + "loss": 0.9517, + "step": 26592 + }, + { + "epoch": 1.71, + "learning_rate": 1.4315034672277208e-05, + "loss": 0.9576, + "step": 26593 + }, + { + "epoch": 1.71, + "learning_rate": 1.4314819983254257e-05, + "loss": 0.9932, + "step": 26594 + }, + { + "epoch": 1.71, + "learning_rate": 1.4314605294231308e-05, + "loss": 1.0209, + "step": 26595 + }, + { + "epoch": 1.71, + "learning_rate": 1.4314390605208357e-05, + "loss": 0.9957, + "step": 26596 + }, + { + "epoch": 1.71, + "learning_rate": 1.4314175916185406e-05, + "loss": 1.0293, + "step": 26597 + }, + { + "epoch": 1.71, + "learning_rate": 1.4313961227162456e-05, + "loss": 1.0293, + "step": 26598 + }, + { + "epoch": 1.71, + "learning_rate": 1.4313746538139505e-05, + "loss": 0.9299, + "step": 26599 + }, + { + "epoch": 1.71, + "learning_rate": 1.4313531849116556e-05, + "loss": 0.9492, + "step": 26600 + }, + { + "epoch": 1.71, + "learning_rate": 1.4313317160093605e-05, + "loss": 1.0033, + "step": 26601 + }, + { + "epoch": 1.71, + "learning_rate": 1.4313102471070654e-05, + "loss": 0.9854, + "step": 26602 + }, + { + "epoch": 1.71, + "learning_rate": 1.4312887782047705e-05, + "loss": 0.9342, + "step": 26603 + }, + { + "epoch": 1.71, + "learning_rate": 1.4312673093024754e-05, + "loss": 0.9441, + "step": 26604 + }, + { + "epoch": 1.71, + "learning_rate": 1.4312458404001806e-05, + "loss": 0.9614, + "step": 26605 + }, + { + "epoch": 1.71, + "learning_rate": 1.4312243714978855e-05, + "loss": 1.0237, + "step": 26606 + }, + { + "epoch": 1.71, + "learning_rate": 1.4312029025955904e-05, + "loss": 1.0248, + "step": 26607 + }, + { + "epoch": 1.71, + "learning_rate": 1.4311814336932955e-05, + "loss": 0.9727, + "step": 26608 + }, + { + "epoch": 1.71, + "learning_rate": 1.4311599647910003e-05, + "loss": 0.897, + "step": 26609 + }, + { + "epoch": 1.71, + "learning_rate": 1.4311384958887054e-05, + "loss": 0.969, + "step": 26610 + }, + { + "epoch": 1.71, + "learning_rate": 1.4311170269864103e-05, + "loss": 1.0138, + "step": 26611 + }, + { + "epoch": 1.71, + "learning_rate": 1.4310955580841152e-05, + "loss": 0.9468, + "step": 26612 + }, + { + "epoch": 1.71, + "learning_rate": 1.4310740891818203e-05, + "loss": 0.9738, + "step": 26613 + }, + { + "epoch": 1.71, + "learning_rate": 1.4310526202795252e-05, + "loss": 0.9326, + "step": 26614 + }, + { + "epoch": 1.71, + "learning_rate": 1.43103115137723e-05, + "loss": 0.9513, + "step": 26615 + }, + { + "epoch": 1.71, + "learning_rate": 1.4310096824749351e-05, + "loss": 1.0437, + "step": 26616 + }, + { + "epoch": 1.71, + "learning_rate": 1.43098821357264e-05, + "loss": 0.9854, + "step": 26617 + }, + { + "epoch": 1.71, + "learning_rate": 1.4309667446703451e-05, + "loss": 1.0002, + "step": 26618 + }, + { + "epoch": 1.71, + "learning_rate": 1.4309452757680502e-05, + "loss": 0.9916, + "step": 26619 + }, + { + "epoch": 1.71, + "learning_rate": 1.4309238068657552e-05, + "loss": 0.9539, + "step": 26620 + }, + { + "epoch": 1.71, + "learning_rate": 1.4309023379634601e-05, + "loss": 0.9659, + "step": 26621 + }, + { + "epoch": 1.71, + "learning_rate": 1.430880869061165e-05, + "loss": 0.9892, + "step": 26622 + }, + { + "epoch": 1.71, + "learning_rate": 1.4308594001588701e-05, + "loss": 0.9428, + "step": 26623 + }, + { + "epoch": 1.71, + "learning_rate": 1.430837931256575e-05, + "loss": 0.9384, + "step": 26624 + }, + { + "epoch": 1.71, + "learning_rate": 1.4308164623542799e-05, + "loss": 0.9239, + "step": 26625 + }, + { + "epoch": 1.71, + "learning_rate": 1.430794993451985e-05, + "loss": 1.0724, + "step": 26626 + }, + { + "epoch": 1.71, + "learning_rate": 1.4307735245496898e-05, + "loss": 1.0493, + "step": 26627 + }, + { + "epoch": 1.71, + "learning_rate": 1.4307520556473949e-05, + "loss": 0.965, + "step": 26628 + }, + { + "epoch": 1.71, + "learning_rate": 1.4307305867450998e-05, + "loss": 0.9455, + "step": 26629 + }, + { + "epoch": 1.71, + "learning_rate": 1.4307091178428047e-05, + "loss": 0.9942, + "step": 26630 + }, + { + "epoch": 1.71, + "learning_rate": 1.4306876489405098e-05, + "loss": 0.9486, + "step": 26631 + }, + { + "epoch": 1.71, + "learning_rate": 1.4306661800382147e-05, + "loss": 0.972, + "step": 26632 + }, + { + "epoch": 1.71, + "learning_rate": 1.4306447111359196e-05, + "loss": 0.9363, + "step": 26633 + }, + { + "epoch": 1.71, + "learning_rate": 1.4306232422336248e-05, + "loss": 0.9932, + "step": 26634 + }, + { + "epoch": 1.71, + "learning_rate": 1.4306017733313297e-05, + "loss": 1.0044, + "step": 26635 + }, + { + "epoch": 1.71, + "learning_rate": 1.4305803044290348e-05, + "loss": 0.9969, + "step": 26636 + }, + { + "epoch": 1.71, + "learning_rate": 1.4305588355267397e-05, + "loss": 1.0593, + "step": 26637 + }, + { + "epoch": 1.71, + "learning_rate": 1.4305373666244447e-05, + "loss": 0.9382, + "step": 26638 + }, + { + "epoch": 1.71, + "learning_rate": 1.4305158977221496e-05, + "loss": 1.0159, + "step": 26639 + }, + { + "epoch": 1.71, + "learning_rate": 1.4304944288198545e-05, + "loss": 0.9575, + "step": 26640 + }, + { + "epoch": 1.71, + "learning_rate": 1.4304729599175596e-05, + "loss": 1.0506, + "step": 26641 + }, + { + "epoch": 1.71, + "learning_rate": 1.4304514910152645e-05, + "loss": 1.0044, + "step": 26642 + }, + { + "epoch": 1.71, + "learning_rate": 1.4304300221129694e-05, + "loss": 1.0184, + "step": 26643 + }, + { + "epoch": 1.71, + "learning_rate": 1.4304085532106744e-05, + "loss": 1.0141, + "step": 26644 + }, + { + "epoch": 1.71, + "learning_rate": 1.4303870843083793e-05, + "loss": 0.956, + "step": 26645 + }, + { + "epoch": 1.71, + "learning_rate": 1.4303656154060844e-05, + "loss": 0.9491, + "step": 26646 + }, + { + "epoch": 1.71, + "learning_rate": 1.4303441465037893e-05, + "loss": 0.9727, + "step": 26647 + }, + { + "epoch": 1.71, + "learning_rate": 1.4303226776014945e-05, + "loss": 1.0216, + "step": 26648 + }, + { + "epoch": 1.71, + "learning_rate": 1.4303012086991994e-05, + "loss": 1.0125, + "step": 26649 + }, + { + "epoch": 1.71, + "learning_rate": 1.4302797397969043e-05, + "loss": 0.9932, + "step": 26650 + }, + { + "epoch": 1.71, + "learning_rate": 1.4302582708946094e-05, + "loss": 0.921, + "step": 26651 + }, + { + "epoch": 1.71, + "learning_rate": 1.4302368019923143e-05, + "loss": 0.9797, + "step": 26652 + }, + { + "epoch": 1.71, + "learning_rate": 1.4302153330900192e-05, + "loss": 1.0126, + "step": 26653 + }, + { + "epoch": 1.71, + "learning_rate": 1.4301938641877242e-05, + "loss": 1.004, + "step": 26654 + }, + { + "epoch": 1.71, + "learning_rate": 1.4301723952854291e-05, + "loss": 0.9388, + "step": 26655 + }, + { + "epoch": 1.71, + "learning_rate": 1.4301509263831342e-05, + "loss": 0.9665, + "step": 26656 + }, + { + "epoch": 1.72, + "learning_rate": 1.4301294574808391e-05, + "loss": 0.9416, + "step": 26657 + }, + { + "epoch": 1.72, + "learning_rate": 1.430107988578544e-05, + "loss": 1.0206, + "step": 26658 + }, + { + "epoch": 1.72, + "learning_rate": 1.430086519676249e-05, + "loss": 0.9801, + "step": 26659 + }, + { + "epoch": 1.72, + "learning_rate": 1.430065050773954e-05, + "loss": 0.972, + "step": 26660 + }, + { + "epoch": 1.72, + "learning_rate": 1.4300435818716589e-05, + "loss": 0.9572, + "step": 26661 + }, + { + "epoch": 1.72, + "learning_rate": 1.430022112969364e-05, + "loss": 0.9945, + "step": 26662 + }, + { + "epoch": 1.72, + "learning_rate": 1.430000644067069e-05, + "loss": 0.9481, + "step": 26663 + }, + { + "epoch": 1.72, + "learning_rate": 1.429979175164774e-05, + "loss": 1.0073, + "step": 26664 + }, + { + "epoch": 1.72, + "learning_rate": 1.429957706262479e-05, + "loss": 1.0173, + "step": 26665 + }, + { + "epoch": 1.72, + "learning_rate": 1.429936237360184e-05, + "loss": 0.9876, + "step": 26666 + }, + { + "epoch": 1.72, + "learning_rate": 1.429914768457889e-05, + "loss": 0.9286, + "step": 26667 + }, + { + "epoch": 1.72, + "learning_rate": 1.4298932995555938e-05, + "loss": 0.9572, + "step": 26668 + }, + { + "epoch": 1.72, + "learning_rate": 1.4298718306532989e-05, + "loss": 0.9865, + "step": 26669 + }, + { + "epoch": 1.72, + "learning_rate": 1.4298503617510038e-05, + "loss": 1.0025, + "step": 26670 + }, + { + "epoch": 1.72, + "learning_rate": 1.4298288928487087e-05, + "loss": 0.8858, + "step": 26671 + }, + { + "epoch": 1.72, + "learning_rate": 1.4298074239464137e-05, + "loss": 0.9602, + "step": 26672 + }, + { + "epoch": 1.72, + "learning_rate": 1.4297859550441186e-05, + "loss": 0.9597, + "step": 26673 + }, + { + "epoch": 1.72, + "learning_rate": 1.4297644861418235e-05, + "loss": 0.9473, + "step": 26674 + }, + { + "epoch": 1.72, + "learning_rate": 1.4297430172395286e-05, + "loss": 1.0045, + "step": 26675 + }, + { + "epoch": 1.72, + "learning_rate": 1.4297215483372335e-05, + "loss": 1.0007, + "step": 26676 + }, + { + "epoch": 1.72, + "learning_rate": 1.4297000794349387e-05, + "loss": 1.0171, + "step": 26677 + }, + { + "epoch": 1.72, + "learning_rate": 1.4296786105326436e-05, + "loss": 0.9265, + "step": 26678 + }, + { + "epoch": 1.72, + "learning_rate": 1.4296571416303487e-05, + "loss": 0.9844, + "step": 26679 + }, + { + "epoch": 1.72, + "learning_rate": 1.4296356727280536e-05, + "loss": 0.9504, + "step": 26680 + }, + { + "epoch": 1.72, + "learning_rate": 1.4296142038257585e-05, + "loss": 1.0073, + "step": 26681 + }, + { + "epoch": 1.72, + "learning_rate": 1.4295927349234636e-05, + "loss": 0.9463, + "step": 26682 + }, + { + "epoch": 1.72, + "learning_rate": 1.4295712660211685e-05, + "loss": 0.9947, + "step": 26683 + }, + { + "epoch": 1.72, + "learning_rate": 1.4295497971188733e-05, + "loss": 0.9506, + "step": 26684 + }, + { + "epoch": 1.72, + "learning_rate": 1.4295283282165784e-05, + "loss": 1.014, + "step": 26685 + }, + { + "epoch": 1.72, + "learning_rate": 1.4295068593142833e-05, + "loss": 0.9261, + "step": 26686 + }, + { + "epoch": 1.72, + "learning_rate": 1.4294853904119884e-05, + "loss": 0.9509, + "step": 26687 + }, + { + "epoch": 1.72, + "learning_rate": 1.4294639215096933e-05, + "loss": 0.9684, + "step": 26688 + }, + { + "epoch": 1.72, + "learning_rate": 1.4294424526073982e-05, + "loss": 0.9821, + "step": 26689 + }, + { + "epoch": 1.72, + "learning_rate": 1.4294209837051032e-05, + "loss": 0.9889, + "step": 26690 + }, + { + "epoch": 1.72, + "learning_rate": 1.4293995148028081e-05, + "loss": 0.9642, + "step": 26691 + }, + { + "epoch": 1.72, + "learning_rate": 1.4293780459005134e-05, + "loss": 0.8877, + "step": 26692 + }, + { + "epoch": 1.72, + "learning_rate": 1.4293565769982183e-05, + "loss": 0.9368, + "step": 26693 + }, + { + "epoch": 1.72, + "learning_rate": 1.4293351080959232e-05, + "loss": 0.983, + "step": 26694 + }, + { + "epoch": 1.72, + "learning_rate": 1.4293136391936282e-05, + "loss": 1.0347, + "step": 26695 + }, + { + "epoch": 1.72, + "learning_rate": 1.4292921702913331e-05, + "loss": 1.0345, + "step": 26696 + }, + { + "epoch": 1.72, + "learning_rate": 1.4292707013890382e-05, + "loss": 0.9893, + "step": 26697 + }, + { + "epoch": 1.72, + "learning_rate": 1.429249232486743e-05, + "loss": 0.9788, + "step": 26698 + }, + { + "epoch": 1.72, + "learning_rate": 1.429227763584448e-05, + "loss": 0.9301, + "step": 26699 + }, + { + "epoch": 1.72, + "learning_rate": 1.429206294682153e-05, + "loss": 0.9256, + "step": 26700 + }, + { + "epoch": 1.72, + "learning_rate": 1.429184825779858e-05, + "loss": 0.9301, + "step": 26701 + }, + { + "epoch": 1.72, + "learning_rate": 1.4291633568775628e-05, + "loss": 0.9765, + "step": 26702 + }, + { + "epoch": 1.72, + "learning_rate": 1.4291418879752679e-05, + "loss": 0.9552, + "step": 26703 + }, + { + "epoch": 1.72, + "learning_rate": 1.4291204190729728e-05, + "loss": 0.884, + "step": 26704 + }, + { + "epoch": 1.72, + "learning_rate": 1.4290989501706779e-05, + "loss": 1.0117, + "step": 26705 + }, + { + "epoch": 1.72, + "learning_rate": 1.429077481268383e-05, + "loss": 0.9759, + "step": 26706 + }, + { + "epoch": 1.72, + "learning_rate": 1.429056012366088e-05, + "loss": 1.0782, + "step": 26707 + }, + { + "epoch": 1.72, + "learning_rate": 1.4290345434637929e-05, + "loss": 1.0303, + "step": 26708 + }, + { + "epoch": 1.72, + "learning_rate": 1.4290130745614978e-05, + "loss": 0.9971, + "step": 26709 + }, + { + "epoch": 1.72, + "learning_rate": 1.4289916056592029e-05, + "loss": 0.911, + "step": 26710 + }, + { + "epoch": 1.72, + "learning_rate": 1.4289701367569078e-05, + "loss": 1.019, + "step": 26711 + }, + { + "epoch": 1.72, + "learning_rate": 1.4289486678546127e-05, + "loss": 1.0075, + "step": 26712 + }, + { + "epoch": 1.72, + "learning_rate": 1.4289271989523177e-05, + "loss": 0.8979, + "step": 26713 + }, + { + "epoch": 1.72, + "learning_rate": 1.4289057300500226e-05, + "loss": 0.9909, + "step": 26714 + }, + { + "epoch": 1.72, + "learning_rate": 1.4288842611477277e-05, + "loss": 1.0569, + "step": 26715 + }, + { + "epoch": 1.72, + "learning_rate": 1.4288627922454326e-05, + "loss": 1.0282, + "step": 26716 + }, + { + "epoch": 1.72, + "learning_rate": 1.4288413233431375e-05, + "loss": 0.9943, + "step": 26717 + }, + { + "epoch": 1.72, + "learning_rate": 1.4288198544408425e-05, + "loss": 0.9629, + "step": 26718 + }, + { + "epoch": 1.72, + "learning_rate": 1.4287983855385474e-05, + "loss": 0.9345, + "step": 26719 + }, + { + "epoch": 1.72, + "learning_rate": 1.4287769166362523e-05, + "loss": 1.0639, + "step": 26720 + }, + { + "epoch": 1.72, + "learning_rate": 1.4287554477339576e-05, + "loss": 1.0241, + "step": 26721 + }, + { + "epoch": 1.72, + "learning_rate": 1.4287339788316625e-05, + "loss": 1.0227, + "step": 26722 + }, + { + "epoch": 1.72, + "learning_rate": 1.4287125099293675e-05, + "loss": 0.9489, + "step": 26723 + }, + { + "epoch": 1.72, + "learning_rate": 1.4286910410270724e-05, + "loss": 0.9613, + "step": 26724 + }, + { + "epoch": 1.72, + "learning_rate": 1.4286695721247775e-05, + "loss": 0.9974, + "step": 26725 + }, + { + "epoch": 1.72, + "learning_rate": 1.4286481032224824e-05, + "loss": 0.9469, + "step": 26726 + }, + { + "epoch": 1.72, + "learning_rate": 1.4286266343201873e-05, + "loss": 0.9745, + "step": 26727 + }, + { + "epoch": 1.72, + "learning_rate": 1.4286051654178923e-05, + "loss": 1.0074, + "step": 26728 + }, + { + "epoch": 1.72, + "learning_rate": 1.4285836965155972e-05, + "loss": 0.9908, + "step": 26729 + }, + { + "epoch": 1.72, + "learning_rate": 1.4285622276133021e-05, + "loss": 0.964, + "step": 26730 + }, + { + "epoch": 1.72, + "learning_rate": 1.4285407587110072e-05, + "loss": 0.9414, + "step": 26731 + }, + { + "epoch": 1.72, + "learning_rate": 1.4285192898087121e-05, + "loss": 0.9496, + "step": 26732 + }, + { + "epoch": 1.72, + "learning_rate": 1.4284978209064172e-05, + "loss": 0.9432, + "step": 26733 + }, + { + "epoch": 1.72, + "learning_rate": 1.428476352004122e-05, + "loss": 1.0083, + "step": 26734 + }, + { + "epoch": 1.72, + "learning_rate": 1.4284548831018273e-05, + "loss": 1.0446, + "step": 26735 + }, + { + "epoch": 1.72, + "learning_rate": 1.4284334141995322e-05, + "loss": 0.9652, + "step": 26736 + }, + { + "epoch": 1.72, + "learning_rate": 1.4284119452972371e-05, + "loss": 1.0133, + "step": 26737 + }, + { + "epoch": 1.72, + "learning_rate": 1.4283904763949422e-05, + "loss": 0.9681, + "step": 26738 + }, + { + "epoch": 1.72, + "learning_rate": 1.428369007492647e-05, + "loss": 1.0033, + "step": 26739 + }, + { + "epoch": 1.72, + "learning_rate": 1.428347538590352e-05, + "loss": 0.9563, + "step": 26740 + }, + { + "epoch": 1.72, + "learning_rate": 1.428326069688057e-05, + "loss": 0.9664, + "step": 26741 + }, + { + "epoch": 1.72, + "learning_rate": 1.428304600785762e-05, + "loss": 1.0182, + "step": 26742 + }, + { + "epoch": 1.72, + "learning_rate": 1.428283131883467e-05, + "loss": 0.9823, + "step": 26743 + }, + { + "epoch": 1.72, + "learning_rate": 1.4282616629811719e-05, + "loss": 0.9369, + "step": 26744 + }, + { + "epoch": 1.72, + "learning_rate": 1.4282401940788768e-05, + "loss": 1.0231, + "step": 26745 + }, + { + "epoch": 1.72, + "learning_rate": 1.4282187251765818e-05, + "loss": 0.943, + "step": 26746 + }, + { + "epoch": 1.72, + "learning_rate": 1.4281972562742867e-05, + "loss": 0.9612, + "step": 26747 + }, + { + "epoch": 1.72, + "learning_rate": 1.4281757873719916e-05, + "loss": 0.99, + "step": 26748 + }, + { + "epoch": 1.72, + "learning_rate": 1.4281543184696967e-05, + "loss": 0.9433, + "step": 26749 + }, + { + "epoch": 1.72, + "learning_rate": 1.4281328495674018e-05, + "loss": 0.9703, + "step": 26750 + }, + { + "epoch": 1.72, + "learning_rate": 1.4281113806651068e-05, + "loss": 0.9011, + "step": 26751 + }, + { + "epoch": 1.72, + "learning_rate": 1.4280899117628117e-05, + "loss": 0.9877, + "step": 26752 + }, + { + "epoch": 1.72, + "learning_rate": 1.4280684428605168e-05, + "loss": 0.9733, + "step": 26753 + }, + { + "epoch": 1.72, + "learning_rate": 1.4280469739582217e-05, + "loss": 0.9402, + "step": 26754 + }, + { + "epoch": 1.72, + "learning_rate": 1.4280255050559266e-05, + "loss": 0.9459, + "step": 26755 + }, + { + "epoch": 1.72, + "learning_rate": 1.4280040361536317e-05, + "loss": 0.9723, + "step": 26756 + }, + { + "epoch": 1.72, + "learning_rate": 1.4279825672513366e-05, + "loss": 1.048, + "step": 26757 + }, + { + "epoch": 1.72, + "learning_rate": 1.4279610983490414e-05, + "loss": 0.9656, + "step": 26758 + }, + { + "epoch": 1.72, + "learning_rate": 1.4279396294467465e-05, + "loss": 0.9233, + "step": 26759 + }, + { + "epoch": 1.72, + "learning_rate": 1.4279181605444514e-05, + "loss": 0.9954, + "step": 26760 + }, + { + "epoch": 1.72, + "learning_rate": 1.4278966916421563e-05, + "loss": 0.9988, + "step": 26761 + }, + { + "epoch": 1.72, + "learning_rate": 1.4278752227398614e-05, + "loss": 0.935, + "step": 26762 + }, + { + "epoch": 1.72, + "learning_rate": 1.4278537538375663e-05, + "loss": 0.8824, + "step": 26763 + }, + { + "epoch": 1.72, + "learning_rate": 1.4278322849352715e-05, + "loss": 0.9613, + "step": 26764 + }, + { + "epoch": 1.72, + "learning_rate": 1.4278108160329764e-05, + "loss": 1.0841, + "step": 26765 + }, + { + "epoch": 1.72, + "learning_rate": 1.4277893471306815e-05, + "loss": 0.9317, + "step": 26766 + }, + { + "epoch": 1.72, + "learning_rate": 1.4277678782283864e-05, + "loss": 0.951, + "step": 26767 + }, + { + "epoch": 1.72, + "learning_rate": 1.4277464093260913e-05, + "loss": 1.0099, + "step": 26768 + }, + { + "epoch": 1.72, + "learning_rate": 1.4277249404237963e-05, + "loss": 0.9411, + "step": 26769 + }, + { + "epoch": 1.72, + "learning_rate": 1.4277034715215012e-05, + "loss": 1.0049, + "step": 26770 + }, + { + "epoch": 1.72, + "learning_rate": 1.4276820026192061e-05, + "loss": 1.0687, + "step": 26771 + }, + { + "epoch": 1.72, + "learning_rate": 1.4276605337169112e-05, + "loss": 1.0283, + "step": 26772 + }, + { + "epoch": 1.72, + "learning_rate": 1.427639064814616e-05, + "loss": 1.0187, + "step": 26773 + }, + { + "epoch": 1.72, + "learning_rate": 1.4276175959123211e-05, + "loss": 0.9963, + "step": 26774 + }, + { + "epoch": 1.72, + "learning_rate": 1.427596127010026e-05, + "loss": 0.9827, + "step": 26775 + }, + { + "epoch": 1.72, + "learning_rate": 1.427574658107731e-05, + "loss": 0.9957, + "step": 26776 + }, + { + "epoch": 1.72, + "learning_rate": 1.427553189205436e-05, + "loss": 1.0202, + "step": 26777 + }, + { + "epoch": 1.72, + "learning_rate": 1.4275317203031409e-05, + "loss": 0.9542, + "step": 26778 + }, + { + "epoch": 1.72, + "learning_rate": 1.4275102514008461e-05, + "loss": 1.0072, + "step": 26779 + }, + { + "epoch": 1.72, + "learning_rate": 1.427488782498551e-05, + "loss": 1.0092, + "step": 26780 + }, + { + "epoch": 1.72, + "learning_rate": 1.427467313596256e-05, + "loss": 0.9639, + "step": 26781 + }, + { + "epoch": 1.72, + "learning_rate": 1.427445844693961e-05, + "loss": 0.9867, + "step": 26782 + }, + { + "epoch": 1.72, + "learning_rate": 1.4274243757916659e-05, + "loss": 0.9834, + "step": 26783 + }, + { + "epoch": 1.72, + "learning_rate": 1.427402906889371e-05, + "loss": 0.9938, + "step": 26784 + }, + { + "epoch": 1.72, + "learning_rate": 1.4273814379870759e-05, + "loss": 1.0568, + "step": 26785 + }, + { + "epoch": 1.72, + "learning_rate": 1.4273599690847808e-05, + "loss": 1.0251, + "step": 26786 + }, + { + "epoch": 1.72, + "learning_rate": 1.4273385001824858e-05, + "loss": 0.9849, + "step": 26787 + }, + { + "epoch": 1.72, + "learning_rate": 1.4273170312801907e-05, + "loss": 0.9705, + "step": 26788 + }, + { + "epoch": 1.72, + "learning_rate": 1.4272955623778956e-05, + "loss": 0.9936, + "step": 26789 + }, + { + "epoch": 1.72, + "learning_rate": 1.4272740934756007e-05, + "loss": 0.9229, + "step": 26790 + }, + { + "epoch": 1.72, + "learning_rate": 1.4272526245733056e-05, + "loss": 0.9209, + "step": 26791 + }, + { + "epoch": 1.72, + "learning_rate": 1.4272311556710106e-05, + "loss": 1.0486, + "step": 26792 + }, + { + "epoch": 1.72, + "learning_rate": 1.4272096867687157e-05, + "loss": 1.024, + "step": 26793 + }, + { + "epoch": 1.72, + "learning_rate": 1.4271882178664208e-05, + "loss": 0.9892, + "step": 26794 + }, + { + "epoch": 1.72, + "learning_rate": 1.4271667489641257e-05, + "loss": 0.9746, + "step": 26795 + }, + { + "epoch": 1.72, + "learning_rate": 1.4271452800618306e-05, + "loss": 1.0292, + "step": 26796 + }, + { + "epoch": 1.72, + "learning_rate": 1.4271238111595356e-05, + "loss": 0.9214, + "step": 26797 + }, + { + "epoch": 1.72, + "learning_rate": 1.4271023422572405e-05, + "loss": 0.9887, + "step": 26798 + }, + { + "epoch": 1.72, + "learning_rate": 1.4270808733549454e-05, + "loss": 0.937, + "step": 26799 + }, + { + "epoch": 1.72, + "learning_rate": 1.4270594044526505e-05, + "loss": 1.0531, + "step": 26800 + }, + { + "epoch": 1.72, + "learning_rate": 1.4270379355503554e-05, + "loss": 1.047, + "step": 26801 + }, + { + "epoch": 1.72, + "learning_rate": 1.4270164666480605e-05, + "loss": 0.9453, + "step": 26802 + }, + { + "epoch": 1.72, + "learning_rate": 1.4269949977457653e-05, + "loss": 0.976, + "step": 26803 + }, + { + "epoch": 1.72, + "learning_rate": 1.4269735288434702e-05, + "loss": 1.0316, + "step": 26804 + }, + { + "epoch": 1.72, + "learning_rate": 1.4269520599411753e-05, + "loss": 1.0094, + "step": 26805 + }, + { + "epoch": 1.72, + "learning_rate": 1.4269305910388802e-05, + "loss": 0.9239, + "step": 26806 + }, + { + "epoch": 1.72, + "learning_rate": 1.4269091221365851e-05, + "loss": 1.0276, + "step": 26807 + }, + { + "epoch": 1.72, + "learning_rate": 1.4268876532342903e-05, + "loss": 0.904, + "step": 26808 + }, + { + "epoch": 1.72, + "learning_rate": 1.4268661843319952e-05, + "loss": 0.9652, + "step": 26809 + }, + { + "epoch": 1.72, + "learning_rate": 1.4268447154297003e-05, + "loss": 0.9496, + "step": 26810 + }, + { + "epoch": 1.72, + "learning_rate": 1.4268232465274052e-05, + "loss": 0.9794, + "step": 26811 + }, + { + "epoch": 1.73, + "learning_rate": 1.4268017776251103e-05, + "loss": 1.0521, + "step": 26812 + }, + { + "epoch": 1.73, + "learning_rate": 1.4267803087228152e-05, + "loss": 0.9843, + "step": 26813 + }, + { + "epoch": 1.73, + "learning_rate": 1.42675883982052e-05, + "loss": 0.9442, + "step": 26814 + }, + { + "epoch": 1.73, + "learning_rate": 1.4267373709182251e-05, + "loss": 1.0288, + "step": 26815 + }, + { + "epoch": 1.73, + "learning_rate": 1.42671590201593e-05, + "loss": 1.0299, + "step": 26816 + }, + { + "epoch": 1.73, + "learning_rate": 1.4266944331136349e-05, + "loss": 0.9813, + "step": 26817 + }, + { + "epoch": 1.73, + "learning_rate": 1.42667296421134e-05, + "loss": 1.0101, + "step": 26818 + }, + { + "epoch": 1.73, + "learning_rate": 1.4266514953090449e-05, + "loss": 0.9661, + "step": 26819 + }, + { + "epoch": 1.73, + "learning_rate": 1.4266300264067498e-05, + "loss": 0.9441, + "step": 26820 + }, + { + "epoch": 1.73, + "learning_rate": 1.4266085575044548e-05, + "loss": 0.9824, + "step": 26821 + }, + { + "epoch": 1.73, + "learning_rate": 1.42658708860216e-05, + "loss": 0.9429, + "step": 26822 + }, + { + "epoch": 1.73, + "learning_rate": 1.426565619699865e-05, + "loss": 1.0088, + "step": 26823 + }, + { + "epoch": 1.73, + "learning_rate": 1.4265441507975699e-05, + "loss": 0.9561, + "step": 26824 + }, + { + "epoch": 1.73, + "learning_rate": 1.426522681895275e-05, + "loss": 0.9389, + "step": 26825 + }, + { + "epoch": 1.73, + "learning_rate": 1.4265012129929798e-05, + "loss": 0.9471, + "step": 26826 + }, + { + "epoch": 1.73, + "learning_rate": 1.4264797440906847e-05, + "loss": 0.9884, + "step": 26827 + }, + { + "epoch": 1.73, + "learning_rate": 1.4264582751883898e-05, + "loss": 0.9611, + "step": 26828 + }, + { + "epoch": 1.73, + "learning_rate": 1.4264368062860947e-05, + "loss": 1.0148, + "step": 26829 + }, + { + "epoch": 1.73, + "learning_rate": 1.4264153373837998e-05, + "loss": 0.9361, + "step": 26830 + }, + { + "epoch": 1.73, + "learning_rate": 1.4263938684815047e-05, + "loss": 0.8921, + "step": 26831 + }, + { + "epoch": 1.73, + "learning_rate": 1.4263723995792095e-05, + "loss": 0.9154, + "step": 26832 + }, + { + "epoch": 1.73, + "learning_rate": 1.4263509306769146e-05, + "loss": 0.9595, + "step": 26833 + }, + { + "epoch": 1.73, + "learning_rate": 1.4263294617746195e-05, + "loss": 1.0568, + "step": 26834 + }, + { + "epoch": 1.73, + "learning_rate": 1.4263079928723244e-05, + "loss": 0.97, + "step": 26835 + }, + { + "epoch": 1.73, + "learning_rate": 1.4262865239700295e-05, + "loss": 0.9343, + "step": 26836 + }, + { + "epoch": 1.73, + "learning_rate": 1.4262650550677345e-05, + "loss": 0.9833, + "step": 26837 + }, + { + "epoch": 1.73, + "learning_rate": 1.4262435861654396e-05, + "loss": 0.9905, + "step": 26838 + }, + { + "epoch": 1.73, + "learning_rate": 1.4262221172631445e-05, + "loss": 0.9578, + "step": 26839 + }, + { + "epoch": 1.73, + "learning_rate": 1.4262006483608496e-05, + "loss": 0.9201, + "step": 26840 + }, + { + "epoch": 1.73, + "learning_rate": 1.4261791794585545e-05, + "loss": 0.9001, + "step": 26841 + }, + { + "epoch": 1.73, + "learning_rate": 1.4261577105562594e-05, + "loss": 0.9549, + "step": 26842 + }, + { + "epoch": 1.73, + "learning_rate": 1.4261362416539644e-05, + "loss": 0.9214, + "step": 26843 + }, + { + "epoch": 1.73, + "learning_rate": 1.4261147727516693e-05, + "loss": 0.9489, + "step": 26844 + }, + { + "epoch": 1.73, + "learning_rate": 1.4260933038493742e-05, + "loss": 0.9798, + "step": 26845 + }, + { + "epoch": 1.73, + "learning_rate": 1.4260718349470793e-05, + "loss": 1.0042, + "step": 26846 + }, + { + "epoch": 1.73, + "learning_rate": 1.4260503660447842e-05, + "loss": 0.8853, + "step": 26847 + }, + { + "epoch": 1.73, + "learning_rate": 1.426028897142489e-05, + "loss": 0.963, + "step": 26848 + }, + { + "epoch": 1.73, + "learning_rate": 1.4260074282401941e-05, + "loss": 1.1162, + "step": 26849 + }, + { + "epoch": 1.73, + "learning_rate": 1.425985959337899e-05, + "loss": 1.0237, + "step": 26850 + }, + { + "epoch": 1.73, + "learning_rate": 1.4259644904356043e-05, + "loss": 1.009, + "step": 26851 + }, + { + "epoch": 1.73, + "learning_rate": 1.4259430215333092e-05, + "loss": 0.917, + "step": 26852 + }, + { + "epoch": 1.73, + "learning_rate": 1.4259215526310142e-05, + "loss": 0.9673, + "step": 26853 + }, + { + "epoch": 1.73, + "learning_rate": 1.4259000837287191e-05, + "loss": 1.008, + "step": 26854 + }, + { + "epoch": 1.73, + "learning_rate": 1.425878614826424e-05, + "loss": 0.9294, + "step": 26855 + }, + { + "epoch": 1.73, + "learning_rate": 1.4258571459241291e-05, + "loss": 0.9857, + "step": 26856 + }, + { + "epoch": 1.73, + "learning_rate": 1.425835677021834e-05, + "loss": 0.9393, + "step": 26857 + }, + { + "epoch": 1.73, + "learning_rate": 1.4258142081195389e-05, + "loss": 0.9999, + "step": 26858 + }, + { + "epoch": 1.73, + "learning_rate": 1.425792739217244e-05, + "loss": 0.9923, + "step": 26859 + }, + { + "epoch": 1.73, + "learning_rate": 1.4257712703149489e-05, + "loss": 1.0218, + "step": 26860 + }, + { + "epoch": 1.73, + "learning_rate": 1.425749801412654e-05, + "loss": 0.9725, + "step": 26861 + }, + { + "epoch": 1.73, + "learning_rate": 1.4257283325103588e-05, + "loss": 0.9609, + "step": 26862 + }, + { + "epoch": 1.73, + "learning_rate": 1.4257068636080637e-05, + "loss": 0.9896, + "step": 26863 + }, + { + "epoch": 1.73, + "learning_rate": 1.4256853947057688e-05, + "loss": 1.0261, + "step": 26864 + }, + { + "epoch": 1.73, + "learning_rate": 1.4256639258034737e-05, + "loss": 0.9144, + "step": 26865 + }, + { + "epoch": 1.73, + "learning_rate": 1.4256424569011789e-05, + "loss": 0.9882, + "step": 26866 + }, + { + "epoch": 1.73, + "learning_rate": 1.4256209879988838e-05, + "loss": 0.9673, + "step": 26867 + }, + { + "epoch": 1.73, + "learning_rate": 1.4255995190965887e-05, + "loss": 0.9475, + "step": 26868 + }, + { + "epoch": 1.73, + "learning_rate": 1.4255780501942938e-05, + "loss": 1.0191, + "step": 26869 + }, + { + "epoch": 1.73, + "learning_rate": 1.4255565812919987e-05, + "loss": 0.9811, + "step": 26870 + }, + { + "epoch": 1.73, + "learning_rate": 1.4255351123897037e-05, + "loss": 1.0038, + "step": 26871 + }, + { + "epoch": 1.73, + "learning_rate": 1.4255136434874086e-05, + "loss": 0.9757, + "step": 26872 + }, + { + "epoch": 1.73, + "learning_rate": 1.4254921745851135e-05, + "loss": 0.9383, + "step": 26873 + }, + { + "epoch": 1.73, + "learning_rate": 1.4254707056828186e-05, + "loss": 0.9459, + "step": 26874 + }, + { + "epoch": 1.73, + "learning_rate": 1.4254492367805235e-05, + "loss": 1.0794, + "step": 26875 + }, + { + "epoch": 1.73, + "learning_rate": 1.4254277678782284e-05, + "loss": 1.0025, + "step": 26876 + }, + { + "epoch": 1.73, + "learning_rate": 1.4254062989759334e-05, + "loss": 0.9099, + "step": 26877 + }, + { + "epoch": 1.73, + "learning_rate": 1.4253848300736383e-05, + "loss": 0.9594, + "step": 26878 + }, + { + "epoch": 1.73, + "learning_rate": 1.4253633611713434e-05, + "loss": 1.0412, + "step": 26879 + }, + { + "epoch": 1.73, + "learning_rate": 1.4253418922690485e-05, + "loss": 0.9834, + "step": 26880 + }, + { + "epoch": 1.73, + "learning_rate": 1.4253204233667535e-05, + "loss": 0.9344, + "step": 26881 + }, + { + "epoch": 1.73, + "learning_rate": 1.4252989544644584e-05, + "loss": 0.9431, + "step": 26882 + }, + { + "epoch": 1.73, + "learning_rate": 1.4252774855621633e-05, + "loss": 0.9596, + "step": 26883 + }, + { + "epoch": 1.73, + "learning_rate": 1.4252560166598684e-05, + "loss": 0.9172, + "step": 26884 + }, + { + "epoch": 1.73, + "learning_rate": 1.4252345477575733e-05, + "loss": 0.9777, + "step": 26885 + }, + { + "epoch": 1.73, + "learning_rate": 1.4252130788552782e-05, + "loss": 0.9745, + "step": 26886 + }, + { + "epoch": 1.73, + "learning_rate": 1.4251916099529833e-05, + "loss": 1.0287, + "step": 26887 + }, + { + "epoch": 1.73, + "learning_rate": 1.4251701410506882e-05, + "loss": 1.0155, + "step": 26888 + }, + { + "epoch": 1.73, + "learning_rate": 1.4251486721483932e-05, + "loss": 0.96, + "step": 26889 + }, + { + "epoch": 1.73, + "learning_rate": 1.4251272032460981e-05, + "loss": 1.0228, + "step": 26890 + }, + { + "epoch": 1.73, + "learning_rate": 1.425105734343803e-05, + "loss": 0.9086, + "step": 26891 + }, + { + "epoch": 1.73, + "learning_rate": 1.425084265441508e-05, + "loss": 0.9406, + "step": 26892 + }, + { + "epoch": 1.73, + "learning_rate": 1.425062796539213e-05, + "loss": 0.8735, + "step": 26893 + }, + { + "epoch": 1.73, + "learning_rate": 1.4250413276369179e-05, + "loss": 0.9694, + "step": 26894 + }, + { + "epoch": 1.73, + "learning_rate": 1.4250198587346231e-05, + "loss": 0.9447, + "step": 26895 + }, + { + "epoch": 1.73, + "learning_rate": 1.424998389832328e-05, + "loss": 0.9414, + "step": 26896 + }, + { + "epoch": 1.73, + "learning_rate": 1.424976920930033e-05, + "loss": 0.9178, + "step": 26897 + }, + { + "epoch": 1.73, + "learning_rate": 1.424955452027738e-05, + "loss": 0.9595, + "step": 26898 + }, + { + "epoch": 1.73, + "learning_rate": 1.424933983125443e-05, + "loss": 1.0318, + "step": 26899 + }, + { + "epoch": 1.73, + "learning_rate": 1.424912514223148e-05, + "loss": 0.9632, + "step": 26900 + }, + { + "epoch": 1.73, + "learning_rate": 1.4248910453208528e-05, + "loss": 1.0825, + "step": 26901 + }, + { + "epoch": 1.73, + "learning_rate": 1.4248695764185579e-05, + "loss": 0.9472, + "step": 26902 + }, + { + "epoch": 1.73, + "learning_rate": 1.4248481075162628e-05, + "loss": 1.0229, + "step": 26903 + }, + { + "epoch": 1.73, + "learning_rate": 1.4248266386139677e-05, + "loss": 0.9388, + "step": 26904 + }, + { + "epoch": 1.73, + "learning_rate": 1.4248051697116728e-05, + "loss": 0.994, + "step": 26905 + }, + { + "epoch": 1.73, + "learning_rate": 1.4247837008093776e-05, + "loss": 0.9443, + "step": 26906 + }, + { + "epoch": 1.73, + "learning_rate": 1.4247622319070825e-05, + "loss": 0.9884, + "step": 26907 + }, + { + "epoch": 1.73, + "learning_rate": 1.4247407630047876e-05, + "loss": 0.9898, + "step": 26908 + }, + { + "epoch": 1.73, + "learning_rate": 1.4247192941024928e-05, + "loss": 0.9952, + "step": 26909 + }, + { + "epoch": 1.73, + "learning_rate": 1.4246978252001977e-05, + "loss": 0.9478, + "step": 26910 + }, + { + "epoch": 1.73, + "learning_rate": 1.4246763562979026e-05, + "loss": 0.9532, + "step": 26911 + }, + { + "epoch": 1.73, + "learning_rate": 1.4246548873956077e-05, + "loss": 0.925, + "step": 26912 + }, + { + "epoch": 1.73, + "learning_rate": 1.4246334184933126e-05, + "loss": 0.9722, + "step": 26913 + }, + { + "epoch": 1.73, + "learning_rate": 1.4246119495910175e-05, + "loss": 0.9448, + "step": 26914 + }, + { + "epoch": 1.73, + "learning_rate": 1.4245904806887226e-05, + "loss": 0.9989, + "step": 26915 + }, + { + "epoch": 1.73, + "learning_rate": 1.4245690117864275e-05, + "loss": 0.9351, + "step": 26916 + }, + { + "epoch": 1.73, + "learning_rate": 1.4245475428841324e-05, + "loss": 0.9196, + "step": 26917 + }, + { + "epoch": 1.73, + "learning_rate": 1.4245260739818374e-05, + "loss": 0.9271, + "step": 26918 + }, + { + "epoch": 1.73, + "learning_rate": 1.4245046050795423e-05, + "loss": 0.9352, + "step": 26919 + }, + { + "epoch": 1.73, + "learning_rate": 1.4244831361772474e-05, + "loss": 1.0594, + "step": 26920 + }, + { + "epoch": 1.73, + "learning_rate": 1.4244616672749523e-05, + "loss": 0.9119, + "step": 26921 + }, + { + "epoch": 1.73, + "learning_rate": 1.4244401983726572e-05, + "loss": 0.9745, + "step": 26922 + }, + { + "epoch": 1.73, + "learning_rate": 1.4244187294703622e-05, + "loss": 1.0306, + "step": 26923 + }, + { + "epoch": 1.73, + "learning_rate": 1.4243972605680673e-05, + "loss": 0.9493, + "step": 26924 + }, + { + "epoch": 1.73, + "learning_rate": 1.4243757916657724e-05, + "loss": 0.9434, + "step": 26925 + }, + { + "epoch": 1.73, + "learning_rate": 1.4243543227634773e-05, + "loss": 0.9492, + "step": 26926 + }, + { + "epoch": 1.73, + "learning_rate": 1.4243328538611822e-05, + "loss": 1.0035, + "step": 26927 + }, + { + "epoch": 1.73, + "learning_rate": 1.4243113849588872e-05, + "loss": 0.9265, + "step": 26928 + }, + { + "epoch": 1.73, + "learning_rate": 1.4242899160565921e-05, + "loss": 1.0227, + "step": 26929 + }, + { + "epoch": 1.73, + "learning_rate": 1.4242684471542972e-05, + "loss": 0.8781, + "step": 26930 + }, + { + "epoch": 1.73, + "learning_rate": 1.4242469782520021e-05, + "loss": 0.9538, + "step": 26931 + }, + { + "epoch": 1.73, + "learning_rate": 1.424225509349707e-05, + "loss": 1.0752, + "step": 26932 + }, + { + "epoch": 1.73, + "learning_rate": 1.424204040447412e-05, + "loss": 0.9892, + "step": 26933 + }, + { + "epoch": 1.73, + "learning_rate": 1.424182571545117e-05, + "loss": 0.971, + "step": 26934 + }, + { + "epoch": 1.73, + "learning_rate": 1.4241611026428218e-05, + "loss": 0.8941, + "step": 26935 + }, + { + "epoch": 1.73, + "learning_rate": 1.4241396337405269e-05, + "loss": 0.9439, + "step": 26936 + }, + { + "epoch": 1.73, + "learning_rate": 1.4241181648382318e-05, + "loss": 0.9847, + "step": 26937 + }, + { + "epoch": 1.73, + "learning_rate": 1.4240966959359369e-05, + "loss": 0.9126, + "step": 26938 + }, + { + "epoch": 1.73, + "learning_rate": 1.424075227033642e-05, + "loss": 0.9853, + "step": 26939 + }, + { + "epoch": 1.73, + "learning_rate": 1.424053758131347e-05, + "loss": 0.9694, + "step": 26940 + }, + { + "epoch": 1.73, + "learning_rate": 1.4240322892290519e-05, + "loss": 1.0438, + "step": 26941 + }, + { + "epoch": 1.73, + "learning_rate": 1.4240108203267568e-05, + "loss": 0.973, + "step": 26942 + }, + { + "epoch": 1.73, + "learning_rate": 1.4239893514244619e-05, + "loss": 0.9537, + "step": 26943 + }, + { + "epoch": 1.73, + "learning_rate": 1.4239678825221668e-05, + "loss": 0.9411, + "step": 26944 + }, + { + "epoch": 1.73, + "learning_rate": 1.4239464136198717e-05, + "loss": 1.0048, + "step": 26945 + }, + { + "epoch": 1.73, + "learning_rate": 1.4239249447175767e-05, + "loss": 0.9495, + "step": 26946 + }, + { + "epoch": 1.73, + "learning_rate": 1.4239034758152816e-05, + "loss": 0.888, + "step": 26947 + }, + { + "epoch": 1.73, + "learning_rate": 1.4238820069129867e-05, + "loss": 0.9659, + "step": 26948 + }, + { + "epoch": 1.73, + "learning_rate": 1.4238605380106916e-05, + "loss": 0.9892, + "step": 26949 + }, + { + "epoch": 1.73, + "learning_rate": 1.4238390691083965e-05, + "loss": 0.9367, + "step": 26950 + }, + { + "epoch": 1.73, + "learning_rate": 1.4238176002061015e-05, + "loss": 1.0068, + "step": 26951 + }, + { + "epoch": 1.73, + "learning_rate": 1.4237961313038064e-05, + "loss": 0.9163, + "step": 26952 + }, + { + "epoch": 1.73, + "learning_rate": 1.4237746624015117e-05, + "loss": 0.9301, + "step": 26953 + }, + { + "epoch": 1.73, + "learning_rate": 1.4237531934992166e-05, + "loss": 1.0846, + "step": 26954 + }, + { + "epoch": 1.73, + "learning_rate": 1.4237317245969215e-05, + "loss": 1.0286, + "step": 26955 + }, + { + "epoch": 1.73, + "learning_rate": 1.4237102556946265e-05, + "loss": 1.0086, + "step": 26956 + }, + { + "epoch": 1.73, + "learning_rate": 1.4236887867923314e-05, + "loss": 0.9184, + "step": 26957 + }, + { + "epoch": 1.73, + "learning_rate": 1.4236673178900365e-05, + "loss": 0.9645, + "step": 26958 + }, + { + "epoch": 1.73, + "learning_rate": 1.4236458489877414e-05, + "loss": 0.9422, + "step": 26959 + }, + { + "epoch": 1.73, + "learning_rate": 1.4236243800854463e-05, + "loss": 0.9293, + "step": 26960 + }, + { + "epoch": 1.73, + "learning_rate": 1.4236029111831514e-05, + "loss": 1.0092, + "step": 26961 + }, + { + "epoch": 1.73, + "learning_rate": 1.4235814422808563e-05, + "loss": 1.038, + "step": 26962 + }, + { + "epoch": 1.73, + "learning_rate": 1.4235599733785612e-05, + "loss": 1.0154, + "step": 26963 + }, + { + "epoch": 1.73, + "learning_rate": 1.4235385044762662e-05, + "loss": 0.9334, + "step": 26964 + }, + { + "epoch": 1.73, + "learning_rate": 1.4235170355739711e-05, + "loss": 0.9593, + "step": 26965 + }, + { + "epoch": 1.73, + "learning_rate": 1.4234955666716762e-05, + "loss": 0.9454, + "step": 26966 + }, + { + "epoch": 1.73, + "learning_rate": 1.423474097769381e-05, + "loss": 0.8931, + "step": 26967 + }, + { + "epoch": 1.74, + "learning_rate": 1.4234526288670863e-05, + "loss": 1.0379, + "step": 26968 + }, + { + "epoch": 1.74, + "learning_rate": 1.4234311599647912e-05, + "loss": 0.99, + "step": 26969 + }, + { + "epoch": 1.74, + "learning_rate": 1.4234096910624961e-05, + "loss": 1.0741, + "step": 26970 + }, + { + "epoch": 1.74, + "learning_rate": 1.4233882221602012e-05, + "loss": 1.0675, + "step": 26971 + }, + { + "epoch": 1.74, + "learning_rate": 1.423366753257906e-05, + "loss": 0.9602, + "step": 26972 + }, + { + "epoch": 1.74, + "learning_rate": 1.423345284355611e-05, + "loss": 0.9565, + "step": 26973 + }, + { + "epoch": 1.74, + "learning_rate": 1.423323815453316e-05, + "loss": 0.9678, + "step": 26974 + }, + { + "epoch": 1.74, + "learning_rate": 1.423302346551021e-05, + "loss": 0.9857, + "step": 26975 + }, + { + "epoch": 1.74, + "learning_rate": 1.423280877648726e-05, + "loss": 0.9726, + "step": 26976 + }, + { + "epoch": 1.74, + "learning_rate": 1.4232594087464309e-05, + "loss": 0.964, + "step": 26977 + }, + { + "epoch": 1.74, + "learning_rate": 1.4232379398441358e-05, + "loss": 0.9363, + "step": 26978 + }, + { + "epoch": 1.74, + "learning_rate": 1.4232164709418409e-05, + "loss": 0.958, + "step": 26979 + }, + { + "epoch": 1.74, + "learning_rate": 1.4231950020395457e-05, + "loss": 0.9954, + "step": 26980 + }, + { + "epoch": 1.74, + "learning_rate": 1.4231735331372506e-05, + "loss": 0.9824, + "step": 26981 + }, + { + "epoch": 1.74, + "learning_rate": 1.4231520642349559e-05, + "loss": 0.9785, + "step": 26982 + }, + { + "epoch": 1.74, + "learning_rate": 1.4231305953326608e-05, + "loss": 0.9393, + "step": 26983 + }, + { + "epoch": 1.74, + "learning_rate": 1.4231091264303658e-05, + "loss": 0.939, + "step": 26984 + }, + { + "epoch": 1.74, + "learning_rate": 1.4230876575280707e-05, + "loss": 0.9292, + "step": 26985 + }, + { + "epoch": 1.74, + "learning_rate": 1.4230661886257758e-05, + "loss": 0.9207, + "step": 26986 + }, + { + "epoch": 1.74, + "learning_rate": 1.4230447197234807e-05, + "loss": 0.9253, + "step": 26987 + }, + { + "epoch": 1.74, + "learning_rate": 1.4230232508211856e-05, + "loss": 0.9757, + "step": 26988 + }, + { + "epoch": 1.74, + "learning_rate": 1.4230017819188907e-05, + "loss": 0.9922, + "step": 26989 + }, + { + "epoch": 1.74, + "learning_rate": 1.4229803130165956e-05, + "loss": 1.0595, + "step": 26990 + }, + { + "epoch": 1.74, + "learning_rate": 1.4229588441143005e-05, + "loss": 0.9555, + "step": 26991 + }, + { + "epoch": 1.74, + "learning_rate": 1.4229373752120055e-05, + "loss": 0.9172, + "step": 26992 + }, + { + "epoch": 1.74, + "learning_rate": 1.4229159063097104e-05, + "loss": 0.973, + "step": 26993 + }, + { + "epoch": 1.74, + "learning_rate": 1.4228944374074153e-05, + "loss": 0.9356, + "step": 26994 + }, + { + "epoch": 1.74, + "learning_rate": 1.4228729685051204e-05, + "loss": 0.9799, + "step": 26995 + }, + { + "epoch": 1.74, + "learning_rate": 1.4228514996028253e-05, + "loss": 0.9547, + "step": 26996 + }, + { + "epoch": 1.74, + "learning_rate": 1.4228300307005305e-05, + "loss": 0.9122, + "step": 26997 + }, + { + "epoch": 1.74, + "learning_rate": 1.4228085617982354e-05, + "loss": 1.0686, + "step": 26998 + }, + { + "epoch": 1.74, + "learning_rate": 1.4227870928959405e-05, + "loss": 1.0478, + "step": 26999 + }, + { + "epoch": 1.74, + "learning_rate": 1.4227656239936454e-05, + "loss": 0.9294, + "step": 27000 + }, + { + "epoch": 1.74, + "eval_laion/cider": 0.49426176376672215, + "eval_laion/meteor": 0.3830627567258297, + "eval_laion/pred_num_words": 36.236328125, + "eval_laion/sacrebleu": 15.246120889019876, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0462486743927002, + "eval_runtime": 96.0585, + "eval_samples_per_second": 5.195, + "eval_steps_per_second": 0.333, + "step": 27000 + }, + { + "epoch": 1.74, + "learning_rate": 1.4227441550913503e-05, + "loss": 0.8931, + "step": 27001 + }, + { + "epoch": 1.74, + "learning_rate": 1.4227226861890553e-05, + "loss": 1.0137, + "step": 27002 + }, + { + "epoch": 1.74, + "learning_rate": 1.4227012172867602e-05, + "loss": 0.9139, + "step": 27003 + }, + { + "epoch": 1.74, + "learning_rate": 1.4226797483844651e-05, + "loss": 0.938, + "step": 27004 + }, + { + "epoch": 1.74, + "learning_rate": 1.4226582794821702e-05, + "loss": 0.8988, + "step": 27005 + }, + { + "epoch": 1.74, + "learning_rate": 1.4226368105798751e-05, + "loss": 1.0262, + "step": 27006 + }, + { + "epoch": 1.74, + "learning_rate": 1.4226153416775802e-05, + "loss": 0.8903, + "step": 27007 + }, + { + "epoch": 1.74, + "learning_rate": 1.422593872775285e-05, + "loss": 0.9854, + "step": 27008 + }, + { + "epoch": 1.74, + "learning_rate": 1.42257240387299e-05, + "loss": 0.9537, + "step": 27009 + }, + { + "epoch": 1.74, + "learning_rate": 1.422550934970695e-05, + "loss": 0.9979, + "step": 27010 + }, + { + "epoch": 1.74, + "learning_rate": 1.4225294660684e-05, + "loss": 0.8922, + "step": 27011 + }, + { + "epoch": 1.74, + "learning_rate": 1.4225079971661051e-05, + "loss": 0.9769, + "step": 27012 + }, + { + "epoch": 1.74, + "learning_rate": 1.42248652826381e-05, + "loss": 0.9785, + "step": 27013 + }, + { + "epoch": 1.74, + "learning_rate": 1.422465059361515e-05, + "loss": 1.0053, + "step": 27014 + }, + { + "epoch": 1.74, + "learning_rate": 1.42244359045922e-05, + "loss": 0.9194, + "step": 27015 + }, + { + "epoch": 1.74, + "learning_rate": 1.4224221215569249e-05, + "loss": 0.9195, + "step": 27016 + }, + { + "epoch": 1.74, + "learning_rate": 1.42240065265463e-05, + "loss": 0.9346, + "step": 27017 + }, + { + "epoch": 1.74, + "learning_rate": 1.4223791837523349e-05, + "loss": 0.919, + "step": 27018 + }, + { + "epoch": 1.74, + "learning_rate": 1.4223577148500398e-05, + "loss": 0.9758, + "step": 27019 + }, + { + "epoch": 1.74, + "learning_rate": 1.4223362459477448e-05, + "loss": 0.929, + "step": 27020 + }, + { + "epoch": 1.74, + "learning_rate": 1.4223147770454497e-05, + "loss": 1.0287, + "step": 27021 + }, + { + "epoch": 1.74, + "learning_rate": 1.4222933081431546e-05, + "loss": 0.9863, + "step": 27022 + }, + { + "epoch": 1.74, + "learning_rate": 1.4222718392408597e-05, + "loss": 0.9591, + "step": 27023 + }, + { + "epoch": 1.74, + "learning_rate": 1.4222503703385646e-05, + "loss": 0.9367, + "step": 27024 + }, + { + "epoch": 1.74, + "learning_rate": 1.4222289014362696e-05, + "loss": 0.9626, + "step": 27025 + }, + { + "epoch": 1.74, + "learning_rate": 1.4222074325339747e-05, + "loss": 0.9828, + "step": 27026 + }, + { + "epoch": 1.74, + "learning_rate": 1.4221859636316798e-05, + "loss": 0.9203, + "step": 27027 + }, + { + "epoch": 1.74, + "learning_rate": 1.4221644947293847e-05, + "loss": 0.917, + "step": 27028 + }, + { + "epoch": 1.74, + "learning_rate": 1.4221430258270896e-05, + "loss": 0.8845, + "step": 27029 + }, + { + "epoch": 1.74, + "learning_rate": 1.4221215569247946e-05, + "loss": 1.0004, + "step": 27030 + }, + { + "epoch": 1.74, + "learning_rate": 1.4221000880224995e-05, + "loss": 1.0543, + "step": 27031 + }, + { + "epoch": 1.74, + "learning_rate": 1.4220786191202044e-05, + "loss": 0.9935, + "step": 27032 + }, + { + "epoch": 1.74, + "learning_rate": 1.4220571502179095e-05, + "loss": 1.0021, + "step": 27033 + }, + { + "epoch": 1.74, + "learning_rate": 1.4220356813156144e-05, + "loss": 1.0656, + "step": 27034 + }, + { + "epoch": 1.74, + "learning_rate": 1.4220142124133195e-05, + "loss": 0.9589, + "step": 27035 + }, + { + "epoch": 1.74, + "learning_rate": 1.4219927435110244e-05, + "loss": 0.9823, + "step": 27036 + }, + { + "epoch": 1.74, + "learning_rate": 1.4219712746087293e-05, + "loss": 0.9103, + "step": 27037 + }, + { + "epoch": 1.74, + "learning_rate": 1.4219498057064343e-05, + "loss": 0.9746, + "step": 27038 + }, + { + "epoch": 1.74, + "learning_rate": 1.4219283368041392e-05, + "loss": 0.9161, + "step": 27039 + }, + { + "epoch": 1.74, + "learning_rate": 1.4219068679018445e-05, + "loss": 0.9676, + "step": 27040 + }, + { + "epoch": 1.74, + "learning_rate": 1.4218853989995493e-05, + "loss": 0.9311, + "step": 27041 + }, + { + "epoch": 1.74, + "learning_rate": 1.4218639300972542e-05, + "loss": 0.9807, + "step": 27042 + }, + { + "epoch": 1.74, + "learning_rate": 1.4218424611949593e-05, + "loss": 0.9706, + "step": 27043 + }, + { + "epoch": 1.74, + "learning_rate": 1.4218209922926642e-05, + "loss": 1.0244, + "step": 27044 + }, + { + "epoch": 1.74, + "learning_rate": 1.4217995233903693e-05, + "loss": 0.9438, + "step": 27045 + }, + { + "epoch": 1.74, + "learning_rate": 1.4217780544880742e-05, + "loss": 1.0006, + "step": 27046 + }, + { + "epoch": 1.74, + "learning_rate": 1.421756585585779e-05, + "loss": 0.9964, + "step": 27047 + }, + { + "epoch": 1.74, + "learning_rate": 1.4217351166834841e-05, + "loss": 0.9173, + "step": 27048 + }, + { + "epoch": 1.74, + "learning_rate": 1.421713647781189e-05, + "loss": 0.9402, + "step": 27049 + }, + { + "epoch": 1.74, + "learning_rate": 1.421692178878894e-05, + "loss": 1.0347, + "step": 27050 + }, + { + "epoch": 1.74, + "learning_rate": 1.421670709976599e-05, + "loss": 0.934, + "step": 27051 + }, + { + "epoch": 1.74, + "learning_rate": 1.4216492410743039e-05, + "loss": 0.9636, + "step": 27052 + }, + { + "epoch": 1.74, + "learning_rate": 1.421627772172009e-05, + "loss": 1.0049, + "step": 27053 + }, + { + "epoch": 1.74, + "learning_rate": 1.4216063032697138e-05, + "loss": 0.9944, + "step": 27054 + }, + { + "epoch": 1.74, + "learning_rate": 1.421584834367419e-05, + "loss": 0.9713, + "step": 27055 + }, + { + "epoch": 1.74, + "learning_rate": 1.421563365465124e-05, + "loss": 0.9642, + "step": 27056 + }, + { + "epoch": 1.74, + "learning_rate": 1.4215418965628289e-05, + "loss": 1.034, + "step": 27057 + }, + { + "epoch": 1.74, + "learning_rate": 1.421520427660534e-05, + "loss": 0.9878, + "step": 27058 + }, + { + "epoch": 1.74, + "learning_rate": 1.4214989587582388e-05, + "loss": 0.883, + "step": 27059 + }, + { + "epoch": 1.74, + "learning_rate": 1.4214774898559437e-05, + "loss": 0.883, + "step": 27060 + }, + { + "epoch": 1.74, + "learning_rate": 1.4214560209536488e-05, + "loss": 0.9606, + "step": 27061 + }, + { + "epoch": 1.74, + "learning_rate": 1.4214345520513537e-05, + "loss": 1.0161, + "step": 27062 + }, + { + "epoch": 1.74, + "learning_rate": 1.4214130831490588e-05, + "loss": 0.9579, + "step": 27063 + }, + { + "epoch": 1.74, + "learning_rate": 1.4213916142467637e-05, + "loss": 0.9812, + "step": 27064 + }, + { + "epoch": 1.74, + "learning_rate": 1.4213701453444686e-05, + "loss": 0.9134, + "step": 27065 + }, + { + "epoch": 1.74, + "learning_rate": 1.4213486764421736e-05, + "loss": 1.0214, + "step": 27066 + }, + { + "epoch": 1.74, + "learning_rate": 1.4213272075398785e-05, + "loss": 0.9644, + "step": 27067 + }, + { + "epoch": 1.74, + "learning_rate": 1.4213057386375834e-05, + "loss": 0.9845, + "step": 27068 + }, + { + "epoch": 1.74, + "learning_rate": 1.4212842697352887e-05, + "loss": 0.8651, + "step": 27069 + }, + { + "epoch": 1.74, + "learning_rate": 1.4212628008329935e-05, + "loss": 0.9203, + "step": 27070 + }, + { + "epoch": 1.74, + "learning_rate": 1.4212413319306986e-05, + "loss": 0.9934, + "step": 27071 + }, + { + "epoch": 1.74, + "learning_rate": 1.4212198630284035e-05, + "loss": 0.9849, + "step": 27072 + }, + { + "epoch": 1.74, + "learning_rate": 1.4211983941261086e-05, + "loss": 0.9821, + "step": 27073 + }, + { + "epoch": 1.74, + "learning_rate": 1.4211769252238135e-05, + "loss": 0.9326, + "step": 27074 + }, + { + "epoch": 1.74, + "learning_rate": 1.4211554563215184e-05, + "loss": 1.0281, + "step": 27075 + }, + { + "epoch": 1.74, + "learning_rate": 1.4211339874192234e-05, + "loss": 0.8877, + "step": 27076 + }, + { + "epoch": 1.74, + "learning_rate": 1.4211125185169283e-05, + "loss": 0.9673, + "step": 27077 + }, + { + "epoch": 1.74, + "learning_rate": 1.4210910496146332e-05, + "loss": 0.9438, + "step": 27078 + }, + { + "epoch": 1.74, + "learning_rate": 1.4210695807123383e-05, + "loss": 0.9236, + "step": 27079 + }, + { + "epoch": 1.74, + "learning_rate": 1.4210481118100432e-05, + "loss": 1.087, + "step": 27080 + }, + { + "epoch": 1.74, + "learning_rate": 1.4210266429077481e-05, + "loss": 0.9186, + "step": 27081 + }, + { + "epoch": 1.74, + "learning_rate": 1.4210051740054532e-05, + "loss": 1.0045, + "step": 27082 + }, + { + "epoch": 1.74, + "learning_rate": 1.420983705103158e-05, + "loss": 0.9726, + "step": 27083 + }, + { + "epoch": 1.74, + "learning_rate": 1.4209622362008633e-05, + "loss": 0.9724, + "step": 27084 + }, + { + "epoch": 1.74, + "learning_rate": 1.4209407672985682e-05, + "loss": 0.9184, + "step": 27085 + }, + { + "epoch": 1.74, + "learning_rate": 1.4209192983962732e-05, + "loss": 0.9793, + "step": 27086 + }, + { + "epoch": 1.74, + "learning_rate": 1.4208978294939781e-05, + "loss": 0.9231, + "step": 27087 + }, + { + "epoch": 1.74, + "learning_rate": 1.420876360591683e-05, + "loss": 1.0244, + "step": 27088 + }, + { + "epoch": 1.74, + "learning_rate": 1.4208548916893881e-05, + "loss": 0.9668, + "step": 27089 + }, + { + "epoch": 1.74, + "learning_rate": 1.420833422787093e-05, + "loss": 0.9952, + "step": 27090 + }, + { + "epoch": 1.74, + "learning_rate": 1.4208119538847979e-05, + "loss": 1.0285, + "step": 27091 + }, + { + "epoch": 1.74, + "learning_rate": 1.420790484982503e-05, + "loss": 0.9081, + "step": 27092 + }, + { + "epoch": 1.74, + "learning_rate": 1.4207690160802079e-05, + "loss": 1.0398, + "step": 27093 + }, + { + "epoch": 1.74, + "learning_rate": 1.420747547177913e-05, + "loss": 0.9212, + "step": 27094 + }, + { + "epoch": 1.74, + "learning_rate": 1.4207260782756178e-05, + "loss": 1.0377, + "step": 27095 + }, + { + "epoch": 1.74, + "learning_rate": 1.4207046093733227e-05, + "loss": 0.9403, + "step": 27096 + }, + { + "epoch": 1.74, + "learning_rate": 1.4206831404710278e-05, + "loss": 0.9115, + "step": 27097 + }, + { + "epoch": 1.74, + "learning_rate": 1.4206616715687329e-05, + "loss": 0.9502, + "step": 27098 + }, + { + "epoch": 1.74, + "learning_rate": 1.420640202666438e-05, + "loss": 0.9636, + "step": 27099 + }, + { + "epoch": 1.74, + "learning_rate": 1.4206187337641428e-05, + "loss": 0.9186, + "step": 27100 + }, + { + "epoch": 1.74, + "learning_rate": 1.4205972648618477e-05, + "loss": 1.0577, + "step": 27101 + }, + { + "epoch": 1.74, + "learning_rate": 1.4205757959595528e-05, + "loss": 1.0492, + "step": 27102 + }, + { + "epoch": 1.74, + "learning_rate": 1.4205543270572577e-05, + "loss": 0.9292, + "step": 27103 + }, + { + "epoch": 1.74, + "learning_rate": 1.4205328581549627e-05, + "loss": 0.9662, + "step": 27104 + }, + { + "epoch": 1.74, + "learning_rate": 1.4205113892526676e-05, + "loss": 0.9515, + "step": 27105 + }, + { + "epoch": 1.74, + "learning_rate": 1.4204899203503725e-05, + "loss": 1.0225, + "step": 27106 + }, + { + "epoch": 1.74, + "learning_rate": 1.4204684514480776e-05, + "loss": 1.0482, + "step": 27107 + }, + { + "epoch": 1.74, + "learning_rate": 1.4204469825457825e-05, + "loss": 1.0339, + "step": 27108 + }, + { + "epoch": 1.74, + "learning_rate": 1.4204255136434874e-05, + "loss": 0.8837, + "step": 27109 + }, + { + "epoch": 1.74, + "learning_rate": 1.4204040447411925e-05, + "loss": 0.894, + "step": 27110 + }, + { + "epoch": 1.74, + "learning_rate": 1.4203825758388974e-05, + "loss": 0.9905, + "step": 27111 + }, + { + "epoch": 1.74, + "learning_rate": 1.4203611069366024e-05, + "loss": 0.9733, + "step": 27112 + }, + { + "epoch": 1.74, + "learning_rate": 1.4203396380343075e-05, + "loss": 1.0144, + "step": 27113 + }, + { + "epoch": 1.74, + "learning_rate": 1.4203181691320126e-05, + "loss": 0.9911, + "step": 27114 + }, + { + "epoch": 1.74, + "learning_rate": 1.4202967002297174e-05, + "loss": 1.0108, + "step": 27115 + }, + { + "epoch": 1.74, + "learning_rate": 1.4202752313274223e-05, + "loss": 1.1237, + "step": 27116 + }, + { + "epoch": 1.74, + "learning_rate": 1.4202537624251274e-05, + "loss": 0.9234, + "step": 27117 + }, + { + "epoch": 1.74, + "learning_rate": 1.4202322935228323e-05, + "loss": 0.9731, + "step": 27118 + }, + { + "epoch": 1.74, + "learning_rate": 1.4202108246205372e-05, + "loss": 0.9813, + "step": 27119 + }, + { + "epoch": 1.74, + "learning_rate": 1.4201893557182423e-05, + "loss": 0.9178, + "step": 27120 + }, + { + "epoch": 1.74, + "learning_rate": 1.4201678868159472e-05, + "loss": 0.973, + "step": 27121 + }, + { + "epoch": 1.74, + "learning_rate": 1.4201464179136522e-05, + "loss": 0.9549, + "step": 27122 + }, + { + "epoch": 1.75, + "learning_rate": 1.4201249490113571e-05, + "loss": 0.9046, + "step": 27123 + }, + { + "epoch": 1.75, + "learning_rate": 1.420103480109062e-05, + "loss": 0.8952, + "step": 27124 + }, + { + "epoch": 1.75, + "learning_rate": 1.4200820112067671e-05, + "loss": 1.052, + "step": 27125 + }, + { + "epoch": 1.75, + "learning_rate": 1.420060542304472e-05, + "loss": 0.9103, + "step": 27126 + }, + { + "epoch": 1.75, + "learning_rate": 1.4200390734021772e-05, + "loss": 1.0348, + "step": 27127 + }, + { + "epoch": 1.75, + "learning_rate": 1.4200176044998821e-05, + "loss": 1.0266, + "step": 27128 + }, + { + "epoch": 1.75, + "learning_rate": 1.419996135597587e-05, + "loss": 0.9449, + "step": 27129 + }, + { + "epoch": 1.75, + "learning_rate": 1.419974666695292e-05, + "loss": 0.9709, + "step": 27130 + }, + { + "epoch": 1.75, + "learning_rate": 1.419953197792997e-05, + "loss": 0.9463, + "step": 27131 + }, + { + "epoch": 1.75, + "learning_rate": 1.419931728890702e-05, + "loss": 0.9231, + "step": 27132 + }, + { + "epoch": 1.75, + "learning_rate": 1.419910259988407e-05, + "loss": 0.9995, + "step": 27133 + }, + { + "epoch": 1.75, + "learning_rate": 1.4198887910861118e-05, + "loss": 0.9542, + "step": 27134 + }, + { + "epoch": 1.75, + "learning_rate": 1.4198673221838169e-05, + "loss": 0.9499, + "step": 27135 + }, + { + "epoch": 1.75, + "learning_rate": 1.4198458532815218e-05, + "loss": 0.9389, + "step": 27136 + }, + { + "epoch": 1.75, + "learning_rate": 1.4198243843792267e-05, + "loss": 0.9584, + "step": 27137 + }, + { + "epoch": 1.75, + "learning_rate": 1.4198029154769318e-05, + "loss": 1.0413, + "step": 27138 + }, + { + "epoch": 1.75, + "learning_rate": 1.4197814465746367e-05, + "loss": 1.0321, + "step": 27139 + }, + { + "epoch": 1.75, + "learning_rate": 1.4197599776723417e-05, + "loss": 0.9323, + "step": 27140 + }, + { + "epoch": 1.75, + "learning_rate": 1.4197385087700466e-05, + "loss": 1.0227, + "step": 27141 + }, + { + "epoch": 1.75, + "learning_rate": 1.4197170398677519e-05, + "loss": 0.9463, + "step": 27142 + }, + { + "epoch": 1.75, + "learning_rate": 1.4196955709654568e-05, + "loss": 0.9488, + "step": 27143 + }, + { + "epoch": 1.75, + "learning_rate": 1.4196741020631616e-05, + "loss": 0.9434, + "step": 27144 + }, + { + "epoch": 1.75, + "learning_rate": 1.4196526331608667e-05, + "loss": 0.9369, + "step": 27145 + }, + { + "epoch": 1.75, + "learning_rate": 1.4196311642585716e-05, + "loss": 1.0438, + "step": 27146 + }, + { + "epoch": 1.75, + "learning_rate": 1.4196096953562765e-05, + "loss": 1.0554, + "step": 27147 + }, + { + "epoch": 1.75, + "learning_rate": 1.4195882264539816e-05, + "loss": 0.9911, + "step": 27148 + }, + { + "epoch": 1.75, + "learning_rate": 1.4195667575516865e-05, + "loss": 0.9571, + "step": 27149 + }, + { + "epoch": 1.75, + "learning_rate": 1.4195452886493915e-05, + "loss": 1.0258, + "step": 27150 + }, + { + "epoch": 1.75, + "learning_rate": 1.4195238197470964e-05, + "loss": 0.9857, + "step": 27151 + }, + { + "epoch": 1.75, + "learning_rate": 1.4195023508448013e-05, + "loss": 0.9838, + "step": 27152 + }, + { + "epoch": 1.75, + "learning_rate": 1.4194808819425064e-05, + "loss": 0.9846, + "step": 27153 + }, + { + "epoch": 1.75, + "learning_rate": 1.4194594130402113e-05, + "loss": 1.0314, + "step": 27154 + }, + { + "epoch": 1.75, + "learning_rate": 1.4194379441379162e-05, + "loss": 1.0505, + "step": 27155 + }, + { + "epoch": 1.75, + "learning_rate": 1.4194164752356214e-05, + "loss": 0.9837, + "step": 27156 + }, + { + "epoch": 1.75, + "learning_rate": 1.4193950063333263e-05, + "loss": 0.9119, + "step": 27157 + }, + { + "epoch": 1.75, + "learning_rate": 1.4193735374310314e-05, + "loss": 0.9927, + "step": 27158 + }, + { + "epoch": 1.75, + "learning_rate": 1.4193520685287363e-05, + "loss": 0.957, + "step": 27159 + }, + { + "epoch": 1.75, + "learning_rate": 1.4193305996264413e-05, + "loss": 1.0283, + "step": 27160 + }, + { + "epoch": 1.75, + "learning_rate": 1.4193091307241462e-05, + "loss": 1.0264, + "step": 27161 + }, + { + "epoch": 1.75, + "learning_rate": 1.4192876618218511e-05, + "loss": 0.9743, + "step": 27162 + }, + { + "epoch": 1.75, + "learning_rate": 1.4192661929195562e-05, + "loss": 0.946, + "step": 27163 + }, + { + "epoch": 1.75, + "learning_rate": 1.4192447240172611e-05, + "loss": 0.954, + "step": 27164 + }, + { + "epoch": 1.75, + "learning_rate": 1.419223255114966e-05, + "loss": 1.0142, + "step": 27165 + }, + { + "epoch": 1.75, + "learning_rate": 1.419201786212671e-05, + "loss": 1.0498, + "step": 27166 + }, + { + "epoch": 1.75, + "learning_rate": 1.419180317310376e-05, + "loss": 0.9977, + "step": 27167 + }, + { + "epoch": 1.75, + "learning_rate": 1.4191588484080809e-05, + "loss": 0.9965, + "step": 27168 + }, + { + "epoch": 1.75, + "learning_rate": 1.419137379505786e-05, + "loss": 1.0294, + "step": 27169 + }, + { + "epoch": 1.75, + "learning_rate": 1.4191159106034908e-05, + "loss": 0.9868, + "step": 27170 + }, + { + "epoch": 1.75, + "learning_rate": 1.419094441701196e-05, + "loss": 0.9867, + "step": 27171 + }, + { + "epoch": 1.75, + "learning_rate": 1.419072972798901e-05, + "loss": 0.9877, + "step": 27172 + }, + { + "epoch": 1.75, + "learning_rate": 1.419051503896606e-05, + "loss": 0.9198, + "step": 27173 + }, + { + "epoch": 1.75, + "learning_rate": 1.4190300349943109e-05, + "loss": 1.0175, + "step": 27174 + }, + { + "epoch": 1.75, + "learning_rate": 1.4190085660920158e-05, + "loss": 0.9605, + "step": 27175 + }, + { + "epoch": 1.75, + "learning_rate": 1.4189870971897209e-05, + "loss": 1.0094, + "step": 27176 + }, + { + "epoch": 1.75, + "learning_rate": 1.4189656282874258e-05, + "loss": 0.9789, + "step": 27177 + }, + { + "epoch": 1.75, + "learning_rate": 1.4189441593851307e-05, + "loss": 1.0067, + "step": 27178 + }, + { + "epoch": 1.75, + "learning_rate": 1.4189226904828357e-05, + "loss": 0.9909, + "step": 27179 + }, + { + "epoch": 1.75, + "learning_rate": 1.4189012215805406e-05, + "loss": 1.0145, + "step": 27180 + }, + { + "epoch": 1.75, + "learning_rate": 1.4188797526782457e-05, + "loss": 0.9342, + "step": 27181 + }, + { + "epoch": 1.75, + "learning_rate": 1.4188582837759506e-05, + "loss": 0.9906, + "step": 27182 + }, + { + "epoch": 1.75, + "learning_rate": 1.4188368148736555e-05, + "loss": 0.9936, + "step": 27183 + }, + { + "epoch": 1.75, + "learning_rate": 1.4188153459713606e-05, + "loss": 0.9244, + "step": 27184 + }, + { + "epoch": 1.75, + "learning_rate": 1.4187938770690656e-05, + "loss": 0.9365, + "step": 27185 + }, + { + "epoch": 1.75, + "learning_rate": 1.4187724081667707e-05, + "loss": 0.9978, + "step": 27186 + }, + { + "epoch": 1.75, + "learning_rate": 1.4187509392644756e-05, + "loss": 0.9623, + "step": 27187 + }, + { + "epoch": 1.75, + "learning_rate": 1.4187294703621805e-05, + "loss": 0.9691, + "step": 27188 + }, + { + "epoch": 1.75, + "learning_rate": 1.4187080014598855e-05, + "loss": 0.9425, + "step": 27189 + }, + { + "epoch": 1.75, + "learning_rate": 1.4186865325575904e-05, + "loss": 0.9143, + "step": 27190 + }, + { + "epoch": 1.75, + "learning_rate": 1.4186650636552955e-05, + "loss": 0.9514, + "step": 27191 + }, + { + "epoch": 1.75, + "learning_rate": 1.4186435947530004e-05, + "loss": 0.9884, + "step": 27192 + }, + { + "epoch": 1.75, + "learning_rate": 1.4186221258507053e-05, + "loss": 0.947, + "step": 27193 + }, + { + "epoch": 1.75, + "learning_rate": 1.4186006569484104e-05, + "loss": 0.9724, + "step": 27194 + }, + { + "epoch": 1.75, + "learning_rate": 1.4185791880461153e-05, + "loss": 0.9507, + "step": 27195 + }, + { + "epoch": 1.75, + "learning_rate": 1.4185577191438202e-05, + "loss": 0.933, + "step": 27196 + }, + { + "epoch": 1.75, + "learning_rate": 1.4185362502415252e-05, + "loss": 0.9614, + "step": 27197 + }, + { + "epoch": 1.75, + "learning_rate": 1.4185147813392301e-05, + "loss": 0.9561, + "step": 27198 + }, + { + "epoch": 1.75, + "learning_rate": 1.4184933124369352e-05, + "loss": 0.9989, + "step": 27199 + }, + { + "epoch": 1.75, + "learning_rate": 1.4184718435346403e-05, + "loss": 0.9564, + "step": 27200 + }, + { + "epoch": 1.75, + "learning_rate": 1.4184503746323453e-05, + "loss": 1.027, + "step": 27201 + }, + { + "epoch": 1.75, + "learning_rate": 1.4184289057300502e-05, + "loss": 0.9611, + "step": 27202 + }, + { + "epoch": 1.75, + "learning_rate": 1.4184074368277551e-05, + "loss": 1.0122, + "step": 27203 + }, + { + "epoch": 1.75, + "learning_rate": 1.4183859679254602e-05, + "loss": 0.9793, + "step": 27204 + }, + { + "epoch": 1.75, + "learning_rate": 1.418364499023165e-05, + "loss": 0.9473, + "step": 27205 + }, + { + "epoch": 1.75, + "learning_rate": 1.41834303012087e-05, + "loss": 1.0084, + "step": 27206 + }, + { + "epoch": 1.75, + "learning_rate": 1.418321561218575e-05, + "loss": 1.026, + "step": 27207 + }, + { + "epoch": 1.75, + "learning_rate": 1.41830009231628e-05, + "loss": 0.8762, + "step": 27208 + }, + { + "epoch": 1.75, + "learning_rate": 1.418278623413985e-05, + "loss": 1.0352, + "step": 27209 + }, + { + "epoch": 1.75, + "learning_rate": 1.4182571545116899e-05, + "loss": 1.0282, + "step": 27210 + }, + { + "epoch": 1.75, + "learning_rate": 1.4182356856093948e-05, + "loss": 0.9692, + "step": 27211 + }, + { + "epoch": 1.75, + "learning_rate": 1.4182142167070999e-05, + "loss": 1.0199, + "step": 27212 + }, + { + "epoch": 1.75, + "learning_rate": 1.4181927478048048e-05, + "loss": 0.9653, + "step": 27213 + }, + { + "epoch": 1.75, + "learning_rate": 1.41817127890251e-05, + "loss": 0.9474, + "step": 27214 + }, + { + "epoch": 1.75, + "learning_rate": 1.4181498100002149e-05, + "loss": 0.9736, + "step": 27215 + }, + { + "epoch": 1.75, + "learning_rate": 1.4181283410979198e-05, + "loss": 0.9425, + "step": 27216 + }, + { + "epoch": 1.75, + "learning_rate": 1.4181068721956249e-05, + "loss": 0.9536, + "step": 27217 + }, + { + "epoch": 1.75, + "learning_rate": 1.4180854032933297e-05, + "loss": 0.9338, + "step": 27218 + }, + { + "epoch": 1.75, + "learning_rate": 1.4180639343910348e-05, + "loss": 0.9313, + "step": 27219 + }, + { + "epoch": 1.75, + "learning_rate": 1.4180424654887397e-05, + "loss": 1.0926, + "step": 27220 + }, + { + "epoch": 1.75, + "learning_rate": 1.4180209965864446e-05, + "loss": 0.9664, + "step": 27221 + }, + { + "epoch": 1.75, + "learning_rate": 1.4179995276841497e-05, + "loss": 0.9439, + "step": 27222 + }, + { + "epoch": 1.75, + "learning_rate": 1.4179780587818546e-05, + "loss": 0.9228, + "step": 27223 + }, + { + "epoch": 1.75, + "learning_rate": 1.4179565898795595e-05, + "loss": 0.9628, + "step": 27224 + }, + { + "epoch": 1.75, + "learning_rate": 1.4179351209772645e-05, + "loss": 1.0446, + "step": 27225 + }, + { + "epoch": 1.75, + "learning_rate": 1.4179136520749694e-05, + "loss": 0.918, + "step": 27226 + }, + { + "epoch": 1.75, + "learning_rate": 1.4178921831726743e-05, + "loss": 0.9974, + "step": 27227 + }, + { + "epoch": 1.75, + "learning_rate": 1.4178707142703794e-05, + "loss": 1.0199, + "step": 27228 + }, + { + "epoch": 1.75, + "learning_rate": 1.4178492453680846e-05, + "loss": 0.9209, + "step": 27229 + }, + { + "epoch": 1.75, + "learning_rate": 1.4178277764657895e-05, + "loss": 0.9741, + "step": 27230 + }, + { + "epoch": 1.75, + "learning_rate": 1.4178063075634944e-05, + "loss": 0.9422, + "step": 27231 + }, + { + "epoch": 1.75, + "learning_rate": 1.4177848386611995e-05, + "loss": 0.9704, + "step": 27232 + }, + { + "epoch": 1.75, + "learning_rate": 1.4177633697589044e-05, + "loss": 0.931, + "step": 27233 + }, + { + "epoch": 1.75, + "learning_rate": 1.4177419008566093e-05, + "loss": 0.9814, + "step": 27234 + }, + { + "epoch": 1.75, + "learning_rate": 1.4177204319543143e-05, + "loss": 0.9435, + "step": 27235 + }, + { + "epoch": 1.75, + "learning_rate": 1.4176989630520192e-05, + "loss": 0.9866, + "step": 27236 + }, + { + "epoch": 1.75, + "learning_rate": 1.4176774941497241e-05, + "loss": 0.966, + "step": 27237 + }, + { + "epoch": 1.75, + "learning_rate": 1.4176560252474292e-05, + "loss": 1.0061, + "step": 27238 + }, + { + "epoch": 1.75, + "learning_rate": 1.4176345563451341e-05, + "loss": 1.0006, + "step": 27239 + }, + { + "epoch": 1.75, + "learning_rate": 1.4176130874428392e-05, + "loss": 0.9924, + "step": 27240 + }, + { + "epoch": 1.75, + "learning_rate": 1.417591618540544e-05, + "loss": 0.981, + "step": 27241 + }, + { + "epoch": 1.75, + "learning_rate": 1.417570149638249e-05, + "loss": 0.9723, + "step": 27242 + }, + { + "epoch": 1.75, + "learning_rate": 1.4175486807359542e-05, + "loss": 0.9752, + "step": 27243 + }, + { + "epoch": 1.75, + "learning_rate": 1.4175272118336591e-05, + "loss": 0.9434, + "step": 27244 + }, + { + "epoch": 1.75, + "learning_rate": 1.4175057429313642e-05, + "loss": 0.9997, + "step": 27245 + }, + { + "epoch": 1.75, + "learning_rate": 1.417484274029069e-05, + "loss": 0.9419, + "step": 27246 + }, + { + "epoch": 1.75, + "learning_rate": 1.417462805126774e-05, + "loss": 0.995, + "step": 27247 + }, + { + "epoch": 1.75, + "learning_rate": 1.417441336224479e-05, + "loss": 0.9518, + "step": 27248 + }, + { + "epoch": 1.75, + "learning_rate": 1.4174198673221839e-05, + "loss": 0.99, + "step": 27249 + }, + { + "epoch": 1.75, + "learning_rate": 1.417398398419889e-05, + "loss": 0.9784, + "step": 27250 + }, + { + "epoch": 1.75, + "learning_rate": 1.4173769295175939e-05, + "loss": 1.0286, + "step": 27251 + }, + { + "epoch": 1.75, + "learning_rate": 1.4173554606152988e-05, + "loss": 0.9071, + "step": 27252 + }, + { + "epoch": 1.75, + "learning_rate": 1.4173339917130038e-05, + "loss": 0.9775, + "step": 27253 + }, + { + "epoch": 1.75, + "learning_rate": 1.4173125228107087e-05, + "loss": 0.99, + "step": 27254 + }, + { + "epoch": 1.75, + "learning_rate": 1.4172910539084136e-05, + "loss": 0.9104, + "step": 27255 + }, + { + "epoch": 1.75, + "learning_rate": 1.4172695850061187e-05, + "loss": 1.0108, + "step": 27256 + }, + { + "epoch": 1.75, + "learning_rate": 1.4172481161038236e-05, + "loss": 0.9951, + "step": 27257 + }, + { + "epoch": 1.75, + "learning_rate": 1.4172266472015288e-05, + "loss": 1.0459, + "step": 27258 + }, + { + "epoch": 1.75, + "learning_rate": 1.4172051782992337e-05, + "loss": 0.9811, + "step": 27259 + }, + { + "epoch": 1.75, + "learning_rate": 1.4171837093969388e-05, + "loss": 0.9652, + "step": 27260 + }, + { + "epoch": 1.75, + "learning_rate": 1.4171622404946437e-05, + "loss": 0.9985, + "step": 27261 + }, + { + "epoch": 1.75, + "learning_rate": 1.4171407715923486e-05, + "loss": 0.995, + "step": 27262 + }, + { + "epoch": 1.75, + "learning_rate": 1.4171193026900536e-05, + "loss": 0.9029, + "step": 27263 + }, + { + "epoch": 1.75, + "learning_rate": 1.4170978337877585e-05, + "loss": 0.9472, + "step": 27264 + }, + { + "epoch": 1.75, + "learning_rate": 1.4170763648854634e-05, + "loss": 0.9698, + "step": 27265 + }, + { + "epoch": 1.75, + "learning_rate": 1.4170548959831685e-05, + "loss": 0.9396, + "step": 27266 + }, + { + "epoch": 1.75, + "learning_rate": 1.4170334270808734e-05, + "loss": 0.9352, + "step": 27267 + }, + { + "epoch": 1.75, + "learning_rate": 1.4170119581785785e-05, + "loss": 1.072, + "step": 27268 + }, + { + "epoch": 1.75, + "learning_rate": 1.4169904892762834e-05, + "loss": 0.9064, + "step": 27269 + }, + { + "epoch": 1.75, + "learning_rate": 1.4169690203739883e-05, + "loss": 0.943, + "step": 27270 + }, + { + "epoch": 1.75, + "learning_rate": 1.4169475514716933e-05, + "loss": 0.9333, + "step": 27271 + }, + { + "epoch": 1.75, + "learning_rate": 1.4169260825693984e-05, + "loss": 0.9515, + "step": 27272 + }, + { + "epoch": 1.75, + "learning_rate": 1.4169046136671035e-05, + "loss": 0.9044, + "step": 27273 + }, + { + "epoch": 1.75, + "learning_rate": 1.4168831447648084e-05, + "loss": 0.9012, + "step": 27274 + }, + { + "epoch": 1.75, + "learning_rate": 1.4168616758625133e-05, + "loss": 1.0014, + "step": 27275 + }, + { + "epoch": 1.75, + "learning_rate": 1.4168402069602183e-05, + "loss": 0.9615, + "step": 27276 + }, + { + "epoch": 1.75, + "learning_rate": 1.4168187380579232e-05, + "loss": 0.9651, + "step": 27277 + }, + { + "epoch": 1.76, + "learning_rate": 1.4167972691556283e-05, + "loss": 1.0005, + "step": 27278 + }, + { + "epoch": 1.76, + "learning_rate": 1.4167758002533332e-05, + "loss": 0.947, + "step": 27279 + }, + { + "epoch": 1.76, + "learning_rate": 1.416754331351038e-05, + "loss": 0.9858, + "step": 27280 + }, + { + "epoch": 1.76, + "learning_rate": 1.4167328624487431e-05, + "loss": 0.9391, + "step": 27281 + }, + { + "epoch": 1.76, + "learning_rate": 1.416711393546448e-05, + "loss": 0.9099, + "step": 27282 + }, + { + "epoch": 1.76, + "learning_rate": 1.416689924644153e-05, + "loss": 0.8531, + "step": 27283 + }, + { + "epoch": 1.76, + "learning_rate": 1.416668455741858e-05, + "loss": 1.0275, + "step": 27284 + }, + { + "epoch": 1.76, + "learning_rate": 1.4166469868395629e-05, + "loss": 1.0048, + "step": 27285 + }, + { + "epoch": 1.76, + "learning_rate": 1.416625517937268e-05, + "loss": 1.0282, + "step": 27286 + }, + { + "epoch": 1.76, + "learning_rate": 1.416604049034973e-05, + "loss": 0.9834, + "step": 27287 + }, + { + "epoch": 1.76, + "learning_rate": 1.4165825801326781e-05, + "loss": 0.9304, + "step": 27288 + }, + { + "epoch": 1.76, + "learning_rate": 1.416561111230383e-05, + "loss": 0.9733, + "step": 27289 + }, + { + "epoch": 1.76, + "learning_rate": 1.4165396423280879e-05, + "loss": 0.9428, + "step": 27290 + }, + { + "epoch": 1.76, + "learning_rate": 1.416518173425793e-05, + "loss": 0.9595, + "step": 27291 + }, + { + "epoch": 1.76, + "learning_rate": 1.4164967045234978e-05, + "loss": 0.9755, + "step": 27292 + }, + { + "epoch": 1.76, + "learning_rate": 1.4164752356212027e-05, + "loss": 1.027, + "step": 27293 + }, + { + "epoch": 1.76, + "learning_rate": 1.4164537667189078e-05, + "loss": 0.9953, + "step": 27294 + }, + { + "epoch": 1.76, + "learning_rate": 1.4164322978166127e-05, + "loss": 0.9355, + "step": 27295 + }, + { + "epoch": 1.76, + "learning_rate": 1.4164108289143178e-05, + "loss": 0.8918, + "step": 27296 + }, + { + "epoch": 1.76, + "learning_rate": 1.4163893600120227e-05, + "loss": 0.9127, + "step": 27297 + }, + { + "epoch": 1.76, + "learning_rate": 1.4163678911097276e-05, + "loss": 0.9186, + "step": 27298 + }, + { + "epoch": 1.76, + "learning_rate": 1.4163464222074326e-05, + "loss": 1.0357, + "step": 27299 + }, + { + "epoch": 1.76, + "learning_rate": 1.4163249533051375e-05, + "loss": 0.993, + "step": 27300 + }, + { + "epoch": 1.76, + "learning_rate": 1.4163034844028424e-05, + "loss": 0.9195, + "step": 27301 + }, + { + "epoch": 1.76, + "learning_rate": 1.4162820155005477e-05, + "loss": 0.9447, + "step": 27302 + }, + { + "epoch": 1.76, + "learning_rate": 1.4162605465982526e-05, + "loss": 0.9824, + "step": 27303 + }, + { + "epoch": 1.76, + "learning_rate": 1.4162390776959576e-05, + "loss": 0.9797, + "step": 27304 + }, + { + "epoch": 1.76, + "learning_rate": 1.4162176087936625e-05, + "loss": 0.9605, + "step": 27305 + }, + { + "epoch": 1.76, + "learning_rate": 1.4161961398913676e-05, + "loss": 0.9565, + "step": 27306 + }, + { + "epoch": 1.76, + "learning_rate": 1.4161746709890725e-05, + "loss": 1.0191, + "step": 27307 + }, + { + "epoch": 1.76, + "learning_rate": 1.4161532020867774e-05, + "loss": 0.9901, + "step": 27308 + }, + { + "epoch": 1.76, + "learning_rate": 1.4161317331844824e-05, + "loss": 1.0188, + "step": 27309 + }, + { + "epoch": 1.76, + "learning_rate": 1.4161102642821873e-05, + "loss": 0.9755, + "step": 27310 + }, + { + "epoch": 1.76, + "learning_rate": 1.4160887953798922e-05, + "loss": 0.9419, + "step": 27311 + }, + { + "epoch": 1.76, + "learning_rate": 1.4160673264775973e-05, + "loss": 0.9751, + "step": 27312 + }, + { + "epoch": 1.76, + "learning_rate": 1.4160458575753022e-05, + "loss": 1.1008, + "step": 27313 + }, + { + "epoch": 1.76, + "learning_rate": 1.4160243886730071e-05, + "loss": 1.0291, + "step": 27314 + }, + { + "epoch": 1.76, + "learning_rate": 1.4160029197707122e-05, + "loss": 0.9988, + "step": 27315 + }, + { + "epoch": 1.76, + "learning_rate": 1.4159814508684174e-05, + "loss": 0.997, + "step": 27316 + }, + { + "epoch": 1.76, + "learning_rate": 1.4159599819661223e-05, + "loss": 0.9005, + "step": 27317 + }, + { + "epoch": 1.76, + "learning_rate": 1.4159385130638272e-05, + "loss": 0.96, + "step": 27318 + }, + { + "epoch": 1.76, + "learning_rate": 1.4159170441615323e-05, + "loss": 1.0151, + "step": 27319 + }, + { + "epoch": 1.76, + "learning_rate": 1.4158955752592372e-05, + "loss": 0.9838, + "step": 27320 + }, + { + "epoch": 1.76, + "learning_rate": 1.415874106356942e-05, + "loss": 0.9755, + "step": 27321 + }, + { + "epoch": 1.76, + "learning_rate": 1.4158526374546471e-05, + "loss": 1.0053, + "step": 27322 + }, + { + "epoch": 1.76, + "learning_rate": 1.415831168552352e-05, + "loss": 1.0614, + "step": 27323 + }, + { + "epoch": 1.76, + "learning_rate": 1.4158096996500569e-05, + "loss": 0.986, + "step": 27324 + }, + { + "epoch": 1.76, + "learning_rate": 1.415788230747762e-05, + "loss": 0.9644, + "step": 27325 + }, + { + "epoch": 1.76, + "learning_rate": 1.4157667618454669e-05, + "loss": 0.8733, + "step": 27326 + }, + { + "epoch": 1.76, + "learning_rate": 1.415745292943172e-05, + "loss": 1.0907, + "step": 27327 + }, + { + "epoch": 1.76, + "learning_rate": 1.4157238240408768e-05, + "loss": 0.9775, + "step": 27328 + }, + { + "epoch": 1.76, + "learning_rate": 1.4157023551385817e-05, + "loss": 1.0257, + "step": 27329 + }, + { + "epoch": 1.76, + "learning_rate": 1.4156808862362868e-05, + "loss": 0.9839, + "step": 27330 + }, + { + "epoch": 1.76, + "learning_rate": 1.4156594173339919e-05, + "loss": 1.0011, + "step": 27331 + }, + { + "epoch": 1.76, + "learning_rate": 1.415637948431697e-05, + "loss": 0.9517, + "step": 27332 + }, + { + "epoch": 1.76, + "learning_rate": 1.4156164795294018e-05, + "loss": 0.9691, + "step": 27333 + }, + { + "epoch": 1.76, + "learning_rate": 1.4155950106271067e-05, + "loss": 0.9499, + "step": 27334 + }, + { + "epoch": 1.76, + "learning_rate": 1.4155735417248118e-05, + "loss": 0.948, + "step": 27335 + }, + { + "epoch": 1.76, + "learning_rate": 1.4155520728225167e-05, + "loss": 0.9904, + "step": 27336 + }, + { + "epoch": 1.76, + "learning_rate": 1.4155306039202217e-05, + "loss": 1.0327, + "step": 27337 + }, + { + "epoch": 1.76, + "learning_rate": 1.4155091350179266e-05, + "loss": 0.9835, + "step": 27338 + }, + { + "epoch": 1.76, + "learning_rate": 1.4154876661156315e-05, + "loss": 0.9472, + "step": 27339 + }, + { + "epoch": 1.76, + "learning_rate": 1.4154661972133366e-05, + "loss": 0.9732, + "step": 27340 + }, + { + "epoch": 1.76, + "learning_rate": 1.4154447283110415e-05, + "loss": 1.0245, + "step": 27341 + }, + { + "epoch": 1.76, + "learning_rate": 1.4154232594087464e-05, + "loss": 0.987, + "step": 27342 + }, + { + "epoch": 1.76, + "learning_rate": 1.4154017905064515e-05, + "loss": 0.9477, + "step": 27343 + }, + { + "epoch": 1.76, + "learning_rate": 1.4153803216041564e-05, + "loss": 1.0135, + "step": 27344 + }, + { + "epoch": 1.76, + "learning_rate": 1.4153588527018616e-05, + "loss": 0.9724, + "step": 27345 + }, + { + "epoch": 1.76, + "learning_rate": 1.4153373837995665e-05, + "loss": 0.979, + "step": 27346 + }, + { + "epoch": 1.76, + "learning_rate": 1.4153159148972716e-05, + "loss": 0.9559, + "step": 27347 + }, + { + "epoch": 1.76, + "learning_rate": 1.4152944459949765e-05, + "loss": 1.0436, + "step": 27348 + }, + { + "epoch": 1.76, + "learning_rate": 1.4152729770926814e-05, + "loss": 1.0078, + "step": 27349 + }, + { + "epoch": 1.76, + "learning_rate": 1.4152515081903864e-05, + "loss": 0.9728, + "step": 27350 + }, + { + "epoch": 1.76, + "learning_rate": 1.4152300392880913e-05, + "loss": 1.0155, + "step": 27351 + }, + { + "epoch": 1.76, + "learning_rate": 1.4152085703857962e-05, + "loss": 0.9363, + "step": 27352 + }, + { + "epoch": 1.76, + "learning_rate": 1.4151871014835013e-05, + "loss": 0.9997, + "step": 27353 + }, + { + "epoch": 1.76, + "learning_rate": 1.4151656325812062e-05, + "loss": 0.9682, + "step": 27354 + }, + { + "epoch": 1.76, + "learning_rate": 1.4151441636789112e-05, + "loss": 0.9069, + "step": 27355 + }, + { + "epoch": 1.76, + "learning_rate": 1.4151226947766161e-05, + "loss": 0.9859, + "step": 27356 + }, + { + "epoch": 1.76, + "learning_rate": 1.415101225874321e-05, + "loss": 0.9657, + "step": 27357 + }, + { + "epoch": 1.76, + "learning_rate": 1.4150797569720261e-05, + "loss": 1.0133, + "step": 27358 + }, + { + "epoch": 1.76, + "learning_rate": 1.415058288069731e-05, + "loss": 0.972, + "step": 27359 + }, + { + "epoch": 1.76, + "learning_rate": 1.4150368191674362e-05, + "loss": 0.9332, + "step": 27360 + }, + { + "epoch": 1.76, + "learning_rate": 1.4150153502651411e-05, + "loss": 0.9371, + "step": 27361 + }, + { + "epoch": 1.76, + "learning_rate": 1.414993881362846e-05, + "loss": 1.0011, + "step": 27362 + }, + { + "epoch": 1.76, + "learning_rate": 1.4149724124605511e-05, + "loss": 0.9489, + "step": 27363 + }, + { + "epoch": 1.76, + "learning_rate": 1.414950943558256e-05, + "loss": 0.9195, + "step": 27364 + }, + { + "epoch": 1.76, + "learning_rate": 1.414929474655961e-05, + "loss": 0.9864, + "step": 27365 + }, + { + "epoch": 1.76, + "learning_rate": 1.414908005753666e-05, + "loss": 0.9529, + "step": 27366 + }, + { + "epoch": 1.76, + "learning_rate": 1.4148865368513708e-05, + "loss": 0.9624, + "step": 27367 + }, + { + "epoch": 1.76, + "learning_rate": 1.4148650679490759e-05, + "loss": 0.9295, + "step": 27368 + }, + { + "epoch": 1.76, + "learning_rate": 1.4148435990467808e-05, + "loss": 0.8697, + "step": 27369 + }, + { + "epoch": 1.76, + "learning_rate": 1.4148221301444857e-05, + "loss": 1.0088, + "step": 27370 + }, + { + "epoch": 1.76, + "learning_rate": 1.4148006612421908e-05, + "loss": 1.0118, + "step": 27371 + }, + { + "epoch": 1.76, + "learning_rate": 1.4147791923398957e-05, + "loss": 1.0272, + "step": 27372 + }, + { + "epoch": 1.76, + "learning_rate": 1.4147577234376007e-05, + "loss": 0.9538, + "step": 27373 + }, + { + "epoch": 1.76, + "learning_rate": 1.4147362545353058e-05, + "loss": 1.0278, + "step": 27374 + }, + { + "epoch": 1.76, + "learning_rate": 1.4147147856330109e-05, + "loss": 0.9606, + "step": 27375 + }, + { + "epoch": 1.76, + "learning_rate": 1.4146933167307158e-05, + "loss": 1.0187, + "step": 27376 + }, + { + "epoch": 1.76, + "learning_rate": 1.4146718478284207e-05, + "loss": 0.9982, + "step": 27377 + }, + { + "epoch": 1.76, + "learning_rate": 1.4146503789261257e-05, + "loss": 1.0183, + "step": 27378 + }, + { + "epoch": 1.76, + "learning_rate": 1.4146289100238306e-05, + "loss": 0.9599, + "step": 27379 + }, + { + "epoch": 1.76, + "learning_rate": 1.4146074411215355e-05, + "loss": 1.021, + "step": 27380 + }, + { + "epoch": 1.76, + "learning_rate": 1.4145859722192406e-05, + "loss": 0.9776, + "step": 27381 + }, + { + "epoch": 1.76, + "learning_rate": 1.4145645033169455e-05, + "loss": 0.986, + "step": 27382 + }, + { + "epoch": 1.76, + "learning_rate": 1.4145430344146505e-05, + "loss": 1.1027, + "step": 27383 + }, + { + "epoch": 1.76, + "learning_rate": 1.4145215655123554e-05, + "loss": 0.9716, + "step": 27384 + }, + { + "epoch": 1.76, + "learning_rate": 1.4145000966100603e-05, + "loss": 0.9531, + "step": 27385 + }, + { + "epoch": 1.76, + "learning_rate": 1.4144786277077654e-05, + "loss": 0.9769, + "step": 27386 + }, + { + "epoch": 1.76, + "learning_rate": 1.4144571588054703e-05, + "loss": 1.0434, + "step": 27387 + }, + { + "epoch": 1.76, + "learning_rate": 1.4144356899031752e-05, + "loss": 0.967, + "step": 27388 + }, + { + "epoch": 1.76, + "learning_rate": 1.4144142210008804e-05, + "loss": 0.9271, + "step": 27389 + }, + { + "epoch": 1.76, + "learning_rate": 1.4143927520985853e-05, + "loss": 0.977, + "step": 27390 + }, + { + "epoch": 1.76, + "learning_rate": 1.4143712831962904e-05, + "loss": 1.0211, + "step": 27391 + }, + { + "epoch": 1.76, + "learning_rate": 1.4143498142939953e-05, + "loss": 0.978, + "step": 27392 + }, + { + "epoch": 1.76, + "learning_rate": 1.4143283453917004e-05, + "loss": 0.9284, + "step": 27393 + }, + { + "epoch": 1.76, + "learning_rate": 1.4143068764894053e-05, + "loss": 0.9893, + "step": 27394 + }, + { + "epoch": 1.76, + "learning_rate": 1.4142854075871101e-05, + "loss": 0.9951, + "step": 27395 + }, + { + "epoch": 1.76, + "learning_rate": 1.4142639386848152e-05, + "loss": 1.0081, + "step": 27396 + }, + { + "epoch": 1.76, + "learning_rate": 1.4142424697825201e-05, + "loss": 0.969, + "step": 27397 + }, + { + "epoch": 1.76, + "learning_rate": 1.414221000880225e-05, + "loss": 0.9223, + "step": 27398 + }, + { + "epoch": 1.76, + "learning_rate": 1.41419953197793e-05, + "loss": 0.9687, + "step": 27399 + }, + { + "epoch": 1.76, + "learning_rate": 1.414178063075635e-05, + "loss": 0.8632, + "step": 27400 + }, + { + "epoch": 1.76, + "learning_rate": 1.4141565941733399e-05, + "loss": 1.0235, + "step": 27401 + }, + { + "epoch": 1.76, + "learning_rate": 1.414135125271045e-05, + "loss": 0.9883, + "step": 27402 + }, + { + "epoch": 1.76, + "learning_rate": 1.4141136563687502e-05, + "loss": 1.0215, + "step": 27403 + }, + { + "epoch": 1.76, + "learning_rate": 1.414092187466455e-05, + "loss": 1.0095, + "step": 27404 + }, + { + "epoch": 1.76, + "learning_rate": 1.41407071856416e-05, + "loss": 0.9681, + "step": 27405 + }, + { + "epoch": 1.76, + "learning_rate": 1.414049249661865e-05, + "loss": 0.8802, + "step": 27406 + }, + { + "epoch": 1.76, + "learning_rate": 1.41402778075957e-05, + "loss": 0.9678, + "step": 27407 + }, + { + "epoch": 1.76, + "learning_rate": 1.4140063118572748e-05, + "loss": 0.9222, + "step": 27408 + }, + { + "epoch": 1.76, + "learning_rate": 1.4139848429549799e-05, + "loss": 1.0483, + "step": 27409 + }, + { + "epoch": 1.76, + "learning_rate": 1.4139633740526848e-05, + "loss": 0.9486, + "step": 27410 + }, + { + "epoch": 1.76, + "learning_rate": 1.4139419051503897e-05, + "loss": 0.9284, + "step": 27411 + }, + { + "epoch": 1.76, + "learning_rate": 1.4139204362480947e-05, + "loss": 0.9278, + "step": 27412 + }, + { + "epoch": 1.76, + "learning_rate": 1.4138989673457996e-05, + "loss": 0.9642, + "step": 27413 + }, + { + "epoch": 1.76, + "learning_rate": 1.4138774984435047e-05, + "loss": 0.9848, + "step": 27414 + }, + { + "epoch": 1.76, + "learning_rate": 1.4138560295412096e-05, + "loss": 1.0338, + "step": 27415 + }, + { + "epoch": 1.76, + "learning_rate": 1.4138345606389145e-05, + "loss": 0.9755, + "step": 27416 + }, + { + "epoch": 1.76, + "learning_rate": 1.4138130917366196e-05, + "loss": 0.962, + "step": 27417 + }, + { + "epoch": 1.76, + "learning_rate": 1.4137916228343246e-05, + "loss": 1.0382, + "step": 27418 + }, + { + "epoch": 1.76, + "learning_rate": 1.4137701539320297e-05, + "loss": 1.0024, + "step": 27419 + }, + { + "epoch": 1.76, + "learning_rate": 1.4137486850297346e-05, + "loss": 0.9331, + "step": 27420 + }, + { + "epoch": 1.76, + "learning_rate": 1.4137272161274395e-05, + "loss": 0.8858, + "step": 27421 + }, + { + "epoch": 1.76, + "learning_rate": 1.4137057472251446e-05, + "loss": 0.9598, + "step": 27422 + }, + { + "epoch": 1.76, + "learning_rate": 1.4136842783228495e-05, + "loss": 1.0364, + "step": 27423 + }, + { + "epoch": 1.76, + "learning_rate": 1.4136628094205545e-05, + "loss": 0.9957, + "step": 27424 + }, + { + "epoch": 1.76, + "learning_rate": 1.4136413405182594e-05, + "loss": 0.993, + "step": 27425 + }, + { + "epoch": 1.76, + "learning_rate": 1.4136198716159643e-05, + "loss": 1.0229, + "step": 27426 + }, + { + "epoch": 1.76, + "learning_rate": 1.4135984027136694e-05, + "loss": 0.9414, + "step": 27427 + }, + { + "epoch": 1.76, + "learning_rate": 1.4135769338113743e-05, + "loss": 0.9462, + "step": 27428 + }, + { + "epoch": 1.76, + "learning_rate": 1.4135554649090792e-05, + "loss": 0.9529, + "step": 27429 + }, + { + "epoch": 1.76, + "learning_rate": 1.4135339960067842e-05, + "loss": 1.025, + "step": 27430 + }, + { + "epoch": 1.76, + "learning_rate": 1.4135125271044891e-05, + "loss": 1.0098, + "step": 27431 + }, + { + "epoch": 1.76, + "learning_rate": 1.4134910582021944e-05, + "loss": 1.0625, + "step": 27432 + }, + { + "epoch": 1.76, + "learning_rate": 1.4134695892998993e-05, + "loss": 0.9021, + "step": 27433 + }, + { + "epoch": 1.77, + "learning_rate": 1.4134481203976043e-05, + "loss": 0.9798, + "step": 27434 + }, + { + "epoch": 1.77, + "learning_rate": 1.4134266514953092e-05, + "loss": 1.0449, + "step": 27435 + }, + { + "epoch": 1.77, + "learning_rate": 1.4134051825930141e-05, + "loss": 0.9291, + "step": 27436 + }, + { + "epoch": 1.77, + "learning_rate": 1.4133837136907192e-05, + "loss": 0.9227, + "step": 27437 + }, + { + "epoch": 1.77, + "learning_rate": 1.4133622447884241e-05, + "loss": 1.0411, + "step": 27438 + }, + { + "epoch": 1.77, + "learning_rate": 1.413340775886129e-05, + "loss": 0.9803, + "step": 27439 + }, + { + "epoch": 1.77, + "learning_rate": 1.413319306983834e-05, + "loss": 0.9621, + "step": 27440 + }, + { + "epoch": 1.77, + "learning_rate": 1.413297838081539e-05, + "loss": 0.9969, + "step": 27441 + }, + { + "epoch": 1.77, + "learning_rate": 1.413276369179244e-05, + "loss": 0.9808, + "step": 27442 + }, + { + "epoch": 1.77, + "learning_rate": 1.4132549002769489e-05, + "loss": 0.9927, + "step": 27443 + }, + { + "epoch": 1.77, + "learning_rate": 1.4132334313746538e-05, + "loss": 1.0593, + "step": 27444 + }, + { + "epoch": 1.77, + "learning_rate": 1.4132119624723589e-05, + "loss": 0.962, + "step": 27445 + }, + { + "epoch": 1.77, + "learning_rate": 1.4131904935700638e-05, + "loss": 1.0121, + "step": 27446 + }, + { + "epoch": 1.77, + "learning_rate": 1.413169024667769e-05, + "loss": 0.9969, + "step": 27447 + }, + { + "epoch": 1.77, + "learning_rate": 1.4131475557654739e-05, + "loss": 1.0015, + "step": 27448 + }, + { + "epoch": 1.77, + "learning_rate": 1.4131260868631788e-05, + "loss": 0.9487, + "step": 27449 + }, + { + "epoch": 1.77, + "learning_rate": 1.4131046179608839e-05, + "loss": 0.9055, + "step": 27450 + }, + { + "epoch": 1.77, + "learning_rate": 1.4130831490585888e-05, + "loss": 0.8679, + "step": 27451 + }, + { + "epoch": 1.77, + "learning_rate": 1.4130616801562938e-05, + "loss": 0.9949, + "step": 27452 + }, + { + "epoch": 1.77, + "learning_rate": 1.4130402112539987e-05, + "loss": 0.9329, + "step": 27453 + }, + { + "epoch": 1.77, + "learning_rate": 1.4130187423517036e-05, + "loss": 0.9607, + "step": 27454 + }, + { + "epoch": 1.77, + "learning_rate": 1.4129972734494087e-05, + "loss": 0.9736, + "step": 27455 + }, + { + "epoch": 1.77, + "learning_rate": 1.4129758045471136e-05, + "loss": 0.8955, + "step": 27456 + }, + { + "epoch": 1.77, + "learning_rate": 1.4129543356448185e-05, + "loss": 1.0307, + "step": 27457 + }, + { + "epoch": 1.77, + "learning_rate": 1.4129328667425235e-05, + "loss": 0.9719, + "step": 27458 + }, + { + "epoch": 1.77, + "learning_rate": 1.4129113978402284e-05, + "loss": 1.0872, + "step": 27459 + }, + { + "epoch": 1.77, + "learning_rate": 1.4128899289379335e-05, + "loss": 0.9865, + "step": 27460 + }, + { + "epoch": 1.77, + "learning_rate": 1.4128684600356386e-05, + "loss": 0.9695, + "step": 27461 + }, + { + "epoch": 1.77, + "learning_rate": 1.4128469911333436e-05, + "loss": 1.0263, + "step": 27462 + }, + { + "epoch": 1.77, + "learning_rate": 1.4128255222310485e-05, + "loss": 1.0202, + "step": 27463 + }, + { + "epoch": 1.77, + "learning_rate": 1.4128040533287534e-05, + "loss": 1.0228, + "step": 27464 + }, + { + "epoch": 1.77, + "learning_rate": 1.4127825844264585e-05, + "loss": 0.9986, + "step": 27465 + }, + { + "epoch": 1.77, + "learning_rate": 1.4127611155241634e-05, + "loss": 0.9689, + "step": 27466 + }, + { + "epoch": 1.77, + "learning_rate": 1.4127396466218683e-05, + "loss": 0.984, + "step": 27467 + }, + { + "epoch": 1.77, + "learning_rate": 1.4127181777195734e-05, + "loss": 0.9109, + "step": 27468 + }, + { + "epoch": 1.77, + "learning_rate": 1.4126967088172782e-05, + "loss": 0.9649, + "step": 27469 + }, + { + "epoch": 1.77, + "learning_rate": 1.4126752399149833e-05, + "loss": 0.9563, + "step": 27470 + }, + { + "epoch": 1.77, + "learning_rate": 1.4126537710126882e-05, + "loss": 0.8905, + "step": 27471 + }, + { + "epoch": 1.77, + "learning_rate": 1.4126323021103931e-05, + "loss": 0.9995, + "step": 27472 + }, + { + "epoch": 1.77, + "learning_rate": 1.4126108332080982e-05, + "loss": 0.9855, + "step": 27473 + }, + { + "epoch": 1.77, + "learning_rate": 1.412589364305803e-05, + "loss": 0.9748, + "step": 27474 + }, + { + "epoch": 1.77, + "learning_rate": 1.412567895403508e-05, + "loss": 0.9795, + "step": 27475 + }, + { + "epoch": 1.77, + "learning_rate": 1.4125464265012132e-05, + "loss": 1.0001, + "step": 27476 + }, + { + "epoch": 1.77, + "learning_rate": 1.4125249575989181e-05, + "loss": 0.9519, + "step": 27477 + }, + { + "epoch": 1.77, + "learning_rate": 1.4125034886966232e-05, + "loss": 0.9966, + "step": 27478 + }, + { + "epoch": 1.77, + "learning_rate": 1.412482019794328e-05, + "loss": 0.9739, + "step": 27479 + }, + { + "epoch": 1.77, + "learning_rate": 1.4124605508920331e-05, + "loss": 0.9233, + "step": 27480 + }, + { + "epoch": 1.77, + "learning_rate": 1.412439081989738e-05, + "loss": 0.9529, + "step": 27481 + }, + { + "epoch": 1.77, + "learning_rate": 1.412417613087443e-05, + "loss": 1.0078, + "step": 27482 + }, + { + "epoch": 1.77, + "learning_rate": 1.412396144185148e-05, + "loss": 0.909, + "step": 27483 + }, + { + "epoch": 1.77, + "learning_rate": 1.4123746752828529e-05, + "loss": 0.966, + "step": 27484 + }, + { + "epoch": 1.77, + "learning_rate": 1.4123532063805578e-05, + "loss": 0.9633, + "step": 27485 + }, + { + "epoch": 1.77, + "learning_rate": 1.4123317374782628e-05, + "loss": 1.0165, + "step": 27486 + }, + { + "epoch": 1.77, + "learning_rate": 1.4123102685759677e-05, + "loss": 1.0224, + "step": 27487 + }, + { + "epoch": 1.77, + "learning_rate": 1.4122887996736726e-05, + "loss": 1.0325, + "step": 27488 + }, + { + "epoch": 1.77, + "learning_rate": 1.4122673307713777e-05, + "loss": 1.0658, + "step": 27489 + }, + { + "epoch": 1.77, + "learning_rate": 1.412245861869083e-05, + "loss": 0.9955, + "step": 27490 + }, + { + "epoch": 1.77, + "learning_rate": 1.4122243929667878e-05, + "loss": 0.9768, + "step": 27491 + }, + { + "epoch": 1.77, + "learning_rate": 1.4122029240644927e-05, + "loss": 1.017, + "step": 27492 + }, + { + "epoch": 1.77, + "learning_rate": 1.4121814551621978e-05, + "loss": 0.9987, + "step": 27493 + }, + { + "epoch": 1.77, + "learning_rate": 1.4121599862599027e-05, + "loss": 1.0379, + "step": 27494 + }, + { + "epoch": 1.77, + "learning_rate": 1.4121385173576076e-05, + "loss": 1.0255, + "step": 27495 + }, + { + "epoch": 1.77, + "learning_rate": 1.4121170484553127e-05, + "loss": 0.9647, + "step": 27496 + }, + { + "epoch": 1.77, + "learning_rate": 1.4120955795530176e-05, + "loss": 1.0093, + "step": 27497 + }, + { + "epoch": 1.77, + "learning_rate": 1.4120741106507225e-05, + "loss": 0.9105, + "step": 27498 + }, + { + "epoch": 1.77, + "learning_rate": 1.4120526417484275e-05, + "loss": 1.0412, + "step": 27499 + }, + { + "epoch": 1.77, + "learning_rate": 1.4120311728461324e-05, + "loss": 0.9177, + "step": 27500 + }, + { + "epoch": 1.77, + "learning_rate": 1.4120097039438375e-05, + "loss": 0.8849, + "step": 27501 + }, + { + "epoch": 1.77, + "learning_rate": 1.4119882350415424e-05, + "loss": 1.015, + "step": 27502 + }, + { + "epoch": 1.77, + "learning_rate": 1.4119667661392473e-05, + "loss": 1.0047, + "step": 27503 + }, + { + "epoch": 1.77, + "learning_rate": 1.4119452972369523e-05, + "loss": 0.9734, + "step": 27504 + }, + { + "epoch": 1.77, + "learning_rate": 1.4119238283346574e-05, + "loss": 0.9015, + "step": 27505 + }, + { + "epoch": 1.77, + "learning_rate": 1.4119023594323625e-05, + "loss": 0.9812, + "step": 27506 + }, + { + "epoch": 1.77, + "learning_rate": 1.4118808905300674e-05, + "loss": 0.9759, + "step": 27507 + }, + { + "epoch": 1.77, + "learning_rate": 1.4118594216277723e-05, + "loss": 1.0205, + "step": 27508 + }, + { + "epoch": 1.77, + "learning_rate": 1.4118379527254773e-05, + "loss": 1.0151, + "step": 27509 + }, + { + "epoch": 1.77, + "learning_rate": 1.4118164838231822e-05, + "loss": 0.991, + "step": 27510 + }, + { + "epoch": 1.77, + "learning_rate": 1.4117950149208873e-05, + "loss": 0.917, + "step": 27511 + }, + { + "epoch": 1.77, + "learning_rate": 1.4117735460185922e-05, + "loss": 0.9303, + "step": 27512 + }, + { + "epoch": 1.77, + "learning_rate": 1.411752077116297e-05, + "loss": 0.9823, + "step": 27513 + }, + { + "epoch": 1.77, + "learning_rate": 1.4117306082140021e-05, + "loss": 0.9974, + "step": 27514 + }, + { + "epoch": 1.77, + "learning_rate": 1.411709139311707e-05, + "loss": 0.932, + "step": 27515 + }, + { + "epoch": 1.77, + "learning_rate": 1.411687670409412e-05, + "loss": 0.9553, + "step": 27516 + }, + { + "epoch": 1.77, + "learning_rate": 1.411666201507117e-05, + "loss": 0.9719, + "step": 27517 + }, + { + "epoch": 1.77, + "learning_rate": 1.4116447326048219e-05, + "loss": 0.9957, + "step": 27518 + }, + { + "epoch": 1.77, + "learning_rate": 1.4116232637025271e-05, + "loss": 0.964, + "step": 27519 + }, + { + "epoch": 1.77, + "learning_rate": 1.411601794800232e-05, + "loss": 0.9615, + "step": 27520 + }, + { + "epoch": 1.77, + "learning_rate": 1.4115803258979371e-05, + "loss": 0.9472, + "step": 27521 + }, + { + "epoch": 1.77, + "learning_rate": 1.411558856995642e-05, + "loss": 0.9723, + "step": 27522 + }, + { + "epoch": 1.77, + "learning_rate": 1.4115373880933469e-05, + "loss": 0.9577, + "step": 27523 + }, + { + "epoch": 1.77, + "learning_rate": 1.411515919191052e-05, + "loss": 0.9742, + "step": 27524 + }, + { + "epoch": 1.77, + "learning_rate": 1.4114944502887569e-05, + "loss": 1.0185, + "step": 27525 + }, + { + "epoch": 1.77, + "learning_rate": 1.4114729813864618e-05, + "loss": 0.9914, + "step": 27526 + }, + { + "epoch": 1.77, + "learning_rate": 1.4114515124841668e-05, + "loss": 0.8479, + "step": 27527 + }, + { + "epoch": 1.77, + "learning_rate": 1.4114300435818717e-05, + "loss": 0.9749, + "step": 27528 + }, + { + "epoch": 1.77, + "learning_rate": 1.4114085746795768e-05, + "loss": 0.912, + "step": 27529 + }, + { + "epoch": 1.77, + "learning_rate": 1.4113871057772817e-05, + "loss": 0.984, + "step": 27530 + }, + { + "epoch": 1.77, + "learning_rate": 1.4113656368749866e-05, + "loss": 1.0074, + "step": 27531 + }, + { + "epoch": 1.77, + "learning_rate": 1.4113441679726916e-05, + "loss": 0.9887, + "step": 27532 + }, + { + "epoch": 1.77, + "learning_rate": 1.4113226990703965e-05, + "loss": 0.9799, + "step": 27533 + }, + { + "epoch": 1.77, + "learning_rate": 1.4113012301681018e-05, + "loss": 0.9588, + "step": 27534 + }, + { + "epoch": 1.77, + "learning_rate": 1.4112797612658067e-05, + "loss": 0.9544, + "step": 27535 + }, + { + "epoch": 1.77, + "learning_rate": 1.4112582923635116e-05, + "loss": 0.976, + "step": 27536 + }, + { + "epoch": 1.77, + "learning_rate": 1.4112368234612166e-05, + "loss": 0.9456, + "step": 27537 + }, + { + "epoch": 1.77, + "learning_rate": 1.4112153545589215e-05, + "loss": 1.0445, + "step": 27538 + }, + { + "epoch": 1.77, + "learning_rate": 1.4111938856566266e-05, + "loss": 0.9684, + "step": 27539 + }, + { + "epoch": 1.77, + "learning_rate": 1.4111724167543315e-05, + "loss": 1.0427, + "step": 27540 + }, + { + "epoch": 1.77, + "learning_rate": 1.4111509478520364e-05, + "loss": 0.9394, + "step": 27541 + }, + { + "epoch": 1.77, + "learning_rate": 1.4111294789497415e-05, + "loss": 0.9832, + "step": 27542 + }, + { + "epoch": 1.77, + "learning_rate": 1.4111080100474464e-05, + "loss": 0.9654, + "step": 27543 + }, + { + "epoch": 1.77, + "learning_rate": 1.4110865411451512e-05, + "loss": 1.0362, + "step": 27544 + }, + { + "epoch": 1.77, + "learning_rate": 1.4110650722428563e-05, + "loss": 0.9328, + "step": 27545 + }, + { + "epoch": 1.77, + "learning_rate": 1.4110436033405612e-05, + "loss": 0.9714, + "step": 27546 + }, + { + "epoch": 1.77, + "learning_rate": 1.4110221344382661e-05, + "loss": 1.0477, + "step": 27547 + }, + { + "epoch": 1.77, + "learning_rate": 1.4110006655359713e-05, + "loss": 0.9895, + "step": 27548 + }, + { + "epoch": 1.77, + "learning_rate": 1.4109791966336764e-05, + "loss": 0.9692, + "step": 27549 + }, + { + "epoch": 1.77, + "learning_rate": 1.4109577277313813e-05, + "loss": 0.9026, + "step": 27550 + }, + { + "epoch": 1.77, + "learning_rate": 1.4109362588290862e-05, + "loss": 0.9066, + "step": 27551 + }, + { + "epoch": 1.77, + "learning_rate": 1.4109147899267913e-05, + "loss": 0.9644, + "step": 27552 + }, + { + "epoch": 1.77, + "learning_rate": 1.4108933210244962e-05, + "loss": 0.965, + "step": 27553 + }, + { + "epoch": 1.77, + "learning_rate": 1.410871852122201e-05, + "loss": 0.9339, + "step": 27554 + }, + { + "epoch": 1.77, + "learning_rate": 1.4108503832199061e-05, + "loss": 0.9135, + "step": 27555 + }, + { + "epoch": 1.77, + "learning_rate": 1.410828914317611e-05, + "loss": 1.0043, + "step": 27556 + }, + { + "epoch": 1.77, + "learning_rate": 1.410807445415316e-05, + "loss": 0.9772, + "step": 27557 + }, + { + "epoch": 1.77, + "learning_rate": 1.410785976513021e-05, + "loss": 0.9779, + "step": 27558 + }, + { + "epoch": 1.77, + "learning_rate": 1.4107645076107259e-05, + "loss": 0.9853, + "step": 27559 + }, + { + "epoch": 1.77, + "learning_rate": 1.410743038708431e-05, + "loss": 0.8847, + "step": 27560 + }, + { + "epoch": 1.77, + "learning_rate": 1.4107215698061358e-05, + "loss": 0.9791, + "step": 27561 + }, + { + "epoch": 1.77, + "learning_rate": 1.4107001009038407e-05, + "loss": 0.9454, + "step": 27562 + }, + { + "epoch": 1.77, + "learning_rate": 1.410678632001546e-05, + "loss": 1.0325, + "step": 27563 + }, + { + "epoch": 1.77, + "learning_rate": 1.4106571630992509e-05, + "loss": 0.9921, + "step": 27564 + }, + { + "epoch": 1.77, + "learning_rate": 1.410635694196956e-05, + "loss": 0.97, + "step": 27565 + }, + { + "epoch": 1.77, + "learning_rate": 1.4106142252946608e-05, + "loss": 0.9968, + "step": 27566 + }, + { + "epoch": 1.77, + "learning_rate": 1.4105927563923659e-05, + "loss": 0.9793, + "step": 27567 + }, + { + "epoch": 1.77, + "learning_rate": 1.4105712874900708e-05, + "loss": 1.0469, + "step": 27568 + }, + { + "epoch": 1.77, + "learning_rate": 1.4105498185877757e-05, + "loss": 0.9604, + "step": 27569 + }, + { + "epoch": 1.77, + "learning_rate": 1.4105283496854808e-05, + "loss": 0.9518, + "step": 27570 + }, + { + "epoch": 1.77, + "learning_rate": 1.4105068807831857e-05, + "loss": 0.9602, + "step": 27571 + }, + { + "epoch": 1.77, + "learning_rate": 1.4104854118808906e-05, + "loss": 1.0043, + "step": 27572 + }, + { + "epoch": 1.77, + "learning_rate": 1.4104639429785956e-05, + "loss": 0.9641, + "step": 27573 + }, + { + "epoch": 1.77, + "learning_rate": 1.4104424740763005e-05, + "loss": 0.9732, + "step": 27574 + }, + { + "epoch": 1.77, + "learning_rate": 1.4104210051740054e-05, + "loss": 1.0336, + "step": 27575 + }, + { + "epoch": 1.77, + "learning_rate": 1.4103995362717105e-05, + "loss": 0.9578, + "step": 27576 + }, + { + "epoch": 1.77, + "learning_rate": 1.4103780673694157e-05, + "loss": 1.084, + "step": 27577 + }, + { + "epoch": 1.77, + "learning_rate": 1.4103565984671206e-05, + "loss": 0.9624, + "step": 27578 + }, + { + "epoch": 1.77, + "learning_rate": 1.4103351295648255e-05, + "loss": 1.0063, + "step": 27579 + }, + { + "epoch": 1.77, + "learning_rate": 1.4103136606625306e-05, + "loss": 0.9587, + "step": 27580 + }, + { + "epoch": 1.77, + "learning_rate": 1.4102921917602355e-05, + "loss": 0.9349, + "step": 27581 + }, + { + "epoch": 1.77, + "learning_rate": 1.4102707228579404e-05, + "loss": 0.9344, + "step": 27582 + }, + { + "epoch": 1.77, + "learning_rate": 1.4102492539556454e-05, + "loss": 0.9687, + "step": 27583 + }, + { + "epoch": 1.77, + "learning_rate": 1.4102277850533503e-05, + "loss": 0.884, + "step": 27584 + }, + { + "epoch": 1.77, + "learning_rate": 1.4102063161510552e-05, + "loss": 0.9012, + "step": 27585 + }, + { + "epoch": 1.77, + "learning_rate": 1.4101848472487603e-05, + "loss": 0.9468, + "step": 27586 + }, + { + "epoch": 1.77, + "learning_rate": 1.4101633783464652e-05, + "loss": 0.8949, + "step": 27587 + }, + { + "epoch": 1.77, + "learning_rate": 1.4101419094441702e-05, + "loss": 1.0244, + "step": 27588 + }, + { + "epoch": 1.78, + "learning_rate": 1.4101204405418751e-05, + "loss": 0.9313, + "step": 27589 + }, + { + "epoch": 1.78, + "learning_rate": 1.41009897163958e-05, + "loss": 0.9395, + "step": 27590 + }, + { + "epoch": 1.78, + "learning_rate": 1.4100775027372851e-05, + "loss": 0.961, + "step": 27591 + }, + { + "epoch": 1.78, + "learning_rate": 1.4100560338349902e-05, + "loss": 0.9286, + "step": 27592 + }, + { + "epoch": 1.78, + "learning_rate": 1.4100345649326952e-05, + "loss": 0.9663, + "step": 27593 + }, + { + "epoch": 1.78, + "learning_rate": 1.4100130960304001e-05, + "loss": 1.0143, + "step": 27594 + }, + { + "epoch": 1.78, + "learning_rate": 1.409991627128105e-05, + "loss": 0.9855, + "step": 27595 + }, + { + "epoch": 1.78, + "learning_rate": 1.4099701582258101e-05, + "loss": 0.9159, + "step": 27596 + }, + { + "epoch": 1.78, + "learning_rate": 1.409948689323515e-05, + "loss": 1.0103, + "step": 27597 + }, + { + "epoch": 1.78, + "learning_rate": 1.40992722042122e-05, + "loss": 1.0152, + "step": 27598 + }, + { + "epoch": 1.78, + "learning_rate": 1.409905751518925e-05, + "loss": 0.9393, + "step": 27599 + }, + { + "epoch": 1.78, + "learning_rate": 1.4098842826166299e-05, + "loss": 0.9157, + "step": 27600 + }, + { + "epoch": 1.78, + "learning_rate": 1.409862813714335e-05, + "loss": 0.9627, + "step": 27601 + }, + { + "epoch": 1.78, + "learning_rate": 1.4098413448120398e-05, + "loss": 1.0307, + "step": 27602 + }, + { + "epoch": 1.78, + "learning_rate": 1.4098198759097447e-05, + "loss": 0.9787, + "step": 27603 + }, + { + "epoch": 1.78, + "learning_rate": 1.4097984070074498e-05, + "loss": 0.9298, + "step": 27604 + }, + { + "epoch": 1.78, + "learning_rate": 1.4097769381051547e-05, + "loss": 0.9571, + "step": 27605 + }, + { + "epoch": 1.78, + "learning_rate": 1.4097554692028599e-05, + "loss": 0.9974, + "step": 27606 + }, + { + "epoch": 1.78, + "learning_rate": 1.4097340003005648e-05, + "loss": 0.9306, + "step": 27607 + }, + { + "epoch": 1.78, + "learning_rate": 1.4097125313982699e-05, + "loss": 1.05, + "step": 27608 + }, + { + "epoch": 1.78, + "learning_rate": 1.4096910624959748e-05, + "loss": 0.958, + "step": 27609 + }, + { + "epoch": 1.78, + "learning_rate": 1.4096695935936797e-05, + "loss": 1.0422, + "step": 27610 + }, + { + "epoch": 1.78, + "learning_rate": 1.4096481246913847e-05, + "loss": 0.973, + "step": 27611 + }, + { + "epoch": 1.78, + "learning_rate": 1.4096266557890896e-05, + "loss": 0.9814, + "step": 27612 + }, + { + "epoch": 1.78, + "learning_rate": 1.4096051868867945e-05, + "loss": 0.922, + "step": 27613 + }, + { + "epoch": 1.78, + "learning_rate": 1.4095837179844996e-05, + "loss": 0.9431, + "step": 27614 + }, + { + "epoch": 1.78, + "learning_rate": 1.4095622490822045e-05, + "loss": 1.0209, + "step": 27615 + }, + { + "epoch": 1.78, + "learning_rate": 1.4095407801799096e-05, + "loss": 1.0517, + "step": 27616 + }, + { + "epoch": 1.78, + "learning_rate": 1.4095193112776145e-05, + "loss": 0.9375, + "step": 27617 + }, + { + "epoch": 1.78, + "learning_rate": 1.4094978423753193e-05, + "loss": 0.9282, + "step": 27618 + }, + { + "epoch": 1.78, + "learning_rate": 1.4094763734730244e-05, + "loss": 0.9855, + "step": 27619 + }, + { + "epoch": 1.78, + "learning_rate": 1.4094549045707293e-05, + "loss": 0.9924, + "step": 27620 + }, + { + "epoch": 1.78, + "learning_rate": 1.4094334356684345e-05, + "loss": 1.039, + "step": 27621 + }, + { + "epoch": 1.78, + "learning_rate": 1.4094119667661394e-05, + "loss": 0.963, + "step": 27622 + }, + { + "epoch": 1.78, + "learning_rate": 1.4093904978638443e-05, + "loss": 0.925, + "step": 27623 + }, + { + "epoch": 1.78, + "learning_rate": 1.4093690289615494e-05, + "loss": 0.9659, + "step": 27624 + }, + { + "epoch": 1.78, + "learning_rate": 1.4093475600592543e-05, + "loss": 0.9719, + "step": 27625 + }, + { + "epoch": 1.78, + "learning_rate": 1.4093260911569594e-05, + "loss": 0.9762, + "step": 27626 + }, + { + "epoch": 1.78, + "learning_rate": 1.4093046222546643e-05, + "loss": 1.0318, + "step": 27627 + }, + { + "epoch": 1.78, + "learning_rate": 1.4092831533523692e-05, + "loss": 1.0213, + "step": 27628 + }, + { + "epoch": 1.78, + "learning_rate": 1.4092616844500742e-05, + "loss": 1.0376, + "step": 27629 + }, + { + "epoch": 1.78, + "learning_rate": 1.4092402155477791e-05, + "loss": 1.0387, + "step": 27630 + }, + { + "epoch": 1.78, + "learning_rate": 1.409218746645484e-05, + "loss": 1.0185, + "step": 27631 + }, + { + "epoch": 1.78, + "learning_rate": 1.409197277743189e-05, + "loss": 1.0135, + "step": 27632 + }, + { + "epoch": 1.78, + "learning_rate": 1.409175808840894e-05, + "loss": 1.0005, + "step": 27633 + }, + { + "epoch": 1.78, + "learning_rate": 1.4091543399385989e-05, + "loss": 0.9065, + "step": 27634 + }, + { + "epoch": 1.78, + "learning_rate": 1.4091328710363041e-05, + "loss": 0.9397, + "step": 27635 + }, + { + "epoch": 1.78, + "learning_rate": 1.4091114021340092e-05, + "loss": 1.0132, + "step": 27636 + }, + { + "epoch": 1.78, + "learning_rate": 1.409089933231714e-05, + "loss": 0.8828, + "step": 27637 + }, + { + "epoch": 1.78, + "learning_rate": 1.409068464329419e-05, + "loss": 0.9868, + "step": 27638 + }, + { + "epoch": 1.78, + "learning_rate": 1.409046995427124e-05, + "loss": 0.9967, + "step": 27639 + }, + { + "epoch": 1.78, + "learning_rate": 1.409025526524829e-05, + "loss": 0.9908, + "step": 27640 + }, + { + "epoch": 1.78, + "learning_rate": 1.4090040576225338e-05, + "loss": 0.9783, + "step": 27641 + }, + { + "epoch": 1.78, + "learning_rate": 1.4089825887202389e-05, + "loss": 0.9699, + "step": 27642 + }, + { + "epoch": 1.78, + "learning_rate": 1.4089611198179438e-05, + "loss": 0.9668, + "step": 27643 + }, + { + "epoch": 1.78, + "learning_rate": 1.4089396509156487e-05, + "loss": 1.009, + "step": 27644 + }, + { + "epoch": 1.78, + "learning_rate": 1.4089181820133538e-05, + "loss": 0.9465, + "step": 27645 + }, + { + "epoch": 1.78, + "learning_rate": 1.4088967131110587e-05, + "loss": 0.9751, + "step": 27646 + }, + { + "epoch": 1.78, + "learning_rate": 1.4088752442087637e-05, + "loss": 0.9452, + "step": 27647 + }, + { + "epoch": 1.78, + "learning_rate": 1.4088537753064686e-05, + "loss": 0.9374, + "step": 27648 + }, + { + "epoch": 1.78, + "learning_rate": 1.4088323064041735e-05, + "loss": 0.964, + "step": 27649 + }, + { + "epoch": 1.78, + "learning_rate": 1.4088108375018787e-05, + "loss": 1.001, + "step": 27650 + }, + { + "epoch": 1.78, + "learning_rate": 1.4087893685995836e-05, + "loss": 0.9346, + "step": 27651 + }, + { + "epoch": 1.78, + "learning_rate": 1.4087678996972887e-05, + "loss": 0.9939, + "step": 27652 + }, + { + "epoch": 1.78, + "learning_rate": 1.4087464307949936e-05, + "loss": 0.9843, + "step": 27653 + }, + { + "epoch": 1.78, + "learning_rate": 1.4087249618926985e-05, + "loss": 1.0052, + "step": 27654 + }, + { + "epoch": 1.78, + "learning_rate": 1.4087034929904036e-05, + "loss": 0.9625, + "step": 27655 + }, + { + "epoch": 1.78, + "learning_rate": 1.4086820240881085e-05, + "loss": 1.0136, + "step": 27656 + }, + { + "epoch": 1.78, + "learning_rate": 1.4086605551858135e-05, + "loss": 0.9659, + "step": 27657 + }, + { + "epoch": 1.78, + "learning_rate": 1.4086390862835184e-05, + "loss": 0.9702, + "step": 27658 + }, + { + "epoch": 1.78, + "learning_rate": 1.4086176173812233e-05, + "loss": 0.9661, + "step": 27659 + }, + { + "epoch": 1.78, + "learning_rate": 1.4085961484789284e-05, + "loss": 0.9525, + "step": 27660 + }, + { + "epoch": 1.78, + "learning_rate": 1.4085746795766333e-05, + "loss": 0.9538, + "step": 27661 + }, + { + "epoch": 1.78, + "learning_rate": 1.4085532106743382e-05, + "loss": 1.0416, + "step": 27662 + }, + { + "epoch": 1.78, + "learning_rate": 1.4085317417720432e-05, + "loss": 0.9297, + "step": 27663 + }, + { + "epoch": 1.78, + "learning_rate": 1.4085102728697481e-05, + "loss": 0.9779, + "step": 27664 + }, + { + "epoch": 1.78, + "learning_rate": 1.4084888039674534e-05, + "loss": 0.9369, + "step": 27665 + }, + { + "epoch": 1.78, + "learning_rate": 1.4084673350651583e-05, + "loss": 0.9629, + "step": 27666 + }, + { + "epoch": 1.78, + "learning_rate": 1.4084458661628633e-05, + "loss": 0.9696, + "step": 27667 + }, + { + "epoch": 1.78, + "learning_rate": 1.4084243972605682e-05, + "loss": 0.9259, + "step": 27668 + }, + { + "epoch": 1.78, + "learning_rate": 1.4084029283582731e-05, + "loss": 1.0051, + "step": 27669 + }, + { + "epoch": 1.78, + "learning_rate": 1.4083814594559782e-05, + "loss": 1.0151, + "step": 27670 + }, + { + "epoch": 1.78, + "learning_rate": 1.4083599905536831e-05, + "loss": 1.0193, + "step": 27671 + }, + { + "epoch": 1.78, + "learning_rate": 1.408338521651388e-05, + "loss": 0.9754, + "step": 27672 + }, + { + "epoch": 1.78, + "learning_rate": 1.408317052749093e-05, + "loss": 0.9634, + "step": 27673 + }, + { + "epoch": 1.78, + "learning_rate": 1.408295583846798e-05, + "loss": 0.9148, + "step": 27674 + }, + { + "epoch": 1.78, + "learning_rate": 1.408274114944503e-05, + "loss": 1.0232, + "step": 27675 + }, + { + "epoch": 1.78, + "learning_rate": 1.408252646042208e-05, + "loss": 0.9438, + "step": 27676 + }, + { + "epoch": 1.78, + "learning_rate": 1.4082311771399128e-05, + "loss": 0.9669, + "step": 27677 + }, + { + "epoch": 1.78, + "learning_rate": 1.4082097082376179e-05, + "loss": 0.9868, + "step": 27678 + }, + { + "epoch": 1.78, + "learning_rate": 1.408188239335323e-05, + "loss": 0.9656, + "step": 27679 + }, + { + "epoch": 1.78, + "learning_rate": 1.408166770433028e-05, + "loss": 1.009, + "step": 27680 + }, + { + "epoch": 1.78, + "learning_rate": 1.4081453015307329e-05, + "loss": 0.933, + "step": 27681 + }, + { + "epoch": 1.78, + "learning_rate": 1.4081238326284378e-05, + "loss": 0.9821, + "step": 27682 + }, + { + "epoch": 1.78, + "learning_rate": 1.4081023637261429e-05, + "loss": 0.9781, + "step": 27683 + }, + { + "epoch": 1.78, + "learning_rate": 1.4080808948238478e-05, + "loss": 1.0349, + "step": 27684 + }, + { + "epoch": 1.78, + "learning_rate": 1.4080594259215528e-05, + "loss": 0.9573, + "step": 27685 + }, + { + "epoch": 1.78, + "learning_rate": 1.4080379570192577e-05, + "loss": 0.9216, + "step": 27686 + }, + { + "epoch": 1.78, + "learning_rate": 1.4080164881169626e-05, + "loss": 1.0112, + "step": 27687 + }, + { + "epoch": 1.78, + "learning_rate": 1.4079950192146677e-05, + "loss": 0.9627, + "step": 27688 + }, + { + "epoch": 1.78, + "learning_rate": 1.4079735503123726e-05, + "loss": 0.9024, + "step": 27689 + }, + { + "epoch": 1.78, + "learning_rate": 1.4079520814100775e-05, + "loss": 0.9341, + "step": 27690 + }, + { + "epoch": 1.78, + "learning_rate": 1.4079306125077826e-05, + "loss": 0.972, + "step": 27691 + }, + { + "epoch": 1.78, + "learning_rate": 1.4079091436054874e-05, + "loss": 0.9799, + "step": 27692 + }, + { + "epoch": 1.78, + "learning_rate": 1.4078876747031925e-05, + "loss": 1.0055, + "step": 27693 + }, + { + "epoch": 1.78, + "learning_rate": 1.4078662058008976e-05, + "loss": 0.9259, + "step": 27694 + }, + { + "epoch": 1.78, + "learning_rate": 1.4078447368986026e-05, + "loss": 0.9586, + "step": 27695 + }, + { + "epoch": 1.78, + "learning_rate": 1.4078232679963075e-05, + "loss": 0.982, + "step": 27696 + }, + { + "epoch": 1.78, + "learning_rate": 1.4078017990940124e-05, + "loss": 0.9411, + "step": 27697 + }, + { + "epoch": 1.78, + "learning_rate": 1.4077803301917175e-05, + "loss": 0.904, + "step": 27698 + }, + { + "epoch": 1.78, + "learning_rate": 1.4077588612894224e-05, + "loss": 1.0074, + "step": 27699 + }, + { + "epoch": 1.78, + "learning_rate": 1.4077373923871273e-05, + "loss": 1.0246, + "step": 27700 + }, + { + "epoch": 1.78, + "learning_rate": 1.4077159234848324e-05, + "loss": 1.0057, + "step": 27701 + }, + { + "epoch": 1.78, + "learning_rate": 1.4076944545825373e-05, + "loss": 0.9858, + "step": 27702 + }, + { + "epoch": 1.78, + "learning_rate": 1.4076729856802423e-05, + "loss": 0.9477, + "step": 27703 + }, + { + "epoch": 1.78, + "learning_rate": 1.4076515167779472e-05, + "loss": 0.9841, + "step": 27704 + }, + { + "epoch": 1.78, + "learning_rate": 1.4076300478756521e-05, + "loss": 0.9437, + "step": 27705 + }, + { + "epoch": 1.78, + "learning_rate": 1.4076085789733572e-05, + "loss": 0.9644, + "step": 27706 + }, + { + "epoch": 1.78, + "learning_rate": 1.407587110071062e-05, + "loss": 0.977, + "step": 27707 + }, + { + "epoch": 1.78, + "learning_rate": 1.4075656411687673e-05, + "loss": 1.0012, + "step": 27708 + }, + { + "epoch": 1.78, + "learning_rate": 1.4075441722664722e-05, + "loss": 0.9444, + "step": 27709 + }, + { + "epoch": 1.78, + "learning_rate": 1.4075227033641771e-05, + "loss": 0.9904, + "step": 27710 + }, + { + "epoch": 1.78, + "learning_rate": 1.4075012344618822e-05, + "loss": 0.9647, + "step": 27711 + }, + { + "epoch": 1.78, + "learning_rate": 1.407479765559587e-05, + "loss": 0.8963, + "step": 27712 + }, + { + "epoch": 1.78, + "learning_rate": 1.4074582966572921e-05, + "loss": 0.963, + "step": 27713 + }, + { + "epoch": 1.78, + "learning_rate": 1.407436827754997e-05, + "loss": 0.9248, + "step": 27714 + }, + { + "epoch": 1.78, + "learning_rate": 1.407415358852702e-05, + "loss": 1.0527, + "step": 27715 + }, + { + "epoch": 1.78, + "learning_rate": 1.407393889950407e-05, + "loss": 0.914, + "step": 27716 + }, + { + "epoch": 1.78, + "learning_rate": 1.4073724210481119e-05, + "loss": 0.9251, + "step": 27717 + }, + { + "epoch": 1.78, + "learning_rate": 1.4073509521458168e-05, + "loss": 0.997, + "step": 27718 + }, + { + "epoch": 1.78, + "learning_rate": 1.4073294832435219e-05, + "loss": 0.9699, + "step": 27719 + }, + { + "epoch": 1.78, + "learning_rate": 1.4073080143412268e-05, + "loss": 0.9777, + "step": 27720 + }, + { + "epoch": 1.78, + "learning_rate": 1.4072865454389316e-05, + "loss": 0.9719, + "step": 27721 + }, + { + "epoch": 1.78, + "learning_rate": 1.4072650765366367e-05, + "loss": 1.0137, + "step": 27722 + }, + { + "epoch": 1.78, + "learning_rate": 1.407243607634342e-05, + "loss": 0.9983, + "step": 27723 + }, + { + "epoch": 1.78, + "learning_rate": 1.4072221387320468e-05, + "loss": 0.9461, + "step": 27724 + }, + { + "epoch": 1.78, + "learning_rate": 1.4072006698297517e-05, + "loss": 0.9896, + "step": 27725 + }, + { + "epoch": 1.78, + "learning_rate": 1.4071792009274568e-05, + "loss": 1.0286, + "step": 27726 + }, + { + "epoch": 1.78, + "learning_rate": 1.4071577320251617e-05, + "loss": 1.0241, + "step": 27727 + }, + { + "epoch": 1.78, + "learning_rate": 1.4071362631228666e-05, + "loss": 1.0208, + "step": 27728 + }, + { + "epoch": 1.78, + "learning_rate": 1.4071147942205717e-05, + "loss": 1.0275, + "step": 27729 + }, + { + "epoch": 1.78, + "learning_rate": 1.4070933253182766e-05, + "loss": 0.9923, + "step": 27730 + }, + { + "epoch": 1.78, + "learning_rate": 1.4070718564159815e-05, + "loss": 0.9263, + "step": 27731 + }, + { + "epoch": 1.78, + "learning_rate": 1.4070503875136865e-05, + "loss": 0.9794, + "step": 27732 + }, + { + "epoch": 1.78, + "learning_rate": 1.4070289186113914e-05, + "loss": 0.9255, + "step": 27733 + }, + { + "epoch": 1.78, + "learning_rate": 1.4070074497090965e-05, + "loss": 0.9546, + "step": 27734 + }, + { + "epoch": 1.78, + "learning_rate": 1.4069859808068014e-05, + "loss": 0.9736, + "step": 27735 + }, + { + "epoch": 1.78, + "learning_rate": 1.4069645119045063e-05, + "loss": 0.963, + "step": 27736 + }, + { + "epoch": 1.78, + "learning_rate": 1.4069430430022115e-05, + "loss": 0.9451, + "step": 27737 + }, + { + "epoch": 1.78, + "learning_rate": 1.4069215740999164e-05, + "loss": 0.9299, + "step": 27738 + }, + { + "epoch": 1.78, + "learning_rate": 1.4069001051976215e-05, + "loss": 0.9329, + "step": 27739 + }, + { + "epoch": 1.78, + "learning_rate": 1.4068786362953264e-05, + "loss": 1.0009, + "step": 27740 + }, + { + "epoch": 1.78, + "learning_rate": 1.4068571673930313e-05, + "loss": 1.0027, + "step": 27741 + }, + { + "epoch": 1.78, + "learning_rate": 1.4068356984907363e-05, + "loss": 0.9742, + "step": 27742 + }, + { + "epoch": 1.78, + "learning_rate": 1.4068142295884412e-05, + "loss": 0.9474, + "step": 27743 + }, + { + "epoch": 1.78, + "learning_rate": 1.4067927606861463e-05, + "loss": 0.9774, + "step": 27744 + }, + { + "epoch": 1.79, + "learning_rate": 1.4067712917838512e-05, + "loss": 1.0193, + "step": 27745 + }, + { + "epoch": 1.79, + "learning_rate": 1.4067498228815561e-05, + "loss": 0.9757, + "step": 27746 + }, + { + "epoch": 1.79, + "learning_rate": 1.4067283539792612e-05, + "loss": 1.0112, + "step": 27747 + }, + { + "epoch": 1.79, + "learning_rate": 1.406706885076966e-05, + "loss": 0.9813, + "step": 27748 + }, + { + "epoch": 1.79, + "learning_rate": 1.406685416174671e-05, + "loss": 1.019, + "step": 27749 + }, + { + "epoch": 1.79, + "learning_rate": 1.406663947272376e-05, + "loss": 0.9581, + "step": 27750 + }, + { + "epoch": 1.79, + "learning_rate": 1.4066424783700809e-05, + "loss": 0.9251, + "step": 27751 + }, + { + "epoch": 1.79, + "learning_rate": 1.4066210094677861e-05, + "loss": 1.0154, + "step": 27752 + }, + { + "epoch": 1.79, + "learning_rate": 1.406599540565491e-05, + "loss": 0.9487, + "step": 27753 + }, + { + "epoch": 1.79, + "learning_rate": 1.4065780716631961e-05, + "loss": 1.0572, + "step": 27754 + }, + { + "epoch": 1.79, + "learning_rate": 1.406556602760901e-05, + "loss": 0.9966, + "step": 27755 + }, + { + "epoch": 1.79, + "learning_rate": 1.4065351338586059e-05, + "loss": 0.9993, + "step": 27756 + }, + { + "epoch": 1.79, + "learning_rate": 1.406513664956311e-05, + "loss": 0.9138, + "step": 27757 + }, + { + "epoch": 1.79, + "learning_rate": 1.4064921960540159e-05, + "loss": 1.0024, + "step": 27758 + }, + { + "epoch": 1.79, + "learning_rate": 1.4064707271517208e-05, + "loss": 0.9445, + "step": 27759 + }, + { + "epoch": 1.79, + "learning_rate": 1.4064492582494258e-05, + "loss": 0.9744, + "step": 27760 + }, + { + "epoch": 1.79, + "learning_rate": 1.4064277893471307e-05, + "loss": 0.9044, + "step": 27761 + }, + { + "epoch": 1.79, + "learning_rate": 1.4064063204448358e-05, + "loss": 0.8994, + "step": 27762 + }, + { + "epoch": 1.79, + "learning_rate": 1.4063848515425407e-05, + "loss": 0.9834, + "step": 27763 + }, + { + "epoch": 1.79, + "learning_rate": 1.4063633826402456e-05, + "loss": 0.9272, + "step": 27764 + }, + { + "epoch": 1.79, + "learning_rate": 1.4063419137379507e-05, + "loss": 1.0146, + "step": 27765 + }, + { + "epoch": 1.79, + "learning_rate": 1.4063204448356557e-05, + "loss": 1.0282, + "step": 27766 + }, + { + "epoch": 1.79, + "learning_rate": 1.4062989759333608e-05, + "loss": 1.011, + "step": 27767 + }, + { + "epoch": 1.79, + "learning_rate": 1.4062775070310657e-05, + "loss": 1.0337, + "step": 27768 + }, + { + "epoch": 1.79, + "learning_rate": 1.4062560381287706e-05, + "loss": 0.9402, + "step": 27769 + }, + { + "epoch": 1.79, + "learning_rate": 1.4062345692264756e-05, + "loss": 1.0678, + "step": 27770 + }, + { + "epoch": 1.79, + "learning_rate": 1.4062131003241805e-05, + "loss": 0.9079, + "step": 27771 + }, + { + "epoch": 1.79, + "learning_rate": 1.4061916314218856e-05, + "loss": 1.0059, + "step": 27772 + }, + { + "epoch": 1.79, + "learning_rate": 1.4061701625195905e-05, + "loss": 0.9794, + "step": 27773 + }, + { + "epoch": 1.79, + "learning_rate": 1.4061486936172954e-05, + "loss": 0.9567, + "step": 27774 + }, + { + "epoch": 1.79, + "learning_rate": 1.4061272247150005e-05, + "loss": 1.0439, + "step": 27775 + }, + { + "epoch": 1.79, + "learning_rate": 1.4061057558127054e-05, + "loss": 0.9768, + "step": 27776 + }, + { + "epoch": 1.79, + "learning_rate": 1.4060842869104103e-05, + "loss": 0.9809, + "step": 27777 + }, + { + "epoch": 1.79, + "learning_rate": 1.4060628180081153e-05, + "loss": 0.9246, + "step": 27778 + }, + { + "epoch": 1.79, + "learning_rate": 1.4060413491058202e-05, + "loss": 0.9491, + "step": 27779 + }, + { + "epoch": 1.79, + "learning_rate": 1.4060198802035253e-05, + "loss": 0.9466, + "step": 27780 + }, + { + "epoch": 1.79, + "learning_rate": 1.4059984113012304e-05, + "loss": 0.9734, + "step": 27781 + }, + { + "epoch": 1.79, + "learning_rate": 1.4059769423989354e-05, + "loss": 1.039, + "step": 27782 + }, + { + "epoch": 1.79, + "learning_rate": 1.4059554734966403e-05, + "loss": 1.0187, + "step": 27783 + }, + { + "epoch": 1.79, + "learning_rate": 1.4059340045943452e-05, + "loss": 0.9331, + "step": 27784 + }, + { + "epoch": 1.79, + "learning_rate": 1.4059125356920503e-05, + "loss": 0.9262, + "step": 27785 + }, + { + "epoch": 1.79, + "learning_rate": 1.4058910667897552e-05, + "loss": 0.9946, + "step": 27786 + }, + { + "epoch": 1.79, + "learning_rate": 1.40586959788746e-05, + "loss": 0.9762, + "step": 27787 + }, + { + "epoch": 1.79, + "learning_rate": 1.4058481289851651e-05, + "loss": 0.9439, + "step": 27788 + }, + { + "epoch": 1.79, + "learning_rate": 1.40582666008287e-05, + "loss": 1.0273, + "step": 27789 + }, + { + "epoch": 1.79, + "learning_rate": 1.4058051911805751e-05, + "loss": 0.9839, + "step": 27790 + }, + { + "epoch": 1.79, + "learning_rate": 1.40578372227828e-05, + "loss": 0.9639, + "step": 27791 + }, + { + "epoch": 1.79, + "learning_rate": 1.4057622533759849e-05, + "loss": 1.025, + "step": 27792 + }, + { + "epoch": 1.79, + "learning_rate": 1.40574078447369e-05, + "loss": 0.9967, + "step": 27793 + }, + { + "epoch": 1.79, + "learning_rate": 1.4057193155713949e-05, + "loss": 0.9545, + "step": 27794 + }, + { + "epoch": 1.79, + "learning_rate": 1.4056978466691001e-05, + "loss": 0.9957, + "step": 27795 + }, + { + "epoch": 1.79, + "learning_rate": 1.405676377766805e-05, + "loss": 0.9216, + "step": 27796 + }, + { + "epoch": 1.79, + "learning_rate": 1.4056549088645099e-05, + "loss": 0.9574, + "step": 27797 + }, + { + "epoch": 1.79, + "learning_rate": 1.405633439962215e-05, + "loss": 0.962, + "step": 27798 + }, + { + "epoch": 1.79, + "learning_rate": 1.4056119710599198e-05, + "loss": 0.9323, + "step": 27799 + }, + { + "epoch": 1.79, + "learning_rate": 1.4055905021576249e-05, + "loss": 0.9431, + "step": 27800 + }, + { + "epoch": 1.79, + "learning_rate": 1.4055690332553298e-05, + "loss": 0.9134, + "step": 27801 + }, + { + "epoch": 1.79, + "learning_rate": 1.4055475643530347e-05, + "loss": 0.9766, + "step": 27802 + }, + { + "epoch": 1.79, + "learning_rate": 1.4055260954507398e-05, + "loss": 0.9696, + "step": 27803 + }, + { + "epoch": 1.79, + "learning_rate": 1.4055046265484447e-05, + "loss": 0.9818, + "step": 27804 + }, + { + "epoch": 1.79, + "learning_rate": 1.4054831576461496e-05, + "loss": 0.9741, + "step": 27805 + }, + { + "epoch": 1.79, + "learning_rate": 1.4054616887438546e-05, + "loss": 1.0434, + "step": 27806 + }, + { + "epoch": 1.79, + "learning_rate": 1.4054402198415595e-05, + "loss": 0.9712, + "step": 27807 + }, + { + "epoch": 1.79, + "learning_rate": 1.4054187509392644e-05, + "loss": 0.892, + "step": 27808 + }, + { + "epoch": 1.79, + "learning_rate": 1.4053972820369695e-05, + "loss": 0.9222, + "step": 27809 + }, + { + "epoch": 1.79, + "learning_rate": 1.4053758131346747e-05, + "loss": 0.9883, + "step": 27810 + }, + { + "epoch": 1.79, + "learning_rate": 1.4053543442323796e-05, + "loss": 0.9612, + "step": 27811 + }, + { + "epoch": 1.79, + "learning_rate": 1.4053328753300845e-05, + "loss": 0.9986, + "step": 27812 + }, + { + "epoch": 1.79, + "learning_rate": 1.4053114064277896e-05, + "loss": 0.9325, + "step": 27813 + }, + { + "epoch": 1.79, + "learning_rate": 1.4052899375254945e-05, + "loss": 0.9334, + "step": 27814 + }, + { + "epoch": 1.79, + "learning_rate": 1.4052684686231994e-05, + "loss": 0.9115, + "step": 27815 + }, + { + "epoch": 1.79, + "learning_rate": 1.4052469997209044e-05, + "loss": 0.9669, + "step": 27816 + }, + { + "epoch": 1.79, + "learning_rate": 1.4052255308186093e-05, + "loss": 0.9944, + "step": 27817 + }, + { + "epoch": 1.79, + "learning_rate": 1.4052040619163142e-05, + "loss": 0.9401, + "step": 27818 + }, + { + "epoch": 1.79, + "learning_rate": 1.4051825930140193e-05, + "loss": 0.979, + "step": 27819 + }, + { + "epoch": 1.79, + "learning_rate": 1.4051611241117242e-05, + "loss": 0.9863, + "step": 27820 + }, + { + "epoch": 1.79, + "learning_rate": 1.4051396552094293e-05, + "loss": 0.9813, + "step": 27821 + }, + { + "epoch": 1.79, + "learning_rate": 1.4051181863071342e-05, + "loss": 0.9774, + "step": 27822 + }, + { + "epoch": 1.79, + "learning_rate": 1.405096717404839e-05, + "loss": 0.9553, + "step": 27823 + }, + { + "epoch": 1.79, + "learning_rate": 1.4050752485025443e-05, + "loss": 1.0116, + "step": 27824 + }, + { + "epoch": 1.79, + "learning_rate": 1.4050537796002492e-05, + "loss": 0.947, + "step": 27825 + }, + { + "epoch": 1.79, + "learning_rate": 1.4050323106979542e-05, + "loss": 0.9686, + "step": 27826 + }, + { + "epoch": 1.79, + "learning_rate": 1.4050108417956591e-05, + "loss": 0.9442, + "step": 27827 + }, + { + "epoch": 1.79, + "learning_rate": 1.404989372893364e-05, + "loss": 0.9308, + "step": 27828 + }, + { + "epoch": 1.79, + "learning_rate": 1.4049679039910691e-05, + "loss": 0.9452, + "step": 27829 + }, + { + "epoch": 1.79, + "learning_rate": 1.404946435088774e-05, + "loss": 0.9645, + "step": 27830 + }, + { + "epoch": 1.79, + "learning_rate": 1.404924966186479e-05, + "loss": 0.85, + "step": 27831 + }, + { + "epoch": 1.79, + "learning_rate": 1.404903497284184e-05, + "loss": 0.9478, + "step": 27832 + }, + { + "epoch": 1.79, + "learning_rate": 1.4048820283818889e-05, + "loss": 0.942, + "step": 27833 + }, + { + "epoch": 1.79, + "learning_rate": 1.404860559479594e-05, + "loss": 1.0241, + "step": 27834 + }, + { + "epoch": 1.79, + "learning_rate": 1.4048390905772988e-05, + "loss": 0.9917, + "step": 27835 + }, + { + "epoch": 1.79, + "learning_rate": 1.4048176216750037e-05, + "loss": 0.9931, + "step": 27836 + }, + { + "epoch": 1.79, + "learning_rate": 1.4047961527727088e-05, + "loss": 0.9084, + "step": 27837 + }, + { + "epoch": 1.79, + "learning_rate": 1.4047746838704137e-05, + "loss": 0.9753, + "step": 27838 + }, + { + "epoch": 1.79, + "learning_rate": 1.404753214968119e-05, + "loss": 0.9497, + "step": 27839 + }, + { + "epoch": 1.79, + "learning_rate": 1.4047317460658238e-05, + "loss": 0.9289, + "step": 27840 + }, + { + "epoch": 1.79, + "learning_rate": 1.4047102771635289e-05, + "loss": 0.9478, + "step": 27841 + }, + { + "epoch": 1.79, + "learning_rate": 1.4046888082612338e-05, + "loss": 0.9561, + "step": 27842 + }, + { + "epoch": 1.79, + "learning_rate": 1.4046673393589387e-05, + "loss": 0.9644, + "step": 27843 + }, + { + "epoch": 1.79, + "learning_rate": 1.4046458704566437e-05, + "loss": 0.9544, + "step": 27844 + }, + { + "epoch": 1.79, + "learning_rate": 1.4046244015543486e-05, + "loss": 1.0211, + "step": 27845 + }, + { + "epoch": 1.79, + "learning_rate": 1.4046029326520535e-05, + "loss": 0.9345, + "step": 27846 + }, + { + "epoch": 1.79, + "learning_rate": 1.4045814637497586e-05, + "loss": 0.9276, + "step": 27847 + }, + { + "epoch": 1.79, + "learning_rate": 1.4045599948474635e-05, + "loss": 0.8728, + "step": 27848 + }, + { + "epoch": 1.79, + "learning_rate": 1.4045385259451686e-05, + "loss": 0.9952, + "step": 27849 + }, + { + "epoch": 1.79, + "learning_rate": 1.4045170570428735e-05, + "loss": 0.9491, + "step": 27850 + }, + { + "epoch": 1.79, + "learning_rate": 1.4044955881405784e-05, + "loss": 0.9953, + "step": 27851 + }, + { + "epoch": 1.79, + "learning_rate": 1.4044741192382834e-05, + "loss": 1.0445, + "step": 27852 + }, + { + "epoch": 1.79, + "learning_rate": 1.4044526503359885e-05, + "loss": 0.955, + "step": 27853 + }, + { + "epoch": 1.79, + "learning_rate": 1.4044311814336936e-05, + "loss": 0.9665, + "step": 27854 + }, + { + "epoch": 1.79, + "learning_rate": 1.4044097125313985e-05, + "loss": 0.9689, + "step": 27855 + }, + { + "epoch": 1.79, + "learning_rate": 1.4043882436291033e-05, + "loss": 0.9808, + "step": 27856 + }, + { + "epoch": 1.79, + "learning_rate": 1.4043667747268084e-05, + "loss": 0.9039, + "step": 27857 + }, + { + "epoch": 1.79, + "learning_rate": 1.4043453058245133e-05, + "loss": 0.97, + "step": 27858 + }, + { + "epoch": 1.79, + "learning_rate": 1.4043238369222184e-05, + "loss": 1.0175, + "step": 27859 + }, + { + "epoch": 1.79, + "learning_rate": 1.4043023680199233e-05, + "loss": 1.0351, + "step": 27860 + }, + { + "epoch": 1.79, + "learning_rate": 1.4042808991176282e-05, + "loss": 0.9228, + "step": 27861 + }, + { + "epoch": 1.79, + "learning_rate": 1.4042594302153332e-05, + "loss": 0.9313, + "step": 27862 + }, + { + "epoch": 1.79, + "learning_rate": 1.4042379613130381e-05, + "loss": 0.9603, + "step": 27863 + }, + { + "epoch": 1.79, + "learning_rate": 1.404216492410743e-05, + "loss": 0.9125, + "step": 27864 + }, + { + "epoch": 1.79, + "learning_rate": 1.4041950235084481e-05, + "loss": 0.9652, + "step": 27865 + }, + { + "epoch": 1.79, + "learning_rate": 1.404173554606153e-05, + "loss": 0.9603, + "step": 27866 + }, + { + "epoch": 1.79, + "learning_rate": 1.4041520857038579e-05, + "loss": 1.0395, + "step": 27867 + }, + { + "epoch": 1.79, + "learning_rate": 1.4041306168015631e-05, + "loss": 0.9307, + "step": 27868 + }, + { + "epoch": 1.79, + "learning_rate": 1.4041091478992682e-05, + "loss": 1.0077, + "step": 27869 + }, + { + "epoch": 1.79, + "learning_rate": 1.404087678996973e-05, + "loss": 1.0153, + "step": 27870 + }, + { + "epoch": 1.79, + "learning_rate": 1.404066210094678e-05, + "loss": 0.9633, + "step": 27871 + }, + { + "epoch": 1.79, + "learning_rate": 1.404044741192383e-05, + "loss": 0.9371, + "step": 27872 + }, + { + "epoch": 1.79, + "learning_rate": 1.404023272290088e-05, + "loss": 0.9446, + "step": 27873 + }, + { + "epoch": 1.79, + "learning_rate": 1.4040018033877928e-05, + "loss": 1.039, + "step": 27874 + }, + { + "epoch": 1.79, + "learning_rate": 1.4039803344854979e-05, + "loss": 0.9706, + "step": 27875 + }, + { + "epoch": 1.79, + "learning_rate": 1.4039588655832028e-05, + "loss": 1.0541, + "step": 27876 + }, + { + "epoch": 1.79, + "learning_rate": 1.4039373966809077e-05, + "loss": 1.005, + "step": 27877 + }, + { + "epoch": 1.79, + "learning_rate": 1.4039159277786128e-05, + "loss": 0.9155, + "step": 27878 + }, + { + "epoch": 1.79, + "learning_rate": 1.4038944588763177e-05, + "loss": 0.9708, + "step": 27879 + }, + { + "epoch": 1.79, + "learning_rate": 1.4038729899740227e-05, + "loss": 0.9941, + "step": 27880 + }, + { + "epoch": 1.79, + "learning_rate": 1.4038515210717276e-05, + "loss": 0.9382, + "step": 27881 + }, + { + "epoch": 1.79, + "learning_rate": 1.4038300521694329e-05, + "loss": 1.0103, + "step": 27882 + }, + { + "epoch": 1.79, + "learning_rate": 1.4038085832671378e-05, + "loss": 0.9813, + "step": 27883 + }, + { + "epoch": 1.79, + "learning_rate": 1.4037871143648427e-05, + "loss": 0.9609, + "step": 27884 + }, + { + "epoch": 1.79, + "learning_rate": 1.4037656454625477e-05, + "loss": 0.9955, + "step": 27885 + }, + { + "epoch": 1.79, + "learning_rate": 1.4037441765602526e-05, + "loss": 0.973, + "step": 27886 + }, + { + "epoch": 1.79, + "learning_rate": 1.4037227076579577e-05, + "loss": 0.8985, + "step": 27887 + }, + { + "epoch": 1.79, + "learning_rate": 1.4037012387556626e-05, + "loss": 1.0526, + "step": 27888 + }, + { + "epoch": 1.79, + "learning_rate": 1.4036797698533675e-05, + "loss": 0.9953, + "step": 27889 + }, + { + "epoch": 1.79, + "learning_rate": 1.4036583009510725e-05, + "loss": 0.9682, + "step": 27890 + }, + { + "epoch": 1.79, + "learning_rate": 1.4036368320487774e-05, + "loss": 0.9706, + "step": 27891 + }, + { + "epoch": 1.79, + "learning_rate": 1.4036153631464823e-05, + "loss": 1.0512, + "step": 27892 + }, + { + "epoch": 1.79, + "learning_rate": 1.4035938942441874e-05, + "loss": 0.9719, + "step": 27893 + }, + { + "epoch": 1.79, + "learning_rate": 1.4035724253418923e-05, + "loss": 0.9392, + "step": 27894 + }, + { + "epoch": 1.79, + "learning_rate": 1.4035509564395972e-05, + "loss": 0.9871, + "step": 27895 + }, + { + "epoch": 1.79, + "learning_rate": 1.4035294875373023e-05, + "loss": 0.9557, + "step": 27896 + }, + { + "epoch": 1.79, + "learning_rate": 1.4035080186350075e-05, + "loss": 0.9263, + "step": 27897 + }, + { + "epoch": 1.79, + "learning_rate": 1.4034865497327124e-05, + "loss": 0.9699, + "step": 27898 + }, + { + "epoch": 1.79, + "learning_rate": 1.4034650808304173e-05, + "loss": 0.9148, + "step": 27899 + }, + { + "epoch": 1.8, + "learning_rate": 1.4034436119281224e-05, + "loss": 0.985, + "step": 27900 + }, + { + "epoch": 1.8, + "learning_rate": 1.4034221430258272e-05, + "loss": 0.9851, + "step": 27901 + }, + { + "epoch": 1.8, + "learning_rate": 1.4034006741235321e-05, + "loss": 0.9576, + "step": 27902 + }, + { + "epoch": 1.8, + "learning_rate": 1.4033792052212372e-05, + "loss": 0.9942, + "step": 27903 + }, + { + "epoch": 1.8, + "learning_rate": 1.4033577363189421e-05, + "loss": 0.9959, + "step": 27904 + }, + { + "epoch": 1.8, + "learning_rate": 1.403336267416647e-05, + "loss": 0.9455, + "step": 27905 + }, + { + "epoch": 1.8, + "learning_rate": 1.403314798514352e-05, + "loss": 0.9682, + "step": 27906 + }, + { + "epoch": 1.8, + "learning_rate": 1.403293329612057e-05, + "loss": 0.9545, + "step": 27907 + }, + { + "epoch": 1.8, + "learning_rate": 1.403271860709762e-05, + "loss": 1.0511, + "step": 27908 + }, + { + "epoch": 1.8, + "learning_rate": 1.403250391807467e-05, + "loss": 1.0167, + "step": 27909 + }, + { + "epoch": 1.8, + "learning_rate": 1.4032289229051718e-05, + "loss": 0.9892, + "step": 27910 + }, + { + "epoch": 1.8, + "learning_rate": 1.403207454002877e-05, + "loss": 0.9052, + "step": 27911 + }, + { + "epoch": 1.8, + "learning_rate": 1.403185985100582e-05, + "loss": 1.014, + "step": 27912 + }, + { + "epoch": 1.8, + "learning_rate": 1.403164516198287e-05, + "loss": 1.0105, + "step": 27913 + }, + { + "epoch": 1.8, + "learning_rate": 1.403143047295992e-05, + "loss": 0.9476, + "step": 27914 + }, + { + "epoch": 1.8, + "learning_rate": 1.4031215783936968e-05, + "loss": 1.0162, + "step": 27915 + }, + { + "epoch": 1.8, + "learning_rate": 1.4031001094914019e-05, + "loss": 0.9585, + "step": 27916 + }, + { + "epoch": 1.8, + "learning_rate": 1.4030786405891068e-05, + "loss": 0.9772, + "step": 27917 + }, + { + "epoch": 1.8, + "learning_rate": 1.4030571716868118e-05, + "loss": 0.9674, + "step": 27918 + }, + { + "epoch": 1.8, + "learning_rate": 1.4030357027845167e-05, + "loss": 0.9063, + "step": 27919 + }, + { + "epoch": 1.8, + "learning_rate": 1.4030142338822216e-05, + "loss": 1.0126, + "step": 27920 + }, + { + "epoch": 1.8, + "learning_rate": 1.4029927649799267e-05, + "loss": 0.9952, + "step": 27921 + }, + { + "epoch": 1.8, + "learning_rate": 1.4029712960776316e-05, + "loss": 1.0869, + "step": 27922 + }, + { + "epoch": 1.8, + "learning_rate": 1.4029498271753365e-05, + "loss": 0.9098, + "step": 27923 + }, + { + "epoch": 1.8, + "learning_rate": 1.4029283582730416e-05, + "loss": 0.9881, + "step": 27924 + }, + { + "epoch": 1.8, + "learning_rate": 1.4029068893707465e-05, + "loss": 1.0125, + "step": 27925 + }, + { + "epoch": 1.8, + "learning_rate": 1.4028854204684517e-05, + "loss": 0.9231, + "step": 27926 + }, + { + "epoch": 1.8, + "learning_rate": 1.4028639515661566e-05, + "loss": 1.0545, + "step": 27927 + }, + { + "epoch": 1.8, + "learning_rate": 1.4028424826638617e-05, + "loss": 1.0129, + "step": 27928 + }, + { + "epoch": 1.8, + "learning_rate": 1.4028210137615666e-05, + "loss": 0.9185, + "step": 27929 + }, + { + "epoch": 1.8, + "learning_rate": 1.4027995448592714e-05, + "loss": 0.9716, + "step": 27930 + }, + { + "epoch": 1.8, + "learning_rate": 1.4027780759569765e-05, + "loss": 0.976, + "step": 27931 + }, + { + "epoch": 1.8, + "learning_rate": 1.4027566070546814e-05, + "loss": 1.0013, + "step": 27932 + }, + { + "epoch": 1.8, + "learning_rate": 1.4027351381523863e-05, + "loss": 0.9371, + "step": 27933 + }, + { + "epoch": 1.8, + "learning_rate": 1.4027136692500914e-05, + "loss": 0.9581, + "step": 27934 + }, + { + "epoch": 1.8, + "learning_rate": 1.4026922003477963e-05, + "loss": 0.9908, + "step": 27935 + }, + { + "epoch": 1.8, + "learning_rate": 1.4026707314455013e-05, + "loss": 0.9366, + "step": 27936 + }, + { + "epoch": 1.8, + "learning_rate": 1.4026492625432062e-05, + "loss": 0.9713, + "step": 27937 + }, + { + "epoch": 1.8, + "learning_rate": 1.4026277936409111e-05, + "loss": 1.0156, + "step": 27938 + }, + { + "epoch": 1.8, + "learning_rate": 1.4026063247386162e-05, + "loss": 0.9564, + "step": 27939 + }, + { + "epoch": 1.8, + "learning_rate": 1.4025848558363213e-05, + "loss": 0.9727, + "step": 27940 + }, + { + "epoch": 1.8, + "learning_rate": 1.4025633869340263e-05, + "loss": 1.0402, + "step": 27941 + }, + { + "epoch": 1.8, + "learning_rate": 1.4025419180317312e-05, + "loss": 0.9532, + "step": 27942 + }, + { + "epoch": 1.8, + "learning_rate": 1.4025204491294361e-05, + "loss": 1.0551, + "step": 27943 + }, + { + "epoch": 1.8, + "learning_rate": 1.4024989802271412e-05, + "loss": 0.9464, + "step": 27944 + }, + { + "epoch": 1.8, + "learning_rate": 1.402477511324846e-05, + "loss": 0.9225, + "step": 27945 + }, + { + "epoch": 1.8, + "learning_rate": 1.4024560424225511e-05, + "loss": 1.0311, + "step": 27946 + }, + { + "epoch": 1.8, + "learning_rate": 1.402434573520256e-05, + "loss": 1.0236, + "step": 27947 + }, + { + "epoch": 1.8, + "learning_rate": 1.402413104617961e-05, + "loss": 0.9223, + "step": 27948 + }, + { + "epoch": 1.8, + "learning_rate": 1.402391635715666e-05, + "loss": 0.9643, + "step": 27949 + }, + { + "epoch": 1.8, + "learning_rate": 1.4023701668133709e-05, + "loss": 0.9173, + "step": 27950 + }, + { + "epoch": 1.8, + "learning_rate": 1.4023486979110758e-05, + "loss": 0.9373, + "step": 27951 + }, + { + "epoch": 1.8, + "learning_rate": 1.4023272290087809e-05, + "loss": 0.9898, + "step": 27952 + }, + { + "epoch": 1.8, + "learning_rate": 1.4023057601064858e-05, + "loss": 0.9582, + "step": 27953 + }, + { + "epoch": 1.8, + "learning_rate": 1.4022842912041907e-05, + "loss": 1.0084, + "step": 27954 + }, + { + "epoch": 1.8, + "learning_rate": 1.4022628223018959e-05, + "loss": 1.0143, + "step": 27955 + }, + { + "epoch": 1.8, + "learning_rate": 1.402241353399601e-05, + "loss": 0.9601, + "step": 27956 + }, + { + "epoch": 1.8, + "learning_rate": 1.4022198844973059e-05, + "loss": 0.9921, + "step": 27957 + }, + { + "epoch": 1.8, + "learning_rate": 1.4021984155950108e-05, + "loss": 0.9422, + "step": 27958 + }, + { + "epoch": 1.8, + "learning_rate": 1.4021769466927158e-05, + "loss": 0.9376, + "step": 27959 + }, + { + "epoch": 1.8, + "learning_rate": 1.4021554777904207e-05, + "loss": 0.9956, + "step": 27960 + }, + { + "epoch": 1.8, + "learning_rate": 1.4021340088881256e-05, + "loss": 1.0594, + "step": 27961 + }, + { + "epoch": 1.8, + "learning_rate": 1.4021125399858307e-05, + "loss": 0.9717, + "step": 27962 + }, + { + "epoch": 1.8, + "learning_rate": 1.4020910710835356e-05, + "loss": 1.0311, + "step": 27963 + }, + { + "epoch": 1.8, + "learning_rate": 1.4020696021812405e-05, + "loss": 0.9762, + "step": 27964 + }, + { + "epoch": 1.8, + "learning_rate": 1.4020481332789455e-05, + "loss": 0.9809, + "step": 27965 + }, + { + "epoch": 1.8, + "learning_rate": 1.4020266643766504e-05, + "loss": 0.9041, + "step": 27966 + }, + { + "epoch": 1.8, + "learning_rate": 1.4020051954743555e-05, + "loss": 0.9511, + "step": 27967 + }, + { + "epoch": 1.8, + "learning_rate": 1.4019837265720604e-05, + "loss": 1.0091, + "step": 27968 + }, + { + "epoch": 1.8, + "learning_rate": 1.4019622576697656e-05, + "loss": 1.022, + "step": 27969 + }, + { + "epoch": 1.8, + "learning_rate": 1.4019407887674705e-05, + "loss": 0.9983, + "step": 27970 + }, + { + "epoch": 1.8, + "learning_rate": 1.4019193198651754e-05, + "loss": 0.9945, + "step": 27971 + }, + { + "epoch": 1.8, + "learning_rate": 1.4018978509628805e-05, + "loss": 0.8891, + "step": 27972 + }, + { + "epoch": 1.8, + "learning_rate": 1.4018763820605854e-05, + "loss": 1.0174, + "step": 27973 + }, + { + "epoch": 1.8, + "learning_rate": 1.4018549131582903e-05, + "loss": 0.9763, + "step": 27974 + }, + { + "epoch": 1.8, + "learning_rate": 1.4018334442559953e-05, + "loss": 0.9311, + "step": 27975 + }, + { + "epoch": 1.8, + "learning_rate": 1.4018119753537002e-05, + "loss": 0.9275, + "step": 27976 + }, + { + "epoch": 1.8, + "learning_rate": 1.4017905064514053e-05, + "loss": 0.9622, + "step": 27977 + }, + { + "epoch": 1.8, + "learning_rate": 1.4017690375491102e-05, + "loss": 0.917, + "step": 27978 + }, + { + "epoch": 1.8, + "learning_rate": 1.4017475686468151e-05, + "loss": 0.9475, + "step": 27979 + }, + { + "epoch": 1.8, + "learning_rate": 1.4017260997445202e-05, + "loss": 0.9379, + "step": 27980 + }, + { + "epoch": 1.8, + "learning_rate": 1.401704630842225e-05, + "loss": 1.0128, + "step": 27981 + }, + { + "epoch": 1.8, + "learning_rate": 1.40168316193993e-05, + "loss": 0.955, + "step": 27982 + }, + { + "epoch": 1.8, + "learning_rate": 1.401661693037635e-05, + "loss": 1.0055, + "step": 27983 + }, + { + "epoch": 1.8, + "learning_rate": 1.4016402241353401e-05, + "loss": 0.9611, + "step": 27984 + }, + { + "epoch": 1.8, + "learning_rate": 1.4016187552330452e-05, + "loss": 0.9919, + "step": 27985 + }, + { + "epoch": 1.8, + "learning_rate": 1.40159728633075e-05, + "loss": 0.938, + "step": 27986 + }, + { + "epoch": 1.8, + "learning_rate": 1.4015758174284551e-05, + "loss": 1.0824, + "step": 27987 + }, + { + "epoch": 1.8, + "learning_rate": 1.40155434852616e-05, + "loss": 0.9954, + "step": 27988 + }, + { + "epoch": 1.8, + "learning_rate": 1.4015328796238649e-05, + "loss": 0.9197, + "step": 27989 + }, + { + "epoch": 1.8, + "learning_rate": 1.40151141072157e-05, + "loss": 1.0698, + "step": 27990 + }, + { + "epoch": 1.8, + "learning_rate": 1.4014899418192749e-05, + "loss": 0.9838, + "step": 27991 + }, + { + "epoch": 1.8, + "learning_rate": 1.4014684729169798e-05, + "loss": 0.9504, + "step": 27992 + }, + { + "epoch": 1.8, + "learning_rate": 1.4014470040146848e-05, + "loss": 0.995, + "step": 27993 + }, + { + "epoch": 1.8, + "learning_rate": 1.4014255351123897e-05, + "loss": 0.9019, + "step": 27994 + }, + { + "epoch": 1.8, + "learning_rate": 1.4014040662100948e-05, + "loss": 1.0536, + "step": 27995 + }, + { + "epoch": 1.8, + "learning_rate": 1.4013825973077997e-05, + "loss": 0.9778, + "step": 27996 + }, + { + "epoch": 1.8, + "learning_rate": 1.4013611284055046e-05, + "loss": 0.9589, + "step": 27997 + }, + { + "epoch": 1.8, + "learning_rate": 1.4013396595032098e-05, + "loss": 0.9473, + "step": 27998 + }, + { + "epoch": 1.8, + "learning_rate": 1.4013181906009147e-05, + "loss": 0.9645, + "step": 27999 + }, + { + "epoch": 1.8, + "learning_rate": 1.4012967216986198e-05, + "loss": 1.0043, + "step": 28000 + }, + { + "epoch": 1.8, + "eval_laion/cider": 0.5509485801196183, + "eval_laion/meteor": 0.3856018114303702, + "eval_laion/pred_num_words": 35.76171875, + "eval_laion/sacrebleu": 15.659789347135883, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.043494462966919, + "eval_runtime": 97.655, + "eval_samples_per_second": 5.11, + "eval_steps_per_second": 0.328, + "step": 28000 + }, + { + "epoch": 1.8, + "learning_rate": 1.4012752527963247e-05, + "loss": 0.9772, + "step": 28001 + }, + { + "epoch": 1.8, + "learning_rate": 1.4012537838940296e-05, + "loss": 1.005, + "step": 28002 + }, + { + "epoch": 1.8, + "learning_rate": 1.4012323149917347e-05, + "loss": 1.0725, + "step": 28003 + }, + { + "epoch": 1.8, + "learning_rate": 1.4012108460894395e-05, + "loss": 0.9724, + "step": 28004 + }, + { + "epoch": 1.8, + "learning_rate": 1.4011893771871446e-05, + "loss": 0.9818, + "step": 28005 + }, + { + "epoch": 1.8, + "learning_rate": 1.4011679082848495e-05, + "loss": 0.9589, + "step": 28006 + }, + { + "epoch": 1.8, + "learning_rate": 1.4011464393825544e-05, + "loss": 0.8944, + "step": 28007 + }, + { + "epoch": 1.8, + "learning_rate": 1.4011249704802595e-05, + "loss": 0.9669, + "step": 28008 + }, + { + "epoch": 1.8, + "learning_rate": 1.4011035015779644e-05, + "loss": 0.9463, + "step": 28009 + }, + { + "epoch": 1.8, + "learning_rate": 1.4010820326756693e-05, + "loss": 0.9352, + "step": 28010 + }, + { + "epoch": 1.8, + "learning_rate": 1.4010605637733743e-05, + "loss": 0.9768, + "step": 28011 + }, + { + "epoch": 1.8, + "learning_rate": 1.4010390948710792e-05, + "loss": 0.9117, + "step": 28012 + }, + { + "epoch": 1.8, + "learning_rate": 1.4010176259687845e-05, + "loss": 1.0376, + "step": 28013 + }, + { + "epoch": 1.8, + "learning_rate": 1.4009961570664894e-05, + "loss": 0.9982, + "step": 28014 + }, + { + "epoch": 1.8, + "learning_rate": 1.4009746881641944e-05, + "loss": 1.0057, + "step": 28015 + }, + { + "epoch": 1.8, + "learning_rate": 1.4009532192618993e-05, + "loss": 1.0195, + "step": 28016 + }, + { + "epoch": 1.8, + "learning_rate": 1.4009317503596042e-05, + "loss": 0.9811, + "step": 28017 + }, + { + "epoch": 1.8, + "learning_rate": 1.4009102814573093e-05, + "loss": 0.9916, + "step": 28018 + }, + { + "epoch": 1.8, + "learning_rate": 1.4008888125550142e-05, + "loss": 0.9563, + "step": 28019 + }, + { + "epoch": 1.8, + "learning_rate": 1.400867343652719e-05, + "loss": 0.9141, + "step": 28020 + }, + { + "epoch": 1.8, + "learning_rate": 1.4008458747504241e-05, + "loss": 0.9482, + "step": 28021 + }, + { + "epoch": 1.8, + "learning_rate": 1.400824405848129e-05, + "loss": 0.9721, + "step": 28022 + }, + { + "epoch": 1.8, + "learning_rate": 1.4008029369458341e-05, + "loss": 0.9658, + "step": 28023 + }, + { + "epoch": 1.8, + "learning_rate": 1.400781468043539e-05, + "loss": 0.9594, + "step": 28024 + }, + { + "epoch": 1.8, + "learning_rate": 1.4007599991412439e-05, + "loss": 0.9939, + "step": 28025 + }, + { + "epoch": 1.8, + "learning_rate": 1.400738530238949e-05, + "loss": 1.0154, + "step": 28026 + }, + { + "epoch": 1.8, + "learning_rate": 1.4007170613366539e-05, + "loss": 1.0065, + "step": 28027 + }, + { + "epoch": 1.8, + "learning_rate": 1.4006955924343591e-05, + "loss": 1.0069, + "step": 28028 + }, + { + "epoch": 1.8, + "learning_rate": 1.400674123532064e-05, + "loss": 0.9843, + "step": 28029 + }, + { + "epoch": 1.8, + "learning_rate": 1.4006526546297689e-05, + "loss": 1.0388, + "step": 28030 + }, + { + "epoch": 1.8, + "learning_rate": 1.400631185727474e-05, + "loss": 0.8957, + "step": 28031 + }, + { + "epoch": 1.8, + "learning_rate": 1.4006097168251789e-05, + "loss": 1.0197, + "step": 28032 + }, + { + "epoch": 1.8, + "learning_rate": 1.400588247922884e-05, + "loss": 0.9877, + "step": 28033 + }, + { + "epoch": 1.8, + "learning_rate": 1.4005667790205888e-05, + "loss": 0.9061, + "step": 28034 + }, + { + "epoch": 1.8, + "learning_rate": 1.4005453101182937e-05, + "loss": 0.9326, + "step": 28035 + }, + { + "epoch": 1.8, + "learning_rate": 1.4005238412159988e-05, + "loss": 0.928, + "step": 28036 + }, + { + "epoch": 1.8, + "learning_rate": 1.4005023723137037e-05, + "loss": 0.9949, + "step": 28037 + }, + { + "epoch": 1.8, + "learning_rate": 1.4004809034114086e-05, + "loss": 0.9524, + "step": 28038 + }, + { + "epoch": 1.8, + "learning_rate": 1.4004594345091136e-05, + "loss": 1.0073, + "step": 28039 + }, + { + "epoch": 1.8, + "learning_rate": 1.4004379656068185e-05, + "loss": 0.9448, + "step": 28040 + }, + { + "epoch": 1.8, + "learning_rate": 1.4004164967045234e-05, + "loss": 0.8969, + "step": 28041 + }, + { + "epoch": 1.8, + "learning_rate": 1.4003950278022287e-05, + "loss": 1.03, + "step": 28042 + }, + { + "epoch": 1.8, + "learning_rate": 1.4003735588999337e-05, + "loss": 1.0781, + "step": 28043 + }, + { + "epoch": 1.8, + "learning_rate": 1.4003520899976386e-05, + "loss": 1.0358, + "step": 28044 + }, + { + "epoch": 1.8, + "learning_rate": 1.4003306210953435e-05, + "loss": 1.0165, + "step": 28045 + }, + { + "epoch": 1.8, + "learning_rate": 1.4003091521930486e-05, + "loss": 0.943, + "step": 28046 + }, + { + "epoch": 1.8, + "learning_rate": 1.4002876832907535e-05, + "loss": 0.9442, + "step": 28047 + }, + { + "epoch": 1.8, + "learning_rate": 1.4002662143884584e-05, + "loss": 0.9538, + "step": 28048 + }, + { + "epoch": 1.8, + "learning_rate": 1.4002447454861634e-05, + "loss": 0.9846, + "step": 28049 + }, + { + "epoch": 1.8, + "learning_rate": 1.4002232765838683e-05, + "loss": 1.016, + "step": 28050 + }, + { + "epoch": 1.8, + "learning_rate": 1.4002018076815732e-05, + "loss": 0.9769, + "step": 28051 + }, + { + "epoch": 1.8, + "learning_rate": 1.4001803387792783e-05, + "loss": 1.0088, + "step": 28052 + }, + { + "epoch": 1.8, + "learning_rate": 1.4001588698769832e-05, + "loss": 0.981, + "step": 28053 + }, + { + "epoch": 1.8, + "learning_rate": 1.4001374009746883e-05, + "loss": 0.9855, + "step": 28054 + }, + { + "epoch": 1.8, + "learning_rate": 1.4001159320723932e-05, + "loss": 0.9507, + "step": 28055 + }, + { + "epoch": 1.81, + "learning_rate": 1.400094463170098e-05, + "loss": 0.9884, + "step": 28056 + }, + { + "epoch": 1.81, + "learning_rate": 1.4000729942678033e-05, + "loss": 0.9178, + "step": 28057 + }, + { + "epoch": 1.81, + "learning_rate": 1.4000515253655082e-05, + "loss": 0.9687, + "step": 28058 + }, + { + "epoch": 1.81, + "learning_rate": 1.4000300564632133e-05, + "loss": 0.9383, + "step": 28059 + }, + { + "epoch": 1.81, + "learning_rate": 1.4000085875609182e-05, + "loss": 0.991, + "step": 28060 + }, + { + "epoch": 1.81, + "learning_rate": 1.399987118658623e-05, + "loss": 0.9004, + "step": 28061 + }, + { + "epoch": 1.81, + "learning_rate": 1.3999656497563281e-05, + "loss": 0.9866, + "step": 28062 + }, + { + "epoch": 1.81, + "learning_rate": 1.399944180854033e-05, + "loss": 0.9518, + "step": 28063 + }, + { + "epoch": 1.81, + "learning_rate": 1.399922711951738e-05, + "loss": 0.9737, + "step": 28064 + }, + { + "epoch": 1.81, + "learning_rate": 1.399901243049443e-05, + "loss": 0.9862, + "step": 28065 + }, + { + "epoch": 1.81, + "learning_rate": 1.3998797741471479e-05, + "loss": 1.0339, + "step": 28066 + }, + { + "epoch": 1.81, + "learning_rate": 1.399858305244853e-05, + "loss": 0.9904, + "step": 28067 + }, + { + "epoch": 1.81, + "learning_rate": 1.3998368363425578e-05, + "loss": 0.9531, + "step": 28068 + }, + { + "epoch": 1.81, + "learning_rate": 1.3998153674402627e-05, + "loss": 0.9758, + "step": 28069 + }, + { + "epoch": 1.81, + "learning_rate": 1.3997938985379678e-05, + "loss": 0.9738, + "step": 28070 + }, + { + "epoch": 1.81, + "learning_rate": 1.3997724296356729e-05, + "loss": 0.947, + "step": 28071 + }, + { + "epoch": 1.81, + "learning_rate": 1.399750960733378e-05, + "loss": 1.0079, + "step": 28072 + }, + { + "epoch": 1.81, + "learning_rate": 1.3997294918310828e-05, + "loss": 0.9732, + "step": 28073 + }, + { + "epoch": 1.81, + "learning_rate": 1.3997080229287879e-05, + "loss": 1.0133, + "step": 28074 + }, + { + "epoch": 1.81, + "learning_rate": 1.3996865540264928e-05, + "loss": 0.9892, + "step": 28075 + }, + { + "epoch": 1.81, + "learning_rate": 1.3996650851241977e-05, + "loss": 1.0061, + "step": 28076 + }, + { + "epoch": 1.81, + "learning_rate": 1.3996436162219028e-05, + "loss": 1.0284, + "step": 28077 + }, + { + "epoch": 1.81, + "learning_rate": 1.3996221473196076e-05, + "loss": 0.9411, + "step": 28078 + }, + { + "epoch": 1.81, + "learning_rate": 1.3996006784173125e-05, + "loss": 0.968, + "step": 28079 + }, + { + "epoch": 1.81, + "learning_rate": 1.3995792095150176e-05, + "loss": 1.013, + "step": 28080 + }, + { + "epoch": 1.81, + "learning_rate": 1.3995577406127225e-05, + "loss": 0.9307, + "step": 28081 + }, + { + "epoch": 1.81, + "learning_rate": 1.3995362717104276e-05, + "loss": 0.9413, + "step": 28082 + }, + { + "epoch": 1.81, + "learning_rate": 1.3995148028081325e-05, + "loss": 0.9634, + "step": 28083 + }, + { + "epoch": 1.81, + "learning_rate": 1.3994933339058374e-05, + "loss": 0.9424, + "step": 28084 + }, + { + "epoch": 1.81, + "learning_rate": 1.3994718650035424e-05, + "loss": 1.0016, + "step": 28085 + }, + { + "epoch": 1.81, + "learning_rate": 1.3994503961012475e-05, + "loss": 1.0273, + "step": 28086 + }, + { + "epoch": 1.81, + "learning_rate": 1.3994289271989526e-05, + "loss": 1.0263, + "step": 28087 + }, + { + "epoch": 1.81, + "learning_rate": 1.3994074582966575e-05, + "loss": 0.9454, + "step": 28088 + }, + { + "epoch": 1.81, + "learning_rate": 1.3993859893943624e-05, + "loss": 0.9886, + "step": 28089 + }, + { + "epoch": 1.81, + "learning_rate": 1.3993645204920674e-05, + "loss": 0.9756, + "step": 28090 + }, + { + "epoch": 1.81, + "learning_rate": 1.3993430515897723e-05, + "loss": 0.9201, + "step": 28091 + }, + { + "epoch": 1.81, + "learning_rate": 1.3993215826874774e-05, + "loss": 1.0797, + "step": 28092 + }, + { + "epoch": 1.81, + "learning_rate": 1.3993001137851823e-05, + "loss": 0.9929, + "step": 28093 + }, + { + "epoch": 1.81, + "learning_rate": 1.3992786448828872e-05, + "loss": 0.9985, + "step": 28094 + }, + { + "epoch": 1.81, + "learning_rate": 1.3992571759805922e-05, + "loss": 1.0519, + "step": 28095 + }, + { + "epoch": 1.81, + "learning_rate": 1.3992357070782971e-05, + "loss": 0.9069, + "step": 28096 + }, + { + "epoch": 1.81, + "learning_rate": 1.399214238176002e-05, + "loss": 0.8739, + "step": 28097 + }, + { + "epoch": 1.81, + "learning_rate": 1.3991927692737071e-05, + "loss": 0.9258, + "step": 28098 + }, + { + "epoch": 1.81, + "learning_rate": 1.399171300371412e-05, + "loss": 1.0261, + "step": 28099 + }, + { + "epoch": 1.81, + "learning_rate": 1.3991498314691172e-05, + "loss": 0.9543, + "step": 28100 + }, + { + "epoch": 1.81, + "learning_rate": 1.3991283625668221e-05, + "loss": 0.9922, + "step": 28101 + }, + { + "epoch": 1.81, + "learning_rate": 1.3991068936645272e-05, + "loss": 0.9469, + "step": 28102 + }, + { + "epoch": 1.81, + "learning_rate": 1.3990854247622321e-05, + "loss": 1.036, + "step": 28103 + }, + { + "epoch": 1.81, + "learning_rate": 1.399063955859937e-05, + "loss": 0.9426, + "step": 28104 + }, + { + "epoch": 1.81, + "learning_rate": 1.399042486957642e-05, + "loss": 1.0552, + "step": 28105 + }, + { + "epoch": 1.81, + "learning_rate": 1.399021018055347e-05, + "loss": 0.9449, + "step": 28106 + }, + { + "epoch": 1.81, + "learning_rate": 1.3989995491530518e-05, + "loss": 0.9654, + "step": 28107 + }, + { + "epoch": 1.81, + "learning_rate": 1.3989780802507569e-05, + "loss": 0.9611, + "step": 28108 + }, + { + "epoch": 1.81, + "learning_rate": 1.3989566113484618e-05, + "loss": 0.9883, + "step": 28109 + }, + { + "epoch": 1.81, + "learning_rate": 1.3989351424461669e-05, + "loss": 1.0018, + "step": 28110 + }, + { + "epoch": 1.81, + "learning_rate": 1.3989136735438718e-05, + "loss": 0.9348, + "step": 28111 + }, + { + "epoch": 1.81, + "learning_rate": 1.3988922046415767e-05, + "loss": 0.9787, + "step": 28112 + }, + { + "epoch": 1.81, + "learning_rate": 1.3988707357392817e-05, + "loss": 1.0376, + "step": 28113 + }, + { + "epoch": 1.81, + "learning_rate": 1.3988492668369866e-05, + "loss": 1.0047, + "step": 28114 + }, + { + "epoch": 1.81, + "learning_rate": 1.3988277979346919e-05, + "loss": 1.0148, + "step": 28115 + }, + { + "epoch": 1.81, + "learning_rate": 1.3988063290323968e-05, + "loss": 0.9754, + "step": 28116 + }, + { + "epoch": 1.81, + "learning_rate": 1.3987848601301017e-05, + "loss": 0.9679, + "step": 28117 + }, + { + "epoch": 1.81, + "learning_rate": 1.3987633912278067e-05, + "loss": 0.9359, + "step": 28118 + }, + { + "epoch": 1.81, + "learning_rate": 1.3987419223255116e-05, + "loss": 0.8999, + "step": 28119 + }, + { + "epoch": 1.81, + "learning_rate": 1.3987204534232167e-05, + "loss": 0.9085, + "step": 28120 + }, + { + "epoch": 1.81, + "learning_rate": 1.3986989845209216e-05, + "loss": 0.9478, + "step": 28121 + }, + { + "epoch": 1.81, + "learning_rate": 1.3986775156186265e-05, + "loss": 0.9395, + "step": 28122 + }, + { + "epoch": 1.81, + "learning_rate": 1.3986560467163315e-05, + "loss": 0.9615, + "step": 28123 + }, + { + "epoch": 1.81, + "learning_rate": 1.3986345778140364e-05, + "loss": 0.8669, + "step": 28124 + }, + { + "epoch": 1.81, + "learning_rate": 1.3986131089117413e-05, + "loss": 0.9975, + "step": 28125 + }, + { + "epoch": 1.81, + "learning_rate": 1.3985916400094464e-05, + "loss": 1.035, + "step": 28126 + }, + { + "epoch": 1.81, + "learning_rate": 1.3985701711071513e-05, + "loss": 0.9455, + "step": 28127 + }, + { + "epoch": 1.81, + "learning_rate": 1.3985487022048562e-05, + "loss": 1.0141, + "step": 28128 + }, + { + "epoch": 1.81, + "learning_rate": 1.3985272333025614e-05, + "loss": 0.9833, + "step": 28129 + }, + { + "epoch": 1.81, + "learning_rate": 1.3985057644002665e-05, + "loss": 0.8974, + "step": 28130 + }, + { + "epoch": 1.81, + "learning_rate": 1.3984842954979714e-05, + "loss": 0.9856, + "step": 28131 + }, + { + "epoch": 1.81, + "learning_rate": 1.3984628265956763e-05, + "loss": 0.9361, + "step": 28132 + }, + { + "epoch": 1.81, + "learning_rate": 1.3984413576933814e-05, + "loss": 0.9973, + "step": 28133 + }, + { + "epoch": 1.81, + "learning_rate": 1.3984198887910863e-05, + "loss": 1.0078, + "step": 28134 + }, + { + "epoch": 1.81, + "learning_rate": 1.3983984198887912e-05, + "loss": 0.8966, + "step": 28135 + }, + { + "epoch": 1.81, + "learning_rate": 1.3983769509864962e-05, + "loss": 0.9735, + "step": 28136 + }, + { + "epoch": 1.81, + "learning_rate": 1.3983554820842011e-05, + "loss": 1.0141, + "step": 28137 + }, + { + "epoch": 1.81, + "learning_rate": 1.398334013181906e-05, + "loss": 0.9126, + "step": 28138 + }, + { + "epoch": 1.81, + "learning_rate": 1.398312544279611e-05, + "loss": 0.9726, + "step": 28139 + }, + { + "epoch": 1.81, + "learning_rate": 1.398291075377316e-05, + "loss": 0.9551, + "step": 28140 + }, + { + "epoch": 1.81, + "learning_rate": 1.398269606475021e-05, + "loss": 0.9773, + "step": 28141 + }, + { + "epoch": 1.81, + "learning_rate": 1.398248137572726e-05, + "loss": 0.9812, + "step": 28142 + }, + { + "epoch": 1.81, + "learning_rate": 1.3982266686704308e-05, + "loss": 0.9814, + "step": 28143 + }, + { + "epoch": 1.81, + "learning_rate": 1.398205199768136e-05, + "loss": 0.9444, + "step": 28144 + }, + { + "epoch": 1.81, + "learning_rate": 1.398183730865841e-05, + "loss": 0.9741, + "step": 28145 + }, + { + "epoch": 1.81, + "learning_rate": 1.398162261963546e-05, + "loss": 0.9323, + "step": 28146 + }, + { + "epoch": 1.81, + "learning_rate": 1.398140793061251e-05, + "loss": 0.95, + "step": 28147 + }, + { + "epoch": 1.81, + "learning_rate": 1.3981193241589558e-05, + "loss": 0.9928, + "step": 28148 + }, + { + "epoch": 1.81, + "learning_rate": 1.3980978552566609e-05, + "loss": 0.9949, + "step": 28149 + }, + { + "epoch": 1.81, + "learning_rate": 1.3980763863543658e-05, + "loss": 0.8608, + "step": 28150 + }, + { + "epoch": 1.81, + "learning_rate": 1.3980549174520709e-05, + "loss": 0.9621, + "step": 28151 + }, + { + "epoch": 1.81, + "learning_rate": 1.3980334485497757e-05, + "loss": 0.9646, + "step": 28152 + }, + { + "epoch": 1.81, + "learning_rate": 1.3980119796474806e-05, + "loss": 0.9326, + "step": 28153 + }, + { + "epoch": 1.81, + "learning_rate": 1.3979905107451857e-05, + "loss": 0.932, + "step": 28154 + }, + { + "epoch": 1.81, + "learning_rate": 1.3979690418428906e-05, + "loss": 1.0374, + "step": 28155 + }, + { + "epoch": 1.81, + "learning_rate": 1.3979475729405955e-05, + "loss": 1.0083, + "step": 28156 + }, + { + "epoch": 1.81, + "learning_rate": 1.3979261040383006e-05, + "loss": 0.978, + "step": 28157 + }, + { + "epoch": 1.81, + "learning_rate": 1.3979046351360056e-05, + "loss": 0.8972, + "step": 28158 + }, + { + "epoch": 1.81, + "learning_rate": 1.3978831662337107e-05, + "loss": 0.9551, + "step": 28159 + }, + { + "epoch": 1.81, + "learning_rate": 1.3978616973314156e-05, + "loss": 0.9109, + "step": 28160 + }, + { + "epoch": 1.81, + "learning_rate": 1.3978402284291207e-05, + "loss": 0.9707, + "step": 28161 + }, + { + "epoch": 1.81, + "learning_rate": 1.3978187595268256e-05, + "loss": 0.9118, + "step": 28162 + }, + { + "epoch": 1.81, + "learning_rate": 1.3977972906245305e-05, + "loss": 0.9925, + "step": 28163 + }, + { + "epoch": 1.81, + "learning_rate": 1.3977758217222355e-05, + "loss": 0.9677, + "step": 28164 + }, + { + "epoch": 1.81, + "learning_rate": 1.3977543528199404e-05, + "loss": 0.9413, + "step": 28165 + }, + { + "epoch": 1.81, + "learning_rate": 1.3977328839176453e-05, + "loss": 0.9907, + "step": 28166 + }, + { + "epoch": 1.81, + "learning_rate": 1.3977114150153504e-05, + "loss": 0.9811, + "step": 28167 + }, + { + "epoch": 1.81, + "learning_rate": 1.3976899461130553e-05, + "loss": 0.9864, + "step": 28168 + }, + { + "epoch": 1.81, + "learning_rate": 1.3976684772107603e-05, + "loss": 0.9807, + "step": 28169 + }, + { + "epoch": 1.81, + "learning_rate": 1.3976470083084652e-05, + "loss": 0.9928, + "step": 28170 + }, + { + "epoch": 1.81, + "learning_rate": 1.3976255394061701e-05, + "loss": 1.0105, + "step": 28171 + }, + { + "epoch": 1.81, + "learning_rate": 1.3976040705038752e-05, + "loss": 0.9659, + "step": 28172 + }, + { + "epoch": 1.81, + "learning_rate": 1.3975826016015803e-05, + "loss": 0.9746, + "step": 28173 + }, + { + "epoch": 1.81, + "learning_rate": 1.3975611326992853e-05, + "loss": 0.8824, + "step": 28174 + }, + { + "epoch": 1.81, + "learning_rate": 1.3975396637969902e-05, + "loss": 0.9325, + "step": 28175 + }, + { + "epoch": 1.81, + "learning_rate": 1.3975181948946951e-05, + "loss": 0.983, + "step": 28176 + }, + { + "epoch": 1.81, + "learning_rate": 1.3974967259924002e-05, + "loss": 0.9524, + "step": 28177 + }, + { + "epoch": 1.81, + "learning_rate": 1.3974752570901051e-05, + "loss": 0.9557, + "step": 28178 + }, + { + "epoch": 1.81, + "learning_rate": 1.3974537881878102e-05, + "loss": 0.9763, + "step": 28179 + }, + { + "epoch": 1.81, + "learning_rate": 1.397432319285515e-05, + "loss": 0.9473, + "step": 28180 + }, + { + "epoch": 1.81, + "learning_rate": 1.39741085038322e-05, + "loss": 0.9404, + "step": 28181 + }, + { + "epoch": 1.81, + "learning_rate": 1.397389381480925e-05, + "loss": 1.0395, + "step": 28182 + }, + { + "epoch": 1.81, + "learning_rate": 1.3973679125786299e-05, + "loss": 1.0158, + "step": 28183 + }, + { + "epoch": 1.81, + "learning_rate": 1.3973464436763348e-05, + "loss": 0.9836, + "step": 28184 + }, + { + "epoch": 1.81, + "learning_rate": 1.3973249747740399e-05, + "loss": 0.9362, + "step": 28185 + }, + { + "epoch": 1.81, + "learning_rate": 1.3973035058717448e-05, + "loss": 0.9463, + "step": 28186 + }, + { + "epoch": 1.81, + "learning_rate": 1.39728203696945e-05, + "loss": 1.0317, + "step": 28187 + }, + { + "epoch": 1.81, + "learning_rate": 1.3972605680671549e-05, + "loss": 1.0075, + "step": 28188 + }, + { + "epoch": 1.81, + "learning_rate": 1.39723909916486e-05, + "loss": 0.9826, + "step": 28189 + }, + { + "epoch": 1.81, + "learning_rate": 1.3972176302625649e-05, + "loss": 0.9739, + "step": 28190 + }, + { + "epoch": 1.81, + "learning_rate": 1.3971961613602698e-05, + "loss": 1.0285, + "step": 28191 + }, + { + "epoch": 1.81, + "learning_rate": 1.3971746924579748e-05, + "loss": 1.0042, + "step": 28192 + }, + { + "epoch": 1.81, + "learning_rate": 1.3971532235556797e-05, + "loss": 0.9536, + "step": 28193 + }, + { + "epoch": 1.81, + "learning_rate": 1.3971317546533846e-05, + "loss": 0.9837, + "step": 28194 + }, + { + "epoch": 1.81, + "learning_rate": 1.3971102857510897e-05, + "loss": 0.9915, + "step": 28195 + }, + { + "epoch": 1.81, + "learning_rate": 1.3970888168487946e-05, + "loss": 0.9438, + "step": 28196 + }, + { + "epoch": 1.81, + "learning_rate": 1.3970673479464996e-05, + "loss": 0.9509, + "step": 28197 + }, + { + "epoch": 1.81, + "learning_rate": 1.3970458790442045e-05, + "loss": 1.0398, + "step": 28198 + }, + { + "epoch": 1.81, + "learning_rate": 1.3970244101419094e-05, + "loss": 0.9411, + "step": 28199 + }, + { + "epoch": 1.81, + "learning_rate": 1.3970029412396145e-05, + "loss": 1.0565, + "step": 28200 + }, + { + "epoch": 1.81, + "learning_rate": 1.3969814723373194e-05, + "loss": 0.9815, + "step": 28201 + }, + { + "epoch": 1.81, + "learning_rate": 1.3969600034350246e-05, + "loss": 0.9529, + "step": 28202 + }, + { + "epoch": 1.81, + "learning_rate": 1.3969385345327295e-05, + "loss": 0.9795, + "step": 28203 + }, + { + "epoch": 1.81, + "learning_rate": 1.3969170656304344e-05, + "loss": 0.9652, + "step": 28204 + }, + { + "epoch": 1.81, + "learning_rate": 1.3968955967281395e-05, + "loss": 0.9972, + "step": 28205 + }, + { + "epoch": 1.81, + "learning_rate": 1.3968741278258444e-05, + "loss": 1.0165, + "step": 28206 + }, + { + "epoch": 1.81, + "learning_rate": 1.3968526589235495e-05, + "loss": 0.9206, + "step": 28207 + }, + { + "epoch": 1.81, + "learning_rate": 1.3968311900212544e-05, + "loss": 0.9815, + "step": 28208 + }, + { + "epoch": 1.81, + "learning_rate": 1.3968097211189593e-05, + "loss": 0.9338, + "step": 28209 + }, + { + "epoch": 1.81, + "learning_rate": 1.3967882522166643e-05, + "loss": 0.9157, + "step": 28210 + }, + { + "epoch": 1.82, + "learning_rate": 1.3967667833143692e-05, + "loss": 0.9728, + "step": 28211 + }, + { + "epoch": 1.82, + "learning_rate": 1.3967453144120741e-05, + "loss": 0.9943, + "step": 28212 + }, + { + "epoch": 1.82, + "learning_rate": 1.3967238455097792e-05, + "loss": 0.989, + "step": 28213 + }, + { + "epoch": 1.82, + "learning_rate": 1.396702376607484e-05, + "loss": 0.9336, + "step": 28214 + }, + { + "epoch": 1.82, + "learning_rate": 1.396680907705189e-05, + "loss": 0.9491, + "step": 28215 + }, + { + "epoch": 1.82, + "learning_rate": 1.3966594388028942e-05, + "loss": 1.0031, + "step": 28216 + }, + { + "epoch": 1.82, + "learning_rate": 1.3966379699005993e-05, + "loss": 0.9448, + "step": 28217 + }, + { + "epoch": 1.82, + "learning_rate": 1.3966165009983042e-05, + "loss": 0.9259, + "step": 28218 + }, + { + "epoch": 1.82, + "learning_rate": 1.396595032096009e-05, + "loss": 1.0072, + "step": 28219 + }, + { + "epoch": 1.82, + "learning_rate": 1.3965735631937141e-05, + "loss": 1.0085, + "step": 28220 + }, + { + "epoch": 1.82, + "learning_rate": 1.396552094291419e-05, + "loss": 0.9858, + "step": 28221 + }, + { + "epoch": 1.82, + "learning_rate": 1.396530625389124e-05, + "loss": 0.9433, + "step": 28222 + }, + { + "epoch": 1.82, + "learning_rate": 1.396509156486829e-05, + "loss": 0.9839, + "step": 28223 + }, + { + "epoch": 1.82, + "learning_rate": 1.3964876875845339e-05, + "loss": 0.9664, + "step": 28224 + }, + { + "epoch": 1.82, + "learning_rate": 1.3964662186822388e-05, + "loss": 0.97, + "step": 28225 + }, + { + "epoch": 1.82, + "learning_rate": 1.3964447497799438e-05, + "loss": 0.9632, + "step": 28226 + }, + { + "epoch": 1.82, + "learning_rate": 1.3964232808776487e-05, + "loss": 0.9248, + "step": 28227 + }, + { + "epoch": 1.82, + "learning_rate": 1.3964018119753538e-05, + "loss": 1.0053, + "step": 28228 + }, + { + "epoch": 1.82, + "learning_rate": 1.3963803430730587e-05, + "loss": 0.9807, + "step": 28229 + }, + { + "epoch": 1.82, + "learning_rate": 1.3963588741707636e-05, + "loss": 0.961, + "step": 28230 + }, + { + "epoch": 1.82, + "learning_rate": 1.3963374052684688e-05, + "loss": 0.9558, + "step": 28231 + }, + { + "epoch": 1.82, + "learning_rate": 1.3963159363661737e-05, + "loss": 1.081, + "step": 28232 + }, + { + "epoch": 1.82, + "learning_rate": 1.3962944674638788e-05, + "loss": 0.9483, + "step": 28233 + }, + { + "epoch": 1.82, + "learning_rate": 1.3962729985615837e-05, + "loss": 0.9586, + "step": 28234 + }, + { + "epoch": 1.82, + "learning_rate": 1.3962515296592886e-05, + "loss": 0.8995, + "step": 28235 + }, + { + "epoch": 1.82, + "learning_rate": 1.3962300607569937e-05, + "loss": 1.0059, + "step": 28236 + }, + { + "epoch": 1.82, + "learning_rate": 1.3962085918546986e-05, + "loss": 0.9534, + "step": 28237 + }, + { + "epoch": 1.82, + "learning_rate": 1.3961871229524036e-05, + "loss": 0.9181, + "step": 28238 + }, + { + "epoch": 1.82, + "learning_rate": 1.3961656540501085e-05, + "loss": 0.9475, + "step": 28239 + }, + { + "epoch": 1.82, + "learning_rate": 1.3961441851478134e-05, + "loss": 0.933, + "step": 28240 + }, + { + "epoch": 1.82, + "learning_rate": 1.3961227162455185e-05, + "loss": 1.0021, + "step": 28241 + }, + { + "epoch": 1.82, + "learning_rate": 1.3961012473432234e-05, + "loss": 0.9772, + "step": 28242 + }, + { + "epoch": 1.82, + "learning_rate": 1.3960797784409283e-05, + "loss": 1.0054, + "step": 28243 + }, + { + "epoch": 1.82, + "learning_rate": 1.3960583095386333e-05, + "loss": 1.0235, + "step": 28244 + }, + { + "epoch": 1.82, + "learning_rate": 1.3960368406363384e-05, + "loss": 0.9605, + "step": 28245 + }, + { + "epoch": 1.82, + "learning_rate": 1.3960153717340435e-05, + "loss": 0.9842, + "step": 28246 + }, + { + "epoch": 1.82, + "learning_rate": 1.3959939028317484e-05, + "loss": 1.0751, + "step": 28247 + }, + { + "epoch": 1.82, + "learning_rate": 1.3959724339294534e-05, + "loss": 1.0097, + "step": 28248 + }, + { + "epoch": 1.82, + "learning_rate": 1.3959509650271583e-05, + "loss": 0.9221, + "step": 28249 + }, + { + "epoch": 1.82, + "learning_rate": 1.3959294961248632e-05, + "loss": 1.014, + "step": 28250 + }, + { + "epoch": 1.82, + "learning_rate": 1.3959080272225683e-05, + "loss": 0.9579, + "step": 28251 + }, + { + "epoch": 1.82, + "learning_rate": 1.3958865583202732e-05, + "loss": 0.9869, + "step": 28252 + }, + { + "epoch": 1.82, + "learning_rate": 1.3958650894179781e-05, + "loss": 1.0131, + "step": 28253 + }, + { + "epoch": 1.82, + "learning_rate": 1.3958436205156832e-05, + "loss": 0.9695, + "step": 28254 + }, + { + "epoch": 1.82, + "learning_rate": 1.395822151613388e-05, + "loss": 0.9383, + "step": 28255 + }, + { + "epoch": 1.82, + "learning_rate": 1.3958006827110931e-05, + "loss": 0.9632, + "step": 28256 + }, + { + "epoch": 1.82, + "learning_rate": 1.395779213808798e-05, + "loss": 0.957, + "step": 28257 + }, + { + "epoch": 1.82, + "learning_rate": 1.3957577449065029e-05, + "loss": 0.9645, + "step": 28258 + }, + { + "epoch": 1.82, + "learning_rate": 1.395736276004208e-05, + "loss": 0.9753, + "step": 28259 + }, + { + "epoch": 1.82, + "learning_rate": 1.395714807101913e-05, + "loss": 0.9554, + "step": 28260 + }, + { + "epoch": 1.82, + "learning_rate": 1.3956933381996181e-05, + "loss": 1.0437, + "step": 28261 + }, + { + "epoch": 1.82, + "learning_rate": 1.395671869297323e-05, + "loss": 0.9766, + "step": 28262 + }, + { + "epoch": 1.82, + "learning_rate": 1.3956504003950279e-05, + "loss": 0.9349, + "step": 28263 + }, + { + "epoch": 1.82, + "learning_rate": 1.395628931492733e-05, + "loss": 1.0173, + "step": 28264 + }, + { + "epoch": 1.82, + "learning_rate": 1.3956074625904379e-05, + "loss": 0.9836, + "step": 28265 + }, + { + "epoch": 1.82, + "learning_rate": 1.395585993688143e-05, + "loss": 0.9422, + "step": 28266 + }, + { + "epoch": 1.82, + "learning_rate": 1.3955645247858478e-05, + "loss": 1.0458, + "step": 28267 + }, + { + "epoch": 1.82, + "learning_rate": 1.3955430558835527e-05, + "loss": 0.9473, + "step": 28268 + }, + { + "epoch": 1.82, + "learning_rate": 1.3955215869812578e-05, + "loss": 1.0227, + "step": 28269 + }, + { + "epoch": 1.82, + "learning_rate": 1.3955001180789627e-05, + "loss": 0.9503, + "step": 28270 + }, + { + "epoch": 1.82, + "learning_rate": 1.3954786491766676e-05, + "loss": 1.0061, + "step": 28271 + }, + { + "epoch": 1.82, + "learning_rate": 1.3954571802743726e-05, + "loss": 0.9182, + "step": 28272 + }, + { + "epoch": 1.82, + "learning_rate": 1.3954357113720775e-05, + "loss": 0.9308, + "step": 28273 + }, + { + "epoch": 1.82, + "learning_rate": 1.3954142424697828e-05, + "loss": 1.011, + "step": 28274 + }, + { + "epoch": 1.82, + "learning_rate": 1.3953927735674877e-05, + "loss": 0.9531, + "step": 28275 + }, + { + "epoch": 1.82, + "learning_rate": 1.3953713046651927e-05, + "loss": 1.0278, + "step": 28276 + }, + { + "epoch": 1.82, + "learning_rate": 1.3953498357628976e-05, + "loss": 0.9587, + "step": 28277 + }, + { + "epoch": 1.82, + "learning_rate": 1.3953283668606025e-05, + "loss": 0.9659, + "step": 28278 + }, + { + "epoch": 1.82, + "learning_rate": 1.3953068979583076e-05, + "loss": 0.994, + "step": 28279 + }, + { + "epoch": 1.82, + "learning_rate": 1.3952854290560125e-05, + "loss": 0.9756, + "step": 28280 + }, + { + "epoch": 1.82, + "learning_rate": 1.3952639601537174e-05, + "loss": 0.9947, + "step": 28281 + }, + { + "epoch": 1.82, + "learning_rate": 1.3952424912514225e-05, + "loss": 0.9608, + "step": 28282 + }, + { + "epoch": 1.82, + "learning_rate": 1.3952210223491274e-05, + "loss": 0.9297, + "step": 28283 + }, + { + "epoch": 1.82, + "learning_rate": 1.3951995534468322e-05, + "loss": 0.9046, + "step": 28284 + }, + { + "epoch": 1.82, + "learning_rate": 1.3951780845445373e-05, + "loss": 1.0095, + "step": 28285 + }, + { + "epoch": 1.82, + "learning_rate": 1.3951566156422422e-05, + "loss": 0.9219, + "step": 28286 + }, + { + "epoch": 1.82, + "learning_rate": 1.3951351467399473e-05, + "loss": 0.8895, + "step": 28287 + }, + { + "epoch": 1.82, + "learning_rate": 1.3951136778376522e-05, + "loss": 0.943, + "step": 28288 + }, + { + "epoch": 1.82, + "learning_rate": 1.3950922089353574e-05, + "loss": 0.9083, + "step": 28289 + }, + { + "epoch": 1.82, + "learning_rate": 1.3950707400330623e-05, + "loss": 0.9956, + "step": 28290 + }, + { + "epoch": 1.82, + "learning_rate": 1.3950492711307672e-05, + "loss": 0.9325, + "step": 28291 + }, + { + "epoch": 1.82, + "learning_rate": 1.3950278022284723e-05, + "loss": 0.9544, + "step": 28292 + }, + { + "epoch": 1.82, + "learning_rate": 1.3950063333261772e-05, + "loss": 1.0369, + "step": 28293 + }, + { + "epoch": 1.82, + "learning_rate": 1.394984864423882e-05, + "loss": 0.9835, + "step": 28294 + }, + { + "epoch": 1.82, + "learning_rate": 1.3949633955215871e-05, + "loss": 1.0077, + "step": 28295 + }, + { + "epoch": 1.82, + "learning_rate": 1.394941926619292e-05, + "loss": 1.0931, + "step": 28296 + }, + { + "epoch": 1.82, + "learning_rate": 1.3949204577169971e-05, + "loss": 0.9662, + "step": 28297 + }, + { + "epoch": 1.82, + "learning_rate": 1.394898988814702e-05, + "loss": 0.9744, + "step": 28298 + }, + { + "epoch": 1.82, + "learning_rate": 1.3948775199124069e-05, + "loss": 0.9968, + "step": 28299 + }, + { + "epoch": 1.82, + "learning_rate": 1.394856051010112e-05, + "loss": 0.9809, + "step": 28300 + }, + { + "epoch": 1.82, + "learning_rate": 1.3948345821078168e-05, + "loss": 0.9252, + "step": 28301 + }, + { + "epoch": 1.82, + "learning_rate": 1.3948131132055217e-05, + "loss": 0.9943, + "step": 28302 + }, + { + "epoch": 1.82, + "learning_rate": 1.394791644303227e-05, + "loss": 0.9271, + "step": 28303 + }, + { + "epoch": 1.82, + "learning_rate": 1.3947701754009319e-05, + "loss": 0.931, + "step": 28304 + }, + { + "epoch": 1.82, + "learning_rate": 1.394748706498637e-05, + "loss": 0.938, + "step": 28305 + }, + { + "epoch": 1.82, + "learning_rate": 1.3947272375963418e-05, + "loss": 0.9247, + "step": 28306 + }, + { + "epoch": 1.82, + "learning_rate": 1.3947057686940469e-05, + "loss": 0.9439, + "step": 28307 + }, + { + "epoch": 1.82, + "learning_rate": 1.3946842997917518e-05, + "loss": 0.9096, + "step": 28308 + }, + { + "epoch": 1.82, + "learning_rate": 1.3946628308894567e-05, + "loss": 1.0237, + "step": 28309 + }, + { + "epoch": 1.82, + "learning_rate": 1.3946413619871618e-05, + "loss": 0.9951, + "step": 28310 + }, + { + "epoch": 1.82, + "learning_rate": 1.3946198930848667e-05, + "loss": 0.9775, + "step": 28311 + }, + { + "epoch": 1.82, + "learning_rate": 1.3945984241825716e-05, + "loss": 1.0166, + "step": 28312 + }, + { + "epoch": 1.82, + "learning_rate": 1.3945769552802766e-05, + "loss": 0.9311, + "step": 28313 + }, + { + "epoch": 1.82, + "learning_rate": 1.3945554863779815e-05, + "loss": 1.035, + "step": 28314 + }, + { + "epoch": 1.82, + "learning_rate": 1.3945340174756866e-05, + "loss": 0.9045, + "step": 28315 + }, + { + "epoch": 1.82, + "learning_rate": 1.3945125485733915e-05, + "loss": 1.0128, + "step": 28316 + }, + { + "epoch": 1.82, + "learning_rate": 1.3944910796710964e-05, + "loss": 1.0084, + "step": 28317 + }, + { + "epoch": 1.82, + "learning_rate": 1.3944696107688016e-05, + "loss": 1.0191, + "step": 28318 + }, + { + "epoch": 1.82, + "learning_rate": 1.3944481418665065e-05, + "loss": 0.9979, + "step": 28319 + }, + { + "epoch": 1.82, + "learning_rate": 1.3944266729642116e-05, + "loss": 0.9761, + "step": 28320 + }, + { + "epoch": 1.82, + "learning_rate": 1.3944052040619165e-05, + "loss": 1.0036, + "step": 28321 + }, + { + "epoch": 1.82, + "learning_rate": 1.3943837351596214e-05, + "loss": 0.8901, + "step": 28322 + }, + { + "epoch": 1.82, + "learning_rate": 1.3943622662573264e-05, + "loss": 1.0276, + "step": 28323 + }, + { + "epoch": 1.82, + "learning_rate": 1.3943407973550313e-05, + "loss": 1.0371, + "step": 28324 + }, + { + "epoch": 1.82, + "learning_rate": 1.3943193284527364e-05, + "loss": 0.983, + "step": 28325 + }, + { + "epoch": 1.82, + "learning_rate": 1.3942978595504413e-05, + "loss": 1.0338, + "step": 28326 + }, + { + "epoch": 1.82, + "learning_rate": 1.3942763906481462e-05, + "loss": 0.9378, + "step": 28327 + }, + { + "epoch": 1.82, + "learning_rate": 1.3942549217458513e-05, + "loss": 0.9054, + "step": 28328 + }, + { + "epoch": 1.82, + "learning_rate": 1.3942334528435561e-05, + "loss": 1.005, + "step": 28329 + }, + { + "epoch": 1.82, + "learning_rate": 1.394211983941261e-05, + "loss": 0.9194, + "step": 28330 + }, + { + "epoch": 1.82, + "learning_rate": 1.3941905150389661e-05, + "loss": 1.0391, + "step": 28331 + }, + { + "epoch": 1.82, + "learning_rate": 1.3941690461366712e-05, + "loss": 1.0059, + "step": 28332 + }, + { + "epoch": 1.82, + "learning_rate": 1.3941475772343762e-05, + "loss": 0.9182, + "step": 28333 + }, + { + "epoch": 1.82, + "learning_rate": 1.3941261083320811e-05, + "loss": 1.001, + "step": 28334 + }, + { + "epoch": 1.82, + "learning_rate": 1.3941046394297862e-05, + "loss": 0.9875, + "step": 28335 + }, + { + "epoch": 1.82, + "learning_rate": 1.3940831705274911e-05, + "loss": 0.966, + "step": 28336 + }, + { + "epoch": 1.82, + "learning_rate": 1.394061701625196e-05, + "loss": 0.9279, + "step": 28337 + }, + { + "epoch": 1.82, + "learning_rate": 1.394040232722901e-05, + "loss": 0.8648, + "step": 28338 + }, + { + "epoch": 1.82, + "learning_rate": 1.394018763820606e-05, + "loss": 0.9347, + "step": 28339 + }, + { + "epoch": 1.82, + "learning_rate": 1.3939972949183109e-05, + "loss": 0.9296, + "step": 28340 + }, + { + "epoch": 1.82, + "learning_rate": 1.393975826016016e-05, + "loss": 0.9793, + "step": 28341 + }, + { + "epoch": 1.82, + "learning_rate": 1.3939543571137208e-05, + "loss": 0.9403, + "step": 28342 + }, + { + "epoch": 1.82, + "learning_rate": 1.3939328882114259e-05, + "loss": 0.9458, + "step": 28343 + }, + { + "epoch": 1.82, + "learning_rate": 1.3939114193091308e-05, + "loss": 0.9535, + "step": 28344 + }, + { + "epoch": 1.82, + "learning_rate": 1.3938899504068357e-05, + "loss": 1.0336, + "step": 28345 + }, + { + "epoch": 1.82, + "learning_rate": 1.3938684815045407e-05, + "loss": 1.0069, + "step": 28346 + }, + { + "epoch": 1.82, + "learning_rate": 1.3938470126022458e-05, + "loss": 0.9776, + "step": 28347 + }, + { + "epoch": 1.82, + "learning_rate": 1.3938255436999509e-05, + "loss": 0.9071, + "step": 28348 + }, + { + "epoch": 1.82, + "learning_rate": 1.3938040747976558e-05, + "loss": 0.9761, + "step": 28349 + }, + { + "epoch": 1.82, + "learning_rate": 1.3937826058953607e-05, + "loss": 0.9874, + "step": 28350 + }, + { + "epoch": 1.82, + "learning_rate": 1.3937611369930657e-05, + "loss": 0.9668, + "step": 28351 + }, + { + "epoch": 1.82, + "learning_rate": 1.3937396680907706e-05, + "loss": 0.9554, + "step": 28352 + }, + { + "epoch": 1.82, + "learning_rate": 1.3937181991884757e-05, + "loss": 1.0301, + "step": 28353 + }, + { + "epoch": 1.82, + "learning_rate": 1.3936967302861806e-05, + "loss": 0.9554, + "step": 28354 + }, + { + "epoch": 1.82, + "learning_rate": 1.3936752613838855e-05, + "loss": 1.0242, + "step": 28355 + }, + { + "epoch": 1.82, + "learning_rate": 1.3936537924815906e-05, + "loss": 0.9687, + "step": 28356 + }, + { + "epoch": 1.82, + "learning_rate": 1.3936323235792955e-05, + "loss": 1.0685, + "step": 28357 + }, + { + "epoch": 1.82, + "learning_rate": 1.3936108546770004e-05, + "loss": 1.0299, + "step": 28358 + }, + { + "epoch": 1.82, + "learning_rate": 1.3935893857747054e-05, + "loss": 1.0305, + "step": 28359 + }, + { + "epoch": 1.82, + "learning_rate": 1.3935679168724103e-05, + "loss": 0.9976, + "step": 28360 + }, + { + "epoch": 1.82, + "learning_rate": 1.3935464479701155e-05, + "loss": 0.9756, + "step": 28361 + }, + { + "epoch": 1.82, + "learning_rate": 1.3935249790678204e-05, + "loss": 0.9498, + "step": 28362 + }, + { + "epoch": 1.82, + "learning_rate": 1.3935035101655255e-05, + "loss": 0.956, + "step": 28363 + }, + { + "epoch": 1.82, + "learning_rate": 1.3934820412632304e-05, + "loss": 0.9209, + "step": 28364 + }, + { + "epoch": 1.82, + "learning_rate": 1.3934605723609353e-05, + "loss": 0.9495, + "step": 28365 + }, + { + "epoch": 1.83, + "learning_rate": 1.3934391034586404e-05, + "loss": 0.9945, + "step": 28366 + }, + { + "epoch": 1.83, + "learning_rate": 1.3934176345563453e-05, + "loss": 0.9619, + "step": 28367 + }, + { + "epoch": 1.83, + "learning_rate": 1.3933961656540502e-05, + "loss": 0.9234, + "step": 28368 + }, + { + "epoch": 1.83, + "learning_rate": 1.3933746967517552e-05, + "loss": 1.0832, + "step": 28369 + }, + { + "epoch": 1.83, + "learning_rate": 1.3933532278494601e-05, + "loss": 0.9554, + "step": 28370 + }, + { + "epoch": 1.83, + "learning_rate": 1.393331758947165e-05, + "loss": 1.0088, + "step": 28371 + }, + { + "epoch": 1.83, + "learning_rate": 1.3933102900448701e-05, + "loss": 1.0008, + "step": 28372 + }, + { + "epoch": 1.83, + "learning_rate": 1.393288821142575e-05, + "loss": 0.9638, + "step": 28373 + }, + { + "epoch": 1.83, + "learning_rate": 1.39326735224028e-05, + "loss": 0.9803, + "step": 28374 + }, + { + "epoch": 1.83, + "learning_rate": 1.393245883337985e-05, + "loss": 1.0448, + "step": 28375 + }, + { + "epoch": 1.83, + "learning_rate": 1.3932244144356902e-05, + "loss": 0.98, + "step": 28376 + }, + { + "epoch": 1.83, + "learning_rate": 1.393202945533395e-05, + "loss": 0.9513, + "step": 28377 + }, + { + "epoch": 1.83, + "learning_rate": 1.3931814766311e-05, + "loss": 0.9953, + "step": 28378 + }, + { + "epoch": 1.83, + "learning_rate": 1.393160007728805e-05, + "loss": 0.9729, + "step": 28379 + }, + { + "epoch": 1.83, + "learning_rate": 1.39313853882651e-05, + "loss": 0.9231, + "step": 28380 + }, + { + "epoch": 1.83, + "learning_rate": 1.3931170699242148e-05, + "loss": 0.9653, + "step": 28381 + }, + { + "epoch": 1.83, + "learning_rate": 1.3930956010219199e-05, + "loss": 0.9329, + "step": 28382 + }, + { + "epoch": 1.83, + "learning_rate": 1.3930741321196248e-05, + "loss": 0.9846, + "step": 28383 + }, + { + "epoch": 1.83, + "learning_rate": 1.3930526632173299e-05, + "loss": 0.9741, + "step": 28384 + }, + { + "epoch": 1.83, + "learning_rate": 1.3930311943150348e-05, + "loss": 0.9775, + "step": 28385 + }, + { + "epoch": 1.83, + "learning_rate": 1.3930097254127397e-05, + "loss": 0.991, + "step": 28386 + }, + { + "epoch": 1.83, + "learning_rate": 1.3929882565104447e-05, + "loss": 1.0356, + "step": 28387 + }, + { + "epoch": 1.83, + "learning_rate": 1.3929667876081496e-05, + "loss": 0.8816, + "step": 28388 + }, + { + "epoch": 1.83, + "learning_rate": 1.3929453187058545e-05, + "loss": 1.0219, + "step": 28389 + }, + { + "epoch": 1.83, + "learning_rate": 1.3929238498035597e-05, + "loss": 1.0198, + "step": 28390 + }, + { + "epoch": 1.83, + "learning_rate": 1.3929023809012646e-05, + "loss": 0.9377, + "step": 28391 + }, + { + "epoch": 1.83, + "learning_rate": 1.3928809119989697e-05, + "loss": 1.0245, + "step": 28392 + }, + { + "epoch": 1.83, + "learning_rate": 1.3928594430966746e-05, + "loss": 0.9561, + "step": 28393 + }, + { + "epoch": 1.83, + "learning_rate": 1.3928379741943797e-05, + "loss": 0.9808, + "step": 28394 + }, + { + "epoch": 1.83, + "learning_rate": 1.3928165052920846e-05, + "loss": 0.9741, + "step": 28395 + }, + { + "epoch": 1.83, + "learning_rate": 1.3927950363897895e-05, + "loss": 0.9697, + "step": 28396 + }, + { + "epoch": 1.83, + "learning_rate": 1.3927735674874945e-05, + "loss": 0.9179, + "step": 28397 + }, + { + "epoch": 1.83, + "learning_rate": 1.3927520985851994e-05, + "loss": 0.9306, + "step": 28398 + }, + { + "epoch": 1.83, + "learning_rate": 1.3927306296829043e-05, + "loss": 0.9547, + "step": 28399 + }, + { + "epoch": 1.83, + "learning_rate": 1.3927091607806094e-05, + "loss": 1.0182, + "step": 28400 + }, + { + "epoch": 1.83, + "learning_rate": 1.3926876918783143e-05, + "loss": 0.9752, + "step": 28401 + }, + { + "epoch": 1.83, + "learning_rate": 1.3926662229760194e-05, + "loss": 0.8921, + "step": 28402 + }, + { + "epoch": 1.83, + "learning_rate": 1.3926447540737243e-05, + "loss": 0.9428, + "step": 28403 + }, + { + "epoch": 1.83, + "learning_rate": 1.3926232851714291e-05, + "loss": 0.9168, + "step": 28404 + }, + { + "epoch": 1.83, + "learning_rate": 1.3926018162691344e-05, + "loss": 0.9983, + "step": 28405 + }, + { + "epoch": 1.83, + "learning_rate": 1.3925803473668393e-05, + "loss": 0.9911, + "step": 28406 + }, + { + "epoch": 1.83, + "learning_rate": 1.3925588784645443e-05, + "loss": 0.955, + "step": 28407 + }, + { + "epoch": 1.83, + "learning_rate": 1.3925374095622492e-05, + "loss": 1.0112, + "step": 28408 + }, + { + "epoch": 1.83, + "learning_rate": 1.3925159406599541e-05, + "loss": 0.9524, + "step": 28409 + }, + { + "epoch": 1.83, + "learning_rate": 1.3924944717576592e-05, + "loss": 1.0043, + "step": 28410 + }, + { + "epoch": 1.83, + "learning_rate": 1.3924730028553641e-05, + "loss": 1.0138, + "step": 28411 + }, + { + "epoch": 1.83, + "learning_rate": 1.3924515339530692e-05, + "loss": 0.8906, + "step": 28412 + }, + { + "epoch": 1.83, + "learning_rate": 1.392430065050774e-05, + "loss": 0.9773, + "step": 28413 + }, + { + "epoch": 1.83, + "learning_rate": 1.392408596148479e-05, + "loss": 0.9779, + "step": 28414 + }, + { + "epoch": 1.83, + "learning_rate": 1.392387127246184e-05, + "loss": 0.9897, + "step": 28415 + }, + { + "epoch": 1.83, + "learning_rate": 1.392365658343889e-05, + "loss": 1.1103, + "step": 28416 + }, + { + "epoch": 1.83, + "learning_rate": 1.3923441894415938e-05, + "loss": 0.9006, + "step": 28417 + }, + { + "epoch": 1.83, + "learning_rate": 1.3923227205392989e-05, + "loss": 0.9355, + "step": 28418 + }, + { + "epoch": 1.83, + "learning_rate": 1.3923012516370038e-05, + "loss": 1.0569, + "step": 28419 + }, + { + "epoch": 1.83, + "learning_rate": 1.392279782734709e-05, + "loss": 0.9473, + "step": 28420 + }, + { + "epoch": 1.83, + "learning_rate": 1.3922583138324139e-05, + "loss": 0.9501, + "step": 28421 + }, + { + "epoch": 1.83, + "learning_rate": 1.392236844930119e-05, + "loss": 0.9579, + "step": 28422 + }, + { + "epoch": 1.83, + "learning_rate": 1.3922153760278239e-05, + "loss": 1.0857, + "step": 28423 + }, + { + "epoch": 1.83, + "learning_rate": 1.3921939071255288e-05, + "loss": 1.0324, + "step": 28424 + }, + { + "epoch": 1.83, + "learning_rate": 1.3921724382232338e-05, + "loss": 1.0106, + "step": 28425 + }, + { + "epoch": 1.83, + "learning_rate": 1.3921509693209387e-05, + "loss": 1.0335, + "step": 28426 + }, + { + "epoch": 1.83, + "learning_rate": 1.3921295004186436e-05, + "loss": 1.0179, + "step": 28427 + }, + { + "epoch": 1.83, + "learning_rate": 1.3921080315163487e-05, + "loss": 1.0115, + "step": 28428 + }, + { + "epoch": 1.83, + "learning_rate": 1.3920865626140536e-05, + "loss": 0.9787, + "step": 28429 + }, + { + "epoch": 1.83, + "learning_rate": 1.3920650937117587e-05, + "loss": 0.9697, + "step": 28430 + }, + { + "epoch": 1.83, + "learning_rate": 1.3920436248094636e-05, + "loss": 1.018, + "step": 28431 + }, + { + "epoch": 1.83, + "learning_rate": 1.3920221559071685e-05, + "loss": 0.9484, + "step": 28432 + }, + { + "epoch": 1.83, + "learning_rate": 1.3920006870048735e-05, + "loss": 0.9541, + "step": 28433 + }, + { + "epoch": 1.83, + "learning_rate": 1.3919792181025786e-05, + "loss": 0.9906, + "step": 28434 + }, + { + "epoch": 1.83, + "learning_rate": 1.3919577492002836e-05, + "loss": 0.9769, + "step": 28435 + }, + { + "epoch": 1.83, + "learning_rate": 1.3919362802979885e-05, + "loss": 0.9389, + "step": 28436 + }, + { + "epoch": 1.83, + "learning_rate": 1.3919148113956934e-05, + "loss": 1.0459, + "step": 28437 + }, + { + "epoch": 1.83, + "learning_rate": 1.3918933424933985e-05, + "loss": 0.9667, + "step": 28438 + }, + { + "epoch": 1.83, + "learning_rate": 1.3918718735911034e-05, + "loss": 0.9395, + "step": 28439 + }, + { + "epoch": 1.83, + "learning_rate": 1.3918504046888085e-05, + "loss": 0.9581, + "step": 28440 + }, + { + "epoch": 1.83, + "learning_rate": 1.3918289357865134e-05, + "loss": 0.9464, + "step": 28441 + }, + { + "epoch": 1.83, + "learning_rate": 1.3918074668842183e-05, + "loss": 1.0046, + "step": 28442 + }, + { + "epoch": 1.83, + "learning_rate": 1.3917859979819233e-05, + "loss": 0.9864, + "step": 28443 + }, + { + "epoch": 1.83, + "learning_rate": 1.3917859979819233e-05, + "loss": 1.0442, + "step": 28444 + }, + { + "epoch": 1.83, + "learning_rate": 1.3917645290796282e-05, + "loss": 0.9989, + "step": 28445 + }, + { + "epoch": 1.83, + "learning_rate": 1.3917430601773331e-05, + "loss": 0.9534, + "step": 28446 + }, + { + "epoch": 1.83, + "learning_rate": 1.3917215912750382e-05, + "loss": 0.9321, + "step": 28447 + }, + { + "epoch": 1.83, + "learning_rate": 1.391700122372743e-05, + "loss": 0.9958, + "step": 28448 + }, + { + "epoch": 1.83, + "learning_rate": 1.391678653470448e-05, + "loss": 0.9638, + "step": 28449 + }, + { + "epoch": 1.83, + "learning_rate": 1.3916571845681532e-05, + "loss": 0.9838, + "step": 28450 + }, + { + "epoch": 1.83, + "learning_rate": 1.3916357156658583e-05, + "loss": 0.9868, + "step": 28451 + }, + { + "epoch": 1.83, + "learning_rate": 1.3916142467635632e-05, + "loss": 0.9529, + "step": 28452 + }, + { + "epoch": 1.83, + "learning_rate": 1.391592777861268e-05, + "loss": 0.9193, + "step": 28453 + }, + { + "epoch": 1.83, + "learning_rate": 1.3915713089589731e-05, + "loss": 0.9396, + "step": 28454 + }, + { + "epoch": 1.83, + "learning_rate": 1.391549840056678e-05, + "loss": 0.9578, + "step": 28455 + }, + { + "epoch": 1.83, + "learning_rate": 1.391528371154383e-05, + "loss": 1.0113, + "step": 28456 + }, + { + "epoch": 1.83, + "learning_rate": 1.391506902252088e-05, + "loss": 1.0429, + "step": 28457 + }, + { + "epoch": 1.83, + "learning_rate": 1.3914854333497929e-05, + "loss": 0.9251, + "step": 28458 + }, + { + "epoch": 1.83, + "learning_rate": 1.3914639644474978e-05, + "loss": 0.9893, + "step": 28459 + }, + { + "epoch": 1.83, + "learning_rate": 1.3914424955452029e-05, + "loss": 0.9509, + "step": 28460 + }, + { + "epoch": 1.83, + "learning_rate": 1.3914210266429078e-05, + "loss": 0.9296, + "step": 28461 + }, + { + "epoch": 1.83, + "learning_rate": 1.3913995577406128e-05, + "loss": 0.9804, + "step": 28462 + }, + { + "epoch": 1.83, + "learning_rate": 1.3913780888383177e-05, + "loss": 0.9812, + "step": 28463 + }, + { + "epoch": 1.83, + "learning_rate": 1.391356619936023e-05, + "loss": 0.9901, + "step": 28464 + }, + { + "epoch": 1.83, + "learning_rate": 1.3913351510337278e-05, + "loss": 0.9408, + "step": 28465 + }, + { + "epoch": 1.83, + "learning_rate": 1.3913136821314327e-05, + "loss": 0.9928, + "step": 28466 + }, + { + "epoch": 1.83, + "learning_rate": 1.3912922132291378e-05, + "loss": 0.9628, + "step": 28467 + }, + { + "epoch": 1.83, + "learning_rate": 1.3912707443268427e-05, + "loss": 1.0492, + "step": 28468 + }, + { + "epoch": 1.83, + "learning_rate": 1.3912492754245476e-05, + "loss": 0.9225, + "step": 28469 + }, + { + "epoch": 1.83, + "learning_rate": 1.3912278065222527e-05, + "loss": 0.9627, + "step": 28470 + }, + { + "epoch": 1.83, + "learning_rate": 1.3912063376199576e-05, + "loss": 1.0597, + "step": 28471 + }, + { + "epoch": 1.83, + "learning_rate": 1.3911848687176626e-05, + "loss": 0.9451, + "step": 28472 + }, + { + "epoch": 1.83, + "learning_rate": 1.3911633998153675e-05, + "loss": 1.0086, + "step": 28473 + }, + { + "epoch": 1.83, + "learning_rate": 1.3911419309130724e-05, + "loss": 0.9749, + "step": 28474 + }, + { + "epoch": 1.83, + "learning_rate": 1.3911204620107775e-05, + "loss": 0.9655, + "step": 28475 + }, + { + "epoch": 1.83, + "learning_rate": 1.3910989931084824e-05, + "loss": 0.9682, + "step": 28476 + }, + { + "epoch": 1.83, + "learning_rate": 1.3910775242061873e-05, + "loss": 0.9846, + "step": 28477 + }, + { + "epoch": 1.83, + "learning_rate": 1.3910560553038924e-05, + "loss": 0.9936, + "step": 28478 + }, + { + "epoch": 1.83, + "learning_rate": 1.3910345864015974e-05, + "loss": 0.9045, + "step": 28479 + }, + { + "epoch": 1.83, + "learning_rate": 1.3910131174993025e-05, + "loss": 1.0674, + "step": 28480 + }, + { + "epoch": 1.83, + "learning_rate": 1.3909916485970074e-05, + "loss": 0.9, + "step": 28481 + }, + { + "epoch": 1.83, + "learning_rate": 1.3909701796947124e-05, + "loss": 0.9577, + "step": 28482 + }, + { + "epoch": 1.83, + "learning_rate": 1.3909487107924173e-05, + "loss": 0.9878, + "step": 28483 + }, + { + "epoch": 1.83, + "learning_rate": 1.3909272418901222e-05, + "loss": 0.9685, + "step": 28484 + }, + { + "epoch": 1.83, + "learning_rate": 1.3909057729878273e-05, + "loss": 0.9608, + "step": 28485 + }, + { + "epoch": 1.83, + "learning_rate": 1.3908843040855322e-05, + "loss": 0.9626, + "step": 28486 + }, + { + "epoch": 1.83, + "learning_rate": 1.3908628351832371e-05, + "loss": 0.999, + "step": 28487 + }, + { + "epoch": 1.83, + "learning_rate": 1.3908413662809422e-05, + "loss": 1.0349, + "step": 28488 + }, + { + "epoch": 1.83, + "learning_rate": 1.390819897378647e-05, + "loss": 0.9439, + "step": 28489 + }, + { + "epoch": 1.83, + "learning_rate": 1.3907984284763521e-05, + "loss": 0.9152, + "step": 28490 + }, + { + "epoch": 1.83, + "learning_rate": 1.390776959574057e-05, + "loss": 0.8697, + "step": 28491 + }, + { + "epoch": 1.83, + "learning_rate": 1.390755490671762e-05, + "loss": 0.9235, + "step": 28492 + }, + { + "epoch": 1.83, + "learning_rate": 1.3907340217694672e-05, + "loss": 1.0403, + "step": 28493 + }, + { + "epoch": 1.83, + "learning_rate": 1.390712552867172e-05, + "loss": 0.9109, + "step": 28494 + }, + { + "epoch": 1.83, + "learning_rate": 1.3906910839648771e-05, + "loss": 0.9901, + "step": 28495 + }, + { + "epoch": 1.83, + "learning_rate": 1.390669615062582e-05, + "loss": 1.0289, + "step": 28496 + }, + { + "epoch": 1.83, + "learning_rate": 1.3906481461602869e-05, + "loss": 1.0165, + "step": 28497 + }, + { + "epoch": 1.83, + "learning_rate": 1.390626677257992e-05, + "loss": 1.0072, + "step": 28498 + }, + { + "epoch": 1.83, + "learning_rate": 1.3906052083556969e-05, + "loss": 0.8911, + "step": 28499 + }, + { + "epoch": 1.83, + "learning_rate": 1.390583739453402e-05, + "loss": 1.0006, + "step": 28500 + }, + { + "epoch": 1.83, + "learning_rate": 1.3905622705511068e-05, + "loss": 0.9555, + "step": 28501 + }, + { + "epoch": 1.83, + "learning_rate": 1.3905408016488117e-05, + "loss": 1.017, + "step": 28502 + }, + { + "epoch": 1.83, + "learning_rate": 1.3905193327465168e-05, + "loss": 0.9473, + "step": 28503 + }, + { + "epoch": 1.83, + "learning_rate": 1.3904978638442217e-05, + "loss": 0.9601, + "step": 28504 + }, + { + "epoch": 1.83, + "learning_rate": 1.3904763949419266e-05, + "loss": 0.9626, + "step": 28505 + }, + { + "epoch": 1.83, + "learning_rate": 1.3904549260396317e-05, + "loss": 0.9469, + "step": 28506 + }, + { + "epoch": 1.83, + "learning_rate": 1.3904334571373366e-05, + "loss": 0.9135, + "step": 28507 + }, + { + "epoch": 1.83, + "learning_rate": 1.3904119882350418e-05, + "loss": 0.9944, + "step": 28508 + }, + { + "epoch": 1.83, + "learning_rate": 1.3903905193327467e-05, + "loss": 0.9305, + "step": 28509 + }, + { + "epoch": 1.83, + "learning_rate": 1.3903690504304517e-05, + "loss": 0.9424, + "step": 28510 + }, + { + "epoch": 1.83, + "learning_rate": 1.3903475815281566e-05, + "loss": 1.0047, + "step": 28511 + }, + { + "epoch": 1.83, + "learning_rate": 1.3903261126258615e-05, + "loss": 1.0819, + "step": 28512 + }, + { + "epoch": 1.83, + "learning_rate": 1.3903046437235666e-05, + "loss": 0.9943, + "step": 28513 + }, + { + "epoch": 1.83, + "learning_rate": 1.3902831748212715e-05, + "loss": 0.9275, + "step": 28514 + }, + { + "epoch": 1.83, + "learning_rate": 1.3902617059189764e-05, + "loss": 1.0182, + "step": 28515 + }, + { + "epoch": 1.83, + "learning_rate": 1.3902402370166815e-05, + "loss": 0.9775, + "step": 28516 + }, + { + "epoch": 1.83, + "learning_rate": 1.3902187681143864e-05, + "loss": 0.9894, + "step": 28517 + }, + { + "epoch": 1.83, + "learning_rate": 1.3901972992120914e-05, + "loss": 0.9743, + "step": 28518 + }, + { + "epoch": 1.83, + "learning_rate": 1.3901758303097963e-05, + "loss": 0.9705, + "step": 28519 + }, + { + "epoch": 1.83, + "learning_rate": 1.3901543614075012e-05, + "loss": 1.0413, + "step": 28520 + }, + { + "epoch": 1.83, + "learning_rate": 1.3901328925052063e-05, + "loss": 0.9359, + "step": 28521 + }, + { + "epoch": 1.84, + "learning_rate": 1.3901114236029114e-05, + "loss": 1.0334, + "step": 28522 + }, + { + "epoch": 1.84, + "learning_rate": 1.3900899547006164e-05, + "loss": 0.9294, + "step": 28523 + }, + { + "epoch": 1.84, + "learning_rate": 1.3900684857983213e-05, + "loss": 0.9164, + "step": 28524 + }, + { + "epoch": 1.84, + "learning_rate": 1.3900470168960262e-05, + "loss": 0.9502, + "step": 28525 + }, + { + "epoch": 1.84, + "learning_rate": 1.3900255479937313e-05, + "loss": 0.9177, + "step": 28526 + }, + { + "epoch": 1.84, + "learning_rate": 1.3900040790914362e-05, + "loss": 1.0123, + "step": 28527 + }, + { + "epoch": 1.84, + "learning_rate": 1.3899826101891412e-05, + "loss": 0.8987, + "step": 28528 + }, + { + "epoch": 1.84, + "learning_rate": 1.3899611412868461e-05, + "loss": 1.0505, + "step": 28529 + }, + { + "epoch": 1.84, + "learning_rate": 1.389939672384551e-05, + "loss": 1.0774, + "step": 28530 + }, + { + "epoch": 1.84, + "learning_rate": 1.3899182034822561e-05, + "loss": 0.9509, + "step": 28531 + }, + { + "epoch": 1.84, + "learning_rate": 1.389896734579961e-05, + "loss": 0.9567, + "step": 28532 + }, + { + "epoch": 1.84, + "learning_rate": 1.3898752656776659e-05, + "loss": 0.89, + "step": 28533 + }, + { + "epoch": 1.84, + "learning_rate": 1.389853796775371e-05, + "loss": 0.9297, + "step": 28534 + }, + { + "epoch": 1.84, + "learning_rate": 1.3898323278730759e-05, + "loss": 0.8848, + "step": 28535 + }, + { + "epoch": 1.84, + "learning_rate": 1.3898108589707808e-05, + "loss": 0.9906, + "step": 28536 + }, + { + "epoch": 1.84, + "learning_rate": 1.389789390068486e-05, + "loss": 0.9819, + "step": 28537 + }, + { + "epoch": 1.84, + "learning_rate": 1.389767921166191e-05, + "loss": 1.0429, + "step": 28538 + }, + { + "epoch": 1.84, + "learning_rate": 1.389746452263896e-05, + "loss": 0.9581, + "step": 28539 + }, + { + "epoch": 1.84, + "learning_rate": 1.3897249833616008e-05, + "loss": 0.9136, + "step": 28540 + }, + { + "epoch": 1.84, + "learning_rate": 1.3897035144593059e-05, + "loss": 0.9741, + "step": 28541 + }, + { + "epoch": 1.84, + "learning_rate": 1.3896820455570108e-05, + "loss": 0.9712, + "step": 28542 + }, + { + "epoch": 1.84, + "learning_rate": 1.3896605766547157e-05, + "loss": 0.9768, + "step": 28543 + }, + { + "epoch": 1.84, + "learning_rate": 1.3896391077524208e-05, + "loss": 0.8733, + "step": 28544 + }, + { + "epoch": 1.84, + "learning_rate": 1.3896176388501257e-05, + "loss": 0.9808, + "step": 28545 + }, + { + "epoch": 1.84, + "learning_rate": 1.3895961699478306e-05, + "loss": 0.9896, + "step": 28546 + }, + { + "epoch": 1.84, + "learning_rate": 1.3895747010455356e-05, + "loss": 0.9529, + "step": 28547 + }, + { + "epoch": 1.84, + "learning_rate": 1.3895532321432405e-05, + "loss": 0.9523, + "step": 28548 + }, + { + "epoch": 1.84, + "learning_rate": 1.3895317632409456e-05, + "loss": 0.9061, + "step": 28549 + }, + { + "epoch": 1.84, + "learning_rate": 1.3895102943386505e-05, + "loss": 0.9449, + "step": 28550 + }, + { + "epoch": 1.84, + "learning_rate": 1.3894888254363557e-05, + "loss": 0.9986, + "step": 28551 + }, + { + "epoch": 1.84, + "learning_rate": 1.3894673565340606e-05, + "loss": 0.9223, + "step": 28552 + }, + { + "epoch": 1.84, + "learning_rate": 1.3894458876317655e-05, + "loss": 0.9768, + "step": 28553 + }, + { + "epoch": 1.84, + "learning_rate": 1.3894244187294706e-05, + "loss": 1.0111, + "step": 28554 + }, + { + "epoch": 1.84, + "learning_rate": 1.3894029498271755e-05, + "loss": 0.9501, + "step": 28555 + }, + { + "epoch": 1.84, + "learning_rate": 1.3893814809248804e-05, + "loss": 0.9781, + "step": 28556 + }, + { + "epoch": 1.84, + "learning_rate": 1.3893600120225854e-05, + "loss": 0.9125, + "step": 28557 + }, + { + "epoch": 1.84, + "learning_rate": 1.3893385431202903e-05, + "loss": 1.0129, + "step": 28558 + }, + { + "epoch": 1.84, + "learning_rate": 1.3893170742179954e-05, + "loss": 1.0425, + "step": 28559 + }, + { + "epoch": 1.84, + "learning_rate": 1.3892956053157003e-05, + "loss": 0.9073, + "step": 28560 + }, + { + "epoch": 1.84, + "learning_rate": 1.3892741364134052e-05, + "loss": 0.9833, + "step": 28561 + }, + { + "epoch": 1.84, + "learning_rate": 1.3892526675111103e-05, + "loss": 0.9946, + "step": 28562 + }, + { + "epoch": 1.84, + "learning_rate": 1.3892311986088152e-05, + "loss": 0.8989, + "step": 28563 + }, + { + "epoch": 1.84, + "learning_rate": 1.38920972970652e-05, + "loss": 0.9306, + "step": 28564 + }, + { + "epoch": 1.84, + "learning_rate": 1.3891882608042251e-05, + "loss": 0.9049, + "step": 28565 + }, + { + "epoch": 1.84, + "learning_rate": 1.3891667919019302e-05, + "loss": 1.0133, + "step": 28566 + }, + { + "epoch": 1.84, + "learning_rate": 1.3891453229996353e-05, + "loss": 1.0066, + "step": 28567 + }, + { + "epoch": 1.84, + "learning_rate": 1.3891238540973401e-05, + "loss": 1.0028, + "step": 28568 + }, + { + "epoch": 1.84, + "learning_rate": 1.3891023851950452e-05, + "loss": 0.9596, + "step": 28569 + }, + { + "epoch": 1.84, + "learning_rate": 1.3890809162927501e-05, + "loss": 1.0066, + "step": 28570 + }, + { + "epoch": 1.84, + "learning_rate": 1.389059447390455e-05, + "loss": 0.9746, + "step": 28571 + }, + { + "epoch": 1.84, + "learning_rate": 1.38903797848816e-05, + "loss": 0.9693, + "step": 28572 + }, + { + "epoch": 1.84, + "learning_rate": 1.389016509585865e-05, + "loss": 0.9882, + "step": 28573 + }, + { + "epoch": 1.84, + "learning_rate": 1.3889950406835699e-05, + "loss": 1.0132, + "step": 28574 + }, + { + "epoch": 1.84, + "learning_rate": 1.388973571781275e-05, + "loss": 0.9158, + "step": 28575 + }, + { + "epoch": 1.84, + "learning_rate": 1.3889521028789798e-05, + "loss": 0.9312, + "step": 28576 + }, + { + "epoch": 1.84, + "learning_rate": 1.3889306339766849e-05, + "loss": 0.9945, + "step": 28577 + }, + { + "epoch": 1.84, + "learning_rate": 1.3889091650743898e-05, + "loss": 0.9911, + "step": 28578 + }, + { + "epoch": 1.84, + "learning_rate": 1.3888876961720947e-05, + "loss": 0.9755, + "step": 28579 + }, + { + "epoch": 1.84, + "learning_rate": 1.3888662272698e-05, + "loss": 0.9643, + "step": 28580 + }, + { + "epoch": 1.84, + "learning_rate": 1.3888447583675048e-05, + "loss": 0.9792, + "step": 28581 + }, + { + "epoch": 1.84, + "learning_rate": 1.3888232894652099e-05, + "loss": 0.9444, + "step": 28582 + }, + { + "epoch": 1.84, + "learning_rate": 1.3888018205629148e-05, + "loss": 0.957, + "step": 28583 + }, + { + "epoch": 1.84, + "learning_rate": 1.3887803516606197e-05, + "loss": 1.0159, + "step": 28584 + }, + { + "epoch": 1.84, + "learning_rate": 1.3887588827583247e-05, + "loss": 0.9945, + "step": 28585 + }, + { + "epoch": 1.84, + "learning_rate": 1.3887374138560296e-05, + "loss": 1.0009, + "step": 28586 + }, + { + "epoch": 1.84, + "learning_rate": 1.3887159449537347e-05, + "loss": 0.9363, + "step": 28587 + }, + { + "epoch": 1.84, + "learning_rate": 1.3886944760514396e-05, + "loss": 0.9305, + "step": 28588 + }, + { + "epoch": 1.84, + "learning_rate": 1.3886730071491445e-05, + "loss": 0.9995, + "step": 28589 + }, + { + "epoch": 1.84, + "learning_rate": 1.3886515382468496e-05, + "loss": 0.8785, + "step": 28590 + }, + { + "epoch": 1.84, + "learning_rate": 1.3886300693445545e-05, + "loss": 0.9416, + "step": 28591 + }, + { + "epoch": 1.84, + "learning_rate": 1.3886086004422594e-05, + "loss": 1.0376, + "step": 28592 + }, + { + "epoch": 1.84, + "learning_rate": 1.3885871315399644e-05, + "loss": 0.9762, + "step": 28593 + }, + { + "epoch": 1.84, + "learning_rate": 1.3885656626376693e-05, + "loss": 0.9707, + "step": 28594 + }, + { + "epoch": 1.84, + "learning_rate": 1.3885441937353746e-05, + "loss": 0.96, + "step": 28595 + }, + { + "epoch": 1.84, + "learning_rate": 1.3885227248330795e-05, + "loss": 0.8678, + "step": 28596 + }, + { + "epoch": 1.84, + "learning_rate": 1.3885012559307845e-05, + "loss": 0.9106, + "step": 28597 + }, + { + "epoch": 1.84, + "learning_rate": 1.3884797870284894e-05, + "loss": 1.0046, + "step": 28598 + }, + { + "epoch": 1.84, + "learning_rate": 1.3884583181261943e-05, + "loss": 0.9748, + "step": 28599 + }, + { + "epoch": 1.84, + "learning_rate": 1.3884368492238994e-05, + "loss": 0.9075, + "step": 28600 + }, + { + "epoch": 1.84, + "learning_rate": 1.3884153803216043e-05, + "loss": 0.9595, + "step": 28601 + }, + { + "epoch": 1.84, + "learning_rate": 1.3883939114193092e-05, + "loss": 0.9532, + "step": 28602 + }, + { + "epoch": 1.84, + "learning_rate": 1.3883724425170142e-05, + "loss": 1.0099, + "step": 28603 + }, + { + "epoch": 1.84, + "learning_rate": 1.3883509736147191e-05, + "loss": 0.9515, + "step": 28604 + }, + { + "epoch": 1.84, + "learning_rate": 1.388329504712424e-05, + "loss": 0.9455, + "step": 28605 + }, + { + "epoch": 1.84, + "learning_rate": 1.3883080358101291e-05, + "loss": 0.9736, + "step": 28606 + }, + { + "epoch": 1.84, + "learning_rate": 1.388286566907834e-05, + "loss": 0.9727, + "step": 28607 + }, + { + "epoch": 1.84, + "learning_rate": 1.388265098005539e-05, + "loss": 1.0668, + "step": 28608 + }, + { + "epoch": 1.84, + "learning_rate": 1.3882436291032441e-05, + "loss": 0.9943, + "step": 28609 + }, + { + "epoch": 1.84, + "learning_rate": 1.3882221602009492e-05, + "loss": 0.9623, + "step": 28610 + }, + { + "epoch": 1.84, + "learning_rate": 1.3882006912986541e-05, + "loss": 0.9809, + "step": 28611 + }, + { + "epoch": 1.84, + "learning_rate": 1.388179222396359e-05, + "loss": 0.9173, + "step": 28612 + }, + { + "epoch": 1.84, + "learning_rate": 1.388157753494064e-05, + "loss": 1.0179, + "step": 28613 + }, + { + "epoch": 1.84, + "learning_rate": 1.388136284591769e-05, + "loss": 0.9373, + "step": 28614 + }, + { + "epoch": 1.84, + "learning_rate": 1.3881148156894738e-05, + "loss": 1.0112, + "step": 28615 + }, + { + "epoch": 1.84, + "learning_rate": 1.3880933467871789e-05, + "loss": 1.0178, + "step": 28616 + }, + { + "epoch": 1.84, + "learning_rate": 1.3880718778848838e-05, + "loss": 0.9762, + "step": 28617 + }, + { + "epoch": 1.84, + "learning_rate": 1.3880504089825889e-05, + "loss": 1.0285, + "step": 28618 + }, + { + "epoch": 1.84, + "learning_rate": 1.3880289400802938e-05, + "loss": 0.9735, + "step": 28619 + }, + { + "epoch": 1.84, + "learning_rate": 1.3880074711779987e-05, + "loss": 0.9151, + "step": 28620 + }, + { + "epoch": 1.84, + "learning_rate": 1.3879860022757037e-05, + "loss": 0.9748, + "step": 28621 + }, + { + "epoch": 1.84, + "learning_rate": 1.3879645333734086e-05, + "loss": 0.9558, + "step": 28622 + }, + { + "epoch": 1.84, + "learning_rate": 1.3879430644711135e-05, + "loss": 1.0758, + "step": 28623 + }, + { + "epoch": 1.84, + "learning_rate": 1.3879215955688188e-05, + "loss": 0.9711, + "step": 28624 + }, + { + "epoch": 1.84, + "learning_rate": 1.3879001266665238e-05, + "loss": 0.9757, + "step": 28625 + }, + { + "epoch": 1.84, + "learning_rate": 1.3878786577642287e-05, + "loss": 0.9642, + "step": 28626 + }, + { + "epoch": 1.84, + "learning_rate": 1.3878571888619336e-05, + "loss": 0.9792, + "step": 28627 + }, + { + "epoch": 1.84, + "learning_rate": 1.3878357199596387e-05, + "loss": 0.9721, + "step": 28628 + }, + { + "epoch": 1.84, + "learning_rate": 1.3878142510573436e-05, + "loss": 0.9625, + "step": 28629 + }, + { + "epoch": 1.84, + "learning_rate": 1.3877927821550485e-05, + "loss": 1.0213, + "step": 28630 + }, + { + "epoch": 1.84, + "learning_rate": 1.3877713132527535e-05, + "loss": 0.9668, + "step": 28631 + }, + { + "epoch": 1.84, + "learning_rate": 1.3877498443504584e-05, + "loss": 0.9399, + "step": 28632 + }, + { + "epoch": 1.84, + "learning_rate": 1.3877283754481633e-05, + "loss": 0.9733, + "step": 28633 + }, + { + "epoch": 1.84, + "learning_rate": 1.3877069065458684e-05, + "loss": 0.9844, + "step": 28634 + }, + { + "epoch": 1.84, + "learning_rate": 1.3876854376435733e-05, + "loss": 0.9142, + "step": 28635 + }, + { + "epoch": 1.84, + "learning_rate": 1.3876639687412784e-05, + "loss": 1.013, + "step": 28636 + }, + { + "epoch": 1.84, + "learning_rate": 1.3876424998389833e-05, + "loss": 0.9938, + "step": 28637 + }, + { + "epoch": 1.84, + "learning_rate": 1.3876210309366885e-05, + "loss": 0.9029, + "step": 28638 + }, + { + "epoch": 1.84, + "learning_rate": 1.3875995620343934e-05, + "loss": 1.081, + "step": 28639 + }, + { + "epoch": 1.84, + "learning_rate": 1.3875780931320983e-05, + "loss": 0.9372, + "step": 28640 + }, + { + "epoch": 1.84, + "learning_rate": 1.3875566242298034e-05, + "loss": 0.971, + "step": 28641 + }, + { + "epoch": 1.84, + "learning_rate": 1.3875351553275083e-05, + "loss": 0.8865, + "step": 28642 + }, + { + "epoch": 1.84, + "learning_rate": 1.3875136864252131e-05, + "loss": 0.9643, + "step": 28643 + }, + { + "epoch": 1.84, + "learning_rate": 1.3874922175229182e-05, + "loss": 0.9421, + "step": 28644 + }, + { + "epoch": 1.84, + "learning_rate": 1.3874707486206231e-05, + "loss": 0.9973, + "step": 28645 + }, + { + "epoch": 1.84, + "learning_rate": 1.3874492797183282e-05, + "loss": 0.9541, + "step": 28646 + }, + { + "epoch": 1.84, + "learning_rate": 1.387427810816033e-05, + "loss": 0.9627, + "step": 28647 + }, + { + "epoch": 1.84, + "learning_rate": 1.387406341913738e-05, + "loss": 1.004, + "step": 28648 + }, + { + "epoch": 1.84, + "learning_rate": 1.387384873011443e-05, + "loss": 0.9841, + "step": 28649 + }, + { + "epoch": 1.84, + "learning_rate": 1.387363404109148e-05, + "loss": 1.0013, + "step": 28650 + }, + { + "epoch": 1.84, + "learning_rate": 1.3873419352068528e-05, + "loss": 0.9594, + "step": 28651 + }, + { + "epoch": 1.84, + "learning_rate": 1.3873204663045579e-05, + "loss": 1.0103, + "step": 28652 + }, + { + "epoch": 1.84, + "learning_rate": 1.387298997402263e-05, + "loss": 0.9502, + "step": 28653 + }, + { + "epoch": 1.84, + "learning_rate": 1.387277528499968e-05, + "loss": 0.9827, + "step": 28654 + }, + { + "epoch": 1.84, + "learning_rate": 1.387256059597673e-05, + "loss": 0.9565, + "step": 28655 + }, + { + "epoch": 1.84, + "learning_rate": 1.387234590695378e-05, + "loss": 1.0018, + "step": 28656 + }, + { + "epoch": 1.84, + "learning_rate": 1.3872131217930829e-05, + "loss": 1.0061, + "step": 28657 + }, + { + "epoch": 1.84, + "learning_rate": 1.3871916528907878e-05, + "loss": 0.9513, + "step": 28658 + }, + { + "epoch": 1.84, + "learning_rate": 1.3871701839884928e-05, + "loss": 0.9667, + "step": 28659 + }, + { + "epoch": 1.84, + "learning_rate": 1.3871487150861977e-05, + "loss": 0.9638, + "step": 28660 + }, + { + "epoch": 1.84, + "learning_rate": 1.3871272461839026e-05, + "loss": 0.9631, + "step": 28661 + }, + { + "epoch": 1.84, + "learning_rate": 1.3871057772816077e-05, + "loss": 0.954, + "step": 28662 + }, + { + "epoch": 1.84, + "learning_rate": 1.3870843083793126e-05, + "loss": 0.9819, + "step": 28663 + }, + { + "epoch": 1.84, + "learning_rate": 1.3870628394770177e-05, + "loss": 0.9432, + "step": 28664 + }, + { + "epoch": 1.84, + "learning_rate": 1.3870413705747226e-05, + "loss": 1.0321, + "step": 28665 + }, + { + "epoch": 1.84, + "learning_rate": 1.3870199016724275e-05, + "loss": 1.0234, + "step": 28666 + }, + { + "epoch": 1.84, + "learning_rate": 1.3869984327701327e-05, + "loss": 0.9862, + "step": 28667 + }, + { + "epoch": 1.84, + "learning_rate": 1.3869769638678376e-05, + "loss": 1.037, + "step": 28668 + }, + { + "epoch": 1.84, + "learning_rate": 1.3869554949655427e-05, + "loss": 0.9575, + "step": 28669 + }, + { + "epoch": 1.84, + "learning_rate": 1.3869340260632476e-05, + "loss": 0.951, + "step": 28670 + }, + { + "epoch": 1.84, + "learning_rate": 1.3869125571609525e-05, + "loss": 0.8393, + "step": 28671 + }, + { + "epoch": 1.84, + "learning_rate": 1.3868910882586575e-05, + "loss": 0.9469, + "step": 28672 + }, + { + "epoch": 1.84, + "learning_rate": 1.3868696193563624e-05, + "loss": 1.0047, + "step": 28673 + }, + { + "epoch": 1.84, + "learning_rate": 1.3868481504540675e-05, + "loss": 0.9642, + "step": 28674 + }, + { + "epoch": 1.84, + "learning_rate": 1.3868266815517724e-05, + "loss": 0.9454, + "step": 28675 + }, + { + "epoch": 1.84, + "learning_rate": 1.3868052126494773e-05, + "loss": 1.0343, + "step": 28676 + }, + { + "epoch": 1.85, + "learning_rate": 1.3867837437471823e-05, + "loss": 0.9957, + "step": 28677 + }, + { + "epoch": 1.85, + "learning_rate": 1.3867622748448872e-05, + "loss": 0.9546, + "step": 28678 + }, + { + "epoch": 1.85, + "learning_rate": 1.3867408059425921e-05, + "loss": 0.9922, + "step": 28679 + }, + { + "epoch": 1.85, + "learning_rate": 1.3867193370402972e-05, + "loss": 1.0198, + "step": 28680 + }, + { + "epoch": 1.85, + "learning_rate": 1.3866978681380021e-05, + "loss": 0.9469, + "step": 28681 + }, + { + "epoch": 1.85, + "learning_rate": 1.3866763992357073e-05, + "loss": 0.9325, + "step": 28682 + }, + { + "epoch": 1.85, + "learning_rate": 1.3866549303334122e-05, + "loss": 0.9663, + "step": 28683 + }, + { + "epoch": 1.85, + "learning_rate": 1.3866334614311173e-05, + "loss": 1.0254, + "step": 28684 + }, + { + "epoch": 1.85, + "learning_rate": 1.3866119925288222e-05, + "loss": 0.9071, + "step": 28685 + }, + { + "epoch": 1.85, + "learning_rate": 1.386590523626527e-05, + "loss": 0.9463, + "step": 28686 + }, + { + "epoch": 1.85, + "learning_rate": 1.3865690547242321e-05, + "loss": 0.9725, + "step": 28687 + }, + { + "epoch": 1.85, + "learning_rate": 1.386547585821937e-05, + "loss": 0.9343, + "step": 28688 + }, + { + "epoch": 1.85, + "learning_rate": 1.386526116919642e-05, + "loss": 0.9927, + "step": 28689 + }, + { + "epoch": 1.85, + "learning_rate": 1.386504648017347e-05, + "loss": 0.9641, + "step": 28690 + }, + { + "epoch": 1.85, + "learning_rate": 1.3864831791150519e-05, + "loss": 0.9694, + "step": 28691 + }, + { + "epoch": 1.85, + "learning_rate": 1.3864617102127568e-05, + "loss": 0.9982, + "step": 28692 + }, + { + "epoch": 1.85, + "learning_rate": 1.3864402413104619e-05, + "loss": 0.9493, + "step": 28693 + }, + { + "epoch": 1.85, + "learning_rate": 1.3864187724081668e-05, + "loss": 1.0115, + "step": 28694 + }, + { + "epoch": 1.85, + "learning_rate": 1.3863973035058718e-05, + "loss": 1.0437, + "step": 28695 + }, + { + "epoch": 1.85, + "learning_rate": 1.3863758346035769e-05, + "loss": 0.9281, + "step": 28696 + }, + { + "epoch": 1.85, + "learning_rate": 1.386354365701282e-05, + "loss": 0.9536, + "step": 28697 + }, + { + "epoch": 1.85, + "learning_rate": 1.3863328967989869e-05, + "loss": 0.9622, + "step": 28698 + }, + { + "epoch": 1.85, + "learning_rate": 1.3863114278966918e-05, + "loss": 0.8874, + "step": 28699 + }, + { + "epoch": 1.85, + "learning_rate": 1.3862899589943968e-05, + "loss": 0.9771, + "step": 28700 + }, + { + "epoch": 1.85, + "learning_rate": 1.3862684900921017e-05, + "loss": 1.0151, + "step": 28701 + }, + { + "epoch": 1.85, + "learning_rate": 1.3862470211898066e-05, + "loss": 1.0244, + "step": 28702 + }, + { + "epoch": 1.85, + "learning_rate": 1.3862255522875117e-05, + "loss": 0.9853, + "step": 28703 + }, + { + "epoch": 1.85, + "learning_rate": 1.3862040833852166e-05, + "loss": 1.062, + "step": 28704 + }, + { + "epoch": 1.85, + "learning_rate": 1.3861826144829216e-05, + "loss": 0.9412, + "step": 28705 + }, + { + "epoch": 1.85, + "learning_rate": 1.3861611455806265e-05, + "loss": 0.9459, + "step": 28706 + }, + { + "epoch": 1.85, + "learning_rate": 1.3861396766783314e-05, + "loss": 0.95, + "step": 28707 + }, + { + "epoch": 1.85, + "learning_rate": 1.3861182077760365e-05, + "loss": 1.0612, + "step": 28708 + }, + { + "epoch": 1.85, + "learning_rate": 1.3860967388737414e-05, + "loss": 0.9916, + "step": 28709 + }, + { + "epoch": 1.85, + "learning_rate": 1.3860752699714463e-05, + "loss": 1.024, + "step": 28710 + }, + { + "epoch": 1.85, + "learning_rate": 1.3860538010691515e-05, + "loss": 0.9304, + "step": 28711 + }, + { + "epoch": 1.85, + "learning_rate": 1.3860323321668564e-05, + "loss": 1.0295, + "step": 28712 + }, + { + "epoch": 1.85, + "learning_rate": 1.3860108632645615e-05, + "loss": 0.9295, + "step": 28713 + }, + { + "epoch": 1.85, + "learning_rate": 1.3859893943622664e-05, + "loss": 0.9083, + "step": 28714 + }, + { + "epoch": 1.85, + "learning_rate": 1.3859679254599715e-05, + "loss": 0.9812, + "step": 28715 + }, + { + "epoch": 1.85, + "learning_rate": 1.3859464565576764e-05, + "loss": 0.9708, + "step": 28716 + }, + { + "epoch": 1.85, + "learning_rate": 1.3859249876553812e-05, + "loss": 0.9987, + "step": 28717 + }, + { + "epoch": 1.85, + "learning_rate": 1.3859035187530863e-05, + "loss": 0.9917, + "step": 28718 + }, + { + "epoch": 1.85, + "learning_rate": 1.3858820498507912e-05, + "loss": 0.9184, + "step": 28719 + }, + { + "epoch": 1.85, + "learning_rate": 1.3858605809484961e-05, + "loss": 0.9505, + "step": 28720 + }, + { + "epoch": 1.85, + "learning_rate": 1.3858391120462012e-05, + "loss": 1.0222, + "step": 28721 + }, + { + "epoch": 1.85, + "learning_rate": 1.385817643143906e-05, + "loss": 1.02, + "step": 28722 + }, + { + "epoch": 1.85, + "learning_rate": 1.3857961742416111e-05, + "loss": 1.0049, + "step": 28723 + }, + { + "epoch": 1.85, + "learning_rate": 1.385774705339316e-05, + "loss": 0.9765, + "step": 28724 + }, + { + "epoch": 1.85, + "learning_rate": 1.3857532364370213e-05, + "loss": 1.0311, + "step": 28725 + }, + { + "epoch": 1.85, + "learning_rate": 1.3857317675347262e-05, + "loss": 1.0165, + "step": 28726 + }, + { + "epoch": 1.85, + "learning_rate": 1.385710298632431e-05, + "loss": 0.9792, + "step": 28727 + }, + { + "epoch": 1.85, + "learning_rate": 1.3856888297301361e-05, + "loss": 1.0766, + "step": 28728 + }, + { + "epoch": 1.85, + "learning_rate": 1.385667360827841e-05, + "loss": 0.9712, + "step": 28729 + }, + { + "epoch": 1.85, + "learning_rate": 1.385645891925546e-05, + "loss": 0.961, + "step": 28730 + }, + { + "epoch": 1.85, + "learning_rate": 1.385624423023251e-05, + "loss": 0.9025, + "step": 28731 + }, + { + "epoch": 1.85, + "learning_rate": 1.3856029541209559e-05, + "loss": 1.0201, + "step": 28732 + }, + { + "epoch": 1.85, + "learning_rate": 1.385581485218661e-05, + "loss": 0.9453, + "step": 28733 + }, + { + "epoch": 1.85, + "learning_rate": 1.3855600163163658e-05, + "loss": 0.951, + "step": 28734 + }, + { + "epoch": 1.85, + "learning_rate": 1.3855385474140707e-05, + "loss": 1.0198, + "step": 28735 + }, + { + "epoch": 1.85, + "learning_rate": 1.3855170785117758e-05, + "loss": 0.9525, + "step": 28736 + }, + { + "epoch": 1.85, + "learning_rate": 1.3854956096094807e-05, + "loss": 0.9183, + "step": 28737 + }, + { + "epoch": 1.85, + "learning_rate": 1.3854741407071856e-05, + "loss": 1.0067, + "step": 28738 + }, + { + "epoch": 1.85, + "learning_rate": 1.3854526718048907e-05, + "loss": 0.9439, + "step": 28739 + }, + { + "epoch": 1.85, + "learning_rate": 1.3854312029025957e-05, + "loss": 0.899, + "step": 28740 + }, + { + "epoch": 1.85, + "learning_rate": 1.3854097340003008e-05, + "loss": 0.919, + "step": 28741 + }, + { + "epoch": 1.85, + "learning_rate": 1.3853882650980057e-05, + "loss": 0.9454, + "step": 28742 + }, + { + "epoch": 1.85, + "learning_rate": 1.3853667961957108e-05, + "loss": 1.0051, + "step": 28743 + }, + { + "epoch": 1.85, + "learning_rate": 1.3853453272934157e-05, + "loss": 0.949, + "step": 28744 + }, + { + "epoch": 1.85, + "learning_rate": 1.3853238583911206e-05, + "loss": 0.959, + "step": 28745 + }, + { + "epoch": 1.85, + "learning_rate": 1.3853023894888256e-05, + "loss": 0.9683, + "step": 28746 + }, + { + "epoch": 1.85, + "learning_rate": 1.3852809205865305e-05, + "loss": 0.967, + "step": 28747 + }, + { + "epoch": 1.85, + "learning_rate": 1.3852594516842354e-05, + "loss": 0.9449, + "step": 28748 + }, + { + "epoch": 1.85, + "learning_rate": 1.3852379827819405e-05, + "loss": 0.9478, + "step": 28749 + }, + { + "epoch": 1.85, + "learning_rate": 1.3852165138796454e-05, + "loss": 1.0033, + "step": 28750 + }, + { + "epoch": 1.85, + "learning_rate": 1.3851950449773504e-05, + "loss": 0.9494, + "step": 28751 + }, + { + "epoch": 1.85, + "learning_rate": 1.3851735760750553e-05, + "loss": 0.9758, + "step": 28752 + }, + { + "epoch": 1.85, + "learning_rate": 1.3851521071727602e-05, + "loss": 0.9544, + "step": 28753 + }, + { + "epoch": 1.85, + "learning_rate": 1.3851306382704655e-05, + "loss": 0.9354, + "step": 28754 + }, + { + "epoch": 1.85, + "learning_rate": 1.3851091693681704e-05, + "loss": 1.0346, + "step": 28755 + }, + { + "epoch": 1.85, + "learning_rate": 1.3850877004658754e-05, + "loss": 0.9055, + "step": 28756 + }, + { + "epoch": 1.85, + "learning_rate": 1.3850662315635803e-05, + "loss": 1.0167, + "step": 28757 + }, + { + "epoch": 1.85, + "learning_rate": 1.3850447626612852e-05, + "loss": 0.9103, + "step": 28758 + }, + { + "epoch": 1.85, + "learning_rate": 1.3850232937589903e-05, + "loss": 0.9887, + "step": 28759 + }, + { + "epoch": 1.85, + "learning_rate": 1.3850018248566952e-05, + "loss": 0.9404, + "step": 28760 + }, + { + "epoch": 1.85, + "learning_rate": 1.3849803559544003e-05, + "loss": 0.9676, + "step": 28761 + }, + { + "epoch": 1.85, + "learning_rate": 1.3849588870521051e-05, + "loss": 0.9824, + "step": 28762 + }, + { + "epoch": 1.85, + "learning_rate": 1.38493741814981e-05, + "loss": 1.0166, + "step": 28763 + }, + { + "epoch": 1.85, + "learning_rate": 1.3849159492475151e-05, + "loss": 0.9188, + "step": 28764 + }, + { + "epoch": 1.85, + "learning_rate": 1.38489448034522e-05, + "loss": 0.989, + "step": 28765 + }, + { + "epoch": 1.85, + "learning_rate": 1.3848730114429249e-05, + "loss": 1.031, + "step": 28766 + }, + { + "epoch": 1.85, + "learning_rate": 1.38485154254063e-05, + "loss": 1.0606, + "step": 28767 + }, + { + "epoch": 1.85, + "learning_rate": 1.3848300736383349e-05, + "loss": 0.949, + "step": 28768 + }, + { + "epoch": 1.85, + "learning_rate": 1.3848086047360401e-05, + "loss": 0.9657, + "step": 28769 + }, + { + "epoch": 1.85, + "learning_rate": 1.384787135833745e-05, + "loss": 0.897, + "step": 28770 + }, + { + "epoch": 1.85, + "learning_rate": 1.38476566693145e-05, + "loss": 0.9798, + "step": 28771 + }, + { + "epoch": 1.85, + "learning_rate": 1.384744198029155e-05, + "loss": 0.9869, + "step": 28772 + }, + { + "epoch": 1.85, + "learning_rate": 1.3847227291268599e-05, + "loss": 0.936, + "step": 28773 + }, + { + "epoch": 1.85, + "learning_rate": 1.384701260224565e-05, + "loss": 1.0217, + "step": 28774 + }, + { + "epoch": 1.85, + "learning_rate": 1.3846797913222698e-05, + "loss": 0.9502, + "step": 28775 + }, + { + "epoch": 1.85, + "learning_rate": 1.3846583224199747e-05, + "loss": 0.9923, + "step": 28776 + }, + { + "epoch": 1.85, + "learning_rate": 1.3846368535176798e-05, + "loss": 0.9326, + "step": 28777 + }, + { + "epoch": 1.85, + "learning_rate": 1.3846153846153847e-05, + "loss": 0.9856, + "step": 28778 + }, + { + "epoch": 1.85, + "learning_rate": 1.3845939157130896e-05, + "loss": 1.008, + "step": 28779 + }, + { + "epoch": 1.85, + "learning_rate": 1.3845724468107946e-05, + "loss": 0.9574, + "step": 28780 + }, + { + "epoch": 1.85, + "learning_rate": 1.3845509779084995e-05, + "loss": 0.9872, + "step": 28781 + }, + { + "epoch": 1.85, + "learning_rate": 1.3845295090062046e-05, + "loss": 0.9458, + "step": 28782 + }, + { + "epoch": 1.85, + "learning_rate": 1.3845080401039095e-05, + "loss": 0.9781, + "step": 28783 + }, + { + "epoch": 1.85, + "learning_rate": 1.3844865712016147e-05, + "loss": 0.9612, + "step": 28784 + }, + { + "epoch": 1.85, + "learning_rate": 1.3844651022993196e-05, + "loss": 0.9569, + "step": 28785 + }, + { + "epoch": 1.85, + "learning_rate": 1.3844436333970245e-05, + "loss": 0.9652, + "step": 28786 + }, + { + "epoch": 1.85, + "learning_rate": 1.3844221644947296e-05, + "loss": 0.982, + "step": 28787 + }, + { + "epoch": 1.85, + "learning_rate": 1.3844006955924345e-05, + "loss": 0.9663, + "step": 28788 + }, + { + "epoch": 1.85, + "learning_rate": 1.3843792266901394e-05, + "loss": 0.992, + "step": 28789 + }, + { + "epoch": 1.85, + "learning_rate": 1.3843577577878445e-05, + "loss": 0.963, + "step": 28790 + }, + { + "epoch": 1.85, + "learning_rate": 1.3843362888855493e-05, + "loss": 0.9163, + "step": 28791 + }, + { + "epoch": 1.85, + "learning_rate": 1.3843148199832544e-05, + "loss": 0.9172, + "step": 28792 + }, + { + "epoch": 1.85, + "learning_rate": 1.3842933510809593e-05, + "loss": 1.0502, + "step": 28793 + }, + { + "epoch": 1.85, + "learning_rate": 1.3842718821786642e-05, + "loss": 0.9893, + "step": 28794 + }, + { + "epoch": 1.85, + "learning_rate": 1.3842504132763693e-05, + "loss": 1.0058, + "step": 28795 + }, + { + "epoch": 1.85, + "learning_rate": 1.3842289443740742e-05, + "loss": 0.9571, + "step": 28796 + }, + { + "epoch": 1.85, + "learning_rate": 1.384207475471779e-05, + "loss": 0.978, + "step": 28797 + }, + { + "epoch": 1.85, + "learning_rate": 1.3841860065694843e-05, + "loss": 0.9098, + "step": 28798 + }, + { + "epoch": 1.85, + "learning_rate": 1.3841645376671892e-05, + "loss": 0.9145, + "step": 28799 + }, + { + "epoch": 1.85, + "learning_rate": 1.3841430687648943e-05, + "loss": 0.9521, + "step": 28800 + }, + { + "epoch": 1.85, + "learning_rate": 1.3841215998625992e-05, + "loss": 0.9512, + "step": 28801 + }, + { + "epoch": 1.85, + "learning_rate": 1.3841001309603042e-05, + "loss": 0.9632, + "step": 28802 + }, + { + "epoch": 1.85, + "learning_rate": 1.3840786620580091e-05, + "loss": 0.9279, + "step": 28803 + }, + { + "epoch": 1.85, + "learning_rate": 1.384057193155714e-05, + "loss": 0.8991, + "step": 28804 + }, + { + "epoch": 1.85, + "learning_rate": 1.384035724253419e-05, + "loss": 0.9988, + "step": 28805 + }, + { + "epoch": 1.85, + "learning_rate": 1.384014255351124e-05, + "loss": 1.0333, + "step": 28806 + }, + { + "epoch": 1.85, + "learning_rate": 1.3839927864488289e-05, + "loss": 0.9545, + "step": 28807 + }, + { + "epoch": 1.85, + "learning_rate": 1.383971317546534e-05, + "loss": 1.0122, + "step": 28808 + }, + { + "epoch": 1.85, + "learning_rate": 1.3839498486442388e-05, + "loss": 0.9362, + "step": 28809 + }, + { + "epoch": 1.85, + "learning_rate": 1.3839283797419439e-05, + "loss": 1.0574, + "step": 28810 + }, + { + "epoch": 1.85, + "learning_rate": 1.3839069108396488e-05, + "loss": 0.9922, + "step": 28811 + }, + { + "epoch": 1.85, + "learning_rate": 1.3838854419373537e-05, + "loss": 0.9489, + "step": 28812 + }, + { + "epoch": 1.85, + "learning_rate": 1.383863973035059e-05, + "loss": 0.9339, + "step": 28813 + }, + { + "epoch": 1.85, + "learning_rate": 1.3838425041327638e-05, + "loss": 0.9931, + "step": 28814 + }, + { + "epoch": 1.85, + "learning_rate": 1.3838210352304689e-05, + "loss": 0.9556, + "step": 28815 + }, + { + "epoch": 1.85, + "learning_rate": 1.3837995663281738e-05, + "loss": 0.9981, + "step": 28816 + }, + { + "epoch": 1.85, + "learning_rate": 1.3837780974258787e-05, + "loss": 1.0176, + "step": 28817 + }, + { + "epoch": 1.85, + "learning_rate": 1.3837566285235838e-05, + "loss": 0.9055, + "step": 28818 + }, + { + "epoch": 1.85, + "learning_rate": 1.3837351596212887e-05, + "loss": 0.9398, + "step": 28819 + }, + { + "epoch": 1.85, + "learning_rate": 1.3837136907189937e-05, + "loss": 0.9566, + "step": 28820 + }, + { + "epoch": 1.85, + "learning_rate": 1.3836922218166986e-05, + "loss": 0.9964, + "step": 28821 + }, + { + "epoch": 1.85, + "learning_rate": 1.3836707529144035e-05, + "loss": 0.9317, + "step": 28822 + }, + { + "epoch": 1.85, + "learning_rate": 1.3836492840121086e-05, + "loss": 0.9208, + "step": 28823 + }, + { + "epoch": 1.85, + "learning_rate": 1.3836278151098135e-05, + "loss": 0.9125, + "step": 28824 + }, + { + "epoch": 1.85, + "learning_rate": 1.3836063462075184e-05, + "loss": 0.9693, + "step": 28825 + }, + { + "epoch": 1.85, + "learning_rate": 1.3835848773052234e-05, + "loss": 0.9209, + "step": 28826 + }, + { + "epoch": 1.85, + "learning_rate": 1.3835634084029285e-05, + "loss": 0.9843, + "step": 28827 + }, + { + "epoch": 1.85, + "learning_rate": 1.3835419395006336e-05, + "loss": 1.0393, + "step": 28828 + }, + { + "epoch": 1.85, + "learning_rate": 1.3835204705983385e-05, + "loss": 0.9261, + "step": 28829 + }, + { + "epoch": 1.85, + "learning_rate": 1.3834990016960435e-05, + "loss": 1.001, + "step": 28830 + }, + { + "epoch": 1.85, + "learning_rate": 1.3834775327937484e-05, + "loss": 0.91, + "step": 28831 + }, + { + "epoch": 1.85, + "learning_rate": 1.3834560638914533e-05, + "loss": 0.9129, + "step": 28832 + }, + { + "epoch": 1.86, + "learning_rate": 1.3834345949891584e-05, + "loss": 0.9632, + "step": 28833 + }, + { + "epoch": 1.86, + "learning_rate": 1.3834131260868633e-05, + "loss": 0.9774, + "step": 28834 + }, + { + "epoch": 1.86, + "learning_rate": 1.3833916571845682e-05, + "loss": 0.9615, + "step": 28835 + }, + { + "epoch": 1.86, + "learning_rate": 1.3833701882822732e-05, + "loss": 0.9269, + "step": 28836 + }, + { + "epoch": 1.86, + "learning_rate": 1.3833487193799781e-05, + "loss": 0.9783, + "step": 28837 + }, + { + "epoch": 1.86, + "learning_rate": 1.3833272504776832e-05, + "loss": 0.9266, + "step": 28838 + }, + { + "epoch": 1.86, + "learning_rate": 1.3833057815753881e-05, + "loss": 1.0264, + "step": 28839 + }, + { + "epoch": 1.86, + "learning_rate": 1.383284312673093e-05, + "loss": 0.927, + "step": 28840 + }, + { + "epoch": 1.86, + "learning_rate": 1.383262843770798e-05, + "loss": 0.9537, + "step": 28841 + }, + { + "epoch": 1.86, + "learning_rate": 1.3832413748685031e-05, + "loss": 0.9867, + "step": 28842 + }, + { + "epoch": 1.86, + "learning_rate": 1.3832199059662082e-05, + "loss": 1.0095, + "step": 28843 + }, + { + "epoch": 1.86, + "learning_rate": 1.3831984370639131e-05, + "loss": 0.9474, + "step": 28844 + }, + { + "epoch": 1.86, + "learning_rate": 1.383176968161618e-05, + "loss": 1.0291, + "step": 28845 + }, + { + "epoch": 1.86, + "learning_rate": 1.383155499259323e-05, + "loss": 0.9581, + "step": 28846 + }, + { + "epoch": 1.86, + "learning_rate": 1.383134030357028e-05, + "loss": 0.96, + "step": 28847 + }, + { + "epoch": 1.86, + "learning_rate": 1.383112561454733e-05, + "loss": 0.9221, + "step": 28848 + }, + { + "epoch": 1.86, + "learning_rate": 1.383091092552438e-05, + "loss": 0.9639, + "step": 28849 + }, + { + "epoch": 1.86, + "learning_rate": 1.3830696236501428e-05, + "loss": 0.9731, + "step": 28850 + }, + { + "epoch": 1.86, + "learning_rate": 1.3830481547478479e-05, + "loss": 0.9931, + "step": 28851 + }, + { + "epoch": 1.86, + "learning_rate": 1.3830266858455528e-05, + "loss": 0.9848, + "step": 28852 + }, + { + "epoch": 1.86, + "learning_rate": 1.3830052169432577e-05, + "loss": 0.9172, + "step": 28853 + }, + { + "epoch": 1.86, + "learning_rate": 1.3829837480409627e-05, + "loss": 0.9157, + "step": 28854 + }, + { + "epoch": 1.86, + "learning_rate": 1.3829622791386676e-05, + "loss": 0.9919, + "step": 28855 + }, + { + "epoch": 1.86, + "learning_rate": 1.3829408102363729e-05, + "loss": 0.9892, + "step": 28856 + }, + { + "epoch": 1.86, + "learning_rate": 1.3829193413340778e-05, + "loss": 0.954, + "step": 28857 + }, + { + "epoch": 1.86, + "learning_rate": 1.3828978724317828e-05, + "loss": 0.9667, + "step": 28858 + }, + { + "epoch": 1.86, + "learning_rate": 1.3828764035294877e-05, + "loss": 0.9699, + "step": 28859 + }, + { + "epoch": 1.86, + "learning_rate": 1.3828549346271926e-05, + "loss": 0.946, + "step": 28860 + }, + { + "epoch": 1.86, + "learning_rate": 1.3828334657248977e-05, + "loss": 0.9335, + "step": 28861 + }, + { + "epoch": 1.86, + "learning_rate": 1.3828119968226026e-05, + "loss": 1.0163, + "step": 28862 + }, + { + "epoch": 1.86, + "learning_rate": 1.3827905279203075e-05, + "loss": 1.0039, + "step": 28863 + }, + { + "epoch": 1.86, + "learning_rate": 1.3827690590180126e-05, + "loss": 1.0016, + "step": 28864 + }, + { + "epoch": 1.86, + "learning_rate": 1.3827475901157174e-05, + "loss": 1.0057, + "step": 28865 + }, + { + "epoch": 1.86, + "learning_rate": 1.3827261212134223e-05, + "loss": 0.9748, + "step": 28866 + }, + { + "epoch": 1.86, + "learning_rate": 1.3827046523111274e-05, + "loss": 0.9591, + "step": 28867 + }, + { + "epoch": 1.86, + "learning_rate": 1.3826831834088323e-05, + "loss": 0.9492, + "step": 28868 + }, + { + "epoch": 1.86, + "learning_rate": 1.3826617145065374e-05, + "loss": 0.9965, + "step": 28869 + }, + { + "epoch": 1.86, + "learning_rate": 1.3826402456042423e-05, + "loss": 0.9793, + "step": 28870 + }, + { + "epoch": 1.86, + "learning_rate": 1.3826187767019475e-05, + "loss": 0.9938, + "step": 28871 + }, + { + "epoch": 1.86, + "learning_rate": 1.3825973077996524e-05, + "loss": 0.9837, + "step": 28872 + }, + { + "epoch": 1.86, + "learning_rate": 1.3825758388973573e-05, + "loss": 0.9623, + "step": 28873 + }, + { + "epoch": 1.86, + "learning_rate": 1.3825543699950624e-05, + "loss": 0.9441, + "step": 28874 + }, + { + "epoch": 1.86, + "learning_rate": 1.3825329010927673e-05, + "loss": 0.999, + "step": 28875 + }, + { + "epoch": 1.86, + "learning_rate": 1.3825114321904722e-05, + "loss": 0.9891, + "step": 28876 + }, + { + "epoch": 1.86, + "learning_rate": 1.3824899632881772e-05, + "loss": 0.975, + "step": 28877 + }, + { + "epoch": 1.86, + "learning_rate": 1.3824684943858821e-05, + "loss": 0.9992, + "step": 28878 + }, + { + "epoch": 1.86, + "learning_rate": 1.3824470254835872e-05, + "loss": 1.0025, + "step": 28879 + }, + { + "epoch": 1.86, + "learning_rate": 1.382425556581292e-05, + "loss": 0.9207, + "step": 28880 + }, + { + "epoch": 1.86, + "learning_rate": 1.382404087678997e-05, + "loss": 0.9721, + "step": 28881 + }, + { + "epoch": 1.86, + "learning_rate": 1.382382618776702e-05, + "loss": 0.949, + "step": 28882 + }, + { + "epoch": 1.86, + "learning_rate": 1.382361149874407e-05, + "loss": 1.0006, + "step": 28883 + }, + { + "epoch": 1.86, + "learning_rate": 1.3823396809721118e-05, + "loss": 0.9685, + "step": 28884 + }, + { + "epoch": 1.86, + "learning_rate": 1.382318212069817e-05, + "loss": 0.9631, + "step": 28885 + }, + { + "epoch": 1.86, + "learning_rate": 1.382296743167522e-05, + "loss": 0.956, + "step": 28886 + }, + { + "epoch": 1.86, + "learning_rate": 1.382275274265227e-05, + "loss": 0.9106, + "step": 28887 + }, + { + "epoch": 1.86, + "learning_rate": 1.382253805362932e-05, + "loss": 0.932, + "step": 28888 + }, + { + "epoch": 1.86, + "learning_rate": 1.382232336460637e-05, + "loss": 1.0431, + "step": 28889 + }, + { + "epoch": 1.86, + "learning_rate": 1.3822108675583419e-05, + "loss": 0.9718, + "step": 28890 + }, + { + "epoch": 1.86, + "learning_rate": 1.3821893986560468e-05, + "loss": 1.0113, + "step": 28891 + }, + { + "epoch": 1.86, + "learning_rate": 1.3821679297537519e-05, + "loss": 0.9725, + "step": 28892 + }, + { + "epoch": 1.86, + "learning_rate": 1.3821464608514568e-05, + "loss": 0.8782, + "step": 28893 + }, + { + "epoch": 1.86, + "learning_rate": 1.3821249919491616e-05, + "loss": 0.8983, + "step": 28894 + }, + { + "epoch": 1.86, + "learning_rate": 1.3821035230468667e-05, + "loss": 0.9358, + "step": 28895 + }, + { + "epoch": 1.86, + "learning_rate": 1.3820820541445716e-05, + "loss": 0.9429, + "step": 28896 + }, + { + "epoch": 1.86, + "learning_rate": 1.3820605852422767e-05, + "loss": 1.0162, + "step": 28897 + }, + { + "epoch": 1.86, + "learning_rate": 1.3820391163399816e-05, + "loss": 0.9839, + "step": 28898 + }, + { + "epoch": 1.86, + "learning_rate": 1.3820176474376865e-05, + "loss": 0.9899, + "step": 28899 + }, + { + "epoch": 1.86, + "learning_rate": 1.3819961785353917e-05, + "loss": 0.9028, + "step": 28900 + }, + { + "epoch": 1.86, + "learning_rate": 1.3819747096330966e-05, + "loss": 0.9946, + "step": 28901 + }, + { + "epoch": 1.86, + "learning_rate": 1.3819532407308017e-05, + "loss": 0.9864, + "step": 28902 + }, + { + "epoch": 1.86, + "learning_rate": 1.3819317718285066e-05, + "loss": 0.9426, + "step": 28903 + }, + { + "epoch": 1.86, + "learning_rate": 1.3819103029262115e-05, + "loss": 0.9648, + "step": 28904 + }, + { + "epoch": 1.86, + "learning_rate": 1.3818888340239165e-05, + "loss": 1.0761, + "step": 28905 + }, + { + "epoch": 1.86, + "learning_rate": 1.3818673651216214e-05, + "loss": 0.984, + "step": 28906 + }, + { + "epoch": 1.86, + "learning_rate": 1.3818458962193265e-05, + "loss": 0.9481, + "step": 28907 + }, + { + "epoch": 1.86, + "learning_rate": 1.3818244273170314e-05, + "loss": 1.0566, + "step": 28908 + }, + { + "epoch": 1.86, + "learning_rate": 1.3818029584147363e-05, + "loss": 0.9082, + "step": 28909 + }, + { + "epoch": 1.86, + "learning_rate": 1.3817814895124413e-05, + "loss": 1.0616, + "step": 28910 + }, + { + "epoch": 1.86, + "learning_rate": 1.3817600206101462e-05, + "loss": 0.9512, + "step": 28911 + }, + { + "epoch": 1.86, + "learning_rate": 1.3817385517078511e-05, + "loss": 0.9671, + "step": 28912 + }, + { + "epoch": 1.86, + "learning_rate": 1.3817170828055562e-05, + "loss": 0.9052, + "step": 28913 + }, + { + "epoch": 1.86, + "learning_rate": 1.3816956139032613e-05, + "loss": 1.0148, + "step": 28914 + }, + { + "epoch": 1.86, + "learning_rate": 1.3816741450009663e-05, + "loss": 0.949, + "step": 28915 + }, + { + "epoch": 1.86, + "learning_rate": 1.3816526760986712e-05, + "loss": 0.9751, + "step": 28916 + }, + { + "epoch": 1.86, + "learning_rate": 1.3816312071963763e-05, + "loss": 0.9484, + "step": 28917 + }, + { + "epoch": 1.86, + "learning_rate": 1.3816097382940812e-05, + "loss": 1.0207, + "step": 28918 + }, + { + "epoch": 1.86, + "learning_rate": 1.3815882693917861e-05, + "loss": 0.9543, + "step": 28919 + }, + { + "epoch": 1.86, + "learning_rate": 1.3815668004894912e-05, + "loss": 0.9757, + "step": 28920 + }, + { + "epoch": 1.86, + "learning_rate": 1.381545331587196e-05, + "loss": 0.958, + "step": 28921 + }, + { + "epoch": 1.86, + "learning_rate": 1.381523862684901e-05, + "loss": 1.0051, + "step": 28922 + }, + { + "epoch": 1.86, + "learning_rate": 1.381502393782606e-05, + "loss": 0.9665, + "step": 28923 + }, + { + "epoch": 1.86, + "learning_rate": 1.3814809248803109e-05, + "loss": 1.0153, + "step": 28924 + }, + { + "epoch": 1.86, + "learning_rate": 1.3814594559780158e-05, + "loss": 0.9633, + "step": 28925 + }, + { + "epoch": 1.86, + "learning_rate": 1.3814379870757209e-05, + "loss": 0.9562, + "step": 28926 + }, + { + "epoch": 1.86, + "learning_rate": 1.3814165181734258e-05, + "loss": 0.9769, + "step": 28927 + }, + { + "epoch": 1.86, + "learning_rate": 1.3813950492711308e-05, + "loss": 0.9209, + "step": 28928 + }, + { + "epoch": 1.86, + "learning_rate": 1.3813735803688359e-05, + "loss": 0.9694, + "step": 28929 + }, + { + "epoch": 1.86, + "learning_rate": 1.381352111466541e-05, + "loss": 0.9818, + "step": 28930 + }, + { + "epoch": 1.86, + "learning_rate": 1.3813306425642459e-05, + "loss": 1.0051, + "step": 28931 + }, + { + "epoch": 1.86, + "learning_rate": 1.3813091736619508e-05, + "loss": 0.9544, + "step": 28932 + }, + { + "epoch": 1.86, + "learning_rate": 1.3812877047596558e-05, + "loss": 1.0098, + "step": 28933 + }, + { + "epoch": 1.86, + "learning_rate": 1.3812662358573607e-05, + "loss": 1.0069, + "step": 28934 + }, + { + "epoch": 1.86, + "learning_rate": 1.3812447669550656e-05, + "loss": 0.9905, + "step": 28935 + }, + { + "epoch": 1.86, + "learning_rate": 1.3812232980527707e-05, + "loss": 0.9763, + "step": 28936 + }, + { + "epoch": 1.86, + "learning_rate": 1.3812018291504756e-05, + "loss": 0.9637, + "step": 28937 + }, + { + "epoch": 1.86, + "learning_rate": 1.3811803602481807e-05, + "loss": 0.9492, + "step": 28938 + }, + { + "epoch": 1.86, + "learning_rate": 1.3811588913458855e-05, + "loss": 0.9504, + "step": 28939 + }, + { + "epoch": 1.86, + "learning_rate": 1.3811374224435904e-05, + "loss": 1.0073, + "step": 28940 + }, + { + "epoch": 1.86, + "learning_rate": 1.3811159535412955e-05, + "loss": 0.9019, + "step": 28941 + }, + { + "epoch": 1.86, + "learning_rate": 1.3810944846390004e-05, + "loss": 0.9711, + "step": 28942 + }, + { + "epoch": 1.86, + "learning_rate": 1.3810730157367056e-05, + "loss": 0.9562, + "step": 28943 + }, + { + "epoch": 1.86, + "learning_rate": 1.3810515468344105e-05, + "loss": 0.9661, + "step": 28944 + }, + { + "epoch": 1.86, + "learning_rate": 1.3810300779321156e-05, + "loss": 0.8881, + "step": 28945 + }, + { + "epoch": 1.86, + "learning_rate": 1.3810086090298205e-05, + "loss": 0.9504, + "step": 28946 + }, + { + "epoch": 1.86, + "learning_rate": 1.3809871401275254e-05, + "loss": 1.0005, + "step": 28947 + }, + { + "epoch": 1.86, + "learning_rate": 1.3809656712252305e-05, + "loss": 1.0297, + "step": 28948 + }, + { + "epoch": 1.86, + "learning_rate": 1.3809442023229354e-05, + "loss": 0.9912, + "step": 28949 + }, + { + "epoch": 1.86, + "learning_rate": 1.3809227334206403e-05, + "loss": 0.9252, + "step": 28950 + }, + { + "epoch": 1.86, + "learning_rate": 1.3809012645183453e-05, + "loss": 1.0149, + "step": 28951 + }, + { + "epoch": 1.86, + "learning_rate": 1.3808797956160502e-05, + "loss": 0.9569, + "step": 28952 + }, + { + "epoch": 1.86, + "learning_rate": 1.3808583267137551e-05, + "loss": 0.9035, + "step": 28953 + }, + { + "epoch": 1.86, + "learning_rate": 1.3808368578114602e-05, + "loss": 0.957, + "step": 28954 + }, + { + "epoch": 1.86, + "learning_rate": 1.380815388909165e-05, + "loss": 0.9088, + "step": 28955 + }, + { + "epoch": 1.86, + "learning_rate": 1.3807939200068701e-05, + "loss": 1.0003, + "step": 28956 + }, + { + "epoch": 1.86, + "learning_rate": 1.380772451104575e-05, + "loss": 0.9337, + "step": 28957 + }, + { + "epoch": 1.86, + "learning_rate": 1.3807509822022803e-05, + "loss": 0.9112, + "step": 28958 + }, + { + "epoch": 1.86, + "learning_rate": 1.3807295132999852e-05, + "loss": 0.9699, + "step": 28959 + }, + { + "epoch": 1.86, + "learning_rate": 1.38070804439769e-05, + "loss": 0.9025, + "step": 28960 + }, + { + "epoch": 1.86, + "learning_rate": 1.3806865754953951e-05, + "loss": 0.969, + "step": 28961 + }, + { + "epoch": 1.86, + "learning_rate": 1.3806651065931e-05, + "loss": 0.951, + "step": 28962 + }, + { + "epoch": 1.86, + "learning_rate": 1.380643637690805e-05, + "loss": 0.9117, + "step": 28963 + }, + { + "epoch": 1.86, + "learning_rate": 1.38062216878851e-05, + "loss": 1.0129, + "step": 28964 + }, + { + "epoch": 1.86, + "learning_rate": 1.3806006998862149e-05, + "loss": 0.9791, + "step": 28965 + }, + { + "epoch": 1.86, + "learning_rate": 1.38057923098392e-05, + "loss": 1.0095, + "step": 28966 + }, + { + "epoch": 1.86, + "learning_rate": 1.3805577620816249e-05, + "loss": 0.9588, + "step": 28967 + }, + { + "epoch": 1.86, + "learning_rate": 1.3805362931793297e-05, + "loss": 0.9786, + "step": 28968 + }, + { + "epoch": 1.86, + "learning_rate": 1.3805148242770348e-05, + "loss": 0.9623, + "step": 28969 + }, + { + "epoch": 1.86, + "learning_rate": 1.3804933553747397e-05, + "loss": 1.017, + "step": 28970 + }, + { + "epoch": 1.86, + "learning_rate": 1.3804718864724446e-05, + "loss": 0.939, + "step": 28971 + }, + { + "epoch": 1.86, + "learning_rate": 1.3804504175701498e-05, + "loss": 0.9968, + "step": 28972 + }, + { + "epoch": 1.86, + "learning_rate": 1.3804289486678547e-05, + "loss": 0.9768, + "step": 28973 + }, + { + "epoch": 1.86, + "learning_rate": 1.3804074797655598e-05, + "loss": 1.0355, + "step": 28974 + }, + { + "epoch": 1.86, + "learning_rate": 1.3803860108632647e-05, + "loss": 0.9711, + "step": 28975 + }, + { + "epoch": 1.86, + "learning_rate": 1.3803645419609698e-05, + "loss": 1.0006, + "step": 28976 + }, + { + "epoch": 1.86, + "learning_rate": 1.3803430730586747e-05, + "loss": 0.922, + "step": 28977 + }, + { + "epoch": 1.86, + "learning_rate": 1.3803216041563796e-05, + "loss": 0.9318, + "step": 28978 + }, + { + "epoch": 1.86, + "learning_rate": 1.3803001352540846e-05, + "loss": 1.0111, + "step": 28979 + }, + { + "epoch": 1.86, + "learning_rate": 1.3802786663517895e-05, + "loss": 0.9023, + "step": 28980 + }, + { + "epoch": 1.86, + "learning_rate": 1.3802571974494944e-05, + "loss": 0.9809, + "step": 28981 + }, + { + "epoch": 1.86, + "learning_rate": 1.3802357285471995e-05, + "loss": 1.0348, + "step": 28982 + }, + { + "epoch": 1.86, + "learning_rate": 1.3802142596449044e-05, + "loss": 0.9859, + "step": 28983 + }, + { + "epoch": 1.86, + "learning_rate": 1.3801927907426094e-05, + "loss": 0.9207, + "step": 28984 + }, + { + "epoch": 1.86, + "learning_rate": 1.3801713218403143e-05, + "loss": 0.9911, + "step": 28985 + }, + { + "epoch": 1.86, + "learning_rate": 1.3801498529380192e-05, + "loss": 0.9518, + "step": 28986 + }, + { + "epoch": 1.86, + "learning_rate": 1.3801283840357245e-05, + "loss": 0.9566, + "step": 28987 + }, + { + "epoch": 1.87, + "learning_rate": 1.3801069151334294e-05, + "loss": 0.9505, + "step": 28988 + }, + { + "epoch": 1.87, + "learning_rate": 1.3800854462311344e-05, + "loss": 1.0876, + "step": 28989 + }, + { + "epoch": 1.87, + "learning_rate": 1.3800639773288393e-05, + "loss": 0.967, + "step": 28990 + }, + { + "epoch": 1.87, + "learning_rate": 1.3800425084265442e-05, + "loss": 0.9414, + "step": 28991 + }, + { + "epoch": 1.87, + "learning_rate": 1.3800210395242493e-05, + "loss": 0.9661, + "step": 28992 + }, + { + "epoch": 1.87, + "learning_rate": 1.3799995706219542e-05, + "loss": 1.0038, + "step": 28993 + }, + { + "epoch": 1.87, + "learning_rate": 1.3799781017196593e-05, + "loss": 0.9816, + "step": 28994 + }, + { + "epoch": 1.87, + "learning_rate": 1.3799566328173642e-05, + "loss": 1.0401, + "step": 28995 + }, + { + "epoch": 1.87, + "learning_rate": 1.379935163915069e-05, + "loss": 0.9553, + "step": 28996 + }, + { + "epoch": 1.87, + "learning_rate": 1.3799136950127741e-05, + "loss": 0.9499, + "step": 28997 + }, + { + "epoch": 1.87, + "learning_rate": 1.379892226110479e-05, + "loss": 0.9923, + "step": 28998 + }, + { + "epoch": 1.87, + "learning_rate": 1.3798707572081839e-05, + "loss": 0.9403, + "step": 28999 + }, + { + "epoch": 1.87, + "learning_rate": 1.379849288305889e-05, + "loss": 0.955, + "step": 29000 + }, + { + "epoch": 1.87, + "eval_laion/cider": 0.5504673159986768, + "eval_laion/meteor": 0.38647651556263074, + "eval_laion/pred_num_words": 36.009765625, + "eval_laion/sacrebleu": 16.045558017954928, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0436437129974365, + "eval_runtime": 99.4636, + "eval_samples_per_second": 5.017, + "eval_steps_per_second": 0.322, + "step": 29000 + }, + { + "epoch": 1.87, + "learning_rate": 1.379827819403594e-05, + "loss": 0.9968, + "step": 29001 + }, + { + "epoch": 1.87, + "learning_rate": 1.3798063505012991e-05, + "loss": 0.9572, + "step": 29002 + }, + { + "epoch": 1.87, + "learning_rate": 1.379784881599004e-05, + "loss": 0.9856, + "step": 29003 + }, + { + "epoch": 1.87, + "learning_rate": 1.379763412696709e-05, + "loss": 1.0135, + "step": 29004 + }, + { + "epoch": 1.87, + "learning_rate": 1.379741943794414e-05, + "loss": 0.9387, + "step": 29005 + }, + { + "epoch": 1.87, + "learning_rate": 1.3797204748921189e-05, + "loss": 0.9733, + "step": 29006 + }, + { + "epoch": 1.87, + "learning_rate": 1.379699005989824e-05, + "loss": 0.9348, + "step": 29007 + }, + { + "epoch": 1.87, + "learning_rate": 1.3796775370875288e-05, + "loss": 1.0026, + "step": 29008 + }, + { + "epoch": 1.87, + "learning_rate": 1.3796560681852337e-05, + "loss": 0.988, + "step": 29009 + }, + { + "epoch": 1.87, + "learning_rate": 1.3796345992829388e-05, + "loss": 0.9638, + "step": 29010 + }, + { + "epoch": 1.87, + "learning_rate": 1.3796131303806437e-05, + "loss": 0.9489, + "step": 29011 + }, + { + "epoch": 1.87, + "learning_rate": 1.3795916614783486e-05, + "loss": 1.0153, + "step": 29012 + }, + { + "epoch": 1.87, + "learning_rate": 1.3795701925760536e-05, + "loss": 0.9338, + "step": 29013 + }, + { + "epoch": 1.87, + "learning_rate": 1.3795487236737585e-05, + "loss": 0.9769, + "step": 29014 + }, + { + "epoch": 1.87, + "learning_rate": 1.3795272547714636e-05, + "loss": 1.0073, + "step": 29015 + }, + { + "epoch": 1.87, + "learning_rate": 1.3795057858691687e-05, + "loss": 0.9645, + "step": 29016 + }, + { + "epoch": 1.87, + "learning_rate": 1.3794843169668737e-05, + "loss": 1.0262, + "step": 29017 + }, + { + "epoch": 1.87, + "learning_rate": 1.3794628480645786e-05, + "loss": 0.986, + "step": 29018 + }, + { + "epoch": 1.87, + "learning_rate": 1.3794413791622835e-05, + "loss": 0.9606, + "step": 29019 + }, + { + "epoch": 1.87, + "learning_rate": 1.3794199102599886e-05, + "loss": 0.9432, + "step": 29020 + }, + { + "epoch": 1.87, + "learning_rate": 1.3793984413576935e-05, + "loss": 0.9724, + "step": 29021 + }, + { + "epoch": 1.87, + "learning_rate": 1.3793769724553984e-05, + "loss": 0.9518, + "step": 29022 + }, + { + "epoch": 1.87, + "learning_rate": 1.3793555035531035e-05, + "loss": 0.9458, + "step": 29023 + }, + { + "epoch": 1.87, + "learning_rate": 1.3793340346508084e-05, + "loss": 0.9457, + "step": 29024 + }, + { + "epoch": 1.87, + "learning_rate": 1.3793125657485134e-05, + "loss": 1.0136, + "step": 29025 + }, + { + "epoch": 1.87, + "learning_rate": 1.3792910968462183e-05, + "loss": 0.9922, + "step": 29026 + }, + { + "epoch": 1.87, + "learning_rate": 1.3792696279439232e-05, + "loss": 0.9615, + "step": 29027 + }, + { + "epoch": 1.87, + "learning_rate": 1.3792481590416283e-05, + "loss": 1.0191, + "step": 29028 + }, + { + "epoch": 1.87, + "learning_rate": 1.3792266901393332e-05, + "loss": 0.9022, + "step": 29029 + }, + { + "epoch": 1.87, + "learning_rate": 1.3792052212370384e-05, + "loss": 1.0054, + "step": 29030 + }, + { + "epoch": 1.87, + "learning_rate": 1.3791837523347433e-05, + "loss": 0.9559, + "step": 29031 + }, + { + "epoch": 1.87, + "learning_rate": 1.3791622834324482e-05, + "loss": 0.9678, + "step": 29032 + }, + { + "epoch": 1.87, + "learning_rate": 1.3791408145301533e-05, + "loss": 0.9995, + "step": 29033 + }, + { + "epoch": 1.87, + "learning_rate": 1.3791193456278582e-05, + "loss": 0.9015, + "step": 29034 + }, + { + "epoch": 1.87, + "learning_rate": 1.3790978767255632e-05, + "loss": 0.9283, + "step": 29035 + }, + { + "epoch": 1.87, + "learning_rate": 1.3790764078232681e-05, + "loss": 0.9986, + "step": 29036 + }, + { + "epoch": 1.87, + "learning_rate": 1.379054938920973e-05, + "loss": 0.9712, + "step": 29037 + }, + { + "epoch": 1.87, + "learning_rate": 1.3790334700186781e-05, + "loss": 0.9467, + "step": 29038 + }, + { + "epoch": 1.87, + "learning_rate": 1.379012001116383e-05, + "loss": 0.9612, + "step": 29039 + }, + { + "epoch": 1.87, + "learning_rate": 1.3789905322140879e-05, + "loss": 0.996, + "step": 29040 + }, + { + "epoch": 1.87, + "learning_rate": 1.378969063311793e-05, + "loss": 0.9332, + "step": 29041 + }, + { + "epoch": 1.87, + "learning_rate": 1.3789475944094978e-05, + "loss": 0.9949, + "step": 29042 + }, + { + "epoch": 1.87, + "learning_rate": 1.3789261255072029e-05, + "loss": 0.9652, + "step": 29043 + }, + { + "epoch": 1.87, + "learning_rate": 1.3789046566049078e-05, + "loss": 0.9862, + "step": 29044 + }, + { + "epoch": 1.87, + "learning_rate": 1.378883187702613e-05, + "loss": 1.0018, + "step": 29045 + }, + { + "epoch": 1.87, + "learning_rate": 1.378861718800318e-05, + "loss": 0.9786, + "step": 29046 + }, + { + "epoch": 1.87, + "learning_rate": 1.3788402498980228e-05, + "loss": 1.0164, + "step": 29047 + }, + { + "epoch": 1.87, + "learning_rate": 1.3788187809957279e-05, + "loss": 0.9458, + "step": 29048 + }, + { + "epoch": 1.87, + "learning_rate": 1.3787973120934328e-05, + "loss": 0.9739, + "step": 29049 + }, + { + "epoch": 1.87, + "learning_rate": 1.3787758431911377e-05, + "loss": 0.9226, + "step": 29050 + }, + { + "epoch": 1.87, + "learning_rate": 1.3787543742888428e-05, + "loss": 0.9684, + "step": 29051 + }, + { + "epoch": 1.87, + "learning_rate": 1.3787329053865477e-05, + "loss": 0.9783, + "step": 29052 + }, + { + "epoch": 1.87, + "learning_rate": 1.3787114364842527e-05, + "loss": 0.9667, + "step": 29053 + }, + { + "epoch": 1.87, + "learning_rate": 1.3786899675819576e-05, + "loss": 0.9357, + "step": 29054 + }, + { + "epoch": 1.87, + "learning_rate": 1.3786684986796625e-05, + "loss": 1.0041, + "step": 29055 + }, + { + "epoch": 1.87, + "learning_rate": 1.3786470297773676e-05, + "loss": 0.9364, + "step": 29056 + }, + { + "epoch": 1.87, + "learning_rate": 1.3786255608750725e-05, + "loss": 0.9833, + "step": 29057 + }, + { + "epoch": 1.87, + "learning_rate": 1.3786040919727774e-05, + "loss": 0.9678, + "step": 29058 + }, + { + "epoch": 1.87, + "learning_rate": 1.3785826230704826e-05, + "loss": 0.9589, + "step": 29059 + }, + { + "epoch": 1.87, + "learning_rate": 1.3785611541681875e-05, + "loss": 0.9873, + "step": 29060 + }, + { + "epoch": 1.87, + "learning_rate": 1.3785396852658926e-05, + "loss": 0.9761, + "step": 29061 + }, + { + "epoch": 1.87, + "learning_rate": 1.3785182163635975e-05, + "loss": 1.0295, + "step": 29062 + }, + { + "epoch": 1.87, + "learning_rate": 1.3784967474613025e-05, + "loss": 0.956, + "step": 29063 + }, + { + "epoch": 1.87, + "learning_rate": 1.3784752785590074e-05, + "loss": 0.9979, + "step": 29064 + }, + { + "epoch": 1.87, + "learning_rate": 1.3784538096567123e-05, + "loss": 0.978, + "step": 29065 + }, + { + "epoch": 1.87, + "learning_rate": 1.3784323407544174e-05, + "loss": 0.9699, + "step": 29066 + }, + { + "epoch": 1.87, + "learning_rate": 1.3784108718521223e-05, + "loss": 0.954, + "step": 29067 + }, + { + "epoch": 1.87, + "learning_rate": 1.3783894029498272e-05, + "loss": 1.0367, + "step": 29068 + }, + { + "epoch": 1.87, + "learning_rate": 1.3783679340475323e-05, + "loss": 0.975, + "step": 29069 + }, + { + "epoch": 1.87, + "learning_rate": 1.3783464651452372e-05, + "loss": 0.9839, + "step": 29070 + }, + { + "epoch": 1.87, + "learning_rate": 1.3783249962429422e-05, + "loss": 1.0162, + "step": 29071 + }, + { + "epoch": 1.87, + "learning_rate": 1.3783035273406471e-05, + "loss": 0.9929, + "step": 29072 + }, + { + "epoch": 1.87, + "learning_rate": 1.378282058438352e-05, + "loss": 0.9718, + "step": 29073 + }, + { + "epoch": 1.87, + "learning_rate": 1.3782605895360572e-05, + "loss": 0.9728, + "step": 29074 + }, + { + "epoch": 1.87, + "learning_rate": 1.3782391206337621e-05, + "loss": 0.938, + "step": 29075 + }, + { + "epoch": 1.87, + "learning_rate": 1.3782176517314672e-05, + "loss": 0.9849, + "step": 29076 + }, + { + "epoch": 1.87, + "learning_rate": 1.3781961828291721e-05, + "loss": 1.0134, + "step": 29077 + }, + { + "epoch": 1.87, + "learning_rate": 1.378174713926877e-05, + "loss": 0.9753, + "step": 29078 + }, + { + "epoch": 1.87, + "learning_rate": 1.378153245024582e-05, + "loss": 0.9847, + "step": 29079 + }, + { + "epoch": 1.87, + "learning_rate": 1.378131776122287e-05, + "loss": 0.9522, + "step": 29080 + }, + { + "epoch": 1.87, + "learning_rate": 1.378110307219992e-05, + "loss": 0.9346, + "step": 29081 + }, + { + "epoch": 1.87, + "learning_rate": 1.378088838317697e-05, + "loss": 0.9491, + "step": 29082 + }, + { + "epoch": 1.87, + "learning_rate": 1.3780673694154018e-05, + "loss": 0.8602, + "step": 29083 + }, + { + "epoch": 1.87, + "learning_rate": 1.3780459005131069e-05, + "loss": 0.9951, + "step": 29084 + }, + { + "epoch": 1.87, + "learning_rate": 1.3780244316108118e-05, + "loss": 0.9537, + "step": 29085 + }, + { + "epoch": 1.87, + "learning_rate": 1.3780029627085167e-05, + "loss": 0.9632, + "step": 29086 + }, + { + "epoch": 1.87, + "learning_rate": 1.3779814938062217e-05, + "loss": 0.9255, + "step": 29087 + }, + { + "epoch": 1.87, + "learning_rate": 1.3779600249039268e-05, + "loss": 0.9184, + "step": 29088 + }, + { + "epoch": 1.87, + "learning_rate": 1.3779385560016319e-05, + "loss": 0.9146, + "step": 29089 + }, + { + "epoch": 1.87, + "learning_rate": 1.3779170870993368e-05, + "loss": 0.9699, + "step": 29090 + }, + { + "epoch": 1.87, + "learning_rate": 1.3778956181970418e-05, + "loss": 0.917, + "step": 29091 + }, + { + "epoch": 1.87, + "learning_rate": 1.3778741492947467e-05, + "loss": 1.0231, + "step": 29092 + }, + { + "epoch": 1.87, + "learning_rate": 1.3778526803924516e-05, + "loss": 0.9798, + "step": 29093 + }, + { + "epoch": 1.87, + "learning_rate": 1.3778312114901567e-05, + "loss": 0.9807, + "step": 29094 + }, + { + "epoch": 1.87, + "learning_rate": 1.3778097425878616e-05, + "loss": 0.9806, + "step": 29095 + }, + { + "epoch": 1.87, + "learning_rate": 1.3777882736855665e-05, + "loss": 0.9676, + "step": 29096 + }, + { + "epoch": 1.87, + "learning_rate": 1.3777668047832716e-05, + "loss": 1.0375, + "step": 29097 + }, + { + "epoch": 1.87, + "learning_rate": 1.3777453358809765e-05, + "loss": 0.9441, + "step": 29098 + }, + { + "epoch": 1.87, + "learning_rate": 1.3777238669786814e-05, + "loss": 1.0027, + "step": 29099 + }, + { + "epoch": 1.87, + "learning_rate": 1.3777023980763864e-05, + "loss": 0.9568, + "step": 29100 + }, + { + "epoch": 1.87, + "learning_rate": 1.3776809291740913e-05, + "loss": 0.9452, + "step": 29101 + }, + { + "epoch": 1.87, + "learning_rate": 1.3776594602717964e-05, + "loss": 0.9084, + "step": 29102 + }, + { + "epoch": 1.87, + "learning_rate": 1.3776379913695014e-05, + "loss": 0.9078, + "step": 29103 + }, + { + "epoch": 1.87, + "learning_rate": 1.3776165224672065e-05, + "loss": 0.9506, + "step": 29104 + }, + { + "epoch": 1.87, + "learning_rate": 1.3775950535649114e-05, + "loss": 0.9674, + "step": 29105 + }, + { + "epoch": 1.87, + "learning_rate": 1.3775735846626163e-05, + "loss": 0.9826, + "step": 29106 + }, + { + "epoch": 1.87, + "learning_rate": 1.3775521157603214e-05, + "loss": 1.0032, + "step": 29107 + }, + { + "epoch": 1.87, + "learning_rate": 1.3775306468580263e-05, + "loss": 0.9551, + "step": 29108 + }, + { + "epoch": 1.87, + "learning_rate": 1.3775091779557312e-05, + "loss": 0.965, + "step": 29109 + }, + { + "epoch": 1.87, + "learning_rate": 1.3774877090534362e-05, + "loss": 0.9757, + "step": 29110 + }, + { + "epoch": 1.87, + "learning_rate": 1.3774662401511411e-05, + "loss": 0.9581, + "step": 29111 + }, + { + "epoch": 1.87, + "learning_rate": 1.3774447712488462e-05, + "loss": 0.9208, + "step": 29112 + }, + { + "epoch": 1.87, + "learning_rate": 1.3774233023465511e-05, + "loss": 0.9613, + "step": 29113 + }, + { + "epoch": 1.87, + "learning_rate": 1.377401833444256e-05, + "loss": 0.8699, + "step": 29114 + }, + { + "epoch": 1.87, + "learning_rate": 1.377380364541961e-05, + "loss": 1.0019, + "step": 29115 + }, + { + "epoch": 1.87, + "learning_rate": 1.377358895639666e-05, + "loss": 1.0063, + "step": 29116 + }, + { + "epoch": 1.87, + "learning_rate": 1.3773374267373712e-05, + "loss": 1.0061, + "step": 29117 + }, + { + "epoch": 1.87, + "learning_rate": 1.377315957835076e-05, + "loss": 0.9591, + "step": 29118 + }, + { + "epoch": 1.87, + "learning_rate": 1.377294488932781e-05, + "loss": 0.8936, + "step": 29119 + }, + { + "epoch": 1.87, + "learning_rate": 1.377273020030486e-05, + "loss": 0.9634, + "step": 29120 + }, + { + "epoch": 1.87, + "learning_rate": 1.377251551128191e-05, + "loss": 0.9715, + "step": 29121 + }, + { + "epoch": 1.87, + "learning_rate": 1.377230082225896e-05, + "loss": 0.9868, + "step": 29122 + }, + { + "epoch": 1.87, + "learning_rate": 1.3772086133236009e-05, + "loss": 0.9509, + "step": 29123 + }, + { + "epoch": 1.87, + "learning_rate": 1.3771871444213058e-05, + "loss": 0.995, + "step": 29124 + }, + { + "epoch": 1.87, + "learning_rate": 1.3771656755190109e-05, + "loss": 1.0342, + "step": 29125 + }, + { + "epoch": 1.87, + "learning_rate": 1.3771442066167158e-05, + "loss": 0.9138, + "step": 29126 + }, + { + "epoch": 1.87, + "learning_rate": 1.3771227377144207e-05, + "loss": 0.8189, + "step": 29127 + }, + { + "epoch": 1.87, + "learning_rate": 1.3771012688121257e-05, + "loss": 0.9611, + "step": 29128 + }, + { + "epoch": 1.87, + "learning_rate": 1.3770797999098306e-05, + "loss": 0.9647, + "step": 29129 + }, + { + "epoch": 1.87, + "learning_rate": 1.3770583310075357e-05, + "loss": 0.8932, + "step": 29130 + }, + { + "epoch": 1.87, + "learning_rate": 1.3770368621052406e-05, + "loss": 1.0186, + "step": 29131 + }, + { + "epoch": 1.87, + "learning_rate": 1.3770153932029458e-05, + "loss": 1.075, + "step": 29132 + }, + { + "epoch": 1.87, + "learning_rate": 1.3769939243006507e-05, + "loss": 0.9825, + "step": 29133 + }, + { + "epoch": 1.87, + "learning_rate": 1.3769724553983556e-05, + "loss": 1.0073, + "step": 29134 + }, + { + "epoch": 1.87, + "learning_rate": 1.3769509864960607e-05, + "loss": 0.92, + "step": 29135 + }, + { + "epoch": 1.87, + "learning_rate": 1.3769295175937656e-05, + "loss": 0.9141, + "step": 29136 + }, + { + "epoch": 1.87, + "learning_rate": 1.3769080486914705e-05, + "loss": 0.9446, + "step": 29137 + }, + { + "epoch": 1.87, + "learning_rate": 1.3768865797891755e-05, + "loss": 0.8883, + "step": 29138 + }, + { + "epoch": 1.87, + "learning_rate": 1.3768651108868804e-05, + "loss": 0.9993, + "step": 29139 + }, + { + "epoch": 1.87, + "learning_rate": 1.3768436419845855e-05, + "loss": 0.9249, + "step": 29140 + }, + { + "epoch": 1.87, + "learning_rate": 1.3768221730822904e-05, + "loss": 1.0033, + "step": 29141 + }, + { + "epoch": 1.87, + "learning_rate": 1.3768007041799953e-05, + "loss": 0.9696, + "step": 29142 + }, + { + "epoch": 1.87, + "learning_rate": 1.3767792352777004e-05, + "loss": 0.928, + "step": 29143 + }, + { + "epoch": 1.88, + "learning_rate": 1.3767577663754053e-05, + "loss": 0.9987, + "step": 29144 + }, + { + "epoch": 1.88, + "learning_rate": 1.3767362974731101e-05, + "loss": 1.0179, + "step": 29145 + }, + { + "epoch": 1.88, + "learning_rate": 1.3767148285708152e-05, + "loss": 0.9455, + "step": 29146 + }, + { + "epoch": 1.88, + "learning_rate": 1.3766933596685203e-05, + "loss": 0.9847, + "step": 29147 + }, + { + "epoch": 1.88, + "learning_rate": 1.3766718907662253e-05, + "loss": 0.9647, + "step": 29148 + }, + { + "epoch": 1.88, + "learning_rate": 1.3766504218639302e-05, + "loss": 0.9065, + "step": 29149 + }, + { + "epoch": 1.88, + "learning_rate": 1.3766289529616353e-05, + "loss": 0.9162, + "step": 29150 + }, + { + "epoch": 1.88, + "learning_rate": 1.3766074840593402e-05, + "loss": 0.9707, + "step": 29151 + }, + { + "epoch": 1.88, + "learning_rate": 1.3765860151570451e-05, + "loss": 0.9647, + "step": 29152 + }, + { + "epoch": 1.88, + "learning_rate": 1.3765645462547502e-05, + "loss": 1.0992, + "step": 29153 + }, + { + "epoch": 1.88, + "learning_rate": 1.376543077352455e-05, + "loss": 0.9978, + "step": 29154 + }, + { + "epoch": 1.88, + "learning_rate": 1.37652160845016e-05, + "loss": 0.9672, + "step": 29155 + }, + { + "epoch": 1.88, + "learning_rate": 1.376500139547865e-05, + "loss": 0.9812, + "step": 29156 + }, + { + "epoch": 1.88, + "learning_rate": 1.37647867064557e-05, + "loss": 0.9472, + "step": 29157 + }, + { + "epoch": 1.88, + "learning_rate": 1.376457201743275e-05, + "loss": 1.0678, + "step": 29158 + }, + { + "epoch": 1.88, + "learning_rate": 1.3764357328409799e-05, + "loss": 0.9465, + "step": 29159 + }, + { + "epoch": 1.88, + "learning_rate": 1.3764142639386848e-05, + "loss": 0.925, + "step": 29160 + }, + { + "epoch": 1.88, + "learning_rate": 1.37639279503639e-05, + "loss": 1.0433, + "step": 29161 + }, + { + "epoch": 1.88, + "learning_rate": 1.3763713261340949e-05, + "loss": 0.9646, + "step": 29162 + }, + { + "epoch": 1.88, + "learning_rate": 1.3763498572318e-05, + "loss": 0.9767, + "step": 29163 + }, + { + "epoch": 1.88, + "learning_rate": 1.3763283883295049e-05, + "loss": 0.9702, + "step": 29164 + }, + { + "epoch": 1.88, + "learning_rate": 1.3763069194272098e-05, + "loss": 0.8846, + "step": 29165 + }, + { + "epoch": 1.88, + "learning_rate": 1.3762854505249148e-05, + "loss": 1.0038, + "step": 29166 + }, + { + "epoch": 1.88, + "learning_rate": 1.3762639816226197e-05, + "loss": 0.9702, + "step": 29167 + }, + { + "epoch": 1.88, + "learning_rate": 1.3762425127203248e-05, + "loss": 0.9754, + "step": 29168 + }, + { + "epoch": 1.88, + "learning_rate": 1.3762210438180297e-05, + "loss": 0.9564, + "step": 29169 + }, + { + "epoch": 1.88, + "learning_rate": 1.3761995749157346e-05, + "loss": 0.948, + "step": 29170 + }, + { + "epoch": 1.88, + "learning_rate": 1.3761781060134397e-05, + "loss": 1.001, + "step": 29171 + }, + { + "epoch": 1.88, + "learning_rate": 1.3761566371111446e-05, + "loss": 0.9951, + "step": 29172 + }, + { + "epoch": 1.88, + "learning_rate": 1.3761351682088495e-05, + "loss": 0.9243, + "step": 29173 + }, + { + "epoch": 1.88, + "learning_rate": 1.3761136993065545e-05, + "loss": 0.9492, + "step": 29174 + }, + { + "epoch": 1.88, + "learning_rate": 1.3760922304042594e-05, + "loss": 0.9479, + "step": 29175 + }, + { + "epoch": 1.88, + "learning_rate": 1.3760707615019647e-05, + "loss": 0.9771, + "step": 29176 + }, + { + "epoch": 1.88, + "learning_rate": 1.3760492925996695e-05, + "loss": 1.002, + "step": 29177 + }, + { + "epoch": 1.88, + "learning_rate": 1.3760278236973746e-05, + "loss": 0.8794, + "step": 29178 + }, + { + "epoch": 1.88, + "learning_rate": 1.3760063547950795e-05, + "loss": 1.0217, + "step": 29179 + }, + { + "epoch": 1.88, + "learning_rate": 1.3759848858927844e-05, + "loss": 0.9412, + "step": 29180 + }, + { + "epoch": 1.88, + "learning_rate": 1.3759634169904895e-05, + "loss": 0.9912, + "step": 29181 + }, + { + "epoch": 1.88, + "learning_rate": 1.3759419480881944e-05, + "loss": 1.0425, + "step": 29182 + }, + { + "epoch": 1.88, + "learning_rate": 1.3759204791858993e-05, + "loss": 0.9432, + "step": 29183 + }, + { + "epoch": 1.88, + "learning_rate": 1.3758990102836043e-05, + "loss": 0.9309, + "step": 29184 + }, + { + "epoch": 1.88, + "learning_rate": 1.3758775413813092e-05, + "loss": 1.026, + "step": 29185 + }, + { + "epoch": 1.88, + "learning_rate": 1.3758560724790141e-05, + "loss": 1.0027, + "step": 29186 + }, + { + "epoch": 1.88, + "learning_rate": 1.3758346035767192e-05, + "loss": 1.0143, + "step": 29187 + }, + { + "epoch": 1.88, + "learning_rate": 1.3758131346744241e-05, + "loss": 0.9794, + "step": 29188 + }, + { + "epoch": 1.88, + "learning_rate": 1.3757916657721292e-05, + "loss": 0.9706, + "step": 29189 + }, + { + "epoch": 1.88, + "learning_rate": 1.3757701968698342e-05, + "loss": 0.993, + "step": 29190 + }, + { + "epoch": 1.88, + "learning_rate": 1.3757487279675393e-05, + "loss": 1.006, + "step": 29191 + }, + { + "epoch": 1.88, + "learning_rate": 1.3757272590652442e-05, + "loss": 0.8906, + "step": 29192 + }, + { + "epoch": 1.88, + "learning_rate": 1.375705790162949e-05, + "loss": 1.0516, + "step": 29193 + }, + { + "epoch": 1.88, + "learning_rate": 1.3756843212606541e-05, + "loss": 0.9072, + "step": 29194 + }, + { + "epoch": 1.88, + "learning_rate": 1.375662852358359e-05, + "loss": 0.979, + "step": 29195 + }, + { + "epoch": 1.88, + "learning_rate": 1.375641383456064e-05, + "loss": 0.9677, + "step": 29196 + }, + { + "epoch": 1.88, + "learning_rate": 1.375619914553769e-05, + "loss": 0.9185, + "step": 29197 + }, + { + "epoch": 1.88, + "learning_rate": 1.3755984456514739e-05, + "loss": 0.961, + "step": 29198 + }, + { + "epoch": 1.88, + "learning_rate": 1.375576976749179e-05, + "loss": 0.9744, + "step": 29199 + }, + { + "epoch": 1.88, + "learning_rate": 1.3755555078468839e-05, + "loss": 0.933, + "step": 29200 + }, + { + "epoch": 1.88, + "learning_rate": 1.3755340389445888e-05, + "loss": 0.9799, + "step": 29201 + }, + { + "epoch": 1.88, + "learning_rate": 1.3755125700422938e-05, + "loss": 0.9296, + "step": 29202 + }, + { + "epoch": 1.88, + "learning_rate": 1.3754911011399987e-05, + "loss": 1.0203, + "step": 29203 + }, + { + "epoch": 1.88, + "learning_rate": 1.3754696322377036e-05, + "loss": 0.8989, + "step": 29204 + }, + { + "epoch": 1.88, + "learning_rate": 1.3754481633354089e-05, + "loss": 0.9802, + "step": 29205 + }, + { + "epoch": 1.88, + "learning_rate": 1.3754266944331137e-05, + "loss": 0.9956, + "step": 29206 + }, + { + "epoch": 1.88, + "learning_rate": 1.3754052255308188e-05, + "loss": 0.9225, + "step": 29207 + }, + { + "epoch": 1.88, + "learning_rate": 1.3753837566285237e-05, + "loss": 0.9903, + "step": 29208 + }, + { + "epoch": 1.88, + "learning_rate": 1.3753622877262288e-05, + "loss": 0.9398, + "step": 29209 + }, + { + "epoch": 1.88, + "learning_rate": 1.3753408188239337e-05, + "loss": 1.0117, + "step": 29210 + }, + { + "epoch": 1.88, + "learning_rate": 1.3753193499216386e-05, + "loss": 0.9658, + "step": 29211 + }, + { + "epoch": 1.88, + "learning_rate": 1.3752978810193436e-05, + "loss": 0.9369, + "step": 29212 + }, + { + "epoch": 1.88, + "learning_rate": 1.3752764121170485e-05, + "loss": 0.9267, + "step": 29213 + }, + { + "epoch": 1.88, + "learning_rate": 1.3752549432147534e-05, + "loss": 1.0539, + "step": 29214 + }, + { + "epoch": 1.88, + "learning_rate": 1.3752334743124585e-05, + "loss": 0.9646, + "step": 29215 + }, + { + "epoch": 1.88, + "learning_rate": 1.3752120054101634e-05, + "loss": 0.9183, + "step": 29216 + }, + { + "epoch": 1.88, + "learning_rate": 1.3751905365078685e-05, + "loss": 0.9088, + "step": 29217 + }, + { + "epoch": 1.88, + "learning_rate": 1.3751690676055734e-05, + "loss": 0.9269, + "step": 29218 + }, + { + "epoch": 1.88, + "learning_rate": 1.3751475987032786e-05, + "loss": 0.9441, + "step": 29219 + }, + { + "epoch": 1.88, + "learning_rate": 1.3751261298009835e-05, + "loss": 0.9833, + "step": 29220 + }, + { + "epoch": 1.88, + "learning_rate": 1.3751046608986884e-05, + "loss": 0.9622, + "step": 29221 + }, + { + "epoch": 1.88, + "learning_rate": 1.3750831919963934e-05, + "loss": 0.9543, + "step": 29222 + }, + { + "epoch": 1.88, + "learning_rate": 1.3750617230940983e-05, + "loss": 1.0482, + "step": 29223 + }, + { + "epoch": 1.88, + "learning_rate": 1.3750402541918032e-05, + "loss": 1.0162, + "step": 29224 + }, + { + "epoch": 1.88, + "learning_rate": 1.3750187852895083e-05, + "loss": 1.0103, + "step": 29225 + }, + { + "epoch": 1.88, + "learning_rate": 1.3749973163872132e-05, + "loss": 0.9825, + "step": 29226 + }, + { + "epoch": 1.88, + "learning_rate": 1.3749758474849183e-05, + "loss": 0.8753, + "step": 29227 + }, + { + "epoch": 1.88, + "learning_rate": 1.3749543785826232e-05, + "loss": 0.9427, + "step": 29228 + }, + { + "epoch": 1.88, + "learning_rate": 1.374932909680328e-05, + "loss": 0.9055, + "step": 29229 + }, + { + "epoch": 1.88, + "learning_rate": 1.3749114407780331e-05, + "loss": 0.9726, + "step": 29230 + }, + { + "epoch": 1.88, + "learning_rate": 1.374889971875738e-05, + "loss": 0.937, + "step": 29231 + }, + { + "epoch": 1.88, + "learning_rate": 1.374868502973443e-05, + "loss": 0.9811, + "step": 29232 + }, + { + "epoch": 1.88, + "learning_rate": 1.374847034071148e-05, + "loss": 1.001, + "step": 29233 + }, + { + "epoch": 1.88, + "learning_rate": 1.374825565168853e-05, + "loss": 1.009, + "step": 29234 + }, + { + "epoch": 1.88, + "learning_rate": 1.3748040962665581e-05, + "loss": 0.8743, + "step": 29235 + }, + { + "epoch": 1.88, + "learning_rate": 1.374782627364263e-05, + "loss": 0.9387, + "step": 29236 + }, + { + "epoch": 1.88, + "learning_rate": 1.374761158461968e-05, + "loss": 1.0059, + "step": 29237 + }, + { + "epoch": 1.88, + "learning_rate": 1.374739689559673e-05, + "loss": 0.9754, + "step": 29238 + }, + { + "epoch": 1.88, + "learning_rate": 1.3747182206573779e-05, + "loss": 1.048, + "step": 29239 + }, + { + "epoch": 1.88, + "learning_rate": 1.374696751755083e-05, + "loss": 0.9512, + "step": 29240 + }, + { + "epoch": 1.88, + "learning_rate": 1.3746752828527878e-05, + "loss": 1.0523, + "step": 29241 + }, + { + "epoch": 1.88, + "learning_rate": 1.3746538139504927e-05, + "loss": 0.9621, + "step": 29242 + }, + { + "epoch": 1.88, + "learning_rate": 1.3746323450481978e-05, + "loss": 0.9023, + "step": 29243 + }, + { + "epoch": 1.88, + "learning_rate": 1.3746108761459027e-05, + "loss": 0.9672, + "step": 29244 + }, + { + "epoch": 1.88, + "learning_rate": 1.3745894072436076e-05, + "loss": 0.9583, + "step": 29245 + }, + { + "epoch": 1.88, + "learning_rate": 1.3745679383413127e-05, + "loss": 1.01, + "step": 29246 + }, + { + "epoch": 1.88, + "learning_rate": 1.3745464694390176e-05, + "loss": 0.9676, + "step": 29247 + }, + { + "epoch": 1.88, + "learning_rate": 1.3745250005367228e-05, + "loss": 0.8867, + "step": 29248 + }, + { + "epoch": 1.88, + "learning_rate": 1.3745035316344277e-05, + "loss": 0.941, + "step": 29249 + }, + { + "epoch": 1.88, + "learning_rate": 1.3744820627321328e-05, + "loss": 0.9627, + "step": 29250 + }, + { + "epoch": 1.88, + "learning_rate": 1.3744605938298376e-05, + "loss": 1.0072, + "step": 29251 + }, + { + "epoch": 1.88, + "learning_rate": 1.3744391249275425e-05, + "loss": 0.8892, + "step": 29252 + }, + { + "epoch": 1.88, + "learning_rate": 1.3744176560252476e-05, + "loss": 1.0282, + "step": 29253 + }, + { + "epoch": 1.88, + "learning_rate": 1.3743961871229525e-05, + "loss": 1.0048, + "step": 29254 + }, + { + "epoch": 1.88, + "learning_rate": 1.3743747182206576e-05, + "loss": 0.9654, + "step": 29255 + }, + { + "epoch": 1.88, + "learning_rate": 1.3743532493183625e-05, + "loss": 1.0363, + "step": 29256 + }, + { + "epoch": 1.88, + "learning_rate": 1.3743317804160674e-05, + "loss": 0.9623, + "step": 29257 + }, + { + "epoch": 1.88, + "learning_rate": 1.3743103115137724e-05, + "loss": 0.9459, + "step": 29258 + }, + { + "epoch": 1.88, + "learning_rate": 1.3742888426114773e-05, + "loss": 1.0055, + "step": 29259 + }, + { + "epoch": 1.88, + "learning_rate": 1.3742673737091822e-05, + "loss": 0.9256, + "step": 29260 + }, + { + "epoch": 1.88, + "learning_rate": 1.3742459048068873e-05, + "loss": 1.0357, + "step": 29261 + }, + { + "epoch": 1.88, + "learning_rate": 1.3742244359045922e-05, + "loss": 0.8907, + "step": 29262 + }, + { + "epoch": 1.88, + "learning_rate": 1.3742029670022974e-05, + "loss": 1.0096, + "step": 29263 + }, + { + "epoch": 1.88, + "learning_rate": 1.3741814981000023e-05, + "loss": 0.9776, + "step": 29264 + }, + { + "epoch": 1.88, + "learning_rate": 1.3741600291977074e-05, + "loss": 0.9461, + "step": 29265 + }, + { + "epoch": 1.88, + "learning_rate": 1.3741385602954123e-05, + "loss": 0.993, + "step": 29266 + }, + { + "epoch": 1.88, + "learning_rate": 1.3741170913931172e-05, + "loss": 0.9199, + "step": 29267 + }, + { + "epoch": 1.88, + "learning_rate": 1.3740956224908222e-05, + "loss": 0.9779, + "step": 29268 + }, + { + "epoch": 1.88, + "learning_rate": 1.3740741535885271e-05, + "loss": 0.978, + "step": 29269 + }, + { + "epoch": 1.88, + "learning_rate": 1.374052684686232e-05, + "loss": 0.9683, + "step": 29270 + }, + { + "epoch": 1.88, + "learning_rate": 1.3740312157839371e-05, + "loss": 0.9728, + "step": 29271 + }, + { + "epoch": 1.88, + "learning_rate": 1.374009746881642e-05, + "loss": 0.9731, + "step": 29272 + }, + { + "epoch": 1.88, + "learning_rate": 1.3739882779793469e-05, + "loss": 0.9203, + "step": 29273 + }, + { + "epoch": 1.88, + "learning_rate": 1.373966809077052e-05, + "loss": 0.9575, + "step": 29274 + }, + { + "epoch": 1.88, + "learning_rate": 1.3739453401747569e-05, + "loss": 1.0261, + "step": 29275 + }, + { + "epoch": 1.88, + "learning_rate": 1.373923871272462e-05, + "loss": 0.9523, + "step": 29276 + }, + { + "epoch": 1.88, + "learning_rate": 1.373902402370167e-05, + "loss": 0.9117, + "step": 29277 + }, + { + "epoch": 1.88, + "learning_rate": 1.373880933467872e-05, + "loss": 0.9504, + "step": 29278 + }, + { + "epoch": 1.88, + "learning_rate": 1.373859464565577e-05, + "loss": 0.9963, + "step": 29279 + }, + { + "epoch": 1.88, + "learning_rate": 1.3738379956632818e-05, + "loss": 0.9109, + "step": 29280 + }, + { + "epoch": 1.88, + "learning_rate": 1.3738165267609869e-05, + "loss": 0.9905, + "step": 29281 + }, + { + "epoch": 1.88, + "learning_rate": 1.3737950578586918e-05, + "loss": 1.0374, + "step": 29282 + }, + { + "epoch": 1.88, + "learning_rate": 1.3737735889563967e-05, + "loss": 0.9786, + "step": 29283 + }, + { + "epoch": 1.88, + "learning_rate": 1.3737521200541018e-05, + "loss": 1.0093, + "step": 29284 + }, + { + "epoch": 1.88, + "learning_rate": 1.3737306511518067e-05, + "loss": 0.9944, + "step": 29285 + }, + { + "epoch": 1.88, + "learning_rate": 1.3737091822495117e-05, + "loss": 1.0179, + "step": 29286 + }, + { + "epoch": 1.88, + "learning_rate": 1.3736877133472166e-05, + "loss": 0.8571, + "step": 29287 + }, + { + "epoch": 1.88, + "learning_rate": 1.3736662444449215e-05, + "loss": 1.043, + "step": 29288 + }, + { + "epoch": 1.88, + "learning_rate": 1.3736447755426266e-05, + "loss": 1.0402, + "step": 29289 + }, + { + "epoch": 1.88, + "learning_rate": 1.3736233066403315e-05, + "loss": 0.9492, + "step": 29290 + }, + { + "epoch": 1.88, + "learning_rate": 1.3736018377380364e-05, + "loss": 1.0042, + "step": 29291 + }, + { + "epoch": 1.88, + "learning_rate": 1.3735803688357416e-05, + "loss": 0.9655, + "step": 29292 + }, + { + "epoch": 1.88, + "learning_rate": 1.3735588999334465e-05, + "loss": 1.0145, + "step": 29293 + }, + { + "epoch": 1.88, + "learning_rate": 1.3735374310311516e-05, + "loss": 0.9757, + "step": 29294 + }, + { + "epoch": 1.88, + "learning_rate": 1.3735159621288565e-05, + "loss": 0.9644, + "step": 29295 + }, + { + "epoch": 1.88, + "learning_rate": 1.3734944932265615e-05, + "loss": 0.966, + "step": 29296 + }, + { + "epoch": 1.88, + "learning_rate": 1.3734730243242664e-05, + "loss": 1.037, + "step": 29297 + }, + { + "epoch": 1.88, + "learning_rate": 1.3734515554219713e-05, + "loss": 0.9378, + "step": 29298 + }, + { + "epoch": 1.89, + "learning_rate": 1.3734300865196764e-05, + "loss": 1.0473, + "step": 29299 + }, + { + "epoch": 1.89, + "learning_rate": 1.3734086176173813e-05, + "loss": 1.0331, + "step": 29300 + }, + { + "epoch": 1.89, + "learning_rate": 1.3733871487150862e-05, + "loss": 0.9921, + "step": 29301 + }, + { + "epoch": 1.89, + "learning_rate": 1.3733656798127913e-05, + "loss": 0.976, + "step": 29302 + }, + { + "epoch": 1.89, + "learning_rate": 1.3733442109104962e-05, + "loss": 1.0075, + "step": 29303 + }, + { + "epoch": 1.89, + "learning_rate": 1.3733227420082012e-05, + "loss": 0.9524, + "step": 29304 + }, + { + "epoch": 1.89, + "learning_rate": 1.3733012731059061e-05, + "loss": 0.9283, + "step": 29305 + }, + { + "epoch": 1.89, + "learning_rate": 1.3732798042036114e-05, + "loss": 0.9535, + "step": 29306 + }, + { + "epoch": 1.89, + "learning_rate": 1.3732583353013163e-05, + "loss": 0.9588, + "step": 29307 + }, + { + "epoch": 1.89, + "learning_rate": 1.3732368663990212e-05, + "loss": 0.9006, + "step": 29308 + }, + { + "epoch": 1.89, + "learning_rate": 1.3732153974967262e-05, + "loss": 0.9519, + "step": 29309 + }, + { + "epoch": 1.89, + "learning_rate": 1.3731939285944311e-05, + "loss": 0.9579, + "step": 29310 + }, + { + "epoch": 1.89, + "learning_rate": 1.373172459692136e-05, + "loss": 0.9943, + "step": 29311 + }, + { + "epoch": 1.89, + "learning_rate": 1.373150990789841e-05, + "loss": 0.9204, + "step": 29312 + }, + { + "epoch": 1.89, + "learning_rate": 1.373129521887546e-05, + "loss": 0.9761, + "step": 29313 + }, + { + "epoch": 1.89, + "learning_rate": 1.373108052985251e-05, + "loss": 0.9475, + "step": 29314 + }, + { + "epoch": 1.89, + "learning_rate": 1.373086584082956e-05, + "loss": 1.0111, + "step": 29315 + }, + { + "epoch": 1.89, + "learning_rate": 1.3730651151806608e-05, + "loss": 0.9759, + "step": 29316 + }, + { + "epoch": 1.89, + "learning_rate": 1.3730436462783659e-05, + "loss": 0.9702, + "step": 29317 + }, + { + "epoch": 1.89, + "learning_rate": 1.3730221773760708e-05, + "loss": 0.8781, + "step": 29318 + }, + { + "epoch": 1.89, + "learning_rate": 1.3730007084737757e-05, + "loss": 0.9032, + "step": 29319 + }, + { + "epoch": 1.89, + "learning_rate": 1.3729792395714808e-05, + "loss": 0.9183, + "step": 29320 + }, + { + "epoch": 1.89, + "learning_rate": 1.3729577706691858e-05, + "loss": 0.9636, + "step": 29321 + }, + { + "epoch": 1.89, + "learning_rate": 1.3729363017668909e-05, + "loss": 0.9762, + "step": 29322 + }, + { + "epoch": 1.89, + "learning_rate": 1.3729148328645958e-05, + "loss": 1.0427, + "step": 29323 + }, + { + "epoch": 1.89, + "learning_rate": 1.3728933639623009e-05, + "loss": 0.903, + "step": 29324 + }, + { + "epoch": 1.89, + "learning_rate": 1.3728718950600057e-05, + "loss": 0.9989, + "step": 29325 + }, + { + "epoch": 1.89, + "learning_rate": 1.3728504261577106e-05, + "loss": 0.8823, + "step": 29326 + }, + { + "epoch": 1.89, + "learning_rate": 1.3728289572554157e-05, + "loss": 0.9531, + "step": 29327 + }, + { + "epoch": 1.89, + "learning_rate": 1.3728074883531206e-05, + "loss": 1.0299, + "step": 29328 + }, + { + "epoch": 1.89, + "learning_rate": 1.3727860194508255e-05, + "loss": 0.9662, + "step": 29329 + }, + { + "epoch": 1.89, + "learning_rate": 1.3727645505485306e-05, + "loss": 0.975, + "step": 29330 + }, + { + "epoch": 1.89, + "learning_rate": 1.3727430816462355e-05, + "loss": 0.9601, + "step": 29331 + }, + { + "epoch": 1.89, + "learning_rate": 1.3727216127439404e-05, + "loss": 0.9597, + "step": 29332 + }, + { + "epoch": 1.89, + "learning_rate": 1.3727001438416454e-05, + "loss": 1.0557, + "step": 29333 + }, + { + "epoch": 1.89, + "learning_rate": 1.3726786749393503e-05, + "loss": 0.9744, + "step": 29334 + }, + { + "epoch": 1.89, + "learning_rate": 1.3726572060370556e-05, + "loss": 0.9809, + "step": 29335 + }, + { + "epoch": 1.89, + "learning_rate": 1.3726357371347605e-05, + "loss": 1.0245, + "step": 29336 + }, + { + "epoch": 1.89, + "learning_rate": 1.3726142682324655e-05, + "loss": 0.9228, + "step": 29337 + }, + { + "epoch": 1.89, + "learning_rate": 1.3725927993301704e-05, + "loss": 0.9036, + "step": 29338 + }, + { + "epoch": 1.89, + "learning_rate": 1.3725713304278753e-05, + "loss": 0.9696, + "step": 29339 + }, + { + "epoch": 1.89, + "learning_rate": 1.3725498615255804e-05, + "loss": 0.9485, + "step": 29340 + }, + { + "epoch": 1.89, + "learning_rate": 1.3725283926232853e-05, + "loss": 0.9915, + "step": 29341 + }, + { + "epoch": 1.89, + "learning_rate": 1.3725069237209902e-05, + "loss": 0.9215, + "step": 29342 + }, + { + "epoch": 1.89, + "learning_rate": 1.3724854548186952e-05, + "loss": 0.9576, + "step": 29343 + }, + { + "epoch": 1.89, + "learning_rate": 1.3724639859164001e-05, + "loss": 0.9892, + "step": 29344 + }, + { + "epoch": 1.89, + "learning_rate": 1.3724425170141052e-05, + "loss": 0.9103, + "step": 29345 + }, + { + "epoch": 1.89, + "learning_rate": 1.3724210481118101e-05, + "loss": 1.0083, + "step": 29346 + }, + { + "epoch": 1.89, + "learning_rate": 1.372399579209515e-05, + "loss": 0.8977, + "step": 29347 + }, + { + "epoch": 1.89, + "learning_rate": 1.37237811030722e-05, + "loss": 0.9411, + "step": 29348 + }, + { + "epoch": 1.89, + "learning_rate": 1.372356641404925e-05, + "loss": 1.0508, + "step": 29349 + }, + { + "epoch": 1.89, + "learning_rate": 1.3723351725026302e-05, + "loss": 1.0077, + "step": 29350 + }, + { + "epoch": 1.89, + "learning_rate": 1.3723137036003351e-05, + "loss": 0.9601, + "step": 29351 + }, + { + "epoch": 1.89, + "learning_rate": 1.37229223469804e-05, + "loss": 0.9912, + "step": 29352 + }, + { + "epoch": 1.89, + "learning_rate": 1.372270765795745e-05, + "loss": 1.0396, + "step": 29353 + }, + { + "epoch": 1.89, + "learning_rate": 1.37224929689345e-05, + "loss": 0.9821, + "step": 29354 + }, + { + "epoch": 1.89, + "learning_rate": 1.372227827991155e-05, + "loss": 0.9873, + "step": 29355 + }, + { + "epoch": 1.89, + "learning_rate": 1.3722063590888599e-05, + "loss": 0.955, + "step": 29356 + }, + { + "epoch": 1.89, + "learning_rate": 1.3721848901865648e-05, + "loss": 0.9931, + "step": 29357 + }, + { + "epoch": 1.89, + "learning_rate": 1.3721634212842699e-05, + "loss": 1.0131, + "step": 29358 + }, + { + "epoch": 1.89, + "learning_rate": 1.3721419523819748e-05, + "loss": 0.9472, + "step": 29359 + }, + { + "epoch": 1.89, + "learning_rate": 1.3721204834796797e-05, + "loss": 0.9318, + "step": 29360 + }, + { + "epoch": 1.89, + "learning_rate": 1.3720990145773847e-05, + "loss": 0.9758, + "step": 29361 + }, + { + "epoch": 1.89, + "learning_rate": 1.3720775456750896e-05, + "loss": 0.9159, + "step": 29362 + }, + { + "epoch": 1.89, + "learning_rate": 1.3720560767727947e-05, + "loss": 0.9394, + "step": 29363 + }, + { + "epoch": 1.89, + "learning_rate": 1.3720346078704998e-05, + "loss": 0.9586, + "step": 29364 + }, + { + "epoch": 1.89, + "learning_rate": 1.3720131389682048e-05, + "loss": 0.9289, + "step": 29365 + }, + { + "epoch": 1.89, + "learning_rate": 1.3719916700659097e-05, + "loss": 0.924, + "step": 29366 + }, + { + "epoch": 1.89, + "learning_rate": 1.3719702011636146e-05, + "loss": 0.9416, + "step": 29367 + }, + { + "epoch": 1.89, + "learning_rate": 1.3719487322613197e-05, + "loss": 1.0371, + "step": 29368 + }, + { + "epoch": 1.89, + "learning_rate": 1.3719272633590246e-05, + "loss": 0.9198, + "step": 29369 + }, + { + "epoch": 1.89, + "learning_rate": 1.3719057944567295e-05, + "loss": 1.0247, + "step": 29370 + }, + { + "epoch": 1.89, + "learning_rate": 1.3718843255544345e-05, + "loss": 0.9158, + "step": 29371 + }, + { + "epoch": 1.89, + "learning_rate": 1.3718628566521394e-05, + "loss": 0.9799, + "step": 29372 + }, + { + "epoch": 1.89, + "learning_rate": 1.3718413877498445e-05, + "loss": 1.04, + "step": 29373 + }, + { + "epoch": 1.89, + "learning_rate": 1.3718199188475494e-05, + "loss": 0.9498, + "step": 29374 + }, + { + "epoch": 1.89, + "learning_rate": 1.3717984499452543e-05, + "loss": 0.9773, + "step": 29375 + }, + { + "epoch": 1.89, + "learning_rate": 1.3717769810429594e-05, + "loss": 0.9674, + "step": 29376 + }, + { + "epoch": 1.89, + "learning_rate": 1.3717555121406643e-05, + "loss": 1.0092, + "step": 29377 + }, + { + "epoch": 1.89, + "learning_rate": 1.3717340432383692e-05, + "loss": 1.0286, + "step": 29378 + }, + { + "epoch": 1.89, + "learning_rate": 1.3717125743360744e-05, + "loss": 0.9535, + "step": 29379 + }, + { + "epoch": 1.89, + "learning_rate": 1.3716911054337793e-05, + "loss": 0.9956, + "step": 29380 + }, + { + "epoch": 1.89, + "learning_rate": 1.3716696365314844e-05, + "loss": 0.9351, + "step": 29381 + }, + { + "epoch": 1.89, + "learning_rate": 1.3716481676291893e-05, + "loss": 0.9737, + "step": 29382 + }, + { + "epoch": 1.89, + "learning_rate": 1.3716266987268943e-05, + "loss": 0.9113, + "step": 29383 + }, + { + "epoch": 1.89, + "learning_rate": 1.3716052298245992e-05, + "loss": 0.9552, + "step": 29384 + }, + { + "epoch": 1.89, + "learning_rate": 1.3715837609223041e-05, + "loss": 0.9505, + "step": 29385 + }, + { + "epoch": 1.89, + "learning_rate": 1.3715622920200092e-05, + "loss": 0.9633, + "step": 29386 + }, + { + "epoch": 1.89, + "learning_rate": 1.371540823117714e-05, + "loss": 1.0154, + "step": 29387 + }, + { + "epoch": 1.89, + "learning_rate": 1.371519354215419e-05, + "loss": 1.0282, + "step": 29388 + }, + { + "epoch": 1.89, + "learning_rate": 1.371497885313124e-05, + "loss": 0.9674, + "step": 29389 + }, + { + "epoch": 1.89, + "learning_rate": 1.371476416410829e-05, + "loss": 0.974, + "step": 29390 + }, + { + "epoch": 1.89, + "learning_rate": 1.371454947508534e-05, + "loss": 0.929, + "step": 29391 + }, + { + "epoch": 1.89, + "learning_rate": 1.3714334786062389e-05, + "loss": 1.0162, + "step": 29392 + }, + { + "epoch": 1.89, + "learning_rate": 1.3714120097039441e-05, + "loss": 0.9713, + "step": 29393 + }, + { + "epoch": 1.89, + "learning_rate": 1.371390540801649e-05, + "loss": 0.9952, + "step": 29394 + }, + { + "epoch": 1.89, + "learning_rate": 1.371369071899354e-05, + "loss": 0.9102, + "step": 29395 + }, + { + "epoch": 1.89, + "learning_rate": 1.371347602997059e-05, + "loss": 0.936, + "step": 29396 + }, + { + "epoch": 1.89, + "learning_rate": 1.3713261340947639e-05, + "loss": 1.0116, + "step": 29397 + }, + { + "epoch": 1.89, + "learning_rate": 1.3713046651924688e-05, + "loss": 0.9975, + "step": 29398 + }, + { + "epoch": 1.89, + "learning_rate": 1.3712831962901738e-05, + "loss": 0.9327, + "step": 29399 + }, + { + "epoch": 1.89, + "learning_rate": 1.3712617273878787e-05, + "loss": 0.9709, + "step": 29400 + }, + { + "epoch": 1.89, + "learning_rate": 1.3712402584855838e-05, + "loss": 1.0417, + "step": 29401 + }, + { + "epoch": 1.89, + "learning_rate": 1.3712187895832887e-05, + "loss": 0.9033, + "step": 29402 + }, + { + "epoch": 1.89, + "learning_rate": 1.3711973206809936e-05, + "loss": 0.999, + "step": 29403 + }, + { + "epoch": 1.89, + "learning_rate": 1.3711758517786987e-05, + "loss": 0.9846, + "step": 29404 + }, + { + "epoch": 1.89, + "learning_rate": 1.3711543828764036e-05, + "loss": 0.9664, + "step": 29405 + }, + { + "epoch": 1.89, + "learning_rate": 1.3711329139741085e-05, + "loss": 0.9462, + "step": 29406 + }, + { + "epoch": 1.89, + "learning_rate": 1.3711114450718135e-05, + "loss": 0.9047, + "step": 29407 + }, + { + "epoch": 1.89, + "learning_rate": 1.3710899761695186e-05, + "loss": 0.9938, + "step": 29408 + }, + { + "epoch": 1.89, + "learning_rate": 1.3710685072672237e-05, + "loss": 1.0351, + "step": 29409 + }, + { + "epoch": 1.89, + "learning_rate": 1.3710470383649286e-05, + "loss": 0.9709, + "step": 29410 + }, + { + "epoch": 1.89, + "learning_rate": 1.3710255694626336e-05, + "loss": 1.0035, + "step": 29411 + }, + { + "epoch": 1.89, + "learning_rate": 1.3710041005603385e-05, + "loss": 0.9193, + "step": 29412 + }, + { + "epoch": 1.89, + "learning_rate": 1.3709826316580434e-05, + "loss": 0.9654, + "step": 29413 + }, + { + "epoch": 1.89, + "learning_rate": 1.3709611627557485e-05, + "loss": 0.8962, + "step": 29414 + }, + { + "epoch": 1.89, + "learning_rate": 1.3709396938534534e-05, + "loss": 0.9667, + "step": 29415 + }, + { + "epoch": 1.89, + "learning_rate": 1.3709182249511583e-05, + "loss": 0.9408, + "step": 29416 + }, + { + "epoch": 1.89, + "learning_rate": 1.3708967560488633e-05, + "loss": 0.9521, + "step": 29417 + }, + { + "epoch": 1.89, + "learning_rate": 1.3708752871465682e-05, + "loss": 0.9109, + "step": 29418 + }, + { + "epoch": 1.89, + "learning_rate": 1.3708538182442731e-05, + "loss": 0.9799, + "step": 29419 + }, + { + "epoch": 1.89, + "learning_rate": 1.3708323493419782e-05, + "loss": 0.9835, + "step": 29420 + }, + { + "epoch": 1.89, + "learning_rate": 1.3708108804396831e-05, + "loss": 1.0667, + "step": 29421 + }, + { + "epoch": 1.89, + "learning_rate": 1.3707894115373883e-05, + "loss": 0.934, + "step": 29422 + }, + { + "epoch": 1.89, + "learning_rate": 1.3707679426350932e-05, + "loss": 0.9637, + "step": 29423 + }, + { + "epoch": 1.89, + "learning_rate": 1.3707464737327983e-05, + "loss": 0.9893, + "step": 29424 + }, + { + "epoch": 1.89, + "learning_rate": 1.3707250048305032e-05, + "loss": 0.9452, + "step": 29425 + }, + { + "epoch": 1.89, + "learning_rate": 1.3707035359282081e-05, + "loss": 0.9584, + "step": 29426 + }, + { + "epoch": 1.89, + "learning_rate": 1.3706820670259132e-05, + "loss": 0.9192, + "step": 29427 + }, + { + "epoch": 1.89, + "learning_rate": 1.370660598123618e-05, + "loss": 0.88, + "step": 29428 + }, + { + "epoch": 1.89, + "learning_rate": 1.370639129221323e-05, + "loss": 0.9443, + "step": 29429 + }, + { + "epoch": 1.89, + "learning_rate": 1.370617660319028e-05, + "loss": 1.0399, + "step": 29430 + }, + { + "epoch": 1.89, + "learning_rate": 1.3705961914167329e-05, + "loss": 1.0139, + "step": 29431 + }, + { + "epoch": 1.89, + "learning_rate": 1.370574722514438e-05, + "loss": 0.9187, + "step": 29432 + }, + { + "epoch": 1.89, + "learning_rate": 1.3705532536121429e-05, + "loss": 0.9091, + "step": 29433 + }, + { + "epoch": 1.89, + "learning_rate": 1.3705317847098478e-05, + "loss": 0.9336, + "step": 29434 + }, + { + "epoch": 1.89, + "learning_rate": 1.3705103158075528e-05, + "loss": 1.0178, + "step": 29435 + }, + { + "epoch": 1.89, + "learning_rate": 1.3704888469052577e-05, + "loss": 0.9569, + "step": 29436 + }, + { + "epoch": 1.89, + "learning_rate": 1.370467378002963e-05, + "loss": 1.0557, + "step": 29437 + }, + { + "epoch": 1.89, + "learning_rate": 1.3704459091006679e-05, + "loss": 0.9487, + "step": 29438 + }, + { + "epoch": 1.89, + "learning_rate": 1.3704244401983728e-05, + "loss": 0.9852, + "step": 29439 + }, + { + "epoch": 1.89, + "learning_rate": 1.3704029712960778e-05, + "loss": 0.8919, + "step": 29440 + }, + { + "epoch": 1.89, + "learning_rate": 1.3703815023937827e-05, + "loss": 0.9477, + "step": 29441 + }, + { + "epoch": 1.89, + "learning_rate": 1.3703600334914878e-05, + "loss": 0.916, + "step": 29442 + }, + { + "epoch": 1.89, + "learning_rate": 1.3703385645891927e-05, + "loss": 0.9381, + "step": 29443 + }, + { + "epoch": 1.89, + "learning_rate": 1.3703170956868976e-05, + "loss": 0.8555, + "step": 29444 + }, + { + "epoch": 1.89, + "learning_rate": 1.3702956267846026e-05, + "loss": 1.0366, + "step": 29445 + }, + { + "epoch": 1.89, + "learning_rate": 1.3702741578823075e-05, + "loss": 0.9316, + "step": 29446 + }, + { + "epoch": 1.89, + "learning_rate": 1.3702526889800124e-05, + "loss": 0.9687, + "step": 29447 + }, + { + "epoch": 1.89, + "learning_rate": 1.3702312200777175e-05, + "loss": 0.9057, + "step": 29448 + }, + { + "epoch": 1.89, + "learning_rate": 1.3702097511754224e-05, + "loss": 1.0004, + "step": 29449 + }, + { + "epoch": 1.89, + "learning_rate": 1.3701882822731275e-05, + "loss": 1.0037, + "step": 29450 + }, + { + "epoch": 1.89, + "learning_rate": 1.3701668133708325e-05, + "loss": 0.9656, + "step": 29451 + }, + { + "epoch": 1.89, + "learning_rate": 1.3701453444685376e-05, + "loss": 0.9872, + "step": 29452 + }, + { + "epoch": 1.89, + "learning_rate": 1.3701238755662425e-05, + "loss": 0.9854, + "step": 29453 + }, + { + "epoch": 1.9, + "learning_rate": 1.3701024066639474e-05, + "loss": 0.9704, + "step": 29454 + }, + { + "epoch": 1.9, + "learning_rate": 1.3700809377616525e-05, + "loss": 1.0225, + "step": 29455 + }, + { + "epoch": 1.9, + "learning_rate": 1.3700594688593574e-05, + "loss": 0.9138, + "step": 29456 + }, + { + "epoch": 1.9, + "learning_rate": 1.3700379999570623e-05, + "loss": 1.0294, + "step": 29457 + }, + { + "epoch": 1.9, + "learning_rate": 1.3700165310547673e-05, + "loss": 0.9655, + "step": 29458 + }, + { + "epoch": 1.9, + "learning_rate": 1.3699950621524722e-05, + "loss": 0.9829, + "step": 29459 + }, + { + "epoch": 1.9, + "learning_rate": 1.3699735932501773e-05, + "loss": 0.9754, + "step": 29460 + }, + { + "epoch": 1.9, + "learning_rate": 1.3699521243478822e-05, + "loss": 0.992, + "step": 29461 + }, + { + "epoch": 1.9, + "learning_rate": 1.369930655445587e-05, + "loss": 0.9652, + "step": 29462 + }, + { + "epoch": 1.9, + "learning_rate": 1.3699091865432921e-05, + "loss": 0.9762, + "step": 29463 + }, + { + "epoch": 1.9, + "learning_rate": 1.369887717640997e-05, + "loss": 0.9572, + "step": 29464 + }, + { + "epoch": 1.9, + "learning_rate": 1.369866248738702e-05, + "loss": 1.024, + "step": 29465 + }, + { + "epoch": 1.9, + "learning_rate": 1.3698447798364072e-05, + "loss": 0.9595, + "step": 29466 + }, + { + "epoch": 1.9, + "learning_rate": 1.369823310934112e-05, + "loss": 0.9753, + "step": 29467 + }, + { + "epoch": 1.9, + "learning_rate": 1.3698018420318171e-05, + "loss": 0.9686, + "step": 29468 + }, + { + "epoch": 1.9, + "learning_rate": 1.369780373129522e-05, + "loss": 0.9545, + "step": 29469 + }, + { + "epoch": 1.9, + "learning_rate": 1.3697589042272271e-05, + "loss": 1.0186, + "step": 29470 + }, + { + "epoch": 1.9, + "learning_rate": 1.369737435324932e-05, + "loss": 0.9509, + "step": 29471 + }, + { + "epoch": 1.9, + "learning_rate": 1.3697159664226369e-05, + "loss": 0.919, + "step": 29472 + }, + { + "epoch": 1.9, + "learning_rate": 1.369694497520342e-05, + "loss": 0.949, + "step": 29473 + }, + { + "epoch": 1.9, + "learning_rate": 1.3696730286180468e-05, + "loss": 1.0384, + "step": 29474 + }, + { + "epoch": 1.9, + "learning_rate": 1.3696515597157517e-05, + "loss": 0.9896, + "step": 29475 + }, + { + "epoch": 1.9, + "learning_rate": 1.3696300908134568e-05, + "loss": 0.9447, + "step": 29476 + }, + { + "epoch": 1.9, + "learning_rate": 1.3696086219111617e-05, + "loss": 0.9162, + "step": 29477 + }, + { + "epoch": 1.9, + "learning_rate": 1.3695871530088668e-05, + "loss": 0.9524, + "step": 29478 + }, + { + "epoch": 1.9, + "learning_rate": 1.3695656841065717e-05, + "loss": 0.9719, + "step": 29479 + }, + { + "epoch": 1.9, + "learning_rate": 1.3695442152042769e-05, + "loss": 0.9083, + "step": 29480 + }, + { + "epoch": 1.9, + "learning_rate": 1.3695227463019818e-05, + "loss": 0.987, + "step": 29481 + }, + { + "epoch": 1.9, + "learning_rate": 1.3695012773996867e-05, + "loss": 0.9579, + "step": 29482 + }, + { + "epoch": 1.9, + "learning_rate": 1.3694798084973918e-05, + "loss": 0.907, + "step": 29483 + }, + { + "epoch": 1.9, + "learning_rate": 1.3694583395950967e-05, + "loss": 1.033, + "step": 29484 + }, + { + "epoch": 1.9, + "learning_rate": 1.3694368706928016e-05, + "loss": 0.9423, + "step": 29485 + }, + { + "epoch": 1.9, + "learning_rate": 1.3694154017905066e-05, + "loss": 0.943, + "step": 29486 + }, + { + "epoch": 1.9, + "learning_rate": 1.3693939328882115e-05, + "loss": 1.0156, + "step": 29487 + }, + { + "epoch": 1.9, + "learning_rate": 1.3693724639859166e-05, + "loss": 0.9371, + "step": 29488 + }, + { + "epoch": 1.9, + "learning_rate": 1.3693509950836215e-05, + "loss": 0.9235, + "step": 29489 + }, + { + "epoch": 1.9, + "learning_rate": 1.3693295261813264e-05, + "loss": 1.0076, + "step": 29490 + }, + { + "epoch": 1.9, + "learning_rate": 1.3693080572790314e-05, + "loss": 1.0449, + "step": 29491 + }, + { + "epoch": 1.9, + "learning_rate": 1.3692865883767363e-05, + "loss": 1.0006, + "step": 29492 + }, + { + "epoch": 1.9, + "learning_rate": 1.3692651194744412e-05, + "loss": 1.0066, + "step": 29493 + }, + { + "epoch": 1.9, + "learning_rate": 1.3692436505721463e-05, + "loss": 0.9928, + "step": 29494 + }, + { + "epoch": 1.9, + "learning_rate": 1.3692221816698514e-05, + "loss": 0.9584, + "step": 29495 + }, + { + "epoch": 1.9, + "learning_rate": 1.3692007127675564e-05, + "loss": 0.9689, + "step": 29496 + }, + { + "epoch": 1.9, + "learning_rate": 1.3691792438652613e-05, + "loss": 0.9996, + "step": 29497 + }, + { + "epoch": 1.9, + "learning_rate": 1.3691577749629664e-05, + "loss": 0.9388, + "step": 29498 + }, + { + "epoch": 1.9, + "learning_rate": 1.3691363060606713e-05, + "loss": 0.9547, + "step": 29499 + }, + { + "epoch": 1.9, + "learning_rate": 1.3691148371583762e-05, + "loss": 0.9738, + "step": 29500 + }, + { + "epoch": 1.9, + "learning_rate": 1.3690933682560813e-05, + "loss": 0.9529, + "step": 29501 + }, + { + "epoch": 1.9, + "learning_rate": 1.3690718993537862e-05, + "loss": 0.9967, + "step": 29502 + }, + { + "epoch": 1.9, + "learning_rate": 1.369050430451491e-05, + "loss": 1.0154, + "step": 29503 + }, + { + "epoch": 1.9, + "learning_rate": 1.3690289615491961e-05, + "loss": 0.9216, + "step": 29504 + }, + { + "epoch": 1.9, + "learning_rate": 1.369007492646901e-05, + "loss": 0.9549, + "step": 29505 + }, + { + "epoch": 1.9, + "learning_rate": 1.3689860237446059e-05, + "loss": 0.9725, + "step": 29506 + }, + { + "epoch": 1.9, + "learning_rate": 1.368964554842311e-05, + "loss": 1.014, + "step": 29507 + }, + { + "epoch": 1.9, + "learning_rate": 1.3689430859400159e-05, + "loss": 0.8986, + "step": 29508 + }, + { + "epoch": 1.9, + "learning_rate": 1.368921617037721e-05, + "loss": 0.9791, + "step": 29509 + }, + { + "epoch": 1.9, + "learning_rate": 1.368900148135426e-05, + "loss": 0.9329, + "step": 29510 + }, + { + "epoch": 1.9, + "learning_rate": 1.368878679233131e-05, + "loss": 0.9155, + "step": 29511 + }, + { + "epoch": 1.9, + "learning_rate": 1.368857210330836e-05, + "loss": 0.9081, + "step": 29512 + }, + { + "epoch": 1.9, + "learning_rate": 1.3688357414285409e-05, + "loss": 1.0108, + "step": 29513 + }, + { + "epoch": 1.9, + "learning_rate": 1.368814272526246e-05, + "loss": 0.943, + "step": 29514 + }, + { + "epoch": 1.9, + "learning_rate": 1.3687928036239508e-05, + "loss": 0.9744, + "step": 29515 + }, + { + "epoch": 1.9, + "learning_rate": 1.3687713347216557e-05, + "loss": 0.9404, + "step": 29516 + }, + { + "epoch": 1.9, + "learning_rate": 1.3687498658193608e-05, + "loss": 1.0127, + "step": 29517 + }, + { + "epoch": 1.9, + "learning_rate": 1.3687283969170657e-05, + "loss": 0.981, + "step": 29518 + }, + { + "epoch": 1.9, + "learning_rate": 1.3687069280147707e-05, + "loss": 0.9351, + "step": 29519 + }, + { + "epoch": 1.9, + "learning_rate": 1.3686854591124756e-05, + "loss": 0.9659, + "step": 29520 + }, + { + "epoch": 1.9, + "learning_rate": 1.3686639902101805e-05, + "loss": 0.9645, + "step": 29521 + }, + { + "epoch": 1.9, + "learning_rate": 1.3686425213078856e-05, + "loss": 1.0162, + "step": 29522 + }, + { + "epoch": 1.9, + "learning_rate": 1.3686210524055905e-05, + "loss": 0.9617, + "step": 29523 + }, + { + "epoch": 1.9, + "learning_rate": 1.3685995835032957e-05, + "loss": 1.0057, + "step": 29524 + }, + { + "epoch": 1.9, + "learning_rate": 1.3685781146010006e-05, + "loss": 0.9558, + "step": 29525 + }, + { + "epoch": 1.9, + "learning_rate": 1.3685566456987055e-05, + "loss": 0.9422, + "step": 29526 + }, + { + "epoch": 1.9, + "learning_rate": 1.3685351767964106e-05, + "loss": 0.9514, + "step": 29527 + }, + { + "epoch": 1.9, + "learning_rate": 1.3685137078941155e-05, + "loss": 0.9126, + "step": 29528 + }, + { + "epoch": 1.9, + "learning_rate": 1.3684922389918206e-05, + "loss": 0.9342, + "step": 29529 + }, + { + "epoch": 1.9, + "learning_rate": 1.3684707700895255e-05, + "loss": 1.0288, + "step": 29530 + }, + { + "epoch": 1.9, + "learning_rate": 1.3684493011872304e-05, + "loss": 0.9421, + "step": 29531 + }, + { + "epoch": 1.9, + "learning_rate": 1.3684278322849354e-05, + "loss": 0.9265, + "step": 29532 + }, + { + "epoch": 1.9, + "learning_rate": 1.3684063633826403e-05, + "loss": 0.9468, + "step": 29533 + }, + { + "epoch": 1.9, + "learning_rate": 1.3683848944803452e-05, + "loss": 0.9982, + "step": 29534 + }, + { + "epoch": 1.9, + "learning_rate": 1.3683634255780503e-05, + "loss": 0.8969, + "step": 29535 + }, + { + "epoch": 1.9, + "learning_rate": 1.3683419566757552e-05, + "loss": 0.9758, + "step": 29536 + }, + { + "epoch": 1.9, + "learning_rate": 1.3683204877734602e-05, + "loss": 0.987, + "step": 29537 + }, + { + "epoch": 1.9, + "learning_rate": 1.3682990188711651e-05, + "loss": 0.9448, + "step": 29538 + }, + { + "epoch": 1.9, + "learning_rate": 1.3682775499688704e-05, + "loss": 0.9719, + "step": 29539 + }, + { + "epoch": 1.9, + "learning_rate": 1.3682560810665753e-05, + "loss": 0.9251, + "step": 29540 + }, + { + "epoch": 1.9, + "learning_rate": 1.3682346121642802e-05, + "loss": 0.8959, + "step": 29541 + }, + { + "epoch": 1.9, + "learning_rate": 1.3682131432619852e-05, + "loss": 0.9891, + "step": 29542 + }, + { + "epoch": 1.9, + "learning_rate": 1.3681916743596901e-05, + "loss": 0.9921, + "step": 29543 + }, + { + "epoch": 1.9, + "learning_rate": 1.368170205457395e-05, + "loss": 0.9854, + "step": 29544 + }, + { + "epoch": 1.9, + "learning_rate": 1.3681487365551001e-05, + "loss": 0.9918, + "step": 29545 + }, + { + "epoch": 1.9, + "learning_rate": 1.368127267652805e-05, + "loss": 0.9952, + "step": 29546 + }, + { + "epoch": 1.9, + "learning_rate": 1.36810579875051e-05, + "loss": 0.986, + "step": 29547 + }, + { + "epoch": 1.9, + "learning_rate": 1.368084329848215e-05, + "loss": 0.9619, + "step": 29548 + }, + { + "epoch": 1.9, + "learning_rate": 1.3680628609459198e-05, + "loss": 0.9504, + "step": 29549 + }, + { + "epoch": 1.9, + "learning_rate": 1.3680413920436249e-05, + "loss": 0.9678, + "step": 29550 + }, + { + "epoch": 1.9, + "learning_rate": 1.3680199231413298e-05, + "loss": 0.9468, + "step": 29551 + }, + { + "epoch": 1.9, + "learning_rate": 1.3679984542390347e-05, + "loss": 0.9694, + "step": 29552 + }, + { + "epoch": 1.9, + "learning_rate": 1.36797698533674e-05, + "loss": 0.9891, + "step": 29553 + }, + { + "epoch": 1.9, + "learning_rate": 1.3679555164344448e-05, + "loss": 1.0083, + "step": 29554 + }, + { + "epoch": 1.9, + "learning_rate": 1.3679340475321499e-05, + "loss": 0.927, + "step": 29555 + }, + { + "epoch": 1.9, + "learning_rate": 1.3679125786298548e-05, + "loss": 0.9006, + "step": 29556 + }, + { + "epoch": 1.9, + "learning_rate": 1.3678911097275599e-05, + "loss": 0.9177, + "step": 29557 + }, + { + "epoch": 1.9, + "learning_rate": 1.3678696408252648e-05, + "loss": 0.9574, + "step": 29558 + }, + { + "epoch": 1.9, + "learning_rate": 1.3678481719229697e-05, + "loss": 0.9297, + "step": 29559 + }, + { + "epoch": 1.9, + "learning_rate": 1.3678267030206747e-05, + "loss": 1.0035, + "step": 29560 + }, + { + "epoch": 1.9, + "learning_rate": 1.3678052341183796e-05, + "loss": 1.0612, + "step": 29561 + }, + { + "epoch": 1.9, + "learning_rate": 1.3677837652160845e-05, + "loss": 1.012, + "step": 29562 + }, + { + "epoch": 1.9, + "learning_rate": 1.3677622963137896e-05, + "loss": 1.0112, + "step": 29563 + }, + { + "epoch": 1.9, + "learning_rate": 1.3677408274114945e-05, + "loss": 1.0716, + "step": 29564 + }, + { + "epoch": 1.9, + "learning_rate": 1.3677193585091995e-05, + "loss": 0.9678, + "step": 29565 + }, + { + "epoch": 1.9, + "learning_rate": 1.3676978896069044e-05, + "loss": 0.9882, + "step": 29566 + }, + { + "epoch": 1.9, + "learning_rate": 1.3676764207046093e-05, + "loss": 0.9987, + "step": 29567 + }, + { + "epoch": 1.9, + "learning_rate": 1.3676549518023146e-05, + "loss": 0.9531, + "step": 29568 + }, + { + "epoch": 1.9, + "learning_rate": 1.3676334829000195e-05, + "loss": 0.9399, + "step": 29569 + }, + { + "epoch": 1.9, + "learning_rate": 1.3676120139977245e-05, + "loss": 1.0365, + "step": 29570 + }, + { + "epoch": 1.9, + "learning_rate": 1.3675905450954294e-05, + "loss": 0.9467, + "step": 29571 + }, + { + "epoch": 1.9, + "learning_rate": 1.3675690761931343e-05, + "loss": 0.8265, + "step": 29572 + }, + { + "epoch": 1.9, + "learning_rate": 1.3675476072908394e-05, + "loss": 0.9792, + "step": 29573 + }, + { + "epoch": 1.9, + "learning_rate": 1.3675261383885443e-05, + "loss": 0.9892, + "step": 29574 + }, + { + "epoch": 1.9, + "learning_rate": 1.3675046694862494e-05, + "loss": 0.8878, + "step": 29575 + }, + { + "epoch": 1.9, + "learning_rate": 1.3674832005839543e-05, + "loss": 0.9394, + "step": 29576 + }, + { + "epoch": 1.9, + "learning_rate": 1.3674617316816591e-05, + "loss": 0.9859, + "step": 29577 + }, + { + "epoch": 1.9, + "learning_rate": 1.3674402627793642e-05, + "loss": 1.0499, + "step": 29578 + }, + { + "epoch": 1.9, + "learning_rate": 1.3674187938770691e-05, + "loss": 1.0399, + "step": 29579 + }, + { + "epoch": 1.9, + "learning_rate": 1.367397324974774e-05, + "loss": 1.0091, + "step": 29580 + }, + { + "epoch": 1.9, + "learning_rate": 1.367375856072479e-05, + "loss": 0.9646, + "step": 29581 + }, + { + "epoch": 1.9, + "learning_rate": 1.3673543871701841e-05, + "loss": 0.8918, + "step": 29582 + }, + { + "epoch": 1.9, + "learning_rate": 1.3673329182678892e-05, + "loss": 0.984, + "step": 29583 + }, + { + "epoch": 1.9, + "learning_rate": 1.3673114493655941e-05, + "loss": 0.9861, + "step": 29584 + }, + { + "epoch": 1.9, + "learning_rate": 1.3672899804632992e-05, + "loss": 0.9929, + "step": 29585 + }, + { + "epoch": 1.9, + "learning_rate": 1.367268511561004e-05, + "loss": 0.9747, + "step": 29586 + }, + { + "epoch": 1.9, + "learning_rate": 1.367247042658709e-05, + "loss": 0.9951, + "step": 29587 + }, + { + "epoch": 1.9, + "learning_rate": 1.367225573756414e-05, + "loss": 0.875, + "step": 29588 + }, + { + "epoch": 1.9, + "learning_rate": 1.367204104854119e-05, + "loss": 0.933, + "step": 29589 + }, + { + "epoch": 1.9, + "learning_rate": 1.3671826359518238e-05, + "loss": 0.9806, + "step": 29590 + }, + { + "epoch": 1.9, + "learning_rate": 1.3671611670495289e-05, + "loss": 0.9925, + "step": 29591 + }, + { + "epoch": 1.9, + "learning_rate": 1.3671396981472338e-05, + "loss": 1.0086, + "step": 29592 + }, + { + "epoch": 1.9, + "learning_rate": 1.3671182292449387e-05, + "loss": 0.9706, + "step": 29593 + }, + { + "epoch": 1.9, + "learning_rate": 1.3670967603426437e-05, + "loss": 0.9819, + "step": 29594 + }, + { + "epoch": 1.9, + "learning_rate": 1.3670752914403486e-05, + "loss": 0.8816, + "step": 29595 + }, + { + "epoch": 1.9, + "learning_rate": 1.3670538225380537e-05, + "loss": 1.0999, + "step": 29596 + }, + { + "epoch": 1.9, + "learning_rate": 1.3670323536357588e-05, + "loss": 0.9799, + "step": 29597 + }, + { + "epoch": 1.9, + "learning_rate": 1.3670108847334638e-05, + "loss": 0.9553, + "step": 29598 + }, + { + "epoch": 1.9, + "learning_rate": 1.3669894158311687e-05, + "loss": 0.9811, + "step": 29599 + }, + { + "epoch": 1.9, + "learning_rate": 1.3669679469288736e-05, + "loss": 0.9737, + "step": 29600 + }, + { + "epoch": 1.9, + "learning_rate": 1.3669464780265787e-05, + "loss": 0.9559, + "step": 29601 + }, + { + "epoch": 1.9, + "learning_rate": 1.3669250091242836e-05, + "loss": 0.9826, + "step": 29602 + }, + { + "epoch": 1.9, + "learning_rate": 1.3669035402219885e-05, + "loss": 0.9275, + "step": 29603 + }, + { + "epoch": 1.9, + "learning_rate": 1.3668820713196936e-05, + "loss": 0.9821, + "step": 29604 + }, + { + "epoch": 1.9, + "learning_rate": 1.3668606024173985e-05, + "loss": 0.9615, + "step": 29605 + }, + { + "epoch": 1.9, + "learning_rate": 1.3668391335151035e-05, + "loss": 0.9551, + "step": 29606 + }, + { + "epoch": 1.9, + "learning_rate": 1.3668176646128084e-05, + "loss": 1.0261, + "step": 29607 + }, + { + "epoch": 1.9, + "learning_rate": 1.3667961957105133e-05, + "loss": 1.0017, + "step": 29608 + }, + { + "epoch": 1.9, + "learning_rate": 1.3667747268082184e-05, + "loss": 0.9011, + "step": 29609 + }, + { + "epoch": 1.91, + "learning_rate": 1.3667532579059233e-05, + "loss": 0.9943, + "step": 29610 + }, + { + "epoch": 1.91, + "learning_rate": 1.3667317890036285e-05, + "loss": 0.9943, + "step": 29611 + }, + { + "epoch": 1.91, + "learning_rate": 1.3667103201013334e-05, + "loss": 0.9996, + "step": 29612 + }, + { + "epoch": 1.91, + "learning_rate": 1.3666888511990383e-05, + "loss": 0.9566, + "step": 29613 + }, + { + "epoch": 1.91, + "learning_rate": 1.3666673822967434e-05, + "loss": 0.9102, + "step": 29614 + }, + { + "epoch": 1.91, + "learning_rate": 1.3666459133944483e-05, + "loss": 1.0407, + "step": 29615 + }, + { + "epoch": 1.91, + "learning_rate": 1.3666244444921533e-05, + "loss": 0.9393, + "step": 29616 + }, + { + "epoch": 1.91, + "learning_rate": 1.3666029755898582e-05, + "loss": 1.0007, + "step": 29617 + }, + { + "epoch": 1.91, + "learning_rate": 1.3665815066875631e-05, + "loss": 0.95, + "step": 29618 + }, + { + "epoch": 1.91, + "learning_rate": 1.3665600377852682e-05, + "loss": 0.9407, + "step": 29619 + }, + { + "epoch": 1.91, + "learning_rate": 1.366538568882973e-05, + "loss": 0.9602, + "step": 29620 + }, + { + "epoch": 1.91, + "learning_rate": 1.366517099980678e-05, + "loss": 0.9719, + "step": 29621 + }, + { + "epoch": 1.91, + "learning_rate": 1.366495631078383e-05, + "loss": 0.9604, + "step": 29622 + }, + { + "epoch": 1.91, + "learning_rate": 1.366474162176088e-05, + "loss": 0.9249, + "step": 29623 + }, + { + "epoch": 1.91, + "learning_rate": 1.366452693273793e-05, + "loss": 1.0117, + "step": 29624 + }, + { + "epoch": 1.91, + "learning_rate": 1.3664312243714979e-05, + "loss": 0.9602, + "step": 29625 + }, + { + "epoch": 1.91, + "learning_rate": 1.3664097554692031e-05, + "loss": 0.9831, + "step": 29626 + }, + { + "epoch": 1.91, + "learning_rate": 1.366388286566908e-05, + "loss": 0.9173, + "step": 29627 + }, + { + "epoch": 1.91, + "learning_rate": 1.366366817664613e-05, + "loss": 0.9482, + "step": 29628 + }, + { + "epoch": 1.91, + "learning_rate": 1.366345348762318e-05, + "loss": 0.9785, + "step": 29629 + }, + { + "epoch": 1.91, + "learning_rate": 1.3663238798600229e-05, + "loss": 0.8977, + "step": 29630 + }, + { + "epoch": 1.91, + "learning_rate": 1.3663024109577278e-05, + "loss": 0.9096, + "step": 29631 + }, + { + "epoch": 1.91, + "learning_rate": 1.3662809420554329e-05, + "loss": 1.0187, + "step": 29632 + }, + { + "epoch": 1.91, + "learning_rate": 1.3662594731531378e-05, + "loss": 0.9982, + "step": 29633 + }, + { + "epoch": 1.91, + "learning_rate": 1.3662380042508428e-05, + "loss": 0.982, + "step": 29634 + }, + { + "epoch": 1.91, + "learning_rate": 1.3662165353485477e-05, + "loss": 0.9367, + "step": 29635 + }, + { + "epoch": 1.91, + "learning_rate": 1.3661950664462526e-05, + "loss": 0.9134, + "step": 29636 + }, + { + "epoch": 1.91, + "learning_rate": 1.3661735975439577e-05, + "loss": 0.9749, + "step": 29637 + }, + { + "epoch": 1.91, + "learning_rate": 1.3661521286416626e-05, + "loss": 0.9447, + "step": 29638 + }, + { + "epoch": 1.91, + "learning_rate": 1.3661306597393675e-05, + "loss": 0.942, + "step": 29639 + }, + { + "epoch": 1.91, + "learning_rate": 1.3661091908370727e-05, + "loss": 0.9476, + "step": 29640 + }, + { + "epoch": 1.91, + "learning_rate": 1.3660877219347776e-05, + "loss": 1.0598, + "step": 29641 + }, + { + "epoch": 1.91, + "learning_rate": 1.3660662530324827e-05, + "loss": 0.9257, + "step": 29642 + }, + { + "epoch": 1.91, + "learning_rate": 1.3660447841301876e-05, + "loss": 1.0031, + "step": 29643 + }, + { + "epoch": 1.91, + "learning_rate": 1.3660233152278926e-05, + "loss": 0.9911, + "step": 29644 + }, + { + "epoch": 1.91, + "learning_rate": 1.3660018463255975e-05, + "loss": 0.9655, + "step": 29645 + }, + { + "epoch": 1.91, + "learning_rate": 1.3659803774233024e-05, + "loss": 1.0168, + "step": 29646 + }, + { + "epoch": 1.91, + "learning_rate": 1.3659589085210075e-05, + "loss": 0.9937, + "step": 29647 + }, + { + "epoch": 1.91, + "learning_rate": 1.3659374396187124e-05, + "loss": 1.0451, + "step": 29648 + }, + { + "epoch": 1.91, + "learning_rate": 1.3659159707164173e-05, + "loss": 0.9922, + "step": 29649 + }, + { + "epoch": 1.91, + "learning_rate": 1.3658945018141224e-05, + "loss": 0.8743, + "step": 29650 + }, + { + "epoch": 1.91, + "learning_rate": 1.3658730329118272e-05, + "loss": 0.9191, + "step": 29651 + }, + { + "epoch": 1.91, + "learning_rate": 1.3658515640095321e-05, + "loss": 1.0225, + "step": 29652 + }, + { + "epoch": 1.91, + "learning_rate": 1.3658300951072372e-05, + "loss": 1.0742, + "step": 29653 + }, + { + "epoch": 1.91, + "learning_rate": 1.3658086262049421e-05, + "loss": 0.9532, + "step": 29654 + }, + { + "epoch": 1.91, + "learning_rate": 1.3657871573026473e-05, + "loss": 1.0255, + "step": 29655 + }, + { + "epoch": 1.91, + "learning_rate": 1.3657656884003522e-05, + "loss": 0.949, + "step": 29656 + }, + { + "epoch": 1.91, + "learning_rate": 1.3657442194980573e-05, + "loss": 0.9546, + "step": 29657 + }, + { + "epoch": 1.91, + "learning_rate": 1.3657227505957622e-05, + "loss": 0.9944, + "step": 29658 + }, + { + "epoch": 1.91, + "learning_rate": 1.3657012816934671e-05, + "loss": 0.9882, + "step": 29659 + }, + { + "epoch": 1.91, + "learning_rate": 1.3656798127911722e-05, + "loss": 0.9668, + "step": 29660 + }, + { + "epoch": 1.91, + "learning_rate": 1.365658343888877e-05, + "loss": 0.9775, + "step": 29661 + }, + { + "epoch": 1.91, + "learning_rate": 1.365636874986582e-05, + "loss": 0.9251, + "step": 29662 + }, + { + "epoch": 1.91, + "learning_rate": 1.365615406084287e-05, + "loss": 0.9451, + "step": 29663 + }, + { + "epoch": 1.91, + "learning_rate": 1.365593937181992e-05, + "loss": 1.05, + "step": 29664 + }, + { + "epoch": 1.91, + "learning_rate": 1.365572468279697e-05, + "loss": 0.9805, + "step": 29665 + }, + { + "epoch": 1.91, + "learning_rate": 1.3655509993774019e-05, + "loss": 1.0454, + "step": 29666 + }, + { + "epoch": 1.91, + "learning_rate": 1.3655295304751068e-05, + "loss": 0.9421, + "step": 29667 + }, + { + "epoch": 1.91, + "learning_rate": 1.3655080615728118e-05, + "loss": 1.0282, + "step": 29668 + }, + { + "epoch": 1.91, + "learning_rate": 1.3654865926705169e-05, + "loss": 1.0613, + "step": 29669 + }, + { + "epoch": 1.91, + "learning_rate": 1.365465123768222e-05, + "loss": 0.9758, + "step": 29670 + }, + { + "epoch": 1.91, + "learning_rate": 1.3654436548659269e-05, + "loss": 0.9996, + "step": 29671 + }, + { + "epoch": 1.91, + "learning_rate": 1.3654221859636318e-05, + "loss": 0.9966, + "step": 29672 + }, + { + "epoch": 1.91, + "learning_rate": 1.3654007170613368e-05, + "loss": 0.9419, + "step": 29673 + }, + { + "epoch": 1.91, + "learning_rate": 1.3653792481590417e-05, + "loss": 0.9226, + "step": 29674 + }, + { + "epoch": 1.91, + "learning_rate": 1.3653577792567468e-05, + "loss": 0.9535, + "step": 29675 + }, + { + "epoch": 1.91, + "learning_rate": 1.3653363103544517e-05, + "loss": 0.962, + "step": 29676 + }, + { + "epoch": 1.91, + "learning_rate": 1.3653148414521566e-05, + "loss": 0.9336, + "step": 29677 + }, + { + "epoch": 1.91, + "learning_rate": 1.3652933725498617e-05, + "loss": 0.9907, + "step": 29678 + }, + { + "epoch": 1.91, + "learning_rate": 1.3652719036475666e-05, + "loss": 1.0103, + "step": 29679 + }, + { + "epoch": 1.91, + "learning_rate": 1.3652504347452714e-05, + "loss": 0.973, + "step": 29680 + }, + { + "epoch": 1.91, + "learning_rate": 1.3652289658429765e-05, + "loss": 0.9385, + "step": 29681 + }, + { + "epoch": 1.91, + "learning_rate": 1.3652074969406814e-05, + "loss": 0.995, + "step": 29682 + }, + { + "epoch": 1.91, + "learning_rate": 1.3651860280383865e-05, + "loss": 0.9776, + "step": 29683 + }, + { + "epoch": 1.91, + "learning_rate": 1.3651645591360915e-05, + "loss": 0.9849, + "step": 29684 + }, + { + "epoch": 1.91, + "learning_rate": 1.3651430902337966e-05, + "loss": 1.0383, + "step": 29685 + }, + { + "epoch": 1.91, + "learning_rate": 1.3651216213315015e-05, + "loss": 0.9719, + "step": 29686 + }, + { + "epoch": 1.91, + "learning_rate": 1.3651001524292064e-05, + "loss": 0.996, + "step": 29687 + }, + { + "epoch": 1.91, + "learning_rate": 1.3650786835269115e-05, + "loss": 0.9418, + "step": 29688 + }, + { + "epoch": 1.91, + "learning_rate": 1.3650572146246164e-05, + "loss": 0.9127, + "step": 29689 + }, + { + "epoch": 1.91, + "learning_rate": 1.3650357457223213e-05, + "loss": 0.8997, + "step": 29690 + }, + { + "epoch": 1.91, + "learning_rate": 1.3650142768200263e-05, + "loss": 1.0019, + "step": 29691 + }, + { + "epoch": 1.91, + "learning_rate": 1.3649928079177312e-05, + "loss": 0.9931, + "step": 29692 + }, + { + "epoch": 1.91, + "learning_rate": 1.3649713390154363e-05, + "loss": 1.057, + "step": 29693 + }, + { + "epoch": 1.91, + "learning_rate": 1.3649498701131412e-05, + "loss": 1.0057, + "step": 29694 + }, + { + "epoch": 1.91, + "learning_rate": 1.364928401210846e-05, + "loss": 0.9226, + "step": 29695 + }, + { + "epoch": 1.91, + "learning_rate": 1.3649069323085511e-05, + "loss": 0.9429, + "step": 29696 + }, + { + "epoch": 1.91, + "learning_rate": 1.364885463406256e-05, + "loss": 0.9031, + "step": 29697 + }, + { + "epoch": 1.91, + "learning_rate": 1.3648639945039613e-05, + "loss": 0.9721, + "step": 29698 + }, + { + "epoch": 1.91, + "learning_rate": 1.3648425256016662e-05, + "loss": 0.9732, + "step": 29699 + }, + { + "epoch": 1.91, + "learning_rate": 1.364821056699371e-05, + "loss": 1.0084, + "step": 29700 + }, + { + "epoch": 1.91, + "learning_rate": 1.3647995877970761e-05, + "loss": 0.9712, + "step": 29701 + }, + { + "epoch": 1.91, + "learning_rate": 1.364778118894781e-05, + "loss": 1.0442, + "step": 29702 + }, + { + "epoch": 1.91, + "learning_rate": 1.3647566499924861e-05, + "loss": 1.0366, + "step": 29703 + }, + { + "epoch": 1.91, + "learning_rate": 1.364735181090191e-05, + "loss": 1.0082, + "step": 29704 + }, + { + "epoch": 1.91, + "learning_rate": 1.3647137121878959e-05, + "loss": 0.9877, + "step": 29705 + }, + { + "epoch": 1.91, + "learning_rate": 1.364692243285601e-05, + "loss": 0.9948, + "step": 29706 + }, + { + "epoch": 1.91, + "learning_rate": 1.3646707743833059e-05, + "loss": 0.9505, + "step": 29707 + }, + { + "epoch": 1.91, + "learning_rate": 1.3646493054810108e-05, + "loss": 1.0478, + "step": 29708 + }, + { + "epoch": 1.91, + "learning_rate": 1.3646278365787158e-05, + "loss": 0.9694, + "step": 29709 + }, + { + "epoch": 1.91, + "learning_rate": 1.3646063676764207e-05, + "loss": 1.0443, + "step": 29710 + }, + { + "epoch": 1.91, + "learning_rate": 1.3645848987741258e-05, + "loss": 0.9343, + "step": 29711 + }, + { + "epoch": 1.91, + "learning_rate": 1.3645634298718307e-05, + "loss": 0.9476, + "step": 29712 + }, + { + "epoch": 1.91, + "learning_rate": 1.3645419609695359e-05, + "loss": 1.0313, + "step": 29713 + }, + { + "epoch": 1.91, + "learning_rate": 1.3645204920672408e-05, + "loss": 0.9797, + "step": 29714 + }, + { + "epoch": 1.91, + "learning_rate": 1.3644990231649457e-05, + "loss": 1.0123, + "step": 29715 + }, + { + "epoch": 1.91, + "learning_rate": 1.3644775542626508e-05, + "loss": 0.9329, + "step": 29716 + }, + { + "epoch": 1.91, + "learning_rate": 1.3644560853603557e-05, + "loss": 1.0037, + "step": 29717 + }, + { + "epoch": 1.91, + "learning_rate": 1.3644346164580606e-05, + "loss": 0.9915, + "step": 29718 + }, + { + "epoch": 1.91, + "learning_rate": 1.3644131475557656e-05, + "loss": 0.9515, + "step": 29719 + }, + { + "epoch": 1.91, + "learning_rate": 1.3643916786534705e-05, + "loss": 1.0185, + "step": 29720 + }, + { + "epoch": 1.91, + "learning_rate": 1.3643702097511756e-05, + "loss": 0.9248, + "step": 29721 + }, + { + "epoch": 1.91, + "learning_rate": 1.3643487408488805e-05, + "loss": 0.9663, + "step": 29722 + }, + { + "epoch": 1.91, + "learning_rate": 1.3643272719465854e-05, + "loss": 0.9288, + "step": 29723 + }, + { + "epoch": 1.91, + "learning_rate": 1.3643058030442905e-05, + "loss": 0.9694, + "step": 29724 + }, + { + "epoch": 1.91, + "learning_rate": 1.3642843341419953e-05, + "loss": 0.9797, + "step": 29725 + }, + { + "epoch": 1.91, + "learning_rate": 1.3642628652397002e-05, + "loss": 0.9783, + "step": 29726 + }, + { + "epoch": 1.91, + "learning_rate": 1.3642413963374055e-05, + "loss": 0.9296, + "step": 29727 + }, + { + "epoch": 1.91, + "learning_rate": 1.3642199274351104e-05, + "loss": 0.9488, + "step": 29728 + }, + { + "epoch": 1.91, + "learning_rate": 1.3641984585328154e-05, + "loss": 0.9504, + "step": 29729 + }, + { + "epoch": 1.91, + "learning_rate": 1.3641769896305203e-05, + "loss": 1.0458, + "step": 29730 + }, + { + "epoch": 1.91, + "learning_rate": 1.3641555207282254e-05, + "loss": 0.9653, + "step": 29731 + }, + { + "epoch": 1.91, + "learning_rate": 1.3641340518259303e-05, + "loss": 0.9579, + "step": 29732 + }, + { + "epoch": 1.91, + "learning_rate": 1.3641125829236352e-05, + "loss": 1.0057, + "step": 29733 + }, + { + "epoch": 1.91, + "learning_rate": 1.3640911140213403e-05, + "loss": 0.9825, + "step": 29734 + }, + { + "epoch": 1.91, + "learning_rate": 1.3640696451190452e-05, + "loss": 1.0065, + "step": 29735 + }, + { + "epoch": 1.91, + "learning_rate": 1.36404817621675e-05, + "loss": 0.9429, + "step": 29736 + }, + { + "epoch": 1.91, + "learning_rate": 1.3640267073144551e-05, + "loss": 1.0358, + "step": 29737 + }, + { + "epoch": 1.91, + "learning_rate": 1.36400523841216e-05, + "loss": 0.9129, + "step": 29738 + }, + { + "epoch": 1.91, + "learning_rate": 1.3639837695098649e-05, + "loss": 1.0382, + "step": 29739 + }, + { + "epoch": 1.91, + "learning_rate": 1.36396230060757e-05, + "loss": 1.027, + "step": 29740 + }, + { + "epoch": 1.91, + "learning_rate": 1.3639408317052749e-05, + "loss": 0.9583, + "step": 29741 + }, + { + "epoch": 1.91, + "learning_rate": 1.3639193628029801e-05, + "loss": 0.9617, + "step": 29742 + }, + { + "epoch": 1.91, + "learning_rate": 1.363897893900685e-05, + "loss": 1.0394, + "step": 29743 + }, + { + "epoch": 1.91, + "learning_rate": 1.36387642499839e-05, + "loss": 0.9265, + "step": 29744 + }, + { + "epoch": 1.91, + "learning_rate": 1.363854956096095e-05, + "loss": 1.0072, + "step": 29745 + }, + { + "epoch": 1.91, + "learning_rate": 1.3638334871937999e-05, + "loss": 1.0195, + "step": 29746 + }, + { + "epoch": 1.91, + "learning_rate": 1.363812018291505e-05, + "loss": 0.9695, + "step": 29747 + }, + { + "epoch": 1.91, + "learning_rate": 1.3637905493892098e-05, + "loss": 0.9263, + "step": 29748 + }, + { + "epoch": 1.91, + "learning_rate": 1.3637690804869147e-05, + "loss": 0.9621, + "step": 29749 + }, + { + "epoch": 1.91, + "learning_rate": 1.3637476115846198e-05, + "loss": 0.9603, + "step": 29750 + }, + { + "epoch": 1.91, + "learning_rate": 1.3637261426823247e-05, + "loss": 1.0283, + "step": 29751 + }, + { + "epoch": 1.91, + "learning_rate": 1.3637046737800298e-05, + "loss": 0.9521, + "step": 29752 + }, + { + "epoch": 1.91, + "learning_rate": 1.3636832048777347e-05, + "loss": 0.9462, + "step": 29753 + }, + { + "epoch": 1.91, + "learning_rate": 1.3636617359754395e-05, + "loss": 0.9824, + "step": 29754 + }, + { + "epoch": 1.91, + "learning_rate": 1.3636402670731446e-05, + "loss": 0.9244, + "step": 29755 + }, + { + "epoch": 1.91, + "learning_rate": 1.3636187981708497e-05, + "loss": 0.9786, + "step": 29756 + }, + { + "epoch": 1.91, + "learning_rate": 1.3635973292685547e-05, + "loss": 0.903, + "step": 29757 + }, + { + "epoch": 1.91, + "learning_rate": 1.3635758603662596e-05, + "loss": 0.9912, + "step": 29758 + }, + { + "epoch": 1.91, + "learning_rate": 1.3635543914639645e-05, + "loss": 0.9881, + "step": 29759 + }, + { + "epoch": 1.91, + "learning_rate": 1.3635329225616696e-05, + "loss": 0.9444, + "step": 29760 + }, + { + "epoch": 1.91, + "learning_rate": 1.3635114536593745e-05, + "loss": 0.9224, + "step": 29761 + }, + { + "epoch": 1.91, + "learning_rate": 1.3634899847570796e-05, + "loss": 0.9986, + "step": 29762 + }, + { + "epoch": 1.91, + "learning_rate": 1.3634685158547845e-05, + "loss": 0.9555, + "step": 29763 + }, + { + "epoch": 1.91, + "learning_rate": 1.3634470469524894e-05, + "loss": 0.9896, + "step": 29764 + }, + { + "epoch": 1.92, + "learning_rate": 1.3634255780501944e-05, + "loss": 0.9623, + "step": 29765 + }, + { + "epoch": 1.92, + "learning_rate": 1.3634041091478993e-05, + "loss": 0.9249, + "step": 29766 + }, + { + "epoch": 1.92, + "learning_rate": 1.3633826402456042e-05, + "loss": 0.9841, + "step": 29767 + }, + { + "epoch": 1.92, + "learning_rate": 1.3633611713433093e-05, + "loss": 0.9542, + "step": 29768 + }, + { + "epoch": 1.92, + "learning_rate": 1.3633397024410142e-05, + "loss": 0.9667, + "step": 29769 + }, + { + "epoch": 1.92, + "learning_rate": 1.3633182335387192e-05, + "loss": 1.0218, + "step": 29770 + }, + { + "epoch": 1.92, + "learning_rate": 1.3632967646364243e-05, + "loss": 0.9387, + "step": 29771 + }, + { + "epoch": 1.92, + "learning_rate": 1.3632752957341294e-05, + "loss": 0.9919, + "step": 29772 + }, + { + "epoch": 1.92, + "learning_rate": 1.3632538268318343e-05, + "loss": 1.0042, + "step": 29773 + }, + { + "epoch": 1.92, + "learning_rate": 1.3632323579295392e-05, + "loss": 0.9666, + "step": 29774 + }, + { + "epoch": 1.92, + "learning_rate": 1.3632108890272442e-05, + "loss": 0.9679, + "step": 29775 + }, + { + "epoch": 1.92, + "learning_rate": 1.3631894201249491e-05, + "loss": 0.9146, + "step": 29776 + }, + { + "epoch": 1.92, + "learning_rate": 1.363167951222654e-05, + "loss": 1.0076, + "step": 29777 + }, + { + "epoch": 1.92, + "learning_rate": 1.3631464823203591e-05, + "loss": 0.9368, + "step": 29778 + }, + { + "epoch": 1.92, + "learning_rate": 1.363125013418064e-05, + "loss": 0.907, + "step": 29779 + }, + { + "epoch": 1.92, + "learning_rate": 1.363103544515769e-05, + "loss": 0.9161, + "step": 29780 + }, + { + "epoch": 1.92, + "learning_rate": 1.363082075613474e-05, + "loss": 0.9759, + "step": 29781 + }, + { + "epoch": 1.92, + "learning_rate": 1.3630606067111789e-05, + "loss": 0.9073, + "step": 29782 + }, + { + "epoch": 1.92, + "learning_rate": 1.363039137808884e-05, + "loss": 0.9239, + "step": 29783 + }, + { + "epoch": 1.92, + "learning_rate": 1.3630176689065888e-05, + "loss": 0.9422, + "step": 29784 + }, + { + "epoch": 1.92, + "learning_rate": 1.362996200004294e-05, + "loss": 0.9362, + "step": 29785 + }, + { + "epoch": 1.92, + "learning_rate": 1.362974731101999e-05, + "loss": 0.9767, + "step": 29786 + }, + { + "epoch": 1.92, + "learning_rate": 1.3629532621997038e-05, + "loss": 0.9202, + "step": 29787 + }, + { + "epoch": 1.92, + "learning_rate": 1.3629317932974089e-05, + "loss": 0.9508, + "step": 29788 + }, + { + "epoch": 1.92, + "learning_rate": 1.3629103243951138e-05, + "loss": 0.9638, + "step": 29789 + }, + { + "epoch": 1.92, + "learning_rate": 1.3628888554928189e-05, + "loss": 1.006, + "step": 29790 + }, + { + "epoch": 1.92, + "learning_rate": 1.3628673865905238e-05, + "loss": 0.9555, + "step": 29791 + }, + { + "epoch": 1.92, + "learning_rate": 1.3628459176882287e-05, + "loss": 0.9985, + "step": 29792 + }, + { + "epoch": 1.92, + "learning_rate": 1.3628244487859337e-05, + "loss": 0.914, + "step": 29793 + }, + { + "epoch": 1.92, + "learning_rate": 1.3628029798836386e-05, + "loss": 1.0157, + "step": 29794 + }, + { + "epoch": 1.92, + "learning_rate": 1.3627815109813435e-05, + "loss": 0.96, + "step": 29795 + }, + { + "epoch": 1.92, + "learning_rate": 1.3627600420790486e-05, + "loss": 1.0075, + "step": 29796 + }, + { + "epoch": 1.92, + "learning_rate": 1.3627385731767535e-05, + "loss": 0.9843, + "step": 29797 + }, + { + "epoch": 1.92, + "learning_rate": 1.3627171042744586e-05, + "loss": 1.0026, + "step": 29798 + }, + { + "epoch": 1.92, + "learning_rate": 1.3626956353721634e-05, + "loss": 1.0045, + "step": 29799 + }, + { + "epoch": 1.92, + "learning_rate": 1.3626741664698687e-05, + "loss": 0.9613, + "step": 29800 + }, + { + "epoch": 1.92, + "learning_rate": 1.3626526975675736e-05, + "loss": 1.0261, + "step": 29801 + }, + { + "epoch": 1.92, + "learning_rate": 1.3626312286652785e-05, + "loss": 0.8578, + "step": 29802 + }, + { + "epoch": 1.92, + "learning_rate": 1.3626097597629835e-05, + "loss": 0.9797, + "step": 29803 + }, + { + "epoch": 1.92, + "learning_rate": 1.3625882908606884e-05, + "loss": 0.883, + "step": 29804 + }, + { + "epoch": 1.92, + "learning_rate": 1.3625668219583933e-05, + "loss": 1.0024, + "step": 29805 + }, + { + "epoch": 1.92, + "learning_rate": 1.3625453530560984e-05, + "loss": 1.0274, + "step": 29806 + }, + { + "epoch": 1.92, + "learning_rate": 1.3625238841538033e-05, + "loss": 0.9697, + "step": 29807 + }, + { + "epoch": 1.92, + "learning_rate": 1.3625024152515084e-05, + "loss": 1.0709, + "step": 29808 + }, + { + "epoch": 1.92, + "learning_rate": 1.3624809463492133e-05, + "loss": 1.0001, + "step": 29809 + }, + { + "epoch": 1.92, + "learning_rate": 1.3624594774469182e-05, + "loss": 1.0307, + "step": 29810 + }, + { + "epoch": 1.92, + "learning_rate": 1.3624380085446232e-05, + "loss": 1.0029, + "step": 29811 + }, + { + "epoch": 1.92, + "learning_rate": 1.3624165396423281e-05, + "loss": 0.946, + "step": 29812 + }, + { + "epoch": 1.92, + "learning_rate": 1.362395070740033e-05, + "loss": 1.0004, + "step": 29813 + }, + { + "epoch": 1.92, + "learning_rate": 1.3623736018377383e-05, + "loss": 0.9834, + "step": 29814 + }, + { + "epoch": 1.92, + "learning_rate": 1.3623521329354431e-05, + "loss": 1.0157, + "step": 29815 + }, + { + "epoch": 1.92, + "learning_rate": 1.3623306640331482e-05, + "loss": 0.9521, + "step": 29816 + }, + { + "epoch": 1.92, + "learning_rate": 1.3623091951308531e-05, + "loss": 0.952, + "step": 29817 + }, + { + "epoch": 1.92, + "learning_rate": 1.3622877262285582e-05, + "loss": 0.9528, + "step": 29818 + }, + { + "epoch": 1.92, + "learning_rate": 1.362266257326263e-05, + "loss": 0.9732, + "step": 29819 + }, + { + "epoch": 1.92, + "learning_rate": 1.362244788423968e-05, + "loss": 0.9397, + "step": 29820 + }, + { + "epoch": 1.92, + "learning_rate": 1.362223319521673e-05, + "loss": 0.9565, + "step": 29821 + }, + { + "epoch": 1.92, + "learning_rate": 1.362201850619378e-05, + "loss": 0.9663, + "step": 29822 + }, + { + "epoch": 1.92, + "learning_rate": 1.3621803817170828e-05, + "loss": 0.9869, + "step": 29823 + }, + { + "epoch": 1.92, + "learning_rate": 1.3621589128147879e-05, + "loss": 0.973, + "step": 29824 + }, + { + "epoch": 1.92, + "learning_rate": 1.3621374439124928e-05, + "loss": 1.0307, + "step": 29825 + }, + { + "epoch": 1.92, + "learning_rate": 1.3621159750101977e-05, + "loss": 0.9114, + "step": 29826 + }, + { + "epoch": 1.92, + "learning_rate": 1.3620945061079028e-05, + "loss": 1.0027, + "step": 29827 + }, + { + "epoch": 1.92, + "learning_rate": 1.3620730372056076e-05, + "loss": 0.9699, + "step": 29828 + }, + { + "epoch": 1.92, + "learning_rate": 1.3620515683033129e-05, + "loss": 0.9885, + "step": 29829 + }, + { + "epoch": 1.92, + "learning_rate": 1.3620300994010178e-05, + "loss": 0.9695, + "step": 29830 + }, + { + "epoch": 1.92, + "learning_rate": 1.3620086304987228e-05, + "loss": 0.9059, + "step": 29831 + }, + { + "epoch": 1.92, + "learning_rate": 1.3619871615964277e-05, + "loss": 0.9511, + "step": 29832 + }, + { + "epoch": 1.92, + "learning_rate": 1.3619656926941326e-05, + "loss": 0.9365, + "step": 29833 + }, + { + "epoch": 1.92, + "learning_rate": 1.3619442237918377e-05, + "loss": 0.9911, + "step": 29834 + }, + { + "epoch": 1.92, + "learning_rate": 1.3619227548895426e-05, + "loss": 1.0033, + "step": 29835 + }, + { + "epoch": 1.92, + "learning_rate": 1.3619012859872475e-05, + "loss": 1.0663, + "step": 29836 + }, + { + "epoch": 1.92, + "learning_rate": 1.3618798170849526e-05, + "loss": 0.8409, + "step": 29837 + }, + { + "epoch": 1.92, + "learning_rate": 1.3618583481826575e-05, + "loss": 0.9136, + "step": 29838 + }, + { + "epoch": 1.92, + "learning_rate": 1.3618368792803625e-05, + "loss": 1.0432, + "step": 29839 + }, + { + "epoch": 1.92, + "learning_rate": 1.3618154103780674e-05, + "loss": 1.0162, + "step": 29840 + }, + { + "epoch": 1.92, + "learning_rate": 1.3617939414757723e-05, + "loss": 0.9392, + "step": 29841 + }, + { + "epoch": 1.92, + "learning_rate": 1.3617724725734774e-05, + "loss": 0.9501, + "step": 29842 + }, + { + "epoch": 1.92, + "learning_rate": 1.3617510036711825e-05, + "loss": 0.8774, + "step": 29843 + }, + { + "epoch": 1.92, + "learning_rate": 1.3617295347688875e-05, + "loss": 0.9071, + "step": 29844 + }, + { + "epoch": 1.92, + "learning_rate": 1.3617080658665924e-05, + "loss": 0.9391, + "step": 29845 + }, + { + "epoch": 1.92, + "learning_rate": 1.3616865969642973e-05, + "loss": 0.9491, + "step": 29846 + }, + { + "epoch": 1.92, + "learning_rate": 1.3616651280620024e-05, + "loss": 0.9981, + "step": 29847 + }, + { + "epoch": 1.92, + "learning_rate": 1.3616436591597073e-05, + "loss": 1.0236, + "step": 29848 + }, + { + "epoch": 1.92, + "learning_rate": 1.3616221902574123e-05, + "loss": 0.8881, + "step": 29849 + }, + { + "epoch": 1.92, + "learning_rate": 1.3616007213551172e-05, + "loss": 0.9444, + "step": 29850 + }, + { + "epoch": 1.92, + "learning_rate": 1.3615792524528221e-05, + "loss": 1.0029, + "step": 29851 + }, + { + "epoch": 1.92, + "learning_rate": 1.3615577835505272e-05, + "loss": 0.9761, + "step": 29852 + }, + { + "epoch": 1.92, + "learning_rate": 1.3615363146482321e-05, + "loss": 1.0025, + "step": 29853 + }, + { + "epoch": 1.92, + "learning_rate": 1.361514845745937e-05, + "loss": 1.0211, + "step": 29854 + }, + { + "epoch": 1.92, + "learning_rate": 1.361493376843642e-05, + "loss": 0.9149, + "step": 29855 + }, + { + "epoch": 1.92, + "learning_rate": 1.361471907941347e-05, + "loss": 0.955, + "step": 29856 + }, + { + "epoch": 1.92, + "learning_rate": 1.361450439039052e-05, + "loss": 0.987, + "step": 29857 + }, + { + "epoch": 1.92, + "learning_rate": 1.361428970136757e-05, + "loss": 0.9743, + "step": 29858 + }, + { + "epoch": 1.92, + "learning_rate": 1.3614075012344622e-05, + "loss": 0.942, + "step": 29859 + }, + { + "epoch": 1.92, + "learning_rate": 1.361386032332167e-05, + "loss": 0.9567, + "step": 29860 + }, + { + "epoch": 1.92, + "learning_rate": 1.361364563429872e-05, + "loss": 0.999, + "step": 29861 + }, + { + "epoch": 1.92, + "learning_rate": 1.361343094527577e-05, + "loss": 0.9795, + "step": 29862 + }, + { + "epoch": 1.92, + "learning_rate": 1.3613216256252819e-05, + "loss": 0.9281, + "step": 29863 + }, + { + "epoch": 1.92, + "learning_rate": 1.3613001567229868e-05, + "loss": 0.9516, + "step": 29864 + }, + { + "epoch": 1.92, + "learning_rate": 1.3612786878206919e-05, + "loss": 0.9725, + "step": 29865 + }, + { + "epoch": 1.92, + "learning_rate": 1.3612572189183968e-05, + "loss": 0.9389, + "step": 29866 + }, + { + "epoch": 1.92, + "learning_rate": 1.3612357500161018e-05, + "loss": 0.9296, + "step": 29867 + }, + { + "epoch": 1.92, + "learning_rate": 1.3612142811138067e-05, + "loss": 0.9285, + "step": 29868 + }, + { + "epoch": 1.92, + "learning_rate": 1.3611928122115116e-05, + "loss": 0.9813, + "step": 29869 + }, + { + "epoch": 1.92, + "learning_rate": 1.3611713433092167e-05, + "loss": 0.9822, + "step": 29870 + }, + { + "epoch": 1.92, + "learning_rate": 1.3611498744069216e-05, + "loss": 0.9073, + "step": 29871 + }, + { + "epoch": 1.92, + "learning_rate": 1.3611284055046268e-05, + "loss": 0.996, + "step": 29872 + }, + { + "epoch": 1.92, + "learning_rate": 1.3611069366023317e-05, + "loss": 0.9038, + "step": 29873 + }, + { + "epoch": 1.92, + "learning_rate": 1.3610854677000366e-05, + "loss": 0.9416, + "step": 29874 + }, + { + "epoch": 1.92, + "learning_rate": 1.3610639987977417e-05, + "loss": 0.8985, + "step": 29875 + }, + { + "epoch": 1.92, + "learning_rate": 1.3610425298954466e-05, + "loss": 0.937, + "step": 29876 + }, + { + "epoch": 1.92, + "learning_rate": 1.3610210609931516e-05, + "loss": 0.9733, + "step": 29877 + }, + { + "epoch": 1.92, + "learning_rate": 1.3609995920908565e-05, + "loss": 0.9611, + "step": 29878 + }, + { + "epoch": 1.92, + "learning_rate": 1.3609781231885614e-05, + "loss": 0.9747, + "step": 29879 + }, + { + "epoch": 1.92, + "learning_rate": 1.3609566542862665e-05, + "loss": 0.8804, + "step": 29880 + }, + { + "epoch": 1.92, + "learning_rate": 1.3609351853839714e-05, + "loss": 0.9547, + "step": 29881 + }, + { + "epoch": 1.92, + "learning_rate": 1.3609137164816763e-05, + "loss": 0.8946, + "step": 29882 + }, + { + "epoch": 1.92, + "learning_rate": 1.3608922475793814e-05, + "loss": 0.9954, + "step": 29883 + }, + { + "epoch": 1.92, + "learning_rate": 1.3608707786770863e-05, + "loss": 0.9589, + "step": 29884 + }, + { + "epoch": 1.92, + "learning_rate": 1.3608493097747913e-05, + "loss": 0.9774, + "step": 29885 + }, + { + "epoch": 1.92, + "learning_rate": 1.3608278408724962e-05, + "loss": 0.9495, + "step": 29886 + }, + { + "epoch": 1.92, + "learning_rate": 1.3608063719702015e-05, + "loss": 1.0092, + "step": 29887 + }, + { + "epoch": 1.92, + "learning_rate": 1.3607849030679064e-05, + "loss": 0.8919, + "step": 29888 + }, + { + "epoch": 1.92, + "learning_rate": 1.3607634341656112e-05, + "loss": 0.9828, + "step": 29889 + }, + { + "epoch": 1.92, + "learning_rate": 1.3607419652633163e-05, + "loss": 0.9441, + "step": 29890 + }, + { + "epoch": 1.92, + "learning_rate": 1.3607204963610212e-05, + "loss": 0.9735, + "step": 29891 + }, + { + "epoch": 1.92, + "learning_rate": 1.3606990274587261e-05, + "loss": 0.974, + "step": 29892 + }, + { + "epoch": 1.92, + "learning_rate": 1.3606775585564312e-05, + "loss": 0.9836, + "step": 29893 + }, + { + "epoch": 1.92, + "learning_rate": 1.360656089654136e-05, + "loss": 0.9195, + "step": 29894 + }, + { + "epoch": 1.92, + "learning_rate": 1.3606346207518411e-05, + "loss": 0.911, + "step": 29895 + }, + { + "epoch": 1.92, + "learning_rate": 1.360613151849546e-05, + "loss": 1.0136, + "step": 29896 + }, + { + "epoch": 1.92, + "learning_rate": 1.360591682947251e-05, + "loss": 0.9403, + "step": 29897 + }, + { + "epoch": 1.92, + "learning_rate": 1.360570214044956e-05, + "loss": 1.0092, + "step": 29898 + }, + { + "epoch": 1.92, + "learning_rate": 1.3605487451426609e-05, + "loss": 1.0188, + "step": 29899 + }, + { + "epoch": 1.92, + "learning_rate": 1.3605272762403658e-05, + "loss": 0.9366, + "step": 29900 + }, + { + "epoch": 1.92, + "learning_rate": 1.3605058073380709e-05, + "loss": 1.0229, + "step": 29901 + }, + { + "epoch": 1.92, + "learning_rate": 1.360484338435776e-05, + "loss": 0.9026, + "step": 29902 + }, + { + "epoch": 1.92, + "learning_rate": 1.360462869533481e-05, + "loss": 0.9414, + "step": 29903 + }, + { + "epoch": 1.92, + "learning_rate": 1.3604414006311859e-05, + "loss": 1.0005, + "step": 29904 + }, + { + "epoch": 1.92, + "learning_rate": 1.360419931728891e-05, + "loss": 0.9351, + "step": 29905 + }, + { + "epoch": 1.92, + "learning_rate": 1.3603984628265958e-05, + "loss": 0.9669, + "step": 29906 + }, + { + "epoch": 1.92, + "learning_rate": 1.3603769939243007e-05, + "loss": 0.9479, + "step": 29907 + }, + { + "epoch": 1.92, + "learning_rate": 1.3603555250220058e-05, + "loss": 0.9651, + "step": 29908 + }, + { + "epoch": 1.92, + "learning_rate": 1.3603340561197107e-05, + "loss": 1.0422, + "step": 29909 + }, + { + "epoch": 1.92, + "learning_rate": 1.3603125872174156e-05, + "loss": 1.0094, + "step": 29910 + }, + { + "epoch": 1.92, + "learning_rate": 1.3602911183151207e-05, + "loss": 0.9817, + "step": 29911 + }, + { + "epoch": 1.92, + "learning_rate": 1.3602696494128256e-05, + "loss": 0.9809, + "step": 29912 + }, + { + "epoch": 1.92, + "learning_rate": 1.3602481805105305e-05, + "loss": 0.9964, + "step": 29913 + }, + { + "epoch": 1.92, + "learning_rate": 1.3602267116082355e-05, + "loss": 0.9751, + "step": 29914 + }, + { + "epoch": 1.92, + "learning_rate": 1.3602052427059404e-05, + "loss": 0.9774, + "step": 29915 + }, + { + "epoch": 1.92, + "learning_rate": 1.3601837738036457e-05, + "loss": 0.9081, + "step": 29916 + }, + { + "epoch": 1.92, + "learning_rate": 1.3601623049013506e-05, + "loss": 0.9716, + "step": 29917 + }, + { + "epoch": 1.92, + "learning_rate": 1.3601408359990556e-05, + "loss": 0.9346, + "step": 29918 + }, + { + "epoch": 1.92, + "learning_rate": 1.3601193670967605e-05, + "loss": 0.9123, + "step": 29919 + }, + { + "epoch": 1.92, + "learning_rate": 1.3600978981944654e-05, + "loss": 0.9411, + "step": 29920 + }, + { + "epoch": 1.93, + "learning_rate": 1.3600764292921705e-05, + "loss": 0.9475, + "step": 29921 + }, + { + "epoch": 1.93, + "learning_rate": 1.3600549603898754e-05, + "loss": 1.0377, + "step": 29922 + }, + { + "epoch": 1.93, + "learning_rate": 1.3600334914875803e-05, + "loss": 0.9669, + "step": 29923 + }, + { + "epoch": 1.93, + "learning_rate": 1.3600120225852853e-05, + "loss": 0.9438, + "step": 29924 + }, + { + "epoch": 1.93, + "learning_rate": 1.3599905536829902e-05, + "loss": 0.9344, + "step": 29925 + }, + { + "epoch": 1.93, + "learning_rate": 1.3599690847806953e-05, + "loss": 0.9994, + "step": 29926 + }, + { + "epoch": 1.93, + "learning_rate": 1.3599476158784002e-05, + "loss": 0.9029, + "step": 29927 + }, + { + "epoch": 1.93, + "learning_rate": 1.3599261469761051e-05, + "loss": 0.8945, + "step": 29928 + }, + { + "epoch": 1.93, + "learning_rate": 1.3599046780738102e-05, + "loss": 0.9025, + "step": 29929 + }, + { + "epoch": 1.93, + "learning_rate": 1.359883209171515e-05, + "loss": 0.8872, + "step": 29930 + }, + { + "epoch": 1.93, + "learning_rate": 1.3598617402692203e-05, + "loss": 0.9972, + "step": 29931 + }, + { + "epoch": 1.93, + "learning_rate": 1.3598402713669252e-05, + "loss": 0.9615, + "step": 29932 + }, + { + "epoch": 1.93, + "learning_rate": 1.35981880246463e-05, + "loss": 0.9679, + "step": 29933 + }, + { + "epoch": 1.93, + "learning_rate": 1.3597973335623351e-05, + "loss": 0.9671, + "step": 29934 + }, + { + "epoch": 1.93, + "learning_rate": 1.35977586466004e-05, + "loss": 0.9651, + "step": 29935 + }, + { + "epoch": 1.93, + "learning_rate": 1.3597543957577451e-05, + "loss": 1.0043, + "step": 29936 + }, + { + "epoch": 1.93, + "learning_rate": 1.35973292685545e-05, + "loss": 1.0142, + "step": 29937 + }, + { + "epoch": 1.93, + "learning_rate": 1.3597114579531549e-05, + "loss": 0.9687, + "step": 29938 + }, + { + "epoch": 1.93, + "learning_rate": 1.35968998905086e-05, + "loss": 0.9327, + "step": 29939 + }, + { + "epoch": 1.93, + "learning_rate": 1.3596685201485649e-05, + "loss": 0.9553, + "step": 29940 + }, + { + "epoch": 1.93, + "learning_rate": 1.3596470512462698e-05, + "loss": 1.0087, + "step": 29941 + }, + { + "epoch": 1.93, + "learning_rate": 1.3596255823439748e-05, + "loss": 0.9636, + "step": 29942 + }, + { + "epoch": 1.93, + "learning_rate": 1.3596041134416797e-05, + "loss": 1.0111, + "step": 29943 + }, + { + "epoch": 1.93, + "learning_rate": 1.3595826445393848e-05, + "loss": 0.9157, + "step": 29944 + }, + { + "epoch": 1.93, + "learning_rate": 1.3595611756370899e-05, + "loss": 0.9017, + "step": 29945 + }, + { + "epoch": 1.93, + "learning_rate": 1.359539706734795e-05, + "loss": 0.9143, + "step": 29946 + }, + { + "epoch": 1.93, + "learning_rate": 1.3595182378324998e-05, + "loss": 1.0042, + "step": 29947 + }, + { + "epoch": 1.93, + "learning_rate": 1.3594967689302047e-05, + "loss": 0.9926, + "step": 29948 + }, + { + "epoch": 1.93, + "learning_rate": 1.3594753000279098e-05, + "loss": 0.9807, + "step": 29949 + }, + { + "epoch": 1.93, + "learning_rate": 1.3594538311256147e-05, + "loss": 0.9762, + "step": 29950 + }, + { + "epoch": 1.93, + "learning_rate": 1.3594323622233196e-05, + "loss": 0.9613, + "step": 29951 + }, + { + "epoch": 1.93, + "learning_rate": 1.3594108933210246e-05, + "loss": 0.9854, + "step": 29952 + }, + { + "epoch": 1.93, + "learning_rate": 1.3593894244187295e-05, + "loss": 0.9436, + "step": 29953 + }, + { + "epoch": 1.93, + "learning_rate": 1.3593679555164346e-05, + "loss": 1.0188, + "step": 29954 + }, + { + "epoch": 1.93, + "learning_rate": 1.3593464866141395e-05, + "loss": 0.9712, + "step": 29955 + }, + { + "epoch": 1.93, + "learning_rate": 1.3593250177118444e-05, + "loss": 0.9571, + "step": 29956 + }, + { + "epoch": 1.93, + "learning_rate": 1.3593035488095495e-05, + "loss": 1.0193, + "step": 29957 + }, + { + "epoch": 1.93, + "learning_rate": 1.3592820799072544e-05, + "loss": 1.007, + "step": 29958 + }, + { + "epoch": 1.93, + "learning_rate": 1.3592606110049593e-05, + "loss": 0.9632, + "step": 29959 + }, + { + "epoch": 1.93, + "learning_rate": 1.3592391421026645e-05, + "loss": 0.9471, + "step": 29960 + }, + { + "epoch": 1.93, + "learning_rate": 1.3592176732003694e-05, + "loss": 0.9321, + "step": 29961 + }, + { + "epoch": 1.93, + "learning_rate": 1.3591962042980745e-05, + "loss": 0.9326, + "step": 29962 + }, + { + "epoch": 1.93, + "learning_rate": 1.3591747353957793e-05, + "loss": 1.0013, + "step": 29963 + }, + { + "epoch": 1.93, + "learning_rate": 1.3591532664934844e-05, + "loss": 0.9432, + "step": 29964 + }, + { + "epoch": 1.93, + "learning_rate": 1.3591317975911893e-05, + "loss": 1.0145, + "step": 29965 + }, + { + "epoch": 1.93, + "learning_rate": 1.3591103286888942e-05, + "loss": 1.0444, + "step": 29966 + }, + { + "epoch": 1.93, + "learning_rate": 1.3590888597865993e-05, + "loss": 0.9247, + "step": 29967 + }, + { + "epoch": 1.93, + "learning_rate": 1.3590673908843042e-05, + "loss": 1.0114, + "step": 29968 + }, + { + "epoch": 1.93, + "learning_rate": 1.359045921982009e-05, + "loss": 0.9344, + "step": 29969 + }, + { + "epoch": 1.93, + "learning_rate": 1.3590244530797141e-05, + "loss": 0.9063, + "step": 29970 + }, + { + "epoch": 1.93, + "learning_rate": 1.359002984177419e-05, + "loss": 0.9818, + "step": 29971 + }, + { + "epoch": 1.93, + "learning_rate": 1.358981515275124e-05, + "loss": 0.9673, + "step": 29972 + }, + { + "epoch": 1.93, + "learning_rate": 1.358960046372829e-05, + "loss": 0.929, + "step": 29973 + }, + { + "epoch": 1.93, + "learning_rate": 1.3589385774705342e-05, + "loss": 0.9564, + "step": 29974 + }, + { + "epoch": 1.93, + "learning_rate": 1.3589171085682391e-05, + "loss": 0.9762, + "step": 29975 + }, + { + "epoch": 1.93, + "learning_rate": 1.358895639665944e-05, + "loss": 0.9466, + "step": 29976 + }, + { + "epoch": 1.93, + "learning_rate": 1.358874170763649e-05, + "loss": 0.9496, + "step": 29977 + }, + { + "epoch": 1.93, + "learning_rate": 1.358852701861354e-05, + "loss": 0.9845, + "step": 29978 + }, + { + "epoch": 1.93, + "learning_rate": 1.3588312329590589e-05, + "loss": 1.084, + "step": 29979 + }, + { + "epoch": 1.93, + "learning_rate": 1.358809764056764e-05, + "loss": 0.8996, + "step": 29980 + }, + { + "epoch": 1.93, + "learning_rate": 1.3587882951544688e-05, + "loss": 0.98, + "step": 29981 + }, + { + "epoch": 1.93, + "learning_rate": 1.3587668262521737e-05, + "loss": 0.921, + "step": 29982 + }, + { + "epoch": 1.93, + "learning_rate": 1.3587453573498788e-05, + "loss": 1.0157, + "step": 29983 + }, + { + "epoch": 1.93, + "learning_rate": 1.3587238884475837e-05, + "loss": 1.0038, + "step": 29984 + }, + { + "epoch": 1.93, + "learning_rate": 1.3587024195452888e-05, + "loss": 1.0352, + "step": 29985 + }, + { + "epoch": 1.93, + "learning_rate": 1.3586809506429937e-05, + "loss": 0.8636, + "step": 29986 + }, + { + "epoch": 1.93, + "learning_rate": 1.3586594817406986e-05, + "loss": 0.9563, + "step": 29987 + }, + { + "epoch": 1.93, + "learning_rate": 1.3586380128384036e-05, + "loss": 0.9758, + "step": 29988 + }, + { + "epoch": 1.93, + "learning_rate": 1.3586165439361087e-05, + "loss": 1.0723, + "step": 29989 + }, + { + "epoch": 1.93, + "learning_rate": 1.3585950750338138e-05, + "loss": 0.9469, + "step": 29990 + }, + { + "epoch": 1.93, + "learning_rate": 1.3585736061315187e-05, + "loss": 0.9161, + "step": 29991 + }, + { + "epoch": 1.93, + "learning_rate": 1.3585521372292235e-05, + "loss": 0.9811, + "step": 29992 + }, + { + "epoch": 1.93, + "learning_rate": 1.3585306683269286e-05, + "loss": 0.9291, + "step": 29993 + }, + { + "epoch": 1.93, + "learning_rate": 1.3585091994246335e-05, + "loss": 1.0002, + "step": 29994 + }, + { + "epoch": 1.93, + "learning_rate": 1.3584877305223386e-05, + "loss": 0.8929, + "step": 29995 + }, + { + "epoch": 1.93, + "learning_rate": 1.3584662616200435e-05, + "loss": 1.0174, + "step": 29996 + }, + { + "epoch": 1.93, + "learning_rate": 1.3584447927177484e-05, + "loss": 0.9873, + "step": 29997 + }, + { + "epoch": 1.93, + "learning_rate": 1.3584233238154534e-05, + "loss": 0.912, + "step": 29998 + }, + { + "epoch": 1.93, + "learning_rate": 1.3584018549131583e-05, + "loss": 1.0005, + "step": 29999 + }, + { + "epoch": 1.93, + "learning_rate": 1.3583803860108632e-05, + "loss": 0.9206, + "step": 30000 + }, + { + "epoch": 1.93, + "eval_laion/cider": 0.585307645656223, + "eval_laion/meteor": 0.40609586353161314, + "eval_laion/pred_num_words": 39.373046875, + "eval_laion/sacrebleu": 16.67262725542426, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0399773120880127, + "eval_runtime": 99.85, + "eval_samples_per_second": 4.997, + "eval_steps_per_second": 0.32, + "step": 30000 + }, + { + "epoch": 1.93, + "learning_rate": 1.3583589171085683e-05, + "loss": 0.9489, + "step": 30001 + }, + { + "epoch": 1.93, + "learning_rate": 1.3583374482062732e-05, + "loss": 0.9957, + "step": 30002 + }, + { + "epoch": 1.93, + "learning_rate": 1.3583159793039784e-05, + "loss": 1.0133, + "step": 30003 + }, + { + "epoch": 1.93, + "learning_rate": 1.3582945104016833e-05, + "loss": 1.0101, + "step": 30004 + }, + { + "epoch": 1.93, + "learning_rate": 1.3582730414993884e-05, + "loss": 0.9933, + "step": 30005 + }, + { + "epoch": 1.93, + "learning_rate": 1.3582515725970933e-05, + "loss": 0.9527, + "step": 30006 + }, + { + "epoch": 1.93, + "learning_rate": 1.3582301036947982e-05, + "loss": 0.9641, + "step": 30007 + }, + { + "epoch": 1.93, + "learning_rate": 1.3582086347925032e-05, + "loss": 0.8727, + "step": 30008 + }, + { + "epoch": 1.93, + "learning_rate": 1.3581871658902081e-05, + "loss": 0.9332, + "step": 30009 + }, + { + "epoch": 1.93, + "learning_rate": 1.358165696987913e-05, + "loss": 1.013, + "step": 30010 + }, + { + "epoch": 1.93, + "learning_rate": 1.3581442280856181e-05, + "loss": 0.949, + "step": 30011 + }, + { + "epoch": 1.93, + "learning_rate": 1.358122759183323e-05, + "loss": 0.941, + "step": 30012 + }, + { + "epoch": 1.93, + "learning_rate": 1.358101290281028e-05, + "loss": 0.9853, + "step": 30013 + }, + { + "epoch": 1.93, + "learning_rate": 1.358079821378733e-05, + "loss": 0.9451, + "step": 30014 + }, + { + "epoch": 1.93, + "learning_rate": 1.3580583524764379e-05, + "loss": 1.0597, + "step": 30015 + }, + { + "epoch": 1.93, + "learning_rate": 1.358036883574143e-05, + "loss": 0.9476, + "step": 30016 + }, + { + "epoch": 1.93, + "learning_rate": 1.3580154146718478e-05, + "loss": 1.028, + "step": 30017 + }, + { + "epoch": 1.93, + "learning_rate": 1.357993945769553e-05, + "loss": 1.0396, + "step": 30018 + }, + { + "epoch": 1.93, + "learning_rate": 1.357972476867258e-05, + "loss": 0.9523, + "step": 30019 + }, + { + "epoch": 1.93, + "learning_rate": 1.3579510079649629e-05, + "loss": 0.9939, + "step": 30020 + }, + { + "epoch": 1.93, + "learning_rate": 1.357929539062668e-05, + "loss": 1.0153, + "step": 30021 + }, + { + "epoch": 1.93, + "learning_rate": 1.3579080701603728e-05, + "loss": 1.035, + "step": 30022 + }, + { + "epoch": 1.93, + "learning_rate": 1.3578866012580779e-05, + "loss": 0.9747, + "step": 30023 + }, + { + "epoch": 1.93, + "learning_rate": 1.3578651323557828e-05, + "loss": 0.9693, + "step": 30024 + }, + { + "epoch": 1.93, + "learning_rate": 1.3578436634534877e-05, + "loss": 0.999, + "step": 30025 + }, + { + "epoch": 1.93, + "learning_rate": 1.3578221945511927e-05, + "loss": 0.975, + "step": 30026 + }, + { + "epoch": 1.93, + "learning_rate": 1.3578007256488976e-05, + "loss": 0.9503, + "step": 30027 + }, + { + "epoch": 1.93, + "learning_rate": 1.3577792567466025e-05, + "loss": 1.0871, + "step": 30028 + }, + { + "epoch": 1.93, + "learning_rate": 1.3577577878443076e-05, + "loss": 1.0247, + "step": 30029 + }, + { + "epoch": 1.93, + "learning_rate": 1.3577363189420125e-05, + "loss": 0.9298, + "step": 30030 + }, + { + "epoch": 1.93, + "learning_rate": 1.3577148500397176e-05, + "loss": 0.9306, + "step": 30031 + }, + { + "epoch": 1.93, + "learning_rate": 1.3576933811374226e-05, + "loss": 0.9232, + "step": 30032 + }, + { + "epoch": 1.93, + "learning_rate": 1.3576719122351277e-05, + "loss": 0.975, + "step": 30033 + }, + { + "epoch": 1.93, + "learning_rate": 1.3576504433328326e-05, + "loss": 1.0095, + "step": 30034 + }, + { + "epoch": 1.93, + "learning_rate": 1.3576289744305375e-05, + "loss": 0.9912, + "step": 30035 + }, + { + "epoch": 1.93, + "learning_rate": 1.3576075055282426e-05, + "loss": 1.0415, + "step": 30036 + }, + { + "epoch": 1.93, + "learning_rate": 1.3575860366259474e-05, + "loss": 1.0373, + "step": 30037 + }, + { + "epoch": 1.93, + "learning_rate": 1.3575645677236523e-05, + "loss": 1.0202, + "step": 30038 + }, + { + "epoch": 1.93, + "learning_rate": 1.3575430988213574e-05, + "loss": 0.9894, + "step": 30039 + }, + { + "epoch": 1.93, + "learning_rate": 1.3575216299190623e-05, + "loss": 1.0221, + "step": 30040 + }, + { + "epoch": 1.93, + "learning_rate": 1.3575001610167674e-05, + "loss": 1.0048, + "step": 30041 + }, + { + "epoch": 1.93, + "learning_rate": 1.3574786921144723e-05, + "loss": 1.0315, + "step": 30042 + }, + { + "epoch": 1.93, + "learning_rate": 1.3574572232121772e-05, + "loss": 0.9329, + "step": 30043 + }, + { + "epoch": 1.93, + "learning_rate": 1.3574357543098822e-05, + "loss": 0.9045, + "step": 30044 + }, + { + "epoch": 1.93, + "learning_rate": 1.3574142854075871e-05, + "loss": 0.9833, + "step": 30045 + }, + { + "epoch": 1.93, + "learning_rate": 1.357392816505292e-05, + "loss": 0.9856, + "step": 30046 + }, + { + "epoch": 1.93, + "learning_rate": 1.3573713476029973e-05, + "loss": 0.9429, + "step": 30047 + }, + { + "epoch": 1.93, + "learning_rate": 1.3573498787007022e-05, + "loss": 1.0088, + "step": 30048 + }, + { + "epoch": 1.93, + "learning_rate": 1.3573284097984072e-05, + "loss": 1.0058, + "step": 30049 + }, + { + "epoch": 1.93, + "learning_rate": 1.3573069408961121e-05, + "loss": 0.9626, + "step": 30050 + }, + { + "epoch": 1.93, + "learning_rate": 1.3572854719938172e-05, + "loss": 0.9231, + "step": 30051 + }, + { + "epoch": 1.93, + "learning_rate": 1.357264003091522e-05, + "loss": 0.9923, + "step": 30052 + }, + { + "epoch": 1.93, + "learning_rate": 1.357242534189227e-05, + "loss": 0.9738, + "step": 30053 + }, + { + "epoch": 1.93, + "learning_rate": 1.357221065286932e-05, + "loss": 0.9611, + "step": 30054 + }, + { + "epoch": 1.93, + "learning_rate": 1.357199596384637e-05, + "loss": 1.0292, + "step": 30055 + }, + { + "epoch": 1.93, + "learning_rate": 1.3571781274823418e-05, + "loss": 0.9911, + "step": 30056 + }, + { + "epoch": 1.93, + "learning_rate": 1.3571566585800469e-05, + "loss": 0.9236, + "step": 30057 + }, + { + "epoch": 1.93, + "learning_rate": 1.3571351896777518e-05, + "loss": 1.0177, + "step": 30058 + }, + { + "epoch": 1.93, + "learning_rate": 1.3571137207754567e-05, + "loss": 0.935, + "step": 30059 + }, + { + "epoch": 1.93, + "learning_rate": 1.3570922518731618e-05, + "loss": 0.9382, + "step": 30060 + }, + { + "epoch": 1.93, + "learning_rate": 1.357070782970867e-05, + "loss": 0.9365, + "step": 30061 + }, + { + "epoch": 1.93, + "learning_rate": 1.3570493140685719e-05, + "loss": 0.9574, + "step": 30062 + }, + { + "epoch": 1.93, + "learning_rate": 1.3570278451662768e-05, + "loss": 0.9366, + "step": 30063 + }, + { + "epoch": 1.93, + "learning_rate": 1.3570063762639819e-05, + "loss": 0.9334, + "step": 30064 + }, + { + "epoch": 1.93, + "learning_rate": 1.3569849073616868e-05, + "loss": 0.9663, + "step": 30065 + }, + { + "epoch": 1.93, + "learning_rate": 1.3569634384593916e-05, + "loss": 1.0201, + "step": 30066 + }, + { + "epoch": 1.93, + "learning_rate": 1.3569419695570967e-05, + "loss": 0.9358, + "step": 30067 + }, + { + "epoch": 1.93, + "learning_rate": 1.3569205006548016e-05, + "loss": 0.9829, + "step": 30068 + }, + { + "epoch": 1.93, + "learning_rate": 1.3568990317525065e-05, + "loss": 0.9951, + "step": 30069 + }, + { + "epoch": 1.93, + "learning_rate": 1.3568775628502116e-05, + "loss": 0.9206, + "step": 30070 + }, + { + "epoch": 1.93, + "learning_rate": 1.3568560939479165e-05, + "loss": 0.9207, + "step": 30071 + }, + { + "epoch": 1.93, + "learning_rate": 1.3568346250456215e-05, + "loss": 0.8895, + "step": 30072 + }, + { + "epoch": 1.93, + "learning_rate": 1.3568131561433264e-05, + "loss": 0.9617, + "step": 30073 + }, + { + "epoch": 1.93, + "learning_rate": 1.3567916872410313e-05, + "loss": 1.0098, + "step": 30074 + }, + { + "epoch": 1.93, + "learning_rate": 1.3567702183387364e-05, + "loss": 0.931, + "step": 30075 + }, + { + "epoch": 1.94, + "learning_rate": 1.3567487494364415e-05, + "loss": 0.9536, + "step": 30076 + }, + { + "epoch": 1.94, + "learning_rate": 1.3567272805341465e-05, + "loss": 0.9855, + "step": 30077 + }, + { + "epoch": 1.94, + "learning_rate": 1.3567058116318514e-05, + "loss": 0.9262, + "step": 30078 + }, + { + "epoch": 1.94, + "learning_rate": 1.3566843427295563e-05, + "loss": 0.9392, + "step": 30079 + }, + { + "epoch": 1.94, + "learning_rate": 1.3566628738272614e-05, + "loss": 0.9921, + "step": 30080 + }, + { + "epoch": 1.94, + "learning_rate": 1.3566414049249663e-05, + "loss": 0.9405, + "step": 30081 + }, + { + "epoch": 1.94, + "learning_rate": 1.3566199360226713e-05, + "loss": 0.9583, + "step": 30082 + }, + { + "epoch": 1.94, + "learning_rate": 1.3565984671203762e-05, + "loss": 0.9071, + "step": 30083 + }, + { + "epoch": 1.94, + "learning_rate": 1.3565769982180811e-05, + "loss": 0.9973, + "step": 30084 + }, + { + "epoch": 1.94, + "learning_rate": 1.3565555293157862e-05, + "loss": 0.9995, + "step": 30085 + }, + { + "epoch": 1.94, + "learning_rate": 1.3565340604134911e-05, + "loss": 1.0065, + "step": 30086 + }, + { + "epoch": 1.94, + "learning_rate": 1.356512591511196e-05, + "loss": 0.9067, + "step": 30087 + }, + { + "epoch": 1.94, + "learning_rate": 1.356491122608901e-05, + "loss": 1.0076, + "step": 30088 + }, + { + "epoch": 1.94, + "learning_rate": 1.356469653706606e-05, + "loss": 1.0029, + "step": 30089 + }, + { + "epoch": 1.94, + "learning_rate": 1.3564481848043112e-05, + "loss": 0.9937, + "step": 30090 + }, + { + "epoch": 1.94, + "learning_rate": 1.3564267159020161e-05, + "loss": 1.0835, + "step": 30091 + }, + { + "epoch": 1.94, + "learning_rate": 1.3564052469997212e-05, + "loss": 0.9253, + "step": 30092 + }, + { + "epoch": 1.94, + "learning_rate": 1.356383778097426e-05, + "loss": 0.9866, + "step": 30093 + }, + { + "epoch": 1.94, + "learning_rate": 1.356362309195131e-05, + "loss": 1.0284, + "step": 30094 + }, + { + "epoch": 1.94, + "learning_rate": 1.356340840292836e-05, + "loss": 0.9193, + "step": 30095 + }, + { + "epoch": 1.94, + "learning_rate": 1.356319371390541e-05, + "loss": 0.9389, + "step": 30096 + }, + { + "epoch": 1.94, + "learning_rate": 1.3562979024882458e-05, + "loss": 0.9724, + "step": 30097 + }, + { + "epoch": 1.94, + "learning_rate": 1.3562764335859509e-05, + "loss": 0.9815, + "step": 30098 + }, + { + "epoch": 1.94, + "learning_rate": 1.3562549646836558e-05, + "loss": 0.9549, + "step": 30099 + }, + { + "epoch": 1.94, + "learning_rate": 1.3562334957813608e-05, + "loss": 0.9208, + "step": 30100 + }, + { + "epoch": 1.94, + "learning_rate": 1.3562120268790657e-05, + "loss": 1.0123, + "step": 30101 + }, + { + "epoch": 1.94, + "learning_rate": 1.3561905579767706e-05, + "loss": 1.031, + "step": 30102 + }, + { + "epoch": 1.94, + "learning_rate": 1.3561690890744757e-05, + "loss": 0.9597, + "step": 30103 + }, + { + "epoch": 1.94, + "learning_rate": 1.3561476201721806e-05, + "loss": 0.9839, + "step": 30104 + }, + { + "epoch": 1.94, + "learning_rate": 1.3561261512698858e-05, + "loss": 0.9672, + "step": 30105 + }, + { + "epoch": 1.94, + "learning_rate": 1.3561046823675907e-05, + "loss": 1.0096, + "step": 30106 + }, + { + "epoch": 1.94, + "learning_rate": 1.3560832134652956e-05, + "loss": 0.9602, + "step": 30107 + }, + { + "epoch": 1.94, + "learning_rate": 1.3560617445630007e-05, + "loss": 0.9467, + "step": 30108 + }, + { + "epoch": 1.94, + "learning_rate": 1.3560402756607056e-05, + "loss": 0.989, + "step": 30109 + }, + { + "epoch": 1.94, + "learning_rate": 1.3560188067584107e-05, + "loss": 0.9252, + "step": 30110 + }, + { + "epoch": 1.94, + "learning_rate": 1.3559973378561155e-05, + "loss": 0.9787, + "step": 30111 + }, + { + "epoch": 1.94, + "learning_rate": 1.3559758689538204e-05, + "loss": 0.9852, + "step": 30112 + }, + { + "epoch": 1.94, + "learning_rate": 1.3559544000515255e-05, + "loss": 1.0374, + "step": 30113 + }, + { + "epoch": 1.94, + "learning_rate": 1.3559329311492304e-05, + "loss": 0.9619, + "step": 30114 + }, + { + "epoch": 1.94, + "learning_rate": 1.3559114622469353e-05, + "loss": 0.9971, + "step": 30115 + }, + { + "epoch": 1.94, + "learning_rate": 1.3558899933446404e-05, + "loss": 0.907, + "step": 30116 + }, + { + "epoch": 1.94, + "learning_rate": 1.3558685244423453e-05, + "loss": 0.9794, + "step": 30117 + }, + { + "epoch": 1.94, + "learning_rate": 1.3558470555400503e-05, + "loss": 0.9658, + "step": 30118 + }, + { + "epoch": 1.94, + "learning_rate": 1.3558255866377554e-05, + "loss": 0.8983, + "step": 30119 + }, + { + "epoch": 1.94, + "learning_rate": 1.3558041177354605e-05, + "loss": 0.9054, + "step": 30120 + }, + { + "epoch": 1.94, + "learning_rate": 1.3557826488331654e-05, + "loss": 0.9039, + "step": 30121 + }, + { + "epoch": 1.94, + "learning_rate": 1.3557611799308703e-05, + "loss": 0.9378, + "step": 30122 + }, + { + "epoch": 1.94, + "learning_rate": 1.3557397110285753e-05, + "loss": 0.9725, + "step": 30123 + }, + { + "epoch": 1.94, + "learning_rate": 1.3557182421262802e-05, + "loss": 1.0157, + "step": 30124 + }, + { + "epoch": 1.94, + "learning_rate": 1.3556967732239851e-05, + "loss": 0.9302, + "step": 30125 + }, + { + "epoch": 1.94, + "learning_rate": 1.3556753043216902e-05, + "loss": 0.9649, + "step": 30126 + }, + { + "epoch": 1.94, + "learning_rate": 1.355653835419395e-05, + "loss": 1.0405, + "step": 30127 + }, + { + "epoch": 1.94, + "learning_rate": 1.3556323665171001e-05, + "loss": 1.0234, + "step": 30128 + }, + { + "epoch": 1.94, + "learning_rate": 1.355610897614805e-05, + "loss": 0.8777, + "step": 30129 + }, + { + "epoch": 1.94, + "learning_rate": 1.35558942871251e-05, + "loss": 0.9253, + "step": 30130 + }, + { + "epoch": 1.94, + "learning_rate": 1.355567959810215e-05, + "loss": 1.0025, + "step": 30131 + }, + { + "epoch": 1.94, + "learning_rate": 1.3555464909079199e-05, + "loss": 0.9496, + "step": 30132 + }, + { + "epoch": 1.94, + "learning_rate": 1.3555250220056248e-05, + "loss": 0.9162, + "step": 30133 + }, + { + "epoch": 1.94, + "learning_rate": 1.35550355310333e-05, + "loss": 0.9752, + "step": 30134 + }, + { + "epoch": 1.94, + "learning_rate": 1.355482084201035e-05, + "loss": 1.004, + "step": 30135 + }, + { + "epoch": 1.94, + "learning_rate": 1.35546061529874e-05, + "loss": 0.9905, + "step": 30136 + }, + { + "epoch": 1.94, + "learning_rate": 1.3554391463964449e-05, + "loss": 1.0247, + "step": 30137 + }, + { + "epoch": 1.94, + "learning_rate": 1.35541767749415e-05, + "loss": 0.9499, + "step": 30138 + }, + { + "epoch": 1.94, + "learning_rate": 1.3553962085918549e-05, + "loss": 0.9732, + "step": 30139 + }, + { + "epoch": 1.94, + "learning_rate": 1.3553747396895597e-05, + "loss": 0.9326, + "step": 30140 + }, + { + "epoch": 1.94, + "learning_rate": 1.3553532707872648e-05, + "loss": 0.9438, + "step": 30141 + }, + { + "epoch": 1.94, + "learning_rate": 1.3553318018849697e-05, + "loss": 0.9202, + "step": 30142 + }, + { + "epoch": 1.94, + "learning_rate": 1.3553103329826746e-05, + "loss": 0.9944, + "step": 30143 + }, + { + "epoch": 1.94, + "learning_rate": 1.3552888640803797e-05, + "loss": 0.988, + "step": 30144 + }, + { + "epoch": 1.94, + "learning_rate": 1.3552673951780846e-05, + "loss": 1.0126, + "step": 30145 + }, + { + "epoch": 1.94, + "learning_rate": 1.3552459262757895e-05, + "loss": 0.9159, + "step": 30146 + }, + { + "epoch": 1.94, + "learning_rate": 1.3552244573734945e-05, + "loss": 1.0474, + "step": 30147 + }, + { + "epoch": 1.94, + "learning_rate": 1.3552029884711998e-05, + "loss": 1.0802, + "step": 30148 + }, + { + "epoch": 1.94, + "learning_rate": 1.3551815195689047e-05, + "loss": 0.9633, + "step": 30149 + }, + { + "epoch": 1.94, + "learning_rate": 1.3551600506666096e-05, + "loss": 0.9894, + "step": 30150 + }, + { + "epoch": 1.94, + "learning_rate": 1.3551385817643146e-05, + "loss": 0.9266, + "step": 30151 + }, + { + "epoch": 1.94, + "learning_rate": 1.3551171128620195e-05, + "loss": 0.9279, + "step": 30152 + }, + { + "epoch": 1.94, + "learning_rate": 1.3550956439597244e-05, + "loss": 0.9489, + "step": 30153 + }, + { + "epoch": 1.94, + "learning_rate": 1.3550741750574295e-05, + "loss": 0.9623, + "step": 30154 + }, + { + "epoch": 1.94, + "learning_rate": 1.3550527061551344e-05, + "loss": 1.0099, + "step": 30155 + }, + { + "epoch": 1.94, + "learning_rate": 1.3550312372528393e-05, + "loss": 1.004, + "step": 30156 + }, + { + "epoch": 1.94, + "learning_rate": 1.3550097683505443e-05, + "loss": 0.9494, + "step": 30157 + }, + { + "epoch": 1.94, + "learning_rate": 1.3549882994482492e-05, + "loss": 0.9736, + "step": 30158 + }, + { + "epoch": 1.94, + "learning_rate": 1.3549668305459543e-05, + "loss": 0.8857, + "step": 30159 + }, + { + "epoch": 1.94, + "learning_rate": 1.3549453616436592e-05, + "loss": 0.9254, + "step": 30160 + }, + { + "epoch": 1.94, + "learning_rate": 1.3549238927413641e-05, + "loss": 0.9693, + "step": 30161 + }, + { + "epoch": 1.94, + "learning_rate": 1.3549024238390692e-05, + "loss": 0.9752, + "step": 30162 + }, + { + "epoch": 1.94, + "learning_rate": 1.3548809549367742e-05, + "loss": 0.9972, + "step": 30163 + }, + { + "epoch": 1.94, + "learning_rate": 1.3548594860344793e-05, + "loss": 0.9827, + "step": 30164 + }, + { + "epoch": 1.94, + "learning_rate": 1.3548380171321842e-05, + "loss": 0.9256, + "step": 30165 + }, + { + "epoch": 1.94, + "learning_rate": 1.3548165482298891e-05, + "loss": 0.9852, + "step": 30166 + }, + { + "epoch": 1.94, + "learning_rate": 1.3547950793275942e-05, + "loss": 0.9369, + "step": 30167 + }, + { + "epoch": 1.94, + "learning_rate": 1.354773610425299e-05, + "loss": 0.9688, + "step": 30168 + }, + { + "epoch": 1.94, + "learning_rate": 1.3547521415230041e-05, + "loss": 1.0199, + "step": 30169 + }, + { + "epoch": 1.94, + "learning_rate": 1.354730672620709e-05, + "loss": 0.9274, + "step": 30170 + }, + { + "epoch": 1.94, + "learning_rate": 1.3547092037184139e-05, + "loss": 1.0141, + "step": 30171 + }, + { + "epoch": 1.94, + "learning_rate": 1.354687734816119e-05, + "loss": 1.0292, + "step": 30172 + }, + { + "epoch": 1.94, + "learning_rate": 1.3546662659138239e-05, + "loss": 1.0117, + "step": 30173 + }, + { + "epoch": 1.94, + "learning_rate": 1.3546447970115288e-05, + "loss": 0.9242, + "step": 30174 + }, + { + "epoch": 1.94, + "learning_rate": 1.3546233281092338e-05, + "loss": 0.9172, + "step": 30175 + }, + { + "epoch": 1.94, + "learning_rate": 1.3546018592069387e-05, + "loss": 0.9662, + "step": 30176 + }, + { + "epoch": 1.94, + "learning_rate": 1.354580390304644e-05, + "loss": 0.9736, + "step": 30177 + }, + { + "epoch": 1.94, + "learning_rate": 1.3545589214023489e-05, + "loss": 0.9867, + "step": 30178 + }, + { + "epoch": 1.94, + "learning_rate": 1.354537452500054e-05, + "loss": 0.9274, + "step": 30179 + }, + { + "epoch": 1.94, + "learning_rate": 1.3545159835977588e-05, + "loss": 0.9791, + "step": 30180 + }, + { + "epoch": 1.94, + "learning_rate": 1.3544945146954637e-05, + "loss": 0.9471, + "step": 30181 + }, + { + "epoch": 1.94, + "learning_rate": 1.3544730457931688e-05, + "loss": 0.9387, + "step": 30182 + }, + { + "epoch": 1.94, + "learning_rate": 1.3544515768908737e-05, + "loss": 0.9955, + "step": 30183 + }, + { + "epoch": 1.94, + "learning_rate": 1.3544301079885786e-05, + "loss": 0.9377, + "step": 30184 + }, + { + "epoch": 1.94, + "learning_rate": 1.3544086390862836e-05, + "loss": 0.9281, + "step": 30185 + }, + { + "epoch": 1.94, + "learning_rate": 1.3543871701839885e-05, + "loss": 0.9633, + "step": 30186 + }, + { + "epoch": 1.94, + "learning_rate": 1.3543657012816936e-05, + "loss": 0.9303, + "step": 30187 + }, + { + "epoch": 1.94, + "learning_rate": 1.3543442323793985e-05, + "loss": 0.9809, + "step": 30188 + }, + { + "epoch": 1.94, + "learning_rate": 1.3543227634771034e-05, + "loss": 0.9456, + "step": 30189 + }, + { + "epoch": 1.94, + "learning_rate": 1.3543012945748085e-05, + "loss": 0.9555, + "step": 30190 + }, + { + "epoch": 1.94, + "learning_rate": 1.3542798256725134e-05, + "loss": 0.9727, + "step": 30191 + }, + { + "epoch": 1.94, + "learning_rate": 1.3542583567702186e-05, + "loss": 1.0925, + "step": 30192 + }, + { + "epoch": 1.94, + "learning_rate": 1.3542368878679235e-05, + "loss": 1.0412, + "step": 30193 + }, + { + "epoch": 1.94, + "learning_rate": 1.3542154189656284e-05, + "loss": 0.9299, + "step": 30194 + }, + { + "epoch": 1.94, + "learning_rate": 1.3541939500633335e-05, + "loss": 0.9267, + "step": 30195 + }, + { + "epoch": 1.94, + "learning_rate": 1.3541724811610384e-05, + "loss": 0.9639, + "step": 30196 + }, + { + "epoch": 1.94, + "learning_rate": 1.3541510122587434e-05, + "loss": 0.9874, + "step": 30197 + }, + { + "epoch": 1.94, + "learning_rate": 1.3541295433564483e-05, + "loss": 0.9541, + "step": 30198 + }, + { + "epoch": 1.94, + "learning_rate": 1.3541080744541532e-05, + "loss": 0.9797, + "step": 30199 + }, + { + "epoch": 1.94, + "learning_rate": 1.3540866055518583e-05, + "loss": 0.942, + "step": 30200 + }, + { + "epoch": 1.94, + "learning_rate": 1.3540651366495632e-05, + "loss": 0.9489, + "step": 30201 + }, + { + "epoch": 1.94, + "learning_rate": 1.354043667747268e-05, + "loss": 1.0325, + "step": 30202 + }, + { + "epoch": 1.94, + "learning_rate": 1.3540221988449731e-05, + "loss": 0.9048, + "step": 30203 + }, + { + "epoch": 1.94, + "learning_rate": 1.354000729942678e-05, + "loss": 0.9895, + "step": 30204 + }, + { + "epoch": 1.94, + "learning_rate": 1.3539792610403831e-05, + "loss": 0.9515, + "step": 30205 + }, + { + "epoch": 1.94, + "learning_rate": 1.3539577921380882e-05, + "loss": 1.0531, + "step": 30206 + }, + { + "epoch": 1.94, + "learning_rate": 1.3539363232357932e-05, + "loss": 0.9053, + "step": 30207 + }, + { + "epoch": 1.94, + "learning_rate": 1.3539148543334981e-05, + "loss": 1.0083, + "step": 30208 + }, + { + "epoch": 1.94, + "learning_rate": 1.353893385431203e-05, + "loss": 0.9597, + "step": 30209 + }, + { + "epoch": 1.94, + "learning_rate": 1.3538719165289081e-05, + "loss": 0.9064, + "step": 30210 + }, + { + "epoch": 1.94, + "learning_rate": 1.353850447626613e-05, + "loss": 0.9777, + "step": 30211 + }, + { + "epoch": 1.94, + "learning_rate": 1.3538289787243179e-05, + "loss": 0.9707, + "step": 30212 + }, + { + "epoch": 1.94, + "learning_rate": 1.353807509822023e-05, + "loss": 0.903, + "step": 30213 + }, + { + "epoch": 1.94, + "learning_rate": 1.3537860409197278e-05, + "loss": 0.8884, + "step": 30214 + }, + { + "epoch": 1.94, + "learning_rate": 1.353764572017433e-05, + "loss": 0.9549, + "step": 30215 + }, + { + "epoch": 1.94, + "learning_rate": 1.3537431031151378e-05, + "loss": 0.8947, + "step": 30216 + }, + { + "epoch": 1.94, + "learning_rate": 1.3537216342128427e-05, + "loss": 0.9061, + "step": 30217 + }, + { + "epoch": 1.94, + "learning_rate": 1.3537001653105478e-05, + "loss": 0.9321, + "step": 30218 + }, + { + "epoch": 1.94, + "learning_rate": 1.3536786964082527e-05, + "loss": 0.9604, + "step": 30219 + }, + { + "epoch": 1.94, + "learning_rate": 1.3536572275059576e-05, + "loss": 1.0348, + "step": 30220 + }, + { + "epoch": 1.94, + "learning_rate": 1.3536357586036628e-05, + "loss": 0.9502, + "step": 30221 + }, + { + "epoch": 1.94, + "learning_rate": 1.3536142897013677e-05, + "loss": 0.9753, + "step": 30222 + }, + { + "epoch": 1.94, + "learning_rate": 1.3535928207990728e-05, + "loss": 0.9147, + "step": 30223 + }, + { + "epoch": 1.94, + "learning_rate": 1.3535713518967777e-05, + "loss": 0.944, + "step": 30224 + }, + { + "epoch": 1.94, + "learning_rate": 1.3535498829944827e-05, + "loss": 0.9915, + "step": 30225 + }, + { + "epoch": 1.94, + "learning_rate": 1.3535284140921876e-05, + "loss": 0.9954, + "step": 30226 + }, + { + "epoch": 1.94, + "learning_rate": 1.3535069451898925e-05, + "loss": 0.9609, + "step": 30227 + }, + { + "epoch": 1.94, + "learning_rate": 1.3534854762875976e-05, + "loss": 0.9785, + "step": 30228 + }, + { + "epoch": 1.94, + "learning_rate": 1.3534640073853025e-05, + "loss": 0.9356, + "step": 30229 + }, + { + "epoch": 1.94, + "learning_rate": 1.3534425384830074e-05, + "loss": 0.9593, + "step": 30230 + }, + { + "epoch": 1.94, + "learning_rate": 1.3534210695807124e-05, + "loss": 0.9556, + "step": 30231 + }, + { + "epoch": 1.95, + "learning_rate": 1.3533996006784173e-05, + "loss": 0.9218, + "step": 30232 + }, + { + "epoch": 1.95, + "learning_rate": 1.3533781317761222e-05, + "loss": 1.0154, + "step": 30233 + }, + { + "epoch": 1.95, + "learning_rate": 1.3533566628738273e-05, + "loss": 1.0159, + "step": 30234 + }, + { + "epoch": 1.95, + "learning_rate": 1.3533351939715325e-05, + "loss": 0.9214, + "step": 30235 + }, + { + "epoch": 1.95, + "learning_rate": 1.3533137250692374e-05, + "loss": 1.0367, + "step": 30236 + }, + { + "epoch": 1.95, + "learning_rate": 1.3532922561669423e-05, + "loss": 0.9388, + "step": 30237 + }, + { + "epoch": 1.95, + "learning_rate": 1.3532707872646474e-05, + "loss": 1.0163, + "step": 30238 + }, + { + "epoch": 1.95, + "learning_rate": 1.3532493183623523e-05, + "loss": 0.9863, + "step": 30239 + }, + { + "epoch": 1.95, + "learning_rate": 1.3532278494600572e-05, + "loss": 0.9352, + "step": 30240 + }, + { + "epoch": 1.95, + "learning_rate": 1.3532063805577623e-05, + "loss": 0.987, + "step": 30241 + }, + { + "epoch": 1.95, + "learning_rate": 1.3531849116554672e-05, + "loss": 0.9212, + "step": 30242 + }, + { + "epoch": 1.95, + "learning_rate": 1.353163442753172e-05, + "loss": 0.9399, + "step": 30243 + }, + { + "epoch": 1.95, + "learning_rate": 1.3531419738508771e-05, + "loss": 0.9417, + "step": 30244 + }, + { + "epoch": 1.95, + "learning_rate": 1.353120504948582e-05, + "loss": 0.9342, + "step": 30245 + }, + { + "epoch": 1.95, + "learning_rate": 1.353099036046287e-05, + "loss": 0.9313, + "step": 30246 + }, + { + "epoch": 1.95, + "learning_rate": 1.353077567143992e-05, + "loss": 0.9134, + "step": 30247 + }, + { + "epoch": 1.95, + "learning_rate": 1.3530560982416969e-05, + "loss": 0.9433, + "step": 30248 + }, + { + "epoch": 1.95, + "learning_rate": 1.353034629339402e-05, + "loss": 1.0051, + "step": 30249 + }, + { + "epoch": 1.95, + "learning_rate": 1.353013160437107e-05, + "loss": 0.964, + "step": 30250 + }, + { + "epoch": 1.95, + "learning_rate": 1.352991691534812e-05, + "loss": 0.9623, + "step": 30251 + }, + { + "epoch": 1.95, + "learning_rate": 1.352970222632517e-05, + "loss": 0.9917, + "step": 30252 + }, + { + "epoch": 1.95, + "learning_rate": 1.3529487537302219e-05, + "loss": 0.9636, + "step": 30253 + }, + { + "epoch": 1.95, + "learning_rate": 1.352927284827927e-05, + "loss": 0.9839, + "step": 30254 + }, + { + "epoch": 1.95, + "learning_rate": 1.3529058159256318e-05, + "loss": 0.9922, + "step": 30255 + }, + { + "epoch": 1.95, + "learning_rate": 1.3528843470233369e-05, + "loss": 0.9951, + "step": 30256 + }, + { + "epoch": 1.95, + "learning_rate": 1.3528628781210418e-05, + "loss": 1.0306, + "step": 30257 + }, + { + "epoch": 1.95, + "learning_rate": 1.3528414092187467e-05, + "loss": 0.9938, + "step": 30258 + }, + { + "epoch": 1.95, + "learning_rate": 1.3528199403164517e-05, + "loss": 1.0309, + "step": 30259 + }, + { + "epoch": 1.95, + "learning_rate": 1.3527984714141566e-05, + "loss": 1.0065, + "step": 30260 + }, + { + "epoch": 1.95, + "learning_rate": 1.3527770025118615e-05, + "loss": 0.9435, + "step": 30261 + }, + { + "epoch": 1.95, + "learning_rate": 1.3527555336095666e-05, + "loss": 0.8899, + "step": 30262 + }, + { + "epoch": 1.95, + "learning_rate": 1.3527340647072715e-05, + "loss": 0.9443, + "step": 30263 + }, + { + "epoch": 1.95, + "learning_rate": 1.3527125958049766e-05, + "loss": 0.9762, + "step": 30264 + }, + { + "epoch": 1.95, + "learning_rate": 1.3526911269026816e-05, + "loss": 0.9667, + "step": 30265 + }, + { + "epoch": 1.95, + "learning_rate": 1.3526696580003867e-05, + "loss": 0.9777, + "step": 30266 + }, + { + "epoch": 1.95, + "learning_rate": 1.3526481890980916e-05, + "loss": 0.9186, + "step": 30267 + }, + { + "epoch": 1.95, + "learning_rate": 1.3526267201957965e-05, + "loss": 0.9304, + "step": 30268 + }, + { + "epoch": 1.95, + "learning_rate": 1.3526052512935016e-05, + "loss": 0.9878, + "step": 30269 + }, + { + "epoch": 1.95, + "learning_rate": 1.3525837823912065e-05, + "loss": 1.0012, + "step": 30270 + }, + { + "epoch": 1.95, + "learning_rate": 1.3525623134889114e-05, + "loss": 0.9007, + "step": 30271 + }, + { + "epoch": 1.95, + "learning_rate": 1.3525408445866164e-05, + "loss": 0.9972, + "step": 30272 + }, + { + "epoch": 1.95, + "learning_rate": 1.3525193756843213e-05, + "loss": 0.9803, + "step": 30273 + }, + { + "epoch": 1.95, + "learning_rate": 1.3524979067820264e-05, + "loss": 0.9959, + "step": 30274 + }, + { + "epoch": 1.95, + "learning_rate": 1.3524764378797313e-05, + "loss": 0.8957, + "step": 30275 + }, + { + "epoch": 1.95, + "learning_rate": 1.3524549689774362e-05, + "loss": 0.9739, + "step": 30276 + }, + { + "epoch": 1.95, + "learning_rate": 1.3524335000751412e-05, + "loss": 0.9204, + "step": 30277 + }, + { + "epoch": 1.95, + "learning_rate": 1.3524120311728461e-05, + "loss": 0.9226, + "step": 30278 + }, + { + "epoch": 1.95, + "learning_rate": 1.3523905622705514e-05, + "loss": 0.9787, + "step": 30279 + }, + { + "epoch": 1.95, + "learning_rate": 1.3523690933682563e-05, + "loss": 1.0158, + "step": 30280 + }, + { + "epoch": 1.95, + "learning_rate": 1.3523476244659612e-05, + "loss": 0.9674, + "step": 30281 + }, + { + "epoch": 1.95, + "learning_rate": 1.3523261555636662e-05, + "loss": 0.9834, + "step": 30282 + }, + { + "epoch": 1.95, + "learning_rate": 1.3523046866613711e-05, + "loss": 1.0558, + "step": 30283 + }, + { + "epoch": 1.95, + "learning_rate": 1.3522832177590762e-05, + "loss": 0.9521, + "step": 30284 + }, + { + "epoch": 1.95, + "learning_rate": 1.3522617488567811e-05, + "loss": 1.0411, + "step": 30285 + }, + { + "epoch": 1.95, + "learning_rate": 1.352240279954486e-05, + "loss": 0.9406, + "step": 30286 + }, + { + "epoch": 1.95, + "learning_rate": 1.352218811052191e-05, + "loss": 0.8789, + "step": 30287 + }, + { + "epoch": 1.95, + "learning_rate": 1.352197342149896e-05, + "loss": 0.9406, + "step": 30288 + }, + { + "epoch": 1.95, + "learning_rate": 1.3521758732476008e-05, + "loss": 0.9971, + "step": 30289 + }, + { + "epoch": 1.95, + "learning_rate": 1.3521544043453059e-05, + "loss": 0.9868, + "step": 30290 + }, + { + "epoch": 1.95, + "learning_rate": 1.3521329354430108e-05, + "loss": 1.0022, + "step": 30291 + }, + { + "epoch": 1.95, + "learning_rate": 1.3521114665407157e-05, + "loss": 0.9739, + "step": 30292 + }, + { + "epoch": 1.95, + "learning_rate": 1.3520899976384208e-05, + "loss": 0.887, + "step": 30293 + }, + { + "epoch": 1.95, + "learning_rate": 1.352068528736126e-05, + "loss": 0.9791, + "step": 30294 + }, + { + "epoch": 1.95, + "learning_rate": 1.3520470598338309e-05, + "loss": 0.9498, + "step": 30295 + }, + { + "epoch": 1.95, + "learning_rate": 1.3520255909315358e-05, + "loss": 0.9862, + "step": 30296 + }, + { + "epoch": 1.95, + "learning_rate": 1.3520041220292409e-05, + "loss": 0.9672, + "step": 30297 + }, + { + "epoch": 1.95, + "learning_rate": 1.3519826531269458e-05, + "loss": 0.9889, + "step": 30298 + }, + { + "epoch": 1.95, + "learning_rate": 1.3519611842246507e-05, + "loss": 0.9785, + "step": 30299 + }, + { + "epoch": 1.95, + "learning_rate": 1.3519397153223557e-05, + "loss": 0.9376, + "step": 30300 + }, + { + "epoch": 1.95, + "learning_rate": 1.3519182464200606e-05, + "loss": 0.9412, + "step": 30301 + }, + { + "epoch": 1.95, + "learning_rate": 1.3518967775177655e-05, + "loss": 1.009, + "step": 30302 + }, + { + "epoch": 1.95, + "learning_rate": 1.3518753086154706e-05, + "loss": 0.927, + "step": 30303 + }, + { + "epoch": 1.95, + "learning_rate": 1.3518538397131755e-05, + "loss": 0.9189, + "step": 30304 + }, + { + "epoch": 1.95, + "learning_rate": 1.3518323708108805e-05, + "loss": 1.034, + "step": 30305 + }, + { + "epoch": 1.95, + "learning_rate": 1.3518109019085854e-05, + "loss": 0.9549, + "step": 30306 + }, + { + "epoch": 1.95, + "learning_rate": 1.3517894330062903e-05, + "loss": 0.9273, + "step": 30307 + }, + { + "epoch": 1.95, + "learning_rate": 1.3517679641039956e-05, + "loss": 0.9645, + "step": 30308 + }, + { + "epoch": 1.95, + "learning_rate": 1.3517464952017005e-05, + "loss": 1.0249, + "step": 30309 + }, + { + "epoch": 1.95, + "learning_rate": 1.3517250262994055e-05, + "loss": 0.9936, + "step": 30310 + }, + { + "epoch": 1.95, + "learning_rate": 1.3517035573971104e-05, + "loss": 0.9096, + "step": 30311 + }, + { + "epoch": 1.95, + "learning_rate": 1.3516820884948155e-05, + "loss": 0.9317, + "step": 30312 + }, + { + "epoch": 1.95, + "learning_rate": 1.3516606195925204e-05, + "loss": 0.9036, + "step": 30313 + }, + { + "epoch": 1.95, + "learning_rate": 1.3516391506902253e-05, + "loss": 1.0449, + "step": 30314 + }, + { + "epoch": 1.95, + "learning_rate": 1.3516176817879304e-05, + "loss": 1.0205, + "step": 30315 + }, + { + "epoch": 1.95, + "learning_rate": 1.3515962128856353e-05, + "loss": 0.9501, + "step": 30316 + }, + { + "epoch": 1.95, + "learning_rate": 1.3515747439833402e-05, + "loss": 1.0185, + "step": 30317 + }, + { + "epoch": 1.95, + "learning_rate": 1.3515532750810452e-05, + "loss": 0.9554, + "step": 30318 + }, + { + "epoch": 1.95, + "learning_rate": 1.3515318061787501e-05, + "loss": 1.0221, + "step": 30319 + }, + { + "epoch": 1.95, + "learning_rate": 1.351510337276455e-05, + "loss": 0.9456, + "step": 30320 + }, + { + "epoch": 1.95, + "learning_rate": 1.35148886837416e-05, + "loss": 0.9558, + "step": 30321 + }, + { + "epoch": 1.95, + "learning_rate": 1.351467399471865e-05, + "loss": 0.9525, + "step": 30322 + }, + { + "epoch": 1.95, + "learning_rate": 1.3514459305695702e-05, + "loss": 0.911, + "step": 30323 + }, + { + "epoch": 1.95, + "learning_rate": 1.3514244616672751e-05, + "loss": 0.9708, + "step": 30324 + }, + { + "epoch": 1.95, + "learning_rate": 1.3514029927649802e-05, + "loss": 0.9455, + "step": 30325 + }, + { + "epoch": 1.95, + "learning_rate": 1.351381523862685e-05, + "loss": 0.9992, + "step": 30326 + }, + { + "epoch": 1.95, + "learning_rate": 1.35136005496039e-05, + "loss": 1.0196, + "step": 30327 + }, + { + "epoch": 1.95, + "learning_rate": 1.351338586058095e-05, + "loss": 0.9894, + "step": 30328 + }, + { + "epoch": 1.95, + "learning_rate": 1.3513171171558e-05, + "loss": 1.0129, + "step": 30329 + }, + { + "epoch": 1.95, + "learning_rate": 1.3512956482535048e-05, + "loss": 1.0414, + "step": 30330 + }, + { + "epoch": 1.95, + "learning_rate": 1.3512741793512099e-05, + "loss": 0.9531, + "step": 30331 + }, + { + "epoch": 1.95, + "learning_rate": 1.3512527104489148e-05, + "loss": 0.9695, + "step": 30332 + }, + { + "epoch": 1.95, + "learning_rate": 1.3512312415466198e-05, + "loss": 0.9627, + "step": 30333 + }, + { + "epoch": 1.95, + "learning_rate": 1.3512097726443247e-05, + "loss": 0.958, + "step": 30334 + }, + { + "epoch": 1.95, + "learning_rate": 1.3511883037420296e-05, + "loss": 0.8849, + "step": 30335 + }, + { + "epoch": 1.95, + "learning_rate": 1.3511668348397347e-05, + "loss": 0.9964, + "step": 30336 + }, + { + "epoch": 1.95, + "learning_rate": 1.3511453659374398e-05, + "loss": 1.0143, + "step": 30337 + }, + { + "epoch": 1.95, + "learning_rate": 1.3511238970351448e-05, + "loss": 0.9654, + "step": 30338 + }, + { + "epoch": 1.95, + "learning_rate": 1.3511024281328497e-05, + "loss": 0.96, + "step": 30339 + }, + { + "epoch": 1.95, + "learning_rate": 1.3510809592305546e-05, + "loss": 0.9476, + "step": 30340 + }, + { + "epoch": 1.95, + "learning_rate": 1.3510594903282597e-05, + "loss": 0.9264, + "step": 30341 + }, + { + "epoch": 1.95, + "learning_rate": 1.3510380214259646e-05, + "loss": 0.9286, + "step": 30342 + }, + { + "epoch": 1.95, + "learning_rate": 1.3510165525236697e-05, + "loss": 0.9259, + "step": 30343 + }, + { + "epoch": 1.95, + "learning_rate": 1.3509950836213746e-05, + "loss": 0.9549, + "step": 30344 + }, + { + "epoch": 1.95, + "learning_rate": 1.3509736147190795e-05, + "loss": 0.95, + "step": 30345 + }, + { + "epoch": 1.95, + "learning_rate": 1.3509521458167845e-05, + "loss": 0.958, + "step": 30346 + }, + { + "epoch": 1.95, + "learning_rate": 1.3509306769144894e-05, + "loss": 0.9742, + "step": 30347 + }, + { + "epoch": 1.95, + "learning_rate": 1.3509092080121943e-05, + "loss": 1.0016, + "step": 30348 + }, + { + "epoch": 1.95, + "learning_rate": 1.3508877391098994e-05, + "loss": 0.9317, + "step": 30349 + }, + { + "epoch": 1.95, + "learning_rate": 1.3508662702076043e-05, + "loss": 0.9626, + "step": 30350 + }, + { + "epoch": 1.95, + "learning_rate": 1.3508448013053093e-05, + "loss": 1.0373, + "step": 30351 + }, + { + "epoch": 1.95, + "learning_rate": 1.3508233324030144e-05, + "loss": 1.0112, + "step": 30352 + }, + { + "epoch": 1.95, + "learning_rate": 1.3508018635007195e-05, + "loss": 1.0458, + "step": 30353 + }, + { + "epoch": 1.95, + "learning_rate": 1.3507803945984244e-05, + "loss": 0.9472, + "step": 30354 + }, + { + "epoch": 1.95, + "learning_rate": 1.3507589256961293e-05, + "loss": 0.9297, + "step": 30355 + }, + { + "epoch": 1.95, + "learning_rate": 1.3507374567938343e-05, + "loss": 0.9197, + "step": 30356 + }, + { + "epoch": 1.95, + "learning_rate": 1.3507159878915392e-05, + "loss": 1.0591, + "step": 30357 + }, + { + "epoch": 1.95, + "learning_rate": 1.3506945189892441e-05, + "loss": 1.0042, + "step": 30358 + }, + { + "epoch": 1.95, + "learning_rate": 1.3506730500869492e-05, + "loss": 0.9461, + "step": 30359 + }, + { + "epoch": 1.95, + "learning_rate": 1.3506515811846541e-05, + "loss": 0.9785, + "step": 30360 + }, + { + "epoch": 1.95, + "learning_rate": 1.3506301122823592e-05, + "loss": 1.0938, + "step": 30361 + }, + { + "epoch": 1.95, + "learning_rate": 1.350608643380064e-05, + "loss": 0.9551, + "step": 30362 + }, + { + "epoch": 1.95, + "learning_rate": 1.350587174477769e-05, + "loss": 1.0591, + "step": 30363 + }, + { + "epoch": 1.95, + "learning_rate": 1.350565705575474e-05, + "loss": 0.9643, + "step": 30364 + }, + { + "epoch": 1.95, + "learning_rate": 1.3505442366731789e-05, + "loss": 0.932, + "step": 30365 + }, + { + "epoch": 1.95, + "learning_rate": 1.3505227677708841e-05, + "loss": 0.8685, + "step": 30366 + }, + { + "epoch": 1.95, + "learning_rate": 1.350501298868589e-05, + "loss": 0.9179, + "step": 30367 + }, + { + "epoch": 1.95, + "learning_rate": 1.350479829966294e-05, + "loss": 0.9735, + "step": 30368 + }, + { + "epoch": 1.95, + "learning_rate": 1.350458361063999e-05, + "loss": 1.0266, + "step": 30369 + }, + { + "epoch": 1.95, + "learning_rate": 1.3504368921617039e-05, + "loss": 0.9924, + "step": 30370 + }, + { + "epoch": 1.95, + "learning_rate": 1.350415423259409e-05, + "loss": 0.9487, + "step": 30371 + }, + { + "epoch": 1.95, + "learning_rate": 1.3503939543571139e-05, + "loss": 0.8553, + "step": 30372 + }, + { + "epoch": 1.95, + "learning_rate": 1.3503724854548188e-05, + "loss": 1.0262, + "step": 30373 + }, + { + "epoch": 1.95, + "learning_rate": 1.3503510165525238e-05, + "loss": 0.9741, + "step": 30374 + }, + { + "epoch": 1.95, + "learning_rate": 1.3503295476502287e-05, + "loss": 0.925, + "step": 30375 + }, + { + "epoch": 1.95, + "learning_rate": 1.3503080787479336e-05, + "loss": 0.9225, + "step": 30376 + }, + { + "epoch": 1.95, + "learning_rate": 1.3502866098456387e-05, + "loss": 0.9413, + "step": 30377 + }, + { + "epoch": 1.95, + "learning_rate": 1.3502651409433436e-05, + "loss": 0.9831, + "step": 30378 + }, + { + "epoch": 1.95, + "learning_rate": 1.3502436720410485e-05, + "loss": 0.9944, + "step": 30379 + }, + { + "epoch": 1.95, + "learning_rate": 1.3502222031387535e-05, + "loss": 0.9948, + "step": 30380 + }, + { + "epoch": 1.95, + "learning_rate": 1.3502007342364588e-05, + "loss": 0.9887, + "step": 30381 + }, + { + "epoch": 1.95, + "learning_rate": 1.3501792653341637e-05, + "loss": 0.8941, + "step": 30382 + }, + { + "epoch": 1.95, + "learning_rate": 1.3501577964318686e-05, + "loss": 1.0552, + "step": 30383 + }, + { + "epoch": 1.95, + "learning_rate": 1.3501363275295736e-05, + "loss": 0.9675, + "step": 30384 + }, + { + "epoch": 1.95, + "learning_rate": 1.3501148586272785e-05, + "loss": 0.9064, + "step": 30385 + }, + { + "epoch": 1.95, + "learning_rate": 1.3500933897249834e-05, + "loss": 0.9762, + "step": 30386 + }, + { + "epoch": 1.96, + "learning_rate": 1.3500719208226885e-05, + "loss": 0.9562, + "step": 30387 + }, + { + "epoch": 1.96, + "learning_rate": 1.3500504519203934e-05, + "loss": 0.8896, + "step": 30388 + }, + { + "epoch": 1.96, + "learning_rate": 1.3500289830180983e-05, + "loss": 0.9508, + "step": 30389 + }, + { + "epoch": 1.96, + "learning_rate": 1.3500075141158034e-05, + "loss": 1.1034, + "step": 30390 + }, + { + "epoch": 1.96, + "learning_rate": 1.3499860452135083e-05, + "loss": 1.0019, + "step": 30391 + }, + { + "epoch": 1.96, + "learning_rate": 1.3499645763112133e-05, + "loss": 1.055, + "step": 30392 + }, + { + "epoch": 1.96, + "learning_rate": 1.3499431074089182e-05, + "loss": 0.9259, + "step": 30393 + }, + { + "epoch": 1.96, + "learning_rate": 1.3499216385066231e-05, + "loss": 0.9727, + "step": 30394 + }, + { + "epoch": 1.96, + "learning_rate": 1.3499001696043283e-05, + "loss": 1.0138, + "step": 30395 + }, + { + "epoch": 1.96, + "learning_rate": 1.3498787007020332e-05, + "loss": 1.0255, + "step": 30396 + }, + { + "epoch": 1.96, + "learning_rate": 1.3498572317997383e-05, + "loss": 0.9485, + "step": 30397 + }, + { + "epoch": 1.96, + "learning_rate": 1.3498357628974432e-05, + "loss": 0.9689, + "step": 30398 + }, + { + "epoch": 1.96, + "learning_rate": 1.3498142939951481e-05, + "loss": 0.9847, + "step": 30399 + }, + { + "epoch": 1.96, + "learning_rate": 1.3497928250928532e-05, + "loss": 0.9204, + "step": 30400 + }, + { + "epoch": 1.96, + "learning_rate": 1.349771356190558e-05, + "loss": 0.941, + "step": 30401 + }, + { + "epoch": 1.96, + "learning_rate": 1.3497498872882631e-05, + "loss": 0.8988, + "step": 30402 + }, + { + "epoch": 1.96, + "learning_rate": 1.349728418385968e-05, + "loss": 0.9554, + "step": 30403 + }, + { + "epoch": 1.96, + "learning_rate": 1.349706949483673e-05, + "loss": 0.9885, + "step": 30404 + }, + { + "epoch": 1.96, + "learning_rate": 1.349685480581378e-05, + "loss": 0.9933, + "step": 30405 + }, + { + "epoch": 1.96, + "learning_rate": 1.3496640116790829e-05, + "loss": 0.9129, + "step": 30406 + }, + { + "epoch": 1.96, + "learning_rate": 1.3496425427767878e-05, + "loss": 0.958, + "step": 30407 + }, + { + "epoch": 1.96, + "learning_rate": 1.3496210738744928e-05, + "loss": 0.8955, + "step": 30408 + }, + { + "epoch": 1.96, + "learning_rate": 1.3495996049721977e-05, + "loss": 0.9312, + "step": 30409 + }, + { + "epoch": 1.96, + "learning_rate": 1.349578136069903e-05, + "loss": 1.0106, + "step": 30410 + }, + { + "epoch": 1.96, + "learning_rate": 1.3495566671676079e-05, + "loss": 1.0529, + "step": 30411 + }, + { + "epoch": 1.96, + "learning_rate": 1.349535198265313e-05, + "loss": 0.9311, + "step": 30412 + }, + { + "epoch": 1.96, + "learning_rate": 1.3495137293630178e-05, + "loss": 1.0124, + "step": 30413 + }, + { + "epoch": 1.96, + "learning_rate": 1.3494922604607227e-05, + "loss": 0.9688, + "step": 30414 + }, + { + "epoch": 1.96, + "learning_rate": 1.3494707915584278e-05, + "loss": 0.9843, + "step": 30415 + }, + { + "epoch": 1.96, + "learning_rate": 1.3494493226561327e-05, + "loss": 1.0246, + "step": 30416 + }, + { + "epoch": 1.96, + "learning_rate": 1.3494278537538376e-05, + "loss": 0.9007, + "step": 30417 + }, + { + "epoch": 1.96, + "learning_rate": 1.3494063848515427e-05, + "loss": 1.041, + "step": 30418 + }, + { + "epoch": 1.96, + "learning_rate": 1.3493849159492476e-05, + "loss": 0.9902, + "step": 30419 + }, + { + "epoch": 1.96, + "learning_rate": 1.3493634470469526e-05, + "loss": 0.9881, + "step": 30420 + }, + { + "epoch": 1.96, + "learning_rate": 1.3493419781446575e-05, + "loss": 0.9985, + "step": 30421 + }, + { + "epoch": 1.96, + "learning_rate": 1.3493205092423624e-05, + "loss": 1.0842, + "step": 30422 + }, + { + "epoch": 1.96, + "learning_rate": 1.3492990403400675e-05, + "loss": 1.0064, + "step": 30423 + }, + { + "epoch": 1.96, + "learning_rate": 1.3492775714377725e-05, + "loss": 1.0236, + "step": 30424 + }, + { + "epoch": 1.96, + "learning_rate": 1.3492561025354776e-05, + "loss": 0.9973, + "step": 30425 + }, + { + "epoch": 1.96, + "learning_rate": 1.3492346336331825e-05, + "loss": 0.8911, + "step": 30426 + }, + { + "epoch": 1.96, + "learning_rate": 1.3492131647308874e-05, + "loss": 0.9223, + "step": 30427 + }, + { + "epoch": 1.96, + "learning_rate": 1.3491916958285925e-05, + "loss": 0.9647, + "step": 30428 + }, + { + "epoch": 1.96, + "learning_rate": 1.3491702269262974e-05, + "loss": 0.913, + "step": 30429 + }, + { + "epoch": 1.96, + "learning_rate": 1.3491487580240024e-05, + "loss": 0.862, + "step": 30430 + }, + { + "epoch": 1.96, + "learning_rate": 1.3491272891217073e-05, + "loss": 0.93, + "step": 30431 + }, + { + "epoch": 1.96, + "learning_rate": 1.3491058202194122e-05, + "loss": 0.9204, + "step": 30432 + }, + { + "epoch": 1.96, + "learning_rate": 1.3490843513171173e-05, + "loss": 1.0483, + "step": 30433 + }, + { + "epoch": 1.96, + "learning_rate": 1.3490628824148222e-05, + "loss": 1.0181, + "step": 30434 + }, + { + "epoch": 1.96, + "learning_rate": 1.349041413512527e-05, + "loss": 0.9838, + "step": 30435 + }, + { + "epoch": 1.96, + "learning_rate": 1.3490199446102322e-05, + "loss": 1.0443, + "step": 30436 + }, + { + "epoch": 1.96, + "learning_rate": 1.348998475707937e-05, + "loss": 0.9906, + "step": 30437 + }, + { + "epoch": 1.96, + "learning_rate": 1.3489770068056421e-05, + "loss": 0.9455, + "step": 30438 + }, + { + "epoch": 1.96, + "learning_rate": 1.3489555379033472e-05, + "loss": 0.9919, + "step": 30439 + }, + { + "epoch": 1.96, + "learning_rate": 1.3489340690010522e-05, + "loss": 0.9617, + "step": 30440 + }, + { + "epoch": 1.96, + "learning_rate": 1.3489126000987571e-05, + "loss": 0.9672, + "step": 30441 + }, + { + "epoch": 1.96, + "learning_rate": 1.348891131196462e-05, + "loss": 0.9444, + "step": 30442 + }, + { + "epoch": 1.96, + "learning_rate": 1.3488696622941671e-05, + "loss": 0.9083, + "step": 30443 + }, + { + "epoch": 1.96, + "learning_rate": 1.348848193391872e-05, + "loss": 0.9591, + "step": 30444 + }, + { + "epoch": 1.96, + "learning_rate": 1.3488267244895769e-05, + "loss": 0.9596, + "step": 30445 + }, + { + "epoch": 1.96, + "learning_rate": 1.348805255587282e-05, + "loss": 0.9766, + "step": 30446 + }, + { + "epoch": 1.96, + "learning_rate": 1.3487837866849869e-05, + "loss": 0.9045, + "step": 30447 + }, + { + "epoch": 1.96, + "learning_rate": 1.348762317782692e-05, + "loss": 1.0077, + "step": 30448 + }, + { + "epoch": 1.96, + "learning_rate": 1.3487408488803968e-05, + "loss": 0.9002, + "step": 30449 + }, + { + "epoch": 1.96, + "learning_rate": 1.3487193799781017e-05, + "loss": 1.0241, + "step": 30450 + }, + { + "epoch": 1.96, + "learning_rate": 1.3486979110758068e-05, + "loss": 0.9204, + "step": 30451 + }, + { + "epoch": 1.96, + "learning_rate": 1.3486764421735117e-05, + "loss": 1.0167, + "step": 30452 + }, + { + "epoch": 1.96, + "learning_rate": 1.348654973271217e-05, + "loss": 0.9722, + "step": 30453 + }, + { + "epoch": 1.96, + "learning_rate": 1.3486335043689218e-05, + "loss": 0.9717, + "step": 30454 + }, + { + "epoch": 1.96, + "learning_rate": 1.3486120354666267e-05, + "loss": 0.9611, + "step": 30455 + }, + { + "epoch": 1.96, + "learning_rate": 1.3485905665643318e-05, + "loss": 0.9798, + "step": 30456 + }, + { + "epoch": 1.96, + "learning_rate": 1.3485690976620367e-05, + "loss": 0.925, + "step": 30457 + }, + { + "epoch": 1.96, + "learning_rate": 1.3485476287597417e-05, + "loss": 0.9522, + "step": 30458 + }, + { + "epoch": 1.96, + "learning_rate": 1.3485261598574466e-05, + "loss": 0.9299, + "step": 30459 + }, + { + "epoch": 1.96, + "learning_rate": 1.3485046909551515e-05, + "loss": 0.9352, + "step": 30460 + }, + { + "epoch": 1.96, + "learning_rate": 1.3484832220528566e-05, + "loss": 0.9485, + "step": 30461 + }, + { + "epoch": 1.96, + "learning_rate": 1.3484617531505615e-05, + "loss": 0.9619, + "step": 30462 + }, + { + "epoch": 1.96, + "learning_rate": 1.3484402842482664e-05, + "loss": 0.9344, + "step": 30463 + }, + { + "epoch": 1.96, + "learning_rate": 1.3484188153459715e-05, + "loss": 0.8906, + "step": 30464 + }, + { + "epoch": 1.96, + "learning_rate": 1.3483973464436764e-05, + "loss": 0.9405, + "step": 30465 + }, + { + "epoch": 1.96, + "learning_rate": 1.3483758775413812e-05, + "loss": 1.0297, + "step": 30466 + }, + { + "epoch": 1.96, + "learning_rate": 1.3483544086390863e-05, + "loss": 0.9403, + "step": 30467 + }, + { + "epoch": 1.96, + "learning_rate": 1.3483329397367915e-05, + "loss": 0.9273, + "step": 30468 + }, + { + "epoch": 1.96, + "learning_rate": 1.3483114708344964e-05, + "loss": 0.9111, + "step": 30469 + }, + { + "epoch": 1.96, + "learning_rate": 1.3482900019322013e-05, + "loss": 0.9474, + "step": 30470 + }, + { + "epoch": 1.96, + "learning_rate": 1.3482685330299064e-05, + "loss": 1.0366, + "step": 30471 + }, + { + "epoch": 1.96, + "learning_rate": 1.3482470641276113e-05, + "loss": 0.9976, + "step": 30472 + }, + { + "epoch": 1.96, + "learning_rate": 1.3482255952253162e-05, + "loss": 0.9638, + "step": 30473 + }, + { + "epoch": 1.96, + "learning_rate": 1.3482041263230213e-05, + "loss": 0.9782, + "step": 30474 + }, + { + "epoch": 1.96, + "learning_rate": 1.3481826574207262e-05, + "loss": 0.9549, + "step": 30475 + }, + { + "epoch": 1.96, + "learning_rate": 1.348161188518431e-05, + "loss": 0.9766, + "step": 30476 + }, + { + "epoch": 1.96, + "learning_rate": 1.3481397196161361e-05, + "loss": 1.0726, + "step": 30477 + }, + { + "epoch": 1.96, + "learning_rate": 1.348118250713841e-05, + "loss": 0.9957, + "step": 30478 + }, + { + "epoch": 1.96, + "learning_rate": 1.3480967818115461e-05, + "loss": 0.9085, + "step": 30479 + }, + { + "epoch": 1.96, + "learning_rate": 1.348075312909251e-05, + "loss": 0.9207, + "step": 30480 + }, + { + "epoch": 1.96, + "learning_rate": 1.3480538440069559e-05, + "loss": 0.9642, + "step": 30481 + }, + { + "epoch": 1.96, + "learning_rate": 1.3480323751046611e-05, + "loss": 0.9709, + "step": 30482 + }, + { + "epoch": 1.96, + "learning_rate": 1.348010906202366e-05, + "loss": 0.9489, + "step": 30483 + }, + { + "epoch": 1.96, + "learning_rate": 1.347989437300071e-05, + "loss": 0.9303, + "step": 30484 + }, + { + "epoch": 1.96, + "learning_rate": 1.347967968397776e-05, + "loss": 0.9088, + "step": 30485 + }, + { + "epoch": 1.96, + "learning_rate": 1.3479464994954809e-05, + "loss": 1.0711, + "step": 30486 + }, + { + "epoch": 1.96, + "learning_rate": 1.347925030593186e-05, + "loss": 0.9254, + "step": 30487 + }, + { + "epoch": 1.96, + "learning_rate": 1.3479035616908908e-05, + "loss": 0.991, + "step": 30488 + }, + { + "epoch": 1.96, + "learning_rate": 1.3478820927885959e-05, + "loss": 1.0512, + "step": 30489 + }, + { + "epoch": 1.96, + "learning_rate": 1.3478606238863008e-05, + "loss": 0.9662, + "step": 30490 + }, + { + "epoch": 1.96, + "learning_rate": 1.3478391549840057e-05, + "loss": 0.9867, + "step": 30491 + }, + { + "epoch": 1.96, + "learning_rate": 1.3478176860817108e-05, + "loss": 1.0149, + "step": 30492 + }, + { + "epoch": 1.96, + "learning_rate": 1.3477962171794157e-05, + "loss": 1.0069, + "step": 30493 + }, + { + "epoch": 1.96, + "learning_rate": 1.3477747482771206e-05, + "loss": 1.0145, + "step": 30494 + }, + { + "epoch": 1.96, + "learning_rate": 1.3477532793748256e-05, + "loss": 0.9814, + "step": 30495 + }, + { + "epoch": 1.96, + "learning_rate": 1.3477318104725305e-05, + "loss": 1.0144, + "step": 30496 + }, + { + "epoch": 1.96, + "learning_rate": 1.3477103415702357e-05, + "loss": 0.9144, + "step": 30497 + }, + { + "epoch": 1.96, + "learning_rate": 1.3476888726679406e-05, + "loss": 0.9786, + "step": 30498 + }, + { + "epoch": 1.96, + "learning_rate": 1.3476674037656457e-05, + "loss": 0.9622, + "step": 30499 + }, + { + "epoch": 1.96, + "learning_rate": 1.3476459348633506e-05, + "loss": 0.952, + "step": 30500 + }, + { + "epoch": 1.96, + "learning_rate": 1.3476244659610555e-05, + "loss": 0.9785, + "step": 30501 + }, + { + "epoch": 1.96, + "learning_rate": 1.3476029970587606e-05, + "loss": 0.8969, + "step": 30502 + }, + { + "epoch": 1.96, + "learning_rate": 1.3475815281564655e-05, + "loss": 1.0189, + "step": 30503 + }, + { + "epoch": 1.96, + "learning_rate": 1.3475600592541704e-05, + "loss": 0.9413, + "step": 30504 + }, + { + "epoch": 1.96, + "learning_rate": 1.3475385903518754e-05, + "loss": 1.0001, + "step": 30505 + }, + { + "epoch": 1.96, + "learning_rate": 1.3475171214495803e-05, + "loss": 0.9788, + "step": 30506 + }, + { + "epoch": 1.96, + "learning_rate": 1.3474956525472854e-05, + "loss": 1.0674, + "step": 30507 + }, + { + "epoch": 1.96, + "learning_rate": 1.3474956525472854e-05, + "loss": 0.8974, + "step": 30508 + }, + { + "epoch": 1.96, + "learning_rate": 1.3474741836449903e-05, + "loss": 0.9903, + "step": 30509 + }, + { + "epoch": 1.96, + "learning_rate": 1.3474527147426952e-05, + "loss": 1.0225, + "step": 30510 + }, + { + "epoch": 1.96, + "learning_rate": 1.3474312458404003e-05, + "loss": 0.9542, + "step": 30511 + }, + { + "epoch": 1.96, + "learning_rate": 1.3474097769381053e-05, + "loss": 0.9379, + "step": 30512 + }, + { + "epoch": 1.96, + "learning_rate": 1.3473883080358104e-05, + "loss": 1.0902, + "step": 30513 + }, + { + "epoch": 1.96, + "learning_rate": 1.3473668391335153e-05, + "loss": 0.957, + "step": 30514 + }, + { + "epoch": 1.96, + "learning_rate": 1.3473453702312202e-05, + "loss": 1.0073, + "step": 30515 + }, + { + "epoch": 1.96, + "learning_rate": 1.3473239013289252e-05, + "loss": 0.9628, + "step": 30516 + }, + { + "epoch": 1.96, + "learning_rate": 1.3473024324266301e-05, + "loss": 0.9476, + "step": 30517 + }, + { + "epoch": 1.96, + "learning_rate": 1.3472809635243352e-05, + "loss": 1.0118, + "step": 30518 + }, + { + "epoch": 1.96, + "learning_rate": 1.3472594946220401e-05, + "loss": 0.9244, + "step": 30519 + }, + { + "epoch": 1.96, + "learning_rate": 1.347238025719745e-05, + "loss": 0.9246, + "step": 30520 + }, + { + "epoch": 1.96, + "learning_rate": 1.34721655681745e-05, + "loss": 0.9432, + "step": 30521 + }, + { + "epoch": 1.96, + "learning_rate": 1.347195087915155e-05, + "loss": 0.9191, + "step": 30522 + }, + { + "epoch": 1.96, + "learning_rate": 1.3471736190128599e-05, + "loss": 0.9989, + "step": 30523 + }, + { + "epoch": 1.96, + "learning_rate": 1.347152150110565e-05, + "loss": 0.9608, + "step": 30524 + }, + { + "epoch": 1.96, + "learning_rate": 1.3471306812082698e-05, + "loss": 0.9554, + "step": 30525 + }, + { + "epoch": 1.96, + "learning_rate": 1.3471092123059749e-05, + "loss": 1.0347, + "step": 30526 + }, + { + "epoch": 1.96, + "learning_rate": 1.34708774340368e-05, + "loss": 0.9494, + "step": 30527 + }, + { + "epoch": 1.96, + "learning_rate": 1.347066274501385e-05, + "loss": 0.967, + "step": 30528 + }, + { + "epoch": 1.96, + "learning_rate": 1.3470448055990899e-05, + "loss": 1.1094, + "step": 30529 + }, + { + "epoch": 1.96, + "learning_rate": 1.3470233366967948e-05, + "loss": 0.9431, + "step": 30530 + }, + { + "epoch": 1.96, + "learning_rate": 1.3470018677944999e-05, + "loss": 0.9321, + "step": 30531 + }, + { + "epoch": 1.96, + "learning_rate": 1.3469803988922048e-05, + "loss": 0.9865, + "step": 30532 + }, + { + "epoch": 1.96, + "learning_rate": 1.3469589299899097e-05, + "loss": 0.9683, + "step": 30533 + }, + { + "epoch": 1.96, + "learning_rate": 1.3469374610876147e-05, + "loss": 0.989, + "step": 30534 + }, + { + "epoch": 1.96, + "learning_rate": 1.3469159921853196e-05, + "loss": 0.9702, + "step": 30535 + }, + { + "epoch": 1.96, + "learning_rate": 1.3468945232830247e-05, + "loss": 1.0105, + "step": 30536 + }, + { + "epoch": 1.96, + "learning_rate": 1.3468730543807296e-05, + "loss": 0.9747, + "step": 30537 + }, + { + "epoch": 1.96, + "learning_rate": 1.3468515854784345e-05, + "loss": 0.9918, + "step": 30538 + }, + { + "epoch": 1.96, + "learning_rate": 1.3468301165761396e-05, + "loss": 0.9363, + "step": 30539 + }, + { + "epoch": 1.96, + "learning_rate": 1.3468086476738445e-05, + "loss": 0.9522, + "step": 30540 + }, + { + "epoch": 1.96, + "learning_rate": 1.3467871787715497e-05, + "loss": 0.9017, + "step": 30541 + }, + { + "epoch": 1.97, + "learning_rate": 1.3467657098692546e-05, + "loss": 1.0076, + "step": 30542 + }, + { + "epoch": 1.97, + "learning_rate": 1.3467442409669595e-05, + "loss": 0.9544, + "step": 30543 + }, + { + "epoch": 1.97, + "learning_rate": 1.3467227720646645e-05, + "loss": 0.987, + "step": 30544 + }, + { + "epoch": 1.97, + "learning_rate": 1.3467013031623694e-05, + "loss": 0.9933, + "step": 30545 + }, + { + "epoch": 1.97, + "learning_rate": 1.3466798342600745e-05, + "loss": 0.9679, + "step": 30546 + }, + { + "epoch": 1.97, + "learning_rate": 1.3466583653577794e-05, + "loss": 0.9706, + "step": 30547 + }, + { + "epoch": 1.97, + "learning_rate": 1.3466368964554843e-05, + "loss": 0.9249, + "step": 30548 + }, + { + "epoch": 1.97, + "learning_rate": 1.3466154275531894e-05, + "loss": 0.9407, + "step": 30549 + }, + { + "epoch": 1.97, + "learning_rate": 1.3465939586508943e-05, + "loss": 1.0147, + "step": 30550 + }, + { + "epoch": 1.97, + "learning_rate": 1.3465724897485992e-05, + "loss": 0.9081, + "step": 30551 + }, + { + "epoch": 1.97, + "learning_rate": 1.3465510208463042e-05, + "loss": 0.9859, + "step": 30552 + }, + { + "epoch": 1.97, + "learning_rate": 1.3465295519440091e-05, + "loss": 0.9998, + "step": 30553 + }, + { + "epoch": 1.97, + "learning_rate": 1.346508083041714e-05, + "loss": 0.9444, + "step": 30554 + }, + { + "epoch": 1.97, + "learning_rate": 1.3464866141394191e-05, + "loss": 0.9482, + "step": 30555 + }, + { + "epoch": 1.97, + "learning_rate": 1.3464651452371243e-05, + "loss": 0.9263, + "step": 30556 + }, + { + "epoch": 1.97, + "learning_rate": 1.3464436763348292e-05, + "loss": 0.9715, + "step": 30557 + }, + { + "epoch": 1.97, + "learning_rate": 1.3464222074325341e-05, + "loss": 0.9251, + "step": 30558 + }, + { + "epoch": 1.97, + "learning_rate": 1.3464007385302392e-05, + "loss": 0.9685, + "step": 30559 + }, + { + "epoch": 1.97, + "learning_rate": 1.346379269627944e-05, + "loss": 0.923, + "step": 30560 + }, + { + "epoch": 1.97, + "learning_rate": 1.346357800725649e-05, + "loss": 1.0019, + "step": 30561 + }, + { + "epoch": 1.97, + "learning_rate": 1.346336331823354e-05, + "loss": 0.981, + "step": 30562 + }, + { + "epoch": 1.97, + "learning_rate": 1.346314862921059e-05, + "loss": 1.013, + "step": 30563 + }, + { + "epoch": 1.97, + "learning_rate": 1.3462933940187638e-05, + "loss": 1.0185, + "step": 30564 + }, + { + "epoch": 1.97, + "learning_rate": 1.3462719251164689e-05, + "loss": 0.9458, + "step": 30565 + }, + { + "epoch": 1.97, + "learning_rate": 1.3462504562141738e-05, + "loss": 0.9879, + "step": 30566 + }, + { + "epoch": 1.97, + "learning_rate": 1.3462289873118789e-05, + "loss": 0.9936, + "step": 30567 + }, + { + "epoch": 1.97, + "learning_rate": 1.3462075184095838e-05, + "loss": 1.012, + "step": 30568 + }, + { + "epoch": 1.97, + "learning_rate": 1.3461860495072887e-05, + "loss": 0.9868, + "step": 30569 + }, + { + "epoch": 1.97, + "learning_rate": 1.3461645806049939e-05, + "loss": 0.9144, + "step": 30570 + }, + { + "epoch": 1.97, + "learning_rate": 1.3461431117026988e-05, + "loss": 0.9046, + "step": 30571 + }, + { + "epoch": 1.97, + "learning_rate": 1.3461216428004038e-05, + "loss": 0.9676, + "step": 30572 + }, + { + "epoch": 1.97, + "learning_rate": 1.3461001738981087e-05, + "loss": 1.01, + "step": 30573 + }, + { + "epoch": 1.97, + "learning_rate": 1.3460787049958136e-05, + "loss": 1.0704, + "step": 30574 + }, + { + "epoch": 1.97, + "learning_rate": 1.3460572360935187e-05, + "loss": 0.9875, + "step": 30575 + }, + { + "epoch": 1.97, + "learning_rate": 1.3460357671912236e-05, + "loss": 0.9348, + "step": 30576 + }, + { + "epoch": 1.97, + "learning_rate": 1.3460142982889287e-05, + "loss": 0.9261, + "step": 30577 + }, + { + "epoch": 1.97, + "learning_rate": 1.3459928293866336e-05, + "loss": 1.0098, + "step": 30578 + }, + { + "epoch": 1.97, + "learning_rate": 1.3459713604843385e-05, + "loss": 0.9449, + "step": 30579 + }, + { + "epoch": 1.97, + "learning_rate": 1.3459498915820435e-05, + "loss": 0.9236, + "step": 30580 + }, + { + "epoch": 1.97, + "learning_rate": 1.3459284226797484e-05, + "loss": 0.9847, + "step": 30581 + }, + { + "epoch": 1.97, + "learning_rate": 1.3459069537774533e-05, + "loss": 0.9293, + "step": 30582 + }, + { + "epoch": 1.97, + "learning_rate": 1.3458854848751584e-05, + "loss": 0.9685, + "step": 30583 + }, + { + "epoch": 1.97, + "learning_rate": 1.3458640159728633e-05, + "loss": 0.9381, + "step": 30584 + }, + { + "epoch": 1.97, + "learning_rate": 1.3458425470705685e-05, + "loss": 0.9948, + "step": 30585 + }, + { + "epoch": 1.97, + "learning_rate": 1.3458210781682734e-05, + "loss": 0.9296, + "step": 30586 + }, + { + "epoch": 1.97, + "learning_rate": 1.3457996092659785e-05, + "loss": 0.9512, + "step": 30587 + }, + { + "epoch": 1.97, + "learning_rate": 1.3457781403636834e-05, + "loss": 1.0012, + "step": 30588 + }, + { + "epoch": 1.97, + "learning_rate": 1.3457566714613883e-05, + "loss": 0.9547, + "step": 30589 + }, + { + "epoch": 1.97, + "learning_rate": 1.3457352025590933e-05, + "loss": 0.9326, + "step": 30590 + }, + { + "epoch": 1.97, + "learning_rate": 1.3457137336567982e-05, + "loss": 1.0381, + "step": 30591 + }, + { + "epoch": 1.97, + "learning_rate": 1.3456922647545031e-05, + "loss": 0.9613, + "step": 30592 + }, + { + "epoch": 1.97, + "learning_rate": 1.3456707958522082e-05, + "loss": 0.9778, + "step": 30593 + }, + { + "epoch": 1.97, + "learning_rate": 1.3456493269499131e-05, + "loss": 0.98, + "step": 30594 + }, + { + "epoch": 1.97, + "learning_rate": 1.3456278580476182e-05, + "loss": 0.9256, + "step": 30595 + }, + { + "epoch": 1.97, + "learning_rate": 1.345606389145323e-05, + "loss": 0.9148, + "step": 30596 + }, + { + "epoch": 1.97, + "learning_rate": 1.345584920243028e-05, + "loss": 0.9692, + "step": 30597 + }, + { + "epoch": 1.97, + "learning_rate": 1.345563451340733e-05, + "loss": 0.972, + "step": 30598 + }, + { + "epoch": 1.97, + "learning_rate": 1.3455419824384381e-05, + "loss": 1.0222, + "step": 30599 + }, + { + "epoch": 1.97, + "learning_rate": 1.3455205135361432e-05, + "loss": 0.9128, + "step": 30600 + }, + { + "epoch": 1.97, + "learning_rate": 1.345499044633848e-05, + "loss": 1.001, + "step": 30601 + }, + { + "epoch": 1.97, + "learning_rate": 1.345477575731553e-05, + "loss": 0.9769, + "step": 30602 + }, + { + "epoch": 1.97, + "learning_rate": 1.345456106829258e-05, + "loss": 1.0178, + "step": 30603 + }, + { + "epoch": 1.97, + "learning_rate": 1.3454346379269629e-05, + "loss": 0.9542, + "step": 30604 + }, + { + "epoch": 1.97, + "learning_rate": 1.345413169024668e-05, + "loss": 0.9493, + "step": 30605 + }, + { + "epoch": 1.97, + "learning_rate": 1.3453917001223729e-05, + "loss": 0.9308, + "step": 30606 + }, + { + "epoch": 1.97, + "learning_rate": 1.3453702312200778e-05, + "loss": 0.9009, + "step": 30607 + }, + { + "epoch": 1.97, + "learning_rate": 1.3453487623177828e-05, + "loss": 0.9722, + "step": 30608 + }, + { + "epoch": 1.97, + "learning_rate": 1.3453272934154877e-05, + "loss": 0.9956, + "step": 30609 + }, + { + "epoch": 1.97, + "learning_rate": 1.3453058245131926e-05, + "loss": 1.0314, + "step": 30610 + }, + { + "epoch": 1.97, + "learning_rate": 1.3452843556108977e-05, + "loss": 0.9562, + "step": 30611 + }, + { + "epoch": 1.97, + "learning_rate": 1.3452628867086026e-05, + "loss": 0.9218, + "step": 30612 + }, + { + "epoch": 1.97, + "learning_rate": 1.3452414178063075e-05, + "loss": 0.92, + "step": 30613 + }, + { + "epoch": 1.97, + "learning_rate": 1.3452199489040127e-05, + "loss": 0.9504, + "step": 30614 + }, + { + "epoch": 1.97, + "learning_rate": 1.3451984800017178e-05, + "loss": 0.9424, + "step": 30615 + }, + { + "epoch": 1.97, + "learning_rate": 1.3451770110994227e-05, + "loss": 0.9413, + "step": 30616 + }, + { + "epoch": 1.97, + "learning_rate": 1.3451555421971276e-05, + "loss": 0.8874, + "step": 30617 + }, + { + "epoch": 1.97, + "learning_rate": 1.3451340732948326e-05, + "loss": 0.962, + "step": 30618 + }, + { + "epoch": 1.97, + "learning_rate": 1.3451126043925375e-05, + "loss": 0.9563, + "step": 30619 + }, + { + "epoch": 1.97, + "learning_rate": 1.3450911354902424e-05, + "loss": 1.0056, + "step": 30620 + }, + { + "epoch": 1.97, + "learning_rate": 1.3450696665879475e-05, + "loss": 0.9049, + "step": 30621 + }, + { + "epoch": 1.97, + "learning_rate": 1.3450481976856524e-05, + "loss": 0.9809, + "step": 30622 + }, + { + "epoch": 1.97, + "learning_rate": 1.3450267287833575e-05, + "loss": 0.9622, + "step": 30623 + }, + { + "epoch": 1.97, + "learning_rate": 1.3450052598810624e-05, + "loss": 0.9899, + "step": 30624 + }, + { + "epoch": 1.97, + "learning_rate": 1.3449837909787673e-05, + "loss": 0.9435, + "step": 30625 + }, + { + "epoch": 1.97, + "learning_rate": 1.3449623220764723e-05, + "loss": 0.8817, + "step": 30626 + }, + { + "epoch": 1.97, + "learning_rate": 1.3449408531741772e-05, + "loss": 1.0038, + "step": 30627 + }, + { + "epoch": 1.97, + "learning_rate": 1.3449193842718821e-05, + "loss": 1.0004, + "step": 30628 + }, + { + "epoch": 1.97, + "learning_rate": 1.3448979153695874e-05, + "loss": 0.9311, + "step": 30629 + }, + { + "epoch": 1.97, + "learning_rate": 1.3448764464672923e-05, + "loss": 0.9052, + "step": 30630 + }, + { + "epoch": 1.97, + "learning_rate": 1.3448549775649973e-05, + "loss": 0.9878, + "step": 30631 + }, + { + "epoch": 1.97, + "learning_rate": 1.3448335086627022e-05, + "loss": 0.9865, + "step": 30632 + }, + { + "epoch": 1.97, + "learning_rate": 1.3448120397604073e-05, + "loss": 1.0092, + "step": 30633 + }, + { + "epoch": 1.97, + "learning_rate": 1.3447905708581122e-05, + "loss": 0.9397, + "step": 30634 + }, + { + "epoch": 1.97, + "learning_rate": 1.344769101955817e-05, + "loss": 1.0363, + "step": 30635 + }, + { + "epoch": 1.97, + "learning_rate": 1.3447476330535221e-05, + "loss": 1.0055, + "step": 30636 + }, + { + "epoch": 1.97, + "learning_rate": 1.344726164151227e-05, + "loss": 0.9763, + "step": 30637 + }, + { + "epoch": 1.97, + "learning_rate": 1.344704695248932e-05, + "loss": 0.9242, + "step": 30638 + }, + { + "epoch": 1.97, + "learning_rate": 1.344683226346637e-05, + "loss": 0.9815, + "step": 30639 + }, + { + "epoch": 1.97, + "learning_rate": 1.3446617574443419e-05, + "loss": 0.9693, + "step": 30640 + }, + { + "epoch": 1.97, + "learning_rate": 1.3446402885420468e-05, + "loss": 0.8809, + "step": 30641 + }, + { + "epoch": 1.97, + "learning_rate": 1.3446188196397519e-05, + "loss": 0.9743, + "step": 30642 + }, + { + "epoch": 1.97, + "learning_rate": 1.3445973507374571e-05, + "loss": 0.9315, + "step": 30643 + }, + { + "epoch": 1.97, + "learning_rate": 1.344575881835162e-05, + "loss": 0.9967, + "step": 30644 + }, + { + "epoch": 1.97, + "learning_rate": 1.3445544129328669e-05, + "loss": 1.025, + "step": 30645 + }, + { + "epoch": 1.97, + "learning_rate": 1.344532944030572e-05, + "loss": 0.9579, + "step": 30646 + }, + { + "epoch": 1.97, + "learning_rate": 1.3445114751282768e-05, + "loss": 0.9946, + "step": 30647 + }, + { + "epoch": 1.97, + "learning_rate": 1.3444900062259817e-05, + "loss": 0.931, + "step": 30648 + }, + { + "epoch": 1.97, + "learning_rate": 1.3444685373236868e-05, + "loss": 0.9483, + "step": 30649 + }, + { + "epoch": 1.97, + "learning_rate": 1.3444470684213917e-05, + "loss": 0.9225, + "step": 30650 + }, + { + "epoch": 1.97, + "learning_rate": 1.3444255995190966e-05, + "loss": 0.9953, + "step": 30651 + }, + { + "epoch": 1.97, + "learning_rate": 1.3444041306168017e-05, + "loss": 1.0173, + "step": 30652 + }, + { + "epoch": 1.97, + "learning_rate": 1.3443826617145066e-05, + "loss": 1.0375, + "step": 30653 + }, + { + "epoch": 1.97, + "learning_rate": 1.3443611928122116e-05, + "loss": 0.9145, + "step": 30654 + }, + { + "epoch": 1.97, + "learning_rate": 1.3443397239099165e-05, + "loss": 0.9084, + "step": 30655 + }, + { + "epoch": 1.97, + "learning_rate": 1.3443182550076214e-05, + "loss": 0.8748, + "step": 30656 + }, + { + "epoch": 1.97, + "learning_rate": 1.3442967861053265e-05, + "loss": 0.9739, + "step": 30657 + }, + { + "epoch": 1.97, + "learning_rate": 1.3442753172030316e-05, + "loss": 0.9745, + "step": 30658 + }, + { + "epoch": 1.97, + "learning_rate": 1.3442538483007366e-05, + "loss": 1.0098, + "step": 30659 + }, + { + "epoch": 1.97, + "learning_rate": 1.3442323793984415e-05, + "loss": 0.9562, + "step": 30660 + }, + { + "epoch": 1.97, + "learning_rate": 1.3442109104961464e-05, + "loss": 1.0091, + "step": 30661 + }, + { + "epoch": 1.97, + "learning_rate": 1.3441894415938515e-05, + "loss": 0.9713, + "step": 30662 + }, + { + "epoch": 1.97, + "learning_rate": 1.3441679726915564e-05, + "loss": 0.9134, + "step": 30663 + }, + { + "epoch": 1.97, + "learning_rate": 1.3441465037892614e-05, + "loss": 1.0496, + "step": 30664 + }, + { + "epoch": 1.97, + "learning_rate": 1.3441250348869663e-05, + "loss": 0.9487, + "step": 30665 + }, + { + "epoch": 1.97, + "learning_rate": 1.3441035659846712e-05, + "loss": 0.9044, + "step": 30666 + }, + { + "epoch": 1.97, + "learning_rate": 1.3440820970823763e-05, + "loss": 0.8962, + "step": 30667 + }, + { + "epoch": 1.97, + "learning_rate": 1.3440606281800812e-05, + "loss": 0.9217, + "step": 30668 + }, + { + "epoch": 1.97, + "learning_rate": 1.3440391592777861e-05, + "loss": 0.944, + "step": 30669 + }, + { + "epoch": 1.97, + "learning_rate": 1.3440176903754912e-05, + "loss": 1.0274, + "step": 30670 + }, + { + "epoch": 1.97, + "learning_rate": 1.343996221473196e-05, + "loss": 0.9698, + "step": 30671 + }, + { + "epoch": 1.97, + "learning_rate": 1.3439747525709013e-05, + "loss": 0.9503, + "step": 30672 + }, + { + "epoch": 1.97, + "learning_rate": 1.3439532836686062e-05, + "loss": 1.0212, + "step": 30673 + }, + { + "epoch": 1.97, + "learning_rate": 1.3439318147663113e-05, + "loss": 0.9058, + "step": 30674 + }, + { + "epoch": 1.97, + "learning_rate": 1.3439103458640162e-05, + "loss": 0.9281, + "step": 30675 + }, + { + "epoch": 1.97, + "learning_rate": 1.343888876961721e-05, + "loss": 0.9747, + "step": 30676 + }, + { + "epoch": 1.97, + "learning_rate": 1.3438674080594261e-05, + "loss": 0.9156, + "step": 30677 + }, + { + "epoch": 1.97, + "learning_rate": 1.343845939157131e-05, + "loss": 0.981, + "step": 30678 + }, + { + "epoch": 1.97, + "learning_rate": 1.3438244702548359e-05, + "loss": 0.9548, + "step": 30679 + }, + { + "epoch": 1.97, + "learning_rate": 1.343803001352541e-05, + "loss": 0.9772, + "step": 30680 + }, + { + "epoch": 1.97, + "learning_rate": 1.3437815324502459e-05, + "loss": 0.9582, + "step": 30681 + }, + { + "epoch": 1.97, + "learning_rate": 1.343760063547951e-05, + "loss": 0.9864, + "step": 30682 + }, + { + "epoch": 1.97, + "learning_rate": 1.3437385946456558e-05, + "loss": 0.9141, + "step": 30683 + }, + { + "epoch": 1.97, + "learning_rate": 1.3437171257433607e-05, + "loss": 1.0164, + "step": 30684 + }, + { + "epoch": 1.97, + "learning_rate": 1.3436956568410658e-05, + "loss": 0.9166, + "step": 30685 + }, + { + "epoch": 1.97, + "learning_rate": 1.3436741879387707e-05, + "loss": 0.9633, + "step": 30686 + }, + { + "epoch": 1.97, + "learning_rate": 1.343652719036476e-05, + "loss": 0.9939, + "step": 30687 + }, + { + "epoch": 1.97, + "learning_rate": 1.3436312501341808e-05, + "loss": 0.9748, + "step": 30688 + }, + { + "epoch": 1.97, + "learning_rate": 1.3436097812318857e-05, + "loss": 1.001, + "step": 30689 + }, + { + "epoch": 1.97, + "learning_rate": 1.3435883123295908e-05, + "loss": 0.9362, + "step": 30690 + }, + { + "epoch": 1.97, + "learning_rate": 1.3435668434272957e-05, + "loss": 0.9422, + "step": 30691 + }, + { + "epoch": 1.97, + "learning_rate": 1.3435453745250007e-05, + "loss": 0.8982, + "step": 30692 + }, + { + "epoch": 1.97, + "learning_rate": 1.3435239056227056e-05, + "loss": 0.928, + "step": 30693 + }, + { + "epoch": 1.97, + "learning_rate": 1.3435024367204105e-05, + "loss": 0.9786, + "step": 30694 + }, + { + "epoch": 1.97, + "learning_rate": 1.3434809678181156e-05, + "loss": 0.9207, + "step": 30695 + }, + { + "epoch": 1.97, + "learning_rate": 1.3434594989158205e-05, + "loss": 0.9327, + "step": 30696 + }, + { + "epoch": 1.97, + "learning_rate": 1.3434380300135254e-05, + "loss": 0.9082, + "step": 30697 + }, + { + "epoch": 1.98, + "learning_rate": 1.3434165611112305e-05, + "loss": 0.9838, + "step": 30698 + }, + { + "epoch": 1.98, + "learning_rate": 1.3433950922089354e-05, + "loss": 1.0294, + "step": 30699 + }, + { + "epoch": 1.98, + "learning_rate": 1.3433736233066403e-05, + "loss": 0.9282, + "step": 30700 + }, + { + "epoch": 1.98, + "learning_rate": 1.3433521544043455e-05, + "loss": 0.984, + "step": 30701 + }, + { + "epoch": 1.98, + "learning_rate": 1.3433306855020506e-05, + "loss": 0.9514, + "step": 30702 + }, + { + "epoch": 1.98, + "learning_rate": 1.3433092165997555e-05, + "loss": 0.9813, + "step": 30703 + }, + { + "epoch": 1.98, + "learning_rate": 1.3432877476974604e-05, + "loss": 0.9769, + "step": 30704 + }, + { + "epoch": 1.98, + "learning_rate": 1.3432662787951654e-05, + "loss": 0.9459, + "step": 30705 + }, + { + "epoch": 1.98, + "learning_rate": 1.3432448098928703e-05, + "loss": 1.0208, + "step": 30706 + }, + { + "epoch": 1.98, + "learning_rate": 1.3432233409905752e-05, + "loss": 0.9557, + "step": 30707 + }, + { + "epoch": 1.98, + "learning_rate": 1.3432018720882803e-05, + "loss": 0.8867, + "step": 30708 + }, + { + "epoch": 1.98, + "learning_rate": 1.3431804031859852e-05, + "loss": 1.0079, + "step": 30709 + }, + { + "epoch": 1.98, + "learning_rate": 1.34315893428369e-05, + "loss": 1.0332, + "step": 30710 + }, + { + "epoch": 1.98, + "learning_rate": 1.3431374653813951e-05, + "loss": 0.9357, + "step": 30711 + }, + { + "epoch": 1.98, + "learning_rate": 1.3431159964791e-05, + "loss": 0.9924, + "step": 30712 + }, + { + "epoch": 1.98, + "learning_rate": 1.3430945275768051e-05, + "loss": 0.9164, + "step": 30713 + }, + { + "epoch": 1.98, + "learning_rate": 1.34307305867451e-05, + "loss": 0.9398, + "step": 30714 + }, + { + "epoch": 1.98, + "learning_rate": 1.3430515897722149e-05, + "loss": 0.9679, + "step": 30715 + }, + { + "epoch": 1.98, + "learning_rate": 1.3430301208699201e-05, + "loss": 0.9154, + "step": 30716 + }, + { + "epoch": 1.98, + "learning_rate": 1.343008651967625e-05, + "loss": 0.9167, + "step": 30717 + }, + { + "epoch": 1.98, + "learning_rate": 1.3429871830653301e-05, + "loss": 0.9367, + "step": 30718 + }, + { + "epoch": 1.98, + "learning_rate": 1.342965714163035e-05, + "loss": 0.9959, + "step": 30719 + }, + { + "epoch": 1.98, + "learning_rate": 1.3429442452607399e-05, + "loss": 0.9815, + "step": 30720 + }, + { + "epoch": 1.98, + "learning_rate": 1.342922776358445e-05, + "loss": 0.9106, + "step": 30721 + }, + { + "epoch": 1.98, + "learning_rate": 1.3429013074561498e-05, + "loss": 0.9883, + "step": 30722 + }, + { + "epoch": 1.98, + "learning_rate": 1.3428798385538549e-05, + "loss": 1.0246, + "step": 30723 + }, + { + "epoch": 1.98, + "learning_rate": 1.3428583696515598e-05, + "loss": 0.9012, + "step": 30724 + }, + { + "epoch": 1.98, + "learning_rate": 1.3428369007492647e-05, + "loss": 1.0311, + "step": 30725 + }, + { + "epoch": 1.98, + "learning_rate": 1.3428154318469698e-05, + "loss": 0.9825, + "step": 30726 + }, + { + "epoch": 1.98, + "learning_rate": 1.3427939629446747e-05, + "loss": 0.9308, + "step": 30727 + }, + { + "epoch": 1.98, + "learning_rate": 1.3427724940423796e-05, + "loss": 0.9049, + "step": 30728 + }, + { + "epoch": 1.98, + "learning_rate": 1.3427510251400846e-05, + "loss": 0.9868, + "step": 30729 + }, + { + "epoch": 1.98, + "learning_rate": 1.3427295562377897e-05, + "loss": 0.9792, + "step": 30730 + }, + { + "epoch": 1.98, + "learning_rate": 1.3427080873354948e-05, + "loss": 0.9936, + "step": 30731 + }, + { + "epoch": 1.98, + "learning_rate": 1.3426866184331997e-05, + "loss": 1.0085, + "step": 30732 + }, + { + "epoch": 1.98, + "learning_rate": 1.3426651495309047e-05, + "loss": 0.9421, + "step": 30733 + }, + { + "epoch": 1.98, + "learning_rate": 1.3426436806286096e-05, + "loss": 0.975, + "step": 30734 + }, + { + "epoch": 1.98, + "learning_rate": 1.3426222117263145e-05, + "loss": 0.8678, + "step": 30735 + }, + { + "epoch": 1.98, + "learning_rate": 1.3426007428240196e-05, + "loss": 1.0147, + "step": 30736 + }, + { + "epoch": 1.98, + "learning_rate": 1.3425792739217245e-05, + "loss": 0.8996, + "step": 30737 + }, + { + "epoch": 1.98, + "learning_rate": 1.3425578050194294e-05, + "loss": 0.9667, + "step": 30738 + }, + { + "epoch": 1.98, + "learning_rate": 1.3425363361171344e-05, + "loss": 0.976, + "step": 30739 + }, + { + "epoch": 1.98, + "learning_rate": 1.3425148672148393e-05, + "loss": 0.9231, + "step": 30740 + }, + { + "epoch": 1.98, + "learning_rate": 1.3424933983125444e-05, + "loss": 1.0152, + "step": 30741 + }, + { + "epoch": 1.98, + "learning_rate": 1.3424719294102493e-05, + "loss": 1.0072, + "step": 30742 + }, + { + "epoch": 1.98, + "learning_rate": 1.3424504605079542e-05, + "loss": 0.917, + "step": 30743 + }, + { + "epoch": 1.98, + "learning_rate": 1.3424289916056593e-05, + "loss": 1.0041, + "step": 30744 + }, + { + "epoch": 1.98, + "learning_rate": 1.3424075227033643e-05, + "loss": 0.9817, + "step": 30745 + }, + { + "epoch": 1.98, + "learning_rate": 1.3423860538010694e-05, + "loss": 0.9478, + "step": 30746 + }, + { + "epoch": 1.98, + "learning_rate": 1.3423645848987743e-05, + "loss": 0.9818, + "step": 30747 + }, + { + "epoch": 1.98, + "learning_rate": 1.3423431159964792e-05, + "loss": 0.9429, + "step": 30748 + }, + { + "epoch": 1.98, + "learning_rate": 1.3423216470941843e-05, + "loss": 1.0216, + "step": 30749 + }, + { + "epoch": 1.98, + "learning_rate": 1.3423001781918891e-05, + "loss": 0.9506, + "step": 30750 + }, + { + "epoch": 1.98, + "learning_rate": 1.3422787092895942e-05, + "loss": 0.9301, + "step": 30751 + }, + { + "epoch": 1.98, + "learning_rate": 1.3422572403872991e-05, + "loss": 0.9915, + "step": 30752 + }, + { + "epoch": 1.98, + "learning_rate": 1.342235771485004e-05, + "loss": 0.9977, + "step": 30753 + }, + { + "epoch": 1.98, + "learning_rate": 1.342214302582709e-05, + "loss": 0.9965, + "step": 30754 + }, + { + "epoch": 1.98, + "learning_rate": 1.342192833680414e-05, + "loss": 1.0192, + "step": 30755 + }, + { + "epoch": 1.98, + "learning_rate": 1.3421713647781189e-05, + "loss": 0.9048, + "step": 30756 + }, + { + "epoch": 1.98, + "learning_rate": 1.342149895875824e-05, + "loss": 0.9641, + "step": 30757 + }, + { + "epoch": 1.98, + "learning_rate": 1.3421284269735288e-05, + "loss": 0.9102, + "step": 30758 + }, + { + "epoch": 1.98, + "learning_rate": 1.342106958071234e-05, + "loss": 0.9857, + "step": 30759 + }, + { + "epoch": 1.98, + "learning_rate": 1.342085489168939e-05, + "loss": 0.9343, + "step": 30760 + }, + { + "epoch": 1.98, + "learning_rate": 1.342064020266644e-05, + "loss": 0.9592, + "step": 30761 + }, + { + "epoch": 1.98, + "learning_rate": 1.342042551364349e-05, + "loss": 0.9203, + "step": 30762 + }, + { + "epoch": 1.98, + "learning_rate": 1.3420210824620538e-05, + "loss": 0.9985, + "step": 30763 + }, + { + "epoch": 1.98, + "learning_rate": 1.3419996135597589e-05, + "loss": 0.9427, + "step": 30764 + }, + { + "epoch": 1.98, + "learning_rate": 1.3419781446574638e-05, + "loss": 0.973, + "step": 30765 + }, + { + "epoch": 1.98, + "learning_rate": 1.3419566757551687e-05, + "loss": 0.9752, + "step": 30766 + }, + { + "epoch": 1.98, + "learning_rate": 1.3419352068528737e-05, + "loss": 0.9459, + "step": 30767 + }, + { + "epoch": 1.98, + "learning_rate": 1.3419137379505786e-05, + "loss": 0.9904, + "step": 30768 + }, + { + "epoch": 1.98, + "learning_rate": 1.3418922690482837e-05, + "loss": 1.0, + "step": 30769 + }, + { + "epoch": 1.98, + "learning_rate": 1.3418708001459886e-05, + "loss": 0.9676, + "step": 30770 + }, + { + "epoch": 1.98, + "learning_rate": 1.3418493312436935e-05, + "loss": 0.9697, + "step": 30771 + }, + { + "epoch": 1.98, + "learning_rate": 1.3418278623413986e-05, + "loss": 0.9989, + "step": 30772 + }, + { + "epoch": 1.98, + "learning_rate": 1.3418063934391035e-05, + "loss": 0.9602, + "step": 30773 + }, + { + "epoch": 1.98, + "learning_rate": 1.3417849245368087e-05, + "loss": 0.9654, + "step": 30774 + }, + { + "epoch": 1.98, + "learning_rate": 1.3417634556345136e-05, + "loss": 1.0412, + "step": 30775 + }, + { + "epoch": 1.98, + "learning_rate": 1.3417419867322185e-05, + "loss": 1.0204, + "step": 30776 + }, + { + "epoch": 1.98, + "learning_rate": 1.3417205178299236e-05, + "loss": 0.9511, + "step": 30777 + }, + { + "epoch": 1.98, + "learning_rate": 1.3416990489276285e-05, + "loss": 0.9213, + "step": 30778 + }, + { + "epoch": 1.98, + "learning_rate": 1.3416775800253335e-05, + "loss": 0.9762, + "step": 30779 + }, + { + "epoch": 1.98, + "learning_rate": 1.3416561111230384e-05, + "loss": 0.9373, + "step": 30780 + }, + { + "epoch": 1.98, + "learning_rate": 1.3416346422207433e-05, + "loss": 0.9801, + "step": 30781 + }, + { + "epoch": 1.98, + "learning_rate": 1.3416131733184484e-05, + "loss": 1.017, + "step": 30782 + }, + { + "epoch": 1.98, + "learning_rate": 1.3415917044161533e-05, + "loss": 0.9676, + "step": 30783 + }, + { + "epoch": 1.98, + "learning_rate": 1.3415702355138582e-05, + "loss": 0.9771, + "step": 30784 + }, + { + "epoch": 1.98, + "learning_rate": 1.3415487666115632e-05, + "loss": 1.0423, + "step": 30785 + }, + { + "epoch": 1.98, + "learning_rate": 1.3415272977092681e-05, + "loss": 0.9297, + "step": 30786 + }, + { + "epoch": 1.98, + "learning_rate": 1.341505828806973e-05, + "loss": 1.006, + "step": 30787 + }, + { + "epoch": 1.98, + "learning_rate": 1.3414843599046783e-05, + "loss": 1.0202, + "step": 30788 + }, + { + "epoch": 1.98, + "learning_rate": 1.3414628910023833e-05, + "loss": 0.8895, + "step": 30789 + }, + { + "epoch": 1.98, + "learning_rate": 1.3414414221000882e-05, + "loss": 0.8888, + "step": 30790 + }, + { + "epoch": 1.98, + "learning_rate": 1.3414199531977931e-05, + "loss": 0.966, + "step": 30791 + }, + { + "epoch": 1.98, + "learning_rate": 1.3413984842954982e-05, + "loss": 0.9869, + "step": 30792 + }, + { + "epoch": 1.98, + "learning_rate": 1.3413770153932031e-05, + "loss": 0.9814, + "step": 30793 + }, + { + "epoch": 1.98, + "learning_rate": 1.341355546490908e-05, + "loss": 1.0243, + "step": 30794 + }, + { + "epoch": 1.98, + "learning_rate": 1.341334077588613e-05, + "loss": 1.027, + "step": 30795 + }, + { + "epoch": 1.98, + "learning_rate": 1.341312608686318e-05, + "loss": 1.041, + "step": 30796 + }, + { + "epoch": 1.98, + "learning_rate": 1.3412911397840228e-05, + "loss": 1.0877, + "step": 30797 + }, + { + "epoch": 1.98, + "learning_rate": 1.3412696708817279e-05, + "loss": 0.9472, + "step": 30798 + }, + { + "epoch": 1.98, + "learning_rate": 1.3412482019794328e-05, + "loss": 1.0081, + "step": 30799 + }, + { + "epoch": 1.98, + "learning_rate": 1.3412267330771379e-05, + "loss": 0.9224, + "step": 30800 + }, + { + "epoch": 1.98, + "learning_rate": 1.3412052641748428e-05, + "loss": 0.9698, + "step": 30801 + }, + { + "epoch": 1.98, + "learning_rate": 1.3411837952725477e-05, + "loss": 0.8824, + "step": 30802 + }, + { + "epoch": 1.98, + "learning_rate": 1.3411623263702529e-05, + "loss": 1.0265, + "step": 30803 + }, + { + "epoch": 1.98, + "learning_rate": 1.3411408574679578e-05, + "loss": 0.9908, + "step": 30804 + }, + { + "epoch": 1.98, + "learning_rate": 1.3411193885656629e-05, + "loss": 0.9768, + "step": 30805 + }, + { + "epoch": 1.98, + "learning_rate": 1.3410979196633678e-05, + "loss": 0.8995, + "step": 30806 + }, + { + "epoch": 1.98, + "learning_rate": 1.3410764507610727e-05, + "loss": 0.9399, + "step": 30807 + }, + { + "epoch": 1.98, + "learning_rate": 1.3410549818587777e-05, + "loss": 0.8839, + "step": 30808 + }, + { + "epoch": 1.98, + "learning_rate": 1.3410335129564826e-05, + "loss": 0.9445, + "step": 30809 + }, + { + "epoch": 1.98, + "learning_rate": 1.3410120440541877e-05, + "loss": 0.9658, + "step": 30810 + }, + { + "epoch": 1.98, + "learning_rate": 1.3409905751518926e-05, + "loss": 0.9575, + "step": 30811 + }, + { + "epoch": 1.98, + "learning_rate": 1.3409691062495975e-05, + "loss": 0.8612, + "step": 30812 + }, + { + "epoch": 1.98, + "learning_rate": 1.3409476373473025e-05, + "loss": 1.0204, + "step": 30813 + }, + { + "epoch": 1.98, + "learning_rate": 1.3409261684450074e-05, + "loss": 0.9908, + "step": 30814 + }, + { + "epoch": 1.98, + "learning_rate": 1.3409046995427123e-05, + "loss": 0.9719, + "step": 30815 + }, + { + "epoch": 1.98, + "learning_rate": 1.3408832306404174e-05, + "loss": 1.0433, + "step": 30816 + }, + { + "epoch": 1.98, + "learning_rate": 1.3408617617381225e-05, + "loss": 1.0111, + "step": 30817 + }, + { + "epoch": 1.98, + "learning_rate": 1.3408402928358275e-05, + "loss": 0.922, + "step": 30818 + }, + { + "epoch": 1.98, + "learning_rate": 1.3408188239335324e-05, + "loss": 0.9396, + "step": 30819 + }, + { + "epoch": 1.98, + "learning_rate": 1.3407973550312375e-05, + "loss": 1.006, + "step": 30820 + }, + { + "epoch": 1.98, + "learning_rate": 1.3407758861289424e-05, + "loss": 1.0227, + "step": 30821 + }, + { + "epoch": 1.98, + "learning_rate": 1.3407544172266473e-05, + "loss": 0.9633, + "step": 30822 + }, + { + "epoch": 1.98, + "learning_rate": 1.3407329483243524e-05, + "loss": 0.9583, + "step": 30823 + }, + { + "epoch": 1.98, + "learning_rate": 1.3407114794220572e-05, + "loss": 0.9669, + "step": 30824 + }, + { + "epoch": 1.98, + "learning_rate": 1.3406900105197621e-05, + "loss": 0.9231, + "step": 30825 + }, + { + "epoch": 1.98, + "learning_rate": 1.3406685416174672e-05, + "loss": 0.9962, + "step": 30826 + }, + { + "epoch": 1.98, + "learning_rate": 1.3406470727151721e-05, + "loss": 0.9638, + "step": 30827 + }, + { + "epoch": 1.98, + "learning_rate": 1.3406256038128772e-05, + "loss": 0.9136, + "step": 30828 + }, + { + "epoch": 1.98, + "learning_rate": 1.340604134910582e-05, + "loss": 0.9302, + "step": 30829 + }, + { + "epoch": 1.98, + "learning_rate": 1.340582666008287e-05, + "loss": 1.0229, + "step": 30830 + }, + { + "epoch": 1.98, + "learning_rate": 1.340561197105992e-05, + "loss": 0.8973, + "step": 30831 + }, + { + "epoch": 1.98, + "learning_rate": 1.3405397282036971e-05, + "loss": 0.9615, + "step": 30832 + }, + { + "epoch": 1.98, + "learning_rate": 1.3405182593014022e-05, + "loss": 1.0076, + "step": 30833 + }, + { + "epoch": 1.98, + "learning_rate": 1.340496790399107e-05, + "loss": 0.8952, + "step": 30834 + }, + { + "epoch": 1.98, + "learning_rate": 1.340475321496812e-05, + "loss": 0.9545, + "step": 30835 + }, + { + "epoch": 1.98, + "learning_rate": 1.340453852594517e-05, + "loss": 0.9827, + "step": 30836 + }, + { + "epoch": 1.98, + "learning_rate": 1.340432383692222e-05, + "loss": 0.923, + "step": 30837 + }, + { + "epoch": 1.98, + "learning_rate": 1.340410914789927e-05, + "loss": 1.0404, + "step": 30838 + }, + { + "epoch": 1.98, + "learning_rate": 1.3403894458876319e-05, + "loss": 1.0492, + "step": 30839 + }, + { + "epoch": 1.98, + "learning_rate": 1.3403679769853368e-05, + "loss": 1.053, + "step": 30840 + }, + { + "epoch": 1.98, + "learning_rate": 1.3403465080830418e-05, + "loss": 0.9063, + "step": 30841 + }, + { + "epoch": 1.98, + "learning_rate": 1.3403250391807467e-05, + "loss": 0.9931, + "step": 30842 + }, + { + "epoch": 1.98, + "learning_rate": 1.3403035702784516e-05, + "loss": 1.0377, + "step": 30843 + }, + { + "epoch": 1.98, + "learning_rate": 1.3402821013761567e-05, + "loss": 0.9058, + "step": 30844 + }, + { + "epoch": 1.98, + "learning_rate": 1.3402606324738616e-05, + "loss": 1.0336, + "step": 30845 + }, + { + "epoch": 1.98, + "learning_rate": 1.3402391635715668e-05, + "loss": 0.9803, + "step": 30846 + }, + { + "epoch": 1.98, + "learning_rate": 1.3402176946692717e-05, + "loss": 0.9008, + "step": 30847 + }, + { + "epoch": 1.98, + "learning_rate": 1.3401962257669768e-05, + "loss": 0.9208, + "step": 30848 + }, + { + "epoch": 1.98, + "learning_rate": 1.3401747568646817e-05, + "loss": 1.0721, + "step": 30849 + }, + { + "epoch": 1.98, + "learning_rate": 1.3401532879623866e-05, + "loss": 0.9979, + "step": 30850 + }, + { + "epoch": 1.98, + "learning_rate": 1.3401318190600917e-05, + "loss": 0.9187, + "step": 30851 + }, + { + "epoch": 1.98, + "learning_rate": 1.3401103501577966e-05, + "loss": 0.9369, + "step": 30852 + }, + { + "epoch": 1.99, + "learning_rate": 1.3400888812555014e-05, + "loss": 0.9222, + "step": 30853 + }, + { + "epoch": 1.99, + "learning_rate": 1.3400674123532065e-05, + "loss": 0.9008, + "step": 30854 + }, + { + "epoch": 1.99, + "learning_rate": 1.3400459434509114e-05, + "loss": 0.9649, + "step": 30855 + }, + { + "epoch": 1.99, + "learning_rate": 1.3400244745486165e-05, + "loss": 0.94, + "step": 30856 + }, + { + "epoch": 1.99, + "learning_rate": 1.3400030056463214e-05, + "loss": 0.9459, + "step": 30857 + }, + { + "epoch": 1.99, + "learning_rate": 1.3399815367440263e-05, + "loss": 0.9756, + "step": 30858 + }, + { + "epoch": 1.99, + "learning_rate": 1.3399600678417313e-05, + "loss": 0.9768, + "step": 30859 + }, + { + "epoch": 1.99, + "learning_rate": 1.3399385989394362e-05, + "loss": 1.0012, + "step": 30860 + }, + { + "epoch": 1.99, + "learning_rate": 1.3399171300371415e-05, + "loss": 0.9602, + "step": 30861 + }, + { + "epoch": 1.99, + "learning_rate": 1.3398956611348464e-05, + "loss": 1.0252, + "step": 30862 + }, + { + "epoch": 1.99, + "learning_rate": 1.3398741922325513e-05, + "loss": 0.9581, + "step": 30863 + }, + { + "epoch": 1.99, + "learning_rate": 1.3398527233302563e-05, + "loss": 0.912, + "step": 30864 + }, + { + "epoch": 1.99, + "learning_rate": 1.3398312544279612e-05, + "loss": 0.9533, + "step": 30865 + }, + { + "epoch": 1.99, + "learning_rate": 1.3398097855256663e-05, + "loss": 1.0026, + "step": 30866 + }, + { + "epoch": 1.99, + "learning_rate": 1.3397883166233712e-05, + "loss": 0.9236, + "step": 30867 + }, + { + "epoch": 1.99, + "learning_rate": 1.339766847721076e-05, + "loss": 1.0346, + "step": 30868 + }, + { + "epoch": 1.99, + "learning_rate": 1.3397453788187811e-05, + "loss": 0.985, + "step": 30869 + }, + { + "epoch": 1.99, + "learning_rate": 1.339723909916486e-05, + "loss": 0.9617, + "step": 30870 + }, + { + "epoch": 1.99, + "learning_rate": 1.339702441014191e-05, + "loss": 0.9544, + "step": 30871 + }, + { + "epoch": 1.99, + "learning_rate": 1.339680972111896e-05, + "loss": 0.9964, + "step": 30872 + }, + { + "epoch": 1.99, + "learning_rate": 1.3396595032096009e-05, + "loss": 0.9161, + "step": 30873 + }, + { + "epoch": 1.99, + "learning_rate": 1.3396380343073058e-05, + "loss": 1.0666, + "step": 30874 + }, + { + "epoch": 1.99, + "learning_rate": 1.339616565405011e-05, + "loss": 0.9528, + "step": 30875 + }, + { + "epoch": 1.99, + "learning_rate": 1.3395950965027161e-05, + "loss": 0.9611, + "step": 30876 + }, + { + "epoch": 1.99, + "learning_rate": 1.339573627600421e-05, + "loss": 1.0087, + "step": 30877 + }, + { + "epoch": 1.99, + "learning_rate": 1.3395521586981259e-05, + "loss": 0.954, + "step": 30878 + }, + { + "epoch": 1.99, + "learning_rate": 1.339530689795831e-05, + "loss": 0.9154, + "step": 30879 + }, + { + "epoch": 1.99, + "learning_rate": 1.3395092208935359e-05, + "loss": 0.9746, + "step": 30880 + }, + { + "epoch": 1.99, + "learning_rate": 1.3394877519912408e-05, + "loss": 0.9528, + "step": 30881 + }, + { + "epoch": 1.99, + "learning_rate": 1.3394662830889458e-05, + "loss": 0.8955, + "step": 30882 + }, + { + "epoch": 1.99, + "learning_rate": 1.3394448141866507e-05, + "loss": 1.0135, + "step": 30883 + }, + { + "epoch": 1.99, + "learning_rate": 1.3394233452843556e-05, + "loss": 0.9792, + "step": 30884 + }, + { + "epoch": 1.99, + "learning_rate": 1.3394018763820607e-05, + "loss": 0.9392, + "step": 30885 + }, + { + "epoch": 1.99, + "learning_rate": 1.3393804074797656e-05, + "loss": 1.0115, + "step": 30886 + }, + { + "epoch": 1.99, + "learning_rate": 1.3393589385774706e-05, + "loss": 0.9626, + "step": 30887 + }, + { + "epoch": 1.99, + "learning_rate": 1.3393374696751755e-05, + "loss": 0.9851, + "step": 30888 + }, + { + "epoch": 1.99, + "learning_rate": 1.3393160007728804e-05, + "loss": 0.953, + "step": 30889 + }, + { + "epoch": 1.99, + "learning_rate": 1.3392945318705857e-05, + "loss": 0.9902, + "step": 30890 + }, + { + "epoch": 1.99, + "learning_rate": 1.3392730629682906e-05, + "loss": 1.0106, + "step": 30891 + }, + { + "epoch": 1.99, + "learning_rate": 1.3392515940659956e-05, + "loss": 0.9373, + "step": 30892 + }, + { + "epoch": 1.99, + "learning_rate": 1.3392301251637005e-05, + "loss": 0.8991, + "step": 30893 + }, + { + "epoch": 1.99, + "learning_rate": 1.3392086562614054e-05, + "loss": 0.9831, + "step": 30894 + }, + { + "epoch": 1.99, + "learning_rate": 1.3391871873591105e-05, + "loss": 0.9716, + "step": 30895 + }, + { + "epoch": 1.99, + "learning_rate": 1.3391657184568154e-05, + "loss": 0.8821, + "step": 30896 + }, + { + "epoch": 1.99, + "learning_rate": 1.3391442495545205e-05, + "loss": 1.0517, + "step": 30897 + }, + { + "epoch": 1.99, + "learning_rate": 1.3391227806522253e-05, + "loss": 1.0409, + "step": 30898 + }, + { + "epoch": 1.99, + "learning_rate": 1.3391013117499302e-05, + "loss": 0.9575, + "step": 30899 + }, + { + "epoch": 1.99, + "learning_rate": 1.3390798428476353e-05, + "loss": 0.9009, + "step": 30900 + }, + { + "epoch": 1.99, + "learning_rate": 1.3390583739453402e-05, + "loss": 1.0371, + "step": 30901 + }, + { + "epoch": 1.99, + "learning_rate": 1.3390369050430451e-05, + "loss": 0.941, + "step": 30902 + }, + { + "epoch": 1.99, + "learning_rate": 1.3390154361407502e-05, + "loss": 0.9236, + "step": 30903 + }, + { + "epoch": 1.99, + "learning_rate": 1.3389939672384552e-05, + "loss": 0.97, + "step": 30904 + }, + { + "epoch": 1.99, + "learning_rate": 1.3389724983361603e-05, + "loss": 1.0084, + "step": 30905 + }, + { + "epoch": 1.99, + "learning_rate": 1.3389510294338652e-05, + "loss": 0.9754, + "step": 30906 + }, + { + "epoch": 1.99, + "learning_rate": 1.3389295605315703e-05, + "loss": 1.0025, + "step": 30907 + }, + { + "epoch": 1.99, + "learning_rate": 1.3389080916292752e-05, + "loss": 0.8968, + "step": 30908 + }, + { + "epoch": 1.99, + "learning_rate": 1.33888662272698e-05, + "loss": 0.9916, + "step": 30909 + }, + { + "epoch": 1.99, + "learning_rate": 1.3388651538246851e-05, + "loss": 1.0008, + "step": 30910 + }, + { + "epoch": 1.99, + "learning_rate": 1.33884368492239e-05, + "loss": 0.9316, + "step": 30911 + }, + { + "epoch": 1.99, + "learning_rate": 1.338822216020095e-05, + "loss": 0.8883, + "step": 30912 + }, + { + "epoch": 1.99, + "learning_rate": 1.3388007471178e-05, + "loss": 1.0078, + "step": 30913 + }, + { + "epoch": 1.99, + "learning_rate": 1.3387792782155049e-05, + "loss": 0.9752, + "step": 30914 + }, + { + "epoch": 1.99, + "learning_rate": 1.33875780931321e-05, + "loss": 0.9645, + "step": 30915 + }, + { + "epoch": 1.99, + "learning_rate": 1.3387363404109148e-05, + "loss": 0.9545, + "step": 30916 + }, + { + "epoch": 1.99, + "learning_rate": 1.3387148715086197e-05, + "loss": 0.9126, + "step": 30917 + }, + { + "epoch": 1.99, + "learning_rate": 1.3386934026063248e-05, + "loss": 0.9869, + "step": 30918 + }, + { + "epoch": 1.99, + "learning_rate": 1.3386719337040299e-05, + "loss": 0.9142, + "step": 30919 + }, + { + "epoch": 1.99, + "learning_rate": 1.338650464801735e-05, + "loss": 0.9756, + "step": 30920 + }, + { + "epoch": 1.99, + "learning_rate": 1.3386289958994398e-05, + "loss": 0.9484, + "step": 30921 + }, + { + "epoch": 1.99, + "learning_rate": 1.3386075269971447e-05, + "loss": 0.997, + "step": 30922 + }, + { + "epoch": 1.99, + "learning_rate": 1.3385860580948498e-05, + "loss": 1.0033, + "step": 30923 + }, + { + "epoch": 1.99, + "learning_rate": 1.3385645891925547e-05, + "loss": 0.917, + "step": 30924 + }, + { + "epoch": 1.99, + "learning_rate": 1.3385431202902598e-05, + "loss": 1.0488, + "step": 30925 + }, + { + "epoch": 1.99, + "learning_rate": 1.3385216513879647e-05, + "loss": 0.9315, + "step": 30926 + }, + { + "epoch": 1.99, + "learning_rate": 1.3385001824856695e-05, + "loss": 0.8747, + "step": 30927 + }, + { + "epoch": 1.99, + "learning_rate": 1.3384787135833746e-05, + "loss": 1.0174, + "step": 30928 + }, + { + "epoch": 1.99, + "learning_rate": 1.3384572446810795e-05, + "loss": 0.9473, + "step": 30929 + }, + { + "epoch": 1.99, + "learning_rate": 1.3384357757787844e-05, + "loss": 0.9046, + "step": 30930 + }, + { + "epoch": 1.99, + "learning_rate": 1.3384143068764895e-05, + "loss": 0.9458, + "step": 30931 + }, + { + "epoch": 1.99, + "learning_rate": 1.3383928379741944e-05, + "loss": 1.0311, + "step": 30932 + }, + { + "epoch": 1.99, + "learning_rate": 1.3383713690718996e-05, + "loss": 0.9447, + "step": 30933 + }, + { + "epoch": 1.99, + "learning_rate": 1.3383499001696045e-05, + "loss": 0.9568, + "step": 30934 + }, + { + "epoch": 1.99, + "learning_rate": 1.3383284312673096e-05, + "loss": 0.9705, + "step": 30935 + }, + { + "epoch": 1.99, + "learning_rate": 1.3383069623650145e-05, + "loss": 0.971, + "step": 30936 + }, + { + "epoch": 1.99, + "learning_rate": 1.3382854934627194e-05, + "loss": 0.9654, + "step": 30937 + }, + { + "epoch": 1.99, + "learning_rate": 1.3382640245604244e-05, + "loss": 0.8944, + "step": 30938 + }, + { + "epoch": 1.99, + "learning_rate": 1.3382425556581293e-05, + "loss": 1.0279, + "step": 30939 + }, + { + "epoch": 1.99, + "learning_rate": 1.3382210867558342e-05, + "loss": 0.9733, + "step": 30940 + }, + { + "epoch": 1.99, + "learning_rate": 1.3381996178535393e-05, + "loss": 0.9466, + "step": 30941 + }, + { + "epoch": 1.99, + "learning_rate": 1.3381781489512442e-05, + "loss": 0.9566, + "step": 30942 + }, + { + "epoch": 1.99, + "learning_rate": 1.3381566800489492e-05, + "loss": 1.0071, + "step": 30943 + }, + { + "epoch": 1.99, + "learning_rate": 1.3381352111466541e-05, + "loss": 1.0264, + "step": 30944 + }, + { + "epoch": 1.99, + "learning_rate": 1.338113742244359e-05, + "loss": 0.9181, + "step": 30945 + }, + { + "epoch": 1.99, + "learning_rate": 1.3380922733420641e-05, + "loss": 0.9301, + "step": 30946 + }, + { + "epoch": 1.99, + "learning_rate": 1.338070804439769e-05, + "loss": 1.0332, + "step": 30947 + }, + { + "epoch": 1.99, + "learning_rate": 1.3380493355374742e-05, + "loss": 0.9938, + "step": 30948 + }, + { + "epoch": 1.99, + "learning_rate": 1.3380278666351791e-05, + "loss": 1.0742, + "step": 30949 + }, + { + "epoch": 1.99, + "learning_rate": 1.338006397732884e-05, + "loss": 0.9488, + "step": 30950 + }, + { + "epoch": 1.99, + "learning_rate": 1.3379849288305891e-05, + "loss": 1.0172, + "step": 30951 + }, + { + "epoch": 1.99, + "learning_rate": 1.337963459928294e-05, + "loss": 0.9176, + "step": 30952 + }, + { + "epoch": 1.99, + "learning_rate": 1.337941991025999e-05, + "loss": 0.9552, + "step": 30953 + }, + { + "epoch": 1.99, + "learning_rate": 1.337920522123704e-05, + "loss": 0.9695, + "step": 30954 + }, + { + "epoch": 1.99, + "learning_rate": 1.3378990532214089e-05, + "loss": 0.9793, + "step": 30955 + }, + { + "epoch": 1.99, + "learning_rate": 1.337877584319114e-05, + "loss": 0.9476, + "step": 30956 + }, + { + "epoch": 1.99, + "learning_rate": 1.3378561154168188e-05, + "loss": 1.0059, + "step": 30957 + }, + { + "epoch": 1.99, + "learning_rate": 1.3378346465145237e-05, + "loss": 0.9492, + "step": 30958 + }, + { + "epoch": 1.99, + "learning_rate": 1.3378131776122288e-05, + "loss": 0.9227, + "step": 30959 + }, + { + "epoch": 1.99, + "learning_rate": 1.3377917087099337e-05, + "loss": 0.9359, + "step": 30960 + }, + { + "epoch": 1.99, + "learning_rate": 1.3377702398076386e-05, + "loss": 0.9526, + "step": 30961 + }, + { + "epoch": 1.99, + "learning_rate": 1.3377487709053438e-05, + "loss": 0.9746, + "step": 30962 + }, + { + "epoch": 1.99, + "learning_rate": 1.3377273020030489e-05, + "loss": 0.943, + "step": 30963 + }, + { + "epoch": 1.99, + "learning_rate": 1.3377058331007538e-05, + "loss": 1.0205, + "step": 30964 + }, + { + "epoch": 1.99, + "learning_rate": 1.3376843641984587e-05, + "loss": 0.9961, + "step": 30965 + }, + { + "epoch": 1.99, + "learning_rate": 1.3376628952961637e-05, + "loss": 0.8981, + "step": 30966 + }, + { + "epoch": 1.99, + "learning_rate": 1.3376414263938686e-05, + "loss": 0.9001, + "step": 30967 + }, + { + "epoch": 1.99, + "learning_rate": 1.3376199574915735e-05, + "loss": 1.0108, + "step": 30968 + }, + { + "epoch": 1.99, + "learning_rate": 1.3375984885892786e-05, + "loss": 0.9799, + "step": 30969 + }, + { + "epoch": 1.99, + "learning_rate": 1.3375770196869835e-05, + "loss": 0.998, + "step": 30970 + }, + { + "epoch": 1.99, + "learning_rate": 1.3375555507846884e-05, + "loss": 0.8871, + "step": 30971 + }, + { + "epoch": 1.99, + "learning_rate": 1.3375340818823934e-05, + "loss": 0.9665, + "step": 30972 + }, + { + "epoch": 1.99, + "learning_rate": 1.3375126129800983e-05, + "loss": 0.9906, + "step": 30973 + }, + { + "epoch": 1.99, + "learning_rate": 1.3374911440778034e-05, + "loss": 0.9833, + "step": 30974 + }, + { + "epoch": 1.99, + "learning_rate": 1.3374696751755083e-05, + "loss": 0.9948, + "step": 30975 + }, + { + "epoch": 1.99, + "learning_rate": 1.3374482062732132e-05, + "loss": 0.9543, + "step": 30976 + }, + { + "epoch": 1.99, + "learning_rate": 1.3374267373709184e-05, + "loss": 1.0268, + "step": 30977 + }, + { + "epoch": 1.99, + "learning_rate": 1.3374052684686233e-05, + "loss": 0.9005, + "step": 30978 + }, + { + "epoch": 1.99, + "learning_rate": 1.3373837995663284e-05, + "loss": 1.0148, + "step": 30979 + }, + { + "epoch": 1.99, + "learning_rate": 1.3373623306640333e-05, + "loss": 0.899, + "step": 30980 + }, + { + "epoch": 1.99, + "learning_rate": 1.3373408617617382e-05, + "loss": 0.9271, + "step": 30981 + }, + { + "epoch": 1.99, + "learning_rate": 1.3373193928594433e-05, + "loss": 0.9525, + "step": 30982 + }, + { + "epoch": 1.99, + "learning_rate": 1.3372979239571482e-05, + "loss": 1.0468, + "step": 30983 + }, + { + "epoch": 1.99, + "learning_rate": 1.3372764550548532e-05, + "loss": 1.0316, + "step": 30984 + }, + { + "epoch": 1.99, + "learning_rate": 1.3372549861525581e-05, + "loss": 0.9918, + "step": 30985 + }, + { + "epoch": 1.99, + "learning_rate": 1.337233517250263e-05, + "loss": 0.9705, + "step": 30986 + }, + { + "epoch": 1.99, + "learning_rate": 1.337212048347968e-05, + "loss": 0.9635, + "step": 30987 + }, + { + "epoch": 1.99, + "learning_rate": 1.337190579445673e-05, + "loss": 0.9708, + "step": 30988 + }, + { + "epoch": 1.99, + "learning_rate": 1.3371691105433779e-05, + "loss": 0.9399, + "step": 30989 + }, + { + "epoch": 1.99, + "learning_rate": 1.337147641641083e-05, + "loss": 1.0204, + "step": 30990 + }, + { + "epoch": 1.99, + "learning_rate": 1.3371261727387878e-05, + "loss": 1.0357, + "step": 30991 + }, + { + "epoch": 1.99, + "learning_rate": 1.337104703836493e-05, + "loss": 0.9961, + "step": 30992 + }, + { + "epoch": 1.99, + "learning_rate": 1.337083234934198e-05, + "loss": 1.0425, + "step": 30993 + }, + { + "epoch": 1.99, + "learning_rate": 1.337061766031903e-05, + "loss": 0.9717, + "step": 30994 + }, + { + "epoch": 1.99, + "learning_rate": 1.337040297129608e-05, + "loss": 0.9194, + "step": 30995 + }, + { + "epoch": 1.99, + "learning_rate": 1.3370188282273128e-05, + "loss": 0.9435, + "step": 30996 + }, + { + "epoch": 1.99, + "learning_rate": 1.3369973593250179e-05, + "loss": 0.9779, + "step": 30997 + }, + { + "epoch": 1.99, + "learning_rate": 1.3369758904227228e-05, + "loss": 0.9798, + "step": 30998 + }, + { + "epoch": 1.99, + "learning_rate": 1.3369544215204277e-05, + "loss": 1.0189, + "step": 30999 + }, + { + "epoch": 1.99, + "learning_rate": 1.3369329526181328e-05, + "loss": 0.9662, + "step": 31000 + }, + { + "epoch": 1.99, + "eval_laion/cider": 0.5624795460370802, + "eval_laion/meteor": 0.39488966510056284, + "eval_laion/pred_num_words": 37.25, + "eval_laion/sacrebleu": 16.13876177460903, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0349810123443604, + "eval_runtime": 121.2915, + "eval_samples_per_second": 4.114, + "eval_steps_per_second": 0.264, + "step": 31000 + }, + { + "epoch": 1.99, + "learning_rate": 1.3369114837158376e-05, + "loss": 0.9576, + "step": 31001 + }, + { + "epoch": 1.99, + "learning_rate": 1.3368900148135427e-05, + "loss": 0.8843, + "step": 31002 + }, + { + "epoch": 1.99, + "learning_rate": 1.3368685459112476e-05, + "loss": 0.9932, + "step": 31003 + }, + { + "epoch": 1.99, + "learning_rate": 1.3368470770089525e-05, + "loss": 1.018, + "step": 31004 + }, + { + "epoch": 1.99, + "learning_rate": 1.3368256081066576e-05, + "loss": 0.8945, + "step": 31005 + }, + { + "epoch": 1.99, + "learning_rate": 1.3368041392043626e-05, + "loss": 0.9853, + "step": 31006 + }, + { + "epoch": 1.99, + "learning_rate": 1.3367826703020677e-05, + "loss": 0.8682, + "step": 31007 + }, + { + "epoch": 1.99, + "learning_rate": 1.3367612013997726e-05, + "loss": 0.9305, + "step": 31008 + }, + { + "epoch": 2.0, + "learning_rate": 1.3367397324974775e-05, + "loss": 0.9973, + "step": 31009 + }, + { + "epoch": 2.0, + "learning_rate": 1.3367182635951826e-05, + "loss": 0.9687, + "step": 31010 + }, + { + "epoch": 2.0, + "learning_rate": 1.3366967946928875e-05, + "loss": 1.0134, + "step": 31011 + }, + { + "epoch": 2.0, + "learning_rate": 1.3366753257905925e-05, + "loss": 0.9682, + "step": 31012 + }, + { + "epoch": 2.0, + "learning_rate": 1.3366538568882974e-05, + "loss": 1.0616, + "step": 31013 + }, + { + "epoch": 2.0, + "learning_rate": 1.3366323879860023e-05, + "loss": 0.9354, + "step": 31014 + }, + { + "epoch": 2.0, + "learning_rate": 1.3366109190837074e-05, + "loss": 0.9408, + "step": 31015 + }, + { + "epoch": 2.0, + "learning_rate": 1.3365894501814123e-05, + "loss": 0.9778, + "step": 31016 + }, + { + "epoch": 2.0, + "learning_rate": 1.3365679812791172e-05, + "loss": 0.8929, + "step": 31017 + }, + { + "epoch": 2.0, + "learning_rate": 1.3365465123768222e-05, + "loss": 1.0111, + "step": 31018 + }, + { + "epoch": 2.0, + "learning_rate": 1.3365250434745271e-05, + "loss": 1.0222, + "step": 31019 + }, + { + "epoch": 2.0, + "learning_rate": 1.336503574572232e-05, + "loss": 1.015, + "step": 31020 + }, + { + "epoch": 2.0, + "learning_rate": 1.3364821056699373e-05, + "loss": 0.8928, + "step": 31021 + }, + { + "epoch": 2.0, + "learning_rate": 1.3364606367676423e-05, + "loss": 0.9873, + "step": 31022 + }, + { + "epoch": 2.0, + "learning_rate": 1.3364391678653472e-05, + "loss": 1.0343, + "step": 31023 + }, + { + "epoch": 2.0, + "learning_rate": 1.3364176989630521e-05, + "loss": 0.9703, + "step": 31024 + }, + { + "epoch": 2.0, + "learning_rate": 1.3363962300607572e-05, + "loss": 0.9419, + "step": 31025 + }, + { + "epoch": 2.0, + "learning_rate": 1.3363747611584621e-05, + "loss": 1.0229, + "step": 31026 + }, + { + "epoch": 2.0, + "learning_rate": 1.336353292256167e-05, + "loss": 0.952, + "step": 31027 + }, + { + "epoch": 2.0, + "learning_rate": 1.336331823353872e-05, + "loss": 0.978, + "step": 31028 + }, + { + "epoch": 2.0, + "learning_rate": 1.336310354451577e-05, + "loss": 0.9533, + "step": 31029 + }, + { + "epoch": 2.0, + "learning_rate": 1.3362888855492818e-05, + "loss": 0.9069, + "step": 31030 + }, + { + "epoch": 2.0, + "learning_rate": 1.336267416646987e-05, + "loss": 0.994, + "step": 31031 + }, + { + "epoch": 2.0, + "learning_rate": 1.3362459477446918e-05, + "loss": 0.9786, + "step": 31032 + }, + { + "epoch": 2.0, + "learning_rate": 1.3362244788423969e-05, + "loss": 0.9413, + "step": 31033 + }, + { + "epoch": 2.0, + "learning_rate": 1.3362030099401018e-05, + "loss": 0.9791, + "step": 31034 + }, + { + "epoch": 2.0, + "learning_rate": 1.336181541037807e-05, + "loss": 0.9483, + "step": 31035 + }, + { + "epoch": 2.0, + "learning_rate": 1.3361600721355119e-05, + "loss": 0.9998, + "step": 31036 + }, + { + "epoch": 2.0, + "learning_rate": 1.3361386032332168e-05, + "loss": 1.0332, + "step": 31037 + }, + { + "epoch": 2.0, + "learning_rate": 1.3361171343309219e-05, + "loss": 1.0259, + "step": 31038 + }, + { + "epoch": 2.0, + "learning_rate": 1.3360956654286268e-05, + "loss": 0.9771, + "step": 31039 + }, + { + "epoch": 2.0, + "learning_rate": 1.3360741965263317e-05, + "loss": 1.0185, + "step": 31040 + }, + { + "epoch": 2.0, + "learning_rate": 1.3360527276240367e-05, + "loss": 0.9196, + "step": 31041 + }, + { + "epoch": 2.0, + "learning_rate": 1.3360312587217416e-05, + "loss": 0.9053, + "step": 31042 + }, + { + "epoch": 2.0, + "learning_rate": 1.3360097898194467e-05, + "loss": 0.9648, + "step": 31043 + }, + { + "epoch": 2.0, + "learning_rate": 1.3359883209171516e-05, + "loss": 0.9386, + "step": 31044 + }, + { + "epoch": 2.0, + "learning_rate": 1.3359668520148565e-05, + "loss": 1.0041, + "step": 31045 + }, + { + "epoch": 2.0, + "learning_rate": 1.3359453831125615e-05, + "loss": 0.9513, + "step": 31046 + }, + { + "epoch": 2.0, + "learning_rate": 1.3359239142102664e-05, + "loss": 0.9622, + "step": 31047 + }, + { + "epoch": 2.0, + "learning_rate": 1.3359024453079713e-05, + "loss": 0.9452, + "step": 31048 + }, + { + "epoch": 2.0, + "learning_rate": 1.3358809764056764e-05, + "loss": 1.0271, + "step": 31049 + }, + { + "epoch": 2.0, + "learning_rate": 1.3358595075033816e-05, + "loss": 0.9928, + "step": 31050 + }, + { + "epoch": 2.0, + "learning_rate": 1.3358380386010865e-05, + "loss": 0.9337, + "step": 31051 + }, + { + "epoch": 2.0, + "learning_rate": 1.3358165696987914e-05, + "loss": 1.018, + "step": 31052 + }, + { + "epoch": 2.0, + "learning_rate": 1.3357951007964965e-05, + "loss": 0.9656, + "step": 31053 + }, + { + "epoch": 2.0, + "learning_rate": 1.3357736318942014e-05, + "loss": 0.9547, + "step": 31054 + }, + { + "epoch": 2.0, + "learning_rate": 1.3357521629919063e-05, + "loss": 0.9289, + "step": 31055 + }, + { + "epoch": 2.0, + "learning_rate": 1.3357306940896114e-05, + "loss": 0.9494, + "step": 31056 + }, + { + "epoch": 2.0, + "learning_rate": 1.3357092251873163e-05, + "loss": 0.9765, + "step": 31057 + }, + { + "epoch": 2.0, + "learning_rate": 1.3356877562850212e-05, + "loss": 0.9427, + "step": 31058 + }, + { + "epoch": 2.0, + "learning_rate": 1.3356662873827262e-05, + "loss": 0.9081, + "step": 31059 + }, + { + "epoch": 2.0, + "learning_rate": 1.3356448184804311e-05, + "loss": 0.9864, + "step": 31060 + }, + { + "epoch": 2.0, + "learning_rate": 1.3356233495781362e-05, + "loss": 0.9924, + "step": 31061 + }, + { + "epoch": 2.0, + "learning_rate": 1.335601880675841e-05, + "loss": 0.9676, + "step": 31062 + }, + { + "epoch": 2.0, + "learning_rate": 1.335580411773546e-05, + "loss": 0.9646, + "step": 31063 + }, + { + "epoch": 2.0, + "learning_rate": 1.3355589428712512e-05, + "loss": 0.962, + "step": 31064 + }, + { + "epoch": 2.0, + "learning_rate": 1.3355374739689561e-05, + "loss": 0.9627, + "step": 31065 + }, + { + "epoch": 2.0, + "learning_rate": 1.3355160050666612e-05, + "loss": 0.9647, + "step": 31066 + }, + { + "epoch": 2.0, + "learning_rate": 1.335494536164366e-05, + "loss": 0.9131, + "step": 31067 + }, + { + "epoch": 2.0, + "learning_rate": 1.335473067262071e-05, + "loss": 0.9783, + "step": 31068 + }, + { + "epoch": 2.0, + "learning_rate": 1.335451598359776e-05, + "loss": 0.8678, + "step": 31069 + }, + { + "epoch": 2.0, + "learning_rate": 1.335430129457481e-05, + "loss": 0.9529, + "step": 31070 + }, + { + "epoch": 2.0, + "learning_rate": 1.335408660555186e-05, + "loss": 0.9841, + "step": 31071 + }, + { + "epoch": 2.0, + "learning_rate": 1.3353871916528909e-05, + "loss": 0.9775, + "step": 31072 + }, + { + "epoch": 2.0, + "learning_rate": 1.3353657227505958e-05, + "loss": 0.942, + "step": 31073 + }, + { + "epoch": 2.0, + "learning_rate": 1.3353442538483009e-05, + "loss": 0.9394, + "step": 31074 + }, + { + "epoch": 2.0, + "learning_rate": 1.3353227849460057e-05, + "loss": 0.9863, + "step": 31075 + }, + { + "epoch": 2.0, + "learning_rate": 1.3353013160437106e-05, + "loss": 0.9195, + "step": 31076 + }, + { + "epoch": 2.0, + "learning_rate": 1.3352798471414157e-05, + "loss": 0.9772, + "step": 31077 + }, + { + "epoch": 2.0, + "learning_rate": 1.3352583782391206e-05, + "loss": 0.9609, + "step": 31078 + }, + { + "epoch": 2.0, + "learning_rate": 1.3352369093368258e-05, + "loss": 0.9476, + "step": 31079 + }, + { + "epoch": 2.0, + "learning_rate": 1.3352154404345307e-05, + "loss": 0.9373, + "step": 31080 + }, + { + "epoch": 2.0, + "learning_rate": 1.3351939715322358e-05, + "loss": 1.0398, + "step": 31081 + }, + { + "epoch": 2.0, + "learning_rate": 1.3351725026299407e-05, + "loss": 0.968, + "step": 31082 + }, + { + "epoch": 2.0, + "learning_rate": 1.3351510337276456e-05, + "loss": 0.9808, + "step": 31083 + }, + { + "epoch": 2.0, + "learning_rate": 1.3351295648253507e-05, + "loss": 0.9501, + "step": 31084 + }, + { + "epoch": 2.0, + "learning_rate": 1.3351080959230556e-05, + "loss": 1.0002, + "step": 31085 + }, + { + "epoch": 2.0, + "learning_rate": 1.3350866270207605e-05, + "loss": 0.9578, + "step": 31086 + }, + { + "epoch": 2.0, + "learning_rate": 1.3350651581184655e-05, + "loss": 0.9118, + "step": 31087 + }, + { + "epoch": 2.0, + "learning_rate": 1.3350436892161704e-05, + "loss": 0.9136, + "step": 31088 + }, + { + "epoch": 2.0, + "learning_rate": 1.3350222203138755e-05, + "loss": 0.8972, + "step": 31089 + }, + { + "epoch": 2.0, + "learning_rate": 1.3350007514115804e-05, + "loss": 0.9843, + "step": 31090 + }, + { + "epoch": 2.0, + "learning_rate": 1.3349792825092853e-05, + "loss": 0.9102, + "step": 31091 + }, + { + "epoch": 2.0, + "learning_rate": 1.3349578136069903e-05, + "loss": 0.9171, + "step": 31092 + }, + { + "epoch": 2.0, + "learning_rate": 1.3349363447046954e-05, + "loss": 0.9465, + "step": 31093 + }, + { + "epoch": 2.0, + "learning_rate": 1.3349148758024005e-05, + "loss": 0.8647, + "step": 31094 + }, + { + "epoch": 2.0, + "learning_rate": 1.3348934069001054e-05, + "loss": 0.9176, + "step": 31095 + }, + { + "epoch": 2.0, + "learning_rate": 1.3348719379978103e-05, + "loss": 0.9154, + "step": 31096 + }, + { + "epoch": 2.0, + "learning_rate": 1.3348504690955153e-05, + "loss": 0.9246, + "step": 31097 + }, + { + "epoch": 2.0, + "learning_rate": 1.3348290001932202e-05, + "loss": 0.9086, + "step": 31098 + }, + { + "epoch": 2.0, + "learning_rate": 1.3348075312909253e-05, + "loss": 0.8918, + "step": 31099 + }, + { + "epoch": 2.0, + "learning_rate": 1.3347860623886302e-05, + "loss": 0.9303, + "step": 31100 + }, + { + "epoch": 2.0, + "learning_rate": 1.3347645934863351e-05, + "loss": 0.8325, + "step": 31101 + }, + { + "epoch": 2.0, + "learning_rate": 1.3347431245840402e-05, + "loss": 0.9047, + "step": 31102 + }, + { + "epoch": 2.0, + "learning_rate": 1.334721655681745e-05, + "loss": 0.8892, + "step": 31103 + }, + { + "epoch": 2.0, + "learning_rate": 1.33470018677945e-05, + "loss": 0.9348, + "step": 31104 + }, + { + "epoch": 2.0, + "learning_rate": 1.334678717877155e-05, + "loss": 0.8346, + "step": 31105 + }, + { + "epoch": 2.0, + "learning_rate": 1.3346572489748599e-05, + "loss": 0.8818, + "step": 31106 + }, + { + "epoch": 2.0, + "learning_rate": 1.3346357800725648e-05, + "loss": 0.8696, + "step": 31107 + }, + { + "epoch": 2.0, + "learning_rate": 1.33461431117027e-05, + "loss": 0.8922, + "step": 31108 + }, + { + "epoch": 2.0, + "learning_rate": 1.3345928422679751e-05, + "loss": 0.8561, + "step": 31109 + }, + { + "epoch": 2.0, + "learning_rate": 1.33457137336568e-05, + "loss": 0.8815, + "step": 31110 + }, + { + "epoch": 2.0, + "learning_rate": 1.3345499044633849e-05, + "loss": 0.9165, + "step": 31111 + }, + { + "epoch": 2.0, + "learning_rate": 1.33452843556109e-05, + "loss": 0.8618, + "step": 31112 + }, + { + "epoch": 2.0, + "learning_rate": 1.3345069666587949e-05, + "loss": 0.8952, + "step": 31113 + }, + { + "epoch": 2.0, + "learning_rate": 1.3344854977564998e-05, + "loss": 0.8522, + "step": 31114 + }, + { + "epoch": 2.0, + "learning_rate": 1.3344640288542048e-05, + "loss": 0.8998, + "step": 31115 + }, + { + "epoch": 2.0, + "learning_rate": 1.3344425599519097e-05, + "loss": 0.9474, + "step": 31116 + }, + { + "epoch": 2.0, + "learning_rate": 1.3344210910496146e-05, + "loss": 0.9063, + "step": 31117 + }, + { + "epoch": 2.0, + "learning_rate": 1.3343996221473197e-05, + "loss": 0.845, + "step": 31118 + }, + { + "epoch": 2.0, + "learning_rate": 1.3343781532450246e-05, + "loss": 0.8954, + "step": 31119 + }, + { + "epoch": 2.0, + "learning_rate": 1.3343566843427296e-05, + "loss": 0.9214, + "step": 31120 + }, + { + "epoch": 2.0, + "learning_rate": 1.3343352154404345e-05, + "loss": 0.9154, + "step": 31121 + }, + { + "epoch": 2.0, + "learning_rate": 1.3343137465381398e-05, + "loss": 0.8544, + "step": 31122 + }, + { + "epoch": 2.0, + "learning_rate": 1.3342922776358447e-05, + "loss": 0.8902, + "step": 31123 + }, + { + "epoch": 2.0, + "learning_rate": 1.3342708087335496e-05, + "loss": 0.8892, + "step": 31124 + }, + { + "epoch": 2.0, + "learning_rate": 1.3342493398312546e-05, + "loss": 0.8377, + "step": 31125 + }, + { + "epoch": 2.0, + "learning_rate": 1.3342278709289595e-05, + "loss": 0.8757, + "step": 31126 + }, + { + "epoch": 2.0, + "learning_rate": 1.3342064020266644e-05, + "loss": 0.8886, + "step": 31127 + }, + { + "epoch": 2.0, + "learning_rate": 1.3341849331243695e-05, + "loss": 0.9435, + "step": 31128 + }, + { + "epoch": 2.0, + "learning_rate": 1.3341634642220744e-05, + "loss": 0.9007, + "step": 31129 + }, + { + "epoch": 2.0, + "learning_rate": 1.3341419953197795e-05, + "loss": 0.9017, + "step": 31130 + }, + { + "epoch": 2.0, + "learning_rate": 1.3341205264174844e-05, + "loss": 0.9163, + "step": 31131 + }, + { + "epoch": 2.0, + "learning_rate": 1.3340990575151893e-05, + "loss": 0.8997, + "step": 31132 + }, + { + "epoch": 2.0, + "learning_rate": 1.3340775886128943e-05, + "loss": 0.8172, + "step": 31133 + }, + { + "epoch": 2.0, + "learning_rate": 1.3340561197105992e-05, + "loss": 0.965, + "step": 31134 + }, + { + "epoch": 2.0, + "learning_rate": 1.3340346508083041e-05, + "loss": 0.8684, + "step": 31135 + }, + { + "epoch": 2.0, + "learning_rate": 1.3340131819060092e-05, + "loss": 0.8931, + "step": 31136 + }, + { + "epoch": 2.0, + "learning_rate": 1.3339917130037142e-05, + "loss": 0.8324, + "step": 31137 + }, + { + "epoch": 2.0, + "learning_rate": 1.3339702441014193e-05, + "loss": 0.8922, + "step": 31138 + }, + { + "epoch": 2.0, + "learning_rate": 1.3339487751991242e-05, + "loss": 0.9371, + "step": 31139 + }, + { + "epoch": 2.0, + "learning_rate": 1.3339273062968293e-05, + "loss": 0.9339, + "step": 31140 + }, + { + "epoch": 2.0, + "learning_rate": 1.3339058373945342e-05, + "loss": 0.8799, + "step": 31141 + }, + { + "epoch": 2.0, + "learning_rate": 1.333884368492239e-05, + "loss": 0.8688, + "step": 31142 + }, + { + "epoch": 2.0, + "learning_rate": 1.3338628995899441e-05, + "loss": 0.9317, + "step": 31143 + }, + { + "epoch": 2.0, + "learning_rate": 1.333841430687649e-05, + "loss": 0.9304, + "step": 31144 + }, + { + "epoch": 2.0, + "learning_rate": 1.333819961785354e-05, + "loss": 0.9325, + "step": 31145 + }, + { + "epoch": 2.0, + "learning_rate": 1.333798492883059e-05, + "loss": 0.8811, + "step": 31146 + }, + { + "epoch": 2.0, + "learning_rate": 1.3337770239807639e-05, + "loss": 0.8588, + "step": 31147 + }, + { + "epoch": 2.0, + "learning_rate": 1.333755555078469e-05, + "loss": 0.8589, + "step": 31148 + }, + { + "epoch": 2.0, + "learning_rate": 1.3337340861761738e-05, + "loss": 0.9174, + "step": 31149 + }, + { + "epoch": 2.0, + "learning_rate": 1.3337126172738787e-05, + "loss": 0.8407, + "step": 31150 + }, + { + "epoch": 2.0, + "learning_rate": 1.333691148371584e-05, + "loss": 0.9212, + "step": 31151 + }, + { + "epoch": 2.0, + "learning_rate": 1.3336696794692889e-05, + "loss": 0.8788, + "step": 31152 + }, + { + "epoch": 2.0, + "learning_rate": 1.333648210566994e-05, + "loss": 0.8976, + "step": 31153 + }, + { + "epoch": 2.0, + "learning_rate": 1.3336267416646988e-05, + "loss": 0.9262, + "step": 31154 + }, + { + "epoch": 2.0, + "learning_rate": 1.3336052727624037e-05, + "loss": 0.8676, + "step": 31155 + }, + { + "epoch": 2.0, + "learning_rate": 1.3335838038601088e-05, + "loss": 0.8588, + "step": 31156 + }, + { + "epoch": 2.0, + "learning_rate": 1.3335623349578137e-05, + "loss": 0.8812, + "step": 31157 + }, + { + "epoch": 2.0, + "learning_rate": 1.3335408660555188e-05, + "loss": 0.8696, + "step": 31158 + }, + { + "epoch": 2.0, + "learning_rate": 1.3335193971532237e-05, + "loss": 0.9206, + "step": 31159 + }, + { + "epoch": 2.0, + "learning_rate": 1.3334979282509286e-05, + "loss": 0.8226, + "step": 31160 + }, + { + "epoch": 2.0, + "learning_rate": 1.3334764593486336e-05, + "loss": 0.9306, + "step": 31161 + }, + { + "epoch": 2.0, + "learning_rate": 1.3334549904463385e-05, + "loss": 0.8684, + "step": 31162 + }, + { + "epoch": 2.0, + "learning_rate": 1.3334335215440434e-05, + "loss": 0.8846, + "step": 31163 + }, + { + "epoch": 2.01, + "learning_rate": 1.3334120526417485e-05, + "loss": 0.9035, + "step": 31164 + }, + { + "epoch": 2.01, + "learning_rate": 1.3333905837394534e-05, + "loss": 0.8375, + "step": 31165 + }, + { + "epoch": 2.01, + "learning_rate": 1.3333691148371586e-05, + "loss": 0.913, + "step": 31166 + }, + { + "epoch": 2.01, + "learning_rate": 1.3333476459348635e-05, + "loss": 0.8591, + "step": 31167 + }, + { + "epoch": 2.01, + "learning_rate": 1.3333261770325686e-05, + "loss": 0.89, + "step": 31168 + }, + { + "epoch": 2.01, + "learning_rate": 1.3333047081302735e-05, + "loss": 0.8747, + "step": 31169 + }, + { + "epoch": 2.01, + "learning_rate": 1.3332832392279784e-05, + "loss": 0.878, + "step": 31170 + }, + { + "epoch": 2.01, + "learning_rate": 1.3332617703256834e-05, + "loss": 0.886, + "step": 31171 + }, + { + "epoch": 2.01, + "learning_rate": 1.3332403014233883e-05, + "loss": 0.8599, + "step": 31172 + }, + { + "epoch": 2.01, + "learning_rate": 1.3332188325210932e-05, + "loss": 0.8561, + "step": 31173 + }, + { + "epoch": 2.01, + "learning_rate": 1.3331973636187983e-05, + "loss": 0.9429, + "step": 31174 + }, + { + "epoch": 2.01, + "learning_rate": 1.3331758947165032e-05, + "loss": 0.9443, + "step": 31175 + }, + { + "epoch": 2.01, + "learning_rate": 1.3331544258142083e-05, + "loss": 0.8817, + "step": 31176 + }, + { + "epoch": 2.01, + "learning_rate": 1.3331329569119132e-05, + "loss": 0.9135, + "step": 31177 + }, + { + "epoch": 2.01, + "learning_rate": 1.333111488009618e-05, + "loss": 0.8764, + "step": 31178 + }, + { + "epoch": 2.01, + "learning_rate": 1.3330900191073231e-05, + "loss": 0.9258, + "step": 31179 + }, + { + "epoch": 2.01, + "learning_rate": 1.3330685502050282e-05, + "loss": 0.95, + "step": 31180 + }, + { + "epoch": 2.01, + "learning_rate": 1.3330470813027332e-05, + "loss": 0.8663, + "step": 31181 + }, + { + "epoch": 2.01, + "learning_rate": 1.3330256124004381e-05, + "loss": 0.8769, + "step": 31182 + }, + { + "epoch": 2.01, + "learning_rate": 1.333004143498143e-05, + "loss": 0.879, + "step": 31183 + }, + { + "epoch": 2.01, + "learning_rate": 1.3329826745958481e-05, + "loss": 0.8595, + "step": 31184 + }, + { + "epoch": 2.01, + "learning_rate": 1.332961205693553e-05, + "loss": 0.9292, + "step": 31185 + }, + { + "epoch": 2.01, + "learning_rate": 1.332939736791258e-05, + "loss": 0.9433, + "step": 31186 + }, + { + "epoch": 2.01, + "learning_rate": 1.332918267888963e-05, + "loss": 0.87, + "step": 31187 + }, + { + "epoch": 2.01, + "learning_rate": 1.3328967989866679e-05, + "loss": 0.9153, + "step": 31188 + }, + { + "epoch": 2.01, + "learning_rate": 1.332875330084373e-05, + "loss": 0.9269, + "step": 31189 + }, + { + "epoch": 2.01, + "learning_rate": 1.3328538611820778e-05, + "loss": 0.9083, + "step": 31190 + }, + { + "epoch": 2.01, + "learning_rate": 1.3328323922797827e-05, + "loss": 0.907, + "step": 31191 + }, + { + "epoch": 2.01, + "learning_rate": 1.3328109233774878e-05, + "loss": 0.7816, + "step": 31192 + }, + { + "epoch": 2.01, + "learning_rate": 1.3327894544751927e-05, + "loss": 0.9302, + "step": 31193 + }, + { + "epoch": 2.01, + "learning_rate": 1.3327679855728976e-05, + "loss": 0.8676, + "step": 31194 + }, + { + "epoch": 2.01, + "learning_rate": 1.3327465166706028e-05, + "loss": 0.933, + "step": 31195 + }, + { + "epoch": 2.01, + "learning_rate": 1.3327250477683079e-05, + "loss": 0.8978, + "step": 31196 + }, + { + "epoch": 2.01, + "learning_rate": 1.3327035788660128e-05, + "loss": 0.9414, + "step": 31197 + }, + { + "epoch": 2.01, + "learning_rate": 1.3326821099637177e-05, + "loss": 0.8759, + "step": 31198 + }, + { + "epoch": 2.01, + "learning_rate": 1.3326606410614227e-05, + "loss": 0.9843, + "step": 31199 + }, + { + "epoch": 2.01, + "learning_rate": 1.3326391721591276e-05, + "loss": 0.8754, + "step": 31200 + }, + { + "epoch": 2.01, + "learning_rate": 1.3326177032568325e-05, + "loss": 0.8544, + "step": 31201 + }, + { + "epoch": 2.01, + "learning_rate": 1.3325962343545376e-05, + "loss": 0.8327, + "step": 31202 + }, + { + "epoch": 2.01, + "learning_rate": 1.3325747654522425e-05, + "loss": 0.8532, + "step": 31203 + }, + { + "epoch": 2.01, + "learning_rate": 1.3325532965499474e-05, + "loss": 0.8879, + "step": 31204 + }, + { + "epoch": 2.01, + "learning_rate": 1.3325318276476525e-05, + "loss": 0.89, + "step": 31205 + }, + { + "epoch": 2.01, + "learning_rate": 1.3325103587453574e-05, + "loss": 0.8888, + "step": 31206 + }, + { + "epoch": 2.01, + "learning_rate": 1.3324888898430624e-05, + "loss": 0.8347, + "step": 31207 + }, + { + "epoch": 2.01, + "learning_rate": 1.3324674209407673e-05, + "loss": 0.8732, + "step": 31208 + }, + { + "epoch": 2.01, + "learning_rate": 1.3324459520384726e-05, + "loss": 0.9164, + "step": 31209 + }, + { + "epoch": 2.01, + "learning_rate": 1.3324244831361774e-05, + "loss": 0.8665, + "step": 31210 + }, + { + "epoch": 2.01, + "learning_rate": 1.3324030142338823e-05, + "loss": 0.9334, + "step": 31211 + }, + { + "epoch": 2.01, + "learning_rate": 1.3323815453315874e-05, + "loss": 0.9533, + "step": 31212 + }, + { + "epoch": 2.01, + "learning_rate": 1.3323600764292923e-05, + "loss": 0.93, + "step": 31213 + }, + { + "epoch": 2.01, + "learning_rate": 1.3323386075269972e-05, + "loss": 0.8487, + "step": 31214 + }, + { + "epoch": 2.01, + "learning_rate": 1.3323171386247023e-05, + "loss": 0.8527, + "step": 31215 + }, + { + "epoch": 2.01, + "learning_rate": 1.3322956697224072e-05, + "loss": 0.914, + "step": 31216 + }, + { + "epoch": 2.01, + "learning_rate": 1.3322742008201122e-05, + "loss": 0.8734, + "step": 31217 + }, + { + "epoch": 2.01, + "learning_rate": 1.3322527319178171e-05, + "loss": 0.9136, + "step": 31218 + }, + { + "epoch": 2.01, + "learning_rate": 1.332231263015522e-05, + "loss": 0.8701, + "step": 31219 + }, + { + "epoch": 2.01, + "learning_rate": 1.3322097941132271e-05, + "loss": 0.9009, + "step": 31220 + }, + { + "epoch": 2.01, + "learning_rate": 1.332188325210932e-05, + "loss": 0.8762, + "step": 31221 + }, + { + "epoch": 2.01, + "learning_rate": 1.3321668563086369e-05, + "loss": 0.8453, + "step": 31222 + }, + { + "epoch": 2.01, + "learning_rate": 1.332145387406342e-05, + "loss": 0.9658, + "step": 31223 + }, + { + "epoch": 2.01, + "learning_rate": 1.332123918504047e-05, + "loss": 0.9086, + "step": 31224 + }, + { + "epoch": 2.01, + "learning_rate": 1.332102449601752e-05, + "loss": 0.8531, + "step": 31225 + }, + { + "epoch": 2.01, + "learning_rate": 1.332080980699457e-05, + "loss": 0.8692, + "step": 31226 + }, + { + "epoch": 2.01, + "learning_rate": 1.332059511797162e-05, + "loss": 0.8848, + "step": 31227 + }, + { + "epoch": 2.01, + "learning_rate": 1.332038042894867e-05, + "loss": 0.8397, + "step": 31228 + }, + { + "epoch": 2.01, + "learning_rate": 1.3320165739925718e-05, + "loss": 0.8695, + "step": 31229 + }, + { + "epoch": 2.01, + "learning_rate": 1.3319951050902769e-05, + "loss": 0.9272, + "step": 31230 + }, + { + "epoch": 2.01, + "learning_rate": 1.3319736361879818e-05, + "loss": 0.8632, + "step": 31231 + }, + { + "epoch": 2.01, + "learning_rate": 1.3319521672856867e-05, + "loss": 0.8505, + "step": 31232 + }, + { + "epoch": 2.01, + "learning_rate": 1.3319306983833918e-05, + "loss": 0.8728, + "step": 31233 + }, + { + "epoch": 2.01, + "learning_rate": 1.3319092294810967e-05, + "loss": 0.8815, + "step": 31234 + }, + { + "epoch": 2.01, + "learning_rate": 1.3318877605788017e-05, + "loss": 0.8671, + "step": 31235 + }, + { + "epoch": 2.01, + "learning_rate": 1.3318662916765066e-05, + "loss": 0.8606, + "step": 31236 + }, + { + "epoch": 2.01, + "learning_rate": 1.3318448227742115e-05, + "loss": 0.9194, + "step": 31237 + }, + { + "epoch": 2.01, + "learning_rate": 1.3318233538719168e-05, + "loss": 0.9015, + "step": 31238 + }, + { + "epoch": 2.01, + "learning_rate": 1.3318018849696216e-05, + "loss": 0.9331, + "step": 31239 + }, + { + "epoch": 2.01, + "learning_rate": 1.3317804160673267e-05, + "loss": 0.9207, + "step": 31240 + }, + { + "epoch": 2.01, + "learning_rate": 1.3317589471650316e-05, + "loss": 0.97, + "step": 31241 + }, + { + "epoch": 2.01, + "learning_rate": 1.3317374782627365e-05, + "loss": 0.8326, + "step": 31242 + }, + { + "epoch": 2.01, + "learning_rate": 1.3317160093604416e-05, + "loss": 0.8351, + "step": 31243 + }, + { + "epoch": 2.01, + "learning_rate": 1.3316945404581465e-05, + "loss": 0.9499, + "step": 31244 + }, + { + "epoch": 2.01, + "learning_rate": 1.3316730715558515e-05, + "loss": 0.8957, + "step": 31245 + }, + { + "epoch": 2.01, + "learning_rate": 1.3316516026535564e-05, + "loss": 0.8938, + "step": 31246 + }, + { + "epoch": 2.01, + "learning_rate": 1.3316301337512613e-05, + "loss": 0.9104, + "step": 31247 + }, + { + "epoch": 2.01, + "learning_rate": 1.3316086648489664e-05, + "loss": 0.8959, + "step": 31248 + }, + { + "epoch": 2.01, + "learning_rate": 1.3315871959466713e-05, + "loss": 0.9044, + "step": 31249 + }, + { + "epoch": 2.01, + "learning_rate": 1.3315657270443762e-05, + "loss": 0.8369, + "step": 31250 + }, + { + "epoch": 2.01, + "learning_rate": 1.3315442581420813e-05, + "loss": 0.8844, + "step": 31251 + }, + { + "epoch": 2.01, + "learning_rate": 1.3315227892397862e-05, + "loss": 0.9571, + "step": 31252 + }, + { + "epoch": 2.01, + "learning_rate": 1.3315013203374914e-05, + "loss": 0.9777, + "step": 31253 + }, + { + "epoch": 2.01, + "learning_rate": 1.3314798514351963e-05, + "loss": 0.9147, + "step": 31254 + }, + { + "epoch": 2.01, + "learning_rate": 1.3314583825329013e-05, + "loss": 0.934, + "step": 31255 + }, + { + "epoch": 2.01, + "learning_rate": 1.3314369136306062e-05, + "loss": 0.947, + "step": 31256 + }, + { + "epoch": 2.01, + "learning_rate": 1.3314154447283111e-05, + "loss": 0.9127, + "step": 31257 + }, + { + "epoch": 2.01, + "learning_rate": 1.3313939758260162e-05, + "loss": 0.9087, + "step": 31258 + }, + { + "epoch": 2.01, + "learning_rate": 1.3313725069237211e-05, + "loss": 0.9273, + "step": 31259 + }, + { + "epoch": 2.01, + "learning_rate": 1.331351038021426e-05, + "loss": 1.021, + "step": 31260 + }, + { + "epoch": 2.01, + "learning_rate": 1.331329569119131e-05, + "loss": 0.9467, + "step": 31261 + }, + { + "epoch": 2.01, + "learning_rate": 1.331308100216836e-05, + "loss": 0.8893, + "step": 31262 + }, + { + "epoch": 2.01, + "learning_rate": 1.331286631314541e-05, + "loss": 0.8946, + "step": 31263 + }, + { + "epoch": 2.01, + "learning_rate": 1.331265162412246e-05, + "loss": 0.8978, + "step": 31264 + }, + { + "epoch": 2.01, + "learning_rate": 1.3312436935099508e-05, + "loss": 0.885, + "step": 31265 + }, + { + "epoch": 2.01, + "learning_rate": 1.3312222246076559e-05, + "loss": 0.9394, + "step": 31266 + }, + { + "epoch": 2.01, + "learning_rate": 1.331200755705361e-05, + "loss": 0.9086, + "step": 31267 + }, + { + "epoch": 2.01, + "learning_rate": 1.331179286803066e-05, + "loss": 0.8486, + "step": 31268 + }, + { + "epoch": 2.01, + "learning_rate": 1.331157817900771e-05, + "loss": 0.8835, + "step": 31269 + }, + { + "epoch": 2.01, + "learning_rate": 1.3311363489984758e-05, + "loss": 0.8534, + "step": 31270 + }, + { + "epoch": 2.01, + "learning_rate": 1.3311148800961809e-05, + "loss": 0.8818, + "step": 31271 + }, + { + "epoch": 2.01, + "learning_rate": 1.3310934111938858e-05, + "loss": 0.9488, + "step": 31272 + }, + { + "epoch": 2.01, + "learning_rate": 1.3310719422915908e-05, + "loss": 0.9231, + "step": 31273 + }, + { + "epoch": 2.01, + "learning_rate": 1.3310504733892957e-05, + "loss": 0.8899, + "step": 31274 + }, + { + "epoch": 2.01, + "learning_rate": 1.3310290044870006e-05, + "loss": 0.8767, + "step": 31275 + }, + { + "epoch": 2.01, + "learning_rate": 1.3310075355847057e-05, + "loss": 0.8584, + "step": 31276 + }, + { + "epoch": 2.01, + "learning_rate": 1.3309860666824106e-05, + "loss": 0.9036, + "step": 31277 + }, + { + "epoch": 2.01, + "learning_rate": 1.3309645977801155e-05, + "loss": 0.8321, + "step": 31278 + }, + { + "epoch": 2.01, + "learning_rate": 1.3309431288778206e-05, + "loss": 0.9303, + "step": 31279 + }, + { + "epoch": 2.01, + "learning_rate": 1.3309216599755255e-05, + "loss": 0.8498, + "step": 31280 + }, + { + "epoch": 2.01, + "learning_rate": 1.3309001910732304e-05, + "loss": 0.8353, + "step": 31281 + }, + { + "epoch": 2.01, + "learning_rate": 1.3308787221709356e-05, + "loss": 0.8671, + "step": 31282 + }, + { + "epoch": 2.01, + "learning_rate": 1.3308572532686407e-05, + "loss": 0.9016, + "step": 31283 + }, + { + "epoch": 2.01, + "learning_rate": 1.3308357843663455e-05, + "loss": 0.9204, + "step": 31284 + }, + { + "epoch": 2.01, + "learning_rate": 1.3308143154640504e-05, + "loss": 0.8898, + "step": 31285 + }, + { + "epoch": 2.01, + "learning_rate": 1.3307928465617555e-05, + "loss": 0.9181, + "step": 31286 + }, + { + "epoch": 2.01, + "learning_rate": 1.3307713776594604e-05, + "loss": 0.9008, + "step": 31287 + }, + { + "epoch": 2.01, + "learning_rate": 1.3307499087571653e-05, + "loss": 0.8709, + "step": 31288 + }, + { + "epoch": 2.01, + "learning_rate": 1.3307284398548704e-05, + "loss": 0.9476, + "step": 31289 + }, + { + "epoch": 2.01, + "learning_rate": 1.3307069709525753e-05, + "loss": 0.8675, + "step": 31290 + }, + { + "epoch": 2.01, + "learning_rate": 1.3306855020502802e-05, + "loss": 0.8417, + "step": 31291 + }, + { + "epoch": 2.01, + "learning_rate": 1.3306640331479852e-05, + "loss": 0.8775, + "step": 31292 + }, + { + "epoch": 2.01, + "learning_rate": 1.3306425642456901e-05, + "loss": 0.9231, + "step": 31293 + }, + { + "epoch": 2.01, + "learning_rate": 1.3306210953433952e-05, + "loss": 0.9258, + "step": 31294 + }, + { + "epoch": 2.01, + "learning_rate": 1.3305996264411001e-05, + "loss": 0.9217, + "step": 31295 + }, + { + "epoch": 2.01, + "learning_rate": 1.3305781575388053e-05, + "loss": 0.8965, + "step": 31296 + }, + { + "epoch": 2.01, + "learning_rate": 1.3305566886365102e-05, + "loss": 0.8716, + "step": 31297 + }, + { + "epoch": 2.01, + "learning_rate": 1.3305352197342151e-05, + "loss": 0.9558, + "step": 31298 + }, + { + "epoch": 2.01, + "learning_rate": 1.3305137508319202e-05, + "loss": 0.9642, + "step": 31299 + }, + { + "epoch": 2.01, + "learning_rate": 1.330492281929625e-05, + "loss": 0.8983, + "step": 31300 + }, + { + "epoch": 2.01, + "learning_rate": 1.33047081302733e-05, + "loss": 0.8735, + "step": 31301 + }, + { + "epoch": 2.01, + "learning_rate": 1.330449344125035e-05, + "loss": 0.9534, + "step": 31302 + }, + { + "epoch": 2.01, + "learning_rate": 1.33042787522274e-05, + "loss": 0.8987, + "step": 31303 + }, + { + "epoch": 2.01, + "learning_rate": 1.330406406320445e-05, + "loss": 0.9059, + "step": 31304 + }, + { + "epoch": 2.01, + "learning_rate": 1.3303849374181499e-05, + "loss": 0.9061, + "step": 31305 + }, + { + "epoch": 2.01, + "learning_rate": 1.3303634685158548e-05, + "loss": 0.8298, + "step": 31306 + }, + { + "epoch": 2.01, + "learning_rate": 1.3303419996135599e-05, + "loss": 0.8924, + "step": 31307 + }, + { + "epoch": 2.01, + "learning_rate": 1.3303205307112648e-05, + "loss": 0.8936, + "step": 31308 + }, + { + "epoch": 2.01, + "learning_rate": 1.3302990618089697e-05, + "loss": 0.8667, + "step": 31309 + }, + { + "epoch": 2.01, + "learning_rate": 1.3302775929066747e-05, + "loss": 0.9023, + "step": 31310 + }, + { + "epoch": 2.01, + "learning_rate": 1.3302561240043798e-05, + "loss": 0.8703, + "step": 31311 + }, + { + "epoch": 2.01, + "learning_rate": 1.3302346551020849e-05, + "loss": 0.9242, + "step": 31312 + }, + { + "epoch": 2.01, + "learning_rate": 1.3302131861997897e-05, + "loss": 0.9196, + "step": 31313 + }, + { + "epoch": 2.01, + "learning_rate": 1.3301917172974948e-05, + "loss": 0.9208, + "step": 31314 + }, + { + "epoch": 2.01, + "learning_rate": 1.3301702483951997e-05, + "loss": 0.8938, + "step": 31315 + }, + { + "epoch": 2.01, + "learning_rate": 1.3301487794929046e-05, + "loss": 0.8767, + "step": 31316 + }, + { + "epoch": 2.01, + "learning_rate": 1.3301273105906097e-05, + "loss": 0.8944, + "step": 31317 + }, + { + "epoch": 2.01, + "learning_rate": 1.3301058416883146e-05, + "loss": 0.9429, + "step": 31318 + }, + { + "epoch": 2.01, + "learning_rate": 1.3300843727860195e-05, + "loss": 0.9319, + "step": 31319 + }, + { + "epoch": 2.02, + "learning_rate": 1.3300629038837245e-05, + "loss": 0.9143, + "step": 31320 + }, + { + "epoch": 2.02, + "learning_rate": 1.3300414349814294e-05, + "loss": 0.9067, + "step": 31321 + }, + { + "epoch": 2.02, + "learning_rate": 1.3300199660791345e-05, + "loss": 0.8874, + "step": 31322 + }, + { + "epoch": 2.02, + "learning_rate": 1.3299984971768394e-05, + "loss": 0.9329, + "step": 31323 + }, + { + "epoch": 2.02, + "learning_rate": 1.3299770282745443e-05, + "loss": 0.8688, + "step": 31324 + }, + { + "epoch": 2.02, + "learning_rate": 1.3299555593722495e-05, + "loss": 0.8344, + "step": 31325 + }, + { + "epoch": 2.02, + "learning_rate": 1.3299340904699544e-05, + "loss": 0.8856, + "step": 31326 + }, + { + "epoch": 2.02, + "learning_rate": 1.3299126215676595e-05, + "loss": 0.9354, + "step": 31327 + }, + { + "epoch": 2.02, + "learning_rate": 1.3298911526653644e-05, + "loss": 0.8911, + "step": 31328 + }, + { + "epoch": 2.02, + "learning_rate": 1.3298696837630693e-05, + "loss": 0.8956, + "step": 31329 + }, + { + "epoch": 2.02, + "learning_rate": 1.3298482148607743e-05, + "loss": 0.9434, + "step": 31330 + }, + { + "epoch": 2.02, + "learning_rate": 1.3298267459584792e-05, + "loss": 0.8937, + "step": 31331 + }, + { + "epoch": 2.02, + "learning_rate": 1.3298052770561843e-05, + "loss": 0.8957, + "step": 31332 + }, + { + "epoch": 2.02, + "learning_rate": 1.3297838081538892e-05, + "loss": 0.8777, + "step": 31333 + }, + { + "epoch": 2.02, + "learning_rate": 1.3297623392515941e-05, + "loss": 0.8827, + "step": 31334 + }, + { + "epoch": 2.02, + "learning_rate": 1.3297408703492992e-05, + "loss": 0.9766, + "step": 31335 + }, + { + "epoch": 2.02, + "learning_rate": 1.329719401447004e-05, + "loss": 0.8744, + "step": 31336 + }, + { + "epoch": 2.02, + "learning_rate": 1.329697932544709e-05, + "loss": 0.8704, + "step": 31337 + }, + { + "epoch": 2.02, + "learning_rate": 1.329676463642414e-05, + "loss": 0.8267, + "step": 31338 + }, + { + "epoch": 2.02, + "learning_rate": 1.329654994740119e-05, + "loss": 0.8745, + "step": 31339 + }, + { + "epoch": 2.02, + "learning_rate": 1.3296335258378242e-05, + "loss": 0.8222, + "step": 31340 + }, + { + "epoch": 2.02, + "learning_rate": 1.329612056935529e-05, + "loss": 0.8955, + "step": 31341 + }, + { + "epoch": 2.02, + "learning_rate": 1.3295905880332341e-05, + "loss": 0.8552, + "step": 31342 + }, + { + "epoch": 2.02, + "learning_rate": 1.329569119130939e-05, + "loss": 0.8076, + "step": 31343 + }, + { + "epoch": 2.02, + "learning_rate": 1.3295476502286439e-05, + "loss": 0.99, + "step": 31344 + }, + { + "epoch": 2.02, + "learning_rate": 1.329526181326349e-05, + "loss": 0.8501, + "step": 31345 + }, + { + "epoch": 2.02, + "learning_rate": 1.3295047124240539e-05, + "loss": 0.8819, + "step": 31346 + }, + { + "epoch": 2.02, + "learning_rate": 1.3294832435217588e-05, + "loss": 0.8758, + "step": 31347 + }, + { + "epoch": 2.02, + "learning_rate": 1.3294617746194638e-05, + "loss": 0.9567, + "step": 31348 + }, + { + "epoch": 2.02, + "learning_rate": 1.3294403057171687e-05, + "loss": 0.8911, + "step": 31349 + }, + { + "epoch": 2.02, + "learning_rate": 1.3294188368148736e-05, + "loss": 0.9046, + "step": 31350 + }, + { + "epoch": 2.02, + "learning_rate": 1.3293973679125787e-05, + "loss": 0.9227, + "step": 31351 + }, + { + "epoch": 2.02, + "learning_rate": 1.3293758990102836e-05, + "loss": 0.8923, + "step": 31352 + }, + { + "epoch": 2.02, + "learning_rate": 1.3293544301079887e-05, + "loss": 0.8922, + "step": 31353 + }, + { + "epoch": 2.02, + "learning_rate": 1.3293329612056936e-05, + "loss": 0.9722, + "step": 31354 + }, + { + "epoch": 2.02, + "learning_rate": 1.3293114923033988e-05, + "loss": 0.8868, + "step": 31355 + }, + { + "epoch": 2.02, + "learning_rate": 1.3292900234011037e-05, + "loss": 0.8616, + "step": 31356 + }, + { + "epoch": 2.02, + "learning_rate": 1.3292685544988086e-05, + "loss": 0.8236, + "step": 31357 + }, + { + "epoch": 2.02, + "learning_rate": 1.3292470855965136e-05, + "loss": 0.9412, + "step": 31358 + }, + { + "epoch": 2.02, + "learning_rate": 1.3292256166942185e-05, + "loss": 0.8372, + "step": 31359 + }, + { + "epoch": 2.02, + "learning_rate": 1.3292041477919234e-05, + "loss": 0.8466, + "step": 31360 + }, + { + "epoch": 2.02, + "learning_rate": 1.3291826788896285e-05, + "loss": 0.8945, + "step": 31361 + }, + { + "epoch": 2.02, + "learning_rate": 1.3291612099873334e-05, + "loss": 0.8811, + "step": 31362 + }, + { + "epoch": 2.02, + "learning_rate": 1.3291397410850385e-05, + "loss": 0.8901, + "step": 31363 + }, + { + "epoch": 2.02, + "learning_rate": 1.3291182721827434e-05, + "loss": 0.9276, + "step": 31364 + }, + { + "epoch": 2.02, + "learning_rate": 1.3290968032804483e-05, + "loss": 0.9056, + "step": 31365 + }, + { + "epoch": 2.02, + "learning_rate": 1.3290753343781533e-05, + "loss": 0.8725, + "step": 31366 + }, + { + "epoch": 2.02, + "learning_rate": 1.3290538654758582e-05, + "loss": 0.8921, + "step": 31367 + }, + { + "epoch": 2.02, + "learning_rate": 1.3290323965735631e-05, + "loss": 0.9315, + "step": 31368 + }, + { + "epoch": 2.02, + "learning_rate": 1.3290109276712684e-05, + "loss": 0.86, + "step": 31369 + }, + { + "epoch": 2.02, + "learning_rate": 1.3289894587689734e-05, + "loss": 0.9203, + "step": 31370 + }, + { + "epoch": 2.02, + "learning_rate": 1.3289679898666783e-05, + "loss": 0.8117, + "step": 31371 + }, + { + "epoch": 2.02, + "learning_rate": 1.3289465209643832e-05, + "loss": 0.8756, + "step": 31372 + }, + { + "epoch": 2.02, + "learning_rate": 1.3289250520620883e-05, + "loss": 0.8938, + "step": 31373 + }, + { + "epoch": 2.02, + "learning_rate": 1.3289035831597932e-05, + "loss": 0.9251, + "step": 31374 + }, + { + "epoch": 2.02, + "learning_rate": 1.328882114257498e-05, + "loss": 0.9067, + "step": 31375 + }, + { + "epoch": 2.02, + "learning_rate": 1.3288606453552031e-05, + "loss": 0.9214, + "step": 31376 + }, + { + "epoch": 2.02, + "learning_rate": 1.328839176452908e-05, + "loss": 0.8861, + "step": 31377 + }, + { + "epoch": 2.02, + "learning_rate": 1.328817707550613e-05, + "loss": 0.961, + "step": 31378 + }, + { + "epoch": 2.02, + "learning_rate": 1.328796238648318e-05, + "loss": 0.7953, + "step": 31379 + }, + { + "epoch": 2.02, + "learning_rate": 1.3287747697460229e-05, + "loss": 0.8975, + "step": 31380 + }, + { + "epoch": 2.02, + "learning_rate": 1.328753300843728e-05, + "loss": 0.8417, + "step": 31381 + }, + { + "epoch": 2.02, + "learning_rate": 1.3287318319414329e-05, + "loss": 0.8172, + "step": 31382 + }, + { + "epoch": 2.02, + "learning_rate": 1.3287103630391378e-05, + "loss": 0.9419, + "step": 31383 + }, + { + "epoch": 2.02, + "learning_rate": 1.328688894136843e-05, + "loss": 0.8905, + "step": 31384 + }, + { + "epoch": 2.02, + "learning_rate": 1.3286674252345479e-05, + "loss": 0.8834, + "step": 31385 + }, + { + "epoch": 2.02, + "learning_rate": 1.328645956332253e-05, + "loss": 0.8875, + "step": 31386 + }, + { + "epoch": 2.02, + "learning_rate": 1.3286244874299579e-05, + "loss": 0.8619, + "step": 31387 + }, + { + "epoch": 2.02, + "learning_rate": 1.3286030185276627e-05, + "loss": 0.9507, + "step": 31388 + }, + { + "epoch": 2.02, + "learning_rate": 1.3285815496253678e-05, + "loss": 0.959, + "step": 31389 + }, + { + "epoch": 2.02, + "learning_rate": 1.3285600807230727e-05, + "loss": 0.9113, + "step": 31390 + }, + { + "epoch": 2.02, + "learning_rate": 1.3285386118207778e-05, + "loss": 0.803, + "step": 31391 + }, + { + "epoch": 2.02, + "learning_rate": 1.3285171429184827e-05, + "loss": 0.9176, + "step": 31392 + }, + { + "epoch": 2.02, + "learning_rate": 1.3284956740161876e-05, + "loss": 0.8934, + "step": 31393 + }, + { + "epoch": 2.02, + "learning_rate": 1.3284742051138926e-05, + "loss": 0.9352, + "step": 31394 + }, + { + "epoch": 2.02, + "learning_rate": 1.3284527362115975e-05, + "loss": 0.9433, + "step": 31395 + }, + { + "epoch": 2.02, + "learning_rate": 1.3284312673093024e-05, + "loss": 0.9328, + "step": 31396 + }, + { + "epoch": 2.02, + "learning_rate": 1.3284097984070075e-05, + "loss": 0.9513, + "step": 31397 + }, + { + "epoch": 2.02, + "learning_rate": 1.3283883295047126e-05, + "loss": 0.8348, + "step": 31398 + }, + { + "epoch": 2.02, + "learning_rate": 1.3283668606024176e-05, + "loss": 0.9341, + "step": 31399 + }, + { + "epoch": 2.02, + "learning_rate": 1.3283453917001225e-05, + "loss": 0.8726, + "step": 31400 + }, + { + "epoch": 2.02, + "learning_rate": 1.3283239227978276e-05, + "loss": 0.897, + "step": 31401 + }, + { + "epoch": 2.02, + "learning_rate": 1.3283024538955325e-05, + "loss": 0.949, + "step": 31402 + }, + { + "epoch": 2.02, + "learning_rate": 1.3282809849932374e-05, + "loss": 0.8996, + "step": 31403 + }, + { + "epoch": 2.02, + "learning_rate": 1.3282595160909424e-05, + "loss": 0.9089, + "step": 31404 + }, + { + "epoch": 2.02, + "learning_rate": 1.3282380471886473e-05, + "loss": 0.8754, + "step": 31405 + }, + { + "epoch": 2.02, + "learning_rate": 1.3282165782863522e-05, + "loss": 0.8927, + "step": 31406 + }, + { + "epoch": 2.02, + "learning_rate": 1.3281951093840573e-05, + "loss": 0.8918, + "step": 31407 + }, + { + "epoch": 2.02, + "learning_rate": 1.3281736404817622e-05, + "loss": 0.7853, + "step": 31408 + }, + { + "epoch": 2.02, + "learning_rate": 1.3281521715794673e-05, + "loss": 0.8688, + "step": 31409 + }, + { + "epoch": 2.02, + "learning_rate": 1.3281307026771722e-05, + "loss": 0.8967, + "step": 31410 + }, + { + "epoch": 2.02, + "learning_rate": 1.328109233774877e-05, + "loss": 0.9317, + "step": 31411 + }, + { + "epoch": 2.02, + "learning_rate": 1.3280877648725821e-05, + "loss": 0.9021, + "step": 31412 + }, + { + "epoch": 2.02, + "learning_rate": 1.3280662959702872e-05, + "loss": 0.921, + "step": 31413 + }, + { + "epoch": 2.02, + "learning_rate": 1.3280448270679923e-05, + "loss": 0.9275, + "step": 31414 + }, + { + "epoch": 2.02, + "learning_rate": 1.3280233581656972e-05, + "loss": 0.8657, + "step": 31415 + }, + { + "epoch": 2.02, + "learning_rate": 1.328001889263402e-05, + "loss": 0.8913, + "step": 31416 + }, + { + "epoch": 2.02, + "learning_rate": 1.3279804203611071e-05, + "loss": 0.8765, + "step": 31417 + }, + { + "epoch": 2.02, + "learning_rate": 1.327958951458812e-05, + "loss": 0.8946, + "step": 31418 + }, + { + "epoch": 2.02, + "learning_rate": 1.327937482556517e-05, + "loss": 0.8751, + "step": 31419 + }, + { + "epoch": 2.02, + "learning_rate": 1.327916013654222e-05, + "loss": 0.9153, + "step": 31420 + }, + { + "epoch": 2.02, + "learning_rate": 1.3278945447519269e-05, + "loss": 0.9346, + "step": 31421 + }, + { + "epoch": 2.02, + "learning_rate": 1.327873075849632e-05, + "loss": 0.8568, + "step": 31422 + }, + { + "epoch": 2.02, + "learning_rate": 1.3278516069473368e-05, + "loss": 0.8745, + "step": 31423 + }, + { + "epoch": 2.02, + "learning_rate": 1.3278301380450417e-05, + "loss": 0.9283, + "step": 31424 + }, + { + "epoch": 2.02, + "learning_rate": 1.3278086691427468e-05, + "loss": 0.9037, + "step": 31425 + }, + { + "epoch": 2.02, + "learning_rate": 1.3277872002404517e-05, + "loss": 0.9305, + "step": 31426 + }, + { + "epoch": 2.02, + "learning_rate": 1.327765731338157e-05, + "loss": 0.8686, + "step": 31427 + }, + { + "epoch": 2.02, + "learning_rate": 1.3277442624358618e-05, + "loss": 0.9004, + "step": 31428 + }, + { + "epoch": 2.02, + "learning_rate": 1.3277227935335669e-05, + "loss": 0.9573, + "step": 31429 + }, + { + "epoch": 2.02, + "learning_rate": 1.3277013246312718e-05, + "loss": 0.8202, + "step": 31430 + }, + { + "epoch": 2.02, + "learning_rate": 1.3276798557289767e-05, + "loss": 0.8756, + "step": 31431 + }, + { + "epoch": 2.02, + "learning_rate": 1.3276583868266817e-05, + "loss": 0.905, + "step": 31432 + }, + { + "epoch": 2.02, + "learning_rate": 1.3276369179243866e-05, + "loss": 0.8593, + "step": 31433 + }, + { + "epoch": 2.02, + "learning_rate": 1.3276154490220915e-05, + "loss": 0.9585, + "step": 31434 + }, + { + "epoch": 2.02, + "learning_rate": 1.3275939801197966e-05, + "loss": 0.8422, + "step": 31435 + }, + { + "epoch": 2.02, + "learning_rate": 1.3275725112175015e-05, + "loss": 0.9279, + "step": 31436 + }, + { + "epoch": 2.02, + "learning_rate": 1.3275510423152064e-05, + "loss": 0.9555, + "step": 31437 + }, + { + "epoch": 2.02, + "learning_rate": 1.3275295734129115e-05, + "loss": 0.9236, + "step": 31438 + }, + { + "epoch": 2.02, + "learning_rate": 1.3275081045106164e-05, + "loss": 0.8645, + "step": 31439 + }, + { + "epoch": 2.02, + "learning_rate": 1.3274866356083214e-05, + "loss": 0.8185, + "step": 31440 + }, + { + "epoch": 2.02, + "learning_rate": 1.3274651667060263e-05, + "loss": 0.8455, + "step": 31441 + }, + { + "epoch": 2.02, + "learning_rate": 1.3274436978037316e-05, + "loss": 0.9169, + "step": 31442 + }, + { + "epoch": 2.02, + "learning_rate": 1.3274222289014365e-05, + "loss": 0.9856, + "step": 31443 + }, + { + "epoch": 2.02, + "learning_rate": 1.3274007599991414e-05, + "loss": 0.9365, + "step": 31444 + }, + { + "epoch": 2.02, + "learning_rate": 1.3273792910968464e-05, + "loss": 0.8935, + "step": 31445 + }, + { + "epoch": 2.02, + "learning_rate": 1.3273578221945513e-05, + "loss": 0.8626, + "step": 31446 + }, + { + "epoch": 2.02, + "learning_rate": 1.3273363532922562e-05, + "loss": 0.8832, + "step": 31447 + }, + { + "epoch": 2.02, + "learning_rate": 1.3273148843899613e-05, + "loss": 0.8499, + "step": 31448 + }, + { + "epoch": 2.02, + "learning_rate": 1.3272934154876662e-05, + "loss": 0.9285, + "step": 31449 + }, + { + "epoch": 2.02, + "learning_rate": 1.3272719465853712e-05, + "loss": 0.9669, + "step": 31450 + }, + { + "epoch": 2.02, + "learning_rate": 1.3272504776830761e-05, + "loss": 0.9318, + "step": 31451 + }, + { + "epoch": 2.02, + "learning_rate": 1.327229008780781e-05, + "loss": 0.8382, + "step": 31452 + }, + { + "epoch": 2.02, + "learning_rate": 1.3272075398784861e-05, + "loss": 0.888, + "step": 31453 + }, + { + "epoch": 2.02, + "learning_rate": 1.327186070976191e-05, + "loss": 0.8927, + "step": 31454 + }, + { + "epoch": 2.02, + "learning_rate": 1.3271646020738959e-05, + "loss": 0.867, + "step": 31455 + }, + { + "epoch": 2.02, + "learning_rate": 1.3271431331716011e-05, + "loss": 0.8718, + "step": 31456 + }, + { + "epoch": 2.02, + "learning_rate": 1.327121664269306e-05, + "loss": 0.8053, + "step": 31457 + }, + { + "epoch": 2.02, + "learning_rate": 1.3271001953670111e-05, + "loss": 0.8419, + "step": 31458 + }, + { + "epoch": 2.02, + "learning_rate": 1.327078726464716e-05, + "loss": 0.8588, + "step": 31459 + }, + { + "epoch": 2.02, + "learning_rate": 1.327057257562421e-05, + "loss": 0.878, + "step": 31460 + }, + { + "epoch": 2.02, + "learning_rate": 1.327035788660126e-05, + "loss": 0.9414, + "step": 31461 + }, + { + "epoch": 2.02, + "learning_rate": 1.3270143197578308e-05, + "loss": 0.9643, + "step": 31462 + }, + { + "epoch": 2.02, + "learning_rate": 1.3269928508555359e-05, + "loss": 0.8661, + "step": 31463 + }, + { + "epoch": 2.02, + "learning_rate": 1.3269713819532408e-05, + "loss": 0.8727, + "step": 31464 + }, + { + "epoch": 2.02, + "learning_rate": 1.3269499130509457e-05, + "loss": 0.8699, + "step": 31465 + }, + { + "epoch": 2.02, + "learning_rate": 1.3269284441486508e-05, + "loss": 0.8562, + "step": 31466 + }, + { + "epoch": 2.02, + "learning_rate": 1.3269069752463557e-05, + "loss": 0.8959, + "step": 31467 + }, + { + "epoch": 2.02, + "learning_rate": 1.3268855063440607e-05, + "loss": 0.7974, + "step": 31468 + }, + { + "epoch": 2.02, + "learning_rate": 1.3268640374417656e-05, + "loss": 0.8904, + "step": 31469 + }, + { + "epoch": 2.02, + "learning_rate": 1.3268425685394705e-05, + "loss": 0.8422, + "step": 31470 + }, + { + "epoch": 2.02, + "learning_rate": 1.3268210996371758e-05, + "loss": 0.871, + "step": 31471 + }, + { + "epoch": 2.02, + "learning_rate": 1.3267996307348807e-05, + "loss": 0.8523, + "step": 31472 + }, + { + "epoch": 2.02, + "learning_rate": 1.3267781618325857e-05, + "loss": 0.9646, + "step": 31473 + }, + { + "epoch": 2.02, + "learning_rate": 1.3267566929302906e-05, + "loss": 0.8832, + "step": 31474 + }, + { + "epoch": 2.03, + "learning_rate": 1.3267352240279955e-05, + "loss": 0.9083, + "step": 31475 + }, + { + "epoch": 2.03, + "learning_rate": 1.3267137551257006e-05, + "loss": 0.8654, + "step": 31476 + }, + { + "epoch": 2.03, + "learning_rate": 1.3266922862234055e-05, + "loss": 0.9399, + "step": 31477 + }, + { + "epoch": 2.03, + "learning_rate": 1.3266708173211105e-05, + "loss": 0.8931, + "step": 31478 + }, + { + "epoch": 2.03, + "learning_rate": 1.3266493484188154e-05, + "loss": 0.9871, + "step": 31479 + }, + { + "epoch": 2.03, + "learning_rate": 1.3266278795165203e-05, + "loss": 0.8716, + "step": 31480 + }, + { + "epoch": 2.03, + "learning_rate": 1.3266064106142254e-05, + "loss": 0.9158, + "step": 31481 + }, + { + "epoch": 2.03, + "learning_rate": 1.3265849417119303e-05, + "loss": 0.8767, + "step": 31482 + }, + { + "epoch": 2.03, + "learning_rate": 1.3265634728096352e-05, + "loss": 0.9723, + "step": 31483 + }, + { + "epoch": 2.03, + "learning_rate": 1.3265420039073403e-05, + "loss": 0.863, + "step": 31484 + }, + { + "epoch": 2.03, + "learning_rate": 1.3265205350050453e-05, + "loss": 0.9208, + "step": 31485 + }, + { + "epoch": 2.03, + "learning_rate": 1.3264990661027504e-05, + "loss": 0.9577, + "step": 31486 + }, + { + "epoch": 2.03, + "learning_rate": 1.3264775972004553e-05, + "loss": 0.89, + "step": 31487 + }, + { + "epoch": 2.03, + "learning_rate": 1.3264561282981604e-05, + "loss": 0.8274, + "step": 31488 + }, + { + "epoch": 2.03, + "learning_rate": 1.3264346593958653e-05, + "loss": 0.9062, + "step": 31489 + }, + { + "epoch": 2.03, + "learning_rate": 1.3264131904935702e-05, + "loss": 0.876, + "step": 31490 + }, + { + "epoch": 2.03, + "learning_rate": 1.3263917215912752e-05, + "loss": 0.874, + "step": 31491 + }, + { + "epoch": 2.03, + "learning_rate": 1.3263702526889801e-05, + "loss": 0.9593, + "step": 31492 + }, + { + "epoch": 2.03, + "learning_rate": 1.326348783786685e-05, + "loss": 0.8529, + "step": 31493 + }, + { + "epoch": 2.03, + "learning_rate": 1.32632731488439e-05, + "loss": 0.842, + "step": 31494 + }, + { + "epoch": 2.03, + "learning_rate": 1.326305845982095e-05, + "loss": 0.8911, + "step": 31495 + }, + { + "epoch": 2.03, + "learning_rate": 1.3262843770798e-05, + "loss": 0.8286, + "step": 31496 + }, + { + "epoch": 2.03, + "learning_rate": 1.326262908177505e-05, + "loss": 0.8673, + "step": 31497 + }, + { + "epoch": 2.03, + "learning_rate": 1.3262414392752098e-05, + "loss": 0.9393, + "step": 31498 + }, + { + "epoch": 2.03, + "learning_rate": 1.3262199703729149e-05, + "loss": 0.8821, + "step": 31499 + }, + { + "epoch": 2.03, + "learning_rate": 1.32619850147062e-05, + "loss": 0.8999, + "step": 31500 + }, + { + "epoch": 2.03, + "learning_rate": 1.326177032568325e-05, + "loss": 0.9538, + "step": 31501 + }, + { + "epoch": 2.03, + "learning_rate": 1.32615556366603e-05, + "loss": 0.9098, + "step": 31502 + }, + { + "epoch": 2.03, + "learning_rate": 1.3261340947637348e-05, + "loss": 0.9179, + "step": 31503 + }, + { + "epoch": 2.03, + "learning_rate": 1.3261126258614399e-05, + "loss": 0.8953, + "step": 31504 + }, + { + "epoch": 2.03, + "learning_rate": 1.3260911569591448e-05, + "loss": 0.8515, + "step": 31505 + }, + { + "epoch": 2.03, + "learning_rate": 1.3260696880568499e-05, + "loss": 0.8592, + "step": 31506 + }, + { + "epoch": 2.03, + "learning_rate": 1.3260482191545547e-05, + "loss": 0.8931, + "step": 31507 + }, + { + "epoch": 2.03, + "learning_rate": 1.3260267502522596e-05, + "loss": 0.9309, + "step": 31508 + }, + { + "epoch": 2.03, + "learning_rate": 1.3260052813499647e-05, + "loss": 0.9391, + "step": 31509 + }, + { + "epoch": 2.03, + "learning_rate": 1.3259838124476696e-05, + "loss": 0.9164, + "step": 31510 + }, + { + "epoch": 2.03, + "learning_rate": 1.3259623435453745e-05, + "loss": 0.865, + "step": 31511 + }, + { + "epoch": 2.03, + "learning_rate": 1.3259408746430796e-05, + "loss": 0.8724, + "step": 31512 + }, + { + "epoch": 2.03, + "learning_rate": 1.3259194057407845e-05, + "loss": 0.9285, + "step": 31513 + }, + { + "epoch": 2.03, + "learning_rate": 1.3258979368384897e-05, + "loss": 0.8932, + "step": 31514 + }, + { + "epoch": 2.03, + "learning_rate": 1.3258764679361946e-05, + "loss": 0.9147, + "step": 31515 + }, + { + "epoch": 2.03, + "learning_rate": 1.3258549990338997e-05, + "loss": 0.8854, + "step": 31516 + }, + { + "epoch": 2.03, + "learning_rate": 1.3258335301316046e-05, + "loss": 0.8912, + "step": 31517 + }, + { + "epoch": 2.03, + "learning_rate": 1.3258120612293095e-05, + "loss": 0.8961, + "step": 31518 + }, + { + "epoch": 2.03, + "learning_rate": 1.3257905923270145e-05, + "loss": 0.921, + "step": 31519 + }, + { + "epoch": 2.03, + "learning_rate": 1.3257691234247194e-05, + "loss": 0.8421, + "step": 31520 + }, + { + "epoch": 2.03, + "learning_rate": 1.3257476545224243e-05, + "loss": 0.8985, + "step": 31521 + }, + { + "epoch": 2.03, + "learning_rate": 1.3257261856201294e-05, + "loss": 0.9025, + "step": 31522 + }, + { + "epoch": 2.03, + "learning_rate": 1.3257047167178343e-05, + "loss": 0.8324, + "step": 31523 + }, + { + "epoch": 2.03, + "learning_rate": 1.3256832478155392e-05, + "loss": 0.874, + "step": 31524 + }, + { + "epoch": 2.03, + "learning_rate": 1.3256617789132442e-05, + "loss": 0.8561, + "step": 31525 + }, + { + "epoch": 2.03, + "learning_rate": 1.3256403100109491e-05, + "loss": 0.9433, + "step": 31526 + }, + { + "epoch": 2.03, + "learning_rate": 1.3256188411086542e-05, + "loss": 0.8347, + "step": 31527 + }, + { + "epoch": 2.03, + "learning_rate": 1.3255973722063591e-05, + "loss": 0.9173, + "step": 31528 + }, + { + "epoch": 2.03, + "learning_rate": 1.3255759033040643e-05, + "loss": 0.8606, + "step": 31529 + }, + { + "epoch": 2.03, + "learning_rate": 1.3255544344017692e-05, + "loss": 0.8577, + "step": 31530 + }, + { + "epoch": 2.03, + "learning_rate": 1.3255329654994741e-05, + "loss": 0.88, + "step": 31531 + }, + { + "epoch": 2.03, + "learning_rate": 1.3255114965971792e-05, + "loss": 0.8456, + "step": 31532 + }, + { + "epoch": 2.03, + "learning_rate": 1.3254900276948841e-05, + "loss": 0.8554, + "step": 31533 + }, + { + "epoch": 2.03, + "learning_rate": 1.325468558792589e-05, + "loss": 0.9726, + "step": 31534 + }, + { + "epoch": 2.03, + "learning_rate": 1.325447089890294e-05, + "loss": 0.8608, + "step": 31535 + }, + { + "epoch": 2.03, + "learning_rate": 1.325425620987999e-05, + "loss": 0.9663, + "step": 31536 + }, + { + "epoch": 2.03, + "learning_rate": 1.325404152085704e-05, + "loss": 0.8498, + "step": 31537 + }, + { + "epoch": 2.03, + "learning_rate": 1.3253826831834089e-05, + "loss": 0.8379, + "step": 31538 + }, + { + "epoch": 2.03, + "learning_rate": 1.3253612142811138e-05, + "loss": 0.8739, + "step": 31539 + }, + { + "epoch": 2.03, + "learning_rate": 1.3253397453788189e-05, + "loss": 0.8051, + "step": 31540 + }, + { + "epoch": 2.03, + "learning_rate": 1.3253182764765238e-05, + "loss": 0.8739, + "step": 31541 + }, + { + "epoch": 2.03, + "learning_rate": 1.3252968075742287e-05, + "loss": 0.8772, + "step": 31542 + }, + { + "epoch": 2.03, + "learning_rate": 1.3252753386719339e-05, + "loss": 0.8717, + "step": 31543 + }, + { + "epoch": 2.03, + "learning_rate": 1.3252538697696388e-05, + "loss": 0.8267, + "step": 31544 + }, + { + "epoch": 2.03, + "learning_rate": 1.3252324008673439e-05, + "loss": 0.901, + "step": 31545 + }, + { + "epoch": 2.03, + "learning_rate": 1.3252109319650488e-05, + "loss": 0.871, + "step": 31546 + }, + { + "epoch": 2.03, + "learning_rate": 1.3251894630627538e-05, + "loss": 0.8286, + "step": 31547 + }, + { + "epoch": 2.03, + "learning_rate": 1.3251679941604587e-05, + "loss": 0.9111, + "step": 31548 + }, + { + "epoch": 2.03, + "learning_rate": 1.3251465252581636e-05, + "loss": 0.9204, + "step": 31549 + }, + { + "epoch": 2.03, + "learning_rate": 1.3251250563558687e-05, + "loss": 0.9133, + "step": 31550 + }, + { + "epoch": 2.03, + "learning_rate": 1.3251035874535736e-05, + "loss": 0.8528, + "step": 31551 + }, + { + "epoch": 2.03, + "learning_rate": 1.3250821185512785e-05, + "loss": 0.9532, + "step": 31552 + }, + { + "epoch": 2.03, + "learning_rate": 1.3250606496489835e-05, + "loss": 0.8509, + "step": 31553 + }, + { + "epoch": 2.03, + "learning_rate": 1.3250391807466884e-05, + "loss": 0.8966, + "step": 31554 + }, + { + "epoch": 2.03, + "learning_rate": 1.3250177118443935e-05, + "loss": 0.8982, + "step": 31555 + }, + { + "epoch": 2.03, + "learning_rate": 1.3249962429420984e-05, + "loss": 0.9213, + "step": 31556 + }, + { + "epoch": 2.03, + "learning_rate": 1.3249747740398033e-05, + "loss": 0.8392, + "step": 31557 + }, + { + "epoch": 2.03, + "learning_rate": 1.3249533051375085e-05, + "loss": 0.9396, + "step": 31558 + }, + { + "epoch": 2.03, + "learning_rate": 1.3249318362352134e-05, + "loss": 0.9189, + "step": 31559 + }, + { + "epoch": 2.03, + "learning_rate": 1.3249103673329185e-05, + "loss": 0.9279, + "step": 31560 + }, + { + "epoch": 2.03, + "learning_rate": 1.3248888984306234e-05, + "loss": 0.8824, + "step": 31561 + }, + { + "epoch": 2.03, + "learning_rate": 1.3248674295283283e-05, + "loss": 0.7958, + "step": 31562 + }, + { + "epoch": 2.03, + "learning_rate": 1.3248459606260334e-05, + "loss": 0.903, + "step": 31563 + }, + { + "epoch": 2.03, + "learning_rate": 1.3248244917237383e-05, + "loss": 0.8911, + "step": 31564 + }, + { + "epoch": 2.03, + "learning_rate": 1.3248030228214433e-05, + "loss": 0.8775, + "step": 31565 + }, + { + "epoch": 2.03, + "learning_rate": 1.3247815539191482e-05, + "loss": 0.8385, + "step": 31566 + }, + { + "epoch": 2.03, + "learning_rate": 1.3247600850168531e-05, + "loss": 0.956, + "step": 31567 + }, + { + "epoch": 2.03, + "learning_rate": 1.3247386161145582e-05, + "loss": 0.9001, + "step": 31568 + }, + { + "epoch": 2.03, + "learning_rate": 1.324717147212263e-05, + "loss": 0.9688, + "step": 31569 + }, + { + "epoch": 2.03, + "learning_rate": 1.324695678309968e-05, + "loss": 0.911, + "step": 31570 + }, + { + "epoch": 2.03, + "learning_rate": 1.324674209407673e-05, + "loss": 0.9628, + "step": 31571 + }, + { + "epoch": 2.03, + "learning_rate": 1.3246527405053781e-05, + "loss": 0.9697, + "step": 31572 + }, + { + "epoch": 2.03, + "learning_rate": 1.3246312716030832e-05, + "loss": 0.9033, + "step": 31573 + }, + { + "epoch": 2.03, + "learning_rate": 1.324609802700788e-05, + "loss": 0.9556, + "step": 31574 + }, + { + "epoch": 2.03, + "learning_rate": 1.3245883337984931e-05, + "loss": 0.884, + "step": 31575 + }, + { + "epoch": 2.03, + "learning_rate": 1.324566864896198e-05, + "loss": 0.8086, + "step": 31576 + }, + { + "epoch": 2.03, + "learning_rate": 1.324545395993903e-05, + "loss": 0.939, + "step": 31577 + }, + { + "epoch": 2.03, + "learning_rate": 1.324523927091608e-05, + "loss": 0.8701, + "step": 31578 + }, + { + "epoch": 2.03, + "learning_rate": 1.3245024581893129e-05, + "loss": 0.8927, + "step": 31579 + }, + { + "epoch": 2.03, + "learning_rate": 1.3244809892870178e-05, + "loss": 0.9544, + "step": 31580 + }, + { + "epoch": 2.03, + "learning_rate": 1.3244595203847228e-05, + "loss": 0.9759, + "step": 31581 + }, + { + "epoch": 2.03, + "learning_rate": 1.3244380514824277e-05, + "loss": 0.8678, + "step": 31582 + }, + { + "epoch": 2.03, + "learning_rate": 1.3244165825801328e-05, + "loss": 0.9023, + "step": 31583 + }, + { + "epoch": 2.03, + "learning_rate": 1.3243951136778377e-05, + "loss": 0.9671, + "step": 31584 + }, + { + "epoch": 2.03, + "learning_rate": 1.3243736447755426e-05, + "loss": 0.8996, + "step": 31585 + }, + { + "epoch": 2.03, + "learning_rate": 1.3243521758732477e-05, + "loss": 0.8881, + "step": 31586 + }, + { + "epoch": 2.03, + "learning_rate": 1.3243307069709527e-05, + "loss": 0.8913, + "step": 31587 + }, + { + "epoch": 2.03, + "learning_rate": 1.3243092380686578e-05, + "loss": 0.8694, + "step": 31588 + }, + { + "epoch": 2.03, + "learning_rate": 1.3242877691663627e-05, + "loss": 0.8106, + "step": 31589 + }, + { + "epoch": 2.03, + "learning_rate": 1.3242663002640676e-05, + "loss": 0.9325, + "step": 31590 + }, + { + "epoch": 2.03, + "learning_rate": 1.3242448313617727e-05, + "loss": 0.8973, + "step": 31591 + }, + { + "epoch": 2.03, + "learning_rate": 1.3242233624594776e-05, + "loss": 0.8919, + "step": 31592 + }, + { + "epoch": 2.03, + "learning_rate": 1.3242018935571826e-05, + "loss": 0.8577, + "step": 31593 + }, + { + "epoch": 2.03, + "learning_rate": 1.3241804246548875e-05, + "loss": 0.9163, + "step": 31594 + }, + { + "epoch": 2.03, + "learning_rate": 1.3241589557525924e-05, + "loss": 0.9182, + "step": 31595 + }, + { + "epoch": 2.03, + "learning_rate": 1.3241374868502975e-05, + "loss": 0.8772, + "step": 31596 + }, + { + "epoch": 2.03, + "learning_rate": 1.3241160179480024e-05, + "loss": 0.9094, + "step": 31597 + }, + { + "epoch": 2.03, + "learning_rate": 1.3240945490457073e-05, + "loss": 0.8696, + "step": 31598 + }, + { + "epoch": 2.03, + "learning_rate": 1.3240730801434123e-05, + "loss": 0.9482, + "step": 31599 + }, + { + "epoch": 2.03, + "learning_rate": 1.3240516112411172e-05, + "loss": 0.8837, + "step": 31600 + }, + { + "epoch": 2.03, + "learning_rate": 1.3240301423388225e-05, + "loss": 0.9168, + "step": 31601 + }, + { + "epoch": 2.03, + "learning_rate": 1.3240086734365274e-05, + "loss": 0.864, + "step": 31602 + }, + { + "epoch": 2.03, + "learning_rate": 1.3239872045342324e-05, + "loss": 0.9251, + "step": 31603 + }, + { + "epoch": 2.03, + "learning_rate": 1.3239657356319373e-05, + "loss": 0.8725, + "step": 31604 + }, + { + "epoch": 2.03, + "learning_rate": 1.3239442667296422e-05, + "loss": 0.9135, + "step": 31605 + }, + { + "epoch": 2.03, + "learning_rate": 1.3239227978273473e-05, + "loss": 0.9307, + "step": 31606 + }, + { + "epoch": 2.03, + "learning_rate": 1.3239013289250522e-05, + "loss": 0.8506, + "step": 31607 + }, + { + "epoch": 2.03, + "learning_rate": 1.3238798600227571e-05, + "loss": 1.0007, + "step": 31608 + }, + { + "epoch": 2.03, + "learning_rate": 1.3238583911204622e-05, + "loss": 0.9411, + "step": 31609 + }, + { + "epoch": 2.03, + "learning_rate": 1.323836922218167e-05, + "loss": 0.9389, + "step": 31610 + }, + { + "epoch": 2.03, + "learning_rate": 1.323815453315872e-05, + "loss": 0.9012, + "step": 31611 + }, + { + "epoch": 2.03, + "learning_rate": 1.323793984413577e-05, + "loss": 0.8465, + "step": 31612 + }, + { + "epoch": 2.03, + "learning_rate": 1.3237725155112819e-05, + "loss": 0.9051, + "step": 31613 + }, + { + "epoch": 2.03, + "learning_rate": 1.323751046608987e-05, + "loss": 0.9323, + "step": 31614 + }, + { + "epoch": 2.03, + "learning_rate": 1.3237295777066919e-05, + "loss": 0.8943, + "step": 31615 + }, + { + "epoch": 2.03, + "learning_rate": 1.3237081088043971e-05, + "loss": 0.8023, + "step": 31616 + }, + { + "epoch": 2.03, + "learning_rate": 1.323686639902102e-05, + "loss": 0.871, + "step": 31617 + }, + { + "epoch": 2.03, + "learning_rate": 1.3236651709998069e-05, + "loss": 0.918, + "step": 31618 + }, + { + "epoch": 2.03, + "learning_rate": 1.323643702097512e-05, + "loss": 0.96, + "step": 31619 + }, + { + "epoch": 2.03, + "learning_rate": 1.3236222331952169e-05, + "loss": 0.9232, + "step": 31620 + }, + { + "epoch": 2.03, + "learning_rate": 1.3236007642929218e-05, + "loss": 0.8297, + "step": 31621 + }, + { + "epoch": 2.03, + "learning_rate": 1.3235792953906268e-05, + "loss": 0.8763, + "step": 31622 + }, + { + "epoch": 2.03, + "learning_rate": 1.3235578264883317e-05, + "loss": 0.8796, + "step": 31623 + }, + { + "epoch": 2.03, + "learning_rate": 1.3235363575860368e-05, + "loss": 0.956, + "step": 31624 + }, + { + "epoch": 2.03, + "learning_rate": 1.3235148886837417e-05, + "loss": 0.8767, + "step": 31625 + }, + { + "epoch": 2.03, + "learning_rate": 1.3234934197814466e-05, + "loss": 0.9352, + "step": 31626 + }, + { + "epoch": 2.03, + "learning_rate": 1.3234719508791516e-05, + "loss": 0.8733, + "step": 31627 + }, + { + "epoch": 2.03, + "learning_rate": 1.3234504819768565e-05, + "loss": 0.9182, + "step": 31628 + }, + { + "epoch": 2.03, + "learning_rate": 1.3234290130745614e-05, + "loss": 0.823, + "step": 31629 + }, + { + "epoch": 2.03, + "learning_rate": 1.3234075441722667e-05, + "loss": 0.8149, + "step": 31630 + }, + { + "epoch": 2.04, + "learning_rate": 1.3233860752699716e-05, + "loss": 0.8283, + "step": 31631 + }, + { + "epoch": 2.04, + "learning_rate": 1.3233646063676766e-05, + "loss": 0.9209, + "step": 31632 + }, + { + "epoch": 2.04, + "learning_rate": 1.3233431374653815e-05, + "loss": 0.9112, + "step": 31633 + }, + { + "epoch": 2.04, + "learning_rate": 1.3233216685630866e-05, + "loss": 0.9121, + "step": 31634 + }, + { + "epoch": 2.04, + "learning_rate": 1.3233001996607915e-05, + "loss": 0.9095, + "step": 31635 + }, + { + "epoch": 2.04, + "learning_rate": 1.3232787307584964e-05, + "loss": 0.9539, + "step": 31636 + }, + { + "epoch": 2.04, + "learning_rate": 1.3232572618562015e-05, + "loss": 0.8684, + "step": 31637 + }, + { + "epoch": 2.04, + "learning_rate": 1.3232357929539064e-05, + "loss": 0.8765, + "step": 31638 + }, + { + "epoch": 2.04, + "learning_rate": 1.3232143240516112e-05, + "loss": 0.9008, + "step": 31639 + }, + { + "epoch": 2.04, + "learning_rate": 1.3231928551493163e-05, + "loss": 0.9088, + "step": 31640 + }, + { + "epoch": 2.04, + "learning_rate": 1.3231713862470212e-05, + "loss": 0.8228, + "step": 31641 + }, + { + "epoch": 2.04, + "learning_rate": 1.3231499173447263e-05, + "loss": 0.9257, + "step": 31642 + }, + { + "epoch": 2.04, + "learning_rate": 1.3231284484424312e-05, + "loss": 0.8704, + "step": 31643 + }, + { + "epoch": 2.04, + "learning_rate": 1.323106979540136e-05, + "loss": 0.9198, + "step": 31644 + }, + { + "epoch": 2.04, + "learning_rate": 1.3230855106378413e-05, + "loss": 0.9225, + "step": 31645 + }, + { + "epoch": 2.04, + "learning_rate": 1.3230640417355462e-05, + "loss": 0.8779, + "step": 31646 + }, + { + "epoch": 2.04, + "learning_rate": 1.3230425728332513e-05, + "loss": 0.8992, + "step": 31647 + }, + { + "epoch": 2.04, + "learning_rate": 1.3230211039309562e-05, + "loss": 0.9168, + "step": 31648 + }, + { + "epoch": 2.04, + "learning_rate": 1.322999635028661e-05, + "loss": 0.8795, + "step": 31649 + }, + { + "epoch": 2.04, + "learning_rate": 1.3229781661263661e-05, + "loss": 0.903, + "step": 31650 + }, + { + "epoch": 2.04, + "learning_rate": 1.322956697224071e-05, + "loss": 0.8633, + "step": 31651 + }, + { + "epoch": 2.04, + "learning_rate": 1.3229352283217761e-05, + "loss": 0.9194, + "step": 31652 + }, + { + "epoch": 2.04, + "learning_rate": 1.322913759419481e-05, + "loss": 0.8822, + "step": 31653 + }, + { + "epoch": 2.04, + "learning_rate": 1.3228922905171859e-05, + "loss": 0.9073, + "step": 31654 + }, + { + "epoch": 2.04, + "learning_rate": 1.322870821614891e-05, + "loss": 0.9077, + "step": 31655 + }, + { + "epoch": 2.04, + "learning_rate": 1.3228493527125958e-05, + "loss": 0.9453, + "step": 31656 + }, + { + "epoch": 2.04, + "learning_rate": 1.3228278838103007e-05, + "loss": 0.881, + "step": 31657 + }, + { + "epoch": 2.04, + "learning_rate": 1.3228064149080058e-05, + "loss": 0.9157, + "step": 31658 + }, + { + "epoch": 2.04, + "learning_rate": 1.3227849460057109e-05, + "loss": 0.9481, + "step": 31659 + }, + { + "epoch": 2.04, + "learning_rate": 1.322763477103416e-05, + "loss": 0.8813, + "step": 31660 + }, + { + "epoch": 2.04, + "learning_rate": 1.3227420082011208e-05, + "loss": 0.8977, + "step": 31661 + }, + { + "epoch": 2.04, + "learning_rate": 1.3227205392988259e-05, + "loss": 0.9011, + "step": 31662 + }, + { + "epoch": 2.04, + "learning_rate": 1.3226990703965308e-05, + "loss": 0.9176, + "step": 31663 + }, + { + "epoch": 2.04, + "learning_rate": 1.3226776014942357e-05, + "loss": 0.8929, + "step": 31664 + }, + { + "epoch": 2.04, + "learning_rate": 1.3226561325919408e-05, + "loss": 0.9206, + "step": 31665 + }, + { + "epoch": 2.04, + "learning_rate": 1.3226346636896457e-05, + "loss": 0.8867, + "step": 31666 + }, + { + "epoch": 2.04, + "learning_rate": 1.3226131947873506e-05, + "loss": 1.0058, + "step": 31667 + }, + { + "epoch": 2.04, + "learning_rate": 1.3225917258850556e-05, + "loss": 0.8539, + "step": 31668 + }, + { + "epoch": 2.04, + "learning_rate": 1.3225702569827605e-05, + "loss": 0.9111, + "step": 31669 + }, + { + "epoch": 2.04, + "learning_rate": 1.3225487880804654e-05, + "loss": 0.8804, + "step": 31670 + }, + { + "epoch": 2.04, + "learning_rate": 1.3225273191781705e-05, + "loss": 0.9173, + "step": 31671 + }, + { + "epoch": 2.04, + "learning_rate": 1.3225058502758754e-05, + "loss": 0.9062, + "step": 31672 + }, + { + "epoch": 2.04, + "learning_rate": 1.3224843813735804e-05, + "loss": 0.8848, + "step": 31673 + }, + { + "epoch": 2.04, + "learning_rate": 1.3224629124712855e-05, + "loss": 0.9234, + "step": 31674 + }, + { + "epoch": 2.04, + "learning_rate": 1.3224414435689906e-05, + "loss": 0.9156, + "step": 31675 + }, + { + "epoch": 2.04, + "learning_rate": 1.3224199746666955e-05, + "loss": 0.8631, + "step": 31676 + }, + { + "epoch": 2.04, + "learning_rate": 1.3223985057644004e-05, + "loss": 0.9204, + "step": 31677 + }, + { + "epoch": 2.04, + "learning_rate": 1.3223770368621054e-05, + "loss": 0.9384, + "step": 31678 + }, + { + "epoch": 2.04, + "learning_rate": 1.3223555679598103e-05, + "loss": 0.8599, + "step": 31679 + }, + { + "epoch": 2.04, + "learning_rate": 1.3223340990575154e-05, + "loss": 0.9566, + "step": 31680 + }, + { + "epoch": 2.04, + "learning_rate": 1.3223126301552203e-05, + "loss": 0.9494, + "step": 31681 + }, + { + "epoch": 2.04, + "learning_rate": 1.3222911612529252e-05, + "loss": 0.8881, + "step": 31682 + }, + { + "epoch": 2.04, + "learning_rate": 1.3222696923506303e-05, + "loss": 0.885, + "step": 31683 + }, + { + "epoch": 2.04, + "learning_rate": 1.3222482234483351e-05, + "loss": 0.8889, + "step": 31684 + }, + { + "epoch": 2.04, + "learning_rate": 1.32222675454604e-05, + "loss": 0.9148, + "step": 31685 + }, + { + "epoch": 2.04, + "learning_rate": 1.3222052856437451e-05, + "loss": 0.8862, + "step": 31686 + }, + { + "epoch": 2.04, + "learning_rate": 1.32218381674145e-05, + "loss": 0.8718, + "step": 31687 + }, + { + "epoch": 2.04, + "learning_rate": 1.3221623478391552e-05, + "loss": 0.9536, + "step": 31688 + }, + { + "epoch": 2.04, + "learning_rate": 1.3221408789368601e-05, + "loss": 0.8102, + "step": 31689 + }, + { + "epoch": 2.04, + "learning_rate": 1.3221194100345652e-05, + "loss": 0.905, + "step": 31690 + }, + { + "epoch": 2.04, + "learning_rate": 1.3220979411322701e-05, + "loss": 0.9066, + "step": 31691 + }, + { + "epoch": 2.04, + "learning_rate": 1.322076472229975e-05, + "loss": 0.8591, + "step": 31692 + }, + { + "epoch": 2.04, + "learning_rate": 1.32205500332768e-05, + "loss": 0.9067, + "step": 31693 + }, + { + "epoch": 2.04, + "learning_rate": 1.322033534425385e-05, + "loss": 0.8887, + "step": 31694 + }, + { + "epoch": 2.04, + "learning_rate": 1.3220120655230899e-05, + "loss": 0.9359, + "step": 31695 + }, + { + "epoch": 2.04, + "learning_rate": 1.321990596620795e-05, + "loss": 0.9004, + "step": 31696 + }, + { + "epoch": 2.04, + "learning_rate": 1.3219691277184998e-05, + "loss": 0.8817, + "step": 31697 + }, + { + "epoch": 2.04, + "learning_rate": 1.3219476588162047e-05, + "loss": 0.9191, + "step": 31698 + }, + { + "epoch": 2.04, + "learning_rate": 1.3219261899139098e-05, + "loss": 0.9942, + "step": 31699 + }, + { + "epoch": 2.04, + "learning_rate": 1.3219047210116147e-05, + "loss": 0.9178, + "step": 31700 + }, + { + "epoch": 2.04, + "learning_rate": 1.3218832521093197e-05, + "loss": 0.8904, + "step": 31701 + }, + { + "epoch": 2.04, + "learning_rate": 1.3218617832070246e-05, + "loss": 0.9085, + "step": 31702 + }, + { + "epoch": 2.04, + "learning_rate": 1.3218403143047299e-05, + "loss": 0.9121, + "step": 31703 + }, + { + "epoch": 2.04, + "learning_rate": 1.3218188454024348e-05, + "loss": 0.909, + "step": 31704 + }, + { + "epoch": 2.04, + "learning_rate": 1.3217973765001397e-05, + "loss": 0.8541, + "step": 31705 + }, + { + "epoch": 2.04, + "learning_rate": 1.3217759075978447e-05, + "loss": 0.856, + "step": 31706 + }, + { + "epoch": 2.04, + "learning_rate": 1.3217544386955496e-05, + "loss": 0.8963, + "step": 31707 + }, + { + "epoch": 2.04, + "learning_rate": 1.3217329697932545e-05, + "loss": 0.9099, + "step": 31708 + }, + { + "epoch": 2.04, + "learning_rate": 1.3217115008909596e-05, + "loss": 0.9107, + "step": 31709 + }, + { + "epoch": 2.04, + "learning_rate": 1.3216900319886645e-05, + "loss": 0.9247, + "step": 31710 + }, + { + "epoch": 2.04, + "learning_rate": 1.3216685630863696e-05, + "loss": 0.8545, + "step": 31711 + }, + { + "epoch": 2.04, + "learning_rate": 1.3216470941840745e-05, + "loss": 0.8915, + "step": 31712 + }, + { + "epoch": 2.04, + "learning_rate": 1.3216256252817793e-05, + "loss": 0.9065, + "step": 31713 + }, + { + "epoch": 2.04, + "learning_rate": 1.3216041563794844e-05, + "loss": 0.9085, + "step": 31714 + }, + { + "epoch": 2.04, + "learning_rate": 1.3215826874771893e-05, + "loss": 0.9455, + "step": 31715 + }, + { + "epoch": 2.04, + "learning_rate": 1.3215612185748942e-05, + "loss": 0.8378, + "step": 31716 + }, + { + "epoch": 2.04, + "learning_rate": 1.3215397496725994e-05, + "loss": 0.8927, + "step": 31717 + }, + { + "epoch": 2.04, + "learning_rate": 1.3215182807703043e-05, + "loss": 0.8771, + "step": 31718 + }, + { + "epoch": 2.04, + "learning_rate": 1.3214968118680094e-05, + "loss": 0.9451, + "step": 31719 + }, + { + "epoch": 2.04, + "learning_rate": 1.3214753429657143e-05, + "loss": 0.8801, + "step": 31720 + }, + { + "epoch": 2.04, + "learning_rate": 1.3214538740634194e-05, + "loss": 0.872, + "step": 31721 + }, + { + "epoch": 2.04, + "learning_rate": 1.3214324051611243e-05, + "loss": 0.9129, + "step": 31722 + }, + { + "epoch": 2.04, + "learning_rate": 1.3214109362588292e-05, + "loss": 0.9085, + "step": 31723 + }, + { + "epoch": 2.04, + "learning_rate": 1.3213894673565342e-05, + "loss": 0.9048, + "step": 31724 + }, + { + "epoch": 2.04, + "learning_rate": 1.3213679984542391e-05, + "loss": 0.8748, + "step": 31725 + }, + { + "epoch": 2.04, + "learning_rate": 1.321346529551944e-05, + "loss": 0.9181, + "step": 31726 + }, + { + "epoch": 2.04, + "learning_rate": 1.3213250606496491e-05, + "loss": 0.8283, + "step": 31727 + }, + { + "epoch": 2.04, + "learning_rate": 1.321303591747354e-05, + "loss": 0.841, + "step": 31728 + }, + { + "epoch": 2.04, + "learning_rate": 1.321282122845059e-05, + "loss": 0.9265, + "step": 31729 + }, + { + "epoch": 2.04, + "learning_rate": 1.321260653942764e-05, + "loss": 0.8684, + "step": 31730 + }, + { + "epoch": 2.04, + "learning_rate": 1.3212391850404688e-05, + "loss": 0.9016, + "step": 31731 + }, + { + "epoch": 2.04, + "learning_rate": 1.321217716138174e-05, + "loss": 0.8194, + "step": 31732 + }, + { + "epoch": 2.04, + "learning_rate": 1.321196247235879e-05, + "loss": 0.8418, + "step": 31733 + }, + { + "epoch": 2.04, + "learning_rate": 1.321174778333584e-05, + "loss": 0.9398, + "step": 31734 + }, + { + "epoch": 2.04, + "learning_rate": 1.321153309431289e-05, + "loss": 0.867, + "step": 31735 + }, + { + "epoch": 2.04, + "learning_rate": 1.3211318405289938e-05, + "loss": 0.8779, + "step": 31736 + }, + { + "epoch": 2.04, + "learning_rate": 1.3211103716266989e-05, + "loss": 0.9135, + "step": 31737 + }, + { + "epoch": 2.04, + "learning_rate": 1.3210889027244038e-05, + "loss": 0.892, + "step": 31738 + }, + { + "epoch": 2.04, + "learning_rate": 1.3210674338221089e-05, + "loss": 0.9449, + "step": 31739 + }, + { + "epoch": 2.04, + "learning_rate": 1.3210459649198138e-05, + "loss": 0.8933, + "step": 31740 + }, + { + "epoch": 2.04, + "learning_rate": 1.3210244960175187e-05, + "loss": 0.9123, + "step": 31741 + }, + { + "epoch": 2.04, + "learning_rate": 1.3210030271152237e-05, + "loss": 0.9216, + "step": 31742 + }, + { + "epoch": 2.04, + "learning_rate": 1.3209815582129286e-05, + "loss": 0.8432, + "step": 31743 + }, + { + "epoch": 2.04, + "learning_rate": 1.3209600893106335e-05, + "loss": 0.9027, + "step": 31744 + }, + { + "epoch": 2.04, + "learning_rate": 1.3209386204083386e-05, + "loss": 0.8221, + "step": 31745 + }, + { + "epoch": 2.04, + "learning_rate": 1.3209171515060435e-05, + "loss": 0.8718, + "step": 31746 + }, + { + "epoch": 2.04, + "learning_rate": 1.3208956826037487e-05, + "loss": 0.846, + "step": 31747 + }, + { + "epoch": 2.04, + "learning_rate": 1.3208742137014536e-05, + "loss": 0.8989, + "step": 31748 + }, + { + "epoch": 2.04, + "learning_rate": 1.3208527447991587e-05, + "loss": 0.8944, + "step": 31749 + }, + { + "epoch": 2.04, + "learning_rate": 1.3208312758968636e-05, + "loss": 0.9585, + "step": 31750 + }, + { + "epoch": 2.04, + "learning_rate": 1.3208098069945685e-05, + "loss": 0.8805, + "step": 31751 + }, + { + "epoch": 2.04, + "learning_rate": 1.3207883380922735e-05, + "loss": 0.8473, + "step": 31752 + }, + { + "epoch": 2.04, + "learning_rate": 1.3207668691899784e-05, + "loss": 0.9299, + "step": 31753 + }, + { + "epoch": 2.04, + "learning_rate": 1.3207454002876833e-05, + "loss": 0.9294, + "step": 31754 + }, + { + "epoch": 2.04, + "learning_rate": 1.3207239313853884e-05, + "loss": 0.8581, + "step": 31755 + }, + { + "epoch": 2.04, + "learning_rate": 1.3207024624830933e-05, + "loss": 0.9372, + "step": 31756 + }, + { + "epoch": 2.04, + "learning_rate": 1.3206809935807982e-05, + "loss": 0.9433, + "step": 31757 + }, + { + "epoch": 2.04, + "learning_rate": 1.3206595246785032e-05, + "loss": 0.9175, + "step": 31758 + }, + { + "epoch": 2.04, + "learning_rate": 1.3206380557762081e-05, + "loss": 0.9096, + "step": 31759 + }, + { + "epoch": 2.04, + "learning_rate": 1.3206165868739132e-05, + "loss": 0.9253, + "step": 31760 + }, + { + "epoch": 2.04, + "learning_rate": 1.3205951179716183e-05, + "loss": 0.8847, + "step": 31761 + }, + { + "epoch": 2.04, + "learning_rate": 1.3205736490693233e-05, + "loss": 0.8458, + "step": 31762 + }, + { + "epoch": 2.04, + "learning_rate": 1.3205521801670282e-05, + "loss": 0.9014, + "step": 31763 + }, + { + "epoch": 2.04, + "learning_rate": 1.3205307112647331e-05, + "loss": 0.8975, + "step": 31764 + }, + { + "epoch": 2.04, + "learning_rate": 1.3205092423624382e-05, + "loss": 0.8933, + "step": 31765 + }, + { + "epoch": 2.04, + "learning_rate": 1.3204877734601431e-05, + "loss": 0.8964, + "step": 31766 + }, + { + "epoch": 2.04, + "learning_rate": 1.320466304557848e-05, + "loss": 0.8749, + "step": 31767 + }, + { + "epoch": 2.04, + "learning_rate": 1.320444835655553e-05, + "loss": 0.7928, + "step": 31768 + }, + { + "epoch": 2.04, + "learning_rate": 1.320423366753258e-05, + "loss": 0.8998, + "step": 31769 + }, + { + "epoch": 2.04, + "learning_rate": 1.320401897850963e-05, + "loss": 0.8909, + "step": 31770 + }, + { + "epoch": 2.04, + "learning_rate": 1.320380428948668e-05, + "loss": 0.8904, + "step": 31771 + }, + { + "epoch": 2.04, + "learning_rate": 1.3203589600463728e-05, + "loss": 0.9271, + "step": 31772 + }, + { + "epoch": 2.04, + "learning_rate": 1.3203374911440779e-05, + "loss": 0.8836, + "step": 31773 + }, + { + "epoch": 2.04, + "learning_rate": 1.3203160222417828e-05, + "loss": 0.8607, + "step": 31774 + }, + { + "epoch": 2.04, + "learning_rate": 1.3202945533394877e-05, + "loss": 0.906, + "step": 31775 + }, + { + "epoch": 2.04, + "learning_rate": 1.3202730844371929e-05, + "loss": 0.8451, + "step": 31776 + }, + { + "epoch": 2.04, + "learning_rate": 1.3202516155348978e-05, + "loss": 0.9352, + "step": 31777 + }, + { + "epoch": 2.04, + "learning_rate": 1.3202301466326029e-05, + "loss": 0.8706, + "step": 31778 + }, + { + "epoch": 2.04, + "learning_rate": 1.3202086777303078e-05, + "loss": 0.9047, + "step": 31779 + }, + { + "epoch": 2.04, + "learning_rate": 1.3201872088280128e-05, + "loss": 0.859, + "step": 31780 + }, + { + "epoch": 2.04, + "learning_rate": 1.3201657399257177e-05, + "loss": 0.845, + "step": 31781 + }, + { + "epoch": 2.04, + "learning_rate": 1.3201442710234226e-05, + "loss": 0.8822, + "step": 31782 + }, + { + "epoch": 2.04, + "learning_rate": 1.3201228021211277e-05, + "loss": 0.8955, + "step": 31783 + }, + { + "epoch": 2.04, + "learning_rate": 1.3201013332188326e-05, + "loss": 0.9397, + "step": 31784 + }, + { + "epoch": 2.04, + "learning_rate": 1.3200798643165375e-05, + "loss": 0.9266, + "step": 31785 + }, + { + "epoch": 2.05, + "learning_rate": 1.3200583954142426e-05, + "loss": 0.8793, + "step": 31786 + }, + { + "epoch": 2.05, + "learning_rate": 1.3200369265119474e-05, + "loss": 0.8316, + "step": 31787 + }, + { + "epoch": 2.05, + "learning_rate": 1.3200154576096525e-05, + "loss": 0.9281, + "step": 31788 + }, + { + "epoch": 2.05, + "learning_rate": 1.3199939887073574e-05, + "loss": 0.8914, + "step": 31789 + }, + { + "epoch": 2.05, + "learning_rate": 1.3199725198050626e-05, + "loss": 0.9232, + "step": 31790 + }, + { + "epoch": 2.05, + "learning_rate": 1.3199510509027675e-05, + "loss": 0.8686, + "step": 31791 + }, + { + "epoch": 2.05, + "learning_rate": 1.3199295820004724e-05, + "loss": 0.9134, + "step": 31792 + }, + { + "epoch": 2.05, + "learning_rate": 1.3199081130981775e-05, + "loss": 0.8509, + "step": 31793 + }, + { + "epoch": 2.05, + "learning_rate": 1.3198866441958824e-05, + "loss": 0.9612, + "step": 31794 + }, + { + "epoch": 2.05, + "learning_rate": 1.3198651752935873e-05, + "loss": 0.8545, + "step": 31795 + }, + { + "epoch": 2.05, + "learning_rate": 1.3198437063912924e-05, + "loss": 0.8694, + "step": 31796 + }, + { + "epoch": 2.05, + "learning_rate": 1.3198222374889973e-05, + "loss": 0.8541, + "step": 31797 + }, + { + "epoch": 2.05, + "learning_rate": 1.3198007685867023e-05, + "loss": 0.9256, + "step": 31798 + }, + { + "epoch": 2.05, + "learning_rate": 1.3197792996844072e-05, + "loss": 0.9137, + "step": 31799 + }, + { + "epoch": 2.05, + "learning_rate": 1.3197578307821121e-05, + "loss": 0.924, + "step": 31800 + }, + { + "epoch": 2.05, + "learning_rate": 1.3197363618798172e-05, + "loss": 0.8702, + "step": 31801 + }, + { + "epoch": 2.05, + "learning_rate": 1.319714892977522e-05, + "loss": 0.9133, + "step": 31802 + }, + { + "epoch": 2.05, + "learning_rate": 1.319693424075227e-05, + "loss": 0.886, + "step": 31803 + }, + { + "epoch": 2.05, + "learning_rate": 1.319671955172932e-05, + "loss": 0.9311, + "step": 31804 + }, + { + "epoch": 2.05, + "learning_rate": 1.3196504862706371e-05, + "loss": 0.8708, + "step": 31805 + }, + { + "epoch": 2.05, + "learning_rate": 1.3196290173683422e-05, + "loss": 0.9656, + "step": 31806 + }, + { + "epoch": 2.05, + "learning_rate": 1.319607548466047e-05, + "loss": 0.8709, + "step": 31807 + }, + { + "epoch": 2.05, + "learning_rate": 1.3195860795637521e-05, + "loss": 0.8797, + "step": 31808 + }, + { + "epoch": 2.05, + "learning_rate": 1.319564610661457e-05, + "loss": 0.8902, + "step": 31809 + }, + { + "epoch": 2.05, + "learning_rate": 1.319543141759162e-05, + "loss": 0.9607, + "step": 31810 + }, + { + "epoch": 2.05, + "learning_rate": 1.319521672856867e-05, + "loss": 0.8752, + "step": 31811 + }, + { + "epoch": 2.05, + "learning_rate": 1.3195002039545719e-05, + "loss": 0.8904, + "step": 31812 + }, + { + "epoch": 2.05, + "learning_rate": 1.3194787350522768e-05, + "loss": 0.8634, + "step": 31813 + }, + { + "epoch": 2.05, + "learning_rate": 1.3194572661499819e-05, + "loss": 0.9263, + "step": 31814 + }, + { + "epoch": 2.05, + "learning_rate": 1.3194357972476868e-05, + "loss": 0.8464, + "step": 31815 + }, + { + "epoch": 2.05, + "learning_rate": 1.3194143283453918e-05, + "loss": 0.8519, + "step": 31816 + }, + { + "epoch": 2.05, + "learning_rate": 1.3193928594430967e-05, + "loss": 0.9179, + "step": 31817 + }, + { + "epoch": 2.05, + "learning_rate": 1.3193713905408016e-05, + "loss": 0.86, + "step": 31818 + }, + { + "epoch": 2.05, + "learning_rate": 1.3193499216385068e-05, + "loss": 0.8699, + "step": 31819 + }, + { + "epoch": 2.05, + "learning_rate": 1.3193284527362117e-05, + "loss": 0.8681, + "step": 31820 + }, + { + "epoch": 2.05, + "learning_rate": 1.3193069838339168e-05, + "loss": 0.8921, + "step": 31821 + }, + { + "epoch": 2.05, + "learning_rate": 1.3192855149316217e-05, + "loss": 0.8877, + "step": 31822 + }, + { + "epoch": 2.05, + "learning_rate": 1.3192640460293266e-05, + "loss": 0.8066, + "step": 31823 + }, + { + "epoch": 2.05, + "learning_rate": 1.3192425771270317e-05, + "loss": 0.8928, + "step": 31824 + }, + { + "epoch": 2.05, + "learning_rate": 1.3192211082247366e-05, + "loss": 0.8854, + "step": 31825 + }, + { + "epoch": 2.05, + "learning_rate": 1.3191996393224416e-05, + "loss": 0.8466, + "step": 31826 + }, + { + "epoch": 2.05, + "learning_rate": 1.3191781704201465e-05, + "loss": 0.8769, + "step": 31827 + }, + { + "epoch": 2.05, + "learning_rate": 1.3191567015178514e-05, + "loss": 0.879, + "step": 31828 + }, + { + "epoch": 2.05, + "learning_rate": 1.3191352326155565e-05, + "loss": 0.9051, + "step": 31829 + }, + { + "epoch": 2.05, + "learning_rate": 1.3191137637132614e-05, + "loss": 0.908, + "step": 31830 + }, + { + "epoch": 2.05, + "learning_rate": 1.3190922948109663e-05, + "loss": 0.8297, + "step": 31831 + }, + { + "epoch": 2.05, + "learning_rate": 1.3190708259086713e-05, + "loss": 0.8863, + "step": 31832 + }, + { + "epoch": 2.05, + "learning_rate": 1.3190493570063762e-05, + "loss": 0.9184, + "step": 31833 + }, + { + "epoch": 2.05, + "learning_rate": 1.3190278881040815e-05, + "loss": 0.8537, + "step": 31834 + }, + { + "epoch": 2.05, + "learning_rate": 1.3190064192017864e-05, + "loss": 0.8745, + "step": 31835 + }, + { + "epoch": 2.05, + "learning_rate": 1.3189849502994914e-05, + "loss": 0.9207, + "step": 31836 + }, + { + "epoch": 2.05, + "learning_rate": 1.3189634813971963e-05, + "loss": 0.8339, + "step": 31837 + }, + { + "epoch": 2.05, + "learning_rate": 1.3189420124949012e-05, + "loss": 0.9177, + "step": 31838 + }, + { + "epoch": 2.05, + "learning_rate": 1.3189205435926063e-05, + "loss": 0.8956, + "step": 31839 + }, + { + "epoch": 2.05, + "learning_rate": 1.3188990746903112e-05, + "loss": 0.9351, + "step": 31840 + }, + { + "epoch": 2.05, + "learning_rate": 1.3188776057880161e-05, + "loss": 0.8508, + "step": 31841 + }, + { + "epoch": 2.05, + "learning_rate": 1.3188561368857212e-05, + "loss": 0.8819, + "step": 31842 + }, + { + "epoch": 2.05, + "learning_rate": 1.318834667983426e-05, + "loss": 0.8858, + "step": 31843 + }, + { + "epoch": 2.05, + "learning_rate": 1.318813199081131e-05, + "loss": 0.8153, + "step": 31844 + }, + { + "epoch": 2.05, + "learning_rate": 1.318791730178836e-05, + "loss": 0.997, + "step": 31845 + }, + { + "epoch": 2.05, + "learning_rate": 1.318770261276541e-05, + "loss": 0.8673, + "step": 31846 + }, + { + "epoch": 2.05, + "learning_rate": 1.318748792374246e-05, + "loss": 0.9016, + "step": 31847 + }, + { + "epoch": 2.05, + "learning_rate": 1.318727323471951e-05, + "loss": 0.8662, + "step": 31848 + }, + { + "epoch": 2.05, + "learning_rate": 1.3187058545696561e-05, + "loss": 0.9422, + "step": 31849 + }, + { + "epoch": 2.05, + "learning_rate": 1.318684385667361e-05, + "loss": 0.9116, + "step": 31850 + }, + { + "epoch": 2.05, + "learning_rate": 1.3186629167650659e-05, + "loss": 0.8474, + "step": 31851 + }, + { + "epoch": 2.05, + "learning_rate": 1.318641447862771e-05, + "loss": 0.8675, + "step": 31852 + }, + { + "epoch": 2.05, + "learning_rate": 1.3186199789604759e-05, + "loss": 0.8769, + "step": 31853 + }, + { + "epoch": 2.05, + "learning_rate": 1.3185985100581808e-05, + "loss": 0.925, + "step": 31854 + }, + { + "epoch": 2.05, + "learning_rate": 1.3185770411558858e-05, + "loss": 0.9411, + "step": 31855 + }, + { + "epoch": 2.05, + "learning_rate": 1.3185555722535907e-05, + "loss": 0.8715, + "step": 31856 + }, + { + "epoch": 2.05, + "learning_rate": 1.3185341033512958e-05, + "loss": 0.8568, + "step": 31857 + }, + { + "epoch": 2.05, + "learning_rate": 1.3185126344490007e-05, + "loss": 0.8789, + "step": 31858 + }, + { + "epoch": 2.05, + "learning_rate": 1.3184911655467056e-05, + "loss": 0.8951, + "step": 31859 + }, + { + "epoch": 2.05, + "learning_rate": 1.3184696966444107e-05, + "loss": 0.9801, + "step": 31860 + }, + { + "epoch": 2.05, + "learning_rate": 1.3184482277421155e-05, + "loss": 0.8799, + "step": 31861 + }, + { + "epoch": 2.05, + "learning_rate": 1.3184267588398204e-05, + "loss": 0.893, + "step": 31862 + }, + { + "epoch": 2.05, + "learning_rate": 1.3184052899375257e-05, + "loss": 0.9025, + "step": 31863 + }, + { + "epoch": 2.05, + "learning_rate": 1.3183838210352306e-05, + "loss": 0.9183, + "step": 31864 + }, + { + "epoch": 2.05, + "learning_rate": 1.3183623521329356e-05, + "loss": 0.8588, + "step": 31865 + }, + { + "epoch": 2.05, + "learning_rate": 1.3183408832306405e-05, + "loss": 0.9185, + "step": 31866 + }, + { + "epoch": 2.05, + "learning_rate": 1.3183194143283456e-05, + "loss": 0.9, + "step": 31867 + }, + { + "epoch": 2.05, + "learning_rate": 1.3182979454260505e-05, + "loss": 0.9759, + "step": 31868 + }, + { + "epoch": 2.05, + "learning_rate": 1.3182764765237554e-05, + "loss": 0.9013, + "step": 31869 + }, + { + "epoch": 2.05, + "learning_rate": 1.3182550076214605e-05, + "loss": 0.8602, + "step": 31870 + }, + { + "epoch": 2.05, + "learning_rate": 1.3182335387191654e-05, + "loss": 0.9411, + "step": 31871 + }, + { + "epoch": 2.05, + "learning_rate": 1.3182120698168703e-05, + "loss": 0.8496, + "step": 31872 + }, + { + "epoch": 2.05, + "learning_rate": 1.3181906009145753e-05, + "loss": 0.8683, + "step": 31873 + }, + { + "epoch": 2.05, + "learning_rate": 1.3181691320122802e-05, + "loss": 0.943, + "step": 31874 + }, + { + "epoch": 2.05, + "learning_rate": 1.3181476631099853e-05, + "loss": 0.8699, + "step": 31875 + }, + { + "epoch": 2.05, + "learning_rate": 1.3181261942076902e-05, + "loss": 0.8691, + "step": 31876 + }, + { + "epoch": 2.05, + "learning_rate": 1.3181047253053954e-05, + "loss": 0.8878, + "step": 31877 + }, + { + "epoch": 2.05, + "learning_rate": 1.3180832564031003e-05, + "loss": 0.8391, + "step": 31878 + }, + { + "epoch": 2.05, + "learning_rate": 1.3180617875008052e-05, + "loss": 0.895, + "step": 31879 + }, + { + "epoch": 2.05, + "learning_rate": 1.3180403185985103e-05, + "loss": 0.8914, + "step": 31880 + }, + { + "epoch": 2.05, + "learning_rate": 1.3180188496962152e-05, + "loss": 0.9224, + "step": 31881 + }, + { + "epoch": 2.05, + "learning_rate": 1.31799738079392e-05, + "loss": 0.8669, + "step": 31882 + }, + { + "epoch": 2.05, + "learning_rate": 1.3179759118916251e-05, + "loss": 0.8263, + "step": 31883 + }, + { + "epoch": 2.05, + "learning_rate": 1.31795444298933e-05, + "loss": 0.8885, + "step": 31884 + }, + { + "epoch": 2.05, + "learning_rate": 1.3179329740870351e-05, + "loss": 0.9523, + "step": 31885 + }, + { + "epoch": 2.05, + "learning_rate": 1.31791150518474e-05, + "loss": 0.9504, + "step": 31886 + }, + { + "epoch": 2.05, + "learning_rate": 1.3178900362824449e-05, + "loss": 0.8933, + "step": 31887 + }, + { + "epoch": 2.05, + "learning_rate": 1.31786856738015e-05, + "loss": 0.9167, + "step": 31888 + }, + { + "epoch": 2.05, + "learning_rate": 1.3178470984778549e-05, + "loss": 0.9031, + "step": 31889 + }, + { + "epoch": 2.05, + "learning_rate": 1.3178256295755597e-05, + "loss": 0.9323, + "step": 31890 + }, + { + "epoch": 2.05, + "learning_rate": 1.3178041606732648e-05, + "loss": 0.8933, + "step": 31891 + }, + { + "epoch": 2.05, + "learning_rate": 1.3177826917709699e-05, + "loss": 0.9896, + "step": 31892 + }, + { + "epoch": 2.05, + "learning_rate": 1.317761222868675e-05, + "loss": 0.7844, + "step": 31893 + }, + { + "epoch": 2.05, + "learning_rate": 1.3177397539663798e-05, + "loss": 0.8961, + "step": 31894 + }, + { + "epoch": 2.05, + "learning_rate": 1.3177182850640849e-05, + "loss": 0.8555, + "step": 31895 + }, + { + "epoch": 2.05, + "learning_rate": 1.3176968161617898e-05, + "loss": 0.8853, + "step": 31896 + }, + { + "epoch": 2.05, + "learning_rate": 1.3176753472594947e-05, + "loss": 0.8499, + "step": 31897 + }, + { + "epoch": 2.05, + "learning_rate": 1.3176538783571998e-05, + "loss": 0.8826, + "step": 31898 + }, + { + "epoch": 2.05, + "learning_rate": 1.3176324094549047e-05, + "loss": 0.8861, + "step": 31899 + }, + { + "epoch": 2.05, + "learning_rate": 1.3176109405526096e-05, + "loss": 0.8617, + "step": 31900 + }, + { + "epoch": 2.05, + "learning_rate": 1.3175894716503146e-05, + "loss": 0.8982, + "step": 31901 + }, + { + "epoch": 2.05, + "learning_rate": 1.3175680027480195e-05, + "loss": 0.8638, + "step": 31902 + }, + { + "epoch": 2.05, + "learning_rate": 1.3175465338457246e-05, + "loss": 0.9504, + "step": 31903 + }, + { + "epoch": 2.05, + "learning_rate": 1.3175250649434295e-05, + "loss": 0.8781, + "step": 31904 + }, + { + "epoch": 2.05, + "learning_rate": 1.3175035960411344e-05, + "loss": 0.8842, + "step": 31905 + }, + { + "epoch": 2.05, + "learning_rate": 1.3174821271388396e-05, + "loss": 0.8751, + "step": 31906 + }, + { + "epoch": 2.05, + "learning_rate": 1.3174606582365445e-05, + "loss": 0.8986, + "step": 31907 + }, + { + "epoch": 2.05, + "learning_rate": 1.3174391893342496e-05, + "loss": 0.8916, + "step": 31908 + }, + { + "epoch": 2.05, + "learning_rate": 1.3174177204319545e-05, + "loss": 0.8413, + "step": 31909 + }, + { + "epoch": 2.05, + "learning_rate": 1.3173962515296594e-05, + "loss": 0.8572, + "step": 31910 + }, + { + "epoch": 2.05, + "learning_rate": 1.3173747826273644e-05, + "loss": 0.8912, + "step": 31911 + }, + { + "epoch": 2.05, + "learning_rate": 1.3173533137250693e-05, + "loss": 0.9422, + "step": 31912 + }, + { + "epoch": 2.05, + "learning_rate": 1.3173318448227744e-05, + "loss": 0.8722, + "step": 31913 + }, + { + "epoch": 2.05, + "learning_rate": 1.3173103759204793e-05, + "loss": 0.8895, + "step": 31914 + }, + { + "epoch": 2.05, + "learning_rate": 1.3172889070181842e-05, + "loss": 0.9121, + "step": 31915 + }, + { + "epoch": 2.05, + "learning_rate": 1.3172674381158893e-05, + "loss": 0.8663, + "step": 31916 + }, + { + "epoch": 2.05, + "learning_rate": 1.3172459692135942e-05, + "loss": 0.9195, + "step": 31917 + }, + { + "epoch": 2.05, + "learning_rate": 1.317224500311299e-05, + "loss": 0.9236, + "step": 31918 + }, + { + "epoch": 2.05, + "learning_rate": 1.3172030314090041e-05, + "loss": 0.8737, + "step": 31919 + }, + { + "epoch": 2.05, + "learning_rate": 1.317181562506709e-05, + "loss": 0.9052, + "step": 31920 + }, + { + "epoch": 2.05, + "learning_rate": 1.3171600936044143e-05, + "loss": 0.868, + "step": 31921 + }, + { + "epoch": 2.05, + "learning_rate": 1.3171386247021191e-05, + "loss": 0.884, + "step": 31922 + }, + { + "epoch": 2.05, + "learning_rate": 1.3171171557998242e-05, + "loss": 0.9196, + "step": 31923 + }, + { + "epoch": 2.05, + "learning_rate": 1.3170956868975291e-05, + "loss": 0.8726, + "step": 31924 + }, + { + "epoch": 2.05, + "learning_rate": 1.317074217995234e-05, + "loss": 0.8586, + "step": 31925 + }, + { + "epoch": 2.05, + "learning_rate": 1.317052749092939e-05, + "loss": 0.8659, + "step": 31926 + }, + { + "epoch": 2.05, + "learning_rate": 1.317031280190644e-05, + "loss": 0.956, + "step": 31927 + }, + { + "epoch": 2.05, + "learning_rate": 1.3170098112883489e-05, + "loss": 0.8503, + "step": 31928 + }, + { + "epoch": 2.05, + "learning_rate": 1.316988342386054e-05, + "loss": 0.8736, + "step": 31929 + }, + { + "epoch": 2.05, + "learning_rate": 1.3169668734837588e-05, + "loss": 0.8093, + "step": 31930 + }, + { + "epoch": 2.05, + "learning_rate": 1.3169454045814637e-05, + "loss": 0.922, + "step": 31931 + }, + { + "epoch": 2.05, + "learning_rate": 1.3169239356791688e-05, + "loss": 0.8772, + "step": 31932 + }, + { + "epoch": 2.05, + "learning_rate": 1.3169024667768737e-05, + "loss": 0.9105, + "step": 31933 + }, + { + "epoch": 2.05, + "learning_rate": 1.3168809978745788e-05, + "loss": 0.9059, + "step": 31934 + }, + { + "epoch": 2.05, + "learning_rate": 1.3168595289722838e-05, + "loss": 0.9235, + "step": 31935 + }, + { + "epoch": 2.05, + "learning_rate": 1.3168380600699889e-05, + "loss": 0.8512, + "step": 31936 + }, + { + "epoch": 2.05, + "learning_rate": 1.3168165911676938e-05, + "loss": 0.9152, + "step": 31937 + }, + { + "epoch": 2.05, + "learning_rate": 1.3167951222653987e-05, + "loss": 0.8704, + "step": 31938 + }, + { + "epoch": 2.05, + "learning_rate": 1.3167736533631037e-05, + "loss": 0.9517, + "step": 31939 + }, + { + "epoch": 2.05, + "learning_rate": 1.3167521844608086e-05, + "loss": 0.8801, + "step": 31940 + }, + { + "epoch": 2.06, + "learning_rate": 1.3167307155585135e-05, + "loss": 0.9339, + "step": 31941 + }, + { + "epoch": 2.06, + "learning_rate": 1.3167092466562186e-05, + "loss": 0.8461, + "step": 31942 + }, + { + "epoch": 2.06, + "learning_rate": 1.3166877777539235e-05, + "loss": 0.8731, + "step": 31943 + }, + { + "epoch": 2.06, + "learning_rate": 1.3166663088516286e-05, + "loss": 0.8956, + "step": 31944 + }, + { + "epoch": 2.06, + "learning_rate": 1.3166448399493335e-05, + "loss": 0.9197, + "step": 31945 + }, + { + "epoch": 2.06, + "learning_rate": 1.3166233710470384e-05, + "loss": 0.902, + "step": 31946 + }, + { + "epoch": 2.06, + "learning_rate": 1.3166019021447434e-05, + "loss": 0.8495, + "step": 31947 + }, + { + "epoch": 2.06, + "learning_rate": 1.3165804332424483e-05, + "loss": 0.9519, + "step": 31948 + }, + { + "epoch": 2.06, + "learning_rate": 1.3165589643401532e-05, + "loss": 0.9493, + "step": 31949 + }, + { + "epoch": 2.06, + "learning_rate": 1.3165374954378585e-05, + "loss": 0.9839, + "step": 31950 + }, + { + "epoch": 2.06, + "learning_rate": 1.3165160265355633e-05, + "loss": 0.8795, + "step": 31951 + }, + { + "epoch": 2.06, + "learning_rate": 1.3164945576332684e-05, + "loss": 0.8963, + "step": 31952 + }, + { + "epoch": 2.06, + "learning_rate": 1.3164730887309733e-05, + "loss": 0.8934, + "step": 31953 + }, + { + "epoch": 2.06, + "learning_rate": 1.3164516198286784e-05, + "loss": 0.8958, + "step": 31954 + }, + { + "epoch": 2.06, + "learning_rate": 1.3164301509263833e-05, + "loss": 0.9591, + "step": 31955 + }, + { + "epoch": 2.06, + "learning_rate": 1.3164086820240882e-05, + "loss": 0.852, + "step": 31956 + }, + { + "epoch": 2.06, + "learning_rate": 1.3163872131217932e-05, + "loss": 0.8778, + "step": 31957 + }, + { + "epoch": 2.06, + "learning_rate": 1.3163657442194981e-05, + "loss": 0.8851, + "step": 31958 + }, + { + "epoch": 2.06, + "learning_rate": 1.316344275317203e-05, + "loss": 0.8959, + "step": 31959 + }, + { + "epoch": 2.06, + "learning_rate": 1.3163228064149081e-05, + "loss": 0.898, + "step": 31960 + }, + { + "epoch": 2.06, + "learning_rate": 1.316301337512613e-05, + "loss": 0.9718, + "step": 31961 + }, + { + "epoch": 2.06, + "learning_rate": 1.316279868610318e-05, + "loss": 0.9003, + "step": 31962 + }, + { + "epoch": 2.06, + "learning_rate": 1.316258399708023e-05, + "loss": 0.9078, + "step": 31963 + }, + { + "epoch": 2.06, + "learning_rate": 1.3162369308057282e-05, + "loss": 0.9319, + "step": 31964 + }, + { + "epoch": 2.06, + "learning_rate": 1.3162154619034331e-05, + "loss": 0.9497, + "step": 31965 + }, + { + "epoch": 2.06, + "learning_rate": 1.316193993001138e-05, + "loss": 0.9332, + "step": 31966 + }, + { + "epoch": 2.06, + "learning_rate": 1.316172524098843e-05, + "loss": 0.9279, + "step": 31967 + }, + { + "epoch": 2.06, + "learning_rate": 1.316151055196548e-05, + "loss": 0.8884, + "step": 31968 + }, + { + "epoch": 2.06, + "learning_rate": 1.3161295862942528e-05, + "loss": 0.8538, + "step": 31969 + }, + { + "epoch": 2.06, + "learning_rate": 1.3161081173919579e-05, + "loss": 0.8631, + "step": 31970 + }, + { + "epoch": 2.06, + "learning_rate": 1.3160866484896628e-05, + "loss": 0.8696, + "step": 31971 + }, + { + "epoch": 2.06, + "learning_rate": 1.3160651795873679e-05, + "loss": 0.9138, + "step": 31972 + }, + { + "epoch": 2.06, + "learning_rate": 1.3160437106850728e-05, + "loss": 0.9035, + "step": 31973 + }, + { + "epoch": 2.06, + "learning_rate": 1.3160222417827777e-05, + "loss": 0.9138, + "step": 31974 + }, + { + "epoch": 2.06, + "learning_rate": 1.3160007728804827e-05, + "loss": 0.8473, + "step": 31975 + }, + { + "epoch": 2.06, + "learning_rate": 1.3159793039781876e-05, + "loss": 0.9359, + "step": 31976 + }, + { + "epoch": 2.06, + "learning_rate": 1.3159578350758925e-05, + "loss": 0.8954, + "step": 31977 + }, + { + "epoch": 2.06, + "learning_rate": 1.3159363661735976e-05, + "loss": 0.9342, + "step": 31978 + }, + { + "epoch": 2.06, + "learning_rate": 1.3159148972713027e-05, + "loss": 0.9667, + "step": 31979 + }, + { + "epoch": 2.06, + "learning_rate": 1.3158934283690077e-05, + "loss": 0.8005, + "step": 31980 + }, + { + "epoch": 2.06, + "learning_rate": 1.3158719594667126e-05, + "loss": 0.8428, + "step": 31981 + }, + { + "epoch": 2.06, + "learning_rate": 1.3158504905644177e-05, + "loss": 0.8345, + "step": 31982 + }, + { + "epoch": 2.06, + "learning_rate": 1.3158290216621226e-05, + "loss": 0.914, + "step": 31983 + }, + { + "epoch": 2.06, + "learning_rate": 1.3158075527598275e-05, + "loss": 0.8525, + "step": 31984 + }, + { + "epoch": 2.06, + "learning_rate": 1.3157860838575325e-05, + "loss": 0.8302, + "step": 31985 + }, + { + "epoch": 2.06, + "learning_rate": 1.3157646149552374e-05, + "loss": 0.9007, + "step": 31986 + }, + { + "epoch": 2.06, + "learning_rate": 1.3157431460529423e-05, + "loss": 0.8201, + "step": 31987 + }, + { + "epoch": 2.06, + "learning_rate": 1.3157216771506474e-05, + "loss": 0.9348, + "step": 31988 + }, + { + "epoch": 2.06, + "learning_rate": 1.3157002082483523e-05, + "loss": 0.9952, + "step": 31989 + }, + { + "epoch": 2.06, + "learning_rate": 1.3156787393460572e-05, + "loss": 0.9281, + "step": 31990 + }, + { + "epoch": 2.06, + "learning_rate": 1.3156572704437623e-05, + "loss": 0.8522, + "step": 31991 + }, + { + "epoch": 2.06, + "learning_rate": 1.3156358015414672e-05, + "loss": 0.896, + "step": 31992 + }, + { + "epoch": 2.06, + "learning_rate": 1.3156143326391724e-05, + "loss": 0.9445, + "step": 31993 + }, + { + "epoch": 2.06, + "learning_rate": 1.3155928637368773e-05, + "loss": 0.9611, + "step": 31994 + }, + { + "epoch": 2.06, + "learning_rate": 1.3155713948345824e-05, + "loss": 0.9152, + "step": 31995 + }, + { + "epoch": 2.06, + "learning_rate": 1.3155499259322872e-05, + "loss": 0.874, + "step": 31996 + }, + { + "epoch": 2.06, + "learning_rate": 1.3155284570299921e-05, + "loss": 0.8672, + "step": 31997 + }, + { + "epoch": 2.06, + "learning_rate": 1.3155069881276972e-05, + "loss": 0.9, + "step": 31998 + }, + { + "epoch": 2.06, + "learning_rate": 1.3154855192254021e-05, + "loss": 0.9438, + "step": 31999 + }, + { + "epoch": 2.06, + "learning_rate": 1.3154640503231072e-05, + "loss": 0.9129, + "step": 32000 + }, + { + "epoch": 2.06, + "eval_laion/cider": 0.5215381077893255, + "eval_laion/meteor": 0.4000185901070932, + "eval_laion/pred_num_words": 37.99609375, + "eval_laion/sacrebleu": 15.985642911731773, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0430983304977417, + "eval_runtime": 96.6379, + "eval_samples_per_second": 5.164, + "eval_steps_per_second": 0.331, + "step": 32000 + }, + { + "epoch": 2.06, + "learning_rate": 1.315442581420812e-05, + "loss": 0.8754, + "step": 32001 + }, + { + "epoch": 2.06, + "learning_rate": 1.315421112518517e-05, + "loss": 0.9492, + "step": 32002 + }, + { + "epoch": 2.06, + "learning_rate": 1.315399643616222e-05, + "loss": 0.9109, + "step": 32003 + }, + { + "epoch": 2.06, + "learning_rate": 1.315378174713927e-05, + "loss": 0.9181, + "step": 32004 + }, + { + "epoch": 2.06, + "learning_rate": 1.3153567058116318e-05, + "loss": 0.8467, + "step": 32005 + }, + { + "epoch": 2.06, + "learning_rate": 1.3153352369093369e-05, + "loss": 0.8326, + "step": 32006 + }, + { + "epoch": 2.06, + "learning_rate": 1.3153137680070418e-05, + "loss": 0.9516, + "step": 32007 + }, + { + "epoch": 2.06, + "learning_rate": 1.315292299104747e-05, + "loss": 0.9578, + "step": 32008 + }, + { + "epoch": 2.06, + "learning_rate": 1.315270830202452e-05, + "loss": 0.9182, + "step": 32009 + }, + { + "epoch": 2.06, + "learning_rate": 1.315249361300157e-05, + "loss": 0.8532, + "step": 32010 + }, + { + "epoch": 2.06, + "learning_rate": 1.3152278923978619e-05, + "loss": 0.8417, + "step": 32011 + }, + { + "epoch": 2.06, + "learning_rate": 1.3152064234955668e-05, + "loss": 0.956, + "step": 32012 + }, + { + "epoch": 2.06, + "learning_rate": 1.3151849545932718e-05, + "loss": 0.9246, + "step": 32013 + }, + { + "epoch": 2.06, + "learning_rate": 1.3151634856909767e-05, + "loss": 0.8836, + "step": 32014 + }, + { + "epoch": 2.06, + "learning_rate": 1.3151420167886816e-05, + "loss": 0.8711, + "step": 32015 + }, + { + "epoch": 2.06, + "learning_rate": 1.3151205478863867e-05, + "loss": 0.9137, + "step": 32016 + }, + { + "epoch": 2.06, + "learning_rate": 1.3150990789840916e-05, + "loss": 0.904, + "step": 32017 + }, + { + "epoch": 2.06, + "learning_rate": 1.3150776100817965e-05, + "loss": 0.8755, + "step": 32018 + }, + { + "epoch": 2.06, + "learning_rate": 1.3150561411795016e-05, + "loss": 0.8731, + "step": 32019 + }, + { + "epoch": 2.06, + "learning_rate": 1.3150346722772065e-05, + "loss": 0.8958, + "step": 32020 + }, + { + "epoch": 2.06, + "learning_rate": 1.3150132033749115e-05, + "loss": 0.8898, + "step": 32021 + }, + { + "epoch": 2.06, + "learning_rate": 1.3149917344726166e-05, + "loss": 0.8459, + "step": 32022 + }, + { + "epoch": 2.06, + "learning_rate": 1.3149702655703217e-05, + "loss": 0.8446, + "step": 32023 + }, + { + "epoch": 2.06, + "learning_rate": 1.3149487966680266e-05, + "loss": 0.9464, + "step": 32024 + }, + { + "epoch": 2.06, + "learning_rate": 1.3149273277657314e-05, + "loss": 0.7936, + "step": 32025 + }, + { + "epoch": 2.06, + "learning_rate": 1.3149058588634365e-05, + "loss": 0.8677, + "step": 32026 + }, + { + "epoch": 2.06, + "learning_rate": 1.3148843899611414e-05, + "loss": 0.9187, + "step": 32027 + }, + { + "epoch": 2.06, + "learning_rate": 1.3148629210588463e-05, + "loss": 0.8528, + "step": 32028 + }, + { + "epoch": 2.06, + "learning_rate": 1.3148414521565514e-05, + "loss": 0.9014, + "step": 32029 + }, + { + "epoch": 2.06, + "learning_rate": 1.3148199832542563e-05, + "loss": 0.8742, + "step": 32030 + }, + { + "epoch": 2.06, + "learning_rate": 1.3147985143519613e-05, + "loss": 0.9199, + "step": 32031 + }, + { + "epoch": 2.06, + "learning_rate": 1.3147770454496662e-05, + "loss": 0.9681, + "step": 32032 + }, + { + "epoch": 2.06, + "learning_rate": 1.3147555765473711e-05, + "loss": 0.8646, + "step": 32033 + }, + { + "epoch": 2.06, + "learning_rate": 1.3147341076450762e-05, + "loss": 0.8825, + "step": 32034 + }, + { + "epoch": 2.06, + "learning_rate": 1.3147126387427811e-05, + "loss": 0.9044, + "step": 32035 + }, + { + "epoch": 2.06, + "learning_rate": 1.314691169840486e-05, + "loss": 0.8242, + "step": 32036 + }, + { + "epoch": 2.06, + "learning_rate": 1.3146697009381912e-05, + "loss": 0.8583, + "step": 32037 + }, + { + "epoch": 2.06, + "learning_rate": 1.3146482320358961e-05, + "loss": 0.8526, + "step": 32038 + }, + { + "epoch": 2.06, + "learning_rate": 1.3146267631336012e-05, + "loss": 0.9198, + "step": 32039 + }, + { + "epoch": 2.06, + "learning_rate": 1.314605294231306e-05, + "loss": 0.9786, + "step": 32040 + }, + { + "epoch": 2.06, + "learning_rate": 1.3145838253290111e-05, + "loss": 0.8797, + "step": 32041 + }, + { + "epoch": 2.06, + "learning_rate": 1.314562356426716e-05, + "loss": 0.9092, + "step": 32042 + }, + { + "epoch": 2.06, + "learning_rate": 1.314540887524421e-05, + "loss": 0.925, + "step": 32043 + }, + { + "epoch": 2.06, + "learning_rate": 1.314519418622126e-05, + "loss": 0.8654, + "step": 32044 + }, + { + "epoch": 2.06, + "learning_rate": 1.3144979497198309e-05, + "loss": 0.8636, + "step": 32045 + }, + { + "epoch": 2.06, + "learning_rate": 1.3144764808175358e-05, + "loss": 0.8805, + "step": 32046 + }, + { + "epoch": 2.06, + "learning_rate": 1.3144550119152409e-05, + "loss": 0.8988, + "step": 32047 + }, + { + "epoch": 2.06, + "learning_rate": 1.3144335430129458e-05, + "loss": 0.9124, + "step": 32048 + }, + { + "epoch": 2.06, + "learning_rate": 1.3144120741106508e-05, + "loss": 0.7964, + "step": 32049 + }, + { + "epoch": 2.06, + "learning_rate": 1.3143906052083557e-05, + "loss": 0.921, + "step": 32050 + }, + { + "epoch": 2.06, + "learning_rate": 1.314369136306061e-05, + "loss": 0.8419, + "step": 32051 + }, + { + "epoch": 2.06, + "learning_rate": 1.3143476674037659e-05, + "loss": 0.8452, + "step": 32052 + }, + { + "epoch": 2.06, + "learning_rate": 1.3143261985014708e-05, + "loss": 0.7979, + "step": 32053 + }, + { + "epoch": 2.06, + "learning_rate": 1.3143047295991758e-05, + "loss": 0.9395, + "step": 32054 + }, + { + "epoch": 2.06, + "learning_rate": 1.3142832606968807e-05, + "loss": 0.8665, + "step": 32055 + }, + { + "epoch": 2.06, + "learning_rate": 1.3142617917945856e-05, + "loss": 0.8221, + "step": 32056 + }, + { + "epoch": 2.06, + "learning_rate": 1.3142403228922907e-05, + "loss": 0.9058, + "step": 32057 + }, + { + "epoch": 2.06, + "learning_rate": 1.3142188539899956e-05, + "loss": 0.921, + "step": 32058 + }, + { + "epoch": 2.06, + "learning_rate": 1.3141973850877006e-05, + "loss": 0.8645, + "step": 32059 + }, + { + "epoch": 2.06, + "learning_rate": 1.3141759161854055e-05, + "loss": 0.9046, + "step": 32060 + }, + { + "epoch": 2.06, + "learning_rate": 1.3141544472831104e-05, + "loss": 0.9432, + "step": 32061 + }, + { + "epoch": 2.06, + "learning_rate": 1.3141329783808155e-05, + "loss": 0.9368, + "step": 32062 + }, + { + "epoch": 2.06, + "learning_rate": 1.3141115094785204e-05, + "loss": 0.8429, + "step": 32063 + }, + { + "epoch": 2.06, + "learning_rate": 1.3140900405762253e-05, + "loss": 0.8461, + "step": 32064 + }, + { + "epoch": 2.06, + "learning_rate": 1.3140685716739304e-05, + "loss": 0.9479, + "step": 32065 + }, + { + "epoch": 2.06, + "learning_rate": 1.3140471027716354e-05, + "loss": 0.9349, + "step": 32066 + }, + { + "epoch": 2.06, + "learning_rate": 1.3140256338693405e-05, + "loss": 0.9094, + "step": 32067 + }, + { + "epoch": 2.06, + "learning_rate": 1.3140041649670454e-05, + "loss": 0.8828, + "step": 32068 + }, + { + "epoch": 2.06, + "learning_rate": 1.3139826960647505e-05, + "loss": 0.8456, + "step": 32069 + }, + { + "epoch": 2.06, + "learning_rate": 1.3139612271624553e-05, + "loss": 0.8797, + "step": 32070 + }, + { + "epoch": 2.06, + "learning_rate": 1.3139397582601602e-05, + "loss": 0.9195, + "step": 32071 + }, + { + "epoch": 2.06, + "learning_rate": 1.3139182893578653e-05, + "loss": 0.9437, + "step": 32072 + }, + { + "epoch": 2.06, + "learning_rate": 1.3138968204555702e-05, + "loss": 0.9341, + "step": 32073 + }, + { + "epoch": 2.06, + "learning_rate": 1.3138753515532751e-05, + "loss": 0.915, + "step": 32074 + }, + { + "epoch": 2.06, + "learning_rate": 1.3138538826509802e-05, + "loss": 0.8447, + "step": 32075 + }, + { + "epoch": 2.06, + "learning_rate": 1.313832413748685e-05, + "loss": 0.8737, + "step": 32076 + }, + { + "epoch": 2.06, + "learning_rate": 1.31381094484639e-05, + "loss": 0.9066, + "step": 32077 + }, + { + "epoch": 2.06, + "learning_rate": 1.313789475944095e-05, + "loss": 0.8773, + "step": 32078 + }, + { + "epoch": 2.06, + "learning_rate": 1.3137680070418e-05, + "loss": 0.8691, + "step": 32079 + }, + { + "epoch": 2.06, + "learning_rate": 1.3137465381395052e-05, + "loss": 0.8653, + "step": 32080 + }, + { + "epoch": 2.06, + "learning_rate": 1.31372506923721e-05, + "loss": 0.9098, + "step": 32081 + }, + { + "epoch": 2.06, + "learning_rate": 1.3137036003349151e-05, + "loss": 0.8663, + "step": 32082 + }, + { + "epoch": 2.06, + "learning_rate": 1.31368213143262e-05, + "loss": 0.9121, + "step": 32083 + }, + { + "epoch": 2.06, + "learning_rate": 1.313660662530325e-05, + "loss": 0.9077, + "step": 32084 + }, + { + "epoch": 2.06, + "learning_rate": 1.31363919362803e-05, + "loss": 0.8835, + "step": 32085 + }, + { + "epoch": 2.06, + "learning_rate": 1.3136177247257349e-05, + "loss": 0.8368, + "step": 32086 + }, + { + "epoch": 2.06, + "learning_rate": 1.3135962558234398e-05, + "loss": 0.8503, + "step": 32087 + }, + { + "epoch": 2.06, + "learning_rate": 1.3135747869211448e-05, + "loss": 0.8154, + "step": 32088 + }, + { + "epoch": 2.06, + "learning_rate": 1.3135533180188497e-05, + "loss": 0.9271, + "step": 32089 + }, + { + "epoch": 2.06, + "learning_rate": 1.3135318491165548e-05, + "loss": 0.8904, + "step": 32090 + }, + { + "epoch": 2.06, + "learning_rate": 1.3135103802142597e-05, + "loss": 0.9, + "step": 32091 + }, + { + "epoch": 2.06, + "learning_rate": 1.3134889113119646e-05, + "loss": 0.884, + "step": 32092 + }, + { + "epoch": 2.06, + "learning_rate": 1.3134674424096697e-05, + "loss": 0.9378, + "step": 32093 + }, + { + "epoch": 2.06, + "learning_rate": 1.3134459735073746e-05, + "loss": 0.9259, + "step": 32094 + }, + { + "epoch": 2.06, + "learning_rate": 1.3134245046050798e-05, + "loss": 0.8591, + "step": 32095 + }, + { + "epoch": 2.06, + "learning_rate": 1.3134030357027847e-05, + "loss": 0.8685, + "step": 32096 + }, + { + "epoch": 2.07, + "learning_rate": 1.3133815668004896e-05, + "loss": 0.8653, + "step": 32097 + }, + { + "epoch": 2.07, + "learning_rate": 1.3133600978981947e-05, + "loss": 0.9075, + "step": 32098 + }, + { + "epoch": 2.07, + "learning_rate": 1.3133386289958995e-05, + "loss": 0.9425, + "step": 32099 + }, + { + "epoch": 2.07, + "learning_rate": 1.3133171600936046e-05, + "loss": 0.8878, + "step": 32100 + }, + { + "epoch": 2.07, + "learning_rate": 1.3132956911913095e-05, + "loss": 0.9482, + "step": 32101 + }, + { + "epoch": 2.07, + "learning_rate": 1.3132742222890144e-05, + "loss": 0.9007, + "step": 32102 + }, + { + "epoch": 2.07, + "learning_rate": 1.3132527533867195e-05, + "loss": 0.8753, + "step": 32103 + }, + { + "epoch": 2.07, + "learning_rate": 1.3132312844844244e-05, + "loss": 0.9109, + "step": 32104 + }, + { + "epoch": 2.07, + "learning_rate": 1.3132098155821293e-05, + "loss": 0.9042, + "step": 32105 + }, + { + "epoch": 2.07, + "learning_rate": 1.3131883466798343e-05, + "loss": 0.9147, + "step": 32106 + }, + { + "epoch": 2.07, + "learning_rate": 1.3131668777775392e-05, + "loss": 0.9539, + "step": 32107 + }, + { + "epoch": 2.07, + "learning_rate": 1.3131454088752443e-05, + "loss": 0.8724, + "step": 32108 + }, + { + "epoch": 2.07, + "learning_rate": 1.3131239399729492e-05, + "loss": 0.9348, + "step": 32109 + }, + { + "epoch": 2.07, + "learning_rate": 1.3131024710706544e-05, + "loss": 0.8996, + "step": 32110 + }, + { + "epoch": 2.07, + "learning_rate": 1.3130810021683593e-05, + "loss": 0.9383, + "step": 32111 + }, + { + "epoch": 2.07, + "learning_rate": 1.3130595332660642e-05, + "loss": 0.8732, + "step": 32112 + }, + { + "epoch": 2.07, + "learning_rate": 1.3130380643637693e-05, + "loss": 0.9587, + "step": 32113 + }, + { + "epoch": 2.07, + "learning_rate": 1.3130165954614742e-05, + "loss": 0.8493, + "step": 32114 + }, + { + "epoch": 2.07, + "learning_rate": 1.312995126559179e-05, + "loss": 0.8539, + "step": 32115 + }, + { + "epoch": 2.07, + "learning_rate": 1.3129736576568841e-05, + "loss": 0.8373, + "step": 32116 + }, + { + "epoch": 2.07, + "learning_rate": 1.312952188754589e-05, + "loss": 0.8909, + "step": 32117 + }, + { + "epoch": 2.07, + "learning_rate": 1.3129307198522941e-05, + "loss": 0.9368, + "step": 32118 + }, + { + "epoch": 2.07, + "learning_rate": 1.312909250949999e-05, + "loss": 0.9037, + "step": 32119 + }, + { + "epoch": 2.07, + "learning_rate": 1.3128877820477039e-05, + "loss": 0.8363, + "step": 32120 + }, + { + "epoch": 2.07, + "learning_rate": 1.312866313145409e-05, + "loss": 0.9452, + "step": 32121 + }, + { + "epoch": 2.07, + "learning_rate": 1.3128448442431139e-05, + "loss": 0.8995, + "step": 32122 + }, + { + "epoch": 2.07, + "learning_rate": 1.3128233753408188e-05, + "loss": 0.8976, + "step": 32123 + }, + { + "epoch": 2.07, + "learning_rate": 1.312801906438524e-05, + "loss": 0.9076, + "step": 32124 + }, + { + "epoch": 2.07, + "learning_rate": 1.3127804375362289e-05, + "loss": 0.884, + "step": 32125 + }, + { + "epoch": 2.07, + "learning_rate": 1.312758968633934e-05, + "loss": 0.8913, + "step": 32126 + }, + { + "epoch": 2.07, + "learning_rate": 1.3127374997316389e-05, + "loss": 0.9414, + "step": 32127 + }, + { + "epoch": 2.07, + "learning_rate": 1.312716030829344e-05, + "loss": 0.9307, + "step": 32128 + }, + { + "epoch": 2.07, + "learning_rate": 1.3126945619270488e-05, + "loss": 0.9458, + "step": 32129 + }, + { + "epoch": 2.07, + "learning_rate": 1.3126730930247537e-05, + "loss": 0.8366, + "step": 32130 + }, + { + "epoch": 2.07, + "learning_rate": 1.3126516241224588e-05, + "loss": 0.8588, + "step": 32131 + }, + { + "epoch": 2.07, + "learning_rate": 1.3126301552201637e-05, + "loss": 0.8831, + "step": 32132 + }, + { + "epoch": 2.07, + "learning_rate": 1.3126086863178686e-05, + "loss": 0.9628, + "step": 32133 + }, + { + "epoch": 2.07, + "learning_rate": 1.3125872174155736e-05, + "loss": 0.91, + "step": 32134 + }, + { + "epoch": 2.07, + "learning_rate": 1.3125657485132785e-05, + "loss": 0.8893, + "step": 32135 + }, + { + "epoch": 2.07, + "learning_rate": 1.3125442796109836e-05, + "loss": 0.8611, + "step": 32136 + }, + { + "epoch": 2.07, + "learning_rate": 1.3125228107086885e-05, + "loss": 0.854, + "step": 32137 + }, + { + "epoch": 2.07, + "learning_rate": 1.3125013418063934e-05, + "loss": 0.8261, + "step": 32138 + }, + { + "epoch": 2.07, + "learning_rate": 1.3124798729040986e-05, + "loss": 0.9566, + "step": 32139 + }, + { + "epoch": 2.07, + "learning_rate": 1.3124584040018035e-05, + "loss": 0.8576, + "step": 32140 + }, + { + "epoch": 2.07, + "learning_rate": 1.3124369350995086e-05, + "loss": 0.8228, + "step": 32141 + }, + { + "epoch": 2.07, + "learning_rate": 1.3124154661972135e-05, + "loss": 0.9242, + "step": 32142 + }, + { + "epoch": 2.07, + "learning_rate": 1.3123939972949184e-05, + "loss": 0.9056, + "step": 32143 + }, + { + "epoch": 2.07, + "learning_rate": 1.3123725283926234e-05, + "loss": 0.875, + "step": 32144 + }, + { + "epoch": 2.07, + "learning_rate": 1.3123510594903283e-05, + "loss": 0.9002, + "step": 32145 + }, + { + "epoch": 2.07, + "learning_rate": 1.3123295905880334e-05, + "loss": 0.8611, + "step": 32146 + }, + { + "epoch": 2.07, + "learning_rate": 1.3123081216857383e-05, + "loss": 0.9214, + "step": 32147 + }, + { + "epoch": 2.07, + "learning_rate": 1.3122866527834432e-05, + "loss": 0.9582, + "step": 32148 + }, + { + "epoch": 2.07, + "learning_rate": 1.3122651838811483e-05, + "loss": 0.8756, + "step": 32149 + }, + { + "epoch": 2.07, + "learning_rate": 1.3122437149788532e-05, + "loss": 0.7997, + "step": 32150 + }, + { + "epoch": 2.07, + "learning_rate": 1.312222246076558e-05, + "loss": 0.9326, + "step": 32151 + }, + { + "epoch": 2.07, + "learning_rate": 1.3122007771742631e-05, + "loss": 0.9097, + "step": 32152 + }, + { + "epoch": 2.07, + "learning_rate": 1.3121793082719682e-05, + "loss": 0.9448, + "step": 32153 + }, + { + "epoch": 2.07, + "learning_rate": 1.3121578393696733e-05, + "loss": 0.8441, + "step": 32154 + }, + { + "epoch": 2.07, + "learning_rate": 1.3121363704673782e-05, + "loss": 0.8831, + "step": 32155 + }, + { + "epoch": 2.07, + "learning_rate": 1.3121149015650832e-05, + "loss": 0.9201, + "step": 32156 + }, + { + "epoch": 2.07, + "learning_rate": 1.3120934326627881e-05, + "loss": 0.8696, + "step": 32157 + }, + { + "epoch": 2.07, + "learning_rate": 1.312071963760493e-05, + "loss": 0.9016, + "step": 32158 + }, + { + "epoch": 2.07, + "learning_rate": 1.312050494858198e-05, + "loss": 0.9557, + "step": 32159 + }, + { + "epoch": 2.07, + "learning_rate": 1.312029025955903e-05, + "loss": 0.8344, + "step": 32160 + }, + { + "epoch": 2.07, + "learning_rate": 1.3120075570536079e-05, + "loss": 0.8357, + "step": 32161 + }, + { + "epoch": 2.07, + "learning_rate": 1.311986088151313e-05, + "loss": 0.8748, + "step": 32162 + }, + { + "epoch": 2.07, + "learning_rate": 1.3119646192490178e-05, + "loss": 0.9572, + "step": 32163 + }, + { + "epoch": 2.07, + "learning_rate": 1.3119431503467227e-05, + "loss": 0.878, + "step": 32164 + }, + { + "epoch": 2.07, + "learning_rate": 1.3119216814444278e-05, + "loss": 0.9115, + "step": 32165 + }, + { + "epoch": 2.07, + "learning_rate": 1.3119002125421327e-05, + "loss": 0.9309, + "step": 32166 + }, + { + "epoch": 2.07, + "learning_rate": 1.3118787436398378e-05, + "loss": 0.9323, + "step": 32167 + }, + { + "epoch": 2.07, + "learning_rate": 1.3118572747375428e-05, + "loss": 0.8953, + "step": 32168 + }, + { + "epoch": 2.07, + "learning_rate": 1.3118358058352479e-05, + "loss": 0.9094, + "step": 32169 + }, + { + "epoch": 2.07, + "learning_rate": 1.3118143369329528e-05, + "loss": 0.8999, + "step": 32170 + }, + { + "epoch": 2.07, + "learning_rate": 1.3117928680306577e-05, + "loss": 0.8709, + "step": 32171 + }, + { + "epoch": 2.07, + "learning_rate": 1.3117713991283628e-05, + "loss": 0.9764, + "step": 32172 + }, + { + "epoch": 2.07, + "learning_rate": 1.3117499302260676e-05, + "loss": 0.9611, + "step": 32173 + }, + { + "epoch": 2.07, + "learning_rate": 1.3117284613237725e-05, + "loss": 0.8123, + "step": 32174 + }, + { + "epoch": 2.07, + "learning_rate": 1.3117069924214776e-05, + "loss": 0.8858, + "step": 32175 + }, + { + "epoch": 2.07, + "learning_rate": 1.3116855235191825e-05, + "loss": 0.8828, + "step": 32176 + }, + { + "epoch": 2.07, + "learning_rate": 1.3116640546168876e-05, + "loss": 0.8921, + "step": 32177 + }, + { + "epoch": 2.07, + "learning_rate": 1.3116425857145925e-05, + "loss": 0.9215, + "step": 32178 + }, + { + "epoch": 2.07, + "learning_rate": 1.3116211168122974e-05, + "loss": 0.8938, + "step": 32179 + }, + { + "epoch": 2.07, + "learning_rate": 1.3115996479100024e-05, + "loss": 0.8721, + "step": 32180 + }, + { + "epoch": 2.07, + "learning_rate": 1.3115781790077073e-05, + "loss": 0.9103, + "step": 32181 + }, + { + "epoch": 2.07, + "learning_rate": 1.3115567101054126e-05, + "loss": 0.896, + "step": 32182 + }, + { + "epoch": 2.07, + "learning_rate": 1.3115352412031175e-05, + "loss": 0.8674, + "step": 32183 + }, + { + "epoch": 2.07, + "learning_rate": 1.3115137723008224e-05, + "loss": 0.8646, + "step": 32184 + }, + { + "epoch": 2.07, + "learning_rate": 1.3114923033985274e-05, + "loss": 0.8893, + "step": 32185 + }, + { + "epoch": 2.07, + "learning_rate": 1.3114708344962323e-05, + "loss": 0.8364, + "step": 32186 + }, + { + "epoch": 2.07, + "learning_rate": 1.3114493655939374e-05, + "loss": 0.8658, + "step": 32187 + }, + { + "epoch": 2.07, + "learning_rate": 1.3114278966916423e-05, + "loss": 0.9195, + "step": 32188 + }, + { + "epoch": 2.07, + "learning_rate": 1.3114064277893472e-05, + "loss": 0.8809, + "step": 32189 + }, + { + "epoch": 2.07, + "learning_rate": 1.3113849588870522e-05, + "loss": 0.8582, + "step": 32190 + }, + { + "epoch": 2.07, + "learning_rate": 1.3113634899847571e-05, + "loss": 0.8505, + "step": 32191 + }, + { + "epoch": 2.07, + "learning_rate": 1.311342021082462e-05, + "loss": 0.892, + "step": 32192 + }, + { + "epoch": 2.07, + "learning_rate": 1.3113205521801671e-05, + "loss": 0.8752, + "step": 32193 + }, + { + "epoch": 2.07, + "learning_rate": 1.311299083277872e-05, + "loss": 0.8996, + "step": 32194 + }, + { + "epoch": 2.07, + "learning_rate": 1.311277614375577e-05, + "loss": 0.8454, + "step": 32195 + }, + { + "epoch": 2.07, + "learning_rate": 1.311256145473282e-05, + "loss": 0.846, + "step": 32196 + }, + { + "epoch": 2.07, + "learning_rate": 1.3112346765709872e-05, + "loss": 0.9438, + "step": 32197 + }, + { + "epoch": 2.07, + "learning_rate": 1.3112132076686921e-05, + "loss": 0.8145, + "step": 32198 + }, + { + "epoch": 2.07, + "learning_rate": 1.311191738766397e-05, + "loss": 0.8491, + "step": 32199 + }, + { + "epoch": 2.07, + "learning_rate": 1.311170269864102e-05, + "loss": 0.873, + "step": 32200 + }, + { + "epoch": 2.07, + "learning_rate": 1.311148800961807e-05, + "loss": 0.8682, + "step": 32201 + }, + { + "epoch": 2.07, + "learning_rate": 1.3111273320595119e-05, + "loss": 0.8927, + "step": 32202 + }, + { + "epoch": 2.07, + "learning_rate": 1.311105863157217e-05, + "loss": 0.9184, + "step": 32203 + }, + { + "epoch": 2.07, + "learning_rate": 1.3110843942549218e-05, + "loss": 0.9055, + "step": 32204 + }, + { + "epoch": 2.07, + "learning_rate": 1.3110629253526269e-05, + "loss": 0.8649, + "step": 32205 + }, + { + "epoch": 2.07, + "learning_rate": 1.3110414564503318e-05, + "loss": 0.9462, + "step": 32206 + }, + { + "epoch": 2.07, + "learning_rate": 1.3110199875480367e-05, + "loss": 0.8812, + "step": 32207 + }, + { + "epoch": 2.07, + "learning_rate": 1.3109985186457417e-05, + "loss": 0.8615, + "step": 32208 + }, + { + "epoch": 2.07, + "learning_rate": 1.3109770497434466e-05, + "loss": 0.9355, + "step": 32209 + }, + { + "epoch": 2.07, + "learning_rate": 1.3109555808411515e-05, + "loss": 0.926, + "step": 32210 + }, + { + "epoch": 2.07, + "learning_rate": 1.3109341119388568e-05, + "loss": 0.8589, + "step": 32211 + }, + { + "epoch": 2.07, + "learning_rate": 1.3109126430365617e-05, + "loss": 0.8014, + "step": 32212 + }, + { + "epoch": 2.07, + "learning_rate": 1.3108911741342667e-05, + "loss": 0.8744, + "step": 32213 + }, + { + "epoch": 2.07, + "learning_rate": 1.3108697052319716e-05, + "loss": 0.8492, + "step": 32214 + }, + { + "epoch": 2.07, + "learning_rate": 1.3108482363296767e-05, + "loss": 0.8725, + "step": 32215 + }, + { + "epoch": 2.07, + "learning_rate": 1.3108267674273816e-05, + "loss": 0.9021, + "step": 32216 + }, + { + "epoch": 2.07, + "learning_rate": 1.3108052985250865e-05, + "loss": 0.9289, + "step": 32217 + }, + { + "epoch": 2.07, + "learning_rate": 1.3107838296227915e-05, + "loss": 0.9763, + "step": 32218 + }, + { + "epoch": 2.07, + "learning_rate": 1.3107623607204964e-05, + "loss": 0.8381, + "step": 32219 + }, + { + "epoch": 2.07, + "learning_rate": 1.3107408918182013e-05, + "loss": 0.9492, + "step": 32220 + }, + { + "epoch": 2.07, + "learning_rate": 1.3107194229159064e-05, + "loss": 0.9253, + "step": 32221 + }, + { + "epoch": 2.07, + "learning_rate": 1.3106979540136113e-05, + "loss": 0.9198, + "step": 32222 + }, + { + "epoch": 2.07, + "learning_rate": 1.3106764851113164e-05, + "loss": 0.8421, + "step": 32223 + }, + { + "epoch": 2.07, + "learning_rate": 1.3106550162090213e-05, + "loss": 0.8732, + "step": 32224 + }, + { + "epoch": 2.07, + "learning_rate": 1.3106335473067262e-05, + "loss": 0.8985, + "step": 32225 + }, + { + "epoch": 2.07, + "learning_rate": 1.3106120784044314e-05, + "loss": 0.9028, + "step": 32226 + }, + { + "epoch": 2.07, + "learning_rate": 1.3105906095021363e-05, + "loss": 0.9532, + "step": 32227 + }, + { + "epoch": 2.07, + "learning_rate": 1.3105691405998414e-05, + "loss": 1.0006, + "step": 32228 + }, + { + "epoch": 2.07, + "learning_rate": 1.3105476716975463e-05, + "loss": 0.9034, + "step": 32229 + }, + { + "epoch": 2.07, + "learning_rate": 1.3105262027952512e-05, + "loss": 0.8974, + "step": 32230 + }, + { + "epoch": 2.07, + "learning_rate": 1.3105047338929562e-05, + "loss": 0.9381, + "step": 32231 + }, + { + "epoch": 2.07, + "learning_rate": 1.3104832649906611e-05, + "loss": 0.9202, + "step": 32232 + }, + { + "epoch": 2.07, + "learning_rate": 1.3104617960883662e-05, + "loss": 0.8552, + "step": 32233 + }, + { + "epoch": 2.07, + "learning_rate": 1.310440327186071e-05, + "loss": 0.9672, + "step": 32234 + }, + { + "epoch": 2.07, + "learning_rate": 1.310418858283776e-05, + "loss": 0.8404, + "step": 32235 + }, + { + "epoch": 2.07, + "learning_rate": 1.310397389381481e-05, + "loss": 0.8579, + "step": 32236 + }, + { + "epoch": 2.07, + "learning_rate": 1.310375920479186e-05, + "loss": 0.8536, + "step": 32237 + }, + { + "epoch": 2.07, + "learning_rate": 1.3103544515768908e-05, + "loss": 0.9806, + "step": 32238 + }, + { + "epoch": 2.07, + "learning_rate": 1.3103329826745959e-05, + "loss": 0.9212, + "step": 32239 + }, + { + "epoch": 2.07, + "learning_rate": 1.310311513772301e-05, + "loss": 0.8515, + "step": 32240 + }, + { + "epoch": 2.07, + "learning_rate": 1.310290044870006e-05, + "loss": 0.8828, + "step": 32241 + }, + { + "epoch": 2.07, + "learning_rate": 1.310268575967711e-05, + "loss": 0.9074, + "step": 32242 + }, + { + "epoch": 2.07, + "learning_rate": 1.310247107065416e-05, + "loss": 0.9047, + "step": 32243 + }, + { + "epoch": 2.07, + "learning_rate": 1.3102256381631209e-05, + "loss": 0.8888, + "step": 32244 + }, + { + "epoch": 2.07, + "learning_rate": 1.3102041692608258e-05, + "loss": 0.8799, + "step": 32245 + }, + { + "epoch": 2.07, + "learning_rate": 1.3101827003585309e-05, + "loss": 0.9398, + "step": 32246 + }, + { + "epoch": 2.07, + "learning_rate": 1.3101612314562358e-05, + "loss": 0.8939, + "step": 32247 + }, + { + "epoch": 2.07, + "learning_rate": 1.3101397625539406e-05, + "loss": 0.9584, + "step": 32248 + }, + { + "epoch": 2.07, + "learning_rate": 1.3101182936516457e-05, + "loss": 0.9072, + "step": 32249 + }, + { + "epoch": 2.07, + "learning_rate": 1.3100968247493506e-05, + "loss": 0.9001, + "step": 32250 + }, + { + "epoch": 2.07, + "learning_rate": 1.3100753558470555e-05, + "loss": 0.9397, + "step": 32251 + }, + { + "epoch": 2.08, + "learning_rate": 1.3100538869447606e-05, + "loss": 0.9056, + "step": 32252 + }, + { + "epoch": 2.08, + "learning_rate": 1.3100324180424655e-05, + "loss": 0.8305, + "step": 32253 + }, + { + "epoch": 2.08, + "learning_rate": 1.3100109491401705e-05, + "loss": 0.9065, + "step": 32254 + }, + { + "epoch": 2.08, + "learning_rate": 1.3099894802378756e-05, + "loss": 0.92, + "step": 32255 + }, + { + "epoch": 2.08, + "learning_rate": 1.3099680113355807e-05, + "loss": 0.9047, + "step": 32256 + }, + { + "epoch": 2.08, + "learning_rate": 1.3099465424332856e-05, + "loss": 0.8669, + "step": 32257 + }, + { + "epoch": 2.08, + "learning_rate": 1.3099250735309905e-05, + "loss": 0.895, + "step": 32258 + }, + { + "epoch": 2.08, + "learning_rate": 1.3099036046286955e-05, + "loss": 0.9046, + "step": 32259 + }, + { + "epoch": 2.08, + "learning_rate": 1.3098821357264004e-05, + "loss": 0.8877, + "step": 32260 + }, + { + "epoch": 2.08, + "learning_rate": 1.3098606668241053e-05, + "loss": 0.9731, + "step": 32261 + }, + { + "epoch": 2.08, + "learning_rate": 1.3098391979218104e-05, + "loss": 0.9423, + "step": 32262 + }, + { + "epoch": 2.08, + "learning_rate": 1.3098177290195153e-05, + "loss": 0.9007, + "step": 32263 + }, + { + "epoch": 2.08, + "learning_rate": 1.3097962601172203e-05, + "loss": 0.9037, + "step": 32264 + }, + { + "epoch": 2.08, + "learning_rate": 1.3097747912149252e-05, + "loss": 0.8059, + "step": 32265 + }, + { + "epoch": 2.08, + "learning_rate": 1.3097533223126301e-05, + "loss": 0.8559, + "step": 32266 + }, + { + "epoch": 2.08, + "learning_rate": 1.3097318534103352e-05, + "loss": 0.866, + "step": 32267 + }, + { + "epoch": 2.08, + "learning_rate": 1.3097103845080401e-05, + "loss": 0.8679, + "step": 32268 + }, + { + "epoch": 2.08, + "learning_rate": 1.3096889156057453e-05, + "loss": 0.938, + "step": 32269 + }, + { + "epoch": 2.08, + "learning_rate": 1.3096674467034502e-05, + "loss": 0.964, + "step": 32270 + }, + { + "epoch": 2.08, + "learning_rate": 1.3096459778011551e-05, + "loss": 0.8497, + "step": 32271 + }, + { + "epoch": 2.08, + "learning_rate": 1.3096245088988602e-05, + "loss": 0.894, + "step": 32272 + }, + { + "epoch": 2.08, + "learning_rate": 1.3096030399965651e-05, + "loss": 0.8658, + "step": 32273 + }, + { + "epoch": 2.08, + "learning_rate": 1.3095815710942702e-05, + "loss": 0.8548, + "step": 32274 + }, + { + "epoch": 2.08, + "learning_rate": 1.309560102191975e-05, + "loss": 0.8975, + "step": 32275 + }, + { + "epoch": 2.08, + "learning_rate": 1.30953863328968e-05, + "loss": 0.9537, + "step": 32276 + }, + { + "epoch": 2.08, + "learning_rate": 1.309517164387385e-05, + "loss": 0.8849, + "step": 32277 + }, + { + "epoch": 2.08, + "learning_rate": 1.3094956954850899e-05, + "loss": 0.8784, + "step": 32278 + }, + { + "epoch": 2.08, + "learning_rate": 1.3094742265827948e-05, + "loss": 0.8517, + "step": 32279 + }, + { + "epoch": 2.08, + "learning_rate": 1.3094527576804999e-05, + "loss": 0.9178, + "step": 32280 + }, + { + "epoch": 2.08, + "learning_rate": 1.3094312887782048e-05, + "loss": 0.8855, + "step": 32281 + }, + { + "epoch": 2.08, + "learning_rate": 1.3094098198759098e-05, + "loss": 0.954, + "step": 32282 + }, + { + "epoch": 2.08, + "learning_rate": 1.3093883509736147e-05, + "loss": 0.7748, + "step": 32283 + }, + { + "epoch": 2.08, + "learning_rate": 1.30936688207132e-05, + "loss": 0.8991, + "step": 32284 + }, + { + "epoch": 2.08, + "learning_rate": 1.3093454131690249e-05, + "loss": 0.92, + "step": 32285 + }, + { + "epoch": 2.08, + "learning_rate": 1.3093239442667298e-05, + "loss": 0.9599, + "step": 32286 + }, + { + "epoch": 2.08, + "learning_rate": 1.3093024753644348e-05, + "loss": 0.8685, + "step": 32287 + }, + { + "epoch": 2.08, + "learning_rate": 1.3092810064621397e-05, + "loss": 0.926, + "step": 32288 + }, + { + "epoch": 2.08, + "learning_rate": 1.3092595375598446e-05, + "loss": 0.8326, + "step": 32289 + }, + { + "epoch": 2.08, + "learning_rate": 1.3092380686575497e-05, + "loss": 0.9333, + "step": 32290 + }, + { + "epoch": 2.08, + "learning_rate": 1.3092165997552546e-05, + "loss": 1.0142, + "step": 32291 + }, + { + "epoch": 2.08, + "learning_rate": 1.3091951308529596e-05, + "loss": 0.8657, + "step": 32292 + }, + { + "epoch": 2.08, + "learning_rate": 1.3091736619506645e-05, + "loss": 0.9562, + "step": 32293 + }, + { + "epoch": 2.08, + "learning_rate": 1.3091521930483694e-05, + "loss": 0.854, + "step": 32294 + }, + { + "epoch": 2.08, + "learning_rate": 1.3091307241460745e-05, + "loss": 0.8876, + "step": 32295 + }, + { + "epoch": 2.08, + "learning_rate": 1.3091092552437794e-05, + "loss": 0.8547, + "step": 32296 + }, + { + "epoch": 2.08, + "learning_rate": 1.3090877863414843e-05, + "loss": 0.8924, + "step": 32297 + }, + { + "epoch": 2.08, + "learning_rate": 1.3090663174391895e-05, + "loss": 0.862, + "step": 32298 + }, + { + "epoch": 2.08, + "learning_rate": 1.3090448485368944e-05, + "loss": 0.9326, + "step": 32299 + }, + { + "epoch": 2.08, + "learning_rate": 1.3090233796345995e-05, + "loss": 0.9724, + "step": 32300 + }, + { + "epoch": 2.08, + "learning_rate": 1.3090019107323044e-05, + "loss": 0.8697, + "step": 32301 + }, + { + "epoch": 2.08, + "learning_rate": 1.3089804418300095e-05, + "loss": 0.8976, + "step": 32302 + }, + { + "epoch": 2.08, + "learning_rate": 1.3089589729277144e-05, + "loss": 0.8937, + "step": 32303 + }, + { + "epoch": 2.08, + "learning_rate": 1.3089375040254193e-05, + "loss": 0.9021, + "step": 32304 + }, + { + "epoch": 2.08, + "learning_rate": 1.3089160351231243e-05, + "loss": 0.8343, + "step": 32305 + }, + { + "epoch": 2.08, + "learning_rate": 1.3088945662208292e-05, + "loss": 0.8662, + "step": 32306 + }, + { + "epoch": 2.08, + "learning_rate": 1.3088730973185341e-05, + "loss": 0.9185, + "step": 32307 + }, + { + "epoch": 2.08, + "learning_rate": 1.3088516284162392e-05, + "loss": 0.9452, + "step": 32308 + }, + { + "epoch": 2.08, + "learning_rate": 1.308830159513944e-05, + "loss": 0.8783, + "step": 32309 + }, + { + "epoch": 2.08, + "learning_rate": 1.3088086906116491e-05, + "loss": 0.8847, + "step": 32310 + }, + { + "epoch": 2.08, + "learning_rate": 1.308787221709354e-05, + "loss": 0.8406, + "step": 32311 + }, + { + "epoch": 2.08, + "learning_rate": 1.308765752807059e-05, + "loss": 0.8671, + "step": 32312 + }, + { + "epoch": 2.08, + "learning_rate": 1.3087442839047642e-05, + "loss": 0.8537, + "step": 32313 + }, + { + "epoch": 2.08, + "learning_rate": 1.308722815002469e-05, + "loss": 0.8837, + "step": 32314 + }, + { + "epoch": 2.08, + "learning_rate": 1.3087013461001741e-05, + "loss": 0.9078, + "step": 32315 + }, + { + "epoch": 2.08, + "learning_rate": 1.308679877197879e-05, + "loss": 0.9377, + "step": 32316 + }, + { + "epoch": 2.08, + "learning_rate": 1.308658408295584e-05, + "loss": 0.8744, + "step": 32317 + }, + { + "epoch": 2.08, + "learning_rate": 1.308636939393289e-05, + "loss": 0.8755, + "step": 32318 + }, + { + "epoch": 2.08, + "learning_rate": 1.3086154704909939e-05, + "loss": 0.8471, + "step": 32319 + }, + { + "epoch": 2.08, + "learning_rate": 1.308594001588699e-05, + "loss": 0.9129, + "step": 32320 + }, + { + "epoch": 2.08, + "learning_rate": 1.3085725326864039e-05, + "loss": 0.9202, + "step": 32321 + }, + { + "epoch": 2.08, + "learning_rate": 1.3085510637841087e-05, + "loss": 0.8117, + "step": 32322 + }, + { + "epoch": 2.08, + "learning_rate": 1.3085295948818138e-05, + "loss": 0.8655, + "step": 32323 + }, + { + "epoch": 2.08, + "learning_rate": 1.3085081259795187e-05, + "loss": 0.9107, + "step": 32324 + }, + { + "epoch": 2.08, + "learning_rate": 1.3084866570772236e-05, + "loss": 0.8772, + "step": 32325 + }, + { + "epoch": 2.08, + "learning_rate": 1.3084651881749287e-05, + "loss": 0.9246, + "step": 32326 + }, + { + "epoch": 2.08, + "learning_rate": 1.3084437192726337e-05, + "loss": 0.9751, + "step": 32327 + }, + { + "epoch": 2.08, + "learning_rate": 1.3084222503703388e-05, + "loss": 0.9181, + "step": 32328 + }, + { + "epoch": 2.08, + "learning_rate": 1.3084007814680437e-05, + "loss": 0.8289, + "step": 32329 + }, + { + "epoch": 2.08, + "learning_rate": 1.3083793125657488e-05, + "loss": 0.8495, + "step": 32330 + }, + { + "epoch": 2.08, + "learning_rate": 1.3083578436634537e-05, + "loss": 0.9045, + "step": 32331 + }, + { + "epoch": 2.08, + "learning_rate": 1.3083363747611586e-05, + "loss": 0.8554, + "step": 32332 + }, + { + "epoch": 2.08, + "learning_rate": 1.3083149058588636e-05, + "loss": 0.9455, + "step": 32333 + }, + { + "epoch": 2.08, + "learning_rate": 1.3082934369565685e-05, + "loss": 0.8409, + "step": 32334 + }, + { + "epoch": 2.08, + "learning_rate": 1.3082719680542734e-05, + "loss": 0.891, + "step": 32335 + }, + { + "epoch": 2.08, + "learning_rate": 1.3082504991519785e-05, + "loss": 0.9612, + "step": 32336 + }, + { + "epoch": 2.08, + "learning_rate": 1.3082290302496834e-05, + "loss": 0.9186, + "step": 32337 + }, + { + "epoch": 2.08, + "learning_rate": 1.3082075613473883e-05, + "loss": 0.8849, + "step": 32338 + }, + { + "epoch": 2.08, + "learning_rate": 1.3081860924450933e-05, + "loss": 0.8936, + "step": 32339 + }, + { + "epoch": 2.08, + "learning_rate": 1.3081646235427982e-05, + "loss": 0.8625, + "step": 32340 + }, + { + "epoch": 2.08, + "learning_rate": 1.3081431546405033e-05, + "loss": 0.9592, + "step": 32341 + }, + { + "epoch": 2.08, + "learning_rate": 1.3081216857382084e-05, + "loss": 0.9158, + "step": 32342 + }, + { + "epoch": 2.08, + "learning_rate": 1.3081002168359134e-05, + "loss": 0.9189, + "step": 32343 + }, + { + "epoch": 2.08, + "learning_rate": 1.3080787479336183e-05, + "loss": 0.8702, + "step": 32344 + }, + { + "epoch": 2.08, + "learning_rate": 1.3080572790313232e-05, + "loss": 0.8916, + "step": 32345 + }, + { + "epoch": 2.08, + "learning_rate": 1.3080358101290283e-05, + "loss": 0.8029, + "step": 32346 + }, + { + "epoch": 2.08, + "learning_rate": 1.3080143412267332e-05, + "loss": 0.8731, + "step": 32347 + }, + { + "epoch": 2.08, + "learning_rate": 1.3079928723244381e-05, + "loss": 0.9044, + "step": 32348 + }, + { + "epoch": 2.08, + "learning_rate": 1.3079714034221432e-05, + "loss": 0.9275, + "step": 32349 + }, + { + "epoch": 2.08, + "learning_rate": 1.307949934519848e-05, + "loss": 0.9178, + "step": 32350 + }, + { + "epoch": 2.08, + "learning_rate": 1.3079284656175531e-05, + "loss": 0.8785, + "step": 32351 + }, + { + "epoch": 2.08, + "learning_rate": 1.307906996715258e-05, + "loss": 0.992, + "step": 32352 + }, + { + "epoch": 2.08, + "learning_rate": 1.3078855278129629e-05, + "loss": 0.9076, + "step": 32353 + }, + { + "epoch": 2.08, + "learning_rate": 1.307864058910668e-05, + "loss": 0.8632, + "step": 32354 + }, + { + "epoch": 2.08, + "learning_rate": 1.3078425900083729e-05, + "loss": 0.8647, + "step": 32355 + }, + { + "epoch": 2.08, + "learning_rate": 1.3078211211060781e-05, + "loss": 0.8384, + "step": 32356 + }, + { + "epoch": 2.08, + "learning_rate": 1.307799652203783e-05, + "loss": 0.8572, + "step": 32357 + }, + { + "epoch": 2.08, + "learning_rate": 1.3077781833014879e-05, + "loss": 0.9435, + "step": 32358 + }, + { + "epoch": 2.08, + "learning_rate": 1.307756714399193e-05, + "loss": 0.9072, + "step": 32359 + }, + { + "epoch": 2.08, + "learning_rate": 1.3077352454968979e-05, + "loss": 0.866, + "step": 32360 + }, + { + "epoch": 2.08, + "learning_rate": 1.307713776594603e-05, + "loss": 0.8801, + "step": 32361 + }, + { + "epoch": 2.08, + "learning_rate": 1.3076923076923078e-05, + "loss": 0.9274, + "step": 32362 + }, + { + "epoch": 2.08, + "learning_rate": 1.3076708387900127e-05, + "loss": 0.7948, + "step": 32363 + }, + { + "epoch": 2.08, + "learning_rate": 1.3076493698877178e-05, + "loss": 0.8694, + "step": 32364 + }, + { + "epoch": 2.08, + "learning_rate": 1.3076279009854227e-05, + "loss": 0.859, + "step": 32365 + }, + { + "epoch": 2.08, + "learning_rate": 1.3076064320831276e-05, + "loss": 0.8675, + "step": 32366 + }, + { + "epoch": 2.08, + "learning_rate": 1.3075849631808326e-05, + "loss": 0.8575, + "step": 32367 + }, + { + "epoch": 2.08, + "learning_rate": 1.3075634942785375e-05, + "loss": 0.9341, + "step": 32368 + }, + { + "epoch": 2.08, + "learning_rate": 1.3075420253762426e-05, + "loss": 0.9354, + "step": 32369 + }, + { + "epoch": 2.08, + "learning_rate": 1.3075205564739475e-05, + "loss": 0.9533, + "step": 32370 + }, + { + "epoch": 2.08, + "learning_rate": 1.3074990875716527e-05, + "loss": 0.9762, + "step": 32371 + }, + { + "epoch": 2.08, + "learning_rate": 1.3074776186693576e-05, + "loss": 0.9236, + "step": 32372 + }, + { + "epoch": 2.08, + "learning_rate": 1.3074561497670625e-05, + "loss": 0.9025, + "step": 32373 + }, + { + "epoch": 2.08, + "learning_rate": 1.3074346808647676e-05, + "loss": 0.8892, + "step": 32374 + }, + { + "epoch": 2.08, + "learning_rate": 1.3074132119624725e-05, + "loss": 0.9283, + "step": 32375 + }, + { + "epoch": 2.08, + "learning_rate": 1.3073917430601774e-05, + "loss": 0.8927, + "step": 32376 + }, + { + "epoch": 2.08, + "learning_rate": 1.3073702741578825e-05, + "loss": 0.8522, + "step": 32377 + }, + { + "epoch": 2.08, + "learning_rate": 1.3073488052555874e-05, + "loss": 0.8633, + "step": 32378 + }, + { + "epoch": 2.08, + "learning_rate": 1.3073273363532924e-05, + "loss": 0.9342, + "step": 32379 + }, + { + "epoch": 2.08, + "learning_rate": 1.3073058674509973e-05, + "loss": 0.9328, + "step": 32380 + }, + { + "epoch": 2.08, + "learning_rate": 1.3072843985487022e-05, + "loss": 0.8826, + "step": 32381 + }, + { + "epoch": 2.08, + "learning_rate": 1.3072629296464073e-05, + "loss": 0.9544, + "step": 32382 + }, + { + "epoch": 2.08, + "learning_rate": 1.3072414607441122e-05, + "loss": 0.8928, + "step": 32383 + }, + { + "epoch": 2.08, + "learning_rate": 1.307219991841817e-05, + "loss": 0.9199, + "step": 32384 + }, + { + "epoch": 2.08, + "learning_rate": 1.3071985229395223e-05, + "loss": 0.9112, + "step": 32385 + }, + { + "epoch": 2.08, + "learning_rate": 1.3071770540372272e-05, + "loss": 0.9178, + "step": 32386 + }, + { + "epoch": 2.08, + "learning_rate": 1.3071555851349323e-05, + "loss": 0.8211, + "step": 32387 + }, + { + "epoch": 2.08, + "learning_rate": 1.3071341162326372e-05, + "loss": 0.8811, + "step": 32388 + }, + { + "epoch": 2.08, + "learning_rate": 1.3071126473303422e-05, + "loss": 0.9585, + "step": 32389 + }, + { + "epoch": 2.08, + "learning_rate": 1.3070911784280471e-05, + "loss": 0.943, + "step": 32390 + }, + { + "epoch": 2.08, + "learning_rate": 1.307069709525752e-05, + "loss": 0.9267, + "step": 32391 + }, + { + "epoch": 2.08, + "learning_rate": 1.3070482406234571e-05, + "loss": 0.9449, + "step": 32392 + }, + { + "epoch": 2.08, + "learning_rate": 1.307026771721162e-05, + "loss": 0.8719, + "step": 32393 + }, + { + "epoch": 2.08, + "learning_rate": 1.3070053028188669e-05, + "loss": 0.9093, + "step": 32394 + }, + { + "epoch": 2.08, + "learning_rate": 1.306983833916572e-05, + "loss": 0.9156, + "step": 32395 + }, + { + "epoch": 2.08, + "learning_rate": 1.3069623650142768e-05, + "loss": 0.8429, + "step": 32396 + }, + { + "epoch": 2.08, + "learning_rate": 1.3069408961119817e-05, + "loss": 0.8614, + "step": 32397 + }, + { + "epoch": 2.08, + "learning_rate": 1.3069194272096868e-05, + "loss": 0.899, + "step": 32398 + }, + { + "epoch": 2.08, + "learning_rate": 1.3068979583073917e-05, + "loss": 0.9328, + "step": 32399 + }, + { + "epoch": 2.08, + "learning_rate": 1.306876489405097e-05, + "loss": 0.9434, + "step": 32400 + }, + { + "epoch": 2.08, + "learning_rate": 1.3068550205028018e-05, + "loss": 0.8779, + "step": 32401 + }, + { + "epoch": 2.08, + "learning_rate": 1.3068335516005069e-05, + "loss": 0.8917, + "step": 32402 + }, + { + "epoch": 2.08, + "learning_rate": 1.3068120826982118e-05, + "loss": 0.9275, + "step": 32403 + }, + { + "epoch": 2.08, + "learning_rate": 1.3067906137959167e-05, + "loss": 0.8381, + "step": 32404 + }, + { + "epoch": 2.08, + "learning_rate": 1.3067691448936218e-05, + "loss": 0.8758, + "step": 32405 + }, + { + "epoch": 2.08, + "learning_rate": 1.3067476759913267e-05, + "loss": 0.8593, + "step": 32406 + }, + { + "epoch": 2.08, + "learning_rate": 1.3067262070890316e-05, + "loss": 0.9036, + "step": 32407 + }, + { + "epoch": 2.09, + "learning_rate": 1.3067047381867366e-05, + "loss": 0.949, + "step": 32408 + }, + { + "epoch": 2.09, + "learning_rate": 1.3066832692844415e-05, + "loss": 0.8688, + "step": 32409 + }, + { + "epoch": 2.09, + "learning_rate": 1.3066618003821466e-05, + "loss": 0.9092, + "step": 32410 + }, + { + "epoch": 2.09, + "learning_rate": 1.3066403314798515e-05, + "loss": 0.9466, + "step": 32411 + }, + { + "epoch": 2.09, + "learning_rate": 1.3066188625775564e-05, + "loss": 0.8837, + "step": 32412 + }, + { + "epoch": 2.09, + "learning_rate": 1.3065973936752614e-05, + "loss": 0.9427, + "step": 32413 + }, + { + "epoch": 2.09, + "learning_rate": 1.3065759247729665e-05, + "loss": 0.8285, + "step": 32414 + }, + { + "epoch": 2.09, + "learning_rate": 1.3065544558706716e-05, + "loss": 0.8812, + "step": 32415 + }, + { + "epoch": 2.09, + "learning_rate": 1.3065329869683765e-05, + "loss": 0.8893, + "step": 32416 + }, + { + "epoch": 2.09, + "learning_rate": 1.3065115180660814e-05, + "loss": 0.8621, + "step": 32417 + }, + { + "epoch": 2.09, + "learning_rate": 1.3064900491637864e-05, + "loss": 0.8888, + "step": 32418 + }, + { + "epoch": 2.09, + "learning_rate": 1.3064685802614913e-05, + "loss": 0.9485, + "step": 32419 + }, + { + "epoch": 2.09, + "learning_rate": 1.3064471113591964e-05, + "loss": 0.9077, + "step": 32420 + }, + { + "epoch": 2.09, + "learning_rate": 1.3064256424569013e-05, + "loss": 0.9606, + "step": 32421 + }, + { + "epoch": 2.09, + "learning_rate": 1.3064041735546062e-05, + "loss": 0.9509, + "step": 32422 + }, + { + "epoch": 2.09, + "learning_rate": 1.3063827046523113e-05, + "loss": 0.8822, + "step": 32423 + }, + { + "epoch": 2.09, + "learning_rate": 1.3063612357500162e-05, + "loss": 0.966, + "step": 32424 + }, + { + "epoch": 2.09, + "learning_rate": 1.306339766847721e-05, + "loss": 0.836, + "step": 32425 + }, + { + "epoch": 2.09, + "learning_rate": 1.3063182979454261e-05, + "loss": 0.8917, + "step": 32426 + }, + { + "epoch": 2.09, + "learning_rate": 1.306296829043131e-05, + "loss": 0.8594, + "step": 32427 + }, + { + "epoch": 2.09, + "learning_rate": 1.306275360140836e-05, + "loss": 0.8533, + "step": 32428 + }, + { + "epoch": 2.09, + "learning_rate": 1.3062538912385411e-05, + "loss": 0.9084, + "step": 32429 + }, + { + "epoch": 2.09, + "learning_rate": 1.3062324223362462e-05, + "loss": 0.8685, + "step": 32430 + }, + { + "epoch": 2.09, + "learning_rate": 1.3062109534339511e-05, + "loss": 0.8692, + "step": 32431 + }, + { + "epoch": 2.09, + "learning_rate": 1.306189484531656e-05, + "loss": 0.8722, + "step": 32432 + }, + { + "epoch": 2.09, + "learning_rate": 1.306168015629361e-05, + "loss": 0.8733, + "step": 32433 + }, + { + "epoch": 2.09, + "learning_rate": 1.306146546727066e-05, + "loss": 0.9451, + "step": 32434 + }, + { + "epoch": 2.09, + "learning_rate": 1.3061250778247709e-05, + "loss": 0.9307, + "step": 32435 + }, + { + "epoch": 2.09, + "learning_rate": 1.306103608922476e-05, + "loss": 0.8794, + "step": 32436 + }, + { + "epoch": 2.09, + "learning_rate": 1.3060821400201808e-05, + "loss": 0.8649, + "step": 32437 + }, + { + "epoch": 2.09, + "learning_rate": 1.3060606711178859e-05, + "loss": 0.8399, + "step": 32438 + }, + { + "epoch": 2.09, + "learning_rate": 1.3060392022155908e-05, + "loss": 0.8439, + "step": 32439 + }, + { + "epoch": 2.09, + "learning_rate": 1.3060177333132957e-05, + "loss": 0.9118, + "step": 32440 + }, + { + "epoch": 2.09, + "learning_rate": 1.3059962644110007e-05, + "loss": 0.8027, + "step": 32441 + }, + { + "epoch": 2.09, + "learning_rate": 1.3059747955087056e-05, + "loss": 0.8496, + "step": 32442 + }, + { + "epoch": 2.09, + "learning_rate": 1.3059533266064109e-05, + "loss": 0.8967, + "step": 32443 + }, + { + "epoch": 2.09, + "learning_rate": 1.3059318577041158e-05, + "loss": 0.9023, + "step": 32444 + }, + { + "epoch": 2.09, + "learning_rate": 1.3059103888018207e-05, + "loss": 0.9185, + "step": 32445 + }, + { + "epoch": 2.09, + "learning_rate": 1.3058889198995257e-05, + "loss": 0.8439, + "step": 32446 + }, + { + "epoch": 2.09, + "learning_rate": 1.3058674509972306e-05, + "loss": 0.8564, + "step": 32447 + }, + { + "epoch": 2.09, + "learning_rate": 1.3058459820949357e-05, + "loss": 0.8963, + "step": 32448 + }, + { + "epoch": 2.09, + "learning_rate": 1.3058245131926406e-05, + "loss": 0.8565, + "step": 32449 + }, + { + "epoch": 2.09, + "learning_rate": 1.3058030442903455e-05, + "loss": 0.9159, + "step": 32450 + }, + { + "epoch": 2.09, + "learning_rate": 1.3057815753880506e-05, + "loss": 0.7857, + "step": 32451 + }, + { + "epoch": 2.09, + "learning_rate": 1.3057601064857555e-05, + "loss": 0.9382, + "step": 32452 + }, + { + "epoch": 2.09, + "learning_rate": 1.3057386375834604e-05, + "loss": 0.8893, + "step": 32453 + }, + { + "epoch": 2.09, + "learning_rate": 1.3057171686811654e-05, + "loss": 0.8695, + "step": 32454 + }, + { + "epoch": 2.09, + "learning_rate": 1.3056956997788703e-05, + "loss": 0.8927, + "step": 32455 + }, + { + "epoch": 2.09, + "learning_rate": 1.3056742308765754e-05, + "loss": 0.8183, + "step": 32456 + }, + { + "epoch": 2.09, + "learning_rate": 1.3056527619742803e-05, + "loss": 0.8767, + "step": 32457 + }, + { + "epoch": 2.09, + "learning_rate": 1.3056312930719855e-05, + "loss": 0.8614, + "step": 32458 + }, + { + "epoch": 2.09, + "learning_rate": 1.3056098241696904e-05, + "loss": 0.8988, + "step": 32459 + }, + { + "epoch": 2.09, + "learning_rate": 1.3055883552673953e-05, + "loss": 0.9674, + "step": 32460 + }, + { + "epoch": 2.09, + "learning_rate": 1.3055668863651004e-05, + "loss": 0.9426, + "step": 32461 + }, + { + "epoch": 2.09, + "learning_rate": 1.3055454174628053e-05, + "loss": 0.8403, + "step": 32462 + }, + { + "epoch": 2.09, + "learning_rate": 1.3055239485605102e-05, + "loss": 0.8726, + "step": 32463 + }, + { + "epoch": 2.09, + "learning_rate": 1.3055024796582152e-05, + "loss": 0.9563, + "step": 32464 + }, + { + "epoch": 2.09, + "learning_rate": 1.3054810107559201e-05, + "loss": 0.9715, + "step": 32465 + }, + { + "epoch": 2.09, + "learning_rate": 1.3054595418536252e-05, + "loss": 0.9107, + "step": 32466 + }, + { + "epoch": 2.09, + "learning_rate": 1.3054380729513301e-05, + "loss": 0.8851, + "step": 32467 + }, + { + "epoch": 2.09, + "learning_rate": 1.305416604049035e-05, + "loss": 0.8392, + "step": 32468 + }, + { + "epoch": 2.09, + "learning_rate": 1.30539513514674e-05, + "loss": 0.9544, + "step": 32469 + }, + { + "epoch": 2.09, + "learning_rate": 1.305373666244445e-05, + "loss": 0.8713, + "step": 32470 + }, + { + "epoch": 2.09, + "learning_rate": 1.3053521973421498e-05, + "loss": 0.9505, + "step": 32471 + }, + { + "epoch": 2.09, + "learning_rate": 1.3053307284398549e-05, + "loss": 0.8708, + "step": 32472 + }, + { + "epoch": 2.09, + "learning_rate": 1.30530925953756e-05, + "loss": 0.9407, + "step": 32473 + }, + { + "epoch": 2.09, + "learning_rate": 1.305287790635265e-05, + "loss": 0.9139, + "step": 32474 + }, + { + "epoch": 2.09, + "learning_rate": 1.30526632173297e-05, + "loss": 0.9088, + "step": 32475 + }, + { + "epoch": 2.09, + "learning_rate": 1.305244852830675e-05, + "loss": 0.9201, + "step": 32476 + }, + { + "epoch": 2.09, + "learning_rate": 1.3052233839283799e-05, + "loss": 0.8256, + "step": 32477 + }, + { + "epoch": 2.09, + "learning_rate": 1.3052019150260848e-05, + "loss": 0.8866, + "step": 32478 + }, + { + "epoch": 2.09, + "learning_rate": 1.3051804461237899e-05, + "loss": 0.9381, + "step": 32479 + }, + { + "epoch": 2.09, + "learning_rate": 1.3051589772214948e-05, + "loss": 0.9632, + "step": 32480 + }, + { + "epoch": 2.09, + "learning_rate": 1.3051375083191997e-05, + "loss": 0.8575, + "step": 32481 + }, + { + "epoch": 2.09, + "learning_rate": 1.3051160394169047e-05, + "loss": 0.8791, + "step": 32482 + }, + { + "epoch": 2.09, + "learning_rate": 1.3050945705146096e-05, + "loss": 0.9478, + "step": 32483 + }, + { + "epoch": 2.09, + "learning_rate": 1.3050731016123145e-05, + "loss": 0.8566, + "step": 32484 + }, + { + "epoch": 2.09, + "learning_rate": 1.3050516327100196e-05, + "loss": 0.9305, + "step": 32485 + }, + { + "epoch": 2.09, + "learning_rate": 1.3050301638077245e-05, + "loss": 0.8731, + "step": 32486 + }, + { + "epoch": 2.09, + "learning_rate": 1.3050086949054297e-05, + "loss": 0.8613, + "step": 32487 + }, + { + "epoch": 2.09, + "learning_rate": 1.3049872260031346e-05, + "loss": 0.9009, + "step": 32488 + }, + { + "epoch": 2.09, + "learning_rate": 1.3049657571008397e-05, + "loss": 0.9413, + "step": 32489 + }, + { + "epoch": 2.09, + "learning_rate": 1.3049442881985446e-05, + "loss": 0.893, + "step": 32490 + }, + { + "epoch": 2.09, + "learning_rate": 1.3049228192962495e-05, + "loss": 0.8671, + "step": 32491 + }, + { + "epoch": 2.09, + "learning_rate": 1.3049013503939545e-05, + "loss": 0.9962, + "step": 32492 + }, + { + "epoch": 2.09, + "learning_rate": 1.3048798814916594e-05, + "loss": 0.9378, + "step": 32493 + }, + { + "epoch": 2.09, + "learning_rate": 1.3048584125893643e-05, + "loss": 0.9499, + "step": 32494 + }, + { + "epoch": 2.09, + "learning_rate": 1.3048369436870694e-05, + "loss": 0.8362, + "step": 32495 + }, + { + "epoch": 2.09, + "learning_rate": 1.3048154747847743e-05, + "loss": 0.8377, + "step": 32496 + }, + { + "epoch": 2.09, + "learning_rate": 1.3047940058824794e-05, + "loss": 0.8634, + "step": 32497 + }, + { + "epoch": 2.09, + "learning_rate": 1.3047725369801843e-05, + "loss": 0.915, + "step": 32498 + }, + { + "epoch": 2.09, + "learning_rate": 1.3047510680778891e-05, + "loss": 0.882, + "step": 32499 + }, + { + "epoch": 2.09, + "learning_rate": 1.3047295991755942e-05, + "loss": 0.8541, + "step": 32500 + }, + { + "epoch": 2.09, + "learning_rate": 1.3047081302732991e-05, + "loss": 0.9266, + "step": 32501 + }, + { + "epoch": 2.09, + "learning_rate": 1.3046866613710043e-05, + "loss": 0.8908, + "step": 32502 + }, + { + "epoch": 2.09, + "learning_rate": 1.3046651924687092e-05, + "loss": 0.8365, + "step": 32503 + }, + { + "epoch": 2.09, + "learning_rate": 1.3046437235664141e-05, + "loss": 0.8941, + "step": 32504 + }, + { + "epoch": 2.09, + "learning_rate": 1.3046222546641192e-05, + "loss": 0.8034, + "step": 32505 + }, + { + "epoch": 2.09, + "learning_rate": 1.3046007857618241e-05, + "loss": 0.8648, + "step": 32506 + }, + { + "epoch": 2.09, + "learning_rate": 1.3045793168595292e-05, + "loss": 0.8615, + "step": 32507 + }, + { + "epoch": 2.09, + "learning_rate": 1.304557847957234e-05, + "loss": 0.8464, + "step": 32508 + }, + { + "epoch": 2.09, + "learning_rate": 1.304536379054939e-05, + "loss": 0.9846, + "step": 32509 + }, + { + "epoch": 2.09, + "learning_rate": 1.304514910152644e-05, + "loss": 0.8424, + "step": 32510 + }, + { + "epoch": 2.09, + "learning_rate": 1.304493441250349e-05, + "loss": 0.8434, + "step": 32511 + }, + { + "epoch": 2.09, + "learning_rate": 1.3044719723480538e-05, + "loss": 0.8655, + "step": 32512 + }, + { + "epoch": 2.09, + "learning_rate": 1.3044505034457589e-05, + "loss": 0.9052, + "step": 32513 + }, + { + "epoch": 2.09, + "learning_rate": 1.3044290345434638e-05, + "loss": 0.7937, + "step": 32514 + }, + { + "epoch": 2.09, + "learning_rate": 1.3044075656411688e-05, + "loss": 0.8833, + "step": 32515 + }, + { + "epoch": 2.09, + "learning_rate": 1.3043860967388739e-05, + "loss": 0.9257, + "step": 32516 + }, + { + "epoch": 2.09, + "learning_rate": 1.304364627836579e-05, + "loss": 0.8546, + "step": 32517 + }, + { + "epoch": 2.09, + "learning_rate": 1.3043431589342839e-05, + "loss": 0.9166, + "step": 32518 + }, + { + "epoch": 2.09, + "learning_rate": 1.3043431589342839e-05, + "loss": 0.8873, + "step": 32519 + }, + { + "epoch": 2.09, + "learning_rate": 1.3043216900319888e-05, + "loss": 0.9038, + "step": 32520 + }, + { + "epoch": 2.09, + "learning_rate": 1.3043002211296938e-05, + "loss": 0.8479, + "step": 32521 + }, + { + "epoch": 2.09, + "learning_rate": 1.3042787522273987e-05, + "loss": 0.9174, + "step": 32522 + }, + { + "epoch": 2.09, + "learning_rate": 1.3042572833251036e-05, + "loss": 0.8903, + "step": 32523 + }, + { + "epoch": 2.09, + "learning_rate": 1.3042358144228087e-05, + "loss": 0.8864, + "step": 32524 + }, + { + "epoch": 2.09, + "learning_rate": 1.3042143455205136e-05, + "loss": 0.9567, + "step": 32525 + }, + { + "epoch": 2.09, + "learning_rate": 1.3041928766182187e-05, + "loss": 0.9317, + "step": 32526 + }, + { + "epoch": 2.09, + "learning_rate": 1.3041714077159236e-05, + "loss": 0.8336, + "step": 32527 + }, + { + "epoch": 2.09, + "learning_rate": 1.3041499388136285e-05, + "loss": 0.9359, + "step": 32528 + }, + { + "epoch": 2.09, + "learning_rate": 1.3041284699113335e-05, + "loss": 0.8481, + "step": 32529 + }, + { + "epoch": 2.09, + "learning_rate": 1.3041070010090384e-05, + "loss": 0.799, + "step": 32530 + }, + { + "epoch": 2.09, + "learning_rate": 1.3040855321067433e-05, + "loss": 0.8664, + "step": 32531 + }, + { + "epoch": 2.09, + "learning_rate": 1.3040640632044485e-05, + "loss": 0.8991, + "step": 32532 + }, + { + "epoch": 2.09, + "learning_rate": 1.3040425943021534e-05, + "loss": 0.8655, + "step": 32533 + }, + { + "epoch": 2.09, + "learning_rate": 1.3040211253998585e-05, + "loss": 0.8428, + "step": 32534 + }, + { + "epoch": 2.09, + "learning_rate": 1.3039996564975634e-05, + "loss": 0.9359, + "step": 32535 + }, + { + "epoch": 2.09, + "learning_rate": 1.3039781875952685e-05, + "loss": 0.9087, + "step": 32536 + }, + { + "epoch": 2.09, + "learning_rate": 1.3039567186929734e-05, + "loss": 0.8979, + "step": 32537 + }, + { + "epoch": 2.09, + "learning_rate": 1.3039352497906783e-05, + "loss": 0.9129, + "step": 32538 + }, + { + "epoch": 2.09, + "learning_rate": 1.3039137808883833e-05, + "loss": 0.9308, + "step": 32539 + }, + { + "epoch": 2.09, + "learning_rate": 1.3038923119860882e-05, + "loss": 0.9337, + "step": 32540 + }, + { + "epoch": 2.09, + "learning_rate": 1.3038708430837931e-05, + "loss": 0.9235, + "step": 32541 + }, + { + "epoch": 2.09, + "learning_rate": 1.3038493741814982e-05, + "loss": 0.8753, + "step": 32542 + }, + { + "epoch": 2.09, + "learning_rate": 1.3038279052792031e-05, + "loss": 0.8406, + "step": 32543 + }, + { + "epoch": 2.09, + "learning_rate": 1.3038064363769082e-05, + "loss": 0.9006, + "step": 32544 + }, + { + "epoch": 2.09, + "learning_rate": 1.303784967474613e-05, + "loss": 0.8172, + "step": 32545 + }, + { + "epoch": 2.09, + "learning_rate": 1.3037634985723183e-05, + "loss": 0.9176, + "step": 32546 + }, + { + "epoch": 2.09, + "learning_rate": 1.3037420296700232e-05, + "loss": 0.8894, + "step": 32547 + }, + { + "epoch": 2.09, + "learning_rate": 1.303720560767728e-05, + "loss": 0.9089, + "step": 32548 + }, + { + "epoch": 2.09, + "learning_rate": 1.3036990918654331e-05, + "loss": 0.9426, + "step": 32549 + }, + { + "epoch": 2.09, + "learning_rate": 1.303677622963138e-05, + "loss": 0.8855, + "step": 32550 + }, + { + "epoch": 2.09, + "learning_rate": 1.303656154060843e-05, + "loss": 0.8857, + "step": 32551 + }, + { + "epoch": 2.09, + "learning_rate": 1.303634685158548e-05, + "loss": 0.8969, + "step": 32552 + }, + { + "epoch": 2.09, + "learning_rate": 1.3036132162562529e-05, + "loss": 0.8502, + "step": 32553 + }, + { + "epoch": 2.09, + "learning_rate": 1.303591747353958e-05, + "loss": 0.8322, + "step": 32554 + }, + { + "epoch": 2.09, + "learning_rate": 1.3035702784516629e-05, + "loss": 0.8633, + "step": 32555 + }, + { + "epoch": 2.09, + "learning_rate": 1.3035488095493678e-05, + "loss": 0.8954, + "step": 32556 + }, + { + "epoch": 2.09, + "learning_rate": 1.3035273406470728e-05, + "loss": 0.9305, + "step": 32557 + }, + { + "epoch": 2.09, + "learning_rate": 1.3035058717447777e-05, + "loss": 0.8577, + "step": 32558 + }, + { + "epoch": 2.09, + "learning_rate": 1.3034844028424826e-05, + "loss": 0.849, + "step": 32559 + }, + { + "epoch": 2.09, + "learning_rate": 1.3034629339401877e-05, + "loss": 0.8938, + "step": 32560 + }, + { + "epoch": 2.09, + "learning_rate": 1.3034414650378927e-05, + "loss": 0.8991, + "step": 32561 + }, + { + "epoch": 2.09, + "learning_rate": 1.3034199961355978e-05, + "loss": 0.8759, + "step": 32562 + }, + { + "epoch": 2.1, + "learning_rate": 1.3033985272333027e-05, + "loss": 0.8958, + "step": 32563 + }, + { + "epoch": 2.1, + "learning_rate": 1.3033770583310078e-05, + "loss": 0.8533, + "step": 32564 + }, + { + "epoch": 2.1, + "learning_rate": 1.3033555894287127e-05, + "loss": 0.8597, + "step": 32565 + }, + { + "epoch": 2.1, + "learning_rate": 1.3033341205264176e-05, + "loss": 0.9156, + "step": 32566 + }, + { + "epoch": 2.1, + "learning_rate": 1.3033126516241226e-05, + "loss": 0.8381, + "step": 32567 + }, + { + "epoch": 2.1, + "learning_rate": 1.3032911827218275e-05, + "loss": 0.8386, + "step": 32568 + }, + { + "epoch": 2.1, + "learning_rate": 1.3032697138195324e-05, + "loss": 0.94, + "step": 32569 + }, + { + "epoch": 2.1, + "learning_rate": 1.3032482449172375e-05, + "loss": 0.8833, + "step": 32570 + }, + { + "epoch": 2.1, + "learning_rate": 1.3032267760149424e-05, + "loss": 0.8748, + "step": 32571 + }, + { + "epoch": 2.1, + "learning_rate": 1.3032053071126473e-05, + "loss": 0.8826, + "step": 32572 + }, + { + "epoch": 2.1, + "learning_rate": 1.3031838382103524e-05, + "loss": 0.9288, + "step": 32573 + }, + { + "epoch": 2.1, + "learning_rate": 1.3031623693080572e-05, + "loss": 0.8919, + "step": 32574 + }, + { + "epoch": 2.1, + "learning_rate": 1.3031409004057625e-05, + "loss": 0.8199, + "step": 32575 + }, + { + "epoch": 2.1, + "learning_rate": 1.3031194315034674e-05, + "loss": 0.887, + "step": 32576 + }, + { + "epoch": 2.1, + "learning_rate": 1.3030979626011724e-05, + "loss": 0.8395, + "step": 32577 + }, + { + "epoch": 2.1, + "learning_rate": 1.3030764936988773e-05, + "loss": 0.8602, + "step": 32578 + }, + { + "epoch": 2.1, + "learning_rate": 1.3030550247965822e-05, + "loss": 0.8722, + "step": 32579 + }, + { + "epoch": 2.1, + "learning_rate": 1.3030335558942873e-05, + "loss": 0.8403, + "step": 32580 + }, + { + "epoch": 2.1, + "learning_rate": 1.3030120869919922e-05, + "loss": 0.8628, + "step": 32581 + }, + { + "epoch": 2.1, + "learning_rate": 1.3029906180896971e-05, + "loss": 0.9501, + "step": 32582 + }, + { + "epoch": 2.1, + "learning_rate": 1.3029691491874022e-05, + "loss": 0.9001, + "step": 32583 + }, + { + "epoch": 2.1, + "learning_rate": 1.302947680285107e-05, + "loss": 0.8835, + "step": 32584 + }, + { + "epoch": 2.1, + "learning_rate": 1.3029262113828121e-05, + "loss": 0.8643, + "step": 32585 + }, + { + "epoch": 2.1, + "learning_rate": 1.302904742480517e-05, + "loss": 0.9004, + "step": 32586 + }, + { + "epoch": 2.1, + "learning_rate": 1.302883273578222e-05, + "loss": 0.8991, + "step": 32587 + }, + { + "epoch": 2.1, + "learning_rate": 1.302861804675927e-05, + "loss": 0.8895, + "step": 32588 + }, + { + "epoch": 2.1, + "learning_rate": 1.3028403357736319e-05, + "loss": 0.9603, + "step": 32589 + }, + { + "epoch": 2.1, + "learning_rate": 1.3028188668713371e-05, + "loss": 0.8991, + "step": 32590 + }, + { + "epoch": 2.1, + "learning_rate": 1.302797397969042e-05, + "loss": 0.909, + "step": 32591 + }, + { + "epoch": 2.1, + "learning_rate": 1.3027759290667469e-05, + "loss": 0.8633, + "step": 32592 + }, + { + "epoch": 2.1, + "learning_rate": 1.302754460164452e-05, + "loss": 0.8272, + "step": 32593 + }, + { + "epoch": 2.1, + "learning_rate": 1.3027329912621569e-05, + "loss": 0.8597, + "step": 32594 + }, + { + "epoch": 2.1, + "learning_rate": 1.302711522359862e-05, + "loss": 0.904, + "step": 32595 + }, + { + "epoch": 2.1, + "learning_rate": 1.3026900534575668e-05, + "loss": 0.9254, + "step": 32596 + }, + { + "epoch": 2.1, + "learning_rate": 1.3026685845552717e-05, + "loss": 0.941, + "step": 32597 + }, + { + "epoch": 2.1, + "learning_rate": 1.3026471156529768e-05, + "loss": 0.8772, + "step": 32598 + }, + { + "epoch": 2.1, + "learning_rate": 1.3026256467506817e-05, + "loss": 0.9288, + "step": 32599 + }, + { + "epoch": 2.1, + "learning_rate": 1.3026041778483866e-05, + "loss": 0.9181, + "step": 32600 + }, + { + "epoch": 2.1, + "learning_rate": 1.3025827089460917e-05, + "loss": 0.917, + "step": 32601 + }, + { + "epoch": 2.1, + "learning_rate": 1.3025612400437966e-05, + "loss": 0.9108, + "step": 32602 + }, + { + "epoch": 2.1, + "learning_rate": 1.3025397711415016e-05, + "loss": 0.887, + "step": 32603 + }, + { + "epoch": 2.1, + "learning_rate": 1.3025183022392067e-05, + "loss": 0.939, + "step": 32604 + }, + { + "epoch": 2.1, + "learning_rate": 1.3024968333369118e-05, + "loss": 0.9183, + "step": 32605 + }, + { + "epoch": 2.1, + "learning_rate": 1.3024753644346166e-05, + "loss": 0.897, + "step": 32606 + }, + { + "epoch": 2.1, + "learning_rate": 1.3024538955323215e-05, + "loss": 0.8894, + "step": 32607 + }, + { + "epoch": 2.1, + "learning_rate": 1.3024324266300266e-05, + "loss": 0.9248, + "step": 32608 + }, + { + "epoch": 2.1, + "learning_rate": 1.3024109577277315e-05, + "loss": 0.8504, + "step": 32609 + }, + { + "epoch": 2.1, + "learning_rate": 1.3023894888254364e-05, + "loss": 0.9201, + "step": 32610 + }, + { + "epoch": 2.1, + "learning_rate": 1.3023680199231415e-05, + "loss": 0.913, + "step": 32611 + }, + { + "epoch": 2.1, + "learning_rate": 1.3023465510208464e-05, + "loss": 0.8743, + "step": 32612 + }, + { + "epoch": 2.1, + "learning_rate": 1.3023250821185514e-05, + "loss": 0.9336, + "step": 32613 + }, + { + "epoch": 2.1, + "learning_rate": 1.3023036132162563e-05, + "loss": 0.9492, + "step": 32614 + }, + { + "epoch": 2.1, + "learning_rate": 1.3022821443139612e-05, + "loss": 0.9068, + "step": 32615 + }, + { + "epoch": 2.1, + "learning_rate": 1.3022606754116663e-05, + "loss": 0.8624, + "step": 32616 + }, + { + "epoch": 2.1, + "learning_rate": 1.3022392065093712e-05, + "loss": 0.9142, + "step": 32617 + }, + { + "epoch": 2.1, + "learning_rate": 1.302217737607076e-05, + "loss": 0.98, + "step": 32618 + }, + { + "epoch": 2.1, + "learning_rate": 1.3021962687047813e-05, + "loss": 0.8874, + "step": 32619 + }, + { + "epoch": 2.1, + "learning_rate": 1.3021747998024862e-05, + "loss": 0.8023, + "step": 32620 + }, + { + "epoch": 2.1, + "learning_rate": 1.3021533309001913e-05, + "loss": 0.8691, + "step": 32621 + }, + { + "epoch": 2.1, + "learning_rate": 1.3021318619978962e-05, + "loss": 0.9285, + "step": 32622 + }, + { + "epoch": 2.1, + "learning_rate": 1.3021103930956012e-05, + "loss": 0.8811, + "step": 32623 + }, + { + "epoch": 2.1, + "learning_rate": 1.3020889241933061e-05, + "loss": 0.8729, + "step": 32624 + }, + { + "epoch": 2.1, + "learning_rate": 1.302067455291011e-05, + "loss": 0.8567, + "step": 32625 + }, + { + "epoch": 2.1, + "learning_rate": 1.3020459863887161e-05, + "loss": 0.8037, + "step": 32626 + }, + { + "epoch": 2.1, + "learning_rate": 1.302024517486421e-05, + "loss": 0.8911, + "step": 32627 + }, + { + "epoch": 2.1, + "learning_rate": 1.3020030485841259e-05, + "loss": 0.9059, + "step": 32628 + }, + { + "epoch": 2.1, + "learning_rate": 1.301981579681831e-05, + "loss": 0.8788, + "step": 32629 + }, + { + "epoch": 2.1, + "learning_rate": 1.3019601107795359e-05, + "loss": 0.9243, + "step": 32630 + }, + { + "epoch": 2.1, + "learning_rate": 1.301938641877241e-05, + "loss": 0.8906, + "step": 32631 + }, + { + "epoch": 2.1, + "learning_rate": 1.3019171729749458e-05, + "loss": 0.9097, + "step": 32632 + }, + { + "epoch": 2.1, + "learning_rate": 1.301895704072651e-05, + "loss": 0.9099, + "step": 32633 + }, + { + "epoch": 2.1, + "learning_rate": 1.301874235170356e-05, + "loss": 0.9149, + "step": 32634 + }, + { + "epoch": 2.1, + "learning_rate": 1.3018527662680608e-05, + "loss": 0.9508, + "step": 32635 + }, + { + "epoch": 2.1, + "learning_rate": 1.3018312973657659e-05, + "loss": 0.943, + "step": 32636 + }, + { + "epoch": 2.1, + "learning_rate": 1.3018098284634708e-05, + "loss": 0.8454, + "step": 32637 + }, + { + "epoch": 2.1, + "learning_rate": 1.3017883595611757e-05, + "loss": 0.8731, + "step": 32638 + }, + { + "epoch": 2.1, + "learning_rate": 1.3017668906588808e-05, + "loss": 0.9083, + "step": 32639 + }, + { + "epoch": 2.1, + "learning_rate": 1.3017454217565857e-05, + "loss": 0.937, + "step": 32640 + }, + { + "epoch": 2.1, + "learning_rate": 1.3017239528542907e-05, + "loss": 0.9514, + "step": 32641 + }, + { + "epoch": 2.1, + "learning_rate": 1.3017024839519956e-05, + "loss": 0.8842, + "step": 32642 + }, + { + "epoch": 2.1, + "learning_rate": 1.3016810150497005e-05, + "loss": 0.837, + "step": 32643 + }, + { + "epoch": 2.1, + "learning_rate": 1.3016595461474056e-05, + "loss": 0.9335, + "step": 32644 + }, + { + "epoch": 2.1, + "learning_rate": 1.3016380772451105e-05, + "loss": 0.8809, + "step": 32645 + }, + { + "epoch": 2.1, + "learning_rate": 1.3016166083428154e-05, + "loss": 0.8054, + "step": 32646 + }, + { + "epoch": 2.1, + "learning_rate": 1.3015951394405205e-05, + "loss": 0.9004, + "step": 32647 + }, + { + "epoch": 2.1, + "learning_rate": 1.3015736705382255e-05, + "loss": 0.8638, + "step": 32648 + }, + { + "epoch": 2.1, + "learning_rate": 1.3015522016359306e-05, + "loss": 0.8716, + "step": 32649 + }, + { + "epoch": 2.1, + "learning_rate": 1.3015307327336355e-05, + "loss": 0.8631, + "step": 32650 + }, + { + "epoch": 2.1, + "learning_rate": 1.3015092638313405e-05, + "loss": 0.9253, + "step": 32651 + }, + { + "epoch": 2.1, + "learning_rate": 1.3014877949290454e-05, + "loss": 0.9151, + "step": 32652 + }, + { + "epoch": 2.1, + "learning_rate": 1.3014663260267503e-05, + "loss": 0.9552, + "step": 32653 + }, + { + "epoch": 2.1, + "learning_rate": 1.3014448571244554e-05, + "loss": 0.9236, + "step": 32654 + }, + { + "epoch": 2.1, + "learning_rate": 1.3014233882221603e-05, + "loss": 0.9322, + "step": 32655 + }, + { + "epoch": 2.1, + "learning_rate": 1.3014019193198652e-05, + "loss": 0.9376, + "step": 32656 + }, + { + "epoch": 2.1, + "learning_rate": 1.3013804504175703e-05, + "loss": 0.9568, + "step": 32657 + }, + { + "epoch": 2.1, + "learning_rate": 1.3013589815152752e-05, + "loss": 0.9458, + "step": 32658 + }, + { + "epoch": 2.1, + "learning_rate": 1.30133751261298e-05, + "loss": 0.8738, + "step": 32659 + }, + { + "epoch": 2.1, + "learning_rate": 1.3013160437106851e-05, + "loss": 0.8937, + "step": 32660 + }, + { + "epoch": 2.1, + "learning_rate": 1.30129457480839e-05, + "loss": 0.9479, + "step": 32661 + }, + { + "epoch": 2.1, + "learning_rate": 1.3012731059060953e-05, + "loss": 0.8488, + "step": 32662 + }, + { + "epoch": 2.1, + "learning_rate": 1.3012516370038002e-05, + "loss": 0.9833, + "step": 32663 + }, + { + "epoch": 2.1, + "learning_rate": 1.3012301681015052e-05, + "loss": 0.8903, + "step": 32664 + }, + { + "epoch": 2.1, + "learning_rate": 1.3012086991992101e-05, + "loss": 0.8726, + "step": 32665 + }, + { + "epoch": 2.1, + "learning_rate": 1.301187230296915e-05, + "loss": 0.8861, + "step": 32666 + }, + { + "epoch": 2.1, + "learning_rate": 1.30116576139462e-05, + "loss": 0.8426, + "step": 32667 + }, + { + "epoch": 2.1, + "learning_rate": 1.301144292492325e-05, + "loss": 0.9108, + "step": 32668 + }, + { + "epoch": 2.1, + "learning_rate": 1.3011228235900299e-05, + "loss": 0.8921, + "step": 32669 + }, + { + "epoch": 2.1, + "learning_rate": 1.301101354687735e-05, + "loss": 0.9652, + "step": 32670 + }, + { + "epoch": 2.1, + "learning_rate": 1.3010798857854398e-05, + "loss": 0.9032, + "step": 32671 + }, + { + "epoch": 2.1, + "learning_rate": 1.3010584168831449e-05, + "loss": 0.8339, + "step": 32672 + }, + { + "epoch": 2.1, + "learning_rate": 1.3010369479808498e-05, + "loss": 0.8877, + "step": 32673 + }, + { + "epoch": 2.1, + "learning_rate": 1.3010154790785547e-05, + "loss": 0.9211, + "step": 32674 + }, + { + "epoch": 2.1, + "learning_rate": 1.3009940101762598e-05, + "loss": 0.9125, + "step": 32675 + }, + { + "epoch": 2.1, + "learning_rate": 1.3009725412739647e-05, + "loss": 0.8799, + "step": 32676 + }, + { + "epoch": 2.1, + "learning_rate": 1.3009510723716699e-05, + "loss": 0.8487, + "step": 32677 + }, + { + "epoch": 2.1, + "learning_rate": 1.3009296034693748e-05, + "loss": 0.8298, + "step": 32678 + }, + { + "epoch": 2.1, + "learning_rate": 1.3009081345670797e-05, + "loss": 0.9151, + "step": 32679 + }, + { + "epoch": 2.1, + "learning_rate": 1.3008866656647847e-05, + "loss": 0.9116, + "step": 32680 + }, + { + "epoch": 2.1, + "learning_rate": 1.3008651967624896e-05, + "loss": 0.914, + "step": 32681 + }, + { + "epoch": 2.1, + "learning_rate": 1.3008437278601947e-05, + "loss": 0.8673, + "step": 32682 + }, + { + "epoch": 2.1, + "learning_rate": 1.3008222589578996e-05, + "loss": 0.9361, + "step": 32683 + }, + { + "epoch": 2.1, + "learning_rate": 1.3008007900556045e-05, + "loss": 0.8825, + "step": 32684 + }, + { + "epoch": 2.1, + "learning_rate": 1.3007793211533096e-05, + "loss": 0.922, + "step": 32685 + }, + { + "epoch": 2.1, + "learning_rate": 1.3007578522510145e-05, + "loss": 0.862, + "step": 32686 + }, + { + "epoch": 2.1, + "learning_rate": 1.3007363833487194e-05, + "loss": 0.9335, + "step": 32687 + }, + { + "epoch": 2.1, + "learning_rate": 1.3007149144464244e-05, + "loss": 0.8792, + "step": 32688 + }, + { + "epoch": 2.1, + "learning_rate": 1.3006934455441293e-05, + "loss": 0.8637, + "step": 32689 + }, + { + "epoch": 2.1, + "learning_rate": 1.3006719766418344e-05, + "loss": 0.8736, + "step": 32690 + }, + { + "epoch": 2.1, + "learning_rate": 1.3006505077395395e-05, + "loss": 0.834, + "step": 32691 + }, + { + "epoch": 2.1, + "learning_rate": 1.3006290388372445e-05, + "loss": 0.9093, + "step": 32692 + }, + { + "epoch": 2.1, + "learning_rate": 1.3006075699349494e-05, + "loss": 0.8429, + "step": 32693 + }, + { + "epoch": 2.1, + "learning_rate": 1.3005861010326543e-05, + "loss": 0.9139, + "step": 32694 + }, + { + "epoch": 2.1, + "learning_rate": 1.3005646321303594e-05, + "loss": 0.9075, + "step": 32695 + }, + { + "epoch": 2.1, + "learning_rate": 1.3005431632280643e-05, + "loss": 0.9527, + "step": 32696 + }, + { + "epoch": 2.1, + "learning_rate": 1.3005216943257692e-05, + "loss": 0.8999, + "step": 32697 + }, + { + "epoch": 2.1, + "learning_rate": 1.3005002254234742e-05, + "loss": 0.8416, + "step": 32698 + }, + { + "epoch": 2.1, + "learning_rate": 1.3004787565211791e-05, + "loss": 0.8751, + "step": 32699 + }, + { + "epoch": 2.1, + "learning_rate": 1.3004572876188842e-05, + "loss": 0.8443, + "step": 32700 + }, + { + "epoch": 2.1, + "learning_rate": 1.3004358187165891e-05, + "loss": 0.8845, + "step": 32701 + }, + { + "epoch": 2.1, + "learning_rate": 1.300414349814294e-05, + "loss": 0.8841, + "step": 32702 + }, + { + "epoch": 2.1, + "learning_rate": 1.300392880911999e-05, + "loss": 0.8743, + "step": 32703 + }, + { + "epoch": 2.1, + "learning_rate": 1.300371412009704e-05, + "loss": 0.8833, + "step": 32704 + }, + { + "epoch": 2.1, + "learning_rate": 1.3003499431074089e-05, + "loss": 0.8997, + "step": 32705 + }, + { + "epoch": 2.1, + "learning_rate": 1.3003284742051141e-05, + "loss": 0.9269, + "step": 32706 + }, + { + "epoch": 2.1, + "learning_rate": 1.300307005302819e-05, + "loss": 0.981, + "step": 32707 + }, + { + "epoch": 2.1, + "learning_rate": 1.300285536400524e-05, + "loss": 0.798, + "step": 32708 + }, + { + "epoch": 2.1, + "learning_rate": 1.300264067498229e-05, + "loss": 0.9135, + "step": 32709 + }, + { + "epoch": 2.1, + "learning_rate": 1.300242598595934e-05, + "loss": 0.8948, + "step": 32710 + }, + { + "epoch": 2.1, + "learning_rate": 1.3002211296936389e-05, + "loss": 0.874, + "step": 32711 + }, + { + "epoch": 2.1, + "learning_rate": 1.3001996607913438e-05, + "loss": 0.8498, + "step": 32712 + }, + { + "epoch": 2.1, + "learning_rate": 1.3001781918890489e-05, + "loss": 0.8963, + "step": 32713 + }, + { + "epoch": 2.1, + "learning_rate": 1.3001567229867538e-05, + "loss": 0.8976, + "step": 32714 + }, + { + "epoch": 2.1, + "learning_rate": 1.3001352540844587e-05, + "loss": 0.8727, + "step": 32715 + }, + { + "epoch": 2.1, + "learning_rate": 1.3001137851821637e-05, + "loss": 0.8276, + "step": 32716 + }, + { + "epoch": 2.1, + "learning_rate": 1.3000923162798686e-05, + "loss": 0.878, + "step": 32717 + }, + { + "epoch": 2.1, + "learning_rate": 1.3000708473775735e-05, + "loss": 0.9912, + "step": 32718 + }, + { + "epoch": 2.11, + "learning_rate": 1.3000493784752786e-05, + "loss": 0.883, + "step": 32719 + }, + { + "epoch": 2.11, + "learning_rate": 1.3000279095729838e-05, + "loss": 0.9063, + "step": 32720 + }, + { + "epoch": 2.11, + "learning_rate": 1.3000064406706887e-05, + "loss": 0.9369, + "step": 32721 + }, + { + "epoch": 2.11, + "learning_rate": 1.2999849717683936e-05, + "loss": 0.8944, + "step": 32722 + }, + { + "epoch": 2.11, + "learning_rate": 1.2999635028660987e-05, + "loss": 0.9209, + "step": 32723 + }, + { + "epoch": 2.11, + "learning_rate": 1.2999420339638036e-05, + "loss": 0.8358, + "step": 32724 + }, + { + "epoch": 2.11, + "learning_rate": 1.2999205650615085e-05, + "loss": 0.8929, + "step": 32725 + }, + { + "epoch": 2.11, + "learning_rate": 1.2998990961592135e-05, + "loss": 0.9734, + "step": 32726 + }, + { + "epoch": 2.11, + "learning_rate": 1.2998776272569184e-05, + "loss": 0.8586, + "step": 32727 + }, + { + "epoch": 2.11, + "learning_rate": 1.2998561583546233e-05, + "loss": 0.8199, + "step": 32728 + }, + { + "epoch": 2.11, + "learning_rate": 1.2998346894523284e-05, + "loss": 0.9295, + "step": 32729 + }, + { + "epoch": 2.11, + "learning_rate": 1.2998132205500333e-05, + "loss": 0.8694, + "step": 32730 + }, + { + "epoch": 2.11, + "learning_rate": 1.2997917516477384e-05, + "loss": 0.8945, + "step": 32731 + }, + { + "epoch": 2.11, + "learning_rate": 1.2997702827454433e-05, + "loss": 0.838, + "step": 32732 + }, + { + "epoch": 2.11, + "learning_rate": 1.2997488138431482e-05, + "loss": 0.8968, + "step": 32733 + }, + { + "epoch": 2.11, + "learning_rate": 1.2997273449408532e-05, + "loss": 0.9188, + "step": 32734 + }, + { + "epoch": 2.11, + "learning_rate": 1.2997058760385583e-05, + "loss": 0.8702, + "step": 32735 + }, + { + "epoch": 2.11, + "learning_rate": 1.2996844071362634e-05, + "loss": 0.9203, + "step": 32736 + }, + { + "epoch": 2.11, + "learning_rate": 1.2996629382339683e-05, + "loss": 0.9209, + "step": 32737 + }, + { + "epoch": 2.11, + "learning_rate": 1.2996414693316733e-05, + "loss": 0.9175, + "step": 32738 + }, + { + "epoch": 2.11, + "learning_rate": 1.2996200004293782e-05, + "loss": 0.8675, + "step": 32739 + }, + { + "epoch": 2.11, + "learning_rate": 1.2995985315270831e-05, + "loss": 0.8628, + "step": 32740 + }, + { + "epoch": 2.11, + "learning_rate": 1.2995770626247882e-05, + "loss": 0.886, + "step": 32741 + }, + { + "epoch": 2.11, + "learning_rate": 1.299555593722493e-05, + "loss": 0.8293, + "step": 32742 + }, + { + "epoch": 2.11, + "learning_rate": 1.299534124820198e-05, + "loss": 0.8417, + "step": 32743 + }, + { + "epoch": 2.11, + "learning_rate": 1.299512655917903e-05, + "loss": 0.8678, + "step": 32744 + }, + { + "epoch": 2.11, + "learning_rate": 1.299491187015608e-05, + "loss": 0.8102, + "step": 32745 + }, + { + "epoch": 2.11, + "learning_rate": 1.2994697181133128e-05, + "loss": 0.9095, + "step": 32746 + }, + { + "epoch": 2.11, + "learning_rate": 1.2994482492110179e-05, + "loss": 0.8164, + "step": 32747 + }, + { + "epoch": 2.11, + "learning_rate": 1.2994267803087228e-05, + "loss": 0.9242, + "step": 32748 + }, + { + "epoch": 2.11, + "learning_rate": 1.299405311406428e-05, + "loss": 0.8887, + "step": 32749 + }, + { + "epoch": 2.11, + "learning_rate": 1.299383842504133e-05, + "loss": 0.8647, + "step": 32750 + }, + { + "epoch": 2.11, + "learning_rate": 1.299362373601838e-05, + "loss": 0.8698, + "step": 32751 + }, + { + "epoch": 2.11, + "learning_rate": 1.2993409046995429e-05, + "loss": 0.8913, + "step": 32752 + }, + { + "epoch": 2.11, + "learning_rate": 1.2993194357972478e-05, + "loss": 0.8507, + "step": 32753 + }, + { + "epoch": 2.11, + "learning_rate": 1.2992979668949528e-05, + "loss": 0.9154, + "step": 32754 + }, + { + "epoch": 2.11, + "learning_rate": 1.2992764979926577e-05, + "loss": 0.8779, + "step": 32755 + }, + { + "epoch": 2.11, + "learning_rate": 1.2992550290903626e-05, + "loss": 0.9303, + "step": 32756 + }, + { + "epoch": 2.11, + "learning_rate": 1.2992335601880677e-05, + "loss": 0.9035, + "step": 32757 + }, + { + "epoch": 2.11, + "learning_rate": 1.2992120912857726e-05, + "loss": 0.9766, + "step": 32758 + }, + { + "epoch": 2.11, + "learning_rate": 1.2991906223834777e-05, + "loss": 0.9409, + "step": 32759 + }, + { + "epoch": 2.11, + "learning_rate": 1.2991691534811826e-05, + "loss": 0.9017, + "step": 32760 + }, + { + "epoch": 2.11, + "learning_rate": 1.2991476845788875e-05, + "loss": 0.9196, + "step": 32761 + }, + { + "epoch": 2.11, + "learning_rate": 1.2991262156765925e-05, + "loss": 0.8885, + "step": 32762 + }, + { + "epoch": 2.11, + "learning_rate": 1.2991047467742974e-05, + "loss": 0.8879, + "step": 32763 + }, + { + "epoch": 2.11, + "learning_rate": 1.2990832778720027e-05, + "loss": 0.9031, + "step": 32764 + }, + { + "epoch": 2.11, + "learning_rate": 1.2990618089697076e-05, + "loss": 0.882, + "step": 32765 + }, + { + "epoch": 2.11, + "learning_rate": 1.2990403400674125e-05, + "loss": 0.833, + "step": 32766 + }, + { + "epoch": 2.11, + "learning_rate": 1.2990188711651175e-05, + "loss": 0.8453, + "step": 32767 + }, + { + "epoch": 2.11, + "learning_rate": 1.2989974022628224e-05, + "loss": 0.8859, + "step": 32768 + }, + { + "epoch": 2.11, + "learning_rate": 1.2989759333605275e-05, + "loss": 0.9439, + "step": 32769 + }, + { + "epoch": 2.11, + "learning_rate": 1.2989544644582324e-05, + "loss": 0.9001, + "step": 32770 + }, + { + "epoch": 2.11, + "learning_rate": 1.2989329955559373e-05, + "loss": 0.9206, + "step": 32771 + }, + { + "epoch": 2.11, + "learning_rate": 1.2989115266536423e-05, + "loss": 0.8632, + "step": 32772 + }, + { + "epoch": 2.11, + "learning_rate": 1.2988900577513472e-05, + "loss": 0.9288, + "step": 32773 + }, + { + "epoch": 2.11, + "learning_rate": 1.2988685888490521e-05, + "loss": 0.862, + "step": 32774 + }, + { + "epoch": 2.11, + "learning_rate": 1.2988471199467572e-05, + "loss": 0.9358, + "step": 32775 + }, + { + "epoch": 2.11, + "learning_rate": 1.2988256510444621e-05, + "loss": 0.9196, + "step": 32776 + }, + { + "epoch": 2.11, + "learning_rate": 1.2988041821421672e-05, + "loss": 0.9011, + "step": 32777 + }, + { + "epoch": 2.11, + "learning_rate": 1.2987827132398722e-05, + "loss": 0.8781, + "step": 32778 + }, + { + "epoch": 2.11, + "learning_rate": 1.2987612443375773e-05, + "loss": 0.9119, + "step": 32779 + }, + { + "epoch": 2.11, + "learning_rate": 1.2987397754352822e-05, + "loss": 0.8639, + "step": 32780 + }, + { + "epoch": 2.11, + "learning_rate": 1.2987183065329871e-05, + "loss": 0.8488, + "step": 32781 + }, + { + "epoch": 2.11, + "learning_rate": 1.2986968376306922e-05, + "loss": 0.9978, + "step": 32782 + }, + { + "epoch": 2.11, + "learning_rate": 1.298675368728397e-05, + "loss": 0.9701, + "step": 32783 + }, + { + "epoch": 2.11, + "learning_rate": 1.298653899826102e-05, + "loss": 0.941, + "step": 32784 + }, + { + "epoch": 2.11, + "learning_rate": 1.298632430923807e-05, + "loss": 0.8511, + "step": 32785 + }, + { + "epoch": 2.11, + "learning_rate": 1.2986109620215119e-05, + "loss": 0.8343, + "step": 32786 + }, + { + "epoch": 2.11, + "learning_rate": 1.298589493119217e-05, + "loss": 0.853, + "step": 32787 + }, + { + "epoch": 2.11, + "learning_rate": 1.2985680242169219e-05, + "loss": 0.7795, + "step": 32788 + }, + { + "epoch": 2.11, + "learning_rate": 1.2985465553146268e-05, + "loss": 0.9401, + "step": 32789 + }, + { + "epoch": 2.11, + "learning_rate": 1.2985250864123318e-05, + "loss": 0.8567, + "step": 32790 + }, + { + "epoch": 2.11, + "learning_rate": 1.2985036175100367e-05, + "loss": 0.92, + "step": 32791 + }, + { + "epoch": 2.11, + "learning_rate": 1.2984821486077416e-05, + "loss": 0.8703, + "step": 32792 + }, + { + "epoch": 2.11, + "learning_rate": 1.2984606797054469e-05, + "loss": 0.9351, + "step": 32793 + }, + { + "epoch": 2.11, + "learning_rate": 1.2984392108031518e-05, + "loss": 0.9301, + "step": 32794 + }, + { + "epoch": 2.11, + "learning_rate": 1.2984177419008568e-05, + "loss": 0.9882, + "step": 32795 + }, + { + "epoch": 2.11, + "learning_rate": 1.2983962729985617e-05, + "loss": 0.8943, + "step": 32796 + }, + { + "epoch": 2.11, + "learning_rate": 1.2983748040962668e-05, + "loss": 0.8903, + "step": 32797 + }, + { + "epoch": 2.11, + "learning_rate": 1.2983533351939717e-05, + "loss": 0.915, + "step": 32798 + }, + { + "epoch": 2.11, + "learning_rate": 1.2983318662916766e-05, + "loss": 0.8784, + "step": 32799 + }, + { + "epoch": 2.11, + "learning_rate": 1.2983103973893816e-05, + "loss": 0.9015, + "step": 32800 + }, + { + "epoch": 2.11, + "learning_rate": 1.2982889284870865e-05, + "loss": 0.9025, + "step": 32801 + }, + { + "epoch": 2.11, + "learning_rate": 1.2982674595847914e-05, + "loss": 0.8406, + "step": 32802 + }, + { + "epoch": 2.11, + "learning_rate": 1.2982459906824965e-05, + "loss": 0.9166, + "step": 32803 + }, + { + "epoch": 2.11, + "learning_rate": 1.2982245217802014e-05, + "loss": 0.8574, + "step": 32804 + }, + { + "epoch": 2.11, + "learning_rate": 1.2982030528779063e-05, + "loss": 0.8733, + "step": 32805 + }, + { + "epoch": 2.11, + "learning_rate": 1.2981815839756114e-05, + "loss": 0.8989, + "step": 32806 + }, + { + "epoch": 2.11, + "learning_rate": 1.2981601150733166e-05, + "loss": 0.8849, + "step": 32807 + }, + { + "epoch": 2.11, + "learning_rate": 1.2981386461710215e-05, + "loss": 0.9242, + "step": 32808 + }, + { + "epoch": 2.11, + "learning_rate": 1.2981171772687264e-05, + "loss": 0.9057, + "step": 32809 + }, + { + "epoch": 2.11, + "learning_rate": 1.2980957083664315e-05, + "loss": 0.9149, + "step": 32810 + }, + { + "epoch": 2.11, + "learning_rate": 1.2980742394641364e-05, + "loss": 0.9496, + "step": 32811 + }, + { + "epoch": 2.11, + "learning_rate": 1.2980527705618412e-05, + "loss": 0.8794, + "step": 32812 + }, + { + "epoch": 2.11, + "learning_rate": 1.2980313016595463e-05, + "loss": 0.8968, + "step": 32813 + }, + { + "epoch": 2.11, + "learning_rate": 1.2980098327572512e-05, + "loss": 0.8453, + "step": 32814 + }, + { + "epoch": 2.11, + "learning_rate": 1.2979883638549561e-05, + "loss": 0.8965, + "step": 32815 + }, + { + "epoch": 2.11, + "learning_rate": 1.2979668949526612e-05, + "loss": 0.8952, + "step": 32816 + }, + { + "epoch": 2.11, + "learning_rate": 1.297945426050366e-05, + "loss": 0.8462, + "step": 32817 + }, + { + "epoch": 2.11, + "learning_rate": 1.2979239571480711e-05, + "loss": 0.8489, + "step": 32818 + }, + { + "epoch": 2.11, + "learning_rate": 1.297902488245776e-05, + "loss": 0.9268, + "step": 32819 + }, + { + "epoch": 2.11, + "learning_rate": 1.297881019343481e-05, + "loss": 0.8732, + "step": 32820 + }, + { + "epoch": 2.11, + "learning_rate": 1.297859550441186e-05, + "loss": 0.8894, + "step": 32821 + }, + { + "epoch": 2.11, + "learning_rate": 1.297838081538891e-05, + "loss": 0.9171, + "step": 32822 + }, + { + "epoch": 2.11, + "learning_rate": 1.2978166126365961e-05, + "loss": 0.8468, + "step": 32823 + }, + { + "epoch": 2.11, + "learning_rate": 1.297795143734301e-05, + "loss": 0.9236, + "step": 32824 + }, + { + "epoch": 2.11, + "learning_rate": 1.297773674832006e-05, + "loss": 0.8672, + "step": 32825 + }, + { + "epoch": 2.11, + "learning_rate": 1.297752205929711e-05, + "loss": 0.9267, + "step": 32826 + }, + { + "epoch": 2.11, + "learning_rate": 1.2977307370274159e-05, + "loss": 0.9197, + "step": 32827 + }, + { + "epoch": 2.11, + "learning_rate": 1.297709268125121e-05, + "loss": 0.9618, + "step": 32828 + }, + { + "epoch": 2.11, + "learning_rate": 1.2976877992228258e-05, + "loss": 0.8135, + "step": 32829 + }, + { + "epoch": 2.11, + "learning_rate": 1.2976663303205307e-05, + "loss": 0.8941, + "step": 32830 + }, + { + "epoch": 2.11, + "learning_rate": 1.2976448614182358e-05, + "loss": 0.822, + "step": 32831 + }, + { + "epoch": 2.11, + "learning_rate": 1.2976233925159407e-05, + "loss": 0.9501, + "step": 32832 + }, + { + "epoch": 2.11, + "learning_rate": 1.2976019236136456e-05, + "loss": 0.8888, + "step": 32833 + }, + { + "epoch": 2.11, + "learning_rate": 1.2975804547113507e-05, + "loss": 0.8426, + "step": 32834 + }, + { + "epoch": 2.11, + "learning_rate": 1.2975589858090556e-05, + "loss": 0.8925, + "step": 32835 + }, + { + "epoch": 2.11, + "learning_rate": 1.2975375169067606e-05, + "loss": 0.9083, + "step": 32836 + }, + { + "epoch": 2.11, + "learning_rate": 1.2975160480044657e-05, + "loss": 0.8924, + "step": 32837 + }, + { + "epoch": 2.11, + "learning_rate": 1.2974945791021708e-05, + "loss": 0.9072, + "step": 32838 + }, + { + "epoch": 2.11, + "learning_rate": 1.2974731101998757e-05, + "loss": 0.9142, + "step": 32839 + }, + { + "epoch": 2.11, + "learning_rate": 1.2974516412975806e-05, + "loss": 0.9093, + "step": 32840 + }, + { + "epoch": 2.11, + "learning_rate": 1.2974301723952856e-05, + "loss": 0.8849, + "step": 32841 + }, + { + "epoch": 2.11, + "learning_rate": 1.2974087034929905e-05, + "loss": 0.8495, + "step": 32842 + }, + { + "epoch": 2.11, + "learning_rate": 1.2973872345906954e-05, + "loss": 0.9647, + "step": 32843 + }, + { + "epoch": 2.11, + "learning_rate": 1.2973657656884005e-05, + "loss": 0.8879, + "step": 32844 + }, + { + "epoch": 2.11, + "learning_rate": 1.2973442967861054e-05, + "loss": 0.8845, + "step": 32845 + }, + { + "epoch": 2.11, + "learning_rate": 1.2973228278838104e-05, + "loss": 0.8689, + "step": 32846 + }, + { + "epoch": 2.11, + "learning_rate": 1.2973013589815153e-05, + "loss": 0.8439, + "step": 32847 + }, + { + "epoch": 2.11, + "learning_rate": 1.2972798900792202e-05, + "loss": 0.929, + "step": 32848 + }, + { + "epoch": 2.11, + "learning_rate": 1.2972584211769253e-05, + "loss": 0.9747, + "step": 32849 + }, + { + "epoch": 2.11, + "learning_rate": 1.2972369522746302e-05, + "loss": 0.8729, + "step": 32850 + }, + { + "epoch": 2.11, + "learning_rate": 1.2972154833723354e-05, + "loss": 0.8539, + "step": 32851 + }, + { + "epoch": 2.11, + "learning_rate": 1.2971940144700403e-05, + "loss": 0.9382, + "step": 32852 + }, + { + "epoch": 2.11, + "learning_rate": 1.2971725455677452e-05, + "loss": 0.8771, + "step": 32853 + }, + { + "epoch": 2.11, + "learning_rate": 1.2971510766654503e-05, + "loss": 0.8414, + "step": 32854 + }, + { + "epoch": 2.11, + "learning_rate": 1.2971296077631552e-05, + "loss": 0.8593, + "step": 32855 + }, + { + "epoch": 2.11, + "learning_rate": 1.2971081388608603e-05, + "loss": 0.8419, + "step": 32856 + }, + { + "epoch": 2.11, + "learning_rate": 1.2970866699585651e-05, + "loss": 0.8652, + "step": 32857 + }, + { + "epoch": 2.11, + "learning_rate": 1.29706520105627e-05, + "loss": 0.8499, + "step": 32858 + }, + { + "epoch": 2.11, + "learning_rate": 1.2970437321539751e-05, + "loss": 0.9001, + "step": 32859 + }, + { + "epoch": 2.11, + "learning_rate": 1.29702226325168e-05, + "loss": 0.9475, + "step": 32860 + }, + { + "epoch": 2.11, + "learning_rate": 1.2970007943493849e-05, + "loss": 0.9007, + "step": 32861 + }, + { + "epoch": 2.11, + "learning_rate": 1.29697932544709e-05, + "loss": 0.8521, + "step": 32862 + }, + { + "epoch": 2.11, + "learning_rate": 1.2969578565447949e-05, + "loss": 0.9135, + "step": 32863 + }, + { + "epoch": 2.11, + "learning_rate": 1.2969363876425e-05, + "loss": 0.9082, + "step": 32864 + }, + { + "epoch": 2.11, + "learning_rate": 1.2969149187402048e-05, + "loss": 0.929, + "step": 32865 + }, + { + "epoch": 2.11, + "learning_rate": 1.29689344983791e-05, + "loss": 0.9079, + "step": 32866 + }, + { + "epoch": 2.11, + "learning_rate": 1.296871980935615e-05, + "loss": 0.9161, + "step": 32867 + }, + { + "epoch": 2.11, + "learning_rate": 1.2968505120333199e-05, + "loss": 0.8385, + "step": 32868 + }, + { + "epoch": 2.11, + "learning_rate": 1.296829043131025e-05, + "loss": 0.8788, + "step": 32869 + }, + { + "epoch": 2.11, + "learning_rate": 1.2968075742287298e-05, + "loss": 0.936, + "step": 32870 + }, + { + "epoch": 2.11, + "learning_rate": 1.2967861053264347e-05, + "loss": 0.9178, + "step": 32871 + }, + { + "epoch": 2.11, + "learning_rate": 1.2967646364241398e-05, + "loss": 0.8926, + "step": 32872 + }, + { + "epoch": 2.11, + "learning_rate": 1.2967431675218447e-05, + "loss": 0.903, + "step": 32873 + }, + { + "epoch": 2.12, + "learning_rate": 1.2967216986195497e-05, + "loss": 0.8213, + "step": 32874 + }, + { + "epoch": 2.12, + "learning_rate": 1.2967002297172546e-05, + "loss": 0.8721, + "step": 32875 + }, + { + "epoch": 2.12, + "learning_rate": 1.2966787608149595e-05, + "loss": 0.8103, + "step": 32876 + }, + { + "epoch": 2.12, + "learning_rate": 1.2966572919126646e-05, + "loss": 0.8714, + "step": 32877 + }, + { + "epoch": 2.12, + "learning_rate": 1.2966358230103695e-05, + "loss": 0.8947, + "step": 32878 + }, + { + "epoch": 2.12, + "learning_rate": 1.2966143541080744e-05, + "loss": 0.912, + "step": 32879 + }, + { + "epoch": 2.12, + "learning_rate": 1.2965928852057796e-05, + "loss": 0.9018, + "step": 32880 + }, + { + "epoch": 2.12, + "learning_rate": 1.2965714163034845e-05, + "loss": 0.8227, + "step": 32881 + }, + { + "epoch": 2.12, + "learning_rate": 1.2965499474011896e-05, + "loss": 0.8429, + "step": 32882 + }, + { + "epoch": 2.12, + "learning_rate": 1.2965284784988945e-05, + "loss": 0.8891, + "step": 32883 + }, + { + "epoch": 2.12, + "learning_rate": 1.2965070095965996e-05, + "loss": 0.9636, + "step": 32884 + }, + { + "epoch": 2.12, + "learning_rate": 1.2964855406943045e-05, + "loss": 0.9067, + "step": 32885 + }, + { + "epoch": 2.12, + "learning_rate": 1.2964640717920093e-05, + "loss": 0.8979, + "step": 32886 + }, + { + "epoch": 2.12, + "learning_rate": 1.2964426028897144e-05, + "loss": 0.9166, + "step": 32887 + }, + { + "epoch": 2.12, + "learning_rate": 1.2964211339874193e-05, + "loss": 0.9195, + "step": 32888 + }, + { + "epoch": 2.12, + "learning_rate": 1.2963996650851242e-05, + "loss": 0.9062, + "step": 32889 + }, + { + "epoch": 2.12, + "learning_rate": 1.2963781961828293e-05, + "loss": 0.8395, + "step": 32890 + }, + { + "epoch": 2.12, + "learning_rate": 1.2963567272805342e-05, + "loss": 0.9055, + "step": 32891 + }, + { + "epoch": 2.12, + "learning_rate": 1.296335258378239e-05, + "loss": 0.9216, + "step": 32892 + }, + { + "epoch": 2.12, + "learning_rate": 1.2963137894759441e-05, + "loss": 0.8962, + "step": 32893 + }, + { + "epoch": 2.12, + "learning_rate": 1.296292320573649e-05, + "loss": 0.9853, + "step": 32894 + }, + { + "epoch": 2.12, + "learning_rate": 1.2962708516713543e-05, + "loss": 0.9107, + "step": 32895 + }, + { + "epoch": 2.12, + "learning_rate": 1.2962493827690592e-05, + "loss": 0.924, + "step": 32896 + }, + { + "epoch": 2.12, + "learning_rate": 1.2962279138667642e-05, + "loss": 0.9904, + "step": 32897 + }, + { + "epoch": 2.12, + "learning_rate": 1.2962064449644691e-05, + "loss": 0.9205, + "step": 32898 + }, + { + "epoch": 2.12, + "learning_rate": 1.296184976062174e-05, + "loss": 0.8597, + "step": 32899 + }, + { + "epoch": 2.12, + "learning_rate": 1.2961635071598791e-05, + "loss": 0.9018, + "step": 32900 + }, + { + "epoch": 2.12, + "learning_rate": 1.296142038257584e-05, + "loss": 0.9495, + "step": 32901 + }, + { + "epoch": 2.12, + "learning_rate": 1.2961205693552889e-05, + "loss": 0.8256, + "step": 32902 + }, + { + "epoch": 2.12, + "learning_rate": 1.296099100452994e-05, + "loss": 0.8617, + "step": 32903 + }, + { + "epoch": 2.12, + "learning_rate": 1.2960776315506988e-05, + "loss": 0.926, + "step": 32904 + }, + { + "epoch": 2.12, + "learning_rate": 1.2960561626484039e-05, + "loss": 0.9288, + "step": 32905 + }, + { + "epoch": 2.12, + "learning_rate": 1.2960346937461088e-05, + "loss": 0.9746, + "step": 32906 + }, + { + "epoch": 2.12, + "learning_rate": 1.2960132248438137e-05, + "loss": 0.978, + "step": 32907 + }, + { + "epoch": 2.12, + "learning_rate": 1.2959917559415188e-05, + "loss": 0.8945, + "step": 32908 + }, + { + "epoch": 2.12, + "learning_rate": 1.2959702870392238e-05, + "loss": 0.8438, + "step": 32909 + }, + { + "epoch": 2.12, + "learning_rate": 1.2959488181369289e-05, + "loss": 0.9091, + "step": 32910 + }, + { + "epoch": 2.12, + "learning_rate": 1.2959273492346338e-05, + "loss": 0.9455, + "step": 32911 + }, + { + "epoch": 2.12, + "learning_rate": 1.2959058803323387e-05, + "loss": 0.8888, + "step": 32912 + }, + { + "epoch": 2.12, + "learning_rate": 1.2958844114300438e-05, + "loss": 0.9079, + "step": 32913 + }, + { + "epoch": 2.12, + "learning_rate": 1.2958629425277487e-05, + "loss": 0.8372, + "step": 32914 + }, + { + "epoch": 2.12, + "learning_rate": 1.2958414736254537e-05, + "loss": 0.8226, + "step": 32915 + }, + { + "epoch": 2.12, + "learning_rate": 1.2958200047231586e-05, + "loss": 0.9118, + "step": 32916 + }, + { + "epoch": 2.12, + "learning_rate": 1.2957985358208635e-05, + "loss": 0.8883, + "step": 32917 + }, + { + "epoch": 2.12, + "learning_rate": 1.2957770669185686e-05, + "loss": 0.9207, + "step": 32918 + }, + { + "epoch": 2.12, + "learning_rate": 1.2957555980162735e-05, + "loss": 0.9937, + "step": 32919 + }, + { + "epoch": 2.12, + "learning_rate": 1.2957341291139784e-05, + "loss": 0.8884, + "step": 32920 + }, + { + "epoch": 2.12, + "learning_rate": 1.2957126602116834e-05, + "loss": 0.8647, + "step": 32921 + }, + { + "epoch": 2.12, + "learning_rate": 1.2956911913093883e-05, + "loss": 0.8559, + "step": 32922 + }, + { + "epoch": 2.12, + "learning_rate": 1.2956697224070934e-05, + "loss": 0.8772, + "step": 32923 + }, + { + "epoch": 2.12, + "learning_rate": 1.2956482535047985e-05, + "loss": 0.7992, + "step": 32924 + }, + { + "epoch": 2.12, + "learning_rate": 1.2956267846025035e-05, + "loss": 0.8891, + "step": 32925 + }, + { + "epoch": 2.12, + "learning_rate": 1.2956053157002084e-05, + "loss": 0.8858, + "step": 32926 + }, + { + "epoch": 2.12, + "learning_rate": 1.2955838467979133e-05, + "loss": 0.8533, + "step": 32927 + }, + { + "epoch": 2.12, + "learning_rate": 1.2955623778956184e-05, + "loss": 0.8486, + "step": 32928 + }, + { + "epoch": 2.12, + "learning_rate": 1.2955409089933233e-05, + "loss": 0.9207, + "step": 32929 + }, + { + "epoch": 2.12, + "learning_rate": 1.2955194400910282e-05, + "loss": 0.9236, + "step": 32930 + }, + { + "epoch": 2.12, + "learning_rate": 1.2954979711887332e-05, + "loss": 0.9115, + "step": 32931 + }, + { + "epoch": 2.12, + "learning_rate": 1.2954765022864381e-05, + "loss": 0.8459, + "step": 32932 + }, + { + "epoch": 2.12, + "learning_rate": 1.2954550333841432e-05, + "loss": 0.883, + "step": 32933 + }, + { + "epoch": 2.12, + "learning_rate": 1.2954335644818481e-05, + "loss": 0.8768, + "step": 32934 + }, + { + "epoch": 2.12, + "learning_rate": 1.295412095579553e-05, + "loss": 0.9205, + "step": 32935 + }, + { + "epoch": 2.12, + "learning_rate": 1.295390626677258e-05, + "loss": 0.8864, + "step": 32936 + }, + { + "epoch": 2.12, + "learning_rate": 1.295369157774963e-05, + "loss": 0.8769, + "step": 32937 + }, + { + "epoch": 2.12, + "learning_rate": 1.2953476888726682e-05, + "loss": 0.8545, + "step": 32938 + }, + { + "epoch": 2.12, + "learning_rate": 1.2953262199703731e-05, + "loss": 0.8683, + "step": 32939 + }, + { + "epoch": 2.12, + "learning_rate": 1.295304751068078e-05, + "loss": 0.9267, + "step": 32940 + }, + { + "epoch": 2.12, + "learning_rate": 1.295283282165783e-05, + "loss": 0.8938, + "step": 32941 + }, + { + "epoch": 2.12, + "learning_rate": 1.295261813263488e-05, + "loss": 0.9283, + "step": 32942 + }, + { + "epoch": 2.12, + "learning_rate": 1.295240344361193e-05, + "loss": 0.8985, + "step": 32943 + }, + { + "epoch": 2.12, + "learning_rate": 1.295218875458898e-05, + "loss": 0.8877, + "step": 32944 + }, + { + "epoch": 2.12, + "learning_rate": 1.2951974065566028e-05, + "loss": 0.9161, + "step": 32945 + }, + { + "epoch": 2.12, + "learning_rate": 1.2951759376543079e-05, + "loss": 0.8717, + "step": 32946 + }, + { + "epoch": 2.12, + "learning_rate": 1.2951544687520128e-05, + "loss": 0.8608, + "step": 32947 + }, + { + "epoch": 2.12, + "learning_rate": 1.2951329998497177e-05, + "loss": 0.8549, + "step": 32948 + }, + { + "epoch": 2.12, + "learning_rate": 1.2951115309474227e-05, + "loss": 0.9003, + "step": 32949 + }, + { + "epoch": 2.12, + "learning_rate": 1.2950900620451276e-05, + "loss": 0.9319, + "step": 32950 + }, + { + "epoch": 2.12, + "learning_rate": 1.2950685931428327e-05, + "loss": 0.8406, + "step": 32951 + }, + { + "epoch": 2.12, + "learning_rate": 1.2950471242405376e-05, + "loss": 0.8912, + "step": 32952 + }, + { + "epoch": 2.12, + "learning_rate": 1.2950256553382428e-05, + "loss": 0.8592, + "step": 32953 + }, + { + "epoch": 2.12, + "learning_rate": 1.2950041864359477e-05, + "loss": 0.8329, + "step": 32954 + }, + { + "epoch": 2.12, + "learning_rate": 1.2949827175336526e-05, + "loss": 0.8937, + "step": 32955 + }, + { + "epoch": 2.12, + "learning_rate": 1.2949612486313577e-05, + "loss": 0.8426, + "step": 32956 + }, + { + "epoch": 2.12, + "learning_rate": 1.2949397797290626e-05, + "loss": 0.9458, + "step": 32957 + }, + { + "epoch": 2.12, + "learning_rate": 1.2949183108267675e-05, + "loss": 0.8097, + "step": 32958 + }, + { + "epoch": 2.12, + "learning_rate": 1.2948968419244726e-05, + "loss": 0.9634, + "step": 32959 + }, + { + "epoch": 2.12, + "learning_rate": 1.2948753730221774e-05, + "loss": 0.821, + "step": 32960 + }, + { + "epoch": 2.12, + "learning_rate": 1.2948539041198825e-05, + "loss": 0.9586, + "step": 32961 + }, + { + "epoch": 2.12, + "learning_rate": 1.2948324352175874e-05, + "loss": 0.9581, + "step": 32962 + }, + { + "epoch": 2.12, + "learning_rate": 1.2948109663152923e-05, + "loss": 0.8914, + "step": 32963 + }, + { + "epoch": 2.12, + "learning_rate": 1.2947894974129974e-05, + "loss": 0.8554, + "step": 32964 + }, + { + "epoch": 2.12, + "learning_rate": 1.2947680285107023e-05, + "loss": 0.8837, + "step": 32965 + }, + { + "epoch": 2.12, + "learning_rate": 1.2947465596084072e-05, + "loss": 0.8484, + "step": 32966 + }, + { + "epoch": 2.12, + "learning_rate": 1.2947250907061124e-05, + "loss": 0.9493, + "step": 32967 + }, + { + "epoch": 2.12, + "learning_rate": 1.2947036218038173e-05, + "loss": 0.8357, + "step": 32968 + }, + { + "epoch": 2.12, + "learning_rate": 1.2946821529015224e-05, + "loss": 0.8903, + "step": 32969 + }, + { + "epoch": 2.12, + "learning_rate": 1.2946606839992273e-05, + "loss": 0.864, + "step": 32970 + }, + { + "epoch": 2.12, + "learning_rate": 1.2946392150969323e-05, + "loss": 0.8903, + "step": 32971 + }, + { + "epoch": 2.12, + "learning_rate": 1.2946177461946372e-05, + "loss": 0.939, + "step": 32972 + }, + { + "epoch": 2.12, + "learning_rate": 1.2945962772923421e-05, + "loss": 0.973, + "step": 32973 + }, + { + "epoch": 2.12, + "learning_rate": 1.2945748083900472e-05, + "loss": 0.8618, + "step": 32974 + }, + { + "epoch": 2.12, + "learning_rate": 1.294553339487752e-05, + "loss": 0.911, + "step": 32975 + }, + { + "epoch": 2.12, + "learning_rate": 1.294531870585457e-05, + "loss": 0.9385, + "step": 32976 + }, + { + "epoch": 2.12, + "learning_rate": 1.294510401683162e-05, + "loss": 0.879, + "step": 32977 + }, + { + "epoch": 2.12, + "learning_rate": 1.294488932780867e-05, + "loss": 0.9181, + "step": 32978 + }, + { + "epoch": 2.12, + "learning_rate": 1.2944674638785718e-05, + "loss": 0.8783, + "step": 32979 + }, + { + "epoch": 2.12, + "learning_rate": 1.2944459949762769e-05, + "loss": 0.9027, + "step": 32980 + }, + { + "epoch": 2.12, + "learning_rate": 1.2944245260739818e-05, + "loss": 0.9087, + "step": 32981 + }, + { + "epoch": 2.12, + "learning_rate": 1.294403057171687e-05, + "loss": 0.8758, + "step": 32982 + }, + { + "epoch": 2.12, + "learning_rate": 1.294381588269392e-05, + "loss": 0.8321, + "step": 32983 + }, + { + "epoch": 2.12, + "learning_rate": 1.294360119367097e-05, + "loss": 0.87, + "step": 32984 + }, + { + "epoch": 2.12, + "learning_rate": 1.2943386504648019e-05, + "loss": 0.9305, + "step": 32985 + }, + { + "epoch": 2.12, + "learning_rate": 1.2943171815625068e-05, + "loss": 0.8893, + "step": 32986 + }, + { + "epoch": 2.12, + "learning_rate": 1.2942957126602119e-05, + "loss": 0.8621, + "step": 32987 + }, + { + "epoch": 2.12, + "learning_rate": 1.2942742437579168e-05, + "loss": 0.8368, + "step": 32988 + }, + { + "epoch": 2.12, + "learning_rate": 1.2942527748556216e-05, + "loss": 0.8661, + "step": 32989 + }, + { + "epoch": 2.12, + "learning_rate": 1.2942313059533267e-05, + "loss": 0.9134, + "step": 32990 + }, + { + "epoch": 2.12, + "learning_rate": 1.2942098370510316e-05, + "loss": 0.9539, + "step": 32991 + }, + { + "epoch": 2.12, + "learning_rate": 1.2941883681487367e-05, + "loss": 0.8686, + "step": 32992 + }, + { + "epoch": 2.12, + "learning_rate": 1.2941668992464416e-05, + "loss": 0.9228, + "step": 32993 + }, + { + "epoch": 2.12, + "learning_rate": 1.2941454303441465e-05, + "loss": 0.8618, + "step": 32994 + }, + { + "epoch": 2.12, + "learning_rate": 1.2941239614418515e-05, + "loss": 0.9215, + "step": 32995 + }, + { + "epoch": 2.12, + "learning_rate": 1.2941024925395566e-05, + "loss": 0.8681, + "step": 32996 + }, + { + "epoch": 2.12, + "learning_rate": 1.2940810236372617e-05, + "loss": 0.8396, + "step": 32997 + }, + { + "epoch": 2.12, + "learning_rate": 1.2940595547349666e-05, + "loss": 0.92, + "step": 32998 + }, + { + "epoch": 2.12, + "learning_rate": 1.2940380858326715e-05, + "loss": 0.9154, + "step": 32999 + }, + { + "epoch": 2.12, + "learning_rate": 1.2940166169303765e-05, + "loss": 0.9422, + "step": 33000 + }, + { + "epoch": 2.12, + "eval_laion/cider": 0.5739246839831045, + "eval_laion/meteor": 0.39572785006641076, + "eval_laion/pred_num_words": 36.693359375, + "eval_laion/sacrebleu": 16.094405243706287, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.044521450996399, + "eval_runtime": 97.6532, + "eval_samples_per_second": 5.11, + "eval_steps_per_second": 0.328, + "step": 33000 + }, + { + "epoch": 2.12, + "learning_rate": 1.2939951480280814e-05, + "loss": 0.8551, + "step": 33001 + }, + { + "epoch": 2.12, + "learning_rate": 1.2939736791257865e-05, + "loss": 0.9107, + "step": 33002 + }, + { + "epoch": 2.12, + "learning_rate": 1.2939522102234914e-05, + "loss": 0.8368, + "step": 33003 + }, + { + "epoch": 2.12, + "learning_rate": 1.2939307413211963e-05, + "loss": 0.946, + "step": 33004 + }, + { + "epoch": 2.12, + "learning_rate": 1.2939092724189013e-05, + "loss": 0.8627, + "step": 33005 + }, + { + "epoch": 2.12, + "learning_rate": 1.2938878035166062e-05, + "loss": 0.9316, + "step": 33006 + }, + { + "epoch": 2.12, + "learning_rate": 1.2938663346143111e-05, + "loss": 1.002, + "step": 33007 + }, + { + "epoch": 2.12, + "learning_rate": 1.2938448657120162e-05, + "loss": 0.9096, + "step": 33008 + }, + { + "epoch": 2.12, + "learning_rate": 1.2938233968097211e-05, + "loss": 0.8757, + "step": 33009 + }, + { + "epoch": 2.12, + "learning_rate": 1.2938019279074262e-05, + "loss": 0.9197, + "step": 33010 + }, + { + "epoch": 2.12, + "learning_rate": 1.2937804590051312e-05, + "loss": 0.9304, + "step": 33011 + }, + { + "epoch": 2.12, + "learning_rate": 1.2937589901028363e-05, + "loss": 0.8976, + "step": 33012 + }, + { + "epoch": 2.12, + "learning_rate": 1.2937375212005412e-05, + "loss": 0.8627, + "step": 33013 + }, + { + "epoch": 2.12, + "learning_rate": 1.2937160522982461e-05, + "loss": 0.9026, + "step": 33014 + }, + { + "epoch": 2.12, + "learning_rate": 1.2936945833959512e-05, + "loss": 0.8296, + "step": 33015 + }, + { + "epoch": 2.12, + "learning_rate": 1.293673114493656e-05, + "loss": 0.8757, + "step": 33016 + }, + { + "epoch": 2.12, + "learning_rate": 1.293651645591361e-05, + "loss": 0.9356, + "step": 33017 + }, + { + "epoch": 2.12, + "learning_rate": 1.293630176689066e-05, + "loss": 0.9815, + "step": 33018 + }, + { + "epoch": 2.12, + "learning_rate": 1.293608707786771e-05, + "loss": 0.8057, + "step": 33019 + }, + { + "epoch": 2.12, + "learning_rate": 1.293587238884476e-05, + "loss": 0.94, + "step": 33020 + }, + { + "epoch": 2.12, + "learning_rate": 1.2935657699821809e-05, + "loss": 0.8834, + "step": 33021 + }, + { + "epoch": 2.12, + "learning_rate": 1.2935443010798858e-05, + "loss": 0.9924, + "step": 33022 + }, + { + "epoch": 2.12, + "learning_rate": 1.2935228321775908e-05, + "loss": 0.9005, + "step": 33023 + }, + { + "epoch": 2.12, + "learning_rate": 1.2935013632752957e-05, + "loss": 0.8912, + "step": 33024 + }, + { + "epoch": 2.12, + "learning_rate": 1.293479894373001e-05, + "loss": 1.008, + "step": 33025 + }, + { + "epoch": 2.12, + "learning_rate": 1.2934584254707059e-05, + "loss": 0.8837, + "step": 33026 + }, + { + "epoch": 2.12, + "learning_rate": 1.2934369565684108e-05, + "loss": 0.8987, + "step": 33027 + }, + { + "epoch": 2.12, + "learning_rate": 1.2934154876661158e-05, + "loss": 0.8366, + "step": 33028 + }, + { + "epoch": 2.13, + "learning_rate": 1.2933940187638207e-05, + "loss": 0.9006, + "step": 33029 + }, + { + "epoch": 2.13, + "learning_rate": 1.2933725498615258e-05, + "loss": 0.8946, + "step": 33030 + }, + { + "epoch": 2.13, + "learning_rate": 1.2933510809592307e-05, + "loss": 0.898, + "step": 33031 + }, + { + "epoch": 2.13, + "learning_rate": 1.2933296120569356e-05, + "loss": 0.8856, + "step": 33032 + }, + { + "epoch": 2.13, + "learning_rate": 1.2933081431546407e-05, + "loss": 0.9181, + "step": 33033 + }, + { + "epoch": 2.13, + "learning_rate": 1.2932866742523455e-05, + "loss": 0.8632, + "step": 33034 + }, + { + "epoch": 2.13, + "learning_rate": 1.2932652053500504e-05, + "loss": 0.9862, + "step": 33035 + }, + { + "epoch": 2.13, + "learning_rate": 1.2932437364477555e-05, + "loss": 0.9976, + "step": 33036 + }, + { + "epoch": 2.13, + "learning_rate": 1.2932222675454604e-05, + "loss": 0.9142, + "step": 33037 + }, + { + "epoch": 2.13, + "learning_rate": 1.2932007986431653e-05, + "loss": 0.918, + "step": 33038 + }, + { + "epoch": 2.13, + "learning_rate": 1.2931793297408704e-05, + "loss": 0.9697, + "step": 33039 + }, + { + "epoch": 2.13, + "learning_rate": 1.2931578608385756e-05, + "loss": 0.8994, + "step": 33040 + }, + { + "epoch": 2.13, + "learning_rate": 1.2931363919362805e-05, + "loss": 0.8556, + "step": 33041 + }, + { + "epoch": 2.13, + "learning_rate": 1.2931149230339854e-05, + "loss": 0.9144, + "step": 33042 + }, + { + "epoch": 2.13, + "learning_rate": 1.2930934541316905e-05, + "loss": 0.8813, + "step": 33043 + }, + { + "epoch": 2.13, + "learning_rate": 1.2930719852293954e-05, + "loss": 0.8986, + "step": 33044 + }, + { + "epoch": 2.13, + "learning_rate": 1.2930505163271003e-05, + "loss": 0.9205, + "step": 33045 + }, + { + "epoch": 2.13, + "learning_rate": 1.2930290474248053e-05, + "loss": 0.928, + "step": 33046 + }, + { + "epoch": 2.13, + "learning_rate": 1.2930075785225102e-05, + "loss": 0.9435, + "step": 33047 + }, + { + "epoch": 2.13, + "learning_rate": 1.2929861096202153e-05, + "loss": 0.8995, + "step": 33048 + }, + { + "epoch": 2.13, + "learning_rate": 1.2929646407179202e-05, + "loss": 0.883, + "step": 33049 + }, + { + "epoch": 2.13, + "learning_rate": 1.292943171815625e-05, + "loss": 0.891, + "step": 33050 + }, + { + "epoch": 2.13, + "learning_rate": 1.2929217029133301e-05, + "loss": 0.8689, + "step": 33051 + }, + { + "epoch": 2.13, + "learning_rate": 1.292900234011035e-05, + "loss": 0.8703, + "step": 33052 + }, + { + "epoch": 2.13, + "learning_rate": 1.29287876510874e-05, + "loss": 0.9368, + "step": 33053 + }, + { + "epoch": 2.13, + "learning_rate": 1.2928572962064452e-05, + "loss": 0.8732, + "step": 33054 + }, + { + "epoch": 2.13, + "learning_rate": 1.29283582730415e-05, + "loss": 0.8225, + "step": 33055 + }, + { + "epoch": 2.13, + "learning_rate": 1.2928143584018551e-05, + "loss": 0.9171, + "step": 33056 + }, + { + "epoch": 2.13, + "learning_rate": 1.29279288949956e-05, + "loss": 0.9024, + "step": 33057 + }, + { + "epoch": 2.13, + "learning_rate": 1.2927714205972651e-05, + "loss": 0.8682, + "step": 33058 + }, + { + "epoch": 2.13, + "learning_rate": 1.29274995169497e-05, + "loss": 0.9047, + "step": 33059 + }, + { + "epoch": 2.13, + "learning_rate": 1.2927284827926749e-05, + "loss": 0.9154, + "step": 33060 + }, + { + "epoch": 2.13, + "learning_rate": 1.29270701389038e-05, + "loss": 0.8509, + "step": 33061 + }, + { + "epoch": 2.13, + "learning_rate": 1.2926855449880849e-05, + "loss": 0.9316, + "step": 33062 + }, + { + "epoch": 2.13, + "learning_rate": 1.2926640760857898e-05, + "loss": 0.8741, + "step": 33063 + }, + { + "epoch": 2.13, + "learning_rate": 1.2926426071834948e-05, + "loss": 0.8977, + "step": 33064 + }, + { + "epoch": 2.13, + "learning_rate": 1.2926211382811997e-05, + "loss": 0.9114, + "step": 33065 + }, + { + "epoch": 2.13, + "learning_rate": 1.2925996693789046e-05, + "loss": 0.9133, + "step": 33066 + }, + { + "epoch": 2.13, + "learning_rate": 1.2925782004766097e-05, + "loss": 0.9509, + "step": 33067 + }, + { + "epoch": 2.13, + "learning_rate": 1.2925567315743146e-05, + "loss": 0.9387, + "step": 33068 + }, + { + "epoch": 2.13, + "learning_rate": 1.2925352626720198e-05, + "loss": 0.9184, + "step": 33069 + }, + { + "epoch": 2.13, + "learning_rate": 1.2925137937697247e-05, + "loss": 0.9843, + "step": 33070 + }, + { + "epoch": 2.13, + "learning_rate": 1.2924923248674298e-05, + "loss": 0.9229, + "step": 33071 + }, + { + "epoch": 2.13, + "learning_rate": 1.2924708559651347e-05, + "loss": 0.8778, + "step": 33072 + }, + { + "epoch": 2.13, + "learning_rate": 1.2924493870628396e-05, + "loss": 0.8926, + "step": 33073 + }, + { + "epoch": 2.13, + "learning_rate": 1.2924279181605446e-05, + "loss": 0.8912, + "step": 33074 + }, + { + "epoch": 2.13, + "learning_rate": 1.2924064492582495e-05, + "loss": 0.8636, + "step": 33075 + }, + { + "epoch": 2.13, + "learning_rate": 1.2923849803559544e-05, + "loss": 0.9611, + "step": 33076 + }, + { + "epoch": 2.13, + "learning_rate": 1.2923635114536595e-05, + "loss": 0.8538, + "step": 33077 + }, + { + "epoch": 2.13, + "learning_rate": 1.2923420425513644e-05, + "loss": 0.9287, + "step": 33078 + }, + { + "epoch": 2.13, + "learning_rate": 1.2923205736490694e-05, + "loss": 0.9254, + "step": 33079 + }, + { + "epoch": 2.13, + "learning_rate": 1.2922991047467743e-05, + "loss": 0.917, + "step": 33080 + }, + { + "epoch": 2.13, + "learning_rate": 1.2922776358444792e-05, + "loss": 0.88, + "step": 33081 + }, + { + "epoch": 2.13, + "learning_rate": 1.2922561669421843e-05, + "loss": 0.8944, + "step": 33082 + }, + { + "epoch": 2.13, + "learning_rate": 1.2922346980398894e-05, + "loss": 0.8948, + "step": 33083 + }, + { + "epoch": 2.13, + "learning_rate": 1.2922132291375944e-05, + "loss": 0.8829, + "step": 33084 + }, + { + "epoch": 2.13, + "learning_rate": 1.2921917602352993e-05, + "loss": 0.8942, + "step": 33085 + }, + { + "epoch": 2.13, + "learning_rate": 1.2921702913330042e-05, + "loss": 0.8845, + "step": 33086 + }, + { + "epoch": 2.13, + "learning_rate": 1.2921488224307093e-05, + "loss": 0.8893, + "step": 33087 + }, + { + "epoch": 2.13, + "learning_rate": 1.2921273535284142e-05, + "loss": 0.8948, + "step": 33088 + }, + { + "epoch": 2.13, + "learning_rate": 1.2921058846261193e-05, + "loss": 0.9302, + "step": 33089 + }, + { + "epoch": 2.13, + "learning_rate": 1.2920844157238242e-05, + "loss": 0.913, + "step": 33090 + }, + { + "epoch": 2.13, + "learning_rate": 1.292062946821529e-05, + "loss": 0.8053, + "step": 33091 + }, + { + "epoch": 2.13, + "learning_rate": 1.2920414779192341e-05, + "loss": 0.9935, + "step": 33092 + }, + { + "epoch": 2.13, + "learning_rate": 1.292020009016939e-05, + "loss": 0.8662, + "step": 33093 + }, + { + "epoch": 2.13, + "learning_rate": 1.2919985401146439e-05, + "loss": 0.8532, + "step": 33094 + }, + { + "epoch": 2.13, + "learning_rate": 1.291977071212349e-05, + "loss": 0.8468, + "step": 33095 + }, + { + "epoch": 2.13, + "learning_rate": 1.2919556023100539e-05, + "loss": 0.9069, + "step": 33096 + }, + { + "epoch": 2.13, + "learning_rate": 1.291934133407759e-05, + "loss": 0.9347, + "step": 33097 + }, + { + "epoch": 2.13, + "learning_rate": 1.291912664505464e-05, + "loss": 0.8904, + "step": 33098 + }, + { + "epoch": 2.13, + "learning_rate": 1.291891195603169e-05, + "loss": 0.9009, + "step": 33099 + }, + { + "epoch": 2.13, + "learning_rate": 1.291869726700874e-05, + "loss": 0.9347, + "step": 33100 + }, + { + "epoch": 2.13, + "learning_rate": 1.2918482577985789e-05, + "loss": 0.8726, + "step": 33101 + }, + { + "epoch": 2.13, + "learning_rate": 1.291826788896284e-05, + "loss": 0.8834, + "step": 33102 + }, + { + "epoch": 2.13, + "learning_rate": 1.2918053199939888e-05, + "loss": 0.8449, + "step": 33103 + }, + { + "epoch": 2.13, + "learning_rate": 1.2917838510916937e-05, + "loss": 0.8893, + "step": 33104 + }, + { + "epoch": 2.13, + "learning_rate": 1.2917623821893988e-05, + "loss": 0.865, + "step": 33105 + }, + { + "epoch": 2.13, + "learning_rate": 1.2917409132871037e-05, + "loss": 0.9235, + "step": 33106 + }, + { + "epoch": 2.13, + "learning_rate": 1.2917194443848088e-05, + "loss": 0.8242, + "step": 33107 + }, + { + "epoch": 2.13, + "learning_rate": 1.2916979754825137e-05, + "loss": 0.9221, + "step": 33108 + }, + { + "epoch": 2.13, + "learning_rate": 1.2916765065802185e-05, + "loss": 0.9571, + "step": 33109 + }, + { + "epoch": 2.13, + "learning_rate": 1.2916550376779236e-05, + "loss": 0.9047, + "step": 33110 + }, + { + "epoch": 2.13, + "learning_rate": 1.2916335687756285e-05, + "loss": 0.9661, + "step": 33111 + }, + { + "epoch": 2.13, + "learning_rate": 1.2916120998733337e-05, + "loss": 0.9187, + "step": 33112 + }, + { + "epoch": 2.13, + "learning_rate": 1.2915906309710386e-05, + "loss": 0.9229, + "step": 33113 + }, + { + "epoch": 2.13, + "learning_rate": 1.2915691620687435e-05, + "loss": 0.8205, + "step": 33114 + }, + { + "epoch": 2.13, + "learning_rate": 1.2915476931664486e-05, + "loss": 0.8593, + "step": 33115 + }, + { + "epoch": 2.13, + "learning_rate": 1.2915262242641535e-05, + "loss": 0.9486, + "step": 33116 + }, + { + "epoch": 2.13, + "learning_rate": 1.2915047553618586e-05, + "loss": 0.8723, + "step": 33117 + }, + { + "epoch": 2.13, + "learning_rate": 1.2914832864595635e-05, + "loss": 0.8912, + "step": 33118 + }, + { + "epoch": 2.13, + "learning_rate": 1.2914618175572684e-05, + "loss": 0.8664, + "step": 33119 + }, + { + "epoch": 2.13, + "learning_rate": 1.2914403486549734e-05, + "loss": 0.9024, + "step": 33120 + }, + { + "epoch": 2.13, + "learning_rate": 1.2914188797526783e-05, + "loss": 0.847, + "step": 33121 + }, + { + "epoch": 2.13, + "learning_rate": 1.2913974108503832e-05, + "loss": 0.9023, + "step": 33122 + }, + { + "epoch": 2.13, + "learning_rate": 1.2913759419480883e-05, + "loss": 0.8949, + "step": 33123 + }, + { + "epoch": 2.13, + "learning_rate": 1.2913544730457932e-05, + "loss": 0.9392, + "step": 33124 + }, + { + "epoch": 2.13, + "learning_rate": 1.291333004143498e-05, + "loss": 0.9024, + "step": 33125 + }, + { + "epoch": 2.13, + "learning_rate": 1.2913115352412031e-05, + "loss": 0.9158, + "step": 33126 + }, + { + "epoch": 2.13, + "learning_rate": 1.2912900663389084e-05, + "loss": 0.9308, + "step": 33127 + }, + { + "epoch": 2.13, + "learning_rate": 1.2912685974366133e-05, + "loss": 0.8505, + "step": 33128 + }, + { + "epoch": 2.13, + "learning_rate": 1.2912471285343182e-05, + "loss": 0.8237, + "step": 33129 + }, + { + "epoch": 2.13, + "learning_rate": 1.2912256596320232e-05, + "loss": 0.8775, + "step": 33130 + }, + { + "epoch": 2.13, + "learning_rate": 1.2912041907297281e-05, + "loss": 0.8662, + "step": 33131 + }, + { + "epoch": 2.13, + "learning_rate": 1.291182721827433e-05, + "loss": 0.9148, + "step": 33132 + }, + { + "epoch": 2.13, + "learning_rate": 1.2911612529251381e-05, + "loss": 0.8469, + "step": 33133 + }, + { + "epoch": 2.13, + "learning_rate": 1.291139784022843e-05, + "loss": 0.8988, + "step": 33134 + }, + { + "epoch": 2.13, + "learning_rate": 1.2911183151205479e-05, + "loss": 0.9574, + "step": 33135 + }, + { + "epoch": 2.13, + "learning_rate": 1.291096846218253e-05, + "loss": 0.8938, + "step": 33136 + }, + { + "epoch": 2.13, + "learning_rate": 1.2910753773159579e-05, + "loss": 0.8067, + "step": 33137 + }, + { + "epoch": 2.13, + "learning_rate": 1.291053908413663e-05, + "loss": 0.8927, + "step": 33138 + }, + { + "epoch": 2.13, + "learning_rate": 1.2910324395113678e-05, + "loss": 0.9246, + "step": 33139 + }, + { + "epoch": 2.13, + "learning_rate": 1.2910109706090727e-05, + "loss": 0.871, + "step": 33140 + }, + { + "epoch": 2.13, + "learning_rate": 1.290989501706778e-05, + "loss": 0.9214, + "step": 33141 + }, + { + "epoch": 2.13, + "learning_rate": 1.2909680328044828e-05, + "loss": 0.9112, + "step": 33142 + }, + { + "epoch": 2.13, + "learning_rate": 1.2909465639021879e-05, + "loss": 0.8572, + "step": 33143 + }, + { + "epoch": 2.13, + "learning_rate": 1.2909250949998928e-05, + "loss": 0.8665, + "step": 33144 + }, + { + "epoch": 2.13, + "learning_rate": 1.2909036260975977e-05, + "loss": 0.8433, + "step": 33145 + }, + { + "epoch": 2.13, + "learning_rate": 1.2908821571953028e-05, + "loss": 0.916, + "step": 33146 + }, + { + "epoch": 2.13, + "learning_rate": 1.2908606882930077e-05, + "loss": 0.9012, + "step": 33147 + }, + { + "epoch": 2.13, + "learning_rate": 1.2908392193907127e-05, + "loss": 0.8731, + "step": 33148 + }, + { + "epoch": 2.13, + "learning_rate": 1.2908177504884176e-05, + "loss": 0.9131, + "step": 33149 + }, + { + "epoch": 2.13, + "learning_rate": 1.2907962815861225e-05, + "loss": 0.9501, + "step": 33150 + }, + { + "epoch": 2.13, + "learning_rate": 1.2907748126838276e-05, + "loss": 0.8675, + "step": 33151 + }, + { + "epoch": 2.13, + "learning_rate": 1.2907533437815325e-05, + "loss": 0.8954, + "step": 33152 + }, + { + "epoch": 2.13, + "learning_rate": 1.2907318748792374e-05, + "loss": 0.9008, + "step": 33153 + }, + { + "epoch": 2.13, + "learning_rate": 1.2907104059769424e-05, + "loss": 0.9557, + "step": 33154 + }, + { + "epoch": 2.13, + "learning_rate": 1.2906889370746473e-05, + "loss": 0.7405, + "step": 33155 + }, + { + "epoch": 2.13, + "learning_rate": 1.2906674681723526e-05, + "loss": 0.7914, + "step": 33156 + }, + { + "epoch": 2.13, + "learning_rate": 1.2906459992700575e-05, + "loss": 0.8698, + "step": 33157 + }, + { + "epoch": 2.13, + "learning_rate": 1.2906245303677625e-05, + "loss": 0.873, + "step": 33158 + }, + { + "epoch": 2.13, + "learning_rate": 1.2906030614654674e-05, + "loss": 0.9937, + "step": 33159 + }, + { + "epoch": 2.13, + "learning_rate": 1.2905815925631723e-05, + "loss": 0.8225, + "step": 33160 + }, + { + "epoch": 2.13, + "learning_rate": 1.2905601236608774e-05, + "loss": 0.8253, + "step": 33161 + }, + { + "epoch": 2.13, + "learning_rate": 1.2905386547585823e-05, + "loss": 0.858, + "step": 33162 + }, + { + "epoch": 2.13, + "learning_rate": 1.2905171858562872e-05, + "loss": 0.9096, + "step": 33163 + }, + { + "epoch": 2.13, + "learning_rate": 1.2904957169539923e-05, + "loss": 0.8883, + "step": 33164 + }, + { + "epoch": 2.13, + "learning_rate": 1.2904742480516972e-05, + "loss": 0.8727, + "step": 33165 + }, + { + "epoch": 2.13, + "learning_rate": 1.2904527791494022e-05, + "loss": 0.8392, + "step": 33166 + }, + { + "epoch": 2.13, + "learning_rate": 1.2904313102471071e-05, + "loss": 0.8775, + "step": 33167 + }, + { + "epoch": 2.13, + "learning_rate": 1.290409841344812e-05, + "loss": 0.8889, + "step": 33168 + }, + { + "epoch": 2.13, + "learning_rate": 1.290388372442517e-05, + "loss": 0.8082, + "step": 33169 + }, + { + "epoch": 2.13, + "learning_rate": 1.2903669035402221e-05, + "loss": 0.8813, + "step": 33170 + }, + { + "epoch": 2.13, + "learning_rate": 1.2903454346379272e-05, + "loss": 0.8129, + "step": 33171 + }, + { + "epoch": 2.13, + "learning_rate": 1.2903239657356321e-05, + "loss": 0.8213, + "step": 33172 + }, + { + "epoch": 2.13, + "learning_rate": 1.290302496833337e-05, + "loss": 0.8655, + "step": 33173 + }, + { + "epoch": 2.13, + "learning_rate": 1.290281027931042e-05, + "loss": 0.8727, + "step": 33174 + }, + { + "epoch": 2.13, + "learning_rate": 1.290259559028747e-05, + "loss": 0.9232, + "step": 33175 + }, + { + "epoch": 2.13, + "learning_rate": 1.290238090126452e-05, + "loss": 0.8798, + "step": 33176 + }, + { + "epoch": 2.13, + "learning_rate": 1.290216621224157e-05, + "loss": 0.9374, + "step": 33177 + }, + { + "epoch": 2.13, + "learning_rate": 1.2901951523218618e-05, + "loss": 0.8583, + "step": 33178 + }, + { + "epoch": 2.13, + "learning_rate": 1.2901736834195669e-05, + "loss": 0.9099, + "step": 33179 + }, + { + "epoch": 2.13, + "learning_rate": 1.2901522145172718e-05, + "loss": 0.8654, + "step": 33180 + }, + { + "epoch": 2.13, + "learning_rate": 1.2901307456149767e-05, + "loss": 0.9324, + "step": 33181 + }, + { + "epoch": 2.13, + "learning_rate": 1.2901092767126818e-05, + "loss": 0.8625, + "step": 33182 + }, + { + "epoch": 2.13, + "learning_rate": 1.2900878078103866e-05, + "loss": 0.8574, + "step": 33183 + }, + { + "epoch": 2.13, + "learning_rate": 1.2900663389080917e-05, + "loss": 0.9033, + "step": 33184 + }, + { + "epoch": 2.14, + "learning_rate": 1.2900448700057968e-05, + "loss": 0.9588, + "step": 33185 + }, + { + "epoch": 2.14, + "learning_rate": 1.2900234011035018e-05, + "loss": 0.8291, + "step": 33186 + }, + { + "epoch": 2.14, + "learning_rate": 1.2900019322012067e-05, + "loss": 0.9094, + "step": 33187 + }, + { + "epoch": 2.14, + "learning_rate": 1.2899804632989116e-05, + "loss": 0.948, + "step": 33188 + }, + { + "epoch": 2.14, + "learning_rate": 1.2899589943966167e-05, + "loss": 0.88, + "step": 33189 + }, + { + "epoch": 2.14, + "learning_rate": 1.2899375254943216e-05, + "loss": 0.8644, + "step": 33190 + }, + { + "epoch": 2.14, + "learning_rate": 1.2899160565920265e-05, + "loss": 0.8461, + "step": 33191 + }, + { + "epoch": 2.14, + "learning_rate": 1.2898945876897316e-05, + "loss": 0.8692, + "step": 33192 + }, + { + "epoch": 2.14, + "learning_rate": 1.2898731187874365e-05, + "loss": 0.9661, + "step": 33193 + }, + { + "epoch": 2.14, + "learning_rate": 1.2898516498851415e-05, + "loss": 0.9323, + "step": 33194 + }, + { + "epoch": 2.14, + "learning_rate": 1.2898301809828464e-05, + "loss": 0.8312, + "step": 33195 + }, + { + "epoch": 2.14, + "learning_rate": 1.2898087120805513e-05, + "loss": 0.9384, + "step": 33196 + }, + { + "epoch": 2.14, + "learning_rate": 1.2897872431782564e-05, + "loss": 0.8984, + "step": 33197 + }, + { + "epoch": 2.14, + "learning_rate": 1.2897657742759613e-05, + "loss": 0.95, + "step": 33198 + }, + { + "epoch": 2.14, + "learning_rate": 1.2897443053736665e-05, + "loss": 0.9389, + "step": 33199 + }, + { + "epoch": 2.14, + "learning_rate": 1.2897228364713714e-05, + "loss": 0.8707, + "step": 33200 + }, + { + "epoch": 2.14, + "learning_rate": 1.2897013675690763e-05, + "loss": 0.8848, + "step": 33201 + }, + { + "epoch": 2.14, + "learning_rate": 1.2896798986667814e-05, + "loss": 0.8847, + "step": 33202 + }, + { + "epoch": 2.14, + "learning_rate": 1.2896584297644863e-05, + "loss": 0.8876, + "step": 33203 + }, + { + "epoch": 2.14, + "learning_rate": 1.2896369608621913e-05, + "loss": 0.837, + "step": 33204 + }, + { + "epoch": 2.14, + "learning_rate": 1.2896154919598962e-05, + "loss": 0.8625, + "step": 33205 + }, + { + "epoch": 2.14, + "learning_rate": 1.2895940230576011e-05, + "loss": 0.9266, + "step": 33206 + }, + { + "epoch": 2.14, + "learning_rate": 1.2895725541553062e-05, + "loss": 0.8579, + "step": 33207 + }, + { + "epoch": 2.14, + "learning_rate": 1.2895510852530111e-05, + "loss": 0.8094, + "step": 33208 + }, + { + "epoch": 2.14, + "learning_rate": 1.289529616350716e-05, + "loss": 0.9037, + "step": 33209 + }, + { + "epoch": 2.14, + "learning_rate": 1.289508147448421e-05, + "loss": 0.9166, + "step": 33210 + }, + { + "epoch": 2.14, + "learning_rate": 1.289486678546126e-05, + "loss": 0.8725, + "step": 33211 + }, + { + "epoch": 2.14, + "learning_rate": 1.2894652096438308e-05, + "loss": 0.8731, + "step": 33212 + }, + { + "epoch": 2.14, + "learning_rate": 1.2894437407415359e-05, + "loss": 0.878, + "step": 33213 + }, + { + "epoch": 2.14, + "learning_rate": 1.2894222718392411e-05, + "loss": 0.8455, + "step": 33214 + }, + { + "epoch": 2.14, + "learning_rate": 1.289400802936946e-05, + "loss": 0.952, + "step": 33215 + }, + { + "epoch": 2.14, + "learning_rate": 1.289379334034651e-05, + "loss": 0.96, + "step": 33216 + }, + { + "epoch": 2.14, + "learning_rate": 1.289357865132356e-05, + "loss": 0.8561, + "step": 33217 + }, + { + "epoch": 2.14, + "learning_rate": 1.2893363962300609e-05, + "loss": 0.9548, + "step": 33218 + }, + { + "epoch": 2.14, + "learning_rate": 1.2893149273277658e-05, + "loss": 0.8443, + "step": 33219 + }, + { + "epoch": 2.14, + "learning_rate": 1.2892934584254709e-05, + "loss": 0.9789, + "step": 33220 + }, + { + "epoch": 2.14, + "learning_rate": 1.2892719895231758e-05, + "loss": 0.927, + "step": 33221 + }, + { + "epoch": 2.14, + "learning_rate": 1.2892505206208807e-05, + "loss": 0.8512, + "step": 33222 + }, + { + "epoch": 2.14, + "learning_rate": 1.2892290517185857e-05, + "loss": 0.9315, + "step": 33223 + }, + { + "epoch": 2.14, + "learning_rate": 1.2892075828162906e-05, + "loss": 0.9335, + "step": 33224 + }, + { + "epoch": 2.14, + "learning_rate": 1.2891861139139957e-05, + "loss": 0.9639, + "step": 33225 + }, + { + "epoch": 2.14, + "learning_rate": 1.2891646450117006e-05, + "loss": 0.8703, + "step": 33226 + }, + { + "epoch": 2.14, + "learning_rate": 1.2891431761094055e-05, + "loss": 0.8859, + "step": 33227 + }, + { + "epoch": 2.14, + "learning_rate": 1.2891217072071105e-05, + "loss": 0.8795, + "step": 33228 + }, + { + "epoch": 2.14, + "learning_rate": 1.2891002383048156e-05, + "loss": 0.8909, + "step": 33229 + }, + { + "epoch": 2.14, + "learning_rate": 1.2890787694025207e-05, + "loss": 0.9068, + "step": 33230 + }, + { + "epoch": 2.14, + "learning_rate": 1.2890573005002256e-05, + "loss": 0.8686, + "step": 33231 + }, + { + "epoch": 2.14, + "learning_rate": 1.2890358315979305e-05, + "loss": 0.8661, + "step": 33232 + }, + { + "epoch": 2.14, + "learning_rate": 1.2890143626956355e-05, + "loss": 0.9192, + "step": 33233 + }, + { + "epoch": 2.14, + "learning_rate": 1.2889928937933404e-05, + "loss": 0.8907, + "step": 33234 + }, + { + "epoch": 2.14, + "learning_rate": 1.2889714248910455e-05, + "loss": 0.8704, + "step": 33235 + }, + { + "epoch": 2.14, + "learning_rate": 1.2889499559887504e-05, + "loss": 0.8827, + "step": 33236 + }, + { + "epoch": 2.14, + "learning_rate": 1.2889284870864553e-05, + "loss": 0.8985, + "step": 33237 + }, + { + "epoch": 2.14, + "learning_rate": 1.2889070181841604e-05, + "loss": 0.9016, + "step": 33238 + }, + { + "epoch": 2.14, + "learning_rate": 1.2888855492818653e-05, + "loss": 0.9226, + "step": 33239 + }, + { + "epoch": 2.14, + "learning_rate": 1.2888640803795702e-05, + "loss": 0.8792, + "step": 33240 + }, + { + "epoch": 2.14, + "learning_rate": 1.2888426114772752e-05, + "loss": 0.8697, + "step": 33241 + }, + { + "epoch": 2.14, + "learning_rate": 1.2888211425749801e-05, + "loss": 0.9048, + "step": 33242 + }, + { + "epoch": 2.14, + "learning_rate": 1.2887996736726853e-05, + "loss": 0.8899, + "step": 33243 + }, + { + "epoch": 2.14, + "learning_rate": 1.2887782047703902e-05, + "loss": 0.8283, + "step": 33244 + }, + { + "epoch": 2.14, + "learning_rate": 1.2887567358680953e-05, + "loss": 0.9558, + "step": 33245 + }, + { + "epoch": 2.14, + "learning_rate": 1.2887352669658002e-05, + "loss": 0.945, + "step": 33246 + }, + { + "epoch": 2.14, + "learning_rate": 1.2887137980635051e-05, + "loss": 0.8887, + "step": 33247 + }, + { + "epoch": 2.14, + "learning_rate": 1.2886923291612102e-05, + "loss": 0.8905, + "step": 33248 + }, + { + "epoch": 2.14, + "learning_rate": 1.288670860258915e-05, + "loss": 0.8652, + "step": 33249 + }, + { + "epoch": 2.14, + "learning_rate": 1.28864939135662e-05, + "loss": 0.9361, + "step": 33250 + }, + { + "epoch": 2.14, + "learning_rate": 1.288627922454325e-05, + "loss": 0.8728, + "step": 33251 + }, + { + "epoch": 2.14, + "learning_rate": 1.28860645355203e-05, + "loss": 0.9164, + "step": 33252 + }, + { + "epoch": 2.14, + "learning_rate": 1.288584984649735e-05, + "loss": 0.9098, + "step": 33253 + }, + { + "epoch": 2.14, + "learning_rate": 1.2885635157474399e-05, + "loss": 0.8789, + "step": 33254 + }, + { + "epoch": 2.14, + "learning_rate": 1.2885420468451448e-05, + "loss": 0.8607, + "step": 33255 + }, + { + "epoch": 2.14, + "learning_rate": 1.2885205779428499e-05, + "loss": 0.9091, + "step": 33256 + }, + { + "epoch": 2.14, + "learning_rate": 1.2884991090405547e-05, + "loss": 0.8952, + "step": 33257 + }, + { + "epoch": 2.14, + "learning_rate": 1.28847764013826e-05, + "loss": 0.8584, + "step": 33258 + }, + { + "epoch": 2.14, + "learning_rate": 1.2884561712359649e-05, + "loss": 0.9423, + "step": 33259 + }, + { + "epoch": 2.14, + "learning_rate": 1.2884347023336698e-05, + "loss": 0.9221, + "step": 33260 + }, + { + "epoch": 2.14, + "learning_rate": 1.2884132334313748e-05, + "loss": 0.9086, + "step": 33261 + }, + { + "epoch": 2.14, + "learning_rate": 1.2883917645290797e-05, + "loss": 0.8927, + "step": 33262 + }, + { + "epoch": 2.14, + "learning_rate": 1.2883702956267848e-05, + "loss": 0.8184, + "step": 33263 + }, + { + "epoch": 2.14, + "learning_rate": 1.2883488267244897e-05, + "loss": 0.9223, + "step": 33264 + }, + { + "epoch": 2.14, + "learning_rate": 1.2883273578221946e-05, + "loss": 0.9086, + "step": 33265 + }, + { + "epoch": 2.14, + "learning_rate": 1.2883058889198997e-05, + "loss": 0.9059, + "step": 33266 + }, + { + "epoch": 2.14, + "learning_rate": 1.2882844200176046e-05, + "loss": 0.8607, + "step": 33267 + }, + { + "epoch": 2.14, + "learning_rate": 1.2882629511153095e-05, + "loss": 0.9433, + "step": 33268 + }, + { + "epoch": 2.14, + "learning_rate": 1.2882414822130145e-05, + "loss": 0.9157, + "step": 33269 + }, + { + "epoch": 2.14, + "learning_rate": 1.2882200133107194e-05, + "loss": 0.9326, + "step": 33270 + }, + { + "epoch": 2.14, + "learning_rate": 1.2881985444084245e-05, + "loss": 0.8341, + "step": 33271 + }, + { + "epoch": 2.14, + "learning_rate": 1.2881770755061295e-05, + "loss": 0.9128, + "step": 33272 + }, + { + "epoch": 2.14, + "learning_rate": 1.2881556066038346e-05, + "loss": 0.9217, + "step": 33273 + }, + { + "epoch": 2.14, + "learning_rate": 1.2881341377015395e-05, + "loss": 0.9303, + "step": 33274 + }, + { + "epoch": 2.14, + "learning_rate": 1.2881126687992444e-05, + "loss": 0.8567, + "step": 33275 + }, + { + "epoch": 2.14, + "learning_rate": 1.2880911998969495e-05, + "loss": 0.9422, + "step": 33276 + }, + { + "epoch": 2.14, + "learning_rate": 1.2880697309946544e-05, + "loss": 0.8969, + "step": 33277 + }, + { + "epoch": 2.14, + "learning_rate": 1.2880482620923593e-05, + "loss": 0.9127, + "step": 33278 + }, + { + "epoch": 2.14, + "learning_rate": 1.2880267931900643e-05, + "loss": 0.8771, + "step": 33279 + }, + { + "epoch": 2.14, + "learning_rate": 1.2880053242877692e-05, + "loss": 0.8841, + "step": 33280 + }, + { + "epoch": 2.14, + "learning_rate": 1.2879838553854743e-05, + "loss": 0.8569, + "step": 33281 + }, + { + "epoch": 2.14, + "learning_rate": 1.2879623864831792e-05, + "loss": 0.8825, + "step": 33282 + }, + { + "epoch": 2.14, + "learning_rate": 1.2879409175808841e-05, + "loss": 0.9188, + "step": 33283 + }, + { + "epoch": 2.14, + "learning_rate": 1.2879194486785892e-05, + "loss": 0.8596, + "step": 33284 + }, + { + "epoch": 2.14, + "learning_rate": 1.287897979776294e-05, + "loss": 0.9696, + "step": 33285 + }, + { + "epoch": 2.14, + "learning_rate": 1.287876510873999e-05, + "loss": 0.8376, + "step": 33286 + }, + { + "epoch": 2.14, + "learning_rate": 1.2878550419717042e-05, + "loss": 0.9037, + "step": 33287 + }, + { + "epoch": 2.14, + "learning_rate": 1.287833573069409e-05, + "loss": 0.8792, + "step": 33288 + }, + { + "epoch": 2.14, + "learning_rate": 1.2878121041671141e-05, + "loss": 0.9403, + "step": 33289 + }, + { + "epoch": 2.14, + "learning_rate": 1.287790635264819e-05, + "loss": 0.9212, + "step": 33290 + }, + { + "epoch": 2.14, + "learning_rate": 1.2877691663625241e-05, + "loss": 1.0077, + "step": 33291 + }, + { + "epoch": 2.14, + "learning_rate": 1.287747697460229e-05, + "loss": 0.9475, + "step": 33292 + }, + { + "epoch": 2.14, + "learning_rate": 1.2877262285579339e-05, + "loss": 0.8629, + "step": 33293 + }, + { + "epoch": 2.14, + "learning_rate": 1.287704759655639e-05, + "loss": 0.8378, + "step": 33294 + }, + { + "epoch": 2.14, + "learning_rate": 1.2876832907533439e-05, + "loss": 0.9143, + "step": 33295 + }, + { + "epoch": 2.14, + "learning_rate": 1.2876618218510488e-05, + "loss": 0.8447, + "step": 33296 + }, + { + "epoch": 2.14, + "learning_rate": 1.2876403529487538e-05, + "loss": 0.9257, + "step": 33297 + }, + { + "epoch": 2.14, + "learning_rate": 1.2876188840464587e-05, + "loss": 0.8824, + "step": 33298 + }, + { + "epoch": 2.14, + "learning_rate": 1.2875974151441636e-05, + "loss": 0.9464, + "step": 33299 + }, + { + "epoch": 2.14, + "learning_rate": 1.2875759462418687e-05, + "loss": 0.8708, + "step": 33300 + }, + { + "epoch": 2.14, + "learning_rate": 1.287554477339574e-05, + "loss": 0.8472, + "step": 33301 + }, + { + "epoch": 2.14, + "learning_rate": 1.2875330084372788e-05, + "loss": 0.9045, + "step": 33302 + }, + { + "epoch": 2.14, + "learning_rate": 1.2875115395349837e-05, + "loss": 0.9074, + "step": 33303 + }, + { + "epoch": 2.14, + "learning_rate": 1.2874900706326888e-05, + "loss": 0.9185, + "step": 33304 + }, + { + "epoch": 2.14, + "learning_rate": 1.2874686017303937e-05, + "loss": 0.8911, + "step": 33305 + }, + { + "epoch": 2.14, + "learning_rate": 1.2874471328280986e-05, + "loss": 0.9102, + "step": 33306 + }, + { + "epoch": 2.14, + "learning_rate": 1.2874256639258036e-05, + "loss": 0.9272, + "step": 33307 + }, + { + "epoch": 2.14, + "learning_rate": 1.2874041950235085e-05, + "loss": 0.8766, + "step": 33308 + }, + { + "epoch": 2.14, + "learning_rate": 1.2873827261212134e-05, + "loss": 0.8742, + "step": 33309 + }, + { + "epoch": 2.14, + "learning_rate": 1.2873612572189185e-05, + "loss": 0.9007, + "step": 33310 + }, + { + "epoch": 2.14, + "learning_rate": 1.2873397883166234e-05, + "loss": 0.8604, + "step": 33311 + }, + { + "epoch": 2.14, + "learning_rate": 1.2873183194143285e-05, + "loss": 0.9118, + "step": 33312 + }, + { + "epoch": 2.14, + "learning_rate": 1.2872968505120334e-05, + "loss": 0.9118, + "step": 33313 + }, + { + "epoch": 2.14, + "learning_rate": 1.2872753816097383e-05, + "loss": 0.9552, + "step": 33314 + }, + { + "epoch": 2.14, + "learning_rate": 1.2872539127074433e-05, + "loss": 0.8686, + "step": 33315 + }, + { + "epoch": 2.14, + "learning_rate": 1.2872324438051484e-05, + "loss": 0.949, + "step": 33316 + }, + { + "epoch": 2.14, + "learning_rate": 1.2872109749028534e-05, + "loss": 0.8609, + "step": 33317 + }, + { + "epoch": 2.14, + "learning_rate": 1.2871895060005583e-05, + "loss": 0.8237, + "step": 33318 + }, + { + "epoch": 2.14, + "learning_rate": 1.2871680370982632e-05, + "loss": 0.8905, + "step": 33319 + }, + { + "epoch": 2.14, + "learning_rate": 1.2871465681959683e-05, + "loss": 0.9232, + "step": 33320 + }, + { + "epoch": 2.14, + "learning_rate": 1.2871250992936732e-05, + "loss": 0.9211, + "step": 33321 + }, + { + "epoch": 2.14, + "learning_rate": 1.2871036303913783e-05, + "loss": 0.902, + "step": 33322 + }, + { + "epoch": 2.14, + "learning_rate": 1.2870821614890832e-05, + "loss": 0.8745, + "step": 33323 + }, + { + "epoch": 2.14, + "learning_rate": 1.287060692586788e-05, + "loss": 0.935, + "step": 33324 + }, + { + "epoch": 2.14, + "learning_rate": 1.2870392236844931e-05, + "loss": 0.789, + "step": 33325 + }, + { + "epoch": 2.14, + "learning_rate": 1.287017754782198e-05, + "loss": 0.8012, + "step": 33326 + }, + { + "epoch": 2.14, + "learning_rate": 1.286996285879903e-05, + "loss": 0.9238, + "step": 33327 + }, + { + "epoch": 2.14, + "learning_rate": 1.286974816977608e-05, + "loss": 0.846, + "step": 33328 + }, + { + "epoch": 2.14, + "learning_rate": 1.2869533480753129e-05, + "loss": 0.8607, + "step": 33329 + }, + { + "epoch": 2.14, + "learning_rate": 1.2869318791730181e-05, + "loss": 0.825, + "step": 33330 + }, + { + "epoch": 2.14, + "learning_rate": 1.286910410270723e-05, + "loss": 0.9286, + "step": 33331 + }, + { + "epoch": 2.14, + "learning_rate": 1.286888941368428e-05, + "loss": 0.9122, + "step": 33332 + }, + { + "epoch": 2.14, + "learning_rate": 1.286867472466133e-05, + "loss": 0.9267, + "step": 33333 + }, + { + "epoch": 2.14, + "learning_rate": 1.2868460035638379e-05, + "loss": 0.8975, + "step": 33334 + }, + { + "epoch": 2.14, + "learning_rate": 1.286824534661543e-05, + "loss": 0.8451, + "step": 33335 + }, + { + "epoch": 2.14, + "learning_rate": 1.2868030657592478e-05, + "loss": 0.9208, + "step": 33336 + }, + { + "epoch": 2.14, + "learning_rate": 1.2867815968569527e-05, + "loss": 0.9065, + "step": 33337 + }, + { + "epoch": 2.14, + "learning_rate": 1.2867601279546578e-05, + "loss": 0.8611, + "step": 33338 + }, + { + "epoch": 2.14, + "learning_rate": 1.2867386590523627e-05, + "loss": 0.9021, + "step": 33339 + }, + { + "epoch": 2.15, + "learning_rate": 1.2867171901500678e-05, + "loss": 0.9209, + "step": 33340 + }, + { + "epoch": 2.15, + "learning_rate": 1.2866957212477727e-05, + "loss": 0.9289, + "step": 33341 + }, + { + "epoch": 2.15, + "learning_rate": 1.2866742523454776e-05, + "loss": 0.8663, + "step": 33342 + }, + { + "epoch": 2.15, + "learning_rate": 1.2866527834431826e-05, + "loss": 0.9187, + "step": 33343 + }, + { + "epoch": 2.15, + "learning_rate": 1.2866313145408875e-05, + "loss": 0.8781, + "step": 33344 + }, + { + "epoch": 2.15, + "learning_rate": 1.2866098456385928e-05, + "loss": 0.9164, + "step": 33345 + }, + { + "epoch": 2.15, + "learning_rate": 1.2865883767362977e-05, + "loss": 0.9251, + "step": 33346 + }, + { + "epoch": 2.15, + "learning_rate": 1.2865669078340025e-05, + "loss": 0.9322, + "step": 33347 + }, + { + "epoch": 2.15, + "learning_rate": 1.2865454389317076e-05, + "loss": 0.9242, + "step": 33348 + }, + { + "epoch": 2.15, + "learning_rate": 1.2865239700294125e-05, + "loss": 0.8446, + "step": 33349 + }, + { + "epoch": 2.15, + "learning_rate": 1.2865025011271176e-05, + "loss": 0.8578, + "step": 33350 + }, + { + "epoch": 2.15, + "learning_rate": 1.2864810322248225e-05, + "loss": 0.8766, + "step": 33351 + }, + { + "epoch": 2.15, + "learning_rate": 1.2864595633225274e-05, + "loss": 0.8855, + "step": 33352 + }, + { + "epoch": 2.15, + "learning_rate": 1.2864380944202324e-05, + "loss": 0.8977, + "step": 33353 + }, + { + "epoch": 2.15, + "learning_rate": 1.2864166255179373e-05, + "loss": 0.8678, + "step": 33354 + }, + { + "epoch": 2.15, + "learning_rate": 1.2863951566156422e-05, + "loss": 0.9211, + "step": 33355 + }, + { + "epoch": 2.15, + "learning_rate": 1.2863736877133473e-05, + "loss": 0.8294, + "step": 33356 + }, + { + "epoch": 2.15, + "learning_rate": 1.2863522188110522e-05, + "loss": 0.9198, + "step": 33357 + }, + { + "epoch": 2.15, + "learning_rate": 1.2863307499087571e-05, + "loss": 0.9136, + "step": 33358 + }, + { + "epoch": 2.15, + "learning_rate": 1.2863092810064623e-05, + "loss": 0.9545, + "step": 33359 + }, + { + "epoch": 2.15, + "learning_rate": 1.2862878121041674e-05, + "loss": 0.8914, + "step": 33360 + }, + { + "epoch": 2.15, + "learning_rate": 1.2862663432018723e-05, + "loss": 0.878, + "step": 33361 + }, + { + "epoch": 2.15, + "learning_rate": 1.2862448742995772e-05, + "loss": 0.8966, + "step": 33362 + }, + { + "epoch": 2.15, + "learning_rate": 1.2862234053972822e-05, + "loss": 0.8444, + "step": 33363 + }, + { + "epoch": 2.15, + "learning_rate": 1.2862019364949871e-05, + "loss": 0.8869, + "step": 33364 + }, + { + "epoch": 2.15, + "learning_rate": 1.286180467592692e-05, + "loss": 0.8149, + "step": 33365 + }, + { + "epoch": 2.15, + "learning_rate": 1.2861589986903971e-05, + "loss": 0.9187, + "step": 33366 + }, + { + "epoch": 2.15, + "learning_rate": 1.286137529788102e-05, + "loss": 0.9247, + "step": 33367 + }, + { + "epoch": 2.15, + "learning_rate": 1.286116060885807e-05, + "loss": 0.9186, + "step": 33368 + }, + { + "epoch": 2.15, + "learning_rate": 1.286094591983512e-05, + "loss": 0.9601, + "step": 33369 + }, + { + "epoch": 2.15, + "learning_rate": 1.2860731230812169e-05, + "loss": 0.8168, + "step": 33370 + }, + { + "epoch": 2.15, + "learning_rate": 1.286051654178922e-05, + "loss": 0.9455, + "step": 33371 + }, + { + "epoch": 2.15, + "learning_rate": 1.2860301852766268e-05, + "loss": 0.9272, + "step": 33372 + }, + { + "epoch": 2.15, + "learning_rate": 1.2860087163743317e-05, + "loss": 0.8655, + "step": 33373 + }, + { + "epoch": 2.15, + "learning_rate": 1.285987247472037e-05, + "loss": 0.9408, + "step": 33374 + }, + { + "epoch": 2.15, + "learning_rate": 1.2859657785697419e-05, + "loss": 0.8499, + "step": 33375 + }, + { + "epoch": 2.15, + "learning_rate": 1.285944309667447e-05, + "loss": 0.9678, + "step": 33376 + }, + { + "epoch": 2.15, + "learning_rate": 1.2859228407651518e-05, + "loss": 0.9384, + "step": 33377 + }, + { + "epoch": 2.15, + "learning_rate": 1.2859013718628569e-05, + "loss": 0.9117, + "step": 33378 + }, + { + "epoch": 2.15, + "learning_rate": 1.2858799029605618e-05, + "loss": 0.8671, + "step": 33379 + }, + { + "epoch": 2.15, + "learning_rate": 1.2858584340582667e-05, + "loss": 0.8795, + "step": 33380 + }, + { + "epoch": 2.15, + "learning_rate": 1.2858369651559717e-05, + "loss": 0.9221, + "step": 33381 + }, + { + "epoch": 2.15, + "learning_rate": 1.2858154962536766e-05, + "loss": 0.9108, + "step": 33382 + }, + { + "epoch": 2.15, + "learning_rate": 1.2857940273513815e-05, + "loss": 0.8743, + "step": 33383 + }, + { + "epoch": 2.15, + "learning_rate": 1.2857725584490866e-05, + "loss": 0.8674, + "step": 33384 + }, + { + "epoch": 2.15, + "learning_rate": 1.2857510895467915e-05, + "loss": 0.9285, + "step": 33385 + }, + { + "epoch": 2.15, + "learning_rate": 1.2857296206444964e-05, + "loss": 0.9376, + "step": 33386 + }, + { + "epoch": 2.15, + "learning_rate": 1.2857081517422015e-05, + "loss": 0.8249, + "step": 33387 + }, + { + "epoch": 2.15, + "learning_rate": 1.2856866828399067e-05, + "loss": 0.8471, + "step": 33388 + }, + { + "epoch": 2.15, + "learning_rate": 1.2856652139376116e-05, + "loss": 0.8813, + "step": 33389 + }, + { + "epoch": 2.15, + "learning_rate": 1.2856437450353165e-05, + "loss": 0.9129, + "step": 33390 + }, + { + "epoch": 2.15, + "learning_rate": 1.2856222761330215e-05, + "loss": 0.8641, + "step": 33391 + }, + { + "epoch": 2.15, + "learning_rate": 1.2856008072307264e-05, + "loss": 0.8415, + "step": 33392 + }, + { + "epoch": 2.15, + "learning_rate": 1.2855793383284313e-05, + "loss": 0.9556, + "step": 33393 + }, + { + "epoch": 2.15, + "learning_rate": 1.2855578694261364e-05, + "loss": 0.8561, + "step": 33394 + }, + { + "epoch": 2.15, + "learning_rate": 1.2855364005238413e-05, + "loss": 0.895, + "step": 33395 + }, + { + "epoch": 2.15, + "learning_rate": 1.2855149316215462e-05, + "loss": 0.8376, + "step": 33396 + }, + { + "epoch": 2.15, + "learning_rate": 1.2854934627192513e-05, + "loss": 0.9179, + "step": 33397 + }, + { + "epoch": 2.15, + "learning_rate": 1.2854719938169562e-05, + "loss": 0.9015, + "step": 33398 + }, + { + "epoch": 2.15, + "learning_rate": 1.2854505249146612e-05, + "loss": 0.9462, + "step": 33399 + }, + { + "epoch": 2.15, + "learning_rate": 1.2854290560123661e-05, + "loss": 0.9175, + "step": 33400 + }, + { + "epoch": 2.15, + "learning_rate": 1.285407587110071e-05, + "loss": 0.8522, + "step": 33401 + }, + { + "epoch": 2.15, + "learning_rate": 1.2853861182077761e-05, + "loss": 0.9052, + "step": 33402 + }, + { + "epoch": 2.15, + "learning_rate": 1.2853646493054812e-05, + "loss": 0.8966, + "step": 33403 + }, + { + "epoch": 2.15, + "learning_rate": 1.2853431804031862e-05, + "loss": 0.9115, + "step": 33404 + }, + { + "epoch": 2.15, + "learning_rate": 1.2853217115008911e-05, + "loss": 0.8962, + "step": 33405 + }, + { + "epoch": 2.15, + "learning_rate": 1.285300242598596e-05, + "loss": 0.9478, + "step": 33406 + }, + { + "epoch": 2.15, + "learning_rate": 1.285278773696301e-05, + "loss": 0.9327, + "step": 33407 + }, + { + "epoch": 2.15, + "learning_rate": 1.285257304794006e-05, + "loss": 0.8671, + "step": 33408 + }, + { + "epoch": 2.15, + "learning_rate": 1.285235835891711e-05, + "loss": 0.9191, + "step": 33409 + }, + { + "epoch": 2.15, + "learning_rate": 1.285214366989416e-05, + "loss": 0.8817, + "step": 33410 + }, + { + "epoch": 2.15, + "learning_rate": 1.2851928980871208e-05, + "loss": 0.8544, + "step": 33411 + }, + { + "epoch": 2.15, + "learning_rate": 1.2851714291848259e-05, + "loss": 0.9302, + "step": 33412 + }, + { + "epoch": 2.15, + "learning_rate": 1.2851499602825308e-05, + "loss": 0.8424, + "step": 33413 + }, + { + "epoch": 2.15, + "learning_rate": 1.2851284913802357e-05, + "loss": 0.9278, + "step": 33414 + }, + { + "epoch": 2.15, + "learning_rate": 1.2851070224779408e-05, + "loss": 0.9328, + "step": 33415 + }, + { + "epoch": 2.15, + "learning_rate": 1.2850855535756457e-05, + "loss": 0.8979, + "step": 33416 + }, + { + "epoch": 2.15, + "learning_rate": 1.2850640846733509e-05, + "loss": 0.9092, + "step": 33417 + }, + { + "epoch": 2.15, + "learning_rate": 1.2850426157710558e-05, + "loss": 0.8272, + "step": 33418 + }, + { + "epoch": 2.15, + "learning_rate": 1.2850211468687609e-05, + "loss": 0.874, + "step": 33419 + }, + { + "epoch": 2.15, + "learning_rate": 1.2849996779664658e-05, + "loss": 0.9063, + "step": 33420 + }, + { + "epoch": 2.15, + "learning_rate": 1.2849782090641706e-05, + "loss": 0.872, + "step": 33421 + }, + { + "epoch": 2.15, + "learning_rate": 1.2849567401618757e-05, + "loss": 0.9399, + "step": 33422 + }, + { + "epoch": 2.15, + "learning_rate": 1.2849352712595806e-05, + "loss": 0.8593, + "step": 33423 + }, + { + "epoch": 2.15, + "learning_rate": 1.2849138023572855e-05, + "loss": 0.945, + "step": 33424 + }, + { + "epoch": 2.15, + "learning_rate": 1.2848923334549906e-05, + "loss": 0.885, + "step": 33425 + }, + { + "epoch": 2.15, + "learning_rate": 1.2848708645526955e-05, + "loss": 0.8756, + "step": 33426 + }, + { + "epoch": 2.15, + "learning_rate": 1.2848493956504005e-05, + "loss": 0.895, + "step": 33427 + }, + { + "epoch": 2.15, + "learning_rate": 1.2848279267481054e-05, + "loss": 0.9016, + "step": 33428 + }, + { + "epoch": 2.15, + "learning_rate": 1.2848064578458103e-05, + "loss": 0.8837, + "step": 33429 + }, + { + "epoch": 2.15, + "learning_rate": 1.2847849889435154e-05, + "loss": 0.9185, + "step": 33430 + }, + { + "epoch": 2.15, + "learning_rate": 1.2847635200412203e-05, + "loss": 0.8616, + "step": 33431 + }, + { + "epoch": 2.15, + "learning_rate": 1.2847420511389255e-05, + "loss": 0.9486, + "step": 33432 + }, + { + "epoch": 2.15, + "learning_rate": 1.2847205822366304e-05, + "loss": 0.8608, + "step": 33433 + }, + { + "epoch": 2.15, + "learning_rate": 1.2846991133343353e-05, + "loss": 0.8834, + "step": 33434 + }, + { + "epoch": 2.15, + "learning_rate": 1.2846776444320404e-05, + "loss": 0.8719, + "step": 33435 + }, + { + "epoch": 2.15, + "learning_rate": 1.2846561755297453e-05, + "loss": 0.8844, + "step": 33436 + }, + { + "epoch": 2.15, + "learning_rate": 1.2846347066274503e-05, + "loss": 0.911, + "step": 33437 + }, + { + "epoch": 2.15, + "learning_rate": 1.2846132377251552e-05, + "loss": 0.8913, + "step": 33438 + }, + { + "epoch": 2.15, + "learning_rate": 1.2845917688228601e-05, + "loss": 0.9242, + "step": 33439 + }, + { + "epoch": 2.15, + "learning_rate": 1.2845702999205652e-05, + "loss": 0.9194, + "step": 33440 + }, + { + "epoch": 2.15, + "learning_rate": 1.2845488310182701e-05, + "loss": 0.927, + "step": 33441 + }, + { + "epoch": 2.15, + "learning_rate": 1.284527362115975e-05, + "loss": 0.9291, + "step": 33442 + }, + { + "epoch": 2.15, + "learning_rate": 1.28450589321368e-05, + "loss": 0.8732, + "step": 33443 + }, + { + "epoch": 2.15, + "learning_rate": 1.284484424311385e-05, + "loss": 0.9009, + "step": 33444 + }, + { + "epoch": 2.15, + "learning_rate": 1.2844629554090899e-05, + "loss": 0.9717, + "step": 33445 + }, + { + "epoch": 2.15, + "learning_rate": 1.2844414865067951e-05, + "loss": 0.8644, + "step": 33446 + }, + { + "epoch": 2.15, + "learning_rate": 1.2844200176045002e-05, + "loss": 0.9321, + "step": 33447 + }, + { + "epoch": 2.15, + "learning_rate": 1.284398548702205e-05, + "loss": 0.7883, + "step": 33448 + }, + { + "epoch": 2.15, + "learning_rate": 1.28437707979991e-05, + "loss": 0.9105, + "step": 33449 + }, + { + "epoch": 2.15, + "learning_rate": 1.284355610897615e-05, + "loss": 1.008, + "step": 33450 + }, + { + "epoch": 2.15, + "learning_rate": 1.2843341419953199e-05, + "loss": 0.8851, + "step": 33451 + }, + { + "epoch": 2.15, + "learning_rate": 1.2843126730930248e-05, + "loss": 0.8189, + "step": 33452 + }, + { + "epoch": 2.15, + "learning_rate": 1.2842912041907299e-05, + "loss": 0.8923, + "step": 33453 + }, + { + "epoch": 2.15, + "learning_rate": 1.2842697352884348e-05, + "loss": 0.9164, + "step": 33454 + }, + { + "epoch": 2.15, + "learning_rate": 1.2842482663861397e-05, + "loss": 0.9828, + "step": 33455 + }, + { + "epoch": 2.15, + "learning_rate": 1.2842267974838447e-05, + "loss": 0.863, + "step": 33456 + }, + { + "epoch": 2.15, + "learning_rate": 1.2842053285815496e-05, + "loss": 0.896, + "step": 33457 + }, + { + "epoch": 2.15, + "learning_rate": 1.2841838596792547e-05, + "loss": 0.9415, + "step": 33458 + }, + { + "epoch": 2.15, + "learning_rate": 1.2841623907769596e-05, + "loss": 0.9233, + "step": 33459 + }, + { + "epoch": 2.15, + "learning_rate": 1.2841409218746645e-05, + "loss": 0.8897, + "step": 33460 + }, + { + "epoch": 2.15, + "learning_rate": 1.2841194529723697e-05, + "loss": 0.8901, + "step": 33461 + }, + { + "epoch": 2.15, + "learning_rate": 1.2840979840700746e-05, + "loss": 0.8361, + "step": 33462 + }, + { + "epoch": 2.15, + "learning_rate": 1.2840765151677797e-05, + "loss": 0.8452, + "step": 33463 + }, + { + "epoch": 2.15, + "learning_rate": 1.2840550462654846e-05, + "loss": 0.8592, + "step": 33464 + }, + { + "epoch": 2.15, + "learning_rate": 1.2840335773631895e-05, + "loss": 0.8428, + "step": 33465 + }, + { + "epoch": 2.15, + "learning_rate": 1.2840121084608945e-05, + "loss": 0.7854, + "step": 33466 + }, + { + "epoch": 2.15, + "learning_rate": 1.2839906395585994e-05, + "loss": 0.8635, + "step": 33467 + }, + { + "epoch": 2.15, + "learning_rate": 1.2839691706563045e-05, + "loss": 0.8604, + "step": 33468 + }, + { + "epoch": 2.15, + "learning_rate": 1.2839477017540094e-05, + "loss": 0.8989, + "step": 33469 + }, + { + "epoch": 2.15, + "learning_rate": 1.2839262328517143e-05, + "loss": 0.8037, + "step": 33470 + }, + { + "epoch": 2.15, + "learning_rate": 1.2839047639494194e-05, + "loss": 0.8373, + "step": 33471 + }, + { + "epoch": 2.15, + "learning_rate": 1.2838832950471243e-05, + "loss": 0.8929, + "step": 33472 + }, + { + "epoch": 2.15, + "learning_rate": 1.2838618261448292e-05, + "loss": 0.8992, + "step": 33473 + }, + { + "epoch": 2.15, + "learning_rate": 1.2838403572425342e-05, + "loss": 0.8795, + "step": 33474 + }, + { + "epoch": 2.15, + "learning_rate": 1.2838188883402395e-05, + "loss": 0.8707, + "step": 33475 + }, + { + "epoch": 2.15, + "learning_rate": 1.2837974194379444e-05, + "loss": 0.9135, + "step": 33476 + }, + { + "epoch": 2.15, + "learning_rate": 1.2837759505356493e-05, + "loss": 0.8558, + "step": 33477 + }, + { + "epoch": 2.15, + "learning_rate": 1.2837544816333543e-05, + "loss": 0.9369, + "step": 33478 + }, + { + "epoch": 2.15, + "learning_rate": 1.2837330127310592e-05, + "loss": 0.881, + "step": 33479 + }, + { + "epoch": 2.15, + "learning_rate": 1.2837115438287641e-05, + "loss": 0.8918, + "step": 33480 + }, + { + "epoch": 2.15, + "learning_rate": 1.2836900749264692e-05, + "loss": 0.8523, + "step": 33481 + }, + { + "epoch": 2.15, + "learning_rate": 1.283668606024174e-05, + "loss": 0.833, + "step": 33482 + }, + { + "epoch": 2.15, + "learning_rate": 1.283647137121879e-05, + "loss": 0.8753, + "step": 33483 + }, + { + "epoch": 2.15, + "learning_rate": 1.283625668219584e-05, + "loss": 0.9082, + "step": 33484 + }, + { + "epoch": 2.15, + "learning_rate": 1.283604199317289e-05, + "loss": 0.8638, + "step": 33485 + }, + { + "epoch": 2.15, + "learning_rate": 1.283582730414994e-05, + "loss": 0.8951, + "step": 33486 + }, + { + "epoch": 2.15, + "learning_rate": 1.2835612615126989e-05, + "loss": 0.9189, + "step": 33487 + }, + { + "epoch": 2.15, + "learning_rate": 1.2835397926104038e-05, + "loss": 0.9986, + "step": 33488 + }, + { + "epoch": 2.15, + "learning_rate": 1.2835183237081089e-05, + "loss": 0.8495, + "step": 33489 + }, + { + "epoch": 2.15, + "learning_rate": 1.283496854805814e-05, + "loss": 0.9054, + "step": 33490 + }, + { + "epoch": 2.15, + "learning_rate": 1.283475385903519e-05, + "loss": 0.8966, + "step": 33491 + }, + { + "epoch": 2.15, + "learning_rate": 1.2834539170012239e-05, + "loss": 0.8815, + "step": 33492 + }, + { + "epoch": 2.15, + "learning_rate": 1.2834324480989288e-05, + "loss": 0.9179, + "step": 33493 + }, + { + "epoch": 2.15, + "learning_rate": 1.2834109791966339e-05, + "loss": 0.8423, + "step": 33494 + }, + { + "epoch": 2.15, + "learning_rate": 1.2833895102943387e-05, + "loss": 0.9317, + "step": 33495 + }, + { + "epoch": 2.16, + "learning_rate": 1.2833680413920438e-05, + "loss": 0.8757, + "step": 33496 + }, + { + "epoch": 2.16, + "learning_rate": 1.2833465724897487e-05, + "loss": 0.8939, + "step": 33497 + }, + { + "epoch": 2.16, + "learning_rate": 1.2833251035874536e-05, + "loss": 0.8973, + "step": 33498 + }, + { + "epoch": 2.16, + "learning_rate": 1.2833036346851587e-05, + "loss": 0.8824, + "step": 33499 + }, + { + "epoch": 2.16, + "learning_rate": 1.2832821657828636e-05, + "loss": 0.8633, + "step": 33500 + }, + { + "epoch": 2.16, + "learning_rate": 1.2832606968805685e-05, + "loss": 0.8756, + "step": 33501 + }, + { + "epoch": 2.16, + "learning_rate": 1.2832392279782735e-05, + "loss": 0.9754, + "step": 33502 + }, + { + "epoch": 2.16, + "learning_rate": 1.2832177590759784e-05, + "loss": 0.9031, + "step": 33503 + }, + { + "epoch": 2.16, + "learning_rate": 1.2831962901736837e-05, + "loss": 0.8606, + "step": 33504 + }, + { + "epoch": 2.16, + "learning_rate": 1.2831748212713886e-05, + "loss": 0.8662, + "step": 33505 + }, + { + "epoch": 2.16, + "learning_rate": 1.2831533523690936e-05, + "loss": 0.8658, + "step": 33506 + }, + { + "epoch": 2.16, + "learning_rate": 1.2831318834667985e-05, + "loss": 0.8455, + "step": 33507 + }, + { + "epoch": 2.16, + "learning_rate": 1.2831104145645034e-05, + "loss": 0.9338, + "step": 33508 + }, + { + "epoch": 2.16, + "learning_rate": 1.2830889456622085e-05, + "loss": 0.8673, + "step": 33509 + }, + { + "epoch": 2.16, + "learning_rate": 1.2830674767599134e-05, + "loss": 0.9718, + "step": 33510 + }, + { + "epoch": 2.16, + "learning_rate": 1.2830460078576183e-05, + "loss": 0.9116, + "step": 33511 + }, + { + "epoch": 2.16, + "learning_rate": 1.2830245389553233e-05, + "loss": 0.9356, + "step": 33512 + }, + { + "epoch": 2.16, + "learning_rate": 1.2830030700530282e-05, + "loss": 0.8646, + "step": 33513 + }, + { + "epoch": 2.16, + "learning_rate": 1.2829816011507333e-05, + "loss": 0.9423, + "step": 33514 + }, + { + "epoch": 2.16, + "learning_rate": 1.2829601322484382e-05, + "loss": 0.8814, + "step": 33515 + }, + { + "epoch": 2.16, + "learning_rate": 1.2829386633461431e-05, + "loss": 0.974, + "step": 33516 + }, + { + "epoch": 2.16, + "learning_rate": 1.2829171944438482e-05, + "loss": 0.8773, + "step": 33517 + }, + { + "epoch": 2.16, + "learning_rate": 1.282895725541553e-05, + "loss": 0.8725, + "step": 33518 + }, + { + "epoch": 2.16, + "learning_rate": 1.2828742566392583e-05, + "loss": 0.9041, + "step": 33519 + }, + { + "epoch": 2.16, + "learning_rate": 1.2828527877369632e-05, + "loss": 0.8764, + "step": 33520 + }, + { + "epoch": 2.16, + "learning_rate": 1.2828313188346681e-05, + "loss": 0.9072, + "step": 33521 + }, + { + "epoch": 2.16, + "learning_rate": 1.2828098499323732e-05, + "loss": 0.8581, + "step": 33522 + }, + { + "epoch": 2.16, + "learning_rate": 1.282788381030078e-05, + "loss": 0.8581, + "step": 33523 + }, + { + "epoch": 2.16, + "learning_rate": 1.2827669121277831e-05, + "loss": 0.8979, + "step": 33524 + }, + { + "epoch": 2.16, + "learning_rate": 1.282745443225488e-05, + "loss": 0.9499, + "step": 33525 + }, + { + "epoch": 2.16, + "learning_rate": 1.2827239743231929e-05, + "loss": 0.9263, + "step": 33526 + }, + { + "epoch": 2.16, + "learning_rate": 1.282702505420898e-05, + "loss": 0.8689, + "step": 33527 + }, + { + "epoch": 2.16, + "learning_rate": 1.2826810365186029e-05, + "loss": 0.8778, + "step": 33528 + }, + { + "epoch": 2.16, + "learning_rate": 1.2826595676163078e-05, + "loss": 0.9064, + "step": 33529 + }, + { + "epoch": 2.16, + "learning_rate": 1.2826380987140128e-05, + "loss": 0.8811, + "step": 33530 + }, + { + "epoch": 2.16, + "learning_rate": 1.2826166298117177e-05, + "loss": 0.9, + "step": 33531 + }, + { + "epoch": 2.16, + "learning_rate": 1.2825951609094226e-05, + "loss": 0.8601, + "step": 33532 + }, + { + "epoch": 2.16, + "learning_rate": 1.2825736920071279e-05, + "loss": 0.9038, + "step": 33533 + }, + { + "epoch": 2.16, + "learning_rate": 1.282552223104833e-05, + "loss": 0.9029, + "step": 33534 + }, + { + "epoch": 2.16, + "learning_rate": 1.2825307542025378e-05, + "loss": 0.9898, + "step": 33535 + }, + { + "epoch": 2.16, + "learning_rate": 1.2825092853002427e-05, + "loss": 0.9137, + "step": 33536 + }, + { + "epoch": 2.16, + "learning_rate": 1.2824878163979478e-05, + "loss": 0.806, + "step": 33537 + }, + { + "epoch": 2.16, + "learning_rate": 1.2824663474956527e-05, + "loss": 0.8679, + "step": 33538 + }, + { + "epoch": 2.16, + "learning_rate": 1.2824448785933576e-05, + "loss": 0.8463, + "step": 33539 + }, + { + "epoch": 2.16, + "learning_rate": 1.2824234096910626e-05, + "loss": 0.8204, + "step": 33540 + }, + { + "epoch": 2.16, + "learning_rate": 1.2824019407887675e-05, + "loss": 0.9355, + "step": 33541 + }, + { + "epoch": 2.16, + "learning_rate": 1.2823804718864724e-05, + "loss": 0.8962, + "step": 33542 + }, + { + "epoch": 2.16, + "learning_rate": 1.2823590029841775e-05, + "loss": 0.9258, + "step": 33543 + }, + { + "epoch": 2.16, + "learning_rate": 1.2823375340818824e-05, + "loss": 0.8602, + "step": 33544 + }, + { + "epoch": 2.16, + "learning_rate": 1.2823160651795875e-05, + "loss": 0.8852, + "step": 33545 + }, + { + "epoch": 2.16, + "learning_rate": 1.2822945962772924e-05, + "loss": 0.9451, + "step": 33546 + }, + { + "epoch": 2.16, + "learning_rate": 1.2822731273749973e-05, + "loss": 0.9613, + "step": 33547 + }, + { + "epoch": 2.16, + "learning_rate": 1.2822516584727025e-05, + "loss": 0.8854, + "step": 33548 + }, + { + "epoch": 2.16, + "learning_rate": 1.2822301895704074e-05, + "loss": 0.9091, + "step": 33549 + }, + { + "epoch": 2.16, + "learning_rate": 1.2822087206681125e-05, + "loss": 0.9558, + "step": 33550 + }, + { + "epoch": 2.16, + "learning_rate": 1.2821872517658174e-05, + "loss": 0.8456, + "step": 33551 + }, + { + "epoch": 2.16, + "learning_rate": 1.2821657828635223e-05, + "loss": 0.908, + "step": 33552 + }, + { + "epoch": 2.16, + "learning_rate": 1.2821443139612273e-05, + "loss": 0.8558, + "step": 33553 + }, + { + "epoch": 2.16, + "learning_rate": 1.2821228450589322e-05, + "loss": 0.9164, + "step": 33554 + }, + { + "epoch": 2.16, + "learning_rate": 1.2821013761566373e-05, + "loss": 0.8675, + "step": 33555 + }, + { + "epoch": 2.16, + "learning_rate": 1.2820799072543422e-05, + "loss": 0.9376, + "step": 33556 + }, + { + "epoch": 2.16, + "learning_rate": 1.282058438352047e-05, + "loss": 0.9776, + "step": 33557 + }, + { + "epoch": 2.16, + "learning_rate": 1.2820369694497521e-05, + "loss": 0.8565, + "step": 33558 + }, + { + "epoch": 2.16, + "learning_rate": 1.282015500547457e-05, + "loss": 0.8895, + "step": 33559 + }, + { + "epoch": 2.16, + "learning_rate": 1.281994031645162e-05, + "loss": 0.8841, + "step": 33560 + }, + { + "epoch": 2.16, + "learning_rate": 1.281972562742867e-05, + "loss": 0.8457, + "step": 33561 + }, + { + "epoch": 2.16, + "learning_rate": 1.281951093840572e-05, + "loss": 0.9328, + "step": 33562 + }, + { + "epoch": 2.16, + "learning_rate": 1.2819296249382771e-05, + "loss": 0.8452, + "step": 33563 + }, + { + "epoch": 2.16, + "learning_rate": 1.281908156035982e-05, + "loss": 0.898, + "step": 33564 + }, + { + "epoch": 2.16, + "learning_rate": 1.2818866871336871e-05, + "loss": 0.9075, + "step": 33565 + }, + { + "epoch": 2.16, + "learning_rate": 1.281865218231392e-05, + "loss": 0.8992, + "step": 33566 + }, + { + "epoch": 2.16, + "learning_rate": 1.2818437493290969e-05, + "loss": 0.8855, + "step": 33567 + }, + { + "epoch": 2.16, + "learning_rate": 1.281822280426802e-05, + "loss": 0.8476, + "step": 33568 + }, + { + "epoch": 2.16, + "learning_rate": 1.2818008115245068e-05, + "loss": 0.9061, + "step": 33569 + }, + { + "epoch": 2.16, + "learning_rate": 1.2817793426222117e-05, + "loss": 0.8673, + "step": 33570 + }, + { + "epoch": 2.16, + "learning_rate": 1.2817578737199168e-05, + "loss": 0.8821, + "step": 33571 + }, + { + "epoch": 2.16, + "learning_rate": 1.2817364048176217e-05, + "loss": 0.9219, + "step": 33572 + }, + { + "epoch": 2.16, + "learning_rate": 1.2817149359153268e-05, + "loss": 0.9139, + "step": 33573 + }, + { + "epoch": 2.16, + "learning_rate": 1.2816934670130317e-05, + "loss": 0.9077, + "step": 33574 + }, + { + "epoch": 2.16, + "learning_rate": 1.2816719981107366e-05, + "loss": 0.8674, + "step": 33575 + }, + { + "epoch": 2.16, + "learning_rate": 1.2816505292084416e-05, + "loss": 0.9962, + "step": 33576 + }, + { + "epoch": 2.16, + "learning_rate": 1.2816290603061467e-05, + "loss": 0.9662, + "step": 33577 + }, + { + "epoch": 2.16, + "learning_rate": 1.2816075914038518e-05, + "loss": 0.9677, + "step": 33578 + }, + { + "epoch": 2.16, + "learning_rate": 1.2815861225015567e-05, + "loss": 0.9046, + "step": 33579 + }, + { + "epoch": 2.16, + "learning_rate": 1.2815646535992616e-05, + "loss": 0.8191, + "step": 33580 + }, + { + "epoch": 2.16, + "learning_rate": 1.2815431846969666e-05, + "loss": 0.8395, + "step": 33581 + }, + { + "epoch": 2.16, + "learning_rate": 1.2815217157946715e-05, + "loss": 0.8898, + "step": 33582 + }, + { + "epoch": 2.16, + "learning_rate": 1.2815002468923766e-05, + "loss": 0.9362, + "step": 33583 + }, + { + "epoch": 2.16, + "learning_rate": 1.2814787779900815e-05, + "loss": 0.9459, + "step": 33584 + }, + { + "epoch": 2.16, + "learning_rate": 1.2814573090877864e-05, + "loss": 0.8712, + "step": 33585 + }, + { + "epoch": 2.16, + "learning_rate": 1.2814358401854914e-05, + "loss": 0.8859, + "step": 33586 + }, + { + "epoch": 2.16, + "learning_rate": 1.2814143712831963e-05, + "loss": 0.8436, + "step": 33587 + }, + { + "epoch": 2.16, + "learning_rate": 1.2813929023809012e-05, + "loss": 0.9032, + "step": 33588 + }, + { + "epoch": 2.16, + "learning_rate": 1.2813714334786063e-05, + "loss": 0.8452, + "step": 33589 + }, + { + "epoch": 2.16, + "learning_rate": 1.2813499645763112e-05, + "loss": 0.8701, + "step": 33590 + }, + { + "epoch": 2.16, + "learning_rate": 1.2813284956740163e-05, + "loss": 0.8939, + "step": 33591 + }, + { + "epoch": 2.16, + "learning_rate": 1.2813070267717213e-05, + "loss": 0.923, + "step": 33592 + }, + { + "epoch": 2.16, + "learning_rate": 1.2812855578694264e-05, + "loss": 0.8152, + "step": 33593 + }, + { + "epoch": 2.16, + "learning_rate": 1.2812640889671313e-05, + "loss": 0.9018, + "step": 33594 + }, + { + "epoch": 2.16, + "learning_rate": 1.2812426200648362e-05, + "loss": 0.8949, + "step": 33595 + }, + { + "epoch": 2.16, + "learning_rate": 1.2812211511625413e-05, + "loss": 0.8482, + "step": 33596 + }, + { + "epoch": 2.16, + "learning_rate": 1.2811996822602462e-05, + "loss": 0.9264, + "step": 33597 + }, + { + "epoch": 2.16, + "learning_rate": 1.281178213357951e-05, + "loss": 0.9307, + "step": 33598 + }, + { + "epoch": 2.16, + "learning_rate": 1.2811567444556561e-05, + "loss": 0.8411, + "step": 33599 + }, + { + "epoch": 2.16, + "learning_rate": 1.281135275553361e-05, + "loss": 0.9067, + "step": 33600 + }, + { + "epoch": 2.16, + "learning_rate": 1.281113806651066e-05, + "loss": 0.8878, + "step": 33601 + }, + { + "epoch": 2.16, + "learning_rate": 1.281092337748771e-05, + "loss": 0.9696, + "step": 33602 + }, + { + "epoch": 2.16, + "learning_rate": 1.2810708688464759e-05, + "loss": 0.8702, + "step": 33603 + }, + { + "epoch": 2.16, + "learning_rate": 1.281049399944181e-05, + "loss": 0.9414, + "step": 33604 + }, + { + "epoch": 2.16, + "learning_rate": 1.2810279310418858e-05, + "loss": 0.8878, + "step": 33605 + }, + { + "epoch": 2.16, + "learning_rate": 1.281006462139591e-05, + "loss": 0.9032, + "step": 33606 + }, + { + "epoch": 2.16, + "learning_rate": 1.280984993237296e-05, + "loss": 0.9169, + "step": 33607 + }, + { + "epoch": 2.16, + "learning_rate": 1.2809635243350009e-05, + "loss": 0.9116, + "step": 33608 + }, + { + "epoch": 2.16, + "learning_rate": 1.280942055432706e-05, + "loss": 0.9032, + "step": 33609 + }, + { + "epoch": 2.16, + "learning_rate": 1.2809205865304108e-05, + "loss": 0.8235, + "step": 33610 + }, + { + "epoch": 2.16, + "learning_rate": 1.2808991176281159e-05, + "loss": 0.9165, + "step": 33611 + }, + { + "epoch": 2.16, + "learning_rate": 1.2808776487258208e-05, + "loss": 0.8921, + "step": 33612 + }, + { + "epoch": 2.16, + "learning_rate": 1.2808561798235257e-05, + "loss": 0.9392, + "step": 33613 + }, + { + "epoch": 2.16, + "learning_rate": 1.2808347109212307e-05, + "loss": 0.9217, + "step": 33614 + }, + { + "epoch": 2.16, + "learning_rate": 1.2808132420189356e-05, + "loss": 0.9012, + "step": 33615 + }, + { + "epoch": 2.16, + "learning_rate": 1.2807917731166405e-05, + "loss": 0.8973, + "step": 33616 + }, + { + "epoch": 2.16, + "learning_rate": 1.2807703042143456e-05, + "loss": 0.8859, + "step": 33617 + }, + { + "epoch": 2.16, + "learning_rate": 1.2807488353120505e-05, + "loss": 0.8701, + "step": 33618 + }, + { + "epoch": 2.16, + "learning_rate": 1.2807273664097554e-05, + "loss": 0.8581, + "step": 33619 + }, + { + "epoch": 2.16, + "learning_rate": 1.2807058975074605e-05, + "loss": 0.9127, + "step": 33620 + }, + { + "epoch": 2.16, + "learning_rate": 1.2806844286051657e-05, + "loss": 0.9377, + "step": 33621 + }, + { + "epoch": 2.16, + "learning_rate": 1.2806629597028706e-05, + "loss": 0.864, + "step": 33622 + }, + { + "epoch": 2.16, + "learning_rate": 1.2806414908005755e-05, + "loss": 0.8698, + "step": 33623 + }, + { + "epoch": 2.16, + "learning_rate": 1.2806200218982806e-05, + "loss": 0.8863, + "step": 33624 + }, + { + "epoch": 2.16, + "learning_rate": 1.2805985529959855e-05, + "loss": 0.8746, + "step": 33625 + }, + { + "epoch": 2.16, + "learning_rate": 1.2805770840936904e-05, + "loss": 0.8952, + "step": 33626 + }, + { + "epoch": 2.16, + "learning_rate": 1.2805556151913954e-05, + "loss": 0.9073, + "step": 33627 + }, + { + "epoch": 2.16, + "learning_rate": 1.2805341462891003e-05, + "loss": 0.8521, + "step": 33628 + }, + { + "epoch": 2.16, + "learning_rate": 1.2805126773868052e-05, + "loss": 0.8416, + "step": 33629 + }, + { + "epoch": 2.16, + "learning_rate": 1.2804912084845103e-05, + "loss": 0.9239, + "step": 33630 + }, + { + "epoch": 2.16, + "learning_rate": 1.2804697395822152e-05, + "loss": 0.8662, + "step": 33631 + }, + { + "epoch": 2.16, + "learning_rate": 1.2804482706799202e-05, + "loss": 0.8416, + "step": 33632 + }, + { + "epoch": 2.16, + "learning_rate": 1.2804268017776251e-05, + "loss": 0.8389, + "step": 33633 + }, + { + "epoch": 2.16, + "learning_rate": 1.28040533287533e-05, + "loss": 0.8581, + "step": 33634 + }, + { + "epoch": 2.16, + "learning_rate": 1.2803838639730353e-05, + "loss": 0.8996, + "step": 33635 + }, + { + "epoch": 2.16, + "learning_rate": 1.2803623950707402e-05, + "loss": 0.8884, + "step": 33636 + }, + { + "epoch": 2.16, + "learning_rate": 1.2803409261684452e-05, + "loss": 0.8325, + "step": 33637 + }, + { + "epoch": 2.16, + "learning_rate": 1.2803194572661501e-05, + "loss": 0.9168, + "step": 33638 + }, + { + "epoch": 2.16, + "learning_rate": 1.280297988363855e-05, + "loss": 0.9482, + "step": 33639 + }, + { + "epoch": 2.16, + "learning_rate": 1.2802765194615601e-05, + "loss": 0.8666, + "step": 33640 + }, + { + "epoch": 2.16, + "learning_rate": 1.280255050559265e-05, + "loss": 0.9026, + "step": 33641 + }, + { + "epoch": 2.16, + "learning_rate": 1.28023358165697e-05, + "loss": 0.8961, + "step": 33642 + }, + { + "epoch": 2.16, + "learning_rate": 1.280212112754675e-05, + "loss": 0.9216, + "step": 33643 + }, + { + "epoch": 2.16, + "learning_rate": 1.2801906438523798e-05, + "loss": 0.9015, + "step": 33644 + }, + { + "epoch": 2.16, + "learning_rate": 1.2801691749500849e-05, + "loss": 0.8552, + "step": 33645 + }, + { + "epoch": 2.16, + "learning_rate": 1.2801477060477898e-05, + "loss": 0.8753, + "step": 33646 + }, + { + "epoch": 2.16, + "learning_rate": 1.2801262371454947e-05, + "loss": 0.8975, + "step": 33647 + }, + { + "epoch": 2.16, + "learning_rate": 1.2801047682431998e-05, + "loss": 0.8628, + "step": 33648 + }, + { + "epoch": 2.16, + "learning_rate": 1.2800832993409047e-05, + "loss": 0.9116, + "step": 33649 + }, + { + "epoch": 2.16, + "learning_rate": 1.2800618304386099e-05, + "loss": 0.9283, + "step": 33650 + }, + { + "epoch": 2.17, + "learning_rate": 1.2800403615363148e-05, + "loss": 0.9102, + "step": 33651 + }, + { + "epoch": 2.17, + "learning_rate": 1.2800188926340199e-05, + "loss": 0.8487, + "step": 33652 + }, + { + "epoch": 2.17, + "learning_rate": 1.2799974237317248e-05, + "loss": 0.877, + "step": 33653 + }, + { + "epoch": 2.17, + "learning_rate": 1.2799759548294297e-05, + "loss": 0.947, + "step": 33654 + }, + { + "epoch": 2.17, + "learning_rate": 1.2799544859271347e-05, + "loss": 0.9137, + "step": 33655 + }, + { + "epoch": 2.17, + "learning_rate": 1.2799330170248396e-05, + "loss": 0.8806, + "step": 33656 + }, + { + "epoch": 2.17, + "learning_rate": 1.2799115481225445e-05, + "loss": 0.9112, + "step": 33657 + }, + { + "epoch": 2.17, + "learning_rate": 1.2798900792202496e-05, + "loss": 0.9614, + "step": 33658 + }, + { + "epoch": 2.17, + "learning_rate": 1.2798686103179545e-05, + "loss": 0.9055, + "step": 33659 + }, + { + "epoch": 2.17, + "learning_rate": 1.2798471414156595e-05, + "loss": 0.9344, + "step": 33660 + }, + { + "epoch": 2.17, + "learning_rate": 1.2798256725133644e-05, + "loss": 0.8713, + "step": 33661 + }, + { + "epoch": 2.17, + "learning_rate": 1.2798042036110693e-05, + "loss": 0.9223, + "step": 33662 + }, + { + "epoch": 2.17, + "learning_rate": 1.2797827347087744e-05, + "loss": 0.8568, + "step": 33663 + }, + { + "epoch": 2.17, + "learning_rate": 1.2797612658064795e-05, + "loss": 0.8304, + "step": 33664 + }, + { + "epoch": 2.17, + "learning_rate": 1.2797397969041845e-05, + "loss": 0.8727, + "step": 33665 + }, + { + "epoch": 2.17, + "learning_rate": 1.2797183280018894e-05, + "loss": 0.8713, + "step": 33666 + }, + { + "epoch": 2.17, + "learning_rate": 1.2796968590995943e-05, + "loss": 0.8769, + "step": 33667 + }, + { + "epoch": 2.17, + "learning_rate": 1.2796753901972994e-05, + "loss": 0.8537, + "step": 33668 + }, + { + "epoch": 2.17, + "learning_rate": 1.2796539212950043e-05, + "loss": 0.9113, + "step": 33669 + }, + { + "epoch": 2.17, + "learning_rate": 1.2796324523927094e-05, + "loss": 0.8718, + "step": 33670 + }, + { + "epoch": 2.17, + "learning_rate": 1.2796109834904143e-05, + "loss": 0.8349, + "step": 33671 + }, + { + "epoch": 2.17, + "learning_rate": 1.2795895145881191e-05, + "loss": 0.9066, + "step": 33672 + }, + { + "epoch": 2.17, + "learning_rate": 1.2795680456858242e-05, + "loss": 0.9309, + "step": 33673 + }, + { + "epoch": 2.17, + "learning_rate": 1.2795465767835291e-05, + "loss": 0.9021, + "step": 33674 + }, + { + "epoch": 2.17, + "learning_rate": 1.279525107881234e-05, + "loss": 0.9115, + "step": 33675 + }, + { + "epoch": 2.17, + "learning_rate": 1.279503638978939e-05, + "loss": 0.9398, + "step": 33676 + }, + { + "epoch": 2.17, + "learning_rate": 1.279482170076644e-05, + "loss": 0.8575, + "step": 33677 + }, + { + "epoch": 2.17, + "learning_rate": 1.279460701174349e-05, + "loss": 0.8742, + "step": 33678 + }, + { + "epoch": 2.17, + "learning_rate": 1.2794392322720541e-05, + "loss": 0.9176, + "step": 33679 + }, + { + "epoch": 2.17, + "learning_rate": 1.2794177633697592e-05, + "loss": 0.9054, + "step": 33680 + }, + { + "epoch": 2.17, + "learning_rate": 1.279396294467464e-05, + "loss": 0.9111, + "step": 33681 + }, + { + "epoch": 2.17, + "learning_rate": 1.279374825565169e-05, + "loss": 0.9043, + "step": 33682 + }, + { + "epoch": 2.17, + "learning_rate": 1.279353356662874e-05, + "loss": 0.8411, + "step": 33683 + }, + { + "epoch": 2.17, + "learning_rate": 1.279331887760579e-05, + "loss": 0.9154, + "step": 33684 + }, + { + "epoch": 2.17, + "learning_rate": 1.2793104188582838e-05, + "loss": 0.9237, + "step": 33685 + }, + { + "epoch": 2.17, + "learning_rate": 1.2792889499559889e-05, + "loss": 0.8932, + "step": 33686 + }, + { + "epoch": 2.17, + "learning_rate": 1.2792674810536938e-05, + "loss": 0.9088, + "step": 33687 + }, + { + "epoch": 2.17, + "learning_rate": 1.2792460121513988e-05, + "loss": 0.9587, + "step": 33688 + }, + { + "epoch": 2.17, + "learning_rate": 1.2792245432491037e-05, + "loss": 0.8194, + "step": 33689 + }, + { + "epoch": 2.17, + "learning_rate": 1.2792030743468086e-05, + "loss": 0.9963, + "step": 33690 + }, + { + "epoch": 2.17, + "learning_rate": 1.2791816054445137e-05, + "loss": 0.8527, + "step": 33691 + }, + { + "epoch": 2.17, + "learning_rate": 1.2791601365422186e-05, + "loss": 0.9822, + "step": 33692 + }, + { + "epoch": 2.17, + "learning_rate": 1.2791386676399238e-05, + "loss": 0.8304, + "step": 33693 + }, + { + "epoch": 2.17, + "learning_rate": 1.2791171987376287e-05, + "loss": 0.8683, + "step": 33694 + }, + { + "epoch": 2.17, + "learning_rate": 1.2790957298353336e-05, + "loss": 0.8653, + "step": 33695 + }, + { + "epoch": 2.17, + "learning_rate": 1.2790742609330387e-05, + "loss": 0.8789, + "step": 33696 + }, + { + "epoch": 2.17, + "learning_rate": 1.2790527920307436e-05, + "loss": 0.8431, + "step": 33697 + }, + { + "epoch": 2.17, + "learning_rate": 1.2790313231284487e-05, + "loss": 0.8039, + "step": 33698 + }, + { + "epoch": 2.17, + "learning_rate": 1.2790098542261536e-05, + "loss": 0.9151, + "step": 33699 + }, + { + "epoch": 2.17, + "learning_rate": 1.2789883853238585e-05, + "loss": 0.9187, + "step": 33700 + }, + { + "epoch": 2.17, + "learning_rate": 1.2789669164215635e-05, + "loss": 0.814, + "step": 33701 + }, + { + "epoch": 2.17, + "learning_rate": 1.2789454475192684e-05, + "loss": 0.9443, + "step": 33702 + }, + { + "epoch": 2.17, + "learning_rate": 1.2789239786169733e-05, + "loss": 0.8725, + "step": 33703 + }, + { + "epoch": 2.17, + "learning_rate": 1.2789025097146784e-05, + "loss": 0.8425, + "step": 33704 + }, + { + "epoch": 2.17, + "learning_rate": 1.2788810408123833e-05, + "loss": 0.8876, + "step": 33705 + }, + { + "epoch": 2.17, + "learning_rate": 1.2788595719100882e-05, + "loss": 0.9527, + "step": 33706 + }, + { + "epoch": 2.17, + "learning_rate": 1.2788381030077932e-05, + "loss": 0.9063, + "step": 33707 + }, + { + "epoch": 2.17, + "learning_rate": 1.2788166341054985e-05, + "loss": 0.8455, + "step": 33708 + }, + { + "epoch": 2.17, + "learning_rate": 1.2787951652032034e-05, + "loss": 0.9375, + "step": 33709 + }, + { + "epoch": 2.17, + "learning_rate": 1.2787736963009083e-05, + "loss": 0.962, + "step": 33710 + }, + { + "epoch": 2.17, + "learning_rate": 1.2787522273986133e-05, + "loss": 0.8959, + "step": 33711 + }, + { + "epoch": 2.17, + "learning_rate": 1.2787307584963182e-05, + "loss": 0.9462, + "step": 33712 + }, + { + "epoch": 2.17, + "learning_rate": 1.2787092895940231e-05, + "loss": 0.9343, + "step": 33713 + }, + { + "epoch": 2.17, + "learning_rate": 1.2786878206917282e-05, + "loss": 0.8756, + "step": 33714 + }, + { + "epoch": 2.17, + "learning_rate": 1.2786663517894331e-05, + "loss": 0.8413, + "step": 33715 + }, + { + "epoch": 2.17, + "learning_rate": 1.278644882887138e-05, + "loss": 0.8947, + "step": 33716 + }, + { + "epoch": 2.17, + "learning_rate": 1.278623413984843e-05, + "loss": 0.9274, + "step": 33717 + }, + { + "epoch": 2.17, + "learning_rate": 1.278601945082548e-05, + "loss": 0.9887, + "step": 33718 + }, + { + "epoch": 2.17, + "learning_rate": 1.278580476180253e-05, + "loss": 0.9753, + "step": 33719 + }, + { + "epoch": 2.17, + "learning_rate": 1.2785590072779579e-05, + "loss": 0.8885, + "step": 33720 + }, + { + "epoch": 2.17, + "learning_rate": 1.2785375383756628e-05, + "loss": 0.9454, + "step": 33721 + }, + { + "epoch": 2.17, + "learning_rate": 1.278516069473368e-05, + "loss": 0.9204, + "step": 33722 + }, + { + "epoch": 2.17, + "learning_rate": 1.278494600571073e-05, + "loss": 0.8737, + "step": 33723 + }, + { + "epoch": 2.17, + "learning_rate": 1.278473131668778e-05, + "loss": 0.9145, + "step": 33724 + }, + { + "epoch": 2.17, + "learning_rate": 1.2784516627664829e-05, + "loss": 0.8594, + "step": 33725 + }, + { + "epoch": 2.17, + "learning_rate": 1.2784301938641878e-05, + "loss": 0.8416, + "step": 33726 + }, + { + "epoch": 2.17, + "learning_rate": 1.2784087249618929e-05, + "loss": 0.9477, + "step": 33727 + }, + { + "epoch": 2.17, + "learning_rate": 1.2783872560595978e-05, + "loss": 0.8884, + "step": 33728 + }, + { + "epoch": 2.17, + "learning_rate": 1.2783657871573028e-05, + "loss": 0.8685, + "step": 33729 + }, + { + "epoch": 2.17, + "learning_rate": 1.2783443182550077e-05, + "loss": 0.9093, + "step": 33730 + }, + { + "epoch": 2.17, + "learning_rate": 1.2783228493527126e-05, + "loss": 0.913, + "step": 33731 + }, + { + "epoch": 2.17, + "learning_rate": 1.2783013804504177e-05, + "loss": 0.9253, + "step": 33732 + }, + { + "epoch": 2.17, + "learning_rate": 1.2782799115481226e-05, + "loss": 0.8519, + "step": 33733 + }, + { + "epoch": 2.17, + "learning_rate": 1.2782584426458275e-05, + "loss": 0.8519, + "step": 33734 + }, + { + "epoch": 2.17, + "learning_rate": 1.2782369737435325e-05, + "loss": 0.9085, + "step": 33735 + }, + { + "epoch": 2.17, + "learning_rate": 1.2782155048412374e-05, + "loss": 0.8829, + "step": 33736 + }, + { + "epoch": 2.17, + "learning_rate": 1.2781940359389427e-05, + "loss": 0.9017, + "step": 33737 + }, + { + "epoch": 2.17, + "learning_rate": 1.2781725670366476e-05, + "loss": 0.8707, + "step": 33738 + }, + { + "epoch": 2.17, + "learning_rate": 1.2781510981343526e-05, + "loss": 0.9286, + "step": 33739 + }, + { + "epoch": 2.17, + "learning_rate": 1.2781296292320575e-05, + "loss": 0.8376, + "step": 33740 + }, + { + "epoch": 2.17, + "learning_rate": 1.2781081603297624e-05, + "loss": 0.8924, + "step": 33741 + }, + { + "epoch": 2.17, + "learning_rate": 1.2780866914274675e-05, + "loss": 0.885, + "step": 33742 + }, + { + "epoch": 2.17, + "learning_rate": 1.2780652225251724e-05, + "loss": 0.89, + "step": 33743 + }, + { + "epoch": 2.17, + "learning_rate": 1.2780437536228773e-05, + "loss": 0.9644, + "step": 33744 + }, + { + "epoch": 2.17, + "learning_rate": 1.2780222847205824e-05, + "loss": 0.8909, + "step": 33745 + }, + { + "epoch": 2.17, + "learning_rate": 1.2780008158182872e-05, + "loss": 0.9206, + "step": 33746 + }, + { + "epoch": 2.17, + "learning_rate": 1.2779793469159923e-05, + "loss": 0.8478, + "step": 33747 + }, + { + "epoch": 2.17, + "learning_rate": 1.2779578780136972e-05, + "loss": 0.9234, + "step": 33748 + }, + { + "epoch": 2.17, + "learning_rate": 1.2779364091114021e-05, + "loss": 0.9323, + "step": 33749 + }, + { + "epoch": 2.17, + "learning_rate": 1.2779149402091072e-05, + "loss": 0.8641, + "step": 33750 + }, + { + "epoch": 2.17, + "learning_rate": 1.2778934713068122e-05, + "loss": 0.8219, + "step": 33751 + }, + { + "epoch": 2.17, + "learning_rate": 1.2778720024045173e-05, + "loss": 0.9505, + "step": 33752 + }, + { + "epoch": 2.17, + "learning_rate": 1.2778505335022222e-05, + "loss": 0.9095, + "step": 33753 + }, + { + "epoch": 2.17, + "learning_rate": 1.2778290645999271e-05, + "loss": 0.9162, + "step": 33754 + }, + { + "epoch": 2.17, + "learning_rate": 1.2778075956976322e-05, + "loss": 0.9012, + "step": 33755 + }, + { + "epoch": 2.17, + "learning_rate": 1.277786126795337e-05, + "loss": 0.9071, + "step": 33756 + }, + { + "epoch": 2.17, + "learning_rate": 1.2777646578930421e-05, + "loss": 0.9231, + "step": 33757 + }, + { + "epoch": 2.17, + "learning_rate": 1.277743188990747e-05, + "loss": 1.0083, + "step": 33758 + }, + { + "epoch": 2.17, + "learning_rate": 1.277721720088452e-05, + "loss": 0.901, + "step": 33759 + }, + { + "epoch": 2.17, + "learning_rate": 1.277700251186157e-05, + "loss": 0.9037, + "step": 33760 + }, + { + "epoch": 2.17, + "learning_rate": 1.2776787822838619e-05, + "loss": 0.9131, + "step": 33761 + }, + { + "epoch": 2.17, + "learning_rate": 1.2776573133815668e-05, + "loss": 0.8879, + "step": 33762 + }, + { + "epoch": 2.17, + "learning_rate": 1.2776358444792718e-05, + "loss": 0.8866, + "step": 33763 + }, + { + "epoch": 2.17, + "learning_rate": 1.2776143755769767e-05, + "loss": 0.9405, + "step": 33764 + }, + { + "epoch": 2.17, + "learning_rate": 1.2775929066746816e-05, + "loss": 0.9408, + "step": 33765 + }, + { + "epoch": 2.17, + "learning_rate": 1.2775714377723869e-05, + "loss": 0.8435, + "step": 33766 + }, + { + "epoch": 2.17, + "learning_rate": 1.277549968870092e-05, + "loss": 0.8825, + "step": 33767 + }, + { + "epoch": 2.17, + "learning_rate": 1.2775284999677968e-05, + "loss": 0.8812, + "step": 33768 + }, + { + "epoch": 2.17, + "learning_rate": 1.2775070310655017e-05, + "loss": 0.9366, + "step": 33769 + }, + { + "epoch": 2.17, + "learning_rate": 1.2774855621632068e-05, + "loss": 0.8591, + "step": 33770 + }, + { + "epoch": 2.17, + "learning_rate": 1.2774640932609117e-05, + "loss": 0.8859, + "step": 33771 + }, + { + "epoch": 2.17, + "learning_rate": 1.2774426243586166e-05, + "loss": 0.9009, + "step": 33772 + }, + { + "epoch": 2.17, + "learning_rate": 1.2774211554563217e-05, + "loss": 0.9362, + "step": 33773 + }, + { + "epoch": 2.17, + "learning_rate": 1.2773996865540266e-05, + "loss": 0.9233, + "step": 33774 + }, + { + "epoch": 2.17, + "learning_rate": 1.2773782176517314e-05, + "loss": 0.9569, + "step": 33775 + }, + { + "epoch": 2.17, + "learning_rate": 1.2773567487494365e-05, + "loss": 0.9324, + "step": 33776 + }, + { + "epoch": 2.17, + "learning_rate": 1.2773352798471414e-05, + "loss": 0.8667, + "step": 33777 + }, + { + "epoch": 2.17, + "learning_rate": 1.2773138109448465e-05, + "loss": 0.9179, + "step": 33778 + }, + { + "epoch": 2.17, + "learning_rate": 1.2772923420425514e-05, + "loss": 0.8944, + "step": 33779 + }, + { + "epoch": 2.17, + "learning_rate": 1.2772708731402566e-05, + "loss": 0.9077, + "step": 33780 + }, + { + "epoch": 2.17, + "learning_rate": 1.2772494042379615e-05, + "loss": 0.8751, + "step": 33781 + }, + { + "epoch": 2.17, + "learning_rate": 1.2772279353356664e-05, + "loss": 0.8506, + "step": 33782 + }, + { + "epoch": 2.17, + "learning_rate": 1.2772064664333715e-05, + "loss": 0.8545, + "step": 33783 + }, + { + "epoch": 2.17, + "learning_rate": 1.2771849975310764e-05, + "loss": 0.9053, + "step": 33784 + }, + { + "epoch": 2.17, + "learning_rate": 1.2771635286287813e-05, + "loss": 0.8397, + "step": 33785 + }, + { + "epoch": 2.17, + "learning_rate": 1.2771420597264863e-05, + "loss": 0.8942, + "step": 33786 + }, + { + "epoch": 2.17, + "learning_rate": 1.2771205908241912e-05, + "loss": 0.8785, + "step": 33787 + }, + { + "epoch": 2.17, + "learning_rate": 1.2770991219218963e-05, + "loss": 0.8866, + "step": 33788 + }, + { + "epoch": 2.17, + "learning_rate": 1.2770776530196012e-05, + "loss": 0.8797, + "step": 33789 + }, + { + "epoch": 2.17, + "learning_rate": 1.277056184117306e-05, + "loss": 0.9166, + "step": 33790 + }, + { + "epoch": 2.17, + "learning_rate": 1.2770347152150111e-05, + "loss": 0.886, + "step": 33791 + }, + { + "epoch": 2.17, + "learning_rate": 1.277013246312716e-05, + "loss": 0.858, + "step": 33792 + }, + { + "epoch": 2.17, + "learning_rate": 1.276991777410421e-05, + "loss": 0.8553, + "step": 33793 + }, + { + "epoch": 2.17, + "learning_rate": 1.276970308508126e-05, + "loss": 0.8878, + "step": 33794 + }, + { + "epoch": 2.17, + "learning_rate": 1.2769488396058312e-05, + "loss": 0.8929, + "step": 33795 + }, + { + "epoch": 2.17, + "learning_rate": 1.2769273707035361e-05, + "loss": 0.8311, + "step": 33796 + }, + { + "epoch": 2.17, + "learning_rate": 1.276905901801241e-05, + "loss": 0.8961, + "step": 33797 + }, + { + "epoch": 2.17, + "learning_rate": 1.2768844328989461e-05, + "loss": 0.8938, + "step": 33798 + }, + { + "epoch": 2.17, + "learning_rate": 1.276862963996651e-05, + "loss": 0.8262, + "step": 33799 + }, + { + "epoch": 2.17, + "learning_rate": 1.2768414950943559e-05, + "loss": 0.8461, + "step": 33800 + }, + { + "epoch": 2.17, + "learning_rate": 1.276820026192061e-05, + "loss": 0.9432, + "step": 33801 + }, + { + "epoch": 2.17, + "learning_rate": 1.2767985572897659e-05, + "loss": 0.8946, + "step": 33802 + }, + { + "epoch": 2.17, + "learning_rate": 1.2767770883874708e-05, + "loss": 0.9225, + "step": 33803 + }, + { + "epoch": 2.17, + "learning_rate": 1.2767556194851758e-05, + "loss": 0.8977, + "step": 33804 + }, + { + "epoch": 2.17, + "learning_rate": 1.2767341505828807e-05, + "loss": 0.8979, + "step": 33805 + }, + { + "epoch": 2.17, + "learning_rate": 1.2767126816805858e-05, + "loss": 0.9381, + "step": 33806 + }, + { + "epoch": 2.18, + "learning_rate": 1.2766912127782907e-05, + "loss": 0.8914, + "step": 33807 + }, + { + "epoch": 2.18, + "learning_rate": 1.2766697438759956e-05, + "loss": 0.8432, + "step": 33808 + }, + { + "epoch": 2.18, + "learning_rate": 1.2766482749737008e-05, + "loss": 0.8818, + "step": 33809 + }, + { + "epoch": 2.18, + "learning_rate": 1.2766268060714057e-05, + "loss": 0.8938, + "step": 33810 + }, + { + "epoch": 2.18, + "learning_rate": 1.2766053371691108e-05, + "loss": 0.9138, + "step": 33811 + }, + { + "epoch": 2.18, + "learning_rate": 1.2765838682668157e-05, + "loss": 0.895, + "step": 33812 + }, + { + "epoch": 2.18, + "learning_rate": 1.2765623993645206e-05, + "loss": 0.8702, + "step": 33813 + }, + { + "epoch": 2.18, + "learning_rate": 1.2765409304622256e-05, + "loss": 0.9254, + "step": 33814 + }, + { + "epoch": 2.18, + "learning_rate": 1.2765194615599305e-05, + "loss": 0.8975, + "step": 33815 + }, + { + "epoch": 2.18, + "learning_rate": 1.2764979926576356e-05, + "loss": 0.9423, + "step": 33816 + }, + { + "epoch": 2.18, + "learning_rate": 1.2764765237553405e-05, + "loss": 0.9108, + "step": 33817 + }, + { + "epoch": 2.18, + "learning_rate": 1.2764550548530454e-05, + "loss": 0.9227, + "step": 33818 + }, + { + "epoch": 2.18, + "learning_rate": 1.2764335859507505e-05, + "loss": 0.9107, + "step": 33819 + }, + { + "epoch": 2.18, + "learning_rate": 1.2764121170484553e-05, + "loss": 0.9422, + "step": 33820 + }, + { + "epoch": 2.18, + "learning_rate": 1.2763906481461602e-05, + "loss": 0.8634, + "step": 33821 + }, + { + "epoch": 2.18, + "learning_rate": 1.2763691792438653e-05, + "loss": 0.8994, + "step": 33822 + }, + { + "epoch": 2.18, + "learning_rate": 1.2763477103415702e-05, + "loss": 0.9645, + "step": 33823 + }, + { + "epoch": 2.18, + "learning_rate": 1.2763262414392754e-05, + "loss": 0.9131, + "step": 33824 + }, + { + "epoch": 2.18, + "learning_rate": 1.2763047725369803e-05, + "loss": 0.8975, + "step": 33825 + }, + { + "epoch": 2.18, + "learning_rate": 1.2762833036346854e-05, + "loss": 0.8758, + "step": 33826 + }, + { + "epoch": 2.18, + "learning_rate": 1.2762618347323903e-05, + "loss": 0.9029, + "step": 33827 + }, + { + "epoch": 2.18, + "learning_rate": 1.2762403658300952e-05, + "loss": 0.8849, + "step": 33828 + }, + { + "epoch": 2.18, + "learning_rate": 1.2762188969278003e-05, + "loss": 0.8805, + "step": 33829 + }, + { + "epoch": 2.18, + "learning_rate": 1.2761974280255052e-05, + "loss": 0.9158, + "step": 33830 + }, + { + "epoch": 2.18, + "learning_rate": 1.27617595912321e-05, + "loss": 0.984, + "step": 33831 + }, + { + "epoch": 2.18, + "learning_rate": 1.2761544902209151e-05, + "loss": 0.8571, + "step": 33832 + }, + { + "epoch": 2.18, + "learning_rate": 1.27613302131862e-05, + "loss": 0.8417, + "step": 33833 + }, + { + "epoch": 2.18, + "learning_rate": 1.2761115524163251e-05, + "loss": 0.9159, + "step": 33834 + }, + { + "epoch": 2.18, + "learning_rate": 1.27609008351403e-05, + "loss": 0.9076, + "step": 33835 + }, + { + "epoch": 2.18, + "learning_rate": 1.2760686146117349e-05, + "loss": 0.8629, + "step": 33836 + }, + { + "epoch": 2.18, + "learning_rate": 1.27604714570944e-05, + "loss": 0.9014, + "step": 33837 + }, + { + "epoch": 2.18, + "learning_rate": 1.276025676807145e-05, + "loss": 0.8172, + "step": 33838 + }, + { + "epoch": 2.18, + "learning_rate": 1.27600420790485e-05, + "loss": 0.9257, + "step": 33839 + }, + { + "epoch": 2.18, + "learning_rate": 1.275982739002555e-05, + "loss": 0.8165, + "step": 33840 + }, + { + "epoch": 2.18, + "learning_rate": 1.2759612701002599e-05, + "loss": 0.9225, + "step": 33841 + }, + { + "epoch": 2.18, + "learning_rate": 1.275939801197965e-05, + "loss": 0.899, + "step": 33842 + }, + { + "epoch": 2.18, + "learning_rate": 1.2759183322956698e-05, + "loss": 0.9265, + "step": 33843 + }, + { + "epoch": 2.18, + "learning_rate": 1.2758968633933749e-05, + "loss": 0.9123, + "step": 33844 + }, + { + "epoch": 2.18, + "learning_rate": 1.2758753944910798e-05, + "loss": 0.8653, + "step": 33845 + }, + { + "epoch": 2.18, + "learning_rate": 1.2758539255887847e-05, + "loss": 0.8791, + "step": 33846 + }, + { + "epoch": 2.18, + "learning_rate": 1.2758324566864898e-05, + "loss": 0.9186, + "step": 33847 + }, + { + "epoch": 2.18, + "learning_rate": 1.2758109877841947e-05, + "loss": 0.8881, + "step": 33848 + }, + { + "epoch": 2.18, + "learning_rate": 1.2757895188818995e-05, + "loss": 0.9011, + "step": 33849 + }, + { + "epoch": 2.18, + "learning_rate": 1.2757680499796046e-05, + "loss": 0.9123, + "step": 33850 + }, + { + "epoch": 2.18, + "learning_rate": 1.2757465810773095e-05, + "loss": 0.9291, + "step": 33851 + }, + { + "epoch": 2.18, + "learning_rate": 1.2757251121750144e-05, + "loss": 0.8767, + "step": 33852 + }, + { + "epoch": 2.18, + "learning_rate": 1.2757036432727196e-05, + "loss": 0.9386, + "step": 33853 + }, + { + "epoch": 2.18, + "learning_rate": 1.2756821743704247e-05, + "loss": 0.9396, + "step": 33854 + }, + { + "epoch": 2.18, + "learning_rate": 1.2756607054681296e-05, + "loss": 0.8346, + "step": 33855 + }, + { + "epoch": 2.18, + "learning_rate": 1.2756392365658345e-05, + "loss": 0.8952, + "step": 33856 + }, + { + "epoch": 2.18, + "learning_rate": 1.2756177676635396e-05, + "loss": 0.8884, + "step": 33857 + }, + { + "epoch": 2.18, + "learning_rate": 1.2755962987612445e-05, + "loss": 0.8997, + "step": 33858 + }, + { + "epoch": 2.18, + "learning_rate": 1.2755748298589494e-05, + "loss": 0.8675, + "step": 33859 + }, + { + "epoch": 2.18, + "learning_rate": 1.2755533609566544e-05, + "loss": 0.8151, + "step": 33860 + }, + { + "epoch": 2.18, + "learning_rate": 1.2755318920543593e-05, + "loss": 0.9071, + "step": 33861 + }, + { + "epoch": 2.18, + "learning_rate": 1.2755104231520642e-05, + "loss": 0.9315, + "step": 33862 + }, + { + "epoch": 2.18, + "learning_rate": 1.2754889542497693e-05, + "loss": 0.9325, + "step": 33863 + }, + { + "epoch": 2.18, + "learning_rate": 1.2754674853474742e-05, + "loss": 0.9341, + "step": 33864 + }, + { + "epoch": 2.18, + "learning_rate": 1.2754460164451792e-05, + "loss": 0.9057, + "step": 33865 + }, + { + "epoch": 2.18, + "learning_rate": 1.2754245475428841e-05, + "loss": 0.9134, + "step": 33866 + }, + { + "epoch": 2.18, + "learning_rate": 1.2754030786405894e-05, + "loss": 0.9043, + "step": 33867 + }, + { + "epoch": 2.18, + "learning_rate": 1.2753816097382943e-05, + "loss": 0.9085, + "step": 33868 + }, + { + "epoch": 2.18, + "learning_rate": 1.2753601408359992e-05, + "loss": 0.853, + "step": 33869 + }, + { + "epoch": 2.18, + "learning_rate": 1.2753386719337042e-05, + "loss": 0.953, + "step": 33870 + }, + { + "epoch": 2.18, + "learning_rate": 1.2753172030314091e-05, + "loss": 0.8759, + "step": 33871 + }, + { + "epoch": 2.18, + "learning_rate": 1.275295734129114e-05, + "loss": 0.9014, + "step": 33872 + }, + { + "epoch": 2.18, + "learning_rate": 1.2752742652268191e-05, + "loss": 0.88, + "step": 33873 + }, + { + "epoch": 2.18, + "learning_rate": 1.275252796324524e-05, + "loss": 0.8519, + "step": 33874 + }, + { + "epoch": 2.18, + "learning_rate": 1.275231327422229e-05, + "loss": 0.8919, + "step": 33875 + }, + { + "epoch": 2.18, + "learning_rate": 1.275209858519934e-05, + "loss": 0.8756, + "step": 33876 + }, + { + "epoch": 2.18, + "learning_rate": 1.2751883896176389e-05, + "loss": 0.8748, + "step": 33877 + }, + { + "epoch": 2.18, + "learning_rate": 1.275166920715344e-05, + "loss": 0.8963, + "step": 33878 + }, + { + "epoch": 2.18, + "learning_rate": 1.2751454518130488e-05, + "loss": 0.8604, + "step": 33879 + }, + { + "epoch": 2.18, + "learning_rate": 1.2751239829107537e-05, + "loss": 0.901, + "step": 33880 + }, + { + "epoch": 2.18, + "learning_rate": 1.2751025140084588e-05, + "loss": 0.8799, + "step": 33881 + }, + { + "epoch": 2.18, + "learning_rate": 1.2750810451061638e-05, + "loss": 0.9028, + "step": 33882 + }, + { + "epoch": 2.18, + "learning_rate": 1.2750595762038689e-05, + "loss": 0.8781, + "step": 33883 + }, + { + "epoch": 2.18, + "learning_rate": 1.2750381073015738e-05, + "loss": 0.9288, + "step": 33884 + }, + { + "epoch": 2.18, + "learning_rate": 1.2750166383992789e-05, + "loss": 0.8905, + "step": 33885 + }, + { + "epoch": 2.18, + "learning_rate": 1.2749951694969838e-05, + "loss": 0.8983, + "step": 33886 + }, + { + "epoch": 2.18, + "learning_rate": 1.2749737005946887e-05, + "loss": 0.8603, + "step": 33887 + }, + { + "epoch": 2.18, + "learning_rate": 1.2749522316923937e-05, + "loss": 0.9294, + "step": 33888 + }, + { + "epoch": 2.18, + "learning_rate": 1.2749307627900986e-05, + "loss": 0.9486, + "step": 33889 + }, + { + "epoch": 2.18, + "learning_rate": 1.2749092938878035e-05, + "loss": 0.9322, + "step": 33890 + }, + { + "epoch": 2.18, + "learning_rate": 1.2748878249855086e-05, + "loss": 0.9114, + "step": 33891 + }, + { + "epoch": 2.18, + "learning_rate": 1.2748663560832135e-05, + "loss": 0.8402, + "step": 33892 + }, + { + "epoch": 2.18, + "learning_rate": 1.2748448871809186e-05, + "loss": 0.9671, + "step": 33893 + }, + { + "epoch": 2.18, + "learning_rate": 1.2748234182786234e-05, + "loss": 0.8621, + "step": 33894 + }, + { + "epoch": 2.18, + "learning_rate": 1.2748019493763283e-05, + "loss": 0.8782, + "step": 33895 + }, + { + "epoch": 2.18, + "learning_rate": 1.2747804804740336e-05, + "loss": 0.8682, + "step": 33896 + }, + { + "epoch": 2.18, + "learning_rate": 1.2747590115717385e-05, + "loss": 0.8984, + "step": 33897 + }, + { + "epoch": 2.18, + "learning_rate": 1.2747375426694435e-05, + "loss": 0.871, + "step": 33898 + }, + { + "epoch": 2.18, + "learning_rate": 1.2747160737671484e-05, + "loss": 0.8436, + "step": 33899 + }, + { + "epoch": 2.18, + "learning_rate": 1.2746946048648533e-05, + "loss": 0.9012, + "step": 33900 + }, + { + "epoch": 2.18, + "learning_rate": 1.2746731359625584e-05, + "loss": 0.9845, + "step": 33901 + }, + { + "epoch": 2.18, + "learning_rate": 1.2746516670602633e-05, + "loss": 0.8912, + "step": 33902 + }, + { + "epoch": 2.18, + "learning_rate": 1.2746301981579684e-05, + "loss": 0.9779, + "step": 33903 + }, + { + "epoch": 2.18, + "learning_rate": 1.2746087292556733e-05, + "loss": 0.8903, + "step": 33904 + }, + { + "epoch": 2.18, + "learning_rate": 1.2745872603533782e-05, + "loss": 0.8902, + "step": 33905 + }, + { + "epoch": 2.18, + "learning_rate": 1.2745657914510832e-05, + "loss": 0.9098, + "step": 33906 + }, + { + "epoch": 2.18, + "learning_rate": 1.2745443225487881e-05, + "loss": 0.9779, + "step": 33907 + }, + { + "epoch": 2.18, + "learning_rate": 1.274522853646493e-05, + "loss": 0.9154, + "step": 33908 + }, + { + "epoch": 2.18, + "learning_rate": 1.274501384744198e-05, + "loss": 0.9206, + "step": 33909 + }, + { + "epoch": 2.18, + "learning_rate": 1.274479915841903e-05, + "loss": 0.8503, + "step": 33910 + }, + { + "epoch": 2.18, + "learning_rate": 1.2744584469396082e-05, + "loss": 0.9171, + "step": 33911 + }, + { + "epoch": 2.18, + "learning_rate": 1.2744369780373131e-05, + "loss": 0.8681, + "step": 33912 + }, + { + "epoch": 2.18, + "learning_rate": 1.2744155091350182e-05, + "loss": 0.8814, + "step": 33913 + }, + { + "epoch": 2.18, + "learning_rate": 1.274394040232723e-05, + "loss": 0.8907, + "step": 33914 + }, + { + "epoch": 2.18, + "learning_rate": 1.274372571330428e-05, + "loss": 0.8744, + "step": 33915 + }, + { + "epoch": 2.18, + "learning_rate": 1.274351102428133e-05, + "loss": 0.8676, + "step": 33916 + }, + { + "epoch": 2.18, + "learning_rate": 1.274329633525838e-05, + "loss": 0.8945, + "step": 33917 + }, + { + "epoch": 2.18, + "learning_rate": 1.2743081646235428e-05, + "loss": 0.9113, + "step": 33918 + }, + { + "epoch": 2.18, + "learning_rate": 1.2742866957212479e-05, + "loss": 0.8811, + "step": 33919 + }, + { + "epoch": 2.18, + "learning_rate": 1.2742652268189528e-05, + "loss": 0.9236, + "step": 33920 + }, + { + "epoch": 2.18, + "learning_rate": 1.2742437579166579e-05, + "loss": 0.8844, + "step": 33921 + }, + { + "epoch": 2.18, + "learning_rate": 1.2742222890143628e-05, + "loss": 0.9088, + "step": 33922 + }, + { + "epoch": 2.18, + "learning_rate": 1.2742008201120677e-05, + "loss": 0.8787, + "step": 33923 + }, + { + "epoch": 2.18, + "learning_rate": 1.2741793512097727e-05, + "loss": 0.8687, + "step": 33924 + }, + { + "epoch": 2.18, + "learning_rate": 1.2741578823074778e-05, + "loss": 0.8635, + "step": 33925 + }, + { + "epoch": 2.18, + "learning_rate": 1.2741364134051828e-05, + "loss": 0.8301, + "step": 33926 + }, + { + "epoch": 2.18, + "learning_rate": 1.2741149445028877e-05, + "loss": 0.9711, + "step": 33927 + }, + { + "epoch": 2.18, + "learning_rate": 1.2740934756005926e-05, + "loss": 0.8816, + "step": 33928 + }, + { + "epoch": 2.18, + "learning_rate": 1.2740720066982977e-05, + "loss": 0.9036, + "step": 33929 + }, + { + "epoch": 2.18, + "learning_rate": 1.2740505377960026e-05, + "loss": 0.8981, + "step": 33930 + }, + { + "epoch": 2.18, + "learning_rate": 1.2740290688937077e-05, + "loss": 0.9287, + "step": 33931 + }, + { + "epoch": 2.18, + "learning_rate": 1.2740075999914126e-05, + "loss": 0.8712, + "step": 33932 + }, + { + "epoch": 2.18, + "learning_rate": 1.2739861310891175e-05, + "loss": 0.8907, + "step": 33933 + }, + { + "epoch": 2.18, + "learning_rate": 1.2739646621868225e-05, + "loss": 0.8895, + "step": 33934 + }, + { + "epoch": 2.18, + "learning_rate": 1.2739431932845274e-05, + "loss": 0.8712, + "step": 33935 + }, + { + "epoch": 2.18, + "learning_rate": 1.2739217243822323e-05, + "loss": 0.7993, + "step": 33936 + }, + { + "epoch": 2.18, + "learning_rate": 1.2739002554799374e-05, + "loss": 0.8699, + "step": 33937 + }, + { + "epoch": 2.18, + "learning_rate": 1.2738787865776423e-05, + "loss": 0.9212, + "step": 33938 + }, + { + "epoch": 2.18, + "learning_rate": 1.2738573176753472e-05, + "loss": 0.8332, + "step": 33939 + }, + { + "epoch": 2.18, + "learning_rate": 1.2738358487730524e-05, + "loss": 0.8767, + "step": 33940 + }, + { + "epoch": 2.18, + "learning_rate": 1.2738143798707575e-05, + "loss": 0.9543, + "step": 33941 + }, + { + "epoch": 2.18, + "learning_rate": 1.2737929109684624e-05, + "loss": 0.9182, + "step": 33942 + }, + { + "epoch": 2.18, + "learning_rate": 1.2737714420661673e-05, + "loss": 0.8974, + "step": 33943 + }, + { + "epoch": 2.18, + "learning_rate": 1.2737499731638723e-05, + "loss": 0.877, + "step": 33944 + }, + { + "epoch": 2.18, + "learning_rate": 1.2737285042615772e-05, + "loss": 0.8495, + "step": 33945 + }, + { + "epoch": 2.18, + "learning_rate": 1.2737070353592821e-05, + "loss": 0.8737, + "step": 33946 + }, + { + "epoch": 2.18, + "learning_rate": 1.2736855664569872e-05, + "loss": 0.9127, + "step": 33947 + }, + { + "epoch": 2.18, + "learning_rate": 1.2736640975546921e-05, + "loss": 0.9022, + "step": 33948 + }, + { + "epoch": 2.18, + "learning_rate": 1.273642628652397e-05, + "loss": 0.8276, + "step": 33949 + }, + { + "epoch": 2.18, + "learning_rate": 1.273621159750102e-05, + "loss": 0.8989, + "step": 33950 + }, + { + "epoch": 2.18, + "learning_rate": 1.273599690847807e-05, + "loss": 0.9865, + "step": 33951 + }, + { + "epoch": 2.18, + "learning_rate": 1.273578221945512e-05, + "loss": 0.9691, + "step": 33952 + }, + { + "epoch": 2.18, + "learning_rate": 1.273556753043217e-05, + "loss": 0.8583, + "step": 33953 + }, + { + "epoch": 2.18, + "learning_rate": 1.2735352841409218e-05, + "loss": 0.958, + "step": 33954 + }, + { + "epoch": 2.18, + "learning_rate": 1.273513815238627e-05, + "loss": 0.9008, + "step": 33955 + }, + { + "epoch": 2.18, + "learning_rate": 1.273492346336332e-05, + "loss": 0.8456, + "step": 33956 + }, + { + "epoch": 2.18, + "learning_rate": 1.273470877434037e-05, + "loss": 0.8417, + "step": 33957 + }, + { + "epoch": 2.18, + "learning_rate": 1.2734494085317419e-05, + "loss": 0.9004, + "step": 33958 + }, + { + "epoch": 2.18, + "learning_rate": 1.2734279396294468e-05, + "loss": 0.8193, + "step": 33959 + }, + { + "epoch": 2.18, + "learning_rate": 1.2734064707271519e-05, + "loss": 0.9073, + "step": 33960 + }, + { + "epoch": 2.18, + "learning_rate": 1.2733850018248568e-05, + "loss": 0.9316, + "step": 33961 + }, + { + "epoch": 2.19, + "learning_rate": 1.2733635329225618e-05, + "loss": 0.9413, + "step": 33962 + }, + { + "epoch": 2.19, + "learning_rate": 1.2733420640202667e-05, + "loss": 0.8889, + "step": 33963 + }, + { + "epoch": 2.19, + "learning_rate": 1.2733205951179716e-05, + "loss": 0.8376, + "step": 33964 + }, + { + "epoch": 2.19, + "learning_rate": 1.2732991262156767e-05, + "loss": 0.9436, + "step": 33965 + }, + { + "epoch": 2.19, + "learning_rate": 1.2732776573133816e-05, + "loss": 0.8854, + "step": 33966 + }, + { + "epoch": 2.19, + "learning_rate": 1.2732561884110865e-05, + "loss": 0.8509, + "step": 33967 + }, + { + "epoch": 2.19, + "learning_rate": 1.2732347195087916e-05, + "loss": 0.8691, + "step": 33968 + }, + { + "epoch": 2.19, + "learning_rate": 1.2732132506064966e-05, + "loss": 0.8919, + "step": 33969 + }, + { + "epoch": 2.19, + "learning_rate": 1.2731917817042017e-05, + "loss": 0.9499, + "step": 33970 + }, + { + "epoch": 2.19, + "learning_rate": 1.2731703128019066e-05, + "loss": 0.8977, + "step": 33971 + }, + { + "epoch": 2.19, + "learning_rate": 1.2731488438996116e-05, + "loss": 0.9399, + "step": 33972 + }, + { + "epoch": 2.19, + "learning_rate": 1.2731273749973165e-05, + "loss": 0.9321, + "step": 33973 + }, + { + "epoch": 2.19, + "learning_rate": 1.2731059060950214e-05, + "loss": 0.8436, + "step": 33974 + }, + { + "epoch": 2.19, + "learning_rate": 1.2730844371927265e-05, + "loss": 0.8683, + "step": 33975 + }, + { + "epoch": 2.19, + "learning_rate": 1.2730629682904314e-05, + "loss": 0.8822, + "step": 33976 + }, + { + "epoch": 2.19, + "learning_rate": 1.2730414993881363e-05, + "loss": 0.9076, + "step": 33977 + }, + { + "epoch": 2.19, + "learning_rate": 1.2730200304858414e-05, + "loss": 0.8874, + "step": 33978 + }, + { + "epoch": 2.19, + "learning_rate": 1.2729985615835463e-05, + "loss": 0.8626, + "step": 33979 + }, + { + "epoch": 2.19, + "learning_rate": 1.2729770926812513e-05, + "loss": 0.8642, + "step": 33980 + }, + { + "epoch": 2.19, + "learning_rate": 1.2729556237789562e-05, + "loss": 0.8851, + "step": 33981 + }, + { + "epoch": 2.19, + "learning_rate": 1.2729341548766611e-05, + "loss": 0.9103, + "step": 33982 + }, + { + "epoch": 2.19, + "learning_rate": 1.2729126859743662e-05, + "loss": 0.9296, + "step": 33983 + }, + { + "epoch": 2.19, + "learning_rate": 1.2728912170720712e-05, + "loss": 0.8099, + "step": 33984 + }, + { + "epoch": 2.19, + "learning_rate": 1.2728697481697763e-05, + "loss": 0.9067, + "step": 33985 + }, + { + "epoch": 2.19, + "learning_rate": 1.2728482792674812e-05, + "loss": 0.9071, + "step": 33986 + }, + { + "epoch": 2.19, + "learning_rate": 1.2728268103651861e-05, + "loss": 0.8607, + "step": 33987 + }, + { + "epoch": 2.19, + "learning_rate": 1.2728053414628912e-05, + "loss": 0.8205, + "step": 33988 + }, + { + "epoch": 2.19, + "learning_rate": 1.272783872560596e-05, + "loss": 0.9505, + "step": 33989 + }, + { + "epoch": 2.19, + "learning_rate": 1.2727624036583011e-05, + "loss": 0.8432, + "step": 33990 + }, + { + "epoch": 2.19, + "learning_rate": 1.272740934756006e-05, + "loss": 0.9063, + "step": 33991 + }, + { + "epoch": 2.19, + "learning_rate": 1.272719465853711e-05, + "loss": 0.9202, + "step": 33992 + }, + { + "epoch": 2.19, + "learning_rate": 1.272697996951416e-05, + "loss": 0.8482, + "step": 33993 + }, + { + "epoch": 2.19, + "learning_rate": 1.2726765280491209e-05, + "loss": 0.9045, + "step": 33994 + }, + { + "epoch": 2.19, + "learning_rate": 1.2726550591468258e-05, + "loss": 0.9264, + "step": 33995 + }, + { + "epoch": 2.19, + "learning_rate": 1.2726335902445309e-05, + "loss": 0.8404, + "step": 33996 + }, + { + "epoch": 2.19, + "learning_rate": 1.2726121213422358e-05, + "loss": 0.813, + "step": 33997 + }, + { + "epoch": 2.19, + "learning_rate": 1.272590652439941e-05, + "loss": 0.905, + "step": 33998 + }, + { + "epoch": 2.19, + "learning_rate": 1.2725691835376459e-05, + "loss": 0.8692, + "step": 33999 + }, + { + "epoch": 2.19, + "learning_rate": 1.272547714635351e-05, + "loss": 0.8637, + "step": 34000 + }, + { + "epoch": 2.19, + "eval_laion/cider": 0.5403510643251865, + "eval_laion/meteor": 0.394125563342911, + "eval_laion/pred_num_words": 36.9140625, + "eval_laion/sacrebleu": 16.077688461132908, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0444412231445312, + "eval_runtime": 96.4282, + "eval_samples_per_second": 5.175, + "eval_steps_per_second": 0.332, + "step": 34000 + }, + { + "epoch": 2.19, + "learning_rate": 1.2725262457330558e-05, + "loss": 0.8818, + "step": 34001 + }, + { + "epoch": 2.19, + "learning_rate": 1.2725047768307607e-05, + "loss": 0.9851, + "step": 34002 + }, + { + "epoch": 2.19, + "learning_rate": 1.2724833079284658e-05, + "loss": 0.8423, + "step": 34003 + }, + { + "epoch": 2.19, + "learning_rate": 1.2724618390261707e-05, + "loss": 0.8867, + "step": 34004 + }, + { + "epoch": 2.19, + "learning_rate": 1.2724403701238756e-05, + "loss": 0.843, + "step": 34005 + }, + { + "epoch": 2.19, + "learning_rate": 1.2724189012215807e-05, + "loss": 0.8682, + "step": 34006 + }, + { + "epoch": 2.19, + "learning_rate": 1.2723974323192856e-05, + "loss": 0.9293, + "step": 34007 + }, + { + "epoch": 2.19, + "learning_rate": 1.2723759634169906e-05, + "loss": 0.8358, + "step": 34008 + }, + { + "epoch": 2.19, + "learning_rate": 1.2723544945146955e-05, + "loss": 0.919, + "step": 34009 + }, + { + "epoch": 2.19, + "learning_rate": 1.2723330256124004e-05, + "loss": 0.8677, + "step": 34010 + }, + { + "epoch": 2.19, + "learning_rate": 1.2723115567101055e-05, + "loss": 0.9224, + "step": 34011 + }, + { + "epoch": 2.19, + "learning_rate": 1.2722900878078104e-05, + "loss": 0.9353, + "step": 34012 + }, + { + "epoch": 2.19, + "learning_rate": 1.2722686189055156e-05, + "loss": 0.8665, + "step": 34013 + }, + { + "epoch": 2.19, + "learning_rate": 1.2722471500032205e-05, + "loss": 0.886, + "step": 34014 + }, + { + "epoch": 2.19, + "learning_rate": 1.2722256811009254e-05, + "loss": 0.8867, + "step": 34015 + }, + { + "epoch": 2.19, + "learning_rate": 1.2722042121986305e-05, + "loss": 0.9276, + "step": 34016 + }, + { + "epoch": 2.19, + "learning_rate": 1.2721827432963354e-05, + "loss": 0.9331, + "step": 34017 + }, + { + "epoch": 2.19, + "learning_rate": 1.2721612743940404e-05, + "loss": 0.8318, + "step": 34018 + }, + { + "epoch": 2.19, + "learning_rate": 1.2721398054917453e-05, + "loss": 0.8317, + "step": 34019 + }, + { + "epoch": 2.19, + "learning_rate": 1.2721183365894502e-05, + "loss": 0.9347, + "step": 34020 + }, + { + "epoch": 2.19, + "learning_rate": 1.2720968676871553e-05, + "loss": 0.8972, + "step": 34021 + }, + { + "epoch": 2.19, + "learning_rate": 1.2720753987848602e-05, + "loss": 0.9644, + "step": 34022 + }, + { + "epoch": 2.19, + "learning_rate": 1.2720539298825651e-05, + "loss": 0.9404, + "step": 34023 + }, + { + "epoch": 2.19, + "learning_rate": 1.2720324609802702e-05, + "loss": 0.9664, + "step": 34024 + }, + { + "epoch": 2.19, + "learning_rate": 1.272010992077975e-05, + "loss": 0.8629, + "step": 34025 + }, + { + "epoch": 2.19, + "learning_rate": 1.27198952317568e-05, + "loss": 0.9589, + "step": 34026 + }, + { + "epoch": 2.19, + "learning_rate": 1.2719680542733852e-05, + "loss": 0.9168, + "step": 34027 + }, + { + "epoch": 2.19, + "learning_rate": 1.2719465853710903e-05, + "loss": 0.8142, + "step": 34028 + }, + { + "epoch": 2.19, + "learning_rate": 1.2719251164687951e-05, + "loss": 0.9003, + "step": 34029 + }, + { + "epoch": 2.19, + "learning_rate": 1.2719036475665e-05, + "loss": 0.8752, + "step": 34030 + }, + { + "epoch": 2.19, + "learning_rate": 1.2718821786642051e-05, + "loss": 0.8519, + "step": 34031 + }, + { + "epoch": 2.19, + "learning_rate": 1.27186070976191e-05, + "loss": 0.8689, + "step": 34032 + }, + { + "epoch": 2.19, + "learning_rate": 1.2718392408596149e-05, + "loss": 0.8234, + "step": 34033 + }, + { + "epoch": 2.19, + "learning_rate": 1.27181777195732e-05, + "loss": 0.956, + "step": 34034 + }, + { + "epoch": 2.19, + "learning_rate": 1.2717963030550249e-05, + "loss": 0.9152, + "step": 34035 + }, + { + "epoch": 2.19, + "learning_rate": 1.2717748341527298e-05, + "loss": 0.8935, + "step": 34036 + }, + { + "epoch": 2.19, + "learning_rate": 1.2717533652504348e-05, + "loss": 0.9389, + "step": 34037 + }, + { + "epoch": 2.19, + "learning_rate": 1.2717318963481397e-05, + "loss": 0.8843, + "step": 34038 + }, + { + "epoch": 2.19, + "learning_rate": 1.2717104274458448e-05, + "loss": 0.9095, + "step": 34039 + }, + { + "epoch": 2.19, + "learning_rate": 1.2716889585435497e-05, + "loss": 0.8452, + "step": 34040 + }, + { + "epoch": 2.19, + "learning_rate": 1.2716674896412546e-05, + "loss": 0.9208, + "step": 34041 + }, + { + "epoch": 2.19, + "learning_rate": 1.2716460207389598e-05, + "loss": 0.8496, + "step": 34042 + }, + { + "epoch": 2.19, + "learning_rate": 1.2716245518366647e-05, + "loss": 0.8217, + "step": 34043 + }, + { + "epoch": 2.19, + "learning_rate": 1.2716030829343698e-05, + "loss": 0.8931, + "step": 34044 + }, + { + "epoch": 2.19, + "learning_rate": 1.2715816140320747e-05, + "loss": 0.9152, + "step": 34045 + }, + { + "epoch": 2.19, + "learning_rate": 1.2715601451297796e-05, + "loss": 0.8534, + "step": 34046 + }, + { + "epoch": 2.19, + "learning_rate": 1.2715386762274846e-05, + "loss": 0.8704, + "step": 34047 + }, + { + "epoch": 2.19, + "learning_rate": 1.2715172073251895e-05, + "loss": 0.8995, + "step": 34048 + }, + { + "epoch": 2.19, + "learning_rate": 1.2714957384228946e-05, + "loss": 0.8488, + "step": 34049 + }, + { + "epoch": 2.19, + "learning_rate": 1.2714742695205995e-05, + "loss": 0.9569, + "step": 34050 + }, + { + "epoch": 2.19, + "learning_rate": 1.2714528006183044e-05, + "loss": 0.924, + "step": 34051 + }, + { + "epoch": 2.19, + "learning_rate": 1.2714313317160095e-05, + "loss": 0.8918, + "step": 34052 + }, + { + "epoch": 2.19, + "learning_rate": 1.2714098628137144e-05, + "loss": 0.9364, + "step": 34053 + }, + { + "epoch": 2.19, + "learning_rate": 1.2713883939114193e-05, + "loss": 0.8807, + "step": 34054 + }, + { + "epoch": 2.19, + "learning_rate": 1.2713669250091243e-05, + "loss": 0.8937, + "step": 34055 + }, + { + "epoch": 2.19, + "learning_rate": 1.2713454561068294e-05, + "loss": 0.8447, + "step": 34056 + }, + { + "epoch": 2.19, + "learning_rate": 1.2713239872045345e-05, + "loss": 0.8767, + "step": 34057 + }, + { + "epoch": 2.19, + "learning_rate": 1.2713025183022393e-05, + "loss": 0.8451, + "step": 34058 + }, + { + "epoch": 2.19, + "learning_rate": 1.2712810493999444e-05, + "loss": 0.938, + "step": 34059 + }, + { + "epoch": 2.19, + "learning_rate": 1.2712595804976493e-05, + "loss": 0.8306, + "step": 34060 + }, + { + "epoch": 2.19, + "learning_rate": 1.2712381115953542e-05, + "loss": 0.8739, + "step": 34061 + }, + { + "epoch": 2.19, + "learning_rate": 1.2712166426930593e-05, + "loss": 0.8649, + "step": 34062 + }, + { + "epoch": 2.19, + "learning_rate": 1.2711951737907642e-05, + "loss": 0.9363, + "step": 34063 + }, + { + "epoch": 2.19, + "learning_rate": 1.271173704888469e-05, + "loss": 0.8441, + "step": 34064 + }, + { + "epoch": 2.19, + "learning_rate": 1.2711522359861741e-05, + "loss": 0.8461, + "step": 34065 + }, + { + "epoch": 2.19, + "learning_rate": 1.271130767083879e-05, + "loss": 0.8883, + "step": 34066 + }, + { + "epoch": 2.19, + "learning_rate": 1.2711092981815841e-05, + "loss": 0.8489, + "step": 34067 + }, + { + "epoch": 2.19, + "learning_rate": 1.271087829279289e-05, + "loss": 0.9004, + "step": 34068 + }, + { + "epoch": 2.19, + "learning_rate": 1.2710663603769939e-05, + "loss": 0.9284, + "step": 34069 + }, + { + "epoch": 2.19, + "learning_rate": 1.271044891474699e-05, + "loss": 0.9157, + "step": 34070 + }, + { + "epoch": 2.19, + "learning_rate": 1.271023422572404e-05, + "loss": 0.9116, + "step": 34071 + }, + { + "epoch": 2.19, + "learning_rate": 1.2710019536701091e-05, + "loss": 0.8475, + "step": 34072 + }, + { + "epoch": 2.19, + "learning_rate": 1.270980484767814e-05, + "loss": 0.9368, + "step": 34073 + }, + { + "epoch": 2.19, + "learning_rate": 1.2709590158655189e-05, + "loss": 0.9647, + "step": 34074 + }, + { + "epoch": 2.19, + "learning_rate": 1.270937546963224e-05, + "loss": 0.9127, + "step": 34075 + }, + { + "epoch": 2.19, + "learning_rate": 1.2709160780609288e-05, + "loss": 0.874, + "step": 34076 + }, + { + "epoch": 2.19, + "learning_rate": 1.2708946091586339e-05, + "loss": 0.8995, + "step": 34077 + }, + { + "epoch": 2.19, + "learning_rate": 1.2708731402563388e-05, + "loss": 0.8805, + "step": 34078 + }, + { + "epoch": 2.19, + "learning_rate": 1.2708516713540437e-05, + "loss": 0.8896, + "step": 34079 + }, + { + "epoch": 2.19, + "learning_rate": 1.2708302024517488e-05, + "loss": 0.9012, + "step": 34080 + }, + { + "epoch": 2.19, + "learning_rate": 1.2708087335494537e-05, + "loss": 0.9057, + "step": 34081 + }, + { + "epoch": 2.19, + "learning_rate": 1.2707872646471586e-05, + "loss": 0.9298, + "step": 34082 + }, + { + "epoch": 2.19, + "learning_rate": 1.2707657957448636e-05, + "loss": 0.9, + "step": 34083 + }, + { + "epoch": 2.19, + "learning_rate": 1.2707443268425685e-05, + "loss": 0.8253, + "step": 34084 + }, + { + "epoch": 2.19, + "learning_rate": 1.2707228579402738e-05, + "loss": 0.8247, + "step": 34085 + }, + { + "epoch": 2.19, + "learning_rate": 1.2707013890379787e-05, + "loss": 0.9115, + "step": 34086 + }, + { + "epoch": 2.19, + "learning_rate": 1.2706799201356837e-05, + "loss": 0.9246, + "step": 34087 + }, + { + "epoch": 2.19, + "learning_rate": 1.2706584512333886e-05, + "loss": 0.8148, + "step": 34088 + }, + { + "epoch": 2.19, + "learning_rate": 1.2706369823310935e-05, + "loss": 0.8524, + "step": 34089 + }, + { + "epoch": 2.19, + "learning_rate": 1.2706155134287986e-05, + "loss": 0.9365, + "step": 34090 + }, + { + "epoch": 2.19, + "learning_rate": 1.2705940445265035e-05, + "loss": 0.9194, + "step": 34091 + }, + { + "epoch": 2.19, + "learning_rate": 1.2705725756242084e-05, + "loss": 0.9369, + "step": 34092 + }, + { + "epoch": 2.19, + "learning_rate": 1.2705511067219134e-05, + "loss": 0.9868, + "step": 34093 + }, + { + "epoch": 2.19, + "learning_rate": 1.2705296378196183e-05, + "loss": 0.9462, + "step": 34094 + }, + { + "epoch": 2.19, + "learning_rate": 1.2705081689173232e-05, + "loss": 0.8632, + "step": 34095 + }, + { + "epoch": 2.19, + "learning_rate": 1.2704867000150283e-05, + "loss": 0.8957, + "step": 34096 + }, + { + "epoch": 2.19, + "learning_rate": 1.2704652311127332e-05, + "loss": 0.9924, + "step": 34097 + }, + { + "epoch": 2.19, + "learning_rate": 1.2704437622104383e-05, + "loss": 0.834, + "step": 34098 + }, + { + "epoch": 2.19, + "learning_rate": 1.2704222933081432e-05, + "loss": 0.9629, + "step": 34099 + }, + { + "epoch": 2.19, + "learning_rate": 1.2704008244058484e-05, + "loss": 0.8721, + "step": 34100 + }, + { + "epoch": 2.19, + "learning_rate": 1.2703793555035533e-05, + "loss": 0.9198, + "step": 34101 + }, + { + "epoch": 2.19, + "learning_rate": 1.2703578866012582e-05, + "loss": 0.8805, + "step": 34102 + }, + { + "epoch": 2.19, + "learning_rate": 1.2703364176989632e-05, + "loss": 0.9103, + "step": 34103 + }, + { + "epoch": 2.19, + "learning_rate": 1.2703149487966681e-05, + "loss": 0.953, + "step": 34104 + }, + { + "epoch": 2.19, + "learning_rate": 1.2702934798943732e-05, + "loss": 0.9167, + "step": 34105 + }, + { + "epoch": 2.19, + "learning_rate": 1.2702720109920781e-05, + "loss": 0.8551, + "step": 34106 + }, + { + "epoch": 2.19, + "learning_rate": 1.270250542089783e-05, + "loss": 0.9001, + "step": 34107 + }, + { + "epoch": 2.19, + "learning_rate": 1.270229073187488e-05, + "loss": 0.9046, + "step": 34108 + }, + { + "epoch": 2.19, + "learning_rate": 1.270207604285193e-05, + "loss": 0.8897, + "step": 34109 + }, + { + "epoch": 2.19, + "learning_rate": 1.2701861353828979e-05, + "loss": 0.8802, + "step": 34110 + }, + { + "epoch": 2.19, + "learning_rate": 1.270164666480603e-05, + "loss": 0.8934, + "step": 34111 + }, + { + "epoch": 2.19, + "learning_rate": 1.2701431975783078e-05, + "loss": 0.8977, + "step": 34112 + }, + { + "epoch": 2.19, + "learning_rate": 1.2701217286760127e-05, + "loss": 0.8726, + "step": 34113 + }, + { + "epoch": 2.19, + "learning_rate": 1.270100259773718e-05, + "loss": 0.8728, + "step": 34114 + }, + { + "epoch": 2.19, + "learning_rate": 1.270078790871423e-05, + "loss": 0.9281, + "step": 34115 + }, + { + "epoch": 2.19, + "learning_rate": 1.270057321969128e-05, + "loss": 0.8754, + "step": 34116 + }, + { + "epoch": 2.2, + "learning_rate": 1.2700358530668328e-05, + "loss": 0.8326, + "step": 34117 + }, + { + "epoch": 2.2, + "learning_rate": 1.2700143841645379e-05, + "loss": 0.8833, + "step": 34118 + }, + { + "epoch": 2.2, + "learning_rate": 1.2699929152622428e-05, + "loss": 0.8982, + "step": 34119 + }, + { + "epoch": 2.2, + "learning_rate": 1.2699714463599477e-05, + "loss": 0.8774, + "step": 34120 + }, + { + "epoch": 2.2, + "learning_rate": 1.2699499774576527e-05, + "loss": 0.899, + "step": 34121 + }, + { + "epoch": 2.2, + "learning_rate": 1.2699285085553576e-05, + "loss": 0.9462, + "step": 34122 + }, + { + "epoch": 2.2, + "learning_rate": 1.2699070396530625e-05, + "loss": 0.9272, + "step": 34123 + }, + { + "epoch": 2.2, + "learning_rate": 1.2698855707507676e-05, + "loss": 0.912, + "step": 34124 + }, + { + "epoch": 2.2, + "learning_rate": 1.2698641018484725e-05, + "loss": 0.9562, + "step": 34125 + }, + { + "epoch": 2.2, + "learning_rate": 1.2698426329461776e-05, + "loss": 0.8624, + "step": 34126 + }, + { + "epoch": 2.2, + "learning_rate": 1.2698211640438825e-05, + "loss": 0.8668, + "step": 34127 + }, + { + "epoch": 2.2, + "learning_rate": 1.2697996951415874e-05, + "loss": 0.9186, + "step": 34128 + }, + { + "epoch": 2.2, + "learning_rate": 1.2697782262392926e-05, + "loss": 0.9154, + "step": 34129 + }, + { + "epoch": 2.2, + "learning_rate": 1.2697567573369975e-05, + "loss": 0.976, + "step": 34130 + }, + { + "epoch": 2.2, + "learning_rate": 1.2697352884347026e-05, + "loss": 0.847, + "step": 34131 + }, + { + "epoch": 2.2, + "learning_rate": 1.2697138195324074e-05, + "loss": 0.9225, + "step": 34132 + }, + { + "epoch": 2.2, + "learning_rate": 1.2696923506301123e-05, + "loss": 0.9303, + "step": 34133 + }, + { + "epoch": 2.2, + "learning_rate": 1.2696708817278174e-05, + "loss": 0.9292, + "step": 34134 + }, + { + "epoch": 2.2, + "learning_rate": 1.2696494128255223e-05, + "loss": 0.8853, + "step": 34135 + }, + { + "epoch": 2.2, + "learning_rate": 1.2696279439232274e-05, + "loss": 0.8725, + "step": 34136 + }, + { + "epoch": 2.2, + "learning_rate": 1.2696064750209323e-05, + "loss": 0.8624, + "step": 34137 + }, + { + "epoch": 2.2, + "learning_rate": 1.2695850061186372e-05, + "loss": 0.9002, + "step": 34138 + }, + { + "epoch": 2.2, + "learning_rate": 1.2695635372163422e-05, + "loss": 0.9174, + "step": 34139 + }, + { + "epoch": 2.2, + "learning_rate": 1.2695420683140471e-05, + "loss": 0.8508, + "step": 34140 + }, + { + "epoch": 2.2, + "learning_rate": 1.269520599411752e-05, + "loss": 0.8337, + "step": 34141 + }, + { + "epoch": 2.2, + "learning_rate": 1.2694991305094571e-05, + "loss": 0.8666, + "step": 34142 + }, + { + "epoch": 2.2, + "learning_rate": 1.2694776616071622e-05, + "loss": 0.9188, + "step": 34143 + }, + { + "epoch": 2.2, + "learning_rate": 1.2694561927048672e-05, + "loss": 0.8252, + "step": 34144 + }, + { + "epoch": 2.2, + "learning_rate": 1.2694347238025721e-05, + "loss": 0.8178, + "step": 34145 + }, + { + "epoch": 2.2, + "learning_rate": 1.2694132549002772e-05, + "loss": 0.9587, + "step": 34146 + }, + { + "epoch": 2.2, + "learning_rate": 1.269391785997982e-05, + "loss": 0.8748, + "step": 34147 + }, + { + "epoch": 2.2, + "learning_rate": 1.269370317095687e-05, + "loss": 0.8777, + "step": 34148 + }, + { + "epoch": 2.2, + "learning_rate": 1.269348848193392e-05, + "loss": 0.9017, + "step": 34149 + }, + { + "epoch": 2.2, + "learning_rate": 1.269327379291097e-05, + "loss": 0.8694, + "step": 34150 + }, + { + "epoch": 2.2, + "learning_rate": 1.2693059103888018e-05, + "loss": 0.8253, + "step": 34151 + }, + { + "epoch": 2.2, + "learning_rate": 1.2692844414865069e-05, + "loss": 0.9124, + "step": 34152 + }, + { + "epoch": 2.2, + "learning_rate": 1.2692629725842118e-05, + "loss": 0.8248, + "step": 34153 + }, + { + "epoch": 2.2, + "learning_rate": 1.2692415036819169e-05, + "loss": 0.8913, + "step": 34154 + }, + { + "epoch": 2.2, + "learning_rate": 1.2692200347796218e-05, + "loss": 0.8608, + "step": 34155 + }, + { + "epoch": 2.2, + "learning_rate": 1.2691985658773267e-05, + "loss": 0.8814, + "step": 34156 + }, + { + "epoch": 2.2, + "learning_rate": 1.2691770969750317e-05, + "loss": 0.9361, + "step": 34157 + }, + { + "epoch": 2.2, + "learning_rate": 1.2691556280727368e-05, + "loss": 0.8666, + "step": 34158 + }, + { + "epoch": 2.2, + "learning_rate": 1.2691341591704419e-05, + "loss": 0.9005, + "step": 34159 + }, + { + "epoch": 2.2, + "learning_rate": 1.2691126902681468e-05, + "loss": 0.9008, + "step": 34160 + }, + { + "epoch": 2.2, + "learning_rate": 1.2690912213658517e-05, + "loss": 0.9292, + "step": 34161 + }, + { + "epoch": 2.2, + "learning_rate": 1.2690697524635567e-05, + "loss": 0.8113, + "step": 34162 + }, + { + "epoch": 2.2, + "learning_rate": 1.2690482835612616e-05, + "loss": 0.8504, + "step": 34163 + }, + { + "epoch": 2.2, + "learning_rate": 1.2690268146589667e-05, + "loss": 0.8876, + "step": 34164 + }, + { + "epoch": 2.2, + "learning_rate": 1.2690053457566716e-05, + "loss": 0.8639, + "step": 34165 + }, + { + "epoch": 2.2, + "learning_rate": 1.2689838768543765e-05, + "loss": 0.8715, + "step": 34166 + }, + { + "epoch": 2.2, + "learning_rate": 1.2689624079520815e-05, + "loss": 0.9289, + "step": 34167 + }, + { + "epoch": 2.2, + "learning_rate": 1.2689409390497864e-05, + "loss": 0.8986, + "step": 34168 + }, + { + "epoch": 2.2, + "learning_rate": 1.2689194701474913e-05, + "loss": 0.9205, + "step": 34169 + }, + { + "epoch": 2.2, + "learning_rate": 1.2688980012451964e-05, + "loss": 0.8703, + "step": 34170 + }, + { + "epoch": 2.2, + "learning_rate": 1.2688765323429013e-05, + "loss": 0.9097, + "step": 34171 + }, + { + "epoch": 2.2, + "learning_rate": 1.2688550634406065e-05, + "loss": 0.9416, + "step": 34172 + }, + { + "epoch": 2.2, + "learning_rate": 1.2688335945383114e-05, + "loss": 0.9476, + "step": 34173 + }, + { + "epoch": 2.2, + "learning_rate": 1.2688121256360165e-05, + "loss": 0.8486, + "step": 34174 + }, + { + "epoch": 2.2, + "learning_rate": 1.2687906567337214e-05, + "loss": 0.8988, + "step": 34175 + }, + { + "epoch": 2.2, + "learning_rate": 1.2687691878314263e-05, + "loss": 0.8675, + "step": 34176 + }, + { + "epoch": 2.2, + "learning_rate": 1.2687477189291313e-05, + "loss": 0.8589, + "step": 34177 + }, + { + "epoch": 2.2, + "learning_rate": 1.2687262500268362e-05, + "loss": 0.9411, + "step": 34178 + }, + { + "epoch": 2.2, + "learning_rate": 1.2687047811245411e-05, + "loss": 0.9222, + "step": 34179 + }, + { + "epoch": 2.2, + "learning_rate": 1.2686833122222462e-05, + "loss": 0.887, + "step": 34180 + }, + { + "epoch": 2.2, + "learning_rate": 1.2686618433199511e-05, + "loss": 0.906, + "step": 34181 + }, + { + "epoch": 2.2, + "learning_rate": 1.268640374417656e-05, + "loss": 0.9285, + "step": 34182 + }, + { + "epoch": 2.2, + "learning_rate": 1.268618905515361e-05, + "loss": 0.8464, + "step": 34183 + }, + { + "epoch": 2.2, + "learning_rate": 1.268597436613066e-05, + "loss": 0.8769, + "step": 34184 + }, + { + "epoch": 2.2, + "learning_rate": 1.268575967710771e-05, + "loss": 0.8879, + "step": 34185 + }, + { + "epoch": 2.2, + "learning_rate": 1.268554498808476e-05, + "loss": 0.8985, + "step": 34186 + }, + { + "epoch": 2.2, + "learning_rate": 1.2685330299061812e-05, + "loss": 0.9044, + "step": 34187 + }, + { + "epoch": 2.2, + "learning_rate": 1.268511561003886e-05, + "loss": 0.9242, + "step": 34188 + }, + { + "epoch": 2.2, + "learning_rate": 1.268490092101591e-05, + "loss": 0.8302, + "step": 34189 + }, + { + "epoch": 2.2, + "learning_rate": 1.268468623199296e-05, + "loss": 0.9243, + "step": 34190 + }, + { + "epoch": 2.2, + "learning_rate": 1.268447154297001e-05, + "loss": 0.8467, + "step": 34191 + }, + { + "epoch": 2.2, + "learning_rate": 1.2684256853947058e-05, + "loss": 0.8978, + "step": 34192 + }, + { + "epoch": 2.2, + "learning_rate": 1.2684042164924109e-05, + "loss": 0.8827, + "step": 34193 + }, + { + "epoch": 2.2, + "learning_rate": 1.2683827475901158e-05, + "loss": 0.8991, + "step": 34194 + }, + { + "epoch": 2.2, + "learning_rate": 1.2683612786878208e-05, + "loss": 0.8648, + "step": 34195 + }, + { + "epoch": 2.2, + "learning_rate": 1.2683398097855257e-05, + "loss": 0.8978, + "step": 34196 + }, + { + "epoch": 2.2, + "learning_rate": 1.2683183408832306e-05, + "loss": 0.8609, + "step": 34197 + }, + { + "epoch": 2.2, + "learning_rate": 1.2682968719809357e-05, + "loss": 0.9298, + "step": 34198 + }, + { + "epoch": 2.2, + "learning_rate": 1.2682754030786406e-05, + "loss": 0.9264, + "step": 34199 + }, + { + "epoch": 2.2, + "learning_rate": 1.2682539341763455e-05, + "loss": 0.8548, + "step": 34200 + }, + { + "epoch": 2.2, + "learning_rate": 1.2682324652740507e-05, + "loss": 0.8974, + "step": 34201 + }, + { + "epoch": 2.2, + "learning_rate": 1.2682109963717556e-05, + "loss": 0.8956, + "step": 34202 + }, + { + "epoch": 2.2, + "learning_rate": 1.2681895274694607e-05, + "loss": 0.9107, + "step": 34203 + }, + { + "epoch": 2.2, + "learning_rate": 1.2681680585671656e-05, + "loss": 0.9742, + "step": 34204 + }, + { + "epoch": 2.2, + "learning_rate": 1.2681465896648707e-05, + "loss": 0.9665, + "step": 34205 + }, + { + "epoch": 2.2, + "learning_rate": 1.2681251207625756e-05, + "loss": 0.9304, + "step": 34206 + }, + { + "epoch": 2.2, + "learning_rate": 1.2681036518602804e-05, + "loss": 0.8555, + "step": 34207 + }, + { + "epoch": 2.2, + "learning_rate": 1.2680821829579855e-05, + "loss": 0.915, + "step": 34208 + }, + { + "epoch": 2.2, + "learning_rate": 1.2680607140556904e-05, + "loss": 0.8471, + "step": 34209 + }, + { + "epoch": 2.2, + "learning_rate": 1.2680392451533953e-05, + "loss": 0.898, + "step": 34210 + }, + { + "epoch": 2.2, + "learning_rate": 1.2680177762511004e-05, + "loss": 0.9164, + "step": 34211 + }, + { + "epoch": 2.2, + "learning_rate": 1.2679963073488053e-05, + "loss": 0.9927, + "step": 34212 + }, + { + "epoch": 2.2, + "learning_rate": 1.2679748384465103e-05, + "loss": 0.8867, + "step": 34213 + }, + { + "epoch": 2.2, + "learning_rate": 1.2679533695442152e-05, + "loss": 0.8769, + "step": 34214 + }, + { + "epoch": 2.2, + "learning_rate": 1.2679319006419201e-05, + "loss": 0.8572, + "step": 34215 + }, + { + "epoch": 2.2, + "learning_rate": 1.2679104317396254e-05, + "loss": 0.9359, + "step": 34216 + }, + { + "epoch": 2.2, + "learning_rate": 1.2678889628373303e-05, + "loss": 0.8658, + "step": 34217 + }, + { + "epoch": 2.2, + "learning_rate": 1.2678674939350353e-05, + "loss": 0.8905, + "step": 34218 + }, + { + "epoch": 2.2, + "learning_rate": 1.2678460250327402e-05, + "loss": 0.9522, + "step": 34219 + }, + { + "epoch": 2.2, + "learning_rate": 1.2678245561304451e-05, + "loss": 0.9322, + "step": 34220 + }, + { + "epoch": 2.2, + "learning_rate": 1.2678030872281502e-05, + "loss": 0.8486, + "step": 34221 + }, + { + "epoch": 2.2, + "learning_rate": 1.267781618325855e-05, + "loss": 0.8471, + "step": 34222 + }, + { + "epoch": 2.2, + "learning_rate": 1.2677601494235601e-05, + "loss": 0.9057, + "step": 34223 + }, + { + "epoch": 2.2, + "learning_rate": 1.267738680521265e-05, + "loss": 0.8825, + "step": 34224 + }, + { + "epoch": 2.2, + "learning_rate": 1.26771721161897e-05, + "loss": 0.825, + "step": 34225 + }, + { + "epoch": 2.2, + "learning_rate": 1.267695742716675e-05, + "loss": 0.9018, + "step": 34226 + }, + { + "epoch": 2.2, + "learning_rate": 1.2676742738143799e-05, + "loss": 0.9847, + "step": 34227 + }, + { + "epoch": 2.2, + "learning_rate": 1.2676528049120848e-05, + "loss": 0.88, + "step": 34228 + }, + { + "epoch": 2.2, + "learning_rate": 1.2676313360097899e-05, + "loss": 0.9514, + "step": 34229 + }, + { + "epoch": 2.2, + "learning_rate": 1.267609867107495e-05, + "loss": 0.9973, + "step": 34230 + }, + { + "epoch": 2.2, + "learning_rate": 1.2675883982052e-05, + "loss": 1.0138, + "step": 34231 + }, + { + "epoch": 2.2, + "learning_rate": 1.2675669293029049e-05, + "loss": 0.8182, + "step": 34232 + }, + { + "epoch": 2.2, + "learning_rate": 1.26754546040061e-05, + "loss": 0.937, + "step": 34233 + }, + { + "epoch": 2.2, + "learning_rate": 1.2675239914983149e-05, + "loss": 0.9427, + "step": 34234 + }, + { + "epoch": 2.2, + "learning_rate": 1.2675025225960198e-05, + "loss": 0.8384, + "step": 34235 + }, + { + "epoch": 2.2, + "learning_rate": 1.2674810536937248e-05, + "loss": 0.8845, + "step": 34236 + }, + { + "epoch": 2.2, + "learning_rate": 1.2674595847914297e-05, + "loss": 0.8834, + "step": 34237 + }, + { + "epoch": 2.2, + "learning_rate": 1.2674381158891346e-05, + "loss": 0.9625, + "step": 34238 + }, + { + "epoch": 2.2, + "learning_rate": 1.2674166469868397e-05, + "loss": 0.8926, + "step": 34239 + }, + { + "epoch": 2.2, + "learning_rate": 1.2673951780845446e-05, + "loss": 0.8541, + "step": 34240 + }, + { + "epoch": 2.2, + "learning_rate": 1.2673737091822496e-05, + "loss": 0.8871, + "step": 34241 + }, + { + "epoch": 2.2, + "learning_rate": 1.2673522402799545e-05, + "loss": 0.8785, + "step": 34242 + }, + { + "epoch": 2.2, + "learning_rate": 1.2673307713776594e-05, + "loss": 0.9612, + "step": 34243 + }, + { + "epoch": 2.2, + "learning_rate": 1.2673093024753645e-05, + "loss": 0.8721, + "step": 34244 + }, + { + "epoch": 2.2, + "learning_rate": 1.2672878335730696e-05, + "loss": 0.93, + "step": 34245 + }, + { + "epoch": 2.2, + "learning_rate": 1.2672663646707746e-05, + "loss": 0.9062, + "step": 34246 + }, + { + "epoch": 2.2, + "learning_rate": 1.2672448957684795e-05, + "loss": 0.8721, + "step": 34247 + }, + { + "epoch": 2.2, + "learning_rate": 1.2672234268661844e-05, + "loss": 0.8604, + "step": 34248 + }, + { + "epoch": 2.2, + "learning_rate": 1.2672019579638895e-05, + "loss": 0.8709, + "step": 34249 + }, + { + "epoch": 2.2, + "learning_rate": 1.2671804890615944e-05, + "loss": 0.9165, + "step": 34250 + }, + { + "epoch": 2.2, + "learning_rate": 1.2671590201592994e-05, + "loss": 0.9056, + "step": 34251 + }, + { + "epoch": 2.2, + "learning_rate": 1.2671375512570043e-05, + "loss": 0.831, + "step": 34252 + }, + { + "epoch": 2.2, + "learning_rate": 1.2671160823547092e-05, + "loss": 0.8822, + "step": 34253 + }, + { + "epoch": 2.2, + "learning_rate": 1.2670946134524143e-05, + "loss": 0.8725, + "step": 34254 + }, + { + "epoch": 2.2, + "learning_rate": 1.2670731445501192e-05, + "loss": 0.9687, + "step": 34255 + }, + { + "epoch": 2.2, + "learning_rate": 1.2670516756478241e-05, + "loss": 0.9894, + "step": 34256 + }, + { + "epoch": 2.2, + "learning_rate": 1.2670302067455292e-05, + "loss": 0.9188, + "step": 34257 + }, + { + "epoch": 2.2, + "learning_rate": 1.267008737843234e-05, + "loss": 0.8571, + "step": 34258 + }, + { + "epoch": 2.2, + "learning_rate": 1.2669872689409393e-05, + "loss": 0.8884, + "step": 34259 + }, + { + "epoch": 2.2, + "learning_rate": 1.2669658000386442e-05, + "loss": 0.8236, + "step": 34260 + }, + { + "epoch": 2.2, + "learning_rate": 1.2669443311363493e-05, + "loss": 0.8565, + "step": 34261 + }, + { + "epoch": 2.2, + "learning_rate": 1.2669228622340542e-05, + "loss": 1.0114, + "step": 34262 + }, + { + "epoch": 2.2, + "learning_rate": 1.266901393331759e-05, + "loss": 0.885, + "step": 34263 + }, + { + "epoch": 2.2, + "learning_rate": 1.2668799244294641e-05, + "loss": 0.9188, + "step": 34264 + }, + { + "epoch": 2.2, + "learning_rate": 1.266858455527169e-05, + "loss": 0.8955, + "step": 34265 + }, + { + "epoch": 2.2, + "learning_rate": 1.2668369866248739e-05, + "loss": 0.8708, + "step": 34266 + }, + { + "epoch": 2.2, + "learning_rate": 1.266815517722579e-05, + "loss": 0.9189, + "step": 34267 + }, + { + "epoch": 2.2, + "learning_rate": 1.2667940488202839e-05, + "loss": 0.8911, + "step": 34268 + }, + { + "epoch": 2.2, + "learning_rate": 1.2667725799179888e-05, + "loss": 0.8791, + "step": 34269 + }, + { + "epoch": 2.2, + "learning_rate": 1.2667511110156938e-05, + "loss": 0.879, + "step": 34270 + }, + { + "epoch": 2.2, + "learning_rate": 1.2667296421133987e-05, + "loss": 1.0068, + "step": 34271 + }, + { + "epoch": 2.2, + "learning_rate": 1.2667081732111038e-05, + "loss": 0.927, + "step": 34272 + }, + { + "epoch": 2.21, + "learning_rate": 1.2666867043088087e-05, + "loss": 0.8328, + "step": 34273 + }, + { + "epoch": 2.21, + "learning_rate": 1.266665235406514e-05, + "loss": 0.9542, + "step": 34274 + }, + { + "epoch": 2.21, + "learning_rate": 1.2666437665042188e-05, + "loss": 0.8273, + "step": 34275 + }, + { + "epoch": 2.21, + "learning_rate": 1.2666222976019237e-05, + "loss": 0.8872, + "step": 34276 + }, + { + "epoch": 2.21, + "learning_rate": 1.2666008286996288e-05, + "loss": 0.9139, + "step": 34277 + }, + { + "epoch": 2.21, + "learning_rate": 1.2665793597973337e-05, + "loss": 0.8104, + "step": 34278 + }, + { + "epoch": 2.21, + "learning_rate": 1.2665578908950386e-05, + "loss": 0.8985, + "step": 34279 + }, + { + "epoch": 2.21, + "learning_rate": 1.2665364219927437e-05, + "loss": 0.8538, + "step": 34280 + }, + { + "epoch": 2.21, + "learning_rate": 1.2665149530904485e-05, + "loss": 0.8878, + "step": 34281 + }, + { + "epoch": 2.21, + "learning_rate": 1.2664934841881536e-05, + "loss": 0.8998, + "step": 34282 + }, + { + "epoch": 2.21, + "learning_rate": 1.2664720152858585e-05, + "loss": 0.8043, + "step": 34283 + }, + { + "epoch": 2.21, + "learning_rate": 1.2664505463835634e-05, + "loss": 0.9152, + "step": 34284 + }, + { + "epoch": 2.21, + "learning_rate": 1.2664290774812685e-05, + "loss": 0.84, + "step": 34285 + }, + { + "epoch": 2.21, + "learning_rate": 1.2664076085789734e-05, + "loss": 0.9215, + "step": 34286 + }, + { + "epoch": 2.21, + "learning_rate": 1.2663861396766783e-05, + "loss": 0.8319, + "step": 34287 + }, + { + "epoch": 2.21, + "learning_rate": 1.2663646707743835e-05, + "loss": 0.9202, + "step": 34288 + }, + { + "epoch": 2.21, + "learning_rate": 1.2663432018720884e-05, + "loss": 0.8846, + "step": 34289 + }, + { + "epoch": 2.21, + "learning_rate": 1.2663217329697935e-05, + "loss": 0.8516, + "step": 34290 + }, + { + "epoch": 2.21, + "learning_rate": 1.2663002640674984e-05, + "loss": 0.8617, + "step": 34291 + }, + { + "epoch": 2.21, + "learning_rate": 1.2662787951652034e-05, + "loss": 0.9054, + "step": 34292 + }, + { + "epoch": 2.21, + "learning_rate": 1.2662573262629083e-05, + "loss": 0.9149, + "step": 34293 + }, + { + "epoch": 2.21, + "learning_rate": 1.2662358573606132e-05, + "loss": 0.9538, + "step": 34294 + }, + { + "epoch": 2.21, + "learning_rate": 1.2662143884583183e-05, + "loss": 0.8945, + "step": 34295 + }, + { + "epoch": 2.21, + "learning_rate": 1.2661929195560232e-05, + "loss": 0.9624, + "step": 34296 + }, + { + "epoch": 2.21, + "learning_rate": 1.266171450653728e-05, + "loss": 0.9148, + "step": 34297 + }, + { + "epoch": 2.21, + "learning_rate": 1.2661499817514331e-05, + "loss": 0.9738, + "step": 34298 + }, + { + "epoch": 2.21, + "learning_rate": 1.266128512849138e-05, + "loss": 0.8919, + "step": 34299 + }, + { + "epoch": 2.21, + "learning_rate": 1.2661070439468431e-05, + "loss": 0.9608, + "step": 34300 + }, + { + "epoch": 2.21, + "learning_rate": 1.266085575044548e-05, + "loss": 0.9313, + "step": 34301 + }, + { + "epoch": 2.21, + "learning_rate": 1.2660641061422529e-05, + "loss": 0.9028, + "step": 34302 + }, + { + "epoch": 2.21, + "learning_rate": 1.2660426372399581e-05, + "loss": 0.9334, + "step": 34303 + }, + { + "epoch": 2.21, + "learning_rate": 1.266021168337663e-05, + "loss": 0.9402, + "step": 34304 + }, + { + "epoch": 2.21, + "learning_rate": 1.2659996994353681e-05, + "loss": 0.9157, + "step": 34305 + }, + { + "epoch": 2.21, + "learning_rate": 1.265978230533073e-05, + "loss": 0.9209, + "step": 34306 + }, + { + "epoch": 2.21, + "learning_rate": 1.2659567616307779e-05, + "loss": 0.9272, + "step": 34307 + }, + { + "epoch": 2.21, + "learning_rate": 1.265935292728483e-05, + "loss": 0.8552, + "step": 34308 + }, + { + "epoch": 2.21, + "learning_rate": 1.2659138238261879e-05, + "loss": 0.8763, + "step": 34309 + }, + { + "epoch": 2.21, + "learning_rate": 1.265892354923893e-05, + "loss": 0.8721, + "step": 34310 + }, + { + "epoch": 2.21, + "learning_rate": 1.2658708860215978e-05, + "loss": 0.8935, + "step": 34311 + }, + { + "epoch": 2.21, + "learning_rate": 1.2658494171193027e-05, + "loss": 0.9198, + "step": 34312 + }, + { + "epoch": 2.21, + "learning_rate": 1.2658279482170078e-05, + "loss": 0.9295, + "step": 34313 + }, + { + "epoch": 2.21, + "learning_rate": 1.2658064793147127e-05, + "loss": 0.8214, + "step": 34314 + }, + { + "epoch": 2.21, + "learning_rate": 1.2657850104124176e-05, + "loss": 0.8943, + "step": 34315 + }, + { + "epoch": 2.21, + "learning_rate": 1.2657635415101226e-05, + "loss": 0.8894, + "step": 34316 + }, + { + "epoch": 2.21, + "learning_rate": 1.2657420726078275e-05, + "loss": 0.9805, + "step": 34317 + }, + { + "epoch": 2.21, + "learning_rate": 1.2657206037055328e-05, + "loss": 0.9033, + "step": 34318 + }, + { + "epoch": 2.21, + "learning_rate": 1.2656991348032377e-05, + "loss": 0.8434, + "step": 34319 + }, + { + "epoch": 2.21, + "learning_rate": 1.2656776659009427e-05, + "loss": 0.9018, + "step": 34320 + }, + { + "epoch": 2.21, + "learning_rate": 1.2656561969986476e-05, + "loss": 0.8494, + "step": 34321 + }, + { + "epoch": 2.21, + "learning_rate": 1.2656347280963525e-05, + "loss": 0.9127, + "step": 34322 + }, + { + "epoch": 2.21, + "learning_rate": 1.2656132591940576e-05, + "loss": 0.8979, + "step": 34323 + }, + { + "epoch": 2.21, + "learning_rate": 1.2655917902917625e-05, + "loss": 0.9218, + "step": 34324 + }, + { + "epoch": 2.21, + "learning_rate": 1.2655703213894674e-05, + "loss": 0.9037, + "step": 34325 + }, + { + "epoch": 2.21, + "learning_rate": 1.2655488524871724e-05, + "loss": 0.8159, + "step": 34326 + }, + { + "epoch": 2.21, + "learning_rate": 1.2655273835848773e-05, + "loss": 0.89, + "step": 34327 + }, + { + "epoch": 2.21, + "learning_rate": 1.2655059146825824e-05, + "loss": 0.8514, + "step": 34328 + }, + { + "epoch": 2.21, + "learning_rate": 1.2654844457802873e-05, + "loss": 0.8727, + "step": 34329 + }, + { + "epoch": 2.21, + "learning_rate": 1.2654629768779922e-05, + "loss": 0.9018, + "step": 34330 + }, + { + "epoch": 2.21, + "learning_rate": 1.2654415079756973e-05, + "loss": 0.9025, + "step": 34331 + }, + { + "epoch": 2.21, + "learning_rate": 1.2654200390734023e-05, + "loss": 0.9311, + "step": 34332 + }, + { + "epoch": 2.21, + "learning_rate": 1.2653985701711074e-05, + "loss": 0.9054, + "step": 34333 + }, + { + "epoch": 2.21, + "learning_rate": 1.2653771012688123e-05, + "loss": 0.8656, + "step": 34334 + }, + { + "epoch": 2.21, + "learning_rate": 1.2653556323665172e-05, + "loss": 0.8921, + "step": 34335 + }, + { + "epoch": 2.21, + "learning_rate": 1.2653341634642223e-05, + "loss": 0.8993, + "step": 34336 + }, + { + "epoch": 2.21, + "learning_rate": 1.2653126945619272e-05, + "loss": 0.8336, + "step": 34337 + }, + { + "epoch": 2.21, + "learning_rate": 1.2652912256596322e-05, + "loss": 0.8452, + "step": 34338 + }, + { + "epoch": 2.21, + "learning_rate": 1.2652697567573371e-05, + "loss": 0.8288, + "step": 34339 + }, + { + "epoch": 2.21, + "learning_rate": 1.265248287855042e-05, + "loss": 0.912, + "step": 34340 + }, + { + "epoch": 2.21, + "learning_rate": 1.265226818952747e-05, + "loss": 0.9445, + "step": 34341 + }, + { + "epoch": 2.21, + "learning_rate": 1.265205350050452e-05, + "loss": 0.8499, + "step": 34342 + }, + { + "epoch": 2.21, + "learning_rate": 1.2651838811481569e-05, + "loss": 0.8839, + "step": 34343 + }, + { + "epoch": 2.21, + "learning_rate": 1.265162412245862e-05, + "loss": 0.8869, + "step": 34344 + }, + { + "epoch": 2.21, + "learning_rate": 1.2651409433435668e-05, + "loss": 0.9212, + "step": 34345 + }, + { + "epoch": 2.21, + "learning_rate": 1.2651194744412717e-05, + "loss": 0.8562, + "step": 34346 + }, + { + "epoch": 2.21, + "learning_rate": 1.265098005538977e-05, + "loss": 0.9904, + "step": 34347 + }, + { + "epoch": 2.21, + "learning_rate": 1.265076536636682e-05, + "loss": 0.9597, + "step": 34348 + }, + { + "epoch": 2.21, + "learning_rate": 1.265055067734387e-05, + "loss": 0.9205, + "step": 34349 + }, + { + "epoch": 2.21, + "learning_rate": 1.2650335988320918e-05, + "loss": 1.0146, + "step": 34350 + }, + { + "epoch": 2.21, + "learning_rate": 1.2650121299297969e-05, + "loss": 0.8906, + "step": 34351 + }, + { + "epoch": 2.21, + "learning_rate": 1.2649906610275018e-05, + "loss": 0.9596, + "step": 34352 + }, + { + "epoch": 2.21, + "learning_rate": 1.2649691921252067e-05, + "loss": 0.8737, + "step": 34353 + }, + { + "epoch": 2.21, + "learning_rate": 1.2649477232229118e-05, + "loss": 0.991, + "step": 34354 + }, + { + "epoch": 2.21, + "learning_rate": 1.2649262543206166e-05, + "loss": 0.8993, + "step": 34355 + }, + { + "epoch": 2.21, + "learning_rate": 1.2649047854183215e-05, + "loss": 0.8775, + "step": 34356 + }, + { + "epoch": 2.21, + "learning_rate": 1.2648833165160266e-05, + "loss": 0.8606, + "step": 34357 + }, + { + "epoch": 2.21, + "learning_rate": 1.2648618476137315e-05, + "loss": 0.9229, + "step": 34358 + }, + { + "epoch": 2.21, + "learning_rate": 1.2648403787114366e-05, + "loss": 0.8479, + "step": 34359 + }, + { + "epoch": 2.21, + "learning_rate": 1.2648189098091415e-05, + "loss": 0.8715, + "step": 34360 + }, + { + "epoch": 2.21, + "learning_rate": 1.2647974409068467e-05, + "loss": 0.8348, + "step": 34361 + }, + { + "epoch": 2.21, + "learning_rate": 1.2647759720045516e-05, + "loss": 0.9225, + "step": 34362 + }, + { + "epoch": 2.21, + "learning_rate": 1.2647545031022565e-05, + "loss": 0.9516, + "step": 34363 + }, + { + "epoch": 2.21, + "learning_rate": 1.2647330341999616e-05, + "loss": 0.8639, + "step": 34364 + }, + { + "epoch": 2.21, + "learning_rate": 1.2647115652976665e-05, + "loss": 0.9255, + "step": 34365 + }, + { + "epoch": 2.21, + "learning_rate": 1.2646900963953714e-05, + "loss": 0.8945, + "step": 34366 + }, + { + "epoch": 2.21, + "learning_rate": 1.2646686274930764e-05, + "loss": 0.9119, + "step": 34367 + }, + { + "epoch": 2.21, + "learning_rate": 1.2646471585907813e-05, + "loss": 0.9303, + "step": 34368 + }, + { + "epoch": 2.21, + "learning_rate": 1.2646256896884864e-05, + "loss": 0.9051, + "step": 34369 + }, + { + "epoch": 2.21, + "learning_rate": 1.2646042207861913e-05, + "loss": 0.8619, + "step": 34370 + }, + { + "epoch": 2.21, + "learning_rate": 1.2645827518838962e-05, + "loss": 0.8758, + "step": 34371 + }, + { + "epoch": 2.21, + "learning_rate": 1.2645612829816012e-05, + "loss": 0.8615, + "step": 34372 + }, + { + "epoch": 2.21, + "learning_rate": 1.2645398140793061e-05, + "loss": 0.916, + "step": 34373 + }, + { + "epoch": 2.21, + "learning_rate": 1.264518345177011e-05, + "loss": 0.8596, + "step": 34374 + }, + { + "epoch": 2.21, + "learning_rate": 1.2644968762747161e-05, + "loss": 0.9039, + "step": 34375 + }, + { + "epoch": 2.21, + "learning_rate": 1.2644754073724212e-05, + "loss": 0.8711, + "step": 34376 + }, + { + "epoch": 2.21, + "learning_rate": 1.2644539384701262e-05, + "loss": 0.9413, + "step": 34377 + }, + { + "epoch": 2.21, + "learning_rate": 1.2644324695678311e-05, + "loss": 0.8489, + "step": 34378 + }, + { + "epoch": 2.21, + "learning_rate": 1.2644110006655362e-05, + "loss": 0.8953, + "step": 34379 + }, + { + "epoch": 2.21, + "learning_rate": 1.2643895317632411e-05, + "loss": 0.9172, + "step": 34380 + }, + { + "epoch": 2.21, + "learning_rate": 1.264368062860946e-05, + "loss": 0.9049, + "step": 34381 + }, + { + "epoch": 2.21, + "learning_rate": 1.264346593958651e-05, + "loss": 0.8741, + "step": 34382 + }, + { + "epoch": 2.21, + "learning_rate": 1.264325125056356e-05, + "loss": 0.9208, + "step": 34383 + }, + { + "epoch": 2.21, + "learning_rate": 1.2643036561540608e-05, + "loss": 0.8518, + "step": 34384 + }, + { + "epoch": 2.21, + "learning_rate": 1.2642821872517659e-05, + "loss": 0.8633, + "step": 34385 + }, + { + "epoch": 2.21, + "learning_rate": 1.2642607183494708e-05, + "loss": 0.965, + "step": 34386 + }, + { + "epoch": 2.21, + "learning_rate": 1.2642392494471759e-05, + "loss": 0.8987, + "step": 34387 + }, + { + "epoch": 2.21, + "learning_rate": 1.2642177805448808e-05, + "loss": 0.9398, + "step": 34388 + }, + { + "epoch": 2.21, + "learning_rate": 1.2641963116425857e-05, + "loss": 0.9085, + "step": 34389 + }, + { + "epoch": 2.21, + "learning_rate": 1.2641748427402909e-05, + "loss": 0.8825, + "step": 34390 + }, + { + "epoch": 2.21, + "learning_rate": 1.2641533738379958e-05, + "loss": 0.8435, + "step": 34391 + }, + { + "epoch": 2.21, + "learning_rate": 1.2641319049357009e-05, + "loss": 0.9669, + "step": 34392 + }, + { + "epoch": 2.21, + "learning_rate": 1.2641104360334058e-05, + "loss": 0.9193, + "step": 34393 + }, + { + "epoch": 2.21, + "learning_rate": 1.2640889671311107e-05, + "loss": 0.8543, + "step": 34394 + }, + { + "epoch": 2.21, + "learning_rate": 1.2640674982288157e-05, + "loss": 0.8421, + "step": 34395 + }, + { + "epoch": 2.21, + "learning_rate": 1.2640460293265206e-05, + "loss": 0.8881, + "step": 34396 + }, + { + "epoch": 2.21, + "learning_rate": 1.2640245604242257e-05, + "loss": 0.9056, + "step": 34397 + }, + { + "epoch": 2.21, + "learning_rate": 1.2640030915219306e-05, + "loss": 0.8543, + "step": 34398 + }, + { + "epoch": 2.21, + "learning_rate": 1.2639816226196355e-05, + "loss": 0.9483, + "step": 34399 + }, + { + "epoch": 2.21, + "learning_rate": 1.2639601537173405e-05, + "loss": 0.957, + "step": 34400 + }, + { + "epoch": 2.21, + "learning_rate": 1.2639386848150454e-05, + "loss": 0.9011, + "step": 34401 + }, + { + "epoch": 2.21, + "learning_rate": 1.2639172159127503e-05, + "loss": 0.8302, + "step": 34402 + }, + { + "epoch": 2.21, + "learning_rate": 1.2638957470104554e-05, + "loss": 0.9033, + "step": 34403 + }, + { + "epoch": 2.21, + "learning_rate": 1.2638742781081603e-05, + "loss": 0.8688, + "step": 34404 + }, + { + "epoch": 2.21, + "learning_rate": 1.2638528092058655e-05, + "loss": 0.9106, + "step": 34405 + }, + { + "epoch": 2.21, + "learning_rate": 1.2638313403035704e-05, + "loss": 0.928, + "step": 34406 + }, + { + "epoch": 2.21, + "learning_rate": 1.2638098714012755e-05, + "loss": 0.8621, + "step": 34407 + }, + { + "epoch": 2.21, + "learning_rate": 1.2637884024989804e-05, + "loss": 0.9441, + "step": 34408 + }, + { + "epoch": 2.21, + "learning_rate": 1.2637669335966853e-05, + "loss": 0.8758, + "step": 34409 + }, + { + "epoch": 2.21, + "learning_rate": 1.2637454646943904e-05, + "loss": 0.9523, + "step": 34410 + }, + { + "epoch": 2.21, + "learning_rate": 1.2637239957920953e-05, + "loss": 0.8731, + "step": 34411 + }, + { + "epoch": 2.21, + "learning_rate": 1.2637025268898002e-05, + "loss": 0.8609, + "step": 34412 + }, + { + "epoch": 2.21, + "learning_rate": 1.2636810579875052e-05, + "loss": 0.8662, + "step": 34413 + }, + { + "epoch": 2.21, + "learning_rate": 1.2636595890852101e-05, + "loss": 0.9132, + "step": 34414 + }, + { + "epoch": 2.21, + "learning_rate": 1.263638120182915e-05, + "loss": 0.8998, + "step": 34415 + }, + { + "epoch": 2.21, + "learning_rate": 1.26361665128062e-05, + "loss": 0.9309, + "step": 34416 + }, + { + "epoch": 2.21, + "learning_rate": 1.263595182378325e-05, + "loss": 0.9043, + "step": 34417 + }, + { + "epoch": 2.21, + "learning_rate": 1.26357371347603e-05, + "loss": 0.9421, + "step": 34418 + }, + { + "epoch": 2.21, + "learning_rate": 1.2635522445737351e-05, + "loss": 0.8315, + "step": 34419 + }, + { + "epoch": 2.21, + "learning_rate": 1.2635307756714402e-05, + "loss": 0.9354, + "step": 34420 + }, + { + "epoch": 2.21, + "learning_rate": 1.263509306769145e-05, + "loss": 0.9408, + "step": 34421 + }, + { + "epoch": 2.21, + "learning_rate": 1.26348783786685e-05, + "loss": 0.8866, + "step": 34422 + }, + { + "epoch": 2.21, + "learning_rate": 1.263466368964555e-05, + "loss": 0.7815, + "step": 34423 + }, + { + "epoch": 2.21, + "learning_rate": 1.26344490006226e-05, + "loss": 0.8594, + "step": 34424 + }, + { + "epoch": 2.21, + "learning_rate": 1.263423431159965e-05, + "loss": 0.918, + "step": 34425 + }, + { + "epoch": 2.21, + "learning_rate": 1.2634019622576699e-05, + "loss": 0.8998, + "step": 34426 + }, + { + "epoch": 2.21, + "learning_rate": 1.2633804933553748e-05, + "loss": 0.8949, + "step": 34427 + }, + { + "epoch": 2.22, + "learning_rate": 1.2633590244530799e-05, + "loss": 0.9649, + "step": 34428 + }, + { + "epoch": 2.22, + "learning_rate": 1.2633375555507847e-05, + "loss": 0.8747, + "step": 34429 + }, + { + "epoch": 2.22, + "learning_rate": 1.2633160866484896e-05, + "loss": 0.8751, + "step": 34430 + }, + { + "epoch": 2.22, + "learning_rate": 1.2632946177461947e-05, + "loss": 0.9609, + "step": 34431 + }, + { + "epoch": 2.22, + "learning_rate": 1.2632731488438996e-05, + "loss": 0.9414, + "step": 34432 + }, + { + "epoch": 2.22, + "learning_rate": 1.2632516799416045e-05, + "loss": 0.87, + "step": 34433 + }, + { + "epoch": 2.22, + "learning_rate": 1.2632302110393097e-05, + "loss": 0.912, + "step": 34434 + }, + { + "epoch": 2.22, + "learning_rate": 1.2632087421370148e-05, + "loss": 0.9865, + "step": 34435 + }, + { + "epoch": 2.22, + "learning_rate": 1.2631872732347197e-05, + "loss": 0.8467, + "step": 34436 + }, + { + "epoch": 2.22, + "learning_rate": 1.2631658043324246e-05, + "loss": 0.8694, + "step": 34437 + }, + { + "epoch": 2.22, + "learning_rate": 1.2631443354301297e-05, + "loss": 0.8645, + "step": 34438 + }, + { + "epoch": 2.22, + "learning_rate": 1.2631228665278346e-05, + "loss": 0.9218, + "step": 34439 + }, + { + "epoch": 2.22, + "learning_rate": 1.2631013976255395e-05, + "loss": 0.9192, + "step": 34440 + }, + { + "epoch": 2.22, + "learning_rate": 1.2630799287232445e-05, + "loss": 0.8974, + "step": 34441 + }, + { + "epoch": 2.22, + "learning_rate": 1.2630584598209494e-05, + "loss": 0.9118, + "step": 34442 + }, + { + "epoch": 2.22, + "learning_rate": 1.2630369909186543e-05, + "loss": 0.911, + "step": 34443 + }, + { + "epoch": 2.22, + "learning_rate": 1.2630155220163594e-05, + "loss": 0.8654, + "step": 34444 + }, + { + "epoch": 2.22, + "learning_rate": 1.2629940531140643e-05, + "loss": 0.8965, + "step": 34445 + }, + { + "epoch": 2.22, + "learning_rate": 1.2629725842117693e-05, + "loss": 0.8939, + "step": 34446 + }, + { + "epoch": 2.22, + "learning_rate": 1.2629511153094742e-05, + "loss": 0.8869, + "step": 34447 + }, + { + "epoch": 2.22, + "learning_rate": 1.2629296464071795e-05, + "loss": 0.8687, + "step": 34448 + }, + { + "epoch": 2.22, + "learning_rate": 1.2629081775048844e-05, + "loss": 0.8572, + "step": 34449 + }, + { + "epoch": 2.22, + "learning_rate": 1.2628867086025893e-05, + "loss": 0.8446, + "step": 34450 + }, + { + "epoch": 2.22, + "learning_rate": 1.2628652397002943e-05, + "loss": 0.9153, + "step": 34451 + }, + { + "epoch": 2.22, + "learning_rate": 1.2628437707979992e-05, + "loss": 0.9042, + "step": 34452 + }, + { + "epoch": 2.22, + "learning_rate": 1.2628223018957041e-05, + "loss": 0.8794, + "step": 34453 + }, + { + "epoch": 2.22, + "learning_rate": 1.2628008329934092e-05, + "loss": 0.8564, + "step": 34454 + }, + { + "epoch": 2.22, + "learning_rate": 1.2627793640911141e-05, + "loss": 0.8868, + "step": 34455 + }, + { + "epoch": 2.22, + "learning_rate": 1.2627578951888192e-05, + "loss": 0.8744, + "step": 34456 + }, + { + "epoch": 2.22, + "learning_rate": 1.262736426286524e-05, + "loss": 0.8361, + "step": 34457 + }, + { + "epoch": 2.22, + "learning_rate": 1.262714957384229e-05, + "loss": 0.8821, + "step": 34458 + }, + { + "epoch": 2.22, + "learning_rate": 1.262693488481934e-05, + "loss": 0.8913, + "step": 34459 + }, + { + "epoch": 2.22, + "learning_rate": 1.2626720195796389e-05, + "loss": 0.8195, + "step": 34460 + }, + { + "epoch": 2.22, + "learning_rate": 1.2626505506773438e-05, + "loss": 0.9051, + "step": 34461 + }, + { + "epoch": 2.22, + "learning_rate": 1.2626290817750489e-05, + "loss": 0.9004, + "step": 34462 + }, + { + "epoch": 2.22, + "learning_rate": 1.262607612872754e-05, + "loss": 0.8574, + "step": 34463 + }, + { + "epoch": 2.22, + "learning_rate": 1.262586143970459e-05, + "loss": 0.8719, + "step": 34464 + }, + { + "epoch": 2.22, + "learning_rate": 1.2625646750681639e-05, + "loss": 0.8991, + "step": 34465 + }, + { + "epoch": 2.22, + "learning_rate": 1.262543206165869e-05, + "loss": 0.8622, + "step": 34466 + }, + { + "epoch": 2.22, + "learning_rate": 1.2625217372635739e-05, + "loss": 0.8503, + "step": 34467 + }, + { + "epoch": 2.22, + "learning_rate": 1.2625002683612788e-05, + "loss": 0.8982, + "step": 34468 + }, + { + "epoch": 2.22, + "learning_rate": 1.2624787994589838e-05, + "loss": 0.8962, + "step": 34469 + }, + { + "epoch": 2.22, + "learning_rate": 1.2624573305566887e-05, + "loss": 0.8581, + "step": 34470 + }, + { + "epoch": 2.22, + "learning_rate": 1.2624358616543936e-05, + "loss": 0.9222, + "step": 34471 + }, + { + "epoch": 2.22, + "learning_rate": 1.2624143927520987e-05, + "loss": 0.8349, + "step": 34472 + }, + { + "epoch": 2.22, + "learning_rate": 1.2623929238498036e-05, + "loss": 0.8527, + "step": 34473 + }, + { + "epoch": 2.22, + "learning_rate": 1.2623714549475086e-05, + "loss": 0.8397, + "step": 34474 + }, + { + "epoch": 2.22, + "learning_rate": 1.2623499860452135e-05, + "loss": 0.8135, + "step": 34475 + }, + { + "epoch": 2.22, + "learning_rate": 1.2623285171429184e-05, + "loss": 0.8705, + "step": 34476 + }, + { + "epoch": 2.22, + "learning_rate": 1.2623070482406237e-05, + "loss": 0.7922, + "step": 34477 + }, + { + "epoch": 2.22, + "learning_rate": 1.2622855793383286e-05, + "loss": 0.899, + "step": 34478 + }, + { + "epoch": 2.22, + "learning_rate": 1.2622641104360336e-05, + "loss": 0.967, + "step": 34479 + }, + { + "epoch": 2.22, + "learning_rate": 1.2622426415337385e-05, + "loss": 0.9244, + "step": 34480 + }, + { + "epoch": 2.22, + "learning_rate": 1.2622211726314434e-05, + "loss": 0.9316, + "step": 34481 + }, + { + "epoch": 2.22, + "learning_rate": 1.2621997037291485e-05, + "loss": 0.9474, + "step": 34482 + }, + { + "epoch": 2.22, + "learning_rate": 1.2621782348268534e-05, + "loss": 0.9369, + "step": 34483 + }, + { + "epoch": 2.22, + "learning_rate": 1.2621567659245585e-05, + "loss": 0.8575, + "step": 34484 + }, + { + "epoch": 2.22, + "learning_rate": 1.2621352970222634e-05, + "loss": 0.9068, + "step": 34485 + }, + { + "epoch": 2.22, + "learning_rate": 1.2621138281199683e-05, + "loss": 0.8249, + "step": 34486 + }, + { + "epoch": 2.22, + "learning_rate": 1.2620923592176733e-05, + "loss": 0.8518, + "step": 34487 + }, + { + "epoch": 2.22, + "learning_rate": 1.2620708903153782e-05, + "loss": 0.9632, + "step": 34488 + }, + { + "epoch": 2.22, + "learning_rate": 1.2620494214130831e-05, + "loss": 0.9205, + "step": 34489 + }, + { + "epoch": 2.22, + "learning_rate": 1.2620279525107882e-05, + "loss": 0.8889, + "step": 34490 + }, + { + "epoch": 2.22, + "learning_rate": 1.262006483608493e-05, + "loss": 0.8879, + "step": 34491 + }, + { + "epoch": 2.22, + "learning_rate": 1.2619850147061983e-05, + "loss": 0.9476, + "step": 34492 + }, + { + "epoch": 2.22, + "learning_rate": 1.2619635458039032e-05, + "loss": 0.9224, + "step": 34493 + }, + { + "epoch": 2.22, + "learning_rate": 1.2619420769016083e-05, + "loss": 0.9123, + "step": 34494 + }, + { + "epoch": 2.22, + "learning_rate": 1.2619206079993132e-05, + "loss": 0.8794, + "step": 34495 + }, + { + "epoch": 2.22, + "learning_rate": 1.261899139097018e-05, + "loss": 0.9015, + "step": 34496 + }, + { + "epoch": 2.22, + "learning_rate": 1.2618776701947231e-05, + "loss": 0.8551, + "step": 34497 + }, + { + "epoch": 2.22, + "learning_rate": 1.261856201292428e-05, + "loss": 0.8809, + "step": 34498 + }, + { + "epoch": 2.22, + "learning_rate": 1.261834732390133e-05, + "loss": 0.8748, + "step": 34499 + }, + { + "epoch": 2.22, + "learning_rate": 1.261813263487838e-05, + "loss": 0.965, + "step": 34500 + }, + { + "epoch": 2.22, + "learning_rate": 1.2617917945855429e-05, + "loss": 0.8882, + "step": 34501 + }, + { + "epoch": 2.22, + "learning_rate": 1.2617703256832478e-05, + "loss": 0.9428, + "step": 34502 + }, + { + "epoch": 2.22, + "learning_rate": 1.2617488567809528e-05, + "loss": 0.8422, + "step": 34503 + }, + { + "epoch": 2.22, + "learning_rate": 1.2617273878786577e-05, + "loss": 0.8835, + "step": 34504 + }, + { + "epoch": 2.22, + "learning_rate": 1.2617059189763628e-05, + "loss": 0.8545, + "step": 34505 + }, + { + "epoch": 2.22, + "learning_rate": 1.2616844500740679e-05, + "loss": 0.9361, + "step": 34506 + }, + { + "epoch": 2.22, + "learning_rate": 1.261662981171773e-05, + "loss": 0.9404, + "step": 34507 + }, + { + "epoch": 2.22, + "learning_rate": 1.2616415122694778e-05, + "loss": 0.871, + "step": 34508 + }, + { + "epoch": 2.22, + "learning_rate": 1.2616200433671827e-05, + "loss": 0.9627, + "step": 34509 + }, + { + "epoch": 2.22, + "learning_rate": 1.2615985744648878e-05, + "loss": 0.9418, + "step": 34510 + }, + { + "epoch": 2.22, + "learning_rate": 1.2615771055625927e-05, + "loss": 0.8752, + "step": 34511 + }, + { + "epoch": 2.22, + "learning_rate": 1.2615556366602976e-05, + "loss": 0.8868, + "step": 34512 + }, + { + "epoch": 2.22, + "learning_rate": 1.2615341677580027e-05, + "loss": 0.8789, + "step": 34513 + }, + { + "epoch": 2.22, + "learning_rate": 1.2615126988557076e-05, + "loss": 0.8651, + "step": 34514 + }, + { + "epoch": 2.22, + "learning_rate": 1.2614912299534126e-05, + "loss": 0.867, + "step": 34515 + }, + { + "epoch": 2.22, + "learning_rate": 1.2614697610511175e-05, + "loss": 0.9154, + "step": 34516 + }, + { + "epoch": 2.22, + "learning_rate": 1.2614482921488224e-05, + "loss": 0.9081, + "step": 34517 + }, + { + "epoch": 2.22, + "learning_rate": 1.2614268232465275e-05, + "loss": 0.8413, + "step": 34518 + }, + { + "epoch": 2.22, + "learning_rate": 1.2614053543442324e-05, + "loss": 0.9187, + "step": 34519 + }, + { + "epoch": 2.22, + "learning_rate": 1.2613838854419373e-05, + "loss": 0.8599, + "step": 34520 + }, + { + "epoch": 2.22, + "learning_rate": 1.2613624165396425e-05, + "loss": 0.9191, + "step": 34521 + }, + { + "epoch": 2.22, + "learning_rate": 1.2613409476373474e-05, + "loss": 0.8538, + "step": 34522 + }, + { + "epoch": 2.22, + "learning_rate": 1.2613194787350525e-05, + "loss": 0.8851, + "step": 34523 + }, + { + "epoch": 2.22, + "learning_rate": 1.2612980098327574e-05, + "loss": 0.9002, + "step": 34524 + }, + { + "epoch": 2.22, + "learning_rate": 1.2612980098327574e-05, + "loss": 0.8514, + "step": 34525 + }, + { + "epoch": 2.22, + "learning_rate": 1.2612765409304624e-05, + "loss": 0.88, + "step": 34526 + }, + { + "epoch": 2.22, + "learning_rate": 1.2612550720281673e-05, + "loss": 0.8493, + "step": 34527 + }, + { + "epoch": 2.22, + "learning_rate": 1.2612336031258722e-05, + "loss": 0.9402, + "step": 34528 + }, + { + "epoch": 2.22, + "learning_rate": 1.2612121342235773e-05, + "loss": 0.8584, + "step": 34529 + }, + { + "epoch": 2.22, + "learning_rate": 1.2611906653212822e-05, + "loss": 0.8728, + "step": 34530 + }, + { + "epoch": 2.22, + "learning_rate": 1.2611691964189871e-05, + "loss": 0.9266, + "step": 34531 + }, + { + "epoch": 2.22, + "learning_rate": 1.2611477275166922e-05, + "loss": 0.8607, + "step": 34532 + }, + { + "epoch": 2.22, + "learning_rate": 1.261126258614397e-05, + "loss": 0.88, + "step": 34533 + }, + { + "epoch": 2.22, + "learning_rate": 1.2611047897121021e-05, + "loss": 0.8687, + "step": 34534 + }, + { + "epoch": 2.22, + "learning_rate": 1.261083320809807e-05, + "loss": 0.8662, + "step": 34535 + }, + { + "epoch": 2.22, + "learning_rate": 1.2610618519075122e-05, + "loss": 0.9012, + "step": 34536 + }, + { + "epoch": 2.22, + "learning_rate": 1.2610403830052171e-05, + "loss": 0.8837, + "step": 34537 + }, + { + "epoch": 2.22, + "learning_rate": 1.261018914102922e-05, + "loss": 0.8435, + "step": 34538 + }, + { + "epoch": 2.22, + "learning_rate": 1.2609974452006271e-05, + "loss": 0.8701, + "step": 34539 + }, + { + "epoch": 2.22, + "learning_rate": 1.260975976298332e-05, + "loss": 0.893, + "step": 34540 + }, + { + "epoch": 2.22, + "learning_rate": 1.2609545073960369e-05, + "loss": 0.9322, + "step": 34541 + }, + { + "epoch": 2.22, + "learning_rate": 1.260933038493742e-05, + "loss": 0.9048, + "step": 34542 + }, + { + "epoch": 2.22, + "learning_rate": 1.2609115695914469e-05, + "loss": 0.9265, + "step": 34543 + }, + { + "epoch": 2.22, + "learning_rate": 1.260890100689152e-05, + "loss": 0.8634, + "step": 34544 + }, + { + "epoch": 2.22, + "learning_rate": 1.2608686317868568e-05, + "loss": 0.9477, + "step": 34545 + }, + { + "epoch": 2.22, + "learning_rate": 1.2608471628845617e-05, + "loss": 0.8507, + "step": 34546 + }, + { + "epoch": 2.22, + "learning_rate": 1.2608256939822668e-05, + "loss": 0.8478, + "step": 34547 + }, + { + "epoch": 2.22, + "learning_rate": 1.2608042250799717e-05, + "loss": 0.9074, + "step": 34548 + }, + { + "epoch": 2.22, + "learning_rate": 1.2607827561776766e-05, + "loss": 0.9076, + "step": 34549 + }, + { + "epoch": 2.22, + "learning_rate": 1.2607612872753816e-05, + "loss": 0.9112, + "step": 34550 + }, + { + "epoch": 2.22, + "learning_rate": 1.2607398183730867e-05, + "loss": 0.9557, + "step": 34551 + }, + { + "epoch": 2.22, + "learning_rate": 1.2607183494707918e-05, + "loss": 0.945, + "step": 34552 + }, + { + "epoch": 2.22, + "learning_rate": 1.2606968805684967e-05, + "loss": 0.8645, + "step": 34553 + }, + { + "epoch": 2.22, + "learning_rate": 1.2606754116662017e-05, + "loss": 0.871, + "step": 34554 + }, + { + "epoch": 2.22, + "learning_rate": 1.2606539427639066e-05, + "loss": 0.8745, + "step": 34555 + }, + { + "epoch": 2.22, + "learning_rate": 1.2606324738616115e-05, + "loss": 0.9038, + "step": 34556 + }, + { + "epoch": 2.22, + "learning_rate": 1.2606110049593166e-05, + "loss": 0.8817, + "step": 34557 + }, + { + "epoch": 2.22, + "learning_rate": 1.2605895360570215e-05, + "loss": 0.8296, + "step": 34558 + }, + { + "epoch": 2.22, + "learning_rate": 1.2605680671547264e-05, + "loss": 0.9405, + "step": 34559 + }, + { + "epoch": 2.22, + "learning_rate": 1.2605465982524315e-05, + "loss": 0.9299, + "step": 34560 + }, + { + "epoch": 2.22, + "learning_rate": 1.2605251293501364e-05, + "loss": 0.9335, + "step": 34561 + }, + { + "epoch": 2.22, + "learning_rate": 1.2605036604478414e-05, + "loss": 0.8741, + "step": 34562 + }, + { + "epoch": 2.22, + "learning_rate": 1.2604821915455463e-05, + "loss": 0.9041, + "step": 34563 + }, + { + "epoch": 2.22, + "learning_rate": 1.2604607226432512e-05, + "loss": 0.8607, + "step": 34564 + }, + { + "epoch": 2.22, + "learning_rate": 1.2604392537409564e-05, + "loss": 0.8981, + "step": 34565 + }, + { + "epoch": 2.22, + "learning_rate": 1.2604177848386613e-05, + "loss": 0.862, + "step": 34566 + }, + { + "epoch": 2.22, + "learning_rate": 1.2603963159363664e-05, + "loss": 0.8363, + "step": 34567 + }, + { + "epoch": 2.22, + "learning_rate": 1.2603748470340713e-05, + "loss": 0.8568, + "step": 34568 + }, + { + "epoch": 2.22, + "learning_rate": 1.2603533781317762e-05, + "loss": 0.9615, + "step": 34569 + }, + { + "epoch": 2.22, + "learning_rate": 1.2603319092294813e-05, + "loss": 0.8881, + "step": 34570 + }, + { + "epoch": 2.22, + "learning_rate": 1.2603104403271862e-05, + "loss": 0.8956, + "step": 34571 + }, + { + "epoch": 2.22, + "learning_rate": 1.2602889714248912e-05, + "loss": 1.0111, + "step": 34572 + }, + { + "epoch": 2.22, + "learning_rate": 1.2602675025225961e-05, + "loss": 0.8743, + "step": 34573 + }, + { + "epoch": 2.22, + "learning_rate": 1.260246033620301e-05, + "loss": 0.8731, + "step": 34574 + }, + { + "epoch": 2.22, + "learning_rate": 1.2602245647180061e-05, + "loss": 0.9042, + "step": 34575 + }, + { + "epoch": 2.22, + "learning_rate": 1.260203095815711e-05, + "loss": 0.9614, + "step": 34576 + }, + { + "epoch": 2.22, + "learning_rate": 1.2601816269134159e-05, + "loss": 0.9256, + "step": 34577 + }, + { + "epoch": 2.22, + "learning_rate": 1.260160158011121e-05, + "loss": 0.8382, + "step": 34578 + }, + { + "epoch": 2.22, + "learning_rate": 1.2601386891088258e-05, + "loss": 0.9744, + "step": 34579 + }, + { + "epoch": 2.22, + "learning_rate": 1.260117220206531e-05, + "loss": 0.9108, + "step": 34580 + }, + { + "epoch": 2.22, + "learning_rate": 1.260095751304236e-05, + "loss": 0.9218, + "step": 34581 + }, + { + "epoch": 2.22, + "learning_rate": 1.260074282401941e-05, + "loss": 0.8992, + "step": 34582 + }, + { + "epoch": 2.22, + "learning_rate": 1.260052813499646e-05, + "loss": 0.8946, + "step": 34583 + }, + { + "epoch": 2.23, + "learning_rate": 1.2600313445973508e-05, + "loss": 0.8753, + "step": 34584 + }, + { + "epoch": 2.23, + "learning_rate": 1.2600098756950559e-05, + "loss": 0.8818, + "step": 34585 + }, + { + "epoch": 2.23, + "learning_rate": 1.2599884067927608e-05, + "loss": 0.8763, + "step": 34586 + }, + { + "epoch": 2.23, + "learning_rate": 1.2599669378904657e-05, + "loss": 0.8843, + "step": 34587 + }, + { + "epoch": 2.23, + "learning_rate": 1.2599454689881708e-05, + "loss": 0.8826, + "step": 34588 + }, + { + "epoch": 2.23, + "learning_rate": 1.2599240000858757e-05, + "loss": 0.9215, + "step": 34589 + }, + { + "epoch": 2.23, + "learning_rate": 1.2599025311835806e-05, + "loss": 0.9204, + "step": 34590 + }, + { + "epoch": 2.23, + "learning_rate": 1.2598810622812856e-05, + "loss": 0.9141, + "step": 34591 + }, + { + "epoch": 2.23, + "learning_rate": 1.2598595933789905e-05, + "loss": 0.859, + "step": 34592 + }, + { + "epoch": 2.23, + "learning_rate": 1.2598381244766956e-05, + "loss": 0.8846, + "step": 34593 + }, + { + "epoch": 2.23, + "learning_rate": 1.2598166555744006e-05, + "loss": 0.8761, + "step": 34594 + }, + { + "epoch": 2.23, + "learning_rate": 1.2597951866721057e-05, + "loss": 0.8707, + "step": 34595 + }, + { + "epoch": 2.23, + "learning_rate": 1.2597737177698106e-05, + "loss": 0.9322, + "step": 34596 + }, + { + "epoch": 2.23, + "learning_rate": 1.2597522488675155e-05, + "loss": 0.9372, + "step": 34597 + }, + { + "epoch": 2.23, + "learning_rate": 1.2597307799652206e-05, + "loss": 0.8779, + "step": 34598 + }, + { + "epoch": 2.23, + "learning_rate": 1.2597093110629255e-05, + "loss": 0.9046, + "step": 34599 + }, + { + "epoch": 2.23, + "learning_rate": 1.2596878421606304e-05, + "loss": 0.8334, + "step": 34600 + }, + { + "epoch": 2.23, + "learning_rate": 1.2596663732583354e-05, + "loss": 0.939, + "step": 34601 + }, + { + "epoch": 2.23, + "learning_rate": 1.2596449043560403e-05, + "loss": 0.8529, + "step": 34602 + }, + { + "epoch": 2.23, + "learning_rate": 1.2596234354537454e-05, + "loss": 0.8508, + "step": 34603 + }, + { + "epoch": 2.23, + "learning_rate": 1.2596019665514503e-05, + "loss": 0.8962, + "step": 34604 + }, + { + "epoch": 2.23, + "learning_rate": 1.2595804976491552e-05, + "loss": 0.9255, + "step": 34605 + }, + { + "epoch": 2.23, + "learning_rate": 1.2595590287468603e-05, + "loss": 0.9236, + "step": 34606 + }, + { + "epoch": 2.23, + "learning_rate": 1.2595375598445651e-05, + "loss": 0.8735, + "step": 34607 + }, + { + "epoch": 2.23, + "learning_rate": 1.25951609094227e-05, + "loss": 0.9181, + "step": 34608 + }, + { + "epoch": 2.23, + "learning_rate": 1.2594946220399753e-05, + "loss": 0.9517, + "step": 34609 + }, + { + "epoch": 2.23, + "learning_rate": 1.2594731531376802e-05, + "loss": 0.8929, + "step": 34610 + }, + { + "epoch": 2.23, + "learning_rate": 1.2594516842353852e-05, + "loss": 0.9499, + "step": 34611 + }, + { + "epoch": 2.23, + "learning_rate": 1.2594302153330901e-05, + "loss": 0.8638, + "step": 34612 + }, + { + "epoch": 2.23, + "learning_rate": 1.2594087464307952e-05, + "loss": 0.9651, + "step": 34613 + }, + { + "epoch": 2.23, + "learning_rate": 1.2593872775285001e-05, + "loss": 0.8972, + "step": 34614 + }, + { + "epoch": 2.23, + "learning_rate": 1.259365808626205e-05, + "loss": 0.8456, + "step": 34615 + }, + { + "epoch": 2.23, + "learning_rate": 1.25934433972391e-05, + "loss": 0.8819, + "step": 34616 + }, + { + "epoch": 2.23, + "learning_rate": 1.259322870821615e-05, + "loss": 0.8933, + "step": 34617 + }, + { + "epoch": 2.23, + "learning_rate": 1.2593014019193199e-05, + "loss": 0.8851, + "step": 34618 + }, + { + "epoch": 2.23, + "learning_rate": 1.259279933017025e-05, + "loss": 0.8794, + "step": 34619 + }, + { + "epoch": 2.23, + "learning_rate": 1.2592584641147298e-05, + "loss": 0.8863, + "step": 34620 + }, + { + "epoch": 2.23, + "learning_rate": 1.2592369952124349e-05, + "loss": 1.0454, + "step": 34621 + }, + { + "epoch": 2.23, + "learning_rate": 1.2592155263101398e-05, + "loss": 0.901, + "step": 34622 + }, + { + "epoch": 2.23, + "learning_rate": 1.259194057407845e-05, + "loss": 0.8828, + "step": 34623 + }, + { + "epoch": 2.23, + "learning_rate": 1.2591725885055499e-05, + "loss": 0.8506, + "step": 34624 + }, + { + "epoch": 2.23, + "learning_rate": 1.2591511196032548e-05, + "loss": 0.9442, + "step": 34625 + }, + { + "epoch": 2.23, + "learning_rate": 1.2591296507009599e-05, + "loss": 0.8948, + "step": 34626 + }, + { + "epoch": 2.23, + "learning_rate": 1.2591081817986648e-05, + "loss": 0.942, + "step": 34627 + }, + { + "epoch": 2.23, + "learning_rate": 1.2590867128963697e-05, + "loss": 0.8918, + "step": 34628 + }, + { + "epoch": 2.23, + "learning_rate": 1.2590652439940747e-05, + "loss": 0.8699, + "step": 34629 + }, + { + "epoch": 2.23, + "learning_rate": 1.2590437750917796e-05, + "loss": 0.8556, + "step": 34630 + }, + { + "epoch": 2.23, + "learning_rate": 1.2590223061894847e-05, + "loss": 0.9599, + "step": 34631 + }, + { + "epoch": 2.23, + "learning_rate": 1.2590008372871896e-05, + "loss": 0.9364, + "step": 34632 + }, + { + "epoch": 2.23, + "learning_rate": 1.2589793683848945e-05, + "loss": 0.9002, + "step": 34633 + }, + { + "epoch": 2.23, + "learning_rate": 1.2589578994825996e-05, + "loss": 0.9089, + "step": 34634 + }, + { + "epoch": 2.23, + "learning_rate": 1.2589364305803045e-05, + "loss": 0.8923, + "step": 34635 + }, + { + "epoch": 2.23, + "learning_rate": 1.2589149616780093e-05, + "loss": 0.9171, + "step": 34636 + }, + { + "epoch": 2.23, + "learning_rate": 1.2588934927757144e-05, + "loss": 0.9285, + "step": 34637 + }, + { + "epoch": 2.23, + "learning_rate": 1.2588720238734195e-05, + "loss": 0.8655, + "step": 34638 + }, + { + "epoch": 2.23, + "learning_rate": 1.2588505549711245e-05, + "loss": 0.8789, + "step": 34639 + }, + { + "epoch": 2.23, + "learning_rate": 1.2588290860688294e-05, + "loss": 0.9385, + "step": 34640 + }, + { + "epoch": 2.23, + "learning_rate": 1.2588076171665345e-05, + "loss": 0.9032, + "step": 34641 + }, + { + "epoch": 2.23, + "learning_rate": 1.2587861482642394e-05, + "loss": 0.8481, + "step": 34642 + }, + { + "epoch": 2.23, + "learning_rate": 1.2587646793619443e-05, + "loss": 0.8924, + "step": 34643 + }, + { + "epoch": 2.23, + "learning_rate": 1.2587432104596494e-05, + "loss": 0.889, + "step": 34644 + }, + { + "epoch": 2.23, + "learning_rate": 1.2587217415573543e-05, + "loss": 0.8766, + "step": 34645 + }, + { + "epoch": 2.23, + "learning_rate": 1.2587002726550592e-05, + "loss": 0.8575, + "step": 34646 + }, + { + "epoch": 2.23, + "learning_rate": 1.2586788037527642e-05, + "loss": 0.8888, + "step": 34647 + }, + { + "epoch": 2.23, + "learning_rate": 1.2586573348504691e-05, + "loss": 0.8914, + "step": 34648 + }, + { + "epoch": 2.23, + "learning_rate": 1.2586358659481742e-05, + "loss": 0.8516, + "step": 34649 + }, + { + "epoch": 2.23, + "learning_rate": 1.2586143970458791e-05, + "loss": 0.8734, + "step": 34650 + }, + { + "epoch": 2.23, + "learning_rate": 1.258592928143584e-05, + "loss": 0.8859, + "step": 34651 + }, + { + "epoch": 2.23, + "learning_rate": 1.2585714592412892e-05, + "loss": 0.8578, + "step": 34652 + }, + { + "epoch": 2.23, + "learning_rate": 1.2585499903389941e-05, + "loss": 0.9021, + "step": 34653 + }, + { + "epoch": 2.23, + "learning_rate": 1.2585285214366992e-05, + "loss": 0.9278, + "step": 34654 + }, + { + "epoch": 2.23, + "learning_rate": 1.258507052534404e-05, + "loss": 0.9157, + "step": 34655 + }, + { + "epoch": 2.23, + "learning_rate": 1.258485583632109e-05, + "loss": 0.8665, + "step": 34656 + }, + { + "epoch": 2.23, + "learning_rate": 1.258464114729814e-05, + "loss": 0.8771, + "step": 34657 + }, + { + "epoch": 2.23, + "learning_rate": 1.258442645827519e-05, + "loss": 0.9188, + "step": 34658 + }, + { + "epoch": 2.23, + "learning_rate": 1.258421176925224e-05, + "loss": 0.9232, + "step": 34659 + }, + { + "epoch": 2.23, + "learning_rate": 1.2583997080229289e-05, + "loss": 0.833, + "step": 34660 + }, + { + "epoch": 2.23, + "learning_rate": 1.2583782391206338e-05, + "loss": 0.937, + "step": 34661 + }, + { + "epoch": 2.23, + "learning_rate": 1.2583567702183389e-05, + "loss": 0.8982, + "step": 34662 + }, + { + "epoch": 2.23, + "learning_rate": 1.2583353013160438e-05, + "loss": 0.8819, + "step": 34663 + }, + { + "epoch": 2.23, + "learning_rate": 1.2583138324137487e-05, + "loss": 0.9003, + "step": 34664 + }, + { + "epoch": 2.23, + "learning_rate": 1.2582923635114537e-05, + "loss": 0.8919, + "step": 34665 + }, + { + "epoch": 2.23, + "learning_rate": 1.2582708946091586e-05, + "loss": 0.8762, + "step": 34666 + }, + { + "epoch": 2.23, + "learning_rate": 1.2582494257068639e-05, + "loss": 0.9056, + "step": 34667 + }, + { + "epoch": 2.23, + "learning_rate": 1.2582279568045687e-05, + "loss": 0.8825, + "step": 34668 + }, + { + "epoch": 2.23, + "learning_rate": 1.2582064879022738e-05, + "loss": 0.9485, + "step": 34669 + }, + { + "epoch": 2.23, + "learning_rate": 1.2581850189999787e-05, + "loss": 0.9819, + "step": 34670 + }, + { + "epoch": 2.23, + "learning_rate": 1.2581635500976836e-05, + "loss": 0.9025, + "step": 34671 + }, + { + "epoch": 2.23, + "learning_rate": 1.2581420811953887e-05, + "loss": 0.8382, + "step": 34672 + }, + { + "epoch": 2.23, + "learning_rate": 1.2581206122930936e-05, + "loss": 0.9084, + "step": 34673 + }, + { + "epoch": 2.23, + "learning_rate": 1.2580991433907985e-05, + "loss": 0.9147, + "step": 34674 + }, + { + "epoch": 2.23, + "learning_rate": 1.2580776744885035e-05, + "loss": 0.9394, + "step": 34675 + }, + { + "epoch": 2.23, + "learning_rate": 1.2580562055862084e-05, + "loss": 0.9092, + "step": 34676 + }, + { + "epoch": 2.23, + "learning_rate": 1.2580347366839133e-05, + "loss": 0.8991, + "step": 34677 + }, + { + "epoch": 2.23, + "learning_rate": 1.2580132677816184e-05, + "loss": 0.9572, + "step": 34678 + }, + { + "epoch": 2.23, + "learning_rate": 1.2579917988793233e-05, + "loss": 0.9705, + "step": 34679 + }, + { + "epoch": 2.23, + "learning_rate": 1.2579703299770284e-05, + "loss": 0.8406, + "step": 34680 + }, + { + "epoch": 2.23, + "learning_rate": 1.2579488610747334e-05, + "loss": 0.9421, + "step": 34681 + }, + { + "epoch": 2.23, + "learning_rate": 1.2579273921724385e-05, + "loss": 0.9166, + "step": 34682 + }, + { + "epoch": 2.23, + "learning_rate": 1.2579059232701434e-05, + "loss": 0.8744, + "step": 34683 + }, + { + "epoch": 2.23, + "learning_rate": 1.2578844543678483e-05, + "loss": 0.8957, + "step": 34684 + }, + { + "epoch": 2.23, + "learning_rate": 1.2578629854655533e-05, + "loss": 0.9043, + "step": 34685 + }, + { + "epoch": 2.23, + "learning_rate": 1.2578415165632582e-05, + "loss": 0.8697, + "step": 34686 + }, + { + "epoch": 2.23, + "learning_rate": 1.2578200476609631e-05, + "loss": 0.9645, + "step": 34687 + }, + { + "epoch": 2.23, + "learning_rate": 1.2577985787586682e-05, + "loss": 0.9047, + "step": 34688 + }, + { + "epoch": 2.23, + "learning_rate": 1.2577771098563731e-05, + "loss": 0.9392, + "step": 34689 + }, + { + "epoch": 2.23, + "learning_rate": 1.2577556409540782e-05, + "loss": 0.8946, + "step": 34690 + }, + { + "epoch": 2.23, + "learning_rate": 1.257734172051783e-05, + "loss": 0.8904, + "step": 34691 + }, + { + "epoch": 2.23, + "learning_rate": 1.257712703149488e-05, + "loss": 0.9179, + "step": 34692 + }, + { + "epoch": 2.23, + "learning_rate": 1.257691234247193e-05, + "loss": 0.9226, + "step": 34693 + }, + { + "epoch": 2.23, + "learning_rate": 1.257669765344898e-05, + "loss": 0.9236, + "step": 34694 + }, + { + "epoch": 2.23, + "learning_rate": 1.2576482964426028e-05, + "loss": 0.9251, + "step": 34695 + }, + { + "epoch": 2.23, + "learning_rate": 1.257626827540308e-05, + "loss": 0.8879, + "step": 34696 + }, + { + "epoch": 2.23, + "learning_rate": 1.257605358638013e-05, + "loss": 0.8289, + "step": 34697 + }, + { + "epoch": 2.23, + "learning_rate": 1.257583889735718e-05, + "loss": 0.9551, + "step": 34698 + }, + { + "epoch": 2.23, + "learning_rate": 1.2575624208334229e-05, + "loss": 0.8869, + "step": 34699 + }, + { + "epoch": 2.23, + "learning_rate": 1.257540951931128e-05, + "loss": 0.8992, + "step": 34700 + }, + { + "epoch": 2.23, + "learning_rate": 1.2575194830288329e-05, + "loss": 0.958, + "step": 34701 + }, + { + "epoch": 2.23, + "learning_rate": 1.2574980141265378e-05, + "loss": 0.8296, + "step": 34702 + }, + { + "epoch": 2.23, + "learning_rate": 1.2574765452242428e-05, + "loss": 0.9273, + "step": 34703 + }, + { + "epoch": 2.23, + "learning_rate": 1.2574550763219477e-05, + "loss": 0.8781, + "step": 34704 + }, + { + "epoch": 2.23, + "learning_rate": 1.2574336074196526e-05, + "loss": 0.8717, + "step": 34705 + }, + { + "epoch": 2.23, + "learning_rate": 1.2574121385173577e-05, + "loss": 0.8362, + "step": 34706 + }, + { + "epoch": 2.23, + "learning_rate": 1.2573906696150626e-05, + "loss": 0.9381, + "step": 34707 + }, + { + "epoch": 2.23, + "learning_rate": 1.2573692007127677e-05, + "loss": 0.8789, + "step": 34708 + }, + { + "epoch": 2.23, + "learning_rate": 1.2573477318104726e-05, + "loss": 0.9078, + "step": 34709 + }, + { + "epoch": 2.23, + "learning_rate": 1.2573262629081774e-05, + "loss": 0.8783, + "step": 34710 + }, + { + "epoch": 2.23, + "learning_rate": 1.2573047940058827e-05, + "loss": 0.8847, + "step": 34711 + }, + { + "epoch": 2.23, + "learning_rate": 1.2572833251035876e-05, + "loss": 0.8927, + "step": 34712 + }, + { + "epoch": 2.23, + "learning_rate": 1.2572618562012926e-05, + "loss": 0.9564, + "step": 34713 + }, + { + "epoch": 2.23, + "learning_rate": 1.2572403872989975e-05, + "loss": 0.9612, + "step": 34714 + }, + { + "epoch": 2.23, + "learning_rate": 1.2572189183967024e-05, + "loss": 0.9292, + "step": 34715 + }, + { + "epoch": 2.23, + "learning_rate": 1.2571974494944075e-05, + "loss": 0.8689, + "step": 34716 + }, + { + "epoch": 2.23, + "learning_rate": 1.2571759805921124e-05, + "loss": 0.8579, + "step": 34717 + }, + { + "epoch": 2.23, + "learning_rate": 1.2571545116898175e-05, + "loss": 0.9269, + "step": 34718 + }, + { + "epoch": 2.23, + "learning_rate": 1.2571330427875224e-05, + "loss": 0.8966, + "step": 34719 + }, + { + "epoch": 2.23, + "learning_rate": 1.2571115738852273e-05, + "loss": 0.9532, + "step": 34720 + }, + { + "epoch": 2.23, + "learning_rate": 1.2570901049829323e-05, + "loss": 0.8862, + "step": 34721 + }, + { + "epoch": 2.23, + "learning_rate": 1.2570686360806372e-05, + "loss": 0.8834, + "step": 34722 + }, + { + "epoch": 2.23, + "learning_rate": 1.2570471671783421e-05, + "loss": 0.8835, + "step": 34723 + }, + { + "epoch": 2.23, + "learning_rate": 1.2570256982760472e-05, + "loss": 0.894, + "step": 34724 + }, + { + "epoch": 2.23, + "learning_rate": 1.2570042293737523e-05, + "loss": 0.9016, + "step": 34725 + }, + { + "epoch": 2.23, + "learning_rate": 1.2569827604714573e-05, + "loss": 0.8985, + "step": 34726 + }, + { + "epoch": 2.23, + "learning_rate": 1.2569612915691622e-05, + "loss": 0.8756, + "step": 34727 + }, + { + "epoch": 2.23, + "learning_rate": 1.2569398226668673e-05, + "loss": 0.8706, + "step": 34728 + }, + { + "epoch": 2.23, + "learning_rate": 1.2569183537645722e-05, + "loss": 0.9233, + "step": 34729 + }, + { + "epoch": 2.23, + "learning_rate": 1.256896884862277e-05, + "loss": 0.8639, + "step": 34730 + }, + { + "epoch": 2.23, + "learning_rate": 1.2568754159599821e-05, + "loss": 0.8484, + "step": 34731 + }, + { + "epoch": 2.23, + "learning_rate": 1.256853947057687e-05, + "loss": 0.9078, + "step": 34732 + }, + { + "epoch": 2.23, + "learning_rate": 1.256832478155392e-05, + "loss": 0.9002, + "step": 34733 + }, + { + "epoch": 2.23, + "learning_rate": 1.256811009253097e-05, + "loss": 0.8461, + "step": 34734 + }, + { + "epoch": 2.23, + "learning_rate": 1.2567895403508019e-05, + "loss": 0.9119, + "step": 34735 + }, + { + "epoch": 2.23, + "learning_rate": 1.256768071448507e-05, + "loss": 0.9184, + "step": 34736 + }, + { + "epoch": 2.23, + "learning_rate": 1.2567466025462119e-05, + "loss": 0.8922, + "step": 34737 + }, + { + "epoch": 2.23, + "learning_rate": 1.2567251336439168e-05, + "loss": 0.9039, + "step": 34738 + }, + { + "epoch": 2.24, + "learning_rate": 1.2567036647416218e-05, + "loss": 0.9027, + "step": 34739 + }, + { + "epoch": 2.24, + "learning_rate": 1.2566821958393269e-05, + "loss": 0.9157, + "step": 34740 + }, + { + "epoch": 2.24, + "learning_rate": 1.256660726937032e-05, + "loss": 0.9145, + "step": 34741 + }, + { + "epoch": 2.24, + "learning_rate": 1.2566392580347368e-05, + "loss": 0.8469, + "step": 34742 + }, + { + "epoch": 2.24, + "learning_rate": 1.2566177891324417e-05, + "loss": 0.9351, + "step": 34743 + }, + { + "epoch": 2.24, + "learning_rate": 1.2565963202301468e-05, + "loss": 0.8493, + "step": 34744 + }, + { + "epoch": 2.24, + "learning_rate": 1.2565748513278517e-05, + "loss": 0.8794, + "step": 34745 + }, + { + "epoch": 2.24, + "learning_rate": 1.2565533824255568e-05, + "loss": 0.9796, + "step": 34746 + }, + { + "epoch": 2.24, + "learning_rate": 1.2565319135232617e-05, + "loss": 0.8893, + "step": 34747 + }, + { + "epoch": 2.24, + "learning_rate": 1.2565104446209666e-05, + "loss": 0.9456, + "step": 34748 + }, + { + "epoch": 2.24, + "learning_rate": 1.2564889757186716e-05, + "loss": 0.897, + "step": 34749 + }, + { + "epoch": 2.24, + "learning_rate": 1.2564675068163765e-05, + "loss": 0.8789, + "step": 34750 + }, + { + "epoch": 2.24, + "learning_rate": 1.2564460379140814e-05, + "loss": 0.8611, + "step": 34751 + }, + { + "epoch": 2.24, + "learning_rate": 1.2564245690117865e-05, + "loss": 0.9694, + "step": 34752 + }, + { + "epoch": 2.24, + "learning_rate": 1.2564031001094914e-05, + "loss": 0.9487, + "step": 34753 + }, + { + "epoch": 2.24, + "learning_rate": 1.2563816312071966e-05, + "loss": 0.9572, + "step": 34754 + }, + { + "epoch": 2.24, + "learning_rate": 1.2563601623049015e-05, + "loss": 0.823, + "step": 34755 + }, + { + "epoch": 2.24, + "learning_rate": 1.2563386934026066e-05, + "loss": 0.8906, + "step": 34756 + }, + { + "epoch": 2.24, + "learning_rate": 1.2563172245003115e-05, + "loss": 0.932, + "step": 34757 + }, + { + "epoch": 2.24, + "learning_rate": 1.2562957555980164e-05, + "loss": 0.8325, + "step": 34758 + }, + { + "epoch": 2.24, + "learning_rate": 1.2562742866957214e-05, + "loss": 0.8687, + "step": 34759 + }, + { + "epoch": 2.24, + "learning_rate": 1.2562528177934263e-05, + "loss": 0.8574, + "step": 34760 + }, + { + "epoch": 2.24, + "learning_rate": 1.2562313488911312e-05, + "loss": 0.9421, + "step": 34761 + }, + { + "epoch": 2.24, + "learning_rate": 1.2562098799888363e-05, + "loss": 0.8531, + "step": 34762 + }, + { + "epoch": 2.24, + "learning_rate": 1.2561884110865412e-05, + "loss": 0.8776, + "step": 34763 + }, + { + "epoch": 2.24, + "learning_rate": 1.2561669421842461e-05, + "loss": 0.919, + "step": 34764 + }, + { + "epoch": 2.24, + "learning_rate": 1.2561454732819512e-05, + "loss": 0.898, + "step": 34765 + }, + { + "epoch": 2.24, + "learning_rate": 1.256124004379656e-05, + "loss": 0.9123, + "step": 34766 + }, + { + "epoch": 2.24, + "learning_rate": 1.2561025354773611e-05, + "loss": 0.8643, + "step": 34767 + }, + { + "epoch": 2.24, + "learning_rate": 1.256081066575066e-05, + "loss": 0.8965, + "step": 34768 + }, + { + "epoch": 2.24, + "learning_rate": 1.2560595976727713e-05, + "loss": 0.8976, + "step": 34769 + }, + { + "epoch": 2.24, + "learning_rate": 1.2560381287704762e-05, + "loss": 0.8804, + "step": 34770 + }, + { + "epoch": 2.24, + "learning_rate": 1.256016659868181e-05, + "loss": 0.9306, + "step": 34771 + }, + { + "epoch": 2.24, + "learning_rate": 1.2559951909658861e-05, + "loss": 0.8864, + "step": 34772 + }, + { + "epoch": 2.24, + "learning_rate": 1.255973722063591e-05, + "loss": 0.907, + "step": 34773 + }, + { + "epoch": 2.24, + "learning_rate": 1.2559522531612959e-05, + "loss": 0.8337, + "step": 34774 + }, + { + "epoch": 2.24, + "learning_rate": 1.255930784259001e-05, + "loss": 0.9368, + "step": 34775 + }, + { + "epoch": 2.24, + "learning_rate": 1.2559093153567059e-05, + "loss": 0.9466, + "step": 34776 + }, + { + "epoch": 2.24, + "learning_rate": 1.255887846454411e-05, + "loss": 0.9125, + "step": 34777 + }, + { + "epoch": 2.24, + "learning_rate": 1.2558663775521158e-05, + "loss": 0.9175, + "step": 34778 + }, + { + "epoch": 2.24, + "learning_rate": 1.2558449086498207e-05, + "loss": 0.9099, + "step": 34779 + }, + { + "epoch": 2.24, + "learning_rate": 1.2558234397475258e-05, + "loss": 0.8801, + "step": 34780 + }, + { + "epoch": 2.24, + "learning_rate": 1.2558019708452307e-05, + "loss": 0.9371, + "step": 34781 + }, + { + "epoch": 2.24, + "learning_rate": 1.2557805019429356e-05, + "loss": 0.8793, + "step": 34782 + }, + { + "epoch": 2.24, + "learning_rate": 1.2557590330406408e-05, + "loss": 0.8987, + "step": 34783 + }, + { + "epoch": 2.24, + "learning_rate": 1.2557375641383457e-05, + "loss": 0.9298, + "step": 34784 + }, + { + "epoch": 2.24, + "learning_rate": 1.2557160952360508e-05, + "loss": 0.8901, + "step": 34785 + }, + { + "epoch": 2.24, + "learning_rate": 1.2556946263337557e-05, + "loss": 0.8598, + "step": 34786 + }, + { + "epoch": 2.24, + "learning_rate": 1.2556731574314607e-05, + "loss": 0.9077, + "step": 34787 + }, + { + "epoch": 2.24, + "learning_rate": 1.2556516885291656e-05, + "loss": 0.8678, + "step": 34788 + }, + { + "epoch": 2.24, + "learning_rate": 1.2556302196268705e-05, + "loss": 0.8949, + "step": 34789 + }, + { + "epoch": 2.24, + "learning_rate": 1.2556087507245756e-05, + "loss": 0.854, + "step": 34790 + }, + { + "epoch": 2.24, + "learning_rate": 1.2555872818222805e-05, + "loss": 0.9345, + "step": 34791 + }, + { + "epoch": 2.24, + "learning_rate": 1.2555658129199854e-05, + "loss": 0.9491, + "step": 34792 + }, + { + "epoch": 2.24, + "learning_rate": 1.2555443440176905e-05, + "loss": 0.926, + "step": 34793 + }, + { + "epoch": 2.24, + "learning_rate": 1.2555228751153954e-05, + "loss": 0.9476, + "step": 34794 + }, + { + "epoch": 2.24, + "learning_rate": 1.2555014062131004e-05, + "loss": 0.8267, + "step": 34795 + }, + { + "epoch": 2.24, + "learning_rate": 1.2554799373108053e-05, + "loss": 0.8672, + "step": 34796 + }, + { + "epoch": 2.24, + "learning_rate": 1.2554584684085102e-05, + "loss": 0.9576, + "step": 34797 + }, + { + "epoch": 2.24, + "learning_rate": 1.2554369995062155e-05, + "loss": 0.909, + "step": 34798 + }, + { + "epoch": 2.24, + "learning_rate": 1.2554155306039204e-05, + "loss": 0.886, + "step": 34799 + }, + { + "epoch": 2.24, + "learning_rate": 1.2553940617016254e-05, + "loss": 0.8718, + "step": 34800 + }, + { + "epoch": 2.24, + "learning_rate": 1.2553725927993303e-05, + "loss": 0.8638, + "step": 34801 + }, + { + "epoch": 2.24, + "learning_rate": 1.2553511238970352e-05, + "loss": 0.7857, + "step": 34802 + }, + { + "epoch": 2.24, + "learning_rate": 1.2553296549947403e-05, + "loss": 0.9033, + "step": 34803 + }, + { + "epoch": 2.24, + "learning_rate": 1.2553081860924452e-05, + "loss": 0.9012, + "step": 34804 + }, + { + "epoch": 2.24, + "learning_rate": 1.2552867171901502e-05, + "loss": 0.9526, + "step": 34805 + }, + { + "epoch": 2.24, + "learning_rate": 1.2552652482878551e-05, + "loss": 0.9209, + "step": 34806 + }, + { + "epoch": 2.24, + "learning_rate": 1.25524377938556e-05, + "loss": 0.9113, + "step": 34807 + }, + { + "epoch": 2.24, + "learning_rate": 1.2552223104832651e-05, + "loss": 0.8353, + "step": 34808 + }, + { + "epoch": 2.24, + "learning_rate": 1.25520084158097e-05, + "loss": 0.8852, + "step": 34809 + }, + { + "epoch": 2.24, + "learning_rate": 1.2551793726786749e-05, + "loss": 0.8792, + "step": 34810 + }, + { + "epoch": 2.24, + "learning_rate": 1.25515790377638e-05, + "loss": 0.9151, + "step": 34811 + }, + { + "epoch": 2.24, + "learning_rate": 1.255136434874085e-05, + "loss": 0.9503, + "step": 34812 + }, + { + "epoch": 2.24, + "learning_rate": 1.2551149659717901e-05, + "loss": 0.9326, + "step": 34813 + }, + { + "epoch": 2.24, + "learning_rate": 1.255093497069495e-05, + "loss": 0.7899, + "step": 34814 + }, + { + "epoch": 2.24, + "learning_rate": 1.2550720281672e-05, + "loss": 0.9019, + "step": 34815 + }, + { + "epoch": 2.24, + "learning_rate": 1.255050559264905e-05, + "loss": 0.8634, + "step": 34816 + }, + { + "epoch": 2.24, + "learning_rate": 1.2550290903626098e-05, + "loss": 0.8553, + "step": 34817 + }, + { + "epoch": 2.24, + "learning_rate": 1.2550076214603149e-05, + "loss": 0.9164, + "step": 34818 + }, + { + "epoch": 2.24, + "learning_rate": 1.2549861525580198e-05, + "loss": 0.9015, + "step": 34819 + }, + { + "epoch": 2.24, + "learning_rate": 1.2549646836557247e-05, + "loss": 0.8981, + "step": 34820 + }, + { + "epoch": 2.24, + "learning_rate": 1.2549432147534298e-05, + "loss": 0.9091, + "step": 34821 + }, + { + "epoch": 2.24, + "learning_rate": 1.2549217458511347e-05, + "loss": 0.8891, + "step": 34822 + }, + { + "epoch": 2.24, + "learning_rate": 1.2549002769488396e-05, + "loss": 0.9223, + "step": 34823 + }, + { + "epoch": 2.24, + "learning_rate": 1.2548788080465446e-05, + "loss": 0.856, + "step": 34824 + }, + { + "epoch": 2.24, + "learning_rate": 1.2548573391442495e-05, + "loss": 0.9176, + "step": 34825 + }, + { + "epoch": 2.24, + "learning_rate": 1.2548358702419546e-05, + "loss": 0.9218, + "step": 34826 + }, + { + "epoch": 2.24, + "learning_rate": 1.2548144013396597e-05, + "loss": 0.8854, + "step": 34827 + }, + { + "epoch": 2.24, + "learning_rate": 1.2547929324373647e-05, + "loss": 0.9066, + "step": 34828 + }, + { + "epoch": 2.24, + "learning_rate": 1.2547714635350696e-05, + "loss": 0.8974, + "step": 34829 + }, + { + "epoch": 2.24, + "learning_rate": 1.2547499946327745e-05, + "loss": 0.9206, + "step": 34830 + }, + { + "epoch": 2.24, + "learning_rate": 1.2547285257304796e-05, + "loss": 0.9335, + "step": 34831 + }, + { + "epoch": 2.24, + "learning_rate": 1.2547070568281845e-05, + "loss": 0.8414, + "step": 34832 + }, + { + "epoch": 2.24, + "learning_rate": 1.2546855879258894e-05, + "loss": 0.8612, + "step": 34833 + }, + { + "epoch": 2.24, + "learning_rate": 1.2546641190235944e-05, + "loss": 0.8562, + "step": 34834 + }, + { + "epoch": 2.24, + "learning_rate": 1.2546426501212993e-05, + "loss": 0.9333, + "step": 34835 + }, + { + "epoch": 2.24, + "learning_rate": 1.2546211812190044e-05, + "loss": 0.8298, + "step": 34836 + }, + { + "epoch": 2.24, + "learning_rate": 1.2545997123167093e-05, + "loss": 0.8953, + "step": 34837 + }, + { + "epoch": 2.24, + "learning_rate": 1.2545782434144142e-05, + "loss": 0.8892, + "step": 34838 + }, + { + "epoch": 2.24, + "learning_rate": 1.2545567745121193e-05, + "loss": 0.8329, + "step": 34839 + }, + { + "epoch": 2.24, + "learning_rate": 1.2545353056098242e-05, + "loss": 0.9256, + "step": 34840 + }, + { + "epoch": 2.24, + "learning_rate": 1.2545138367075294e-05, + "loss": 0.8915, + "step": 34841 + }, + { + "epoch": 2.24, + "learning_rate": 1.2544923678052343e-05, + "loss": 0.8953, + "step": 34842 + }, + { + "epoch": 2.24, + "learning_rate": 1.2544708989029392e-05, + "loss": 0.8538, + "step": 34843 + }, + { + "epoch": 2.24, + "learning_rate": 1.2544494300006443e-05, + "loss": 0.9157, + "step": 34844 + }, + { + "epoch": 2.24, + "learning_rate": 1.2544279610983491e-05, + "loss": 0.8645, + "step": 34845 + }, + { + "epoch": 2.24, + "learning_rate": 1.2544064921960542e-05, + "loss": 0.9002, + "step": 34846 + }, + { + "epoch": 2.24, + "learning_rate": 1.2543850232937591e-05, + "loss": 0.8606, + "step": 34847 + }, + { + "epoch": 2.24, + "learning_rate": 1.254363554391464e-05, + "loss": 0.887, + "step": 34848 + }, + { + "epoch": 2.24, + "learning_rate": 1.254342085489169e-05, + "loss": 0.8659, + "step": 34849 + }, + { + "epoch": 2.24, + "learning_rate": 1.254320616586874e-05, + "loss": 0.9104, + "step": 34850 + }, + { + "epoch": 2.24, + "learning_rate": 1.2542991476845789e-05, + "loss": 0.9306, + "step": 34851 + }, + { + "epoch": 2.24, + "learning_rate": 1.254277678782284e-05, + "loss": 0.8633, + "step": 34852 + }, + { + "epoch": 2.24, + "learning_rate": 1.2542562098799888e-05, + "loss": 0.8903, + "step": 34853 + }, + { + "epoch": 2.24, + "learning_rate": 1.2542347409776939e-05, + "loss": 0.8949, + "step": 34854 + }, + { + "epoch": 2.24, + "learning_rate": 1.2542132720753988e-05, + "loss": 0.9279, + "step": 34855 + }, + { + "epoch": 2.24, + "learning_rate": 1.254191803173104e-05, + "loss": 0.8865, + "step": 34856 + }, + { + "epoch": 2.24, + "learning_rate": 1.254170334270809e-05, + "loss": 0.9083, + "step": 34857 + }, + { + "epoch": 2.24, + "learning_rate": 1.2541488653685138e-05, + "loss": 0.8841, + "step": 34858 + }, + { + "epoch": 2.24, + "learning_rate": 1.2541273964662189e-05, + "loss": 0.9085, + "step": 34859 + }, + { + "epoch": 2.24, + "learning_rate": 1.2541059275639238e-05, + "loss": 0.8241, + "step": 34860 + }, + { + "epoch": 2.24, + "learning_rate": 1.2540844586616287e-05, + "loss": 0.8776, + "step": 34861 + }, + { + "epoch": 2.24, + "learning_rate": 1.2540629897593337e-05, + "loss": 0.8354, + "step": 34862 + }, + { + "epoch": 2.24, + "learning_rate": 1.2540415208570386e-05, + "loss": 0.9006, + "step": 34863 + }, + { + "epoch": 2.24, + "learning_rate": 1.2540200519547437e-05, + "loss": 0.8112, + "step": 34864 + }, + { + "epoch": 2.24, + "learning_rate": 1.2539985830524486e-05, + "loss": 0.9109, + "step": 34865 + }, + { + "epoch": 2.24, + "learning_rate": 1.2539771141501535e-05, + "loss": 0.934, + "step": 34866 + }, + { + "epoch": 2.24, + "learning_rate": 1.2539556452478586e-05, + "loss": 0.9175, + "step": 34867 + }, + { + "epoch": 2.24, + "learning_rate": 1.2539341763455635e-05, + "loss": 0.8682, + "step": 34868 + }, + { + "epoch": 2.24, + "learning_rate": 1.2539127074432684e-05, + "loss": 0.8854, + "step": 34869 + }, + { + "epoch": 2.24, + "learning_rate": 1.2538912385409736e-05, + "loss": 0.9029, + "step": 34870 + }, + { + "epoch": 2.24, + "learning_rate": 1.2538697696386785e-05, + "loss": 0.9304, + "step": 34871 + }, + { + "epoch": 2.24, + "learning_rate": 1.2538483007363836e-05, + "loss": 0.8387, + "step": 34872 + }, + { + "epoch": 2.24, + "learning_rate": 1.2538268318340885e-05, + "loss": 0.8969, + "step": 34873 + }, + { + "epoch": 2.24, + "learning_rate": 1.2538053629317935e-05, + "loss": 0.8235, + "step": 34874 + }, + { + "epoch": 2.24, + "learning_rate": 1.2537838940294984e-05, + "loss": 0.8252, + "step": 34875 + }, + { + "epoch": 2.24, + "learning_rate": 1.2537624251272033e-05, + "loss": 0.895, + "step": 34876 + }, + { + "epoch": 2.24, + "learning_rate": 1.2537409562249084e-05, + "loss": 0.9254, + "step": 34877 + }, + { + "epoch": 2.24, + "learning_rate": 1.2537194873226133e-05, + "loss": 0.8406, + "step": 34878 + }, + { + "epoch": 2.24, + "learning_rate": 1.2536980184203182e-05, + "loss": 0.9311, + "step": 34879 + }, + { + "epoch": 2.24, + "learning_rate": 1.2536765495180232e-05, + "loss": 0.9102, + "step": 34880 + }, + { + "epoch": 2.24, + "learning_rate": 1.2536550806157281e-05, + "loss": 0.8974, + "step": 34881 + }, + { + "epoch": 2.24, + "learning_rate": 1.2536336117134332e-05, + "loss": 0.8913, + "step": 34882 + }, + { + "epoch": 2.24, + "learning_rate": 1.2536121428111381e-05, + "loss": 0.9447, + "step": 34883 + }, + { + "epoch": 2.24, + "learning_rate": 1.253590673908843e-05, + "loss": 0.9247, + "step": 34884 + }, + { + "epoch": 2.24, + "learning_rate": 1.2535692050065482e-05, + "loss": 0.8261, + "step": 34885 + }, + { + "epoch": 2.24, + "learning_rate": 1.2535477361042531e-05, + "loss": 0.8814, + "step": 34886 + }, + { + "epoch": 2.24, + "learning_rate": 1.2535262672019582e-05, + "loss": 0.9213, + "step": 34887 + }, + { + "epoch": 2.24, + "learning_rate": 1.2535047982996631e-05, + "loss": 0.926, + "step": 34888 + }, + { + "epoch": 2.24, + "learning_rate": 1.253483329397368e-05, + "loss": 0.9315, + "step": 34889 + }, + { + "epoch": 2.24, + "learning_rate": 1.253461860495073e-05, + "loss": 0.9082, + "step": 34890 + }, + { + "epoch": 2.24, + "learning_rate": 1.253440391592778e-05, + "loss": 0.8705, + "step": 34891 + }, + { + "epoch": 2.24, + "learning_rate": 1.253418922690483e-05, + "loss": 0.9392, + "step": 34892 + }, + { + "epoch": 2.24, + "learning_rate": 1.2533974537881879e-05, + "loss": 0.9175, + "step": 34893 + }, + { + "epoch": 2.24, + "learning_rate": 1.2533759848858928e-05, + "loss": 0.9389, + "step": 34894 + }, + { + "epoch": 2.25, + "learning_rate": 1.2533545159835979e-05, + "loss": 0.8787, + "step": 34895 + }, + { + "epoch": 2.25, + "learning_rate": 1.2533330470813028e-05, + "loss": 0.8379, + "step": 34896 + }, + { + "epoch": 2.25, + "learning_rate": 1.2533115781790077e-05, + "loss": 0.952, + "step": 34897 + }, + { + "epoch": 2.25, + "learning_rate": 1.2532901092767127e-05, + "loss": 0.8623, + "step": 34898 + }, + { + "epoch": 2.25, + "learning_rate": 1.2532686403744178e-05, + "loss": 0.8928, + "step": 34899 + }, + { + "epoch": 2.25, + "learning_rate": 1.2532471714721229e-05, + "loss": 0.9084, + "step": 34900 + }, + { + "epoch": 2.25, + "learning_rate": 1.2532257025698278e-05, + "loss": 0.9577, + "step": 34901 + }, + { + "epoch": 2.25, + "learning_rate": 1.2532042336675328e-05, + "loss": 1.0008, + "step": 34902 + }, + { + "epoch": 2.25, + "learning_rate": 1.2531827647652377e-05, + "loss": 0.8702, + "step": 34903 + }, + { + "epoch": 2.25, + "learning_rate": 1.2531612958629426e-05, + "loss": 0.9014, + "step": 34904 + }, + { + "epoch": 2.25, + "learning_rate": 1.2531398269606477e-05, + "loss": 0.8771, + "step": 34905 + }, + { + "epoch": 2.25, + "learning_rate": 1.2531183580583526e-05, + "loss": 0.8983, + "step": 34906 + }, + { + "epoch": 2.25, + "learning_rate": 1.2530968891560575e-05, + "loss": 0.8625, + "step": 34907 + }, + { + "epoch": 2.25, + "learning_rate": 1.2530754202537625e-05, + "loss": 0.8383, + "step": 34908 + }, + { + "epoch": 2.25, + "learning_rate": 1.2530539513514674e-05, + "loss": 0.9584, + "step": 34909 + }, + { + "epoch": 2.25, + "learning_rate": 1.2530324824491723e-05, + "loss": 0.8845, + "step": 34910 + }, + { + "epoch": 2.25, + "learning_rate": 1.2530110135468774e-05, + "loss": 0.9374, + "step": 34911 + }, + { + "epoch": 2.25, + "learning_rate": 1.2529895446445823e-05, + "loss": 0.871, + "step": 34912 + }, + { + "epoch": 2.25, + "learning_rate": 1.2529680757422874e-05, + "loss": 0.8797, + "step": 34913 + }, + { + "epoch": 2.25, + "learning_rate": 1.2529466068399924e-05, + "loss": 0.99, + "step": 34914 + }, + { + "epoch": 2.25, + "learning_rate": 1.2529251379376975e-05, + "loss": 0.9117, + "step": 34915 + }, + { + "epoch": 2.25, + "learning_rate": 1.2529036690354024e-05, + "loss": 0.9485, + "step": 34916 + }, + { + "epoch": 2.25, + "learning_rate": 1.2528822001331073e-05, + "loss": 0.8538, + "step": 34917 + }, + { + "epoch": 2.25, + "learning_rate": 1.2528607312308124e-05, + "loss": 0.9222, + "step": 34918 + }, + { + "epoch": 2.25, + "learning_rate": 1.2528392623285172e-05, + "loss": 0.847, + "step": 34919 + }, + { + "epoch": 2.25, + "learning_rate": 1.2528177934262221e-05, + "loss": 0.9207, + "step": 34920 + }, + { + "epoch": 2.25, + "learning_rate": 1.2527963245239272e-05, + "loss": 0.8222, + "step": 34921 + }, + { + "epoch": 2.25, + "learning_rate": 1.2527748556216321e-05, + "loss": 0.9066, + "step": 34922 + }, + { + "epoch": 2.25, + "learning_rate": 1.2527533867193372e-05, + "loss": 0.917, + "step": 34923 + }, + { + "epoch": 2.25, + "learning_rate": 1.252731917817042e-05, + "loss": 0.9515, + "step": 34924 + }, + { + "epoch": 2.25, + "learning_rate": 1.252710448914747e-05, + "loss": 0.8589, + "step": 34925 + }, + { + "epoch": 2.25, + "learning_rate": 1.252688980012452e-05, + "loss": 0.9128, + "step": 34926 + }, + { + "epoch": 2.25, + "learning_rate": 1.252667511110157e-05, + "loss": 0.8991, + "step": 34927 + }, + { + "epoch": 2.25, + "learning_rate": 1.2526460422078622e-05, + "loss": 0.9092, + "step": 34928 + }, + { + "epoch": 2.25, + "learning_rate": 1.252624573305567e-05, + "loss": 0.8967, + "step": 34929 + }, + { + "epoch": 2.25, + "learning_rate": 1.252603104403272e-05, + "loss": 0.8895, + "step": 34930 + }, + { + "epoch": 2.25, + "learning_rate": 1.252581635500977e-05, + "loss": 0.7869, + "step": 34931 + }, + { + "epoch": 2.25, + "learning_rate": 1.252560166598682e-05, + "loss": 0.8711, + "step": 34932 + }, + { + "epoch": 2.25, + "learning_rate": 1.252538697696387e-05, + "loss": 0.9118, + "step": 34933 + }, + { + "epoch": 2.25, + "learning_rate": 1.2525172287940919e-05, + "loss": 0.9239, + "step": 34934 + }, + { + "epoch": 2.25, + "learning_rate": 1.2524957598917968e-05, + "loss": 0.9428, + "step": 34935 + }, + { + "epoch": 2.25, + "learning_rate": 1.2524742909895018e-05, + "loss": 0.9228, + "step": 34936 + }, + { + "epoch": 2.25, + "learning_rate": 1.2524528220872067e-05, + "loss": 0.9061, + "step": 34937 + }, + { + "epoch": 2.25, + "learning_rate": 1.2524313531849116e-05, + "loss": 0.87, + "step": 34938 + }, + { + "epoch": 2.25, + "learning_rate": 1.2524098842826167e-05, + "loss": 0.8838, + "step": 34939 + }, + { + "epoch": 2.25, + "learning_rate": 1.2523884153803216e-05, + "loss": 0.8201, + "step": 34940 + }, + { + "epoch": 2.25, + "learning_rate": 1.2523669464780267e-05, + "loss": 0.854, + "step": 34941 + }, + { + "epoch": 2.25, + "learning_rate": 1.2523454775757316e-05, + "loss": 0.9422, + "step": 34942 + }, + { + "epoch": 2.25, + "learning_rate": 1.2523240086734368e-05, + "loss": 0.8796, + "step": 34943 + }, + { + "epoch": 2.25, + "learning_rate": 1.2523025397711417e-05, + "loss": 0.8781, + "step": 34944 + }, + { + "epoch": 2.25, + "learning_rate": 1.2522810708688466e-05, + "loss": 0.9638, + "step": 34945 + }, + { + "epoch": 2.25, + "learning_rate": 1.2522596019665517e-05, + "loss": 0.9071, + "step": 34946 + }, + { + "epoch": 2.25, + "learning_rate": 1.2522381330642566e-05, + "loss": 0.9416, + "step": 34947 + }, + { + "epoch": 2.25, + "learning_rate": 1.2522166641619615e-05, + "loss": 0.9255, + "step": 34948 + }, + { + "epoch": 2.25, + "learning_rate": 1.2521951952596665e-05, + "loss": 0.8913, + "step": 34949 + }, + { + "epoch": 2.25, + "learning_rate": 1.2521737263573714e-05, + "loss": 0.9661, + "step": 34950 + }, + { + "epoch": 2.25, + "learning_rate": 1.2521522574550765e-05, + "loss": 0.834, + "step": 34951 + }, + { + "epoch": 2.25, + "learning_rate": 1.2521307885527814e-05, + "loss": 0.8524, + "step": 34952 + }, + { + "epoch": 2.25, + "learning_rate": 1.2521093196504863e-05, + "loss": 0.8976, + "step": 34953 + }, + { + "epoch": 2.25, + "learning_rate": 1.2520878507481913e-05, + "loss": 0.8879, + "step": 34954 + }, + { + "epoch": 2.25, + "learning_rate": 1.2520663818458962e-05, + "loss": 0.9603, + "step": 34955 + }, + { + "epoch": 2.25, + "learning_rate": 1.2520449129436011e-05, + "loss": 0.8581, + "step": 34956 + }, + { + "epoch": 2.25, + "learning_rate": 1.2520234440413064e-05, + "loss": 0.9387, + "step": 34957 + }, + { + "epoch": 2.25, + "learning_rate": 1.2520019751390113e-05, + "loss": 0.9178, + "step": 34958 + }, + { + "epoch": 2.25, + "learning_rate": 1.2519805062367163e-05, + "loss": 0.8496, + "step": 34959 + }, + { + "epoch": 2.25, + "learning_rate": 1.2519590373344212e-05, + "loss": 0.9651, + "step": 34960 + }, + { + "epoch": 2.25, + "learning_rate": 1.2519375684321263e-05, + "loss": 0.9574, + "step": 34961 + }, + { + "epoch": 2.25, + "learning_rate": 1.2519160995298312e-05, + "loss": 0.9055, + "step": 34962 + }, + { + "epoch": 2.25, + "learning_rate": 1.251894630627536e-05, + "loss": 0.911, + "step": 34963 + }, + { + "epoch": 2.25, + "learning_rate": 1.2518731617252411e-05, + "loss": 0.869, + "step": 34964 + }, + { + "epoch": 2.25, + "learning_rate": 1.251851692822946e-05, + "loss": 0.9115, + "step": 34965 + }, + { + "epoch": 2.25, + "learning_rate": 1.251830223920651e-05, + "loss": 0.9459, + "step": 34966 + }, + { + "epoch": 2.25, + "learning_rate": 1.251808755018356e-05, + "loss": 0.8317, + "step": 34967 + }, + { + "epoch": 2.25, + "learning_rate": 1.2517872861160609e-05, + "loss": 0.9104, + "step": 34968 + }, + { + "epoch": 2.25, + "learning_rate": 1.251765817213766e-05, + "loss": 0.9666, + "step": 34969 + }, + { + "epoch": 2.25, + "learning_rate": 1.2517443483114709e-05, + "loss": 0.9418, + "step": 34970 + }, + { + "epoch": 2.25, + "learning_rate": 1.2517228794091758e-05, + "loss": 0.851, + "step": 34971 + }, + { + "epoch": 2.25, + "learning_rate": 1.251701410506881e-05, + "loss": 0.9385, + "step": 34972 + }, + { + "epoch": 2.25, + "learning_rate": 1.2516799416045859e-05, + "loss": 0.8938, + "step": 34973 + }, + { + "epoch": 2.25, + "learning_rate": 1.251658472702291e-05, + "loss": 0.8634, + "step": 34974 + }, + { + "epoch": 2.25, + "learning_rate": 1.2516370037999959e-05, + "loss": 0.9173, + "step": 34975 + }, + { + "epoch": 2.25, + "learning_rate": 1.2516155348977008e-05, + "loss": 0.9172, + "step": 34976 + }, + { + "epoch": 2.25, + "learning_rate": 1.2515940659954058e-05, + "loss": 0.8672, + "step": 34977 + }, + { + "epoch": 2.25, + "learning_rate": 1.2515725970931107e-05, + "loss": 0.8748, + "step": 34978 + }, + { + "epoch": 2.25, + "learning_rate": 1.2515511281908158e-05, + "loss": 0.93, + "step": 34979 + }, + { + "epoch": 2.25, + "learning_rate": 1.2515296592885207e-05, + "loss": 0.8671, + "step": 34980 + }, + { + "epoch": 2.25, + "learning_rate": 1.2515081903862256e-05, + "loss": 0.9202, + "step": 34981 + }, + { + "epoch": 2.25, + "learning_rate": 1.2514867214839306e-05, + "loss": 0.9255, + "step": 34982 + }, + { + "epoch": 2.25, + "learning_rate": 1.2514652525816355e-05, + "loss": 0.8674, + "step": 34983 + }, + { + "epoch": 2.25, + "learning_rate": 1.2514437836793404e-05, + "loss": 0.8819, + "step": 34984 + }, + { + "epoch": 2.25, + "learning_rate": 1.2514223147770455e-05, + "loss": 0.892, + "step": 34985 + }, + { + "epoch": 2.25, + "learning_rate": 1.2514008458747506e-05, + "loss": 0.8996, + "step": 34986 + }, + { + "epoch": 2.25, + "learning_rate": 1.2513793769724556e-05, + "loss": 0.9533, + "step": 34987 + }, + { + "epoch": 2.25, + "learning_rate": 1.2513579080701605e-05, + "loss": 0.8827, + "step": 34988 + }, + { + "epoch": 2.25, + "learning_rate": 1.2513364391678656e-05, + "loss": 0.9357, + "step": 34989 + }, + { + "epoch": 2.25, + "learning_rate": 1.2513149702655705e-05, + "loss": 0.8567, + "step": 34990 + }, + { + "epoch": 2.25, + "learning_rate": 1.2512935013632754e-05, + "loss": 0.9129, + "step": 34991 + }, + { + "epoch": 2.25, + "learning_rate": 1.2512720324609805e-05, + "loss": 0.9271, + "step": 34992 + }, + { + "epoch": 2.25, + "learning_rate": 1.2512505635586853e-05, + "loss": 0.8277, + "step": 34993 + }, + { + "epoch": 2.25, + "learning_rate": 1.2512290946563902e-05, + "loss": 0.8569, + "step": 34994 + }, + { + "epoch": 2.25, + "learning_rate": 1.2512076257540953e-05, + "loss": 0.8596, + "step": 34995 + }, + { + "epoch": 2.25, + "learning_rate": 1.2511861568518002e-05, + "loss": 0.9424, + "step": 34996 + }, + { + "epoch": 2.25, + "learning_rate": 1.2511646879495051e-05, + "loss": 0.8931, + "step": 34997 + }, + { + "epoch": 2.25, + "learning_rate": 1.2511432190472102e-05, + "loss": 0.8563, + "step": 34998 + }, + { + "epoch": 2.25, + "learning_rate": 1.251121750144915e-05, + "loss": 0.8903, + "step": 34999 + }, + { + "epoch": 2.25, + "learning_rate": 1.2511002812426201e-05, + "loss": 0.8222, + "step": 35000 + }, + { + "epoch": 2.25, + "eval_laion/cider": 0.5755084141301785, + "eval_laion/meteor": 0.3934124971665931, + "eval_laion/pred_num_words": 37.41015625, + "eval_laion/sacrebleu": 16.1391791789547, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0408897399902344, + "eval_runtime": 101.53, + "eval_samples_per_second": 4.915, + "eval_steps_per_second": 0.315, + "step": 35000 + }, + { + "epoch": 2.25, + "learning_rate": 1.2510788123403252e-05, + "loss": 0.9153, + "step": 35001 + }, + { + "epoch": 2.25, + "learning_rate": 1.2510573434380303e-05, + "loss": 0.9065, + "step": 35002 + }, + { + "epoch": 2.25, + "learning_rate": 1.2510358745357352e-05, + "loss": 0.8367, + "step": 35003 + }, + { + "epoch": 2.25, + "learning_rate": 1.25101440563344e-05, + "loss": 0.8572, + "step": 35004 + }, + { + "epoch": 2.25, + "learning_rate": 1.2509929367311451e-05, + "loss": 0.9633, + "step": 35005 + }, + { + "epoch": 2.25, + "learning_rate": 1.25097146782885e-05, + "loss": 0.8958, + "step": 35006 + }, + { + "epoch": 2.25, + "learning_rate": 1.250949998926555e-05, + "loss": 0.931, + "step": 35007 + }, + { + "epoch": 2.25, + "learning_rate": 1.25092853002426e-05, + "loss": 0.9102, + "step": 35008 + }, + { + "epoch": 2.25, + "learning_rate": 1.2509070611219649e-05, + "loss": 0.8982, + "step": 35009 + }, + { + "epoch": 2.25, + "learning_rate": 1.25088559221967e-05, + "loss": 0.9376, + "step": 35010 + }, + { + "epoch": 2.25, + "learning_rate": 1.2508641233173748e-05, + "loss": 0.9195, + "step": 35011 + }, + { + "epoch": 2.25, + "learning_rate": 1.2508426544150797e-05, + "loss": 0.8133, + "step": 35012 + }, + { + "epoch": 2.25, + "learning_rate": 1.2508211855127848e-05, + "loss": 0.9105, + "step": 35013 + }, + { + "epoch": 2.25, + "learning_rate": 1.2507997166104897e-05, + "loss": 0.9101, + "step": 35014 + }, + { + "epoch": 2.25, + "learning_rate": 1.250778247708195e-05, + "loss": 0.8751, + "step": 35015 + }, + { + "epoch": 2.25, + "learning_rate": 1.2507567788058998e-05, + "loss": 0.8461, + "step": 35016 + }, + { + "epoch": 2.25, + "learning_rate": 1.2507353099036047e-05, + "loss": 0.94, + "step": 35017 + }, + { + "epoch": 2.25, + "learning_rate": 1.2507138410013098e-05, + "loss": 0.8158, + "step": 35018 + }, + { + "epoch": 2.25, + "learning_rate": 1.2506923720990147e-05, + "loss": 0.8764, + "step": 35019 + }, + { + "epoch": 2.25, + "learning_rate": 1.2506709031967198e-05, + "loss": 0.8834, + "step": 35020 + }, + { + "epoch": 2.25, + "learning_rate": 1.2506494342944247e-05, + "loss": 0.9274, + "step": 35021 + }, + { + "epoch": 2.25, + "learning_rate": 1.2506279653921296e-05, + "loss": 0.8435, + "step": 35022 + }, + { + "epoch": 2.25, + "learning_rate": 1.2506064964898346e-05, + "loss": 0.8693, + "step": 35023 + }, + { + "epoch": 2.25, + "learning_rate": 1.2505850275875395e-05, + "loss": 0.8827, + "step": 35024 + }, + { + "epoch": 2.25, + "learning_rate": 1.2505635586852444e-05, + "loss": 0.9053, + "step": 35025 + }, + { + "epoch": 2.25, + "learning_rate": 1.2505420897829495e-05, + "loss": 0.8964, + "step": 35026 + }, + { + "epoch": 2.25, + "learning_rate": 1.2505206208806544e-05, + "loss": 0.8455, + "step": 35027 + }, + { + "epoch": 2.25, + "learning_rate": 1.2504991519783594e-05, + "loss": 0.8637, + "step": 35028 + }, + { + "epoch": 2.25, + "learning_rate": 1.2504776830760643e-05, + "loss": 0.8892, + "step": 35029 + }, + { + "epoch": 2.25, + "learning_rate": 1.2504562141737696e-05, + "loss": 0.8949, + "step": 35030 + }, + { + "epoch": 2.25, + "learning_rate": 1.2504347452714745e-05, + "loss": 0.921, + "step": 35031 + }, + { + "epoch": 2.25, + "learning_rate": 1.2504132763691794e-05, + "loss": 0.8598, + "step": 35032 + }, + { + "epoch": 2.25, + "learning_rate": 1.2503918074668844e-05, + "loss": 0.9436, + "step": 35033 + }, + { + "epoch": 2.25, + "learning_rate": 1.2503703385645893e-05, + "loss": 0.8967, + "step": 35034 + }, + { + "epoch": 2.25, + "learning_rate": 1.2503488696622942e-05, + "loss": 0.8414, + "step": 35035 + }, + { + "epoch": 2.25, + "learning_rate": 1.2503274007599993e-05, + "loss": 0.9067, + "step": 35036 + }, + { + "epoch": 2.25, + "learning_rate": 1.2503059318577042e-05, + "loss": 0.9005, + "step": 35037 + }, + { + "epoch": 2.25, + "learning_rate": 1.2502844629554092e-05, + "loss": 0.834, + "step": 35038 + }, + { + "epoch": 2.25, + "learning_rate": 1.2502629940531141e-05, + "loss": 0.914, + "step": 35039 + }, + { + "epoch": 2.25, + "learning_rate": 1.250241525150819e-05, + "loss": 0.8509, + "step": 35040 + }, + { + "epoch": 2.25, + "learning_rate": 1.2502200562485241e-05, + "loss": 0.8718, + "step": 35041 + }, + { + "epoch": 2.25, + "learning_rate": 1.250198587346229e-05, + "loss": 0.8772, + "step": 35042 + }, + { + "epoch": 2.25, + "learning_rate": 1.2501771184439339e-05, + "loss": 0.9205, + "step": 35043 + }, + { + "epoch": 2.25, + "learning_rate": 1.2501556495416391e-05, + "loss": 0.8886, + "step": 35044 + }, + { + "epoch": 2.25, + "learning_rate": 1.250134180639344e-05, + "loss": 0.9564, + "step": 35045 + }, + { + "epoch": 2.25, + "learning_rate": 1.2501127117370491e-05, + "loss": 0.869, + "step": 35046 + }, + { + "epoch": 2.25, + "learning_rate": 1.250091242834754e-05, + "loss": 0.9131, + "step": 35047 + }, + { + "epoch": 2.25, + "learning_rate": 1.250069773932459e-05, + "loss": 0.9513, + "step": 35048 + }, + { + "epoch": 2.25, + "learning_rate": 1.250048305030164e-05, + "loss": 0.8886, + "step": 35049 + }, + { + "epoch": 2.26, + "learning_rate": 1.2500268361278689e-05, + "loss": 0.9216, + "step": 35050 + }, + { + "epoch": 2.26, + "learning_rate": 1.250005367225574e-05, + "loss": 0.9049, + "step": 35051 + }, + { + "epoch": 2.26, + "learning_rate": 1.2499838983232788e-05, + "loss": 0.8969, + "step": 35052 + }, + { + "epoch": 2.26, + "learning_rate": 1.2499624294209837e-05, + "loss": 0.8828, + "step": 35053 + }, + { + "epoch": 2.26, + "learning_rate": 1.2499409605186888e-05, + "loss": 0.9578, + "step": 35054 + }, + { + "epoch": 2.26, + "learning_rate": 1.2499194916163937e-05, + "loss": 0.8735, + "step": 35055 + }, + { + "epoch": 2.26, + "learning_rate": 1.2498980227140987e-05, + "loss": 0.9341, + "step": 35056 + }, + { + "epoch": 2.26, + "learning_rate": 1.2498765538118036e-05, + "loss": 0.9041, + "step": 35057 + }, + { + "epoch": 2.26, + "learning_rate": 1.2498550849095085e-05, + "loss": 0.8651, + "step": 35058 + }, + { + "epoch": 2.26, + "learning_rate": 1.2498336160072138e-05, + "loss": 0.9118, + "step": 35059 + }, + { + "epoch": 2.26, + "learning_rate": 1.2498121471049187e-05, + "loss": 0.9484, + "step": 35060 + }, + { + "epoch": 2.26, + "learning_rate": 1.2497906782026237e-05, + "loss": 0.8919, + "step": 35061 + }, + { + "epoch": 2.26, + "learning_rate": 1.2497692093003286e-05, + "loss": 0.899, + "step": 35062 + }, + { + "epoch": 2.26, + "learning_rate": 1.2497477403980335e-05, + "loss": 0.9481, + "step": 35063 + }, + { + "epoch": 2.26, + "learning_rate": 1.2497262714957386e-05, + "loss": 0.9702, + "step": 35064 + }, + { + "epoch": 2.26, + "learning_rate": 1.2497048025934435e-05, + "loss": 0.9487, + "step": 35065 + }, + { + "epoch": 2.26, + "learning_rate": 1.2496833336911486e-05, + "loss": 0.9148, + "step": 35066 + }, + { + "epoch": 2.26, + "learning_rate": 1.2496618647888535e-05, + "loss": 0.8841, + "step": 35067 + }, + { + "epoch": 2.26, + "learning_rate": 1.2496403958865583e-05, + "loss": 0.8467, + "step": 35068 + }, + { + "epoch": 2.26, + "learning_rate": 1.2496189269842634e-05, + "loss": 0.8757, + "step": 35069 + }, + { + "epoch": 2.26, + "learning_rate": 1.2495974580819683e-05, + "loss": 0.9155, + "step": 35070 + }, + { + "epoch": 2.26, + "learning_rate": 1.2495759891796732e-05, + "loss": 0.8668, + "step": 35071 + }, + { + "epoch": 2.26, + "learning_rate": 1.2495545202773783e-05, + "loss": 0.9301, + "step": 35072 + }, + { + "epoch": 2.26, + "learning_rate": 1.2495330513750832e-05, + "loss": 0.8691, + "step": 35073 + }, + { + "epoch": 2.26, + "learning_rate": 1.2495115824727884e-05, + "loss": 0.8763, + "step": 35074 + }, + { + "epoch": 2.26, + "learning_rate": 1.2494901135704933e-05, + "loss": 0.8717, + "step": 35075 + }, + { + "epoch": 2.26, + "learning_rate": 1.2494686446681984e-05, + "loss": 0.9199, + "step": 35076 + }, + { + "epoch": 2.26, + "learning_rate": 1.2494471757659033e-05, + "loss": 0.9003, + "step": 35077 + }, + { + "epoch": 2.26, + "learning_rate": 1.2494257068636082e-05, + "loss": 0.8537, + "step": 35078 + }, + { + "epoch": 2.26, + "learning_rate": 1.2494042379613132e-05, + "loss": 0.8419, + "step": 35079 + }, + { + "epoch": 2.26, + "learning_rate": 1.2493827690590181e-05, + "loss": 0.8819, + "step": 35080 + }, + { + "epoch": 2.26, + "learning_rate": 1.249361300156723e-05, + "loss": 0.903, + "step": 35081 + }, + { + "epoch": 2.26, + "learning_rate": 1.249339831254428e-05, + "loss": 0.8684, + "step": 35082 + }, + { + "epoch": 2.26, + "learning_rate": 1.249318362352133e-05, + "loss": 0.8662, + "step": 35083 + }, + { + "epoch": 2.26, + "learning_rate": 1.2492968934498379e-05, + "loss": 0.9129, + "step": 35084 + }, + { + "epoch": 2.26, + "learning_rate": 1.249275424547543e-05, + "loss": 0.8438, + "step": 35085 + }, + { + "epoch": 2.26, + "learning_rate": 1.2492539556452478e-05, + "loss": 0.8899, + "step": 35086 + }, + { + "epoch": 2.26, + "learning_rate": 1.2492324867429529e-05, + "loss": 0.9047, + "step": 35087 + }, + { + "epoch": 2.26, + "learning_rate": 1.249211017840658e-05, + "loss": 0.8477, + "step": 35088 + }, + { + "epoch": 2.26, + "learning_rate": 1.249189548938363e-05, + "loss": 0.8504, + "step": 35089 + }, + { + "epoch": 2.26, + "learning_rate": 1.249168080036068e-05, + "loss": 0.8723, + "step": 35090 + }, + { + "epoch": 2.26, + "learning_rate": 1.2491466111337728e-05, + "loss": 0.8862, + "step": 35091 + }, + { + "epoch": 2.26, + "learning_rate": 1.2491251422314779e-05, + "loss": 0.8586, + "step": 35092 + }, + { + "epoch": 2.26, + "learning_rate": 1.2491036733291828e-05, + "loss": 0.9238, + "step": 35093 + }, + { + "epoch": 2.26, + "learning_rate": 1.2490822044268877e-05, + "loss": 0.9389, + "step": 35094 + }, + { + "epoch": 2.26, + "learning_rate": 1.2490607355245928e-05, + "loss": 0.8905, + "step": 35095 + }, + { + "epoch": 2.26, + "learning_rate": 1.2490392666222977e-05, + "loss": 0.8834, + "step": 35096 + }, + { + "epoch": 2.26, + "learning_rate": 1.2490177977200027e-05, + "loss": 0.8769, + "step": 35097 + }, + { + "epoch": 2.26, + "learning_rate": 1.2489963288177076e-05, + "loss": 0.8611, + "step": 35098 + }, + { + "epoch": 2.26, + "learning_rate": 1.2489748599154125e-05, + "loss": 0.9238, + "step": 35099 + }, + { + "epoch": 2.26, + "learning_rate": 1.2489533910131176e-05, + "loss": 0.8348, + "step": 35100 + }, + { + "epoch": 2.26, + "learning_rate": 1.2489319221108225e-05, + "loss": 0.848, + "step": 35101 + }, + { + "epoch": 2.26, + "learning_rate": 1.2489104532085274e-05, + "loss": 0.8697, + "step": 35102 + }, + { + "epoch": 2.26, + "learning_rate": 1.2488889843062326e-05, + "loss": 0.9049, + "step": 35103 + }, + { + "epoch": 2.26, + "learning_rate": 1.2488675154039375e-05, + "loss": 0.9281, + "step": 35104 + }, + { + "epoch": 2.26, + "learning_rate": 1.2488460465016426e-05, + "loss": 0.8871, + "step": 35105 + }, + { + "epoch": 2.26, + "learning_rate": 1.2488245775993475e-05, + "loss": 0.9214, + "step": 35106 + }, + { + "epoch": 2.26, + "learning_rate": 1.2488031086970525e-05, + "loss": 0.9093, + "step": 35107 + }, + { + "epoch": 2.26, + "learning_rate": 1.2487816397947574e-05, + "loss": 0.8872, + "step": 35108 + }, + { + "epoch": 2.26, + "learning_rate": 1.2487601708924623e-05, + "loss": 0.9113, + "step": 35109 + }, + { + "epoch": 2.26, + "learning_rate": 1.2487387019901674e-05, + "loss": 0.9031, + "step": 35110 + }, + { + "epoch": 2.26, + "learning_rate": 1.2487172330878723e-05, + "loss": 0.8293, + "step": 35111 + }, + { + "epoch": 2.26, + "learning_rate": 1.2486957641855772e-05, + "loss": 0.896, + "step": 35112 + }, + { + "epoch": 2.26, + "learning_rate": 1.2486742952832822e-05, + "loss": 0.883, + "step": 35113 + }, + { + "epoch": 2.26, + "learning_rate": 1.2486528263809871e-05, + "loss": 0.9226, + "step": 35114 + }, + { + "epoch": 2.26, + "learning_rate": 1.2486313574786922e-05, + "loss": 0.9342, + "step": 35115 + }, + { + "epoch": 2.26, + "learning_rate": 1.2486098885763971e-05, + "loss": 0.9399, + "step": 35116 + }, + { + "epoch": 2.26, + "learning_rate": 1.2485884196741023e-05, + "loss": 0.8656, + "step": 35117 + }, + { + "epoch": 2.26, + "learning_rate": 1.2485669507718072e-05, + "loss": 0.9096, + "step": 35118 + }, + { + "epoch": 2.26, + "learning_rate": 1.2485454818695121e-05, + "loss": 0.9363, + "step": 35119 + }, + { + "epoch": 2.26, + "learning_rate": 1.2485240129672172e-05, + "loss": 0.8822, + "step": 35120 + }, + { + "epoch": 2.26, + "learning_rate": 1.2485025440649221e-05, + "loss": 0.8634, + "step": 35121 + }, + { + "epoch": 2.26, + "learning_rate": 1.248481075162627e-05, + "loss": 0.8684, + "step": 35122 + }, + { + "epoch": 2.26, + "learning_rate": 1.248459606260332e-05, + "loss": 0.9458, + "step": 35123 + }, + { + "epoch": 2.26, + "learning_rate": 1.248438137358037e-05, + "loss": 0.8351, + "step": 35124 + }, + { + "epoch": 2.26, + "learning_rate": 1.248416668455742e-05, + "loss": 0.9408, + "step": 35125 + }, + { + "epoch": 2.26, + "learning_rate": 1.248395199553447e-05, + "loss": 0.8811, + "step": 35126 + }, + { + "epoch": 2.26, + "learning_rate": 1.2483737306511518e-05, + "loss": 0.9348, + "step": 35127 + }, + { + "epoch": 2.26, + "learning_rate": 1.2483522617488569e-05, + "loss": 0.8508, + "step": 35128 + }, + { + "epoch": 2.26, + "learning_rate": 1.2483307928465618e-05, + "loss": 0.8593, + "step": 35129 + }, + { + "epoch": 2.26, + "learning_rate": 1.2483093239442667e-05, + "loss": 0.9213, + "step": 35130 + }, + { + "epoch": 2.26, + "learning_rate": 1.2482878550419717e-05, + "loss": 0.9019, + "step": 35131 + }, + { + "epoch": 2.26, + "learning_rate": 1.2482663861396768e-05, + "loss": 0.9409, + "step": 35132 + }, + { + "epoch": 2.26, + "learning_rate": 1.2482449172373819e-05, + "loss": 0.9308, + "step": 35133 + }, + { + "epoch": 2.26, + "learning_rate": 1.2482234483350868e-05, + "loss": 0.9551, + "step": 35134 + }, + { + "epoch": 2.26, + "learning_rate": 1.2482019794327918e-05, + "loss": 0.8698, + "step": 35135 + }, + { + "epoch": 2.26, + "learning_rate": 1.2481805105304967e-05, + "loss": 0.8622, + "step": 35136 + }, + { + "epoch": 2.26, + "learning_rate": 1.2481590416282016e-05, + "loss": 0.8524, + "step": 35137 + }, + { + "epoch": 2.26, + "learning_rate": 1.2481375727259067e-05, + "loss": 0.8401, + "step": 35138 + }, + { + "epoch": 2.26, + "learning_rate": 1.2481161038236116e-05, + "loss": 0.9949, + "step": 35139 + }, + { + "epoch": 2.26, + "learning_rate": 1.2480946349213165e-05, + "loss": 0.8187, + "step": 35140 + }, + { + "epoch": 2.26, + "learning_rate": 1.2480731660190216e-05, + "loss": 0.8995, + "step": 35141 + }, + { + "epoch": 2.26, + "learning_rate": 1.2480516971167264e-05, + "loss": 0.8402, + "step": 35142 + }, + { + "epoch": 2.26, + "learning_rate": 1.2480302282144313e-05, + "loss": 0.8762, + "step": 35143 + }, + { + "epoch": 2.26, + "learning_rate": 1.2480087593121364e-05, + "loss": 0.912, + "step": 35144 + }, + { + "epoch": 2.26, + "learning_rate": 1.2479872904098413e-05, + "loss": 0.8851, + "step": 35145 + }, + { + "epoch": 2.26, + "learning_rate": 1.2479658215075465e-05, + "loss": 0.895, + "step": 35146 + }, + { + "epoch": 2.26, + "learning_rate": 1.2479443526052514e-05, + "loss": 0.9149, + "step": 35147 + }, + { + "epoch": 2.26, + "learning_rate": 1.2479228837029565e-05, + "loss": 0.8974, + "step": 35148 + }, + { + "epoch": 2.26, + "learning_rate": 1.2479014148006614e-05, + "loss": 0.9063, + "step": 35149 + }, + { + "epoch": 2.26, + "learning_rate": 1.2478799458983663e-05, + "loss": 0.9466, + "step": 35150 + }, + { + "epoch": 2.26, + "learning_rate": 1.2478584769960714e-05, + "loss": 0.8876, + "step": 35151 + }, + { + "epoch": 2.26, + "learning_rate": 1.2478370080937763e-05, + "loss": 0.9844, + "step": 35152 + }, + { + "epoch": 2.26, + "learning_rate": 1.2478155391914812e-05, + "loss": 0.7938, + "step": 35153 + }, + { + "epoch": 2.26, + "learning_rate": 1.2477940702891862e-05, + "loss": 0.9467, + "step": 35154 + }, + { + "epoch": 2.26, + "learning_rate": 1.2477726013868911e-05, + "loss": 0.8756, + "step": 35155 + }, + { + "epoch": 2.26, + "learning_rate": 1.2477511324845962e-05, + "loss": 0.9134, + "step": 35156 + }, + { + "epoch": 2.26, + "learning_rate": 1.247729663582301e-05, + "loss": 0.8794, + "step": 35157 + }, + { + "epoch": 2.26, + "learning_rate": 1.247708194680006e-05, + "loss": 0.8498, + "step": 35158 + }, + { + "epoch": 2.26, + "learning_rate": 1.247686725777711e-05, + "loss": 0.8651, + "step": 35159 + }, + { + "epoch": 2.26, + "learning_rate": 1.247665256875416e-05, + "loss": 0.94, + "step": 35160 + }, + { + "epoch": 2.26, + "learning_rate": 1.2476437879731212e-05, + "loss": 0.9465, + "step": 35161 + }, + { + "epoch": 2.26, + "learning_rate": 1.247622319070826e-05, + "loss": 0.9079, + "step": 35162 + }, + { + "epoch": 2.26, + "learning_rate": 1.2476008501685311e-05, + "loss": 0.8758, + "step": 35163 + }, + { + "epoch": 2.26, + "learning_rate": 1.247579381266236e-05, + "loss": 0.8647, + "step": 35164 + }, + { + "epoch": 2.26, + "learning_rate": 1.247557912363941e-05, + "loss": 0.8968, + "step": 35165 + }, + { + "epoch": 2.26, + "learning_rate": 1.247536443461646e-05, + "loss": 0.8881, + "step": 35166 + }, + { + "epoch": 2.26, + "learning_rate": 1.2475149745593509e-05, + "loss": 0.9404, + "step": 35167 + }, + { + "epoch": 2.26, + "learning_rate": 1.2474935056570558e-05, + "loss": 0.95, + "step": 35168 + }, + { + "epoch": 2.26, + "learning_rate": 1.2474720367547609e-05, + "loss": 0.9227, + "step": 35169 + }, + { + "epoch": 2.26, + "learning_rate": 1.2474505678524658e-05, + "loss": 0.8436, + "step": 35170 + }, + { + "epoch": 2.26, + "learning_rate": 1.2474290989501706e-05, + "loss": 0.8933, + "step": 35171 + }, + { + "epoch": 2.26, + "learning_rate": 1.2474076300478757e-05, + "loss": 0.8577, + "step": 35172 + }, + { + "epoch": 2.26, + "learning_rate": 1.2473861611455806e-05, + "loss": 0.8991, + "step": 35173 + }, + { + "epoch": 2.26, + "learning_rate": 1.2473646922432857e-05, + "loss": 0.9659, + "step": 35174 + }, + { + "epoch": 2.26, + "learning_rate": 1.2473432233409907e-05, + "loss": 0.8976, + "step": 35175 + }, + { + "epoch": 2.26, + "learning_rate": 1.2473217544386958e-05, + "loss": 0.921, + "step": 35176 + }, + { + "epoch": 2.26, + "learning_rate": 1.2473002855364007e-05, + "loss": 0.8974, + "step": 35177 + }, + { + "epoch": 2.26, + "learning_rate": 1.2472788166341056e-05, + "loss": 0.9101, + "step": 35178 + }, + { + "epoch": 2.26, + "learning_rate": 1.2472573477318107e-05, + "loss": 0.8727, + "step": 35179 + }, + { + "epoch": 2.26, + "learning_rate": 1.2472358788295156e-05, + "loss": 0.8744, + "step": 35180 + }, + { + "epoch": 2.26, + "learning_rate": 1.2472144099272205e-05, + "loss": 0.9215, + "step": 35181 + }, + { + "epoch": 2.26, + "learning_rate": 1.2471929410249255e-05, + "loss": 0.9045, + "step": 35182 + }, + { + "epoch": 2.26, + "learning_rate": 1.2471714721226304e-05, + "loss": 0.927, + "step": 35183 + }, + { + "epoch": 2.26, + "learning_rate": 1.2471500032203355e-05, + "loss": 0.8764, + "step": 35184 + }, + { + "epoch": 2.26, + "learning_rate": 1.2471285343180404e-05, + "loss": 0.8948, + "step": 35185 + }, + { + "epoch": 2.26, + "learning_rate": 1.2471070654157453e-05, + "loss": 0.8815, + "step": 35186 + }, + { + "epoch": 2.26, + "learning_rate": 1.2470855965134503e-05, + "loss": 0.823, + "step": 35187 + }, + { + "epoch": 2.26, + "learning_rate": 1.2470641276111552e-05, + "loss": 0.8695, + "step": 35188 + }, + { + "epoch": 2.26, + "learning_rate": 1.2470426587088601e-05, + "loss": 0.9482, + "step": 35189 + }, + { + "epoch": 2.26, + "learning_rate": 1.2470211898065654e-05, + "loss": 0.8575, + "step": 35190 + }, + { + "epoch": 2.26, + "learning_rate": 1.2469997209042703e-05, + "loss": 0.928, + "step": 35191 + }, + { + "epoch": 2.26, + "learning_rate": 1.2469782520019753e-05, + "loss": 0.9266, + "step": 35192 + }, + { + "epoch": 2.26, + "learning_rate": 1.2469567830996802e-05, + "loss": 0.8703, + "step": 35193 + }, + { + "epoch": 2.26, + "learning_rate": 1.2469353141973853e-05, + "loss": 0.8453, + "step": 35194 + }, + { + "epoch": 2.26, + "learning_rate": 1.2469138452950902e-05, + "loss": 0.8284, + "step": 35195 + }, + { + "epoch": 2.26, + "learning_rate": 1.2468923763927951e-05, + "loss": 0.9152, + "step": 35196 + }, + { + "epoch": 2.26, + "learning_rate": 1.2468709074905002e-05, + "loss": 0.8815, + "step": 35197 + }, + { + "epoch": 2.26, + "learning_rate": 1.246849438588205e-05, + "loss": 0.9448, + "step": 35198 + }, + { + "epoch": 2.26, + "learning_rate": 1.24682796968591e-05, + "loss": 0.9021, + "step": 35199 + }, + { + "epoch": 2.26, + "learning_rate": 1.246806500783615e-05, + "loss": 0.9279, + "step": 35200 + }, + { + "epoch": 2.26, + "learning_rate": 1.2467850318813199e-05, + "loss": 0.9026, + "step": 35201 + }, + { + "epoch": 2.26, + "learning_rate": 1.246763562979025e-05, + "loss": 0.8298, + "step": 35202 + }, + { + "epoch": 2.26, + "learning_rate": 1.2467420940767299e-05, + "loss": 0.9581, + "step": 35203 + }, + { + "epoch": 2.26, + "learning_rate": 1.2467206251744351e-05, + "loss": 0.9037, + "step": 35204 + }, + { + "epoch": 2.27, + "learning_rate": 1.24669915627214e-05, + "loss": 0.8899, + "step": 35205 + }, + { + "epoch": 2.27, + "learning_rate": 1.2466776873698449e-05, + "loss": 0.9252, + "step": 35206 + }, + { + "epoch": 2.27, + "learning_rate": 1.24665621846755e-05, + "loss": 0.9408, + "step": 35207 + }, + { + "epoch": 2.27, + "learning_rate": 1.2466347495652549e-05, + "loss": 0.8373, + "step": 35208 + }, + { + "epoch": 2.27, + "learning_rate": 1.2466132806629598e-05, + "loss": 0.9854, + "step": 35209 + }, + { + "epoch": 2.27, + "learning_rate": 1.2465918117606648e-05, + "loss": 0.857, + "step": 35210 + }, + { + "epoch": 2.27, + "learning_rate": 1.2465703428583697e-05, + "loss": 0.8934, + "step": 35211 + }, + { + "epoch": 2.27, + "learning_rate": 1.2465488739560748e-05, + "loss": 0.9191, + "step": 35212 + }, + { + "epoch": 2.27, + "learning_rate": 1.2465274050537797e-05, + "loss": 0.8901, + "step": 35213 + }, + { + "epoch": 2.27, + "learning_rate": 1.2465059361514846e-05, + "loss": 0.8977, + "step": 35214 + }, + { + "epoch": 2.27, + "learning_rate": 1.2464844672491897e-05, + "loss": 0.9807, + "step": 35215 + }, + { + "epoch": 2.27, + "learning_rate": 1.2464629983468945e-05, + "loss": 0.8702, + "step": 35216 + }, + { + "epoch": 2.27, + "learning_rate": 1.2464415294445994e-05, + "loss": 0.9062, + "step": 35217 + }, + { + "epoch": 2.27, + "learning_rate": 1.2464200605423045e-05, + "loss": 0.8492, + "step": 35218 + }, + { + "epoch": 2.27, + "learning_rate": 1.2463985916400096e-05, + "loss": 0.9039, + "step": 35219 + }, + { + "epoch": 2.27, + "learning_rate": 1.2463771227377146e-05, + "loss": 0.8591, + "step": 35220 + }, + { + "epoch": 2.27, + "learning_rate": 1.2463556538354195e-05, + "loss": 0.9296, + "step": 35221 + }, + { + "epoch": 2.27, + "learning_rate": 1.2463341849331246e-05, + "loss": 0.9421, + "step": 35222 + }, + { + "epoch": 2.27, + "learning_rate": 1.2463127160308295e-05, + "loss": 0.8307, + "step": 35223 + }, + { + "epoch": 2.27, + "learning_rate": 1.2462912471285344e-05, + "loss": 0.8392, + "step": 35224 + }, + { + "epoch": 2.27, + "learning_rate": 1.2462697782262395e-05, + "loss": 0.8603, + "step": 35225 + }, + { + "epoch": 2.27, + "learning_rate": 1.2462483093239444e-05, + "loss": 0.9808, + "step": 35226 + }, + { + "epoch": 2.27, + "learning_rate": 1.2462268404216493e-05, + "loss": 0.9089, + "step": 35227 + }, + { + "epoch": 2.27, + "learning_rate": 1.2462053715193543e-05, + "loss": 0.8562, + "step": 35228 + }, + { + "epoch": 2.27, + "learning_rate": 1.2461839026170592e-05, + "loss": 0.8568, + "step": 35229 + }, + { + "epoch": 2.27, + "learning_rate": 1.2461624337147641e-05, + "loss": 0.8876, + "step": 35230 + }, + { + "epoch": 2.27, + "learning_rate": 1.2461409648124692e-05, + "loss": 0.8981, + "step": 35231 + }, + { + "epoch": 2.27, + "learning_rate": 1.246119495910174e-05, + "loss": 0.8833, + "step": 35232 + }, + { + "epoch": 2.27, + "learning_rate": 1.2460980270078793e-05, + "loss": 0.9422, + "step": 35233 + }, + { + "epoch": 2.27, + "learning_rate": 1.2460765581055842e-05, + "loss": 0.9352, + "step": 35234 + }, + { + "epoch": 2.27, + "learning_rate": 1.2460550892032893e-05, + "loss": 0.8919, + "step": 35235 + }, + { + "epoch": 2.27, + "learning_rate": 1.2460336203009942e-05, + "loss": 0.8303, + "step": 35236 + }, + { + "epoch": 2.27, + "learning_rate": 1.246012151398699e-05, + "loss": 0.8537, + "step": 35237 + }, + { + "epoch": 2.27, + "learning_rate": 1.2459906824964041e-05, + "loss": 0.8377, + "step": 35238 + }, + { + "epoch": 2.27, + "learning_rate": 1.245969213594109e-05, + "loss": 0.9035, + "step": 35239 + }, + { + "epoch": 2.27, + "learning_rate": 1.245947744691814e-05, + "loss": 0.8612, + "step": 35240 + }, + { + "epoch": 2.27, + "learning_rate": 1.245926275789519e-05, + "loss": 0.932, + "step": 35241 + }, + { + "epoch": 2.27, + "learning_rate": 1.2459048068872239e-05, + "loss": 0.9302, + "step": 35242 + }, + { + "epoch": 2.27, + "learning_rate": 1.245883337984929e-05, + "loss": 0.8935, + "step": 35243 + }, + { + "epoch": 2.27, + "learning_rate": 1.2458618690826339e-05, + "loss": 0.8506, + "step": 35244 + }, + { + "epoch": 2.27, + "learning_rate": 1.2458404001803387e-05, + "loss": 0.9672, + "step": 35245 + }, + { + "epoch": 2.27, + "learning_rate": 1.2458189312780438e-05, + "loss": 0.8718, + "step": 35246 + }, + { + "epoch": 2.27, + "learning_rate": 1.2457974623757487e-05, + "loss": 0.8453, + "step": 35247 + }, + { + "epoch": 2.27, + "learning_rate": 1.245775993473454e-05, + "loss": 0.9068, + "step": 35248 + }, + { + "epoch": 2.27, + "learning_rate": 1.2457545245711588e-05, + "loss": 0.9474, + "step": 35249 + }, + { + "epoch": 2.27, + "learning_rate": 1.2457330556688637e-05, + "loss": 0.8214, + "step": 35250 + }, + { + "epoch": 2.27, + "learning_rate": 1.2457115867665688e-05, + "loss": 0.8791, + "step": 35251 + }, + { + "epoch": 2.27, + "learning_rate": 1.2456901178642737e-05, + "loss": 0.9413, + "step": 35252 + }, + { + "epoch": 2.27, + "learning_rate": 1.2456686489619788e-05, + "loss": 0.9184, + "step": 35253 + }, + { + "epoch": 2.27, + "learning_rate": 1.2456471800596837e-05, + "loss": 0.8867, + "step": 35254 + }, + { + "epoch": 2.27, + "learning_rate": 1.2456257111573886e-05, + "loss": 0.9328, + "step": 35255 + }, + { + "epoch": 2.27, + "learning_rate": 1.2456042422550936e-05, + "loss": 0.9699, + "step": 35256 + }, + { + "epoch": 2.27, + "learning_rate": 1.2455827733527985e-05, + "loss": 0.8599, + "step": 35257 + }, + { + "epoch": 2.27, + "learning_rate": 1.2455613044505034e-05, + "loss": 0.9068, + "step": 35258 + }, + { + "epoch": 2.27, + "learning_rate": 1.2455398355482085e-05, + "loss": 0.918, + "step": 35259 + }, + { + "epoch": 2.27, + "learning_rate": 1.2455183666459134e-05, + "loss": 0.9112, + "step": 35260 + }, + { + "epoch": 2.27, + "learning_rate": 1.2454968977436184e-05, + "loss": 0.9387, + "step": 35261 + }, + { + "epoch": 2.27, + "learning_rate": 1.2454754288413235e-05, + "loss": 0.898, + "step": 35262 + }, + { + "epoch": 2.27, + "learning_rate": 1.2454539599390286e-05, + "loss": 0.8357, + "step": 35263 + }, + { + "epoch": 2.27, + "learning_rate": 1.2454324910367335e-05, + "loss": 0.8559, + "step": 35264 + }, + { + "epoch": 2.27, + "learning_rate": 1.2454110221344384e-05, + "loss": 0.9065, + "step": 35265 + }, + { + "epoch": 2.27, + "learning_rate": 1.2453895532321434e-05, + "loss": 0.969, + "step": 35266 + }, + { + "epoch": 2.27, + "learning_rate": 1.2453680843298483e-05, + "loss": 0.9384, + "step": 35267 + }, + { + "epoch": 2.27, + "learning_rate": 1.2453466154275532e-05, + "loss": 0.9108, + "step": 35268 + }, + { + "epoch": 2.27, + "learning_rate": 1.2453251465252583e-05, + "loss": 0.8593, + "step": 35269 + }, + { + "epoch": 2.27, + "learning_rate": 1.2453036776229632e-05, + "loss": 0.9308, + "step": 35270 + }, + { + "epoch": 2.27, + "learning_rate": 1.2452822087206683e-05, + "loss": 0.9355, + "step": 35271 + }, + { + "epoch": 2.27, + "learning_rate": 1.2452607398183732e-05, + "loss": 0.8532, + "step": 35272 + }, + { + "epoch": 2.27, + "learning_rate": 1.245239270916078e-05, + "loss": 0.876, + "step": 35273 + }, + { + "epoch": 2.27, + "learning_rate": 1.2452178020137831e-05, + "loss": 0.8968, + "step": 35274 + }, + { + "epoch": 2.27, + "learning_rate": 1.245196333111488e-05, + "loss": 0.9086, + "step": 35275 + }, + { + "epoch": 2.27, + "learning_rate": 1.2451748642091929e-05, + "loss": 0.8946, + "step": 35276 + }, + { + "epoch": 2.27, + "learning_rate": 1.2451533953068981e-05, + "loss": 0.9346, + "step": 35277 + }, + { + "epoch": 2.27, + "learning_rate": 1.245131926404603e-05, + "loss": 0.9434, + "step": 35278 + }, + { + "epoch": 2.27, + "learning_rate": 1.2451104575023081e-05, + "loss": 0.9218, + "step": 35279 + }, + { + "epoch": 2.27, + "learning_rate": 1.245088988600013e-05, + "loss": 0.8956, + "step": 35280 + }, + { + "epoch": 2.27, + "learning_rate": 1.245067519697718e-05, + "loss": 0.8976, + "step": 35281 + }, + { + "epoch": 2.27, + "learning_rate": 1.245046050795423e-05, + "loss": 0.8854, + "step": 35282 + }, + { + "epoch": 2.27, + "learning_rate": 1.2450245818931279e-05, + "loss": 0.9106, + "step": 35283 + }, + { + "epoch": 2.27, + "learning_rate": 1.245003112990833e-05, + "loss": 0.9166, + "step": 35284 + }, + { + "epoch": 2.27, + "learning_rate": 1.2449816440885378e-05, + "loss": 0.9277, + "step": 35285 + }, + { + "epoch": 2.27, + "learning_rate": 1.2449601751862427e-05, + "loss": 0.9628, + "step": 35286 + }, + { + "epoch": 2.27, + "learning_rate": 1.2449387062839478e-05, + "loss": 0.9307, + "step": 35287 + }, + { + "epoch": 2.27, + "learning_rate": 1.2449172373816527e-05, + "loss": 0.8836, + "step": 35288 + }, + { + "epoch": 2.27, + "learning_rate": 1.2448957684793578e-05, + "loss": 0.941, + "step": 35289 + }, + { + "epoch": 2.27, + "learning_rate": 1.2448742995770626e-05, + "loss": 0.8486, + "step": 35290 + }, + { + "epoch": 2.27, + "learning_rate": 1.2448528306747679e-05, + "loss": 0.8803, + "step": 35291 + }, + { + "epoch": 2.27, + "learning_rate": 1.2448313617724728e-05, + "loss": 0.9584, + "step": 35292 + }, + { + "epoch": 2.27, + "learning_rate": 1.2448098928701777e-05, + "loss": 0.9326, + "step": 35293 + }, + { + "epoch": 2.27, + "learning_rate": 1.2447884239678827e-05, + "loss": 0.8988, + "step": 35294 + }, + { + "epoch": 2.27, + "learning_rate": 1.2447669550655876e-05, + "loss": 0.8968, + "step": 35295 + }, + { + "epoch": 2.27, + "learning_rate": 1.2447454861632925e-05, + "loss": 0.8436, + "step": 35296 + }, + { + "epoch": 2.27, + "learning_rate": 1.2447240172609976e-05, + "loss": 0.8911, + "step": 35297 + }, + { + "epoch": 2.27, + "learning_rate": 1.2447025483587025e-05, + "loss": 0.874, + "step": 35298 + }, + { + "epoch": 2.27, + "learning_rate": 1.2446810794564076e-05, + "loss": 0.958, + "step": 35299 + }, + { + "epoch": 2.27, + "learning_rate": 1.2446596105541125e-05, + "loss": 0.9494, + "step": 35300 + }, + { + "epoch": 2.27, + "learning_rate": 1.2446381416518174e-05, + "loss": 0.8558, + "step": 35301 + }, + { + "epoch": 2.27, + "learning_rate": 1.2446166727495224e-05, + "loss": 0.9498, + "step": 35302 + }, + { + "epoch": 2.27, + "learning_rate": 1.2445952038472273e-05, + "loss": 0.9274, + "step": 35303 + }, + { + "epoch": 2.27, + "learning_rate": 1.2445737349449322e-05, + "loss": 0.8646, + "step": 35304 + }, + { + "epoch": 2.27, + "learning_rate": 1.2445522660426373e-05, + "loss": 0.935, + "step": 35305 + }, + { + "epoch": 2.27, + "learning_rate": 1.2445307971403423e-05, + "loss": 0.8977, + "step": 35306 + }, + { + "epoch": 2.27, + "learning_rate": 1.2445093282380474e-05, + "loss": 0.8456, + "step": 35307 + }, + { + "epoch": 2.27, + "learning_rate": 1.2444878593357523e-05, + "loss": 0.9359, + "step": 35308 + }, + { + "epoch": 2.27, + "learning_rate": 1.2444663904334574e-05, + "loss": 0.9218, + "step": 35309 + }, + { + "epoch": 2.27, + "learning_rate": 1.2444449215311623e-05, + "loss": 0.9119, + "step": 35310 + }, + { + "epoch": 2.27, + "learning_rate": 1.2444234526288672e-05, + "loss": 0.9093, + "step": 35311 + }, + { + "epoch": 2.27, + "learning_rate": 1.2444019837265722e-05, + "loss": 0.914, + "step": 35312 + }, + { + "epoch": 2.27, + "learning_rate": 1.2443805148242771e-05, + "loss": 0.9333, + "step": 35313 + }, + { + "epoch": 2.27, + "learning_rate": 1.244359045921982e-05, + "loss": 0.8739, + "step": 35314 + }, + { + "epoch": 2.27, + "learning_rate": 1.2443375770196871e-05, + "loss": 0.8894, + "step": 35315 + }, + { + "epoch": 2.27, + "learning_rate": 1.244316108117392e-05, + "loss": 0.9156, + "step": 35316 + }, + { + "epoch": 2.27, + "learning_rate": 1.2442946392150969e-05, + "loss": 0.9285, + "step": 35317 + }, + { + "epoch": 2.27, + "learning_rate": 1.244273170312802e-05, + "loss": 0.8661, + "step": 35318 + }, + { + "epoch": 2.27, + "learning_rate": 1.2442517014105068e-05, + "loss": 0.9559, + "step": 35319 + }, + { + "epoch": 2.27, + "learning_rate": 1.244230232508212e-05, + "loss": 0.8527, + "step": 35320 + }, + { + "epoch": 2.27, + "learning_rate": 1.244208763605917e-05, + "loss": 0.9217, + "step": 35321 + }, + { + "epoch": 2.27, + "learning_rate": 1.244187294703622e-05, + "loss": 0.9436, + "step": 35322 + }, + { + "epoch": 2.27, + "learning_rate": 1.244165825801327e-05, + "loss": 0.8481, + "step": 35323 + }, + { + "epoch": 2.27, + "learning_rate": 1.2441443568990318e-05, + "loss": 0.8799, + "step": 35324 + }, + { + "epoch": 2.27, + "learning_rate": 1.2441228879967369e-05, + "loss": 0.8978, + "step": 35325 + }, + { + "epoch": 2.27, + "learning_rate": 1.2441014190944418e-05, + "loss": 0.8934, + "step": 35326 + }, + { + "epoch": 2.27, + "learning_rate": 1.2440799501921467e-05, + "loss": 0.8499, + "step": 35327 + }, + { + "epoch": 2.27, + "learning_rate": 1.2440584812898518e-05, + "loss": 0.9463, + "step": 35328 + }, + { + "epoch": 2.27, + "learning_rate": 1.2440370123875567e-05, + "loss": 0.8645, + "step": 35329 + }, + { + "epoch": 2.27, + "learning_rate": 1.2440155434852617e-05, + "loss": 0.8509, + "step": 35330 + }, + { + "epoch": 2.27, + "learning_rate": 1.2439940745829666e-05, + "loss": 0.9289, + "step": 35331 + }, + { + "epoch": 2.27, + "learning_rate": 1.2439726056806715e-05, + "loss": 1.014, + "step": 35332 + }, + { + "epoch": 2.27, + "learning_rate": 1.2439511367783766e-05, + "loss": 0.9172, + "step": 35333 + }, + { + "epoch": 2.27, + "learning_rate": 1.2439296678760815e-05, + "loss": 0.8754, + "step": 35334 + }, + { + "epoch": 2.27, + "learning_rate": 1.2439081989737867e-05, + "loss": 0.876, + "step": 35335 + }, + { + "epoch": 2.27, + "learning_rate": 1.2438867300714916e-05, + "loss": 0.9045, + "step": 35336 + }, + { + "epoch": 2.27, + "learning_rate": 1.2438652611691965e-05, + "loss": 0.9164, + "step": 35337 + }, + { + "epoch": 2.27, + "learning_rate": 1.2438437922669016e-05, + "loss": 0.8891, + "step": 35338 + }, + { + "epoch": 2.27, + "learning_rate": 1.2438223233646065e-05, + "loss": 0.8563, + "step": 35339 + }, + { + "epoch": 2.27, + "learning_rate": 1.2438008544623115e-05, + "loss": 0.9691, + "step": 35340 + }, + { + "epoch": 2.27, + "learning_rate": 1.2437793855600164e-05, + "loss": 0.9179, + "step": 35341 + }, + { + "epoch": 2.27, + "learning_rate": 1.2437579166577213e-05, + "loss": 0.9356, + "step": 35342 + }, + { + "epoch": 2.27, + "learning_rate": 1.2437364477554264e-05, + "loss": 0.9796, + "step": 35343 + }, + { + "epoch": 2.27, + "learning_rate": 1.2437149788531313e-05, + "loss": 0.922, + "step": 35344 + }, + { + "epoch": 2.27, + "learning_rate": 1.2436935099508362e-05, + "loss": 0.8345, + "step": 35345 + }, + { + "epoch": 2.27, + "learning_rate": 1.2436720410485413e-05, + "loss": 0.9297, + "step": 35346 + }, + { + "epoch": 2.27, + "learning_rate": 1.2436505721462462e-05, + "loss": 0.9089, + "step": 35347 + }, + { + "epoch": 2.27, + "learning_rate": 1.2436291032439512e-05, + "loss": 0.9531, + "step": 35348 + }, + { + "epoch": 2.27, + "learning_rate": 1.2436076343416563e-05, + "loss": 0.9087, + "step": 35349 + }, + { + "epoch": 2.27, + "learning_rate": 1.2435861654393614e-05, + "loss": 0.8191, + "step": 35350 + }, + { + "epoch": 2.27, + "learning_rate": 1.2435646965370662e-05, + "loss": 0.8672, + "step": 35351 + }, + { + "epoch": 2.27, + "learning_rate": 1.2435432276347711e-05, + "loss": 0.9455, + "step": 35352 + }, + { + "epoch": 2.27, + "learning_rate": 1.2435217587324762e-05, + "loss": 0.7893, + "step": 35353 + }, + { + "epoch": 2.27, + "learning_rate": 1.2435002898301811e-05, + "loss": 0.9638, + "step": 35354 + }, + { + "epoch": 2.27, + "learning_rate": 1.243478820927886e-05, + "loss": 0.9723, + "step": 35355 + }, + { + "epoch": 2.27, + "learning_rate": 1.243457352025591e-05, + "loss": 0.9072, + "step": 35356 + }, + { + "epoch": 2.27, + "learning_rate": 1.243435883123296e-05, + "loss": 0.919, + "step": 35357 + }, + { + "epoch": 2.27, + "learning_rate": 1.243414414221001e-05, + "loss": 0.8593, + "step": 35358 + }, + { + "epoch": 2.27, + "learning_rate": 1.243392945318706e-05, + "loss": 0.9396, + "step": 35359 + }, + { + "epoch": 2.27, + "learning_rate": 1.2433714764164108e-05, + "loss": 0.8615, + "step": 35360 + }, + { + "epoch": 2.28, + "learning_rate": 1.2433500075141159e-05, + "loss": 0.87, + "step": 35361 + }, + { + "epoch": 2.28, + "learning_rate": 1.2433285386118208e-05, + "loss": 0.9435, + "step": 35362 + }, + { + "epoch": 2.28, + "learning_rate": 1.2433070697095257e-05, + "loss": 0.8625, + "step": 35363 + }, + { + "epoch": 2.28, + "learning_rate": 1.243285600807231e-05, + "loss": 0.9128, + "step": 35364 + }, + { + "epoch": 2.28, + "learning_rate": 1.2432641319049358e-05, + "loss": 0.8892, + "step": 35365 + }, + { + "epoch": 2.28, + "learning_rate": 1.2432426630026409e-05, + "loss": 0.8562, + "step": 35366 + }, + { + "epoch": 2.28, + "learning_rate": 1.2432211941003458e-05, + "loss": 0.901, + "step": 35367 + }, + { + "epoch": 2.28, + "learning_rate": 1.2431997251980508e-05, + "loss": 0.7992, + "step": 35368 + }, + { + "epoch": 2.28, + "learning_rate": 1.2431782562957557e-05, + "loss": 0.9089, + "step": 35369 + }, + { + "epoch": 2.28, + "learning_rate": 1.2431567873934606e-05, + "loss": 0.8239, + "step": 35370 + }, + { + "epoch": 2.28, + "learning_rate": 1.2431353184911657e-05, + "loss": 0.855, + "step": 35371 + }, + { + "epoch": 2.28, + "learning_rate": 1.2431138495888706e-05, + "loss": 0.8997, + "step": 35372 + }, + { + "epoch": 2.28, + "learning_rate": 1.2430923806865755e-05, + "loss": 0.9, + "step": 35373 + }, + { + "epoch": 2.28, + "learning_rate": 1.2430709117842806e-05, + "loss": 0.8858, + "step": 35374 + }, + { + "epoch": 2.28, + "learning_rate": 1.2430494428819855e-05, + "loss": 0.8809, + "step": 35375 + }, + { + "epoch": 2.28, + "learning_rate": 1.2430279739796905e-05, + "loss": 0.9078, + "step": 35376 + }, + { + "epoch": 2.28, + "learning_rate": 1.2430065050773954e-05, + "loss": 0.9327, + "step": 35377 + }, + { + "epoch": 2.28, + "learning_rate": 1.2429850361751007e-05, + "loss": 0.9326, + "step": 35378 + }, + { + "epoch": 2.28, + "learning_rate": 1.2429635672728056e-05, + "loss": 0.9355, + "step": 35379 + }, + { + "epoch": 2.28, + "learning_rate": 1.2429420983705104e-05, + "loss": 0.922, + "step": 35380 + }, + { + "epoch": 2.28, + "learning_rate": 1.2429206294682155e-05, + "loss": 0.9672, + "step": 35381 + }, + { + "epoch": 2.28, + "learning_rate": 1.2428991605659204e-05, + "loss": 0.8797, + "step": 35382 + }, + { + "epoch": 2.28, + "learning_rate": 1.2428776916636253e-05, + "loss": 0.8988, + "step": 35383 + }, + { + "epoch": 2.28, + "learning_rate": 1.2428562227613304e-05, + "loss": 0.8668, + "step": 35384 + }, + { + "epoch": 2.28, + "learning_rate": 1.2428347538590353e-05, + "loss": 0.8815, + "step": 35385 + }, + { + "epoch": 2.28, + "learning_rate": 1.2428132849567403e-05, + "loss": 0.8941, + "step": 35386 + }, + { + "epoch": 2.28, + "learning_rate": 1.2427918160544452e-05, + "loss": 0.8943, + "step": 35387 + }, + { + "epoch": 2.28, + "learning_rate": 1.2427703471521501e-05, + "loss": 0.9238, + "step": 35388 + }, + { + "epoch": 2.28, + "learning_rate": 1.2427488782498552e-05, + "loss": 0.8787, + "step": 35389 + }, + { + "epoch": 2.28, + "learning_rate": 1.2427274093475601e-05, + "loss": 0.9799, + "step": 35390 + }, + { + "epoch": 2.28, + "learning_rate": 1.242705940445265e-05, + "loss": 0.828, + "step": 35391 + }, + { + "epoch": 2.28, + "learning_rate": 1.24268447154297e-05, + "loss": 0.8635, + "step": 35392 + }, + { + "epoch": 2.28, + "learning_rate": 1.2426630026406751e-05, + "loss": 0.9235, + "step": 35393 + }, + { + "epoch": 2.28, + "learning_rate": 1.2426415337383802e-05, + "loss": 0.8528, + "step": 35394 + }, + { + "epoch": 2.28, + "learning_rate": 1.242620064836085e-05, + "loss": 0.8941, + "step": 35395 + }, + { + "epoch": 2.28, + "learning_rate": 1.2425985959337901e-05, + "loss": 0.8857, + "step": 35396 + }, + { + "epoch": 2.28, + "learning_rate": 1.242577127031495e-05, + "loss": 0.8461, + "step": 35397 + }, + { + "epoch": 2.28, + "learning_rate": 1.2425556581292e-05, + "loss": 0.9354, + "step": 35398 + }, + { + "epoch": 2.28, + "learning_rate": 1.242534189226905e-05, + "loss": 0.8794, + "step": 35399 + }, + { + "epoch": 2.28, + "learning_rate": 1.2425127203246099e-05, + "loss": 0.833, + "step": 35400 + }, + { + "epoch": 2.28, + "learning_rate": 1.2424912514223148e-05, + "loss": 0.8829, + "step": 35401 + }, + { + "epoch": 2.28, + "learning_rate": 1.2424697825200199e-05, + "loss": 0.9181, + "step": 35402 + }, + { + "epoch": 2.28, + "learning_rate": 1.2424483136177248e-05, + "loss": 0.9462, + "step": 35403 + }, + { + "epoch": 2.28, + "learning_rate": 1.2424268447154297e-05, + "loss": 0.9824, + "step": 35404 + }, + { + "epoch": 2.28, + "learning_rate": 1.2424053758131347e-05, + "loss": 0.8558, + "step": 35405 + }, + { + "epoch": 2.28, + "learning_rate": 1.2423839069108396e-05, + "loss": 0.9284, + "step": 35406 + }, + { + "epoch": 2.28, + "learning_rate": 1.2423624380085449e-05, + "loss": 0.8518, + "step": 35407 + }, + { + "epoch": 2.28, + "learning_rate": 1.2423409691062498e-05, + "loss": 0.9012, + "step": 35408 + }, + { + "epoch": 2.28, + "learning_rate": 1.2423195002039548e-05, + "loss": 0.9215, + "step": 35409 + }, + { + "epoch": 2.28, + "learning_rate": 1.2422980313016597e-05, + "loss": 0.8973, + "step": 35410 + }, + { + "epoch": 2.28, + "learning_rate": 1.2422765623993646e-05, + "loss": 0.8893, + "step": 35411 + }, + { + "epoch": 2.28, + "learning_rate": 1.2422550934970697e-05, + "loss": 0.8641, + "step": 35412 + }, + { + "epoch": 2.28, + "learning_rate": 1.2422336245947746e-05, + "loss": 0.8852, + "step": 35413 + }, + { + "epoch": 2.28, + "learning_rate": 1.2422121556924795e-05, + "loss": 0.9554, + "step": 35414 + }, + { + "epoch": 2.28, + "learning_rate": 1.2421906867901845e-05, + "loss": 0.9376, + "step": 35415 + }, + { + "epoch": 2.28, + "learning_rate": 1.2421692178878894e-05, + "loss": 0.8859, + "step": 35416 + }, + { + "epoch": 2.28, + "learning_rate": 1.2421477489855945e-05, + "loss": 0.8863, + "step": 35417 + }, + { + "epoch": 2.28, + "learning_rate": 1.2421262800832994e-05, + "loss": 0.8402, + "step": 35418 + }, + { + "epoch": 2.28, + "learning_rate": 1.2421048111810043e-05, + "loss": 0.8634, + "step": 35419 + }, + { + "epoch": 2.28, + "learning_rate": 1.2420833422787094e-05, + "loss": 0.832, + "step": 35420 + }, + { + "epoch": 2.28, + "learning_rate": 1.2420618733764143e-05, + "loss": 0.8849, + "step": 35421 + }, + { + "epoch": 2.28, + "learning_rate": 1.2420404044741195e-05, + "loss": 0.8738, + "step": 35422 + }, + { + "epoch": 2.28, + "learning_rate": 1.2420189355718244e-05, + "loss": 0.901, + "step": 35423 + }, + { + "epoch": 2.28, + "learning_rate": 1.2419974666695293e-05, + "loss": 0.8653, + "step": 35424 + }, + { + "epoch": 2.28, + "learning_rate": 1.2419759977672343e-05, + "loss": 0.8816, + "step": 35425 + }, + { + "epoch": 2.28, + "learning_rate": 1.2419545288649392e-05, + "loss": 0.9455, + "step": 35426 + }, + { + "epoch": 2.28, + "learning_rate": 1.2419330599626443e-05, + "loss": 0.8597, + "step": 35427 + }, + { + "epoch": 2.28, + "learning_rate": 1.2419115910603492e-05, + "loss": 0.8849, + "step": 35428 + }, + { + "epoch": 2.28, + "learning_rate": 1.2418901221580541e-05, + "loss": 0.8631, + "step": 35429 + }, + { + "epoch": 2.28, + "learning_rate": 1.2418686532557592e-05, + "loss": 0.9426, + "step": 35430 + }, + { + "epoch": 2.28, + "learning_rate": 1.241847184353464e-05, + "loss": 0.8885, + "step": 35431 + }, + { + "epoch": 2.28, + "learning_rate": 1.241825715451169e-05, + "loss": 0.9437, + "step": 35432 + }, + { + "epoch": 2.28, + "learning_rate": 1.241804246548874e-05, + "loss": 0.8642, + "step": 35433 + }, + { + "epoch": 2.28, + "learning_rate": 1.241782777646579e-05, + "loss": 0.867, + "step": 35434 + }, + { + "epoch": 2.28, + "learning_rate": 1.241761308744284e-05, + "loss": 0.8522, + "step": 35435 + }, + { + "epoch": 2.28, + "learning_rate": 1.2417398398419889e-05, + "loss": 0.8622, + "step": 35436 + }, + { + "epoch": 2.28, + "learning_rate": 1.2417183709396941e-05, + "loss": 0.8614, + "step": 35437 + }, + { + "epoch": 2.28, + "learning_rate": 1.241696902037399e-05, + "loss": 0.8444, + "step": 35438 + }, + { + "epoch": 2.28, + "learning_rate": 1.2416754331351039e-05, + "loss": 0.8942, + "step": 35439 + }, + { + "epoch": 2.28, + "learning_rate": 1.241653964232809e-05, + "loss": 0.8913, + "step": 35440 + }, + { + "epoch": 2.28, + "learning_rate": 1.2416324953305139e-05, + "loss": 0.8609, + "step": 35441 + }, + { + "epoch": 2.28, + "learning_rate": 1.2416110264282188e-05, + "loss": 0.9007, + "step": 35442 + }, + { + "epoch": 2.28, + "learning_rate": 1.2415895575259238e-05, + "loss": 0.834, + "step": 35443 + }, + { + "epoch": 2.28, + "learning_rate": 1.2415680886236287e-05, + "loss": 0.8673, + "step": 35444 + }, + { + "epoch": 2.28, + "learning_rate": 1.2415466197213338e-05, + "loss": 0.8977, + "step": 35445 + }, + { + "epoch": 2.28, + "learning_rate": 1.2415251508190387e-05, + "loss": 0.9587, + "step": 35446 + }, + { + "epoch": 2.28, + "learning_rate": 1.2415036819167436e-05, + "loss": 0.8412, + "step": 35447 + }, + { + "epoch": 2.28, + "learning_rate": 1.2414822130144487e-05, + "loss": 0.887, + "step": 35448 + }, + { + "epoch": 2.28, + "learning_rate": 1.2414607441121536e-05, + "loss": 0.9322, + "step": 35449 + }, + { + "epoch": 2.28, + "learning_rate": 1.2414392752098585e-05, + "loss": 0.9223, + "step": 35450 + }, + { + "epoch": 2.28, + "learning_rate": 1.2414178063075637e-05, + "loss": 0.942, + "step": 35451 + }, + { + "epoch": 2.28, + "learning_rate": 1.2413963374052686e-05, + "loss": 0.9472, + "step": 35452 + }, + { + "epoch": 2.28, + "learning_rate": 1.2413748685029737e-05, + "loss": 0.885, + "step": 35453 + }, + { + "epoch": 2.28, + "learning_rate": 1.2413533996006785e-05, + "loss": 0.8879, + "step": 35454 + }, + { + "epoch": 2.28, + "learning_rate": 1.2413319306983836e-05, + "loss": 0.8942, + "step": 35455 + }, + { + "epoch": 2.28, + "learning_rate": 1.2413104617960885e-05, + "loss": 0.9506, + "step": 35456 + }, + { + "epoch": 2.28, + "learning_rate": 1.2412889928937934e-05, + "loss": 0.8853, + "step": 35457 + }, + { + "epoch": 2.28, + "learning_rate": 1.2412675239914985e-05, + "loss": 0.9398, + "step": 35458 + }, + { + "epoch": 2.28, + "learning_rate": 1.2412460550892034e-05, + "loss": 0.9198, + "step": 35459 + }, + { + "epoch": 2.28, + "learning_rate": 1.2412245861869083e-05, + "loss": 0.944, + "step": 35460 + }, + { + "epoch": 2.28, + "learning_rate": 1.2412031172846133e-05, + "loss": 0.894, + "step": 35461 + }, + { + "epoch": 2.28, + "learning_rate": 1.2411816483823182e-05, + "loss": 0.8397, + "step": 35462 + }, + { + "epoch": 2.28, + "learning_rate": 1.2411601794800231e-05, + "loss": 0.9073, + "step": 35463 + }, + { + "epoch": 2.28, + "learning_rate": 1.2411387105777282e-05, + "loss": 0.8759, + "step": 35464 + }, + { + "epoch": 2.28, + "learning_rate": 1.2411172416754331e-05, + "loss": 0.8718, + "step": 35465 + }, + { + "epoch": 2.28, + "learning_rate": 1.2410957727731383e-05, + "loss": 0.8264, + "step": 35466 + }, + { + "epoch": 2.28, + "learning_rate": 1.2410743038708432e-05, + "loss": 0.9193, + "step": 35467 + }, + { + "epoch": 2.28, + "learning_rate": 1.2410528349685483e-05, + "loss": 0.8292, + "step": 35468 + }, + { + "epoch": 2.28, + "learning_rate": 1.2410313660662532e-05, + "loss": 0.9291, + "step": 35469 + }, + { + "epoch": 2.28, + "learning_rate": 1.241009897163958e-05, + "loss": 0.9095, + "step": 35470 + }, + { + "epoch": 2.28, + "learning_rate": 1.2409884282616631e-05, + "loss": 0.9509, + "step": 35471 + }, + { + "epoch": 2.28, + "learning_rate": 1.240966959359368e-05, + "loss": 0.8944, + "step": 35472 + }, + { + "epoch": 2.28, + "learning_rate": 1.2409454904570731e-05, + "loss": 0.9304, + "step": 35473 + }, + { + "epoch": 2.28, + "learning_rate": 1.240924021554778e-05, + "loss": 0.866, + "step": 35474 + }, + { + "epoch": 2.28, + "learning_rate": 1.2409025526524829e-05, + "loss": 0.8626, + "step": 35475 + }, + { + "epoch": 2.28, + "learning_rate": 1.240881083750188e-05, + "loss": 0.8586, + "step": 35476 + }, + { + "epoch": 2.28, + "learning_rate": 1.2408596148478929e-05, + "loss": 0.8944, + "step": 35477 + }, + { + "epoch": 2.28, + "learning_rate": 1.2408381459455978e-05, + "loss": 0.9038, + "step": 35478 + }, + { + "epoch": 2.28, + "learning_rate": 1.2408166770433028e-05, + "loss": 0.9102, + "step": 35479 + }, + { + "epoch": 2.28, + "learning_rate": 1.2407952081410079e-05, + "loss": 0.8958, + "step": 35480 + }, + { + "epoch": 2.28, + "learning_rate": 1.240773739238713e-05, + "loss": 0.8535, + "step": 35481 + }, + { + "epoch": 2.28, + "learning_rate": 1.2407522703364179e-05, + "loss": 0.9479, + "step": 35482 + }, + { + "epoch": 2.28, + "learning_rate": 1.240730801434123e-05, + "loss": 0.9126, + "step": 35483 + }, + { + "epoch": 2.28, + "learning_rate": 1.2407093325318278e-05, + "loss": 0.8603, + "step": 35484 + }, + { + "epoch": 2.28, + "learning_rate": 1.2406878636295327e-05, + "loss": 0.9059, + "step": 35485 + }, + { + "epoch": 2.28, + "learning_rate": 1.2406663947272378e-05, + "loss": 0.8644, + "step": 35486 + }, + { + "epoch": 2.28, + "learning_rate": 1.2406449258249427e-05, + "loss": 0.9011, + "step": 35487 + }, + { + "epoch": 2.28, + "learning_rate": 1.2406234569226476e-05, + "loss": 0.9157, + "step": 35488 + }, + { + "epoch": 2.28, + "learning_rate": 1.2406019880203526e-05, + "loss": 0.9125, + "step": 35489 + }, + { + "epoch": 2.28, + "learning_rate": 1.2405805191180575e-05, + "loss": 0.8611, + "step": 35490 + }, + { + "epoch": 2.28, + "learning_rate": 1.2405590502157624e-05, + "loss": 0.8621, + "step": 35491 + }, + { + "epoch": 2.28, + "learning_rate": 1.2405375813134675e-05, + "loss": 0.9436, + "step": 35492 + }, + { + "epoch": 2.28, + "learning_rate": 1.2405161124111724e-05, + "loss": 0.9323, + "step": 35493 + }, + { + "epoch": 2.28, + "learning_rate": 1.2404946435088775e-05, + "loss": 0.892, + "step": 35494 + }, + { + "epoch": 2.28, + "learning_rate": 1.2404731746065825e-05, + "loss": 0.9027, + "step": 35495 + }, + { + "epoch": 2.28, + "learning_rate": 1.2404517057042876e-05, + "loss": 0.8951, + "step": 35496 + }, + { + "epoch": 2.28, + "learning_rate": 1.2404302368019925e-05, + "loss": 0.9409, + "step": 35497 + }, + { + "epoch": 2.28, + "learning_rate": 1.2404087678996974e-05, + "loss": 0.8913, + "step": 35498 + }, + { + "epoch": 2.28, + "learning_rate": 1.2403872989974024e-05, + "loss": 0.9398, + "step": 35499 + }, + { + "epoch": 2.28, + "learning_rate": 1.2403658300951073e-05, + "loss": 0.9814, + "step": 35500 + }, + { + "epoch": 2.28, + "learning_rate": 1.2403443611928122e-05, + "loss": 0.951, + "step": 35501 + }, + { + "epoch": 2.28, + "learning_rate": 1.2403228922905173e-05, + "loss": 0.8744, + "step": 35502 + }, + { + "epoch": 2.28, + "learning_rate": 1.2403014233882222e-05, + "loss": 0.9469, + "step": 35503 + }, + { + "epoch": 2.28, + "learning_rate": 1.2402799544859273e-05, + "loss": 0.8524, + "step": 35504 + }, + { + "epoch": 2.28, + "learning_rate": 1.2402584855836322e-05, + "loss": 0.9257, + "step": 35505 + }, + { + "epoch": 2.28, + "learning_rate": 1.240237016681337e-05, + "loss": 0.879, + "step": 35506 + }, + { + "epoch": 2.28, + "learning_rate": 1.2402155477790421e-05, + "loss": 0.8167, + "step": 35507 + }, + { + "epoch": 2.28, + "learning_rate": 1.240194078876747e-05, + "loss": 0.9224, + "step": 35508 + }, + { + "epoch": 2.28, + "learning_rate": 1.2401726099744523e-05, + "loss": 0.9147, + "step": 35509 + }, + { + "epoch": 2.28, + "learning_rate": 1.2401511410721572e-05, + "loss": 0.9461, + "step": 35510 + }, + { + "epoch": 2.28, + "learning_rate": 1.240129672169862e-05, + "loss": 0.9176, + "step": 35511 + }, + { + "epoch": 2.28, + "learning_rate": 1.2401082032675671e-05, + "loss": 0.9468, + "step": 35512 + }, + { + "epoch": 2.28, + "learning_rate": 1.240086734365272e-05, + "loss": 0.8883, + "step": 35513 + }, + { + "epoch": 2.28, + "learning_rate": 1.240065265462977e-05, + "loss": 0.8335, + "step": 35514 + }, + { + "epoch": 2.28, + "learning_rate": 1.240043796560682e-05, + "loss": 0.8508, + "step": 35515 + }, + { + "epoch": 2.29, + "learning_rate": 1.2400223276583869e-05, + "loss": 0.9148, + "step": 35516 + }, + { + "epoch": 2.29, + "learning_rate": 1.240000858756092e-05, + "loss": 0.8565, + "step": 35517 + }, + { + "epoch": 2.29, + "learning_rate": 1.2399793898537968e-05, + "loss": 0.8812, + "step": 35518 + }, + { + "epoch": 2.29, + "learning_rate": 1.2399579209515017e-05, + "loss": 0.8878, + "step": 35519 + }, + { + "epoch": 2.29, + "learning_rate": 1.2399364520492068e-05, + "loss": 0.9304, + "step": 35520 + }, + { + "epoch": 2.29, + "learning_rate": 1.2399149831469117e-05, + "loss": 0.8427, + "step": 35521 + }, + { + "epoch": 2.29, + "learning_rate": 1.2398935142446168e-05, + "loss": 0.8803, + "step": 35522 + }, + { + "epoch": 2.29, + "learning_rate": 1.2398720453423217e-05, + "loss": 0.7989, + "step": 35523 + }, + { + "epoch": 2.29, + "learning_rate": 1.2398505764400269e-05, + "loss": 0.9414, + "step": 35524 + }, + { + "epoch": 2.29, + "learning_rate": 1.2398291075377318e-05, + "loss": 0.8619, + "step": 35525 + }, + { + "epoch": 2.29, + "learning_rate": 1.2398076386354367e-05, + "loss": 0.8816, + "step": 35526 + }, + { + "epoch": 2.29, + "learning_rate": 1.2397861697331418e-05, + "loss": 0.8869, + "step": 35527 + }, + { + "epoch": 2.29, + "learning_rate": 1.2397647008308466e-05, + "loss": 0.9097, + "step": 35528 + }, + { + "epoch": 2.29, + "learning_rate": 1.2397432319285515e-05, + "loss": 0.8376, + "step": 35529 + }, + { + "epoch": 2.29, + "learning_rate": 1.2397217630262566e-05, + "loss": 0.84, + "step": 35530 + }, + { + "epoch": 2.29, + "learning_rate": 1.2397002941239615e-05, + "loss": 0.8831, + "step": 35531 + }, + { + "epoch": 2.29, + "learning_rate": 1.2396788252216666e-05, + "loss": 0.8081, + "step": 35532 + }, + { + "epoch": 2.29, + "learning_rate": 1.2396573563193715e-05, + "loss": 0.9467, + "step": 35533 + }, + { + "epoch": 2.29, + "learning_rate": 1.2396358874170764e-05, + "loss": 0.8969, + "step": 35534 + }, + { + "epoch": 2.29, + "learning_rate": 1.2396144185147814e-05, + "loss": 0.948, + "step": 35535 + }, + { + "epoch": 2.29, + "learning_rate": 1.2395929496124863e-05, + "loss": 0.91, + "step": 35536 + }, + { + "epoch": 2.29, + "learning_rate": 1.2395714807101912e-05, + "loss": 0.8766, + "step": 35537 + }, + { + "epoch": 2.29, + "learning_rate": 1.2395500118078965e-05, + "loss": 0.9009, + "step": 35538 + }, + { + "epoch": 2.29, + "learning_rate": 1.2395285429056014e-05, + "loss": 0.9383, + "step": 35539 + }, + { + "epoch": 2.29, + "learning_rate": 1.2395070740033064e-05, + "loss": 0.9191, + "step": 35540 + }, + { + "epoch": 2.29, + "learning_rate": 1.2394856051010113e-05, + "loss": 0.9013, + "step": 35541 + }, + { + "epoch": 2.29, + "learning_rate": 1.2394641361987164e-05, + "loss": 0.956, + "step": 35542 + }, + { + "epoch": 2.29, + "learning_rate": 1.2394426672964213e-05, + "loss": 0.9246, + "step": 35543 + }, + { + "epoch": 2.29, + "learning_rate": 1.2394211983941262e-05, + "loss": 0.8902, + "step": 35544 + }, + { + "epoch": 2.29, + "learning_rate": 1.2393997294918312e-05, + "loss": 0.7727, + "step": 35545 + }, + { + "epoch": 2.29, + "learning_rate": 1.2393782605895361e-05, + "loss": 0.8916, + "step": 35546 + }, + { + "epoch": 2.29, + "learning_rate": 1.239356791687241e-05, + "loss": 0.8531, + "step": 35547 + }, + { + "epoch": 2.29, + "learning_rate": 1.2393353227849461e-05, + "loss": 0.8465, + "step": 35548 + }, + { + "epoch": 2.29, + "learning_rate": 1.239313853882651e-05, + "loss": 0.914, + "step": 35549 + }, + { + "epoch": 2.29, + "learning_rate": 1.2392923849803559e-05, + "loss": 0.9121, + "step": 35550 + }, + { + "epoch": 2.29, + "learning_rate": 1.239270916078061e-05, + "loss": 0.8306, + "step": 35551 + }, + { + "epoch": 2.29, + "learning_rate": 1.2392494471757659e-05, + "loss": 0.9532, + "step": 35552 + }, + { + "epoch": 2.29, + "learning_rate": 1.2392279782734711e-05, + "loss": 0.8276, + "step": 35553 + }, + { + "epoch": 2.29, + "learning_rate": 1.239206509371176e-05, + "loss": 0.9083, + "step": 35554 + }, + { + "epoch": 2.29, + "learning_rate": 1.239185040468881e-05, + "loss": 0.9064, + "step": 35555 + }, + { + "epoch": 2.29, + "learning_rate": 1.239163571566586e-05, + "loss": 0.9152, + "step": 35556 + }, + { + "epoch": 2.29, + "learning_rate": 1.2391421026642908e-05, + "loss": 0.9688, + "step": 35557 + }, + { + "epoch": 2.29, + "learning_rate": 1.2391206337619959e-05, + "loss": 0.9184, + "step": 35558 + }, + { + "epoch": 2.29, + "learning_rate": 1.2390991648597008e-05, + "loss": 0.8654, + "step": 35559 + }, + { + "epoch": 2.29, + "learning_rate": 1.2390776959574057e-05, + "loss": 0.8556, + "step": 35560 + }, + { + "epoch": 2.29, + "learning_rate": 1.2390562270551108e-05, + "loss": 0.9132, + "step": 35561 + }, + { + "epoch": 2.29, + "learning_rate": 1.2390347581528157e-05, + "loss": 0.8936, + "step": 35562 + }, + { + "epoch": 2.29, + "learning_rate": 1.2390132892505207e-05, + "loss": 0.8607, + "step": 35563 + }, + { + "epoch": 2.29, + "learning_rate": 1.2389918203482256e-05, + "loss": 0.9016, + "step": 35564 + }, + { + "epoch": 2.29, + "learning_rate": 1.2389703514459305e-05, + "loss": 0.876, + "step": 35565 + }, + { + "epoch": 2.29, + "learning_rate": 1.2389488825436356e-05, + "loss": 0.9377, + "step": 35566 + }, + { + "epoch": 2.29, + "learning_rate": 1.2389274136413407e-05, + "loss": 0.9457, + "step": 35567 + }, + { + "epoch": 2.29, + "learning_rate": 1.2389059447390457e-05, + "loss": 0.9281, + "step": 35568 + }, + { + "epoch": 2.29, + "learning_rate": 1.2388844758367506e-05, + "loss": 0.8359, + "step": 35569 + }, + { + "epoch": 2.29, + "learning_rate": 1.2388630069344555e-05, + "loss": 0.9366, + "step": 35570 + }, + { + "epoch": 2.29, + "learning_rate": 1.2388415380321606e-05, + "loss": 0.8884, + "step": 35571 + }, + { + "epoch": 2.29, + "learning_rate": 1.2388200691298655e-05, + "loss": 0.9133, + "step": 35572 + }, + { + "epoch": 2.29, + "learning_rate": 1.2387986002275705e-05, + "loss": 0.865, + "step": 35573 + }, + { + "epoch": 2.29, + "learning_rate": 1.2387771313252754e-05, + "loss": 0.9212, + "step": 35574 + }, + { + "epoch": 2.29, + "learning_rate": 1.2387556624229803e-05, + "loss": 0.9024, + "step": 35575 + }, + { + "epoch": 2.29, + "learning_rate": 1.2387341935206854e-05, + "loss": 0.886, + "step": 35576 + }, + { + "epoch": 2.29, + "learning_rate": 1.2387127246183903e-05, + "loss": 0.9682, + "step": 35577 + }, + { + "epoch": 2.29, + "learning_rate": 1.2386912557160952e-05, + "loss": 0.9461, + "step": 35578 + }, + { + "epoch": 2.29, + "learning_rate": 1.2386697868138003e-05, + "loss": 0.8952, + "step": 35579 + }, + { + "epoch": 2.29, + "learning_rate": 1.2386483179115052e-05, + "loss": 0.8666, + "step": 35580 + }, + { + "epoch": 2.29, + "learning_rate": 1.2386268490092102e-05, + "loss": 0.9501, + "step": 35581 + }, + { + "epoch": 2.29, + "learning_rate": 1.2386053801069153e-05, + "loss": 0.9598, + "step": 35582 + }, + { + "epoch": 2.29, + "learning_rate": 1.2385839112046204e-05, + "loss": 0.9449, + "step": 35583 + }, + { + "epoch": 2.29, + "learning_rate": 1.2385624423023253e-05, + "loss": 0.8933, + "step": 35584 + }, + { + "epoch": 2.29, + "learning_rate": 1.2385409734000302e-05, + "loss": 0.8725, + "step": 35585 + }, + { + "epoch": 2.29, + "learning_rate": 1.2385195044977352e-05, + "loss": 0.9315, + "step": 35586 + }, + { + "epoch": 2.29, + "learning_rate": 1.2384980355954401e-05, + "loss": 0.9216, + "step": 35587 + }, + { + "epoch": 2.29, + "learning_rate": 1.238476566693145e-05, + "loss": 0.9026, + "step": 35588 + }, + { + "epoch": 2.29, + "learning_rate": 1.23845509779085e-05, + "loss": 0.9039, + "step": 35589 + }, + { + "epoch": 2.29, + "learning_rate": 1.238433628888555e-05, + "loss": 0.8575, + "step": 35590 + }, + { + "epoch": 2.29, + "learning_rate": 1.23841215998626e-05, + "loss": 0.9224, + "step": 35591 + }, + { + "epoch": 2.29, + "learning_rate": 1.238390691083965e-05, + "loss": 0.937, + "step": 35592 + }, + { + "epoch": 2.29, + "learning_rate": 1.2383692221816698e-05, + "loss": 0.9217, + "step": 35593 + }, + { + "epoch": 2.29, + "learning_rate": 1.2383477532793749e-05, + "loss": 0.9011, + "step": 35594 + }, + { + "epoch": 2.29, + "learning_rate": 1.2383262843770798e-05, + "loss": 0.8158, + "step": 35595 + }, + { + "epoch": 2.29, + "learning_rate": 1.238304815474785e-05, + "loss": 0.8739, + "step": 35596 + }, + { + "epoch": 2.29, + "learning_rate": 1.23828334657249e-05, + "loss": 0.901, + "step": 35597 + }, + { + "epoch": 2.29, + "learning_rate": 1.2382618776701948e-05, + "loss": 0.8761, + "step": 35598 + }, + { + "epoch": 2.29, + "learning_rate": 1.2382404087678999e-05, + "loss": 0.9217, + "step": 35599 + }, + { + "epoch": 2.29, + "learning_rate": 1.2382189398656048e-05, + "loss": 0.9062, + "step": 35600 + }, + { + "epoch": 2.29, + "learning_rate": 1.2381974709633099e-05, + "loss": 0.9246, + "step": 35601 + }, + { + "epoch": 2.29, + "learning_rate": 1.2381760020610147e-05, + "loss": 0.9048, + "step": 35602 + }, + { + "epoch": 2.29, + "learning_rate": 1.2381545331587196e-05, + "loss": 0.8471, + "step": 35603 + }, + { + "epoch": 2.29, + "learning_rate": 1.2381330642564247e-05, + "loss": 0.8587, + "step": 35604 + }, + { + "epoch": 2.29, + "learning_rate": 1.2381115953541296e-05, + "loss": 0.9685, + "step": 35605 + }, + { + "epoch": 2.29, + "learning_rate": 1.2380901264518345e-05, + "loss": 0.8767, + "step": 35606 + }, + { + "epoch": 2.29, + "learning_rate": 1.2380686575495396e-05, + "loss": 0.9688, + "step": 35607 + }, + { + "epoch": 2.29, + "learning_rate": 1.2380471886472445e-05, + "loss": 0.9701, + "step": 35608 + }, + { + "epoch": 2.29, + "learning_rate": 1.2380257197449495e-05, + "loss": 0.9239, + "step": 35609 + }, + { + "epoch": 2.29, + "learning_rate": 1.2380042508426544e-05, + "loss": 0.9016, + "step": 35610 + }, + { + "epoch": 2.29, + "learning_rate": 1.2379827819403597e-05, + "loss": 0.9113, + "step": 35611 + }, + { + "epoch": 2.29, + "learning_rate": 1.2379613130380646e-05, + "loss": 0.9386, + "step": 35612 + }, + { + "epoch": 2.29, + "learning_rate": 1.2379398441357695e-05, + "loss": 0.8922, + "step": 35613 + }, + { + "epoch": 2.29, + "learning_rate": 1.2379183752334745e-05, + "loss": 0.88, + "step": 35614 + }, + { + "epoch": 2.29, + "learning_rate": 1.2378969063311794e-05, + "loss": 0.8894, + "step": 35615 + }, + { + "epoch": 2.29, + "learning_rate": 1.2378754374288843e-05, + "loss": 0.9349, + "step": 35616 + }, + { + "epoch": 2.29, + "learning_rate": 1.2378539685265894e-05, + "loss": 0.8762, + "step": 35617 + }, + { + "epoch": 2.29, + "learning_rate": 1.2378324996242943e-05, + "loss": 0.9935, + "step": 35618 + }, + { + "epoch": 2.29, + "learning_rate": 1.2378110307219993e-05, + "loss": 0.9358, + "step": 35619 + }, + { + "epoch": 2.29, + "learning_rate": 1.2377895618197042e-05, + "loss": 0.9255, + "step": 35620 + }, + { + "epoch": 2.29, + "learning_rate": 1.2377680929174091e-05, + "loss": 0.9386, + "step": 35621 + }, + { + "epoch": 2.29, + "learning_rate": 1.2377466240151142e-05, + "loss": 0.9183, + "step": 35622 + }, + { + "epoch": 2.29, + "learning_rate": 1.2377251551128191e-05, + "loss": 0.8927, + "step": 35623 + }, + { + "epoch": 2.29, + "learning_rate": 1.237703686210524e-05, + "loss": 0.9393, + "step": 35624 + }, + { + "epoch": 2.29, + "learning_rate": 1.2376822173082292e-05, + "loss": 0.9356, + "step": 35625 + }, + { + "epoch": 2.29, + "learning_rate": 1.2376607484059341e-05, + "loss": 0.9041, + "step": 35626 + }, + { + "epoch": 2.29, + "learning_rate": 1.2376392795036392e-05, + "loss": 0.9448, + "step": 35627 + }, + { + "epoch": 2.29, + "learning_rate": 1.2376178106013441e-05, + "loss": 0.8883, + "step": 35628 + }, + { + "epoch": 2.29, + "learning_rate": 1.2375963416990492e-05, + "loss": 0.8588, + "step": 35629 + }, + { + "epoch": 2.29, + "learning_rate": 1.237574872796754e-05, + "loss": 0.9576, + "step": 35630 + }, + { + "epoch": 2.29, + "learning_rate": 1.237553403894459e-05, + "loss": 0.9068, + "step": 35631 + }, + { + "epoch": 2.29, + "learning_rate": 1.237531934992164e-05, + "loss": 0.8894, + "step": 35632 + }, + { + "epoch": 2.29, + "learning_rate": 1.2375104660898689e-05, + "loss": 0.9147, + "step": 35633 + }, + { + "epoch": 2.29, + "learning_rate": 1.2374889971875738e-05, + "loss": 0.9275, + "step": 35634 + }, + { + "epoch": 2.29, + "learning_rate": 1.2374675282852789e-05, + "loss": 0.815, + "step": 35635 + }, + { + "epoch": 2.29, + "learning_rate": 1.2374460593829838e-05, + "loss": 0.8847, + "step": 35636 + }, + { + "epoch": 2.29, + "learning_rate": 1.2374245904806887e-05, + "loss": 0.9195, + "step": 35637 + }, + { + "epoch": 2.29, + "learning_rate": 1.2374031215783937e-05, + "loss": 0.9717, + "step": 35638 + }, + { + "epoch": 2.29, + "learning_rate": 1.2373816526760986e-05, + "loss": 0.8936, + "step": 35639 + }, + { + "epoch": 2.29, + "learning_rate": 1.2373601837738039e-05, + "loss": 0.9492, + "step": 35640 + }, + { + "epoch": 2.29, + "learning_rate": 1.2373387148715088e-05, + "loss": 0.8718, + "step": 35641 + }, + { + "epoch": 2.29, + "learning_rate": 1.2373172459692138e-05, + "loss": 0.8571, + "step": 35642 + }, + { + "epoch": 2.29, + "learning_rate": 1.2372957770669187e-05, + "loss": 0.836, + "step": 35643 + }, + { + "epoch": 2.29, + "learning_rate": 1.2372743081646236e-05, + "loss": 0.8845, + "step": 35644 + }, + { + "epoch": 2.29, + "learning_rate": 1.2372528392623287e-05, + "loss": 0.9571, + "step": 35645 + }, + { + "epoch": 2.29, + "learning_rate": 1.2372313703600336e-05, + "loss": 0.8836, + "step": 35646 + }, + { + "epoch": 2.29, + "learning_rate": 1.2372099014577385e-05, + "loss": 0.9427, + "step": 35647 + }, + { + "epoch": 2.29, + "learning_rate": 1.2371884325554435e-05, + "loss": 0.9066, + "step": 35648 + }, + { + "epoch": 2.29, + "learning_rate": 1.2371669636531484e-05, + "loss": 0.9092, + "step": 35649 + }, + { + "epoch": 2.29, + "learning_rate": 1.2371454947508535e-05, + "loss": 0.9105, + "step": 35650 + }, + { + "epoch": 2.29, + "learning_rate": 1.2371240258485584e-05, + "loss": 0.8362, + "step": 35651 + }, + { + "epoch": 2.29, + "learning_rate": 1.2371025569462633e-05, + "loss": 0.8947, + "step": 35652 + }, + { + "epoch": 2.29, + "learning_rate": 1.2370810880439684e-05, + "loss": 0.9009, + "step": 35653 + }, + { + "epoch": 2.29, + "learning_rate": 1.2370596191416734e-05, + "loss": 0.8745, + "step": 35654 + }, + { + "epoch": 2.29, + "learning_rate": 1.2370381502393785e-05, + "loss": 0.8556, + "step": 35655 + }, + { + "epoch": 2.29, + "learning_rate": 1.2370166813370834e-05, + "loss": 0.9495, + "step": 35656 + }, + { + "epoch": 2.29, + "learning_rate": 1.2369952124347883e-05, + "loss": 0.8679, + "step": 35657 + }, + { + "epoch": 2.29, + "learning_rate": 1.2369737435324934e-05, + "loss": 0.8809, + "step": 35658 + }, + { + "epoch": 2.29, + "learning_rate": 1.2369522746301983e-05, + "loss": 0.8076, + "step": 35659 + }, + { + "epoch": 2.29, + "learning_rate": 1.2369308057279033e-05, + "loss": 0.9177, + "step": 35660 + }, + { + "epoch": 2.29, + "learning_rate": 1.2369093368256082e-05, + "loss": 0.9184, + "step": 35661 + }, + { + "epoch": 2.29, + "learning_rate": 1.2368878679233131e-05, + "loss": 0.8807, + "step": 35662 + }, + { + "epoch": 2.29, + "learning_rate": 1.2368663990210182e-05, + "loss": 0.9645, + "step": 35663 + }, + { + "epoch": 2.29, + "learning_rate": 1.236844930118723e-05, + "loss": 0.8826, + "step": 35664 + }, + { + "epoch": 2.29, + "learning_rate": 1.236823461216428e-05, + "loss": 0.8659, + "step": 35665 + }, + { + "epoch": 2.29, + "learning_rate": 1.236801992314133e-05, + "loss": 0.8654, + "step": 35666 + }, + { + "epoch": 2.29, + "learning_rate": 1.236780523411838e-05, + "loss": 0.8751, + "step": 35667 + }, + { + "epoch": 2.29, + "learning_rate": 1.236759054509543e-05, + "loss": 0.957, + "step": 35668 + }, + { + "epoch": 2.29, + "learning_rate": 1.236737585607248e-05, + "loss": 0.9125, + "step": 35669 + }, + { + "epoch": 2.29, + "learning_rate": 1.2367161167049531e-05, + "loss": 0.8967, + "step": 35670 + }, + { + "epoch": 2.29, + "learning_rate": 1.236694647802658e-05, + "loss": 0.8515, + "step": 35671 + }, + { + "epoch": 2.3, + "learning_rate": 1.236673178900363e-05, + "loss": 0.8602, + "step": 35672 + }, + { + "epoch": 2.3, + "learning_rate": 1.236651709998068e-05, + "loss": 0.8943, + "step": 35673 + }, + { + "epoch": 2.3, + "learning_rate": 1.2366302410957729e-05, + "loss": 0.8978, + "step": 35674 + }, + { + "epoch": 2.3, + "learning_rate": 1.2366087721934778e-05, + "loss": 0.92, + "step": 35675 + }, + { + "epoch": 2.3, + "learning_rate": 1.2365873032911828e-05, + "loss": 0.9615, + "step": 35676 + }, + { + "epoch": 2.3, + "learning_rate": 1.2365658343888877e-05, + "loss": 0.8905, + "step": 35677 + }, + { + "epoch": 2.3, + "learning_rate": 1.2365443654865928e-05, + "loss": 0.923, + "step": 35678 + }, + { + "epoch": 2.3, + "learning_rate": 1.2365228965842977e-05, + "loss": 0.8487, + "step": 35679 + }, + { + "epoch": 2.3, + "learning_rate": 1.2365014276820026e-05, + "loss": 0.9423, + "step": 35680 + }, + { + "epoch": 2.3, + "learning_rate": 1.2364799587797077e-05, + "loss": 0.9005, + "step": 35681 + }, + { + "epoch": 2.3, + "learning_rate": 1.2364584898774126e-05, + "loss": 0.8787, + "step": 35682 + }, + { + "epoch": 2.3, + "learning_rate": 1.2364370209751178e-05, + "loss": 0.8446, + "step": 35683 + }, + { + "epoch": 2.3, + "learning_rate": 1.2364155520728227e-05, + "loss": 0.8979, + "step": 35684 + }, + { + "epoch": 2.3, + "learning_rate": 1.2363940831705276e-05, + "loss": 0.872, + "step": 35685 + }, + { + "epoch": 2.3, + "learning_rate": 1.2363726142682327e-05, + "loss": 0.8583, + "step": 35686 + }, + { + "epoch": 2.3, + "learning_rate": 1.2363511453659376e-05, + "loss": 0.8735, + "step": 35687 + }, + { + "epoch": 2.3, + "learning_rate": 1.2363296764636426e-05, + "loss": 0.908, + "step": 35688 + }, + { + "epoch": 2.3, + "learning_rate": 1.2363082075613475e-05, + "loss": 0.8452, + "step": 35689 + }, + { + "epoch": 2.3, + "learning_rate": 1.2362867386590524e-05, + "loss": 0.8318, + "step": 35690 + }, + { + "epoch": 2.3, + "learning_rate": 1.2362652697567575e-05, + "loss": 0.8724, + "step": 35691 + }, + { + "epoch": 2.3, + "learning_rate": 1.2362438008544624e-05, + "loss": 0.9672, + "step": 35692 + }, + { + "epoch": 2.3, + "learning_rate": 1.2362223319521673e-05, + "loss": 0.8714, + "step": 35693 + }, + { + "epoch": 2.3, + "learning_rate": 1.2362008630498723e-05, + "loss": 0.9314, + "step": 35694 + }, + { + "epoch": 2.3, + "learning_rate": 1.2361793941475772e-05, + "loss": 0.9102, + "step": 35695 + }, + { + "epoch": 2.3, + "learning_rate": 1.2361579252452823e-05, + "loss": 0.84, + "step": 35696 + }, + { + "epoch": 2.3, + "learning_rate": 1.2361364563429872e-05, + "loss": 0.9363, + "step": 35697 + }, + { + "epoch": 2.3, + "learning_rate": 1.2361149874406924e-05, + "loss": 0.896, + "step": 35698 + }, + { + "epoch": 2.3, + "learning_rate": 1.2360935185383973e-05, + "loss": 0.8559, + "step": 35699 + }, + { + "epoch": 2.3, + "learning_rate": 1.2360720496361022e-05, + "loss": 0.9122, + "step": 35700 + }, + { + "epoch": 2.3, + "learning_rate": 1.2360505807338073e-05, + "loss": 0.8805, + "step": 35701 + }, + { + "epoch": 2.3, + "learning_rate": 1.2360291118315122e-05, + "loss": 0.9, + "step": 35702 + }, + { + "epoch": 2.3, + "learning_rate": 1.2360076429292171e-05, + "loss": 0.8977, + "step": 35703 + }, + { + "epoch": 2.3, + "learning_rate": 1.2359861740269222e-05, + "loss": 0.8817, + "step": 35704 + }, + { + "epoch": 2.3, + "learning_rate": 1.235964705124627e-05, + "loss": 0.9139, + "step": 35705 + }, + { + "epoch": 2.3, + "learning_rate": 1.2359432362223321e-05, + "loss": 0.9055, + "step": 35706 + }, + { + "epoch": 2.3, + "learning_rate": 1.235921767320037e-05, + "loss": 0.8412, + "step": 35707 + }, + { + "epoch": 2.3, + "learning_rate": 1.2359002984177419e-05, + "loss": 0.902, + "step": 35708 + }, + { + "epoch": 2.3, + "learning_rate": 1.235878829515447e-05, + "loss": 0.9319, + "step": 35709 + }, + { + "epoch": 2.3, + "learning_rate": 1.2358573606131519e-05, + "loss": 0.8716, + "step": 35710 + }, + { + "epoch": 2.3, + "learning_rate": 1.2358358917108568e-05, + "loss": 0.9112, + "step": 35711 + }, + { + "epoch": 2.3, + "learning_rate": 1.235814422808562e-05, + "loss": 0.8633, + "step": 35712 + }, + { + "epoch": 2.3, + "learning_rate": 1.2357929539062669e-05, + "loss": 0.8686, + "step": 35713 + }, + { + "epoch": 2.3, + "learning_rate": 1.235771485003972e-05, + "loss": 0.8944, + "step": 35714 + }, + { + "epoch": 2.3, + "learning_rate": 1.2357500161016769e-05, + "loss": 0.9888, + "step": 35715 + }, + { + "epoch": 2.3, + "learning_rate": 1.235728547199382e-05, + "loss": 0.9015, + "step": 35716 + }, + { + "epoch": 2.3, + "learning_rate": 1.2357070782970868e-05, + "loss": 0.8773, + "step": 35717 + }, + { + "epoch": 2.3, + "learning_rate": 1.2356856093947917e-05, + "loss": 0.9452, + "step": 35718 + }, + { + "epoch": 2.3, + "learning_rate": 1.2356641404924968e-05, + "loss": 0.8761, + "step": 35719 + }, + { + "epoch": 2.3, + "learning_rate": 1.2356426715902017e-05, + "loss": 0.9631, + "step": 35720 + }, + { + "epoch": 2.3, + "learning_rate": 1.2356212026879066e-05, + "loss": 0.8694, + "step": 35721 + }, + { + "epoch": 2.3, + "learning_rate": 1.2355997337856116e-05, + "loss": 0.982, + "step": 35722 + }, + { + "epoch": 2.3, + "learning_rate": 1.2355782648833165e-05, + "loss": 0.8536, + "step": 35723 + }, + { + "epoch": 2.3, + "learning_rate": 1.2355567959810214e-05, + "loss": 0.828, + "step": 35724 + }, + { + "epoch": 2.3, + "learning_rate": 1.2355353270787265e-05, + "loss": 0.8679, + "step": 35725 + }, + { + "epoch": 2.3, + "learning_rate": 1.2355138581764314e-05, + "loss": 0.9423, + "step": 35726 + }, + { + "epoch": 2.3, + "learning_rate": 1.2354923892741366e-05, + "loss": 0.8486, + "step": 35727 + }, + { + "epoch": 2.3, + "learning_rate": 1.2354709203718415e-05, + "loss": 0.9498, + "step": 35728 + }, + { + "epoch": 2.3, + "learning_rate": 1.2354494514695466e-05, + "loss": 0.9437, + "step": 35729 + }, + { + "epoch": 2.3, + "learning_rate": 1.2354279825672515e-05, + "loss": 0.8703, + "step": 35730 + }, + { + "epoch": 2.3, + "learning_rate": 1.2354065136649564e-05, + "loss": 0.9314, + "step": 35731 + }, + { + "epoch": 2.3, + "learning_rate": 1.2353850447626615e-05, + "loss": 0.9047, + "step": 35732 + }, + { + "epoch": 2.3, + "learning_rate": 1.2353635758603664e-05, + "loss": 0.8767, + "step": 35733 + }, + { + "epoch": 2.3, + "learning_rate": 1.2353421069580712e-05, + "loss": 0.9659, + "step": 35734 + }, + { + "epoch": 2.3, + "learning_rate": 1.2353206380557763e-05, + "loss": 0.8326, + "step": 35735 + }, + { + "epoch": 2.3, + "learning_rate": 1.2352991691534812e-05, + "loss": 0.9246, + "step": 35736 + }, + { + "epoch": 2.3, + "learning_rate": 1.2352777002511863e-05, + "loss": 0.9394, + "step": 35737 + }, + { + "epoch": 2.3, + "learning_rate": 1.2352562313488912e-05, + "loss": 0.9435, + "step": 35738 + }, + { + "epoch": 2.3, + "learning_rate": 1.235234762446596e-05, + "loss": 0.8942, + "step": 35739 + }, + { + "epoch": 2.3, + "learning_rate": 1.2352132935443011e-05, + "loss": 0.9027, + "step": 35740 + }, + { + "epoch": 2.3, + "learning_rate": 1.2351918246420062e-05, + "loss": 0.9111, + "step": 35741 + }, + { + "epoch": 2.3, + "learning_rate": 1.2351703557397113e-05, + "loss": 0.8783, + "step": 35742 + }, + { + "epoch": 2.3, + "learning_rate": 1.2351488868374162e-05, + "loss": 0.9432, + "step": 35743 + }, + { + "epoch": 2.3, + "learning_rate": 1.235127417935121e-05, + "loss": 0.9365, + "step": 35744 + }, + { + "epoch": 2.3, + "learning_rate": 1.2351059490328261e-05, + "loss": 0.8975, + "step": 35745 + }, + { + "epoch": 2.3, + "learning_rate": 1.235084480130531e-05, + "loss": 0.9252, + "step": 35746 + }, + { + "epoch": 2.3, + "learning_rate": 1.2350630112282361e-05, + "loss": 0.9065, + "step": 35747 + }, + { + "epoch": 2.3, + "learning_rate": 1.235041542325941e-05, + "loss": 0.8696, + "step": 35748 + }, + { + "epoch": 2.3, + "learning_rate": 1.2350200734236459e-05, + "loss": 0.9192, + "step": 35749 + }, + { + "epoch": 2.3, + "learning_rate": 1.234998604521351e-05, + "loss": 0.8794, + "step": 35750 + }, + { + "epoch": 2.3, + "learning_rate": 1.2349771356190558e-05, + "loss": 0.9032, + "step": 35751 + }, + { + "epoch": 2.3, + "learning_rate": 1.2349556667167607e-05, + "loss": 0.8542, + "step": 35752 + }, + { + "epoch": 2.3, + "learning_rate": 1.2349341978144658e-05, + "loss": 0.9293, + "step": 35753 + }, + { + "epoch": 2.3, + "learning_rate": 1.2349127289121707e-05, + "loss": 0.8818, + "step": 35754 + }, + { + "epoch": 2.3, + "learning_rate": 1.2348912600098758e-05, + "loss": 0.8848, + "step": 35755 + }, + { + "epoch": 2.3, + "learning_rate": 1.2348697911075808e-05, + "loss": 0.8585, + "step": 35756 + }, + { + "epoch": 2.3, + "learning_rate": 1.2348483222052859e-05, + "loss": 0.8667, + "step": 35757 + }, + { + "epoch": 2.3, + "learning_rate": 1.2348268533029908e-05, + "loss": 0.8921, + "step": 35758 + }, + { + "epoch": 2.3, + "learning_rate": 1.2348053844006957e-05, + "loss": 0.9285, + "step": 35759 + }, + { + "epoch": 2.3, + "learning_rate": 1.2347839154984008e-05, + "loss": 0.871, + "step": 35760 + }, + { + "epoch": 2.3, + "learning_rate": 1.2347624465961057e-05, + "loss": 0.9186, + "step": 35761 + }, + { + "epoch": 2.3, + "learning_rate": 1.2347409776938106e-05, + "loss": 0.8498, + "step": 35762 + }, + { + "epoch": 2.3, + "learning_rate": 1.2347195087915156e-05, + "loss": 0.8423, + "step": 35763 + }, + { + "epoch": 2.3, + "learning_rate": 1.2346980398892205e-05, + "loss": 0.9105, + "step": 35764 + }, + { + "epoch": 2.3, + "learning_rate": 1.2346765709869256e-05, + "loss": 0.8898, + "step": 35765 + }, + { + "epoch": 2.3, + "learning_rate": 1.2346551020846305e-05, + "loss": 0.8293, + "step": 35766 + }, + { + "epoch": 2.3, + "learning_rate": 1.2346336331823354e-05, + "loss": 0.9011, + "step": 35767 + }, + { + "epoch": 2.3, + "learning_rate": 1.2346121642800404e-05, + "loss": 0.8969, + "step": 35768 + }, + { + "epoch": 2.3, + "learning_rate": 1.2345906953777453e-05, + "loss": 0.9443, + "step": 35769 + }, + { + "epoch": 2.3, + "learning_rate": 1.2345692264754506e-05, + "loss": 0.8345, + "step": 35770 + }, + { + "epoch": 2.3, + "learning_rate": 1.2345477575731555e-05, + "loss": 0.8665, + "step": 35771 + }, + { + "epoch": 2.3, + "learning_rate": 1.2345262886708604e-05, + "loss": 0.8907, + "step": 35772 + }, + { + "epoch": 2.3, + "learning_rate": 1.2345048197685654e-05, + "loss": 0.8988, + "step": 35773 + }, + { + "epoch": 2.3, + "learning_rate": 1.2344833508662703e-05, + "loss": 0.9413, + "step": 35774 + }, + { + "epoch": 2.3, + "learning_rate": 1.2344618819639754e-05, + "loss": 0.9317, + "step": 35775 + }, + { + "epoch": 2.3, + "learning_rate": 1.2344404130616803e-05, + "loss": 0.8859, + "step": 35776 + }, + { + "epoch": 2.3, + "learning_rate": 1.2344189441593852e-05, + "loss": 0.981, + "step": 35777 + }, + { + "epoch": 2.3, + "learning_rate": 1.2343974752570903e-05, + "loss": 0.9432, + "step": 35778 + }, + { + "epoch": 2.3, + "learning_rate": 1.2343760063547951e-05, + "loss": 0.9086, + "step": 35779 + }, + { + "epoch": 2.3, + "learning_rate": 1.2343545374525e-05, + "loss": 0.8727, + "step": 35780 + }, + { + "epoch": 2.3, + "learning_rate": 1.2343330685502051e-05, + "loss": 0.8896, + "step": 35781 + }, + { + "epoch": 2.3, + "learning_rate": 1.23431159964791e-05, + "loss": 0.8734, + "step": 35782 + }, + { + "epoch": 2.3, + "learning_rate": 1.2342901307456149e-05, + "loss": 0.938, + "step": 35783 + }, + { + "epoch": 2.3, + "learning_rate": 1.23426866184332e-05, + "loss": 0.8847, + "step": 35784 + }, + { + "epoch": 2.3, + "learning_rate": 1.2342471929410252e-05, + "loss": 0.8583, + "step": 35785 + }, + { + "epoch": 2.3, + "learning_rate": 1.2342257240387301e-05, + "loss": 0.8944, + "step": 35786 + }, + { + "epoch": 2.3, + "learning_rate": 1.234204255136435e-05, + "loss": 0.8751, + "step": 35787 + }, + { + "epoch": 2.3, + "learning_rate": 1.23418278623414e-05, + "loss": 0.9112, + "step": 35788 + }, + { + "epoch": 2.3, + "learning_rate": 1.234161317331845e-05, + "loss": 0.8346, + "step": 35789 + }, + { + "epoch": 2.3, + "learning_rate": 1.2341398484295499e-05, + "loss": 0.8935, + "step": 35790 + }, + { + "epoch": 2.3, + "learning_rate": 1.234118379527255e-05, + "loss": 0.8802, + "step": 35791 + }, + { + "epoch": 2.3, + "learning_rate": 1.2340969106249598e-05, + "loss": 1.0349, + "step": 35792 + }, + { + "epoch": 2.3, + "learning_rate": 1.2340754417226649e-05, + "loss": 0.9087, + "step": 35793 + }, + { + "epoch": 2.3, + "learning_rate": 1.2340539728203698e-05, + "loss": 0.8919, + "step": 35794 + }, + { + "epoch": 2.3, + "learning_rate": 1.2340325039180747e-05, + "loss": 0.9847, + "step": 35795 + }, + { + "epoch": 2.3, + "learning_rate": 1.2340110350157797e-05, + "loss": 0.8029, + "step": 35796 + }, + { + "epoch": 2.3, + "learning_rate": 1.2339895661134846e-05, + "loss": 0.9176, + "step": 35797 + }, + { + "epoch": 2.3, + "learning_rate": 1.2339680972111895e-05, + "loss": 0.8666, + "step": 35798 + }, + { + "epoch": 2.3, + "learning_rate": 1.2339466283088946e-05, + "loss": 0.9012, + "step": 35799 + }, + { + "epoch": 2.3, + "learning_rate": 1.2339251594065997e-05, + "loss": 0.9091, + "step": 35800 + }, + { + "epoch": 2.3, + "learning_rate": 1.2339036905043047e-05, + "loss": 0.8133, + "step": 35801 + }, + { + "epoch": 2.3, + "learning_rate": 1.2338822216020096e-05, + "loss": 0.8848, + "step": 35802 + }, + { + "epoch": 2.3, + "learning_rate": 1.2338607526997147e-05, + "loss": 0.8779, + "step": 35803 + }, + { + "epoch": 2.3, + "learning_rate": 1.2338392837974196e-05, + "loss": 0.8874, + "step": 35804 + }, + { + "epoch": 2.3, + "learning_rate": 1.2338178148951245e-05, + "loss": 0.8981, + "step": 35805 + }, + { + "epoch": 2.3, + "learning_rate": 1.2337963459928296e-05, + "loss": 0.8516, + "step": 35806 + }, + { + "epoch": 2.3, + "learning_rate": 1.2337748770905345e-05, + "loss": 0.9261, + "step": 35807 + }, + { + "epoch": 2.3, + "learning_rate": 1.2337534081882394e-05, + "loss": 0.8288, + "step": 35808 + }, + { + "epoch": 2.3, + "learning_rate": 1.2337319392859444e-05, + "loss": 0.915, + "step": 35809 + }, + { + "epoch": 2.3, + "learning_rate": 1.2337104703836493e-05, + "loss": 0.8544, + "step": 35810 + }, + { + "epoch": 2.3, + "learning_rate": 1.2336890014813542e-05, + "loss": 0.9013, + "step": 35811 + }, + { + "epoch": 2.3, + "learning_rate": 1.2336675325790593e-05, + "loss": 0.9587, + "step": 35812 + }, + { + "epoch": 2.3, + "learning_rate": 1.2336460636767642e-05, + "loss": 0.8787, + "step": 35813 + }, + { + "epoch": 2.3, + "learning_rate": 1.2336245947744694e-05, + "loss": 0.8829, + "step": 35814 + }, + { + "epoch": 2.3, + "learning_rate": 1.2336031258721743e-05, + "loss": 0.9601, + "step": 35815 + }, + { + "epoch": 2.3, + "learning_rate": 1.2335816569698794e-05, + "loss": 0.8662, + "step": 35816 + }, + { + "epoch": 2.3, + "learning_rate": 1.2335601880675843e-05, + "loss": 0.8845, + "step": 35817 + }, + { + "epoch": 2.3, + "learning_rate": 1.2335387191652892e-05, + "loss": 0.8757, + "step": 35818 + }, + { + "epoch": 2.3, + "learning_rate": 1.2335172502629942e-05, + "loss": 0.9619, + "step": 35819 + }, + { + "epoch": 2.3, + "learning_rate": 1.2334957813606991e-05, + "loss": 0.9148, + "step": 35820 + }, + { + "epoch": 2.3, + "learning_rate": 1.233474312458404e-05, + "loss": 0.8525, + "step": 35821 + }, + { + "epoch": 2.3, + "learning_rate": 1.2334528435561091e-05, + "loss": 0.849, + "step": 35822 + }, + { + "epoch": 2.3, + "learning_rate": 1.233431374653814e-05, + "loss": 0.9389, + "step": 35823 + }, + { + "epoch": 2.3, + "learning_rate": 1.233409905751519e-05, + "loss": 0.9398, + "step": 35824 + }, + { + "epoch": 2.3, + "learning_rate": 1.233388436849224e-05, + "loss": 0.8701, + "step": 35825 + }, + { + "epoch": 2.3, + "learning_rate": 1.2333669679469288e-05, + "loss": 0.8606, + "step": 35826 + }, + { + "epoch": 2.31, + "learning_rate": 1.2333454990446339e-05, + "loss": 0.8824, + "step": 35827 + }, + { + "epoch": 2.31, + "learning_rate": 1.2333240301423388e-05, + "loss": 0.8837, + "step": 35828 + }, + { + "epoch": 2.31, + "learning_rate": 1.233302561240044e-05, + "loss": 0.9069, + "step": 35829 + }, + { + "epoch": 2.31, + "learning_rate": 1.233281092337749e-05, + "loss": 0.8652, + "step": 35830 + }, + { + "epoch": 2.31, + "learning_rate": 1.2332596234354538e-05, + "loss": 0.9323, + "step": 35831 + }, + { + "epoch": 2.31, + "learning_rate": 1.2332381545331589e-05, + "loss": 0.8876, + "step": 35832 + }, + { + "epoch": 2.31, + "learning_rate": 1.2332166856308638e-05, + "loss": 0.867, + "step": 35833 + }, + { + "epoch": 2.31, + "learning_rate": 1.2331952167285689e-05, + "loss": 0.97, + "step": 35834 + }, + { + "epoch": 2.31, + "learning_rate": 1.2331737478262738e-05, + "loss": 0.9391, + "step": 35835 + }, + { + "epoch": 2.31, + "learning_rate": 1.2331522789239787e-05, + "loss": 0.8814, + "step": 35836 + }, + { + "epoch": 2.31, + "learning_rate": 1.2331308100216837e-05, + "loss": 0.8537, + "step": 35837 + }, + { + "epoch": 2.31, + "learning_rate": 1.2331093411193886e-05, + "loss": 0.8456, + "step": 35838 + }, + { + "epoch": 2.31, + "learning_rate": 1.2330878722170935e-05, + "loss": 0.8867, + "step": 35839 + }, + { + "epoch": 2.31, + "learning_rate": 1.2330664033147986e-05, + "loss": 0.841, + "step": 35840 + }, + { + "epoch": 2.31, + "learning_rate": 1.2330449344125035e-05, + "loss": 0.8016, + "step": 35841 + }, + { + "epoch": 2.31, + "learning_rate": 1.2330234655102085e-05, + "loss": 0.9379, + "step": 35842 + }, + { + "epoch": 2.31, + "learning_rate": 1.2330019966079136e-05, + "loss": 0.8771, + "step": 35843 + }, + { + "epoch": 2.31, + "learning_rate": 1.2329805277056187e-05, + "loss": 0.8854, + "step": 35844 + }, + { + "epoch": 2.31, + "learning_rate": 1.2329590588033236e-05, + "loss": 0.8574, + "step": 35845 + }, + { + "epoch": 2.31, + "learning_rate": 1.2329375899010285e-05, + "loss": 0.9072, + "step": 35846 + }, + { + "epoch": 2.31, + "learning_rate": 1.2329161209987335e-05, + "loss": 0.8401, + "step": 35847 + }, + { + "epoch": 2.31, + "learning_rate": 1.2328946520964384e-05, + "loss": 0.9402, + "step": 35848 + }, + { + "epoch": 2.31, + "learning_rate": 1.2328731831941433e-05, + "loss": 0.8422, + "step": 35849 + }, + { + "epoch": 2.31, + "learning_rate": 1.2328517142918484e-05, + "loss": 0.9664, + "step": 35850 + }, + { + "epoch": 2.31, + "learning_rate": 1.2328302453895533e-05, + "loss": 0.8508, + "step": 35851 + }, + { + "epoch": 2.31, + "learning_rate": 1.2328087764872584e-05, + "loss": 0.9137, + "step": 35852 + }, + { + "epoch": 2.31, + "learning_rate": 1.2327873075849632e-05, + "loss": 0.9318, + "step": 35853 + }, + { + "epoch": 2.31, + "learning_rate": 1.2327658386826681e-05, + "loss": 0.8808, + "step": 35854 + }, + { + "epoch": 2.31, + "learning_rate": 1.2327443697803732e-05, + "loss": 0.8546, + "step": 35855 + }, + { + "epoch": 2.31, + "learning_rate": 1.2327229008780781e-05, + "loss": 0.8942, + "step": 35856 + }, + { + "epoch": 2.31, + "learning_rate": 1.232701431975783e-05, + "loss": 0.8967, + "step": 35857 + }, + { + "epoch": 2.31, + "learning_rate": 1.2326799630734882e-05, + "loss": 0.8989, + "step": 35858 + }, + { + "epoch": 2.31, + "learning_rate": 1.2326584941711931e-05, + "loss": 0.8502, + "step": 35859 + }, + { + "epoch": 2.31, + "learning_rate": 1.2326370252688982e-05, + "loss": 0.8934, + "step": 35860 + }, + { + "epoch": 2.31, + "learning_rate": 1.2326155563666031e-05, + "loss": 0.8579, + "step": 35861 + }, + { + "epoch": 2.31, + "learning_rate": 1.2325940874643082e-05, + "loss": 0.8835, + "step": 35862 + }, + { + "epoch": 2.31, + "learning_rate": 1.232572618562013e-05, + "loss": 0.8302, + "step": 35863 + }, + { + "epoch": 2.31, + "learning_rate": 1.232551149659718e-05, + "loss": 0.8136, + "step": 35864 + }, + { + "epoch": 2.31, + "learning_rate": 1.232529680757423e-05, + "loss": 0.9719, + "step": 35865 + }, + { + "epoch": 2.31, + "learning_rate": 1.232508211855128e-05, + "loss": 0.8454, + "step": 35866 + }, + { + "epoch": 2.31, + "learning_rate": 1.2324867429528328e-05, + "loss": 0.9717, + "step": 35867 + }, + { + "epoch": 2.31, + "learning_rate": 1.2324652740505379e-05, + "loss": 0.8894, + "step": 35868 + }, + { + "epoch": 2.31, + "learning_rate": 1.2324438051482428e-05, + "loss": 0.9694, + "step": 35869 + }, + { + "epoch": 2.31, + "learning_rate": 1.2324223362459477e-05, + "loss": 0.8394, + "step": 35870 + }, + { + "epoch": 2.31, + "learning_rate": 1.2324008673436527e-05, + "loss": 0.9048, + "step": 35871 + }, + { + "epoch": 2.31, + "learning_rate": 1.232379398441358e-05, + "loss": 0.8949, + "step": 35872 + }, + { + "epoch": 2.31, + "learning_rate": 1.2323579295390629e-05, + "loss": 0.9396, + "step": 35873 + }, + { + "epoch": 2.31, + "learning_rate": 1.2323364606367678e-05, + "loss": 0.8998, + "step": 35874 + }, + { + "epoch": 2.31, + "learning_rate": 1.2323149917344728e-05, + "loss": 1.0083, + "step": 35875 + }, + { + "epoch": 2.31, + "learning_rate": 1.2322935228321777e-05, + "loss": 0.8521, + "step": 35876 + }, + { + "epoch": 2.31, + "learning_rate": 1.2322720539298826e-05, + "loss": 0.9203, + "step": 35877 + }, + { + "epoch": 2.31, + "learning_rate": 1.2322505850275877e-05, + "loss": 0.8496, + "step": 35878 + }, + { + "epoch": 2.31, + "learning_rate": 1.2322291161252926e-05, + "loss": 0.9236, + "step": 35879 + }, + { + "epoch": 2.31, + "learning_rate": 1.2322076472229975e-05, + "loss": 0.8917, + "step": 35880 + }, + { + "epoch": 2.31, + "learning_rate": 1.2321861783207026e-05, + "loss": 0.9117, + "step": 35881 + }, + { + "epoch": 2.31, + "learning_rate": 1.2321647094184075e-05, + "loss": 0.8411, + "step": 35882 + }, + { + "epoch": 2.31, + "learning_rate": 1.2321432405161125e-05, + "loss": 0.8326, + "step": 35883 + }, + { + "epoch": 2.31, + "learning_rate": 1.2321217716138174e-05, + "loss": 0.8928, + "step": 35884 + }, + { + "epoch": 2.31, + "learning_rate": 1.2321003027115223e-05, + "loss": 0.8784, + "step": 35885 + }, + { + "epoch": 2.31, + "learning_rate": 1.2320788338092274e-05, + "loss": 0.925, + "step": 35886 + }, + { + "epoch": 2.31, + "learning_rate": 1.2320573649069324e-05, + "loss": 0.9464, + "step": 35887 + }, + { + "epoch": 2.31, + "learning_rate": 1.2320358960046375e-05, + "loss": 0.8204, + "step": 35888 + }, + { + "epoch": 2.31, + "learning_rate": 1.2320144271023424e-05, + "loss": 0.9177, + "step": 35889 + }, + { + "epoch": 2.31, + "learning_rate": 1.2319929582000473e-05, + "loss": 0.8551, + "step": 35890 + }, + { + "epoch": 2.31, + "learning_rate": 1.2319714892977524e-05, + "loss": 0.8895, + "step": 35891 + }, + { + "epoch": 2.31, + "learning_rate": 1.2319500203954573e-05, + "loss": 0.8957, + "step": 35892 + }, + { + "epoch": 2.31, + "learning_rate": 1.2319285514931623e-05, + "loss": 0.8628, + "step": 35893 + }, + { + "epoch": 2.31, + "learning_rate": 1.2319070825908672e-05, + "loss": 0.9115, + "step": 35894 + }, + { + "epoch": 2.31, + "learning_rate": 1.2318856136885721e-05, + "loss": 0.9091, + "step": 35895 + }, + { + "epoch": 2.31, + "learning_rate": 1.2318641447862772e-05, + "loss": 0.8653, + "step": 35896 + }, + { + "epoch": 2.31, + "learning_rate": 1.231842675883982e-05, + "loss": 0.921, + "step": 35897 + }, + { + "epoch": 2.31, + "learning_rate": 1.231821206981687e-05, + "loss": 0.8215, + "step": 35898 + }, + { + "epoch": 2.31, + "learning_rate": 1.231799738079392e-05, + "loss": 0.9409, + "step": 35899 + }, + { + "epoch": 2.31, + "learning_rate": 1.231778269177097e-05, + "loss": 0.7855, + "step": 35900 + }, + { + "epoch": 2.31, + "learning_rate": 1.2317568002748022e-05, + "loss": 0.8394, + "step": 35901 + }, + { + "epoch": 2.31, + "learning_rate": 1.231735331372507e-05, + "loss": 0.9882, + "step": 35902 + }, + { + "epoch": 2.31, + "learning_rate": 1.2317138624702121e-05, + "loss": 0.8734, + "step": 35903 + }, + { + "epoch": 2.31, + "learning_rate": 1.231692393567917e-05, + "loss": 0.9177, + "step": 35904 + }, + { + "epoch": 2.31, + "learning_rate": 1.231670924665622e-05, + "loss": 0.9182, + "step": 35905 + }, + { + "epoch": 2.31, + "learning_rate": 1.231649455763327e-05, + "loss": 0.8852, + "step": 35906 + }, + { + "epoch": 2.31, + "learning_rate": 1.2316279868610319e-05, + "loss": 0.8233, + "step": 35907 + }, + { + "epoch": 2.31, + "learning_rate": 1.2316065179587368e-05, + "loss": 0.9055, + "step": 35908 + }, + { + "epoch": 2.31, + "learning_rate": 1.2315850490564419e-05, + "loss": 0.9408, + "step": 35909 + }, + { + "epoch": 2.31, + "learning_rate": 1.2315635801541468e-05, + "loss": 0.9219, + "step": 35910 + }, + { + "epoch": 2.31, + "learning_rate": 1.2315421112518518e-05, + "loss": 0.9585, + "step": 35911 + }, + { + "epoch": 2.31, + "learning_rate": 1.2315206423495567e-05, + "loss": 0.8898, + "step": 35912 + }, + { + "epoch": 2.31, + "learning_rate": 1.2314991734472616e-05, + "loss": 0.8478, + "step": 35913 + }, + { + "epoch": 2.31, + "learning_rate": 1.2314777045449667e-05, + "loss": 0.9058, + "step": 35914 + }, + { + "epoch": 2.31, + "learning_rate": 1.2314562356426716e-05, + "loss": 0.8359, + "step": 35915 + }, + { + "epoch": 2.31, + "learning_rate": 1.2314347667403768e-05, + "loss": 0.9069, + "step": 35916 + }, + { + "epoch": 2.31, + "learning_rate": 1.2314132978380817e-05, + "loss": 0.8051, + "step": 35917 + }, + { + "epoch": 2.31, + "learning_rate": 1.2313918289357866e-05, + "loss": 0.8948, + "step": 35918 + }, + { + "epoch": 2.31, + "learning_rate": 1.2313703600334917e-05, + "loss": 0.9531, + "step": 35919 + }, + { + "epoch": 2.31, + "learning_rate": 1.2313488911311966e-05, + "loss": 0.8256, + "step": 35920 + }, + { + "epoch": 2.31, + "learning_rate": 1.2313274222289016e-05, + "loss": 0.9198, + "step": 35921 + }, + { + "epoch": 2.31, + "learning_rate": 1.2313059533266065e-05, + "loss": 0.9354, + "step": 35922 + }, + { + "epoch": 2.31, + "learning_rate": 1.2312844844243114e-05, + "loss": 0.843, + "step": 35923 + }, + { + "epoch": 2.31, + "learning_rate": 1.2312630155220165e-05, + "loss": 0.9381, + "step": 35924 + }, + { + "epoch": 2.31, + "learning_rate": 1.2312415466197214e-05, + "loss": 0.8948, + "step": 35925 + }, + { + "epoch": 2.31, + "learning_rate": 1.2312200777174263e-05, + "loss": 0.8343, + "step": 35926 + }, + { + "epoch": 2.31, + "learning_rate": 1.2311986088151314e-05, + "loss": 0.8524, + "step": 35927 + }, + { + "epoch": 2.31, + "learning_rate": 1.2311771399128362e-05, + "loss": 0.9106, + "step": 35928 + }, + { + "epoch": 2.31, + "learning_rate": 1.2311556710105413e-05, + "loss": 0.9478, + "step": 35929 + }, + { + "epoch": 2.31, + "learning_rate": 1.2311342021082464e-05, + "loss": 0.8296, + "step": 35930 + }, + { + "epoch": 2.31, + "learning_rate": 1.2311127332059514e-05, + "loss": 0.8121, + "step": 35931 + }, + { + "epoch": 2.31, + "learning_rate": 1.2310912643036563e-05, + "loss": 0.8098, + "step": 35932 + }, + { + "epoch": 2.31, + "learning_rate": 1.2310697954013612e-05, + "loss": 0.8335, + "step": 35933 + }, + { + "epoch": 2.31, + "learning_rate": 1.2310483264990663e-05, + "loss": 0.8688, + "step": 35934 + }, + { + "epoch": 2.31, + "learning_rate": 1.2310268575967712e-05, + "loss": 0.9174, + "step": 35935 + }, + { + "epoch": 2.31, + "learning_rate": 1.2310053886944761e-05, + "loss": 0.8379, + "step": 35936 + }, + { + "epoch": 2.31, + "learning_rate": 1.2309839197921812e-05, + "loss": 0.9147, + "step": 35937 + }, + { + "epoch": 2.31, + "learning_rate": 1.230962450889886e-05, + "loss": 0.8547, + "step": 35938 + }, + { + "epoch": 2.31, + "learning_rate": 1.2309409819875911e-05, + "loss": 0.9111, + "step": 35939 + }, + { + "epoch": 2.31, + "learning_rate": 1.230919513085296e-05, + "loss": 0.9854, + "step": 35940 + }, + { + "epoch": 2.31, + "learning_rate": 1.230898044183001e-05, + "loss": 0.8346, + "step": 35941 + }, + { + "epoch": 2.31, + "learning_rate": 1.230876575280706e-05, + "loss": 0.8171, + "step": 35942 + }, + { + "epoch": 2.31, + "learning_rate": 1.2308551063784109e-05, + "loss": 0.9013, + "step": 35943 + }, + { + "epoch": 2.31, + "learning_rate": 1.2308336374761158e-05, + "loss": 0.9708, + "step": 35944 + }, + { + "epoch": 2.31, + "learning_rate": 1.230812168573821e-05, + "loss": 0.92, + "step": 35945 + }, + { + "epoch": 2.31, + "learning_rate": 1.2307906996715259e-05, + "loss": 0.9167, + "step": 35946 + }, + { + "epoch": 2.31, + "learning_rate": 1.230769230769231e-05, + "loss": 0.8915, + "step": 35947 + }, + { + "epoch": 2.31, + "learning_rate": 1.2307477618669359e-05, + "loss": 0.8811, + "step": 35948 + }, + { + "epoch": 2.31, + "learning_rate": 1.230726292964641e-05, + "loss": 0.9396, + "step": 35949 + }, + { + "epoch": 2.31, + "learning_rate": 1.2307048240623458e-05, + "loss": 0.8893, + "step": 35950 + }, + { + "epoch": 2.31, + "learning_rate": 1.2306833551600507e-05, + "loss": 0.8276, + "step": 35951 + }, + { + "epoch": 2.31, + "learning_rate": 1.2306618862577558e-05, + "loss": 0.9074, + "step": 35952 + }, + { + "epoch": 2.31, + "learning_rate": 1.2306404173554607e-05, + "loss": 0.8673, + "step": 35953 + }, + { + "epoch": 2.31, + "learning_rate": 1.2306189484531656e-05, + "loss": 0.8177, + "step": 35954 + }, + { + "epoch": 2.31, + "learning_rate": 1.2305974795508707e-05, + "loss": 0.9273, + "step": 35955 + }, + { + "epoch": 2.31, + "learning_rate": 1.2305760106485756e-05, + "loss": 0.9005, + "step": 35956 + }, + { + "epoch": 2.31, + "learning_rate": 1.2305545417462804e-05, + "loss": 0.9104, + "step": 35957 + }, + { + "epoch": 2.31, + "learning_rate": 1.2305330728439855e-05, + "loss": 0.9515, + "step": 35958 + }, + { + "epoch": 2.31, + "learning_rate": 1.2305116039416907e-05, + "loss": 0.892, + "step": 35959 + }, + { + "epoch": 2.31, + "learning_rate": 1.2304901350393956e-05, + "loss": 0.8971, + "step": 35960 + }, + { + "epoch": 2.31, + "learning_rate": 1.2304686661371005e-05, + "loss": 0.8586, + "step": 35961 + }, + { + "epoch": 2.31, + "learning_rate": 1.2304471972348056e-05, + "loss": 0.8586, + "step": 35962 + }, + { + "epoch": 2.31, + "learning_rate": 1.2304257283325105e-05, + "loss": 0.8696, + "step": 35963 + }, + { + "epoch": 2.31, + "learning_rate": 1.2304042594302154e-05, + "loss": 0.8679, + "step": 35964 + }, + { + "epoch": 2.31, + "learning_rate": 1.2303827905279205e-05, + "loss": 0.9942, + "step": 35965 + }, + { + "epoch": 2.31, + "learning_rate": 1.2303613216256254e-05, + "loss": 0.8595, + "step": 35966 + }, + { + "epoch": 2.31, + "learning_rate": 1.2303398527233303e-05, + "loss": 0.9164, + "step": 35967 + }, + { + "epoch": 2.31, + "learning_rate": 1.2303183838210353e-05, + "loss": 0.909, + "step": 35968 + }, + { + "epoch": 2.31, + "learning_rate": 1.2302969149187402e-05, + "loss": 0.8953, + "step": 35969 + }, + { + "epoch": 2.31, + "learning_rate": 1.2302754460164453e-05, + "loss": 0.9561, + "step": 35970 + }, + { + "epoch": 2.31, + "learning_rate": 1.2302539771141502e-05, + "loss": 0.8663, + "step": 35971 + }, + { + "epoch": 2.31, + "learning_rate": 1.230232508211855e-05, + "loss": 0.9138, + "step": 35972 + }, + { + "epoch": 2.31, + "learning_rate": 1.2302110393095601e-05, + "loss": 0.9018, + "step": 35973 + }, + { + "epoch": 2.31, + "learning_rate": 1.2301895704072652e-05, + "loss": 0.913, + "step": 35974 + }, + { + "epoch": 2.31, + "learning_rate": 1.2301681015049703e-05, + "loss": 0.9584, + "step": 35975 + }, + { + "epoch": 2.31, + "learning_rate": 1.2301466326026752e-05, + "loss": 0.8436, + "step": 35976 + }, + { + "epoch": 2.31, + "learning_rate": 1.23012516370038e-05, + "loss": 0.9456, + "step": 35977 + }, + { + "epoch": 2.31, + "learning_rate": 1.2301036947980851e-05, + "loss": 0.9403, + "step": 35978 + }, + { + "epoch": 2.31, + "learning_rate": 1.23008222589579e-05, + "loss": 0.8803, + "step": 35979 + }, + { + "epoch": 2.31, + "learning_rate": 1.2300607569934951e-05, + "loss": 0.9295, + "step": 35980 + }, + { + "epoch": 2.31, + "learning_rate": 1.2300392880912e-05, + "loss": 0.8901, + "step": 35981 + }, + { + "epoch": 2.31, + "learning_rate": 1.2300178191889049e-05, + "loss": 0.9046, + "step": 35982 + }, + { + "epoch": 2.32, + "learning_rate": 1.22999635028661e-05, + "loss": 0.8699, + "step": 35983 + }, + { + "epoch": 2.32, + "learning_rate": 1.2299748813843149e-05, + "loss": 0.8868, + "step": 35984 + }, + { + "epoch": 2.32, + "learning_rate": 1.2299534124820198e-05, + "loss": 0.9018, + "step": 35985 + }, + { + "epoch": 2.32, + "learning_rate": 1.2299319435797248e-05, + "loss": 0.873, + "step": 35986 + }, + { + "epoch": 2.32, + "learning_rate": 1.2299104746774297e-05, + "loss": 0.8914, + "step": 35987 + }, + { + "epoch": 2.32, + "learning_rate": 1.229889005775135e-05, + "loss": 0.8745, + "step": 35988 + }, + { + "epoch": 2.32, + "learning_rate": 1.2298675368728398e-05, + "loss": 0.9037, + "step": 35989 + }, + { + "epoch": 2.32, + "learning_rate": 1.2298460679705449e-05, + "loss": 0.8364, + "step": 35990 + }, + { + "epoch": 2.32, + "learning_rate": 1.2298245990682498e-05, + "loss": 0.8288, + "step": 35991 + }, + { + "epoch": 2.32, + "learning_rate": 1.2298031301659547e-05, + "loss": 0.9551, + "step": 35992 + }, + { + "epoch": 2.32, + "learning_rate": 1.2297816612636598e-05, + "loss": 0.9743, + "step": 35993 + }, + { + "epoch": 2.32, + "learning_rate": 1.2297601923613647e-05, + "loss": 0.9471, + "step": 35994 + }, + { + "epoch": 2.32, + "learning_rate": 1.2297387234590696e-05, + "loss": 0.9229, + "step": 35995 + }, + { + "epoch": 2.32, + "learning_rate": 1.2297172545567746e-05, + "loss": 0.8919, + "step": 35996 + }, + { + "epoch": 2.32, + "learning_rate": 1.2296957856544795e-05, + "loss": 0.9651, + "step": 35997 + }, + { + "epoch": 2.32, + "learning_rate": 1.2296743167521846e-05, + "loss": 0.9301, + "step": 35998 + }, + { + "epoch": 2.32, + "learning_rate": 1.2296528478498895e-05, + "loss": 0.8887, + "step": 35999 + }, + { + "epoch": 2.32, + "learning_rate": 1.2296313789475944e-05, + "loss": 0.8995, + "step": 36000 + }, + { + "epoch": 2.32, + "eval_laion/cider": 0.5257531603109513, + "eval_laion/meteor": 0.39813432385708836, + "eval_laion/pred_num_words": 38.970703125, + "eval_laion/sacrebleu": 16.365987575182242, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0420863628387451, + "eval_runtime": 101.616, + "eval_samples_per_second": 4.911, + "eval_steps_per_second": 0.315, + "step": 36000 + }, + { + "epoch": 2.32, + "learning_rate": 1.2296099100452995e-05, + "loss": 0.8627, + "step": 36001 + }, + { + "epoch": 2.32, + "learning_rate": 1.2295884411430043e-05, + "loss": 0.8839, + "step": 36002 + }, + { + "epoch": 2.32, + "learning_rate": 1.2295669722407096e-05, + "loss": 0.8535, + "step": 36003 + }, + { + "epoch": 2.32, + "learning_rate": 1.2295455033384145e-05, + "loss": 0.9722, + "step": 36004 + }, + { + "epoch": 2.32, + "learning_rate": 1.2295240344361194e-05, + "loss": 0.8757, + "step": 36005 + }, + { + "epoch": 2.32, + "learning_rate": 1.2295025655338244e-05, + "loss": 0.9282, + "step": 36006 + }, + { + "epoch": 2.32, + "learning_rate": 1.2294810966315293e-05, + "loss": 0.9262, + "step": 36007 + }, + { + "epoch": 2.32, + "learning_rate": 1.2294596277292344e-05, + "loss": 0.9072, + "step": 36008 + }, + { + "epoch": 2.32, + "learning_rate": 1.2294381588269393e-05, + "loss": 0.9534, + "step": 36009 + }, + { + "epoch": 2.32, + "learning_rate": 1.2294166899246442e-05, + "loss": 0.9263, + "step": 36010 + }, + { + "epoch": 2.32, + "learning_rate": 1.2293952210223493e-05, + "loss": 0.9151, + "step": 36011 + }, + { + "epoch": 2.32, + "learning_rate": 1.2293737521200542e-05, + "loss": 0.8762, + "step": 36012 + }, + { + "epoch": 2.32, + "learning_rate": 1.229352283217759e-05, + "loss": 0.8733, + "step": 36013 + }, + { + "epoch": 2.32, + "learning_rate": 1.2293308143154641e-05, + "loss": 0.8941, + "step": 36014 + }, + { + "epoch": 2.32, + "learning_rate": 1.229309345413169e-05, + "loss": 0.8629, + "step": 36015 + }, + { + "epoch": 2.32, + "learning_rate": 1.229287876510874e-05, + "loss": 0.8566, + "step": 36016 + }, + { + "epoch": 2.32, + "learning_rate": 1.2292664076085791e-05, + "loss": 0.8483, + "step": 36017 + }, + { + "epoch": 2.32, + "learning_rate": 1.2292449387062842e-05, + "loss": 0.8857, + "step": 36018 + }, + { + "epoch": 2.32, + "learning_rate": 1.2292234698039891e-05, + "loss": 0.8849, + "step": 36019 + }, + { + "epoch": 2.32, + "learning_rate": 1.229202000901694e-05, + "loss": 0.9051, + "step": 36020 + }, + { + "epoch": 2.32, + "learning_rate": 1.229180531999399e-05, + "loss": 0.9756, + "step": 36021 + }, + { + "epoch": 2.32, + "learning_rate": 1.229159063097104e-05, + "loss": 0.8895, + "step": 36022 + }, + { + "epoch": 2.32, + "learning_rate": 1.2291375941948089e-05, + "loss": 0.8596, + "step": 36023 + }, + { + "epoch": 2.32, + "learning_rate": 1.229116125292514e-05, + "loss": 0.9763, + "step": 36024 + }, + { + "epoch": 2.32, + "learning_rate": 1.2290946563902188e-05, + "loss": 0.8966, + "step": 36025 + }, + { + "epoch": 2.32, + "learning_rate": 1.2290731874879239e-05, + "loss": 0.9226, + "step": 36026 + }, + { + "epoch": 2.32, + "learning_rate": 1.2290517185856288e-05, + "loss": 0.8726, + "step": 36027 + }, + { + "epoch": 2.32, + "learning_rate": 1.2290302496833337e-05, + "loss": 0.9263, + "step": 36028 + }, + { + "epoch": 2.32, + "learning_rate": 1.2290087807810388e-05, + "loss": 0.9079, + "step": 36029 + }, + { + "epoch": 2.32, + "learning_rate": 1.2289873118787437e-05, + "loss": 0.8822, + "step": 36030 + }, + { + "epoch": 2.32, + "learning_rate": 1.2289658429764485e-05, + "loss": 0.8725, + "step": 36031 + }, + { + "epoch": 2.32, + "learning_rate": 1.2289443740741538e-05, + "loss": 0.8326, + "step": 36032 + }, + { + "epoch": 2.32, + "learning_rate": 1.2289229051718587e-05, + "loss": 0.8691, + "step": 36033 + }, + { + "epoch": 2.32, + "learning_rate": 1.2289014362695637e-05, + "loss": 0.8759, + "step": 36034 + }, + { + "epoch": 2.32, + "learning_rate": 1.2288799673672686e-05, + "loss": 0.8304, + "step": 36035 + }, + { + "epoch": 2.32, + "learning_rate": 1.2288584984649737e-05, + "loss": 0.934, + "step": 36036 + }, + { + "epoch": 2.32, + "learning_rate": 1.2288370295626786e-05, + "loss": 0.9395, + "step": 36037 + }, + { + "epoch": 2.32, + "learning_rate": 1.2288155606603835e-05, + "loss": 0.8925, + "step": 36038 + }, + { + "epoch": 2.32, + "learning_rate": 1.2287940917580886e-05, + "loss": 0.9133, + "step": 36039 + }, + { + "epoch": 2.32, + "learning_rate": 1.2287726228557935e-05, + "loss": 0.8653, + "step": 36040 + }, + { + "epoch": 2.32, + "learning_rate": 1.2287511539534984e-05, + "loss": 0.8737, + "step": 36041 + }, + { + "epoch": 2.32, + "learning_rate": 1.2287296850512034e-05, + "loss": 0.822, + "step": 36042 + }, + { + "epoch": 2.32, + "learning_rate": 1.2287082161489083e-05, + "loss": 0.9592, + "step": 36043 + }, + { + "epoch": 2.32, + "learning_rate": 1.2286867472466132e-05, + "loss": 1.0048, + "step": 36044 + }, + { + "epoch": 2.32, + "learning_rate": 1.2286652783443183e-05, + "loss": 0.8711, + "step": 36045 + }, + { + "epoch": 2.32, + "learning_rate": 1.2286438094420235e-05, + "loss": 0.9271, + "step": 36046 + }, + { + "epoch": 2.32, + "learning_rate": 1.2286223405397284e-05, + "loss": 0.9173, + "step": 36047 + }, + { + "epoch": 2.32, + "learning_rate": 1.2286008716374333e-05, + "loss": 0.8408, + "step": 36048 + }, + { + "epoch": 2.32, + "learning_rate": 1.2285794027351384e-05, + "loss": 0.931, + "step": 36049 + }, + { + "epoch": 2.32, + "learning_rate": 1.2285579338328433e-05, + "loss": 0.902, + "step": 36050 + }, + { + "epoch": 2.32, + "learning_rate": 1.2285364649305482e-05, + "loss": 0.9124, + "step": 36051 + }, + { + "epoch": 2.32, + "learning_rate": 1.2285149960282532e-05, + "loss": 0.9552, + "step": 36052 + }, + { + "epoch": 2.32, + "learning_rate": 1.2284935271259581e-05, + "loss": 0.9144, + "step": 36053 + }, + { + "epoch": 2.32, + "learning_rate": 1.228472058223663e-05, + "loss": 0.899, + "step": 36054 + }, + { + "epoch": 2.32, + "learning_rate": 1.2284505893213681e-05, + "loss": 0.9983, + "step": 36055 + }, + { + "epoch": 2.32, + "learning_rate": 1.228429120419073e-05, + "loss": 0.8807, + "step": 36056 + }, + { + "epoch": 2.32, + "learning_rate": 1.228407651516778e-05, + "loss": 0.9529, + "step": 36057 + }, + { + "epoch": 2.32, + "learning_rate": 1.228386182614483e-05, + "loss": 0.9126, + "step": 36058 + }, + { + "epoch": 2.32, + "learning_rate": 1.2283647137121879e-05, + "loss": 0.9066, + "step": 36059 + }, + { + "epoch": 2.32, + "learning_rate": 1.228343244809893e-05, + "loss": 0.919, + "step": 36060 + }, + { + "epoch": 2.32, + "learning_rate": 1.228321775907598e-05, + "loss": 0.9263, + "step": 36061 + }, + { + "epoch": 2.32, + "learning_rate": 1.228300307005303e-05, + "loss": 0.9055, + "step": 36062 + }, + { + "epoch": 2.32, + "learning_rate": 1.228278838103008e-05, + "loss": 0.8968, + "step": 36063 + }, + { + "epoch": 2.32, + "learning_rate": 1.2282573692007128e-05, + "loss": 0.9029, + "step": 36064 + }, + { + "epoch": 2.32, + "learning_rate": 1.2282359002984179e-05, + "loss": 0.9105, + "step": 36065 + }, + { + "epoch": 2.32, + "learning_rate": 1.2282144313961228e-05, + "loss": 0.9, + "step": 36066 + }, + { + "epoch": 2.32, + "learning_rate": 1.2281929624938279e-05, + "loss": 0.8992, + "step": 36067 + }, + { + "epoch": 2.32, + "learning_rate": 1.2281714935915328e-05, + "loss": 1.014, + "step": 36068 + }, + { + "epoch": 2.32, + "learning_rate": 1.2281500246892377e-05, + "loss": 0.8666, + "step": 36069 + }, + { + "epoch": 2.32, + "learning_rate": 1.2281285557869427e-05, + "loss": 0.915, + "step": 36070 + }, + { + "epoch": 2.32, + "learning_rate": 1.2281070868846476e-05, + "loss": 0.8835, + "step": 36071 + }, + { + "epoch": 2.32, + "learning_rate": 1.2280856179823525e-05, + "loss": 0.9661, + "step": 36072 + }, + { + "epoch": 2.32, + "learning_rate": 1.2280641490800576e-05, + "loss": 0.8812, + "step": 36073 + }, + { + "epoch": 2.32, + "learning_rate": 1.2280426801777625e-05, + "loss": 0.8905, + "step": 36074 + }, + { + "epoch": 2.32, + "learning_rate": 1.2280212112754677e-05, + "loss": 0.888, + "step": 36075 + }, + { + "epoch": 2.32, + "learning_rate": 1.2279997423731726e-05, + "loss": 0.8572, + "step": 36076 + }, + { + "epoch": 2.32, + "learning_rate": 1.2279782734708777e-05, + "loss": 0.9134, + "step": 36077 + }, + { + "epoch": 2.32, + "learning_rate": 1.2279568045685826e-05, + "loss": 0.8331, + "step": 36078 + }, + { + "epoch": 2.32, + "learning_rate": 1.2279353356662875e-05, + "loss": 0.9156, + "step": 36079 + }, + { + "epoch": 2.32, + "learning_rate": 1.2279138667639925e-05, + "loss": 0.9035, + "step": 36080 + }, + { + "epoch": 2.32, + "learning_rate": 1.2278923978616974e-05, + "loss": 0.8973, + "step": 36081 + }, + { + "epoch": 2.32, + "learning_rate": 1.2278709289594023e-05, + "loss": 0.8041, + "step": 36082 + }, + { + "epoch": 2.32, + "learning_rate": 1.2278494600571074e-05, + "loss": 0.8929, + "step": 36083 + }, + { + "epoch": 2.32, + "learning_rate": 1.2278279911548123e-05, + "loss": 0.8847, + "step": 36084 + }, + { + "epoch": 2.32, + "learning_rate": 1.2278065222525174e-05, + "loss": 0.8787, + "step": 36085 + }, + { + "epoch": 2.32, + "learning_rate": 1.2277850533502223e-05, + "loss": 0.92, + "step": 36086 + }, + { + "epoch": 2.32, + "learning_rate": 1.2277635844479272e-05, + "loss": 0.8905, + "step": 36087 + }, + { + "epoch": 2.32, + "learning_rate": 1.2277421155456322e-05, + "loss": 0.8825, + "step": 36088 + }, + { + "epoch": 2.32, + "learning_rate": 1.2277206466433371e-05, + "loss": 0.864, + "step": 36089 + }, + { + "epoch": 2.32, + "learning_rate": 1.2276991777410424e-05, + "loss": 0.8573, + "step": 36090 + }, + { + "epoch": 2.32, + "learning_rate": 1.2276777088387473e-05, + "loss": 0.8464, + "step": 36091 + }, + { + "epoch": 2.32, + "learning_rate": 1.2276562399364521e-05, + "loss": 0.8902, + "step": 36092 + }, + { + "epoch": 2.32, + "learning_rate": 1.2276347710341572e-05, + "loss": 0.9145, + "step": 36093 + }, + { + "epoch": 2.32, + "learning_rate": 1.2276133021318621e-05, + "loss": 0.8721, + "step": 36094 + }, + { + "epoch": 2.32, + "learning_rate": 1.2275918332295672e-05, + "loss": 0.8767, + "step": 36095 + }, + { + "epoch": 2.32, + "learning_rate": 1.227570364327272e-05, + "loss": 0.8408, + "step": 36096 + }, + { + "epoch": 2.32, + "learning_rate": 1.227548895424977e-05, + "loss": 0.844, + "step": 36097 + }, + { + "epoch": 2.32, + "learning_rate": 1.227527426522682e-05, + "loss": 0.9248, + "step": 36098 + }, + { + "epoch": 2.32, + "learning_rate": 1.227505957620387e-05, + "loss": 0.9044, + "step": 36099 + }, + { + "epoch": 2.32, + "learning_rate": 1.2274844887180918e-05, + "loss": 0.9027, + "step": 36100 + }, + { + "epoch": 2.32, + "learning_rate": 1.2274630198157969e-05, + "loss": 0.8978, + "step": 36101 + }, + { + "epoch": 2.32, + "learning_rate": 1.2274415509135018e-05, + "loss": 0.9887, + "step": 36102 + }, + { + "epoch": 2.32, + "learning_rate": 1.2274200820112069e-05, + "loss": 0.8715, + "step": 36103 + }, + { + "epoch": 2.32, + "learning_rate": 1.227398613108912e-05, + "loss": 0.9006, + "step": 36104 + }, + { + "epoch": 2.32, + "learning_rate": 1.227377144206617e-05, + "loss": 0.9005, + "step": 36105 + }, + { + "epoch": 2.32, + "learning_rate": 1.2273556753043219e-05, + "loss": 0.862, + "step": 36106 + }, + { + "epoch": 2.32, + "learning_rate": 1.2273342064020268e-05, + "loss": 0.9125, + "step": 36107 + }, + { + "epoch": 2.32, + "learning_rate": 1.2273127374997318e-05, + "loss": 0.9076, + "step": 36108 + }, + { + "epoch": 2.32, + "learning_rate": 1.2272912685974367e-05, + "loss": 0.9459, + "step": 36109 + }, + { + "epoch": 2.32, + "learning_rate": 1.2272697996951416e-05, + "loss": 0.8613, + "step": 36110 + }, + { + "epoch": 2.32, + "learning_rate": 1.2272483307928467e-05, + "loss": 0.9045, + "step": 36111 + }, + { + "epoch": 2.32, + "learning_rate": 1.2272268618905516e-05, + "loss": 0.9102, + "step": 36112 + }, + { + "epoch": 2.32, + "learning_rate": 1.2272053929882567e-05, + "loss": 0.9222, + "step": 36113 + }, + { + "epoch": 2.32, + "learning_rate": 1.2271839240859616e-05, + "loss": 0.875, + "step": 36114 + }, + { + "epoch": 2.32, + "learning_rate": 1.2271624551836665e-05, + "loss": 0.8872, + "step": 36115 + }, + { + "epoch": 2.32, + "learning_rate": 1.2271409862813715e-05, + "loss": 0.898, + "step": 36116 + }, + { + "epoch": 2.32, + "learning_rate": 1.2271195173790764e-05, + "loss": 0.9062, + "step": 36117 + }, + { + "epoch": 2.32, + "learning_rate": 1.2270980484767813e-05, + "loss": 0.9522, + "step": 36118 + }, + { + "epoch": 2.32, + "learning_rate": 1.2270765795744866e-05, + "loss": 0.8815, + "step": 36119 + }, + { + "epoch": 2.32, + "learning_rate": 1.2270551106721915e-05, + "loss": 0.8912, + "step": 36120 + }, + { + "epoch": 2.32, + "learning_rate": 1.2270336417698965e-05, + "loss": 0.9002, + "step": 36121 + }, + { + "epoch": 2.32, + "learning_rate": 1.2270121728676014e-05, + "loss": 0.984, + "step": 36122 + }, + { + "epoch": 2.32, + "learning_rate": 1.2269907039653065e-05, + "loss": 0.9634, + "step": 36123 + }, + { + "epoch": 2.32, + "learning_rate": 1.2269692350630114e-05, + "loss": 0.9012, + "step": 36124 + }, + { + "epoch": 2.32, + "learning_rate": 1.2269477661607163e-05, + "loss": 0.881, + "step": 36125 + }, + { + "epoch": 2.32, + "learning_rate": 1.2269262972584213e-05, + "loss": 0.8752, + "step": 36126 + }, + { + "epoch": 2.32, + "learning_rate": 1.2269048283561262e-05, + "loss": 0.9107, + "step": 36127 + }, + { + "epoch": 2.32, + "learning_rate": 1.2268833594538311e-05, + "loss": 0.9419, + "step": 36128 + }, + { + "epoch": 2.32, + "learning_rate": 1.2268618905515362e-05, + "loss": 0.8514, + "step": 36129 + }, + { + "epoch": 2.32, + "learning_rate": 1.2268404216492411e-05, + "loss": 0.8437, + "step": 36130 + }, + { + "epoch": 2.32, + "learning_rate": 1.226818952746946e-05, + "loss": 0.9351, + "step": 36131 + }, + { + "epoch": 2.32, + "learning_rate": 1.226797483844651e-05, + "loss": 0.8752, + "step": 36132 + }, + { + "epoch": 2.32, + "learning_rate": 1.2267760149423563e-05, + "loss": 0.8933, + "step": 36133 + }, + { + "epoch": 2.32, + "learning_rate": 1.2267545460400612e-05, + "loss": 1.0132, + "step": 36134 + }, + { + "epoch": 2.32, + "learning_rate": 1.226733077137766e-05, + "loss": 0.8531, + "step": 36135 + }, + { + "epoch": 2.32, + "learning_rate": 1.2267116082354711e-05, + "loss": 0.8353, + "step": 36136 + }, + { + "epoch": 2.32, + "learning_rate": 1.226690139333176e-05, + "loss": 0.9035, + "step": 36137 + }, + { + "epoch": 2.33, + "learning_rate": 1.226668670430881e-05, + "loss": 0.9259, + "step": 36138 + }, + { + "epoch": 2.33, + "learning_rate": 1.226647201528586e-05, + "loss": 0.887, + "step": 36139 + }, + { + "epoch": 2.33, + "learning_rate": 1.2266257326262909e-05, + "loss": 0.9091, + "step": 36140 + }, + { + "epoch": 2.33, + "learning_rate": 1.2266042637239958e-05, + "loss": 0.8457, + "step": 36141 + }, + { + "epoch": 2.33, + "learning_rate": 1.2265827948217009e-05, + "loss": 0.8767, + "step": 36142 + }, + { + "epoch": 2.33, + "learning_rate": 1.2265613259194058e-05, + "loss": 0.8634, + "step": 36143 + }, + { + "epoch": 2.33, + "learning_rate": 1.2265398570171108e-05, + "loss": 0.8985, + "step": 36144 + }, + { + "epoch": 2.33, + "learning_rate": 1.2265183881148157e-05, + "loss": 0.8967, + "step": 36145 + }, + { + "epoch": 2.33, + "learning_rate": 1.2264969192125206e-05, + "loss": 0.8736, + "step": 36146 + }, + { + "epoch": 2.33, + "learning_rate": 1.2264754503102257e-05, + "loss": 0.9432, + "step": 36147 + }, + { + "epoch": 2.33, + "learning_rate": 1.2264539814079308e-05, + "loss": 0.851, + "step": 36148 + }, + { + "epoch": 2.33, + "learning_rate": 1.2264325125056358e-05, + "loss": 0.8909, + "step": 36149 + }, + { + "epoch": 2.33, + "learning_rate": 1.2264110436033407e-05, + "loss": 0.8037, + "step": 36150 + }, + { + "epoch": 2.33, + "learning_rate": 1.2263895747010456e-05, + "loss": 0.8975, + "step": 36151 + }, + { + "epoch": 2.33, + "learning_rate": 1.2263681057987507e-05, + "loss": 0.8678, + "step": 36152 + }, + { + "epoch": 2.33, + "learning_rate": 1.2263466368964556e-05, + "loss": 0.9184, + "step": 36153 + }, + { + "epoch": 2.33, + "learning_rate": 1.2263251679941606e-05, + "loss": 0.9056, + "step": 36154 + }, + { + "epoch": 2.33, + "learning_rate": 1.2263036990918655e-05, + "loss": 0.9527, + "step": 36155 + }, + { + "epoch": 2.33, + "learning_rate": 1.2262822301895704e-05, + "loss": 0.9002, + "step": 36156 + }, + { + "epoch": 2.33, + "learning_rate": 1.2262607612872755e-05, + "loss": 0.9083, + "step": 36157 + }, + { + "epoch": 2.33, + "learning_rate": 1.2262392923849804e-05, + "loss": 0.8798, + "step": 36158 + }, + { + "epoch": 2.33, + "learning_rate": 1.2262178234826853e-05, + "loss": 0.9459, + "step": 36159 + }, + { + "epoch": 2.33, + "learning_rate": 1.2261963545803904e-05, + "loss": 0.8626, + "step": 36160 + }, + { + "epoch": 2.33, + "learning_rate": 1.2261748856780953e-05, + "loss": 1.0462, + "step": 36161 + }, + { + "epoch": 2.33, + "learning_rate": 1.2261534167758003e-05, + "loss": 0.9311, + "step": 36162 + }, + { + "epoch": 2.33, + "learning_rate": 1.2261319478735054e-05, + "loss": 0.8355, + "step": 36163 + }, + { + "epoch": 2.33, + "learning_rate": 1.2261104789712105e-05, + "loss": 1.0523, + "step": 36164 + }, + { + "epoch": 2.33, + "learning_rate": 1.2260890100689154e-05, + "loss": 0.8924, + "step": 36165 + }, + { + "epoch": 2.33, + "learning_rate": 1.2260675411666202e-05, + "loss": 0.8827, + "step": 36166 + }, + { + "epoch": 2.33, + "learning_rate": 1.2260460722643253e-05, + "loss": 0.9826, + "step": 36167 + }, + { + "epoch": 2.33, + "learning_rate": 1.2260246033620302e-05, + "loss": 0.8438, + "step": 36168 + }, + { + "epoch": 2.33, + "learning_rate": 1.2260031344597351e-05, + "loss": 0.9248, + "step": 36169 + }, + { + "epoch": 2.33, + "learning_rate": 1.2259816655574402e-05, + "loss": 0.8744, + "step": 36170 + }, + { + "epoch": 2.33, + "learning_rate": 1.225960196655145e-05, + "loss": 0.8849, + "step": 36171 + }, + { + "epoch": 2.33, + "learning_rate": 1.2259387277528501e-05, + "loss": 0.8692, + "step": 36172 + }, + { + "epoch": 2.33, + "learning_rate": 1.225917258850555e-05, + "loss": 0.8583, + "step": 36173 + }, + { + "epoch": 2.33, + "learning_rate": 1.22589578994826e-05, + "loss": 0.8803, + "step": 36174 + }, + { + "epoch": 2.33, + "learning_rate": 1.225874321045965e-05, + "loss": 0.9468, + "step": 36175 + }, + { + "epoch": 2.33, + "learning_rate": 1.2258528521436699e-05, + "loss": 0.8944, + "step": 36176 + }, + { + "epoch": 2.33, + "learning_rate": 1.2258313832413751e-05, + "loss": 0.8705, + "step": 36177 + }, + { + "epoch": 2.33, + "learning_rate": 1.22580991433908e-05, + "loss": 0.8603, + "step": 36178 + }, + { + "epoch": 2.33, + "learning_rate": 1.225788445436785e-05, + "loss": 0.8914, + "step": 36179 + }, + { + "epoch": 2.33, + "learning_rate": 1.22576697653449e-05, + "loss": 0.8925, + "step": 36180 + }, + { + "epoch": 2.33, + "learning_rate": 1.2257455076321949e-05, + "loss": 0.8541, + "step": 36181 + }, + { + "epoch": 2.33, + "learning_rate": 1.2257240387299e-05, + "loss": 0.8448, + "step": 36182 + }, + { + "epoch": 2.33, + "learning_rate": 1.2257025698276048e-05, + "loss": 0.8648, + "step": 36183 + }, + { + "epoch": 2.33, + "learning_rate": 1.2256811009253097e-05, + "loss": 0.874, + "step": 36184 + }, + { + "epoch": 2.33, + "learning_rate": 1.2256596320230148e-05, + "loss": 0.8667, + "step": 36185 + }, + { + "epoch": 2.33, + "learning_rate": 1.2256381631207197e-05, + "loss": 0.886, + "step": 36186 + }, + { + "epoch": 2.33, + "learning_rate": 1.2256166942184246e-05, + "loss": 0.8991, + "step": 36187 + }, + { + "epoch": 2.33, + "learning_rate": 1.2255952253161297e-05, + "loss": 0.8796, + "step": 36188 + }, + { + "epoch": 2.33, + "learning_rate": 1.2255737564138346e-05, + "loss": 0.8898, + "step": 36189 + }, + { + "epoch": 2.33, + "learning_rate": 1.2255522875115395e-05, + "loss": 0.8706, + "step": 36190 + }, + { + "epoch": 2.33, + "learning_rate": 1.2255308186092445e-05, + "loss": 0.8425, + "step": 36191 + }, + { + "epoch": 2.33, + "learning_rate": 1.2255093497069498e-05, + "loss": 0.9288, + "step": 36192 + }, + { + "epoch": 2.33, + "learning_rate": 1.2254878808046547e-05, + "loss": 0.81, + "step": 36193 + }, + { + "epoch": 2.33, + "learning_rate": 1.2254664119023596e-05, + "loss": 0.9034, + "step": 36194 + }, + { + "epoch": 2.33, + "learning_rate": 1.2254449430000646e-05, + "loss": 0.9008, + "step": 36195 + }, + { + "epoch": 2.33, + "learning_rate": 1.2254234740977695e-05, + "loss": 0.933, + "step": 36196 + }, + { + "epoch": 2.33, + "learning_rate": 1.2254020051954744e-05, + "loss": 0.8509, + "step": 36197 + }, + { + "epoch": 2.33, + "learning_rate": 1.2253805362931795e-05, + "loss": 0.8838, + "step": 36198 + }, + { + "epoch": 2.33, + "learning_rate": 1.2253590673908844e-05, + "loss": 0.8363, + "step": 36199 + }, + { + "epoch": 2.33, + "learning_rate": 1.2253375984885893e-05, + "loss": 0.9133, + "step": 36200 + }, + { + "epoch": 2.33, + "learning_rate": 1.2253161295862943e-05, + "loss": 0.8944, + "step": 36201 + }, + { + "epoch": 2.33, + "learning_rate": 1.2252946606839992e-05, + "loss": 0.8689, + "step": 36202 + }, + { + "epoch": 2.33, + "learning_rate": 1.2252731917817043e-05, + "loss": 0.8369, + "step": 36203 + }, + { + "epoch": 2.33, + "learning_rate": 1.2252517228794092e-05, + "loss": 0.8485, + "step": 36204 + }, + { + "epoch": 2.33, + "learning_rate": 1.2252302539771141e-05, + "loss": 0.9193, + "step": 36205 + }, + { + "epoch": 2.33, + "learning_rate": 1.2252087850748193e-05, + "loss": 0.9207, + "step": 36206 + }, + { + "epoch": 2.33, + "learning_rate": 1.2251873161725242e-05, + "loss": 0.9003, + "step": 36207 + }, + { + "epoch": 2.33, + "learning_rate": 1.2251658472702293e-05, + "loss": 0.9289, + "step": 36208 + }, + { + "epoch": 2.33, + "learning_rate": 1.2251443783679342e-05, + "loss": 0.9161, + "step": 36209 + }, + { + "epoch": 2.33, + "learning_rate": 1.225122909465639e-05, + "loss": 0.8804, + "step": 36210 + }, + { + "epoch": 2.33, + "learning_rate": 1.2251014405633441e-05, + "loss": 0.9116, + "step": 36211 + }, + { + "epoch": 2.33, + "learning_rate": 1.225079971661049e-05, + "loss": 0.8689, + "step": 36212 + }, + { + "epoch": 2.33, + "learning_rate": 1.2250585027587541e-05, + "loss": 0.9301, + "step": 36213 + }, + { + "epoch": 2.33, + "learning_rate": 1.225037033856459e-05, + "loss": 0.9449, + "step": 36214 + }, + { + "epoch": 2.33, + "learning_rate": 1.2250155649541639e-05, + "loss": 0.9476, + "step": 36215 + }, + { + "epoch": 2.33, + "learning_rate": 1.224994096051869e-05, + "loss": 0.8733, + "step": 36216 + }, + { + "epoch": 2.33, + "learning_rate": 1.2249726271495739e-05, + "loss": 0.848, + "step": 36217 + }, + { + "epoch": 2.33, + "learning_rate": 1.2249511582472788e-05, + "loss": 0.9281, + "step": 36218 + }, + { + "epoch": 2.33, + "learning_rate": 1.2249296893449838e-05, + "loss": 0.9361, + "step": 36219 + }, + { + "epoch": 2.33, + "learning_rate": 1.2249082204426887e-05, + "loss": 0.8517, + "step": 36220 + }, + { + "epoch": 2.33, + "learning_rate": 1.224886751540394e-05, + "loss": 0.8812, + "step": 36221 + }, + { + "epoch": 2.33, + "learning_rate": 1.2248652826380989e-05, + "loss": 0.9112, + "step": 36222 + }, + { + "epoch": 2.33, + "learning_rate": 1.224843813735804e-05, + "loss": 0.8597, + "step": 36223 + }, + { + "epoch": 2.33, + "learning_rate": 1.2248223448335088e-05, + "loss": 0.8508, + "step": 36224 + }, + { + "epoch": 2.33, + "learning_rate": 1.2248008759312137e-05, + "loss": 0.8599, + "step": 36225 + }, + { + "epoch": 2.33, + "learning_rate": 1.2247794070289188e-05, + "loss": 0.8891, + "step": 36226 + }, + { + "epoch": 2.33, + "learning_rate": 1.2247579381266237e-05, + "loss": 0.9977, + "step": 36227 + }, + { + "epoch": 2.33, + "learning_rate": 1.2247364692243286e-05, + "loss": 0.9392, + "step": 36228 + }, + { + "epoch": 2.33, + "learning_rate": 1.2247150003220336e-05, + "loss": 0.8607, + "step": 36229 + }, + { + "epoch": 2.33, + "learning_rate": 1.2246935314197385e-05, + "loss": 0.9407, + "step": 36230 + }, + { + "epoch": 2.33, + "learning_rate": 1.2246720625174436e-05, + "loss": 0.8545, + "step": 36231 + }, + { + "epoch": 2.33, + "learning_rate": 1.2246505936151485e-05, + "loss": 0.8469, + "step": 36232 + }, + { + "epoch": 2.33, + "learning_rate": 1.2246291247128534e-05, + "loss": 0.8992, + "step": 36233 + }, + { + "epoch": 2.33, + "learning_rate": 1.2246076558105585e-05, + "loss": 0.9027, + "step": 36234 + }, + { + "epoch": 2.33, + "learning_rate": 1.2245861869082635e-05, + "loss": 0.8258, + "step": 36235 + }, + { + "epoch": 2.33, + "learning_rate": 1.2245647180059686e-05, + "loss": 0.893, + "step": 36236 + }, + { + "epoch": 2.33, + "learning_rate": 1.2245432491036735e-05, + "loss": 0.9074, + "step": 36237 + }, + { + "epoch": 2.33, + "learning_rate": 1.2245217802013784e-05, + "loss": 0.8846, + "step": 36238 + }, + { + "epoch": 2.33, + "learning_rate": 1.2245003112990835e-05, + "loss": 0.9262, + "step": 36239 + }, + { + "epoch": 2.33, + "learning_rate": 1.2244788423967883e-05, + "loss": 0.8334, + "step": 36240 + }, + { + "epoch": 2.33, + "learning_rate": 1.2244573734944934e-05, + "loss": 0.9244, + "step": 36241 + }, + { + "epoch": 2.33, + "learning_rate": 1.2244359045921983e-05, + "loss": 0.8821, + "step": 36242 + }, + { + "epoch": 2.33, + "learning_rate": 1.2244144356899032e-05, + "loss": 0.8499, + "step": 36243 + }, + { + "epoch": 2.33, + "learning_rate": 1.2243929667876083e-05, + "loss": 0.8915, + "step": 36244 + }, + { + "epoch": 2.33, + "learning_rate": 1.2243714978853132e-05, + "loss": 0.8672, + "step": 36245 + }, + { + "epoch": 2.33, + "learning_rate": 1.224350028983018e-05, + "loss": 0.8775, + "step": 36246 + }, + { + "epoch": 2.33, + "learning_rate": 1.2243285600807231e-05, + "loss": 0.9205, + "step": 36247 + }, + { + "epoch": 2.33, + "learning_rate": 1.224307091178428e-05, + "loss": 0.9339, + "step": 36248 + }, + { + "epoch": 2.33, + "learning_rate": 1.2242856222761331e-05, + "loss": 0.8994, + "step": 36249 + }, + { + "epoch": 2.33, + "learning_rate": 1.2242641533738382e-05, + "loss": 0.8591, + "step": 36250 + }, + { + "epoch": 2.33, + "learning_rate": 1.2242426844715432e-05, + "loss": 0.9708, + "step": 36251 + }, + { + "epoch": 2.33, + "learning_rate": 1.2242212155692481e-05, + "loss": 0.8935, + "step": 36252 + }, + { + "epoch": 2.33, + "learning_rate": 1.224199746666953e-05, + "loss": 0.8761, + "step": 36253 + }, + { + "epoch": 2.33, + "learning_rate": 1.224178277764658e-05, + "loss": 0.8923, + "step": 36254 + }, + { + "epoch": 2.33, + "learning_rate": 1.224156808862363e-05, + "loss": 0.9372, + "step": 36255 + }, + { + "epoch": 2.33, + "learning_rate": 1.2241353399600679e-05, + "loss": 0.8999, + "step": 36256 + }, + { + "epoch": 2.33, + "learning_rate": 1.224113871057773e-05, + "loss": 0.8604, + "step": 36257 + }, + { + "epoch": 2.33, + "learning_rate": 1.2240924021554778e-05, + "loss": 0.9551, + "step": 36258 + }, + { + "epoch": 2.33, + "learning_rate": 1.2240709332531829e-05, + "loss": 0.8786, + "step": 36259 + }, + { + "epoch": 2.33, + "learning_rate": 1.2240494643508878e-05, + "loss": 0.9006, + "step": 36260 + }, + { + "epoch": 2.33, + "learning_rate": 1.2240279954485927e-05, + "loss": 0.8755, + "step": 36261 + }, + { + "epoch": 2.33, + "learning_rate": 1.2240065265462978e-05, + "loss": 0.8566, + "step": 36262 + }, + { + "epoch": 2.33, + "learning_rate": 1.2239850576440027e-05, + "loss": 0.8977, + "step": 36263 + }, + { + "epoch": 2.33, + "learning_rate": 1.2239635887417079e-05, + "loss": 0.9747, + "step": 36264 + }, + { + "epoch": 2.33, + "learning_rate": 1.2239421198394128e-05, + "loss": 0.9102, + "step": 36265 + }, + { + "epoch": 2.33, + "learning_rate": 1.2239206509371177e-05, + "loss": 0.9533, + "step": 36266 + }, + { + "epoch": 2.33, + "learning_rate": 1.2238991820348228e-05, + "loss": 0.9837, + "step": 36267 + }, + { + "epoch": 2.33, + "learning_rate": 1.2238777131325277e-05, + "loss": 0.8405, + "step": 36268 + }, + { + "epoch": 2.33, + "learning_rate": 1.2238562442302327e-05, + "loss": 0.9603, + "step": 36269 + }, + { + "epoch": 2.33, + "learning_rate": 1.2238347753279376e-05, + "loss": 0.8415, + "step": 36270 + }, + { + "epoch": 2.33, + "learning_rate": 1.2238133064256425e-05, + "loss": 0.9246, + "step": 36271 + }, + { + "epoch": 2.33, + "learning_rate": 1.2237918375233476e-05, + "loss": 0.9141, + "step": 36272 + }, + { + "epoch": 2.33, + "learning_rate": 1.2237703686210525e-05, + "loss": 0.8853, + "step": 36273 + }, + { + "epoch": 2.33, + "learning_rate": 1.2237488997187574e-05, + "loss": 0.7871, + "step": 36274 + }, + { + "epoch": 2.33, + "learning_rate": 1.2237274308164624e-05, + "loss": 0.9804, + "step": 36275 + }, + { + "epoch": 2.33, + "learning_rate": 1.2237059619141673e-05, + "loss": 0.9214, + "step": 36276 + }, + { + "epoch": 2.33, + "learning_rate": 1.2236844930118722e-05, + "loss": 0.9553, + "step": 36277 + }, + { + "epoch": 2.33, + "learning_rate": 1.2236630241095773e-05, + "loss": 0.8393, + "step": 36278 + }, + { + "epoch": 2.33, + "learning_rate": 1.2236415552072825e-05, + "loss": 0.9719, + "step": 36279 + }, + { + "epoch": 2.33, + "learning_rate": 1.2236200863049874e-05, + "loss": 0.9249, + "step": 36280 + }, + { + "epoch": 2.33, + "learning_rate": 1.2235986174026923e-05, + "loss": 0.8537, + "step": 36281 + }, + { + "epoch": 2.33, + "learning_rate": 1.2235771485003974e-05, + "loss": 0.8786, + "step": 36282 + }, + { + "epoch": 2.33, + "learning_rate": 1.2235556795981023e-05, + "loss": 0.8285, + "step": 36283 + }, + { + "epoch": 2.33, + "learning_rate": 1.2235342106958072e-05, + "loss": 0.921, + "step": 36284 + }, + { + "epoch": 2.33, + "learning_rate": 1.2235127417935122e-05, + "loss": 0.8471, + "step": 36285 + }, + { + "epoch": 2.33, + "learning_rate": 1.2234912728912171e-05, + "loss": 0.8997, + "step": 36286 + }, + { + "epoch": 2.33, + "learning_rate": 1.223469803988922e-05, + "loss": 0.8918, + "step": 36287 + }, + { + "epoch": 2.33, + "learning_rate": 1.2234483350866271e-05, + "loss": 0.8653, + "step": 36288 + }, + { + "epoch": 2.33, + "learning_rate": 1.223426866184332e-05, + "loss": 0.8795, + "step": 36289 + }, + { + "epoch": 2.33, + "learning_rate": 1.223405397282037e-05, + "loss": 0.9287, + "step": 36290 + }, + { + "epoch": 2.33, + "learning_rate": 1.223383928379742e-05, + "loss": 0.884, + "step": 36291 + }, + { + "epoch": 2.33, + "learning_rate": 1.2233624594774469e-05, + "loss": 0.8136, + "step": 36292 + }, + { + "epoch": 2.34, + "learning_rate": 1.2233409905751521e-05, + "loss": 0.9078, + "step": 36293 + }, + { + "epoch": 2.34, + "learning_rate": 1.223319521672857e-05, + "loss": 0.8981, + "step": 36294 + }, + { + "epoch": 2.34, + "learning_rate": 1.223298052770562e-05, + "loss": 0.9035, + "step": 36295 + }, + { + "epoch": 2.34, + "learning_rate": 1.223276583868267e-05, + "loss": 0.8978, + "step": 36296 + }, + { + "epoch": 2.34, + "learning_rate": 1.2232551149659719e-05, + "loss": 0.9333, + "step": 36297 + }, + { + "epoch": 2.34, + "learning_rate": 1.223233646063677e-05, + "loss": 0.8435, + "step": 36298 + }, + { + "epoch": 2.34, + "learning_rate": 1.2232121771613818e-05, + "loss": 0.9423, + "step": 36299 + }, + { + "epoch": 2.34, + "learning_rate": 1.2231907082590869e-05, + "loss": 0.8787, + "step": 36300 + }, + { + "epoch": 2.34, + "learning_rate": 1.2231692393567918e-05, + "loss": 0.9394, + "step": 36301 + }, + { + "epoch": 2.34, + "learning_rate": 1.2231477704544967e-05, + "loss": 0.9251, + "step": 36302 + }, + { + "epoch": 2.34, + "learning_rate": 1.2231263015522017e-05, + "loss": 0.9003, + "step": 36303 + }, + { + "epoch": 2.34, + "learning_rate": 1.2231048326499066e-05, + "loss": 0.9078, + "step": 36304 + }, + { + "epoch": 2.34, + "learning_rate": 1.2230833637476115e-05, + "loss": 0.9248, + "step": 36305 + }, + { + "epoch": 2.34, + "learning_rate": 1.2230618948453166e-05, + "loss": 0.8918, + "step": 36306 + }, + { + "epoch": 2.34, + "learning_rate": 1.2230404259430215e-05, + "loss": 0.861, + "step": 36307 + }, + { + "epoch": 2.34, + "learning_rate": 1.2230189570407267e-05, + "loss": 0.8306, + "step": 36308 + }, + { + "epoch": 2.34, + "learning_rate": 1.2229974881384316e-05, + "loss": 0.8635, + "step": 36309 + }, + { + "epoch": 2.34, + "learning_rate": 1.2229760192361367e-05, + "loss": 0.8762, + "step": 36310 + }, + { + "epoch": 2.34, + "learning_rate": 1.2229545503338416e-05, + "loss": 0.9161, + "step": 36311 + }, + { + "epoch": 2.34, + "learning_rate": 1.2229330814315465e-05, + "loss": 0.9137, + "step": 36312 + }, + { + "epoch": 2.34, + "learning_rate": 1.2229116125292516e-05, + "loss": 0.8655, + "step": 36313 + }, + { + "epoch": 2.34, + "learning_rate": 1.2228901436269564e-05, + "loss": 0.884, + "step": 36314 + }, + { + "epoch": 2.34, + "learning_rate": 1.2228686747246613e-05, + "loss": 0.8745, + "step": 36315 + }, + { + "epoch": 2.34, + "learning_rate": 1.2228472058223664e-05, + "loss": 0.9404, + "step": 36316 + }, + { + "epoch": 2.34, + "learning_rate": 1.2228257369200713e-05, + "loss": 0.912, + "step": 36317 + }, + { + "epoch": 2.34, + "learning_rate": 1.2228042680177764e-05, + "loss": 0.8829, + "step": 36318 + }, + { + "epoch": 2.34, + "learning_rate": 1.2227827991154813e-05, + "loss": 0.9196, + "step": 36319 + }, + { + "epoch": 2.34, + "learning_rate": 1.2227613302131862e-05, + "loss": 0.9375, + "step": 36320 + }, + { + "epoch": 2.34, + "learning_rate": 1.2227398613108912e-05, + "loss": 0.9242, + "step": 36321 + }, + { + "epoch": 2.34, + "learning_rate": 1.2227183924085963e-05, + "loss": 0.9143, + "step": 36322 + }, + { + "epoch": 2.34, + "learning_rate": 1.2226969235063014e-05, + "loss": 0.9015, + "step": 36323 + }, + { + "epoch": 2.34, + "learning_rate": 1.2226754546040063e-05, + "loss": 0.8141, + "step": 36324 + }, + { + "epoch": 2.34, + "learning_rate": 1.2226539857017112e-05, + "loss": 0.8893, + "step": 36325 + }, + { + "epoch": 2.34, + "learning_rate": 1.2226325167994162e-05, + "loss": 0.8564, + "step": 36326 + }, + { + "epoch": 2.34, + "learning_rate": 1.2226110478971211e-05, + "loss": 0.9069, + "step": 36327 + }, + { + "epoch": 2.34, + "learning_rate": 1.2225895789948262e-05, + "loss": 0.8644, + "step": 36328 + }, + { + "epoch": 2.34, + "learning_rate": 1.222568110092531e-05, + "loss": 0.8732, + "step": 36329 + }, + { + "epoch": 2.34, + "learning_rate": 1.222546641190236e-05, + "loss": 0.8545, + "step": 36330 + }, + { + "epoch": 2.34, + "learning_rate": 1.222525172287941e-05, + "loss": 0.8764, + "step": 36331 + }, + { + "epoch": 2.34, + "learning_rate": 1.222503703385646e-05, + "loss": 0.9069, + "step": 36332 + }, + { + "epoch": 2.34, + "learning_rate": 1.2224822344833508e-05, + "loss": 0.8861, + "step": 36333 + }, + { + "epoch": 2.34, + "learning_rate": 1.2224607655810559e-05, + "loss": 0.9021, + "step": 36334 + }, + { + "epoch": 2.34, + "learning_rate": 1.2224392966787608e-05, + "loss": 0.9302, + "step": 36335 + }, + { + "epoch": 2.34, + "learning_rate": 1.2224178277764659e-05, + "loss": 0.8956, + "step": 36336 + }, + { + "epoch": 2.34, + "learning_rate": 1.222396358874171e-05, + "loss": 0.9497, + "step": 36337 + }, + { + "epoch": 2.34, + "learning_rate": 1.222374889971876e-05, + "loss": 0.9568, + "step": 36338 + }, + { + "epoch": 2.34, + "learning_rate": 1.2223534210695809e-05, + "loss": 0.9183, + "step": 36339 + }, + { + "epoch": 2.34, + "learning_rate": 1.2223319521672858e-05, + "loss": 0.833, + "step": 36340 + }, + { + "epoch": 2.34, + "learning_rate": 1.2223104832649909e-05, + "loss": 0.8161, + "step": 36341 + }, + { + "epoch": 2.34, + "learning_rate": 1.2222890143626958e-05, + "loss": 0.9001, + "step": 36342 + }, + { + "epoch": 2.34, + "learning_rate": 1.2222675454604006e-05, + "loss": 0.8896, + "step": 36343 + }, + { + "epoch": 2.34, + "learning_rate": 1.2222460765581057e-05, + "loss": 0.9219, + "step": 36344 + }, + { + "epoch": 2.34, + "learning_rate": 1.2222246076558106e-05, + "loss": 0.9281, + "step": 36345 + }, + { + "epoch": 2.34, + "learning_rate": 1.2222031387535157e-05, + "loss": 0.9665, + "step": 36346 + }, + { + "epoch": 2.34, + "learning_rate": 1.2221816698512206e-05, + "loss": 0.9027, + "step": 36347 + }, + { + "epoch": 2.34, + "learning_rate": 1.2221602009489255e-05, + "loss": 0.8746, + "step": 36348 + }, + { + "epoch": 2.34, + "learning_rate": 1.2221387320466305e-05, + "loss": 0.8575, + "step": 36349 + }, + { + "epoch": 2.34, + "learning_rate": 1.2221172631443354e-05, + "loss": 0.9232, + "step": 36350 + }, + { + "epoch": 2.34, + "learning_rate": 1.2220957942420407e-05, + "loss": 0.8687, + "step": 36351 + }, + { + "epoch": 2.34, + "learning_rate": 1.2220743253397456e-05, + "loss": 0.882, + "step": 36352 + }, + { + "epoch": 2.34, + "learning_rate": 1.2220528564374505e-05, + "loss": 0.8603, + "step": 36353 + }, + { + "epoch": 2.34, + "learning_rate": 1.2220313875351555e-05, + "loss": 0.8955, + "step": 36354 + }, + { + "epoch": 2.34, + "learning_rate": 1.2220099186328604e-05, + "loss": 0.9164, + "step": 36355 + }, + { + "epoch": 2.34, + "learning_rate": 1.2219884497305655e-05, + "loss": 0.9417, + "step": 36356 + }, + { + "epoch": 2.34, + "learning_rate": 1.2219669808282704e-05, + "loss": 0.8195, + "step": 36357 + }, + { + "epoch": 2.34, + "learning_rate": 1.2219455119259753e-05, + "loss": 0.9172, + "step": 36358 + }, + { + "epoch": 2.34, + "learning_rate": 1.2219240430236803e-05, + "loss": 0.9073, + "step": 36359 + }, + { + "epoch": 2.34, + "learning_rate": 1.2219025741213852e-05, + "loss": 0.8376, + "step": 36360 + }, + { + "epoch": 2.34, + "learning_rate": 1.2218811052190901e-05, + "loss": 0.8753, + "step": 36361 + }, + { + "epoch": 2.34, + "learning_rate": 1.2218596363167952e-05, + "loss": 0.8843, + "step": 36362 + }, + { + "epoch": 2.34, + "learning_rate": 1.2218381674145001e-05, + "loss": 0.836, + "step": 36363 + }, + { + "epoch": 2.34, + "learning_rate": 1.221816698512205e-05, + "loss": 0.956, + "step": 36364 + }, + { + "epoch": 2.34, + "learning_rate": 1.22179522960991e-05, + "loss": 0.8439, + "step": 36365 + }, + { + "epoch": 2.34, + "learning_rate": 1.2217737607076153e-05, + "loss": 0.9126, + "step": 36366 + }, + { + "epoch": 2.34, + "learning_rate": 1.2217522918053202e-05, + "loss": 0.8783, + "step": 36367 + }, + { + "epoch": 2.34, + "learning_rate": 1.2217308229030251e-05, + "loss": 0.8976, + "step": 36368 + }, + { + "epoch": 2.34, + "learning_rate": 1.2217093540007302e-05, + "loss": 0.9029, + "step": 36369 + }, + { + "epoch": 2.34, + "learning_rate": 1.221687885098435e-05, + "loss": 0.9073, + "step": 36370 + }, + { + "epoch": 2.34, + "learning_rate": 1.22166641619614e-05, + "loss": 0.8905, + "step": 36371 + }, + { + "epoch": 2.34, + "learning_rate": 1.221644947293845e-05, + "loss": 0.9498, + "step": 36372 + }, + { + "epoch": 2.34, + "learning_rate": 1.2216234783915499e-05, + "loss": 0.9055, + "step": 36373 + }, + { + "epoch": 2.34, + "learning_rate": 1.2216020094892548e-05, + "loss": 0.871, + "step": 36374 + }, + { + "epoch": 2.34, + "learning_rate": 1.2215805405869599e-05, + "loss": 0.8829, + "step": 36375 + }, + { + "epoch": 2.34, + "learning_rate": 1.2215590716846648e-05, + "loss": 0.8833, + "step": 36376 + }, + { + "epoch": 2.34, + "learning_rate": 1.2215376027823698e-05, + "loss": 0.8214, + "step": 36377 + }, + { + "epoch": 2.34, + "learning_rate": 1.2215161338800747e-05, + "loss": 0.8974, + "step": 36378 + }, + { + "epoch": 2.34, + "learning_rate": 1.2214946649777796e-05, + "loss": 0.9071, + "step": 36379 + }, + { + "epoch": 2.34, + "learning_rate": 1.2214731960754849e-05, + "loss": 0.9301, + "step": 36380 + }, + { + "epoch": 2.34, + "learning_rate": 1.2214517271731898e-05, + "loss": 1.0301, + "step": 36381 + }, + { + "epoch": 2.34, + "learning_rate": 1.2214302582708948e-05, + "loss": 0.8956, + "step": 36382 + }, + { + "epoch": 2.34, + "learning_rate": 1.2214087893685997e-05, + "loss": 0.894, + "step": 36383 + }, + { + "epoch": 2.34, + "learning_rate": 1.2213873204663046e-05, + "loss": 0.8916, + "step": 36384 + }, + { + "epoch": 2.34, + "learning_rate": 1.2213658515640097e-05, + "loss": 0.8986, + "step": 36385 + }, + { + "epoch": 2.34, + "learning_rate": 1.2213443826617146e-05, + "loss": 0.8201, + "step": 36386 + }, + { + "epoch": 2.34, + "learning_rate": 1.2213229137594197e-05, + "loss": 0.9009, + "step": 36387 + }, + { + "epoch": 2.34, + "learning_rate": 1.2213014448571245e-05, + "loss": 0.9595, + "step": 36388 + }, + { + "epoch": 2.34, + "learning_rate": 1.2212799759548294e-05, + "loss": 0.8429, + "step": 36389 + }, + { + "epoch": 2.34, + "learning_rate": 1.2212585070525345e-05, + "loss": 0.86, + "step": 36390 + }, + { + "epoch": 2.34, + "learning_rate": 1.2212370381502394e-05, + "loss": 0.8556, + "step": 36391 + }, + { + "epoch": 2.34, + "learning_rate": 1.2212155692479443e-05, + "loss": 0.8658, + "step": 36392 + }, + { + "epoch": 2.34, + "learning_rate": 1.2211941003456494e-05, + "loss": 0.8654, + "step": 36393 + }, + { + "epoch": 2.34, + "learning_rate": 1.2211726314433543e-05, + "loss": 0.8942, + "step": 36394 + }, + { + "epoch": 2.34, + "learning_rate": 1.2211511625410595e-05, + "loss": 0.8953, + "step": 36395 + }, + { + "epoch": 2.34, + "learning_rate": 1.2211296936387644e-05, + "loss": 0.8393, + "step": 36396 + }, + { + "epoch": 2.34, + "learning_rate": 1.2211082247364695e-05, + "loss": 0.8394, + "step": 36397 + }, + { + "epoch": 2.34, + "learning_rate": 1.2210867558341744e-05, + "loss": 0.9204, + "step": 36398 + }, + { + "epoch": 2.34, + "learning_rate": 1.2210652869318793e-05, + "loss": 0.8705, + "step": 36399 + }, + { + "epoch": 2.34, + "learning_rate": 1.2210438180295843e-05, + "loss": 0.8731, + "step": 36400 + }, + { + "epoch": 2.34, + "learning_rate": 1.2210223491272892e-05, + "loss": 0.9063, + "step": 36401 + }, + { + "epoch": 2.34, + "learning_rate": 1.2210008802249941e-05, + "loss": 0.9338, + "step": 36402 + }, + { + "epoch": 2.34, + "learning_rate": 1.2209794113226992e-05, + "loss": 0.8392, + "step": 36403 + }, + { + "epoch": 2.34, + "learning_rate": 1.220957942420404e-05, + "loss": 0.8385, + "step": 36404 + }, + { + "epoch": 2.34, + "learning_rate": 1.2209364735181091e-05, + "loss": 0.9112, + "step": 36405 + }, + { + "epoch": 2.34, + "learning_rate": 1.220915004615814e-05, + "loss": 0.9388, + "step": 36406 + }, + { + "epoch": 2.34, + "learning_rate": 1.220893535713519e-05, + "loss": 0.8983, + "step": 36407 + }, + { + "epoch": 2.34, + "learning_rate": 1.220872066811224e-05, + "loss": 0.9479, + "step": 36408 + }, + { + "epoch": 2.34, + "learning_rate": 1.220850597908929e-05, + "loss": 0.9153, + "step": 36409 + }, + { + "epoch": 2.34, + "learning_rate": 1.2208291290066341e-05, + "loss": 0.9539, + "step": 36410 + }, + { + "epoch": 2.34, + "learning_rate": 1.220807660104339e-05, + "loss": 0.8603, + "step": 36411 + }, + { + "epoch": 2.34, + "learning_rate": 1.220786191202044e-05, + "loss": 0.9256, + "step": 36412 + }, + { + "epoch": 2.34, + "learning_rate": 1.220764722299749e-05, + "loss": 0.8311, + "step": 36413 + }, + { + "epoch": 2.34, + "learning_rate": 1.2207432533974539e-05, + "loss": 0.9077, + "step": 36414 + }, + { + "epoch": 2.34, + "learning_rate": 1.220721784495159e-05, + "loss": 0.8452, + "step": 36415 + }, + { + "epoch": 2.34, + "learning_rate": 1.2207003155928639e-05, + "loss": 0.8624, + "step": 36416 + }, + { + "epoch": 2.34, + "learning_rate": 1.2206788466905687e-05, + "loss": 0.9163, + "step": 36417 + }, + { + "epoch": 2.34, + "learning_rate": 1.2206573777882738e-05, + "loss": 0.8776, + "step": 36418 + }, + { + "epoch": 2.34, + "learning_rate": 1.2206359088859787e-05, + "loss": 0.8133, + "step": 36419 + }, + { + "epoch": 2.34, + "learning_rate": 1.2206144399836836e-05, + "loss": 0.8829, + "step": 36420 + }, + { + "epoch": 2.34, + "learning_rate": 1.2205929710813887e-05, + "loss": 0.993, + "step": 36421 + }, + { + "epoch": 2.34, + "learning_rate": 1.2205715021790936e-05, + "loss": 0.8591, + "step": 36422 + }, + { + "epoch": 2.34, + "learning_rate": 1.2205500332767986e-05, + "loss": 0.9743, + "step": 36423 + }, + { + "epoch": 2.34, + "learning_rate": 1.2205285643745037e-05, + "loss": 0.8655, + "step": 36424 + }, + { + "epoch": 2.34, + "learning_rate": 1.2205070954722088e-05, + "loss": 0.8627, + "step": 36425 + }, + { + "epoch": 2.34, + "learning_rate": 1.2204856265699137e-05, + "loss": 0.8391, + "step": 36426 + }, + { + "epoch": 2.34, + "learning_rate": 1.2204641576676186e-05, + "loss": 0.9019, + "step": 36427 + }, + { + "epoch": 2.34, + "learning_rate": 1.2204426887653236e-05, + "loss": 0.9324, + "step": 36428 + }, + { + "epoch": 2.34, + "learning_rate": 1.2204212198630285e-05, + "loss": 0.9211, + "step": 36429 + }, + { + "epoch": 2.34, + "learning_rate": 1.2203997509607334e-05, + "loss": 0.9357, + "step": 36430 + }, + { + "epoch": 2.34, + "learning_rate": 1.2203782820584385e-05, + "loss": 0.8248, + "step": 36431 + }, + { + "epoch": 2.34, + "learning_rate": 1.2203568131561434e-05, + "loss": 0.8844, + "step": 36432 + }, + { + "epoch": 2.34, + "learning_rate": 1.2203353442538484e-05, + "loss": 0.8452, + "step": 36433 + }, + { + "epoch": 2.34, + "learning_rate": 1.2203138753515533e-05, + "loss": 0.8393, + "step": 36434 + }, + { + "epoch": 2.34, + "learning_rate": 1.2202924064492582e-05, + "loss": 0.9394, + "step": 36435 + }, + { + "epoch": 2.34, + "learning_rate": 1.2202709375469633e-05, + "loss": 0.8037, + "step": 36436 + }, + { + "epoch": 2.34, + "learning_rate": 1.2202494686446682e-05, + "loss": 0.8836, + "step": 36437 + }, + { + "epoch": 2.34, + "learning_rate": 1.2202279997423734e-05, + "loss": 0.8932, + "step": 36438 + }, + { + "epoch": 2.34, + "learning_rate": 1.2202065308400783e-05, + "loss": 0.9141, + "step": 36439 + }, + { + "epoch": 2.34, + "learning_rate": 1.2201850619377832e-05, + "loss": 0.8352, + "step": 36440 + }, + { + "epoch": 2.34, + "learning_rate": 1.2201635930354883e-05, + "loss": 0.8847, + "step": 36441 + }, + { + "epoch": 2.34, + "learning_rate": 1.2201421241331932e-05, + "loss": 0.895, + "step": 36442 + }, + { + "epoch": 2.34, + "learning_rate": 1.2201206552308983e-05, + "loss": 0.8439, + "step": 36443 + }, + { + "epoch": 2.34, + "learning_rate": 1.2200991863286032e-05, + "loss": 0.9342, + "step": 36444 + }, + { + "epoch": 2.34, + "learning_rate": 1.220077717426308e-05, + "loss": 0.9426, + "step": 36445 + }, + { + "epoch": 2.34, + "learning_rate": 1.2200562485240131e-05, + "loss": 0.9264, + "step": 36446 + }, + { + "epoch": 2.34, + "learning_rate": 1.220034779621718e-05, + "loss": 0.8548, + "step": 36447 + }, + { + "epoch": 2.34, + "learning_rate": 1.2200133107194229e-05, + "loss": 1.0514, + "step": 36448 + }, + { + "epoch": 2.35, + "learning_rate": 1.219991841817128e-05, + "loss": 0.8955, + "step": 36449 + }, + { + "epoch": 2.35, + "learning_rate": 1.2199703729148329e-05, + "loss": 0.9211, + "step": 36450 + }, + { + "epoch": 2.35, + "learning_rate": 1.2199489040125378e-05, + "loss": 0.8823, + "step": 36451 + }, + { + "epoch": 2.35, + "learning_rate": 1.2199274351102428e-05, + "loss": 0.9698, + "step": 36452 + }, + { + "epoch": 2.35, + "learning_rate": 1.219905966207948e-05, + "loss": 0.8838, + "step": 36453 + }, + { + "epoch": 2.35, + "learning_rate": 1.219884497305653e-05, + "loss": 0.816, + "step": 36454 + }, + { + "epoch": 2.35, + "learning_rate": 1.2198630284033579e-05, + "loss": 0.8116, + "step": 36455 + }, + { + "epoch": 2.35, + "learning_rate": 1.219841559501063e-05, + "loss": 0.8731, + "step": 36456 + }, + { + "epoch": 2.35, + "learning_rate": 1.2198200905987678e-05, + "loss": 0.8569, + "step": 36457 + }, + { + "epoch": 2.35, + "learning_rate": 1.2197986216964727e-05, + "loss": 0.8936, + "step": 36458 + }, + { + "epoch": 2.35, + "learning_rate": 1.2197771527941778e-05, + "loss": 0.9139, + "step": 36459 + }, + { + "epoch": 2.35, + "learning_rate": 1.2197556838918827e-05, + "loss": 0.869, + "step": 36460 + }, + { + "epoch": 2.35, + "learning_rate": 1.2197342149895876e-05, + "loss": 0.8182, + "step": 36461 + }, + { + "epoch": 2.35, + "learning_rate": 1.2197127460872926e-05, + "loss": 0.9378, + "step": 36462 + }, + { + "epoch": 2.35, + "learning_rate": 1.2196912771849975e-05, + "loss": 0.928, + "step": 36463 + }, + { + "epoch": 2.35, + "learning_rate": 1.2196698082827026e-05, + "loss": 0.9108, + "step": 36464 + }, + { + "epoch": 2.35, + "learning_rate": 1.2196483393804075e-05, + "loss": 0.9282, + "step": 36465 + }, + { + "epoch": 2.35, + "learning_rate": 1.2196268704781124e-05, + "loss": 0.9333, + "step": 36466 + }, + { + "epoch": 2.35, + "learning_rate": 1.2196054015758176e-05, + "loss": 0.8336, + "step": 36467 + }, + { + "epoch": 2.35, + "learning_rate": 1.2195839326735225e-05, + "loss": 0.9187, + "step": 36468 + }, + { + "epoch": 2.35, + "learning_rate": 1.2195624637712276e-05, + "loss": 0.8801, + "step": 36469 + }, + { + "epoch": 2.35, + "learning_rate": 1.2195409948689325e-05, + "loss": 0.9183, + "step": 36470 + }, + { + "epoch": 2.35, + "learning_rate": 1.2195195259666374e-05, + "loss": 0.9365, + "step": 36471 + }, + { + "epoch": 2.35, + "learning_rate": 1.2194980570643425e-05, + "loss": 0.9724, + "step": 36472 + }, + { + "epoch": 2.35, + "learning_rate": 1.2194765881620474e-05, + "loss": 0.8679, + "step": 36473 + }, + { + "epoch": 2.35, + "learning_rate": 1.2194551192597524e-05, + "loss": 0.8534, + "step": 36474 + }, + { + "epoch": 2.35, + "learning_rate": 1.2194336503574573e-05, + "loss": 0.8916, + "step": 36475 + }, + { + "epoch": 2.35, + "learning_rate": 1.2194121814551622e-05, + "loss": 0.9391, + "step": 36476 + }, + { + "epoch": 2.35, + "learning_rate": 1.2193907125528673e-05, + "loss": 0.9552, + "step": 36477 + }, + { + "epoch": 2.35, + "learning_rate": 1.2193692436505722e-05, + "loss": 0.9153, + "step": 36478 + }, + { + "epoch": 2.35, + "learning_rate": 1.219347774748277e-05, + "loss": 0.8127, + "step": 36479 + }, + { + "epoch": 2.35, + "learning_rate": 1.2193263058459821e-05, + "loss": 0.9226, + "step": 36480 + }, + { + "epoch": 2.35, + "learning_rate": 1.219304836943687e-05, + "loss": 0.8769, + "step": 36481 + }, + { + "epoch": 2.35, + "learning_rate": 1.2192833680413923e-05, + "loss": 0.9084, + "step": 36482 + }, + { + "epoch": 2.35, + "learning_rate": 1.2192618991390972e-05, + "loss": 0.8987, + "step": 36483 + }, + { + "epoch": 2.35, + "learning_rate": 1.2192404302368022e-05, + "loss": 0.828, + "step": 36484 + }, + { + "epoch": 2.35, + "learning_rate": 1.2192189613345071e-05, + "loss": 0.9226, + "step": 36485 + }, + { + "epoch": 2.35, + "learning_rate": 1.219197492432212e-05, + "loss": 0.8897, + "step": 36486 + }, + { + "epoch": 2.35, + "learning_rate": 1.2191760235299171e-05, + "loss": 0.9007, + "step": 36487 + }, + { + "epoch": 2.35, + "learning_rate": 1.219154554627622e-05, + "loss": 0.9015, + "step": 36488 + }, + { + "epoch": 2.35, + "learning_rate": 1.2191330857253269e-05, + "loss": 0.9861, + "step": 36489 + }, + { + "epoch": 2.35, + "learning_rate": 1.219111616823032e-05, + "loss": 0.9006, + "step": 36490 + }, + { + "epoch": 2.35, + "learning_rate": 1.2190901479207368e-05, + "loss": 0.916, + "step": 36491 + }, + { + "epoch": 2.35, + "learning_rate": 1.2190686790184419e-05, + "loss": 0.9067, + "step": 36492 + }, + { + "epoch": 2.35, + "learning_rate": 1.2190472101161468e-05, + "loss": 0.9558, + "step": 36493 + }, + { + "epoch": 2.35, + "learning_rate": 1.2190257412138517e-05, + "loss": 0.8769, + "step": 36494 + }, + { + "epoch": 2.35, + "learning_rate": 1.2190042723115568e-05, + "loss": 0.8924, + "step": 36495 + }, + { + "epoch": 2.35, + "learning_rate": 1.2189828034092618e-05, + "loss": 0.9225, + "step": 36496 + }, + { + "epoch": 2.35, + "learning_rate": 1.2189613345069669e-05, + "loss": 0.9192, + "step": 36497 + }, + { + "epoch": 2.35, + "learning_rate": 1.2189398656046718e-05, + "loss": 0.8919, + "step": 36498 + }, + { + "epoch": 2.35, + "learning_rate": 1.2189183967023767e-05, + "loss": 0.7995, + "step": 36499 + }, + { + "epoch": 2.35, + "learning_rate": 1.2188969278000818e-05, + "loss": 0.9531, + "step": 36500 + }, + { + "epoch": 2.35, + "learning_rate": 1.2188754588977867e-05, + "loss": 0.9011, + "step": 36501 + }, + { + "epoch": 2.35, + "learning_rate": 1.2188539899954917e-05, + "loss": 0.8416, + "step": 36502 + }, + { + "epoch": 2.35, + "learning_rate": 1.2188325210931966e-05, + "loss": 0.9135, + "step": 36503 + }, + { + "epoch": 2.35, + "learning_rate": 1.2188110521909015e-05, + "loss": 0.9208, + "step": 36504 + }, + { + "epoch": 2.35, + "learning_rate": 1.2187895832886066e-05, + "loss": 0.947, + "step": 36505 + }, + { + "epoch": 2.35, + "learning_rate": 1.2187681143863115e-05, + "loss": 0.922, + "step": 36506 + }, + { + "epoch": 2.35, + "learning_rate": 1.2187466454840164e-05, + "loss": 0.8872, + "step": 36507 + }, + { + "epoch": 2.35, + "learning_rate": 1.2187251765817214e-05, + "loss": 0.8846, + "step": 36508 + }, + { + "epoch": 2.35, + "learning_rate": 1.2187037076794263e-05, + "loss": 0.9214, + "step": 36509 + }, + { + "epoch": 2.35, + "learning_rate": 1.2186822387771312e-05, + "loss": 0.8463, + "step": 36510 + }, + { + "epoch": 2.35, + "learning_rate": 1.2186607698748365e-05, + "loss": 0.9182, + "step": 36511 + }, + { + "epoch": 2.35, + "learning_rate": 1.2186393009725415e-05, + "loss": 0.9127, + "step": 36512 + }, + { + "epoch": 2.35, + "learning_rate": 1.2186178320702464e-05, + "loss": 0.9469, + "step": 36513 + }, + { + "epoch": 2.35, + "learning_rate": 1.2185963631679513e-05, + "loss": 0.8716, + "step": 36514 + }, + { + "epoch": 2.35, + "learning_rate": 1.2185748942656564e-05, + "loss": 0.8977, + "step": 36515 + }, + { + "epoch": 2.35, + "learning_rate": 1.2185534253633613e-05, + "loss": 0.8987, + "step": 36516 + }, + { + "epoch": 2.35, + "learning_rate": 1.2185319564610662e-05, + "loss": 0.9325, + "step": 36517 + }, + { + "epoch": 2.35, + "learning_rate": 1.2185104875587713e-05, + "loss": 0.886, + "step": 36518 + }, + { + "epoch": 2.35, + "learning_rate": 1.2184890186564762e-05, + "loss": 0.9111, + "step": 36519 + }, + { + "epoch": 2.35, + "learning_rate": 1.218467549754181e-05, + "loss": 0.9413, + "step": 36520 + }, + { + "epoch": 2.35, + "learning_rate": 1.2184460808518861e-05, + "loss": 0.8909, + "step": 36521 + }, + { + "epoch": 2.35, + "learning_rate": 1.218424611949591e-05, + "loss": 0.8864, + "step": 36522 + }, + { + "epoch": 2.35, + "learning_rate": 1.218403143047296e-05, + "loss": 0.9678, + "step": 36523 + }, + { + "epoch": 2.35, + "learning_rate": 1.218381674145001e-05, + "loss": 0.8781, + "step": 36524 + }, + { + "epoch": 2.35, + "learning_rate": 1.2183602052427062e-05, + "loss": 0.956, + "step": 36525 + }, + { + "epoch": 2.35, + "learning_rate": 1.2183387363404111e-05, + "loss": 0.925, + "step": 36526 + }, + { + "epoch": 2.35, + "learning_rate": 1.218317267438116e-05, + "loss": 0.9021, + "step": 36527 + }, + { + "epoch": 2.35, + "learning_rate": 1.218295798535821e-05, + "loss": 0.8618, + "step": 36528 + }, + { + "epoch": 2.35, + "learning_rate": 1.218274329633526e-05, + "loss": 0.893, + "step": 36529 + }, + { + "epoch": 2.35, + "learning_rate": 1.218252860731231e-05, + "loss": 0.9002, + "step": 36530 + }, + { + "epoch": 2.35, + "learning_rate": 1.218231391828936e-05, + "loss": 0.9574, + "step": 36531 + }, + { + "epoch": 2.35, + "learning_rate": 1.2182099229266408e-05, + "loss": 0.9508, + "step": 36532 + }, + { + "epoch": 2.35, + "learning_rate": 1.2181884540243459e-05, + "loss": 0.8863, + "step": 36533 + }, + { + "epoch": 2.35, + "learning_rate": 1.2181669851220508e-05, + "loss": 0.8713, + "step": 36534 + }, + { + "epoch": 2.35, + "learning_rate": 1.2181455162197557e-05, + "loss": 0.9172, + "step": 36535 + }, + { + "epoch": 2.35, + "learning_rate": 1.2181455162197557e-05, + "loss": 0.8833, + "step": 36536 + }, + { + "epoch": 2.35, + "learning_rate": 1.2181240473174607e-05, + "loss": 0.9068, + "step": 36537 + }, + { + "epoch": 2.35, + "learning_rate": 1.2181025784151656e-05, + "loss": 0.8497, + "step": 36538 + }, + { + "epoch": 2.35, + "learning_rate": 1.2180811095128705e-05, + "loss": 0.9802, + "step": 36539 + }, + { + "epoch": 2.35, + "learning_rate": 1.2180596406105756e-05, + "loss": 0.9667, + "step": 36540 + }, + { + "epoch": 2.35, + "learning_rate": 1.2180381717082808e-05, + "loss": 0.8841, + "step": 36541 + }, + { + "epoch": 2.35, + "learning_rate": 1.2180167028059857e-05, + "loss": 0.887, + "step": 36542 + }, + { + "epoch": 2.35, + "learning_rate": 1.2179952339036906e-05, + "loss": 0.884, + "step": 36543 + }, + { + "epoch": 2.35, + "learning_rate": 1.2179737650013957e-05, + "loss": 0.8919, + "step": 36544 + }, + { + "epoch": 2.35, + "learning_rate": 1.2179522960991006e-05, + "loss": 0.9223, + "step": 36545 + }, + { + "epoch": 2.35, + "learning_rate": 1.2179308271968055e-05, + "loss": 0.9021, + "step": 36546 + }, + { + "epoch": 2.35, + "learning_rate": 1.2179093582945106e-05, + "loss": 0.8893, + "step": 36547 + }, + { + "epoch": 2.35, + "learning_rate": 1.2178878893922155e-05, + "loss": 0.9235, + "step": 36548 + }, + { + "epoch": 2.35, + "learning_rate": 1.2178664204899204e-05, + "loss": 0.9485, + "step": 36549 + }, + { + "epoch": 2.35, + "learning_rate": 1.2178449515876254e-05, + "loss": 0.9667, + "step": 36550 + }, + { + "epoch": 2.35, + "learning_rate": 1.2178234826853303e-05, + "loss": 0.941, + "step": 36551 + }, + { + "epoch": 2.35, + "learning_rate": 1.2178020137830354e-05, + "loss": 0.9173, + "step": 36552 + }, + { + "epoch": 2.35, + "learning_rate": 1.2177805448807403e-05, + "loss": 0.9022, + "step": 36553 + }, + { + "epoch": 2.35, + "learning_rate": 1.2177590759784452e-05, + "loss": 0.8699, + "step": 36554 + }, + { + "epoch": 2.35, + "learning_rate": 1.2177376070761502e-05, + "loss": 0.8687, + "step": 36555 + }, + { + "epoch": 2.35, + "learning_rate": 1.2177161381738553e-05, + "loss": 0.8568, + "step": 36556 + }, + { + "epoch": 2.35, + "learning_rate": 1.2176946692715604e-05, + "loss": 0.8594, + "step": 36557 + }, + { + "epoch": 2.35, + "learning_rate": 1.2176732003692653e-05, + "loss": 0.8788, + "step": 36558 + }, + { + "epoch": 2.35, + "learning_rate": 1.2176517314669702e-05, + "loss": 0.9236, + "step": 36559 + }, + { + "epoch": 2.35, + "learning_rate": 1.2176302625646752e-05, + "loss": 0.9286, + "step": 36560 + }, + { + "epoch": 2.35, + "learning_rate": 1.2176087936623801e-05, + "loss": 0.9671, + "step": 36561 + }, + { + "epoch": 2.35, + "learning_rate": 1.2175873247600852e-05, + "loss": 0.8664, + "step": 36562 + }, + { + "epoch": 2.35, + "learning_rate": 1.2175658558577901e-05, + "loss": 0.8933, + "step": 36563 + }, + { + "epoch": 2.35, + "learning_rate": 1.217544386955495e-05, + "loss": 0.9333, + "step": 36564 + }, + { + "epoch": 2.35, + "learning_rate": 1.2175229180532e-05, + "loss": 0.8696, + "step": 36565 + }, + { + "epoch": 2.35, + "learning_rate": 1.217501449150905e-05, + "loss": 0.8716, + "step": 36566 + }, + { + "epoch": 2.35, + "learning_rate": 1.2174799802486098e-05, + "loss": 0.9398, + "step": 36567 + }, + { + "epoch": 2.35, + "learning_rate": 1.2174585113463149e-05, + "loss": 0.9407, + "step": 36568 + }, + { + "epoch": 2.35, + "learning_rate": 1.2174370424440198e-05, + "loss": 0.9621, + "step": 36569 + }, + { + "epoch": 2.35, + "learning_rate": 1.217415573541725e-05, + "loss": 0.949, + "step": 36570 + }, + { + "epoch": 2.35, + "learning_rate": 1.21739410463943e-05, + "loss": 0.8484, + "step": 36571 + }, + { + "epoch": 2.35, + "learning_rate": 1.217372635737135e-05, + "loss": 0.8711, + "step": 36572 + }, + { + "epoch": 2.35, + "learning_rate": 1.2173511668348399e-05, + "loss": 0.8396, + "step": 36573 + }, + { + "epoch": 2.35, + "learning_rate": 1.2173296979325448e-05, + "loss": 0.9223, + "step": 36574 + }, + { + "epoch": 2.35, + "learning_rate": 1.2173082290302499e-05, + "loss": 0.8664, + "step": 36575 + }, + { + "epoch": 2.35, + "learning_rate": 1.2172867601279548e-05, + "loss": 0.9231, + "step": 36576 + }, + { + "epoch": 2.35, + "learning_rate": 1.2172652912256597e-05, + "loss": 0.8467, + "step": 36577 + }, + { + "epoch": 2.35, + "learning_rate": 1.2172438223233647e-05, + "loss": 0.8857, + "step": 36578 + }, + { + "epoch": 2.35, + "learning_rate": 1.2172223534210696e-05, + "loss": 0.9291, + "step": 36579 + }, + { + "epoch": 2.35, + "learning_rate": 1.2172008845187747e-05, + "loss": 0.886, + "step": 36580 + }, + { + "epoch": 2.35, + "learning_rate": 1.2171794156164796e-05, + "loss": 0.8327, + "step": 36581 + }, + { + "epoch": 2.35, + "learning_rate": 1.2171579467141845e-05, + "loss": 0.8457, + "step": 36582 + }, + { + "epoch": 2.35, + "learning_rate": 1.2171364778118895e-05, + "loss": 0.8897, + "step": 36583 + }, + { + "epoch": 2.35, + "learning_rate": 1.2171150089095944e-05, + "loss": 0.8648, + "step": 36584 + }, + { + "epoch": 2.35, + "learning_rate": 1.2170935400072997e-05, + "loss": 0.8728, + "step": 36585 + }, + { + "epoch": 2.35, + "learning_rate": 1.2170720711050046e-05, + "loss": 0.8727, + "step": 36586 + }, + { + "epoch": 2.35, + "learning_rate": 1.2170506022027095e-05, + "loss": 0.8876, + "step": 36587 + }, + { + "epoch": 2.35, + "learning_rate": 1.2170291333004145e-05, + "loss": 0.8932, + "step": 36588 + }, + { + "epoch": 2.35, + "learning_rate": 1.2170076643981194e-05, + "loss": 0.877, + "step": 36589 + }, + { + "epoch": 2.35, + "learning_rate": 1.2169861954958245e-05, + "loss": 0.9222, + "step": 36590 + }, + { + "epoch": 2.35, + "learning_rate": 1.2169647265935294e-05, + "loss": 0.9802, + "step": 36591 + }, + { + "epoch": 2.35, + "learning_rate": 1.2169432576912343e-05, + "loss": 0.8976, + "step": 36592 + }, + { + "epoch": 2.35, + "learning_rate": 1.2169217887889394e-05, + "loss": 0.8499, + "step": 36593 + }, + { + "epoch": 2.35, + "learning_rate": 1.2169003198866443e-05, + "loss": 0.8443, + "step": 36594 + }, + { + "epoch": 2.35, + "learning_rate": 1.2168788509843491e-05, + "loss": 0.8964, + "step": 36595 + }, + { + "epoch": 2.35, + "learning_rate": 1.2168573820820542e-05, + "loss": 0.9354, + "step": 36596 + }, + { + "epoch": 2.35, + "learning_rate": 1.2168359131797591e-05, + "loss": 0.8701, + "step": 36597 + }, + { + "epoch": 2.35, + "learning_rate": 1.216814444277464e-05, + "loss": 0.9331, + "step": 36598 + }, + { + "epoch": 2.35, + "learning_rate": 1.2167929753751692e-05, + "loss": 0.9414, + "step": 36599 + }, + { + "epoch": 2.35, + "learning_rate": 1.2167715064728743e-05, + "loss": 0.8133, + "step": 36600 + }, + { + "epoch": 2.35, + "learning_rate": 1.2167500375705792e-05, + "loss": 0.883, + "step": 36601 + }, + { + "epoch": 2.35, + "learning_rate": 1.2167285686682841e-05, + "loss": 0.902, + "step": 36602 + }, + { + "epoch": 2.35, + "learning_rate": 1.2167070997659892e-05, + "loss": 0.8242, + "step": 36603 + }, + { + "epoch": 2.36, + "learning_rate": 1.216685630863694e-05, + "loss": 0.9098, + "step": 36604 + }, + { + "epoch": 2.36, + "learning_rate": 1.216664161961399e-05, + "loss": 0.8616, + "step": 36605 + }, + { + "epoch": 2.36, + "learning_rate": 1.216642693059104e-05, + "loss": 0.8395, + "step": 36606 + }, + { + "epoch": 2.36, + "learning_rate": 1.216621224156809e-05, + "loss": 0.8968, + "step": 36607 + }, + { + "epoch": 2.36, + "learning_rate": 1.2165997552545138e-05, + "loss": 0.8934, + "step": 36608 + }, + { + "epoch": 2.36, + "learning_rate": 1.2165782863522189e-05, + "loss": 0.9822, + "step": 36609 + }, + { + "epoch": 2.36, + "learning_rate": 1.2165568174499238e-05, + "loss": 0.9247, + "step": 36610 + }, + { + "epoch": 2.36, + "learning_rate": 1.2165353485476288e-05, + "loss": 0.9233, + "step": 36611 + }, + { + "epoch": 2.36, + "learning_rate": 1.2165138796453337e-05, + "loss": 0.8141, + "step": 36612 + }, + { + "epoch": 2.36, + "learning_rate": 1.2164924107430386e-05, + "loss": 0.8358, + "step": 36613 + }, + { + "epoch": 2.36, + "learning_rate": 1.2164709418407439e-05, + "loss": 0.8898, + "step": 36614 + }, + { + "epoch": 2.36, + "learning_rate": 1.2164494729384488e-05, + "loss": 0.9406, + "step": 36615 + }, + { + "epoch": 2.36, + "learning_rate": 1.2164280040361538e-05, + "loss": 0.9345, + "step": 36616 + }, + { + "epoch": 2.36, + "learning_rate": 1.2164065351338587e-05, + "loss": 0.9047, + "step": 36617 + }, + { + "epoch": 2.36, + "learning_rate": 1.2163850662315636e-05, + "loss": 0.906, + "step": 36618 + }, + { + "epoch": 2.36, + "learning_rate": 1.2163635973292687e-05, + "loss": 0.9021, + "step": 36619 + }, + { + "epoch": 2.36, + "learning_rate": 1.2163421284269736e-05, + "loss": 0.9099, + "step": 36620 + }, + { + "epoch": 2.36, + "learning_rate": 1.2163206595246787e-05, + "loss": 0.9181, + "step": 36621 + }, + { + "epoch": 2.36, + "learning_rate": 1.2162991906223836e-05, + "loss": 0.9076, + "step": 36622 + }, + { + "epoch": 2.36, + "learning_rate": 1.2162777217200885e-05, + "loss": 0.8682, + "step": 36623 + }, + { + "epoch": 2.36, + "learning_rate": 1.2162562528177935e-05, + "loss": 0.906, + "step": 36624 + }, + { + "epoch": 2.36, + "learning_rate": 1.2162347839154984e-05, + "loss": 0.9076, + "step": 36625 + }, + { + "epoch": 2.36, + "learning_rate": 1.2162133150132033e-05, + "loss": 0.9368, + "step": 36626 + }, + { + "epoch": 2.36, + "learning_rate": 1.2161918461109084e-05, + "loss": 0.8733, + "step": 36627 + }, + { + "epoch": 2.36, + "learning_rate": 1.2161703772086134e-05, + "loss": 0.8833, + "step": 36628 + }, + { + "epoch": 2.36, + "learning_rate": 1.2161489083063185e-05, + "loss": 0.7802, + "step": 36629 + }, + { + "epoch": 2.36, + "learning_rate": 1.2161274394040234e-05, + "loss": 0.8947, + "step": 36630 + }, + { + "epoch": 2.36, + "learning_rate": 1.2161059705017285e-05, + "loss": 0.8994, + "step": 36631 + }, + { + "epoch": 2.36, + "learning_rate": 1.2160845015994334e-05, + "loss": 0.8431, + "step": 36632 + }, + { + "epoch": 2.36, + "learning_rate": 1.2160630326971383e-05, + "loss": 0.8687, + "step": 36633 + }, + { + "epoch": 2.36, + "learning_rate": 1.2160415637948433e-05, + "loss": 0.8984, + "step": 36634 + }, + { + "epoch": 2.36, + "learning_rate": 1.2160200948925482e-05, + "loss": 0.8546, + "step": 36635 + }, + { + "epoch": 2.36, + "learning_rate": 1.2159986259902531e-05, + "loss": 0.8627, + "step": 36636 + }, + { + "epoch": 2.36, + "learning_rate": 1.2159771570879582e-05, + "loss": 0.901, + "step": 36637 + }, + { + "epoch": 2.36, + "learning_rate": 1.2159556881856631e-05, + "loss": 0.843, + "step": 36638 + }, + { + "epoch": 2.36, + "learning_rate": 1.2159342192833682e-05, + "loss": 0.9499, + "step": 36639 + }, + { + "epoch": 2.36, + "learning_rate": 1.215912750381073e-05, + "loss": 0.9713, + "step": 36640 + }, + { + "epoch": 2.36, + "learning_rate": 1.215891281478778e-05, + "loss": 0.8933, + "step": 36641 + }, + { + "epoch": 2.36, + "learning_rate": 1.215869812576483e-05, + "loss": 0.849, + "step": 36642 + }, + { + "epoch": 2.36, + "learning_rate": 1.215848343674188e-05, + "loss": 0.8591, + "step": 36643 + }, + { + "epoch": 2.36, + "learning_rate": 1.2158268747718931e-05, + "loss": 0.8703, + "step": 36644 + }, + { + "epoch": 2.36, + "learning_rate": 1.215805405869598e-05, + "loss": 0.9159, + "step": 36645 + }, + { + "epoch": 2.36, + "learning_rate": 1.215783936967303e-05, + "loss": 0.9367, + "step": 36646 + }, + { + "epoch": 2.36, + "learning_rate": 1.215762468065008e-05, + "loss": 0.8948, + "step": 36647 + }, + { + "epoch": 2.36, + "learning_rate": 1.2157409991627129e-05, + "loss": 0.8183, + "step": 36648 + }, + { + "epoch": 2.36, + "learning_rate": 1.215719530260418e-05, + "loss": 0.908, + "step": 36649 + }, + { + "epoch": 2.36, + "learning_rate": 1.2156980613581229e-05, + "loss": 0.8306, + "step": 36650 + }, + { + "epoch": 2.36, + "learning_rate": 1.2156765924558278e-05, + "loss": 0.9096, + "step": 36651 + }, + { + "epoch": 2.36, + "learning_rate": 1.2156551235535328e-05, + "loss": 0.879, + "step": 36652 + }, + { + "epoch": 2.36, + "learning_rate": 1.2156336546512377e-05, + "loss": 0.9106, + "step": 36653 + }, + { + "epoch": 2.36, + "learning_rate": 1.2156121857489426e-05, + "loss": 0.9445, + "step": 36654 + }, + { + "epoch": 2.36, + "learning_rate": 1.2155907168466477e-05, + "loss": 0.8946, + "step": 36655 + }, + { + "epoch": 2.36, + "learning_rate": 1.2155692479443526e-05, + "loss": 0.8976, + "step": 36656 + }, + { + "epoch": 2.36, + "learning_rate": 1.2155477790420578e-05, + "loss": 0.8522, + "step": 36657 + }, + { + "epoch": 2.36, + "learning_rate": 1.2155263101397627e-05, + "loss": 0.8933, + "step": 36658 + }, + { + "epoch": 2.36, + "learning_rate": 1.2155048412374678e-05, + "loss": 0.9529, + "step": 36659 + }, + { + "epoch": 2.36, + "learning_rate": 1.2154833723351727e-05, + "loss": 0.8815, + "step": 36660 + }, + { + "epoch": 2.36, + "learning_rate": 1.2154619034328776e-05, + "loss": 0.9222, + "step": 36661 + }, + { + "epoch": 2.36, + "learning_rate": 1.2154404345305826e-05, + "loss": 0.8992, + "step": 36662 + }, + { + "epoch": 2.36, + "learning_rate": 1.2154189656282875e-05, + "loss": 0.8738, + "step": 36663 + }, + { + "epoch": 2.36, + "learning_rate": 1.2153974967259924e-05, + "loss": 0.9316, + "step": 36664 + }, + { + "epoch": 2.36, + "learning_rate": 1.2153760278236975e-05, + "loss": 0.9412, + "step": 36665 + }, + { + "epoch": 2.36, + "learning_rate": 1.2153545589214024e-05, + "loss": 0.9397, + "step": 36666 + }, + { + "epoch": 2.36, + "learning_rate": 1.2153330900191075e-05, + "loss": 0.8475, + "step": 36667 + }, + { + "epoch": 2.36, + "learning_rate": 1.2153116211168124e-05, + "loss": 0.9438, + "step": 36668 + }, + { + "epoch": 2.36, + "learning_rate": 1.2152901522145173e-05, + "loss": 0.8818, + "step": 36669 + }, + { + "epoch": 2.36, + "learning_rate": 1.2152686833122223e-05, + "loss": 0.933, + "step": 36670 + }, + { + "epoch": 2.36, + "learning_rate": 1.2152472144099272e-05, + "loss": 0.8972, + "step": 36671 + }, + { + "epoch": 2.36, + "learning_rate": 1.2152257455076324e-05, + "loss": 0.8853, + "step": 36672 + }, + { + "epoch": 2.36, + "learning_rate": 1.2152042766053373e-05, + "loss": 0.84, + "step": 36673 + }, + { + "epoch": 2.36, + "learning_rate": 1.2151828077030422e-05, + "loss": 0.8739, + "step": 36674 + }, + { + "epoch": 2.36, + "learning_rate": 1.2151613388007473e-05, + "loss": 0.9185, + "step": 36675 + }, + { + "epoch": 2.36, + "learning_rate": 1.2151398698984522e-05, + "loss": 0.9446, + "step": 36676 + }, + { + "epoch": 2.36, + "learning_rate": 1.2151184009961573e-05, + "loss": 0.9279, + "step": 36677 + }, + { + "epoch": 2.36, + "learning_rate": 1.2150969320938622e-05, + "loss": 0.8493, + "step": 36678 + }, + { + "epoch": 2.36, + "learning_rate": 1.215075463191567e-05, + "loss": 0.9499, + "step": 36679 + }, + { + "epoch": 2.36, + "learning_rate": 1.2150539942892721e-05, + "loss": 0.9271, + "step": 36680 + }, + { + "epoch": 2.36, + "learning_rate": 1.215032525386977e-05, + "loss": 0.89, + "step": 36681 + }, + { + "epoch": 2.36, + "learning_rate": 1.215011056484682e-05, + "loss": 0.9148, + "step": 36682 + }, + { + "epoch": 2.36, + "learning_rate": 1.214989587582387e-05, + "loss": 0.8721, + "step": 36683 + }, + { + "epoch": 2.36, + "learning_rate": 1.2149681186800919e-05, + "loss": 0.8824, + "step": 36684 + }, + { + "epoch": 2.36, + "learning_rate": 1.2149466497777968e-05, + "loss": 0.8767, + "step": 36685 + }, + { + "epoch": 2.36, + "learning_rate": 1.214925180875502e-05, + "loss": 0.8653, + "step": 36686 + }, + { + "epoch": 2.36, + "learning_rate": 1.214903711973207e-05, + "loss": 0.8423, + "step": 36687 + }, + { + "epoch": 2.36, + "learning_rate": 1.214882243070912e-05, + "loss": 0.8427, + "step": 36688 + }, + { + "epoch": 2.36, + "learning_rate": 1.2148607741686169e-05, + "loss": 1.0, + "step": 36689 + }, + { + "epoch": 2.36, + "learning_rate": 1.214839305266322e-05, + "loss": 0.9099, + "step": 36690 + }, + { + "epoch": 2.36, + "learning_rate": 1.2148178363640268e-05, + "loss": 0.8868, + "step": 36691 + }, + { + "epoch": 2.36, + "learning_rate": 1.2147963674617317e-05, + "loss": 0.9365, + "step": 36692 + }, + { + "epoch": 2.36, + "learning_rate": 1.2147748985594368e-05, + "loss": 0.8897, + "step": 36693 + }, + { + "epoch": 2.36, + "learning_rate": 1.2147534296571417e-05, + "loss": 0.9499, + "step": 36694 + }, + { + "epoch": 2.36, + "learning_rate": 1.2147319607548466e-05, + "loss": 0.9001, + "step": 36695 + }, + { + "epoch": 2.36, + "learning_rate": 1.2147104918525517e-05, + "loss": 0.8638, + "step": 36696 + }, + { + "epoch": 2.36, + "learning_rate": 1.2146890229502566e-05, + "loss": 0.9032, + "step": 36697 + }, + { + "epoch": 2.36, + "learning_rate": 1.2146675540479616e-05, + "loss": 0.8455, + "step": 36698 + }, + { + "epoch": 2.36, + "learning_rate": 1.2146460851456665e-05, + "loss": 0.9248, + "step": 36699 + }, + { + "epoch": 2.36, + "learning_rate": 1.2146246162433714e-05, + "loss": 0.8583, + "step": 36700 + }, + { + "epoch": 2.36, + "learning_rate": 1.2146031473410766e-05, + "loss": 0.9212, + "step": 36701 + }, + { + "epoch": 2.36, + "learning_rate": 1.2145816784387815e-05, + "loss": 0.904, + "step": 36702 + }, + { + "epoch": 2.36, + "learning_rate": 1.2145602095364866e-05, + "loss": 0.8767, + "step": 36703 + }, + { + "epoch": 2.36, + "learning_rate": 1.2145387406341915e-05, + "loss": 0.8715, + "step": 36704 + }, + { + "epoch": 2.36, + "learning_rate": 1.2145172717318964e-05, + "loss": 0.8673, + "step": 36705 + }, + { + "epoch": 2.36, + "learning_rate": 1.2144958028296015e-05, + "loss": 0.9184, + "step": 36706 + }, + { + "epoch": 2.36, + "learning_rate": 1.2144743339273064e-05, + "loss": 0.9225, + "step": 36707 + }, + { + "epoch": 2.36, + "learning_rate": 1.2144528650250114e-05, + "loss": 0.8861, + "step": 36708 + }, + { + "epoch": 2.36, + "learning_rate": 1.2144313961227163e-05, + "loss": 0.9379, + "step": 36709 + }, + { + "epoch": 2.36, + "learning_rate": 1.2144099272204212e-05, + "loss": 0.8616, + "step": 36710 + }, + { + "epoch": 2.36, + "learning_rate": 1.2143884583181263e-05, + "loss": 0.8591, + "step": 36711 + }, + { + "epoch": 2.36, + "learning_rate": 1.2143669894158312e-05, + "loss": 0.8003, + "step": 36712 + }, + { + "epoch": 2.36, + "learning_rate": 1.214345520513536e-05, + "loss": 0.9158, + "step": 36713 + }, + { + "epoch": 2.36, + "learning_rate": 1.2143240516112411e-05, + "loss": 0.9882, + "step": 36714 + }, + { + "epoch": 2.36, + "learning_rate": 1.2143025827089462e-05, + "loss": 0.8588, + "step": 36715 + }, + { + "epoch": 2.36, + "learning_rate": 1.2142811138066513e-05, + "loss": 0.9736, + "step": 36716 + }, + { + "epoch": 2.36, + "learning_rate": 1.2142596449043562e-05, + "loss": 0.9315, + "step": 36717 + }, + { + "epoch": 2.36, + "learning_rate": 1.2142381760020612e-05, + "loss": 0.916, + "step": 36718 + }, + { + "epoch": 2.36, + "learning_rate": 1.2142167070997661e-05, + "loss": 0.8786, + "step": 36719 + }, + { + "epoch": 2.36, + "learning_rate": 1.214195238197471e-05, + "loss": 0.9007, + "step": 36720 + }, + { + "epoch": 2.36, + "learning_rate": 1.2141737692951761e-05, + "loss": 0.9062, + "step": 36721 + }, + { + "epoch": 2.36, + "learning_rate": 1.214152300392881e-05, + "loss": 0.8705, + "step": 36722 + }, + { + "epoch": 2.36, + "learning_rate": 1.2141308314905859e-05, + "loss": 0.9054, + "step": 36723 + }, + { + "epoch": 2.36, + "learning_rate": 1.214109362588291e-05, + "loss": 0.9438, + "step": 36724 + }, + { + "epoch": 2.36, + "learning_rate": 1.2140878936859959e-05, + "loss": 0.9191, + "step": 36725 + }, + { + "epoch": 2.36, + "learning_rate": 1.214066424783701e-05, + "loss": 0.9821, + "step": 36726 + }, + { + "epoch": 2.36, + "learning_rate": 1.2140449558814058e-05, + "loss": 0.9255, + "step": 36727 + }, + { + "epoch": 2.36, + "learning_rate": 1.2140234869791107e-05, + "loss": 0.9091, + "step": 36728 + }, + { + "epoch": 2.36, + "learning_rate": 1.2140020180768158e-05, + "loss": 0.8324, + "step": 36729 + }, + { + "epoch": 2.36, + "learning_rate": 1.2139805491745208e-05, + "loss": 0.9188, + "step": 36730 + }, + { + "epoch": 2.36, + "learning_rate": 1.2139590802722259e-05, + "loss": 0.8385, + "step": 36731 + }, + { + "epoch": 2.36, + "learning_rate": 1.2139376113699308e-05, + "loss": 0.8969, + "step": 36732 + }, + { + "epoch": 2.36, + "learning_rate": 1.2139161424676357e-05, + "loss": 0.8685, + "step": 36733 + }, + { + "epoch": 2.36, + "learning_rate": 1.2138946735653408e-05, + "loss": 0.8615, + "step": 36734 + }, + { + "epoch": 2.36, + "learning_rate": 1.2138732046630457e-05, + "loss": 0.9712, + "step": 36735 + }, + { + "epoch": 2.36, + "learning_rate": 1.2138517357607507e-05, + "loss": 0.8612, + "step": 36736 + }, + { + "epoch": 2.36, + "learning_rate": 1.2138302668584556e-05, + "loss": 0.8852, + "step": 36737 + }, + { + "epoch": 2.36, + "learning_rate": 1.2138087979561605e-05, + "loss": 0.9148, + "step": 36738 + }, + { + "epoch": 2.36, + "learning_rate": 1.2137873290538656e-05, + "loss": 0.925, + "step": 36739 + }, + { + "epoch": 2.36, + "learning_rate": 1.2137658601515705e-05, + "loss": 0.9412, + "step": 36740 + }, + { + "epoch": 2.36, + "learning_rate": 1.2137443912492754e-05, + "loss": 0.8393, + "step": 36741 + }, + { + "epoch": 2.36, + "learning_rate": 1.2137229223469805e-05, + "loss": 0.9081, + "step": 36742 + }, + { + "epoch": 2.36, + "learning_rate": 1.2137014534446854e-05, + "loss": 0.911, + "step": 36743 + }, + { + "epoch": 2.36, + "learning_rate": 1.2136799845423906e-05, + "loss": 0.9171, + "step": 36744 + }, + { + "epoch": 2.36, + "learning_rate": 1.2136585156400955e-05, + "loss": 0.8805, + "step": 36745 + }, + { + "epoch": 2.36, + "learning_rate": 1.2136370467378005e-05, + "loss": 0.9488, + "step": 36746 + }, + { + "epoch": 2.36, + "learning_rate": 1.2136155778355054e-05, + "loss": 0.9418, + "step": 36747 + }, + { + "epoch": 2.36, + "learning_rate": 1.2135941089332103e-05, + "loss": 0.867, + "step": 36748 + }, + { + "epoch": 2.36, + "learning_rate": 1.2135726400309154e-05, + "loss": 0.917, + "step": 36749 + }, + { + "epoch": 2.36, + "learning_rate": 1.2135511711286203e-05, + "loss": 0.9174, + "step": 36750 + }, + { + "epoch": 2.36, + "learning_rate": 1.2135297022263252e-05, + "loss": 0.9671, + "step": 36751 + }, + { + "epoch": 2.36, + "learning_rate": 1.2135082333240303e-05, + "loss": 0.932, + "step": 36752 + }, + { + "epoch": 2.36, + "learning_rate": 1.2134867644217352e-05, + "loss": 0.8642, + "step": 36753 + }, + { + "epoch": 2.36, + "learning_rate": 1.2134652955194402e-05, + "loss": 0.8526, + "step": 36754 + }, + { + "epoch": 2.36, + "learning_rate": 1.2134438266171451e-05, + "loss": 0.8659, + "step": 36755 + }, + { + "epoch": 2.36, + "learning_rate": 1.21342235771485e-05, + "loss": 0.9232, + "step": 36756 + }, + { + "epoch": 2.36, + "learning_rate": 1.2134008888125551e-05, + "loss": 0.8404, + "step": 36757 + }, + { + "epoch": 2.36, + "learning_rate": 1.21337941991026e-05, + "loss": 0.8681, + "step": 36758 + }, + { + "epoch": 2.36, + "learning_rate": 1.2133579510079652e-05, + "loss": 0.9229, + "step": 36759 + }, + { + "epoch": 2.37, + "learning_rate": 1.2133364821056701e-05, + "loss": 0.9256, + "step": 36760 + }, + { + "epoch": 2.37, + "learning_rate": 1.213315013203375e-05, + "loss": 0.9435, + "step": 36761 + }, + { + "epoch": 2.37, + "learning_rate": 1.21329354430108e-05, + "loss": 0.8703, + "step": 36762 + }, + { + "epoch": 2.37, + "learning_rate": 1.213272075398785e-05, + "loss": 0.8914, + "step": 36763 + }, + { + "epoch": 2.37, + "learning_rate": 1.21325060649649e-05, + "loss": 0.8821, + "step": 36764 + }, + { + "epoch": 2.37, + "learning_rate": 1.213229137594195e-05, + "loss": 0.858, + "step": 36765 + }, + { + "epoch": 2.37, + "learning_rate": 1.2132076686918998e-05, + "loss": 1.0103, + "step": 36766 + }, + { + "epoch": 2.37, + "learning_rate": 1.2131861997896049e-05, + "loss": 0.8951, + "step": 36767 + }, + { + "epoch": 2.37, + "learning_rate": 1.2131647308873098e-05, + "loss": 0.9072, + "step": 36768 + }, + { + "epoch": 2.37, + "learning_rate": 1.2131432619850147e-05, + "loss": 0.9359, + "step": 36769 + }, + { + "epoch": 2.37, + "learning_rate": 1.2131217930827198e-05, + "loss": 0.8489, + "step": 36770 + }, + { + "epoch": 2.37, + "learning_rate": 1.2131003241804247e-05, + "loss": 0.9106, + "step": 36771 + }, + { + "epoch": 2.37, + "learning_rate": 1.2130788552781296e-05, + "loss": 0.9943, + "step": 36772 + }, + { + "epoch": 2.37, + "learning_rate": 1.2130573863758348e-05, + "loss": 0.9268, + "step": 36773 + }, + { + "epoch": 2.37, + "learning_rate": 1.2130359174735399e-05, + "loss": 0.8649, + "step": 36774 + }, + { + "epoch": 2.37, + "learning_rate": 1.2130144485712447e-05, + "loss": 0.8526, + "step": 36775 + }, + { + "epoch": 2.37, + "learning_rate": 1.2129929796689496e-05, + "loss": 0.9511, + "step": 36776 + }, + { + "epoch": 2.37, + "learning_rate": 1.2129715107666547e-05, + "loss": 0.8459, + "step": 36777 + }, + { + "epoch": 2.37, + "learning_rate": 1.2129500418643596e-05, + "loss": 0.8063, + "step": 36778 + }, + { + "epoch": 2.37, + "learning_rate": 1.2129285729620645e-05, + "loss": 0.8786, + "step": 36779 + }, + { + "epoch": 2.37, + "learning_rate": 1.2129071040597696e-05, + "loss": 0.9188, + "step": 36780 + }, + { + "epoch": 2.37, + "learning_rate": 1.2128856351574745e-05, + "loss": 0.9453, + "step": 36781 + }, + { + "epoch": 2.37, + "learning_rate": 1.2128641662551794e-05, + "loss": 0.9607, + "step": 36782 + }, + { + "epoch": 2.37, + "learning_rate": 1.2128426973528844e-05, + "loss": 0.9249, + "step": 36783 + }, + { + "epoch": 2.37, + "learning_rate": 1.2128212284505893e-05, + "loss": 0.9457, + "step": 36784 + }, + { + "epoch": 2.37, + "learning_rate": 1.2127997595482944e-05, + "loss": 0.9305, + "step": 36785 + }, + { + "epoch": 2.37, + "learning_rate": 1.2127782906459993e-05, + "loss": 0.985, + "step": 36786 + }, + { + "epoch": 2.37, + "learning_rate": 1.2127568217437042e-05, + "loss": 0.9578, + "step": 36787 + }, + { + "epoch": 2.37, + "learning_rate": 1.2127353528414094e-05, + "loss": 0.9411, + "step": 36788 + }, + { + "epoch": 2.37, + "learning_rate": 1.2127138839391143e-05, + "loss": 0.8732, + "step": 36789 + }, + { + "epoch": 2.37, + "learning_rate": 1.2126924150368194e-05, + "loss": 0.9275, + "step": 36790 + }, + { + "epoch": 2.37, + "learning_rate": 1.2126709461345243e-05, + "loss": 0.9183, + "step": 36791 + }, + { + "epoch": 2.37, + "learning_rate": 1.2126494772322292e-05, + "loss": 0.8732, + "step": 36792 + }, + { + "epoch": 2.37, + "learning_rate": 1.2126280083299342e-05, + "loss": 0.9205, + "step": 36793 + }, + { + "epoch": 2.37, + "learning_rate": 1.2126065394276391e-05, + "loss": 0.9234, + "step": 36794 + }, + { + "epoch": 2.37, + "learning_rate": 1.2125850705253442e-05, + "loss": 0.8987, + "step": 36795 + }, + { + "epoch": 2.37, + "learning_rate": 1.2125636016230491e-05, + "loss": 0.913, + "step": 36796 + }, + { + "epoch": 2.37, + "learning_rate": 1.212542132720754e-05, + "loss": 0.9893, + "step": 36797 + }, + { + "epoch": 2.37, + "learning_rate": 1.212520663818459e-05, + "loss": 0.795, + "step": 36798 + }, + { + "epoch": 2.37, + "learning_rate": 1.212499194916164e-05, + "loss": 0.8539, + "step": 36799 + }, + { + "epoch": 2.37, + "learning_rate": 1.2124777260138689e-05, + "loss": 0.9278, + "step": 36800 + }, + { + "epoch": 2.37, + "learning_rate": 1.212456257111574e-05, + "loss": 0.8882, + "step": 36801 + }, + { + "epoch": 2.37, + "learning_rate": 1.212434788209279e-05, + "loss": 0.9002, + "step": 36802 + }, + { + "epoch": 2.37, + "learning_rate": 1.212413319306984e-05, + "loss": 0.8442, + "step": 36803 + }, + { + "epoch": 2.37, + "learning_rate": 1.212391850404689e-05, + "loss": 0.8948, + "step": 36804 + }, + { + "epoch": 2.37, + "learning_rate": 1.212370381502394e-05, + "loss": 0.8758, + "step": 36805 + }, + { + "epoch": 2.37, + "learning_rate": 1.2123489126000989e-05, + "loss": 0.885, + "step": 36806 + }, + { + "epoch": 2.37, + "learning_rate": 1.2123274436978038e-05, + "loss": 0.9714, + "step": 36807 + }, + { + "epoch": 2.37, + "learning_rate": 1.2123059747955089e-05, + "loss": 0.9054, + "step": 36808 + }, + { + "epoch": 2.37, + "learning_rate": 1.2122845058932138e-05, + "loss": 0.9474, + "step": 36809 + }, + { + "epoch": 2.37, + "learning_rate": 1.2122630369909187e-05, + "loss": 0.948, + "step": 36810 + }, + { + "epoch": 2.37, + "learning_rate": 1.2122415680886237e-05, + "loss": 0.9094, + "step": 36811 + }, + { + "epoch": 2.37, + "learning_rate": 1.2122200991863286e-05, + "loss": 0.9738, + "step": 36812 + }, + { + "epoch": 2.37, + "learning_rate": 1.2121986302840337e-05, + "loss": 0.8921, + "step": 36813 + }, + { + "epoch": 2.37, + "learning_rate": 1.2121771613817386e-05, + "loss": 0.8911, + "step": 36814 + }, + { + "epoch": 2.37, + "learning_rate": 1.2121556924794435e-05, + "loss": 0.955, + "step": 36815 + }, + { + "epoch": 2.37, + "learning_rate": 1.2121342235771486e-05, + "loss": 0.9641, + "step": 36816 + }, + { + "epoch": 2.37, + "learning_rate": 1.2121127546748536e-05, + "loss": 0.7957, + "step": 36817 + }, + { + "epoch": 2.37, + "learning_rate": 1.2120912857725587e-05, + "loss": 0.9224, + "step": 36818 + }, + { + "epoch": 2.37, + "learning_rate": 1.2120698168702636e-05, + "loss": 0.8909, + "step": 36819 + }, + { + "epoch": 2.37, + "learning_rate": 1.2120483479679685e-05, + "loss": 0.8899, + "step": 36820 + }, + { + "epoch": 2.37, + "learning_rate": 1.2120268790656735e-05, + "loss": 0.9262, + "step": 36821 + }, + { + "epoch": 2.37, + "learning_rate": 1.2120054101633784e-05, + "loss": 0.9061, + "step": 36822 + }, + { + "epoch": 2.37, + "learning_rate": 1.2119839412610835e-05, + "loss": 0.9377, + "step": 36823 + }, + { + "epoch": 2.37, + "learning_rate": 1.2119624723587884e-05, + "loss": 0.9611, + "step": 36824 + }, + { + "epoch": 2.37, + "learning_rate": 1.2119410034564933e-05, + "loss": 0.9065, + "step": 36825 + }, + { + "epoch": 2.37, + "learning_rate": 1.2119195345541984e-05, + "loss": 0.9191, + "step": 36826 + }, + { + "epoch": 2.37, + "learning_rate": 1.2118980656519033e-05, + "loss": 0.9049, + "step": 36827 + }, + { + "epoch": 2.37, + "learning_rate": 1.2118765967496082e-05, + "loss": 0.9435, + "step": 36828 + }, + { + "epoch": 2.37, + "learning_rate": 1.2118551278473132e-05, + "loss": 0.8305, + "step": 36829 + }, + { + "epoch": 2.37, + "learning_rate": 1.2118336589450181e-05, + "loss": 0.8402, + "step": 36830 + }, + { + "epoch": 2.37, + "learning_rate": 1.2118121900427234e-05, + "loss": 0.9097, + "step": 36831 + }, + { + "epoch": 2.37, + "learning_rate": 1.2117907211404283e-05, + "loss": 0.9451, + "step": 36832 + }, + { + "epoch": 2.37, + "learning_rate": 1.2117692522381333e-05, + "loss": 0.8823, + "step": 36833 + }, + { + "epoch": 2.37, + "learning_rate": 1.2117477833358382e-05, + "loss": 0.854, + "step": 36834 + }, + { + "epoch": 2.37, + "learning_rate": 1.2117263144335431e-05, + "loss": 0.9564, + "step": 36835 + }, + { + "epoch": 2.37, + "learning_rate": 1.2117048455312482e-05, + "loss": 0.8713, + "step": 36836 + }, + { + "epoch": 2.37, + "learning_rate": 1.211683376628953e-05, + "loss": 0.9487, + "step": 36837 + }, + { + "epoch": 2.37, + "learning_rate": 1.211661907726658e-05, + "loss": 0.9439, + "step": 36838 + }, + { + "epoch": 2.37, + "learning_rate": 1.211640438824363e-05, + "loss": 0.8617, + "step": 36839 + }, + { + "epoch": 2.37, + "learning_rate": 1.211618969922068e-05, + "loss": 0.9396, + "step": 36840 + }, + { + "epoch": 2.37, + "learning_rate": 1.2115975010197728e-05, + "loss": 0.9015, + "step": 36841 + }, + { + "epoch": 2.37, + "learning_rate": 1.2115760321174779e-05, + "loss": 0.9482, + "step": 36842 + }, + { + "epoch": 2.37, + "learning_rate": 1.2115545632151828e-05, + "loss": 0.8293, + "step": 36843 + }, + { + "epoch": 2.37, + "learning_rate": 1.2115330943128879e-05, + "loss": 0.9539, + "step": 36844 + }, + { + "epoch": 2.37, + "learning_rate": 1.2115116254105928e-05, + "loss": 0.917, + "step": 36845 + }, + { + "epoch": 2.37, + "learning_rate": 1.211490156508298e-05, + "loss": 0.9147, + "step": 36846 + }, + { + "epoch": 2.37, + "learning_rate": 1.2114686876060029e-05, + "loss": 0.8737, + "step": 36847 + }, + { + "epoch": 2.37, + "learning_rate": 1.2114472187037078e-05, + "loss": 0.9228, + "step": 36848 + }, + { + "epoch": 2.37, + "learning_rate": 1.2114257498014128e-05, + "loss": 0.994, + "step": 36849 + }, + { + "epoch": 2.37, + "learning_rate": 1.2114042808991177e-05, + "loss": 0.9369, + "step": 36850 + }, + { + "epoch": 2.37, + "learning_rate": 1.2113828119968228e-05, + "loss": 0.9025, + "step": 36851 + }, + { + "epoch": 2.37, + "learning_rate": 1.2113613430945277e-05, + "loss": 0.8703, + "step": 36852 + }, + { + "epoch": 2.37, + "learning_rate": 1.2113398741922326e-05, + "loss": 0.8892, + "step": 36853 + }, + { + "epoch": 2.37, + "learning_rate": 1.2113184052899377e-05, + "loss": 0.8507, + "step": 36854 + }, + { + "epoch": 2.37, + "learning_rate": 1.2112969363876426e-05, + "loss": 0.9007, + "step": 36855 + }, + { + "epoch": 2.37, + "learning_rate": 1.2112754674853475e-05, + "loss": 0.8343, + "step": 36856 + }, + { + "epoch": 2.37, + "learning_rate": 1.2112539985830525e-05, + "loss": 0.9022, + "step": 36857 + }, + { + "epoch": 2.37, + "learning_rate": 1.2112325296807574e-05, + "loss": 0.876, + "step": 36858 + }, + { + "epoch": 2.37, + "learning_rate": 1.2112110607784623e-05, + "loss": 0.8213, + "step": 36859 + }, + { + "epoch": 2.37, + "learning_rate": 1.2111895918761676e-05, + "loss": 0.8529, + "step": 36860 + }, + { + "epoch": 2.37, + "learning_rate": 1.2111681229738726e-05, + "loss": 0.9303, + "step": 36861 + }, + { + "epoch": 2.37, + "learning_rate": 1.2111466540715775e-05, + "loss": 0.873, + "step": 36862 + }, + { + "epoch": 2.37, + "learning_rate": 1.2111251851692824e-05, + "loss": 0.883, + "step": 36863 + }, + { + "epoch": 2.37, + "learning_rate": 1.2111037162669875e-05, + "loss": 0.8886, + "step": 36864 + }, + { + "epoch": 2.37, + "learning_rate": 1.2110822473646924e-05, + "loss": 0.9325, + "step": 36865 + }, + { + "epoch": 2.37, + "learning_rate": 1.2110607784623973e-05, + "loss": 0.9084, + "step": 36866 + }, + { + "epoch": 2.37, + "learning_rate": 1.2110393095601023e-05, + "loss": 0.9143, + "step": 36867 + }, + { + "epoch": 2.37, + "learning_rate": 1.2110178406578072e-05, + "loss": 0.8783, + "step": 36868 + }, + { + "epoch": 2.37, + "learning_rate": 1.2109963717555121e-05, + "loss": 0.9653, + "step": 36869 + }, + { + "epoch": 2.37, + "learning_rate": 1.2109749028532172e-05, + "loss": 0.8595, + "step": 36870 + }, + { + "epoch": 2.37, + "learning_rate": 1.2109534339509221e-05, + "loss": 0.8919, + "step": 36871 + }, + { + "epoch": 2.37, + "learning_rate": 1.2109319650486272e-05, + "loss": 0.9138, + "step": 36872 + }, + { + "epoch": 2.37, + "learning_rate": 1.210910496146332e-05, + "loss": 0.8468, + "step": 36873 + }, + { + "epoch": 2.37, + "learning_rate": 1.210889027244037e-05, + "loss": 0.9501, + "step": 36874 + }, + { + "epoch": 2.37, + "learning_rate": 1.2108675583417422e-05, + "loss": 0.884, + "step": 36875 + }, + { + "epoch": 2.37, + "learning_rate": 1.2108460894394471e-05, + "loss": 0.985, + "step": 36876 + }, + { + "epoch": 2.37, + "learning_rate": 1.2108246205371522e-05, + "loss": 0.9621, + "step": 36877 + }, + { + "epoch": 2.37, + "learning_rate": 1.210803151634857e-05, + "loss": 0.894, + "step": 36878 + }, + { + "epoch": 2.37, + "learning_rate": 1.210781682732562e-05, + "loss": 0.8657, + "step": 36879 + }, + { + "epoch": 2.37, + "learning_rate": 1.210760213830267e-05, + "loss": 0.9308, + "step": 36880 + }, + { + "epoch": 2.37, + "learning_rate": 1.2107387449279719e-05, + "loss": 0.9349, + "step": 36881 + }, + { + "epoch": 2.37, + "learning_rate": 1.210717276025677e-05, + "loss": 0.9547, + "step": 36882 + }, + { + "epoch": 2.37, + "learning_rate": 1.2106958071233819e-05, + "loss": 0.8548, + "step": 36883 + }, + { + "epoch": 2.37, + "learning_rate": 1.2106743382210868e-05, + "loss": 0.834, + "step": 36884 + }, + { + "epoch": 2.37, + "learning_rate": 1.2106528693187918e-05, + "loss": 0.8296, + "step": 36885 + }, + { + "epoch": 2.37, + "learning_rate": 1.2106314004164967e-05, + "loss": 0.9184, + "step": 36886 + }, + { + "epoch": 2.37, + "learning_rate": 1.2106099315142016e-05, + "loss": 0.9307, + "step": 36887 + }, + { + "epoch": 2.37, + "learning_rate": 1.2105884626119067e-05, + "loss": 0.7757, + "step": 36888 + }, + { + "epoch": 2.37, + "learning_rate": 1.2105669937096118e-05, + "loss": 1.0055, + "step": 36889 + }, + { + "epoch": 2.37, + "learning_rate": 1.2105455248073168e-05, + "loss": 0.9065, + "step": 36890 + }, + { + "epoch": 2.37, + "learning_rate": 1.2105240559050217e-05, + "loss": 0.9178, + "step": 36891 + }, + { + "epoch": 2.37, + "learning_rate": 1.2105025870027268e-05, + "loss": 0.8352, + "step": 36892 + }, + { + "epoch": 2.37, + "learning_rate": 1.2104811181004317e-05, + "loss": 0.8627, + "step": 36893 + }, + { + "epoch": 2.37, + "learning_rate": 1.2104596491981366e-05, + "loss": 0.8695, + "step": 36894 + }, + { + "epoch": 2.37, + "learning_rate": 1.2104381802958416e-05, + "loss": 0.9141, + "step": 36895 + }, + { + "epoch": 2.37, + "learning_rate": 1.2104167113935465e-05, + "loss": 0.9128, + "step": 36896 + }, + { + "epoch": 2.37, + "learning_rate": 1.2103952424912514e-05, + "loss": 0.9004, + "step": 36897 + }, + { + "epoch": 2.37, + "learning_rate": 1.2103737735889565e-05, + "loss": 0.995, + "step": 36898 + }, + { + "epoch": 2.37, + "learning_rate": 1.2103523046866614e-05, + "loss": 0.8471, + "step": 36899 + }, + { + "epoch": 2.37, + "learning_rate": 1.2103308357843665e-05, + "loss": 0.8932, + "step": 36900 + }, + { + "epoch": 2.37, + "learning_rate": 1.2103093668820714e-05, + "loss": 0.8897, + "step": 36901 + }, + { + "epoch": 2.37, + "learning_rate": 1.2102878979797763e-05, + "loss": 0.9423, + "step": 36902 + }, + { + "epoch": 2.37, + "learning_rate": 1.2102664290774813e-05, + "loss": 0.9063, + "step": 36903 + }, + { + "epoch": 2.37, + "learning_rate": 1.2102449601751864e-05, + "loss": 0.9217, + "step": 36904 + }, + { + "epoch": 2.37, + "learning_rate": 1.2102234912728915e-05, + "loss": 0.8349, + "step": 36905 + }, + { + "epoch": 2.37, + "learning_rate": 1.2102020223705964e-05, + "loss": 0.8995, + "step": 36906 + }, + { + "epoch": 2.37, + "learning_rate": 1.2101805534683013e-05, + "loss": 0.9226, + "step": 36907 + }, + { + "epoch": 2.37, + "learning_rate": 1.2101590845660063e-05, + "loss": 0.8853, + "step": 36908 + }, + { + "epoch": 2.37, + "learning_rate": 1.2101376156637112e-05, + "loss": 0.9232, + "step": 36909 + }, + { + "epoch": 2.37, + "learning_rate": 1.2101161467614163e-05, + "loss": 0.8714, + "step": 36910 + }, + { + "epoch": 2.37, + "learning_rate": 1.2100946778591212e-05, + "loss": 0.9747, + "step": 36911 + }, + { + "epoch": 2.37, + "learning_rate": 1.210073208956826e-05, + "loss": 0.8686, + "step": 36912 + }, + { + "epoch": 2.37, + "learning_rate": 1.2100517400545311e-05, + "loss": 0.8218, + "step": 36913 + }, + { + "epoch": 2.37, + "learning_rate": 1.210030271152236e-05, + "loss": 0.8609, + "step": 36914 + }, + { + "epoch": 2.38, + "learning_rate": 1.210008802249941e-05, + "loss": 0.9356, + "step": 36915 + }, + { + "epoch": 2.38, + "learning_rate": 1.209987333347646e-05, + "loss": 0.8997, + "step": 36916 + }, + { + "epoch": 2.38, + "learning_rate": 1.2099658644453509e-05, + "loss": 0.9414, + "step": 36917 + }, + { + "epoch": 2.38, + "learning_rate": 1.2099443955430558e-05, + "loss": 1.0001, + "step": 36918 + }, + { + "epoch": 2.38, + "learning_rate": 1.209922926640761e-05, + "loss": 0.9128, + "step": 36919 + }, + { + "epoch": 2.38, + "learning_rate": 1.2099014577384661e-05, + "loss": 0.9328, + "step": 36920 + }, + { + "epoch": 2.38, + "learning_rate": 1.209879988836171e-05, + "loss": 0.8603, + "step": 36921 + }, + { + "epoch": 2.38, + "learning_rate": 1.2098585199338759e-05, + "loss": 0.8798, + "step": 36922 + }, + { + "epoch": 2.38, + "learning_rate": 1.209837051031581e-05, + "loss": 0.8353, + "step": 36923 + }, + { + "epoch": 2.38, + "learning_rate": 1.2098155821292858e-05, + "loss": 0.9776, + "step": 36924 + }, + { + "epoch": 2.38, + "learning_rate": 1.2097941132269907e-05, + "loss": 0.9246, + "step": 36925 + }, + { + "epoch": 2.38, + "learning_rate": 1.2097726443246958e-05, + "loss": 0.887, + "step": 36926 + }, + { + "epoch": 2.38, + "learning_rate": 1.2097511754224007e-05, + "loss": 0.8952, + "step": 36927 + }, + { + "epoch": 2.38, + "learning_rate": 1.2097297065201056e-05, + "loss": 0.8271, + "step": 36928 + }, + { + "epoch": 2.38, + "learning_rate": 1.2097082376178107e-05, + "loss": 0.9038, + "step": 36929 + }, + { + "epoch": 2.38, + "learning_rate": 1.2096867687155156e-05, + "loss": 0.8715, + "step": 36930 + }, + { + "epoch": 2.38, + "learning_rate": 1.2096652998132206e-05, + "loss": 0.8736, + "step": 36931 + }, + { + "epoch": 2.38, + "learning_rate": 1.2096438309109255e-05, + "loss": 0.88, + "step": 36932 + }, + { + "epoch": 2.38, + "learning_rate": 1.2096223620086308e-05, + "loss": 0.8749, + "step": 36933 + }, + { + "epoch": 2.38, + "learning_rate": 1.2096008931063357e-05, + "loss": 0.9172, + "step": 36934 + }, + { + "epoch": 2.38, + "learning_rate": 1.2095794242040406e-05, + "loss": 0.9455, + "step": 36935 + }, + { + "epoch": 2.38, + "learning_rate": 1.2095579553017456e-05, + "loss": 0.9257, + "step": 36936 + }, + { + "epoch": 2.38, + "learning_rate": 1.2095364863994505e-05, + "loss": 0.8357, + "step": 36937 + }, + { + "epoch": 2.38, + "learning_rate": 1.2095150174971554e-05, + "loss": 0.8702, + "step": 36938 + }, + { + "epoch": 2.38, + "learning_rate": 1.2094935485948605e-05, + "loss": 0.9334, + "step": 36939 + }, + { + "epoch": 2.38, + "learning_rate": 1.2094720796925654e-05, + "loss": 0.9456, + "step": 36940 + }, + { + "epoch": 2.38, + "learning_rate": 1.2094506107902704e-05, + "loss": 0.8842, + "step": 36941 + }, + { + "epoch": 2.38, + "learning_rate": 1.2094291418879753e-05, + "loss": 0.8728, + "step": 36942 + }, + { + "epoch": 2.38, + "learning_rate": 1.2094076729856802e-05, + "loss": 0.9048, + "step": 36943 + }, + { + "epoch": 2.38, + "learning_rate": 1.2093862040833853e-05, + "loss": 0.8929, + "step": 36944 + }, + { + "epoch": 2.38, + "learning_rate": 1.2093647351810902e-05, + "loss": 0.94, + "step": 36945 + }, + { + "epoch": 2.38, + "learning_rate": 1.2093432662787951e-05, + "loss": 0.8556, + "step": 36946 + }, + { + "epoch": 2.38, + "learning_rate": 1.2093217973765002e-05, + "loss": 0.872, + "step": 36947 + }, + { + "epoch": 2.38, + "learning_rate": 1.2093003284742052e-05, + "loss": 0.9027, + "step": 36948 + }, + { + "epoch": 2.38, + "learning_rate": 1.2092788595719103e-05, + "loss": 0.9266, + "step": 36949 + }, + { + "epoch": 2.38, + "learning_rate": 1.2092573906696152e-05, + "loss": 0.8958, + "step": 36950 + }, + { + "epoch": 2.38, + "learning_rate": 1.2092359217673203e-05, + "loss": 0.9113, + "step": 36951 + }, + { + "epoch": 2.38, + "learning_rate": 1.2092144528650252e-05, + "loss": 0.8948, + "step": 36952 + }, + { + "epoch": 2.38, + "learning_rate": 1.20919298396273e-05, + "loss": 0.9491, + "step": 36953 + }, + { + "epoch": 2.38, + "learning_rate": 1.2091715150604351e-05, + "loss": 0.8899, + "step": 36954 + }, + { + "epoch": 2.38, + "learning_rate": 1.20915004615814e-05, + "loss": 0.9036, + "step": 36955 + }, + { + "epoch": 2.38, + "learning_rate": 1.2091285772558449e-05, + "loss": 0.8634, + "step": 36956 + }, + { + "epoch": 2.38, + "learning_rate": 1.20910710835355e-05, + "loss": 0.9059, + "step": 36957 + }, + { + "epoch": 2.38, + "learning_rate": 1.2090856394512549e-05, + "loss": 0.843, + "step": 36958 + }, + { + "epoch": 2.38, + "learning_rate": 1.20906417054896e-05, + "loss": 0.8378, + "step": 36959 + }, + { + "epoch": 2.38, + "learning_rate": 1.2090427016466648e-05, + "loss": 0.8716, + "step": 36960 + }, + { + "epoch": 2.38, + "learning_rate": 1.2090212327443697e-05, + "loss": 0.8881, + "step": 36961 + }, + { + "epoch": 2.38, + "learning_rate": 1.208999763842075e-05, + "loss": 0.9137, + "step": 36962 + }, + { + "epoch": 2.38, + "learning_rate": 1.2089782949397799e-05, + "loss": 0.9394, + "step": 36963 + }, + { + "epoch": 2.38, + "learning_rate": 1.208956826037485e-05, + "loss": 0.8897, + "step": 36964 + }, + { + "epoch": 2.38, + "learning_rate": 1.2089353571351898e-05, + "loss": 0.9204, + "step": 36965 + }, + { + "epoch": 2.38, + "learning_rate": 1.2089138882328947e-05, + "loss": 0.8434, + "step": 36966 + }, + { + "epoch": 2.38, + "learning_rate": 1.2088924193305998e-05, + "loss": 0.8875, + "step": 36967 + }, + { + "epoch": 2.38, + "learning_rate": 1.2088709504283047e-05, + "loss": 0.9219, + "step": 36968 + }, + { + "epoch": 2.38, + "learning_rate": 1.2088494815260097e-05, + "loss": 0.938, + "step": 36969 + }, + { + "epoch": 2.38, + "learning_rate": 1.2088280126237146e-05, + "loss": 0.9576, + "step": 36970 + }, + { + "epoch": 2.38, + "learning_rate": 1.2088065437214195e-05, + "loss": 0.9038, + "step": 36971 + }, + { + "epoch": 2.38, + "learning_rate": 1.2087850748191246e-05, + "loss": 0.8535, + "step": 36972 + }, + { + "epoch": 2.38, + "learning_rate": 1.2087636059168295e-05, + "loss": 0.8716, + "step": 36973 + }, + { + "epoch": 2.38, + "learning_rate": 1.2087421370145344e-05, + "loss": 0.8459, + "step": 36974 + }, + { + "epoch": 2.38, + "learning_rate": 1.2087206681122395e-05, + "loss": 0.8983, + "step": 36975 + }, + { + "epoch": 2.38, + "learning_rate": 1.2086991992099444e-05, + "loss": 0.9433, + "step": 36976 + }, + { + "epoch": 2.38, + "learning_rate": 1.2086777303076496e-05, + "loss": 0.8442, + "step": 36977 + }, + { + "epoch": 2.38, + "learning_rate": 1.2086562614053545e-05, + "loss": 0.9439, + "step": 36978 + }, + { + "epoch": 2.38, + "learning_rate": 1.2086347925030596e-05, + "loss": 0.8556, + "step": 36979 + }, + { + "epoch": 2.38, + "learning_rate": 1.2086133236007645e-05, + "loss": 0.9151, + "step": 36980 + }, + { + "epoch": 2.38, + "learning_rate": 1.2085918546984694e-05, + "loss": 0.8588, + "step": 36981 + }, + { + "epoch": 2.38, + "learning_rate": 1.2085703857961744e-05, + "loss": 0.943, + "step": 36982 + }, + { + "epoch": 2.38, + "learning_rate": 1.2085489168938793e-05, + "loss": 0.9157, + "step": 36983 + }, + { + "epoch": 2.38, + "learning_rate": 1.2085274479915842e-05, + "loss": 0.9382, + "step": 36984 + }, + { + "epoch": 2.38, + "learning_rate": 1.2085059790892893e-05, + "loss": 0.8875, + "step": 36985 + }, + { + "epoch": 2.38, + "learning_rate": 1.2084845101869942e-05, + "loss": 0.923, + "step": 36986 + }, + { + "epoch": 2.38, + "learning_rate": 1.2084630412846992e-05, + "loss": 0.8772, + "step": 36987 + }, + { + "epoch": 2.38, + "learning_rate": 1.2084415723824041e-05, + "loss": 0.914, + "step": 36988 + }, + { + "epoch": 2.38, + "learning_rate": 1.208420103480109e-05, + "loss": 0.9488, + "step": 36989 + }, + { + "epoch": 2.38, + "learning_rate": 1.2083986345778141e-05, + "loss": 0.8959, + "step": 36990 + }, + { + "epoch": 2.38, + "learning_rate": 1.2083771656755192e-05, + "loss": 0.9426, + "step": 36991 + }, + { + "epoch": 2.38, + "learning_rate": 1.2083556967732242e-05, + "loss": 0.8752, + "step": 36992 + }, + { + "epoch": 2.38, + "learning_rate": 1.2083342278709291e-05, + "loss": 0.8117, + "step": 36993 + }, + { + "epoch": 2.38, + "learning_rate": 1.208312758968634e-05, + "loss": 0.8576, + "step": 36994 + }, + { + "epoch": 2.38, + "learning_rate": 1.2082912900663391e-05, + "loss": 0.9707, + "step": 36995 + }, + { + "epoch": 2.38, + "learning_rate": 1.208269821164044e-05, + "loss": 0.9129, + "step": 36996 + }, + { + "epoch": 2.38, + "learning_rate": 1.208248352261749e-05, + "loss": 0.9214, + "step": 36997 + }, + { + "epoch": 2.38, + "learning_rate": 1.208226883359454e-05, + "loss": 0.8099, + "step": 36998 + }, + { + "epoch": 2.38, + "learning_rate": 1.2082054144571588e-05, + "loss": 0.8594, + "step": 36999 + }, + { + "epoch": 2.38, + "learning_rate": 1.2081839455548639e-05, + "loss": 0.9711, + "step": 37000 + }, + { + "epoch": 2.38, + "eval_laion/cider": 0.4973755300081719, + "eval_laion/meteor": 0.3860635809698547, + "eval_laion/pred_num_words": 36.615234375, + "eval_laion/sacrebleu": 15.854540533267889, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0436550378799438, + "eval_runtime": 95.9727, + "eval_samples_per_second": 5.199, + "eval_steps_per_second": 0.333, + "step": 37000 + }, + { + "epoch": 2.38, + "learning_rate": 1.2081624766525688e-05, + "loss": 0.906, + "step": 37001 + }, + { + "epoch": 2.38, + "learning_rate": 1.2081410077502737e-05, + "loss": 0.893, + "step": 37002 + }, + { + "epoch": 2.38, + "learning_rate": 1.2081195388479788e-05, + "loss": 0.8791, + "step": 37003 + }, + { + "epoch": 2.38, + "learning_rate": 1.2080980699456837e-05, + "loss": 0.8602, + "step": 37004 + }, + { + "epoch": 2.38, + "learning_rate": 1.2080766010433886e-05, + "loss": 0.9036, + "step": 37005 + }, + { + "epoch": 2.38, + "learning_rate": 1.2080551321410938e-05, + "loss": 0.9525, + "step": 37006 + }, + { + "epoch": 2.38, + "learning_rate": 1.2080336632387989e-05, + "loss": 0.8463, + "step": 37007 + }, + { + "epoch": 2.38, + "learning_rate": 1.2080121943365038e-05, + "loss": 0.8854, + "step": 37008 + }, + { + "epoch": 2.38, + "learning_rate": 1.2079907254342087e-05, + "loss": 0.8256, + "step": 37009 + }, + { + "epoch": 2.38, + "learning_rate": 1.2079692565319137e-05, + "loss": 0.8721, + "step": 37010 + }, + { + "epoch": 2.38, + "learning_rate": 1.2079477876296186e-05, + "loss": 0.8851, + "step": 37011 + }, + { + "epoch": 2.38, + "learning_rate": 1.2079263187273235e-05, + "loss": 0.9516, + "step": 37012 + }, + { + "epoch": 2.38, + "learning_rate": 1.2079048498250286e-05, + "loss": 0.8412, + "step": 37013 + }, + { + "epoch": 2.38, + "learning_rate": 1.2078833809227335e-05, + "loss": 0.8826, + "step": 37014 + }, + { + "epoch": 2.38, + "learning_rate": 1.2078619120204384e-05, + "loss": 0.9313, + "step": 37015 + }, + { + "epoch": 2.38, + "learning_rate": 1.2078404431181434e-05, + "loss": 0.9166, + "step": 37016 + }, + { + "epoch": 2.38, + "learning_rate": 1.2078189742158483e-05, + "loss": 0.9406, + "step": 37017 + }, + { + "epoch": 2.38, + "learning_rate": 1.2077975053135534e-05, + "loss": 0.9264, + "step": 37018 + }, + { + "epoch": 2.38, + "learning_rate": 1.2077760364112583e-05, + "loss": 0.8739, + "step": 37019 + }, + { + "epoch": 2.38, + "learning_rate": 1.2077545675089635e-05, + "loss": 0.9321, + "step": 37020 + }, + { + "epoch": 2.38, + "learning_rate": 1.2077330986066684e-05, + "loss": 0.9565, + "step": 37021 + }, + { + "epoch": 2.38, + "learning_rate": 1.2077116297043733e-05, + "loss": 0.9052, + "step": 37022 + }, + { + "epoch": 2.38, + "learning_rate": 1.2076901608020784e-05, + "loss": 0.8567, + "step": 37023 + }, + { + "epoch": 2.38, + "learning_rate": 1.2076686918997833e-05, + "loss": 0.8754, + "step": 37024 + }, + { + "epoch": 2.38, + "learning_rate": 1.2076472229974882e-05, + "loss": 0.966, + "step": 37025 + }, + { + "epoch": 2.38, + "learning_rate": 1.2076257540951933e-05, + "loss": 0.9023, + "step": 37026 + }, + { + "epoch": 2.38, + "learning_rate": 1.2076042851928981e-05, + "loss": 0.8907, + "step": 37027 + }, + { + "epoch": 2.38, + "learning_rate": 1.2075828162906032e-05, + "loss": 0.9093, + "step": 37028 + }, + { + "epoch": 2.38, + "learning_rate": 1.2075613473883081e-05, + "loss": 0.9237, + "step": 37029 + }, + { + "epoch": 2.38, + "learning_rate": 1.207539878486013e-05, + "loss": 0.8574, + "step": 37030 + }, + { + "epoch": 2.38, + "learning_rate": 1.207518409583718e-05, + "loss": 0.8805, + "step": 37031 + }, + { + "epoch": 2.38, + "learning_rate": 1.207496940681423e-05, + "loss": 0.9006, + "step": 37032 + }, + { + "epoch": 2.38, + "learning_rate": 1.2074754717791279e-05, + "loss": 0.9491, + "step": 37033 + }, + { + "epoch": 2.38, + "learning_rate": 1.207454002876833e-05, + "loss": 0.8688, + "step": 37034 + }, + { + "epoch": 2.38, + "learning_rate": 1.207432533974538e-05, + "loss": 0.8183, + "step": 37035 + }, + { + "epoch": 2.38, + "learning_rate": 1.207411065072243e-05, + "loss": 0.8857, + "step": 37036 + }, + { + "epoch": 2.38, + "learning_rate": 1.207389596169948e-05, + "loss": 0.8606, + "step": 37037 + }, + { + "epoch": 2.38, + "learning_rate": 1.207368127267653e-05, + "loss": 0.9051, + "step": 37038 + }, + { + "epoch": 2.38, + "learning_rate": 1.207346658365358e-05, + "loss": 0.9156, + "step": 37039 + }, + { + "epoch": 2.38, + "learning_rate": 1.2073251894630628e-05, + "loss": 0.9105, + "step": 37040 + }, + { + "epoch": 2.38, + "learning_rate": 1.2073037205607679e-05, + "loss": 0.9388, + "step": 37041 + }, + { + "epoch": 2.38, + "learning_rate": 1.2072822516584728e-05, + "loss": 0.986, + "step": 37042 + }, + { + "epoch": 2.38, + "learning_rate": 1.2072607827561777e-05, + "loss": 0.8783, + "step": 37043 + }, + { + "epoch": 2.38, + "learning_rate": 1.2072393138538827e-05, + "loss": 0.9398, + "step": 37044 + }, + { + "epoch": 2.38, + "learning_rate": 1.2072178449515876e-05, + "loss": 0.8397, + "step": 37045 + }, + { + "epoch": 2.38, + "learning_rate": 1.2071963760492927e-05, + "loss": 0.9606, + "step": 37046 + }, + { + "epoch": 2.38, + "learning_rate": 1.2071749071469976e-05, + "loss": 0.8675, + "step": 37047 + }, + { + "epoch": 2.38, + "learning_rate": 1.2071534382447025e-05, + "loss": 0.8587, + "step": 37048 + }, + { + "epoch": 2.38, + "learning_rate": 1.2071319693424077e-05, + "loss": 0.9407, + "step": 37049 + }, + { + "epoch": 2.38, + "learning_rate": 1.2071105004401126e-05, + "loss": 0.934, + "step": 37050 + }, + { + "epoch": 2.38, + "learning_rate": 1.2070890315378177e-05, + "loss": 0.8647, + "step": 37051 + }, + { + "epoch": 2.38, + "learning_rate": 1.2070675626355226e-05, + "loss": 0.8499, + "step": 37052 + }, + { + "epoch": 2.38, + "learning_rate": 1.2070460937332275e-05, + "loss": 0.8574, + "step": 37053 + }, + { + "epoch": 2.38, + "learning_rate": 1.2070246248309326e-05, + "loss": 0.8484, + "step": 37054 + }, + { + "epoch": 2.38, + "learning_rate": 1.2070031559286375e-05, + "loss": 0.9214, + "step": 37055 + }, + { + "epoch": 2.38, + "learning_rate": 1.2069816870263425e-05, + "loss": 0.8698, + "step": 37056 + }, + { + "epoch": 2.38, + "learning_rate": 1.2069602181240474e-05, + "loss": 0.877, + "step": 37057 + }, + { + "epoch": 2.38, + "learning_rate": 1.2069387492217523e-05, + "loss": 0.8186, + "step": 37058 + }, + { + "epoch": 2.38, + "learning_rate": 1.2069172803194574e-05, + "loss": 0.834, + "step": 37059 + }, + { + "epoch": 2.38, + "learning_rate": 1.2068958114171623e-05, + "loss": 0.9237, + "step": 37060 + }, + { + "epoch": 2.38, + "learning_rate": 1.2068743425148672e-05, + "loss": 0.8579, + "step": 37061 + }, + { + "epoch": 2.38, + "learning_rate": 1.2068528736125722e-05, + "loss": 0.9958, + "step": 37062 + }, + { + "epoch": 2.38, + "learning_rate": 1.2068314047102771e-05, + "loss": 0.8856, + "step": 37063 + }, + { + "epoch": 2.38, + "learning_rate": 1.2068099358079824e-05, + "loss": 0.9134, + "step": 37064 + }, + { + "epoch": 2.38, + "learning_rate": 1.2067884669056873e-05, + "loss": 0.8982, + "step": 37065 + }, + { + "epoch": 2.38, + "learning_rate": 1.2067669980033923e-05, + "loss": 0.8795, + "step": 37066 + }, + { + "epoch": 2.38, + "learning_rate": 1.2067455291010972e-05, + "loss": 0.806, + "step": 37067 + }, + { + "epoch": 2.38, + "learning_rate": 1.2067240601988021e-05, + "loss": 0.8834, + "step": 37068 + }, + { + "epoch": 2.38, + "learning_rate": 1.2067025912965072e-05, + "loss": 0.9421, + "step": 37069 + }, + { + "epoch": 2.38, + "learning_rate": 1.206681122394212e-05, + "loss": 0.9216, + "step": 37070 + }, + { + "epoch": 2.39, + "learning_rate": 1.206659653491917e-05, + "loss": 0.8923, + "step": 37071 + }, + { + "epoch": 2.39, + "learning_rate": 1.206638184589622e-05, + "loss": 0.9213, + "step": 37072 + }, + { + "epoch": 2.39, + "learning_rate": 1.206616715687327e-05, + "loss": 0.9404, + "step": 37073 + }, + { + "epoch": 2.39, + "learning_rate": 1.206595246785032e-05, + "loss": 0.9423, + "step": 37074 + }, + { + "epoch": 2.39, + "learning_rate": 1.2065737778827369e-05, + "loss": 0.8689, + "step": 37075 + }, + { + "epoch": 2.39, + "learning_rate": 1.2065523089804418e-05, + "loss": 0.8961, + "step": 37076 + }, + { + "epoch": 2.39, + "learning_rate": 1.2065308400781469e-05, + "loss": 0.8102, + "step": 37077 + }, + { + "epoch": 2.39, + "learning_rate": 1.206509371175852e-05, + "loss": 0.9262, + "step": 37078 + }, + { + "epoch": 2.39, + "learning_rate": 1.206487902273557e-05, + "loss": 0.8139, + "step": 37079 + }, + { + "epoch": 2.39, + "learning_rate": 1.2064664333712619e-05, + "loss": 0.9236, + "step": 37080 + }, + { + "epoch": 2.39, + "learning_rate": 1.2064449644689668e-05, + "loss": 0.8575, + "step": 37081 + }, + { + "epoch": 2.39, + "learning_rate": 1.2064234955666719e-05, + "loss": 0.8992, + "step": 37082 + }, + { + "epoch": 2.39, + "learning_rate": 1.2064020266643768e-05, + "loss": 0.8363, + "step": 37083 + }, + { + "epoch": 2.39, + "learning_rate": 1.2063805577620818e-05, + "loss": 0.9008, + "step": 37084 + }, + { + "epoch": 2.39, + "learning_rate": 1.2063590888597867e-05, + "loss": 0.9209, + "step": 37085 + }, + { + "epoch": 2.39, + "learning_rate": 1.2063376199574916e-05, + "loss": 0.8421, + "step": 37086 + }, + { + "epoch": 2.39, + "learning_rate": 1.2063161510551967e-05, + "loss": 0.9209, + "step": 37087 + }, + { + "epoch": 2.39, + "learning_rate": 1.2062946821529016e-05, + "loss": 0.9221, + "step": 37088 + }, + { + "epoch": 2.39, + "learning_rate": 1.2062732132506065e-05, + "loss": 0.923, + "step": 37089 + }, + { + "epoch": 2.39, + "learning_rate": 1.2062517443483115e-05, + "loss": 0.8367, + "step": 37090 + }, + { + "epoch": 2.39, + "learning_rate": 1.2062302754460164e-05, + "loss": 0.854, + "step": 37091 + }, + { + "epoch": 2.39, + "learning_rate": 1.2062088065437213e-05, + "loss": 0.8905, + "step": 37092 + }, + { + "epoch": 2.39, + "learning_rate": 1.2061873376414266e-05, + "loss": 0.8745, + "step": 37093 + }, + { + "epoch": 2.39, + "learning_rate": 1.2061658687391316e-05, + "loss": 0.8774, + "step": 37094 + }, + { + "epoch": 2.39, + "learning_rate": 1.2061443998368365e-05, + "loss": 0.9076, + "step": 37095 + }, + { + "epoch": 2.39, + "learning_rate": 1.2061229309345414e-05, + "loss": 0.8888, + "step": 37096 + }, + { + "epoch": 2.39, + "learning_rate": 1.2061014620322465e-05, + "loss": 0.8411, + "step": 37097 + }, + { + "epoch": 2.39, + "learning_rate": 1.2060799931299514e-05, + "loss": 0.9029, + "step": 37098 + }, + { + "epoch": 2.39, + "learning_rate": 1.2060585242276563e-05, + "loss": 0.8732, + "step": 37099 + }, + { + "epoch": 2.39, + "learning_rate": 1.2060370553253614e-05, + "loss": 0.8385, + "step": 37100 + }, + { + "epoch": 2.39, + "learning_rate": 1.2060155864230662e-05, + "loss": 0.8821, + "step": 37101 + }, + { + "epoch": 2.39, + "learning_rate": 1.2059941175207711e-05, + "loss": 0.9389, + "step": 37102 + }, + { + "epoch": 2.39, + "learning_rate": 1.2059726486184762e-05, + "loss": 0.9535, + "step": 37103 + }, + { + "epoch": 2.39, + "learning_rate": 1.2059511797161811e-05, + "loss": 0.8667, + "step": 37104 + }, + { + "epoch": 2.39, + "learning_rate": 1.2059297108138862e-05, + "loss": 0.8689, + "step": 37105 + }, + { + "epoch": 2.39, + "learning_rate": 1.205908241911591e-05, + "loss": 0.8568, + "step": 37106 + }, + { + "epoch": 2.39, + "learning_rate": 1.2058867730092963e-05, + "loss": 0.8487, + "step": 37107 + }, + { + "epoch": 2.39, + "learning_rate": 1.2058653041070012e-05, + "loss": 0.8566, + "step": 37108 + }, + { + "epoch": 2.39, + "learning_rate": 1.2058438352047061e-05, + "loss": 0.934, + "step": 37109 + }, + { + "epoch": 2.39, + "learning_rate": 1.2058223663024112e-05, + "loss": 0.8954, + "step": 37110 + }, + { + "epoch": 2.39, + "learning_rate": 1.205800897400116e-05, + "loss": 0.8468, + "step": 37111 + }, + { + "epoch": 2.39, + "learning_rate": 1.205779428497821e-05, + "loss": 0.8443, + "step": 37112 + }, + { + "epoch": 2.39, + "learning_rate": 1.205757959595526e-05, + "loss": 0.8976, + "step": 37113 + }, + { + "epoch": 2.39, + "learning_rate": 1.205736490693231e-05, + "loss": 0.9133, + "step": 37114 + }, + { + "epoch": 2.39, + "learning_rate": 1.205715021790936e-05, + "loss": 0.9233, + "step": 37115 + }, + { + "epoch": 2.39, + "learning_rate": 1.2056935528886409e-05, + "loss": 0.8977, + "step": 37116 + }, + { + "epoch": 2.39, + "learning_rate": 1.2056720839863458e-05, + "loss": 0.9204, + "step": 37117 + }, + { + "epoch": 2.39, + "learning_rate": 1.2056506150840508e-05, + "loss": 0.892, + "step": 37118 + }, + { + "epoch": 2.39, + "learning_rate": 1.2056291461817557e-05, + "loss": 0.8663, + "step": 37119 + }, + { + "epoch": 2.39, + "learning_rate": 1.2056076772794606e-05, + "loss": 0.8613, + "step": 37120 + }, + { + "epoch": 2.39, + "learning_rate": 1.2055862083771657e-05, + "loss": 0.892, + "step": 37121 + }, + { + "epoch": 2.39, + "learning_rate": 1.2055647394748708e-05, + "loss": 0.9523, + "step": 37122 + }, + { + "epoch": 2.39, + "learning_rate": 1.2055432705725758e-05, + "loss": 0.8772, + "step": 37123 + }, + { + "epoch": 2.39, + "learning_rate": 1.2055218016702807e-05, + "loss": 0.9109, + "step": 37124 + }, + { + "epoch": 2.39, + "learning_rate": 1.2055003327679858e-05, + "loss": 0.8948, + "step": 37125 + }, + { + "epoch": 2.39, + "learning_rate": 1.2054788638656907e-05, + "loss": 0.8751, + "step": 37126 + }, + { + "epoch": 2.39, + "learning_rate": 1.2054573949633956e-05, + "loss": 0.8573, + "step": 37127 + }, + { + "epoch": 2.39, + "learning_rate": 1.2054359260611007e-05, + "loss": 0.8695, + "step": 37128 + }, + { + "epoch": 2.39, + "learning_rate": 1.2054144571588056e-05, + "loss": 0.8975, + "step": 37129 + }, + { + "epoch": 2.39, + "learning_rate": 1.2053929882565104e-05, + "loss": 0.8516, + "step": 37130 + }, + { + "epoch": 2.39, + "learning_rate": 1.2053715193542155e-05, + "loss": 0.8691, + "step": 37131 + }, + { + "epoch": 2.39, + "learning_rate": 1.2053500504519204e-05, + "loss": 0.8658, + "step": 37132 + }, + { + "epoch": 2.39, + "learning_rate": 1.2053285815496255e-05, + "loss": 0.8936, + "step": 37133 + }, + { + "epoch": 2.39, + "learning_rate": 1.2053071126473304e-05, + "loss": 0.8385, + "step": 37134 + }, + { + "epoch": 2.39, + "learning_rate": 1.2052856437450353e-05, + "loss": 0.884, + "step": 37135 + }, + { + "epoch": 2.39, + "learning_rate": 1.2052641748427405e-05, + "loss": 0.818, + "step": 37136 + }, + { + "epoch": 2.39, + "learning_rate": 1.2052427059404454e-05, + "loss": 0.9447, + "step": 37137 + }, + { + "epoch": 2.39, + "learning_rate": 1.2052212370381505e-05, + "loss": 0.9115, + "step": 37138 + }, + { + "epoch": 2.39, + "learning_rate": 1.2051997681358554e-05, + "loss": 0.8855, + "step": 37139 + }, + { + "epoch": 2.39, + "learning_rate": 1.2051782992335603e-05, + "loss": 0.925, + "step": 37140 + }, + { + "epoch": 2.39, + "learning_rate": 1.2051568303312653e-05, + "loss": 0.885, + "step": 37141 + }, + { + "epoch": 2.39, + "learning_rate": 1.2051353614289702e-05, + "loss": 0.834, + "step": 37142 + }, + { + "epoch": 2.39, + "learning_rate": 1.2051138925266753e-05, + "loss": 1.0164, + "step": 37143 + }, + { + "epoch": 2.39, + "learning_rate": 1.2050924236243802e-05, + "loss": 0.9357, + "step": 37144 + }, + { + "epoch": 2.39, + "learning_rate": 1.205070954722085e-05, + "loss": 0.9236, + "step": 37145 + }, + { + "epoch": 2.39, + "learning_rate": 1.2050494858197901e-05, + "loss": 0.8723, + "step": 37146 + }, + { + "epoch": 2.39, + "learning_rate": 1.205028016917495e-05, + "loss": 0.9431, + "step": 37147 + }, + { + "epoch": 2.39, + "learning_rate": 1.2050065480152e-05, + "loss": 0.8981, + "step": 37148 + }, + { + "epoch": 2.39, + "learning_rate": 1.204985079112905e-05, + "loss": 0.8611, + "step": 37149 + }, + { + "epoch": 2.39, + "learning_rate": 1.2049636102106099e-05, + "loss": 0.9462, + "step": 37150 + }, + { + "epoch": 2.39, + "learning_rate": 1.2049421413083151e-05, + "loss": 0.936, + "step": 37151 + }, + { + "epoch": 2.39, + "learning_rate": 1.20492067240602e-05, + "loss": 0.8047, + "step": 37152 + }, + { + "epoch": 2.39, + "learning_rate": 1.2048992035037251e-05, + "loss": 0.8549, + "step": 37153 + }, + { + "epoch": 2.39, + "learning_rate": 1.20487773460143e-05, + "loss": 0.8362, + "step": 37154 + }, + { + "epoch": 2.39, + "learning_rate": 1.2048562656991349e-05, + "loss": 0.8687, + "step": 37155 + }, + { + "epoch": 2.39, + "learning_rate": 1.20483479679684e-05, + "loss": 0.9157, + "step": 37156 + }, + { + "epoch": 2.39, + "learning_rate": 1.2048133278945449e-05, + "loss": 0.8889, + "step": 37157 + }, + { + "epoch": 2.39, + "learning_rate": 1.2047918589922498e-05, + "loss": 0.9353, + "step": 37158 + }, + { + "epoch": 2.39, + "learning_rate": 1.2047703900899548e-05, + "loss": 0.8784, + "step": 37159 + }, + { + "epoch": 2.39, + "learning_rate": 1.2047489211876597e-05, + "loss": 0.8795, + "step": 37160 + }, + { + "epoch": 2.39, + "learning_rate": 1.2047274522853648e-05, + "loss": 0.8784, + "step": 37161 + }, + { + "epoch": 2.39, + "learning_rate": 1.2047059833830697e-05, + "loss": 0.8968, + "step": 37162 + }, + { + "epoch": 2.39, + "learning_rate": 1.2046845144807746e-05, + "loss": 0.929, + "step": 37163 + }, + { + "epoch": 2.39, + "learning_rate": 1.2046630455784796e-05, + "loss": 0.9065, + "step": 37164 + }, + { + "epoch": 2.39, + "learning_rate": 1.2046415766761847e-05, + "loss": 0.9048, + "step": 37165 + }, + { + "epoch": 2.39, + "learning_rate": 1.2046201077738898e-05, + "loss": 0.8946, + "step": 37166 + }, + { + "epoch": 2.39, + "learning_rate": 1.2045986388715947e-05, + "loss": 0.8962, + "step": 37167 + }, + { + "epoch": 2.39, + "learning_rate": 1.2045771699692996e-05, + "loss": 0.8922, + "step": 37168 + }, + { + "epoch": 2.39, + "learning_rate": 1.2045557010670046e-05, + "loss": 0.8312, + "step": 37169 + }, + { + "epoch": 2.39, + "learning_rate": 1.2045342321647095e-05, + "loss": 0.9081, + "step": 37170 + }, + { + "epoch": 2.39, + "learning_rate": 1.2045127632624146e-05, + "loss": 0.9115, + "step": 37171 + }, + { + "epoch": 2.39, + "learning_rate": 1.2044912943601195e-05, + "loss": 0.8873, + "step": 37172 + }, + { + "epoch": 2.39, + "learning_rate": 1.2044698254578244e-05, + "loss": 0.9389, + "step": 37173 + }, + { + "epoch": 2.39, + "learning_rate": 1.2044483565555295e-05, + "loss": 0.9287, + "step": 37174 + }, + { + "epoch": 2.39, + "learning_rate": 1.2044268876532343e-05, + "loss": 0.8771, + "step": 37175 + }, + { + "epoch": 2.39, + "learning_rate": 1.2044054187509392e-05, + "loss": 0.9447, + "step": 37176 + }, + { + "epoch": 2.39, + "learning_rate": 1.2043839498486443e-05, + "loss": 0.9921, + "step": 37177 + }, + { + "epoch": 2.39, + "learning_rate": 1.2043624809463492e-05, + "loss": 0.9679, + "step": 37178 + }, + { + "epoch": 2.39, + "learning_rate": 1.2043410120440541e-05, + "loss": 0.8681, + "step": 37179 + }, + { + "epoch": 2.39, + "learning_rate": 1.2043195431417593e-05, + "loss": 0.9063, + "step": 37180 + }, + { + "epoch": 2.39, + "learning_rate": 1.2042980742394644e-05, + "loss": 0.9892, + "step": 37181 + }, + { + "epoch": 2.39, + "learning_rate": 1.2042766053371693e-05, + "loss": 0.9433, + "step": 37182 + }, + { + "epoch": 2.39, + "learning_rate": 1.2042551364348742e-05, + "loss": 0.9505, + "step": 37183 + }, + { + "epoch": 2.39, + "learning_rate": 1.2042336675325793e-05, + "loss": 0.8655, + "step": 37184 + }, + { + "epoch": 2.39, + "learning_rate": 1.2042121986302842e-05, + "loss": 0.9193, + "step": 37185 + }, + { + "epoch": 2.39, + "learning_rate": 1.204190729727989e-05, + "loss": 0.8913, + "step": 37186 + }, + { + "epoch": 2.39, + "learning_rate": 1.2041692608256941e-05, + "loss": 0.8275, + "step": 37187 + }, + { + "epoch": 2.39, + "learning_rate": 1.204147791923399e-05, + "loss": 0.8752, + "step": 37188 + }, + { + "epoch": 2.39, + "learning_rate": 1.2041263230211039e-05, + "loss": 0.9191, + "step": 37189 + }, + { + "epoch": 2.39, + "learning_rate": 1.204104854118809e-05, + "loss": 0.814, + "step": 37190 + }, + { + "epoch": 2.39, + "learning_rate": 1.2040833852165139e-05, + "loss": 0.9369, + "step": 37191 + }, + { + "epoch": 2.39, + "learning_rate": 1.204061916314219e-05, + "loss": 0.8563, + "step": 37192 + }, + { + "epoch": 2.39, + "learning_rate": 1.2040404474119238e-05, + "loss": 0.871, + "step": 37193 + }, + { + "epoch": 2.39, + "learning_rate": 1.204018978509629e-05, + "loss": 0.9241, + "step": 37194 + }, + { + "epoch": 2.39, + "learning_rate": 1.203997509607334e-05, + "loss": 0.8535, + "step": 37195 + }, + { + "epoch": 2.39, + "learning_rate": 1.2039760407050389e-05, + "loss": 0.902, + "step": 37196 + }, + { + "epoch": 2.39, + "learning_rate": 1.203954571802744e-05, + "loss": 0.8952, + "step": 37197 + }, + { + "epoch": 2.39, + "learning_rate": 1.2039331029004488e-05, + "loss": 0.8582, + "step": 37198 + }, + { + "epoch": 2.39, + "learning_rate": 1.2039116339981537e-05, + "loss": 0.9285, + "step": 37199 + }, + { + "epoch": 2.39, + "learning_rate": 1.2038901650958588e-05, + "loss": 0.8681, + "step": 37200 + }, + { + "epoch": 2.39, + "learning_rate": 1.2038686961935637e-05, + "loss": 0.8469, + "step": 37201 + }, + { + "epoch": 2.39, + "learning_rate": 1.2038472272912688e-05, + "loss": 0.8643, + "step": 37202 + }, + { + "epoch": 2.39, + "learning_rate": 1.2038257583889737e-05, + "loss": 0.942, + "step": 37203 + }, + { + "epoch": 2.39, + "learning_rate": 1.2038042894866785e-05, + "loss": 0.877, + "step": 37204 + }, + { + "epoch": 2.39, + "learning_rate": 1.2037828205843836e-05, + "loss": 0.8507, + "step": 37205 + }, + { + "epoch": 2.39, + "learning_rate": 1.2037613516820885e-05, + "loss": 0.9151, + "step": 37206 + }, + { + "epoch": 2.39, + "learning_rate": 1.2037398827797934e-05, + "loss": 0.8938, + "step": 37207 + }, + { + "epoch": 2.39, + "learning_rate": 1.2037184138774985e-05, + "loss": 0.8468, + "step": 37208 + }, + { + "epoch": 2.39, + "learning_rate": 1.2036969449752035e-05, + "loss": 0.8829, + "step": 37209 + }, + { + "epoch": 2.39, + "learning_rate": 1.2036754760729086e-05, + "loss": 0.8582, + "step": 37210 + }, + { + "epoch": 2.39, + "learning_rate": 1.2036540071706135e-05, + "loss": 0.9036, + "step": 37211 + }, + { + "epoch": 2.39, + "learning_rate": 1.2036325382683186e-05, + "loss": 0.8734, + "step": 37212 + }, + { + "epoch": 2.39, + "learning_rate": 1.2036110693660235e-05, + "loss": 0.8463, + "step": 37213 + }, + { + "epoch": 2.39, + "learning_rate": 1.2035896004637284e-05, + "loss": 0.8047, + "step": 37214 + }, + { + "epoch": 2.39, + "learning_rate": 1.2035681315614334e-05, + "loss": 0.9246, + "step": 37215 + }, + { + "epoch": 2.39, + "learning_rate": 1.2035466626591383e-05, + "loss": 0.8793, + "step": 37216 + }, + { + "epoch": 2.39, + "learning_rate": 1.2035251937568432e-05, + "loss": 0.8783, + "step": 37217 + }, + { + "epoch": 2.39, + "learning_rate": 1.2035037248545483e-05, + "loss": 0.9021, + "step": 37218 + }, + { + "epoch": 2.39, + "learning_rate": 1.2034822559522532e-05, + "loss": 0.9207, + "step": 37219 + }, + { + "epoch": 2.39, + "learning_rate": 1.2034607870499582e-05, + "loss": 0.925, + "step": 37220 + }, + { + "epoch": 2.39, + "learning_rate": 1.2034393181476631e-05, + "loss": 0.8664, + "step": 37221 + }, + { + "epoch": 2.39, + "learning_rate": 1.203417849245368e-05, + "loss": 0.932, + "step": 37222 + }, + { + "epoch": 2.39, + "learning_rate": 1.2033963803430733e-05, + "loss": 0.9153, + "step": 37223 + }, + { + "epoch": 2.39, + "learning_rate": 1.2033749114407782e-05, + "loss": 0.8802, + "step": 37224 + }, + { + "epoch": 2.39, + "learning_rate": 1.2033534425384832e-05, + "loss": 0.8211, + "step": 37225 + }, + { + "epoch": 2.4, + "learning_rate": 1.2033319736361881e-05, + "loss": 0.9271, + "step": 37226 + }, + { + "epoch": 2.4, + "learning_rate": 1.203310504733893e-05, + "loss": 0.9184, + "step": 37227 + }, + { + "epoch": 2.4, + "learning_rate": 1.2032890358315981e-05, + "loss": 0.8978, + "step": 37228 + }, + { + "epoch": 2.4, + "learning_rate": 1.203267566929303e-05, + "loss": 0.839, + "step": 37229 + }, + { + "epoch": 2.4, + "learning_rate": 1.203246098027008e-05, + "loss": 0.8696, + "step": 37230 + }, + { + "epoch": 2.4, + "learning_rate": 1.203224629124713e-05, + "loss": 0.8616, + "step": 37231 + }, + { + "epoch": 2.4, + "learning_rate": 1.2032031602224179e-05, + "loss": 0.8838, + "step": 37232 + }, + { + "epoch": 2.4, + "learning_rate": 1.203181691320123e-05, + "loss": 0.8863, + "step": 37233 + }, + { + "epoch": 2.4, + "learning_rate": 1.2031602224178278e-05, + "loss": 0.8941, + "step": 37234 + }, + { + "epoch": 2.4, + "learning_rate": 1.2031387535155327e-05, + "loss": 0.9093, + "step": 37235 + }, + { + "epoch": 2.4, + "learning_rate": 1.2031172846132378e-05, + "loss": 0.9298, + "step": 37236 + }, + { + "epoch": 2.4, + "learning_rate": 1.2030958157109427e-05, + "loss": 0.9403, + "step": 37237 + }, + { + "epoch": 2.4, + "learning_rate": 1.2030743468086479e-05, + "loss": 0.9735, + "step": 37238 + }, + { + "epoch": 2.4, + "learning_rate": 1.2030528779063528e-05, + "loss": 0.88, + "step": 37239 + }, + { + "epoch": 2.4, + "learning_rate": 1.2030314090040579e-05, + "loss": 0.8679, + "step": 37240 + }, + { + "epoch": 2.4, + "learning_rate": 1.2030099401017628e-05, + "loss": 0.9975, + "step": 37241 + }, + { + "epoch": 2.4, + "learning_rate": 1.2029884711994677e-05, + "loss": 0.8924, + "step": 37242 + }, + { + "epoch": 2.4, + "learning_rate": 1.2029670022971727e-05, + "loss": 0.9187, + "step": 37243 + }, + { + "epoch": 2.4, + "learning_rate": 1.2029455333948776e-05, + "loss": 0.9499, + "step": 37244 + }, + { + "epoch": 2.4, + "learning_rate": 1.2029240644925825e-05, + "loss": 0.8701, + "step": 37245 + }, + { + "epoch": 2.4, + "learning_rate": 1.2029025955902876e-05, + "loss": 0.8769, + "step": 37246 + }, + { + "epoch": 2.4, + "learning_rate": 1.2028811266879925e-05, + "loss": 0.9496, + "step": 37247 + }, + { + "epoch": 2.4, + "learning_rate": 1.2028596577856974e-05, + "loss": 0.9035, + "step": 37248 + }, + { + "epoch": 2.4, + "learning_rate": 1.2028381888834024e-05, + "loss": 0.9131, + "step": 37249 + }, + { + "epoch": 2.4, + "learning_rate": 1.2028167199811073e-05, + "loss": 0.9647, + "step": 37250 + }, + { + "epoch": 2.4, + "learning_rate": 1.2027952510788124e-05, + "loss": 0.8808, + "step": 37251 + }, + { + "epoch": 2.4, + "learning_rate": 1.2027737821765175e-05, + "loss": 0.8857, + "step": 37252 + }, + { + "epoch": 2.4, + "learning_rate": 1.2027523132742225e-05, + "loss": 0.9105, + "step": 37253 + }, + { + "epoch": 2.4, + "learning_rate": 1.2027308443719274e-05, + "loss": 0.9044, + "step": 37254 + }, + { + "epoch": 2.4, + "learning_rate": 1.2027093754696323e-05, + "loss": 0.8735, + "step": 37255 + }, + { + "epoch": 2.4, + "learning_rate": 1.2026879065673374e-05, + "loss": 0.9059, + "step": 37256 + }, + { + "epoch": 2.4, + "learning_rate": 1.2026664376650423e-05, + "loss": 0.8875, + "step": 37257 + }, + { + "epoch": 2.4, + "learning_rate": 1.2026449687627472e-05, + "loss": 0.8545, + "step": 37258 + }, + { + "epoch": 2.4, + "learning_rate": 1.2026234998604523e-05, + "loss": 0.9342, + "step": 37259 + }, + { + "epoch": 2.4, + "learning_rate": 1.2026020309581572e-05, + "loss": 0.8931, + "step": 37260 + }, + { + "epoch": 2.4, + "learning_rate": 1.2025805620558622e-05, + "loss": 0.876, + "step": 37261 + }, + { + "epoch": 2.4, + "learning_rate": 1.2025590931535671e-05, + "loss": 0.9655, + "step": 37262 + }, + { + "epoch": 2.4, + "learning_rate": 1.202537624251272e-05, + "loss": 0.9561, + "step": 37263 + }, + { + "epoch": 2.4, + "learning_rate": 1.202516155348977e-05, + "loss": 0.8695, + "step": 37264 + }, + { + "epoch": 2.4, + "learning_rate": 1.202494686446682e-05, + "loss": 0.8253, + "step": 37265 + }, + { + "epoch": 2.4, + "learning_rate": 1.2024732175443869e-05, + "loss": 0.9276, + "step": 37266 + }, + { + "epoch": 2.4, + "learning_rate": 1.2024517486420921e-05, + "loss": 0.8547, + "step": 37267 + }, + { + "epoch": 2.4, + "learning_rate": 1.202430279739797e-05, + "loss": 0.9217, + "step": 37268 + }, + { + "epoch": 2.4, + "learning_rate": 1.202408810837502e-05, + "loss": 0.9072, + "step": 37269 + }, + { + "epoch": 2.4, + "learning_rate": 1.202387341935207e-05, + "loss": 0.908, + "step": 37270 + }, + { + "epoch": 2.4, + "learning_rate": 1.202365873032912e-05, + "loss": 1.0253, + "step": 37271 + }, + { + "epoch": 2.4, + "learning_rate": 1.202344404130617e-05, + "loss": 0.9028, + "step": 37272 + }, + { + "epoch": 2.4, + "learning_rate": 1.2023229352283218e-05, + "loss": 0.8335, + "step": 37273 + }, + { + "epoch": 2.4, + "learning_rate": 1.2023014663260269e-05, + "loss": 0.9102, + "step": 37274 + }, + { + "epoch": 2.4, + "learning_rate": 1.2022799974237318e-05, + "loss": 0.9048, + "step": 37275 + }, + { + "epoch": 2.4, + "learning_rate": 1.2022585285214367e-05, + "loss": 0.8574, + "step": 37276 + }, + { + "epoch": 2.4, + "learning_rate": 1.2022370596191418e-05, + "loss": 0.9855, + "step": 37277 + }, + { + "epoch": 2.4, + "learning_rate": 1.2022155907168466e-05, + "loss": 0.891, + "step": 37278 + }, + { + "epoch": 2.4, + "learning_rate": 1.2021941218145517e-05, + "loss": 0.8951, + "step": 37279 + }, + { + "epoch": 2.4, + "learning_rate": 1.2021726529122566e-05, + "loss": 0.8768, + "step": 37280 + }, + { + "epoch": 2.4, + "learning_rate": 1.2021511840099615e-05, + "loss": 0.8728, + "step": 37281 + }, + { + "epoch": 2.4, + "learning_rate": 1.2021297151076667e-05, + "loss": 0.8706, + "step": 37282 + }, + { + "epoch": 2.4, + "learning_rate": 1.2021082462053716e-05, + "loss": 0.8646, + "step": 37283 + }, + { + "epoch": 2.4, + "learning_rate": 1.2020867773030767e-05, + "loss": 0.8864, + "step": 37284 + }, + { + "epoch": 2.4, + "learning_rate": 1.2020653084007816e-05, + "loss": 0.9156, + "step": 37285 + }, + { + "epoch": 2.4, + "learning_rate": 1.2020438394984865e-05, + "loss": 1.0254, + "step": 37286 + }, + { + "epoch": 2.4, + "learning_rate": 1.2020223705961916e-05, + "loss": 0.9184, + "step": 37287 + }, + { + "epoch": 2.4, + "learning_rate": 1.2020009016938965e-05, + "loss": 0.9465, + "step": 37288 + }, + { + "epoch": 2.4, + "learning_rate": 1.2019794327916015e-05, + "loss": 0.8867, + "step": 37289 + }, + { + "epoch": 2.4, + "learning_rate": 1.2019579638893064e-05, + "loss": 0.8381, + "step": 37290 + }, + { + "epoch": 2.4, + "learning_rate": 1.2019364949870113e-05, + "loss": 0.937, + "step": 37291 + }, + { + "epoch": 2.4, + "learning_rate": 1.2019150260847164e-05, + "loss": 0.9161, + "step": 37292 + }, + { + "epoch": 2.4, + "learning_rate": 1.2018935571824213e-05, + "loss": 0.8869, + "step": 37293 + }, + { + "epoch": 2.4, + "learning_rate": 1.2018720882801262e-05, + "loss": 0.8953, + "step": 37294 + }, + { + "epoch": 2.4, + "learning_rate": 1.2018506193778312e-05, + "loss": 0.8608, + "step": 37295 + }, + { + "epoch": 2.4, + "learning_rate": 1.2018291504755363e-05, + "loss": 0.8664, + "step": 37296 + }, + { + "epoch": 2.4, + "learning_rate": 1.2018076815732414e-05, + "loss": 0.9075, + "step": 37297 + }, + { + "epoch": 2.4, + "learning_rate": 1.2017862126709463e-05, + "loss": 0.8476, + "step": 37298 + }, + { + "epoch": 2.4, + "learning_rate": 1.2017647437686513e-05, + "loss": 0.8934, + "step": 37299 + }, + { + "epoch": 2.4, + "learning_rate": 1.2017432748663562e-05, + "loss": 0.8726, + "step": 37300 + }, + { + "epoch": 2.4, + "learning_rate": 1.2017218059640611e-05, + "loss": 0.9741, + "step": 37301 + }, + { + "epoch": 2.4, + "learning_rate": 1.2017003370617662e-05, + "loss": 0.8599, + "step": 37302 + }, + { + "epoch": 2.4, + "learning_rate": 1.2016788681594711e-05, + "loss": 0.8886, + "step": 37303 + }, + { + "epoch": 2.4, + "learning_rate": 1.201657399257176e-05, + "loss": 0.9209, + "step": 37304 + }, + { + "epoch": 2.4, + "learning_rate": 1.201635930354881e-05, + "loss": 0.8509, + "step": 37305 + }, + { + "epoch": 2.4, + "learning_rate": 1.201614461452586e-05, + "loss": 0.9486, + "step": 37306 + }, + { + "epoch": 2.4, + "learning_rate": 1.201592992550291e-05, + "loss": 0.9182, + "step": 37307 + }, + { + "epoch": 2.4, + "learning_rate": 1.2015715236479959e-05, + "loss": 0.935, + "step": 37308 + }, + { + "epoch": 2.4, + "learning_rate": 1.2015500547457008e-05, + "loss": 0.9266, + "step": 37309 + }, + { + "epoch": 2.4, + "learning_rate": 1.2015285858434059e-05, + "loss": 0.8678, + "step": 37310 + }, + { + "epoch": 2.4, + "learning_rate": 1.201507116941111e-05, + "loss": 0.8817, + "step": 37311 + }, + { + "epoch": 2.4, + "learning_rate": 1.201485648038816e-05, + "loss": 0.9392, + "step": 37312 + }, + { + "epoch": 2.4, + "learning_rate": 1.2014641791365209e-05, + "loss": 0.8486, + "step": 37313 + }, + { + "epoch": 2.4, + "learning_rate": 1.2014427102342258e-05, + "loss": 0.8825, + "step": 37314 + }, + { + "epoch": 2.4, + "learning_rate": 1.2014212413319309e-05, + "loss": 0.9006, + "step": 37315 + }, + { + "epoch": 2.4, + "learning_rate": 1.2013997724296358e-05, + "loss": 0.95, + "step": 37316 + }, + { + "epoch": 2.4, + "learning_rate": 1.2013783035273408e-05, + "loss": 0.9887, + "step": 37317 + }, + { + "epoch": 2.4, + "learning_rate": 1.2013568346250457e-05, + "loss": 0.9502, + "step": 37318 + }, + { + "epoch": 2.4, + "learning_rate": 1.2013353657227506e-05, + "loss": 0.9237, + "step": 37319 + }, + { + "epoch": 2.4, + "learning_rate": 1.2013138968204557e-05, + "loss": 0.9079, + "step": 37320 + }, + { + "epoch": 2.4, + "learning_rate": 1.2012924279181606e-05, + "loss": 0.891, + "step": 37321 + }, + { + "epoch": 2.4, + "learning_rate": 1.2012709590158655e-05, + "loss": 0.8837, + "step": 37322 + }, + { + "epoch": 2.4, + "learning_rate": 1.2012494901135705e-05, + "loss": 0.9103, + "step": 37323 + }, + { + "epoch": 2.4, + "learning_rate": 1.2012280212112754e-05, + "loss": 0.951, + "step": 37324 + }, + { + "epoch": 2.4, + "learning_rate": 1.2012065523089807e-05, + "loss": 0.9243, + "step": 37325 + }, + { + "epoch": 2.4, + "learning_rate": 1.2011850834066856e-05, + "loss": 0.9408, + "step": 37326 + }, + { + "epoch": 2.4, + "learning_rate": 1.2011636145043906e-05, + "loss": 0.8674, + "step": 37327 + }, + { + "epoch": 2.4, + "learning_rate": 1.2011421456020955e-05, + "loss": 0.9106, + "step": 37328 + }, + { + "epoch": 2.4, + "learning_rate": 1.2011206766998004e-05, + "loss": 0.9679, + "step": 37329 + }, + { + "epoch": 2.4, + "learning_rate": 1.2010992077975055e-05, + "loss": 0.8788, + "step": 37330 + }, + { + "epoch": 2.4, + "learning_rate": 1.2010777388952104e-05, + "loss": 0.923, + "step": 37331 + }, + { + "epoch": 2.4, + "learning_rate": 1.2010562699929153e-05, + "loss": 0.8789, + "step": 37332 + }, + { + "epoch": 2.4, + "learning_rate": 1.2010348010906204e-05, + "loss": 0.9315, + "step": 37333 + }, + { + "epoch": 2.4, + "learning_rate": 1.2010133321883253e-05, + "loss": 0.8493, + "step": 37334 + }, + { + "epoch": 2.4, + "learning_rate": 1.2009918632860302e-05, + "loss": 0.8686, + "step": 37335 + }, + { + "epoch": 2.4, + "learning_rate": 1.2009703943837352e-05, + "loss": 0.9365, + "step": 37336 + }, + { + "epoch": 2.4, + "learning_rate": 1.2009489254814401e-05, + "loss": 0.878, + "step": 37337 + }, + { + "epoch": 2.4, + "learning_rate": 1.2009274565791452e-05, + "loss": 0.9024, + "step": 37338 + }, + { + "epoch": 2.4, + "learning_rate": 1.20090598767685e-05, + "loss": 0.876, + "step": 37339 + }, + { + "epoch": 2.4, + "learning_rate": 1.2008845187745553e-05, + "loss": 0.8913, + "step": 37340 + }, + { + "epoch": 2.4, + "learning_rate": 1.2008630498722602e-05, + "loss": 0.8822, + "step": 37341 + }, + { + "epoch": 2.4, + "learning_rate": 1.2008415809699651e-05, + "loss": 0.8011, + "step": 37342 + }, + { + "epoch": 2.4, + "learning_rate": 1.2008201120676702e-05, + "loss": 0.8837, + "step": 37343 + }, + { + "epoch": 2.4, + "learning_rate": 1.200798643165375e-05, + "loss": 0.8735, + "step": 37344 + }, + { + "epoch": 2.4, + "learning_rate": 1.20077717426308e-05, + "loss": 0.898, + "step": 37345 + }, + { + "epoch": 2.4, + "learning_rate": 1.200755705360785e-05, + "loss": 0.9933, + "step": 37346 + }, + { + "epoch": 2.4, + "learning_rate": 1.20073423645849e-05, + "loss": 0.8948, + "step": 37347 + }, + { + "epoch": 2.4, + "learning_rate": 1.200712767556195e-05, + "loss": 0.9108, + "step": 37348 + }, + { + "epoch": 2.4, + "learning_rate": 1.2006912986538999e-05, + "loss": 0.9118, + "step": 37349 + }, + { + "epoch": 2.4, + "learning_rate": 1.2006698297516048e-05, + "loss": 0.8902, + "step": 37350 + }, + { + "epoch": 2.4, + "learning_rate": 1.2006483608493099e-05, + "loss": 0.9184, + "step": 37351 + }, + { + "epoch": 2.4, + "learning_rate": 1.2006268919470147e-05, + "loss": 0.9142, + "step": 37352 + }, + { + "epoch": 2.4, + "learning_rate": 1.2006054230447196e-05, + "loss": 0.9168, + "step": 37353 + }, + { + "epoch": 2.4, + "learning_rate": 1.2005839541424249e-05, + "loss": 0.8873, + "step": 37354 + }, + { + "epoch": 2.4, + "learning_rate": 1.2005624852401298e-05, + "loss": 0.9152, + "step": 37355 + }, + { + "epoch": 2.4, + "learning_rate": 1.2005410163378348e-05, + "loss": 0.8077, + "step": 37356 + }, + { + "epoch": 2.4, + "learning_rate": 1.2005195474355397e-05, + "loss": 0.8877, + "step": 37357 + }, + { + "epoch": 2.4, + "learning_rate": 1.2004980785332448e-05, + "loss": 0.9582, + "step": 37358 + }, + { + "epoch": 2.4, + "learning_rate": 1.2004766096309497e-05, + "loss": 0.8932, + "step": 37359 + }, + { + "epoch": 2.4, + "learning_rate": 1.2004551407286546e-05, + "loss": 0.8694, + "step": 37360 + }, + { + "epoch": 2.4, + "learning_rate": 1.2004336718263597e-05, + "loss": 0.9034, + "step": 37361 + }, + { + "epoch": 2.4, + "learning_rate": 1.2004122029240646e-05, + "loss": 0.8982, + "step": 37362 + }, + { + "epoch": 2.4, + "learning_rate": 1.2003907340217695e-05, + "loss": 0.8865, + "step": 37363 + }, + { + "epoch": 2.4, + "learning_rate": 1.2003692651194745e-05, + "loss": 0.9628, + "step": 37364 + }, + { + "epoch": 2.4, + "learning_rate": 1.2003477962171794e-05, + "loss": 0.9556, + "step": 37365 + }, + { + "epoch": 2.4, + "learning_rate": 1.2003263273148845e-05, + "loss": 0.8845, + "step": 37366 + }, + { + "epoch": 2.4, + "learning_rate": 1.2003048584125894e-05, + "loss": 0.8678, + "step": 37367 + }, + { + "epoch": 2.4, + "learning_rate": 1.2002833895102943e-05, + "loss": 0.9051, + "step": 37368 + }, + { + "epoch": 2.4, + "learning_rate": 1.2002619206079995e-05, + "loss": 0.891, + "step": 37369 + }, + { + "epoch": 2.4, + "learning_rate": 1.2002404517057044e-05, + "loss": 0.8636, + "step": 37370 + }, + { + "epoch": 2.4, + "learning_rate": 1.2002189828034095e-05, + "loss": 0.9888, + "step": 37371 + }, + { + "epoch": 2.4, + "learning_rate": 1.2001975139011144e-05, + "loss": 0.9658, + "step": 37372 + }, + { + "epoch": 2.4, + "learning_rate": 1.2001760449988193e-05, + "loss": 0.877, + "step": 37373 + }, + { + "epoch": 2.4, + "learning_rate": 1.2001545760965243e-05, + "loss": 0.913, + "step": 37374 + }, + { + "epoch": 2.4, + "learning_rate": 1.2001331071942292e-05, + "loss": 0.9209, + "step": 37375 + }, + { + "epoch": 2.4, + "learning_rate": 1.2001116382919343e-05, + "loss": 0.8296, + "step": 37376 + }, + { + "epoch": 2.4, + "learning_rate": 1.2000901693896392e-05, + "loss": 0.9093, + "step": 37377 + }, + { + "epoch": 2.4, + "learning_rate": 1.2000687004873441e-05, + "loss": 0.9354, + "step": 37378 + }, + { + "epoch": 2.4, + "learning_rate": 1.2000472315850492e-05, + "loss": 0.8834, + "step": 37379 + }, + { + "epoch": 2.4, + "learning_rate": 1.200025762682754e-05, + "loss": 0.9035, + "step": 37380 + }, + { + "epoch": 2.41, + "learning_rate": 1.200004293780459e-05, + "loss": 0.8887, + "step": 37381 + }, + { + "epoch": 2.41, + "learning_rate": 1.199982824878164e-05, + "loss": 0.9274, + "step": 37382 + }, + { + "epoch": 2.41, + "learning_rate": 1.199961355975869e-05, + "loss": 0.9135, + "step": 37383 + }, + { + "epoch": 2.41, + "learning_rate": 1.1999398870735741e-05, + "loss": 0.8836, + "step": 37384 + }, + { + "epoch": 2.41, + "learning_rate": 1.199918418171279e-05, + "loss": 0.9038, + "step": 37385 + }, + { + "epoch": 2.41, + "learning_rate": 1.1998969492689841e-05, + "loss": 0.8387, + "step": 37386 + }, + { + "epoch": 2.41, + "learning_rate": 1.199875480366689e-05, + "loss": 0.7909, + "step": 37387 + }, + { + "epoch": 2.41, + "learning_rate": 1.1998540114643939e-05, + "loss": 0.841, + "step": 37388 + }, + { + "epoch": 2.41, + "learning_rate": 1.199832542562099e-05, + "loss": 0.8512, + "step": 37389 + }, + { + "epoch": 2.41, + "learning_rate": 1.1998110736598039e-05, + "loss": 0.9377, + "step": 37390 + }, + { + "epoch": 2.41, + "learning_rate": 1.1997896047575088e-05, + "loss": 0.9135, + "step": 37391 + }, + { + "epoch": 2.41, + "learning_rate": 1.1997681358552138e-05, + "loss": 0.79, + "step": 37392 + }, + { + "epoch": 2.41, + "learning_rate": 1.1997466669529187e-05, + "loss": 0.8874, + "step": 37393 + }, + { + "epoch": 2.41, + "learning_rate": 1.1997251980506238e-05, + "loss": 0.9368, + "step": 37394 + }, + { + "epoch": 2.41, + "learning_rate": 1.1997037291483287e-05, + "loss": 0.9173, + "step": 37395 + }, + { + "epoch": 2.41, + "learning_rate": 1.1996822602460336e-05, + "loss": 0.8779, + "step": 37396 + }, + { + "epoch": 2.41, + "learning_rate": 1.1996607913437386e-05, + "loss": 0.8819, + "step": 37397 + }, + { + "epoch": 2.41, + "learning_rate": 1.1996393224414437e-05, + "loss": 0.9064, + "step": 37398 + }, + { + "epoch": 2.41, + "learning_rate": 1.1996178535391488e-05, + "loss": 0.8588, + "step": 37399 + }, + { + "epoch": 2.41, + "learning_rate": 1.1995963846368537e-05, + "loss": 0.8378, + "step": 37400 + }, + { + "epoch": 2.41, + "learning_rate": 1.1995749157345586e-05, + "loss": 0.8675, + "step": 37401 + }, + { + "epoch": 2.41, + "learning_rate": 1.1995534468322636e-05, + "loss": 0.8109, + "step": 37402 + }, + { + "epoch": 2.41, + "learning_rate": 1.1995319779299685e-05, + "loss": 0.8874, + "step": 37403 + }, + { + "epoch": 2.41, + "learning_rate": 1.1995105090276736e-05, + "loss": 0.9297, + "step": 37404 + }, + { + "epoch": 2.41, + "learning_rate": 1.1994890401253785e-05, + "loss": 0.8589, + "step": 37405 + }, + { + "epoch": 2.41, + "learning_rate": 1.1994675712230834e-05, + "loss": 0.9654, + "step": 37406 + }, + { + "epoch": 2.41, + "learning_rate": 1.1994461023207885e-05, + "loss": 0.9282, + "step": 37407 + }, + { + "epoch": 2.41, + "learning_rate": 1.1994246334184934e-05, + "loss": 0.8633, + "step": 37408 + }, + { + "epoch": 2.41, + "learning_rate": 1.1994031645161983e-05, + "loss": 0.9072, + "step": 37409 + }, + { + "epoch": 2.41, + "learning_rate": 1.1993816956139033e-05, + "loss": 0.8918, + "step": 37410 + }, + { + "epoch": 2.41, + "learning_rate": 1.1993602267116082e-05, + "loss": 0.8916, + "step": 37411 + }, + { + "epoch": 2.41, + "learning_rate": 1.1993387578093135e-05, + "loss": 0.9606, + "step": 37412 + }, + { + "epoch": 2.41, + "learning_rate": 1.1993172889070183e-05, + "loss": 0.9057, + "step": 37413 + }, + { + "epoch": 2.41, + "learning_rate": 1.1992958200047234e-05, + "loss": 0.837, + "step": 37414 + }, + { + "epoch": 2.41, + "learning_rate": 1.1992743511024283e-05, + "loss": 0.922, + "step": 37415 + }, + { + "epoch": 2.41, + "learning_rate": 1.1992528822001332e-05, + "loss": 0.9186, + "step": 37416 + }, + { + "epoch": 2.41, + "learning_rate": 1.1992314132978383e-05, + "loss": 0.8149, + "step": 37417 + }, + { + "epoch": 2.41, + "learning_rate": 1.1992099443955432e-05, + "loss": 0.9029, + "step": 37418 + }, + { + "epoch": 2.41, + "learning_rate": 1.199188475493248e-05, + "loss": 0.9351, + "step": 37419 + }, + { + "epoch": 2.41, + "learning_rate": 1.1991670065909531e-05, + "loss": 0.9277, + "step": 37420 + }, + { + "epoch": 2.41, + "learning_rate": 1.199145537688658e-05, + "loss": 0.9481, + "step": 37421 + }, + { + "epoch": 2.41, + "learning_rate": 1.199124068786363e-05, + "loss": 0.8677, + "step": 37422 + }, + { + "epoch": 2.41, + "learning_rate": 1.199102599884068e-05, + "loss": 0.9491, + "step": 37423 + }, + { + "epoch": 2.41, + "learning_rate": 1.1990811309817729e-05, + "loss": 0.8585, + "step": 37424 + }, + { + "epoch": 2.41, + "learning_rate": 1.199059662079478e-05, + "loss": 0.9927, + "step": 37425 + }, + { + "epoch": 2.41, + "learning_rate": 1.1990381931771828e-05, + "loss": 0.9526, + "step": 37426 + }, + { + "epoch": 2.41, + "learning_rate": 1.199016724274888e-05, + "loss": 0.8888, + "step": 37427 + }, + { + "epoch": 2.41, + "learning_rate": 1.198995255372593e-05, + "loss": 0.8515, + "step": 37428 + }, + { + "epoch": 2.41, + "learning_rate": 1.1989737864702979e-05, + "loss": 0.9097, + "step": 37429 + }, + { + "epoch": 2.41, + "learning_rate": 1.198952317568003e-05, + "loss": 0.9205, + "step": 37430 + }, + { + "epoch": 2.41, + "learning_rate": 1.1989308486657078e-05, + "loss": 0.9056, + "step": 37431 + }, + { + "epoch": 2.41, + "learning_rate": 1.1989093797634127e-05, + "loss": 0.8424, + "step": 37432 + }, + { + "epoch": 2.41, + "learning_rate": 1.1988879108611178e-05, + "loss": 0.8589, + "step": 37433 + }, + { + "epoch": 2.41, + "learning_rate": 1.1988664419588227e-05, + "loss": 0.8028, + "step": 37434 + }, + { + "epoch": 2.41, + "learning_rate": 1.1988449730565278e-05, + "loss": 0.9175, + "step": 37435 + }, + { + "epoch": 2.41, + "learning_rate": 1.1988235041542327e-05, + "loss": 0.9272, + "step": 37436 + }, + { + "epoch": 2.41, + "learning_rate": 1.1988020352519376e-05, + "loss": 0.8773, + "step": 37437 + }, + { + "epoch": 2.41, + "learning_rate": 1.1987805663496426e-05, + "loss": 0.8774, + "step": 37438 + }, + { + "epoch": 2.41, + "learning_rate": 1.1987590974473475e-05, + "loss": 0.8694, + "step": 37439 + }, + { + "epoch": 2.41, + "learning_rate": 1.1987376285450524e-05, + "loss": 0.8705, + "step": 37440 + }, + { + "epoch": 2.41, + "learning_rate": 1.1987161596427577e-05, + "loss": 0.8706, + "step": 37441 + }, + { + "epoch": 2.41, + "learning_rate": 1.1986946907404625e-05, + "loss": 0.872, + "step": 37442 + }, + { + "epoch": 2.41, + "learning_rate": 1.1986732218381676e-05, + "loss": 0.9677, + "step": 37443 + }, + { + "epoch": 2.41, + "learning_rate": 1.1986517529358725e-05, + "loss": 0.9135, + "step": 37444 + }, + { + "epoch": 2.41, + "learning_rate": 1.1986302840335776e-05, + "loss": 0.8379, + "step": 37445 + }, + { + "epoch": 2.41, + "learning_rate": 1.1986088151312825e-05, + "loss": 0.9399, + "step": 37446 + }, + { + "epoch": 2.41, + "learning_rate": 1.1985873462289874e-05, + "loss": 0.8697, + "step": 37447 + }, + { + "epoch": 2.41, + "learning_rate": 1.1985658773266924e-05, + "loss": 0.9042, + "step": 37448 + }, + { + "epoch": 2.41, + "learning_rate": 1.1985444084243973e-05, + "loss": 0.9716, + "step": 37449 + }, + { + "epoch": 2.41, + "learning_rate": 1.1985229395221022e-05, + "loss": 0.86, + "step": 37450 + }, + { + "epoch": 2.41, + "learning_rate": 1.1985014706198073e-05, + "loss": 0.9337, + "step": 37451 + }, + { + "epoch": 2.41, + "learning_rate": 1.1984800017175122e-05, + "loss": 0.902, + "step": 37452 + }, + { + "epoch": 2.41, + "learning_rate": 1.1984585328152173e-05, + "loss": 0.9189, + "step": 37453 + }, + { + "epoch": 2.41, + "learning_rate": 1.1984370639129222e-05, + "loss": 0.9104, + "step": 37454 + }, + { + "epoch": 2.41, + "learning_rate": 1.198415595010627e-05, + "loss": 0.8873, + "step": 37455 + }, + { + "epoch": 2.41, + "learning_rate": 1.1983941261083323e-05, + "loss": 0.8833, + "step": 37456 + }, + { + "epoch": 2.41, + "learning_rate": 1.1983726572060372e-05, + "loss": 0.8964, + "step": 37457 + }, + { + "epoch": 2.41, + "learning_rate": 1.1983511883037422e-05, + "loss": 0.9244, + "step": 37458 + }, + { + "epoch": 2.41, + "learning_rate": 1.1983297194014471e-05, + "loss": 0.9343, + "step": 37459 + }, + { + "epoch": 2.41, + "learning_rate": 1.198308250499152e-05, + "loss": 0.8896, + "step": 37460 + }, + { + "epoch": 2.41, + "learning_rate": 1.1982867815968571e-05, + "loss": 0.9341, + "step": 37461 + }, + { + "epoch": 2.41, + "learning_rate": 1.198265312694562e-05, + "loss": 0.9449, + "step": 37462 + }, + { + "epoch": 2.41, + "learning_rate": 1.198243843792267e-05, + "loss": 0.9244, + "step": 37463 + }, + { + "epoch": 2.41, + "learning_rate": 1.198222374889972e-05, + "loss": 0.9064, + "step": 37464 + }, + { + "epoch": 2.41, + "learning_rate": 1.1982009059876769e-05, + "loss": 0.9588, + "step": 37465 + }, + { + "epoch": 2.41, + "learning_rate": 1.198179437085382e-05, + "loss": 0.9305, + "step": 37466 + }, + { + "epoch": 2.41, + "learning_rate": 1.1981579681830868e-05, + "loss": 0.9473, + "step": 37467 + }, + { + "epoch": 2.41, + "learning_rate": 1.1981364992807917e-05, + "loss": 1.0069, + "step": 37468 + }, + { + "epoch": 2.41, + "learning_rate": 1.1981150303784968e-05, + "loss": 0.9025, + "step": 37469 + }, + { + "epoch": 2.41, + "learning_rate": 1.1980935614762019e-05, + "loss": 0.8395, + "step": 37470 + }, + { + "epoch": 2.41, + "learning_rate": 1.198072092573907e-05, + "loss": 0.8879, + "step": 37471 + }, + { + "epoch": 2.41, + "learning_rate": 1.1980506236716118e-05, + "loss": 0.9223, + "step": 37472 + }, + { + "epoch": 2.41, + "learning_rate": 1.1980291547693169e-05, + "loss": 0.9068, + "step": 37473 + }, + { + "epoch": 2.41, + "learning_rate": 1.1980076858670218e-05, + "loss": 0.9646, + "step": 37474 + }, + { + "epoch": 2.41, + "learning_rate": 1.1979862169647267e-05, + "loss": 0.846, + "step": 37475 + }, + { + "epoch": 2.41, + "learning_rate": 1.1979647480624317e-05, + "loss": 0.8578, + "step": 37476 + }, + { + "epoch": 2.41, + "learning_rate": 1.1979432791601366e-05, + "loss": 0.9481, + "step": 37477 + }, + { + "epoch": 2.41, + "learning_rate": 1.1979218102578415e-05, + "loss": 0.8585, + "step": 37478 + }, + { + "epoch": 2.41, + "learning_rate": 1.1979003413555466e-05, + "loss": 0.8684, + "step": 37479 + }, + { + "epoch": 2.41, + "learning_rate": 1.1978788724532515e-05, + "loss": 0.9026, + "step": 37480 + }, + { + "epoch": 2.41, + "learning_rate": 1.1978574035509566e-05, + "loss": 1.0017, + "step": 37481 + }, + { + "epoch": 2.41, + "learning_rate": 1.1978359346486615e-05, + "loss": 0.8827, + "step": 37482 + }, + { + "epoch": 2.41, + "learning_rate": 1.1978144657463664e-05, + "loss": 0.8512, + "step": 37483 + }, + { + "epoch": 2.41, + "learning_rate": 1.1977929968440714e-05, + "loss": 0.8887, + "step": 37484 + }, + { + "epoch": 2.41, + "learning_rate": 1.1977715279417765e-05, + "loss": 0.9323, + "step": 37485 + }, + { + "epoch": 2.41, + "learning_rate": 1.1977500590394816e-05, + "loss": 0.9264, + "step": 37486 + }, + { + "epoch": 2.41, + "learning_rate": 1.1977285901371864e-05, + "loss": 0.9605, + "step": 37487 + }, + { + "epoch": 2.41, + "learning_rate": 1.1977071212348913e-05, + "loss": 0.9121, + "step": 37488 + }, + { + "epoch": 2.41, + "learning_rate": 1.1976856523325964e-05, + "loss": 0.872, + "step": 37489 + }, + { + "epoch": 2.41, + "learning_rate": 1.1976641834303013e-05, + "loss": 0.8401, + "step": 37490 + }, + { + "epoch": 2.41, + "learning_rate": 1.1976427145280064e-05, + "loss": 0.9181, + "step": 37491 + }, + { + "epoch": 2.41, + "learning_rate": 1.1976212456257113e-05, + "loss": 0.8449, + "step": 37492 + }, + { + "epoch": 2.41, + "learning_rate": 1.1975997767234162e-05, + "loss": 0.8051, + "step": 37493 + }, + { + "epoch": 2.41, + "learning_rate": 1.1975783078211212e-05, + "loss": 0.9124, + "step": 37494 + }, + { + "epoch": 2.41, + "learning_rate": 1.1975568389188261e-05, + "loss": 0.8657, + "step": 37495 + }, + { + "epoch": 2.41, + "learning_rate": 1.197535370016531e-05, + "loss": 0.897, + "step": 37496 + }, + { + "epoch": 2.41, + "learning_rate": 1.1975139011142361e-05, + "loss": 0.8376, + "step": 37497 + }, + { + "epoch": 2.41, + "learning_rate": 1.197492432211941e-05, + "loss": 0.839, + "step": 37498 + }, + { + "epoch": 2.41, + "learning_rate": 1.1974709633096462e-05, + "loss": 0.9368, + "step": 37499 + }, + { + "epoch": 2.41, + "learning_rate": 1.1974494944073511e-05, + "loss": 0.9079, + "step": 37500 + }, + { + "epoch": 2.41, + "learning_rate": 1.1974280255050562e-05, + "loss": 0.8633, + "step": 37501 + }, + { + "epoch": 2.41, + "learning_rate": 1.197406556602761e-05, + "loss": 0.8905, + "step": 37502 + }, + { + "epoch": 2.41, + "learning_rate": 1.197385087700466e-05, + "loss": 0.9482, + "step": 37503 + }, + { + "epoch": 2.41, + "learning_rate": 1.197363618798171e-05, + "loss": 0.9174, + "step": 37504 + }, + { + "epoch": 2.41, + "learning_rate": 1.197342149895876e-05, + "loss": 0.8417, + "step": 37505 + }, + { + "epoch": 2.41, + "learning_rate": 1.1973206809935808e-05, + "loss": 0.8726, + "step": 37506 + }, + { + "epoch": 2.41, + "learning_rate": 1.1972992120912859e-05, + "loss": 0.8783, + "step": 37507 + }, + { + "epoch": 2.41, + "learning_rate": 1.1972777431889908e-05, + "loss": 0.8925, + "step": 37508 + }, + { + "epoch": 2.41, + "learning_rate": 1.1972562742866957e-05, + "loss": 0.9308, + "step": 37509 + }, + { + "epoch": 2.41, + "learning_rate": 1.1972348053844008e-05, + "loss": 0.9232, + "step": 37510 + }, + { + "epoch": 2.41, + "learning_rate": 1.1972133364821057e-05, + "loss": 0.8705, + "step": 37511 + }, + { + "epoch": 2.41, + "learning_rate": 1.1971918675798107e-05, + "loss": 0.8528, + "step": 37512 + }, + { + "epoch": 2.41, + "learning_rate": 1.1971703986775156e-05, + "loss": 0.8675, + "step": 37513 + }, + { + "epoch": 2.41, + "learning_rate": 1.1971489297752209e-05, + "loss": 0.861, + "step": 37514 + }, + { + "epoch": 2.41, + "learning_rate": 1.1971274608729258e-05, + "loss": 0.897, + "step": 37515 + }, + { + "epoch": 2.41, + "learning_rate": 1.1971059919706306e-05, + "loss": 0.8711, + "step": 37516 + }, + { + "epoch": 2.41, + "learning_rate": 1.1970845230683357e-05, + "loss": 0.9073, + "step": 37517 + }, + { + "epoch": 2.41, + "learning_rate": 1.1970630541660406e-05, + "loss": 0.8941, + "step": 37518 + }, + { + "epoch": 2.41, + "learning_rate": 1.1970415852637455e-05, + "loss": 0.842, + "step": 37519 + }, + { + "epoch": 2.41, + "learning_rate": 1.1970201163614506e-05, + "loss": 0.9413, + "step": 37520 + }, + { + "epoch": 2.41, + "learning_rate": 1.1969986474591555e-05, + "loss": 0.9576, + "step": 37521 + }, + { + "epoch": 2.41, + "learning_rate": 1.1969771785568605e-05, + "loss": 0.937, + "step": 37522 + }, + { + "epoch": 2.41, + "learning_rate": 1.1969557096545654e-05, + "loss": 0.8937, + "step": 37523 + }, + { + "epoch": 2.41, + "learning_rate": 1.1969342407522703e-05, + "loss": 0.9362, + "step": 37524 + }, + { + "epoch": 2.41, + "learning_rate": 1.1969127718499754e-05, + "loss": 0.9517, + "step": 37525 + }, + { + "epoch": 2.41, + "learning_rate": 1.1968913029476803e-05, + "loss": 0.8745, + "step": 37526 + }, + { + "epoch": 2.41, + "learning_rate": 1.1968698340453852e-05, + "loss": 0.8911, + "step": 37527 + }, + { + "epoch": 2.41, + "learning_rate": 1.1968483651430904e-05, + "loss": 0.8953, + "step": 37528 + }, + { + "epoch": 2.41, + "learning_rate": 1.1968268962407953e-05, + "loss": 0.8814, + "step": 37529 + }, + { + "epoch": 2.41, + "learning_rate": 1.1968054273385004e-05, + "loss": 0.9241, + "step": 37530 + }, + { + "epoch": 2.41, + "learning_rate": 1.1967839584362053e-05, + "loss": 0.9653, + "step": 37531 + }, + { + "epoch": 2.41, + "learning_rate": 1.1967624895339103e-05, + "loss": 0.9011, + "step": 37532 + }, + { + "epoch": 2.41, + "learning_rate": 1.1967410206316152e-05, + "loss": 0.9337, + "step": 37533 + }, + { + "epoch": 2.41, + "learning_rate": 1.1967195517293201e-05, + "loss": 0.9178, + "step": 37534 + }, + { + "epoch": 2.41, + "learning_rate": 1.1966980828270252e-05, + "loss": 0.9209, + "step": 37535 + }, + { + "epoch": 2.41, + "learning_rate": 1.1966766139247301e-05, + "loss": 1.0045, + "step": 37536 + }, + { + "epoch": 2.42, + "learning_rate": 1.196655145022435e-05, + "loss": 0.8667, + "step": 37537 + }, + { + "epoch": 2.42, + "learning_rate": 1.19663367612014e-05, + "loss": 0.8706, + "step": 37538 + }, + { + "epoch": 2.42, + "learning_rate": 1.196612207217845e-05, + "loss": 0.8556, + "step": 37539 + }, + { + "epoch": 2.42, + "learning_rate": 1.19659073831555e-05, + "loss": 0.9775, + "step": 37540 + }, + { + "epoch": 2.42, + "learning_rate": 1.196569269413255e-05, + "loss": 0.8993, + "step": 37541 + }, + { + "epoch": 2.42, + "learning_rate": 1.1965478005109598e-05, + "loss": 0.9183, + "step": 37542 + }, + { + "epoch": 2.42, + "learning_rate": 1.196526331608665e-05, + "loss": 0.9099, + "step": 37543 + }, + { + "epoch": 2.42, + "learning_rate": 1.19650486270637e-05, + "loss": 0.9099, + "step": 37544 + }, + { + "epoch": 2.42, + "learning_rate": 1.196483393804075e-05, + "loss": 0.8873, + "step": 37545 + }, + { + "epoch": 2.42, + "learning_rate": 1.1964619249017799e-05, + "loss": 0.8691, + "step": 37546 + }, + { + "epoch": 2.42, + "learning_rate": 1.1964404559994848e-05, + "loss": 0.8253, + "step": 37547 + }, + { + "epoch": 2.42, + "learning_rate": 1.1964189870971899e-05, + "loss": 0.9141, + "step": 37548 + }, + { + "epoch": 2.42, + "learning_rate": 1.1963975181948948e-05, + "loss": 0.9027, + "step": 37549 + }, + { + "epoch": 2.42, + "learning_rate": 1.1963760492925998e-05, + "loss": 0.9222, + "step": 37550 + }, + { + "epoch": 2.42, + "learning_rate": 1.1963545803903047e-05, + "loss": 0.8724, + "step": 37551 + }, + { + "epoch": 2.42, + "learning_rate": 1.1963331114880096e-05, + "loss": 0.852, + "step": 37552 + }, + { + "epoch": 2.42, + "learning_rate": 1.1963116425857147e-05, + "loss": 0.9444, + "step": 37553 + }, + { + "epoch": 2.42, + "learning_rate": 1.1962901736834196e-05, + "loss": 0.9016, + "step": 37554 + }, + { + "epoch": 2.42, + "learning_rate": 1.1962687047811245e-05, + "loss": 0.9058, + "step": 37555 + }, + { + "epoch": 2.42, + "learning_rate": 1.1962472358788296e-05, + "loss": 0.8776, + "step": 37556 + }, + { + "epoch": 2.42, + "learning_rate": 1.1962257669765346e-05, + "loss": 0.9062, + "step": 37557 + }, + { + "epoch": 2.42, + "learning_rate": 1.1962042980742397e-05, + "loss": 0.8912, + "step": 37558 + }, + { + "epoch": 2.42, + "learning_rate": 1.1961828291719446e-05, + "loss": 0.9312, + "step": 37559 + }, + { + "epoch": 2.42, + "learning_rate": 1.1961613602696497e-05, + "loss": 0.9277, + "step": 37560 + }, + { + "epoch": 2.42, + "learning_rate": 1.1961398913673545e-05, + "loss": 0.8729, + "step": 37561 + }, + { + "epoch": 2.42, + "learning_rate": 1.1961184224650594e-05, + "loss": 0.8748, + "step": 37562 + }, + { + "epoch": 2.42, + "learning_rate": 1.1960969535627645e-05, + "loss": 0.8953, + "step": 37563 + }, + { + "epoch": 2.42, + "learning_rate": 1.1960754846604694e-05, + "loss": 0.8414, + "step": 37564 + }, + { + "epoch": 2.42, + "learning_rate": 1.1960540157581743e-05, + "loss": 0.8731, + "step": 37565 + }, + { + "epoch": 2.42, + "learning_rate": 1.1960325468558794e-05, + "loss": 0.8694, + "step": 37566 + }, + { + "epoch": 2.42, + "learning_rate": 1.1960110779535843e-05, + "loss": 0.8495, + "step": 37567 + }, + { + "epoch": 2.42, + "learning_rate": 1.1959896090512892e-05, + "loss": 0.8831, + "step": 37568 + }, + { + "epoch": 2.42, + "learning_rate": 1.1959681401489942e-05, + "loss": 0.9549, + "step": 37569 + }, + { + "epoch": 2.42, + "learning_rate": 1.1959466712466991e-05, + "loss": 0.8606, + "step": 37570 + }, + { + "epoch": 2.42, + "learning_rate": 1.1959252023444042e-05, + "loss": 0.9512, + "step": 37571 + }, + { + "epoch": 2.42, + "learning_rate": 1.1959037334421093e-05, + "loss": 0.8422, + "step": 37572 + }, + { + "epoch": 2.42, + "learning_rate": 1.1958822645398143e-05, + "loss": 0.8738, + "step": 37573 + }, + { + "epoch": 2.42, + "learning_rate": 1.1958607956375192e-05, + "loss": 0.8758, + "step": 37574 + }, + { + "epoch": 2.42, + "learning_rate": 1.1958393267352241e-05, + "loss": 0.9239, + "step": 37575 + }, + { + "epoch": 2.42, + "learning_rate": 1.1958178578329292e-05, + "loss": 0.9429, + "step": 37576 + }, + { + "epoch": 2.42, + "learning_rate": 1.195796388930634e-05, + "loss": 0.8843, + "step": 37577 + }, + { + "epoch": 2.42, + "learning_rate": 1.195774920028339e-05, + "loss": 0.8671, + "step": 37578 + }, + { + "epoch": 2.42, + "learning_rate": 1.195753451126044e-05, + "loss": 0.9007, + "step": 37579 + }, + { + "epoch": 2.42, + "learning_rate": 1.195731982223749e-05, + "loss": 0.8521, + "step": 37580 + }, + { + "epoch": 2.42, + "learning_rate": 1.195710513321454e-05, + "loss": 0.8638, + "step": 37581 + }, + { + "epoch": 2.42, + "learning_rate": 1.1956890444191589e-05, + "loss": 1.0048, + "step": 37582 + }, + { + "epoch": 2.42, + "learning_rate": 1.1956675755168638e-05, + "loss": 0.9052, + "step": 37583 + }, + { + "epoch": 2.42, + "learning_rate": 1.1956461066145689e-05, + "loss": 0.9228, + "step": 37584 + }, + { + "epoch": 2.42, + "learning_rate": 1.1956246377122738e-05, + "loss": 0.8288, + "step": 37585 + }, + { + "epoch": 2.42, + "learning_rate": 1.195603168809979e-05, + "loss": 0.9021, + "step": 37586 + }, + { + "epoch": 2.42, + "learning_rate": 1.1955816999076839e-05, + "loss": 0.8263, + "step": 37587 + }, + { + "epoch": 2.42, + "learning_rate": 1.195560231005389e-05, + "loss": 0.8987, + "step": 37588 + }, + { + "epoch": 2.42, + "learning_rate": 1.1955387621030939e-05, + "loss": 0.9284, + "step": 37589 + }, + { + "epoch": 2.42, + "learning_rate": 1.1955172932007987e-05, + "loss": 0.8844, + "step": 37590 + }, + { + "epoch": 2.42, + "learning_rate": 1.1954958242985038e-05, + "loss": 0.9345, + "step": 37591 + }, + { + "epoch": 2.42, + "learning_rate": 1.1954743553962087e-05, + "loss": 0.9056, + "step": 37592 + }, + { + "epoch": 2.42, + "learning_rate": 1.1954528864939136e-05, + "loss": 0.9895, + "step": 37593 + }, + { + "epoch": 2.42, + "learning_rate": 1.1954314175916187e-05, + "loss": 0.8401, + "step": 37594 + }, + { + "epoch": 2.42, + "learning_rate": 1.1954099486893236e-05, + "loss": 0.8071, + "step": 37595 + }, + { + "epoch": 2.42, + "learning_rate": 1.1953884797870285e-05, + "loss": 0.9409, + "step": 37596 + }, + { + "epoch": 2.42, + "learning_rate": 1.1953670108847335e-05, + "loss": 0.8843, + "step": 37597 + }, + { + "epoch": 2.42, + "learning_rate": 1.1953455419824384e-05, + "loss": 0.8656, + "step": 37598 + }, + { + "epoch": 2.42, + "learning_rate": 1.1953240730801435e-05, + "loss": 1.0064, + "step": 37599 + }, + { + "epoch": 2.42, + "learning_rate": 1.1953026041778484e-05, + "loss": 0.8561, + "step": 37600 + }, + { + "epoch": 2.42, + "learning_rate": 1.1952811352755536e-05, + "loss": 0.835, + "step": 37601 + }, + { + "epoch": 2.42, + "learning_rate": 1.1952596663732585e-05, + "loss": 0.8836, + "step": 37602 + }, + { + "epoch": 2.42, + "learning_rate": 1.1952381974709634e-05, + "loss": 0.908, + "step": 37603 + }, + { + "epoch": 2.42, + "learning_rate": 1.1952167285686685e-05, + "loss": 0.9478, + "step": 37604 + }, + { + "epoch": 2.42, + "learning_rate": 1.1951952596663734e-05, + "loss": 0.8999, + "step": 37605 + }, + { + "epoch": 2.42, + "learning_rate": 1.1951737907640783e-05, + "loss": 0.8722, + "step": 37606 + }, + { + "epoch": 2.42, + "learning_rate": 1.1951523218617833e-05, + "loss": 0.9283, + "step": 37607 + }, + { + "epoch": 2.42, + "learning_rate": 1.1951308529594882e-05, + "loss": 0.9065, + "step": 37608 + }, + { + "epoch": 2.42, + "learning_rate": 1.1951093840571933e-05, + "loss": 0.8866, + "step": 37609 + }, + { + "epoch": 2.42, + "learning_rate": 1.1950879151548982e-05, + "loss": 0.8811, + "step": 37610 + }, + { + "epoch": 2.42, + "learning_rate": 1.1950664462526031e-05, + "loss": 0.8767, + "step": 37611 + }, + { + "epoch": 2.42, + "learning_rate": 1.1950449773503082e-05, + "loss": 0.9503, + "step": 37612 + }, + { + "epoch": 2.42, + "learning_rate": 1.195023508448013e-05, + "loss": 0.8143, + "step": 37613 + }, + { + "epoch": 2.42, + "learning_rate": 1.195002039545718e-05, + "loss": 0.9313, + "step": 37614 + }, + { + "epoch": 2.42, + "learning_rate": 1.1949805706434232e-05, + "loss": 0.8841, + "step": 37615 + }, + { + "epoch": 2.42, + "learning_rate": 1.1949591017411281e-05, + "loss": 0.8777, + "step": 37616 + }, + { + "epoch": 2.42, + "learning_rate": 1.1949376328388332e-05, + "loss": 0.9309, + "step": 37617 + }, + { + "epoch": 2.42, + "learning_rate": 1.194916163936538e-05, + "loss": 0.8629, + "step": 37618 + }, + { + "epoch": 2.42, + "learning_rate": 1.1948946950342431e-05, + "loss": 0.8609, + "step": 37619 + }, + { + "epoch": 2.42, + "learning_rate": 1.194873226131948e-05, + "loss": 0.9448, + "step": 37620 + }, + { + "epoch": 2.42, + "learning_rate": 1.1948517572296529e-05, + "loss": 0.9782, + "step": 37621 + }, + { + "epoch": 2.42, + "learning_rate": 1.194830288327358e-05, + "loss": 0.8397, + "step": 37622 + }, + { + "epoch": 2.42, + "learning_rate": 1.1948088194250629e-05, + "loss": 0.8415, + "step": 37623 + }, + { + "epoch": 2.42, + "learning_rate": 1.1947873505227678e-05, + "loss": 0.9946, + "step": 37624 + }, + { + "epoch": 2.42, + "learning_rate": 1.1947658816204728e-05, + "loss": 0.9596, + "step": 37625 + }, + { + "epoch": 2.42, + "learning_rate": 1.1947444127181777e-05, + "loss": 0.8459, + "step": 37626 + }, + { + "epoch": 2.42, + "learning_rate": 1.1947229438158828e-05, + "loss": 0.8591, + "step": 37627 + }, + { + "epoch": 2.42, + "learning_rate": 1.1947014749135877e-05, + "loss": 0.921, + "step": 37628 + }, + { + "epoch": 2.42, + "learning_rate": 1.1946800060112926e-05, + "loss": 0.9009, + "step": 37629 + }, + { + "epoch": 2.42, + "learning_rate": 1.1946585371089978e-05, + "loss": 0.8122, + "step": 37630 + }, + { + "epoch": 2.42, + "learning_rate": 1.1946370682067027e-05, + "loss": 0.9121, + "step": 37631 + }, + { + "epoch": 2.42, + "learning_rate": 1.1946155993044078e-05, + "loss": 0.8801, + "step": 37632 + }, + { + "epoch": 2.42, + "learning_rate": 1.1945941304021127e-05, + "loss": 0.8546, + "step": 37633 + }, + { + "epoch": 2.42, + "learning_rate": 1.1945726614998176e-05, + "loss": 0.8966, + "step": 37634 + }, + { + "epoch": 2.42, + "learning_rate": 1.1945511925975226e-05, + "loss": 0.8544, + "step": 37635 + }, + { + "epoch": 2.42, + "learning_rate": 1.1945297236952275e-05, + "loss": 0.9337, + "step": 37636 + }, + { + "epoch": 2.42, + "learning_rate": 1.1945082547929326e-05, + "loss": 0.8645, + "step": 37637 + }, + { + "epoch": 2.42, + "learning_rate": 1.1944867858906375e-05, + "loss": 0.875, + "step": 37638 + }, + { + "epoch": 2.42, + "learning_rate": 1.1944653169883424e-05, + "loss": 0.9794, + "step": 37639 + }, + { + "epoch": 2.42, + "learning_rate": 1.1944438480860475e-05, + "loss": 0.8207, + "step": 37640 + }, + { + "epoch": 2.42, + "learning_rate": 1.1944223791837524e-05, + "loss": 0.8654, + "step": 37641 + }, + { + "epoch": 2.42, + "learning_rate": 1.1944009102814573e-05, + "loss": 0.8336, + "step": 37642 + }, + { + "epoch": 2.42, + "learning_rate": 1.1943794413791623e-05, + "loss": 0.8702, + "step": 37643 + }, + { + "epoch": 2.42, + "learning_rate": 1.1943579724768672e-05, + "loss": 0.9211, + "step": 37644 + }, + { + "epoch": 2.42, + "learning_rate": 1.1943365035745725e-05, + "loss": 0.8908, + "step": 37645 + }, + { + "epoch": 2.42, + "learning_rate": 1.1943150346722774e-05, + "loss": 0.871, + "step": 37646 + }, + { + "epoch": 2.42, + "learning_rate": 1.1942935657699824e-05, + "loss": 0.9473, + "step": 37647 + }, + { + "epoch": 2.42, + "learning_rate": 1.1942720968676873e-05, + "loss": 0.8749, + "step": 37648 + }, + { + "epoch": 2.42, + "learning_rate": 1.1942506279653922e-05, + "loss": 0.982, + "step": 37649 + }, + { + "epoch": 2.42, + "learning_rate": 1.1942291590630973e-05, + "loss": 0.8456, + "step": 37650 + }, + { + "epoch": 2.42, + "learning_rate": 1.1942076901608022e-05, + "loss": 0.9035, + "step": 37651 + }, + { + "epoch": 2.42, + "learning_rate": 1.194186221258507e-05, + "loss": 0.8836, + "step": 37652 + }, + { + "epoch": 2.42, + "learning_rate": 1.1941647523562121e-05, + "loss": 0.8226, + "step": 37653 + }, + { + "epoch": 2.42, + "learning_rate": 1.194143283453917e-05, + "loss": 0.8754, + "step": 37654 + }, + { + "epoch": 2.42, + "learning_rate": 1.194121814551622e-05, + "loss": 0.977, + "step": 37655 + }, + { + "epoch": 2.42, + "learning_rate": 1.194100345649327e-05, + "loss": 0.8424, + "step": 37656 + }, + { + "epoch": 2.42, + "learning_rate": 1.1940788767470319e-05, + "loss": 0.8928, + "step": 37657 + }, + { + "epoch": 2.42, + "learning_rate": 1.194057407844737e-05, + "loss": 0.8437, + "step": 37658 + }, + { + "epoch": 2.42, + "learning_rate": 1.194035938942442e-05, + "loss": 0.8853, + "step": 37659 + }, + { + "epoch": 2.42, + "learning_rate": 1.1940144700401471e-05, + "loss": 0.9032, + "step": 37660 + }, + { + "epoch": 2.42, + "learning_rate": 1.193993001137852e-05, + "loss": 0.9007, + "step": 37661 + }, + { + "epoch": 2.42, + "learning_rate": 1.1939715322355569e-05, + "loss": 0.9018, + "step": 37662 + }, + { + "epoch": 2.42, + "learning_rate": 1.193950063333262e-05, + "loss": 0.8867, + "step": 37663 + }, + { + "epoch": 2.42, + "learning_rate": 1.1939285944309668e-05, + "loss": 0.8732, + "step": 37664 + }, + { + "epoch": 2.42, + "learning_rate": 1.1939071255286717e-05, + "loss": 0.8758, + "step": 37665 + }, + { + "epoch": 2.42, + "learning_rate": 1.1938856566263768e-05, + "loss": 0.8937, + "step": 37666 + }, + { + "epoch": 2.42, + "learning_rate": 1.1938641877240817e-05, + "loss": 0.8842, + "step": 37667 + }, + { + "epoch": 2.42, + "learning_rate": 1.1938427188217868e-05, + "loss": 0.9361, + "step": 37668 + }, + { + "epoch": 2.42, + "learning_rate": 1.1938212499194917e-05, + "loss": 0.9324, + "step": 37669 + }, + { + "epoch": 2.42, + "learning_rate": 1.1937997810171966e-05, + "loss": 0.889, + "step": 37670 + }, + { + "epoch": 2.42, + "learning_rate": 1.1937783121149016e-05, + "loss": 0.8777, + "step": 37671 + }, + { + "epoch": 2.42, + "learning_rate": 1.1937568432126065e-05, + "loss": 0.9214, + "step": 37672 + }, + { + "epoch": 2.42, + "learning_rate": 1.1937353743103114e-05, + "loss": 0.897, + "step": 37673 + }, + { + "epoch": 2.42, + "learning_rate": 1.1937139054080167e-05, + "loss": 0.8307, + "step": 37674 + }, + { + "epoch": 2.42, + "learning_rate": 1.1936924365057216e-05, + "loss": 0.9594, + "step": 37675 + }, + { + "epoch": 2.42, + "learning_rate": 1.1936709676034266e-05, + "loss": 0.8425, + "step": 37676 + }, + { + "epoch": 2.42, + "learning_rate": 1.1936494987011315e-05, + "loss": 0.8567, + "step": 37677 + }, + { + "epoch": 2.42, + "learning_rate": 1.1936280297988366e-05, + "loss": 0.944, + "step": 37678 + }, + { + "epoch": 2.42, + "learning_rate": 1.1936065608965415e-05, + "loss": 0.8383, + "step": 37679 + }, + { + "epoch": 2.42, + "learning_rate": 1.1935850919942464e-05, + "loss": 0.9505, + "step": 37680 + }, + { + "epoch": 2.42, + "learning_rate": 1.1935636230919514e-05, + "loss": 0.8025, + "step": 37681 + }, + { + "epoch": 2.42, + "learning_rate": 1.1935421541896563e-05, + "loss": 0.8905, + "step": 37682 + }, + { + "epoch": 2.42, + "learning_rate": 1.1935206852873612e-05, + "loss": 0.8402, + "step": 37683 + }, + { + "epoch": 2.42, + "learning_rate": 1.1934992163850663e-05, + "loss": 0.8894, + "step": 37684 + }, + { + "epoch": 2.42, + "learning_rate": 1.1934777474827712e-05, + "loss": 0.9167, + "step": 37685 + }, + { + "epoch": 2.42, + "learning_rate": 1.1934562785804763e-05, + "loss": 0.9082, + "step": 37686 + }, + { + "epoch": 2.42, + "learning_rate": 1.1934348096781812e-05, + "loss": 0.8449, + "step": 37687 + }, + { + "epoch": 2.42, + "learning_rate": 1.1934133407758864e-05, + "loss": 0.9776, + "step": 37688 + }, + { + "epoch": 2.42, + "learning_rate": 1.1933918718735913e-05, + "loss": 0.8871, + "step": 37689 + }, + { + "epoch": 2.42, + "learning_rate": 1.1933704029712962e-05, + "loss": 0.901, + "step": 37690 + }, + { + "epoch": 2.42, + "learning_rate": 1.1933489340690013e-05, + "loss": 0.8977, + "step": 37691 + }, + { + "epoch": 2.43, + "learning_rate": 1.1933274651667062e-05, + "loss": 0.8776, + "step": 37692 + }, + { + "epoch": 2.43, + "learning_rate": 1.193305996264411e-05, + "loss": 0.929, + "step": 37693 + }, + { + "epoch": 2.43, + "learning_rate": 1.1932845273621161e-05, + "loss": 0.8923, + "step": 37694 + }, + { + "epoch": 2.43, + "learning_rate": 1.193263058459821e-05, + "loss": 0.9216, + "step": 37695 + }, + { + "epoch": 2.43, + "learning_rate": 1.193241589557526e-05, + "loss": 0.89, + "step": 37696 + }, + { + "epoch": 2.43, + "learning_rate": 1.193220120655231e-05, + "loss": 0.93, + "step": 37697 + }, + { + "epoch": 2.43, + "learning_rate": 1.1931986517529359e-05, + "loss": 0.8515, + "step": 37698 + }, + { + "epoch": 2.43, + "learning_rate": 1.193177182850641e-05, + "loss": 0.8816, + "step": 37699 + }, + { + "epoch": 2.43, + "learning_rate": 1.1931557139483458e-05, + "loss": 0.9398, + "step": 37700 + }, + { + "epoch": 2.43, + "learning_rate": 1.1931342450460507e-05, + "loss": 0.9032, + "step": 37701 + }, + { + "epoch": 2.43, + "learning_rate": 1.1931127761437558e-05, + "loss": 0.8977, + "step": 37702 + }, + { + "epoch": 2.43, + "learning_rate": 1.1930913072414609e-05, + "loss": 0.9753, + "step": 37703 + }, + { + "epoch": 2.43, + "learning_rate": 1.193069838339166e-05, + "loss": 0.8563, + "step": 37704 + }, + { + "epoch": 2.43, + "learning_rate": 1.1930483694368708e-05, + "loss": 0.9147, + "step": 37705 + }, + { + "epoch": 2.43, + "learning_rate": 1.1930269005345759e-05, + "loss": 0.8857, + "step": 37706 + }, + { + "epoch": 2.43, + "learning_rate": 1.1930054316322808e-05, + "loss": 0.8766, + "step": 37707 + }, + { + "epoch": 2.43, + "learning_rate": 1.1929839627299857e-05, + "loss": 0.91, + "step": 37708 + }, + { + "epoch": 2.43, + "learning_rate": 1.1929624938276907e-05, + "loss": 0.8915, + "step": 37709 + }, + { + "epoch": 2.43, + "learning_rate": 1.1929410249253956e-05, + "loss": 0.8284, + "step": 37710 + }, + { + "epoch": 2.43, + "learning_rate": 1.1929195560231005e-05, + "loss": 0.9181, + "step": 37711 + }, + { + "epoch": 2.43, + "learning_rate": 1.1928980871208056e-05, + "loss": 0.8505, + "step": 37712 + }, + { + "epoch": 2.43, + "learning_rate": 1.1928766182185105e-05, + "loss": 0.8864, + "step": 37713 + }, + { + "epoch": 2.43, + "learning_rate": 1.1928551493162156e-05, + "loss": 0.9385, + "step": 37714 + }, + { + "epoch": 2.43, + "learning_rate": 1.1928336804139205e-05, + "loss": 0.8657, + "step": 37715 + }, + { + "epoch": 2.43, + "learning_rate": 1.1928122115116254e-05, + "loss": 0.9222, + "step": 37716 + }, + { + "epoch": 2.43, + "learning_rate": 1.1927907426093306e-05, + "loss": 0.9169, + "step": 37717 + }, + { + "epoch": 2.43, + "learning_rate": 1.1927692737070355e-05, + "loss": 0.8406, + "step": 37718 + }, + { + "epoch": 2.43, + "learning_rate": 1.1927478048047406e-05, + "loss": 0.8493, + "step": 37719 + }, + { + "epoch": 2.43, + "learning_rate": 1.1927263359024455e-05, + "loss": 0.8794, + "step": 37720 + }, + { + "epoch": 2.43, + "learning_rate": 1.1927048670001504e-05, + "loss": 0.8801, + "step": 37721 + }, + { + "epoch": 2.43, + "learning_rate": 1.1926833980978554e-05, + "loss": 0.7935, + "step": 37722 + }, + { + "epoch": 2.43, + "learning_rate": 1.1926619291955603e-05, + "loss": 0.9358, + "step": 37723 + }, + { + "epoch": 2.43, + "learning_rate": 1.1926404602932654e-05, + "loss": 0.9007, + "step": 37724 + }, + { + "epoch": 2.43, + "learning_rate": 1.1926189913909703e-05, + "loss": 0.8393, + "step": 37725 + }, + { + "epoch": 2.43, + "learning_rate": 1.1925975224886752e-05, + "loss": 0.9297, + "step": 37726 + }, + { + "epoch": 2.43, + "learning_rate": 1.1925760535863802e-05, + "loss": 0.9228, + "step": 37727 + }, + { + "epoch": 2.43, + "learning_rate": 1.1925545846840851e-05, + "loss": 0.9762, + "step": 37728 + }, + { + "epoch": 2.43, + "learning_rate": 1.19253311578179e-05, + "loss": 0.8416, + "step": 37729 + }, + { + "epoch": 2.43, + "learning_rate": 1.1925116468794951e-05, + "loss": 0.962, + "step": 37730 + }, + { + "epoch": 2.43, + "learning_rate": 1.1924901779772e-05, + "loss": 0.8986, + "step": 37731 + }, + { + "epoch": 2.43, + "learning_rate": 1.1924687090749052e-05, + "loss": 0.8729, + "step": 37732 + }, + { + "epoch": 2.43, + "learning_rate": 1.1924472401726101e-05, + "loss": 0.8745, + "step": 37733 + }, + { + "epoch": 2.43, + "learning_rate": 1.1924257712703152e-05, + "loss": 0.8972, + "step": 37734 + }, + { + "epoch": 2.43, + "learning_rate": 1.1924043023680201e-05, + "loss": 0.8652, + "step": 37735 + }, + { + "epoch": 2.43, + "learning_rate": 1.192382833465725e-05, + "loss": 0.8412, + "step": 37736 + }, + { + "epoch": 2.43, + "learning_rate": 1.19236136456343e-05, + "loss": 0.838, + "step": 37737 + }, + { + "epoch": 2.43, + "learning_rate": 1.192339895661135e-05, + "loss": 0.8583, + "step": 37738 + }, + { + "epoch": 2.43, + "learning_rate": 1.1923184267588398e-05, + "loss": 0.8408, + "step": 37739 + }, + { + "epoch": 2.43, + "learning_rate": 1.1922969578565449e-05, + "loss": 0.9376, + "step": 37740 + }, + { + "epoch": 2.43, + "learning_rate": 1.1922754889542498e-05, + "loss": 0.8834, + "step": 37741 + }, + { + "epoch": 2.43, + "learning_rate": 1.1922540200519547e-05, + "loss": 0.894, + "step": 37742 + }, + { + "epoch": 2.43, + "learning_rate": 1.1922325511496598e-05, + "loss": 0.8804, + "step": 37743 + }, + { + "epoch": 2.43, + "learning_rate": 1.1922110822473647e-05, + "loss": 0.8312, + "step": 37744 + }, + { + "epoch": 2.43, + "learning_rate": 1.1921896133450697e-05, + "loss": 0.7972, + "step": 37745 + }, + { + "epoch": 2.43, + "learning_rate": 1.1921681444427748e-05, + "loss": 0.9766, + "step": 37746 + }, + { + "epoch": 2.43, + "learning_rate": 1.1921466755404799e-05, + "loss": 0.8479, + "step": 37747 + }, + { + "epoch": 2.43, + "learning_rate": 1.1921252066381848e-05, + "loss": 0.936, + "step": 37748 + }, + { + "epoch": 2.43, + "learning_rate": 1.1921037377358897e-05, + "loss": 0.9143, + "step": 37749 + }, + { + "epoch": 2.43, + "learning_rate": 1.1920822688335947e-05, + "loss": 0.8342, + "step": 37750 + }, + { + "epoch": 2.43, + "learning_rate": 1.1920607999312996e-05, + "loss": 0.9234, + "step": 37751 + }, + { + "epoch": 2.43, + "learning_rate": 1.1920393310290045e-05, + "loss": 0.8545, + "step": 37752 + }, + { + "epoch": 2.43, + "learning_rate": 1.1920178621267096e-05, + "loss": 0.8241, + "step": 37753 + }, + { + "epoch": 2.43, + "learning_rate": 1.1919963932244145e-05, + "loss": 0.9039, + "step": 37754 + }, + { + "epoch": 2.43, + "learning_rate": 1.1919749243221195e-05, + "loss": 0.9334, + "step": 37755 + }, + { + "epoch": 2.43, + "learning_rate": 1.1919534554198244e-05, + "loss": 0.9357, + "step": 37756 + }, + { + "epoch": 2.43, + "learning_rate": 1.1919319865175293e-05, + "loss": 0.843, + "step": 37757 + }, + { + "epoch": 2.43, + "learning_rate": 1.1919105176152344e-05, + "loss": 0.8503, + "step": 37758 + }, + { + "epoch": 2.43, + "learning_rate": 1.1918890487129393e-05, + "loss": 1.0186, + "step": 37759 + }, + { + "epoch": 2.43, + "learning_rate": 1.1918675798106442e-05, + "loss": 0.8729, + "step": 37760 + }, + { + "epoch": 2.43, + "learning_rate": 1.1918461109083494e-05, + "loss": 0.8261, + "step": 37761 + }, + { + "epoch": 2.43, + "learning_rate": 1.1918246420060543e-05, + "loss": 0.8861, + "step": 37762 + }, + { + "epoch": 2.43, + "learning_rate": 1.1918031731037594e-05, + "loss": 0.8737, + "step": 37763 + }, + { + "epoch": 2.43, + "learning_rate": 1.1917817042014643e-05, + "loss": 0.8716, + "step": 37764 + }, + { + "epoch": 2.43, + "learning_rate": 1.1917602352991694e-05, + "loss": 0.8967, + "step": 37765 + }, + { + "epoch": 2.43, + "learning_rate": 1.1917387663968743e-05, + "loss": 0.9293, + "step": 37766 + }, + { + "epoch": 2.43, + "learning_rate": 1.1917172974945792e-05, + "loss": 0.8965, + "step": 37767 + }, + { + "epoch": 2.43, + "learning_rate": 1.1916958285922842e-05, + "loss": 0.8945, + "step": 37768 + }, + { + "epoch": 2.43, + "learning_rate": 1.1916743596899891e-05, + "loss": 0.8621, + "step": 37769 + }, + { + "epoch": 2.43, + "learning_rate": 1.191652890787694e-05, + "loss": 0.9024, + "step": 37770 + }, + { + "epoch": 2.43, + "learning_rate": 1.191631421885399e-05, + "loss": 0.8848, + "step": 37771 + }, + { + "epoch": 2.43, + "learning_rate": 1.191609952983104e-05, + "loss": 0.9913, + "step": 37772 + }, + { + "epoch": 2.43, + "learning_rate": 1.191588484080809e-05, + "loss": 0.9567, + "step": 37773 + }, + { + "epoch": 2.43, + "learning_rate": 1.191567015178514e-05, + "loss": 0.9521, + "step": 37774 + }, + { + "epoch": 2.43, + "learning_rate": 1.1915455462762192e-05, + "loss": 0.9051, + "step": 37775 + }, + { + "epoch": 2.43, + "learning_rate": 1.191524077373924e-05, + "loss": 0.855, + "step": 37776 + }, + { + "epoch": 2.43, + "learning_rate": 1.191502608471629e-05, + "loss": 0.9447, + "step": 37777 + }, + { + "epoch": 2.43, + "learning_rate": 1.191481139569334e-05, + "loss": 0.9067, + "step": 37778 + }, + { + "epoch": 2.43, + "learning_rate": 1.191459670667039e-05, + "loss": 0.9077, + "step": 37779 + }, + { + "epoch": 2.43, + "learning_rate": 1.1914382017647438e-05, + "loss": 0.8714, + "step": 37780 + }, + { + "epoch": 2.43, + "learning_rate": 1.1914167328624489e-05, + "loss": 0.866, + "step": 37781 + }, + { + "epoch": 2.43, + "learning_rate": 1.1913952639601538e-05, + "loss": 0.8996, + "step": 37782 + }, + { + "epoch": 2.43, + "learning_rate": 1.1913737950578588e-05, + "loss": 0.907, + "step": 37783 + }, + { + "epoch": 2.43, + "learning_rate": 1.1913523261555637e-05, + "loss": 0.9066, + "step": 37784 + }, + { + "epoch": 2.43, + "learning_rate": 1.1913308572532686e-05, + "loss": 0.9206, + "step": 37785 + }, + { + "epoch": 2.43, + "learning_rate": 1.1913093883509737e-05, + "loss": 0.9772, + "step": 37786 + }, + { + "epoch": 2.43, + "learning_rate": 1.1912879194486786e-05, + "loss": 0.8852, + "step": 37787 + }, + { + "epoch": 2.43, + "learning_rate": 1.1912664505463835e-05, + "loss": 0.879, + "step": 37788 + }, + { + "epoch": 2.43, + "learning_rate": 1.1912449816440886e-05, + "loss": 0.9508, + "step": 37789 + }, + { + "epoch": 2.43, + "learning_rate": 1.1912235127417936e-05, + "loss": 0.9717, + "step": 37790 + }, + { + "epoch": 2.43, + "learning_rate": 1.1912020438394987e-05, + "loss": 0.9035, + "step": 37791 + }, + { + "epoch": 2.43, + "learning_rate": 1.1911805749372036e-05, + "loss": 0.9167, + "step": 37792 + }, + { + "epoch": 2.43, + "learning_rate": 1.1911591060349087e-05, + "loss": 0.8796, + "step": 37793 + }, + { + "epoch": 2.43, + "learning_rate": 1.1911376371326136e-05, + "loss": 0.9289, + "step": 37794 + }, + { + "epoch": 2.43, + "learning_rate": 1.1911161682303185e-05, + "loss": 0.8619, + "step": 37795 + }, + { + "epoch": 2.43, + "learning_rate": 1.1910946993280235e-05, + "loss": 0.904, + "step": 37796 + }, + { + "epoch": 2.43, + "learning_rate": 1.1910732304257284e-05, + "loss": 0.8198, + "step": 37797 + }, + { + "epoch": 2.43, + "learning_rate": 1.1910517615234333e-05, + "loss": 0.9217, + "step": 37798 + }, + { + "epoch": 2.43, + "learning_rate": 1.1910302926211384e-05, + "loss": 0.9082, + "step": 37799 + }, + { + "epoch": 2.43, + "learning_rate": 1.1910088237188433e-05, + "loss": 0.8778, + "step": 37800 + }, + { + "epoch": 2.43, + "learning_rate": 1.1909873548165483e-05, + "loss": 0.863, + "step": 37801 + }, + { + "epoch": 2.43, + "learning_rate": 1.1909658859142532e-05, + "loss": 0.9417, + "step": 37802 + }, + { + "epoch": 2.43, + "learning_rate": 1.1909444170119581e-05, + "loss": 0.8361, + "step": 37803 + }, + { + "epoch": 2.43, + "learning_rate": 1.1909229481096634e-05, + "loss": 0.8739, + "step": 37804 + }, + { + "epoch": 2.43, + "learning_rate": 1.1909014792073683e-05, + "loss": 0.8615, + "step": 37805 + }, + { + "epoch": 2.43, + "learning_rate": 1.1908800103050733e-05, + "loss": 0.8847, + "step": 37806 + }, + { + "epoch": 2.43, + "learning_rate": 1.1908585414027782e-05, + "loss": 0.9172, + "step": 37807 + }, + { + "epoch": 2.43, + "learning_rate": 1.1908370725004831e-05, + "loss": 0.8938, + "step": 37808 + }, + { + "epoch": 2.43, + "learning_rate": 1.1908156035981882e-05, + "loss": 0.8671, + "step": 37809 + }, + { + "epoch": 2.43, + "learning_rate": 1.1907941346958931e-05, + "loss": 0.9225, + "step": 37810 + }, + { + "epoch": 2.43, + "learning_rate": 1.1907726657935982e-05, + "loss": 0.9767, + "step": 37811 + }, + { + "epoch": 2.43, + "learning_rate": 1.190751196891303e-05, + "loss": 0.9703, + "step": 37812 + }, + { + "epoch": 2.43, + "learning_rate": 1.190729727989008e-05, + "loss": 0.9161, + "step": 37813 + }, + { + "epoch": 2.43, + "learning_rate": 1.190708259086713e-05, + "loss": 0.8581, + "step": 37814 + }, + { + "epoch": 2.43, + "learning_rate": 1.1906867901844179e-05, + "loss": 0.8596, + "step": 37815 + }, + { + "epoch": 2.43, + "learning_rate": 1.1906653212821228e-05, + "loss": 0.9196, + "step": 37816 + }, + { + "epoch": 2.43, + "learning_rate": 1.1906438523798279e-05, + "loss": 0.8532, + "step": 37817 + }, + { + "epoch": 2.43, + "learning_rate": 1.1906223834775328e-05, + "loss": 0.9474, + "step": 37818 + }, + { + "epoch": 2.43, + "learning_rate": 1.190600914575238e-05, + "loss": 0.8534, + "step": 37819 + }, + { + "epoch": 2.43, + "learning_rate": 1.1905794456729429e-05, + "loss": 0.8887, + "step": 37820 + }, + { + "epoch": 2.43, + "learning_rate": 1.190557976770648e-05, + "loss": 0.8805, + "step": 37821 + }, + { + "epoch": 2.43, + "learning_rate": 1.1905365078683529e-05, + "loss": 0.8984, + "step": 37822 + }, + { + "epoch": 2.43, + "learning_rate": 1.1905150389660578e-05, + "loss": 0.9072, + "step": 37823 + }, + { + "epoch": 2.43, + "learning_rate": 1.1904935700637628e-05, + "loss": 0.8757, + "step": 37824 + }, + { + "epoch": 2.43, + "learning_rate": 1.1904721011614677e-05, + "loss": 0.9105, + "step": 37825 + }, + { + "epoch": 2.43, + "learning_rate": 1.1904506322591726e-05, + "loss": 0.8775, + "step": 37826 + }, + { + "epoch": 2.43, + "learning_rate": 1.1904291633568777e-05, + "loss": 0.9084, + "step": 37827 + }, + { + "epoch": 2.43, + "learning_rate": 1.1904076944545826e-05, + "loss": 0.8694, + "step": 37828 + }, + { + "epoch": 2.43, + "learning_rate": 1.1903862255522875e-05, + "loss": 0.9451, + "step": 37829 + }, + { + "epoch": 2.43, + "learning_rate": 1.1903647566499925e-05, + "loss": 0.8961, + "step": 37830 + }, + { + "epoch": 2.43, + "learning_rate": 1.1903432877476974e-05, + "loss": 0.9348, + "step": 37831 + }, + { + "epoch": 2.43, + "learning_rate": 1.1903218188454025e-05, + "loss": 0.9523, + "step": 37832 + }, + { + "epoch": 2.43, + "learning_rate": 1.1903003499431076e-05, + "loss": 0.8973, + "step": 37833 + }, + { + "epoch": 2.43, + "learning_rate": 1.1902788810408126e-05, + "loss": 0.8594, + "step": 37834 + }, + { + "epoch": 2.43, + "learning_rate": 1.1902574121385175e-05, + "loss": 0.922, + "step": 37835 + }, + { + "epoch": 2.43, + "learning_rate": 1.1902359432362224e-05, + "loss": 0.9253, + "step": 37836 + }, + { + "epoch": 2.43, + "learning_rate": 1.1902144743339275e-05, + "loss": 0.8336, + "step": 37837 + }, + { + "epoch": 2.43, + "learning_rate": 1.1901930054316324e-05, + "loss": 0.8591, + "step": 37838 + }, + { + "epoch": 2.43, + "learning_rate": 1.1901715365293373e-05, + "loss": 0.8726, + "step": 37839 + }, + { + "epoch": 2.43, + "learning_rate": 1.1901500676270424e-05, + "loss": 1.005, + "step": 37840 + }, + { + "epoch": 2.43, + "learning_rate": 1.1901285987247473e-05, + "loss": 0.8964, + "step": 37841 + }, + { + "epoch": 2.43, + "learning_rate": 1.1901071298224523e-05, + "loss": 0.9074, + "step": 37842 + }, + { + "epoch": 2.43, + "learning_rate": 1.1900856609201572e-05, + "loss": 0.8952, + "step": 37843 + }, + { + "epoch": 2.43, + "learning_rate": 1.1900641920178621e-05, + "loss": 0.9316, + "step": 37844 + }, + { + "epoch": 2.43, + "learning_rate": 1.1900427231155672e-05, + "loss": 0.9029, + "step": 37845 + }, + { + "epoch": 2.43, + "learning_rate": 1.190021254213272e-05, + "loss": 0.8493, + "step": 37846 + }, + { + "epoch": 2.43, + "learning_rate": 1.189999785310977e-05, + "loss": 0.921, + "step": 37847 + }, + { + "epoch": 2.44, + "learning_rate": 1.1899783164086822e-05, + "loss": 0.7808, + "step": 37848 + }, + { + "epoch": 2.44, + "learning_rate": 1.1899568475063871e-05, + "loss": 0.8664, + "step": 37849 + }, + { + "epoch": 2.44, + "learning_rate": 1.1899353786040922e-05, + "loss": 0.9466, + "step": 37850 + }, + { + "epoch": 2.44, + "learning_rate": 1.189913909701797e-05, + "loss": 0.9185, + "step": 37851 + }, + { + "epoch": 2.44, + "learning_rate": 1.1898924407995021e-05, + "loss": 0.9313, + "step": 37852 + }, + { + "epoch": 2.44, + "learning_rate": 1.189870971897207e-05, + "loss": 0.8304, + "step": 37853 + }, + { + "epoch": 2.44, + "learning_rate": 1.189849502994912e-05, + "loss": 0.881, + "step": 37854 + }, + { + "epoch": 2.44, + "learning_rate": 1.189828034092617e-05, + "loss": 0.8884, + "step": 37855 + }, + { + "epoch": 2.44, + "learning_rate": 1.1898065651903219e-05, + "loss": 0.8964, + "step": 37856 + }, + { + "epoch": 2.44, + "learning_rate": 1.1897850962880268e-05, + "loss": 0.8498, + "step": 37857 + }, + { + "epoch": 2.44, + "learning_rate": 1.1897636273857318e-05, + "loss": 0.8767, + "step": 37858 + }, + { + "epoch": 2.44, + "learning_rate": 1.1897421584834367e-05, + "loss": 0.9301, + "step": 37859 + }, + { + "epoch": 2.44, + "learning_rate": 1.1897206895811418e-05, + "loss": 0.9724, + "step": 37860 + }, + { + "epoch": 2.44, + "learning_rate": 1.1896992206788467e-05, + "loss": 0.9031, + "step": 37861 + }, + { + "epoch": 2.44, + "learning_rate": 1.189677751776552e-05, + "loss": 0.8621, + "step": 37862 + }, + { + "epoch": 2.44, + "learning_rate": 1.1896562828742568e-05, + "loss": 0.91, + "step": 37863 + }, + { + "epoch": 2.44, + "learning_rate": 1.1896348139719617e-05, + "loss": 0.8719, + "step": 37864 + }, + { + "epoch": 2.44, + "learning_rate": 1.1896133450696668e-05, + "loss": 0.9442, + "step": 37865 + }, + { + "epoch": 2.44, + "learning_rate": 1.1895918761673717e-05, + "loss": 0.9665, + "step": 37866 + }, + { + "epoch": 2.44, + "learning_rate": 1.1895704072650766e-05, + "loss": 0.8812, + "step": 37867 + }, + { + "epoch": 2.44, + "learning_rate": 1.1895489383627817e-05, + "loss": 0.8824, + "step": 37868 + }, + { + "epoch": 2.44, + "learning_rate": 1.1895274694604866e-05, + "loss": 0.9084, + "step": 37869 + }, + { + "epoch": 2.44, + "learning_rate": 1.1895060005581916e-05, + "loss": 0.909, + "step": 37870 + }, + { + "epoch": 2.44, + "learning_rate": 1.1894845316558965e-05, + "loss": 0.8863, + "step": 37871 + }, + { + "epoch": 2.44, + "learning_rate": 1.1894630627536014e-05, + "loss": 0.8804, + "step": 37872 + }, + { + "epoch": 2.44, + "learning_rate": 1.1894415938513065e-05, + "loss": 0.847, + "step": 37873 + }, + { + "epoch": 2.44, + "learning_rate": 1.1894201249490114e-05, + "loss": 0.8887, + "step": 37874 + }, + { + "epoch": 2.44, + "learning_rate": 1.1893986560467163e-05, + "loss": 0.9611, + "step": 37875 + }, + { + "epoch": 2.44, + "learning_rate": 1.1893771871444213e-05, + "loss": 0.9436, + "step": 37876 + }, + { + "epoch": 2.44, + "learning_rate": 1.1893557182421264e-05, + "loss": 0.8845, + "step": 37877 + }, + { + "epoch": 2.44, + "learning_rate": 1.1893342493398315e-05, + "loss": 0.8619, + "step": 37878 + }, + { + "epoch": 2.44, + "learning_rate": 1.1893127804375364e-05, + "loss": 0.9152, + "step": 37879 + }, + { + "epoch": 2.44, + "learning_rate": 1.1892913115352414e-05, + "loss": 0.8698, + "step": 37880 + }, + { + "epoch": 2.44, + "learning_rate": 1.1892698426329463e-05, + "loss": 0.9238, + "step": 37881 + }, + { + "epoch": 2.44, + "learning_rate": 1.1892483737306512e-05, + "loss": 0.9071, + "step": 37882 + }, + { + "epoch": 2.44, + "learning_rate": 1.1892269048283563e-05, + "loss": 0.9258, + "step": 37883 + }, + { + "epoch": 2.44, + "learning_rate": 1.1892054359260612e-05, + "loss": 0.9148, + "step": 37884 + }, + { + "epoch": 2.44, + "learning_rate": 1.189183967023766e-05, + "loss": 0.9199, + "step": 37885 + }, + { + "epoch": 2.44, + "learning_rate": 1.1891624981214712e-05, + "loss": 0.8427, + "step": 37886 + }, + { + "epoch": 2.44, + "learning_rate": 1.189141029219176e-05, + "loss": 0.8997, + "step": 37887 + }, + { + "epoch": 2.44, + "learning_rate": 1.189119560316881e-05, + "loss": 0.9536, + "step": 37888 + }, + { + "epoch": 2.44, + "learning_rate": 1.189098091414586e-05, + "loss": 0.9231, + "step": 37889 + }, + { + "epoch": 2.44, + "learning_rate": 1.1890766225122909e-05, + "loss": 0.9849, + "step": 37890 + }, + { + "epoch": 2.44, + "learning_rate": 1.1890551536099961e-05, + "loss": 0.8517, + "step": 37891 + }, + { + "epoch": 2.44, + "learning_rate": 1.189033684707701e-05, + "loss": 0.9335, + "step": 37892 + }, + { + "epoch": 2.44, + "learning_rate": 1.1890122158054061e-05, + "loss": 0.9084, + "step": 37893 + }, + { + "epoch": 2.44, + "learning_rate": 1.188990746903111e-05, + "loss": 0.9099, + "step": 37894 + }, + { + "epoch": 2.44, + "learning_rate": 1.1889692780008159e-05, + "loss": 0.9251, + "step": 37895 + }, + { + "epoch": 2.44, + "learning_rate": 1.188947809098521e-05, + "loss": 0.8041, + "step": 37896 + }, + { + "epoch": 2.44, + "learning_rate": 1.1889263401962259e-05, + "loss": 0.884, + "step": 37897 + }, + { + "epoch": 2.44, + "learning_rate": 1.188904871293931e-05, + "loss": 0.9087, + "step": 37898 + }, + { + "epoch": 2.44, + "learning_rate": 1.1888834023916358e-05, + "loss": 0.8471, + "step": 37899 + }, + { + "epoch": 2.44, + "learning_rate": 1.1888619334893407e-05, + "loss": 0.9248, + "step": 37900 + }, + { + "epoch": 2.44, + "learning_rate": 1.1888404645870458e-05, + "loss": 0.8903, + "step": 37901 + }, + { + "epoch": 2.44, + "learning_rate": 1.1888189956847507e-05, + "loss": 0.9183, + "step": 37902 + }, + { + "epoch": 2.44, + "learning_rate": 1.1887975267824556e-05, + "loss": 0.8997, + "step": 37903 + }, + { + "epoch": 2.44, + "learning_rate": 1.1887760578801606e-05, + "loss": 0.914, + "step": 37904 + }, + { + "epoch": 2.44, + "learning_rate": 1.1887545889778655e-05, + "loss": 0.8607, + "step": 37905 + }, + { + "epoch": 2.44, + "learning_rate": 1.1887331200755708e-05, + "loss": 0.9106, + "step": 37906 + }, + { + "epoch": 2.44, + "learning_rate": 1.1887116511732757e-05, + "loss": 0.9111, + "step": 37907 + }, + { + "epoch": 2.44, + "learning_rate": 1.1886901822709807e-05, + "loss": 0.9037, + "step": 37908 + }, + { + "epoch": 2.44, + "learning_rate": 1.1886687133686856e-05, + "loss": 0.8847, + "step": 37909 + }, + { + "epoch": 2.44, + "learning_rate": 1.1886472444663905e-05, + "loss": 0.8378, + "step": 37910 + }, + { + "epoch": 2.44, + "learning_rate": 1.1886257755640956e-05, + "loss": 0.8955, + "step": 37911 + }, + { + "epoch": 2.44, + "learning_rate": 1.1886043066618005e-05, + "loss": 0.8943, + "step": 37912 + }, + { + "epoch": 2.44, + "learning_rate": 1.1885828377595054e-05, + "loss": 0.9373, + "step": 37913 + }, + { + "epoch": 2.44, + "learning_rate": 1.1885613688572105e-05, + "loss": 0.8878, + "step": 37914 + }, + { + "epoch": 2.44, + "learning_rate": 1.1885398999549154e-05, + "loss": 0.8425, + "step": 37915 + }, + { + "epoch": 2.44, + "learning_rate": 1.1885184310526202e-05, + "loss": 0.8794, + "step": 37916 + }, + { + "epoch": 2.44, + "learning_rate": 1.1884969621503253e-05, + "loss": 0.9141, + "step": 37917 + }, + { + "epoch": 2.44, + "learning_rate": 1.1884754932480302e-05, + "loss": 0.9303, + "step": 37918 + }, + { + "epoch": 2.44, + "learning_rate": 1.1884540243457353e-05, + "loss": 0.8142, + "step": 37919 + }, + { + "epoch": 2.44, + "learning_rate": 1.1884325554434403e-05, + "loss": 0.9532, + "step": 37920 + }, + { + "epoch": 2.44, + "learning_rate": 1.1884110865411454e-05, + "loss": 0.877, + "step": 37921 + }, + { + "epoch": 2.44, + "learning_rate": 1.1883896176388503e-05, + "loss": 0.8827, + "step": 37922 + }, + { + "epoch": 2.44, + "learning_rate": 1.1883681487365552e-05, + "loss": 0.8977, + "step": 37923 + }, + { + "epoch": 2.44, + "learning_rate": 1.1883466798342603e-05, + "loss": 0.9133, + "step": 37924 + }, + { + "epoch": 2.44, + "learning_rate": 1.1883252109319652e-05, + "loss": 0.9491, + "step": 37925 + }, + { + "epoch": 2.44, + "learning_rate": 1.18830374202967e-05, + "loss": 0.9345, + "step": 37926 + }, + { + "epoch": 2.44, + "learning_rate": 1.1882822731273751e-05, + "loss": 0.8741, + "step": 37927 + }, + { + "epoch": 2.44, + "learning_rate": 1.18826080422508e-05, + "loss": 0.8555, + "step": 37928 + }, + { + "epoch": 2.44, + "learning_rate": 1.1882393353227851e-05, + "loss": 0.9488, + "step": 37929 + }, + { + "epoch": 2.44, + "learning_rate": 1.18821786642049e-05, + "loss": 0.8391, + "step": 37930 + }, + { + "epoch": 2.44, + "learning_rate": 1.1881963975181949e-05, + "loss": 0.8647, + "step": 37931 + }, + { + "epoch": 2.44, + "learning_rate": 1.1881749286159e-05, + "loss": 0.9163, + "step": 37932 + }, + { + "epoch": 2.44, + "learning_rate": 1.1881534597136048e-05, + "loss": 0.8701, + "step": 37933 + }, + { + "epoch": 2.44, + "learning_rate": 1.1881319908113097e-05, + "loss": 0.8568, + "step": 37934 + }, + { + "epoch": 2.44, + "learning_rate": 1.188110521909015e-05, + "loss": 0.8516, + "step": 37935 + }, + { + "epoch": 2.44, + "learning_rate": 1.1880890530067199e-05, + "loss": 0.877, + "step": 37936 + }, + { + "epoch": 2.44, + "learning_rate": 1.188067584104425e-05, + "loss": 0.794, + "step": 37937 + }, + { + "epoch": 2.44, + "learning_rate": 1.1880461152021298e-05, + "loss": 0.8901, + "step": 37938 + }, + { + "epoch": 2.44, + "learning_rate": 1.1880246462998349e-05, + "loss": 0.8876, + "step": 37939 + }, + { + "epoch": 2.44, + "learning_rate": 1.1880031773975398e-05, + "loss": 0.9374, + "step": 37940 + }, + { + "epoch": 2.44, + "learning_rate": 1.1879817084952447e-05, + "loss": 0.9265, + "step": 37941 + }, + { + "epoch": 2.44, + "learning_rate": 1.1879602395929498e-05, + "loss": 0.8924, + "step": 37942 + }, + { + "epoch": 2.44, + "learning_rate": 1.1879387706906547e-05, + "loss": 0.8922, + "step": 37943 + }, + { + "epoch": 2.44, + "learning_rate": 1.1879173017883596e-05, + "loss": 0.909, + "step": 37944 + }, + { + "epoch": 2.44, + "learning_rate": 1.1878958328860646e-05, + "loss": 0.9386, + "step": 37945 + }, + { + "epoch": 2.44, + "learning_rate": 1.1878743639837695e-05, + "loss": 0.8833, + "step": 37946 + }, + { + "epoch": 2.44, + "learning_rate": 1.1878528950814746e-05, + "loss": 0.9078, + "step": 37947 + }, + { + "epoch": 2.44, + "learning_rate": 1.1878314261791795e-05, + "loss": 0.8643, + "step": 37948 + }, + { + "epoch": 2.44, + "learning_rate": 1.1878099572768847e-05, + "loss": 0.893, + "step": 37949 + }, + { + "epoch": 2.44, + "learning_rate": 1.1877884883745896e-05, + "loss": 0.8982, + "step": 37950 + }, + { + "epoch": 2.44, + "learning_rate": 1.1877670194722945e-05, + "loss": 0.8858, + "step": 37951 + }, + { + "epoch": 2.44, + "learning_rate": 1.1877455505699996e-05, + "loss": 0.9064, + "step": 37952 + }, + { + "epoch": 2.44, + "learning_rate": 1.1877240816677045e-05, + "loss": 0.8846, + "step": 37953 + }, + { + "epoch": 2.44, + "learning_rate": 1.1877026127654094e-05, + "loss": 0.9875, + "step": 37954 + }, + { + "epoch": 2.44, + "learning_rate": 1.1876811438631144e-05, + "loss": 0.8922, + "step": 37955 + }, + { + "epoch": 2.44, + "learning_rate": 1.1876596749608193e-05, + "loss": 0.834, + "step": 37956 + }, + { + "epoch": 2.44, + "learning_rate": 1.1876382060585244e-05, + "loss": 0.8895, + "step": 37957 + }, + { + "epoch": 2.44, + "learning_rate": 1.1876167371562293e-05, + "loss": 0.9584, + "step": 37958 + }, + { + "epoch": 2.44, + "learning_rate": 1.1875952682539342e-05, + "loss": 0.9297, + "step": 37959 + }, + { + "epoch": 2.44, + "learning_rate": 1.1875737993516393e-05, + "loss": 0.8913, + "step": 37960 + }, + { + "epoch": 2.44, + "learning_rate": 1.1875523304493441e-05, + "loss": 0.9087, + "step": 37961 + }, + { + "epoch": 2.44, + "learning_rate": 1.187530861547049e-05, + "loss": 0.967, + "step": 37962 + }, + { + "epoch": 2.44, + "learning_rate": 1.1875093926447541e-05, + "loss": 0.9112, + "step": 37963 + }, + { + "epoch": 2.44, + "learning_rate": 1.1874879237424592e-05, + "loss": 0.9485, + "step": 37964 + }, + { + "epoch": 2.44, + "learning_rate": 1.1874664548401642e-05, + "loss": 0.8929, + "step": 37965 + }, + { + "epoch": 2.44, + "learning_rate": 1.1874449859378691e-05, + "loss": 0.8521, + "step": 37966 + }, + { + "epoch": 2.44, + "learning_rate": 1.1874235170355742e-05, + "loss": 0.9635, + "step": 37967 + }, + { + "epoch": 2.44, + "learning_rate": 1.1874020481332791e-05, + "loss": 0.8755, + "step": 37968 + }, + { + "epoch": 2.44, + "learning_rate": 1.187380579230984e-05, + "loss": 0.8743, + "step": 37969 + }, + { + "epoch": 2.44, + "learning_rate": 1.187359110328689e-05, + "loss": 0.8352, + "step": 37970 + }, + { + "epoch": 2.44, + "learning_rate": 1.187337641426394e-05, + "loss": 0.9005, + "step": 37971 + }, + { + "epoch": 2.44, + "learning_rate": 1.1873161725240989e-05, + "loss": 0.9379, + "step": 37972 + }, + { + "epoch": 2.44, + "learning_rate": 1.187294703621804e-05, + "loss": 0.8363, + "step": 37973 + }, + { + "epoch": 2.44, + "learning_rate": 1.1872732347195088e-05, + "loss": 0.9165, + "step": 37974 + }, + { + "epoch": 2.44, + "learning_rate": 1.1872517658172137e-05, + "loss": 0.9222, + "step": 37975 + }, + { + "epoch": 2.44, + "learning_rate": 1.1872302969149188e-05, + "loss": 0.8777, + "step": 37976 + }, + { + "epoch": 2.44, + "learning_rate": 1.1872088280126237e-05, + "loss": 0.8776, + "step": 37977 + }, + { + "epoch": 2.44, + "learning_rate": 1.1871873591103289e-05, + "loss": 0.8792, + "step": 37978 + }, + { + "epoch": 2.44, + "learning_rate": 1.1871658902080338e-05, + "loss": 0.8529, + "step": 37979 + }, + { + "epoch": 2.44, + "learning_rate": 1.1871444213057389e-05, + "loss": 0.9194, + "step": 37980 + }, + { + "epoch": 2.44, + "learning_rate": 1.1871229524034438e-05, + "loss": 0.8759, + "step": 37981 + }, + { + "epoch": 2.44, + "learning_rate": 1.1871014835011487e-05, + "loss": 0.8964, + "step": 37982 + }, + { + "epoch": 2.44, + "learning_rate": 1.1870800145988537e-05, + "loss": 0.8635, + "step": 37983 + }, + { + "epoch": 2.44, + "learning_rate": 1.1870585456965586e-05, + "loss": 0.8987, + "step": 37984 + }, + { + "epoch": 2.44, + "learning_rate": 1.1870370767942635e-05, + "loss": 0.934, + "step": 37985 + }, + { + "epoch": 2.44, + "learning_rate": 1.1870156078919686e-05, + "loss": 0.8759, + "step": 37986 + }, + { + "epoch": 2.44, + "learning_rate": 1.1869941389896735e-05, + "loss": 0.9136, + "step": 37987 + }, + { + "epoch": 2.44, + "learning_rate": 1.1869726700873786e-05, + "loss": 1.0063, + "step": 37988 + }, + { + "epoch": 2.44, + "learning_rate": 1.1869512011850835e-05, + "loss": 0.929, + "step": 37989 + }, + { + "epoch": 2.44, + "learning_rate": 1.1869297322827883e-05, + "loss": 0.8406, + "step": 37990 + }, + { + "epoch": 2.44, + "learning_rate": 1.1869082633804934e-05, + "loss": 0.8855, + "step": 37991 + }, + { + "epoch": 2.44, + "learning_rate": 1.1868867944781983e-05, + "loss": 0.8304, + "step": 37992 + }, + { + "epoch": 2.44, + "learning_rate": 1.1868653255759035e-05, + "loss": 0.9773, + "step": 37993 + }, + { + "epoch": 2.44, + "learning_rate": 1.1868438566736084e-05, + "loss": 0.8358, + "step": 37994 + }, + { + "epoch": 2.44, + "learning_rate": 1.1868223877713133e-05, + "loss": 0.8294, + "step": 37995 + }, + { + "epoch": 2.44, + "learning_rate": 1.1868009188690184e-05, + "loss": 0.8203, + "step": 37996 + }, + { + "epoch": 2.44, + "learning_rate": 1.1867794499667233e-05, + "loss": 0.9253, + "step": 37997 + }, + { + "epoch": 2.44, + "learning_rate": 1.1867579810644284e-05, + "loss": 0.9086, + "step": 37998 + }, + { + "epoch": 2.44, + "learning_rate": 1.1867365121621333e-05, + "loss": 0.9346, + "step": 37999 + }, + { + "epoch": 2.44, + "learning_rate": 1.1867150432598382e-05, + "loss": 0.9319, + "step": 38000 + }, + { + "epoch": 2.44, + "eval_laion/cider": 0.49617624099770713, + "eval_laion/meteor": 0.3914140569660207, + "eval_laion/pred_num_words": 36.833984375, + "eval_laion/sacrebleu": 15.889290610348763, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.037583351135254, + "eval_runtime": 99.7703, + "eval_samples_per_second": 5.001, + "eval_steps_per_second": 0.321, + "step": 38000 + }, + { + "epoch": 2.44, + "learning_rate": 1.1866935743575432e-05, + "loss": 0.9281, + "step": 38001 + }, + { + "epoch": 2.44, + "learning_rate": 1.1866721054552481e-05, + "loss": 0.8338, + "step": 38002 + }, + { + "epoch": 2.45, + "learning_rate": 1.186650636552953e-05, + "loss": 0.9875, + "step": 38003 + }, + { + "epoch": 2.45, + "learning_rate": 1.186629167650658e-05, + "loss": 0.8973, + "step": 38004 + }, + { + "epoch": 2.45, + "learning_rate": 1.186607698748363e-05, + "loss": 0.9023, + "step": 38005 + }, + { + "epoch": 2.45, + "learning_rate": 1.186586229846068e-05, + "loss": 0.9323, + "step": 38006 + }, + { + "epoch": 2.45, + "learning_rate": 1.1865647609437731e-05, + "loss": 0.851, + "step": 38007 + }, + { + "epoch": 2.45, + "learning_rate": 1.1865432920414782e-05, + "loss": 0.9125, + "step": 38008 + }, + { + "epoch": 2.45, + "learning_rate": 1.186521823139183e-05, + "loss": 0.9172, + "step": 38009 + }, + { + "epoch": 2.45, + "learning_rate": 1.186500354236888e-05, + "loss": 0.8421, + "step": 38010 + }, + { + "epoch": 2.45, + "learning_rate": 1.186478885334593e-05, + "loss": 0.9434, + "step": 38011 + }, + { + "epoch": 2.45, + "learning_rate": 1.186457416432298e-05, + "loss": 0.9727, + "step": 38012 + }, + { + "epoch": 2.45, + "learning_rate": 1.1864359475300028e-05, + "loss": 0.859, + "step": 38013 + }, + { + "epoch": 2.45, + "learning_rate": 1.1864144786277079e-05, + "loss": 0.9389, + "step": 38014 + }, + { + "epoch": 2.45, + "learning_rate": 1.1863930097254128e-05, + "loss": 0.8625, + "step": 38015 + }, + { + "epoch": 2.45, + "learning_rate": 1.1863715408231179e-05, + "loss": 0.9173, + "step": 38016 + }, + { + "epoch": 2.45, + "learning_rate": 1.1863500719208228e-05, + "loss": 0.8461, + "step": 38017 + }, + { + "epoch": 2.45, + "learning_rate": 1.1863286030185277e-05, + "loss": 0.9667, + "step": 38018 + }, + { + "epoch": 2.45, + "learning_rate": 1.1863071341162327e-05, + "loss": 0.8301, + "step": 38019 + }, + { + "epoch": 2.45, + "learning_rate": 1.1862856652139376e-05, + "loss": 0.9413, + "step": 38020 + }, + { + "epoch": 2.45, + "learning_rate": 1.1862641963116425e-05, + "loss": 0.8839, + "step": 38021 + }, + { + "epoch": 2.45, + "learning_rate": 1.1862427274093477e-05, + "loss": 0.9043, + "step": 38022 + }, + { + "epoch": 2.45, + "learning_rate": 1.1862212585070526e-05, + "loss": 0.9038, + "step": 38023 + }, + { + "epoch": 2.45, + "learning_rate": 1.1861997896047577e-05, + "loss": 0.9007, + "step": 38024 + }, + { + "epoch": 2.45, + "learning_rate": 1.1861783207024626e-05, + "loss": 0.8486, + "step": 38025 + }, + { + "epoch": 2.45, + "learning_rate": 1.1861568518001677e-05, + "loss": 0.868, + "step": 38026 + }, + { + "epoch": 2.45, + "learning_rate": 1.1861353828978726e-05, + "loss": 0.9046, + "step": 38027 + }, + { + "epoch": 2.45, + "learning_rate": 1.1861139139955775e-05, + "loss": 0.8884, + "step": 38028 + }, + { + "epoch": 2.45, + "learning_rate": 1.1860924450932825e-05, + "loss": 0.9923, + "step": 38029 + }, + { + "epoch": 2.45, + "learning_rate": 1.1860709761909874e-05, + "loss": 0.8924, + "step": 38030 + }, + { + "epoch": 2.45, + "learning_rate": 1.1860495072886923e-05, + "loss": 0.8599, + "step": 38031 + }, + { + "epoch": 2.45, + "learning_rate": 1.1860280383863974e-05, + "loss": 0.9076, + "step": 38032 + }, + { + "epoch": 2.45, + "learning_rate": 1.1860065694841023e-05, + "loss": 0.905, + "step": 38033 + }, + { + "epoch": 2.45, + "learning_rate": 1.1859851005818074e-05, + "loss": 0.8732, + "step": 38034 + }, + { + "epoch": 2.45, + "learning_rate": 1.1859636316795122e-05, + "loss": 0.94, + "step": 38035 + }, + { + "epoch": 2.45, + "learning_rate": 1.1859421627772171e-05, + "loss": 0.8518, + "step": 38036 + }, + { + "epoch": 2.45, + "learning_rate": 1.1859206938749224e-05, + "loss": 0.936, + "step": 38037 + }, + { + "epoch": 2.45, + "learning_rate": 1.1858992249726273e-05, + "loss": 0.8897, + "step": 38038 + }, + { + "epoch": 2.45, + "learning_rate": 1.1858777560703323e-05, + "loss": 0.9044, + "step": 38039 + }, + { + "epoch": 2.45, + "learning_rate": 1.1858562871680372e-05, + "loss": 0.9413, + "step": 38040 + }, + { + "epoch": 2.45, + "learning_rate": 1.1858348182657421e-05, + "loss": 0.8431, + "step": 38041 + }, + { + "epoch": 2.45, + "learning_rate": 1.1858133493634472e-05, + "loss": 0.9292, + "step": 38042 + }, + { + "epoch": 2.45, + "learning_rate": 1.1857918804611521e-05, + "loss": 0.8848, + "step": 38043 + }, + { + "epoch": 2.45, + "learning_rate": 1.1857704115588572e-05, + "loss": 0.9154, + "step": 38044 + }, + { + "epoch": 2.45, + "learning_rate": 1.185748942656562e-05, + "loss": 0.9141, + "step": 38045 + }, + { + "epoch": 2.45, + "learning_rate": 1.185727473754267e-05, + "loss": 0.8474, + "step": 38046 + }, + { + "epoch": 2.45, + "learning_rate": 1.185706004851972e-05, + "loss": 0.8989, + "step": 38047 + }, + { + "epoch": 2.45, + "learning_rate": 1.185684535949677e-05, + "loss": 0.8227, + "step": 38048 + }, + { + "epoch": 2.45, + "learning_rate": 1.1856630670473818e-05, + "loss": 0.8281, + "step": 38049 + }, + { + "epoch": 2.45, + "learning_rate": 1.1856415981450869e-05, + "loss": 0.9369, + "step": 38050 + }, + { + "epoch": 2.45, + "learning_rate": 1.185620129242792e-05, + "loss": 0.9171, + "step": 38051 + }, + { + "epoch": 2.45, + "learning_rate": 1.185598660340497e-05, + "loss": 0.8353, + "step": 38052 + }, + { + "epoch": 2.45, + "learning_rate": 1.1855771914382019e-05, + "loss": 0.8406, + "step": 38053 + }, + { + "epoch": 2.45, + "learning_rate": 1.185555722535907e-05, + "loss": 0.8646, + "step": 38054 + }, + { + "epoch": 2.45, + "learning_rate": 1.1855342536336119e-05, + "loss": 0.894, + "step": 38055 + }, + { + "epoch": 2.45, + "learning_rate": 1.1855127847313168e-05, + "loss": 0.878, + "step": 38056 + }, + { + "epoch": 2.45, + "learning_rate": 1.1854913158290218e-05, + "loss": 0.9514, + "step": 38057 + }, + { + "epoch": 2.45, + "learning_rate": 1.1854698469267267e-05, + "loss": 0.8932, + "step": 38058 + }, + { + "epoch": 2.45, + "learning_rate": 1.1854483780244316e-05, + "loss": 0.8645, + "step": 38059 + }, + { + "epoch": 2.45, + "learning_rate": 1.1854269091221367e-05, + "loss": 0.8809, + "step": 38060 + }, + { + "epoch": 2.45, + "learning_rate": 1.1854054402198416e-05, + "loss": 0.8752, + "step": 38061 + }, + { + "epoch": 2.45, + "learning_rate": 1.1853839713175465e-05, + "loss": 0.8903, + "step": 38062 + }, + { + "epoch": 2.45, + "learning_rate": 1.1853625024152516e-05, + "loss": 0.9685, + "step": 38063 + }, + { + "epoch": 2.45, + "learning_rate": 1.1853410335129564e-05, + "loss": 0.896, + "step": 38064 + }, + { + "epoch": 2.45, + "learning_rate": 1.1853195646106615e-05, + "loss": 0.8658, + "step": 38065 + }, + { + "epoch": 2.45, + "learning_rate": 1.1852980957083666e-05, + "loss": 0.8905, + "step": 38066 + }, + { + "epoch": 2.45, + "learning_rate": 1.1852766268060716e-05, + "loss": 0.8975, + "step": 38067 + }, + { + "epoch": 2.45, + "learning_rate": 1.1852551579037765e-05, + "loss": 0.9139, + "step": 38068 + }, + { + "epoch": 2.45, + "learning_rate": 1.1852336890014814e-05, + "loss": 0.9042, + "step": 38069 + }, + { + "epoch": 2.45, + "learning_rate": 1.1852122200991865e-05, + "loss": 0.8739, + "step": 38070 + }, + { + "epoch": 2.45, + "learning_rate": 1.1851907511968914e-05, + "loss": 0.9846, + "step": 38071 + }, + { + "epoch": 2.45, + "learning_rate": 1.1851692822945963e-05, + "loss": 0.8624, + "step": 38072 + }, + { + "epoch": 2.45, + "learning_rate": 1.1851478133923014e-05, + "loss": 0.8178, + "step": 38073 + }, + { + "epoch": 2.45, + "learning_rate": 1.1851263444900063e-05, + "loss": 0.8657, + "step": 38074 + }, + { + "epoch": 2.45, + "learning_rate": 1.1851048755877113e-05, + "loss": 0.8932, + "step": 38075 + }, + { + "epoch": 2.45, + "learning_rate": 1.1850834066854162e-05, + "loss": 0.885, + "step": 38076 + }, + { + "epoch": 2.45, + "learning_rate": 1.1850619377831211e-05, + "loss": 0.9145, + "step": 38077 + }, + { + "epoch": 2.45, + "learning_rate": 1.1850404688808262e-05, + "loss": 0.9092, + "step": 38078 + }, + { + "epoch": 2.45, + "learning_rate": 1.185018999978531e-05, + "loss": 0.9171, + "step": 38079 + }, + { + "epoch": 2.45, + "learning_rate": 1.1849975310762363e-05, + "loss": 0.8577, + "step": 38080 + }, + { + "epoch": 2.45, + "learning_rate": 1.1849760621739412e-05, + "loss": 0.8969, + "step": 38081 + }, + { + "epoch": 2.45, + "learning_rate": 1.1849545932716461e-05, + "loss": 0.8531, + "step": 38082 + }, + { + "epoch": 2.45, + "learning_rate": 1.1849331243693512e-05, + "loss": 0.9636, + "step": 38083 + }, + { + "epoch": 2.45, + "learning_rate": 1.184911655467056e-05, + "loss": 0.8732, + "step": 38084 + }, + { + "epoch": 2.45, + "learning_rate": 1.1848901865647611e-05, + "loss": 0.9209, + "step": 38085 + }, + { + "epoch": 2.45, + "learning_rate": 1.184868717662466e-05, + "loss": 0.9517, + "step": 38086 + }, + { + "epoch": 2.45, + "learning_rate": 1.184847248760171e-05, + "loss": 0.884, + "step": 38087 + }, + { + "epoch": 2.45, + "learning_rate": 1.184825779857876e-05, + "loss": 0.8894, + "step": 38088 + }, + { + "epoch": 2.45, + "learning_rate": 1.1848043109555809e-05, + "loss": 0.9071, + "step": 38089 + }, + { + "epoch": 2.45, + "learning_rate": 1.1847828420532858e-05, + "loss": 0.8848, + "step": 38090 + }, + { + "epoch": 2.45, + "learning_rate": 1.1847613731509909e-05, + "loss": 0.8458, + "step": 38091 + }, + { + "epoch": 2.45, + "learning_rate": 1.1847399042486958e-05, + "loss": 0.934, + "step": 38092 + }, + { + "epoch": 2.45, + "learning_rate": 1.1847184353464008e-05, + "loss": 0.8952, + "step": 38093 + }, + { + "epoch": 2.45, + "learning_rate": 1.1846969664441057e-05, + "loss": 0.9059, + "step": 38094 + }, + { + "epoch": 2.45, + "learning_rate": 1.184675497541811e-05, + "loss": 0.9173, + "step": 38095 + }, + { + "epoch": 2.45, + "learning_rate": 1.1846540286395158e-05, + "loss": 0.8542, + "step": 38096 + }, + { + "epoch": 2.45, + "learning_rate": 1.1846325597372207e-05, + "loss": 0.9438, + "step": 38097 + }, + { + "epoch": 2.45, + "learning_rate": 1.1846110908349258e-05, + "loss": 0.9539, + "step": 38098 + }, + { + "epoch": 2.45, + "learning_rate": 1.1845896219326307e-05, + "loss": 0.9655, + "step": 38099 + }, + { + "epoch": 2.45, + "learning_rate": 1.1845681530303356e-05, + "loss": 0.8905, + "step": 38100 + }, + { + "epoch": 2.45, + "learning_rate": 1.1845466841280407e-05, + "loss": 0.8868, + "step": 38101 + }, + { + "epoch": 2.45, + "learning_rate": 1.1845252152257456e-05, + "loss": 0.8777, + "step": 38102 + }, + { + "epoch": 2.45, + "learning_rate": 1.1845037463234506e-05, + "loss": 0.8833, + "step": 38103 + }, + { + "epoch": 2.45, + "learning_rate": 1.1844822774211555e-05, + "loss": 0.9463, + "step": 38104 + }, + { + "epoch": 2.45, + "learning_rate": 1.1844608085188604e-05, + "loss": 0.8937, + "step": 38105 + }, + { + "epoch": 2.45, + "learning_rate": 1.1844393396165655e-05, + "loss": 0.9175, + "step": 38106 + }, + { + "epoch": 2.45, + "learning_rate": 1.1844178707142704e-05, + "loss": 0.9301, + "step": 38107 + }, + { + "epoch": 2.45, + "learning_rate": 1.1843964018119753e-05, + "loss": 0.9211, + "step": 38108 + }, + { + "epoch": 2.45, + "learning_rate": 1.1843749329096805e-05, + "loss": 0.9009, + "step": 38109 + }, + { + "epoch": 2.45, + "learning_rate": 1.1843534640073854e-05, + "loss": 0.8431, + "step": 38110 + }, + { + "epoch": 2.45, + "learning_rate": 1.1843319951050905e-05, + "loss": 0.8574, + "step": 38111 + }, + { + "epoch": 2.45, + "learning_rate": 1.1843105262027954e-05, + "loss": 0.8691, + "step": 38112 + }, + { + "epoch": 2.45, + "learning_rate": 1.1842890573005004e-05, + "loss": 0.8896, + "step": 38113 + }, + { + "epoch": 2.45, + "learning_rate": 1.1842675883982053e-05, + "loss": 0.9074, + "step": 38114 + }, + { + "epoch": 2.45, + "learning_rate": 1.1842461194959102e-05, + "loss": 0.9216, + "step": 38115 + }, + { + "epoch": 2.45, + "learning_rate": 1.1842246505936153e-05, + "loss": 0.9447, + "step": 38116 + }, + { + "epoch": 2.45, + "learning_rate": 1.1842031816913202e-05, + "loss": 0.8811, + "step": 38117 + }, + { + "epoch": 2.45, + "learning_rate": 1.1841817127890251e-05, + "loss": 0.8995, + "step": 38118 + }, + { + "epoch": 2.45, + "learning_rate": 1.1841602438867302e-05, + "loss": 0.8978, + "step": 38119 + }, + { + "epoch": 2.45, + "learning_rate": 1.184138774984435e-05, + "loss": 0.8963, + "step": 38120 + }, + { + "epoch": 2.45, + "learning_rate": 1.1841173060821401e-05, + "loss": 0.8755, + "step": 38121 + }, + { + "epoch": 2.45, + "learning_rate": 1.184095837179845e-05, + "loss": 0.9045, + "step": 38122 + }, + { + "epoch": 2.45, + "learning_rate": 1.1840743682775499e-05, + "loss": 0.8956, + "step": 38123 + }, + { + "epoch": 2.45, + "learning_rate": 1.1840528993752552e-05, + "loss": 0.8274, + "step": 38124 + }, + { + "epoch": 2.45, + "learning_rate": 1.18403143047296e-05, + "loss": 0.9635, + "step": 38125 + }, + { + "epoch": 2.45, + "learning_rate": 1.1840099615706651e-05, + "loss": 0.9283, + "step": 38126 + }, + { + "epoch": 2.45, + "learning_rate": 1.18398849266837e-05, + "loss": 0.9968, + "step": 38127 + }, + { + "epoch": 2.45, + "learning_rate": 1.1839670237660749e-05, + "loss": 0.9432, + "step": 38128 + }, + { + "epoch": 2.45, + "learning_rate": 1.18394555486378e-05, + "loss": 0.8344, + "step": 38129 + }, + { + "epoch": 2.45, + "learning_rate": 1.1839240859614849e-05, + "loss": 0.884, + "step": 38130 + }, + { + "epoch": 2.45, + "learning_rate": 1.18390261705919e-05, + "loss": 0.8544, + "step": 38131 + }, + { + "epoch": 2.45, + "learning_rate": 1.1838811481568948e-05, + "loss": 0.9764, + "step": 38132 + }, + { + "epoch": 2.45, + "learning_rate": 1.1838596792545997e-05, + "loss": 0.9215, + "step": 38133 + }, + { + "epoch": 2.45, + "learning_rate": 1.1838382103523048e-05, + "loss": 0.8715, + "step": 38134 + }, + { + "epoch": 2.45, + "learning_rate": 1.1838167414500097e-05, + "loss": 0.9176, + "step": 38135 + }, + { + "epoch": 2.45, + "learning_rate": 1.1837952725477146e-05, + "loss": 0.8887, + "step": 38136 + }, + { + "epoch": 2.45, + "learning_rate": 1.1837738036454197e-05, + "loss": 0.963, + "step": 38137 + }, + { + "epoch": 2.45, + "learning_rate": 1.1837523347431247e-05, + "loss": 1.0101, + "step": 38138 + }, + { + "epoch": 2.45, + "learning_rate": 1.1837308658408298e-05, + "loss": 0.8376, + "step": 38139 + }, + { + "epoch": 2.45, + "learning_rate": 1.1837093969385347e-05, + "loss": 0.8951, + "step": 38140 + }, + { + "epoch": 2.45, + "learning_rate": 1.1836879280362397e-05, + "loss": 0.9265, + "step": 38141 + }, + { + "epoch": 2.45, + "learning_rate": 1.1836664591339446e-05, + "loss": 0.891, + "step": 38142 + }, + { + "epoch": 2.45, + "learning_rate": 1.1836449902316495e-05, + "loss": 0.8222, + "step": 38143 + }, + { + "epoch": 2.45, + "learning_rate": 1.1836235213293546e-05, + "loss": 0.9397, + "step": 38144 + }, + { + "epoch": 2.45, + "learning_rate": 1.1836020524270595e-05, + "loss": 0.9503, + "step": 38145 + }, + { + "epoch": 2.45, + "learning_rate": 1.1835805835247644e-05, + "loss": 0.8632, + "step": 38146 + }, + { + "epoch": 2.45, + "learning_rate": 1.1835591146224695e-05, + "loss": 0.8357, + "step": 38147 + }, + { + "epoch": 2.45, + "learning_rate": 1.1835376457201744e-05, + "loss": 0.9441, + "step": 38148 + }, + { + "epoch": 2.45, + "learning_rate": 1.1835161768178793e-05, + "loss": 0.9904, + "step": 38149 + }, + { + "epoch": 2.45, + "learning_rate": 1.1834947079155843e-05, + "loss": 0.9137, + "step": 38150 + }, + { + "epoch": 2.45, + "learning_rate": 1.1834732390132892e-05, + "loss": 0.839, + "step": 38151 + }, + { + "epoch": 2.45, + "learning_rate": 1.1834517701109943e-05, + "loss": 0.8364, + "step": 38152 + }, + { + "epoch": 2.45, + "learning_rate": 1.1834303012086994e-05, + "loss": 0.9192, + "step": 38153 + }, + { + "epoch": 2.45, + "learning_rate": 1.1834088323064044e-05, + "loss": 0.8929, + "step": 38154 + }, + { + "epoch": 2.45, + "learning_rate": 1.1833873634041093e-05, + "loss": 0.8824, + "step": 38155 + }, + { + "epoch": 2.45, + "learning_rate": 1.1833658945018142e-05, + "loss": 0.8881, + "step": 38156 + }, + { + "epoch": 2.45, + "learning_rate": 1.1833444255995193e-05, + "loss": 0.8972, + "step": 38157 + }, + { + "epoch": 2.45, + "learning_rate": 1.1833229566972242e-05, + "loss": 0.8663, + "step": 38158 + }, + { + "epoch": 2.46, + "learning_rate": 1.183301487794929e-05, + "loss": 0.8894, + "step": 38159 + }, + { + "epoch": 2.46, + "learning_rate": 1.1832800188926341e-05, + "loss": 0.9405, + "step": 38160 + }, + { + "epoch": 2.46, + "learning_rate": 1.183258549990339e-05, + "loss": 0.8701, + "step": 38161 + }, + { + "epoch": 2.46, + "learning_rate": 1.1832370810880441e-05, + "loss": 0.8936, + "step": 38162 + }, + { + "epoch": 2.46, + "learning_rate": 1.183215612185749e-05, + "loss": 0.8786, + "step": 38163 + }, + { + "epoch": 2.46, + "learning_rate": 1.1831941432834539e-05, + "loss": 0.8405, + "step": 38164 + }, + { + "epoch": 2.46, + "learning_rate": 1.183172674381159e-05, + "loss": 0.8658, + "step": 38165 + }, + { + "epoch": 2.46, + "learning_rate": 1.1831512054788639e-05, + "loss": 0.8719, + "step": 38166 + }, + { + "epoch": 2.46, + "learning_rate": 1.1831297365765691e-05, + "loss": 0.8571, + "step": 38167 + }, + { + "epoch": 2.46, + "learning_rate": 1.183108267674274e-05, + "loss": 0.9307, + "step": 38168 + }, + { + "epoch": 2.46, + "learning_rate": 1.1830867987719789e-05, + "loss": 0.8647, + "step": 38169 + }, + { + "epoch": 2.46, + "learning_rate": 1.183065329869684e-05, + "loss": 0.8807, + "step": 38170 + }, + { + "epoch": 2.46, + "learning_rate": 1.1830438609673888e-05, + "loss": 0.8194, + "step": 38171 + }, + { + "epoch": 2.46, + "learning_rate": 1.1830223920650939e-05, + "loss": 0.9224, + "step": 38172 + }, + { + "epoch": 2.46, + "learning_rate": 1.1830009231627988e-05, + "loss": 0.8928, + "step": 38173 + }, + { + "epoch": 2.46, + "learning_rate": 1.1829794542605037e-05, + "loss": 0.8728, + "step": 38174 + }, + { + "epoch": 2.46, + "learning_rate": 1.1829579853582088e-05, + "loss": 0.9289, + "step": 38175 + }, + { + "epoch": 2.46, + "learning_rate": 1.1829365164559137e-05, + "loss": 0.8901, + "step": 38176 + }, + { + "epoch": 2.46, + "learning_rate": 1.1829150475536186e-05, + "loss": 0.925, + "step": 38177 + }, + { + "epoch": 2.46, + "learning_rate": 1.1828935786513236e-05, + "loss": 0.9409, + "step": 38178 + }, + { + "epoch": 2.46, + "learning_rate": 1.1828721097490285e-05, + "loss": 0.8948, + "step": 38179 + }, + { + "epoch": 2.46, + "learning_rate": 1.1828506408467336e-05, + "loss": 0.9045, + "step": 38180 + }, + { + "epoch": 2.46, + "learning_rate": 1.1828291719444385e-05, + "loss": 0.8635, + "step": 38181 + }, + { + "epoch": 2.46, + "learning_rate": 1.1828077030421437e-05, + "loss": 0.8661, + "step": 38182 + }, + { + "epoch": 2.46, + "learning_rate": 1.1827862341398486e-05, + "loss": 0.8846, + "step": 38183 + }, + { + "epoch": 2.46, + "learning_rate": 1.1827647652375535e-05, + "loss": 0.8406, + "step": 38184 + }, + { + "epoch": 2.46, + "learning_rate": 1.1827432963352586e-05, + "loss": 0.9513, + "step": 38185 + }, + { + "epoch": 2.46, + "learning_rate": 1.1827218274329635e-05, + "loss": 0.9126, + "step": 38186 + }, + { + "epoch": 2.46, + "learning_rate": 1.1827003585306684e-05, + "loss": 0.9018, + "step": 38187 + }, + { + "epoch": 2.46, + "learning_rate": 1.1826788896283734e-05, + "loss": 0.8833, + "step": 38188 + }, + { + "epoch": 2.46, + "learning_rate": 1.1826574207260783e-05, + "loss": 0.9013, + "step": 38189 + }, + { + "epoch": 2.46, + "learning_rate": 1.1826359518237834e-05, + "loss": 0.8868, + "step": 38190 + }, + { + "epoch": 2.46, + "learning_rate": 1.1826144829214883e-05, + "loss": 0.8055, + "step": 38191 + }, + { + "epoch": 2.46, + "learning_rate": 1.1825930140191932e-05, + "loss": 0.9208, + "step": 38192 + }, + { + "epoch": 2.46, + "learning_rate": 1.1825715451168983e-05, + "loss": 0.8768, + "step": 38193 + }, + { + "epoch": 2.46, + "learning_rate": 1.1825500762146032e-05, + "loss": 0.8621, + "step": 38194 + }, + { + "epoch": 2.46, + "learning_rate": 1.182528607312308e-05, + "loss": 0.9149, + "step": 38195 + }, + { + "epoch": 2.46, + "learning_rate": 1.1825071384100133e-05, + "loss": 0.8946, + "step": 38196 + }, + { + "epoch": 2.46, + "learning_rate": 1.1824856695077182e-05, + "loss": 0.8602, + "step": 38197 + }, + { + "epoch": 2.46, + "learning_rate": 1.1824642006054233e-05, + "loss": 0.9443, + "step": 38198 + }, + { + "epoch": 2.46, + "learning_rate": 1.1824427317031281e-05, + "loss": 0.898, + "step": 38199 + }, + { + "epoch": 2.46, + "learning_rate": 1.1824212628008332e-05, + "loss": 0.8303, + "step": 38200 + }, + { + "epoch": 2.46, + "learning_rate": 1.1823997938985381e-05, + "loss": 0.9458, + "step": 38201 + }, + { + "epoch": 2.46, + "learning_rate": 1.182378324996243e-05, + "loss": 0.853, + "step": 38202 + }, + { + "epoch": 2.46, + "learning_rate": 1.182356856093948e-05, + "loss": 0.866, + "step": 38203 + }, + { + "epoch": 2.46, + "learning_rate": 1.182335387191653e-05, + "loss": 0.8908, + "step": 38204 + }, + { + "epoch": 2.46, + "learning_rate": 1.1823139182893579e-05, + "loss": 0.9347, + "step": 38205 + }, + { + "epoch": 2.46, + "learning_rate": 1.182292449387063e-05, + "loss": 0.9037, + "step": 38206 + }, + { + "epoch": 2.46, + "learning_rate": 1.1822709804847678e-05, + "loss": 0.8478, + "step": 38207 + }, + { + "epoch": 2.46, + "learning_rate": 1.1822495115824727e-05, + "loss": 0.8271, + "step": 38208 + }, + { + "epoch": 2.46, + "learning_rate": 1.1822280426801778e-05, + "loss": 0.8255, + "step": 38209 + }, + { + "epoch": 2.46, + "learning_rate": 1.1822065737778827e-05, + "loss": 0.9583, + "step": 38210 + }, + { + "epoch": 2.46, + "learning_rate": 1.182185104875588e-05, + "loss": 0.8932, + "step": 38211 + }, + { + "epoch": 2.46, + "learning_rate": 1.1821636359732928e-05, + "loss": 0.8471, + "step": 38212 + }, + { + "epoch": 2.46, + "learning_rate": 1.1821421670709979e-05, + "loss": 0.964, + "step": 38213 + }, + { + "epoch": 2.46, + "learning_rate": 1.1821206981687028e-05, + "loss": 0.9246, + "step": 38214 + }, + { + "epoch": 2.46, + "learning_rate": 1.1820992292664077e-05, + "loss": 0.8665, + "step": 38215 + }, + { + "epoch": 2.46, + "learning_rate": 1.1820777603641127e-05, + "loss": 0.8498, + "step": 38216 + }, + { + "epoch": 2.46, + "learning_rate": 1.1820562914618176e-05, + "loss": 0.8609, + "step": 38217 + }, + { + "epoch": 2.46, + "learning_rate": 1.1820348225595227e-05, + "loss": 0.9565, + "step": 38218 + }, + { + "epoch": 2.46, + "learning_rate": 1.1820133536572276e-05, + "loss": 0.921, + "step": 38219 + }, + { + "epoch": 2.46, + "learning_rate": 1.1819918847549325e-05, + "loss": 0.908, + "step": 38220 + }, + { + "epoch": 2.46, + "learning_rate": 1.1819704158526376e-05, + "loss": 0.8741, + "step": 38221 + }, + { + "epoch": 2.46, + "learning_rate": 1.1819489469503425e-05, + "loss": 0.8401, + "step": 38222 + }, + { + "epoch": 2.46, + "learning_rate": 1.1819274780480474e-05, + "loss": 0.8809, + "step": 38223 + }, + { + "epoch": 2.46, + "learning_rate": 1.1819060091457524e-05, + "loss": 0.8606, + "step": 38224 + }, + { + "epoch": 2.46, + "learning_rate": 1.1818845402434575e-05, + "loss": 0.9081, + "step": 38225 + }, + { + "epoch": 2.46, + "learning_rate": 1.1818630713411626e-05, + "loss": 0.917, + "step": 38226 + }, + { + "epoch": 2.46, + "learning_rate": 1.1818416024388675e-05, + "loss": 0.8759, + "step": 38227 + }, + { + "epoch": 2.46, + "learning_rate": 1.1818201335365725e-05, + "loss": 0.8793, + "step": 38228 + }, + { + "epoch": 2.46, + "learning_rate": 1.1817986646342774e-05, + "loss": 0.8804, + "step": 38229 + }, + { + "epoch": 2.46, + "learning_rate": 1.1817771957319823e-05, + "loss": 0.8714, + "step": 38230 + }, + { + "epoch": 2.46, + "learning_rate": 1.1817557268296874e-05, + "loss": 0.9216, + "step": 38231 + }, + { + "epoch": 2.46, + "learning_rate": 1.1817342579273923e-05, + "loss": 0.8748, + "step": 38232 + }, + { + "epoch": 2.46, + "learning_rate": 1.1817127890250972e-05, + "loss": 0.8764, + "step": 38233 + }, + { + "epoch": 2.46, + "learning_rate": 1.1816913201228022e-05, + "loss": 0.9045, + "step": 38234 + }, + { + "epoch": 2.46, + "learning_rate": 1.1816698512205071e-05, + "loss": 0.8607, + "step": 38235 + }, + { + "epoch": 2.46, + "learning_rate": 1.181648382318212e-05, + "loss": 0.8945, + "step": 38236 + }, + { + "epoch": 2.46, + "learning_rate": 1.1816269134159171e-05, + "loss": 0.8924, + "step": 38237 + }, + { + "epoch": 2.46, + "learning_rate": 1.181605444513622e-05, + "loss": 0.9481, + "step": 38238 + }, + { + "epoch": 2.46, + "learning_rate": 1.181583975611327e-05, + "loss": 0.9379, + "step": 38239 + }, + { + "epoch": 2.46, + "learning_rate": 1.1815625067090321e-05, + "loss": 0.963, + "step": 38240 + }, + { + "epoch": 2.46, + "learning_rate": 1.1815410378067372e-05, + "loss": 0.8633, + "step": 38241 + }, + { + "epoch": 2.46, + "learning_rate": 1.1815195689044421e-05, + "loss": 0.8681, + "step": 38242 + }, + { + "epoch": 2.46, + "learning_rate": 1.181498100002147e-05, + "loss": 0.8981, + "step": 38243 + }, + { + "epoch": 2.46, + "learning_rate": 1.181476631099852e-05, + "loss": 0.8962, + "step": 38244 + }, + { + "epoch": 2.46, + "learning_rate": 1.181455162197557e-05, + "loss": 0.8767, + "step": 38245 + }, + { + "epoch": 2.46, + "learning_rate": 1.1814336932952618e-05, + "loss": 0.8696, + "step": 38246 + }, + { + "epoch": 2.46, + "learning_rate": 1.1814122243929669e-05, + "loss": 0.8979, + "step": 38247 + }, + { + "epoch": 2.46, + "learning_rate": 1.1813907554906718e-05, + "loss": 0.9054, + "step": 38248 + }, + { + "epoch": 2.46, + "learning_rate": 1.1813692865883769e-05, + "loss": 0.9259, + "step": 38249 + }, + { + "epoch": 2.46, + "learning_rate": 1.1813478176860818e-05, + "loss": 0.9286, + "step": 38250 + }, + { + "epoch": 2.46, + "learning_rate": 1.1813263487837867e-05, + "loss": 0.8387, + "step": 38251 + }, + { + "epoch": 2.46, + "learning_rate": 1.1813048798814917e-05, + "loss": 0.9473, + "step": 38252 + }, + { + "epoch": 2.46, + "learning_rate": 1.1812834109791966e-05, + "loss": 0.8911, + "step": 38253 + }, + { + "epoch": 2.46, + "learning_rate": 1.1812619420769019e-05, + "loss": 1.0038, + "step": 38254 + }, + { + "epoch": 2.46, + "learning_rate": 1.1812404731746068e-05, + "loss": 0.921, + "step": 38255 + }, + { + "epoch": 2.46, + "learning_rate": 1.1812190042723117e-05, + "loss": 0.8667, + "step": 38256 + }, + { + "epoch": 2.46, + "learning_rate": 1.1811975353700167e-05, + "loss": 0.9864, + "step": 38257 + }, + { + "epoch": 2.46, + "learning_rate": 1.1811760664677216e-05, + "loss": 0.8972, + "step": 38258 + }, + { + "epoch": 2.46, + "learning_rate": 1.1811545975654267e-05, + "loss": 0.8828, + "step": 38259 + }, + { + "epoch": 2.46, + "learning_rate": 1.1811331286631316e-05, + "loss": 0.9227, + "step": 38260 + }, + { + "epoch": 2.46, + "learning_rate": 1.1811116597608365e-05, + "loss": 0.9635, + "step": 38261 + }, + { + "epoch": 2.46, + "learning_rate": 1.1810901908585415e-05, + "loss": 0.9009, + "step": 38262 + }, + { + "epoch": 2.46, + "learning_rate": 1.1810687219562464e-05, + "loss": 0.8651, + "step": 38263 + }, + { + "epoch": 2.46, + "learning_rate": 1.1810472530539513e-05, + "loss": 0.9474, + "step": 38264 + }, + { + "epoch": 2.46, + "learning_rate": 1.1810257841516564e-05, + "loss": 0.8365, + "step": 38265 + }, + { + "epoch": 2.46, + "learning_rate": 1.1810043152493613e-05, + "loss": 0.8649, + "step": 38266 + }, + { + "epoch": 2.46, + "learning_rate": 1.1809828463470664e-05, + "loss": 0.8797, + "step": 38267 + }, + { + "epoch": 2.46, + "learning_rate": 1.1809613774447713e-05, + "loss": 0.875, + "step": 38268 + }, + { + "epoch": 2.46, + "learning_rate": 1.1809399085424765e-05, + "loss": 0.9144, + "step": 38269 + }, + { + "epoch": 2.46, + "learning_rate": 1.1809184396401814e-05, + "loss": 0.8787, + "step": 38270 + }, + { + "epoch": 2.46, + "learning_rate": 1.1808969707378863e-05, + "loss": 0.9235, + "step": 38271 + }, + { + "epoch": 2.46, + "learning_rate": 1.1808755018355914e-05, + "loss": 0.8784, + "step": 38272 + }, + { + "epoch": 2.46, + "learning_rate": 1.1808540329332962e-05, + "loss": 0.9004, + "step": 38273 + }, + { + "epoch": 2.46, + "learning_rate": 1.1808325640310011e-05, + "loss": 0.8712, + "step": 38274 + }, + { + "epoch": 2.46, + "learning_rate": 1.1808110951287062e-05, + "loss": 0.8764, + "step": 38275 + }, + { + "epoch": 2.46, + "learning_rate": 1.1807896262264111e-05, + "loss": 0.8862, + "step": 38276 + }, + { + "epoch": 2.46, + "learning_rate": 1.1807681573241162e-05, + "loss": 0.8502, + "step": 38277 + }, + { + "epoch": 2.46, + "learning_rate": 1.180746688421821e-05, + "loss": 0.8855, + "step": 38278 + }, + { + "epoch": 2.46, + "learning_rate": 1.180725219519526e-05, + "loss": 0.9245, + "step": 38279 + }, + { + "epoch": 2.46, + "learning_rate": 1.180703750617231e-05, + "loss": 0.8983, + "step": 38280 + }, + { + "epoch": 2.46, + "learning_rate": 1.180682281714936e-05, + "loss": 0.8945, + "step": 38281 + }, + { + "epoch": 2.46, + "learning_rate": 1.1806608128126408e-05, + "loss": 0.9185, + "step": 38282 + }, + { + "epoch": 2.46, + "learning_rate": 1.180639343910346e-05, + "loss": 0.9111, + "step": 38283 + }, + { + "epoch": 2.46, + "learning_rate": 1.180617875008051e-05, + "loss": 0.8721, + "step": 38284 + }, + { + "epoch": 2.46, + "learning_rate": 1.180596406105756e-05, + "loss": 0.865, + "step": 38285 + }, + { + "epoch": 2.46, + "learning_rate": 1.180574937203461e-05, + "loss": 0.8926, + "step": 38286 + }, + { + "epoch": 2.46, + "learning_rate": 1.180553468301166e-05, + "loss": 0.9796, + "step": 38287 + }, + { + "epoch": 2.46, + "learning_rate": 1.1805319993988709e-05, + "loss": 0.8899, + "step": 38288 + }, + { + "epoch": 2.46, + "learning_rate": 1.1805105304965758e-05, + "loss": 0.8548, + "step": 38289 + }, + { + "epoch": 2.46, + "learning_rate": 1.1804890615942808e-05, + "loss": 0.9234, + "step": 38290 + }, + { + "epoch": 2.46, + "learning_rate": 1.1804675926919857e-05, + "loss": 0.9424, + "step": 38291 + }, + { + "epoch": 2.46, + "learning_rate": 1.1804461237896906e-05, + "loss": 0.9576, + "step": 38292 + }, + { + "epoch": 2.46, + "learning_rate": 1.1804246548873957e-05, + "loss": 0.9749, + "step": 38293 + }, + { + "epoch": 2.46, + "learning_rate": 1.1804031859851006e-05, + "loss": 0.9129, + "step": 38294 + }, + { + "epoch": 2.46, + "learning_rate": 1.1803817170828055e-05, + "loss": 0.9017, + "step": 38295 + }, + { + "epoch": 2.46, + "learning_rate": 1.1803602481805106e-05, + "loss": 0.8728, + "step": 38296 + }, + { + "epoch": 2.46, + "learning_rate": 1.1803387792782155e-05, + "loss": 0.92, + "step": 38297 + }, + { + "epoch": 2.46, + "learning_rate": 1.1803173103759207e-05, + "loss": 0.9155, + "step": 38298 + }, + { + "epoch": 2.46, + "learning_rate": 1.1802958414736256e-05, + "loss": 0.8714, + "step": 38299 + }, + { + "epoch": 2.46, + "learning_rate": 1.1802743725713307e-05, + "loss": 0.9783, + "step": 38300 + }, + { + "epoch": 2.46, + "learning_rate": 1.1802529036690356e-05, + "loss": 0.8783, + "step": 38301 + }, + { + "epoch": 2.46, + "learning_rate": 1.1802314347667404e-05, + "loss": 0.9042, + "step": 38302 + }, + { + "epoch": 2.46, + "learning_rate": 1.1802099658644455e-05, + "loss": 0.9114, + "step": 38303 + }, + { + "epoch": 2.46, + "learning_rate": 1.1801884969621504e-05, + "loss": 0.9343, + "step": 38304 + }, + { + "epoch": 2.46, + "learning_rate": 1.1801670280598553e-05, + "loss": 0.9416, + "step": 38305 + }, + { + "epoch": 2.46, + "learning_rate": 1.1801455591575604e-05, + "loss": 0.8558, + "step": 38306 + }, + { + "epoch": 2.46, + "learning_rate": 1.1801240902552653e-05, + "loss": 0.9015, + "step": 38307 + }, + { + "epoch": 2.46, + "learning_rate": 1.1801026213529703e-05, + "loss": 0.9274, + "step": 38308 + }, + { + "epoch": 2.46, + "learning_rate": 1.1800811524506752e-05, + "loss": 0.9541, + "step": 38309 + }, + { + "epoch": 2.46, + "learning_rate": 1.1800596835483801e-05, + "loss": 0.9189, + "step": 38310 + }, + { + "epoch": 2.46, + "learning_rate": 1.1800382146460852e-05, + "loss": 0.8683, + "step": 38311 + }, + { + "epoch": 2.46, + "learning_rate": 1.1800167457437903e-05, + "loss": 0.8954, + "step": 38312 + }, + { + "epoch": 2.46, + "learning_rate": 1.1799952768414953e-05, + "loss": 0.9, + "step": 38313 + }, + { + "epoch": 2.47, + "learning_rate": 1.1799738079392002e-05, + "loss": 0.8132, + "step": 38314 + }, + { + "epoch": 2.47, + "learning_rate": 1.1799523390369051e-05, + "loss": 0.9266, + "step": 38315 + }, + { + "epoch": 2.47, + "learning_rate": 1.1799308701346102e-05, + "loss": 0.8752, + "step": 38316 + }, + { + "epoch": 2.47, + "learning_rate": 1.179909401232315e-05, + "loss": 0.941, + "step": 38317 + }, + { + "epoch": 2.47, + "learning_rate": 1.1798879323300201e-05, + "loss": 0.8948, + "step": 38318 + }, + { + "epoch": 2.47, + "learning_rate": 1.179866463427725e-05, + "loss": 0.8518, + "step": 38319 + }, + { + "epoch": 2.47, + "learning_rate": 1.17984499452543e-05, + "loss": 0.9596, + "step": 38320 + }, + { + "epoch": 2.47, + "learning_rate": 1.179823525623135e-05, + "loss": 0.9266, + "step": 38321 + }, + { + "epoch": 2.47, + "learning_rate": 1.1798020567208399e-05, + "loss": 0.8831, + "step": 38322 + }, + { + "epoch": 2.47, + "learning_rate": 1.1797805878185448e-05, + "loss": 0.8682, + "step": 38323 + }, + { + "epoch": 2.47, + "learning_rate": 1.1797591189162499e-05, + "loss": 0.9589, + "step": 38324 + }, + { + "epoch": 2.47, + "learning_rate": 1.1797376500139548e-05, + "loss": 0.9157, + "step": 38325 + }, + { + "epoch": 2.47, + "learning_rate": 1.1797161811116598e-05, + "loss": 0.9188, + "step": 38326 + }, + { + "epoch": 2.47, + "learning_rate": 1.1796947122093649e-05, + "loss": 0.7795, + "step": 38327 + }, + { + "epoch": 2.47, + "learning_rate": 1.17967324330707e-05, + "loss": 0.8916, + "step": 38328 + }, + { + "epoch": 2.47, + "learning_rate": 1.1796517744047749e-05, + "loss": 0.89, + "step": 38329 + }, + { + "epoch": 2.47, + "learning_rate": 1.1796303055024798e-05, + "loss": 0.8259, + "step": 38330 + }, + { + "epoch": 2.47, + "learning_rate": 1.1796088366001848e-05, + "loss": 0.9164, + "step": 38331 + }, + { + "epoch": 2.47, + "learning_rate": 1.1795873676978897e-05, + "loss": 0.9395, + "step": 38332 + }, + { + "epoch": 2.47, + "learning_rate": 1.1795658987955946e-05, + "loss": 0.9589, + "step": 38333 + }, + { + "epoch": 2.47, + "learning_rate": 1.1795444298932997e-05, + "loss": 0.8851, + "step": 38334 + }, + { + "epoch": 2.47, + "learning_rate": 1.1795229609910046e-05, + "loss": 0.9461, + "step": 38335 + }, + { + "epoch": 2.47, + "learning_rate": 1.1795014920887096e-05, + "loss": 0.8821, + "step": 38336 + }, + { + "epoch": 2.47, + "learning_rate": 1.1794800231864145e-05, + "loss": 0.9099, + "step": 38337 + }, + { + "epoch": 2.47, + "learning_rate": 1.1794585542841194e-05, + "loss": 0.9396, + "step": 38338 + }, + { + "epoch": 2.47, + "learning_rate": 1.1794370853818245e-05, + "loss": 0.8852, + "step": 38339 + }, + { + "epoch": 2.47, + "learning_rate": 1.1794156164795294e-05, + "loss": 0.9654, + "step": 38340 + }, + { + "epoch": 2.47, + "learning_rate": 1.1793941475772346e-05, + "loss": 0.9501, + "step": 38341 + }, + { + "epoch": 2.47, + "learning_rate": 1.1793726786749395e-05, + "loss": 0.8985, + "step": 38342 + }, + { + "epoch": 2.47, + "learning_rate": 1.1793512097726444e-05, + "loss": 0.9296, + "step": 38343 + }, + { + "epoch": 2.47, + "learning_rate": 1.1793297408703495e-05, + "loss": 0.8946, + "step": 38344 + }, + { + "epoch": 2.47, + "learning_rate": 1.1793082719680544e-05, + "loss": 0.8577, + "step": 38345 + }, + { + "epoch": 2.47, + "learning_rate": 1.1792868030657595e-05, + "loss": 0.9487, + "step": 38346 + }, + { + "epoch": 2.47, + "learning_rate": 1.1792653341634643e-05, + "loss": 0.9297, + "step": 38347 + }, + { + "epoch": 2.47, + "learning_rate": 1.1792438652611692e-05, + "loss": 0.8537, + "step": 38348 + }, + { + "epoch": 2.47, + "learning_rate": 1.1792223963588743e-05, + "loss": 0.9007, + "step": 38349 + }, + { + "epoch": 2.47, + "learning_rate": 1.1792009274565792e-05, + "loss": 0.8534, + "step": 38350 + }, + { + "epoch": 2.47, + "learning_rate": 1.1791794585542841e-05, + "loss": 0.9738, + "step": 38351 + }, + { + "epoch": 2.47, + "learning_rate": 1.1791579896519892e-05, + "loss": 0.8468, + "step": 38352 + }, + { + "epoch": 2.47, + "learning_rate": 1.179136520749694e-05, + "loss": 0.9592, + "step": 38353 + }, + { + "epoch": 2.47, + "learning_rate": 1.1791150518473991e-05, + "loss": 0.9047, + "step": 38354 + }, + { + "epoch": 2.47, + "learning_rate": 1.179093582945104e-05, + "loss": 0.8754, + "step": 38355 + }, + { + "epoch": 2.47, + "learning_rate": 1.1790721140428093e-05, + "loss": 0.9225, + "step": 38356 + }, + { + "epoch": 2.47, + "learning_rate": 1.1790506451405142e-05, + "loss": 0.8329, + "step": 38357 + }, + { + "epoch": 2.47, + "learning_rate": 1.179029176238219e-05, + "loss": 0.9114, + "step": 38358 + }, + { + "epoch": 2.47, + "learning_rate": 1.1790077073359241e-05, + "loss": 0.8951, + "step": 38359 + }, + { + "epoch": 2.47, + "learning_rate": 1.178986238433629e-05, + "loss": 0.8755, + "step": 38360 + }, + { + "epoch": 2.47, + "learning_rate": 1.178964769531334e-05, + "loss": 0.9544, + "step": 38361 + }, + { + "epoch": 2.47, + "learning_rate": 1.178943300629039e-05, + "loss": 0.9197, + "step": 38362 + }, + { + "epoch": 2.47, + "learning_rate": 1.1789218317267439e-05, + "loss": 0.8793, + "step": 38363 + }, + { + "epoch": 2.47, + "learning_rate": 1.178900362824449e-05, + "loss": 0.9204, + "step": 38364 + }, + { + "epoch": 2.47, + "learning_rate": 1.1788788939221538e-05, + "loss": 0.9166, + "step": 38365 + }, + { + "epoch": 2.47, + "learning_rate": 1.1788574250198587e-05, + "loss": 0.8691, + "step": 38366 + }, + { + "epoch": 2.47, + "learning_rate": 1.1788359561175638e-05, + "loss": 0.939, + "step": 38367 + }, + { + "epoch": 2.47, + "learning_rate": 1.1788144872152687e-05, + "loss": 0.9339, + "step": 38368 + }, + { + "epoch": 2.47, + "learning_rate": 1.1787930183129736e-05, + "loss": 0.8466, + "step": 38369 + }, + { + "epoch": 2.47, + "learning_rate": 1.1787715494106788e-05, + "loss": 0.9212, + "step": 38370 + }, + { + "epoch": 2.47, + "learning_rate": 1.1787500805083837e-05, + "loss": 0.9307, + "step": 38371 + }, + { + "epoch": 2.47, + "learning_rate": 1.1787286116060888e-05, + "loss": 0.9079, + "step": 38372 + }, + { + "epoch": 2.47, + "learning_rate": 1.1787071427037937e-05, + "loss": 0.8547, + "step": 38373 + }, + { + "epoch": 2.47, + "learning_rate": 1.1786856738014988e-05, + "loss": 0.8721, + "step": 38374 + }, + { + "epoch": 2.47, + "learning_rate": 1.1786642048992037e-05, + "loss": 0.8958, + "step": 38375 + }, + { + "epoch": 2.47, + "learning_rate": 1.1786427359969085e-05, + "loss": 0.8862, + "step": 38376 + }, + { + "epoch": 2.47, + "learning_rate": 1.1786212670946136e-05, + "loss": 0.9119, + "step": 38377 + }, + { + "epoch": 2.47, + "learning_rate": 1.1785997981923185e-05, + "loss": 0.8926, + "step": 38378 + }, + { + "epoch": 2.47, + "learning_rate": 1.1785783292900234e-05, + "loss": 0.8433, + "step": 38379 + }, + { + "epoch": 2.47, + "learning_rate": 1.1785568603877285e-05, + "loss": 0.9042, + "step": 38380 + }, + { + "epoch": 2.47, + "learning_rate": 1.1785353914854334e-05, + "loss": 0.8818, + "step": 38381 + }, + { + "epoch": 2.47, + "learning_rate": 1.1785139225831383e-05, + "loss": 0.8924, + "step": 38382 + }, + { + "epoch": 2.47, + "learning_rate": 1.1784924536808433e-05, + "loss": 0.9005, + "step": 38383 + }, + { + "epoch": 2.47, + "learning_rate": 1.1784709847785482e-05, + "loss": 0.9414, + "step": 38384 + }, + { + "epoch": 2.47, + "learning_rate": 1.1784495158762535e-05, + "loss": 0.8537, + "step": 38385 + }, + { + "epoch": 2.47, + "learning_rate": 1.1784280469739584e-05, + "loss": 0.9553, + "step": 38386 + }, + { + "epoch": 2.47, + "learning_rate": 1.1784065780716634e-05, + "loss": 0.9099, + "step": 38387 + }, + { + "epoch": 2.47, + "learning_rate": 1.1783851091693683e-05, + "loss": 0.8777, + "step": 38388 + }, + { + "epoch": 2.47, + "learning_rate": 1.1783636402670732e-05, + "loss": 0.8773, + "step": 38389 + }, + { + "epoch": 2.47, + "learning_rate": 1.1783421713647783e-05, + "loss": 0.9256, + "step": 38390 + }, + { + "epoch": 2.47, + "learning_rate": 1.1783207024624832e-05, + "loss": 0.871, + "step": 38391 + }, + { + "epoch": 2.47, + "learning_rate": 1.178299233560188e-05, + "loss": 0.888, + "step": 38392 + }, + { + "epoch": 2.47, + "learning_rate": 1.1782777646578931e-05, + "loss": 0.8829, + "step": 38393 + }, + { + "epoch": 2.47, + "learning_rate": 1.178256295755598e-05, + "loss": 0.8357, + "step": 38394 + }, + { + "epoch": 2.47, + "learning_rate": 1.1782348268533031e-05, + "loss": 0.9211, + "step": 38395 + }, + { + "epoch": 2.47, + "learning_rate": 1.178213357951008e-05, + "loss": 0.9417, + "step": 38396 + }, + { + "epoch": 2.47, + "learning_rate": 1.1781918890487129e-05, + "loss": 0.8949, + "step": 38397 + }, + { + "epoch": 2.47, + "learning_rate": 1.178170420146418e-05, + "loss": 0.9574, + "step": 38398 + }, + { + "epoch": 2.47, + "learning_rate": 1.1781489512441229e-05, + "loss": 0.8723, + "step": 38399 + }, + { + "epoch": 2.47, + "learning_rate": 1.1781274823418281e-05, + "loss": 0.9043, + "step": 38400 + }, + { + "epoch": 2.47, + "learning_rate": 1.178106013439533e-05, + "loss": 0.9958, + "step": 38401 + }, + { + "epoch": 2.47, + "learning_rate": 1.1780845445372379e-05, + "loss": 1.0168, + "step": 38402 + }, + { + "epoch": 2.47, + "learning_rate": 1.178063075634943e-05, + "loss": 0.8636, + "step": 38403 + }, + { + "epoch": 2.47, + "learning_rate": 1.1780416067326479e-05, + "loss": 0.8999, + "step": 38404 + }, + { + "epoch": 2.47, + "learning_rate": 1.178020137830353e-05, + "loss": 0.941, + "step": 38405 + }, + { + "epoch": 2.47, + "learning_rate": 1.1779986689280578e-05, + "loss": 0.9078, + "step": 38406 + }, + { + "epoch": 2.47, + "learning_rate": 1.1779772000257627e-05, + "loss": 0.8849, + "step": 38407 + }, + { + "epoch": 2.47, + "learning_rate": 1.1779557311234678e-05, + "loss": 0.9543, + "step": 38408 + }, + { + "epoch": 2.47, + "learning_rate": 1.1779342622211727e-05, + "loss": 0.8797, + "step": 38409 + }, + { + "epoch": 2.47, + "learning_rate": 1.1779127933188776e-05, + "loss": 0.8782, + "step": 38410 + }, + { + "epoch": 2.47, + "learning_rate": 1.1778913244165826e-05, + "loss": 0.8977, + "step": 38411 + }, + { + "epoch": 2.47, + "learning_rate": 1.1778698555142875e-05, + "loss": 0.8653, + "step": 38412 + }, + { + "epoch": 2.47, + "learning_rate": 1.1778483866119926e-05, + "loss": 0.9644, + "step": 38413 + }, + { + "epoch": 2.47, + "learning_rate": 1.1778269177096977e-05, + "loss": 0.8813, + "step": 38414 + }, + { + "epoch": 2.47, + "learning_rate": 1.1778054488074027e-05, + "loss": 0.8724, + "step": 38415 + }, + { + "epoch": 2.47, + "learning_rate": 1.1777839799051076e-05, + "loss": 0.9201, + "step": 38416 + }, + { + "epoch": 2.47, + "learning_rate": 1.1777625110028125e-05, + "loss": 0.8639, + "step": 38417 + }, + { + "epoch": 2.47, + "learning_rate": 1.1777410421005176e-05, + "loss": 0.9054, + "step": 38418 + }, + { + "epoch": 2.47, + "learning_rate": 1.1777195731982225e-05, + "loss": 0.8547, + "step": 38419 + }, + { + "epoch": 2.47, + "learning_rate": 1.1776981042959274e-05, + "loss": 0.9681, + "step": 38420 + }, + { + "epoch": 2.47, + "learning_rate": 1.1776766353936324e-05, + "loss": 0.9224, + "step": 38421 + }, + { + "epoch": 2.47, + "learning_rate": 1.1776551664913373e-05, + "loss": 0.8484, + "step": 38422 + }, + { + "epoch": 2.47, + "learning_rate": 1.1776336975890424e-05, + "loss": 0.9549, + "step": 38423 + }, + { + "epoch": 2.47, + "learning_rate": 1.1776122286867473e-05, + "loss": 0.8767, + "step": 38424 + }, + { + "epoch": 2.47, + "learning_rate": 1.1775907597844522e-05, + "loss": 0.8969, + "step": 38425 + }, + { + "epoch": 2.47, + "learning_rate": 1.1775692908821573e-05, + "loss": 0.9097, + "step": 38426 + }, + { + "epoch": 2.47, + "learning_rate": 1.1775478219798622e-05, + "loss": 0.9151, + "step": 38427 + }, + { + "epoch": 2.47, + "learning_rate": 1.177526353077567e-05, + "loss": 0.8973, + "step": 38428 + }, + { + "epoch": 2.47, + "learning_rate": 1.1775048841752723e-05, + "loss": 0.8726, + "step": 38429 + }, + { + "epoch": 2.47, + "learning_rate": 1.1774834152729772e-05, + "loss": 0.9162, + "step": 38430 + }, + { + "epoch": 2.47, + "learning_rate": 1.1774619463706823e-05, + "loss": 0.8603, + "step": 38431 + }, + { + "epoch": 2.47, + "learning_rate": 1.1774404774683872e-05, + "loss": 0.8185, + "step": 38432 + }, + { + "epoch": 2.47, + "learning_rate": 1.1774190085660922e-05, + "loss": 0.8299, + "step": 38433 + }, + { + "epoch": 2.47, + "learning_rate": 1.1773975396637971e-05, + "loss": 0.8534, + "step": 38434 + }, + { + "epoch": 2.47, + "learning_rate": 1.177376070761502e-05, + "loss": 0.8835, + "step": 38435 + }, + { + "epoch": 2.47, + "learning_rate": 1.177354601859207e-05, + "loss": 0.8616, + "step": 38436 + }, + { + "epoch": 2.47, + "learning_rate": 1.177333132956912e-05, + "loss": 0.9104, + "step": 38437 + }, + { + "epoch": 2.47, + "learning_rate": 1.1773116640546169e-05, + "loss": 0.9345, + "step": 38438 + }, + { + "epoch": 2.47, + "learning_rate": 1.177290195152322e-05, + "loss": 0.8831, + "step": 38439 + }, + { + "epoch": 2.47, + "learning_rate": 1.1772687262500268e-05, + "loss": 0.8843, + "step": 38440 + }, + { + "epoch": 2.47, + "learning_rate": 1.1772472573477319e-05, + "loss": 0.9579, + "step": 38441 + }, + { + "epoch": 2.47, + "learning_rate": 1.1772257884454368e-05, + "loss": 0.8488, + "step": 38442 + }, + { + "epoch": 2.47, + "learning_rate": 1.177204319543142e-05, + "loss": 0.8785, + "step": 38443 + }, + { + "epoch": 2.47, + "learning_rate": 1.177182850640847e-05, + "loss": 0.981, + "step": 38444 + }, + { + "epoch": 2.47, + "learning_rate": 1.1771613817385518e-05, + "loss": 0.8492, + "step": 38445 + }, + { + "epoch": 2.47, + "learning_rate": 1.1771399128362569e-05, + "loss": 0.8928, + "step": 38446 + }, + { + "epoch": 2.47, + "learning_rate": 1.1771184439339618e-05, + "loss": 0.9364, + "step": 38447 + }, + { + "epoch": 2.47, + "learning_rate": 1.1770969750316667e-05, + "loss": 0.9132, + "step": 38448 + }, + { + "epoch": 2.47, + "learning_rate": 1.1770755061293718e-05, + "loss": 0.9036, + "step": 38449 + }, + { + "epoch": 2.47, + "learning_rate": 1.1770540372270766e-05, + "loss": 0.8916, + "step": 38450 + }, + { + "epoch": 2.47, + "learning_rate": 1.1770325683247817e-05, + "loss": 0.9244, + "step": 38451 + }, + { + "epoch": 2.47, + "learning_rate": 1.1770110994224866e-05, + "loss": 0.9419, + "step": 38452 + }, + { + "epoch": 2.47, + "learning_rate": 1.1769896305201915e-05, + "loss": 0.8909, + "step": 38453 + }, + { + "epoch": 2.47, + "learning_rate": 1.1769681616178966e-05, + "loss": 0.8804, + "step": 38454 + }, + { + "epoch": 2.47, + "learning_rate": 1.1769466927156015e-05, + "loss": 0.864, + "step": 38455 + }, + { + "epoch": 2.47, + "learning_rate": 1.1769252238133064e-05, + "loss": 0.8798, + "step": 38456 + }, + { + "epoch": 2.47, + "learning_rate": 1.1769037549110114e-05, + "loss": 0.8818, + "step": 38457 + }, + { + "epoch": 2.47, + "learning_rate": 1.1768822860087165e-05, + "loss": 0.9111, + "step": 38458 + }, + { + "epoch": 2.47, + "learning_rate": 1.1768608171064216e-05, + "loss": 0.9281, + "step": 38459 + }, + { + "epoch": 2.47, + "learning_rate": 1.1768393482041265e-05, + "loss": 0.9248, + "step": 38460 + }, + { + "epoch": 2.47, + "learning_rate": 1.1768178793018315e-05, + "loss": 0.8933, + "step": 38461 + }, + { + "epoch": 2.47, + "learning_rate": 1.1767964103995364e-05, + "loss": 0.8719, + "step": 38462 + }, + { + "epoch": 2.47, + "learning_rate": 1.1767749414972413e-05, + "loss": 0.8503, + "step": 38463 + }, + { + "epoch": 2.47, + "learning_rate": 1.1767534725949464e-05, + "loss": 0.8434, + "step": 38464 + }, + { + "epoch": 2.47, + "learning_rate": 1.1767320036926513e-05, + "loss": 0.8716, + "step": 38465 + }, + { + "epoch": 2.47, + "learning_rate": 1.1767105347903562e-05, + "loss": 0.9284, + "step": 38466 + }, + { + "epoch": 2.47, + "learning_rate": 1.1766890658880612e-05, + "loss": 0.7751, + "step": 38467 + }, + { + "epoch": 2.47, + "learning_rate": 1.1766675969857661e-05, + "loss": 0.9141, + "step": 38468 + }, + { + "epoch": 2.48, + "learning_rate": 1.176646128083471e-05, + "loss": 0.8636, + "step": 38469 + }, + { + "epoch": 2.48, + "learning_rate": 1.1766246591811761e-05, + "loss": 0.9204, + "step": 38470 + }, + { + "epoch": 2.48, + "learning_rate": 1.176603190278881e-05, + "loss": 0.8253, + "step": 38471 + }, + { + "epoch": 2.48, + "learning_rate": 1.1765817213765862e-05, + "loss": 0.9364, + "step": 38472 + }, + { + "epoch": 2.48, + "learning_rate": 1.1765602524742911e-05, + "loss": 0.8744, + "step": 38473 + }, + { + "epoch": 2.48, + "learning_rate": 1.1765387835719962e-05, + "loss": 0.9008, + "step": 38474 + }, + { + "epoch": 2.48, + "learning_rate": 1.1765173146697011e-05, + "loss": 0.9165, + "step": 38475 + }, + { + "epoch": 2.48, + "learning_rate": 1.176495845767406e-05, + "loss": 0.9216, + "step": 38476 + }, + { + "epoch": 2.48, + "learning_rate": 1.176474376865111e-05, + "loss": 0.8991, + "step": 38477 + }, + { + "epoch": 2.48, + "learning_rate": 1.176452907962816e-05, + "loss": 0.8659, + "step": 38478 + }, + { + "epoch": 2.48, + "learning_rate": 1.1764314390605208e-05, + "loss": 0.8792, + "step": 38479 + }, + { + "epoch": 2.48, + "learning_rate": 1.176409970158226e-05, + "loss": 0.9757, + "step": 38480 + }, + { + "epoch": 2.48, + "learning_rate": 1.1763885012559308e-05, + "loss": 0.8529, + "step": 38481 + }, + { + "epoch": 2.48, + "learning_rate": 1.1763670323536359e-05, + "loss": 0.8677, + "step": 38482 + }, + { + "epoch": 2.48, + "learning_rate": 1.1763455634513408e-05, + "loss": 0.9156, + "step": 38483 + }, + { + "epoch": 2.48, + "learning_rate": 1.1763240945490457e-05, + "loss": 0.8992, + "step": 38484 + }, + { + "epoch": 2.48, + "learning_rate": 1.1763026256467507e-05, + "loss": 0.9064, + "step": 38485 + }, + { + "epoch": 2.48, + "learning_rate": 1.1762811567444556e-05, + "loss": 0.8252, + "step": 38486 + }, + { + "epoch": 2.48, + "learning_rate": 1.1762596878421609e-05, + "loss": 0.9033, + "step": 38487 + }, + { + "epoch": 2.48, + "learning_rate": 1.1762382189398658e-05, + "loss": 0.8869, + "step": 38488 + }, + { + "epoch": 2.48, + "learning_rate": 1.1762167500375707e-05, + "loss": 0.9318, + "step": 38489 + }, + { + "epoch": 2.48, + "learning_rate": 1.1761952811352757e-05, + "loss": 0.8753, + "step": 38490 + }, + { + "epoch": 2.48, + "learning_rate": 1.1761738122329806e-05, + "loss": 0.8867, + "step": 38491 + }, + { + "epoch": 2.48, + "learning_rate": 1.1761523433306857e-05, + "loss": 0.8098, + "step": 38492 + }, + { + "epoch": 2.48, + "learning_rate": 1.1761308744283906e-05, + "loss": 0.8439, + "step": 38493 + }, + { + "epoch": 2.48, + "learning_rate": 1.1761094055260955e-05, + "loss": 0.8889, + "step": 38494 + }, + { + "epoch": 2.48, + "learning_rate": 1.1760879366238005e-05, + "loss": 0.8996, + "step": 38495 + }, + { + "epoch": 2.48, + "learning_rate": 1.1760664677215054e-05, + "loss": 0.8594, + "step": 38496 + }, + { + "epoch": 2.48, + "learning_rate": 1.1760449988192103e-05, + "loss": 0.8742, + "step": 38497 + }, + { + "epoch": 2.48, + "learning_rate": 1.1760235299169154e-05, + "loss": 0.9288, + "step": 38498 + }, + { + "epoch": 2.48, + "learning_rate": 1.1760020610146203e-05, + "loss": 0.9401, + "step": 38499 + }, + { + "epoch": 2.48, + "learning_rate": 1.1759805921123254e-05, + "loss": 0.891, + "step": 38500 + }, + { + "epoch": 2.48, + "learning_rate": 1.1759591232100304e-05, + "loss": 0.86, + "step": 38501 + }, + { + "epoch": 2.48, + "learning_rate": 1.1759376543077355e-05, + "loss": 0.8861, + "step": 38502 + }, + { + "epoch": 2.48, + "learning_rate": 1.1759161854054404e-05, + "loss": 0.8613, + "step": 38503 + }, + { + "epoch": 2.48, + "learning_rate": 1.1758947165031453e-05, + "loss": 0.9049, + "step": 38504 + }, + { + "epoch": 2.48, + "learning_rate": 1.1758732476008504e-05, + "loss": 0.9123, + "step": 38505 + }, + { + "epoch": 2.48, + "learning_rate": 1.1758517786985553e-05, + "loss": 0.8298, + "step": 38506 + }, + { + "epoch": 2.48, + "learning_rate": 1.1758303097962602e-05, + "loss": 0.7732, + "step": 38507 + }, + { + "epoch": 2.48, + "learning_rate": 1.1758088408939652e-05, + "loss": 0.8098, + "step": 38508 + }, + { + "epoch": 2.48, + "learning_rate": 1.1757873719916701e-05, + "loss": 0.9223, + "step": 38509 + }, + { + "epoch": 2.48, + "learning_rate": 1.1757659030893752e-05, + "loss": 0.8702, + "step": 38510 + }, + { + "epoch": 2.48, + "learning_rate": 1.17574443418708e-05, + "loss": 0.9148, + "step": 38511 + }, + { + "epoch": 2.48, + "learning_rate": 1.175722965284785e-05, + "loss": 0.9179, + "step": 38512 + }, + { + "epoch": 2.48, + "learning_rate": 1.17570149638249e-05, + "loss": 0.9098, + "step": 38513 + }, + { + "epoch": 2.48, + "learning_rate": 1.175680027480195e-05, + "loss": 0.9002, + "step": 38514 + }, + { + "epoch": 2.48, + "learning_rate": 1.1756585585778998e-05, + "loss": 0.9486, + "step": 38515 + }, + { + "epoch": 2.48, + "learning_rate": 1.175637089675605e-05, + "loss": 0.9442, + "step": 38516 + }, + { + "epoch": 2.48, + "learning_rate": 1.17561562077331e-05, + "loss": 0.9523, + "step": 38517 + }, + { + "epoch": 2.48, + "learning_rate": 1.175594151871015e-05, + "loss": 0.8578, + "step": 38518 + }, + { + "epoch": 2.48, + "learning_rate": 1.17557268296872e-05, + "loss": 0.8727, + "step": 38519 + }, + { + "epoch": 2.48, + "learning_rate": 1.175551214066425e-05, + "loss": 0.909, + "step": 38520 + }, + { + "epoch": 2.48, + "learning_rate": 1.1755297451641299e-05, + "loss": 0.904, + "step": 38521 + }, + { + "epoch": 2.48, + "learning_rate": 1.1755082762618348e-05, + "loss": 0.857, + "step": 38522 + }, + { + "epoch": 2.48, + "learning_rate": 1.1754868073595399e-05, + "loss": 0.8899, + "step": 38523 + }, + { + "epoch": 2.48, + "learning_rate": 1.1754653384572447e-05, + "loss": 0.9276, + "step": 38524 + }, + { + "epoch": 2.48, + "learning_rate": 1.1754438695549496e-05, + "loss": 1.0031, + "step": 38525 + }, + { + "epoch": 2.48, + "learning_rate": 1.1754224006526547e-05, + "loss": 0.7983, + "step": 38526 + }, + { + "epoch": 2.48, + "learning_rate": 1.1754009317503596e-05, + "loss": 0.8435, + "step": 38527 + }, + { + "epoch": 2.48, + "learning_rate": 1.1753794628480647e-05, + "loss": 0.8477, + "step": 38528 + }, + { + "epoch": 2.48, + "learning_rate": 1.1753579939457696e-05, + "loss": 0.9363, + "step": 38529 + }, + { + "epoch": 2.48, + "learning_rate": 1.1753365250434748e-05, + "loss": 0.8951, + "step": 38530 + }, + { + "epoch": 2.48, + "learning_rate": 1.1753150561411797e-05, + "loss": 0.9037, + "step": 38531 + }, + { + "epoch": 2.48, + "learning_rate": 1.1752935872388846e-05, + "loss": 0.9083, + "step": 38532 + }, + { + "epoch": 2.48, + "learning_rate": 1.1752721183365897e-05, + "loss": 0.8858, + "step": 38533 + }, + { + "epoch": 2.48, + "learning_rate": 1.1752506494342946e-05, + "loss": 0.9072, + "step": 38534 + }, + { + "epoch": 2.48, + "learning_rate": 1.1752291805319995e-05, + "loss": 0.8527, + "step": 38535 + }, + { + "epoch": 2.48, + "learning_rate": 1.1752077116297045e-05, + "loss": 0.8968, + "step": 38536 + }, + { + "epoch": 2.48, + "learning_rate": 1.1751862427274094e-05, + "loss": 0.9227, + "step": 38537 + }, + { + "epoch": 2.48, + "learning_rate": 1.1751647738251145e-05, + "loss": 0.9156, + "step": 38538 + }, + { + "epoch": 2.48, + "learning_rate": 1.1751433049228194e-05, + "loss": 0.8532, + "step": 38539 + }, + { + "epoch": 2.48, + "learning_rate": 1.1751218360205243e-05, + "loss": 0.8963, + "step": 38540 + }, + { + "epoch": 2.48, + "learning_rate": 1.1751003671182293e-05, + "loss": 0.8799, + "step": 38541 + }, + { + "epoch": 2.48, + "learning_rate": 1.1750788982159342e-05, + "loss": 0.9412, + "step": 38542 + }, + { + "epoch": 2.48, + "learning_rate": 1.1750574293136391e-05, + "loss": 0.8655, + "step": 38543 + }, + { + "epoch": 2.48, + "learning_rate": 1.1750574293136391e-05, + "loss": 0.9546, + "step": 38544 + }, + { + "epoch": 2.48, + "learning_rate": 1.1750359604113442e-05, + "loss": 0.846, + "step": 38545 + }, + { + "epoch": 2.48, + "learning_rate": 1.1750144915090493e-05, + "loss": 0.8925, + "step": 38546 + }, + { + "epoch": 2.48, + "learning_rate": 1.1749930226067543e-05, + "loss": 0.888, + "step": 38547 + }, + { + "epoch": 2.48, + "learning_rate": 1.1749715537044592e-05, + "loss": 0.8983, + "step": 38548 + }, + { + "epoch": 2.48, + "learning_rate": 1.1749500848021643e-05, + "loss": 0.9145, + "step": 38549 + }, + { + "epoch": 2.48, + "learning_rate": 1.1749286158998692e-05, + "loss": 0.9048, + "step": 38550 + }, + { + "epoch": 2.48, + "learning_rate": 1.1749071469975741e-05, + "loss": 0.9277, + "step": 38551 + }, + { + "epoch": 2.48, + "learning_rate": 1.1748856780952792e-05, + "loss": 0.8399, + "step": 38552 + }, + { + "epoch": 2.48, + "learning_rate": 1.174864209192984e-05, + "loss": 0.8925, + "step": 38553 + }, + { + "epoch": 2.48, + "learning_rate": 1.174842740290689e-05, + "loss": 0.8764, + "step": 38554 + }, + { + "epoch": 2.48, + "learning_rate": 1.174821271388394e-05, + "loss": 0.8648, + "step": 38555 + }, + { + "epoch": 2.48, + "learning_rate": 1.1747998024860989e-05, + "loss": 0.9298, + "step": 38556 + }, + { + "epoch": 2.48, + "learning_rate": 1.1747783335838038e-05, + "loss": 0.8713, + "step": 38557 + }, + { + "epoch": 2.48, + "learning_rate": 1.1747568646815089e-05, + "loss": 0.9159, + "step": 38558 + }, + { + "epoch": 2.48, + "learning_rate": 1.1747353957792138e-05, + "loss": 0.9085, + "step": 38559 + }, + { + "epoch": 2.48, + "learning_rate": 1.174713926876919e-05, + "loss": 0.9263, + "step": 38560 + }, + { + "epoch": 2.48, + "learning_rate": 1.1746924579746239e-05, + "loss": 0.9787, + "step": 38561 + }, + { + "epoch": 2.48, + "learning_rate": 1.174670989072329e-05, + "loss": 0.9, + "step": 38562 + }, + { + "epoch": 2.48, + "learning_rate": 1.1746495201700339e-05, + "loss": 0.8865, + "step": 38563 + }, + { + "epoch": 2.48, + "learning_rate": 1.1746280512677388e-05, + "loss": 0.8202, + "step": 38564 + }, + { + "epoch": 2.48, + "learning_rate": 1.1746065823654438e-05, + "loss": 0.935, + "step": 38565 + }, + { + "epoch": 2.48, + "learning_rate": 1.1745851134631487e-05, + "loss": 0.862, + "step": 38566 + }, + { + "epoch": 2.48, + "learning_rate": 1.1745636445608536e-05, + "loss": 0.9152, + "step": 38567 + }, + { + "epoch": 2.48, + "learning_rate": 1.1745421756585587e-05, + "loss": 0.918, + "step": 38568 + }, + { + "epoch": 2.48, + "learning_rate": 1.1745207067562636e-05, + "loss": 0.8632, + "step": 38569 + }, + { + "epoch": 2.48, + "learning_rate": 1.1744992378539686e-05, + "loss": 0.869, + "step": 38570 + }, + { + "epoch": 2.48, + "learning_rate": 1.1744777689516735e-05, + "loss": 0.9143, + "step": 38571 + }, + { + "epoch": 2.48, + "learning_rate": 1.1744563000493784e-05, + "loss": 0.9907, + "step": 38572 + }, + { + "epoch": 2.48, + "learning_rate": 1.1744348311470835e-05, + "loss": 0.8175, + "step": 38573 + }, + { + "epoch": 2.48, + "learning_rate": 1.1744133622447884e-05, + "loss": 0.9094, + "step": 38574 + }, + { + "epoch": 2.48, + "learning_rate": 1.1743918933424936e-05, + "loss": 0.8746, + "step": 38575 + }, + { + "epoch": 2.48, + "learning_rate": 1.1743704244401985e-05, + "loss": 0.8652, + "step": 38576 + }, + { + "epoch": 2.48, + "learning_rate": 1.1743489555379034e-05, + "loss": 0.9024, + "step": 38577 + }, + { + "epoch": 2.48, + "learning_rate": 1.1743274866356085e-05, + "loss": 0.8861, + "step": 38578 + }, + { + "epoch": 2.48, + "learning_rate": 1.1743060177333134e-05, + "loss": 0.8911, + "step": 38579 + }, + { + "epoch": 2.48, + "learning_rate": 1.1742845488310185e-05, + "loss": 0.9162, + "step": 38580 + }, + { + "epoch": 2.48, + "learning_rate": 1.1742630799287234e-05, + "loss": 0.8674, + "step": 38581 + }, + { + "epoch": 2.48, + "learning_rate": 1.1742416110264283e-05, + "loss": 0.9273, + "step": 38582 + }, + { + "epoch": 2.48, + "learning_rate": 1.1742201421241333e-05, + "loss": 0.908, + "step": 38583 + }, + { + "epoch": 2.48, + "learning_rate": 1.1741986732218382e-05, + "loss": 0.9655, + "step": 38584 + }, + { + "epoch": 2.48, + "learning_rate": 1.1741772043195431e-05, + "loss": 0.9956, + "step": 38585 + }, + { + "epoch": 2.48, + "learning_rate": 1.1741557354172482e-05, + "loss": 0.9582, + "step": 38586 + }, + { + "epoch": 2.48, + "learning_rate": 1.174134266514953e-05, + "loss": 0.8543, + "step": 38587 + }, + { + "epoch": 2.48, + "learning_rate": 1.1741127976126581e-05, + "loss": 0.8652, + "step": 38588 + }, + { + "epoch": 2.48, + "learning_rate": 1.1740913287103632e-05, + "loss": 0.8287, + "step": 38589 + }, + { + "epoch": 2.48, + "learning_rate": 1.1740698598080683e-05, + "loss": 0.9238, + "step": 38590 + }, + { + "epoch": 2.48, + "learning_rate": 1.1740483909057732e-05, + "loss": 0.8521, + "step": 38591 + }, + { + "epoch": 2.48, + "learning_rate": 1.174026922003478e-05, + "loss": 0.9472, + "step": 38592 + }, + { + "epoch": 2.48, + "learning_rate": 1.1740054531011831e-05, + "loss": 0.8598, + "step": 38593 + }, + { + "epoch": 2.48, + "learning_rate": 1.173983984198888e-05, + "loss": 0.8741, + "step": 38594 + }, + { + "epoch": 2.48, + "learning_rate": 1.173962515296593e-05, + "loss": 0.9088, + "step": 38595 + }, + { + "epoch": 2.48, + "learning_rate": 1.173941046394298e-05, + "loss": 0.8811, + "step": 38596 + }, + { + "epoch": 2.48, + "learning_rate": 1.1739195774920029e-05, + "loss": 0.8476, + "step": 38597 + }, + { + "epoch": 2.48, + "learning_rate": 1.173898108589708e-05, + "loss": 0.8772, + "step": 38598 + }, + { + "epoch": 2.48, + "learning_rate": 1.1738766396874128e-05, + "loss": 0.8855, + "step": 38599 + }, + { + "epoch": 2.48, + "learning_rate": 1.1738551707851177e-05, + "loss": 0.871, + "step": 38600 + }, + { + "epoch": 2.48, + "learning_rate": 1.1738337018828228e-05, + "loss": 0.8801, + "step": 38601 + }, + { + "epoch": 2.48, + "learning_rate": 1.1738122329805277e-05, + "loss": 0.9231, + "step": 38602 + }, + { + "epoch": 2.48, + "learning_rate": 1.1737907640782326e-05, + "loss": 0.8415, + "step": 38603 + }, + { + "epoch": 2.48, + "learning_rate": 1.1737692951759378e-05, + "loss": 0.8865, + "step": 38604 + }, + { + "epoch": 2.48, + "learning_rate": 1.1737478262736427e-05, + "loss": 0.8122, + "step": 38605 + }, + { + "epoch": 2.48, + "learning_rate": 1.1737263573713478e-05, + "loss": 0.9046, + "step": 38606 + }, + { + "epoch": 2.48, + "learning_rate": 1.1737048884690527e-05, + "loss": 0.9293, + "step": 38607 + }, + { + "epoch": 2.48, + "learning_rate": 1.1736834195667578e-05, + "loss": 0.8422, + "step": 38608 + }, + { + "epoch": 2.48, + "learning_rate": 1.1736619506644627e-05, + "loss": 0.9097, + "step": 38609 + }, + { + "epoch": 2.48, + "learning_rate": 1.1736404817621676e-05, + "loss": 0.9363, + "step": 38610 + }, + { + "epoch": 2.48, + "learning_rate": 1.1736190128598726e-05, + "loss": 0.8753, + "step": 38611 + }, + { + "epoch": 2.48, + "learning_rate": 1.1735975439575775e-05, + "loss": 0.852, + "step": 38612 + }, + { + "epoch": 2.48, + "learning_rate": 1.1735760750552824e-05, + "loss": 0.9062, + "step": 38613 + }, + { + "epoch": 2.48, + "learning_rate": 1.1735546061529875e-05, + "loss": 0.8938, + "step": 38614 + }, + { + "epoch": 2.48, + "learning_rate": 1.1735331372506924e-05, + "loss": 0.8574, + "step": 38615 + }, + { + "epoch": 2.48, + "learning_rate": 1.1735116683483973e-05, + "loss": 0.9089, + "step": 38616 + }, + { + "epoch": 2.48, + "learning_rate": 1.1734901994461023e-05, + "loss": 0.9166, + "step": 38617 + }, + { + "epoch": 2.48, + "learning_rate": 1.1734687305438076e-05, + "loss": 0.8758, + "step": 38618 + }, + { + "epoch": 2.48, + "learning_rate": 1.1734472616415125e-05, + "loss": 0.8501, + "step": 38619 + }, + { + "epoch": 2.48, + "learning_rate": 1.1734257927392174e-05, + "loss": 0.9996, + "step": 38620 + }, + { + "epoch": 2.48, + "learning_rate": 1.1734043238369224e-05, + "loss": 0.9089, + "step": 38621 + }, + { + "epoch": 2.48, + "learning_rate": 1.1733828549346273e-05, + "loss": 0.9786, + "step": 38622 + }, + { + "epoch": 2.48, + "learning_rate": 1.1733613860323322e-05, + "loss": 0.9266, + "step": 38623 + }, + { + "epoch": 2.48, + "learning_rate": 1.1733399171300373e-05, + "loss": 0.8553, + "step": 38624 + }, + { + "epoch": 2.49, + "learning_rate": 1.1733184482277422e-05, + "loss": 0.9476, + "step": 38625 + }, + { + "epoch": 2.49, + "learning_rate": 1.1732969793254471e-05, + "loss": 0.8897, + "step": 38626 + }, + { + "epoch": 2.49, + "learning_rate": 1.1732755104231522e-05, + "loss": 0.9385, + "step": 38627 + }, + { + "epoch": 2.49, + "learning_rate": 1.173254041520857e-05, + "loss": 0.8995, + "step": 38628 + }, + { + "epoch": 2.49, + "learning_rate": 1.1732325726185621e-05, + "loss": 0.8918, + "step": 38629 + }, + { + "epoch": 2.49, + "learning_rate": 1.173211103716267e-05, + "loss": 0.8668, + "step": 38630 + }, + { + "epoch": 2.49, + "learning_rate": 1.1731896348139719e-05, + "loss": 0.9393, + "step": 38631 + }, + { + "epoch": 2.49, + "learning_rate": 1.173168165911677e-05, + "loss": 0.9469, + "step": 38632 + }, + { + "epoch": 2.49, + "learning_rate": 1.173146697009382e-05, + "loss": 0.8369, + "step": 38633 + }, + { + "epoch": 2.49, + "learning_rate": 1.1731252281070871e-05, + "loss": 0.8941, + "step": 38634 + }, + { + "epoch": 2.49, + "learning_rate": 1.173103759204792e-05, + "loss": 0.8518, + "step": 38635 + }, + { + "epoch": 2.49, + "learning_rate": 1.1730822903024969e-05, + "loss": 0.8622, + "step": 38636 + }, + { + "epoch": 2.49, + "learning_rate": 1.173060821400202e-05, + "loss": 0.9249, + "step": 38637 + }, + { + "epoch": 2.49, + "learning_rate": 1.1730393524979069e-05, + "loss": 0.9325, + "step": 38638 + }, + { + "epoch": 2.49, + "learning_rate": 1.173017883595612e-05, + "loss": 0.8429, + "step": 38639 + }, + { + "epoch": 2.49, + "learning_rate": 1.1729964146933168e-05, + "loss": 0.9215, + "step": 38640 + }, + { + "epoch": 2.49, + "learning_rate": 1.1729749457910217e-05, + "loss": 0.8593, + "step": 38641 + }, + { + "epoch": 2.49, + "learning_rate": 1.1729534768887268e-05, + "loss": 0.9538, + "step": 38642 + }, + { + "epoch": 2.49, + "learning_rate": 1.1729320079864317e-05, + "loss": 0.933, + "step": 38643 + }, + { + "epoch": 2.49, + "learning_rate": 1.1729105390841366e-05, + "loss": 0.9436, + "step": 38644 + }, + { + "epoch": 2.49, + "learning_rate": 1.1728890701818416e-05, + "loss": 0.9366, + "step": 38645 + }, + { + "epoch": 2.49, + "learning_rate": 1.1728676012795465e-05, + "loss": 0.8176, + "step": 38646 + }, + { + "epoch": 2.49, + "learning_rate": 1.1728461323772518e-05, + "loss": 0.8744, + "step": 38647 + }, + { + "epoch": 2.49, + "learning_rate": 1.1728246634749567e-05, + "loss": 0.9075, + "step": 38648 + }, + { + "epoch": 2.49, + "learning_rate": 1.1728031945726617e-05, + "loss": 0.8892, + "step": 38649 + }, + { + "epoch": 2.49, + "learning_rate": 1.1727817256703666e-05, + "loss": 0.8973, + "step": 38650 + }, + { + "epoch": 2.49, + "learning_rate": 1.1727602567680715e-05, + "loss": 0.8798, + "step": 38651 + }, + { + "epoch": 2.49, + "learning_rate": 1.1727387878657766e-05, + "loss": 0.9229, + "step": 38652 + }, + { + "epoch": 2.49, + "learning_rate": 1.1727173189634815e-05, + "loss": 0.8183, + "step": 38653 + }, + { + "epoch": 2.49, + "learning_rate": 1.1726958500611864e-05, + "loss": 0.928, + "step": 38654 + }, + { + "epoch": 2.49, + "learning_rate": 1.1726743811588915e-05, + "loss": 1.0012, + "step": 38655 + }, + { + "epoch": 2.49, + "learning_rate": 1.1726529122565964e-05, + "loss": 0.9166, + "step": 38656 + }, + { + "epoch": 2.49, + "learning_rate": 1.1726314433543014e-05, + "loss": 0.9483, + "step": 38657 + }, + { + "epoch": 2.49, + "learning_rate": 1.1726099744520063e-05, + "loss": 0.8893, + "step": 38658 + }, + { + "epoch": 2.49, + "learning_rate": 1.1725885055497112e-05, + "loss": 0.8499, + "step": 38659 + }, + { + "epoch": 2.49, + "learning_rate": 1.1725670366474163e-05, + "loss": 0.8941, + "step": 38660 + }, + { + "epoch": 2.49, + "learning_rate": 1.1725455677451212e-05, + "loss": 0.9489, + "step": 38661 + }, + { + "epoch": 2.49, + "learning_rate": 1.1725240988428264e-05, + "loss": 0.8626, + "step": 38662 + }, + { + "epoch": 2.49, + "learning_rate": 1.1725026299405313e-05, + "loss": 0.8553, + "step": 38663 + }, + { + "epoch": 2.49, + "learning_rate": 1.1724811610382362e-05, + "loss": 0.8613, + "step": 38664 + }, + { + "epoch": 2.49, + "learning_rate": 1.1724596921359413e-05, + "loss": 0.9006, + "step": 38665 + }, + { + "epoch": 2.49, + "learning_rate": 1.1724382232336462e-05, + "loss": 0.8265, + "step": 38666 + }, + { + "epoch": 2.49, + "learning_rate": 1.1724167543313512e-05, + "loss": 0.8855, + "step": 38667 + }, + { + "epoch": 2.49, + "learning_rate": 1.1723952854290561e-05, + "loss": 0.9359, + "step": 38668 + }, + { + "epoch": 2.49, + "learning_rate": 1.172373816526761e-05, + "loss": 0.8834, + "step": 38669 + }, + { + "epoch": 2.49, + "learning_rate": 1.1723523476244661e-05, + "loss": 0.8638, + "step": 38670 + }, + { + "epoch": 2.49, + "learning_rate": 1.172330878722171e-05, + "loss": 0.8751, + "step": 38671 + }, + { + "epoch": 2.49, + "learning_rate": 1.1723094098198759e-05, + "loss": 0.9435, + "step": 38672 + }, + { + "epoch": 2.49, + "learning_rate": 1.172287940917581e-05, + "loss": 0.8455, + "step": 38673 + }, + { + "epoch": 2.49, + "learning_rate": 1.1722664720152858e-05, + "loss": 0.8563, + "step": 38674 + }, + { + "epoch": 2.49, + "learning_rate": 1.1722450031129909e-05, + "loss": 0.9229, + "step": 38675 + }, + { + "epoch": 2.49, + "learning_rate": 1.172223534210696e-05, + "loss": 0.8846, + "step": 38676 + }, + { + "epoch": 2.49, + "learning_rate": 1.172202065308401e-05, + "loss": 0.889, + "step": 38677 + }, + { + "epoch": 2.49, + "learning_rate": 1.172180596406106e-05, + "loss": 0.8987, + "step": 38678 + }, + { + "epoch": 2.49, + "learning_rate": 1.1721591275038108e-05, + "loss": 0.902, + "step": 38679 + }, + { + "epoch": 2.49, + "learning_rate": 1.1721376586015159e-05, + "loss": 0.8177, + "step": 38680 + }, + { + "epoch": 2.49, + "learning_rate": 1.1721161896992208e-05, + "loss": 0.8235, + "step": 38681 + }, + { + "epoch": 2.49, + "learning_rate": 1.1720947207969257e-05, + "loss": 0.872, + "step": 38682 + }, + { + "epoch": 2.49, + "learning_rate": 1.1720732518946308e-05, + "loss": 0.9084, + "step": 38683 + }, + { + "epoch": 2.49, + "learning_rate": 1.1720517829923357e-05, + "loss": 0.9492, + "step": 38684 + }, + { + "epoch": 2.49, + "learning_rate": 1.1720303140900407e-05, + "loss": 0.9151, + "step": 38685 + }, + { + "epoch": 2.49, + "learning_rate": 1.1720088451877456e-05, + "loss": 0.8425, + "step": 38686 + }, + { + "epoch": 2.49, + "learning_rate": 1.1719873762854505e-05, + "loss": 0.8032, + "step": 38687 + }, + { + "epoch": 2.49, + "learning_rate": 1.1719659073831556e-05, + "loss": 0.8737, + "step": 38688 + }, + { + "epoch": 2.49, + "learning_rate": 1.1719444384808605e-05, + "loss": 0.9033, + "step": 38689 + }, + { + "epoch": 2.49, + "learning_rate": 1.1719229695785654e-05, + "loss": 0.8949, + "step": 38690 + }, + { + "epoch": 2.49, + "learning_rate": 1.1719015006762706e-05, + "loss": 0.8767, + "step": 38691 + }, + { + "epoch": 2.49, + "learning_rate": 1.1718800317739755e-05, + "loss": 0.8971, + "step": 38692 + }, + { + "epoch": 2.49, + "learning_rate": 1.1718585628716806e-05, + "loss": 0.8916, + "step": 38693 + }, + { + "epoch": 2.49, + "learning_rate": 1.1718370939693855e-05, + "loss": 1.0028, + "step": 38694 + }, + { + "epoch": 2.49, + "learning_rate": 1.1718156250670905e-05, + "loss": 0.9047, + "step": 38695 + }, + { + "epoch": 2.49, + "learning_rate": 1.1717941561647954e-05, + "loss": 0.8946, + "step": 38696 + }, + { + "epoch": 2.49, + "learning_rate": 1.1717726872625003e-05, + "loss": 0.9193, + "step": 38697 + }, + { + "epoch": 2.49, + "learning_rate": 1.1717512183602054e-05, + "loss": 0.8787, + "step": 38698 + }, + { + "epoch": 2.49, + "learning_rate": 1.1717297494579103e-05, + "loss": 0.9243, + "step": 38699 + }, + { + "epoch": 2.49, + "learning_rate": 1.1717082805556152e-05, + "loss": 0.9235, + "step": 38700 + }, + { + "epoch": 2.49, + "learning_rate": 1.1716868116533203e-05, + "loss": 0.8532, + "step": 38701 + }, + { + "epoch": 2.49, + "learning_rate": 1.1716653427510252e-05, + "loss": 0.9274, + "step": 38702 + }, + { + "epoch": 2.49, + "learning_rate": 1.17164387384873e-05, + "loss": 0.9245, + "step": 38703 + }, + { + "epoch": 2.49, + "learning_rate": 1.1716224049464351e-05, + "loss": 0.8422, + "step": 38704 + }, + { + "epoch": 2.49, + "learning_rate": 1.1716009360441403e-05, + "loss": 0.8941, + "step": 38705 + }, + { + "epoch": 2.49, + "learning_rate": 1.1715794671418452e-05, + "loss": 0.9158, + "step": 38706 + }, + { + "epoch": 2.49, + "learning_rate": 1.1715579982395501e-05, + "loss": 0.8875, + "step": 38707 + }, + { + "epoch": 2.49, + "learning_rate": 1.1715365293372552e-05, + "loss": 0.8957, + "step": 38708 + }, + { + "epoch": 2.49, + "learning_rate": 1.1715150604349601e-05, + "loss": 0.9047, + "step": 38709 + }, + { + "epoch": 2.49, + "learning_rate": 1.171493591532665e-05, + "loss": 0.8829, + "step": 38710 + }, + { + "epoch": 2.49, + "learning_rate": 1.17147212263037e-05, + "loss": 0.8933, + "step": 38711 + }, + { + "epoch": 2.49, + "learning_rate": 1.171450653728075e-05, + "loss": 0.9726, + "step": 38712 + }, + { + "epoch": 2.49, + "learning_rate": 1.1714291848257799e-05, + "loss": 0.9618, + "step": 38713 + }, + { + "epoch": 2.49, + "learning_rate": 1.171407715923485e-05, + "loss": 0.9819, + "step": 38714 + }, + { + "epoch": 2.49, + "learning_rate": 1.1713862470211898e-05, + "loss": 0.9204, + "step": 38715 + }, + { + "epoch": 2.49, + "learning_rate": 1.1713647781188949e-05, + "loss": 0.9209, + "step": 38716 + }, + { + "epoch": 2.49, + "learning_rate": 1.1713433092165998e-05, + "loss": 0.8529, + "step": 38717 + }, + { + "epoch": 2.49, + "learning_rate": 1.1713218403143047e-05, + "loss": 0.9438, + "step": 38718 + }, + { + "epoch": 2.49, + "learning_rate": 1.1713003714120097e-05, + "loss": 0.9069, + "step": 38719 + }, + { + "epoch": 2.49, + "learning_rate": 1.1712789025097148e-05, + "loss": 0.9043, + "step": 38720 + }, + { + "epoch": 2.49, + "learning_rate": 1.1712574336074199e-05, + "loss": 0.8804, + "step": 38721 + }, + { + "epoch": 2.49, + "learning_rate": 1.1712359647051248e-05, + "loss": 0.8611, + "step": 38722 + }, + { + "epoch": 2.49, + "learning_rate": 1.1712144958028297e-05, + "loss": 0.9415, + "step": 38723 + }, + { + "epoch": 2.49, + "learning_rate": 1.1711930269005347e-05, + "loss": 0.8639, + "step": 38724 + }, + { + "epoch": 2.49, + "learning_rate": 1.1711715579982396e-05, + "loss": 0.8878, + "step": 38725 + }, + { + "epoch": 2.49, + "learning_rate": 1.1711500890959447e-05, + "loss": 0.9187, + "step": 38726 + }, + { + "epoch": 2.49, + "learning_rate": 1.1711286201936496e-05, + "loss": 0.9028, + "step": 38727 + }, + { + "epoch": 2.49, + "learning_rate": 1.1711071512913545e-05, + "loss": 0.9257, + "step": 38728 + }, + { + "epoch": 2.49, + "learning_rate": 1.1710856823890596e-05, + "loss": 0.9258, + "step": 38729 + }, + { + "epoch": 2.49, + "learning_rate": 1.1710642134867645e-05, + "loss": 0.904, + "step": 38730 + }, + { + "epoch": 2.49, + "learning_rate": 1.1710427445844694e-05, + "loss": 0.9159, + "step": 38731 + }, + { + "epoch": 2.49, + "learning_rate": 1.1710212756821744e-05, + "loss": 0.9129, + "step": 38732 + }, + { + "epoch": 2.49, + "learning_rate": 1.1709998067798793e-05, + "loss": 0.958, + "step": 38733 + }, + { + "epoch": 2.49, + "learning_rate": 1.1709783378775845e-05, + "loss": 0.943, + "step": 38734 + }, + { + "epoch": 2.49, + "learning_rate": 1.1709568689752894e-05, + "loss": 0.9114, + "step": 38735 + }, + { + "epoch": 2.49, + "learning_rate": 1.1709354000729945e-05, + "loss": 0.8518, + "step": 38736 + }, + { + "epoch": 2.49, + "learning_rate": 1.1709139311706994e-05, + "loss": 0.8458, + "step": 38737 + }, + { + "epoch": 2.49, + "learning_rate": 1.1708924622684043e-05, + "loss": 0.8395, + "step": 38738 + }, + { + "epoch": 2.49, + "learning_rate": 1.1708709933661094e-05, + "loss": 0.9041, + "step": 38739 + }, + { + "epoch": 2.49, + "learning_rate": 1.1708495244638143e-05, + "loss": 0.9208, + "step": 38740 + }, + { + "epoch": 2.49, + "learning_rate": 1.1708280555615192e-05, + "loss": 0.8653, + "step": 38741 + }, + { + "epoch": 2.49, + "learning_rate": 1.1708065866592242e-05, + "loss": 0.8606, + "step": 38742 + }, + { + "epoch": 2.49, + "learning_rate": 1.1707851177569291e-05, + "loss": 0.8625, + "step": 38743 + }, + { + "epoch": 2.49, + "learning_rate": 1.1707636488546342e-05, + "loss": 0.9042, + "step": 38744 + }, + { + "epoch": 2.49, + "learning_rate": 1.1707421799523391e-05, + "loss": 0.9477, + "step": 38745 + }, + { + "epoch": 2.49, + "learning_rate": 1.170720711050044e-05, + "loss": 0.9191, + "step": 38746 + }, + { + "epoch": 2.49, + "learning_rate": 1.170699242147749e-05, + "loss": 0.8906, + "step": 38747 + }, + { + "epoch": 2.49, + "learning_rate": 1.170677773245454e-05, + "loss": 0.9172, + "step": 38748 + }, + { + "epoch": 2.49, + "learning_rate": 1.1706563043431592e-05, + "loss": 0.8759, + "step": 38749 + }, + { + "epoch": 2.49, + "learning_rate": 1.170634835440864e-05, + "loss": 0.9283, + "step": 38750 + }, + { + "epoch": 2.49, + "learning_rate": 1.170613366538569e-05, + "loss": 0.9092, + "step": 38751 + }, + { + "epoch": 2.49, + "learning_rate": 1.170591897636274e-05, + "loss": 0.851, + "step": 38752 + }, + { + "epoch": 2.49, + "learning_rate": 1.170570428733979e-05, + "loss": 0.9642, + "step": 38753 + }, + { + "epoch": 2.49, + "learning_rate": 1.170548959831684e-05, + "loss": 0.9548, + "step": 38754 + }, + { + "epoch": 2.49, + "learning_rate": 1.1705274909293889e-05, + "loss": 0.9067, + "step": 38755 + }, + { + "epoch": 2.49, + "learning_rate": 1.1705060220270938e-05, + "loss": 0.9334, + "step": 38756 + }, + { + "epoch": 2.49, + "learning_rate": 1.1704845531247989e-05, + "loss": 0.9119, + "step": 38757 + }, + { + "epoch": 2.49, + "learning_rate": 1.1704630842225038e-05, + "loss": 0.8572, + "step": 38758 + }, + { + "epoch": 2.49, + "learning_rate": 1.1704416153202087e-05, + "loss": 0.8813, + "step": 38759 + }, + { + "epoch": 2.49, + "learning_rate": 1.1704201464179137e-05, + "loss": 0.8466, + "step": 38760 + }, + { + "epoch": 2.49, + "learning_rate": 1.1703986775156186e-05, + "loss": 0.8689, + "step": 38761 + }, + { + "epoch": 2.49, + "learning_rate": 1.1703772086133237e-05, + "loss": 0.86, + "step": 38762 + }, + { + "epoch": 2.49, + "learning_rate": 1.1703557397110286e-05, + "loss": 0.9034, + "step": 38763 + }, + { + "epoch": 2.49, + "learning_rate": 1.1703342708087338e-05, + "loss": 0.8982, + "step": 38764 + }, + { + "epoch": 2.49, + "learning_rate": 1.1703128019064387e-05, + "loss": 0.8863, + "step": 38765 + }, + { + "epoch": 2.49, + "learning_rate": 1.1702913330041436e-05, + "loss": 0.8657, + "step": 38766 + }, + { + "epoch": 2.49, + "learning_rate": 1.1702698641018487e-05, + "loss": 0.8057, + "step": 38767 + }, + { + "epoch": 2.49, + "learning_rate": 1.1702483951995536e-05, + "loss": 0.9161, + "step": 38768 + }, + { + "epoch": 2.49, + "learning_rate": 1.1702269262972585e-05, + "loss": 0.8711, + "step": 38769 + }, + { + "epoch": 2.49, + "learning_rate": 1.1702054573949635e-05, + "loss": 0.887, + "step": 38770 + }, + { + "epoch": 2.49, + "learning_rate": 1.1701839884926684e-05, + "loss": 0.9429, + "step": 38771 + }, + { + "epoch": 2.49, + "learning_rate": 1.1701625195903735e-05, + "loss": 0.9203, + "step": 38772 + }, + { + "epoch": 2.49, + "learning_rate": 1.1701410506880784e-05, + "loss": 0.8524, + "step": 38773 + }, + { + "epoch": 2.49, + "learning_rate": 1.1701195817857833e-05, + "loss": 0.8928, + "step": 38774 + }, + { + "epoch": 2.49, + "learning_rate": 1.1700981128834884e-05, + "loss": 0.9156, + "step": 38775 + }, + { + "epoch": 2.49, + "learning_rate": 1.1700766439811933e-05, + "loss": 0.967, + "step": 38776 + }, + { + "epoch": 2.49, + "learning_rate": 1.1700551750788981e-05, + "loss": 0.975, + "step": 38777 + }, + { + "epoch": 2.49, + "learning_rate": 1.1700337061766034e-05, + "loss": 0.9588, + "step": 38778 + }, + { + "epoch": 2.49, + "learning_rate": 1.1700122372743083e-05, + "loss": 0.9129, + "step": 38779 + }, + { + "epoch": 2.5, + "learning_rate": 1.1699907683720133e-05, + "loss": 0.897, + "step": 38780 + }, + { + "epoch": 2.5, + "learning_rate": 1.1699692994697182e-05, + "loss": 0.8704, + "step": 38781 + }, + { + "epoch": 2.5, + "learning_rate": 1.1699478305674233e-05, + "loss": 0.8897, + "step": 38782 + }, + { + "epoch": 2.5, + "learning_rate": 1.1699263616651282e-05, + "loss": 0.8938, + "step": 38783 + }, + { + "epoch": 2.5, + "learning_rate": 1.1699048927628331e-05, + "loss": 0.9562, + "step": 38784 + }, + { + "epoch": 2.5, + "learning_rate": 1.1698834238605382e-05, + "loss": 0.9374, + "step": 38785 + }, + { + "epoch": 2.5, + "learning_rate": 1.169861954958243e-05, + "loss": 0.8829, + "step": 38786 + }, + { + "epoch": 2.5, + "learning_rate": 1.169840486055948e-05, + "loss": 0.8503, + "step": 38787 + }, + { + "epoch": 2.5, + "learning_rate": 1.169819017153653e-05, + "loss": 1.0345, + "step": 38788 + }, + { + "epoch": 2.5, + "learning_rate": 1.169797548251358e-05, + "loss": 0.8157, + "step": 38789 + }, + { + "epoch": 2.5, + "learning_rate": 1.1697760793490628e-05, + "loss": 0.8848, + "step": 38790 + }, + { + "epoch": 2.5, + "learning_rate": 1.1697546104467679e-05, + "loss": 0.8912, + "step": 38791 + }, + { + "epoch": 2.5, + "learning_rate": 1.1697331415444728e-05, + "loss": 0.9209, + "step": 38792 + }, + { + "epoch": 2.5, + "learning_rate": 1.169711672642178e-05, + "loss": 0.9031, + "step": 38793 + }, + { + "epoch": 2.5, + "learning_rate": 1.1696902037398829e-05, + "loss": 0.8627, + "step": 38794 + }, + { + "epoch": 2.5, + "learning_rate": 1.169668734837588e-05, + "loss": 0.8446, + "step": 38795 + }, + { + "epoch": 2.5, + "learning_rate": 1.1696472659352929e-05, + "loss": 0.8889, + "step": 38796 + }, + { + "epoch": 2.5, + "learning_rate": 1.1696257970329978e-05, + "loss": 0.8545, + "step": 38797 + }, + { + "epoch": 2.5, + "learning_rate": 1.1696043281307028e-05, + "loss": 0.9191, + "step": 38798 + }, + { + "epoch": 2.5, + "learning_rate": 1.1695828592284077e-05, + "loss": 0.8835, + "step": 38799 + }, + { + "epoch": 2.5, + "learning_rate": 1.1695613903261126e-05, + "loss": 0.8648, + "step": 38800 + }, + { + "epoch": 2.5, + "learning_rate": 1.1695399214238177e-05, + "loss": 0.8448, + "step": 38801 + }, + { + "epoch": 2.5, + "learning_rate": 1.1695184525215226e-05, + "loss": 0.9421, + "step": 38802 + }, + { + "epoch": 2.5, + "learning_rate": 1.1694969836192277e-05, + "loss": 0.9206, + "step": 38803 + }, + { + "epoch": 2.5, + "learning_rate": 1.1694755147169326e-05, + "loss": 0.8724, + "step": 38804 + }, + { + "epoch": 2.5, + "learning_rate": 1.1694540458146375e-05, + "loss": 0.9233, + "step": 38805 + }, + { + "epoch": 2.5, + "learning_rate": 1.1694325769123425e-05, + "loss": 0.8019, + "step": 38806 + }, + { + "epoch": 2.5, + "learning_rate": 1.1694111080100476e-05, + "loss": 0.8631, + "step": 38807 + }, + { + "epoch": 2.5, + "learning_rate": 1.1693896391077526e-05, + "loss": 0.9931, + "step": 38808 + }, + { + "epoch": 2.5, + "learning_rate": 1.1693681702054575e-05, + "loss": 0.9358, + "step": 38809 + }, + { + "epoch": 2.5, + "learning_rate": 1.1693467013031624e-05, + "loss": 0.848, + "step": 38810 + }, + { + "epoch": 2.5, + "learning_rate": 1.1693252324008675e-05, + "loss": 0.8013, + "step": 38811 + }, + { + "epoch": 2.5, + "learning_rate": 1.1693037634985724e-05, + "loss": 0.9077, + "step": 38812 + }, + { + "epoch": 2.5, + "learning_rate": 1.1692822945962775e-05, + "loss": 0.8498, + "step": 38813 + }, + { + "epoch": 2.5, + "learning_rate": 1.1692608256939824e-05, + "loss": 0.8819, + "step": 38814 + }, + { + "epoch": 2.5, + "learning_rate": 1.1692393567916873e-05, + "loss": 0.9431, + "step": 38815 + }, + { + "epoch": 2.5, + "learning_rate": 1.1692178878893923e-05, + "loss": 0.8768, + "step": 38816 + }, + { + "epoch": 2.5, + "learning_rate": 1.1691964189870972e-05, + "loss": 0.8383, + "step": 38817 + }, + { + "epoch": 2.5, + "learning_rate": 1.1691749500848021e-05, + "loss": 0.9221, + "step": 38818 + }, + { + "epoch": 2.5, + "learning_rate": 1.1691534811825072e-05, + "loss": 0.8572, + "step": 38819 + }, + { + "epoch": 2.5, + "learning_rate": 1.1691320122802121e-05, + "loss": 0.9113, + "step": 38820 + }, + { + "epoch": 2.5, + "learning_rate": 1.1691105433779172e-05, + "loss": 0.8844, + "step": 38821 + }, + { + "epoch": 2.5, + "learning_rate": 1.1690890744756222e-05, + "loss": 0.8487, + "step": 38822 + }, + { + "epoch": 2.5, + "learning_rate": 1.1690676055733273e-05, + "loss": 1.015, + "step": 38823 + }, + { + "epoch": 2.5, + "learning_rate": 1.1690461366710322e-05, + "loss": 0.9047, + "step": 38824 + }, + { + "epoch": 2.5, + "learning_rate": 1.169024667768737e-05, + "loss": 0.9471, + "step": 38825 + }, + { + "epoch": 2.5, + "learning_rate": 1.1690031988664421e-05, + "loss": 0.9369, + "step": 38826 + }, + { + "epoch": 2.5, + "learning_rate": 1.168981729964147e-05, + "loss": 0.8794, + "step": 38827 + }, + { + "epoch": 2.5, + "learning_rate": 1.168960261061852e-05, + "loss": 0.906, + "step": 38828 + }, + { + "epoch": 2.5, + "learning_rate": 1.168938792159557e-05, + "loss": 0.816, + "step": 38829 + }, + { + "epoch": 2.5, + "learning_rate": 1.1689173232572619e-05, + "loss": 0.9385, + "step": 38830 + }, + { + "epoch": 2.5, + "learning_rate": 1.168895854354967e-05, + "loss": 0.8693, + "step": 38831 + }, + { + "epoch": 2.5, + "learning_rate": 1.1688743854526719e-05, + "loss": 0.9174, + "step": 38832 + }, + { + "epoch": 2.5, + "learning_rate": 1.1688529165503768e-05, + "loss": 0.9263, + "step": 38833 + }, + { + "epoch": 2.5, + "learning_rate": 1.1688314476480818e-05, + "loss": 0.8849, + "step": 38834 + }, + { + "epoch": 2.5, + "learning_rate": 1.1688099787457867e-05, + "loss": 0.9261, + "step": 38835 + }, + { + "epoch": 2.5, + "learning_rate": 1.168788509843492e-05, + "loss": 0.8553, + "step": 38836 + }, + { + "epoch": 2.5, + "learning_rate": 1.1687670409411968e-05, + "loss": 0.854, + "step": 38837 + }, + { + "epoch": 2.5, + "learning_rate": 1.1687455720389017e-05, + "loss": 0.8456, + "step": 38838 + }, + { + "epoch": 2.5, + "learning_rate": 1.1687241031366068e-05, + "loss": 0.885, + "step": 38839 + }, + { + "epoch": 2.5, + "learning_rate": 1.1687026342343117e-05, + "loss": 0.8424, + "step": 38840 + }, + { + "epoch": 2.5, + "learning_rate": 1.1686811653320168e-05, + "loss": 0.9069, + "step": 38841 + }, + { + "epoch": 2.5, + "learning_rate": 1.1686596964297217e-05, + "loss": 0.8967, + "step": 38842 + }, + { + "epoch": 2.5, + "learning_rate": 1.1686382275274266e-05, + "loss": 0.8819, + "step": 38843 + }, + { + "epoch": 2.5, + "learning_rate": 1.1686167586251316e-05, + "loss": 0.8585, + "step": 38844 + }, + { + "epoch": 2.5, + "learning_rate": 1.1685952897228365e-05, + "loss": 0.9261, + "step": 38845 + }, + { + "epoch": 2.5, + "learning_rate": 1.1685738208205414e-05, + "loss": 0.9164, + "step": 38846 + }, + { + "epoch": 2.5, + "learning_rate": 1.1685523519182465e-05, + "loss": 0.9462, + "step": 38847 + }, + { + "epoch": 2.5, + "learning_rate": 1.1685308830159514e-05, + "loss": 0.9123, + "step": 38848 + }, + { + "epoch": 2.5, + "learning_rate": 1.1685094141136565e-05, + "loss": 0.8558, + "step": 38849 + }, + { + "epoch": 2.5, + "learning_rate": 1.1684879452113614e-05, + "loss": 0.8924, + "step": 38850 + }, + { + "epoch": 2.5, + "learning_rate": 1.1684664763090666e-05, + "loss": 0.9153, + "step": 38851 + }, + { + "epoch": 2.5, + "learning_rate": 1.1684450074067715e-05, + "loss": 0.9555, + "step": 38852 + }, + { + "epoch": 2.5, + "learning_rate": 1.1684235385044764e-05, + "loss": 0.9019, + "step": 38853 + }, + { + "epoch": 2.5, + "learning_rate": 1.1684020696021814e-05, + "loss": 0.8213, + "step": 38854 + }, + { + "epoch": 2.5, + "learning_rate": 1.1683806006998863e-05, + "loss": 0.9779, + "step": 38855 + }, + { + "epoch": 2.5, + "learning_rate": 1.1683591317975912e-05, + "loss": 0.8916, + "step": 38856 + }, + { + "epoch": 2.5, + "learning_rate": 1.1683376628952963e-05, + "loss": 0.9059, + "step": 38857 + }, + { + "epoch": 2.5, + "learning_rate": 1.1683161939930012e-05, + "loss": 0.924, + "step": 38858 + }, + { + "epoch": 2.5, + "learning_rate": 1.1682947250907063e-05, + "loss": 0.8983, + "step": 38859 + }, + { + "epoch": 2.5, + "learning_rate": 1.1682732561884112e-05, + "loss": 0.9278, + "step": 38860 + }, + { + "epoch": 2.5, + "learning_rate": 1.168251787286116e-05, + "loss": 0.9455, + "step": 38861 + }, + { + "epoch": 2.5, + "learning_rate": 1.1682303183838211e-05, + "loss": 0.9223, + "step": 38862 + }, + { + "epoch": 2.5, + "learning_rate": 1.168208849481526e-05, + "loss": 0.9392, + "step": 38863 + }, + { + "epoch": 2.5, + "learning_rate": 1.168187380579231e-05, + "loss": 0.8853, + "step": 38864 + }, + { + "epoch": 2.5, + "learning_rate": 1.1681659116769362e-05, + "loss": 0.8532, + "step": 38865 + }, + { + "epoch": 2.5, + "learning_rate": 1.168144442774641e-05, + "loss": 0.8947, + "step": 38866 + }, + { + "epoch": 2.5, + "learning_rate": 1.1681229738723461e-05, + "loss": 0.8739, + "step": 38867 + }, + { + "epoch": 2.5, + "learning_rate": 1.168101504970051e-05, + "loss": 0.9332, + "step": 38868 + }, + { + "epoch": 2.5, + "learning_rate": 1.168080036067756e-05, + "loss": 0.9377, + "step": 38869 + }, + { + "epoch": 2.5, + "learning_rate": 1.168058567165461e-05, + "loss": 0.846, + "step": 38870 + }, + { + "epoch": 2.5, + "learning_rate": 1.1680370982631659e-05, + "loss": 0.8812, + "step": 38871 + }, + { + "epoch": 2.5, + "learning_rate": 1.168015629360871e-05, + "loss": 0.8706, + "step": 38872 + }, + { + "epoch": 2.5, + "learning_rate": 1.1679941604585758e-05, + "loss": 0.9182, + "step": 38873 + }, + { + "epoch": 2.5, + "learning_rate": 1.1679726915562807e-05, + "loss": 0.9091, + "step": 38874 + }, + { + "epoch": 2.5, + "learning_rate": 1.1679512226539858e-05, + "loss": 0.8614, + "step": 38875 + }, + { + "epoch": 2.5, + "learning_rate": 1.1679297537516907e-05, + "loss": 0.9373, + "step": 38876 + }, + { + "epoch": 2.5, + "learning_rate": 1.1679082848493956e-05, + "loss": 0.9596, + "step": 38877 + }, + { + "epoch": 2.5, + "learning_rate": 1.1678868159471007e-05, + "loss": 0.926, + "step": 38878 + }, + { + "epoch": 2.5, + "learning_rate": 1.1678653470448056e-05, + "loss": 0.9549, + "step": 38879 + }, + { + "epoch": 2.5, + "learning_rate": 1.1678438781425108e-05, + "loss": 0.8488, + "step": 38880 + }, + { + "epoch": 2.5, + "learning_rate": 1.1678224092402157e-05, + "loss": 0.9413, + "step": 38881 + }, + { + "epoch": 2.5, + "learning_rate": 1.1678009403379207e-05, + "loss": 0.9238, + "step": 38882 + }, + { + "epoch": 2.5, + "learning_rate": 1.1677794714356256e-05, + "loss": 0.8827, + "step": 38883 + }, + { + "epoch": 2.5, + "learning_rate": 1.1677580025333305e-05, + "loss": 0.8804, + "step": 38884 + }, + { + "epoch": 2.5, + "learning_rate": 1.1677365336310356e-05, + "loss": 0.8897, + "step": 38885 + }, + { + "epoch": 2.5, + "learning_rate": 1.1677150647287405e-05, + "loss": 0.9584, + "step": 38886 + }, + { + "epoch": 2.5, + "learning_rate": 1.1676935958264454e-05, + "loss": 0.9089, + "step": 38887 + }, + { + "epoch": 2.5, + "learning_rate": 1.1676721269241505e-05, + "loss": 0.9216, + "step": 38888 + }, + { + "epoch": 2.5, + "learning_rate": 1.1676506580218554e-05, + "loss": 0.904, + "step": 38889 + }, + { + "epoch": 2.5, + "learning_rate": 1.1676291891195604e-05, + "loss": 0.9093, + "step": 38890 + }, + { + "epoch": 2.5, + "learning_rate": 1.1676077202172653e-05, + "loss": 0.9428, + "step": 38891 + }, + { + "epoch": 2.5, + "learning_rate": 1.1675862513149702e-05, + "loss": 0.9222, + "step": 38892 + }, + { + "epoch": 2.5, + "learning_rate": 1.1675647824126753e-05, + "loss": 0.9022, + "step": 38893 + }, + { + "epoch": 2.5, + "learning_rate": 1.1675433135103804e-05, + "loss": 0.8918, + "step": 38894 + }, + { + "epoch": 2.5, + "learning_rate": 1.1675218446080854e-05, + "loss": 0.7514, + "step": 38895 + }, + { + "epoch": 2.5, + "learning_rate": 1.1675003757057903e-05, + "loss": 0.934, + "step": 38896 + }, + { + "epoch": 2.5, + "learning_rate": 1.1674789068034952e-05, + "loss": 0.9178, + "step": 38897 + }, + { + "epoch": 2.5, + "learning_rate": 1.1674574379012003e-05, + "loss": 0.8565, + "step": 38898 + }, + { + "epoch": 2.5, + "learning_rate": 1.1674359689989052e-05, + "loss": 0.8951, + "step": 38899 + }, + { + "epoch": 2.5, + "learning_rate": 1.1674145000966102e-05, + "loss": 0.8945, + "step": 38900 + }, + { + "epoch": 2.5, + "learning_rate": 1.1673930311943151e-05, + "loss": 0.8828, + "step": 38901 + }, + { + "epoch": 2.5, + "learning_rate": 1.16737156229202e-05, + "loss": 0.8821, + "step": 38902 + }, + { + "epoch": 2.5, + "learning_rate": 1.1673500933897251e-05, + "loss": 0.8931, + "step": 38903 + }, + { + "epoch": 2.5, + "learning_rate": 1.16732862448743e-05, + "loss": 0.8761, + "step": 38904 + }, + { + "epoch": 2.5, + "learning_rate": 1.1673071555851349e-05, + "loss": 0.9193, + "step": 38905 + }, + { + "epoch": 2.5, + "learning_rate": 1.16728568668284e-05, + "loss": 0.8953, + "step": 38906 + }, + { + "epoch": 2.5, + "learning_rate": 1.1672642177805449e-05, + "loss": 0.9648, + "step": 38907 + }, + { + "epoch": 2.5, + "learning_rate": 1.16724274887825e-05, + "loss": 0.9048, + "step": 38908 + }, + { + "epoch": 2.5, + "learning_rate": 1.167221279975955e-05, + "loss": 0.9051, + "step": 38909 + }, + { + "epoch": 2.5, + "learning_rate": 1.16719981107366e-05, + "loss": 0.8808, + "step": 38910 + }, + { + "epoch": 2.5, + "learning_rate": 1.167178342171365e-05, + "loss": 0.8642, + "step": 38911 + }, + { + "epoch": 2.5, + "learning_rate": 1.1671568732690698e-05, + "loss": 0.8757, + "step": 38912 + }, + { + "epoch": 2.5, + "learning_rate": 1.1671354043667749e-05, + "loss": 0.948, + "step": 38913 + }, + { + "epoch": 2.5, + "learning_rate": 1.1671139354644798e-05, + "loss": 0.8848, + "step": 38914 + }, + { + "epoch": 2.5, + "learning_rate": 1.1670924665621847e-05, + "loss": 0.9071, + "step": 38915 + }, + { + "epoch": 2.5, + "learning_rate": 1.1670709976598898e-05, + "loss": 0.8611, + "step": 38916 + }, + { + "epoch": 2.5, + "learning_rate": 1.1670495287575947e-05, + "loss": 0.9069, + "step": 38917 + }, + { + "epoch": 2.5, + "learning_rate": 1.1670280598552997e-05, + "loss": 0.8573, + "step": 38918 + }, + { + "epoch": 2.5, + "learning_rate": 1.1670065909530046e-05, + "loss": 0.8879, + "step": 38919 + }, + { + "epoch": 2.5, + "learning_rate": 1.1669851220507095e-05, + "loss": 0.8786, + "step": 38920 + }, + { + "epoch": 2.5, + "learning_rate": 1.1669636531484146e-05, + "loss": 0.8548, + "step": 38921 + }, + { + "epoch": 2.5, + "learning_rate": 1.1669421842461195e-05, + "loss": 0.8071, + "step": 38922 + }, + { + "epoch": 2.5, + "learning_rate": 1.1669207153438247e-05, + "loss": 0.8531, + "step": 38923 + }, + { + "epoch": 2.5, + "learning_rate": 1.1668992464415296e-05, + "loss": 0.8474, + "step": 38924 + }, + { + "epoch": 2.5, + "learning_rate": 1.1668777775392345e-05, + "loss": 0.8768, + "step": 38925 + }, + { + "epoch": 2.5, + "learning_rate": 1.1668563086369396e-05, + "loss": 0.8926, + "step": 38926 + }, + { + "epoch": 2.5, + "learning_rate": 1.1668348397346445e-05, + "loss": 0.9368, + "step": 38927 + }, + { + "epoch": 2.5, + "learning_rate": 1.1668133708323495e-05, + "loss": 0.882, + "step": 38928 + }, + { + "epoch": 2.5, + "learning_rate": 1.1667919019300544e-05, + "loss": 0.9197, + "step": 38929 + }, + { + "epoch": 2.5, + "learning_rate": 1.1667704330277593e-05, + "loss": 0.9615, + "step": 38930 + }, + { + "epoch": 2.5, + "learning_rate": 1.1667489641254644e-05, + "loss": 0.9183, + "step": 38931 + }, + { + "epoch": 2.5, + "learning_rate": 1.1667274952231693e-05, + "loss": 0.8727, + "step": 38932 + }, + { + "epoch": 2.5, + "learning_rate": 1.1667060263208742e-05, + "loss": 0.8372, + "step": 38933 + }, + { + "epoch": 2.5, + "learning_rate": 1.1666845574185793e-05, + "loss": 0.9063, + "step": 38934 + }, + { + "epoch": 2.5, + "learning_rate": 1.1666630885162842e-05, + "loss": 0.9455, + "step": 38935 + }, + { + "epoch": 2.51, + "learning_rate": 1.166641619613989e-05, + "loss": 0.9471, + "step": 38936 + }, + { + "epoch": 2.51, + "learning_rate": 1.1666201507116941e-05, + "loss": 0.9043, + "step": 38937 + }, + { + "epoch": 2.51, + "learning_rate": 1.1665986818093994e-05, + "loss": 0.9405, + "step": 38938 + }, + { + "epoch": 2.51, + "learning_rate": 1.1665772129071043e-05, + "loss": 0.79, + "step": 38939 + }, + { + "epoch": 2.51, + "learning_rate": 1.1665557440048092e-05, + "loss": 0.9415, + "step": 38940 + }, + { + "epoch": 2.51, + "learning_rate": 1.1665342751025142e-05, + "loss": 0.8449, + "step": 38941 + }, + { + "epoch": 2.51, + "learning_rate": 1.1665128062002191e-05, + "loss": 0.8855, + "step": 38942 + }, + { + "epoch": 2.51, + "learning_rate": 1.166491337297924e-05, + "loss": 0.8805, + "step": 38943 + }, + { + "epoch": 2.51, + "learning_rate": 1.166469868395629e-05, + "loss": 0.952, + "step": 38944 + }, + { + "epoch": 2.51, + "learning_rate": 1.166448399493334e-05, + "loss": 0.8799, + "step": 38945 + }, + { + "epoch": 2.51, + "learning_rate": 1.1664269305910389e-05, + "loss": 0.9028, + "step": 38946 + }, + { + "epoch": 2.51, + "learning_rate": 1.166405461688744e-05, + "loss": 0.8134, + "step": 38947 + }, + { + "epoch": 2.51, + "learning_rate": 1.1663839927864488e-05, + "loss": 0.9168, + "step": 38948 + }, + { + "epoch": 2.51, + "learning_rate": 1.1663625238841539e-05, + "loss": 0.932, + "step": 38949 + }, + { + "epoch": 2.51, + "learning_rate": 1.1663410549818588e-05, + "loss": 0.931, + "step": 38950 + }, + { + "epoch": 2.51, + "learning_rate": 1.1663195860795637e-05, + "loss": 0.9273, + "step": 38951 + }, + { + "epoch": 2.51, + "learning_rate": 1.166298117177269e-05, + "loss": 0.883, + "step": 38952 + }, + { + "epoch": 2.51, + "learning_rate": 1.1662766482749738e-05, + "loss": 0.9261, + "step": 38953 + }, + { + "epoch": 2.51, + "learning_rate": 1.1662551793726789e-05, + "loss": 0.9525, + "step": 38954 + }, + { + "epoch": 2.51, + "learning_rate": 1.1662337104703838e-05, + "loss": 0.9511, + "step": 38955 + }, + { + "epoch": 2.51, + "learning_rate": 1.1662122415680888e-05, + "loss": 0.8769, + "step": 38956 + }, + { + "epoch": 2.51, + "learning_rate": 1.1661907726657937e-05, + "loss": 0.9463, + "step": 38957 + }, + { + "epoch": 2.51, + "learning_rate": 1.1661693037634986e-05, + "loss": 0.8687, + "step": 38958 + }, + { + "epoch": 2.51, + "learning_rate": 1.1661478348612037e-05, + "loss": 0.854, + "step": 38959 + }, + { + "epoch": 2.51, + "learning_rate": 1.1661263659589086e-05, + "loss": 0.899, + "step": 38960 + }, + { + "epoch": 2.51, + "learning_rate": 1.1661048970566135e-05, + "loss": 0.8866, + "step": 38961 + }, + { + "epoch": 2.51, + "learning_rate": 1.1660834281543186e-05, + "loss": 0.974, + "step": 38962 + }, + { + "epoch": 2.51, + "learning_rate": 1.1660619592520235e-05, + "loss": 0.8367, + "step": 38963 + }, + { + "epoch": 2.51, + "learning_rate": 1.1660404903497284e-05, + "loss": 0.8832, + "step": 38964 + }, + { + "epoch": 2.51, + "learning_rate": 1.1660190214474334e-05, + "loss": 0.9228, + "step": 38965 + }, + { + "epoch": 2.51, + "learning_rate": 1.1659975525451383e-05, + "loss": 0.9129, + "step": 38966 + }, + { + "epoch": 2.51, + "learning_rate": 1.1659760836428436e-05, + "loss": 0.8771, + "step": 38967 + }, + { + "epoch": 2.51, + "learning_rate": 1.1659546147405485e-05, + "loss": 0.9051, + "step": 38968 + }, + { + "epoch": 2.51, + "learning_rate": 1.1659331458382535e-05, + "loss": 0.9367, + "step": 38969 + }, + { + "epoch": 2.51, + "learning_rate": 1.1659116769359584e-05, + "loss": 0.938, + "step": 38970 + }, + { + "epoch": 2.51, + "learning_rate": 1.1658902080336633e-05, + "loss": 0.903, + "step": 38971 + }, + { + "epoch": 2.51, + "learning_rate": 1.1658687391313684e-05, + "loss": 0.8872, + "step": 38972 + }, + { + "epoch": 2.51, + "learning_rate": 1.1658472702290733e-05, + "loss": 0.8859, + "step": 38973 + }, + { + "epoch": 2.51, + "learning_rate": 1.1658258013267782e-05, + "loss": 0.8862, + "step": 38974 + }, + { + "epoch": 2.51, + "learning_rate": 1.1658043324244832e-05, + "loss": 0.8938, + "step": 38975 + }, + { + "epoch": 2.51, + "learning_rate": 1.1657828635221881e-05, + "loss": 0.9114, + "step": 38976 + }, + { + "epoch": 2.51, + "learning_rate": 1.1657613946198932e-05, + "loss": 0.9075, + "step": 38977 + }, + { + "epoch": 2.51, + "learning_rate": 1.1657399257175981e-05, + "loss": 0.8734, + "step": 38978 + }, + { + "epoch": 2.51, + "learning_rate": 1.165718456815303e-05, + "loss": 0.8804, + "step": 38979 + }, + { + "epoch": 2.51, + "learning_rate": 1.165696987913008e-05, + "loss": 0.9292, + "step": 38980 + }, + { + "epoch": 2.51, + "learning_rate": 1.1656755190107131e-05, + "loss": 0.9509, + "step": 38981 + }, + { + "epoch": 2.51, + "learning_rate": 1.1656540501084182e-05, + "loss": 0.8479, + "step": 38982 + }, + { + "epoch": 2.51, + "learning_rate": 1.1656325812061231e-05, + "loss": 0.8324, + "step": 38983 + }, + { + "epoch": 2.51, + "learning_rate": 1.165611112303828e-05, + "loss": 0.8992, + "step": 38984 + }, + { + "epoch": 2.51, + "learning_rate": 1.165589643401533e-05, + "loss": 0.7941, + "step": 38985 + }, + { + "epoch": 2.51, + "learning_rate": 1.165568174499238e-05, + "loss": 0.9055, + "step": 38986 + }, + { + "epoch": 2.51, + "learning_rate": 1.165546705596943e-05, + "loss": 0.91, + "step": 38987 + }, + { + "epoch": 2.51, + "learning_rate": 1.1655252366946479e-05, + "loss": 0.8673, + "step": 38988 + }, + { + "epoch": 2.51, + "learning_rate": 1.1655037677923528e-05, + "loss": 0.923, + "step": 38989 + }, + { + "epoch": 2.51, + "learning_rate": 1.1654822988900579e-05, + "loss": 0.8498, + "step": 38990 + }, + { + "epoch": 2.51, + "learning_rate": 1.1654608299877628e-05, + "loss": 0.8966, + "step": 38991 + }, + { + "epoch": 2.51, + "learning_rate": 1.1654393610854677e-05, + "loss": 0.8475, + "step": 38992 + }, + { + "epoch": 2.51, + "learning_rate": 1.1654178921831727e-05, + "loss": 0.9208, + "step": 38993 + }, + { + "epoch": 2.51, + "learning_rate": 1.1653964232808776e-05, + "loss": 0.8237, + "step": 38994 + }, + { + "epoch": 2.51, + "learning_rate": 1.1653749543785827e-05, + "loss": 0.9432, + "step": 38995 + }, + { + "epoch": 2.51, + "learning_rate": 1.1653534854762878e-05, + "loss": 0.9257, + "step": 38996 + }, + { + "epoch": 2.51, + "learning_rate": 1.1653320165739928e-05, + "loss": 0.884, + "step": 38997 + }, + { + "epoch": 2.51, + "learning_rate": 1.1653105476716977e-05, + "loss": 0.9028, + "step": 38998 + }, + { + "epoch": 2.51, + "learning_rate": 1.1652890787694026e-05, + "loss": 0.9278, + "step": 38999 + }, + { + "epoch": 2.51, + "learning_rate": 1.1652676098671077e-05, + "loss": 0.954, + "step": 39000 + }, + { + "epoch": 2.51, + "eval_laion/cider": 0.575874138577386, + "eval_laion/meteor": 0.39722556501965295, + "eval_laion/pred_num_words": 37.802734375, + "eval_laion/sacrebleu": 16.42834375874169, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0377085208892822, + "eval_runtime": 102.8838, + "eval_samples_per_second": 4.85, + "eval_steps_per_second": 0.311, + "step": 39000 + }, + { + "epoch": 2.51, + "learning_rate": 1.1652461409648126e-05, + "loss": 0.8685, + "step": 39001 + }, + { + "epoch": 2.51, + "learning_rate": 1.1652246720625175e-05, + "loss": 0.9628, + "step": 39002 + }, + { + "epoch": 2.51, + "learning_rate": 1.1652032031602225e-05, + "loss": 0.9483, + "step": 39003 + }, + { + "epoch": 2.51, + "learning_rate": 1.1651817342579274e-05, + "loss": 0.9274, + "step": 39004 + }, + { + "epoch": 2.51, + "learning_rate": 1.1651602653556325e-05, + "loss": 0.9651, + "step": 39005 + }, + { + "epoch": 2.51, + "learning_rate": 1.1651387964533374e-05, + "loss": 0.8827, + "step": 39006 + }, + { + "epoch": 2.51, + "learning_rate": 1.1651173275510423e-05, + "loss": 0.8771, + "step": 39007 + }, + { + "epoch": 2.51, + "learning_rate": 1.1650958586487474e-05, + "loss": 0.9144, + "step": 39008 + }, + { + "epoch": 2.51, + "learning_rate": 1.1650743897464523e-05, + "loss": 0.8806, + "step": 39009 + }, + { + "epoch": 2.51, + "learning_rate": 1.1650529208441575e-05, + "loss": 0.8648, + "step": 39010 + }, + { + "epoch": 2.51, + "learning_rate": 1.1650314519418624e-05, + "loss": 0.9227, + "step": 39011 + }, + { + "epoch": 2.51, + "learning_rate": 1.1650099830395673e-05, + "loss": 0.9224, + "step": 39012 + }, + { + "epoch": 2.51, + "learning_rate": 1.1649885141372724e-05, + "loss": 1.0091, + "step": 39013 + }, + { + "epoch": 2.51, + "learning_rate": 1.1649670452349773e-05, + "loss": 0.8691, + "step": 39014 + }, + { + "epoch": 2.51, + "learning_rate": 1.1649455763326823e-05, + "loss": 0.9388, + "step": 39015 + }, + { + "epoch": 2.51, + "learning_rate": 1.1649241074303872e-05, + "loss": 0.885, + "step": 39016 + }, + { + "epoch": 2.51, + "learning_rate": 1.1649026385280921e-05, + "loss": 0.9584, + "step": 39017 + }, + { + "epoch": 2.51, + "learning_rate": 1.1648811696257972e-05, + "loss": 0.8828, + "step": 39018 + }, + { + "epoch": 2.51, + "learning_rate": 1.164859700723502e-05, + "loss": 0.8978, + "step": 39019 + }, + { + "epoch": 2.51, + "learning_rate": 1.164838231821207e-05, + "loss": 0.8875, + "step": 39020 + }, + { + "epoch": 2.51, + "learning_rate": 1.164816762918912e-05, + "loss": 0.8675, + "step": 39021 + }, + { + "epoch": 2.51, + "learning_rate": 1.164795294016617e-05, + "loss": 0.848, + "step": 39022 + }, + { + "epoch": 2.51, + "learning_rate": 1.1647738251143218e-05, + "loss": 0.9618, + "step": 39023 + }, + { + "epoch": 2.51, + "learning_rate": 1.1647523562120269e-05, + "loss": 0.9356, + "step": 39024 + }, + { + "epoch": 2.51, + "learning_rate": 1.1647308873097321e-05, + "loss": 0.9233, + "step": 39025 + }, + { + "epoch": 2.51, + "learning_rate": 1.164709418407437e-05, + "loss": 0.9035, + "step": 39026 + }, + { + "epoch": 2.51, + "learning_rate": 1.164687949505142e-05, + "loss": 0.8996, + "step": 39027 + }, + { + "epoch": 2.51, + "learning_rate": 1.164666480602847e-05, + "loss": 0.8795, + "step": 39028 + }, + { + "epoch": 2.51, + "learning_rate": 1.1646450117005519e-05, + "loss": 0.9233, + "step": 39029 + }, + { + "epoch": 2.51, + "learning_rate": 1.1646235427982568e-05, + "loss": 0.9171, + "step": 39030 + }, + { + "epoch": 2.51, + "learning_rate": 1.1646020738959618e-05, + "loss": 0.89, + "step": 39031 + }, + { + "epoch": 2.51, + "learning_rate": 1.1645806049936667e-05, + "loss": 0.91, + "step": 39032 + }, + { + "epoch": 2.51, + "learning_rate": 1.1645591360913716e-05, + "loss": 0.8224, + "step": 39033 + }, + { + "epoch": 2.51, + "learning_rate": 1.1645376671890767e-05, + "loss": 0.8771, + "step": 39034 + }, + { + "epoch": 2.51, + "learning_rate": 1.1645161982867816e-05, + "loss": 0.8119, + "step": 39035 + }, + { + "epoch": 2.51, + "learning_rate": 1.1644947293844867e-05, + "loss": 0.9004, + "step": 39036 + }, + { + "epoch": 2.51, + "learning_rate": 1.1644732604821916e-05, + "loss": 0.8647, + "step": 39037 + }, + { + "epoch": 2.51, + "learning_rate": 1.1644517915798965e-05, + "loss": 0.9259, + "step": 39038 + }, + { + "epoch": 2.51, + "learning_rate": 1.1644303226776017e-05, + "loss": 0.8711, + "step": 39039 + }, + { + "epoch": 2.51, + "learning_rate": 1.1644088537753066e-05, + "loss": 0.8311, + "step": 39040 + }, + { + "epoch": 2.51, + "learning_rate": 1.1643873848730117e-05, + "loss": 0.8811, + "step": 39041 + }, + { + "epoch": 2.51, + "learning_rate": 1.1643659159707166e-05, + "loss": 0.8632, + "step": 39042 + }, + { + "epoch": 2.51, + "learning_rate": 1.1643444470684215e-05, + "loss": 0.9168, + "step": 39043 + }, + { + "epoch": 2.51, + "learning_rate": 1.1643229781661265e-05, + "loss": 0.8887, + "step": 39044 + }, + { + "epoch": 2.51, + "learning_rate": 1.1643015092638314e-05, + "loss": 0.8619, + "step": 39045 + }, + { + "epoch": 2.51, + "learning_rate": 1.1642800403615365e-05, + "loss": 0.9391, + "step": 39046 + }, + { + "epoch": 2.51, + "learning_rate": 1.1642585714592414e-05, + "loss": 0.8854, + "step": 39047 + }, + { + "epoch": 2.51, + "learning_rate": 1.1642371025569463e-05, + "loss": 0.8952, + "step": 39048 + }, + { + "epoch": 2.51, + "learning_rate": 1.1642156336546513e-05, + "loss": 0.9469, + "step": 39049 + }, + { + "epoch": 2.51, + "learning_rate": 1.1641941647523562e-05, + "loss": 0.9291, + "step": 39050 + }, + { + "epoch": 2.51, + "learning_rate": 1.1641726958500611e-05, + "loss": 0.9405, + "step": 39051 + }, + { + "epoch": 2.51, + "learning_rate": 1.1641512269477662e-05, + "loss": 0.9316, + "step": 39052 + }, + { + "epoch": 2.51, + "learning_rate": 1.1641297580454711e-05, + "loss": 0.9257, + "step": 39053 + }, + { + "epoch": 2.51, + "learning_rate": 1.1641082891431763e-05, + "loss": 0.9362, + "step": 39054 + }, + { + "epoch": 2.51, + "learning_rate": 1.1640868202408812e-05, + "loss": 0.9009, + "step": 39055 + }, + { + "epoch": 2.51, + "learning_rate": 1.1640653513385863e-05, + "loss": 0.9372, + "step": 39056 + }, + { + "epoch": 2.51, + "learning_rate": 1.1640438824362912e-05, + "loss": 0.9276, + "step": 39057 + }, + { + "epoch": 2.51, + "learning_rate": 1.1640224135339961e-05, + "loss": 0.8583, + "step": 39058 + }, + { + "epoch": 2.51, + "learning_rate": 1.1640009446317012e-05, + "loss": 0.8795, + "step": 39059 + }, + { + "epoch": 2.51, + "learning_rate": 1.163979475729406e-05, + "loss": 0.8844, + "step": 39060 + }, + { + "epoch": 2.51, + "learning_rate": 1.163958006827111e-05, + "loss": 0.8748, + "step": 39061 + }, + { + "epoch": 2.51, + "learning_rate": 1.163936537924816e-05, + "loss": 0.8794, + "step": 39062 + }, + { + "epoch": 2.51, + "learning_rate": 1.1639150690225209e-05, + "loss": 0.9045, + "step": 39063 + }, + { + "epoch": 2.51, + "learning_rate": 1.163893600120226e-05, + "loss": 0.8623, + "step": 39064 + }, + { + "epoch": 2.51, + "learning_rate": 1.1638721312179309e-05, + "loss": 0.9584, + "step": 39065 + }, + { + "epoch": 2.51, + "learning_rate": 1.1638506623156358e-05, + "loss": 0.939, + "step": 39066 + }, + { + "epoch": 2.51, + "learning_rate": 1.1638291934133408e-05, + "loss": 0.9499, + "step": 39067 + }, + { + "epoch": 2.51, + "learning_rate": 1.1638077245110459e-05, + "loss": 0.8981, + "step": 39068 + }, + { + "epoch": 2.51, + "learning_rate": 1.163786255608751e-05, + "loss": 0.934, + "step": 39069 + }, + { + "epoch": 2.51, + "learning_rate": 1.1637647867064559e-05, + "loss": 0.9064, + "step": 39070 + }, + { + "epoch": 2.51, + "learning_rate": 1.1637433178041608e-05, + "loss": 0.9173, + "step": 39071 + }, + { + "epoch": 2.51, + "learning_rate": 1.1637218489018658e-05, + "loss": 0.8344, + "step": 39072 + }, + { + "epoch": 2.51, + "learning_rate": 1.1637003799995707e-05, + "loss": 0.8985, + "step": 39073 + }, + { + "epoch": 2.51, + "learning_rate": 1.1636789110972758e-05, + "loss": 0.9173, + "step": 39074 + }, + { + "epoch": 2.51, + "learning_rate": 1.1636574421949807e-05, + "loss": 0.8928, + "step": 39075 + }, + { + "epoch": 2.51, + "learning_rate": 1.1636359732926856e-05, + "loss": 0.9208, + "step": 39076 + }, + { + "epoch": 2.51, + "learning_rate": 1.1636145043903906e-05, + "loss": 0.9546, + "step": 39077 + }, + { + "epoch": 2.51, + "learning_rate": 1.1635930354880955e-05, + "loss": 0.929, + "step": 39078 + }, + { + "epoch": 2.51, + "learning_rate": 1.1635715665858004e-05, + "loss": 0.9468, + "step": 39079 + }, + { + "epoch": 2.51, + "learning_rate": 1.1635500976835055e-05, + "loss": 0.9208, + "step": 39080 + }, + { + "epoch": 2.51, + "learning_rate": 1.1635286287812104e-05, + "loss": 0.8584, + "step": 39081 + }, + { + "epoch": 2.51, + "learning_rate": 1.1635071598789155e-05, + "loss": 0.8961, + "step": 39082 + }, + { + "epoch": 2.51, + "learning_rate": 1.1634856909766205e-05, + "loss": 0.8385, + "step": 39083 + }, + { + "epoch": 2.51, + "learning_rate": 1.1634642220743256e-05, + "loss": 0.8602, + "step": 39084 + }, + { + "epoch": 2.51, + "learning_rate": 1.1634427531720305e-05, + "loss": 0.948, + "step": 39085 + }, + { + "epoch": 2.51, + "learning_rate": 1.1634212842697354e-05, + "loss": 0.9167, + "step": 39086 + }, + { + "epoch": 2.51, + "learning_rate": 1.1633998153674405e-05, + "loss": 0.8617, + "step": 39087 + }, + { + "epoch": 2.51, + "learning_rate": 1.1633783464651454e-05, + "loss": 0.8861, + "step": 39088 + }, + { + "epoch": 2.51, + "learning_rate": 1.1633568775628502e-05, + "loss": 0.9082, + "step": 39089 + }, + { + "epoch": 2.51, + "learning_rate": 1.1633354086605553e-05, + "loss": 0.8244, + "step": 39090 + }, + { + "epoch": 2.52, + "learning_rate": 1.1633139397582602e-05, + "loss": 0.9462, + "step": 39091 + }, + { + "epoch": 2.52, + "learning_rate": 1.1632924708559653e-05, + "loss": 0.932, + "step": 39092 + }, + { + "epoch": 2.52, + "learning_rate": 1.1632710019536702e-05, + "loss": 0.8271, + "step": 39093 + }, + { + "epoch": 2.52, + "learning_rate": 1.163249533051375e-05, + "loss": 0.9236, + "step": 39094 + }, + { + "epoch": 2.52, + "learning_rate": 1.1632280641490801e-05, + "loss": 0.9733, + "step": 39095 + }, + { + "epoch": 2.52, + "learning_rate": 1.163206595246785e-05, + "loss": 0.9441, + "step": 39096 + }, + { + "epoch": 2.52, + "learning_rate": 1.1631851263444903e-05, + "loss": 0.8162, + "step": 39097 + }, + { + "epoch": 2.52, + "learning_rate": 1.1631636574421952e-05, + "loss": 0.9036, + "step": 39098 + }, + { + "epoch": 2.52, + "learning_rate": 1.1631421885399e-05, + "loss": 0.9251, + "step": 39099 + }, + { + "epoch": 2.52, + "learning_rate": 1.1631207196376051e-05, + "loss": 0.8801, + "step": 39100 + }, + { + "epoch": 2.52, + "learning_rate": 1.16309925073531e-05, + "loss": 0.9528, + "step": 39101 + }, + { + "epoch": 2.52, + "learning_rate": 1.1630777818330151e-05, + "loss": 0.8673, + "step": 39102 + }, + { + "epoch": 2.52, + "learning_rate": 1.16305631293072e-05, + "loss": 0.9764, + "step": 39103 + }, + { + "epoch": 2.52, + "learning_rate": 1.1630348440284249e-05, + "loss": 0.8994, + "step": 39104 + }, + { + "epoch": 2.52, + "learning_rate": 1.16301337512613e-05, + "loss": 0.8897, + "step": 39105 + }, + { + "epoch": 2.52, + "learning_rate": 1.1629919062238348e-05, + "loss": 0.8816, + "step": 39106 + }, + { + "epoch": 2.52, + "learning_rate": 1.1629704373215397e-05, + "loss": 0.9346, + "step": 39107 + }, + { + "epoch": 2.52, + "learning_rate": 1.1629489684192448e-05, + "loss": 0.9073, + "step": 39108 + }, + { + "epoch": 2.52, + "learning_rate": 1.1629274995169497e-05, + "loss": 0.9345, + "step": 39109 + }, + { + "epoch": 2.52, + "learning_rate": 1.1629060306146546e-05, + "loss": 0.9461, + "step": 39110 + }, + { + "epoch": 2.52, + "learning_rate": 1.1628845617123597e-05, + "loss": 0.7988, + "step": 39111 + }, + { + "epoch": 2.52, + "learning_rate": 1.1628630928100649e-05, + "loss": 0.8959, + "step": 39112 + }, + { + "epoch": 2.52, + "learning_rate": 1.1628416239077698e-05, + "loss": 0.9064, + "step": 39113 + }, + { + "epoch": 2.52, + "learning_rate": 1.1628201550054747e-05, + "loss": 0.8166, + "step": 39114 + }, + { + "epoch": 2.52, + "learning_rate": 1.1627986861031798e-05, + "loss": 0.9174, + "step": 39115 + }, + { + "epoch": 2.52, + "learning_rate": 1.1627772172008847e-05, + "loss": 0.9106, + "step": 39116 + }, + { + "epoch": 2.52, + "learning_rate": 1.1627557482985896e-05, + "loss": 0.845, + "step": 39117 + }, + { + "epoch": 2.52, + "learning_rate": 1.1627342793962946e-05, + "loss": 0.8884, + "step": 39118 + }, + { + "epoch": 2.52, + "learning_rate": 1.1627128104939995e-05, + "loss": 0.9588, + "step": 39119 + }, + { + "epoch": 2.52, + "learning_rate": 1.1626913415917044e-05, + "loss": 0.9192, + "step": 39120 + }, + { + "epoch": 2.52, + "learning_rate": 1.1626698726894095e-05, + "loss": 0.8645, + "step": 39121 + }, + { + "epoch": 2.52, + "learning_rate": 1.1626484037871144e-05, + "loss": 0.9126, + "step": 39122 + }, + { + "epoch": 2.52, + "learning_rate": 1.1626269348848194e-05, + "loss": 0.9231, + "step": 39123 + }, + { + "epoch": 2.52, + "learning_rate": 1.1626054659825243e-05, + "loss": 0.8609, + "step": 39124 + }, + { + "epoch": 2.52, + "learning_rate": 1.1625839970802292e-05, + "loss": 0.9157, + "step": 39125 + }, + { + "epoch": 2.52, + "learning_rate": 1.1625625281779343e-05, + "loss": 0.9524, + "step": 39126 + }, + { + "epoch": 2.52, + "learning_rate": 1.1625410592756394e-05, + "loss": 0.8526, + "step": 39127 + }, + { + "epoch": 2.52, + "learning_rate": 1.1625195903733444e-05, + "loss": 0.8862, + "step": 39128 + }, + { + "epoch": 2.52, + "learning_rate": 1.1624981214710493e-05, + "loss": 0.9157, + "step": 39129 + }, + { + "epoch": 2.52, + "learning_rate": 1.1624766525687542e-05, + "loss": 0.9486, + "step": 39130 + }, + { + "epoch": 2.52, + "learning_rate": 1.1624551836664593e-05, + "loss": 0.8806, + "step": 39131 + }, + { + "epoch": 2.52, + "learning_rate": 1.1624337147641642e-05, + "loss": 0.8222, + "step": 39132 + }, + { + "epoch": 2.52, + "learning_rate": 1.1624122458618693e-05, + "loss": 0.8625, + "step": 39133 + }, + { + "epoch": 2.52, + "learning_rate": 1.1623907769595741e-05, + "loss": 0.928, + "step": 39134 + }, + { + "epoch": 2.52, + "learning_rate": 1.162369308057279e-05, + "loss": 0.8836, + "step": 39135 + }, + { + "epoch": 2.52, + "learning_rate": 1.1623478391549841e-05, + "loss": 0.8771, + "step": 39136 + }, + { + "epoch": 2.52, + "learning_rate": 1.162326370252689e-05, + "loss": 0.9567, + "step": 39137 + }, + { + "epoch": 2.52, + "learning_rate": 1.1623049013503939e-05, + "loss": 0.8823, + "step": 39138 + }, + { + "epoch": 2.52, + "learning_rate": 1.162283432448099e-05, + "loss": 0.8186, + "step": 39139 + }, + { + "epoch": 2.52, + "learning_rate": 1.1622619635458039e-05, + "loss": 0.8701, + "step": 39140 + }, + { + "epoch": 2.52, + "learning_rate": 1.1622404946435091e-05, + "loss": 1.0025, + "step": 39141 + }, + { + "epoch": 2.52, + "learning_rate": 1.162219025741214e-05, + "loss": 0.7898, + "step": 39142 + }, + { + "epoch": 2.52, + "learning_rate": 1.162197556838919e-05, + "loss": 0.9036, + "step": 39143 + }, + { + "epoch": 2.52, + "learning_rate": 1.162176087936624e-05, + "loss": 0.8735, + "step": 39144 + }, + { + "epoch": 2.52, + "learning_rate": 1.1621546190343289e-05, + "loss": 0.9417, + "step": 39145 + }, + { + "epoch": 2.52, + "learning_rate": 1.162133150132034e-05, + "loss": 0.8717, + "step": 39146 + }, + { + "epoch": 2.52, + "learning_rate": 1.1621116812297388e-05, + "loss": 0.8695, + "step": 39147 + }, + { + "epoch": 2.52, + "learning_rate": 1.1620902123274437e-05, + "loss": 0.8732, + "step": 39148 + }, + { + "epoch": 2.52, + "learning_rate": 1.1620687434251488e-05, + "loss": 0.8909, + "step": 39149 + }, + { + "epoch": 2.52, + "learning_rate": 1.1620472745228537e-05, + "loss": 0.9164, + "step": 39150 + }, + { + "epoch": 2.52, + "learning_rate": 1.1620258056205587e-05, + "loss": 0.8714, + "step": 39151 + }, + { + "epoch": 2.52, + "learning_rate": 1.1620043367182636e-05, + "loss": 0.9097, + "step": 39152 + }, + { + "epoch": 2.52, + "learning_rate": 1.1619828678159685e-05, + "loss": 0.8823, + "step": 39153 + }, + { + "epoch": 2.52, + "learning_rate": 1.1619613989136736e-05, + "loss": 0.8492, + "step": 39154 + }, + { + "epoch": 2.52, + "learning_rate": 1.1619399300113785e-05, + "loss": 0.8788, + "step": 39155 + }, + { + "epoch": 2.52, + "learning_rate": 1.1619184611090837e-05, + "loss": 0.8706, + "step": 39156 + }, + { + "epoch": 2.52, + "learning_rate": 1.1618969922067886e-05, + "loss": 0.9062, + "step": 39157 + }, + { + "epoch": 2.52, + "learning_rate": 1.1618755233044935e-05, + "loss": 0.886, + "step": 39158 + }, + { + "epoch": 2.52, + "learning_rate": 1.1618540544021986e-05, + "loss": 0.8594, + "step": 39159 + }, + { + "epoch": 2.52, + "learning_rate": 1.1618325854999035e-05, + "loss": 0.9466, + "step": 39160 + }, + { + "epoch": 2.52, + "learning_rate": 1.1618111165976086e-05, + "loss": 0.9087, + "step": 39161 + }, + { + "epoch": 2.52, + "learning_rate": 1.1617896476953135e-05, + "loss": 0.8594, + "step": 39162 + }, + { + "epoch": 2.52, + "learning_rate": 1.1617681787930183e-05, + "loss": 0.87, + "step": 39163 + }, + { + "epoch": 2.52, + "learning_rate": 1.1617467098907234e-05, + "loss": 0.9422, + "step": 39164 + }, + { + "epoch": 2.52, + "learning_rate": 1.1617252409884283e-05, + "loss": 0.8693, + "step": 39165 + }, + { + "epoch": 2.52, + "learning_rate": 1.1617037720861332e-05, + "loss": 0.9065, + "step": 39166 + }, + { + "epoch": 2.52, + "learning_rate": 1.1616823031838383e-05, + "loss": 0.9281, + "step": 39167 + }, + { + "epoch": 2.52, + "learning_rate": 1.1616608342815432e-05, + "loss": 0.9391, + "step": 39168 + }, + { + "epoch": 2.52, + "learning_rate": 1.1616393653792482e-05, + "loss": 0.8422, + "step": 39169 + }, + { + "epoch": 2.52, + "learning_rate": 1.1616178964769533e-05, + "loss": 0.9175, + "step": 39170 + }, + { + "epoch": 2.52, + "learning_rate": 1.1615964275746584e-05, + "loss": 0.8584, + "step": 39171 + }, + { + "epoch": 2.52, + "learning_rate": 1.1615749586723633e-05, + "loss": 0.8762, + "step": 39172 + }, + { + "epoch": 2.52, + "learning_rate": 1.1615534897700682e-05, + "loss": 0.8812, + "step": 39173 + }, + { + "epoch": 2.52, + "learning_rate": 1.1615320208677732e-05, + "loss": 0.8936, + "step": 39174 + }, + { + "epoch": 2.52, + "learning_rate": 1.1615105519654781e-05, + "loss": 0.8979, + "step": 39175 + }, + { + "epoch": 2.52, + "learning_rate": 1.161489083063183e-05, + "loss": 0.9281, + "step": 39176 + }, + { + "epoch": 2.52, + "learning_rate": 1.1614676141608881e-05, + "loss": 0.8283, + "step": 39177 + }, + { + "epoch": 2.52, + "learning_rate": 1.161446145258593e-05, + "loss": 0.9022, + "step": 39178 + }, + { + "epoch": 2.52, + "learning_rate": 1.161424676356298e-05, + "loss": 0.8772, + "step": 39179 + }, + { + "epoch": 2.52, + "learning_rate": 1.161403207454003e-05, + "loss": 0.941, + "step": 39180 + }, + { + "epoch": 2.52, + "learning_rate": 1.1613817385517078e-05, + "loss": 0.8852, + "step": 39181 + }, + { + "epoch": 2.52, + "learning_rate": 1.1613602696494129e-05, + "loss": 0.8205, + "step": 39182 + }, + { + "epoch": 2.52, + "learning_rate": 1.1613388007471178e-05, + "loss": 0.878, + "step": 39183 + }, + { + "epoch": 2.52, + "learning_rate": 1.1613173318448227e-05, + "loss": 0.9185, + "step": 39184 + }, + { + "epoch": 2.52, + "learning_rate": 1.161295862942528e-05, + "loss": 0.8448, + "step": 39185 + }, + { + "epoch": 2.52, + "learning_rate": 1.1612743940402328e-05, + "loss": 0.9552, + "step": 39186 + }, + { + "epoch": 2.52, + "learning_rate": 1.1612529251379379e-05, + "loss": 0.8953, + "step": 39187 + }, + { + "epoch": 2.52, + "learning_rate": 1.1612314562356428e-05, + "loss": 0.8413, + "step": 39188 + }, + { + "epoch": 2.52, + "learning_rate": 1.1612099873333479e-05, + "loss": 0.8748, + "step": 39189 + }, + { + "epoch": 2.52, + "learning_rate": 1.1611885184310528e-05, + "loss": 0.9489, + "step": 39190 + }, + { + "epoch": 2.52, + "learning_rate": 1.1611670495287577e-05, + "loss": 0.9494, + "step": 39191 + }, + { + "epoch": 2.52, + "learning_rate": 1.1611455806264627e-05, + "loss": 0.9519, + "step": 39192 + }, + { + "epoch": 2.52, + "learning_rate": 1.1611241117241676e-05, + "loss": 0.8285, + "step": 39193 + }, + { + "epoch": 2.52, + "learning_rate": 1.1611026428218725e-05, + "loss": 0.8862, + "step": 39194 + }, + { + "epoch": 2.52, + "learning_rate": 1.1610811739195776e-05, + "loss": 0.8578, + "step": 39195 + }, + { + "epoch": 2.52, + "learning_rate": 1.1610597050172825e-05, + "loss": 0.8955, + "step": 39196 + }, + { + "epoch": 2.52, + "learning_rate": 1.1610382361149874e-05, + "loss": 0.9504, + "step": 39197 + }, + { + "epoch": 2.52, + "learning_rate": 1.1610167672126924e-05, + "loss": 0.8748, + "step": 39198 + }, + { + "epoch": 2.52, + "learning_rate": 1.1609952983103977e-05, + "loss": 0.8929, + "step": 39199 + }, + { + "epoch": 2.52, + "learning_rate": 1.1609738294081026e-05, + "loss": 0.8742, + "step": 39200 + }, + { + "epoch": 2.52, + "learning_rate": 1.1609523605058075e-05, + "loss": 0.9563, + "step": 39201 + }, + { + "epoch": 2.52, + "learning_rate": 1.1609308916035125e-05, + "loss": 0.9183, + "step": 39202 + }, + { + "epoch": 2.52, + "learning_rate": 1.1609094227012174e-05, + "loss": 0.8933, + "step": 39203 + }, + { + "epoch": 2.52, + "learning_rate": 1.1608879537989223e-05, + "loss": 0.9752, + "step": 39204 + }, + { + "epoch": 2.52, + "learning_rate": 1.1608664848966274e-05, + "loss": 0.8955, + "step": 39205 + }, + { + "epoch": 2.52, + "learning_rate": 1.1608450159943323e-05, + "loss": 0.9729, + "step": 39206 + }, + { + "epoch": 2.52, + "learning_rate": 1.1608235470920372e-05, + "loss": 0.9361, + "step": 39207 + }, + { + "epoch": 2.52, + "learning_rate": 1.1608020781897422e-05, + "loss": 0.8516, + "step": 39208 + }, + { + "epoch": 2.52, + "learning_rate": 1.1607806092874471e-05, + "loss": 0.9625, + "step": 39209 + }, + { + "epoch": 2.52, + "learning_rate": 1.1607591403851522e-05, + "loss": 0.9568, + "step": 39210 + }, + { + "epoch": 2.52, + "learning_rate": 1.1607376714828571e-05, + "loss": 0.8413, + "step": 39211 + }, + { + "epoch": 2.52, + "learning_rate": 1.160716202580562e-05, + "loss": 0.8611, + "step": 39212 + }, + { + "epoch": 2.52, + "learning_rate": 1.160694733678267e-05, + "loss": 0.8797, + "step": 39213 + }, + { + "epoch": 2.52, + "learning_rate": 1.1606732647759721e-05, + "loss": 0.8578, + "step": 39214 + }, + { + "epoch": 2.52, + "learning_rate": 1.1606517958736772e-05, + "loss": 0.9047, + "step": 39215 + }, + { + "epoch": 2.52, + "learning_rate": 1.1606303269713821e-05, + "loss": 0.8823, + "step": 39216 + }, + { + "epoch": 2.52, + "learning_rate": 1.160608858069087e-05, + "loss": 0.8779, + "step": 39217 + }, + { + "epoch": 2.52, + "learning_rate": 1.160587389166792e-05, + "loss": 0.9147, + "step": 39218 + }, + { + "epoch": 2.52, + "learning_rate": 1.160565920264497e-05, + "loss": 0.8926, + "step": 39219 + }, + { + "epoch": 2.52, + "learning_rate": 1.160544451362202e-05, + "loss": 0.9241, + "step": 39220 + }, + { + "epoch": 2.52, + "learning_rate": 1.160522982459907e-05, + "loss": 0.8768, + "step": 39221 + }, + { + "epoch": 2.52, + "learning_rate": 1.1605015135576118e-05, + "loss": 0.8438, + "step": 39222 + }, + { + "epoch": 2.52, + "learning_rate": 1.1604800446553169e-05, + "loss": 0.8503, + "step": 39223 + }, + { + "epoch": 2.52, + "learning_rate": 1.1604585757530218e-05, + "loss": 0.9418, + "step": 39224 + }, + { + "epoch": 2.52, + "learning_rate": 1.1604371068507267e-05, + "loss": 0.9281, + "step": 39225 + }, + { + "epoch": 2.52, + "learning_rate": 1.1604156379484317e-05, + "loss": 0.8933, + "step": 39226 + }, + { + "epoch": 2.52, + "learning_rate": 1.1603941690461366e-05, + "loss": 0.91, + "step": 39227 + }, + { + "epoch": 2.52, + "learning_rate": 1.1603727001438419e-05, + "loss": 0.8739, + "step": 39228 + }, + { + "epoch": 2.52, + "learning_rate": 1.1603512312415468e-05, + "loss": 0.906, + "step": 39229 + }, + { + "epoch": 2.52, + "learning_rate": 1.1603297623392518e-05, + "loss": 0.91, + "step": 39230 + }, + { + "epoch": 2.52, + "learning_rate": 1.1603082934369567e-05, + "loss": 0.897, + "step": 39231 + }, + { + "epoch": 2.52, + "learning_rate": 1.1602868245346616e-05, + "loss": 0.8351, + "step": 39232 + }, + { + "epoch": 2.52, + "learning_rate": 1.1602653556323667e-05, + "loss": 0.8479, + "step": 39233 + }, + { + "epoch": 2.52, + "learning_rate": 1.1602438867300716e-05, + "loss": 0.916, + "step": 39234 + }, + { + "epoch": 2.52, + "learning_rate": 1.1602224178277765e-05, + "loss": 0.87, + "step": 39235 + }, + { + "epoch": 2.52, + "learning_rate": 1.1602009489254816e-05, + "loss": 0.8507, + "step": 39236 + }, + { + "epoch": 2.52, + "learning_rate": 1.1601794800231864e-05, + "loss": 0.9055, + "step": 39237 + }, + { + "epoch": 2.52, + "learning_rate": 1.1601580111208915e-05, + "loss": 0.8234, + "step": 39238 + }, + { + "epoch": 2.52, + "learning_rate": 1.1601365422185964e-05, + "loss": 0.919, + "step": 39239 + }, + { + "epoch": 2.52, + "learning_rate": 1.1601150733163013e-05, + "loss": 0.9014, + "step": 39240 + }, + { + "epoch": 2.52, + "learning_rate": 1.1600936044140064e-05, + "loss": 0.8668, + "step": 39241 + }, + { + "epoch": 2.52, + "learning_rate": 1.1600721355117113e-05, + "loss": 0.941, + "step": 39242 + }, + { + "epoch": 2.52, + "learning_rate": 1.1600506666094165e-05, + "loss": 0.9653, + "step": 39243 + }, + { + "epoch": 2.52, + "learning_rate": 1.1600291977071214e-05, + "loss": 0.8826, + "step": 39244 + }, + { + "epoch": 2.52, + "learning_rate": 1.1600077288048263e-05, + "loss": 0.7945, + "step": 39245 + }, + { + "epoch": 2.52, + "learning_rate": 1.1599862599025314e-05, + "loss": 0.8558, + "step": 39246 + }, + { + "epoch": 2.53, + "learning_rate": 1.1599647910002363e-05, + "loss": 0.9073, + "step": 39247 + }, + { + "epoch": 2.53, + "learning_rate": 1.1599433220979413e-05, + "loss": 0.8471, + "step": 39248 + }, + { + "epoch": 2.53, + "learning_rate": 1.1599218531956462e-05, + "loss": 0.8439, + "step": 39249 + }, + { + "epoch": 2.53, + "learning_rate": 1.1599003842933511e-05, + "loss": 0.9008, + "step": 39250 + }, + { + "epoch": 2.53, + "learning_rate": 1.1598789153910562e-05, + "loss": 0.8585, + "step": 39251 + }, + { + "epoch": 2.53, + "learning_rate": 1.159857446488761e-05, + "loss": 0.8995, + "step": 39252 + }, + { + "epoch": 2.53, + "learning_rate": 1.159835977586466e-05, + "loss": 0.9009, + "step": 39253 + }, + { + "epoch": 2.53, + "learning_rate": 1.159814508684171e-05, + "loss": 0.9284, + "step": 39254 + }, + { + "epoch": 2.53, + "learning_rate": 1.159793039781876e-05, + "loss": 0.9621, + "step": 39255 + }, + { + "epoch": 2.53, + "learning_rate": 1.1597715708795808e-05, + "loss": 0.8151, + "step": 39256 + }, + { + "epoch": 2.53, + "learning_rate": 1.159750101977286e-05, + "loss": 0.8687, + "step": 39257 + }, + { + "epoch": 2.53, + "learning_rate": 1.1597286330749911e-05, + "loss": 0.8742, + "step": 39258 + }, + { + "epoch": 2.53, + "learning_rate": 1.159707164172696e-05, + "loss": 0.9025, + "step": 39259 + }, + { + "epoch": 2.53, + "learning_rate": 1.159685695270401e-05, + "loss": 0.9325, + "step": 39260 + }, + { + "epoch": 2.53, + "learning_rate": 1.159664226368106e-05, + "loss": 0.9242, + "step": 39261 + }, + { + "epoch": 2.53, + "learning_rate": 1.1596427574658109e-05, + "loss": 0.8654, + "step": 39262 + }, + { + "epoch": 2.53, + "learning_rate": 1.1596212885635158e-05, + "loss": 0.848, + "step": 39263 + }, + { + "epoch": 2.53, + "learning_rate": 1.1595998196612209e-05, + "loss": 0.9548, + "step": 39264 + }, + { + "epoch": 2.53, + "learning_rate": 1.1595783507589258e-05, + "loss": 0.9009, + "step": 39265 + }, + { + "epoch": 2.53, + "learning_rate": 1.1595568818566306e-05, + "loss": 0.9071, + "step": 39266 + }, + { + "epoch": 2.53, + "learning_rate": 1.1595354129543357e-05, + "loss": 0.9086, + "step": 39267 + }, + { + "epoch": 2.53, + "learning_rate": 1.1595139440520406e-05, + "loss": 0.927, + "step": 39268 + }, + { + "epoch": 2.53, + "learning_rate": 1.1594924751497457e-05, + "loss": 0.8305, + "step": 39269 + }, + { + "epoch": 2.53, + "learning_rate": 1.1594710062474506e-05, + "loss": 0.8632, + "step": 39270 + }, + { + "epoch": 2.53, + "learning_rate": 1.1594495373451555e-05, + "loss": 0.8662, + "step": 39271 + }, + { + "epoch": 2.53, + "learning_rate": 1.1594280684428607e-05, + "loss": 0.8792, + "step": 39272 + }, + { + "epoch": 2.53, + "learning_rate": 1.1594065995405656e-05, + "loss": 0.9069, + "step": 39273 + }, + { + "epoch": 2.53, + "learning_rate": 1.1593851306382707e-05, + "loss": 0.9003, + "step": 39274 + }, + { + "epoch": 2.53, + "learning_rate": 1.1593636617359756e-05, + "loss": 0.9035, + "step": 39275 + }, + { + "epoch": 2.53, + "learning_rate": 1.1593421928336806e-05, + "loss": 0.9917, + "step": 39276 + }, + { + "epoch": 2.53, + "learning_rate": 1.1593207239313855e-05, + "loss": 0.8854, + "step": 39277 + }, + { + "epoch": 2.53, + "learning_rate": 1.1592992550290904e-05, + "loss": 0.9259, + "step": 39278 + }, + { + "epoch": 2.53, + "learning_rate": 1.1592777861267955e-05, + "loss": 0.8913, + "step": 39279 + }, + { + "epoch": 2.53, + "learning_rate": 1.1592563172245004e-05, + "loss": 0.8435, + "step": 39280 + }, + { + "epoch": 2.53, + "learning_rate": 1.1592348483222053e-05, + "loss": 0.8927, + "step": 39281 + }, + { + "epoch": 2.53, + "learning_rate": 1.1592133794199103e-05, + "loss": 0.9271, + "step": 39282 + }, + { + "epoch": 2.53, + "learning_rate": 1.1591919105176152e-05, + "loss": 0.8739, + "step": 39283 + }, + { + "epoch": 2.53, + "learning_rate": 1.1591704416153201e-05, + "loss": 0.9205, + "step": 39284 + }, + { + "epoch": 2.53, + "learning_rate": 1.1591489727130252e-05, + "loss": 0.9415, + "step": 39285 + }, + { + "epoch": 2.53, + "learning_rate": 1.1591275038107304e-05, + "loss": 0.9671, + "step": 39286 + }, + { + "epoch": 2.53, + "learning_rate": 1.1591060349084353e-05, + "loss": 0.9206, + "step": 39287 + }, + { + "epoch": 2.53, + "learning_rate": 1.1590845660061402e-05, + "loss": 0.9016, + "step": 39288 + }, + { + "epoch": 2.53, + "learning_rate": 1.1590630971038453e-05, + "loss": 0.8988, + "step": 39289 + }, + { + "epoch": 2.53, + "learning_rate": 1.1590416282015502e-05, + "loss": 0.9486, + "step": 39290 + }, + { + "epoch": 2.53, + "learning_rate": 1.1590201592992551e-05, + "loss": 0.9277, + "step": 39291 + }, + { + "epoch": 2.53, + "learning_rate": 1.1589986903969602e-05, + "loss": 0.8976, + "step": 39292 + }, + { + "epoch": 2.53, + "learning_rate": 1.158977221494665e-05, + "loss": 0.8718, + "step": 39293 + }, + { + "epoch": 2.53, + "learning_rate": 1.15895575259237e-05, + "loss": 0.9341, + "step": 39294 + }, + { + "epoch": 2.53, + "learning_rate": 1.158934283690075e-05, + "loss": 0.8588, + "step": 39295 + }, + { + "epoch": 2.53, + "learning_rate": 1.15891281478778e-05, + "loss": 0.8621, + "step": 39296 + }, + { + "epoch": 2.53, + "learning_rate": 1.158891345885485e-05, + "loss": 0.9399, + "step": 39297 + }, + { + "epoch": 2.53, + "learning_rate": 1.1588698769831899e-05, + "loss": 0.9344, + "step": 39298 + }, + { + "epoch": 2.53, + "learning_rate": 1.1588484080808948e-05, + "loss": 0.8956, + "step": 39299 + }, + { + "epoch": 2.53, + "learning_rate": 1.1588269391785998e-05, + "loss": 0.9745, + "step": 39300 + }, + { + "epoch": 2.53, + "learning_rate": 1.1588054702763049e-05, + "loss": 0.8906, + "step": 39301 + }, + { + "epoch": 2.53, + "learning_rate": 1.15878400137401e-05, + "loss": 0.9146, + "step": 39302 + }, + { + "epoch": 2.53, + "learning_rate": 1.1587625324717149e-05, + "loss": 0.8733, + "step": 39303 + }, + { + "epoch": 2.53, + "learning_rate": 1.1587410635694198e-05, + "loss": 0.945, + "step": 39304 + }, + { + "epoch": 2.53, + "learning_rate": 1.1587195946671248e-05, + "loss": 0.8441, + "step": 39305 + }, + { + "epoch": 2.53, + "learning_rate": 1.1586981257648297e-05, + "loss": 0.9049, + "step": 39306 + }, + { + "epoch": 2.53, + "learning_rate": 1.1586766568625348e-05, + "loss": 0.9017, + "step": 39307 + }, + { + "epoch": 2.53, + "learning_rate": 1.1586551879602397e-05, + "loss": 0.8893, + "step": 39308 + }, + { + "epoch": 2.53, + "learning_rate": 1.1586337190579446e-05, + "loss": 0.9101, + "step": 39309 + }, + { + "epoch": 2.53, + "learning_rate": 1.1586122501556497e-05, + "loss": 0.8989, + "step": 39310 + }, + { + "epoch": 2.53, + "learning_rate": 1.1585907812533545e-05, + "loss": 0.9335, + "step": 39311 + }, + { + "epoch": 2.53, + "learning_rate": 1.1585693123510594e-05, + "loss": 0.8738, + "step": 39312 + }, + { + "epoch": 2.53, + "learning_rate": 1.1585478434487645e-05, + "loss": 0.9017, + "step": 39313 + }, + { + "epoch": 2.53, + "learning_rate": 1.1585263745464694e-05, + "loss": 0.8716, + "step": 39314 + }, + { + "epoch": 2.53, + "learning_rate": 1.1585049056441746e-05, + "loss": 0.9325, + "step": 39315 + }, + { + "epoch": 2.53, + "learning_rate": 1.1584834367418795e-05, + "loss": 0.8837, + "step": 39316 + }, + { + "epoch": 2.53, + "learning_rate": 1.1584619678395846e-05, + "loss": 0.8713, + "step": 39317 + }, + { + "epoch": 2.53, + "learning_rate": 1.1584404989372895e-05, + "loss": 0.873, + "step": 39318 + }, + { + "epoch": 2.53, + "learning_rate": 1.1584190300349944e-05, + "loss": 0.9497, + "step": 39319 + }, + { + "epoch": 2.53, + "learning_rate": 1.1583975611326995e-05, + "loss": 0.9455, + "step": 39320 + }, + { + "epoch": 2.53, + "learning_rate": 1.1583760922304044e-05, + "loss": 0.9054, + "step": 39321 + }, + { + "epoch": 2.53, + "learning_rate": 1.1583546233281093e-05, + "loss": 0.9003, + "step": 39322 + }, + { + "epoch": 2.53, + "learning_rate": 1.1583331544258143e-05, + "loss": 0.8403, + "step": 39323 + }, + { + "epoch": 2.53, + "learning_rate": 1.1583116855235192e-05, + "loss": 0.9033, + "step": 39324 + }, + { + "epoch": 2.53, + "learning_rate": 1.1582902166212243e-05, + "loss": 0.9354, + "step": 39325 + }, + { + "epoch": 2.53, + "learning_rate": 1.1582687477189292e-05, + "loss": 0.9099, + "step": 39326 + }, + { + "epoch": 2.53, + "learning_rate": 1.158247278816634e-05, + "loss": 0.8803, + "step": 39327 + }, + { + "epoch": 2.53, + "learning_rate": 1.1582258099143391e-05, + "loss": 0.9023, + "step": 39328 + }, + { + "epoch": 2.53, + "learning_rate": 1.158204341012044e-05, + "loss": 0.9016, + "step": 39329 + }, + { + "epoch": 2.53, + "learning_rate": 1.1581828721097493e-05, + "loss": 0.8715, + "step": 39330 + }, + { + "epoch": 2.53, + "learning_rate": 1.1581614032074542e-05, + "loss": 0.9949, + "step": 39331 + }, + { + "epoch": 2.53, + "learning_rate": 1.158139934305159e-05, + "loss": 0.8984, + "step": 39332 + }, + { + "epoch": 2.53, + "learning_rate": 1.1581184654028641e-05, + "loss": 0.8944, + "step": 39333 + }, + { + "epoch": 2.53, + "learning_rate": 1.158096996500569e-05, + "loss": 0.8877, + "step": 39334 + }, + { + "epoch": 2.53, + "learning_rate": 1.1580755275982741e-05, + "loss": 0.8621, + "step": 39335 + }, + { + "epoch": 2.53, + "learning_rate": 1.158054058695979e-05, + "loss": 0.9331, + "step": 39336 + }, + { + "epoch": 2.53, + "learning_rate": 1.1580325897936839e-05, + "loss": 0.8515, + "step": 39337 + }, + { + "epoch": 2.53, + "learning_rate": 1.158011120891389e-05, + "loss": 0.8508, + "step": 39338 + }, + { + "epoch": 2.53, + "learning_rate": 1.1579896519890939e-05, + "loss": 0.9381, + "step": 39339 + }, + { + "epoch": 2.53, + "learning_rate": 1.1579681830867987e-05, + "loss": 0.8751, + "step": 39340 + }, + { + "epoch": 2.53, + "learning_rate": 1.1579467141845038e-05, + "loss": 0.8831, + "step": 39341 + }, + { + "epoch": 2.53, + "learning_rate": 1.1579252452822087e-05, + "loss": 0.8872, + "step": 39342 + }, + { + "epoch": 2.53, + "learning_rate": 1.1579037763799136e-05, + "loss": 0.9185, + "step": 39343 + }, + { + "epoch": 2.53, + "learning_rate": 1.1578823074776188e-05, + "loss": 0.8933, + "step": 39344 + }, + { + "epoch": 2.53, + "learning_rate": 1.1578608385753239e-05, + "loss": 0.9425, + "step": 39345 + }, + { + "epoch": 2.53, + "learning_rate": 1.1578393696730288e-05, + "loss": 0.9206, + "step": 39346 + }, + { + "epoch": 2.53, + "learning_rate": 1.1578179007707337e-05, + "loss": 0.9136, + "step": 39347 + }, + { + "epoch": 2.53, + "learning_rate": 1.1577964318684388e-05, + "loss": 0.9023, + "step": 39348 + }, + { + "epoch": 2.53, + "learning_rate": 1.1577749629661437e-05, + "loss": 0.9052, + "step": 39349 + }, + { + "epoch": 2.53, + "learning_rate": 1.1577534940638486e-05, + "loss": 0.8904, + "step": 39350 + }, + { + "epoch": 2.53, + "learning_rate": 1.1577320251615536e-05, + "loss": 0.8805, + "step": 39351 + }, + { + "epoch": 2.53, + "learning_rate": 1.1577105562592585e-05, + "loss": 0.9647, + "step": 39352 + }, + { + "epoch": 2.53, + "learning_rate": 1.1576890873569634e-05, + "loss": 0.9433, + "step": 39353 + }, + { + "epoch": 2.53, + "learning_rate": 1.1576676184546685e-05, + "loss": 0.8846, + "step": 39354 + }, + { + "epoch": 2.53, + "learning_rate": 1.1576461495523734e-05, + "loss": 0.8634, + "step": 39355 + }, + { + "epoch": 2.53, + "learning_rate": 1.1576246806500784e-05, + "loss": 0.9137, + "step": 39356 + }, + { + "epoch": 2.53, + "learning_rate": 1.1576032117477833e-05, + "loss": 0.8696, + "step": 39357 + }, + { + "epoch": 2.53, + "learning_rate": 1.1575817428454882e-05, + "loss": 0.9525, + "step": 39358 + }, + { + "epoch": 2.53, + "learning_rate": 1.1575602739431935e-05, + "loss": 0.8201, + "step": 39359 + }, + { + "epoch": 2.53, + "learning_rate": 1.1575388050408984e-05, + "loss": 0.8572, + "step": 39360 + }, + { + "epoch": 2.53, + "learning_rate": 1.1575173361386034e-05, + "loss": 0.9476, + "step": 39361 + }, + { + "epoch": 2.53, + "learning_rate": 1.1574958672363083e-05, + "loss": 0.8881, + "step": 39362 + }, + { + "epoch": 2.53, + "learning_rate": 1.1574743983340132e-05, + "loss": 0.8888, + "step": 39363 + }, + { + "epoch": 2.53, + "learning_rate": 1.1574529294317183e-05, + "loss": 0.8848, + "step": 39364 + }, + { + "epoch": 2.53, + "learning_rate": 1.1574314605294232e-05, + "loss": 0.8702, + "step": 39365 + }, + { + "epoch": 2.53, + "learning_rate": 1.1574099916271283e-05, + "loss": 0.9065, + "step": 39366 + }, + { + "epoch": 2.53, + "learning_rate": 1.1573885227248332e-05, + "loss": 0.8118, + "step": 39367 + }, + { + "epoch": 2.53, + "learning_rate": 1.157367053822538e-05, + "loss": 0.8861, + "step": 39368 + }, + { + "epoch": 2.53, + "learning_rate": 1.1573455849202431e-05, + "loss": 0.8714, + "step": 39369 + }, + { + "epoch": 2.53, + "learning_rate": 1.157324116017948e-05, + "loss": 0.9609, + "step": 39370 + }, + { + "epoch": 2.53, + "learning_rate": 1.1573026471156529e-05, + "loss": 0.9421, + "step": 39371 + }, + { + "epoch": 2.53, + "learning_rate": 1.157281178213358e-05, + "loss": 0.8325, + "step": 39372 + }, + { + "epoch": 2.53, + "learning_rate": 1.157259709311063e-05, + "loss": 0.9027, + "step": 39373 + }, + { + "epoch": 2.53, + "learning_rate": 1.1572382404087681e-05, + "loss": 1.0082, + "step": 39374 + }, + { + "epoch": 2.53, + "learning_rate": 1.157216771506473e-05, + "loss": 0.9011, + "step": 39375 + }, + { + "epoch": 2.53, + "learning_rate": 1.157195302604178e-05, + "loss": 0.9045, + "step": 39376 + }, + { + "epoch": 2.53, + "learning_rate": 1.157173833701883e-05, + "loss": 0.9116, + "step": 39377 + }, + { + "epoch": 2.53, + "learning_rate": 1.1571523647995879e-05, + "loss": 0.9008, + "step": 39378 + }, + { + "epoch": 2.53, + "learning_rate": 1.157130895897293e-05, + "loss": 0.9439, + "step": 39379 + }, + { + "epoch": 2.53, + "learning_rate": 1.1571094269949978e-05, + "loss": 0.8817, + "step": 39380 + }, + { + "epoch": 2.53, + "learning_rate": 1.1570879580927027e-05, + "loss": 0.9134, + "step": 39381 + }, + { + "epoch": 2.53, + "learning_rate": 1.1570664891904078e-05, + "loss": 0.9981, + "step": 39382 + }, + { + "epoch": 2.53, + "learning_rate": 1.1570450202881127e-05, + "loss": 0.8888, + "step": 39383 + }, + { + "epoch": 2.53, + "learning_rate": 1.1570235513858178e-05, + "loss": 0.9608, + "step": 39384 + }, + { + "epoch": 2.53, + "learning_rate": 1.1570020824835226e-05, + "loss": 0.8651, + "step": 39385 + }, + { + "epoch": 2.53, + "learning_rate": 1.1569806135812275e-05, + "loss": 0.8474, + "step": 39386 + }, + { + "epoch": 2.53, + "learning_rate": 1.1569591446789326e-05, + "loss": 0.9129, + "step": 39387 + }, + { + "epoch": 2.53, + "learning_rate": 1.1569376757766377e-05, + "loss": 0.9144, + "step": 39388 + }, + { + "epoch": 2.53, + "learning_rate": 1.1569162068743427e-05, + "loss": 0.884, + "step": 39389 + }, + { + "epoch": 2.53, + "learning_rate": 1.1568947379720476e-05, + "loss": 0.9086, + "step": 39390 + }, + { + "epoch": 2.53, + "learning_rate": 1.1568732690697525e-05, + "loss": 0.9075, + "step": 39391 + }, + { + "epoch": 2.53, + "learning_rate": 1.1568518001674576e-05, + "loss": 0.8486, + "step": 39392 + }, + { + "epoch": 2.53, + "learning_rate": 1.1568303312651625e-05, + "loss": 0.9343, + "step": 39393 + }, + { + "epoch": 2.53, + "learning_rate": 1.1568088623628676e-05, + "loss": 0.8622, + "step": 39394 + }, + { + "epoch": 2.53, + "learning_rate": 1.1567873934605725e-05, + "loss": 0.8708, + "step": 39395 + }, + { + "epoch": 2.53, + "learning_rate": 1.1567659245582774e-05, + "loss": 0.8739, + "step": 39396 + }, + { + "epoch": 2.53, + "learning_rate": 1.1567444556559824e-05, + "loss": 0.9622, + "step": 39397 + }, + { + "epoch": 2.53, + "learning_rate": 1.1567229867536873e-05, + "loss": 0.9118, + "step": 39398 + }, + { + "epoch": 2.53, + "learning_rate": 1.1567015178513922e-05, + "loss": 0.8909, + "step": 39399 + }, + { + "epoch": 2.53, + "learning_rate": 1.1566800489490973e-05, + "loss": 0.897, + "step": 39400 + }, + { + "epoch": 2.53, + "learning_rate": 1.1566585800468022e-05, + "loss": 0.849, + "step": 39401 + }, + { + "epoch": 2.54, + "learning_rate": 1.1566371111445074e-05, + "loss": 0.9181, + "step": 39402 + }, + { + "epoch": 2.54, + "learning_rate": 1.1566156422422123e-05, + "loss": 0.9177, + "step": 39403 + }, + { + "epoch": 2.54, + "learning_rate": 1.1565941733399174e-05, + "loss": 0.8153, + "step": 39404 + }, + { + "epoch": 2.54, + "learning_rate": 1.1565727044376223e-05, + "loss": 0.8144, + "step": 39405 + }, + { + "epoch": 2.54, + "learning_rate": 1.1565512355353272e-05, + "loss": 0.8933, + "step": 39406 + }, + { + "epoch": 2.54, + "learning_rate": 1.1565297666330322e-05, + "loss": 0.9071, + "step": 39407 + }, + { + "epoch": 2.54, + "learning_rate": 1.1565082977307371e-05, + "loss": 0.9218, + "step": 39408 + }, + { + "epoch": 2.54, + "learning_rate": 1.156486828828442e-05, + "loss": 0.9277, + "step": 39409 + }, + { + "epoch": 2.54, + "learning_rate": 1.1564653599261471e-05, + "loss": 0.8336, + "step": 39410 + }, + { + "epoch": 2.54, + "learning_rate": 1.156443891023852e-05, + "loss": 0.9098, + "step": 39411 + }, + { + "epoch": 2.54, + "learning_rate": 1.156422422121557e-05, + "loss": 0.9146, + "step": 39412 + }, + { + "epoch": 2.54, + "learning_rate": 1.156400953219262e-05, + "loss": 0.9203, + "step": 39413 + }, + { + "epoch": 2.54, + "learning_rate": 1.1563794843169668e-05, + "loss": 0.8428, + "step": 39414 + }, + { + "epoch": 2.54, + "learning_rate": 1.156358015414672e-05, + "loss": 0.9029, + "step": 39415 + }, + { + "epoch": 2.54, + "learning_rate": 1.1563365465123768e-05, + "loss": 0.923, + "step": 39416 + }, + { + "epoch": 2.54, + "learning_rate": 1.156315077610082e-05, + "loss": 0.9212, + "step": 39417 + }, + { + "epoch": 2.54, + "learning_rate": 1.156293608707787e-05, + "loss": 0.9408, + "step": 39418 + }, + { + "epoch": 2.54, + "learning_rate": 1.1562721398054918e-05, + "loss": 0.9182, + "step": 39419 + }, + { + "epoch": 2.54, + "learning_rate": 1.1562506709031969e-05, + "loss": 0.905, + "step": 39420 + }, + { + "epoch": 2.54, + "learning_rate": 1.1562292020009018e-05, + "loss": 0.9106, + "step": 39421 + }, + { + "epoch": 2.54, + "learning_rate": 1.1562077330986069e-05, + "loss": 0.8973, + "step": 39422 + }, + { + "epoch": 2.54, + "learning_rate": 1.1561862641963118e-05, + "loss": 1.0321, + "step": 39423 + }, + { + "epoch": 2.54, + "learning_rate": 1.1561647952940167e-05, + "loss": 0.9058, + "step": 39424 + }, + { + "epoch": 2.54, + "learning_rate": 1.1561433263917217e-05, + "loss": 0.9292, + "step": 39425 + }, + { + "epoch": 2.54, + "learning_rate": 1.1561218574894266e-05, + "loss": 0.8699, + "step": 39426 + }, + { + "epoch": 2.54, + "learning_rate": 1.1561003885871315e-05, + "loss": 0.9635, + "step": 39427 + }, + { + "epoch": 2.54, + "learning_rate": 1.1560789196848366e-05, + "loss": 0.8742, + "step": 39428 + }, + { + "epoch": 2.54, + "learning_rate": 1.1560574507825415e-05, + "loss": 0.9103, + "step": 39429 + }, + { + "epoch": 2.54, + "learning_rate": 1.1560359818802464e-05, + "loss": 0.9078, + "step": 39430 + }, + { + "epoch": 2.54, + "learning_rate": 1.1560145129779516e-05, + "loss": 0.8391, + "step": 39431 + }, + { + "epoch": 2.54, + "learning_rate": 1.1559930440756567e-05, + "loss": 0.9519, + "step": 39432 + }, + { + "epoch": 2.54, + "learning_rate": 1.1559715751733616e-05, + "loss": 0.8805, + "step": 39433 + }, + { + "epoch": 2.54, + "learning_rate": 1.1559501062710665e-05, + "loss": 0.8158, + "step": 39434 + }, + { + "epoch": 2.54, + "learning_rate": 1.1559286373687715e-05, + "loss": 0.8985, + "step": 39435 + }, + { + "epoch": 2.54, + "learning_rate": 1.1559071684664764e-05, + "loss": 0.8916, + "step": 39436 + }, + { + "epoch": 2.54, + "learning_rate": 1.1558856995641813e-05, + "loss": 0.8662, + "step": 39437 + }, + { + "epoch": 2.54, + "learning_rate": 1.1558642306618864e-05, + "loss": 0.8863, + "step": 39438 + }, + { + "epoch": 2.54, + "learning_rate": 1.1558427617595913e-05, + "loss": 0.8785, + "step": 39439 + }, + { + "epoch": 2.54, + "learning_rate": 1.1558212928572962e-05, + "loss": 0.9234, + "step": 39440 + }, + { + "epoch": 2.54, + "learning_rate": 1.1557998239550013e-05, + "loss": 0.8891, + "step": 39441 + }, + { + "epoch": 2.54, + "learning_rate": 1.1557783550527062e-05, + "loss": 0.8591, + "step": 39442 + }, + { + "epoch": 2.54, + "learning_rate": 1.1557568861504112e-05, + "loss": 0.8861, + "step": 39443 + }, + { + "epoch": 2.54, + "learning_rate": 1.1557354172481161e-05, + "loss": 0.9458, + "step": 39444 + }, + { + "epoch": 2.54, + "learning_rate": 1.155713948345821e-05, + "loss": 0.8977, + "step": 39445 + }, + { + "epoch": 2.54, + "learning_rate": 1.1556924794435262e-05, + "loss": 0.8475, + "step": 39446 + }, + { + "epoch": 2.54, + "learning_rate": 1.1556710105412311e-05, + "loss": 0.8591, + "step": 39447 + }, + { + "epoch": 2.54, + "learning_rate": 1.1556495416389362e-05, + "loss": 0.8634, + "step": 39448 + }, + { + "epoch": 2.54, + "learning_rate": 1.1556280727366411e-05, + "loss": 0.8864, + "step": 39449 + }, + { + "epoch": 2.54, + "learning_rate": 1.155606603834346e-05, + "loss": 0.88, + "step": 39450 + }, + { + "epoch": 2.54, + "learning_rate": 1.155585134932051e-05, + "loss": 0.9175, + "step": 39451 + }, + { + "epoch": 2.54, + "learning_rate": 1.155563666029756e-05, + "loss": 0.9344, + "step": 39452 + }, + { + "epoch": 2.54, + "learning_rate": 1.155542197127461e-05, + "loss": 0.9233, + "step": 39453 + }, + { + "epoch": 2.54, + "learning_rate": 1.155520728225166e-05, + "loss": 0.9022, + "step": 39454 + }, + { + "epoch": 2.54, + "learning_rate": 1.1554992593228708e-05, + "loss": 0.8902, + "step": 39455 + }, + { + "epoch": 2.54, + "learning_rate": 1.1554777904205759e-05, + "loss": 0.8875, + "step": 39456 + }, + { + "epoch": 2.54, + "learning_rate": 1.1554563215182808e-05, + "loss": 0.8591, + "step": 39457 + }, + { + "epoch": 2.54, + "learning_rate": 1.1554348526159857e-05, + "loss": 0.9471, + "step": 39458 + }, + { + "epoch": 2.54, + "learning_rate": 1.1554133837136907e-05, + "loss": 0.9747, + "step": 39459 + }, + { + "epoch": 2.54, + "learning_rate": 1.1553919148113958e-05, + "loss": 0.9163, + "step": 39460 + }, + { + "epoch": 2.54, + "learning_rate": 1.1553704459091009e-05, + "loss": 0.8384, + "step": 39461 + }, + { + "epoch": 2.54, + "learning_rate": 1.1553489770068058e-05, + "loss": 0.8673, + "step": 39462 + }, + { + "epoch": 2.54, + "learning_rate": 1.1553275081045108e-05, + "loss": 0.9045, + "step": 39463 + }, + { + "epoch": 2.54, + "learning_rate": 1.1553060392022157e-05, + "loss": 0.9825, + "step": 39464 + }, + { + "epoch": 2.54, + "learning_rate": 1.1552845702999206e-05, + "loss": 0.8843, + "step": 39465 + }, + { + "epoch": 2.54, + "learning_rate": 1.1552631013976257e-05, + "loss": 0.9042, + "step": 39466 + }, + { + "epoch": 2.54, + "learning_rate": 1.1552416324953306e-05, + "loss": 0.8577, + "step": 39467 + }, + { + "epoch": 2.54, + "learning_rate": 1.1552201635930355e-05, + "loss": 0.91, + "step": 39468 + }, + { + "epoch": 2.54, + "learning_rate": 1.1551986946907406e-05, + "loss": 0.8991, + "step": 39469 + }, + { + "epoch": 2.54, + "learning_rate": 1.1551772257884455e-05, + "loss": 0.8287, + "step": 39470 + }, + { + "epoch": 2.54, + "learning_rate": 1.1551557568861505e-05, + "loss": 0.9625, + "step": 39471 + }, + { + "epoch": 2.54, + "learning_rate": 1.1551342879838554e-05, + "loss": 0.9861, + "step": 39472 + }, + { + "epoch": 2.54, + "learning_rate": 1.1551128190815603e-05, + "loss": 0.8702, + "step": 39473 + }, + { + "epoch": 2.54, + "learning_rate": 1.1550913501792654e-05, + "loss": 0.8501, + "step": 39474 + }, + { + "epoch": 2.54, + "learning_rate": 1.1550698812769704e-05, + "loss": 0.9647, + "step": 39475 + }, + { + "epoch": 2.54, + "learning_rate": 1.1550484123746755e-05, + "loss": 0.8853, + "step": 39476 + }, + { + "epoch": 2.54, + "learning_rate": 1.1550269434723804e-05, + "loss": 0.917, + "step": 39477 + }, + { + "epoch": 2.54, + "learning_rate": 1.1550054745700853e-05, + "loss": 0.8731, + "step": 39478 + }, + { + "epoch": 2.54, + "learning_rate": 1.1549840056677904e-05, + "loss": 0.8934, + "step": 39479 + }, + { + "epoch": 2.54, + "learning_rate": 1.1549625367654953e-05, + "loss": 0.9097, + "step": 39480 + }, + { + "epoch": 2.54, + "learning_rate": 1.1549410678632003e-05, + "loss": 0.9357, + "step": 39481 + }, + { + "epoch": 2.54, + "learning_rate": 1.1549195989609052e-05, + "loss": 0.9373, + "step": 39482 + }, + { + "epoch": 2.54, + "learning_rate": 1.1548981300586101e-05, + "loss": 0.9217, + "step": 39483 + }, + { + "epoch": 2.54, + "learning_rate": 1.1548766611563152e-05, + "loss": 0.9123, + "step": 39484 + }, + { + "epoch": 2.54, + "learning_rate": 1.1548551922540201e-05, + "loss": 0.9493, + "step": 39485 + }, + { + "epoch": 2.54, + "learning_rate": 1.154833723351725e-05, + "loss": 0.9351, + "step": 39486 + }, + { + "epoch": 2.54, + "learning_rate": 1.15481225444943e-05, + "loss": 0.8411, + "step": 39487 + }, + { + "epoch": 2.54, + "learning_rate": 1.154790785547135e-05, + "loss": 0.8427, + "step": 39488 + }, + { + "epoch": 2.54, + "learning_rate": 1.1547693166448402e-05, + "loss": 0.9578, + "step": 39489 + }, + { + "epoch": 2.54, + "learning_rate": 1.154747847742545e-05, + "loss": 0.8929, + "step": 39490 + }, + { + "epoch": 2.54, + "learning_rate": 1.1547263788402501e-05, + "loss": 0.83, + "step": 39491 + }, + { + "epoch": 2.54, + "learning_rate": 1.154704909937955e-05, + "loss": 0.8893, + "step": 39492 + }, + { + "epoch": 2.54, + "learning_rate": 1.15468344103566e-05, + "loss": 0.8526, + "step": 39493 + }, + { + "epoch": 2.54, + "learning_rate": 1.154661972133365e-05, + "loss": 0.9053, + "step": 39494 + }, + { + "epoch": 2.54, + "learning_rate": 1.1546405032310699e-05, + "loss": 0.9099, + "step": 39495 + }, + { + "epoch": 2.54, + "learning_rate": 1.1546190343287748e-05, + "loss": 0.948, + "step": 39496 + }, + { + "epoch": 2.54, + "learning_rate": 1.1545975654264799e-05, + "loss": 0.8219, + "step": 39497 + }, + { + "epoch": 2.54, + "learning_rate": 1.1545760965241848e-05, + "loss": 0.8666, + "step": 39498 + }, + { + "epoch": 2.54, + "learning_rate": 1.1545546276218898e-05, + "loss": 0.9714, + "step": 39499 + }, + { + "epoch": 2.54, + "learning_rate": 1.1545331587195947e-05, + "loss": 0.8987, + "step": 39500 + }, + { + "epoch": 2.54, + "learning_rate": 1.1545116898172996e-05, + "loss": 0.9216, + "step": 39501 + }, + { + "epoch": 2.54, + "learning_rate": 1.1544902209150047e-05, + "loss": 0.8798, + "step": 39502 + }, + { + "epoch": 2.54, + "learning_rate": 1.1544687520127096e-05, + "loss": 0.8739, + "step": 39503 + }, + { + "epoch": 2.54, + "learning_rate": 1.1544472831104148e-05, + "loss": 0.8906, + "step": 39504 + }, + { + "epoch": 2.54, + "learning_rate": 1.1544258142081197e-05, + "loss": 0.8271, + "step": 39505 + }, + { + "epoch": 2.54, + "learning_rate": 1.1544043453058246e-05, + "loss": 0.8678, + "step": 39506 + }, + { + "epoch": 2.54, + "learning_rate": 1.1543828764035297e-05, + "loss": 0.9117, + "step": 39507 + }, + { + "epoch": 2.54, + "learning_rate": 1.1543614075012346e-05, + "loss": 0.8973, + "step": 39508 + }, + { + "epoch": 2.54, + "learning_rate": 1.1543399385989396e-05, + "loss": 0.9003, + "step": 39509 + }, + { + "epoch": 2.54, + "learning_rate": 1.1543184696966445e-05, + "loss": 0.9253, + "step": 39510 + }, + { + "epoch": 2.54, + "learning_rate": 1.1542970007943494e-05, + "loss": 0.9483, + "step": 39511 + }, + { + "epoch": 2.54, + "learning_rate": 1.1542755318920545e-05, + "loss": 0.8931, + "step": 39512 + }, + { + "epoch": 2.54, + "learning_rate": 1.1542540629897594e-05, + "loss": 0.9353, + "step": 39513 + }, + { + "epoch": 2.54, + "learning_rate": 1.1542325940874643e-05, + "loss": 0.8756, + "step": 39514 + }, + { + "epoch": 2.54, + "learning_rate": 1.1542111251851694e-05, + "loss": 0.8295, + "step": 39515 + }, + { + "epoch": 2.54, + "learning_rate": 1.1541896562828743e-05, + "loss": 0.878, + "step": 39516 + }, + { + "epoch": 2.54, + "learning_rate": 1.1541681873805792e-05, + "loss": 0.8341, + "step": 39517 + }, + { + "epoch": 2.54, + "learning_rate": 1.1541467184782842e-05, + "loss": 0.9104, + "step": 39518 + }, + { + "epoch": 2.54, + "learning_rate": 1.1541252495759895e-05, + "loss": 0.9146, + "step": 39519 + }, + { + "epoch": 2.54, + "learning_rate": 1.1541037806736943e-05, + "loss": 0.8865, + "step": 39520 + }, + { + "epoch": 2.54, + "learning_rate": 1.1540823117713992e-05, + "loss": 0.9038, + "step": 39521 + }, + { + "epoch": 2.54, + "learning_rate": 1.1540608428691043e-05, + "loss": 0.8592, + "step": 39522 + }, + { + "epoch": 2.54, + "learning_rate": 1.1540393739668092e-05, + "loss": 0.8997, + "step": 39523 + }, + { + "epoch": 2.54, + "learning_rate": 1.1540179050645141e-05, + "loss": 0.9212, + "step": 39524 + }, + { + "epoch": 2.54, + "learning_rate": 1.1539964361622192e-05, + "loss": 0.8404, + "step": 39525 + }, + { + "epoch": 2.54, + "learning_rate": 1.153974967259924e-05, + "loss": 0.9109, + "step": 39526 + }, + { + "epoch": 2.54, + "learning_rate": 1.153953498357629e-05, + "loss": 0.8879, + "step": 39527 + }, + { + "epoch": 2.54, + "learning_rate": 1.153932029455334e-05, + "loss": 0.83, + "step": 39528 + }, + { + "epoch": 2.54, + "learning_rate": 1.153910560553039e-05, + "loss": 0.8845, + "step": 39529 + }, + { + "epoch": 2.54, + "learning_rate": 1.153889091650744e-05, + "loss": 0.9453, + "step": 39530 + }, + { + "epoch": 2.54, + "learning_rate": 1.1538676227484489e-05, + "loss": 0.9072, + "step": 39531 + }, + { + "epoch": 2.54, + "learning_rate": 1.1538461538461538e-05, + "loss": 0.9321, + "step": 39532 + }, + { + "epoch": 2.54, + "learning_rate": 1.153824684943859e-05, + "loss": 0.9481, + "step": 39533 + }, + { + "epoch": 2.54, + "learning_rate": 1.153803216041564e-05, + "loss": 0.966, + "step": 39534 + }, + { + "epoch": 2.54, + "learning_rate": 1.153781747139269e-05, + "loss": 0.9608, + "step": 39535 + }, + { + "epoch": 2.54, + "learning_rate": 1.1537602782369739e-05, + "loss": 0.9179, + "step": 39536 + }, + { + "epoch": 2.54, + "learning_rate": 1.1537388093346788e-05, + "loss": 0.8906, + "step": 39537 + }, + { + "epoch": 2.54, + "learning_rate": 1.1537173404323838e-05, + "loss": 0.888, + "step": 39538 + }, + { + "epoch": 2.54, + "learning_rate": 1.1536958715300887e-05, + "loss": 0.8653, + "step": 39539 + }, + { + "epoch": 2.54, + "learning_rate": 1.1536744026277938e-05, + "loss": 0.9069, + "step": 39540 + }, + { + "epoch": 2.54, + "learning_rate": 1.1536529337254987e-05, + "loss": 0.9175, + "step": 39541 + }, + { + "epoch": 2.54, + "learning_rate": 1.1536314648232036e-05, + "loss": 0.9219, + "step": 39542 + }, + { + "epoch": 2.54, + "learning_rate": 1.1536099959209087e-05, + "loss": 0.9282, + "step": 39543 + }, + { + "epoch": 2.54, + "learning_rate": 1.1535885270186136e-05, + "loss": 0.8529, + "step": 39544 + }, + { + "epoch": 2.54, + "learning_rate": 1.1535670581163185e-05, + "loss": 0.9231, + "step": 39545 + }, + { + "epoch": 2.54, + "learning_rate": 1.1535455892140235e-05, + "loss": 0.8892, + "step": 39546 + }, + { + "epoch": 2.54, + "learning_rate": 1.1535241203117284e-05, + "loss": 0.8619, + "step": 39547 + }, + { + "epoch": 2.54, + "learning_rate": 1.1535026514094337e-05, + "loss": 0.8817, + "step": 39548 + }, + { + "epoch": 2.54, + "learning_rate": 1.1534811825071385e-05, + "loss": 0.9001, + "step": 39549 + }, + { + "epoch": 2.54, + "learning_rate": 1.1534597136048436e-05, + "loss": 0.8486, + "step": 39550 + }, + { + "epoch": 2.54, + "learning_rate": 1.1534382447025485e-05, + "loss": 0.9165, + "step": 39551 + }, + { + "epoch": 2.54, + "learning_rate": 1.1534167758002534e-05, + "loss": 0.9011, + "step": 39552 + }, + { + "epoch": 2.54, + "learning_rate": 1.1533953068979585e-05, + "loss": 0.8915, + "step": 39553 + }, + { + "epoch": 2.54, + "learning_rate": 1.1533738379956634e-05, + "loss": 0.9126, + "step": 39554 + }, + { + "epoch": 2.54, + "learning_rate": 1.1533523690933683e-05, + "loss": 0.8325, + "step": 39555 + }, + { + "epoch": 2.54, + "learning_rate": 1.1533309001910733e-05, + "loss": 0.9078, + "step": 39556 + }, + { + "epoch": 2.55, + "learning_rate": 1.1533094312887782e-05, + "loss": 0.9209, + "step": 39557 + }, + { + "epoch": 2.55, + "learning_rate": 1.1532879623864833e-05, + "loss": 0.9116, + "step": 39558 + }, + { + "epoch": 2.55, + "learning_rate": 1.1532664934841882e-05, + "loss": 0.9333, + "step": 39559 + }, + { + "epoch": 2.55, + "learning_rate": 1.1532450245818931e-05, + "loss": 0.8775, + "step": 39560 + }, + { + "epoch": 2.55, + "learning_rate": 1.1532235556795982e-05, + "loss": 0.8897, + "step": 39561 + }, + { + "epoch": 2.55, + "learning_rate": 1.1532020867773032e-05, + "loss": 0.8701, + "step": 39562 + }, + { + "epoch": 2.55, + "learning_rate": 1.1531806178750083e-05, + "loss": 0.9434, + "step": 39563 + }, + { + "epoch": 2.55, + "learning_rate": 1.1531591489727132e-05, + "loss": 0.916, + "step": 39564 + }, + { + "epoch": 2.55, + "learning_rate": 1.153137680070418e-05, + "loss": 0.8793, + "step": 39565 + }, + { + "epoch": 2.55, + "learning_rate": 1.1531162111681231e-05, + "loss": 0.8898, + "step": 39566 + }, + { + "epoch": 2.55, + "learning_rate": 1.153094742265828e-05, + "loss": 0.9731, + "step": 39567 + }, + { + "epoch": 2.55, + "learning_rate": 1.1530732733635331e-05, + "loss": 0.8795, + "step": 39568 + }, + { + "epoch": 2.55, + "learning_rate": 1.153051804461238e-05, + "loss": 0.87, + "step": 39569 + }, + { + "epoch": 2.55, + "learning_rate": 1.1530303355589429e-05, + "loss": 0.9208, + "step": 39570 + }, + { + "epoch": 2.55, + "learning_rate": 1.153008866656648e-05, + "loss": 0.8405, + "step": 39571 + }, + { + "epoch": 2.55, + "learning_rate": 1.1529873977543529e-05, + "loss": 0.8568, + "step": 39572 + }, + { + "epoch": 2.55, + "learning_rate": 1.1529659288520578e-05, + "loss": 0.9819, + "step": 39573 + }, + { + "epoch": 2.55, + "learning_rate": 1.1529444599497628e-05, + "loss": 0.8998, + "step": 39574 + }, + { + "epoch": 2.55, + "learning_rate": 1.1529229910474677e-05, + "loss": 0.9486, + "step": 39575 + }, + { + "epoch": 2.55, + "learning_rate": 1.1529015221451726e-05, + "loss": 0.91, + "step": 39576 + }, + { + "epoch": 2.55, + "learning_rate": 1.1528800532428779e-05, + "loss": 0.9181, + "step": 39577 + }, + { + "epoch": 2.55, + "learning_rate": 1.152858584340583e-05, + "loss": 0.9377, + "step": 39578 + }, + { + "epoch": 2.55, + "learning_rate": 1.1528371154382878e-05, + "loss": 0.918, + "step": 39579 + }, + { + "epoch": 2.55, + "learning_rate": 1.1528156465359927e-05, + "loss": 0.9252, + "step": 39580 + }, + { + "epoch": 2.55, + "learning_rate": 1.1527941776336978e-05, + "loss": 0.9213, + "step": 39581 + }, + { + "epoch": 2.55, + "learning_rate": 1.1527727087314027e-05, + "loss": 0.8802, + "step": 39582 + }, + { + "epoch": 2.55, + "learning_rate": 1.1527512398291076e-05, + "loss": 1.0088, + "step": 39583 + }, + { + "epoch": 2.55, + "learning_rate": 1.1527297709268126e-05, + "loss": 0.8521, + "step": 39584 + }, + { + "epoch": 2.55, + "learning_rate": 1.1527083020245175e-05, + "loss": 0.8785, + "step": 39585 + }, + { + "epoch": 2.55, + "learning_rate": 1.1526868331222226e-05, + "loss": 0.8727, + "step": 39586 + }, + { + "epoch": 2.55, + "learning_rate": 1.1526653642199275e-05, + "loss": 0.9194, + "step": 39587 + }, + { + "epoch": 2.55, + "learning_rate": 1.1526438953176324e-05, + "loss": 0.8402, + "step": 39588 + }, + { + "epoch": 2.55, + "learning_rate": 1.1526224264153375e-05, + "loss": 0.8493, + "step": 39589 + }, + { + "epoch": 2.55, + "learning_rate": 1.1526009575130424e-05, + "loss": 0.8719, + "step": 39590 + }, + { + "epoch": 2.55, + "learning_rate": 1.1525794886107476e-05, + "loss": 0.8822, + "step": 39591 + }, + { + "epoch": 2.55, + "learning_rate": 1.1525580197084525e-05, + "loss": 0.976, + "step": 39592 + }, + { + "epoch": 2.55, + "learning_rate": 1.1525365508061574e-05, + "loss": 0.8907, + "step": 39593 + }, + { + "epoch": 2.55, + "learning_rate": 1.1525150819038624e-05, + "loss": 0.8817, + "step": 39594 + }, + { + "epoch": 2.55, + "learning_rate": 1.1524936130015673e-05, + "loss": 0.8796, + "step": 39595 + }, + { + "epoch": 2.55, + "learning_rate": 1.1524721440992724e-05, + "loss": 0.8781, + "step": 39596 + }, + { + "epoch": 2.55, + "learning_rate": 1.1524506751969773e-05, + "loss": 0.9099, + "step": 39597 + }, + { + "epoch": 2.55, + "learning_rate": 1.1524292062946822e-05, + "loss": 0.919, + "step": 39598 + }, + { + "epoch": 2.55, + "learning_rate": 1.1524077373923873e-05, + "loss": 0.9549, + "step": 39599 + }, + { + "epoch": 2.55, + "learning_rate": 1.1523862684900922e-05, + "loss": 0.9428, + "step": 39600 + }, + { + "epoch": 2.55, + "learning_rate": 1.152364799587797e-05, + "loss": 0.8732, + "step": 39601 + }, + { + "epoch": 2.55, + "learning_rate": 1.1523433306855021e-05, + "loss": 0.926, + "step": 39602 + }, + { + "epoch": 2.55, + "learning_rate": 1.152321861783207e-05, + "loss": 0.9181, + "step": 39603 + }, + { + "epoch": 2.55, + "learning_rate": 1.152300392880912e-05, + "loss": 0.939, + "step": 39604 + }, + { + "epoch": 2.55, + "learning_rate": 1.152278923978617e-05, + "loss": 0.878, + "step": 39605 + }, + { + "epoch": 2.55, + "learning_rate": 1.1522574550763222e-05, + "loss": 0.8726, + "step": 39606 + }, + { + "epoch": 2.55, + "learning_rate": 1.1522359861740271e-05, + "loss": 0.931, + "step": 39607 + }, + { + "epoch": 2.55, + "learning_rate": 1.152214517271732e-05, + "loss": 0.9049, + "step": 39608 + }, + { + "epoch": 2.55, + "learning_rate": 1.152193048369437e-05, + "loss": 0.9455, + "step": 39609 + }, + { + "epoch": 2.55, + "learning_rate": 1.152171579467142e-05, + "loss": 0.9942, + "step": 39610 + }, + { + "epoch": 2.55, + "learning_rate": 1.1521501105648469e-05, + "loss": 0.8824, + "step": 39611 + }, + { + "epoch": 2.55, + "learning_rate": 1.152128641662552e-05, + "loss": 0.9459, + "step": 39612 + }, + { + "epoch": 2.55, + "learning_rate": 1.1521071727602568e-05, + "loss": 0.9102, + "step": 39613 + }, + { + "epoch": 2.55, + "learning_rate": 1.1520857038579617e-05, + "loss": 0.9215, + "step": 39614 + }, + { + "epoch": 2.55, + "learning_rate": 1.1520642349556668e-05, + "loss": 0.9233, + "step": 39615 + }, + { + "epoch": 2.55, + "learning_rate": 1.1520427660533717e-05, + "loss": 0.9026, + "step": 39616 + }, + { + "epoch": 2.55, + "learning_rate": 1.1520212971510768e-05, + "loss": 0.8783, + "step": 39617 + }, + { + "epoch": 2.55, + "learning_rate": 1.1519998282487817e-05, + "loss": 0.8798, + "step": 39618 + }, + { + "epoch": 2.55, + "learning_rate": 1.1519783593464866e-05, + "loss": 0.8834, + "step": 39619 + }, + { + "epoch": 2.55, + "learning_rate": 1.1519568904441918e-05, + "loss": 0.8806, + "step": 39620 + }, + { + "epoch": 2.55, + "learning_rate": 1.1519354215418967e-05, + "loss": 0.8727, + "step": 39621 + }, + { + "epoch": 2.55, + "learning_rate": 1.1519139526396018e-05, + "loss": 0.921, + "step": 39622 + }, + { + "epoch": 2.55, + "learning_rate": 1.1518924837373066e-05, + "loss": 0.8958, + "step": 39623 + }, + { + "epoch": 2.55, + "learning_rate": 1.1518710148350115e-05, + "loss": 0.8804, + "step": 39624 + }, + { + "epoch": 2.55, + "learning_rate": 1.1518495459327166e-05, + "loss": 0.9282, + "step": 39625 + }, + { + "epoch": 2.55, + "learning_rate": 1.1518280770304215e-05, + "loss": 0.8527, + "step": 39626 + }, + { + "epoch": 2.55, + "learning_rate": 1.1518066081281266e-05, + "loss": 0.8888, + "step": 39627 + }, + { + "epoch": 2.55, + "learning_rate": 1.1517851392258315e-05, + "loss": 0.8479, + "step": 39628 + }, + { + "epoch": 2.55, + "learning_rate": 1.1517636703235364e-05, + "loss": 0.8596, + "step": 39629 + }, + { + "epoch": 2.55, + "learning_rate": 1.1517422014212414e-05, + "loss": 0.8421, + "step": 39630 + }, + { + "epoch": 2.55, + "learning_rate": 1.1517207325189463e-05, + "loss": 0.8809, + "step": 39631 + }, + { + "epoch": 2.55, + "learning_rate": 1.1516992636166512e-05, + "loss": 0.8617, + "step": 39632 + }, + { + "epoch": 2.55, + "learning_rate": 1.1516777947143563e-05, + "loss": 0.9392, + "step": 39633 + }, + { + "epoch": 2.55, + "learning_rate": 1.1516563258120612e-05, + "loss": 0.9305, + "step": 39634 + }, + { + "epoch": 2.55, + "learning_rate": 1.1516348569097664e-05, + "loss": 0.9161, + "step": 39635 + }, + { + "epoch": 2.55, + "learning_rate": 1.1516133880074713e-05, + "loss": 0.9323, + "step": 39636 + }, + { + "epoch": 2.55, + "learning_rate": 1.1515919191051764e-05, + "loss": 0.8825, + "step": 39637 + }, + { + "epoch": 2.55, + "learning_rate": 1.1515704502028813e-05, + "loss": 0.9041, + "step": 39638 + }, + { + "epoch": 2.55, + "learning_rate": 1.1515489813005862e-05, + "loss": 0.8839, + "step": 39639 + }, + { + "epoch": 2.55, + "learning_rate": 1.1515275123982912e-05, + "loss": 0.8783, + "step": 39640 + }, + { + "epoch": 2.55, + "learning_rate": 1.1515060434959961e-05, + "loss": 0.923, + "step": 39641 + }, + { + "epoch": 2.55, + "learning_rate": 1.151484574593701e-05, + "loss": 0.9111, + "step": 39642 + }, + { + "epoch": 2.55, + "learning_rate": 1.1514631056914061e-05, + "loss": 0.8959, + "step": 39643 + }, + { + "epoch": 2.55, + "learning_rate": 1.151441636789111e-05, + "loss": 0.8912, + "step": 39644 + }, + { + "epoch": 2.55, + "learning_rate": 1.151420167886816e-05, + "loss": 0.8725, + "step": 39645 + }, + { + "epoch": 2.55, + "learning_rate": 1.151398698984521e-05, + "loss": 0.813, + "step": 39646 + }, + { + "epoch": 2.55, + "learning_rate": 1.1513772300822259e-05, + "loss": 0.9473, + "step": 39647 + }, + { + "epoch": 2.55, + "learning_rate": 1.151355761179931e-05, + "loss": 0.8449, + "step": 39648 + }, + { + "epoch": 2.55, + "learning_rate": 1.151334292277636e-05, + "loss": 0.8445, + "step": 39649 + }, + { + "epoch": 2.55, + "learning_rate": 1.151312823375341e-05, + "loss": 0.9264, + "step": 39650 + }, + { + "epoch": 2.55, + "learning_rate": 1.151291354473046e-05, + "loss": 0.8578, + "step": 39651 + }, + { + "epoch": 2.55, + "learning_rate": 1.1512698855707509e-05, + "loss": 0.8826, + "step": 39652 + }, + { + "epoch": 2.55, + "learning_rate": 1.151248416668456e-05, + "loss": 0.8898, + "step": 39653 + }, + { + "epoch": 2.55, + "learning_rate": 1.1512269477661608e-05, + "loss": 0.9154, + "step": 39654 + }, + { + "epoch": 2.55, + "learning_rate": 1.1512054788638659e-05, + "loss": 0.9017, + "step": 39655 + }, + { + "epoch": 2.55, + "learning_rate": 1.1511840099615708e-05, + "loss": 0.9809, + "step": 39656 + }, + { + "epoch": 2.55, + "learning_rate": 1.1511625410592757e-05, + "loss": 0.9396, + "step": 39657 + }, + { + "epoch": 2.55, + "learning_rate": 1.1511410721569807e-05, + "loss": 0.8311, + "step": 39658 + }, + { + "epoch": 2.55, + "learning_rate": 1.1511196032546856e-05, + "loss": 0.9085, + "step": 39659 + }, + { + "epoch": 2.55, + "learning_rate": 1.1510981343523905e-05, + "loss": 0.8972, + "step": 39660 + }, + { + "epoch": 2.55, + "learning_rate": 1.1510766654500956e-05, + "loss": 0.8862, + "step": 39661 + }, + { + "epoch": 2.55, + "learning_rate": 1.1510551965478005e-05, + "loss": 0.8461, + "step": 39662 + }, + { + "epoch": 2.55, + "learning_rate": 1.1510337276455054e-05, + "loss": 0.8863, + "step": 39663 + }, + { + "epoch": 2.55, + "learning_rate": 1.1510122587432106e-05, + "loss": 0.9036, + "step": 39664 + }, + { + "epoch": 2.55, + "learning_rate": 1.1509907898409157e-05, + "loss": 0.9012, + "step": 39665 + }, + { + "epoch": 2.55, + "learning_rate": 1.1509693209386206e-05, + "loss": 0.8539, + "step": 39666 + }, + { + "epoch": 2.55, + "learning_rate": 1.1509478520363255e-05, + "loss": 0.8725, + "step": 39667 + }, + { + "epoch": 2.55, + "learning_rate": 1.1509263831340305e-05, + "loss": 0.8235, + "step": 39668 + }, + { + "epoch": 2.55, + "learning_rate": 1.1509049142317354e-05, + "loss": 0.8728, + "step": 39669 + }, + { + "epoch": 2.55, + "learning_rate": 1.1508834453294403e-05, + "loss": 1.0006, + "step": 39670 + }, + { + "epoch": 2.55, + "learning_rate": 1.1508619764271454e-05, + "loss": 0.9095, + "step": 39671 + }, + { + "epoch": 2.55, + "learning_rate": 1.1508405075248503e-05, + "loss": 0.8706, + "step": 39672 + }, + { + "epoch": 2.55, + "learning_rate": 1.1508190386225552e-05, + "loss": 0.8041, + "step": 39673 + }, + { + "epoch": 2.55, + "learning_rate": 1.1507975697202603e-05, + "loss": 0.9394, + "step": 39674 + }, + { + "epoch": 2.55, + "learning_rate": 1.1507761008179652e-05, + "loss": 0.8881, + "step": 39675 + }, + { + "epoch": 2.55, + "learning_rate": 1.1507546319156702e-05, + "loss": 0.8377, + "step": 39676 + }, + { + "epoch": 2.55, + "learning_rate": 1.1507331630133751e-05, + "loss": 0.8278, + "step": 39677 + }, + { + "epoch": 2.55, + "learning_rate": 1.1507116941110804e-05, + "loss": 0.9346, + "step": 39678 + }, + { + "epoch": 2.55, + "learning_rate": 1.1506902252087853e-05, + "loss": 0.8869, + "step": 39679 + }, + { + "epoch": 2.55, + "learning_rate": 1.1506687563064902e-05, + "loss": 0.9022, + "step": 39680 + }, + { + "epoch": 2.55, + "learning_rate": 1.1506472874041952e-05, + "loss": 0.8805, + "step": 39681 + }, + { + "epoch": 2.55, + "learning_rate": 1.1506258185019001e-05, + "loss": 0.9079, + "step": 39682 + }, + { + "epoch": 2.55, + "learning_rate": 1.150604349599605e-05, + "loss": 0.915, + "step": 39683 + }, + { + "epoch": 2.55, + "learning_rate": 1.15058288069731e-05, + "loss": 0.9195, + "step": 39684 + }, + { + "epoch": 2.55, + "learning_rate": 1.150561411795015e-05, + "loss": 0.958, + "step": 39685 + }, + { + "epoch": 2.55, + "learning_rate": 1.15053994289272e-05, + "loss": 0.85, + "step": 39686 + }, + { + "epoch": 2.55, + "learning_rate": 1.150518473990425e-05, + "loss": 0.896, + "step": 39687 + }, + { + "epoch": 2.55, + "learning_rate": 1.1504970050881298e-05, + "loss": 0.931, + "step": 39688 + }, + { + "epoch": 2.55, + "learning_rate": 1.1504755361858349e-05, + "loss": 0.8867, + "step": 39689 + }, + { + "epoch": 2.55, + "learning_rate": 1.1504540672835398e-05, + "loss": 0.9033, + "step": 39690 + }, + { + "epoch": 2.55, + "learning_rate": 1.1504325983812447e-05, + "loss": 0.8847, + "step": 39691 + }, + { + "epoch": 2.55, + "learning_rate": 1.1504111294789498e-05, + "loss": 0.898, + "step": 39692 + }, + { + "epoch": 2.55, + "learning_rate": 1.1503896605766548e-05, + "loss": 0.9161, + "step": 39693 + }, + { + "epoch": 2.55, + "learning_rate": 1.1503681916743599e-05, + "loss": 0.8858, + "step": 39694 + }, + { + "epoch": 2.55, + "learning_rate": 1.1503467227720648e-05, + "loss": 0.8711, + "step": 39695 + }, + { + "epoch": 2.55, + "learning_rate": 1.1503252538697699e-05, + "loss": 0.8804, + "step": 39696 + }, + { + "epoch": 2.55, + "learning_rate": 1.1503037849674747e-05, + "loss": 0.8862, + "step": 39697 + }, + { + "epoch": 2.55, + "learning_rate": 1.1502823160651796e-05, + "loss": 0.9052, + "step": 39698 + }, + { + "epoch": 2.55, + "learning_rate": 1.1502608471628847e-05, + "loss": 0.8214, + "step": 39699 + }, + { + "epoch": 2.55, + "learning_rate": 1.1502393782605896e-05, + "loss": 0.915, + "step": 39700 + }, + { + "epoch": 2.55, + "learning_rate": 1.1502179093582945e-05, + "loss": 0.8808, + "step": 39701 + }, + { + "epoch": 2.55, + "learning_rate": 1.1501964404559996e-05, + "loss": 0.9349, + "step": 39702 + }, + { + "epoch": 2.55, + "learning_rate": 1.1501749715537045e-05, + "loss": 0.9088, + "step": 39703 + }, + { + "epoch": 2.55, + "learning_rate": 1.1501535026514095e-05, + "loss": 0.8837, + "step": 39704 + }, + { + "epoch": 2.55, + "learning_rate": 1.1501320337491144e-05, + "loss": 0.8947, + "step": 39705 + }, + { + "epoch": 2.55, + "learning_rate": 1.1501105648468193e-05, + "loss": 0.9212, + "step": 39706 + }, + { + "epoch": 2.55, + "learning_rate": 1.1500890959445246e-05, + "loss": 0.8915, + "step": 39707 + }, + { + "epoch": 2.55, + "learning_rate": 1.1500676270422295e-05, + "loss": 0.8802, + "step": 39708 + }, + { + "epoch": 2.55, + "learning_rate": 1.1500461581399345e-05, + "loss": 0.9141, + "step": 39709 + }, + { + "epoch": 2.55, + "learning_rate": 1.1500246892376394e-05, + "loss": 0.899, + "step": 39710 + }, + { + "epoch": 2.55, + "learning_rate": 1.1500032203353443e-05, + "loss": 0.9049, + "step": 39711 + }, + { + "epoch": 2.55, + "learning_rate": 1.1499817514330494e-05, + "loss": 0.8609, + "step": 39712 + }, + { + "epoch": 2.56, + "learning_rate": 1.1499602825307543e-05, + "loss": 0.9277, + "step": 39713 + }, + { + "epoch": 2.56, + "learning_rate": 1.1499388136284593e-05, + "loss": 0.9337, + "step": 39714 + }, + { + "epoch": 2.56, + "learning_rate": 1.1499173447261642e-05, + "loss": 0.8589, + "step": 39715 + }, + { + "epoch": 2.56, + "learning_rate": 1.1498958758238691e-05, + "loss": 0.9788, + "step": 39716 + }, + { + "epoch": 2.56, + "learning_rate": 1.1498744069215742e-05, + "loss": 0.9056, + "step": 39717 + }, + { + "epoch": 2.56, + "learning_rate": 1.1498529380192791e-05, + "loss": 0.8864, + "step": 39718 + }, + { + "epoch": 2.56, + "learning_rate": 1.149831469116984e-05, + "loss": 0.8913, + "step": 39719 + }, + { + "epoch": 2.56, + "learning_rate": 1.149810000214689e-05, + "loss": 0.8936, + "step": 39720 + }, + { + "epoch": 2.56, + "learning_rate": 1.149788531312394e-05, + "loss": 0.8886, + "step": 39721 + }, + { + "epoch": 2.56, + "learning_rate": 1.1497670624100992e-05, + "loss": 0.8932, + "step": 39722 + }, + { + "epoch": 2.56, + "learning_rate": 1.1497455935078041e-05, + "loss": 0.8313, + "step": 39723 + }, + { + "epoch": 2.56, + "learning_rate": 1.1497241246055092e-05, + "loss": 0.9223, + "step": 39724 + }, + { + "epoch": 2.56, + "learning_rate": 1.149702655703214e-05, + "loss": 0.9114, + "step": 39725 + }, + { + "epoch": 2.56, + "learning_rate": 1.149681186800919e-05, + "loss": 0.9365, + "step": 39726 + }, + { + "epoch": 2.56, + "learning_rate": 1.149659717898624e-05, + "loss": 0.8414, + "step": 39727 + }, + { + "epoch": 2.56, + "learning_rate": 1.1496382489963289e-05, + "loss": 0.8956, + "step": 39728 + }, + { + "epoch": 2.56, + "learning_rate": 1.1496167800940338e-05, + "loss": 0.8718, + "step": 39729 + }, + { + "epoch": 2.56, + "learning_rate": 1.1495953111917389e-05, + "loss": 0.9519, + "step": 39730 + }, + { + "epoch": 2.56, + "learning_rate": 1.1495738422894438e-05, + "loss": 0.8699, + "step": 39731 + }, + { + "epoch": 2.56, + "learning_rate": 1.1495523733871488e-05, + "loss": 0.9161, + "step": 39732 + }, + { + "epoch": 2.56, + "learning_rate": 1.1495309044848537e-05, + "loss": 0.9297, + "step": 39733 + }, + { + "epoch": 2.56, + "learning_rate": 1.1495094355825586e-05, + "loss": 0.9249, + "step": 39734 + }, + { + "epoch": 2.56, + "learning_rate": 1.1494879666802637e-05, + "loss": 0.9102, + "step": 39735 + }, + { + "epoch": 2.56, + "learning_rate": 1.1494664977779688e-05, + "loss": 0.8811, + "step": 39736 + }, + { + "epoch": 2.56, + "learning_rate": 1.1494450288756738e-05, + "loss": 0.8865, + "step": 39737 + }, + { + "epoch": 2.56, + "learning_rate": 1.1494235599733787e-05, + "loss": 0.9187, + "step": 39738 + }, + { + "epoch": 2.56, + "learning_rate": 1.1494020910710836e-05, + "loss": 0.8813, + "step": 39739 + }, + { + "epoch": 2.56, + "learning_rate": 1.1493806221687887e-05, + "loss": 0.9515, + "step": 39740 + }, + { + "epoch": 2.56, + "learning_rate": 1.1493591532664936e-05, + "loss": 0.9338, + "step": 39741 + }, + { + "epoch": 2.56, + "learning_rate": 1.1493376843641986e-05, + "loss": 0.8828, + "step": 39742 + }, + { + "epoch": 2.56, + "learning_rate": 1.1493162154619035e-05, + "loss": 0.8089, + "step": 39743 + }, + { + "epoch": 2.56, + "learning_rate": 1.1492947465596084e-05, + "loss": 0.9114, + "step": 39744 + }, + { + "epoch": 2.56, + "learning_rate": 1.1492732776573135e-05, + "loss": 0.8708, + "step": 39745 + }, + { + "epoch": 2.56, + "learning_rate": 1.1492518087550184e-05, + "loss": 0.8576, + "step": 39746 + }, + { + "epoch": 2.56, + "learning_rate": 1.1492303398527233e-05, + "loss": 0.9535, + "step": 39747 + }, + { + "epoch": 2.56, + "learning_rate": 1.1492088709504284e-05, + "loss": 0.973, + "step": 39748 + }, + { + "epoch": 2.56, + "learning_rate": 1.1491874020481333e-05, + "loss": 0.915, + "step": 39749 + }, + { + "epoch": 2.56, + "learning_rate": 1.1491659331458382e-05, + "loss": 0.8954, + "step": 39750 + }, + { + "epoch": 2.56, + "learning_rate": 1.1491444642435434e-05, + "loss": 0.8874, + "step": 39751 + }, + { + "epoch": 2.56, + "learning_rate": 1.1491229953412485e-05, + "loss": 0.9161, + "step": 39752 + }, + { + "epoch": 2.56, + "learning_rate": 1.1491015264389534e-05, + "loss": 0.8851, + "step": 39753 + }, + { + "epoch": 2.56, + "learning_rate": 1.1490800575366583e-05, + "loss": 0.8276, + "step": 39754 + }, + { + "epoch": 2.56, + "learning_rate": 1.1490585886343633e-05, + "loss": 0.9196, + "step": 39755 + }, + { + "epoch": 2.56, + "learning_rate": 1.1490371197320682e-05, + "loss": 0.9144, + "step": 39756 + }, + { + "epoch": 2.56, + "learning_rate": 1.1490156508297731e-05, + "loss": 0.8283, + "step": 39757 + }, + { + "epoch": 2.56, + "learning_rate": 1.1489941819274782e-05, + "loss": 0.8952, + "step": 39758 + }, + { + "epoch": 2.56, + "learning_rate": 1.148972713025183e-05, + "loss": 0.8495, + "step": 39759 + }, + { + "epoch": 2.56, + "learning_rate": 1.148951244122888e-05, + "loss": 0.8833, + "step": 39760 + }, + { + "epoch": 2.56, + "learning_rate": 1.148929775220593e-05, + "loss": 0.9247, + "step": 39761 + }, + { + "epoch": 2.56, + "learning_rate": 1.148908306318298e-05, + "loss": 0.9413, + "step": 39762 + }, + { + "epoch": 2.56, + "learning_rate": 1.148886837416003e-05, + "loss": 0.9026, + "step": 39763 + }, + { + "epoch": 2.56, + "learning_rate": 1.1488653685137079e-05, + "loss": 0.9182, + "step": 39764 + }, + { + "epoch": 2.56, + "learning_rate": 1.1488438996114131e-05, + "loss": 0.927, + "step": 39765 + }, + { + "epoch": 2.56, + "learning_rate": 1.148822430709118e-05, + "loss": 0.9103, + "step": 39766 + }, + { + "epoch": 2.56, + "learning_rate": 1.148800961806823e-05, + "loss": 0.8835, + "step": 39767 + }, + { + "epoch": 2.56, + "learning_rate": 1.148779492904528e-05, + "loss": 0.8662, + "step": 39768 + }, + { + "epoch": 2.56, + "learning_rate": 1.1487580240022329e-05, + "loss": 0.9566, + "step": 39769 + }, + { + "epoch": 2.56, + "learning_rate": 1.1487365550999378e-05, + "loss": 0.9482, + "step": 39770 + }, + { + "epoch": 2.56, + "learning_rate": 1.1487150861976429e-05, + "loss": 0.8803, + "step": 39771 + }, + { + "epoch": 2.56, + "learning_rate": 1.1486936172953477e-05, + "loss": 0.9167, + "step": 39772 + }, + { + "epoch": 2.56, + "learning_rate": 1.1486721483930528e-05, + "loss": 0.9076, + "step": 39773 + }, + { + "epoch": 2.56, + "learning_rate": 1.1486506794907577e-05, + "loss": 0.8336, + "step": 39774 + }, + { + "epoch": 2.56, + "learning_rate": 1.1486292105884626e-05, + "loss": 0.9253, + "step": 39775 + }, + { + "epoch": 2.56, + "learning_rate": 1.1486077416861677e-05, + "loss": 0.8799, + "step": 39776 + }, + { + "epoch": 2.56, + "learning_rate": 1.1485862727838726e-05, + "loss": 0.9401, + "step": 39777 + }, + { + "epoch": 2.56, + "learning_rate": 1.1485648038815775e-05, + "loss": 0.8771, + "step": 39778 + }, + { + "epoch": 2.56, + "learning_rate": 1.1485433349792825e-05, + "loss": 0.9023, + "step": 39779 + }, + { + "epoch": 2.56, + "learning_rate": 1.1485218660769876e-05, + "loss": 0.8472, + "step": 39780 + }, + { + "epoch": 2.56, + "learning_rate": 1.1485003971746927e-05, + "loss": 0.855, + "step": 39781 + }, + { + "epoch": 2.56, + "learning_rate": 1.1484789282723976e-05, + "loss": 0.9425, + "step": 39782 + }, + { + "epoch": 2.56, + "learning_rate": 1.1484574593701026e-05, + "loss": 0.9249, + "step": 39783 + }, + { + "epoch": 2.56, + "learning_rate": 1.1484359904678075e-05, + "loss": 0.9179, + "step": 39784 + }, + { + "epoch": 2.56, + "learning_rate": 1.1484145215655124e-05, + "loss": 0.8384, + "step": 39785 + }, + { + "epoch": 2.56, + "learning_rate": 1.1483930526632175e-05, + "loss": 0.9293, + "step": 39786 + }, + { + "epoch": 2.56, + "learning_rate": 1.1483715837609224e-05, + "loss": 0.902, + "step": 39787 + }, + { + "epoch": 2.56, + "learning_rate": 1.1483501148586273e-05, + "loss": 0.8751, + "step": 39788 + }, + { + "epoch": 2.56, + "learning_rate": 1.1483286459563323e-05, + "loss": 0.8836, + "step": 39789 + }, + { + "epoch": 2.56, + "learning_rate": 1.1483071770540372e-05, + "loss": 0.8525, + "step": 39790 + }, + { + "epoch": 2.56, + "learning_rate": 1.1482857081517423e-05, + "loss": 0.9057, + "step": 39791 + }, + { + "epoch": 2.56, + "learning_rate": 1.1482642392494472e-05, + "loss": 0.8776, + "step": 39792 + }, + { + "epoch": 2.56, + "learning_rate": 1.1482427703471521e-05, + "loss": 0.9156, + "step": 39793 + }, + { + "epoch": 2.56, + "learning_rate": 1.1482213014448573e-05, + "loss": 0.8688, + "step": 39794 + }, + { + "epoch": 2.56, + "learning_rate": 1.1481998325425622e-05, + "loss": 0.8484, + "step": 39795 + }, + { + "epoch": 2.56, + "learning_rate": 1.1481783636402673e-05, + "loss": 0.9257, + "step": 39796 + }, + { + "epoch": 2.56, + "learning_rate": 1.1481568947379722e-05, + "loss": 0.8923, + "step": 39797 + }, + { + "epoch": 2.56, + "learning_rate": 1.1481354258356771e-05, + "loss": 0.9089, + "step": 39798 + }, + { + "epoch": 2.56, + "learning_rate": 1.1481139569333822e-05, + "loss": 0.9266, + "step": 39799 + }, + { + "epoch": 2.56, + "learning_rate": 1.148092488031087e-05, + "loss": 0.8492, + "step": 39800 + }, + { + "epoch": 2.56, + "learning_rate": 1.1480710191287921e-05, + "loss": 0.8925, + "step": 39801 + }, + { + "epoch": 2.56, + "learning_rate": 1.148049550226497e-05, + "loss": 0.893, + "step": 39802 + }, + { + "epoch": 2.56, + "learning_rate": 1.1480280813242019e-05, + "loss": 0.8165, + "step": 39803 + }, + { + "epoch": 2.56, + "learning_rate": 1.148006612421907e-05, + "loss": 0.9327, + "step": 39804 + }, + { + "epoch": 2.56, + "learning_rate": 1.1479851435196119e-05, + "loss": 0.9234, + "step": 39805 + }, + { + "epoch": 2.56, + "learning_rate": 1.1479636746173168e-05, + "loss": 0.9044, + "step": 39806 + }, + { + "epoch": 2.56, + "learning_rate": 1.1479422057150218e-05, + "loss": 0.8501, + "step": 39807 + }, + { + "epoch": 2.56, + "learning_rate": 1.1479207368127267e-05, + "loss": 1.0079, + "step": 39808 + }, + { + "epoch": 2.56, + "learning_rate": 1.147899267910432e-05, + "loss": 0.8384, + "step": 39809 + }, + { + "epoch": 2.56, + "learning_rate": 1.1478777990081369e-05, + "loss": 0.8932, + "step": 39810 + }, + { + "epoch": 2.56, + "learning_rate": 1.147856330105842e-05, + "loss": 0.8841, + "step": 39811 + }, + { + "epoch": 2.56, + "learning_rate": 1.1478348612035468e-05, + "loss": 1.0017, + "step": 39812 + }, + { + "epoch": 2.56, + "learning_rate": 1.1478133923012517e-05, + "loss": 0.8786, + "step": 39813 + }, + { + "epoch": 2.56, + "learning_rate": 1.1477919233989568e-05, + "loss": 0.8947, + "step": 39814 + }, + { + "epoch": 2.56, + "learning_rate": 1.1477704544966617e-05, + "loss": 0.894, + "step": 39815 + }, + { + "epoch": 2.56, + "learning_rate": 1.1477489855943666e-05, + "loss": 0.8471, + "step": 39816 + }, + { + "epoch": 2.56, + "learning_rate": 1.1477275166920716e-05, + "loss": 0.8701, + "step": 39817 + }, + { + "epoch": 2.56, + "learning_rate": 1.1477060477897765e-05, + "loss": 0.9356, + "step": 39818 + }, + { + "epoch": 2.56, + "learning_rate": 1.1476845788874816e-05, + "loss": 0.9455, + "step": 39819 + }, + { + "epoch": 2.56, + "learning_rate": 1.1476631099851865e-05, + "loss": 0.8849, + "step": 39820 + }, + { + "epoch": 2.56, + "learning_rate": 1.1476416410828914e-05, + "loss": 0.8751, + "step": 39821 + }, + { + "epoch": 2.56, + "learning_rate": 1.1476201721805965e-05, + "loss": 0.8641, + "step": 39822 + }, + { + "epoch": 2.56, + "learning_rate": 1.1475987032783015e-05, + "loss": 0.8855, + "step": 39823 + }, + { + "epoch": 2.56, + "learning_rate": 1.1475772343760066e-05, + "loss": 0.865, + "step": 39824 + }, + { + "epoch": 2.56, + "learning_rate": 1.1475557654737115e-05, + "loss": 0.8933, + "step": 39825 + }, + { + "epoch": 2.56, + "learning_rate": 1.1475342965714164e-05, + "loss": 0.9309, + "step": 39826 + }, + { + "epoch": 2.56, + "learning_rate": 1.1475128276691215e-05, + "loss": 0.9046, + "step": 39827 + }, + { + "epoch": 2.56, + "learning_rate": 1.1474913587668264e-05, + "loss": 0.8915, + "step": 39828 + }, + { + "epoch": 2.56, + "learning_rate": 1.1474698898645314e-05, + "loss": 0.9008, + "step": 39829 + }, + { + "epoch": 2.56, + "learning_rate": 1.1474484209622363e-05, + "loss": 0.8839, + "step": 39830 + }, + { + "epoch": 2.56, + "learning_rate": 1.1474269520599412e-05, + "loss": 0.8496, + "step": 39831 + }, + { + "epoch": 2.56, + "learning_rate": 1.1474054831576463e-05, + "loss": 0.8987, + "step": 39832 + }, + { + "epoch": 2.56, + "learning_rate": 1.1473840142553512e-05, + "loss": 0.8634, + "step": 39833 + }, + { + "epoch": 2.56, + "learning_rate": 1.147362545353056e-05, + "loss": 0.8906, + "step": 39834 + }, + { + "epoch": 2.56, + "learning_rate": 1.1473410764507611e-05, + "loss": 0.9891, + "step": 39835 + }, + { + "epoch": 2.56, + "learning_rate": 1.147319607548466e-05, + "loss": 0.8365, + "step": 39836 + }, + { + "epoch": 2.56, + "learning_rate": 1.147298138646171e-05, + "loss": 0.8893, + "step": 39837 + }, + { + "epoch": 2.56, + "learning_rate": 1.1472766697438762e-05, + "loss": 0.8703, + "step": 39838 + }, + { + "epoch": 2.56, + "learning_rate": 1.1472552008415812e-05, + "loss": 0.7989, + "step": 39839 + }, + { + "epoch": 2.56, + "learning_rate": 1.1472337319392861e-05, + "loss": 0.8409, + "step": 39840 + }, + { + "epoch": 2.56, + "learning_rate": 1.147212263036991e-05, + "loss": 0.9689, + "step": 39841 + }, + { + "epoch": 2.56, + "learning_rate": 1.1471907941346961e-05, + "loss": 0.9816, + "step": 39842 + }, + { + "epoch": 2.56, + "learning_rate": 1.147169325232401e-05, + "loss": 0.9444, + "step": 39843 + }, + { + "epoch": 2.56, + "learning_rate": 1.1471478563301059e-05, + "loss": 0.8851, + "step": 39844 + }, + { + "epoch": 2.56, + "learning_rate": 1.147126387427811e-05, + "loss": 0.8969, + "step": 39845 + }, + { + "epoch": 2.56, + "learning_rate": 1.1471049185255158e-05, + "loss": 0.8962, + "step": 39846 + }, + { + "epoch": 2.56, + "learning_rate": 1.1470834496232207e-05, + "loss": 0.9872, + "step": 39847 + }, + { + "epoch": 2.56, + "learning_rate": 1.1470619807209258e-05, + "loss": 0.9141, + "step": 39848 + }, + { + "epoch": 2.56, + "learning_rate": 1.1470405118186307e-05, + "loss": 0.9123, + "step": 39849 + }, + { + "epoch": 2.56, + "learning_rate": 1.1470190429163358e-05, + "loss": 0.873, + "step": 39850 + }, + { + "epoch": 2.56, + "learning_rate": 1.1469975740140407e-05, + "loss": 0.8493, + "step": 39851 + }, + { + "epoch": 2.56, + "learning_rate": 1.1469761051117459e-05, + "loss": 0.9055, + "step": 39852 + }, + { + "epoch": 2.56, + "learning_rate": 1.1469546362094508e-05, + "loss": 0.8813, + "step": 39853 + }, + { + "epoch": 2.56, + "learning_rate": 1.1469331673071557e-05, + "loss": 0.9128, + "step": 39854 + }, + { + "epoch": 2.56, + "learning_rate": 1.1469116984048608e-05, + "loss": 0.8059, + "step": 39855 + }, + { + "epoch": 2.56, + "learning_rate": 1.1468902295025657e-05, + "loss": 0.889, + "step": 39856 + }, + { + "epoch": 2.56, + "learning_rate": 1.1468687606002706e-05, + "loss": 0.9, + "step": 39857 + }, + { + "epoch": 2.56, + "learning_rate": 1.1468472916979756e-05, + "loss": 0.9079, + "step": 39858 + }, + { + "epoch": 2.56, + "learning_rate": 1.1468258227956805e-05, + "loss": 0.9185, + "step": 39859 + }, + { + "epoch": 2.56, + "learning_rate": 1.1468043538933856e-05, + "loss": 0.869, + "step": 39860 + }, + { + "epoch": 2.56, + "learning_rate": 1.1467828849910905e-05, + "loss": 0.93, + "step": 39861 + }, + { + "epoch": 2.56, + "learning_rate": 1.1467614160887954e-05, + "loss": 0.8553, + "step": 39862 + }, + { + "epoch": 2.56, + "learning_rate": 1.1467399471865004e-05, + "loss": 0.8899, + "step": 39863 + }, + { + "epoch": 2.56, + "learning_rate": 1.1467184782842053e-05, + "loss": 0.8652, + "step": 39864 + }, + { + "epoch": 2.56, + "learning_rate": 1.1466970093819102e-05, + "loss": 0.8518, + "step": 39865 + }, + { + "epoch": 2.56, + "learning_rate": 1.1466755404796153e-05, + "loss": 0.901, + "step": 39866 + }, + { + "epoch": 2.56, + "learning_rate": 1.1466540715773204e-05, + "loss": 0.9188, + "step": 39867 + }, + { + "epoch": 2.57, + "learning_rate": 1.1466326026750254e-05, + "loss": 0.8605, + "step": 39868 + }, + { + "epoch": 2.57, + "learning_rate": 1.1466111337727303e-05, + "loss": 0.8351, + "step": 39869 + }, + { + "epoch": 2.57, + "learning_rate": 1.1465896648704354e-05, + "loss": 0.9063, + "step": 39870 + }, + { + "epoch": 2.57, + "learning_rate": 1.1465681959681403e-05, + "loss": 0.8706, + "step": 39871 + }, + { + "epoch": 2.57, + "learning_rate": 1.1465467270658452e-05, + "loss": 0.872, + "step": 39872 + }, + { + "epoch": 2.57, + "learning_rate": 1.1465252581635503e-05, + "loss": 0.8926, + "step": 39873 + }, + { + "epoch": 2.57, + "learning_rate": 1.1465037892612552e-05, + "loss": 0.8648, + "step": 39874 + }, + { + "epoch": 2.57, + "learning_rate": 1.14648232035896e-05, + "loss": 0.8594, + "step": 39875 + }, + { + "epoch": 2.57, + "learning_rate": 1.1464608514566651e-05, + "loss": 0.8665, + "step": 39876 + }, + { + "epoch": 2.57, + "learning_rate": 1.14643938255437e-05, + "loss": 0.8877, + "step": 39877 + }, + { + "epoch": 2.57, + "learning_rate": 1.146417913652075e-05, + "loss": 0.9078, + "step": 39878 + }, + { + "epoch": 2.57, + "learning_rate": 1.14639644474978e-05, + "loss": 0.8262, + "step": 39879 + }, + { + "epoch": 2.57, + "learning_rate": 1.1463749758474849e-05, + "loss": 0.891, + "step": 39880 + }, + { + "epoch": 2.57, + "learning_rate": 1.14635350694519e-05, + "loss": 0.8388, + "step": 39881 + }, + { + "epoch": 2.57, + "learning_rate": 1.146332038042895e-05, + "loss": 0.7986, + "step": 39882 + }, + { + "epoch": 2.57, + "learning_rate": 1.1463105691406e-05, + "loss": 0.8558, + "step": 39883 + }, + { + "epoch": 2.57, + "learning_rate": 1.146289100238305e-05, + "loss": 0.9514, + "step": 39884 + }, + { + "epoch": 2.57, + "learning_rate": 1.1462676313360099e-05, + "loss": 0.9224, + "step": 39885 + }, + { + "epoch": 2.57, + "learning_rate": 1.146246162433715e-05, + "loss": 0.9018, + "step": 39886 + }, + { + "epoch": 2.57, + "learning_rate": 1.1462246935314198e-05, + "loss": 0.9189, + "step": 39887 + }, + { + "epoch": 2.57, + "learning_rate": 1.1462032246291249e-05, + "loss": 0.8361, + "step": 39888 + }, + { + "epoch": 2.57, + "learning_rate": 1.1461817557268298e-05, + "loss": 0.9039, + "step": 39889 + }, + { + "epoch": 2.57, + "learning_rate": 1.1461602868245347e-05, + "loss": 0.8643, + "step": 39890 + }, + { + "epoch": 2.57, + "learning_rate": 1.1461388179222397e-05, + "loss": 0.9382, + "step": 39891 + }, + { + "epoch": 2.57, + "learning_rate": 1.1461173490199446e-05, + "loss": 0.8862, + "step": 39892 + }, + { + "epoch": 2.57, + "learning_rate": 1.1460958801176495e-05, + "loss": 0.9114, + "step": 39893 + }, + { + "epoch": 2.57, + "learning_rate": 1.1460744112153546e-05, + "loss": 0.9043, + "step": 39894 + }, + { + "epoch": 2.57, + "learning_rate": 1.1460529423130595e-05, + "loss": 0.906, + "step": 39895 + }, + { + "epoch": 2.57, + "learning_rate": 1.1460314734107647e-05, + "loss": 0.8476, + "step": 39896 + }, + { + "epoch": 2.57, + "learning_rate": 1.1460100045084696e-05, + "loss": 0.8994, + "step": 39897 + }, + { + "epoch": 2.57, + "learning_rate": 1.1459885356061747e-05, + "loss": 0.9485, + "step": 39898 + }, + { + "epoch": 2.57, + "learning_rate": 1.1459670667038796e-05, + "loss": 0.9036, + "step": 39899 + }, + { + "epoch": 2.57, + "learning_rate": 1.1459455978015845e-05, + "loss": 0.8744, + "step": 39900 + }, + { + "epoch": 2.57, + "learning_rate": 1.1459241288992896e-05, + "loss": 0.9636, + "step": 39901 + }, + { + "epoch": 2.57, + "learning_rate": 1.1459026599969945e-05, + "loss": 0.9962, + "step": 39902 + }, + { + "epoch": 2.57, + "learning_rate": 1.1458811910946994e-05, + "loss": 0.8924, + "step": 39903 + }, + { + "epoch": 2.57, + "learning_rate": 1.1458597221924044e-05, + "loss": 0.8051, + "step": 39904 + }, + { + "epoch": 2.57, + "learning_rate": 1.1458382532901093e-05, + "loss": 0.8871, + "step": 39905 + }, + { + "epoch": 2.57, + "learning_rate": 1.1458167843878144e-05, + "loss": 0.8843, + "step": 39906 + }, + { + "epoch": 2.57, + "learning_rate": 1.1457953154855193e-05, + "loss": 0.9514, + "step": 39907 + }, + { + "epoch": 2.57, + "learning_rate": 1.1457738465832242e-05, + "loss": 0.8374, + "step": 39908 + }, + { + "epoch": 2.57, + "learning_rate": 1.1457523776809292e-05, + "loss": 0.8948, + "step": 39909 + }, + { + "epoch": 2.57, + "learning_rate": 1.1457309087786341e-05, + "loss": 0.9812, + "step": 39910 + }, + { + "epoch": 2.57, + "learning_rate": 1.1457094398763394e-05, + "loss": 0.9074, + "step": 39911 + }, + { + "epoch": 2.57, + "learning_rate": 1.1456879709740443e-05, + "loss": 0.8786, + "step": 39912 + }, + { + "epoch": 2.57, + "learning_rate": 1.1456665020717492e-05, + "loss": 0.8515, + "step": 39913 + }, + { + "epoch": 2.57, + "learning_rate": 1.1456450331694542e-05, + "loss": 0.8721, + "step": 39914 + }, + { + "epoch": 2.57, + "learning_rate": 1.1456235642671591e-05, + "loss": 0.935, + "step": 39915 + }, + { + "epoch": 2.57, + "learning_rate": 1.1456020953648642e-05, + "loss": 0.8815, + "step": 39916 + }, + { + "epoch": 2.57, + "learning_rate": 1.1455806264625691e-05, + "loss": 0.9166, + "step": 39917 + }, + { + "epoch": 2.57, + "learning_rate": 1.145559157560274e-05, + "loss": 0.9027, + "step": 39918 + }, + { + "epoch": 2.57, + "learning_rate": 1.145537688657979e-05, + "loss": 0.8728, + "step": 39919 + }, + { + "epoch": 2.57, + "learning_rate": 1.145516219755684e-05, + "loss": 0.9145, + "step": 39920 + }, + { + "epoch": 2.57, + "learning_rate": 1.1454947508533888e-05, + "loss": 0.8671, + "step": 39921 + }, + { + "epoch": 2.57, + "learning_rate": 1.1454732819510939e-05, + "loss": 0.8782, + "step": 39922 + }, + { + "epoch": 2.57, + "learning_rate": 1.1454518130487988e-05, + "loss": 0.9247, + "step": 39923 + }, + { + "epoch": 2.57, + "learning_rate": 1.1454303441465037e-05, + "loss": 0.9095, + "step": 39924 + }, + { + "epoch": 2.57, + "learning_rate": 1.145408875244209e-05, + "loss": 0.9237, + "step": 39925 + }, + { + "epoch": 2.57, + "learning_rate": 1.145387406341914e-05, + "loss": 0.8101, + "step": 39926 + }, + { + "epoch": 2.57, + "learning_rate": 1.1453659374396189e-05, + "loss": 0.8861, + "step": 39927 + }, + { + "epoch": 2.57, + "learning_rate": 1.1453444685373238e-05, + "loss": 0.9668, + "step": 39928 + }, + { + "epoch": 2.57, + "learning_rate": 1.1453229996350289e-05, + "loss": 0.9701, + "step": 39929 + }, + { + "epoch": 2.57, + "learning_rate": 1.1453015307327338e-05, + "loss": 0.7737, + "step": 39930 + }, + { + "epoch": 2.57, + "learning_rate": 1.1452800618304387e-05, + "loss": 0.9135, + "step": 39931 + }, + { + "epoch": 2.57, + "learning_rate": 1.1452585929281437e-05, + "loss": 0.8437, + "step": 39932 + }, + { + "epoch": 2.57, + "learning_rate": 1.1452371240258486e-05, + "loss": 0.8567, + "step": 39933 + }, + { + "epoch": 2.57, + "learning_rate": 1.1452156551235535e-05, + "loss": 0.8999, + "step": 39934 + }, + { + "epoch": 2.57, + "learning_rate": 1.1451941862212586e-05, + "loss": 0.8988, + "step": 39935 + }, + { + "epoch": 2.57, + "learning_rate": 1.1451727173189635e-05, + "loss": 0.9015, + "step": 39936 + }, + { + "epoch": 2.57, + "learning_rate": 1.1451512484166685e-05, + "loss": 0.9664, + "step": 39937 + }, + { + "epoch": 2.57, + "learning_rate": 1.1451297795143734e-05, + "loss": 0.8687, + "step": 39938 + }, + { + "epoch": 2.57, + "learning_rate": 1.1451083106120783e-05, + "loss": 0.8544, + "step": 39939 + }, + { + "epoch": 2.57, + "learning_rate": 1.1450868417097836e-05, + "loss": 0.863, + "step": 39940 + }, + { + "epoch": 2.57, + "learning_rate": 1.1450653728074885e-05, + "loss": 0.9386, + "step": 39941 + }, + { + "epoch": 2.57, + "learning_rate": 1.1450439039051935e-05, + "loss": 0.9005, + "step": 39942 + }, + { + "epoch": 2.57, + "learning_rate": 1.1450224350028984e-05, + "loss": 0.8932, + "step": 39943 + }, + { + "epoch": 2.57, + "learning_rate": 1.1450009661006033e-05, + "loss": 0.9017, + "step": 39944 + }, + { + "epoch": 2.57, + "learning_rate": 1.1449794971983084e-05, + "loss": 0.8349, + "step": 39945 + }, + { + "epoch": 2.57, + "learning_rate": 1.1449580282960133e-05, + "loss": 0.941, + "step": 39946 + }, + { + "epoch": 2.57, + "learning_rate": 1.1449365593937184e-05, + "loss": 0.912, + "step": 39947 + }, + { + "epoch": 2.57, + "learning_rate": 1.1449150904914233e-05, + "loss": 0.8696, + "step": 39948 + }, + { + "epoch": 2.57, + "learning_rate": 1.1448936215891281e-05, + "loss": 0.8767, + "step": 39949 + }, + { + "epoch": 2.57, + "learning_rate": 1.1448721526868332e-05, + "loss": 0.9154, + "step": 39950 + }, + { + "epoch": 2.57, + "learning_rate": 1.1448506837845381e-05, + "loss": 0.9119, + "step": 39951 + }, + { + "epoch": 2.57, + "learning_rate": 1.144829214882243e-05, + "loss": 0.8682, + "step": 39952 + }, + { + "epoch": 2.57, + "learning_rate": 1.144807745979948e-05, + "loss": 0.9213, + "step": 39953 + }, + { + "epoch": 2.57, + "learning_rate": 1.1447862770776531e-05, + "loss": 0.8621, + "step": 39954 + }, + { + "epoch": 2.57, + "learning_rate": 1.1447648081753582e-05, + "loss": 0.8517, + "step": 39955 + }, + { + "epoch": 2.57, + "learning_rate": 1.1447433392730631e-05, + "loss": 0.9356, + "step": 39956 + }, + { + "epoch": 2.57, + "learning_rate": 1.1447218703707682e-05, + "loss": 0.8794, + "step": 39957 + }, + { + "epoch": 2.57, + "learning_rate": 1.144700401468473e-05, + "loss": 0.8508, + "step": 39958 + }, + { + "epoch": 2.57, + "learning_rate": 1.144678932566178e-05, + "loss": 0.966, + "step": 39959 + }, + { + "epoch": 2.57, + "learning_rate": 1.144657463663883e-05, + "loss": 0.8911, + "step": 39960 + }, + { + "epoch": 2.57, + "learning_rate": 1.144635994761588e-05, + "loss": 0.9112, + "step": 39961 + }, + { + "epoch": 2.57, + "learning_rate": 1.1446145258592928e-05, + "loss": 0.8766, + "step": 39962 + }, + { + "epoch": 2.57, + "learning_rate": 1.1445930569569979e-05, + "loss": 0.8411, + "step": 39963 + }, + { + "epoch": 2.57, + "learning_rate": 1.1445715880547028e-05, + "loss": 0.8209, + "step": 39964 + }, + { + "epoch": 2.57, + "learning_rate": 1.1445501191524078e-05, + "loss": 0.9328, + "step": 39965 + }, + { + "epoch": 2.57, + "learning_rate": 1.1445286502501127e-05, + "loss": 0.8483, + "step": 39966 + }, + { + "epoch": 2.57, + "learning_rate": 1.1445071813478176e-05, + "loss": 0.9054, + "step": 39967 + }, + { + "epoch": 2.57, + "learning_rate": 1.1444857124455227e-05, + "loss": 0.8662, + "step": 39968 + }, + { + "epoch": 2.57, + "learning_rate": 1.1444642435432278e-05, + "loss": 0.8736, + "step": 39969 + }, + { + "epoch": 2.57, + "learning_rate": 1.1444427746409328e-05, + "loss": 0.8859, + "step": 39970 + }, + { + "epoch": 2.57, + "learning_rate": 1.1444213057386377e-05, + "loss": 0.9635, + "step": 39971 + }, + { + "epoch": 2.57, + "learning_rate": 1.1443998368363426e-05, + "loss": 0.8852, + "step": 39972 + }, + { + "epoch": 2.57, + "learning_rate": 1.1443783679340477e-05, + "loss": 0.9237, + "step": 39973 + }, + { + "epoch": 2.57, + "learning_rate": 1.1443568990317526e-05, + "loss": 0.8985, + "step": 39974 + }, + { + "epoch": 2.57, + "learning_rate": 1.1443354301294577e-05, + "loss": 0.9593, + "step": 39975 + }, + { + "epoch": 2.57, + "learning_rate": 1.1443139612271626e-05, + "loss": 0.8959, + "step": 39976 + }, + { + "epoch": 2.57, + "learning_rate": 1.1442924923248675e-05, + "loss": 0.8716, + "step": 39977 + }, + { + "epoch": 2.57, + "learning_rate": 1.1442710234225725e-05, + "loss": 0.8808, + "step": 39978 + }, + { + "epoch": 2.57, + "learning_rate": 1.1442495545202774e-05, + "loss": 0.9013, + "step": 39979 + }, + { + "epoch": 2.57, + "learning_rate": 1.1442280856179823e-05, + "loss": 0.8957, + "step": 39980 + }, + { + "epoch": 2.57, + "learning_rate": 1.1442066167156874e-05, + "loss": 0.9024, + "step": 39981 + }, + { + "epoch": 2.57, + "learning_rate": 1.1441851478133923e-05, + "loss": 0.8786, + "step": 39982 + }, + { + "epoch": 2.57, + "learning_rate": 1.1441636789110975e-05, + "loss": 0.8822, + "step": 39983 + }, + { + "epoch": 2.57, + "learning_rate": 1.1441422100088024e-05, + "loss": 0.8944, + "step": 39984 + }, + { + "epoch": 2.57, + "learning_rate": 1.1441207411065075e-05, + "loss": 0.9359, + "step": 39985 + }, + { + "epoch": 2.57, + "learning_rate": 1.1440992722042124e-05, + "loss": 0.9276, + "step": 39986 + }, + { + "epoch": 2.57, + "learning_rate": 1.1440778033019173e-05, + "loss": 0.8687, + "step": 39987 + }, + { + "epoch": 2.57, + "learning_rate": 1.1440563343996223e-05, + "loss": 0.8104, + "step": 39988 + }, + { + "epoch": 2.57, + "learning_rate": 1.1440348654973272e-05, + "loss": 0.8402, + "step": 39989 + }, + { + "epoch": 2.57, + "learning_rate": 1.1440133965950321e-05, + "loss": 0.9221, + "step": 39990 + }, + { + "epoch": 2.57, + "learning_rate": 1.1439919276927372e-05, + "loss": 0.8189, + "step": 39991 + }, + { + "epoch": 2.57, + "learning_rate": 1.1439704587904421e-05, + "loss": 0.8973, + "step": 39992 + }, + { + "epoch": 2.57, + "learning_rate": 1.143948989888147e-05, + "loss": 0.9059, + "step": 39993 + }, + { + "epoch": 2.57, + "learning_rate": 1.143927520985852e-05, + "loss": 0.9162, + "step": 39994 + }, + { + "epoch": 2.57, + "learning_rate": 1.143906052083557e-05, + "loss": 0.892, + "step": 39995 + }, + { + "epoch": 2.57, + "learning_rate": 1.143884583181262e-05, + "loss": 0.9005, + "step": 39996 + }, + { + "epoch": 2.57, + "learning_rate": 1.1438631142789669e-05, + "loss": 0.8515, + "step": 39997 + }, + { + "epoch": 2.57, + "learning_rate": 1.1438416453766721e-05, + "loss": 0.8805, + "step": 39998 + }, + { + "epoch": 2.57, + "learning_rate": 1.143820176474377e-05, + "loss": 0.8878, + "step": 39999 + }, + { + "epoch": 2.57, + "learning_rate": 1.143798707572082e-05, + "loss": 0.8912, + "step": 40000 + }, + { + "epoch": 2.57, + "eval_laion/cider": 0.5748159716689011, + "eval_laion/meteor": 0.40120212579056846, + "eval_laion/pred_num_words": 37.2265625, + "eval_laion/sacrebleu": 16.494159136885276, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0385785102844238, + "eval_runtime": 100.4699, + "eval_samples_per_second": 4.967, + "eval_steps_per_second": 0.319, + "step": 40000 + }, + { + "epoch": 2.57, + "learning_rate": 1.143777238669787e-05, + "loss": 0.9161, + "step": 40001 + }, + { + "epoch": 2.57, + "learning_rate": 1.1437557697674919e-05, + "loss": 0.9173, + "step": 40002 + }, + { + "epoch": 2.57, + "learning_rate": 1.1437343008651968e-05, + "loss": 0.9114, + "step": 40003 + }, + { + "epoch": 2.57, + "learning_rate": 1.1437128319629019e-05, + "loss": 0.8237, + "step": 40004 + }, + { + "epoch": 2.57, + "learning_rate": 1.1436913630606068e-05, + "loss": 0.8558, + "step": 40005 + }, + { + "epoch": 2.57, + "learning_rate": 1.1436698941583118e-05, + "loss": 0.9434, + "step": 40006 + }, + { + "epoch": 2.57, + "learning_rate": 1.1436484252560167e-05, + "loss": 0.8739, + "step": 40007 + }, + { + "epoch": 2.57, + "learning_rate": 1.1436269563537216e-05, + "loss": 0.8821, + "step": 40008 + }, + { + "epoch": 2.57, + "learning_rate": 1.1436054874514267e-05, + "loss": 0.8977, + "step": 40009 + }, + { + "epoch": 2.57, + "learning_rate": 1.1435840185491316e-05, + "loss": 0.8416, + "step": 40010 + }, + { + "epoch": 2.57, + "learning_rate": 1.1435625496468365e-05, + "loss": 0.8994, + "step": 40011 + }, + { + "epoch": 2.57, + "learning_rate": 1.1435410807445417e-05, + "loss": 0.9013, + "step": 40012 + }, + { + "epoch": 2.57, + "learning_rate": 1.1435196118422468e-05, + "loss": 0.8791, + "step": 40013 + }, + { + "epoch": 2.57, + "learning_rate": 1.1434981429399517e-05, + "loss": 0.8806, + "step": 40014 + }, + { + "epoch": 2.57, + "learning_rate": 1.1434766740376566e-05, + "loss": 0.847, + "step": 40015 + }, + { + "epoch": 2.57, + "learning_rate": 1.1434552051353616e-05, + "loss": 0.985, + "step": 40016 + }, + { + "epoch": 2.57, + "learning_rate": 1.1434337362330665e-05, + "loss": 0.9458, + "step": 40017 + }, + { + "epoch": 2.57, + "learning_rate": 1.1434122673307714e-05, + "loss": 0.7807, + "step": 40018 + }, + { + "epoch": 2.57, + "learning_rate": 1.1433907984284765e-05, + "loss": 0.9777, + "step": 40019 + }, + { + "epoch": 2.57, + "learning_rate": 1.1433693295261814e-05, + "loss": 0.833, + "step": 40020 + }, + { + "epoch": 2.57, + "learning_rate": 1.1433478606238863e-05, + "loss": 0.9032, + "step": 40021 + }, + { + "epoch": 2.57, + "learning_rate": 1.1433263917215914e-05, + "loss": 0.89, + "step": 40022 + }, + { + "epoch": 2.57, + "learning_rate": 1.1433049228192962e-05, + "loss": 0.8623, + "step": 40023 + }, + { + "epoch": 2.58, + "learning_rate": 1.1432834539170013e-05, + "loss": 0.9389, + "step": 40024 + }, + { + "epoch": 2.58, + "learning_rate": 1.1432619850147062e-05, + "loss": 0.9014, + "step": 40025 + }, + { + "epoch": 2.58, + "learning_rate": 1.1432405161124111e-05, + "loss": 0.8976, + "step": 40026 + }, + { + "epoch": 2.58, + "learning_rate": 1.1432190472101163e-05, + "loss": 0.9692, + "step": 40027 + }, + { + "epoch": 2.58, + "learning_rate": 1.1431975783078212e-05, + "loss": 0.9415, + "step": 40028 + }, + { + "epoch": 2.58, + "learning_rate": 1.1431761094055263e-05, + "loss": 0.8973, + "step": 40029 + }, + { + "epoch": 2.58, + "learning_rate": 1.1431546405032312e-05, + "loss": 0.9851, + "step": 40030 + }, + { + "epoch": 2.58, + "learning_rate": 1.1431331716009361e-05, + "loss": 0.927, + "step": 40031 + }, + { + "epoch": 2.58, + "learning_rate": 1.1431117026986412e-05, + "loss": 0.8683, + "step": 40032 + }, + { + "epoch": 2.58, + "learning_rate": 1.143090233796346e-05, + "loss": 0.9318, + "step": 40033 + }, + { + "epoch": 2.58, + "learning_rate": 1.1430687648940511e-05, + "loss": 0.9189, + "step": 40034 + }, + { + "epoch": 2.58, + "learning_rate": 1.143047295991756e-05, + "loss": 0.9609, + "step": 40035 + }, + { + "epoch": 2.58, + "learning_rate": 1.143025827089461e-05, + "loss": 0.8783, + "step": 40036 + }, + { + "epoch": 2.58, + "learning_rate": 1.143004358187166e-05, + "loss": 0.8578, + "step": 40037 + }, + { + "epoch": 2.58, + "learning_rate": 1.1429828892848709e-05, + "loss": 0.939, + "step": 40038 + }, + { + "epoch": 2.58, + "learning_rate": 1.1429614203825758e-05, + "loss": 0.9023, + "step": 40039 + }, + { + "epoch": 2.58, + "learning_rate": 1.1429399514802808e-05, + "loss": 0.8502, + "step": 40040 + }, + { + "epoch": 2.58, + "learning_rate": 1.1429184825779859e-05, + "loss": 0.847, + "step": 40041 + }, + { + "epoch": 2.58, + "learning_rate": 1.142897013675691e-05, + "loss": 0.8694, + "step": 40042 + }, + { + "epoch": 2.58, + "learning_rate": 1.1428755447733959e-05, + "loss": 0.8528, + "step": 40043 + }, + { + "epoch": 2.58, + "learning_rate": 1.142854075871101e-05, + "loss": 0.8194, + "step": 40044 + }, + { + "epoch": 2.58, + "learning_rate": 1.1428326069688058e-05, + "loss": 0.8994, + "step": 40045 + }, + { + "epoch": 2.58, + "learning_rate": 1.1428111380665107e-05, + "loss": 0.928, + "step": 40046 + }, + { + "epoch": 2.58, + "learning_rate": 1.1427896691642158e-05, + "loss": 0.938, + "step": 40047 + }, + { + "epoch": 2.58, + "learning_rate": 1.1427682002619207e-05, + "loss": 0.8923, + "step": 40048 + }, + { + "epoch": 2.58, + "learning_rate": 1.1427467313596256e-05, + "loss": 0.8813, + "step": 40049 + }, + { + "epoch": 2.58, + "learning_rate": 1.1427252624573307e-05, + "loss": 0.8597, + "step": 40050 + }, + { + "epoch": 2.58, + "learning_rate": 1.1427037935550356e-05, + "loss": 0.9736, + "step": 40051 + }, + { + "epoch": 2.58, + "learning_rate": 1.1426823246527406e-05, + "loss": 0.9185, + "step": 40052 + }, + { + "epoch": 2.58, + "learning_rate": 1.1426608557504455e-05, + "loss": 0.9065, + "step": 40053 + }, + { + "epoch": 2.58, + "learning_rate": 1.1426393868481504e-05, + "loss": 0.853, + "step": 40054 + }, + { + "epoch": 2.58, + "learning_rate": 1.1426179179458555e-05, + "loss": 0.9808, + "step": 40055 + }, + { + "epoch": 2.58, + "learning_rate": 1.1425964490435605e-05, + "loss": 0.9002, + "step": 40056 + }, + { + "epoch": 2.58, + "learning_rate": 1.1425749801412656e-05, + "loss": 0.8701, + "step": 40057 + }, + { + "epoch": 2.58, + "learning_rate": 1.1425535112389705e-05, + "loss": 0.8543, + "step": 40058 + }, + { + "epoch": 2.58, + "learning_rate": 1.1425320423366754e-05, + "loss": 0.9312, + "step": 40059 + }, + { + "epoch": 2.58, + "learning_rate": 1.1425105734343805e-05, + "loss": 0.8777, + "step": 40060 + }, + { + "epoch": 2.58, + "learning_rate": 1.1424891045320854e-05, + "loss": 0.9237, + "step": 40061 + }, + { + "epoch": 2.58, + "learning_rate": 1.1424676356297904e-05, + "loss": 0.8918, + "step": 40062 + }, + { + "epoch": 2.58, + "learning_rate": 1.1424461667274953e-05, + "loss": 0.8305, + "step": 40063 + }, + { + "epoch": 2.58, + "learning_rate": 1.1424246978252002e-05, + "loss": 0.8936, + "step": 40064 + }, + { + "epoch": 2.58, + "learning_rate": 1.1424032289229053e-05, + "loss": 0.9358, + "step": 40065 + }, + { + "epoch": 2.58, + "learning_rate": 1.1423817600206102e-05, + "loss": 0.8648, + "step": 40066 + }, + { + "epoch": 2.58, + "learning_rate": 1.142360291118315e-05, + "loss": 0.9348, + "step": 40067 + }, + { + "epoch": 2.58, + "learning_rate": 1.1423388222160201e-05, + "loss": 0.9501, + "step": 40068 + }, + { + "epoch": 2.58, + "learning_rate": 1.142317353313725e-05, + "loss": 0.8678, + "step": 40069 + }, + { + "epoch": 2.58, + "learning_rate": 1.1422958844114303e-05, + "loss": 0.8331, + "step": 40070 + }, + { + "epoch": 2.58, + "learning_rate": 1.1422744155091352e-05, + "loss": 0.8921, + "step": 40071 + }, + { + "epoch": 2.58, + "learning_rate": 1.1422529466068402e-05, + "loss": 1.0035, + "step": 40072 + }, + { + "epoch": 2.58, + "learning_rate": 1.1422314777045451e-05, + "loss": 0.843, + "step": 40073 + }, + { + "epoch": 2.58, + "learning_rate": 1.14221000880225e-05, + "loss": 0.9706, + "step": 40074 + }, + { + "epoch": 2.58, + "learning_rate": 1.1421885398999551e-05, + "loss": 0.9268, + "step": 40075 + }, + { + "epoch": 2.58, + "learning_rate": 1.14216707099766e-05, + "loss": 0.9674, + "step": 40076 + }, + { + "epoch": 2.58, + "learning_rate": 1.1421456020953649e-05, + "loss": 0.897, + "step": 40077 + }, + { + "epoch": 2.58, + "learning_rate": 1.14212413319307e-05, + "loss": 0.8965, + "step": 40078 + }, + { + "epoch": 2.58, + "learning_rate": 1.1421026642907749e-05, + "loss": 0.9348, + "step": 40079 + }, + { + "epoch": 2.58, + "learning_rate": 1.1420811953884798e-05, + "loss": 0.9187, + "step": 40080 + }, + { + "epoch": 2.58, + "learning_rate": 1.1420597264861848e-05, + "loss": 0.8663, + "step": 40081 + }, + { + "epoch": 2.58, + "learning_rate": 1.1420382575838897e-05, + "loss": 0.9501, + "step": 40082 + }, + { + "epoch": 2.58, + "learning_rate": 1.1420167886815948e-05, + "loss": 0.8922, + "step": 40083 + }, + { + "epoch": 2.58, + "learning_rate": 1.1419953197792997e-05, + "loss": 0.888, + "step": 40084 + }, + { + "epoch": 2.58, + "learning_rate": 1.1419738508770049e-05, + "loss": 0.9081, + "step": 40085 + }, + { + "epoch": 2.58, + "learning_rate": 1.1419523819747098e-05, + "loss": 0.8943, + "step": 40086 + }, + { + "epoch": 2.58, + "learning_rate": 1.1419309130724147e-05, + "loss": 0.9165, + "step": 40087 + }, + { + "epoch": 2.58, + "learning_rate": 1.1419094441701198e-05, + "loss": 0.8977, + "step": 40088 + }, + { + "epoch": 2.58, + "learning_rate": 1.1418879752678247e-05, + "loss": 0.9161, + "step": 40089 + }, + { + "epoch": 2.58, + "learning_rate": 1.1418665063655296e-05, + "loss": 0.8711, + "step": 40090 + }, + { + "epoch": 2.58, + "learning_rate": 1.1418450374632346e-05, + "loss": 0.9079, + "step": 40091 + }, + { + "epoch": 2.58, + "learning_rate": 1.1418235685609395e-05, + "loss": 0.9378, + "step": 40092 + }, + { + "epoch": 2.58, + "learning_rate": 1.1418020996586446e-05, + "loss": 0.8594, + "step": 40093 + }, + { + "epoch": 2.58, + "learning_rate": 1.1417806307563495e-05, + "loss": 0.9402, + "step": 40094 + }, + { + "epoch": 2.58, + "learning_rate": 1.1417591618540544e-05, + "loss": 0.9188, + "step": 40095 + }, + { + "epoch": 2.58, + "learning_rate": 1.1417376929517595e-05, + "loss": 0.9348, + "step": 40096 + }, + { + "epoch": 2.58, + "learning_rate": 1.1417162240494643e-05, + "loss": 0.9263, + "step": 40097 + }, + { + "epoch": 2.58, + "learning_rate": 1.1416947551471692e-05, + "loss": 0.9527, + "step": 40098 + }, + { + "epoch": 2.58, + "learning_rate": 1.1416732862448745e-05, + "loss": 0.8754, + "step": 40099 + }, + { + "epoch": 2.58, + "learning_rate": 1.1416518173425794e-05, + "loss": 0.9416, + "step": 40100 + }, + { + "epoch": 2.58, + "learning_rate": 1.1416303484402844e-05, + "loss": 0.8482, + "step": 40101 + }, + { + "epoch": 2.58, + "learning_rate": 1.1416088795379893e-05, + "loss": 0.9144, + "step": 40102 + }, + { + "epoch": 2.58, + "learning_rate": 1.1415874106356944e-05, + "loss": 0.8874, + "step": 40103 + }, + { + "epoch": 2.58, + "learning_rate": 1.1415659417333993e-05, + "loss": 0.8732, + "step": 40104 + }, + { + "epoch": 2.58, + "learning_rate": 1.1415444728311042e-05, + "loss": 0.9154, + "step": 40105 + }, + { + "epoch": 2.58, + "learning_rate": 1.1415230039288093e-05, + "loss": 0.8589, + "step": 40106 + }, + { + "epoch": 2.58, + "learning_rate": 1.1415015350265142e-05, + "loss": 0.8787, + "step": 40107 + }, + { + "epoch": 2.58, + "learning_rate": 1.141480066124219e-05, + "loss": 0.9596, + "step": 40108 + }, + { + "epoch": 2.58, + "learning_rate": 1.1414585972219241e-05, + "loss": 0.8862, + "step": 40109 + }, + { + "epoch": 2.58, + "learning_rate": 1.141437128319629e-05, + "loss": 0.9687, + "step": 40110 + }, + { + "epoch": 2.58, + "learning_rate": 1.1414156594173341e-05, + "loss": 0.9299, + "step": 40111 + }, + { + "epoch": 2.58, + "learning_rate": 1.141394190515039e-05, + "loss": 0.8901, + "step": 40112 + }, + { + "epoch": 2.58, + "learning_rate": 1.1413727216127439e-05, + "loss": 0.9087, + "step": 40113 + }, + { + "epoch": 2.58, + "learning_rate": 1.1413512527104491e-05, + "loss": 0.8629, + "step": 40114 + }, + { + "epoch": 2.58, + "learning_rate": 1.141329783808154e-05, + "loss": 0.9007, + "step": 40115 + }, + { + "epoch": 2.58, + "learning_rate": 1.141308314905859e-05, + "loss": 0.9218, + "step": 40116 + }, + { + "epoch": 2.58, + "learning_rate": 1.141286846003564e-05, + "loss": 0.88, + "step": 40117 + }, + { + "epoch": 2.58, + "learning_rate": 1.1412653771012689e-05, + "loss": 0.9366, + "step": 40118 + }, + { + "epoch": 2.58, + "learning_rate": 1.141243908198974e-05, + "loss": 0.978, + "step": 40119 + }, + { + "epoch": 2.58, + "learning_rate": 1.1412224392966788e-05, + "loss": 0.9725, + "step": 40120 + }, + { + "epoch": 2.58, + "learning_rate": 1.1412009703943839e-05, + "loss": 0.9372, + "step": 40121 + }, + { + "epoch": 2.58, + "learning_rate": 1.1411795014920888e-05, + "loss": 0.9205, + "step": 40122 + }, + { + "epoch": 2.58, + "learning_rate": 1.1411580325897937e-05, + "loss": 0.897, + "step": 40123 + }, + { + "epoch": 2.58, + "learning_rate": 1.1411365636874988e-05, + "loss": 0.911, + "step": 40124 + }, + { + "epoch": 2.58, + "learning_rate": 1.1411150947852037e-05, + "loss": 0.8785, + "step": 40125 + }, + { + "epoch": 2.58, + "learning_rate": 1.1410936258829085e-05, + "loss": 0.9147, + "step": 40126 + }, + { + "epoch": 2.58, + "learning_rate": 1.1410721569806136e-05, + "loss": 0.9252, + "step": 40127 + }, + { + "epoch": 2.58, + "learning_rate": 1.1410506880783187e-05, + "loss": 0.9074, + "step": 40128 + }, + { + "epoch": 2.58, + "learning_rate": 1.1410292191760237e-05, + "loss": 0.9256, + "step": 40129 + }, + { + "epoch": 2.58, + "learning_rate": 1.1410077502737286e-05, + "loss": 0.8014, + "step": 40130 + }, + { + "epoch": 2.58, + "learning_rate": 1.1409862813714337e-05, + "loss": 0.8947, + "step": 40131 + }, + { + "epoch": 2.58, + "learning_rate": 1.1409648124691386e-05, + "loss": 0.923, + "step": 40132 + }, + { + "epoch": 2.58, + "learning_rate": 1.1409433435668435e-05, + "loss": 0.9419, + "step": 40133 + }, + { + "epoch": 2.58, + "learning_rate": 1.1409218746645486e-05, + "loss": 0.9459, + "step": 40134 + }, + { + "epoch": 2.58, + "learning_rate": 1.1409004057622535e-05, + "loss": 0.9676, + "step": 40135 + }, + { + "epoch": 2.58, + "learning_rate": 1.1408789368599584e-05, + "loss": 0.9137, + "step": 40136 + }, + { + "epoch": 2.58, + "learning_rate": 1.1408574679576634e-05, + "loss": 0.8814, + "step": 40137 + }, + { + "epoch": 2.58, + "learning_rate": 1.1408359990553683e-05, + "loss": 0.8982, + "step": 40138 + }, + { + "epoch": 2.58, + "learning_rate": 1.1408145301530734e-05, + "loss": 0.8764, + "step": 40139 + }, + { + "epoch": 2.58, + "learning_rate": 1.1407930612507783e-05, + "loss": 0.9148, + "step": 40140 + }, + { + "epoch": 2.58, + "learning_rate": 1.1407715923484832e-05, + "loss": 0.9243, + "step": 40141 + }, + { + "epoch": 2.58, + "learning_rate": 1.1407501234461882e-05, + "loss": 0.8614, + "step": 40142 + }, + { + "epoch": 2.58, + "learning_rate": 1.1407286545438933e-05, + "loss": 0.9112, + "step": 40143 + }, + { + "epoch": 2.58, + "learning_rate": 1.1407071856415984e-05, + "loss": 0.9316, + "step": 40144 + }, + { + "epoch": 2.58, + "learning_rate": 1.1406857167393033e-05, + "loss": 0.9229, + "step": 40145 + }, + { + "epoch": 2.58, + "learning_rate": 1.1406642478370082e-05, + "loss": 0.8941, + "step": 40146 + }, + { + "epoch": 2.58, + "learning_rate": 1.1406427789347132e-05, + "loss": 0.873, + "step": 40147 + }, + { + "epoch": 2.58, + "learning_rate": 1.1406213100324181e-05, + "loss": 0.9555, + "step": 40148 + }, + { + "epoch": 2.58, + "learning_rate": 1.1405998411301232e-05, + "loss": 0.8127, + "step": 40149 + }, + { + "epoch": 2.58, + "learning_rate": 1.1405783722278281e-05, + "loss": 0.8791, + "step": 40150 + }, + { + "epoch": 2.58, + "learning_rate": 1.140556903325533e-05, + "loss": 0.8759, + "step": 40151 + }, + { + "epoch": 2.58, + "learning_rate": 1.140535434423238e-05, + "loss": 0.8473, + "step": 40152 + }, + { + "epoch": 2.58, + "learning_rate": 1.140513965520943e-05, + "loss": 0.8914, + "step": 40153 + }, + { + "epoch": 2.58, + "learning_rate": 1.1404924966186479e-05, + "loss": 0.904, + "step": 40154 + }, + { + "epoch": 2.58, + "learning_rate": 1.140471027716353e-05, + "loss": 0.8816, + "step": 40155 + }, + { + "epoch": 2.58, + "learning_rate": 1.1404495588140578e-05, + "loss": 0.8893, + "step": 40156 + }, + { + "epoch": 2.58, + "learning_rate": 1.140428089911763e-05, + "loss": 0.9428, + "step": 40157 + }, + { + "epoch": 2.58, + "learning_rate": 1.140406621009468e-05, + "loss": 0.923, + "step": 40158 + }, + { + "epoch": 2.58, + "learning_rate": 1.140385152107173e-05, + "loss": 0.8554, + "step": 40159 + }, + { + "epoch": 2.58, + "learning_rate": 1.1403636832048779e-05, + "loss": 0.8425, + "step": 40160 + }, + { + "epoch": 2.58, + "learning_rate": 1.1403422143025828e-05, + "loss": 0.8856, + "step": 40161 + }, + { + "epoch": 2.58, + "learning_rate": 1.1403207454002879e-05, + "loss": 0.8836, + "step": 40162 + }, + { + "epoch": 2.58, + "learning_rate": 1.1402992764979928e-05, + "loss": 0.8713, + "step": 40163 + }, + { + "epoch": 2.58, + "learning_rate": 1.1402778075956977e-05, + "loss": 0.9458, + "step": 40164 + }, + { + "epoch": 2.58, + "learning_rate": 1.1402563386934027e-05, + "loss": 0.9116, + "step": 40165 + }, + { + "epoch": 2.58, + "learning_rate": 1.1402348697911076e-05, + "loss": 0.9333, + "step": 40166 + }, + { + "epoch": 2.58, + "learning_rate": 1.1402134008888125e-05, + "loss": 0.931, + "step": 40167 + }, + { + "epoch": 2.58, + "learning_rate": 1.1401919319865176e-05, + "loss": 0.8632, + "step": 40168 + }, + { + "epoch": 2.58, + "learning_rate": 1.1401704630842225e-05, + "loss": 0.8544, + "step": 40169 + }, + { + "epoch": 2.58, + "learning_rate": 1.1401489941819276e-05, + "loss": 0.8793, + "step": 40170 + }, + { + "epoch": 2.58, + "learning_rate": 1.1401275252796324e-05, + "loss": 0.923, + "step": 40171 + }, + { + "epoch": 2.58, + "learning_rate": 1.1401060563773377e-05, + "loss": 0.876, + "step": 40172 + }, + { + "epoch": 2.58, + "learning_rate": 1.1400845874750426e-05, + "loss": 0.9587, + "step": 40173 + }, + { + "epoch": 2.58, + "learning_rate": 1.1400631185727475e-05, + "loss": 0.8656, + "step": 40174 + }, + { + "epoch": 2.58, + "learning_rate": 1.1400416496704525e-05, + "loss": 0.897, + "step": 40175 + }, + { + "epoch": 2.58, + "learning_rate": 1.1400201807681574e-05, + "loss": 0.9004, + "step": 40176 + }, + { + "epoch": 2.58, + "learning_rate": 1.1399987118658623e-05, + "loss": 0.917, + "step": 40177 + }, + { + "epoch": 2.58, + "learning_rate": 1.1399772429635674e-05, + "loss": 0.8852, + "step": 40178 + }, + { + "epoch": 2.59, + "learning_rate": 1.1399557740612723e-05, + "loss": 0.8703, + "step": 40179 + }, + { + "epoch": 2.59, + "learning_rate": 1.1399343051589774e-05, + "loss": 0.884, + "step": 40180 + }, + { + "epoch": 2.59, + "learning_rate": 1.1399128362566823e-05, + "loss": 0.8841, + "step": 40181 + }, + { + "epoch": 2.59, + "learning_rate": 1.1398913673543872e-05, + "loss": 0.8783, + "step": 40182 + }, + { + "epoch": 2.59, + "learning_rate": 1.1398698984520922e-05, + "loss": 0.9186, + "step": 40183 + }, + { + "epoch": 2.59, + "learning_rate": 1.1398484295497971e-05, + "loss": 0.9328, + "step": 40184 + }, + { + "epoch": 2.59, + "learning_rate": 1.139826960647502e-05, + "loss": 0.8576, + "step": 40185 + }, + { + "epoch": 2.59, + "learning_rate": 1.1398054917452073e-05, + "loss": 0.8823, + "step": 40186 + }, + { + "epoch": 2.59, + "learning_rate": 1.1397840228429121e-05, + "loss": 0.8569, + "step": 40187 + }, + { + "epoch": 2.59, + "learning_rate": 1.1397625539406172e-05, + "loss": 0.853, + "step": 40188 + }, + { + "epoch": 2.59, + "learning_rate": 1.1397410850383221e-05, + "loss": 0.8975, + "step": 40189 + }, + { + "epoch": 2.59, + "learning_rate": 1.1397196161360272e-05, + "loss": 0.8713, + "step": 40190 + }, + { + "epoch": 2.59, + "learning_rate": 1.139698147233732e-05, + "loss": 0.8792, + "step": 40191 + }, + { + "epoch": 2.59, + "learning_rate": 1.139676678331437e-05, + "loss": 0.8996, + "step": 40192 + }, + { + "epoch": 2.59, + "learning_rate": 1.139655209429142e-05, + "loss": 0.8978, + "step": 40193 + }, + { + "epoch": 2.59, + "learning_rate": 1.139633740526847e-05, + "loss": 0.9179, + "step": 40194 + }, + { + "epoch": 2.59, + "learning_rate": 1.1396122716245518e-05, + "loss": 0.9124, + "step": 40195 + }, + { + "epoch": 2.59, + "learning_rate": 1.1395908027222569e-05, + "loss": 0.9071, + "step": 40196 + }, + { + "epoch": 2.59, + "learning_rate": 1.1395693338199618e-05, + "loss": 0.8483, + "step": 40197 + }, + { + "epoch": 2.59, + "learning_rate": 1.1395478649176669e-05, + "loss": 0.9081, + "step": 40198 + }, + { + "epoch": 2.59, + "learning_rate": 1.1395263960153718e-05, + "loss": 0.9247, + "step": 40199 + }, + { + "epoch": 2.59, + "learning_rate": 1.1395049271130766e-05, + "loss": 0.9305, + "step": 40200 + }, + { + "epoch": 2.59, + "learning_rate": 1.1394834582107819e-05, + "loss": 0.8678, + "step": 40201 + }, + { + "epoch": 2.59, + "learning_rate": 1.1394619893084868e-05, + "loss": 0.8242, + "step": 40202 + }, + { + "epoch": 2.59, + "learning_rate": 1.1394405204061918e-05, + "loss": 0.9468, + "step": 40203 + }, + { + "epoch": 2.59, + "learning_rate": 1.1394190515038967e-05, + "loss": 0.8915, + "step": 40204 + }, + { + "epoch": 2.59, + "learning_rate": 1.1393975826016016e-05, + "loss": 0.9171, + "step": 40205 + }, + { + "epoch": 2.59, + "learning_rate": 1.1393761136993067e-05, + "loss": 0.913, + "step": 40206 + }, + { + "epoch": 2.59, + "learning_rate": 1.1393546447970116e-05, + "loss": 0.9542, + "step": 40207 + }, + { + "epoch": 2.59, + "learning_rate": 1.1393331758947167e-05, + "loss": 0.898, + "step": 40208 + }, + { + "epoch": 2.59, + "learning_rate": 1.1393117069924216e-05, + "loss": 0.8313, + "step": 40209 + }, + { + "epoch": 2.59, + "learning_rate": 1.1392902380901265e-05, + "loss": 0.9283, + "step": 40210 + }, + { + "epoch": 2.59, + "learning_rate": 1.1392687691878315e-05, + "loss": 0.9359, + "step": 40211 + }, + { + "epoch": 2.59, + "learning_rate": 1.1392473002855364e-05, + "loss": 0.8851, + "step": 40212 + }, + { + "epoch": 2.59, + "learning_rate": 1.1392258313832413e-05, + "loss": 0.9205, + "step": 40213 + }, + { + "epoch": 2.59, + "learning_rate": 1.1392043624809464e-05, + "loss": 0.8415, + "step": 40214 + }, + { + "epoch": 2.59, + "learning_rate": 1.1391828935786515e-05, + "loss": 0.8223, + "step": 40215 + }, + { + "epoch": 2.59, + "learning_rate": 1.1391614246763565e-05, + "loss": 0.8601, + "step": 40216 + }, + { + "epoch": 2.59, + "learning_rate": 1.1391399557740614e-05, + "loss": 0.8705, + "step": 40217 + }, + { + "epoch": 2.59, + "learning_rate": 1.1391184868717665e-05, + "loss": 0.8609, + "step": 40218 + }, + { + "epoch": 2.59, + "learning_rate": 1.1390970179694714e-05, + "loss": 0.8582, + "step": 40219 + }, + { + "epoch": 2.59, + "learning_rate": 1.1390755490671763e-05, + "loss": 0.9004, + "step": 40220 + }, + { + "epoch": 2.59, + "learning_rate": 1.1390540801648813e-05, + "loss": 0.9272, + "step": 40221 + }, + { + "epoch": 2.59, + "learning_rate": 1.1390326112625862e-05, + "loss": 0.8531, + "step": 40222 + }, + { + "epoch": 2.59, + "learning_rate": 1.1390111423602911e-05, + "loss": 0.8261, + "step": 40223 + }, + { + "epoch": 2.59, + "learning_rate": 1.1389896734579962e-05, + "loss": 0.8649, + "step": 40224 + }, + { + "epoch": 2.59, + "learning_rate": 1.1389682045557011e-05, + "loss": 0.9547, + "step": 40225 + }, + { + "epoch": 2.59, + "learning_rate": 1.1389467356534062e-05, + "loss": 0.8975, + "step": 40226 + }, + { + "epoch": 2.59, + "learning_rate": 1.138925266751111e-05, + "loss": 0.9488, + "step": 40227 + }, + { + "epoch": 2.59, + "learning_rate": 1.138903797848816e-05, + "loss": 0.8295, + "step": 40228 + }, + { + "epoch": 2.59, + "learning_rate": 1.138882328946521e-05, + "loss": 0.929, + "step": 40229 + }, + { + "epoch": 2.59, + "learning_rate": 1.1388608600442261e-05, + "loss": 0.8985, + "step": 40230 + }, + { + "epoch": 2.59, + "learning_rate": 1.1388393911419312e-05, + "loss": 0.8671, + "step": 40231 + }, + { + "epoch": 2.59, + "learning_rate": 1.138817922239636e-05, + "loss": 0.8726, + "step": 40232 + }, + { + "epoch": 2.59, + "learning_rate": 1.138796453337341e-05, + "loss": 0.9022, + "step": 40233 + }, + { + "epoch": 2.59, + "learning_rate": 1.138774984435046e-05, + "loss": 0.9012, + "step": 40234 + }, + { + "epoch": 2.59, + "learning_rate": 1.1387535155327509e-05, + "loss": 0.9173, + "step": 40235 + }, + { + "epoch": 2.59, + "learning_rate": 1.138732046630456e-05, + "loss": 0.9042, + "step": 40236 + }, + { + "epoch": 2.59, + "learning_rate": 1.1387105777281609e-05, + "loss": 0.8618, + "step": 40237 + }, + { + "epoch": 2.59, + "learning_rate": 1.1386891088258658e-05, + "loss": 0.9456, + "step": 40238 + }, + { + "epoch": 2.59, + "learning_rate": 1.1386676399235708e-05, + "loss": 0.9208, + "step": 40239 + }, + { + "epoch": 2.59, + "learning_rate": 1.1386461710212757e-05, + "loss": 0.8592, + "step": 40240 + }, + { + "epoch": 2.59, + "learning_rate": 1.1386247021189806e-05, + "loss": 0.8727, + "step": 40241 + }, + { + "epoch": 2.59, + "learning_rate": 1.1386032332166857e-05, + "loss": 0.9197, + "step": 40242 + }, + { + "epoch": 2.59, + "learning_rate": 1.1385817643143906e-05, + "loss": 0.8815, + "step": 40243 + }, + { + "epoch": 2.59, + "learning_rate": 1.1385602954120955e-05, + "loss": 0.9093, + "step": 40244 + }, + { + "epoch": 2.59, + "learning_rate": 1.1385388265098007e-05, + "loss": 0.9573, + "step": 40245 + }, + { + "epoch": 2.59, + "learning_rate": 1.1385173576075058e-05, + "loss": 0.8779, + "step": 40246 + }, + { + "epoch": 2.59, + "learning_rate": 1.1384958887052107e-05, + "loss": 0.9078, + "step": 40247 + }, + { + "epoch": 2.59, + "learning_rate": 1.1384744198029156e-05, + "loss": 0.8219, + "step": 40248 + }, + { + "epoch": 2.59, + "learning_rate": 1.1384529509006206e-05, + "loss": 0.8715, + "step": 40249 + }, + { + "epoch": 2.59, + "learning_rate": 1.1384314819983255e-05, + "loss": 0.8338, + "step": 40250 + }, + { + "epoch": 2.59, + "learning_rate": 1.1384100130960304e-05, + "loss": 0.8878, + "step": 40251 + }, + { + "epoch": 2.59, + "learning_rate": 1.1383885441937355e-05, + "loss": 0.8996, + "step": 40252 + }, + { + "epoch": 2.59, + "learning_rate": 1.1383670752914404e-05, + "loss": 0.8867, + "step": 40253 + }, + { + "epoch": 2.59, + "learning_rate": 1.1383456063891453e-05, + "loss": 0.863, + "step": 40254 + }, + { + "epoch": 2.59, + "learning_rate": 1.1383241374868504e-05, + "loss": 0.9944, + "step": 40255 + }, + { + "epoch": 2.59, + "learning_rate": 1.1383026685845553e-05, + "loss": 0.8875, + "step": 40256 + }, + { + "epoch": 2.59, + "learning_rate": 1.1382811996822603e-05, + "loss": 0.8803, + "step": 40257 + }, + { + "epoch": 2.59, + "learning_rate": 1.1382597307799652e-05, + "loss": 0.8979, + "step": 40258 + }, + { + "epoch": 2.59, + "learning_rate": 1.1382382618776705e-05, + "loss": 0.8414, + "step": 40259 + }, + { + "epoch": 2.59, + "learning_rate": 1.1382167929753754e-05, + "loss": 0.956, + "step": 40260 + }, + { + "epoch": 2.59, + "learning_rate": 1.1381953240730802e-05, + "loss": 0.8981, + "step": 40261 + }, + { + "epoch": 2.59, + "learning_rate": 1.1381738551707853e-05, + "loss": 0.9279, + "step": 40262 + }, + { + "epoch": 2.59, + "learning_rate": 1.1381523862684902e-05, + "loss": 0.9376, + "step": 40263 + }, + { + "epoch": 2.59, + "learning_rate": 1.1381309173661951e-05, + "loss": 0.8512, + "step": 40264 + }, + { + "epoch": 2.59, + "learning_rate": 1.1381094484639002e-05, + "loss": 0.9266, + "step": 40265 + }, + { + "epoch": 2.59, + "learning_rate": 1.138087979561605e-05, + "loss": 0.9085, + "step": 40266 + }, + { + "epoch": 2.59, + "learning_rate": 1.1380665106593101e-05, + "loss": 0.8821, + "step": 40267 + }, + { + "epoch": 2.59, + "learning_rate": 1.138045041757015e-05, + "loss": 0.919, + "step": 40268 + }, + { + "epoch": 2.59, + "learning_rate": 1.13802357285472e-05, + "loss": 0.8284, + "step": 40269 + }, + { + "epoch": 2.59, + "learning_rate": 1.138002103952425e-05, + "loss": 0.9069, + "step": 40270 + }, + { + "epoch": 2.59, + "learning_rate": 1.1379806350501299e-05, + "loss": 0.8812, + "step": 40271 + }, + { + "epoch": 2.59, + "learning_rate": 1.1379591661478348e-05, + "loss": 0.8673, + "step": 40272 + }, + { + "epoch": 2.59, + "learning_rate": 1.1379376972455399e-05, + "loss": 0.8675, + "step": 40273 + }, + { + "epoch": 2.59, + "learning_rate": 1.137916228343245e-05, + "loss": 0.9401, + "step": 40274 + }, + { + "epoch": 2.59, + "learning_rate": 1.13789475944095e-05, + "loss": 0.8303, + "step": 40275 + }, + { + "epoch": 2.59, + "learning_rate": 1.1378732905386549e-05, + "loss": 0.8583, + "step": 40276 + }, + { + "epoch": 2.59, + "learning_rate": 1.13785182163636e-05, + "loss": 0.9591, + "step": 40277 + }, + { + "epoch": 2.59, + "learning_rate": 1.1378303527340648e-05, + "loss": 0.8846, + "step": 40278 + }, + { + "epoch": 2.59, + "learning_rate": 1.1378088838317697e-05, + "loss": 0.8265, + "step": 40279 + }, + { + "epoch": 2.59, + "learning_rate": 1.1377874149294748e-05, + "loss": 0.8801, + "step": 40280 + }, + { + "epoch": 2.59, + "learning_rate": 1.1377659460271797e-05, + "loss": 0.8815, + "step": 40281 + }, + { + "epoch": 2.59, + "learning_rate": 1.1377444771248846e-05, + "loss": 0.9484, + "step": 40282 + }, + { + "epoch": 2.59, + "learning_rate": 1.1377230082225897e-05, + "loss": 0.8793, + "step": 40283 + }, + { + "epoch": 2.59, + "learning_rate": 1.1377015393202946e-05, + "loss": 0.9143, + "step": 40284 + }, + { + "epoch": 2.59, + "learning_rate": 1.1376800704179996e-05, + "loss": 0.9445, + "step": 40285 + }, + { + "epoch": 2.59, + "learning_rate": 1.1376586015157045e-05, + "loss": 0.8549, + "step": 40286 + }, + { + "epoch": 2.59, + "learning_rate": 1.1376371326134094e-05, + "loss": 0.8974, + "step": 40287 + }, + { + "epoch": 2.59, + "learning_rate": 1.1376156637111147e-05, + "loss": 0.865, + "step": 40288 + }, + { + "epoch": 2.59, + "learning_rate": 1.1375941948088196e-05, + "loss": 0.8531, + "step": 40289 + }, + { + "epoch": 2.59, + "learning_rate": 1.1375727259065246e-05, + "loss": 0.9193, + "step": 40290 + }, + { + "epoch": 2.59, + "learning_rate": 1.1375512570042295e-05, + "loss": 0.8418, + "step": 40291 + }, + { + "epoch": 2.59, + "learning_rate": 1.1375297881019344e-05, + "loss": 0.8462, + "step": 40292 + }, + { + "epoch": 2.59, + "learning_rate": 1.1375083191996395e-05, + "loss": 0.9041, + "step": 40293 + }, + { + "epoch": 2.59, + "learning_rate": 1.1374868502973444e-05, + "loss": 0.9267, + "step": 40294 + }, + { + "epoch": 2.59, + "learning_rate": 1.1374653813950494e-05, + "loss": 0.9025, + "step": 40295 + }, + { + "epoch": 2.59, + "learning_rate": 1.1374439124927543e-05, + "loss": 0.8737, + "step": 40296 + }, + { + "epoch": 2.59, + "learning_rate": 1.1374224435904592e-05, + "loss": 0.9538, + "step": 40297 + }, + { + "epoch": 2.59, + "learning_rate": 1.1374009746881643e-05, + "loss": 0.8872, + "step": 40298 + }, + { + "epoch": 2.59, + "learning_rate": 1.1373795057858692e-05, + "loss": 0.96, + "step": 40299 + }, + { + "epoch": 2.59, + "learning_rate": 1.1373580368835741e-05, + "loss": 0.8279, + "step": 40300 + }, + { + "epoch": 2.59, + "learning_rate": 1.1373365679812792e-05, + "loss": 0.8942, + "step": 40301 + }, + { + "epoch": 2.59, + "learning_rate": 1.137315099078984e-05, + "loss": 0.9895, + "step": 40302 + }, + { + "epoch": 2.59, + "learning_rate": 1.1372936301766893e-05, + "loss": 0.9291, + "step": 40303 + }, + { + "epoch": 2.59, + "learning_rate": 1.1372721612743942e-05, + "loss": 0.8735, + "step": 40304 + }, + { + "epoch": 2.59, + "learning_rate": 1.1372506923720993e-05, + "loss": 0.8816, + "step": 40305 + }, + { + "epoch": 2.59, + "learning_rate": 1.1372292234698041e-05, + "loss": 0.9117, + "step": 40306 + }, + { + "epoch": 2.59, + "learning_rate": 1.137207754567509e-05, + "loss": 0.8895, + "step": 40307 + }, + { + "epoch": 2.59, + "learning_rate": 1.1371862856652141e-05, + "loss": 0.8672, + "step": 40308 + }, + { + "epoch": 2.59, + "learning_rate": 1.137164816762919e-05, + "loss": 0.9182, + "step": 40309 + }, + { + "epoch": 2.59, + "learning_rate": 1.1371433478606239e-05, + "loss": 1.0185, + "step": 40310 + }, + { + "epoch": 2.59, + "learning_rate": 1.137121878958329e-05, + "loss": 0.8227, + "step": 40311 + }, + { + "epoch": 2.59, + "learning_rate": 1.1371004100560339e-05, + "loss": 0.8638, + "step": 40312 + }, + { + "epoch": 2.59, + "learning_rate": 1.1370789411537388e-05, + "loss": 0.9066, + "step": 40313 + }, + { + "epoch": 2.59, + "learning_rate": 1.1370574722514438e-05, + "loss": 0.8322, + "step": 40314 + }, + { + "epoch": 2.59, + "learning_rate": 1.1370360033491487e-05, + "loss": 0.8782, + "step": 40315 + }, + { + "epoch": 2.59, + "learning_rate": 1.1370145344468538e-05, + "loss": 0.8709, + "step": 40316 + }, + { + "epoch": 2.59, + "learning_rate": 1.1369930655445589e-05, + "loss": 0.8964, + "step": 40317 + }, + { + "epoch": 2.59, + "learning_rate": 1.136971596642264e-05, + "loss": 0.8909, + "step": 40318 + }, + { + "epoch": 2.59, + "learning_rate": 1.1369501277399688e-05, + "loss": 0.7999, + "step": 40319 + }, + { + "epoch": 2.59, + "learning_rate": 1.1369286588376737e-05, + "loss": 1.0032, + "step": 40320 + }, + { + "epoch": 2.59, + "learning_rate": 1.1369071899353788e-05, + "loss": 0.9233, + "step": 40321 + }, + { + "epoch": 2.59, + "learning_rate": 1.1368857210330837e-05, + "loss": 0.8534, + "step": 40322 + }, + { + "epoch": 2.59, + "learning_rate": 1.1368642521307887e-05, + "loss": 0.8708, + "step": 40323 + }, + { + "epoch": 2.59, + "learning_rate": 1.1368427832284936e-05, + "loss": 0.9358, + "step": 40324 + }, + { + "epoch": 2.59, + "learning_rate": 1.1368213143261985e-05, + "loss": 0.9362, + "step": 40325 + }, + { + "epoch": 2.59, + "learning_rate": 1.1367998454239036e-05, + "loss": 0.883, + "step": 40326 + }, + { + "epoch": 2.59, + "learning_rate": 1.1367783765216085e-05, + "loss": 0.9045, + "step": 40327 + }, + { + "epoch": 2.59, + "learning_rate": 1.1367569076193134e-05, + "loss": 0.9824, + "step": 40328 + }, + { + "epoch": 2.59, + "learning_rate": 1.1367354387170185e-05, + "loss": 0.8344, + "step": 40329 + }, + { + "epoch": 2.59, + "learning_rate": 1.1367139698147234e-05, + "loss": 0.8436, + "step": 40330 + }, + { + "epoch": 2.59, + "learning_rate": 1.1366925009124283e-05, + "loss": 0.8242, + "step": 40331 + }, + { + "epoch": 2.59, + "learning_rate": 1.1366710320101335e-05, + "loss": 0.9563, + "step": 40332 + }, + { + "epoch": 2.59, + "learning_rate": 1.1366495631078386e-05, + "loss": 0.874, + "step": 40333 + }, + { + "epoch": 2.59, + "learning_rate": 1.1366280942055435e-05, + "loss": 0.9828, + "step": 40334 + }, + { + "epoch": 2.6, + "learning_rate": 1.1366066253032483e-05, + "loss": 0.9206, + "step": 40335 + }, + { + "epoch": 2.6, + "learning_rate": 1.1365851564009534e-05, + "loss": 0.92, + "step": 40336 + }, + { + "epoch": 2.6, + "learning_rate": 1.1365636874986583e-05, + "loss": 0.8384, + "step": 40337 + }, + { + "epoch": 2.6, + "learning_rate": 1.1365422185963632e-05, + "loss": 0.8377, + "step": 40338 + }, + { + "epoch": 2.6, + "learning_rate": 1.1365207496940683e-05, + "loss": 0.893, + "step": 40339 + }, + { + "epoch": 2.6, + "learning_rate": 1.1364992807917732e-05, + "loss": 0.9404, + "step": 40340 + }, + { + "epoch": 2.6, + "learning_rate": 1.136477811889478e-05, + "loss": 0.9453, + "step": 40341 + }, + { + "epoch": 2.6, + "learning_rate": 1.1364563429871831e-05, + "loss": 0.848, + "step": 40342 + }, + { + "epoch": 2.6, + "learning_rate": 1.136434874084888e-05, + "loss": 0.8746, + "step": 40343 + }, + { + "epoch": 2.6, + "learning_rate": 1.1364134051825931e-05, + "loss": 0.848, + "step": 40344 + }, + { + "epoch": 2.6, + "learning_rate": 1.136391936280298e-05, + "loss": 0.9934, + "step": 40345 + }, + { + "epoch": 2.6, + "learning_rate": 1.1363704673780032e-05, + "loss": 0.9146, + "step": 40346 + }, + { + "epoch": 2.6, + "learning_rate": 1.1363489984757081e-05, + "loss": 0.9112, + "step": 40347 + }, + { + "epoch": 2.6, + "learning_rate": 1.136327529573413e-05, + "loss": 0.9018, + "step": 40348 + }, + { + "epoch": 2.6, + "learning_rate": 1.1363060606711181e-05, + "loss": 0.9633, + "step": 40349 + }, + { + "epoch": 2.6, + "learning_rate": 1.136284591768823e-05, + "loss": 0.8824, + "step": 40350 + }, + { + "epoch": 2.6, + "learning_rate": 1.1362631228665279e-05, + "loss": 0.8866, + "step": 40351 + }, + { + "epoch": 2.6, + "learning_rate": 1.136241653964233e-05, + "loss": 0.8707, + "step": 40352 + }, + { + "epoch": 2.6, + "learning_rate": 1.1362201850619378e-05, + "loss": 0.8403, + "step": 40353 + }, + { + "epoch": 2.6, + "learning_rate": 1.1361987161596429e-05, + "loss": 0.8194, + "step": 40354 + }, + { + "epoch": 2.6, + "learning_rate": 1.1361772472573478e-05, + "loss": 0.9267, + "step": 40355 + }, + { + "epoch": 2.6, + "learning_rate": 1.1361557783550527e-05, + "loss": 0.9233, + "step": 40356 + }, + { + "epoch": 2.6, + "learning_rate": 1.1361343094527578e-05, + "loss": 0.889, + "step": 40357 + }, + { + "epoch": 2.6, + "learning_rate": 1.1361128405504627e-05, + "loss": 0.871, + "step": 40358 + }, + { + "epoch": 2.6, + "learning_rate": 1.1360913716481676e-05, + "loss": 0.947, + "step": 40359 + }, + { + "epoch": 2.6, + "learning_rate": 1.1360699027458726e-05, + "loss": 0.9056, + "step": 40360 + }, + { + "epoch": 2.6, + "learning_rate": 1.1360484338435777e-05, + "loss": 0.8527, + "step": 40361 + }, + { + "epoch": 2.6, + "learning_rate": 1.1360269649412828e-05, + "loss": 0.9395, + "step": 40362 + }, + { + "epoch": 2.6, + "learning_rate": 1.1360054960389877e-05, + "loss": 0.99, + "step": 40363 + }, + { + "epoch": 2.6, + "learning_rate": 1.1359840271366927e-05, + "loss": 0.8274, + "step": 40364 + }, + { + "epoch": 2.6, + "learning_rate": 1.1359625582343976e-05, + "loss": 0.9009, + "step": 40365 + }, + { + "epoch": 2.6, + "learning_rate": 1.1359410893321025e-05, + "loss": 0.876, + "step": 40366 + }, + { + "epoch": 2.6, + "learning_rate": 1.1359196204298076e-05, + "loss": 0.8683, + "step": 40367 + }, + { + "epoch": 2.6, + "learning_rate": 1.1358981515275125e-05, + "loss": 0.8717, + "step": 40368 + }, + { + "epoch": 2.6, + "learning_rate": 1.1358766826252174e-05, + "loss": 0.8584, + "step": 40369 + }, + { + "epoch": 2.6, + "learning_rate": 1.1358552137229224e-05, + "loss": 0.9251, + "step": 40370 + }, + { + "epoch": 2.6, + "learning_rate": 1.1358337448206273e-05, + "loss": 0.9135, + "step": 40371 + }, + { + "epoch": 2.6, + "learning_rate": 1.1358122759183324e-05, + "loss": 0.9393, + "step": 40372 + }, + { + "epoch": 2.6, + "learning_rate": 1.1357908070160373e-05, + "loss": 0.8304, + "step": 40373 + }, + { + "epoch": 2.6, + "learning_rate": 1.1357693381137422e-05, + "loss": 0.9805, + "step": 40374 + }, + { + "epoch": 2.6, + "learning_rate": 1.1357478692114474e-05, + "loss": 0.9322, + "step": 40375 + }, + { + "epoch": 2.6, + "learning_rate": 1.1357264003091523e-05, + "loss": 0.8546, + "step": 40376 + }, + { + "epoch": 2.6, + "learning_rate": 1.1357049314068574e-05, + "loss": 0.9178, + "step": 40377 + }, + { + "epoch": 2.6, + "learning_rate": 1.1356834625045623e-05, + "loss": 0.9394, + "step": 40378 + }, + { + "epoch": 2.6, + "learning_rate": 1.1356619936022672e-05, + "loss": 0.897, + "step": 40379 + }, + { + "epoch": 2.6, + "learning_rate": 1.1356405246999722e-05, + "loss": 0.9348, + "step": 40380 + }, + { + "epoch": 2.6, + "learning_rate": 1.1356190557976771e-05, + "loss": 0.8128, + "step": 40381 + }, + { + "epoch": 2.6, + "learning_rate": 1.1355975868953822e-05, + "loss": 0.8944, + "step": 40382 + }, + { + "epoch": 2.6, + "learning_rate": 1.1355761179930871e-05, + "loss": 0.8963, + "step": 40383 + }, + { + "epoch": 2.6, + "learning_rate": 1.135554649090792e-05, + "loss": 0.9078, + "step": 40384 + }, + { + "epoch": 2.6, + "learning_rate": 1.135533180188497e-05, + "loss": 0.9477, + "step": 40385 + }, + { + "epoch": 2.6, + "learning_rate": 1.135511711286202e-05, + "loss": 0.8555, + "step": 40386 + }, + { + "epoch": 2.6, + "learning_rate": 1.1354902423839069e-05, + "loss": 0.8214, + "step": 40387 + }, + { + "epoch": 2.6, + "learning_rate": 1.135468773481612e-05, + "loss": 0.8654, + "step": 40388 + }, + { + "epoch": 2.6, + "learning_rate": 1.1354473045793168e-05, + "loss": 0.8873, + "step": 40389 + }, + { + "epoch": 2.6, + "learning_rate": 1.135425835677022e-05, + "loss": 0.8629, + "step": 40390 + }, + { + "epoch": 2.6, + "learning_rate": 1.135404366774727e-05, + "loss": 0.9559, + "step": 40391 + }, + { + "epoch": 2.6, + "learning_rate": 1.135382897872432e-05, + "loss": 0.9584, + "step": 40392 + }, + { + "epoch": 2.6, + "learning_rate": 1.135361428970137e-05, + "loss": 0.8743, + "step": 40393 + }, + { + "epoch": 2.6, + "learning_rate": 1.1353399600678418e-05, + "loss": 0.8413, + "step": 40394 + }, + { + "epoch": 2.6, + "learning_rate": 1.1353184911655469e-05, + "loss": 0.9399, + "step": 40395 + }, + { + "epoch": 2.6, + "learning_rate": 1.1352970222632518e-05, + "loss": 0.8497, + "step": 40396 + }, + { + "epoch": 2.6, + "learning_rate": 1.1352755533609567e-05, + "loss": 0.8937, + "step": 40397 + }, + { + "epoch": 2.6, + "learning_rate": 1.1352540844586617e-05, + "loss": 0.9008, + "step": 40398 + }, + { + "epoch": 2.6, + "learning_rate": 1.1352326155563666e-05, + "loss": 0.8578, + "step": 40399 + }, + { + "epoch": 2.6, + "learning_rate": 1.1352111466540715e-05, + "loss": 0.8531, + "step": 40400 + }, + { + "epoch": 2.6, + "learning_rate": 1.1351896777517766e-05, + "loss": 0.8952, + "step": 40401 + }, + { + "epoch": 2.6, + "learning_rate": 1.1351682088494815e-05, + "loss": 0.9293, + "step": 40402 + }, + { + "epoch": 2.6, + "learning_rate": 1.1351467399471866e-05, + "loss": 0.9014, + "step": 40403 + }, + { + "epoch": 2.6, + "learning_rate": 1.1351252710448916e-05, + "loss": 0.9148, + "step": 40404 + }, + { + "epoch": 2.6, + "learning_rate": 1.1351038021425967e-05, + "loss": 0.8593, + "step": 40405 + }, + { + "epoch": 2.6, + "learning_rate": 1.1350823332403016e-05, + "loss": 0.899, + "step": 40406 + }, + { + "epoch": 2.6, + "learning_rate": 1.1350608643380065e-05, + "loss": 0.843, + "step": 40407 + }, + { + "epoch": 2.6, + "learning_rate": 1.1350393954357116e-05, + "loss": 0.9382, + "step": 40408 + }, + { + "epoch": 2.6, + "learning_rate": 1.1350179265334164e-05, + "loss": 1.0185, + "step": 40409 + }, + { + "epoch": 2.6, + "learning_rate": 1.1349964576311213e-05, + "loss": 0.9038, + "step": 40410 + }, + { + "epoch": 2.6, + "learning_rate": 1.1349749887288264e-05, + "loss": 0.8792, + "step": 40411 + }, + { + "epoch": 2.6, + "learning_rate": 1.1349535198265313e-05, + "loss": 0.8937, + "step": 40412 + }, + { + "epoch": 2.6, + "learning_rate": 1.1349320509242364e-05, + "loss": 0.9939, + "step": 40413 + }, + { + "epoch": 2.6, + "learning_rate": 1.1349105820219413e-05, + "loss": 0.9176, + "step": 40414 + }, + { + "epoch": 2.6, + "learning_rate": 1.1348891131196462e-05, + "loss": 0.8841, + "step": 40415 + }, + { + "epoch": 2.6, + "learning_rate": 1.1348676442173512e-05, + "loss": 0.933, + "step": 40416 + }, + { + "epoch": 2.6, + "learning_rate": 1.1348461753150561e-05, + "loss": 0.935, + "step": 40417 + }, + { + "epoch": 2.6, + "learning_rate": 1.134824706412761e-05, + "loss": 0.8815, + "step": 40418 + }, + { + "epoch": 2.6, + "learning_rate": 1.1348032375104663e-05, + "loss": 0.874, + "step": 40419 + }, + { + "epoch": 2.6, + "learning_rate": 1.1347817686081712e-05, + "loss": 0.8975, + "step": 40420 + }, + { + "epoch": 2.6, + "learning_rate": 1.1347602997058762e-05, + "loss": 0.9336, + "step": 40421 + }, + { + "epoch": 2.6, + "learning_rate": 1.1347388308035811e-05, + "loss": 0.9172, + "step": 40422 + }, + { + "epoch": 2.6, + "learning_rate": 1.1347173619012862e-05, + "loss": 0.9232, + "step": 40423 + }, + { + "epoch": 2.6, + "learning_rate": 1.134695892998991e-05, + "loss": 0.8794, + "step": 40424 + }, + { + "epoch": 2.6, + "learning_rate": 1.134674424096696e-05, + "loss": 0.873, + "step": 40425 + }, + { + "epoch": 2.6, + "learning_rate": 1.134652955194401e-05, + "loss": 0.8812, + "step": 40426 + }, + { + "epoch": 2.6, + "learning_rate": 1.134631486292106e-05, + "loss": 0.8942, + "step": 40427 + }, + { + "epoch": 2.6, + "learning_rate": 1.1346100173898108e-05, + "loss": 0.9669, + "step": 40428 + }, + { + "epoch": 2.6, + "learning_rate": 1.1345885484875159e-05, + "loss": 0.8428, + "step": 40429 + }, + { + "epoch": 2.6, + "learning_rate": 1.1345670795852208e-05, + "loss": 0.9842, + "step": 40430 + }, + { + "epoch": 2.6, + "learning_rate": 1.1345456106829259e-05, + "loss": 0.9084, + "step": 40431 + }, + { + "epoch": 2.6, + "learning_rate": 1.1345241417806308e-05, + "loss": 0.8804, + "step": 40432 + }, + { + "epoch": 2.6, + "learning_rate": 1.134502672878336e-05, + "loss": 0.8771, + "step": 40433 + }, + { + "epoch": 2.6, + "learning_rate": 1.1344812039760409e-05, + "loss": 0.8613, + "step": 40434 + }, + { + "epoch": 2.6, + "learning_rate": 1.1344597350737458e-05, + "loss": 0.9565, + "step": 40435 + }, + { + "epoch": 2.6, + "learning_rate": 1.1344382661714509e-05, + "loss": 0.8497, + "step": 40436 + }, + { + "epoch": 2.6, + "learning_rate": 1.1344167972691558e-05, + "loss": 0.9351, + "step": 40437 + }, + { + "epoch": 2.6, + "learning_rate": 1.1343953283668606e-05, + "loss": 0.8713, + "step": 40438 + }, + { + "epoch": 2.6, + "learning_rate": 1.1343738594645657e-05, + "loss": 0.9015, + "step": 40439 + }, + { + "epoch": 2.6, + "learning_rate": 1.1343523905622706e-05, + "loss": 0.8538, + "step": 40440 + }, + { + "epoch": 2.6, + "learning_rate": 1.1343309216599757e-05, + "loss": 0.9389, + "step": 40441 + }, + { + "epoch": 2.6, + "learning_rate": 1.1343094527576806e-05, + "loss": 0.922, + "step": 40442 + }, + { + "epoch": 2.6, + "learning_rate": 1.1342879838553855e-05, + "loss": 0.8889, + "step": 40443 + }, + { + "epoch": 2.6, + "learning_rate": 1.1342665149530905e-05, + "loss": 0.8374, + "step": 40444 + }, + { + "epoch": 2.6, + "learning_rate": 1.1342450460507954e-05, + "loss": 0.929, + "step": 40445 + }, + { + "epoch": 2.6, + "learning_rate": 1.1342235771485003e-05, + "loss": 0.915, + "step": 40446 + }, + { + "epoch": 2.6, + "learning_rate": 1.1342021082462054e-05, + "loss": 0.8589, + "step": 40447 + }, + { + "epoch": 2.6, + "learning_rate": 1.1341806393439105e-05, + "loss": 0.8579, + "step": 40448 + }, + { + "epoch": 2.6, + "learning_rate": 1.1341591704416155e-05, + "loss": 0.8662, + "step": 40449 + }, + { + "epoch": 2.6, + "learning_rate": 1.1341377015393204e-05, + "loss": 0.894, + "step": 40450 + }, + { + "epoch": 2.6, + "learning_rate": 1.1341162326370255e-05, + "loss": 0.9334, + "step": 40451 + }, + { + "epoch": 2.6, + "learning_rate": 1.1340947637347304e-05, + "loss": 0.864, + "step": 40452 + }, + { + "epoch": 2.6, + "learning_rate": 1.1340732948324353e-05, + "loss": 0.9362, + "step": 40453 + }, + { + "epoch": 2.6, + "learning_rate": 1.1340518259301403e-05, + "loss": 0.9887, + "step": 40454 + }, + { + "epoch": 2.6, + "learning_rate": 1.1340303570278452e-05, + "loss": 0.9286, + "step": 40455 + }, + { + "epoch": 2.6, + "learning_rate": 1.1340088881255501e-05, + "loss": 0.9522, + "step": 40456 + }, + { + "epoch": 2.6, + "learning_rate": 1.1339874192232552e-05, + "loss": 0.8495, + "step": 40457 + }, + { + "epoch": 2.6, + "learning_rate": 1.1339659503209601e-05, + "loss": 0.8927, + "step": 40458 + }, + { + "epoch": 2.6, + "learning_rate": 1.1339444814186652e-05, + "loss": 0.9329, + "step": 40459 + }, + { + "epoch": 2.6, + "learning_rate": 1.13392301251637e-05, + "loss": 0.9266, + "step": 40460 + }, + { + "epoch": 2.6, + "learning_rate": 1.133901543614075e-05, + "loss": 0.8656, + "step": 40461 + }, + { + "epoch": 2.6, + "learning_rate": 1.1338800747117802e-05, + "loss": 0.9004, + "step": 40462 + }, + { + "epoch": 2.6, + "learning_rate": 1.1338586058094851e-05, + "loss": 0.9165, + "step": 40463 + }, + { + "epoch": 2.6, + "learning_rate": 1.1338371369071902e-05, + "loss": 0.9416, + "step": 40464 + }, + { + "epoch": 2.6, + "learning_rate": 1.133815668004895e-05, + "loss": 0.9062, + "step": 40465 + }, + { + "epoch": 2.6, + "learning_rate": 1.1337941991026e-05, + "loss": 0.9121, + "step": 40466 + }, + { + "epoch": 2.6, + "learning_rate": 1.133772730200305e-05, + "loss": 0.9417, + "step": 40467 + }, + { + "epoch": 2.6, + "learning_rate": 1.13375126129801e-05, + "loss": 0.8836, + "step": 40468 + }, + { + "epoch": 2.6, + "learning_rate": 1.133729792395715e-05, + "loss": 0.8347, + "step": 40469 + }, + { + "epoch": 2.6, + "learning_rate": 1.1337083234934199e-05, + "loss": 0.9496, + "step": 40470 + }, + { + "epoch": 2.6, + "learning_rate": 1.1336868545911248e-05, + "loss": 0.9927, + "step": 40471 + }, + { + "epoch": 2.6, + "learning_rate": 1.1336653856888298e-05, + "loss": 0.8524, + "step": 40472 + }, + { + "epoch": 2.6, + "learning_rate": 1.1336439167865347e-05, + "loss": 0.9282, + "step": 40473 + }, + { + "epoch": 2.6, + "learning_rate": 1.1336224478842396e-05, + "loss": 0.871, + "step": 40474 + }, + { + "epoch": 2.6, + "learning_rate": 1.1336009789819447e-05, + "loss": 0.8939, + "step": 40475 + }, + { + "epoch": 2.6, + "learning_rate": 1.1335795100796496e-05, + "loss": 0.9434, + "step": 40476 + }, + { + "epoch": 2.6, + "learning_rate": 1.1335580411773548e-05, + "loss": 0.8753, + "step": 40477 + }, + { + "epoch": 2.6, + "learning_rate": 1.1335365722750597e-05, + "loss": 0.8595, + "step": 40478 + }, + { + "epoch": 2.6, + "learning_rate": 1.1335151033727648e-05, + "loss": 0.9899, + "step": 40479 + }, + { + "epoch": 2.6, + "learning_rate": 1.1334936344704697e-05, + "loss": 0.8923, + "step": 40480 + }, + { + "epoch": 2.6, + "learning_rate": 1.1334721655681746e-05, + "loss": 0.9238, + "step": 40481 + }, + { + "epoch": 2.6, + "learning_rate": 1.1334506966658797e-05, + "loss": 0.9135, + "step": 40482 + }, + { + "epoch": 2.6, + "learning_rate": 1.1334292277635845e-05, + "loss": 0.9089, + "step": 40483 + }, + { + "epoch": 2.6, + "learning_rate": 1.1334077588612894e-05, + "loss": 0.8946, + "step": 40484 + }, + { + "epoch": 2.6, + "learning_rate": 1.1333862899589945e-05, + "loss": 0.8617, + "step": 40485 + }, + { + "epoch": 2.6, + "learning_rate": 1.1333648210566994e-05, + "loss": 0.848, + "step": 40486 + }, + { + "epoch": 2.6, + "learning_rate": 1.1333433521544043e-05, + "loss": 0.8999, + "step": 40487 + }, + { + "epoch": 2.6, + "learning_rate": 1.1333218832521094e-05, + "loss": 0.8595, + "step": 40488 + }, + { + "epoch": 2.6, + "learning_rate": 1.1333004143498143e-05, + "loss": 0.8655, + "step": 40489 + }, + { + "epoch": 2.61, + "learning_rate": 1.1332789454475193e-05, + "loss": 0.8533, + "step": 40490 + }, + { + "epoch": 2.61, + "learning_rate": 1.1332574765452244e-05, + "loss": 0.8338, + "step": 40491 + }, + { + "epoch": 2.61, + "learning_rate": 1.1332360076429295e-05, + "loss": 0.8875, + "step": 40492 + }, + { + "epoch": 2.61, + "learning_rate": 1.1332145387406344e-05, + "loss": 0.8628, + "step": 40493 + }, + { + "epoch": 2.61, + "learning_rate": 1.1331930698383393e-05, + "loss": 0.9534, + "step": 40494 + }, + { + "epoch": 2.61, + "learning_rate": 1.1331716009360443e-05, + "loss": 0.9005, + "step": 40495 + }, + { + "epoch": 2.61, + "learning_rate": 1.1331501320337492e-05, + "loss": 0.862, + "step": 40496 + }, + { + "epoch": 2.61, + "learning_rate": 1.1331286631314541e-05, + "loss": 0.9098, + "step": 40497 + }, + { + "epoch": 2.61, + "learning_rate": 1.1331071942291592e-05, + "loss": 0.8688, + "step": 40498 + }, + { + "epoch": 2.61, + "learning_rate": 1.133085725326864e-05, + "loss": 0.8016, + "step": 40499 + }, + { + "epoch": 2.61, + "learning_rate": 1.1330642564245691e-05, + "loss": 0.9133, + "step": 40500 + }, + { + "epoch": 2.61, + "learning_rate": 1.133042787522274e-05, + "loss": 0.9023, + "step": 40501 + }, + { + "epoch": 2.61, + "learning_rate": 1.133021318619979e-05, + "loss": 0.9566, + "step": 40502 + }, + { + "epoch": 2.61, + "learning_rate": 1.132999849717684e-05, + "loss": 0.8804, + "step": 40503 + }, + { + "epoch": 2.61, + "learning_rate": 1.1329783808153889e-05, + "loss": 0.9158, + "step": 40504 + }, + { + "epoch": 2.61, + "learning_rate": 1.1329569119130938e-05, + "loss": 0.9111, + "step": 40505 + }, + { + "epoch": 2.61, + "learning_rate": 1.132935443010799e-05, + "loss": 0.8993, + "step": 40506 + }, + { + "epoch": 2.61, + "learning_rate": 1.132913974108504e-05, + "loss": 0.8848, + "step": 40507 + }, + { + "epoch": 2.61, + "learning_rate": 1.132892505206209e-05, + "loss": 0.9091, + "step": 40508 + }, + { + "epoch": 2.61, + "learning_rate": 1.1328710363039139e-05, + "loss": 0.8865, + "step": 40509 + }, + { + "epoch": 2.61, + "learning_rate": 1.132849567401619e-05, + "loss": 0.913, + "step": 40510 + }, + { + "epoch": 2.61, + "learning_rate": 1.1328280984993239e-05, + "loss": 0.8786, + "step": 40511 + }, + { + "epoch": 2.61, + "learning_rate": 1.1328066295970288e-05, + "loss": 0.8335, + "step": 40512 + }, + { + "epoch": 2.61, + "learning_rate": 1.1327851606947338e-05, + "loss": 0.9007, + "step": 40513 + }, + { + "epoch": 2.61, + "learning_rate": 1.1327636917924387e-05, + "loss": 0.9474, + "step": 40514 + }, + { + "epoch": 2.61, + "learning_rate": 1.1327422228901436e-05, + "loss": 0.9273, + "step": 40515 + }, + { + "epoch": 2.61, + "learning_rate": 1.1327207539878487e-05, + "loss": 0.829, + "step": 40516 + }, + { + "epoch": 2.61, + "learning_rate": 1.1326992850855536e-05, + "loss": 0.9199, + "step": 40517 + }, + { + "epoch": 2.61, + "learning_rate": 1.1326778161832586e-05, + "loss": 0.862, + "step": 40518 + }, + { + "epoch": 2.61, + "learning_rate": 1.1326563472809635e-05, + "loss": 0.8972, + "step": 40519 + }, + { + "epoch": 2.61, + "learning_rate": 1.1326348783786688e-05, + "loss": 0.9428, + "step": 40520 + }, + { + "epoch": 2.61, + "learning_rate": 1.1326134094763737e-05, + "loss": 0.928, + "step": 40521 + }, + { + "epoch": 2.61, + "learning_rate": 1.1325919405740786e-05, + "loss": 0.898, + "step": 40522 + }, + { + "epoch": 2.61, + "learning_rate": 1.1325704716717836e-05, + "loss": 0.864, + "step": 40523 + }, + { + "epoch": 2.61, + "learning_rate": 1.1325490027694885e-05, + "loss": 0.9051, + "step": 40524 + }, + { + "epoch": 2.61, + "learning_rate": 1.1325275338671934e-05, + "loss": 0.8785, + "step": 40525 + }, + { + "epoch": 2.61, + "learning_rate": 1.1325060649648985e-05, + "loss": 0.9091, + "step": 40526 + }, + { + "epoch": 2.61, + "learning_rate": 1.1324845960626034e-05, + "loss": 0.931, + "step": 40527 + }, + { + "epoch": 2.61, + "learning_rate": 1.1324631271603084e-05, + "loss": 1.0182, + "step": 40528 + }, + { + "epoch": 2.61, + "learning_rate": 1.1324416582580133e-05, + "loss": 0.867, + "step": 40529 + }, + { + "epoch": 2.61, + "learning_rate": 1.1324201893557182e-05, + "loss": 0.9207, + "step": 40530 + }, + { + "epoch": 2.61, + "learning_rate": 1.1323987204534233e-05, + "loss": 0.9007, + "step": 40531 + }, + { + "epoch": 2.61, + "learning_rate": 1.1323772515511282e-05, + "loss": 0.9478, + "step": 40532 + }, + { + "epoch": 2.61, + "learning_rate": 1.1323557826488331e-05, + "loss": 0.8964, + "step": 40533 + }, + { + "epoch": 2.61, + "learning_rate": 1.1323343137465382e-05, + "loss": 0.873, + "step": 40534 + }, + { + "epoch": 2.61, + "learning_rate": 1.1323128448442432e-05, + "loss": 0.8847, + "step": 40535 + }, + { + "epoch": 2.61, + "learning_rate": 1.1322913759419483e-05, + "loss": 0.9688, + "step": 40536 + }, + { + "epoch": 2.61, + "learning_rate": 1.1322699070396532e-05, + "loss": 0.8904, + "step": 40537 + }, + { + "epoch": 2.61, + "learning_rate": 1.1322484381373583e-05, + "loss": 0.8509, + "step": 40538 + }, + { + "epoch": 2.61, + "learning_rate": 1.1322269692350632e-05, + "loss": 0.8821, + "step": 40539 + }, + { + "epoch": 2.61, + "learning_rate": 1.132205500332768e-05, + "loss": 0.804, + "step": 40540 + }, + { + "epoch": 2.61, + "learning_rate": 1.1321840314304731e-05, + "loss": 0.8466, + "step": 40541 + }, + { + "epoch": 2.61, + "learning_rate": 1.132162562528178e-05, + "loss": 0.8791, + "step": 40542 + }, + { + "epoch": 2.61, + "learning_rate": 1.1321410936258829e-05, + "loss": 0.8836, + "step": 40543 + }, + { + "epoch": 2.61, + "learning_rate": 1.132119624723588e-05, + "loss": 0.8797, + "step": 40544 + }, + { + "epoch": 2.61, + "learning_rate": 1.1320981558212929e-05, + "loss": 0.8658, + "step": 40545 + }, + { + "epoch": 2.61, + "learning_rate": 1.132076686918998e-05, + "loss": 0.916, + "step": 40546 + }, + { + "epoch": 2.61, + "learning_rate": 1.1320552180167028e-05, + "loss": 0.9187, + "step": 40547 + }, + { + "epoch": 2.61, + "learning_rate": 1.1320337491144077e-05, + "loss": 0.917, + "step": 40548 + }, + { + "epoch": 2.61, + "learning_rate": 1.132012280212113e-05, + "loss": 0.9052, + "step": 40549 + }, + { + "epoch": 2.61, + "learning_rate": 1.1319908113098179e-05, + "loss": 0.8882, + "step": 40550 + }, + { + "epoch": 2.61, + "learning_rate": 1.131969342407523e-05, + "loss": 0.9562, + "step": 40551 + }, + { + "epoch": 2.61, + "learning_rate": 1.1319478735052278e-05, + "loss": 0.9378, + "step": 40552 + }, + { + "epoch": 2.61, + "learning_rate": 1.1319264046029327e-05, + "loss": 0.8634, + "step": 40553 + }, + { + "epoch": 2.61, + "learning_rate": 1.1319049357006378e-05, + "loss": 0.9317, + "step": 40554 + }, + { + "epoch": 2.61, + "learning_rate": 1.1319049357006378e-05, + "loss": 0.8967, + "step": 40555 + }, + { + "epoch": 2.61, + "learning_rate": 1.1318834667983427e-05, + "loss": 0.9007, + "step": 40556 + }, + { + "epoch": 2.61, + "learning_rate": 1.1318619978960478e-05, + "loss": 0.9471, + "step": 40557 + }, + { + "epoch": 2.61, + "learning_rate": 1.1318405289937526e-05, + "loss": 0.9677, + "step": 40558 + }, + { + "epoch": 2.61, + "learning_rate": 1.1318190600914575e-05, + "loss": 0.8949, + "step": 40559 + }, + { + "epoch": 2.61, + "learning_rate": 1.1317975911891626e-05, + "loss": 0.8737, + "step": 40560 + }, + { + "epoch": 2.61, + "learning_rate": 1.1317761222868675e-05, + "loss": 0.8731, + "step": 40561 + }, + { + "epoch": 2.61, + "learning_rate": 1.1317546533845724e-05, + "loss": 0.9372, + "step": 40562 + }, + { + "epoch": 2.61, + "learning_rate": 1.1317331844822775e-05, + "loss": 0.9613, + "step": 40563 + }, + { + "epoch": 2.61, + "learning_rate": 1.1317117155799824e-05, + "loss": 0.9069, + "step": 40564 + }, + { + "epoch": 2.61, + "learning_rate": 1.1316902466776876e-05, + "loss": 0.9446, + "step": 40565 + }, + { + "epoch": 2.61, + "learning_rate": 1.1316687777753925e-05, + "loss": 0.8833, + "step": 40566 + }, + { + "epoch": 2.61, + "learning_rate": 1.1316473088730976e-05, + "loss": 0.9492, + "step": 40567 + }, + { + "epoch": 2.61, + "learning_rate": 1.1316258399708025e-05, + "loss": 0.8712, + "step": 40568 + }, + { + "epoch": 2.61, + "learning_rate": 1.1316043710685074e-05, + "loss": 0.8823, + "step": 40569 + }, + { + "epoch": 2.61, + "learning_rate": 1.1315829021662124e-05, + "loss": 0.8667, + "step": 40570 + }, + { + "epoch": 2.61, + "learning_rate": 1.1315614332639173e-05, + "loss": 0.9158, + "step": 40571 + }, + { + "epoch": 2.61, + "learning_rate": 1.1315399643616222e-05, + "loss": 0.9015, + "step": 40572 + }, + { + "epoch": 2.61, + "learning_rate": 1.1315184954593273e-05, + "loss": 0.8418, + "step": 40573 + }, + { + "epoch": 2.61, + "learning_rate": 1.1314970265570322e-05, + "loss": 0.9387, + "step": 40574 + }, + { + "epoch": 2.61, + "learning_rate": 1.131475557654737e-05, + "loss": 0.8928, + "step": 40575 + }, + { + "epoch": 2.61, + "learning_rate": 1.1314540887524421e-05, + "loss": 0.9292, + "step": 40576 + }, + { + "epoch": 2.61, + "learning_rate": 1.131432619850147e-05, + "loss": 0.9376, + "step": 40577 + }, + { + "epoch": 2.61, + "learning_rate": 1.1314111509478521e-05, + "loss": 0.8896, + "step": 40578 + }, + { + "epoch": 2.61, + "learning_rate": 1.1313896820455572e-05, + "loss": 0.8824, + "step": 40579 + }, + { + "epoch": 2.61, + "learning_rate": 1.1313682131432622e-05, + "loss": 0.8873, + "step": 40580 + }, + { + "epoch": 2.61, + "learning_rate": 1.1313467442409671e-05, + "loss": 0.9041, + "step": 40581 + }, + { + "epoch": 2.61, + "learning_rate": 1.131325275338672e-05, + "loss": 0.9044, + "step": 40582 + }, + { + "epoch": 2.61, + "learning_rate": 1.1313038064363771e-05, + "loss": 0.836, + "step": 40583 + }, + { + "epoch": 2.61, + "learning_rate": 1.131282337534082e-05, + "loss": 0.8837, + "step": 40584 + }, + { + "epoch": 2.61, + "learning_rate": 1.1312608686317869e-05, + "loss": 0.8687, + "step": 40585 + }, + { + "epoch": 2.61, + "learning_rate": 1.131239399729492e-05, + "loss": 0.8704, + "step": 40586 + }, + { + "epoch": 2.61, + "learning_rate": 1.1312179308271969e-05, + "loss": 0.9414, + "step": 40587 + }, + { + "epoch": 2.61, + "learning_rate": 1.131196461924902e-05, + "loss": 0.8801, + "step": 40588 + }, + { + "epoch": 2.61, + "learning_rate": 1.1311749930226068e-05, + "loss": 0.8489, + "step": 40589 + }, + { + "epoch": 2.61, + "learning_rate": 1.1311535241203117e-05, + "loss": 0.8228, + "step": 40590 + }, + { + "epoch": 2.61, + "learning_rate": 1.1311320552180168e-05, + "loss": 0.7912, + "step": 40591 + }, + { + "epoch": 2.61, + "learning_rate": 1.1311105863157217e-05, + "loss": 0.922, + "step": 40592 + }, + { + "epoch": 2.61, + "learning_rate": 1.1310891174134266e-05, + "loss": 0.8487, + "step": 40593 + }, + { + "epoch": 2.61, + "learning_rate": 1.1310676485111318e-05, + "loss": 0.9375, + "step": 40594 + }, + { + "epoch": 2.61, + "learning_rate": 1.1310461796088367e-05, + "loss": 0.8344, + "step": 40595 + }, + { + "epoch": 2.61, + "learning_rate": 1.1310247107065418e-05, + "loss": 0.902, + "step": 40596 + }, + { + "epoch": 2.61, + "learning_rate": 1.1310032418042467e-05, + "loss": 0.8359, + "step": 40597 + }, + { + "epoch": 2.61, + "learning_rate": 1.1309817729019517e-05, + "loss": 0.8176, + "step": 40598 + }, + { + "epoch": 2.61, + "learning_rate": 1.1309603039996566e-05, + "loss": 0.9324, + "step": 40599 + }, + { + "epoch": 2.61, + "learning_rate": 1.1309388350973615e-05, + "loss": 0.8614, + "step": 40600 + }, + { + "epoch": 2.61, + "learning_rate": 1.1309173661950666e-05, + "loss": 0.9016, + "step": 40601 + }, + { + "epoch": 2.61, + "learning_rate": 1.1308958972927715e-05, + "loss": 0.8875, + "step": 40602 + }, + { + "epoch": 2.61, + "learning_rate": 1.1308744283904764e-05, + "loss": 0.8655, + "step": 40603 + }, + { + "epoch": 2.61, + "learning_rate": 1.1308529594881814e-05, + "loss": 0.907, + "step": 40604 + }, + { + "epoch": 2.61, + "learning_rate": 1.1308314905858863e-05, + "loss": 0.8992, + "step": 40605 + }, + { + "epoch": 2.61, + "learning_rate": 1.1308100216835914e-05, + "loss": 0.888, + "step": 40606 + }, + { + "epoch": 2.61, + "learning_rate": 1.1307885527812963e-05, + "loss": 0.9386, + "step": 40607 + }, + { + "epoch": 2.61, + "learning_rate": 1.1307670838790012e-05, + "loss": 0.8809, + "step": 40608 + }, + { + "epoch": 2.61, + "learning_rate": 1.1307456149767064e-05, + "loss": 0.9423, + "step": 40609 + }, + { + "epoch": 2.61, + "learning_rate": 1.1307241460744113e-05, + "loss": 0.9104, + "step": 40610 + }, + { + "epoch": 2.61, + "learning_rate": 1.1307026771721164e-05, + "loss": 0.8697, + "step": 40611 + }, + { + "epoch": 2.61, + "learning_rate": 1.1306812082698213e-05, + "loss": 0.8911, + "step": 40612 + }, + { + "epoch": 2.61, + "learning_rate": 1.1306597393675262e-05, + "loss": 0.8964, + "step": 40613 + }, + { + "epoch": 2.61, + "learning_rate": 1.1306382704652313e-05, + "loss": 0.9226, + "step": 40614 + }, + { + "epoch": 2.61, + "learning_rate": 1.1306168015629362e-05, + "loss": 0.865, + "step": 40615 + }, + { + "epoch": 2.61, + "learning_rate": 1.1305953326606412e-05, + "loss": 0.829, + "step": 40616 + }, + { + "epoch": 2.61, + "learning_rate": 1.1305738637583461e-05, + "loss": 0.881, + "step": 40617 + }, + { + "epoch": 2.61, + "learning_rate": 1.130552394856051e-05, + "loss": 0.9176, + "step": 40618 + }, + { + "epoch": 2.61, + "learning_rate": 1.130530925953756e-05, + "loss": 0.8791, + "step": 40619 + }, + { + "epoch": 2.61, + "learning_rate": 1.130509457051461e-05, + "loss": 0.9173, + "step": 40620 + }, + { + "epoch": 2.61, + "learning_rate": 1.1304879881491659e-05, + "loss": 0.8699, + "step": 40621 + }, + { + "epoch": 2.61, + "learning_rate": 1.130466519246871e-05, + "loss": 0.9234, + "step": 40622 + }, + { + "epoch": 2.61, + "learning_rate": 1.130445050344576e-05, + "loss": 0.8536, + "step": 40623 + }, + { + "epoch": 2.61, + "learning_rate": 1.130423581442281e-05, + "loss": 0.926, + "step": 40624 + }, + { + "epoch": 2.61, + "learning_rate": 1.130402112539986e-05, + "loss": 0.8692, + "step": 40625 + }, + { + "epoch": 2.61, + "learning_rate": 1.130380643637691e-05, + "loss": 0.856, + "step": 40626 + }, + { + "epoch": 2.61, + "learning_rate": 1.130359174735396e-05, + "loss": 1.0003, + "step": 40627 + }, + { + "epoch": 2.61, + "learning_rate": 1.1303377058331008e-05, + "loss": 0.9092, + "step": 40628 + }, + { + "epoch": 2.61, + "learning_rate": 1.1303162369308059e-05, + "loss": 0.7984, + "step": 40629 + }, + { + "epoch": 2.61, + "learning_rate": 1.1302947680285108e-05, + "loss": 0.8653, + "step": 40630 + }, + { + "epoch": 2.61, + "learning_rate": 1.1302732991262157e-05, + "loss": 0.9575, + "step": 40631 + }, + { + "epoch": 2.61, + "learning_rate": 1.1302518302239208e-05, + "loss": 0.9126, + "step": 40632 + }, + { + "epoch": 2.61, + "learning_rate": 1.1302303613216256e-05, + "loss": 0.9712, + "step": 40633 + }, + { + "epoch": 2.61, + "learning_rate": 1.1302088924193305e-05, + "loss": 0.8932, + "step": 40634 + }, + { + "epoch": 2.61, + "learning_rate": 1.1301874235170356e-05, + "loss": 0.8792, + "step": 40635 + }, + { + "epoch": 2.61, + "learning_rate": 1.1301659546147405e-05, + "loss": 0.9427, + "step": 40636 + }, + { + "epoch": 2.61, + "learning_rate": 1.1301444857124456e-05, + "loss": 0.8806, + "step": 40637 + }, + { + "epoch": 2.61, + "learning_rate": 1.1301230168101506e-05, + "loss": 0.9057, + "step": 40638 + }, + { + "epoch": 2.61, + "learning_rate": 1.1301015479078557e-05, + "loss": 0.8972, + "step": 40639 + }, + { + "epoch": 2.61, + "learning_rate": 1.1300800790055606e-05, + "loss": 0.881, + "step": 40640 + }, + { + "epoch": 2.61, + "learning_rate": 1.1300586101032655e-05, + "loss": 0.8601, + "step": 40641 + }, + { + "epoch": 2.61, + "learning_rate": 1.1300371412009706e-05, + "loss": 0.9293, + "step": 40642 + }, + { + "epoch": 2.61, + "learning_rate": 1.1300156722986755e-05, + "loss": 0.8795, + "step": 40643 + }, + { + "epoch": 2.61, + "learning_rate": 1.1299942033963805e-05, + "loss": 0.8573, + "step": 40644 + }, + { + "epoch": 2.62, + "learning_rate": 1.1299727344940854e-05, + "loss": 0.9517, + "step": 40645 + }, + { + "epoch": 2.62, + "learning_rate": 1.1299512655917903e-05, + "loss": 0.8192, + "step": 40646 + }, + { + "epoch": 2.62, + "learning_rate": 1.1299297966894954e-05, + "loss": 0.922, + "step": 40647 + }, + { + "epoch": 2.62, + "learning_rate": 1.1299083277872003e-05, + "loss": 0.8981, + "step": 40648 + }, + { + "epoch": 2.62, + "learning_rate": 1.1298868588849052e-05, + "loss": 0.9706, + "step": 40649 + }, + { + "epoch": 2.62, + "learning_rate": 1.1298653899826102e-05, + "loss": 0.837, + "step": 40650 + }, + { + "epoch": 2.62, + "learning_rate": 1.1298439210803151e-05, + "loss": 0.9137, + "step": 40651 + }, + { + "epoch": 2.62, + "learning_rate": 1.1298224521780204e-05, + "loss": 0.9028, + "step": 40652 + }, + { + "epoch": 2.62, + "learning_rate": 1.1298009832757253e-05, + "loss": 0.9205, + "step": 40653 + }, + { + "epoch": 2.62, + "learning_rate": 1.1297795143734303e-05, + "loss": 0.9571, + "step": 40654 + }, + { + "epoch": 2.62, + "learning_rate": 1.1297580454711352e-05, + "loss": 0.8802, + "step": 40655 + }, + { + "epoch": 2.62, + "learning_rate": 1.1297365765688401e-05, + "loss": 0.8956, + "step": 40656 + }, + { + "epoch": 2.62, + "learning_rate": 1.1297151076665452e-05, + "loss": 0.8137, + "step": 40657 + }, + { + "epoch": 2.62, + "learning_rate": 1.1296936387642501e-05, + "loss": 0.887, + "step": 40658 + }, + { + "epoch": 2.62, + "learning_rate": 1.129672169861955e-05, + "loss": 0.9023, + "step": 40659 + }, + { + "epoch": 2.62, + "learning_rate": 1.12965070095966e-05, + "loss": 0.8991, + "step": 40660 + }, + { + "epoch": 2.62, + "learning_rate": 1.129629232057365e-05, + "loss": 0.7659, + "step": 40661 + }, + { + "epoch": 2.62, + "learning_rate": 1.1296077631550698e-05, + "loss": 0.8575, + "step": 40662 + }, + { + "epoch": 2.62, + "learning_rate": 1.1295862942527749e-05, + "loss": 0.8794, + "step": 40663 + }, + { + "epoch": 2.62, + "learning_rate": 1.1295648253504798e-05, + "loss": 0.8466, + "step": 40664 + }, + { + "epoch": 2.62, + "learning_rate": 1.1295433564481849e-05, + "loss": 0.9384, + "step": 40665 + }, + { + "epoch": 2.62, + "learning_rate": 1.1295218875458898e-05, + "loss": 0.8961, + "step": 40666 + }, + { + "epoch": 2.62, + "learning_rate": 1.129500418643595e-05, + "loss": 0.9323, + "step": 40667 + }, + { + "epoch": 2.62, + "learning_rate": 1.1294789497412999e-05, + "loss": 0.8119, + "step": 40668 + }, + { + "epoch": 2.62, + "learning_rate": 1.1294574808390048e-05, + "loss": 0.9594, + "step": 40669 + }, + { + "epoch": 2.62, + "learning_rate": 1.1294360119367099e-05, + "loss": 0.9087, + "step": 40670 + }, + { + "epoch": 2.62, + "learning_rate": 1.1294145430344148e-05, + "loss": 0.9517, + "step": 40671 + }, + { + "epoch": 2.62, + "learning_rate": 1.1293930741321197e-05, + "loss": 0.8345, + "step": 40672 + }, + { + "epoch": 2.62, + "learning_rate": 1.1293716052298247e-05, + "loss": 0.9194, + "step": 40673 + }, + { + "epoch": 2.62, + "learning_rate": 1.1293501363275296e-05, + "loss": 0.891, + "step": 40674 + }, + { + "epoch": 2.62, + "learning_rate": 1.1293286674252347e-05, + "loss": 0.9053, + "step": 40675 + }, + { + "epoch": 2.62, + "learning_rate": 1.1293071985229396e-05, + "loss": 0.8475, + "step": 40676 + }, + { + "epoch": 2.62, + "learning_rate": 1.1292857296206445e-05, + "loss": 0.8886, + "step": 40677 + }, + { + "epoch": 2.62, + "learning_rate": 1.1292642607183495e-05, + "loss": 0.8887, + "step": 40678 + }, + { + "epoch": 2.62, + "learning_rate": 1.1292427918160544e-05, + "loss": 0.8829, + "step": 40679 + }, + { + "epoch": 2.62, + "learning_rate": 1.1292213229137593e-05, + "loss": 0.8952, + "step": 40680 + }, + { + "epoch": 2.62, + "learning_rate": 1.1291998540114646e-05, + "loss": 0.8794, + "step": 40681 + }, + { + "epoch": 2.62, + "learning_rate": 1.1291783851091695e-05, + "loss": 0.918, + "step": 40682 + }, + { + "epoch": 2.62, + "learning_rate": 1.1291569162068745e-05, + "loss": 0.8839, + "step": 40683 + }, + { + "epoch": 2.62, + "learning_rate": 1.1291354473045794e-05, + "loss": 0.8813, + "step": 40684 + }, + { + "epoch": 2.62, + "learning_rate": 1.1291139784022845e-05, + "loss": 0.9055, + "step": 40685 + }, + { + "epoch": 2.62, + "learning_rate": 1.1290925094999894e-05, + "loss": 0.9178, + "step": 40686 + }, + { + "epoch": 2.62, + "learning_rate": 1.1290710405976943e-05, + "loss": 0.8652, + "step": 40687 + }, + { + "epoch": 2.62, + "learning_rate": 1.1290495716953994e-05, + "loss": 0.8787, + "step": 40688 + }, + { + "epoch": 2.62, + "learning_rate": 1.1290281027931043e-05, + "loss": 0.8201, + "step": 40689 + }, + { + "epoch": 2.62, + "learning_rate": 1.1290066338908092e-05, + "loss": 0.9191, + "step": 40690 + }, + { + "epoch": 2.62, + "learning_rate": 1.1289851649885142e-05, + "loss": 0.929, + "step": 40691 + }, + { + "epoch": 2.62, + "learning_rate": 1.1289636960862191e-05, + "loss": 0.8281, + "step": 40692 + }, + { + "epoch": 2.62, + "learning_rate": 1.1289422271839242e-05, + "loss": 0.9019, + "step": 40693 + }, + { + "epoch": 2.62, + "learning_rate": 1.128920758281629e-05, + "loss": 0.8894, + "step": 40694 + }, + { + "epoch": 2.62, + "learning_rate": 1.128899289379334e-05, + "loss": 0.8998, + "step": 40695 + }, + { + "epoch": 2.62, + "learning_rate": 1.1288778204770392e-05, + "loss": 0.8653, + "step": 40696 + }, + { + "epoch": 2.62, + "learning_rate": 1.1288563515747441e-05, + "loss": 0.8266, + "step": 40697 + }, + { + "epoch": 2.62, + "learning_rate": 1.1288348826724492e-05, + "loss": 0.8868, + "step": 40698 + }, + { + "epoch": 2.62, + "learning_rate": 1.128813413770154e-05, + "loss": 0.8663, + "step": 40699 + }, + { + "epoch": 2.62, + "learning_rate": 1.128791944867859e-05, + "loss": 0.8635, + "step": 40700 + }, + { + "epoch": 2.62, + "learning_rate": 1.128770475965564e-05, + "loss": 0.911, + "step": 40701 + }, + { + "epoch": 2.62, + "learning_rate": 1.128749007063269e-05, + "loss": 0.8689, + "step": 40702 + }, + { + "epoch": 2.62, + "learning_rate": 1.128727538160974e-05, + "loss": 0.9388, + "step": 40703 + }, + { + "epoch": 2.62, + "learning_rate": 1.1287060692586789e-05, + "loss": 0.958, + "step": 40704 + }, + { + "epoch": 2.62, + "learning_rate": 1.1286846003563838e-05, + "loss": 0.9559, + "step": 40705 + }, + { + "epoch": 2.62, + "learning_rate": 1.1286631314540889e-05, + "loss": 0.9157, + "step": 40706 + }, + { + "epoch": 2.62, + "learning_rate": 1.1286416625517937e-05, + "loss": 0.8974, + "step": 40707 + }, + { + "epoch": 2.62, + "learning_rate": 1.1286201936494986e-05, + "loss": 0.8933, + "step": 40708 + }, + { + "epoch": 2.62, + "learning_rate": 1.1285987247472037e-05, + "loss": 0.8535, + "step": 40709 + }, + { + "epoch": 2.62, + "learning_rate": 1.1285772558449088e-05, + "loss": 0.8624, + "step": 40710 + }, + { + "epoch": 2.62, + "learning_rate": 1.1285557869426138e-05, + "loss": 0.9053, + "step": 40711 + }, + { + "epoch": 2.62, + "learning_rate": 1.1285343180403187e-05, + "loss": 0.819, + "step": 40712 + }, + { + "epoch": 2.62, + "learning_rate": 1.1285128491380238e-05, + "loss": 0.8917, + "step": 40713 + }, + { + "epoch": 2.62, + "learning_rate": 1.1284913802357287e-05, + "loss": 0.9574, + "step": 40714 + }, + { + "epoch": 2.62, + "learning_rate": 1.1284699113334336e-05, + "loss": 0.9051, + "step": 40715 + }, + { + "epoch": 2.62, + "learning_rate": 1.1284484424311387e-05, + "loss": 0.8674, + "step": 40716 + }, + { + "epoch": 2.62, + "learning_rate": 1.1284269735288436e-05, + "loss": 0.8754, + "step": 40717 + }, + { + "epoch": 2.62, + "learning_rate": 1.1284055046265485e-05, + "loss": 0.9346, + "step": 40718 + }, + { + "epoch": 2.62, + "learning_rate": 1.1283840357242535e-05, + "loss": 0.9153, + "step": 40719 + }, + { + "epoch": 2.62, + "learning_rate": 1.1283625668219584e-05, + "loss": 0.9433, + "step": 40720 + }, + { + "epoch": 2.62, + "learning_rate": 1.1283410979196633e-05, + "loss": 0.9202, + "step": 40721 + }, + { + "epoch": 2.62, + "learning_rate": 1.1283196290173684e-05, + "loss": 0.9184, + "step": 40722 + }, + { + "epoch": 2.62, + "learning_rate": 1.1282981601150733e-05, + "loss": 0.8849, + "step": 40723 + }, + { + "epoch": 2.62, + "learning_rate": 1.1282766912127783e-05, + "loss": 0.9628, + "step": 40724 + }, + { + "epoch": 2.62, + "learning_rate": 1.1282552223104834e-05, + "loss": 0.879, + "step": 40725 + }, + { + "epoch": 2.62, + "learning_rate": 1.1282337534081885e-05, + "loss": 0.935, + "step": 40726 + }, + { + "epoch": 2.62, + "learning_rate": 1.1282122845058934e-05, + "loss": 0.9076, + "step": 40727 + }, + { + "epoch": 2.62, + "learning_rate": 1.1281908156035983e-05, + "loss": 0.909, + "step": 40728 + }, + { + "epoch": 2.62, + "learning_rate": 1.1281693467013033e-05, + "loss": 0.9368, + "step": 40729 + }, + { + "epoch": 2.62, + "learning_rate": 1.1281478777990082e-05, + "loss": 0.925, + "step": 40730 + }, + { + "epoch": 2.62, + "learning_rate": 1.1281264088967131e-05, + "loss": 0.9412, + "step": 40731 + }, + { + "epoch": 2.62, + "learning_rate": 1.1281049399944182e-05, + "loss": 0.8766, + "step": 40732 + }, + { + "epoch": 2.62, + "learning_rate": 1.1280834710921231e-05, + "loss": 0.8946, + "step": 40733 + }, + { + "epoch": 2.62, + "learning_rate": 1.1280620021898282e-05, + "loss": 0.8417, + "step": 40734 + }, + { + "epoch": 2.62, + "learning_rate": 1.128040533287533e-05, + "loss": 0.9106, + "step": 40735 + }, + { + "epoch": 2.62, + "learning_rate": 1.128019064385238e-05, + "loss": 0.9093, + "step": 40736 + }, + { + "epoch": 2.62, + "learning_rate": 1.127997595482943e-05, + "loss": 0.8661, + "step": 40737 + }, + { + "epoch": 2.62, + "learning_rate": 1.1279761265806479e-05, + "loss": 0.9299, + "step": 40738 + }, + { + "epoch": 2.62, + "learning_rate": 1.1279546576783531e-05, + "loss": 0.8811, + "step": 40739 + }, + { + "epoch": 2.62, + "learning_rate": 1.127933188776058e-05, + "loss": 0.8837, + "step": 40740 + }, + { + "epoch": 2.62, + "learning_rate": 1.127911719873763e-05, + "loss": 0.9028, + "step": 40741 + }, + { + "epoch": 2.62, + "learning_rate": 1.127890250971468e-05, + "loss": 0.8347, + "step": 40742 + }, + { + "epoch": 2.62, + "learning_rate": 1.1278687820691729e-05, + "loss": 0.8737, + "step": 40743 + }, + { + "epoch": 2.62, + "learning_rate": 1.127847313166878e-05, + "loss": 0.8436, + "step": 40744 + }, + { + "epoch": 2.62, + "learning_rate": 1.1278258442645829e-05, + "loss": 0.8745, + "step": 40745 + }, + { + "epoch": 2.62, + "learning_rate": 1.1278043753622878e-05, + "loss": 0.9294, + "step": 40746 + }, + { + "epoch": 2.62, + "learning_rate": 1.1277829064599928e-05, + "loss": 0.8815, + "step": 40747 + }, + { + "epoch": 2.62, + "learning_rate": 1.1277614375576977e-05, + "loss": 0.9333, + "step": 40748 + }, + { + "epoch": 2.62, + "learning_rate": 1.1277399686554026e-05, + "loss": 0.9355, + "step": 40749 + }, + { + "epoch": 2.62, + "learning_rate": 1.1277184997531077e-05, + "loss": 0.8139, + "step": 40750 + }, + { + "epoch": 2.62, + "learning_rate": 1.1276970308508126e-05, + "loss": 0.839, + "step": 40751 + }, + { + "epoch": 2.62, + "learning_rate": 1.1276755619485176e-05, + "loss": 0.9194, + "step": 40752 + }, + { + "epoch": 2.62, + "learning_rate": 1.1276540930462225e-05, + "loss": 0.8365, + "step": 40753 + }, + { + "epoch": 2.62, + "learning_rate": 1.1276326241439278e-05, + "loss": 0.9769, + "step": 40754 + }, + { + "epoch": 2.62, + "learning_rate": 1.1276111552416327e-05, + "loss": 0.9943, + "step": 40755 + }, + { + "epoch": 2.62, + "learning_rate": 1.1275896863393376e-05, + "loss": 0.9207, + "step": 40756 + }, + { + "epoch": 2.62, + "learning_rate": 1.1275682174370426e-05, + "loss": 0.8729, + "step": 40757 + }, + { + "epoch": 2.62, + "learning_rate": 1.1275467485347475e-05, + "loss": 0.8948, + "step": 40758 + }, + { + "epoch": 2.62, + "learning_rate": 1.1275252796324524e-05, + "loss": 0.8491, + "step": 40759 + }, + { + "epoch": 2.62, + "learning_rate": 1.1275038107301575e-05, + "loss": 0.8829, + "step": 40760 + }, + { + "epoch": 2.62, + "learning_rate": 1.1274823418278624e-05, + "loss": 0.9397, + "step": 40761 + }, + { + "epoch": 2.62, + "learning_rate": 1.1274608729255675e-05, + "loss": 0.8509, + "step": 40762 + }, + { + "epoch": 2.62, + "learning_rate": 1.1274394040232724e-05, + "loss": 0.8807, + "step": 40763 + }, + { + "epoch": 2.62, + "learning_rate": 1.1274179351209773e-05, + "loss": 0.8697, + "step": 40764 + }, + { + "epoch": 2.62, + "learning_rate": 1.1273964662186823e-05, + "loss": 0.8786, + "step": 40765 + }, + { + "epoch": 2.62, + "learning_rate": 1.1273749973163872e-05, + "loss": 0.9593, + "step": 40766 + }, + { + "epoch": 2.62, + "learning_rate": 1.1273535284140921e-05, + "loss": 0.8221, + "step": 40767 + }, + { + "epoch": 2.62, + "learning_rate": 1.1273320595117973e-05, + "loss": 0.9087, + "step": 40768 + }, + { + "epoch": 2.62, + "learning_rate": 1.1273105906095022e-05, + "loss": 0.8939, + "step": 40769 + }, + { + "epoch": 2.62, + "learning_rate": 1.1272891217072073e-05, + "loss": 0.955, + "step": 40770 + }, + { + "epoch": 2.62, + "learning_rate": 1.1272676528049122e-05, + "loss": 0.906, + "step": 40771 + }, + { + "epoch": 2.62, + "learning_rate": 1.1272461839026173e-05, + "loss": 0.9023, + "step": 40772 + }, + { + "epoch": 2.62, + "learning_rate": 1.1272247150003222e-05, + "loss": 0.9041, + "step": 40773 + }, + { + "epoch": 2.62, + "learning_rate": 1.127203246098027e-05, + "loss": 0.9067, + "step": 40774 + }, + { + "epoch": 2.62, + "learning_rate": 1.1271817771957321e-05, + "loss": 0.9716, + "step": 40775 + }, + { + "epoch": 2.62, + "learning_rate": 1.127160308293437e-05, + "loss": 0.9132, + "step": 40776 + }, + { + "epoch": 2.62, + "learning_rate": 1.127138839391142e-05, + "loss": 0.8745, + "step": 40777 + }, + { + "epoch": 2.62, + "learning_rate": 1.127117370488847e-05, + "loss": 0.8721, + "step": 40778 + }, + { + "epoch": 2.62, + "learning_rate": 1.1270959015865519e-05, + "loss": 0.9869, + "step": 40779 + }, + { + "epoch": 2.62, + "learning_rate": 1.127074432684257e-05, + "loss": 0.932, + "step": 40780 + }, + { + "epoch": 2.62, + "learning_rate": 1.1270529637819618e-05, + "loss": 0.8983, + "step": 40781 + }, + { + "epoch": 2.62, + "learning_rate": 1.1270314948796667e-05, + "loss": 0.9199, + "step": 40782 + }, + { + "epoch": 2.62, + "learning_rate": 1.127010025977372e-05, + "loss": 0.9393, + "step": 40783 + }, + { + "epoch": 2.62, + "learning_rate": 1.1269885570750769e-05, + "loss": 0.8526, + "step": 40784 + }, + { + "epoch": 2.62, + "learning_rate": 1.126967088172782e-05, + "loss": 0.888, + "step": 40785 + }, + { + "epoch": 2.62, + "learning_rate": 1.1269456192704868e-05, + "loss": 0.9373, + "step": 40786 + }, + { + "epoch": 2.62, + "learning_rate": 1.1269241503681917e-05, + "loss": 0.9198, + "step": 40787 + }, + { + "epoch": 2.62, + "learning_rate": 1.1269026814658968e-05, + "loss": 0.9375, + "step": 40788 + }, + { + "epoch": 2.62, + "learning_rate": 1.1268812125636017e-05, + "loss": 0.8332, + "step": 40789 + }, + { + "epoch": 2.62, + "learning_rate": 1.1268597436613068e-05, + "loss": 0.8547, + "step": 40790 + }, + { + "epoch": 2.62, + "learning_rate": 1.1268382747590117e-05, + "loss": 0.8585, + "step": 40791 + }, + { + "epoch": 2.62, + "learning_rate": 1.1268168058567166e-05, + "loss": 0.8737, + "step": 40792 + }, + { + "epoch": 2.62, + "learning_rate": 1.1267953369544216e-05, + "loss": 0.9271, + "step": 40793 + }, + { + "epoch": 2.62, + "learning_rate": 1.1267738680521265e-05, + "loss": 0.9402, + "step": 40794 + }, + { + "epoch": 2.62, + "learning_rate": 1.1267523991498314e-05, + "loss": 0.9618, + "step": 40795 + }, + { + "epoch": 2.62, + "learning_rate": 1.1267309302475365e-05, + "loss": 0.9445, + "step": 40796 + }, + { + "epoch": 2.62, + "learning_rate": 1.1267094613452415e-05, + "loss": 0.9229, + "step": 40797 + }, + { + "epoch": 2.62, + "learning_rate": 1.1266879924429466e-05, + "loss": 0.9757, + "step": 40798 + }, + { + "epoch": 2.62, + "learning_rate": 1.1266665235406515e-05, + "loss": 0.8615, + "step": 40799 + }, + { + "epoch": 2.62, + "learning_rate": 1.1266450546383566e-05, + "loss": 0.8534, + "step": 40800 + }, + { + "epoch": 2.63, + "learning_rate": 1.1266235857360615e-05, + "loss": 0.884, + "step": 40801 + }, + { + "epoch": 2.63, + "learning_rate": 1.1266021168337664e-05, + "loss": 0.8721, + "step": 40802 + }, + { + "epoch": 2.63, + "learning_rate": 1.1265806479314714e-05, + "loss": 0.9041, + "step": 40803 + }, + { + "epoch": 2.63, + "learning_rate": 1.1265591790291763e-05, + "loss": 0.8623, + "step": 40804 + }, + { + "epoch": 2.63, + "learning_rate": 1.1265377101268812e-05, + "loss": 0.8552, + "step": 40805 + }, + { + "epoch": 2.63, + "learning_rate": 1.1265162412245863e-05, + "loss": 0.8639, + "step": 40806 + }, + { + "epoch": 2.63, + "learning_rate": 1.1264947723222912e-05, + "loss": 0.8838, + "step": 40807 + }, + { + "epoch": 2.63, + "learning_rate": 1.1264733034199961e-05, + "loss": 0.9299, + "step": 40808 + }, + { + "epoch": 2.63, + "learning_rate": 1.1264518345177012e-05, + "loss": 0.9034, + "step": 40809 + }, + { + "epoch": 2.63, + "learning_rate": 1.126430365615406e-05, + "loss": 0.9135, + "step": 40810 + }, + { + "epoch": 2.63, + "learning_rate": 1.1264088967131111e-05, + "loss": 0.9259, + "step": 40811 + }, + { + "epoch": 2.63, + "learning_rate": 1.1263874278108162e-05, + "loss": 0.9013, + "step": 40812 + }, + { + "epoch": 2.63, + "learning_rate": 1.1263659589085212e-05, + "loss": 0.8611, + "step": 40813 + }, + { + "epoch": 2.63, + "learning_rate": 1.1263444900062261e-05, + "loss": 0.8284, + "step": 40814 + }, + { + "epoch": 2.63, + "learning_rate": 1.126323021103931e-05, + "loss": 0.9793, + "step": 40815 + }, + { + "epoch": 2.63, + "learning_rate": 1.1263015522016361e-05, + "loss": 0.9049, + "step": 40816 + }, + { + "epoch": 2.63, + "learning_rate": 1.126280083299341e-05, + "loss": 0.9291, + "step": 40817 + }, + { + "epoch": 2.63, + "learning_rate": 1.1262586143970459e-05, + "loss": 0.8793, + "step": 40818 + }, + { + "epoch": 2.63, + "learning_rate": 1.126237145494751e-05, + "loss": 0.8791, + "step": 40819 + }, + { + "epoch": 2.63, + "learning_rate": 1.1262156765924559e-05, + "loss": 0.93, + "step": 40820 + }, + { + "epoch": 2.63, + "learning_rate": 1.126194207690161e-05, + "loss": 0.958, + "step": 40821 + }, + { + "epoch": 2.63, + "learning_rate": 1.1261727387878658e-05, + "loss": 0.972, + "step": 40822 + }, + { + "epoch": 2.63, + "learning_rate": 1.1261512698855707e-05, + "loss": 0.8759, + "step": 40823 + }, + { + "epoch": 2.63, + "learning_rate": 1.1261298009832758e-05, + "loss": 0.8414, + "step": 40824 + }, + { + "epoch": 2.63, + "learning_rate": 1.1261083320809807e-05, + "loss": 0.9243, + "step": 40825 + }, + { + "epoch": 2.63, + "learning_rate": 1.126086863178686e-05, + "loss": 0.8999, + "step": 40826 + }, + { + "epoch": 2.63, + "learning_rate": 1.1260653942763908e-05, + "loss": 0.8869, + "step": 40827 + }, + { + "epoch": 2.63, + "learning_rate": 1.1260439253740957e-05, + "loss": 0.8729, + "step": 40828 + }, + { + "epoch": 2.63, + "learning_rate": 1.1260224564718008e-05, + "loss": 0.8332, + "step": 40829 + }, + { + "epoch": 2.63, + "learning_rate": 1.1260009875695057e-05, + "loss": 0.8705, + "step": 40830 + }, + { + "epoch": 2.63, + "learning_rate": 1.1259795186672107e-05, + "loss": 0.9011, + "step": 40831 + }, + { + "epoch": 2.63, + "learning_rate": 1.1259580497649156e-05, + "loss": 0.9053, + "step": 40832 + }, + { + "epoch": 2.63, + "learning_rate": 1.1259365808626205e-05, + "loss": 0.9239, + "step": 40833 + }, + { + "epoch": 2.63, + "learning_rate": 1.1259151119603256e-05, + "loss": 0.9268, + "step": 40834 + }, + { + "epoch": 2.63, + "learning_rate": 1.1258936430580305e-05, + "loss": 0.8485, + "step": 40835 + }, + { + "epoch": 2.63, + "learning_rate": 1.1258721741557354e-05, + "loss": 0.8704, + "step": 40836 + }, + { + "epoch": 2.63, + "learning_rate": 1.1258507052534405e-05, + "loss": 0.9039, + "step": 40837 + }, + { + "epoch": 2.63, + "learning_rate": 1.1258292363511454e-05, + "loss": 0.9668, + "step": 40838 + }, + { + "epoch": 2.63, + "learning_rate": 1.1258077674488504e-05, + "loss": 0.9003, + "step": 40839 + }, + { + "epoch": 2.63, + "learning_rate": 1.1257862985465553e-05, + "loss": 0.9323, + "step": 40840 + }, + { + "epoch": 2.63, + "learning_rate": 1.1257648296442605e-05, + "loss": 0.8224, + "step": 40841 + }, + { + "epoch": 2.63, + "learning_rate": 1.1257433607419654e-05, + "loss": 0.8864, + "step": 40842 + }, + { + "epoch": 2.63, + "learning_rate": 1.1257218918396703e-05, + "loss": 0.913, + "step": 40843 + }, + { + "epoch": 2.63, + "learning_rate": 1.1257004229373754e-05, + "loss": 0.8752, + "step": 40844 + }, + { + "epoch": 2.63, + "learning_rate": 1.1256789540350803e-05, + "loss": 1.0455, + "step": 40845 + }, + { + "epoch": 2.63, + "learning_rate": 1.1256574851327852e-05, + "loss": 0.9258, + "step": 40846 + }, + { + "epoch": 2.63, + "learning_rate": 1.1256360162304903e-05, + "loss": 0.9506, + "step": 40847 + }, + { + "epoch": 2.63, + "learning_rate": 1.1256145473281952e-05, + "loss": 0.9715, + "step": 40848 + }, + { + "epoch": 2.63, + "learning_rate": 1.1255930784259002e-05, + "loss": 0.8899, + "step": 40849 + }, + { + "epoch": 2.63, + "learning_rate": 1.1255716095236051e-05, + "loss": 0.9452, + "step": 40850 + }, + { + "epoch": 2.63, + "learning_rate": 1.12555014062131e-05, + "loss": 0.9233, + "step": 40851 + }, + { + "epoch": 2.63, + "learning_rate": 1.1255286717190151e-05, + "loss": 0.8628, + "step": 40852 + }, + { + "epoch": 2.63, + "learning_rate": 1.12550720281672e-05, + "loss": 0.9134, + "step": 40853 + }, + { + "epoch": 2.63, + "learning_rate": 1.1254857339144249e-05, + "loss": 0.9191, + "step": 40854 + }, + { + "epoch": 2.63, + "learning_rate": 1.1254642650121301e-05, + "loss": 0.8322, + "step": 40855 + }, + { + "epoch": 2.63, + "learning_rate": 1.125442796109835e-05, + "loss": 0.9243, + "step": 40856 + }, + { + "epoch": 2.63, + "learning_rate": 1.12542132720754e-05, + "loss": 0.8883, + "step": 40857 + }, + { + "epoch": 2.63, + "learning_rate": 1.125399858305245e-05, + "loss": 0.947, + "step": 40858 + }, + { + "epoch": 2.63, + "learning_rate": 1.12537838940295e-05, + "loss": 0.9446, + "step": 40859 + }, + { + "epoch": 2.63, + "learning_rate": 1.125356920500655e-05, + "loss": 0.911, + "step": 40860 + }, + { + "epoch": 2.63, + "learning_rate": 1.1253354515983598e-05, + "loss": 0.9346, + "step": 40861 + }, + { + "epoch": 2.63, + "learning_rate": 1.1253139826960649e-05, + "loss": 0.9642, + "step": 40862 + }, + { + "epoch": 2.63, + "learning_rate": 1.1252925137937698e-05, + "loss": 0.9733, + "step": 40863 + }, + { + "epoch": 2.63, + "learning_rate": 1.1252710448914747e-05, + "loss": 0.8274, + "step": 40864 + }, + { + "epoch": 2.63, + "learning_rate": 1.1252495759891798e-05, + "loss": 0.9116, + "step": 40865 + }, + { + "epoch": 2.63, + "learning_rate": 1.1252281070868847e-05, + "loss": 0.783, + "step": 40866 + }, + { + "epoch": 2.63, + "learning_rate": 1.1252066381845897e-05, + "loss": 0.8711, + "step": 40867 + }, + { + "epoch": 2.63, + "learning_rate": 1.1251851692822946e-05, + "loss": 0.9216, + "step": 40868 + }, + { + "epoch": 2.63, + "learning_rate": 1.1251637003799995e-05, + "loss": 0.8763, + "step": 40869 + }, + { + "epoch": 2.63, + "learning_rate": 1.1251422314777048e-05, + "loss": 0.936, + "step": 40870 + }, + { + "epoch": 2.63, + "learning_rate": 1.1251207625754096e-05, + "loss": 0.9512, + "step": 40871 + }, + { + "epoch": 2.63, + "learning_rate": 1.1250992936731147e-05, + "loss": 0.8737, + "step": 40872 + }, + { + "epoch": 2.63, + "learning_rate": 1.1250778247708196e-05, + "loss": 0.8756, + "step": 40873 + }, + { + "epoch": 2.63, + "learning_rate": 1.1250563558685245e-05, + "loss": 0.9478, + "step": 40874 + }, + { + "epoch": 2.63, + "learning_rate": 1.1250348869662296e-05, + "loss": 0.881, + "step": 40875 + }, + { + "epoch": 2.63, + "learning_rate": 1.1250134180639345e-05, + "loss": 0.8572, + "step": 40876 + }, + { + "epoch": 2.63, + "learning_rate": 1.1249919491616395e-05, + "loss": 0.8825, + "step": 40877 + }, + { + "epoch": 2.63, + "learning_rate": 1.1249704802593444e-05, + "loss": 0.8586, + "step": 40878 + }, + { + "epoch": 2.63, + "learning_rate": 1.1249490113570493e-05, + "loss": 0.9227, + "step": 40879 + }, + { + "epoch": 2.63, + "learning_rate": 1.1249275424547544e-05, + "loss": 0.881, + "step": 40880 + }, + { + "epoch": 2.63, + "learning_rate": 1.1249060735524593e-05, + "loss": 0.9497, + "step": 40881 + }, + { + "epoch": 2.63, + "learning_rate": 1.1248846046501642e-05, + "loss": 0.8959, + "step": 40882 + }, + { + "epoch": 2.63, + "learning_rate": 1.1248631357478693e-05, + "loss": 0.8636, + "step": 40883 + }, + { + "epoch": 2.63, + "learning_rate": 1.1248416668455743e-05, + "loss": 0.9041, + "step": 40884 + }, + { + "epoch": 2.63, + "learning_rate": 1.1248201979432794e-05, + "loss": 0.9267, + "step": 40885 + }, + { + "epoch": 2.63, + "learning_rate": 1.1247987290409843e-05, + "loss": 0.8825, + "step": 40886 + }, + { + "epoch": 2.63, + "learning_rate": 1.1247772601386893e-05, + "loss": 0.8886, + "step": 40887 + }, + { + "epoch": 2.63, + "learning_rate": 1.1247557912363942e-05, + "loss": 0.9134, + "step": 40888 + }, + { + "epoch": 2.63, + "learning_rate": 1.1247343223340991e-05, + "loss": 0.8781, + "step": 40889 + }, + { + "epoch": 2.63, + "learning_rate": 1.1247128534318042e-05, + "loss": 0.9366, + "step": 40890 + }, + { + "epoch": 2.63, + "learning_rate": 1.1246913845295091e-05, + "loss": 0.8785, + "step": 40891 + }, + { + "epoch": 2.63, + "learning_rate": 1.124669915627214e-05, + "loss": 0.9281, + "step": 40892 + }, + { + "epoch": 2.63, + "learning_rate": 1.124648446724919e-05, + "loss": 0.9482, + "step": 40893 + }, + { + "epoch": 2.63, + "learning_rate": 1.124626977822624e-05, + "loss": 0.831, + "step": 40894 + }, + { + "epoch": 2.63, + "learning_rate": 1.1246055089203289e-05, + "loss": 0.9373, + "step": 40895 + }, + { + "epoch": 2.63, + "learning_rate": 1.124584040018034e-05, + "loss": 0.9216, + "step": 40896 + }, + { + "epoch": 2.63, + "learning_rate": 1.1245625711157388e-05, + "loss": 0.8657, + "step": 40897 + }, + { + "epoch": 2.63, + "learning_rate": 1.1245411022134439e-05, + "loss": 0.9096, + "step": 40898 + }, + { + "epoch": 2.63, + "learning_rate": 1.124519633311149e-05, + "loss": 0.9275, + "step": 40899 + }, + { + "epoch": 2.63, + "learning_rate": 1.124498164408854e-05, + "loss": 0.9298, + "step": 40900 + }, + { + "epoch": 2.63, + "learning_rate": 1.1244766955065589e-05, + "loss": 0.8326, + "step": 40901 + }, + { + "epoch": 2.63, + "learning_rate": 1.1244552266042638e-05, + "loss": 0.8974, + "step": 40902 + }, + { + "epoch": 2.63, + "learning_rate": 1.1244337577019689e-05, + "loss": 0.899, + "step": 40903 + }, + { + "epoch": 2.63, + "learning_rate": 1.1244122887996738e-05, + "loss": 0.892, + "step": 40904 + }, + { + "epoch": 2.63, + "learning_rate": 1.1243908198973787e-05, + "loss": 0.9597, + "step": 40905 + }, + { + "epoch": 2.63, + "learning_rate": 1.1243693509950837e-05, + "loss": 0.8227, + "step": 40906 + }, + { + "epoch": 2.63, + "learning_rate": 1.1243478820927886e-05, + "loss": 0.8853, + "step": 40907 + }, + { + "epoch": 2.63, + "learning_rate": 1.1243264131904937e-05, + "loss": 0.8359, + "step": 40908 + }, + { + "epoch": 2.63, + "learning_rate": 1.1243049442881986e-05, + "loss": 0.9255, + "step": 40909 + }, + { + "epoch": 2.63, + "learning_rate": 1.1242834753859035e-05, + "loss": 0.9145, + "step": 40910 + }, + { + "epoch": 2.63, + "learning_rate": 1.1242620064836086e-05, + "loss": 0.9496, + "step": 40911 + }, + { + "epoch": 2.63, + "learning_rate": 1.1242405375813135e-05, + "loss": 0.8681, + "step": 40912 + }, + { + "epoch": 2.63, + "learning_rate": 1.1242190686790187e-05, + "loss": 0.9666, + "step": 40913 + }, + { + "epoch": 2.63, + "learning_rate": 1.1241975997767236e-05, + "loss": 0.8747, + "step": 40914 + }, + { + "epoch": 2.63, + "learning_rate": 1.1241761308744285e-05, + "loss": 0.8635, + "step": 40915 + }, + { + "epoch": 2.63, + "learning_rate": 1.1241546619721335e-05, + "loss": 0.9089, + "step": 40916 + }, + { + "epoch": 2.63, + "learning_rate": 1.1241331930698384e-05, + "loss": 0.9584, + "step": 40917 + }, + { + "epoch": 2.63, + "learning_rate": 1.1241117241675435e-05, + "loss": 0.9442, + "step": 40918 + }, + { + "epoch": 2.63, + "learning_rate": 1.1240902552652484e-05, + "loss": 0.8827, + "step": 40919 + }, + { + "epoch": 2.63, + "learning_rate": 1.1240687863629533e-05, + "loss": 0.8431, + "step": 40920 + }, + { + "epoch": 2.63, + "learning_rate": 1.1240473174606584e-05, + "loss": 0.8739, + "step": 40921 + }, + { + "epoch": 2.63, + "learning_rate": 1.1240258485583633e-05, + "loss": 0.8523, + "step": 40922 + }, + { + "epoch": 2.63, + "learning_rate": 1.1240043796560682e-05, + "loss": 0.8789, + "step": 40923 + }, + { + "epoch": 2.63, + "learning_rate": 1.1239829107537732e-05, + "loss": 0.9017, + "step": 40924 + }, + { + "epoch": 2.63, + "learning_rate": 1.1239614418514781e-05, + "loss": 0.9085, + "step": 40925 + }, + { + "epoch": 2.63, + "learning_rate": 1.1239399729491832e-05, + "loss": 0.8367, + "step": 40926 + }, + { + "epoch": 2.63, + "learning_rate": 1.1239185040468881e-05, + "loss": 0.941, + "step": 40927 + }, + { + "epoch": 2.63, + "learning_rate": 1.1238970351445933e-05, + "loss": 0.9358, + "step": 40928 + }, + { + "epoch": 2.63, + "learning_rate": 1.1238755662422982e-05, + "loss": 0.8906, + "step": 40929 + }, + { + "epoch": 2.63, + "learning_rate": 1.1238540973400031e-05, + "loss": 0.9569, + "step": 40930 + }, + { + "epoch": 2.63, + "learning_rate": 1.1238326284377082e-05, + "loss": 0.9511, + "step": 40931 + }, + { + "epoch": 2.63, + "learning_rate": 1.123811159535413e-05, + "loss": 0.8961, + "step": 40932 + }, + { + "epoch": 2.63, + "learning_rate": 1.123789690633118e-05, + "loss": 0.9173, + "step": 40933 + }, + { + "epoch": 2.63, + "learning_rate": 1.123768221730823e-05, + "loss": 0.9117, + "step": 40934 + }, + { + "epoch": 2.63, + "learning_rate": 1.123746752828528e-05, + "loss": 0.8254, + "step": 40935 + }, + { + "epoch": 2.63, + "learning_rate": 1.123725283926233e-05, + "loss": 0.878, + "step": 40936 + }, + { + "epoch": 2.63, + "learning_rate": 1.1237038150239379e-05, + "loss": 0.8514, + "step": 40937 + }, + { + "epoch": 2.63, + "learning_rate": 1.1236823461216428e-05, + "loss": 0.8313, + "step": 40938 + }, + { + "epoch": 2.63, + "learning_rate": 1.1236608772193479e-05, + "loss": 0.8476, + "step": 40939 + }, + { + "epoch": 2.63, + "learning_rate": 1.1236394083170528e-05, + "loss": 0.9615, + "step": 40940 + }, + { + "epoch": 2.63, + "learning_rate": 1.1236179394147577e-05, + "loss": 0.9361, + "step": 40941 + }, + { + "epoch": 2.63, + "learning_rate": 1.1235964705124629e-05, + "loss": 0.908, + "step": 40942 + }, + { + "epoch": 2.63, + "learning_rate": 1.1235750016101678e-05, + "loss": 0.7959, + "step": 40943 + }, + { + "epoch": 2.63, + "learning_rate": 1.1235535327078729e-05, + "loss": 0.9212, + "step": 40944 + }, + { + "epoch": 2.63, + "learning_rate": 1.1235320638055777e-05, + "loss": 0.8923, + "step": 40945 + }, + { + "epoch": 2.63, + "learning_rate": 1.1235105949032828e-05, + "loss": 0.9176, + "step": 40946 + }, + { + "epoch": 2.63, + "learning_rate": 1.1234891260009877e-05, + "loss": 0.8788, + "step": 40947 + }, + { + "epoch": 2.63, + "learning_rate": 1.1234676570986926e-05, + "loss": 0.8784, + "step": 40948 + }, + { + "epoch": 2.63, + "learning_rate": 1.1234461881963977e-05, + "loss": 0.8414, + "step": 40949 + }, + { + "epoch": 2.63, + "learning_rate": 1.1234247192941026e-05, + "loss": 0.8877, + "step": 40950 + }, + { + "epoch": 2.63, + "learning_rate": 1.1234032503918075e-05, + "loss": 0.9148, + "step": 40951 + }, + { + "epoch": 2.63, + "learning_rate": 1.1233817814895125e-05, + "loss": 0.8569, + "step": 40952 + }, + { + "epoch": 2.63, + "learning_rate": 1.1233603125872174e-05, + "loss": 0.8882, + "step": 40953 + }, + { + "epoch": 2.63, + "learning_rate": 1.1233388436849225e-05, + "loss": 0.8375, + "step": 40954 + }, + { + "epoch": 2.63, + "learning_rate": 1.1233173747826274e-05, + "loss": 0.9686, + "step": 40955 + }, + { + "epoch": 2.64, + "learning_rate": 1.1232959058803323e-05, + "loss": 0.9417, + "step": 40956 + }, + { + "epoch": 2.64, + "learning_rate": 1.1232744369780375e-05, + "loss": 0.9283, + "step": 40957 + }, + { + "epoch": 2.64, + "learning_rate": 1.1232529680757424e-05, + "loss": 0.9086, + "step": 40958 + }, + { + "epoch": 2.64, + "learning_rate": 1.1232314991734475e-05, + "loss": 0.9127, + "step": 40959 + }, + { + "epoch": 2.64, + "learning_rate": 1.1232100302711524e-05, + "loss": 0.8471, + "step": 40960 + }, + { + "epoch": 2.64, + "learning_rate": 1.1231885613688573e-05, + "loss": 0.8408, + "step": 40961 + }, + { + "epoch": 2.64, + "learning_rate": 1.1231670924665623e-05, + "loss": 0.9257, + "step": 40962 + }, + { + "epoch": 2.64, + "learning_rate": 1.1231456235642672e-05, + "loss": 0.9185, + "step": 40963 + }, + { + "epoch": 2.64, + "learning_rate": 1.1231241546619723e-05, + "loss": 0.9022, + "step": 40964 + }, + { + "epoch": 2.64, + "learning_rate": 1.1231026857596772e-05, + "loss": 0.915, + "step": 40965 + }, + { + "epoch": 2.64, + "learning_rate": 1.1230812168573821e-05, + "loss": 0.9144, + "step": 40966 + }, + { + "epoch": 2.64, + "learning_rate": 1.1230597479550872e-05, + "loss": 0.9613, + "step": 40967 + }, + { + "epoch": 2.64, + "learning_rate": 1.123038279052792e-05, + "loss": 0.9144, + "step": 40968 + }, + { + "epoch": 2.64, + "learning_rate": 1.123016810150497e-05, + "loss": 0.8417, + "step": 40969 + }, + { + "epoch": 2.64, + "learning_rate": 1.122995341248202e-05, + "loss": 0.9791, + "step": 40970 + }, + { + "epoch": 2.64, + "learning_rate": 1.122973872345907e-05, + "loss": 0.8658, + "step": 40971 + }, + { + "epoch": 2.64, + "learning_rate": 1.1229524034436122e-05, + "loss": 0.9103, + "step": 40972 + }, + { + "epoch": 2.64, + "learning_rate": 1.122930934541317e-05, + "loss": 0.884, + "step": 40973 + }, + { + "epoch": 2.64, + "learning_rate": 1.1229094656390221e-05, + "loss": 0.9244, + "step": 40974 + }, + { + "epoch": 2.64, + "learning_rate": 1.122887996736727e-05, + "loss": 0.8892, + "step": 40975 + }, + { + "epoch": 2.64, + "learning_rate": 1.1228665278344319e-05, + "loss": 0.8522, + "step": 40976 + }, + { + "epoch": 2.64, + "learning_rate": 1.122845058932137e-05, + "loss": 0.9371, + "step": 40977 + }, + { + "epoch": 2.64, + "learning_rate": 1.1228235900298419e-05, + "loss": 0.9351, + "step": 40978 + }, + { + "epoch": 2.64, + "learning_rate": 1.1228021211275468e-05, + "loss": 0.8817, + "step": 40979 + }, + { + "epoch": 2.64, + "learning_rate": 1.1227806522252518e-05, + "loss": 0.8943, + "step": 40980 + }, + { + "epoch": 2.64, + "learning_rate": 1.1227591833229567e-05, + "loss": 0.8499, + "step": 40981 + }, + { + "epoch": 2.64, + "learning_rate": 1.1227377144206616e-05, + "loss": 0.9091, + "step": 40982 + }, + { + "epoch": 2.64, + "learning_rate": 1.1227162455183667e-05, + "loss": 0.9553, + "step": 40983 + }, + { + "epoch": 2.64, + "learning_rate": 1.1226947766160716e-05, + "loss": 0.8893, + "step": 40984 + }, + { + "epoch": 2.64, + "learning_rate": 1.1226733077137767e-05, + "loss": 0.9201, + "step": 40985 + }, + { + "epoch": 2.64, + "learning_rate": 1.1226518388114817e-05, + "loss": 0.8481, + "step": 40986 + }, + { + "epoch": 2.64, + "learning_rate": 1.1226303699091868e-05, + "loss": 0.8701, + "step": 40987 + }, + { + "epoch": 2.64, + "learning_rate": 1.1226089010068917e-05, + "loss": 0.93, + "step": 40988 + }, + { + "epoch": 2.64, + "learning_rate": 1.1225874321045966e-05, + "loss": 0.9673, + "step": 40989 + }, + { + "epoch": 2.64, + "learning_rate": 1.1225659632023016e-05, + "loss": 0.907, + "step": 40990 + }, + { + "epoch": 2.64, + "learning_rate": 1.1225444943000065e-05, + "loss": 0.9245, + "step": 40991 + }, + { + "epoch": 2.64, + "learning_rate": 1.1225230253977114e-05, + "loss": 0.8595, + "step": 40992 + }, + { + "epoch": 2.64, + "learning_rate": 1.1225015564954165e-05, + "loss": 0.887, + "step": 40993 + }, + { + "epoch": 2.64, + "learning_rate": 1.1224800875931214e-05, + "loss": 0.9028, + "step": 40994 + }, + { + "epoch": 2.64, + "learning_rate": 1.1224586186908265e-05, + "loss": 0.8908, + "step": 40995 + }, + { + "epoch": 2.64, + "learning_rate": 1.1224371497885314e-05, + "loss": 0.9041, + "step": 40996 + }, + { + "epoch": 2.64, + "learning_rate": 1.1224156808862363e-05, + "loss": 0.9324, + "step": 40997 + }, + { + "epoch": 2.64, + "learning_rate": 1.1223942119839413e-05, + "loss": 0.8879, + "step": 40998 + }, + { + "epoch": 2.64, + "learning_rate": 1.1223727430816462e-05, + "loss": 0.9153, + "step": 40999 + }, + { + "epoch": 2.64, + "learning_rate": 1.1223512741793511e-05, + "loss": 0.8799, + "step": 41000 + }, + { + "epoch": 2.64, + "eval_laion/cider": 0.529824596404683, + "eval_laion/meteor": 0.3896557477461897, + "eval_laion/pred_num_words": 38.359375, + "eval_laion/sacrebleu": 15.664230187387789, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0398929119110107, + "eval_runtime": 97.338, + "eval_samples_per_second": 5.126, + "eval_steps_per_second": 0.329, + "step": 41000 + }, + { + "epoch": 2.64, + "learning_rate": 1.1223298052770564e-05, + "loss": 0.8339, + "step": 41001 + }, + { + "epoch": 2.64, + "learning_rate": 1.1223083363747613e-05, + "loss": 0.8592, + "step": 41002 + }, + { + "epoch": 2.64, + "learning_rate": 1.1222868674724663e-05, + "loss": 0.9408, + "step": 41003 + }, + { + "epoch": 2.64, + "learning_rate": 1.1222653985701712e-05, + "loss": 0.8968, + "step": 41004 + }, + { + "epoch": 2.64, + "learning_rate": 1.1222439296678763e-05, + "loss": 0.8146, + "step": 41005 + }, + { + "epoch": 2.64, + "learning_rate": 1.1222224607655812e-05, + "loss": 0.889, + "step": 41006 + }, + { + "epoch": 2.64, + "learning_rate": 1.122200991863286e-05, + "loss": 0.8696, + "step": 41007 + }, + { + "epoch": 2.64, + "learning_rate": 1.1221795229609911e-05, + "loss": 0.9269, + "step": 41008 + }, + { + "epoch": 2.64, + "learning_rate": 1.122158054058696e-05, + "loss": 0.8353, + "step": 41009 + }, + { + "epoch": 2.64, + "learning_rate": 1.122136585156401e-05, + "loss": 0.8656, + "step": 41010 + }, + { + "epoch": 2.64, + "learning_rate": 1.122115116254106e-05, + "loss": 0.8656, + "step": 41011 + }, + { + "epoch": 2.64, + "learning_rate": 1.1220936473518109e-05, + "loss": 0.9287, + "step": 41012 + }, + { + "epoch": 2.64, + "learning_rate": 1.122072178449516e-05, + "loss": 0.8422, + "step": 41013 + }, + { + "epoch": 2.64, + "learning_rate": 1.1220507095472209e-05, + "loss": 0.9059, + "step": 41014 + }, + { + "epoch": 2.64, + "learning_rate": 1.1220292406449261e-05, + "loss": 0.8885, + "step": 41015 + }, + { + "epoch": 2.64, + "learning_rate": 1.122007771742631e-05, + "loss": 0.9155, + "step": 41016 + }, + { + "epoch": 2.64, + "learning_rate": 1.1219863028403359e-05, + "loss": 0.8933, + "step": 41017 + }, + { + "epoch": 2.64, + "learning_rate": 1.121964833938041e-05, + "loss": 0.9327, + "step": 41018 + }, + { + "epoch": 2.64, + "learning_rate": 1.1219433650357458e-05, + "loss": 0.8454, + "step": 41019 + }, + { + "epoch": 2.64, + "learning_rate": 1.1219218961334507e-05, + "loss": 0.9208, + "step": 41020 + }, + { + "epoch": 2.64, + "learning_rate": 1.1219004272311558e-05, + "loss": 0.9423, + "step": 41021 + }, + { + "epoch": 2.64, + "learning_rate": 1.1218789583288607e-05, + "loss": 0.9534, + "step": 41022 + }, + { + "epoch": 2.64, + "learning_rate": 1.1218574894265658e-05, + "loss": 0.8865, + "step": 41023 + }, + { + "epoch": 2.64, + "learning_rate": 1.1218360205242707e-05, + "loss": 0.8985, + "step": 41024 + }, + { + "epoch": 2.64, + "learning_rate": 1.1218145516219756e-05, + "loss": 0.9265, + "step": 41025 + }, + { + "epoch": 2.64, + "learning_rate": 1.1217930827196806e-05, + "loss": 0.9203, + "step": 41026 + }, + { + "epoch": 2.64, + "learning_rate": 1.1217716138173855e-05, + "loss": 0.9243, + "step": 41027 + }, + { + "epoch": 2.64, + "learning_rate": 1.1217501449150904e-05, + "loss": 0.883, + "step": 41028 + }, + { + "epoch": 2.64, + "learning_rate": 1.1217286760127955e-05, + "loss": 0.8864, + "step": 41029 + }, + { + "epoch": 2.64, + "learning_rate": 1.1217072071105006e-05, + "loss": 0.8762, + "step": 41030 + }, + { + "epoch": 2.64, + "learning_rate": 1.1216857382082056e-05, + "loss": 0.9408, + "step": 41031 + }, + { + "epoch": 2.64, + "learning_rate": 1.1216642693059105e-05, + "loss": 0.908, + "step": 41032 + }, + { + "epoch": 2.64, + "learning_rate": 1.1216428004036156e-05, + "loss": 0.8639, + "step": 41033 + }, + { + "epoch": 2.64, + "learning_rate": 1.1216213315013205e-05, + "loss": 0.8633, + "step": 41034 + }, + { + "epoch": 2.64, + "learning_rate": 1.1215998625990254e-05, + "loss": 0.9418, + "step": 41035 + }, + { + "epoch": 2.64, + "learning_rate": 1.1215783936967304e-05, + "loss": 0.9161, + "step": 41036 + }, + { + "epoch": 2.64, + "learning_rate": 1.1215569247944353e-05, + "loss": 0.9312, + "step": 41037 + }, + { + "epoch": 2.64, + "learning_rate": 1.1215354558921402e-05, + "loss": 0.892, + "step": 41038 + }, + { + "epoch": 2.64, + "learning_rate": 1.1215139869898453e-05, + "loss": 0.9384, + "step": 41039 + }, + { + "epoch": 2.64, + "learning_rate": 1.1214925180875502e-05, + "loss": 0.9474, + "step": 41040 + }, + { + "epoch": 2.64, + "learning_rate": 1.1214710491852551e-05, + "loss": 0.8139, + "step": 41041 + }, + { + "epoch": 2.64, + "learning_rate": 1.1214495802829602e-05, + "loss": 0.9087, + "step": 41042 + }, + { + "epoch": 2.64, + "learning_rate": 1.121428111380665e-05, + "loss": 0.9022, + "step": 41043 + }, + { + "epoch": 2.64, + "learning_rate": 1.1214066424783703e-05, + "loss": 0.849, + "step": 41044 + }, + { + "epoch": 2.64, + "learning_rate": 1.1213851735760752e-05, + "loss": 0.9282, + "step": 41045 + }, + { + "epoch": 2.64, + "learning_rate": 1.1213637046737803e-05, + "loss": 0.9137, + "step": 41046 + }, + { + "epoch": 2.64, + "learning_rate": 1.1213422357714852e-05, + "loss": 0.9316, + "step": 41047 + }, + { + "epoch": 2.64, + "learning_rate": 1.12132076686919e-05, + "loss": 0.9339, + "step": 41048 + }, + { + "epoch": 2.64, + "learning_rate": 1.1212992979668951e-05, + "loss": 0.942, + "step": 41049 + }, + { + "epoch": 2.64, + "learning_rate": 1.1212778290646e-05, + "loss": 0.9186, + "step": 41050 + }, + { + "epoch": 2.64, + "learning_rate": 1.1212563601623049e-05, + "loss": 0.938, + "step": 41051 + }, + { + "epoch": 2.64, + "learning_rate": 1.12123489126001e-05, + "loss": 0.9075, + "step": 41052 + }, + { + "epoch": 2.64, + "learning_rate": 1.1212134223577149e-05, + "loss": 0.9027, + "step": 41053 + }, + { + "epoch": 2.64, + "learning_rate": 1.12119195345542e-05, + "loss": 0.8765, + "step": 41054 + }, + { + "epoch": 2.64, + "learning_rate": 1.1211704845531248e-05, + "loss": 0.8408, + "step": 41055 + }, + { + "epoch": 2.64, + "learning_rate": 1.1211490156508297e-05, + "loss": 0.8705, + "step": 41056 + }, + { + "epoch": 2.64, + "learning_rate": 1.1211275467485348e-05, + "loss": 0.8995, + "step": 41057 + }, + { + "epoch": 2.64, + "learning_rate": 1.1211060778462397e-05, + "loss": 0.9513, + "step": 41058 + }, + { + "epoch": 2.64, + "learning_rate": 1.121084608943945e-05, + "loss": 0.8599, + "step": 41059 + }, + { + "epoch": 2.64, + "learning_rate": 1.1210631400416498e-05, + "loss": 0.8649, + "step": 41060 + }, + { + "epoch": 2.64, + "learning_rate": 1.1210416711393547e-05, + "loss": 0.9342, + "step": 41061 + }, + { + "epoch": 2.64, + "learning_rate": 1.1210202022370598e-05, + "loss": 0.8387, + "step": 41062 + }, + { + "epoch": 2.64, + "learning_rate": 1.1209987333347647e-05, + "loss": 0.898, + "step": 41063 + }, + { + "epoch": 2.64, + "learning_rate": 1.1209772644324697e-05, + "loss": 0.9186, + "step": 41064 + }, + { + "epoch": 2.64, + "learning_rate": 1.1209557955301746e-05, + "loss": 0.8788, + "step": 41065 + }, + { + "epoch": 2.64, + "learning_rate": 1.1209343266278795e-05, + "loss": 0.8736, + "step": 41066 + }, + { + "epoch": 2.64, + "learning_rate": 1.1209128577255846e-05, + "loss": 0.8805, + "step": 41067 + }, + { + "epoch": 2.64, + "learning_rate": 1.1208913888232895e-05, + "loss": 0.8577, + "step": 41068 + }, + { + "epoch": 2.64, + "learning_rate": 1.1208699199209944e-05, + "loss": 0.9417, + "step": 41069 + }, + { + "epoch": 2.64, + "learning_rate": 1.1208484510186995e-05, + "loss": 0.9717, + "step": 41070 + }, + { + "epoch": 2.64, + "learning_rate": 1.1208269821164044e-05, + "loss": 0.921, + "step": 41071 + }, + { + "epoch": 2.64, + "learning_rate": 1.1208055132141094e-05, + "loss": 0.9138, + "step": 41072 + }, + { + "epoch": 2.64, + "learning_rate": 1.1207840443118145e-05, + "loss": 0.9229, + "step": 41073 + }, + { + "epoch": 2.64, + "learning_rate": 1.1207625754095196e-05, + "loss": 0.8816, + "step": 41074 + }, + { + "epoch": 2.64, + "learning_rate": 1.1207411065072245e-05, + "loss": 0.9336, + "step": 41075 + }, + { + "epoch": 2.64, + "learning_rate": 1.1207196376049294e-05, + "loss": 0.8667, + "step": 41076 + }, + { + "epoch": 2.64, + "learning_rate": 1.1206981687026344e-05, + "loss": 0.98, + "step": 41077 + }, + { + "epoch": 2.64, + "learning_rate": 1.1206766998003393e-05, + "loss": 0.884, + "step": 41078 + }, + { + "epoch": 2.64, + "learning_rate": 1.1206552308980442e-05, + "loss": 0.9085, + "step": 41079 + }, + { + "epoch": 2.64, + "learning_rate": 1.1206337619957493e-05, + "loss": 0.8914, + "step": 41080 + }, + { + "epoch": 2.64, + "learning_rate": 1.1206122930934542e-05, + "loss": 0.9414, + "step": 41081 + }, + { + "epoch": 2.64, + "learning_rate": 1.1205908241911592e-05, + "loss": 0.8591, + "step": 41082 + }, + { + "epoch": 2.64, + "learning_rate": 1.1205693552888641e-05, + "loss": 0.9652, + "step": 41083 + }, + { + "epoch": 2.64, + "learning_rate": 1.120547886386569e-05, + "loss": 0.9602, + "step": 41084 + }, + { + "epoch": 2.64, + "learning_rate": 1.1205264174842741e-05, + "loss": 0.9156, + "step": 41085 + }, + { + "epoch": 2.64, + "learning_rate": 1.120504948581979e-05, + "loss": 0.8618, + "step": 41086 + }, + { + "epoch": 2.64, + "learning_rate": 1.1204834796796839e-05, + "loss": 0.9469, + "step": 41087 + }, + { + "epoch": 2.64, + "learning_rate": 1.1204620107773891e-05, + "loss": 0.8595, + "step": 41088 + }, + { + "epoch": 2.64, + "learning_rate": 1.120440541875094e-05, + "loss": 0.9044, + "step": 41089 + }, + { + "epoch": 2.64, + "learning_rate": 1.1204190729727991e-05, + "loss": 0.9014, + "step": 41090 + }, + { + "epoch": 2.64, + "learning_rate": 1.120397604070504e-05, + "loss": 0.9044, + "step": 41091 + }, + { + "epoch": 2.64, + "learning_rate": 1.120376135168209e-05, + "loss": 0.9567, + "step": 41092 + }, + { + "epoch": 2.64, + "learning_rate": 1.120354666265914e-05, + "loss": 0.9428, + "step": 41093 + }, + { + "epoch": 2.64, + "learning_rate": 1.1203331973636188e-05, + "loss": 0.8265, + "step": 41094 + }, + { + "epoch": 2.64, + "learning_rate": 1.1203117284613239e-05, + "loss": 0.9043, + "step": 41095 + }, + { + "epoch": 2.64, + "learning_rate": 1.1202902595590288e-05, + "loss": 0.9507, + "step": 41096 + }, + { + "epoch": 2.64, + "learning_rate": 1.1202687906567337e-05, + "loss": 0.8639, + "step": 41097 + }, + { + "epoch": 2.64, + "learning_rate": 1.1202473217544388e-05, + "loss": 0.9, + "step": 41098 + }, + { + "epoch": 2.64, + "learning_rate": 1.1202258528521437e-05, + "loss": 0.9171, + "step": 41099 + }, + { + "epoch": 2.64, + "learning_rate": 1.1202043839498487e-05, + "loss": 0.9733, + "step": 41100 + }, + { + "epoch": 2.64, + "learning_rate": 1.1201829150475536e-05, + "loss": 0.8917, + "step": 41101 + }, + { + "epoch": 2.64, + "learning_rate": 1.1201614461452589e-05, + "loss": 0.9164, + "step": 41102 + }, + { + "epoch": 2.64, + "learning_rate": 1.1201399772429638e-05, + "loss": 0.9287, + "step": 41103 + }, + { + "epoch": 2.64, + "learning_rate": 1.1201185083406687e-05, + "loss": 0.929, + "step": 41104 + }, + { + "epoch": 2.64, + "learning_rate": 1.1200970394383737e-05, + "loss": 0.8818, + "step": 41105 + }, + { + "epoch": 2.64, + "learning_rate": 1.1200755705360786e-05, + "loss": 0.8297, + "step": 41106 + }, + { + "epoch": 2.64, + "learning_rate": 1.1200541016337835e-05, + "loss": 0.9044, + "step": 41107 + }, + { + "epoch": 2.64, + "learning_rate": 1.1200326327314886e-05, + "loss": 0.9283, + "step": 41108 + }, + { + "epoch": 2.64, + "learning_rate": 1.1200111638291935e-05, + "loss": 0.8838, + "step": 41109 + }, + { + "epoch": 2.64, + "learning_rate": 1.1199896949268985e-05, + "loss": 0.8916, + "step": 41110 + }, + { + "epoch": 2.64, + "learning_rate": 1.1199682260246034e-05, + "loss": 0.8803, + "step": 41111 + }, + { + "epoch": 2.65, + "learning_rate": 1.1199467571223083e-05, + "loss": 0.8577, + "step": 41112 + }, + { + "epoch": 2.65, + "learning_rate": 1.1199252882200134e-05, + "loss": 0.8963, + "step": 41113 + }, + { + "epoch": 2.65, + "learning_rate": 1.1199038193177183e-05, + "loss": 0.938, + "step": 41114 + }, + { + "epoch": 2.65, + "learning_rate": 1.1198823504154232e-05, + "loss": 0.8987, + "step": 41115 + }, + { + "epoch": 2.65, + "learning_rate": 1.1198608815131283e-05, + "loss": 0.9002, + "step": 41116 + }, + { + "epoch": 2.65, + "learning_rate": 1.1198394126108333e-05, + "loss": 0.966, + "step": 41117 + }, + { + "epoch": 2.65, + "learning_rate": 1.1198179437085384e-05, + "loss": 0.9468, + "step": 41118 + }, + { + "epoch": 2.65, + "learning_rate": 1.1197964748062433e-05, + "loss": 0.8668, + "step": 41119 + }, + { + "epoch": 2.65, + "learning_rate": 1.1197750059039484e-05, + "loss": 0.8589, + "step": 41120 + }, + { + "epoch": 2.65, + "learning_rate": 1.1197535370016533e-05, + "loss": 0.7927, + "step": 41121 + }, + { + "epoch": 2.65, + "learning_rate": 1.1197320680993581e-05, + "loss": 0.8592, + "step": 41122 + }, + { + "epoch": 2.65, + "learning_rate": 1.1197105991970632e-05, + "loss": 0.9235, + "step": 41123 + }, + { + "epoch": 2.65, + "learning_rate": 1.1196891302947681e-05, + "loss": 0.8571, + "step": 41124 + }, + { + "epoch": 2.65, + "learning_rate": 1.119667661392473e-05, + "loss": 0.8903, + "step": 41125 + }, + { + "epoch": 2.65, + "learning_rate": 1.119646192490178e-05, + "loss": 0.8315, + "step": 41126 + }, + { + "epoch": 2.65, + "learning_rate": 1.119624723587883e-05, + "loss": 0.863, + "step": 41127 + }, + { + "epoch": 2.65, + "learning_rate": 1.1196032546855879e-05, + "loss": 1.0243, + "step": 41128 + }, + { + "epoch": 2.65, + "learning_rate": 1.119581785783293e-05, + "loss": 0.8534, + "step": 41129 + }, + { + "epoch": 2.65, + "learning_rate": 1.1195603168809978e-05, + "loss": 0.8729, + "step": 41130 + }, + { + "epoch": 2.65, + "learning_rate": 1.119538847978703e-05, + "loss": 0.9235, + "step": 41131 + }, + { + "epoch": 2.65, + "learning_rate": 1.119517379076408e-05, + "loss": 0.8993, + "step": 41132 + }, + { + "epoch": 2.65, + "learning_rate": 1.119495910174113e-05, + "loss": 0.8752, + "step": 41133 + }, + { + "epoch": 2.65, + "learning_rate": 1.119474441271818e-05, + "loss": 0.8504, + "step": 41134 + }, + { + "epoch": 2.65, + "learning_rate": 1.1194529723695228e-05, + "loss": 0.8892, + "step": 41135 + }, + { + "epoch": 2.65, + "learning_rate": 1.1194315034672279e-05, + "loss": 0.8474, + "step": 41136 + }, + { + "epoch": 2.65, + "learning_rate": 1.1194100345649328e-05, + "loss": 0.8688, + "step": 41137 + }, + { + "epoch": 2.65, + "learning_rate": 1.1193885656626377e-05, + "loss": 0.9616, + "step": 41138 + }, + { + "epoch": 2.65, + "learning_rate": 1.1193670967603427e-05, + "loss": 0.9024, + "step": 41139 + }, + { + "epoch": 2.65, + "learning_rate": 1.1193456278580476e-05, + "loss": 0.9011, + "step": 41140 + }, + { + "epoch": 2.65, + "learning_rate": 1.1193241589557527e-05, + "loss": 0.9127, + "step": 41141 + }, + { + "epoch": 2.65, + "learning_rate": 1.1193026900534576e-05, + "loss": 0.8955, + "step": 41142 + }, + { + "epoch": 2.65, + "learning_rate": 1.1192812211511625e-05, + "loss": 0.9372, + "step": 41143 + }, + { + "epoch": 2.65, + "learning_rate": 1.1192597522488676e-05, + "loss": 0.9091, + "step": 41144 + }, + { + "epoch": 2.65, + "learning_rate": 1.1192382833465725e-05, + "loss": 0.9461, + "step": 41145 + }, + { + "epoch": 2.65, + "learning_rate": 1.1192168144442777e-05, + "loss": 0.9196, + "step": 41146 + }, + { + "epoch": 2.65, + "learning_rate": 1.1191953455419826e-05, + "loss": 0.9335, + "step": 41147 + }, + { + "epoch": 2.65, + "learning_rate": 1.1191738766396875e-05, + "loss": 0.911, + "step": 41148 + }, + { + "epoch": 2.65, + "learning_rate": 1.1191524077373926e-05, + "loss": 0.8656, + "step": 41149 + }, + { + "epoch": 2.65, + "learning_rate": 1.1191309388350975e-05, + "loss": 0.8689, + "step": 41150 + }, + { + "epoch": 2.65, + "learning_rate": 1.1191094699328025e-05, + "loss": 0.8936, + "step": 41151 + }, + { + "epoch": 2.65, + "learning_rate": 1.1190880010305074e-05, + "loss": 0.9583, + "step": 41152 + }, + { + "epoch": 2.65, + "learning_rate": 1.1190665321282123e-05, + "loss": 0.8537, + "step": 41153 + }, + { + "epoch": 2.65, + "learning_rate": 1.1190450632259174e-05, + "loss": 0.8447, + "step": 41154 + }, + { + "epoch": 2.65, + "learning_rate": 1.1190235943236223e-05, + "loss": 0.9372, + "step": 41155 + }, + { + "epoch": 2.65, + "learning_rate": 1.1190021254213272e-05, + "loss": 0.9636, + "step": 41156 + }, + { + "epoch": 2.65, + "learning_rate": 1.1189806565190322e-05, + "loss": 0.9285, + "step": 41157 + }, + { + "epoch": 2.65, + "learning_rate": 1.1189591876167371e-05, + "loss": 0.862, + "step": 41158 + }, + { + "epoch": 2.65, + "learning_rate": 1.1189377187144422e-05, + "loss": 0.9216, + "step": 41159 + }, + { + "epoch": 2.65, + "learning_rate": 1.1189162498121473e-05, + "loss": 0.8911, + "step": 41160 + }, + { + "epoch": 2.65, + "learning_rate": 1.1188947809098523e-05, + "loss": 0.8623, + "step": 41161 + }, + { + "epoch": 2.65, + "learning_rate": 1.1188733120075572e-05, + "loss": 0.8731, + "step": 41162 + }, + { + "epoch": 2.65, + "learning_rate": 1.1188518431052621e-05, + "loss": 0.858, + "step": 41163 + }, + { + "epoch": 2.65, + "learning_rate": 1.1188303742029672e-05, + "loss": 0.908, + "step": 41164 + }, + { + "epoch": 2.65, + "learning_rate": 1.1188089053006721e-05, + "loss": 0.9208, + "step": 41165 + }, + { + "epoch": 2.65, + "learning_rate": 1.118787436398377e-05, + "loss": 0.9534, + "step": 41166 + }, + { + "epoch": 2.65, + "learning_rate": 1.118765967496082e-05, + "loss": 0.9392, + "step": 41167 + }, + { + "epoch": 2.65, + "learning_rate": 1.118744498593787e-05, + "loss": 0.9396, + "step": 41168 + }, + { + "epoch": 2.65, + "learning_rate": 1.118723029691492e-05, + "loss": 0.8535, + "step": 41169 + }, + { + "epoch": 2.65, + "learning_rate": 1.1187015607891969e-05, + "loss": 0.9276, + "step": 41170 + }, + { + "epoch": 2.65, + "learning_rate": 1.1186800918869018e-05, + "loss": 0.8647, + "step": 41171 + }, + { + "epoch": 2.65, + "learning_rate": 1.1186586229846069e-05, + "loss": 0.8943, + "step": 41172 + }, + { + "epoch": 2.65, + "learning_rate": 1.1186371540823118e-05, + "loss": 0.9424, + "step": 41173 + }, + { + "epoch": 2.65, + "learning_rate": 1.1186156851800167e-05, + "loss": 0.946, + "step": 41174 + }, + { + "epoch": 2.65, + "learning_rate": 1.1185942162777219e-05, + "loss": 0.853, + "step": 41175 + }, + { + "epoch": 2.65, + "learning_rate": 1.1185727473754268e-05, + "loss": 0.8565, + "step": 41176 + }, + { + "epoch": 2.65, + "learning_rate": 1.1185512784731319e-05, + "loss": 0.7997, + "step": 41177 + }, + { + "epoch": 2.65, + "learning_rate": 1.1185298095708368e-05, + "loss": 0.8386, + "step": 41178 + }, + { + "epoch": 2.65, + "learning_rate": 1.1185083406685418e-05, + "loss": 0.816, + "step": 41179 + }, + { + "epoch": 2.65, + "learning_rate": 1.1184868717662467e-05, + "loss": 0.9952, + "step": 41180 + }, + { + "epoch": 2.65, + "learning_rate": 1.1184654028639516e-05, + "loss": 0.8779, + "step": 41181 + }, + { + "epoch": 2.65, + "learning_rate": 1.1184439339616567e-05, + "loss": 0.9194, + "step": 41182 + }, + { + "epoch": 2.65, + "learning_rate": 1.1184224650593616e-05, + "loss": 0.9542, + "step": 41183 + }, + { + "epoch": 2.65, + "learning_rate": 1.1184009961570665e-05, + "loss": 0.8653, + "step": 41184 + }, + { + "epoch": 2.65, + "learning_rate": 1.1183795272547715e-05, + "loss": 0.9234, + "step": 41185 + }, + { + "epoch": 2.65, + "learning_rate": 1.1183580583524764e-05, + "loss": 0.84, + "step": 41186 + }, + { + "epoch": 2.65, + "learning_rate": 1.1183365894501815e-05, + "loss": 0.8611, + "step": 41187 + }, + { + "epoch": 2.65, + "learning_rate": 1.1183151205478864e-05, + "loss": 0.9023, + "step": 41188 + }, + { + "epoch": 2.65, + "learning_rate": 1.1182936516455916e-05, + "loss": 0.9263, + "step": 41189 + }, + { + "epoch": 2.65, + "learning_rate": 1.1182721827432965e-05, + "loss": 0.8612, + "step": 41190 + }, + { + "epoch": 2.65, + "learning_rate": 1.1182507138410014e-05, + "loss": 0.8719, + "step": 41191 + }, + { + "epoch": 2.65, + "learning_rate": 1.1182292449387065e-05, + "loss": 0.9401, + "step": 41192 + }, + { + "epoch": 2.65, + "learning_rate": 1.1182077760364114e-05, + "loss": 0.9214, + "step": 41193 + }, + { + "epoch": 2.65, + "learning_rate": 1.1181863071341163e-05, + "loss": 0.8816, + "step": 41194 + }, + { + "epoch": 2.65, + "learning_rate": 1.1181648382318214e-05, + "loss": 0.8859, + "step": 41195 + }, + { + "epoch": 2.65, + "learning_rate": 1.1181433693295262e-05, + "loss": 0.8556, + "step": 41196 + }, + { + "epoch": 2.65, + "learning_rate": 1.1181219004272313e-05, + "loss": 0.9685, + "step": 41197 + }, + { + "epoch": 2.65, + "learning_rate": 1.1181004315249362e-05, + "loss": 0.8609, + "step": 41198 + }, + { + "epoch": 2.65, + "learning_rate": 1.1180789626226411e-05, + "loss": 0.9057, + "step": 41199 + }, + { + "epoch": 2.65, + "learning_rate": 1.1180574937203462e-05, + "loss": 0.8331, + "step": 41200 + }, + { + "epoch": 2.65, + "learning_rate": 1.118036024818051e-05, + "loss": 0.9652, + "step": 41201 + }, + { + "epoch": 2.65, + "learning_rate": 1.118014555915756e-05, + "loss": 0.9182, + "step": 41202 + }, + { + "epoch": 2.65, + "learning_rate": 1.117993087013461e-05, + "loss": 0.9128, + "step": 41203 + }, + { + "epoch": 2.65, + "learning_rate": 1.1179716181111661e-05, + "loss": 0.8771, + "step": 41204 + }, + { + "epoch": 2.65, + "learning_rate": 1.1179501492088712e-05, + "loss": 0.9064, + "step": 41205 + }, + { + "epoch": 2.65, + "learning_rate": 1.117928680306576e-05, + "loss": 0.9023, + "step": 41206 + }, + { + "epoch": 2.65, + "learning_rate": 1.1179072114042811e-05, + "loss": 0.936, + "step": 41207 + }, + { + "epoch": 2.65, + "learning_rate": 1.117885742501986e-05, + "loss": 0.7729, + "step": 41208 + }, + { + "epoch": 2.65, + "learning_rate": 1.117864273599691e-05, + "loss": 0.9207, + "step": 41209 + }, + { + "epoch": 2.65, + "learning_rate": 1.117842804697396e-05, + "loss": 0.938, + "step": 41210 + }, + { + "epoch": 2.65, + "learning_rate": 1.1178213357951009e-05, + "loss": 0.8693, + "step": 41211 + }, + { + "epoch": 2.65, + "learning_rate": 1.1177998668928058e-05, + "loss": 0.8851, + "step": 41212 + }, + { + "epoch": 2.65, + "learning_rate": 1.1177783979905108e-05, + "loss": 0.8971, + "step": 41213 + }, + { + "epoch": 2.65, + "learning_rate": 1.1177569290882157e-05, + "loss": 0.9308, + "step": 41214 + }, + { + "epoch": 2.65, + "learning_rate": 1.1177354601859206e-05, + "loss": 0.9356, + "step": 41215 + }, + { + "epoch": 2.65, + "learning_rate": 1.1177139912836257e-05, + "loss": 0.942, + "step": 41216 + }, + { + "epoch": 2.65, + "learning_rate": 1.1176925223813306e-05, + "loss": 0.8583, + "step": 41217 + }, + { + "epoch": 2.65, + "learning_rate": 1.1176710534790358e-05, + "loss": 0.9225, + "step": 41218 + }, + { + "epoch": 2.65, + "learning_rate": 1.1176495845767407e-05, + "loss": 0.9292, + "step": 41219 + }, + { + "epoch": 2.65, + "learning_rate": 1.1176281156744458e-05, + "loss": 0.9201, + "step": 41220 + }, + { + "epoch": 2.65, + "learning_rate": 1.1176066467721507e-05, + "loss": 0.8695, + "step": 41221 + }, + { + "epoch": 2.65, + "learning_rate": 1.1175851778698556e-05, + "loss": 0.848, + "step": 41222 + }, + { + "epoch": 2.65, + "learning_rate": 1.1175637089675607e-05, + "loss": 0.8607, + "step": 41223 + }, + { + "epoch": 2.65, + "learning_rate": 1.1175422400652656e-05, + "loss": 0.8647, + "step": 41224 + }, + { + "epoch": 2.65, + "learning_rate": 1.1175207711629704e-05, + "loss": 0.8497, + "step": 41225 + }, + { + "epoch": 2.65, + "learning_rate": 1.1174993022606755e-05, + "loss": 0.8585, + "step": 41226 + }, + { + "epoch": 2.65, + "learning_rate": 1.1174778333583804e-05, + "loss": 0.9303, + "step": 41227 + }, + { + "epoch": 2.65, + "learning_rate": 1.1174563644560855e-05, + "loss": 0.8978, + "step": 41228 + }, + { + "epoch": 2.65, + "learning_rate": 1.1174348955537904e-05, + "loss": 0.8788, + "step": 41229 + }, + { + "epoch": 2.65, + "learning_rate": 1.1174134266514953e-05, + "loss": 0.8867, + "step": 41230 + }, + { + "epoch": 2.65, + "learning_rate": 1.1173919577492003e-05, + "loss": 0.9004, + "step": 41231 + }, + { + "epoch": 2.65, + "learning_rate": 1.1173704888469052e-05, + "loss": 0.8885, + "step": 41232 + }, + { + "epoch": 2.65, + "learning_rate": 1.1173490199446105e-05, + "loss": 0.9148, + "step": 41233 + }, + { + "epoch": 2.65, + "learning_rate": 1.1173275510423154e-05, + "loss": 0.8655, + "step": 41234 + }, + { + "epoch": 2.65, + "learning_rate": 1.1173060821400203e-05, + "loss": 0.864, + "step": 41235 + }, + { + "epoch": 2.65, + "learning_rate": 1.1172846132377253e-05, + "loss": 0.8342, + "step": 41236 + }, + { + "epoch": 2.65, + "learning_rate": 1.1172631443354302e-05, + "loss": 0.8497, + "step": 41237 + }, + { + "epoch": 2.65, + "learning_rate": 1.1172416754331353e-05, + "loss": 0.8734, + "step": 41238 + }, + { + "epoch": 2.65, + "learning_rate": 1.1172202065308402e-05, + "loss": 0.8135, + "step": 41239 + }, + { + "epoch": 2.65, + "learning_rate": 1.117198737628545e-05, + "loss": 0.9121, + "step": 41240 + }, + { + "epoch": 2.65, + "learning_rate": 1.1171772687262501e-05, + "loss": 0.9019, + "step": 41241 + }, + { + "epoch": 2.65, + "learning_rate": 1.117155799823955e-05, + "loss": 0.9605, + "step": 41242 + }, + { + "epoch": 2.65, + "learning_rate": 1.11713433092166e-05, + "loss": 0.8886, + "step": 41243 + }, + { + "epoch": 2.65, + "learning_rate": 1.117112862019365e-05, + "loss": 0.8298, + "step": 41244 + }, + { + "epoch": 2.65, + "learning_rate": 1.1170913931170699e-05, + "loss": 0.9594, + "step": 41245 + }, + { + "epoch": 2.65, + "learning_rate": 1.117069924214775e-05, + "loss": 0.8974, + "step": 41246 + }, + { + "epoch": 2.65, + "learning_rate": 1.11704845531248e-05, + "loss": 0.9213, + "step": 41247 + }, + { + "epoch": 2.65, + "learning_rate": 1.1170269864101851e-05, + "loss": 0.9476, + "step": 41248 + }, + { + "epoch": 2.65, + "learning_rate": 1.11700551750789e-05, + "loss": 0.8773, + "step": 41249 + }, + { + "epoch": 2.65, + "learning_rate": 1.1169840486055949e-05, + "loss": 0.9149, + "step": 41250 + }, + { + "epoch": 2.65, + "learning_rate": 1.1169625797033e-05, + "loss": 0.9074, + "step": 41251 + }, + { + "epoch": 2.65, + "learning_rate": 1.1169411108010049e-05, + "loss": 0.8995, + "step": 41252 + }, + { + "epoch": 2.65, + "learning_rate": 1.1169196418987098e-05, + "loss": 0.9862, + "step": 41253 + }, + { + "epoch": 2.65, + "learning_rate": 1.1168981729964148e-05, + "loss": 0.8592, + "step": 41254 + }, + { + "epoch": 2.65, + "learning_rate": 1.1168767040941197e-05, + "loss": 0.8359, + "step": 41255 + }, + { + "epoch": 2.65, + "learning_rate": 1.1168552351918248e-05, + "loss": 0.8809, + "step": 41256 + }, + { + "epoch": 2.65, + "learning_rate": 1.1168337662895297e-05, + "loss": 0.8678, + "step": 41257 + }, + { + "epoch": 2.65, + "learning_rate": 1.1168122973872346e-05, + "loss": 0.8892, + "step": 41258 + }, + { + "epoch": 2.65, + "learning_rate": 1.1167908284849396e-05, + "loss": 0.9142, + "step": 41259 + }, + { + "epoch": 2.65, + "learning_rate": 1.1167693595826445e-05, + "loss": 0.9705, + "step": 41260 + }, + { + "epoch": 2.65, + "learning_rate": 1.1167478906803494e-05, + "loss": 0.8608, + "step": 41261 + }, + { + "epoch": 2.65, + "learning_rate": 1.1167264217780547e-05, + "loss": 0.8799, + "step": 41262 + }, + { + "epoch": 2.65, + "learning_rate": 1.1167049528757596e-05, + "loss": 0.9475, + "step": 41263 + }, + { + "epoch": 2.65, + "learning_rate": 1.1166834839734646e-05, + "loss": 0.911, + "step": 41264 + }, + { + "epoch": 2.65, + "learning_rate": 1.1166620150711695e-05, + "loss": 0.9216, + "step": 41265 + }, + { + "epoch": 2.65, + "learning_rate": 1.1166405461688746e-05, + "loss": 0.8478, + "step": 41266 + }, + { + "epoch": 2.66, + "learning_rate": 1.1166190772665795e-05, + "loss": 0.9379, + "step": 41267 + }, + { + "epoch": 2.66, + "learning_rate": 1.1165976083642844e-05, + "loss": 0.9237, + "step": 41268 + }, + { + "epoch": 2.66, + "learning_rate": 1.1165761394619895e-05, + "loss": 0.8796, + "step": 41269 + }, + { + "epoch": 2.66, + "learning_rate": 1.1165546705596943e-05, + "loss": 0.9206, + "step": 41270 + }, + { + "epoch": 2.66, + "learning_rate": 1.1165332016573992e-05, + "loss": 0.9545, + "step": 41271 + }, + { + "epoch": 2.66, + "learning_rate": 1.1165117327551043e-05, + "loss": 0.9603, + "step": 41272 + }, + { + "epoch": 2.66, + "learning_rate": 1.1164902638528092e-05, + "loss": 0.9097, + "step": 41273 + }, + { + "epoch": 2.66, + "learning_rate": 1.1164687949505143e-05, + "loss": 0.888, + "step": 41274 + }, + { + "epoch": 2.66, + "learning_rate": 1.1164473260482192e-05, + "loss": 0.8931, + "step": 41275 + }, + { + "epoch": 2.66, + "learning_rate": 1.1164258571459244e-05, + "loss": 0.863, + "step": 41276 + }, + { + "epoch": 2.66, + "learning_rate": 1.1164043882436293e-05, + "loss": 0.9205, + "step": 41277 + }, + { + "epoch": 2.66, + "learning_rate": 1.1163829193413342e-05, + "loss": 0.8428, + "step": 41278 + }, + { + "epoch": 2.66, + "learning_rate": 1.1163614504390393e-05, + "loss": 0.9198, + "step": 41279 + }, + { + "epoch": 2.66, + "learning_rate": 1.1163399815367442e-05, + "loss": 0.8299, + "step": 41280 + }, + { + "epoch": 2.66, + "learning_rate": 1.116318512634449e-05, + "loss": 0.9351, + "step": 41281 + }, + { + "epoch": 2.66, + "learning_rate": 1.1162970437321541e-05, + "loss": 0.8529, + "step": 41282 + }, + { + "epoch": 2.66, + "learning_rate": 1.116275574829859e-05, + "loss": 0.9719, + "step": 41283 + }, + { + "epoch": 2.66, + "learning_rate": 1.1162541059275641e-05, + "loss": 0.914, + "step": 41284 + }, + { + "epoch": 2.66, + "learning_rate": 1.116232637025269e-05, + "loss": 0.8776, + "step": 41285 + }, + { + "epoch": 2.66, + "learning_rate": 1.1162111681229739e-05, + "loss": 0.8944, + "step": 41286 + }, + { + "epoch": 2.66, + "learning_rate": 1.116189699220679e-05, + "loss": 0.9761, + "step": 41287 + }, + { + "epoch": 2.66, + "learning_rate": 1.1161682303183838e-05, + "loss": 0.9189, + "step": 41288 + }, + { + "epoch": 2.66, + "learning_rate": 1.1161467614160887e-05, + "loss": 0.8251, + "step": 41289 + }, + { + "epoch": 2.66, + "learning_rate": 1.1161252925137938e-05, + "loss": 0.9586, + "step": 41290 + }, + { + "epoch": 2.66, + "learning_rate": 1.1161038236114989e-05, + "loss": 0.9167, + "step": 41291 + }, + { + "epoch": 2.66, + "learning_rate": 1.116082354709204e-05, + "loss": 0.8751, + "step": 41292 + }, + { + "epoch": 2.66, + "learning_rate": 1.1160608858069088e-05, + "loss": 0.8756, + "step": 41293 + }, + { + "epoch": 2.66, + "learning_rate": 1.1160394169046139e-05, + "loss": 0.9097, + "step": 41294 + }, + { + "epoch": 2.66, + "learning_rate": 1.1160179480023188e-05, + "loss": 0.8933, + "step": 41295 + }, + { + "epoch": 2.66, + "learning_rate": 1.1159964791000237e-05, + "loss": 0.9044, + "step": 41296 + }, + { + "epoch": 2.66, + "learning_rate": 1.1159750101977288e-05, + "loss": 0.8499, + "step": 41297 + }, + { + "epoch": 2.66, + "learning_rate": 1.1159535412954337e-05, + "loss": 0.8429, + "step": 41298 + }, + { + "epoch": 2.66, + "learning_rate": 1.1159320723931385e-05, + "loss": 0.8857, + "step": 41299 + }, + { + "epoch": 2.66, + "learning_rate": 1.1159106034908436e-05, + "loss": 0.8434, + "step": 41300 + }, + { + "epoch": 2.66, + "learning_rate": 1.1158891345885485e-05, + "loss": 0.9393, + "step": 41301 + }, + { + "epoch": 2.66, + "learning_rate": 1.1158676656862534e-05, + "loss": 0.8371, + "step": 41302 + }, + { + "epoch": 2.66, + "learning_rate": 1.1158461967839585e-05, + "loss": 0.8516, + "step": 41303 + }, + { + "epoch": 2.66, + "learning_rate": 1.1158247278816634e-05, + "loss": 0.8437, + "step": 41304 + }, + { + "epoch": 2.66, + "learning_rate": 1.1158032589793686e-05, + "loss": 0.8464, + "step": 41305 + }, + { + "epoch": 2.66, + "learning_rate": 1.1157817900770735e-05, + "loss": 0.8752, + "step": 41306 + }, + { + "epoch": 2.66, + "learning_rate": 1.1157603211747786e-05, + "loss": 0.9022, + "step": 41307 + }, + { + "epoch": 2.66, + "learning_rate": 1.1157388522724835e-05, + "loss": 0.8492, + "step": 41308 + }, + { + "epoch": 2.66, + "learning_rate": 1.1157173833701884e-05, + "loss": 0.8463, + "step": 41309 + }, + { + "epoch": 2.66, + "learning_rate": 1.1156959144678934e-05, + "loss": 0.8862, + "step": 41310 + }, + { + "epoch": 2.66, + "learning_rate": 1.1156744455655983e-05, + "loss": 0.8746, + "step": 41311 + }, + { + "epoch": 2.66, + "learning_rate": 1.1156529766633032e-05, + "loss": 0.8711, + "step": 41312 + }, + { + "epoch": 2.66, + "learning_rate": 1.1156315077610083e-05, + "loss": 0.8675, + "step": 41313 + }, + { + "epoch": 2.66, + "learning_rate": 1.1156100388587132e-05, + "loss": 0.8541, + "step": 41314 + }, + { + "epoch": 2.66, + "learning_rate": 1.1155885699564182e-05, + "loss": 0.8865, + "step": 41315 + }, + { + "epoch": 2.66, + "learning_rate": 1.1155671010541231e-05, + "loss": 0.9209, + "step": 41316 + }, + { + "epoch": 2.66, + "learning_rate": 1.115545632151828e-05, + "loss": 0.9626, + "step": 41317 + }, + { + "epoch": 2.66, + "learning_rate": 1.1155241632495331e-05, + "loss": 0.9476, + "step": 41318 + }, + { + "epoch": 2.66, + "learning_rate": 1.115502694347238e-05, + "loss": 0.9411, + "step": 41319 + }, + { + "epoch": 2.66, + "learning_rate": 1.1154812254449432e-05, + "loss": 0.9118, + "step": 41320 + }, + { + "epoch": 2.66, + "learning_rate": 1.1154597565426481e-05, + "loss": 0.8903, + "step": 41321 + }, + { + "epoch": 2.66, + "learning_rate": 1.115438287640353e-05, + "loss": 0.9516, + "step": 41322 + }, + { + "epoch": 2.66, + "learning_rate": 1.1154168187380581e-05, + "loss": 0.8504, + "step": 41323 + }, + { + "epoch": 2.66, + "learning_rate": 1.115395349835763e-05, + "loss": 0.8877, + "step": 41324 + }, + { + "epoch": 2.66, + "learning_rate": 1.115373880933468e-05, + "loss": 0.9005, + "step": 41325 + }, + { + "epoch": 2.66, + "learning_rate": 1.115352412031173e-05, + "loss": 0.8987, + "step": 41326 + }, + { + "epoch": 2.66, + "learning_rate": 1.1153309431288779e-05, + "loss": 0.9129, + "step": 41327 + }, + { + "epoch": 2.66, + "learning_rate": 1.115309474226583e-05, + "loss": 0.9143, + "step": 41328 + }, + { + "epoch": 2.66, + "learning_rate": 1.1152880053242878e-05, + "loss": 0.9165, + "step": 41329 + }, + { + "epoch": 2.66, + "learning_rate": 1.1152665364219927e-05, + "loss": 1.0071, + "step": 41330 + }, + { + "epoch": 2.66, + "learning_rate": 1.1152450675196978e-05, + "loss": 0.9602, + "step": 41331 + }, + { + "epoch": 2.66, + "learning_rate": 1.1152235986174027e-05, + "loss": 0.8746, + "step": 41332 + }, + { + "epoch": 2.66, + "learning_rate": 1.1152021297151077e-05, + "loss": 0.8961, + "step": 41333 + }, + { + "epoch": 2.66, + "learning_rate": 1.1151806608128128e-05, + "loss": 0.8806, + "step": 41334 + }, + { + "epoch": 2.66, + "learning_rate": 1.1151591919105179e-05, + "loss": 0.8887, + "step": 41335 + }, + { + "epoch": 2.66, + "learning_rate": 1.1151377230082228e-05, + "loss": 0.9463, + "step": 41336 + }, + { + "epoch": 2.66, + "learning_rate": 1.1151162541059277e-05, + "loss": 0.9023, + "step": 41337 + }, + { + "epoch": 2.66, + "learning_rate": 1.1150947852036327e-05, + "loss": 0.8759, + "step": 41338 + }, + { + "epoch": 2.66, + "learning_rate": 1.1150733163013376e-05, + "loss": 0.8823, + "step": 41339 + }, + { + "epoch": 2.66, + "learning_rate": 1.1150518473990425e-05, + "loss": 0.9118, + "step": 41340 + }, + { + "epoch": 2.66, + "learning_rate": 1.1150303784967476e-05, + "loss": 0.9075, + "step": 41341 + }, + { + "epoch": 2.66, + "learning_rate": 1.1150089095944525e-05, + "loss": 0.8949, + "step": 41342 + }, + { + "epoch": 2.66, + "learning_rate": 1.1149874406921576e-05, + "loss": 0.8924, + "step": 41343 + }, + { + "epoch": 2.66, + "learning_rate": 1.1149659717898624e-05, + "loss": 0.9558, + "step": 41344 + }, + { + "epoch": 2.66, + "learning_rate": 1.1149445028875673e-05, + "loss": 0.9115, + "step": 41345 + }, + { + "epoch": 2.66, + "learning_rate": 1.1149230339852724e-05, + "loss": 0.8432, + "step": 41346 + }, + { + "epoch": 2.66, + "learning_rate": 1.1149015650829773e-05, + "loss": 0.8982, + "step": 41347 + }, + { + "epoch": 2.66, + "learning_rate": 1.1148800961806822e-05, + "loss": 0.9538, + "step": 41348 + }, + { + "epoch": 2.66, + "learning_rate": 1.1148586272783874e-05, + "loss": 0.8939, + "step": 41349 + }, + { + "epoch": 2.66, + "learning_rate": 1.1148371583760923e-05, + "loss": 0.8913, + "step": 41350 + }, + { + "epoch": 2.66, + "learning_rate": 1.1148156894737974e-05, + "loss": 0.8468, + "step": 41351 + }, + { + "epoch": 2.66, + "learning_rate": 1.1147942205715023e-05, + "loss": 0.8891, + "step": 41352 + }, + { + "epoch": 2.66, + "learning_rate": 1.1147727516692074e-05, + "loss": 0.8887, + "step": 41353 + }, + { + "epoch": 2.66, + "learning_rate": 1.1147512827669123e-05, + "loss": 0.9267, + "step": 41354 + }, + { + "epoch": 2.66, + "learning_rate": 1.1147298138646172e-05, + "loss": 0.8188, + "step": 41355 + }, + { + "epoch": 2.66, + "learning_rate": 1.1147083449623222e-05, + "loss": 0.9173, + "step": 41356 + }, + { + "epoch": 2.66, + "learning_rate": 1.1146868760600271e-05, + "loss": 0.8429, + "step": 41357 + }, + { + "epoch": 2.66, + "learning_rate": 1.114665407157732e-05, + "loss": 0.9918, + "step": 41358 + }, + { + "epoch": 2.66, + "learning_rate": 1.114643938255437e-05, + "loss": 0.8467, + "step": 41359 + }, + { + "epoch": 2.66, + "learning_rate": 1.114622469353142e-05, + "loss": 0.9812, + "step": 41360 + }, + { + "epoch": 2.66, + "learning_rate": 1.1146010004508469e-05, + "loss": 0.9155, + "step": 41361 + }, + { + "epoch": 2.66, + "learning_rate": 1.114579531548552e-05, + "loss": 0.902, + "step": 41362 + }, + { + "epoch": 2.66, + "learning_rate": 1.1145580626462568e-05, + "loss": 0.9512, + "step": 41363 + }, + { + "epoch": 2.66, + "learning_rate": 1.114536593743962e-05, + "loss": 0.8884, + "step": 41364 + }, + { + "epoch": 2.66, + "learning_rate": 1.114515124841667e-05, + "loss": 0.9093, + "step": 41365 + }, + { + "epoch": 2.66, + "learning_rate": 1.114493655939372e-05, + "loss": 0.9065, + "step": 41366 + }, + { + "epoch": 2.66, + "learning_rate": 1.114472187037077e-05, + "loss": 0.9275, + "step": 41367 + }, + { + "epoch": 2.66, + "learning_rate": 1.1144507181347818e-05, + "loss": 0.9158, + "step": 41368 + }, + { + "epoch": 2.66, + "learning_rate": 1.1144292492324869e-05, + "loss": 0.8923, + "step": 41369 + }, + { + "epoch": 2.66, + "learning_rate": 1.1144077803301918e-05, + "loss": 0.8388, + "step": 41370 + }, + { + "epoch": 2.66, + "learning_rate": 1.1143863114278967e-05, + "loss": 0.9335, + "step": 41371 + }, + { + "epoch": 2.66, + "learning_rate": 1.1143648425256018e-05, + "loss": 0.8637, + "step": 41372 + }, + { + "epoch": 2.66, + "learning_rate": 1.1143433736233067e-05, + "loss": 0.9283, + "step": 41373 + }, + { + "epoch": 2.66, + "learning_rate": 1.1143219047210117e-05, + "loss": 0.9252, + "step": 41374 + }, + { + "epoch": 2.66, + "learning_rate": 1.1143004358187166e-05, + "loss": 0.9245, + "step": 41375 + }, + { + "epoch": 2.66, + "learning_rate": 1.1142789669164215e-05, + "loss": 0.8938, + "step": 41376 + }, + { + "epoch": 2.66, + "learning_rate": 1.1142574980141266e-05, + "loss": 0.8663, + "step": 41377 + }, + { + "epoch": 2.66, + "learning_rate": 1.1142360291118316e-05, + "loss": 0.8645, + "step": 41378 + }, + { + "epoch": 2.66, + "learning_rate": 1.1142145602095367e-05, + "loss": 0.9187, + "step": 41379 + }, + { + "epoch": 2.66, + "learning_rate": 1.1141930913072416e-05, + "loss": 0.939, + "step": 41380 + }, + { + "epoch": 2.66, + "learning_rate": 1.1141716224049467e-05, + "loss": 0.8975, + "step": 41381 + }, + { + "epoch": 2.66, + "learning_rate": 1.1141501535026516e-05, + "loss": 0.874, + "step": 41382 + }, + { + "epoch": 2.66, + "learning_rate": 1.1141286846003565e-05, + "loss": 0.858, + "step": 41383 + }, + { + "epoch": 2.66, + "learning_rate": 1.1141072156980615e-05, + "loss": 1.0011, + "step": 41384 + }, + { + "epoch": 2.66, + "learning_rate": 1.1140857467957664e-05, + "loss": 0.8496, + "step": 41385 + }, + { + "epoch": 2.66, + "learning_rate": 1.1140642778934713e-05, + "loss": 0.9204, + "step": 41386 + }, + { + "epoch": 2.66, + "learning_rate": 1.1140428089911764e-05, + "loss": 0.9608, + "step": 41387 + }, + { + "epoch": 2.66, + "learning_rate": 1.1140213400888813e-05, + "loss": 0.9275, + "step": 41388 + }, + { + "epoch": 2.66, + "learning_rate": 1.1139998711865862e-05, + "loss": 0.8418, + "step": 41389 + }, + { + "epoch": 2.66, + "learning_rate": 1.1139784022842912e-05, + "loss": 0.898, + "step": 41390 + }, + { + "epoch": 2.66, + "learning_rate": 1.1139569333819961e-05, + "loss": 0.921, + "step": 41391 + }, + { + "epoch": 2.66, + "learning_rate": 1.1139354644797012e-05, + "loss": 0.8767, + "step": 41392 + }, + { + "epoch": 2.66, + "learning_rate": 1.1139139955774063e-05, + "loss": 0.934, + "step": 41393 + }, + { + "epoch": 2.66, + "learning_rate": 1.1138925266751113e-05, + "loss": 0.8329, + "step": 41394 + }, + { + "epoch": 2.66, + "learning_rate": 1.1138710577728162e-05, + "loss": 0.8985, + "step": 41395 + }, + { + "epoch": 2.66, + "learning_rate": 1.1138495888705211e-05, + "loss": 0.9116, + "step": 41396 + }, + { + "epoch": 2.66, + "learning_rate": 1.1138281199682262e-05, + "loss": 0.9025, + "step": 41397 + }, + { + "epoch": 2.66, + "learning_rate": 1.1138066510659311e-05, + "loss": 0.9021, + "step": 41398 + }, + { + "epoch": 2.66, + "learning_rate": 1.113785182163636e-05, + "loss": 0.8477, + "step": 41399 + }, + { + "epoch": 2.66, + "learning_rate": 1.113763713261341e-05, + "loss": 0.7856, + "step": 41400 + }, + { + "epoch": 2.66, + "learning_rate": 1.113742244359046e-05, + "loss": 0.9148, + "step": 41401 + }, + { + "epoch": 2.66, + "learning_rate": 1.113720775456751e-05, + "loss": 0.9127, + "step": 41402 + }, + { + "epoch": 2.66, + "learning_rate": 1.113699306554456e-05, + "loss": 0.9077, + "step": 41403 + }, + { + "epoch": 2.66, + "learning_rate": 1.1136778376521608e-05, + "loss": 0.9506, + "step": 41404 + }, + { + "epoch": 2.66, + "learning_rate": 1.1136563687498659e-05, + "loss": 0.9408, + "step": 41405 + }, + { + "epoch": 2.66, + "learning_rate": 1.1136348998475708e-05, + "loss": 0.894, + "step": 41406 + }, + { + "epoch": 2.66, + "learning_rate": 1.113613430945276e-05, + "loss": 0.9023, + "step": 41407 + }, + { + "epoch": 2.66, + "learning_rate": 1.1135919620429809e-05, + "loss": 0.875, + "step": 41408 + }, + { + "epoch": 2.66, + "learning_rate": 1.1135704931406858e-05, + "loss": 0.9422, + "step": 41409 + }, + { + "epoch": 2.66, + "learning_rate": 1.1135490242383909e-05, + "loss": 0.8511, + "step": 41410 + }, + { + "epoch": 2.66, + "learning_rate": 1.1135275553360958e-05, + "loss": 0.9771, + "step": 41411 + }, + { + "epoch": 2.66, + "learning_rate": 1.1135060864338008e-05, + "loss": 0.9198, + "step": 41412 + }, + { + "epoch": 2.66, + "learning_rate": 1.1134846175315057e-05, + "loss": 0.94, + "step": 41413 + }, + { + "epoch": 2.66, + "learning_rate": 1.1134631486292106e-05, + "loss": 0.9374, + "step": 41414 + }, + { + "epoch": 2.66, + "learning_rate": 1.1134416797269157e-05, + "loss": 0.8659, + "step": 41415 + }, + { + "epoch": 2.66, + "learning_rate": 1.1134202108246206e-05, + "loss": 0.8728, + "step": 41416 + }, + { + "epoch": 2.66, + "learning_rate": 1.1133987419223255e-05, + "loss": 0.9431, + "step": 41417 + }, + { + "epoch": 2.66, + "learning_rate": 1.1133772730200305e-05, + "loss": 0.8983, + "step": 41418 + }, + { + "epoch": 2.66, + "learning_rate": 1.1133558041177354e-05, + "loss": 0.8421, + "step": 41419 + }, + { + "epoch": 2.66, + "learning_rate": 1.1133343352154405e-05, + "loss": 0.8391, + "step": 41420 + }, + { + "epoch": 2.66, + "learning_rate": 1.1133128663131454e-05, + "loss": 0.8616, + "step": 41421 + }, + { + "epoch": 2.66, + "learning_rate": 1.1132913974108506e-05, + "loss": 0.9113, + "step": 41422 + }, + { + "epoch": 2.67, + "learning_rate": 1.1132699285085555e-05, + "loss": 0.8493, + "step": 41423 + }, + { + "epoch": 2.67, + "learning_rate": 1.1132484596062604e-05, + "loss": 0.8674, + "step": 41424 + }, + { + "epoch": 2.67, + "learning_rate": 1.1132269907039655e-05, + "loss": 0.9736, + "step": 41425 + }, + { + "epoch": 2.67, + "learning_rate": 1.1132055218016704e-05, + "loss": 0.9214, + "step": 41426 + }, + { + "epoch": 2.67, + "learning_rate": 1.1131840528993753e-05, + "loss": 0.9219, + "step": 41427 + }, + { + "epoch": 2.67, + "learning_rate": 1.1131625839970804e-05, + "loss": 0.9595, + "step": 41428 + }, + { + "epoch": 2.67, + "learning_rate": 1.1131411150947853e-05, + "loss": 0.9289, + "step": 41429 + }, + { + "epoch": 2.67, + "learning_rate": 1.1131196461924903e-05, + "loss": 0.9203, + "step": 41430 + }, + { + "epoch": 2.67, + "learning_rate": 1.1130981772901952e-05, + "loss": 0.873, + "step": 41431 + }, + { + "epoch": 2.67, + "learning_rate": 1.1130767083879001e-05, + "loss": 0.9, + "step": 41432 + }, + { + "epoch": 2.67, + "learning_rate": 1.1130552394856052e-05, + "loss": 0.8716, + "step": 41433 + }, + { + "epoch": 2.67, + "learning_rate": 1.11303377058331e-05, + "loss": 0.8496, + "step": 41434 + }, + { + "epoch": 2.67, + "learning_rate": 1.113012301681015e-05, + "loss": 0.9662, + "step": 41435 + }, + { + "epoch": 2.67, + "learning_rate": 1.1129908327787202e-05, + "loss": 0.8313, + "step": 41436 + }, + { + "epoch": 2.67, + "learning_rate": 1.1129693638764251e-05, + "loss": 0.9126, + "step": 41437 + }, + { + "epoch": 2.67, + "learning_rate": 1.1129478949741302e-05, + "loss": 0.8382, + "step": 41438 + }, + { + "epoch": 2.67, + "learning_rate": 1.112926426071835e-05, + "loss": 0.9274, + "step": 41439 + }, + { + "epoch": 2.67, + "learning_rate": 1.1129049571695401e-05, + "loss": 0.8627, + "step": 41440 + }, + { + "epoch": 2.67, + "learning_rate": 1.112883488267245e-05, + "loss": 0.9074, + "step": 41441 + }, + { + "epoch": 2.67, + "learning_rate": 1.11286201936495e-05, + "loss": 0.9022, + "step": 41442 + }, + { + "epoch": 2.67, + "learning_rate": 1.112840550462655e-05, + "loss": 0.9426, + "step": 41443 + }, + { + "epoch": 2.67, + "learning_rate": 1.1128190815603599e-05, + "loss": 0.8925, + "step": 41444 + }, + { + "epoch": 2.67, + "learning_rate": 1.1127976126580648e-05, + "loss": 0.9454, + "step": 41445 + }, + { + "epoch": 2.67, + "learning_rate": 1.1127761437557699e-05, + "loss": 0.8743, + "step": 41446 + }, + { + "epoch": 2.67, + "learning_rate": 1.1127546748534748e-05, + "loss": 0.9967, + "step": 41447 + }, + { + "epoch": 2.67, + "learning_rate": 1.1127332059511796e-05, + "loss": 0.9787, + "step": 41448 + }, + { + "epoch": 2.67, + "learning_rate": 1.1127117370488847e-05, + "loss": 0.9077, + "step": 41449 + }, + { + "epoch": 2.67, + "learning_rate": 1.1126902681465896e-05, + "loss": 0.968, + "step": 41450 + }, + { + "epoch": 2.67, + "learning_rate": 1.1126687992442948e-05, + "loss": 0.8284, + "step": 41451 + }, + { + "epoch": 2.67, + "learning_rate": 1.1126473303419997e-05, + "loss": 0.9523, + "step": 41452 + }, + { + "epoch": 2.67, + "learning_rate": 1.1126258614397048e-05, + "loss": 0.8843, + "step": 41453 + }, + { + "epoch": 2.67, + "learning_rate": 1.1126043925374097e-05, + "loss": 0.969, + "step": 41454 + }, + { + "epoch": 2.67, + "learning_rate": 1.1125829236351146e-05, + "loss": 0.8766, + "step": 41455 + }, + { + "epoch": 2.67, + "learning_rate": 1.1125614547328197e-05, + "loss": 0.9235, + "step": 41456 + }, + { + "epoch": 2.67, + "learning_rate": 1.1125399858305246e-05, + "loss": 0.8933, + "step": 41457 + }, + { + "epoch": 2.67, + "learning_rate": 1.1125185169282295e-05, + "loss": 0.844, + "step": 41458 + }, + { + "epoch": 2.67, + "learning_rate": 1.1124970480259345e-05, + "loss": 0.9341, + "step": 41459 + }, + { + "epoch": 2.67, + "learning_rate": 1.1124755791236394e-05, + "loss": 0.9447, + "step": 41460 + }, + { + "epoch": 2.67, + "learning_rate": 1.1124541102213445e-05, + "loss": 0.8808, + "step": 41461 + }, + { + "epoch": 2.67, + "learning_rate": 1.1124326413190494e-05, + "loss": 0.9094, + "step": 41462 + }, + { + "epoch": 2.67, + "learning_rate": 1.1124111724167543e-05, + "loss": 0.9175, + "step": 41463 + }, + { + "epoch": 2.67, + "learning_rate": 1.1123897035144593e-05, + "loss": 0.8858, + "step": 41464 + }, + { + "epoch": 2.67, + "learning_rate": 1.1123682346121644e-05, + "loss": 0.8842, + "step": 41465 + }, + { + "epoch": 2.67, + "learning_rate": 1.1123467657098695e-05, + "loss": 0.8822, + "step": 41466 + }, + { + "epoch": 2.67, + "learning_rate": 1.1123252968075744e-05, + "loss": 0.896, + "step": 41467 + }, + { + "epoch": 2.67, + "learning_rate": 1.1123038279052793e-05, + "loss": 0.8706, + "step": 41468 + }, + { + "epoch": 2.67, + "learning_rate": 1.1122823590029843e-05, + "loss": 0.9152, + "step": 41469 + }, + { + "epoch": 2.67, + "learning_rate": 1.1122608901006892e-05, + "loss": 0.9266, + "step": 41470 + }, + { + "epoch": 2.67, + "learning_rate": 1.1122394211983943e-05, + "loss": 0.901, + "step": 41471 + }, + { + "epoch": 2.67, + "learning_rate": 1.1122179522960992e-05, + "loss": 0.8545, + "step": 41472 + }, + { + "epoch": 2.67, + "learning_rate": 1.1121964833938041e-05, + "loss": 0.8765, + "step": 41473 + }, + { + "epoch": 2.67, + "learning_rate": 1.1121750144915092e-05, + "loss": 0.8972, + "step": 41474 + }, + { + "epoch": 2.67, + "learning_rate": 1.112153545589214e-05, + "loss": 0.8999, + "step": 41475 + }, + { + "epoch": 2.67, + "learning_rate": 1.112132076686919e-05, + "loss": 0.872, + "step": 41476 + }, + { + "epoch": 2.67, + "learning_rate": 1.112110607784624e-05, + "loss": 0.9627, + "step": 41477 + }, + { + "epoch": 2.67, + "learning_rate": 1.1120891388823289e-05, + "loss": 0.9267, + "step": 41478 + }, + { + "epoch": 2.67, + "learning_rate": 1.112067669980034e-05, + "loss": 0.8485, + "step": 41479 + }, + { + "epoch": 2.67, + "learning_rate": 1.112046201077739e-05, + "loss": 0.897, + "step": 41480 + }, + { + "epoch": 2.67, + "learning_rate": 1.1120247321754441e-05, + "loss": 0.8416, + "step": 41481 + }, + { + "epoch": 2.67, + "learning_rate": 1.112003263273149e-05, + "loss": 0.9444, + "step": 41482 + }, + { + "epoch": 2.67, + "learning_rate": 1.1119817943708539e-05, + "loss": 0.8808, + "step": 41483 + }, + { + "epoch": 2.67, + "learning_rate": 1.111960325468559e-05, + "loss": 0.9245, + "step": 41484 + }, + { + "epoch": 2.67, + "learning_rate": 1.1119388565662639e-05, + "loss": 0.8884, + "step": 41485 + }, + { + "epoch": 2.67, + "learning_rate": 1.1119173876639688e-05, + "loss": 0.8841, + "step": 41486 + }, + { + "epoch": 2.67, + "learning_rate": 1.1118959187616738e-05, + "loss": 0.9269, + "step": 41487 + }, + { + "epoch": 2.67, + "learning_rate": 1.1118744498593787e-05, + "loss": 0.924, + "step": 41488 + }, + { + "epoch": 2.67, + "learning_rate": 1.1118529809570838e-05, + "loss": 0.8858, + "step": 41489 + }, + { + "epoch": 2.67, + "learning_rate": 1.1118315120547887e-05, + "loss": 0.9442, + "step": 41490 + }, + { + "epoch": 2.67, + "learning_rate": 1.1118100431524936e-05, + "loss": 0.882, + "step": 41491 + }, + { + "epoch": 2.67, + "learning_rate": 1.1117885742501987e-05, + "loss": 0.9347, + "step": 41492 + }, + { + "epoch": 2.67, + "learning_rate": 1.1117671053479035e-05, + "loss": 0.8959, + "step": 41493 + }, + { + "epoch": 2.67, + "learning_rate": 1.1117456364456088e-05, + "loss": 0.9429, + "step": 41494 + }, + { + "epoch": 2.67, + "learning_rate": 1.1117241675433137e-05, + "loss": 0.8881, + "step": 41495 + }, + { + "epoch": 2.67, + "learning_rate": 1.1117026986410186e-05, + "loss": 0.8958, + "step": 41496 + }, + { + "epoch": 2.67, + "learning_rate": 1.1116812297387236e-05, + "loss": 0.873, + "step": 41497 + }, + { + "epoch": 2.67, + "learning_rate": 1.1116597608364285e-05, + "loss": 0.8782, + "step": 41498 + }, + { + "epoch": 2.67, + "learning_rate": 1.1116382919341336e-05, + "loss": 0.8592, + "step": 41499 + }, + { + "epoch": 2.67, + "learning_rate": 1.1116168230318385e-05, + "loss": 0.9279, + "step": 41500 + }, + { + "epoch": 2.67, + "learning_rate": 1.1115953541295434e-05, + "loss": 0.8296, + "step": 41501 + }, + { + "epoch": 2.67, + "learning_rate": 1.1115738852272485e-05, + "loss": 0.9068, + "step": 41502 + }, + { + "epoch": 2.67, + "learning_rate": 1.1115524163249534e-05, + "loss": 0.8595, + "step": 41503 + }, + { + "epoch": 2.67, + "learning_rate": 1.1115309474226583e-05, + "loss": 0.8882, + "step": 41504 + }, + { + "epoch": 2.67, + "learning_rate": 1.1115094785203633e-05, + "loss": 0.862, + "step": 41505 + }, + { + "epoch": 2.67, + "learning_rate": 1.1114880096180682e-05, + "loss": 0.8592, + "step": 41506 + }, + { + "epoch": 2.67, + "learning_rate": 1.1114665407157733e-05, + "loss": 0.9419, + "step": 41507 + }, + { + "epoch": 2.67, + "learning_rate": 1.1114450718134782e-05, + "loss": 0.9469, + "step": 41508 + }, + { + "epoch": 2.67, + "learning_rate": 1.1114236029111834e-05, + "loss": 0.8828, + "step": 41509 + }, + { + "epoch": 2.67, + "learning_rate": 1.1114021340088883e-05, + "loss": 0.9119, + "step": 41510 + }, + { + "epoch": 2.67, + "learning_rate": 1.1113806651065932e-05, + "loss": 0.9457, + "step": 41511 + }, + { + "epoch": 2.67, + "learning_rate": 1.1113591962042983e-05, + "loss": 0.8538, + "step": 41512 + }, + { + "epoch": 2.67, + "learning_rate": 1.1113377273020032e-05, + "loss": 0.8431, + "step": 41513 + }, + { + "epoch": 2.67, + "learning_rate": 1.111316258399708e-05, + "loss": 0.8368, + "step": 41514 + }, + { + "epoch": 2.67, + "learning_rate": 1.1112947894974131e-05, + "loss": 0.9639, + "step": 41515 + }, + { + "epoch": 2.67, + "learning_rate": 1.111273320595118e-05, + "loss": 0.8829, + "step": 41516 + }, + { + "epoch": 2.67, + "learning_rate": 1.1112518516928231e-05, + "loss": 0.9341, + "step": 41517 + }, + { + "epoch": 2.67, + "learning_rate": 1.111230382790528e-05, + "loss": 0.8943, + "step": 41518 + }, + { + "epoch": 2.67, + "learning_rate": 1.1112089138882329e-05, + "loss": 0.9116, + "step": 41519 + }, + { + "epoch": 2.67, + "learning_rate": 1.111187444985938e-05, + "loss": 0.9489, + "step": 41520 + }, + { + "epoch": 2.67, + "learning_rate": 1.1111659760836429e-05, + "loss": 0.9295, + "step": 41521 + }, + { + "epoch": 2.67, + "learning_rate": 1.1111445071813477e-05, + "loss": 0.9146, + "step": 41522 + }, + { + "epoch": 2.67, + "learning_rate": 1.111123038279053e-05, + "loss": 0.9066, + "step": 41523 + }, + { + "epoch": 2.67, + "learning_rate": 1.1111015693767579e-05, + "loss": 0.8521, + "step": 41524 + }, + { + "epoch": 2.67, + "learning_rate": 1.111080100474463e-05, + "loss": 0.9089, + "step": 41525 + }, + { + "epoch": 2.67, + "learning_rate": 1.1110586315721678e-05, + "loss": 0.9201, + "step": 41526 + }, + { + "epoch": 2.67, + "learning_rate": 1.1110371626698729e-05, + "loss": 0.9154, + "step": 41527 + }, + { + "epoch": 2.67, + "learning_rate": 1.1110156937675778e-05, + "loss": 0.8979, + "step": 41528 + }, + { + "epoch": 2.67, + "learning_rate": 1.1109942248652827e-05, + "loss": 0.8773, + "step": 41529 + }, + { + "epoch": 2.67, + "learning_rate": 1.1109727559629878e-05, + "loss": 0.9189, + "step": 41530 + }, + { + "epoch": 2.67, + "learning_rate": 1.1109512870606927e-05, + "loss": 0.849, + "step": 41531 + }, + { + "epoch": 2.67, + "learning_rate": 1.1109298181583976e-05, + "loss": 0.854, + "step": 41532 + }, + { + "epoch": 2.67, + "learning_rate": 1.1109083492561026e-05, + "loss": 0.8731, + "step": 41533 + }, + { + "epoch": 2.67, + "learning_rate": 1.1108868803538075e-05, + "loss": 0.8534, + "step": 41534 + }, + { + "epoch": 2.67, + "learning_rate": 1.1108654114515124e-05, + "loss": 0.8948, + "step": 41535 + }, + { + "epoch": 2.67, + "learning_rate": 1.1108439425492175e-05, + "loss": 0.8299, + "step": 41536 + }, + { + "epoch": 2.67, + "learning_rate": 1.1108224736469224e-05, + "loss": 0.9144, + "step": 41537 + }, + { + "epoch": 2.67, + "learning_rate": 1.1108010047446276e-05, + "loss": 0.8602, + "step": 41538 + }, + { + "epoch": 2.67, + "learning_rate": 1.1107795358423325e-05, + "loss": 0.8806, + "step": 41539 + }, + { + "epoch": 2.67, + "learning_rate": 1.1107580669400376e-05, + "loss": 0.9026, + "step": 41540 + }, + { + "epoch": 2.67, + "learning_rate": 1.1107365980377425e-05, + "loss": 0.8736, + "step": 41541 + }, + { + "epoch": 2.67, + "learning_rate": 1.1107151291354474e-05, + "loss": 0.8565, + "step": 41542 + }, + { + "epoch": 2.67, + "learning_rate": 1.1106936602331524e-05, + "loss": 0.903, + "step": 41543 + }, + { + "epoch": 2.67, + "learning_rate": 1.1106721913308573e-05, + "loss": 0.9182, + "step": 41544 + }, + { + "epoch": 2.67, + "learning_rate": 1.1106507224285622e-05, + "loss": 0.9079, + "step": 41545 + }, + { + "epoch": 2.67, + "learning_rate": 1.1106292535262673e-05, + "loss": 0.8862, + "step": 41546 + }, + { + "epoch": 2.67, + "learning_rate": 1.1106077846239722e-05, + "loss": 0.9036, + "step": 41547 + }, + { + "epoch": 2.67, + "learning_rate": 1.1105863157216773e-05, + "loss": 0.8998, + "step": 41548 + }, + { + "epoch": 2.67, + "learning_rate": 1.1105648468193822e-05, + "loss": 0.879, + "step": 41549 + }, + { + "epoch": 2.67, + "learning_rate": 1.110543377917087e-05, + "loss": 0.8803, + "step": 41550 + }, + { + "epoch": 2.67, + "learning_rate": 1.1105219090147921e-05, + "loss": 0.9512, + "step": 41551 + }, + { + "epoch": 2.67, + "learning_rate": 1.1105004401124972e-05, + "loss": 0.9339, + "step": 41552 + }, + { + "epoch": 2.67, + "learning_rate": 1.1104789712102022e-05, + "loss": 0.864, + "step": 41553 + }, + { + "epoch": 2.67, + "learning_rate": 1.1104575023079071e-05, + "loss": 0.8777, + "step": 41554 + }, + { + "epoch": 2.67, + "learning_rate": 1.110436033405612e-05, + "loss": 0.8919, + "step": 41555 + }, + { + "epoch": 2.67, + "learning_rate": 1.1104145645033171e-05, + "loss": 0.9448, + "step": 41556 + }, + { + "epoch": 2.67, + "learning_rate": 1.110393095601022e-05, + "loss": 0.8868, + "step": 41557 + }, + { + "epoch": 2.67, + "learning_rate": 1.110371626698727e-05, + "loss": 0.8983, + "step": 41558 + }, + { + "epoch": 2.67, + "learning_rate": 1.110350157796432e-05, + "loss": 0.9522, + "step": 41559 + }, + { + "epoch": 2.67, + "learning_rate": 1.1103286888941369e-05, + "loss": 0.9621, + "step": 41560 + }, + { + "epoch": 2.67, + "learning_rate": 1.110307219991842e-05, + "loss": 0.9182, + "step": 41561 + }, + { + "epoch": 2.67, + "learning_rate": 1.1102857510895468e-05, + "loss": 0.8531, + "step": 41562 + }, + { + "epoch": 2.67, + "learning_rate": 1.1102642821872517e-05, + "loss": 0.9648, + "step": 41563 + }, + { + "epoch": 2.67, + "learning_rate": 1.1102428132849568e-05, + "loss": 0.9218, + "step": 41564 + }, + { + "epoch": 2.67, + "learning_rate": 1.1102213443826617e-05, + "loss": 0.8678, + "step": 41565 + }, + { + "epoch": 2.67, + "learning_rate": 1.1101998754803668e-05, + "loss": 0.8962, + "step": 41566 + }, + { + "epoch": 2.67, + "learning_rate": 1.1101784065780718e-05, + "loss": 0.9074, + "step": 41567 + }, + { + "epoch": 2.67, + "learning_rate": 1.1101569376757769e-05, + "loss": 0.8648, + "step": 41568 + }, + { + "epoch": 2.67, + "learning_rate": 1.1101354687734818e-05, + "loss": 0.8664, + "step": 41569 + }, + { + "epoch": 2.67, + "learning_rate": 1.1101139998711867e-05, + "loss": 0.9533, + "step": 41570 + }, + { + "epoch": 2.67, + "learning_rate": 1.1100925309688917e-05, + "loss": 0.8983, + "step": 41571 + }, + { + "epoch": 2.67, + "learning_rate": 1.1100710620665966e-05, + "loss": 0.9268, + "step": 41572 + }, + { + "epoch": 2.67, + "learning_rate": 1.1100495931643015e-05, + "loss": 0.8196, + "step": 41573 + }, + { + "epoch": 2.67, + "learning_rate": 1.1100281242620066e-05, + "loss": 0.9328, + "step": 41574 + }, + { + "epoch": 2.67, + "learning_rate": 1.1100066553597115e-05, + "loss": 0.9349, + "step": 41575 + }, + { + "epoch": 2.67, + "learning_rate": 1.1099851864574166e-05, + "loss": 0.8542, + "step": 41576 + }, + { + "epoch": 2.67, + "learning_rate": 1.1099637175551215e-05, + "loss": 0.8926, + "step": 41577 + }, + { + "epoch": 2.68, + "learning_rate": 1.1099422486528264e-05, + "loss": 0.8832, + "step": 41578 + }, + { + "epoch": 2.68, + "learning_rate": 1.1099207797505314e-05, + "loss": 0.9026, + "step": 41579 + }, + { + "epoch": 2.68, + "learning_rate": 1.1098993108482363e-05, + "loss": 0.9092, + "step": 41580 + }, + { + "epoch": 2.68, + "learning_rate": 1.1098778419459416e-05, + "loss": 0.8536, + "step": 41581 + }, + { + "epoch": 2.68, + "learning_rate": 1.1098563730436464e-05, + "loss": 0.9278, + "step": 41582 + }, + { + "epoch": 2.68, + "learning_rate": 1.1098349041413513e-05, + "loss": 0.9634, + "step": 41583 + }, + { + "epoch": 2.68, + "learning_rate": 1.1098134352390564e-05, + "loss": 0.8974, + "step": 41584 + }, + { + "epoch": 2.68, + "learning_rate": 1.1097919663367613e-05, + "loss": 0.8535, + "step": 41585 + }, + { + "epoch": 2.68, + "learning_rate": 1.1097704974344664e-05, + "loss": 0.9028, + "step": 41586 + }, + { + "epoch": 2.68, + "learning_rate": 1.1097490285321713e-05, + "loss": 0.9151, + "step": 41587 + }, + { + "epoch": 2.68, + "learning_rate": 1.1097275596298762e-05, + "loss": 0.8577, + "step": 41588 + }, + { + "epoch": 2.68, + "learning_rate": 1.1097060907275812e-05, + "loss": 0.93, + "step": 41589 + }, + { + "epoch": 2.68, + "learning_rate": 1.1096846218252861e-05, + "loss": 0.9513, + "step": 41590 + }, + { + "epoch": 2.68, + "learning_rate": 1.109663152922991e-05, + "loss": 0.8685, + "step": 41591 + }, + { + "epoch": 2.68, + "learning_rate": 1.1096416840206961e-05, + "loss": 0.859, + "step": 41592 + }, + { + "epoch": 2.68, + "learning_rate": 1.109620215118401e-05, + "loss": 0.929, + "step": 41593 + }, + { + "epoch": 2.68, + "learning_rate": 1.109598746216106e-05, + "loss": 0.9017, + "step": 41594 + }, + { + "epoch": 2.68, + "learning_rate": 1.109577277313811e-05, + "loss": 0.8812, + "step": 41595 + }, + { + "epoch": 2.68, + "learning_rate": 1.1095558084115162e-05, + "loss": 0.8961, + "step": 41596 + }, + { + "epoch": 2.68, + "learning_rate": 1.109534339509221e-05, + "loss": 0.913, + "step": 41597 + }, + { + "epoch": 2.68, + "learning_rate": 1.109512870606926e-05, + "loss": 0.8765, + "step": 41598 + }, + { + "epoch": 2.68, + "learning_rate": 1.109491401704631e-05, + "loss": 0.9345, + "step": 41599 + }, + { + "epoch": 2.68, + "learning_rate": 1.109469932802336e-05, + "loss": 0.9145, + "step": 41600 + }, + { + "epoch": 2.68, + "learning_rate": 1.1094484639000408e-05, + "loss": 0.9173, + "step": 41601 + }, + { + "epoch": 2.68, + "learning_rate": 1.1094269949977459e-05, + "loss": 0.9707, + "step": 41602 + }, + { + "epoch": 2.68, + "learning_rate": 1.1094055260954508e-05, + "loss": 0.9117, + "step": 41603 + }, + { + "epoch": 2.68, + "learning_rate": 1.1093840571931559e-05, + "loss": 0.8625, + "step": 41604 + }, + { + "epoch": 2.68, + "learning_rate": 1.1093625882908608e-05, + "loss": 0.8089, + "step": 41605 + }, + { + "epoch": 2.68, + "learning_rate": 1.1093411193885657e-05, + "loss": 0.9404, + "step": 41606 + }, + { + "epoch": 2.68, + "learning_rate": 1.1093196504862707e-05, + "loss": 0.8983, + "step": 41607 + }, + { + "epoch": 2.68, + "learning_rate": 1.1092981815839756e-05, + "loss": 0.8919, + "step": 41608 + }, + { + "epoch": 2.68, + "learning_rate": 1.1092767126816805e-05, + "loss": 0.8802, + "step": 41609 + }, + { + "epoch": 2.68, + "learning_rate": 1.1092552437793858e-05, + "loss": 0.8873, + "step": 41610 + }, + { + "epoch": 2.68, + "learning_rate": 1.1092337748770907e-05, + "loss": 0.8745, + "step": 41611 + }, + { + "epoch": 2.68, + "learning_rate": 1.1092123059747957e-05, + "loss": 0.9154, + "step": 41612 + }, + { + "epoch": 2.68, + "learning_rate": 1.1091908370725006e-05, + "loss": 1.0083, + "step": 41613 + }, + { + "epoch": 2.68, + "learning_rate": 1.1091693681702057e-05, + "loss": 0.9081, + "step": 41614 + }, + { + "epoch": 2.68, + "learning_rate": 1.1091478992679106e-05, + "loss": 0.95, + "step": 41615 + }, + { + "epoch": 2.68, + "learning_rate": 1.1091264303656155e-05, + "loss": 0.939, + "step": 41616 + }, + { + "epoch": 2.68, + "learning_rate": 1.1091049614633205e-05, + "loss": 0.9253, + "step": 41617 + }, + { + "epoch": 2.68, + "learning_rate": 1.1090834925610254e-05, + "loss": 0.9014, + "step": 41618 + }, + { + "epoch": 2.68, + "learning_rate": 1.1090620236587303e-05, + "loss": 0.8584, + "step": 41619 + }, + { + "epoch": 2.68, + "learning_rate": 1.1090405547564354e-05, + "loss": 0.8565, + "step": 41620 + }, + { + "epoch": 2.68, + "learning_rate": 1.1090190858541403e-05, + "loss": 0.9067, + "step": 41621 + }, + { + "epoch": 2.68, + "learning_rate": 1.1089976169518452e-05, + "loss": 0.9338, + "step": 41622 + }, + { + "epoch": 2.68, + "learning_rate": 1.1089761480495503e-05, + "loss": 0.8767, + "step": 41623 + }, + { + "epoch": 2.68, + "learning_rate": 1.1089546791472552e-05, + "loss": 0.8784, + "step": 41624 + }, + { + "epoch": 2.68, + "learning_rate": 1.1089332102449604e-05, + "loss": 0.8481, + "step": 41625 + }, + { + "epoch": 2.68, + "learning_rate": 1.1089117413426653e-05, + "loss": 0.8782, + "step": 41626 + }, + { + "epoch": 2.68, + "learning_rate": 1.1088902724403703e-05, + "loss": 0.8872, + "step": 41627 + }, + { + "epoch": 2.68, + "learning_rate": 1.1088688035380752e-05, + "loss": 0.8931, + "step": 41628 + }, + { + "epoch": 2.68, + "learning_rate": 1.1088473346357801e-05, + "loss": 0.903, + "step": 41629 + }, + { + "epoch": 2.68, + "learning_rate": 1.1088258657334852e-05, + "loss": 0.8585, + "step": 41630 + }, + { + "epoch": 2.68, + "learning_rate": 1.1088043968311901e-05, + "loss": 0.9002, + "step": 41631 + }, + { + "epoch": 2.68, + "learning_rate": 1.108782927928895e-05, + "loss": 0.8339, + "step": 41632 + }, + { + "epoch": 2.68, + "learning_rate": 1.1087614590266e-05, + "loss": 0.939, + "step": 41633 + }, + { + "epoch": 2.68, + "learning_rate": 1.108739990124305e-05, + "loss": 0.9301, + "step": 41634 + }, + { + "epoch": 2.68, + "learning_rate": 1.10871852122201e-05, + "loss": 0.9165, + "step": 41635 + }, + { + "epoch": 2.68, + "learning_rate": 1.108697052319715e-05, + "loss": 0.9281, + "step": 41636 + }, + { + "epoch": 2.68, + "learning_rate": 1.1086755834174198e-05, + "loss": 0.8846, + "step": 41637 + }, + { + "epoch": 2.68, + "learning_rate": 1.1086541145151249e-05, + "loss": 0.842, + "step": 41638 + }, + { + "epoch": 2.68, + "learning_rate": 1.10863264561283e-05, + "loss": 0.9535, + "step": 41639 + }, + { + "epoch": 2.68, + "learning_rate": 1.108611176710535e-05, + "loss": 0.9462, + "step": 41640 + }, + { + "epoch": 2.68, + "learning_rate": 1.10858970780824e-05, + "loss": 0.8711, + "step": 41641 + }, + { + "epoch": 2.68, + "learning_rate": 1.1085682389059448e-05, + "loss": 0.8644, + "step": 41642 + }, + { + "epoch": 2.68, + "learning_rate": 1.1085467700036499e-05, + "loss": 0.9424, + "step": 41643 + }, + { + "epoch": 2.68, + "learning_rate": 1.1085253011013548e-05, + "loss": 0.8321, + "step": 41644 + }, + { + "epoch": 2.68, + "learning_rate": 1.1085038321990598e-05, + "loss": 0.9124, + "step": 41645 + }, + { + "epoch": 2.68, + "learning_rate": 1.1084823632967647e-05, + "loss": 0.8557, + "step": 41646 + }, + { + "epoch": 2.68, + "learning_rate": 1.1084608943944696e-05, + "loss": 0.9194, + "step": 41647 + }, + { + "epoch": 2.68, + "learning_rate": 1.1084394254921747e-05, + "loss": 0.8694, + "step": 41648 + }, + { + "epoch": 2.68, + "learning_rate": 1.1084179565898796e-05, + "loss": 0.9221, + "step": 41649 + }, + { + "epoch": 2.68, + "learning_rate": 1.1083964876875845e-05, + "loss": 0.9066, + "step": 41650 + }, + { + "epoch": 2.68, + "learning_rate": 1.1083750187852896e-05, + "loss": 0.9028, + "step": 41651 + }, + { + "epoch": 2.68, + "learning_rate": 1.1083535498829945e-05, + "loss": 0.8545, + "step": 41652 + }, + { + "epoch": 2.68, + "learning_rate": 1.1083320809806995e-05, + "loss": 0.8553, + "step": 41653 + }, + { + "epoch": 2.68, + "learning_rate": 1.1083106120784046e-05, + "loss": 0.8144, + "step": 41654 + }, + { + "epoch": 2.68, + "learning_rate": 1.1082891431761097e-05, + "loss": 0.878, + "step": 41655 + }, + { + "epoch": 2.68, + "learning_rate": 1.1082676742738146e-05, + "loss": 0.8374, + "step": 41656 + }, + { + "epoch": 2.68, + "learning_rate": 1.1082462053715194e-05, + "loss": 0.8287, + "step": 41657 + }, + { + "epoch": 2.68, + "learning_rate": 1.1082247364692245e-05, + "loss": 0.8843, + "step": 41658 + }, + { + "epoch": 2.68, + "learning_rate": 1.1082032675669294e-05, + "loss": 0.8852, + "step": 41659 + }, + { + "epoch": 2.68, + "learning_rate": 1.1081817986646343e-05, + "loss": 1.0007, + "step": 41660 + }, + { + "epoch": 2.68, + "learning_rate": 1.1081603297623394e-05, + "loss": 0.8366, + "step": 41661 + }, + { + "epoch": 2.68, + "learning_rate": 1.1081388608600443e-05, + "loss": 0.8992, + "step": 41662 + }, + { + "epoch": 2.68, + "learning_rate": 1.1081173919577493e-05, + "loss": 0.8755, + "step": 41663 + }, + { + "epoch": 2.68, + "learning_rate": 1.1080959230554542e-05, + "loss": 0.9262, + "step": 41664 + }, + { + "epoch": 2.68, + "learning_rate": 1.1080744541531591e-05, + "loss": 0.9198, + "step": 41665 + }, + { + "epoch": 2.68, + "learning_rate": 1.1080529852508642e-05, + "loss": 0.8717, + "step": 41666 + }, + { + "epoch": 2.68, + "learning_rate": 1.1080315163485691e-05, + "loss": 0.8609, + "step": 41667 + }, + { + "epoch": 2.68, + "learning_rate": 1.1080100474462743e-05, + "loss": 0.9418, + "step": 41668 + }, + { + "epoch": 2.68, + "learning_rate": 1.1079885785439792e-05, + "loss": 0.8824, + "step": 41669 + }, + { + "epoch": 2.68, + "learning_rate": 1.1079671096416841e-05, + "loss": 0.8592, + "step": 41670 + }, + { + "epoch": 2.68, + "learning_rate": 1.1079456407393892e-05, + "loss": 0.9356, + "step": 41671 + }, + { + "epoch": 2.68, + "learning_rate": 1.107924171837094e-05, + "loss": 0.9003, + "step": 41672 + }, + { + "epoch": 2.68, + "learning_rate": 1.1079027029347991e-05, + "loss": 0.9387, + "step": 41673 + }, + { + "epoch": 2.68, + "learning_rate": 1.107881234032504e-05, + "loss": 0.8639, + "step": 41674 + }, + { + "epoch": 2.68, + "learning_rate": 1.107859765130209e-05, + "loss": 0.8707, + "step": 41675 + }, + { + "epoch": 2.68, + "learning_rate": 1.107838296227914e-05, + "loss": 0.962, + "step": 41676 + }, + { + "epoch": 2.68, + "learning_rate": 1.1078168273256189e-05, + "loss": 0.8868, + "step": 41677 + }, + { + "epoch": 2.68, + "learning_rate": 1.1077953584233238e-05, + "loss": 0.9379, + "step": 41678 + }, + { + "epoch": 2.68, + "learning_rate": 1.1077738895210289e-05, + "loss": 0.9263, + "step": 41679 + }, + { + "epoch": 2.68, + "learning_rate": 1.1077524206187338e-05, + "loss": 0.8816, + "step": 41680 + }, + { + "epoch": 2.68, + "learning_rate": 1.1077309517164387e-05, + "loss": 0.8871, + "step": 41681 + }, + { + "epoch": 2.68, + "learning_rate": 1.1077094828141437e-05, + "loss": 0.947, + "step": 41682 + }, + { + "epoch": 2.68, + "learning_rate": 1.107688013911849e-05, + "loss": 0.9107, + "step": 41683 + }, + { + "epoch": 2.68, + "learning_rate": 1.1076665450095539e-05, + "loss": 0.9038, + "step": 41684 + }, + { + "epoch": 2.68, + "learning_rate": 1.1076450761072588e-05, + "loss": 0.834, + "step": 41685 + }, + { + "epoch": 2.68, + "learning_rate": 1.1076236072049638e-05, + "loss": 0.8841, + "step": 41686 + }, + { + "epoch": 2.68, + "learning_rate": 1.1076021383026687e-05, + "loss": 0.8767, + "step": 41687 + }, + { + "epoch": 2.68, + "learning_rate": 1.1075806694003736e-05, + "loss": 0.913, + "step": 41688 + }, + { + "epoch": 2.68, + "learning_rate": 1.1075592004980787e-05, + "loss": 0.8971, + "step": 41689 + }, + { + "epoch": 2.68, + "learning_rate": 1.1075377315957836e-05, + "loss": 0.8932, + "step": 41690 + }, + { + "epoch": 2.68, + "learning_rate": 1.1075162626934885e-05, + "loss": 0.8064, + "step": 41691 + }, + { + "epoch": 2.68, + "learning_rate": 1.1074947937911935e-05, + "loss": 0.8339, + "step": 41692 + }, + { + "epoch": 2.68, + "learning_rate": 1.1074733248888984e-05, + "loss": 0.8863, + "step": 41693 + }, + { + "epoch": 2.68, + "learning_rate": 1.1074518559866035e-05, + "loss": 0.9001, + "step": 41694 + }, + { + "epoch": 2.68, + "learning_rate": 1.1074303870843084e-05, + "loss": 0.8833, + "step": 41695 + }, + { + "epoch": 2.68, + "learning_rate": 1.1074089181820133e-05, + "loss": 0.8721, + "step": 41696 + }, + { + "epoch": 2.68, + "learning_rate": 1.1073874492797185e-05, + "loss": 0.9313, + "step": 41697 + }, + { + "epoch": 2.68, + "learning_rate": 1.1073659803774234e-05, + "loss": 0.9432, + "step": 41698 + }, + { + "epoch": 2.68, + "learning_rate": 1.1073445114751285e-05, + "loss": 0.9653, + "step": 41699 + }, + { + "epoch": 2.68, + "learning_rate": 1.1073230425728334e-05, + "loss": 0.8758, + "step": 41700 + }, + { + "epoch": 2.68, + "learning_rate": 1.1073015736705384e-05, + "loss": 0.9743, + "step": 41701 + }, + { + "epoch": 2.68, + "learning_rate": 1.1072801047682433e-05, + "loss": 0.88, + "step": 41702 + }, + { + "epoch": 2.68, + "learning_rate": 1.1072586358659482e-05, + "loss": 0.8846, + "step": 41703 + }, + { + "epoch": 2.68, + "learning_rate": 1.1072371669636533e-05, + "loss": 0.9208, + "step": 41704 + }, + { + "epoch": 2.68, + "learning_rate": 1.1072156980613582e-05, + "loss": 0.8362, + "step": 41705 + }, + { + "epoch": 2.68, + "learning_rate": 1.1071942291590631e-05, + "loss": 0.9536, + "step": 41706 + }, + { + "epoch": 2.68, + "learning_rate": 1.1071727602567682e-05, + "loss": 0.8482, + "step": 41707 + }, + { + "epoch": 2.68, + "learning_rate": 1.107151291354473e-05, + "loss": 0.9416, + "step": 41708 + }, + { + "epoch": 2.68, + "learning_rate": 1.107129822452178e-05, + "loss": 0.9102, + "step": 41709 + }, + { + "epoch": 2.68, + "learning_rate": 1.107108353549883e-05, + "loss": 0.9485, + "step": 41710 + }, + { + "epoch": 2.68, + "learning_rate": 1.107086884647588e-05, + "loss": 0.886, + "step": 41711 + }, + { + "epoch": 2.68, + "learning_rate": 1.1070654157452932e-05, + "loss": 0.8913, + "step": 41712 + }, + { + "epoch": 2.68, + "learning_rate": 1.107043946842998e-05, + "loss": 0.8516, + "step": 41713 + }, + { + "epoch": 2.68, + "learning_rate": 1.1070224779407031e-05, + "loss": 0.878, + "step": 41714 + }, + { + "epoch": 2.68, + "learning_rate": 1.107001009038408e-05, + "loss": 0.8855, + "step": 41715 + }, + { + "epoch": 2.68, + "learning_rate": 1.1069795401361129e-05, + "loss": 0.9352, + "step": 41716 + }, + { + "epoch": 2.68, + "learning_rate": 1.106958071233818e-05, + "loss": 0.8452, + "step": 41717 + }, + { + "epoch": 2.68, + "learning_rate": 1.1069366023315229e-05, + "loss": 0.8969, + "step": 41718 + }, + { + "epoch": 2.68, + "learning_rate": 1.1069151334292278e-05, + "loss": 0.9424, + "step": 41719 + }, + { + "epoch": 2.68, + "learning_rate": 1.1068936645269328e-05, + "loss": 0.8304, + "step": 41720 + }, + { + "epoch": 2.68, + "learning_rate": 1.1068721956246377e-05, + "loss": 0.9232, + "step": 41721 + }, + { + "epoch": 2.68, + "learning_rate": 1.1068507267223428e-05, + "loss": 0.9387, + "step": 41722 + }, + { + "epoch": 2.68, + "learning_rate": 1.1068292578200477e-05, + "loss": 0.9194, + "step": 41723 + }, + { + "epoch": 2.68, + "learning_rate": 1.1068077889177526e-05, + "loss": 0.8761, + "step": 41724 + }, + { + "epoch": 2.68, + "learning_rate": 1.1067863200154577e-05, + "loss": 0.9513, + "step": 41725 + }, + { + "epoch": 2.68, + "learning_rate": 1.1067648511131626e-05, + "loss": 0.9476, + "step": 41726 + }, + { + "epoch": 2.68, + "learning_rate": 1.1067433822108678e-05, + "loss": 0.8467, + "step": 41727 + }, + { + "epoch": 2.68, + "learning_rate": 1.1067219133085727e-05, + "loss": 0.8992, + "step": 41728 + }, + { + "epoch": 2.68, + "learning_rate": 1.1067004444062776e-05, + "loss": 0.9301, + "step": 41729 + }, + { + "epoch": 2.68, + "learning_rate": 1.1066789755039827e-05, + "loss": 0.8708, + "step": 41730 + }, + { + "epoch": 2.68, + "learning_rate": 1.1066575066016875e-05, + "loss": 0.9165, + "step": 41731 + }, + { + "epoch": 2.68, + "learning_rate": 1.1066360376993926e-05, + "loss": 0.8972, + "step": 41732 + }, + { + "epoch": 2.69, + "learning_rate": 1.1066145687970975e-05, + "loss": 0.8634, + "step": 41733 + }, + { + "epoch": 2.69, + "learning_rate": 1.1065930998948024e-05, + "loss": 0.8537, + "step": 41734 + }, + { + "epoch": 2.69, + "learning_rate": 1.1065716309925075e-05, + "loss": 0.9499, + "step": 41735 + }, + { + "epoch": 2.69, + "learning_rate": 1.1065501620902124e-05, + "loss": 0.8742, + "step": 41736 + }, + { + "epoch": 2.69, + "learning_rate": 1.1065286931879173e-05, + "loss": 0.914, + "step": 41737 + }, + { + "epoch": 2.69, + "learning_rate": 1.1065072242856223e-05, + "loss": 0.9141, + "step": 41738 + }, + { + "epoch": 2.69, + "learning_rate": 1.1064857553833272e-05, + "loss": 0.8557, + "step": 41739 + }, + { + "epoch": 2.69, + "learning_rate": 1.1064642864810323e-05, + "loss": 0.9394, + "step": 41740 + }, + { + "epoch": 2.69, + "learning_rate": 1.1064428175787374e-05, + "loss": 0.9165, + "step": 41741 + }, + { + "epoch": 2.69, + "learning_rate": 1.1064213486764424e-05, + "loss": 0.8817, + "step": 41742 + }, + { + "epoch": 2.69, + "learning_rate": 1.1063998797741473e-05, + "loss": 0.9285, + "step": 41743 + }, + { + "epoch": 2.69, + "learning_rate": 1.1063784108718522e-05, + "loss": 0.8695, + "step": 41744 + }, + { + "epoch": 2.69, + "learning_rate": 1.1063569419695573e-05, + "loss": 0.8732, + "step": 41745 + }, + { + "epoch": 2.69, + "learning_rate": 1.1063354730672622e-05, + "loss": 0.9058, + "step": 41746 + }, + { + "epoch": 2.69, + "learning_rate": 1.106314004164967e-05, + "loss": 0.9236, + "step": 41747 + }, + { + "epoch": 2.69, + "learning_rate": 1.1062925352626721e-05, + "loss": 0.8551, + "step": 41748 + }, + { + "epoch": 2.69, + "learning_rate": 1.106271066360377e-05, + "loss": 0.9515, + "step": 41749 + }, + { + "epoch": 2.69, + "learning_rate": 1.1062495974580821e-05, + "loss": 0.9107, + "step": 41750 + }, + { + "epoch": 2.69, + "learning_rate": 1.106228128555787e-05, + "loss": 0.9652, + "step": 41751 + }, + { + "epoch": 2.69, + "learning_rate": 1.1062066596534919e-05, + "loss": 0.8881, + "step": 41752 + }, + { + "epoch": 2.69, + "learning_rate": 1.106185190751197e-05, + "loss": 0.8779, + "step": 41753 + }, + { + "epoch": 2.69, + "learning_rate": 1.1061637218489019e-05, + "loss": 0.8712, + "step": 41754 + }, + { + "epoch": 2.69, + "learning_rate": 1.1061422529466068e-05, + "loss": 0.8927, + "step": 41755 + }, + { + "epoch": 2.69, + "learning_rate": 1.106120784044312e-05, + "loss": 0.9323, + "step": 41756 + }, + { + "epoch": 2.69, + "learning_rate": 1.1060993151420169e-05, + "loss": 0.8485, + "step": 41757 + }, + { + "epoch": 2.69, + "learning_rate": 1.106077846239722e-05, + "loss": 0.8289, + "step": 41758 + }, + { + "epoch": 2.69, + "learning_rate": 1.1060563773374269e-05, + "loss": 0.88, + "step": 41759 + }, + { + "epoch": 2.69, + "learning_rate": 1.106034908435132e-05, + "loss": 0.9064, + "step": 41760 + }, + { + "epoch": 2.69, + "learning_rate": 1.1060134395328368e-05, + "loss": 0.944, + "step": 41761 + }, + { + "epoch": 2.69, + "learning_rate": 1.1059919706305417e-05, + "loss": 0.9147, + "step": 41762 + }, + { + "epoch": 2.69, + "learning_rate": 1.1059705017282468e-05, + "loss": 0.8915, + "step": 41763 + }, + { + "epoch": 2.69, + "learning_rate": 1.1059490328259517e-05, + "loss": 0.8848, + "step": 41764 + }, + { + "epoch": 2.69, + "learning_rate": 1.1059275639236566e-05, + "loss": 0.9125, + "step": 41765 + }, + { + "epoch": 2.69, + "learning_rate": 1.1059060950213616e-05, + "loss": 0.8151, + "step": 41766 + }, + { + "epoch": 2.69, + "learning_rate": 1.1058846261190665e-05, + "loss": 0.9833, + "step": 41767 + }, + { + "epoch": 2.69, + "learning_rate": 1.1058631572167714e-05, + "loss": 0.8536, + "step": 41768 + }, + { + "epoch": 2.69, + "learning_rate": 1.1058416883144765e-05, + "loss": 0.8988, + "step": 41769 + }, + { + "epoch": 2.69, + "learning_rate": 1.1058202194121817e-05, + "loss": 0.927, + "step": 41770 + }, + { + "epoch": 2.69, + "learning_rate": 1.1057987505098866e-05, + "loss": 0.892, + "step": 41771 + }, + { + "epoch": 2.69, + "learning_rate": 1.1057772816075915e-05, + "loss": 0.8511, + "step": 41772 + }, + { + "epoch": 2.69, + "learning_rate": 1.1057558127052966e-05, + "loss": 0.8979, + "step": 41773 + }, + { + "epoch": 2.69, + "learning_rate": 1.1057343438030015e-05, + "loss": 0.949, + "step": 41774 + }, + { + "epoch": 2.69, + "learning_rate": 1.1057128749007064e-05, + "loss": 0.9076, + "step": 41775 + }, + { + "epoch": 2.69, + "learning_rate": 1.1056914059984114e-05, + "loss": 0.8831, + "step": 41776 + }, + { + "epoch": 2.69, + "learning_rate": 1.1056699370961163e-05, + "loss": 0.8514, + "step": 41777 + }, + { + "epoch": 2.69, + "learning_rate": 1.1056484681938212e-05, + "loss": 0.9107, + "step": 41778 + }, + { + "epoch": 2.69, + "learning_rate": 1.1056269992915263e-05, + "loss": 0.945, + "step": 41779 + }, + { + "epoch": 2.69, + "learning_rate": 1.1056055303892312e-05, + "loss": 0.9299, + "step": 41780 + }, + { + "epoch": 2.69, + "learning_rate": 1.1055840614869363e-05, + "loss": 0.8787, + "step": 41781 + }, + { + "epoch": 2.69, + "learning_rate": 1.1055625925846412e-05, + "loss": 0.9731, + "step": 41782 + }, + { + "epoch": 2.69, + "learning_rate": 1.105541123682346e-05, + "loss": 0.9172, + "step": 41783 + }, + { + "epoch": 2.69, + "learning_rate": 1.1055196547800511e-05, + "loss": 0.9217, + "step": 41784 + }, + { + "epoch": 2.69, + "learning_rate": 1.1054981858777562e-05, + "loss": 0.8144, + "step": 41785 + }, + { + "epoch": 2.69, + "learning_rate": 1.1054767169754613e-05, + "loss": 0.9567, + "step": 41786 + }, + { + "epoch": 2.69, + "learning_rate": 1.1054552480731662e-05, + "loss": 0.9601, + "step": 41787 + }, + { + "epoch": 2.69, + "learning_rate": 1.105433779170871e-05, + "loss": 0.8819, + "step": 41788 + }, + { + "epoch": 2.69, + "learning_rate": 1.1054123102685761e-05, + "loss": 0.826, + "step": 41789 + }, + { + "epoch": 2.69, + "learning_rate": 1.105390841366281e-05, + "loss": 0.8884, + "step": 41790 + }, + { + "epoch": 2.69, + "learning_rate": 1.105369372463986e-05, + "loss": 0.9834, + "step": 41791 + }, + { + "epoch": 2.69, + "learning_rate": 1.105347903561691e-05, + "loss": 0.9775, + "step": 41792 + }, + { + "epoch": 2.69, + "learning_rate": 1.1053264346593959e-05, + "loss": 0.887, + "step": 41793 + }, + { + "epoch": 2.69, + "learning_rate": 1.105304965757101e-05, + "loss": 0.9238, + "step": 41794 + }, + { + "epoch": 2.69, + "learning_rate": 1.1052834968548058e-05, + "loss": 0.8924, + "step": 41795 + }, + { + "epoch": 2.69, + "learning_rate": 1.1052620279525107e-05, + "loss": 0.9089, + "step": 41796 + }, + { + "epoch": 2.69, + "learning_rate": 1.1052405590502158e-05, + "loss": 0.9248, + "step": 41797 + }, + { + "epoch": 2.69, + "learning_rate": 1.1052190901479207e-05, + "loss": 0.8703, + "step": 41798 + }, + { + "epoch": 2.69, + "learning_rate": 1.105197621245626e-05, + "loss": 0.8306, + "step": 41799 + }, + { + "epoch": 2.69, + "learning_rate": 1.1051761523433308e-05, + "loss": 0.8891, + "step": 41800 + }, + { + "epoch": 2.69, + "learning_rate": 1.1051546834410359e-05, + "loss": 0.9271, + "step": 41801 + }, + { + "epoch": 2.69, + "learning_rate": 1.1051332145387408e-05, + "loss": 0.9214, + "step": 41802 + }, + { + "epoch": 2.69, + "learning_rate": 1.1051117456364457e-05, + "loss": 0.8888, + "step": 41803 + }, + { + "epoch": 2.69, + "learning_rate": 1.1050902767341508e-05, + "loss": 0.8819, + "step": 41804 + }, + { + "epoch": 2.69, + "learning_rate": 1.1050688078318556e-05, + "loss": 0.884, + "step": 41805 + }, + { + "epoch": 2.69, + "learning_rate": 1.1050473389295605e-05, + "loss": 0.8877, + "step": 41806 + }, + { + "epoch": 2.69, + "learning_rate": 1.1050258700272656e-05, + "loss": 0.9349, + "step": 41807 + }, + { + "epoch": 2.69, + "learning_rate": 1.1050044011249705e-05, + "loss": 0.9201, + "step": 41808 + }, + { + "epoch": 2.69, + "learning_rate": 1.1049829322226756e-05, + "loss": 0.8352, + "step": 41809 + }, + { + "epoch": 2.69, + "learning_rate": 1.1049614633203805e-05, + "loss": 0.9256, + "step": 41810 + }, + { + "epoch": 2.69, + "learning_rate": 1.1049399944180854e-05, + "loss": 0.9058, + "step": 41811 + }, + { + "epoch": 2.69, + "learning_rate": 1.1049185255157904e-05, + "loss": 0.8471, + "step": 41812 + }, + { + "epoch": 2.69, + "learning_rate": 1.1048970566134953e-05, + "loss": 0.8582, + "step": 41813 + }, + { + "epoch": 2.69, + "learning_rate": 1.1048755877112006e-05, + "loss": 0.9716, + "step": 41814 + }, + { + "epoch": 2.69, + "learning_rate": 1.1048541188089055e-05, + "loss": 0.8925, + "step": 41815 + }, + { + "epoch": 2.69, + "learning_rate": 1.1048326499066104e-05, + "loss": 0.9169, + "step": 41816 + }, + { + "epoch": 2.69, + "learning_rate": 1.1048111810043154e-05, + "loss": 0.8821, + "step": 41817 + }, + { + "epoch": 2.69, + "learning_rate": 1.1047897121020203e-05, + "loss": 0.8682, + "step": 41818 + }, + { + "epoch": 2.69, + "learning_rate": 1.1047682431997254e-05, + "loss": 0.8632, + "step": 41819 + }, + { + "epoch": 2.69, + "learning_rate": 1.1047467742974303e-05, + "loss": 0.8756, + "step": 41820 + }, + { + "epoch": 2.69, + "learning_rate": 1.1047253053951352e-05, + "loss": 0.9025, + "step": 41821 + }, + { + "epoch": 2.69, + "learning_rate": 1.1047038364928402e-05, + "loss": 0.8446, + "step": 41822 + }, + { + "epoch": 2.69, + "learning_rate": 1.1046823675905451e-05, + "loss": 0.8674, + "step": 41823 + }, + { + "epoch": 2.69, + "learning_rate": 1.10466089868825e-05, + "loss": 0.8683, + "step": 41824 + }, + { + "epoch": 2.69, + "learning_rate": 1.1046394297859551e-05, + "loss": 1.0466, + "step": 41825 + }, + { + "epoch": 2.69, + "learning_rate": 1.10461796088366e-05, + "loss": 0.9068, + "step": 41826 + }, + { + "epoch": 2.69, + "learning_rate": 1.104596491981365e-05, + "loss": 0.8961, + "step": 41827 + }, + { + "epoch": 2.69, + "learning_rate": 1.1045750230790701e-05, + "loss": 0.9013, + "step": 41828 + }, + { + "epoch": 2.69, + "learning_rate": 1.1045535541767752e-05, + "loss": 0.9025, + "step": 41829 + }, + { + "epoch": 2.69, + "learning_rate": 1.1045320852744801e-05, + "loss": 0.8419, + "step": 41830 + }, + { + "epoch": 2.69, + "learning_rate": 1.104510616372185e-05, + "loss": 0.9265, + "step": 41831 + }, + { + "epoch": 2.69, + "learning_rate": 1.10448914746989e-05, + "loss": 0.9267, + "step": 41832 + }, + { + "epoch": 2.69, + "learning_rate": 1.104467678567595e-05, + "loss": 0.8657, + "step": 41833 + }, + { + "epoch": 2.69, + "learning_rate": 1.1044462096652998e-05, + "loss": 0.9581, + "step": 41834 + }, + { + "epoch": 2.69, + "learning_rate": 1.1044247407630049e-05, + "loss": 0.9255, + "step": 41835 + }, + { + "epoch": 2.69, + "learning_rate": 1.1044032718607098e-05, + "loss": 0.952, + "step": 41836 + }, + { + "epoch": 2.69, + "learning_rate": 1.1043818029584149e-05, + "loss": 0.8758, + "step": 41837 + }, + { + "epoch": 2.69, + "learning_rate": 1.1043603340561198e-05, + "loss": 0.8418, + "step": 41838 + }, + { + "epoch": 2.69, + "learning_rate": 1.1043388651538247e-05, + "loss": 0.8783, + "step": 41839 + }, + { + "epoch": 2.69, + "learning_rate": 1.1043173962515297e-05, + "loss": 0.9129, + "step": 41840 + }, + { + "epoch": 2.69, + "learning_rate": 1.1042959273492346e-05, + "loss": 0.9254, + "step": 41841 + }, + { + "epoch": 2.69, + "learning_rate": 1.1042744584469395e-05, + "loss": 0.9064, + "step": 41842 + }, + { + "epoch": 2.69, + "learning_rate": 1.1042529895446448e-05, + "loss": 0.9243, + "step": 41843 + }, + { + "epoch": 2.69, + "learning_rate": 1.1042315206423497e-05, + "loss": 0.9111, + "step": 41844 + }, + { + "epoch": 2.69, + "learning_rate": 1.1042100517400547e-05, + "loss": 0.9286, + "step": 41845 + }, + { + "epoch": 2.69, + "learning_rate": 1.1041885828377596e-05, + "loss": 0.8943, + "step": 41846 + }, + { + "epoch": 2.69, + "learning_rate": 1.1041671139354647e-05, + "loss": 0.8682, + "step": 41847 + }, + { + "epoch": 2.69, + "learning_rate": 1.1041456450331696e-05, + "loss": 0.8457, + "step": 41848 + }, + { + "epoch": 2.69, + "learning_rate": 1.1041241761308745e-05, + "loss": 0.9487, + "step": 41849 + }, + { + "epoch": 2.69, + "learning_rate": 1.1041027072285795e-05, + "loss": 0.9156, + "step": 41850 + }, + { + "epoch": 2.69, + "learning_rate": 1.1040812383262844e-05, + "loss": 0.9442, + "step": 41851 + }, + { + "epoch": 2.69, + "learning_rate": 1.1040597694239893e-05, + "loss": 0.9184, + "step": 41852 + }, + { + "epoch": 2.69, + "learning_rate": 1.1040383005216944e-05, + "loss": 0.8751, + "step": 41853 + }, + { + "epoch": 2.69, + "learning_rate": 1.1040168316193993e-05, + "loss": 0.8874, + "step": 41854 + }, + { + "epoch": 2.69, + "learning_rate": 1.1039953627171042e-05, + "loss": 0.8944, + "step": 41855 + }, + { + "epoch": 2.69, + "learning_rate": 1.1039738938148093e-05, + "loss": 0.8632, + "step": 41856 + }, + { + "epoch": 2.69, + "learning_rate": 1.1039524249125145e-05, + "loss": 0.8713, + "step": 41857 + }, + { + "epoch": 2.69, + "learning_rate": 1.1039309560102194e-05, + "loss": 0.9463, + "step": 41858 + }, + { + "epoch": 2.69, + "learning_rate": 1.1039094871079243e-05, + "loss": 0.8995, + "step": 41859 + }, + { + "epoch": 2.69, + "learning_rate": 1.1038880182056294e-05, + "loss": 0.8701, + "step": 41860 + }, + { + "epoch": 2.69, + "learning_rate": 1.1038665493033343e-05, + "loss": 0.9092, + "step": 41861 + }, + { + "epoch": 2.69, + "learning_rate": 1.1038450804010392e-05, + "loss": 0.91, + "step": 41862 + }, + { + "epoch": 2.69, + "learning_rate": 1.1038236114987442e-05, + "loss": 0.9387, + "step": 41863 + }, + { + "epoch": 2.69, + "learning_rate": 1.1038021425964491e-05, + "loss": 0.9049, + "step": 41864 + }, + { + "epoch": 2.69, + "learning_rate": 1.103780673694154e-05, + "loss": 0.8627, + "step": 41865 + }, + { + "epoch": 2.69, + "learning_rate": 1.103759204791859e-05, + "loss": 0.8892, + "step": 41866 + }, + { + "epoch": 2.69, + "learning_rate": 1.103737735889564e-05, + "loss": 0.9524, + "step": 41867 + }, + { + "epoch": 2.69, + "learning_rate": 1.103716266987269e-05, + "loss": 0.873, + "step": 41868 + }, + { + "epoch": 2.69, + "learning_rate": 1.103694798084974e-05, + "loss": 0.9439, + "step": 41869 + }, + { + "epoch": 2.69, + "learning_rate": 1.1036733291826788e-05, + "loss": 0.9067, + "step": 41870 + }, + { + "epoch": 2.69, + "learning_rate": 1.1036518602803839e-05, + "loss": 0.8265, + "step": 41871 + }, + { + "epoch": 2.69, + "learning_rate": 1.103630391378089e-05, + "loss": 0.8521, + "step": 41872 + }, + { + "epoch": 2.69, + "learning_rate": 1.103608922475794e-05, + "loss": 0.8988, + "step": 41873 + }, + { + "epoch": 2.69, + "learning_rate": 1.103587453573499e-05, + "loss": 0.8358, + "step": 41874 + }, + { + "epoch": 2.69, + "learning_rate": 1.1035659846712038e-05, + "loss": 0.8832, + "step": 41875 + }, + { + "epoch": 2.69, + "learning_rate": 1.1035445157689089e-05, + "loss": 0.9067, + "step": 41876 + }, + { + "epoch": 2.69, + "learning_rate": 1.1035230468666138e-05, + "loss": 0.8611, + "step": 41877 + }, + { + "epoch": 2.69, + "learning_rate": 1.1035015779643189e-05, + "loss": 0.8705, + "step": 41878 + }, + { + "epoch": 2.69, + "learning_rate": 1.1034801090620237e-05, + "loss": 0.8496, + "step": 41879 + }, + { + "epoch": 2.69, + "learning_rate": 1.1034586401597286e-05, + "loss": 0.91, + "step": 41880 + }, + { + "epoch": 2.69, + "learning_rate": 1.1034371712574337e-05, + "loss": 0.8782, + "step": 41881 + }, + { + "epoch": 2.69, + "learning_rate": 1.1034157023551386e-05, + "loss": 0.8196, + "step": 41882 + }, + { + "epoch": 2.69, + "learning_rate": 1.1033942334528435e-05, + "loss": 0.8901, + "step": 41883 + }, + { + "epoch": 2.69, + "learning_rate": 1.1033727645505486e-05, + "loss": 0.8738, + "step": 41884 + }, + { + "epoch": 2.69, + "learning_rate": 1.1033512956482535e-05, + "loss": 0.9278, + "step": 41885 + }, + { + "epoch": 2.69, + "learning_rate": 1.1033298267459587e-05, + "loss": 0.8982, + "step": 41886 + }, + { + "epoch": 2.69, + "learning_rate": 1.1033083578436636e-05, + "loss": 0.928, + "step": 41887 + }, + { + "epoch": 2.69, + "learning_rate": 1.1032868889413687e-05, + "loss": 0.9594, + "step": 41888 + }, + { + "epoch": 2.7, + "learning_rate": 1.1032654200390736e-05, + "loss": 0.9058, + "step": 41889 + }, + { + "epoch": 2.7, + "learning_rate": 1.1032439511367785e-05, + "loss": 0.8981, + "step": 41890 + }, + { + "epoch": 2.7, + "learning_rate": 1.1032224822344835e-05, + "loss": 0.8951, + "step": 41891 + }, + { + "epoch": 2.7, + "learning_rate": 1.1032010133321884e-05, + "loss": 0.9529, + "step": 41892 + }, + { + "epoch": 2.7, + "learning_rate": 1.1031795444298933e-05, + "loss": 0.9574, + "step": 41893 + }, + { + "epoch": 2.7, + "learning_rate": 1.1031580755275984e-05, + "loss": 0.8773, + "step": 41894 + }, + { + "epoch": 2.7, + "learning_rate": 1.1031366066253033e-05, + "loss": 1.0043, + "step": 41895 + }, + { + "epoch": 2.7, + "learning_rate": 1.1031151377230083e-05, + "loss": 0.9089, + "step": 41896 + }, + { + "epoch": 2.7, + "learning_rate": 1.1030936688207132e-05, + "loss": 0.928, + "step": 41897 + }, + { + "epoch": 2.7, + "learning_rate": 1.1030721999184181e-05, + "loss": 0.931, + "step": 41898 + }, + { + "epoch": 2.7, + "learning_rate": 1.1030507310161232e-05, + "loss": 0.9366, + "step": 41899 + }, + { + "epoch": 2.7, + "learning_rate": 1.1030292621138281e-05, + "loss": 0.905, + "step": 41900 + }, + { + "epoch": 2.7, + "learning_rate": 1.1030077932115333e-05, + "loss": 0.8126, + "step": 41901 + }, + { + "epoch": 2.7, + "learning_rate": 1.1029863243092382e-05, + "loss": 0.8553, + "step": 41902 + }, + { + "epoch": 2.7, + "learning_rate": 1.1029648554069431e-05, + "loss": 0.8827, + "step": 41903 + }, + { + "epoch": 2.7, + "learning_rate": 1.1029433865046482e-05, + "loss": 0.885, + "step": 41904 + }, + { + "epoch": 2.7, + "learning_rate": 1.1029219176023531e-05, + "loss": 0.9107, + "step": 41905 + }, + { + "epoch": 2.7, + "learning_rate": 1.1029004487000582e-05, + "loss": 0.8687, + "step": 41906 + }, + { + "epoch": 2.7, + "learning_rate": 1.102878979797763e-05, + "loss": 0.8296, + "step": 41907 + }, + { + "epoch": 2.7, + "learning_rate": 1.102857510895468e-05, + "loss": 0.8782, + "step": 41908 + }, + { + "epoch": 2.7, + "learning_rate": 1.102836041993173e-05, + "loss": 0.9168, + "step": 41909 + }, + { + "epoch": 2.7, + "learning_rate": 1.1028145730908779e-05, + "loss": 0.8559, + "step": 41910 + }, + { + "epoch": 2.7, + "learning_rate": 1.1027931041885828e-05, + "loss": 0.8983, + "step": 41911 + }, + { + "epoch": 2.7, + "learning_rate": 1.1027716352862879e-05, + "loss": 0.8602, + "step": 41912 + }, + { + "epoch": 2.7, + "learning_rate": 1.1027501663839928e-05, + "loss": 0.9197, + "step": 41913 + }, + { + "epoch": 2.7, + "learning_rate": 1.1027286974816978e-05, + "loss": 0.9067, + "step": 41914 + }, + { + "epoch": 2.7, + "learning_rate": 1.1027072285794029e-05, + "loss": 0.8663, + "step": 41915 + }, + { + "epoch": 2.7, + "learning_rate": 1.102685759677108e-05, + "loss": 0.8587, + "step": 41916 + }, + { + "epoch": 2.7, + "learning_rate": 1.1026642907748129e-05, + "loss": 0.8823, + "step": 41917 + }, + { + "epoch": 2.7, + "learning_rate": 1.1026428218725178e-05, + "loss": 0.8522, + "step": 41918 + }, + { + "epoch": 2.7, + "learning_rate": 1.1026213529702228e-05, + "loss": 0.9411, + "step": 41919 + }, + { + "epoch": 2.7, + "learning_rate": 1.1025998840679277e-05, + "loss": 0.9543, + "step": 41920 + }, + { + "epoch": 2.7, + "learning_rate": 1.1025784151656326e-05, + "loss": 0.9847, + "step": 41921 + }, + { + "epoch": 2.7, + "learning_rate": 1.1025569462633377e-05, + "loss": 0.873, + "step": 41922 + }, + { + "epoch": 2.7, + "learning_rate": 1.1025354773610426e-05, + "loss": 0.9379, + "step": 41923 + }, + { + "epoch": 2.7, + "learning_rate": 1.1025140084587476e-05, + "loss": 0.8596, + "step": 41924 + }, + { + "epoch": 2.7, + "learning_rate": 1.1024925395564525e-05, + "loss": 0.9731, + "step": 41925 + }, + { + "epoch": 2.7, + "learning_rate": 1.1024710706541574e-05, + "loss": 0.9252, + "step": 41926 + }, + { + "epoch": 2.7, + "learning_rate": 1.1024496017518625e-05, + "loss": 0.8999, + "step": 41927 + }, + { + "epoch": 2.7, + "learning_rate": 1.1024281328495674e-05, + "loss": 0.9427, + "step": 41928 + }, + { + "epoch": 2.7, + "learning_rate": 1.1024066639472723e-05, + "loss": 0.9007, + "step": 41929 + }, + { + "epoch": 2.7, + "learning_rate": 1.1023851950449775e-05, + "loss": 0.932, + "step": 41930 + }, + { + "epoch": 2.7, + "learning_rate": 1.1023637261426824e-05, + "loss": 0.8328, + "step": 41931 + }, + { + "epoch": 2.7, + "learning_rate": 1.1023422572403875e-05, + "loss": 0.8823, + "step": 41932 + }, + { + "epoch": 2.7, + "learning_rate": 1.1023207883380924e-05, + "loss": 0.9647, + "step": 41933 + }, + { + "epoch": 2.7, + "learning_rate": 1.1022993194357975e-05, + "loss": 0.9538, + "step": 41934 + }, + { + "epoch": 2.7, + "learning_rate": 1.1022778505335024e-05, + "loss": 0.8724, + "step": 41935 + }, + { + "epoch": 2.7, + "learning_rate": 1.1022563816312073e-05, + "loss": 0.8972, + "step": 41936 + }, + { + "epoch": 2.7, + "learning_rate": 1.1022349127289123e-05, + "loss": 0.9164, + "step": 41937 + }, + { + "epoch": 2.7, + "learning_rate": 1.1022134438266172e-05, + "loss": 0.8648, + "step": 41938 + }, + { + "epoch": 2.7, + "learning_rate": 1.1021919749243221e-05, + "loss": 0.9019, + "step": 41939 + }, + { + "epoch": 2.7, + "learning_rate": 1.1021705060220272e-05, + "loss": 0.8651, + "step": 41940 + }, + { + "epoch": 2.7, + "learning_rate": 1.102149037119732e-05, + "loss": 0.8886, + "step": 41941 + }, + { + "epoch": 2.7, + "learning_rate": 1.102127568217437e-05, + "loss": 0.8672, + "step": 41942 + }, + { + "epoch": 2.7, + "learning_rate": 1.102106099315142e-05, + "loss": 0.8623, + "step": 41943 + }, + { + "epoch": 2.7, + "learning_rate": 1.1020846304128473e-05, + "loss": 0.9313, + "step": 41944 + }, + { + "epoch": 2.7, + "learning_rate": 1.1020631615105522e-05, + "loss": 0.9131, + "step": 41945 + }, + { + "epoch": 2.7, + "learning_rate": 1.102041692608257e-05, + "loss": 0.938, + "step": 41946 + }, + { + "epoch": 2.7, + "learning_rate": 1.1020202237059621e-05, + "loss": 0.8423, + "step": 41947 + }, + { + "epoch": 2.7, + "learning_rate": 1.101998754803667e-05, + "loss": 0.9101, + "step": 41948 + }, + { + "epoch": 2.7, + "learning_rate": 1.101977285901372e-05, + "loss": 0.8776, + "step": 41949 + }, + { + "epoch": 2.7, + "learning_rate": 1.101955816999077e-05, + "loss": 0.9184, + "step": 41950 + }, + { + "epoch": 2.7, + "learning_rate": 1.1019343480967819e-05, + "loss": 0.8765, + "step": 41951 + }, + { + "epoch": 2.7, + "learning_rate": 1.1019128791944868e-05, + "loss": 0.9312, + "step": 41952 + }, + { + "epoch": 2.7, + "learning_rate": 1.1018914102921918e-05, + "loss": 0.8805, + "step": 41953 + }, + { + "epoch": 2.7, + "learning_rate": 1.1018699413898967e-05, + "loss": 0.9012, + "step": 41954 + }, + { + "epoch": 2.7, + "learning_rate": 1.1018484724876018e-05, + "loss": 0.9041, + "step": 41955 + }, + { + "epoch": 2.7, + "learning_rate": 1.1018270035853067e-05, + "loss": 0.8829, + "step": 41956 + }, + { + "epoch": 2.7, + "learning_rate": 1.1018055346830116e-05, + "loss": 0.8762, + "step": 41957 + }, + { + "epoch": 2.7, + "learning_rate": 1.1017840657807167e-05, + "loss": 0.8565, + "step": 41958 + }, + { + "epoch": 2.7, + "learning_rate": 1.1017625968784217e-05, + "loss": 0.894, + "step": 41959 + }, + { + "epoch": 2.7, + "learning_rate": 1.1017411279761268e-05, + "loss": 0.8878, + "step": 41960 + }, + { + "epoch": 2.7, + "learning_rate": 1.1017196590738317e-05, + "loss": 0.8993, + "step": 41961 + }, + { + "epoch": 2.7, + "learning_rate": 1.1016981901715366e-05, + "loss": 0.9076, + "step": 41962 + }, + { + "epoch": 2.7, + "learning_rate": 1.1016767212692417e-05, + "loss": 0.9298, + "step": 41963 + }, + { + "epoch": 2.7, + "learning_rate": 1.1016552523669466e-05, + "loss": 0.8967, + "step": 41964 + }, + { + "epoch": 2.7, + "learning_rate": 1.1016337834646516e-05, + "loss": 0.848, + "step": 41965 + }, + { + "epoch": 2.7, + "learning_rate": 1.1016123145623565e-05, + "loss": 0.9908, + "step": 41966 + }, + { + "epoch": 2.7, + "learning_rate": 1.1015908456600614e-05, + "loss": 0.9527, + "step": 41967 + }, + { + "epoch": 2.7, + "learning_rate": 1.1015693767577665e-05, + "loss": 0.8627, + "step": 41968 + }, + { + "epoch": 2.7, + "learning_rate": 1.1015479078554714e-05, + "loss": 0.9531, + "step": 41969 + }, + { + "epoch": 2.7, + "learning_rate": 1.1015264389531763e-05, + "loss": 0.8501, + "step": 41970 + }, + { + "epoch": 2.7, + "learning_rate": 1.1015049700508813e-05, + "loss": 0.9446, + "step": 41971 + }, + { + "epoch": 2.7, + "learning_rate": 1.1014835011485862e-05, + "loss": 0.9344, + "step": 41972 + }, + { + "epoch": 2.7, + "learning_rate": 1.1014620322462915e-05, + "loss": 0.98, + "step": 41973 + }, + { + "epoch": 2.7, + "learning_rate": 1.1014405633439964e-05, + "loss": 0.9059, + "step": 41974 + }, + { + "epoch": 2.7, + "learning_rate": 1.1014190944417014e-05, + "loss": 0.8871, + "step": 41975 + }, + { + "epoch": 2.7, + "learning_rate": 1.1013976255394063e-05, + "loss": 0.8945, + "step": 41976 + }, + { + "epoch": 2.7, + "learning_rate": 1.1013761566371112e-05, + "loss": 0.9059, + "step": 41977 + }, + { + "epoch": 2.7, + "learning_rate": 1.1013546877348163e-05, + "loss": 0.9784, + "step": 41978 + }, + { + "epoch": 2.7, + "learning_rate": 1.1013332188325212e-05, + "loss": 0.8597, + "step": 41979 + }, + { + "epoch": 2.7, + "learning_rate": 1.1013117499302261e-05, + "loss": 0.9054, + "step": 41980 + }, + { + "epoch": 2.7, + "learning_rate": 1.1012902810279312e-05, + "loss": 0.8856, + "step": 41981 + }, + { + "epoch": 2.7, + "learning_rate": 1.101268812125636e-05, + "loss": 0.9209, + "step": 41982 + }, + { + "epoch": 2.7, + "learning_rate": 1.1012473432233411e-05, + "loss": 0.9502, + "step": 41983 + }, + { + "epoch": 2.7, + "learning_rate": 1.101225874321046e-05, + "loss": 0.831, + "step": 41984 + }, + { + "epoch": 2.7, + "learning_rate": 1.1012044054187509e-05, + "loss": 0.9766, + "step": 41985 + }, + { + "epoch": 2.7, + "learning_rate": 1.101182936516456e-05, + "loss": 0.9075, + "step": 41986 + }, + { + "epoch": 2.7, + "learning_rate": 1.1011614676141609e-05, + "loss": 0.9011, + "step": 41987 + }, + { + "epoch": 2.7, + "learning_rate": 1.1011399987118661e-05, + "loss": 0.9462, + "step": 41988 + }, + { + "epoch": 2.7, + "learning_rate": 1.101118529809571e-05, + "loss": 0.9146, + "step": 41989 + }, + { + "epoch": 2.7, + "learning_rate": 1.1010970609072759e-05, + "loss": 0.9557, + "step": 41990 + }, + { + "epoch": 2.7, + "learning_rate": 1.101075592004981e-05, + "loss": 0.9685, + "step": 41991 + }, + { + "epoch": 2.7, + "learning_rate": 1.1010541231026859e-05, + "loss": 0.8758, + "step": 41992 + }, + { + "epoch": 2.7, + "learning_rate": 1.101032654200391e-05, + "loss": 0.7755, + "step": 41993 + }, + { + "epoch": 2.7, + "learning_rate": 1.1010111852980958e-05, + "loss": 0.8588, + "step": 41994 + }, + { + "epoch": 2.7, + "learning_rate": 1.1009897163958007e-05, + "loss": 0.8738, + "step": 41995 + }, + { + "epoch": 2.7, + "learning_rate": 1.1009682474935058e-05, + "loss": 0.8964, + "step": 41996 + }, + { + "epoch": 2.7, + "learning_rate": 1.1009467785912107e-05, + "loss": 0.8957, + "step": 41997 + }, + { + "epoch": 2.7, + "learning_rate": 1.1009253096889156e-05, + "loss": 0.8852, + "step": 41998 + }, + { + "epoch": 2.7, + "learning_rate": 1.1009038407866206e-05, + "loss": 0.9027, + "step": 41999 + }, + { + "epoch": 2.7, + "learning_rate": 1.1008823718843255e-05, + "loss": 0.8134, + "step": 42000 + }, + { + "epoch": 2.7, + "eval_laion/cider": 0.5806063759748071, + "eval_laion/meteor": 0.3916025760140512, + "eval_laion/pred_num_words": 37.396484375, + "eval_laion/sacrebleu": 16.072485808384997, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.038920283317566, + "eval_runtime": 103.0742, + "eval_samples_per_second": 4.841, + "eval_steps_per_second": 0.31, + "step": 42000 + }, + { + "epoch": 2.7, + "learning_rate": 1.1008609029820304e-05, + "loss": 0.8657, + "step": 42001 + }, + { + "epoch": 2.7, + "learning_rate": 1.1008394340797357e-05, + "loss": 0.8748, + "step": 42002 + }, + { + "epoch": 2.7, + "learning_rate": 1.1008179651774407e-05, + "loss": 0.9269, + "step": 42003 + }, + { + "epoch": 2.7, + "learning_rate": 1.1007964962751456e-05, + "loss": 0.8208, + "step": 42004 + }, + { + "epoch": 2.7, + "learning_rate": 1.1007750273728505e-05, + "loss": 0.8662, + "step": 42005 + }, + { + "epoch": 2.7, + "learning_rate": 1.1007535584705556e-05, + "loss": 0.9235, + "step": 42006 + }, + { + "epoch": 2.7, + "learning_rate": 1.1007320895682605e-05, + "loss": 0.9159, + "step": 42007 + }, + { + "epoch": 2.7, + "learning_rate": 1.1007106206659654e-05, + "loss": 0.9636, + "step": 42008 + }, + { + "epoch": 2.7, + "learning_rate": 1.1006891517636705e-05, + "loss": 0.9326, + "step": 42009 + }, + { + "epoch": 2.7, + "learning_rate": 1.1006676828613754e-05, + "loss": 0.8458, + "step": 42010 + }, + { + "epoch": 2.7, + "learning_rate": 1.1006462139590804e-05, + "loss": 0.8947, + "step": 42011 + }, + { + "epoch": 2.7, + "learning_rate": 1.1006247450567853e-05, + "loss": 0.9075, + "step": 42012 + }, + { + "epoch": 2.7, + "learning_rate": 1.1006032761544902e-05, + "loss": 0.8639, + "step": 42013 + }, + { + "epoch": 2.7, + "learning_rate": 1.1005818072521953e-05, + "loss": 0.8319, + "step": 42014 + }, + { + "epoch": 2.7, + "learning_rate": 1.1005603383499002e-05, + "loss": 0.8882, + "step": 42015 + }, + { + "epoch": 2.7, + "learning_rate": 1.100538869447605e-05, + "loss": 0.9251, + "step": 42016 + }, + { + "epoch": 2.7, + "learning_rate": 1.1005174005453103e-05, + "loss": 0.841, + "step": 42017 + }, + { + "epoch": 2.7, + "learning_rate": 1.1004959316430152e-05, + "loss": 0.8576, + "step": 42018 + }, + { + "epoch": 2.7, + "learning_rate": 1.1004744627407203e-05, + "loss": 0.9429, + "step": 42019 + }, + { + "epoch": 2.7, + "learning_rate": 1.1004529938384252e-05, + "loss": 0.8822, + "step": 42020 + }, + { + "epoch": 2.7, + "learning_rate": 1.1004315249361302e-05, + "loss": 0.8829, + "step": 42021 + }, + { + "epoch": 2.7, + "learning_rate": 1.1004100560338351e-05, + "loss": 0.8705, + "step": 42022 + }, + { + "epoch": 2.7, + "learning_rate": 1.10038858713154e-05, + "loss": 0.9098, + "step": 42023 + }, + { + "epoch": 2.7, + "learning_rate": 1.1003671182292451e-05, + "loss": 0.8847, + "step": 42024 + }, + { + "epoch": 2.7, + "learning_rate": 1.10034564932695e-05, + "loss": 0.8843, + "step": 42025 + }, + { + "epoch": 2.7, + "learning_rate": 1.1003241804246549e-05, + "loss": 0.8893, + "step": 42026 + }, + { + "epoch": 2.7, + "learning_rate": 1.10030271152236e-05, + "loss": 0.861, + "step": 42027 + }, + { + "epoch": 2.7, + "learning_rate": 1.1002812426200648e-05, + "loss": 0.9232, + "step": 42028 + }, + { + "epoch": 2.7, + "learning_rate": 1.1002597737177697e-05, + "loss": 0.923, + "step": 42029 + }, + { + "epoch": 2.7, + "learning_rate": 1.1002383048154748e-05, + "loss": 0.8355, + "step": 42030 + }, + { + "epoch": 2.7, + "learning_rate": 1.10021683591318e-05, + "loss": 0.8882, + "step": 42031 + }, + { + "epoch": 2.7, + "learning_rate": 1.100195367010885e-05, + "loss": 0.8505, + "step": 42032 + }, + { + "epoch": 2.7, + "learning_rate": 1.1001738981085898e-05, + "loss": 0.9204, + "step": 42033 + }, + { + "epoch": 2.7, + "learning_rate": 1.1001524292062949e-05, + "loss": 0.8374, + "step": 42034 + }, + { + "epoch": 2.7, + "learning_rate": 1.1001309603039998e-05, + "loss": 0.9291, + "step": 42035 + }, + { + "epoch": 2.7, + "learning_rate": 1.1001094914017047e-05, + "loss": 0.9073, + "step": 42036 + }, + { + "epoch": 2.7, + "learning_rate": 1.1000880224994098e-05, + "loss": 0.9665, + "step": 42037 + }, + { + "epoch": 2.7, + "learning_rate": 1.1000665535971147e-05, + "loss": 0.8661, + "step": 42038 + }, + { + "epoch": 2.7, + "learning_rate": 1.1000450846948196e-05, + "loss": 0.906, + "step": 42039 + }, + { + "epoch": 2.7, + "learning_rate": 1.1000236157925246e-05, + "loss": 0.8665, + "step": 42040 + }, + { + "epoch": 2.7, + "learning_rate": 1.1000021468902295e-05, + "loss": 0.9141, + "step": 42041 + }, + { + "epoch": 2.7, + "learning_rate": 1.0999806779879346e-05, + "loss": 0.9361, + "step": 42042 + }, + { + "epoch": 2.7, + "learning_rate": 1.0999592090856395e-05, + "loss": 0.9228, + "step": 42043 + }, + { + "epoch": 2.71, + "learning_rate": 1.0999377401833444e-05, + "loss": 0.8908, + "step": 42044 + }, + { + "epoch": 2.71, + "learning_rate": 1.0999162712810494e-05, + "loss": 0.8835, + "step": 42045 + }, + { + "epoch": 2.71, + "learning_rate": 1.0998948023787545e-05, + "loss": 0.8514, + "step": 42046 + }, + { + "epoch": 2.71, + "learning_rate": 1.0998733334764596e-05, + "loss": 0.904, + "step": 42047 + }, + { + "epoch": 2.71, + "learning_rate": 1.0998518645741645e-05, + "loss": 0.8461, + "step": 42048 + }, + { + "epoch": 2.71, + "learning_rate": 1.0998303956718694e-05, + "loss": 0.9166, + "step": 42049 + }, + { + "epoch": 2.71, + "learning_rate": 1.0998089267695744e-05, + "loss": 0.9723, + "step": 42050 + }, + { + "epoch": 2.71, + "learning_rate": 1.0997874578672793e-05, + "loss": 0.8401, + "step": 42051 + }, + { + "epoch": 2.71, + "learning_rate": 1.0997659889649844e-05, + "loss": 0.9232, + "step": 42052 + }, + { + "epoch": 2.71, + "learning_rate": 1.0997445200626893e-05, + "loss": 0.9431, + "step": 42053 + }, + { + "epoch": 2.71, + "learning_rate": 1.0997230511603942e-05, + "loss": 0.9065, + "step": 42054 + }, + { + "epoch": 2.71, + "learning_rate": 1.0997015822580993e-05, + "loss": 0.7939, + "step": 42055 + }, + { + "epoch": 2.71, + "learning_rate": 1.0996801133558041e-05, + "loss": 0.8974, + "step": 42056 + }, + { + "epoch": 2.71, + "learning_rate": 1.099658644453509e-05, + "loss": 0.8346, + "step": 42057 + }, + { + "epoch": 2.71, + "learning_rate": 1.0996371755512141e-05, + "loss": 0.8995, + "step": 42058 + }, + { + "epoch": 2.71, + "learning_rate": 1.099615706648919e-05, + "loss": 0.9032, + "step": 42059 + }, + { + "epoch": 2.71, + "learning_rate": 1.0995942377466242e-05, + "loss": 0.8726, + "step": 42060 + }, + { + "epoch": 2.71, + "learning_rate": 1.0995727688443291e-05, + "loss": 0.9531, + "step": 42061 + }, + { + "epoch": 2.71, + "learning_rate": 1.0995512999420342e-05, + "loss": 0.8068, + "step": 42062 + }, + { + "epoch": 2.71, + "learning_rate": 1.0995298310397391e-05, + "loss": 0.9021, + "step": 42063 + }, + { + "epoch": 2.71, + "learning_rate": 1.099508362137444e-05, + "loss": 0.8918, + "step": 42064 + }, + { + "epoch": 2.71, + "learning_rate": 1.099486893235149e-05, + "loss": 0.9166, + "step": 42065 + }, + { + "epoch": 2.71, + "learning_rate": 1.099465424332854e-05, + "loss": 0.8961, + "step": 42066 + }, + { + "epoch": 2.71, + "learning_rate": 1.0994439554305589e-05, + "loss": 0.8718, + "step": 42067 + }, + { + "epoch": 2.71, + "learning_rate": 1.099422486528264e-05, + "loss": 0.8851, + "step": 42068 + }, + { + "epoch": 2.71, + "learning_rate": 1.0994010176259688e-05, + "loss": 0.88, + "step": 42069 + }, + { + "epoch": 2.71, + "learning_rate": 1.0993795487236739e-05, + "loss": 0.8691, + "step": 42070 + }, + { + "epoch": 2.71, + "learning_rate": 1.0993580798213788e-05, + "loss": 0.8626, + "step": 42071 + }, + { + "epoch": 2.71, + "learning_rate": 1.0993366109190837e-05, + "loss": 0.8541, + "step": 42072 + }, + { + "epoch": 2.71, + "learning_rate": 1.0993151420167887e-05, + "loss": 0.9466, + "step": 42073 + }, + { + "epoch": 2.71, + "learning_rate": 1.0992936731144936e-05, + "loss": 0.9214, + "step": 42074 + }, + { + "epoch": 2.71, + "learning_rate": 1.0992722042121989e-05, + "loss": 0.8718, + "step": 42075 + }, + { + "epoch": 2.71, + "learning_rate": 1.0992507353099038e-05, + "loss": 0.9895, + "step": 42076 + }, + { + "epoch": 2.71, + "learning_rate": 1.0992292664076087e-05, + "loss": 0.8656, + "step": 42077 + }, + { + "epoch": 2.71, + "learning_rate": 1.0992077975053137e-05, + "loss": 0.896, + "step": 42078 + }, + { + "epoch": 2.71, + "learning_rate": 1.0991863286030186e-05, + "loss": 0.9071, + "step": 42079 + }, + { + "epoch": 2.71, + "learning_rate": 1.0991648597007237e-05, + "loss": 0.9044, + "step": 42080 + }, + { + "epoch": 2.71, + "learning_rate": 1.0991433907984286e-05, + "loss": 0.89, + "step": 42081 + }, + { + "epoch": 2.71, + "learning_rate": 1.0991219218961335e-05, + "loss": 0.9284, + "step": 42082 + }, + { + "epoch": 2.71, + "learning_rate": 1.0991004529938386e-05, + "loss": 0.899, + "step": 42083 + }, + { + "epoch": 2.71, + "learning_rate": 1.0990789840915435e-05, + "loss": 0.8882, + "step": 42084 + }, + { + "epoch": 2.71, + "learning_rate": 1.0990575151892483e-05, + "loss": 0.8797, + "step": 42085 + }, + { + "epoch": 2.71, + "learning_rate": 1.0990360462869534e-05, + "loss": 0.8463, + "step": 42086 + }, + { + "epoch": 2.71, + "learning_rate": 1.0990145773846583e-05, + "loss": 0.8981, + "step": 42087 + }, + { + "epoch": 2.71, + "learning_rate": 1.0989931084823632e-05, + "loss": 0.9277, + "step": 42088 + }, + { + "epoch": 2.71, + "learning_rate": 1.0989716395800683e-05, + "loss": 0.8442, + "step": 42089 + }, + { + "epoch": 2.71, + "learning_rate": 1.0989501706777735e-05, + "loss": 0.8323, + "step": 42090 + }, + { + "epoch": 2.71, + "learning_rate": 1.0989287017754784e-05, + "loss": 0.9035, + "step": 42091 + }, + { + "epoch": 2.71, + "learning_rate": 1.0989072328731833e-05, + "loss": 0.8525, + "step": 42092 + }, + { + "epoch": 2.71, + "learning_rate": 1.0988857639708884e-05, + "loss": 0.8617, + "step": 42093 + }, + { + "epoch": 2.71, + "learning_rate": 1.0988642950685933e-05, + "loss": 0.8845, + "step": 42094 + }, + { + "epoch": 2.71, + "learning_rate": 1.0988428261662982e-05, + "loss": 0.9019, + "step": 42095 + }, + { + "epoch": 2.71, + "learning_rate": 1.0988213572640032e-05, + "loss": 0.9056, + "step": 42096 + }, + { + "epoch": 2.71, + "learning_rate": 1.0987998883617081e-05, + "loss": 0.8643, + "step": 42097 + }, + { + "epoch": 2.71, + "learning_rate": 1.098778419459413e-05, + "loss": 0.9109, + "step": 42098 + }, + { + "epoch": 2.71, + "learning_rate": 1.0987569505571181e-05, + "loss": 0.8588, + "step": 42099 + }, + { + "epoch": 2.71, + "learning_rate": 1.098735481654823e-05, + "loss": 0.9034, + "step": 42100 + }, + { + "epoch": 2.71, + "learning_rate": 1.098714012752528e-05, + "loss": 0.9424, + "step": 42101 + }, + { + "epoch": 2.71, + "learning_rate": 1.098692543850233e-05, + "loss": 0.885, + "step": 42102 + }, + { + "epoch": 2.71, + "learning_rate": 1.0986710749479378e-05, + "loss": 0.8945, + "step": 42103 + }, + { + "epoch": 2.71, + "learning_rate": 1.098649606045643e-05, + "loss": 0.9094, + "step": 42104 + }, + { + "epoch": 2.71, + "learning_rate": 1.098628137143348e-05, + "loss": 0.8967, + "step": 42105 + }, + { + "epoch": 2.71, + "learning_rate": 1.098606668241053e-05, + "loss": 0.8799, + "step": 42106 + }, + { + "epoch": 2.71, + "learning_rate": 1.098585199338758e-05, + "loss": 0.8663, + "step": 42107 + }, + { + "epoch": 2.71, + "learning_rate": 1.0985637304364628e-05, + "loss": 0.8896, + "step": 42108 + }, + { + "epoch": 2.71, + "learning_rate": 1.0985422615341679e-05, + "loss": 0.8986, + "step": 42109 + }, + { + "epoch": 2.71, + "learning_rate": 1.0985207926318728e-05, + "loss": 0.9635, + "step": 42110 + }, + { + "epoch": 2.71, + "learning_rate": 1.0984993237295779e-05, + "loss": 0.9079, + "step": 42111 + }, + { + "epoch": 2.71, + "learning_rate": 1.0984778548272828e-05, + "loss": 0.8655, + "step": 42112 + }, + { + "epoch": 2.71, + "learning_rate": 1.0984563859249877e-05, + "loss": 0.9034, + "step": 42113 + }, + { + "epoch": 2.71, + "learning_rate": 1.0984349170226927e-05, + "loss": 0.902, + "step": 42114 + }, + { + "epoch": 2.71, + "learning_rate": 1.0984134481203976e-05, + "loss": 0.8609, + "step": 42115 + }, + { + "epoch": 2.71, + "learning_rate": 1.0983919792181025e-05, + "loss": 0.9024, + "step": 42116 + }, + { + "epoch": 2.71, + "learning_rate": 1.0983705103158076e-05, + "loss": 0.8947, + "step": 42117 + }, + { + "epoch": 2.71, + "learning_rate": 1.0983490414135125e-05, + "loss": 0.8466, + "step": 42118 + }, + { + "epoch": 2.71, + "learning_rate": 1.0983275725112177e-05, + "loss": 0.9059, + "step": 42119 + }, + { + "epoch": 2.71, + "learning_rate": 1.0983061036089226e-05, + "loss": 0.898, + "step": 42120 + }, + { + "epoch": 2.71, + "learning_rate": 1.0982846347066277e-05, + "loss": 0.8434, + "step": 42121 + }, + { + "epoch": 2.71, + "learning_rate": 1.0982631658043326e-05, + "loss": 0.8679, + "step": 42122 + }, + { + "epoch": 2.71, + "learning_rate": 1.0982416969020375e-05, + "loss": 0.924, + "step": 42123 + }, + { + "epoch": 2.71, + "learning_rate": 1.0982202279997425e-05, + "loss": 0.8782, + "step": 42124 + }, + { + "epoch": 2.71, + "learning_rate": 1.0981987590974474e-05, + "loss": 0.8955, + "step": 42125 + }, + { + "epoch": 2.71, + "learning_rate": 1.0981772901951523e-05, + "loss": 0.873, + "step": 42126 + }, + { + "epoch": 2.71, + "learning_rate": 1.0981558212928574e-05, + "loss": 0.8898, + "step": 42127 + }, + { + "epoch": 2.71, + "learning_rate": 1.0981343523905623e-05, + "loss": 0.8569, + "step": 42128 + }, + { + "epoch": 2.71, + "learning_rate": 1.0981128834882674e-05, + "loss": 0.8674, + "step": 42129 + }, + { + "epoch": 2.71, + "learning_rate": 1.0980914145859722e-05, + "loss": 0.848, + "step": 42130 + }, + { + "epoch": 2.71, + "learning_rate": 1.0980699456836771e-05, + "loss": 0.9031, + "step": 42131 + }, + { + "epoch": 2.71, + "learning_rate": 1.0980484767813822e-05, + "loss": 0.8535, + "step": 42132 + }, + { + "epoch": 2.71, + "learning_rate": 1.0980270078790873e-05, + "loss": 0.9833, + "step": 42133 + }, + { + "epoch": 2.71, + "learning_rate": 1.0980055389767923e-05, + "loss": 0.8978, + "step": 42134 + }, + { + "epoch": 2.71, + "learning_rate": 1.0979840700744972e-05, + "loss": 0.8832, + "step": 42135 + }, + { + "epoch": 2.71, + "learning_rate": 1.0979626011722021e-05, + "loss": 0.927, + "step": 42136 + }, + { + "epoch": 2.71, + "learning_rate": 1.0979411322699072e-05, + "loss": 0.8947, + "step": 42137 + }, + { + "epoch": 2.71, + "learning_rate": 1.0979196633676121e-05, + "loss": 0.8884, + "step": 42138 + }, + { + "epoch": 2.71, + "learning_rate": 1.0978981944653172e-05, + "loss": 0.8775, + "step": 42139 + }, + { + "epoch": 2.71, + "learning_rate": 1.097876725563022e-05, + "loss": 0.8607, + "step": 42140 + }, + { + "epoch": 2.71, + "learning_rate": 1.097855256660727e-05, + "loss": 0.9243, + "step": 42141 + }, + { + "epoch": 2.71, + "learning_rate": 1.097833787758432e-05, + "loss": 0.8807, + "step": 42142 + }, + { + "epoch": 2.71, + "learning_rate": 1.097812318856137e-05, + "loss": 0.8503, + "step": 42143 + }, + { + "epoch": 2.71, + "learning_rate": 1.0977908499538418e-05, + "loss": 0.9027, + "step": 42144 + }, + { + "epoch": 2.71, + "learning_rate": 1.0977693810515469e-05, + "loss": 0.8967, + "step": 42145 + }, + { + "epoch": 2.71, + "learning_rate": 1.0977479121492518e-05, + "loss": 0.9261, + "step": 42146 + }, + { + "epoch": 2.71, + "learning_rate": 1.0977264432469568e-05, + "loss": 0.8155, + "step": 42147 + }, + { + "epoch": 2.71, + "learning_rate": 1.0977049743446619e-05, + "loss": 0.9145, + "step": 42148 + }, + { + "epoch": 2.71, + "learning_rate": 1.097683505442367e-05, + "loss": 0.8969, + "step": 42149 + }, + { + "epoch": 2.71, + "learning_rate": 1.0976620365400719e-05, + "loss": 0.8866, + "step": 42150 + }, + { + "epoch": 2.71, + "learning_rate": 1.0976405676377768e-05, + "loss": 0.9203, + "step": 42151 + }, + { + "epoch": 2.71, + "learning_rate": 1.0976190987354818e-05, + "loss": 0.9323, + "step": 42152 + }, + { + "epoch": 2.71, + "learning_rate": 1.0975976298331867e-05, + "loss": 0.8625, + "step": 42153 + }, + { + "epoch": 2.71, + "learning_rate": 1.0975761609308916e-05, + "loss": 0.8785, + "step": 42154 + }, + { + "epoch": 2.71, + "learning_rate": 1.0975546920285967e-05, + "loss": 0.9069, + "step": 42155 + }, + { + "epoch": 2.71, + "learning_rate": 1.0975332231263016e-05, + "loss": 0.898, + "step": 42156 + }, + { + "epoch": 2.71, + "learning_rate": 1.0975117542240067e-05, + "loss": 0.9253, + "step": 42157 + }, + { + "epoch": 2.71, + "learning_rate": 1.0974902853217116e-05, + "loss": 0.8597, + "step": 42158 + }, + { + "epoch": 2.71, + "learning_rate": 1.0974688164194164e-05, + "loss": 0.869, + "step": 42159 + }, + { + "epoch": 2.71, + "learning_rate": 1.0974473475171215e-05, + "loss": 0.9165, + "step": 42160 + }, + { + "epoch": 2.71, + "learning_rate": 1.0974258786148264e-05, + "loss": 0.9469, + "step": 42161 + }, + { + "epoch": 2.71, + "learning_rate": 1.0974044097125316e-05, + "loss": 0.9044, + "step": 42162 + }, + { + "epoch": 2.71, + "learning_rate": 1.0973829408102365e-05, + "loss": 0.871, + "step": 42163 + }, + { + "epoch": 2.71, + "learning_rate": 1.0973614719079414e-05, + "loss": 0.8996, + "step": 42164 + }, + { + "epoch": 2.71, + "learning_rate": 1.0973400030056465e-05, + "loss": 0.9176, + "step": 42165 + }, + { + "epoch": 2.71, + "learning_rate": 1.0973185341033514e-05, + "loss": 0.9378, + "step": 42166 + }, + { + "epoch": 2.71, + "learning_rate": 1.0972970652010565e-05, + "loss": 0.9194, + "step": 42167 + }, + { + "epoch": 2.71, + "learning_rate": 1.0972755962987614e-05, + "loss": 0.9848, + "step": 42168 + }, + { + "epoch": 2.71, + "learning_rate": 1.0972541273964663e-05, + "loss": 0.9271, + "step": 42169 + }, + { + "epoch": 2.71, + "learning_rate": 1.0972326584941713e-05, + "loss": 0.8733, + "step": 42170 + }, + { + "epoch": 2.71, + "learning_rate": 1.0972111895918762e-05, + "loss": 0.8778, + "step": 42171 + }, + { + "epoch": 2.71, + "learning_rate": 1.0971897206895811e-05, + "loss": 0.8992, + "step": 42172 + }, + { + "epoch": 2.71, + "learning_rate": 1.0971682517872862e-05, + "loss": 0.887, + "step": 42173 + }, + { + "epoch": 2.71, + "learning_rate": 1.097146782884991e-05, + "loss": 0.8811, + "step": 42174 + }, + { + "epoch": 2.71, + "learning_rate": 1.097125313982696e-05, + "loss": 0.897, + "step": 42175 + }, + { + "epoch": 2.71, + "learning_rate": 1.097103845080401e-05, + "loss": 0.8866, + "step": 42176 + }, + { + "epoch": 2.71, + "learning_rate": 1.0970823761781063e-05, + "loss": 0.9044, + "step": 42177 + }, + { + "epoch": 2.71, + "learning_rate": 1.0970609072758112e-05, + "loss": 0.8868, + "step": 42178 + }, + { + "epoch": 2.71, + "learning_rate": 1.097039438373516e-05, + "loss": 0.9307, + "step": 42179 + }, + { + "epoch": 2.71, + "learning_rate": 1.0970179694712211e-05, + "loss": 0.9315, + "step": 42180 + }, + { + "epoch": 2.71, + "learning_rate": 1.096996500568926e-05, + "loss": 0.883, + "step": 42181 + }, + { + "epoch": 2.71, + "learning_rate": 1.096975031666631e-05, + "loss": 0.8954, + "step": 42182 + }, + { + "epoch": 2.71, + "learning_rate": 1.096953562764336e-05, + "loss": 0.9352, + "step": 42183 + }, + { + "epoch": 2.71, + "learning_rate": 1.0969320938620409e-05, + "loss": 0.9871, + "step": 42184 + }, + { + "epoch": 2.71, + "learning_rate": 1.0969106249597458e-05, + "loss": 0.8874, + "step": 42185 + }, + { + "epoch": 2.71, + "learning_rate": 1.0968891560574509e-05, + "loss": 0.8802, + "step": 42186 + }, + { + "epoch": 2.71, + "learning_rate": 1.0968676871551558e-05, + "loss": 0.9184, + "step": 42187 + }, + { + "epoch": 2.71, + "learning_rate": 1.0968462182528608e-05, + "loss": 0.9554, + "step": 42188 + }, + { + "epoch": 2.71, + "learning_rate": 1.0968247493505657e-05, + "loss": 0.8907, + "step": 42189 + }, + { + "epoch": 2.71, + "learning_rate": 1.0968032804482706e-05, + "loss": 0.9642, + "step": 42190 + }, + { + "epoch": 2.71, + "learning_rate": 1.0967818115459758e-05, + "loss": 0.8902, + "step": 42191 + }, + { + "epoch": 2.71, + "learning_rate": 1.0967603426436807e-05, + "loss": 0.9062, + "step": 42192 + }, + { + "epoch": 2.71, + "learning_rate": 1.0967388737413858e-05, + "loss": 0.9225, + "step": 42193 + }, + { + "epoch": 2.71, + "learning_rate": 1.0967174048390907e-05, + "loss": 0.9261, + "step": 42194 + }, + { + "epoch": 2.71, + "learning_rate": 1.0966959359367956e-05, + "loss": 0.9136, + "step": 42195 + }, + { + "epoch": 2.71, + "learning_rate": 1.0966744670345007e-05, + "loss": 0.8692, + "step": 42196 + }, + { + "epoch": 2.71, + "learning_rate": 1.0966529981322056e-05, + "loss": 0.8918, + "step": 42197 + }, + { + "epoch": 2.71, + "learning_rate": 1.0966315292299106e-05, + "loss": 0.9222, + "step": 42198 + }, + { + "epoch": 2.71, + "learning_rate": 1.0966100603276155e-05, + "loss": 0.9248, + "step": 42199 + }, + { + "epoch": 2.72, + "learning_rate": 1.0965885914253204e-05, + "loss": 0.9369, + "step": 42200 + }, + { + "epoch": 2.72, + "learning_rate": 1.0965671225230255e-05, + "loss": 0.8632, + "step": 42201 + }, + { + "epoch": 2.72, + "learning_rate": 1.0965456536207304e-05, + "loss": 0.9156, + "step": 42202 + }, + { + "epoch": 2.72, + "learning_rate": 1.0965241847184353e-05, + "loss": 0.943, + "step": 42203 + }, + { + "epoch": 2.72, + "learning_rate": 1.0965027158161403e-05, + "loss": 0.841, + "step": 42204 + }, + { + "epoch": 2.72, + "learning_rate": 1.0964812469138452e-05, + "loss": 0.9208, + "step": 42205 + }, + { + "epoch": 2.72, + "learning_rate": 1.0964597780115505e-05, + "loss": 0.9176, + "step": 42206 + }, + { + "epoch": 2.72, + "learning_rate": 1.0964383091092554e-05, + "loss": 0.83, + "step": 42207 + }, + { + "epoch": 2.72, + "learning_rate": 1.0964168402069604e-05, + "loss": 0.8574, + "step": 42208 + }, + { + "epoch": 2.72, + "learning_rate": 1.0963953713046653e-05, + "loss": 0.8959, + "step": 42209 + }, + { + "epoch": 2.72, + "learning_rate": 1.0963739024023702e-05, + "loss": 0.8376, + "step": 42210 + }, + { + "epoch": 2.72, + "learning_rate": 1.0963524335000753e-05, + "loss": 0.9037, + "step": 42211 + }, + { + "epoch": 2.72, + "learning_rate": 1.0963309645977802e-05, + "loss": 0.9034, + "step": 42212 + }, + { + "epoch": 2.72, + "learning_rate": 1.0963094956954851e-05, + "loss": 0.9512, + "step": 42213 + }, + { + "epoch": 2.72, + "learning_rate": 1.0962880267931902e-05, + "loss": 0.8724, + "step": 42214 + }, + { + "epoch": 2.72, + "learning_rate": 1.096266557890895e-05, + "loss": 0.826, + "step": 42215 + }, + { + "epoch": 2.72, + "learning_rate": 1.0962450889886001e-05, + "loss": 0.9119, + "step": 42216 + }, + { + "epoch": 2.72, + "learning_rate": 1.096223620086305e-05, + "loss": 0.8624, + "step": 42217 + }, + { + "epoch": 2.72, + "learning_rate": 1.09620215118401e-05, + "loss": 0.8656, + "step": 42218 + }, + { + "epoch": 2.72, + "learning_rate": 1.096180682281715e-05, + "loss": 0.8439, + "step": 42219 + }, + { + "epoch": 2.72, + "learning_rate": 1.09615921337942e-05, + "loss": 0.8454, + "step": 42220 + }, + { + "epoch": 2.72, + "learning_rate": 1.0961377444771251e-05, + "loss": 0.8831, + "step": 42221 + }, + { + "epoch": 2.72, + "learning_rate": 1.09611627557483e-05, + "loss": 0.8817, + "step": 42222 + }, + { + "epoch": 2.72, + "learning_rate": 1.0960948066725349e-05, + "loss": 0.889, + "step": 42223 + }, + { + "epoch": 2.72, + "learning_rate": 1.09607333777024e-05, + "loss": 0.9144, + "step": 42224 + }, + { + "epoch": 2.72, + "learning_rate": 1.0960518688679449e-05, + "loss": 0.9627, + "step": 42225 + }, + { + "epoch": 2.72, + "learning_rate": 1.09603039996565e-05, + "loss": 0.864, + "step": 42226 + }, + { + "epoch": 2.72, + "learning_rate": 1.0960089310633548e-05, + "loss": 0.9456, + "step": 42227 + }, + { + "epoch": 2.72, + "learning_rate": 1.0959874621610597e-05, + "loss": 0.8076, + "step": 42228 + }, + { + "epoch": 2.72, + "learning_rate": 1.0959659932587648e-05, + "loss": 0.9115, + "step": 42229 + }, + { + "epoch": 2.72, + "learning_rate": 1.0959445243564697e-05, + "loss": 0.9067, + "step": 42230 + }, + { + "epoch": 2.72, + "learning_rate": 1.0959230554541746e-05, + "loss": 0.8894, + "step": 42231 + }, + { + "epoch": 2.72, + "learning_rate": 1.0959015865518797e-05, + "loss": 0.8986, + "step": 42232 + }, + { + "epoch": 2.72, + "learning_rate": 1.0958801176495846e-05, + "loss": 0.8329, + "step": 42233 + }, + { + "epoch": 2.72, + "learning_rate": 1.0958586487472896e-05, + "loss": 0.9129, + "step": 42234 + }, + { + "epoch": 2.72, + "learning_rate": 1.0958371798449947e-05, + "loss": 0.8309, + "step": 42235 + }, + { + "epoch": 2.72, + "learning_rate": 1.0958157109426997e-05, + "loss": 0.9532, + "step": 42236 + }, + { + "epoch": 2.72, + "learning_rate": 1.0957942420404046e-05, + "loss": 0.9036, + "step": 42237 + }, + { + "epoch": 2.72, + "learning_rate": 1.0957727731381095e-05, + "loss": 0.9174, + "step": 42238 + }, + { + "epoch": 2.72, + "learning_rate": 1.0957513042358146e-05, + "loss": 0.8805, + "step": 42239 + }, + { + "epoch": 2.72, + "learning_rate": 1.0957298353335195e-05, + "loss": 0.9928, + "step": 42240 + }, + { + "epoch": 2.72, + "learning_rate": 1.0957083664312244e-05, + "loss": 0.9309, + "step": 42241 + }, + { + "epoch": 2.72, + "learning_rate": 1.0956868975289295e-05, + "loss": 0.9426, + "step": 42242 + }, + { + "epoch": 2.72, + "learning_rate": 1.0956654286266344e-05, + "loss": 0.9251, + "step": 42243 + }, + { + "epoch": 2.72, + "learning_rate": 1.0956439597243394e-05, + "loss": 0.8946, + "step": 42244 + }, + { + "epoch": 2.72, + "learning_rate": 1.0956224908220443e-05, + "loss": 0.8536, + "step": 42245 + }, + { + "epoch": 2.72, + "learning_rate": 1.0956010219197492e-05, + "loss": 0.8694, + "step": 42246 + }, + { + "epoch": 2.72, + "learning_rate": 1.0955795530174543e-05, + "loss": 0.9214, + "step": 42247 + }, + { + "epoch": 2.72, + "learning_rate": 1.0955580841151592e-05, + "loss": 0.9103, + "step": 42248 + }, + { + "epoch": 2.72, + "learning_rate": 1.0955366152128644e-05, + "loss": 0.8607, + "step": 42249 + }, + { + "epoch": 2.72, + "learning_rate": 1.0955151463105693e-05, + "loss": 0.8824, + "step": 42250 + }, + { + "epoch": 2.72, + "learning_rate": 1.0954936774082742e-05, + "loss": 0.8881, + "step": 42251 + }, + { + "epoch": 2.72, + "learning_rate": 1.0954722085059793e-05, + "loss": 0.8712, + "step": 42252 + }, + { + "epoch": 2.72, + "learning_rate": 1.0954507396036842e-05, + "loss": 0.897, + "step": 42253 + }, + { + "epoch": 2.72, + "learning_rate": 1.0954292707013892e-05, + "loss": 0.962, + "step": 42254 + }, + { + "epoch": 2.72, + "learning_rate": 1.0954078017990941e-05, + "loss": 0.8918, + "step": 42255 + }, + { + "epoch": 2.72, + "learning_rate": 1.095386332896799e-05, + "loss": 0.8797, + "step": 42256 + }, + { + "epoch": 2.72, + "learning_rate": 1.0953648639945041e-05, + "loss": 0.9317, + "step": 42257 + }, + { + "epoch": 2.72, + "learning_rate": 1.095343395092209e-05, + "loss": 0.8318, + "step": 42258 + }, + { + "epoch": 2.72, + "learning_rate": 1.0953219261899139e-05, + "loss": 0.859, + "step": 42259 + }, + { + "epoch": 2.72, + "learning_rate": 1.095300457287619e-05, + "loss": 0.7916, + "step": 42260 + }, + { + "epoch": 2.72, + "learning_rate": 1.0952789883853239e-05, + "loss": 0.9179, + "step": 42261 + }, + { + "epoch": 2.72, + "learning_rate": 1.0952575194830288e-05, + "loss": 0.8492, + "step": 42262 + }, + { + "epoch": 2.72, + "learning_rate": 1.0952360505807338e-05, + "loss": 0.9237, + "step": 42263 + }, + { + "epoch": 2.72, + "learning_rate": 1.095214581678439e-05, + "loss": 0.911, + "step": 42264 + }, + { + "epoch": 2.72, + "learning_rate": 1.095193112776144e-05, + "loss": 0.9018, + "step": 42265 + }, + { + "epoch": 2.72, + "learning_rate": 1.0951716438738488e-05, + "loss": 0.8454, + "step": 42266 + }, + { + "epoch": 2.72, + "learning_rate": 1.0951501749715539e-05, + "loss": 0.8676, + "step": 42267 + }, + { + "epoch": 2.72, + "learning_rate": 1.0951287060692588e-05, + "loss": 0.9108, + "step": 42268 + }, + { + "epoch": 2.72, + "learning_rate": 1.0951072371669637e-05, + "loss": 0.9171, + "step": 42269 + }, + { + "epoch": 2.72, + "learning_rate": 1.0950857682646688e-05, + "loss": 0.9297, + "step": 42270 + }, + { + "epoch": 2.72, + "learning_rate": 1.0950642993623737e-05, + "loss": 0.9729, + "step": 42271 + }, + { + "epoch": 2.72, + "learning_rate": 1.0950428304600786e-05, + "loss": 0.9141, + "step": 42272 + }, + { + "epoch": 2.72, + "learning_rate": 1.0950213615577836e-05, + "loss": 0.905, + "step": 42273 + }, + { + "epoch": 2.72, + "learning_rate": 1.0949998926554885e-05, + "loss": 0.9591, + "step": 42274 + }, + { + "epoch": 2.72, + "learning_rate": 1.0949784237531936e-05, + "loss": 0.9104, + "step": 42275 + }, + { + "epoch": 2.72, + "learning_rate": 1.0949569548508985e-05, + "loss": 0.8487, + "step": 42276 + }, + { + "epoch": 2.72, + "learning_rate": 1.0949354859486034e-05, + "loss": 0.8799, + "step": 42277 + }, + { + "epoch": 2.72, + "learning_rate": 1.0949140170463086e-05, + "loss": 0.8965, + "step": 42278 + }, + { + "epoch": 2.72, + "learning_rate": 1.0948925481440135e-05, + "loss": 0.9776, + "step": 42279 + }, + { + "epoch": 2.72, + "learning_rate": 1.0948710792417186e-05, + "loss": 0.8974, + "step": 42280 + }, + { + "epoch": 2.72, + "learning_rate": 1.0948496103394235e-05, + "loss": 0.9232, + "step": 42281 + }, + { + "epoch": 2.72, + "learning_rate": 1.0948281414371284e-05, + "loss": 0.9296, + "step": 42282 + }, + { + "epoch": 2.72, + "learning_rate": 1.0948066725348334e-05, + "loss": 0.8818, + "step": 42283 + }, + { + "epoch": 2.72, + "learning_rate": 1.0947852036325383e-05, + "loss": 0.8835, + "step": 42284 + }, + { + "epoch": 2.72, + "learning_rate": 1.0947637347302434e-05, + "loss": 0.8724, + "step": 42285 + }, + { + "epoch": 2.72, + "learning_rate": 1.0947422658279483e-05, + "loss": 0.9217, + "step": 42286 + }, + { + "epoch": 2.72, + "learning_rate": 1.0947207969256532e-05, + "loss": 0.9185, + "step": 42287 + }, + { + "epoch": 2.72, + "learning_rate": 1.0946993280233583e-05, + "loss": 0.9261, + "step": 42288 + }, + { + "epoch": 2.72, + "learning_rate": 1.0946778591210632e-05, + "loss": 0.9098, + "step": 42289 + }, + { + "epoch": 2.72, + "learning_rate": 1.094656390218768e-05, + "loss": 0.9044, + "step": 42290 + }, + { + "epoch": 2.72, + "learning_rate": 1.0946349213164731e-05, + "loss": 0.9685, + "step": 42291 + }, + { + "epoch": 2.72, + "learning_rate": 1.094613452414178e-05, + "loss": 0.8839, + "step": 42292 + }, + { + "epoch": 2.72, + "learning_rate": 1.0945919835118833e-05, + "loss": 0.8947, + "step": 42293 + }, + { + "epoch": 2.72, + "learning_rate": 1.0945705146095881e-05, + "loss": 0.9118, + "step": 42294 + }, + { + "epoch": 2.72, + "learning_rate": 1.0945490457072932e-05, + "loss": 0.8975, + "step": 42295 + }, + { + "epoch": 2.72, + "learning_rate": 1.0945275768049981e-05, + "loss": 0.9201, + "step": 42296 + }, + { + "epoch": 2.72, + "learning_rate": 1.094506107902703e-05, + "loss": 0.9968, + "step": 42297 + }, + { + "epoch": 2.72, + "learning_rate": 1.094484639000408e-05, + "loss": 0.916, + "step": 42298 + }, + { + "epoch": 2.72, + "learning_rate": 1.094463170098113e-05, + "loss": 0.9342, + "step": 42299 + }, + { + "epoch": 2.72, + "learning_rate": 1.0944417011958179e-05, + "loss": 0.8563, + "step": 42300 + }, + { + "epoch": 2.72, + "learning_rate": 1.094420232293523e-05, + "loss": 0.9419, + "step": 42301 + }, + { + "epoch": 2.72, + "learning_rate": 1.0943987633912278e-05, + "loss": 0.8478, + "step": 42302 + }, + { + "epoch": 2.72, + "learning_rate": 1.0943772944889329e-05, + "loss": 0.9161, + "step": 42303 + }, + { + "epoch": 2.72, + "learning_rate": 1.0943558255866378e-05, + "loss": 0.8196, + "step": 42304 + }, + { + "epoch": 2.72, + "learning_rate": 1.0943343566843427e-05, + "loss": 0.9428, + "step": 42305 + }, + { + "epoch": 2.72, + "learning_rate": 1.0943128877820478e-05, + "loss": 0.8477, + "step": 42306 + }, + { + "epoch": 2.72, + "learning_rate": 1.0942914188797528e-05, + "loss": 0.9259, + "step": 42307 + }, + { + "epoch": 2.72, + "learning_rate": 1.0942699499774579e-05, + "loss": 0.9027, + "step": 42308 + }, + { + "epoch": 2.72, + "learning_rate": 1.0942484810751628e-05, + "loss": 0.9379, + "step": 42309 + }, + { + "epoch": 2.72, + "learning_rate": 1.0942270121728677e-05, + "loss": 0.8724, + "step": 42310 + }, + { + "epoch": 2.72, + "learning_rate": 1.0942055432705727e-05, + "loss": 0.8845, + "step": 42311 + }, + { + "epoch": 2.72, + "learning_rate": 1.0941840743682776e-05, + "loss": 0.8591, + "step": 42312 + }, + { + "epoch": 2.72, + "learning_rate": 1.0941626054659827e-05, + "loss": 0.8664, + "step": 42313 + }, + { + "epoch": 2.72, + "learning_rate": 1.0941411365636876e-05, + "loss": 0.8838, + "step": 42314 + }, + { + "epoch": 2.72, + "learning_rate": 1.0941196676613925e-05, + "loss": 0.8394, + "step": 42315 + }, + { + "epoch": 2.72, + "learning_rate": 1.0940981987590976e-05, + "loss": 0.8476, + "step": 42316 + }, + { + "epoch": 2.72, + "learning_rate": 1.0940767298568025e-05, + "loss": 0.9039, + "step": 42317 + }, + { + "epoch": 2.72, + "learning_rate": 1.0940552609545074e-05, + "loss": 0.9429, + "step": 42318 + }, + { + "epoch": 2.72, + "learning_rate": 1.0940337920522124e-05, + "loss": 0.8566, + "step": 42319 + }, + { + "epoch": 2.72, + "learning_rate": 1.0940123231499173e-05, + "loss": 0.8791, + "step": 42320 + }, + { + "epoch": 2.72, + "learning_rate": 1.0939908542476224e-05, + "loss": 0.9088, + "step": 42321 + }, + { + "epoch": 2.72, + "learning_rate": 1.0939693853453275e-05, + "loss": 0.9142, + "step": 42322 + }, + { + "epoch": 2.72, + "learning_rate": 1.0939479164430325e-05, + "loss": 0.9005, + "step": 42323 + }, + { + "epoch": 2.72, + "learning_rate": 1.0939264475407374e-05, + "loss": 0.8648, + "step": 42324 + }, + { + "epoch": 2.72, + "learning_rate": 1.0939049786384423e-05, + "loss": 0.9421, + "step": 42325 + }, + { + "epoch": 2.72, + "learning_rate": 1.0938835097361474e-05, + "loss": 0.8239, + "step": 42326 + }, + { + "epoch": 2.72, + "learning_rate": 1.0938620408338523e-05, + "loss": 0.8825, + "step": 42327 + }, + { + "epoch": 2.72, + "learning_rate": 1.0938405719315572e-05, + "loss": 0.8767, + "step": 42328 + }, + { + "epoch": 2.72, + "learning_rate": 1.0938191030292622e-05, + "loss": 0.9242, + "step": 42329 + }, + { + "epoch": 2.72, + "learning_rate": 1.0937976341269671e-05, + "loss": 0.9224, + "step": 42330 + }, + { + "epoch": 2.72, + "learning_rate": 1.0937761652246722e-05, + "loss": 0.9309, + "step": 42331 + }, + { + "epoch": 2.72, + "learning_rate": 1.0937546963223771e-05, + "loss": 0.9098, + "step": 42332 + }, + { + "epoch": 2.72, + "learning_rate": 1.093733227420082e-05, + "loss": 0.9178, + "step": 42333 + }, + { + "epoch": 2.72, + "learning_rate": 1.093711758517787e-05, + "loss": 0.8947, + "step": 42334 + }, + { + "epoch": 2.72, + "learning_rate": 1.093690289615492e-05, + "loss": 0.8621, + "step": 42335 + }, + { + "epoch": 2.72, + "learning_rate": 1.0936688207131972e-05, + "loss": 0.8851, + "step": 42336 + }, + { + "epoch": 2.72, + "learning_rate": 1.0936473518109021e-05, + "loss": 0.8722, + "step": 42337 + }, + { + "epoch": 2.72, + "learning_rate": 1.093625882908607e-05, + "loss": 0.9237, + "step": 42338 + }, + { + "epoch": 2.72, + "learning_rate": 1.093604414006312e-05, + "loss": 0.8291, + "step": 42339 + }, + { + "epoch": 2.72, + "learning_rate": 1.093582945104017e-05, + "loss": 0.9449, + "step": 42340 + }, + { + "epoch": 2.72, + "learning_rate": 1.093561476201722e-05, + "loss": 0.9015, + "step": 42341 + }, + { + "epoch": 2.72, + "learning_rate": 1.0935400072994269e-05, + "loss": 0.9084, + "step": 42342 + }, + { + "epoch": 2.72, + "learning_rate": 1.0935185383971318e-05, + "loss": 0.9029, + "step": 42343 + }, + { + "epoch": 2.72, + "learning_rate": 1.0934970694948369e-05, + "loss": 0.9278, + "step": 42344 + }, + { + "epoch": 2.72, + "learning_rate": 1.0934756005925418e-05, + "loss": 0.8758, + "step": 42345 + }, + { + "epoch": 2.72, + "learning_rate": 1.0934541316902467e-05, + "loss": 0.9235, + "step": 42346 + }, + { + "epoch": 2.72, + "learning_rate": 1.0934326627879517e-05, + "loss": 0.9625, + "step": 42347 + }, + { + "epoch": 2.72, + "learning_rate": 1.0934111938856566e-05, + "loss": 0.8858, + "step": 42348 + }, + { + "epoch": 2.72, + "learning_rate": 1.0933897249833615e-05, + "loss": 0.9009, + "step": 42349 + }, + { + "epoch": 2.72, + "learning_rate": 1.0933682560810666e-05, + "loss": 0.9644, + "step": 42350 + }, + { + "epoch": 2.72, + "learning_rate": 1.0933467871787718e-05, + "loss": 0.9922, + "step": 42351 + }, + { + "epoch": 2.72, + "learning_rate": 1.0933253182764767e-05, + "loss": 0.9571, + "step": 42352 + }, + { + "epoch": 2.72, + "learning_rate": 1.0933038493741816e-05, + "loss": 0.8368, + "step": 42353 + }, + { + "epoch": 2.72, + "learning_rate": 1.0932823804718867e-05, + "loss": 0.8837, + "step": 42354 + }, + { + "epoch": 2.73, + "learning_rate": 1.0932609115695916e-05, + "loss": 0.9442, + "step": 42355 + }, + { + "epoch": 2.73, + "learning_rate": 1.0932394426672965e-05, + "loss": 0.9303, + "step": 42356 + }, + { + "epoch": 2.73, + "learning_rate": 1.0932179737650015e-05, + "loss": 0.8704, + "step": 42357 + }, + { + "epoch": 2.73, + "learning_rate": 1.0931965048627064e-05, + "loss": 0.885, + "step": 42358 + }, + { + "epoch": 2.73, + "learning_rate": 1.0931750359604113e-05, + "loss": 0.9008, + "step": 42359 + }, + { + "epoch": 2.73, + "learning_rate": 1.0931535670581164e-05, + "loss": 0.8996, + "step": 42360 + }, + { + "epoch": 2.73, + "learning_rate": 1.0931320981558213e-05, + "loss": 0.8791, + "step": 42361 + }, + { + "epoch": 2.73, + "learning_rate": 1.0931106292535264e-05, + "loss": 0.876, + "step": 42362 + }, + { + "epoch": 2.73, + "learning_rate": 1.0930891603512313e-05, + "loss": 0.8623, + "step": 42363 + }, + { + "epoch": 2.73, + "learning_rate": 1.0930676914489362e-05, + "loss": 0.8471, + "step": 42364 + }, + { + "epoch": 2.73, + "learning_rate": 1.0930462225466414e-05, + "loss": 0.9182, + "step": 42365 + }, + { + "epoch": 2.73, + "learning_rate": 1.0930247536443463e-05, + "loss": 0.8803, + "step": 42366 + }, + { + "epoch": 2.73, + "learning_rate": 1.0930032847420514e-05, + "loss": 0.9027, + "step": 42367 + }, + { + "epoch": 2.73, + "learning_rate": 1.0929818158397562e-05, + "loss": 0.8533, + "step": 42368 + }, + { + "epoch": 2.73, + "learning_rate": 1.0929603469374611e-05, + "loss": 0.8834, + "step": 42369 + }, + { + "epoch": 2.73, + "learning_rate": 1.0929388780351662e-05, + "loss": 0.8521, + "step": 42370 + }, + { + "epoch": 2.73, + "learning_rate": 1.0929174091328711e-05, + "loss": 0.9684, + "step": 42371 + }, + { + "epoch": 2.73, + "learning_rate": 1.0928959402305762e-05, + "loss": 0.904, + "step": 42372 + }, + { + "epoch": 2.73, + "learning_rate": 1.092874471328281e-05, + "loss": 0.8772, + "step": 42373 + }, + { + "epoch": 2.73, + "learning_rate": 1.092853002425986e-05, + "loss": 0.8764, + "step": 42374 + }, + { + "epoch": 2.73, + "learning_rate": 1.092831533523691e-05, + "loss": 0.8152, + "step": 42375 + }, + { + "epoch": 2.73, + "learning_rate": 1.092810064621396e-05, + "loss": 0.9219, + "step": 42376 + }, + { + "epoch": 2.73, + "learning_rate": 1.0927885957191008e-05, + "loss": 0.93, + "step": 42377 + }, + { + "epoch": 2.73, + "learning_rate": 1.0927671268168059e-05, + "loss": 0.8654, + "step": 42378 + }, + { + "epoch": 2.73, + "learning_rate": 1.0927456579145108e-05, + "loss": 0.8853, + "step": 42379 + }, + { + "epoch": 2.73, + "learning_rate": 1.092724189012216e-05, + "loss": 0.9584, + "step": 42380 + }, + { + "epoch": 2.73, + "learning_rate": 1.092702720109921e-05, + "loss": 0.9007, + "step": 42381 + }, + { + "epoch": 2.73, + "learning_rate": 1.092681251207626e-05, + "loss": 0.8485, + "step": 42382 + }, + { + "epoch": 2.73, + "learning_rate": 1.0926597823053309e-05, + "loss": 0.9532, + "step": 42383 + }, + { + "epoch": 2.73, + "learning_rate": 1.0926383134030358e-05, + "loss": 0.929, + "step": 42384 + }, + { + "epoch": 2.73, + "learning_rate": 1.0926168445007408e-05, + "loss": 0.7865, + "step": 42385 + }, + { + "epoch": 2.73, + "learning_rate": 1.0925953755984457e-05, + "loss": 0.8542, + "step": 42386 + }, + { + "epoch": 2.73, + "learning_rate": 1.0925739066961506e-05, + "loss": 0.9145, + "step": 42387 + }, + { + "epoch": 2.73, + "learning_rate": 1.0925524377938557e-05, + "loss": 0.8997, + "step": 42388 + }, + { + "epoch": 2.73, + "learning_rate": 1.0925309688915606e-05, + "loss": 0.7982, + "step": 42389 + }, + { + "epoch": 2.73, + "learning_rate": 1.0925094999892657e-05, + "loss": 0.9029, + "step": 42390 + }, + { + "epoch": 2.73, + "learning_rate": 1.0924880310869706e-05, + "loss": 0.9254, + "step": 42391 + }, + { + "epoch": 2.73, + "learning_rate": 1.0924665621846755e-05, + "loss": 0.8882, + "step": 42392 + }, + { + "epoch": 2.73, + "learning_rate": 1.0924450932823805e-05, + "loss": 0.9157, + "step": 42393 + }, + { + "epoch": 2.73, + "learning_rate": 1.0924236243800856e-05, + "loss": 0.9404, + "step": 42394 + }, + { + "epoch": 2.73, + "learning_rate": 1.0924021554777907e-05, + "loss": 0.8986, + "step": 42395 + }, + { + "epoch": 2.73, + "learning_rate": 1.0923806865754956e-05, + "loss": 0.8754, + "step": 42396 + }, + { + "epoch": 2.73, + "learning_rate": 1.0923592176732004e-05, + "loss": 0.942, + "step": 42397 + }, + { + "epoch": 2.73, + "learning_rate": 1.0923377487709055e-05, + "loss": 0.9045, + "step": 42398 + }, + { + "epoch": 2.73, + "learning_rate": 1.0923162798686104e-05, + "loss": 0.8949, + "step": 42399 + }, + { + "epoch": 2.73, + "learning_rate": 1.0922948109663155e-05, + "loss": 0.9127, + "step": 42400 + }, + { + "epoch": 2.73, + "learning_rate": 1.0922733420640204e-05, + "loss": 0.8488, + "step": 42401 + }, + { + "epoch": 2.73, + "learning_rate": 1.0922518731617253e-05, + "loss": 0.8897, + "step": 42402 + }, + { + "epoch": 2.73, + "learning_rate": 1.0922304042594303e-05, + "loss": 0.8307, + "step": 42403 + }, + { + "epoch": 2.73, + "learning_rate": 1.0922089353571352e-05, + "loss": 0.8835, + "step": 42404 + }, + { + "epoch": 2.73, + "learning_rate": 1.0921874664548401e-05, + "loss": 0.8842, + "step": 42405 + }, + { + "epoch": 2.73, + "learning_rate": 1.0921659975525452e-05, + "loss": 0.9453, + "step": 42406 + }, + { + "epoch": 2.73, + "learning_rate": 1.0921445286502501e-05, + "loss": 0.8906, + "step": 42407 + }, + { + "epoch": 2.73, + "learning_rate": 1.092123059747955e-05, + "loss": 0.7778, + "step": 42408 + }, + { + "epoch": 2.73, + "learning_rate": 1.0921015908456602e-05, + "loss": 0.9194, + "step": 42409 + }, + { + "epoch": 2.73, + "learning_rate": 1.0920801219433653e-05, + "loss": 0.9136, + "step": 42410 + }, + { + "epoch": 2.73, + "learning_rate": 1.0920586530410702e-05, + "loss": 0.8995, + "step": 42411 + }, + { + "epoch": 2.73, + "learning_rate": 1.092037184138775e-05, + "loss": 0.9206, + "step": 42412 + }, + { + "epoch": 2.73, + "learning_rate": 1.0920157152364801e-05, + "loss": 0.8805, + "step": 42413 + }, + { + "epoch": 2.73, + "learning_rate": 1.091994246334185e-05, + "loss": 0.8876, + "step": 42414 + }, + { + "epoch": 2.73, + "learning_rate": 1.09197277743189e-05, + "loss": 0.8465, + "step": 42415 + }, + { + "epoch": 2.73, + "learning_rate": 1.091951308529595e-05, + "loss": 0.8867, + "step": 42416 + }, + { + "epoch": 2.73, + "learning_rate": 1.0919298396272999e-05, + "loss": 0.8808, + "step": 42417 + }, + { + "epoch": 2.73, + "learning_rate": 1.0919083707250048e-05, + "loss": 0.9, + "step": 42418 + }, + { + "epoch": 2.73, + "learning_rate": 1.0918869018227099e-05, + "loss": 0.8666, + "step": 42419 + }, + { + "epoch": 2.73, + "learning_rate": 1.0918654329204148e-05, + "loss": 0.8639, + "step": 42420 + }, + { + "epoch": 2.73, + "learning_rate": 1.0918439640181198e-05, + "loss": 0.8654, + "step": 42421 + }, + { + "epoch": 2.73, + "learning_rate": 1.0918224951158247e-05, + "loss": 0.8912, + "step": 42422 + }, + { + "epoch": 2.73, + "learning_rate": 1.09180102621353e-05, + "loss": 0.8776, + "step": 42423 + }, + { + "epoch": 2.73, + "learning_rate": 1.0917795573112349e-05, + "loss": 0.9068, + "step": 42424 + }, + { + "epoch": 2.73, + "learning_rate": 1.0917580884089398e-05, + "loss": 0.9556, + "step": 42425 + }, + { + "epoch": 2.73, + "learning_rate": 1.0917366195066448e-05, + "loss": 0.9232, + "step": 42426 + }, + { + "epoch": 2.73, + "learning_rate": 1.0917151506043497e-05, + "loss": 0.9647, + "step": 42427 + }, + { + "epoch": 2.73, + "learning_rate": 1.0916936817020546e-05, + "loss": 0.8774, + "step": 42428 + }, + { + "epoch": 2.73, + "learning_rate": 1.0916722127997597e-05, + "loss": 0.8683, + "step": 42429 + }, + { + "epoch": 2.73, + "learning_rate": 1.0916507438974646e-05, + "loss": 0.9416, + "step": 42430 + }, + { + "epoch": 2.73, + "learning_rate": 1.0916292749951696e-05, + "loss": 0.963, + "step": 42431 + }, + { + "epoch": 2.73, + "learning_rate": 1.0916078060928745e-05, + "loss": 0.8462, + "step": 42432 + }, + { + "epoch": 2.73, + "learning_rate": 1.0915863371905794e-05, + "loss": 0.8347, + "step": 42433 + }, + { + "epoch": 2.73, + "learning_rate": 1.0915648682882845e-05, + "loss": 0.9145, + "step": 42434 + }, + { + "epoch": 2.73, + "learning_rate": 1.0915433993859894e-05, + "loss": 0.8389, + "step": 42435 + }, + { + "epoch": 2.73, + "learning_rate": 1.0915219304836943e-05, + "loss": 0.8543, + "step": 42436 + }, + { + "epoch": 2.73, + "learning_rate": 1.0915004615813994e-05, + "loss": 0.8436, + "step": 42437 + }, + { + "epoch": 2.73, + "learning_rate": 1.0914789926791046e-05, + "loss": 0.8896, + "step": 42438 + }, + { + "epoch": 2.73, + "learning_rate": 1.0914575237768095e-05, + "loss": 0.9529, + "step": 42439 + }, + { + "epoch": 2.73, + "learning_rate": 1.0914360548745144e-05, + "loss": 0.958, + "step": 42440 + }, + { + "epoch": 2.73, + "learning_rate": 1.0914145859722195e-05, + "loss": 0.931, + "step": 42441 + }, + { + "epoch": 2.73, + "learning_rate": 1.0913931170699243e-05, + "loss": 0.9147, + "step": 42442 + }, + { + "epoch": 2.73, + "learning_rate": 1.0913716481676292e-05, + "loss": 0.8687, + "step": 42443 + }, + { + "epoch": 2.73, + "learning_rate": 1.0913501792653343e-05, + "loss": 0.8931, + "step": 42444 + }, + { + "epoch": 2.73, + "learning_rate": 1.0913287103630392e-05, + "loss": 0.8998, + "step": 42445 + }, + { + "epoch": 2.73, + "learning_rate": 1.0913072414607441e-05, + "loss": 0.9303, + "step": 42446 + }, + { + "epoch": 2.73, + "learning_rate": 1.0912857725584492e-05, + "loss": 0.9201, + "step": 42447 + }, + { + "epoch": 2.73, + "learning_rate": 1.091264303656154e-05, + "loss": 0.8715, + "step": 42448 + }, + { + "epoch": 2.73, + "learning_rate": 1.0912428347538591e-05, + "loss": 0.9745, + "step": 42449 + }, + { + "epoch": 2.73, + "learning_rate": 1.091221365851564e-05, + "loss": 0.9292, + "step": 42450 + }, + { + "epoch": 2.73, + "learning_rate": 1.091199896949269e-05, + "loss": 0.9725, + "step": 42451 + }, + { + "epoch": 2.73, + "learning_rate": 1.091178428046974e-05, + "loss": 0.8951, + "step": 42452 + }, + { + "epoch": 2.73, + "learning_rate": 1.091156959144679e-05, + "loss": 0.8714, + "step": 42453 + }, + { + "epoch": 2.73, + "learning_rate": 1.0911354902423841e-05, + "loss": 0.9365, + "step": 42454 + }, + { + "epoch": 2.73, + "learning_rate": 1.091114021340089e-05, + "loss": 0.8839, + "step": 42455 + }, + { + "epoch": 2.73, + "learning_rate": 1.091092552437794e-05, + "loss": 0.8754, + "step": 42456 + }, + { + "epoch": 2.73, + "learning_rate": 1.091071083535499e-05, + "loss": 0.9412, + "step": 42457 + }, + { + "epoch": 2.73, + "learning_rate": 1.0910496146332039e-05, + "loss": 0.8523, + "step": 42458 + }, + { + "epoch": 2.73, + "learning_rate": 1.091028145730909e-05, + "loss": 0.9236, + "step": 42459 + }, + { + "epoch": 2.73, + "learning_rate": 1.0910066768286138e-05, + "loss": 0.9457, + "step": 42460 + }, + { + "epoch": 2.73, + "learning_rate": 1.0909852079263187e-05, + "loss": 0.8686, + "step": 42461 + }, + { + "epoch": 2.73, + "learning_rate": 1.0909637390240238e-05, + "loss": 0.8968, + "step": 42462 + }, + { + "epoch": 2.73, + "learning_rate": 1.0909422701217287e-05, + "loss": 0.9019, + "step": 42463 + }, + { + "epoch": 2.73, + "learning_rate": 1.0909208012194336e-05, + "loss": 0.934, + "step": 42464 + }, + { + "epoch": 2.73, + "learning_rate": 1.0908993323171387e-05, + "loss": 0.8853, + "step": 42465 + }, + { + "epoch": 2.73, + "learning_rate": 1.0908778634148436e-05, + "loss": 0.8924, + "step": 42466 + }, + { + "epoch": 2.73, + "learning_rate": 1.0908563945125488e-05, + "loss": 0.8663, + "step": 42467 + }, + { + "epoch": 2.73, + "learning_rate": 1.0908349256102537e-05, + "loss": 0.8897, + "step": 42468 + }, + { + "epoch": 2.73, + "learning_rate": 1.0908134567079588e-05, + "loss": 0.8719, + "step": 42469 + }, + { + "epoch": 2.73, + "learning_rate": 1.0907919878056637e-05, + "loss": 0.8748, + "step": 42470 + }, + { + "epoch": 2.73, + "learning_rate": 1.0907705189033686e-05, + "loss": 0.9067, + "step": 42471 + }, + { + "epoch": 2.73, + "learning_rate": 1.0907490500010736e-05, + "loss": 0.8588, + "step": 42472 + }, + { + "epoch": 2.73, + "learning_rate": 1.0907275810987785e-05, + "loss": 0.8677, + "step": 42473 + }, + { + "epoch": 2.73, + "learning_rate": 1.0907061121964834e-05, + "loss": 0.9345, + "step": 42474 + }, + { + "epoch": 2.73, + "learning_rate": 1.0906846432941885e-05, + "loss": 0.9104, + "step": 42475 + }, + { + "epoch": 2.73, + "learning_rate": 1.0906631743918934e-05, + "loss": 0.8701, + "step": 42476 + }, + { + "epoch": 2.73, + "learning_rate": 1.0906417054895984e-05, + "loss": 0.8959, + "step": 42477 + }, + { + "epoch": 2.73, + "learning_rate": 1.0906202365873033e-05, + "loss": 0.889, + "step": 42478 + }, + { + "epoch": 2.73, + "learning_rate": 1.0905987676850082e-05, + "loss": 0.8105, + "step": 42479 + }, + { + "epoch": 2.73, + "learning_rate": 1.0905772987827133e-05, + "loss": 0.8833, + "step": 42480 + }, + { + "epoch": 2.73, + "learning_rate": 1.0905558298804182e-05, + "loss": 0.8575, + "step": 42481 + }, + { + "epoch": 2.73, + "learning_rate": 1.0905343609781234e-05, + "loss": 0.9252, + "step": 42482 + }, + { + "epoch": 2.73, + "learning_rate": 1.0905128920758283e-05, + "loss": 0.8033, + "step": 42483 + }, + { + "epoch": 2.73, + "learning_rate": 1.0904914231735332e-05, + "loss": 0.8859, + "step": 42484 + }, + { + "epoch": 2.73, + "learning_rate": 1.0904699542712383e-05, + "loss": 0.9057, + "step": 42485 + }, + { + "epoch": 2.73, + "learning_rate": 1.0904484853689432e-05, + "loss": 0.9387, + "step": 42486 + }, + { + "epoch": 2.73, + "learning_rate": 1.0904270164666482e-05, + "loss": 0.9063, + "step": 42487 + }, + { + "epoch": 2.73, + "learning_rate": 1.0904055475643531e-05, + "loss": 0.8759, + "step": 42488 + }, + { + "epoch": 2.73, + "learning_rate": 1.090384078662058e-05, + "loss": 0.8802, + "step": 42489 + }, + { + "epoch": 2.73, + "learning_rate": 1.0903626097597631e-05, + "loss": 0.8011, + "step": 42490 + }, + { + "epoch": 2.73, + "learning_rate": 1.090341140857468e-05, + "loss": 0.8763, + "step": 42491 + }, + { + "epoch": 2.73, + "learning_rate": 1.0903196719551729e-05, + "loss": 0.8868, + "step": 42492 + }, + { + "epoch": 2.73, + "learning_rate": 1.090298203052878e-05, + "loss": 0.8518, + "step": 42493 + }, + { + "epoch": 2.73, + "learning_rate": 1.0902767341505829e-05, + "loss": 0.9209, + "step": 42494 + }, + { + "epoch": 2.73, + "learning_rate": 1.0902552652482878e-05, + "loss": 0.9126, + "step": 42495 + }, + { + "epoch": 2.73, + "learning_rate": 1.090233796345993e-05, + "loss": 0.8574, + "step": 42496 + }, + { + "epoch": 2.73, + "learning_rate": 1.090212327443698e-05, + "loss": 0.9132, + "step": 42497 + }, + { + "epoch": 2.73, + "learning_rate": 1.090190858541403e-05, + "loss": 0.9381, + "step": 42498 + }, + { + "epoch": 2.73, + "learning_rate": 1.0901693896391079e-05, + "loss": 0.8958, + "step": 42499 + }, + { + "epoch": 2.73, + "learning_rate": 1.090147920736813e-05, + "loss": 0.8734, + "step": 42500 + }, + { + "epoch": 2.73, + "learning_rate": 1.0901264518345178e-05, + "loss": 0.9045, + "step": 42501 + }, + { + "epoch": 2.73, + "learning_rate": 1.0901049829322227e-05, + "loss": 0.9193, + "step": 42502 + }, + { + "epoch": 2.73, + "learning_rate": 1.0900835140299278e-05, + "loss": 0.875, + "step": 42503 + }, + { + "epoch": 2.73, + "learning_rate": 1.0900620451276327e-05, + "loss": 0.9516, + "step": 42504 + }, + { + "epoch": 2.73, + "learning_rate": 1.0900405762253376e-05, + "loss": 0.8691, + "step": 42505 + }, + { + "epoch": 2.73, + "learning_rate": 1.0900191073230426e-05, + "loss": 0.9282, + "step": 42506 + }, + { + "epoch": 2.73, + "learning_rate": 1.0899976384207475e-05, + "loss": 0.9071, + "step": 42507 + }, + { + "epoch": 2.73, + "learning_rate": 1.0899761695184526e-05, + "loss": 0.8296, + "step": 42508 + }, + { + "epoch": 2.73, + "learning_rate": 1.0899547006161575e-05, + "loss": 0.9157, + "step": 42509 + }, + { + "epoch": 2.73, + "learning_rate": 1.0899332317138624e-05, + "loss": 0.911, + "step": 42510 + }, + { + "epoch": 2.74, + "learning_rate": 1.0899117628115676e-05, + "loss": 0.8999, + "step": 42511 + }, + { + "epoch": 2.74, + "learning_rate": 1.0898902939092725e-05, + "loss": 0.9005, + "step": 42512 + }, + { + "epoch": 2.74, + "learning_rate": 1.0898688250069776e-05, + "loss": 0.8428, + "step": 42513 + }, + { + "epoch": 2.74, + "learning_rate": 1.0898473561046825e-05, + "loss": 0.9159, + "step": 42514 + }, + { + "epoch": 2.74, + "learning_rate": 1.0898258872023874e-05, + "loss": 0.8586, + "step": 42515 + }, + { + "epoch": 2.74, + "learning_rate": 1.0898044183000925e-05, + "loss": 0.8868, + "step": 42516 + }, + { + "epoch": 2.74, + "learning_rate": 1.0897829493977973e-05, + "loss": 0.9815, + "step": 42517 + }, + { + "epoch": 2.74, + "learning_rate": 1.0897614804955024e-05, + "loss": 0.9042, + "step": 42518 + }, + { + "epoch": 2.74, + "learning_rate": 1.0897400115932073e-05, + "loss": 0.8715, + "step": 42519 + }, + { + "epoch": 2.74, + "learning_rate": 1.0897185426909122e-05, + "loss": 0.9361, + "step": 42520 + }, + { + "epoch": 2.74, + "learning_rate": 1.0896970737886173e-05, + "loss": 0.8817, + "step": 42521 + }, + { + "epoch": 2.74, + "learning_rate": 1.0896756048863222e-05, + "loss": 0.8673, + "step": 42522 + }, + { + "epoch": 2.74, + "learning_rate": 1.089654135984027e-05, + "loss": 0.9189, + "step": 42523 + }, + { + "epoch": 2.74, + "learning_rate": 1.0896326670817321e-05, + "loss": 0.9914, + "step": 42524 + }, + { + "epoch": 2.74, + "learning_rate": 1.0896111981794372e-05, + "loss": 0.8749, + "step": 42525 + }, + { + "epoch": 2.74, + "learning_rate": 1.0895897292771423e-05, + "loss": 0.9919, + "step": 42526 + }, + { + "epoch": 2.74, + "learning_rate": 1.0895682603748472e-05, + "loss": 0.8719, + "step": 42527 + }, + { + "epoch": 2.74, + "learning_rate": 1.0895467914725522e-05, + "loss": 0.9614, + "step": 42528 + }, + { + "epoch": 2.74, + "learning_rate": 1.0895253225702571e-05, + "loss": 0.9339, + "step": 42529 + }, + { + "epoch": 2.74, + "learning_rate": 1.089503853667962e-05, + "loss": 0.9138, + "step": 42530 + }, + { + "epoch": 2.74, + "learning_rate": 1.089482384765667e-05, + "loss": 0.8873, + "step": 42531 + }, + { + "epoch": 2.74, + "learning_rate": 1.089460915863372e-05, + "loss": 0.8718, + "step": 42532 + }, + { + "epoch": 2.74, + "learning_rate": 1.0894394469610769e-05, + "loss": 0.8719, + "step": 42533 + }, + { + "epoch": 2.74, + "learning_rate": 1.089417978058782e-05, + "loss": 0.8819, + "step": 42534 + }, + { + "epoch": 2.74, + "learning_rate": 1.0893965091564868e-05, + "loss": 0.8967, + "step": 42535 + }, + { + "epoch": 2.74, + "learning_rate": 1.0893750402541919e-05, + "loss": 0.9444, + "step": 42536 + }, + { + "epoch": 2.74, + "learning_rate": 1.0893535713518968e-05, + "loss": 0.8926, + "step": 42537 + }, + { + "epoch": 2.74, + "learning_rate": 1.0893321024496017e-05, + "loss": 0.9435, + "step": 42538 + }, + { + "epoch": 2.74, + "learning_rate": 1.0893106335473068e-05, + "loss": 0.8814, + "step": 42539 + }, + { + "epoch": 2.74, + "learning_rate": 1.0892891646450118e-05, + "loss": 0.8971, + "step": 42540 + }, + { + "epoch": 2.74, + "learning_rate": 1.0892676957427169e-05, + "loss": 0.8358, + "step": 42541 + }, + { + "epoch": 2.74, + "learning_rate": 1.0892462268404218e-05, + "loss": 0.892, + "step": 42542 + }, + { + "epoch": 2.74, + "learning_rate": 1.0892247579381267e-05, + "loss": 0.8631, + "step": 42543 + }, + { + "epoch": 2.74, + "learning_rate": 1.0892032890358318e-05, + "loss": 0.925, + "step": 42544 + }, + { + "epoch": 2.74, + "learning_rate": 1.0891818201335367e-05, + "loss": 0.8039, + "step": 42545 + }, + { + "epoch": 2.74, + "learning_rate": 1.0891603512312417e-05, + "loss": 0.82, + "step": 42546 + }, + { + "epoch": 2.74, + "learning_rate": 1.0891388823289466e-05, + "loss": 0.9896, + "step": 42547 + }, + { + "epoch": 2.74, + "learning_rate": 1.0891174134266515e-05, + "loss": 1.0201, + "step": 42548 + }, + { + "epoch": 2.74, + "learning_rate": 1.0890959445243566e-05, + "loss": 0.9245, + "step": 42549 + }, + { + "epoch": 2.74, + "learning_rate": 1.0890744756220615e-05, + "loss": 0.9452, + "step": 42550 + }, + { + "epoch": 2.74, + "learning_rate": 1.0890530067197664e-05, + "loss": 0.9127, + "step": 42551 + }, + { + "epoch": 2.74, + "learning_rate": 1.0890315378174714e-05, + "loss": 0.8865, + "step": 42552 + }, + { + "epoch": 2.74, + "learning_rate": 1.0890100689151763e-05, + "loss": 0.8964, + "step": 42553 + }, + { + "epoch": 2.74, + "learning_rate": 1.0889886000128816e-05, + "loss": 0.9219, + "step": 42554 + }, + { + "epoch": 2.74, + "learning_rate": 1.0889671311105865e-05, + "loss": 0.9264, + "step": 42555 + }, + { + "epoch": 2.74, + "learning_rate": 1.0889456622082915e-05, + "loss": 0.8435, + "step": 42556 + }, + { + "epoch": 2.74, + "learning_rate": 1.0889241933059964e-05, + "loss": 0.8868, + "step": 42557 + }, + { + "epoch": 2.74, + "learning_rate": 1.0889027244037013e-05, + "loss": 0.8316, + "step": 42558 + }, + { + "epoch": 2.74, + "learning_rate": 1.0888812555014064e-05, + "loss": 0.8898, + "step": 42559 + }, + { + "epoch": 2.74, + "learning_rate": 1.0888597865991113e-05, + "loss": 0.8653, + "step": 42560 + }, + { + "epoch": 2.74, + "learning_rate": 1.0888383176968162e-05, + "loss": 0.8625, + "step": 42561 + }, + { + "epoch": 2.74, + "learning_rate": 1.0888168487945212e-05, + "loss": 0.8939, + "step": 42562 + }, + { + "epoch": 2.74, + "learning_rate": 1.0887953798922261e-05, + "loss": 0.888, + "step": 42563 + }, + { + "epoch": 2.74, + "learning_rate": 1.0887739109899312e-05, + "loss": 0.8017, + "step": 42564 + }, + { + "epoch": 2.74, + "learning_rate": 1.0887524420876361e-05, + "loss": 1.0009, + "step": 42565 + }, + { + "epoch": 2.74, + "learning_rate": 1.088730973185341e-05, + "loss": 0.893, + "step": 42566 + }, + { + "epoch": 2.74, + "learning_rate": 1.088709504283046e-05, + "loss": 0.8824, + "step": 42567 + }, + { + "epoch": 2.74, + "learning_rate": 1.088688035380751e-05, + "loss": 0.8919, + "step": 42568 + }, + { + "epoch": 2.74, + "learning_rate": 1.0886665664784562e-05, + "loss": 1.0133, + "step": 42569 + }, + { + "epoch": 2.74, + "learning_rate": 1.0886450975761611e-05, + "loss": 0.9192, + "step": 42570 + }, + { + "epoch": 2.74, + "learning_rate": 1.088623628673866e-05, + "loss": 0.9432, + "step": 42571 + }, + { + "epoch": 2.74, + "learning_rate": 1.088602159771571e-05, + "loss": 0.9423, + "step": 42572 + }, + { + "epoch": 2.74, + "learning_rate": 1.088602159771571e-05, + "loss": 0.864, + "step": 42573 + }, + { + "epoch": 2.74, + "learning_rate": 1.088580690869276e-05, + "loss": 0.8644, + "step": 42574 + }, + { + "epoch": 2.74, + "learning_rate": 1.088559221966981e-05, + "loss": 0.9133, + "step": 42575 + }, + { + "epoch": 2.74, + "learning_rate": 1.088537753064686e-05, + "loss": 0.9034, + "step": 42576 + }, + { + "epoch": 2.74, + "learning_rate": 1.0885162841623908e-05, + "loss": 0.9563, + "step": 42577 + }, + { + "epoch": 2.74, + "learning_rate": 1.0884948152600959e-05, + "loss": 0.8399, + "step": 42578 + }, + { + "epoch": 2.74, + "learning_rate": 1.0884733463578008e-05, + "loss": 0.8834, + "step": 42579 + }, + { + "epoch": 2.74, + "learning_rate": 1.0884518774555057e-05, + "loss": 1.0314, + "step": 42580 + }, + { + "epoch": 2.74, + "learning_rate": 1.0884304085532107e-05, + "loss": 0.9067, + "step": 42581 + }, + { + "epoch": 2.74, + "learning_rate": 1.0884089396509156e-05, + "loss": 0.9094, + "step": 42582 + }, + { + "epoch": 2.74, + "learning_rate": 1.0883874707486205e-05, + "loss": 0.9661, + "step": 42583 + }, + { + "epoch": 2.74, + "learning_rate": 1.0883660018463258e-05, + "loss": 0.9135, + "step": 42584 + }, + { + "epoch": 2.74, + "learning_rate": 1.0883445329440308e-05, + "loss": 0.8865, + "step": 42585 + }, + { + "epoch": 2.74, + "learning_rate": 1.0883230640417357e-05, + "loss": 0.8584, + "step": 42586 + }, + { + "epoch": 2.74, + "learning_rate": 1.0883015951394406e-05, + "loss": 0.8844, + "step": 42587 + }, + { + "epoch": 2.74, + "learning_rate": 1.0882801262371457e-05, + "loss": 0.8791, + "step": 42588 + }, + { + "epoch": 2.74, + "learning_rate": 1.0882586573348506e-05, + "loss": 0.874, + "step": 42589 + }, + { + "epoch": 2.74, + "learning_rate": 1.0882371884325555e-05, + "loss": 0.9183, + "step": 42590 + }, + { + "epoch": 2.74, + "learning_rate": 1.0882157195302606e-05, + "loss": 0.8525, + "step": 42591 + }, + { + "epoch": 2.74, + "learning_rate": 1.0881942506279654e-05, + "loss": 0.8542, + "step": 42592 + }, + { + "epoch": 2.74, + "learning_rate": 1.0881727817256703e-05, + "loss": 0.8577, + "step": 42593 + }, + { + "epoch": 2.74, + "learning_rate": 1.0881513128233754e-05, + "loss": 0.9016, + "step": 42594 + }, + { + "epoch": 2.74, + "learning_rate": 1.0881298439210803e-05, + "loss": 0.9178, + "step": 42595 + }, + { + "epoch": 2.74, + "learning_rate": 1.0881083750187854e-05, + "loss": 0.9598, + "step": 42596 + }, + { + "epoch": 2.74, + "learning_rate": 1.0880869061164903e-05, + "loss": 0.8672, + "step": 42597 + }, + { + "epoch": 2.74, + "learning_rate": 1.0880654372141952e-05, + "loss": 0.9144, + "step": 42598 + }, + { + "epoch": 2.74, + "learning_rate": 1.0880439683119004e-05, + "loss": 0.9077, + "step": 42599 + }, + { + "epoch": 2.74, + "learning_rate": 1.0880224994096053e-05, + "loss": 0.8817, + "step": 42600 + }, + { + "epoch": 2.74, + "learning_rate": 1.0880010305073104e-05, + "loss": 1.042, + "step": 42601 + }, + { + "epoch": 2.74, + "learning_rate": 1.0879795616050153e-05, + "loss": 0.8906, + "step": 42602 + }, + { + "epoch": 2.74, + "learning_rate": 1.0879580927027202e-05, + "loss": 0.9087, + "step": 42603 + }, + { + "epoch": 2.74, + "learning_rate": 1.0879366238004252e-05, + "loss": 0.8985, + "step": 42604 + }, + { + "epoch": 2.74, + "learning_rate": 1.0879151548981301e-05, + "loss": 0.8643, + "step": 42605 + }, + { + "epoch": 2.74, + "learning_rate": 1.0878936859958352e-05, + "loss": 0.9071, + "step": 42606 + }, + { + "epoch": 2.74, + "learning_rate": 1.08787221709354e-05, + "loss": 0.9228, + "step": 42607 + }, + { + "epoch": 2.74, + "learning_rate": 1.087850748191245e-05, + "loss": 0.8677, + "step": 42608 + }, + { + "epoch": 2.74, + "learning_rate": 1.08782927928895e-05, + "loss": 0.9455, + "step": 42609 + }, + { + "epoch": 2.74, + "learning_rate": 1.087807810386655e-05, + "loss": 0.941, + "step": 42610 + }, + { + "epoch": 2.74, + "learning_rate": 1.0877863414843598e-05, + "loss": 0.8428, + "step": 42611 + }, + { + "epoch": 2.74, + "learning_rate": 1.0877648725820649e-05, + "loss": 0.9129, + "step": 42612 + }, + { + "epoch": 2.74, + "learning_rate": 1.08774340367977e-05, + "loss": 0.8653, + "step": 42613 + }, + { + "epoch": 2.74, + "learning_rate": 1.087721934777475e-05, + "loss": 0.9538, + "step": 42614 + }, + { + "epoch": 2.74, + "learning_rate": 1.08770046587518e-05, + "loss": 0.9044, + "step": 42615 + }, + { + "epoch": 2.74, + "learning_rate": 1.087678996972885e-05, + "loss": 0.9319, + "step": 42616 + }, + { + "epoch": 2.74, + "learning_rate": 1.0876575280705899e-05, + "loss": 0.8909, + "step": 42617 + }, + { + "epoch": 2.74, + "learning_rate": 1.0876360591682948e-05, + "loss": 0.9043, + "step": 42618 + }, + { + "epoch": 2.74, + "learning_rate": 1.0876145902659999e-05, + "loss": 0.8978, + "step": 42619 + }, + { + "epoch": 2.74, + "learning_rate": 1.0875931213637048e-05, + "loss": 0.8965, + "step": 42620 + }, + { + "epoch": 2.74, + "learning_rate": 1.0875716524614096e-05, + "loss": 0.8235, + "step": 42621 + }, + { + "epoch": 2.74, + "learning_rate": 1.0875501835591147e-05, + "loss": 0.8716, + "step": 42622 + }, + { + "epoch": 2.74, + "learning_rate": 1.0875287146568196e-05, + "loss": 0.9488, + "step": 42623 + }, + { + "epoch": 2.74, + "learning_rate": 1.0875072457545247e-05, + "loss": 0.9402, + "step": 42624 + }, + { + "epoch": 2.74, + "learning_rate": 1.0874857768522296e-05, + "loss": 0.8999, + "step": 42625 + }, + { + "epoch": 2.74, + "learning_rate": 1.0874643079499345e-05, + "loss": 0.9422, + "step": 42626 + }, + { + "epoch": 2.74, + "learning_rate": 1.0874428390476395e-05, + "loss": 0.9296, + "step": 42627 + }, + { + "epoch": 2.74, + "learning_rate": 1.0874213701453446e-05, + "loss": 0.9662, + "step": 42628 + }, + { + "epoch": 2.74, + "learning_rate": 1.0873999012430497e-05, + "loss": 0.8763, + "step": 42629 + }, + { + "epoch": 2.74, + "learning_rate": 1.0873784323407546e-05, + "loss": 0.9346, + "step": 42630 + }, + { + "epoch": 2.74, + "learning_rate": 1.0873569634384595e-05, + "loss": 0.8944, + "step": 42631 + }, + { + "epoch": 2.74, + "learning_rate": 1.0873354945361645e-05, + "loss": 0.9521, + "step": 42632 + }, + { + "epoch": 2.74, + "learning_rate": 1.0873140256338694e-05, + "loss": 0.9527, + "step": 42633 + }, + { + "epoch": 2.74, + "learning_rate": 1.0872925567315745e-05, + "loss": 0.8916, + "step": 42634 + }, + { + "epoch": 2.74, + "learning_rate": 1.0872710878292794e-05, + "loss": 0.9329, + "step": 42635 + }, + { + "epoch": 2.74, + "learning_rate": 1.0872496189269843e-05, + "loss": 0.9019, + "step": 42636 + }, + { + "epoch": 2.74, + "learning_rate": 1.0872281500246893e-05, + "loss": 0.942, + "step": 42637 + }, + { + "epoch": 2.74, + "learning_rate": 1.0872066811223942e-05, + "loss": 0.8584, + "step": 42638 + }, + { + "epoch": 2.74, + "learning_rate": 1.0871852122200991e-05, + "loss": 0.8447, + "step": 42639 + }, + { + "epoch": 2.74, + "learning_rate": 1.0871637433178042e-05, + "loss": 0.8989, + "step": 42640 + }, + { + "epoch": 2.74, + "learning_rate": 1.0871422744155091e-05, + "loss": 0.8297, + "step": 42641 + }, + { + "epoch": 2.74, + "learning_rate": 1.0871208055132143e-05, + "loss": 0.8854, + "step": 42642 + }, + { + "epoch": 2.74, + "learning_rate": 1.0870993366109192e-05, + "loss": 0.9091, + "step": 42643 + }, + { + "epoch": 2.74, + "learning_rate": 1.0870778677086243e-05, + "loss": 0.9417, + "step": 42644 + }, + { + "epoch": 2.74, + "learning_rate": 1.0870563988063292e-05, + "loss": 0.8673, + "step": 42645 + }, + { + "epoch": 2.74, + "learning_rate": 1.0870349299040341e-05, + "loss": 0.9649, + "step": 42646 + }, + { + "epoch": 2.74, + "learning_rate": 1.0870134610017392e-05, + "loss": 0.8626, + "step": 42647 + }, + { + "epoch": 2.74, + "learning_rate": 1.086991992099444e-05, + "loss": 0.8989, + "step": 42648 + }, + { + "epoch": 2.74, + "learning_rate": 1.086970523197149e-05, + "loss": 0.9432, + "step": 42649 + }, + { + "epoch": 2.74, + "learning_rate": 1.086949054294854e-05, + "loss": 0.8999, + "step": 42650 + }, + { + "epoch": 2.74, + "learning_rate": 1.0869275853925589e-05, + "loss": 0.9685, + "step": 42651 + }, + { + "epoch": 2.74, + "learning_rate": 1.086906116490264e-05, + "loss": 0.9305, + "step": 42652 + }, + { + "epoch": 2.74, + "learning_rate": 1.0868846475879689e-05, + "loss": 0.9006, + "step": 42653 + }, + { + "epoch": 2.74, + "learning_rate": 1.0868631786856738e-05, + "loss": 0.8532, + "step": 42654 + }, + { + "epoch": 2.74, + "learning_rate": 1.0868417097833788e-05, + "loss": 0.8425, + "step": 42655 + }, + { + "epoch": 2.74, + "learning_rate": 1.0868202408810837e-05, + "loss": 0.8971, + "step": 42656 + }, + { + "epoch": 2.74, + "learning_rate": 1.086798771978789e-05, + "loss": 0.9133, + "step": 42657 + }, + { + "epoch": 2.74, + "learning_rate": 1.0867773030764939e-05, + "loss": 0.8942, + "step": 42658 + }, + { + "epoch": 2.74, + "learning_rate": 1.0867558341741988e-05, + "loss": 0.8723, + "step": 42659 + }, + { + "epoch": 2.74, + "learning_rate": 1.0867343652719038e-05, + "loss": 0.8894, + "step": 42660 + }, + { + "epoch": 2.74, + "learning_rate": 1.0867128963696087e-05, + "loss": 0.9037, + "step": 42661 + }, + { + "epoch": 2.74, + "learning_rate": 1.0866914274673138e-05, + "loss": 0.8326, + "step": 42662 + }, + { + "epoch": 2.74, + "learning_rate": 1.0866699585650187e-05, + "loss": 0.9332, + "step": 42663 + }, + { + "epoch": 2.74, + "learning_rate": 1.0866484896627236e-05, + "loss": 0.8699, + "step": 42664 + }, + { + "epoch": 2.74, + "learning_rate": 1.0866270207604287e-05, + "loss": 0.9135, + "step": 42665 + }, + { + "epoch": 2.75, + "learning_rate": 1.0866055518581335e-05, + "loss": 0.9013, + "step": 42666 + }, + { + "epoch": 2.75, + "learning_rate": 1.0865840829558384e-05, + "loss": 0.9552, + "step": 42667 + }, + { + "epoch": 2.75, + "learning_rate": 1.0865626140535435e-05, + "loss": 0.8709, + "step": 42668 + }, + { + "epoch": 2.75, + "learning_rate": 1.0865411451512484e-05, + "loss": 0.912, + "step": 42669 + }, + { + "epoch": 2.75, + "learning_rate": 1.0865196762489533e-05, + "loss": 0.9317, + "step": 42670 + }, + { + "epoch": 2.75, + "learning_rate": 1.0864982073466585e-05, + "loss": 0.933, + "step": 42671 + }, + { + "epoch": 2.75, + "learning_rate": 1.0864767384443636e-05, + "loss": 0.8996, + "step": 42672 + }, + { + "epoch": 2.75, + "learning_rate": 1.0864552695420685e-05, + "loss": 0.8627, + "step": 42673 + }, + { + "epoch": 2.75, + "learning_rate": 1.0864338006397734e-05, + "loss": 0.9051, + "step": 42674 + }, + { + "epoch": 2.75, + "learning_rate": 1.0864123317374785e-05, + "loss": 0.937, + "step": 42675 + }, + { + "epoch": 2.75, + "learning_rate": 1.0863908628351834e-05, + "loss": 0.9576, + "step": 42676 + }, + { + "epoch": 2.75, + "learning_rate": 1.0863693939328883e-05, + "loss": 0.8645, + "step": 42677 + }, + { + "epoch": 2.75, + "learning_rate": 1.0863479250305933e-05, + "loss": 0.9137, + "step": 42678 + }, + { + "epoch": 2.75, + "learning_rate": 1.0863264561282982e-05, + "loss": 0.8424, + "step": 42679 + }, + { + "epoch": 2.75, + "learning_rate": 1.0863049872260031e-05, + "loss": 0.893, + "step": 42680 + }, + { + "epoch": 2.75, + "learning_rate": 1.0862835183237082e-05, + "loss": 0.9042, + "step": 42681 + }, + { + "epoch": 2.75, + "learning_rate": 1.086262049421413e-05, + "loss": 0.846, + "step": 42682 + }, + { + "epoch": 2.75, + "learning_rate": 1.0862405805191181e-05, + "loss": 0.905, + "step": 42683 + }, + { + "epoch": 2.75, + "learning_rate": 1.086219111616823e-05, + "loss": 0.8952, + "step": 42684 + }, + { + "epoch": 2.75, + "learning_rate": 1.086197642714528e-05, + "loss": 0.9154, + "step": 42685 + }, + { + "epoch": 2.75, + "learning_rate": 1.0861761738122332e-05, + "loss": 0.8769, + "step": 42686 + }, + { + "epoch": 2.75, + "learning_rate": 1.086154704909938e-05, + "loss": 0.9277, + "step": 42687 + }, + { + "epoch": 2.75, + "learning_rate": 1.0861332360076431e-05, + "loss": 0.9397, + "step": 42688 + }, + { + "epoch": 2.75, + "learning_rate": 1.086111767105348e-05, + "loss": 0.846, + "step": 42689 + }, + { + "epoch": 2.75, + "learning_rate": 1.086090298203053e-05, + "loss": 0.9172, + "step": 42690 + }, + { + "epoch": 2.75, + "learning_rate": 1.086068829300758e-05, + "loss": 0.9102, + "step": 42691 + }, + { + "epoch": 2.75, + "learning_rate": 1.0860473603984629e-05, + "loss": 0.9417, + "step": 42692 + }, + { + "epoch": 2.75, + "learning_rate": 1.086025891496168e-05, + "loss": 0.869, + "step": 42693 + }, + { + "epoch": 2.75, + "learning_rate": 1.0860044225938729e-05, + "loss": 0.8906, + "step": 42694 + }, + { + "epoch": 2.75, + "learning_rate": 1.0859829536915777e-05, + "loss": 0.9438, + "step": 42695 + }, + { + "epoch": 2.75, + "learning_rate": 1.0859614847892828e-05, + "loss": 0.9111, + "step": 42696 + }, + { + "epoch": 2.75, + "learning_rate": 1.0859400158869877e-05, + "loss": 0.8497, + "step": 42697 + }, + { + "epoch": 2.75, + "learning_rate": 1.0859185469846926e-05, + "loss": 0.8674, + "step": 42698 + }, + { + "epoch": 2.75, + "learning_rate": 1.0858970780823977e-05, + "loss": 0.892, + "step": 42699 + }, + { + "epoch": 2.75, + "learning_rate": 1.0858756091801027e-05, + "loss": 0.8546, + "step": 42700 + }, + { + "epoch": 2.75, + "learning_rate": 1.0858541402778078e-05, + "loss": 0.9496, + "step": 42701 + }, + { + "epoch": 2.75, + "learning_rate": 1.0858326713755127e-05, + "loss": 0.8794, + "step": 42702 + }, + { + "epoch": 2.75, + "learning_rate": 1.0858112024732178e-05, + "loss": 0.9044, + "step": 42703 + }, + { + "epoch": 2.75, + "learning_rate": 1.0857897335709227e-05, + "loss": 0.8718, + "step": 42704 + }, + { + "epoch": 2.75, + "learning_rate": 1.0857682646686276e-05, + "loss": 0.8969, + "step": 42705 + }, + { + "epoch": 2.75, + "learning_rate": 1.0857467957663326e-05, + "loss": 0.8912, + "step": 42706 + }, + { + "epoch": 2.75, + "learning_rate": 1.0857253268640375e-05, + "loss": 0.9158, + "step": 42707 + }, + { + "epoch": 2.75, + "learning_rate": 1.0857038579617424e-05, + "loss": 0.8619, + "step": 42708 + }, + { + "epoch": 2.75, + "learning_rate": 1.0856823890594475e-05, + "loss": 0.8459, + "step": 42709 + }, + { + "epoch": 2.75, + "learning_rate": 1.0856609201571524e-05, + "loss": 0.9457, + "step": 42710 + }, + { + "epoch": 2.75, + "learning_rate": 1.0856394512548574e-05, + "loss": 0.8439, + "step": 42711 + }, + { + "epoch": 2.75, + "learning_rate": 1.0856179823525623e-05, + "loss": 0.9102, + "step": 42712 + }, + { + "epoch": 2.75, + "learning_rate": 1.0855965134502672e-05, + "loss": 0.9279, + "step": 42713 + }, + { + "epoch": 2.75, + "learning_rate": 1.0855750445479723e-05, + "loss": 0.9037, + "step": 42714 + }, + { + "epoch": 2.75, + "learning_rate": 1.0855535756456774e-05, + "loss": 0.9144, + "step": 42715 + }, + { + "epoch": 2.75, + "learning_rate": 1.0855321067433824e-05, + "loss": 0.9037, + "step": 42716 + }, + { + "epoch": 2.75, + "learning_rate": 1.0855106378410873e-05, + "loss": 0.9201, + "step": 42717 + }, + { + "epoch": 2.75, + "learning_rate": 1.0854891689387922e-05, + "loss": 0.9142, + "step": 42718 + }, + { + "epoch": 2.75, + "learning_rate": 1.0854677000364973e-05, + "loss": 0.8903, + "step": 42719 + }, + { + "epoch": 2.75, + "learning_rate": 1.0854462311342022e-05, + "loss": 0.8835, + "step": 42720 + }, + { + "epoch": 2.75, + "learning_rate": 1.0854247622319073e-05, + "loss": 0.8561, + "step": 42721 + }, + { + "epoch": 2.75, + "learning_rate": 1.0854032933296122e-05, + "loss": 0.8501, + "step": 42722 + }, + { + "epoch": 2.75, + "learning_rate": 1.085381824427317e-05, + "loss": 0.8946, + "step": 42723 + }, + { + "epoch": 2.75, + "learning_rate": 1.0853603555250221e-05, + "loss": 0.9509, + "step": 42724 + }, + { + "epoch": 2.75, + "learning_rate": 1.085338886622727e-05, + "loss": 0.9165, + "step": 42725 + }, + { + "epoch": 2.75, + "learning_rate": 1.0853174177204319e-05, + "loss": 0.8866, + "step": 42726 + }, + { + "epoch": 2.75, + "learning_rate": 1.085295948818137e-05, + "loss": 0.9233, + "step": 42727 + }, + { + "epoch": 2.75, + "learning_rate": 1.0852744799158419e-05, + "loss": 0.8307, + "step": 42728 + }, + { + "epoch": 2.75, + "learning_rate": 1.0852530110135471e-05, + "loss": 0.8936, + "step": 42729 + }, + { + "epoch": 2.75, + "learning_rate": 1.085231542111252e-05, + "loss": 0.9317, + "step": 42730 + }, + { + "epoch": 2.75, + "learning_rate": 1.085210073208957e-05, + "loss": 0.9017, + "step": 42731 + }, + { + "epoch": 2.75, + "learning_rate": 1.085188604306662e-05, + "loss": 0.9102, + "step": 42732 + }, + { + "epoch": 2.75, + "learning_rate": 1.0851671354043669e-05, + "loss": 0.862, + "step": 42733 + }, + { + "epoch": 2.75, + "learning_rate": 1.085145666502072e-05, + "loss": 0.9106, + "step": 42734 + }, + { + "epoch": 2.75, + "learning_rate": 1.0851241975997768e-05, + "loss": 0.8841, + "step": 42735 + }, + { + "epoch": 2.75, + "learning_rate": 1.0851027286974817e-05, + "loss": 0.8041, + "step": 42736 + }, + { + "epoch": 2.75, + "learning_rate": 1.0850812597951868e-05, + "loss": 0.8657, + "step": 42737 + }, + { + "epoch": 2.75, + "learning_rate": 1.0850597908928917e-05, + "loss": 0.9256, + "step": 42738 + }, + { + "epoch": 2.75, + "learning_rate": 1.0850383219905966e-05, + "loss": 0.9465, + "step": 42739 + }, + { + "epoch": 2.75, + "learning_rate": 1.0850168530883016e-05, + "loss": 0.8513, + "step": 42740 + }, + { + "epoch": 2.75, + "learning_rate": 1.0849953841860065e-05, + "loss": 0.9056, + "step": 42741 + }, + { + "epoch": 2.75, + "learning_rate": 1.0849739152837116e-05, + "loss": 0.8959, + "step": 42742 + }, + { + "epoch": 2.75, + "learning_rate": 1.0849524463814165e-05, + "loss": 0.9178, + "step": 42743 + }, + { + "epoch": 2.75, + "learning_rate": 1.0849309774791217e-05, + "loss": 0.8959, + "step": 42744 + }, + { + "epoch": 2.75, + "learning_rate": 1.0849095085768266e-05, + "loss": 0.8953, + "step": 42745 + }, + { + "epoch": 2.75, + "learning_rate": 1.0848880396745315e-05, + "loss": 0.9165, + "step": 42746 + }, + { + "epoch": 2.75, + "learning_rate": 1.0848665707722366e-05, + "loss": 0.8787, + "step": 42747 + }, + { + "epoch": 2.75, + "learning_rate": 1.0848451018699415e-05, + "loss": 0.8652, + "step": 42748 + }, + { + "epoch": 2.75, + "learning_rate": 1.0848236329676464e-05, + "loss": 1.0023, + "step": 42749 + }, + { + "epoch": 2.75, + "learning_rate": 1.0848021640653515e-05, + "loss": 0.9096, + "step": 42750 + }, + { + "epoch": 2.75, + "learning_rate": 1.0847806951630564e-05, + "loss": 0.9613, + "step": 42751 + }, + { + "epoch": 2.75, + "learning_rate": 1.0847592262607614e-05, + "loss": 0.8442, + "step": 42752 + }, + { + "epoch": 2.75, + "learning_rate": 1.0847377573584663e-05, + "loss": 0.8818, + "step": 42753 + }, + { + "epoch": 2.75, + "learning_rate": 1.0847162884561712e-05, + "loss": 0.8821, + "step": 42754 + }, + { + "epoch": 2.75, + "learning_rate": 1.0846948195538763e-05, + "loss": 0.9, + "step": 42755 + }, + { + "epoch": 2.75, + "learning_rate": 1.0846733506515812e-05, + "loss": 0.9353, + "step": 42756 + }, + { + "epoch": 2.75, + "learning_rate": 1.084651881749286e-05, + "loss": 0.95, + "step": 42757 + }, + { + "epoch": 2.75, + "learning_rate": 1.0846304128469913e-05, + "loss": 0.8677, + "step": 42758 + }, + { + "epoch": 2.75, + "learning_rate": 1.0846089439446964e-05, + "loss": 0.9007, + "step": 42759 + }, + { + "epoch": 2.75, + "learning_rate": 1.0845874750424013e-05, + "loss": 0.936, + "step": 42760 + }, + { + "epoch": 2.75, + "learning_rate": 1.0845660061401062e-05, + "loss": 0.8927, + "step": 42761 + }, + { + "epoch": 2.75, + "learning_rate": 1.0845445372378112e-05, + "loss": 0.8947, + "step": 42762 + }, + { + "epoch": 2.75, + "learning_rate": 1.0845230683355161e-05, + "loss": 0.954, + "step": 42763 + }, + { + "epoch": 2.75, + "learning_rate": 1.084501599433221e-05, + "loss": 0.9647, + "step": 42764 + }, + { + "epoch": 2.75, + "learning_rate": 1.0844801305309261e-05, + "loss": 0.9555, + "step": 42765 + }, + { + "epoch": 2.75, + "learning_rate": 1.084458661628631e-05, + "loss": 0.901, + "step": 42766 + }, + { + "epoch": 2.75, + "learning_rate": 1.0844371927263359e-05, + "loss": 0.8761, + "step": 42767 + }, + { + "epoch": 2.75, + "learning_rate": 1.084415723824041e-05, + "loss": 0.9491, + "step": 42768 + }, + { + "epoch": 2.75, + "learning_rate": 1.0843942549217458e-05, + "loss": 0.8956, + "step": 42769 + }, + { + "epoch": 2.75, + "learning_rate": 1.0843727860194509e-05, + "loss": 0.9034, + "step": 42770 + }, + { + "epoch": 2.75, + "learning_rate": 1.0843513171171558e-05, + "loss": 0.8901, + "step": 42771 + }, + { + "epoch": 2.75, + "learning_rate": 1.0843298482148607e-05, + "loss": 0.9091, + "step": 42772 + }, + { + "epoch": 2.75, + "learning_rate": 1.084308379312566e-05, + "loss": 0.8681, + "step": 42773 + }, + { + "epoch": 2.75, + "learning_rate": 1.0842869104102708e-05, + "loss": 0.8901, + "step": 42774 + }, + { + "epoch": 2.75, + "learning_rate": 1.0842654415079759e-05, + "loss": 0.9226, + "step": 42775 + }, + { + "epoch": 2.75, + "learning_rate": 1.0842439726056808e-05, + "loss": 0.9074, + "step": 42776 + }, + { + "epoch": 2.75, + "learning_rate": 1.0842225037033857e-05, + "loss": 0.8857, + "step": 42777 + }, + { + "epoch": 2.75, + "learning_rate": 1.0842010348010908e-05, + "loss": 0.8823, + "step": 42778 + }, + { + "epoch": 2.75, + "learning_rate": 1.0841795658987957e-05, + "loss": 0.8311, + "step": 42779 + }, + { + "epoch": 2.75, + "learning_rate": 1.0841580969965007e-05, + "loss": 0.8619, + "step": 42780 + }, + { + "epoch": 2.75, + "learning_rate": 1.0841366280942056e-05, + "loss": 0.8489, + "step": 42781 + }, + { + "epoch": 2.75, + "learning_rate": 1.0841151591919105e-05, + "loss": 0.8579, + "step": 42782 + }, + { + "epoch": 2.75, + "learning_rate": 1.0840936902896156e-05, + "loss": 0.8998, + "step": 42783 + }, + { + "epoch": 2.75, + "learning_rate": 1.0840722213873205e-05, + "loss": 0.8826, + "step": 42784 + }, + { + "epoch": 2.75, + "learning_rate": 1.0840507524850254e-05, + "loss": 0.919, + "step": 42785 + }, + { + "epoch": 2.75, + "learning_rate": 1.0840292835827304e-05, + "loss": 0.9129, + "step": 42786 + }, + { + "epoch": 2.75, + "learning_rate": 1.0840078146804355e-05, + "loss": 0.8282, + "step": 42787 + }, + { + "epoch": 2.75, + "learning_rate": 1.0839863457781406e-05, + "loss": 0.8871, + "step": 42788 + }, + { + "epoch": 2.75, + "learning_rate": 1.0839648768758455e-05, + "loss": 0.8159, + "step": 42789 + }, + { + "epoch": 2.75, + "learning_rate": 1.0839434079735505e-05, + "loss": 0.8192, + "step": 42790 + }, + { + "epoch": 2.75, + "learning_rate": 1.0839219390712554e-05, + "loss": 0.8376, + "step": 42791 + }, + { + "epoch": 2.75, + "learning_rate": 1.0839004701689603e-05, + "loss": 0.8751, + "step": 42792 + }, + { + "epoch": 2.75, + "learning_rate": 1.0838790012666654e-05, + "loss": 0.883, + "step": 42793 + }, + { + "epoch": 2.75, + "learning_rate": 1.0838575323643703e-05, + "loss": 0.9501, + "step": 42794 + }, + { + "epoch": 2.75, + "learning_rate": 1.0838360634620752e-05, + "loss": 0.8794, + "step": 42795 + }, + { + "epoch": 2.75, + "learning_rate": 1.0838145945597803e-05, + "loss": 0.9737, + "step": 42796 + }, + { + "epoch": 2.75, + "learning_rate": 1.0837931256574852e-05, + "loss": 0.8751, + "step": 42797 + }, + { + "epoch": 2.75, + "learning_rate": 1.0837716567551902e-05, + "loss": 0.9151, + "step": 42798 + }, + { + "epoch": 2.75, + "learning_rate": 1.0837501878528951e-05, + "loss": 0.8755, + "step": 42799 + }, + { + "epoch": 2.75, + "learning_rate": 1.0837287189506e-05, + "loss": 0.9052, + "step": 42800 + }, + { + "epoch": 2.75, + "learning_rate": 1.083707250048305e-05, + "loss": 0.8663, + "step": 42801 + }, + { + "epoch": 2.75, + "learning_rate": 1.0836857811460101e-05, + "loss": 0.8504, + "step": 42802 + }, + { + "epoch": 2.75, + "learning_rate": 1.0836643122437152e-05, + "loss": 0.9342, + "step": 42803 + }, + { + "epoch": 2.75, + "learning_rate": 1.0836428433414201e-05, + "loss": 0.8964, + "step": 42804 + }, + { + "epoch": 2.75, + "learning_rate": 1.083621374439125e-05, + "loss": 0.938, + "step": 42805 + }, + { + "epoch": 2.75, + "learning_rate": 1.08359990553683e-05, + "loss": 0.8813, + "step": 42806 + }, + { + "epoch": 2.75, + "learning_rate": 1.083578436634535e-05, + "loss": 0.9014, + "step": 42807 + }, + { + "epoch": 2.75, + "learning_rate": 1.08355696773224e-05, + "loss": 0.9392, + "step": 42808 + }, + { + "epoch": 2.75, + "learning_rate": 1.083535498829945e-05, + "loss": 0.8877, + "step": 42809 + }, + { + "epoch": 2.75, + "learning_rate": 1.0835140299276498e-05, + "loss": 0.9582, + "step": 42810 + }, + { + "epoch": 2.75, + "learning_rate": 1.0834925610253549e-05, + "loss": 0.9298, + "step": 42811 + }, + { + "epoch": 2.75, + "learning_rate": 1.0834710921230598e-05, + "loss": 0.9474, + "step": 42812 + }, + { + "epoch": 2.75, + "learning_rate": 1.0834496232207647e-05, + "loss": 0.9019, + "step": 42813 + }, + { + "epoch": 2.75, + "learning_rate": 1.0834281543184697e-05, + "loss": 0.8185, + "step": 42814 + }, + { + "epoch": 2.75, + "learning_rate": 1.0834066854161746e-05, + "loss": 0.8861, + "step": 42815 + }, + { + "epoch": 2.75, + "learning_rate": 1.0833852165138799e-05, + "loss": 0.8253, + "step": 42816 + }, + { + "epoch": 2.75, + "learning_rate": 1.0833637476115848e-05, + "loss": 0.9139, + "step": 42817 + }, + { + "epoch": 2.75, + "learning_rate": 1.0833422787092898e-05, + "loss": 0.8389, + "step": 42818 + }, + { + "epoch": 2.75, + "learning_rate": 1.0833208098069947e-05, + "loss": 0.8759, + "step": 42819 + }, + { + "epoch": 2.75, + "learning_rate": 1.0832993409046996e-05, + "loss": 0.8852, + "step": 42820 + }, + { + "epoch": 2.76, + "learning_rate": 1.0832778720024047e-05, + "loss": 0.8653, + "step": 42821 + }, + { + "epoch": 2.76, + "learning_rate": 1.0832564031001096e-05, + "loss": 0.9122, + "step": 42822 + }, + { + "epoch": 2.76, + "learning_rate": 1.0832349341978145e-05, + "loss": 0.8737, + "step": 42823 + }, + { + "epoch": 2.76, + "learning_rate": 1.0832134652955196e-05, + "loss": 0.8802, + "step": 42824 + }, + { + "epoch": 2.76, + "learning_rate": 1.0831919963932245e-05, + "loss": 0.9189, + "step": 42825 + }, + { + "epoch": 2.76, + "learning_rate": 1.0831705274909294e-05, + "loss": 0.8921, + "step": 42826 + }, + { + "epoch": 2.76, + "learning_rate": 1.0831490585886344e-05, + "loss": 0.871, + "step": 42827 + }, + { + "epoch": 2.76, + "learning_rate": 1.0831275896863393e-05, + "loss": 0.9083, + "step": 42828 + }, + { + "epoch": 2.76, + "learning_rate": 1.0831061207840444e-05, + "loss": 0.8767, + "step": 42829 + }, + { + "epoch": 2.76, + "learning_rate": 1.0830846518817493e-05, + "loss": 0.9129, + "step": 42830 + }, + { + "epoch": 2.76, + "learning_rate": 1.0830631829794545e-05, + "loss": 0.886, + "step": 42831 + }, + { + "epoch": 2.76, + "learning_rate": 1.0830417140771594e-05, + "loss": 0.8721, + "step": 42832 + }, + { + "epoch": 2.76, + "learning_rate": 1.0830202451748643e-05, + "loss": 0.8777, + "step": 42833 + }, + { + "epoch": 2.76, + "learning_rate": 1.0829987762725694e-05, + "loss": 0.915, + "step": 42834 + }, + { + "epoch": 2.76, + "learning_rate": 1.0829773073702743e-05, + "loss": 0.882, + "step": 42835 + }, + { + "epoch": 2.76, + "learning_rate": 1.0829558384679792e-05, + "loss": 0.8624, + "step": 42836 + }, + { + "epoch": 2.76, + "learning_rate": 1.0829343695656842e-05, + "loss": 0.8975, + "step": 42837 + }, + { + "epoch": 2.76, + "learning_rate": 1.0829129006633891e-05, + "loss": 0.8969, + "step": 42838 + }, + { + "epoch": 2.76, + "learning_rate": 1.0828914317610942e-05, + "loss": 0.9153, + "step": 42839 + }, + { + "epoch": 2.76, + "learning_rate": 1.0828699628587991e-05, + "loss": 0.934, + "step": 42840 + }, + { + "epoch": 2.76, + "learning_rate": 1.082848493956504e-05, + "loss": 0.9182, + "step": 42841 + }, + { + "epoch": 2.76, + "learning_rate": 1.082827025054209e-05, + "loss": 0.9166, + "step": 42842 + }, + { + "epoch": 2.76, + "learning_rate": 1.082805556151914e-05, + "loss": 0.8323, + "step": 42843 + }, + { + "epoch": 2.76, + "learning_rate": 1.0827840872496188e-05, + "loss": 0.8955, + "step": 42844 + }, + { + "epoch": 2.76, + "learning_rate": 1.0827626183473239e-05, + "loss": 0.8372, + "step": 42845 + }, + { + "epoch": 2.76, + "learning_rate": 1.082741149445029e-05, + "loss": 0.88, + "step": 42846 + }, + { + "epoch": 2.76, + "learning_rate": 1.082719680542734e-05, + "loss": 0.8761, + "step": 42847 + }, + { + "epoch": 2.76, + "learning_rate": 1.082698211640439e-05, + "loss": 0.8564, + "step": 42848 + }, + { + "epoch": 2.76, + "learning_rate": 1.082676742738144e-05, + "loss": 0.9585, + "step": 42849 + }, + { + "epoch": 2.76, + "learning_rate": 1.0826552738358489e-05, + "loss": 0.9112, + "step": 42850 + }, + { + "epoch": 2.76, + "learning_rate": 1.0826338049335538e-05, + "loss": 0.9362, + "step": 42851 + }, + { + "epoch": 2.76, + "learning_rate": 1.0826123360312589e-05, + "loss": 0.8798, + "step": 42852 + }, + { + "epoch": 2.76, + "learning_rate": 1.0825908671289638e-05, + "loss": 0.8367, + "step": 42853 + }, + { + "epoch": 2.76, + "learning_rate": 1.0825693982266687e-05, + "loss": 0.8407, + "step": 42854 + }, + { + "epoch": 2.76, + "learning_rate": 1.0825479293243737e-05, + "loss": 0.9352, + "step": 42855 + }, + { + "epoch": 2.76, + "learning_rate": 1.0825264604220786e-05, + "loss": 0.9048, + "step": 42856 + }, + { + "epoch": 2.76, + "learning_rate": 1.0825049915197837e-05, + "loss": 0.9299, + "step": 42857 + }, + { + "epoch": 2.76, + "learning_rate": 1.0824835226174886e-05, + "loss": 0.9658, + "step": 42858 + }, + { + "epoch": 2.76, + "learning_rate": 1.0824620537151935e-05, + "loss": 0.8798, + "step": 42859 + }, + { + "epoch": 2.76, + "learning_rate": 1.0824405848128987e-05, + "loss": 0.9635, + "step": 42860 + }, + { + "epoch": 2.76, + "learning_rate": 1.0824191159106036e-05, + "loss": 0.9092, + "step": 42861 + }, + { + "epoch": 2.76, + "learning_rate": 1.0823976470083087e-05, + "loss": 0.9088, + "step": 42862 + }, + { + "epoch": 2.76, + "learning_rate": 1.0823761781060136e-05, + "loss": 0.8361, + "step": 42863 + }, + { + "epoch": 2.76, + "learning_rate": 1.0823547092037185e-05, + "loss": 0.8988, + "step": 42864 + }, + { + "epoch": 2.76, + "learning_rate": 1.0823332403014235e-05, + "loss": 0.8314, + "step": 42865 + }, + { + "epoch": 2.76, + "learning_rate": 1.0823117713991284e-05, + "loss": 0.9138, + "step": 42866 + }, + { + "epoch": 2.76, + "learning_rate": 1.0822903024968335e-05, + "loss": 0.9223, + "step": 42867 + }, + { + "epoch": 2.76, + "learning_rate": 1.0822688335945384e-05, + "loss": 0.8629, + "step": 42868 + }, + { + "epoch": 2.76, + "learning_rate": 1.0822473646922433e-05, + "loss": 0.8852, + "step": 42869 + }, + { + "epoch": 2.76, + "learning_rate": 1.0822258957899484e-05, + "loss": 0.8515, + "step": 42870 + }, + { + "epoch": 2.76, + "learning_rate": 1.0822044268876533e-05, + "loss": 0.8959, + "step": 42871 + }, + { + "epoch": 2.76, + "learning_rate": 1.0821829579853581e-05, + "loss": 0.8781, + "step": 42872 + }, + { + "epoch": 2.76, + "learning_rate": 1.0821614890830632e-05, + "loss": 0.8636, + "step": 42873 + }, + { + "epoch": 2.76, + "learning_rate": 1.0821400201807681e-05, + "loss": 0.8806, + "step": 42874 + }, + { + "epoch": 2.76, + "learning_rate": 1.0821185512784733e-05, + "loss": 0.9434, + "step": 42875 + }, + { + "epoch": 2.76, + "learning_rate": 1.0820970823761782e-05, + "loss": 0.8987, + "step": 42876 + }, + { + "epoch": 2.76, + "learning_rate": 1.0820756134738833e-05, + "loss": 0.9994, + "step": 42877 + }, + { + "epoch": 2.76, + "learning_rate": 1.0820541445715882e-05, + "loss": 0.8361, + "step": 42878 + }, + { + "epoch": 2.76, + "learning_rate": 1.0820326756692931e-05, + "loss": 0.7957, + "step": 42879 + }, + { + "epoch": 2.76, + "learning_rate": 1.0820112067669982e-05, + "loss": 0.9284, + "step": 42880 + }, + { + "epoch": 2.76, + "learning_rate": 1.081989737864703e-05, + "loss": 0.9605, + "step": 42881 + }, + { + "epoch": 2.76, + "learning_rate": 1.081968268962408e-05, + "loss": 0.9802, + "step": 42882 + }, + { + "epoch": 2.76, + "learning_rate": 1.081946800060113e-05, + "loss": 0.8815, + "step": 42883 + }, + { + "epoch": 2.76, + "learning_rate": 1.081925331157818e-05, + "loss": 0.8883, + "step": 42884 + }, + { + "epoch": 2.76, + "learning_rate": 1.081903862255523e-05, + "loss": 0.8934, + "step": 42885 + }, + { + "epoch": 2.76, + "learning_rate": 1.0818823933532279e-05, + "loss": 0.9284, + "step": 42886 + }, + { + "epoch": 2.76, + "learning_rate": 1.0818609244509328e-05, + "loss": 0.8881, + "step": 42887 + }, + { + "epoch": 2.76, + "learning_rate": 1.0818394555486378e-05, + "loss": 0.8287, + "step": 42888 + }, + { + "epoch": 2.76, + "learning_rate": 1.0818179866463429e-05, + "loss": 0.8966, + "step": 42889 + }, + { + "epoch": 2.76, + "learning_rate": 1.081796517744048e-05, + "loss": 0.9292, + "step": 42890 + }, + { + "epoch": 2.76, + "learning_rate": 1.0817750488417529e-05, + "loss": 0.8614, + "step": 42891 + }, + { + "epoch": 2.76, + "learning_rate": 1.0817535799394578e-05, + "loss": 0.9059, + "step": 42892 + }, + { + "epoch": 2.76, + "learning_rate": 1.0817321110371628e-05, + "loss": 0.8519, + "step": 42893 + }, + { + "epoch": 2.76, + "learning_rate": 1.0817106421348677e-05, + "loss": 0.8977, + "step": 42894 + }, + { + "epoch": 2.76, + "learning_rate": 1.0816891732325728e-05, + "loss": 0.8767, + "step": 42895 + }, + { + "epoch": 2.76, + "learning_rate": 1.0816677043302777e-05, + "loss": 0.9032, + "step": 42896 + }, + { + "epoch": 2.76, + "learning_rate": 1.0816462354279826e-05, + "loss": 0.8994, + "step": 42897 + }, + { + "epoch": 2.76, + "learning_rate": 1.0816247665256877e-05, + "loss": 0.8751, + "step": 42898 + }, + { + "epoch": 2.76, + "learning_rate": 1.0816032976233926e-05, + "loss": 0.9071, + "step": 42899 + }, + { + "epoch": 2.76, + "learning_rate": 1.0815818287210975e-05, + "loss": 0.8158, + "step": 42900 + }, + { + "epoch": 2.76, + "learning_rate": 1.0815603598188025e-05, + "loss": 0.8971, + "step": 42901 + }, + { + "epoch": 2.76, + "learning_rate": 1.0815388909165074e-05, + "loss": 0.9205, + "step": 42902 + }, + { + "epoch": 2.76, + "learning_rate": 1.0815174220142123e-05, + "loss": 0.926, + "step": 42903 + }, + { + "epoch": 2.76, + "learning_rate": 1.0814959531119175e-05, + "loss": 0.8687, + "step": 42904 + }, + { + "epoch": 2.76, + "learning_rate": 1.0814744842096226e-05, + "loss": 0.8852, + "step": 42905 + }, + { + "epoch": 2.76, + "learning_rate": 1.0814530153073275e-05, + "loss": 0.901, + "step": 42906 + }, + { + "epoch": 2.76, + "learning_rate": 1.0814315464050324e-05, + "loss": 0.882, + "step": 42907 + }, + { + "epoch": 2.76, + "learning_rate": 1.0814100775027375e-05, + "loss": 0.9405, + "step": 42908 + }, + { + "epoch": 2.76, + "learning_rate": 1.0813886086004424e-05, + "loss": 0.9108, + "step": 42909 + }, + { + "epoch": 2.76, + "learning_rate": 1.0813671396981473e-05, + "loss": 1.0146, + "step": 42910 + }, + { + "epoch": 2.76, + "learning_rate": 1.0813456707958523e-05, + "loss": 0.9453, + "step": 42911 + }, + { + "epoch": 2.76, + "learning_rate": 1.0813242018935572e-05, + "loss": 0.8861, + "step": 42912 + }, + { + "epoch": 2.76, + "learning_rate": 1.0813027329912621e-05, + "loss": 0.8285, + "step": 42913 + }, + { + "epoch": 2.76, + "learning_rate": 1.0812812640889672e-05, + "loss": 0.9024, + "step": 42914 + }, + { + "epoch": 2.76, + "learning_rate": 1.0812597951866721e-05, + "loss": 0.9279, + "step": 42915 + }, + { + "epoch": 2.76, + "learning_rate": 1.0812383262843772e-05, + "loss": 0.885, + "step": 42916 + }, + { + "epoch": 2.76, + "learning_rate": 1.081216857382082e-05, + "loss": 0.8483, + "step": 42917 + }, + { + "epoch": 2.76, + "learning_rate": 1.0811953884797873e-05, + "loss": 0.9274, + "step": 42918 + }, + { + "epoch": 2.76, + "learning_rate": 1.0811739195774922e-05, + "loss": 0.8847, + "step": 42919 + }, + { + "epoch": 2.76, + "learning_rate": 1.081152450675197e-05, + "loss": 0.9393, + "step": 42920 + }, + { + "epoch": 2.76, + "learning_rate": 1.0811309817729021e-05, + "loss": 0.943, + "step": 42921 + }, + { + "epoch": 2.76, + "learning_rate": 1.081109512870607e-05, + "loss": 0.9248, + "step": 42922 + }, + { + "epoch": 2.76, + "learning_rate": 1.081088043968312e-05, + "loss": 0.9149, + "step": 42923 + }, + { + "epoch": 2.76, + "learning_rate": 1.081066575066017e-05, + "loss": 0.8538, + "step": 42924 + }, + { + "epoch": 2.76, + "learning_rate": 1.0810451061637219e-05, + "loss": 0.883, + "step": 42925 + }, + { + "epoch": 2.76, + "learning_rate": 1.081023637261427e-05, + "loss": 0.9143, + "step": 42926 + }, + { + "epoch": 2.76, + "learning_rate": 1.0810021683591319e-05, + "loss": 0.8331, + "step": 42927 + }, + { + "epoch": 2.76, + "learning_rate": 1.0809806994568368e-05, + "loss": 0.9101, + "step": 42928 + }, + { + "epoch": 2.76, + "learning_rate": 1.0809592305545418e-05, + "loss": 0.8859, + "step": 42929 + }, + { + "epoch": 2.76, + "learning_rate": 1.0809377616522467e-05, + "loss": 0.9134, + "step": 42930 + }, + { + "epoch": 2.76, + "learning_rate": 1.0809162927499516e-05, + "loss": 0.9033, + "step": 42931 + }, + { + "epoch": 2.76, + "learning_rate": 1.0808948238476567e-05, + "loss": 0.9, + "step": 42932 + }, + { + "epoch": 2.76, + "learning_rate": 1.0808733549453617e-05, + "loss": 0.9346, + "step": 42933 + }, + { + "epoch": 2.76, + "learning_rate": 1.0808518860430668e-05, + "loss": 0.9543, + "step": 42934 + }, + { + "epoch": 2.76, + "learning_rate": 1.0808304171407717e-05, + "loss": 0.9096, + "step": 42935 + }, + { + "epoch": 2.76, + "learning_rate": 1.0808089482384768e-05, + "loss": 0.9424, + "step": 42936 + }, + { + "epoch": 2.76, + "learning_rate": 1.0807874793361817e-05, + "loss": 0.8312, + "step": 42937 + }, + { + "epoch": 2.76, + "learning_rate": 1.0807660104338866e-05, + "loss": 0.9369, + "step": 42938 + }, + { + "epoch": 2.76, + "learning_rate": 1.0807445415315916e-05, + "loss": 0.8548, + "step": 42939 + }, + { + "epoch": 2.76, + "learning_rate": 1.0807230726292965e-05, + "loss": 0.8988, + "step": 42940 + }, + { + "epoch": 2.76, + "learning_rate": 1.0807016037270014e-05, + "loss": 0.9009, + "step": 42941 + }, + { + "epoch": 2.76, + "learning_rate": 1.0806801348247065e-05, + "loss": 0.8837, + "step": 42942 + }, + { + "epoch": 2.76, + "learning_rate": 1.0806586659224114e-05, + "loss": 0.9582, + "step": 42943 + }, + { + "epoch": 2.76, + "learning_rate": 1.0806371970201165e-05, + "loss": 0.886, + "step": 42944 + }, + { + "epoch": 2.76, + "learning_rate": 1.0806157281178214e-05, + "loss": 0.8513, + "step": 42945 + }, + { + "epoch": 2.76, + "learning_rate": 1.0805942592155262e-05, + "loss": 0.8528, + "step": 42946 + }, + { + "epoch": 2.76, + "learning_rate": 1.0805727903132315e-05, + "loss": 0.922, + "step": 42947 + }, + { + "epoch": 2.76, + "learning_rate": 1.0805513214109364e-05, + "loss": 0.9146, + "step": 42948 + }, + { + "epoch": 2.76, + "learning_rate": 1.0805298525086414e-05, + "loss": 0.872, + "step": 42949 + }, + { + "epoch": 2.76, + "learning_rate": 1.0805083836063463e-05, + "loss": 0.9084, + "step": 42950 + }, + { + "epoch": 2.76, + "learning_rate": 1.0804869147040512e-05, + "loss": 0.904, + "step": 42951 + }, + { + "epoch": 2.76, + "learning_rate": 1.0804654458017563e-05, + "loss": 1.0003, + "step": 42952 + }, + { + "epoch": 2.76, + "learning_rate": 1.0804439768994612e-05, + "loss": 0.9311, + "step": 42953 + }, + { + "epoch": 2.76, + "learning_rate": 1.0804225079971663e-05, + "loss": 0.8778, + "step": 42954 + }, + { + "epoch": 2.76, + "learning_rate": 1.0804010390948712e-05, + "loss": 0.9588, + "step": 42955 + }, + { + "epoch": 2.76, + "learning_rate": 1.080379570192576e-05, + "loss": 0.8974, + "step": 42956 + }, + { + "epoch": 2.76, + "learning_rate": 1.0803581012902811e-05, + "loss": 0.9074, + "step": 42957 + }, + { + "epoch": 2.76, + "learning_rate": 1.080336632387986e-05, + "loss": 0.8711, + "step": 42958 + }, + { + "epoch": 2.76, + "learning_rate": 1.080315163485691e-05, + "loss": 0.9311, + "step": 42959 + }, + { + "epoch": 2.76, + "learning_rate": 1.080293694583396e-05, + "loss": 0.9057, + "step": 42960 + }, + { + "epoch": 2.76, + "learning_rate": 1.0802722256811009e-05, + "loss": 0.8752, + "step": 42961 + }, + { + "epoch": 2.76, + "learning_rate": 1.0802507567788061e-05, + "loss": 0.8776, + "step": 42962 + }, + { + "epoch": 2.76, + "learning_rate": 1.080229287876511e-05, + "loss": 0.9223, + "step": 42963 + }, + { + "epoch": 2.76, + "learning_rate": 1.080207818974216e-05, + "loss": 0.9283, + "step": 42964 + }, + { + "epoch": 2.76, + "learning_rate": 1.080186350071921e-05, + "loss": 0.9019, + "step": 42965 + }, + { + "epoch": 2.76, + "learning_rate": 1.0801648811696259e-05, + "loss": 0.8631, + "step": 42966 + }, + { + "epoch": 2.76, + "learning_rate": 1.080143412267331e-05, + "loss": 0.9054, + "step": 42967 + }, + { + "epoch": 2.76, + "learning_rate": 1.0801219433650358e-05, + "loss": 0.8451, + "step": 42968 + }, + { + "epoch": 2.76, + "learning_rate": 1.0801004744627407e-05, + "loss": 0.8669, + "step": 42969 + }, + { + "epoch": 2.76, + "learning_rate": 1.0800790055604458e-05, + "loss": 0.9387, + "step": 42970 + }, + { + "epoch": 2.76, + "learning_rate": 1.0800575366581507e-05, + "loss": 0.8343, + "step": 42971 + }, + { + "epoch": 2.76, + "learning_rate": 1.0800360677558558e-05, + "loss": 0.9161, + "step": 42972 + }, + { + "epoch": 2.76, + "learning_rate": 1.0800145988535607e-05, + "loss": 0.9549, + "step": 42973 + }, + { + "epoch": 2.76, + "learning_rate": 1.0799931299512656e-05, + "loss": 0.9117, + "step": 42974 + }, + { + "epoch": 2.76, + "learning_rate": 1.0799716610489706e-05, + "loss": 0.8937, + "step": 42975 + }, + { + "epoch": 2.76, + "learning_rate": 1.0799501921466757e-05, + "loss": 1.0046, + "step": 42976 + }, + { + "epoch": 2.77, + "learning_rate": 1.0799287232443808e-05, + "loss": 0.8314, + "step": 42977 + }, + { + "epoch": 2.77, + "learning_rate": 1.0799072543420856e-05, + "loss": 0.9447, + "step": 42978 + }, + { + "epoch": 2.77, + "learning_rate": 1.0798857854397905e-05, + "loss": 0.9303, + "step": 42979 + }, + { + "epoch": 2.77, + "learning_rate": 1.0798643165374956e-05, + "loss": 0.9181, + "step": 42980 + }, + { + "epoch": 2.77, + "learning_rate": 1.0798428476352005e-05, + "loss": 0.8962, + "step": 42981 + }, + { + "epoch": 2.77, + "learning_rate": 1.0798213787329056e-05, + "loss": 0.9166, + "step": 42982 + }, + { + "epoch": 2.77, + "learning_rate": 1.0797999098306105e-05, + "loss": 0.8411, + "step": 42983 + }, + { + "epoch": 2.77, + "learning_rate": 1.0797784409283154e-05, + "loss": 0.8819, + "step": 42984 + }, + { + "epoch": 2.77, + "learning_rate": 1.0797569720260204e-05, + "loss": 0.8816, + "step": 42985 + }, + { + "epoch": 2.77, + "learning_rate": 1.0797355031237253e-05, + "loss": 0.8931, + "step": 42986 + }, + { + "epoch": 2.77, + "learning_rate": 1.0797140342214302e-05, + "loss": 0.8893, + "step": 42987 + }, + { + "epoch": 2.77, + "learning_rate": 1.0796925653191353e-05, + "loss": 0.8899, + "step": 42988 + }, + { + "epoch": 2.77, + "learning_rate": 1.0796710964168402e-05, + "loss": 0.9699, + "step": 42989 + }, + { + "epoch": 2.77, + "learning_rate": 1.079649627514545e-05, + "loss": 0.9626, + "step": 42990 + }, + { + "epoch": 2.77, + "learning_rate": 1.0796281586122503e-05, + "loss": 0.9554, + "step": 42991 + }, + { + "epoch": 2.77, + "learning_rate": 1.0796066897099554e-05, + "loss": 0.9283, + "step": 42992 + }, + { + "epoch": 2.77, + "learning_rate": 1.0795852208076603e-05, + "loss": 0.8776, + "step": 42993 + }, + { + "epoch": 2.77, + "learning_rate": 1.0795637519053652e-05, + "loss": 0.9538, + "step": 42994 + }, + { + "epoch": 2.77, + "learning_rate": 1.0795422830030702e-05, + "loss": 0.8823, + "step": 42995 + }, + { + "epoch": 2.77, + "learning_rate": 1.0795208141007751e-05, + "loss": 0.8523, + "step": 42996 + }, + { + "epoch": 2.77, + "learning_rate": 1.07949934519848e-05, + "loss": 0.9788, + "step": 42997 + }, + { + "epoch": 2.77, + "learning_rate": 1.0794778762961851e-05, + "loss": 0.8991, + "step": 42998 + }, + { + "epoch": 2.77, + "learning_rate": 1.07945640739389e-05, + "loss": 0.915, + "step": 42999 + }, + { + "epoch": 2.77, + "learning_rate": 1.0794349384915949e-05, + "loss": 0.9003, + "step": 43000 + }, + { + "epoch": 2.77, + "eval_laion/cider": 0.545078297623235, + "eval_laion/meteor": 0.3984140568744658, + "eval_laion/pred_num_words": 38.541015625, + "eval_laion/sacrebleu": 16.371039428949306, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0390197038650513, + "eval_runtime": 100.0088, + "eval_samples_per_second": 4.99, + "eval_steps_per_second": 0.32, + "step": 43000 + }, + { + "epoch": 2.77, + "learning_rate": 1.0794134695893e-05, + "loss": 0.9299, + "step": 43001 + }, + { + "epoch": 2.77, + "learning_rate": 1.0793920006870049e-05, + "loss": 0.9513, + "step": 43002 + }, + { + "epoch": 2.77, + "learning_rate": 1.07937053178471e-05, + "loss": 0.9549, + "step": 43003 + }, + { + "epoch": 2.77, + "learning_rate": 1.0793490628824148e-05, + "loss": 0.9656, + "step": 43004 + }, + { + "epoch": 2.77, + "learning_rate": 1.07932759398012e-05, + "loss": 0.9013, + "step": 43005 + }, + { + "epoch": 2.77, + "learning_rate": 1.079306125077825e-05, + "loss": 0.8616, + "step": 43006 + }, + { + "epoch": 2.77, + "learning_rate": 1.0792846561755298e-05, + "loss": 0.8691, + "step": 43007 + }, + { + "epoch": 2.77, + "learning_rate": 1.0792631872732349e-05, + "loss": 0.8918, + "step": 43008 + }, + { + "epoch": 2.77, + "learning_rate": 1.0792417183709398e-05, + "loss": 0.8951, + "step": 43009 + }, + { + "epoch": 2.77, + "learning_rate": 1.0792202494686447e-05, + "loss": 0.8469, + "step": 43010 + }, + { + "epoch": 2.77, + "learning_rate": 1.0791987805663498e-05, + "loss": 0.901, + "step": 43011 + }, + { + "epoch": 2.77, + "learning_rate": 1.0791773116640547e-05, + "loss": 0.9268, + "step": 43012 + }, + { + "epoch": 2.77, + "learning_rate": 1.0791558427617597e-05, + "loss": 0.9422, + "step": 43013 + }, + { + "epoch": 2.77, + "learning_rate": 1.0791343738594646e-05, + "loss": 0.8634, + "step": 43014 + }, + { + "epoch": 2.77, + "learning_rate": 1.0791129049571695e-05, + "loss": 0.9442, + "step": 43015 + }, + { + "epoch": 2.77, + "learning_rate": 1.0790914360548746e-05, + "loss": 0.9483, + "step": 43016 + }, + { + "epoch": 2.77, + "learning_rate": 1.0790699671525795e-05, + "loss": 0.901, + "step": 43017 + }, + { + "epoch": 2.77, + "learning_rate": 1.0790484982502844e-05, + "loss": 0.925, + "step": 43018 + }, + { + "epoch": 2.77, + "learning_rate": 1.0790270293479895e-05, + "loss": 0.8757, + "step": 43019 + }, + { + "epoch": 2.77, + "learning_rate": 1.0790055604456945e-05, + "loss": 0.9716, + "step": 43020 + }, + { + "epoch": 2.77, + "learning_rate": 1.0789840915433996e-05, + "loss": 0.9408, + "step": 43021 + }, + { + "epoch": 2.77, + "learning_rate": 1.0789626226411045e-05, + "loss": 0.8991, + "step": 43022 + }, + { + "epoch": 2.77, + "learning_rate": 1.0789411537388095e-05, + "loss": 0.9217, + "step": 43023 + }, + { + "epoch": 2.77, + "learning_rate": 1.0789196848365144e-05, + "loss": 0.9105, + "step": 43024 + }, + { + "epoch": 2.77, + "learning_rate": 1.0788982159342193e-05, + "loss": 0.9143, + "step": 43025 + }, + { + "epoch": 2.77, + "learning_rate": 1.0788767470319244e-05, + "loss": 0.9401, + "step": 43026 + }, + { + "epoch": 2.77, + "learning_rate": 1.0788552781296293e-05, + "loss": 0.8913, + "step": 43027 + }, + { + "epoch": 2.77, + "learning_rate": 1.0788338092273342e-05, + "loss": 0.8847, + "step": 43028 + }, + { + "epoch": 2.77, + "learning_rate": 1.0788123403250393e-05, + "loss": 0.9235, + "step": 43029 + }, + { + "epoch": 2.77, + "learning_rate": 1.0787908714227442e-05, + "loss": 0.9264, + "step": 43030 + }, + { + "epoch": 2.77, + "learning_rate": 1.0787694025204492e-05, + "loss": 0.8974, + "step": 43031 + }, + { + "epoch": 2.77, + "learning_rate": 1.0787479336181541e-05, + "loss": 0.8521, + "step": 43032 + }, + { + "epoch": 2.77, + "learning_rate": 1.078726464715859e-05, + "loss": 0.8932, + "step": 43033 + }, + { + "epoch": 2.77, + "learning_rate": 1.0787049958135643e-05, + "loss": 0.8894, + "step": 43034 + }, + { + "epoch": 2.77, + "learning_rate": 1.0786835269112692e-05, + "loss": 0.9081, + "step": 43035 + }, + { + "epoch": 2.77, + "learning_rate": 1.0786620580089742e-05, + "loss": 0.9373, + "step": 43036 + }, + { + "epoch": 2.77, + "learning_rate": 1.0786405891066791e-05, + "loss": 0.8392, + "step": 43037 + }, + { + "epoch": 2.77, + "learning_rate": 1.078619120204384e-05, + "loss": 0.8705, + "step": 43038 + }, + { + "epoch": 2.77, + "learning_rate": 1.078597651302089e-05, + "loss": 0.8837, + "step": 43039 + }, + { + "epoch": 2.77, + "learning_rate": 1.078576182399794e-05, + "loss": 0.9299, + "step": 43040 + }, + { + "epoch": 2.77, + "learning_rate": 1.078554713497499e-05, + "loss": 0.9188, + "step": 43041 + }, + { + "epoch": 2.77, + "learning_rate": 1.078533244595204e-05, + "loss": 0.9125, + "step": 43042 + }, + { + "epoch": 2.77, + "learning_rate": 1.0785117756929088e-05, + "loss": 0.9048, + "step": 43043 + }, + { + "epoch": 2.77, + "learning_rate": 1.0784903067906139e-05, + "loss": 0.8758, + "step": 43044 + }, + { + "epoch": 2.77, + "learning_rate": 1.0784688378883188e-05, + "loss": 0.8376, + "step": 43045 + }, + { + "epoch": 2.77, + "learning_rate": 1.0784473689860237e-05, + "loss": 0.9006, + "step": 43046 + }, + { + "epoch": 2.77, + "learning_rate": 1.0784259000837288e-05, + "loss": 0.8799, + "step": 43047 + }, + { + "epoch": 2.77, + "learning_rate": 1.0784044311814337e-05, + "loss": 0.9127, + "step": 43048 + }, + { + "epoch": 2.77, + "learning_rate": 1.0783829622791389e-05, + "loss": 0.9247, + "step": 43049 + }, + { + "epoch": 2.77, + "learning_rate": 1.0783614933768438e-05, + "loss": 0.8337, + "step": 43050 + }, + { + "epoch": 2.77, + "learning_rate": 1.0783400244745489e-05, + "loss": 0.9211, + "step": 43051 + }, + { + "epoch": 2.77, + "learning_rate": 1.0783185555722537e-05, + "loss": 0.9874, + "step": 43052 + }, + { + "epoch": 2.77, + "learning_rate": 1.0782970866699586e-05, + "loss": 0.8599, + "step": 43053 + }, + { + "epoch": 2.77, + "learning_rate": 1.0782756177676637e-05, + "loss": 0.8587, + "step": 43054 + }, + { + "epoch": 2.77, + "learning_rate": 1.0782541488653686e-05, + "loss": 0.8863, + "step": 43055 + }, + { + "epoch": 2.77, + "learning_rate": 1.0782326799630735e-05, + "loss": 0.838, + "step": 43056 + }, + { + "epoch": 2.77, + "learning_rate": 1.0782112110607786e-05, + "loss": 0.9715, + "step": 43057 + }, + { + "epoch": 2.77, + "learning_rate": 1.0781897421584835e-05, + "loss": 0.9021, + "step": 43058 + }, + { + "epoch": 2.77, + "learning_rate": 1.0781682732561884e-05, + "loss": 0.8532, + "step": 43059 + }, + { + "epoch": 2.77, + "learning_rate": 1.0781468043538934e-05, + "loss": 0.9672, + "step": 43060 + }, + { + "epoch": 2.77, + "learning_rate": 1.0781253354515983e-05, + "loss": 0.8736, + "step": 43061 + }, + { + "epoch": 2.77, + "learning_rate": 1.0781038665493034e-05, + "loss": 0.8867, + "step": 43062 + }, + { + "epoch": 2.77, + "learning_rate": 1.0780823976470085e-05, + "loss": 0.9027, + "step": 43063 + }, + { + "epoch": 2.77, + "learning_rate": 1.0780609287447135e-05, + "loss": 0.8633, + "step": 43064 + }, + { + "epoch": 2.77, + "learning_rate": 1.0780394598424184e-05, + "loss": 0.865, + "step": 43065 + }, + { + "epoch": 2.77, + "learning_rate": 1.0780179909401233e-05, + "loss": 0.8523, + "step": 43066 + }, + { + "epoch": 2.77, + "learning_rate": 1.0779965220378284e-05, + "loss": 0.9004, + "step": 43067 + }, + { + "epoch": 2.77, + "learning_rate": 1.0779750531355333e-05, + "loss": 0.9162, + "step": 43068 + }, + { + "epoch": 2.77, + "learning_rate": 1.0779535842332383e-05, + "loss": 0.8867, + "step": 43069 + }, + { + "epoch": 2.77, + "learning_rate": 1.0779321153309432e-05, + "loss": 0.9045, + "step": 43070 + }, + { + "epoch": 2.77, + "learning_rate": 1.0779106464286481e-05, + "loss": 0.9622, + "step": 43071 + }, + { + "epoch": 2.77, + "learning_rate": 1.0778891775263532e-05, + "loss": 0.9021, + "step": 43072 + }, + { + "epoch": 2.77, + "learning_rate": 1.0778677086240581e-05, + "loss": 0.9378, + "step": 43073 + }, + { + "epoch": 2.77, + "learning_rate": 1.077846239721763e-05, + "loss": 0.9178, + "step": 43074 + }, + { + "epoch": 2.77, + "learning_rate": 1.077824770819468e-05, + "loss": 0.8493, + "step": 43075 + }, + { + "epoch": 2.77, + "learning_rate": 1.077803301917173e-05, + "loss": 0.9299, + "step": 43076 + }, + { + "epoch": 2.77, + "learning_rate": 1.0777818330148779e-05, + "loss": 0.9566, + "step": 43077 + }, + { + "epoch": 2.77, + "learning_rate": 1.0777603641125831e-05, + "loss": 0.916, + "step": 43078 + }, + { + "epoch": 2.77, + "learning_rate": 1.0777388952102882e-05, + "loss": 0.8471, + "step": 43079 + }, + { + "epoch": 2.77, + "learning_rate": 1.077717426307993e-05, + "loss": 0.96, + "step": 43080 + }, + { + "epoch": 2.77, + "learning_rate": 1.077695957405698e-05, + "loss": 0.8601, + "step": 43081 + }, + { + "epoch": 2.77, + "learning_rate": 1.077674488503403e-05, + "loss": 0.8974, + "step": 43082 + }, + { + "epoch": 2.77, + "learning_rate": 1.0776530196011079e-05, + "loss": 0.9021, + "step": 43083 + }, + { + "epoch": 2.77, + "learning_rate": 1.0776315506988128e-05, + "loss": 0.8651, + "step": 43084 + }, + { + "epoch": 2.77, + "learning_rate": 1.0776100817965179e-05, + "loss": 0.8572, + "step": 43085 + }, + { + "epoch": 2.77, + "learning_rate": 1.0775886128942228e-05, + "loss": 0.899, + "step": 43086 + }, + { + "epoch": 2.77, + "learning_rate": 1.0775671439919277e-05, + "loss": 0.8258, + "step": 43087 + }, + { + "epoch": 2.77, + "learning_rate": 1.0775456750896327e-05, + "loss": 0.9045, + "step": 43088 + }, + { + "epoch": 2.77, + "learning_rate": 1.0775242061873376e-05, + "loss": 0.8556, + "step": 43089 + }, + { + "epoch": 2.77, + "learning_rate": 1.0775027372850427e-05, + "loss": 0.932, + "step": 43090 + }, + { + "epoch": 2.77, + "learning_rate": 1.0774812683827476e-05, + "loss": 0.9536, + "step": 43091 + }, + { + "epoch": 2.77, + "learning_rate": 1.0774597994804528e-05, + "loss": 0.8959, + "step": 43092 + }, + { + "epoch": 2.77, + "learning_rate": 1.0774383305781577e-05, + "loss": 0.855, + "step": 43093 + }, + { + "epoch": 2.77, + "learning_rate": 1.0774168616758626e-05, + "loss": 0.8709, + "step": 43094 + }, + { + "epoch": 2.77, + "learning_rate": 1.0773953927735677e-05, + "loss": 0.927, + "step": 43095 + }, + { + "epoch": 2.77, + "learning_rate": 1.0773739238712726e-05, + "loss": 0.8667, + "step": 43096 + }, + { + "epoch": 2.77, + "learning_rate": 1.0773524549689775e-05, + "loss": 0.9209, + "step": 43097 + }, + { + "epoch": 2.77, + "learning_rate": 1.0773309860666825e-05, + "loss": 0.8414, + "step": 43098 + }, + { + "epoch": 2.77, + "learning_rate": 1.0773095171643874e-05, + "loss": 0.8683, + "step": 43099 + }, + { + "epoch": 2.77, + "learning_rate": 1.0772880482620925e-05, + "loss": 0.9452, + "step": 43100 + }, + { + "epoch": 2.77, + "learning_rate": 1.0772665793597974e-05, + "loss": 0.8112, + "step": 43101 + }, + { + "epoch": 2.77, + "learning_rate": 1.0772451104575023e-05, + "loss": 0.9136, + "step": 43102 + }, + { + "epoch": 2.77, + "learning_rate": 1.0772236415552074e-05, + "loss": 0.8617, + "step": 43103 + }, + { + "epoch": 2.77, + "learning_rate": 1.0772021726529123e-05, + "loss": 0.884, + "step": 43104 + }, + { + "epoch": 2.77, + "learning_rate": 1.0771807037506172e-05, + "loss": 0.892, + "step": 43105 + }, + { + "epoch": 2.77, + "learning_rate": 1.0771592348483222e-05, + "loss": 0.8794, + "step": 43106 + }, + { + "epoch": 2.77, + "learning_rate": 1.0771377659460273e-05, + "loss": 0.9152, + "step": 43107 + }, + { + "epoch": 2.77, + "learning_rate": 1.0771162970437324e-05, + "loss": 0.8621, + "step": 43108 + }, + { + "epoch": 2.77, + "learning_rate": 1.0770948281414373e-05, + "loss": 0.9103, + "step": 43109 + }, + { + "epoch": 2.77, + "learning_rate": 1.0770733592391423e-05, + "loss": 0.8422, + "step": 43110 + }, + { + "epoch": 2.77, + "learning_rate": 1.0770518903368472e-05, + "loss": 0.883, + "step": 43111 + }, + { + "epoch": 2.77, + "learning_rate": 1.0770304214345521e-05, + "loss": 0.922, + "step": 43112 + }, + { + "epoch": 2.77, + "learning_rate": 1.0770089525322572e-05, + "loss": 0.9055, + "step": 43113 + }, + { + "epoch": 2.77, + "learning_rate": 1.076987483629962e-05, + "loss": 0.8644, + "step": 43114 + }, + { + "epoch": 2.77, + "learning_rate": 1.076966014727667e-05, + "loss": 0.8385, + "step": 43115 + }, + { + "epoch": 2.77, + "learning_rate": 1.076944545825372e-05, + "loss": 0.8802, + "step": 43116 + }, + { + "epoch": 2.77, + "learning_rate": 1.076923076923077e-05, + "loss": 0.9333, + "step": 43117 + }, + { + "epoch": 2.77, + "learning_rate": 1.076901608020782e-05, + "loss": 0.8974, + "step": 43118 + }, + { + "epoch": 2.77, + "learning_rate": 1.0768801391184869e-05, + "loss": 0.8877, + "step": 43119 + }, + { + "epoch": 2.77, + "learning_rate": 1.0768586702161918e-05, + "loss": 0.8847, + "step": 43120 + }, + { + "epoch": 2.77, + "learning_rate": 1.076837201313897e-05, + "loss": 0.8717, + "step": 43121 + }, + { + "epoch": 2.77, + "learning_rate": 1.076815732411602e-05, + "loss": 0.8939, + "step": 43122 + }, + { + "epoch": 2.77, + "learning_rate": 1.076794263509307e-05, + "loss": 0.929, + "step": 43123 + }, + { + "epoch": 2.77, + "learning_rate": 1.0767727946070119e-05, + "loss": 0.8691, + "step": 43124 + }, + { + "epoch": 2.77, + "learning_rate": 1.0767513257047168e-05, + "loss": 0.918, + "step": 43125 + }, + { + "epoch": 2.77, + "learning_rate": 1.0767298568024218e-05, + "loss": 0.8105, + "step": 43126 + }, + { + "epoch": 2.77, + "learning_rate": 1.0767083879001267e-05, + "loss": 0.8939, + "step": 43127 + }, + { + "epoch": 2.77, + "learning_rate": 1.0766869189978318e-05, + "loss": 0.9109, + "step": 43128 + }, + { + "epoch": 2.77, + "learning_rate": 1.0766654500955367e-05, + "loss": 0.8665, + "step": 43129 + }, + { + "epoch": 2.77, + "learning_rate": 1.0766439811932416e-05, + "loss": 0.9248, + "step": 43130 + }, + { + "epoch": 2.77, + "learning_rate": 1.0766225122909467e-05, + "loss": 0.9123, + "step": 43131 + }, + { + "epoch": 2.78, + "learning_rate": 1.0766010433886516e-05, + "loss": 0.9582, + "step": 43132 + }, + { + "epoch": 2.78, + "learning_rate": 1.0765795744863565e-05, + "loss": 0.9373, + "step": 43133 + }, + { + "epoch": 2.78, + "learning_rate": 1.0765581055840615e-05, + "loss": 0.9406, + "step": 43134 + }, + { + "epoch": 2.78, + "learning_rate": 1.0765366366817664e-05, + "loss": 0.9695, + "step": 43135 + }, + { + "epoch": 2.78, + "learning_rate": 1.0765151677794717e-05, + "loss": 0.8974, + "step": 43136 + }, + { + "epoch": 2.78, + "learning_rate": 1.0764936988771766e-05, + "loss": 0.8811, + "step": 43137 + }, + { + "epoch": 2.78, + "learning_rate": 1.0764722299748816e-05, + "loss": 0.8688, + "step": 43138 + }, + { + "epoch": 2.78, + "learning_rate": 1.0764507610725865e-05, + "loss": 0.9246, + "step": 43139 + }, + { + "epoch": 2.78, + "learning_rate": 1.0764292921702914e-05, + "loss": 0.9251, + "step": 43140 + }, + { + "epoch": 2.78, + "learning_rate": 1.0764078232679965e-05, + "loss": 0.9381, + "step": 43141 + }, + { + "epoch": 2.78, + "learning_rate": 1.0763863543657014e-05, + "loss": 0.8768, + "step": 43142 + }, + { + "epoch": 2.78, + "learning_rate": 1.0763648854634063e-05, + "loss": 0.8535, + "step": 43143 + }, + { + "epoch": 2.78, + "learning_rate": 1.0763434165611113e-05, + "loss": 0.9218, + "step": 43144 + }, + { + "epoch": 2.78, + "learning_rate": 1.0763219476588162e-05, + "loss": 0.9201, + "step": 43145 + }, + { + "epoch": 2.78, + "learning_rate": 1.0763004787565211e-05, + "loss": 0.8583, + "step": 43146 + }, + { + "epoch": 2.78, + "learning_rate": 1.0762790098542262e-05, + "loss": 0.856, + "step": 43147 + }, + { + "epoch": 2.78, + "learning_rate": 1.0762575409519311e-05, + "loss": 0.9627, + "step": 43148 + }, + { + "epoch": 2.78, + "learning_rate": 1.0762360720496362e-05, + "loss": 0.988, + "step": 43149 + }, + { + "epoch": 2.78, + "learning_rate": 1.0762146031473412e-05, + "loss": 0.8757, + "step": 43150 + }, + { + "epoch": 2.78, + "learning_rate": 1.0761931342450463e-05, + "loss": 0.8576, + "step": 43151 + }, + { + "epoch": 2.78, + "learning_rate": 1.0761716653427512e-05, + "loss": 0.9842, + "step": 43152 + }, + { + "epoch": 2.78, + "learning_rate": 1.0761501964404561e-05, + "loss": 0.8744, + "step": 43153 + }, + { + "epoch": 2.78, + "learning_rate": 1.0761287275381612e-05, + "loss": 0.8472, + "step": 43154 + }, + { + "epoch": 2.78, + "learning_rate": 1.076107258635866e-05, + "loss": 0.8588, + "step": 43155 + }, + { + "epoch": 2.78, + "learning_rate": 1.076085789733571e-05, + "loss": 0.8974, + "step": 43156 + }, + { + "epoch": 2.78, + "learning_rate": 1.076064320831276e-05, + "loss": 0.9651, + "step": 43157 + }, + { + "epoch": 2.78, + "learning_rate": 1.0760428519289809e-05, + "loss": 0.863, + "step": 43158 + }, + { + "epoch": 2.78, + "learning_rate": 1.076021383026686e-05, + "loss": 0.8729, + "step": 43159 + }, + { + "epoch": 2.78, + "learning_rate": 1.0759999141243909e-05, + "loss": 0.8749, + "step": 43160 + }, + { + "epoch": 2.78, + "learning_rate": 1.0759784452220958e-05, + "loss": 0.8856, + "step": 43161 + }, + { + "epoch": 2.78, + "learning_rate": 1.0759569763198008e-05, + "loss": 0.9252, + "step": 43162 + }, + { + "epoch": 2.78, + "learning_rate": 1.0759355074175057e-05, + "loss": 0.9091, + "step": 43163 + }, + { + "epoch": 2.78, + "learning_rate": 1.0759140385152106e-05, + "loss": 0.8774, + "step": 43164 + }, + { + "epoch": 2.78, + "learning_rate": 1.0758925696129159e-05, + "loss": 0.7952, + "step": 43165 + }, + { + "epoch": 2.78, + "learning_rate": 1.0758711007106208e-05, + "loss": 0.9165, + "step": 43166 + }, + { + "epoch": 2.78, + "learning_rate": 1.0758496318083258e-05, + "loss": 0.8559, + "step": 43167 + }, + { + "epoch": 2.78, + "learning_rate": 1.0758281629060307e-05, + "loss": 0.882, + "step": 43168 + }, + { + "epoch": 2.78, + "learning_rate": 1.0758066940037358e-05, + "loss": 0.9729, + "step": 43169 + }, + { + "epoch": 2.78, + "learning_rate": 1.0757852251014407e-05, + "loss": 0.909, + "step": 43170 + }, + { + "epoch": 2.78, + "learning_rate": 1.0757637561991456e-05, + "loss": 0.9158, + "step": 43171 + }, + { + "epoch": 2.78, + "learning_rate": 1.0757422872968506e-05, + "loss": 0.8618, + "step": 43172 + }, + { + "epoch": 2.78, + "learning_rate": 1.0757208183945555e-05, + "loss": 0.9288, + "step": 43173 + }, + { + "epoch": 2.78, + "learning_rate": 1.0756993494922604e-05, + "loss": 0.8968, + "step": 43174 + }, + { + "epoch": 2.78, + "learning_rate": 1.0756778805899655e-05, + "loss": 0.9218, + "step": 43175 + }, + { + "epoch": 2.78, + "learning_rate": 1.0756564116876704e-05, + "loss": 0.8746, + "step": 43176 + }, + { + "epoch": 2.78, + "learning_rate": 1.0756349427853755e-05, + "loss": 0.9259, + "step": 43177 + }, + { + "epoch": 2.78, + "learning_rate": 1.0756134738830804e-05, + "loss": 0.869, + "step": 43178 + }, + { + "epoch": 2.78, + "learning_rate": 1.0755920049807856e-05, + "loss": 0.9203, + "step": 43179 + }, + { + "epoch": 2.78, + "learning_rate": 1.0755705360784905e-05, + "loss": 0.8543, + "step": 43180 + }, + { + "epoch": 2.78, + "learning_rate": 1.0755490671761954e-05, + "loss": 0.9497, + "step": 43181 + }, + { + "epoch": 2.78, + "learning_rate": 1.0755275982739005e-05, + "loss": 0.853, + "step": 43182 + }, + { + "epoch": 2.78, + "learning_rate": 1.0755061293716054e-05, + "loss": 0.9098, + "step": 43183 + }, + { + "epoch": 2.78, + "learning_rate": 1.0754846604693102e-05, + "loss": 0.9245, + "step": 43184 + }, + { + "epoch": 2.78, + "learning_rate": 1.0754631915670153e-05, + "loss": 0.791, + "step": 43185 + }, + { + "epoch": 2.78, + "learning_rate": 1.0754417226647202e-05, + "loss": 0.9319, + "step": 43186 + }, + { + "epoch": 2.78, + "learning_rate": 1.0754202537624253e-05, + "loss": 0.8786, + "step": 43187 + }, + { + "epoch": 2.78, + "learning_rate": 1.0753987848601302e-05, + "loss": 0.9328, + "step": 43188 + }, + { + "epoch": 2.78, + "learning_rate": 1.075377315957835e-05, + "loss": 0.9214, + "step": 43189 + }, + { + "epoch": 2.78, + "learning_rate": 1.0753558470555401e-05, + "loss": 0.9146, + "step": 43190 + }, + { + "epoch": 2.78, + "learning_rate": 1.075334378153245e-05, + "loss": 0.9094, + "step": 43191 + }, + { + "epoch": 2.78, + "learning_rate": 1.07531290925095e-05, + "loss": 0.8528, + "step": 43192 + }, + { + "epoch": 2.78, + "learning_rate": 1.075291440348655e-05, + "loss": 0.8839, + "step": 43193 + }, + { + "epoch": 2.78, + "learning_rate": 1.07526997144636e-05, + "loss": 0.8723, + "step": 43194 + }, + { + "epoch": 2.78, + "learning_rate": 1.0752485025440651e-05, + "loss": 0.8745, + "step": 43195 + }, + { + "epoch": 2.78, + "learning_rate": 1.07522703364177e-05, + "loss": 0.8559, + "step": 43196 + }, + { + "epoch": 2.78, + "learning_rate": 1.0752055647394751e-05, + "loss": 0.8591, + "step": 43197 + }, + { + "epoch": 2.78, + "learning_rate": 1.07518409583718e-05, + "loss": 0.8339, + "step": 43198 + }, + { + "epoch": 2.78, + "learning_rate": 1.0751626269348849e-05, + "loss": 0.8796, + "step": 43199 + }, + { + "epoch": 2.78, + "learning_rate": 1.07514115803259e-05, + "loss": 0.9045, + "step": 43200 + }, + { + "epoch": 2.78, + "learning_rate": 1.0751196891302948e-05, + "loss": 0.8831, + "step": 43201 + }, + { + "epoch": 2.78, + "learning_rate": 1.0750982202279997e-05, + "loss": 0.8754, + "step": 43202 + }, + { + "epoch": 2.78, + "learning_rate": 1.0750767513257048e-05, + "loss": 0.9167, + "step": 43203 + }, + { + "epoch": 2.78, + "learning_rate": 1.0750552824234097e-05, + "loss": 0.8814, + "step": 43204 + }, + { + "epoch": 2.78, + "learning_rate": 1.0750338135211148e-05, + "loss": 0.812, + "step": 43205 + }, + { + "epoch": 2.78, + "learning_rate": 1.0750123446188197e-05, + "loss": 0.924, + "step": 43206 + }, + { + "epoch": 2.78, + "learning_rate": 1.0749908757165246e-05, + "loss": 0.9204, + "step": 43207 + }, + { + "epoch": 2.78, + "learning_rate": 1.0749694068142296e-05, + "loss": 0.9204, + "step": 43208 + }, + { + "epoch": 2.78, + "learning_rate": 1.0749479379119347e-05, + "loss": 0.9554, + "step": 43209 + }, + { + "epoch": 2.78, + "learning_rate": 1.0749264690096398e-05, + "loss": 0.9196, + "step": 43210 + }, + { + "epoch": 2.78, + "learning_rate": 1.0749050001073447e-05, + "loss": 0.8213, + "step": 43211 + }, + { + "epoch": 2.78, + "learning_rate": 1.0748835312050496e-05, + "loss": 0.9442, + "step": 43212 + }, + { + "epoch": 2.78, + "learning_rate": 1.0748620623027546e-05, + "loss": 0.8708, + "step": 43213 + }, + { + "epoch": 2.78, + "learning_rate": 1.0748405934004595e-05, + "loss": 0.8594, + "step": 43214 + }, + { + "epoch": 2.78, + "learning_rate": 1.0748191244981646e-05, + "loss": 0.8815, + "step": 43215 + }, + { + "epoch": 2.78, + "learning_rate": 1.0747976555958695e-05, + "loss": 0.894, + "step": 43216 + }, + { + "epoch": 2.78, + "learning_rate": 1.0747761866935744e-05, + "loss": 0.9017, + "step": 43217 + }, + { + "epoch": 2.78, + "learning_rate": 1.0747547177912794e-05, + "loss": 0.8772, + "step": 43218 + }, + { + "epoch": 2.78, + "learning_rate": 1.0747332488889843e-05, + "loss": 0.8942, + "step": 43219 + }, + { + "epoch": 2.78, + "learning_rate": 1.0747117799866892e-05, + "loss": 0.9234, + "step": 43220 + }, + { + "epoch": 2.78, + "learning_rate": 1.0746903110843943e-05, + "loss": 0.9186, + "step": 43221 + }, + { + "epoch": 2.78, + "learning_rate": 1.0746688421820992e-05, + "loss": 0.8733, + "step": 43222 + }, + { + "epoch": 2.78, + "learning_rate": 1.0746473732798044e-05, + "loss": 0.9038, + "step": 43223 + }, + { + "epoch": 2.78, + "learning_rate": 1.0746259043775093e-05, + "loss": 0.9306, + "step": 43224 + }, + { + "epoch": 2.78, + "learning_rate": 1.0746044354752144e-05, + "loss": 0.8991, + "step": 43225 + }, + { + "epoch": 2.78, + "learning_rate": 1.0745829665729193e-05, + "loss": 0.8723, + "step": 43226 + }, + { + "epoch": 2.78, + "learning_rate": 1.0745614976706242e-05, + "loss": 0.8956, + "step": 43227 + }, + { + "epoch": 2.78, + "learning_rate": 1.0745400287683293e-05, + "loss": 0.9788, + "step": 43228 + }, + { + "epoch": 2.78, + "learning_rate": 1.0745185598660341e-05, + "loss": 0.9155, + "step": 43229 + }, + { + "epoch": 2.78, + "learning_rate": 1.074497090963739e-05, + "loss": 0.9342, + "step": 43230 + }, + { + "epoch": 2.78, + "learning_rate": 1.0744756220614441e-05, + "loss": 0.9707, + "step": 43231 + }, + { + "epoch": 2.78, + "learning_rate": 1.074454153159149e-05, + "loss": 0.8915, + "step": 43232 + }, + { + "epoch": 2.78, + "learning_rate": 1.0744326842568539e-05, + "loss": 0.8203, + "step": 43233 + }, + { + "epoch": 2.78, + "learning_rate": 1.074411215354559e-05, + "loss": 0.9369, + "step": 43234 + }, + { + "epoch": 2.78, + "learning_rate": 1.0743897464522639e-05, + "loss": 0.896, + "step": 43235 + }, + { + "epoch": 2.78, + "learning_rate": 1.074368277549969e-05, + "loss": 0.9266, + "step": 43236 + }, + { + "epoch": 2.78, + "learning_rate": 1.0743468086476738e-05, + "loss": 0.8135, + "step": 43237 + }, + { + "epoch": 2.78, + "learning_rate": 1.074325339745379e-05, + "loss": 0.832, + "step": 43238 + }, + { + "epoch": 2.78, + "learning_rate": 1.074303870843084e-05, + "loss": 0.8951, + "step": 43239 + }, + { + "epoch": 2.78, + "learning_rate": 1.0742824019407889e-05, + "loss": 0.8519, + "step": 43240 + }, + { + "epoch": 2.78, + "learning_rate": 1.074260933038494e-05, + "loss": 0.8591, + "step": 43241 + }, + { + "epoch": 2.78, + "learning_rate": 1.0742394641361988e-05, + "loss": 0.8909, + "step": 43242 + }, + { + "epoch": 2.78, + "learning_rate": 1.0742179952339037e-05, + "loss": 0.8472, + "step": 43243 + }, + { + "epoch": 2.78, + "learning_rate": 1.0741965263316088e-05, + "loss": 0.8939, + "step": 43244 + }, + { + "epoch": 2.78, + "learning_rate": 1.0741750574293137e-05, + "loss": 0.8452, + "step": 43245 + }, + { + "epoch": 2.78, + "learning_rate": 1.0741535885270187e-05, + "loss": 0.8817, + "step": 43246 + }, + { + "epoch": 2.78, + "learning_rate": 1.0741321196247236e-05, + "loss": 0.9745, + "step": 43247 + }, + { + "epoch": 2.78, + "learning_rate": 1.0741106507224285e-05, + "loss": 0.8761, + "step": 43248 + }, + { + "epoch": 2.78, + "learning_rate": 1.0740891818201336e-05, + "loss": 0.8762, + "step": 43249 + }, + { + "epoch": 2.78, + "learning_rate": 1.0740677129178385e-05, + "loss": 0.8815, + "step": 43250 + }, + { + "epoch": 2.78, + "learning_rate": 1.0740462440155434e-05, + "loss": 0.8737, + "step": 43251 + }, + { + "epoch": 2.78, + "learning_rate": 1.0740247751132486e-05, + "loss": 0.9619, + "step": 43252 + }, + { + "epoch": 2.78, + "learning_rate": 1.0740033062109535e-05, + "loss": 0.9408, + "step": 43253 + }, + { + "epoch": 2.78, + "learning_rate": 1.0739818373086586e-05, + "loss": 0.9062, + "step": 43254 + }, + { + "epoch": 2.78, + "learning_rate": 1.0739603684063635e-05, + "loss": 0.8976, + "step": 43255 + }, + { + "epoch": 2.78, + "learning_rate": 1.0739388995040686e-05, + "loss": 0.9012, + "step": 43256 + }, + { + "epoch": 2.78, + "learning_rate": 1.0739174306017735e-05, + "loss": 0.8814, + "step": 43257 + }, + { + "epoch": 2.78, + "learning_rate": 1.0738959616994783e-05, + "loss": 0.9094, + "step": 43258 + }, + { + "epoch": 2.78, + "learning_rate": 1.0738744927971834e-05, + "loss": 0.8531, + "step": 43259 + }, + { + "epoch": 2.78, + "learning_rate": 1.0738530238948883e-05, + "loss": 0.904, + "step": 43260 + }, + { + "epoch": 2.78, + "learning_rate": 1.0738315549925932e-05, + "loss": 0.9407, + "step": 43261 + }, + { + "epoch": 2.78, + "learning_rate": 1.0738100860902983e-05, + "loss": 0.8565, + "step": 43262 + }, + { + "epoch": 2.78, + "learning_rate": 1.0737886171880032e-05, + "loss": 0.9111, + "step": 43263 + }, + { + "epoch": 2.78, + "learning_rate": 1.0737671482857082e-05, + "loss": 0.8995, + "step": 43264 + }, + { + "epoch": 2.78, + "learning_rate": 1.0737456793834131e-05, + "loss": 0.8482, + "step": 43265 + }, + { + "epoch": 2.78, + "learning_rate": 1.073724210481118e-05, + "loss": 0.9169, + "step": 43266 + }, + { + "epoch": 2.78, + "learning_rate": 1.0737027415788233e-05, + "loss": 0.8502, + "step": 43267 + }, + { + "epoch": 2.78, + "learning_rate": 1.0736812726765282e-05, + "loss": 0.9038, + "step": 43268 + }, + { + "epoch": 2.78, + "learning_rate": 1.0736598037742332e-05, + "loss": 0.8695, + "step": 43269 + }, + { + "epoch": 2.78, + "learning_rate": 1.0736383348719381e-05, + "loss": 0.8797, + "step": 43270 + }, + { + "epoch": 2.78, + "learning_rate": 1.073616865969643e-05, + "loss": 0.8929, + "step": 43271 + }, + { + "epoch": 2.78, + "learning_rate": 1.0735953970673481e-05, + "loss": 0.8925, + "step": 43272 + }, + { + "epoch": 2.78, + "learning_rate": 1.073573928165053e-05, + "loss": 0.9175, + "step": 43273 + }, + { + "epoch": 2.78, + "learning_rate": 1.073552459262758e-05, + "loss": 0.9604, + "step": 43274 + }, + { + "epoch": 2.78, + "learning_rate": 1.073530990360463e-05, + "loss": 0.9039, + "step": 43275 + }, + { + "epoch": 2.78, + "learning_rate": 1.0735095214581678e-05, + "loss": 0.9185, + "step": 43276 + }, + { + "epoch": 2.78, + "learning_rate": 1.0734880525558729e-05, + "loss": 0.8953, + "step": 43277 + }, + { + "epoch": 2.78, + "learning_rate": 1.0734665836535778e-05, + "loss": 0.9121, + "step": 43278 + }, + { + "epoch": 2.78, + "learning_rate": 1.0734451147512827e-05, + "loss": 0.8586, + "step": 43279 + }, + { + "epoch": 2.78, + "learning_rate": 1.0734236458489878e-05, + "loss": 0.8646, + "step": 43280 + }, + { + "epoch": 2.78, + "learning_rate": 1.0734021769466928e-05, + "loss": 0.8916, + "step": 43281 + }, + { + "epoch": 2.78, + "learning_rate": 1.0733807080443979e-05, + "loss": 0.8594, + "step": 43282 + }, + { + "epoch": 2.78, + "learning_rate": 1.0733592391421028e-05, + "loss": 0.9225, + "step": 43283 + }, + { + "epoch": 2.78, + "learning_rate": 1.0733377702398079e-05, + "loss": 0.9271, + "step": 43284 + }, + { + "epoch": 2.78, + "learning_rate": 1.0733163013375128e-05, + "loss": 0.8399, + "step": 43285 + }, + { + "epoch": 2.78, + "learning_rate": 1.0732948324352177e-05, + "loss": 0.8409, + "step": 43286 + }, + { + "epoch": 2.78, + "learning_rate": 1.0732733635329227e-05, + "loss": 0.8913, + "step": 43287 + }, + { + "epoch": 2.79, + "learning_rate": 1.0732518946306276e-05, + "loss": 0.8864, + "step": 43288 + }, + { + "epoch": 2.79, + "learning_rate": 1.0732304257283325e-05, + "loss": 0.8782, + "step": 43289 + }, + { + "epoch": 2.79, + "learning_rate": 1.0732089568260376e-05, + "loss": 0.8938, + "step": 43290 + }, + { + "epoch": 2.79, + "learning_rate": 1.0731874879237425e-05, + "loss": 0.8492, + "step": 43291 + }, + { + "epoch": 2.79, + "learning_rate": 1.0731660190214475e-05, + "loss": 0.9044, + "step": 43292 + }, + { + "epoch": 2.79, + "learning_rate": 1.0731445501191524e-05, + "loss": 0.8938, + "step": 43293 + }, + { + "epoch": 2.79, + "learning_rate": 1.0731230812168573e-05, + "loss": 0.8982, + "step": 43294 + }, + { + "epoch": 2.79, + "learning_rate": 1.0731016123145624e-05, + "loss": 0.9134, + "step": 43295 + }, + { + "epoch": 2.79, + "learning_rate": 1.0730801434122675e-05, + "loss": 0.874, + "step": 43296 + }, + { + "epoch": 2.79, + "learning_rate": 1.0730586745099725e-05, + "loss": 0.8459, + "step": 43297 + }, + { + "epoch": 2.79, + "learning_rate": 1.0730372056076774e-05, + "loss": 0.9431, + "step": 43298 + }, + { + "epoch": 2.79, + "learning_rate": 1.0730157367053823e-05, + "loss": 0.8954, + "step": 43299 + }, + { + "epoch": 2.79, + "learning_rate": 1.0729942678030874e-05, + "loss": 0.8718, + "step": 43300 + }, + { + "epoch": 2.79, + "learning_rate": 1.0729727989007923e-05, + "loss": 0.9058, + "step": 43301 + }, + { + "epoch": 2.79, + "learning_rate": 1.0729513299984974e-05, + "loss": 0.9085, + "step": 43302 + }, + { + "epoch": 2.79, + "learning_rate": 1.0729298610962022e-05, + "loss": 0.8636, + "step": 43303 + }, + { + "epoch": 2.79, + "learning_rate": 1.0729083921939071e-05, + "loss": 0.8394, + "step": 43304 + }, + { + "epoch": 2.79, + "learning_rate": 1.0728869232916122e-05, + "loss": 0.8415, + "step": 43305 + }, + { + "epoch": 2.79, + "learning_rate": 1.0728654543893171e-05, + "loss": 0.8657, + "step": 43306 + }, + { + "epoch": 2.79, + "learning_rate": 1.072843985487022e-05, + "loss": 0.8536, + "step": 43307 + }, + { + "epoch": 2.79, + "learning_rate": 1.072822516584727e-05, + "loss": 0.931, + "step": 43308 + }, + { + "epoch": 2.79, + "learning_rate": 1.072801047682432e-05, + "loss": 0.8672, + "step": 43309 + }, + { + "epoch": 2.79, + "learning_rate": 1.0727795787801372e-05, + "loss": 0.8944, + "step": 43310 + }, + { + "epoch": 2.79, + "learning_rate": 1.0727581098778421e-05, + "loss": 0.8302, + "step": 43311 + }, + { + "epoch": 2.79, + "learning_rate": 1.0727366409755472e-05, + "loss": 0.8786, + "step": 43312 + }, + { + "epoch": 2.79, + "learning_rate": 1.072715172073252e-05, + "loss": 0.8592, + "step": 43313 + }, + { + "epoch": 2.79, + "learning_rate": 1.072693703170957e-05, + "loss": 0.879, + "step": 43314 + }, + { + "epoch": 2.79, + "learning_rate": 1.072672234268662e-05, + "loss": 0.8474, + "step": 43315 + }, + { + "epoch": 2.79, + "learning_rate": 1.072650765366367e-05, + "loss": 0.8248, + "step": 43316 + }, + { + "epoch": 2.79, + "learning_rate": 1.0726292964640718e-05, + "loss": 0.9707, + "step": 43317 + }, + { + "epoch": 2.79, + "learning_rate": 1.0726078275617769e-05, + "loss": 0.8661, + "step": 43318 + }, + { + "epoch": 2.79, + "learning_rate": 1.0725863586594818e-05, + "loss": 0.9451, + "step": 43319 + }, + { + "epoch": 2.79, + "learning_rate": 1.0725648897571867e-05, + "loss": 0.9243, + "step": 43320 + }, + { + "epoch": 2.79, + "learning_rate": 1.0725434208548917e-05, + "loss": 0.8409, + "step": 43321 + }, + { + "epoch": 2.79, + "learning_rate": 1.0725219519525966e-05, + "loss": 0.925, + "step": 43322 + }, + { + "epoch": 2.79, + "learning_rate": 1.0725004830503017e-05, + "loss": 0.9067, + "step": 43323 + }, + { + "epoch": 2.79, + "learning_rate": 1.0724790141480066e-05, + "loss": 0.9395, + "step": 43324 + }, + { + "epoch": 2.79, + "learning_rate": 1.0724575452457118e-05, + "loss": 0.8401, + "step": 43325 + }, + { + "epoch": 2.79, + "learning_rate": 1.0724360763434167e-05, + "loss": 0.9249, + "step": 43326 + }, + { + "epoch": 2.79, + "learning_rate": 1.0724146074411216e-05, + "loss": 0.8478, + "step": 43327 + }, + { + "epoch": 2.79, + "learning_rate": 1.0723931385388267e-05, + "loss": 0.8817, + "step": 43328 + }, + { + "epoch": 2.79, + "learning_rate": 1.0723716696365316e-05, + "loss": 0.9381, + "step": 43329 + }, + { + "epoch": 2.79, + "learning_rate": 1.0723502007342365e-05, + "loss": 0.8711, + "step": 43330 + }, + { + "epoch": 2.79, + "learning_rate": 1.0723287318319416e-05, + "loss": 0.8679, + "step": 43331 + }, + { + "epoch": 2.79, + "learning_rate": 1.0723072629296465e-05, + "loss": 0.8582, + "step": 43332 + }, + { + "epoch": 2.79, + "learning_rate": 1.0722857940273515e-05, + "loss": 0.8661, + "step": 43333 + }, + { + "epoch": 2.79, + "learning_rate": 1.0722643251250564e-05, + "loss": 0.9246, + "step": 43334 + }, + { + "epoch": 2.79, + "learning_rate": 1.0722428562227613e-05, + "loss": 0.8686, + "step": 43335 + }, + { + "epoch": 2.79, + "learning_rate": 1.0722213873204664e-05, + "loss": 0.8887, + "step": 43336 + }, + { + "epoch": 2.79, + "learning_rate": 1.0721999184181713e-05, + "loss": 0.8167, + "step": 43337 + }, + { + "epoch": 2.79, + "learning_rate": 1.0721784495158762e-05, + "loss": 0.8608, + "step": 43338 + }, + { + "epoch": 2.79, + "learning_rate": 1.0721569806135814e-05, + "loss": 0.8792, + "step": 43339 + }, + { + "epoch": 2.79, + "learning_rate": 1.0721355117112863e-05, + "loss": 0.8867, + "step": 43340 + }, + { + "epoch": 2.79, + "learning_rate": 1.0721140428089914e-05, + "loss": 0.9942, + "step": 43341 + }, + { + "epoch": 2.79, + "learning_rate": 1.0720925739066963e-05, + "loss": 0.8638, + "step": 43342 + }, + { + "epoch": 2.79, + "learning_rate": 1.0720711050044013e-05, + "loss": 0.9021, + "step": 43343 + }, + { + "epoch": 2.79, + "learning_rate": 1.0720496361021062e-05, + "loss": 0.9002, + "step": 43344 + }, + { + "epoch": 2.79, + "learning_rate": 1.0720281671998111e-05, + "loss": 0.935, + "step": 43345 + }, + { + "epoch": 2.79, + "learning_rate": 1.0720066982975162e-05, + "loss": 0.9222, + "step": 43346 + }, + { + "epoch": 2.79, + "learning_rate": 1.071985229395221e-05, + "loss": 0.8625, + "step": 43347 + }, + { + "epoch": 2.79, + "learning_rate": 1.071963760492926e-05, + "loss": 0.8972, + "step": 43348 + }, + { + "epoch": 2.79, + "learning_rate": 1.071942291590631e-05, + "loss": 0.9648, + "step": 43349 + }, + { + "epoch": 2.79, + "learning_rate": 1.071920822688336e-05, + "loss": 0.9143, + "step": 43350 + }, + { + "epoch": 2.79, + "learning_rate": 1.071899353786041e-05, + "loss": 0.905, + "step": 43351 + }, + { + "epoch": 2.79, + "learning_rate": 1.0718778848837459e-05, + "loss": 0.909, + "step": 43352 + }, + { + "epoch": 2.79, + "learning_rate": 1.0718564159814508e-05, + "loss": 0.9739, + "step": 43353 + }, + { + "epoch": 2.79, + "learning_rate": 1.071834947079156e-05, + "loss": 0.8889, + "step": 43354 + }, + { + "epoch": 2.79, + "learning_rate": 1.071813478176861e-05, + "loss": 0.953, + "step": 43355 + }, + { + "epoch": 2.79, + "learning_rate": 1.071792009274566e-05, + "loss": 0.9646, + "step": 43356 + }, + { + "epoch": 2.79, + "learning_rate": 1.0717705403722709e-05, + "loss": 0.903, + "step": 43357 + }, + { + "epoch": 2.79, + "learning_rate": 1.0717490714699758e-05, + "loss": 0.9277, + "step": 43358 + }, + { + "epoch": 2.79, + "learning_rate": 1.0717276025676809e-05, + "loss": 0.8616, + "step": 43359 + }, + { + "epoch": 2.79, + "learning_rate": 1.0717061336653858e-05, + "loss": 0.8425, + "step": 43360 + }, + { + "epoch": 2.79, + "learning_rate": 1.0716846647630908e-05, + "loss": 0.9419, + "step": 43361 + }, + { + "epoch": 2.79, + "learning_rate": 1.0716631958607957e-05, + "loss": 0.972, + "step": 43362 + }, + { + "epoch": 2.79, + "learning_rate": 1.0716417269585006e-05, + "loss": 0.8603, + "step": 43363 + }, + { + "epoch": 2.79, + "learning_rate": 1.0716202580562057e-05, + "loss": 0.8601, + "step": 43364 + }, + { + "epoch": 2.79, + "learning_rate": 1.0715987891539106e-05, + "loss": 0.8844, + "step": 43365 + }, + { + "epoch": 2.79, + "learning_rate": 1.0715773202516155e-05, + "loss": 0.8689, + "step": 43366 + }, + { + "epoch": 2.79, + "learning_rate": 1.0715558513493205e-05, + "loss": 0.9234, + "step": 43367 + }, + { + "epoch": 2.79, + "learning_rate": 1.0715343824470256e-05, + "loss": 0.8601, + "step": 43368 + }, + { + "epoch": 2.79, + "learning_rate": 1.0715129135447307e-05, + "loss": 0.8843, + "step": 43369 + }, + { + "epoch": 2.79, + "learning_rate": 1.0714914446424356e-05, + "loss": 0.918, + "step": 43370 + }, + { + "epoch": 2.79, + "learning_rate": 1.0714699757401406e-05, + "loss": 0.8277, + "step": 43371 + }, + { + "epoch": 2.79, + "learning_rate": 1.0714485068378455e-05, + "loss": 0.9684, + "step": 43372 + }, + { + "epoch": 2.79, + "learning_rate": 1.0714270379355504e-05, + "loss": 0.9257, + "step": 43373 + }, + { + "epoch": 2.79, + "learning_rate": 1.0714055690332555e-05, + "loss": 0.8193, + "step": 43374 + }, + { + "epoch": 2.79, + "learning_rate": 1.0713841001309604e-05, + "loss": 0.9043, + "step": 43375 + }, + { + "epoch": 2.79, + "learning_rate": 1.0713626312286653e-05, + "loss": 0.923, + "step": 43376 + }, + { + "epoch": 2.79, + "learning_rate": 1.0713411623263704e-05, + "loss": 0.8863, + "step": 43377 + }, + { + "epoch": 2.79, + "learning_rate": 1.0713196934240752e-05, + "loss": 0.8788, + "step": 43378 + }, + { + "epoch": 2.79, + "learning_rate": 1.0712982245217803e-05, + "loss": 0.9006, + "step": 43379 + }, + { + "epoch": 2.79, + "learning_rate": 1.0712767556194852e-05, + "loss": 0.873, + "step": 43380 + }, + { + "epoch": 2.79, + "learning_rate": 1.0712552867171901e-05, + "loss": 0.8599, + "step": 43381 + }, + { + "epoch": 2.79, + "learning_rate": 1.0712338178148952e-05, + "loss": 0.8079, + "step": 43382 + }, + { + "epoch": 2.79, + "learning_rate": 1.0712123489126002e-05, + "loss": 0.84, + "step": 43383 + }, + { + "epoch": 2.79, + "learning_rate": 1.0711908800103053e-05, + "loss": 0.9319, + "step": 43384 + }, + { + "epoch": 2.79, + "learning_rate": 1.0711694111080102e-05, + "loss": 0.8944, + "step": 43385 + }, + { + "epoch": 2.79, + "learning_rate": 1.0711479422057151e-05, + "loss": 0.88, + "step": 43386 + }, + { + "epoch": 2.79, + "learning_rate": 1.0711264733034202e-05, + "loss": 0.8933, + "step": 43387 + }, + { + "epoch": 2.79, + "learning_rate": 1.071105004401125e-05, + "loss": 0.9113, + "step": 43388 + }, + { + "epoch": 2.79, + "learning_rate": 1.0710835354988301e-05, + "loss": 0.8807, + "step": 43389 + }, + { + "epoch": 2.79, + "learning_rate": 1.071062066596535e-05, + "loss": 0.8178, + "step": 43390 + }, + { + "epoch": 2.79, + "learning_rate": 1.07104059769424e-05, + "loss": 0.8596, + "step": 43391 + }, + { + "epoch": 2.79, + "learning_rate": 1.071019128791945e-05, + "loss": 0.9052, + "step": 43392 + }, + { + "epoch": 2.79, + "learning_rate": 1.0709976598896499e-05, + "loss": 0.8835, + "step": 43393 + }, + { + "epoch": 2.79, + "learning_rate": 1.0709761909873548e-05, + "loss": 0.8775, + "step": 43394 + }, + { + "epoch": 2.79, + "learning_rate": 1.0709547220850598e-05, + "loss": 0.9996, + "step": 43395 + }, + { + "epoch": 2.79, + "learning_rate": 1.0709332531827647e-05, + "loss": 0.9175, + "step": 43396 + }, + { + "epoch": 2.79, + "learning_rate": 1.07091178428047e-05, + "loss": 0.9016, + "step": 43397 + }, + { + "epoch": 2.79, + "learning_rate": 1.0708903153781749e-05, + "loss": 0.8941, + "step": 43398 + }, + { + "epoch": 2.79, + "learning_rate": 1.07086884647588e-05, + "loss": 0.8703, + "step": 43399 + }, + { + "epoch": 2.79, + "learning_rate": 1.0708473775735848e-05, + "loss": 0.9168, + "step": 43400 + }, + { + "epoch": 2.79, + "learning_rate": 1.0708259086712897e-05, + "loss": 0.9124, + "step": 43401 + }, + { + "epoch": 2.79, + "learning_rate": 1.0708044397689948e-05, + "loss": 0.9208, + "step": 43402 + }, + { + "epoch": 2.79, + "learning_rate": 1.0707829708666997e-05, + "loss": 0.8671, + "step": 43403 + }, + { + "epoch": 2.79, + "learning_rate": 1.0707615019644046e-05, + "loss": 0.8968, + "step": 43404 + }, + { + "epoch": 2.79, + "learning_rate": 1.0707400330621097e-05, + "loss": 0.9299, + "step": 43405 + }, + { + "epoch": 2.79, + "learning_rate": 1.0707185641598146e-05, + "loss": 0.8826, + "step": 43406 + }, + { + "epoch": 2.79, + "learning_rate": 1.0706970952575194e-05, + "loss": 0.9314, + "step": 43407 + }, + { + "epoch": 2.79, + "learning_rate": 1.0706756263552245e-05, + "loss": 0.835, + "step": 43408 + }, + { + "epoch": 2.79, + "learning_rate": 1.0706541574529294e-05, + "loss": 0.9154, + "step": 43409 + }, + { + "epoch": 2.79, + "learning_rate": 1.0706326885506345e-05, + "loss": 0.9036, + "step": 43410 + }, + { + "epoch": 2.79, + "learning_rate": 1.0706112196483394e-05, + "loss": 0.9354, + "step": 43411 + }, + { + "epoch": 2.79, + "learning_rate": 1.0705897507460446e-05, + "loss": 0.8953, + "step": 43412 + }, + { + "epoch": 2.79, + "learning_rate": 1.0705682818437495e-05, + "loss": 0.9235, + "step": 43413 + }, + { + "epoch": 2.79, + "learning_rate": 1.0705468129414544e-05, + "loss": 0.933, + "step": 43414 + }, + { + "epoch": 2.79, + "learning_rate": 1.0705253440391595e-05, + "loss": 0.9684, + "step": 43415 + }, + { + "epoch": 2.79, + "learning_rate": 1.0705038751368644e-05, + "loss": 0.9047, + "step": 43416 + }, + { + "epoch": 2.79, + "learning_rate": 1.0704824062345693e-05, + "loss": 0.8774, + "step": 43417 + }, + { + "epoch": 2.79, + "learning_rate": 1.0704609373322743e-05, + "loss": 0.921, + "step": 43418 + }, + { + "epoch": 2.79, + "learning_rate": 1.0704394684299792e-05, + "loss": 0.858, + "step": 43419 + }, + { + "epoch": 2.79, + "learning_rate": 1.0704179995276843e-05, + "loss": 0.8694, + "step": 43420 + }, + { + "epoch": 2.79, + "learning_rate": 1.0703965306253892e-05, + "loss": 0.9046, + "step": 43421 + }, + { + "epoch": 2.79, + "learning_rate": 1.070375061723094e-05, + "loss": 0.8914, + "step": 43422 + }, + { + "epoch": 2.79, + "learning_rate": 1.0703535928207991e-05, + "loss": 0.9174, + "step": 43423 + }, + { + "epoch": 2.79, + "learning_rate": 1.070332123918504e-05, + "loss": 0.8403, + "step": 43424 + }, + { + "epoch": 2.79, + "learning_rate": 1.070310655016209e-05, + "loss": 0.8684, + "step": 43425 + }, + { + "epoch": 2.79, + "learning_rate": 1.0702891861139142e-05, + "loss": 1.0, + "step": 43426 + }, + { + "epoch": 2.79, + "learning_rate": 1.070267717211619e-05, + "loss": 0.9398, + "step": 43427 + }, + { + "epoch": 2.79, + "learning_rate": 1.0702462483093241e-05, + "loss": 0.841, + "step": 43428 + }, + { + "epoch": 2.79, + "learning_rate": 1.070224779407029e-05, + "loss": 0.9116, + "step": 43429 + }, + { + "epoch": 2.79, + "learning_rate": 1.0702033105047341e-05, + "loss": 0.8812, + "step": 43430 + }, + { + "epoch": 2.79, + "learning_rate": 1.070181841602439e-05, + "loss": 0.9694, + "step": 43431 + }, + { + "epoch": 2.79, + "learning_rate": 1.0701603727001439e-05, + "loss": 0.8457, + "step": 43432 + }, + { + "epoch": 2.79, + "learning_rate": 1.070138903797849e-05, + "loss": 0.8579, + "step": 43433 + }, + { + "epoch": 2.79, + "learning_rate": 1.0701174348955539e-05, + "loss": 0.892, + "step": 43434 + }, + { + "epoch": 2.79, + "learning_rate": 1.0700959659932588e-05, + "loss": 0.8701, + "step": 43435 + }, + { + "epoch": 2.79, + "learning_rate": 1.0700744970909638e-05, + "loss": 0.8763, + "step": 43436 + }, + { + "epoch": 2.79, + "learning_rate": 1.0700530281886687e-05, + "loss": 0.8563, + "step": 43437 + }, + { + "epoch": 2.79, + "learning_rate": 1.0700315592863738e-05, + "loss": 0.84, + "step": 43438 + }, + { + "epoch": 2.79, + "learning_rate": 1.0700100903840787e-05, + "loss": 0.9237, + "step": 43439 + }, + { + "epoch": 2.79, + "learning_rate": 1.0699886214817836e-05, + "loss": 0.9033, + "step": 43440 + }, + { + "epoch": 2.79, + "learning_rate": 1.0699671525794888e-05, + "loss": 0.9393, + "step": 43441 + }, + { + "epoch": 2.79, + "learning_rate": 1.0699456836771937e-05, + "loss": 0.8803, + "step": 43442 + }, + { + "epoch": 2.8, + "learning_rate": 1.0699242147748988e-05, + "loss": 0.9524, + "step": 43443 + }, + { + "epoch": 2.8, + "learning_rate": 1.0699027458726037e-05, + "loss": 0.8978, + "step": 43444 + }, + { + "epoch": 2.8, + "learning_rate": 1.0698812769703086e-05, + "loss": 0.8908, + "step": 43445 + }, + { + "epoch": 2.8, + "learning_rate": 1.0698598080680136e-05, + "loss": 0.9262, + "step": 43446 + }, + { + "epoch": 2.8, + "learning_rate": 1.0698383391657185e-05, + "loss": 0.8245, + "step": 43447 + }, + { + "epoch": 2.8, + "learning_rate": 1.0698168702634236e-05, + "loss": 0.9038, + "step": 43448 + }, + { + "epoch": 2.8, + "learning_rate": 1.0697954013611285e-05, + "loss": 0.8574, + "step": 43449 + }, + { + "epoch": 2.8, + "learning_rate": 1.0697739324588334e-05, + "loss": 0.9097, + "step": 43450 + }, + { + "epoch": 2.8, + "learning_rate": 1.0697524635565385e-05, + "loss": 0.8599, + "step": 43451 + }, + { + "epoch": 2.8, + "learning_rate": 1.0697309946542433e-05, + "loss": 0.8938, + "step": 43452 + }, + { + "epoch": 2.8, + "learning_rate": 1.0697095257519482e-05, + "loss": 0.8991, + "step": 43453 + }, + { + "epoch": 2.8, + "learning_rate": 1.0696880568496533e-05, + "loss": 0.9077, + "step": 43454 + }, + { + "epoch": 2.8, + "learning_rate": 1.0696665879473584e-05, + "loss": 0.9569, + "step": 43455 + }, + { + "epoch": 2.8, + "learning_rate": 1.0696451190450634e-05, + "loss": 0.9183, + "step": 43456 + }, + { + "epoch": 2.8, + "learning_rate": 1.0696236501427683e-05, + "loss": 0.9167, + "step": 43457 + }, + { + "epoch": 2.8, + "learning_rate": 1.0696021812404734e-05, + "loss": 0.9279, + "step": 43458 + }, + { + "epoch": 2.8, + "learning_rate": 1.0695807123381783e-05, + "loss": 0.9226, + "step": 43459 + }, + { + "epoch": 2.8, + "learning_rate": 1.0695592434358832e-05, + "loss": 0.8847, + "step": 43460 + }, + { + "epoch": 2.8, + "learning_rate": 1.0695377745335883e-05, + "loss": 0.8909, + "step": 43461 + }, + { + "epoch": 2.8, + "learning_rate": 1.0695163056312932e-05, + "loss": 0.8597, + "step": 43462 + }, + { + "epoch": 2.8, + "learning_rate": 1.069494836728998e-05, + "loss": 0.9283, + "step": 43463 + }, + { + "epoch": 2.8, + "learning_rate": 1.0694733678267031e-05, + "loss": 0.8098, + "step": 43464 + }, + { + "epoch": 2.8, + "learning_rate": 1.069451898924408e-05, + "loss": 0.8422, + "step": 43465 + }, + { + "epoch": 2.8, + "learning_rate": 1.0694304300221129e-05, + "loss": 0.8532, + "step": 43466 + }, + { + "epoch": 2.8, + "learning_rate": 1.069408961119818e-05, + "loss": 0.9732, + "step": 43467 + }, + { + "epoch": 2.8, + "learning_rate": 1.0693874922175229e-05, + "loss": 0.8941, + "step": 43468 + }, + { + "epoch": 2.8, + "learning_rate": 1.069366023315228e-05, + "loss": 0.9428, + "step": 43469 + }, + { + "epoch": 2.8, + "learning_rate": 1.069344554412933e-05, + "loss": 0.9178, + "step": 43470 + }, + { + "epoch": 2.8, + "learning_rate": 1.069323085510638e-05, + "loss": 0.8723, + "step": 43471 + }, + { + "epoch": 2.8, + "learning_rate": 1.069301616608343e-05, + "loss": 0.8275, + "step": 43472 + }, + { + "epoch": 2.8, + "learning_rate": 1.0692801477060479e-05, + "loss": 0.934, + "step": 43473 + }, + { + "epoch": 2.8, + "learning_rate": 1.069258678803753e-05, + "loss": 0.7861, + "step": 43474 + }, + { + "epoch": 2.8, + "learning_rate": 1.0692372099014578e-05, + "loss": 0.9219, + "step": 43475 + }, + { + "epoch": 2.8, + "learning_rate": 1.0692157409991627e-05, + "loss": 0.9254, + "step": 43476 + }, + { + "epoch": 2.8, + "learning_rate": 1.0691942720968678e-05, + "loss": 0.9367, + "step": 43477 + }, + { + "epoch": 2.8, + "learning_rate": 1.0691728031945727e-05, + "loss": 0.903, + "step": 43478 + }, + { + "epoch": 2.8, + "learning_rate": 1.0691513342922778e-05, + "loss": 0.8979, + "step": 43479 + }, + { + "epoch": 2.8, + "learning_rate": 1.0691298653899827e-05, + "loss": 0.9561, + "step": 43480 + }, + { + "epoch": 2.8, + "learning_rate": 1.0691083964876875e-05, + "loss": 0.9384, + "step": 43481 + }, + { + "epoch": 2.8, + "learning_rate": 1.0690869275853926e-05, + "loss": 0.8968, + "step": 43482 + }, + { + "epoch": 2.8, + "learning_rate": 1.0690654586830975e-05, + "loss": 0.9056, + "step": 43483 + }, + { + "epoch": 2.8, + "learning_rate": 1.0690439897808027e-05, + "loss": 1.0017, + "step": 43484 + }, + { + "epoch": 2.8, + "learning_rate": 1.0690225208785076e-05, + "loss": 1.0017, + "step": 43485 + }, + { + "epoch": 2.8, + "learning_rate": 1.0690010519762125e-05, + "loss": 0.8626, + "step": 43486 + }, + { + "epoch": 2.8, + "learning_rate": 1.0689795830739176e-05, + "loss": 0.7825, + "step": 43487 + }, + { + "epoch": 2.8, + "learning_rate": 1.0689581141716225e-05, + "loss": 0.925, + "step": 43488 + }, + { + "epoch": 2.8, + "learning_rate": 1.0689366452693276e-05, + "loss": 0.8555, + "step": 43489 + }, + { + "epoch": 2.8, + "learning_rate": 1.0689151763670325e-05, + "loss": 0.8173, + "step": 43490 + }, + { + "epoch": 2.8, + "learning_rate": 1.0688937074647374e-05, + "loss": 0.8752, + "step": 43491 + }, + { + "epoch": 2.8, + "learning_rate": 1.0688722385624424e-05, + "loss": 0.8527, + "step": 43492 + }, + { + "epoch": 2.8, + "learning_rate": 1.0688507696601473e-05, + "loss": 0.8937, + "step": 43493 + }, + { + "epoch": 2.8, + "learning_rate": 1.0688293007578522e-05, + "loss": 0.9751, + "step": 43494 + }, + { + "epoch": 2.8, + "learning_rate": 1.0688078318555573e-05, + "loss": 0.8895, + "step": 43495 + }, + { + "epoch": 2.8, + "learning_rate": 1.0687863629532622e-05, + "loss": 0.9574, + "step": 43496 + }, + { + "epoch": 2.8, + "learning_rate": 1.0687648940509672e-05, + "loss": 0.8447, + "step": 43497 + }, + { + "epoch": 2.8, + "learning_rate": 1.0687434251486721e-05, + "loss": 0.8189, + "step": 43498 + }, + { + "epoch": 2.8, + "learning_rate": 1.0687219562463774e-05, + "loss": 0.8758, + "step": 43499 + }, + { + "epoch": 2.8, + "learning_rate": 1.0687004873440823e-05, + "loss": 0.9861, + "step": 43500 + }, + { + "epoch": 2.8, + "learning_rate": 1.0686790184417872e-05, + "loss": 0.8838, + "step": 43501 + }, + { + "epoch": 2.8, + "learning_rate": 1.0686575495394922e-05, + "loss": 0.8688, + "step": 43502 + }, + { + "epoch": 2.8, + "learning_rate": 1.0686360806371971e-05, + "loss": 0.9254, + "step": 43503 + }, + { + "epoch": 2.8, + "learning_rate": 1.068614611734902e-05, + "loss": 0.926, + "step": 43504 + }, + { + "epoch": 2.8, + "learning_rate": 1.0685931428326071e-05, + "loss": 0.9248, + "step": 43505 + }, + { + "epoch": 2.8, + "learning_rate": 1.068571673930312e-05, + "loss": 0.9084, + "step": 43506 + }, + { + "epoch": 2.8, + "learning_rate": 1.068550205028017e-05, + "loss": 0.8766, + "step": 43507 + }, + { + "epoch": 2.8, + "learning_rate": 1.068528736125722e-05, + "loss": 0.9059, + "step": 43508 + }, + { + "epoch": 2.8, + "learning_rate": 1.0685072672234269e-05, + "loss": 0.8913, + "step": 43509 + }, + { + "epoch": 2.8, + "learning_rate": 1.068485798321132e-05, + "loss": 0.8783, + "step": 43510 + }, + { + "epoch": 2.8, + "learning_rate": 1.0684643294188368e-05, + "loss": 0.8775, + "step": 43511 + }, + { + "epoch": 2.8, + "learning_rate": 1.0684428605165417e-05, + "loss": 0.8528, + "step": 43512 + }, + { + "epoch": 2.8, + "learning_rate": 1.068421391614247e-05, + "loss": 0.8776, + "step": 43513 + }, + { + "epoch": 2.8, + "learning_rate": 1.0683999227119518e-05, + "loss": 0.9393, + "step": 43514 + }, + { + "epoch": 2.8, + "learning_rate": 1.0683784538096569e-05, + "loss": 0.9587, + "step": 43515 + }, + { + "epoch": 2.8, + "learning_rate": 1.0683569849073618e-05, + "loss": 0.9461, + "step": 43516 + }, + { + "epoch": 2.8, + "learning_rate": 1.0683355160050669e-05, + "loss": 0.8108, + "step": 43517 + }, + { + "epoch": 2.8, + "learning_rate": 1.0683140471027718e-05, + "loss": 0.8845, + "step": 43518 + }, + { + "epoch": 2.8, + "learning_rate": 1.0682925782004767e-05, + "loss": 0.9295, + "step": 43519 + }, + { + "epoch": 2.8, + "learning_rate": 1.0682711092981817e-05, + "loss": 0.8725, + "step": 43520 + }, + { + "epoch": 2.8, + "learning_rate": 1.0682496403958866e-05, + "loss": 0.8608, + "step": 43521 + }, + { + "epoch": 2.8, + "learning_rate": 1.0682281714935915e-05, + "loss": 0.8578, + "step": 43522 + }, + { + "epoch": 2.8, + "learning_rate": 1.0682067025912966e-05, + "loss": 0.9108, + "step": 43523 + }, + { + "epoch": 2.8, + "learning_rate": 1.0681852336890015e-05, + "loss": 0.9254, + "step": 43524 + }, + { + "epoch": 2.8, + "learning_rate": 1.0681637647867066e-05, + "loss": 0.8981, + "step": 43525 + }, + { + "epoch": 2.8, + "learning_rate": 1.0681422958844114e-05, + "loss": 0.8642, + "step": 43526 + }, + { + "epoch": 2.8, + "learning_rate": 1.0681208269821163e-05, + "loss": 0.8699, + "step": 43527 + }, + { + "epoch": 2.8, + "learning_rate": 1.0680993580798216e-05, + "loss": 0.8557, + "step": 43528 + }, + { + "epoch": 2.8, + "learning_rate": 1.0680778891775265e-05, + "loss": 0.8773, + "step": 43529 + }, + { + "epoch": 2.8, + "learning_rate": 1.0680564202752315e-05, + "loss": 0.8633, + "step": 43530 + }, + { + "epoch": 2.8, + "learning_rate": 1.0680349513729364e-05, + "loss": 0.8336, + "step": 43531 + }, + { + "epoch": 2.8, + "learning_rate": 1.0680134824706413e-05, + "loss": 0.9121, + "step": 43532 + }, + { + "epoch": 2.8, + "learning_rate": 1.0679920135683464e-05, + "loss": 0.9125, + "step": 43533 + }, + { + "epoch": 2.8, + "learning_rate": 1.0679705446660513e-05, + "loss": 0.8592, + "step": 43534 + }, + { + "epoch": 2.8, + "learning_rate": 1.0679490757637564e-05, + "loss": 0.8726, + "step": 43535 + }, + { + "epoch": 2.8, + "learning_rate": 1.0679276068614613e-05, + "loss": 0.7869, + "step": 43536 + }, + { + "epoch": 2.8, + "learning_rate": 1.0679061379591662e-05, + "loss": 0.9194, + "step": 43537 + }, + { + "epoch": 2.8, + "learning_rate": 1.0678846690568712e-05, + "loss": 0.918, + "step": 43538 + }, + { + "epoch": 2.8, + "learning_rate": 1.0678632001545761e-05, + "loss": 0.9023, + "step": 43539 + }, + { + "epoch": 2.8, + "learning_rate": 1.067841731252281e-05, + "loss": 0.8864, + "step": 43540 + }, + { + "epoch": 2.8, + "learning_rate": 1.067820262349986e-05, + "loss": 0.8856, + "step": 43541 + }, + { + "epoch": 2.8, + "learning_rate": 1.0677987934476911e-05, + "loss": 0.9114, + "step": 43542 + }, + { + "epoch": 2.8, + "learning_rate": 1.0677773245453962e-05, + "loss": 0.8888, + "step": 43543 + }, + { + "epoch": 2.8, + "learning_rate": 1.0677558556431011e-05, + "loss": 0.8998, + "step": 43544 + }, + { + "epoch": 2.8, + "learning_rate": 1.0677343867408062e-05, + "loss": 0.9032, + "step": 43545 + }, + { + "epoch": 2.8, + "learning_rate": 1.067712917838511e-05, + "loss": 0.8961, + "step": 43546 + }, + { + "epoch": 2.8, + "learning_rate": 1.067691448936216e-05, + "loss": 0.8622, + "step": 43547 + }, + { + "epoch": 2.8, + "learning_rate": 1.067669980033921e-05, + "loss": 0.9372, + "step": 43548 + }, + { + "epoch": 2.8, + "learning_rate": 1.067648511131626e-05, + "loss": 0.9054, + "step": 43549 + }, + { + "epoch": 2.8, + "learning_rate": 1.0676270422293308e-05, + "loss": 0.9121, + "step": 43550 + }, + { + "epoch": 2.8, + "learning_rate": 1.0676055733270359e-05, + "loss": 0.9333, + "step": 43551 + }, + { + "epoch": 2.8, + "learning_rate": 1.0675841044247408e-05, + "loss": 0.8948, + "step": 43552 + }, + { + "epoch": 2.8, + "learning_rate": 1.0675626355224457e-05, + "loss": 0.925, + "step": 43553 + }, + { + "epoch": 2.8, + "learning_rate": 1.0675411666201508e-05, + "loss": 0.9127, + "step": 43554 + }, + { + "epoch": 2.8, + "learning_rate": 1.0675196977178556e-05, + "loss": 0.9148, + "step": 43555 + }, + { + "epoch": 2.8, + "learning_rate": 1.0674982288155607e-05, + "loss": 0.9185, + "step": 43556 + }, + { + "epoch": 2.8, + "learning_rate": 1.0674767599132658e-05, + "loss": 0.8953, + "step": 43557 + }, + { + "epoch": 2.8, + "learning_rate": 1.0674552910109708e-05, + "loss": 0.875, + "step": 43558 + }, + { + "epoch": 2.8, + "learning_rate": 1.0674338221086757e-05, + "loss": 0.8911, + "step": 43559 + }, + { + "epoch": 2.8, + "learning_rate": 1.0674123532063806e-05, + "loss": 0.8445, + "step": 43560 + }, + { + "epoch": 2.8, + "learning_rate": 1.0673908843040857e-05, + "loss": 0.8571, + "step": 43561 + }, + { + "epoch": 2.8, + "learning_rate": 1.0673694154017906e-05, + "loss": 0.8727, + "step": 43562 + }, + { + "epoch": 2.8, + "learning_rate": 1.0673479464994955e-05, + "loss": 0.9045, + "step": 43563 + }, + { + "epoch": 2.8, + "learning_rate": 1.0673264775972006e-05, + "loss": 0.9294, + "step": 43564 + }, + { + "epoch": 2.8, + "learning_rate": 1.0673050086949055e-05, + "loss": 0.8823, + "step": 43565 + }, + { + "epoch": 2.8, + "learning_rate": 1.0672835397926105e-05, + "loss": 0.927, + "step": 43566 + }, + { + "epoch": 2.8, + "learning_rate": 1.0672620708903154e-05, + "loss": 0.8518, + "step": 43567 + }, + { + "epoch": 2.8, + "learning_rate": 1.0672406019880203e-05, + "loss": 0.9578, + "step": 43568 + }, + { + "epoch": 2.8, + "learning_rate": 1.0672191330857254e-05, + "loss": 0.9206, + "step": 43569 + }, + { + "epoch": 2.8, + "learning_rate": 1.0671976641834303e-05, + "loss": 0.85, + "step": 43570 + }, + { + "epoch": 2.8, + "learning_rate": 1.0671761952811352e-05, + "loss": 0.9408, + "step": 43571 + }, + { + "epoch": 2.8, + "learning_rate": 1.0671547263788404e-05, + "loss": 0.8724, + "step": 43572 + }, + { + "epoch": 2.8, + "learning_rate": 1.0671332574765453e-05, + "loss": 0.9027, + "step": 43573 + }, + { + "epoch": 2.8, + "learning_rate": 1.0671117885742504e-05, + "loss": 0.9151, + "step": 43574 + }, + { + "epoch": 2.8, + "learning_rate": 1.0670903196719553e-05, + "loss": 0.8381, + "step": 43575 + }, + { + "epoch": 2.8, + "learning_rate": 1.0670688507696603e-05, + "loss": 0.9039, + "step": 43576 + }, + { + "epoch": 2.8, + "learning_rate": 1.0670473818673652e-05, + "loss": 0.9184, + "step": 43577 + }, + { + "epoch": 2.8, + "learning_rate": 1.0670259129650701e-05, + "loss": 0.8629, + "step": 43578 + }, + { + "epoch": 2.8, + "learning_rate": 1.0670044440627752e-05, + "loss": 0.87, + "step": 43579 + }, + { + "epoch": 2.8, + "learning_rate": 1.0669829751604801e-05, + "loss": 0.8575, + "step": 43580 + }, + { + "epoch": 2.8, + "learning_rate": 1.066961506258185e-05, + "loss": 0.945, + "step": 43581 + }, + { + "epoch": 2.8, + "learning_rate": 1.06694003735589e-05, + "loss": 0.9367, + "step": 43582 + }, + { + "epoch": 2.8, + "learning_rate": 1.066918568453595e-05, + "loss": 0.9088, + "step": 43583 + }, + { + "epoch": 2.8, + "learning_rate": 1.0668970995513e-05, + "loss": 0.8733, + "step": 43584 + }, + { + "epoch": 2.8, + "learning_rate": 1.0668756306490049e-05, + "loss": 0.8767, + "step": 43585 + }, + { + "epoch": 2.8, + "learning_rate": 1.0668541617467101e-05, + "loss": 0.9768, + "step": 43586 + }, + { + "epoch": 2.8, + "learning_rate": 1.066832692844415e-05, + "loss": 0.9018, + "step": 43587 + }, + { + "epoch": 2.8, + "learning_rate": 1.06681122394212e-05, + "loss": 0.9266, + "step": 43588 + }, + { + "epoch": 2.8, + "learning_rate": 1.066789755039825e-05, + "loss": 0.9329, + "step": 43589 + }, + { + "epoch": 2.8, + "learning_rate": 1.0667682861375299e-05, + "loss": 0.8974, + "step": 43590 + }, + { + "epoch": 2.8, + "learning_rate": 1.0667468172352348e-05, + "loss": 0.9461, + "step": 43591 + }, + { + "epoch": 2.8, + "learning_rate": 1.0667253483329399e-05, + "loss": 0.8361, + "step": 43592 + }, + { + "epoch": 2.8, + "learning_rate": 1.0667038794306448e-05, + "loss": 0.8401, + "step": 43593 + }, + { + "epoch": 2.8, + "learning_rate": 1.0666824105283498e-05, + "loss": 0.8791, + "step": 43594 + }, + { + "epoch": 2.8, + "learning_rate": 1.0666609416260547e-05, + "loss": 0.9101, + "step": 43595 + }, + { + "epoch": 2.8, + "learning_rate": 1.0666394727237596e-05, + "loss": 0.8916, + "step": 43596 + }, + { + "epoch": 2.8, + "learning_rate": 1.0666180038214647e-05, + "loss": 0.8691, + "step": 43597 + }, + { + "epoch": 2.8, + "learning_rate": 1.0665965349191696e-05, + "loss": 0.9576, + "step": 43598 + }, + { + "epoch": 2.81, + "learning_rate": 1.0665750660168745e-05, + "loss": 0.8132, + "step": 43599 + }, + { + "epoch": 2.81, + "learning_rate": 1.0665535971145795e-05, + "loss": 0.8742, + "step": 43600 + }, + { + "epoch": 2.81, + "learning_rate": 1.0665321282122846e-05, + "loss": 0.9227, + "step": 43601 + }, + { + "epoch": 2.81, + "learning_rate": 1.0665106593099897e-05, + "loss": 0.9047, + "step": 43602 + }, + { + "epoch": 2.81, + "learning_rate": 1.0664891904076946e-05, + "loss": 0.9031, + "step": 43603 + }, + { + "epoch": 2.81, + "learning_rate": 1.0664677215053996e-05, + "loss": 0.865, + "step": 43604 + }, + { + "epoch": 2.81, + "learning_rate": 1.0664462526031045e-05, + "loss": 0.8872, + "step": 43605 + }, + { + "epoch": 2.81, + "learning_rate": 1.0664247837008094e-05, + "loss": 0.8789, + "step": 43606 + }, + { + "epoch": 2.81, + "learning_rate": 1.0664033147985145e-05, + "loss": 0.8869, + "step": 43607 + }, + { + "epoch": 2.81, + "learning_rate": 1.0663818458962194e-05, + "loss": 0.949, + "step": 43608 + }, + { + "epoch": 2.81, + "learning_rate": 1.0663603769939243e-05, + "loss": 0.8846, + "step": 43609 + }, + { + "epoch": 2.81, + "learning_rate": 1.0663389080916294e-05, + "loss": 0.9215, + "step": 43610 + }, + { + "epoch": 2.81, + "learning_rate": 1.0663174391893343e-05, + "loss": 0.8581, + "step": 43611 + }, + { + "epoch": 2.81, + "learning_rate": 1.0662959702870393e-05, + "loss": 0.8299, + "step": 43612 + }, + { + "epoch": 2.81, + "learning_rate": 1.0662745013847442e-05, + "loss": 0.9164, + "step": 43613 + }, + { + "epoch": 2.81, + "learning_rate": 1.0662530324824491e-05, + "loss": 0.9047, + "step": 43614 + }, + { + "epoch": 2.81, + "learning_rate": 1.0662315635801544e-05, + "loss": 0.9697, + "step": 43615 + }, + { + "epoch": 2.81, + "learning_rate": 1.0662100946778592e-05, + "loss": 0.8839, + "step": 43616 + }, + { + "epoch": 2.81, + "learning_rate": 1.0661886257755643e-05, + "loss": 0.9126, + "step": 43617 + }, + { + "epoch": 2.81, + "learning_rate": 1.0661671568732692e-05, + "loss": 0.906, + "step": 43618 + }, + { + "epoch": 2.81, + "learning_rate": 1.0661456879709741e-05, + "loss": 0.8603, + "step": 43619 + }, + { + "epoch": 2.81, + "learning_rate": 1.0661242190686792e-05, + "loss": 0.8592, + "step": 43620 + }, + { + "epoch": 2.81, + "learning_rate": 1.066102750166384e-05, + "loss": 0.8997, + "step": 43621 + }, + { + "epoch": 2.81, + "learning_rate": 1.0660812812640891e-05, + "loss": 0.9297, + "step": 43622 + }, + { + "epoch": 2.81, + "learning_rate": 1.066059812361794e-05, + "loss": 0.8933, + "step": 43623 + }, + { + "epoch": 2.81, + "learning_rate": 1.066038343459499e-05, + "loss": 0.9208, + "step": 43624 + }, + { + "epoch": 2.81, + "learning_rate": 1.066016874557204e-05, + "loss": 0.8397, + "step": 43625 + }, + { + "epoch": 2.81, + "learning_rate": 1.0659954056549089e-05, + "loss": 0.896, + "step": 43626 + }, + { + "epoch": 2.81, + "learning_rate": 1.0659739367526138e-05, + "loss": 0.8736, + "step": 43627 + }, + { + "epoch": 2.81, + "learning_rate": 1.0659524678503189e-05, + "loss": 0.905, + "step": 43628 + }, + { + "epoch": 2.81, + "learning_rate": 1.0659309989480237e-05, + "loss": 0.8986, + "step": 43629 + }, + { + "epoch": 2.81, + "learning_rate": 1.065909530045729e-05, + "loss": 0.9379, + "step": 43630 + }, + { + "epoch": 2.81, + "learning_rate": 1.0658880611434339e-05, + "loss": 0.9151, + "step": 43631 + }, + { + "epoch": 2.81, + "learning_rate": 1.065866592241139e-05, + "loss": 0.8305, + "step": 43632 + }, + { + "epoch": 2.81, + "learning_rate": 1.0658451233388438e-05, + "loss": 0.8723, + "step": 43633 + }, + { + "epoch": 2.81, + "learning_rate": 1.0658236544365487e-05, + "loss": 0.9069, + "step": 43634 + }, + { + "epoch": 2.81, + "learning_rate": 1.0658021855342538e-05, + "loss": 0.9, + "step": 43635 + }, + { + "epoch": 2.81, + "learning_rate": 1.0657807166319587e-05, + "loss": 0.9076, + "step": 43636 + }, + { + "epoch": 2.81, + "learning_rate": 1.0657592477296636e-05, + "loss": 0.9285, + "step": 43637 + }, + { + "epoch": 2.81, + "learning_rate": 1.0657377788273687e-05, + "loss": 0.9682, + "step": 43638 + }, + { + "epoch": 2.81, + "learning_rate": 1.0657163099250736e-05, + "loss": 0.8915, + "step": 43639 + }, + { + "epoch": 2.81, + "learning_rate": 1.0656948410227785e-05, + "loss": 0.9002, + "step": 43640 + }, + { + "epoch": 2.81, + "learning_rate": 1.0656733721204835e-05, + "loss": 0.9024, + "step": 43641 + }, + { + "epoch": 2.81, + "learning_rate": 1.0656519032181884e-05, + "loss": 0.9045, + "step": 43642 + }, + { + "epoch": 2.81, + "learning_rate": 1.0656304343158935e-05, + "loss": 0.8845, + "step": 43643 + }, + { + "epoch": 2.81, + "learning_rate": 1.0656089654135986e-05, + "loss": 0.8824, + "step": 43644 + }, + { + "epoch": 2.81, + "learning_rate": 1.0655874965113036e-05, + "loss": 0.8653, + "step": 43645 + }, + { + "epoch": 2.81, + "learning_rate": 1.0655660276090085e-05, + "loss": 0.9398, + "step": 43646 + }, + { + "epoch": 2.81, + "learning_rate": 1.0655445587067134e-05, + "loss": 0.9124, + "step": 43647 + }, + { + "epoch": 2.81, + "learning_rate": 1.0655230898044185e-05, + "loss": 0.8949, + "step": 43648 + }, + { + "epoch": 2.81, + "learning_rate": 1.0655016209021234e-05, + "loss": 0.9099, + "step": 43649 + }, + { + "epoch": 2.81, + "learning_rate": 1.0654801519998283e-05, + "loss": 0.8719, + "step": 43650 + }, + { + "epoch": 2.81, + "learning_rate": 1.0654586830975333e-05, + "loss": 0.8778, + "step": 43651 + }, + { + "epoch": 2.81, + "learning_rate": 1.0654372141952382e-05, + "loss": 0.9339, + "step": 43652 + }, + { + "epoch": 2.81, + "learning_rate": 1.0654157452929433e-05, + "loss": 0.8907, + "step": 43653 + }, + { + "epoch": 2.81, + "learning_rate": 1.0653942763906482e-05, + "loss": 0.9735, + "step": 43654 + }, + { + "epoch": 2.81, + "learning_rate": 1.0653728074883531e-05, + "loss": 0.9143, + "step": 43655 + }, + { + "epoch": 2.81, + "learning_rate": 1.0653513385860582e-05, + "loss": 0.8579, + "step": 43656 + }, + { + "epoch": 2.81, + "learning_rate": 1.065329869683763e-05, + "loss": 0.8098, + "step": 43657 + }, + { + "epoch": 2.81, + "learning_rate": 1.065308400781468e-05, + "loss": 0.8109, + "step": 43658 + }, + { + "epoch": 2.81, + "learning_rate": 1.0652869318791732e-05, + "loss": 0.9075, + "step": 43659 + }, + { + "epoch": 2.81, + "learning_rate": 1.065265462976878e-05, + "loss": 0.9273, + "step": 43660 + }, + { + "epoch": 2.81, + "learning_rate": 1.0652439940745831e-05, + "loss": 0.8974, + "step": 43661 + }, + { + "epoch": 2.81, + "learning_rate": 1.065222525172288e-05, + "loss": 0.9398, + "step": 43662 + }, + { + "epoch": 2.81, + "learning_rate": 1.0652010562699931e-05, + "loss": 0.8867, + "step": 43663 + }, + { + "epoch": 2.81, + "learning_rate": 1.065179587367698e-05, + "loss": 0.9093, + "step": 43664 + }, + { + "epoch": 2.81, + "learning_rate": 1.0651581184654029e-05, + "loss": 0.8919, + "step": 43665 + }, + { + "epoch": 2.81, + "learning_rate": 1.065136649563108e-05, + "loss": 0.9026, + "step": 43666 + }, + { + "epoch": 2.81, + "learning_rate": 1.0651151806608129e-05, + "loss": 0.8741, + "step": 43667 + }, + { + "epoch": 2.81, + "learning_rate": 1.0650937117585178e-05, + "loss": 0.9465, + "step": 43668 + }, + { + "epoch": 2.81, + "learning_rate": 1.0650722428562228e-05, + "loss": 0.8801, + "step": 43669 + }, + { + "epoch": 2.81, + "learning_rate": 1.0650507739539277e-05, + "loss": 0.8391, + "step": 43670 + }, + { + "epoch": 2.81, + "learning_rate": 1.0650293050516328e-05, + "loss": 0.879, + "step": 43671 + }, + { + "epoch": 2.81, + "learning_rate": 1.0650078361493377e-05, + "loss": 0.9016, + "step": 43672 + }, + { + "epoch": 2.81, + "learning_rate": 1.064986367247043e-05, + "loss": 0.9555, + "step": 43673 + }, + { + "epoch": 2.81, + "learning_rate": 1.0649648983447478e-05, + "loss": 0.8716, + "step": 43674 + }, + { + "epoch": 2.81, + "learning_rate": 1.0649434294424527e-05, + "loss": 0.9728, + "step": 43675 + }, + { + "epoch": 2.81, + "learning_rate": 1.0649219605401578e-05, + "loss": 0.8502, + "step": 43676 + }, + { + "epoch": 2.81, + "learning_rate": 1.0649004916378627e-05, + "loss": 0.9212, + "step": 43677 + }, + { + "epoch": 2.81, + "learning_rate": 1.0648790227355676e-05, + "loss": 0.9573, + "step": 43678 + }, + { + "epoch": 2.81, + "learning_rate": 1.0648575538332726e-05, + "loss": 0.9354, + "step": 43679 + }, + { + "epoch": 2.81, + "learning_rate": 1.0648360849309775e-05, + "loss": 0.9092, + "step": 43680 + }, + { + "epoch": 2.81, + "learning_rate": 1.0648146160286826e-05, + "loss": 0.9365, + "step": 43681 + }, + { + "epoch": 2.81, + "learning_rate": 1.0647931471263875e-05, + "loss": 0.9053, + "step": 43682 + }, + { + "epoch": 2.81, + "learning_rate": 1.0647716782240924e-05, + "loss": 0.8807, + "step": 43683 + }, + { + "epoch": 2.81, + "learning_rate": 1.0647502093217975e-05, + "loss": 0.9376, + "step": 43684 + }, + { + "epoch": 2.81, + "learning_rate": 1.0647287404195024e-05, + "loss": 0.8801, + "step": 43685 + }, + { + "epoch": 2.81, + "learning_rate": 1.0647072715172073e-05, + "loss": 0.8982, + "step": 43686 + }, + { + "epoch": 2.81, + "learning_rate": 1.0646858026149123e-05, + "loss": 0.9023, + "step": 43687 + }, + { + "epoch": 2.81, + "learning_rate": 1.0646643337126174e-05, + "loss": 0.8892, + "step": 43688 + }, + { + "epoch": 2.81, + "learning_rate": 1.0646428648103225e-05, + "loss": 0.987, + "step": 43689 + }, + { + "epoch": 2.81, + "learning_rate": 1.0646213959080273e-05, + "loss": 0.8982, + "step": 43690 + }, + { + "epoch": 2.81, + "learning_rate": 1.0645999270057324e-05, + "loss": 0.9356, + "step": 43691 + }, + { + "epoch": 2.81, + "learning_rate": 1.0645784581034373e-05, + "loss": 0.8734, + "step": 43692 + }, + { + "epoch": 2.81, + "learning_rate": 1.0645569892011422e-05, + "loss": 0.8812, + "step": 43693 + }, + { + "epoch": 2.81, + "learning_rate": 1.0645355202988473e-05, + "loss": 0.8815, + "step": 43694 + }, + { + "epoch": 2.81, + "learning_rate": 1.0645140513965522e-05, + "loss": 0.876, + "step": 43695 + }, + { + "epoch": 2.81, + "learning_rate": 1.064492582494257e-05, + "loss": 0.9311, + "step": 43696 + }, + { + "epoch": 2.81, + "learning_rate": 1.0644711135919621e-05, + "loss": 0.9254, + "step": 43697 + }, + { + "epoch": 2.81, + "learning_rate": 1.064449644689667e-05, + "loss": 0.9411, + "step": 43698 + }, + { + "epoch": 2.81, + "learning_rate": 1.0644281757873721e-05, + "loss": 0.8915, + "step": 43699 + }, + { + "epoch": 2.81, + "learning_rate": 1.064406706885077e-05, + "loss": 0.8813, + "step": 43700 + }, + { + "epoch": 2.81, + "learning_rate": 1.0643852379827819e-05, + "loss": 0.9103, + "step": 43701 + }, + { + "epoch": 2.81, + "learning_rate": 1.0643637690804871e-05, + "loss": 0.9025, + "step": 43702 + }, + { + "epoch": 2.81, + "learning_rate": 1.064342300178192e-05, + "loss": 0.9189, + "step": 43703 + }, + { + "epoch": 2.81, + "learning_rate": 1.064320831275897e-05, + "loss": 0.9557, + "step": 43704 + }, + { + "epoch": 2.81, + "learning_rate": 1.064299362373602e-05, + "loss": 0.8727, + "step": 43705 + }, + { + "epoch": 2.81, + "learning_rate": 1.0642778934713069e-05, + "loss": 0.8608, + "step": 43706 + }, + { + "epoch": 2.81, + "learning_rate": 1.064256424569012e-05, + "loss": 0.9324, + "step": 43707 + }, + { + "epoch": 2.81, + "learning_rate": 1.0642349556667168e-05, + "loss": 0.8928, + "step": 43708 + }, + { + "epoch": 2.81, + "learning_rate": 1.0642134867644219e-05, + "loss": 0.9041, + "step": 43709 + }, + { + "epoch": 2.81, + "learning_rate": 1.0641920178621268e-05, + "loss": 0.8818, + "step": 43710 + }, + { + "epoch": 2.81, + "learning_rate": 1.0641705489598317e-05, + "loss": 0.9147, + "step": 43711 + }, + { + "epoch": 2.81, + "learning_rate": 1.0641490800575368e-05, + "loss": 0.9009, + "step": 43712 + }, + { + "epoch": 2.81, + "learning_rate": 1.0641276111552417e-05, + "loss": 0.8989, + "step": 43713 + }, + { + "epoch": 2.81, + "learning_rate": 1.0641061422529466e-05, + "loss": 0.9056, + "step": 43714 + }, + { + "epoch": 2.81, + "learning_rate": 1.0640846733506516e-05, + "loss": 0.8377, + "step": 43715 + }, + { + "epoch": 2.81, + "learning_rate": 1.0640632044483565e-05, + "loss": 0.9196, + "step": 43716 + }, + { + "epoch": 2.81, + "learning_rate": 1.0640417355460618e-05, + "loss": 0.8812, + "step": 43717 + }, + { + "epoch": 2.81, + "learning_rate": 1.0640202666437667e-05, + "loss": 0.9237, + "step": 43718 + }, + { + "epoch": 2.81, + "learning_rate": 1.0639987977414717e-05, + "loss": 0.8942, + "step": 43719 + }, + { + "epoch": 2.81, + "learning_rate": 1.0639773288391766e-05, + "loss": 0.8601, + "step": 43720 + }, + { + "epoch": 2.81, + "learning_rate": 1.0639558599368815e-05, + "loss": 0.9415, + "step": 43721 + }, + { + "epoch": 2.81, + "learning_rate": 1.0639343910345866e-05, + "loss": 0.8762, + "step": 43722 + }, + { + "epoch": 2.81, + "learning_rate": 1.0639129221322915e-05, + "loss": 0.9004, + "step": 43723 + }, + { + "epoch": 2.81, + "learning_rate": 1.0638914532299964e-05, + "loss": 0.8802, + "step": 43724 + }, + { + "epoch": 2.81, + "learning_rate": 1.0638699843277014e-05, + "loss": 0.8861, + "step": 43725 + }, + { + "epoch": 2.81, + "learning_rate": 1.0638485154254063e-05, + "loss": 0.9144, + "step": 43726 + }, + { + "epoch": 2.81, + "learning_rate": 1.0638270465231112e-05, + "loss": 0.8906, + "step": 43727 + }, + { + "epoch": 2.81, + "learning_rate": 1.0638055776208163e-05, + "loss": 0.8603, + "step": 43728 + }, + { + "epoch": 2.81, + "learning_rate": 1.0637841087185212e-05, + "loss": 0.8991, + "step": 43729 + }, + { + "epoch": 2.81, + "learning_rate": 1.0637626398162263e-05, + "loss": 0.9788, + "step": 43730 + }, + { + "epoch": 2.81, + "learning_rate": 1.0637411709139313e-05, + "loss": 0.86, + "step": 43731 + }, + { + "epoch": 2.81, + "learning_rate": 1.0637197020116364e-05, + "loss": 0.8868, + "step": 43732 + }, + { + "epoch": 2.81, + "learning_rate": 1.0636982331093413e-05, + "loss": 0.9005, + "step": 43733 + }, + { + "epoch": 2.81, + "learning_rate": 1.0636767642070462e-05, + "loss": 0.9083, + "step": 43734 + }, + { + "epoch": 2.81, + "learning_rate": 1.0636552953047512e-05, + "loss": 0.8952, + "step": 43735 + }, + { + "epoch": 2.81, + "learning_rate": 1.0636338264024561e-05, + "loss": 0.9207, + "step": 43736 + }, + { + "epoch": 2.81, + "learning_rate": 1.063612357500161e-05, + "loss": 0.8977, + "step": 43737 + }, + { + "epoch": 2.81, + "learning_rate": 1.0635908885978661e-05, + "loss": 0.8336, + "step": 43738 + }, + { + "epoch": 2.81, + "learning_rate": 1.063569419695571e-05, + "loss": 0.958, + "step": 43739 + }, + { + "epoch": 2.81, + "learning_rate": 1.063547950793276e-05, + "loss": 0.9391, + "step": 43740 + }, + { + "epoch": 2.81, + "learning_rate": 1.063526481890981e-05, + "loss": 0.939, + "step": 43741 + }, + { + "epoch": 2.81, + "learning_rate": 1.0635050129886859e-05, + "loss": 0.9204, + "step": 43742 + }, + { + "epoch": 2.81, + "learning_rate": 1.063483544086391e-05, + "loss": 0.9137, + "step": 43743 + }, + { + "epoch": 2.81, + "learning_rate": 1.0634620751840958e-05, + "loss": 0.9071, + "step": 43744 + }, + { + "epoch": 2.81, + "learning_rate": 1.0634406062818007e-05, + "loss": 0.8967, + "step": 43745 + }, + { + "epoch": 2.81, + "learning_rate": 1.063419137379506e-05, + "loss": 0.8767, + "step": 43746 + }, + { + "epoch": 2.81, + "learning_rate": 1.0633976684772109e-05, + "loss": 0.8868, + "step": 43747 + }, + { + "epoch": 2.81, + "learning_rate": 1.063376199574916e-05, + "loss": 0.9314, + "step": 43748 + }, + { + "epoch": 2.81, + "learning_rate": 1.0633547306726208e-05, + "loss": 0.8328, + "step": 43749 + }, + { + "epoch": 2.81, + "learning_rate": 1.0633332617703259e-05, + "loss": 0.9272, + "step": 43750 + }, + { + "epoch": 2.81, + "learning_rate": 1.0633117928680308e-05, + "loss": 0.8944, + "step": 43751 + }, + { + "epoch": 2.81, + "learning_rate": 1.0632903239657357e-05, + "loss": 0.9393, + "step": 43752 + }, + { + "epoch": 2.81, + "learning_rate": 1.0632688550634407e-05, + "loss": 0.9045, + "step": 43753 + }, + { + "epoch": 2.82, + "learning_rate": 1.0632473861611456e-05, + "loss": 0.8679, + "step": 43754 + }, + { + "epoch": 2.82, + "learning_rate": 1.0632259172588505e-05, + "loss": 0.9479, + "step": 43755 + }, + { + "epoch": 2.82, + "learning_rate": 1.0632044483565556e-05, + "loss": 0.9277, + "step": 43756 + }, + { + "epoch": 2.82, + "learning_rate": 1.0631829794542605e-05, + "loss": 0.9317, + "step": 43757 + }, + { + "epoch": 2.82, + "learning_rate": 1.0631615105519656e-05, + "loss": 0.9087, + "step": 43758 + }, + { + "epoch": 2.82, + "learning_rate": 1.0631400416496705e-05, + "loss": 0.8859, + "step": 43759 + }, + { + "epoch": 2.82, + "learning_rate": 1.0631185727473757e-05, + "loss": 0.8845, + "step": 43760 + }, + { + "epoch": 2.82, + "learning_rate": 1.0630971038450806e-05, + "loss": 0.9376, + "step": 43761 + }, + { + "epoch": 2.82, + "learning_rate": 1.0630756349427855e-05, + "loss": 0.8042, + "step": 43762 + }, + { + "epoch": 2.82, + "learning_rate": 1.0630541660404906e-05, + "loss": 0.9216, + "step": 43763 + }, + { + "epoch": 2.82, + "learning_rate": 1.0630326971381954e-05, + "loss": 0.8617, + "step": 43764 + }, + { + "epoch": 2.82, + "learning_rate": 1.0630112282359003e-05, + "loss": 0.9673, + "step": 43765 + }, + { + "epoch": 2.82, + "learning_rate": 1.0629897593336054e-05, + "loss": 0.9376, + "step": 43766 + }, + { + "epoch": 2.82, + "learning_rate": 1.0629682904313103e-05, + "loss": 0.888, + "step": 43767 + }, + { + "epoch": 2.82, + "learning_rate": 1.0629468215290154e-05, + "loss": 0.9282, + "step": 43768 + }, + { + "epoch": 2.82, + "learning_rate": 1.0629253526267203e-05, + "loss": 0.9468, + "step": 43769 + }, + { + "epoch": 2.82, + "learning_rate": 1.0629038837244252e-05, + "loss": 0.9161, + "step": 43770 + }, + { + "epoch": 2.82, + "learning_rate": 1.0628824148221302e-05, + "loss": 0.885, + "step": 43771 + }, + { + "epoch": 2.82, + "learning_rate": 1.0628609459198351e-05, + "loss": 0.8753, + "step": 43772 + }, + { + "epoch": 2.82, + "learning_rate": 1.06283947701754e-05, + "loss": 0.8456, + "step": 43773 + }, + { + "epoch": 2.82, + "learning_rate": 1.0628180081152451e-05, + "loss": 0.9104, + "step": 43774 + }, + { + "epoch": 2.82, + "learning_rate": 1.0627965392129502e-05, + "loss": 0.9192, + "step": 43775 + }, + { + "epoch": 2.82, + "learning_rate": 1.0627750703106552e-05, + "loss": 0.9502, + "step": 43776 + }, + { + "epoch": 2.82, + "learning_rate": 1.0627536014083601e-05, + "loss": 0.8515, + "step": 43777 + }, + { + "epoch": 2.82, + "learning_rate": 1.0627321325060652e-05, + "loss": 0.9152, + "step": 43778 + }, + { + "epoch": 2.82, + "learning_rate": 1.06271066360377e-05, + "loss": 0.8674, + "step": 43779 + }, + { + "epoch": 2.82, + "learning_rate": 1.062689194701475e-05, + "loss": 0.9447, + "step": 43780 + }, + { + "epoch": 2.82, + "learning_rate": 1.06266772579918e-05, + "loss": 0.8848, + "step": 43781 + }, + { + "epoch": 2.82, + "learning_rate": 1.062646256896885e-05, + "loss": 0.8582, + "step": 43782 + }, + { + "epoch": 2.82, + "learning_rate": 1.0626247879945898e-05, + "loss": 0.9235, + "step": 43783 + }, + { + "epoch": 2.82, + "learning_rate": 1.0626033190922949e-05, + "loss": 0.9207, + "step": 43784 + }, + { + "epoch": 2.82, + "learning_rate": 1.0625818501899998e-05, + "loss": 0.8897, + "step": 43785 + }, + { + "epoch": 2.82, + "learning_rate": 1.0625603812877047e-05, + "loss": 0.8934, + "step": 43786 + }, + { + "epoch": 2.82, + "learning_rate": 1.0625389123854098e-05, + "loss": 0.8386, + "step": 43787 + }, + { + "epoch": 2.82, + "learning_rate": 1.0625174434831147e-05, + "loss": 0.9043, + "step": 43788 + }, + { + "epoch": 2.82, + "learning_rate": 1.0624959745808199e-05, + "loss": 0.9691, + "step": 43789 + }, + { + "epoch": 2.82, + "learning_rate": 1.0624745056785248e-05, + "loss": 0.9158, + "step": 43790 + }, + { + "epoch": 2.82, + "learning_rate": 1.0624530367762299e-05, + "loss": 0.894, + "step": 43791 + }, + { + "epoch": 2.82, + "learning_rate": 1.0624315678739348e-05, + "loss": 0.9087, + "step": 43792 + }, + { + "epoch": 2.82, + "learning_rate": 1.0624100989716396e-05, + "loss": 0.8881, + "step": 43793 + }, + { + "epoch": 2.82, + "learning_rate": 1.0623886300693447e-05, + "loss": 0.9125, + "step": 43794 + }, + { + "epoch": 2.82, + "learning_rate": 1.0623671611670496e-05, + "loss": 0.8557, + "step": 43795 + }, + { + "epoch": 2.82, + "learning_rate": 1.0623456922647545e-05, + "loss": 0.9161, + "step": 43796 + }, + { + "epoch": 2.82, + "learning_rate": 1.0623242233624596e-05, + "loss": 0.9087, + "step": 43797 + }, + { + "epoch": 2.82, + "learning_rate": 1.0623027544601645e-05, + "loss": 0.8734, + "step": 43798 + }, + { + "epoch": 2.82, + "learning_rate": 1.0622812855578695e-05, + "loss": 0.9669, + "step": 43799 + }, + { + "epoch": 2.82, + "learning_rate": 1.0622598166555744e-05, + "loss": 0.9452, + "step": 43800 + }, + { + "epoch": 2.82, + "learning_rate": 1.0622383477532793e-05, + "loss": 0.859, + "step": 43801 + }, + { + "epoch": 2.82, + "learning_rate": 1.0622168788509844e-05, + "loss": 0.8919, + "step": 43802 + }, + { + "epoch": 2.82, + "learning_rate": 1.0621954099486893e-05, + "loss": 0.9222, + "step": 43803 + }, + { + "epoch": 2.82, + "learning_rate": 1.0621739410463945e-05, + "loss": 0.8837, + "step": 43804 + }, + { + "epoch": 2.82, + "learning_rate": 1.0621524721440994e-05, + "loss": 0.9065, + "step": 43805 + }, + { + "epoch": 2.82, + "learning_rate": 1.0621310032418045e-05, + "loss": 0.8694, + "step": 43806 + }, + { + "epoch": 2.82, + "learning_rate": 1.0621095343395094e-05, + "loss": 0.9024, + "step": 43807 + }, + { + "epoch": 2.82, + "learning_rate": 1.0620880654372143e-05, + "loss": 0.8392, + "step": 43808 + }, + { + "epoch": 2.82, + "learning_rate": 1.0620665965349193e-05, + "loss": 0.8861, + "step": 43809 + }, + { + "epoch": 2.82, + "learning_rate": 1.0620451276326242e-05, + "loss": 0.9447, + "step": 43810 + }, + { + "epoch": 2.82, + "learning_rate": 1.0620236587303291e-05, + "loss": 0.8571, + "step": 43811 + }, + { + "epoch": 2.82, + "learning_rate": 1.0620021898280342e-05, + "loss": 0.8544, + "step": 43812 + }, + { + "epoch": 2.82, + "learning_rate": 1.0619807209257391e-05, + "loss": 0.8948, + "step": 43813 + }, + { + "epoch": 2.82, + "learning_rate": 1.061959252023444e-05, + "loss": 0.941, + "step": 43814 + }, + { + "epoch": 2.82, + "learning_rate": 1.061937783121149e-05, + "loss": 0.8611, + "step": 43815 + }, + { + "epoch": 2.82, + "learning_rate": 1.061916314218854e-05, + "loss": 0.8849, + "step": 43816 + }, + { + "epoch": 2.82, + "learning_rate": 1.061894845316559e-05, + "loss": 0.9392, + "step": 43817 + }, + { + "epoch": 2.82, + "learning_rate": 1.0618733764142641e-05, + "loss": 0.9218, + "step": 43818 + }, + { + "epoch": 2.82, + "learning_rate": 1.0618519075119692e-05, + "loss": 0.9137, + "step": 43819 + }, + { + "epoch": 2.82, + "learning_rate": 1.061830438609674e-05, + "loss": 0.9027, + "step": 43820 + }, + { + "epoch": 2.82, + "learning_rate": 1.061808969707379e-05, + "loss": 0.924, + "step": 43821 + }, + { + "epoch": 2.82, + "learning_rate": 1.061787500805084e-05, + "loss": 0.9077, + "step": 43822 + }, + { + "epoch": 2.82, + "learning_rate": 1.0617660319027889e-05, + "loss": 0.8808, + "step": 43823 + }, + { + "epoch": 2.82, + "learning_rate": 1.0617445630004938e-05, + "loss": 0.8947, + "step": 43824 + }, + { + "epoch": 2.82, + "learning_rate": 1.0617230940981989e-05, + "loss": 0.9217, + "step": 43825 + }, + { + "epoch": 2.82, + "learning_rate": 1.0617016251959038e-05, + "loss": 0.9427, + "step": 43826 + }, + { + "epoch": 2.82, + "learning_rate": 1.0616801562936088e-05, + "loss": 0.8795, + "step": 43827 + }, + { + "epoch": 2.82, + "learning_rate": 1.0616586873913137e-05, + "loss": 0.8933, + "step": 43828 + }, + { + "epoch": 2.82, + "learning_rate": 1.0616372184890186e-05, + "loss": 0.8925, + "step": 43829 + }, + { + "epoch": 2.82, + "learning_rate": 1.0616157495867237e-05, + "loss": 0.9376, + "step": 43830 + }, + { + "epoch": 2.82, + "learning_rate": 1.0615942806844286e-05, + "loss": 0.896, + "step": 43831 + }, + { + "epoch": 2.82, + "learning_rate": 1.0615728117821335e-05, + "loss": 0.9106, + "step": 43832 + }, + { + "epoch": 2.82, + "learning_rate": 1.0615513428798387e-05, + "loss": 0.8174, + "step": 43833 + }, + { + "epoch": 2.82, + "learning_rate": 1.0615298739775436e-05, + "loss": 0.8596, + "step": 43834 + }, + { + "epoch": 2.82, + "learning_rate": 1.0615084050752487e-05, + "loss": 0.8876, + "step": 43835 + }, + { + "epoch": 2.82, + "learning_rate": 1.0614869361729536e-05, + "loss": 0.8502, + "step": 43836 + }, + { + "epoch": 2.82, + "learning_rate": 1.0614654672706587e-05, + "loss": 0.8921, + "step": 43837 + }, + { + "epoch": 2.82, + "learning_rate": 1.0614439983683635e-05, + "loss": 0.9329, + "step": 43838 + }, + { + "epoch": 2.82, + "learning_rate": 1.0614225294660684e-05, + "loss": 0.8505, + "step": 43839 + }, + { + "epoch": 2.82, + "learning_rate": 1.0614010605637735e-05, + "loss": 0.9489, + "step": 43840 + }, + { + "epoch": 2.82, + "learning_rate": 1.0613795916614784e-05, + "loss": 0.9165, + "step": 43841 + }, + { + "epoch": 2.82, + "learning_rate": 1.0613581227591833e-05, + "loss": 0.8478, + "step": 43842 + }, + { + "epoch": 2.82, + "learning_rate": 1.0613366538568884e-05, + "loss": 0.882, + "step": 43843 + }, + { + "epoch": 2.82, + "learning_rate": 1.0613151849545933e-05, + "loss": 0.8927, + "step": 43844 + }, + { + "epoch": 2.82, + "learning_rate": 1.0612937160522983e-05, + "loss": 0.8165, + "step": 43845 + }, + { + "epoch": 2.82, + "learning_rate": 1.0612722471500032e-05, + "loss": 0.8934, + "step": 43846 + }, + { + "epoch": 2.82, + "learning_rate": 1.0612507782477085e-05, + "loss": 0.8468, + "step": 43847 + }, + { + "epoch": 2.82, + "learning_rate": 1.0612293093454134e-05, + "loss": 0.9342, + "step": 43848 + }, + { + "epoch": 2.82, + "learning_rate": 1.0612078404431183e-05, + "loss": 0.9194, + "step": 43849 + }, + { + "epoch": 2.82, + "learning_rate": 1.0611863715408233e-05, + "loss": 0.9568, + "step": 43850 + }, + { + "epoch": 2.82, + "learning_rate": 1.0611649026385282e-05, + "loss": 0.9136, + "step": 43851 + }, + { + "epoch": 2.82, + "learning_rate": 1.0611434337362331e-05, + "loss": 0.9487, + "step": 43852 + }, + { + "epoch": 2.82, + "learning_rate": 1.0611219648339382e-05, + "loss": 0.8934, + "step": 43853 + }, + { + "epoch": 2.82, + "learning_rate": 1.061100495931643e-05, + "loss": 0.9043, + "step": 43854 + }, + { + "epoch": 2.82, + "learning_rate": 1.0610790270293481e-05, + "loss": 0.8778, + "step": 43855 + }, + { + "epoch": 2.82, + "learning_rate": 1.061057558127053e-05, + "loss": 0.861, + "step": 43856 + }, + { + "epoch": 2.82, + "learning_rate": 1.061036089224758e-05, + "loss": 0.9012, + "step": 43857 + }, + { + "epoch": 2.82, + "learning_rate": 1.061014620322463e-05, + "loss": 0.8853, + "step": 43858 + }, + { + "epoch": 2.82, + "learning_rate": 1.0609931514201679e-05, + "loss": 0.8933, + "step": 43859 + }, + { + "epoch": 2.82, + "learning_rate": 1.0609716825178728e-05, + "loss": 0.8827, + "step": 43860 + }, + { + "epoch": 2.82, + "learning_rate": 1.0609502136155779e-05, + "loss": 0.9519, + "step": 43861 + }, + { + "epoch": 2.82, + "learning_rate": 1.060928744713283e-05, + "loss": 0.9334, + "step": 43862 + }, + { + "epoch": 2.82, + "learning_rate": 1.060907275810988e-05, + "loss": 0.8668, + "step": 43863 + }, + { + "epoch": 2.82, + "learning_rate": 1.0608858069086929e-05, + "loss": 0.8668, + "step": 43864 + }, + { + "epoch": 2.82, + "learning_rate": 1.060864338006398e-05, + "loss": 0.9195, + "step": 43865 + }, + { + "epoch": 2.82, + "learning_rate": 1.0608428691041029e-05, + "loss": 0.8759, + "step": 43866 + }, + { + "epoch": 2.82, + "learning_rate": 1.0608214002018077e-05, + "loss": 0.8726, + "step": 43867 + }, + { + "epoch": 2.82, + "learning_rate": 1.0607999312995128e-05, + "loss": 0.8341, + "step": 43868 + }, + { + "epoch": 2.82, + "learning_rate": 1.0607784623972177e-05, + "loss": 0.947, + "step": 43869 + }, + { + "epoch": 2.82, + "learning_rate": 1.0607569934949226e-05, + "loss": 0.9376, + "step": 43870 + }, + { + "epoch": 2.82, + "learning_rate": 1.0607355245926277e-05, + "loss": 0.976, + "step": 43871 + }, + { + "epoch": 2.82, + "learning_rate": 1.0607140556903326e-05, + "loss": 0.8722, + "step": 43872 + }, + { + "epoch": 2.82, + "learning_rate": 1.0606925867880375e-05, + "loss": 0.8428, + "step": 43873 + }, + { + "epoch": 2.82, + "learning_rate": 1.0606711178857425e-05, + "loss": 0.948, + "step": 43874 + }, + { + "epoch": 2.82, + "learning_rate": 1.0606496489834474e-05, + "loss": 0.9331, + "step": 43875 + }, + { + "epoch": 2.82, + "learning_rate": 1.0606281800811527e-05, + "loss": 0.9515, + "step": 43876 + }, + { + "epoch": 2.82, + "learning_rate": 1.0606067111788576e-05, + "loss": 0.9033, + "step": 43877 + }, + { + "epoch": 2.82, + "learning_rate": 1.0605852422765626e-05, + "loss": 0.9565, + "step": 43878 + }, + { + "epoch": 2.82, + "learning_rate": 1.0605637733742675e-05, + "loss": 0.9016, + "step": 43879 + }, + { + "epoch": 2.82, + "learning_rate": 1.0605423044719724e-05, + "loss": 0.8223, + "step": 43880 + }, + { + "epoch": 2.82, + "learning_rate": 1.0605208355696775e-05, + "loss": 0.8357, + "step": 43881 + }, + { + "epoch": 2.82, + "learning_rate": 1.0604993666673824e-05, + "loss": 0.8217, + "step": 43882 + }, + { + "epoch": 2.82, + "learning_rate": 1.0604778977650873e-05, + "loss": 0.8974, + "step": 43883 + }, + { + "epoch": 2.82, + "learning_rate": 1.0604564288627923e-05, + "loss": 0.8768, + "step": 43884 + }, + { + "epoch": 2.82, + "learning_rate": 1.0604349599604972e-05, + "loss": 0.8718, + "step": 43885 + }, + { + "epoch": 2.82, + "learning_rate": 1.0604134910582023e-05, + "loss": 0.9434, + "step": 43886 + }, + { + "epoch": 2.82, + "learning_rate": 1.0603920221559072e-05, + "loss": 0.8806, + "step": 43887 + }, + { + "epoch": 2.82, + "learning_rate": 1.0603705532536121e-05, + "loss": 0.8673, + "step": 43888 + }, + { + "epoch": 2.82, + "learning_rate": 1.0603490843513172e-05, + "loss": 0.9004, + "step": 43889 + }, + { + "epoch": 2.82, + "learning_rate": 1.060327615449022e-05, + "loss": 0.9082, + "step": 43890 + }, + { + "epoch": 2.82, + "learning_rate": 1.0603061465467273e-05, + "loss": 0.8613, + "step": 43891 + }, + { + "epoch": 2.82, + "learning_rate": 1.0602846776444322e-05, + "loss": 0.8684, + "step": 43892 + }, + { + "epoch": 2.82, + "learning_rate": 1.0602632087421371e-05, + "loss": 0.9651, + "step": 43893 + }, + { + "epoch": 2.82, + "learning_rate": 1.0602417398398422e-05, + "loss": 0.8823, + "step": 43894 + }, + { + "epoch": 2.82, + "learning_rate": 1.060220270937547e-05, + "loss": 0.8425, + "step": 43895 + }, + { + "epoch": 2.82, + "learning_rate": 1.0601988020352521e-05, + "loss": 0.9324, + "step": 43896 + }, + { + "epoch": 2.82, + "learning_rate": 1.060177333132957e-05, + "loss": 0.892, + "step": 43897 + }, + { + "epoch": 2.82, + "learning_rate": 1.0601558642306619e-05, + "loss": 0.8208, + "step": 43898 + }, + { + "epoch": 2.82, + "learning_rate": 1.060134395328367e-05, + "loss": 0.9289, + "step": 43899 + }, + { + "epoch": 2.82, + "learning_rate": 1.0601129264260719e-05, + "loss": 0.8393, + "step": 43900 + }, + { + "epoch": 2.82, + "learning_rate": 1.0600914575237768e-05, + "loss": 0.9162, + "step": 43901 + }, + { + "epoch": 2.82, + "learning_rate": 1.0600699886214818e-05, + "loss": 0.9107, + "step": 43902 + }, + { + "epoch": 2.82, + "learning_rate": 1.0600485197191867e-05, + "loss": 0.8643, + "step": 43903 + }, + { + "epoch": 2.82, + "learning_rate": 1.0600270508168918e-05, + "loss": 0.8934, + "step": 43904 + }, + { + "epoch": 2.82, + "learning_rate": 1.0600055819145969e-05, + "loss": 0.9616, + "step": 43905 + }, + { + "epoch": 2.82, + "learning_rate": 1.059984113012302e-05, + "loss": 0.8794, + "step": 43906 + }, + { + "epoch": 2.82, + "learning_rate": 1.0599626441100068e-05, + "loss": 0.902, + "step": 43907 + }, + { + "epoch": 2.82, + "learning_rate": 1.0599411752077117e-05, + "loss": 0.8914, + "step": 43908 + }, + { + "epoch": 2.83, + "learning_rate": 1.0599197063054168e-05, + "loss": 0.8898, + "step": 43909 + }, + { + "epoch": 2.83, + "learning_rate": 1.0598982374031217e-05, + "loss": 0.939, + "step": 43910 + }, + { + "epoch": 2.83, + "learning_rate": 1.0598767685008266e-05, + "loss": 0.9001, + "step": 43911 + }, + { + "epoch": 2.83, + "learning_rate": 1.0598552995985316e-05, + "loss": 0.9007, + "step": 43912 + }, + { + "epoch": 2.83, + "learning_rate": 1.0598338306962365e-05, + "loss": 0.8756, + "step": 43913 + }, + { + "epoch": 2.83, + "learning_rate": 1.0598123617939416e-05, + "loss": 0.9217, + "step": 43914 + }, + { + "epoch": 2.83, + "learning_rate": 1.0597908928916465e-05, + "loss": 0.8843, + "step": 43915 + }, + { + "epoch": 2.83, + "learning_rate": 1.0597694239893514e-05, + "loss": 0.8934, + "step": 43916 + }, + { + "epoch": 2.83, + "learning_rate": 1.0597479550870565e-05, + "loss": 0.9344, + "step": 43917 + }, + { + "epoch": 2.83, + "learning_rate": 1.0597264861847614e-05, + "loss": 0.863, + "step": 43918 + }, + { + "epoch": 2.83, + "learning_rate": 1.0597050172824663e-05, + "loss": 0.7387, + "step": 43919 + }, + { + "epoch": 2.83, + "learning_rate": 1.0596835483801715e-05, + "loss": 0.8621, + "step": 43920 + }, + { + "epoch": 2.83, + "learning_rate": 1.0596620794778764e-05, + "loss": 0.9239, + "step": 43921 + }, + { + "epoch": 2.83, + "learning_rate": 1.0596406105755815e-05, + "loss": 0.9263, + "step": 43922 + }, + { + "epoch": 2.83, + "learning_rate": 1.0596191416732864e-05, + "loss": 0.8807, + "step": 43923 + }, + { + "epoch": 2.83, + "learning_rate": 1.0595976727709914e-05, + "loss": 0.9353, + "step": 43924 + }, + { + "epoch": 2.83, + "learning_rate": 1.0595762038686963e-05, + "loss": 0.8726, + "step": 43925 + }, + { + "epoch": 2.83, + "learning_rate": 1.0595547349664012e-05, + "loss": 0.9286, + "step": 43926 + }, + { + "epoch": 2.83, + "learning_rate": 1.0595332660641063e-05, + "loss": 0.8672, + "step": 43927 + }, + { + "epoch": 2.83, + "learning_rate": 1.0595117971618112e-05, + "loss": 0.8889, + "step": 43928 + }, + { + "epoch": 2.83, + "learning_rate": 1.059490328259516e-05, + "loss": 0.8751, + "step": 43929 + }, + { + "epoch": 2.83, + "learning_rate": 1.0594688593572211e-05, + "loss": 0.8778, + "step": 43930 + }, + { + "epoch": 2.83, + "learning_rate": 1.059447390454926e-05, + "loss": 0.9281, + "step": 43931 + }, + { + "epoch": 2.83, + "learning_rate": 1.0594259215526311e-05, + "loss": 1.018, + "step": 43932 + }, + { + "epoch": 2.83, + "learning_rate": 1.059404452650336e-05, + "loss": 0.9377, + "step": 43933 + }, + { + "epoch": 2.83, + "learning_rate": 1.0593829837480409e-05, + "loss": 0.9244, + "step": 43934 + }, + { + "epoch": 2.83, + "learning_rate": 1.0593615148457461e-05, + "loss": 0.9921, + "step": 43935 + }, + { + "epoch": 2.83, + "learning_rate": 1.059340045943451e-05, + "loss": 0.9462, + "step": 43936 + }, + { + "epoch": 2.83, + "learning_rate": 1.0593185770411561e-05, + "loss": 0.9164, + "step": 43937 + }, + { + "epoch": 2.83, + "learning_rate": 1.059297108138861e-05, + "loss": 0.8808, + "step": 43938 + }, + { + "epoch": 2.83, + "learning_rate": 1.0592756392365659e-05, + "loss": 0.8519, + "step": 43939 + }, + { + "epoch": 2.83, + "learning_rate": 1.059254170334271e-05, + "loss": 0.8594, + "step": 43940 + }, + { + "epoch": 2.83, + "learning_rate": 1.0592327014319758e-05, + "loss": 0.8837, + "step": 43941 + }, + { + "epoch": 2.83, + "learning_rate": 1.0592112325296809e-05, + "loss": 0.8726, + "step": 43942 + }, + { + "epoch": 2.83, + "learning_rate": 1.0591897636273858e-05, + "loss": 0.8892, + "step": 43943 + }, + { + "epoch": 2.83, + "learning_rate": 1.0591682947250907e-05, + "loss": 0.9377, + "step": 43944 + }, + { + "epoch": 2.83, + "learning_rate": 1.0591468258227958e-05, + "loss": 0.886, + "step": 43945 + }, + { + "epoch": 2.83, + "learning_rate": 1.0591253569205007e-05, + "loss": 0.8995, + "step": 43946 + }, + { + "epoch": 2.83, + "learning_rate": 1.0591038880182056e-05, + "loss": 0.8691, + "step": 43947 + }, + { + "epoch": 2.83, + "learning_rate": 1.0590824191159106e-05, + "loss": 0.9655, + "step": 43948 + }, + { + "epoch": 2.83, + "learning_rate": 1.0590609502136157e-05, + "loss": 0.8276, + "step": 43949 + }, + { + "epoch": 2.83, + "learning_rate": 1.0590394813113208e-05, + "loss": 0.8906, + "step": 43950 + }, + { + "epoch": 2.83, + "learning_rate": 1.0590180124090257e-05, + "loss": 0.8163, + "step": 43951 + }, + { + "epoch": 2.83, + "learning_rate": 1.0589965435067307e-05, + "loss": 0.9151, + "step": 43952 + }, + { + "epoch": 2.83, + "learning_rate": 1.0589750746044356e-05, + "loss": 0.8637, + "step": 43953 + }, + { + "epoch": 2.83, + "learning_rate": 1.0589536057021405e-05, + "loss": 0.8799, + "step": 43954 + }, + { + "epoch": 2.83, + "learning_rate": 1.0589321367998456e-05, + "loss": 0.8505, + "step": 43955 + }, + { + "epoch": 2.83, + "learning_rate": 1.0589106678975505e-05, + "loss": 0.812, + "step": 43956 + }, + { + "epoch": 2.83, + "learning_rate": 1.0588891989952554e-05, + "loss": 0.9655, + "step": 43957 + }, + { + "epoch": 2.83, + "learning_rate": 1.0588677300929604e-05, + "loss": 0.9016, + "step": 43958 + }, + { + "epoch": 2.83, + "learning_rate": 1.0588462611906653e-05, + "loss": 0.851, + "step": 43959 + }, + { + "epoch": 2.83, + "learning_rate": 1.0588247922883702e-05, + "loss": 0.8366, + "step": 43960 + }, + { + "epoch": 2.83, + "learning_rate": 1.0588033233860753e-05, + "loss": 0.8927, + "step": 43961 + }, + { + "epoch": 2.83, + "learning_rate": 1.0587818544837802e-05, + "loss": 0.9003, + "step": 43962 + }, + { + "epoch": 2.83, + "learning_rate": 1.0587603855814853e-05, + "loss": 0.9157, + "step": 43963 + }, + { + "epoch": 2.83, + "learning_rate": 1.0587389166791903e-05, + "loss": 0.8787, + "step": 43964 + }, + { + "epoch": 2.83, + "learning_rate": 1.0587174477768954e-05, + "loss": 0.86, + "step": 43965 + }, + { + "epoch": 2.83, + "learning_rate": 1.0586959788746003e-05, + "loss": 0.873, + "step": 43966 + }, + { + "epoch": 2.83, + "learning_rate": 1.0586745099723052e-05, + "loss": 0.919, + "step": 43967 + }, + { + "epoch": 2.83, + "learning_rate": 1.0586530410700103e-05, + "loss": 0.8698, + "step": 43968 + }, + { + "epoch": 2.83, + "learning_rate": 1.0586315721677152e-05, + "loss": 0.848, + "step": 43969 + }, + { + "epoch": 2.83, + "learning_rate": 1.05861010326542e-05, + "loss": 0.9109, + "step": 43970 + }, + { + "epoch": 2.83, + "learning_rate": 1.0585886343631251e-05, + "loss": 0.8775, + "step": 43971 + }, + { + "epoch": 2.83, + "learning_rate": 1.05856716546083e-05, + "loss": 0.8707, + "step": 43972 + }, + { + "epoch": 2.83, + "learning_rate": 1.058545696558535e-05, + "loss": 0.9233, + "step": 43973 + }, + { + "epoch": 2.83, + "learning_rate": 1.05852422765624e-05, + "loss": 0.9211, + "step": 43974 + }, + { + "epoch": 2.83, + "learning_rate": 1.0585027587539449e-05, + "loss": 0.9238, + "step": 43975 + }, + { + "epoch": 2.83, + "learning_rate": 1.05848128985165e-05, + "loss": 0.9185, + "step": 43976 + }, + { + "epoch": 2.83, + "learning_rate": 1.0584598209493548e-05, + "loss": 0.899, + "step": 43977 + }, + { + "epoch": 2.83, + "learning_rate": 1.05843835204706e-05, + "loss": 0.8843, + "step": 43978 + }, + { + "epoch": 2.83, + "learning_rate": 1.058416883144765e-05, + "loss": 0.8625, + "step": 43979 + }, + { + "epoch": 2.83, + "learning_rate": 1.0583954142424699e-05, + "loss": 0.9329, + "step": 43980 + }, + { + "epoch": 2.83, + "learning_rate": 1.058373945340175e-05, + "loss": 0.8924, + "step": 43981 + }, + { + "epoch": 2.83, + "learning_rate": 1.0583524764378798e-05, + "loss": 0.9191, + "step": 43982 + }, + { + "epoch": 2.83, + "learning_rate": 1.0583310075355849e-05, + "loss": 0.973, + "step": 43983 + }, + { + "epoch": 2.83, + "learning_rate": 1.0583095386332898e-05, + "loss": 0.9621, + "step": 43984 + }, + { + "epoch": 2.83, + "learning_rate": 1.0582880697309947e-05, + "loss": 0.8763, + "step": 43985 + }, + { + "epoch": 2.83, + "learning_rate": 1.0582666008286997e-05, + "loss": 0.9435, + "step": 43986 + }, + { + "epoch": 2.83, + "learning_rate": 1.0582451319264046e-05, + "loss": 0.8788, + "step": 43987 + }, + { + "epoch": 2.83, + "learning_rate": 1.0582236630241095e-05, + "loss": 0.875, + "step": 43988 + }, + { + "epoch": 2.83, + "learning_rate": 1.0582021941218146e-05, + "loss": 0.951, + "step": 43989 + }, + { + "epoch": 2.83, + "learning_rate": 1.0581807252195195e-05, + "loss": 0.8956, + "step": 43990 + }, + { + "epoch": 2.83, + "learning_rate": 1.0581592563172246e-05, + "loss": 0.851, + "step": 43991 + }, + { + "epoch": 2.83, + "learning_rate": 1.0581377874149295e-05, + "loss": 0.9136, + "step": 43992 + }, + { + "epoch": 2.83, + "learning_rate": 1.0581163185126347e-05, + "loss": 0.8656, + "step": 43993 + }, + { + "epoch": 2.83, + "learning_rate": 1.0580948496103396e-05, + "loss": 0.8584, + "step": 43994 + }, + { + "epoch": 2.83, + "learning_rate": 1.0580733807080445e-05, + "loss": 0.9083, + "step": 43995 + }, + { + "epoch": 2.83, + "learning_rate": 1.0580519118057496e-05, + "loss": 0.9288, + "step": 43996 + }, + { + "epoch": 2.83, + "learning_rate": 1.0580304429034545e-05, + "loss": 0.8513, + "step": 43997 + }, + { + "epoch": 2.83, + "learning_rate": 1.0580089740011594e-05, + "loss": 0.9027, + "step": 43998 + }, + { + "epoch": 2.83, + "learning_rate": 1.0579875050988644e-05, + "loss": 0.8928, + "step": 43999 + }, + { + "epoch": 2.83, + "learning_rate": 1.0579660361965693e-05, + "loss": 0.8827, + "step": 44000 + }, + { + "epoch": 2.83, + "eval_laion/cider": 0.4869535858906943, + "eval_laion/meteor": 0.38434351041549497, + "eval_laion/pred_num_words": 36.306640625, + "eval_laion/sacrebleu": 15.311592117377781, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0340217351913452, + "eval_runtime": 97.0121, + "eval_samples_per_second": 5.144, + "eval_steps_per_second": 0.33, + "step": 44000 + }, + { + "epoch": 2.83, + "learning_rate": 1.0579445672942744e-05, + "loss": 0.9092, + "step": 44001 + }, + { + "epoch": 2.83, + "learning_rate": 1.0579230983919793e-05, + "loss": 0.8942, + "step": 44002 + }, + { + "epoch": 2.83, + "learning_rate": 1.0579016294896842e-05, + "loss": 0.8603, + "step": 44003 + }, + { + "epoch": 2.83, + "learning_rate": 1.0578801605873892e-05, + "loss": 0.8933, + "step": 44004 + }, + { + "epoch": 2.83, + "learning_rate": 1.0578586916850941e-05, + "loss": 0.9628, + "step": 44005 + }, + { + "epoch": 2.83, + "learning_rate": 1.057837222782799e-05, + "loss": 0.8824, + "step": 44006 + }, + { + "epoch": 2.83, + "learning_rate": 1.0578157538805043e-05, + "loss": 0.8761, + "step": 44007 + }, + { + "epoch": 2.83, + "learning_rate": 1.0577942849782092e-05, + "loss": 0.8316, + "step": 44008 + }, + { + "epoch": 2.83, + "learning_rate": 1.0577728160759142e-05, + "loss": 0.8988, + "step": 44009 + }, + { + "epoch": 2.83, + "learning_rate": 1.0577513471736191e-05, + "loss": 0.8644, + "step": 44010 + }, + { + "epoch": 2.83, + "learning_rate": 1.0577298782713242e-05, + "loss": 0.8919, + "step": 44011 + }, + { + "epoch": 2.83, + "learning_rate": 1.0577084093690291e-05, + "loss": 0.8914, + "step": 44012 + }, + { + "epoch": 2.83, + "learning_rate": 1.057686940466734e-05, + "loss": 0.8573, + "step": 44013 + }, + { + "epoch": 2.83, + "learning_rate": 1.057665471564439e-05, + "loss": 0.8939, + "step": 44014 + }, + { + "epoch": 2.83, + "learning_rate": 1.057644002662144e-05, + "loss": 0.9194, + "step": 44015 + }, + { + "epoch": 2.83, + "learning_rate": 1.0576225337598488e-05, + "loss": 0.8657, + "step": 44016 + }, + { + "epoch": 2.83, + "learning_rate": 1.0576010648575539e-05, + "loss": 0.8905, + "step": 44017 + }, + { + "epoch": 2.83, + "learning_rate": 1.0575795959552588e-05, + "loss": 0.9237, + "step": 44018 + }, + { + "epoch": 2.83, + "learning_rate": 1.0575581270529639e-05, + "loss": 0.9399, + "step": 44019 + }, + { + "epoch": 2.83, + "learning_rate": 1.0575366581506688e-05, + "loss": 0.9527, + "step": 44020 + }, + { + "epoch": 2.83, + "learning_rate": 1.0575151892483737e-05, + "loss": 0.9633, + "step": 44021 + }, + { + "epoch": 2.83, + "learning_rate": 1.0574937203460789e-05, + "loss": 0.9503, + "step": 44022 + }, + { + "epoch": 2.83, + "learning_rate": 1.0574722514437838e-05, + "loss": 0.9019, + "step": 44023 + }, + { + "epoch": 2.83, + "learning_rate": 1.0574507825414889e-05, + "loss": 0.8583, + "step": 44024 + }, + { + "epoch": 2.83, + "learning_rate": 1.0574293136391938e-05, + "loss": 0.9064, + "step": 44025 + }, + { + "epoch": 2.83, + "learning_rate": 1.0574078447368987e-05, + "loss": 0.8682, + "step": 44026 + }, + { + "epoch": 2.83, + "learning_rate": 1.0573863758346037e-05, + "loss": 0.9162, + "step": 44027 + }, + { + "epoch": 2.83, + "learning_rate": 1.0573649069323086e-05, + "loss": 0.8523, + "step": 44028 + }, + { + "epoch": 2.83, + "learning_rate": 1.0573434380300137e-05, + "loss": 0.9356, + "step": 44029 + }, + { + "epoch": 2.83, + "learning_rate": 1.0573219691277186e-05, + "loss": 0.8977, + "step": 44030 + }, + { + "epoch": 2.83, + "learning_rate": 1.0573005002254235e-05, + "loss": 0.86, + "step": 44031 + }, + { + "epoch": 2.83, + "learning_rate": 1.0572790313231285e-05, + "loss": 0.9948, + "step": 44032 + }, + { + "epoch": 2.83, + "learning_rate": 1.0572575624208334e-05, + "loss": 0.9456, + "step": 44033 + }, + { + "epoch": 2.83, + "learning_rate": 1.0572360935185383e-05, + "loss": 0.9614, + "step": 44034 + }, + { + "epoch": 2.83, + "learning_rate": 1.0572146246162434e-05, + "loss": 0.8644, + "step": 44035 + }, + { + "epoch": 2.83, + "learning_rate": 1.0571931557139485e-05, + "loss": 0.9172, + "step": 44036 + }, + { + "epoch": 2.83, + "learning_rate": 1.0571716868116535e-05, + "loss": 0.8114, + "step": 44037 + }, + { + "epoch": 2.83, + "learning_rate": 1.0571502179093584e-05, + "loss": 0.9298, + "step": 44038 + }, + { + "epoch": 2.83, + "learning_rate": 1.0571287490070635e-05, + "loss": 0.9155, + "step": 44039 + }, + { + "epoch": 2.83, + "learning_rate": 1.0571072801047684e-05, + "loss": 0.917, + "step": 44040 + }, + { + "epoch": 2.83, + "learning_rate": 1.0570858112024733e-05, + "loss": 0.8703, + "step": 44041 + }, + { + "epoch": 2.83, + "learning_rate": 1.0570643423001784e-05, + "loss": 0.8752, + "step": 44042 + }, + { + "epoch": 2.83, + "learning_rate": 1.0570428733978833e-05, + "loss": 0.826, + "step": 44043 + }, + { + "epoch": 2.83, + "learning_rate": 1.0570214044955881e-05, + "loss": 0.8278, + "step": 44044 + }, + { + "epoch": 2.83, + "learning_rate": 1.0569999355932932e-05, + "loss": 0.8521, + "step": 44045 + }, + { + "epoch": 2.83, + "learning_rate": 1.0569784666909981e-05, + "loss": 0.9214, + "step": 44046 + }, + { + "epoch": 2.83, + "learning_rate": 1.056956997788703e-05, + "loss": 0.8768, + "step": 44047 + }, + { + "epoch": 2.83, + "learning_rate": 1.056935528886408e-05, + "loss": 0.8798, + "step": 44048 + }, + { + "epoch": 2.83, + "learning_rate": 1.056914059984113e-05, + "loss": 0.9456, + "step": 44049 + }, + { + "epoch": 2.83, + "learning_rate": 1.056892591081818e-05, + "loss": 0.8408, + "step": 44050 + }, + { + "epoch": 2.83, + "learning_rate": 1.0568711221795231e-05, + "loss": 0.8519, + "step": 44051 + }, + { + "epoch": 2.83, + "learning_rate": 1.0568496532772282e-05, + "loss": 0.9172, + "step": 44052 + }, + { + "epoch": 2.83, + "learning_rate": 1.056828184374933e-05, + "loss": 0.8697, + "step": 44053 + }, + { + "epoch": 2.83, + "learning_rate": 1.056806715472638e-05, + "loss": 0.8495, + "step": 44054 + }, + { + "epoch": 2.83, + "learning_rate": 1.056785246570343e-05, + "loss": 0.9109, + "step": 44055 + }, + { + "epoch": 2.83, + "learning_rate": 1.056763777668048e-05, + "loss": 0.8947, + "step": 44056 + }, + { + "epoch": 2.83, + "learning_rate": 1.0567423087657528e-05, + "loss": 0.9111, + "step": 44057 + }, + { + "epoch": 2.83, + "learning_rate": 1.0567208398634579e-05, + "loss": 0.9153, + "step": 44058 + }, + { + "epoch": 2.83, + "learning_rate": 1.0566993709611628e-05, + "loss": 0.8032, + "step": 44059 + }, + { + "epoch": 2.83, + "learning_rate": 1.0566779020588678e-05, + "loss": 0.8917, + "step": 44060 + }, + { + "epoch": 2.83, + "learning_rate": 1.0566564331565727e-05, + "loss": 0.8828, + "step": 44061 + }, + { + "epoch": 2.83, + "learning_rate": 1.0566349642542776e-05, + "loss": 0.9326, + "step": 44062 + }, + { + "epoch": 2.83, + "learning_rate": 1.0566134953519827e-05, + "loss": 0.8664, + "step": 44063 + }, + { + "epoch": 2.83, + "learning_rate": 1.0565920264496876e-05, + "loss": 0.955, + "step": 44064 + }, + { + "epoch": 2.84, + "learning_rate": 1.0565705575473928e-05, + "loss": 0.9275, + "step": 44065 + }, + { + "epoch": 2.84, + "learning_rate": 1.0565490886450977e-05, + "loss": 0.8601, + "step": 44066 + }, + { + "epoch": 2.84, + "learning_rate": 1.0565276197428026e-05, + "loss": 0.8023, + "step": 44067 + }, + { + "epoch": 2.84, + "learning_rate": 1.0565061508405077e-05, + "loss": 0.9413, + "step": 44068 + }, + { + "epoch": 2.84, + "learning_rate": 1.0564846819382126e-05, + "loss": 0.8507, + "step": 44069 + }, + { + "epoch": 2.84, + "learning_rate": 1.0564632130359177e-05, + "loss": 0.8628, + "step": 44070 + }, + { + "epoch": 2.84, + "learning_rate": 1.0564417441336226e-05, + "loss": 0.916, + "step": 44071 + }, + { + "epoch": 2.84, + "learning_rate": 1.0564202752313275e-05, + "loss": 0.9123, + "step": 44072 + }, + { + "epoch": 2.84, + "learning_rate": 1.0563988063290325e-05, + "loss": 0.8834, + "step": 44073 + }, + { + "epoch": 2.84, + "learning_rate": 1.0563773374267374e-05, + "loss": 0.8925, + "step": 44074 + }, + { + "epoch": 2.84, + "learning_rate": 1.0563558685244423e-05, + "loss": 0.9266, + "step": 44075 + }, + { + "epoch": 2.84, + "learning_rate": 1.0563343996221474e-05, + "loss": 0.8742, + "step": 44076 + }, + { + "epoch": 2.84, + "learning_rate": 1.0563129307198523e-05, + "loss": 0.8862, + "step": 44077 + }, + { + "epoch": 2.84, + "learning_rate": 1.0562914618175573e-05, + "loss": 0.9104, + "step": 44078 + }, + { + "epoch": 2.84, + "learning_rate": 1.0562699929152622e-05, + "loss": 0.8848, + "step": 44079 + }, + { + "epoch": 2.84, + "learning_rate": 1.0562485240129675e-05, + "loss": 0.9184, + "step": 44080 + }, + { + "epoch": 2.84, + "learning_rate": 1.0562270551106724e-05, + "loss": 0.8813, + "step": 44081 + }, + { + "epoch": 2.84, + "learning_rate": 1.0562055862083773e-05, + "loss": 0.8541, + "step": 44082 + }, + { + "epoch": 2.84, + "learning_rate": 1.0561841173060823e-05, + "loss": 0.9482, + "step": 44083 + }, + { + "epoch": 2.84, + "learning_rate": 1.0561626484037872e-05, + "loss": 0.874, + "step": 44084 + }, + { + "epoch": 2.84, + "learning_rate": 1.0561411795014921e-05, + "loss": 0.9297, + "step": 44085 + }, + { + "epoch": 2.84, + "learning_rate": 1.0561197105991972e-05, + "loss": 0.9135, + "step": 44086 + }, + { + "epoch": 2.84, + "learning_rate": 1.0560982416969021e-05, + "loss": 0.907, + "step": 44087 + }, + { + "epoch": 2.84, + "learning_rate": 1.0560767727946072e-05, + "loss": 0.8206, + "step": 44088 + }, + { + "epoch": 2.84, + "learning_rate": 1.056055303892312e-05, + "loss": 0.8595, + "step": 44089 + }, + { + "epoch": 2.84, + "learning_rate": 1.056033834990017e-05, + "loss": 0.9032, + "step": 44090 + }, + { + "epoch": 2.84, + "learning_rate": 1.056012366087722e-05, + "loss": 0.9228, + "step": 44091 + }, + { + "epoch": 2.84, + "learning_rate": 1.0559908971854269e-05, + "loss": 0.8925, + "step": 44092 + }, + { + "epoch": 2.84, + "learning_rate": 1.0559694282831318e-05, + "loss": 0.8344, + "step": 44093 + }, + { + "epoch": 2.84, + "learning_rate": 1.055947959380837e-05, + "loss": 0.8825, + "step": 44094 + }, + { + "epoch": 2.84, + "learning_rate": 1.055926490478542e-05, + "loss": 0.886, + "step": 44095 + }, + { + "epoch": 2.84, + "learning_rate": 1.055905021576247e-05, + "loss": 0.892, + "step": 44096 + }, + { + "epoch": 2.84, + "learning_rate": 1.0558835526739519e-05, + "loss": 0.907, + "step": 44097 + }, + { + "epoch": 2.84, + "learning_rate": 1.055862083771657e-05, + "loss": 0.9134, + "step": 44098 + }, + { + "epoch": 2.84, + "learning_rate": 1.0558406148693619e-05, + "loss": 0.9185, + "step": 44099 + }, + { + "epoch": 2.84, + "learning_rate": 1.0558191459670668e-05, + "loss": 0.8683, + "step": 44100 + }, + { + "epoch": 2.84, + "learning_rate": 1.0557976770647718e-05, + "loss": 0.9033, + "step": 44101 + }, + { + "epoch": 2.84, + "learning_rate": 1.0557762081624767e-05, + "loss": 0.9422, + "step": 44102 + }, + { + "epoch": 2.84, + "learning_rate": 1.0557547392601816e-05, + "loss": 0.9323, + "step": 44103 + }, + { + "epoch": 2.84, + "learning_rate": 1.0557332703578867e-05, + "loss": 0.9494, + "step": 44104 + }, + { + "epoch": 2.84, + "learning_rate": 1.0557118014555916e-05, + "loss": 0.8396, + "step": 44105 + }, + { + "epoch": 2.84, + "learning_rate": 1.0556903325532965e-05, + "loss": 0.9333, + "step": 44106 + }, + { + "epoch": 2.84, + "learning_rate": 1.0556688636510015e-05, + "loss": 0.8845, + "step": 44107 + }, + { + "epoch": 2.84, + "learning_rate": 1.0556473947487064e-05, + "loss": 0.862, + "step": 44108 + }, + { + "epoch": 2.84, + "learning_rate": 1.0556259258464117e-05, + "loss": 0.8839, + "step": 44109 + }, + { + "epoch": 2.84, + "learning_rate": 1.0556044569441166e-05, + "loss": 0.9115, + "step": 44110 + }, + { + "epoch": 2.84, + "learning_rate": 1.0555829880418216e-05, + "loss": 0.8407, + "step": 44111 + }, + { + "epoch": 2.84, + "learning_rate": 1.0555615191395265e-05, + "loss": 0.834, + "step": 44112 + }, + { + "epoch": 2.84, + "learning_rate": 1.0555400502372314e-05, + "loss": 0.9006, + "step": 44113 + }, + { + "epoch": 2.84, + "learning_rate": 1.0555185813349365e-05, + "loss": 0.9805, + "step": 44114 + }, + { + "epoch": 2.84, + "learning_rate": 1.0554971124326414e-05, + "loss": 0.8561, + "step": 44115 + }, + { + "epoch": 2.84, + "learning_rate": 1.0554756435303463e-05, + "loss": 0.9113, + "step": 44116 + }, + { + "epoch": 2.84, + "learning_rate": 1.0554541746280514e-05, + "loss": 1.0407, + "step": 44117 + }, + { + "epoch": 2.84, + "learning_rate": 1.0554327057257562e-05, + "loss": 0.8758, + "step": 44118 + }, + { + "epoch": 2.84, + "learning_rate": 1.0554112368234613e-05, + "loss": 0.903, + "step": 44119 + }, + { + "epoch": 2.84, + "learning_rate": 1.0553897679211662e-05, + "loss": 0.9523, + "step": 44120 + }, + { + "epoch": 2.84, + "learning_rate": 1.0553682990188711e-05, + "loss": 0.9116, + "step": 44121 + }, + { + "epoch": 2.84, + "learning_rate": 1.0553468301165762e-05, + "loss": 0.8759, + "step": 44122 + }, + { + "epoch": 2.84, + "learning_rate": 1.0553253612142812e-05, + "loss": 0.9486, + "step": 44123 + }, + { + "epoch": 2.84, + "learning_rate": 1.0553038923119863e-05, + "loss": 0.8074, + "step": 44124 + }, + { + "epoch": 2.84, + "learning_rate": 1.0552824234096912e-05, + "loss": 0.8949, + "step": 44125 + }, + { + "epoch": 2.84, + "learning_rate": 1.0552609545073963e-05, + "loss": 0.925, + "step": 44126 + }, + { + "epoch": 2.84, + "learning_rate": 1.0552394856051012e-05, + "loss": 0.9113, + "step": 44127 + }, + { + "epoch": 2.84, + "learning_rate": 1.055218016702806e-05, + "loss": 0.8824, + "step": 44128 + }, + { + "epoch": 2.84, + "learning_rate": 1.0551965478005111e-05, + "loss": 0.9446, + "step": 44129 + }, + { + "epoch": 2.84, + "learning_rate": 1.055175078898216e-05, + "loss": 0.8943, + "step": 44130 + }, + { + "epoch": 2.84, + "learning_rate": 1.055153609995921e-05, + "loss": 0.8272, + "step": 44131 + }, + { + "epoch": 2.84, + "learning_rate": 1.055132141093626e-05, + "loss": 0.9247, + "step": 44132 + }, + { + "epoch": 2.84, + "learning_rate": 1.0551106721913309e-05, + "loss": 0.8647, + "step": 44133 + }, + { + "epoch": 2.84, + "learning_rate": 1.0550892032890358e-05, + "loss": 0.8801, + "step": 44134 + }, + { + "epoch": 2.84, + "learning_rate": 1.0550677343867408e-05, + "loss": 0.934, + "step": 44135 + }, + { + "epoch": 2.84, + "learning_rate": 1.0550462654844457e-05, + "loss": 0.9263, + "step": 44136 + }, + { + "epoch": 2.84, + "learning_rate": 1.0550247965821508e-05, + "loss": 0.957, + "step": 44137 + }, + { + "epoch": 2.84, + "learning_rate": 1.0550033276798559e-05, + "loss": 0.9162, + "step": 44138 + }, + { + "epoch": 2.84, + "learning_rate": 1.054981858777561e-05, + "loss": 0.8968, + "step": 44139 + }, + { + "epoch": 2.84, + "learning_rate": 1.0549603898752658e-05, + "loss": 0.8797, + "step": 44140 + }, + { + "epoch": 2.84, + "learning_rate": 1.0549389209729707e-05, + "loss": 0.8981, + "step": 44141 + }, + { + "epoch": 2.84, + "learning_rate": 1.0549174520706758e-05, + "loss": 0.8903, + "step": 44142 + }, + { + "epoch": 2.84, + "learning_rate": 1.0548959831683807e-05, + "loss": 0.8827, + "step": 44143 + }, + { + "epoch": 2.84, + "learning_rate": 1.0548745142660856e-05, + "loss": 0.9892, + "step": 44144 + }, + { + "epoch": 2.84, + "learning_rate": 1.0548530453637907e-05, + "loss": 0.911, + "step": 44145 + }, + { + "epoch": 2.84, + "learning_rate": 1.0548315764614956e-05, + "loss": 0.9213, + "step": 44146 + }, + { + "epoch": 2.84, + "learning_rate": 1.0548101075592006e-05, + "loss": 0.8493, + "step": 44147 + }, + { + "epoch": 2.84, + "learning_rate": 1.0547886386569055e-05, + "loss": 0.9221, + "step": 44148 + }, + { + "epoch": 2.84, + "learning_rate": 1.0547671697546104e-05, + "loss": 0.8758, + "step": 44149 + }, + { + "epoch": 2.84, + "learning_rate": 1.0547457008523155e-05, + "loss": 0.8692, + "step": 44150 + }, + { + "epoch": 2.84, + "learning_rate": 1.0547242319500204e-05, + "loss": 0.9272, + "step": 44151 + }, + { + "epoch": 2.84, + "learning_rate": 1.0547027630477256e-05, + "loss": 0.8191, + "step": 44152 + }, + { + "epoch": 2.84, + "learning_rate": 1.0546812941454305e-05, + "loss": 0.8877, + "step": 44153 + }, + { + "epoch": 2.84, + "learning_rate": 1.0546598252431354e-05, + "loss": 0.9156, + "step": 44154 + }, + { + "epoch": 2.84, + "learning_rate": 1.0546383563408405e-05, + "loss": 0.8767, + "step": 44155 + }, + { + "epoch": 2.84, + "learning_rate": 1.0546168874385454e-05, + "loss": 0.8991, + "step": 44156 + }, + { + "epoch": 2.84, + "learning_rate": 1.0545954185362504e-05, + "loss": 0.8752, + "step": 44157 + }, + { + "epoch": 2.84, + "learning_rate": 1.0545739496339553e-05, + "loss": 0.8761, + "step": 44158 + }, + { + "epoch": 2.84, + "learning_rate": 1.0545524807316602e-05, + "loss": 0.8572, + "step": 44159 + }, + { + "epoch": 2.84, + "learning_rate": 1.0545310118293653e-05, + "loss": 0.9437, + "step": 44160 + }, + { + "epoch": 2.84, + "learning_rate": 1.0545095429270702e-05, + "loss": 0.9596, + "step": 44161 + }, + { + "epoch": 2.84, + "learning_rate": 1.054488074024775e-05, + "loss": 0.9416, + "step": 44162 + }, + { + "epoch": 2.84, + "learning_rate": 1.0544666051224801e-05, + "loss": 0.9261, + "step": 44163 + }, + { + "epoch": 2.84, + "learning_rate": 1.054445136220185e-05, + "loss": 0.9141, + "step": 44164 + }, + { + "epoch": 2.84, + "learning_rate": 1.0544236673178901e-05, + "loss": 0.8751, + "step": 44165 + }, + { + "epoch": 2.84, + "learning_rate": 1.054402198415595e-05, + "loss": 0.8547, + "step": 44166 + }, + { + "epoch": 2.84, + "learning_rate": 1.0543807295133002e-05, + "loss": 0.892, + "step": 44167 + }, + { + "epoch": 2.84, + "learning_rate": 1.0543592606110051e-05, + "loss": 0.923, + "step": 44168 + }, + { + "epoch": 2.84, + "learning_rate": 1.05433779170871e-05, + "loss": 0.9203, + "step": 44169 + }, + { + "epoch": 2.84, + "learning_rate": 1.0543163228064151e-05, + "loss": 0.8976, + "step": 44170 + }, + { + "epoch": 2.84, + "learning_rate": 1.05429485390412e-05, + "loss": 0.9113, + "step": 44171 + }, + { + "epoch": 2.84, + "learning_rate": 1.0542733850018249e-05, + "loss": 0.9148, + "step": 44172 + }, + { + "epoch": 2.84, + "learning_rate": 1.05425191609953e-05, + "loss": 0.8196, + "step": 44173 + }, + { + "epoch": 2.84, + "learning_rate": 1.0542304471972349e-05, + "loss": 0.8934, + "step": 44174 + }, + { + "epoch": 2.84, + "learning_rate": 1.05420897829494e-05, + "loss": 0.9587, + "step": 44175 + }, + { + "epoch": 2.84, + "learning_rate": 1.0541875093926448e-05, + "loss": 0.9373, + "step": 44176 + }, + { + "epoch": 2.84, + "learning_rate": 1.0541660404903497e-05, + "loss": 0.8603, + "step": 44177 + }, + { + "epoch": 2.84, + "learning_rate": 1.0541445715880548e-05, + "loss": 0.8301, + "step": 44178 + }, + { + "epoch": 2.84, + "learning_rate": 1.0541231026857597e-05, + "loss": 0.9288, + "step": 44179 + }, + { + "epoch": 2.84, + "learning_rate": 1.0541016337834646e-05, + "loss": 0.8882, + "step": 44180 + }, + { + "epoch": 2.84, + "learning_rate": 1.0540801648811698e-05, + "loss": 0.8893, + "step": 44181 + }, + { + "epoch": 2.84, + "learning_rate": 1.0540586959788747e-05, + "loss": 0.9159, + "step": 44182 + }, + { + "epoch": 2.84, + "learning_rate": 1.0540372270765798e-05, + "loss": 0.8568, + "step": 44183 + }, + { + "epoch": 2.84, + "learning_rate": 1.0540157581742847e-05, + "loss": 0.9086, + "step": 44184 + }, + { + "epoch": 2.84, + "learning_rate": 1.0539942892719897e-05, + "loss": 0.9299, + "step": 44185 + }, + { + "epoch": 2.84, + "learning_rate": 1.0539728203696946e-05, + "loss": 0.8663, + "step": 44186 + }, + { + "epoch": 2.84, + "learning_rate": 1.0539513514673995e-05, + "loss": 0.898, + "step": 44187 + }, + { + "epoch": 2.84, + "learning_rate": 1.0539298825651046e-05, + "loss": 0.9548, + "step": 44188 + }, + { + "epoch": 2.84, + "learning_rate": 1.0539084136628095e-05, + "loss": 0.8315, + "step": 44189 + }, + { + "epoch": 2.84, + "learning_rate": 1.0538869447605144e-05, + "loss": 0.8759, + "step": 44190 + }, + { + "epoch": 2.84, + "learning_rate": 1.0538654758582195e-05, + "loss": 0.8982, + "step": 44191 + }, + { + "epoch": 2.84, + "learning_rate": 1.0538440069559244e-05, + "loss": 0.8562, + "step": 44192 + }, + { + "epoch": 2.84, + "learning_rate": 1.0538225380536292e-05, + "loss": 0.9007, + "step": 44193 + }, + { + "epoch": 2.84, + "learning_rate": 1.0538010691513343e-05, + "loss": 0.8843, + "step": 44194 + }, + { + "epoch": 2.84, + "learning_rate": 1.0537796002490392e-05, + "loss": 0.8966, + "step": 44195 + }, + { + "epoch": 2.84, + "learning_rate": 1.0537581313467444e-05, + "loss": 0.9545, + "step": 44196 + }, + { + "epoch": 2.84, + "learning_rate": 1.0537366624444493e-05, + "loss": 0.8945, + "step": 44197 + }, + { + "epoch": 2.84, + "learning_rate": 1.0537151935421544e-05, + "loss": 0.9089, + "step": 44198 + }, + { + "epoch": 2.84, + "learning_rate": 1.0536937246398593e-05, + "loss": 0.8813, + "step": 44199 + }, + { + "epoch": 2.84, + "learning_rate": 1.0536722557375642e-05, + "loss": 0.9505, + "step": 44200 + }, + { + "epoch": 2.84, + "learning_rate": 1.0536507868352693e-05, + "loss": 0.9394, + "step": 44201 + }, + { + "epoch": 2.84, + "learning_rate": 1.0536293179329742e-05, + "loss": 0.9125, + "step": 44202 + }, + { + "epoch": 2.84, + "learning_rate": 1.053607849030679e-05, + "loss": 0.8392, + "step": 44203 + }, + { + "epoch": 2.84, + "learning_rate": 1.0535863801283841e-05, + "loss": 0.8661, + "step": 44204 + }, + { + "epoch": 2.84, + "learning_rate": 1.053564911226089e-05, + "loss": 0.8443, + "step": 44205 + }, + { + "epoch": 2.84, + "learning_rate": 1.0535434423237941e-05, + "loss": 0.9366, + "step": 44206 + }, + { + "epoch": 2.84, + "learning_rate": 1.053521973421499e-05, + "loss": 1.0073, + "step": 44207 + }, + { + "epoch": 2.84, + "learning_rate": 1.0535005045192039e-05, + "loss": 0.894, + "step": 44208 + }, + { + "epoch": 2.84, + "learning_rate": 1.053479035616909e-05, + "loss": 0.9426, + "step": 44209 + }, + { + "epoch": 2.84, + "learning_rate": 1.053457566714614e-05, + "loss": 0.858, + "step": 44210 + }, + { + "epoch": 2.84, + "learning_rate": 1.053436097812319e-05, + "loss": 0.8577, + "step": 44211 + }, + { + "epoch": 2.84, + "learning_rate": 1.053414628910024e-05, + "loss": 0.8935, + "step": 44212 + }, + { + "epoch": 2.84, + "learning_rate": 1.0533931600077289e-05, + "loss": 0.8945, + "step": 44213 + }, + { + "epoch": 2.84, + "learning_rate": 1.053371691105434e-05, + "loss": 0.8928, + "step": 44214 + }, + { + "epoch": 2.84, + "learning_rate": 1.0533502222031388e-05, + "loss": 1.0269, + "step": 44215 + }, + { + "epoch": 2.84, + "learning_rate": 1.0533287533008439e-05, + "loss": 0.9465, + "step": 44216 + }, + { + "epoch": 2.84, + "learning_rate": 1.0533072843985488e-05, + "loss": 0.9435, + "step": 44217 + }, + { + "epoch": 2.84, + "learning_rate": 1.0532858154962537e-05, + "loss": 0.8531, + "step": 44218 + }, + { + "epoch": 2.84, + "learning_rate": 1.0532643465939588e-05, + "loss": 0.9139, + "step": 44219 + }, + { + "epoch": 2.85, + "learning_rate": 1.0532428776916637e-05, + "loss": 0.8406, + "step": 44220 + }, + { + "epoch": 2.85, + "learning_rate": 1.0532214087893686e-05, + "loss": 0.8918, + "step": 44221 + }, + { + "epoch": 2.85, + "learning_rate": 1.0531999398870736e-05, + "loss": 0.8865, + "step": 44222 + }, + { + "epoch": 2.85, + "learning_rate": 1.0531784709847785e-05, + "loss": 0.8794, + "step": 44223 + }, + { + "epoch": 2.85, + "learning_rate": 1.0531570020824836e-05, + "loss": 0.9231, + "step": 44224 + }, + { + "epoch": 2.85, + "learning_rate": 1.0531355331801886e-05, + "loss": 0.9352, + "step": 44225 + }, + { + "epoch": 2.85, + "learning_rate": 1.0531140642778937e-05, + "loss": 0.8483, + "step": 44226 + }, + { + "epoch": 2.85, + "learning_rate": 1.0530925953755986e-05, + "loss": 0.9103, + "step": 44227 + }, + { + "epoch": 2.85, + "learning_rate": 1.0530711264733035e-05, + "loss": 0.8704, + "step": 44228 + }, + { + "epoch": 2.85, + "learning_rate": 1.0530496575710086e-05, + "loss": 0.9515, + "step": 44229 + }, + { + "epoch": 2.85, + "learning_rate": 1.0530281886687135e-05, + "loss": 0.9, + "step": 44230 + }, + { + "epoch": 2.85, + "learning_rate": 1.0530067197664184e-05, + "loss": 0.8684, + "step": 44231 + }, + { + "epoch": 2.85, + "learning_rate": 1.0529852508641234e-05, + "loss": 0.85, + "step": 44232 + }, + { + "epoch": 2.85, + "learning_rate": 1.0529637819618283e-05, + "loss": 0.9186, + "step": 44233 + }, + { + "epoch": 2.85, + "learning_rate": 1.0529423130595334e-05, + "loss": 0.9225, + "step": 44234 + }, + { + "epoch": 2.85, + "learning_rate": 1.0529208441572383e-05, + "loss": 0.8703, + "step": 44235 + }, + { + "epoch": 2.85, + "learning_rate": 1.0528993752549432e-05, + "loss": 0.9086, + "step": 44236 + }, + { + "epoch": 2.85, + "learning_rate": 1.0528779063526483e-05, + "loss": 0.9246, + "step": 44237 + }, + { + "epoch": 2.85, + "learning_rate": 1.0528564374503531e-05, + "loss": 0.8494, + "step": 44238 + }, + { + "epoch": 2.85, + "learning_rate": 1.0528349685480584e-05, + "loss": 0.941, + "step": 44239 + }, + { + "epoch": 2.85, + "learning_rate": 1.0528134996457633e-05, + "loss": 0.8661, + "step": 44240 + }, + { + "epoch": 2.85, + "learning_rate": 1.0527920307434682e-05, + "loss": 0.9669, + "step": 44241 + }, + { + "epoch": 2.85, + "learning_rate": 1.0527705618411732e-05, + "loss": 0.8613, + "step": 44242 + }, + { + "epoch": 2.85, + "learning_rate": 1.0527490929388781e-05, + "loss": 0.8653, + "step": 44243 + }, + { + "epoch": 2.85, + "learning_rate": 1.0527276240365832e-05, + "loss": 0.9469, + "step": 44244 + }, + { + "epoch": 2.85, + "learning_rate": 1.0527061551342881e-05, + "loss": 0.954, + "step": 44245 + }, + { + "epoch": 2.85, + "learning_rate": 1.052684686231993e-05, + "loss": 0.7543, + "step": 44246 + }, + { + "epoch": 2.85, + "learning_rate": 1.052663217329698e-05, + "loss": 0.9423, + "step": 44247 + }, + { + "epoch": 2.85, + "learning_rate": 1.052641748427403e-05, + "loss": 0.8972, + "step": 44248 + }, + { + "epoch": 2.85, + "learning_rate": 1.0526202795251079e-05, + "loss": 0.8723, + "step": 44249 + }, + { + "epoch": 2.85, + "learning_rate": 1.052598810622813e-05, + "loss": 0.8963, + "step": 44250 + }, + { + "epoch": 2.85, + "learning_rate": 1.0525773417205178e-05, + "loss": 0.9548, + "step": 44251 + }, + { + "epoch": 2.85, + "learning_rate": 1.0525558728182229e-05, + "loss": 0.8954, + "step": 44252 + }, + { + "epoch": 2.85, + "learning_rate": 1.0525344039159278e-05, + "loss": 0.9102, + "step": 44253 + }, + { + "epoch": 2.85, + "learning_rate": 1.052512935013633e-05, + "loss": 0.9712, + "step": 44254 + }, + { + "epoch": 2.85, + "learning_rate": 1.0524914661113379e-05, + "loss": 0.9562, + "step": 44255 + }, + { + "epoch": 2.85, + "learning_rate": 1.0524699972090428e-05, + "loss": 0.947, + "step": 44256 + }, + { + "epoch": 2.85, + "learning_rate": 1.0524485283067479e-05, + "loss": 0.9239, + "step": 44257 + }, + { + "epoch": 2.85, + "learning_rate": 1.0524270594044528e-05, + "loss": 0.8457, + "step": 44258 + }, + { + "epoch": 2.85, + "learning_rate": 1.0524055905021577e-05, + "loss": 0.8698, + "step": 44259 + }, + { + "epoch": 2.85, + "learning_rate": 1.0523841215998627e-05, + "loss": 0.9035, + "step": 44260 + }, + { + "epoch": 2.85, + "learning_rate": 1.0523626526975676e-05, + "loss": 0.9162, + "step": 44261 + }, + { + "epoch": 2.85, + "learning_rate": 1.0523411837952727e-05, + "loss": 0.8626, + "step": 44262 + }, + { + "epoch": 2.85, + "learning_rate": 1.0523197148929776e-05, + "loss": 0.905, + "step": 44263 + }, + { + "epoch": 2.85, + "learning_rate": 1.0522982459906825e-05, + "loss": 0.8906, + "step": 44264 + }, + { + "epoch": 2.85, + "learning_rate": 1.0522767770883876e-05, + "loss": 0.9232, + "step": 44265 + }, + { + "epoch": 2.85, + "learning_rate": 1.0522553081860925e-05, + "loss": 0.8704, + "step": 44266 + }, + { + "epoch": 2.85, + "learning_rate": 1.0522338392837973e-05, + "loss": 0.8452, + "step": 44267 + }, + { + "epoch": 2.85, + "learning_rate": 1.0522123703815026e-05, + "loss": 0.8982, + "step": 44268 + }, + { + "epoch": 2.85, + "learning_rate": 1.0521909014792075e-05, + "loss": 0.8799, + "step": 44269 + }, + { + "epoch": 2.85, + "learning_rate": 1.0521694325769125e-05, + "loss": 0.9198, + "step": 44270 + }, + { + "epoch": 2.85, + "learning_rate": 1.0521479636746174e-05, + "loss": 0.8561, + "step": 44271 + }, + { + "epoch": 2.85, + "learning_rate": 1.0521264947723225e-05, + "loss": 0.8487, + "step": 44272 + }, + { + "epoch": 2.85, + "learning_rate": 1.0521050258700274e-05, + "loss": 0.8819, + "step": 44273 + }, + { + "epoch": 2.85, + "learning_rate": 1.0520835569677323e-05, + "loss": 0.894, + "step": 44274 + }, + { + "epoch": 2.85, + "learning_rate": 1.0520620880654374e-05, + "loss": 0.9175, + "step": 44275 + }, + { + "epoch": 2.85, + "learning_rate": 1.0520406191631423e-05, + "loss": 0.8954, + "step": 44276 + }, + { + "epoch": 2.85, + "learning_rate": 1.0520191502608472e-05, + "loss": 0.8773, + "step": 44277 + }, + { + "epoch": 2.85, + "learning_rate": 1.0519976813585522e-05, + "loss": 0.8346, + "step": 44278 + }, + { + "epoch": 2.85, + "learning_rate": 1.0519762124562571e-05, + "loss": 0.9422, + "step": 44279 + }, + { + "epoch": 2.85, + "learning_rate": 1.051954743553962e-05, + "loss": 0.9386, + "step": 44280 + }, + { + "epoch": 2.85, + "learning_rate": 1.051933274651667e-05, + "loss": 0.8852, + "step": 44281 + }, + { + "epoch": 2.85, + "learning_rate": 1.051911805749372e-05, + "loss": 0.9656, + "step": 44282 + }, + { + "epoch": 2.85, + "learning_rate": 1.0518903368470772e-05, + "loss": 0.9426, + "step": 44283 + }, + { + "epoch": 2.85, + "learning_rate": 1.0518688679447821e-05, + "loss": 0.9002, + "step": 44284 + }, + { + "epoch": 2.85, + "learning_rate": 1.0518473990424872e-05, + "loss": 0.87, + "step": 44285 + }, + { + "epoch": 2.85, + "learning_rate": 1.051825930140192e-05, + "loss": 0.971, + "step": 44286 + }, + { + "epoch": 2.85, + "learning_rate": 1.051804461237897e-05, + "loss": 0.8728, + "step": 44287 + }, + { + "epoch": 2.85, + "learning_rate": 1.051782992335602e-05, + "loss": 0.9126, + "step": 44288 + }, + { + "epoch": 2.85, + "learning_rate": 1.051761523433307e-05, + "loss": 0.9111, + "step": 44289 + }, + { + "epoch": 2.85, + "learning_rate": 1.0517400545310118e-05, + "loss": 0.8772, + "step": 44290 + }, + { + "epoch": 2.85, + "learning_rate": 1.0517185856287169e-05, + "loss": 0.9036, + "step": 44291 + }, + { + "epoch": 2.85, + "learning_rate": 1.0516971167264218e-05, + "loss": 0.8633, + "step": 44292 + }, + { + "epoch": 2.85, + "learning_rate": 1.0516756478241269e-05, + "loss": 0.8403, + "step": 44293 + }, + { + "epoch": 2.85, + "learning_rate": 1.0516541789218318e-05, + "loss": 0.9468, + "step": 44294 + }, + { + "epoch": 2.85, + "learning_rate": 1.0516327100195367e-05, + "loss": 0.9006, + "step": 44295 + }, + { + "epoch": 2.85, + "learning_rate": 1.0516112411172417e-05, + "loss": 0.9222, + "step": 44296 + }, + { + "epoch": 2.85, + "learning_rate": 1.0515897722149468e-05, + "loss": 0.8858, + "step": 44297 + }, + { + "epoch": 2.85, + "learning_rate": 1.0515683033126518e-05, + "loss": 0.9827, + "step": 44298 + }, + { + "epoch": 2.85, + "learning_rate": 1.0515468344103567e-05, + "loss": 0.9887, + "step": 44299 + }, + { + "epoch": 2.85, + "learning_rate": 1.0515253655080616e-05, + "loss": 0.9022, + "step": 44300 + }, + { + "epoch": 2.85, + "learning_rate": 1.0515038966057667e-05, + "loss": 0.8745, + "step": 44301 + }, + { + "epoch": 2.85, + "learning_rate": 1.0514824277034716e-05, + "loss": 0.8984, + "step": 44302 + }, + { + "epoch": 2.85, + "learning_rate": 1.0514609588011767e-05, + "loss": 0.9358, + "step": 44303 + }, + { + "epoch": 2.85, + "learning_rate": 1.0514394898988816e-05, + "loss": 0.9051, + "step": 44304 + }, + { + "epoch": 2.85, + "learning_rate": 1.0514180209965865e-05, + "loss": 0.874, + "step": 44305 + }, + { + "epoch": 2.85, + "learning_rate": 1.0513965520942915e-05, + "loss": 0.8931, + "step": 44306 + }, + { + "epoch": 2.85, + "learning_rate": 1.0513750831919964e-05, + "loss": 0.8194, + "step": 44307 + }, + { + "epoch": 2.85, + "learning_rate": 1.0513536142897013e-05, + "loss": 0.8742, + "step": 44308 + }, + { + "epoch": 2.85, + "learning_rate": 1.0513321453874064e-05, + "loss": 0.9254, + "step": 44309 + }, + { + "epoch": 2.85, + "learning_rate": 1.0513106764851113e-05, + "loss": 0.8895, + "step": 44310 + }, + { + "epoch": 2.85, + "learning_rate": 1.0512892075828164e-05, + "loss": 0.9555, + "step": 44311 + }, + { + "epoch": 2.85, + "learning_rate": 1.0512677386805214e-05, + "loss": 0.915, + "step": 44312 + }, + { + "epoch": 2.85, + "learning_rate": 1.0512462697782265e-05, + "loss": 0.8683, + "step": 44313 + }, + { + "epoch": 2.85, + "learning_rate": 1.0512248008759314e-05, + "loss": 0.8855, + "step": 44314 + }, + { + "epoch": 2.85, + "learning_rate": 1.0512033319736363e-05, + "loss": 0.8629, + "step": 44315 + }, + { + "epoch": 2.85, + "learning_rate": 1.0511818630713413e-05, + "loss": 0.9786, + "step": 44316 + }, + { + "epoch": 2.85, + "learning_rate": 1.0511603941690462e-05, + "loss": 0.8369, + "step": 44317 + }, + { + "epoch": 2.85, + "learning_rate": 1.0511389252667511e-05, + "loss": 0.8567, + "step": 44318 + }, + { + "epoch": 2.85, + "learning_rate": 1.0511174563644562e-05, + "loss": 0.922, + "step": 44319 + }, + { + "epoch": 2.85, + "learning_rate": 1.0510959874621611e-05, + "loss": 0.8489, + "step": 44320 + }, + { + "epoch": 2.85, + "learning_rate": 1.0510745185598662e-05, + "loss": 0.8655, + "step": 44321 + }, + { + "epoch": 2.85, + "learning_rate": 1.051053049657571e-05, + "loss": 0.8926, + "step": 44322 + }, + { + "epoch": 2.85, + "learning_rate": 1.051031580755276e-05, + "loss": 0.8805, + "step": 44323 + }, + { + "epoch": 2.85, + "learning_rate": 1.051010111852981e-05, + "loss": 0.8597, + "step": 44324 + }, + { + "epoch": 2.85, + "learning_rate": 1.050988642950686e-05, + "loss": 0.8885, + "step": 44325 + }, + { + "epoch": 2.85, + "learning_rate": 1.0509671740483908e-05, + "loss": 0.8576, + "step": 44326 + }, + { + "epoch": 2.85, + "learning_rate": 1.050945705146096e-05, + "loss": 0.8932, + "step": 44327 + }, + { + "epoch": 2.85, + "learning_rate": 1.050924236243801e-05, + "loss": 0.912, + "step": 44328 + }, + { + "epoch": 2.85, + "learning_rate": 1.050902767341506e-05, + "loss": 0.8786, + "step": 44329 + }, + { + "epoch": 2.85, + "learning_rate": 1.0508812984392109e-05, + "loss": 0.9407, + "step": 44330 + }, + { + "epoch": 2.85, + "learning_rate": 1.050859829536916e-05, + "loss": 0.898, + "step": 44331 + }, + { + "epoch": 2.85, + "learning_rate": 1.0508383606346209e-05, + "loss": 0.8713, + "step": 44332 + }, + { + "epoch": 2.85, + "learning_rate": 1.0508168917323258e-05, + "loss": 0.8369, + "step": 44333 + }, + { + "epoch": 2.85, + "learning_rate": 1.0507954228300308e-05, + "loss": 0.939, + "step": 44334 + }, + { + "epoch": 2.85, + "learning_rate": 1.0507739539277357e-05, + "loss": 0.861, + "step": 44335 + }, + { + "epoch": 2.85, + "learning_rate": 1.0507524850254406e-05, + "loss": 0.8776, + "step": 44336 + }, + { + "epoch": 2.85, + "learning_rate": 1.0507310161231457e-05, + "loss": 0.9602, + "step": 44337 + }, + { + "epoch": 2.85, + "learning_rate": 1.0507095472208506e-05, + "loss": 0.9215, + "step": 44338 + }, + { + "epoch": 2.85, + "learning_rate": 1.0506880783185557e-05, + "loss": 0.8911, + "step": 44339 + }, + { + "epoch": 2.85, + "learning_rate": 1.0506666094162606e-05, + "loss": 0.9178, + "step": 44340 + }, + { + "epoch": 2.85, + "learning_rate": 1.0506451405139658e-05, + "loss": 0.897, + "step": 44341 + }, + { + "epoch": 2.85, + "learning_rate": 1.0506236716116707e-05, + "loss": 0.9134, + "step": 44342 + }, + { + "epoch": 2.85, + "learning_rate": 1.0506022027093756e-05, + "loss": 0.91, + "step": 44343 + }, + { + "epoch": 2.85, + "learning_rate": 1.0505807338070806e-05, + "loss": 0.8866, + "step": 44344 + }, + { + "epoch": 2.85, + "learning_rate": 1.0505592649047855e-05, + "loss": 0.8486, + "step": 44345 + }, + { + "epoch": 2.85, + "learning_rate": 1.0505377960024904e-05, + "loss": 0.9402, + "step": 44346 + }, + { + "epoch": 2.85, + "learning_rate": 1.0505163271001955e-05, + "loss": 0.9065, + "step": 44347 + }, + { + "epoch": 2.85, + "learning_rate": 1.0504948581979004e-05, + "loss": 0.941, + "step": 44348 + }, + { + "epoch": 2.85, + "learning_rate": 1.0504733892956055e-05, + "loss": 0.9019, + "step": 44349 + }, + { + "epoch": 2.85, + "learning_rate": 1.0504519203933104e-05, + "loss": 0.8676, + "step": 44350 + }, + { + "epoch": 2.85, + "learning_rate": 1.0504304514910153e-05, + "loss": 0.8767, + "step": 44351 + }, + { + "epoch": 2.85, + "learning_rate": 1.0504089825887203e-05, + "loss": 0.7943, + "step": 44352 + }, + { + "epoch": 2.85, + "learning_rate": 1.0503875136864252e-05, + "loss": 0.925, + "step": 44353 + }, + { + "epoch": 2.85, + "learning_rate": 1.0503660447841301e-05, + "loss": 0.9073, + "step": 44354 + }, + { + "epoch": 2.85, + "learning_rate": 1.0503445758818352e-05, + "loss": 0.9443, + "step": 44355 + }, + { + "epoch": 2.85, + "learning_rate": 1.0503231069795403e-05, + "loss": 0.82, + "step": 44356 + }, + { + "epoch": 2.85, + "learning_rate": 1.0503016380772453e-05, + "loss": 0.9181, + "step": 44357 + }, + { + "epoch": 2.85, + "learning_rate": 1.0502801691749502e-05, + "loss": 0.908, + "step": 44358 + }, + { + "epoch": 2.85, + "learning_rate": 1.0502587002726553e-05, + "loss": 0.8113, + "step": 44359 + }, + { + "epoch": 2.85, + "learning_rate": 1.0502372313703602e-05, + "loss": 0.9341, + "step": 44360 + }, + { + "epoch": 2.85, + "learning_rate": 1.050215762468065e-05, + "loss": 0.8818, + "step": 44361 + }, + { + "epoch": 2.85, + "learning_rate": 1.0501942935657701e-05, + "loss": 0.906, + "step": 44362 + }, + { + "epoch": 2.85, + "learning_rate": 1.050172824663475e-05, + "loss": 0.8779, + "step": 44363 + }, + { + "epoch": 2.85, + "learning_rate": 1.05015135576118e-05, + "loss": 0.9446, + "step": 44364 + }, + { + "epoch": 2.85, + "learning_rate": 1.050129886858885e-05, + "loss": 0.8496, + "step": 44365 + }, + { + "epoch": 2.85, + "learning_rate": 1.0501084179565899e-05, + "loss": 0.9084, + "step": 44366 + }, + { + "epoch": 2.85, + "learning_rate": 1.0500869490542948e-05, + "loss": 0.9083, + "step": 44367 + }, + { + "epoch": 2.85, + "learning_rate": 1.0500654801519999e-05, + "loss": 0.9348, + "step": 44368 + }, + { + "epoch": 2.85, + "learning_rate": 1.0500440112497048e-05, + "loss": 0.8609, + "step": 44369 + }, + { + "epoch": 2.85, + "learning_rate": 1.05002254234741e-05, + "loss": 0.8846, + "step": 44370 + }, + { + "epoch": 2.85, + "learning_rate": 1.0500010734451149e-05, + "loss": 0.8835, + "step": 44371 + }, + { + "epoch": 2.85, + "learning_rate": 1.04997960454282e-05, + "loss": 0.9287, + "step": 44372 + }, + { + "epoch": 2.85, + "learning_rate": 1.0499581356405248e-05, + "loss": 0.9734, + "step": 44373 + }, + { + "epoch": 2.85, + "learning_rate": 1.0499366667382297e-05, + "loss": 0.8885, + "step": 44374 + }, + { + "epoch": 2.85, + "learning_rate": 1.0499151978359348e-05, + "loss": 0.8488, + "step": 44375 + }, + { + "epoch": 2.86, + "learning_rate": 1.0498937289336397e-05, + "loss": 0.8995, + "step": 44376 + }, + { + "epoch": 2.86, + "learning_rate": 1.0498722600313446e-05, + "loss": 0.9024, + "step": 44377 + }, + { + "epoch": 2.86, + "learning_rate": 1.0498507911290497e-05, + "loss": 0.9086, + "step": 44378 + }, + { + "epoch": 2.86, + "learning_rate": 1.0498293222267546e-05, + "loss": 0.9516, + "step": 44379 + }, + { + "epoch": 2.86, + "learning_rate": 1.0498078533244596e-05, + "loss": 0.8325, + "step": 44380 + }, + { + "epoch": 2.86, + "learning_rate": 1.0497863844221645e-05, + "loss": 0.9086, + "step": 44381 + }, + { + "epoch": 2.86, + "learning_rate": 1.0497649155198694e-05, + "loss": 0.9179, + "step": 44382 + }, + { + "epoch": 2.86, + "learning_rate": 1.0497434466175745e-05, + "loss": 0.9563, + "step": 44383 + }, + { + "epoch": 2.86, + "learning_rate": 1.0497219777152794e-05, + "loss": 0.8867, + "step": 44384 + }, + { + "epoch": 2.86, + "learning_rate": 1.0497005088129846e-05, + "loss": 0.9112, + "step": 44385 + }, + { + "epoch": 2.86, + "learning_rate": 1.0496790399106895e-05, + "loss": 0.8949, + "step": 44386 + }, + { + "epoch": 2.86, + "learning_rate": 1.0496575710083944e-05, + "loss": 0.9163, + "step": 44387 + }, + { + "epoch": 2.86, + "learning_rate": 1.0496361021060995e-05, + "loss": 0.8476, + "step": 44388 + }, + { + "epoch": 2.86, + "learning_rate": 1.0496146332038044e-05, + "loss": 0.8865, + "step": 44389 + }, + { + "epoch": 2.86, + "learning_rate": 1.0495931643015094e-05, + "loss": 0.8601, + "step": 44390 + }, + { + "epoch": 2.86, + "learning_rate": 1.0495716953992143e-05, + "loss": 0.8525, + "step": 44391 + }, + { + "epoch": 2.86, + "learning_rate": 1.0495502264969192e-05, + "loss": 0.8804, + "step": 44392 + }, + { + "epoch": 2.86, + "learning_rate": 1.0495287575946243e-05, + "loss": 0.9021, + "step": 44393 + }, + { + "epoch": 2.86, + "learning_rate": 1.0495072886923292e-05, + "loss": 0.8953, + "step": 44394 + }, + { + "epoch": 2.86, + "learning_rate": 1.0494858197900341e-05, + "loss": 0.9178, + "step": 44395 + }, + { + "epoch": 2.86, + "learning_rate": 1.0494643508877392e-05, + "loss": 0.8277, + "step": 44396 + }, + { + "epoch": 2.86, + "learning_rate": 1.049442881985444e-05, + "loss": 0.9013, + "step": 44397 + }, + { + "epoch": 2.86, + "learning_rate": 1.0494214130831491e-05, + "loss": 0.9327, + "step": 44398 + }, + { + "epoch": 2.86, + "learning_rate": 1.0493999441808542e-05, + "loss": 0.8564, + "step": 44399 + }, + { + "epoch": 2.86, + "learning_rate": 1.0493784752785593e-05, + "loss": 0.9377, + "step": 44400 + }, + { + "epoch": 2.86, + "learning_rate": 1.0493570063762641e-05, + "loss": 0.9508, + "step": 44401 + }, + { + "epoch": 2.86, + "learning_rate": 1.049335537473969e-05, + "loss": 0.9014, + "step": 44402 + }, + { + "epoch": 2.86, + "learning_rate": 1.0493140685716741e-05, + "loss": 0.8901, + "step": 44403 + }, + { + "epoch": 2.86, + "learning_rate": 1.049292599669379e-05, + "loss": 0.9688, + "step": 44404 + }, + { + "epoch": 2.86, + "learning_rate": 1.0492711307670839e-05, + "loss": 0.8778, + "step": 44405 + }, + { + "epoch": 2.86, + "learning_rate": 1.049249661864789e-05, + "loss": 0.9281, + "step": 44406 + }, + { + "epoch": 2.86, + "learning_rate": 1.0492281929624939e-05, + "loss": 0.8832, + "step": 44407 + }, + { + "epoch": 2.86, + "learning_rate": 1.049206724060199e-05, + "loss": 0.9024, + "step": 44408 + }, + { + "epoch": 2.86, + "learning_rate": 1.0491852551579038e-05, + "loss": 0.79, + "step": 44409 + }, + { + "epoch": 2.86, + "learning_rate": 1.0491637862556087e-05, + "loss": 0.9048, + "step": 44410 + }, + { + "epoch": 2.86, + "learning_rate": 1.0491423173533138e-05, + "loss": 0.8753, + "step": 44411 + }, + { + "epoch": 2.86, + "learning_rate": 1.0491208484510187e-05, + "loss": 0.924, + "step": 44412 + }, + { + "epoch": 2.86, + "learning_rate": 1.0490993795487236e-05, + "loss": 0.9339, + "step": 44413 + }, + { + "epoch": 2.86, + "learning_rate": 1.0490779106464288e-05, + "loss": 0.9504, + "step": 44414 + }, + { + "epoch": 2.86, + "learning_rate": 1.0490564417441337e-05, + "loss": 0.862, + "step": 44415 + }, + { + "epoch": 2.86, + "learning_rate": 1.0490349728418388e-05, + "loss": 0.8651, + "step": 44416 + }, + { + "epoch": 2.86, + "learning_rate": 1.0490135039395437e-05, + "loss": 0.9081, + "step": 44417 + }, + { + "epoch": 2.86, + "learning_rate": 1.0489920350372487e-05, + "loss": 0.921, + "step": 44418 + }, + { + "epoch": 2.86, + "learning_rate": 1.0489705661349536e-05, + "loss": 0.8951, + "step": 44419 + }, + { + "epoch": 2.86, + "learning_rate": 1.0489490972326585e-05, + "loss": 0.8858, + "step": 44420 + }, + { + "epoch": 2.86, + "learning_rate": 1.0489276283303636e-05, + "loss": 0.9324, + "step": 44421 + }, + { + "epoch": 2.86, + "learning_rate": 1.0489061594280685e-05, + "loss": 0.9059, + "step": 44422 + }, + { + "epoch": 2.86, + "learning_rate": 1.0488846905257734e-05, + "loss": 0.9041, + "step": 44423 + }, + { + "epoch": 2.86, + "learning_rate": 1.0488632216234785e-05, + "loss": 0.9434, + "step": 44424 + }, + { + "epoch": 2.86, + "learning_rate": 1.0488417527211834e-05, + "loss": 0.8624, + "step": 44425 + }, + { + "epoch": 2.86, + "learning_rate": 1.0488202838188883e-05, + "loss": 0.9249, + "step": 44426 + }, + { + "epoch": 2.86, + "learning_rate": 1.0487988149165933e-05, + "loss": 0.9238, + "step": 44427 + }, + { + "epoch": 2.86, + "learning_rate": 1.0487773460142986e-05, + "loss": 0.9083, + "step": 44428 + }, + { + "epoch": 2.86, + "learning_rate": 1.0487558771120035e-05, + "loss": 0.8479, + "step": 44429 + }, + { + "epoch": 2.86, + "learning_rate": 1.0487344082097084e-05, + "loss": 0.9318, + "step": 44430 + }, + { + "epoch": 2.86, + "learning_rate": 1.0487129393074134e-05, + "loss": 0.9809, + "step": 44431 + }, + { + "epoch": 2.86, + "learning_rate": 1.0486914704051183e-05, + "loss": 0.8764, + "step": 44432 + }, + { + "epoch": 2.86, + "learning_rate": 1.0486700015028232e-05, + "loss": 0.8252, + "step": 44433 + }, + { + "epoch": 2.86, + "learning_rate": 1.0486485326005283e-05, + "loss": 0.8523, + "step": 44434 + }, + { + "epoch": 2.86, + "learning_rate": 1.0486270636982332e-05, + "loss": 0.9002, + "step": 44435 + }, + { + "epoch": 2.86, + "learning_rate": 1.0486055947959382e-05, + "loss": 0.8706, + "step": 44436 + }, + { + "epoch": 2.86, + "learning_rate": 1.0485841258936431e-05, + "loss": 0.8812, + "step": 44437 + }, + { + "epoch": 2.86, + "learning_rate": 1.048562656991348e-05, + "loss": 0.9208, + "step": 44438 + }, + { + "epoch": 2.86, + "learning_rate": 1.0485411880890531e-05, + "loss": 0.8747, + "step": 44439 + }, + { + "epoch": 2.86, + "learning_rate": 1.048519719186758e-05, + "loss": 0.9322, + "step": 44440 + }, + { + "epoch": 2.86, + "learning_rate": 1.0484982502844629e-05, + "loss": 0.8924, + "step": 44441 + }, + { + "epoch": 2.86, + "learning_rate": 1.048476781382168e-05, + "loss": 0.9522, + "step": 44442 + }, + { + "epoch": 2.86, + "learning_rate": 1.048455312479873e-05, + "loss": 0.8172, + "step": 44443 + }, + { + "epoch": 2.86, + "learning_rate": 1.0484338435775781e-05, + "loss": 0.9589, + "step": 44444 + }, + { + "epoch": 2.86, + "learning_rate": 1.048412374675283e-05, + "loss": 1.0352, + "step": 44445 + }, + { + "epoch": 2.86, + "learning_rate": 1.048390905772988e-05, + "loss": 0.9205, + "step": 44446 + }, + { + "epoch": 2.86, + "learning_rate": 1.048369436870693e-05, + "loss": 0.9262, + "step": 44447 + }, + { + "epoch": 2.86, + "learning_rate": 1.0483479679683978e-05, + "loss": 0.8926, + "step": 44448 + }, + { + "epoch": 2.86, + "learning_rate": 1.0483264990661029e-05, + "loss": 0.9391, + "step": 44449 + }, + { + "epoch": 2.86, + "learning_rate": 1.0483050301638078e-05, + "loss": 0.9178, + "step": 44450 + }, + { + "epoch": 2.86, + "learning_rate": 1.0482835612615127e-05, + "loss": 0.9226, + "step": 44451 + }, + { + "epoch": 2.86, + "learning_rate": 1.0482620923592178e-05, + "loss": 0.8808, + "step": 44452 + }, + { + "epoch": 2.86, + "learning_rate": 1.0482406234569227e-05, + "loss": 0.932, + "step": 44453 + }, + { + "epoch": 2.86, + "learning_rate": 1.0482191545546276e-05, + "loss": 0.9193, + "step": 44454 + }, + { + "epoch": 2.86, + "learning_rate": 1.0481976856523326e-05, + "loss": 0.9457, + "step": 44455 + }, + { + "epoch": 2.86, + "learning_rate": 1.0481762167500375e-05, + "loss": 0.9023, + "step": 44456 + }, + { + "epoch": 2.86, + "learning_rate": 1.0481547478477428e-05, + "loss": 0.85, + "step": 44457 + }, + { + "epoch": 2.86, + "learning_rate": 1.0481332789454477e-05, + "loss": 0.9221, + "step": 44458 + }, + { + "epoch": 2.86, + "learning_rate": 1.0481118100431527e-05, + "loss": 0.9526, + "step": 44459 + }, + { + "epoch": 2.86, + "learning_rate": 1.0480903411408576e-05, + "loss": 0.9085, + "step": 44460 + }, + { + "epoch": 2.86, + "learning_rate": 1.0480688722385625e-05, + "loss": 0.8797, + "step": 44461 + }, + { + "epoch": 2.86, + "learning_rate": 1.0480474033362676e-05, + "loss": 0.9086, + "step": 44462 + }, + { + "epoch": 2.86, + "learning_rate": 1.0480259344339725e-05, + "loss": 0.8761, + "step": 44463 + }, + { + "epoch": 2.86, + "learning_rate": 1.0480044655316774e-05, + "loss": 0.8732, + "step": 44464 + }, + { + "epoch": 2.86, + "learning_rate": 1.0479829966293824e-05, + "loss": 0.9264, + "step": 44465 + }, + { + "epoch": 2.86, + "learning_rate": 1.0479615277270873e-05, + "loss": 0.8844, + "step": 44466 + }, + { + "epoch": 2.86, + "learning_rate": 1.0479400588247924e-05, + "loss": 0.898, + "step": 44467 + }, + { + "epoch": 2.86, + "learning_rate": 1.0479185899224973e-05, + "loss": 0.9381, + "step": 44468 + }, + { + "epoch": 2.86, + "learning_rate": 1.0478971210202022e-05, + "loss": 0.8811, + "step": 44469 + }, + { + "epoch": 2.86, + "learning_rate": 1.0478756521179073e-05, + "loss": 0.9191, + "step": 44470 + }, + { + "epoch": 2.86, + "learning_rate": 1.0478541832156122e-05, + "loss": 0.8907, + "step": 44471 + }, + { + "epoch": 2.86, + "learning_rate": 1.0478327143133174e-05, + "loss": 0.8663, + "step": 44472 + }, + { + "epoch": 2.86, + "learning_rate": 1.0478112454110223e-05, + "loss": 0.8677, + "step": 44473 + }, + { + "epoch": 2.86, + "learning_rate": 1.0477897765087272e-05, + "loss": 0.9384, + "step": 44474 + }, + { + "epoch": 2.86, + "learning_rate": 1.0477683076064323e-05, + "loss": 0.8783, + "step": 44475 + }, + { + "epoch": 2.86, + "learning_rate": 1.0477468387041371e-05, + "loss": 0.8859, + "step": 44476 + }, + { + "epoch": 2.86, + "learning_rate": 1.0477253698018422e-05, + "loss": 0.8757, + "step": 44477 + }, + { + "epoch": 2.86, + "learning_rate": 1.0477039008995471e-05, + "loss": 0.9025, + "step": 44478 + }, + { + "epoch": 2.86, + "learning_rate": 1.047682431997252e-05, + "loss": 0.937, + "step": 44479 + }, + { + "epoch": 2.86, + "learning_rate": 1.047660963094957e-05, + "loss": 0.9328, + "step": 44480 + }, + { + "epoch": 2.86, + "learning_rate": 1.047639494192662e-05, + "loss": 0.9481, + "step": 44481 + }, + { + "epoch": 2.86, + "learning_rate": 1.0476180252903669e-05, + "loss": 0.8894, + "step": 44482 + }, + { + "epoch": 2.86, + "learning_rate": 1.047596556388072e-05, + "loss": 0.9134, + "step": 44483 + }, + { + "epoch": 2.86, + "learning_rate": 1.0475750874857768e-05, + "loss": 0.8077, + "step": 44484 + }, + { + "epoch": 2.86, + "learning_rate": 1.0475536185834819e-05, + "loss": 0.8788, + "step": 44485 + }, + { + "epoch": 2.86, + "learning_rate": 1.047532149681187e-05, + "loss": 0.9048, + "step": 44486 + }, + { + "epoch": 2.86, + "learning_rate": 1.047510680778892e-05, + "loss": 0.8814, + "step": 44487 + }, + { + "epoch": 2.86, + "learning_rate": 1.047489211876597e-05, + "loss": 0.921, + "step": 44488 + }, + { + "epoch": 2.86, + "learning_rate": 1.0474677429743018e-05, + "loss": 0.8087, + "step": 44489 + }, + { + "epoch": 2.86, + "learning_rate": 1.0474462740720069e-05, + "loss": 0.8947, + "step": 44490 + }, + { + "epoch": 2.86, + "learning_rate": 1.0474248051697118e-05, + "loss": 0.9109, + "step": 44491 + }, + { + "epoch": 2.86, + "learning_rate": 1.0474033362674167e-05, + "loss": 0.8551, + "step": 44492 + }, + { + "epoch": 2.86, + "learning_rate": 1.0473818673651217e-05, + "loss": 0.8799, + "step": 44493 + }, + { + "epoch": 2.86, + "learning_rate": 1.0473603984628266e-05, + "loss": 0.8981, + "step": 44494 + }, + { + "epoch": 2.86, + "learning_rate": 1.0473389295605317e-05, + "loss": 0.8521, + "step": 44495 + }, + { + "epoch": 2.86, + "learning_rate": 1.0473174606582366e-05, + "loss": 0.9063, + "step": 44496 + }, + { + "epoch": 2.86, + "learning_rate": 1.0472959917559415e-05, + "loss": 0.8783, + "step": 44497 + }, + { + "epoch": 2.86, + "learning_rate": 1.0472745228536466e-05, + "loss": 0.8967, + "step": 44498 + }, + { + "epoch": 2.86, + "learning_rate": 1.0472530539513515e-05, + "loss": 0.9184, + "step": 44499 + }, + { + "epoch": 2.86, + "learning_rate": 1.0472315850490564e-05, + "loss": 0.8766, + "step": 44500 + }, + { + "epoch": 2.86, + "learning_rate": 1.0472101161467616e-05, + "loss": 0.8999, + "step": 44501 + }, + { + "epoch": 2.86, + "learning_rate": 1.0471886472444665e-05, + "loss": 0.9551, + "step": 44502 + }, + { + "epoch": 2.86, + "learning_rate": 1.0471671783421716e-05, + "loss": 0.8961, + "step": 44503 + }, + { + "epoch": 2.86, + "learning_rate": 1.0471457094398765e-05, + "loss": 0.9275, + "step": 44504 + }, + { + "epoch": 2.86, + "learning_rate": 1.0471242405375815e-05, + "loss": 1.0024, + "step": 44505 + }, + { + "epoch": 2.86, + "learning_rate": 1.0471027716352864e-05, + "loss": 0.8634, + "step": 44506 + }, + { + "epoch": 2.86, + "learning_rate": 1.0470813027329913e-05, + "loss": 0.8644, + "step": 44507 + }, + { + "epoch": 2.86, + "learning_rate": 1.0470598338306964e-05, + "loss": 0.9207, + "step": 44508 + }, + { + "epoch": 2.86, + "learning_rate": 1.0470383649284013e-05, + "loss": 0.8702, + "step": 44509 + }, + { + "epoch": 2.86, + "learning_rate": 1.0470168960261062e-05, + "loss": 0.8745, + "step": 44510 + }, + { + "epoch": 2.86, + "learning_rate": 1.0469954271238112e-05, + "loss": 0.8292, + "step": 44511 + }, + { + "epoch": 2.86, + "learning_rate": 1.0469739582215161e-05, + "loss": 0.8754, + "step": 44512 + }, + { + "epoch": 2.86, + "learning_rate": 1.046952489319221e-05, + "loss": 0.9561, + "step": 44513 + }, + { + "epoch": 2.86, + "learning_rate": 1.0469310204169261e-05, + "loss": 0.9455, + "step": 44514 + }, + { + "epoch": 2.86, + "learning_rate": 1.0469095515146313e-05, + "loss": 0.9148, + "step": 44515 + }, + { + "epoch": 2.86, + "learning_rate": 1.0468880826123362e-05, + "loss": 0.8543, + "step": 44516 + }, + { + "epoch": 2.86, + "learning_rate": 1.0468666137100411e-05, + "loss": 0.9155, + "step": 44517 + }, + { + "epoch": 2.86, + "learning_rate": 1.0468451448077462e-05, + "loss": 0.913, + "step": 44518 + }, + { + "epoch": 2.86, + "learning_rate": 1.046823675905451e-05, + "loss": 0.9384, + "step": 44519 + }, + { + "epoch": 2.86, + "learning_rate": 1.046802207003156e-05, + "loss": 0.8674, + "step": 44520 + }, + { + "epoch": 2.86, + "learning_rate": 1.046780738100861e-05, + "loss": 0.9053, + "step": 44521 + }, + { + "epoch": 2.86, + "learning_rate": 1.046759269198566e-05, + "loss": 0.8851, + "step": 44522 + }, + { + "epoch": 2.86, + "learning_rate": 1.0467378002962708e-05, + "loss": 0.8701, + "step": 44523 + }, + { + "epoch": 2.86, + "learning_rate": 1.0467163313939759e-05, + "loss": 0.8816, + "step": 44524 + }, + { + "epoch": 2.86, + "learning_rate": 1.0466948624916808e-05, + "loss": 0.8812, + "step": 44525 + }, + { + "epoch": 2.86, + "learning_rate": 1.0466733935893859e-05, + "loss": 0.8982, + "step": 44526 + }, + { + "epoch": 2.86, + "learning_rate": 1.0466519246870908e-05, + "loss": 0.9346, + "step": 44527 + }, + { + "epoch": 2.86, + "learning_rate": 1.0466304557847957e-05, + "loss": 0.904, + "step": 44528 + }, + { + "epoch": 2.86, + "learning_rate": 1.0466089868825007e-05, + "loss": 0.8819, + "step": 44529 + }, + { + "epoch": 2.86, + "learning_rate": 1.0465875179802058e-05, + "loss": 0.9618, + "step": 44530 + }, + { + "epoch": 2.87, + "learning_rate": 1.0465660490779109e-05, + "loss": 0.9438, + "step": 44531 + }, + { + "epoch": 2.87, + "learning_rate": 1.0465445801756158e-05, + "loss": 0.9162, + "step": 44532 + }, + { + "epoch": 2.87, + "learning_rate": 1.0465231112733207e-05, + "loss": 0.8324, + "step": 44533 + }, + { + "epoch": 2.87, + "learning_rate": 1.0465016423710257e-05, + "loss": 0.8712, + "step": 44534 + }, + { + "epoch": 2.87, + "learning_rate": 1.0464801734687306e-05, + "loss": 0.8766, + "step": 44535 + }, + { + "epoch": 2.87, + "learning_rate": 1.0464587045664357e-05, + "loss": 0.8796, + "step": 44536 + }, + { + "epoch": 2.87, + "learning_rate": 1.0464372356641406e-05, + "loss": 0.8656, + "step": 44537 + }, + { + "epoch": 2.87, + "learning_rate": 1.0464157667618455e-05, + "loss": 0.8698, + "step": 44538 + }, + { + "epoch": 2.87, + "learning_rate": 1.0463942978595505e-05, + "loss": 0.9841, + "step": 44539 + }, + { + "epoch": 2.87, + "learning_rate": 1.0463728289572554e-05, + "loss": 0.9717, + "step": 44540 + }, + { + "epoch": 2.87, + "learning_rate": 1.0463513600549603e-05, + "loss": 0.9284, + "step": 44541 + }, + { + "epoch": 2.87, + "learning_rate": 1.0463298911526654e-05, + "loss": 0.9658, + "step": 44542 + }, + { + "epoch": 2.87, + "learning_rate": 1.0463084222503703e-05, + "loss": 0.9093, + "step": 44543 + }, + { + "epoch": 2.87, + "learning_rate": 1.0462869533480755e-05, + "loss": 0.9455, + "step": 44544 + }, + { + "epoch": 2.87, + "learning_rate": 1.0462654844457804e-05, + "loss": 0.8317, + "step": 44545 + }, + { + "epoch": 2.87, + "learning_rate": 1.0462440155434855e-05, + "loss": 0.9631, + "step": 44546 + }, + { + "epoch": 2.87, + "learning_rate": 1.0462225466411904e-05, + "loss": 0.8962, + "step": 44547 + }, + { + "epoch": 2.87, + "learning_rate": 1.0462010777388953e-05, + "loss": 0.9358, + "step": 44548 + }, + { + "epoch": 2.87, + "learning_rate": 1.0461796088366004e-05, + "loss": 0.982, + "step": 44549 + }, + { + "epoch": 2.87, + "learning_rate": 1.0461581399343052e-05, + "loss": 0.9425, + "step": 44550 + }, + { + "epoch": 2.87, + "learning_rate": 1.0461366710320101e-05, + "loss": 0.9382, + "step": 44551 + }, + { + "epoch": 2.87, + "learning_rate": 1.0461152021297152e-05, + "loss": 0.9182, + "step": 44552 + }, + { + "epoch": 2.87, + "learning_rate": 1.0460937332274201e-05, + "loss": 0.9143, + "step": 44553 + }, + { + "epoch": 2.87, + "learning_rate": 1.0460722643251252e-05, + "loss": 0.9224, + "step": 44554 + }, + { + "epoch": 2.87, + "learning_rate": 1.04605079542283e-05, + "loss": 0.9775, + "step": 44555 + }, + { + "epoch": 2.87, + "learning_rate": 1.046029326520535e-05, + "loss": 0.8517, + "step": 44556 + }, + { + "epoch": 2.87, + "learning_rate": 1.04600785761824e-05, + "loss": 0.9613, + "step": 44557 + }, + { + "epoch": 2.87, + "learning_rate": 1.045986388715945e-05, + "loss": 0.9355, + "step": 44558 + }, + { + "epoch": 2.87, + "learning_rate": 1.0459649198136502e-05, + "loss": 0.8301, + "step": 44559 + }, + { + "epoch": 2.87, + "learning_rate": 1.045943450911355e-05, + "loss": 0.9338, + "step": 44560 + }, + { + "epoch": 2.87, + "learning_rate": 1.04592198200906e-05, + "loss": 0.8058, + "step": 44561 + }, + { + "epoch": 2.87, + "learning_rate": 1.045900513106765e-05, + "loss": 0.9158, + "step": 44562 + }, + { + "epoch": 2.87, + "learning_rate": 1.04587904420447e-05, + "loss": 0.8955, + "step": 44563 + }, + { + "epoch": 2.87, + "learning_rate": 1.045857575302175e-05, + "loss": 0.8686, + "step": 44564 + }, + { + "epoch": 2.87, + "learning_rate": 1.0458361063998799e-05, + "loss": 0.8612, + "step": 44565 + }, + { + "epoch": 2.87, + "learning_rate": 1.0458146374975848e-05, + "loss": 0.8228, + "step": 44566 + }, + { + "epoch": 2.87, + "learning_rate": 1.0457931685952898e-05, + "loss": 0.8393, + "step": 44567 + }, + { + "epoch": 2.87, + "learning_rate": 1.0457716996929947e-05, + "loss": 0.9558, + "step": 44568 + }, + { + "epoch": 2.87, + "learning_rate": 1.0457502307906996e-05, + "loss": 0.8185, + "step": 44569 + }, + { + "epoch": 2.87, + "learning_rate": 1.0457287618884047e-05, + "loss": 0.8712, + "step": 44570 + }, + { + "epoch": 2.87, + "learning_rate": 1.0457072929861096e-05, + "loss": 0.8652, + "step": 44571 + }, + { + "epoch": 2.87, + "learning_rate": 1.0456858240838147e-05, + "loss": 0.9254, + "step": 44572 + }, + { + "epoch": 2.87, + "learning_rate": 1.0456643551815197e-05, + "loss": 0.9303, + "step": 44573 + }, + { + "epoch": 2.87, + "learning_rate": 1.0456428862792248e-05, + "loss": 0.9599, + "step": 44574 + }, + { + "epoch": 2.87, + "learning_rate": 1.0456214173769297e-05, + "loss": 0.9471, + "step": 44575 + }, + { + "epoch": 2.87, + "learning_rate": 1.0455999484746346e-05, + "loss": 0.8063, + "step": 44576 + }, + { + "epoch": 2.87, + "learning_rate": 1.0455784795723397e-05, + "loss": 0.9271, + "step": 44577 + }, + { + "epoch": 2.87, + "learning_rate": 1.0455570106700446e-05, + "loss": 0.8731, + "step": 44578 + }, + { + "epoch": 2.87, + "learning_rate": 1.0455355417677494e-05, + "loss": 0.9157, + "step": 44579 + }, + { + "epoch": 2.87, + "learning_rate": 1.0455140728654545e-05, + "loss": 0.8889, + "step": 44580 + }, + { + "epoch": 2.87, + "learning_rate": 1.0454926039631594e-05, + "loss": 0.9135, + "step": 44581 + }, + { + "epoch": 2.87, + "learning_rate": 1.0454711350608645e-05, + "loss": 0.8706, + "step": 44582 + }, + { + "epoch": 2.87, + "learning_rate": 1.0454496661585694e-05, + "loss": 0.9289, + "step": 44583 + }, + { + "epoch": 2.87, + "learning_rate": 1.0454281972562743e-05, + "loss": 0.9763, + "step": 44584 + }, + { + "epoch": 2.87, + "learning_rate": 1.0454067283539793e-05, + "loss": 0.8671, + "step": 44585 + }, + { + "epoch": 2.87, + "learning_rate": 1.0453852594516842e-05, + "loss": 0.8056, + "step": 44586 + }, + { + "epoch": 2.87, + "learning_rate": 1.0453637905493891e-05, + "loss": 0.9063, + "step": 44587 + }, + { + "epoch": 2.87, + "learning_rate": 1.0453637905493891e-05, + "loss": 0.9347, + "step": 44588 + }, + { + "epoch": 2.87, + "learning_rate": 1.0453423216470944e-05, + "loss": 0.8325, + "step": 44589 + }, + { + "epoch": 2.87, + "learning_rate": 1.0453208527447993e-05, + "loss": 0.8384, + "step": 44590 + }, + { + "epoch": 2.87, + "learning_rate": 1.0452993838425043e-05, + "loss": 0.8852, + "step": 44591 + }, + { + "epoch": 2.87, + "learning_rate": 1.0452779149402092e-05, + "loss": 0.8852, + "step": 44592 + }, + { + "epoch": 2.87, + "learning_rate": 1.0452564460379143e-05, + "loss": 0.8848, + "step": 44593 + }, + { + "epoch": 2.87, + "learning_rate": 1.0452349771356192e-05, + "loss": 0.8579, + "step": 44594 + }, + { + "epoch": 2.87, + "learning_rate": 1.045213508233324e-05, + "loss": 0.9053, + "step": 44595 + }, + { + "epoch": 2.87, + "learning_rate": 1.0451920393310291e-05, + "loss": 0.9264, + "step": 44596 + }, + { + "epoch": 2.87, + "learning_rate": 1.045170570428734e-05, + "loss": 0.8389, + "step": 44597 + }, + { + "epoch": 2.87, + "learning_rate": 1.045149101526439e-05, + "loss": 0.8885, + "step": 44598 + }, + { + "epoch": 2.87, + "learning_rate": 1.045127632624144e-05, + "loss": 0.8868, + "step": 44599 + }, + { + "epoch": 2.87, + "learning_rate": 1.0451061637218489e-05, + "loss": 0.8618, + "step": 44600 + }, + { + "epoch": 2.87, + "learning_rate": 1.0450846948195538e-05, + "loss": 0.8764, + "step": 44601 + }, + { + "epoch": 2.87, + "learning_rate": 1.0450632259172589e-05, + "loss": 0.8977, + "step": 44602 + }, + { + "epoch": 2.87, + "learning_rate": 1.0450417570149641e-05, + "loss": 0.8882, + "step": 44603 + }, + { + "epoch": 2.87, + "learning_rate": 1.045020288112669e-05, + "loss": 0.9327, + "step": 44604 + }, + { + "epoch": 2.87, + "learning_rate": 1.0449988192103739e-05, + "loss": 0.9313, + "step": 44605 + }, + { + "epoch": 2.87, + "learning_rate": 1.044977350308079e-05, + "loss": 0.9509, + "step": 44606 + }, + { + "epoch": 2.87, + "learning_rate": 1.0449558814057839e-05, + "loss": 0.8635, + "step": 44607 + }, + { + "epoch": 2.87, + "learning_rate": 1.0449344125034888e-05, + "loss": 0.8719, + "step": 44608 + }, + { + "epoch": 2.87, + "learning_rate": 1.0449129436011938e-05, + "loss": 0.975, + "step": 44609 + }, + { + "epoch": 2.87, + "learning_rate": 1.0448914746988987e-05, + "loss": 0.7935, + "step": 44610 + }, + { + "epoch": 2.87, + "learning_rate": 1.0448700057966036e-05, + "loss": 0.8965, + "step": 44611 + }, + { + "epoch": 2.87, + "learning_rate": 1.0448485368943087e-05, + "loss": 0.8881, + "step": 44612 + }, + { + "epoch": 2.87, + "learning_rate": 1.0448270679920136e-05, + "loss": 0.9452, + "step": 44613 + }, + { + "epoch": 2.87, + "learning_rate": 1.0448055990897186e-05, + "loss": 0.8922, + "step": 44614 + }, + { + "epoch": 2.87, + "learning_rate": 1.0447841301874235e-05, + "loss": 0.9332, + "step": 44615 + }, + { + "epoch": 2.87, + "learning_rate": 1.0447626612851284e-05, + "loss": 0.8789, + "step": 44616 + }, + { + "epoch": 2.87, + "learning_rate": 1.0447411923828335e-05, + "loss": 0.9752, + "step": 44617 + }, + { + "epoch": 2.87, + "learning_rate": 1.0447197234805386e-05, + "loss": 0.8828, + "step": 44618 + }, + { + "epoch": 2.87, + "learning_rate": 1.0446982545782436e-05, + "loss": 0.8767, + "step": 44619 + }, + { + "epoch": 2.87, + "learning_rate": 1.0446767856759485e-05, + "loss": 0.8866, + "step": 44620 + }, + { + "epoch": 2.87, + "learning_rate": 1.0446553167736534e-05, + "loss": 0.9023, + "step": 44621 + }, + { + "epoch": 2.87, + "learning_rate": 1.0446338478713585e-05, + "loss": 0.9135, + "step": 44622 + }, + { + "epoch": 2.87, + "learning_rate": 1.0446123789690634e-05, + "loss": 0.8702, + "step": 44623 + }, + { + "epoch": 2.87, + "learning_rate": 1.0445909100667685e-05, + "loss": 0.9206, + "step": 44624 + }, + { + "epoch": 2.87, + "learning_rate": 1.0445694411644733e-05, + "loss": 0.8807, + "step": 44625 + }, + { + "epoch": 2.87, + "learning_rate": 1.0445479722621782e-05, + "loss": 0.8318, + "step": 44626 + }, + { + "epoch": 2.87, + "learning_rate": 1.0445265033598833e-05, + "loss": 0.8957, + "step": 44627 + }, + { + "epoch": 2.87, + "learning_rate": 1.0445050344575882e-05, + "loss": 0.8803, + "step": 44628 + }, + { + "epoch": 2.87, + "learning_rate": 1.0444835655552931e-05, + "loss": 0.9381, + "step": 44629 + }, + { + "epoch": 2.87, + "learning_rate": 1.0444620966529982e-05, + "loss": 0.86, + "step": 44630 + }, + { + "epoch": 2.87, + "learning_rate": 1.044440627750703e-05, + "loss": 0.907, + "step": 44631 + }, + { + "epoch": 2.87, + "learning_rate": 1.0444191588484083e-05, + "loss": 0.8754, + "step": 44632 + }, + { + "epoch": 2.87, + "learning_rate": 1.0443976899461132e-05, + "loss": 0.9017, + "step": 44633 + }, + { + "epoch": 2.87, + "learning_rate": 1.0443762210438183e-05, + "loss": 0.8526, + "step": 44634 + }, + { + "epoch": 2.87, + "learning_rate": 1.0443547521415232e-05, + "loss": 0.8448, + "step": 44635 + }, + { + "epoch": 2.87, + "learning_rate": 1.044333283239228e-05, + "loss": 0.8457, + "step": 44636 + }, + { + "epoch": 2.87, + "learning_rate": 1.0443118143369331e-05, + "loss": 0.8679, + "step": 44637 + }, + { + "epoch": 2.87, + "learning_rate": 1.044290345434638e-05, + "loss": 0.8557, + "step": 44638 + }, + { + "epoch": 2.87, + "learning_rate": 1.0442688765323429e-05, + "loss": 0.9085, + "step": 44639 + }, + { + "epoch": 2.87, + "learning_rate": 1.044247407630048e-05, + "loss": 0.8343, + "step": 44640 + }, + { + "epoch": 2.87, + "learning_rate": 1.0442259387277529e-05, + "loss": 0.8961, + "step": 44641 + }, + { + "epoch": 2.87, + "learning_rate": 1.044204469825458e-05, + "loss": 0.9198, + "step": 44642 + }, + { + "epoch": 2.87, + "learning_rate": 1.0441830009231628e-05, + "loss": 0.9877, + "step": 44643 + }, + { + "epoch": 2.87, + "learning_rate": 1.0441615320208677e-05, + "loss": 0.9047, + "step": 44644 + }, + { + "epoch": 2.87, + "learning_rate": 1.0441400631185728e-05, + "loss": 0.9313, + "step": 44645 + }, + { + "epoch": 2.87, + "learning_rate": 1.0441185942162777e-05, + "loss": 0.8754, + "step": 44646 + }, + { + "epoch": 2.87, + "learning_rate": 1.044097125313983e-05, + "loss": 0.9398, + "step": 44647 + }, + { + "epoch": 2.87, + "learning_rate": 1.0440756564116878e-05, + "loss": 0.8636, + "step": 44648 + }, + { + "epoch": 2.87, + "learning_rate": 1.0440541875093927e-05, + "loss": 0.8817, + "step": 44649 + }, + { + "epoch": 2.87, + "learning_rate": 1.0440327186070978e-05, + "loss": 0.9246, + "step": 44650 + }, + { + "epoch": 2.87, + "learning_rate": 1.0440112497048027e-05, + "loss": 0.8189, + "step": 44651 + }, + { + "epoch": 2.87, + "learning_rate": 1.0439897808025078e-05, + "loss": 0.9193, + "step": 44652 + }, + { + "epoch": 2.87, + "learning_rate": 1.0439683119002127e-05, + "loss": 1.0161, + "step": 44653 + }, + { + "epoch": 2.87, + "learning_rate": 1.0439468429979175e-05, + "loss": 0.8884, + "step": 44654 + }, + { + "epoch": 2.87, + "learning_rate": 1.0439253740956226e-05, + "loss": 0.9007, + "step": 44655 + }, + { + "epoch": 2.87, + "learning_rate": 1.0439039051933275e-05, + "loss": 0.8889, + "step": 44656 + }, + { + "epoch": 2.87, + "learning_rate": 1.0438824362910324e-05, + "loss": 0.8501, + "step": 44657 + }, + { + "epoch": 2.87, + "learning_rate": 1.0438609673887375e-05, + "loss": 0.9347, + "step": 44658 + }, + { + "epoch": 2.87, + "learning_rate": 1.0438394984864424e-05, + "loss": 0.9322, + "step": 44659 + }, + { + "epoch": 2.87, + "learning_rate": 1.0438180295841474e-05, + "loss": 0.8863, + "step": 44660 + }, + { + "epoch": 2.87, + "learning_rate": 1.0437965606818525e-05, + "loss": 0.9427, + "step": 44661 + }, + { + "epoch": 2.87, + "learning_rate": 1.0437750917795576e-05, + "loss": 0.8475, + "step": 44662 + }, + { + "epoch": 2.87, + "learning_rate": 1.0437536228772625e-05, + "loss": 0.9454, + "step": 44663 + }, + { + "epoch": 2.87, + "learning_rate": 1.0437321539749674e-05, + "loss": 0.8746, + "step": 44664 + }, + { + "epoch": 2.87, + "learning_rate": 1.0437106850726724e-05, + "loss": 0.8394, + "step": 44665 + }, + { + "epoch": 2.87, + "learning_rate": 1.0436892161703773e-05, + "loss": 0.9445, + "step": 44666 + }, + { + "epoch": 2.87, + "learning_rate": 1.0436677472680822e-05, + "loss": 0.8817, + "step": 44667 + }, + { + "epoch": 2.87, + "learning_rate": 1.0436462783657873e-05, + "loss": 0.8664, + "step": 44668 + }, + { + "epoch": 2.87, + "learning_rate": 1.0436248094634922e-05, + "loss": 0.9362, + "step": 44669 + }, + { + "epoch": 2.87, + "learning_rate": 1.0436033405611972e-05, + "loss": 0.9382, + "step": 44670 + }, + { + "epoch": 2.87, + "learning_rate": 1.0435818716589021e-05, + "loss": 0.8991, + "step": 44671 + }, + { + "epoch": 2.87, + "learning_rate": 1.043560402756607e-05, + "loss": 0.9891, + "step": 44672 + }, + { + "epoch": 2.87, + "learning_rate": 1.0435389338543121e-05, + "loss": 0.9193, + "step": 44673 + }, + { + "epoch": 2.87, + "learning_rate": 1.043517464952017e-05, + "loss": 0.8901, + "step": 44674 + }, + { + "epoch": 2.87, + "learning_rate": 1.0434959960497219e-05, + "loss": 0.8828, + "step": 44675 + }, + { + "epoch": 2.87, + "learning_rate": 1.0434745271474271e-05, + "loss": 0.8845, + "step": 44676 + }, + { + "epoch": 2.87, + "learning_rate": 1.043453058245132e-05, + "loss": 0.9507, + "step": 44677 + }, + { + "epoch": 2.87, + "learning_rate": 1.0434315893428371e-05, + "loss": 0.908, + "step": 44678 + }, + { + "epoch": 2.87, + "learning_rate": 1.043410120440542e-05, + "loss": 0.9683, + "step": 44679 + }, + { + "epoch": 2.87, + "learning_rate": 1.043388651538247e-05, + "loss": 0.974, + "step": 44680 + }, + { + "epoch": 2.87, + "learning_rate": 1.043367182635952e-05, + "loss": 0.8395, + "step": 44681 + }, + { + "epoch": 2.87, + "learning_rate": 1.0433457137336569e-05, + "loss": 0.9181, + "step": 44682 + }, + { + "epoch": 2.87, + "learning_rate": 1.043324244831362e-05, + "loss": 0.8909, + "step": 44683 + }, + { + "epoch": 2.87, + "learning_rate": 1.0433027759290668e-05, + "loss": 0.9047, + "step": 44684 + }, + { + "epoch": 2.87, + "learning_rate": 1.0432813070267717e-05, + "loss": 0.8828, + "step": 44685 + }, + { + "epoch": 2.87, + "learning_rate": 1.0432598381244768e-05, + "loss": 0.9456, + "step": 44686 + }, + { + "epoch": 2.88, + "learning_rate": 1.0432383692221817e-05, + "loss": 0.891, + "step": 44687 + }, + { + "epoch": 2.88, + "learning_rate": 1.0432169003198866e-05, + "loss": 0.9478, + "step": 44688 + }, + { + "epoch": 2.88, + "learning_rate": 1.0431954314175916e-05, + "loss": 0.8014, + "step": 44689 + }, + { + "epoch": 2.88, + "learning_rate": 1.0431739625152965e-05, + "loss": 0.8888, + "step": 44690 + }, + { + "epoch": 2.88, + "learning_rate": 1.0431524936130018e-05, + "loss": 0.8899, + "step": 44691 + }, + { + "epoch": 2.88, + "learning_rate": 1.0431310247107067e-05, + "loss": 0.8658, + "step": 44692 + }, + { + "epoch": 2.88, + "learning_rate": 1.0431095558084117e-05, + "loss": 0.7762, + "step": 44693 + }, + { + "epoch": 2.88, + "learning_rate": 1.0430880869061166e-05, + "loss": 0.9076, + "step": 44694 + }, + { + "epoch": 2.88, + "learning_rate": 1.0430666180038215e-05, + "loss": 0.8809, + "step": 44695 + }, + { + "epoch": 2.88, + "learning_rate": 1.0430451491015266e-05, + "loss": 0.8673, + "step": 44696 + }, + { + "epoch": 2.88, + "learning_rate": 1.0430236801992315e-05, + "loss": 0.9046, + "step": 44697 + }, + { + "epoch": 2.88, + "learning_rate": 1.0430022112969364e-05, + "loss": 0.8791, + "step": 44698 + }, + { + "epoch": 2.88, + "learning_rate": 1.0429807423946414e-05, + "loss": 0.8663, + "step": 44699 + }, + { + "epoch": 2.88, + "learning_rate": 1.0429592734923463e-05, + "loss": 0.8955, + "step": 44700 + }, + { + "epoch": 2.88, + "learning_rate": 1.0429378045900514e-05, + "loss": 0.8751, + "step": 44701 + }, + { + "epoch": 2.88, + "learning_rate": 1.0429163356877563e-05, + "loss": 0.9081, + "step": 44702 + }, + { + "epoch": 2.88, + "learning_rate": 1.0428948667854612e-05, + "loss": 0.8797, + "step": 44703 + }, + { + "epoch": 2.88, + "learning_rate": 1.0428733978831663e-05, + "loss": 0.9191, + "step": 44704 + }, + { + "epoch": 2.88, + "learning_rate": 1.0428519289808713e-05, + "loss": 0.9375, + "step": 44705 + }, + { + "epoch": 2.88, + "learning_rate": 1.0428304600785764e-05, + "loss": 0.8593, + "step": 44706 + }, + { + "epoch": 2.88, + "learning_rate": 1.0428089911762813e-05, + "loss": 0.9243, + "step": 44707 + }, + { + "epoch": 2.88, + "learning_rate": 1.0427875222739862e-05, + "loss": 0.8659, + "step": 44708 + }, + { + "epoch": 2.88, + "learning_rate": 1.0427660533716913e-05, + "loss": 0.9039, + "step": 44709 + }, + { + "epoch": 2.88, + "learning_rate": 1.0427445844693962e-05, + "loss": 0.9075, + "step": 44710 + }, + { + "epoch": 2.88, + "learning_rate": 1.0427231155671012e-05, + "loss": 0.9323, + "step": 44711 + }, + { + "epoch": 2.88, + "learning_rate": 1.0427016466648061e-05, + "loss": 0.919, + "step": 44712 + }, + { + "epoch": 2.88, + "learning_rate": 1.042680177762511e-05, + "loss": 0.9411, + "step": 44713 + }, + { + "epoch": 2.88, + "learning_rate": 1.042658708860216e-05, + "loss": 0.8964, + "step": 44714 + }, + { + "epoch": 2.88, + "learning_rate": 1.042637239957921e-05, + "loss": 0.797, + "step": 44715 + }, + { + "epoch": 2.88, + "learning_rate": 1.0426157710556259e-05, + "loss": 0.8528, + "step": 44716 + }, + { + "epoch": 2.88, + "learning_rate": 1.042594302153331e-05, + "loss": 0.883, + "step": 44717 + }, + { + "epoch": 2.88, + "learning_rate": 1.0425728332510358e-05, + "loss": 0.8788, + "step": 44718 + }, + { + "epoch": 2.88, + "learning_rate": 1.0425513643487409e-05, + "loss": 0.9539, + "step": 44719 + }, + { + "epoch": 2.88, + "learning_rate": 1.042529895446446e-05, + "loss": 0.8425, + "step": 44720 + }, + { + "epoch": 2.88, + "learning_rate": 1.042508426544151e-05, + "loss": 0.8328, + "step": 44721 + }, + { + "epoch": 2.88, + "learning_rate": 1.042486957641856e-05, + "loss": 0.8834, + "step": 44722 + }, + { + "epoch": 2.88, + "learning_rate": 1.0424654887395608e-05, + "loss": 0.901, + "step": 44723 + }, + { + "epoch": 2.88, + "learning_rate": 1.0424440198372659e-05, + "loss": 0.8163, + "step": 44724 + }, + { + "epoch": 2.88, + "learning_rate": 1.0424225509349708e-05, + "loss": 0.9011, + "step": 44725 + }, + { + "epoch": 2.88, + "learning_rate": 1.0424010820326757e-05, + "loss": 0.9196, + "step": 44726 + }, + { + "epoch": 2.88, + "learning_rate": 1.0423796131303808e-05, + "loss": 0.9507, + "step": 44727 + }, + { + "epoch": 2.88, + "learning_rate": 1.0423581442280856e-05, + "loss": 0.8901, + "step": 44728 + }, + { + "epoch": 2.88, + "learning_rate": 1.0423366753257907e-05, + "loss": 0.9503, + "step": 44729 + }, + { + "epoch": 2.88, + "learning_rate": 1.0423152064234956e-05, + "loss": 0.9704, + "step": 44730 + }, + { + "epoch": 2.88, + "learning_rate": 1.0422937375212005e-05, + "loss": 0.8855, + "step": 44731 + }, + { + "epoch": 2.88, + "learning_rate": 1.0422722686189056e-05, + "loss": 0.9204, + "step": 44732 + }, + { + "epoch": 2.88, + "learning_rate": 1.0422507997166105e-05, + "loss": 0.9028, + "step": 44733 + }, + { + "epoch": 2.88, + "learning_rate": 1.0422293308143157e-05, + "loss": 0.9383, + "step": 44734 + }, + { + "epoch": 2.88, + "learning_rate": 1.0422078619120206e-05, + "loss": 0.8377, + "step": 44735 + }, + { + "epoch": 2.88, + "learning_rate": 1.0421863930097255e-05, + "loss": 0.9854, + "step": 44736 + }, + { + "epoch": 2.88, + "learning_rate": 1.0421649241074306e-05, + "loss": 0.9484, + "step": 44737 + }, + { + "epoch": 2.88, + "learning_rate": 1.0421434552051355e-05, + "loss": 0.8757, + "step": 44738 + }, + { + "epoch": 2.88, + "learning_rate": 1.0421219863028405e-05, + "loss": 0.9152, + "step": 44739 + }, + { + "epoch": 2.88, + "learning_rate": 1.0421005174005454e-05, + "loss": 0.9751, + "step": 44740 + }, + { + "epoch": 2.88, + "learning_rate": 1.0420790484982503e-05, + "loss": 0.8792, + "step": 44741 + }, + { + "epoch": 2.88, + "learning_rate": 1.0420575795959554e-05, + "loss": 0.8282, + "step": 44742 + }, + { + "epoch": 2.88, + "learning_rate": 1.0420361106936603e-05, + "loss": 0.8008, + "step": 44743 + }, + { + "epoch": 2.88, + "learning_rate": 1.0420146417913652e-05, + "loss": 0.855, + "step": 44744 + }, + { + "epoch": 2.88, + "learning_rate": 1.0419931728890702e-05, + "loss": 0.9243, + "step": 44745 + }, + { + "epoch": 2.88, + "learning_rate": 1.0419717039867751e-05, + "loss": 0.8908, + "step": 44746 + }, + { + "epoch": 2.88, + "learning_rate": 1.04195023508448e-05, + "loss": 0.8932, + "step": 44747 + }, + { + "epoch": 2.88, + "learning_rate": 1.0419287661821851e-05, + "loss": 0.8323, + "step": 44748 + }, + { + "epoch": 2.88, + "learning_rate": 1.0419072972798903e-05, + "loss": 0.9147, + "step": 44749 + }, + { + "epoch": 2.88, + "learning_rate": 1.0418858283775952e-05, + "loss": 0.9651, + "step": 44750 + }, + { + "epoch": 2.88, + "learning_rate": 1.0418643594753001e-05, + "loss": 0.891, + "step": 44751 + }, + { + "epoch": 2.88, + "learning_rate": 1.0418428905730052e-05, + "loss": 0.8606, + "step": 44752 + }, + { + "epoch": 2.88, + "learning_rate": 1.0418214216707101e-05, + "loss": 0.903, + "step": 44753 + }, + { + "epoch": 2.88, + "learning_rate": 1.041799952768415e-05, + "loss": 0.9308, + "step": 44754 + }, + { + "epoch": 2.88, + "learning_rate": 1.04177848386612e-05, + "loss": 0.8481, + "step": 44755 + }, + { + "epoch": 2.88, + "learning_rate": 1.041757014963825e-05, + "loss": 0.8795, + "step": 44756 + }, + { + "epoch": 2.88, + "learning_rate": 1.04173554606153e-05, + "loss": 0.901, + "step": 44757 + }, + { + "epoch": 2.88, + "learning_rate": 1.0417140771592349e-05, + "loss": 0.8555, + "step": 44758 + }, + { + "epoch": 2.88, + "learning_rate": 1.0416926082569398e-05, + "loss": 0.8786, + "step": 44759 + }, + { + "epoch": 2.88, + "learning_rate": 1.0416711393546449e-05, + "loss": 0.9583, + "step": 44760 + }, + { + "epoch": 2.88, + "learning_rate": 1.0416496704523498e-05, + "loss": 0.971, + "step": 44761 + }, + { + "epoch": 2.88, + "learning_rate": 1.0416282015500547e-05, + "loss": 0.9162, + "step": 44762 + }, + { + "epoch": 2.88, + "learning_rate": 1.0416067326477599e-05, + "loss": 0.8971, + "step": 44763 + }, + { + "epoch": 2.88, + "learning_rate": 1.0415852637454648e-05, + "loss": 0.8851, + "step": 44764 + }, + { + "epoch": 2.88, + "learning_rate": 1.0415637948431699e-05, + "loss": 0.8675, + "step": 44765 + }, + { + "epoch": 2.88, + "learning_rate": 1.0415423259408748e-05, + "loss": 0.8854, + "step": 44766 + }, + { + "epoch": 2.88, + "learning_rate": 1.0415208570385798e-05, + "loss": 0.9082, + "step": 44767 + }, + { + "epoch": 2.88, + "learning_rate": 1.0414993881362847e-05, + "loss": 0.9114, + "step": 44768 + }, + { + "epoch": 2.88, + "learning_rate": 1.0414779192339896e-05, + "loss": 0.9326, + "step": 44769 + }, + { + "epoch": 2.88, + "learning_rate": 1.0414564503316947e-05, + "loss": 0.8104, + "step": 44770 + }, + { + "epoch": 2.88, + "learning_rate": 1.0414349814293996e-05, + "loss": 0.8884, + "step": 44771 + }, + { + "epoch": 2.88, + "learning_rate": 1.0414135125271045e-05, + "loss": 0.8506, + "step": 44772 + }, + { + "epoch": 2.88, + "learning_rate": 1.0413920436248095e-05, + "loss": 0.8222, + "step": 44773 + }, + { + "epoch": 2.88, + "learning_rate": 1.0413705747225144e-05, + "loss": 0.9432, + "step": 44774 + }, + { + "epoch": 2.88, + "learning_rate": 1.0413491058202193e-05, + "loss": 0.9248, + "step": 44775 + }, + { + "epoch": 2.88, + "learning_rate": 1.0413276369179244e-05, + "loss": 0.8951, + "step": 44776 + }, + { + "epoch": 2.88, + "learning_rate": 1.0413061680156293e-05, + "loss": 0.8834, + "step": 44777 + }, + { + "epoch": 2.88, + "learning_rate": 1.0412846991133345e-05, + "loss": 0.9497, + "step": 44778 + }, + { + "epoch": 2.88, + "learning_rate": 1.0412632302110394e-05, + "loss": 0.9051, + "step": 44779 + }, + { + "epoch": 2.88, + "learning_rate": 1.0412417613087445e-05, + "loss": 0.9621, + "step": 44780 + }, + { + "epoch": 2.88, + "learning_rate": 1.0412202924064494e-05, + "loss": 0.854, + "step": 44781 + }, + { + "epoch": 2.88, + "learning_rate": 1.0411988235041543e-05, + "loss": 0.8819, + "step": 44782 + }, + { + "epoch": 2.88, + "learning_rate": 1.0411773546018594e-05, + "loss": 0.8846, + "step": 44783 + }, + { + "epoch": 2.88, + "learning_rate": 1.0411558856995643e-05, + "loss": 1.0066, + "step": 44784 + }, + { + "epoch": 2.88, + "learning_rate": 1.0411344167972692e-05, + "loss": 0.9465, + "step": 44785 + }, + { + "epoch": 2.88, + "learning_rate": 1.0411129478949742e-05, + "loss": 0.8219, + "step": 44786 + }, + { + "epoch": 2.88, + "learning_rate": 1.0410914789926791e-05, + "loss": 0.9238, + "step": 44787 + }, + { + "epoch": 2.88, + "learning_rate": 1.0410700100903842e-05, + "loss": 0.8511, + "step": 44788 + }, + { + "epoch": 2.88, + "learning_rate": 1.041048541188089e-05, + "loss": 0.9303, + "step": 44789 + }, + { + "epoch": 2.88, + "learning_rate": 1.041027072285794e-05, + "loss": 0.8222, + "step": 44790 + }, + { + "epoch": 2.88, + "learning_rate": 1.041005603383499e-05, + "loss": 0.9579, + "step": 44791 + }, + { + "epoch": 2.88, + "learning_rate": 1.0409841344812041e-05, + "loss": 0.9139, + "step": 44792 + }, + { + "epoch": 2.88, + "learning_rate": 1.0409626655789092e-05, + "loss": 0.8659, + "step": 44793 + }, + { + "epoch": 2.88, + "learning_rate": 1.040941196676614e-05, + "loss": 0.8698, + "step": 44794 + }, + { + "epoch": 2.88, + "learning_rate": 1.040919727774319e-05, + "loss": 0.8679, + "step": 44795 + }, + { + "epoch": 2.88, + "learning_rate": 1.040898258872024e-05, + "loss": 0.9703, + "step": 44796 + }, + { + "epoch": 2.88, + "learning_rate": 1.040876789969729e-05, + "loss": 0.8809, + "step": 44797 + }, + { + "epoch": 2.88, + "learning_rate": 1.040855321067434e-05, + "loss": 0.97, + "step": 44798 + }, + { + "epoch": 2.88, + "learning_rate": 1.0408338521651389e-05, + "loss": 0.9362, + "step": 44799 + }, + { + "epoch": 2.88, + "learning_rate": 1.0408123832628438e-05, + "loss": 0.8836, + "step": 44800 + }, + { + "epoch": 2.88, + "learning_rate": 1.0407909143605489e-05, + "loss": 0.9226, + "step": 44801 + }, + { + "epoch": 2.88, + "learning_rate": 1.0407694454582537e-05, + "loss": 0.8892, + "step": 44802 + }, + { + "epoch": 2.88, + "learning_rate": 1.0407479765559586e-05, + "loss": 0.8617, + "step": 44803 + }, + { + "epoch": 2.88, + "learning_rate": 1.0407265076536637e-05, + "loss": 0.8594, + "step": 44804 + }, + { + "epoch": 2.88, + "learning_rate": 1.0407050387513686e-05, + "loss": 0.877, + "step": 44805 + }, + { + "epoch": 2.88, + "learning_rate": 1.0406835698490737e-05, + "loss": 0.9587, + "step": 44806 + }, + { + "epoch": 2.88, + "learning_rate": 1.0406621009467787e-05, + "loss": 0.9221, + "step": 44807 + }, + { + "epoch": 2.88, + "learning_rate": 1.0406406320444838e-05, + "loss": 0.8796, + "step": 44808 + }, + { + "epoch": 2.88, + "learning_rate": 1.0406191631421887e-05, + "loss": 0.8959, + "step": 44809 + }, + { + "epoch": 2.88, + "learning_rate": 1.0405976942398936e-05, + "loss": 0.8192, + "step": 44810 + }, + { + "epoch": 2.88, + "learning_rate": 1.0405762253375987e-05, + "loss": 0.9144, + "step": 44811 + }, + { + "epoch": 2.88, + "learning_rate": 1.0405547564353036e-05, + "loss": 0.8406, + "step": 44812 + }, + { + "epoch": 2.88, + "learning_rate": 1.0405332875330085e-05, + "loss": 0.9392, + "step": 44813 + }, + { + "epoch": 2.88, + "learning_rate": 1.0405118186307135e-05, + "loss": 0.9112, + "step": 44814 + }, + { + "epoch": 2.88, + "learning_rate": 1.0404903497284184e-05, + "loss": 0.8855, + "step": 44815 + }, + { + "epoch": 2.88, + "learning_rate": 1.0404688808261235e-05, + "loss": 0.9281, + "step": 44816 + }, + { + "epoch": 2.88, + "learning_rate": 1.0404474119238284e-05, + "loss": 0.9134, + "step": 44817 + }, + { + "epoch": 2.88, + "learning_rate": 1.0404259430215333e-05, + "loss": 0.8686, + "step": 44818 + }, + { + "epoch": 2.88, + "learning_rate": 1.0404044741192383e-05, + "loss": 0.9352, + "step": 44819 + }, + { + "epoch": 2.88, + "learning_rate": 1.0403830052169432e-05, + "loss": 0.8591, + "step": 44820 + }, + { + "epoch": 2.88, + "learning_rate": 1.0403615363146485e-05, + "loss": 0.947, + "step": 44821 + }, + { + "epoch": 2.88, + "learning_rate": 1.0403400674123534e-05, + "loss": 0.8882, + "step": 44822 + }, + { + "epoch": 2.88, + "learning_rate": 1.0403185985100583e-05, + "loss": 0.83, + "step": 44823 + }, + { + "epoch": 2.88, + "learning_rate": 1.0402971296077633e-05, + "loss": 0.9286, + "step": 44824 + }, + { + "epoch": 2.88, + "learning_rate": 1.0402756607054682e-05, + "loss": 0.9473, + "step": 44825 + }, + { + "epoch": 2.88, + "learning_rate": 1.0402541918031733e-05, + "loss": 0.9098, + "step": 44826 + }, + { + "epoch": 2.88, + "learning_rate": 1.0402327229008782e-05, + "loss": 0.8493, + "step": 44827 + }, + { + "epoch": 2.88, + "learning_rate": 1.0402112539985831e-05, + "loss": 0.8958, + "step": 44828 + }, + { + "epoch": 2.88, + "learning_rate": 1.0401897850962882e-05, + "loss": 0.9508, + "step": 44829 + }, + { + "epoch": 2.88, + "learning_rate": 1.040168316193993e-05, + "loss": 0.8789, + "step": 44830 + }, + { + "epoch": 2.88, + "learning_rate": 1.040146847291698e-05, + "loss": 0.8829, + "step": 44831 + }, + { + "epoch": 2.88, + "learning_rate": 1.040125378389403e-05, + "loss": 0.9911, + "step": 44832 + }, + { + "epoch": 2.88, + "learning_rate": 1.0401039094871079e-05, + "loss": 0.9074, + "step": 44833 + }, + { + "epoch": 2.88, + "learning_rate": 1.0400824405848128e-05, + "loss": 0.8789, + "step": 44834 + }, + { + "epoch": 2.88, + "learning_rate": 1.0400609716825179e-05, + "loss": 0.9227, + "step": 44835 + }, + { + "epoch": 2.88, + "learning_rate": 1.0400395027802231e-05, + "loss": 0.8849, + "step": 44836 + }, + { + "epoch": 2.88, + "learning_rate": 1.040018033877928e-05, + "loss": 0.887, + "step": 44837 + }, + { + "epoch": 2.88, + "learning_rate": 1.0399965649756329e-05, + "loss": 0.9277, + "step": 44838 + }, + { + "epoch": 2.88, + "learning_rate": 1.039975096073338e-05, + "loss": 0.9261, + "step": 44839 + }, + { + "epoch": 2.88, + "learning_rate": 1.0399536271710429e-05, + "loss": 0.8818, + "step": 44840 + }, + { + "epoch": 2.88, + "learning_rate": 1.0399321582687478e-05, + "loss": 0.8844, + "step": 44841 + }, + { + "epoch": 2.89, + "learning_rate": 1.0399106893664528e-05, + "loss": 0.88, + "step": 44842 + }, + { + "epoch": 2.89, + "learning_rate": 1.0398892204641577e-05, + "loss": 0.8672, + "step": 44843 + }, + { + "epoch": 2.89, + "learning_rate": 1.0398677515618626e-05, + "loss": 0.9162, + "step": 44844 + }, + { + "epoch": 2.89, + "learning_rate": 1.0398462826595677e-05, + "loss": 0.9026, + "step": 44845 + }, + { + "epoch": 2.89, + "learning_rate": 1.0398248137572726e-05, + "loss": 0.9055, + "step": 44846 + }, + { + "epoch": 2.89, + "learning_rate": 1.0398033448549776e-05, + "loss": 0.8459, + "step": 44847 + }, + { + "epoch": 2.89, + "learning_rate": 1.0397818759526825e-05, + "loss": 0.8673, + "step": 44848 + }, + { + "epoch": 2.89, + "learning_rate": 1.0397604070503874e-05, + "loss": 0.8947, + "step": 44849 + }, + { + "epoch": 2.89, + "learning_rate": 1.0397389381480927e-05, + "loss": 0.8997, + "step": 44850 + }, + { + "epoch": 2.89, + "learning_rate": 1.0397174692457976e-05, + "loss": 0.8733, + "step": 44851 + }, + { + "epoch": 2.89, + "learning_rate": 1.0396960003435026e-05, + "loss": 0.8942, + "step": 44852 + }, + { + "epoch": 2.89, + "learning_rate": 1.0396745314412075e-05, + "loss": 0.9217, + "step": 44853 + }, + { + "epoch": 2.89, + "learning_rate": 1.0396530625389124e-05, + "loss": 0.9478, + "step": 44854 + }, + { + "epoch": 2.89, + "learning_rate": 1.0396315936366175e-05, + "loss": 0.9669, + "step": 44855 + }, + { + "epoch": 2.89, + "learning_rate": 1.0396101247343224e-05, + "loss": 0.8505, + "step": 44856 + }, + { + "epoch": 2.89, + "learning_rate": 1.0395886558320275e-05, + "loss": 0.9692, + "step": 44857 + }, + { + "epoch": 2.89, + "learning_rate": 1.0395671869297324e-05, + "loss": 0.9304, + "step": 44858 + }, + { + "epoch": 2.89, + "learning_rate": 1.0395457180274373e-05, + "loss": 0.9399, + "step": 44859 + }, + { + "epoch": 2.89, + "learning_rate": 1.0395242491251423e-05, + "loss": 0.9648, + "step": 44860 + }, + { + "epoch": 2.89, + "learning_rate": 1.0395027802228472e-05, + "loss": 0.85, + "step": 44861 + }, + { + "epoch": 2.89, + "learning_rate": 1.0394813113205521e-05, + "loss": 0.9342, + "step": 44862 + }, + { + "epoch": 2.89, + "learning_rate": 1.0394598424182572e-05, + "loss": 0.8366, + "step": 44863 + }, + { + "epoch": 2.89, + "learning_rate": 1.039438373515962e-05, + "loss": 0.9142, + "step": 44864 + }, + { + "epoch": 2.89, + "learning_rate": 1.0394169046136673e-05, + "loss": 0.8992, + "step": 44865 + }, + { + "epoch": 2.89, + "learning_rate": 1.0393954357113722e-05, + "loss": 0.8865, + "step": 44866 + }, + { + "epoch": 2.89, + "learning_rate": 1.0393739668090773e-05, + "loss": 0.8897, + "step": 44867 + }, + { + "epoch": 2.89, + "learning_rate": 1.0393524979067822e-05, + "loss": 0.946, + "step": 44868 + }, + { + "epoch": 2.89, + "learning_rate": 1.039331029004487e-05, + "loss": 0.888, + "step": 44869 + }, + { + "epoch": 2.89, + "learning_rate": 1.0393095601021921e-05, + "loss": 0.9439, + "step": 44870 + }, + { + "epoch": 2.89, + "learning_rate": 1.039288091199897e-05, + "loss": 0.8804, + "step": 44871 + }, + { + "epoch": 2.89, + "learning_rate": 1.039266622297602e-05, + "loss": 0.9777, + "step": 44872 + }, + { + "epoch": 2.89, + "learning_rate": 1.039245153395307e-05, + "loss": 0.937, + "step": 44873 + }, + { + "epoch": 2.89, + "learning_rate": 1.0392236844930119e-05, + "loss": 0.8546, + "step": 44874 + }, + { + "epoch": 2.89, + "learning_rate": 1.039202215590717e-05, + "loss": 0.8756, + "step": 44875 + }, + { + "epoch": 2.89, + "learning_rate": 1.0391807466884218e-05, + "loss": 0.8782, + "step": 44876 + }, + { + "epoch": 2.89, + "learning_rate": 1.0391592777861267e-05, + "loss": 0.9304, + "step": 44877 + }, + { + "epoch": 2.89, + "learning_rate": 1.0391378088838318e-05, + "loss": 0.8456, + "step": 44878 + }, + { + "epoch": 2.89, + "learning_rate": 1.0391163399815369e-05, + "loss": 0.928, + "step": 44879 + }, + { + "epoch": 2.89, + "learning_rate": 1.039094871079242e-05, + "loss": 0.8567, + "step": 44880 + }, + { + "epoch": 2.89, + "learning_rate": 1.0390734021769468e-05, + "loss": 0.9138, + "step": 44881 + }, + { + "epoch": 2.89, + "learning_rate": 1.0390519332746517e-05, + "loss": 0.8889, + "step": 44882 + }, + { + "epoch": 2.89, + "learning_rate": 1.0390304643723568e-05, + "loss": 0.8567, + "step": 44883 + }, + { + "epoch": 2.89, + "learning_rate": 1.0390089954700617e-05, + "loss": 0.9652, + "step": 44884 + }, + { + "epoch": 2.89, + "learning_rate": 1.0389875265677668e-05, + "loss": 0.8695, + "step": 44885 + }, + { + "epoch": 2.89, + "learning_rate": 1.0389660576654717e-05, + "loss": 0.835, + "step": 44886 + }, + { + "epoch": 2.89, + "learning_rate": 1.0389445887631766e-05, + "loss": 0.9482, + "step": 44887 + }, + { + "epoch": 2.89, + "learning_rate": 1.0389231198608816e-05, + "loss": 0.8558, + "step": 44888 + }, + { + "epoch": 2.89, + "learning_rate": 1.0389016509585865e-05, + "loss": 0.9375, + "step": 44889 + }, + { + "epoch": 2.89, + "learning_rate": 1.0388801820562914e-05, + "loss": 0.8702, + "step": 44890 + }, + { + "epoch": 2.89, + "learning_rate": 1.0388587131539965e-05, + "loss": 0.9332, + "step": 44891 + }, + { + "epoch": 2.89, + "learning_rate": 1.0388372442517014e-05, + "loss": 0.9233, + "step": 44892 + }, + { + "epoch": 2.89, + "learning_rate": 1.0388157753494064e-05, + "loss": 0.8584, + "step": 44893 + }, + { + "epoch": 2.89, + "learning_rate": 1.0387943064471115e-05, + "loss": 0.9878, + "step": 44894 + }, + { + "epoch": 2.89, + "learning_rate": 1.0387728375448166e-05, + "loss": 0.8247, + "step": 44895 + }, + { + "epoch": 2.89, + "learning_rate": 1.0387513686425215e-05, + "loss": 0.8525, + "step": 44896 + }, + { + "epoch": 2.89, + "learning_rate": 1.0387298997402264e-05, + "loss": 0.85, + "step": 44897 + }, + { + "epoch": 2.89, + "learning_rate": 1.0387084308379314e-05, + "loss": 0.9973, + "step": 44898 + }, + { + "epoch": 2.89, + "learning_rate": 1.0386869619356363e-05, + "loss": 0.9322, + "step": 44899 + }, + { + "epoch": 2.89, + "learning_rate": 1.0386654930333412e-05, + "loss": 0.9355, + "step": 44900 + }, + { + "epoch": 2.89, + "learning_rate": 1.0386440241310463e-05, + "loss": 0.9444, + "step": 44901 + }, + { + "epoch": 2.89, + "learning_rate": 1.0386225552287512e-05, + "loss": 0.9033, + "step": 44902 + }, + { + "epoch": 2.89, + "learning_rate": 1.0386010863264563e-05, + "loss": 0.8858, + "step": 44903 + }, + { + "epoch": 2.89, + "learning_rate": 1.0385796174241612e-05, + "loss": 0.9095, + "step": 44904 + }, + { + "epoch": 2.89, + "learning_rate": 1.038558148521866e-05, + "loss": 0.8283, + "step": 44905 + }, + { + "epoch": 2.89, + "learning_rate": 1.0385366796195711e-05, + "loss": 0.8957, + "step": 44906 + }, + { + "epoch": 2.89, + "learning_rate": 1.038515210717276e-05, + "loss": 0.8555, + "step": 44907 + }, + { + "epoch": 2.89, + "learning_rate": 1.0384937418149812e-05, + "loss": 0.9347, + "step": 44908 + }, + { + "epoch": 2.89, + "learning_rate": 1.0384722729126861e-05, + "loss": 0.8573, + "step": 44909 + }, + { + "epoch": 2.89, + "learning_rate": 1.038450804010391e-05, + "loss": 0.9208, + "step": 44910 + }, + { + "epoch": 2.89, + "learning_rate": 1.0384293351080961e-05, + "loss": 0.86, + "step": 44911 + }, + { + "epoch": 2.89, + "learning_rate": 1.038407866205801e-05, + "loss": 0.8286, + "step": 44912 + }, + { + "epoch": 2.89, + "learning_rate": 1.038386397303506e-05, + "loss": 0.8881, + "step": 44913 + }, + { + "epoch": 2.89, + "learning_rate": 1.038364928401211e-05, + "loss": 0.8958, + "step": 44914 + }, + { + "epoch": 2.89, + "learning_rate": 1.0383434594989159e-05, + "loss": 0.9189, + "step": 44915 + }, + { + "epoch": 2.89, + "learning_rate": 1.038321990596621e-05, + "loss": 0.9498, + "step": 44916 + }, + { + "epoch": 2.89, + "learning_rate": 1.0383005216943258e-05, + "loss": 0.8903, + "step": 44917 + }, + { + "epoch": 2.89, + "learning_rate": 1.0382790527920307e-05, + "loss": 0.9468, + "step": 44918 + }, + { + "epoch": 2.89, + "learning_rate": 1.0382575838897358e-05, + "loss": 0.9332, + "step": 44919 + }, + { + "epoch": 2.89, + "learning_rate": 1.0382361149874407e-05, + "loss": 0.9267, + "step": 44920 + }, + { + "epoch": 2.89, + "learning_rate": 1.0382146460851456e-05, + "loss": 0.8923, + "step": 44921 + }, + { + "epoch": 2.89, + "learning_rate": 1.0381931771828506e-05, + "loss": 0.843, + "step": 44922 + }, + { + "epoch": 2.89, + "learning_rate": 1.0381717082805559e-05, + "loss": 0.8511, + "step": 44923 + }, + { + "epoch": 2.89, + "learning_rate": 1.0381502393782608e-05, + "loss": 0.8658, + "step": 44924 + }, + { + "epoch": 2.89, + "learning_rate": 1.0381287704759657e-05, + "loss": 0.8408, + "step": 44925 + }, + { + "epoch": 2.89, + "learning_rate": 1.0381073015736707e-05, + "loss": 0.8855, + "step": 44926 + }, + { + "epoch": 2.89, + "learning_rate": 1.0380858326713756e-05, + "loss": 0.9321, + "step": 44927 + }, + { + "epoch": 2.89, + "learning_rate": 1.0380643637690805e-05, + "loss": 0.8183, + "step": 44928 + }, + { + "epoch": 2.89, + "learning_rate": 1.0380428948667856e-05, + "loss": 0.8954, + "step": 44929 + }, + { + "epoch": 2.89, + "learning_rate": 1.0380214259644905e-05, + "loss": 0.9653, + "step": 44930 + }, + { + "epoch": 2.89, + "learning_rate": 1.0379999570621954e-05, + "loss": 0.8983, + "step": 44931 + }, + { + "epoch": 2.89, + "learning_rate": 1.0379784881599005e-05, + "loss": 0.8885, + "step": 44932 + }, + { + "epoch": 2.89, + "learning_rate": 1.0379570192576054e-05, + "loss": 0.9582, + "step": 44933 + }, + { + "epoch": 2.89, + "learning_rate": 1.0379355503553104e-05, + "loss": 0.9387, + "step": 44934 + }, + { + "epoch": 2.89, + "learning_rate": 1.0379140814530153e-05, + "loss": 0.8665, + "step": 44935 + }, + { + "epoch": 2.89, + "learning_rate": 1.0378926125507202e-05, + "loss": 0.8634, + "step": 44936 + }, + { + "epoch": 2.89, + "learning_rate": 1.0378711436484254e-05, + "loss": 0.8275, + "step": 44937 + }, + { + "epoch": 2.89, + "learning_rate": 1.0378496747461303e-05, + "loss": 0.9667, + "step": 44938 + }, + { + "epoch": 2.89, + "learning_rate": 1.0378282058438354e-05, + "loss": 0.8798, + "step": 44939 + }, + { + "epoch": 2.89, + "learning_rate": 1.0378067369415403e-05, + "loss": 0.9611, + "step": 44940 + }, + { + "epoch": 2.89, + "learning_rate": 1.0377852680392452e-05, + "loss": 0.954, + "step": 44941 + }, + { + "epoch": 2.89, + "learning_rate": 1.0377637991369503e-05, + "loss": 0.9203, + "step": 44942 + }, + { + "epoch": 2.89, + "learning_rate": 1.0377423302346552e-05, + "loss": 0.8368, + "step": 44943 + }, + { + "epoch": 2.89, + "learning_rate": 1.0377208613323602e-05, + "loss": 0.8844, + "step": 44944 + }, + { + "epoch": 2.89, + "learning_rate": 1.0376993924300651e-05, + "loss": 0.9274, + "step": 44945 + }, + { + "epoch": 2.89, + "learning_rate": 1.03767792352777e-05, + "loss": 0.9206, + "step": 44946 + }, + { + "epoch": 2.89, + "learning_rate": 1.0376564546254751e-05, + "loss": 0.8484, + "step": 44947 + }, + { + "epoch": 2.89, + "learning_rate": 1.03763498572318e-05, + "loss": 0.8571, + "step": 44948 + }, + { + "epoch": 2.89, + "learning_rate": 1.0376135168208849e-05, + "loss": 0.934, + "step": 44949 + }, + { + "epoch": 2.89, + "learning_rate": 1.03759204791859e-05, + "loss": 0.8793, + "step": 44950 + }, + { + "epoch": 2.89, + "learning_rate": 1.0375705790162948e-05, + "loss": 0.9473, + "step": 44951 + }, + { + "epoch": 2.89, + "learning_rate": 1.037549110114e-05, + "loss": 0.9577, + "step": 44952 + }, + { + "epoch": 2.89, + "learning_rate": 1.037527641211705e-05, + "loss": 0.8819, + "step": 44953 + }, + { + "epoch": 2.89, + "learning_rate": 1.03750617230941e-05, + "loss": 0.8961, + "step": 44954 + }, + { + "epoch": 2.89, + "learning_rate": 1.037484703407115e-05, + "loss": 0.9531, + "step": 44955 + }, + { + "epoch": 2.89, + "learning_rate": 1.0374632345048198e-05, + "loss": 0.8522, + "step": 44956 + }, + { + "epoch": 2.89, + "learning_rate": 1.0374417656025249e-05, + "loss": 0.8895, + "step": 44957 + }, + { + "epoch": 2.89, + "learning_rate": 1.0374202967002298e-05, + "loss": 0.8906, + "step": 44958 + }, + { + "epoch": 2.89, + "learning_rate": 1.0373988277979347e-05, + "loss": 0.904, + "step": 44959 + }, + { + "epoch": 2.89, + "learning_rate": 1.0373773588956398e-05, + "loss": 0.9239, + "step": 44960 + }, + { + "epoch": 2.89, + "learning_rate": 1.0373558899933447e-05, + "loss": 0.8947, + "step": 44961 + }, + { + "epoch": 2.89, + "learning_rate": 1.0373344210910497e-05, + "loss": 0.9081, + "step": 44962 + }, + { + "epoch": 2.89, + "learning_rate": 1.0373129521887546e-05, + "loss": 0.862, + "step": 44963 + }, + { + "epoch": 2.89, + "learning_rate": 1.0372914832864595e-05, + "loss": 0.86, + "step": 44964 + }, + { + "epoch": 2.89, + "learning_rate": 1.0372700143841646e-05, + "loss": 0.9588, + "step": 44965 + }, + { + "epoch": 2.89, + "learning_rate": 1.0372485454818696e-05, + "loss": 0.9667, + "step": 44966 + }, + { + "epoch": 2.89, + "learning_rate": 1.0372270765795747e-05, + "loss": 0.8799, + "step": 44967 + }, + { + "epoch": 2.89, + "learning_rate": 1.0372056076772796e-05, + "loss": 0.9307, + "step": 44968 + }, + { + "epoch": 2.89, + "learning_rate": 1.0371841387749845e-05, + "loss": 0.9135, + "step": 44969 + }, + { + "epoch": 2.89, + "learning_rate": 1.0371626698726896e-05, + "loss": 0.8917, + "step": 44970 + }, + { + "epoch": 2.89, + "learning_rate": 1.0371412009703945e-05, + "loss": 0.904, + "step": 44971 + }, + { + "epoch": 2.89, + "learning_rate": 1.0371197320680995e-05, + "loss": 0.9193, + "step": 44972 + }, + { + "epoch": 2.89, + "learning_rate": 1.0370982631658044e-05, + "loss": 0.9263, + "step": 44973 + }, + { + "epoch": 2.89, + "learning_rate": 1.0370767942635093e-05, + "loss": 0.8866, + "step": 44974 + }, + { + "epoch": 2.89, + "learning_rate": 1.0370553253612144e-05, + "loss": 0.8846, + "step": 44975 + }, + { + "epoch": 2.89, + "learning_rate": 1.0370338564589193e-05, + "loss": 0.9076, + "step": 44976 + }, + { + "epoch": 2.89, + "learning_rate": 1.0370123875566242e-05, + "loss": 0.8401, + "step": 44977 + }, + { + "epoch": 2.89, + "learning_rate": 1.0369909186543293e-05, + "loss": 0.842, + "step": 44978 + }, + { + "epoch": 2.89, + "learning_rate": 1.0369694497520341e-05, + "loss": 0.8835, + "step": 44979 + }, + { + "epoch": 2.89, + "learning_rate": 1.0369479808497392e-05, + "loss": 0.9855, + "step": 44980 + }, + { + "epoch": 2.89, + "learning_rate": 1.0369265119474443e-05, + "loss": 0.9478, + "step": 44981 + }, + { + "epoch": 2.89, + "learning_rate": 1.0369050430451493e-05, + "loss": 0.9939, + "step": 44982 + }, + { + "epoch": 2.89, + "learning_rate": 1.0368835741428542e-05, + "loss": 0.8952, + "step": 44983 + }, + { + "epoch": 2.89, + "learning_rate": 1.0368621052405591e-05, + "loss": 0.879, + "step": 44984 + }, + { + "epoch": 2.89, + "learning_rate": 1.0368406363382642e-05, + "loss": 0.8556, + "step": 44985 + }, + { + "epoch": 2.89, + "learning_rate": 1.0368191674359691e-05, + "loss": 0.8776, + "step": 44986 + }, + { + "epoch": 2.89, + "learning_rate": 1.036797698533674e-05, + "loss": 0.876, + "step": 44987 + }, + { + "epoch": 2.89, + "learning_rate": 1.036776229631379e-05, + "loss": 0.8425, + "step": 44988 + }, + { + "epoch": 2.89, + "learning_rate": 1.036754760729084e-05, + "loss": 0.8807, + "step": 44989 + }, + { + "epoch": 2.89, + "learning_rate": 1.036733291826789e-05, + "loss": 0.9284, + "step": 44990 + }, + { + "epoch": 2.89, + "learning_rate": 1.036711822924494e-05, + "loss": 0.8871, + "step": 44991 + }, + { + "epoch": 2.89, + "learning_rate": 1.0366903540221988e-05, + "loss": 0.8694, + "step": 44992 + }, + { + "epoch": 2.89, + "learning_rate": 1.0366688851199039e-05, + "loss": 0.9113, + "step": 44993 + }, + { + "epoch": 2.89, + "learning_rate": 1.0366474162176088e-05, + "loss": 0.9374, + "step": 44994 + }, + { + "epoch": 2.89, + "learning_rate": 1.036625947315314e-05, + "loss": 0.9783, + "step": 44995 + }, + { + "epoch": 2.89, + "learning_rate": 1.0366044784130189e-05, + "loss": 0.8633, + "step": 44996 + }, + { + "epoch": 2.9, + "learning_rate": 1.0365830095107238e-05, + "loss": 0.8692, + "step": 44997 + }, + { + "epoch": 2.9, + "learning_rate": 1.0365615406084289e-05, + "loss": 0.9268, + "step": 44998 + }, + { + "epoch": 2.9, + "learning_rate": 1.0365400717061338e-05, + "loss": 0.9309, + "step": 44999 + }, + { + "epoch": 2.9, + "learning_rate": 1.0365186028038388e-05, + "loss": 0.8435, + "step": 45000 + }, + { + "epoch": 2.9, + "eval_laion/cider": 0.5533484638884172, + "eval_laion/meteor": 0.3962713423942672, + "eval_laion/pred_num_words": 37.640625, + "eval_laion/sacrebleu": 16.10243161216638, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0353193283081055, + "eval_runtime": 102.809, + "eval_samples_per_second": 4.854, + "eval_steps_per_second": 0.311, + "step": 45000 + }, + { + "epoch": 2.9, + "learning_rate": 1.0364971339015437e-05, + "loss": 0.8678, + "step": 45001 + }, + { + "epoch": 2.9, + "learning_rate": 1.0364756649992486e-05, + "loss": 0.8876, + "step": 45002 + }, + { + "epoch": 2.9, + "learning_rate": 1.0364541960969537e-05, + "loss": 0.8672, + "step": 45003 + }, + { + "epoch": 2.9, + "learning_rate": 1.0364327271946586e-05, + "loss": 0.9196, + "step": 45004 + }, + { + "epoch": 2.9, + "learning_rate": 1.0364112582923635e-05, + "loss": 0.907, + "step": 45005 + }, + { + "epoch": 2.9, + "learning_rate": 1.0363897893900686e-05, + "loss": 0.9088, + "step": 45006 + }, + { + "epoch": 2.9, + "learning_rate": 1.0363683204877735e-05, + "loss": 0.8936, + "step": 45007 + }, + { + "epoch": 2.9, + "learning_rate": 1.0363468515854784e-05, + "loss": 0.9212, + "step": 45008 + }, + { + "epoch": 2.9, + "learning_rate": 1.0363253826831834e-05, + "loss": 0.935, + "step": 45009 + }, + { + "epoch": 2.9, + "learning_rate": 1.0363039137808887e-05, + "loss": 0.8656, + "step": 45010 + }, + { + "epoch": 2.9, + "learning_rate": 1.0362824448785935e-05, + "loss": 0.9305, + "step": 45011 + }, + { + "epoch": 2.9, + "learning_rate": 1.0362609759762984e-05, + "loss": 0.9129, + "step": 45012 + }, + { + "epoch": 2.9, + "learning_rate": 1.0362395070740035e-05, + "loss": 0.9623, + "step": 45013 + }, + { + "epoch": 2.9, + "learning_rate": 1.0362180381717084e-05, + "loss": 0.9601, + "step": 45014 + }, + { + "epoch": 2.9, + "learning_rate": 1.0361965692694133e-05, + "loss": 0.9593, + "step": 45015 + }, + { + "epoch": 2.9, + "learning_rate": 1.0361751003671184e-05, + "loss": 0.8249, + "step": 45016 + }, + { + "epoch": 2.9, + "learning_rate": 1.0361536314648233e-05, + "loss": 0.8864, + "step": 45017 + }, + { + "epoch": 2.9, + "learning_rate": 1.0361321625625282e-05, + "loss": 0.8825, + "step": 45018 + }, + { + "epoch": 2.9, + "learning_rate": 1.0361106936602332e-05, + "loss": 0.8891, + "step": 45019 + }, + { + "epoch": 2.9, + "learning_rate": 1.0360892247579381e-05, + "loss": 0.8922, + "step": 45020 + }, + { + "epoch": 2.9, + "learning_rate": 1.0360677558556432e-05, + "loss": 0.9392, + "step": 45021 + }, + { + "epoch": 2.9, + "learning_rate": 1.0360462869533481e-05, + "loss": 0.8742, + "step": 45022 + }, + { + "epoch": 2.9, + "learning_rate": 1.036024818051053e-05, + "loss": 0.8912, + "step": 45023 + }, + { + "epoch": 2.9, + "learning_rate": 1.0360033491487582e-05, + "loss": 0.8756, + "step": 45024 + }, + { + "epoch": 2.9, + "learning_rate": 1.0359818802464631e-05, + "loss": 0.9313, + "step": 45025 + }, + { + "epoch": 2.9, + "learning_rate": 1.0359604113441682e-05, + "loss": 0.8371, + "step": 45026 + }, + { + "epoch": 2.9, + "learning_rate": 1.035938942441873e-05, + "loss": 0.9194, + "step": 45027 + }, + { + "epoch": 2.9, + "learning_rate": 1.035917473539578e-05, + "loss": 0.8973, + "step": 45028 + }, + { + "epoch": 2.9, + "learning_rate": 1.035896004637283e-05, + "loss": 0.9347, + "step": 45029 + }, + { + "epoch": 2.9, + "learning_rate": 1.035874535734988e-05, + "loss": 0.9, + "step": 45030 + }, + { + "epoch": 2.9, + "learning_rate": 1.035853066832693e-05, + "loss": 0.8455, + "step": 45031 + }, + { + "epoch": 2.9, + "learning_rate": 1.0358315979303979e-05, + "loss": 0.9206, + "step": 45032 + }, + { + "epoch": 2.9, + "learning_rate": 1.0358101290281028e-05, + "loss": 0.8306, + "step": 45033 + }, + { + "epoch": 2.9, + "learning_rate": 1.0357886601258079e-05, + "loss": 0.8806, + "step": 45034 + }, + { + "epoch": 2.9, + "learning_rate": 1.0357671912235128e-05, + "loss": 0.9159, + "step": 45035 + }, + { + "epoch": 2.9, + "learning_rate": 1.0357457223212177e-05, + "loss": 0.8748, + "step": 45036 + }, + { + "epoch": 2.9, + "learning_rate": 1.0357242534189227e-05, + "loss": 0.9046, + "step": 45037 + }, + { + "epoch": 2.9, + "learning_rate": 1.0357027845166276e-05, + "loss": 0.9548, + "step": 45038 + }, + { + "epoch": 2.9, + "learning_rate": 1.0356813156143329e-05, + "loss": 0.8344, + "step": 45039 + }, + { + "epoch": 2.9, + "learning_rate": 1.0356598467120377e-05, + "loss": 0.9299, + "step": 45040 + }, + { + "epoch": 2.9, + "learning_rate": 1.0356383778097428e-05, + "loss": 0.912, + "step": 45041 + }, + { + "epoch": 2.9, + "learning_rate": 1.0356169089074477e-05, + "loss": 0.8398, + "step": 45042 + }, + { + "epoch": 2.9, + "learning_rate": 1.0355954400051526e-05, + "loss": 0.8674, + "step": 45043 + }, + { + "epoch": 2.9, + "learning_rate": 1.0355739711028577e-05, + "loss": 0.8849, + "step": 45044 + }, + { + "epoch": 2.9, + "learning_rate": 1.0355525022005626e-05, + "loss": 0.9117, + "step": 45045 + }, + { + "epoch": 2.9, + "learning_rate": 1.0355310332982675e-05, + "loss": 0.8338, + "step": 45046 + }, + { + "epoch": 2.9, + "learning_rate": 1.0355095643959725e-05, + "loss": 0.8791, + "step": 45047 + }, + { + "epoch": 2.9, + "learning_rate": 1.0354880954936774e-05, + "loss": 0.9006, + "step": 45048 + }, + { + "epoch": 2.9, + "learning_rate": 1.0354666265913825e-05, + "loss": 0.835, + "step": 45049 + }, + { + "epoch": 2.9, + "learning_rate": 1.0354451576890874e-05, + "loss": 1.001, + "step": 45050 + }, + { + "epoch": 2.9, + "learning_rate": 1.0354236887867923e-05, + "loss": 0.9275, + "step": 45051 + }, + { + "epoch": 2.9, + "learning_rate": 1.0354022198844974e-05, + "loss": 0.8559, + "step": 45052 + }, + { + "epoch": 2.9, + "learning_rate": 1.0353807509822023e-05, + "loss": 0.9649, + "step": 45053 + }, + { + "epoch": 2.9, + "learning_rate": 1.0353592820799075e-05, + "loss": 0.9431, + "step": 45054 + }, + { + "epoch": 2.9, + "learning_rate": 1.0353378131776124e-05, + "loss": 0.8755, + "step": 45055 + }, + { + "epoch": 2.9, + "learning_rate": 1.0353163442753173e-05, + "loss": 0.9372, + "step": 45056 + }, + { + "epoch": 2.9, + "learning_rate": 1.0352948753730223e-05, + "loss": 0.9285, + "step": 45057 + }, + { + "epoch": 2.9, + "learning_rate": 1.0352734064707272e-05, + "loss": 0.8657, + "step": 45058 + }, + { + "epoch": 2.9, + "learning_rate": 1.0352519375684323e-05, + "loss": 0.9188, + "step": 45059 + }, + { + "epoch": 2.9, + "learning_rate": 1.0352304686661372e-05, + "loss": 0.9265, + "step": 45060 + }, + { + "epoch": 2.9, + "learning_rate": 1.0352089997638421e-05, + "loss": 0.9258, + "step": 45061 + }, + { + "epoch": 2.9, + "learning_rate": 1.0351875308615472e-05, + "loss": 0.9128, + "step": 45062 + }, + { + "epoch": 2.9, + "learning_rate": 1.035166061959252e-05, + "loss": 0.8677, + "step": 45063 + }, + { + "epoch": 2.9, + "learning_rate": 1.035144593056957e-05, + "loss": 0.9175, + "step": 45064 + }, + { + "epoch": 2.9, + "learning_rate": 1.035123124154662e-05, + "loss": 0.925, + "step": 45065 + }, + { + "epoch": 2.9, + "learning_rate": 1.035101655252367e-05, + "loss": 0.9883, + "step": 45066 + }, + { + "epoch": 2.9, + "learning_rate": 1.035080186350072e-05, + "loss": 0.9206, + "step": 45067 + }, + { + "epoch": 2.9, + "learning_rate": 1.035058717447777e-05, + "loss": 0.928, + "step": 45068 + }, + { + "epoch": 2.9, + "learning_rate": 1.0350372485454821e-05, + "loss": 0.8612, + "step": 45069 + }, + { + "epoch": 2.9, + "learning_rate": 1.035015779643187e-05, + "loss": 0.9155, + "step": 45070 + }, + { + "epoch": 2.9, + "learning_rate": 1.0349943107408919e-05, + "loss": 0.9184, + "step": 45071 + }, + { + "epoch": 2.9, + "learning_rate": 1.034972841838597e-05, + "loss": 0.9397, + "step": 45072 + }, + { + "epoch": 2.9, + "learning_rate": 1.0349513729363019e-05, + "loss": 0.8559, + "step": 45073 + }, + { + "epoch": 2.9, + "learning_rate": 1.0349299040340068e-05, + "loss": 0.9277, + "step": 45074 + }, + { + "epoch": 2.9, + "learning_rate": 1.0349084351317118e-05, + "loss": 0.9196, + "step": 45075 + }, + { + "epoch": 2.9, + "learning_rate": 1.0348869662294167e-05, + "loss": 0.9407, + "step": 45076 + }, + { + "epoch": 2.9, + "learning_rate": 1.0348654973271218e-05, + "loss": 0.9168, + "step": 45077 + }, + { + "epoch": 2.9, + "learning_rate": 1.0348440284248267e-05, + "loss": 0.8746, + "step": 45078 + }, + { + "epoch": 2.9, + "learning_rate": 1.0348225595225316e-05, + "loss": 0.8575, + "step": 45079 + }, + { + "epoch": 2.9, + "learning_rate": 1.0348010906202367e-05, + "loss": 0.8687, + "step": 45080 + }, + { + "epoch": 2.9, + "learning_rate": 1.0347796217179416e-05, + "loss": 0.9088, + "step": 45081 + }, + { + "epoch": 2.9, + "learning_rate": 1.0347581528156465e-05, + "loss": 0.9059, + "step": 45082 + }, + { + "epoch": 2.9, + "learning_rate": 1.0347366839133517e-05, + "loss": 0.8596, + "step": 45083 + }, + { + "epoch": 2.9, + "learning_rate": 1.0347152150110566e-05, + "loss": 0.9558, + "step": 45084 + }, + { + "epoch": 2.9, + "learning_rate": 1.0346937461087616e-05, + "loss": 0.8698, + "step": 45085 + }, + { + "epoch": 2.9, + "learning_rate": 1.0346722772064665e-05, + "loss": 0.8935, + "step": 45086 + }, + { + "epoch": 2.9, + "learning_rate": 1.0346508083041716e-05, + "loss": 0.8818, + "step": 45087 + }, + { + "epoch": 2.9, + "learning_rate": 1.0346293394018765e-05, + "loss": 0.9329, + "step": 45088 + }, + { + "epoch": 2.9, + "learning_rate": 1.0346078704995814e-05, + "loss": 1.0111, + "step": 45089 + }, + { + "epoch": 2.9, + "learning_rate": 1.0345864015972865e-05, + "loss": 0.9249, + "step": 45090 + }, + { + "epoch": 2.9, + "learning_rate": 1.0345649326949914e-05, + "loss": 0.8905, + "step": 45091 + }, + { + "epoch": 2.9, + "learning_rate": 1.0345434637926963e-05, + "loss": 0.9072, + "step": 45092 + }, + { + "epoch": 2.9, + "learning_rate": 1.0345219948904013e-05, + "loss": 0.9113, + "step": 45093 + }, + { + "epoch": 2.9, + "learning_rate": 1.0345005259881062e-05, + "loss": 0.8462, + "step": 45094 + }, + { + "epoch": 2.9, + "learning_rate": 1.0344790570858111e-05, + "loss": 0.8403, + "step": 45095 + }, + { + "epoch": 2.9, + "learning_rate": 1.0344575881835162e-05, + "loss": 0.9087, + "step": 45096 + }, + { + "epoch": 2.9, + "learning_rate": 1.0344361192812214e-05, + "loss": 0.907, + "step": 45097 + }, + { + "epoch": 2.9, + "learning_rate": 1.0344146503789263e-05, + "loss": 0.8624, + "step": 45098 + }, + { + "epoch": 2.9, + "learning_rate": 1.0343931814766312e-05, + "loss": 0.9, + "step": 45099 + }, + { + "epoch": 2.9, + "learning_rate": 1.0343717125743363e-05, + "loss": 0.8852, + "step": 45100 + }, + { + "epoch": 2.9, + "learning_rate": 1.0343502436720412e-05, + "loss": 0.9431, + "step": 45101 + }, + { + "epoch": 2.9, + "learning_rate": 1.034328774769746e-05, + "loss": 0.899, + "step": 45102 + }, + { + "epoch": 2.9, + "learning_rate": 1.0343073058674511e-05, + "loss": 0.9022, + "step": 45103 + }, + { + "epoch": 2.9, + "learning_rate": 1.034285836965156e-05, + "loss": 0.9424, + "step": 45104 + }, + { + "epoch": 2.9, + "learning_rate": 1.034264368062861e-05, + "loss": 0.8539, + "step": 45105 + }, + { + "epoch": 2.9, + "learning_rate": 1.034242899160566e-05, + "loss": 0.9384, + "step": 45106 + }, + { + "epoch": 2.9, + "learning_rate": 1.0342214302582709e-05, + "loss": 0.8968, + "step": 45107 + }, + { + "epoch": 2.9, + "learning_rate": 1.034199961355976e-05, + "loss": 0.9191, + "step": 45108 + }, + { + "epoch": 2.9, + "learning_rate": 1.0341784924536809e-05, + "loss": 0.9069, + "step": 45109 + }, + { + "epoch": 2.9, + "learning_rate": 1.0341570235513858e-05, + "loss": 0.8563, + "step": 45110 + }, + { + "epoch": 2.9, + "learning_rate": 1.0341355546490908e-05, + "loss": 0.9329, + "step": 45111 + }, + { + "epoch": 2.9, + "learning_rate": 1.0341140857467959e-05, + "loss": 0.8845, + "step": 45112 + }, + { + "epoch": 2.9, + "learning_rate": 1.034092616844501e-05, + "loss": 0.8859, + "step": 45113 + }, + { + "epoch": 2.9, + "learning_rate": 1.0340711479422058e-05, + "loss": 0.8536, + "step": 45114 + }, + { + "epoch": 2.9, + "learning_rate": 1.0340496790399107e-05, + "loss": 0.8874, + "step": 45115 + }, + { + "epoch": 2.9, + "learning_rate": 1.0340282101376158e-05, + "loss": 0.8366, + "step": 45116 + }, + { + "epoch": 2.9, + "learning_rate": 1.0340067412353207e-05, + "loss": 0.9414, + "step": 45117 + }, + { + "epoch": 2.9, + "learning_rate": 1.0339852723330258e-05, + "loss": 0.8979, + "step": 45118 + }, + { + "epoch": 2.9, + "learning_rate": 1.0339638034307307e-05, + "loss": 0.847, + "step": 45119 + }, + { + "epoch": 2.9, + "learning_rate": 1.0339423345284356e-05, + "loss": 0.889, + "step": 45120 + }, + { + "epoch": 2.9, + "learning_rate": 1.0339208656261406e-05, + "loss": 0.8204, + "step": 45121 + }, + { + "epoch": 2.9, + "learning_rate": 1.0338993967238455e-05, + "loss": 0.8346, + "step": 45122 + }, + { + "epoch": 2.9, + "learning_rate": 1.0338779278215504e-05, + "loss": 0.9092, + "step": 45123 + }, + { + "epoch": 2.9, + "learning_rate": 1.0338564589192555e-05, + "loss": 0.9246, + "step": 45124 + }, + { + "epoch": 2.9, + "learning_rate": 1.0338349900169604e-05, + "loss": 0.9205, + "step": 45125 + }, + { + "epoch": 2.9, + "learning_rate": 1.0338135211146656e-05, + "loss": 0.9145, + "step": 45126 + }, + { + "epoch": 2.9, + "learning_rate": 1.0337920522123705e-05, + "loss": 0.8537, + "step": 45127 + }, + { + "epoch": 2.9, + "learning_rate": 1.0337705833100756e-05, + "loss": 0.9468, + "step": 45128 + }, + { + "epoch": 2.9, + "learning_rate": 1.0337491144077805e-05, + "loss": 0.9623, + "step": 45129 + }, + { + "epoch": 2.9, + "learning_rate": 1.0337276455054854e-05, + "loss": 0.8612, + "step": 45130 + }, + { + "epoch": 2.9, + "learning_rate": 1.0337061766031904e-05, + "loss": 0.889, + "step": 45131 + }, + { + "epoch": 2.9, + "learning_rate": 1.0336847077008953e-05, + "loss": 0.9051, + "step": 45132 + }, + { + "epoch": 2.9, + "learning_rate": 1.0336632387986002e-05, + "loss": 0.8677, + "step": 45133 + }, + { + "epoch": 2.9, + "learning_rate": 1.0336417698963053e-05, + "loss": 0.9095, + "step": 45134 + }, + { + "epoch": 2.9, + "learning_rate": 1.0336203009940102e-05, + "loss": 0.9613, + "step": 45135 + }, + { + "epoch": 2.9, + "learning_rate": 1.0335988320917153e-05, + "loss": 0.8295, + "step": 45136 + }, + { + "epoch": 2.9, + "learning_rate": 1.0335773631894202e-05, + "loss": 0.8472, + "step": 45137 + }, + { + "epoch": 2.9, + "learning_rate": 1.033555894287125e-05, + "loss": 0.8515, + "step": 45138 + }, + { + "epoch": 2.9, + "learning_rate": 1.0335344253848301e-05, + "loss": 0.9062, + "step": 45139 + }, + { + "epoch": 2.9, + "learning_rate": 1.033512956482535e-05, + "loss": 0.8533, + "step": 45140 + }, + { + "epoch": 2.9, + "learning_rate": 1.0334914875802403e-05, + "loss": 0.9341, + "step": 45141 + }, + { + "epoch": 2.9, + "learning_rate": 1.0334700186779452e-05, + "loss": 0.8884, + "step": 45142 + }, + { + "epoch": 2.9, + "learning_rate": 1.03344854977565e-05, + "loss": 0.8382, + "step": 45143 + }, + { + "epoch": 2.9, + "learning_rate": 1.0334270808733551e-05, + "loss": 0.9413, + "step": 45144 + }, + { + "epoch": 2.9, + "learning_rate": 1.03340561197106e-05, + "loss": 0.8267, + "step": 45145 + }, + { + "epoch": 2.9, + "learning_rate": 1.033384143068765e-05, + "loss": 0.9306, + "step": 45146 + }, + { + "epoch": 2.9, + "learning_rate": 1.03336267416647e-05, + "loss": 0.8907, + "step": 45147 + }, + { + "epoch": 2.9, + "learning_rate": 1.0333412052641749e-05, + "loss": 0.8549, + "step": 45148 + }, + { + "epoch": 2.9, + "learning_rate": 1.03331973636188e-05, + "loss": 0.8475, + "step": 45149 + }, + { + "epoch": 2.9, + "learning_rate": 1.0332982674595848e-05, + "loss": 0.8974, + "step": 45150 + }, + { + "epoch": 2.9, + "learning_rate": 1.0332767985572897e-05, + "loss": 0.9297, + "step": 45151 + }, + { + "epoch": 2.9, + "learning_rate": 1.0332553296549948e-05, + "loss": 0.7888, + "step": 45152 + }, + { + "epoch": 2.91, + "learning_rate": 1.0332338607526997e-05, + "loss": 0.8392, + "step": 45153 + }, + { + "epoch": 2.91, + "learning_rate": 1.0332123918504046e-05, + "loss": 0.8301, + "step": 45154 + }, + { + "epoch": 2.91, + "learning_rate": 1.0331909229481098e-05, + "loss": 0.8901, + "step": 45155 + }, + { + "epoch": 2.91, + "learning_rate": 1.0331694540458149e-05, + "loss": 0.9143, + "step": 45156 + }, + { + "epoch": 2.91, + "learning_rate": 1.0331479851435198e-05, + "loss": 0.9058, + "step": 45157 + }, + { + "epoch": 2.91, + "learning_rate": 1.0331265162412247e-05, + "loss": 0.9127, + "step": 45158 + }, + { + "epoch": 2.91, + "learning_rate": 1.0331050473389297e-05, + "loss": 0.9855, + "step": 45159 + }, + { + "epoch": 2.91, + "learning_rate": 1.0330835784366346e-05, + "loss": 0.9837, + "step": 45160 + }, + { + "epoch": 2.91, + "learning_rate": 1.0330621095343395e-05, + "loss": 0.8796, + "step": 45161 + }, + { + "epoch": 2.91, + "learning_rate": 1.0330406406320446e-05, + "loss": 0.9245, + "step": 45162 + }, + { + "epoch": 2.91, + "learning_rate": 1.0330191717297495e-05, + "loss": 0.9286, + "step": 45163 + }, + { + "epoch": 2.91, + "learning_rate": 1.0329977028274544e-05, + "loss": 1.0213, + "step": 45164 + }, + { + "epoch": 2.91, + "learning_rate": 1.0329762339251595e-05, + "loss": 0.9474, + "step": 45165 + }, + { + "epoch": 2.91, + "learning_rate": 1.0329547650228644e-05, + "loss": 0.9302, + "step": 45166 + }, + { + "epoch": 2.91, + "learning_rate": 1.0329332961205694e-05, + "loss": 0.8936, + "step": 45167 + }, + { + "epoch": 2.91, + "learning_rate": 1.0329118272182743e-05, + "loss": 0.8786, + "step": 45168 + }, + { + "epoch": 2.91, + "learning_rate": 1.0328903583159792e-05, + "loss": 0.8385, + "step": 45169 + }, + { + "epoch": 2.91, + "learning_rate": 1.0328688894136845e-05, + "loss": 0.9048, + "step": 45170 + }, + { + "epoch": 2.91, + "learning_rate": 1.0328474205113894e-05, + "loss": 0.952, + "step": 45171 + }, + { + "epoch": 2.91, + "learning_rate": 1.0328259516090944e-05, + "loss": 0.861, + "step": 45172 + }, + { + "epoch": 2.91, + "learning_rate": 1.0328044827067993e-05, + "loss": 0.9572, + "step": 45173 + }, + { + "epoch": 2.91, + "learning_rate": 1.0327830138045042e-05, + "loss": 0.8875, + "step": 45174 + }, + { + "epoch": 2.91, + "learning_rate": 1.0327615449022093e-05, + "loss": 0.8654, + "step": 45175 + }, + { + "epoch": 2.91, + "learning_rate": 1.0327400759999142e-05, + "loss": 0.8975, + "step": 45176 + }, + { + "epoch": 2.91, + "learning_rate": 1.0327186070976192e-05, + "loss": 0.8813, + "step": 45177 + }, + { + "epoch": 2.91, + "learning_rate": 1.0326971381953241e-05, + "loss": 0.9132, + "step": 45178 + }, + { + "epoch": 2.91, + "learning_rate": 1.032675669293029e-05, + "loss": 0.8726, + "step": 45179 + }, + { + "epoch": 2.91, + "learning_rate": 1.0326542003907341e-05, + "loss": 0.8139, + "step": 45180 + }, + { + "epoch": 2.91, + "learning_rate": 1.032632731488439e-05, + "loss": 0.847, + "step": 45181 + }, + { + "epoch": 2.91, + "learning_rate": 1.0326112625861439e-05, + "loss": 0.8986, + "step": 45182 + }, + { + "epoch": 2.91, + "learning_rate": 1.032589793683849e-05, + "loss": 0.9319, + "step": 45183 + }, + { + "epoch": 2.91, + "learning_rate": 1.0325683247815542e-05, + "loss": 0.896, + "step": 45184 + }, + { + "epoch": 2.91, + "learning_rate": 1.0325468558792591e-05, + "loss": 0.8142, + "step": 45185 + }, + { + "epoch": 2.91, + "learning_rate": 1.032525386976964e-05, + "loss": 0.8653, + "step": 45186 + }, + { + "epoch": 2.91, + "learning_rate": 1.032503918074669e-05, + "loss": 0.8899, + "step": 45187 + }, + { + "epoch": 2.91, + "learning_rate": 1.032482449172374e-05, + "loss": 0.8373, + "step": 45188 + }, + { + "epoch": 2.91, + "learning_rate": 1.0324609802700788e-05, + "loss": 0.8729, + "step": 45189 + }, + { + "epoch": 2.91, + "learning_rate": 1.0324395113677839e-05, + "loss": 0.9097, + "step": 45190 + }, + { + "epoch": 2.91, + "learning_rate": 1.0324180424654888e-05, + "loss": 0.9273, + "step": 45191 + }, + { + "epoch": 2.91, + "learning_rate": 1.0323965735631937e-05, + "loss": 0.9903, + "step": 45192 + }, + { + "epoch": 2.91, + "learning_rate": 1.0323751046608988e-05, + "loss": 0.8492, + "step": 45193 + }, + { + "epoch": 2.91, + "learning_rate": 1.0323536357586037e-05, + "loss": 0.926, + "step": 45194 + }, + { + "epoch": 2.91, + "learning_rate": 1.0323321668563087e-05, + "loss": 0.8161, + "step": 45195 + }, + { + "epoch": 2.91, + "learning_rate": 1.0323106979540136e-05, + "loss": 0.8452, + "step": 45196 + }, + { + "epoch": 2.91, + "learning_rate": 1.0322892290517185e-05, + "loss": 0.9002, + "step": 45197 + }, + { + "epoch": 2.91, + "learning_rate": 1.0322677601494236e-05, + "loss": 0.9105, + "step": 45198 + }, + { + "epoch": 2.91, + "learning_rate": 1.0322462912471287e-05, + "loss": 0.8596, + "step": 45199 + }, + { + "epoch": 2.91, + "learning_rate": 1.0322248223448337e-05, + "loss": 0.9122, + "step": 45200 + }, + { + "epoch": 2.91, + "learning_rate": 1.0322033534425386e-05, + "loss": 0.9499, + "step": 45201 + }, + { + "epoch": 2.91, + "learning_rate": 1.0321818845402435e-05, + "loss": 0.9302, + "step": 45202 + }, + { + "epoch": 2.91, + "learning_rate": 1.0321604156379486e-05, + "loss": 0.9205, + "step": 45203 + }, + { + "epoch": 2.91, + "learning_rate": 1.0321389467356535e-05, + "loss": 0.8635, + "step": 45204 + }, + { + "epoch": 2.91, + "learning_rate": 1.0321174778333585e-05, + "loss": 0.9332, + "step": 45205 + }, + { + "epoch": 2.91, + "learning_rate": 1.0320960089310634e-05, + "loss": 0.9506, + "step": 45206 + }, + { + "epoch": 2.91, + "learning_rate": 1.0320745400287683e-05, + "loss": 0.8485, + "step": 45207 + }, + { + "epoch": 2.91, + "learning_rate": 1.0320530711264734e-05, + "loss": 0.8728, + "step": 45208 + }, + { + "epoch": 2.91, + "learning_rate": 1.0320316022241783e-05, + "loss": 0.8464, + "step": 45209 + }, + { + "epoch": 2.91, + "learning_rate": 1.0320101333218832e-05, + "loss": 0.8343, + "step": 45210 + }, + { + "epoch": 2.91, + "learning_rate": 1.0319886644195883e-05, + "loss": 0.8942, + "step": 45211 + }, + { + "epoch": 2.91, + "learning_rate": 1.0319671955172932e-05, + "loss": 0.8688, + "step": 45212 + }, + { + "epoch": 2.91, + "learning_rate": 1.0319457266149984e-05, + "loss": 0.8762, + "step": 45213 + }, + { + "epoch": 2.91, + "learning_rate": 1.0319242577127033e-05, + "loss": 0.9321, + "step": 45214 + }, + { + "epoch": 2.91, + "learning_rate": 1.0319027888104084e-05, + "loss": 0.9369, + "step": 45215 + }, + { + "epoch": 2.91, + "learning_rate": 1.0318813199081133e-05, + "loss": 0.9176, + "step": 45216 + }, + { + "epoch": 2.91, + "learning_rate": 1.0318598510058182e-05, + "loss": 0.8349, + "step": 45217 + }, + { + "epoch": 2.91, + "learning_rate": 1.0318383821035232e-05, + "loss": 0.9433, + "step": 45218 + }, + { + "epoch": 2.91, + "learning_rate": 1.0318169132012281e-05, + "loss": 0.8969, + "step": 45219 + }, + { + "epoch": 2.91, + "learning_rate": 1.031795444298933e-05, + "loss": 0.9047, + "step": 45220 + }, + { + "epoch": 2.91, + "learning_rate": 1.031773975396638e-05, + "loss": 0.8928, + "step": 45221 + }, + { + "epoch": 2.91, + "learning_rate": 1.031752506494343e-05, + "loss": 0.8719, + "step": 45222 + }, + { + "epoch": 2.91, + "learning_rate": 1.031731037592048e-05, + "loss": 0.9105, + "step": 45223 + }, + { + "epoch": 2.91, + "learning_rate": 1.031709568689753e-05, + "loss": 0.8804, + "step": 45224 + }, + { + "epoch": 2.91, + "learning_rate": 1.0316880997874578e-05, + "loss": 0.8926, + "step": 45225 + }, + { + "epoch": 2.91, + "learning_rate": 1.0316666308851629e-05, + "loss": 0.9139, + "step": 45226 + }, + { + "epoch": 2.91, + "learning_rate": 1.0316451619828678e-05, + "loss": 0.8711, + "step": 45227 + }, + { + "epoch": 2.91, + "learning_rate": 1.031623693080573e-05, + "loss": 0.8607, + "step": 45228 + }, + { + "epoch": 2.91, + "learning_rate": 1.031602224178278e-05, + "loss": 0.9274, + "step": 45229 + }, + { + "epoch": 2.91, + "learning_rate": 1.0315807552759828e-05, + "loss": 0.8813, + "step": 45230 + }, + { + "epoch": 2.91, + "learning_rate": 1.0315592863736879e-05, + "loss": 0.9278, + "step": 45231 + }, + { + "epoch": 2.91, + "learning_rate": 1.0315378174713928e-05, + "loss": 0.8977, + "step": 45232 + }, + { + "epoch": 2.91, + "learning_rate": 1.0315163485690978e-05, + "loss": 0.8444, + "step": 45233 + }, + { + "epoch": 2.91, + "learning_rate": 1.0314948796668027e-05, + "loss": 0.8804, + "step": 45234 + }, + { + "epoch": 2.91, + "learning_rate": 1.0314734107645076e-05, + "loss": 0.934, + "step": 45235 + }, + { + "epoch": 2.91, + "learning_rate": 1.0314519418622127e-05, + "loss": 0.895, + "step": 45236 + }, + { + "epoch": 2.91, + "learning_rate": 1.0314304729599176e-05, + "loss": 0.8502, + "step": 45237 + }, + { + "epoch": 2.91, + "learning_rate": 1.0314090040576225e-05, + "loss": 0.8559, + "step": 45238 + }, + { + "epoch": 2.91, + "learning_rate": 1.0313875351553276e-05, + "loss": 0.9125, + "step": 45239 + }, + { + "epoch": 2.91, + "learning_rate": 1.0313660662530325e-05, + "loss": 0.8696, + "step": 45240 + }, + { + "epoch": 2.91, + "learning_rate": 1.0313445973507374e-05, + "loss": 0.9766, + "step": 45241 + }, + { + "epoch": 2.91, + "learning_rate": 1.0313231284484426e-05, + "loss": 0.9041, + "step": 45242 + }, + { + "epoch": 2.91, + "learning_rate": 1.0313016595461477e-05, + "loss": 0.9098, + "step": 45243 + }, + { + "epoch": 2.91, + "learning_rate": 1.0312801906438526e-05, + "loss": 0.8184, + "step": 45244 + }, + { + "epoch": 2.91, + "learning_rate": 1.0312587217415575e-05, + "loss": 0.8905, + "step": 45245 + }, + { + "epoch": 2.91, + "learning_rate": 1.0312372528392625e-05, + "loss": 0.8947, + "step": 45246 + }, + { + "epoch": 2.91, + "learning_rate": 1.0312157839369674e-05, + "loss": 0.8772, + "step": 45247 + }, + { + "epoch": 2.91, + "learning_rate": 1.0311943150346723e-05, + "loss": 0.8778, + "step": 45248 + }, + { + "epoch": 2.91, + "learning_rate": 1.0311728461323774e-05, + "loss": 0.8183, + "step": 45249 + }, + { + "epoch": 2.91, + "learning_rate": 1.0311513772300823e-05, + "loss": 0.8914, + "step": 45250 + }, + { + "epoch": 2.91, + "learning_rate": 1.0311299083277872e-05, + "loss": 0.9304, + "step": 45251 + }, + { + "epoch": 2.91, + "learning_rate": 1.0311084394254922e-05, + "loss": 0.8603, + "step": 45252 + }, + { + "epoch": 2.91, + "learning_rate": 1.0310869705231971e-05, + "loss": 0.9089, + "step": 45253 + }, + { + "epoch": 2.91, + "learning_rate": 1.0310655016209022e-05, + "loss": 0.8579, + "step": 45254 + }, + { + "epoch": 2.91, + "learning_rate": 1.0310440327186071e-05, + "loss": 0.8987, + "step": 45255 + }, + { + "epoch": 2.91, + "learning_rate": 1.031022563816312e-05, + "loss": 0.8872, + "step": 45256 + }, + { + "epoch": 2.91, + "learning_rate": 1.0310010949140172e-05, + "loss": 0.8813, + "step": 45257 + }, + { + "epoch": 2.91, + "learning_rate": 1.0309796260117221e-05, + "loss": 0.8568, + "step": 45258 + }, + { + "epoch": 2.91, + "learning_rate": 1.0309581571094272e-05, + "loss": 0.9007, + "step": 45259 + }, + { + "epoch": 2.91, + "learning_rate": 1.0309366882071321e-05, + "loss": 0.8762, + "step": 45260 + }, + { + "epoch": 2.91, + "learning_rate": 1.030915219304837e-05, + "loss": 0.8863, + "step": 45261 + }, + { + "epoch": 2.91, + "learning_rate": 1.030893750402542e-05, + "loss": 0.886, + "step": 45262 + }, + { + "epoch": 2.91, + "learning_rate": 1.030872281500247e-05, + "loss": 0.9342, + "step": 45263 + }, + { + "epoch": 2.91, + "learning_rate": 1.030850812597952e-05, + "loss": 0.9264, + "step": 45264 + }, + { + "epoch": 2.91, + "learning_rate": 1.0308293436956569e-05, + "loss": 0.8968, + "step": 45265 + }, + { + "epoch": 2.91, + "learning_rate": 1.0308078747933618e-05, + "loss": 0.8148, + "step": 45266 + }, + { + "epoch": 2.91, + "learning_rate": 1.0307864058910669e-05, + "loss": 0.9223, + "step": 45267 + }, + { + "epoch": 2.91, + "learning_rate": 1.0307649369887718e-05, + "loss": 0.8746, + "step": 45268 + }, + { + "epoch": 2.91, + "learning_rate": 1.0307434680864767e-05, + "loss": 0.8997, + "step": 45269 + }, + { + "epoch": 2.91, + "learning_rate": 1.0307219991841817e-05, + "loss": 0.8633, + "step": 45270 + }, + { + "epoch": 2.91, + "learning_rate": 1.0307005302818868e-05, + "loss": 0.9165, + "step": 45271 + }, + { + "epoch": 2.91, + "learning_rate": 1.0306790613795919e-05, + "loss": 0.9571, + "step": 45272 + }, + { + "epoch": 2.91, + "learning_rate": 1.0306575924772968e-05, + "loss": 0.8907, + "step": 45273 + }, + { + "epoch": 2.91, + "learning_rate": 1.0306361235750018e-05, + "loss": 0.8526, + "step": 45274 + }, + { + "epoch": 2.91, + "learning_rate": 1.0306146546727067e-05, + "loss": 0.8997, + "step": 45275 + }, + { + "epoch": 2.91, + "learning_rate": 1.0305931857704116e-05, + "loss": 0.896, + "step": 45276 + }, + { + "epoch": 2.91, + "learning_rate": 1.0305717168681167e-05, + "loss": 0.9208, + "step": 45277 + }, + { + "epoch": 2.91, + "learning_rate": 1.0305502479658216e-05, + "loss": 0.8381, + "step": 45278 + }, + { + "epoch": 2.91, + "learning_rate": 1.0305287790635265e-05, + "loss": 0.8871, + "step": 45279 + }, + { + "epoch": 2.91, + "learning_rate": 1.0305073101612315e-05, + "loss": 0.974, + "step": 45280 + }, + { + "epoch": 2.91, + "learning_rate": 1.0304858412589364e-05, + "loss": 0.9238, + "step": 45281 + }, + { + "epoch": 2.91, + "learning_rate": 1.0304643723566415e-05, + "loss": 0.922, + "step": 45282 + }, + { + "epoch": 2.91, + "learning_rate": 1.0304429034543464e-05, + "loss": 0.9755, + "step": 45283 + }, + { + "epoch": 2.91, + "learning_rate": 1.0304214345520513e-05, + "loss": 0.9143, + "step": 45284 + }, + { + "epoch": 2.91, + "learning_rate": 1.0303999656497564e-05, + "loss": 0.9141, + "step": 45285 + }, + { + "epoch": 2.91, + "learning_rate": 1.0303784967474614e-05, + "loss": 0.944, + "step": 45286 + }, + { + "epoch": 2.91, + "learning_rate": 1.0303570278451665e-05, + "loss": 0.9268, + "step": 45287 + }, + { + "epoch": 2.91, + "learning_rate": 1.0303355589428714e-05, + "loss": 0.9065, + "step": 45288 + }, + { + "epoch": 2.91, + "learning_rate": 1.0303140900405763e-05, + "loss": 0.9666, + "step": 45289 + }, + { + "epoch": 2.91, + "learning_rate": 1.0302926211382814e-05, + "loss": 0.8994, + "step": 45290 + }, + { + "epoch": 2.91, + "learning_rate": 1.0302711522359863e-05, + "loss": 0.8818, + "step": 45291 + }, + { + "epoch": 2.91, + "learning_rate": 1.0302496833336913e-05, + "loss": 0.9574, + "step": 45292 + }, + { + "epoch": 2.91, + "learning_rate": 1.0302282144313962e-05, + "loss": 0.8483, + "step": 45293 + }, + { + "epoch": 2.91, + "learning_rate": 1.0302067455291011e-05, + "loss": 0.8881, + "step": 45294 + }, + { + "epoch": 2.91, + "learning_rate": 1.0301852766268062e-05, + "loss": 0.8831, + "step": 45295 + }, + { + "epoch": 2.91, + "learning_rate": 1.030163807724511e-05, + "loss": 0.8857, + "step": 45296 + }, + { + "epoch": 2.91, + "learning_rate": 1.030142338822216e-05, + "loss": 0.9627, + "step": 45297 + }, + { + "epoch": 2.91, + "learning_rate": 1.030120869919921e-05, + "loss": 0.8433, + "step": 45298 + }, + { + "epoch": 2.91, + "learning_rate": 1.030099401017626e-05, + "loss": 0.8994, + "step": 45299 + }, + { + "epoch": 2.91, + "learning_rate": 1.0300779321153312e-05, + "loss": 0.8795, + "step": 45300 + }, + { + "epoch": 2.91, + "learning_rate": 1.030056463213036e-05, + "loss": 0.894, + "step": 45301 + }, + { + "epoch": 2.91, + "learning_rate": 1.0300349943107411e-05, + "loss": 0.8646, + "step": 45302 + }, + { + "epoch": 2.91, + "learning_rate": 1.030013525408446e-05, + "loss": 0.9365, + "step": 45303 + }, + { + "epoch": 2.91, + "learning_rate": 1.029992056506151e-05, + "loss": 0.856, + "step": 45304 + }, + { + "epoch": 2.91, + "learning_rate": 1.029970587603856e-05, + "loss": 0.9091, + "step": 45305 + }, + { + "epoch": 2.91, + "learning_rate": 1.0299491187015609e-05, + "loss": 0.9543, + "step": 45306 + }, + { + "epoch": 2.91, + "learning_rate": 1.0299276497992658e-05, + "loss": 0.9107, + "step": 45307 + }, + { + "epoch": 2.92, + "learning_rate": 1.0299061808969708e-05, + "loss": 0.9014, + "step": 45308 + }, + { + "epoch": 2.92, + "learning_rate": 1.0298847119946757e-05, + "loss": 0.8579, + "step": 45309 + }, + { + "epoch": 2.92, + "learning_rate": 1.0298632430923808e-05, + "loss": 0.9322, + "step": 45310 + }, + { + "epoch": 2.92, + "learning_rate": 1.0298417741900857e-05, + "loss": 0.8527, + "step": 45311 + }, + { + "epoch": 2.92, + "learning_rate": 1.0298203052877906e-05, + "loss": 0.9277, + "step": 45312 + }, + { + "epoch": 2.92, + "learning_rate": 1.0297988363854957e-05, + "loss": 0.9646, + "step": 45313 + }, + { + "epoch": 2.92, + "learning_rate": 1.0297773674832006e-05, + "loss": 0.9054, + "step": 45314 + }, + { + "epoch": 2.92, + "learning_rate": 1.0297558985809058e-05, + "loss": 0.8413, + "step": 45315 + }, + { + "epoch": 2.92, + "learning_rate": 1.0297344296786107e-05, + "loss": 0.9018, + "step": 45316 + }, + { + "epoch": 2.92, + "learning_rate": 1.0297129607763156e-05, + "loss": 0.9471, + "step": 45317 + }, + { + "epoch": 2.92, + "learning_rate": 1.0296914918740207e-05, + "loss": 0.8981, + "step": 45318 + }, + { + "epoch": 2.92, + "learning_rate": 1.0296700229717256e-05, + "loss": 0.8569, + "step": 45319 + }, + { + "epoch": 2.92, + "learning_rate": 1.0296485540694306e-05, + "loss": 0.9053, + "step": 45320 + }, + { + "epoch": 2.92, + "learning_rate": 1.0296270851671355e-05, + "loss": 0.8571, + "step": 45321 + }, + { + "epoch": 2.92, + "learning_rate": 1.0296056162648404e-05, + "loss": 0.8774, + "step": 45322 + }, + { + "epoch": 2.92, + "learning_rate": 1.0295841473625455e-05, + "loss": 0.8717, + "step": 45323 + }, + { + "epoch": 2.92, + "learning_rate": 1.0295626784602504e-05, + "loss": 0.8705, + "step": 45324 + }, + { + "epoch": 2.92, + "learning_rate": 1.0295412095579553e-05, + "loss": 0.8824, + "step": 45325 + }, + { + "epoch": 2.92, + "learning_rate": 1.0295197406556603e-05, + "loss": 0.9023, + "step": 45326 + }, + { + "epoch": 2.92, + "learning_rate": 1.0294982717533652e-05, + "loss": 0.8709, + "step": 45327 + }, + { + "epoch": 2.92, + "learning_rate": 1.0294768028510701e-05, + "loss": 0.881, + "step": 45328 + }, + { + "epoch": 2.92, + "learning_rate": 1.0294553339487754e-05, + "loss": 0.9431, + "step": 45329 + }, + { + "epoch": 2.92, + "learning_rate": 1.0294338650464804e-05, + "loss": 0.9491, + "step": 45330 + }, + { + "epoch": 2.92, + "learning_rate": 1.0294123961441853e-05, + "loss": 0.9734, + "step": 45331 + }, + { + "epoch": 2.92, + "learning_rate": 1.0293909272418902e-05, + "loss": 0.9165, + "step": 45332 + }, + { + "epoch": 2.92, + "learning_rate": 1.0293694583395953e-05, + "loss": 0.9204, + "step": 45333 + }, + { + "epoch": 2.92, + "learning_rate": 1.0293479894373002e-05, + "loss": 0.8835, + "step": 45334 + }, + { + "epoch": 2.92, + "learning_rate": 1.029326520535005e-05, + "loss": 0.8353, + "step": 45335 + }, + { + "epoch": 2.92, + "learning_rate": 1.0293050516327102e-05, + "loss": 0.8512, + "step": 45336 + }, + { + "epoch": 2.92, + "learning_rate": 1.029283582730415e-05, + "loss": 0.867, + "step": 45337 + }, + { + "epoch": 2.92, + "learning_rate": 1.02926211382812e-05, + "loss": 0.857, + "step": 45338 + }, + { + "epoch": 2.92, + "learning_rate": 1.029240644925825e-05, + "loss": 0.9255, + "step": 45339 + }, + { + "epoch": 2.92, + "learning_rate": 1.0292191760235299e-05, + "loss": 0.8936, + "step": 45340 + }, + { + "epoch": 2.92, + "learning_rate": 1.029197707121235e-05, + "loss": 0.9109, + "step": 45341 + }, + { + "epoch": 2.92, + "learning_rate": 1.0291762382189399e-05, + "loss": 0.8767, + "step": 45342 + }, + { + "epoch": 2.92, + "learning_rate": 1.0291547693166448e-05, + "loss": 0.9412, + "step": 45343 + }, + { + "epoch": 2.92, + "learning_rate": 1.02913330041435e-05, + "loss": 0.9455, + "step": 45344 + }, + { + "epoch": 2.92, + "learning_rate": 1.0291118315120549e-05, + "loss": 0.9045, + "step": 45345 + }, + { + "epoch": 2.92, + "learning_rate": 1.02909036260976e-05, + "loss": 0.9111, + "step": 45346 + }, + { + "epoch": 2.92, + "learning_rate": 1.0290688937074649e-05, + "loss": 0.8592, + "step": 45347 + }, + { + "epoch": 2.92, + "learning_rate": 1.0290474248051698e-05, + "loss": 0.8497, + "step": 45348 + }, + { + "epoch": 2.92, + "learning_rate": 1.0290259559028748e-05, + "loss": 0.8801, + "step": 45349 + }, + { + "epoch": 2.92, + "learning_rate": 1.0290044870005797e-05, + "loss": 0.9403, + "step": 45350 + }, + { + "epoch": 2.92, + "learning_rate": 1.0289830180982848e-05, + "loss": 0.9215, + "step": 45351 + }, + { + "epoch": 2.92, + "learning_rate": 1.0289615491959897e-05, + "loss": 0.8633, + "step": 45352 + }, + { + "epoch": 2.92, + "learning_rate": 1.0289400802936946e-05, + "loss": 0.9946, + "step": 45353 + }, + { + "epoch": 2.92, + "learning_rate": 1.0289186113913996e-05, + "loss": 0.9027, + "step": 45354 + }, + { + "epoch": 2.92, + "learning_rate": 1.0288971424891045e-05, + "loss": 0.8944, + "step": 45355 + }, + { + "epoch": 2.92, + "learning_rate": 1.0288756735868094e-05, + "loss": 0.9594, + "step": 45356 + }, + { + "epoch": 2.92, + "learning_rate": 1.0288542046845145e-05, + "loss": 0.8503, + "step": 45357 + }, + { + "epoch": 2.92, + "learning_rate": 1.0288327357822196e-05, + "loss": 0.8533, + "step": 45358 + }, + { + "epoch": 2.92, + "learning_rate": 1.0288112668799246e-05, + "loss": 0.9334, + "step": 45359 + }, + { + "epoch": 2.92, + "learning_rate": 1.0287897979776295e-05, + "loss": 0.9306, + "step": 45360 + }, + { + "epoch": 2.92, + "learning_rate": 1.0287683290753346e-05, + "loss": 0.8988, + "step": 45361 + }, + { + "epoch": 2.92, + "learning_rate": 1.0287468601730395e-05, + "loss": 0.8487, + "step": 45362 + }, + { + "epoch": 2.92, + "learning_rate": 1.0287253912707444e-05, + "loss": 0.9741, + "step": 45363 + }, + { + "epoch": 2.92, + "learning_rate": 1.0287039223684495e-05, + "loss": 0.9502, + "step": 45364 + }, + { + "epoch": 2.92, + "learning_rate": 1.0286824534661544e-05, + "loss": 0.8818, + "step": 45365 + }, + { + "epoch": 2.92, + "learning_rate": 1.0286609845638592e-05, + "loss": 0.8792, + "step": 45366 + }, + { + "epoch": 2.92, + "learning_rate": 1.0286395156615643e-05, + "loss": 0.8675, + "step": 45367 + }, + { + "epoch": 2.92, + "learning_rate": 1.0286180467592692e-05, + "loss": 0.8575, + "step": 45368 + }, + { + "epoch": 2.92, + "learning_rate": 1.0285965778569743e-05, + "loss": 0.942, + "step": 45369 + }, + { + "epoch": 2.92, + "learning_rate": 1.0285751089546792e-05, + "loss": 0.9065, + "step": 45370 + }, + { + "epoch": 2.92, + "learning_rate": 1.028553640052384e-05, + "loss": 0.9566, + "step": 45371 + }, + { + "epoch": 2.92, + "learning_rate": 1.0285321711500891e-05, + "loss": 0.9198, + "step": 45372 + }, + { + "epoch": 2.92, + "learning_rate": 1.0285107022477942e-05, + "loss": 0.8524, + "step": 45373 + }, + { + "epoch": 2.92, + "learning_rate": 1.0284892333454993e-05, + "loss": 0.9183, + "step": 45374 + }, + { + "epoch": 2.92, + "learning_rate": 1.0284677644432042e-05, + "loss": 0.8716, + "step": 45375 + }, + { + "epoch": 2.92, + "learning_rate": 1.028446295540909e-05, + "loss": 0.9273, + "step": 45376 + }, + { + "epoch": 2.92, + "learning_rate": 1.0284248266386141e-05, + "loss": 0.8975, + "step": 45377 + }, + { + "epoch": 2.92, + "learning_rate": 1.028403357736319e-05, + "loss": 0.9181, + "step": 45378 + }, + { + "epoch": 2.92, + "learning_rate": 1.0283818888340241e-05, + "loss": 0.9175, + "step": 45379 + }, + { + "epoch": 2.92, + "learning_rate": 1.028360419931729e-05, + "loss": 0.9237, + "step": 45380 + }, + { + "epoch": 2.92, + "learning_rate": 1.0283389510294339e-05, + "loss": 0.9133, + "step": 45381 + }, + { + "epoch": 2.92, + "learning_rate": 1.028317482127139e-05, + "loss": 0.8816, + "step": 45382 + }, + { + "epoch": 2.92, + "learning_rate": 1.0282960132248438e-05, + "loss": 0.9017, + "step": 45383 + }, + { + "epoch": 2.92, + "learning_rate": 1.0282745443225487e-05, + "loss": 0.9069, + "step": 45384 + }, + { + "epoch": 2.92, + "learning_rate": 1.0282530754202538e-05, + "loss": 0.9423, + "step": 45385 + }, + { + "epoch": 2.92, + "learning_rate": 1.0282316065179587e-05, + "loss": 0.8675, + "step": 45386 + }, + { + "epoch": 2.92, + "learning_rate": 1.028210137615664e-05, + "loss": 0.886, + "step": 45387 + }, + { + "epoch": 2.92, + "learning_rate": 1.0281886687133688e-05, + "loss": 0.903, + "step": 45388 + }, + { + "epoch": 2.92, + "learning_rate": 1.0281671998110739e-05, + "loss": 0.9174, + "step": 45389 + }, + { + "epoch": 2.92, + "learning_rate": 1.0281457309087788e-05, + "loss": 0.9541, + "step": 45390 + }, + { + "epoch": 2.92, + "learning_rate": 1.0281242620064837e-05, + "loss": 0.8558, + "step": 45391 + }, + { + "epoch": 2.92, + "learning_rate": 1.0281027931041888e-05, + "loss": 0.8243, + "step": 45392 + }, + { + "epoch": 2.92, + "learning_rate": 1.0280813242018937e-05, + "loss": 0.9111, + "step": 45393 + }, + { + "epoch": 2.92, + "learning_rate": 1.0280598552995986e-05, + "loss": 0.8662, + "step": 45394 + }, + { + "epoch": 2.92, + "learning_rate": 1.0280383863973036e-05, + "loss": 0.9031, + "step": 45395 + }, + { + "epoch": 2.92, + "learning_rate": 1.0280169174950085e-05, + "loss": 0.8143, + "step": 45396 + }, + { + "epoch": 2.92, + "learning_rate": 1.0279954485927136e-05, + "loss": 0.9025, + "step": 45397 + }, + { + "epoch": 2.92, + "learning_rate": 1.0279739796904185e-05, + "loss": 0.8595, + "step": 45398 + }, + { + "epoch": 2.92, + "learning_rate": 1.0279525107881234e-05, + "loss": 0.8677, + "step": 45399 + }, + { + "epoch": 2.92, + "learning_rate": 1.0279310418858284e-05, + "loss": 0.9474, + "step": 45400 + }, + { + "epoch": 2.92, + "learning_rate": 1.0279095729835333e-05, + "loss": 0.8155, + "step": 45401 + }, + { + "epoch": 2.92, + "learning_rate": 1.0278881040812386e-05, + "loss": 0.9467, + "step": 45402 + }, + { + "epoch": 2.92, + "learning_rate": 1.0278666351789435e-05, + "loss": 0.8825, + "step": 45403 + }, + { + "epoch": 2.92, + "learning_rate": 1.0278451662766484e-05, + "loss": 0.9472, + "step": 45404 + }, + { + "epoch": 2.92, + "learning_rate": 1.0278236973743534e-05, + "loss": 0.91, + "step": 45405 + }, + { + "epoch": 2.92, + "learning_rate": 1.0278022284720583e-05, + "loss": 0.8445, + "step": 45406 + }, + { + "epoch": 2.92, + "learning_rate": 1.0277807595697634e-05, + "loss": 0.9638, + "step": 45407 + }, + { + "epoch": 2.92, + "learning_rate": 1.0277592906674683e-05, + "loss": 0.8472, + "step": 45408 + }, + { + "epoch": 2.92, + "learning_rate": 1.0277378217651732e-05, + "loss": 0.9649, + "step": 45409 + }, + { + "epoch": 2.92, + "learning_rate": 1.0277163528628783e-05, + "loss": 0.9713, + "step": 45410 + }, + { + "epoch": 2.92, + "learning_rate": 1.0276948839605831e-05, + "loss": 0.888, + "step": 45411 + }, + { + "epoch": 2.92, + "learning_rate": 1.027673415058288e-05, + "loss": 0.8495, + "step": 45412 + }, + { + "epoch": 2.92, + "learning_rate": 1.0276519461559931e-05, + "loss": 0.8093, + "step": 45413 + }, + { + "epoch": 2.92, + "learning_rate": 1.027630477253698e-05, + "loss": 0.9189, + "step": 45414 + }, + { + "epoch": 2.92, + "learning_rate": 1.0276090083514029e-05, + "loss": 0.9316, + "step": 45415 + }, + { + "epoch": 2.92, + "learning_rate": 1.027587539449108e-05, + "loss": 0.9251, + "step": 45416 + }, + { + "epoch": 2.92, + "learning_rate": 1.0275660705468132e-05, + "loss": 0.8713, + "step": 45417 + }, + { + "epoch": 2.92, + "learning_rate": 1.0275446016445181e-05, + "loss": 0.9225, + "step": 45418 + }, + { + "epoch": 2.92, + "learning_rate": 1.027523132742223e-05, + "loss": 0.8262, + "step": 45419 + }, + { + "epoch": 2.92, + "learning_rate": 1.027501663839928e-05, + "loss": 0.9063, + "step": 45420 + }, + { + "epoch": 2.92, + "learning_rate": 1.027480194937633e-05, + "loss": 0.9014, + "step": 45421 + }, + { + "epoch": 2.92, + "learning_rate": 1.0274587260353379e-05, + "loss": 0.9119, + "step": 45422 + }, + { + "epoch": 2.92, + "learning_rate": 1.027437257133043e-05, + "loss": 0.8991, + "step": 45423 + }, + { + "epoch": 2.92, + "learning_rate": 1.0274157882307478e-05, + "loss": 0.9466, + "step": 45424 + }, + { + "epoch": 2.92, + "learning_rate": 1.0273943193284527e-05, + "loss": 0.8841, + "step": 45425 + }, + { + "epoch": 2.92, + "learning_rate": 1.0273728504261578e-05, + "loss": 0.8788, + "step": 45426 + }, + { + "epoch": 2.92, + "learning_rate": 1.0273513815238627e-05, + "loss": 0.9599, + "step": 45427 + }, + { + "epoch": 2.92, + "learning_rate": 1.0273299126215677e-05, + "loss": 0.8913, + "step": 45428 + }, + { + "epoch": 2.92, + "learning_rate": 1.0273084437192726e-05, + "loss": 0.9202, + "step": 45429 + }, + { + "epoch": 2.92, + "learning_rate": 1.0272869748169775e-05, + "loss": 0.9159, + "step": 45430 + }, + { + "epoch": 2.92, + "learning_rate": 1.0272655059146828e-05, + "loss": 0.9247, + "step": 45431 + }, + { + "epoch": 2.92, + "learning_rate": 1.0272440370123877e-05, + "loss": 0.9061, + "step": 45432 + }, + { + "epoch": 2.92, + "learning_rate": 1.0272225681100927e-05, + "loss": 0.8309, + "step": 45433 + }, + { + "epoch": 2.92, + "learning_rate": 1.0272010992077976e-05, + "loss": 0.8865, + "step": 45434 + }, + { + "epoch": 2.92, + "learning_rate": 1.0271796303055025e-05, + "loss": 0.8769, + "step": 45435 + }, + { + "epoch": 2.92, + "learning_rate": 1.0271581614032076e-05, + "loss": 0.8683, + "step": 45436 + }, + { + "epoch": 2.92, + "learning_rate": 1.0271366925009125e-05, + "loss": 0.8323, + "step": 45437 + }, + { + "epoch": 2.92, + "learning_rate": 1.0271152235986176e-05, + "loss": 0.8589, + "step": 45438 + }, + { + "epoch": 2.92, + "learning_rate": 1.0270937546963225e-05, + "loss": 0.9346, + "step": 45439 + }, + { + "epoch": 2.92, + "learning_rate": 1.0270722857940273e-05, + "loss": 0.9398, + "step": 45440 + }, + { + "epoch": 2.92, + "learning_rate": 1.0270508168917324e-05, + "loss": 0.9076, + "step": 45441 + }, + { + "epoch": 2.92, + "learning_rate": 1.0270293479894373e-05, + "loss": 0.9315, + "step": 45442 + }, + { + "epoch": 2.92, + "learning_rate": 1.0270078790871422e-05, + "loss": 0.87, + "step": 45443 + }, + { + "epoch": 2.92, + "learning_rate": 1.0269864101848473e-05, + "loss": 0.8498, + "step": 45444 + }, + { + "epoch": 2.92, + "learning_rate": 1.0269649412825522e-05, + "loss": 0.9079, + "step": 45445 + }, + { + "epoch": 2.92, + "learning_rate": 1.0269434723802574e-05, + "loss": 0.885, + "step": 45446 + }, + { + "epoch": 2.92, + "learning_rate": 1.0269220034779623e-05, + "loss": 0.8878, + "step": 45447 + }, + { + "epoch": 2.92, + "learning_rate": 1.0269005345756674e-05, + "loss": 0.8846, + "step": 45448 + }, + { + "epoch": 2.92, + "learning_rate": 1.0268790656733723e-05, + "loss": 0.8869, + "step": 45449 + }, + { + "epoch": 2.92, + "learning_rate": 1.0268575967710772e-05, + "loss": 0.8728, + "step": 45450 + }, + { + "epoch": 2.92, + "learning_rate": 1.0268361278687822e-05, + "loss": 0.9315, + "step": 45451 + }, + { + "epoch": 2.92, + "learning_rate": 1.0268146589664871e-05, + "loss": 0.9456, + "step": 45452 + }, + { + "epoch": 2.92, + "learning_rate": 1.026793190064192e-05, + "loss": 0.8699, + "step": 45453 + }, + { + "epoch": 2.92, + "learning_rate": 1.026771721161897e-05, + "loss": 0.864, + "step": 45454 + }, + { + "epoch": 2.92, + "learning_rate": 1.026750252259602e-05, + "loss": 0.8881, + "step": 45455 + }, + { + "epoch": 2.92, + "learning_rate": 1.026728783357307e-05, + "loss": 0.8853, + "step": 45456 + }, + { + "epoch": 2.92, + "learning_rate": 1.026707314455012e-05, + "loss": 0.8952, + "step": 45457 + }, + { + "epoch": 2.92, + "learning_rate": 1.0266858455527168e-05, + "loss": 0.8561, + "step": 45458 + }, + { + "epoch": 2.92, + "learning_rate": 1.0266643766504219e-05, + "loss": 0.9254, + "step": 45459 + }, + { + "epoch": 2.92, + "learning_rate": 1.026642907748127e-05, + "loss": 0.9345, + "step": 45460 + }, + { + "epoch": 2.92, + "learning_rate": 1.026621438845832e-05, + "loss": 0.9034, + "step": 45461 + }, + { + "epoch": 2.92, + "learning_rate": 1.026599969943537e-05, + "loss": 0.9543, + "step": 45462 + }, + { + "epoch": 2.92, + "learning_rate": 1.0265785010412418e-05, + "loss": 0.8896, + "step": 45463 + }, + { + "epoch": 2.93, + "learning_rate": 1.0265570321389469e-05, + "loss": 0.8927, + "step": 45464 + }, + { + "epoch": 2.93, + "learning_rate": 1.0265355632366518e-05, + "loss": 0.9395, + "step": 45465 + }, + { + "epoch": 2.93, + "learning_rate": 1.0265140943343569e-05, + "loss": 0.9319, + "step": 45466 + }, + { + "epoch": 2.93, + "learning_rate": 1.0264926254320618e-05, + "loss": 0.9211, + "step": 45467 + }, + { + "epoch": 2.93, + "learning_rate": 1.0264711565297667e-05, + "loss": 0.8814, + "step": 45468 + }, + { + "epoch": 2.93, + "learning_rate": 1.0264496876274717e-05, + "loss": 0.8555, + "step": 45469 + }, + { + "epoch": 2.93, + "learning_rate": 1.0264282187251766e-05, + "loss": 0.9299, + "step": 45470 + }, + { + "epoch": 2.93, + "learning_rate": 1.0264067498228815e-05, + "loss": 0.8809, + "step": 45471 + }, + { + "epoch": 2.93, + "learning_rate": 1.0263852809205866e-05, + "loss": 0.8489, + "step": 45472 + }, + { + "epoch": 2.93, + "learning_rate": 1.0263638120182915e-05, + "loss": 0.9092, + "step": 45473 + }, + { + "epoch": 2.93, + "learning_rate": 1.0263423431159964e-05, + "loss": 0.8919, + "step": 45474 + }, + { + "epoch": 2.93, + "learning_rate": 1.0263208742137016e-05, + "loss": 0.8849, + "step": 45475 + }, + { + "epoch": 2.93, + "learning_rate": 1.0262994053114067e-05, + "loss": 0.888, + "step": 45476 + }, + { + "epoch": 2.93, + "learning_rate": 1.0262779364091116e-05, + "loss": 0.8557, + "step": 45477 + }, + { + "epoch": 2.93, + "learning_rate": 1.0262564675068165e-05, + "loss": 0.8751, + "step": 45478 + }, + { + "epoch": 2.93, + "learning_rate": 1.0262349986045215e-05, + "loss": 0.8385, + "step": 45479 + }, + { + "epoch": 2.93, + "learning_rate": 1.0262135297022264e-05, + "loss": 0.8871, + "step": 45480 + }, + { + "epoch": 2.93, + "learning_rate": 1.0261920607999313e-05, + "loss": 0.8779, + "step": 45481 + }, + { + "epoch": 2.93, + "learning_rate": 1.0261705918976364e-05, + "loss": 0.8837, + "step": 45482 + }, + { + "epoch": 2.93, + "learning_rate": 1.0261491229953413e-05, + "loss": 0.8458, + "step": 45483 + }, + { + "epoch": 2.93, + "learning_rate": 1.0261276540930462e-05, + "loss": 0.865, + "step": 45484 + }, + { + "epoch": 2.93, + "learning_rate": 1.0261061851907512e-05, + "loss": 0.882, + "step": 45485 + }, + { + "epoch": 2.93, + "learning_rate": 1.0260847162884561e-05, + "loss": 0.9243, + "step": 45486 + }, + { + "epoch": 2.93, + "learning_rate": 1.0260632473861612e-05, + "loss": 0.9676, + "step": 45487 + }, + { + "epoch": 2.93, + "learning_rate": 1.0260417784838661e-05, + "loss": 0.8819, + "step": 45488 + }, + { + "epoch": 2.93, + "learning_rate": 1.0260203095815713e-05, + "loss": 0.8897, + "step": 45489 + }, + { + "epoch": 2.93, + "learning_rate": 1.0259988406792762e-05, + "loss": 0.9908, + "step": 45490 + }, + { + "epoch": 2.93, + "learning_rate": 1.0259773717769811e-05, + "loss": 0.9573, + "step": 45491 + }, + { + "epoch": 2.93, + "learning_rate": 1.0259559028746862e-05, + "loss": 0.8672, + "step": 45492 + }, + { + "epoch": 2.93, + "learning_rate": 1.0259344339723911e-05, + "loss": 0.8593, + "step": 45493 + }, + { + "epoch": 2.93, + "learning_rate": 1.0259129650700962e-05, + "loss": 0.8591, + "step": 45494 + }, + { + "epoch": 2.93, + "learning_rate": 1.025891496167801e-05, + "loss": 0.8785, + "step": 45495 + }, + { + "epoch": 2.93, + "learning_rate": 1.025870027265506e-05, + "loss": 0.9073, + "step": 45496 + }, + { + "epoch": 2.93, + "learning_rate": 1.025848558363211e-05, + "loss": 0.8924, + "step": 45497 + }, + { + "epoch": 2.93, + "learning_rate": 1.025827089460916e-05, + "loss": 0.8476, + "step": 45498 + }, + { + "epoch": 2.93, + "learning_rate": 1.0258056205586208e-05, + "loss": 0.9051, + "step": 45499 + }, + { + "epoch": 2.93, + "learning_rate": 1.0257841516563259e-05, + "loss": 0.8629, + "step": 45500 + }, + { + "epoch": 2.93, + "learning_rate": 1.0257626827540308e-05, + "loss": 0.8875, + "step": 45501 + }, + { + "epoch": 2.93, + "learning_rate": 1.0257412138517357e-05, + "loss": 0.9635, + "step": 45502 + }, + { + "epoch": 2.93, + "learning_rate": 1.0257197449494407e-05, + "loss": 0.865, + "step": 45503 + }, + { + "epoch": 2.93, + "learning_rate": 1.025698276047146e-05, + "loss": 0.9303, + "step": 45504 + }, + { + "epoch": 2.93, + "learning_rate": 1.0256768071448509e-05, + "loss": 0.925, + "step": 45505 + }, + { + "epoch": 2.93, + "learning_rate": 1.0256553382425558e-05, + "loss": 0.8619, + "step": 45506 + }, + { + "epoch": 2.93, + "learning_rate": 1.0256338693402608e-05, + "loss": 0.8805, + "step": 45507 + }, + { + "epoch": 2.93, + "learning_rate": 1.0256124004379657e-05, + "loss": 0.9223, + "step": 45508 + }, + { + "epoch": 2.93, + "learning_rate": 1.0255909315356706e-05, + "loss": 0.8865, + "step": 45509 + }, + { + "epoch": 2.93, + "learning_rate": 1.0255694626333757e-05, + "loss": 0.8773, + "step": 45510 + }, + { + "epoch": 2.93, + "learning_rate": 1.0255479937310806e-05, + "loss": 0.9886, + "step": 45511 + }, + { + "epoch": 2.93, + "learning_rate": 1.0255265248287855e-05, + "loss": 0.9255, + "step": 45512 + }, + { + "epoch": 2.93, + "learning_rate": 1.0255050559264906e-05, + "loss": 0.8722, + "step": 45513 + }, + { + "epoch": 2.93, + "learning_rate": 1.0254835870241954e-05, + "loss": 0.8793, + "step": 45514 + }, + { + "epoch": 2.93, + "learning_rate": 1.0254621181219005e-05, + "loss": 0.8639, + "step": 45515 + }, + { + "epoch": 2.93, + "learning_rate": 1.0254406492196054e-05, + "loss": 0.9889, + "step": 45516 + }, + { + "epoch": 2.93, + "learning_rate": 1.0254191803173103e-05, + "loss": 0.9285, + "step": 45517 + }, + { + "epoch": 2.93, + "learning_rate": 1.0253977114150155e-05, + "loss": 0.9503, + "step": 45518 + }, + { + "epoch": 2.93, + "learning_rate": 1.0253762425127204e-05, + "loss": 0.8884, + "step": 45519 + }, + { + "epoch": 2.93, + "learning_rate": 1.0253547736104255e-05, + "loss": 0.9956, + "step": 45520 + }, + { + "epoch": 2.93, + "learning_rate": 1.0253333047081304e-05, + "loss": 0.9237, + "step": 45521 + }, + { + "epoch": 2.93, + "learning_rate": 1.0253118358058353e-05, + "loss": 0.8029, + "step": 45522 + }, + { + "epoch": 2.93, + "learning_rate": 1.0252903669035404e-05, + "loss": 0.8741, + "step": 45523 + }, + { + "epoch": 2.93, + "learning_rate": 1.0252688980012453e-05, + "loss": 0.9053, + "step": 45524 + }, + { + "epoch": 2.93, + "learning_rate": 1.0252474290989503e-05, + "loss": 0.8765, + "step": 45525 + }, + { + "epoch": 2.93, + "learning_rate": 1.0252259601966552e-05, + "loss": 1.0093, + "step": 45526 + }, + { + "epoch": 2.93, + "learning_rate": 1.0252044912943601e-05, + "loss": 0.8839, + "step": 45527 + }, + { + "epoch": 2.93, + "learning_rate": 1.0251830223920652e-05, + "loss": 0.8429, + "step": 45528 + }, + { + "epoch": 2.93, + "learning_rate": 1.02516155348977e-05, + "loss": 0.8328, + "step": 45529 + }, + { + "epoch": 2.93, + "learning_rate": 1.025140084587475e-05, + "loss": 0.9289, + "step": 45530 + }, + { + "epoch": 2.93, + "learning_rate": 1.02511861568518e-05, + "loss": 0.9135, + "step": 45531 + }, + { + "epoch": 2.93, + "learning_rate": 1.025097146782885e-05, + "loss": 0.9243, + "step": 45532 + }, + { + "epoch": 2.93, + "learning_rate": 1.0250756778805902e-05, + "loss": 0.8675, + "step": 45533 + }, + { + "epoch": 2.93, + "learning_rate": 1.025054208978295e-05, + "loss": 0.9609, + "step": 45534 + }, + { + "epoch": 2.93, + "learning_rate": 1.0250327400760001e-05, + "loss": 0.9219, + "step": 45535 + }, + { + "epoch": 2.93, + "learning_rate": 1.025011271173705e-05, + "loss": 0.8484, + "step": 45536 + }, + { + "epoch": 2.93, + "learning_rate": 1.02498980227141e-05, + "loss": 0.8701, + "step": 45537 + }, + { + "epoch": 2.93, + "learning_rate": 1.024968333369115e-05, + "loss": 0.8504, + "step": 45538 + }, + { + "epoch": 2.93, + "learning_rate": 1.0249468644668199e-05, + "loss": 0.9309, + "step": 45539 + }, + { + "epoch": 2.93, + "learning_rate": 1.0249253955645248e-05, + "loss": 0.8774, + "step": 45540 + }, + { + "epoch": 2.93, + "learning_rate": 1.0249039266622299e-05, + "loss": 0.9607, + "step": 45541 + }, + { + "epoch": 2.93, + "learning_rate": 1.0248824577599348e-05, + "loss": 0.9232, + "step": 45542 + }, + { + "epoch": 2.93, + "learning_rate": 1.0248609888576398e-05, + "loss": 0.9277, + "step": 45543 + }, + { + "epoch": 2.93, + "learning_rate": 1.0248395199553447e-05, + "loss": 0.8836, + "step": 45544 + }, + { + "epoch": 2.93, + "learning_rate": 1.0248180510530496e-05, + "loss": 0.8967, + "step": 45545 + }, + { + "epoch": 2.93, + "learning_rate": 1.0247965821507547e-05, + "loss": 0.876, + "step": 45546 + }, + { + "epoch": 2.93, + "learning_rate": 1.0247751132484597e-05, + "loss": 0.884, + "step": 45547 + }, + { + "epoch": 2.93, + "learning_rate": 1.0247536443461648e-05, + "loss": 0.9237, + "step": 45548 + }, + { + "epoch": 2.93, + "learning_rate": 1.0247321754438697e-05, + "loss": 0.8667, + "step": 45549 + }, + { + "epoch": 2.93, + "learning_rate": 1.0247107065415746e-05, + "loss": 0.8632, + "step": 45550 + }, + { + "epoch": 2.93, + "learning_rate": 1.0246892376392797e-05, + "loss": 0.9274, + "step": 45551 + }, + { + "epoch": 2.93, + "learning_rate": 1.0246677687369846e-05, + "loss": 0.8341, + "step": 45552 + }, + { + "epoch": 2.93, + "learning_rate": 1.0246462998346896e-05, + "loss": 0.8354, + "step": 45553 + }, + { + "epoch": 2.93, + "learning_rate": 1.0246248309323945e-05, + "loss": 0.9008, + "step": 45554 + }, + { + "epoch": 2.93, + "learning_rate": 1.0246033620300994e-05, + "loss": 0.842, + "step": 45555 + }, + { + "epoch": 2.93, + "learning_rate": 1.0245818931278045e-05, + "loss": 0.8759, + "step": 45556 + }, + { + "epoch": 2.93, + "learning_rate": 1.0245604242255094e-05, + "loss": 0.8692, + "step": 45557 + }, + { + "epoch": 2.93, + "learning_rate": 1.0245389553232143e-05, + "loss": 0.8549, + "step": 45558 + }, + { + "epoch": 2.93, + "learning_rate": 1.0245174864209193e-05, + "loss": 0.8722, + "step": 45559 + }, + { + "epoch": 2.93, + "learning_rate": 1.0244960175186242e-05, + "loss": 0.9203, + "step": 45560 + }, + { + "epoch": 2.93, + "learning_rate": 1.0244745486163291e-05, + "loss": 0.8984, + "step": 45561 + }, + { + "epoch": 2.93, + "learning_rate": 1.0244530797140344e-05, + "loss": 0.8904, + "step": 45562 + }, + { + "epoch": 2.93, + "learning_rate": 1.0244316108117394e-05, + "loss": 0.8291, + "step": 45563 + }, + { + "epoch": 2.93, + "learning_rate": 1.0244101419094443e-05, + "loss": 0.9288, + "step": 45564 + }, + { + "epoch": 2.93, + "learning_rate": 1.0243886730071492e-05, + "loss": 0.8553, + "step": 45565 + }, + { + "epoch": 2.93, + "learning_rate": 1.0243672041048543e-05, + "loss": 0.9742, + "step": 45566 + }, + { + "epoch": 2.93, + "learning_rate": 1.0243457352025592e-05, + "loss": 0.9892, + "step": 45567 + }, + { + "epoch": 2.93, + "learning_rate": 1.0243242663002641e-05, + "loss": 0.8772, + "step": 45568 + }, + { + "epoch": 2.93, + "learning_rate": 1.0243027973979692e-05, + "loss": 0.8562, + "step": 45569 + }, + { + "epoch": 2.93, + "learning_rate": 1.024281328495674e-05, + "loss": 0.8025, + "step": 45570 + }, + { + "epoch": 2.93, + "learning_rate": 1.024259859593379e-05, + "loss": 0.9314, + "step": 45571 + }, + { + "epoch": 2.93, + "learning_rate": 1.024238390691084e-05, + "loss": 0.8735, + "step": 45572 + }, + { + "epoch": 2.93, + "learning_rate": 1.0242169217887889e-05, + "loss": 0.8812, + "step": 45573 + }, + { + "epoch": 2.93, + "learning_rate": 1.024195452886494e-05, + "loss": 0.8527, + "step": 45574 + }, + { + "epoch": 2.93, + "learning_rate": 1.0241739839841989e-05, + "loss": 0.8862, + "step": 45575 + }, + { + "epoch": 2.93, + "learning_rate": 1.0241525150819041e-05, + "loss": 0.9324, + "step": 45576 + }, + { + "epoch": 2.93, + "learning_rate": 1.024131046179609e-05, + "loss": 0.8893, + "step": 45577 + }, + { + "epoch": 2.93, + "learning_rate": 1.0241095772773139e-05, + "loss": 0.9089, + "step": 45578 + }, + { + "epoch": 2.93, + "learning_rate": 1.024088108375019e-05, + "loss": 0.9669, + "step": 45579 + }, + { + "epoch": 2.93, + "learning_rate": 1.0240666394727239e-05, + "loss": 0.9016, + "step": 45580 + }, + { + "epoch": 2.93, + "learning_rate": 1.0240451705704288e-05, + "loss": 0.8959, + "step": 45581 + }, + { + "epoch": 2.93, + "learning_rate": 1.0240237016681338e-05, + "loss": 0.9376, + "step": 45582 + }, + { + "epoch": 2.93, + "learning_rate": 1.0240022327658387e-05, + "loss": 0.8843, + "step": 45583 + }, + { + "epoch": 2.93, + "learning_rate": 1.0239807638635438e-05, + "loss": 0.9319, + "step": 45584 + }, + { + "epoch": 2.93, + "learning_rate": 1.0239592949612487e-05, + "loss": 0.849, + "step": 45585 + }, + { + "epoch": 2.93, + "learning_rate": 1.0239378260589536e-05, + "loss": 0.8632, + "step": 45586 + }, + { + "epoch": 2.93, + "learning_rate": 1.0239163571566587e-05, + "loss": 0.9304, + "step": 45587 + }, + { + "epoch": 2.93, + "learning_rate": 1.0238948882543635e-05, + "loss": 0.89, + "step": 45588 + }, + { + "epoch": 2.93, + "learning_rate": 1.0238734193520684e-05, + "loss": 0.8808, + "step": 45589 + }, + { + "epoch": 2.93, + "learning_rate": 1.0238519504497735e-05, + "loss": 0.9078, + "step": 45590 + }, + { + "epoch": 2.93, + "learning_rate": 1.0238304815474786e-05, + "loss": 0.8827, + "step": 45591 + }, + { + "epoch": 2.93, + "learning_rate": 1.0238090126451836e-05, + "loss": 0.8775, + "step": 45592 + }, + { + "epoch": 2.93, + "learning_rate": 1.0237875437428885e-05, + "loss": 0.8785, + "step": 45593 + }, + { + "epoch": 2.93, + "learning_rate": 1.0237660748405936e-05, + "loss": 0.8809, + "step": 45594 + }, + { + "epoch": 2.93, + "learning_rate": 1.0237446059382985e-05, + "loss": 0.8675, + "step": 45595 + }, + { + "epoch": 2.93, + "learning_rate": 1.0237231370360034e-05, + "loss": 0.8708, + "step": 45596 + }, + { + "epoch": 2.93, + "learning_rate": 1.0237016681337085e-05, + "loss": 0.9086, + "step": 45597 + }, + { + "epoch": 2.93, + "learning_rate": 1.0236801992314134e-05, + "loss": 0.9584, + "step": 45598 + }, + { + "epoch": 2.93, + "learning_rate": 1.0236587303291183e-05, + "loss": 0.8789, + "step": 45599 + }, + { + "epoch": 2.93, + "learning_rate": 1.0236372614268233e-05, + "loss": 0.8751, + "step": 45600 + }, + { + "epoch": 2.93, + "learning_rate": 1.0236157925245282e-05, + "loss": 0.9761, + "step": 45601 + }, + { + "epoch": 2.93, + "learning_rate": 1.0235943236222333e-05, + "loss": 0.9669, + "step": 45602 + }, + { + "epoch": 2.93, + "learning_rate": 1.0235728547199382e-05, + "loss": 0.9362, + "step": 45603 + }, + { + "epoch": 2.93, + "learning_rate": 1.023551385817643e-05, + "loss": 0.818, + "step": 45604 + }, + { + "epoch": 2.93, + "learning_rate": 1.0235299169153483e-05, + "loss": 0.846, + "step": 45605 + }, + { + "epoch": 2.93, + "learning_rate": 1.0235084480130532e-05, + "loss": 0.8383, + "step": 45606 + }, + { + "epoch": 2.93, + "learning_rate": 1.0234869791107583e-05, + "loss": 0.8934, + "step": 45607 + }, + { + "epoch": 2.93, + "learning_rate": 1.0234655102084632e-05, + "loss": 0.9725, + "step": 45608 + }, + { + "epoch": 2.93, + "learning_rate": 1.023444041306168e-05, + "loss": 0.7969, + "step": 45609 + }, + { + "epoch": 2.93, + "learning_rate": 1.0234225724038731e-05, + "loss": 0.9829, + "step": 45610 + }, + { + "epoch": 2.93, + "learning_rate": 1.023401103501578e-05, + "loss": 0.8279, + "step": 45611 + }, + { + "epoch": 2.93, + "learning_rate": 1.0233796345992831e-05, + "loss": 0.9361, + "step": 45612 + }, + { + "epoch": 2.93, + "learning_rate": 1.023358165696988e-05, + "loss": 0.8767, + "step": 45613 + }, + { + "epoch": 2.93, + "learning_rate": 1.0233366967946929e-05, + "loss": 0.9559, + "step": 45614 + }, + { + "epoch": 2.93, + "learning_rate": 1.023315227892398e-05, + "loss": 0.8604, + "step": 45615 + }, + { + "epoch": 2.93, + "learning_rate": 1.0232937589901029e-05, + "loss": 0.9392, + "step": 45616 + }, + { + "epoch": 2.93, + "learning_rate": 1.0232722900878077e-05, + "loss": 0.9697, + "step": 45617 + }, + { + "epoch": 2.93, + "learning_rate": 1.0232508211855128e-05, + "loss": 0.9329, + "step": 45618 + }, + { + "epoch": 2.94, + "learning_rate": 1.0232293522832177e-05, + "loss": 0.9027, + "step": 45619 + }, + { + "epoch": 2.94, + "learning_rate": 1.023207883380923e-05, + "loss": 1.0052, + "step": 45620 + }, + { + "epoch": 2.94, + "learning_rate": 1.0231864144786278e-05, + "loss": 0.9271, + "step": 45621 + }, + { + "epoch": 2.94, + "learning_rate": 1.0231649455763329e-05, + "loss": 0.9362, + "step": 45622 + }, + { + "epoch": 2.94, + "learning_rate": 1.0231434766740378e-05, + "loss": 0.9624, + "step": 45623 + }, + { + "epoch": 2.94, + "learning_rate": 1.0231220077717427e-05, + "loss": 0.9238, + "step": 45624 + }, + { + "epoch": 2.94, + "learning_rate": 1.0231005388694478e-05, + "loss": 0.9137, + "step": 45625 + }, + { + "epoch": 2.94, + "learning_rate": 1.0230790699671527e-05, + "loss": 0.8789, + "step": 45626 + }, + { + "epoch": 2.94, + "learning_rate": 1.0230576010648576e-05, + "loss": 0.8422, + "step": 45627 + }, + { + "epoch": 2.94, + "learning_rate": 1.0230361321625626e-05, + "loss": 0.8341, + "step": 45628 + }, + { + "epoch": 2.94, + "learning_rate": 1.0230146632602675e-05, + "loss": 0.8735, + "step": 45629 + }, + { + "epoch": 2.94, + "learning_rate": 1.0229931943579726e-05, + "loss": 0.8945, + "step": 45630 + }, + { + "epoch": 2.94, + "learning_rate": 1.0229717254556775e-05, + "loss": 0.9166, + "step": 45631 + }, + { + "epoch": 2.94, + "learning_rate": 1.0229502565533824e-05, + "loss": 0.857, + "step": 45632 + }, + { + "epoch": 2.94, + "learning_rate": 1.0229287876510874e-05, + "loss": 0.8903, + "step": 45633 + }, + { + "epoch": 2.94, + "learning_rate": 1.0229073187487925e-05, + "loss": 0.9186, + "step": 45634 + }, + { + "epoch": 2.94, + "learning_rate": 1.0228858498464976e-05, + "loss": 0.8758, + "step": 45635 + }, + { + "epoch": 2.94, + "learning_rate": 1.0228643809442025e-05, + "loss": 0.917, + "step": 45636 + }, + { + "epoch": 2.94, + "learning_rate": 1.0228429120419074e-05, + "loss": 0.9427, + "step": 45637 + }, + { + "epoch": 2.94, + "learning_rate": 1.0228214431396124e-05, + "loss": 0.9295, + "step": 45638 + }, + { + "epoch": 2.94, + "learning_rate": 1.0227999742373173e-05, + "loss": 0.8471, + "step": 45639 + }, + { + "epoch": 2.94, + "learning_rate": 1.0227785053350224e-05, + "loss": 0.8506, + "step": 45640 + }, + { + "epoch": 2.94, + "learning_rate": 1.0227570364327273e-05, + "loss": 0.8699, + "step": 45641 + }, + { + "epoch": 2.94, + "learning_rate": 1.0227355675304322e-05, + "loss": 0.8527, + "step": 45642 + }, + { + "epoch": 2.94, + "learning_rate": 1.0227140986281373e-05, + "loss": 0.9416, + "step": 45643 + }, + { + "epoch": 2.94, + "learning_rate": 1.0226926297258422e-05, + "loss": 0.9479, + "step": 45644 + }, + { + "epoch": 2.94, + "learning_rate": 1.022671160823547e-05, + "loss": 0.8653, + "step": 45645 + }, + { + "epoch": 2.94, + "learning_rate": 1.0226496919212521e-05, + "loss": 0.8943, + "step": 45646 + }, + { + "epoch": 2.94, + "learning_rate": 1.022628223018957e-05, + "loss": 0.8706, + "step": 45647 + }, + { + "epoch": 2.94, + "learning_rate": 1.0226067541166619e-05, + "loss": 0.9895, + "step": 45648 + }, + { + "epoch": 2.94, + "learning_rate": 1.0225852852143671e-05, + "loss": 0.8519, + "step": 45649 + }, + { + "epoch": 2.94, + "learning_rate": 1.0225638163120722e-05, + "loss": 0.931, + "step": 45650 + }, + { + "epoch": 2.94, + "learning_rate": 1.0225423474097771e-05, + "loss": 0.9593, + "step": 45651 + }, + { + "epoch": 2.94, + "learning_rate": 1.022520878507482e-05, + "loss": 0.93, + "step": 45652 + }, + { + "epoch": 2.94, + "learning_rate": 1.022499409605187e-05, + "loss": 0.8899, + "step": 45653 + }, + { + "epoch": 2.94, + "learning_rate": 1.022477940702892e-05, + "loss": 0.8958, + "step": 45654 + }, + { + "epoch": 2.94, + "learning_rate": 1.0224564718005969e-05, + "loss": 0.9192, + "step": 45655 + }, + { + "epoch": 2.94, + "learning_rate": 1.022435002898302e-05, + "loss": 0.8718, + "step": 45656 + }, + { + "epoch": 2.94, + "learning_rate": 1.0224135339960068e-05, + "loss": 0.9085, + "step": 45657 + }, + { + "epoch": 2.94, + "learning_rate": 1.0223920650937117e-05, + "loss": 0.9437, + "step": 45658 + }, + { + "epoch": 2.94, + "learning_rate": 1.0223705961914168e-05, + "loss": 0.9051, + "step": 45659 + }, + { + "epoch": 2.94, + "learning_rate": 1.0223491272891217e-05, + "loss": 0.8852, + "step": 45660 + }, + { + "epoch": 2.94, + "learning_rate": 1.0223276583868268e-05, + "loss": 0.7957, + "step": 45661 + }, + { + "epoch": 2.94, + "learning_rate": 1.0223061894845316e-05, + "loss": 0.9142, + "step": 45662 + }, + { + "epoch": 2.94, + "learning_rate": 1.0222847205822369e-05, + "loss": 0.8679, + "step": 45663 + }, + { + "epoch": 2.94, + "learning_rate": 1.0222632516799418e-05, + "loss": 0.9456, + "step": 45664 + }, + { + "epoch": 2.94, + "learning_rate": 1.0222417827776467e-05, + "loss": 0.8202, + "step": 45665 + }, + { + "epoch": 2.94, + "learning_rate": 1.0222203138753517e-05, + "loss": 0.9272, + "step": 45666 + }, + { + "epoch": 2.94, + "learning_rate": 1.0221988449730566e-05, + "loss": 0.926, + "step": 45667 + }, + { + "epoch": 2.94, + "learning_rate": 1.0221773760707615e-05, + "loss": 0.9431, + "step": 45668 + }, + { + "epoch": 2.94, + "learning_rate": 1.0221559071684666e-05, + "loss": 0.8213, + "step": 45669 + }, + { + "epoch": 2.94, + "learning_rate": 1.0221344382661715e-05, + "loss": 0.9141, + "step": 45670 + }, + { + "epoch": 2.94, + "learning_rate": 1.0221129693638766e-05, + "loss": 0.954, + "step": 45671 + }, + { + "epoch": 2.94, + "learning_rate": 1.0220915004615815e-05, + "loss": 0.9629, + "step": 45672 + }, + { + "epoch": 2.94, + "learning_rate": 1.0220700315592864e-05, + "loss": 0.8766, + "step": 45673 + }, + { + "epoch": 2.94, + "learning_rate": 1.0220485626569914e-05, + "loss": 0.9238, + "step": 45674 + }, + { + "epoch": 2.94, + "learning_rate": 1.0220270937546963e-05, + "loss": 0.8455, + "step": 45675 + }, + { + "epoch": 2.94, + "learning_rate": 1.0220056248524012e-05, + "loss": 0.9397, + "step": 45676 + }, + { + "epoch": 2.94, + "learning_rate": 1.0219841559501063e-05, + "loss": 0.8898, + "step": 45677 + }, + { + "epoch": 2.94, + "learning_rate": 1.0219626870478113e-05, + "loss": 0.9274, + "step": 45678 + }, + { + "epoch": 2.94, + "learning_rate": 1.0219412181455164e-05, + "loss": 0.8987, + "step": 45679 + }, + { + "epoch": 2.94, + "learning_rate": 1.0219197492432213e-05, + "loss": 0.8666, + "step": 45680 + }, + { + "epoch": 2.94, + "learning_rate": 1.0218982803409264e-05, + "loss": 0.8758, + "step": 45681 + }, + { + "epoch": 2.94, + "learning_rate": 1.0218768114386313e-05, + "loss": 0.8922, + "step": 45682 + }, + { + "epoch": 2.94, + "learning_rate": 1.0218553425363362e-05, + "loss": 0.9276, + "step": 45683 + }, + { + "epoch": 2.94, + "learning_rate": 1.0218338736340412e-05, + "loss": 0.9236, + "step": 45684 + }, + { + "epoch": 2.94, + "learning_rate": 1.0218124047317461e-05, + "loss": 0.8986, + "step": 45685 + }, + { + "epoch": 2.94, + "learning_rate": 1.021790935829451e-05, + "loss": 0.9163, + "step": 45686 + }, + { + "epoch": 2.94, + "learning_rate": 1.0217694669271561e-05, + "loss": 0.8756, + "step": 45687 + }, + { + "epoch": 2.94, + "learning_rate": 1.021747998024861e-05, + "loss": 0.9062, + "step": 45688 + }, + { + "epoch": 2.94, + "learning_rate": 1.021726529122566e-05, + "loss": 0.9921, + "step": 45689 + }, + { + "epoch": 2.94, + "learning_rate": 1.021705060220271e-05, + "loss": 0.8753, + "step": 45690 + }, + { + "epoch": 2.94, + "learning_rate": 1.0216835913179758e-05, + "loss": 0.8985, + "step": 45691 + }, + { + "epoch": 2.94, + "learning_rate": 1.021662122415681e-05, + "loss": 0.8787, + "step": 45692 + }, + { + "epoch": 2.94, + "learning_rate": 1.021640653513386e-05, + "loss": 0.8543, + "step": 45693 + }, + { + "epoch": 2.94, + "learning_rate": 1.021619184611091e-05, + "loss": 0.8894, + "step": 45694 + }, + { + "epoch": 2.94, + "learning_rate": 1.021597715708796e-05, + "loss": 0.8718, + "step": 45695 + }, + { + "epoch": 2.94, + "learning_rate": 1.0215762468065008e-05, + "loss": 0.8602, + "step": 45696 + }, + { + "epoch": 2.94, + "learning_rate": 1.0215547779042059e-05, + "loss": 0.917, + "step": 45697 + }, + { + "epoch": 2.94, + "learning_rate": 1.0215333090019108e-05, + "loss": 0.954, + "step": 45698 + }, + { + "epoch": 2.94, + "learning_rate": 1.0215118400996159e-05, + "loss": 0.9278, + "step": 45699 + }, + { + "epoch": 2.94, + "learning_rate": 1.0214903711973208e-05, + "loss": 0.8844, + "step": 45700 + }, + { + "epoch": 2.94, + "learning_rate": 1.0214689022950257e-05, + "loss": 0.9188, + "step": 45701 + }, + { + "epoch": 2.94, + "learning_rate": 1.0214474333927307e-05, + "loss": 0.8519, + "step": 45702 + }, + { + "epoch": 2.94, + "learning_rate": 1.0214259644904356e-05, + "loss": 0.8963, + "step": 45703 + }, + { + "epoch": 2.94, + "learning_rate": 1.0214044955881405e-05, + "loss": 0.9218, + "step": 45704 + }, + { + "epoch": 2.94, + "learning_rate": 1.0213830266858456e-05, + "loss": 0.9187, + "step": 45705 + }, + { + "epoch": 2.94, + "learning_rate": 1.0213615577835505e-05, + "loss": 0.9225, + "step": 45706 + }, + { + "epoch": 2.94, + "learning_rate": 1.0213400888812557e-05, + "loss": 0.862, + "step": 45707 + }, + { + "epoch": 2.94, + "learning_rate": 1.0213186199789606e-05, + "loss": 0.9019, + "step": 45708 + }, + { + "epoch": 2.94, + "learning_rate": 1.0212971510766657e-05, + "loss": 0.8747, + "step": 45709 + }, + { + "epoch": 2.94, + "learning_rate": 1.0212756821743706e-05, + "loss": 0.8818, + "step": 45710 + }, + { + "epoch": 2.94, + "learning_rate": 1.0212542132720755e-05, + "loss": 0.9153, + "step": 45711 + }, + { + "epoch": 2.94, + "learning_rate": 1.0212327443697805e-05, + "loss": 0.9179, + "step": 45712 + }, + { + "epoch": 2.94, + "learning_rate": 1.0212112754674854e-05, + "loss": 0.9278, + "step": 45713 + }, + { + "epoch": 2.94, + "learning_rate": 1.0211898065651903e-05, + "loss": 0.9136, + "step": 45714 + }, + { + "epoch": 2.94, + "learning_rate": 1.0211683376628954e-05, + "loss": 0.9372, + "step": 45715 + }, + { + "epoch": 2.94, + "learning_rate": 1.0211468687606003e-05, + "loss": 0.9334, + "step": 45716 + }, + { + "epoch": 2.94, + "learning_rate": 1.0211253998583054e-05, + "loss": 0.8551, + "step": 45717 + }, + { + "epoch": 2.94, + "learning_rate": 1.0211039309560103e-05, + "loss": 0.8789, + "step": 45718 + }, + { + "epoch": 2.94, + "learning_rate": 1.0210824620537152e-05, + "loss": 0.8913, + "step": 45719 + }, + { + "epoch": 2.94, + "learning_rate": 1.0210609931514202e-05, + "loss": 0.9006, + "step": 45720 + }, + { + "epoch": 2.94, + "learning_rate": 1.0210395242491253e-05, + "loss": 0.9117, + "step": 45721 + }, + { + "epoch": 2.94, + "learning_rate": 1.0210180553468304e-05, + "loss": 0.9899, + "step": 45722 + }, + { + "epoch": 2.94, + "learning_rate": 1.0209965864445352e-05, + "loss": 0.8803, + "step": 45723 + }, + { + "epoch": 2.94, + "learning_rate": 1.0209751175422401e-05, + "loss": 0.9431, + "step": 45724 + }, + { + "epoch": 2.94, + "learning_rate": 1.0209536486399452e-05, + "loss": 0.9017, + "step": 45725 + }, + { + "epoch": 2.94, + "learning_rate": 1.0209321797376501e-05, + "loss": 0.8896, + "step": 45726 + }, + { + "epoch": 2.94, + "learning_rate": 1.0209107108353552e-05, + "loss": 0.9069, + "step": 45727 + }, + { + "epoch": 2.94, + "learning_rate": 1.02088924193306e-05, + "loss": 0.8654, + "step": 45728 + }, + { + "epoch": 2.94, + "learning_rate": 1.020867773030765e-05, + "loss": 0.8703, + "step": 45729 + }, + { + "epoch": 2.94, + "learning_rate": 1.02084630412847e-05, + "loss": 0.8833, + "step": 45730 + }, + { + "epoch": 2.94, + "learning_rate": 1.020824835226175e-05, + "loss": 0.9025, + "step": 45731 + }, + { + "epoch": 2.94, + "learning_rate": 1.0208033663238798e-05, + "loss": 0.8795, + "step": 45732 + }, + { + "epoch": 2.94, + "learning_rate": 1.0207818974215849e-05, + "loss": 0.9026, + "step": 45733 + }, + { + "epoch": 2.94, + "learning_rate": 1.0207604285192898e-05, + "loss": 0.9564, + "step": 45734 + }, + { + "epoch": 2.94, + "learning_rate": 1.0207389596169947e-05, + "loss": 0.8726, + "step": 45735 + }, + { + "epoch": 2.94, + "learning_rate": 1.0207174907147e-05, + "loss": 0.8852, + "step": 45736 + }, + { + "epoch": 2.94, + "learning_rate": 1.020696021812405e-05, + "loss": 0.874, + "step": 45737 + }, + { + "epoch": 2.94, + "learning_rate": 1.0206745529101099e-05, + "loss": 0.8871, + "step": 45738 + }, + { + "epoch": 2.94, + "learning_rate": 1.0206530840078148e-05, + "loss": 0.8783, + "step": 45739 + }, + { + "epoch": 2.94, + "learning_rate": 1.0206316151055198e-05, + "loss": 1.0066, + "step": 45740 + }, + { + "epoch": 2.94, + "learning_rate": 1.0206101462032247e-05, + "loss": 0.9029, + "step": 45741 + }, + { + "epoch": 2.94, + "learning_rate": 1.0205886773009296e-05, + "loss": 0.86, + "step": 45742 + }, + { + "epoch": 2.94, + "learning_rate": 1.0205672083986347e-05, + "loss": 0.8923, + "step": 45743 + }, + { + "epoch": 2.94, + "learning_rate": 1.0205457394963396e-05, + "loss": 0.8991, + "step": 45744 + }, + { + "epoch": 2.94, + "learning_rate": 1.0205242705940445e-05, + "loss": 0.8703, + "step": 45745 + }, + { + "epoch": 2.94, + "learning_rate": 1.0205028016917496e-05, + "loss": 0.8275, + "step": 45746 + }, + { + "epoch": 2.94, + "learning_rate": 1.0204813327894545e-05, + "loss": 0.858, + "step": 45747 + }, + { + "epoch": 2.94, + "learning_rate": 1.0204598638871595e-05, + "loss": 0.9036, + "step": 45748 + }, + { + "epoch": 2.94, + "learning_rate": 1.0204383949848644e-05, + "loss": 0.893, + "step": 45749 + }, + { + "epoch": 2.94, + "learning_rate": 1.0204169260825697e-05, + "loss": 0.9668, + "step": 45750 + }, + { + "epoch": 2.94, + "learning_rate": 1.0203954571802746e-05, + "loss": 0.92, + "step": 45751 + }, + { + "epoch": 2.94, + "learning_rate": 1.0203739882779794e-05, + "loss": 0.902, + "step": 45752 + }, + { + "epoch": 2.94, + "learning_rate": 1.0203525193756845e-05, + "loss": 0.8535, + "step": 45753 + }, + { + "epoch": 2.94, + "learning_rate": 1.0203310504733894e-05, + "loss": 0.9898, + "step": 45754 + }, + { + "epoch": 2.94, + "learning_rate": 1.0203095815710943e-05, + "loss": 0.8846, + "step": 45755 + }, + { + "epoch": 2.94, + "learning_rate": 1.0202881126687994e-05, + "loss": 0.9225, + "step": 45756 + }, + { + "epoch": 2.94, + "learning_rate": 1.0202666437665043e-05, + "loss": 0.8409, + "step": 45757 + }, + { + "epoch": 2.94, + "learning_rate": 1.0202451748642093e-05, + "loss": 0.8702, + "step": 45758 + }, + { + "epoch": 2.94, + "learning_rate": 1.0202237059619142e-05, + "loss": 0.8878, + "step": 45759 + }, + { + "epoch": 2.94, + "learning_rate": 1.0202022370596191e-05, + "loss": 0.9725, + "step": 45760 + }, + { + "epoch": 2.94, + "learning_rate": 1.0201807681573242e-05, + "loss": 0.9087, + "step": 45761 + }, + { + "epoch": 2.94, + "learning_rate": 1.0201592992550291e-05, + "loss": 0.8695, + "step": 45762 + }, + { + "epoch": 2.94, + "learning_rate": 1.020137830352734e-05, + "loss": 0.8898, + "step": 45763 + }, + { + "epoch": 2.94, + "learning_rate": 1.020116361450439e-05, + "loss": 0.9249, + "step": 45764 + }, + { + "epoch": 2.94, + "learning_rate": 1.0200948925481441e-05, + "loss": 0.932, + "step": 45765 + }, + { + "epoch": 2.94, + "learning_rate": 1.0200734236458492e-05, + "loss": 0.8287, + "step": 45766 + }, + { + "epoch": 2.94, + "learning_rate": 1.020051954743554e-05, + "loss": 0.8144, + "step": 45767 + }, + { + "epoch": 2.94, + "learning_rate": 1.0200304858412591e-05, + "loss": 0.8865, + "step": 45768 + }, + { + "epoch": 2.94, + "learning_rate": 1.020009016938964e-05, + "loss": 0.8955, + "step": 45769 + }, + { + "epoch": 2.94, + "learning_rate": 1.019987548036669e-05, + "loss": 0.9652, + "step": 45770 + }, + { + "epoch": 2.94, + "learning_rate": 1.019966079134374e-05, + "loss": 0.927, + "step": 45771 + }, + { + "epoch": 2.94, + "learning_rate": 1.0199446102320789e-05, + "loss": 0.9865, + "step": 45772 + }, + { + "epoch": 2.94, + "learning_rate": 1.0199231413297838e-05, + "loss": 0.9981, + "step": 45773 + }, + { + "epoch": 2.94, + "learning_rate": 1.0199016724274889e-05, + "loss": 0.9211, + "step": 45774 + }, + { + "epoch": 2.95, + "learning_rate": 1.0198802035251938e-05, + "loss": 0.8381, + "step": 45775 + }, + { + "epoch": 2.95, + "learning_rate": 1.0198587346228988e-05, + "loss": 0.9369, + "step": 45776 + }, + { + "epoch": 2.95, + "learning_rate": 1.0198372657206037e-05, + "loss": 0.9471, + "step": 45777 + }, + { + "epoch": 2.95, + "learning_rate": 1.0198157968183086e-05, + "loss": 0.8906, + "step": 45778 + }, + { + "epoch": 2.95, + "learning_rate": 1.0197943279160137e-05, + "loss": 0.9253, + "step": 45779 + }, + { + "epoch": 2.95, + "learning_rate": 1.0197728590137188e-05, + "loss": 0.8611, + "step": 45780 + }, + { + "epoch": 2.95, + "learning_rate": 1.0197513901114238e-05, + "loss": 0.8912, + "step": 45781 + }, + { + "epoch": 2.95, + "learning_rate": 1.0197299212091287e-05, + "loss": 0.8536, + "step": 45782 + }, + { + "epoch": 2.95, + "learning_rate": 1.0197084523068336e-05, + "loss": 0.9049, + "step": 45783 + }, + { + "epoch": 2.95, + "learning_rate": 1.0196869834045387e-05, + "loss": 0.9108, + "step": 45784 + }, + { + "epoch": 2.95, + "learning_rate": 1.0196655145022436e-05, + "loss": 0.9332, + "step": 45785 + }, + { + "epoch": 2.95, + "learning_rate": 1.0196440455999486e-05, + "loss": 0.927, + "step": 45786 + }, + { + "epoch": 2.95, + "learning_rate": 1.0196225766976535e-05, + "loss": 0.9289, + "step": 45787 + }, + { + "epoch": 2.95, + "learning_rate": 1.0196011077953584e-05, + "loss": 0.9358, + "step": 45788 + }, + { + "epoch": 2.95, + "learning_rate": 1.0195796388930635e-05, + "loss": 0.8709, + "step": 45789 + }, + { + "epoch": 2.95, + "learning_rate": 1.0195581699907684e-05, + "loss": 0.8564, + "step": 45790 + }, + { + "epoch": 2.95, + "learning_rate": 1.0195367010884733e-05, + "loss": 0.8995, + "step": 45791 + }, + { + "epoch": 2.95, + "learning_rate": 1.0195152321861784e-05, + "loss": 0.9217, + "step": 45792 + }, + { + "epoch": 2.95, + "learning_rate": 1.0194937632838833e-05, + "loss": 0.8274, + "step": 45793 + }, + { + "epoch": 2.95, + "learning_rate": 1.0194722943815885e-05, + "loss": 0.9886, + "step": 45794 + }, + { + "epoch": 2.95, + "learning_rate": 1.0194508254792934e-05, + "loss": 0.8513, + "step": 45795 + }, + { + "epoch": 2.95, + "learning_rate": 1.0194293565769985e-05, + "loss": 0.8913, + "step": 45796 + }, + { + "epoch": 2.95, + "learning_rate": 1.0194078876747033e-05, + "loss": 0.8936, + "step": 45797 + }, + { + "epoch": 2.95, + "learning_rate": 1.0193864187724082e-05, + "loss": 0.9148, + "step": 45798 + }, + { + "epoch": 2.95, + "learning_rate": 1.0193649498701133e-05, + "loss": 0.8597, + "step": 45799 + }, + { + "epoch": 2.95, + "learning_rate": 1.0193434809678182e-05, + "loss": 0.9151, + "step": 45800 + }, + { + "epoch": 2.95, + "learning_rate": 1.0193220120655231e-05, + "loss": 0.8788, + "step": 45801 + }, + { + "epoch": 2.95, + "learning_rate": 1.0193005431632282e-05, + "loss": 0.9443, + "step": 45802 + }, + { + "epoch": 2.95, + "learning_rate": 1.019279074260933e-05, + "loss": 0.958, + "step": 45803 + }, + { + "epoch": 2.95, + "learning_rate": 1.0192576053586381e-05, + "loss": 0.8746, + "step": 45804 + }, + { + "epoch": 2.95, + "learning_rate": 1.019236136456343e-05, + "loss": 0.9289, + "step": 45805 + }, + { + "epoch": 2.95, + "learning_rate": 1.019214667554048e-05, + "loss": 0.9017, + "step": 45806 + }, + { + "epoch": 2.95, + "learning_rate": 1.019193198651753e-05, + "loss": 0.88, + "step": 45807 + }, + { + "epoch": 2.95, + "learning_rate": 1.0191717297494579e-05, + "loss": 0.9272, + "step": 45808 + }, + { + "epoch": 2.95, + "learning_rate": 1.0191502608471631e-05, + "loss": 0.9056, + "step": 45809 + }, + { + "epoch": 2.95, + "learning_rate": 1.019128791944868e-05, + "loss": 0.9532, + "step": 45810 + }, + { + "epoch": 2.95, + "learning_rate": 1.019107323042573e-05, + "loss": 0.8345, + "step": 45811 + }, + { + "epoch": 2.95, + "learning_rate": 1.019085854140278e-05, + "loss": 0.9052, + "step": 45812 + }, + { + "epoch": 2.95, + "learning_rate": 1.0190643852379829e-05, + "loss": 0.881, + "step": 45813 + }, + { + "epoch": 2.95, + "learning_rate": 1.019042916335688e-05, + "loss": 0.9775, + "step": 45814 + }, + { + "epoch": 2.95, + "learning_rate": 1.0190214474333928e-05, + "loss": 0.9403, + "step": 45815 + }, + { + "epoch": 2.95, + "learning_rate": 1.0189999785310977e-05, + "loss": 0.8935, + "step": 45816 + }, + { + "epoch": 2.95, + "learning_rate": 1.0189785096288028e-05, + "loss": 0.9009, + "step": 45817 + }, + { + "epoch": 2.95, + "learning_rate": 1.0189570407265077e-05, + "loss": 0.9667, + "step": 45818 + }, + { + "epoch": 2.95, + "learning_rate": 1.0189355718242126e-05, + "loss": 0.9314, + "step": 45819 + }, + { + "epoch": 2.95, + "learning_rate": 1.0189141029219177e-05, + "loss": 0.898, + "step": 45820 + }, + { + "epoch": 2.95, + "learning_rate": 1.0188926340196226e-05, + "loss": 0.851, + "step": 45821 + }, + { + "epoch": 2.95, + "learning_rate": 1.0188711651173275e-05, + "loss": 0.8705, + "step": 45822 + }, + { + "epoch": 2.95, + "learning_rate": 1.0188496962150327e-05, + "loss": 0.8605, + "step": 45823 + }, + { + "epoch": 2.95, + "learning_rate": 1.0188282273127378e-05, + "loss": 0.9316, + "step": 45824 + }, + { + "epoch": 2.95, + "learning_rate": 1.0188067584104427e-05, + "loss": 0.9448, + "step": 45825 + }, + { + "epoch": 2.95, + "learning_rate": 1.0187852895081475e-05, + "loss": 0.8608, + "step": 45826 + }, + { + "epoch": 2.95, + "learning_rate": 1.0187638206058526e-05, + "loss": 0.8961, + "step": 45827 + }, + { + "epoch": 2.95, + "learning_rate": 1.0187423517035575e-05, + "loss": 0.8858, + "step": 45828 + }, + { + "epoch": 2.95, + "learning_rate": 1.0187208828012624e-05, + "loss": 0.8399, + "step": 45829 + }, + { + "epoch": 2.95, + "learning_rate": 1.0186994138989675e-05, + "loss": 0.9948, + "step": 45830 + }, + { + "epoch": 2.95, + "learning_rate": 1.0186779449966724e-05, + "loss": 0.9451, + "step": 45831 + }, + { + "epoch": 2.95, + "learning_rate": 1.0186564760943773e-05, + "loss": 0.9329, + "step": 45832 + }, + { + "epoch": 2.95, + "learning_rate": 1.0186350071920823e-05, + "loss": 0.9475, + "step": 45833 + }, + { + "epoch": 2.95, + "learning_rate": 1.0186135382897872e-05, + "loss": 0.8699, + "step": 45834 + }, + { + "epoch": 2.95, + "learning_rate": 1.0185920693874923e-05, + "loss": 0.8564, + "step": 45835 + }, + { + "epoch": 2.95, + "learning_rate": 1.0185706004851972e-05, + "loss": 0.9249, + "step": 45836 + }, + { + "epoch": 2.95, + "learning_rate": 1.0185491315829021e-05, + "loss": 0.9099, + "step": 45837 + }, + { + "epoch": 2.95, + "learning_rate": 1.0185276626806073e-05, + "loss": 0.8962, + "step": 45838 + }, + { + "epoch": 2.95, + "learning_rate": 1.0185061937783122e-05, + "loss": 0.8756, + "step": 45839 + }, + { + "epoch": 2.95, + "learning_rate": 1.0184847248760173e-05, + "loss": 0.8611, + "step": 45840 + }, + { + "epoch": 2.95, + "learning_rate": 1.0184632559737222e-05, + "loss": 0.92, + "step": 45841 + }, + { + "epoch": 2.95, + "learning_rate": 1.018441787071427e-05, + "loss": 0.8858, + "step": 45842 + }, + { + "epoch": 2.95, + "learning_rate": 1.0184203181691321e-05, + "loss": 0.8676, + "step": 45843 + }, + { + "epoch": 2.95, + "learning_rate": 1.018398849266837e-05, + "loss": 0.9561, + "step": 45844 + }, + { + "epoch": 2.95, + "learning_rate": 1.0183773803645421e-05, + "loss": 0.964, + "step": 45845 + }, + { + "epoch": 2.95, + "learning_rate": 1.018355911462247e-05, + "loss": 0.8218, + "step": 45846 + }, + { + "epoch": 2.95, + "learning_rate": 1.0183344425599519e-05, + "loss": 0.8706, + "step": 45847 + }, + { + "epoch": 2.95, + "learning_rate": 1.018312973657657e-05, + "loss": 0.8302, + "step": 45848 + }, + { + "epoch": 2.95, + "learning_rate": 1.0182915047553619e-05, + "loss": 1.0056, + "step": 45849 + }, + { + "epoch": 2.95, + "learning_rate": 1.0182700358530668e-05, + "loss": 0.8977, + "step": 45850 + }, + { + "epoch": 2.95, + "learning_rate": 1.0182485669507718e-05, + "loss": 0.8764, + "step": 45851 + }, + { + "epoch": 2.95, + "learning_rate": 1.0182270980484769e-05, + "loss": 0.8778, + "step": 45852 + }, + { + "epoch": 2.95, + "learning_rate": 1.018205629146182e-05, + "loss": 0.9295, + "step": 45853 + }, + { + "epoch": 2.95, + "learning_rate": 1.0181841602438869e-05, + "loss": 0.8153, + "step": 45854 + }, + { + "epoch": 2.95, + "learning_rate": 1.018162691341592e-05, + "loss": 0.9424, + "step": 45855 + }, + { + "epoch": 2.95, + "learning_rate": 1.0181412224392968e-05, + "loss": 0.9081, + "step": 45856 + }, + { + "epoch": 2.95, + "learning_rate": 1.0181197535370017e-05, + "loss": 0.9605, + "step": 45857 + }, + { + "epoch": 2.95, + "learning_rate": 1.0180982846347068e-05, + "loss": 0.9467, + "step": 45858 + }, + { + "epoch": 2.95, + "learning_rate": 1.0180768157324117e-05, + "loss": 0.8685, + "step": 45859 + }, + { + "epoch": 2.95, + "learning_rate": 1.0180553468301166e-05, + "loss": 0.9221, + "step": 45860 + }, + { + "epoch": 2.95, + "learning_rate": 1.0180338779278216e-05, + "loss": 0.9195, + "step": 45861 + }, + { + "epoch": 2.95, + "learning_rate": 1.0180124090255265e-05, + "loss": 0.8526, + "step": 45862 + }, + { + "epoch": 2.95, + "learning_rate": 1.0179909401232316e-05, + "loss": 0.8909, + "step": 45863 + }, + { + "epoch": 2.95, + "learning_rate": 1.0179694712209365e-05, + "loss": 0.9274, + "step": 45864 + }, + { + "epoch": 2.95, + "learning_rate": 1.0179480023186414e-05, + "loss": 0.9001, + "step": 45865 + }, + { + "epoch": 2.95, + "learning_rate": 1.0179265334163465e-05, + "loss": 0.9292, + "step": 45866 + }, + { + "epoch": 2.95, + "learning_rate": 1.0179050645140515e-05, + "loss": 0.8877, + "step": 45867 + }, + { + "epoch": 2.95, + "learning_rate": 1.0178835956117566e-05, + "loss": 0.9647, + "step": 45868 + }, + { + "epoch": 2.95, + "learning_rate": 1.0178621267094615e-05, + "loss": 0.9297, + "step": 45869 + }, + { + "epoch": 2.95, + "learning_rate": 1.0178406578071664e-05, + "loss": 0.8991, + "step": 45870 + }, + { + "epoch": 2.95, + "learning_rate": 1.0178191889048714e-05, + "loss": 0.9243, + "step": 45871 + }, + { + "epoch": 2.95, + "learning_rate": 1.0177977200025763e-05, + "loss": 0.8452, + "step": 45872 + }, + { + "epoch": 2.95, + "learning_rate": 1.0177762511002814e-05, + "loss": 0.9009, + "step": 45873 + }, + { + "epoch": 2.95, + "learning_rate": 1.0177547821979863e-05, + "loss": 0.9194, + "step": 45874 + }, + { + "epoch": 2.95, + "learning_rate": 1.0177333132956912e-05, + "loss": 0.8778, + "step": 45875 + }, + { + "epoch": 2.95, + "learning_rate": 1.0177118443933963e-05, + "loss": 0.9437, + "step": 45876 + }, + { + "epoch": 2.95, + "learning_rate": 1.0176903754911012e-05, + "loss": 0.9321, + "step": 45877 + }, + { + "epoch": 2.95, + "learning_rate": 1.017668906588806e-05, + "loss": 0.9447, + "step": 45878 + }, + { + "epoch": 2.95, + "learning_rate": 1.0176474376865111e-05, + "loss": 0.9353, + "step": 45879 + }, + { + "epoch": 2.95, + "learning_rate": 1.017625968784216e-05, + "loss": 0.8481, + "step": 45880 + }, + { + "epoch": 2.95, + "learning_rate": 1.0176044998819213e-05, + "loss": 0.917, + "step": 45881 + }, + { + "epoch": 2.95, + "learning_rate": 1.0175830309796262e-05, + "loss": 0.9431, + "step": 45882 + }, + { + "epoch": 2.95, + "learning_rate": 1.0175615620773312e-05, + "loss": 0.9239, + "step": 45883 + }, + { + "epoch": 2.95, + "learning_rate": 1.0175400931750361e-05, + "loss": 0.9117, + "step": 45884 + }, + { + "epoch": 2.95, + "learning_rate": 1.017518624272741e-05, + "loss": 0.8857, + "step": 45885 + }, + { + "epoch": 2.95, + "learning_rate": 1.017497155370446e-05, + "loss": 0.9286, + "step": 45886 + }, + { + "epoch": 2.95, + "learning_rate": 1.017475686468151e-05, + "loss": 0.9211, + "step": 45887 + }, + { + "epoch": 2.95, + "learning_rate": 1.0174542175658559e-05, + "loss": 0.8928, + "step": 45888 + }, + { + "epoch": 2.95, + "learning_rate": 1.017432748663561e-05, + "loss": 0.9462, + "step": 45889 + }, + { + "epoch": 2.95, + "learning_rate": 1.0174112797612658e-05, + "loss": 0.9114, + "step": 45890 + }, + { + "epoch": 2.95, + "learning_rate": 1.0173898108589707e-05, + "loss": 0.9952, + "step": 45891 + }, + { + "epoch": 2.95, + "learning_rate": 1.0173683419566758e-05, + "loss": 0.9066, + "step": 45892 + }, + { + "epoch": 2.95, + "learning_rate": 1.0173468730543807e-05, + "loss": 0.9283, + "step": 45893 + }, + { + "epoch": 2.95, + "learning_rate": 1.0173254041520858e-05, + "loss": 0.8854, + "step": 45894 + }, + { + "epoch": 2.95, + "learning_rate": 1.0173039352497907e-05, + "loss": 0.8963, + "step": 45895 + }, + { + "epoch": 2.95, + "learning_rate": 1.0172824663474959e-05, + "loss": 0.8226, + "step": 45896 + }, + { + "epoch": 2.95, + "learning_rate": 1.0172609974452008e-05, + "loss": 0.9022, + "step": 45897 + }, + { + "epoch": 2.95, + "learning_rate": 1.0172395285429057e-05, + "loss": 0.863, + "step": 45898 + }, + { + "epoch": 2.95, + "learning_rate": 1.0172180596406108e-05, + "loss": 0.8972, + "step": 45899 + }, + { + "epoch": 2.95, + "learning_rate": 1.0171965907383156e-05, + "loss": 0.9183, + "step": 45900 + }, + { + "epoch": 2.95, + "learning_rate": 1.0171751218360205e-05, + "loss": 0.9372, + "step": 45901 + }, + { + "epoch": 2.95, + "learning_rate": 1.0171536529337256e-05, + "loss": 0.8549, + "step": 45902 + }, + { + "epoch": 2.95, + "learning_rate": 1.0171321840314305e-05, + "loss": 0.863, + "step": 45903 + }, + { + "epoch": 2.95, + "learning_rate": 1.0171107151291356e-05, + "loss": 0.8235, + "step": 45904 + }, + { + "epoch": 2.95, + "learning_rate": 1.0170892462268405e-05, + "loss": 0.9061, + "step": 45905 + }, + { + "epoch": 2.95, + "learning_rate": 1.0170677773245454e-05, + "loss": 0.9163, + "step": 45906 + }, + { + "epoch": 2.95, + "learning_rate": 1.0170463084222504e-05, + "loss": 0.9236, + "step": 45907 + }, + { + "epoch": 2.95, + "learning_rate": 1.0170248395199553e-05, + "loss": 0.8244, + "step": 45908 + }, + { + "epoch": 2.95, + "learning_rate": 1.0170033706176602e-05, + "loss": 0.8862, + "step": 45909 + }, + { + "epoch": 2.95, + "learning_rate": 1.0169819017153655e-05, + "loss": 0.8794, + "step": 45910 + }, + { + "epoch": 2.95, + "learning_rate": 1.0169604328130704e-05, + "loss": 0.8575, + "step": 45911 + }, + { + "epoch": 2.95, + "learning_rate": 1.0169389639107754e-05, + "loss": 1.0067, + "step": 45912 + }, + { + "epoch": 2.95, + "learning_rate": 1.0169174950084803e-05, + "loss": 0.8696, + "step": 45913 + }, + { + "epoch": 2.95, + "learning_rate": 1.0168960261061854e-05, + "loss": 0.8495, + "step": 45914 + }, + { + "epoch": 2.95, + "learning_rate": 1.0168745572038903e-05, + "loss": 0.8512, + "step": 45915 + }, + { + "epoch": 2.95, + "learning_rate": 1.0168530883015952e-05, + "loss": 0.9324, + "step": 45916 + }, + { + "epoch": 2.95, + "learning_rate": 1.0168316193993002e-05, + "loss": 0.9323, + "step": 45917 + }, + { + "epoch": 2.95, + "learning_rate": 1.0168101504970051e-05, + "loss": 0.9227, + "step": 45918 + }, + { + "epoch": 2.95, + "learning_rate": 1.01678868159471e-05, + "loss": 0.8862, + "step": 45919 + }, + { + "epoch": 2.95, + "learning_rate": 1.0167672126924151e-05, + "loss": 0.9232, + "step": 45920 + }, + { + "epoch": 2.95, + "learning_rate": 1.01674574379012e-05, + "loss": 0.8694, + "step": 45921 + }, + { + "epoch": 2.95, + "learning_rate": 1.016724274887825e-05, + "loss": 0.909, + "step": 45922 + }, + { + "epoch": 2.95, + "learning_rate": 1.01670280598553e-05, + "loss": 0.8971, + "step": 45923 + }, + { + "epoch": 2.95, + "learning_rate": 1.0166813370832349e-05, + "loss": 0.8815, + "step": 45924 + }, + { + "epoch": 2.95, + "learning_rate": 1.0166598681809401e-05, + "loss": 0.8985, + "step": 45925 + }, + { + "epoch": 2.95, + "learning_rate": 1.016638399278645e-05, + "loss": 0.9393, + "step": 45926 + }, + { + "epoch": 2.95, + "learning_rate": 1.01661693037635e-05, + "loss": 0.9029, + "step": 45927 + }, + { + "epoch": 2.95, + "learning_rate": 1.016595461474055e-05, + "loss": 0.932, + "step": 45928 + }, + { + "epoch": 2.95, + "learning_rate": 1.0165739925717598e-05, + "loss": 0.9099, + "step": 45929 + }, + { + "epoch": 2.96, + "learning_rate": 1.016552523669465e-05, + "loss": 0.8994, + "step": 45930 + }, + { + "epoch": 2.96, + "learning_rate": 1.0165310547671698e-05, + "loss": 0.8891, + "step": 45931 + }, + { + "epoch": 2.96, + "learning_rate": 1.0165095858648749e-05, + "loss": 0.8765, + "step": 45932 + }, + { + "epoch": 2.96, + "learning_rate": 1.0164881169625798e-05, + "loss": 0.9252, + "step": 45933 + }, + { + "epoch": 2.96, + "learning_rate": 1.0164666480602847e-05, + "loss": 0.9212, + "step": 45934 + }, + { + "epoch": 2.96, + "learning_rate": 1.0164451791579897e-05, + "loss": 0.8564, + "step": 45935 + }, + { + "epoch": 2.96, + "learning_rate": 1.0164237102556946e-05, + "loss": 0.9457, + "step": 45936 + }, + { + "epoch": 2.96, + "learning_rate": 1.0164022413533995e-05, + "loss": 0.9309, + "step": 45937 + }, + { + "epoch": 2.96, + "learning_rate": 1.0163807724511046e-05, + "loss": 0.8846, + "step": 45938 + }, + { + "epoch": 2.96, + "learning_rate": 1.0163593035488097e-05, + "loss": 0.9629, + "step": 45939 + }, + { + "epoch": 2.96, + "learning_rate": 1.0163378346465147e-05, + "loss": 0.8557, + "step": 45940 + }, + { + "epoch": 2.96, + "learning_rate": 1.0163163657442196e-05, + "loss": 0.8477, + "step": 45941 + }, + { + "epoch": 2.96, + "learning_rate": 1.0162948968419247e-05, + "loss": 0.9801, + "step": 45942 + }, + { + "epoch": 2.96, + "learning_rate": 1.0162734279396296e-05, + "loss": 0.8648, + "step": 45943 + }, + { + "epoch": 2.96, + "learning_rate": 1.0162519590373345e-05, + "loss": 0.9023, + "step": 45944 + }, + { + "epoch": 2.96, + "learning_rate": 1.0162304901350395e-05, + "loss": 0.9604, + "step": 45945 + }, + { + "epoch": 2.96, + "learning_rate": 1.0162090212327444e-05, + "loss": 0.8534, + "step": 45946 + }, + { + "epoch": 2.96, + "learning_rate": 1.0161875523304493e-05, + "loss": 0.8936, + "step": 45947 + }, + { + "epoch": 2.96, + "learning_rate": 1.0161660834281544e-05, + "loss": 0.8641, + "step": 45948 + }, + { + "epoch": 2.96, + "learning_rate": 1.0161446145258593e-05, + "loss": 0.9142, + "step": 45949 + }, + { + "epoch": 2.96, + "learning_rate": 1.0161231456235644e-05, + "loss": 0.8633, + "step": 45950 + }, + { + "epoch": 2.96, + "learning_rate": 1.0161016767212693e-05, + "loss": 0.889, + "step": 45951 + }, + { + "epoch": 2.96, + "learning_rate": 1.0160802078189742e-05, + "loss": 0.9472, + "step": 45952 + }, + { + "epoch": 2.96, + "learning_rate": 1.0160587389166792e-05, + "loss": 0.8681, + "step": 45953 + }, + { + "epoch": 2.96, + "learning_rate": 1.0160372700143843e-05, + "loss": 0.8829, + "step": 45954 + }, + { + "epoch": 2.96, + "learning_rate": 1.0160158011120894e-05, + "loss": 0.9047, + "step": 45955 + }, + { + "epoch": 2.96, + "learning_rate": 1.0159943322097943e-05, + "loss": 0.9081, + "step": 45956 + }, + { + "epoch": 2.96, + "learning_rate": 1.0159728633074992e-05, + "loss": 0.9547, + "step": 45957 + }, + { + "epoch": 2.96, + "learning_rate": 1.0159513944052042e-05, + "loss": 0.7831, + "step": 45958 + }, + { + "epoch": 2.96, + "learning_rate": 1.0159299255029091e-05, + "loss": 0.9034, + "step": 45959 + }, + { + "epoch": 2.96, + "learning_rate": 1.0159084566006142e-05, + "loss": 0.8771, + "step": 45960 + }, + { + "epoch": 2.96, + "learning_rate": 1.015886987698319e-05, + "loss": 0.9043, + "step": 45961 + }, + { + "epoch": 2.96, + "learning_rate": 1.015865518796024e-05, + "loss": 0.8827, + "step": 45962 + }, + { + "epoch": 2.96, + "learning_rate": 1.015844049893729e-05, + "loss": 0.8953, + "step": 45963 + }, + { + "epoch": 2.96, + "learning_rate": 1.015822580991434e-05, + "loss": 0.8544, + "step": 45964 + }, + { + "epoch": 2.96, + "learning_rate": 1.0158011120891388e-05, + "loss": 0.8961, + "step": 45965 + }, + { + "epoch": 2.96, + "learning_rate": 1.0157796431868439e-05, + "loss": 0.836, + "step": 45966 + }, + { + "epoch": 2.96, + "learning_rate": 1.0157581742845488e-05, + "loss": 0.8524, + "step": 45967 + }, + { + "epoch": 2.96, + "learning_rate": 1.015736705382254e-05, + "loss": 0.8413, + "step": 45968 + }, + { + "epoch": 2.96, + "learning_rate": 1.015715236479959e-05, + "loss": 0.875, + "step": 45969 + }, + { + "epoch": 2.96, + "learning_rate": 1.015693767577664e-05, + "loss": 0.9709, + "step": 45970 + }, + { + "epoch": 2.96, + "learning_rate": 1.0156722986753689e-05, + "loss": 0.9025, + "step": 45971 + }, + { + "epoch": 2.96, + "learning_rate": 1.0156508297730738e-05, + "loss": 0.9434, + "step": 45972 + }, + { + "epoch": 2.96, + "learning_rate": 1.0156293608707789e-05, + "loss": 0.9472, + "step": 45973 + }, + { + "epoch": 2.96, + "learning_rate": 1.0156078919684837e-05, + "loss": 0.904, + "step": 45974 + }, + { + "epoch": 2.96, + "learning_rate": 1.0155864230661886e-05, + "loss": 0.9727, + "step": 45975 + }, + { + "epoch": 2.96, + "learning_rate": 1.0155649541638937e-05, + "loss": 0.8799, + "step": 45976 + }, + { + "epoch": 2.96, + "learning_rate": 1.0155434852615986e-05, + "loss": 0.8684, + "step": 45977 + }, + { + "epoch": 2.96, + "learning_rate": 1.0155220163593035e-05, + "loss": 0.8344, + "step": 45978 + }, + { + "epoch": 2.96, + "learning_rate": 1.0155005474570086e-05, + "loss": 0.8276, + "step": 45979 + }, + { + "epoch": 2.96, + "learning_rate": 1.0154790785547135e-05, + "loss": 0.8834, + "step": 45980 + }, + { + "epoch": 2.96, + "learning_rate": 1.0154576096524185e-05, + "loss": 0.9283, + "step": 45981 + }, + { + "epoch": 2.96, + "learning_rate": 1.0154361407501234e-05, + "loss": 0.9274, + "step": 45982 + }, + { + "epoch": 2.96, + "learning_rate": 1.0154146718478287e-05, + "loss": 0.8838, + "step": 45983 + }, + { + "epoch": 2.96, + "learning_rate": 1.0153932029455336e-05, + "loss": 0.923, + "step": 45984 + }, + { + "epoch": 2.96, + "learning_rate": 1.0153717340432385e-05, + "loss": 0.8953, + "step": 45985 + }, + { + "epoch": 2.96, + "learning_rate": 1.0153502651409435e-05, + "loss": 0.9033, + "step": 45986 + }, + { + "epoch": 2.96, + "learning_rate": 1.0153287962386484e-05, + "loss": 0.8588, + "step": 45987 + }, + { + "epoch": 2.96, + "learning_rate": 1.0153073273363533e-05, + "loss": 0.8681, + "step": 45988 + }, + { + "epoch": 2.96, + "learning_rate": 1.0152858584340584e-05, + "loss": 0.8764, + "step": 45989 + }, + { + "epoch": 2.96, + "learning_rate": 1.0152643895317633e-05, + "loss": 0.8844, + "step": 45990 + }, + { + "epoch": 2.96, + "learning_rate": 1.0152429206294683e-05, + "loss": 0.8378, + "step": 45991 + }, + { + "epoch": 2.96, + "learning_rate": 1.0152214517271732e-05, + "loss": 0.8693, + "step": 45992 + }, + { + "epoch": 2.96, + "learning_rate": 1.0151999828248781e-05, + "loss": 0.9464, + "step": 45993 + }, + { + "epoch": 2.96, + "learning_rate": 1.0151785139225832e-05, + "loss": 0.8703, + "step": 45994 + }, + { + "epoch": 2.96, + "learning_rate": 1.0151570450202881e-05, + "loss": 0.9659, + "step": 45995 + }, + { + "epoch": 2.96, + "learning_rate": 1.015135576117993e-05, + "loss": 0.8613, + "step": 45996 + }, + { + "epoch": 2.96, + "learning_rate": 1.0151141072156982e-05, + "loss": 0.8818, + "step": 45997 + }, + { + "epoch": 2.96, + "learning_rate": 1.0150926383134031e-05, + "loss": 0.876, + "step": 45998 + }, + { + "epoch": 2.96, + "learning_rate": 1.0150711694111082e-05, + "loss": 0.823, + "step": 45999 + }, + { + "epoch": 2.96, + "learning_rate": 1.0150497005088131e-05, + "loss": 0.9188, + "step": 46000 + }, + { + "epoch": 2.96, + "eval_laion/cider": 0.5316418362592665, + "eval_laion/meteor": 0.3961539880324666, + "eval_laion/pred_num_words": 38.10546875, + "eval_laion/sacrebleu": 15.769917909563947, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0350570678710938, + "eval_runtime": 102.6414, + "eval_samples_per_second": 4.862, + "eval_steps_per_second": 0.312, + "step": 46000 + }, + { + "epoch": 2.96, + "learning_rate": 1.0150282316065182e-05, + "loss": 0.8447, + "step": 46001 + }, + { + "epoch": 2.96, + "learning_rate": 1.015006762704223e-05, + "loss": 0.9209, + "step": 46002 + }, + { + "epoch": 2.96, + "learning_rate": 1.014985293801928e-05, + "loss": 0.828, + "step": 46003 + }, + { + "epoch": 2.96, + "learning_rate": 1.014963824899633e-05, + "loss": 0.8766, + "step": 46004 + }, + { + "epoch": 2.96, + "learning_rate": 1.0149423559973379e-05, + "loss": 0.8339, + "step": 46005 + }, + { + "epoch": 2.96, + "learning_rate": 1.0149208870950428e-05, + "loss": 0.8905, + "step": 46006 + }, + { + "epoch": 2.96, + "learning_rate": 1.0148994181927479e-05, + "loss": 0.9549, + "step": 46007 + }, + { + "epoch": 2.96, + "learning_rate": 1.0148779492904528e-05, + "loss": 0.8991, + "step": 46008 + }, + { + "epoch": 2.96, + "learning_rate": 1.0148564803881578e-05, + "loss": 0.8617, + "step": 46009 + }, + { + "epoch": 2.96, + "learning_rate": 1.0148350114858627e-05, + "loss": 0.9191, + "step": 46010 + }, + { + "epoch": 2.96, + "learning_rate": 1.0148135425835676e-05, + "loss": 0.9985, + "step": 46011 + }, + { + "epoch": 2.96, + "learning_rate": 1.0147920736812729e-05, + "loss": 0.9121, + "step": 46012 + }, + { + "epoch": 2.96, + "learning_rate": 1.0147706047789778e-05, + "loss": 0.9271, + "step": 46013 + }, + { + "epoch": 2.96, + "learning_rate": 1.0147491358766828e-05, + "loss": 0.8853, + "step": 46014 + }, + { + "epoch": 2.96, + "learning_rate": 1.0147276669743877e-05, + "loss": 0.9528, + "step": 46015 + }, + { + "epoch": 2.96, + "learning_rate": 1.0147061980720926e-05, + "loss": 0.8391, + "step": 46016 + }, + { + "epoch": 2.96, + "learning_rate": 1.0146847291697977e-05, + "loss": 0.9139, + "step": 46017 + }, + { + "epoch": 2.96, + "learning_rate": 1.0146632602675026e-05, + "loss": 0.8943, + "step": 46018 + }, + { + "epoch": 2.96, + "learning_rate": 1.0146417913652076e-05, + "loss": 0.8463, + "step": 46019 + }, + { + "epoch": 2.96, + "learning_rate": 1.0146203224629125e-05, + "loss": 0.8862, + "step": 46020 + }, + { + "epoch": 2.96, + "learning_rate": 1.0145988535606174e-05, + "loss": 0.8596, + "step": 46021 + }, + { + "epoch": 2.96, + "learning_rate": 1.0145773846583225e-05, + "loss": 0.8564, + "step": 46022 + }, + { + "epoch": 2.96, + "learning_rate": 1.0145559157560274e-05, + "loss": 0.9399, + "step": 46023 + }, + { + "epoch": 2.96, + "learning_rate": 1.0145344468537323e-05, + "loss": 0.9567, + "step": 46024 + }, + { + "epoch": 2.96, + "learning_rate": 1.0145129779514374e-05, + "loss": 0.901, + "step": 46025 + }, + { + "epoch": 2.96, + "learning_rate": 1.0144915090491424e-05, + "loss": 0.9178, + "step": 46026 + }, + { + "epoch": 2.96, + "learning_rate": 1.0144700401468475e-05, + "loss": 0.8257, + "step": 46027 + }, + { + "epoch": 2.96, + "learning_rate": 1.0144485712445524e-05, + "loss": 0.866, + "step": 46028 + }, + { + "epoch": 2.96, + "learning_rate": 1.0144271023422575e-05, + "loss": 0.8305, + "step": 46029 + }, + { + "epoch": 2.96, + "learning_rate": 1.0144056334399624e-05, + "loss": 0.8434, + "step": 46030 + }, + { + "epoch": 2.96, + "learning_rate": 1.0143841645376673e-05, + "loss": 0.9295, + "step": 46031 + }, + { + "epoch": 2.96, + "learning_rate": 1.0143626956353723e-05, + "loss": 0.8781, + "step": 46032 + }, + { + "epoch": 2.96, + "learning_rate": 1.0143412267330772e-05, + "loss": 0.8741, + "step": 46033 + }, + { + "epoch": 2.96, + "learning_rate": 1.0143197578307821e-05, + "loss": 0.9083, + "step": 46034 + }, + { + "epoch": 2.96, + "learning_rate": 1.0142982889284872e-05, + "loss": 0.867, + "step": 46035 + }, + { + "epoch": 2.96, + "learning_rate": 1.014276820026192e-05, + "loss": 0.9129, + "step": 46036 + }, + { + "epoch": 2.96, + "learning_rate": 1.0142553511238971e-05, + "loss": 0.902, + "step": 46037 + }, + { + "epoch": 2.96, + "learning_rate": 1.014233882221602e-05, + "loss": 0.8289, + "step": 46038 + }, + { + "epoch": 2.96, + "learning_rate": 1.014212413319307e-05, + "loss": 0.8207, + "step": 46039 + }, + { + "epoch": 2.96, + "learning_rate": 1.014190944417012e-05, + "loss": 0.881, + "step": 46040 + }, + { + "epoch": 2.96, + "learning_rate": 1.014169475514717e-05, + "loss": 0.8996, + "step": 46041 + }, + { + "epoch": 2.96, + "learning_rate": 1.0141480066124221e-05, + "loss": 0.9031, + "step": 46042 + }, + { + "epoch": 2.96, + "learning_rate": 1.014126537710127e-05, + "loss": 0.9446, + "step": 46043 + }, + { + "epoch": 2.96, + "learning_rate": 1.014105068807832e-05, + "loss": 0.9299, + "step": 46044 + }, + { + "epoch": 2.96, + "learning_rate": 1.014083599905537e-05, + "loss": 0.9855, + "step": 46045 + }, + { + "epoch": 2.96, + "learning_rate": 1.0140621310032419e-05, + "loss": 0.865, + "step": 46046 + }, + { + "epoch": 2.96, + "learning_rate": 1.014040662100947e-05, + "loss": 0.9497, + "step": 46047 + }, + { + "epoch": 2.96, + "learning_rate": 1.0140191931986518e-05, + "loss": 0.8649, + "step": 46048 + }, + { + "epoch": 2.96, + "learning_rate": 1.0139977242963567e-05, + "loss": 0.913, + "step": 46049 + }, + { + "epoch": 2.96, + "learning_rate": 1.0139762553940618e-05, + "loss": 0.8903, + "step": 46050 + }, + { + "epoch": 2.96, + "learning_rate": 1.0139547864917667e-05, + "loss": 0.9501, + "step": 46051 + }, + { + "epoch": 2.96, + "learning_rate": 1.0139333175894716e-05, + "loss": 0.9915, + "step": 46052 + }, + { + "epoch": 2.96, + "learning_rate": 1.0139118486871767e-05, + "loss": 0.847, + "step": 46053 + }, + { + "epoch": 2.96, + "learning_rate": 1.0138903797848816e-05, + "loss": 0.9071, + "step": 46054 + }, + { + "epoch": 2.96, + "learning_rate": 1.0138689108825868e-05, + "loss": 0.8419, + "step": 46055 + }, + { + "epoch": 2.96, + "learning_rate": 1.0138474419802917e-05, + "loss": 0.9228, + "step": 46056 + }, + { + "epoch": 2.96, + "learning_rate": 1.0138259730779968e-05, + "loss": 0.8863, + "step": 46057 + }, + { + "epoch": 2.96, + "learning_rate": 1.0138045041757017e-05, + "loss": 0.9257, + "step": 46058 + }, + { + "epoch": 2.96, + "learning_rate": 1.0137830352734066e-05, + "loss": 0.8761, + "step": 46059 + }, + { + "epoch": 2.96, + "learning_rate": 1.0137615663711116e-05, + "loss": 0.8645, + "step": 46060 + }, + { + "epoch": 2.96, + "learning_rate": 1.0137400974688165e-05, + "loss": 0.835, + "step": 46061 + }, + { + "epoch": 2.96, + "learning_rate": 1.0137186285665214e-05, + "loss": 0.8671, + "step": 46062 + }, + { + "epoch": 2.96, + "learning_rate": 1.0136971596642265e-05, + "loss": 0.9121, + "step": 46063 + }, + { + "epoch": 2.96, + "learning_rate": 1.0136756907619314e-05, + "loss": 0.9024, + "step": 46064 + }, + { + "epoch": 2.96, + "learning_rate": 1.0136542218596363e-05, + "loss": 0.8793, + "step": 46065 + }, + { + "epoch": 2.96, + "learning_rate": 1.0136327529573413e-05, + "loss": 0.9252, + "step": 46066 + }, + { + "epoch": 2.96, + "learning_rate": 1.0136112840550462e-05, + "loss": 0.8797, + "step": 46067 + }, + { + "epoch": 2.96, + "learning_rate": 1.0135898151527513e-05, + "loss": 0.8814, + "step": 46068 + }, + { + "epoch": 2.96, + "learning_rate": 1.0135683462504562e-05, + "loss": 0.8623, + "step": 46069 + }, + { + "epoch": 2.96, + "learning_rate": 1.0135468773481614e-05, + "loss": 0.9291, + "step": 46070 + }, + { + "epoch": 2.96, + "learning_rate": 1.0135254084458663e-05, + "loss": 0.9587, + "step": 46071 + }, + { + "epoch": 2.96, + "learning_rate": 1.0135039395435712e-05, + "loss": 0.9319, + "step": 46072 + }, + { + "epoch": 2.96, + "learning_rate": 1.0134824706412763e-05, + "loss": 0.8647, + "step": 46073 + }, + { + "epoch": 2.96, + "learning_rate": 1.0134610017389812e-05, + "loss": 0.9095, + "step": 46074 + }, + { + "epoch": 2.96, + "learning_rate": 1.0134395328366861e-05, + "loss": 0.9338, + "step": 46075 + }, + { + "epoch": 2.96, + "learning_rate": 1.0134180639343912e-05, + "loss": 0.9392, + "step": 46076 + }, + { + "epoch": 2.96, + "learning_rate": 1.013396595032096e-05, + "loss": 0.8322, + "step": 46077 + }, + { + "epoch": 2.96, + "learning_rate": 1.0133751261298011e-05, + "loss": 0.9109, + "step": 46078 + }, + { + "epoch": 2.96, + "learning_rate": 1.013353657227506e-05, + "loss": 0.9352, + "step": 46079 + }, + { + "epoch": 2.96, + "learning_rate": 1.0133321883252109e-05, + "loss": 0.9193, + "step": 46080 + }, + { + "epoch": 2.96, + "learning_rate": 1.013310719422916e-05, + "loss": 0.9123, + "step": 46081 + }, + { + "epoch": 2.96, + "learning_rate": 1.0132892505206209e-05, + "loss": 0.8646, + "step": 46082 + }, + { + "epoch": 2.96, + "learning_rate": 1.0132677816183258e-05, + "loss": 0.8109, + "step": 46083 + }, + { + "epoch": 2.96, + "learning_rate": 1.013246312716031e-05, + "loss": 0.997, + "step": 46084 + }, + { + "epoch": 2.97, + "learning_rate": 1.0132248438137359e-05, + "loss": 0.8982, + "step": 46085 + }, + { + "epoch": 2.97, + "learning_rate": 1.013203374911441e-05, + "loss": 0.8944, + "step": 46086 + }, + { + "epoch": 2.97, + "learning_rate": 1.0131819060091459e-05, + "loss": 0.8387, + "step": 46087 + }, + { + "epoch": 2.97, + "learning_rate": 1.013160437106851e-05, + "loss": 0.8603, + "step": 46088 + }, + { + "epoch": 2.97, + "learning_rate": 1.0131389682045558e-05, + "loss": 0.9121, + "step": 46089 + }, + { + "epoch": 2.97, + "learning_rate": 1.0131174993022607e-05, + "loss": 0.9471, + "step": 46090 + }, + { + "epoch": 2.97, + "learning_rate": 1.0130960303999658e-05, + "loss": 0.8591, + "step": 46091 + }, + { + "epoch": 2.97, + "learning_rate": 1.0130745614976707e-05, + "loss": 0.9177, + "step": 46092 + }, + { + "epoch": 2.97, + "learning_rate": 1.0130530925953756e-05, + "loss": 0.854, + "step": 46093 + }, + { + "epoch": 2.97, + "learning_rate": 1.0130316236930806e-05, + "loss": 0.8772, + "step": 46094 + }, + { + "epoch": 2.97, + "learning_rate": 1.0130101547907855e-05, + "loss": 0.8992, + "step": 46095 + }, + { + "epoch": 2.97, + "learning_rate": 1.0129886858884906e-05, + "loss": 0.8551, + "step": 46096 + }, + { + "epoch": 2.97, + "learning_rate": 1.0129672169861955e-05, + "loss": 0.9066, + "step": 46097 + }, + { + "epoch": 2.97, + "learning_rate": 1.0129457480839004e-05, + "loss": 0.8935, + "step": 46098 + }, + { + "epoch": 2.97, + "learning_rate": 1.0129242791816056e-05, + "loss": 0.8615, + "step": 46099 + }, + { + "epoch": 2.97, + "learning_rate": 1.0129028102793105e-05, + "loss": 0.936, + "step": 46100 + }, + { + "epoch": 2.97, + "learning_rate": 1.0128813413770156e-05, + "loss": 0.9206, + "step": 46101 + }, + { + "epoch": 2.97, + "learning_rate": 1.0128598724747205e-05, + "loss": 0.8954, + "step": 46102 + }, + { + "epoch": 2.97, + "learning_rate": 1.0128384035724254e-05, + "loss": 0.8707, + "step": 46103 + }, + { + "epoch": 2.97, + "learning_rate": 1.0128169346701305e-05, + "loss": 0.8884, + "step": 46104 + }, + { + "epoch": 2.97, + "learning_rate": 1.0127954657678354e-05, + "loss": 0.8689, + "step": 46105 + }, + { + "epoch": 2.97, + "learning_rate": 1.0127739968655404e-05, + "loss": 0.9068, + "step": 46106 + }, + { + "epoch": 2.97, + "learning_rate": 1.0127525279632453e-05, + "loss": 0.8924, + "step": 46107 + }, + { + "epoch": 2.97, + "learning_rate": 1.0127310590609502e-05, + "loss": 0.8893, + "step": 46108 + }, + { + "epoch": 2.97, + "learning_rate": 1.0127095901586553e-05, + "loss": 0.8588, + "step": 46109 + }, + { + "epoch": 2.97, + "learning_rate": 1.0126881212563602e-05, + "loss": 0.9258, + "step": 46110 + }, + { + "epoch": 2.97, + "learning_rate": 1.012666652354065e-05, + "loss": 0.868, + "step": 46111 + }, + { + "epoch": 2.97, + "learning_rate": 1.0126451834517701e-05, + "loss": 0.8821, + "step": 46112 + }, + { + "epoch": 2.97, + "learning_rate": 1.0126237145494752e-05, + "loss": 0.8683, + "step": 46113 + }, + { + "epoch": 2.97, + "learning_rate": 1.0126022456471803e-05, + "loss": 0.9148, + "step": 46114 + }, + { + "epoch": 2.97, + "learning_rate": 1.0125807767448852e-05, + "loss": 0.9269, + "step": 46115 + }, + { + "epoch": 2.97, + "learning_rate": 1.0125593078425902e-05, + "loss": 0.9189, + "step": 46116 + }, + { + "epoch": 2.97, + "learning_rate": 1.0125378389402951e-05, + "loss": 0.8849, + "step": 46117 + }, + { + "epoch": 2.97, + "learning_rate": 1.012516370038e-05, + "loss": 0.9179, + "step": 46118 + }, + { + "epoch": 2.97, + "learning_rate": 1.0124949011357051e-05, + "loss": 0.9082, + "step": 46119 + }, + { + "epoch": 2.97, + "learning_rate": 1.01247343223341e-05, + "loss": 0.9248, + "step": 46120 + }, + { + "epoch": 2.97, + "learning_rate": 1.0124519633311149e-05, + "loss": 0.8511, + "step": 46121 + }, + { + "epoch": 2.97, + "learning_rate": 1.01243049442882e-05, + "loss": 0.9442, + "step": 46122 + }, + { + "epoch": 2.97, + "learning_rate": 1.0124090255265248e-05, + "loss": 0.8753, + "step": 46123 + }, + { + "epoch": 2.97, + "learning_rate": 1.0123875566242299e-05, + "loss": 0.8465, + "step": 46124 + }, + { + "epoch": 2.97, + "learning_rate": 1.0123660877219348e-05, + "loss": 0.9359, + "step": 46125 + }, + { + "epoch": 2.97, + "learning_rate": 1.0123446188196397e-05, + "loss": 0.918, + "step": 46126 + }, + { + "epoch": 2.97, + "learning_rate": 1.0123231499173448e-05, + "loss": 0.9076, + "step": 46127 + }, + { + "epoch": 2.97, + "learning_rate": 1.0123016810150498e-05, + "loss": 0.8824, + "step": 46128 + }, + { + "epoch": 2.97, + "learning_rate": 1.0122802121127549e-05, + "loss": 0.8864, + "step": 46129 + }, + { + "epoch": 2.97, + "learning_rate": 1.0122587432104598e-05, + "loss": 0.907, + "step": 46130 + }, + { + "epoch": 2.97, + "learning_rate": 1.0122372743081647e-05, + "loss": 0.8037, + "step": 46131 + }, + { + "epoch": 2.97, + "learning_rate": 1.0122158054058698e-05, + "loss": 0.9225, + "step": 46132 + }, + { + "epoch": 2.97, + "learning_rate": 1.0121943365035747e-05, + "loss": 0.8799, + "step": 46133 + }, + { + "epoch": 2.97, + "learning_rate": 1.0121728676012797e-05, + "loss": 0.9339, + "step": 46134 + }, + { + "epoch": 2.97, + "learning_rate": 1.0121513986989846e-05, + "loss": 0.9033, + "step": 46135 + }, + { + "epoch": 2.97, + "learning_rate": 1.0121299297966895e-05, + "loss": 0.9446, + "step": 46136 + }, + { + "epoch": 2.97, + "learning_rate": 1.0121084608943946e-05, + "loss": 0.9062, + "step": 46137 + }, + { + "epoch": 2.97, + "learning_rate": 1.0120869919920995e-05, + "loss": 0.9117, + "step": 46138 + }, + { + "epoch": 2.97, + "learning_rate": 1.0120655230898044e-05, + "loss": 0.9154, + "step": 46139 + }, + { + "epoch": 2.97, + "learning_rate": 1.0120440541875094e-05, + "loss": 1.0032, + "step": 46140 + }, + { + "epoch": 2.97, + "learning_rate": 1.0120225852852143e-05, + "loss": 0.9114, + "step": 46141 + }, + { + "epoch": 2.97, + "learning_rate": 1.0120011163829196e-05, + "loss": 0.9587, + "step": 46142 + }, + { + "epoch": 2.97, + "learning_rate": 1.0119796474806245e-05, + "loss": 0.8773, + "step": 46143 + }, + { + "epoch": 2.97, + "learning_rate": 1.0119581785783295e-05, + "loss": 0.8611, + "step": 46144 + }, + { + "epoch": 2.97, + "learning_rate": 1.0119367096760344e-05, + "loss": 0.9076, + "step": 46145 + }, + { + "epoch": 2.97, + "learning_rate": 1.0119152407737393e-05, + "loss": 0.8569, + "step": 46146 + }, + { + "epoch": 2.97, + "learning_rate": 1.0118937718714444e-05, + "loss": 0.8543, + "step": 46147 + }, + { + "epoch": 2.97, + "learning_rate": 1.0118723029691493e-05, + "loss": 0.8905, + "step": 46148 + }, + { + "epoch": 2.97, + "learning_rate": 1.0118508340668542e-05, + "loss": 0.9402, + "step": 46149 + }, + { + "epoch": 2.97, + "learning_rate": 1.0118293651645593e-05, + "loss": 0.8642, + "step": 46150 + }, + { + "epoch": 2.97, + "learning_rate": 1.0118078962622642e-05, + "loss": 0.9133, + "step": 46151 + }, + { + "epoch": 2.97, + "learning_rate": 1.011786427359969e-05, + "loss": 0.9665, + "step": 46152 + }, + { + "epoch": 2.97, + "learning_rate": 1.0117649584576741e-05, + "loss": 0.9326, + "step": 46153 + }, + { + "epoch": 2.97, + "learning_rate": 1.011743489555379e-05, + "loss": 0.9658, + "step": 46154 + }, + { + "epoch": 2.97, + "learning_rate": 1.011722020653084e-05, + "loss": 0.9581, + "step": 46155 + }, + { + "epoch": 2.97, + "learning_rate": 1.011700551750789e-05, + "loss": 0.9295, + "step": 46156 + }, + { + "epoch": 2.97, + "learning_rate": 1.0116790828484942e-05, + "loss": 0.8548, + "step": 46157 + }, + { + "epoch": 2.97, + "learning_rate": 1.0116576139461991e-05, + "loss": 0.8439, + "step": 46158 + }, + { + "epoch": 2.97, + "learning_rate": 1.011636145043904e-05, + "loss": 0.8817, + "step": 46159 + }, + { + "epoch": 2.97, + "learning_rate": 1.011614676141609e-05, + "loss": 0.8579, + "step": 46160 + }, + { + "epoch": 2.97, + "learning_rate": 1.011593207239314e-05, + "loss": 0.9016, + "step": 46161 + }, + { + "epoch": 2.97, + "learning_rate": 1.0115717383370189e-05, + "loss": 0.8929, + "step": 46162 + }, + { + "epoch": 2.97, + "learning_rate": 1.011550269434724e-05, + "loss": 0.859, + "step": 46163 + }, + { + "epoch": 2.97, + "learning_rate": 1.0115288005324288e-05, + "loss": 0.9018, + "step": 46164 + }, + { + "epoch": 2.97, + "learning_rate": 1.0115073316301339e-05, + "loss": 0.9341, + "step": 46165 + }, + { + "epoch": 2.97, + "learning_rate": 1.0114858627278388e-05, + "loss": 0.8547, + "step": 46166 + }, + { + "epoch": 2.97, + "learning_rate": 1.0114643938255437e-05, + "loss": 0.9828, + "step": 46167 + }, + { + "epoch": 2.97, + "learning_rate": 1.0114429249232487e-05, + "loss": 0.8871, + "step": 46168 + }, + { + "epoch": 2.97, + "learning_rate": 1.0114214560209536e-05, + "loss": 0.8653, + "step": 46169 + }, + { + "epoch": 2.97, + "learning_rate": 1.0113999871186585e-05, + "loss": 0.869, + "step": 46170 + }, + { + "epoch": 2.97, + "learning_rate": 1.0113785182163636e-05, + "loss": 0.8687, + "step": 46171 + }, + { + "epoch": 2.97, + "learning_rate": 1.0113570493140687e-05, + "loss": 0.9216, + "step": 46172 + }, + { + "epoch": 2.97, + "learning_rate": 1.0113355804117737e-05, + "loss": 0.8961, + "step": 46173 + }, + { + "epoch": 2.97, + "learning_rate": 1.0113141115094786e-05, + "loss": 0.8964, + "step": 46174 + }, + { + "epoch": 2.97, + "learning_rate": 1.0112926426071837e-05, + "loss": 0.9362, + "step": 46175 + }, + { + "epoch": 2.97, + "learning_rate": 1.0112711737048886e-05, + "loss": 0.8124, + "step": 46176 + }, + { + "epoch": 2.97, + "learning_rate": 1.0112497048025935e-05, + "loss": 0.9461, + "step": 46177 + }, + { + "epoch": 2.97, + "learning_rate": 1.0112282359002986e-05, + "loss": 0.8614, + "step": 46178 + }, + { + "epoch": 2.97, + "learning_rate": 1.0112067669980035e-05, + "loss": 0.9818, + "step": 46179 + }, + { + "epoch": 2.97, + "learning_rate": 1.0111852980957084e-05, + "loss": 0.8934, + "step": 46180 + }, + { + "epoch": 2.97, + "learning_rate": 1.0111638291934134e-05, + "loss": 0.8865, + "step": 46181 + }, + { + "epoch": 2.97, + "learning_rate": 1.0111423602911183e-05, + "loss": 0.8883, + "step": 46182 + }, + { + "epoch": 2.97, + "learning_rate": 1.0111208913888234e-05, + "loss": 0.848, + "step": 46183 + }, + { + "epoch": 2.97, + "learning_rate": 1.0110994224865283e-05, + "loss": 0.9879, + "step": 46184 + }, + { + "epoch": 2.97, + "learning_rate": 1.0110779535842332e-05, + "loss": 0.9325, + "step": 46185 + }, + { + "epoch": 2.97, + "learning_rate": 1.0110564846819384e-05, + "loss": 0.9355, + "step": 46186 + }, + { + "epoch": 2.97, + "learning_rate": 1.0110350157796433e-05, + "loss": 0.8631, + "step": 46187 + }, + { + "epoch": 2.97, + "learning_rate": 1.0110135468773484e-05, + "loss": 0.9146, + "step": 46188 + }, + { + "epoch": 2.97, + "learning_rate": 1.0109920779750533e-05, + "loss": 0.8634, + "step": 46189 + }, + { + "epoch": 2.97, + "learning_rate": 1.0109706090727582e-05, + "loss": 0.9291, + "step": 46190 + }, + { + "epoch": 2.97, + "learning_rate": 1.0109491401704632e-05, + "loss": 0.8834, + "step": 46191 + }, + { + "epoch": 2.97, + "learning_rate": 1.0109276712681681e-05, + "loss": 0.8664, + "step": 46192 + }, + { + "epoch": 2.97, + "learning_rate": 1.0109062023658732e-05, + "loss": 0.8682, + "step": 46193 + }, + { + "epoch": 2.97, + "learning_rate": 1.0108847334635781e-05, + "loss": 0.8328, + "step": 46194 + }, + { + "epoch": 2.97, + "learning_rate": 1.010863264561283e-05, + "loss": 0.8416, + "step": 46195 + }, + { + "epoch": 2.97, + "learning_rate": 1.010841795658988e-05, + "loss": 0.8981, + "step": 46196 + }, + { + "epoch": 2.97, + "learning_rate": 1.010820326756693e-05, + "loss": 0.9037, + "step": 46197 + }, + { + "epoch": 2.97, + "learning_rate": 1.0107988578543978e-05, + "loss": 0.9262, + "step": 46198 + }, + { + "epoch": 2.97, + "learning_rate": 1.0107773889521029e-05, + "loss": 0.9428, + "step": 46199 + }, + { + "epoch": 2.97, + "learning_rate": 1.0107559200498078e-05, + "loss": 0.8836, + "step": 46200 + }, + { + "epoch": 2.97, + "learning_rate": 1.010734451147513e-05, + "loss": 0.9402, + "step": 46201 + }, + { + "epoch": 2.97, + "learning_rate": 1.010712982245218e-05, + "loss": 0.9303, + "step": 46202 + }, + { + "epoch": 2.97, + "learning_rate": 1.010691513342923e-05, + "loss": 0.8736, + "step": 46203 + }, + { + "epoch": 2.97, + "learning_rate": 1.0106700444406279e-05, + "loss": 0.924, + "step": 46204 + }, + { + "epoch": 2.97, + "learning_rate": 1.0106485755383328e-05, + "loss": 0.847, + "step": 46205 + }, + { + "epoch": 2.97, + "learning_rate": 1.0106271066360379e-05, + "loss": 0.888, + "step": 46206 + }, + { + "epoch": 2.97, + "learning_rate": 1.0106056377337428e-05, + "loss": 0.8162, + "step": 46207 + }, + { + "epoch": 2.97, + "learning_rate": 1.0105841688314477e-05, + "loss": 0.9239, + "step": 46208 + }, + { + "epoch": 2.97, + "learning_rate": 1.0105626999291527e-05, + "loss": 0.9122, + "step": 46209 + }, + { + "epoch": 2.97, + "learning_rate": 1.0105412310268576e-05, + "loss": 0.8911, + "step": 46210 + }, + { + "epoch": 2.97, + "learning_rate": 1.0105197621245625e-05, + "loss": 0.9042, + "step": 46211 + }, + { + "epoch": 2.97, + "learning_rate": 1.0104982932222676e-05, + "loss": 0.8843, + "step": 46212 + }, + { + "epoch": 2.97, + "learning_rate": 1.0104768243199725e-05, + "loss": 0.8567, + "step": 46213 + }, + { + "epoch": 2.97, + "learning_rate": 1.0104553554176775e-05, + "loss": 0.9231, + "step": 46214 + }, + { + "epoch": 2.97, + "learning_rate": 1.0104338865153826e-05, + "loss": 0.9384, + "step": 46215 + }, + { + "epoch": 2.97, + "learning_rate": 1.0104124176130877e-05, + "loss": 0.8371, + "step": 46216 + }, + { + "epoch": 2.97, + "learning_rate": 1.0103909487107926e-05, + "loss": 0.8833, + "step": 46217 + }, + { + "epoch": 2.97, + "learning_rate": 1.0103694798084975e-05, + "loss": 0.9228, + "step": 46218 + }, + { + "epoch": 2.97, + "learning_rate": 1.0103480109062025e-05, + "loss": 0.857, + "step": 46219 + }, + { + "epoch": 2.97, + "learning_rate": 1.0103265420039074e-05, + "loss": 0.8425, + "step": 46220 + }, + { + "epoch": 2.97, + "learning_rate": 1.0103050731016123e-05, + "loss": 0.8802, + "step": 46221 + }, + { + "epoch": 2.97, + "learning_rate": 1.0102836041993174e-05, + "loss": 0.9107, + "step": 46222 + }, + { + "epoch": 2.97, + "learning_rate": 1.0102621352970223e-05, + "loss": 0.8625, + "step": 46223 + }, + { + "epoch": 2.97, + "learning_rate": 1.0102406663947274e-05, + "loss": 0.8499, + "step": 46224 + }, + { + "epoch": 2.97, + "learning_rate": 1.0102191974924323e-05, + "loss": 0.8618, + "step": 46225 + }, + { + "epoch": 2.97, + "learning_rate": 1.0101977285901371e-05, + "loss": 0.8798, + "step": 46226 + }, + { + "epoch": 2.97, + "learning_rate": 1.0101762596878422e-05, + "loss": 0.8944, + "step": 46227 + }, + { + "epoch": 2.97, + "learning_rate": 1.0101547907855471e-05, + "loss": 0.9429, + "step": 46228 + }, + { + "epoch": 2.97, + "learning_rate": 1.010133321883252e-05, + "loss": 0.9044, + "step": 46229 + }, + { + "epoch": 2.97, + "learning_rate": 1.0101118529809572e-05, + "loss": 0.8894, + "step": 46230 + }, + { + "epoch": 2.97, + "learning_rate": 1.0100903840786623e-05, + "loss": 0.8514, + "step": 46231 + }, + { + "epoch": 2.97, + "learning_rate": 1.0100689151763672e-05, + "loss": 0.9935, + "step": 46232 + }, + { + "epoch": 2.97, + "learning_rate": 1.0100474462740721e-05, + "loss": 0.8891, + "step": 46233 + }, + { + "epoch": 2.97, + "learning_rate": 1.0100259773717772e-05, + "loss": 0.9455, + "step": 46234 + }, + { + "epoch": 2.97, + "learning_rate": 1.010004508469482e-05, + "loss": 0.9351, + "step": 46235 + }, + { + "epoch": 2.97, + "learning_rate": 1.009983039567187e-05, + "loss": 0.8908, + "step": 46236 + }, + { + "epoch": 2.97, + "learning_rate": 1.009961570664892e-05, + "loss": 0.8784, + "step": 46237 + }, + { + "epoch": 2.97, + "learning_rate": 1.009940101762597e-05, + "loss": 0.8807, + "step": 46238 + }, + { + "epoch": 2.97, + "learning_rate": 1.0099186328603018e-05, + "loss": 0.964, + "step": 46239 + }, + { + "epoch": 2.97, + "learning_rate": 1.0098971639580069e-05, + "loss": 0.8833, + "step": 46240 + }, + { + "epoch": 2.98, + "learning_rate": 1.0098756950557118e-05, + "loss": 0.9407, + "step": 46241 + }, + { + "epoch": 2.98, + "learning_rate": 1.0098542261534168e-05, + "loss": 0.8785, + "step": 46242 + }, + { + "epoch": 2.98, + "learning_rate": 1.0098327572511217e-05, + "loss": 0.9378, + "step": 46243 + }, + { + "epoch": 2.98, + "learning_rate": 1.009811288348827e-05, + "loss": 0.9392, + "step": 46244 + }, + { + "epoch": 2.98, + "learning_rate": 1.0097898194465319e-05, + "loss": 0.9018, + "step": 46245 + }, + { + "epoch": 2.98, + "learning_rate": 1.0097683505442368e-05, + "loss": 0.8449, + "step": 46246 + }, + { + "epoch": 2.98, + "learning_rate": 1.0097468816419418e-05, + "loss": 0.9509, + "step": 46247 + }, + { + "epoch": 2.98, + "learning_rate": 1.0097254127396467e-05, + "loss": 0.8451, + "step": 46248 + }, + { + "epoch": 2.98, + "learning_rate": 1.0097039438373516e-05, + "loss": 0.9987, + "step": 46249 + }, + { + "epoch": 2.98, + "learning_rate": 1.0096824749350567e-05, + "loss": 0.8698, + "step": 46250 + }, + { + "epoch": 2.98, + "learning_rate": 1.0096610060327616e-05, + "loss": 0.9418, + "step": 46251 + }, + { + "epoch": 2.98, + "learning_rate": 1.0096395371304667e-05, + "loss": 0.8907, + "step": 46252 + }, + { + "epoch": 2.98, + "learning_rate": 1.0096180682281716e-05, + "loss": 0.8838, + "step": 46253 + }, + { + "epoch": 2.98, + "learning_rate": 1.0095965993258765e-05, + "loss": 0.9335, + "step": 46254 + }, + { + "epoch": 2.98, + "learning_rate": 1.0095751304235815e-05, + "loss": 1.019, + "step": 46255 + }, + { + "epoch": 2.98, + "learning_rate": 1.0095536615212864e-05, + "loss": 0.8236, + "step": 46256 + }, + { + "epoch": 2.98, + "learning_rate": 1.0095321926189913e-05, + "loss": 0.8555, + "step": 46257 + }, + { + "epoch": 2.98, + "learning_rate": 1.0095107237166964e-05, + "loss": 0.8988, + "step": 46258 + }, + { + "epoch": 2.98, + "learning_rate": 1.0094892548144014e-05, + "loss": 0.9213, + "step": 46259 + }, + { + "epoch": 2.98, + "learning_rate": 1.0094677859121065e-05, + "loss": 0.8437, + "step": 46260 + }, + { + "epoch": 2.98, + "learning_rate": 1.0094463170098114e-05, + "loss": 0.8983, + "step": 46261 + }, + { + "epoch": 2.98, + "learning_rate": 1.0094248481075165e-05, + "loss": 0.9276, + "step": 46262 + }, + { + "epoch": 2.98, + "learning_rate": 1.0094033792052214e-05, + "loss": 0.8832, + "step": 46263 + }, + { + "epoch": 2.98, + "learning_rate": 1.0093819103029263e-05, + "loss": 0.8622, + "step": 46264 + }, + { + "epoch": 2.98, + "learning_rate": 1.0093604414006313e-05, + "loss": 0.8952, + "step": 46265 + }, + { + "epoch": 2.98, + "learning_rate": 1.0093389724983362e-05, + "loss": 0.8638, + "step": 46266 + }, + { + "epoch": 2.98, + "learning_rate": 1.0093175035960411e-05, + "loss": 0.9007, + "step": 46267 + }, + { + "epoch": 2.98, + "learning_rate": 1.0092960346937462e-05, + "loss": 0.9159, + "step": 46268 + }, + { + "epoch": 2.98, + "learning_rate": 1.0092745657914511e-05, + "loss": 0.8877, + "step": 46269 + }, + { + "epoch": 2.98, + "learning_rate": 1.0092530968891562e-05, + "loss": 0.9143, + "step": 46270 + }, + { + "epoch": 2.98, + "learning_rate": 1.009231627986861e-05, + "loss": 0.8764, + "step": 46271 + }, + { + "epoch": 2.98, + "learning_rate": 1.009210159084566e-05, + "loss": 0.9884, + "step": 46272 + }, + { + "epoch": 2.98, + "learning_rate": 1.0091886901822712e-05, + "loss": 0.9218, + "step": 46273 + }, + { + "epoch": 2.98, + "learning_rate": 1.009167221279976e-05, + "loss": 0.8886, + "step": 46274 + }, + { + "epoch": 2.98, + "learning_rate": 1.0091457523776811e-05, + "loss": 0.9311, + "step": 46275 + }, + { + "epoch": 2.98, + "learning_rate": 1.009124283475386e-05, + "loss": 0.9096, + "step": 46276 + }, + { + "epoch": 2.98, + "learning_rate": 1.009102814573091e-05, + "loss": 0.8614, + "step": 46277 + }, + { + "epoch": 2.98, + "learning_rate": 1.009081345670796e-05, + "loss": 0.8996, + "step": 46278 + }, + { + "epoch": 2.98, + "learning_rate": 1.0090598767685009e-05, + "loss": 0.925, + "step": 46279 + }, + { + "epoch": 2.98, + "learning_rate": 1.009038407866206e-05, + "loss": 0.9209, + "step": 46280 + }, + { + "epoch": 2.98, + "learning_rate": 1.0090169389639109e-05, + "loss": 0.8554, + "step": 46281 + }, + { + "epoch": 2.98, + "learning_rate": 1.0089954700616158e-05, + "loss": 0.9172, + "step": 46282 + }, + { + "epoch": 2.98, + "learning_rate": 1.0089740011593208e-05, + "loss": 0.8665, + "step": 46283 + }, + { + "epoch": 2.98, + "learning_rate": 1.0089525322570257e-05, + "loss": 0.9201, + "step": 46284 + }, + { + "epoch": 2.98, + "learning_rate": 1.0089310633547306e-05, + "loss": 0.8898, + "step": 46285 + }, + { + "epoch": 2.98, + "learning_rate": 1.0089095944524357e-05, + "loss": 0.9263, + "step": 46286 + }, + { + "epoch": 2.98, + "learning_rate": 1.0088881255501406e-05, + "loss": 0.9435, + "step": 46287 + }, + { + "epoch": 2.98, + "learning_rate": 1.0088666566478458e-05, + "loss": 0.939, + "step": 46288 + }, + { + "epoch": 2.98, + "learning_rate": 1.0088451877455507e-05, + "loss": 0.9328, + "step": 46289 + }, + { + "epoch": 2.98, + "learning_rate": 1.0088237188432558e-05, + "loss": 0.839, + "step": 46290 + }, + { + "epoch": 2.98, + "learning_rate": 1.0088022499409607e-05, + "loss": 0.8294, + "step": 46291 + }, + { + "epoch": 2.98, + "learning_rate": 1.0087807810386656e-05, + "loss": 0.9182, + "step": 46292 + }, + { + "epoch": 2.98, + "learning_rate": 1.0087593121363706e-05, + "loss": 0.8169, + "step": 46293 + }, + { + "epoch": 2.98, + "learning_rate": 1.0087378432340755e-05, + "loss": 0.9684, + "step": 46294 + }, + { + "epoch": 2.98, + "learning_rate": 1.0087163743317804e-05, + "loss": 0.9181, + "step": 46295 + }, + { + "epoch": 2.98, + "learning_rate": 1.0086949054294855e-05, + "loss": 0.8765, + "step": 46296 + }, + { + "epoch": 2.98, + "learning_rate": 1.0086734365271904e-05, + "loss": 0.8169, + "step": 46297 + }, + { + "epoch": 2.98, + "learning_rate": 1.0086519676248953e-05, + "loss": 0.9115, + "step": 46298 + }, + { + "epoch": 2.98, + "learning_rate": 1.0086304987226004e-05, + "loss": 0.8655, + "step": 46299 + }, + { + "epoch": 2.98, + "learning_rate": 1.0086090298203052e-05, + "loss": 0.8442, + "step": 46300 + }, + { + "epoch": 2.98, + "learning_rate": 1.0085875609180103e-05, + "loss": 0.8725, + "step": 46301 + }, + { + "epoch": 2.98, + "learning_rate": 1.0085660920157154e-05, + "loss": 0.8653, + "step": 46302 + }, + { + "epoch": 2.98, + "learning_rate": 1.0085446231134204e-05, + "loss": 0.8459, + "step": 46303 + }, + { + "epoch": 2.98, + "learning_rate": 1.0085231542111253e-05, + "loss": 0.9472, + "step": 46304 + }, + { + "epoch": 2.98, + "learning_rate": 1.0085016853088302e-05, + "loss": 0.9082, + "step": 46305 + }, + { + "epoch": 2.98, + "learning_rate": 1.0084802164065353e-05, + "loss": 0.8822, + "step": 46306 + }, + { + "epoch": 2.98, + "learning_rate": 1.0084587475042402e-05, + "loss": 0.85, + "step": 46307 + }, + { + "epoch": 2.98, + "learning_rate": 1.0084372786019451e-05, + "loss": 0.9069, + "step": 46308 + }, + { + "epoch": 2.98, + "learning_rate": 1.0084158096996502e-05, + "loss": 0.9357, + "step": 46309 + }, + { + "epoch": 2.98, + "learning_rate": 1.008394340797355e-05, + "loss": 0.9162, + "step": 46310 + }, + { + "epoch": 2.98, + "learning_rate": 1.0083728718950601e-05, + "loss": 0.9151, + "step": 46311 + }, + { + "epoch": 2.98, + "learning_rate": 1.008351402992765e-05, + "loss": 0.9532, + "step": 46312 + }, + { + "epoch": 2.98, + "learning_rate": 1.00832993409047e-05, + "loss": 0.8753, + "step": 46313 + }, + { + "epoch": 2.98, + "learning_rate": 1.008308465188175e-05, + "loss": 0.8825, + "step": 46314 + }, + { + "epoch": 2.98, + "learning_rate": 1.0082869962858799e-05, + "loss": 0.9617, + "step": 46315 + }, + { + "epoch": 2.98, + "learning_rate": 1.0082655273835848e-05, + "loss": 0.9496, + "step": 46316 + }, + { + "epoch": 2.98, + "learning_rate": 1.00824405848129e-05, + "loss": 0.8535, + "step": 46317 + }, + { + "epoch": 2.98, + "learning_rate": 1.0082225895789949e-05, + "loss": 0.8444, + "step": 46318 + }, + { + "epoch": 2.98, + "learning_rate": 1.0082011206767e-05, + "loss": 0.9618, + "step": 46319 + }, + { + "epoch": 2.98, + "learning_rate": 1.0081796517744049e-05, + "loss": 0.9397, + "step": 46320 + }, + { + "epoch": 2.98, + "learning_rate": 1.00815818287211e-05, + "loss": 0.8696, + "step": 46321 + }, + { + "epoch": 2.98, + "learning_rate": 1.0081367139698148e-05, + "loss": 0.9526, + "step": 46322 + }, + { + "epoch": 2.98, + "learning_rate": 1.0081152450675197e-05, + "loss": 0.958, + "step": 46323 + }, + { + "epoch": 2.98, + "learning_rate": 1.0080937761652248e-05, + "loss": 0.909, + "step": 46324 + }, + { + "epoch": 2.98, + "learning_rate": 1.0080723072629297e-05, + "loss": 0.9123, + "step": 46325 + }, + { + "epoch": 2.98, + "learning_rate": 1.0080508383606346e-05, + "loss": 0.9234, + "step": 46326 + }, + { + "epoch": 2.98, + "learning_rate": 1.0080293694583397e-05, + "loss": 0.9468, + "step": 46327 + }, + { + "epoch": 2.98, + "learning_rate": 1.0080079005560446e-05, + "loss": 0.8548, + "step": 46328 + }, + { + "epoch": 2.98, + "learning_rate": 1.0079864316537496e-05, + "loss": 0.8547, + "step": 46329 + }, + { + "epoch": 2.98, + "learning_rate": 1.0079649627514545e-05, + "loss": 0.9473, + "step": 46330 + }, + { + "epoch": 2.98, + "learning_rate": 1.0079434938491597e-05, + "loss": 0.9016, + "step": 46331 + }, + { + "epoch": 2.98, + "learning_rate": 1.0079220249468646e-05, + "loss": 0.9564, + "step": 46332 + }, + { + "epoch": 2.98, + "learning_rate": 1.0079005560445695e-05, + "loss": 0.9529, + "step": 46333 + }, + { + "epoch": 2.98, + "learning_rate": 1.0078790871422746e-05, + "loss": 0.9313, + "step": 46334 + }, + { + "epoch": 2.98, + "learning_rate": 1.0078576182399795e-05, + "loss": 0.8649, + "step": 46335 + }, + { + "epoch": 2.98, + "learning_rate": 1.0078361493376844e-05, + "loss": 0.8874, + "step": 46336 + }, + { + "epoch": 2.98, + "learning_rate": 1.0078146804353895e-05, + "loss": 0.8839, + "step": 46337 + }, + { + "epoch": 2.98, + "learning_rate": 1.0077932115330944e-05, + "loss": 0.917, + "step": 46338 + }, + { + "epoch": 2.98, + "learning_rate": 1.0077717426307994e-05, + "loss": 0.8574, + "step": 46339 + }, + { + "epoch": 2.98, + "learning_rate": 1.0077502737285043e-05, + "loss": 0.9348, + "step": 46340 + }, + { + "epoch": 2.98, + "learning_rate": 1.0077288048262092e-05, + "loss": 0.9152, + "step": 46341 + }, + { + "epoch": 2.98, + "learning_rate": 1.0077073359239143e-05, + "loss": 0.929, + "step": 46342 + }, + { + "epoch": 2.98, + "learning_rate": 1.0076858670216192e-05, + "loss": 0.8458, + "step": 46343 + }, + { + "epoch": 2.98, + "learning_rate": 1.007664398119324e-05, + "loss": 0.8456, + "step": 46344 + }, + { + "epoch": 2.98, + "learning_rate": 1.0076429292170291e-05, + "loss": 0.8586, + "step": 46345 + }, + { + "epoch": 2.98, + "learning_rate": 1.0076214603147342e-05, + "loss": 0.9004, + "step": 46346 + }, + { + "epoch": 2.98, + "learning_rate": 1.0075999914124393e-05, + "loss": 0.911, + "step": 46347 + }, + { + "epoch": 2.98, + "learning_rate": 1.0075785225101442e-05, + "loss": 0.9078, + "step": 46348 + }, + { + "epoch": 2.98, + "learning_rate": 1.0075570536078492e-05, + "loss": 0.9054, + "step": 46349 + }, + { + "epoch": 2.98, + "learning_rate": 1.0075355847055541e-05, + "loss": 0.8657, + "step": 46350 + }, + { + "epoch": 2.98, + "learning_rate": 1.007514115803259e-05, + "loss": 0.869, + "step": 46351 + }, + { + "epoch": 2.98, + "learning_rate": 1.0074926469009641e-05, + "loss": 0.8266, + "step": 46352 + }, + { + "epoch": 2.98, + "learning_rate": 1.007471177998669e-05, + "loss": 0.8407, + "step": 46353 + }, + { + "epoch": 2.98, + "learning_rate": 1.0074497090963739e-05, + "loss": 0.8992, + "step": 46354 + }, + { + "epoch": 2.98, + "learning_rate": 1.007428240194079e-05, + "loss": 0.9024, + "step": 46355 + }, + { + "epoch": 2.98, + "learning_rate": 1.0074067712917839e-05, + "loss": 0.9309, + "step": 46356 + }, + { + "epoch": 2.98, + "learning_rate": 1.007385302389489e-05, + "loss": 0.9658, + "step": 46357 + }, + { + "epoch": 2.98, + "learning_rate": 1.0073638334871938e-05, + "loss": 0.9344, + "step": 46358 + }, + { + "epoch": 2.98, + "learning_rate": 1.0073423645848987e-05, + "loss": 0.8958, + "step": 46359 + }, + { + "epoch": 2.98, + "learning_rate": 1.007320895682604e-05, + "loss": 0.8989, + "step": 46360 + }, + { + "epoch": 2.98, + "learning_rate": 1.0072994267803088e-05, + "loss": 0.8865, + "step": 46361 + }, + { + "epoch": 2.98, + "learning_rate": 1.0072779578780139e-05, + "loss": 0.9217, + "step": 46362 + }, + { + "epoch": 2.98, + "learning_rate": 1.0072564889757188e-05, + "loss": 0.9066, + "step": 46363 + }, + { + "epoch": 2.98, + "learning_rate": 1.0072350200734237e-05, + "loss": 0.8678, + "step": 46364 + }, + { + "epoch": 2.98, + "learning_rate": 1.0072135511711288e-05, + "loss": 0.8988, + "step": 46365 + }, + { + "epoch": 2.98, + "learning_rate": 1.0071920822688337e-05, + "loss": 0.863, + "step": 46366 + }, + { + "epoch": 2.98, + "learning_rate": 1.0071706133665387e-05, + "loss": 0.9619, + "step": 46367 + }, + { + "epoch": 2.98, + "learning_rate": 1.0071491444642436e-05, + "loss": 0.8928, + "step": 46368 + }, + { + "epoch": 2.98, + "learning_rate": 1.0071276755619485e-05, + "loss": 0.8978, + "step": 46369 + }, + { + "epoch": 2.98, + "learning_rate": 1.0071062066596536e-05, + "loss": 0.8998, + "step": 46370 + }, + { + "epoch": 2.98, + "learning_rate": 1.0070847377573585e-05, + "loss": 0.9215, + "step": 46371 + }, + { + "epoch": 2.98, + "learning_rate": 1.0070632688550634e-05, + "loss": 0.8978, + "step": 46372 + }, + { + "epoch": 2.98, + "learning_rate": 1.0070417999527685e-05, + "loss": 0.9289, + "step": 46373 + }, + { + "epoch": 2.98, + "learning_rate": 1.0070203310504733e-05, + "loss": 0.9028, + "step": 46374 + }, + { + "epoch": 2.98, + "learning_rate": 1.0069988621481786e-05, + "loss": 0.9147, + "step": 46375 + }, + { + "epoch": 2.98, + "learning_rate": 1.0069773932458835e-05, + "loss": 0.8851, + "step": 46376 + }, + { + "epoch": 2.98, + "learning_rate": 1.0069559243435885e-05, + "loss": 0.9204, + "step": 46377 + }, + { + "epoch": 2.98, + "learning_rate": 1.0069344554412934e-05, + "loss": 0.9259, + "step": 46378 + }, + { + "epoch": 2.98, + "learning_rate": 1.0069129865389983e-05, + "loss": 0.8987, + "step": 46379 + }, + { + "epoch": 2.98, + "learning_rate": 1.0068915176367034e-05, + "loss": 0.9517, + "step": 46380 + }, + { + "epoch": 2.98, + "learning_rate": 1.0068700487344083e-05, + "loss": 0.9236, + "step": 46381 + }, + { + "epoch": 2.98, + "learning_rate": 1.0068485798321132e-05, + "loss": 0.9461, + "step": 46382 + }, + { + "epoch": 2.98, + "learning_rate": 1.0068271109298183e-05, + "loss": 0.8626, + "step": 46383 + }, + { + "epoch": 2.98, + "learning_rate": 1.0068056420275232e-05, + "loss": 0.9062, + "step": 46384 + }, + { + "epoch": 2.98, + "learning_rate": 1.006784173125228e-05, + "loss": 0.9484, + "step": 46385 + }, + { + "epoch": 2.98, + "learning_rate": 1.0067627042229331e-05, + "loss": 0.8775, + "step": 46386 + }, + { + "epoch": 2.98, + "learning_rate": 1.006741235320638e-05, + "loss": 0.8871, + "step": 46387 + }, + { + "epoch": 2.98, + "learning_rate": 1.0067197664183431e-05, + "loss": 0.9061, + "step": 46388 + }, + { + "epoch": 2.98, + "learning_rate": 1.0066982975160482e-05, + "loss": 0.8529, + "step": 46389 + }, + { + "epoch": 2.98, + "learning_rate": 1.0066768286137532e-05, + "loss": 0.8925, + "step": 46390 + }, + { + "epoch": 2.98, + "learning_rate": 1.0066553597114581e-05, + "loss": 0.8932, + "step": 46391 + }, + { + "epoch": 2.98, + "learning_rate": 1.006633890809163e-05, + "loss": 0.9089, + "step": 46392 + }, + { + "epoch": 2.98, + "learning_rate": 1.006612421906868e-05, + "loss": 0.9458, + "step": 46393 + }, + { + "epoch": 2.98, + "learning_rate": 1.006590953004573e-05, + "loss": 0.9451, + "step": 46394 + }, + { + "epoch": 2.98, + "learning_rate": 1.0065694841022779e-05, + "loss": 0.9085, + "step": 46395 + }, + { + "epoch": 2.99, + "learning_rate": 1.006548015199983e-05, + "loss": 0.9154, + "step": 46396 + }, + { + "epoch": 2.99, + "learning_rate": 1.0065265462976878e-05, + "loss": 0.8862, + "step": 46397 + }, + { + "epoch": 2.99, + "learning_rate": 1.0065050773953929e-05, + "loss": 0.8812, + "step": 46398 + }, + { + "epoch": 2.99, + "learning_rate": 1.0064836084930978e-05, + "loss": 0.9358, + "step": 46399 + }, + { + "epoch": 2.99, + "learning_rate": 1.0064621395908027e-05, + "loss": 0.8934, + "step": 46400 + }, + { + "epoch": 2.99, + "learning_rate": 1.0064406706885078e-05, + "loss": 0.8845, + "step": 46401 + }, + { + "epoch": 2.99, + "learning_rate": 1.0064192017862127e-05, + "loss": 0.8716, + "step": 46402 + }, + { + "epoch": 2.99, + "learning_rate": 1.0063977328839175e-05, + "loss": 0.8286, + "step": 46403 + }, + { + "epoch": 2.99, + "learning_rate": 1.0063762639816228e-05, + "loss": 0.8651, + "step": 46404 + }, + { + "epoch": 2.99, + "learning_rate": 1.0063547950793277e-05, + "loss": 0.9188, + "step": 46405 + }, + { + "epoch": 2.99, + "learning_rate": 1.0063333261770327e-05, + "loss": 0.8909, + "step": 46406 + }, + { + "epoch": 2.99, + "learning_rate": 1.0063118572747376e-05, + "loss": 0.8761, + "step": 46407 + }, + { + "epoch": 2.99, + "learning_rate": 1.0062903883724427e-05, + "loss": 0.8761, + "step": 46408 + }, + { + "epoch": 2.99, + "learning_rate": 1.0062689194701476e-05, + "loss": 0.8913, + "step": 46409 + }, + { + "epoch": 2.99, + "learning_rate": 1.0062474505678525e-05, + "loss": 0.9292, + "step": 46410 + }, + { + "epoch": 2.99, + "learning_rate": 1.0062259816655576e-05, + "loss": 0.8798, + "step": 46411 + }, + { + "epoch": 2.99, + "learning_rate": 1.0062045127632625e-05, + "loss": 0.8728, + "step": 46412 + }, + { + "epoch": 2.99, + "learning_rate": 1.0061830438609674e-05, + "loss": 0.8555, + "step": 46413 + }, + { + "epoch": 2.99, + "learning_rate": 1.0061615749586724e-05, + "loss": 0.8801, + "step": 46414 + }, + { + "epoch": 2.99, + "learning_rate": 1.0061401060563773e-05, + "loss": 0.8907, + "step": 46415 + }, + { + "epoch": 2.99, + "learning_rate": 1.0061186371540824e-05, + "loss": 0.8629, + "step": 46416 + }, + { + "epoch": 2.99, + "learning_rate": 1.0060971682517873e-05, + "loss": 0.9133, + "step": 46417 + }, + { + "epoch": 2.99, + "learning_rate": 1.0060756993494925e-05, + "loss": 0.8967, + "step": 46418 + }, + { + "epoch": 2.99, + "learning_rate": 1.0060542304471974e-05, + "loss": 0.8778, + "step": 46419 + }, + { + "epoch": 2.99, + "learning_rate": 1.0060327615449023e-05, + "loss": 0.8833, + "step": 46420 + }, + { + "epoch": 2.99, + "learning_rate": 1.0060112926426074e-05, + "loss": 0.9334, + "step": 46421 + }, + { + "epoch": 2.99, + "learning_rate": 1.0059898237403123e-05, + "loss": 0.9393, + "step": 46422 + }, + { + "epoch": 2.99, + "learning_rate": 1.0059683548380172e-05, + "loss": 0.873, + "step": 46423 + }, + { + "epoch": 2.99, + "learning_rate": 1.0059468859357222e-05, + "loss": 0.938, + "step": 46424 + }, + { + "epoch": 2.99, + "learning_rate": 1.0059254170334271e-05, + "loss": 0.904, + "step": 46425 + }, + { + "epoch": 2.99, + "learning_rate": 1.0059039481311322e-05, + "loss": 0.8826, + "step": 46426 + }, + { + "epoch": 2.99, + "learning_rate": 1.0058824792288371e-05, + "loss": 0.8559, + "step": 46427 + }, + { + "epoch": 2.99, + "learning_rate": 1.005861010326542e-05, + "loss": 0.8315, + "step": 46428 + }, + { + "epoch": 2.99, + "learning_rate": 1.005839541424247e-05, + "loss": 0.8816, + "step": 46429 + }, + { + "epoch": 2.99, + "learning_rate": 1.005818072521952e-05, + "loss": 0.9724, + "step": 46430 + }, + { + "epoch": 2.99, + "learning_rate": 1.0057966036196569e-05, + "loss": 0.8803, + "step": 46431 + }, + { + "epoch": 2.99, + "learning_rate": 1.005775134717362e-05, + "loss": 0.9131, + "step": 46432 + }, + { + "epoch": 2.99, + "learning_rate": 1.005753665815067e-05, + "loss": 0.8906, + "step": 46433 + }, + { + "epoch": 2.99, + "learning_rate": 1.005732196912772e-05, + "loss": 0.945, + "step": 46434 + }, + { + "epoch": 2.99, + "learning_rate": 1.005710728010477e-05, + "loss": 0.9027, + "step": 46435 + }, + { + "epoch": 2.99, + "learning_rate": 1.005689259108182e-05, + "loss": 0.8853, + "step": 46436 + }, + { + "epoch": 2.99, + "learning_rate": 1.0056677902058869e-05, + "loss": 0.8782, + "step": 46437 + }, + { + "epoch": 2.99, + "learning_rate": 1.0056463213035918e-05, + "loss": 1.0061, + "step": 46438 + }, + { + "epoch": 2.99, + "learning_rate": 1.0056248524012969e-05, + "loss": 0.8672, + "step": 46439 + }, + { + "epoch": 2.99, + "learning_rate": 1.0056033834990018e-05, + "loss": 0.9156, + "step": 46440 + }, + { + "epoch": 2.99, + "learning_rate": 1.0055819145967067e-05, + "loss": 0.8728, + "step": 46441 + }, + { + "epoch": 2.99, + "learning_rate": 1.0055604456944117e-05, + "loss": 0.8246, + "step": 46442 + }, + { + "epoch": 2.99, + "learning_rate": 1.0055389767921166e-05, + "loss": 0.8661, + "step": 46443 + }, + { + "epoch": 2.99, + "learning_rate": 1.0055175078898217e-05, + "loss": 0.9217, + "step": 46444 + }, + { + "epoch": 2.99, + "learning_rate": 1.0054960389875266e-05, + "loss": 0.9086, + "step": 46445 + }, + { + "epoch": 2.99, + "learning_rate": 1.0054745700852315e-05, + "loss": 0.9007, + "step": 46446 + }, + { + "epoch": 2.99, + "learning_rate": 1.0054531011829367e-05, + "loss": 0.9349, + "step": 46447 + }, + { + "epoch": 2.99, + "learning_rate": 1.0054316322806416e-05, + "loss": 0.9032, + "step": 46448 + }, + { + "epoch": 2.99, + "learning_rate": 1.0054101633783467e-05, + "loss": 0.9272, + "step": 46449 + }, + { + "epoch": 2.99, + "learning_rate": 1.0053886944760516e-05, + "loss": 0.9282, + "step": 46450 + }, + { + "epoch": 2.99, + "learning_rate": 1.0053672255737565e-05, + "loss": 0.8794, + "step": 46451 + }, + { + "epoch": 2.99, + "learning_rate": 1.0053457566714615e-05, + "loss": 0.9304, + "step": 46452 + }, + { + "epoch": 2.99, + "learning_rate": 1.0053242877691664e-05, + "loss": 0.856, + "step": 46453 + }, + { + "epoch": 2.99, + "learning_rate": 1.0053028188668715e-05, + "loss": 0.8823, + "step": 46454 + }, + { + "epoch": 2.99, + "learning_rate": 1.0052813499645764e-05, + "loss": 0.8917, + "step": 46455 + }, + { + "epoch": 2.99, + "learning_rate": 1.0052598810622813e-05, + "loss": 0.9195, + "step": 46456 + }, + { + "epoch": 2.99, + "learning_rate": 1.0052384121599864e-05, + "loss": 1.0, + "step": 46457 + }, + { + "epoch": 2.99, + "learning_rate": 1.0052169432576913e-05, + "loss": 0.8832, + "step": 46458 + }, + { + "epoch": 2.99, + "learning_rate": 1.0051954743553962e-05, + "loss": 0.9084, + "step": 46459 + }, + { + "epoch": 2.99, + "learning_rate": 1.0051740054531012e-05, + "loss": 0.9326, + "step": 46460 + }, + { + "epoch": 2.99, + "learning_rate": 1.0051525365508061e-05, + "loss": 0.9179, + "step": 46461 + }, + { + "epoch": 2.99, + "learning_rate": 1.0051310676485114e-05, + "loss": 0.898, + "step": 46462 + }, + { + "epoch": 2.99, + "learning_rate": 1.0051095987462163e-05, + "loss": 0.8927, + "step": 46463 + }, + { + "epoch": 2.99, + "learning_rate": 1.0050881298439213e-05, + "loss": 0.9027, + "step": 46464 + }, + { + "epoch": 2.99, + "learning_rate": 1.0050666609416262e-05, + "loss": 0.8426, + "step": 46465 + }, + { + "epoch": 2.99, + "learning_rate": 1.0050451920393311e-05, + "loss": 0.914, + "step": 46466 + }, + { + "epoch": 2.99, + "learning_rate": 1.0050237231370362e-05, + "loss": 0.8201, + "step": 46467 + }, + { + "epoch": 2.99, + "learning_rate": 1.005002254234741e-05, + "loss": 0.9199, + "step": 46468 + }, + { + "epoch": 2.99, + "learning_rate": 1.004980785332446e-05, + "loss": 0.8744, + "step": 46469 + }, + { + "epoch": 2.99, + "learning_rate": 1.004959316430151e-05, + "loss": 0.9095, + "step": 46470 + }, + { + "epoch": 2.99, + "learning_rate": 1.004937847527856e-05, + "loss": 0.9225, + "step": 46471 + }, + { + "epoch": 2.99, + "learning_rate": 1.0049163786255608e-05, + "loss": 0.894, + "step": 46472 + }, + { + "epoch": 2.99, + "learning_rate": 1.0048949097232659e-05, + "loss": 0.8396, + "step": 46473 + }, + { + "epoch": 2.99, + "learning_rate": 1.0048734408209708e-05, + "loss": 0.8995, + "step": 46474 + }, + { + "epoch": 2.99, + "learning_rate": 1.0048519719186759e-05, + "loss": 0.9239, + "step": 46475 + }, + { + "epoch": 2.99, + "learning_rate": 1.004830503016381e-05, + "loss": 0.8984, + "step": 46476 + }, + { + "epoch": 2.99, + "learning_rate": 1.004809034114086e-05, + "loss": 0.8993, + "step": 46477 + }, + { + "epoch": 2.99, + "learning_rate": 1.0047875652117909e-05, + "loss": 0.8614, + "step": 46478 + }, + { + "epoch": 2.99, + "learning_rate": 1.0047660963094958e-05, + "loss": 0.8578, + "step": 46479 + }, + { + "epoch": 2.99, + "learning_rate": 1.0047446274072008e-05, + "loss": 0.9435, + "step": 46480 + }, + { + "epoch": 2.99, + "learning_rate": 1.0047231585049057e-05, + "loss": 0.886, + "step": 46481 + }, + { + "epoch": 2.99, + "learning_rate": 1.0047016896026106e-05, + "loss": 0.9238, + "step": 46482 + }, + { + "epoch": 2.99, + "learning_rate": 1.0046802207003157e-05, + "loss": 0.8635, + "step": 46483 + }, + { + "epoch": 2.99, + "learning_rate": 1.0046587517980206e-05, + "loss": 0.9007, + "step": 46484 + }, + { + "epoch": 2.99, + "learning_rate": 1.0046372828957257e-05, + "loss": 0.8531, + "step": 46485 + }, + { + "epoch": 2.99, + "learning_rate": 1.0046158139934306e-05, + "loss": 0.8841, + "step": 46486 + }, + { + "epoch": 2.99, + "learning_rate": 1.0045943450911355e-05, + "loss": 0.8883, + "step": 46487 + }, + { + "epoch": 2.99, + "learning_rate": 1.0045728761888405e-05, + "loss": 0.9268, + "step": 46488 + }, + { + "epoch": 2.99, + "learning_rate": 1.0045514072865454e-05, + "loss": 0.9004, + "step": 46489 + }, + { + "epoch": 2.99, + "learning_rate": 1.0045299383842503e-05, + "loss": 0.8798, + "step": 46490 + }, + { + "epoch": 2.99, + "learning_rate": 1.0045084694819556e-05, + "loss": 0.9689, + "step": 46491 + }, + { + "epoch": 2.99, + "learning_rate": 1.0044870005796605e-05, + "loss": 0.8538, + "step": 46492 + }, + { + "epoch": 2.99, + "learning_rate": 1.0044655316773655e-05, + "loss": 0.9071, + "step": 46493 + }, + { + "epoch": 2.99, + "learning_rate": 1.0044440627750704e-05, + "loss": 0.9101, + "step": 46494 + }, + { + "epoch": 2.99, + "learning_rate": 1.0044225938727755e-05, + "loss": 0.8787, + "step": 46495 + }, + { + "epoch": 2.99, + "learning_rate": 1.0044011249704804e-05, + "loss": 0.9381, + "step": 46496 + }, + { + "epoch": 2.99, + "learning_rate": 1.0043796560681853e-05, + "loss": 0.9259, + "step": 46497 + }, + { + "epoch": 2.99, + "learning_rate": 1.0043581871658903e-05, + "loss": 0.9307, + "step": 46498 + }, + { + "epoch": 2.99, + "learning_rate": 1.0043367182635952e-05, + "loss": 0.8843, + "step": 46499 + }, + { + "epoch": 2.99, + "learning_rate": 1.0043152493613001e-05, + "loss": 0.8866, + "step": 46500 + }, + { + "epoch": 2.99, + "learning_rate": 1.0042937804590052e-05, + "loss": 0.8454, + "step": 46501 + }, + { + "epoch": 2.99, + "learning_rate": 1.0042723115567101e-05, + "loss": 0.9675, + "step": 46502 + }, + { + "epoch": 2.99, + "learning_rate": 1.0042508426544152e-05, + "loss": 0.8561, + "step": 46503 + }, + { + "epoch": 2.99, + "learning_rate": 1.00422937375212e-05, + "loss": 0.949, + "step": 46504 + }, + { + "epoch": 2.99, + "learning_rate": 1.0042079048498253e-05, + "loss": 0.8789, + "step": 46505 + }, + { + "epoch": 2.99, + "learning_rate": 1.0041864359475302e-05, + "loss": 0.8857, + "step": 46506 + }, + { + "epoch": 2.99, + "learning_rate": 1.0041649670452351e-05, + "loss": 0.9039, + "step": 46507 + }, + { + "epoch": 2.99, + "learning_rate": 1.0041434981429402e-05, + "loss": 0.8011, + "step": 46508 + }, + { + "epoch": 2.99, + "learning_rate": 1.004122029240645e-05, + "loss": 0.8506, + "step": 46509 + }, + { + "epoch": 2.99, + "learning_rate": 1.00410056033835e-05, + "loss": 0.9204, + "step": 46510 + }, + { + "epoch": 2.99, + "learning_rate": 1.004079091436055e-05, + "loss": 0.9339, + "step": 46511 + }, + { + "epoch": 2.99, + "learning_rate": 1.0040576225337599e-05, + "loss": 0.8547, + "step": 46512 + }, + { + "epoch": 2.99, + "learning_rate": 1.004036153631465e-05, + "loss": 0.9133, + "step": 46513 + }, + { + "epoch": 2.99, + "learning_rate": 1.0040146847291699e-05, + "loss": 0.9376, + "step": 46514 + }, + { + "epoch": 2.99, + "learning_rate": 1.0039932158268748e-05, + "loss": 0.8224, + "step": 46515 + }, + { + "epoch": 2.99, + "learning_rate": 1.0039717469245798e-05, + "loss": 0.9243, + "step": 46516 + }, + { + "epoch": 2.99, + "learning_rate": 1.0039502780222847e-05, + "loss": 0.8982, + "step": 46517 + }, + { + "epoch": 2.99, + "learning_rate": 1.0039288091199896e-05, + "loss": 0.8519, + "step": 46518 + }, + { + "epoch": 2.99, + "learning_rate": 1.0039073402176947e-05, + "loss": 0.8931, + "step": 46519 + }, + { + "epoch": 2.99, + "learning_rate": 1.0038858713153998e-05, + "loss": 0.9352, + "step": 46520 + }, + { + "epoch": 2.99, + "learning_rate": 1.0038644024131048e-05, + "loss": 0.8521, + "step": 46521 + }, + { + "epoch": 2.99, + "learning_rate": 1.0038429335108097e-05, + "loss": 0.9332, + "step": 46522 + }, + { + "epoch": 2.99, + "learning_rate": 1.0038214646085148e-05, + "loss": 0.8861, + "step": 46523 + }, + { + "epoch": 2.99, + "learning_rate": 1.0037999957062197e-05, + "loss": 0.883, + "step": 46524 + }, + { + "epoch": 2.99, + "learning_rate": 1.0037785268039246e-05, + "loss": 0.916, + "step": 46525 + }, + { + "epoch": 2.99, + "learning_rate": 1.0037570579016296e-05, + "loss": 0.8121, + "step": 46526 + }, + { + "epoch": 2.99, + "learning_rate": 1.0037355889993345e-05, + "loss": 0.9346, + "step": 46527 + }, + { + "epoch": 2.99, + "learning_rate": 1.0037141200970394e-05, + "loss": 0.9086, + "step": 46528 + }, + { + "epoch": 2.99, + "learning_rate": 1.0036926511947445e-05, + "loss": 0.791, + "step": 46529 + }, + { + "epoch": 2.99, + "learning_rate": 1.0036711822924494e-05, + "loss": 0.9223, + "step": 46530 + }, + { + "epoch": 2.99, + "learning_rate": 1.0036497133901543e-05, + "loss": 0.9074, + "step": 46531 + }, + { + "epoch": 2.99, + "learning_rate": 1.0036282444878594e-05, + "loss": 0.8577, + "step": 46532 + }, + { + "epoch": 2.99, + "learning_rate": 1.0036067755855643e-05, + "loss": 0.8986, + "step": 46533 + }, + { + "epoch": 2.99, + "learning_rate": 1.0035853066832693e-05, + "loss": 0.9797, + "step": 46534 + }, + { + "epoch": 2.99, + "learning_rate": 1.0035638377809744e-05, + "loss": 0.9293, + "step": 46535 + }, + { + "epoch": 2.99, + "learning_rate": 1.0035423688786795e-05, + "loss": 0.9412, + "step": 46536 + }, + { + "epoch": 2.99, + "learning_rate": 1.0035208999763844e-05, + "loss": 0.9349, + "step": 46537 + }, + { + "epoch": 2.99, + "learning_rate": 1.0034994310740892e-05, + "loss": 0.8565, + "step": 46538 + }, + { + "epoch": 2.99, + "learning_rate": 1.0034779621717943e-05, + "loss": 0.8637, + "step": 46539 + }, + { + "epoch": 2.99, + "learning_rate": 1.0034564932694992e-05, + "loss": 0.8486, + "step": 46540 + }, + { + "epoch": 2.99, + "learning_rate": 1.0034350243672041e-05, + "loss": 0.9114, + "step": 46541 + }, + { + "epoch": 2.99, + "learning_rate": 1.0034135554649092e-05, + "loss": 0.9426, + "step": 46542 + }, + { + "epoch": 2.99, + "learning_rate": 1.003392086562614e-05, + "loss": 0.9566, + "step": 46543 + }, + { + "epoch": 2.99, + "learning_rate": 1.0033706176603191e-05, + "loss": 0.8746, + "step": 46544 + }, + { + "epoch": 2.99, + "learning_rate": 1.003349148758024e-05, + "loss": 0.8964, + "step": 46545 + }, + { + "epoch": 2.99, + "learning_rate": 1.003327679855729e-05, + "loss": 0.8993, + "step": 46546 + }, + { + "epoch": 2.99, + "learning_rate": 1.003306210953434e-05, + "loss": 0.8649, + "step": 46547 + }, + { + "epoch": 2.99, + "learning_rate": 1.0032847420511389e-05, + "loss": 0.904, + "step": 46548 + }, + { + "epoch": 2.99, + "learning_rate": 1.0032632731488441e-05, + "loss": 0.9027, + "step": 46549 + }, + { + "epoch": 2.99, + "learning_rate": 1.003241804246549e-05, + "loss": 0.9372, + "step": 46550 + }, + { + "epoch": 2.99, + "learning_rate": 1.0032203353442541e-05, + "loss": 0.9192, + "step": 46551 + }, + { + "epoch": 3.0, + "learning_rate": 1.003198866441959e-05, + "loss": 0.9869, + "step": 46552 + }, + { + "epoch": 3.0, + "learning_rate": 1.0031773975396639e-05, + "loss": 0.8642, + "step": 46553 + }, + { + "epoch": 3.0, + "learning_rate": 1.003155928637369e-05, + "loss": 0.9211, + "step": 46554 + }, + { + "epoch": 3.0, + "learning_rate": 1.0031344597350738e-05, + "loss": 0.9047, + "step": 46555 + }, + { + "epoch": 3.0, + "learning_rate": 1.0031129908327787e-05, + "loss": 0.887, + "step": 46556 + }, + { + "epoch": 3.0, + "learning_rate": 1.0030915219304838e-05, + "loss": 0.8939, + "step": 46557 + }, + { + "epoch": 3.0, + "learning_rate": 1.0030700530281887e-05, + "loss": 0.9219, + "step": 46558 + }, + { + "epoch": 3.0, + "learning_rate": 1.0030485841258936e-05, + "loss": 0.9375, + "step": 46559 + }, + { + "epoch": 3.0, + "learning_rate": 1.0030271152235987e-05, + "loss": 0.8959, + "step": 46560 + }, + { + "epoch": 3.0, + "learning_rate": 1.0030056463213036e-05, + "loss": 0.878, + "step": 46561 + }, + { + "epoch": 3.0, + "learning_rate": 1.0029841774190086e-05, + "loss": 0.8933, + "step": 46562 + }, + { + "epoch": 3.0, + "learning_rate": 1.0029627085167135e-05, + "loss": 0.8557, + "step": 46563 + }, + { + "epoch": 3.0, + "learning_rate": 1.0029412396144188e-05, + "loss": 0.8768, + "step": 46564 + }, + { + "epoch": 3.0, + "learning_rate": 1.0029197707121237e-05, + "loss": 0.9157, + "step": 46565 + }, + { + "epoch": 3.0, + "learning_rate": 1.0028983018098286e-05, + "loss": 0.8765, + "step": 46566 + }, + { + "epoch": 3.0, + "learning_rate": 1.0028768329075336e-05, + "loss": 0.8508, + "step": 46567 + }, + { + "epoch": 3.0, + "learning_rate": 1.0028553640052385e-05, + "loss": 0.9293, + "step": 46568 + }, + { + "epoch": 3.0, + "learning_rate": 1.0028338951029434e-05, + "loss": 0.8422, + "step": 46569 + }, + { + "epoch": 3.0, + "learning_rate": 1.0028124262006485e-05, + "loss": 0.8522, + "step": 46570 + }, + { + "epoch": 3.0, + "learning_rate": 1.0027909572983534e-05, + "loss": 0.9238, + "step": 46571 + }, + { + "epoch": 3.0, + "learning_rate": 1.0027694883960584e-05, + "loss": 0.9352, + "step": 46572 + }, + { + "epoch": 3.0, + "learning_rate": 1.0027480194937633e-05, + "loss": 0.9194, + "step": 46573 + }, + { + "epoch": 3.0, + "learning_rate": 1.0027265505914682e-05, + "loss": 0.8517, + "step": 46574 + }, + { + "epoch": 3.0, + "learning_rate": 1.0027050816891733e-05, + "loss": 0.9371, + "step": 46575 + }, + { + "epoch": 3.0, + "learning_rate": 1.0026836127868782e-05, + "loss": 0.9048, + "step": 46576 + }, + { + "epoch": 3.0, + "learning_rate": 1.0026621438845831e-05, + "loss": 0.8082, + "step": 46577 + }, + { + "epoch": 3.0, + "learning_rate": 1.0026406749822883e-05, + "loss": 0.9422, + "step": 46578 + }, + { + "epoch": 3.0, + "learning_rate": 1.0026192060799932e-05, + "loss": 0.9369, + "step": 46579 + }, + { + "epoch": 3.0, + "learning_rate": 1.0025977371776983e-05, + "loss": 0.8556, + "step": 46580 + }, + { + "epoch": 3.0, + "learning_rate": 1.0025762682754032e-05, + "loss": 0.8691, + "step": 46581 + }, + { + "epoch": 3.0, + "learning_rate": 1.0025547993731083e-05, + "loss": 0.8421, + "step": 46582 + }, + { + "epoch": 3.0, + "learning_rate": 1.0025333304708131e-05, + "loss": 0.8984, + "step": 46583 + }, + { + "epoch": 3.0, + "learning_rate": 1.002511861568518e-05, + "loss": 0.8825, + "step": 46584 + }, + { + "epoch": 3.0, + "learning_rate": 1.0024903926662231e-05, + "loss": 0.8771, + "step": 46585 + }, + { + "epoch": 3.0, + "learning_rate": 1.002468923763928e-05, + "loss": 0.8806, + "step": 46586 + }, + { + "epoch": 3.0, + "learning_rate": 1.0024474548616329e-05, + "loss": 0.8802, + "step": 46587 + }, + { + "epoch": 3.0, + "learning_rate": 1.002425985959338e-05, + "loss": 0.8593, + "step": 46588 + }, + { + "epoch": 3.0, + "learning_rate": 1.0024045170570429e-05, + "loss": 0.8815, + "step": 46589 + }, + { + "epoch": 3.0, + "learning_rate": 1.002383048154748e-05, + "loss": 0.893, + "step": 46590 + }, + { + "epoch": 3.0, + "learning_rate": 1.002383048154748e-05, + "loss": 0.9015, + "step": 46591 + }, + { + "epoch": 3.0, + "learning_rate": 1.0023615792524528e-05, + "loss": 0.8728, + "step": 46592 + }, + { + "epoch": 3.0, + "learning_rate": 1.0023401103501577e-05, + "loss": 0.9132, + "step": 46593 + }, + { + "epoch": 3.0, + "learning_rate": 1.002318641447863e-05, + "loss": 0.8387, + "step": 46594 + }, + { + "epoch": 3.0, + "learning_rate": 1.0022971725455679e-05, + "loss": 0.8836, + "step": 46595 + }, + { + "epoch": 3.0, + "learning_rate": 1.002275703643273e-05, + "loss": 0.8805, + "step": 46596 + }, + { + "epoch": 3.0, + "learning_rate": 1.0022542347409778e-05, + "loss": 0.8917, + "step": 46597 + }, + { + "epoch": 3.0, + "learning_rate": 1.0022327658386827e-05, + "loss": 0.8638, + "step": 46598 + }, + { + "epoch": 3.0, + "learning_rate": 1.0022112969363878e-05, + "loss": 0.889, + "step": 46599 + }, + { + "epoch": 3.0, + "learning_rate": 1.0021898280340927e-05, + "loss": 0.9041, + "step": 46600 + }, + { + "epoch": 3.0, + "learning_rate": 1.0021683591317977e-05, + "loss": 0.9238, + "step": 46601 + }, + { + "epoch": 3.0, + "learning_rate": 1.0021468902295026e-05, + "loss": 0.8664, + "step": 46602 + }, + { + "epoch": 3.0, + "learning_rate": 1.0021254213272075e-05, + "loss": 0.9024, + "step": 46603 + }, + { + "epoch": 3.0, + "learning_rate": 1.0021039524249126e-05, + "loss": 0.8888, + "step": 46604 + }, + { + "epoch": 3.0, + "learning_rate": 1.0020824835226175e-05, + "loss": 0.8625, + "step": 46605 + }, + { + "epoch": 3.0, + "learning_rate": 1.0020610146203224e-05, + "loss": 0.8539, + "step": 46606 + }, + { + "epoch": 3.0, + "learning_rate": 1.0020395457180275e-05, + "loss": 0.8965, + "step": 46607 + }, + { + "epoch": 3.0, + "learning_rate": 1.0020180768157325e-05, + "loss": 0.8232, + "step": 46608 + }, + { + "epoch": 3.0, + "learning_rate": 1.0019966079134376e-05, + "loss": 0.9004, + "step": 46609 + }, + { + "epoch": 3.0, + "learning_rate": 1.0019751390111425e-05, + "loss": 0.9711, + "step": 46610 + }, + { + "epoch": 3.0, + "learning_rate": 1.0019536701088476e-05, + "loss": 0.8204, + "step": 46611 + }, + { + "epoch": 3.0, + "learning_rate": 1.0019322012065525e-05, + "loss": 0.9241, + "step": 46612 + }, + { + "epoch": 3.0, + "learning_rate": 1.0019107323042573e-05, + "loss": 0.9024, + "step": 46613 + }, + { + "epoch": 3.0, + "learning_rate": 1.0018892634019624e-05, + "loss": 0.8799, + "step": 46614 + }, + { + "epoch": 3.0, + "learning_rate": 1.0018677944996673e-05, + "loss": 0.8679, + "step": 46615 + }, + { + "epoch": 3.0, + "learning_rate": 1.0018463255973722e-05, + "loss": 0.9032, + "step": 46616 + }, + { + "epoch": 3.0, + "learning_rate": 1.0018248566950773e-05, + "loss": 0.8915, + "step": 46617 + }, + { + "epoch": 3.0, + "learning_rate": 1.0018033877927822e-05, + "loss": 0.8648, + "step": 46618 + }, + { + "epoch": 3.0, + "learning_rate": 1.001781918890487e-05, + "loss": 0.8748, + "step": 46619 + }, + { + "epoch": 3.0, + "learning_rate": 1.0017604499881921e-05, + "loss": 0.9142, + "step": 46620 + }, + { + "epoch": 3.0, + "learning_rate": 1.001738981085897e-05, + "loss": 0.9253, + "step": 46621 + }, + { + "epoch": 3.0, + "learning_rate": 1.0017175121836021e-05, + "loss": 0.9429, + "step": 46622 + }, + { + "epoch": 3.0, + "learning_rate": 1.0016960432813072e-05, + "loss": 0.9265, + "step": 46623 + }, + { + "epoch": 3.0, + "learning_rate": 1.0016745743790122e-05, + "loss": 0.8715, + "step": 46624 + }, + { + "epoch": 3.0, + "learning_rate": 1.0016531054767171e-05, + "loss": 0.8791, + "step": 46625 + }, + { + "epoch": 3.0, + "learning_rate": 1.001631636574422e-05, + "loss": 0.9042, + "step": 46626 + }, + { + "epoch": 3.0, + "learning_rate": 1.0016101676721271e-05, + "loss": 0.8752, + "step": 46627 + }, + { + "epoch": 3.0, + "learning_rate": 1.001588698769832e-05, + "loss": 0.8619, + "step": 46628 + }, + { + "epoch": 3.0, + "learning_rate": 1.0015672298675369e-05, + "loss": 0.8547, + "step": 46629 + }, + { + "epoch": 3.0, + "learning_rate": 1.001545760965242e-05, + "loss": 0.8034, + "step": 46630 + }, + { + "epoch": 3.0, + "learning_rate": 1.0015242920629468e-05, + "loss": 0.8331, + "step": 46631 + }, + { + "epoch": 3.0, + "learning_rate": 1.0015028231606519e-05, + "loss": 0.8937, + "step": 46632 + }, + { + "epoch": 3.0, + "learning_rate": 1.0014813542583568e-05, + "loss": 0.8262, + "step": 46633 + }, + { + "epoch": 3.0, + "learning_rate": 1.0014598853560617e-05, + "loss": 0.7754, + "step": 46634 + }, + { + "epoch": 3.0, + "learning_rate": 1.0014384164537668e-05, + "loss": 0.8164, + "step": 46635 + }, + { + "epoch": 3.0, + "learning_rate": 1.0014169475514717e-05, + "loss": 0.8243, + "step": 46636 + }, + { + "epoch": 3.0, + "learning_rate": 1.0013954786491769e-05, + "loss": 0.8189, + "step": 46637 + }, + { + "epoch": 3.0, + "learning_rate": 1.0013740097468818e-05, + "loss": 0.7728, + "step": 46638 + }, + { + "epoch": 3.0, + "learning_rate": 1.0013525408445867e-05, + "loss": 0.9125, + "step": 46639 + }, + { + "epoch": 3.0, + "learning_rate": 1.0013310719422918e-05, + "loss": 0.8693, + "step": 46640 + }, + { + "epoch": 3.0, + "learning_rate": 1.0013096030399967e-05, + "loss": 0.7595, + "step": 46641 + }, + { + "epoch": 3.0, + "learning_rate": 1.0012881341377017e-05, + "loss": 0.8556, + "step": 46642 + }, + { + "epoch": 3.0, + "learning_rate": 1.0012666652354066e-05, + "loss": 0.831, + "step": 46643 + }, + { + "epoch": 3.0, + "learning_rate": 1.0012451963331115e-05, + "loss": 0.849, + "step": 46644 + }, + { + "epoch": 3.0, + "learning_rate": 1.0012237274308166e-05, + "loss": 0.8415, + "step": 46645 + }, + { + "epoch": 3.0, + "learning_rate": 1.0012022585285215e-05, + "loss": 0.8383, + "step": 46646 + }, + { + "epoch": 3.0, + "learning_rate": 1.0011807896262264e-05, + "loss": 0.7899, + "step": 46647 + }, + { + "epoch": 3.0, + "learning_rate": 1.0011593207239314e-05, + "loss": 0.7831, + "step": 46648 + }, + { + "epoch": 3.0, + "learning_rate": 1.0011378518216363e-05, + "loss": 0.7709, + "step": 46649 + }, + { + "epoch": 3.0, + "learning_rate": 1.0011163829193414e-05, + "loss": 0.837, + "step": 46650 + }, + { + "epoch": 3.0, + "learning_rate": 1.0010949140170463e-05, + "loss": 0.8842, + "step": 46651 + }, + { + "epoch": 3.0, + "learning_rate": 1.0010734451147515e-05, + "loss": 0.8927, + "step": 46652 + }, + { + "epoch": 3.0, + "learning_rate": 1.0010519762124564e-05, + "loss": 0.8143, + "step": 46653 + }, + { + "epoch": 3.0, + "learning_rate": 1.0010305073101613e-05, + "loss": 0.8231, + "step": 46654 + }, + { + "epoch": 3.0, + "learning_rate": 1.0010090384078664e-05, + "loss": 0.8157, + "step": 46655 + }, + { + "epoch": 3.0, + "learning_rate": 1.0009875695055713e-05, + "loss": 0.8353, + "step": 46656 + }, + { + "epoch": 3.0, + "learning_rate": 1.0009661006032762e-05, + "loss": 0.799, + "step": 46657 + }, + { + "epoch": 3.0, + "learning_rate": 1.0009446317009812e-05, + "loss": 0.7766, + "step": 46658 + }, + { + "epoch": 3.0, + "learning_rate": 1.0009231627986861e-05, + "loss": 0.8436, + "step": 46659 + }, + { + "epoch": 3.0, + "learning_rate": 1.0009016938963912e-05, + "loss": 0.8308, + "step": 46660 + }, + { + "epoch": 3.0, + "learning_rate": 1.0008802249940961e-05, + "loss": 0.8353, + "step": 46661 + }, + { + "epoch": 3.0, + "learning_rate": 1.000858756091801e-05, + "loss": 0.865, + "step": 46662 + }, + { + "epoch": 3.0, + "learning_rate": 1.000837287189506e-05, + "loss": 0.893, + "step": 46663 + }, + { + "epoch": 3.0, + "learning_rate": 1.000815818287211e-05, + "loss": 0.8411, + "step": 46664 + }, + { + "epoch": 3.0, + "learning_rate": 1.0007943493849159e-05, + "loss": 0.8095, + "step": 46665 + }, + { + "epoch": 3.0, + "learning_rate": 1.0007728804826211e-05, + "loss": 0.835, + "step": 46666 + }, + { + "epoch": 3.0, + "learning_rate": 1.000751411580326e-05, + "loss": 0.8457, + "step": 46667 + }, + { + "epoch": 3.0, + "learning_rate": 1.000729942678031e-05, + "loss": 0.8313, + "step": 46668 + }, + { + "epoch": 3.0, + "learning_rate": 1.000708473775736e-05, + "loss": 0.8227, + "step": 46669 + }, + { + "epoch": 3.0, + "learning_rate": 1.000687004873441e-05, + "loss": 0.8634, + "step": 46670 + }, + { + "epoch": 3.0, + "learning_rate": 1.000665535971146e-05, + "loss": 0.8268, + "step": 46671 + }, + { + "epoch": 3.0, + "learning_rate": 1.0006440670688508e-05, + "loss": 0.8708, + "step": 46672 + }, + { + "epoch": 3.0, + "learning_rate": 1.0006225981665559e-05, + "loss": 0.7566, + "step": 46673 + }, + { + "epoch": 3.0, + "learning_rate": 1.0006011292642608e-05, + "loss": 0.8494, + "step": 46674 + }, + { + "epoch": 3.0, + "learning_rate": 1.0005796603619657e-05, + "loss": 0.8311, + "step": 46675 + }, + { + "epoch": 3.0, + "learning_rate": 1.0005581914596707e-05, + "loss": 0.859, + "step": 46676 + }, + { + "epoch": 3.0, + "learning_rate": 1.0005367225573756e-05, + "loss": 0.8325, + "step": 46677 + }, + { + "epoch": 3.0, + "learning_rate": 1.0005152536550807e-05, + "loss": 0.8223, + "step": 46678 + }, + { + "epoch": 3.0, + "learning_rate": 1.0004937847527856e-05, + "loss": 0.7472, + "step": 46679 + }, + { + "epoch": 3.0, + "learning_rate": 1.0004723158504905e-05, + "loss": 0.824, + "step": 46680 + }, + { + "epoch": 3.0, + "learning_rate": 1.0004508469481957e-05, + "loss": 0.7761, + "step": 46681 + }, + { + "epoch": 3.0, + "learning_rate": 1.0004293780459006e-05, + "loss": 0.8223, + "step": 46682 + }, + { + "epoch": 3.0, + "learning_rate": 1.0004079091436057e-05, + "loss": 0.8192, + "step": 46683 + }, + { + "epoch": 3.0, + "learning_rate": 1.0003864402413106e-05, + "loss": 0.8166, + "step": 46684 + }, + { + "epoch": 3.0, + "learning_rate": 1.0003649713390155e-05, + "loss": 0.842, + "step": 46685 + }, + { + "epoch": 3.0, + "learning_rate": 1.0003435024367206e-05, + "loss": 0.7925, + "step": 46686 + }, + { + "epoch": 3.0, + "learning_rate": 1.0003220335344254e-05, + "loss": 0.7858, + "step": 46687 + }, + { + "epoch": 3.0, + "learning_rate": 1.0003005646321305e-05, + "loss": 0.8006, + "step": 46688 + }, + { + "epoch": 3.0, + "learning_rate": 1.0002790957298354e-05, + "loss": 0.8657, + "step": 46689 + }, + { + "epoch": 3.0, + "learning_rate": 1.0002576268275403e-05, + "loss": 0.8065, + "step": 46690 + }, + { + "epoch": 3.0, + "learning_rate": 1.0002361579252454e-05, + "loss": 0.8308, + "step": 46691 + }, + { + "epoch": 3.0, + "learning_rate": 1.0002146890229503e-05, + "loss": 0.7525, + "step": 46692 + }, + { + "epoch": 3.0, + "learning_rate": 1.0001932201206552e-05, + "loss": 0.8261, + "step": 46693 + }, + { + "epoch": 3.0, + "learning_rate": 1.0001717512183602e-05, + "loss": 0.7813, + "step": 46694 + }, + { + "epoch": 3.0, + "learning_rate": 1.0001502823160653e-05, + "loss": 0.8117, + "step": 46695 + }, + { + "epoch": 3.0, + "learning_rate": 1.0001288134137704e-05, + "loss": 0.7951, + "step": 46696 + }, + { + "epoch": 3.0, + "learning_rate": 1.0001073445114753e-05, + "loss": 0.8579, + "step": 46697 + }, + { + "epoch": 3.0, + "learning_rate": 1.0000858756091803e-05, + "loss": 0.8036, + "step": 46698 + }, + { + "epoch": 3.0, + "learning_rate": 1.0000644067068852e-05, + "loss": 0.7425, + "step": 46699 + }, + { + "epoch": 3.0, + "learning_rate": 1.0000429378045901e-05, + "loss": 0.82, + "step": 46700 + }, + { + "epoch": 3.0, + "learning_rate": 1.0000214689022952e-05, + "loss": 0.8199, + "step": 46701 + }, + { + "epoch": 3.0, + "learning_rate": 1e-05, + "loss": 0.8468, + "step": 46702 + }, + { + "epoch": 3.0, + "learning_rate": 9.99978531097705e-06, + "loss": 0.8185, + "step": 46703 + }, + { + "epoch": 3.0, + "learning_rate": 9.9995706219541e-06, + "loss": 0.8232, + "step": 46704 + }, + { + "epoch": 3.0, + "learning_rate": 9.99935593293115e-06, + "loss": 0.8463, + "step": 46705 + }, + { + "epoch": 3.0, + "learning_rate": 9.9991412439082e-06, + "loss": 0.8424, + "step": 46706 + }, + { + "epoch": 3.01, + "learning_rate": 9.99892655488525e-06, + "loss": 0.796, + "step": 46707 + }, + { + "epoch": 3.01, + "learning_rate": 9.9987118658623e-06, + "loss": 0.7179, + "step": 46708 + }, + { + "epoch": 3.01, + "learning_rate": 9.998497176839349e-06, + "loss": 0.8847, + "step": 46709 + }, + { + "epoch": 3.01, + "learning_rate": 9.9982824878164e-06, + "loss": 0.8482, + "step": 46710 + }, + { + "epoch": 3.01, + "learning_rate": 9.998067798793448e-06, + "loss": 0.807, + "step": 46711 + }, + { + "epoch": 3.01, + "learning_rate": 9.997853109770497e-06, + "loss": 0.8688, + "step": 46712 + }, + { + "epoch": 3.01, + "learning_rate": 9.997638420747548e-06, + "loss": 0.7748, + "step": 46713 + }, + { + "epoch": 3.01, + "learning_rate": 9.997423731724599e-06, + "loss": 0.8148, + "step": 46714 + }, + { + "epoch": 3.01, + "learning_rate": 9.997209042701648e-06, + "loss": 0.7514, + "step": 46715 + }, + { + "epoch": 3.01, + "learning_rate": 9.996994353678696e-06, + "loss": 0.7867, + "step": 46716 + }, + { + "epoch": 3.01, + "learning_rate": 9.996779664655747e-06, + "loss": 0.8556, + "step": 46717 + }, + { + "epoch": 3.01, + "learning_rate": 9.996564975632796e-06, + "loss": 0.7892, + "step": 46718 + }, + { + "epoch": 3.01, + "learning_rate": 9.996350286609847e-06, + "loss": 0.8555, + "step": 46719 + }, + { + "epoch": 3.01, + "learning_rate": 9.996135597586896e-06, + "loss": 0.8457, + "step": 46720 + }, + { + "epoch": 3.01, + "learning_rate": 9.995920908563946e-06, + "loss": 0.8462, + "step": 46721 + }, + { + "epoch": 3.01, + "learning_rate": 9.995706219540995e-06, + "loss": 0.813, + "step": 46722 + }, + { + "epoch": 3.01, + "learning_rate": 9.995491530518046e-06, + "loss": 0.9241, + "step": 46723 + }, + { + "epoch": 3.01, + "learning_rate": 9.995276841495095e-06, + "loss": 0.7873, + "step": 46724 + }, + { + "epoch": 3.01, + "learning_rate": 9.995062152472144e-06, + "loss": 0.81, + "step": 46725 + }, + { + "epoch": 3.01, + "learning_rate": 9.994847463449195e-06, + "loss": 0.8573, + "step": 46726 + }, + { + "epoch": 3.01, + "learning_rate": 9.994632774426244e-06, + "loss": 0.7992, + "step": 46727 + }, + { + "epoch": 3.01, + "learning_rate": 9.994418085403294e-06, + "loss": 0.8008, + "step": 46728 + }, + { + "epoch": 3.01, + "learning_rate": 9.994203396380345e-06, + "loss": 0.8084, + "step": 46729 + }, + { + "epoch": 3.01, + "learning_rate": 9.993988707357394e-06, + "loss": 0.8407, + "step": 46730 + }, + { + "epoch": 3.01, + "learning_rate": 9.993774018334443e-06, + "loss": 0.7737, + "step": 46731 + }, + { + "epoch": 3.01, + "learning_rate": 9.993559329311493e-06, + "loss": 0.8188, + "step": 46732 + }, + { + "epoch": 3.01, + "learning_rate": 9.993344640288542e-06, + "loss": 0.8434, + "step": 46733 + }, + { + "epoch": 3.01, + "learning_rate": 9.993129951265591e-06, + "loss": 0.8446, + "step": 46734 + }, + { + "epoch": 3.01, + "learning_rate": 9.992915262242642e-06, + "loss": 0.7954, + "step": 46735 + }, + { + "epoch": 3.01, + "learning_rate": 9.992700573219693e-06, + "loss": 0.777, + "step": 46736 + }, + { + "epoch": 3.01, + "learning_rate": 9.992485884196742e-06, + "loss": 0.8382, + "step": 46737 + }, + { + "epoch": 3.01, + "learning_rate": 9.992271195173792e-06, + "loss": 0.8537, + "step": 46738 + }, + { + "epoch": 3.01, + "learning_rate": 9.992056506150841e-06, + "loss": 0.7875, + "step": 46739 + }, + { + "epoch": 3.01, + "learning_rate": 9.99184181712789e-06, + "loss": 0.814, + "step": 46740 + }, + { + "epoch": 3.01, + "learning_rate": 9.991627128104941e-06, + "loss": 0.8216, + "step": 46741 + }, + { + "epoch": 3.01, + "learning_rate": 9.991412439081992e-06, + "loss": 0.8236, + "step": 46742 + }, + { + "epoch": 3.01, + "learning_rate": 9.99119775005904e-06, + "loss": 0.7861, + "step": 46743 + }, + { + "epoch": 3.01, + "learning_rate": 9.99098306103609e-06, + "loss": 0.7828, + "step": 46744 + }, + { + "epoch": 3.01, + "learning_rate": 9.99076837201314e-06, + "loss": 0.8608, + "step": 46745 + }, + { + "epoch": 3.01, + "learning_rate": 9.99055368299019e-06, + "loss": 0.7793, + "step": 46746 + }, + { + "epoch": 3.01, + "learning_rate": 9.99033899396724e-06, + "loss": 0.8457, + "step": 46747 + }, + { + "epoch": 3.01, + "learning_rate": 9.990124304944289e-06, + "loss": 0.8526, + "step": 46748 + }, + { + "epoch": 3.01, + "learning_rate": 9.989909615921338e-06, + "loss": 0.806, + "step": 46749 + }, + { + "epoch": 3.01, + "learning_rate": 9.989694926898388e-06, + "loss": 0.8226, + "step": 46750 + }, + { + "epoch": 3.01, + "learning_rate": 9.989480237875439e-06, + "loss": 0.861, + "step": 46751 + }, + { + "epoch": 3.01, + "learning_rate": 9.989265548852488e-06, + "loss": 0.8238, + "step": 46752 + }, + { + "epoch": 3.01, + "learning_rate": 9.989050859829537e-06, + "loss": 0.8022, + "step": 46753 + }, + { + "epoch": 3.01, + "learning_rate": 9.988836170806588e-06, + "loss": 0.815, + "step": 46754 + }, + { + "epoch": 3.01, + "learning_rate": 9.988621481783637e-06, + "loss": 0.8297, + "step": 46755 + }, + { + "epoch": 3.01, + "learning_rate": 9.988406792760687e-06, + "loss": 0.8242, + "step": 46756 + }, + { + "epoch": 3.01, + "learning_rate": 9.988192103737738e-06, + "loss": 0.8654, + "step": 46757 + }, + { + "epoch": 3.01, + "learning_rate": 9.987977414714787e-06, + "loss": 0.8294, + "step": 46758 + }, + { + "epoch": 3.01, + "learning_rate": 9.987762725691836e-06, + "loss": 0.8306, + "step": 46759 + }, + { + "epoch": 3.01, + "learning_rate": 9.987548036668887e-06, + "loss": 0.8354, + "step": 46760 + }, + { + "epoch": 3.01, + "learning_rate": 9.987333347645935e-06, + "loss": 0.8156, + "step": 46761 + }, + { + "epoch": 3.01, + "learning_rate": 9.987118658622984e-06, + "loss": 0.8339, + "step": 46762 + }, + { + "epoch": 3.01, + "learning_rate": 9.986903969600035e-06, + "loss": 0.8537, + "step": 46763 + }, + { + "epoch": 3.01, + "learning_rate": 9.986689280577086e-06, + "loss": 0.7843, + "step": 46764 + }, + { + "epoch": 3.01, + "learning_rate": 9.986474591554135e-06, + "loss": 0.8443, + "step": 46765 + }, + { + "epoch": 3.01, + "learning_rate": 9.986259902531185e-06, + "loss": 0.8002, + "step": 46766 + }, + { + "epoch": 3.01, + "learning_rate": 9.986045213508234e-06, + "loss": 0.805, + "step": 46767 + }, + { + "epoch": 3.01, + "learning_rate": 9.985830524485283e-06, + "loss": 0.7992, + "step": 46768 + }, + { + "epoch": 3.01, + "learning_rate": 9.985615835462334e-06, + "loss": 0.7531, + "step": 46769 + }, + { + "epoch": 3.01, + "learning_rate": 9.985401146439383e-06, + "loss": 0.8328, + "step": 46770 + }, + { + "epoch": 3.01, + "learning_rate": 9.985186457416434e-06, + "loss": 0.8008, + "step": 46771 + }, + { + "epoch": 3.01, + "learning_rate": 9.984971768393483e-06, + "loss": 0.8343, + "step": 46772 + }, + { + "epoch": 3.01, + "learning_rate": 9.984757079370533e-06, + "loss": 0.8308, + "step": 46773 + }, + { + "epoch": 3.01, + "learning_rate": 9.984542390347582e-06, + "loss": 0.829, + "step": 46774 + }, + { + "epoch": 3.01, + "learning_rate": 9.984327701324633e-06, + "loss": 0.7858, + "step": 46775 + }, + { + "epoch": 3.01, + "learning_rate": 9.984113012301682e-06, + "loss": 0.8019, + "step": 46776 + }, + { + "epoch": 3.01, + "learning_rate": 9.98389832327873e-06, + "loss": 0.8058, + "step": 46777 + }, + { + "epoch": 3.01, + "learning_rate": 9.983683634255781e-06, + "loss": 0.8222, + "step": 46778 + }, + { + "epoch": 3.01, + "learning_rate": 9.983468945232832e-06, + "loss": 0.7496, + "step": 46779 + }, + { + "epoch": 3.01, + "learning_rate": 9.983254256209881e-06, + "loss": 0.8029, + "step": 46780 + }, + { + "epoch": 3.01, + "learning_rate": 9.98303956718693e-06, + "loss": 0.8425, + "step": 46781 + }, + { + "epoch": 3.01, + "learning_rate": 9.98282487816398e-06, + "loss": 0.8415, + "step": 46782 + }, + { + "epoch": 3.01, + "learning_rate": 9.98261018914103e-06, + "loss": 0.7699, + "step": 46783 + }, + { + "epoch": 3.01, + "learning_rate": 9.982395500118079e-06, + "loss": 0.8192, + "step": 46784 + }, + { + "epoch": 3.01, + "learning_rate": 9.98218081109513e-06, + "loss": 0.8112, + "step": 46785 + }, + { + "epoch": 3.01, + "learning_rate": 9.98196612207218e-06, + "loss": 0.8847, + "step": 46786 + }, + { + "epoch": 3.01, + "learning_rate": 9.981751433049229e-06, + "loss": 0.7331, + "step": 46787 + }, + { + "epoch": 3.01, + "learning_rate": 9.98153674402628e-06, + "loss": 0.857, + "step": 46788 + }, + { + "epoch": 3.01, + "learning_rate": 9.981322055003329e-06, + "loss": 0.8933, + "step": 46789 + }, + { + "epoch": 3.01, + "learning_rate": 9.981107365980377e-06, + "loss": 0.7826, + "step": 46790 + }, + { + "epoch": 3.01, + "learning_rate": 9.980892676957428e-06, + "loss": 0.7733, + "step": 46791 + }, + { + "epoch": 3.01, + "learning_rate": 9.980677987934477e-06, + "loss": 0.8446, + "step": 46792 + }, + { + "epoch": 3.01, + "learning_rate": 9.980463298911528e-06, + "loss": 0.8205, + "step": 46793 + }, + { + "epoch": 3.01, + "learning_rate": 9.980248609888577e-06, + "loss": 0.8016, + "step": 46794 + }, + { + "epoch": 3.01, + "learning_rate": 9.980033920865627e-06, + "loss": 0.8273, + "step": 46795 + }, + { + "epoch": 3.01, + "learning_rate": 9.979819231842676e-06, + "loss": 0.8519, + "step": 46796 + }, + { + "epoch": 3.01, + "learning_rate": 9.979604542819727e-06, + "loss": 0.8252, + "step": 46797 + }, + { + "epoch": 3.01, + "learning_rate": 9.979389853796776e-06, + "loss": 0.7885, + "step": 46798 + }, + { + "epoch": 3.01, + "learning_rate": 9.979175164773825e-06, + "loss": 0.7489, + "step": 46799 + }, + { + "epoch": 3.01, + "learning_rate": 9.978960475750876e-06, + "loss": 0.7602, + "step": 46800 + }, + { + "epoch": 3.01, + "learning_rate": 9.978745786727926e-06, + "loss": 0.7939, + "step": 46801 + }, + { + "epoch": 3.01, + "learning_rate": 9.978531097704975e-06, + "loss": 0.8316, + "step": 46802 + }, + { + "epoch": 3.01, + "learning_rate": 9.978316408682024e-06, + "loss": 0.792, + "step": 46803 + }, + { + "epoch": 3.01, + "learning_rate": 9.978101719659075e-06, + "loss": 0.8348, + "step": 46804 + }, + { + "epoch": 3.01, + "learning_rate": 9.977887030636124e-06, + "loss": 0.9011, + "step": 46805 + }, + { + "epoch": 3.01, + "learning_rate": 9.977672341613174e-06, + "loss": 0.8247, + "step": 46806 + }, + { + "epoch": 3.01, + "learning_rate": 9.977457652590223e-06, + "loss": 0.8543, + "step": 46807 + }, + { + "epoch": 3.01, + "learning_rate": 9.977242963567274e-06, + "loss": 0.7955, + "step": 46808 + }, + { + "epoch": 3.01, + "learning_rate": 9.977028274544323e-06, + "loss": 0.8487, + "step": 46809 + }, + { + "epoch": 3.01, + "learning_rate": 9.976813585521374e-06, + "loss": 0.8084, + "step": 46810 + }, + { + "epoch": 3.01, + "learning_rate": 9.976598896498423e-06, + "loss": 0.8307, + "step": 46811 + }, + { + "epoch": 3.01, + "learning_rate": 9.976384207475472e-06, + "loss": 0.8548, + "step": 46812 + }, + { + "epoch": 3.01, + "learning_rate": 9.976169518452522e-06, + "loss": 0.8514, + "step": 46813 + }, + { + "epoch": 3.01, + "learning_rate": 9.975954829429571e-06, + "loss": 0.7898, + "step": 46814 + }, + { + "epoch": 3.01, + "learning_rate": 9.975740140406622e-06, + "loss": 0.866, + "step": 46815 + }, + { + "epoch": 3.01, + "learning_rate": 9.975525451383673e-06, + "loss": 0.7936, + "step": 46816 + }, + { + "epoch": 3.01, + "learning_rate": 9.975310762360722e-06, + "loss": 0.8597, + "step": 46817 + }, + { + "epoch": 3.01, + "learning_rate": 9.97509607333777e-06, + "loss": 0.8667, + "step": 46818 + }, + { + "epoch": 3.01, + "learning_rate": 9.974881384314821e-06, + "loss": 0.7938, + "step": 46819 + }, + { + "epoch": 3.01, + "learning_rate": 9.97466669529187e-06, + "loss": 0.7668, + "step": 46820 + }, + { + "epoch": 3.01, + "learning_rate": 9.974452006268919e-06, + "loss": 0.7845, + "step": 46821 + }, + { + "epoch": 3.01, + "learning_rate": 9.97423731724597e-06, + "loss": 0.8263, + "step": 46822 + }, + { + "epoch": 3.01, + "learning_rate": 9.97402262822302e-06, + "loss": 0.9085, + "step": 46823 + }, + { + "epoch": 3.01, + "learning_rate": 9.97380793920007e-06, + "loss": 0.837, + "step": 46824 + }, + { + "epoch": 3.01, + "learning_rate": 9.97359325017712e-06, + "loss": 0.867, + "step": 46825 + }, + { + "epoch": 3.01, + "learning_rate": 9.973378561154169e-06, + "loss": 0.7752, + "step": 46826 + }, + { + "epoch": 3.01, + "learning_rate": 9.973163872131218e-06, + "loss": 0.7993, + "step": 46827 + }, + { + "epoch": 3.01, + "learning_rate": 9.972949183108269e-06, + "loss": 0.8696, + "step": 46828 + }, + { + "epoch": 3.01, + "learning_rate": 9.972734494085318e-06, + "loss": 0.8319, + "step": 46829 + }, + { + "epoch": 3.01, + "learning_rate": 9.972519805062368e-06, + "loss": 0.7988, + "step": 46830 + }, + { + "epoch": 3.01, + "learning_rate": 9.972305116039417e-06, + "loss": 0.7853, + "step": 46831 + }, + { + "epoch": 3.01, + "learning_rate": 9.972090427016468e-06, + "loss": 0.7732, + "step": 46832 + }, + { + "epoch": 3.01, + "learning_rate": 9.971875737993517e-06, + "loss": 0.7943, + "step": 46833 + }, + { + "epoch": 3.01, + "learning_rate": 9.971661048970568e-06, + "loss": 0.8235, + "step": 46834 + }, + { + "epoch": 3.01, + "learning_rate": 9.971446359947616e-06, + "loss": 0.7992, + "step": 46835 + }, + { + "epoch": 3.01, + "learning_rate": 9.971231670924665e-06, + "loss": 0.8401, + "step": 46836 + }, + { + "epoch": 3.01, + "learning_rate": 9.971016981901716e-06, + "loss": 0.8383, + "step": 46837 + }, + { + "epoch": 3.01, + "learning_rate": 9.970802292878767e-06, + "loss": 0.8537, + "step": 46838 + }, + { + "epoch": 3.01, + "learning_rate": 9.970587603855816e-06, + "loss": 0.8208, + "step": 46839 + }, + { + "epoch": 3.01, + "learning_rate": 9.970372914832865e-06, + "loss": 0.9012, + "step": 46840 + }, + { + "epoch": 3.01, + "learning_rate": 9.970158225809915e-06, + "loss": 0.8019, + "step": 46841 + }, + { + "epoch": 3.01, + "learning_rate": 9.969943536786964e-06, + "loss": 0.8081, + "step": 46842 + }, + { + "epoch": 3.01, + "learning_rate": 9.969728847764015e-06, + "loss": 0.8957, + "step": 46843 + }, + { + "epoch": 3.01, + "learning_rate": 9.969514158741066e-06, + "loss": 0.8035, + "step": 46844 + }, + { + "epoch": 3.01, + "learning_rate": 9.969299469718115e-06, + "loss": 0.8578, + "step": 46845 + }, + { + "epoch": 3.01, + "learning_rate": 9.969084780695164e-06, + "loss": 0.8753, + "step": 46846 + }, + { + "epoch": 3.01, + "learning_rate": 9.968870091672214e-06, + "loss": 0.7402, + "step": 46847 + }, + { + "epoch": 3.01, + "learning_rate": 9.968655402649263e-06, + "loss": 0.8452, + "step": 46848 + }, + { + "epoch": 3.01, + "learning_rate": 9.968440713626312e-06, + "loss": 0.8215, + "step": 46849 + }, + { + "epoch": 3.01, + "learning_rate": 9.968226024603363e-06, + "loss": 0.907, + "step": 46850 + }, + { + "epoch": 3.01, + "learning_rate": 9.968011335580413e-06, + "loss": 0.7686, + "step": 46851 + }, + { + "epoch": 3.01, + "learning_rate": 9.967796646557462e-06, + "loss": 0.8632, + "step": 46852 + }, + { + "epoch": 3.01, + "learning_rate": 9.967581957534513e-06, + "loss": 0.8527, + "step": 46853 + }, + { + "epoch": 3.01, + "learning_rate": 9.967367268511562e-06, + "loss": 0.8613, + "step": 46854 + }, + { + "epoch": 3.01, + "learning_rate": 9.967152579488611e-06, + "loss": 0.8432, + "step": 46855 + }, + { + "epoch": 3.01, + "learning_rate": 9.966937890465662e-06, + "loss": 0.8511, + "step": 46856 + }, + { + "epoch": 3.01, + "learning_rate": 9.96672320144271e-06, + "loss": 0.8122, + "step": 46857 + }, + { + "epoch": 3.01, + "learning_rate": 9.96650851241976e-06, + "loss": 0.8959, + "step": 46858 + }, + { + "epoch": 3.01, + "learning_rate": 9.96629382339681e-06, + "loss": 0.7845, + "step": 46859 + }, + { + "epoch": 3.01, + "learning_rate": 9.966079134373861e-06, + "loss": 0.8526, + "step": 46860 + }, + { + "epoch": 3.01, + "learning_rate": 9.96586444535091e-06, + "loss": 0.8638, + "step": 46861 + }, + { + "epoch": 3.01, + "learning_rate": 9.96564975632796e-06, + "loss": 0.7695, + "step": 46862 + }, + { + "epoch": 3.02, + "learning_rate": 9.96543506730501e-06, + "loss": 0.8024, + "step": 46863 + }, + { + "epoch": 3.02, + "learning_rate": 9.965220378282058e-06, + "loss": 0.7733, + "step": 46864 + }, + { + "epoch": 3.02, + "learning_rate": 9.96500568925911e-06, + "loss": 0.7898, + "step": 46865 + }, + { + "epoch": 3.02, + "learning_rate": 9.96479100023616e-06, + "loss": 0.8443, + "step": 46866 + }, + { + "epoch": 3.02, + "learning_rate": 9.964576311213209e-06, + "loss": 0.8087, + "step": 46867 + }, + { + "epoch": 3.02, + "learning_rate": 9.964361622190258e-06, + "loss": 0.8669, + "step": 46868 + }, + { + "epoch": 3.02, + "learning_rate": 9.964146933167308e-06, + "loss": 0.8325, + "step": 46869 + }, + { + "epoch": 3.02, + "learning_rate": 9.963932244144357e-06, + "loss": 0.841, + "step": 46870 + }, + { + "epoch": 3.02, + "learning_rate": 9.963717555121406e-06, + "loss": 0.7894, + "step": 46871 + }, + { + "epoch": 3.02, + "learning_rate": 9.963502866098457e-06, + "loss": 0.8149, + "step": 46872 + }, + { + "epoch": 3.02, + "learning_rate": 9.963288177075508e-06, + "loss": 0.905, + "step": 46873 + }, + { + "epoch": 3.02, + "learning_rate": 9.963073488052557e-06, + "loss": 0.7739, + "step": 46874 + }, + { + "epoch": 3.02, + "learning_rate": 9.962858799029607e-06, + "loss": 0.7748, + "step": 46875 + }, + { + "epoch": 3.02, + "learning_rate": 9.962644110006656e-06, + "loss": 0.773, + "step": 46876 + }, + { + "epoch": 3.02, + "learning_rate": 9.962429420983705e-06, + "loss": 0.9103, + "step": 46877 + }, + { + "epoch": 3.02, + "learning_rate": 9.962214731960756e-06, + "loss": 0.813, + "step": 46878 + }, + { + "epoch": 3.02, + "learning_rate": 9.962000042937805e-06, + "loss": 0.8693, + "step": 46879 + }, + { + "epoch": 3.02, + "learning_rate": 9.961785353914855e-06, + "loss": 0.7911, + "step": 46880 + }, + { + "epoch": 3.02, + "learning_rate": 9.961570664891904e-06, + "loss": 0.8317, + "step": 46881 + }, + { + "epoch": 3.02, + "learning_rate": 9.961355975868955e-06, + "loss": 0.9302, + "step": 46882 + }, + { + "epoch": 3.02, + "learning_rate": 9.961141286846004e-06, + "loss": 0.8098, + "step": 46883 + }, + { + "epoch": 3.02, + "learning_rate": 9.960926597823055e-06, + "loss": 0.8043, + "step": 46884 + }, + { + "epoch": 3.02, + "learning_rate": 9.960711908800104e-06, + "loss": 0.8502, + "step": 46885 + }, + { + "epoch": 3.02, + "learning_rate": 9.960497219777153e-06, + "loss": 0.845, + "step": 46886 + }, + { + "epoch": 3.02, + "learning_rate": 9.960282530754203e-06, + "loss": 0.8628, + "step": 46887 + }, + { + "epoch": 3.02, + "learning_rate": 9.960067841731254e-06, + "loss": 0.8179, + "step": 46888 + }, + { + "epoch": 3.02, + "learning_rate": 9.959853152708303e-06, + "loss": 0.8347, + "step": 46889 + }, + { + "epoch": 3.02, + "learning_rate": 9.959638463685352e-06, + "loss": 0.822, + "step": 46890 + }, + { + "epoch": 3.02, + "learning_rate": 9.959423774662403e-06, + "loss": 0.7986, + "step": 46891 + }, + { + "epoch": 3.02, + "learning_rate": 9.959209085639452e-06, + "loss": 0.7811, + "step": 46892 + }, + { + "epoch": 3.02, + "learning_rate": 9.958994396616502e-06, + "loss": 0.815, + "step": 46893 + }, + { + "epoch": 3.02, + "learning_rate": 9.958779707593551e-06, + "loss": 0.8145, + "step": 46894 + }, + { + "epoch": 3.02, + "learning_rate": 9.958565018570602e-06, + "loss": 0.8384, + "step": 46895 + }, + { + "epoch": 3.02, + "learning_rate": 9.95835032954765e-06, + "loss": 0.7707, + "step": 46896 + }, + { + "epoch": 3.02, + "learning_rate": 9.958135640524701e-06, + "loss": 0.8038, + "step": 46897 + }, + { + "epoch": 3.02, + "learning_rate": 9.95792095150175e-06, + "loss": 0.8709, + "step": 46898 + }, + { + "epoch": 3.02, + "learning_rate": 9.9577062624788e-06, + "loss": 0.7444, + "step": 46899 + }, + { + "epoch": 3.02, + "learning_rate": 9.95749157345585e-06, + "loss": 0.8647, + "step": 46900 + }, + { + "epoch": 3.02, + "learning_rate": 9.957276884432899e-06, + "loss": 0.8038, + "step": 46901 + }, + { + "epoch": 3.02, + "learning_rate": 9.95706219540995e-06, + "loss": 0.7883, + "step": 46902 + }, + { + "epoch": 3.02, + "learning_rate": 9.956847506387e-06, + "loss": 0.8618, + "step": 46903 + }, + { + "epoch": 3.02, + "learning_rate": 9.95663281736405e-06, + "loss": 0.8666, + "step": 46904 + }, + { + "epoch": 3.02, + "learning_rate": 9.956418128341098e-06, + "loss": 0.8044, + "step": 46905 + }, + { + "epoch": 3.02, + "learning_rate": 9.956203439318149e-06, + "loss": 0.8591, + "step": 46906 + }, + { + "epoch": 3.02, + "learning_rate": 9.955988750295198e-06, + "loss": 0.8436, + "step": 46907 + }, + { + "epoch": 3.02, + "learning_rate": 9.955774061272247e-06, + "loss": 0.7882, + "step": 46908 + }, + { + "epoch": 3.02, + "learning_rate": 9.955559372249297e-06, + "loss": 0.8031, + "step": 46909 + }, + { + "epoch": 3.02, + "learning_rate": 9.955344683226348e-06, + "loss": 0.8667, + "step": 46910 + }, + { + "epoch": 3.02, + "learning_rate": 9.955129994203397e-06, + "loss": 0.8425, + "step": 46911 + }, + { + "epoch": 3.02, + "learning_rate": 9.954915305180448e-06, + "loss": 0.8401, + "step": 46912 + }, + { + "epoch": 3.02, + "learning_rate": 9.954700616157497e-06, + "loss": 0.8298, + "step": 46913 + }, + { + "epoch": 3.02, + "learning_rate": 9.954485927134546e-06, + "loss": 0.835, + "step": 46914 + }, + { + "epoch": 3.02, + "learning_rate": 9.954271238111596e-06, + "loss": 0.8433, + "step": 46915 + }, + { + "epoch": 3.02, + "learning_rate": 9.954056549088645e-06, + "loss": 0.8354, + "step": 46916 + }, + { + "epoch": 3.02, + "learning_rate": 9.953841860065696e-06, + "loss": 0.7761, + "step": 46917 + }, + { + "epoch": 3.02, + "learning_rate": 9.953627171042745e-06, + "loss": 0.8219, + "step": 46918 + }, + { + "epoch": 3.02, + "learning_rate": 9.953412482019796e-06, + "loss": 0.8357, + "step": 46919 + }, + { + "epoch": 3.02, + "learning_rate": 9.953197792996845e-06, + "loss": 0.8587, + "step": 46920 + }, + { + "epoch": 3.02, + "learning_rate": 9.952983103973895e-06, + "loss": 0.834, + "step": 46921 + }, + { + "epoch": 3.02, + "learning_rate": 9.952768414950944e-06, + "loss": 0.8737, + "step": 46922 + }, + { + "epoch": 3.02, + "learning_rate": 9.952553725927993e-06, + "loss": 0.8224, + "step": 46923 + }, + { + "epoch": 3.02, + "learning_rate": 9.952339036905044e-06, + "loss": 0.7794, + "step": 46924 + }, + { + "epoch": 3.02, + "learning_rate": 9.952124347882094e-06, + "loss": 0.8603, + "step": 46925 + }, + { + "epoch": 3.02, + "learning_rate": 9.951909658859143e-06, + "loss": 0.8107, + "step": 46926 + }, + { + "epoch": 3.02, + "learning_rate": 9.951694969836192e-06, + "loss": 0.799, + "step": 46927 + }, + { + "epoch": 3.02, + "learning_rate": 9.951480280813243e-06, + "loss": 0.8559, + "step": 46928 + }, + { + "epoch": 3.02, + "learning_rate": 9.951265591790292e-06, + "loss": 0.8197, + "step": 46929 + }, + { + "epoch": 3.02, + "learning_rate": 9.951050902767343e-06, + "loss": 0.8366, + "step": 46930 + }, + { + "epoch": 3.02, + "learning_rate": 9.950836213744393e-06, + "loss": 0.8552, + "step": 46931 + }, + { + "epoch": 3.02, + "learning_rate": 9.950621524721442e-06, + "loss": 0.8582, + "step": 46932 + }, + { + "epoch": 3.02, + "learning_rate": 9.950406835698491e-06, + "loss": 0.791, + "step": 46933 + }, + { + "epoch": 3.02, + "learning_rate": 9.950192146675542e-06, + "loss": 0.7884, + "step": 46934 + }, + { + "epoch": 3.02, + "learning_rate": 9.949977457652591e-06, + "loss": 0.7716, + "step": 46935 + }, + { + "epoch": 3.02, + "learning_rate": 9.94976276862964e-06, + "loss": 0.8125, + "step": 46936 + }, + { + "epoch": 3.02, + "learning_rate": 9.94954807960669e-06, + "loss": 0.7964, + "step": 46937 + }, + { + "epoch": 3.02, + "learning_rate": 9.949333390583741e-06, + "loss": 0.8173, + "step": 46938 + }, + { + "epoch": 3.02, + "learning_rate": 9.94911870156079e-06, + "loss": 0.7955, + "step": 46939 + }, + { + "epoch": 3.02, + "learning_rate": 9.94890401253784e-06, + "loss": 0.8403, + "step": 46940 + }, + { + "epoch": 3.02, + "learning_rate": 9.94868932351489e-06, + "loss": 0.7622, + "step": 46941 + }, + { + "epoch": 3.02, + "learning_rate": 9.948474634491939e-06, + "loss": 0.8741, + "step": 46942 + }, + { + "epoch": 3.02, + "learning_rate": 9.94825994546899e-06, + "loss": 0.8397, + "step": 46943 + }, + { + "epoch": 3.02, + "learning_rate": 9.948045256446038e-06, + "loss": 0.7808, + "step": 46944 + }, + { + "epoch": 3.02, + "learning_rate": 9.947830567423087e-06, + "loss": 0.8321, + "step": 46945 + }, + { + "epoch": 3.02, + "learning_rate": 9.947615878400138e-06, + "loss": 0.8349, + "step": 46946 + }, + { + "epoch": 3.02, + "learning_rate": 9.947401189377189e-06, + "loss": 0.7979, + "step": 46947 + }, + { + "epoch": 3.02, + "learning_rate": 9.947186500354238e-06, + "loss": 0.8764, + "step": 46948 + }, + { + "epoch": 3.02, + "learning_rate": 9.946971811331287e-06, + "loss": 0.7808, + "step": 46949 + }, + { + "epoch": 3.02, + "learning_rate": 9.946757122308337e-06, + "loss": 0.8146, + "step": 46950 + }, + { + "epoch": 3.02, + "learning_rate": 9.946542433285386e-06, + "loss": 0.8203, + "step": 46951 + }, + { + "epoch": 3.02, + "learning_rate": 9.946327744262437e-06, + "loss": 0.8425, + "step": 46952 + }, + { + "epoch": 3.02, + "learning_rate": 9.946113055239488e-06, + "loss": 0.8686, + "step": 46953 + }, + { + "epoch": 3.02, + "learning_rate": 9.945898366216536e-06, + "loss": 0.798, + "step": 46954 + }, + { + "epoch": 3.02, + "learning_rate": 9.945683677193585e-06, + "loss": 0.8076, + "step": 46955 + }, + { + "epoch": 3.02, + "learning_rate": 9.945468988170636e-06, + "loss": 0.7696, + "step": 46956 + }, + { + "epoch": 3.02, + "learning_rate": 9.945254299147685e-06, + "loss": 0.8652, + "step": 46957 + }, + { + "epoch": 3.02, + "learning_rate": 9.945039610124734e-06, + "loss": 0.8395, + "step": 46958 + }, + { + "epoch": 3.02, + "learning_rate": 9.944824921101785e-06, + "loss": 0.8599, + "step": 46959 + }, + { + "epoch": 3.02, + "learning_rate": 9.944610232078835e-06, + "loss": 0.8477, + "step": 46960 + }, + { + "epoch": 3.02, + "learning_rate": 9.944395543055884e-06, + "loss": 0.8048, + "step": 46961 + }, + { + "epoch": 3.02, + "learning_rate": 9.944180854032935e-06, + "loss": 0.8862, + "step": 46962 + }, + { + "epoch": 3.02, + "learning_rate": 9.943966165009984e-06, + "loss": 0.7687, + "step": 46963 + }, + { + "epoch": 3.02, + "learning_rate": 9.943751475987033e-06, + "loss": 0.8334, + "step": 46964 + }, + { + "epoch": 3.02, + "learning_rate": 9.943536786964084e-06, + "loss": 0.8133, + "step": 46965 + }, + { + "epoch": 3.02, + "learning_rate": 9.943322097941133e-06, + "loss": 0.755, + "step": 46966 + }, + { + "epoch": 3.02, + "learning_rate": 9.943107408918183e-06, + "loss": 0.8512, + "step": 46967 + }, + { + "epoch": 3.02, + "learning_rate": 9.942892719895232e-06, + "loss": 0.8313, + "step": 46968 + }, + { + "epoch": 3.02, + "learning_rate": 9.942678030872283e-06, + "loss": 0.8671, + "step": 46969 + }, + { + "epoch": 3.02, + "learning_rate": 9.942463341849332e-06, + "loss": 0.82, + "step": 46970 + }, + { + "epoch": 3.02, + "learning_rate": 9.942248652826382e-06, + "loss": 0.8131, + "step": 46971 + }, + { + "epoch": 3.02, + "learning_rate": 9.942033963803431e-06, + "loss": 0.8009, + "step": 46972 + }, + { + "epoch": 3.02, + "learning_rate": 9.94181927478048e-06, + "loss": 0.7622, + "step": 46973 + }, + { + "epoch": 3.02, + "learning_rate": 9.941604585757531e-06, + "loss": 0.7839, + "step": 46974 + }, + { + "epoch": 3.02, + "learning_rate": 9.941389896734582e-06, + "loss": 0.8337, + "step": 46975 + }, + { + "epoch": 3.02, + "learning_rate": 9.94117520771163e-06, + "loss": 0.7896, + "step": 46976 + }, + { + "epoch": 3.02, + "learning_rate": 9.94096051868868e-06, + "loss": 0.8064, + "step": 46977 + }, + { + "epoch": 3.02, + "learning_rate": 9.94074582966573e-06, + "loss": 0.8188, + "step": 46978 + }, + { + "epoch": 3.02, + "learning_rate": 9.94053114064278e-06, + "loss": 0.7657, + "step": 46979 + }, + { + "epoch": 3.02, + "learning_rate": 9.94031645161983e-06, + "loss": 0.8377, + "step": 46980 + }, + { + "epoch": 3.02, + "learning_rate": 9.940101762596879e-06, + "loss": 0.8339, + "step": 46981 + }, + { + "epoch": 3.02, + "learning_rate": 9.93988707357393e-06, + "loss": 0.8339, + "step": 46982 + }, + { + "epoch": 3.02, + "learning_rate": 9.939672384550978e-06, + "loss": 0.7929, + "step": 46983 + }, + { + "epoch": 3.02, + "learning_rate": 9.93945769552803e-06, + "loss": 0.8256, + "step": 46984 + }, + { + "epoch": 3.02, + "learning_rate": 9.939243006505078e-06, + "loss": 0.7556, + "step": 46985 + }, + { + "epoch": 3.02, + "learning_rate": 9.939028317482127e-06, + "loss": 0.7792, + "step": 46986 + }, + { + "epoch": 3.02, + "learning_rate": 9.938813628459178e-06, + "loss": 0.8432, + "step": 46987 + }, + { + "epoch": 3.02, + "learning_rate": 9.938598939436227e-06, + "loss": 0.7957, + "step": 46988 + }, + { + "epoch": 3.02, + "learning_rate": 9.938384250413277e-06, + "loss": 0.8652, + "step": 46989 + }, + { + "epoch": 3.02, + "learning_rate": 9.938169561390328e-06, + "loss": 0.8054, + "step": 46990 + }, + { + "epoch": 3.02, + "learning_rate": 9.937954872367377e-06, + "loss": 0.7813, + "step": 46991 + }, + { + "epoch": 3.02, + "learning_rate": 9.937740183344426e-06, + "loss": 0.7697, + "step": 46992 + }, + { + "epoch": 3.02, + "learning_rate": 9.937525494321477e-06, + "loss": 0.781, + "step": 46993 + }, + { + "epoch": 3.02, + "learning_rate": 9.937310805298526e-06, + "loss": 0.8611, + "step": 46994 + }, + { + "epoch": 3.02, + "learning_rate": 9.937096116275575e-06, + "loss": 0.7542, + "step": 46995 + }, + { + "epoch": 3.02, + "learning_rate": 9.936881427252625e-06, + "loss": 0.8345, + "step": 46996 + }, + { + "epoch": 3.02, + "learning_rate": 9.936666738229676e-06, + "loss": 0.8671, + "step": 46997 + }, + { + "epoch": 3.02, + "learning_rate": 9.936452049206725e-06, + "loss": 0.7977, + "step": 46998 + }, + { + "epoch": 3.02, + "learning_rate": 9.936237360183775e-06, + "loss": 0.8008, + "step": 46999 + }, + { + "epoch": 3.02, + "learning_rate": 9.936022671160824e-06, + "loss": 0.823, + "step": 47000 + }, + { + "epoch": 3.02, + "eval_laion/cider": 0.5462685782906329, + "eval_laion/meteor": 0.3990934305244386, + "eval_laion/pred_num_words": 36.15625, + "eval_laion/sacrebleu": 16.13506259798906, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0455368757247925, + "eval_runtime": 96.992, + "eval_samples_per_second": 5.145, + "eval_steps_per_second": 0.33, + "step": 47000 + }, + { + "epoch": 3.02, + "learning_rate": 9.935807982137873e-06, + "loss": 0.8413, + "step": 47001 + }, + { + "epoch": 3.02, + "learning_rate": 9.935593293114924e-06, + "loss": 0.7887, + "step": 47002 + }, + { + "epoch": 3.02, + "learning_rate": 9.935378604091973e-06, + "loss": 0.8415, + "step": 47003 + }, + { + "epoch": 3.02, + "learning_rate": 9.935163915069024e-06, + "loss": 0.8686, + "step": 47004 + }, + { + "epoch": 3.02, + "learning_rate": 9.934949226046073e-06, + "loss": 0.8075, + "step": 47005 + }, + { + "epoch": 3.02, + "learning_rate": 9.934734537023123e-06, + "loss": 0.8182, + "step": 47006 + }, + { + "epoch": 3.02, + "learning_rate": 9.934519848000172e-06, + "loss": 0.8327, + "step": 47007 + }, + { + "epoch": 3.02, + "learning_rate": 9.934305158977223e-06, + "loss": 0.8575, + "step": 47008 + }, + { + "epoch": 3.02, + "learning_rate": 9.934090469954272e-06, + "loss": 0.7836, + "step": 47009 + }, + { + "epoch": 3.02, + "learning_rate": 9.933875780931321e-06, + "loss": 0.8573, + "step": 47010 + }, + { + "epoch": 3.02, + "learning_rate": 9.933661091908372e-06, + "loss": 0.8187, + "step": 47011 + }, + { + "epoch": 3.02, + "learning_rate": 9.933446402885422e-06, + "loss": 0.7468, + "step": 47012 + }, + { + "epoch": 3.02, + "learning_rate": 9.933231713862471e-06, + "loss": 0.8145, + "step": 47013 + }, + { + "epoch": 3.02, + "learning_rate": 9.93301702483952e-06, + "loss": 0.8149, + "step": 47014 + }, + { + "epoch": 3.02, + "learning_rate": 9.93280233581657e-06, + "loss": 0.8173, + "step": 47015 + }, + { + "epoch": 3.02, + "learning_rate": 9.93258764679362e-06, + "loss": 0.8077, + "step": 47016 + }, + { + "epoch": 3.02, + "learning_rate": 9.93237295777067e-06, + "loss": 0.8658, + "step": 47017 + }, + { + "epoch": 3.03, + "learning_rate": 9.932158268747721e-06, + "loss": 0.8293, + "step": 47018 + }, + { + "epoch": 3.03, + "learning_rate": 9.93194357972477e-06, + "loss": 0.8697, + "step": 47019 + }, + { + "epoch": 3.03, + "learning_rate": 9.931728890701819e-06, + "loss": 0.8528, + "step": 47020 + }, + { + "epoch": 3.03, + "learning_rate": 9.93151420167887e-06, + "loss": 0.8169, + "step": 47021 + }, + { + "epoch": 3.03, + "learning_rate": 9.931299512655919e-06, + "loss": 0.7465, + "step": 47022 + }, + { + "epoch": 3.03, + "learning_rate": 9.931084823632968e-06, + "loss": 0.8224, + "step": 47023 + }, + { + "epoch": 3.03, + "learning_rate": 9.930870134610018e-06, + "loss": 0.7959, + "step": 47024 + }, + { + "epoch": 3.03, + "learning_rate": 9.930655445587067e-06, + "loss": 0.8112, + "step": 47025 + }, + { + "epoch": 3.03, + "learning_rate": 9.930440756564118e-06, + "loss": 0.7818, + "step": 47026 + }, + { + "epoch": 3.03, + "learning_rate": 9.930226067541169e-06, + "loss": 0.8154, + "step": 47027 + }, + { + "epoch": 3.03, + "learning_rate": 9.930011378518217e-06, + "loss": 0.8425, + "step": 47028 + }, + { + "epoch": 3.03, + "learning_rate": 9.929796689495266e-06, + "loss": 0.7966, + "step": 47029 + }, + { + "epoch": 3.03, + "learning_rate": 9.929582000472317e-06, + "loss": 0.7985, + "step": 47030 + }, + { + "epoch": 3.03, + "learning_rate": 9.929367311449366e-06, + "loss": 0.8383, + "step": 47031 + }, + { + "epoch": 3.03, + "learning_rate": 9.929152622426415e-06, + "loss": 0.8093, + "step": 47032 + }, + { + "epoch": 3.03, + "learning_rate": 9.928937933403466e-06, + "loss": 0.836, + "step": 47033 + }, + { + "epoch": 3.03, + "learning_rate": 9.928723244380516e-06, + "loss": 0.8454, + "step": 47034 + }, + { + "epoch": 3.03, + "learning_rate": 9.928508555357565e-06, + "loss": 0.7385, + "step": 47035 + }, + { + "epoch": 3.03, + "learning_rate": 9.928293866334614e-06, + "loss": 0.7793, + "step": 47036 + }, + { + "epoch": 3.03, + "learning_rate": 9.928079177311665e-06, + "loss": 0.8507, + "step": 47037 + }, + { + "epoch": 3.03, + "learning_rate": 9.927864488288714e-06, + "loss": 0.8007, + "step": 47038 + }, + { + "epoch": 3.03, + "learning_rate": 9.927649799265765e-06, + "loss": 0.8306, + "step": 47039 + }, + { + "epoch": 3.03, + "learning_rate": 9.927435110242815e-06, + "loss": 0.8465, + "step": 47040 + }, + { + "epoch": 3.03, + "learning_rate": 9.927220421219864e-06, + "loss": 0.7494, + "step": 47041 + }, + { + "epoch": 3.03, + "learning_rate": 9.927005732196913e-06, + "loss": 0.8802, + "step": 47042 + }, + { + "epoch": 3.03, + "learning_rate": 9.926791043173964e-06, + "loss": 0.8398, + "step": 47043 + }, + { + "epoch": 3.03, + "learning_rate": 9.926576354151013e-06, + "loss": 0.8245, + "step": 47044 + }, + { + "epoch": 3.03, + "learning_rate": 9.926361665128062e-06, + "loss": 0.7319, + "step": 47045 + }, + { + "epoch": 3.03, + "learning_rate": 9.926146976105112e-06, + "loss": 0.8037, + "step": 47046 + }, + { + "epoch": 3.03, + "learning_rate": 9.925932287082163e-06, + "loss": 0.8297, + "step": 47047 + }, + { + "epoch": 3.03, + "learning_rate": 9.925717598059212e-06, + "loss": 0.7993, + "step": 47048 + }, + { + "epoch": 3.03, + "learning_rate": 9.925502909036263e-06, + "loss": 0.8273, + "step": 47049 + }, + { + "epoch": 3.03, + "learning_rate": 9.925288220013312e-06, + "loss": 0.7731, + "step": 47050 + }, + { + "epoch": 3.03, + "learning_rate": 9.92507353099036e-06, + "loss": 0.7915, + "step": 47051 + }, + { + "epoch": 3.03, + "learning_rate": 9.924858841967411e-06, + "loss": 0.8282, + "step": 47052 + }, + { + "epoch": 3.03, + "learning_rate": 9.92464415294446e-06, + "loss": 0.7595, + "step": 47053 + }, + { + "epoch": 3.03, + "learning_rate": 9.92442946392151e-06, + "loss": 0.8034, + "step": 47054 + }, + { + "epoch": 3.03, + "learning_rate": 9.92421477489856e-06, + "loss": 0.8323, + "step": 47055 + }, + { + "epoch": 3.03, + "learning_rate": 9.92400008587561e-06, + "loss": 0.908, + "step": 47056 + }, + { + "epoch": 3.03, + "learning_rate": 9.92378539685266e-06, + "loss": 0.749, + "step": 47057 + }, + { + "epoch": 3.03, + "learning_rate": 9.92357070782971e-06, + "loss": 0.7958, + "step": 47058 + }, + { + "epoch": 3.03, + "learning_rate": 9.923356018806759e-06, + "loss": 0.8405, + "step": 47059 + }, + { + "epoch": 3.03, + "learning_rate": 9.923141329783808e-06, + "loss": 0.7818, + "step": 47060 + }, + { + "epoch": 3.03, + "learning_rate": 9.922926640760859e-06, + "loss": 0.84, + "step": 47061 + }, + { + "epoch": 3.03, + "learning_rate": 9.92271195173791e-06, + "loss": 0.8184, + "step": 47062 + }, + { + "epoch": 3.03, + "learning_rate": 9.922497262714958e-06, + "loss": 0.8167, + "step": 47063 + }, + { + "epoch": 3.03, + "learning_rate": 9.922282573692007e-06, + "loss": 0.8155, + "step": 47064 + }, + { + "epoch": 3.03, + "learning_rate": 9.922067884669058e-06, + "loss": 0.8465, + "step": 47065 + }, + { + "epoch": 3.03, + "learning_rate": 9.921853195646107e-06, + "loss": 0.8375, + "step": 47066 + }, + { + "epoch": 3.03, + "learning_rate": 9.921638506623158e-06, + "loss": 0.8076, + "step": 47067 + }, + { + "epoch": 3.03, + "learning_rate": 9.921423817600207e-06, + "loss": 0.8073, + "step": 47068 + }, + { + "epoch": 3.03, + "learning_rate": 9.921209128577257e-06, + "loss": 0.8149, + "step": 47069 + }, + { + "epoch": 3.03, + "learning_rate": 9.920994439554306e-06, + "loss": 0.8471, + "step": 47070 + }, + { + "epoch": 3.03, + "learning_rate": 9.920779750531357e-06, + "loss": 0.8487, + "step": 47071 + }, + { + "epoch": 3.03, + "learning_rate": 9.920565061508406e-06, + "loss": 0.837, + "step": 47072 + }, + { + "epoch": 3.03, + "learning_rate": 9.920350372485455e-06, + "loss": 0.8488, + "step": 47073 + }, + { + "epoch": 3.03, + "learning_rate": 9.920135683462505e-06, + "loss": 0.8141, + "step": 47074 + }, + { + "epoch": 3.03, + "learning_rate": 9.919920994439554e-06, + "loss": 0.8213, + "step": 47075 + }, + { + "epoch": 3.03, + "learning_rate": 9.919706305416605e-06, + "loss": 0.7386, + "step": 47076 + }, + { + "epoch": 3.03, + "learning_rate": 9.919491616393656e-06, + "loss": 0.8271, + "step": 47077 + }, + { + "epoch": 3.03, + "learning_rate": 9.919276927370705e-06, + "loss": 0.8209, + "step": 47078 + }, + { + "epoch": 3.03, + "learning_rate": 9.919062238347754e-06, + "loss": 0.7951, + "step": 47079 + }, + { + "epoch": 3.03, + "learning_rate": 9.918847549324804e-06, + "loss": 0.8112, + "step": 47080 + }, + { + "epoch": 3.03, + "learning_rate": 9.918632860301853e-06, + "loss": 0.8298, + "step": 47081 + }, + { + "epoch": 3.03, + "learning_rate": 9.918418171278902e-06, + "loss": 0.8649, + "step": 47082 + }, + { + "epoch": 3.03, + "learning_rate": 9.918203482255953e-06, + "loss": 0.7872, + "step": 47083 + }, + { + "epoch": 3.03, + "learning_rate": 9.917988793233004e-06, + "loss": 0.7657, + "step": 47084 + }, + { + "epoch": 3.03, + "learning_rate": 9.917774104210053e-06, + "loss": 0.7731, + "step": 47085 + }, + { + "epoch": 3.03, + "learning_rate": 9.917559415187103e-06, + "loss": 0.8626, + "step": 47086 + }, + { + "epoch": 3.03, + "learning_rate": 9.917344726164152e-06, + "loss": 0.818, + "step": 47087 + }, + { + "epoch": 3.03, + "learning_rate": 9.917130037141201e-06, + "loss": 0.8587, + "step": 47088 + }, + { + "epoch": 3.03, + "learning_rate": 9.916915348118252e-06, + "loss": 0.8189, + "step": 47089 + }, + { + "epoch": 3.03, + "learning_rate": 9.9167006590953e-06, + "loss": 0.8091, + "step": 47090 + }, + { + "epoch": 3.03, + "learning_rate": 9.916485970072351e-06, + "loss": 0.8951, + "step": 47091 + }, + { + "epoch": 3.03, + "learning_rate": 9.9162712810494e-06, + "loss": 0.8046, + "step": 47092 + }, + { + "epoch": 3.03, + "learning_rate": 9.916056592026451e-06, + "loss": 0.774, + "step": 47093 + }, + { + "epoch": 3.03, + "learning_rate": 9.9158419030035e-06, + "loss": 0.8038, + "step": 47094 + }, + { + "epoch": 3.03, + "learning_rate": 9.91562721398055e-06, + "loss": 0.7733, + "step": 47095 + }, + { + "epoch": 3.03, + "learning_rate": 9.9154125249576e-06, + "loss": 0.7854, + "step": 47096 + }, + { + "epoch": 3.03, + "learning_rate": 9.915197835934649e-06, + "loss": 0.7929, + "step": 47097 + }, + { + "epoch": 3.03, + "learning_rate": 9.9149831469117e-06, + "loss": 0.7699, + "step": 47098 + }, + { + "epoch": 3.03, + "learning_rate": 9.91476845788875e-06, + "loss": 0.7461, + "step": 47099 + }, + { + "epoch": 3.03, + "learning_rate": 9.914553768865799e-06, + "loss": 0.7847, + "step": 47100 + }, + { + "epoch": 3.03, + "learning_rate": 9.914339079842848e-06, + "loss": 0.8257, + "step": 47101 + }, + { + "epoch": 3.03, + "learning_rate": 9.914124390819899e-06, + "loss": 0.8167, + "step": 47102 + }, + { + "epoch": 3.03, + "learning_rate": 9.913909701796947e-06, + "loss": 0.7978, + "step": 47103 + }, + { + "epoch": 3.03, + "learning_rate": 9.913695012773996e-06, + "loss": 0.8465, + "step": 47104 + }, + { + "epoch": 3.03, + "learning_rate": 9.913480323751049e-06, + "loss": 0.7887, + "step": 47105 + }, + { + "epoch": 3.03, + "learning_rate": 9.913265634728098e-06, + "loss": 0.7738, + "step": 47106 + }, + { + "epoch": 3.03, + "learning_rate": 9.913050945705147e-06, + "loss": 0.7488, + "step": 47107 + }, + { + "epoch": 3.03, + "learning_rate": 9.912836256682197e-06, + "loss": 0.8331, + "step": 47108 + }, + { + "epoch": 3.03, + "learning_rate": 9.912621567659246e-06, + "loss": 0.8335, + "step": 47109 + }, + { + "epoch": 3.03, + "learning_rate": 9.912406878636295e-06, + "loss": 0.8014, + "step": 47110 + }, + { + "epoch": 3.03, + "learning_rate": 9.912192189613346e-06, + "loss": 0.8042, + "step": 47111 + }, + { + "epoch": 3.03, + "learning_rate": 9.911977500590395e-06, + "loss": 0.7896, + "step": 47112 + }, + { + "epoch": 3.03, + "learning_rate": 9.911762811567446e-06, + "loss": 0.8452, + "step": 47113 + }, + { + "epoch": 3.03, + "learning_rate": 9.911548122544495e-06, + "loss": 0.7951, + "step": 47114 + }, + { + "epoch": 3.03, + "learning_rate": 9.911333433521545e-06, + "loss": 0.7771, + "step": 47115 + }, + { + "epoch": 3.03, + "learning_rate": 9.911118744498594e-06, + "loss": 0.8365, + "step": 47116 + }, + { + "epoch": 3.03, + "learning_rate": 9.910904055475645e-06, + "loss": 0.8081, + "step": 47117 + }, + { + "epoch": 3.03, + "learning_rate": 9.910689366452694e-06, + "loss": 0.7673, + "step": 47118 + }, + { + "epoch": 3.03, + "learning_rate": 9.910474677429743e-06, + "loss": 0.8286, + "step": 47119 + }, + { + "epoch": 3.03, + "learning_rate": 9.910259988406793e-06, + "loss": 0.8014, + "step": 47120 + }, + { + "epoch": 3.03, + "learning_rate": 9.910045299383844e-06, + "loss": 0.8013, + "step": 47121 + }, + { + "epoch": 3.03, + "learning_rate": 9.909830610360893e-06, + "loss": 0.8474, + "step": 47122 + }, + { + "epoch": 3.03, + "learning_rate": 9.909615921337942e-06, + "loss": 0.859, + "step": 47123 + }, + { + "epoch": 3.03, + "learning_rate": 9.909401232314993e-06, + "loss": 0.811, + "step": 47124 + }, + { + "epoch": 3.03, + "learning_rate": 9.909186543292042e-06, + "loss": 0.7987, + "step": 47125 + }, + { + "epoch": 3.03, + "learning_rate": 9.908971854269092e-06, + "loss": 0.8645, + "step": 47126 + }, + { + "epoch": 3.03, + "learning_rate": 9.908757165246143e-06, + "loss": 0.7923, + "step": 47127 + }, + { + "epoch": 3.03, + "learning_rate": 9.908542476223192e-06, + "loss": 0.8597, + "step": 47128 + }, + { + "epoch": 3.03, + "learning_rate": 9.908327787200241e-06, + "loss": 0.867, + "step": 47129 + }, + { + "epoch": 3.03, + "learning_rate": 9.908113098177292e-06, + "loss": 0.7924, + "step": 47130 + }, + { + "epoch": 3.03, + "learning_rate": 9.90789840915434e-06, + "loss": 0.8129, + "step": 47131 + }, + { + "epoch": 3.03, + "learning_rate": 9.90768372013139e-06, + "loss": 0.7585, + "step": 47132 + }, + { + "epoch": 3.03, + "learning_rate": 9.90746903110844e-06, + "loss": 0.8598, + "step": 47133 + }, + { + "epoch": 3.03, + "learning_rate": 9.90725434208549e-06, + "loss": 0.8163, + "step": 47134 + }, + { + "epoch": 3.03, + "learning_rate": 9.90703965306254e-06, + "loss": 0.8035, + "step": 47135 + }, + { + "epoch": 3.03, + "learning_rate": 9.90682496403959e-06, + "loss": 0.8912, + "step": 47136 + }, + { + "epoch": 3.03, + "learning_rate": 9.90661027501664e-06, + "loss": 0.8605, + "step": 47137 + }, + { + "epoch": 3.03, + "learning_rate": 9.906395585993688e-06, + "loss": 0.834, + "step": 47138 + }, + { + "epoch": 3.03, + "learning_rate": 9.906180896970739e-06, + "loss": 0.7779, + "step": 47139 + }, + { + "epoch": 3.03, + "learning_rate": 9.905966207947788e-06, + "loss": 0.8854, + "step": 47140 + }, + { + "epoch": 3.03, + "learning_rate": 9.905751518924837e-06, + "loss": 0.8716, + "step": 47141 + }, + { + "epoch": 3.03, + "learning_rate": 9.905536829901888e-06, + "loss": 0.8314, + "step": 47142 + }, + { + "epoch": 3.03, + "learning_rate": 9.905322140878938e-06, + "loss": 0.8538, + "step": 47143 + }, + { + "epoch": 3.03, + "learning_rate": 9.905107451855987e-06, + "loss": 0.8375, + "step": 47144 + }, + { + "epoch": 3.03, + "learning_rate": 9.904892762833038e-06, + "loss": 0.7564, + "step": 47145 + }, + { + "epoch": 3.03, + "learning_rate": 9.904678073810087e-06, + "loss": 0.7988, + "step": 47146 + }, + { + "epoch": 3.03, + "learning_rate": 9.904463384787136e-06, + "loss": 0.8813, + "step": 47147 + }, + { + "epoch": 3.03, + "learning_rate": 9.904248695764186e-06, + "loss": 0.7847, + "step": 47148 + }, + { + "epoch": 3.03, + "learning_rate": 9.904034006741237e-06, + "loss": 0.8324, + "step": 47149 + }, + { + "epoch": 3.03, + "learning_rate": 9.903819317718286e-06, + "loss": 0.7595, + "step": 47150 + }, + { + "epoch": 3.03, + "learning_rate": 9.903604628695335e-06, + "loss": 0.8069, + "step": 47151 + }, + { + "epoch": 3.03, + "learning_rate": 9.903389939672386e-06, + "loss": 0.8485, + "step": 47152 + }, + { + "epoch": 3.03, + "learning_rate": 9.903175250649435e-06, + "loss": 0.7947, + "step": 47153 + }, + { + "epoch": 3.03, + "learning_rate": 9.902960561626485e-06, + "loss": 0.8105, + "step": 47154 + }, + { + "epoch": 3.03, + "learning_rate": 9.902745872603534e-06, + "loss": 0.8612, + "step": 47155 + }, + { + "epoch": 3.03, + "learning_rate": 9.902531183580585e-06, + "loss": 0.8556, + "step": 47156 + }, + { + "epoch": 3.03, + "learning_rate": 9.902316494557634e-06, + "loss": 0.8518, + "step": 47157 + }, + { + "epoch": 3.03, + "learning_rate": 9.902101805534685e-06, + "loss": 0.7518, + "step": 47158 + }, + { + "epoch": 3.03, + "learning_rate": 9.901887116511734e-06, + "loss": 0.8216, + "step": 47159 + }, + { + "epoch": 3.03, + "learning_rate": 9.901672427488783e-06, + "loss": 0.7306, + "step": 47160 + }, + { + "epoch": 3.03, + "learning_rate": 9.901457738465833e-06, + "loss": 0.8525, + "step": 47161 + }, + { + "epoch": 3.03, + "learning_rate": 9.901243049442882e-06, + "loss": 0.7568, + "step": 47162 + }, + { + "epoch": 3.03, + "learning_rate": 9.901028360419933e-06, + "loss": 0.8636, + "step": 47163 + }, + { + "epoch": 3.03, + "learning_rate": 9.900813671396983e-06, + "loss": 0.8169, + "step": 47164 + }, + { + "epoch": 3.03, + "learning_rate": 9.900598982374032e-06, + "loss": 0.7633, + "step": 47165 + }, + { + "epoch": 3.03, + "learning_rate": 9.900384293351081e-06, + "loss": 0.8373, + "step": 47166 + }, + { + "epoch": 3.03, + "learning_rate": 9.900169604328132e-06, + "loss": 0.7528, + "step": 47167 + }, + { + "epoch": 3.03, + "learning_rate": 9.899954915305181e-06, + "loss": 0.8455, + "step": 47168 + }, + { + "epoch": 3.03, + "learning_rate": 9.89974022628223e-06, + "loss": 0.8637, + "step": 47169 + }, + { + "epoch": 3.03, + "learning_rate": 9.89952553725928e-06, + "loss": 0.798, + "step": 47170 + }, + { + "epoch": 3.03, + "learning_rate": 9.899310848236331e-06, + "loss": 0.8574, + "step": 47171 + }, + { + "epoch": 3.03, + "learning_rate": 9.89909615921338e-06, + "loss": 0.8128, + "step": 47172 + }, + { + "epoch": 3.03, + "learning_rate": 9.898881470190431e-06, + "loss": 0.8426, + "step": 47173 + }, + { + "epoch": 3.04, + "learning_rate": 9.89866678116748e-06, + "loss": 0.7918, + "step": 47174 + }, + { + "epoch": 3.04, + "learning_rate": 9.898452092144529e-06, + "loss": 0.774, + "step": 47175 + }, + { + "epoch": 3.04, + "learning_rate": 9.89823740312158e-06, + "loss": 0.8337, + "step": 47176 + }, + { + "epoch": 3.04, + "learning_rate": 9.898022714098628e-06, + "loss": 0.8143, + "step": 47177 + }, + { + "epoch": 3.04, + "learning_rate": 9.897808025075679e-06, + "loss": 0.8363, + "step": 47178 + }, + { + "epoch": 3.04, + "learning_rate": 9.897593336052728e-06, + "loss": 0.8209, + "step": 47179 + }, + { + "epoch": 3.04, + "learning_rate": 9.897378647029779e-06, + "loss": 0.8869, + "step": 47180 + }, + { + "epoch": 3.04, + "learning_rate": 9.897163958006828e-06, + "loss": 0.8564, + "step": 47181 + }, + { + "epoch": 3.04, + "learning_rate": 9.896949268983878e-06, + "loss": 0.8589, + "step": 47182 + }, + { + "epoch": 3.04, + "learning_rate": 9.896734579960927e-06, + "loss": 0.8403, + "step": 47183 + }, + { + "epoch": 3.04, + "learning_rate": 9.896519890937976e-06, + "loss": 0.7791, + "step": 47184 + }, + { + "epoch": 3.04, + "learning_rate": 9.896305201915027e-06, + "loss": 0.7993, + "step": 47185 + }, + { + "epoch": 3.04, + "learning_rate": 9.896090512892078e-06, + "loss": 0.8239, + "step": 47186 + }, + { + "epoch": 3.04, + "learning_rate": 9.895875823869127e-06, + "loss": 0.8256, + "step": 47187 + }, + { + "epoch": 3.04, + "learning_rate": 9.895661134846176e-06, + "loss": 0.8652, + "step": 47188 + }, + { + "epoch": 3.04, + "learning_rate": 9.895446445823226e-06, + "loss": 0.8614, + "step": 47189 + }, + { + "epoch": 3.04, + "learning_rate": 9.895231756800275e-06, + "loss": 0.8353, + "step": 47190 + }, + { + "epoch": 3.04, + "learning_rate": 9.895017067777324e-06, + "loss": 0.8176, + "step": 47191 + }, + { + "epoch": 3.04, + "learning_rate": 9.894802378754375e-06, + "loss": 0.812, + "step": 47192 + }, + { + "epoch": 3.04, + "learning_rate": 9.894587689731425e-06, + "loss": 0.7939, + "step": 47193 + }, + { + "epoch": 3.04, + "learning_rate": 9.894373000708474e-06, + "loss": 0.883, + "step": 47194 + }, + { + "epoch": 3.04, + "learning_rate": 9.894158311685525e-06, + "loss": 0.8535, + "step": 47195 + }, + { + "epoch": 3.04, + "learning_rate": 9.893943622662574e-06, + "loss": 0.8245, + "step": 47196 + }, + { + "epoch": 3.04, + "learning_rate": 9.893728933639623e-06, + "loss": 0.8206, + "step": 47197 + }, + { + "epoch": 3.04, + "learning_rate": 9.893514244616674e-06, + "loss": 0.7929, + "step": 47198 + }, + { + "epoch": 3.04, + "learning_rate": 9.893299555593723e-06, + "loss": 0.8257, + "step": 47199 + }, + { + "epoch": 3.04, + "learning_rate": 9.893084866570773e-06, + "loss": 0.8199, + "step": 47200 + }, + { + "epoch": 3.04, + "learning_rate": 9.892870177547822e-06, + "loss": 0.8203, + "step": 47201 + }, + { + "epoch": 3.04, + "learning_rate": 9.892655488524873e-06, + "loss": 0.7815, + "step": 47202 + }, + { + "epoch": 3.04, + "learning_rate": 9.892440799501922e-06, + "loss": 0.7932, + "step": 47203 + }, + { + "epoch": 3.04, + "learning_rate": 9.892226110478973e-06, + "loss": 0.8565, + "step": 47204 + }, + { + "epoch": 3.04, + "learning_rate": 9.892011421456022e-06, + "loss": 0.8462, + "step": 47205 + }, + { + "epoch": 3.04, + "learning_rate": 9.89179673243307e-06, + "loss": 0.8209, + "step": 47206 + }, + { + "epoch": 3.04, + "learning_rate": 9.891582043410121e-06, + "loss": 0.8615, + "step": 47207 + }, + { + "epoch": 3.04, + "learning_rate": 9.891367354387172e-06, + "loss": 0.8353, + "step": 47208 + }, + { + "epoch": 3.04, + "learning_rate": 9.89115266536422e-06, + "loss": 0.852, + "step": 47209 + }, + { + "epoch": 3.04, + "learning_rate": 9.89093797634127e-06, + "loss": 0.8187, + "step": 47210 + }, + { + "epoch": 3.04, + "learning_rate": 9.89072328731832e-06, + "loss": 0.8466, + "step": 47211 + }, + { + "epoch": 3.04, + "learning_rate": 9.89050859829537e-06, + "loss": 0.7755, + "step": 47212 + }, + { + "epoch": 3.04, + "learning_rate": 9.89029390927242e-06, + "loss": 0.9003, + "step": 47213 + }, + { + "epoch": 3.04, + "learning_rate": 9.89007922024947e-06, + "loss": 0.8623, + "step": 47214 + }, + { + "epoch": 3.04, + "learning_rate": 9.88986453122652e-06, + "loss": 0.8385, + "step": 47215 + }, + { + "epoch": 3.04, + "learning_rate": 9.889649842203569e-06, + "loss": 0.8262, + "step": 47216 + }, + { + "epoch": 3.04, + "learning_rate": 9.88943515318062e-06, + "loss": 0.811, + "step": 47217 + }, + { + "epoch": 3.04, + "learning_rate": 9.889220464157668e-06, + "loss": 0.8624, + "step": 47218 + }, + { + "epoch": 3.04, + "learning_rate": 9.889005775134717e-06, + "loss": 0.8656, + "step": 47219 + }, + { + "epoch": 3.04, + "learning_rate": 9.888791086111768e-06, + "loss": 0.8673, + "step": 47220 + }, + { + "epoch": 3.04, + "learning_rate": 9.888576397088817e-06, + "loss": 0.8504, + "step": 47221 + }, + { + "epoch": 3.04, + "learning_rate": 9.888361708065867e-06, + "loss": 0.858, + "step": 47222 + }, + { + "epoch": 3.04, + "learning_rate": 9.888147019042918e-06, + "loss": 0.7935, + "step": 47223 + }, + { + "epoch": 3.04, + "learning_rate": 9.887932330019967e-06, + "loss": 0.8364, + "step": 47224 + }, + { + "epoch": 3.04, + "learning_rate": 9.887717640997016e-06, + "loss": 0.7999, + "step": 47225 + }, + { + "epoch": 3.04, + "learning_rate": 9.887502951974067e-06, + "loss": 0.8348, + "step": 47226 + }, + { + "epoch": 3.04, + "learning_rate": 9.887288262951116e-06, + "loss": 0.819, + "step": 47227 + }, + { + "epoch": 3.04, + "learning_rate": 9.887073573928165e-06, + "loss": 0.8423, + "step": 47228 + }, + { + "epoch": 3.04, + "learning_rate": 9.886858884905215e-06, + "loss": 0.7915, + "step": 47229 + }, + { + "epoch": 3.04, + "learning_rate": 9.886644195882266e-06, + "loss": 0.8632, + "step": 47230 + }, + { + "epoch": 3.04, + "learning_rate": 9.886429506859315e-06, + "loss": 0.7904, + "step": 47231 + }, + { + "epoch": 3.04, + "learning_rate": 9.886214817836366e-06, + "loss": 0.7455, + "step": 47232 + }, + { + "epoch": 3.04, + "learning_rate": 9.886000128813415e-06, + "loss": 0.8496, + "step": 47233 + }, + { + "epoch": 3.04, + "learning_rate": 9.885785439790464e-06, + "loss": 0.8491, + "step": 47234 + }, + { + "epoch": 3.04, + "learning_rate": 9.885570750767514e-06, + "loss": 0.768, + "step": 47235 + }, + { + "epoch": 3.04, + "learning_rate": 9.885356061744565e-06, + "loss": 0.8007, + "step": 47236 + }, + { + "epoch": 3.04, + "learning_rate": 9.885141372721614e-06, + "loss": 0.8321, + "step": 47237 + }, + { + "epoch": 3.04, + "learning_rate": 9.884926683698663e-06, + "loss": 0.8197, + "step": 47238 + }, + { + "epoch": 3.04, + "learning_rate": 9.884711994675713e-06, + "loss": 0.7688, + "step": 47239 + }, + { + "epoch": 3.04, + "learning_rate": 9.884497305652762e-06, + "loss": 0.8085, + "step": 47240 + }, + { + "epoch": 3.04, + "learning_rate": 9.884282616629813e-06, + "loss": 0.8625, + "step": 47241 + }, + { + "epoch": 3.04, + "learning_rate": 9.884067927606862e-06, + "loss": 0.7899, + "step": 47242 + }, + { + "epoch": 3.04, + "learning_rate": 9.883853238583913e-06, + "loss": 0.8707, + "step": 47243 + }, + { + "epoch": 3.04, + "learning_rate": 9.883638549560962e-06, + "loss": 0.8183, + "step": 47244 + }, + { + "epoch": 3.04, + "learning_rate": 9.883423860538012e-06, + "loss": 0.8652, + "step": 47245 + }, + { + "epoch": 3.04, + "learning_rate": 9.883209171515061e-06, + "loss": 0.8455, + "step": 47246 + }, + { + "epoch": 3.04, + "learning_rate": 9.88299448249211e-06, + "loss": 0.8451, + "step": 47247 + }, + { + "epoch": 3.04, + "learning_rate": 9.882779793469161e-06, + "loss": 0.823, + "step": 47248 + }, + { + "epoch": 3.04, + "learning_rate": 9.88256510444621e-06, + "loss": 0.7879, + "step": 47249 + }, + { + "epoch": 3.04, + "learning_rate": 9.88235041542326e-06, + "loss": 0.7901, + "step": 47250 + }, + { + "epoch": 3.04, + "learning_rate": 9.882135726400311e-06, + "loss": 0.8347, + "step": 47251 + }, + { + "epoch": 3.04, + "learning_rate": 9.88192103737736e-06, + "loss": 0.8079, + "step": 47252 + }, + { + "epoch": 3.04, + "learning_rate": 9.881706348354409e-06, + "loss": 0.8436, + "step": 47253 + }, + { + "epoch": 3.04, + "learning_rate": 9.88149165933146e-06, + "loss": 0.8852, + "step": 47254 + }, + { + "epoch": 3.04, + "learning_rate": 9.881276970308509e-06, + "loss": 0.8425, + "step": 47255 + }, + { + "epoch": 3.04, + "learning_rate": 9.881062281285558e-06, + "loss": 0.8014, + "step": 47256 + }, + { + "epoch": 3.04, + "learning_rate": 9.880847592262608e-06, + "loss": 0.8072, + "step": 47257 + }, + { + "epoch": 3.04, + "learning_rate": 9.880632903239659e-06, + "loss": 0.7644, + "step": 47258 + }, + { + "epoch": 3.04, + "learning_rate": 9.880418214216708e-06, + "loss": 0.8204, + "step": 47259 + }, + { + "epoch": 3.04, + "learning_rate": 9.880203525193759e-06, + "loss": 0.7937, + "step": 47260 + }, + { + "epoch": 3.04, + "learning_rate": 9.879988836170808e-06, + "loss": 0.8535, + "step": 47261 + }, + { + "epoch": 3.04, + "learning_rate": 9.879774147147857e-06, + "loss": 0.8476, + "step": 47262 + }, + { + "epoch": 3.04, + "learning_rate": 9.879559458124907e-06, + "loss": 0.9227, + "step": 47263 + }, + { + "epoch": 3.04, + "learning_rate": 9.879344769101956e-06, + "loss": 0.8024, + "step": 47264 + }, + { + "epoch": 3.04, + "learning_rate": 9.879130080079007e-06, + "loss": 0.8412, + "step": 47265 + }, + { + "epoch": 3.04, + "learning_rate": 9.878915391056056e-06, + "loss": 0.8509, + "step": 47266 + }, + { + "epoch": 3.04, + "learning_rate": 9.878700702033106e-06, + "loss": 0.8454, + "step": 47267 + }, + { + "epoch": 3.04, + "learning_rate": 9.878486013010155e-06, + "loss": 0.7928, + "step": 47268 + }, + { + "epoch": 3.04, + "learning_rate": 9.878271323987204e-06, + "loss": 0.8904, + "step": 47269 + }, + { + "epoch": 3.04, + "learning_rate": 9.878056634964255e-06, + "loss": 0.7864, + "step": 47270 + }, + { + "epoch": 3.04, + "learning_rate": 9.877841945941304e-06, + "loss": 0.829, + "step": 47271 + }, + { + "epoch": 3.04, + "learning_rate": 9.877627256918355e-06, + "loss": 0.8215, + "step": 47272 + }, + { + "epoch": 3.04, + "learning_rate": 9.877412567895405e-06, + "loss": 0.829, + "step": 47273 + }, + { + "epoch": 3.04, + "learning_rate": 9.877197878872454e-06, + "loss": 0.7976, + "step": 47274 + }, + { + "epoch": 3.04, + "learning_rate": 9.876983189849503e-06, + "loss": 0.7789, + "step": 47275 + }, + { + "epoch": 3.04, + "learning_rate": 9.876768500826554e-06, + "loss": 0.8088, + "step": 47276 + }, + { + "epoch": 3.04, + "learning_rate": 9.876553811803603e-06, + "loss": 0.8858, + "step": 47277 + }, + { + "epoch": 3.04, + "learning_rate": 9.876339122780652e-06, + "loss": 0.7743, + "step": 47278 + }, + { + "epoch": 3.04, + "learning_rate": 9.876124433757703e-06, + "loss": 0.8234, + "step": 47279 + }, + { + "epoch": 3.04, + "learning_rate": 9.875909744734753e-06, + "loss": 0.7903, + "step": 47280 + }, + { + "epoch": 3.04, + "learning_rate": 9.875695055711802e-06, + "loss": 0.8946, + "step": 47281 + }, + { + "epoch": 3.04, + "learning_rate": 9.875480366688853e-06, + "loss": 0.797, + "step": 47282 + }, + { + "epoch": 3.04, + "learning_rate": 9.875265677665902e-06, + "loss": 0.7908, + "step": 47283 + }, + { + "epoch": 3.04, + "learning_rate": 9.87505098864295e-06, + "loss": 0.8677, + "step": 47284 + }, + { + "epoch": 3.04, + "learning_rate": 9.874836299620001e-06, + "loss": 0.8071, + "step": 47285 + }, + { + "epoch": 3.04, + "learning_rate": 9.87462161059705e-06, + "loss": 0.796, + "step": 47286 + }, + { + "epoch": 3.04, + "learning_rate": 9.874406921574101e-06, + "loss": 0.7875, + "step": 47287 + }, + { + "epoch": 3.04, + "learning_rate": 9.87419223255115e-06, + "loss": 0.8285, + "step": 47288 + }, + { + "epoch": 3.04, + "learning_rate": 9.8739775435282e-06, + "loss": 0.8005, + "step": 47289 + }, + { + "epoch": 3.04, + "learning_rate": 9.87376285450525e-06, + "loss": 0.873, + "step": 47290 + }, + { + "epoch": 3.04, + "learning_rate": 9.8735481654823e-06, + "loss": 0.85, + "step": 47291 + }, + { + "epoch": 3.04, + "learning_rate": 9.87333347645935e-06, + "loss": 0.7803, + "step": 47292 + }, + { + "epoch": 3.04, + "learning_rate": 9.873118787436398e-06, + "loss": 0.8019, + "step": 47293 + }, + { + "epoch": 3.04, + "learning_rate": 9.872904098413449e-06, + "loss": 0.8158, + "step": 47294 + }, + { + "epoch": 3.04, + "learning_rate": 9.8726894093905e-06, + "loss": 0.8531, + "step": 47295 + }, + { + "epoch": 3.04, + "learning_rate": 9.872474720367548e-06, + "loss": 0.7341, + "step": 47296 + }, + { + "epoch": 3.04, + "learning_rate": 9.872260031344597e-06, + "loss": 0.8173, + "step": 47297 + }, + { + "epoch": 3.04, + "learning_rate": 9.872045342321648e-06, + "loss": 0.8114, + "step": 47298 + }, + { + "epoch": 3.04, + "learning_rate": 9.871830653298697e-06, + "loss": 0.7722, + "step": 47299 + }, + { + "epoch": 3.04, + "learning_rate": 9.871615964275748e-06, + "loss": 0.8422, + "step": 47300 + }, + { + "epoch": 3.04, + "learning_rate": 9.871401275252798e-06, + "loss": 0.7866, + "step": 47301 + }, + { + "epoch": 3.04, + "learning_rate": 9.871186586229847e-06, + "loss": 0.8149, + "step": 47302 + }, + { + "epoch": 3.04, + "learning_rate": 9.870971897206896e-06, + "loss": 0.7926, + "step": 47303 + }, + { + "epoch": 3.04, + "learning_rate": 9.870757208183947e-06, + "loss": 0.8182, + "step": 47304 + }, + { + "epoch": 3.04, + "learning_rate": 9.870542519160996e-06, + "loss": 0.8617, + "step": 47305 + }, + { + "epoch": 3.04, + "learning_rate": 9.870327830138045e-06, + "loss": 0.8168, + "step": 47306 + }, + { + "epoch": 3.04, + "learning_rate": 9.870113141115096e-06, + "loss": 0.7737, + "step": 47307 + }, + { + "epoch": 3.04, + "learning_rate": 9.869898452092145e-06, + "loss": 0.8953, + "step": 47308 + }, + { + "epoch": 3.04, + "learning_rate": 9.869683763069195e-06, + "loss": 0.8365, + "step": 47309 + }, + { + "epoch": 3.04, + "learning_rate": 9.869469074046246e-06, + "loss": 0.8058, + "step": 47310 + }, + { + "epoch": 3.04, + "learning_rate": 9.869254385023295e-06, + "loss": 0.8382, + "step": 47311 + }, + { + "epoch": 3.04, + "learning_rate": 9.869039696000344e-06, + "loss": 0.8893, + "step": 47312 + }, + { + "epoch": 3.04, + "learning_rate": 9.868825006977394e-06, + "loss": 0.8025, + "step": 47313 + }, + { + "epoch": 3.04, + "learning_rate": 9.868610317954443e-06, + "loss": 0.7989, + "step": 47314 + }, + { + "epoch": 3.04, + "learning_rate": 9.868395628931492e-06, + "loss": 0.7986, + "step": 47315 + }, + { + "epoch": 3.04, + "learning_rate": 9.868180939908543e-06, + "loss": 0.7883, + "step": 47316 + }, + { + "epoch": 3.04, + "learning_rate": 9.867966250885594e-06, + "loss": 0.8288, + "step": 47317 + }, + { + "epoch": 3.04, + "learning_rate": 9.867751561862643e-06, + "loss": 0.8164, + "step": 47318 + }, + { + "epoch": 3.04, + "learning_rate": 9.867536872839693e-06, + "loss": 0.8255, + "step": 47319 + }, + { + "epoch": 3.04, + "learning_rate": 9.867322183816742e-06, + "loss": 0.8149, + "step": 47320 + }, + { + "epoch": 3.04, + "learning_rate": 9.867107494793791e-06, + "loss": 0.8785, + "step": 47321 + }, + { + "epoch": 3.04, + "learning_rate": 9.866892805770842e-06, + "loss": 0.7294, + "step": 47322 + }, + { + "epoch": 3.04, + "learning_rate": 9.866678116747893e-06, + "loss": 0.8611, + "step": 47323 + }, + { + "epoch": 3.04, + "learning_rate": 9.866463427724942e-06, + "loss": 0.8785, + "step": 47324 + }, + { + "epoch": 3.04, + "learning_rate": 9.86624873870199e-06, + "loss": 0.7506, + "step": 47325 + }, + { + "epoch": 3.04, + "learning_rate": 9.866034049679041e-06, + "loss": 0.827, + "step": 47326 + }, + { + "epoch": 3.04, + "learning_rate": 9.86581936065609e-06, + "loss": 0.841, + "step": 47327 + }, + { + "epoch": 3.04, + "learning_rate": 9.86560467163314e-06, + "loss": 0.853, + "step": 47328 + }, + { + "epoch": 3.05, + "learning_rate": 9.86538998261019e-06, + "loss": 0.8405, + "step": 47329 + }, + { + "epoch": 3.05, + "learning_rate": 9.86517529358724e-06, + "loss": 0.7819, + "step": 47330 + }, + { + "epoch": 3.05, + "learning_rate": 9.86496060456429e-06, + "loss": 0.8436, + "step": 47331 + }, + { + "epoch": 3.05, + "learning_rate": 9.86474591554134e-06, + "loss": 0.7756, + "step": 47332 + }, + { + "epoch": 3.05, + "learning_rate": 9.864531226518389e-06, + "loss": 0.8387, + "step": 47333 + }, + { + "epoch": 3.05, + "learning_rate": 9.864316537495438e-06, + "loss": 0.8328, + "step": 47334 + }, + { + "epoch": 3.05, + "learning_rate": 9.864101848472489e-06, + "loss": 0.8275, + "step": 47335 + }, + { + "epoch": 3.05, + "learning_rate": 9.863887159449538e-06, + "loss": 0.8774, + "step": 47336 + }, + { + "epoch": 3.05, + "learning_rate": 9.863672470426588e-06, + "loss": 0.8475, + "step": 47337 + }, + { + "epoch": 3.05, + "learning_rate": 9.863457781403639e-06, + "loss": 0.8178, + "step": 47338 + }, + { + "epoch": 3.05, + "learning_rate": 9.863243092380688e-06, + "loss": 0.8246, + "step": 47339 + }, + { + "epoch": 3.05, + "learning_rate": 9.863028403357737e-06, + "loss": 0.8212, + "step": 47340 + }, + { + "epoch": 3.05, + "learning_rate": 9.862813714334787e-06, + "loss": 0.7895, + "step": 47341 + }, + { + "epoch": 3.05, + "learning_rate": 9.862599025311836e-06, + "loss": 0.7583, + "step": 47342 + }, + { + "epoch": 3.05, + "learning_rate": 9.862384336288885e-06, + "loss": 0.819, + "step": 47343 + }, + { + "epoch": 3.05, + "learning_rate": 9.862169647265936e-06, + "loss": 0.7643, + "step": 47344 + }, + { + "epoch": 3.05, + "learning_rate": 9.861954958242987e-06, + "loss": 0.8157, + "step": 47345 + }, + { + "epoch": 3.05, + "learning_rate": 9.861740269220036e-06, + "loss": 0.7764, + "step": 47346 + }, + { + "epoch": 3.05, + "learning_rate": 9.861525580197086e-06, + "loss": 0.8239, + "step": 47347 + }, + { + "epoch": 3.05, + "learning_rate": 9.861310891174135e-06, + "loss": 0.8628, + "step": 47348 + }, + { + "epoch": 3.05, + "learning_rate": 9.861096202151184e-06, + "loss": 0.7995, + "step": 47349 + }, + { + "epoch": 3.05, + "learning_rate": 9.860881513128235e-06, + "loss": 0.8635, + "step": 47350 + }, + { + "epoch": 3.05, + "learning_rate": 9.860666824105284e-06, + "loss": 0.7729, + "step": 47351 + }, + { + "epoch": 3.05, + "learning_rate": 9.860452135082335e-06, + "loss": 0.7878, + "step": 47352 + }, + { + "epoch": 3.05, + "learning_rate": 9.860237446059384e-06, + "loss": 0.8146, + "step": 47353 + }, + { + "epoch": 3.05, + "learning_rate": 9.860022757036434e-06, + "loss": 0.8037, + "step": 47354 + }, + { + "epoch": 3.05, + "learning_rate": 9.859808068013483e-06, + "loss": 0.8327, + "step": 47355 + }, + { + "epoch": 3.05, + "learning_rate": 9.859593378990532e-06, + "loss": 0.8649, + "step": 47356 + }, + { + "epoch": 3.05, + "learning_rate": 9.859378689967583e-06, + "loss": 0.8164, + "step": 47357 + }, + { + "epoch": 3.05, + "learning_rate": 9.859164000944632e-06, + "loss": 0.9204, + "step": 47358 + }, + { + "epoch": 3.05, + "learning_rate": 9.858949311921682e-06, + "loss": 0.8306, + "step": 47359 + }, + { + "epoch": 3.05, + "learning_rate": 9.858734622898733e-06, + "loss": 0.8126, + "step": 47360 + }, + { + "epoch": 3.05, + "learning_rate": 9.858519933875782e-06, + "loss": 0.8423, + "step": 47361 + }, + { + "epoch": 3.05, + "learning_rate": 9.858305244852831e-06, + "loss": 0.7821, + "step": 47362 + }, + { + "epoch": 3.05, + "learning_rate": 9.858090555829882e-06, + "loss": 0.8467, + "step": 47363 + }, + { + "epoch": 3.05, + "learning_rate": 9.85787586680693e-06, + "loss": 0.7905, + "step": 47364 + }, + { + "epoch": 3.05, + "learning_rate": 9.85766117778398e-06, + "loss": 0.8236, + "step": 47365 + }, + { + "epoch": 3.05, + "learning_rate": 9.85744648876103e-06, + "loss": 0.8122, + "step": 47366 + }, + { + "epoch": 3.05, + "learning_rate": 9.857231799738081e-06, + "loss": 0.7968, + "step": 47367 + }, + { + "epoch": 3.05, + "learning_rate": 9.85701711071513e-06, + "loss": 0.7678, + "step": 47368 + }, + { + "epoch": 3.05, + "learning_rate": 9.85680242169218e-06, + "loss": 0.8357, + "step": 47369 + }, + { + "epoch": 3.05, + "learning_rate": 9.85658773266923e-06, + "loss": 0.7754, + "step": 47370 + }, + { + "epoch": 3.05, + "learning_rate": 9.856373043646278e-06, + "loss": 0.8221, + "step": 47371 + }, + { + "epoch": 3.05, + "learning_rate": 9.856158354623329e-06, + "loss": 0.8027, + "step": 47372 + }, + { + "epoch": 3.05, + "learning_rate": 9.855943665600378e-06, + "loss": 0.7455, + "step": 47373 + }, + { + "epoch": 3.05, + "learning_rate": 9.855728976577429e-06, + "loss": 0.8581, + "step": 47374 + }, + { + "epoch": 3.05, + "learning_rate": 9.855514287554478e-06, + "loss": 0.7301, + "step": 47375 + }, + { + "epoch": 3.05, + "learning_rate": 9.855299598531528e-06, + "loss": 0.7483, + "step": 47376 + }, + { + "epoch": 3.05, + "learning_rate": 9.855084909508577e-06, + "loss": 0.8862, + "step": 47377 + }, + { + "epoch": 3.05, + "learning_rate": 9.854870220485628e-06, + "loss": 0.795, + "step": 47378 + }, + { + "epoch": 3.05, + "learning_rate": 9.854655531462677e-06, + "loss": 0.7549, + "step": 47379 + }, + { + "epoch": 3.05, + "learning_rate": 9.854440842439726e-06, + "loss": 0.873, + "step": 47380 + }, + { + "epoch": 3.05, + "learning_rate": 9.854226153416777e-06, + "loss": 0.8533, + "step": 47381 + }, + { + "epoch": 3.05, + "learning_rate": 9.854011464393827e-06, + "loss": 0.7931, + "step": 47382 + }, + { + "epoch": 3.05, + "learning_rate": 9.853796775370876e-06, + "loss": 0.8319, + "step": 47383 + }, + { + "epoch": 3.05, + "learning_rate": 9.853582086347925e-06, + "loss": 0.8301, + "step": 47384 + }, + { + "epoch": 3.05, + "learning_rate": 9.853367397324976e-06, + "loss": 0.8385, + "step": 47385 + }, + { + "epoch": 3.05, + "learning_rate": 9.853152708302025e-06, + "loss": 0.8534, + "step": 47386 + }, + { + "epoch": 3.05, + "learning_rate": 9.852938019279075e-06, + "loss": 0.8479, + "step": 47387 + }, + { + "epoch": 3.05, + "learning_rate": 9.852723330256124e-06, + "loss": 0.8337, + "step": 47388 + }, + { + "epoch": 3.05, + "learning_rate": 9.852508641233175e-06, + "loss": 0.8542, + "step": 47389 + }, + { + "epoch": 3.05, + "learning_rate": 9.852293952210224e-06, + "loss": 0.8378, + "step": 47390 + }, + { + "epoch": 3.05, + "learning_rate": 9.852079263187275e-06, + "loss": 0.8628, + "step": 47391 + }, + { + "epoch": 3.05, + "learning_rate": 9.851864574164324e-06, + "loss": 0.8259, + "step": 47392 + }, + { + "epoch": 3.05, + "learning_rate": 9.851649885141373e-06, + "loss": 0.7802, + "step": 47393 + }, + { + "epoch": 3.05, + "learning_rate": 9.851435196118423e-06, + "loss": 0.7651, + "step": 47394 + }, + { + "epoch": 3.05, + "learning_rate": 9.851220507095472e-06, + "loss": 0.7925, + "step": 47395 + }, + { + "epoch": 3.05, + "learning_rate": 9.851005818072523e-06, + "loss": 0.7875, + "step": 47396 + }, + { + "epoch": 3.05, + "learning_rate": 9.850791129049574e-06, + "loss": 0.8155, + "step": 47397 + }, + { + "epoch": 3.05, + "learning_rate": 9.850576440026623e-06, + "loss": 0.9091, + "step": 47398 + }, + { + "epoch": 3.05, + "learning_rate": 9.850361751003671e-06, + "loss": 0.821, + "step": 47399 + }, + { + "epoch": 3.05, + "learning_rate": 9.850147061980722e-06, + "loss": 0.8072, + "step": 47400 + }, + { + "epoch": 3.05, + "learning_rate": 9.849932372957771e-06, + "loss": 0.7998, + "step": 47401 + }, + { + "epoch": 3.05, + "learning_rate": 9.84971768393482e-06, + "loss": 0.8013, + "step": 47402 + }, + { + "epoch": 3.05, + "learning_rate": 9.84950299491187e-06, + "loss": 0.8513, + "step": 47403 + }, + { + "epoch": 3.05, + "learning_rate": 9.849288305888921e-06, + "loss": 0.7954, + "step": 47404 + }, + { + "epoch": 3.05, + "learning_rate": 9.84907361686597e-06, + "loss": 0.8404, + "step": 47405 + }, + { + "epoch": 3.05, + "learning_rate": 9.848858927843021e-06, + "loss": 0.7424, + "step": 47406 + }, + { + "epoch": 3.05, + "learning_rate": 9.84864423882007e-06, + "loss": 0.8249, + "step": 47407 + }, + { + "epoch": 3.05, + "learning_rate": 9.848429549797119e-06, + "loss": 0.7501, + "step": 47408 + }, + { + "epoch": 3.05, + "learning_rate": 9.84821486077417e-06, + "loss": 0.8189, + "step": 47409 + }, + { + "epoch": 3.05, + "learning_rate": 9.84800017175122e-06, + "loss": 0.7779, + "step": 47410 + }, + { + "epoch": 3.05, + "learning_rate": 9.84778548272827e-06, + "loss": 0.8019, + "step": 47411 + }, + { + "epoch": 3.05, + "learning_rate": 9.847570793705318e-06, + "loss": 0.8445, + "step": 47412 + }, + { + "epoch": 3.05, + "learning_rate": 9.847356104682369e-06, + "loss": 0.7718, + "step": 47413 + }, + { + "epoch": 3.05, + "learning_rate": 9.847141415659418e-06, + "loss": 0.8292, + "step": 47414 + }, + { + "epoch": 3.05, + "learning_rate": 9.846926726636468e-06, + "loss": 0.8518, + "step": 47415 + }, + { + "epoch": 3.05, + "learning_rate": 9.846712037613517e-06, + "loss": 0.8755, + "step": 47416 + }, + { + "epoch": 3.05, + "learning_rate": 9.846497348590566e-06, + "loss": 0.8314, + "step": 47417 + }, + { + "epoch": 3.05, + "learning_rate": 9.846282659567617e-06, + "loss": 0.8866, + "step": 47418 + }, + { + "epoch": 3.05, + "learning_rate": 9.846067970544668e-06, + "loss": 0.8036, + "step": 47419 + }, + { + "epoch": 3.05, + "learning_rate": 9.845853281521717e-06, + "loss": 0.8354, + "step": 47420 + }, + { + "epoch": 3.05, + "learning_rate": 9.845638592498766e-06, + "loss": 0.7584, + "step": 47421 + }, + { + "epoch": 3.05, + "learning_rate": 9.845423903475816e-06, + "loss": 0.8299, + "step": 47422 + }, + { + "epoch": 3.05, + "learning_rate": 9.845209214452865e-06, + "loss": 0.8407, + "step": 47423 + }, + { + "epoch": 3.05, + "learning_rate": 9.844994525429914e-06, + "loss": 0.8641, + "step": 47424 + }, + { + "epoch": 3.05, + "learning_rate": 9.844779836406967e-06, + "loss": 0.8702, + "step": 47425 + }, + { + "epoch": 3.05, + "learning_rate": 9.844565147384016e-06, + "loss": 0.7862, + "step": 47426 + }, + { + "epoch": 3.05, + "learning_rate": 9.844350458361065e-06, + "loss": 0.8368, + "step": 47427 + }, + { + "epoch": 3.05, + "learning_rate": 9.844135769338115e-06, + "loss": 0.7596, + "step": 47428 + }, + { + "epoch": 3.05, + "learning_rate": 9.843921080315164e-06, + "loss": 0.8137, + "step": 47429 + }, + { + "epoch": 3.05, + "learning_rate": 9.843706391292213e-06, + "loss": 0.7659, + "step": 47430 + }, + { + "epoch": 3.05, + "learning_rate": 9.843491702269264e-06, + "loss": 0.8396, + "step": 47431 + }, + { + "epoch": 3.05, + "learning_rate": 9.843277013246314e-06, + "loss": 0.8566, + "step": 47432 + }, + { + "epoch": 3.05, + "learning_rate": 9.843062324223363e-06, + "loss": 0.8299, + "step": 47433 + }, + { + "epoch": 3.05, + "learning_rate": 9.842847635200412e-06, + "loss": 0.8641, + "step": 47434 + }, + { + "epoch": 3.05, + "learning_rate": 9.842632946177463e-06, + "loss": 0.7464, + "step": 47435 + }, + { + "epoch": 3.05, + "learning_rate": 9.842418257154512e-06, + "loss": 0.8484, + "step": 47436 + }, + { + "epoch": 3.05, + "learning_rate": 9.842203568131563e-06, + "loss": 0.8098, + "step": 47437 + }, + { + "epoch": 3.05, + "learning_rate": 9.841988879108612e-06, + "loss": 0.7794, + "step": 47438 + }, + { + "epoch": 3.05, + "learning_rate": 9.841774190085662e-06, + "loss": 0.7634, + "step": 47439 + }, + { + "epoch": 3.05, + "learning_rate": 9.841559501062711e-06, + "loss": 0.8107, + "step": 47440 + }, + { + "epoch": 3.05, + "learning_rate": 9.841344812039762e-06, + "loss": 0.7952, + "step": 47441 + }, + { + "epoch": 3.05, + "learning_rate": 9.841130123016811e-06, + "loss": 0.8406, + "step": 47442 + }, + { + "epoch": 3.05, + "learning_rate": 9.84091543399386e-06, + "loss": 0.8307, + "step": 47443 + }, + { + "epoch": 3.05, + "learning_rate": 9.84070074497091e-06, + "loss": 0.8448, + "step": 47444 + }, + { + "epoch": 3.05, + "learning_rate": 9.84048605594796e-06, + "loss": 0.8264, + "step": 47445 + }, + { + "epoch": 3.05, + "learning_rate": 9.84027136692501e-06, + "loss": 0.82, + "step": 47446 + }, + { + "epoch": 3.05, + "learning_rate": 9.84005667790206e-06, + "loss": 0.883, + "step": 47447 + }, + { + "epoch": 3.05, + "learning_rate": 9.83984198887911e-06, + "loss": 0.7789, + "step": 47448 + }, + { + "epoch": 3.05, + "learning_rate": 9.839627299856159e-06, + "loss": 0.8749, + "step": 47449 + }, + { + "epoch": 3.05, + "learning_rate": 9.83941261083321e-06, + "loss": 0.7966, + "step": 47450 + }, + { + "epoch": 3.05, + "learning_rate": 9.839197921810258e-06, + "loss": 0.7947, + "step": 47451 + }, + { + "epoch": 3.05, + "learning_rate": 9.838983232787307e-06, + "loss": 0.819, + "step": 47452 + }, + { + "epoch": 3.05, + "learning_rate": 9.838768543764358e-06, + "loss": 0.8535, + "step": 47453 + }, + { + "epoch": 3.05, + "learning_rate": 9.838553854741409e-06, + "loss": 0.8564, + "step": 47454 + }, + { + "epoch": 3.05, + "learning_rate": 9.838339165718458e-06, + "loss": 0.8328, + "step": 47455 + }, + { + "epoch": 3.05, + "learning_rate": 9.838124476695508e-06, + "loss": 0.8247, + "step": 47456 + }, + { + "epoch": 3.05, + "learning_rate": 9.837909787672557e-06, + "loss": 0.8181, + "step": 47457 + }, + { + "epoch": 3.05, + "learning_rate": 9.837695098649606e-06, + "loss": 0.7798, + "step": 47458 + }, + { + "epoch": 3.05, + "learning_rate": 9.837480409626657e-06, + "loss": 0.8326, + "step": 47459 + }, + { + "epoch": 3.05, + "learning_rate": 9.837265720603706e-06, + "loss": 0.884, + "step": 47460 + }, + { + "epoch": 3.05, + "learning_rate": 9.837051031580756e-06, + "loss": 0.839, + "step": 47461 + }, + { + "epoch": 3.05, + "learning_rate": 9.836836342557805e-06, + "loss": 0.8244, + "step": 47462 + }, + { + "epoch": 3.05, + "learning_rate": 9.836621653534856e-06, + "loss": 0.823, + "step": 47463 + }, + { + "epoch": 3.05, + "learning_rate": 9.836406964511905e-06, + "loss": 0.8723, + "step": 47464 + }, + { + "epoch": 3.05, + "learning_rate": 9.836192275488956e-06, + "loss": 0.8819, + "step": 47465 + }, + { + "epoch": 3.05, + "learning_rate": 9.835977586466005e-06, + "loss": 0.8014, + "step": 47466 + }, + { + "epoch": 3.05, + "learning_rate": 9.835762897443054e-06, + "loss": 0.7993, + "step": 47467 + }, + { + "epoch": 3.05, + "learning_rate": 9.835548208420104e-06, + "loss": 0.8047, + "step": 47468 + }, + { + "epoch": 3.05, + "learning_rate": 9.835333519397155e-06, + "loss": 0.8587, + "step": 47469 + }, + { + "epoch": 3.05, + "learning_rate": 9.835118830374204e-06, + "loss": 0.7787, + "step": 47470 + }, + { + "epoch": 3.05, + "learning_rate": 9.834904141351253e-06, + "loss": 0.8404, + "step": 47471 + }, + { + "epoch": 3.05, + "learning_rate": 9.834689452328304e-06, + "loss": 0.7988, + "step": 47472 + }, + { + "epoch": 3.05, + "learning_rate": 9.834474763305352e-06, + "loss": 0.8638, + "step": 47473 + }, + { + "epoch": 3.05, + "learning_rate": 9.834260074282403e-06, + "loss": 0.7833, + "step": 47474 + }, + { + "epoch": 3.05, + "learning_rate": 9.834045385259452e-06, + "loss": 0.809, + "step": 47475 + }, + { + "epoch": 3.05, + "learning_rate": 9.833830696236503e-06, + "loss": 0.8245, + "step": 47476 + }, + { + "epoch": 3.05, + "learning_rate": 9.833616007213552e-06, + "loss": 0.8467, + "step": 47477 + }, + { + "epoch": 3.05, + "learning_rate": 9.833401318190602e-06, + "loss": 0.8352, + "step": 47478 + }, + { + "epoch": 3.05, + "learning_rate": 9.833186629167651e-06, + "loss": 0.8699, + "step": 47479 + }, + { + "epoch": 3.05, + "learning_rate": 9.8329719401447e-06, + "loss": 0.8197, + "step": 47480 + }, + { + "epoch": 3.05, + "learning_rate": 9.832757251121751e-06, + "loss": 0.7835, + "step": 47481 + }, + { + "epoch": 3.05, + "learning_rate": 9.8325425620988e-06, + "loss": 0.7946, + "step": 47482 + }, + { + "epoch": 3.05, + "learning_rate": 9.83232787307585e-06, + "loss": 0.8292, + "step": 47483 + }, + { + "epoch": 3.06, + "learning_rate": 9.832113184052901e-06, + "loss": 0.8211, + "step": 47484 + }, + { + "epoch": 3.06, + "learning_rate": 9.83189849502995e-06, + "loss": 0.8025, + "step": 47485 + }, + { + "epoch": 3.06, + "learning_rate": 9.831683806007e-06, + "loss": 0.8366, + "step": 47486 + }, + { + "epoch": 3.06, + "learning_rate": 9.83146911698405e-06, + "loss": 0.8652, + "step": 47487 + }, + { + "epoch": 3.06, + "learning_rate": 9.831254427961099e-06, + "loss": 0.8452, + "step": 47488 + }, + { + "epoch": 3.06, + "learning_rate": 9.831039738938148e-06, + "loss": 0.8348, + "step": 47489 + }, + { + "epoch": 3.06, + "learning_rate": 9.830825049915198e-06, + "loss": 0.8218, + "step": 47490 + }, + { + "epoch": 3.06, + "learning_rate": 9.830610360892249e-06, + "loss": 0.8665, + "step": 47491 + }, + { + "epoch": 3.06, + "learning_rate": 9.830395671869298e-06, + "loss": 0.8782, + "step": 47492 + }, + { + "epoch": 3.06, + "learning_rate": 9.830180982846349e-06, + "loss": 0.8155, + "step": 47493 + }, + { + "epoch": 3.06, + "learning_rate": 9.829966293823398e-06, + "loss": 0.8125, + "step": 47494 + }, + { + "epoch": 3.06, + "learning_rate": 9.829751604800447e-06, + "loss": 0.8113, + "step": 47495 + }, + { + "epoch": 3.06, + "learning_rate": 9.829536915777497e-06, + "loss": 0.8268, + "step": 47496 + }, + { + "epoch": 3.06, + "learning_rate": 9.829322226754548e-06, + "loss": 0.8003, + "step": 47497 + }, + { + "epoch": 3.06, + "learning_rate": 9.829107537731597e-06, + "loss": 0.8263, + "step": 47498 + }, + { + "epoch": 3.06, + "learning_rate": 9.828892848708646e-06, + "loss": 0.8888, + "step": 47499 + }, + { + "epoch": 3.06, + "learning_rate": 9.828678159685697e-06, + "loss": 0.844, + "step": 47500 + }, + { + "epoch": 3.06, + "learning_rate": 9.828463470662746e-06, + "loss": 0.7787, + "step": 47501 + }, + { + "epoch": 3.06, + "learning_rate": 9.828248781639796e-06, + "loss": 0.815, + "step": 47502 + }, + { + "epoch": 3.06, + "learning_rate": 9.828034092616845e-06, + "loss": 0.7706, + "step": 47503 + }, + { + "epoch": 3.06, + "learning_rate": 9.827819403593894e-06, + "loss": 0.7925, + "step": 47504 + }, + { + "epoch": 3.06, + "learning_rate": 9.827604714570945e-06, + "loss": 0.8559, + "step": 47505 + }, + { + "epoch": 3.06, + "learning_rate": 9.827390025547995e-06, + "loss": 0.8151, + "step": 47506 + }, + { + "epoch": 3.06, + "learning_rate": 9.827175336525044e-06, + "loss": 0.7722, + "step": 47507 + }, + { + "epoch": 3.06, + "learning_rate": 9.826960647502093e-06, + "loss": 0.8018, + "step": 47508 + }, + { + "epoch": 3.06, + "learning_rate": 9.826745958479144e-06, + "loss": 0.8623, + "step": 47509 + }, + { + "epoch": 3.06, + "learning_rate": 9.826531269456193e-06, + "loss": 0.8138, + "step": 47510 + }, + { + "epoch": 3.06, + "learning_rate": 9.826316580433242e-06, + "loss": 0.8472, + "step": 47511 + }, + { + "epoch": 3.06, + "learning_rate": 9.826101891410294e-06, + "loss": 0.7635, + "step": 47512 + }, + { + "epoch": 3.06, + "learning_rate": 9.825887202387343e-06, + "loss": 0.8748, + "step": 47513 + }, + { + "epoch": 3.06, + "learning_rate": 9.825672513364392e-06, + "loss": 0.7704, + "step": 47514 + }, + { + "epoch": 3.06, + "learning_rate": 9.825457824341443e-06, + "loss": 0.8063, + "step": 47515 + }, + { + "epoch": 3.06, + "learning_rate": 9.825243135318492e-06, + "loss": 0.7985, + "step": 47516 + }, + { + "epoch": 3.06, + "learning_rate": 9.82502844629554e-06, + "loss": 0.8491, + "step": 47517 + }, + { + "epoch": 3.06, + "learning_rate": 9.824813757272591e-06, + "loss": 0.8599, + "step": 47518 + }, + { + "epoch": 3.06, + "learning_rate": 9.824599068249642e-06, + "loss": 0.8065, + "step": 47519 + }, + { + "epoch": 3.06, + "learning_rate": 9.824384379226691e-06, + "loss": 0.795, + "step": 47520 + }, + { + "epoch": 3.06, + "learning_rate": 9.82416969020374e-06, + "loss": 0.9183, + "step": 47521 + }, + { + "epoch": 3.06, + "learning_rate": 9.82395500118079e-06, + "loss": 0.8648, + "step": 47522 + }, + { + "epoch": 3.06, + "learning_rate": 9.82374031215784e-06, + "loss": 0.7868, + "step": 47523 + }, + { + "epoch": 3.06, + "learning_rate": 9.82352562313489e-06, + "loss": 0.7999, + "step": 47524 + }, + { + "epoch": 3.06, + "learning_rate": 9.82331093411194e-06, + "loss": 0.7802, + "step": 47525 + }, + { + "epoch": 3.06, + "learning_rate": 9.82309624508899e-06, + "loss": 0.8613, + "step": 47526 + }, + { + "epoch": 3.06, + "learning_rate": 9.822881556066039e-06, + "loss": 0.8696, + "step": 47527 + }, + { + "epoch": 3.06, + "learning_rate": 9.82266686704309e-06, + "loss": 0.8159, + "step": 47528 + }, + { + "epoch": 3.06, + "learning_rate": 9.822452178020139e-06, + "loss": 0.7925, + "step": 47529 + }, + { + "epoch": 3.06, + "learning_rate": 9.822237488997188e-06, + "loss": 0.8208, + "step": 47530 + }, + { + "epoch": 3.06, + "learning_rate": 9.822022799974238e-06, + "loss": 0.8357, + "step": 47531 + }, + { + "epoch": 3.06, + "learning_rate": 9.821808110951287e-06, + "loss": 0.9126, + "step": 47532 + }, + { + "epoch": 3.06, + "learning_rate": 9.821593421928338e-06, + "loss": 0.8788, + "step": 47533 + }, + { + "epoch": 3.06, + "learning_rate": 9.821378732905388e-06, + "loss": 0.7545, + "step": 47534 + }, + { + "epoch": 3.06, + "learning_rate": 9.821164043882437e-06, + "loss": 0.913, + "step": 47535 + }, + { + "epoch": 3.06, + "learning_rate": 9.820949354859486e-06, + "loss": 0.7985, + "step": 47536 + }, + { + "epoch": 3.06, + "learning_rate": 9.820734665836537e-06, + "loss": 0.7551, + "step": 47537 + }, + { + "epoch": 3.06, + "learning_rate": 9.820519976813586e-06, + "loss": 0.8647, + "step": 47538 + }, + { + "epoch": 3.06, + "learning_rate": 9.820305287790635e-06, + "loss": 0.8002, + "step": 47539 + }, + { + "epoch": 3.06, + "learning_rate": 9.820090598767686e-06, + "loss": 0.8064, + "step": 47540 + }, + { + "epoch": 3.06, + "learning_rate": 9.819875909744736e-06, + "loss": 0.8154, + "step": 47541 + }, + { + "epoch": 3.06, + "learning_rate": 9.819661220721785e-06, + "loss": 0.8367, + "step": 47542 + }, + { + "epoch": 3.06, + "learning_rate": 9.819446531698836e-06, + "loss": 0.7901, + "step": 47543 + }, + { + "epoch": 3.06, + "learning_rate": 9.819231842675885e-06, + "loss": 0.7746, + "step": 47544 + }, + { + "epoch": 3.06, + "learning_rate": 9.819017153652934e-06, + "loss": 0.8215, + "step": 47545 + }, + { + "epoch": 3.06, + "learning_rate": 9.818802464629985e-06, + "loss": 0.7641, + "step": 47546 + }, + { + "epoch": 3.06, + "learning_rate": 9.818587775607033e-06, + "loss": 0.8225, + "step": 47547 + }, + { + "epoch": 3.06, + "learning_rate": 9.818373086584084e-06, + "loss": 0.8817, + "step": 47548 + }, + { + "epoch": 3.06, + "learning_rate": 9.818158397561133e-06, + "loss": 0.772, + "step": 47549 + }, + { + "epoch": 3.06, + "learning_rate": 9.817943708538184e-06, + "loss": 0.8013, + "step": 47550 + }, + { + "epoch": 3.06, + "learning_rate": 9.817729019515233e-06, + "loss": 0.8321, + "step": 47551 + }, + { + "epoch": 3.06, + "learning_rate": 9.817514330492283e-06, + "loss": 0.8188, + "step": 47552 + }, + { + "epoch": 3.06, + "learning_rate": 9.817299641469332e-06, + "loss": 0.8328, + "step": 47553 + }, + { + "epoch": 3.06, + "learning_rate": 9.817084952446381e-06, + "loss": 0.804, + "step": 47554 + }, + { + "epoch": 3.06, + "learning_rate": 9.816870263423432e-06, + "loss": 0.8038, + "step": 47555 + }, + { + "epoch": 3.06, + "learning_rate": 9.816655574400483e-06, + "loss": 0.8014, + "step": 47556 + }, + { + "epoch": 3.06, + "learning_rate": 9.816440885377532e-06, + "loss": 0.8118, + "step": 47557 + }, + { + "epoch": 3.06, + "learning_rate": 9.81622619635458e-06, + "loss": 0.7713, + "step": 47558 + }, + { + "epoch": 3.06, + "learning_rate": 9.816011507331631e-06, + "loss": 0.8652, + "step": 47559 + }, + { + "epoch": 3.06, + "learning_rate": 9.81579681830868e-06, + "loss": 0.8414, + "step": 47560 + }, + { + "epoch": 3.06, + "learning_rate": 9.815582129285731e-06, + "loss": 0.864, + "step": 47561 + }, + { + "epoch": 3.06, + "learning_rate": 9.81536744026278e-06, + "loss": 0.8169, + "step": 47562 + }, + { + "epoch": 3.06, + "learning_rate": 9.81515275123983e-06, + "loss": 0.8124, + "step": 47563 + }, + { + "epoch": 3.06, + "learning_rate": 9.81493806221688e-06, + "loss": 0.8317, + "step": 47564 + }, + { + "epoch": 3.06, + "learning_rate": 9.81472337319393e-06, + "loss": 0.7967, + "step": 47565 + }, + { + "epoch": 3.06, + "learning_rate": 9.814508684170979e-06, + "loss": 0.821, + "step": 47566 + }, + { + "epoch": 3.06, + "learning_rate": 9.814293995148028e-06, + "loss": 0.8347, + "step": 47567 + }, + { + "epoch": 3.06, + "learning_rate": 9.814079306125079e-06, + "loss": 0.7751, + "step": 47568 + }, + { + "epoch": 3.06, + "learning_rate": 9.813864617102128e-06, + "loss": 0.813, + "step": 47569 + }, + { + "epoch": 3.06, + "learning_rate": 9.813649928079178e-06, + "loss": 0.8411, + "step": 47570 + }, + { + "epoch": 3.06, + "learning_rate": 9.813435239056229e-06, + "loss": 0.7762, + "step": 47571 + }, + { + "epoch": 3.06, + "learning_rate": 9.813220550033278e-06, + "loss": 0.7857, + "step": 47572 + }, + { + "epoch": 3.06, + "learning_rate": 9.813005861010327e-06, + "loss": 0.8084, + "step": 47573 + }, + { + "epoch": 3.06, + "learning_rate": 9.812791171987378e-06, + "loss": 0.7914, + "step": 47574 + }, + { + "epoch": 3.06, + "learning_rate": 9.812576482964427e-06, + "loss": 0.8247, + "step": 47575 + }, + { + "epoch": 3.06, + "learning_rate": 9.812361793941475e-06, + "loss": 0.8634, + "step": 47576 + }, + { + "epoch": 3.06, + "learning_rate": 9.812147104918526e-06, + "loss": 0.8451, + "step": 47577 + }, + { + "epoch": 3.06, + "learning_rate": 9.811932415895577e-06, + "loss": 0.8204, + "step": 47578 + }, + { + "epoch": 3.06, + "learning_rate": 9.811717726872626e-06, + "loss": 0.8065, + "step": 47579 + }, + { + "epoch": 3.06, + "learning_rate": 9.811503037849676e-06, + "loss": 0.7792, + "step": 47580 + }, + { + "epoch": 3.06, + "learning_rate": 9.811288348826725e-06, + "loss": 0.7912, + "step": 47581 + }, + { + "epoch": 3.06, + "learning_rate": 9.811073659803774e-06, + "loss": 0.7899, + "step": 47582 + }, + { + "epoch": 3.06, + "learning_rate": 9.810858970780825e-06, + "loss": 0.8246, + "step": 47583 + }, + { + "epoch": 3.06, + "learning_rate": 9.810644281757874e-06, + "loss": 0.7495, + "step": 47584 + }, + { + "epoch": 3.06, + "learning_rate": 9.810429592734925e-06, + "loss": 0.8145, + "step": 47585 + }, + { + "epoch": 3.06, + "learning_rate": 9.810214903711974e-06, + "loss": 0.8079, + "step": 47586 + }, + { + "epoch": 3.06, + "learning_rate": 9.810000214689024e-06, + "loss": 0.8061, + "step": 47587 + }, + { + "epoch": 3.06, + "learning_rate": 9.809785525666073e-06, + "loss": 0.8303, + "step": 47588 + }, + { + "epoch": 3.06, + "learning_rate": 9.809570836643122e-06, + "loss": 0.8065, + "step": 47589 + }, + { + "epoch": 3.06, + "learning_rate": 9.809356147620173e-06, + "loss": 0.8233, + "step": 47590 + }, + { + "epoch": 3.06, + "learning_rate": 9.809141458597222e-06, + "loss": 0.8254, + "step": 47591 + }, + { + "epoch": 3.06, + "learning_rate": 9.808926769574272e-06, + "loss": 0.8299, + "step": 47592 + }, + { + "epoch": 3.06, + "learning_rate": 9.808712080551323e-06, + "loss": 0.8073, + "step": 47593 + }, + { + "epoch": 3.06, + "learning_rate": 9.808497391528372e-06, + "loss": 0.7823, + "step": 47594 + }, + { + "epoch": 3.06, + "learning_rate": 9.808282702505421e-06, + "loss": 0.854, + "step": 47595 + }, + { + "epoch": 3.06, + "learning_rate": 9.808068013482472e-06, + "loss": 0.8507, + "step": 47596 + }, + { + "epoch": 3.06, + "learning_rate": 9.80785332445952e-06, + "loss": 0.7842, + "step": 47597 + }, + { + "epoch": 3.06, + "learning_rate": 9.80763863543657e-06, + "loss": 0.8619, + "step": 47598 + }, + { + "epoch": 3.06, + "learning_rate": 9.80742394641362e-06, + "loss": 0.8007, + "step": 47599 + }, + { + "epoch": 3.06, + "learning_rate": 9.807209257390671e-06, + "loss": 0.8164, + "step": 47600 + }, + { + "epoch": 3.06, + "learning_rate": 9.80699456836772e-06, + "loss": 0.8003, + "step": 47601 + }, + { + "epoch": 3.06, + "learning_rate": 9.80677987934477e-06, + "loss": 0.813, + "step": 47602 + }, + { + "epoch": 3.06, + "learning_rate": 9.80656519032182e-06, + "loss": 0.8836, + "step": 47603 + }, + { + "epoch": 3.06, + "learning_rate": 9.806350501298869e-06, + "loss": 0.8473, + "step": 47604 + }, + { + "epoch": 3.06, + "learning_rate": 9.80613581227592e-06, + "loss": 0.8328, + "step": 47605 + }, + { + "epoch": 3.06, + "learning_rate": 9.80592112325297e-06, + "loss": 0.8014, + "step": 47606 + }, + { + "epoch": 3.06, + "learning_rate": 9.805706434230019e-06, + "loss": 0.8591, + "step": 47607 + }, + { + "epoch": 3.06, + "learning_rate": 9.805491745207068e-06, + "loss": 0.8543, + "step": 47608 + }, + { + "epoch": 3.06, + "learning_rate": 9.805277056184118e-06, + "loss": 0.8635, + "step": 47609 + }, + { + "epoch": 3.06, + "learning_rate": 9.805062367161167e-06, + "loss": 0.8284, + "step": 47610 + }, + { + "epoch": 3.06, + "learning_rate": 9.804847678138218e-06, + "loss": 0.855, + "step": 47611 + }, + { + "epoch": 3.06, + "learning_rate": 9.804632989115267e-06, + "loss": 0.8226, + "step": 47612 + }, + { + "epoch": 3.06, + "learning_rate": 9.804418300092316e-06, + "loss": 0.837, + "step": 47613 + }, + { + "epoch": 3.06, + "learning_rate": 9.804203611069367e-06, + "loss": 0.8142, + "step": 47614 + }, + { + "epoch": 3.06, + "learning_rate": 9.803988922046417e-06, + "loss": 0.7688, + "step": 47615 + }, + { + "epoch": 3.06, + "learning_rate": 9.803774233023466e-06, + "loss": 0.8672, + "step": 47616 + }, + { + "epoch": 3.06, + "learning_rate": 9.803559544000515e-06, + "loss": 0.7855, + "step": 47617 + }, + { + "epoch": 3.06, + "learning_rate": 9.803344854977566e-06, + "loss": 0.8359, + "step": 47618 + }, + { + "epoch": 3.06, + "learning_rate": 9.803130165954615e-06, + "loss": 0.8064, + "step": 47619 + }, + { + "epoch": 3.06, + "learning_rate": 9.802915476931666e-06, + "loss": 0.8059, + "step": 47620 + }, + { + "epoch": 3.06, + "learning_rate": 9.802700787908716e-06, + "loss": 0.8009, + "step": 47621 + }, + { + "epoch": 3.06, + "learning_rate": 9.802486098885765e-06, + "loss": 0.8485, + "step": 47622 + }, + { + "epoch": 3.06, + "learning_rate": 9.802271409862814e-06, + "loss": 0.7796, + "step": 47623 + }, + { + "epoch": 3.06, + "learning_rate": 9.802056720839865e-06, + "loss": 0.8854, + "step": 47624 + }, + { + "epoch": 3.06, + "learning_rate": 9.801842031816914e-06, + "loss": 0.854, + "step": 47625 + }, + { + "epoch": 3.06, + "learning_rate": 9.801627342793963e-06, + "loss": 0.8115, + "step": 47626 + }, + { + "epoch": 3.06, + "learning_rate": 9.801412653771013e-06, + "loss": 0.7998, + "step": 47627 + }, + { + "epoch": 3.06, + "learning_rate": 9.801197964748064e-06, + "loss": 0.857, + "step": 47628 + }, + { + "epoch": 3.06, + "learning_rate": 9.800983275725113e-06, + "loss": 0.7798, + "step": 47629 + }, + { + "epoch": 3.06, + "learning_rate": 9.800768586702164e-06, + "loss": 0.8692, + "step": 47630 + }, + { + "epoch": 3.06, + "learning_rate": 9.800553897679213e-06, + "loss": 0.8006, + "step": 47631 + }, + { + "epoch": 3.06, + "learning_rate": 9.800339208656262e-06, + "loss": 0.8335, + "step": 47632 + }, + { + "epoch": 3.06, + "learning_rate": 9.800124519633312e-06, + "loss": 0.8577, + "step": 47633 + }, + { + "epoch": 3.06, + "learning_rate": 9.799909830610361e-06, + "loss": 0.8396, + "step": 47634 + }, + { + "epoch": 3.06, + "learning_rate": 9.799695141587412e-06, + "loss": 0.8667, + "step": 47635 + }, + { + "epoch": 3.06, + "learning_rate": 9.79948045256446e-06, + "loss": 0.8735, + "step": 47636 + }, + { + "epoch": 3.06, + "learning_rate": 9.799265763541511e-06, + "loss": 0.7857, + "step": 47637 + }, + { + "epoch": 3.06, + "learning_rate": 9.79905107451856e-06, + "loss": 0.8755, + "step": 47638 + }, + { + "epoch": 3.06, + "learning_rate": 9.798836385495611e-06, + "loss": 0.8878, + "step": 47639 + }, + { + "epoch": 3.07, + "learning_rate": 9.79862169647266e-06, + "loss": 0.8123, + "step": 47640 + }, + { + "epoch": 3.07, + "learning_rate": 9.798407007449709e-06, + "loss": 0.8099, + "step": 47641 + }, + { + "epoch": 3.07, + "learning_rate": 9.79819231842676e-06, + "loss": 0.8006, + "step": 47642 + }, + { + "epoch": 3.07, + "learning_rate": 9.79797762940381e-06, + "loss": 0.8151, + "step": 47643 + }, + { + "epoch": 3.07, + "learning_rate": 9.79776294038086e-06, + "loss": 0.7905, + "step": 47644 + }, + { + "epoch": 3.07, + "learning_rate": 9.797548251357908e-06, + "loss": 0.8704, + "step": 47645 + }, + { + "epoch": 3.07, + "learning_rate": 9.797333562334959e-06, + "loss": 0.8253, + "step": 47646 + }, + { + "epoch": 3.07, + "learning_rate": 9.797118873312008e-06, + "loss": 0.7336, + "step": 47647 + }, + { + "epoch": 3.07, + "learning_rate": 9.796904184289059e-06, + "loss": 0.8299, + "step": 47648 + }, + { + "epoch": 3.07, + "learning_rate": 9.796689495266108e-06, + "loss": 0.863, + "step": 47649 + }, + { + "epoch": 3.07, + "learning_rate": 9.796474806243158e-06, + "loss": 0.8163, + "step": 47650 + }, + { + "epoch": 3.07, + "learning_rate": 9.796260117220207e-06, + "loss": 0.8305, + "step": 47651 + }, + { + "epoch": 3.07, + "learning_rate": 9.796045428197258e-06, + "loss": 0.8605, + "step": 47652 + }, + { + "epoch": 3.07, + "learning_rate": 9.795830739174307e-06, + "loss": 0.8776, + "step": 47653 + }, + { + "epoch": 3.07, + "learning_rate": 9.795616050151356e-06, + "loss": 0.7836, + "step": 47654 + }, + { + "epoch": 3.07, + "learning_rate": 9.795401361128406e-06, + "loss": 0.8338, + "step": 47655 + }, + { + "epoch": 3.07, + "learning_rate": 9.795186672105455e-06, + "loss": 0.8636, + "step": 47656 + }, + { + "epoch": 3.07, + "learning_rate": 9.794971983082506e-06, + "loss": 0.8454, + "step": 47657 + }, + { + "epoch": 3.07, + "learning_rate": 9.794757294059557e-06, + "loss": 0.8579, + "step": 47658 + }, + { + "epoch": 3.07, + "learning_rate": 9.794542605036606e-06, + "loss": 0.8056, + "step": 47659 + }, + { + "epoch": 3.07, + "learning_rate": 9.794327916013655e-06, + "loss": 0.7922, + "step": 47660 + }, + { + "epoch": 3.07, + "learning_rate": 9.794113226990705e-06, + "loss": 0.7903, + "step": 47661 + }, + { + "epoch": 3.07, + "learning_rate": 9.793898537967754e-06, + "loss": 0.8288, + "step": 47662 + }, + { + "epoch": 3.07, + "learning_rate": 9.793683848944803e-06, + "loss": 0.8537, + "step": 47663 + }, + { + "epoch": 3.07, + "learning_rate": 9.793469159921854e-06, + "loss": 0.8541, + "step": 47664 + }, + { + "epoch": 3.07, + "learning_rate": 9.793254470898905e-06, + "loss": 0.8045, + "step": 47665 + }, + { + "epoch": 3.07, + "learning_rate": 9.793039781875953e-06, + "loss": 0.8107, + "step": 47666 + }, + { + "epoch": 3.07, + "learning_rate": 9.792825092853004e-06, + "loss": 0.8706, + "step": 47667 + }, + { + "epoch": 3.07, + "learning_rate": 9.792610403830053e-06, + "loss": 0.877, + "step": 47668 + }, + { + "epoch": 3.07, + "learning_rate": 9.792395714807102e-06, + "loss": 0.8204, + "step": 47669 + }, + { + "epoch": 3.07, + "learning_rate": 9.792181025784153e-06, + "loss": 0.8226, + "step": 47670 + }, + { + "epoch": 3.07, + "learning_rate": 9.791966336761202e-06, + "loss": 0.8035, + "step": 47671 + }, + { + "epoch": 3.07, + "learning_rate": 9.791751647738252e-06, + "loss": 0.8224, + "step": 47672 + }, + { + "epoch": 3.07, + "learning_rate": 9.791536958715301e-06, + "loss": 0.8142, + "step": 47673 + }, + { + "epoch": 3.07, + "learning_rate": 9.791322269692352e-06, + "loss": 0.8506, + "step": 47674 + }, + { + "epoch": 3.07, + "learning_rate": 9.791107580669401e-06, + "loss": 0.7758, + "step": 47675 + }, + { + "epoch": 3.07, + "learning_rate": 9.79089289164645e-06, + "loss": 0.8451, + "step": 47676 + }, + { + "epoch": 3.07, + "learning_rate": 9.7906782026235e-06, + "loss": 0.8018, + "step": 47677 + }, + { + "epoch": 3.07, + "learning_rate": 9.79046351360055e-06, + "loss": 0.8347, + "step": 47678 + }, + { + "epoch": 3.07, + "learning_rate": 9.7902488245776e-06, + "loss": 0.8261, + "step": 47679 + }, + { + "epoch": 3.07, + "learning_rate": 9.790034135554651e-06, + "loss": 0.8204, + "step": 47680 + }, + { + "epoch": 3.07, + "learning_rate": 9.7898194465317e-06, + "loss": 0.7791, + "step": 47681 + }, + { + "epoch": 3.07, + "learning_rate": 9.789604757508749e-06, + "loss": 0.853, + "step": 47682 + }, + { + "epoch": 3.07, + "learning_rate": 9.7893900684858e-06, + "loss": 0.8234, + "step": 47683 + }, + { + "epoch": 3.07, + "learning_rate": 9.789175379462848e-06, + "loss": 0.7896, + "step": 47684 + }, + { + "epoch": 3.07, + "learning_rate": 9.788960690439897e-06, + "loss": 0.773, + "step": 47685 + }, + { + "epoch": 3.07, + "learning_rate": 9.788746001416948e-06, + "loss": 0.8257, + "step": 47686 + }, + { + "epoch": 3.07, + "learning_rate": 9.788531312393999e-06, + "loss": 0.8362, + "step": 47687 + }, + { + "epoch": 3.07, + "learning_rate": 9.788316623371048e-06, + "loss": 0.7806, + "step": 47688 + }, + { + "epoch": 3.07, + "learning_rate": 9.788101934348098e-06, + "loss": 0.7936, + "step": 47689 + }, + { + "epoch": 3.07, + "learning_rate": 9.787887245325147e-06, + "loss": 0.8718, + "step": 47690 + }, + { + "epoch": 3.07, + "learning_rate": 9.787672556302196e-06, + "loss": 0.7986, + "step": 47691 + }, + { + "epoch": 3.07, + "learning_rate": 9.787457867279247e-06, + "loss": 0.7776, + "step": 47692 + }, + { + "epoch": 3.07, + "learning_rate": 9.787243178256298e-06, + "loss": 0.8373, + "step": 47693 + }, + { + "epoch": 3.07, + "learning_rate": 9.787028489233347e-06, + "loss": 0.815, + "step": 47694 + }, + { + "epoch": 3.07, + "learning_rate": 9.786813800210395e-06, + "loss": 0.8369, + "step": 47695 + }, + { + "epoch": 3.07, + "learning_rate": 9.786599111187446e-06, + "loss": 0.8249, + "step": 47696 + }, + { + "epoch": 3.07, + "learning_rate": 9.786384422164495e-06, + "loss": 0.864, + "step": 47697 + }, + { + "epoch": 3.07, + "learning_rate": 9.786169733141546e-06, + "loss": 0.7938, + "step": 47698 + }, + { + "epoch": 3.07, + "learning_rate": 9.785955044118595e-06, + "loss": 0.8059, + "step": 47699 + }, + { + "epoch": 3.07, + "learning_rate": 9.785740355095644e-06, + "loss": 0.885, + "step": 47700 + }, + { + "epoch": 3.07, + "learning_rate": 9.785525666072694e-06, + "loss": 0.7897, + "step": 47701 + }, + { + "epoch": 3.07, + "learning_rate": 9.785310977049745e-06, + "loss": 0.8153, + "step": 47702 + }, + { + "epoch": 3.07, + "learning_rate": 9.785096288026794e-06, + "loss": 0.814, + "step": 47703 + }, + { + "epoch": 3.07, + "learning_rate": 9.784881599003843e-06, + "loss": 0.8827, + "step": 47704 + }, + { + "epoch": 3.07, + "learning_rate": 9.784666909980894e-06, + "loss": 0.8365, + "step": 47705 + }, + { + "epoch": 3.07, + "learning_rate": 9.784452220957943e-06, + "loss": 0.7401, + "step": 47706 + }, + { + "epoch": 3.07, + "learning_rate": 9.784237531934993e-06, + "loss": 0.833, + "step": 47707 + }, + { + "epoch": 3.07, + "learning_rate": 9.784022842912044e-06, + "loss": 0.8778, + "step": 47708 + }, + { + "epoch": 3.07, + "learning_rate": 9.783808153889093e-06, + "loss": 0.8246, + "step": 47709 + }, + { + "epoch": 3.07, + "learning_rate": 9.783593464866142e-06, + "loss": 0.8159, + "step": 47710 + }, + { + "epoch": 3.07, + "learning_rate": 9.783378775843192e-06, + "loss": 0.8401, + "step": 47711 + }, + { + "epoch": 3.07, + "learning_rate": 9.783164086820241e-06, + "loss": 0.7718, + "step": 47712 + }, + { + "epoch": 3.07, + "learning_rate": 9.78294939779729e-06, + "loss": 0.7778, + "step": 47713 + }, + { + "epoch": 3.07, + "learning_rate": 9.782734708774341e-06, + "loss": 0.834, + "step": 47714 + }, + { + "epoch": 3.07, + "learning_rate": 9.782520019751392e-06, + "loss": 0.7513, + "step": 47715 + }, + { + "epoch": 3.07, + "learning_rate": 9.78230533072844e-06, + "loss": 0.8297, + "step": 47716 + }, + { + "epoch": 3.07, + "learning_rate": 9.782090641705491e-06, + "loss": 0.817, + "step": 47717 + }, + { + "epoch": 3.07, + "learning_rate": 9.78187595268254e-06, + "loss": 0.875, + "step": 47718 + }, + { + "epoch": 3.07, + "learning_rate": 9.78166126365959e-06, + "loss": 0.8223, + "step": 47719 + }, + { + "epoch": 3.07, + "learning_rate": 9.78144657463664e-06, + "loss": 0.7847, + "step": 47720 + }, + { + "epoch": 3.07, + "learning_rate": 9.781231885613689e-06, + "loss": 0.7764, + "step": 47721 + }, + { + "epoch": 3.07, + "learning_rate": 9.781017196590738e-06, + "loss": 0.8256, + "step": 47722 + }, + { + "epoch": 3.07, + "learning_rate": 9.780802507567789e-06, + "loss": 0.8415, + "step": 47723 + }, + { + "epoch": 3.07, + "learning_rate": 9.78058781854484e-06, + "loss": 0.8565, + "step": 47724 + }, + { + "epoch": 3.07, + "learning_rate": 9.780373129521888e-06, + "loss": 0.7783, + "step": 47725 + }, + { + "epoch": 3.07, + "learning_rate": 9.780158440498939e-06, + "loss": 0.8568, + "step": 47726 + }, + { + "epoch": 3.07, + "learning_rate": 9.779943751475988e-06, + "loss": 0.7683, + "step": 47727 + }, + { + "epoch": 3.07, + "learning_rate": 9.779729062453037e-06, + "loss": 0.7988, + "step": 47728 + }, + { + "epoch": 3.07, + "learning_rate": 9.779514373430087e-06, + "loss": 0.8204, + "step": 47729 + }, + { + "epoch": 3.07, + "learning_rate": 9.779299684407138e-06, + "loss": 0.7644, + "step": 47730 + }, + { + "epoch": 3.07, + "learning_rate": 9.779084995384187e-06, + "loss": 0.8561, + "step": 47731 + }, + { + "epoch": 3.07, + "learning_rate": 9.778870306361236e-06, + "loss": 0.8628, + "step": 47732 + }, + { + "epoch": 3.07, + "learning_rate": 9.778655617338287e-06, + "loss": 0.7629, + "step": 47733 + }, + { + "epoch": 3.07, + "learning_rate": 9.778440928315336e-06, + "loss": 0.7897, + "step": 47734 + }, + { + "epoch": 3.07, + "learning_rate": 9.778226239292386e-06, + "loss": 0.8722, + "step": 47735 + }, + { + "epoch": 3.07, + "learning_rate": 9.778011550269435e-06, + "loss": 0.821, + "step": 47736 + }, + { + "epoch": 3.07, + "learning_rate": 9.777796861246486e-06, + "loss": 0.835, + "step": 47737 + }, + { + "epoch": 3.07, + "learning_rate": 9.777582172223535e-06, + "loss": 0.7778, + "step": 47738 + }, + { + "epoch": 3.07, + "learning_rate": 9.777367483200586e-06, + "loss": 0.7802, + "step": 47739 + }, + { + "epoch": 3.07, + "learning_rate": 9.777152794177634e-06, + "loss": 0.8497, + "step": 47740 + }, + { + "epoch": 3.07, + "learning_rate": 9.776938105154683e-06, + "loss": 0.8387, + "step": 47741 + }, + { + "epoch": 3.07, + "learning_rate": 9.776723416131734e-06, + "loss": 0.8699, + "step": 47742 + }, + { + "epoch": 3.07, + "learning_rate": 9.776508727108783e-06, + "loss": 0.777, + "step": 47743 + }, + { + "epoch": 3.07, + "learning_rate": 9.776294038085834e-06, + "loss": 0.8126, + "step": 47744 + }, + { + "epoch": 3.07, + "learning_rate": 9.776079349062884e-06, + "loss": 0.852, + "step": 47745 + }, + { + "epoch": 3.07, + "learning_rate": 9.775864660039933e-06, + "loss": 0.8373, + "step": 47746 + }, + { + "epoch": 3.07, + "learning_rate": 9.775649971016982e-06, + "loss": 0.7762, + "step": 47747 + }, + { + "epoch": 3.07, + "learning_rate": 9.775435281994033e-06, + "loss": 0.7517, + "step": 47748 + }, + { + "epoch": 3.07, + "learning_rate": 9.775220592971082e-06, + "loss": 0.8332, + "step": 47749 + }, + { + "epoch": 3.07, + "learning_rate": 9.775005903948131e-06, + "loss": 0.8046, + "step": 47750 + }, + { + "epoch": 3.07, + "learning_rate": 9.774791214925182e-06, + "loss": 0.8288, + "step": 47751 + }, + { + "epoch": 3.07, + "learning_rate": 9.774576525902232e-06, + "loss": 0.7762, + "step": 47752 + }, + { + "epoch": 3.07, + "learning_rate": 9.774361836879281e-06, + "loss": 0.824, + "step": 47753 + }, + { + "epoch": 3.07, + "learning_rate": 9.77414714785633e-06, + "loss": 0.7179, + "step": 47754 + }, + { + "epoch": 3.07, + "learning_rate": 9.77393245883338e-06, + "loss": 0.8492, + "step": 47755 + }, + { + "epoch": 3.07, + "learning_rate": 9.77371776981043e-06, + "loss": 0.8524, + "step": 47756 + }, + { + "epoch": 3.07, + "learning_rate": 9.77350308078748e-06, + "loss": 0.8186, + "step": 47757 + }, + { + "epoch": 3.07, + "learning_rate": 9.77328839176453e-06, + "loss": 0.875, + "step": 47758 + }, + { + "epoch": 3.07, + "learning_rate": 9.77307370274158e-06, + "loss": 0.8036, + "step": 47759 + }, + { + "epoch": 3.07, + "learning_rate": 9.772859013718629e-06, + "loss": 0.7826, + "step": 47760 + }, + { + "epoch": 3.07, + "learning_rate": 9.77264432469568e-06, + "loss": 0.8174, + "step": 47761 + }, + { + "epoch": 3.07, + "learning_rate": 9.772429635672729e-06, + "loss": 0.8147, + "step": 47762 + }, + { + "epoch": 3.07, + "learning_rate": 9.772214946649778e-06, + "loss": 0.8389, + "step": 47763 + }, + { + "epoch": 3.07, + "learning_rate": 9.772000257626828e-06, + "loss": 0.7638, + "step": 47764 + }, + { + "epoch": 3.07, + "learning_rate": 9.771785568603877e-06, + "loss": 0.8385, + "step": 47765 + }, + { + "epoch": 3.07, + "learning_rate": 9.771570879580928e-06, + "loss": 0.8119, + "step": 47766 + }, + { + "epoch": 3.07, + "learning_rate": 9.771356190557979e-06, + "loss": 0.8666, + "step": 47767 + }, + { + "epoch": 3.07, + "learning_rate": 9.771141501535028e-06, + "loss": 0.8946, + "step": 47768 + }, + { + "epoch": 3.07, + "learning_rate": 9.770926812512076e-06, + "loss": 0.7989, + "step": 47769 + }, + { + "epoch": 3.07, + "learning_rate": 9.770712123489127e-06, + "loss": 0.8157, + "step": 47770 + }, + { + "epoch": 3.07, + "learning_rate": 9.770497434466176e-06, + "loss": 0.8346, + "step": 47771 + }, + { + "epoch": 3.07, + "learning_rate": 9.770282745443225e-06, + "loss": 0.8367, + "step": 47772 + }, + { + "epoch": 3.07, + "learning_rate": 9.770068056420276e-06, + "loss": 0.8578, + "step": 47773 + }, + { + "epoch": 3.07, + "learning_rate": 9.769853367397326e-06, + "loss": 0.8189, + "step": 47774 + }, + { + "epoch": 3.07, + "learning_rate": 9.769638678374375e-06, + "loss": 0.904, + "step": 47775 + }, + { + "epoch": 3.07, + "learning_rate": 9.769423989351426e-06, + "loss": 0.8292, + "step": 47776 + }, + { + "epoch": 3.07, + "learning_rate": 9.769209300328475e-06, + "loss": 0.8102, + "step": 47777 + }, + { + "epoch": 3.07, + "learning_rate": 9.768994611305524e-06, + "loss": 0.8114, + "step": 47778 + }, + { + "epoch": 3.07, + "learning_rate": 9.768779922282575e-06, + "loss": 0.8263, + "step": 47779 + }, + { + "epoch": 3.07, + "learning_rate": 9.768565233259624e-06, + "loss": 0.8482, + "step": 47780 + }, + { + "epoch": 3.07, + "learning_rate": 9.768350544236674e-06, + "loss": 0.7439, + "step": 47781 + }, + { + "epoch": 3.07, + "learning_rate": 9.768135855213723e-06, + "loss": 0.8226, + "step": 47782 + }, + { + "epoch": 3.07, + "learning_rate": 9.767921166190774e-06, + "loss": 0.8175, + "step": 47783 + }, + { + "epoch": 3.07, + "learning_rate": 9.767706477167823e-06, + "loss": 0.789, + "step": 47784 + }, + { + "epoch": 3.07, + "learning_rate": 9.767491788144873e-06, + "loss": 0.7803, + "step": 47785 + }, + { + "epoch": 3.07, + "learning_rate": 9.767277099121922e-06, + "loss": 0.7961, + "step": 47786 + }, + { + "epoch": 3.07, + "learning_rate": 9.767062410098971e-06, + "loss": 0.8682, + "step": 47787 + }, + { + "epoch": 3.07, + "learning_rate": 9.766847721076022e-06, + "loss": 0.7918, + "step": 47788 + }, + { + "epoch": 3.07, + "learning_rate": 9.766633032053073e-06, + "loss": 0.792, + "step": 47789 + }, + { + "epoch": 3.07, + "learning_rate": 9.766418343030122e-06, + "loss": 0.8343, + "step": 47790 + }, + { + "epoch": 3.07, + "learning_rate": 9.76620365400717e-06, + "loss": 0.7998, + "step": 47791 + }, + { + "epoch": 3.07, + "learning_rate": 9.765988964984221e-06, + "loss": 0.8874, + "step": 47792 + }, + { + "epoch": 3.07, + "learning_rate": 9.76577427596127e-06, + "loss": 0.8074, + "step": 47793 + }, + { + "epoch": 3.07, + "learning_rate": 9.765559586938321e-06, + "loss": 0.8054, + "step": 47794 + }, + { + "epoch": 3.08, + "learning_rate": 9.765344897915372e-06, + "loss": 0.8644, + "step": 47795 + }, + { + "epoch": 3.08, + "learning_rate": 9.76513020889242e-06, + "loss": 0.9311, + "step": 47796 + }, + { + "epoch": 3.08, + "learning_rate": 9.76491551986947e-06, + "loss": 0.8464, + "step": 47797 + }, + { + "epoch": 3.08, + "learning_rate": 9.76470083084652e-06, + "loss": 0.8383, + "step": 47798 + }, + { + "epoch": 3.08, + "learning_rate": 9.76448614182357e-06, + "loss": 0.8664, + "step": 47799 + }, + { + "epoch": 3.08, + "learning_rate": 9.764271452800618e-06, + "loss": 0.8321, + "step": 47800 + }, + { + "epoch": 3.08, + "learning_rate": 9.764056763777669e-06, + "loss": 0.876, + "step": 47801 + }, + { + "epoch": 3.08, + "learning_rate": 9.76384207475472e-06, + "loss": 0.8421, + "step": 47802 + }, + { + "epoch": 3.08, + "learning_rate": 9.763627385731768e-06, + "loss": 0.8543, + "step": 47803 + }, + { + "epoch": 3.08, + "learning_rate": 9.763412696708819e-06, + "loss": 0.7675, + "step": 47804 + }, + { + "epoch": 3.08, + "learning_rate": 9.763198007685868e-06, + "loss": 0.7939, + "step": 47805 + }, + { + "epoch": 3.08, + "learning_rate": 9.762983318662917e-06, + "loss": 0.8436, + "step": 47806 + }, + { + "epoch": 3.08, + "learning_rate": 9.762768629639968e-06, + "loss": 0.7782, + "step": 47807 + }, + { + "epoch": 3.08, + "learning_rate": 9.762553940617017e-06, + "loss": 0.8847, + "step": 47808 + }, + { + "epoch": 3.08, + "learning_rate": 9.762339251594066e-06, + "loss": 0.8224, + "step": 47809 + }, + { + "epoch": 3.08, + "learning_rate": 9.762124562571116e-06, + "loss": 0.8218, + "step": 47810 + }, + { + "epoch": 3.08, + "learning_rate": 9.761909873548167e-06, + "loss": 0.8506, + "step": 47811 + }, + { + "epoch": 3.08, + "learning_rate": 9.761695184525216e-06, + "loss": 0.8753, + "step": 47812 + }, + { + "epoch": 3.08, + "learning_rate": 9.761480495502267e-06, + "loss": 0.8022, + "step": 47813 + }, + { + "epoch": 3.08, + "learning_rate": 9.761265806479315e-06, + "loss": 0.8097, + "step": 47814 + }, + { + "epoch": 3.08, + "learning_rate": 9.761051117456364e-06, + "loss": 0.8458, + "step": 47815 + }, + { + "epoch": 3.08, + "learning_rate": 9.760836428433415e-06, + "loss": 0.9041, + "step": 47816 + }, + { + "epoch": 3.08, + "learning_rate": 9.760621739410466e-06, + "loss": 0.868, + "step": 47817 + }, + { + "epoch": 3.08, + "learning_rate": 9.760407050387515e-06, + "loss": 0.8216, + "step": 47818 + }, + { + "epoch": 3.08, + "learning_rate": 9.760192361364564e-06, + "loss": 0.7762, + "step": 47819 + }, + { + "epoch": 3.08, + "learning_rate": 9.759977672341614e-06, + "loss": 0.8517, + "step": 47820 + }, + { + "epoch": 3.08, + "learning_rate": 9.759762983318663e-06, + "loss": 0.7808, + "step": 47821 + }, + { + "epoch": 3.08, + "learning_rate": 9.759548294295714e-06, + "loss": 0.7822, + "step": 47822 + }, + { + "epoch": 3.08, + "learning_rate": 9.759333605272763e-06, + "loss": 0.8036, + "step": 47823 + }, + { + "epoch": 3.08, + "learning_rate": 9.759118916249814e-06, + "loss": 0.7974, + "step": 47824 + }, + { + "epoch": 3.08, + "learning_rate": 9.758904227226863e-06, + "loss": 0.8273, + "step": 47825 + }, + { + "epoch": 3.08, + "learning_rate": 9.758689538203913e-06, + "loss": 0.8378, + "step": 47826 + }, + { + "epoch": 3.08, + "learning_rate": 9.758474849180962e-06, + "loss": 0.8641, + "step": 47827 + }, + { + "epoch": 3.08, + "learning_rate": 9.758260160158011e-06, + "loss": 0.8497, + "step": 47828 + }, + { + "epoch": 3.08, + "learning_rate": 9.758045471135062e-06, + "loss": 0.7527, + "step": 47829 + }, + { + "epoch": 3.08, + "learning_rate": 9.75783078211211e-06, + "loss": 0.8363, + "step": 47830 + }, + { + "epoch": 3.08, + "learning_rate": 9.757616093089161e-06, + "loss": 0.7734, + "step": 47831 + }, + { + "epoch": 3.08, + "learning_rate": 9.757401404066212e-06, + "loss": 0.7571, + "step": 47832 + }, + { + "epoch": 3.08, + "learning_rate": 9.757186715043261e-06, + "loss": 0.816, + "step": 47833 + }, + { + "epoch": 3.08, + "learning_rate": 9.75697202602031e-06, + "loss": 0.874, + "step": 47834 + }, + { + "epoch": 3.08, + "learning_rate": 9.75675733699736e-06, + "loss": 0.7961, + "step": 47835 + }, + { + "epoch": 3.08, + "learning_rate": 9.75654264797441e-06, + "loss": 0.8616, + "step": 47836 + }, + { + "epoch": 3.08, + "learning_rate": 9.756327958951459e-06, + "loss": 0.8221, + "step": 47837 + }, + { + "epoch": 3.08, + "learning_rate": 9.75611326992851e-06, + "loss": 0.8298, + "step": 47838 + }, + { + "epoch": 3.08, + "learning_rate": 9.75589858090556e-06, + "loss": 0.8222, + "step": 47839 + }, + { + "epoch": 3.08, + "learning_rate": 9.755683891882609e-06, + "loss": 0.8268, + "step": 47840 + }, + { + "epoch": 3.08, + "learning_rate": 9.755469202859658e-06, + "loss": 0.8167, + "step": 47841 + }, + { + "epoch": 3.08, + "learning_rate": 9.755254513836709e-06, + "loss": 0.8764, + "step": 47842 + }, + { + "epoch": 3.08, + "learning_rate": 9.755039824813757e-06, + "loss": 0.8334, + "step": 47843 + }, + { + "epoch": 3.08, + "learning_rate": 9.754825135790808e-06, + "loss": 0.7911, + "step": 47844 + }, + { + "epoch": 3.08, + "learning_rate": 9.754610446767857e-06, + "loss": 0.7765, + "step": 47845 + }, + { + "epoch": 3.08, + "learning_rate": 9.754395757744908e-06, + "loss": 0.8197, + "step": 47846 + }, + { + "epoch": 3.08, + "learning_rate": 9.754181068721957e-06, + "loss": 0.7752, + "step": 47847 + }, + { + "epoch": 3.08, + "learning_rate": 9.753966379699007e-06, + "loss": 0.7911, + "step": 47848 + }, + { + "epoch": 3.08, + "learning_rate": 9.753751690676056e-06, + "loss": 0.8102, + "step": 47849 + }, + { + "epoch": 3.08, + "learning_rate": 9.753537001653105e-06, + "loss": 0.8, + "step": 47850 + }, + { + "epoch": 3.08, + "learning_rate": 9.753322312630156e-06, + "loss": 0.8416, + "step": 47851 + }, + { + "epoch": 3.08, + "learning_rate": 9.753107623607205e-06, + "loss": 0.837, + "step": 47852 + }, + { + "epoch": 3.08, + "learning_rate": 9.752892934584256e-06, + "loss": 0.8163, + "step": 47853 + }, + { + "epoch": 3.08, + "learning_rate": 9.752678245561306e-06, + "loss": 0.816, + "step": 47854 + }, + { + "epoch": 3.08, + "learning_rate": 9.752463556538355e-06, + "loss": 0.7988, + "step": 47855 + }, + { + "epoch": 3.08, + "learning_rate": 9.752248867515404e-06, + "loss": 0.7915, + "step": 47856 + }, + { + "epoch": 3.08, + "learning_rate": 9.752034178492455e-06, + "loss": 0.8746, + "step": 47857 + }, + { + "epoch": 3.08, + "learning_rate": 9.751819489469504e-06, + "loss": 0.8008, + "step": 47858 + }, + { + "epoch": 3.08, + "learning_rate": 9.751604800446553e-06, + "loss": 0.8776, + "step": 47859 + }, + { + "epoch": 3.08, + "learning_rate": 9.751390111423603e-06, + "loss": 0.8278, + "step": 47860 + }, + { + "epoch": 3.08, + "learning_rate": 9.751175422400654e-06, + "loss": 0.8535, + "step": 47861 + }, + { + "epoch": 3.08, + "learning_rate": 9.750960733377703e-06, + "loss": 0.8093, + "step": 47862 + }, + { + "epoch": 3.08, + "learning_rate": 9.750746044354754e-06, + "loss": 0.7667, + "step": 47863 + }, + { + "epoch": 3.08, + "learning_rate": 9.750531355331803e-06, + "loss": 0.7831, + "step": 47864 + }, + { + "epoch": 3.08, + "learning_rate": 9.750316666308852e-06, + "loss": 0.8359, + "step": 47865 + }, + { + "epoch": 3.08, + "learning_rate": 9.750101977285902e-06, + "loss": 0.8651, + "step": 47866 + }, + { + "epoch": 3.08, + "learning_rate": 9.749887288262951e-06, + "loss": 0.7804, + "step": 47867 + }, + { + "epoch": 3.08, + "learning_rate": 9.749672599240002e-06, + "loss": 0.7925, + "step": 47868 + }, + { + "epoch": 3.08, + "learning_rate": 9.749457910217051e-06, + "loss": 0.8098, + "step": 47869 + }, + { + "epoch": 3.08, + "learning_rate": 9.749243221194102e-06, + "loss": 0.7941, + "step": 47870 + }, + { + "epoch": 3.08, + "learning_rate": 9.74902853217115e-06, + "loss": 0.8495, + "step": 47871 + }, + { + "epoch": 3.08, + "learning_rate": 9.748813843148201e-06, + "loss": 0.8118, + "step": 47872 + }, + { + "epoch": 3.08, + "learning_rate": 9.74859915412525e-06, + "loss": 0.7712, + "step": 47873 + }, + { + "epoch": 3.08, + "learning_rate": 9.748384465102299e-06, + "loss": 0.8077, + "step": 47874 + }, + { + "epoch": 3.08, + "learning_rate": 9.74816977607935e-06, + "loss": 0.83, + "step": 47875 + }, + { + "epoch": 3.08, + "learning_rate": 9.7479550870564e-06, + "loss": 0.87, + "step": 47876 + }, + { + "epoch": 3.08, + "learning_rate": 9.74774039803345e-06, + "loss": 0.8459, + "step": 47877 + }, + { + "epoch": 3.08, + "learning_rate": 9.747525709010498e-06, + "loss": 0.7992, + "step": 47878 + }, + { + "epoch": 3.08, + "learning_rate": 9.747311019987549e-06, + "loss": 0.8361, + "step": 47879 + }, + { + "epoch": 3.08, + "learning_rate": 9.747096330964598e-06, + "loss": 0.8305, + "step": 47880 + }, + { + "epoch": 3.08, + "learning_rate": 9.746881641941649e-06, + "loss": 0.816, + "step": 47881 + }, + { + "epoch": 3.08, + "learning_rate": 9.7466669529187e-06, + "loss": 0.8443, + "step": 47882 + }, + { + "epoch": 3.08, + "learning_rate": 9.746452263895748e-06, + "loss": 0.7426, + "step": 47883 + }, + { + "epoch": 3.08, + "learning_rate": 9.746237574872797e-06, + "loss": 0.8008, + "step": 47884 + }, + { + "epoch": 3.08, + "learning_rate": 9.746022885849848e-06, + "loss": 0.7492, + "step": 47885 + }, + { + "epoch": 3.08, + "learning_rate": 9.745808196826897e-06, + "loss": 0.8157, + "step": 47886 + }, + { + "epoch": 3.08, + "learning_rate": 9.745593507803946e-06, + "loss": 0.7934, + "step": 47887 + }, + { + "epoch": 3.08, + "learning_rate": 9.745378818780996e-06, + "loss": 0.8505, + "step": 47888 + }, + { + "epoch": 3.08, + "learning_rate": 9.745164129758047e-06, + "loss": 0.7965, + "step": 47889 + }, + { + "epoch": 3.08, + "learning_rate": 9.744949440735096e-06, + "loss": 0.8697, + "step": 47890 + }, + { + "epoch": 3.08, + "learning_rate": 9.744734751712147e-06, + "loss": 0.8448, + "step": 47891 + }, + { + "epoch": 3.08, + "learning_rate": 9.744520062689196e-06, + "loss": 0.8605, + "step": 47892 + }, + { + "epoch": 3.08, + "learning_rate": 9.744305373666245e-06, + "loss": 0.809, + "step": 47893 + }, + { + "epoch": 3.08, + "learning_rate": 9.744090684643295e-06, + "loss": 0.8443, + "step": 47894 + }, + { + "epoch": 3.08, + "learning_rate": 9.743875995620344e-06, + "loss": 0.7957, + "step": 47895 + }, + { + "epoch": 3.08, + "learning_rate": 9.743661306597393e-06, + "loss": 0.7931, + "step": 47896 + }, + { + "epoch": 3.08, + "learning_rate": 9.743446617574444e-06, + "loss": 0.8312, + "step": 47897 + }, + { + "epoch": 3.08, + "learning_rate": 9.743231928551495e-06, + "loss": 0.9044, + "step": 47898 + }, + { + "epoch": 3.08, + "learning_rate": 9.743017239528544e-06, + "loss": 0.8231, + "step": 47899 + }, + { + "epoch": 3.08, + "learning_rate": 9.742802550505594e-06, + "loss": 0.7893, + "step": 47900 + }, + { + "epoch": 3.08, + "learning_rate": 9.742587861482643e-06, + "loss": 0.8118, + "step": 47901 + }, + { + "epoch": 3.08, + "learning_rate": 9.742373172459692e-06, + "loss": 0.8281, + "step": 47902 + }, + { + "epoch": 3.08, + "learning_rate": 9.742158483436743e-06, + "loss": 0.8867, + "step": 47903 + }, + { + "epoch": 3.08, + "learning_rate": 9.741943794413793e-06, + "loss": 0.7693, + "step": 47904 + }, + { + "epoch": 3.08, + "learning_rate": 9.741729105390842e-06, + "loss": 0.7766, + "step": 47905 + }, + { + "epoch": 3.08, + "learning_rate": 9.741514416367891e-06, + "loss": 0.7832, + "step": 47906 + }, + { + "epoch": 3.08, + "learning_rate": 9.741299727344942e-06, + "loss": 0.7512, + "step": 47907 + }, + { + "epoch": 3.08, + "learning_rate": 9.741085038321991e-06, + "loss": 0.7946, + "step": 47908 + }, + { + "epoch": 3.08, + "learning_rate": 9.74087034929904e-06, + "loss": 0.8494, + "step": 47909 + }, + { + "epoch": 3.08, + "learning_rate": 9.74065566027609e-06, + "loss": 0.7406, + "step": 47910 + }, + { + "epoch": 3.08, + "learning_rate": 9.740440971253141e-06, + "loss": 0.8379, + "step": 47911 + }, + { + "epoch": 3.08, + "learning_rate": 9.74022628223019e-06, + "loss": 0.8199, + "step": 47912 + }, + { + "epoch": 3.08, + "learning_rate": 9.740011593207241e-06, + "loss": 0.8512, + "step": 47913 + }, + { + "epoch": 3.08, + "learning_rate": 9.73979690418429e-06, + "loss": 0.7779, + "step": 47914 + }, + { + "epoch": 3.08, + "learning_rate": 9.739582215161339e-06, + "loss": 0.8057, + "step": 47915 + }, + { + "epoch": 3.08, + "learning_rate": 9.73936752613839e-06, + "loss": 0.8032, + "step": 47916 + }, + { + "epoch": 3.08, + "learning_rate": 9.739152837115439e-06, + "loss": 0.8661, + "step": 47917 + }, + { + "epoch": 3.08, + "learning_rate": 9.738938148092487e-06, + "loss": 0.795, + "step": 47918 + }, + { + "epoch": 3.08, + "learning_rate": 9.73872345906954e-06, + "loss": 0.7671, + "step": 47919 + }, + { + "epoch": 3.08, + "learning_rate": 9.738508770046589e-06, + "loss": 0.8182, + "step": 47920 + }, + { + "epoch": 3.08, + "learning_rate": 9.738294081023638e-06, + "loss": 0.7855, + "step": 47921 + }, + { + "epoch": 3.08, + "learning_rate": 9.738079392000688e-06, + "loss": 0.8218, + "step": 47922 + }, + { + "epoch": 3.08, + "learning_rate": 9.737864702977737e-06, + "loss": 0.8509, + "step": 47923 + }, + { + "epoch": 3.08, + "learning_rate": 9.737650013954786e-06, + "loss": 0.8736, + "step": 47924 + }, + { + "epoch": 3.08, + "learning_rate": 9.737435324931837e-06, + "loss": 0.8393, + "step": 47925 + }, + { + "epoch": 3.08, + "learning_rate": 9.737220635908888e-06, + "loss": 0.9234, + "step": 47926 + }, + { + "epoch": 3.08, + "learning_rate": 9.737005946885937e-06, + "loss": 0.8149, + "step": 47927 + }, + { + "epoch": 3.08, + "learning_rate": 9.736791257862986e-06, + "loss": 0.8203, + "step": 47928 + }, + { + "epoch": 3.08, + "learning_rate": 9.736576568840036e-06, + "loss": 0.8701, + "step": 47929 + }, + { + "epoch": 3.08, + "learning_rate": 9.736361879817085e-06, + "loss": 0.8477, + "step": 47930 + }, + { + "epoch": 3.08, + "learning_rate": 9.736147190794136e-06, + "loss": 0.785, + "step": 47931 + }, + { + "epoch": 3.08, + "learning_rate": 9.735932501771185e-06, + "loss": 0.8315, + "step": 47932 + }, + { + "epoch": 3.08, + "learning_rate": 9.735717812748235e-06, + "loss": 0.8485, + "step": 47933 + }, + { + "epoch": 3.08, + "learning_rate": 9.735503123725284e-06, + "loss": 0.8095, + "step": 47934 + }, + { + "epoch": 3.08, + "learning_rate": 9.735288434702335e-06, + "loss": 0.8111, + "step": 47935 + }, + { + "epoch": 3.08, + "learning_rate": 9.735073745679384e-06, + "loss": 0.7807, + "step": 47936 + }, + { + "epoch": 3.08, + "learning_rate": 9.734859056656433e-06, + "loss": 0.7941, + "step": 47937 + }, + { + "epoch": 3.08, + "learning_rate": 9.734644367633484e-06, + "loss": 0.7956, + "step": 47938 + }, + { + "epoch": 3.08, + "learning_rate": 9.734429678610533e-06, + "loss": 0.7881, + "step": 47939 + }, + { + "epoch": 3.08, + "learning_rate": 9.734214989587583e-06, + "loss": 0.8162, + "step": 47940 + }, + { + "epoch": 3.08, + "learning_rate": 9.734000300564634e-06, + "loss": 0.879, + "step": 47941 + }, + { + "epoch": 3.08, + "learning_rate": 9.733785611541683e-06, + "loss": 0.7913, + "step": 47942 + }, + { + "epoch": 3.08, + "learning_rate": 9.733570922518732e-06, + "loss": 0.8378, + "step": 47943 + }, + { + "epoch": 3.08, + "learning_rate": 9.733356233495783e-06, + "loss": 0.8172, + "step": 47944 + }, + { + "epoch": 3.08, + "learning_rate": 9.733141544472832e-06, + "loss": 0.8279, + "step": 47945 + }, + { + "epoch": 3.08, + "learning_rate": 9.73292685544988e-06, + "loss": 0.8455, + "step": 47946 + }, + { + "epoch": 3.08, + "learning_rate": 9.732712166426931e-06, + "loss": 0.8293, + "step": 47947 + }, + { + "epoch": 3.08, + "learning_rate": 9.732497477403982e-06, + "loss": 0.7786, + "step": 47948 + }, + { + "epoch": 3.08, + "learning_rate": 9.73228278838103e-06, + "loss": 0.785, + "step": 47949 + }, + { + "epoch": 3.08, + "learning_rate": 9.732068099358081e-06, + "loss": 0.7952, + "step": 47950 + }, + { + "epoch": 3.09, + "learning_rate": 9.73185341033513e-06, + "loss": 0.8691, + "step": 47951 + }, + { + "epoch": 3.09, + "learning_rate": 9.73163872131218e-06, + "loss": 0.8721, + "step": 47952 + }, + { + "epoch": 3.09, + "learning_rate": 9.73142403228923e-06, + "loss": 0.8448, + "step": 47953 + }, + { + "epoch": 3.09, + "learning_rate": 9.731209343266279e-06, + "loss": 0.7931, + "step": 47954 + }, + { + "epoch": 3.09, + "learning_rate": 9.73099465424333e-06, + "loss": 0.8139, + "step": 47955 + }, + { + "epoch": 3.09, + "learning_rate": 9.730779965220379e-06, + "loss": 0.88, + "step": 47956 + }, + { + "epoch": 3.09, + "learning_rate": 9.73056527619743e-06, + "loss": 0.8256, + "step": 47957 + }, + { + "epoch": 3.09, + "learning_rate": 9.730350587174478e-06, + "loss": 0.8318, + "step": 47958 + }, + { + "epoch": 3.09, + "learning_rate": 9.730135898151529e-06, + "loss": 0.7684, + "step": 47959 + }, + { + "epoch": 3.09, + "learning_rate": 9.729921209128578e-06, + "loss": 0.789, + "step": 47960 + }, + { + "epoch": 3.09, + "learning_rate": 9.729706520105627e-06, + "loss": 0.7486, + "step": 47961 + }, + { + "epoch": 3.09, + "learning_rate": 9.729491831082678e-06, + "loss": 0.7329, + "step": 47962 + }, + { + "epoch": 3.09, + "learning_rate": 9.729277142059728e-06, + "loss": 0.8586, + "step": 47963 + }, + { + "epoch": 3.09, + "learning_rate": 9.729062453036777e-06, + "loss": 0.8311, + "step": 47964 + }, + { + "epoch": 3.09, + "learning_rate": 9.728847764013826e-06, + "loss": 0.813, + "step": 47965 + }, + { + "epoch": 3.09, + "learning_rate": 9.728633074990877e-06, + "loss": 0.7795, + "step": 47966 + }, + { + "epoch": 3.09, + "learning_rate": 9.728418385967926e-06, + "loss": 0.805, + "step": 47967 + }, + { + "epoch": 3.09, + "learning_rate": 9.728203696944976e-06, + "loss": 0.8885, + "step": 47968 + }, + { + "epoch": 3.09, + "learning_rate": 9.727989007922027e-06, + "loss": 0.8298, + "step": 47969 + }, + { + "epoch": 3.09, + "learning_rate": 9.727774318899076e-06, + "loss": 0.8242, + "step": 47970 + }, + { + "epoch": 3.09, + "learning_rate": 9.727559629876125e-06, + "loss": 0.7698, + "step": 47971 + }, + { + "epoch": 3.09, + "learning_rate": 9.727344940853176e-06, + "loss": 0.8345, + "step": 47972 + }, + { + "epoch": 3.09, + "learning_rate": 9.727130251830225e-06, + "loss": 0.7967, + "step": 47973 + }, + { + "epoch": 3.09, + "learning_rate": 9.726915562807274e-06, + "loss": 0.8112, + "step": 47974 + }, + { + "epoch": 3.09, + "learning_rate": 9.726700873784324e-06, + "loss": 0.8491, + "step": 47975 + }, + { + "epoch": 3.09, + "learning_rate": 9.726486184761373e-06, + "loss": 0.8315, + "step": 47976 + }, + { + "epoch": 3.09, + "learning_rate": 9.726271495738424e-06, + "loss": 0.825, + "step": 47977 + }, + { + "epoch": 3.09, + "learning_rate": 9.726056806715474e-06, + "loss": 0.8015, + "step": 47978 + }, + { + "epoch": 3.09, + "learning_rate": 9.725842117692523e-06, + "loss": 0.8382, + "step": 47979 + }, + { + "epoch": 3.09, + "learning_rate": 9.725627428669572e-06, + "loss": 0.7924, + "step": 47980 + }, + { + "epoch": 3.09, + "learning_rate": 9.725412739646623e-06, + "loss": 0.942, + "step": 47981 + }, + { + "epoch": 3.09, + "learning_rate": 9.725198050623672e-06, + "loss": 0.8404, + "step": 47982 + }, + { + "epoch": 3.09, + "learning_rate": 9.724983361600721e-06, + "loss": 0.85, + "step": 47983 + }, + { + "epoch": 3.09, + "learning_rate": 9.724768672577772e-06, + "loss": 0.8073, + "step": 47984 + }, + { + "epoch": 3.09, + "learning_rate": 9.724553983554822e-06, + "loss": 0.803, + "step": 47985 + }, + { + "epoch": 3.09, + "learning_rate": 9.724339294531871e-06, + "loss": 0.8409, + "step": 47986 + }, + { + "epoch": 3.09, + "learning_rate": 9.724124605508922e-06, + "loss": 0.7796, + "step": 47987 + }, + { + "epoch": 3.09, + "learning_rate": 9.723909916485971e-06, + "loss": 0.8719, + "step": 47988 + }, + { + "epoch": 3.09, + "learning_rate": 9.72369522746302e-06, + "loss": 0.8038, + "step": 47989 + }, + { + "epoch": 3.09, + "learning_rate": 9.72348053844007e-06, + "loss": 0.8401, + "step": 47990 + }, + { + "epoch": 3.09, + "learning_rate": 9.723265849417121e-06, + "loss": 0.7987, + "step": 47991 + }, + { + "epoch": 3.09, + "learning_rate": 9.72305116039417e-06, + "loss": 0.8289, + "step": 47992 + }, + { + "epoch": 3.09, + "learning_rate": 9.722836471371219e-06, + "loss": 0.8157, + "step": 47993 + }, + { + "epoch": 3.09, + "learning_rate": 9.72262178234827e-06, + "loss": 0.847, + "step": 47994 + }, + { + "epoch": 3.09, + "learning_rate": 9.722407093325319e-06, + "loss": 0.8363, + "step": 47995 + }, + { + "epoch": 3.09, + "learning_rate": 9.722192404302368e-06, + "loss": 0.8035, + "step": 47996 + }, + { + "epoch": 3.09, + "learning_rate": 9.721977715279418e-06, + "loss": 0.8375, + "step": 47997 + }, + { + "epoch": 3.09, + "learning_rate": 9.721763026256469e-06, + "loss": 0.8384, + "step": 47998 + }, + { + "epoch": 3.09, + "learning_rate": 9.721548337233518e-06, + "loss": 0.8469, + "step": 47999 + }, + { + "epoch": 3.09, + "learning_rate": 9.721333648210569e-06, + "loss": 0.8491, + "step": 48000 + }, + { + "epoch": 3.09, + "eval_laion/cider": 0.5220728164178372, + "eval_laion/meteor": 0.3974905910678318, + "eval_laion/pred_num_words": 37.072265625, + "eval_laion/sacrebleu": 16.099413047084727, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.050874948501587, + "eval_runtime": 95.4826, + "eval_samples_per_second": 5.226, + "eval_steps_per_second": 0.335, + "step": 48000 + }, + { + "epoch": 3.09, + "learning_rate": 9.721118959187618e-06, + "loss": 0.7547, + "step": 48001 + }, + { + "epoch": 3.09, + "learning_rate": 9.720904270164667e-06, + "loss": 0.8523, + "step": 48002 + }, + { + "epoch": 3.09, + "learning_rate": 9.720689581141717e-06, + "loss": 0.8703, + "step": 48003 + }, + { + "epoch": 3.09, + "learning_rate": 9.720474892118766e-06, + "loss": 0.8446, + "step": 48004 + }, + { + "epoch": 3.09, + "learning_rate": 9.720260203095815e-06, + "loss": 0.8274, + "step": 48005 + }, + { + "epoch": 3.09, + "learning_rate": 9.720045514072866e-06, + "loss": 0.8031, + "step": 48006 + }, + { + "epoch": 3.09, + "learning_rate": 9.719830825049916e-06, + "loss": 0.742, + "step": 48007 + }, + { + "epoch": 3.09, + "learning_rate": 9.719616136026965e-06, + "loss": 0.8587, + "step": 48008 + }, + { + "epoch": 3.09, + "learning_rate": 9.719401447004016e-06, + "loss": 0.8547, + "step": 48009 + }, + { + "epoch": 3.09, + "learning_rate": 9.719186757981065e-06, + "loss": 0.7951, + "step": 48010 + }, + { + "epoch": 3.09, + "learning_rate": 9.718972068958114e-06, + "loss": 0.7687, + "step": 48011 + }, + { + "epoch": 3.09, + "learning_rate": 9.718757379935165e-06, + "loss": 0.8431, + "step": 48012 + }, + { + "epoch": 3.09, + "learning_rate": 9.718542690912215e-06, + "loss": 0.8438, + "step": 48013 + }, + { + "epoch": 3.09, + "learning_rate": 9.718328001889264e-06, + "loss": 0.8014, + "step": 48014 + }, + { + "epoch": 3.09, + "learning_rate": 9.718113312866313e-06, + "loss": 0.8177, + "step": 48015 + }, + { + "epoch": 3.09, + "learning_rate": 9.717898623843364e-06, + "loss": 0.8855, + "step": 48016 + }, + { + "epoch": 3.09, + "learning_rate": 9.717683934820413e-06, + "loss": 0.8098, + "step": 48017 + }, + { + "epoch": 3.09, + "learning_rate": 9.717469245797464e-06, + "loss": 0.829, + "step": 48018 + }, + { + "epoch": 3.09, + "learning_rate": 9.717254556774513e-06, + "loss": 0.8156, + "step": 48019 + }, + { + "epoch": 3.09, + "learning_rate": 9.717039867751563e-06, + "loss": 0.8085, + "step": 48020 + }, + { + "epoch": 3.09, + "learning_rate": 9.716825178728612e-06, + "loss": 0.8874, + "step": 48021 + }, + { + "epoch": 3.09, + "learning_rate": 9.716610489705663e-06, + "loss": 0.8135, + "step": 48022 + }, + { + "epoch": 3.09, + "learning_rate": 9.716395800682712e-06, + "loss": 0.8055, + "step": 48023 + }, + { + "epoch": 3.09, + "learning_rate": 9.71618111165976e-06, + "loss": 0.7832, + "step": 48024 + }, + { + "epoch": 3.09, + "learning_rate": 9.715966422636811e-06, + "loss": 0.7414, + "step": 48025 + }, + { + "epoch": 3.09, + "learning_rate": 9.71575173361386e-06, + "loss": 0.8651, + "step": 48026 + }, + { + "epoch": 3.09, + "learning_rate": 9.715537044590911e-06, + "loss": 0.8338, + "step": 48027 + }, + { + "epoch": 3.09, + "learning_rate": 9.715322355567962e-06, + "loss": 0.8363, + "step": 48028 + }, + { + "epoch": 3.09, + "learning_rate": 9.71510766654501e-06, + "loss": 0.8754, + "step": 48029 + }, + { + "epoch": 3.09, + "learning_rate": 9.71489297752206e-06, + "loss": 0.7871, + "step": 48030 + }, + { + "epoch": 3.09, + "learning_rate": 9.71467828849911e-06, + "loss": 0.8534, + "step": 48031 + }, + { + "epoch": 3.09, + "learning_rate": 9.71446359947616e-06, + "loss": 0.7948, + "step": 48032 + }, + { + "epoch": 3.09, + "learning_rate": 9.714248910453208e-06, + "loss": 0.7539, + "step": 48033 + }, + { + "epoch": 3.09, + "learning_rate": 9.714034221430259e-06, + "loss": 0.8665, + "step": 48034 + }, + { + "epoch": 3.09, + "learning_rate": 9.71381953240731e-06, + "loss": 0.8049, + "step": 48035 + }, + { + "epoch": 3.09, + "learning_rate": 9.713604843384359e-06, + "loss": 0.8292, + "step": 48036 + }, + { + "epoch": 3.09, + "learning_rate": 9.71339015436141e-06, + "loss": 0.8445, + "step": 48037 + }, + { + "epoch": 3.09, + "learning_rate": 9.713175465338458e-06, + "loss": 0.8296, + "step": 48038 + }, + { + "epoch": 3.09, + "learning_rate": 9.712960776315507e-06, + "loss": 0.8063, + "step": 48039 + }, + { + "epoch": 3.09, + "learning_rate": 9.712746087292558e-06, + "loss": 0.8292, + "step": 48040 + }, + { + "epoch": 3.09, + "learning_rate": 9.712531398269607e-06, + "loss": 0.8462, + "step": 48041 + }, + { + "epoch": 3.09, + "learning_rate": 9.712316709246657e-06, + "loss": 0.7717, + "step": 48042 + }, + { + "epoch": 3.09, + "learning_rate": 9.712102020223706e-06, + "loss": 0.8285, + "step": 48043 + }, + { + "epoch": 3.09, + "learning_rate": 9.711887331200757e-06, + "loss": 0.7915, + "step": 48044 + }, + { + "epoch": 3.09, + "learning_rate": 9.711672642177806e-06, + "loss": 0.8361, + "step": 48045 + }, + { + "epoch": 3.09, + "learning_rate": 9.711457953154857e-06, + "loss": 0.8659, + "step": 48046 + }, + { + "epoch": 3.09, + "learning_rate": 9.711243264131906e-06, + "loss": 0.8052, + "step": 48047 + }, + { + "epoch": 3.09, + "learning_rate": 9.711028575108955e-06, + "loss": 0.785, + "step": 48048 + }, + { + "epoch": 3.09, + "learning_rate": 9.710813886086005e-06, + "loss": 0.8475, + "step": 48049 + }, + { + "epoch": 3.09, + "learning_rate": 9.710599197063056e-06, + "loss": 0.8303, + "step": 48050 + }, + { + "epoch": 3.09, + "learning_rate": 9.710384508040105e-06, + "loss": 0.786, + "step": 48051 + }, + { + "epoch": 3.09, + "learning_rate": 9.710169819017154e-06, + "loss": 0.8562, + "step": 48052 + }, + { + "epoch": 3.09, + "learning_rate": 9.709955129994204e-06, + "loss": 0.8514, + "step": 48053 + }, + { + "epoch": 3.09, + "learning_rate": 9.709740440971253e-06, + "loss": 0.8637, + "step": 48054 + }, + { + "epoch": 3.09, + "learning_rate": 9.709525751948304e-06, + "loss": 0.8276, + "step": 48055 + }, + { + "epoch": 3.09, + "learning_rate": 9.709311062925355e-06, + "loss": 0.7866, + "step": 48056 + }, + { + "epoch": 3.09, + "learning_rate": 9.709096373902404e-06, + "loss": 0.8174, + "step": 48057 + }, + { + "epoch": 3.09, + "learning_rate": 9.708881684879453e-06, + "loss": 0.8461, + "step": 48058 + }, + { + "epoch": 3.09, + "learning_rate": 9.708666995856503e-06, + "loss": 0.735, + "step": 48059 + }, + { + "epoch": 3.09, + "learning_rate": 9.708452306833552e-06, + "loss": 0.7591, + "step": 48060 + }, + { + "epoch": 3.09, + "learning_rate": 9.708237617810601e-06, + "loss": 0.7776, + "step": 48061 + }, + { + "epoch": 3.09, + "learning_rate": 9.708022928787652e-06, + "loss": 0.7953, + "step": 48062 + }, + { + "epoch": 3.09, + "learning_rate": 9.707808239764701e-06, + "loss": 0.8958, + "step": 48063 + }, + { + "epoch": 3.09, + "learning_rate": 9.707593550741752e-06, + "loss": 0.8365, + "step": 48064 + }, + { + "epoch": 3.09, + "learning_rate": 9.707378861718802e-06, + "loss": 0.7759, + "step": 48065 + }, + { + "epoch": 3.09, + "learning_rate": 9.707164172695851e-06, + "loss": 0.8705, + "step": 48066 + }, + { + "epoch": 3.09, + "learning_rate": 9.7069494836729e-06, + "loss": 0.8211, + "step": 48067 + }, + { + "epoch": 3.09, + "learning_rate": 9.70673479464995e-06, + "loss": 0.842, + "step": 48068 + }, + { + "epoch": 3.09, + "learning_rate": 9.706520105627e-06, + "loss": 0.8019, + "step": 48069 + }, + { + "epoch": 3.09, + "learning_rate": 9.706305416604049e-06, + "loss": 0.8362, + "step": 48070 + }, + { + "epoch": 3.09, + "learning_rate": 9.7060907275811e-06, + "loss": 0.8235, + "step": 48071 + }, + { + "epoch": 3.09, + "learning_rate": 9.70587603855815e-06, + "loss": 0.8643, + "step": 48072 + }, + { + "epoch": 3.09, + "learning_rate": 9.705661349535199e-06, + "loss": 0.8248, + "step": 48073 + }, + { + "epoch": 3.09, + "learning_rate": 9.70544666051225e-06, + "loss": 0.8462, + "step": 48074 + }, + { + "epoch": 3.09, + "learning_rate": 9.705231971489299e-06, + "loss": 0.8243, + "step": 48075 + }, + { + "epoch": 3.09, + "learning_rate": 9.705017282466348e-06, + "loss": 0.8132, + "step": 48076 + }, + { + "epoch": 3.09, + "learning_rate": 9.704802593443398e-06, + "loss": 0.7648, + "step": 48077 + }, + { + "epoch": 3.09, + "learning_rate": 9.704587904420449e-06, + "loss": 0.8887, + "step": 48078 + }, + { + "epoch": 3.09, + "learning_rate": 9.704373215397498e-06, + "loss": 0.7956, + "step": 48079 + }, + { + "epoch": 3.09, + "learning_rate": 9.704158526374547e-06, + "loss": 0.8076, + "step": 48080 + }, + { + "epoch": 3.09, + "learning_rate": 9.703943837351598e-06, + "loss": 0.8635, + "step": 48081 + }, + { + "epoch": 3.09, + "learning_rate": 9.703729148328646e-06, + "loss": 0.7921, + "step": 48082 + }, + { + "epoch": 3.09, + "learning_rate": 9.703514459305695e-06, + "loss": 0.786, + "step": 48083 + }, + { + "epoch": 3.09, + "learning_rate": 9.703299770282746e-06, + "loss": 0.8513, + "step": 48084 + }, + { + "epoch": 3.09, + "learning_rate": 9.703085081259797e-06, + "loss": 0.808, + "step": 48085 + }, + { + "epoch": 3.09, + "learning_rate": 9.702870392236846e-06, + "loss": 0.8285, + "step": 48086 + }, + { + "epoch": 3.09, + "learning_rate": 9.702655703213896e-06, + "loss": 0.8458, + "step": 48087 + }, + { + "epoch": 3.09, + "learning_rate": 9.702441014190945e-06, + "loss": 0.8012, + "step": 48088 + }, + { + "epoch": 3.09, + "learning_rate": 9.702226325167994e-06, + "loss": 0.8204, + "step": 48089 + }, + { + "epoch": 3.09, + "learning_rate": 9.702011636145045e-06, + "loss": 0.8552, + "step": 48090 + }, + { + "epoch": 3.09, + "learning_rate": 9.701796947122094e-06, + "loss": 0.8576, + "step": 48091 + }, + { + "epoch": 3.09, + "learning_rate": 9.701582258099143e-06, + "loss": 0.8134, + "step": 48092 + }, + { + "epoch": 3.09, + "learning_rate": 9.701367569076194e-06, + "loss": 0.816, + "step": 48093 + }, + { + "epoch": 3.09, + "learning_rate": 9.701152880053244e-06, + "loss": 0.9009, + "step": 48094 + }, + { + "epoch": 3.09, + "learning_rate": 9.700938191030293e-06, + "loss": 0.8112, + "step": 48095 + }, + { + "epoch": 3.09, + "learning_rate": 9.700723502007344e-06, + "loss": 0.7929, + "step": 48096 + }, + { + "epoch": 3.09, + "learning_rate": 9.700508812984393e-06, + "loss": 0.7853, + "step": 48097 + }, + { + "epoch": 3.09, + "learning_rate": 9.700294123961442e-06, + "loss": 0.8058, + "step": 48098 + }, + { + "epoch": 3.09, + "learning_rate": 9.700079434938492e-06, + "loss": 0.7963, + "step": 48099 + }, + { + "epoch": 3.09, + "learning_rate": 9.699864745915543e-06, + "loss": 0.8363, + "step": 48100 + }, + { + "epoch": 3.09, + "learning_rate": 9.699650056892592e-06, + "loss": 0.8546, + "step": 48101 + }, + { + "epoch": 3.09, + "learning_rate": 9.699435367869641e-06, + "loss": 0.8211, + "step": 48102 + }, + { + "epoch": 3.09, + "learning_rate": 9.699220678846692e-06, + "loss": 0.7805, + "step": 48103 + }, + { + "epoch": 3.09, + "learning_rate": 9.69900598982374e-06, + "loss": 0.86, + "step": 48104 + }, + { + "epoch": 3.09, + "learning_rate": 9.698791300800791e-06, + "loss": 0.8616, + "step": 48105 + }, + { + "epoch": 3.1, + "learning_rate": 9.69857661177784e-06, + "loss": 0.8303, + "step": 48106 + }, + { + "epoch": 3.1, + "learning_rate": 9.698361922754891e-06, + "loss": 0.8179, + "step": 48107 + }, + { + "epoch": 3.1, + "learning_rate": 9.69814723373194e-06, + "loss": 0.8296, + "step": 48108 + }, + { + "epoch": 3.1, + "learning_rate": 9.69793254470899e-06, + "loss": 0.8798, + "step": 48109 + }, + { + "epoch": 3.1, + "learning_rate": 9.69771785568604e-06, + "loss": 0.8325, + "step": 48110 + }, + { + "epoch": 3.1, + "learning_rate": 9.697503166663088e-06, + "loss": 0.7359, + "step": 48111 + }, + { + "epoch": 3.1, + "learning_rate": 9.697288477640139e-06, + "loss": 0.765, + "step": 48112 + }, + { + "epoch": 3.1, + "learning_rate": 9.697073788617188e-06, + "loss": 0.7909, + "step": 48113 + }, + { + "epoch": 3.1, + "learning_rate": 9.696859099594239e-06, + "loss": 0.8429, + "step": 48114 + }, + { + "epoch": 3.1, + "learning_rate": 9.69664441057129e-06, + "loss": 0.8147, + "step": 48115 + }, + { + "epoch": 3.1, + "learning_rate": 9.696429721548338e-06, + "loss": 0.7772, + "step": 48116 + }, + { + "epoch": 3.1, + "learning_rate": 9.696215032525387e-06, + "loss": 0.7701, + "step": 48117 + }, + { + "epoch": 3.1, + "learning_rate": 9.696000343502438e-06, + "loss": 0.8872, + "step": 48118 + }, + { + "epoch": 3.1, + "learning_rate": 9.695785654479487e-06, + "loss": 0.883, + "step": 48119 + }, + { + "epoch": 3.1, + "learning_rate": 9.695570965456536e-06, + "loss": 0.8138, + "step": 48120 + }, + { + "epoch": 3.1, + "learning_rate": 9.695356276433587e-06, + "loss": 0.8292, + "step": 48121 + }, + { + "epoch": 3.1, + "learning_rate": 9.695141587410637e-06, + "loss": 0.8231, + "step": 48122 + }, + { + "epoch": 3.1, + "learning_rate": 9.694926898387686e-06, + "loss": 0.8593, + "step": 48123 + }, + { + "epoch": 3.1, + "learning_rate": 9.694712209364737e-06, + "loss": 0.8755, + "step": 48124 + }, + { + "epoch": 3.1, + "learning_rate": 9.694497520341786e-06, + "loss": 0.7809, + "step": 48125 + }, + { + "epoch": 3.1, + "learning_rate": 9.694282831318835e-06, + "loss": 0.8239, + "step": 48126 + }, + { + "epoch": 3.1, + "learning_rate": 9.694068142295885e-06, + "loss": 0.8029, + "step": 48127 + }, + { + "epoch": 3.1, + "learning_rate": 9.693853453272934e-06, + "loss": 0.8406, + "step": 48128 + }, + { + "epoch": 3.1, + "learning_rate": 9.693638764249985e-06, + "loss": 0.8178, + "step": 48129 + }, + { + "epoch": 3.1, + "learning_rate": 9.693424075227034e-06, + "loss": 0.8167, + "step": 48130 + }, + { + "epoch": 3.1, + "learning_rate": 9.693209386204085e-06, + "loss": 0.8312, + "step": 48131 + }, + { + "epoch": 3.1, + "learning_rate": 9.692994697181134e-06, + "loss": 0.7422, + "step": 48132 + }, + { + "epoch": 3.1, + "learning_rate": 9.692780008158184e-06, + "loss": 0.7935, + "step": 48133 + }, + { + "epoch": 3.1, + "learning_rate": 9.692565319135233e-06, + "loss": 0.8112, + "step": 48134 + }, + { + "epoch": 3.1, + "learning_rate": 9.692350630112282e-06, + "loss": 0.8163, + "step": 48135 + }, + { + "epoch": 3.1, + "learning_rate": 9.692135941089333e-06, + "loss": 0.8435, + "step": 48136 + }, + { + "epoch": 3.1, + "learning_rate": 9.691921252066384e-06, + "loss": 0.8204, + "step": 48137 + }, + { + "epoch": 3.1, + "learning_rate": 9.691706563043433e-06, + "loss": 0.7804, + "step": 48138 + }, + { + "epoch": 3.1, + "learning_rate": 9.691491874020482e-06, + "loss": 0.8439, + "step": 48139 + }, + { + "epoch": 3.1, + "learning_rate": 9.691277184997532e-06, + "loss": 0.854, + "step": 48140 + }, + { + "epoch": 3.1, + "learning_rate": 9.691062495974581e-06, + "loss": 0.8477, + "step": 48141 + }, + { + "epoch": 3.1, + "learning_rate": 9.690847806951632e-06, + "loss": 0.83, + "step": 48142 + }, + { + "epoch": 3.1, + "learning_rate": 9.69063311792868e-06, + "loss": 0.8174, + "step": 48143 + }, + { + "epoch": 3.1, + "learning_rate": 9.690418428905731e-06, + "loss": 0.8218, + "step": 48144 + }, + { + "epoch": 3.1, + "learning_rate": 9.69020373988278e-06, + "loss": 0.8061, + "step": 48145 + }, + { + "epoch": 3.1, + "learning_rate": 9.689989050859831e-06, + "loss": 0.8765, + "step": 48146 + }, + { + "epoch": 3.1, + "learning_rate": 9.68977436183688e-06, + "loss": 0.8494, + "step": 48147 + }, + { + "epoch": 3.1, + "learning_rate": 9.689559672813929e-06, + "loss": 0.8565, + "step": 48148 + }, + { + "epoch": 3.1, + "learning_rate": 9.68934498379098e-06, + "loss": 0.894, + "step": 48149 + }, + { + "epoch": 3.1, + "learning_rate": 9.689130294768029e-06, + "loss": 0.8568, + "step": 48150 + }, + { + "epoch": 3.1, + "learning_rate": 9.68891560574508e-06, + "loss": 0.76, + "step": 48151 + }, + { + "epoch": 3.1, + "learning_rate": 9.68870091672213e-06, + "loss": 0.8059, + "step": 48152 + }, + { + "epoch": 3.1, + "learning_rate": 9.688486227699179e-06, + "loss": 0.7955, + "step": 48153 + }, + { + "epoch": 3.1, + "learning_rate": 9.688271538676228e-06, + "loss": 0.8352, + "step": 48154 + }, + { + "epoch": 3.1, + "learning_rate": 9.688056849653279e-06, + "loss": 0.8047, + "step": 48155 + }, + { + "epoch": 3.1, + "learning_rate": 9.687842160630327e-06, + "loss": 0.8327, + "step": 48156 + }, + { + "epoch": 3.1, + "learning_rate": 9.687627471607376e-06, + "loss": 0.8839, + "step": 48157 + }, + { + "epoch": 3.1, + "learning_rate": 9.687412782584427e-06, + "loss": 0.829, + "step": 48158 + }, + { + "epoch": 3.1, + "learning_rate": 9.687198093561478e-06, + "loss": 0.8606, + "step": 48159 + }, + { + "epoch": 3.1, + "learning_rate": 9.686983404538527e-06, + "loss": 0.8218, + "step": 48160 + }, + { + "epoch": 3.1, + "learning_rate": 9.686768715515576e-06, + "loss": 0.8142, + "step": 48161 + }, + { + "epoch": 3.1, + "learning_rate": 9.686554026492626e-06, + "loss": 0.8621, + "step": 48162 + }, + { + "epoch": 3.1, + "learning_rate": 9.686339337469675e-06, + "loss": 0.846, + "step": 48163 + }, + { + "epoch": 3.1, + "learning_rate": 9.686124648446726e-06, + "loss": 0.8977, + "step": 48164 + }, + { + "epoch": 3.1, + "learning_rate": 9.685909959423777e-06, + "loss": 0.8165, + "step": 48165 + }, + { + "epoch": 3.1, + "learning_rate": 9.685695270400826e-06, + "loss": 0.7683, + "step": 48166 + }, + { + "epoch": 3.1, + "learning_rate": 9.685480581377875e-06, + "loss": 0.7904, + "step": 48167 + }, + { + "epoch": 3.1, + "learning_rate": 9.685265892354925e-06, + "loss": 0.7672, + "step": 48168 + }, + { + "epoch": 3.1, + "learning_rate": 9.685051203331974e-06, + "loss": 0.8113, + "step": 48169 + }, + { + "epoch": 3.1, + "learning_rate": 9.684836514309023e-06, + "loss": 0.8407, + "step": 48170 + }, + { + "epoch": 3.1, + "learning_rate": 9.684621825286074e-06, + "loss": 0.761, + "step": 48171 + }, + { + "epoch": 3.1, + "learning_rate": 9.684407136263123e-06, + "loss": 0.7848, + "step": 48172 + }, + { + "epoch": 3.1, + "learning_rate": 9.684192447240173e-06, + "loss": 0.8724, + "step": 48173 + }, + { + "epoch": 3.1, + "learning_rate": 9.683977758217224e-06, + "loss": 0.7691, + "step": 48174 + }, + { + "epoch": 3.1, + "learning_rate": 9.683763069194273e-06, + "loss": 0.822, + "step": 48175 + }, + { + "epoch": 3.1, + "learning_rate": 9.683548380171322e-06, + "loss": 0.8099, + "step": 48176 + }, + { + "epoch": 3.1, + "learning_rate": 9.683333691148373e-06, + "loss": 0.8028, + "step": 48177 + }, + { + "epoch": 3.1, + "learning_rate": 9.683119002125422e-06, + "loss": 0.7837, + "step": 48178 + }, + { + "epoch": 3.1, + "learning_rate": 9.68290431310247e-06, + "loss": 0.7868, + "step": 48179 + }, + { + "epoch": 3.1, + "learning_rate": 9.682689624079521e-06, + "loss": 0.8472, + "step": 48180 + }, + { + "epoch": 3.1, + "learning_rate": 9.682474935056572e-06, + "loss": 0.8586, + "step": 48181 + }, + { + "epoch": 3.1, + "learning_rate": 9.682260246033621e-06, + "loss": 0.8489, + "step": 48182 + }, + { + "epoch": 3.1, + "learning_rate": 9.682045557010672e-06, + "loss": 0.7903, + "step": 48183 + }, + { + "epoch": 3.1, + "learning_rate": 9.68183086798772e-06, + "loss": 0.823, + "step": 48184 + }, + { + "epoch": 3.1, + "learning_rate": 9.68161617896477e-06, + "loss": 0.8354, + "step": 48185 + }, + { + "epoch": 3.1, + "learning_rate": 9.68140148994182e-06, + "loss": 0.8139, + "step": 48186 + }, + { + "epoch": 3.1, + "learning_rate": 9.68118680091887e-06, + "loss": 0.7984, + "step": 48187 + }, + { + "epoch": 3.1, + "learning_rate": 9.68097211189592e-06, + "loss": 0.7982, + "step": 48188 + }, + { + "epoch": 3.1, + "learning_rate": 9.680757422872969e-06, + "loss": 0.8164, + "step": 48189 + }, + { + "epoch": 3.1, + "learning_rate": 9.68054273385002e-06, + "loss": 0.7976, + "step": 48190 + }, + { + "epoch": 3.1, + "learning_rate": 9.680328044827068e-06, + "loss": 0.7698, + "step": 48191 + }, + { + "epoch": 3.1, + "learning_rate": 9.680113355804119e-06, + "loss": 0.7754, + "step": 48192 + }, + { + "epoch": 3.1, + "learning_rate": 9.679898666781168e-06, + "loss": 0.8491, + "step": 48193 + }, + { + "epoch": 3.1, + "learning_rate": 9.679683977758219e-06, + "loss": 0.8601, + "step": 48194 + }, + { + "epoch": 3.1, + "learning_rate": 9.679469288735268e-06, + "loss": 0.8501, + "step": 48195 + }, + { + "epoch": 3.1, + "learning_rate": 9.679254599712318e-06, + "loss": 0.8204, + "step": 48196 + }, + { + "epoch": 3.1, + "learning_rate": 9.679039910689367e-06, + "loss": 0.8379, + "step": 48197 + }, + { + "epoch": 3.1, + "learning_rate": 9.678825221666416e-06, + "loss": 0.8578, + "step": 48198 + }, + { + "epoch": 3.1, + "learning_rate": 9.678610532643467e-06, + "loss": 0.873, + "step": 48199 + }, + { + "epoch": 3.1, + "learning_rate": 9.678395843620516e-06, + "loss": 0.8425, + "step": 48200 + }, + { + "epoch": 3.1, + "learning_rate": 9.678181154597566e-06, + "loss": 0.8295, + "step": 48201 + }, + { + "epoch": 3.1, + "learning_rate": 9.677966465574617e-06, + "loss": 0.806, + "step": 48202 + }, + { + "epoch": 3.1, + "learning_rate": 9.677751776551666e-06, + "loss": 0.8798, + "step": 48203 + }, + { + "epoch": 3.1, + "learning_rate": 9.677537087528715e-06, + "loss": 0.8201, + "step": 48204 + }, + { + "epoch": 3.1, + "learning_rate": 9.677322398505766e-06, + "loss": 0.8191, + "step": 48205 + }, + { + "epoch": 3.1, + "learning_rate": 9.677107709482815e-06, + "loss": 0.8103, + "step": 48206 + }, + { + "epoch": 3.1, + "learning_rate": 9.676893020459864e-06, + "loss": 0.8028, + "step": 48207 + }, + { + "epoch": 3.1, + "learning_rate": 9.676678331436914e-06, + "loss": 0.758, + "step": 48208 + }, + { + "epoch": 3.1, + "learning_rate": 9.676463642413965e-06, + "loss": 0.8437, + "step": 48209 + }, + { + "epoch": 3.1, + "learning_rate": 9.676248953391014e-06, + "loss": 0.8518, + "step": 48210 + }, + { + "epoch": 3.1, + "learning_rate": 9.676034264368065e-06, + "loss": 0.8375, + "step": 48211 + }, + { + "epoch": 3.1, + "learning_rate": 9.675819575345114e-06, + "loss": 0.83, + "step": 48212 + }, + { + "epoch": 3.1, + "learning_rate": 9.675604886322163e-06, + "loss": 0.8822, + "step": 48213 + }, + { + "epoch": 3.1, + "learning_rate": 9.675390197299213e-06, + "loss": 0.8587, + "step": 48214 + }, + { + "epoch": 3.1, + "learning_rate": 9.675175508276262e-06, + "loss": 0.8472, + "step": 48215 + }, + { + "epoch": 3.1, + "learning_rate": 9.674960819253313e-06, + "loss": 0.8422, + "step": 48216 + }, + { + "epoch": 3.1, + "learning_rate": 9.674746130230362e-06, + "loss": 0.8267, + "step": 48217 + }, + { + "epoch": 3.1, + "learning_rate": 9.674531441207412e-06, + "loss": 0.8297, + "step": 48218 + }, + { + "epoch": 3.1, + "learning_rate": 9.674316752184461e-06, + "loss": 0.833, + "step": 48219 + }, + { + "epoch": 3.1, + "learning_rate": 9.674102063161512e-06, + "loss": 0.8366, + "step": 48220 + }, + { + "epoch": 3.1, + "learning_rate": 9.673887374138561e-06, + "loss": 0.8156, + "step": 48221 + }, + { + "epoch": 3.1, + "learning_rate": 9.67367268511561e-06, + "loss": 0.8616, + "step": 48222 + }, + { + "epoch": 3.1, + "learning_rate": 9.67345799609266e-06, + "loss": 0.812, + "step": 48223 + }, + { + "epoch": 3.1, + "learning_rate": 9.673243307069711e-06, + "loss": 0.8896, + "step": 48224 + }, + { + "epoch": 3.1, + "learning_rate": 9.67302861804676e-06, + "loss": 0.8846, + "step": 48225 + }, + { + "epoch": 3.1, + "learning_rate": 9.67281392902381e-06, + "loss": 0.78, + "step": 48226 + }, + { + "epoch": 3.1, + "learning_rate": 9.67259924000086e-06, + "loss": 0.839, + "step": 48227 + }, + { + "epoch": 3.1, + "learning_rate": 9.672384550977909e-06, + "loss": 0.8359, + "step": 48228 + }, + { + "epoch": 3.1, + "learning_rate": 9.67216986195496e-06, + "loss": 0.7802, + "step": 48229 + }, + { + "epoch": 3.1, + "learning_rate": 9.671955172932008e-06, + "loss": 0.8226, + "step": 48230 + }, + { + "epoch": 3.1, + "learning_rate": 9.671740483909059e-06, + "loss": 0.8089, + "step": 48231 + }, + { + "epoch": 3.1, + "learning_rate": 9.671525794886108e-06, + "loss": 0.8418, + "step": 48232 + }, + { + "epoch": 3.1, + "learning_rate": 9.671311105863159e-06, + "loss": 0.7999, + "step": 48233 + }, + { + "epoch": 3.1, + "learning_rate": 9.671096416840208e-06, + "loss": 0.8004, + "step": 48234 + }, + { + "epoch": 3.1, + "learning_rate": 9.670881727817257e-06, + "loss": 0.8034, + "step": 48235 + }, + { + "epoch": 3.1, + "learning_rate": 9.670667038794307e-06, + "loss": 0.7775, + "step": 48236 + }, + { + "epoch": 3.1, + "learning_rate": 9.670452349771356e-06, + "loss": 0.805, + "step": 48237 + }, + { + "epoch": 3.1, + "learning_rate": 9.670237660748407e-06, + "loss": 0.7882, + "step": 48238 + }, + { + "epoch": 3.1, + "learning_rate": 9.670022971725458e-06, + "loss": 0.8004, + "step": 48239 + }, + { + "epoch": 3.1, + "learning_rate": 9.669808282702507e-06, + "loss": 0.8526, + "step": 48240 + }, + { + "epoch": 3.1, + "learning_rate": 9.669593593679556e-06, + "loss": 0.7668, + "step": 48241 + }, + { + "epoch": 3.1, + "learning_rate": 9.669378904656606e-06, + "loss": 0.8519, + "step": 48242 + }, + { + "epoch": 3.1, + "learning_rate": 9.669164215633655e-06, + "loss": 0.8097, + "step": 48243 + }, + { + "epoch": 3.1, + "learning_rate": 9.668949526610704e-06, + "loss": 0.8137, + "step": 48244 + }, + { + "epoch": 3.1, + "learning_rate": 9.668734837587755e-06, + "loss": 0.8044, + "step": 48245 + }, + { + "epoch": 3.1, + "learning_rate": 9.668520148564805e-06, + "loss": 0.7994, + "step": 48246 + }, + { + "epoch": 3.1, + "learning_rate": 9.668305459541854e-06, + "loss": 0.8439, + "step": 48247 + }, + { + "epoch": 3.1, + "learning_rate": 9.668090770518903e-06, + "loss": 0.7372, + "step": 48248 + }, + { + "epoch": 3.1, + "learning_rate": 9.667876081495954e-06, + "loss": 0.796, + "step": 48249 + }, + { + "epoch": 3.1, + "learning_rate": 9.667661392473003e-06, + "loss": 0.8715, + "step": 48250 + }, + { + "epoch": 3.1, + "learning_rate": 9.667446703450054e-06, + "loss": 0.7914, + "step": 48251 + }, + { + "epoch": 3.1, + "learning_rate": 9.667232014427104e-06, + "loss": 0.8053, + "step": 48252 + }, + { + "epoch": 3.1, + "learning_rate": 9.667017325404153e-06, + "loss": 0.848, + "step": 48253 + }, + { + "epoch": 3.1, + "learning_rate": 9.666802636381202e-06, + "loss": 0.8211, + "step": 48254 + }, + { + "epoch": 3.1, + "learning_rate": 9.666587947358253e-06, + "loss": 0.8206, + "step": 48255 + }, + { + "epoch": 3.1, + "learning_rate": 9.666373258335302e-06, + "loss": 0.8159, + "step": 48256 + }, + { + "epoch": 3.1, + "learning_rate": 9.666158569312351e-06, + "loss": 0.8099, + "step": 48257 + }, + { + "epoch": 3.1, + "learning_rate": 9.665943880289402e-06, + "loss": 0.8162, + "step": 48258 + }, + { + "epoch": 3.1, + "learning_rate": 9.66572919126645e-06, + "loss": 0.761, + "step": 48259 + }, + { + "epoch": 3.1, + "learning_rate": 9.665514502243501e-06, + "loss": 0.8202, + "step": 48260 + }, + { + "epoch": 3.1, + "learning_rate": 9.665299813220552e-06, + "loss": 0.8126, + "step": 48261 + }, + { + "epoch": 3.11, + "learning_rate": 9.6650851241976e-06, + "loss": 0.8541, + "step": 48262 + }, + { + "epoch": 3.11, + "learning_rate": 9.66487043517465e-06, + "loss": 0.7989, + "step": 48263 + }, + { + "epoch": 3.11, + "learning_rate": 9.6646557461517e-06, + "loss": 0.8325, + "step": 48264 + }, + { + "epoch": 3.11, + "learning_rate": 9.66444105712875e-06, + "loss": 0.9072, + "step": 48265 + }, + { + "epoch": 3.11, + "learning_rate": 9.664226368105798e-06, + "loss": 0.7939, + "step": 48266 + }, + { + "epoch": 3.11, + "learning_rate": 9.664011679082849e-06, + "loss": 0.862, + "step": 48267 + }, + { + "epoch": 3.11, + "learning_rate": 9.6637969900599e-06, + "loss": 0.8036, + "step": 48268 + }, + { + "epoch": 3.11, + "learning_rate": 9.663582301036949e-06, + "loss": 0.8247, + "step": 48269 + }, + { + "epoch": 3.11, + "learning_rate": 9.663367612014e-06, + "loss": 0.7784, + "step": 48270 + }, + { + "epoch": 3.11, + "learning_rate": 9.663152922991048e-06, + "loss": 0.862, + "step": 48271 + }, + { + "epoch": 3.11, + "learning_rate": 9.662938233968097e-06, + "loss": 0.8045, + "step": 48272 + }, + { + "epoch": 3.11, + "learning_rate": 9.662723544945148e-06, + "loss": 0.8098, + "step": 48273 + }, + { + "epoch": 3.11, + "learning_rate": 9.662508855922199e-06, + "loss": 0.7905, + "step": 48274 + }, + { + "epoch": 3.11, + "learning_rate": 9.662294166899247e-06, + "loss": 0.8304, + "step": 48275 + }, + { + "epoch": 3.11, + "learning_rate": 9.662079477876296e-06, + "loss": 0.8058, + "step": 48276 + }, + { + "epoch": 3.11, + "learning_rate": 9.661864788853347e-06, + "loss": 0.8516, + "step": 48277 + }, + { + "epoch": 3.11, + "learning_rate": 9.661650099830396e-06, + "loss": 0.8202, + "step": 48278 + }, + { + "epoch": 3.11, + "learning_rate": 9.661435410807447e-06, + "loss": 0.8431, + "step": 48279 + }, + { + "epoch": 3.11, + "learning_rate": 9.661220721784496e-06, + "loss": 0.7988, + "step": 48280 + }, + { + "epoch": 3.11, + "learning_rate": 9.661006032761545e-06, + "loss": 0.8474, + "step": 48281 + }, + { + "epoch": 3.11, + "learning_rate": 9.660791343738595e-06, + "loss": 0.7955, + "step": 48282 + }, + { + "epoch": 3.11, + "learning_rate": 9.660576654715646e-06, + "loss": 0.7869, + "step": 48283 + }, + { + "epoch": 3.11, + "learning_rate": 9.660361965692695e-06, + "loss": 0.7951, + "step": 48284 + }, + { + "epoch": 3.11, + "learning_rate": 9.660147276669744e-06, + "loss": 0.8108, + "step": 48285 + }, + { + "epoch": 3.11, + "learning_rate": 9.659932587646795e-06, + "loss": 0.7955, + "step": 48286 + }, + { + "epoch": 3.11, + "learning_rate": 9.659717898623844e-06, + "loss": 0.7585, + "step": 48287 + }, + { + "epoch": 3.11, + "learning_rate": 9.659503209600894e-06, + "loss": 0.8251, + "step": 48288 + }, + { + "epoch": 3.11, + "learning_rate": 9.659288520577945e-06, + "loss": 0.861, + "step": 48289 + }, + { + "epoch": 3.11, + "learning_rate": 9.659073831554994e-06, + "loss": 0.862, + "step": 48290 + }, + { + "epoch": 3.11, + "learning_rate": 9.658859142532043e-06, + "loss": 0.7937, + "step": 48291 + }, + { + "epoch": 3.11, + "learning_rate": 9.658644453509093e-06, + "loss": 0.8429, + "step": 48292 + }, + { + "epoch": 3.11, + "learning_rate": 9.658429764486142e-06, + "loss": 0.8554, + "step": 48293 + }, + { + "epoch": 3.11, + "learning_rate": 9.658215075463191e-06, + "loss": 0.7872, + "step": 48294 + }, + { + "epoch": 3.11, + "learning_rate": 9.658000386440242e-06, + "loss": 0.8305, + "step": 48295 + }, + { + "epoch": 3.11, + "learning_rate": 9.657785697417293e-06, + "loss": 0.8274, + "step": 48296 + }, + { + "epoch": 3.11, + "learning_rate": 9.657571008394342e-06, + "loss": 0.868, + "step": 48297 + }, + { + "epoch": 3.11, + "learning_rate": 9.657356319371392e-06, + "loss": 0.795, + "step": 48298 + }, + { + "epoch": 3.11, + "learning_rate": 9.657141630348441e-06, + "loss": 0.7748, + "step": 48299 + }, + { + "epoch": 3.11, + "learning_rate": 9.65692694132549e-06, + "loss": 0.8316, + "step": 48300 + }, + { + "epoch": 3.11, + "learning_rate": 9.656712252302541e-06, + "loss": 0.7425, + "step": 48301 + }, + { + "epoch": 3.11, + "learning_rate": 9.65649756327959e-06, + "loss": 0.9267, + "step": 48302 + }, + { + "epoch": 3.11, + "learning_rate": 9.65628287425664e-06, + "loss": 0.8025, + "step": 48303 + }, + { + "epoch": 3.11, + "learning_rate": 9.65606818523369e-06, + "loss": 0.7913, + "step": 48304 + }, + { + "epoch": 3.11, + "learning_rate": 9.65585349621074e-06, + "loss": 0.8996, + "step": 48305 + }, + { + "epoch": 3.11, + "learning_rate": 9.655638807187789e-06, + "loss": 0.8195, + "step": 48306 + }, + { + "epoch": 3.11, + "learning_rate": 9.65542411816484e-06, + "loss": 0.8307, + "step": 48307 + }, + { + "epoch": 3.11, + "learning_rate": 9.655209429141889e-06, + "loss": 0.8142, + "step": 48308 + }, + { + "epoch": 3.11, + "learning_rate": 9.654994740118938e-06, + "loss": 0.7978, + "step": 48309 + }, + { + "epoch": 3.11, + "learning_rate": 9.654780051095988e-06, + "loss": 0.8345, + "step": 48310 + }, + { + "epoch": 3.11, + "learning_rate": 9.654565362073039e-06, + "loss": 0.8353, + "step": 48311 + }, + { + "epoch": 3.11, + "learning_rate": 9.654350673050088e-06, + "loss": 0.7797, + "step": 48312 + }, + { + "epoch": 3.11, + "learning_rate": 9.654135984027137e-06, + "loss": 0.8717, + "step": 48313 + }, + { + "epoch": 3.11, + "learning_rate": 9.653921295004188e-06, + "loss": 0.8087, + "step": 48314 + }, + { + "epoch": 3.11, + "learning_rate": 9.653706605981237e-06, + "loss": 0.8472, + "step": 48315 + }, + { + "epoch": 3.11, + "learning_rate": 9.653491916958286e-06, + "loss": 0.7589, + "step": 48316 + }, + { + "epoch": 3.11, + "learning_rate": 9.653277227935336e-06, + "loss": 0.8119, + "step": 48317 + }, + { + "epoch": 3.11, + "learning_rate": 9.653062538912387e-06, + "loss": 0.7895, + "step": 48318 + }, + { + "epoch": 3.11, + "learning_rate": 9.652847849889436e-06, + "loss": 0.8704, + "step": 48319 + }, + { + "epoch": 3.11, + "learning_rate": 9.652633160866486e-06, + "loss": 0.8089, + "step": 48320 + }, + { + "epoch": 3.11, + "learning_rate": 9.652418471843535e-06, + "loss": 0.8219, + "step": 48321 + }, + { + "epoch": 3.11, + "learning_rate": 9.652203782820584e-06, + "loss": 0.7944, + "step": 48322 + }, + { + "epoch": 3.11, + "learning_rate": 9.651989093797635e-06, + "loss": 0.8309, + "step": 48323 + }, + { + "epoch": 3.11, + "learning_rate": 9.651774404774684e-06, + "loss": 0.7737, + "step": 48324 + }, + { + "epoch": 3.11, + "learning_rate": 9.651559715751735e-06, + "loss": 0.8503, + "step": 48325 + }, + { + "epoch": 3.11, + "learning_rate": 9.651345026728784e-06, + "loss": 0.8456, + "step": 48326 + }, + { + "epoch": 3.11, + "learning_rate": 9.651130337705834e-06, + "loss": 0.8341, + "step": 48327 + }, + { + "epoch": 3.11, + "learning_rate": 9.650915648682883e-06, + "loss": 0.7899, + "step": 48328 + }, + { + "epoch": 3.11, + "learning_rate": 9.650700959659934e-06, + "loss": 0.8797, + "step": 48329 + }, + { + "epoch": 3.11, + "learning_rate": 9.650486270636983e-06, + "loss": 0.8139, + "step": 48330 + }, + { + "epoch": 3.11, + "learning_rate": 9.650271581614032e-06, + "loss": 0.8411, + "step": 48331 + }, + { + "epoch": 3.11, + "learning_rate": 9.650056892591083e-06, + "loss": 0.8799, + "step": 48332 + }, + { + "epoch": 3.11, + "learning_rate": 9.649842203568133e-06, + "loss": 0.8574, + "step": 48333 + }, + { + "epoch": 3.11, + "learning_rate": 9.649627514545182e-06, + "loss": 0.8787, + "step": 48334 + }, + { + "epoch": 3.11, + "learning_rate": 9.649412825522231e-06, + "loss": 0.8287, + "step": 48335 + }, + { + "epoch": 3.11, + "learning_rate": 9.649198136499282e-06, + "loss": 0.8209, + "step": 48336 + }, + { + "epoch": 3.11, + "learning_rate": 9.64898344747633e-06, + "loss": 0.8384, + "step": 48337 + }, + { + "epoch": 3.11, + "learning_rate": 9.648768758453381e-06, + "loss": 0.8095, + "step": 48338 + }, + { + "epoch": 3.11, + "learning_rate": 9.64855406943043e-06, + "loss": 0.7935, + "step": 48339 + }, + { + "epoch": 3.11, + "learning_rate": 9.648339380407481e-06, + "loss": 0.8207, + "step": 48340 + }, + { + "epoch": 3.11, + "learning_rate": 9.64812469138453e-06, + "loss": 0.8064, + "step": 48341 + }, + { + "epoch": 3.11, + "learning_rate": 9.64791000236158e-06, + "loss": 0.7958, + "step": 48342 + }, + { + "epoch": 3.11, + "learning_rate": 9.64769531333863e-06, + "loss": 0.853, + "step": 48343 + }, + { + "epoch": 3.11, + "learning_rate": 9.647480624315679e-06, + "loss": 0.8216, + "step": 48344 + }, + { + "epoch": 3.11, + "learning_rate": 9.64726593529273e-06, + "loss": 0.8484, + "step": 48345 + }, + { + "epoch": 3.11, + "learning_rate": 9.647051246269778e-06, + "loss": 0.8121, + "step": 48346 + }, + { + "epoch": 3.11, + "learning_rate": 9.646836557246829e-06, + "loss": 0.8023, + "step": 48347 + }, + { + "epoch": 3.11, + "learning_rate": 9.64662186822388e-06, + "loss": 0.8418, + "step": 48348 + }, + { + "epoch": 3.11, + "learning_rate": 9.646407179200928e-06, + "loss": 0.8092, + "step": 48349 + }, + { + "epoch": 3.11, + "learning_rate": 9.646192490177977e-06, + "loss": 0.7794, + "step": 48350 + }, + { + "epoch": 3.11, + "learning_rate": 9.645977801155028e-06, + "loss": 0.8311, + "step": 48351 + }, + { + "epoch": 3.11, + "learning_rate": 9.645763112132077e-06, + "loss": 0.8036, + "step": 48352 + }, + { + "epoch": 3.11, + "learning_rate": 9.645548423109126e-06, + "loss": 0.8138, + "step": 48353 + }, + { + "epoch": 3.11, + "learning_rate": 9.645333734086177e-06, + "loss": 0.8294, + "step": 48354 + }, + { + "epoch": 3.11, + "learning_rate": 9.645119045063227e-06, + "loss": 0.8521, + "step": 48355 + }, + { + "epoch": 3.11, + "learning_rate": 9.644904356040276e-06, + "loss": 0.8153, + "step": 48356 + }, + { + "epoch": 3.11, + "learning_rate": 9.644689667017327e-06, + "loss": 0.814, + "step": 48357 + }, + { + "epoch": 3.11, + "learning_rate": 9.644474977994376e-06, + "loss": 0.848, + "step": 48358 + }, + { + "epoch": 3.11, + "learning_rate": 9.644260288971425e-06, + "loss": 0.8679, + "step": 48359 + }, + { + "epoch": 3.11, + "learning_rate": 9.644045599948476e-06, + "loss": 0.8309, + "step": 48360 + }, + { + "epoch": 3.11, + "learning_rate": 9.643830910925526e-06, + "loss": 0.8608, + "step": 48361 + }, + { + "epoch": 3.11, + "learning_rate": 9.643616221902575e-06, + "loss": 0.8398, + "step": 48362 + }, + { + "epoch": 3.11, + "learning_rate": 9.643401532879624e-06, + "loss": 0.9416, + "step": 48363 + }, + { + "epoch": 3.11, + "learning_rate": 9.643186843856675e-06, + "loss": 0.7815, + "step": 48364 + }, + { + "epoch": 3.11, + "learning_rate": 9.642972154833724e-06, + "loss": 0.8401, + "step": 48365 + }, + { + "epoch": 3.11, + "learning_rate": 9.642757465810774e-06, + "loss": 0.8273, + "step": 48366 + }, + { + "epoch": 3.11, + "learning_rate": 9.642542776787823e-06, + "loss": 0.8641, + "step": 48367 + }, + { + "epoch": 3.11, + "learning_rate": 9.642328087764872e-06, + "loss": 0.8199, + "step": 48368 + }, + { + "epoch": 3.11, + "learning_rate": 9.642113398741923e-06, + "loss": 0.8084, + "step": 48369 + }, + { + "epoch": 3.11, + "learning_rate": 9.641898709718974e-06, + "loss": 0.804, + "step": 48370 + }, + { + "epoch": 3.11, + "learning_rate": 9.641684020696023e-06, + "loss": 0.8966, + "step": 48371 + }, + { + "epoch": 3.11, + "learning_rate": 9.641469331673072e-06, + "loss": 0.8012, + "step": 48372 + }, + { + "epoch": 3.11, + "learning_rate": 9.641254642650122e-06, + "loss": 0.7567, + "step": 48373 + }, + { + "epoch": 3.11, + "learning_rate": 9.641039953627171e-06, + "loss": 0.871, + "step": 48374 + }, + { + "epoch": 3.11, + "learning_rate": 9.640825264604222e-06, + "loss": 0.8036, + "step": 48375 + }, + { + "epoch": 3.11, + "learning_rate": 9.640610575581273e-06, + "loss": 0.834, + "step": 48376 + }, + { + "epoch": 3.11, + "learning_rate": 9.640395886558322e-06, + "loss": 0.8197, + "step": 48377 + }, + { + "epoch": 3.11, + "learning_rate": 9.64018119753537e-06, + "loss": 0.7801, + "step": 48378 + }, + { + "epoch": 3.11, + "learning_rate": 9.639966508512421e-06, + "loss": 0.7796, + "step": 48379 + }, + { + "epoch": 3.11, + "learning_rate": 9.63975181948947e-06, + "loss": 0.8727, + "step": 48380 + }, + { + "epoch": 3.11, + "learning_rate": 9.639537130466519e-06, + "loss": 0.852, + "step": 48381 + }, + { + "epoch": 3.11, + "learning_rate": 9.63932244144357e-06, + "loss": 0.8379, + "step": 48382 + }, + { + "epoch": 3.11, + "learning_rate": 9.63910775242062e-06, + "loss": 0.7977, + "step": 48383 + }, + { + "epoch": 3.11, + "learning_rate": 9.63889306339767e-06, + "loss": 0.8673, + "step": 48384 + }, + { + "epoch": 3.11, + "learning_rate": 9.63867837437472e-06, + "loss": 0.7987, + "step": 48385 + }, + { + "epoch": 3.11, + "learning_rate": 9.638463685351769e-06, + "loss": 0.7953, + "step": 48386 + }, + { + "epoch": 3.11, + "learning_rate": 9.638248996328818e-06, + "loss": 0.8464, + "step": 48387 + }, + { + "epoch": 3.11, + "learning_rate": 9.638034307305869e-06, + "loss": 0.831, + "step": 48388 + }, + { + "epoch": 3.11, + "learning_rate": 9.637819618282918e-06, + "loss": 0.8237, + "step": 48389 + }, + { + "epoch": 3.11, + "learning_rate": 9.637604929259968e-06, + "loss": 0.8048, + "step": 48390 + }, + { + "epoch": 3.11, + "learning_rate": 9.637390240237017e-06, + "loss": 0.7773, + "step": 48391 + }, + { + "epoch": 3.11, + "learning_rate": 9.637175551214068e-06, + "loss": 0.8213, + "step": 48392 + }, + { + "epoch": 3.11, + "learning_rate": 9.636960862191117e-06, + "loss": 0.8983, + "step": 48393 + }, + { + "epoch": 3.11, + "learning_rate": 9.636746173168167e-06, + "loss": 0.8605, + "step": 48394 + }, + { + "epoch": 3.11, + "learning_rate": 9.636531484145216e-06, + "loss": 0.787, + "step": 48395 + }, + { + "epoch": 3.11, + "learning_rate": 9.636316795122265e-06, + "loss": 0.8457, + "step": 48396 + }, + { + "epoch": 3.11, + "learning_rate": 9.636102106099316e-06, + "loss": 0.7731, + "step": 48397 + }, + { + "epoch": 3.11, + "learning_rate": 9.635887417076367e-06, + "loss": 0.7979, + "step": 48398 + }, + { + "epoch": 3.11, + "learning_rate": 9.635672728053416e-06, + "loss": 0.8912, + "step": 48399 + }, + { + "epoch": 3.11, + "learning_rate": 9.635458039030465e-06, + "loss": 0.8322, + "step": 48400 + }, + { + "epoch": 3.11, + "learning_rate": 9.635243350007515e-06, + "loss": 0.8342, + "step": 48401 + }, + { + "epoch": 3.11, + "learning_rate": 9.635028660984564e-06, + "loss": 0.829, + "step": 48402 + }, + { + "epoch": 3.11, + "learning_rate": 9.634813971961613e-06, + "loss": 0.8561, + "step": 48403 + }, + { + "epoch": 3.11, + "learning_rate": 9.634599282938664e-06, + "loss": 0.7968, + "step": 48404 + }, + { + "epoch": 3.11, + "learning_rate": 9.634384593915715e-06, + "loss": 0.8852, + "step": 48405 + }, + { + "epoch": 3.11, + "learning_rate": 9.634169904892764e-06, + "loss": 0.803, + "step": 48406 + }, + { + "epoch": 3.11, + "learning_rate": 9.633955215869814e-06, + "loss": 0.8285, + "step": 48407 + }, + { + "epoch": 3.11, + "learning_rate": 9.633740526846863e-06, + "loss": 0.8995, + "step": 48408 + }, + { + "epoch": 3.11, + "learning_rate": 9.633525837823912e-06, + "loss": 0.8645, + "step": 48409 + }, + { + "epoch": 3.11, + "learning_rate": 9.633311148800963e-06, + "loss": 0.8257, + "step": 48410 + }, + { + "epoch": 3.11, + "learning_rate": 9.633096459778012e-06, + "loss": 0.7806, + "step": 48411 + }, + { + "epoch": 3.11, + "learning_rate": 9.632881770755062e-06, + "loss": 0.8657, + "step": 48412 + }, + { + "epoch": 3.11, + "learning_rate": 9.632667081732111e-06, + "loss": 0.8426, + "step": 48413 + }, + { + "epoch": 3.11, + "learning_rate": 9.632452392709162e-06, + "loss": 0.8282, + "step": 48414 + }, + { + "epoch": 3.11, + "learning_rate": 9.632237703686211e-06, + "loss": 0.8611, + "step": 48415 + }, + { + "epoch": 3.11, + "learning_rate": 9.632023014663262e-06, + "loss": 0.8055, + "step": 48416 + }, + { + "epoch": 3.12, + "learning_rate": 9.63180832564031e-06, + "loss": 0.8163, + "step": 48417 + }, + { + "epoch": 3.12, + "learning_rate": 9.63159363661736e-06, + "loss": 0.8304, + "step": 48418 + }, + { + "epoch": 3.12, + "learning_rate": 9.63137894759441e-06, + "loss": 0.8437, + "step": 48419 + }, + { + "epoch": 3.12, + "learning_rate": 9.631164258571461e-06, + "loss": 0.754, + "step": 48420 + }, + { + "epoch": 3.12, + "learning_rate": 9.63094956954851e-06, + "loss": 0.8074, + "step": 48421 + }, + { + "epoch": 3.12, + "learning_rate": 9.630734880525559e-06, + "loss": 0.8491, + "step": 48422 + }, + { + "epoch": 3.12, + "learning_rate": 9.63052019150261e-06, + "loss": 0.8138, + "step": 48423 + }, + { + "epoch": 3.12, + "learning_rate": 9.630305502479658e-06, + "loss": 0.8181, + "step": 48424 + }, + { + "epoch": 3.12, + "learning_rate": 9.630090813456709e-06, + "loss": 0.8302, + "step": 48425 + }, + { + "epoch": 3.12, + "learning_rate": 9.629876124433758e-06, + "loss": 0.8102, + "step": 48426 + }, + { + "epoch": 3.12, + "learning_rate": 9.629661435410809e-06, + "loss": 0.7986, + "step": 48427 + }, + { + "epoch": 3.12, + "learning_rate": 9.629446746387858e-06, + "loss": 0.8984, + "step": 48428 + }, + { + "epoch": 3.12, + "learning_rate": 9.629232057364908e-06, + "loss": 0.8014, + "step": 48429 + }, + { + "epoch": 3.12, + "learning_rate": 9.629017368341957e-06, + "loss": 0.8965, + "step": 48430 + }, + { + "epoch": 3.12, + "learning_rate": 9.628802679319006e-06, + "loss": 0.8008, + "step": 48431 + }, + { + "epoch": 3.12, + "learning_rate": 9.628587990296057e-06, + "loss": 0.8512, + "step": 48432 + }, + { + "epoch": 3.12, + "learning_rate": 9.628373301273106e-06, + "loss": 0.7904, + "step": 48433 + }, + { + "epoch": 3.12, + "learning_rate": 9.628158612250157e-06, + "loss": 0.8105, + "step": 48434 + }, + { + "epoch": 3.12, + "learning_rate": 9.627943923227207e-06, + "loss": 0.8253, + "step": 48435 + }, + { + "epoch": 3.12, + "learning_rate": 9.627729234204256e-06, + "loss": 0.8348, + "step": 48436 + }, + { + "epoch": 3.12, + "learning_rate": 9.627514545181305e-06, + "loss": 0.8904, + "step": 48437 + }, + { + "epoch": 3.12, + "learning_rate": 9.627299856158356e-06, + "loss": 0.849, + "step": 48438 + }, + { + "epoch": 3.12, + "learning_rate": 9.627085167135405e-06, + "loss": 0.7855, + "step": 48439 + }, + { + "epoch": 3.12, + "learning_rate": 9.626870478112454e-06, + "loss": 0.86, + "step": 48440 + }, + { + "epoch": 3.12, + "learning_rate": 9.626655789089504e-06, + "loss": 0.7835, + "step": 48441 + }, + { + "epoch": 3.12, + "learning_rate": 9.626441100066555e-06, + "loss": 0.8168, + "step": 48442 + }, + { + "epoch": 3.12, + "learning_rate": 9.626226411043604e-06, + "loss": 0.8284, + "step": 48443 + }, + { + "epoch": 3.12, + "learning_rate": 9.626011722020655e-06, + "loss": 0.8493, + "step": 48444 + }, + { + "epoch": 3.12, + "learning_rate": 9.625797032997704e-06, + "loss": 0.8152, + "step": 48445 + }, + { + "epoch": 3.12, + "learning_rate": 9.625582343974753e-06, + "loss": 0.884, + "step": 48446 + }, + { + "epoch": 3.12, + "learning_rate": 9.625367654951803e-06, + "loss": 0.812, + "step": 48447 + }, + { + "epoch": 3.12, + "learning_rate": 9.625152965928854e-06, + "loss": 0.8131, + "step": 48448 + }, + { + "epoch": 3.12, + "learning_rate": 9.624938276905903e-06, + "loss": 0.7722, + "step": 48449 + }, + { + "epoch": 3.12, + "learning_rate": 9.624723587882952e-06, + "loss": 0.84, + "step": 48450 + }, + { + "epoch": 3.12, + "learning_rate": 9.624508898860003e-06, + "loss": 0.8305, + "step": 48451 + }, + { + "epoch": 3.12, + "learning_rate": 9.624294209837051e-06, + "loss": 0.8084, + "step": 48452 + }, + { + "epoch": 3.12, + "learning_rate": 9.624079520814102e-06, + "loss": 0.8473, + "step": 48453 + }, + { + "epoch": 3.12, + "learning_rate": 9.623864831791151e-06, + "loss": 0.8706, + "step": 48454 + }, + { + "epoch": 3.12, + "learning_rate": 9.6236501427682e-06, + "loss": 0.7865, + "step": 48455 + }, + { + "epoch": 3.12, + "learning_rate": 9.62343545374525e-06, + "loss": 0.8264, + "step": 48456 + }, + { + "epoch": 3.12, + "learning_rate": 9.623220764722301e-06, + "loss": 0.8149, + "step": 48457 + }, + { + "epoch": 3.12, + "learning_rate": 9.62300607569935e-06, + "loss": 0.8121, + "step": 48458 + }, + { + "epoch": 3.12, + "learning_rate": 9.6227913866764e-06, + "loss": 0.7731, + "step": 48459 + }, + { + "epoch": 3.12, + "learning_rate": 9.62257669765345e-06, + "loss": 0.7747, + "step": 48460 + }, + { + "epoch": 3.12, + "learning_rate": 9.622362008630499e-06, + "loss": 0.8083, + "step": 48461 + }, + { + "epoch": 3.12, + "learning_rate": 9.62214731960755e-06, + "loss": 0.7912, + "step": 48462 + }, + { + "epoch": 3.12, + "learning_rate": 9.6219326305846e-06, + "loss": 0.8111, + "step": 48463 + }, + { + "epoch": 3.12, + "learning_rate": 9.62171794156165e-06, + "loss": 0.7837, + "step": 48464 + }, + { + "epoch": 3.12, + "learning_rate": 9.621503252538698e-06, + "loss": 0.7989, + "step": 48465 + }, + { + "epoch": 3.12, + "learning_rate": 9.621288563515749e-06, + "loss": 0.8145, + "step": 48466 + }, + { + "epoch": 3.12, + "learning_rate": 9.621073874492798e-06, + "loss": 0.8231, + "step": 48467 + }, + { + "epoch": 3.12, + "learning_rate": 9.620859185469847e-06, + "loss": 0.7951, + "step": 48468 + }, + { + "epoch": 3.12, + "learning_rate": 9.620644496446897e-06, + "loss": 0.8478, + "step": 48469 + }, + { + "epoch": 3.12, + "learning_rate": 9.620429807423948e-06, + "loss": 0.7828, + "step": 48470 + }, + { + "epoch": 3.12, + "learning_rate": 9.620215118400997e-06, + "loss": 0.8236, + "step": 48471 + }, + { + "epoch": 3.12, + "learning_rate": 9.620000429378048e-06, + "loss": 0.8419, + "step": 48472 + }, + { + "epoch": 3.12, + "learning_rate": 9.619785740355097e-06, + "loss": 0.8237, + "step": 48473 + }, + { + "epoch": 3.12, + "learning_rate": 9.619571051332146e-06, + "loss": 0.8107, + "step": 48474 + }, + { + "epoch": 3.12, + "learning_rate": 9.619356362309196e-06, + "loss": 0.8518, + "step": 48475 + }, + { + "epoch": 3.12, + "learning_rate": 9.619141673286245e-06, + "loss": 0.798, + "step": 48476 + }, + { + "epoch": 3.12, + "learning_rate": 9.618926984263294e-06, + "loss": 0.8075, + "step": 48477 + }, + { + "epoch": 3.12, + "learning_rate": 9.618712295240345e-06, + "loss": 0.8604, + "step": 48478 + }, + { + "epoch": 3.12, + "learning_rate": 9.618497606217396e-06, + "loss": 0.9274, + "step": 48479 + }, + { + "epoch": 3.12, + "learning_rate": 9.618282917194445e-06, + "loss": 0.8558, + "step": 48480 + }, + { + "epoch": 3.12, + "learning_rate": 9.618068228171493e-06, + "loss": 0.8109, + "step": 48481 + }, + { + "epoch": 3.12, + "learning_rate": 9.617853539148544e-06, + "loss": 0.7911, + "step": 48482 + }, + { + "epoch": 3.12, + "learning_rate": 9.617638850125593e-06, + "loss": 0.8427, + "step": 48483 + }, + { + "epoch": 3.12, + "learning_rate": 9.617424161102644e-06, + "loss": 0.8092, + "step": 48484 + }, + { + "epoch": 3.12, + "learning_rate": 9.617209472079694e-06, + "loss": 0.8483, + "step": 48485 + }, + { + "epoch": 3.12, + "learning_rate": 9.616994783056743e-06, + "loss": 0.8456, + "step": 48486 + }, + { + "epoch": 3.12, + "learning_rate": 9.616780094033792e-06, + "loss": 0.8772, + "step": 48487 + }, + { + "epoch": 3.12, + "learning_rate": 9.616565405010843e-06, + "loss": 0.8656, + "step": 48488 + }, + { + "epoch": 3.12, + "learning_rate": 9.616350715987892e-06, + "loss": 0.9053, + "step": 48489 + }, + { + "epoch": 3.12, + "learning_rate": 9.616136026964941e-06, + "loss": 0.8006, + "step": 48490 + }, + { + "epoch": 3.12, + "learning_rate": 9.615921337941992e-06, + "loss": 0.8249, + "step": 48491 + }, + { + "epoch": 3.12, + "learning_rate": 9.615706648919042e-06, + "loss": 0.7839, + "step": 48492 + }, + { + "epoch": 3.12, + "learning_rate": 9.615491959896091e-06, + "loss": 0.8575, + "step": 48493 + }, + { + "epoch": 3.12, + "learning_rate": 9.615277270873142e-06, + "loss": 0.8089, + "step": 48494 + }, + { + "epoch": 3.12, + "learning_rate": 9.615062581850191e-06, + "loss": 0.8001, + "step": 48495 + }, + { + "epoch": 3.12, + "learning_rate": 9.61484789282724e-06, + "loss": 0.8476, + "step": 48496 + }, + { + "epoch": 3.12, + "learning_rate": 9.61463320380429e-06, + "loss": 0.8521, + "step": 48497 + }, + { + "epoch": 3.12, + "learning_rate": 9.61441851478134e-06, + "loss": 0.7958, + "step": 48498 + }, + { + "epoch": 3.12, + "learning_rate": 9.61420382575839e-06, + "loss": 0.8897, + "step": 48499 + }, + { + "epoch": 3.12, + "learning_rate": 9.613989136735439e-06, + "loss": 0.7885, + "step": 48500 + }, + { + "epoch": 3.12, + "learning_rate": 9.61377444771249e-06, + "loss": 0.8342, + "step": 48501 + }, + { + "epoch": 3.12, + "learning_rate": 9.613559758689539e-06, + "loss": 0.8055, + "step": 48502 + }, + { + "epoch": 3.12, + "learning_rate": 9.61334506966659e-06, + "loss": 0.8279, + "step": 48503 + }, + { + "epoch": 3.12, + "learning_rate": 9.613130380643638e-06, + "loss": 0.8382, + "step": 48504 + }, + { + "epoch": 3.12, + "learning_rate": 9.612915691620687e-06, + "loss": 0.7381, + "step": 48505 + }, + { + "epoch": 3.12, + "learning_rate": 9.612701002597738e-06, + "loss": 0.8645, + "step": 48506 + }, + { + "epoch": 3.12, + "learning_rate": 9.612486313574789e-06, + "loss": 0.8262, + "step": 48507 + }, + { + "epoch": 3.12, + "learning_rate": 9.612271624551838e-06, + "loss": 0.8389, + "step": 48508 + }, + { + "epoch": 3.12, + "learning_rate": 9.612056935528887e-06, + "loss": 0.7107, + "step": 48509 + }, + { + "epoch": 3.12, + "learning_rate": 9.611842246505937e-06, + "loss": 0.8297, + "step": 48510 + }, + { + "epoch": 3.12, + "learning_rate": 9.611627557482986e-06, + "loss": 0.7849, + "step": 48511 + }, + { + "epoch": 3.12, + "learning_rate": 9.611412868460037e-06, + "loss": 0.8448, + "step": 48512 + }, + { + "epoch": 3.12, + "learning_rate": 9.611198179437086e-06, + "loss": 0.88, + "step": 48513 + }, + { + "epoch": 3.12, + "learning_rate": 9.610983490414136e-06, + "loss": 0.8225, + "step": 48514 + }, + { + "epoch": 3.12, + "learning_rate": 9.610768801391185e-06, + "loss": 0.7788, + "step": 48515 + }, + { + "epoch": 3.12, + "learning_rate": 9.610554112368236e-06, + "loss": 0.8539, + "step": 48516 + }, + { + "epoch": 3.12, + "learning_rate": 9.610339423345285e-06, + "loss": 0.8669, + "step": 48517 + }, + { + "epoch": 3.12, + "learning_rate": 9.610124734322334e-06, + "loss": 0.7776, + "step": 48518 + }, + { + "epoch": 3.12, + "learning_rate": 9.609910045299385e-06, + "loss": 0.8351, + "step": 48519 + }, + { + "epoch": 3.12, + "learning_rate": 9.609695356276434e-06, + "loss": 0.82, + "step": 48520 + }, + { + "epoch": 3.12, + "learning_rate": 9.609480667253484e-06, + "loss": 0.8755, + "step": 48521 + }, + { + "epoch": 3.12, + "learning_rate": 9.609265978230535e-06, + "loss": 0.812, + "step": 48522 + }, + { + "epoch": 3.12, + "learning_rate": 9.609051289207584e-06, + "loss": 0.8804, + "step": 48523 + }, + { + "epoch": 3.12, + "learning_rate": 9.608836600184633e-06, + "loss": 0.8651, + "step": 48524 + }, + { + "epoch": 3.12, + "learning_rate": 9.608621911161684e-06, + "loss": 0.8018, + "step": 48525 + }, + { + "epoch": 3.12, + "learning_rate": 9.608407222138732e-06, + "loss": 0.8212, + "step": 48526 + }, + { + "epoch": 3.12, + "learning_rate": 9.608192533115781e-06, + "loss": 0.8695, + "step": 48527 + }, + { + "epoch": 3.12, + "learning_rate": 9.607977844092832e-06, + "loss": 0.795, + "step": 48528 + }, + { + "epoch": 3.12, + "learning_rate": 9.607763155069883e-06, + "loss": 0.8309, + "step": 48529 + }, + { + "epoch": 3.12, + "learning_rate": 9.607548466046932e-06, + "loss": 0.8537, + "step": 48530 + }, + { + "epoch": 3.12, + "learning_rate": 9.607333777023982e-06, + "loss": 0.7858, + "step": 48531 + }, + { + "epoch": 3.12, + "learning_rate": 9.607119088001031e-06, + "loss": 0.819, + "step": 48532 + }, + { + "epoch": 3.12, + "learning_rate": 9.60690439897808e-06, + "loss": 0.8476, + "step": 48533 + }, + { + "epoch": 3.12, + "learning_rate": 9.606689709955131e-06, + "loss": 0.8053, + "step": 48534 + }, + { + "epoch": 3.12, + "learning_rate": 9.60647502093218e-06, + "loss": 0.8814, + "step": 48535 + }, + { + "epoch": 3.12, + "learning_rate": 9.60626033190923e-06, + "loss": 0.7897, + "step": 48536 + }, + { + "epoch": 3.12, + "learning_rate": 9.60604564288628e-06, + "loss": 0.8847, + "step": 48537 + }, + { + "epoch": 3.12, + "learning_rate": 9.60583095386333e-06, + "loss": 0.8816, + "step": 48538 + }, + { + "epoch": 3.12, + "learning_rate": 9.60561626484038e-06, + "loss": 0.8058, + "step": 48539 + }, + { + "epoch": 3.12, + "learning_rate": 9.60540157581743e-06, + "loss": 0.8407, + "step": 48540 + }, + { + "epoch": 3.12, + "learning_rate": 9.605186886794479e-06, + "loss": 0.7804, + "step": 48541 + }, + { + "epoch": 3.12, + "learning_rate": 9.604972197771528e-06, + "loss": 0.8215, + "step": 48542 + }, + { + "epoch": 3.12, + "learning_rate": 9.604757508748578e-06, + "loss": 0.8234, + "step": 48543 + }, + { + "epoch": 3.12, + "learning_rate": 9.604542819725629e-06, + "loss": 0.9054, + "step": 48544 + }, + { + "epoch": 3.12, + "learning_rate": 9.604328130702678e-06, + "loss": 0.8455, + "step": 48545 + }, + { + "epoch": 3.12, + "learning_rate": 9.604113441679727e-06, + "loss": 0.8458, + "step": 48546 + }, + { + "epoch": 3.12, + "learning_rate": 9.603898752656778e-06, + "loss": 0.8291, + "step": 48547 + }, + { + "epoch": 3.12, + "learning_rate": 9.603684063633827e-06, + "loss": 0.8583, + "step": 48548 + }, + { + "epoch": 3.12, + "learning_rate": 9.603469374610877e-06, + "loss": 0.8016, + "step": 48549 + }, + { + "epoch": 3.12, + "learning_rate": 9.603254685587928e-06, + "loss": 0.8012, + "step": 48550 + }, + { + "epoch": 3.12, + "learning_rate": 9.603039996564977e-06, + "loss": 0.8294, + "step": 48551 + }, + { + "epoch": 3.12, + "learning_rate": 9.602825307542026e-06, + "loss": 0.8202, + "step": 48552 + }, + { + "epoch": 3.12, + "learning_rate": 9.602610618519077e-06, + "loss": 0.7796, + "step": 48553 + }, + { + "epoch": 3.12, + "learning_rate": 9.602395929496126e-06, + "loss": 0.8446, + "step": 48554 + }, + { + "epoch": 3.12, + "learning_rate": 9.602181240473174e-06, + "loss": 0.8042, + "step": 48555 + }, + { + "epoch": 3.12, + "learning_rate": 9.601966551450225e-06, + "loss": 0.8353, + "step": 48556 + }, + { + "epoch": 3.12, + "learning_rate": 9.601751862427276e-06, + "loss": 0.8441, + "step": 48557 + }, + { + "epoch": 3.12, + "learning_rate": 9.601537173404325e-06, + "loss": 0.8032, + "step": 48558 + }, + { + "epoch": 3.12, + "learning_rate": 9.601322484381375e-06, + "loss": 0.8451, + "step": 48559 + }, + { + "epoch": 3.12, + "learning_rate": 9.601107795358424e-06, + "loss": 0.8214, + "step": 48560 + }, + { + "epoch": 3.12, + "learning_rate": 9.600893106335473e-06, + "loss": 0.732, + "step": 48561 + }, + { + "epoch": 3.12, + "learning_rate": 9.600678417312524e-06, + "loss": 0.8061, + "step": 48562 + }, + { + "epoch": 3.12, + "learning_rate": 9.600463728289573e-06, + "loss": 0.7969, + "step": 48563 + }, + { + "epoch": 3.12, + "learning_rate": 9.600249039266622e-06, + "loss": 0.8214, + "step": 48564 + }, + { + "epoch": 3.12, + "learning_rate": 9.600034350243673e-06, + "loss": 0.8571, + "step": 48565 + }, + { + "epoch": 3.12, + "learning_rate": 9.599819661220723e-06, + "loss": 0.7767, + "step": 48566 + }, + { + "epoch": 3.12, + "learning_rate": 9.599604972197772e-06, + "loss": 0.8699, + "step": 48567 + }, + { + "epoch": 3.12, + "learning_rate": 9.599390283174821e-06, + "loss": 0.8104, + "step": 48568 + }, + { + "epoch": 3.12, + "learning_rate": 9.599175594151872e-06, + "loss": 0.8426, + "step": 48569 + }, + { + "epoch": 3.12, + "learning_rate": 9.59896090512892e-06, + "loss": 0.8323, + "step": 48570 + }, + { + "epoch": 3.12, + "learning_rate": 9.598746216105971e-06, + "loss": 0.7414, + "step": 48571 + }, + { + "epoch": 3.13, + "learning_rate": 9.598531527083022e-06, + "loss": 0.7865, + "step": 48572 + }, + { + "epoch": 3.13, + "learning_rate": 9.598316838060071e-06, + "loss": 0.8251, + "step": 48573 + }, + { + "epoch": 3.13, + "learning_rate": 9.59810214903712e-06, + "loss": 0.8955, + "step": 48574 + }, + { + "epoch": 3.13, + "learning_rate": 9.59788746001417e-06, + "loss": 0.7606, + "step": 48575 + }, + { + "epoch": 3.13, + "learning_rate": 9.59767277099122e-06, + "loss": 0.7375, + "step": 48576 + }, + { + "epoch": 3.13, + "learning_rate": 9.597458081968269e-06, + "loss": 0.8527, + "step": 48577 + }, + { + "epoch": 3.13, + "learning_rate": 9.59724339294532e-06, + "loss": 0.7789, + "step": 48578 + }, + { + "epoch": 3.13, + "learning_rate": 9.59702870392237e-06, + "loss": 0.8049, + "step": 48579 + }, + { + "epoch": 3.13, + "learning_rate": 9.596814014899419e-06, + "loss": 0.8345, + "step": 48580 + }, + { + "epoch": 3.13, + "learning_rate": 9.59659932587647e-06, + "loss": 0.7848, + "step": 48581 + }, + { + "epoch": 3.13, + "learning_rate": 9.596384636853519e-06, + "loss": 0.8755, + "step": 48582 + }, + { + "epoch": 3.13, + "learning_rate": 9.596169947830568e-06, + "loss": 0.8568, + "step": 48583 + }, + { + "epoch": 3.13, + "learning_rate": 9.595955258807618e-06, + "loss": 0.8403, + "step": 48584 + }, + { + "epoch": 3.13, + "learning_rate": 9.595740569784667e-06, + "loss": 0.8059, + "step": 48585 + }, + { + "epoch": 3.13, + "learning_rate": 9.595525880761718e-06, + "loss": 0.8806, + "step": 48586 + }, + { + "epoch": 3.13, + "learning_rate": 9.595311191738767e-06, + "loss": 0.8063, + "step": 48587 + }, + { + "epoch": 3.13, + "learning_rate": 9.595096502715817e-06, + "loss": 0.8066, + "step": 48588 + }, + { + "epoch": 3.13, + "learning_rate": 9.594881813692866e-06, + "loss": 0.8519, + "step": 48589 + }, + { + "epoch": 3.13, + "learning_rate": 9.594667124669917e-06, + "loss": 0.8355, + "step": 48590 + }, + { + "epoch": 3.13, + "learning_rate": 9.594452435646966e-06, + "loss": 0.8578, + "step": 48591 + }, + { + "epoch": 3.13, + "learning_rate": 9.594237746624015e-06, + "loss": 0.852, + "step": 48592 + }, + { + "epoch": 3.13, + "learning_rate": 9.594237746624015e-06, + "loss": 0.8526, + "step": 48593 + }, + { + "epoch": 3.13, + "learning_rate": 9.594023057601066e-06, + "loss": 0.7902, + "step": 48594 + }, + { + "epoch": 3.13, + "learning_rate": 9.593808368578116e-06, + "loss": 0.7723, + "step": 48595 + }, + { + "epoch": 3.13, + "learning_rate": 9.593593679555165e-06, + "loss": 0.7929, + "step": 48596 + }, + { + "epoch": 3.13, + "learning_rate": 9.593378990532214e-06, + "loss": 0.8766, + "step": 48597 + }, + { + "epoch": 3.13, + "learning_rate": 9.593164301509265e-06, + "loss": 0.8304, + "step": 48598 + }, + { + "epoch": 3.13, + "learning_rate": 9.592949612486314e-06, + "loss": 0.7579, + "step": 48599 + }, + { + "epoch": 3.13, + "learning_rate": 9.592734923463365e-06, + "loss": 0.7799, + "step": 48600 + }, + { + "epoch": 3.13, + "learning_rate": 9.592520234440413e-06, + "loss": 0.8353, + "step": 48601 + }, + { + "epoch": 3.13, + "learning_rate": 9.592305545417464e-06, + "loss": 0.8001, + "step": 48602 + }, + { + "epoch": 3.13, + "learning_rate": 9.592090856394513e-06, + "loss": 0.7774, + "step": 48603 + }, + { + "epoch": 3.13, + "learning_rate": 9.591876167371564e-06, + "loss": 0.8511, + "step": 48604 + }, + { + "epoch": 3.13, + "learning_rate": 9.591661478348613e-06, + "loss": 0.8451, + "step": 48605 + }, + { + "epoch": 3.13, + "learning_rate": 9.591446789325662e-06, + "loss": 0.8331, + "step": 48606 + }, + { + "epoch": 3.13, + "learning_rate": 9.591232100302712e-06, + "loss": 0.7858, + "step": 48607 + }, + { + "epoch": 3.13, + "learning_rate": 9.591017411279761e-06, + "loss": 0.7955, + "step": 48608 + }, + { + "epoch": 3.13, + "learning_rate": 9.590802722256812e-06, + "loss": 0.7759, + "step": 48609 + }, + { + "epoch": 3.13, + "learning_rate": 9.590588033233863e-06, + "loss": 0.7656, + "step": 48610 + }, + { + "epoch": 3.13, + "learning_rate": 9.590373344210912e-06, + "loss": 0.9053, + "step": 48611 + }, + { + "epoch": 3.13, + "learning_rate": 9.59015865518796e-06, + "loss": 0.8407, + "step": 48612 + }, + { + "epoch": 3.13, + "learning_rate": 9.589943966165011e-06, + "loss": 0.8457, + "step": 48613 + }, + { + "epoch": 3.13, + "learning_rate": 9.58972927714206e-06, + "loss": 0.8131, + "step": 48614 + }, + { + "epoch": 3.13, + "learning_rate": 9.58951458811911e-06, + "loss": 0.8265, + "step": 48615 + }, + { + "epoch": 3.13, + "learning_rate": 9.58929989909616e-06, + "loss": 0.7967, + "step": 48616 + }, + { + "epoch": 3.13, + "learning_rate": 9.58908521007321e-06, + "loss": 0.8392, + "step": 48617 + }, + { + "epoch": 3.13, + "learning_rate": 9.58887052105026e-06, + "loss": 0.8516, + "step": 48618 + }, + { + "epoch": 3.13, + "learning_rate": 9.58865583202731e-06, + "loss": 0.8745, + "step": 48619 + }, + { + "epoch": 3.13, + "learning_rate": 9.588441143004359e-06, + "loss": 0.7721, + "step": 48620 + }, + { + "epoch": 3.13, + "learning_rate": 9.588226453981408e-06, + "loss": 0.8894, + "step": 48621 + }, + { + "epoch": 3.13, + "learning_rate": 9.588011764958459e-06, + "loss": 0.8366, + "step": 48622 + }, + { + "epoch": 3.13, + "learning_rate": 9.587797075935508e-06, + "loss": 0.8572, + "step": 48623 + }, + { + "epoch": 3.13, + "learning_rate": 9.587582386912558e-06, + "loss": 0.8812, + "step": 48624 + }, + { + "epoch": 3.13, + "learning_rate": 9.587367697889607e-06, + "loss": 0.7085, + "step": 48625 + }, + { + "epoch": 3.13, + "learning_rate": 9.587153008866658e-06, + "loss": 0.8263, + "step": 48626 + }, + { + "epoch": 3.13, + "learning_rate": 9.586938319843707e-06, + "loss": 0.8422, + "step": 48627 + }, + { + "epoch": 3.13, + "learning_rate": 9.586723630820758e-06, + "loss": 0.7923, + "step": 48628 + }, + { + "epoch": 3.13, + "learning_rate": 9.586508941797807e-06, + "loss": 0.8434, + "step": 48629 + }, + { + "epoch": 3.13, + "learning_rate": 9.586294252774855e-06, + "loss": 0.854, + "step": 48630 + }, + { + "epoch": 3.13, + "learning_rate": 9.586079563751906e-06, + "loss": 0.8064, + "step": 48631 + }, + { + "epoch": 3.13, + "learning_rate": 9.585864874728957e-06, + "loss": 0.7844, + "step": 48632 + }, + { + "epoch": 3.13, + "learning_rate": 9.585650185706006e-06, + "loss": 0.8543, + "step": 48633 + }, + { + "epoch": 3.13, + "learning_rate": 9.585435496683055e-06, + "loss": 0.8547, + "step": 48634 + }, + { + "epoch": 3.13, + "learning_rate": 9.585220807660105e-06, + "loss": 0.7992, + "step": 48635 + }, + { + "epoch": 3.13, + "learning_rate": 9.585006118637154e-06, + "loss": 0.8575, + "step": 48636 + }, + { + "epoch": 3.13, + "learning_rate": 9.584791429614203e-06, + "loss": 0.833, + "step": 48637 + }, + { + "epoch": 3.13, + "learning_rate": 9.584576740591256e-06, + "loss": 0.8169, + "step": 48638 + }, + { + "epoch": 3.13, + "learning_rate": 9.584362051568305e-06, + "loss": 0.8235, + "step": 48639 + }, + { + "epoch": 3.13, + "learning_rate": 9.584147362545354e-06, + "loss": 0.9047, + "step": 48640 + }, + { + "epoch": 3.13, + "learning_rate": 9.583932673522404e-06, + "loss": 0.8445, + "step": 48641 + }, + { + "epoch": 3.13, + "learning_rate": 9.583717984499453e-06, + "loss": 0.8514, + "step": 48642 + }, + { + "epoch": 3.13, + "learning_rate": 9.583503295476502e-06, + "loss": 0.7921, + "step": 48643 + }, + { + "epoch": 3.13, + "learning_rate": 9.583288606453553e-06, + "loss": 0.8547, + "step": 48644 + }, + { + "epoch": 3.13, + "learning_rate": 9.583073917430602e-06, + "loss": 0.8132, + "step": 48645 + }, + { + "epoch": 3.13, + "learning_rate": 9.582859228407652e-06, + "loss": 0.8331, + "step": 48646 + }, + { + "epoch": 3.13, + "learning_rate": 9.582644539384701e-06, + "loss": 0.7266, + "step": 48647 + }, + { + "epoch": 3.13, + "learning_rate": 9.582429850361752e-06, + "loss": 0.7912, + "step": 48648 + }, + { + "epoch": 3.13, + "learning_rate": 9.582215161338801e-06, + "loss": 0.8845, + "step": 48649 + }, + { + "epoch": 3.13, + "learning_rate": 9.582000472315852e-06, + "loss": 0.8079, + "step": 48650 + }, + { + "epoch": 3.13, + "learning_rate": 9.5817857832929e-06, + "loss": 0.7503, + "step": 48651 + }, + { + "epoch": 3.13, + "learning_rate": 9.58157109426995e-06, + "loss": 0.8511, + "step": 48652 + }, + { + "epoch": 3.13, + "learning_rate": 9.581356405247e-06, + "loss": 0.8049, + "step": 48653 + }, + { + "epoch": 3.13, + "learning_rate": 9.581141716224051e-06, + "loss": 0.8186, + "step": 48654 + }, + { + "epoch": 3.13, + "learning_rate": 9.5809270272011e-06, + "loss": 0.8321, + "step": 48655 + }, + { + "epoch": 3.13, + "learning_rate": 9.580712338178149e-06, + "loss": 0.836, + "step": 48656 + }, + { + "epoch": 3.13, + "learning_rate": 9.5804976491552e-06, + "loss": 0.8221, + "step": 48657 + }, + { + "epoch": 3.13, + "learning_rate": 9.580282960132249e-06, + "loss": 0.7919, + "step": 48658 + }, + { + "epoch": 3.13, + "learning_rate": 9.5800682711093e-06, + "loss": 0.8569, + "step": 48659 + }, + { + "epoch": 3.13, + "learning_rate": 9.57985358208635e-06, + "loss": 0.7858, + "step": 48660 + }, + { + "epoch": 3.13, + "learning_rate": 9.579638893063399e-06, + "loss": 0.8068, + "step": 48661 + }, + { + "epoch": 3.13, + "learning_rate": 9.579424204040448e-06, + "loss": 0.8155, + "step": 48662 + }, + { + "epoch": 3.13, + "learning_rate": 9.579209515017498e-06, + "loss": 0.8461, + "step": 48663 + }, + { + "epoch": 3.13, + "learning_rate": 9.578994825994547e-06, + "loss": 0.8332, + "step": 48664 + }, + { + "epoch": 3.13, + "learning_rate": 9.578780136971596e-06, + "loss": 0.7904, + "step": 48665 + }, + { + "epoch": 3.13, + "learning_rate": 9.578565447948647e-06, + "loss": 0.8186, + "step": 48666 + }, + { + "epoch": 3.13, + "learning_rate": 9.578350758925698e-06, + "loss": 0.8642, + "step": 48667 + }, + { + "epoch": 3.13, + "learning_rate": 9.578136069902747e-06, + "loss": 0.7918, + "step": 48668 + }, + { + "epoch": 3.13, + "learning_rate": 9.577921380879797e-06, + "loss": 0.8148, + "step": 48669 + }, + { + "epoch": 3.13, + "learning_rate": 9.577706691856846e-06, + "loss": 0.8008, + "step": 48670 + }, + { + "epoch": 3.13, + "learning_rate": 9.577492002833895e-06, + "loss": 0.8004, + "step": 48671 + }, + { + "epoch": 3.13, + "learning_rate": 9.577277313810946e-06, + "loss": 0.8578, + "step": 48672 + }, + { + "epoch": 3.13, + "learning_rate": 9.577062624787995e-06, + "loss": 0.843, + "step": 48673 + }, + { + "epoch": 3.13, + "learning_rate": 9.576847935765044e-06, + "loss": 0.831, + "step": 48674 + }, + { + "epoch": 3.13, + "learning_rate": 9.576633246742094e-06, + "loss": 0.792, + "step": 48675 + }, + { + "epoch": 3.13, + "learning_rate": 9.576418557719145e-06, + "loss": 0.8116, + "step": 48676 + }, + { + "epoch": 3.13, + "learning_rate": 9.576203868696194e-06, + "loss": 0.746, + "step": 48677 + }, + { + "epoch": 3.13, + "learning_rate": 9.575989179673245e-06, + "loss": 0.8543, + "step": 48678 + }, + { + "epoch": 3.13, + "learning_rate": 9.575774490650294e-06, + "loss": 0.8704, + "step": 48679 + }, + { + "epoch": 3.13, + "learning_rate": 9.575559801627343e-06, + "loss": 0.8924, + "step": 48680 + }, + { + "epoch": 3.13, + "learning_rate": 9.575345112604393e-06, + "loss": 0.8589, + "step": 48681 + }, + { + "epoch": 3.13, + "learning_rate": 9.575130423581444e-06, + "loss": 0.7799, + "step": 48682 + }, + { + "epoch": 3.13, + "learning_rate": 9.574915734558493e-06, + "loss": 0.7806, + "step": 48683 + }, + { + "epoch": 3.13, + "learning_rate": 9.574701045535542e-06, + "loss": 0.8233, + "step": 48684 + }, + { + "epoch": 3.13, + "learning_rate": 9.574486356512593e-06, + "loss": 0.805, + "step": 48685 + }, + { + "epoch": 3.13, + "learning_rate": 9.574271667489642e-06, + "loss": 0.8917, + "step": 48686 + }, + { + "epoch": 3.13, + "learning_rate": 9.574056978466692e-06, + "loss": 0.799, + "step": 48687 + }, + { + "epoch": 3.13, + "learning_rate": 9.573842289443741e-06, + "loss": 0.8248, + "step": 48688 + }, + { + "epoch": 3.13, + "learning_rate": 9.573627600420792e-06, + "loss": 0.8392, + "step": 48689 + }, + { + "epoch": 3.13, + "learning_rate": 9.57341291139784e-06, + "loss": 0.8528, + "step": 48690 + }, + { + "epoch": 3.13, + "learning_rate": 9.573198222374891e-06, + "loss": 0.8586, + "step": 48691 + }, + { + "epoch": 3.13, + "learning_rate": 9.57298353335194e-06, + "loss": 0.8275, + "step": 48692 + }, + { + "epoch": 3.13, + "learning_rate": 9.57276884432899e-06, + "loss": 0.8521, + "step": 48693 + }, + { + "epoch": 3.13, + "learning_rate": 9.57255415530604e-06, + "loss": 0.7781, + "step": 48694 + }, + { + "epoch": 3.13, + "learning_rate": 9.572339466283089e-06, + "loss": 0.8036, + "step": 48695 + }, + { + "epoch": 3.13, + "learning_rate": 9.57212477726014e-06, + "loss": 0.8281, + "step": 48696 + }, + { + "epoch": 3.13, + "learning_rate": 9.57191008823719e-06, + "loss": 0.8508, + "step": 48697 + }, + { + "epoch": 3.13, + "learning_rate": 9.57169539921424e-06, + "loss": 0.7506, + "step": 48698 + }, + { + "epoch": 3.13, + "learning_rate": 9.571480710191288e-06, + "loss": 0.865, + "step": 48699 + }, + { + "epoch": 3.13, + "learning_rate": 9.571266021168339e-06, + "loss": 0.7888, + "step": 48700 + }, + { + "epoch": 3.13, + "learning_rate": 9.571051332145388e-06, + "loss": 0.8622, + "step": 48701 + }, + { + "epoch": 3.13, + "learning_rate": 9.570836643122437e-06, + "loss": 0.849, + "step": 48702 + }, + { + "epoch": 3.13, + "learning_rate": 9.570621954099488e-06, + "loss": 0.8737, + "step": 48703 + }, + { + "epoch": 3.13, + "learning_rate": 9.570407265076538e-06, + "loss": 0.7829, + "step": 48704 + }, + { + "epoch": 3.13, + "learning_rate": 9.570192576053587e-06, + "loss": 0.7922, + "step": 48705 + }, + { + "epoch": 3.13, + "learning_rate": 9.569977887030638e-06, + "loss": 0.8491, + "step": 48706 + }, + { + "epoch": 3.13, + "learning_rate": 9.569763198007687e-06, + "loss": 0.8616, + "step": 48707 + }, + { + "epoch": 3.13, + "learning_rate": 9.569548508984736e-06, + "loss": 0.8229, + "step": 48708 + }, + { + "epoch": 3.13, + "learning_rate": 9.569333819961786e-06, + "loss": 0.7982, + "step": 48709 + }, + { + "epoch": 3.13, + "learning_rate": 9.569119130938835e-06, + "loss": 0.9305, + "step": 48710 + }, + { + "epoch": 3.13, + "learning_rate": 9.568904441915886e-06, + "loss": 0.8314, + "step": 48711 + }, + { + "epoch": 3.13, + "learning_rate": 9.568689752892935e-06, + "loss": 0.8271, + "step": 48712 + }, + { + "epoch": 3.13, + "learning_rate": 9.568475063869986e-06, + "loss": 0.8239, + "step": 48713 + }, + { + "epoch": 3.13, + "learning_rate": 9.568260374847035e-06, + "loss": 0.88, + "step": 48714 + }, + { + "epoch": 3.13, + "learning_rate": 9.568045685824085e-06, + "loss": 0.8238, + "step": 48715 + }, + { + "epoch": 3.13, + "learning_rate": 9.567830996801134e-06, + "loss": 0.8105, + "step": 48716 + }, + { + "epoch": 3.13, + "learning_rate": 9.567616307778183e-06, + "loss": 0.8446, + "step": 48717 + }, + { + "epoch": 3.13, + "learning_rate": 9.567401618755234e-06, + "loss": 0.7905, + "step": 48718 + }, + { + "epoch": 3.13, + "learning_rate": 9.567186929732285e-06, + "loss": 0.8472, + "step": 48719 + }, + { + "epoch": 3.13, + "learning_rate": 9.566972240709333e-06, + "loss": 0.7942, + "step": 48720 + }, + { + "epoch": 3.13, + "learning_rate": 9.566757551686382e-06, + "loss": 0.7649, + "step": 48721 + }, + { + "epoch": 3.13, + "learning_rate": 9.566542862663433e-06, + "loss": 0.7958, + "step": 48722 + }, + { + "epoch": 3.13, + "learning_rate": 9.566328173640482e-06, + "loss": 0.8278, + "step": 48723 + }, + { + "epoch": 3.13, + "learning_rate": 9.566113484617531e-06, + "loss": 0.8233, + "step": 48724 + }, + { + "epoch": 3.13, + "learning_rate": 9.565898795594583e-06, + "loss": 0.7979, + "step": 48725 + }, + { + "epoch": 3.13, + "learning_rate": 9.565684106571632e-06, + "loss": 0.8415, + "step": 48726 + }, + { + "epoch": 3.13, + "learning_rate": 9.565469417548681e-06, + "loss": 0.8393, + "step": 48727 + }, + { + "epoch": 3.14, + "learning_rate": 9.565254728525732e-06, + "loss": 0.7804, + "step": 48728 + }, + { + "epoch": 3.14, + "learning_rate": 9.565040039502781e-06, + "loss": 0.8152, + "step": 48729 + }, + { + "epoch": 3.14, + "learning_rate": 9.56482535047983e-06, + "loss": 0.7938, + "step": 48730 + }, + { + "epoch": 3.14, + "learning_rate": 9.56461066145688e-06, + "loss": 0.8566, + "step": 48731 + }, + { + "epoch": 3.14, + "learning_rate": 9.56439597243393e-06, + "loss": 0.7805, + "step": 48732 + }, + { + "epoch": 3.14, + "learning_rate": 9.56418128341098e-06, + "loss": 0.7828, + "step": 48733 + }, + { + "epoch": 3.14, + "learning_rate": 9.56396659438803e-06, + "loss": 0.8343, + "step": 48734 + }, + { + "epoch": 3.14, + "learning_rate": 9.56375190536508e-06, + "loss": 0.8126, + "step": 48735 + }, + { + "epoch": 3.14, + "learning_rate": 9.563537216342129e-06, + "loss": 0.8159, + "step": 48736 + }, + { + "epoch": 3.14, + "learning_rate": 9.56332252731918e-06, + "loss": 0.8538, + "step": 48737 + }, + { + "epoch": 3.14, + "learning_rate": 9.563107838296228e-06, + "loss": 0.7986, + "step": 48738 + }, + { + "epoch": 3.14, + "learning_rate": 9.562893149273277e-06, + "loss": 0.86, + "step": 48739 + }, + { + "epoch": 3.14, + "learning_rate": 9.562678460250328e-06, + "loss": 0.8644, + "step": 48740 + }, + { + "epoch": 3.14, + "learning_rate": 9.562463771227379e-06, + "loss": 0.8296, + "step": 48741 + }, + { + "epoch": 3.14, + "learning_rate": 9.562249082204428e-06, + "loss": 0.8192, + "step": 48742 + }, + { + "epoch": 3.14, + "learning_rate": 9.562034393181477e-06, + "loss": 0.8136, + "step": 48743 + }, + { + "epoch": 3.14, + "learning_rate": 9.561819704158527e-06, + "loss": 0.8113, + "step": 48744 + }, + { + "epoch": 3.14, + "learning_rate": 9.561605015135576e-06, + "loss": 0.8145, + "step": 48745 + }, + { + "epoch": 3.14, + "learning_rate": 9.561390326112627e-06, + "loss": 0.7822, + "step": 48746 + }, + { + "epoch": 3.14, + "learning_rate": 9.561175637089678e-06, + "loss": 0.7946, + "step": 48747 + }, + { + "epoch": 3.14, + "learning_rate": 9.560960948066727e-06, + "loss": 0.8509, + "step": 48748 + }, + { + "epoch": 3.14, + "learning_rate": 9.560746259043775e-06, + "loss": 0.8565, + "step": 48749 + }, + { + "epoch": 3.14, + "learning_rate": 9.560531570020826e-06, + "loss": 0.8803, + "step": 48750 + }, + { + "epoch": 3.14, + "learning_rate": 9.560316880997875e-06, + "loss": 0.8096, + "step": 48751 + }, + { + "epoch": 3.14, + "learning_rate": 9.560102191974924e-06, + "loss": 0.7544, + "step": 48752 + }, + { + "epoch": 3.14, + "learning_rate": 9.559887502951975e-06, + "loss": 0.8124, + "step": 48753 + }, + { + "epoch": 3.14, + "learning_rate": 9.559672813929025e-06, + "loss": 0.8625, + "step": 48754 + }, + { + "epoch": 3.14, + "learning_rate": 9.559458124906074e-06, + "loss": 0.8358, + "step": 48755 + }, + { + "epoch": 3.14, + "learning_rate": 9.559243435883125e-06, + "loss": 0.8212, + "step": 48756 + }, + { + "epoch": 3.14, + "learning_rate": 9.559028746860174e-06, + "loss": 0.7921, + "step": 48757 + }, + { + "epoch": 3.14, + "learning_rate": 9.558814057837223e-06, + "loss": 0.836, + "step": 48758 + }, + { + "epoch": 3.14, + "learning_rate": 9.558599368814274e-06, + "loss": 0.7939, + "step": 48759 + }, + { + "epoch": 3.14, + "learning_rate": 9.558384679791323e-06, + "loss": 0.8239, + "step": 48760 + }, + { + "epoch": 3.14, + "learning_rate": 9.558169990768372e-06, + "loss": 0.843, + "step": 48761 + }, + { + "epoch": 3.14, + "learning_rate": 9.557955301745422e-06, + "loss": 0.8067, + "step": 48762 + }, + { + "epoch": 3.14, + "learning_rate": 9.557740612722473e-06, + "loss": 0.8308, + "step": 48763 + }, + { + "epoch": 3.14, + "learning_rate": 9.557525923699522e-06, + "loss": 0.8193, + "step": 48764 + }, + { + "epoch": 3.14, + "learning_rate": 9.557311234676572e-06, + "loss": 0.8463, + "step": 48765 + }, + { + "epoch": 3.14, + "learning_rate": 9.557096545653621e-06, + "loss": 0.8169, + "step": 48766 + }, + { + "epoch": 3.14, + "learning_rate": 9.55688185663067e-06, + "loss": 0.7854, + "step": 48767 + }, + { + "epoch": 3.14, + "learning_rate": 9.556667167607721e-06, + "loss": 0.8083, + "step": 48768 + }, + { + "epoch": 3.14, + "learning_rate": 9.556452478584772e-06, + "loss": 0.7654, + "step": 48769 + }, + { + "epoch": 3.14, + "learning_rate": 9.55623778956182e-06, + "loss": 0.8298, + "step": 48770 + }, + { + "epoch": 3.14, + "learning_rate": 9.55602310053887e-06, + "loss": 0.8223, + "step": 48771 + }, + { + "epoch": 3.14, + "learning_rate": 9.55580841151592e-06, + "loss": 0.8577, + "step": 48772 + }, + { + "epoch": 3.14, + "learning_rate": 9.55559372249297e-06, + "loss": 0.8811, + "step": 48773 + }, + { + "epoch": 3.14, + "learning_rate": 9.55537903347002e-06, + "loss": 0.7845, + "step": 48774 + }, + { + "epoch": 3.14, + "learning_rate": 9.555164344447069e-06, + "loss": 0.8643, + "step": 48775 + }, + { + "epoch": 3.14, + "learning_rate": 9.55494965542412e-06, + "loss": 0.8679, + "step": 48776 + }, + { + "epoch": 3.14, + "learning_rate": 9.554734966401169e-06, + "loss": 0.8769, + "step": 48777 + }, + { + "epoch": 3.14, + "learning_rate": 9.55452027737822e-06, + "loss": 0.8404, + "step": 48778 + }, + { + "epoch": 3.14, + "learning_rate": 9.554305588355268e-06, + "loss": 0.8367, + "step": 48779 + }, + { + "epoch": 3.14, + "learning_rate": 9.554090899332317e-06, + "loss": 0.7843, + "step": 48780 + }, + { + "epoch": 3.14, + "learning_rate": 9.553876210309368e-06, + "loss": 0.8846, + "step": 48781 + }, + { + "epoch": 3.14, + "learning_rate": 9.553661521286417e-06, + "loss": 0.8347, + "step": 48782 + }, + { + "epoch": 3.14, + "learning_rate": 9.553446832263467e-06, + "loss": 0.8198, + "step": 48783 + }, + { + "epoch": 3.14, + "learning_rate": 9.553232143240518e-06, + "loss": 0.7789, + "step": 48784 + }, + { + "epoch": 3.14, + "learning_rate": 9.553017454217567e-06, + "loss": 0.8933, + "step": 48785 + }, + { + "epoch": 3.14, + "learning_rate": 9.552802765194616e-06, + "loss": 0.8327, + "step": 48786 + }, + { + "epoch": 3.14, + "learning_rate": 9.552588076171667e-06, + "loss": 0.8059, + "step": 48787 + }, + { + "epoch": 3.14, + "learning_rate": 9.552373387148716e-06, + "loss": 0.8183, + "step": 48788 + }, + { + "epoch": 3.14, + "learning_rate": 9.552158698125765e-06, + "loss": 0.8405, + "step": 48789 + }, + { + "epoch": 3.14, + "learning_rate": 9.551944009102815e-06, + "loss": 0.8194, + "step": 48790 + }, + { + "epoch": 3.14, + "learning_rate": 9.551729320079866e-06, + "loss": 0.8684, + "step": 48791 + }, + { + "epoch": 3.14, + "learning_rate": 9.551514631056915e-06, + "loss": 0.7992, + "step": 48792 + }, + { + "epoch": 3.14, + "learning_rate": 9.551299942033966e-06, + "loss": 0.792, + "step": 48793 + }, + { + "epoch": 3.14, + "learning_rate": 9.551085253011014e-06, + "loss": 0.8292, + "step": 48794 + }, + { + "epoch": 3.14, + "learning_rate": 9.550870563988063e-06, + "loss": 0.8906, + "step": 48795 + }, + { + "epoch": 3.14, + "learning_rate": 9.550655874965114e-06, + "loss": 0.8319, + "step": 48796 + }, + { + "epoch": 3.14, + "learning_rate": 9.550441185942163e-06, + "loss": 0.7857, + "step": 48797 + }, + { + "epoch": 3.14, + "learning_rate": 9.550226496919214e-06, + "loss": 0.7964, + "step": 48798 + }, + { + "epoch": 3.14, + "learning_rate": 9.550011807896263e-06, + "loss": 0.8281, + "step": 48799 + }, + { + "epoch": 3.14, + "learning_rate": 9.549797118873313e-06, + "loss": 0.7622, + "step": 48800 + }, + { + "epoch": 3.14, + "learning_rate": 9.549582429850362e-06, + "loss": 0.7941, + "step": 48801 + }, + { + "epoch": 3.14, + "learning_rate": 9.549367740827411e-06, + "loss": 0.8413, + "step": 48802 + }, + { + "epoch": 3.14, + "learning_rate": 9.549153051804462e-06, + "loss": 0.7892, + "step": 48803 + }, + { + "epoch": 3.14, + "learning_rate": 9.548938362781511e-06, + "loss": 0.8308, + "step": 48804 + }, + { + "epoch": 3.14, + "learning_rate": 9.548723673758562e-06, + "loss": 0.8523, + "step": 48805 + }, + { + "epoch": 3.14, + "learning_rate": 9.548508984735612e-06, + "loss": 0.783, + "step": 48806 + }, + { + "epoch": 3.14, + "learning_rate": 9.548294295712661e-06, + "loss": 0.812, + "step": 48807 + }, + { + "epoch": 3.14, + "learning_rate": 9.54807960668971e-06, + "loss": 0.8257, + "step": 48808 + }, + { + "epoch": 3.14, + "learning_rate": 9.54786491766676e-06, + "loss": 0.8576, + "step": 48809 + }, + { + "epoch": 3.14, + "learning_rate": 9.54765022864381e-06, + "loss": 0.8641, + "step": 48810 + }, + { + "epoch": 3.14, + "learning_rate": 9.547435539620859e-06, + "loss": 0.7944, + "step": 48811 + }, + { + "epoch": 3.14, + "learning_rate": 9.54722085059791e-06, + "loss": 0.8192, + "step": 48812 + }, + { + "epoch": 3.14, + "learning_rate": 9.54700616157496e-06, + "loss": 0.843, + "step": 48813 + }, + { + "epoch": 3.14, + "learning_rate": 9.546791472552009e-06, + "loss": 0.7494, + "step": 48814 + }, + { + "epoch": 3.14, + "learning_rate": 9.54657678352906e-06, + "loss": 0.8878, + "step": 48815 + }, + { + "epoch": 3.14, + "learning_rate": 9.546362094506109e-06, + "loss": 0.8349, + "step": 48816 + }, + { + "epoch": 3.14, + "learning_rate": 9.546147405483158e-06, + "loss": 0.8048, + "step": 48817 + }, + { + "epoch": 3.14, + "learning_rate": 9.545932716460208e-06, + "loss": 0.8071, + "step": 48818 + }, + { + "epoch": 3.14, + "learning_rate": 9.545718027437257e-06, + "loss": 0.8015, + "step": 48819 + }, + { + "epoch": 3.14, + "learning_rate": 9.545503338414308e-06, + "loss": 0.8285, + "step": 48820 + }, + { + "epoch": 3.14, + "learning_rate": 9.545288649391357e-06, + "loss": 0.8322, + "step": 48821 + }, + { + "epoch": 3.14, + "learning_rate": 9.545073960368408e-06, + "loss": 0.8336, + "step": 48822 + }, + { + "epoch": 3.14, + "learning_rate": 9.544859271345457e-06, + "loss": 0.8561, + "step": 48823 + }, + { + "epoch": 3.14, + "learning_rate": 9.544644582322507e-06, + "loss": 0.8438, + "step": 48824 + }, + { + "epoch": 3.14, + "learning_rate": 9.544429893299556e-06, + "loss": 0.829, + "step": 48825 + }, + { + "epoch": 3.14, + "learning_rate": 9.544215204276605e-06, + "loss": 0.879, + "step": 48826 + }, + { + "epoch": 3.14, + "learning_rate": 9.544000515253656e-06, + "loss": 0.7965, + "step": 48827 + }, + { + "epoch": 3.14, + "learning_rate": 9.543785826230706e-06, + "loss": 0.7948, + "step": 48828 + }, + { + "epoch": 3.14, + "learning_rate": 9.543571137207755e-06, + "loss": 0.8142, + "step": 48829 + }, + { + "epoch": 3.14, + "learning_rate": 9.543356448184804e-06, + "loss": 0.8376, + "step": 48830 + }, + { + "epoch": 3.14, + "learning_rate": 9.543141759161855e-06, + "loss": 0.8425, + "step": 48831 + }, + { + "epoch": 3.14, + "learning_rate": 9.542927070138904e-06, + "loss": 0.8118, + "step": 48832 + }, + { + "epoch": 3.14, + "learning_rate": 9.542712381115955e-06, + "loss": 0.8844, + "step": 48833 + }, + { + "epoch": 3.14, + "learning_rate": 9.542497692093005e-06, + "loss": 0.9075, + "step": 48834 + }, + { + "epoch": 3.14, + "learning_rate": 9.542283003070054e-06, + "loss": 0.8246, + "step": 48835 + }, + { + "epoch": 3.14, + "learning_rate": 9.542068314047103e-06, + "loss": 0.7959, + "step": 48836 + }, + { + "epoch": 3.14, + "learning_rate": 9.541853625024154e-06, + "loss": 0.8084, + "step": 48837 + }, + { + "epoch": 3.14, + "learning_rate": 9.541638936001203e-06, + "loss": 0.8325, + "step": 48838 + }, + { + "epoch": 3.14, + "learning_rate": 9.541424246978252e-06, + "loss": 0.7975, + "step": 48839 + }, + { + "epoch": 3.14, + "learning_rate": 9.541209557955302e-06, + "loss": 0.7914, + "step": 48840 + }, + { + "epoch": 3.14, + "learning_rate": 9.540994868932351e-06, + "loss": 0.831, + "step": 48841 + }, + { + "epoch": 3.14, + "learning_rate": 9.540780179909402e-06, + "loss": 0.7975, + "step": 48842 + }, + { + "epoch": 3.14, + "learning_rate": 9.540565490886453e-06, + "loss": 0.828, + "step": 48843 + }, + { + "epoch": 3.14, + "learning_rate": 9.540350801863502e-06, + "loss": 0.822, + "step": 48844 + }, + { + "epoch": 3.14, + "learning_rate": 9.54013611284055e-06, + "loss": 0.8215, + "step": 48845 + }, + { + "epoch": 3.14, + "learning_rate": 9.539921423817601e-06, + "loss": 0.8192, + "step": 48846 + }, + { + "epoch": 3.14, + "learning_rate": 9.53970673479465e-06, + "loss": 0.8084, + "step": 48847 + }, + { + "epoch": 3.14, + "learning_rate": 9.5394920457717e-06, + "loss": 0.8285, + "step": 48848 + }, + { + "epoch": 3.14, + "learning_rate": 9.53927735674875e-06, + "loss": 0.8306, + "step": 48849 + }, + { + "epoch": 3.14, + "learning_rate": 9.5390626677258e-06, + "loss": 0.8251, + "step": 48850 + }, + { + "epoch": 3.14, + "learning_rate": 9.53884797870285e-06, + "loss": 0.8113, + "step": 48851 + }, + { + "epoch": 3.14, + "learning_rate": 9.5386332896799e-06, + "loss": 0.8553, + "step": 48852 + }, + { + "epoch": 3.14, + "learning_rate": 9.53841860065695e-06, + "loss": 0.8377, + "step": 48853 + }, + { + "epoch": 3.14, + "learning_rate": 9.538203911633998e-06, + "loss": 0.8344, + "step": 48854 + }, + { + "epoch": 3.14, + "learning_rate": 9.537989222611049e-06, + "loss": 0.8006, + "step": 48855 + }, + { + "epoch": 3.14, + "learning_rate": 9.5377745335881e-06, + "loss": 0.8485, + "step": 48856 + }, + { + "epoch": 3.14, + "learning_rate": 9.537559844565148e-06, + "loss": 0.7908, + "step": 48857 + }, + { + "epoch": 3.14, + "learning_rate": 9.537345155542197e-06, + "loss": 0.8521, + "step": 48858 + }, + { + "epoch": 3.14, + "learning_rate": 9.537130466519248e-06, + "loss": 0.837, + "step": 48859 + }, + { + "epoch": 3.14, + "learning_rate": 9.536915777496297e-06, + "loss": 0.8211, + "step": 48860 + }, + { + "epoch": 3.14, + "learning_rate": 9.536701088473348e-06, + "loss": 0.7877, + "step": 48861 + }, + { + "epoch": 3.14, + "learning_rate": 9.536486399450397e-06, + "loss": 0.8232, + "step": 48862 + }, + { + "epoch": 3.14, + "learning_rate": 9.536271710427447e-06, + "loss": 0.8389, + "step": 48863 + }, + { + "epoch": 3.14, + "learning_rate": 9.536057021404496e-06, + "loss": 0.773, + "step": 48864 + }, + { + "epoch": 3.14, + "learning_rate": 9.535842332381547e-06, + "loss": 0.868, + "step": 48865 + }, + { + "epoch": 3.14, + "learning_rate": 9.535627643358596e-06, + "loss": 0.8894, + "step": 48866 + }, + { + "epoch": 3.14, + "learning_rate": 9.535412954335645e-06, + "loss": 0.8407, + "step": 48867 + }, + { + "epoch": 3.14, + "learning_rate": 9.535198265312695e-06, + "loss": 0.7774, + "step": 48868 + }, + { + "epoch": 3.14, + "learning_rate": 9.534983576289744e-06, + "loss": 0.7935, + "step": 48869 + }, + { + "epoch": 3.14, + "learning_rate": 9.534768887266795e-06, + "loss": 0.7696, + "step": 48870 + }, + { + "epoch": 3.14, + "learning_rate": 9.534554198243846e-06, + "loss": 0.8487, + "step": 48871 + }, + { + "epoch": 3.14, + "learning_rate": 9.534339509220895e-06, + "loss": 0.8277, + "step": 48872 + }, + { + "epoch": 3.14, + "learning_rate": 9.534124820197944e-06, + "loss": 0.7876, + "step": 48873 + }, + { + "epoch": 3.14, + "learning_rate": 9.533910131174994e-06, + "loss": 0.8224, + "step": 48874 + }, + { + "epoch": 3.14, + "learning_rate": 9.533695442152043e-06, + "loss": 0.868, + "step": 48875 + }, + { + "epoch": 3.14, + "learning_rate": 9.533480753129092e-06, + "loss": 0.8373, + "step": 48876 + }, + { + "epoch": 3.14, + "learning_rate": 9.533266064106143e-06, + "loss": 0.7687, + "step": 48877 + }, + { + "epoch": 3.14, + "learning_rate": 9.533051375083194e-06, + "loss": 0.8383, + "step": 48878 + }, + { + "epoch": 3.14, + "learning_rate": 9.532836686060243e-06, + "loss": 0.77, + "step": 48879 + }, + { + "epoch": 3.14, + "learning_rate": 9.532621997037293e-06, + "loss": 0.7925, + "step": 48880 + }, + { + "epoch": 3.14, + "learning_rate": 9.532407308014342e-06, + "loss": 0.8021, + "step": 48881 + }, + { + "epoch": 3.14, + "learning_rate": 9.532192618991391e-06, + "loss": 0.7983, + "step": 48882 + }, + { + "epoch": 3.15, + "learning_rate": 9.531977929968442e-06, + "loss": 0.83, + "step": 48883 + }, + { + "epoch": 3.15, + "learning_rate": 9.53176324094549e-06, + "loss": 0.8285, + "step": 48884 + }, + { + "epoch": 3.15, + "learning_rate": 9.531548551922541e-06, + "loss": 0.8228, + "step": 48885 + }, + { + "epoch": 3.15, + "learning_rate": 9.53133386289959e-06, + "loss": 0.8459, + "step": 48886 + }, + { + "epoch": 3.15, + "learning_rate": 9.531119173876641e-06, + "loss": 0.7875, + "step": 48887 + }, + { + "epoch": 3.15, + "learning_rate": 9.53090448485369e-06, + "loss": 0.7762, + "step": 48888 + }, + { + "epoch": 3.15, + "learning_rate": 9.530689795830739e-06, + "loss": 0.8119, + "step": 48889 + }, + { + "epoch": 3.15, + "learning_rate": 9.53047510680779e-06, + "loss": 0.8291, + "step": 48890 + }, + { + "epoch": 3.15, + "learning_rate": 9.530260417784839e-06, + "loss": 0.8406, + "step": 48891 + }, + { + "epoch": 3.15, + "learning_rate": 9.53004572876189e-06, + "loss": 0.8547, + "step": 48892 + }, + { + "epoch": 3.15, + "learning_rate": 9.52983103973894e-06, + "loss": 0.7528, + "step": 48893 + }, + { + "epoch": 3.15, + "learning_rate": 9.529616350715989e-06, + "loss": 0.8089, + "step": 48894 + }, + { + "epoch": 3.15, + "learning_rate": 9.529401661693038e-06, + "loss": 0.8003, + "step": 48895 + }, + { + "epoch": 3.15, + "learning_rate": 9.529186972670089e-06, + "loss": 0.7728, + "step": 48896 + }, + { + "epoch": 3.15, + "learning_rate": 9.528972283647138e-06, + "loss": 0.9104, + "step": 48897 + }, + { + "epoch": 3.15, + "learning_rate": 9.528757594624186e-06, + "loss": 0.8461, + "step": 48898 + }, + { + "epoch": 3.15, + "learning_rate": 9.528542905601237e-06, + "loss": 0.8349, + "step": 48899 + }, + { + "epoch": 3.15, + "learning_rate": 9.528328216578288e-06, + "loss": 0.83, + "step": 48900 + }, + { + "epoch": 3.15, + "learning_rate": 9.528113527555337e-06, + "loss": 0.802, + "step": 48901 + }, + { + "epoch": 3.15, + "learning_rate": 9.527898838532387e-06, + "loss": 0.8374, + "step": 48902 + }, + { + "epoch": 3.15, + "learning_rate": 9.527684149509436e-06, + "loss": 0.7956, + "step": 48903 + }, + { + "epoch": 3.15, + "learning_rate": 9.527469460486485e-06, + "loss": 0.8066, + "step": 48904 + }, + { + "epoch": 3.15, + "learning_rate": 9.527254771463536e-06, + "loss": 0.8947, + "step": 48905 + }, + { + "epoch": 3.15, + "learning_rate": 9.527040082440585e-06, + "loss": 0.7859, + "step": 48906 + }, + { + "epoch": 3.15, + "learning_rate": 9.526825393417636e-06, + "loss": 0.7673, + "step": 48907 + }, + { + "epoch": 3.15, + "learning_rate": 9.526610704394685e-06, + "loss": 0.8343, + "step": 48908 + }, + { + "epoch": 3.15, + "learning_rate": 9.526396015371735e-06, + "loss": 0.8473, + "step": 48909 + }, + { + "epoch": 3.15, + "learning_rate": 9.526181326348784e-06, + "loss": 0.8456, + "step": 48910 + }, + { + "epoch": 3.15, + "learning_rate": 9.525966637325835e-06, + "loss": 0.8359, + "step": 48911 + }, + { + "epoch": 3.15, + "learning_rate": 9.525751948302884e-06, + "loss": 0.8629, + "step": 48912 + }, + { + "epoch": 3.15, + "learning_rate": 9.525537259279933e-06, + "loss": 0.8042, + "step": 48913 + }, + { + "epoch": 3.15, + "learning_rate": 9.525322570256983e-06, + "loss": 0.8688, + "step": 48914 + }, + { + "epoch": 3.15, + "learning_rate": 9.525107881234034e-06, + "loss": 0.8599, + "step": 48915 + }, + { + "epoch": 3.15, + "learning_rate": 9.524893192211083e-06, + "loss": 0.8236, + "step": 48916 + }, + { + "epoch": 3.15, + "learning_rate": 9.524678503188132e-06, + "loss": 0.8128, + "step": 48917 + }, + { + "epoch": 3.15, + "learning_rate": 9.524463814165183e-06, + "loss": 0.8342, + "step": 48918 + }, + { + "epoch": 3.15, + "learning_rate": 9.524249125142232e-06, + "loss": 0.8145, + "step": 48919 + }, + { + "epoch": 3.15, + "learning_rate": 9.524034436119282e-06, + "loss": 0.8122, + "step": 48920 + }, + { + "epoch": 3.15, + "learning_rate": 9.523819747096333e-06, + "loss": 0.8779, + "step": 48921 + }, + { + "epoch": 3.15, + "learning_rate": 9.523605058073382e-06, + "loss": 0.848, + "step": 48922 + }, + { + "epoch": 3.15, + "learning_rate": 9.523390369050431e-06, + "loss": 0.7831, + "step": 48923 + }, + { + "epoch": 3.15, + "learning_rate": 9.523175680027482e-06, + "loss": 0.8205, + "step": 48924 + }, + { + "epoch": 3.15, + "learning_rate": 9.52296099100453e-06, + "loss": 0.8966, + "step": 48925 + }, + { + "epoch": 3.15, + "learning_rate": 9.52274630198158e-06, + "loss": 0.7522, + "step": 48926 + }, + { + "epoch": 3.15, + "learning_rate": 9.52253161295863e-06, + "loss": 0.7571, + "step": 48927 + }, + { + "epoch": 3.15, + "learning_rate": 9.522316923935679e-06, + "loss": 0.8093, + "step": 48928 + }, + { + "epoch": 3.15, + "learning_rate": 9.52210223491273e-06, + "loss": 0.838, + "step": 48929 + }, + { + "epoch": 3.15, + "learning_rate": 9.52188754588978e-06, + "loss": 0.8078, + "step": 48930 + }, + { + "epoch": 3.15, + "learning_rate": 9.52167285686683e-06, + "loss": 0.7931, + "step": 48931 + }, + { + "epoch": 3.15, + "learning_rate": 9.521458167843878e-06, + "loss": 0.8113, + "step": 48932 + }, + { + "epoch": 3.15, + "learning_rate": 9.521243478820929e-06, + "loss": 0.8105, + "step": 48933 + }, + { + "epoch": 3.15, + "learning_rate": 9.521028789797978e-06, + "loss": 0.7877, + "step": 48934 + }, + { + "epoch": 3.15, + "learning_rate": 9.520814100775027e-06, + "loss": 0.8379, + "step": 48935 + }, + { + "epoch": 3.15, + "learning_rate": 9.520599411752078e-06, + "loss": 0.7688, + "step": 48936 + }, + { + "epoch": 3.15, + "learning_rate": 9.520384722729128e-06, + "loss": 0.828, + "step": 48937 + }, + { + "epoch": 3.15, + "learning_rate": 9.520170033706177e-06, + "loss": 0.7904, + "step": 48938 + }, + { + "epoch": 3.15, + "learning_rate": 9.519955344683228e-06, + "loss": 0.8091, + "step": 48939 + }, + { + "epoch": 3.15, + "learning_rate": 9.519740655660277e-06, + "loss": 0.8263, + "step": 48940 + }, + { + "epoch": 3.15, + "learning_rate": 9.519525966637326e-06, + "loss": 0.8095, + "step": 48941 + }, + { + "epoch": 3.15, + "learning_rate": 9.519311277614377e-06, + "loss": 0.8186, + "step": 48942 + }, + { + "epoch": 3.15, + "learning_rate": 9.519096588591427e-06, + "loss": 0.8118, + "step": 48943 + }, + { + "epoch": 3.15, + "learning_rate": 9.518881899568476e-06, + "loss": 0.8697, + "step": 48944 + }, + { + "epoch": 3.15, + "learning_rate": 9.518667210545525e-06, + "loss": 0.7971, + "step": 48945 + }, + { + "epoch": 3.15, + "learning_rate": 9.518452521522576e-06, + "loss": 0.7286, + "step": 48946 + }, + { + "epoch": 3.15, + "learning_rate": 9.518237832499625e-06, + "loss": 0.819, + "step": 48947 + }, + { + "epoch": 3.15, + "learning_rate": 9.518023143476675e-06, + "loss": 0.8391, + "step": 48948 + }, + { + "epoch": 3.15, + "learning_rate": 9.517808454453724e-06, + "loss": 0.8402, + "step": 48949 + }, + { + "epoch": 3.15, + "learning_rate": 9.517593765430775e-06, + "loss": 0.8114, + "step": 48950 + }, + { + "epoch": 3.15, + "learning_rate": 9.517379076407824e-06, + "loss": 0.8902, + "step": 48951 + }, + { + "epoch": 3.15, + "learning_rate": 9.517164387384875e-06, + "loss": 0.8423, + "step": 48952 + }, + { + "epoch": 3.15, + "learning_rate": 9.516949698361924e-06, + "loss": 0.7964, + "step": 48953 + }, + { + "epoch": 3.15, + "learning_rate": 9.516735009338973e-06, + "loss": 0.8437, + "step": 48954 + }, + { + "epoch": 3.15, + "learning_rate": 9.516520320316023e-06, + "loss": 0.8132, + "step": 48955 + }, + { + "epoch": 3.15, + "learning_rate": 9.516305631293072e-06, + "loss": 0.767, + "step": 48956 + }, + { + "epoch": 3.15, + "learning_rate": 9.516090942270121e-06, + "loss": 0.7906, + "step": 48957 + }, + { + "epoch": 3.15, + "learning_rate": 9.515876253247173e-06, + "loss": 0.8187, + "step": 48958 + }, + { + "epoch": 3.15, + "learning_rate": 9.515661564224222e-06, + "loss": 0.8234, + "step": 48959 + }, + { + "epoch": 3.15, + "learning_rate": 9.515446875201271e-06, + "loss": 0.8621, + "step": 48960 + }, + { + "epoch": 3.15, + "learning_rate": 9.515232186178322e-06, + "loss": 0.8449, + "step": 48961 + }, + { + "epoch": 3.15, + "learning_rate": 9.515017497155371e-06, + "loss": 0.8272, + "step": 48962 + }, + { + "epoch": 3.15, + "learning_rate": 9.51480280813242e-06, + "loss": 0.8052, + "step": 48963 + }, + { + "epoch": 3.15, + "learning_rate": 9.51458811910947e-06, + "loss": 0.7562, + "step": 48964 + }, + { + "epoch": 3.15, + "learning_rate": 9.514373430086521e-06, + "loss": 0.8436, + "step": 48965 + }, + { + "epoch": 3.15, + "learning_rate": 9.51415874106357e-06, + "loss": 0.8093, + "step": 48966 + }, + { + "epoch": 3.15, + "learning_rate": 9.51394405204062e-06, + "loss": 0.8009, + "step": 48967 + }, + { + "epoch": 3.15, + "learning_rate": 9.51372936301767e-06, + "loss": 0.8214, + "step": 48968 + }, + { + "epoch": 3.15, + "learning_rate": 9.513514673994719e-06, + "loss": 0.8498, + "step": 48969 + }, + { + "epoch": 3.15, + "learning_rate": 9.51329998497177e-06, + "loss": 0.7897, + "step": 48970 + }, + { + "epoch": 3.15, + "learning_rate": 9.513085295948819e-06, + "loss": 0.83, + "step": 48971 + }, + { + "epoch": 3.15, + "learning_rate": 9.51287060692587e-06, + "loss": 0.7618, + "step": 48972 + }, + { + "epoch": 3.15, + "learning_rate": 9.512655917902918e-06, + "loss": 0.7624, + "step": 48973 + }, + { + "epoch": 3.15, + "learning_rate": 9.512441228879969e-06, + "loss": 0.8183, + "step": 48974 + }, + { + "epoch": 3.15, + "learning_rate": 9.512226539857018e-06, + "loss": 0.8262, + "step": 48975 + }, + { + "epoch": 3.15, + "learning_rate": 9.512011850834067e-06, + "loss": 0.8021, + "step": 48976 + }, + { + "epoch": 3.15, + "learning_rate": 9.511797161811117e-06, + "loss": 0.7935, + "step": 48977 + }, + { + "epoch": 3.15, + "learning_rate": 9.511582472788166e-06, + "loss": 0.8136, + "step": 48978 + }, + { + "epoch": 3.15, + "learning_rate": 9.511367783765217e-06, + "loss": 0.8464, + "step": 48979 + }, + { + "epoch": 3.15, + "learning_rate": 9.511153094742268e-06, + "loss": 0.8301, + "step": 48980 + }, + { + "epoch": 3.15, + "learning_rate": 9.510938405719317e-06, + "loss": 0.8073, + "step": 48981 + }, + { + "epoch": 3.15, + "learning_rate": 9.510723716696366e-06, + "loss": 0.8243, + "step": 48982 + }, + { + "epoch": 3.15, + "learning_rate": 9.510509027673416e-06, + "loss": 0.8067, + "step": 48983 + }, + { + "epoch": 3.15, + "learning_rate": 9.510294338650465e-06, + "loss": 0.7603, + "step": 48984 + }, + { + "epoch": 3.15, + "learning_rate": 9.510079649627514e-06, + "loss": 0.7926, + "step": 48985 + }, + { + "epoch": 3.15, + "learning_rate": 9.509864960604565e-06, + "loss": 0.8358, + "step": 48986 + }, + { + "epoch": 3.15, + "learning_rate": 9.509650271581615e-06, + "loss": 0.813, + "step": 48987 + }, + { + "epoch": 3.15, + "learning_rate": 9.509435582558664e-06, + "loss": 0.8214, + "step": 48988 + }, + { + "epoch": 3.15, + "learning_rate": 9.509220893535715e-06, + "loss": 0.8924, + "step": 48989 + }, + { + "epoch": 3.15, + "learning_rate": 9.509006204512764e-06, + "loss": 0.856, + "step": 48990 + }, + { + "epoch": 3.15, + "learning_rate": 9.508791515489813e-06, + "loss": 0.8366, + "step": 48991 + }, + { + "epoch": 3.15, + "learning_rate": 9.508576826466864e-06, + "loss": 0.8292, + "step": 48992 + }, + { + "epoch": 3.15, + "learning_rate": 9.508362137443913e-06, + "loss": 0.8269, + "step": 48993 + }, + { + "epoch": 3.15, + "learning_rate": 9.508147448420963e-06, + "loss": 0.814, + "step": 48994 + }, + { + "epoch": 3.15, + "learning_rate": 9.507932759398012e-06, + "loss": 0.8112, + "step": 48995 + }, + { + "epoch": 3.15, + "learning_rate": 9.507718070375063e-06, + "loss": 0.8618, + "step": 48996 + }, + { + "epoch": 3.15, + "learning_rate": 9.507503381352112e-06, + "loss": 0.7972, + "step": 48997 + }, + { + "epoch": 3.15, + "learning_rate": 9.507288692329163e-06, + "loss": 0.8258, + "step": 48998 + }, + { + "epoch": 3.15, + "learning_rate": 9.507074003306212e-06, + "loss": 0.8345, + "step": 48999 + }, + { + "epoch": 3.15, + "learning_rate": 9.50685931428326e-06, + "loss": 0.8214, + "step": 49000 + }, + { + "epoch": 3.15, + "eval_laion/cider": 0.573724993007589, + "eval_laion/meteor": 0.40491326278548845, + "eval_laion/pred_num_words": 38.216796875, + "eval_laion/sacrebleu": 16.45091013190023, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0530622005462646, + "eval_runtime": 97.1273, + "eval_samples_per_second": 5.138, + "eval_steps_per_second": 0.329, + "step": 49000 + }, + { + "epoch": 3.15, + "learning_rate": 9.506644625260311e-06, + "loss": 0.846, + "step": 49001 + }, + { + "epoch": 3.15, + "learning_rate": 9.506429936237362e-06, + "loss": 0.8247, + "step": 49002 + }, + { + "epoch": 3.15, + "learning_rate": 9.50621524721441e-06, + "loss": 0.7744, + "step": 49003 + }, + { + "epoch": 3.15, + "learning_rate": 9.50600055819146e-06, + "loss": 0.8398, + "step": 49004 + }, + { + "epoch": 3.15, + "learning_rate": 9.50578586916851e-06, + "loss": 0.8079, + "step": 49005 + }, + { + "epoch": 3.15, + "learning_rate": 9.50557118014556e-06, + "loss": 0.8119, + "step": 49006 + }, + { + "epoch": 3.15, + "learning_rate": 9.50535649112261e-06, + "loss": 0.8706, + "step": 49007 + }, + { + "epoch": 3.15, + "learning_rate": 9.50514180209966e-06, + "loss": 0.841, + "step": 49008 + }, + { + "epoch": 3.15, + "learning_rate": 9.50492711307671e-06, + "loss": 0.7817, + "step": 49009 + }, + { + "epoch": 3.15, + "learning_rate": 9.504712424053759e-06, + "loss": 0.8261, + "step": 49010 + }, + { + "epoch": 3.15, + "learning_rate": 9.50449773503081e-06, + "loss": 0.8629, + "step": 49011 + }, + { + "epoch": 3.15, + "learning_rate": 9.504283046007858e-06, + "loss": 0.8139, + "step": 49012 + }, + { + "epoch": 3.15, + "learning_rate": 9.504068356984907e-06, + "loss": 0.8196, + "step": 49013 + }, + { + "epoch": 3.15, + "learning_rate": 9.503853667961958e-06, + "loss": 0.8173, + "step": 49014 + }, + { + "epoch": 3.15, + "learning_rate": 9.503638978939007e-06, + "loss": 0.8289, + "step": 49015 + }, + { + "epoch": 3.15, + "learning_rate": 9.503424289916058e-06, + "loss": 0.842, + "step": 49016 + }, + { + "epoch": 3.15, + "learning_rate": 9.503209600893108e-06, + "loss": 0.8884, + "step": 49017 + }, + { + "epoch": 3.15, + "learning_rate": 9.502994911870157e-06, + "loss": 0.7823, + "step": 49018 + }, + { + "epoch": 3.15, + "learning_rate": 9.502780222847206e-06, + "loss": 0.8806, + "step": 49019 + }, + { + "epoch": 3.15, + "learning_rate": 9.502565533824257e-06, + "loss": 0.8042, + "step": 49020 + }, + { + "epoch": 3.15, + "learning_rate": 9.502350844801306e-06, + "loss": 0.7944, + "step": 49021 + }, + { + "epoch": 3.15, + "learning_rate": 9.502136155778355e-06, + "loss": 0.8527, + "step": 49022 + }, + { + "epoch": 3.15, + "learning_rate": 9.501921466755405e-06, + "loss": 0.8072, + "step": 49023 + }, + { + "epoch": 3.15, + "learning_rate": 9.501706777732456e-06, + "loss": 0.9036, + "step": 49024 + }, + { + "epoch": 3.15, + "learning_rate": 9.501492088709505e-06, + "loss": 0.8382, + "step": 49025 + }, + { + "epoch": 3.15, + "learning_rate": 9.501277399686556e-06, + "loss": 0.7639, + "step": 49026 + }, + { + "epoch": 3.15, + "learning_rate": 9.501062710663605e-06, + "loss": 0.8186, + "step": 49027 + }, + { + "epoch": 3.15, + "learning_rate": 9.500848021640654e-06, + "loss": 0.8528, + "step": 49028 + }, + { + "epoch": 3.15, + "learning_rate": 9.500633332617704e-06, + "loss": 0.7779, + "step": 49029 + }, + { + "epoch": 3.15, + "learning_rate": 9.500418643594755e-06, + "loss": 0.821, + "step": 49030 + }, + { + "epoch": 3.15, + "learning_rate": 9.500203954571804e-06, + "loss": 0.8083, + "step": 49031 + }, + { + "epoch": 3.15, + "learning_rate": 9.499989265548853e-06, + "loss": 0.8191, + "step": 49032 + }, + { + "epoch": 3.15, + "learning_rate": 9.499774576525903e-06, + "loss": 0.8202, + "step": 49033 + }, + { + "epoch": 3.15, + "learning_rate": 9.499559887502952e-06, + "loss": 0.8229, + "step": 49034 + }, + { + "epoch": 3.15, + "learning_rate": 9.499345198480003e-06, + "loss": 0.7714, + "step": 49035 + }, + { + "epoch": 3.15, + "learning_rate": 9.499130509457052e-06, + "loss": 0.8192, + "step": 49036 + }, + { + "epoch": 3.15, + "learning_rate": 9.498915820434101e-06, + "loss": 0.8442, + "step": 49037 + }, + { + "epoch": 3.15, + "learning_rate": 9.498701131411152e-06, + "loss": 0.8243, + "step": 49038 + }, + { + "epoch": 3.16, + "learning_rate": 9.498486442388202e-06, + "loss": 0.7797, + "step": 49039 + }, + { + "epoch": 3.16, + "learning_rate": 9.498271753365251e-06, + "loss": 0.7762, + "step": 49040 + }, + { + "epoch": 3.16, + "learning_rate": 9.4980570643423e-06, + "loss": 0.8139, + "step": 49041 + }, + { + "epoch": 3.16, + "learning_rate": 9.497842375319351e-06, + "loss": 0.7685, + "step": 49042 + }, + { + "epoch": 3.16, + "learning_rate": 9.4976276862964e-06, + "loss": 0.8542, + "step": 49043 + }, + { + "epoch": 3.16, + "learning_rate": 9.497412997273449e-06, + "loss": 0.8241, + "step": 49044 + }, + { + "epoch": 3.16, + "learning_rate": 9.497198308250501e-06, + "loss": 0.7656, + "step": 49045 + }, + { + "epoch": 3.16, + "learning_rate": 9.49698361922755e-06, + "loss": 0.8125, + "step": 49046 + }, + { + "epoch": 3.16, + "learning_rate": 9.496768930204599e-06, + "loss": 0.8444, + "step": 49047 + }, + { + "epoch": 3.16, + "learning_rate": 9.49655424118165e-06, + "loss": 0.8767, + "step": 49048 + }, + { + "epoch": 3.16, + "learning_rate": 9.496339552158699e-06, + "loss": 0.7985, + "step": 49049 + }, + { + "epoch": 3.16, + "learning_rate": 9.496124863135748e-06, + "loss": 0.7751, + "step": 49050 + }, + { + "epoch": 3.16, + "learning_rate": 9.495910174112798e-06, + "loss": 0.8034, + "step": 49051 + }, + { + "epoch": 3.16, + "learning_rate": 9.495695485089849e-06, + "loss": 0.8385, + "step": 49052 + }, + { + "epoch": 3.16, + "learning_rate": 9.495480796066898e-06, + "loss": 0.817, + "step": 49053 + }, + { + "epoch": 3.16, + "learning_rate": 9.495266107043947e-06, + "loss": 0.8329, + "step": 49054 + }, + { + "epoch": 3.16, + "learning_rate": 9.495051418020998e-06, + "loss": 0.8088, + "step": 49055 + }, + { + "epoch": 3.16, + "learning_rate": 9.494836728998047e-06, + "loss": 0.8298, + "step": 49056 + }, + { + "epoch": 3.16, + "learning_rate": 9.494622039975097e-06, + "loss": 0.8423, + "step": 49057 + }, + { + "epoch": 3.16, + "learning_rate": 9.494407350952146e-06, + "loss": 0.8786, + "step": 49058 + }, + { + "epoch": 3.16, + "learning_rate": 9.494192661929197e-06, + "loss": 0.8324, + "step": 49059 + }, + { + "epoch": 3.16, + "learning_rate": 9.493977972906246e-06, + "loss": 0.7868, + "step": 49060 + }, + { + "epoch": 3.16, + "learning_rate": 9.493763283883297e-06, + "loss": 0.8541, + "step": 49061 + }, + { + "epoch": 3.16, + "learning_rate": 9.493548594860345e-06, + "loss": 0.8227, + "step": 49062 + }, + { + "epoch": 3.16, + "learning_rate": 9.493333905837394e-06, + "loss": 0.8122, + "step": 49063 + }, + { + "epoch": 3.16, + "learning_rate": 9.493119216814445e-06, + "loss": 0.8303, + "step": 49064 + }, + { + "epoch": 3.16, + "learning_rate": 9.492904527791494e-06, + "loss": 0.7834, + "step": 49065 + }, + { + "epoch": 3.16, + "learning_rate": 9.492689838768545e-06, + "loss": 0.7779, + "step": 49066 + }, + { + "epoch": 3.16, + "learning_rate": 9.492475149745595e-06, + "loss": 0.8221, + "step": 49067 + }, + { + "epoch": 3.16, + "learning_rate": 9.492260460722644e-06, + "loss": 0.8261, + "step": 49068 + }, + { + "epoch": 3.16, + "learning_rate": 9.492045771699693e-06, + "loss": 0.7713, + "step": 49069 + }, + { + "epoch": 3.16, + "learning_rate": 9.491831082676744e-06, + "loss": 0.7915, + "step": 49070 + }, + { + "epoch": 3.16, + "learning_rate": 9.491616393653793e-06, + "loss": 0.7686, + "step": 49071 + }, + { + "epoch": 3.16, + "learning_rate": 9.491401704630842e-06, + "loss": 0.8782, + "step": 49072 + }, + { + "epoch": 3.16, + "learning_rate": 9.491187015607893e-06, + "loss": 0.8566, + "step": 49073 + }, + { + "epoch": 3.16, + "learning_rate": 9.490972326584943e-06, + "loss": 0.8168, + "step": 49074 + }, + { + "epoch": 3.16, + "learning_rate": 9.490757637561992e-06, + "loss": 0.8231, + "step": 49075 + }, + { + "epoch": 3.16, + "learning_rate": 9.490542948539043e-06, + "loss": 0.7528, + "step": 49076 + }, + { + "epoch": 3.16, + "learning_rate": 9.490328259516092e-06, + "loss": 0.8312, + "step": 49077 + }, + { + "epoch": 3.16, + "learning_rate": 9.49011357049314e-06, + "loss": 0.7804, + "step": 49078 + }, + { + "epoch": 3.16, + "learning_rate": 9.489898881470191e-06, + "loss": 0.8573, + "step": 49079 + }, + { + "epoch": 3.16, + "learning_rate": 9.48968419244724e-06, + "loss": 0.8276, + "step": 49080 + }, + { + "epoch": 3.16, + "learning_rate": 9.489469503424291e-06, + "loss": 0.8056, + "step": 49081 + }, + { + "epoch": 3.16, + "learning_rate": 9.48925481440134e-06, + "loss": 0.8035, + "step": 49082 + }, + { + "epoch": 3.16, + "learning_rate": 9.48904012537839e-06, + "loss": 0.8113, + "step": 49083 + }, + { + "epoch": 3.16, + "learning_rate": 9.48882543635544e-06, + "loss": 0.7519, + "step": 49084 + }, + { + "epoch": 3.16, + "learning_rate": 9.48861074733249e-06, + "loss": 0.7961, + "step": 49085 + }, + { + "epoch": 3.16, + "learning_rate": 9.48839605830954e-06, + "loss": 0.8174, + "step": 49086 + }, + { + "epoch": 3.16, + "learning_rate": 9.488181369286588e-06, + "loss": 0.8561, + "step": 49087 + }, + { + "epoch": 3.16, + "learning_rate": 9.487966680263639e-06, + "loss": 0.858, + "step": 49088 + }, + { + "epoch": 3.16, + "learning_rate": 9.48775199124069e-06, + "loss": 0.7613, + "step": 49089 + }, + { + "epoch": 3.16, + "learning_rate": 9.487537302217739e-06, + "loss": 0.7724, + "step": 49090 + }, + { + "epoch": 3.16, + "learning_rate": 9.487322613194787e-06, + "loss": 0.8101, + "step": 49091 + }, + { + "epoch": 3.16, + "learning_rate": 9.487107924171838e-06, + "loss": 0.8147, + "step": 49092 + }, + { + "epoch": 3.16, + "learning_rate": 9.486893235148887e-06, + "loss": 0.8199, + "step": 49093 + }, + { + "epoch": 3.16, + "learning_rate": 9.486678546125938e-06, + "loss": 0.7703, + "step": 49094 + }, + { + "epoch": 3.16, + "learning_rate": 9.486463857102987e-06, + "loss": 0.7987, + "step": 49095 + }, + { + "epoch": 3.16, + "learning_rate": 9.486249168080037e-06, + "loss": 0.8386, + "step": 49096 + }, + { + "epoch": 3.16, + "learning_rate": 9.486034479057086e-06, + "loss": 0.8325, + "step": 49097 + }, + { + "epoch": 3.16, + "learning_rate": 9.485819790034137e-06, + "loss": 0.8454, + "step": 49098 + }, + { + "epoch": 3.16, + "learning_rate": 9.485605101011186e-06, + "loss": 0.8424, + "step": 49099 + }, + { + "epoch": 3.16, + "learning_rate": 9.485390411988235e-06, + "loss": 0.7729, + "step": 49100 + }, + { + "epoch": 3.16, + "learning_rate": 9.485175722965286e-06, + "loss": 0.8005, + "step": 49101 + }, + { + "epoch": 3.16, + "learning_rate": 9.484961033942335e-06, + "loss": 0.8959, + "step": 49102 + }, + { + "epoch": 3.16, + "learning_rate": 9.484746344919385e-06, + "loss": 0.8306, + "step": 49103 + }, + { + "epoch": 3.16, + "learning_rate": 9.484531655896436e-06, + "loss": 0.8348, + "step": 49104 + }, + { + "epoch": 3.16, + "learning_rate": 9.484316966873485e-06, + "loss": 0.7713, + "step": 49105 + }, + { + "epoch": 3.16, + "learning_rate": 9.484102277850534e-06, + "loss": 0.7506, + "step": 49106 + }, + { + "epoch": 3.16, + "learning_rate": 9.483887588827584e-06, + "loss": 0.8452, + "step": 49107 + }, + { + "epoch": 3.16, + "learning_rate": 9.483672899804633e-06, + "loss": 0.8591, + "step": 49108 + }, + { + "epoch": 3.16, + "learning_rate": 9.483458210781682e-06, + "loss": 0.8039, + "step": 49109 + }, + { + "epoch": 3.16, + "learning_rate": 9.483243521758733e-06, + "loss": 0.8462, + "step": 49110 + }, + { + "epoch": 3.16, + "learning_rate": 9.483028832735784e-06, + "loss": 0.8954, + "step": 49111 + }, + { + "epoch": 3.16, + "learning_rate": 9.482814143712833e-06, + "loss": 0.8268, + "step": 49112 + }, + { + "epoch": 3.16, + "learning_rate": 9.482599454689883e-06, + "loss": 0.7808, + "step": 49113 + }, + { + "epoch": 3.16, + "learning_rate": 9.482384765666932e-06, + "loss": 0.7666, + "step": 49114 + }, + { + "epoch": 3.16, + "learning_rate": 9.482170076643981e-06, + "loss": 0.8191, + "step": 49115 + }, + { + "epoch": 3.16, + "learning_rate": 9.481955387621032e-06, + "loss": 0.7972, + "step": 49116 + }, + { + "epoch": 3.16, + "learning_rate": 9.481740698598083e-06, + "loss": 0.8765, + "step": 49117 + }, + { + "epoch": 3.16, + "learning_rate": 9.481526009575132e-06, + "loss": 0.8178, + "step": 49118 + }, + { + "epoch": 3.16, + "learning_rate": 9.48131132055218e-06, + "loss": 0.7596, + "step": 49119 + }, + { + "epoch": 3.16, + "learning_rate": 9.481096631529231e-06, + "loss": 0.8363, + "step": 49120 + }, + { + "epoch": 3.16, + "learning_rate": 9.48088194250628e-06, + "loss": 0.913, + "step": 49121 + }, + { + "epoch": 3.16, + "learning_rate": 9.480667253483329e-06, + "loss": 0.7853, + "step": 49122 + }, + { + "epoch": 3.16, + "learning_rate": 9.48045256446038e-06, + "loss": 0.8497, + "step": 49123 + }, + { + "epoch": 3.16, + "learning_rate": 9.480237875437429e-06, + "loss": 0.8196, + "step": 49124 + }, + { + "epoch": 3.16, + "learning_rate": 9.48002318641448e-06, + "loss": 0.7929, + "step": 49125 + }, + { + "epoch": 3.16, + "learning_rate": 9.47980849739153e-06, + "loss": 0.8419, + "step": 49126 + }, + { + "epoch": 3.16, + "learning_rate": 9.479593808368579e-06, + "loss": 0.8288, + "step": 49127 + }, + { + "epoch": 3.16, + "learning_rate": 9.479379119345628e-06, + "loss": 0.7648, + "step": 49128 + }, + { + "epoch": 3.16, + "learning_rate": 9.479164430322679e-06, + "loss": 0.8324, + "step": 49129 + }, + { + "epoch": 3.16, + "learning_rate": 9.478949741299728e-06, + "loss": 0.8928, + "step": 49130 + }, + { + "epoch": 3.16, + "learning_rate": 9.478735052276777e-06, + "loss": 0.7887, + "step": 49131 + }, + { + "epoch": 3.16, + "learning_rate": 9.478520363253829e-06, + "loss": 0.8447, + "step": 49132 + }, + { + "epoch": 3.16, + "learning_rate": 9.478305674230878e-06, + "loss": 0.8105, + "step": 49133 + }, + { + "epoch": 3.16, + "learning_rate": 9.478090985207927e-06, + "loss": 0.8292, + "step": 49134 + }, + { + "epoch": 3.16, + "learning_rate": 9.477876296184978e-06, + "loss": 0.8211, + "step": 49135 + }, + { + "epoch": 3.16, + "learning_rate": 9.477661607162026e-06, + "loss": 0.8158, + "step": 49136 + }, + { + "epoch": 3.16, + "learning_rate": 9.477446918139075e-06, + "loss": 0.7668, + "step": 49137 + }, + { + "epoch": 3.16, + "learning_rate": 9.477232229116126e-06, + "loss": 0.8138, + "step": 49138 + }, + { + "epoch": 3.16, + "learning_rate": 9.477017540093177e-06, + "loss": 0.7912, + "step": 49139 + }, + { + "epoch": 3.16, + "learning_rate": 9.476802851070226e-06, + "loss": 0.8273, + "step": 49140 + }, + { + "epoch": 3.16, + "learning_rate": 9.476588162047275e-06, + "loss": 0.823, + "step": 49141 + }, + { + "epoch": 3.16, + "learning_rate": 9.476373473024325e-06, + "loss": 0.8751, + "step": 49142 + }, + { + "epoch": 3.16, + "learning_rate": 9.476158784001374e-06, + "loss": 0.8203, + "step": 49143 + }, + { + "epoch": 3.16, + "learning_rate": 9.475944094978425e-06, + "loss": 0.8164, + "step": 49144 + }, + { + "epoch": 3.16, + "learning_rate": 9.475729405955474e-06, + "loss": 0.832, + "step": 49145 + }, + { + "epoch": 3.16, + "learning_rate": 9.475514716932525e-06, + "loss": 0.8129, + "step": 49146 + }, + { + "epoch": 3.16, + "learning_rate": 9.475300027909574e-06, + "loss": 0.8516, + "step": 49147 + }, + { + "epoch": 3.16, + "learning_rate": 9.475085338886624e-06, + "loss": 0.8784, + "step": 49148 + }, + { + "epoch": 3.16, + "learning_rate": 9.474870649863673e-06, + "loss": 0.8247, + "step": 49149 + }, + { + "epoch": 3.16, + "learning_rate": 9.474655960840722e-06, + "loss": 0.802, + "step": 49150 + }, + { + "epoch": 3.16, + "learning_rate": 9.474441271817773e-06, + "loss": 0.846, + "step": 49151 + }, + { + "epoch": 3.16, + "learning_rate": 9.474226582794822e-06, + "loss": 0.7741, + "step": 49152 + }, + { + "epoch": 3.16, + "learning_rate": 9.474011893771872e-06, + "loss": 0.8041, + "step": 49153 + }, + { + "epoch": 3.16, + "learning_rate": 9.473797204748923e-06, + "loss": 0.8507, + "step": 49154 + }, + { + "epoch": 3.16, + "learning_rate": 9.473582515725972e-06, + "loss": 0.8514, + "step": 49155 + }, + { + "epoch": 3.16, + "learning_rate": 9.473367826703021e-06, + "loss": 0.8324, + "step": 49156 + }, + { + "epoch": 3.16, + "learning_rate": 9.473153137680072e-06, + "loss": 0.8296, + "step": 49157 + }, + { + "epoch": 3.16, + "learning_rate": 9.47293844865712e-06, + "loss": 0.8746, + "step": 49158 + }, + { + "epoch": 3.16, + "learning_rate": 9.47272375963417e-06, + "loss": 0.8133, + "step": 49159 + }, + { + "epoch": 3.16, + "learning_rate": 9.47250907061122e-06, + "loss": 0.9454, + "step": 49160 + }, + { + "epoch": 3.16, + "learning_rate": 9.472294381588271e-06, + "loss": 0.7776, + "step": 49161 + }, + { + "epoch": 3.16, + "learning_rate": 9.47207969256532e-06, + "loss": 0.7916, + "step": 49162 + }, + { + "epoch": 3.16, + "learning_rate": 9.47186500354237e-06, + "loss": 0.7318, + "step": 49163 + }, + { + "epoch": 3.16, + "learning_rate": 9.47165031451942e-06, + "loss": 0.7976, + "step": 49164 + }, + { + "epoch": 3.16, + "learning_rate": 9.471435625496468e-06, + "loss": 0.8152, + "step": 49165 + }, + { + "epoch": 3.16, + "learning_rate": 9.471220936473519e-06, + "loss": 0.7775, + "step": 49166 + }, + { + "epoch": 3.16, + "learning_rate": 9.471006247450568e-06, + "loss": 0.8517, + "step": 49167 + }, + { + "epoch": 3.16, + "learning_rate": 9.470791558427619e-06, + "loss": 0.8032, + "step": 49168 + }, + { + "epoch": 3.16, + "learning_rate": 9.470576869404668e-06, + "loss": 0.8409, + "step": 49169 + }, + { + "epoch": 3.16, + "learning_rate": 9.470362180381718e-06, + "loss": 0.814, + "step": 49170 + }, + { + "epoch": 3.16, + "learning_rate": 9.470147491358767e-06, + "loss": 0.8856, + "step": 49171 + }, + { + "epoch": 3.16, + "learning_rate": 9.469932802335818e-06, + "loss": 0.888, + "step": 49172 + }, + { + "epoch": 3.16, + "learning_rate": 9.469718113312867e-06, + "loss": 0.8329, + "step": 49173 + }, + { + "epoch": 3.16, + "learning_rate": 9.469503424289916e-06, + "loss": 0.8481, + "step": 49174 + }, + { + "epoch": 3.16, + "learning_rate": 9.469288735266967e-06, + "loss": 0.8763, + "step": 49175 + }, + { + "epoch": 3.16, + "learning_rate": 9.469074046244017e-06, + "loss": 0.8453, + "step": 49176 + }, + { + "epoch": 3.16, + "learning_rate": 9.468859357221066e-06, + "loss": 0.7746, + "step": 49177 + }, + { + "epoch": 3.16, + "learning_rate": 9.468644668198115e-06, + "loss": 0.8588, + "step": 49178 + }, + { + "epoch": 3.16, + "learning_rate": 9.468429979175166e-06, + "loss": 0.8254, + "step": 49179 + }, + { + "epoch": 3.16, + "learning_rate": 9.468215290152215e-06, + "loss": 0.7739, + "step": 49180 + }, + { + "epoch": 3.16, + "learning_rate": 9.468000601129265e-06, + "loss": 0.8659, + "step": 49181 + }, + { + "epoch": 3.16, + "learning_rate": 9.467785912106314e-06, + "loss": 0.8255, + "step": 49182 + }, + { + "epoch": 3.16, + "learning_rate": 9.467571223083365e-06, + "loss": 0.7783, + "step": 49183 + }, + { + "epoch": 3.16, + "learning_rate": 9.467356534060414e-06, + "loss": 0.7681, + "step": 49184 + }, + { + "epoch": 3.16, + "learning_rate": 9.467141845037465e-06, + "loss": 0.8184, + "step": 49185 + }, + { + "epoch": 3.16, + "learning_rate": 9.466927156014514e-06, + "loss": 0.8501, + "step": 49186 + }, + { + "epoch": 3.16, + "learning_rate": 9.466712466991563e-06, + "loss": 0.8028, + "step": 49187 + }, + { + "epoch": 3.16, + "learning_rate": 9.466497777968613e-06, + "loss": 0.8617, + "step": 49188 + }, + { + "epoch": 3.16, + "learning_rate": 9.466283088945662e-06, + "loss": 0.828, + "step": 49189 + }, + { + "epoch": 3.16, + "learning_rate": 9.466068399922713e-06, + "loss": 0.7606, + "step": 49190 + }, + { + "epoch": 3.16, + "learning_rate": 9.465853710899764e-06, + "loss": 0.8815, + "step": 49191 + }, + { + "epoch": 3.16, + "learning_rate": 9.465639021876813e-06, + "loss": 0.7723, + "step": 49192 + }, + { + "epoch": 3.16, + "learning_rate": 9.465424332853862e-06, + "loss": 0.8149, + "step": 49193 + }, + { + "epoch": 3.17, + "learning_rate": 9.465209643830912e-06, + "loss": 0.8317, + "step": 49194 + }, + { + "epoch": 3.17, + "learning_rate": 9.464994954807961e-06, + "loss": 0.8804, + "step": 49195 + }, + { + "epoch": 3.17, + "learning_rate": 9.46478026578501e-06, + "loss": 0.785, + "step": 49196 + }, + { + "epoch": 3.17, + "learning_rate": 9.46456557676206e-06, + "loss": 0.7653, + "step": 49197 + }, + { + "epoch": 3.17, + "learning_rate": 9.464350887739111e-06, + "loss": 0.8234, + "step": 49198 + }, + { + "epoch": 3.17, + "learning_rate": 9.46413619871616e-06, + "loss": 0.8789, + "step": 49199 + }, + { + "epoch": 3.17, + "learning_rate": 9.463921509693211e-06, + "loss": 0.764, + "step": 49200 + }, + { + "epoch": 3.17, + "learning_rate": 9.46370682067026e-06, + "loss": 0.8017, + "step": 49201 + }, + { + "epoch": 3.17, + "learning_rate": 9.463492131647309e-06, + "loss": 0.88, + "step": 49202 + }, + { + "epoch": 3.17, + "learning_rate": 9.46327744262436e-06, + "loss": 0.8887, + "step": 49203 + }, + { + "epoch": 3.17, + "learning_rate": 9.463062753601409e-06, + "loss": 0.784, + "step": 49204 + }, + { + "epoch": 3.17, + "learning_rate": 9.46284806457846e-06, + "loss": 0.7922, + "step": 49205 + }, + { + "epoch": 3.17, + "learning_rate": 9.462633375555508e-06, + "loss": 0.7915, + "step": 49206 + }, + { + "epoch": 3.17, + "learning_rate": 9.462418686532559e-06, + "loss": 0.9004, + "step": 49207 + }, + { + "epoch": 3.17, + "learning_rate": 9.462203997509608e-06, + "loss": 0.7707, + "step": 49208 + }, + { + "epoch": 3.17, + "learning_rate": 9.461989308486657e-06, + "loss": 0.8486, + "step": 49209 + }, + { + "epoch": 3.17, + "learning_rate": 9.461774619463707e-06, + "loss": 0.8315, + "step": 49210 + }, + { + "epoch": 3.17, + "learning_rate": 9.461559930440756e-06, + "loss": 0.834, + "step": 49211 + }, + { + "epoch": 3.17, + "learning_rate": 9.461345241417807e-06, + "loss": 0.8837, + "step": 49212 + }, + { + "epoch": 3.17, + "learning_rate": 9.461130552394858e-06, + "loss": 0.8605, + "step": 49213 + }, + { + "epoch": 3.17, + "learning_rate": 9.460915863371907e-06, + "loss": 0.8145, + "step": 49214 + }, + { + "epoch": 3.17, + "learning_rate": 9.460701174348956e-06, + "loss": 0.7793, + "step": 49215 + }, + { + "epoch": 3.17, + "learning_rate": 9.460486485326006e-06, + "loss": 0.8378, + "step": 49216 + }, + { + "epoch": 3.17, + "learning_rate": 9.460271796303055e-06, + "loss": 0.7832, + "step": 49217 + }, + { + "epoch": 3.17, + "learning_rate": 9.460057107280104e-06, + "loss": 0.786, + "step": 49218 + }, + { + "epoch": 3.17, + "learning_rate": 9.459842418257155e-06, + "loss": 0.8533, + "step": 49219 + }, + { + "epoch": 3.17, + "learning_rate": 9.459627729234206e-06, + "loss": 0.8398, + "step": 49220 + }, + { + "epoch": 3.17, + "learning_rate": 9.459413040211255e-06, + "loss": 0.8293, + "step": 49221 + }, + { + "epoch": 3.17, + "learning_rate": 9.459198351188305e-06, + "loss": 0.8164, + "step": 49222 + }, + { + "epoch": 3.17, + "learning_rate": 9.458983662165354e-06, + "loss": 0.8158, + "step": 49223 + }, + { + "epoch": 3.17, + "learning_rate": 9.458768973142403e-06, + "loss": 0.808, + "step": 49224 + }, + { + "epoch": 3.17, + "learning_rate": 9.458554284119454e-06, + "loss": 0.7913, + "step": 49225 + }, + { + "epoch": 3.17, + "learning_rate": 9.458339595096504e-06, + "loss": 0.8665, + "step": 49226 + }, + { + "epoch": 3.17, + "learning_rate": 9.458124906073553e-06, + "loss": 0.8263, + "step": 49227 + }, + { + "epoch": 3.17, + "learning_rate": 9.457910217050602e-06, + "loss": 0.8025, + "step": 49228 + }, + { + "epoch": 3.17, + "learning_rate": 9.457695528027653e-06, + "loss": 0.7959, + "step": 49229 + }, + { + "epoch": 3.17, + "learning_rate": 9.457480839004702e-06, + "loss": 0.8617, + "step": 49230 + }, + { + "epoch": 3.17, + "learning_rate": 9.457266149981753e-06, + "loss": 0.8184, + "step": 49231 + }, + { + "epoch": 3.17, + "learning_rate": 9.457051460958802e-06, + "loss": 0.8325, + "step": 49232 + }, + { + "epoch": 3.17, + "learning_rate": 9.45683677193585e-06, + "loss": 0.8024, + "step": 49233 + }, + { + "epoch": 3.17, + "learning_rate": 9.456622082912901e-06, + "loss": 0.7705, + "step": 49234 + }, + { + "epoch": 3.17, + "learning_rate": 9.456407393889952e-06, + "loss": 0.7628, + "step": 49235 + }, + { + "epoch": 3.17, + "learning_rate": 9.456192704867001e-06, + "loss": 0.8863, + "step": 49236 + }, + { + "epoch": 3.17, + "learning_rate": 9.45597801584405e-06, + "loss": 0.7783, + "step": 49237 + }, + { + "epoch": 3.17, + "learning_rate": 9.4557633268211e-06, + "loss": 0.7773, + "step": 49238 + }, + { + "epoch": 3.17, + "learning_rate": 9.45554863779815e-06, + "loss": 0.8229, + "step": 49239 + }, + { + "epoch": 3.17, + "learning_rate": 9.4553339487752e-06, + "loss": 0.8105, + "step": 49240 + }, + { + "epoch": 3.17, + "learning_rate": 9.45511925975225e-06, + "loss": 0.8338, + "step": 49241 + }, + { + "epoch": 3.17, + "learning_rate": 9.4549045707293e-06, + "loss": 0.815, + "step": 49242 + }, + { + "epoch": 3.17, + "learning_rate": 9.454689881706349e-06, + "loss": 0.7836, + "step": 49243 + }, + { + "epoch": 3.17, + "learning_rate": 9.4544751926834e-06, + "loss": 0.8682, + "step": 49244 + }, + { + "epoch": 3.17, + "learning_rate": 9.454260503660448e-06, + "loss": 0.7724, + "step": 49245 + }, + { + "epoch": 3.17, + "learning_rate": 9.454045814637497e-06, + "loss": 0.7675, + "step": 49246 + }, + { + "epoch": 3.17, + "learning_rate": 9.453831125614548e-06, + "loss": 0.8643, + "step": 49247 + }, + { + "epoch": 3.17, + "learning_rate": 9.453616436591599e-06, + "loss": 0.7281, + "step": 49248 + }, + { + "epoch": 3.17, + "learning_rate": 9.453401747568648e-06, + "loss": 0.8094, + "step": 49249 + }, + { + "epoch": 3.17, + "learning_rate": 9.453187058545698e-06, + "loss": 0.7849, + "step": 49250 + }, + { + "epoch": 3.17, + "learning_rate": 9.452972369522747e-06, + "loss": 0.8747, + "step": 49251 + }, + { + "epoch": 3.17, + "learning_rate": 9.452757680499796e-06, + "loss": 0.8181, + "step": 49252 + }, + { + "epoch": 3.17, + "learning_rate": 9.452542991476847e-06, + "loss": 0.8025, + "step": 49253 + }, + { + "epoch": 3.17, + "learning_rate": 9.452328302453896e-06, + "loss": 0.7634, + "step": 49254 + }, + { + "epoch": 3.17, + "learning_rate": 9.452113613430946e-06, + "loss": 0.8219, + "step": 49255 + }, + { + "epoch": 3.17, + "learning_rate": 9.451898924407995e-06, + "loss": 0.792, + "step": 49256 + }, + { + "epoch": 3.17, + "learning_rate": 9.451684235385046e-06, + "loss": 0.7889, + "step": 49257 + }, + { + "epoch": 3.17, + "learning_rate": 9.451469546362095e-06, + "loss": 0.8235, + "step": 49258 + }, + { + "epoch": 3.17, + "learning_rate": 9.451254857339146e-06, + "loss": 0.8131, + "step": 49259 + }, + { + "epoch": 3.17, + "learning_rate": 9.451040168316195e-06, + "loss": 0.7871, + "step": 49260 + }, + { + "epoch": 3.17, + "learning_rate": 9.450825479293244e-06, + "loss": 0.8226, + "step": 49261 + }, + { + "epoch": 3.17, + "learning_rate": 9.450610790270294e-06, + "loss": 0.8185, + "step": 49262 + }, + { + "epoch": 3.17, + "learning_rate": 9.450396101247345e-06, + "loss": 0.8508, + "step": 49263 + }, + { + "epoch": 3.17, + "learning_rate": 9.450181412224394e-06, + "loss": 0.8172, + "step": 49264 + }, + { + "epoch": 3.17, + "learning_rate": 9.449966723201443e-06, + "loss": 0.8565, + "step": 49265 + }, + { + "epoch": 3.17, + "learning_rate": 9.449752034178494e-06, + "loss": 0.9058, + "step": 49266 + }, + { + "epoch": 3.17, + "learning_rate": 9.449537345155543e-06, + "loss": 0.834, + "step": 49267 + }, + { + "epoch": 3.17, + "learning_rate": 9.449322656132593e-06, + "loss": 0.7666, + "step": 49268 + }, + { + "epoch": 3.17, + "learning_rate": 9.449107967109642e-06, + "loss": 0.8686, + "step": 49269 + }, + { + "epoch": 3.17, + "learning_rate": 9.448893278086693e-06, + "loss": 0.8439, + "step": 49270 + }, + { + "epoch": 3.17, + "learning_rate": 9.448678589063742e-06, + "loss": 0.8299, + "step": 49271 + }, + { + "epoch": 3.17, + "learning_rate": 9.448463900040792e-06, + "loss": 0.7765, + "step": 49272 + }, + { + "epoch": 3.17, + "learning_rate": 9.448249211017841e-06, + "loss": 0.8349, + "step": 49273 + }, + { + "epoch": 3.17, + "learning_rate": 9.44803452199489e-06, + "loss": 0.8672, + "step": 49274 + }, + { + "epoch": 3.17, + "learning_rate": 9.447819832971941e-06, + "loss": 0.8683, + "step": 49275 + }, + { + "epoch": 3.17, + "learning_rate": 9.44760514394899e-06, + "loss": 0.8007, + "step": 49276 + }, + { + "epoch": 3.17, + "learning_rate": 9.44739045492604e-06, + "loss": 0.7997, + "step": 49277 + }, + { + "epoch": 3.17, + "learning_rate": 9.447175765903091e-06, + "loss": 0.8481, + "step": 49278 + }, + { + "epoch": 3.17, + "learning_rate": 9.44696107688014e-06, + "loss": 0.8027, + "step": 49279 + }, + { + "epoch": 3.17, + "learning_rate": 9.44674638785719e-06, + "loss": 0.7951, + "step": 49280 + }, + { + "epoch": 3.17, + "learning_rate": 9.44653169883424e-06, + "loss": 0.8572, + "step": 49281 + }, + { + "epoch": 3.17, + "learning_rate": 9.446317009811289e-06, + "loss": 0.8395, + "step": 49282 + }, + { + "epoch": 3.17, + "learning_rate": 9.446102320788338e-06, + "loss": 0.8097, + "step": 49283 + }, + { + "epoch": 3.17, + "learning_rate": 9.445887631765388e-06, + "loss": 0.8513, + "step": 49284 + }, + { + "epoch": 3.17, + "learning_rate": 9.445672942742439e-06, + "loss": 0.7695, + "step": 49285 + }, + { + "epoch": 3.17, + "learning_rate": 9.445458253719488e-06, + "loss": 0.7968, + "step": 49286 + }, + { + "epoch": 3.17, + "learning_rate": 9.445243564696539e-06, + "loss": 0.7949, + "step": 49287 + }, + { + "epoch": 3.17, + "learning_rate": 9.445028875673588e-06, + "loss": 0.8063, + "step": 49288 + }, + { + "epoch": 3.17, + "learning_rate": 9.444814186650637e-06, + "loss": 0.7941, + "step": 49289 + }, + { + "epoch": 3.17, + "learning_rate": 9.444599497627687e-06, + "loss": 0.8389, + "step": 49290 + }, + { + "epoch": 3.17, + "learning_rate": 9.444384808604736e-06, + "loss": 0.7878, + "step": 49291 + }, + { + "epoch": 3.17, + "learning_rate": 9.444170119581787e-06, + "loss": 0.8596, + "step": 49292 + }, + { + "epoch": 3.17, + "learning_rate": 9.443955430558836e-06, + "loss": 0.7591, + "step": 49293 + }, + { + "epoch": 3.17, + "learning_rate": 9.443740741535887e-06, + "loss": 0.7992, + "step": 49294 + }, + { + "epoch": 3.17, + "learning_rate": 9.443526052512936e-06, + "loss": 0.823, + "step": 49295 + }, + { + "epoch": 3.17, + "learning_rate": 9.443311363489985e-06, + "loss": 0.7915, + "step": 49296 + }, + { + "epoch": 3.17, + "learning_rate": 9.443096674467035e-06, + "loss": 0.8188, + "step": 49297 + }, + { + "epoch": 3.17, + "learning_rate": 9.442881985444084e-06, + "loss": 0.8859, + "step": 49298 + }, + { + "epoch": 3.17, + "learning_rate": 9.442667296421135e-06, + "loss": 0.7524, + "step": 49299 + }, + { + "epoch": 3.17, + "learning_rate": 9.442452607398185e-06, + "loss": 0.8036, + "step": 49300 + }, + { + "epoch": 3.17, + "learning_rate": 9.442237918375234e-06, + "loss": 0.8078, + "step": 49301 + }, + { + "epoch": 3.17, + "learning_rate": 9.442023229352283e-06, + "loss": 0.7483, + "step": 49302 + }, + { + "epoch": 3.17, + "learning_rate": 9.441808540329334e-06, + "loss": 0.8341, + "step": 49303 + }, + { + "epoch": 3.17, + "learning_rate": 9.441593851306383e-06, + "loss": 0.887, + "step": 49304 + }, + { + "epoch": 3.17, + "learning_rate": 9.441379162283432e-06, + "loss": 0.8113, + "step": 49305 + }, + { + "epoch": 3.17, + "learning_rate": 9.441164473260483e-06, + "loss": 0.766, + "step": 49306 + }, + { + "epoch": 3.17, + "learning_rate": 9.440949784237533e-06, + "loss": 0.8529, + "step": 49307 + }, + { + "epoch": 3.17, + "learning_rate": 9.440735095214582e-06, + "loss": 0.8117, + "step": 49308 + }, + { + "epoch": 3.17, + "learning_rate": 9.440520406191633e-06, + "loss": 0.8625, + "step": 49309 + }, + { + "epoch": 3.17, + "learning_rate": 9.440305717168682e-06, + "loss": 0.8069, + "step": 49310 + }, + { + "epoch": 3.17, + "learning_rate": 9.440091028145731e-06, + "loss": 0.8225, + "step": 49311 + }, + { + "epoch": 3.17, + "learning_rate": 9.439876339122782e-06, + "loss": 0.8387, + "step": 49312 + }, + { + "epoch": 3.17, + "learning_rate": 9.439661650099832e-06, + "loss": 0.7948, + "step": 49313 + }, + { + "epoch": 3.17, + "learning_rate": 9.439446961076881e-06, + "loss": 0.865, + "step": 49314 + }, + { + "epoch": 3.17, + "learning_rate": 9.43923227205393e-06, + "loss": 0.8112, + "step": 49315 + }, + { + "epoch": 3.17, + "learning_rate": 9.43901758303098e-06, + "loss": 0.7996, + "step": 49316 + }, + { + "epoch": 3.17, + "learning_rate": 9.43880289400803e-06, + "loss": 0.8037, + "step": 49317 + }, + { + "epoch": 3.17, + "learning_rate": 9.43858820498508e-06, + "loss": 0.8633, + "step": 49318 + }, + { + "epoch": 3.17, + "learning_rate": 9.43837351596213e-06, + "loss": 0.8027, + "step": 49319 + }, + { + "epoch": 3.17, + "learning_rate": 9.438158826939178e-06, + "loss": 0.8401, + "step": 49320 + }, + { + "epoch": 3.17, + "learning_rate": 9.437944137916229e-06, + "loss": 0.8294, + "step": 49321 + }, + { + "epoch": 3.17, + "learning_rate": 9.43772944889328e-06, + "loss": 0.7948, + "step": 49322 + }, + { + "epoch": 3.17, + "learning_rate": 9.437514759870329e-06, + "loss": 0.8286, + "step": 49323 + }, + { + "epoch": 3.17, + "learning_rate": 9.437300070847378e-06, + "loss": 0.8004, + "step": 49324 + }, + { + "epoch": 3.17, + "learning_rate": 9.437085381824428e-06, + "loss": 0.8015, + "step": 49325 + }, + { + "epoch": 3.17, + "learning_rate": 9.436870692801477e-06, + "loss": 0.823, + "step": 49326 + }, + { + "epoch": 3.17, + "learning_rate": 9.436656003778528e-06, + "loss": 0.8097, + "step": 49327 + }, + { + "epoch": 3.17, + "learning_rate": 9.436441314755579e-06, + "loss": 0.8644, + "step": 49328 + }, + { + "epoch": 3.17, + "learning_rate": 9.436226625732627e-06, + "loss": 0.8563, + "step": 49329 + }, + { + "epoch": 3.17, + "learning_rate": 9.436011936709676e-06, + "loss": 0.8623, + "step": 49330 + }, + { + "epoch": 3.17, + "learning_rate": 9.435797247686727e-06, + "loss": 0.8389, + "step": 49331 + }, + { + "epoch": 3.17, + "learning_rate": 9.435582558663776e-06, + "loss": 0.8709, + "step": 49332 + }, + { + "epoch": 3.17, + "learning_rate": 9.435367869640825e-06, + "loss": 0.8587, + "step": 49333 + }, + { + "epoch": 3.17, + "learning_rate": 9.435153180617876e-06, + "loss": 0.8573, + "step": 49334 + }, + { + "epoch": 3.17, + "learning_rate": 9.434938491594926e-06, + "loss": 0.816, + "step": 49335 + }, + { + "epoch": 3.17, + "learning_rate": 9.434723802571975e-06, + "loss": 0.9229, + "step": 49336 + }, + { + "epoch": 3.17, + "learning_rate": 9.434509113549026e-06, + "loss": 0.8555, + "step": 49337 + }, + { + "epoch": 3.17, + "learning_rate": 9.434294424526075e-06, + "loss": 0.7925, + "step": 49338 + }, + { + "epoch": 3.17, + "learning_rate": 9.434079735503124e-06, + "loss": 0.8715, + "step": 49339 + }, + { + "epoch": 3.17, + "learning_rate": 9.433865046480175e-06, + "loss": 0.8036, + "step": 49340 + }, + { + "epoch": 3.17, + "learning_rate": 9.433650357457224e-06, + "loss": 0.7984, + "step": 49341 + }, + { + "epoch": 3.17, + "learning_rate": 9.433435668434274e-06, + "loss": 0.8373, + "step": 49342 + }, + { + "epoch": 3.17, + "learning_rate": 9.433220979411323e-06, + "loss": 0.8262, + "step": 49343 + }, + { + "epoch": 3.17, + "learning_rate": 9.433006290388374e-06, + "loss": 0.8559, + "step": 49344 + }, + { + "epoch": 3.17, + "learning_rate": 9.432791601365423e-06, + "loss": 0.8059, + "step": 49345 + }, + { + "epoch": 3.17, + "learning_rate": 9.432576912342473e-06, + "loss": 0.8507, + "step": 49346 + }, + { + "epoch": 3.17, + "learning_rate": 9.432362223319522e-06, + "loss": 0.7898, + "step": 49347 + }, + { + "epoch": 3.17, + "learning_rate": 9.432147534296571e-06, + "loss": 0.8097, + "step": 49348 + }, + { + "epoch": 3.17, + "learning_rate": 9.431932845273622e-06, + "loss": 0.8787, + "step": 49349 + }, + { + "epoch": 3.18, + "learning_rate": 9.431718156250673e-06, + "loss": 0.8375, + "step": 49350 + }, + { + "epoch": 3.18, + "learning_rate": 9.431503467227722e-06, + "loss": 0.8258, + "step": 49351 + }, + { + "epoch": 3.18, + "learning_rate": 9.43128877820477e-06, + "loss": 0.8507, + "step": 49352 + }, + { + "epoch": 3.18, + "learning_rate": 9.431074089181821e-06, + "loss": 0.7504, + "step": 49353 + }, + { + "epoch": 3.18, + "learning_rate": 9.43085940015887e-06, + "loss": 0.7815, + "step": 49354 + }, + { + "epoch": 3.18, + "learning_rate": 9.430644711135921e-06, + "loss": 0.8, + "step": 49355 + }, + { + "epoch": 3.18, + "learning_rate": 9.43043002211297e-06, + "loss": 0.804, + "step": 49356 + }, + { + "epoch": 3.18, + "learning_rate": 9.43021533309002e-06, + "loss": 0.856, + "step": 49357 + }, + { + "epoch": 3.18, + "learning_rate": 9.43000064406707e-06, + "loss": 0.8372, + "step": 49358 + }, + { + "epoch": 3.18, + "learning_rate": 9.42978595504412e-06, + "loss": 0.8485, + "step": 49359 + }, + { + "epoch": 3.18, + "learning_rate": 9.429571266021169e-06, + "loss": 0.8935, + "step": 49360 + }, + { + "epoch": 3.18, + "learning_rate": 9.429356576998218e-06, + "loss": 0.8456, + "step": 49361 + }, + { + "epoch": 3.18, + "learning_rate": 9.429141887975269e-06, + "loss": 0.8956, + "step": 49362 + }, + { + "epoch": 3.18, + "learning_rate": 9.428927198952318e-06, + "loss": 0.8618, + "step": 49363 + }, + { + "epoch": 3.18, + "learning_rate": 9.428712509929368e-06, + "loss": 0.7593, + "step": 49364 + }, + { + "epoch": 3.18, + "learning_rate": 9.428497820906419e-06, + "loss": 0.7475, + "step": 49365 + }, + { + "epoch": 3.18, + "learning_rate": 9.428283131883468e-06, + "loss": 0.8463, + "step": 49366 + }, + { + "epoch": 3.18, + "learning_rate": 9.428068442860517e-06, + "loss": 0.8225, + "step": 49367 + }, + { + "epoch": 3.18, + "learning_rate": 9.427853753837568e-06, + "loss": 0.8648, + "step": 49368 + }, + { + "epoch": 3.18, + "learning_rate": 9.427639064814617e-06, + "loss": 0.8745, + "step": 49369 + }, + { + "epoch": 3.18, + "learning_rate": 9.427424375791666e-06, + "loss": 0.8206, + "step": 49370 + }, + { + "epoch": 3.18, + "learning_rate": 9.427209686768716e-06, + "loss": 0.7999, + "step": 49371 + }, + { + "epoch": 3.18, + "learning_rate": 9.426994997745767e-06, + "loss": 0.8137, + "step": 49372 + }, + { + "epoch": 3.18, + "learning_rate": 9.426780308722816e-06, + "loss": 0.8368, + "step": 49373 + }, + { + "epoch": 3.18, + "learning_rate": 9.426565619699865e-06, + "loss": 0.7691, + "step": 49374 + }, + { + "epoch": 3.18, + "learning_rate": 9.426350930676915e-06, + "loss": 0.865, + "step": 49375 + }, + { + "epoch": 3.18, + "learning_rate": 9.426136241653964e-06, + "loss": 0.7814, + "step": 49376 + }, + { + "epoch": 3.18, + "learning_rate": 9.425921552631015e-06, + "loss": 0.8404, + "step": 49377 + }, + { + "epoch": 3.18, + "learning_rate": 9.425706863608064e-06, + "loss": 0.8451, + "step": 49378 + }, + { + "epoch": 3.18, + "learning_rate": 9.425492174585115e-06, + "loss": 0.7931, + "step": 49379 + }, + { + "epoch": 3.18, + "learning_rate": 9.425277485562164e-06, + "loss": 0.7919, + "step": 49380 + }, + { + "epoch": 3.18, + "learning_rate": 9.425062796539214e-06, + "loss": 0.7832, + "step": 49381 + }, + { + "epoch": 3.18, + "learning_rate": 9.424848107516263e-06, + "loss": 0.7753, + "step": 49382 + }, + { + "epoch": 3.18, + "learning_rate": 9.424633418493312e-06, + "loss": 0.8751, + "step": 49383 + }, + { + "epoch": 3.18, + "learning_rate": 9.424418729470363e-06, + "loss": 0.8041, + "step": 49384 + }, + { + "epoch": 3.18, + "learning_rate": 9.424204040447412e-06, + "loss": 0.8232, + "step": 49385 + }, + { + "epoch": 3.18, + "learning_rate": 9.423989351424463e-06, + "loss": 0.7722, + "step": 49386 + }, + { + "epoch": 3.18, + "learning_rate": 9.423774662401513e-06, + "loss": 0.8068, + "step": 49387 + }, + { + "epoch": 3.18, + "learning_rate": 9.423559973378562e-06, + "loss": 0.8091, + "step": 49388 + }, + { + "epoch": 3.18, + "learning_rate": 9.423345284355611e-06, + "loss": 0.8028, + "step": 49389 + }, + { + "epoch": 3.18, + "learning_rate": 9.423130595332662e-06, + "loss": 0.8048, + "step": 49390 + }, + { + "epoch": 3.18, + "learning_rate": 9.42291590630971e-06, + "loss": 0.8701, + "step": 49391 + }, + { + "epoch": 3.18, + "learning_rate": 9.42270121728676e-06, + "loss": 0.7897, + "step": 49392 + }, + { + "epoch": 3.18, + "learning_rate": 9.42248652826381e-06, + "loss": 0.8784, + "step": 49393 + }, + { + "epoch": 3.18, + "learning_rate": 9.422271839240861e-06, + "loss": 0.878, + "step": 49394 + }, + { + "epoch": 3.18, + "learning_rate": 9.42205715021791e-06, + "loss": 0.8545, + "step": 49395 + }, + { + "epoch": 3.18, + "learning_rate": 9.42184246119496e-06, + "loss": 0.8458, + "step": 49396 + }, + { + "epoch": 3.18, + "learning_rate": 9.42162777217201e-06, + "loss": 0.8364, + "step": 49397 + }, + { + "epoch": 3.18, + "learning_rate": 9.421413083149059e-06, + "loss": 0.8459, + "step": 49398 + }, + { + "epoch": 3.18, + "learning_rate": 9.42119839412611e-06, + "loss": 0.8551, + "step": 49399 + }, + { + "epoch": 3.18, + "learning_rate": 9.420983705103158e-06, + "loss": 0.7784, + "step": 49400 + }, + { + "epoch": 3.18, + "learning_rate": 9.420769016080209e-06, + "loss": 0.7882, + "step": 49401 + }, + { + "epoch": 3.18, + "learning_rate": 9.420554327057258e-06, + "loss": 0.8777, + "step": 49402 + }, + { + "epoch": 3.18, + "learning_rate": 9.420339638034308e-06, + "loss": 0.8788, + "step": 49403 + }, + { + "epoch": 3.18, + "learning_rate": 9.420124949011357e-06, + "loss": 0.835, + "step": 49404 + }, + { + "epoch": 3.18, + "learning_rate": 9.419910259988408e-06, + "loss": 0.8064, + "step": 49405 + }, + { + "epoch": 3.18, + "learning_rate": 9.419695570965457e-06, + "loss": 0.7864, + "step": 49406 + }, + { + "epoch": 3.18, + "learning_rate": 9.419480881942506e-06, + "loss": 0.871, + "step": 49407 + }, + { + "epoch": 3.18, + "learning_rate": 9.419266192919557e-06, + "loss": 0.8478, + "step": 49408 + }, + { + "epoch": 3.18, + "learning_rate": 9.419051503896607e-06, + "loss": 0.7758, + "step": 49409 + }, + { + "epoch": 3.18, + "learning_rate": 9.418836814873656e-06, + "loss": 0.858, + "step": 49410 + }, + { + "epoch": 3.18, + "learning_rate": 9.418622125850705e-06, + "loss": 0.8248, + "step": 49411 + }, + { + "epoch": 3.18, + "learning_rate": 9.418407436827756e-06, + "loss": 0.7427, + "step": 49412 + }, + { + "epoch": 3.18, + "learning_rate": 9.418192747804805e-06, + "loss": 0.8026, + "step": 49413 + }, + { + "epoch": 3.18, + "learning_rate": 9.417978058781856e-06, + "loss": 0.7781, + "step": 49414 + }, + { + "epoch": 3.18, + "learning_rate": 9.417763369758906e-06, + "loss": 0.7994, + "step": 49415 + }, + { + "epoch": 3.18, + "learning_rate": 9.417548680735955e-06, + "loss": 0.8322, + "step": 49416 + }, + { + "epoch": 3.18, + "learning_rate": 9.417333991713004e-06, + "loss": 0.7575, + "step": 49417 + }, + { + "epoch": 3.18, + "learning_rate": 9.417119302690055e-06, + "loss": 0.787, + "step": 49418 + }, + { + "epoch": 3.18, + "learning_rate": 9.416904613667104e-06, + "loss": 0.7474, + "step": 49419 + }, + { + "epoch": 3.18, + "learning_rate": 9.416689924644153e-06, + "loss": 0.8622, + "step": 49420 + }, + { + "epoch": 3.18, + "learning_rate": 9.416475235621203e-06, + "loss": 0.8891, + "step": 49421 + }, + { + "epoch": 3.18, + "learning_rate": 9.416260546598254e-06, + "loss": 0.8283, + "step": 49422 + }, + { + "epoch": 3.18, + "learning_rate": 9.416045857575303e-06, + "loss": 0.8254, + "step": 49423 + }, + { + "epoch": 3.18, + "learning_rate": 9.415831168552354e-06, + "loss": 0.8319, + "step": 49424 + }, + { + "epoch": 3.18, + "learning_rate": 9.415616479529403e-06, + "loss": 0.8066, + "step": 49425 + }, + { + "epoch": 3.18, + "learning_rate": 9.415401790506452e-06, + "loss": 0.8719, + "step": 49426 + }, + { + "epoch": 3.18, + "learning_rate": 9.415187101483502e-06, + "loss": 0.8298, + "step": 49427 + }, + { + "epoch": 3.18, + "learning_rate": 9.414972412460551e-06, + "loss": 0.8493, + "step": 49428 + }, + { + "epoch": 3.18, + "learning_rate": 9.4147577234376e-06, + "loss": 0.7945, + "step": 49429 + }, + { + "epoch": 3.18, + "learning_rate": 9.414543034414651e-06, + "loss": 0.853, + "step": 49430 + }, + { + "epoch": 3.18, + "learning_rate": 9.414328345391702e-06, + "loss": 0.7881, + "step": 49431 + }, + { + "epoch": 3.18, + "learning_rate": 9.41411365636875e-06, + "loss": 0.7877, + "step": 49432 + }, + { + "epoch": 3.18, + "learning_rate": 9.413898967345801e-06, + "loss": 0.8214, + "step": 49433 + }, + { + "epoch": 3.18, + "learning_rate": 9.41368427832285e-06, + "loss": 0.8206, + "step": 49434 + }, + { + "epoch": 3.18, + "learning_rate": 9.413469589299899e-06, + "loss": 0.8361, + "step": 49435 + }, + { + "epoch": 3.18, + "learning_rate": 9.41325490027695e-06, + "loss": 0.7179, + "step": 49436 + }, + { + "epoch": 3.18, + "learning_rate": 9.413040211254e-06, + "loss": 0.7648, + "step": 49437 + }, + { + "epoch": 3.18, + "learning_rate": 9.41282552223105e-06, + "loss": 0.8198, + "step": 49438 + }, + { + "epoch": 3.18, + "learning_rate": 9.412610833208098e-06, + "loss": 0.8354, + "step": 49439 + }, + { + "epoch": 3.18, + "learning_rate": 9.412396144185149e-06, + "loss": 0.8381, + "step": 49440 + }, + { + "epoch": 3.18, + "learning_rate": 9.412181455162198e-06, + "loss": 0.8115, + "step": 49441 + }, + { + "epoch": 3.18, + "learning_rate": 9.411966766139249e-06, + "loss": 0.7827, + "step": 49442 + }, + { + "epoch": 3.18, + "learning_rate": 9.411752077116298e-06, + "loss": 0.8401, + "step": 49443 + }, + { + "epoch": 3.18, + "learning_rate": 9.411537388093348e-06, + "loss": 0.8078, + "step": 49444 + }, + { + "epoch": 3.18, + "learning_rate": 9.411322699070397e-06, + "loss": 0.8487, + "step": 49445 + }, + { + "epoch": 3.18, + "learning_rate": 9.411108010047448e-06, + "loss": 0.8221, + "step": 49446 + }, + { + "epoch": 3.18, + "learning_rate": 9.410893321024497e-06, + "loss": 0.8681, + "step": 49447 + }, + { + "epoch": 3.18, + "learning_rate": 9.410678632001546e-06, + "loss": 0.7034, + "step": 49448 + }, + { + "epoch": 3.18, + "learning_rate": 9.410463942978596e-06, + "loss": 0.7913, + "step": 49449 + }, + { + "epoch": 3.18, + "learning_rate": 9.410249253955645e-06, + "loss": 0.8107, + "step": 49450 + }, + { + "epoch": 3.18, + "learning_rate": 9.410034564932696e-06, + "loss": 0.8334, + "step": 49451 + }, + { + "epoch": 3.18, + "learning_rate": 9.409819875909747e-06, + "loss": 0.7832, + "step": 49452 + }, + { + "epoch": 3.18, + "learning_rate": 9.409605186886796e-06, + "loss": 0.8674, + "step": 49453 + }, + { + "epoch": 3.18, + "learning_rate": 9.409390497863845e-06, + "loss": 0.8126, + "step": 49454 + }, + { + "epoch": 3.18, + "learning_rate": 9.409175808840895e-06, + "loss": 0.8316, + "step": 49455 + }, + { + "epoch": 3.18, + "learning_rate": 9.408961119817944e-06, + "loss": 0.8886, + "step": 49456 + }, + { + "epoch": 3.18, + "learning_rate": 9.408746430794993e-06, + "loss": 0.846, + "step": 49457 + }, + { + "epoch": 3.18, + "learning_rate": 9.408531741772044e-06, + "loss": 0.8317, + "step": 49458 + }, + { + "epoch": 3.18, + "learning_rate": 9.408317052749095e-06, + "loss": 0.7607, + "step": 49459 + }, + { + "epoch": 3.18, + "learning_rate": 9.408102363726144e-06, + "loss": 0.8187, + "step": 49460 + }, + { + "epoch": 3.18, + "learning_rate": 9.407887674703192e-06, + "loss": 0.8628, + "step": 49461 + }, + { + "epoch": 3.18, + "learning_rate": 9.407672985680243e-06, + "loss": 0.8637, + "step": 49462 + }, + { + "epoch": 3.18, + "learning_rate": 9.407458296657292e-06, + "loss": 0.7804, + "step": 49463 + }, + { + "epoch": 3.18, + "learning_rate": 9.407243607634343e-06, + "loss": 0.7818, + "step": 49464 + }, + { + "epoch": 3.18, + "learning_rate": 9.407028918611392e-06, + "loss": 0.7809, + "step": 49465 + }, + { + "epoch": 3.18, + "learning_rate": 9.406814229588442e-06, + "loss": 0.817, + "step": 49466 + }, + { + "epoch": 3.18, + "learning_rate": 9.406599540565491e-06, + "loss": 0.8305, + "step": 49467 + }, + { + "epoch": 3.18, + "learning_rate": 9.406384851542542e-06, + "loss": 0.8738, + "step": 49468 + }, + { + "epoch": 3.18, + "learning_rate": 9.406170162519591e-06, + "loss": 0.8132, + "step": 49469 + }, + { + "epoch": 3.18, + "learning_rate": 9.40595547349664e-06, + "loss": 0.8585, + "step": 49470 + }, + { + "epoch": 3.18, + "learning_rate": 9.40574078447369e-06, + "loss": 0.7678, + "step": 49471 + }, + { + "epoch": 3.18, + "learning_rate": 9.40552609545074e-06, + "loss": 0.8049, + "step": 49472 + }, + { + "epoch": 3.18, + "learning_rate": 9.40531140642779e-06, + "loss": 0.8781, + "step": 49473 + }, + { + "epoch": 3.18, + "learning_rate": 9.405096717404841e-06, + "loss": 0.7865, + "step": 49474 + }, + { + "epoch": 3.18, + "learning_rate": 9.40488202838189e-06, + "loss": 0.8495, + "step": 49475 + }, + { + "epoch": 3.18, + "learning_rate": 9.404667339358939e-06, + "loss": 0.8431, + "step": 49476 + }, + { + "epoch": 3.18, + "learning_rate": 9.40445265033599e-06, + "loss": 0.8299, + "step": 49477 + }, + { + "epoch": 3.18, + "learning_rate": 9.404237961313038e-06, + "loss": 0.8052, + "step": 49478 + }, + { + "epoch": 3.18, + "learning_rate": 9.404023272290087e-06, + "loss": 0.8748, + "step": 49479 + }, + { + "epoch": 3.18, + "learning_rate": 9.403808583267138e-06, + "loss": 0.8288, + "step": 49480 + }, + { + "epoch": 3.18, + "learning_rate": 9.403593894244189e-06, + "loss": 0.8267, + "step": 49481 + }, + { + "epoch": 3.18, + "learning_rate": 9.403379205221238e-06, + "loss": 0.8073, + "step": 49482 + }, + { + "epoch": 3.18, + "learning_rate": 9.403164516198288e-06, + "loss": 0.7996, + "step": 49483 + }, + { + "epoch": 3.18, + "learning_rate": 9.402949827175337e-06, + "loss": 0.8216, + "step": 49484 + }, + { + "epoch": 3.18, + "learning_rate": 9.402735138152386e-06, + "loss": 0.8337, + "step": 49485 + }, + { + "epoch": 3.18, + "learning_rate": 9.402520449129437e-06, + "loss": 0.7578, + "step": 49486 + }, + { + "epoch": 3.18, + "learning_rate": 9.402305760106486e-06, + "loss": 0.817, + "step": 49487 + }, + { + "epoch": 3.18, + "learning_rate": 9.402091071083537e-06, + "loss": 0.7829, + "step": 49488 + }, + { + "epoch": 3.18, + "learning_rate": 9.401876382060586e-06, + "loss": 0.8459, + "step": 49489 + }, + { + "epoch": 3.18, + "learning_rate": 9.401661693037636e-06, + "loss": 0.8105, + "step": 49490 + }, + { + "epoch": 3.18, + "learning_rate": 9.401447004014685e-06, + "loss": 0.8361, + "step": 49491 + }, + { + "epoch": 3.18, + "learning_rate": 9.401232314991736e-06, + "loss": 0.8072, + "step": 49492 + }, + { + "epoch": 3.18, + "learning_rate": 9.401017625968785e-06, + "loss": 0.8213, + "step": 49493 + }, + { + "epoch": 3.18, + "learning_rate": 9.400802936945834e-06, + "loss": 0.8173, + "step": 49494 + }, + { + "epoch": 3.18, + "learning_rate": 9.400588247922884e-06, + "loss": 0.8362, + "step": 49495 + }, + { + "epoch": 3.18, + "learning_rate": 9.400373558899935e-06, + "loss": 0.9019, + "step": 49496 + }, + { + "epoch": 3.18, + "learning_rate": 9.400158869876984e-06, + "loss": 0.8153, + "step": 49497 + }, + { + "epoch": 3.18, + "learning_rate": 9.399944180854033e-06, + "loss": 0.808, + "step": 49498 + }, + { + "epoch": 3.18, + "learning_rate": 9.399729491831084e-06, + "loss": 0.8413, + "step": 49499 + }, + { + "epoch": 3.18, + "learning_rate": 9.399514802808133e-06, + "loss": 0.8109, + "step": 49500 + }, + { + "epoch": 3.18, + "learning_rate": 9.399300113785183e-06, + "loss": 0.8255, + "step": 49501 + }, + { + "epoch": 3.18, + "learning_rate": 9.399085424762234e-06, + "loss": 0.8078, + "step": 49502 + }, + { + "epoch": 3.18, + "learning_rate": 9.398870735739283e-06, + "loss": 0.8007, + "step": 49503 + }, + { + "epoch": 3.18, + "learning_rate": 9.398656046716332e-06, + "loss": 0.8228, + "step": 49504 + }, + { + "epoch": 3.19, + "learning_rate": 9.398441357693383e-06, + "loss": 0.8929, + "step": 49505 + }, + { + "epoch": 3.19, + "learning_rate": 9.398226668670431e-06, + "loss": 0.7752, + "step": 49506 + }, + { + "epoch": 3.19, + "learning_rate": 9.39801197964748e-06, + "loss": 0.8347, + "step": 49507 + }, + { + "epoch": 3.19, + "learning_rate": 9.397797290624531e-06, + "loss": 0.8063, + "step": 49508 + }, + { + "epoch": 3.19, + "learning_rate": 9.397582601601582e-06, + "loss": 0.7924, + "step": 49509 + }, + { + "epoch": 3.19, + "learning_rate": 9.39736791257863e-06, + "loss": 0.8485, + "step": 49510 + }, + { + "epoch": 3.19, + "learning_rate": 9.397153223555681e-06, + "loss": 0.8024, + "step": 49511 + }, + { + "epoch": 3.19, + "learning_rate": 9.39693853453273e-06, + "loss": 0.76, + "step": 49512 + }, + { + "epoch": 3.19, + "learning_rate": 9.39672384550978e-06, + "loss": 0.8572, + "step": 49513 + }, + { + "epoch": 3.19, + "learning_rate": 9.39650915648683e-06, + "loss": 0.8074, + "step": 49514 + }, + { + "epoch": 3.19, + "learning_rate": 9.396294467463879e-06, + "loss": 0.8017, + "step": 49515 + }, + { + "epoch": 3.19, + "learning_rate": 9.396079778440928e-06, + "loss": 0.8121, + "step": 49516 + }, + { + "epoch": 3.19, + "learning_rate": 9.395865089417979e-06, + "loss": 0.8426, + "step": 49517 + }, + { + "epoch": 3.19, + "learning_rate": 9.39565040039503e-06, + "loss": 0.7604, + "step": 49518 + }, + { + "epoch": 3.19, + "learning_rate": 9.395435711372078e-06, + "loss": 0.7765, + "step": 49519 + }, + { + "epoch": 3.19, + "learning_rate": 9.395221022349129e-06, + "loss": 0.7839, + "step": 49520 + }, + { + "epoch": 3.19, + "learning_rate": 9.395006333326178e-06, + "loss": 0.8186, + "step": 49521 + }, + { + "epoch": 3.19, + "learning_rate": 9.394791644303227e-06, + "loss": 0.8009, + "step": 49522 + }, + { + "epoch": 3.19, + "learning_rate": 9.394576955280277e-06, + "loss": 0.8652, + "step": 49523 + }, + { + "epoch": 3.19, + "learning_rate": 9.394362266257328e-06, + "loss": 0.7075, + "step": 49524 + }, + { + "epoch": 3.19, + "learning_rate": 9.394147577234377e-06, + "loss": 0.8512, + "step": 49525 + }, + { + "epoch": 3.19, + "learning_rate": 9.393932888211426e-06, + "loss": 0.7521, + "step": 49526 + }, + { + "epoch": 3.19, + "learning_rate": 9.393718199188477e-06, + "loss": 0.7946, + "step": 49527 + }, + { + "epoch": 3.19, + "learning_rate": 9.393503510165526e-06, + "loss": 0.8836, + "step": 49528 + }, + { + "epoch": 3.19, + "learning_rate": 9.393288821142575e-06, + "loss": 0.826, + "step": 49529 + }, + { + "epoch": 3.19, + "learning_rate": 9.393074132119625e-06, + "loss": 0.9, + "step": 49530 + }, + { + "epoch": 3.19, + "learning_rate": 9.392859443096676e-06, + "loss": 0.8625, + "step": 49531 + }, + { + "epoch": 3.19, + "learning_rate": 9.392644754073725e-06, + "loss": 0.799, + "step": 49532 + }, + { + "epoch": 3.19, + "learning_rate": 9.392430065050776e-06, + "loss": 0.8296, + "step": 49533 + }, + { + "epoch": 3.19, + "learning_rate": 9.392215376027825e-06, + "loss": 0.8042, + "step": 49534 + }, + { + "epoch": 3.19, + "learning_rate": 9.392000687004873e-06, + "loss": 0.8254, + "step": 49535 + }, + { + "epoch": 3.19, + "learning_rate": 9.391785997981924e-06, + "loss": 0.8217, + "step": 49536 + }, + { + "epoch": 3.19, + "learning_rate": 9.391571308958973e-06, + "loss": 0.8201, + "step": 49537 + }, + { + "epoch": 3.19, + "learning_rate": 9.391356619936024e-06, + "loss": 0.7801, + "step": 49538 + }, + { + "epoch": 3.19, + "learning_rate": 9.391141930913073e-06, + "loss": 0.8463, + "step": 49539 + }, + { + "epoch": 3.19, + "learning_rate": 9.390927241890123e-06, + "loss": 0.8241, + "step": 49540 + }, + { + "epoch": 3.19, + "learning_rate": 9.390712552867172e-06, + "loss": 0.7907, + "step": 49541 + }, + { + "epoch": 3.19, + "learning_rate": 9.390497863844223e-06, + "loss": 0.8341, + "step": 49542 + }, + { + "epoch": 3.19, + "learning_rate": 9.390283174821272e-06, + "loss": 0.8753, + "step": 49543 + }, + { + "epoch": 3.19, + "learning_rate": 9.390068485798321e-06, + "loss": 0.8439, + "step": 49544 + }, + { + "epoch": 3.19, + "learning_rate": 9.389853796775372e-06, + "loss": 0.8273, + "step": 49545 + }, + { + "epoch": 3.19, + "learning_rate": 9.389639107752422e-06, + "loss": 0.8659, + "step": 49546 + }, + { + "epoch": 3.19, + "learning_rate": 9.389424418729471e-06, + "loss": 0.8345, + "step": 49547 + }, + { + "epoch": 3.19, + "learning_rate": 9.38920972970652e-06, + "loss": 0.7813, + "step": 49548 + }, + { + "epoch": 3.19, + "learning_rate": 9.388995040683571e-06, + "loss": 0.8469, + "step": 49549 + }, + { + "epoch": 3.19, + "learning_rate": 9.38878035166062e-06, + "loss": 0.806, + "step": 49550 + }, + { + "epoch": 3.19, + "learning_rate": 9.38856566263767e-06, + "loss": 0.9139, + "step": 49551 + }, + { + "epoch": 3.19, + "learning_rate": 9.38835097361472e-06, + "loss": 0.7972, + "step": 49552 + }, + { + "epoch": 3.19, + "learning_rate": 9.38813628459177e-06, + "loss": 0.7724, + "step": 49553 + }, + { + "epoch": 3.19, + "learning_rate": 9.387921595568819e-06, + "loss": 0.7547, + "step": 49554 + }, + { + "epoch": 3.19, + "learning_rate": 9.38770690654587e-06, + "loss": 0.8391, + "step": 49555 + }, + { + "epoch": 3.19, + "learning_rate": 9.387492217522919e-06, + "loss": 0.8059, + "step": 49556 + }, + { + "epoch": 3.19, + "learning_rate": 9.387277528499968e-06, + "loss": 0.7996, + "step": 49557 + }, + { + "epoch": 3.19, + "learning_rate": 9.387062839477018e-06, + "loss": 0.8333, + "step": 49558 + }, + { + "epoch": 3.19, + "learning_rate": 9.386848150454067e-06, + "loss": 0.8615, + "step": 49559 + }, + { + "epoch": 3.19, + "learning_rate": 9.386633461431118e-06, + "loss": 0.8652, + "step": 49560 + }, + { + "epoch": 3.19, + "learning_rate": 9.386418772408169e-06, + "loss": 0.8701, + "step": 49561 + }, + { + "epoch": 3.19, + "learning_rate": 9.386204083385218e-06, + "loss": 0.8556, + "step": 49562 + }, + { + "epoch": 3.19, + "learning_rate": 9.385989394362267e-06, + "loss": 0.8635, + "step": 49563 + }, + { + "epoch": 3.19, + "learning_rate": 9.385774705339317e-06, + "loss": 0.7885, + "step": 49564 + }, + { + "epoch": 3.19, + "learning_rate": 9.385560016316366e-06, + "loss": 0.8244, + "step": 49565 + }, + { + "epoch": 3.19, + "learning_rate": 9.385345327293415e-06, + "loss": 0.8261, + "step": 49566 + }, + { + "epoch": 3.19, + "learning_rate": 9.385130638270466e-06, + "loss": 0.7333, + "step": 49567 + }, + { + "epoch": 3.19, + "learning_rate": 9.384915949247516e-06, + "loss": 0.8311, + "step": 49568 + }, + { + "epoch": 3.19, + "learning_rate": 9.384701260224565e-06, + "loss": 0.8713, + "step": 49569 + }, + { + "epoch": 3.19, + "learning_rate": 9.384486571201616e-06, + "loss": 0.8085, + "step": 49570 + }, + { + "epoch": 3.19, + "learning_rate": 9.384271882178665e-06, + "loss": 0.8351, + "step": 49571 + }, + { + "epoch": 3.19, + "learning_rate": 9.384057193155714e-06, + "loss": 0.8439, + "step": 49572 + }, + { + "epoch": 3.19, + "learning_rate": 9.383842504132765e-06, + "loss": 0.793, + "step": 49573 + }, + { + "epoch": 3.19, + "learning_rate": 9.383627815109814e-06, + "loss": 0.7799, + "step": 49574 + }, + { + "epoch": 3.19, + "learning_rate": 9.383413126086864e-06, + "loss": 0.8359, + "step": 49575 + }, + { + "epoch": 3.19, + "learning_rate": 9.383198437063913e-06, + "loss": 0.8769, + "step": 49576 + }, + { + "epoch": 3.19, + "learning_rate": 9.382983748040964e-06, + "loss": 0.7796, + "step": 49577 + }, + { + "epoch": 3.19, + "learning_rate": 9.382769059018013e-06, + "loss": 0.8166, + "step": 49578 + }, + { + "epoch": 3.19, + "learning_rate": 9.382554369995064e-06, + "loss": 0.8757, + "step": 49579 + }, + { + "epoch": 3.19, + "learning_rate": 9.382339680972112e-06, + "loss": 0.7614, + "step": 49580 + }, + { + "epoch": 3.19, + "learning_rate": 9.382124991949161e-06, + "loss": 0.8319, + "step": 49581 + }, + { + "epoch": 3.19, + "learning_rate": 9.381910302926212e-06, + "loss": 0.8194, + "step": 49582 + }, + { + "epoch": 3.19, + "learning_rate": 9.381695613903263e-06, + "loss": 0.8183, + "step": 49583 + }, + { + "epoch": 3.19, + "learning_rate": 9.381480924880312e-06, + "loss": 0.8266, + "step": 49584 + }, + { + "epoch": 3.19, + "learning_rate": 9.38126623585736e-06, + "loss": 0.7997, + "step": 49585 + }, + { + "epoch": 3.19, + "learning_rate": 9.381051546834411e-06, + "loss": 0.787, + "step": 49586 + }, + { + "epoch": 3.19, + "learning_rate": 9.38083685781146e-06, + "loss": 0.8224, + "step": 49587 + }, + { + "epoch": 3.19, + "learning_rate": 9.380622168788511e-06, + "loss": 0.8156, + "step": 49588 + }, + { + "epoch": 3.19, + "learning_rate": 9.380407479765562e-06, + "loss": 0.8164, + "step": 49589 + }, + { + "epoch": 3.19, + "learning_rate": 9.38019279074261e-06, + "loss": 0.8235, + "step": 49590 + }, + { + "epoch": 3.19, + "learning_rate": 9.37997810171966e-06, + "loss": 0.8209, + "step": 49591 + }, + { + "epoch": 3.19, + "learning_rate": 9.37976341269671e-06, + "loss": 0.8469, + "step": 49592 + }, + { + "epoch": 3.19, + "learning_rate": 9.37954872367376e-06, + "loss": 0.8757, + "step": 49593 + }, + { + "epoch": 3.19, + "learning_rate": 9.379334034650808e-06, + "loss": 0.7877, + "step": 49594 + }, + { + "epoch": 3.19, + "learning_rate": 9.379119345627859e-06, + "loss": 0.886, + "step": 49595 + }, + { + "epoch": 3.19, + "learning_rate": 9.378904656604908e-06, + "loss": 0.7733, + "step": 49596 + }, + { + "epoch": 3.19, + "learning_rate": 9.378689967581958e-06, + "loss": 0.8007, + "step": 49597 + }, + { + "epoch": 3.19, + "learning_rate": 9.378475278559009e-06, + "loss": 0.9181, + "step": 49598 + }, + { + "epoch": 3.19, + "learning_rate": 9.378260589536058e-06, + "loss": 0.7762, + "step": 49599 + }, + { + "epoch": 3.19, + "learning_rate": 9.378045900513107e-06, + "loss": 0.8214, + "step": 49600 + }, + { + "epoch": 3.19, + "learning_rate": 9.377831211490158e-06, + "loss": 0.8108, + "step": 49601 + }, + { + "epoch": 3.19, + "learning_rate": 9.377616522467207e-06, + "loss": 0.7874, + "step": 49602 + }, + { + "epoch": 3.19, + "learning_rate": 9.377401833444256e-06, + "loss": 0.7499, + "step": 49603 + }, + { + "epoch": 3.19, + "learning_rate": 9.377187144421306e-06, + "loss": 0.7636, + "step": 49604 + }, + { + "epoch": 3.19, + "learning_rate": 9.376972455398357e-06, + "loss": 0.783, + "step": 49605 + }, + { + "epoch": 3.19, + "learning_rate": 9.376757766375406e-06, + "loss": 0.8016, + "step": 49606 + }, + { + "epoch": 3.19, + "learning_rate": 9.376543077352457e-06, + "loss": 0.864, + "step": 49607 + }, + { + "epoch": 3.19, + "learning_rate": 9.376328388329506e-06, + "loss": 0.8292, + "step": 49608 + }, + { + "epoch": 3.19, + "learning_rate": 9.376113699306554e-06, + "loss": 0.792, + "step": 49609 + }, + { + "epoch": 3.19, + "learning_rate": 9.375899010283605e-06, + "loss": 0.8648, + "step": 49610 + }, + { + "epoch": 3.19, + "learning_rate": 9.375684321260656e-06, + "loss": 0.7747, + "step": 49611 + }, + { + "epoch": 3.19, + "learning_rate": 9.375469632237705e-06, + "loss": 0.8255, + "step": 49612 + }, + { + "epoch": 3.19, + "learning_rate": 9.375254943214754e-06, + "loss": 0.8412, + "step": 49613 + }, + { + "epoch": 3.19, + "learning_rate": 9.375040254191804e-06, + "loss": 0.8011, + "step": 49614 + }, + { + "epoch": 3.19, + "learning_rate": 9.374825565168853e-06, + "loss": 0.7975, + "step": 49615 + }, + { + "epoch": 3.19, + "learning_rate": 9.374610876145902e-06, + "loss": 0.8031, + "step": 49616 + }, + { + "epoch": 3.19, + "learning_rate": 9.374396187122953e-06, + "loss": 0.8094, + "step": 49617 + }, + { + "epoch": 3.19, + "learning_rate": 9.374181498100004e-06, + "loss": 0.8515, + "step": 49618 + }, + { + "epoch": 3.19, + "learning_rate": 9.373966809077053e-06, + "loss": 0.9247, + "step": 49619 + }, + { + "epoch": 3.19, + "learning_rate": 9.373752120054103e-06, + "loss": 0.8134, + "step": 49620 + }, + { + "epoch": 3.19, + "learning_rate": 9.373537431031152e-06, + "loss": 0.8754, + "step": 49621 + }, + { + "epoch": 3.19, + "learning_rate": 9.373322742008201e-06, + "loss": 0.8132, + "step": 49622 + }, + { + "epoch": 3.19, + "learning_rate": 9.373108052985252e-06, + "loss": 0.8455, + "step": 49623 + }, + { + "epoch": 3.19, + "learning_rate": 9.3728933639623e-06, + "loss": 0.7341, + "step": 49624 + }, + { + "epoch": 3.19, + "learning_rate": 9.37267867493935e-06, + "loss": 0.8416, + "step": 49625 + }, + { + "epoch": 3.19, + "learning_rate": 9.3724639859164e-06, + "loss": 0.7646, + "step": 49626 + }, + { + "epoch": 3.19, + "learning_rate": 9.372249296893451e-06, + "loss": 0.8322, + "step": 49627 + }, + { + "epoch": 3.19, + "learning_rate": 9.3720346078705e-06, + "loss": 0.8161, + "step": 49628 + }, + { + "epoch": 3.19, + "learning_rate": 9.37181991884755e-06, + "loss": 0.8158, + "step": 49629 + }, + { + "epoch": 3.19, + "learning_rate": 9.3716052298246e-06, + "loss": 0.8246, + "step": 49630 + }, + { + "epoch": 3.19, + "learning_rate": 9.371390540801649e-06, + "loss": 0.7735, + "step": 49631 + }, + { + "epoch": 3.19, + "learning_rate": 9.3711758517787e-06, + "loss": 0.8878, + "step": 49632 + }, + { + "epoch": 3.19, + "learning_rate": 9.37096116275575e-06, + "loss": 0.8553, + "step": 49633 + }, + { + "epoch": 3.19, + "learning_rate": 9.370746473732799e-06, + "loss": 0.7754, + "step": 49634 + }, + { + "epoch": 3.19, + "learning_rate": 9.370531784709848e-06, + "loss": 0.7604, + "step": 49635 + }, + { + "epoch": 3.19, + "learning_rate": 9.370317095686899e-06, + "loss": 0.8004, + "step": 49636 + }, + { + "epoch": 3.19, + "learning_rate": 9.370102406663948e-06, + "loss": 0.7843, + "step": 49637 + }, + { + "epoch": 3.19, + "learning_rate": 9.369887717640998e-06, + "loss": 0.8163, + "step": 49638 + }, + { + "epoch": 3.19, + "learning_rate": 9.369673028618047e-06, + "loss": 0.8039, + "step": 49639 + }, + { + "epoch": 3.19, + "learning_rate": 9.369458339595098e-06, + "loss": 0.704, + "step": 49640 + }, + { + "epoch": 3.19, + "learning_rate": 9.369243650572147e-06, + "loss": 0.7888, + "step": 49641 + }, + { + "epoch": 3.19, + "learning_rate": 9.369028961549197e-06, + "loss": 0.7942, + "step": 49642 + }, + { + "epoch": 3.19, + "learning_rate": 9.368814272526246e-06, + "loss": 0.8366, + "step": 49643 + }, + { + "epoch": 3.19, + "learning_rate": 9.368599583503295e-06, + "loss": 0.8347, + "step": 49644 + }, + { + "epoch": 3.19, + "learning_rate": 9.368384894480346e-06, + "loss": 0.9247, + "step": 49645 + }, + { + "epoch": 3.19, + "learning_rate": 9.368170205457395e-06, + "loss": 0.806, + "step": 49646 + }, + { + "epoch": 3.19, + "learning_rate": 9.367955516434446e-06, + "loss": 0.7455, + "step": 49647 + }, + { + "epoch": 3.19, + "learning_rate": 9.367740827411496e-06, + "loss": 0.8431, + "step": 49648 + }, + { + "epoch": 3.19, + "learning_rate": 9.367526138388545e-06, + "loss": 0.887, + "step": 49649 + }, + { + "epoch": 3.19, + "learning_rate": 9.367311449365594e-06, + "loss": 0.8674, + "step": 49650 + }, + { + "epoch": 3.19, + "learning_rate": 9.367096760342645e-06, + "loss": 0.9017, + "step": 49651 + }, + { + "epoch": 3.19, + "learning_rate": 9.366882071319694e-06, + "loss": 0.7943, + "step": 49652 + }, + { + "epoch": 3.19, + "learning_rate": 9.366667382296743e-06, + "loss": 0.7753, + "step": 49653 + }, + { + "epoch": 3.19, + "learning_rate": 9.366452693273793e-06, + "loss": 0.8135, + "step": 49654 + }, + { + "epoch": 3.19, + "learning_rate": 9.366238004250844e-06, + "loss": 0.8597, + "step": 49655 + }, + { + "epoch": 3.19, + "learning_rate": 9.366023315227893e-06, + "loss": 0.842, + "step": 49656 + }, + { + "epoch": 3.19, + "learning_rate": 9.365808626204944e-06, + "loss": 0.8368, + "step": 49657 + }, + { + "epoch": 3.19, + "learning_rate": 9.365593937181993e-06, + "loss": 0.8458, + "step": 49658 + }, + { + "epoch": 3.19, + "learning_rate": 9.365379248159042e-06, + "loss": 0.8339, + "step": 49659 + }, + { + "epoch": 3.2, + "learning_rate": 9.365164559136092e-06, + "loss": 0.8072, + "step": 49660 + }, + { + "epoch": 3.2, + "learning_rate": 9.364949870113141e-06, + "loss": 0.8311, + "step": 49661 + }, + { + "epoch": 3.2, + "learning_rate": 9.364735181090192e-06, + "loss": 0.8072, + "step": 49662 + }, + { + "epoch": 3.2, + "learning_rate": 9.364520492067241e-06, + "loss": 0.829, + "step": 49663 + }, + { + "epoch": 3.2, + "learning_rate": 9.364305803044292e-06, + "loss": 0.8744, + "step": 49664 + }, + { + "epoch": 3.2, + "learning_rate": 9.36409111402134e-06, + "loss": 0.8755, + "step": 49665 + }, + { + "epoch": 3.2, + "learning_rate": 9.363876424998391e-06, + "loss": 0.7455, + "step": 49666 + }, + { + "epoch": 3.2, + "learning_rate": 9.36366173597544e-06, + "loss": 0.8259, + "step": 49667 + }, + { + "epoch": 3.2, + "learning_rate": 9.36344704695249e-06, + "loss": 0.7324, + "step": 49668 + }, + { + "epoch": 3.2, + "learning_rate": 9.36323235792954e-06, + "loss": 0.7867, + "step": 49669 + }, + { + "epoch": 3.2, + "learning_rate": 9.36301766890659e-06, + "loss": 0.8014, + "step": 49670 + }, + { + "epoch": 3.2, + "learning_rate": 9.36280297988364e-06, + "loss": 0.7944, + "step": 49671 + }, + { + "epoch": 3.2, + "learning_rate": 9.362588290860688e-06, + "loss": 0.802, + "step": 49672 + }, + { + "epoch": 3.2, + "learning_rate": 9.362373601837739e-06, + "loss": 0.7821, + "step": 49673 + }, + { + "epoch": 3.2, + "learning_rate": 9.362158912814788e-06, + "loss": 0.8505, + "step": 49674 + }, + { + "epoch": 3.2, + "learning_rate": 9.361944223791839e-06, + "loss": 0.8086, + "step": 49675 + }, + { + "epoch": 3.2, + "learning_rate": 9.36172953476889e-06, + "loss": 0.7574, + "step": 49676 + }, + { + "epoch": 3.2, + "learning_rate": 9.361514845745938e-06, + "loss": 0.8325, + "step": 49677 + }, + { + "epoch": 3.2, + "learning_rate": 9.361300156722987e-06, + "loss": 0.7922, + "step": 49678 + }, + { + "epoch": 3.2, + "learning_rate": 9.361085467700038e-06, + "loss": 0.8737, + "step": 49679 + }, + { + "epoch": 3.2, + "learning_rate": 9.360870778677087e-06, + "loss": 0.7952, + "step": 49680 + }, + { + "epoch": 3.2, + "learning_rate": 9.360656089654136e-06, + "loss": 0.8224, + "step": 49681 + }, + { + "epoch": 3.2, + "learning_rate": 9.360441400631187e-06, + "loss": 0.8772, + "step": 49682 + }, + { + "epoch": 3.2, + "learning_rate": 9.360226711608236e-06, + "loss": 0.7709, + "step": 49683 + }, + { + "epoch": 3.2, + "learning_rate": 9.360012022585286e-06, + "loss": 0.8429, + "step": 49684 + }, + { + "epoch": 3.2, + "learning_rate": 9.359797333562337e-06, + "loss": 0.7647, + "step": 49685 + }, + { + "epoch": 3.2, + "learning_rate": 9.359582644539386e-06, + "loss": 0.8095, + "step": 49686 + }, + { + "epoch": 3.2, + "learning_rate": 9.359367955516435e-06, + "loss": 0.8145, + "step": 49687 + }, + { + "epoch": 3.2, + "learning_rate": 9.359153266493485e-06, + "loss": 0.8173, + "step": 49688 + }, + { + "epoch": 3.2, + "learning_rate": 9.358938577470534e-06, + "loss": 0.8653, + "step": 49689 + }, + { + "epoch": 3.2, + "learning_rate": 9.358723888447583e-06, + "loss": 0.8089, + "step": 49690 + }, + { + "epoch": 3.2, + "learning_rate": 9.358509199424634e-06, + "loss": 0.8586, + "step": 49691 + }, + { + "epoch": 3.2, + "learning_rate": 9.358294510401685e-06, + "loss": 0.7372, + "step": 49692 + }, + { + "epoch": 3.2, + "learning_rate": 9.358079821378734e-06, + "loss": 0.9053, + "step": 49693 + }, + { + "epoch": 3.2, + "learning_rate": 9.357865132355783e-06, + "loss": 0.8371, + "step": 49694 + }, + { + "epoch": 3.2, + "learning_rate": 9.357650443332833e-06, + "loss": 0.8197, + "step": 49695 + }, + { + "epoch": 3.2, + "learning_rate": 9.357435754309882e-06, + "loss": 0.8074, + "step": 49696 + }, + { + "epoch": 3.2, + "learning_rate": 9.357221065286933e-06, + "loss": 0.8377, + "step": 49697 + }, + { + "epoch": 3.2, + "learning_rate": 9.357006376263984e-06, + "loss": 0.8333, + "step": 49698 + }, + { + "epoch": 3.2, + "learning_rate": 9.356791687241032e-06, + "loss": 0.7987, + "step": 49699 + }, + { + "epoch": 3.2, + "learning_rate": 9.356576998218081e-06, + "loss": 0.8582, + "step": 49700 + }, + { + "epoch": 3.2, + "learning_rate": 9.356362309195132e-06, + "loss": 0.8253, + "step": 49701 + }, + { + "epoch": 3.2, + "learning_rate": 9.356147620172181e-06, + "loss": 0.7615, + "step": 49702 + }, + { + "epoch": 3.2, + "learning_rate": 9.35593293114923e-06, + "loss": 0.8318, + "step": 49703 + }, + { + "epoch": 3.2, + "learning_rate": 9.35571824212628e-06, + "loss": 0.8193, + "step": 49704 + }, + { + "epoch": 3.2, + "learning_rate": 9.355503553103331e-06, + "loss": 0.9011, + "step": 49705 + }, + { + "epoch": 3.2, + "learning_rate": 9.35528886408038e-06, + "loss": 0.8627, + "step": 49706 + }, + { + "epoch": 3.2, + "learning_rate": 9.355074175057431e-06, + "loss": 0.8912, + "step": 49707 + }, + { + "epoch": 3.2, + "learning_rate": 9.35485948603448e-06, + "loss": 0.8129, + "step": 49708 + }, + { + "epoch": 3.2, + "learning_rate": 9.354644797011529e-06, + "loss": 0.8235, + "step": 49709 + }, + { + "epoch": 3.2, + "learning_rate": 9.35443010798858e-06, + "loss": 0.7808, + "step": 49710 + }, + { + "epoch": 3.2, + "learning_rate": 9.354215418965629e-06, + "loss": 0.836, + "step": 49711 + }, + { + "epoch": 3.2, + "learning_rate": 9.354000729942678e-06, + "loss": 0.7747, + "step": 49712 + }, + { + "epoch": 3.2, + "learning_rate": 9.353786040919728e-06, + "loss": 0.8311, + "step": 49713 + }, + { + "epoch": 3.2, + "learning_rate": 9.353571351896779e-06, + "loss": 0.8534, + "step": 49714 + }, + { + "epoch": 3.2, + "learning_rate": 9.353356662873828e-06, + "loss": 0.858, + "step": 49715 + }, + { + "epoch": 3.2, + "learning_rate": 9.353141973850878e-06, + "loss": 0.8223, + "step": 49716 + }, + { + "epoch": 3.2, + "learning_rate": 9.352927284827927e-06, + "loss": 0.8125, + "step": 49717 + }, + { + "epoch": 3.2, + "learning_rate": 9.352712595804976e-06, + "loss": 0.8403, + "step": 49718 + }, + { + "epoch": 3.2, + "learning_rate": 9.352497906782027e-06, + "loss": 0.8171, + "step": 49719 + }, + { + "epoch": 3.2, + "learning_rate": 9.352283217759078e-06, + "loss": 0.7469, + "step": 49720 + }, + { + "epoch": 3.2, + "learning_rate": 9.352068528736127e-06, + "loss": 0.8667, + "step": 49721 + }, + { + "epoch": 3.2, + "learning_rate": 9.351853839713176e-06, + "loss": 0.8645, + "step": 49722 + }, + { + "epoch": 3.2, + "learning_rate": 9.351639150690226e-06, + "loss": 0.8128, + "step": 49723 + }, + { + "epoch": 3.2, + "learning_rate": 9.351424461667275e-06, + "loss": 0.8335, + "step": 49724 + }, + { + "epoch": 3.2, + "learning_rate": 9.351209772644326e-06, + "loss": 0.8084, + "step": 49725 + }, + { + "epoch": 3.2, + "learning_rate": 9.350995083621375e-06, + "loss": 0.7443, + "step": 49726 + }, + { + "epoch": 3.2, + "learning_rate": 9.350780394598426e-06, + "loss": 0.8115, + "step": 49727 + }, + { + "epoch": 3.2, + "learning_rate": 9.350565705575474e-06, + "loss": 0.7613, + "step": 49728 + }, + { + "epoch": 3.2, + "learning_rate": 9.350351016552525e-06, + "loss": 0.8077, + "step": 49729 + }, + { + "epoch": 3.2, + "learning_rate": 9.350136327529574e-06, + "loss": 0.8433, + "step": 49730 + }, + { + "epoch": 3.2, + "learning_rate": 9.349921638506623e-06, + "loss": 0.8264, + "step": 49731 + }, + { + "epoch": 3.2, + "learning_rate": 9.349706949483674e-06, + "loss": 0.8459, + "step": 49732 + }, + { + "epoch": 3.2, + "learning_rate": 9.349492260460723e-06, + "loss": 0.826, + "step": 49733 + }, + { + "epoch": 3.2, + "learning_rate": 9.349277571437773e-06, + "loss": 0.7529, + "step": 49734 + }, + { + "epoch": 3.2, + "learning_rate": 9.349062882414824e-06, + "loss": 0.7582, + "step": 49735 + }, + { + "epoch": 3.2, + "learning_rate": 9.348848193391873e-06, + "loss": 0.8541, + "step": 49736 + }, + { + "epoch": 3.2, + "learning_rate": 9.348633504368922e-06, + "loss": 0.8814, + "step": 49737 + }, + { + "epoch": 3.2, + "learning_rate": 9.348418815345973e-06, + "loss": 0.8904, + "step": 49738 + }, + { + "epoch": 3.2, + "learning_rate": 9.348204126323022e-06, + "loss": 0.8391, + "step": 49739 + }, + { + "epoch": 3.2, + "learning_rate": 9.34798943730007e-06, + "loss": 0.8173, + "step": 49740 + }, + { + "epoch": 3.2, + "learning_rate": 9.347774748277121e-06, + "loss": 0.7873, + "step": 49741 + }, + { + "epoch": 3.2, + "learning_rate": 9.347560059254172e-06, + "loss": 0.8355, + "step": 49742 + }, + { + "epoch": 3.2, + "learning_rate": 9.34734537023122e-06, + "loss": 0.8219, + "step": 49743 + }, + { + "epoch": 3.2, + "learning_rate": 9.347130681208271e-06, + "loss": 0.8796, + "step": 49744 + }, + { + "epoch": 3.2, + "learning_rate": 9.34691599218532e-06, + "loss": 0.7913, + "step": 49745 + }, + { + "epoch": 3.2, + "learning_rate": 9.34670130316237e-06, + "loss": 0.7931, + "step": 49746 + }, + { + "epoch": 3.2, + "learning_rate": 9.34648661413942e-06, + "loss": 0.8285, + "step": 49747 + }, + { + "epoch": 3.2, + "learning_rate": 9.346271925116469e-06, + "loss": 0.8271, + "step": 49748 + }, + { + "epoch": 3.2, + "learning_rate": 9.34605723609352e-06, + "loss": 0.9116, + "step": 49749 + }, + { + "epoch": 3.2, + "learning_rate": 9.345842547070569e-06, + "loss": 0.8553, + "step": 49750 + }, + { + "epoch": 3.2, + "learning_rate": 9.34562785804762e-06, + "loss": 0.8088, + "step": 49751 + }, + { + "epoch": 3.2, + "learning_rate": 9.345413169024668e-06, + "loss": 0.8204, + "step": 49752 + }, + { + "epoch": 3.2, + "learning_rate": 9.345198480001719e-06, + "loss": 0.8157, + "step": 49753 + }, + { + "epoch": 3.2, + "learning_rate": 9.344983790978768e-06, + "loss": 0.7883, + "step": 49754 + }, + { + "epoch": 3.2, + "learning_rate": 9.344769101955817e-06, + "loss": 0.8408, + "step": 49755 + }, + { + "epoch": 3.2, + "learning_rate": 9.344554412932868e-06, + "loss": 0.7815, + "step": 49756 + }, + { + "epoch": 3.2, + "learning_rate": 9.344339723909918e-06, + "loss": 0.8075, + "step": 49757 + }, + { + "epoch": 3.2, + "learning_rate": 9.344125034886967e-06, + "loss": 0.8111, + "step": 49758 + }, + { + "epoch": 3.2, + "learning_rate": 9.343910345864016e-06, + "loss": 0.8132, + "step": 49759 + }, + { + "epoch": 3.2, + "learning_rate": 9.343695656841067e-06, + "loss": 0.8645, + "step": 49760 + }, + { + "epoch": 3.2, + "learning_rate": 9.343480967818116e-06, + "loss": 0.8112, + "step": 49761 + }, + { + "epoch": 3.2, + "learning_rate": 9.343266278795166e-06, + "loss": 0.8089, + "step": 49762 + }, + { + "epoch": 3.2, + "learning_rate": 9.343051589772215e-06, + "loss": 0.8328, + "step": 49763 + }, + { + "epoch": 3.2, + "learning_rate": 9.342836900749266e-06, + "loss": 0.9294, + "step": 49764 + }, + { + "epoch": 3.2, + "learning_rate": 9.342622211726315e-06, + "loss": 0.7766, + "step": 49765 + }, + { + "epoch": 3.2, + "learning_rate": 9.342407522703366e-06, + "loss": 0.821, + "step": 49766 + }, + { + "epoch": 3.2, + "learning_rate": 9.342192833680415e-06, + "loss": 0.8215, + "step": 49767 + }, + { + "epoch": 3.2, + "learning_rate": 9.341978144657464e-06, + "loss": 0.8154, + "step": 49768 + }, + { + "epoch": 3.2, + "learning_rate": 9.341763455634514e-06, + "loss": 0.8197, + "step": 49769 + }, + { + "epoch": 3.2, + "learning_rate": 9.341548766611563e-06, + "loss": 0.7675, + "step": 49770 + }, + { + "epoch": 3.2, + "learning_rate": 9.341334077588614e-06, + "loss": 0.8409, + "step": 49771 + }, + { + "epoch": 3.2, + "learning_rate": 9.341119388565665e-06, + "loss": 0.7269, + "step": 49772 + }, + { + "epoch": 3.2, + "learning_rate": 9.340904699542713e-06, + "loss": 0.8289, + "step": 49773 + }, + { + "epoch": 3.2, + "learning_rate": 9.340690010519762e-06, + "loss": 0.8214, + "step": 49774 + }, + { + "epoch": 3.2, + "learning_rate": 9.340475321496813e-06, + "loss": 0.7772, + "step": 49775 + }, + { + "epoch": 3.2, + "learning_rate": 9.340260632473862e-06, + "loss": 0.8367, + "step": 49776 + }, + { + "epoch": 3.2, + "learning_rate": 9.340045943450911e-06, + "loss": 0.836, + "step": 49777 + }, + { + "epoch": 3.2, + "learning_rate": 9.339831254427962e-06, + "loss": 0.8411, + "step": 49778 + }, + { + "epoch": 3.2, + "learning_rate": 9.339616565405012e-06, + "loss": 0.842, + "step": 49779 + }, + { + "epoch": 3.2, + "learning_rate": 9.339401876382061e-06, + "loss": 0.8305, + "step": 49780 + }, + { + "epoch": 3.2, + "learning_rate": 9.33918718735911e-06, + "loss": 0.7973, + "step": 49781 + }, + { + "epoch": 3.2, + "learning_rate": 9.338972498336161e-06, + "loss": 0.8279, + "step": 49782 + }, + { + "epoch": 3.2, + "learning_rate": 9.33875780931321e-06, + "loss": 0.8209, + "step": 49783 + }, + { + "epoch": 3.2, + "learning_rate": 9.33854312029026e-06, + "loss": 0.8102, + "step": 49784 + }, + { + "epoch": 3.2, + "learning_rate": 9.338328431267311e-06, + "loss": 0.7959, + "step": 49785 + }, + { + "epoch": 3.2, + "learning_rate": 9.33811374224436e-06, + "loss": 0.8492, + "step": 49786 + }, + { + "epoch": 3.2, + "learning_rate": 9.33789905322141e-06, + "loss": 0.9017, + "step": 49787 + }, + { + "epoch": 3.2, + "learning_rate": 9.33768436419846e-06, + "loss": 0.8075, + "step": 49788 + }, + { + "epoch": 3.2, + "learning_rate": 9.337469675175509e-06, + "loss": 0.7995, + "step": 49789 + }, + { + "epoch": 3.2, + "learning_rate": 9.337254986152558e-06, + "loss": 0.8174, + "step": 49790 + }, + { + "epoch": 3.2, + "learning_rate": 9.337040297129608e-06, + "loss": 0.8387, + "step": 49791 + }, + { + "epoch": 3.2, + "learning_rate": 9.336825608106657e-06, + "loss": 0.7481, + "step": 49792 + }, + { + "epoch": 3.2, + "learning_rate": 9.336610919083708e-06, + "loss": 0.8383, + "step": 49793 + }, + { + "epoch": 3.2, + "learning_rate": 9.336396230060759e-06, + "loss": 0.7829, + "step": 49794 + }, + { + "epoch": 3.2, + "learning_rate": 9.336181541037808e-06, + "loss": 0.818, + "step": 49795 + }, + { + "epoch": 3.2, + "learning_rate": 9.335966852014857e-06, + "loss": 0.7564, + "step": 49796 + }, + { + "epoch": 3.2, + "learning_rate": 9.335752162991907e-06, + "loss": 0.7705, + "step": 49797 + }, + { + "epoch": 3.2, + "learning_rate": 9.335537473968956e-06, + "loss": 0.7975, + "step": 49798 + }, + { + "epoch": 3.2, + "learning_rate": 9.335322784946005e-06, + "loss": 0.8206, + "step": 49799 + }, + { + "epoch": 3.2, + "learning_rate": 9.335108095923056e-06, + "loss": 0.8316, + "step": 49800 + }, + { + "epoch": 3.2, + "learning_rate": 9.334893406900107e-06, + "loss": 0.8395, + "step": 49801 + }, + { + "epoch": 3.2, + "learning_rate": 9.334678717877156e-06, + "loss": 0.8284, + "step": 49802 + }, + { + "epoch": 3.2, + "learning_rate": 9.334464028854206e-06, + "loss": 0.8384, + "step": 49803 + }, + { + "epoch": 3.2, + "learning_rate": 9.334249339831255e-06, + "loss": 0.8092, + "step": 49804 + }, + { + "epoch": 3.2, + "learning_rate": 9.334034650808304e-06, + "loss": 0.8062, + "step": 49805 + }, + { + "epoch": 3.2, + "learning_rate": 9.333819961785355e-06, + "loss": 0.7543, + "step": 49806 + }, + { + "epoch": 3.2, + "learning_rate": 9.333605272762405e-06, + "loss": 0.8592, + "step": 49807 + }, + { + "epoch": 3.2, + "learning_rate": 9.333390583739454e-06, + "loss": 0.7647, + "step": 49808 + }, + { + "epoch": 3.2, + "learning_rate": 9.333175894716503e-06, + "loss": 0.8567, + "step": 49809 + }, + { + "epoch": 3.2, + "learning_rate": 9.332961205693554e-06, + "loss": 0.8346, + "step": 49810 + }, + { + "epoch": 3.2, + "learning_rate": 9.332746516670603e-06, + "loss": 0.8307, + "step": 49811 + }, + { + "epoch": 3.2, + "learning_rate": 9.332531827647654e-06, + "loss": 0.8589, + "step": 49812 + }, + { + "epoch": 3.2, + "learning_rate": 9.332317138624703e-06, + "loss": 0.8041, + "step": 49813 + }, + { + "epoch": 3.2, + "learning_rate": 9.332102449601753e-06, + "loss": 0.8168, + "step": 49814 + }, + { + "epoch": 3.2, + "learning_rate": 9.331887760578802e-06, + "loss": 0.8452, + "step": 49815 + }, + { + "epoch": 3.21, + "learning_rate": 9.331673071555853e-06, + "loss": 0.8784, + "step": 49816 + }, + { + "epoch": 3.21, + "learning_rate": 9.331458382532902e-06, + "loss": 0.7945, + "step": 49817 + }, + { + "epoch": 3.21, + "learning_rate": 9.33124369350995e-06, + "loss": 0.7302, + "step": 49818 + }, + { + "epoch": 3.21, + "learning_rate": 9.331029004487001e-06, + "loss": 0.8492, + "step": 49819 + }, + { + "epoch": 3.21, + "learning_rate": 9.33081431546405e-06, + "loss": 0.747, + "step": 49820 + }, + { + "epoch": 3.21, + "learning_rate": 9.330599626441101e-06, + "loss": 0.8923, + "step": 49821 + }, + { + "epoch": 3.21, + "learning_rate": 9.330384937418152e-06, + "loss": 0.7933, + "step": 49822 + }, + { + "epoch": 3.21, + "learning_rate": 9.3301702483952e-06, + "loss": 0.7865, + "step": 49823 + }, + { + "epoch": 3.21, + "learning_rate": 9.32995555937225e-06, + "loss": 0.8418, + "step": 49824 + }, + { + "epoch": 3.21, + "learning_rate": 9.3297408703493e-06, + "loss": 0.7964, + "step": 49825 + }, + { + "epoch": 3.21, + "learning_rate": 9.32952618132635e-06, + "loss": 0.7964, + "step": 49826 + }, + { + "epoch": 3.21, + "learning_rate": 9.329311492303398e-06, + "loss": 0.81, + "step": 49827 + }, + { + "epoch": 3.21, + "learning_rate": 9.329096803280449e-06, + "loss": 0.8076, + "step": 49828 + }, + { + "epoch": 3.21, + "learning_rate": 9.3288821142575e-06, + "loss": 0.7867, + "step": 49829 + }, + { + "epoch": 3.21, + "learning_rate": 9.328667425234549e-06, + "loss": 0.8112, + "step": 49830 + }, + { + "epoch": 3.21, + "learning_rate": 9.3284527362116e-06, + "loss": 0.8255, + "step": 49831 + }, + { + "epoch": 3.21, + "learning_rate": 9.328238047188648e-06, + "loss": 0.7822, + "step": 49832 + }, + { + "epoch": 3.21, + "learning_rate": 9.328023358165697e-06, + "loss": 0.9196, + "step": 49833 + }, + { + "epoch": 3.21, + "learning_rate": 9.327808669142748e-06, + "loss": 0.8145, + "step": 49834 + }, + { + "epoch": 3.21, + "learning_rate": 9.327593980119797e-06, + "loss": 0.8477, + "step": 49835 + }, + { + "epoch": 3.21, + "learning_rate": 9.327379291096847e-06, + "loss": 0.8148, + "step": 49836 + }, + { + "epoch": 3.21, + "learning_rate": 9.327164602073896e-06, + "loss": 0.8244, + "step": 49837 + }, + { + "epoch": 3.21, + "learning_rate": 9.326949913050947e-06, + "loss": 0.9157, + "step": 49838 + }, + { + "epoch": 3.21, + "learning_rate": 9.326735224027996e-06, + "loss": 0.7513, + "step": 49839 + }, + { + "epoch": 3.21, + "learning_rate": 9.326520535005047e-06, + "loss": 0.7981, + "step": 49840 + }, + { + "epoch": 3.21, + "learning_rate": 9.326305845982096e-06, + "loss": 0.7873, + "step": 49841 + }, + { + "epoch": 3.21, + "learning_rate": 9.326091156959145e-06, + "loss": 0.8038, + "step": 49842 + }, + { + "epoch": 3.21, + "learning_rate": 9.325876467936195e-06, + "loss": 0.7862, + "step": 49843 + }, + { + "epoch": 3.21, + "learning_rate": 9.325661778913246e-06, + "loss": 0.7933, + "step": 49844 + }, + { + "epoch": 3.21, + "learning_rate": 9.325447089890295e-06, + "loss": 0.7963, + "step": 49845 + }, + { + "epoch": 3.21, + "learning_rate": 9.325232400867344e-06, + "loss": 0.7827, + "step": 49846 + }, + { + "epoch": 3.21, + "learning_rate": 9.325017711844394e-06, + "loss": 0.8357, + "step": 49847 + }, + { + "epoch": 3.21, + "learning_rate": 9.324803022821443e-06, + "loss": 0.8068, + "step": 49848 + }, + { + "epoch": 3.21, + "learning_rate": 9.324588333798492e-06, + "loss": 0.7617, + "step": 49849 + }, + { + "epoch": 3.21, + "learning_rate": 9.324373644775543e-06, + "loss": 0.8425, + "step": 49850 + }, + { + "epoch": 3.21, + "learning_rate": 9.324158955752594e-06, + "loss": 0.8192, + "step": 49851 + }, + { + "epoch": 3.21, + "learning_rate": 9.323944266729643e-06, + "loss": 0.7946, + "step": 49852 + }, + { + "epoch": 3.21, + "learning_rate": 9.323729577706693e-06, + "loss": 0.7856, + "step": 49853 + }, + { + "epoch": 3.21, + "learning_rate": 9.323514888683742e-06, + "loss": 0.814, + "step": 49854 + }, + { + "epoch": 3.21, + "learning_rate": 9.323300199660791e-06, + "loss": 0.8599, + "step": 49855 + }, + { + "epoch": 3.21, + "learning_rate": 9.323085510637842e-06, + "loss": 0.8366, + "step": 49856 + }, + { + "epoch": 3.21, + "learning_rate": 9.322870821614891e-06, + "loss": 0.8739, + "step": 49857 + }, + { + "epoch": 3.21, + "learning_rate": 9.322656132591942e-06, + "loss": 0.8626, + "step": 49858 + }, + { + "epoch": 3.21, + "learning_rate": 9.32244144356899e-06, + "loss": 0.7899, + "step": 49859 + }, + { + "epoch": 3.21, + "learning_rate": 9.322226754546041e-06, + "loss": 0.8487, + "step": 49860 + }, + { + "epoch": 3.21, + "learning_rate": 9.32201206552309e-06, + "loss": 0.8187, + "step": 49861 + }, + { + "epoch": 3.21, + "learning_rate": 9.32179737650014e-06, + "loss": 0.8225, + "step": 49862 + }, + { + "epoch": 3.21, + "learning_rate": 9.32158268747719e-06, + "loss": 0.7806, + "step": 49863 + }, + { + "epoch": 3.21, + "learning_rate": 9.321367998454239e-06, + "loss": 0.8381, + "step": 49864 + }, + { + "epoch": 3.21, + "learning_rate": 9.32115330943129e-06, + "loss": 0.855, + "step": 49865 + }, + { + "epoch": 3.21, + "learning_rate": 9.32093862040834e-06, + "loss": 0.8424, + "step": 49866 + }, + { + "epoch": 3.21, + "learning_rate": 9.320723931385389e-06, + "loss": 0.8926, + "step": 49867 + }, + { + "epoch": 3.21, + "learning_rate": 9.320509242362438e-06, + "loss": 0.7628, + "step": 49868 + }, + { + "epoch": 3.21, + "learning_rate": 9.320294553339489e-06, + "loss": 0.8595, + "step": 49869 + }, + { + "epoch": 3.21, + "learning_rate": 9.320079864316538e-06, + "loss": 0.8118, + "step": 49870 + }, + { + "epoch": 3.21, + "learning_rate": 9.319865175293588e-06, + "loss": 0.8704, + "step": 49871 + }, + { + "epoch": 3.21, + "learning_rate": 9.319650486270639e-06, + "loss": 0.8957, + "step": 49872 + }, + { + "epoch": 3.21, + "learning_rate": 9.319435797247688e-06, + "loss": 0.7942, + "step": 49873 + }, + { + "epoch": 3.21, + "learning_rate": 9.319221108224737e-06, + "loss": 0.8347, + "step": 49874 + }, + { + "epoch": 3.21, + "learning_rate": 9.319006419201788e-06, + "loss": 0.907, + "step": 49875 + }, + { + "epoch": 3.21, + "learning_rate": 9.318791730178837e-06, + "loss": 0.8381, + "step": 49876 + }, + { + "epoch": 3.21, + "learning_rate": 9.318577041155885e-06, + "loss": 0.8171, + "step": 49877 + }, + { + "epoch": 3.21, + "learning_rate": 9.318362352132936e-06, + "loss": 0.7794, + "step": 49878 + }, + { + "epoch": 3.21, + "learning_rate": 9.318147663109985e-06, + "loss": 0.795, + "step": 49879 + }, + { + "epoch": 3.21, + "learning_rate": 9.317932974087036e-06, + "loss": 0.8102, + "step": 49880 + }, + { + "epoch": 3.21, + "learning_rate": 9.317718285064086e-06, + "loss": 0.9283, + "step": 49881 + }, + { + "epoch": 3.21, + "learning_rate": 9.317503596041135e-06, + "loss": 0.8378, + "step": 49882 + }, + { + "epoch": 3.21, + "learning_rate": 9.317288907018184e-06, + "loss": 0.7886, + "step": 49883 + }, + { + "epoch": 3.21, + "learning_rate": 9.317074217995235e-06, + "loss": 0.7769, + "step": 49884 + }, + { + "epoch": 3.21, + "learning_rate": 9.316859528972284e-06, + "loss": 0.8353, + "step": 49885 + }, + { + "epoch": 3.21, + "learning_rate": 9.316644839949333e-06, + "loss": 0.8832, + "step": 49886 + }, + { + "epoch": 3.21, + "learning_rate": 9.316430150926384e-06, + "loss": 0.7934, + "step": 49887 + }, + { + "epoch": 3.21, + "learning_rate": 9.316215461903434e-06, + "loss": 0.8726, + "step": 49888 + }, + { + "epoch": 3.21, + "learning_rate": 9.316000772880483e-06, + "loss": 0.8439, + "step": 49889 + }, + { + "epoch": 3.21, + "learning_rate": 9.315786083857534e-06, + "loss": 0.8408, + "step": 49890 + }, + { + "epoch": 3.21, + "learning_rate": 9.315571394834583e-06, + "loss": 0.7875, + "step": 49891 + }, + { + "epoch": 3.21, + "learning_rate": 9.315356705811632e-06, + "loss": 0.8449, + "step": 49892 + }, + { + "epoch": 3.21, + "learning_rate": 9.315142016788682e-06, + "loss": 0.8603, + "step": 49893 + }, + { + "epoch": 3.21, + "learning_rate": 9.314927327765733e-06, + "loss": 0.8144, + "step": 49894 + }, + { + "epoch": 3.21, + "learning_rate": 9.314712638742782e-06, + "loss": 0.881, + "step": 49895 + }, + { + "epoch": 3.21, + "learning_rate": 9.314497949719831e-06, + "loss": 0.8047, + "step": 49896 + }, + { + "epoch": 3.21, + "learning_rate": 9.314283260696882e-06, + "loss": 0.8763, + "step": 49897 + }, + { + "epoch": 3.21, + "learning_rate": 9.31406857167393e-06, + "loss": 0.8385, + "step": 49898 + }, + { + "epoch": 3.21, + "learning_rate": 9.313853882650981e-06, + "loss": 0.8266, + "step": 49899 + }, + { + "epoch": 3.21, + "learning_rate": 9.31363919362803e-06, + "loss": 0.8209, + "step": 49900 + }, + { + "epoch": 3.21, + "learning_rate": 9.313424504605081e-06, + "loss": 0.838, + "step": 49901 + }, + { + "epoch": 3.21, + "learning_rate": 9.31320981558213e-06, + "loss": 0.815, + "step": 49902 + }, + { + "epoch": 3.21, + "learning_rate": 9.31299512655918e-06, + "loss": 0.8798, + "step": 49903 + }, + { + "epoch": 3.21, + "learning_rate": 9.31278043753623e-06, + "loss": 0.8503, + "step": 49904 + }, + { + "epoch": 3.21, + "learning_rate": 9.312565748513279e-06, + "loss": 0.7444, + "step": 49905 + }, + { + "epoch": 3.21, + "learning_rate": 9.31235105949033e-06, + "loss": 0.7993, + "step": 49906 + }, + { + "epoch": 3.21, + "learning_rate": 9.312136370467378e-06, + "loss": 0.8446, + "step": 49907 + }, + { + "epoch": 3.21, + "learning_rate": 9.311921681444429e-06, + "loss": 0.8663, + "step": 49908 + }, + { + "epoch": 3.21, + "learning_rate": 9.31170699242148e-06, + "loss": 0.8347, + "step": 49909 + }, + { + "epoch": 3.21, + "learning_rate": 9.311492303398528e-06, + "loss": 0.8238, + "step": 49910 + }, + { + "epoch": 3.21, + "learning_rate": 9.311277614375577e-06, + "loss": 0.6922, + "step": 49911 + }, + { + "epoch": 3.21, + "learning_rate": 9.311062925352628e-06, + "loss": 0.7732, + "step": 49912 + }, + { + "epoch": 3.21, + "learning_rate": 9.310848236329677e-06, + "loss": 0.7916, + "step": 49913 + }, + { + "epoch": 3.21, + "learning_rate": 9.310633547306726e-06, + "loss": 0.8737, + "step": 49914 + }, + { + "epoch": 3.21, + "learning_rate": 9.310418858283777e-06, + "loss": 0.8399, + "step": 49915 + }, + { + "epoch": 3.21, + "learning_rate": 9.310204169260827e-06, + "loss": 0.9009, + "step": 49916 + }, + { + "epoch": 3.21, + "learning_rate": 9.309989480237876e-06, + "loss": 0.8002, + "step": 49917 + }, + { + "epoch": 3.21, + "learning_rate": 9.309774791214927e-06, + "loss": 0.867, + "step": 49918 + }, + { + "epoch": 3.21, + "learning_rate": 9.309560102191976e-06, + "loss": 0.7934, + "step": 49919 + }, + { + "epoch": 3.21, + "learning_rate": 9.309345413169025e-06, + "loss": 0.8166, + "step": 49920 + }, + { + "epoch": 3.21, + "learning_rate": 9.309130724146076e-06, + "loss": 0.8741, + "step": 49921 + }, + { + "epoch": 3.21, + "learning_rate": 9.308916035123124e-06, + "loss": 0.8587, + "step": 49922 + }, + { + "epoch": 3.21, + "learning_rate": 9.308701346100175e-06, + "loss": 0.8301, + "step": 49923 + }, + { + "epoch": 3.21, + "learning_rate": 9.308486657077224e-06, + "loss": 0.7889, + "step": 49924 + }, + { + "epoch": 3.21, + "learning_rate": 9.308271968054275e-06, + "loss": 0.7896, + "step": 49925 + }, + { + "epoch": 3.21, + "learning_rate": 9.308057279031324e-06, + "loss": 0.8655, + "step": 49926 + }, + { + "epoch": 3.21, + "learning_rate": 9.307842590008374e-06, + "loss": 0.8631, + "step": 49927 + }, + { + "epoch": 3.21, + "learning_rate": 9.307627900985423e-06, + "loss": 0.7754, + "step": 49928 + }, + { + "epoch": 3.21, + "learning_rate": 9.307413211962472e-06, + "loss": 0.7797, + "step": 49929 + }, + { + "epoch": 3.21, + "learning_rate": 9.307198522939523e-06, + "loss": 0.832, + "step": 49930 + }, + { + "epoch": 3.21, + "learning_rate": 9.306983833916574e-06, + "loss": 0.8575, + "step": 49931 + }, + { + "epoch": 3.21, + "learning_rate": 9.306769144893623e-06, + "loss": 0.8347, + "step": 49932 + }, + { + "epoch": 3.21, + "learning_rate": 9.306554455870672e-06, + "loss": 0.8345, + "step": 49933 + }, + { + "epoch": 3.21, + "learning_rate": 9.306339766847722e-06, + "loss": 0.8227, + "step": 49934 + }, + { + "epoch": 3.21, + "learning_rate": 9.306125077824771e-06, + "loss": 0.8051, + "step": 49935 + }, + { + "epoch": 3.21, + "learning_rate": 9.30591038880182e-06, + "loss": 0.8499, + "step": 49936 + }, + { + "epoch": 3.21, + "learning_rate": 9.30569569977887e-06, + "loss": 0.7967, + "step": 49937 + }, + { + "epoch": 3.21, + "learning_rate": 9.305481010755921e-06, + "loss": 0.7483, + "step": 49938 + }, + { + "epoch": 3.21, + "learning_rate": 9.30526632173297e-06, + "loss": 0.8551, + "step": 49939 + }, + { + "epoch": 3.21, + "learning_rate": 9.305051632710021e-06, + "loss": 0.7915, + "step": 49940 + }, + { + "epoch": 3.21, + "learning_rate": 9.30483694368707e-06, + "loss": 0.8194, + "step": 49941 + }, + { + "epoch": 3.21, + "learning_rate": 9.304622254664119e-06, + "loss": 0.8036, + "step": 49942 + }, + { + "epoch": 3.21, + "learning_rate": 9.30440756564117e-06, + "loss": 0.8448, + "step": 49943 + }, + { + "epoch": 3.21, + "learning_rate": 9.304192876618219e-06, + "loss": 0.8206, + "step": 49944 + }, + { + "epoch": 3.21, + "learning_rate": 9.30397818759527e-06, + "loss": 0.8158, + "step": 49945 + }, + { + "epoch": 3.21, + "learning_rate": 9.303763498572318e-06, + "loss": 0.7779, + "step": 49946 + }, + { + "epoch": 3.21, + "learning_rate": 9.303548809549369e-06, + "loss": 0.8503, + "step": 49947 + }, + { + "epoch": 3.21, + "learning_rate": 9.303334120526418e-06, + "loss": 0.8178, + "step": 49948 + }, + { + "epoch": 3.21, + "learning_rate": 9.303119431503469e-06, + "loss": 0.813, + "step": 49949 + }, + { + "epoch": 3.21, + "learning_rate": 9.302904742480518e-06, + "loss": 0.8032, + "step": 49950 + }, + { + "epoch": 3.21, + "learning_rate": 9.302690053457566e-06, + "loss": 0.8561, + "step": 49951 + }, + { + "epoch": 3.21, + "learning_rate": 9.302475364434617e-06, + "loss": 0.8053, + "step": 49952 + }, + { + "epoch": 3.21, + "learning_rate": 9.302260675411668e-06, + "loss": 0.8102, + "step": 49953 + }, + { + "epoch": 3.21, + "learning_rate": 9.302045986388717e-06, + "loss": 0.7989, + "step": 49954 + }, + { + "epoch": 3.21, + "learning_rate": 9.301831297365766e-06, + "loss": 0.8189, + "step": 49955 + }, + { + "epoch": 3.21, + "learning_rate": 9.301616608342816e-06, + "loss": 0.8887, + "step": 49956 + }, + { + "epoch": 3.21, + "learning_rate": 9.301401919319865e-06, + "loss": 0.7932, + "step": 49957 + }, + { + "epoch": 3.21, + "learning_rate": 9.301187230296916e-06, + "loss": 0.7911, + "step": 49958 + }, + { + "epoch": 3.21, + "learning_rate": 9.300972541273965e-06, + "loss": 0.8452, + "step": 49959 + }, + { + "epoch": 3.21, + "learning_rate": 9.300757852251016e-06, + "loss": 0.8479, + "step": 49960 + }, + { + "epoch": 3.21, + "learning_rate": 9.300543163228065e-06, + "loss": 0.7813, + "step": 49961 + }, + { + "epoch": 3.21, + "learning_rate": 9.300328474205115e-06, + "loss": 0.817, + "step": 49962 + }, + { + "epoch": 3.21, + "learning_rate": 9.300113785182164e-06, + "loss": 0.8423, + "step": 49963 + }, + { + "epoch": 3.21, + "learning_rate": 9.299899096159213e-06, + "loss": 0.8837, + "step": 49964 + }, + { + "epoch": 3.21, + "learning_rate": 9.299684407136264e-06, + "loss": 0.8127, + "step": 49965 + }, + { + "epoch": 3.21, + "learning_rate": 9.299469718113313e-06, + "loss": 0.825, + "step": 49966 + }, + { + "epoch": 3.21, + "learning_rate": 9.299255029090363e-06, + "loss": 0.8853, + "step": 49967 + }, + { + "epoch": 3.21, + "learning_rate": 9.299040340067414e-06, + "loss": 0.8256, + "step": 49968 + }, + { + "epoch": 3.21, + "learning_rate": 9.298825651044463e-06, + "loss": 0.8514, + "step": 49969 + }, + { + "epoch": 3.21, + "learning_rate": 9.298610962021512e-06, + "loss": 0.8555, + "step": 49970 + }, + { + "epoch": 3.22, + "learning_rate": 9.298396272998563e-06, + "loss": 0.8792, + "step": 49971 + }, + { + "epoch": 3.22, + "learning_rate": 9.298181583975612e-06, + "loss": 0.7902, + "step": 49972 + }, + { + "epoch": 3.22, + "learning_rate": 9.29796689495266e-06, + "loss": 0.8525, + "step": 49973 + }, + { + "epoch": 3.22, + "learning_rate": 9.297752205929711e-06, + "loss": 0.8338, + "step": 49974 + }, + { + "epoch": 3.22, + "learning_rate": 9.297537516906762e-06, + "loss": 0.8174, + "step": 49975 + }, + { + "epoch": 3.22, + "learning_rate": 9.297322827883811e-06, + "loss": 0.8124, + "step": 49976 + }, + { + "epoch": 3.22, + "learning_rate": 9.297108138860862e-06, + "loss": 0.9181, + "step": 49977 + }, + { + "epoch": 3.22, + "learning_rate": 9.29689344983791e-06, + "loss": 0.8609, + "step": 49978 + }, + { + "epoch": 3.22, + "learning_rate": 9.29667876081496e-06, + "loss": 0.8396, + "step": 49979 + }, + { + "epoch": 3.22, + "learning_rate": 9.29646407179201e-06, + "loss": 0.7576, + "step": 49980 + }, + { + "epoch": 3.22, + "learning_rate": 9.29624938276906e-06, + "loss": 0.8266, + "step": 49981 + }, + { + "epoch": 3.22, + "learning_rate": 9.29603469374611e-06, + "loss": 0.8534, + "step": 49982 + }, + { + "epoch": 3.22, + "learning_rate": 9.295820004723159e-06, + "loss": 0.7783, + "step": 49983 + }, + { + "epoch": 3.22, + "learning_rate": 9.29560531570021e-06, + "loss": 0.8044, + "step": 49984 + }, + { + "epoch": 3.22, + "learning_rate": 9.295390626677258e-06, + "loss": 0.8392, + "step": 49985 + }, + { + "epoch": 3.22, + "learning_rate": 9.295175937654309e-06, + "loss": 0.8419, + "step": 49986 + }, + { + "epoch": 3.22, + "learning_rate": 9.294961248631358e-06, + "loss": 0.7895, + "step": 49987 + }, + { + "epoch": 3.22, + "learning_rate": 9.294746559608407e-06, + "loss": 0.8176, + "step": 49988 + }, + { + "epoch": 3.22, + "learning_rate": 9.294531870585458e-06, + "loss": 0.8457, + "step": 49989 + }, + { + "epoch": 3.22, + "learning_rate": 9.294317181562508e-06, + "loss": 0.7916, + "step": 49990 + }, + { + "epoch": 3.22, + "learning_rate": 9.294102492539557e-06, + "loss": 0.838, + "step": 49991 + }, + { + "epoch": 3.22, + "learning_rate": 9.293887803516606e-06, + "loss": 0.7654, + "step": 49992 + }, + { + "epoch": 3.22, + "learning_rate": 9.293673114493657e-06, + "loss": 0.8438, + "step": 49993 + }, + { + "epoch": 3.22, + "learning_rate": 9.293458425470706e-06, + "loss": 0.8668, + "step": 49994 + }, + { + "epoch": 3.22, + "learning_rate": 9.293243736447757e-06, + "loss": 0.8181, + "step": 49995 + }, + { + "epoch": 3.22, + "learning_rate": 9.293029047424807e-06, + "loss": 0.8123, + "step": 49996 + }, + { + "epoch": 3.22, + "learning_rate": 9.292814358401856e-06, + "loss": 0.7938, + "step": 49997 + }, + { + "epoch": 3.22, + "learning_rate": 9.292599669378905e-06, + "loss": 0.8036, + "step": 49998 + }, + { + "epoch": 3.22, + "learning_rate": 9.292384980355956e-06, + "loss": 0.826, + "step": 49999 + }, + { + "epoch": 3.22, + "learning_rate": 9.292170291333005e-06, + "loss": 0.8691, + "step": 50000 + }, + { + "epoch": 3.22, + "eval_laion/cider": 0.525816737111265, + "eval_laion/meteor": 0.3928087864157689, + "eval_laion/pred_num_words": 38.62890625, + "eval_laion/sacrebleu": 16.149836537568337, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.054405927658081, + "eval_runtime": 98.6443, + "eval_samples_per_second": 5.059, + "eval_steps_per_second": 0.324, + "step": 50000 + }, + { + "epoch": 3.22, + "learning_rate": 9.291955602310054e-06, + "loss": 0.8181, + "step": 50001 + }, + { + "epoch": 3.22, + "learning_rate": 9.291740913287104e-06, + "loss": 0.7889, + "step": 50002 + }, + { + "epoch": 3.22, + "learning_rate": 9.291526224264155e-06, + "loss": 0.8233, + "step": 50003 + }, + { + "epoch": 3.22, + "learning_rate": 9.291311535241204e-06, + "loss": 0.8181, + "step": 50004 + }, + { + "epoch": 3.22, + "learning_rate": 9.291096846218255e-06, + "loss": 0.8877, + "step": 50005 + }, + { + "epoch": 3.22, + "learning_rate": 9.290882157195304e-06, + "loss": 0.8407, + "step": 50006 + }, + { + "epoch": 3.22, + "learning_rate": 9.290667468172353e-06, + "loss": 0.8648, + "step": 50007 + }, + { + "epoch": 3.22, + "learning_rate": 9.290452779149403e-06, + "loss": 0.8311, + "step": 50008 + }, + { + "epoch": 3.22, + "learning_rate": 9.290238090126452e-06, + "loss": 0.8364, + "step": 50009 + }, + { + "epoch": 3.22, + "learning_rate": 9.290023401103503e-06, + "loss": 0.8316, + "step": 50010 + }, + { + "epoch": 3.22, + "learning_rate": 9.289808712080552e-06, + "loss": 0.8232, + "step": 50011 + }, + { + "epoch": 3.22, + "learning_rate": 9.289594023057602e-06, + "loss": 0.8222, + "step": 50012 + }, + { + "epoch": 3.22, + "learning_rate": 9.289379334034651e-06, + "loss": 0.7771, + "step": 50013 + }, + { + "epoch": 3.22, + "learning_rate": 9.2891646450117e-06, + "loss": 0.8465, + "step": 50014 + }, + { + "epoch": 3.22, + "learning_rate": 9.288949955988751e-06, + "loss": 0.8252, + "step": 50015 + }, + { + "epoch": 3.22, + "learning_rate": 9.2887352669658e-06, + "loss": 0.8741, + "step": 50016 + }, + { + "epoch": 3.22, + "learning_rate": 9.28852057794285e-06, + "loss": 0.7904, + "step": 50017 + }, + { + "epoch": 3.22, + "learning_rate": 9.288305888919901e-06, + "loss": 0.8286, + "step": 50018 + }, + { + "epoch": 3.22, + "learning_rate": 9.28809119989695e-06, + "loss": 0.7798, + "step": 50019 + }, + { + "epoch": 3.22, + "learning_rate": 9.287876510874e-06, + "loss": 0.7714, + "step": 50020 + }, + { + "epoch": 3.22, + "learning_rate": 9.28766182185105e-06, + "loss": 0.8595, + "step": 50021 + }, + { + "epoch": 3.22, + "learning_rate": 9.287447132828099e-06, + "loss": 0.8267, + "step": 50022 + }, + { + "epoch": 3.22, + "learning_rate": 9.287232443805148e-06, + "loss": 0.7467, + "step": 50023 + }, + { + "epoch": 3.22, + "learning_rate": 9.287017754782199e-06, + "loss": 0.8459, + "step": 50024 + }, + { + "epoch": 3.22, + "learning_rate": 9.28680306575925e-06, + "loss": 0.8252, + "step": 50025 + }, + { + "epoch": 3.22, + "learning_rate": 9.286588376736298e-06, + "loss": 0.7546, + "step": 50026 + }, + { + "epoch": 3.22, + "learning_rate": 9.286373687713349e-06, + "loss": 0.8091, + "step": 50027 + }, + { + "epoch": 3.22, + "learning_rate": 9.286158998690398e-06, + "loss": 0.8045, + "step": 50028 + }, + { + "epoch": 3.22, + "learning_rate": 9.285944309667447e-06, + "loss": 0.8195, + "step": 50029 + }, + { + "epoch": 3.22, + "learning_rate": 9.285729620644497e-06, + "loss": 0.8206, + "step": 50030 + }, + { + "epoch": 3.22, + "learning_rate": 9.285514931621546e-06, + "loss": 0.8279, + "step": 50031 + }, + { + "epoch": 3.22, + "learning_rate": 9.285300242598597e-06, + "loss": 0.907, + "step": 50032 + }, + { + "epoch": 3.22, + "learning_rate": 9.285085553575646e-06, + "loss": 0.8931, + "step": 50033 + }, + { + "epoch": 3.22, + "learning_rate": 9.284870864552697e-06, + "loss": 0.7805, + "step": 50034 + }, + { + "epoch": 3.22, + "learning_rate": 9.284656175529746e-06, + "loss": 0.8515, + "step": 50035 + }, + { + "epoch": 3.22, + "learning_rate": 9.284441486506796e-06, + "loss": 0.7653, + "step": 50036 + }, + { + "epoch": 3.22, + "learning_rate": 9.284226797483845e-06, + "loss": 0.8162, + "step": 50037 + }, + { + "epoch": 3.22, + "learning_rate": 9.284012108460894e-06, + "loss": 0.7891, + "step": 50038 + }, + { + "epoch": 3.22, + "learning_rate": 9.283797419437945e-06, + "loss": 0.8333, + "step": 50039 + }, + { + "epoch": 3.22, + "learning_rate": 9.283582730414996e-06, + "loss": 0.822, + "step": 50040 + }, + { + "epoch": 3.22, + "learning_rate": 9.283368041392044e-06, + "loss": 0.8515, + "step": 50041 + }, + { + "epoch": 3.22, + "learning_rate": 9.283153352369093e-06, + "loss": 0.8093, + "step": 50042 + }, + { + "epoch": 3.22, + "learning_rate": 9.282938663346144e-06, + "loss": 0.8609, + "step": 50043 + }, + { + "epoch": 3.22, + "learning_rate": 9.282723974323193e-06, + "loss": 0.7741, + "step": 50044 + }, + { + "epoch": 3.22, + "learning_rate": 9.282509285300244e-06, + "loss": 0.8578, + "step": 50045 + }, + { + "epoch": 3.22, + "learning_rate": 9.282294596277293e-06, + "loss": 0.8174, + "step": 50046 + }, + { + "epoch": 3.22, + "learning_rate": 9.282079907254343e-06, + "loss": 0.7942, + "step": 50047 + }, + { + "epoch": 3.22, + "learning_rate": 9.281865218231392e-06, + "loss": 0.8638, + "step": 50048 + }, + { + "epoch": 3.22, + "learning_rate": 9.281650529208443e-06, + "loss": 0.8398, + "step": 50049 + }, + { + "epoch": 3.22, + "learning_rate": 9.281435840185492e-06, + "loss": 0.8263, + "step": 50050 + }, + { + "epoch": 3.22, + "learning_rate": 9.281221151162541e-06, + "loss": 0.8069, + "step": 50051 + }, + { + "epoch": 3.22, + "learning_rate": 9.281006462139592e-06, + "loss": 0.8471, + "step": 50052 + }, + { + "epoch": 3.22, + "learning_rate": 9.28079177311664e-06, + "loss": 0.8233, + "step": 50053 + }, + { + "epoch": 3.22, + "learning_rate": 9.280577084093691e-06, + "loss": 0.803, + "step": 50054 + }, + { + "epoch": 3.22, + "learning_rate": 9.280362395070742e-06, + "loss": 0.8062, + "step": 50055 + }, + { + "epoch": 3.22, + "learning_rate": 9.28014770604779e-06, + "loss": 0.8064, + "step": 50056 + }, + { + "epoch": 3.22, + "learning_rate": 9.27993301702484e-06, + "loss": 0.8387, + "step": 50057 + }, + { + "epoch": 3.22, + "learning_rate": 9.27971832800189e-06, + "loss": 0.8348, + "step": 50058 + }, + { + "epoch": 3.22, + "learning_rate": 9.27950363897894e-06, + "loss": 0.7838, + "step": 50059 + }, + { + "epoch": 3.22, + "learning_rate": 9.279288949955988e-06, + "loss": 0.8326, + "step": 50060 + }, + { + "epoch": 3.22, + "learning_rate": 9.279074260933039e-06, + "loss": 0.7873, + "step": 50061 + }, + { + "epoch": 3.22, + "learning_rate": 9.27885957191009e-06, + "loss": 0.8314, + "step": 50062 + }, + { + "epoch": 3.22, + "learning_rate": 9.278644882887139e-06, + "loss": 0.8214, + "step": 50063 + }, + { + "epoch": 3.22, + "learning_rate": 9.27843019386419e-06, + "loss": 0.7992, + "step": 50064 + }, + { + "epoch": 3.22, + "learning_rate": 9.278215504841238e-06, + "loss": 0.8379, + "step": 50065 + }, + { + "epoch": 3.22, + "learning_rate": 9.278000815818287e-06, + "loss": 0.8187, + "step": 50066 + }, + { + "epoch": 3.22, + "learning_rate": 9.277786126795338e-06, + "loss": 0.7895, + "step": 50067 + }, + { + "epoch": 3.22, + "learning_rate": 9.277571437772389e-06, + "loss": 0.8617, + "step": 50068 + }, + { + "epoch": 3.22, + "learning_rate": 9.277356748749438e-06, + "loss": 0.8324, + "step": 50069 + }, + { + "epoch": 3.22, + "learning_rate": 9.277142059726486e-06, + "loss": 0.8421, + "step": 50070 + }, + { + "epoch": 3.22, + "learning_rate": 9.276927370703537e-06, + "loss": 0.8502, + "step": 50071 + }, + { + "epoch": 3.22, + "learning_rate": 9.276712681680586e-06, + "loss": 0.8193, + "step": 50072 + }, + { + "epoch": 3.22, + "learning_rate": 9.276497992657637e-06, + "loss": 0.7909, + "step": 50073 + }, + { + "epoch": 3.22, + "learning_rate": 9.276283303634686e-06, + "loss": 0.776, + "step": 50074 + }, + { + "epoch": 3.22, + "learning_rate": 9.276068614611735e-06, + "loss": 0.9231, + "step": 50075 + }, + { + "epoch": 3.22, + "learning_rate": 9.275853925588785e-06, + "loss": 0.8165, + "step": 50076 + }, + { + "epoch": 3.22, + "learning_rate": 9.275639236565836e-06, + "loss": 0.8377, + "step": 50077 + }, + { + "epoch": 3.22, + "learning_rate": 9.275424547542885e-06, + "loss": 0.8015, + "step": 50078 + }, + { + "epoch": 3.22, + "learning_rate": 9.275209858519934e-06, + "loss": 0.7943, + "step": 50079 + }, + { + "epoch": 3.22, + "learning_rate": 9.274995169496985e-06, + "loss": 0.7798, + "step": 50080 + }, + { + "epoch": 3.22, + "learning_rate": 9.274780480474034e-06, + "loss": 0.7699, + "step": 50081 + }, + { + "epoch": 3.22, + "learning_rate": 9.274565791451084e-06, + "loss": 0.7807, + "step": 50082 + }, + { + "epoch": 3.22, + "learning_rate": 9.274351102428135e-06, + "loss": 0.837, + "step": 50083 + }, + { + "epoch": 3.22, + "learning_rate": 9.274136413405184e-06, + "loss": 0.7943, + "step": 50084 + }, + { + "epoch": 3.22, + "learning_rate": 9.273921724382233e-06, + "loss": 0.8204, + "step": 50085 + }, + { + "epoch": 3.22, + "learning_rate": 9.273707035359283e-06, + "loss": 0.8146, + "step": 50086 + }, + { + "epoch": 3.22, + "learning_rate": 9.273492346336332e-06, + "loss": 0.8033, + "step": 50087 + }, + { + "epoch": 3.22, + "learning_rate": 9.273277657313381e-06, + "loss": 0.8659, + "step": 50088 + }, + { + "epoch": 3.22, + "learning_rate": 9.273062968290432e-06, + "loss": 0.8313, + "step": 50089 + }, + { + "epoch": 3.22, + "learning_rate": 9.272848279267483e-06, + "loss": 0.8338, + "step": 50090 + }, + { + "epoch": 3.22, + "learning_rate": 9.272633590244532e-06, + "loss": 0.8234, + "step": 50091 + }, + { + "epoch": 3.22, + "learning_rate": 9.272418901221582e-06, + "loss": 0.7978, + "step": 50092 + }, + { + "epoch": 3.22, + "learning_rate": 9.272204212198631e-06, + "loss": 0.7785, + "step": 50093 + }, + { + "epoch": 3.22, + "learning_rate": 9.27198952317568e-06, + "loss": 0.8658, + "step": 50094 + }, + { + "epoch": 3.22, + "learning_rate": 9.271774834152731e-06, + "loss": 0.8286, + "step": 50095 + }, + { + "epoch": 3.22, + "learning_rate": 9.27156014512978e-06, + "loss": 0.8487, + "step": 50096 + }, + { + "epoch": 3.22, + "learning_rate": 9.27134545610683e-06, + "loss": 0.8362, + "step": 50097 + }, + { + "epoch": 3.22, + "learning_rate": 9.27113076708388e-06, + "loss": 0.8471, + "step": 50098 + }, + { + "epoch": 3.22, + "learning_rate": 9.27091607806093e-06, + "loss": 0.8291, + "step": 50099 + }, + { + "epoch": 3.22, + "learning_rate": 9.270701389037979e-06, + "loss": 0.821, + "step": 50100 + }, + { + "epoch": 3.22, + "learning_rate": 9.270486700015028e-06, + "loss": 0.7851, + "step": 50101 + }, + { + "epoch": 3.22, + "learning_rate": 9.270272010992079e-06, + "loss": 0.844, + "step": 50102 + }, + { + "epoch": 3.22, + "learning_rate": 9.270057321969128e-06, + "loss": 0.8421, + "step": 50103 + }, + { + "epoch": 3.22, + "learning_rate": 9.269842632946178e-06, + "loss": 0.8202, + "step": 50104 + }, + { + "epoch": 3.22, + "learning_rate": 9.269627943923229e-06, + "loss": 0.8315, + "step": 50105 + }, + { + "epoch": 3.22, + "learning_rate": 9.269413254900278e-06, + "loss": 0.8136, + "step": 50106 + }, + { + "epoch": 3.22, + "learning_rate": 9.269198565877327e-06, + "loss": 0.8114, + "step": 50107 + }, + { + "epoch": 3.22, + "learning_rate": 9.268983876854378e-06, + "loss": 0.85, + "step": 50108 + }, + { + "epoch": 3.22, + "learning_rate": 9.268769187831427e-06, + "loss": 0.8372, + "step": 50109 + }, + { + "epoch": 3.22, + "learning_rate": 9.268554498808476e-06, + "loss": 0.8396, + "step": 50110 + }, + { + "epoch": 3.22, + "learning_rate": 9.268339809785526e-06, + "loss": 0.8027, + "step": 50111 + }, + { + "epoch": 3.22, + "learning_rate": 9.268125120762577e-06, + "loss": 0.8416, + "step": 50112 + }, + { + "epoch": 3.22, + "learning_rate": 9.267910431739626e-06, + "loss": 0.827, + "step": 50113 + }, + { + "epoch": 3.22, + "learning_rate": 9.267695742716677e-06, + "loss": 0.9306, + "step": 50114 + }, + { + "epoch": 3.22, + "learning_rate": 9.267481053693725e-06, + "loss": 0.7652, + "step": 50115 + }, + { + "epoch": 3.22, + "learning_rate": 9.267266364670774e-06, + "loss": 0.8455, + "step": 50116 + }, + { + "epoch": 3.22, + "learning_rate": 9.267051675647825e-06, + "loss": 0.8381, + "step": 50117 + }, + { + "epoch": 3.22, + "learning_rate": 9.266836986624874e-06, + "loss": 0.8597, + "step": 50118 + }, + { + "epoch": 3.22, + "learning_rate": 9.266622297601925e-06, + "loss": 0.8613, + "step": 50119 + }, + { + "epoch": 3.22, + "learning_rate": 9.266407608578974e-06, + "loss": 0.8388, + "step": 50120 + }, + { + "epoch": 3.22, + "learning_rate": 9.266192919556024e-06, + "loss": 0.8763, + "step": 50121 + }, + { + "epoch": 3.22, + "learning_rate": 9.265978230533073e-06, + "loss": 0.8893, + "step": 50122 + }, + { + "epoch": 3.22, + "learning_rate": 9.265763541510124e-06, + "loss": 0.8131, + "step": 50123 + }, + { + "epoch": 3.22, + "learning_rate": 9.265548852487173e-06, + "loss": 0.8496, + "step": 50124 + }, + { + "epoch": 3.22, + "learning_rate": 9.265334163464222e-06, + "loss": 0.8085, + "step": 50125 + }, + { + "epoch": 3.22, + "learning_rate": 9.265119474441273e-06, + "loss": 0.8157, + "step": 50126 + }, + { + "epoch": 3.23, + "learning_rate": 9.264904785418323e-06, + "loss": 0.8623, + "step": 50127 + }, + { + "epoch": 3.23, + "learning_rate": 9.264690096395372e-06, + "loss": 0.8513, + "step": 50128 + }, + { + "epoch": 3.23, + "learning_rate": 9.264475407372421e-06, + "loss": 0.8123, + "step": 50129 + }, + { + "epoch": 3.23, + "learning_rate": 9.264260718349472e-06, + "loss": 0.8204, + "step": 50130 + }, + { + "epoch": 3.23, + "learning_rate": 9.26404602932652e-06, + "loss": 0.8169, + "step": 50131 + }, + { + "epoch": 3.23, + "learning_rate": 9.263831340303571e-06, + "loss": 0.8313, + "step": 50132 + }, + { + "epoch": 3.23, + "learning_rate": 9.26361665128062e-06, + "loss": 0.8275, + "step": 50133 + }, + { + "epoch": 3.23, + "learning_rate": 9.263401962257671e-06, + "loss": 0.8642, + "step": 50134 + }, + { + "epoch": 3.23, + "learning_rate": 9.26318727323472e-06, + "loss": 0.7975, + "step": 50135 + }, + { + "epoch": 3.23, + "learning_rate": 9.26297258421177e-06, + "loss": 0.8253, + "step": 50136 + }, + { + "epoch": 3.23, + "learning_rate": 9.26275789518882e-06, + "loss": 0.823, + "step": 50137 + }, + { + "epoch": 3.23, + "learning_rate": 9.262543206165869e-06, + "loss": 0.8498, + "step": 50138 + }, + { + "epoch": 3.23, + "learning_rate": 9.26232851714292e-06, + "loss": 0.7726, + "step": 50139 + }, + { + "epoch": 3.23, + "learning_rate": 9.262113828119968e-06, + "loss": 0.8163, + "step": 50140 + }, + { + "epoch": 3.23, + "learning_rate": 9.261899139097019e-06, + "loss": 0.7997, + "step": 50141 + }, + { + "epoch": 3.23, + "learning_rate": 9.26168445007407e-06, + "loss": 0.7816, + "step": 50142 + }, + { + "epoch": 3.23, + "learning_rate": 9.261469761051119e-06, + "loss": 0.8006, + "step": 50143 + }, + { + "epoch": 3.23, + "learning_rate": 9.261255072028167e-06, + "loss": 0.8449, + "step": 50144 + }, + { + "epoch": 3.23, + "learning_rate": 9.261040383005218e-06, + "loss": 0.8525, + "step": 50145 + }, + { + "epoch": 3.23, + "learning_rate": 9.260825693982267e-06, + "loss": 0.7945, + "step": 50146 + }, + { + "epoch": 3.23, + "learning_rate": 9.260611004959316e-06, + "loss": 0.8085, + "step": 50147 + }, + { + "epoch": 3.23, + "learning_rate": 9.260396315936367e-06, + "loss": 0.8321, + "step": 50148 + }, + { + "epoch": 3.23, + "learning_rate": 9.260181626913417e-06, + "loss": 0.8088, + "step": 50149 + }, + { + "epoch": 3.23, + "learning_rate": 9.259966937890466e-06, + "loss": 0.8367, + "step": 50150 + }, + { + "epoch": 3.23, + "learning_rate": 9.259752248867517e-06, + "loss": 0.8053, + "step": 50151 + }, + { + "epoch": 3.23, + "learning_rate": 9.259537559844566e-06, + "loss": 0.8894, + "step": 50152 + }, + { + "epoch": 3.23, + "learning_rate": 9.259322870821615e-06, + "loss": 0.768, + "step": 50153 + }, + { + "epoch": 3.23, + "learning_rate": 9.259108181798666e-06, + "loss": 0.8581, + "step": 50154 + }, + { + "epoch": 3.23, + "learning_rate": 9.258893492775715e-06, + "loss": 0.8676, + "step": 50155 + }, + { + "epoch": 3.23, + "learning_rate": 9.258678803752765e-06, + "loss": 0.7819, + "step": 50156 + }, + { + "epoch": 3.23, + "learning_rate": 9.258464114729814e-06, + "loss": 0.8178, + "step": 50157 + }, + { + "epoch": 3.23, + "learning_rate": 9.258249425706865e-06, + "loss": 0.8226, + "step": 50158 + }, + { + "epoch": 3.23, + "learning_rate": 9.258034736683914e-06, + "loss": 0.8424, + "step": 50159 + }, + { + "epoch": 3.23, + "learning_rate": 9.257820047660964e-06, + "loss": 0.8209, + "step": 50160 + }, + { + "epoch": 3.23, + "learning_rate": 9.257605358638013e-06, + "loss": 0.8528, + "step": 50161 + }, + { + "epoch": 3.23, + "learning_rate": 9.257390669615062e-06, + "loss": 0.8616, + "step": 50162 + }, + { + "epoch": 3.23, + "learning_rate": 9.257175980592113e-06, + "loss": 0.8058, + "step": 50163 + }, + { + "epoch": 3.23, + "learning_rate": 9.256961291569164e-06, + "loss": 0.858, + "step": 50164 + }, + { + "epoch": 3.23, + "learning_rate": 9.256746602546213e-06, + "loss": 0.7948, + "step": 50165 + }, + { + "epoch": 3.23, + "learning_rate": 9.256531913523262e-06, + "loss": 0.7912, + "step": 50166 + }, + { + "epoch": 3.23, + "learning_rate": 9.256317224500312e-06, + "loss": 0.8185, + "step": 50167 + }, + { + "epoch": 3.23, + "learning_rate": 9.256102535477361e-06, + "loss": 0.8011, + "step": 50168 + }, + { + "epoch": 3.23, + "learning_rate": 9.25588784645441e-06, + "loss": 0.8041, + "step": 50169 + }, + { + "epoch": 3.23, + "learning_rate": 9.255673157431463e-06, + "loss": 0.8685, + "step": 50170 + }, + { + "epoch": 3.23, + "learning_rate": 9.255458468408512e-06, + "loss": 0.8575, + "step": 50171 + }, + { + "epoch": 3.23, + "learning_rate": 9.25524377938556e-06, + "loss": 0.8107, + "step": 50172 + }, + { + "epoch": 3.23, + "learning_rate": 9.255029090362611e-06, + "loss": 0.7737, + "step": 50173 + }, + { + "epoch": 3.23, + "learning_rate": 9.25481440133966e-06, + "loss": 0.8439, + "step": 50174 + }, + { + "epoch": 3.23, + "learning_rate": 9.254599712316709e-06, + "loss": 0.7776, + "step": 50175 + }, + { + "epoch": 3.23, + "learning_rate": 9.25438502329376e-06, + "loss": 0.7966, + "step": 50176 + }, + { + "epoch": 3.23, + "learning_rate": 9.25417033427081e-06, + "loss": 0.7931, + "step": 50177 + }, + { + "epoch": 3.23, + "learning_rate": 9.25395564524786e-06, + "loss": 0.783, + "step": 50178 + }, + { + "epoch": 3.23, + "learning_rate": 9.253740956224908e-06, + "loss": 0.8388, + "step": 50179 + }, + { + "epoch": 3.23, + "learning_rate": 9.253526267201959e-06, + "loss": 0.8231, + "step": 50180 + }, + { + "epoch": 3.23, + "learning_rate": 9.253311578179008e-06, + "loss": 0.8302, + "step": 50181 + }, + { + "epoch": 3.23, + "learning_rate": 9.253096889156059e-06, + "loss": 0.834, + "step": 50182 + }, + { + "epoch": 3.23, + "learning_rate": 9.252882200133108e-06, + "loss": 0.8876, + "step": 50183 + }, + { + "epoch": 3.23, + "learning_rate": 9.252667511110157e-06, + "loss": 0.8417, + "step": 50184 + }, + { + "epoch": 3.23, + "learning_rate": 9.252452822087207e-06, + "loss": 0.8178, + "step": 50185 + }, + { + "epoch": 3.23, + "learning_rate": 9.252238133064258e-06, + "loss": 0.8651, + "step": 50186 + }, + { + "epoch": 3.23, + "learning_rate": 9.252023444041307e-06, + "loss": 0.8259, + "step": 50187 + }, + { + "epoch": 3.23, + "learning_rate": 9.251808755018356e-06, + "loss": 0.8488, + "step": 50188 + }, + { + "epoch": 3.23, + "learning_rate": 9.251594065995406e-06, + "loss": 0.7741, + "step": 50189 + }, + { + "epoch": 3.23, + "learning_rate": 9.251379376972455e-06, + "loss": 0.8589, + "step": 50190 + }, + { + "epoch": 3.23, + "learning_rate": 9.251164687949506e-06, + "loss": 0.7439, + "step": 50191 + }, + { + "epoch": 3.23, + "learning_rate": 9.250949998926557e-06, + "loss": 0.8509, + "step": 50192 + }, + { + "epoch": 3.23, + "learning_rate": 9.250735309903606e-06, + "loss": 0.8698, + "step": 50193 + }, + { + "epoch": 3.23, + "learning_rate": 9.250520620880655e-06, + "loss": 0.8245, + "step": 50194 + }, + { + "epoch": 3.23, + "learning_rate": 9.250305931857705e-06, + "loss": 0.8278, + "step": 50195 + }, + { + "epoch": 3.23, + "learning_rate": 9.250091242834754e-06, + "loss": 0.838, + "step": 50196 + }, + { + "epoch": 3.23, + "learning_rate": 9.249876553811803e-06, + "loss": 0.7955, + "step": 50197 + }, + { + "epoch": 3.23, + "learning_rate": 9.249661864788854e-06, + "loss": 0.8253, + "step": 50198 + }, + { + "epoch": 3.23, + "learning_rate": 9.249447175765905e-06, + "loss": 0.8577, + "step": 50199 + }, + { + "epoch": 3.23, + "learning_rate": 9.249232486742954e-06, + "loss": 0.7739, + "step": 50200 + }, + { + "epoch": 3.23, + "learning_rate": 9.249017797720004e-06, + "loss": 0.8702, + "step": 50201 + }, + { + "epoch": 3.23, + "learning_rate": 9.248803108697053e-06, + "loss": 0.7419, + "step": 50202 + }, + { + "epoch": 3.23, + "learning_rate": 9.248588419674102e-06, + "loss": 0.8652, + "step": 50203 + }, + { + "epoch": 3.23, + "learning_rate": 9.248373730651153e-06, + "loss": 0.8138, + "step": 50204 + }, + { + "epoch": 3.23, + "learning_rate": 9.248159041628202e-06, + "loss": 0.8375, + "step": 50205 + }, + { + "epoch": 3.23, + "learning_rate": 9.247944352605252e-06, + "loss": 0.8216, + "step": 50206 + }, + { + "epoch": 3.23, + "learning_rate": 9.247729663582301e-06, + "loss": 0.799, + "step": 50207 + }, + { + "epoch": 3.23, + "learning_rate": 9.247514974559352e-06, + "loss": 0.8213, + "step": 50208 + }, + { + "epoch": 3.23, + "learning_rate": 9.247300285536401e-06, + "loss": 0.828, + "step": 50209 + }, + { + "epoch": 3.23, + "learning_rate": 9.247085596513452e-06, + "loss": 0.822, + "step": 50210 + }, + { + "epoch": 3.23, + "learning_rate": 9.2468709074905e-06, + "loss": 0.8257, + "step": 50211 + }, + { + "epoch": 3.23, + "learning_rate": 9.24665621846755e-06, + "loss": 0.7986, + "step": 50212 + }, + { + "epoch": 3.23, + "learning_rate": 9.2464415294446e-06, + "loss": 0.8337, + "step": 50213 + }, + { + "epoch": 3.23, + "learning_rate": 9.246226840421651e-06, + "loss": 0.8253, + "step": 50214 + }, + { + "epoch": 3.23, + "learning_rate": 9.2460121513987e-06, + "loss": 0.8383, + "step": 50215 + }, + { + "epoch": 3.23, + "learning_rate": 9.245797462375749e-06, + "loss": 0.8483, + "step": 50216 + }, + { + "epoch": 3.23, + "learning_rate": 9.2455827733528e-06, + "loss": 0.8077, + "step": 50217 + }, + { + "epoch": 3.23, + "learning_rate": 9.245368084329848e-06, + "loss": 0.77, + "step": 50218 + }, + { + "epoch": 3.23, + "learning_rate": 9.245153395306899e-06, + "loss": 0.9066, + "step": 50219 + }, + { + "epoch": 3.23, + "learning_rate": 9.244938706283948e-06, + "loss": 0.8361, + "step": 50220 + }, + { + "epoch": 3.23, + "learning_rate": 9.244724017260999e-06, + "loss": 0.7565, + "step": 50221 + }, + { + "epoch": 3.23, + "learning_rate": 9.244509328238048e-06, + "loss": 0.8222, + "step": 50222 + }, + { + "epoch": 3.23, + "learning_rate": 9.244294639215098e-06, + "loss": 0.8293, + "step": 50223 + }, + { + "epoch": 3.23, + "learning_rate": 9.244079950192147e-06, + "loss": 0.8607, + "step": 50224 + }, + { + "epoch": 3.23, + "learning_rate": 9.243865261169196e-06, + "loss": 0.8034, + "step": 50225 + }, + { + "epoch": 3.23, + "learning_rate": 9.243650572146247e-06, + "loss": 0.8303, + "step": 50226 + }, + { + "epoch": 3.23, + "learning_rate": 9.243435883123296e-06, + "loss": 0.8264, + "step": 50227 + }, + { + "epoch": 3.23, + "learning_rate": 9.243221194100347e-06, + "loss": 0.8029, + "step": 50228 + }, + { + "epoch": 3.23, + "learning_rate": 9.243006505077397e-06, + "loss": 0.8107, + "step": 50229 + }, + { + "epoch": 3.23, + "learning_rate": 9.242791816054446e-06, + "loss": 0.8627, + "step": 50230 + }, + { + "epoch": 3.23, + "learning_rate": 9.242577127031495e-06, + "loss": 0.8741, + "step": 50231 + }, + { + "epoch": 3.23, + "learning_rate": 9.242362438008546e-06, + "loss": 0.8271, + "step": 50232 + }, + { + "epoch": 3.23, + "learning_rate": 9.242147748985595e-06, + "loss": 0.8197, + "step": 50233 + }, + { + "epoch": 3.23, + "learning_rate": 9.241933059962644e-06, + "loss": 0.8822, + "step": 50234 + }, + { + "epoch": 3.23, + "learning_rate": 9.241718370939694e-06, + "loss": 0.7928, + "step": 50235 + }, + { + "epoch": 3.23, + "learning_rate": 9.241503681916745e-06, + "loss": 0.7925, + "step": 50236 + }, + { + "epoch": 3.23, + "learning_rate": 9.241288992893794e-06, + "loss": 0.7984, + "step": 50237 + }, + { + "epoch": 3.23, + "learning_rate": 9.241074303870845e-06, + "loss": 0.861, + "step": 50238 + }, + { + "epoch": 3.23, + "learning_rate": 9.240859614847894e-06, + "loss": 0.8465, + "step": 50239 + }, + { + "epoch": 3.23, + "learning_rate": 9.240644925824943e-06, + "loss": 0.9258, + "step": 50240 + }, + { + "epoch": 3.23, + "learning_rate": 9.240430236801993e-06, + "loss": 0.8274, + "step": 50241 + }, + { + "epoch": 3.23, + "learning_rate": 9.240215547779042e-06, + "loss": 0.8483, + "step": 50242 + }, + { + "epoch": 3.23, + "learning_rate": 9.240000858756093e-06, + "loss": 0.7925, + "step": 50243 + }, + { + "epoch": 3.23, + "learning_rate": 9.239786169733142e-06, + "loss": 0.8331, + "step": 50244 + }, + { + "epoch": 3.23, + "learning_rate": 9.239571480710193e-06, + "loss": 0.8026, + "step": 50245 + }, + { + "epoch": 3.23, + "learning_rate": 9.239356791687242e-06, + "loss": 0.8047, + "step": 50246 + }, + { + "epoch": 3.23, + "learning_rate": 9.239142102664292e-06, + "loss": 0.8207, + "step": 50247 + }, + { + "epoch": 3.23, + "learning_rate": 9.238927413641341e-06, + "loss": 0.8539, + "step": 50248 + }, + { + "epoch": 3.23, + "learning_rate": 9.23871272461839e-06, + "loss": 0.7959, + "step": 50249 + }, + { + "epoch": 3.23, + "learning_rate": 9.23849803559544e-06, + "loss": 0.7977, + "step": 50250 + }, + { + "epoch": 3.23, + "learning_rate": 9.238283346572491e-06, + "loss": 0.8273, + "step": 50251 + }, + { + "epoch": 3.23, + "learning_rate": 9.23806865754954e-06, + "loss": 0.8182, + "step": 50252 + }, + { + "epoch": 3.23, + "learning_rate": 9.23785396852659e-06, + "loss": 0.8219, + "step": 50253 + }, + { + "epoch": 3.23, + "learning_rate": 9.23763927950364e-06, + "loss": 0.7891, + "step": 50254 + }, + { + "epoch": 3.23, + "learning_rate": 9.237424590480689e-06, + "loss": 0.8529, + "step": 50255 + }, + { + "epoch": 3.23, + "learning_rate": 9.237209901457738e-06, + "loss": 0.7615, + "step": 50256 + }, + { + "epoch": 3.23, + "learning_rate": 9.23699521243479e-06, + "loss": 0.8124, + "step": 50257 + }, + { + "epoch": 3.23, + "learning_rate": 9.23678052341184e-06, + "loss": 0.8606, + "step": 50258 + }, + { + "epoch": 3.23, + "learning_rate": 9.236565834388888e-06, + "loss": 0.8529, + "step": 50259 + }, + { + "epoch": 3.23, + "learning_rate": 9.236351145365939e-06, + "loss": 0.8005, + "step": 50260 + }, + { + "epoch": 3.23, + "learning_rate": 9.236136456342988e-06, + "loss": 0.759, + "step": 50261 + }, + { + "epoch": 3.23, + "learning_rate": 9.235921767320037e-06, + "loss": 0.8329, + "step": 50262 + }, + { + "epoch": 3.23, + "learning_rate": 9.235707078297087e-06, + "loss": 0.8026, + "step": 50263 + }, + { + "epoch": 3.23, + "learning_rate": 9.235492389274138e-06, + "loss": 0.835, + "step": 50264 + }, + { + "epoch": 3.23, + "learning_rate": 9.235277700251187e-06, + "loss": 0.8245, + "step": 50265 + }, + { + "epoch": 3.23, + "learning_rate": 9.235063011228236e-06, + "loss": 0.8492, + "step": 50266 + }, + { + "epoch": 3.23, + "learning_rate": 9.234848322205287e-06, + "loss": 0.8791, + "step": 50267 + }, + { + "epoch": 3.23, + "learning_rate": 9.234633633182336e-06, + "loss": 0.8642, + "step": 50268 + }, + { + "epoch": 3.23, + "learning_rate": 9.234418944159386e-06, + "loss": 0.8477, + "step": 50269 + }, + { + "epoch": 3.23, + "learning_rate": 9.234204255136435e-06, + "loss": 0.8617, + "step": 50270 + }, + { + "epoch": 3.23, + "learning_rate": 9.233989566113484e-06, + "loss": 0.7755, + "step": 50271 + }, + { + "epoch": 3.23, + "learning_rate": 9.233774877090535e-06, + "loss": 0.7938, + "step": 50272 + }, + { + "epoch": 3.23, + "learning_rate": 9.233560188067586e-06, + "loss": 0.8413, + "step": 50273 + }, + { + "epoch": 3.23, + "learning_rate": 9.233345499044635e-06, + "loss": 0.8652, + "step": 50274 + }, + { + "epoch": 3.23, + "learning_rate": 9.233130810021684e-06, + "loss": 0.8373, + "step": 50275 + }, + { + "epoch": 3.23, + "learning_rate": 9.232916120998734e-06, + "loss": 0.8399, + "step": 50276 + }, + { + "epoch": 3.23, + "learning_rate": 9.232701431975783e-06, + "loss": 0.8084, + "step": 50277 + }, + { + "epoch": 3.23, + "learning_rate": 9.232486742952834e-06, + "loss": 0.8525, + "step": 50278 + }, + { + "epoch": 3.23, + "learning_rate": 9.232272053929884e-06, + "loss": 0.7978, + "step": 50279 + }, + { + "epoch": 3.23, + "learning_rate": 9.232057364906933e-06, + "loss": 0.8823, + "step": 50280 + }, + { + "epoch": 3.23, + "learning_rate": 9.231842675883982e-06, + "loss": 0.914, + "step": 50281 + }, + { + "epoch": 3.24, + "learning_rate": 9.231627986861033e-06, + "loss": 0.823, + "step": 50282 + }, + { + "epoch": 3.24, + "learning_rate": 9.231413297838082e-06, + "loss": 0.8311, + "step": 50283 + }, + { + "epoch": 3.24, + "learning_rate": 9.231198608815131e-06, + "loss": 0.8593, + "step": 50284 + }, + { + "epoch": 3.24, + "learning_rate": 9.230983919792182e-06, + "loss": 0.8325, + "step": 50285 + }, + { + "epoch": 3.24, + "learning_rate": 9.230769230769232e-06, + "loss": 0.8049, + "step": 50286 + }, + { + "epoch": 3.24, + "learning_rate": 9.230554541746281e-06, + "loss": 0.8229, + "step": 50287 + }, + { + "epoch": 3.24, + "learning_rate": 9.230339852723332e-06, + "loss": 0.8748, + "step": 50288 + }, + { + "epoch": 3.24, + "learning_rate": 9.230125163700381e-06, + "loss": 0.8865, + "step": 50289 + }, + { + "epoch": 3.24, + "learning_rate": 9.22991047467743e-06, + "loss": 0.8349, + "step": 50290 + }, + { + "epoch": 3.24, + "learning_rate": 9.22969578565448e-06, + "loss": 0.7939, + "step": 50291 + }, + { + "epoch": 3.24, + "learning_rate": 9.22948109663153e-06, + "loss": 0.7646, + "step": 50292 + }, + { + "epoch": 3.24, + "learning_rate": 9.22926640760858e-06, + "loss": 0.8018, + "step": 50293 + }, + { + "epoch": 3.24, + "learning_rate": 9.229051718585629e-06, + "loss": 0.7504, + "step": 50294 + }, + { + "epoch": 3.24, + "learning_rate": 9.22883702956268e-06, + "loss": 0.8201, + "step": 50295 + }, + { + "epoch": 3.24, + "learning_rate": 9.228622340539729e-06, + "loss": 0.8649, + "step": 50296 + }, + { + "epoch": 3.24, + "learning_rate": 9.22840765151678e-06, + "loss": 0.8175, + "step": 50297 + }, + { + "epoch": 3.24, + "learning_rate": 9.228192962493828e-06, + "loss": 0.8689, + "step": 50298 + }, + { + "epoch": 3.24, + "learning_rate": 9.227978273470877e-06, + "loss": 0.8775, + "step": 50299 + }, + { + "epoch": 3.24, + "learning_rate": 9.227763584447928e-06, + "loss": 0.8344, + "step": 50300 + }, + { + "epoch": 3.24, + "learning_rate": 9.227548895424979e-06, + "loss": 0.8271, + "step": 50301 + }, + { + "epoch": 3.24, + "learning_rate": 9.227334206402028e-06, + "loss": 0.9033, + "step": 50302 + }, + { + "epoch": 3.24, + "learning_rate": 9.227119517379077e-06, + "loss": 0.8603, + "step": 50303 + }, + { + "epoch": 3.24, + "learning_rate": 9.226904828356127e-06, + "loss": 0.853, + "step": 50304 + }, + { + "epoch": 3.24, + "learning_rate": 9.226690139333176e-06, + "loss": 0.8887, + "step": 50305 + }, + { + "epoch": 3.24, + "learning_rate": 9.226475450310227e-06, + "loss": 0.8149, + "step": 50306 + }, + { + "epoch": 3.24, + "learning_rate": 9.226260761287276e-06, + "loss": 0.8695, + "step": 50307 + }, + { + "epoch": 3.24, + "learning_rate": 9.226046072264326e-06, + "loss": 0.8313, + "step": 50308 + }, + { + "epoch": 3.24, + "learning_rate": 9.225831383241375e-06, + "loss": 0.8017, + "step": 50309 + }, + { + "epoch": 3.24, + "learning_rate": 9.225616694218426e-06, + "loss": 0.8598, + "step": 50310 + }, + { + "epoch": 3.24, + "learning_rate": 9.225402005195475e-06, + "loss": 0.8256, + "step": 50311 + }, + { + "epoch": 3.24, + "learning_rate": 9.225187316172524e-06, + "loss": 0.8781, + "step": 50312 + }, + { + "epoch": 3.24, + "learning_rate": 9.224972627149575e-06, + "loss": 0.7916, + "step": 50313 + }, + { + "epoch": 3.24, + "learning_rate": 9.224757938126624e-06, + "loss": 0.7903, + "step": 50314 + }, + { + "epoch": 3.24, + "learning_rate": 9.224543249103674e-06, + "loss": 0.8907, + "step": 50315 + }, + { + "epoch": 3.24, + "learning_rate": 9.224328560080725e-06, + "loss": 0.8073, + "step": 50316 + }, + { + "epoch": 3.24, + "learning_rate": 9.224113871057774e-06, + "loss": 0.7642, + "step": 50317 + }, + { + "epoch": 3.24, + "learning_rate": 9.223899182034823e-06, + "loss": 0.8489, + "step": 50318 + }, + { + "epoch": 3.24, + "learning_rate": 9.223684493011874e-06, + "loss": 0.7716, + "step": 50319 + }, + { + "epoch": 3.24, + "learning_rate": 9.223469803988923e-06, + "loss": 0.811, + "step": 50320 + }, + { + "epoch": 3.24, + "learning_rate": 9.223255114965971e-06, + "loss": 0.8568, + "step": 50321 + }, + { + "epoch": 3.24, + "learning_rate": 9.223040425943022e-06, + "loss": 0.8499, + "step": 50322 + }, + { + "epoch": 3.24, + "learning_rate": 9.222825736920073e-06, + "loss": 0.831, + "step": 50323 + }, + { + "epoch": 3.24, + "learning_rate": 9.222611047897122e-06, + "loss": 0.7586, + "step": 50324 + }, + { + "epoch": 3.24, + "learning_rate": 9.222396358874172e-06, + "loss": 0.8409, + "step": 50325 + }, + { + "epoch": 3.24, + "learning_rate": 9.222181669851221e-06, + "loss": 0.7691, + "step": 50326 + }, + { + "epoch": 3.24, + "learning_rate": 9.22196698082827e-06, + "loss": 0.8478, + "step": 50327 + }, + { + "epoch": 3.24, + "learning_rate": 9.221752291805321e-06, + "loss": 0.8459, + "step": 50328 + }, + { + "epoch": 3.24, + "learning_rate": 9.22153760278237e-06, + "loss": 0.8423, + "step": 50329 + }, + { + "epoch": 3.24, + "learning_rate": 9.22132291375942e-06, + "loss": 0.7588, + "step": 50330 + }, + { + "epoch": 3.24, + "learning_rate": 9.22110822473647e-06, + "loss": 0.8651, + "step": 50331 + }, + { + "epoch": 3.24, + "learning_rate": 9.22089353571352e-06, + "loss": 0.8203, + "step": 50332 + }, + { + "epoch": 3.24, + "learning_rate": 9.22067884669057e-06, + "loss": 0.868, + "step": 50333 + }, + { + "epoch": 3.24, + "learning_rate": 9.220464157667618e-06, + "loss": 0.807, + "step": 50334 + }, + { + "epoch": 3.24, + "learning_rate": 9.220249468644669e-06, + "loss": 0.7557, + "step": 50335 + }, + { + "epoch": 3.24, + "learning_rate": 9.220034779621718e-06, + "loss": 0.7894, + "step": 50336 + }, + { + "epoch": 3.24, + "learning_rate": 9.219820090598768e-06, + "loss": 0.8307, + "step": 50337 + }, + { + "epoch": 3.24, + "learning_rate": 9.219605401575819e-06, + "loss": 0.7873, + "step": 50338 + }, + { + "epoch": 3.24, + "learning_rate": 9.219390712552868e-06, + "loss": 0.8252, + "step": 50339 + }, + { + "epoch": 3.24, + "learning_rate": 9.219176023529917e-06, + "loss": 0.8227, + "step": 50340 + }, + { + "epoch": 3.24, + "learning_rate": 9.218961334506968e-06, + "loss": 0.8387, + "step": 50341 + }, + { + "epoch": 3.24, + "learning_rate": 9.218746645484017e-06, + "loss": 0.8722, + "step": 50342 + }, + { + "epoch": 3.24, + "learning_rate": 9.218531956461066e-06, + "loss": 0.8278, + "step": 50343 + }, + { + "epoch": 3.24, + "learning_rate": 9.218317267438118e-06, + "loss": 0.8329, + "step": 50344 + }, + { + "epoch": 3.24, + "learning_rate": 9.218102578415167e-06, + "loss": 0.7602, + "step": 50345 + }, + { + "epoch": 3.24, + "learning_rate": 9.217887889392216e-06, + "loss": 0.78, + "step": 50346 + }, + { + "epoch": 3.24, + "learning_rate": 9.217673200369267e-06, + "loss": 0.783, + "step": 50347 + }, + { + "epoch": 3.24, + "learning_rate": 9.217458511346316e-06, + "loss": 0.8412, + "step": 50348 + }, + { + "epoch": 3.24, + "learning_rate": 9.217243822323365e-06, + "loss": 0.7857, + "step": 50349 + }, + { + "epoch": 3.24, + "learning_rate": 9.217029133300415e-06, + "loss": 0.7837, + "step": 50350 + }, + { + "epoch": 3.24, + "learning_rate": 9.216814444277464e-06, + "loss": 0.8169, + "step": 50351 + }, + { + "epoch": 3.24, + "learning_rate": 9.216599755254515e-06, + "loss": 0.7768, + "step": 50352 + }, + { + "epoch": 3.24, + "learning_rate": 9.216385066231564e-06, + "loss": 0.7868, + "step": 50353 + }, + { + "epoch": 3.24, + "learning_rate": 9.216170377208614e-06, + "loss": 0.8583, + "step": 50354 + }, + { + "epoch": 3.24, + "learning_rate": 9.215955688185663e-06, + "loss": 0.9668, + "step": 50355 + }, + { + "epoch": 3.24, + "learning_rate": 9.215740999162714e-06, + "loss": 0.8151, + "step": 50356 + }, + { + "epoch": 3.24, + "learning_rate": 9.215526310139763e-06, + "loss": 0.7711, + "step": 50357 + }, + { + "epoch": 3.24, + "learning_rate": 9.215311621116812e-06, + "loss": 0.7953, + "step": 50358 + }, + { + "epoch": 3.24, + "learning_rate": 9.215096932093863e-06, + "loss": 0.7487, + "step": 50359 + }, + { + "epoch": 3.24, + "learning_rate": 9.214882243070913e-06, + "loss": 0.838, + "step": 50360 + }, + { + "epoch": 3.24, + "learning_rate": 9.214667554047962e-06, + "loss": 0.8677, + "step": 50361 + }, + { + "epoch": 3.24, + "learning_rate": 9.214452865025011e-06, + "loss": 0.8268, + "step": 50362 + }, + { + "epoch": 3.24, + "learning_rate": 9.214238176002062e-06, + "loss": 0.8258, + "step": 50363 + }, + { + "epoch": 3.24, + "learning_rate": 9.214023486979111e-06, + "loss": 0.8371, + "step": 50364 + }, + { + "epoch": 3.24, + "learning_rate": 9.213808797956162e-06, + "loss": 0.8157, + "step": 50365 + }, + { + "epoch": 3.24, + "learning_rate": 9.213594108933212e-06, + "loss": 0.8066, + "step": 50366 + }, + { + "epoch": 3.24, + "learning_rate": 9.213379419910261e-06, + "loss": 0.9037, + "step": 50367 + }, + { + "epoch": 3.24, + "learning_rate": 9.21316473088731e-06, + "loss": 0.7712, + "step": 50368 + }, + { + "epoch": 3.24, + "learning_rate": 9.21295004186436e-06, + "loss": 0.8323, + "step": 50369 + }, + { + "epoch": 3.24, + "learning_rate": 9.21273535284141e-06, + "loss": 0.868, + "step": 50370 + }, + { + "epoch": 3.24, + "learning_rate": 9.212520663818459e-06, + "loss": 0.7336, + "step": 50371 + }, + { + "epoch": 3.24, + "learning_rate": 9.21230597479551e-06, + "loss": 0.7408, + "step": 50372 + }, + { + "epoch": 3.24, + "learning_rate": 9.21209128577256e-06, + "loss": 0.8273, + "step": 50373 + }, + { + "epoch": 3.24, + "learning_rate": 9.211876596749609e-06, + "loss": 0.8772, + "step": 50374 + }, + { + "epoch": 3.24, + "learning_rate": 9.21166190772666e-06, + "loss": 0.8576, + "step": 50375 + }, + { + "epoch": 3.24, + "learning_rate": 9.211447218703709e-06, + "loss": 0.8199, + "step": 50376 + }, + { + "epoch": 3.24, + "learning_rate": 9.211232529680758e-06, + "loss": 0.8709, + "step": 50377 + }, + { + "epoch": 3.24, + "learning_rate": 9.211017840657808e-06, + "loss": 0.8365, + "step": 50378 + }, + { + "epoch": 3.24, + "learning_rate": 9.210803151634857e-06, + "loss": 0.844, + "step": 50379 + }, + { + "epoch": 3.24, + "learning_rate": 9.210588462611906e-06, + "loss": 0.7906, + "step": 50380 + }, + { + "epoch": 3.24, + "learning_rate": 9.210373773588957e-06, + "loss": 0.7883, + "step": 50381 + }, + { + "epoch": 3.24, + "learning_rate": 9.210159084566007e-06, + "loss": 0.9361, + "step": 50382 + }, + { + "epoch": 3.24, + "learning_rate": 9.209944395543056e-06, + "loss": 0.8237, + "step": 50383 + }, + { + "epoch": 3.24, + "learning_rate": 9.209729706520107e-06, + "loss": 0.917, + "step": 50384 + }, + { + "epoch": 3.24, + "learning_rate": 9.209515017497156e-06, + "loss": 0.8068, + "step": 50385 + }, + { + "epoch": 3.24, + "learning_rate": 9.209300328474205e-06, + "loss": 0.8389, + "step": 50386 + }, + { + "epoch": 3.24, + "learning_rate": 9.209085639451256e-06, + "loss": 0.7822, + "step": 50387 + }, + { + "epoch": 3.24, + "learning_rate": 9.208870950428306e-06, + "loss": 0.8592, + "step": 50388 + }, + { + "epoch": 3.24, + "learning_rate": 9.208656261405355e-06, + "loss": 0.8465, + "step": 50389 + }, + { + "epoch": 3.24, + "learning_rate": 9.208441572382404e-06, + "loss": 0.824, + "step": 50390 + }, + { + "epoch": 3.24, + "learning_rate": 9.208226883359455e-06, + "loss": 0.8993, + "step": 50391 + }, + { + "epoch": 3.24, + "learning_rate": 9.208012194336504e-06, + "loss": 0.7625, + "step": 50392 + }, + { + "epoch": 3.24, + "learning_rate": 9.207797505313555e-06, + "loss": 0.8499, + "step": 50393 + }, + { + "epoch": 3.24, + "learning_rate": 9.207582816290604e-06, + "loss": 0.8298, + "step": 50394 + }, + { + "epoch": 3.24, + "learning_rate": 9.207368127267654e-06, + "loss": 0.8894, + "step": 50395 + }, + { + "epoch": 3.24, + "learning_rate": 9.207153438244703e-06, + "loss": 0.8768, + "step": 50396 + }, + { + "epoch": 3.24, + "learning_rate": 9.206938749221754e-06, + "loss": 0.8236, + "step": 50397 + }, + { + "epoch": 3.24, + "learning_rate": 9.206724060198803e-06, + "loss": 0.7491, + "step": 50398 + }, + { + "epoch": 3.24, + "learning_rate": 9.206509371175852e-06, + "loss": 0.7662, + "step": 50399 + }, + { + "epoch": 3.24, + "learning_rate": 9.206294682152902e-06, + "loss": 0.8363, + "step": 50400 + }, + { + "epoch": 3.24, + "learning_rate": 9.206079993129951e-06, + "loss": 0.7729, + "step": 50401 + }, + { + "epoch": 3.24, + "learning_rate": 9.205865304107002e-06, + "loss": 0.8402, + "step": 50402 + }, + { + "epoch": 3.24, + "learning_rate": 9.205650615084053e-06, + "loss": 0.7744, + "step": 50403 + }, + { + "epoch": 3.24, + "learning_rate": 9.205435926061102e-06, + "loss": 0.8249, + "step": 50404 + }, + { + "epoch": 3.24, + "learning_rate": 9.20522123703815e-06, + "loss": 0.786, + "step": 50405 + }, + { + "epoch": 3.24, + "learning_rate": 9.205006548015201e-06, + "loss": 0.7443, + "step": 50406 + }, + { + "epoch": 3.24, + "learning_rate": 9.20479185899225e-06, + "loss": 0.8468, + "step": 50407 + }, + { + "epoch": 3.24, + "learning_rate": 9.2045771699693e-06, + "loss": 0.794, + "step": 50408 + }, + { + "epoch": 3.24, + "learning_rate": 9.20436248094635e-06, + "loss": 0.8678, + "step": 50409 + }, + { + "epoch": 3.24, + "learning_rate": 9.2041477919234e-06, + "loss": 0.8515, + "step": 50410 + }, + { + "epoch": 3.24, + "learning_rate": 9.20393310290045e-06, + "loss": 0.8459, + "step": 50411 + }, + { + "epoch": 3.24, + "learning_rate": 9.2037184138775e-06, + "loss": 0.8268, + "step": 50412 + }, + { + "epoch": 3.24, + "learning_rate": 9.203503724854549e-06, + "loss": 0.7747, + "step": 50413 + }, + { + "epoch": 3.24, + "learning_rate": 9.203289035831598e-06, + "loss": 0.87, + "step": 50414 + }, + { + "epoch": 3.24, + "learning_rate": 9.203074346808649e-06, + "loss": 0.8965, + "step": 50415 + }, + { + "epoch": 3.24, + "learning_rate": 9.202859657785698e-06, + "loss": 0.8311, + "step": 50416 + }, + { + "epoch": 3.24, + "learning_rate": 9.202644968762748e-06, + "loss": 0.8382, + "step": 50417 + }, + { + "epoch": 3.24, + "learning_rate": 9.202430279739797e-06, + "loss": 0.8226, + "step": 50418 + }, + { + "epoch": 3.24, + "learning_rate": 9.202215590716848e-06, + "loss": 0.8206, + "step": 50419 + }, + { + "epoch": 3.24, + "learning_rate": 9.202000901693897e-06, + "loss": 0.8535, + "step": 50420 + }, + { + "epoch": 3.24, + "learning_rate": 9.201786212670946e-06, + "loss": 0.8638, + "step": 50421 + }, + { + "epoch": 3.24, + "learning_rate": 9.201571523647997e-06, + "loss": 0.8253, + "step": 50422 + }, + { + "epoch": 3.24, + "learning_rate": 9.201356834625046e-06, + "loss": 0.7936, + "step": 50423 + }, + { + "epoch": 3.24, + "learning_rate": 9.201142145602096e-06, + "loss": 0.8447, + "step": 50424 + }, + { + "epoch": 3.24, + "learning_rate": 9.200927456579147e-06, + "loss": 0.7883, + "step": 50425 + }, + { + "epoch": 3.24, + "learning_rate": 9.200712767556196e-06, + "loss": 0.8427, + "step": 50426 + }, + { + "epoch": 3.24, + "learning_rate": 9.200498078533245e-06, + "loss": 0.7687, + "step": 50427 + }, + { + "epoch": 3.24, + "learning_rate": 9.200283389510295e-06, + "loss": 0.8597, + "step": 50428 + }, + { + "epoch": 3.24, + "learning_rate": 9.200068700487344e-06, + "loss": 0.7868, + "step": 50429 + }, + { + "epoch": 3.24, + "learning_rate": 9.199854011464393e-06, + "loss": 0.8121, + "step": 50430 + }, + { + "epoch": 3.24, + "learning_rate": 9.199639322441444e-06, + "loss": 0.8111, + "step": 50431 + }, + { + "epoch": 3.24, + "learning_rate": 9.199424633418495e-06, + "loss": 0.7911, + "step": 50432 + }, + { + "epoch": 3.24, + "learning_rate": 9.199209944395544e-06, + "loss": 0.7613, + "step": 50433 + }, + { + "epoch": 3.24, + "learning_rate": 9.198995255372594e-06, + "loss": 0.8674, + "step": 50434 + }, + { + "epoch": 3.24, + "learning_rate": 9.198780566349643e-06, + "loss": 0.815, + "step": 50435 + }, + { + "epoch": 3.24, + "learning_rate": 9.198565877326692e-06, + "loss": 0.8618, + "step": 50436 + }, + { + "epoch": 3.24, + "learning_rate": 9.198351188303743e-06, + "loss": 0.8217, + "step": 50437 + }, + { + "epoch": 3.25, + "learning_rate": 9.198136499280792e-06, + "loss": 0.8471, + "step": 50438 + }, + { + "epoch": 3.25, + "learning_rate": 9.197921810257843e-06, + "loss": 0.8696, + "step": 50439 + }, + { + "epoch": 3.25, + "learning_rate": 9.197707121234891e-06, + "loss": 0.826, + "step": 50440 + }, + { + "epoch": 3.25, + "learning_rate": 9.197492432211942e-06, + "loss": 0.7592, + "step": 50441 + }, + { + "epoch": 3.25, + "learning_rate": 9.197277743188991e-06, + "loss": 0.8334, + "step": 50442 + }, + { + "epoch": 3.25, + "learning_rate": 9.197063054166042e-06, + "loss": 0.8504, + "step": 50443 + }, + { + "epoch": 3.25, + "learning_rate": 9.19684836514309e-06, + "loss": 0.7951, + "step": 50444 + }, + { + "epoch": 3.25, + "learning_rate": 9.19663367612014e-06, + "loss": 0.8462, + "step": 50445 + }, + { + "epoch": 3.25, + "learning_rate": 9.19641898709719e-06, + "loss": 0.8382, + "step": 50446 + }, + { + "epoch": 3.25, + "learning_rate": 9.196204298074241e-06, + "loss": 0.8402, + "step": 50447 + }, + { + "epoch": 3.25, + "learning_rate": 9.19598960905129e-06, + "loss": 0.8437, + "step": 50448 + }, + { + "epoch": 3.25, + "learning_rate": 9.195774920028339e-06, + "loss": 0.8315, + "step": 50449 + }, + { + "epoch": 3.25, + "learning_rate": 9.19556023100539e-06, + "loss": 0.7883, + "step": 50450 + }, + { + "epoch": 3.25, + "learning_rate": 9.195345541982439e-06, + "loss": 0.7936, + "step": 50451 + }, + { + "epoch": 3.25, + "learning_rate": 9.19513085295949e-06, + "loss": 0.8018, + "step": 50452 + }, + { + "epoch": 3.25, + "learning_rate": 9.19491616393654e-06, + "loss": 0.7949, + "step": 50453 + }, + { + "epoch": 3.25, + "learning_rate": 9.194701474913589e-06, + "loss": 0.8511, + "step": 50454 + }, + { + "epoch": 3.25, + "learning_rate": 9.194486785890638e-06, + "loss": 0.8346, + "step": 50455 + }, + { + "epoch": 3.25, + "learning_rate": 9.194272096867688e-06, + "loss": 0.87, + "step": 50456 + }, + { + "epoch": 3.25, + "learning_rate": 9.194057407844737e-06, + "loss": 0.8519, + "step": 50457 + }, + { + "epoch": 3.25, + "learning_rate": 9.193842718821786e-06, + "loss": 0.817, + "step": 50458 + }, + { + "epoch": 3.25, + "learning_rate": 9.193628029798837e-06, + "loss": 0.8171, + "step": 50459 + }, + { + "epoch": 3.25, + "learning_rate": 9.193413340775888e-06, + "loss": 0.7975, + "step": 50460 + }, + { + "epoch": 3.25, + "learning_rate": 9.193198651752937e-06, + "loss": 0.8187, + "step": 50461 + }, + { + "epoch": 3.25, + "learning_rate": 9.192983962729987e-06, + "loss": 0.853, + "step": 50462 + }, + { + "epoch": 3.25, + "learning_rate": 9.192769273707036e-06, + "loss": 0.8914, + "step": 50463 + }, + { + "epoch": 3.25, + "learning_rate": 9.192554584684085e-06, + "loss": 0.7998, + "step": 50464 + }, + { + "epoch": 3.25, + "learning_rate": 9.192339895661136e-06, + "loss": 0.8054, + "step": 50465 + }, + { + "epoch": 3.25, + "learning_rate": 9.192125206638185e-06, + "loss": 0.8287, + "step": 50466 + }, + { + "epoch": 3.25, + "learning_rate": 9.191910517615234e-06, + "loss": 0.7854, + "step": 50467 + }, + { + "epoch": 3.25, + "learning_rate": 9.191695828592285e-06, + "loss": 0.8329, + "step": 50468 + }, + { + "epoch": 3.25, + "learning_rate": 9.191481139569335e-06, + "loss": 0.8267, + "step": 50469 + }, + { + "epoch": 3.25, + "learning_rate": 9.191266450546384e-06, + "loss": 0.8563, + "step": 50470 + }, + { + "epoch": 3.25, + "learning_rate": 9.191051761523435e-06, + "loss": 0.7936, + "step": 50471 + }, + { + "epoch": 3.25, + "learning_rate": 9.190837072500484e-06, + "loss": 0.7564, + "step": 50472 + }, + { + "epoch": 3.25, + "learning_rate": 9.190622383477533e-06, + "loss": 0.8333, + "step": 50473 + }, + { + "epoch": 3.25, + "learning_rate": 9.190407694454583e-06, + "loss": 0.7735, + "step": 50474 + }, + { + "epoch": 3.25, + "learning_rate": 9.190193005431634e-06, + "loss": 0.7773, + "step": 50475 + }, + { + "epoch": 3.25, + "learning_rate": 9.189978316408683e-06, + "loss": 0.8511, + "step": 50476 + }, + { + "epoch": 3.25, + "learning_rate": 9.189763627385732e-06, + "loss": 0.7744, + "step": 50477 + }, + { + "epoch": 3.25, + "learning_rate": 9.189548938362783e-06, + "loss": 0.8016, + "step": 50478 + }, + { + "epoch": 3.25, + "learning_rate": 9.189334249339832e-06, + "loss": 0.8643, + "step": 50479 + }, + { + "epoch": 3.25, + "learning_rate": 9.189119560316882e-06, + "loss": 0.8354, + "step": 50480 + }, + { + "epoch": 3.25, + "learning_rate": 9.188904871293931e-06, + "loss": 0.8134, + "step": 50481 + }, + { + "epoch": 3.25, + "learning_rate": 9.188690182270982e-06, + "loss": 0.7865, + "step": 50482 + }, + { + "epoch": 3.25, + "learning_rate": 9.188475493248031e-06, + "loss": 0.8389, + "step": 50483 + }, + { + "epoch": 3.25, + "learning_rate": 9.188260804225082e-06, + "loss": 0.8207, + "step": 50484 + }, + { + "epoch": 3.25, + "learning_rate": 9.18804611520213e-06, + "loss": 0.8295, + "step": 50485 + }, + { + "epoch": 3.25, + "learning_rate": 9.18783142617918e-06, + "loss": 0.8131, + "step": 50486 + }, + { + "epoch": 3.25, + "learning_rate": 9.18761673715623e-06, + "loss": 0.8415, + "step": 50487 + }, + { + "epoch": 3.25, + "learning_rate": 9.187402048133279e-06, + "loss": 0.8434, + "step": 50488 + }, + { + "epoch": 3.25, + "learning_rate": 9.18718735911033e-06, + "loss": 0.8051, + "step": 50489 + }, + { + "epoch": 3.25, + "learning_rate": 9.18697267008738e-06, + "loss": 0.7816, + "step": 50490 + }, + { + "epoch": 3.25, + "learning_rate": 9.18675798106443e-06, + "loss": 0.8071, + "step": 50491 + }, + { + "epoch": 3.25, + "learning_rate": 9.186543292041478e-06, + "loss": 0.8359, + "step": 50492 + }, + { + "epoch": 3.25, + "learning_rate": 9.186328603018529e-06, + "loss": 0.7775, + "step": 50493 + }, + { + "epoch": 3.25, + "learning_rate": 9.186113913995578e-06, + "loss": 0.7602, + "step": 50494 + }, + { + "epoch": 3.25, + "learning_rate": 9.185899224972627e-06, + "loss": 0.8346, + "step": 50495 + }, + { + "epoch": 3.25, + "learning_rate": 9.185684535949678e-06, + "loss": 0.7809, + "step": 50496 + }, + { + "epoch": 3.25, + "learning_rate": 9.185469846926728e-06, + "loss": 0.859, + "step": 50497 + }, + { + "epoch": 3.25, + "learning_rate": 9.185255157903777e-06, + "loss": 0.858, + "step": 50498 + }, + { + "epoch": 3.25, + "learning_rate": 9.185040468880828e-06, + "loss": 0.7924, + "step": 50499 + }, + { + "epoch": 3.25, + "learning_rate": 9.184825779857877e-06, + "loss": 0.8124, + "step": 50500 + }, + { + "epoch": 3.25, + "learning_rate": 9.184611090834926e-06, + "loss": 0.8308, + "step": 50501 + }, + { + "epoch": 3.25, + "learning_rate": 9.184396401811976e-06, + "loss": 0.7359, + "step": 50502 + }, + { + "epoch": 3.25, + "learning_rate": 9.184181712789025e-06, + "loss": 0.7701, + "step": 50503 + }, + { + "epoch": 3.25, + "learning_rate": 9.183967023766076e-06, + "loss": 0.8501, + "step": 50504 + }, + { + "epoch": 3.25, + "learning_rate": 9.183752334743125e-06, + "loss": 0.8186, + "step": 50505 + }, + { + "epoch": 3.25, + "learning_rate": 9.183537645720176e-06, + "loss": 0.8401, + "step": 50506 + }, + { + "epoch": 3.25, + "learning_rate": 9.183322956697225e-06, + "loss": 0.817, + "step": 50507 + }, + { + "epoch": 3.25, + "learning_rate": 9.183108267674274e-06, + "loss": 0.8132, + "step": 50508 + }, + { + "epoch": 3.25, + "learning_rate": 9.182893578651324e-06, + "loss": 0.8655, + "step": 50509 + }, + { + "epoch": 3.25, + "learning_rate": 9.182678889628373e-06, + "loss": 0.8546, + "step": 50510 + }, + { + "epoch": 3.25, + "learning_rate": 9.182464200605424e-06, + "loss": 0.7574, + "step": 50511 + }, + { + "epoch": 3.25, + "learning_rate": 9.182249511582475e-06, + "loss": 0.825, + "step": 50512 + }, + { + "epoch": 3.25, + "learning_rate": 9.182034822559524e-06, + "loss": 0.8508, + "step": 50513 + }, + { + "epoch": 3.25, + "learning_rate": 9.181820133536572e-06, + "loss": 0.7926, + "step": 50514 + }, + { + "epoch": 3.25, + "learning_rate": 9.181605444513623e-06, + "loss": 0.8441, + "step": 50515 + }, + { + "epoch": 3.25, + "learning_rate": 9.181390755490672e-06, + "loss": 0.8715, + "step": 50516 + }, + { + "epoch": 3.25, + "learning_rate": 9.181176066467721e-06, + "loss": 0.7992, + "step": 50517 + }, + { + "epoch": 3.25, + "learning_rate": 9.180961377444772e-06, + "loss": 0.8333, + "step": 50518 + }, + { + "epoch": 3.25, + "learning_rate": 9.180746688421822e-06, + "loss": 0.8286, + "step": 50519 + }, + { + "epoch": 3.25, + "learning_rate": 9.180531999398871e-06, + "loss": 0.8214, + "step": 50520 + }, + { + "epoch": 3.25, + "learning_rate": 9.180317310375922e-06, + "loss": 0.754, + "step": 50521 + }, + { + "epoch": 3.25, + "learning_rate": 9.180102621352971e-06, + "loss": 0.7912, + "step": 50522 + }, + { + "epoch": 3.25, + "learning_rate": 9.17988793233002e-06, + "loss": 0.8346, + "step": 50523 + }, + { + "epoch": 3.25, + "learning_rate": 9.17967324330707e-06, + "loss": 0.8511, + "step": 50524 + }, + { + "epoch": 3.25, + "learning_rate": 9.17945855428412e-06, + "loss": 0.8041, + "step": 50525 + }, + { + "epoch": 3.25, + "learning_rate": 9.17924386526117e-06, + "loss": 0.8246, + "step": 50526 + }, + { + "epoch": 3.25, + "learning_rate": 9.17902917623822e-06, + "loss": 0.7934, + "step": 50527 + }, + { + "epoch": 3.25, + "learning_rate": 9.17881448721527e-06, + "loss": 0.879, + "step": 50528 + }, + { + "epoch": 3.25, + "learning_rate": 9.178599798192319e-06, + "loss": 0.7492, + "step": 50529 + }, + { + "epoch": 3.25, + "learning_rate": 9.17838510916937e-06, + "loss": 0.8981, + "step": 50530 + }, + { + "epoch": 3.25, + "learning_rate": 9.178170420146418e-06, + "loss": 0.8324, + "step": 50531 + }, + { + "epoch": 3.25, + "learning_rate": 9.177955731123467e-06, + "loss": 0.77, + "step": 50532 + }, + { + "epoch": 3.25, + "learning_rate": 9.177741042100518e-06, + "loss": 0.8523, + "step": 50533 + }, + { + "epoch": 3.25, + "learning_rate": 9.177526353077569e-06, + "loss": 0.8184, + "step": 50534 + }, + { + "epoch": 3.25, + "learning_rate": 9.177311664054618e-06, + "loss": 0.8525, + "step": 50535 + }, + { + "epoch": 3.25, + "learning_rate": 9.177096975031667e-06, + "loss": 0.7927, + "step": 50536 + }, + { + "epoch": 3.25, + "learning_rate": 9.176882286008717e-06, + "loss": 0.8162, + "step": 50537 + }, + { + "epoch": 3.25, + "learning_rate": 9.176667596985766e-06, + "loss": 0.8428, + "step": 50538 + }, + { + "epoch": 3.25, + "learning_rate": 9.176452907962817e-06, + "loss": 0.8085, + "step": 50539 + }, + { + "epoch": 3.25, + "learning_rate": 9.176238218939868e-06, + "loss": 0.8314, + "step": 50540 + }, + { + "epoch": 3.25, + "learning_rate": 9.176023529916917e-06, + "loss": 0.8332, + "step": 50541 + }, + { + "epoch": 3.25, + "learning_rate": 9.175808840893966e-06, + "loss": 0.8398, + "step": 50542 + }, + { + "epoch": 3.25, + "learning_rate": 9.175594151871016e-06, + "loss": 0.804, + "step": 50543 + }, + { + "epoch": 3.25, + "learning_rate": 9.175379462848065e-06, + "loss": 0.8431, + "step": 50544 + }, + { + "epoch": 3.25, + "learning_rate": 9.175164773825114e-06, + "loss": 0.8638, + "step": 50545 + }, + { + "epoch": 3.25, + "learning_rate": 9.174950084802165e-06, + "loss": 0.7603, + "step": 50546 + }, + { + "epoch": 3.25, + "learning_rate": 9.174735395779214e-06, + "loss": 0.8632, + "step": 50547 + }, + { + "epoch": 3.25, + "learning_rate": 9.174520706756264e-06, + "loss": 0.8249, + "step": 50548 + }, + { + "epoch": 3.25, + "learning_rate": 9.174306017733315e-06, + "loss": 0.8234, + "step": 50549 + }, + { + "epoch": 3.25, + "learning_rate": 9.174091328710364e-06, + "loss": 0.8363, + "step": 50550 + }, + { + "epoch": 3.25, + "learning_rate": 9.173876639687413e-06, + "loss": 0.8058, + "step": 50551 + }, + { + "epoch": 3.25, + "learning_rate": 9.173661950664464e-06, + "loss": 0.7954, + "step": 50552 + }, + { + "epoch": 3.25, + "learning_rate": 9.173447261641513e-06, + "loss": 0.8706, + "step": 50553 + }, + { + "epoch": 3.25, + "learning_rate": 9.173232572618562e-06, + "loss": 0.8279, + "step": 50554 + }, + { + "epoch": 3.25, + "learning_rate": 9.173017883595612e-06, + "loss": 0.8006, + "step": 50555 + }, + { + "epoch": 3.25, + "learning_rate": 9.172803194572663e-06, + "loss": 0.8187, + "step": 50556 + }, + { + "epoch": 3.25, + "learning_rate": 9.172588505549712e-06, + "loss": 0.7944, + "step": 50557 + }, + { + "epoch": 3.25, + "learning_rate": 9.172373816526763e-06, + "loss": 0.8036, + "step": 50558 + }, + { + "epoch": 3.25, + "learning_rate": 9.172159127503811e-06, + "loss": 0.8457, + "step": 50559 + }, + { + "epoch": 3.25, + "learning_rate": 9.17194443848086e-06, + "loss": 0.8671, + "step": 50560 + }, + { + "epoch": 3.25, + "learning_rate": 9.171729749457911e-06, + "loss": 0.8581, + "step": 50561 + }, + { + "epoch": 3.25, + "learning_rate": 9.171515060434962e-06, + "loss": 0.8059, + "step": 50562 + }, + { + "epoch": 3.25, + "learning_rate": 9.17130037141201e-06, + "loss": 0.8445, + "step": 50563 + }, + { + "epoch": 3.25, + "learning_rate": 9.17108568238906e-06, + "loss": 0.863, + "step": 50564 + }, + { + "epoch": 3.25, + "learning_rate": 9.17087099336611e-06, + "loss": 0.7925, + "step": 50565 + }, + { + "epoch": 3.25, + "learning_rate": 9.17065630434316e-06, + "loss": 0.7841, + "step": 50566 + }, + { + "epoch": 3.25, + "learning_rate": 9.17044161532021e-06, + "loss": 0.8616, + "step": 50567 + }, + { + "epoch": 3.25, + "learning_rate": 9.170226926297259e-06, + "loss": 0.8519, + "step": 50568 + }, + { + "epoch": 3.25, + "learning_rate": 9.17001223727431e-06, + "loss": 0.8527, + "step": 50569 + }, + { + "epoch": 3.25, + "learning_rate": 9.169797548251359e-06, + "loss": 0.8458, + "step": 50570 + }, + { + "epoch": 3.25, + "learning_rate": 9.16958285922841e-06, + "loss": 0.8622, + "step": 50571 + }, + { + "epoch": 3.25, + "learning_rate": 9.169368170205458e-06, + "loss": 0.7648, + "step": 50572 + }, + { + "epoch": 3.25, + "learning_rate": 9.169153481182507e-06, + "loss": 0.7892, + "step": 50573 + }, + { + "epoch": 3.25, + "learning_rate": 9.168938792159558e-06, + "loss": 0.7792, + "step": 50574 + }, + { + "epoch": 3.25, + "learning_rate": 9.168724103136607e-06, + "loss": 0.8009, + "step": 50575 + }, + { + "epoch": 3.25, + "learning_rate": 9.168509414113656e-06, + "loss": 0.7768, + "step": 50576 + }, + { + "epoch": 3.25, + "learning_rate": 9.168294725090708e-06, + "loss": 0.7898, + "step": 50577 + }, + { + "epoch": 3.25, + "learning_rate": 9.168080036067757e-06, + "loss": 0.8207, + "step": 50578 + }, + { + "epoch": 3.25, + "learning_rate": 9.167865347044806e-06, + "loss": 0.8257, + "step": 50579 + }, + { + "epoch": 3.25, + "learning_rate": 9.167650658021857e-06, + "loss": 0.8579, + "step": 50580 + }, + { + "epoch": 3.25, + "learning_rate": 9.167435968998906e-06, + "loss": 0.8599, + "step": 50581 + }, + { + "epoch": 3.25, + "learning_rate": 9.167221279975955e-06, + "loss": 0.8102, + "step": 50582 + }, + { + "epoch": 3.25, + "learning_rate": 9.167006590953005e-06, + "loss": 0.8308, + "step": 50583 + }, + { + "epoch": 3.25, + "learning_rate": 9.166791901930056e-06, + "loss": 0.8494, + "step": 50584 + }, + { + "epoch": 3.25, + "learning_rate": 9.166577212907105e-06, + "loss": 0.8158, + "step": 50585 + }, + { + "epoch": 3.25, + "learning_rate": 9.166362523884154e-06, + "loss": 0.8068, + "step": 50586 + }, + { + "epoch": 3.25, + "learning_rate": 9.166147834861205e-06, + "loss": 0.8822, + "step": 50587 + }, + { + "epoch": 3.25, + "learning_rate": 9.165933145838253e-06, + "loss": 0.8322, + "step": 50588 + }, + { + "epoch": 3.25, + "learning_rate": 9.165718456815304e-06, + "loss": 0.8706, + "step": 50589 + }, + { + "epoch": 3.25, + "learning_rate": 9.165503767792353e-06, + "loss": 0.874, + "step": 50590 + }, + { + "epoch": 3.25, + "learning_rate": 9.165289078769404e-06, + "loss": 0.818, + "step": 50591 + }, + { + "epoch": 3.25, + "learning_rate": 9.165074389746453e-06, + "loss": 0.8343, + "step": 50592 + }, + { + "epoch": 3.26, + "learning_rate": 9.164859700723503e-06, + "loss": 0.8285, + "step": 50593 + }, + { + "epoch": 3.26, + "learning_rate": 9.164645011700552e-06, + "loss": 0.7837, + "step": 50594 + }, + { + "epoch": 3.26, + "learning_rate": 9.164430322677601e-06, + "loss": 0.8851, + "step": 50595 + }, + { + "epoch": 3.26, + "learning_rate": 9.164215633654652e-06, + "loss": 0.8259, + "step": 50596 + }, + { + "epoch": 3.26, + "learning_rate": 9.164000944631701e-06, + "loss": 0.8237, + "step": 50597 + }, + { + "epoch": 3.26, + "learning_rate": 9.163786255608752e-06, + "loss": 0.8365, + "step": 50598 + }, + { + "epoch": 3.26, + "learning_rate": 9.163571566585802e-06, + "loss": 0.7944, + "step": 50599 + }, + { + "epoch": 3.26, + "learning_rate": 9.163356877562851e-06, + "loss": 0.8125, + "step": 50600 + }, + { + "epoch": 3.26, + "learning_rate": 9.1631421885399e-06, + "loss": 0.8429, + "step": 50601 + }, + { + "epoch": 3.26, + "learning_rate": 9.162927499516951e-06, + "loss": 0.8135, + "step": 50602 + }, + { + "epoch": 3.26, + "learning_rate": 9.162712810494e-06, + "loss": 0.8579, + "step": 50603 + }, + { + "epoch": 3.26, + "learning_rate": 9.162498121471049e-06, + "loss": 0.834, + "step": 50604 + }, + { + "epoch": 3.26, + "learning_rate": 9.1622834324481e-06, + "loss": 0.8413, + "step": 50605 + }, + { + "epoch": 3.26, + "learning_rate": 9.16206874342515e-06, + "loss": 0.8198, + "step": 50606 + }, + { + "epoch": 3.26, + "learning_rate": 9.161854054402199e-06, + "loss": 0.7769, + "step": 50607 + }, + { + "epoch": 3.26, + "learning_rate": 9.16163936537925e-06, + "loss": 0.8206, + "step": 50608 + }, + { + "epoch": 3.26, + "learning_rate": 9.161424676356299e-06, + "loss": 0.8221, + "step": 50609 + }, + { + "epoch": 3.26, + "learning_rate": 9.161209987333348e-06, + "loss": 0.7819, + "step": 50610 + }, + { + "epoch": 3.26, + "learning_rate": 9.160995298310398e-06, + "loss": 0.7756, + "step": 50611 + }, + { + "epoch": 3.26, + "learning_rate": 9.160995298310398e-06, + "loss": 0.8567, + "step": 50612 + }, + { + "epoch": 3.26, + "learning_rate": 9.160780609287447e-06, + "loss": 0.8406, + "step": 50613 + }, + { + "epoch": 3.26, + "learning_rate": 9.160565920264498e-06, + "loss": 0.8226, + "step": 50614 + }, + { + "epoch": 3.26, + "learning_rate": 9.160351231241547e-06, + "loss": 0.8353, + "step": 50615 + }, + { + "epoch": 3.26, + "learning_rate": 9.160136542218598e-06, + "loss": 0.8541, + "step": 50616 + }, + { + "epoch": 3.26, + "learning_rate": 9.159921853195647e-06, + "loss": 0.8297, + "step": 50617 + }, + { + "epoch": 3.26, + "learning_rate": 9.159707164172697e-06, + "loss": 0.811, + "step": 50618 + }, + { + "epoch": 3.26, + "learning_rate": 9.159492475149746e-06, + "loss": 0.8561, + "step": 50619 + }, + { + "epoch": 3.26, + "learning_rate": 9.159277786126795e-06, + "loss": 0.8311, + "step": 50620 + }, + { + "epoch": 3.26, + "learning_rate": 9.159063097103846e-06, + "loss": 0.8131, + "step": 50621 + }, + { + "epoch": 3.26, + "learning_rate": 9.158848408080896e-06, + "loss": 0.8112, + "step": 50622 + }, + { + "epoch": 3.26, + "learning_rate": 9.158633719057945e-06, + "loss": 0.8005, + "step": 50623 + }, + { + "epoch": 3.26, + "learning_rate": 9.158419030034994e-06, + "loss": 0.7609, + "step": 50624 + }, + { + "epoch": 3.26, + "learning_rate": 9.158204341012045e-06, + "loss": 0.8091, + "step": 50625 + }, + { + "epoch": 3.26, + "learning_rate": 9.157989651989094e-06, + "loss": 0.7879, + "step": 50626 + }, + { + "epoch": 3.26, + "learning_rate": 9.157774962966145e-06, + "loss": 0.7844, + "step": 50627 + }, + { + "epoch": 3.26, + "learning_rate": 9.157560273943195e-06, + "loss": 0.8026, + "step": 50628 + }, + { + "epoch": 3.26, + "learning_rate": 9.157345584920244e-06, + "loss": 0.8395, + "step": 50629 + }, + { + "epoch": 3.26, + "learning_rate": 9.157130895897293e-06, + "loss": 0.8727, + "step": 50630 + }, + { + "epoch": 3.26, + "learning_rate": 9.156916206874344e-06, + "loss": 0.8879, + "step": 50631 + }, + { + "epoch": 3.26, + "learning_rate": 9.156701517851393e-06, + "loss": 0.8289, + "step": 50632 + }, + { + "epoch": 3.26, + "learning_rate": 9.156486828828442e-06, + "loss": 0.7262, + "step": 50633 + }, + { + "epoch": 3.26, + "learning_rate": 9.156272139805492e-06, + "loss": 0.7737, + "step": 50634 + }, + { + "epoch": 3.26, + "learning_rate": 9.156057450782541e-06, + "loss": 0.8291, + "step": 50635 + }, + { + "epoch": 3.26, + "learning_rate": 9.155842761759592e-06, + "loss": 0.8418, + "step": 50636 + }, + { + "epoch": 3.26, + "learning_rate": 9.155628072736643e-06, + "loss": 0.8247, + "step": 50637 + }, + { + "epoch": 3.26, + "learning_rate": 9.155413383713692e-06, + "loss": 0.8196, + "step": 50638 + }, + { + "epoch": 3.26, + "learning_rate": 9.15519869469074e-06, + "loss": 0.7605, + "step": 50639 + }, + { + "epoch": 3.26, + "learning_rate": 9.154984005667791e-06, + "loss": 0.7371, + "step": 50640 + }, + { + "epoch": 3.26, + "learning_rate": 9.15476931664484e-06, + "loss": 0.8703, + "step": 50641 + }, + { + "epoch": 3.26, + "learning_rate": 9.15455462762189e-06, + "loss": 0.8706, + "step": 50642 + }, + { + "epoch": 3.26, + "learning_rate": 9.15433993859894e-06, + "loss": 0.872, + "step": 50643 + }, + { + "epoch": 3.26, + "learning_rate": 9.15412524957599e-06, + "loss": 0.8173, + "step": 50644 + }, + { + "epoch": 3.26, + "learning_rate": 9.15391056055304e-06, + "loss": 0.8387, + "step": 50645 + }, + { + "epoch": 3.26, + "learning_rate": 9.15369587153009e-06, + "loss": 0.8288, + "step": 50646 + }, + { + "epoch": 3.26, + "learning_rate": 9.15348118250714e-06, + "loss": 0.8325, + "step": 50647 + }, + { + "epoch": 3.26, + "learning_rate": 9.153266493484188e-06, + "loss": 0.7861, + "step": 50648 + }, + { + "epoch": 3.26, + "learning_rate": 9.153051804461239e-06, + "loss": 0.8179, + "step": 50649 + }, + { + "epoch": 3.26, + "learning_rate": 9.15283711543829e-06, + "loss": 0.8172, + "step": 50650 + }, + { + "epoch": 3.26, + "learning_rate": 9.152622426415338e-06, + "loss": 0.8182, + "step": 50651 + }, + { + "epoch": 3.26, + "learning_rate": 9.152407737392387e-06, + "loss": 0.8451, + "step": 50652 + }, + { + "epoch": 3.26, + "learning_rate": 9.152193048369438e-06, + "loss": 0.8085, + "step": 50653 + }, + { + "epoch": 3.26, + "learning_rate": 9.151978359346487e-06, + "loss": 0.8055, + "step": 50654 + }, + { + "epoch": 3.26, + "learning_rate": 9.151763670323538e-06, + "loss": 0.785, + "step": 50655 + }, + { + "epoch": 3.26, + "learning_rate": 9.151548981300587e-06, + "loss": 0.9283, + "step": 50656 + }, + { + "epoch": 3.26, + "learning_rate": 9.151334292277637e-06, + "loss": 0.8842, + "step": 50657 + }, + { + "epoch": 3.26, + "learning_rate": 9.151119603254686e-06, + "loss": 0.8285, + "step": 50658 + }, + { + "epoch": 3.26, + "learning_rate": 9.150904914231737e-06, + "loss": 0.8392, + "step": 50659 + }, + { + "epoch": 3.26, + "learning_rate": 9.150690225208786e-06, + "loss": 0.8076, + "step": 50660 + }, + { + "epoch": 3.26, + "learning_rate": 9.150475536185835e-06, + "loss": 0.8229, + "step": 50661 + }, + { + "epoch": 3.26, + "learning_rate": 9.150260847162886e-06, + "loss": 0.8082, + "step": 50662 + }, + { + "epoch": 3.26, + "learning_rate": 9.150046158139935e-06, + "loss": 0.8299, + "step": 50663 + }, + { + "epoch": 3.26, + "learning_rate": 9.149831469116983e-06, + "loss": 0.7894, + "step": 50664 + }, + { + "epoch": 3.26, + "learning_rate": 9.149616780094036e-06, + "loss": 0.8529, + "step": 50665 + }, + { + "epoch": 3.26, + "learning_rate": 9.149402091071085e-06, + "loss": 0.9106, + "step": 50666 + }, + { + "epoch": 3.26, + "learning_rate": 9.149187402048134e-06, + "loss": 0.8567, + "step": 50667 + }, + { + "epoch": 3.26, + "learning_rate": 9.148972713025184e-06, + "loss": 0.8163, + "step": 50668 + }, + { + "epoch": 3.26, + "learning_rate": 9.148758024002233e-06, + "loss": 0.8737, + "step": 50669 + }, + { + "epoch": 3.26, + "learning_rate": 9.148543334979282e-06, + "loss": 0.8325, + "step": 50670 + }, + { + "epoch": 3.26, + "learning_rate": 9.148328645956333e-06, + "loss": 0.8511, + "step": 50671 + }, + { + "epoch": 3.26, + "learning_rate": 9.148113956933384e-06, + "loss": 0.8507, + "step": 50672 + }, + { + "epoch": 3.26, + "learning_rate": 9.147899267910433e-06, + "loss": 0.7804, + "step": 50673 + }, + { + "epoch": 3.26, + "learning_rate": 9.147684578887482e-06, + "loss": 0.762, + "step": 50674 + }, + { + "epoch": 3.26, + "learning_rate": 9.147469889864532e-06, + "loss": 0.8499, + "step": 50675 + }, + { + "epoch": 3.26, + "learning_rate": 9.147255200841581e-06, + "loss": 0.7911, + "step": 50676 + }, + { + "epoch": 3.26, + "learning_rate": 9.147040511818632e-06, + "loss": 0.8401, + "step": 50677 + }, + { + "epoch": 3.26, + "learning_rate": 9.14682582279568e-06, + "loss": 0.8469, + "step": 50678 + }, + { + "epoch": 3.26, + "learning_rate": 9.146611133772731e-06, + "loss": 0.7467, + "step": 50679 + }, + { + "epoch": 3.26, + "learning_rate": 9.14639644474978e-06, + "loss": 0.8197, + "step": 50680 + }, + { + "epoch": 3.26, + "learning_rate": 9.146181755726831e-06, + "loss": 0.853, + "step": 50681 + }, + { + "epoch": 3.26, + "learning_rate": 9.14596706670388e-06, + "loss": 0.8482, + "step": 50682 + }, + { + "epoch": 3.26, + "learning_rate": 9.145752377680929e-06, + "loss": 0.8492, + "step": 50683 + }, + { + "epoch": 3.26, + "learning_rate": 9.14553768865798e-06, + "loss": 0.7868, + "step": 50684 + }, + { + "epoch": 3.26, + "learning_rate": 9.145322999635029e-06, + "loss": 0.8331, + "step": 50685 + }, + { + "epoch": 3.26, + "learning_rate": 9.14510831061208e-06, + "loss": 0.819, + "step": 50686 + }, + { + "epoch": 3.26, + "learning_rate": 9.14489362158913e-06, + "loss": 0.8561, + "step": 50687 + }, + { + "epoch": 3.26, + "learning_rate": 9.144678932566179e-06, + "loss": 0.8815, + "step": 50688 + }, + { + "epoch": 3.26, + "learning_rate": 9.144464243543228e-06, + "loss": 0.7199, + "step": 50689 + }, + { + "epoch": 3.26, + "learning_rate": 9.144249554520279e-06, + "loss": 0.802, + "step": 50690 + }, + { + "epoch": 3.26, + "learning_rate": 9.144034865497328e-06, + "loss": 0.8637, + "step": 50691 + }, + { + "epoch": 3.26, + "learning_rate": 9.143820176474377e-06, + "loss": 0.7938, + "step": 50692 + }, + { + "epoch": 3.26, + "learning_rate": 9.143605487451427e-06, + "loss": 0.8391, + "step": 50693 + }, + { + "epoch": 3.26, + "learning_rate": 9.143390798428478e-06, + "loss": 0.8221, + "step": 50694 + }, + { + "epoch": 3.26, + "learning_rate": 9.143176109405527e-06, + "loss": 0.7627, + "step": 50695 + }, + { + "epoch": 3.26, + "learning_rate": 9.142961420382577e-06, + "loss": 0.8693, + "step": 50696 + }, + { + "epoch": 3.26, + "learning_rate": 9.142746731359626e-06, + "loss": 0.8857, + "step": 50697 + }, + { + "epoch": 3.26, + "learning_rate": 9.142532042336675e-06, + "loss": 0.8114, + "step": 50698 + }, + { + "epoch": 3.26, + "learning_rate": 9.142317353313726e-06, + "loss": 0.7938, + "step": 50699 + }, + { + "epoch": 3.26, + "learning_rate": 9.142102664290775e-06, + "loss": 0.7116, + "step": 50700 + }, + { + "epoch": 3.26, + "learning_rate": 9.141887975267826e-06, + "loss": 0.7902, + "step": 50701 + }, + { + "epoch": 3.26, + "learning_rate": 9.141673286244875e-06, + "loss": 0.9051, + "step": 50702 + }, + { + "epoch": 3.26, + "learning_rate": 9.141458597221925e-06, + "loss": 0.8257, + "step": 50703 + }, + { + "epoch": 3.26, + "learning_rate": 9.141243908198974e-06, + "loss": 0.8108, + "step": 50704 + }, + { + "epoch": 3.26, + "learning_rate": 9.141029219176025e-06, + "loss": 0.7713, + "step": 50705 + }, + { + "epoch": 3.26, + "learning_rate": 9.140814530153074e-06, + "loss": 0.8666, + "step": 50706 + }, + { + "epoch": 3.26, + "learning_rate": 9.140599841130123e-06, + "loss": 0.8401, + "step": 50707 + }, + { + "epoch": 3.26, + "learning_rate": 9.140385152107173e-06, + "loss": 0.8298, + "step": 50708 + }, + { + "epoch": 3.26, + "learning_rate": 9.140170463084224e-06, + "loss": 0.7922, + "step": 50709 + }, + { + "epoch": 3.26, + "learning_rate": 9.139955774061273e-06, + "loss": 0.7823, + "step": 50710 + }, + { + "epoch": 3.26, + "learning_rate": 9.139741085038322e-06, + "loss": 0.7781, + "step": 50711 + }, + { + "epoch": 3.26, + "learning_rate": 9.139526396015373e-06, + "loss": 0.9107, + "step": 50712 + }, + { + "epoch": 3.26, + "learning_rate": 9.139311706992422e-06, + "loss": 0.8932, + "step": 50713 + }, + { + "epoch": 3.26, + "learning_rate": 9.139097017969472e-06, + "loss": 0.7565, + "step": 50714 + }, + { + "epoch": 3.26, + "learning_rate": 9.138882328946521e-06, + "loss": 0.8508, + "step": 50715 + }, + { + "epoch": 3.26, + "learning_rate": 9.138667639923572e-06, + "loss": 0.8309, + "step": 50716 + }, + { + "epoch": 3.26, + "learning_rate": 9.138452950900621e-06, + "loss": 0.8362, + "step": 50717 + }, + { + "epoch": 3.26, + "learning_rate": 9.138238261877672e-06, + "loss": 0.7963, + "step": 50718 + }, + { + "epoch": 3.26, + "learning_rate": 9.13802357285472e-06, + "loss": 0.8221, + "step": 50719 + }, + { + "epoch": 3.26, + "learning_rate": 9.13780888383177e-06, + "loss": 0.8089, + "step": 50720 + }, + { + "epoch": 3.26, + "learning_rate": 9.13759419480882e-06, + "loss": 0.7968, + "step": 50721 + }, + { + "epoch": 3.26, + "learning_rate": 9.13737950578587e-06, + "loss": 0.815, + "step": 50722 + }, + { + "epoch": 3.26, + "learning_rate": 9.13716481676292e-06, + "loss": 0.8382, + "step": 50723 + }, + { + "epoch": 3.26, + "learning_rate": 9.13695012773997e-06, + "loss": 0.8579, + "step": 50724 + }, + { + "epoch": 3.26, + "learning_rate": 9.13673543871702e-06, + "loss": 0.8831, + "step": 50725 + }, + { + "epoch": 3.26, + "learning_rate": 9.136520749694068e-06, + "loss": 0.8235, + "step": 50726 + }, + { + "epoch": 3.26, + "learning_rate": 9.136306060671119e-06, + "loss": 0.832, + "step": 50727 + }, + { + "epoch": 3.26, + "learning_rate": 9.136091371648168e-06, + "loss": 0.8593, + "step": 50728 + }, + { + "epoch": 3.26, + "learning_rate": 9.135876682625217e-06, + "loss": 0.8152, + "step": 50729 + }, + { + "epoch": 3.26, + "learning_rate": 9.135661993602268e-06, + "loss": 0.8222, + "step": 50730 + }, + { + "epoch": 3.26, + "learning_rate": 9.135447304579318e-06, + "loss": 0.8151, + "step": 50731 + }, + { + "epoch": 3.26, + "learning_rate": 9.135232615556367e-06, + "loss": 0.7771, + "step": 50732 + }, + { + "epoch": 3.26, + "learning_rate": 9.135017926533418e-06, + "loss": 0.8224, + "step": 50733 + }, + { + "epoch": 3.26, + "learning_rate": 9.134803237510467e-06, + "loss": 0.8096, + "step": 50734 + }, + { + "epoch": 3.26, + "learning_rate": 9.134588548487516e-06, + "loss": 0.8525, + "step": 50735 + }, + { + "epoch": 3.26, + "learning_rate": 9.134373859464567e-06, + "loss": 0.7656, + "step": 50736 + }, + { + "epoch": 3.26, + "learning_rate": 9.134159170441617e-06, + "loss": 0.7751, + "step": 50737 + }, + { + "epoch": 3.26, + "learning_rate": 9.133944481418666e-06, + "loss": 0.7977, + "step": 50738 + }, + { + "epoch": 3.26, + "learning_rate": 9.133729792395715e-06, + "loss": 0.7577, + "step": 50739 + }, + { + "epoch": 3.26, + "learning_rate": 9.133515103372766e-06, + "loss": 0.8748, + "step": 50740 + }, + { + "epoch": 3.26, + "learning_rate": 9.133300414349815e-06, + "loss": 0.8321, + "step": 50741 + }, + { + "epoch": 3.26, + "learning_rate": 9.133085725326864e-06, + "loss": 0.8434, + "step": 50742 + }, + { + "epoch": 3.26, + "learning_rate": 9.132871036303914e-06, + "loss": 0.8376, + "step": 50743 + }, + { + "epoch": 3.26, + "learning_rate": 9.132656347280963e-06, + "loss": 0.8917, + "step": 50744 + }, + { + "epoch": 3.26, + "learning_rate": 9.132441658258014e-06, + "loss": 0.8218, + "step": 50745 + }, + { + "epoch": 3.26, + "learning_rate": 9.132226969235065e-06, + "loss": 0.7629, + "step": 50746 + }, + { + "epoch": 3.26, + "learning_rate": 9.132012280212114e-06, + "loss": 0.8416, + "step": 50747 + }, + { + "epoch": 3.27, + "learning_rate": 9.131797591189163e-06, + "loss": 0.8131, + "step": 50748 + }, + { + "epoch": 3.27, + "learning_rate": 9.131582902166213e-06, + "loss": 0.8574, + "step": 50749 + }, + { + "epoch": 3.27, + "learning_rate": 9.131368213143262e-06, + "loss": 0.8607, + "step": 50750 + }, + { + "epoch": 3.27, + "learning_rate": 9.131153524120311e-06, + "loss": 0.7919, + "step": 50751 + }, + { + "epoch": 3.27, + "learning_rate": 9.130938835097362e-06, + "loss": 0.8958, + "step": 50752 + }, + { + "epoch": 3.27, + "learning_rate": 9.130724146074412e-06, + "loss": 0.8657, + "step": 50753 + }, + { + "epoch": 3.27, + "learning_rate": 9.130509457051461e-06, + "loss": 0.7912, + "step": 50754 + }, + { + "epoch": 3.27, + "learning_rate": 9.130294768028512e-06, + "loss": 0.7962, + "step": 50755 + }, + { + "epoch": 3.27, + "learning_rate": 9.130080079005561e-06, + "loss": 0.8505, + "step": 50756 + }, + { + "epoch": 3.27, + "learning_rate": 9.12986538998261e-06, + "loss": 0.8716, + "step": 50757 + }, + { + "epoch": 3.27, + "learning_rate": 9.12965070095966e-06, + "loss": 0.8303, + "step": 50758 + }, + { + "epoch": 3.27, + "learning_rate": 9.129436011936711e-06, + "loss": 0.8061, + "step": 50759 + }, + { + "epoch": 3.27, + "learning_rate": 9.12922132291376e-06, + "loss": 0.8242, + "step": 50760 + }, + { + "epoch": 3.27, + "learning_rate": 9.12900663389081e-06, + "loss": 0.7942, + "step": 50761 + }, + { + "epoch": 3.27, + "learning_rate": 9.12879194486786e-06, + "loss": 0.8149, + "step": 50762 + }, + { + "epoch": 3.27, + "learning_rate": 9.128577255844909e-06, + "loss": 0.8187, + "step": 50763 + }, + { + "epoch": 3.27, + "learning_rate": 9.12836256682196e-06, + "loss": 0.79, + "step": 50764 + }, + { + "epoch": 3.27, + "learning_rate": 9.128147877799009e-06, + "loss": 0.8557, + "step": 50765 + }, + { + "epoch": 3.27, + "learning_rate": 9.12793318877606e-06, + "loss": 0.7571, + "step": 50766 + }, + { + "epoch": 3.27, + "learning_rate": 9.127718499753108e-06, + "loss": 0.8794, + "step": 50767 + }, + { + "epoch": 3.27, + "learning_rate": 9.127503810730159e-06, + "loss": 0.733, + "step": 50768 + }, + { + "epoch": 3.27, + "learning_rate": 9.127289121707208e-06, + "loss": 0.817, + "step": 50769 + }, + { + "epoch": 3.27, + "learning_rate": 9.127074432684257e-06, + "loss": 0.808, + "step": 50770 + }, + { + "epoch": 3.27, + "learning_rate": 9.126859743661307e-06, + "loss": 0.8123, + "step": 50771 + }, + { + "epoch": 3.27, + "learning_rate": 9.126645054638356e-06, + "loss": 0.8302, + "step": 50772 + }, + { + "epoch": 3.27, + "learning_rate": 9.126430365615407e-06, + "loss": 0.876, + "step": 50773 + }, + { + "epoch": 3.27, + "learning_rate": 9.126215676592458e-06, + "loss": 0.8632, + "step": 50774 + }, + { + "epoch": 3.27, + "learning_rate": 9.126000987569507e-06, + "loss": 0.7321, + "step": 50775 + }, + { + "epoch": 3.27, + "learning_rate": 9.125786298546556e-06, + "loss": 0.7734, + "step": 50776 + }, + { + "epoch": 3.27, + "learning_rate": 9.125571609523606e-06, + "loss": 0.8154, + "step": 50777 + }, + { + "epoch": 3.27, + "learning_rate": 9.125356920500655e-06, + "loss": 0.8143, + "step": 50778 + }, + { + "epoch": 3.27, + "learning_rate": 9.125142231477704e-06, + "loss": 0.7632, + "step": 50779 + }, + { + "epoch": 3.27, + "learning_rate": 9.124927542454755e-06, + "loss": 0.8258, + "step": 50780 + }, + { + "epoch": 3.27, + "learning_rate": 9.124712853431806e-06, + "loss": 0.8151, + "step": 50781 + }, + { + "epoch": 3.27, + "learning_rate": 9.124498164408855e-06, + "loss": 0.7596, + "step": 50782 + }, + { + "epoch": 3.27, + "learning_rate": 9.124283475385905e-06, + "loss": 0.8026, + "step": 50783 + }, + { + "epoch": 3.27, + "learning_rate": 9.124068786362954e-06, + "loss": 0.8326, + "step": 50784 + }, + { + "epoch": 3.27, + "learning_rate": 9.123854097340003e-06, + "loss": 0.8321, + "step": 50785 + }, + { + "epoch": 3.27, + "learning_rate": 9.123639408317054e-06, + "loss": 0.8424, + "step": 50786 + }, + { + "epoch": 3.27, + "learning_rate": 9.123424719294103e-06, + "loss": 0.8311, + "step": 50787 + }, + { + "epoch": 3.27, + "learning_rate": 9.123210030271153e-06, + "loss": 0.7988, + "step": 50788 + }, + { + "epoch": 3.27, + "learning_rate": 9.122995341248202e-06, + "loss": 0.8331, + "step": 50789 + }, + { + "epoch": 3.27, + "learning_rate": 9.122780652225253e-06, + "loss": 0.8473, + "step": 50790 + }, + { + "epoch": 3.27, + "learning_rate": 9.122565963202302e-06, + "loss": 0.8636, + "step": 50791 + }, + { + "epoch": 3.27, + "learning_rate": 9.122351274179353e-06, + "loss": 0.8363, + "step": 50792 + }, + { + "epoch": 3.27, + "learning_rate": 9.122136585156402e-06, + "loss": 0.7955, + "step": 50793 + }, + { + "epoch": 3.27, + "learning_rate": 9.12192189613345e-06, + "loss": 0.8745, + "step": 50794 + }, + { + "epoch": 3.27, + "learning_rate": 9.121707207110501e-06, + "loss": 0.8329, + "step": 50795 + }, + { + "epoch": 3.27, + "learning_rate": 9.121492518087552e-06, + "loss": 0.8312, + "step": 50796 + }, + { + "epoch": 3.27, + "learning_rate": 9.1212778290646e-06, + "loss": 0.8734, + "step": 50797 + }, + { + "epoch": 3.27, + "learning_rate": 9.12106314004165e-06, + "loss": 0.801, + "step": 50798 + }, + { + "epoch": 3.27, + "learning_rate": 9.1208484510187e-06, + "loss": 0.8116, + "step": 50799 + }, + { + "epoch": 3.27, + "learning_rate": 9.12063376199575e-06, + "loss": 0.8029, + "step": 50800 + }, + { + "epoch": 3.27, + "learning_rate": 9.1204190729728e-06, + "loss": 0.8533, + "step": 50801 + }, + { + "epoch": 3.27, + "learning_rate": 9.120204383949849e-06, + "loss": 0.8588, + "step": 50802 + }, + { + "epoch": 3.27, + "learning_rate": 9.1199896949269e-06, + "loss": 0.8142, + "step": 50803 + }, + { + "epoch": 3.27, + "learning_rate": 9.119775005903949e-06, + "loss": 0.8404, + "step": 50804 + }, + { + "epoch": 3.27, + "learning_rate": 9.119560316881e-06, + "loss": 0.913, + "step": 50805 + }, + { + "epoch": 3.27, + "learning_rate": 9.119345627858048e-06, + "loss": 0.8287, + "step": 50806 + }, + { + "epoch": 3.27, + "learning_rate": 9.119130938835097e-06, + "loss": 0.8043, + "step": 50807 + }, + { + "epoch": 3.27, + "learning_rate": 9.118916249812148e-06, + "loss": 0.8299, + "step": 50808 + }, + { + "epoch": 3.27, + "learning_rate": 9.118701560789197e-06, + "loss": 0.8602, + "step": 50809 + }, + { + "epoch": 3.27, + "learning_rate": 9.118486871766248e-06, + "loss": 0.8367, + "step": 50810 + }, + { + "epoch": 3.27, + "learning_rate": 9.118272182743298e-06, + "loss": 0.8105, + "step": 50811 + }, + { + "epoch": 3.27, + "learning_rate": 9.118057493720347e-06, + "loss": 0.8167, + "step": 50812 + }, + { + "epoch": 3.27, + "learning_rate": 9.117842804697396e-06, + "loss": 0.8624, + "step": 50813 + }, + { + "epoch": 3.27, + "learning_rate": 9.117628115674447e-06, + "loss": 0.7761, + "step": 50814 + }, + { + "epoch": 3.27, + "learning_rate": 9.117413426651496e-06, + "loss": 0.8257, + "step": 50815 + }, + { + "epoch": 3.27, + "learning_rate": 9.117198737628545e-06, + "loss": 0.8472, + "step": 50816 + }, + { + "epoch": 3.27, + "learning_rate": 9.116984048605595e-06, + "loss": 0.782, + "step": 50817 + }, + { + "epoch": 3.27, + "learning_rate": 9.116769359582646e-06, + "loss": 0.8098, + "step": 50818 + }, + { + "epoch": 3.27, + "learning_rate": 9.116554670559695e-06, + "loss": 0.8181, + "step": 50819 + }, + { + "epoch": 3.27, + "learning_rate": 9.116339981536746e-06, + "loss": 0.7773, + "step": 50820 + }, + { + "epoch": 3.27, + "learning_rate": 9.116125292513795e-06, + "loss": 0.7921, + "step": 50821 + }, + { + "epoch": 3.27, + "learning_rate": 9.115910603490844e-06, + "loss": 0.8021, + "step": 50822 + }, + { + "epoch": 3.27, + "learning_rate": 9.115695914467894e-06, + "loss": 0.8197, + "step": 50823 + }, + { + "epoch": 3.27, + "learning_rate": 9.115481225444945e-06, + "loss": 0.7878, + "step": 50824 + }, + { + "epoch": 3.27, + "learning_rate": 9.115266536421994e-06, + "loss": 0.8722, + "step": 50825 + }, + { + "epoch": 3.27, + "learning_rate": 9.115051847399043e-06, + "loss": 0.7793, + "step": 50826 + }, + { + "epoch": 3.27, + "learning_rate": 9.114837158376093e-06, + "loss": 0.8603, + "step": 50827 + }, + { + "epoch": 3.27, + "learning_rate": 9.114622469353142e-06, + "loss": 0.8804, + "step": 50828 + }, + { + "epoch": 3.27, + "learning_rate": 9.114407780330191e-06, + "loss": 0.9438, + "step": 50829 + }, + { + "epoch": 3.27, + "learning_rate": 9.114193091307242e-06, + "loss": 0.8946, + "step": 50830 + }, + { + "epoch": 3.27, + "learning_rate": 9.113978402284291e-06, + "loss": 0.7683, + "step": 50831 + }, + { + "epoch": 3.27, + "learning_rate": 9.113763713261342e-06, + "loss": 0.7451, + "step": 50832 + }, + { + "epoch": 3.27, + "learning_rate": 9.113549024238392e-06, + "loss": 0.8008, + "step": 50833 + }, + { + "epoch": 3.27, + "learning_rate": 9.113334335215441e-06, + "loss": 0.8176, + "step": 50834 + }, + { + "epoch": 3.27, + "learning_rate": 9.11311964619249e-06, + "loss": 0.8155, + "step": 50835 + }, + { + "epoch": 3.27, + "learning_rate": 9.112904957169541e-06, + "loss": 0.7942, + "step": 50836 + }, + { + "epoch": 3.27, + "learning_rate": 9.11269026814659e-06, + "loss": 0.8024, + "step": 50837 + }, + { + "epoch": 3.27, + "learning_rate": 9.112475579123639e-06, + "loss": 0.8255, + "step": 50838 + }, + { + "epoch": 3.27, + "learning_rate": 9.11226089010069e-06, + "loss": 0.7921, + "step": 50839 + }, + { + "epoch": 3.27, + "learning_rate": 9.11204620107774e-06, + "loss": 0.8008, + "step": 50840 + }, + { + "epoch": 3.27, + "learning_rate": 9.11183151205479e-06, + "loss": 0.8676, + "step": 50841 + }, + { + "epoch": 3.27, + "learning_rate": 9.11161682303184e-06, + "loss": 0.8118, + "step": 50842 + }, + { + "epoch": 3.27, + "learning_rate": 9.111402134008889e-06, + "loss": 0.8285, + "step": 50843 + }, + { + "epoch": 3.27, + "learning_rate": 9.111187444985938e-06, + "loss": 0.7713, + "step": 50844 + }, + { + "epoch": 3.27, + "learning_rate": 9.110972755962988e-06, + "loss": 0.8483, + "step": 50845 + }, + { + "epoch": 3.27, + "learning_rate": 9.110758066940039e-06, + "loss": 0.8252, + "step": 50846 + }, + { + "epoch": 3.27, + "learning_rate": 9.110543377917088e-06, + "loss": 0.8167, + "step": 50847 + }, + { + "epoch": 3.27, + "learning_rate": 9.110328688894137e-06, + "loss": 0.7752, + "step": 50848 + }, + { + "epoch": 3.27, + "learning_rate": 9.110113999871188e-06, + "loss": 0.7735, + "step": 50849 + }, + { + "epoch": 3.27, + "learning_rate": 9.109899310848237e-06, + "loss": 0.7964, + "step": 50850 + }, + { + "epoch": 3.27, + "learning_rate": 9.109684621825287e-06, + "loss": 0.8055, + "step": 50851 + }, + { + "epoch": 3.27, + "learning_rate": 9.109469932802336e-06, + "loss": 0.8067, + "step": 50852 + }, + { + "epoch": 3.27, + "learning_rate": 9.109255243779387e-06, + "loss": 0.7591, + "step": 50853 + }, + { + "epoch": 3.27, + "learning_rate": 9.109040554756436e-06, + "loss": 0.817, + "step": 50854 + }, + { + "epoch": 3.27, + "learning_rate": 9.108825865733487e-06, + "loss": 0.8559, + "step": 50855 + }, + { + "epoch": 3.27, + "learning_rate": 9.108611176710536e-06, + "loss": 0.8288, + "step": 50856 + }, + { + "epoch": 3.27, + "learning_rate": 9.108396487687584e-06, + "loss": 0.7935, + "step": 50857 + }, + { + "epoch": 3.27, + "learning_rate": 9.108181798664635e-06, + "loss": 0.8246, + "step": 50858 + }, + { + "epoch": 3.27, + "learning_rate": 9.107967109641684e-06, + "loss": 0.7917, + "step": 50859 + }, + { + "epoch": 3.27, + "learning_rate": 9.107752420618735e-06, + "loss": 0.8269, + "step": 50860 + }, + { + "epoch": 3.27, + "learning_rate": 9.107537731595785e-06, + "loss": 0.8039, + "step": 50861 + }, + { + "epoch": 3.27, + "learning_rate": 9.107323042572834e-06, + "loss": 0.8507, + "step": 50862 + }, + { + "epoch": 3.27, + "learning_rate": 9.107108353549883e-06, + "loss": 0.8115, + "step": 50863 + }, + { + "epoch": 3.27, + "learning_rate": 9.106893664526934e-06, + "loss": 0.8128, + "step": 50864 + }, + { + "epoch": 3.27, + "learning_rate": 9.106678975503983e-06, + "loss": 0.8173, + "step": 50865 + }, + { + "epoch": 3.27, + "learning_rate": 9.106464286481032e-06, + "loss": 0.8362, + "step": 50866 + }, + { + "epoch": 3.27, + "learning_rate": 9.106249597458083e-06, + "loss": 0.8096, + "step": 50867 + }, + { + "epoch": 3.27, + "learning_rate": 9.106034908435133e-06, + "loss": 0.8012, + "step": 50868 + }, + { + "epoch": 3.27, + "learning_rate": 9.105820219412182e-06, + "loss": 0.8054, + "step": 50869 + }, + { + "epoch": 3.27, + "learning_rate": 9.105605530389233e-06, + "loss": 0.8574, + "step": 50870 + }, + { + "epoch": 3.27, + "learning_rate": 9.105390841366282e-06, + "loss": 0.8566, + "step": 50871 + }, + { + "epoch": 3.27, + "learning_rate": 9.10517615234333e-06, + "loss": 0.896, + "step": 50872 + }, + { + "epoch": 3.27, + "learning_rate": 9.104961463320381e-06, + "loss": 0.7672, + "step": 50873 + }, + { + "epoch": 3.27, + "learning_rate": 9.10474677429743e-06, + "loss": 0.8013, + "step": 50874 + }, + { + "epoch": 3.27, + "learning_rate": 9.104532085274481e-06, + "loss": 0.863, + "step": 50875 + }, + { + "epoch": 3.27, + "learning_rate": 9.10431739625153e-06, + "loss": 0.7979, + "step": 50876 + }, + { + "epoch": 3.27, + "learning_rate": 9.10410270722858e-06, + "loss": 0.7634, + "step": 50877 + }, + { + "epoch": 3.27, + "learning_rate": 9.10388801820563e-06, + "loss": 0.8734, + "step": 50878 + }, + { + "epoch": 3.27, + "learning_rate": 9.10367332918268e-06, + "loss": 0.7801, + "step": 50879 + }, + { + "epoch": 3.27, + "learning_rate": 9.10345864015973e-06, + "loss": 0.7946, + "step": 50880 + }, + { + "epoch": 3.27, + "learning_rate": 9.103243951136778e-06, + "loss": 0.8308, + "step": 50881 + }, + { + "epoch": 3.27, + "learning_rate": 9.103029262113829e-06, + "loss": 0.8291, + "step": 50882 + }, + { + "epoch": 3.27, + "learning_rate": 9.10281457309088e-06, + "loss": 0.7562, + "step": 50883 + }, + { + "epoch": 3.27, + "learning_rate": 9.102599884067929e-06, + "loss": 0.7998, + "step": 50884 + }, + { + "epoch": 3.27, + "learning_rate": 9.102385195044978e-06, + "loss": 0.8242, + "step": 50885 + }, + { + "epoch": 3.27, + "learning_rate": 9.102170506022028e-06, + "loss": 0.8128, + "step": 50886 + }, + { + "epoch": 3.27, + "learning_rate": 9.101955816999077e-06, + "loss": 0.7961, + "step": 50887 + }, + { + "epoch": 3.27, + "learning_rate": 9.101741127976128e-06, + "loss": 0.8272, + "step": 50888 + }, + { + "epoch": 3.27, + "learning_rate": 9.101526438953177e-06, + "loss": 0.8429, + "step": 50889 + }, + { + "epoch": 3.27, + "learning_rate": 9.101311749930227e-06, + "loss": 0.7571, + "step": 50890 + }, + { + "epoch": 3.27, + "learning_rate": 9.101097060907276e-06, + "loss": 0.8123, + "step": 50891 + }, + { + "epoch": 3.27, + "learning_rate": 9.100882371884327e-06, + "loss": 0.8273, + "step": 50892 + }, + { + "epoch": 3.27, + "learning_rate": 9.100667682861376e-06, + "loss": 0.8512, + "step": 50893 + }, + { + "epoch": 3.27, + "learning_rate": 9.100452993838425e-06, + "loss": 0.8442, + "step": 50894 + }, + { + "epoch": 3.27, + "learning_rate": 9.100238304815476e-06, + "loss": 0.8238, + "step": 50895 + }, + { + "epoch": 3.27, + "learning_rate": 9.100023615792525e-06, + "loss": 0.8539, + "step": 50896 + }, + { + "epoch": 3.27, + "learning_rate": 9.099808926769575e-06, + "loss": 0.7711, + "step": 50897 + }, + { + "epoch": 3.27, + "learning_rate": 9.099594237746626e-06, + "loss": 0.8182, + "step": 50898 + }, + { + "epoch": 3.27, + "learning_rate": 9.099379548723675e-06, + "loss": 0.7817, + "step": 50899 + }, + { + "epoch": 3.27, + "learning_rate": 9.099164859700724e-06, + "loss": 0.831, + "step": 50900 + }, + { + "epoch": 3.27, + "learning_rate": 9.098950170677775e-06, + "loss": 0.8429, + "step": 50901 + }, + { + "epoch": 3.27, + "learning_rate": 9.098735481654823e-06, + "loss": 0.8657, + "step": 50902 + }, + { + "epoch": 3.27, + "learning_rate": 9.098520792631872e-06, + "loss": 0.8775, + "step": 50903 + }, + { + "epoch": 3.28, + "learning_rate": 9.098306103608923e-06, + "loss": 0.8056, + "step": 50904 + }, + { + "epoch": 3.28, + "learning_rate": 9.098091414585974e-06, + "loss": 0.8857, + "step": 50905 + }, + { + "epoch": 3.28, + "learning_rate": 9.097876725563023e-06, + "loss": 0.8314, + "step": 50906 + }, + { + "epoch": 3.28, + "learning_rate": 9.097662036540072e-06, + "loss": 0.8157, + "step": 50907 + }, + { + "epoch": 3.28, + "learning_rate": 9.097447347517122e-06, + "loss": 0.8084, + "step": 50908 + }, + { + "epoch": 3.28, + "learning_rate": 9.097232658494171e-06, + "loss": 0.8454, + "step": 50909 + }, + { + "epoch": 3.28, + "learning_rate": 9.097017969471222e-06, + "loss": 0.7521, + "step": 50910 + }, + { + "epoch": 3.28, + "learning_rate": 9.096803280448271e-06, + "loss": 0.8354, + "step": 50911 + }, + { + "epoch": 3.28, + "learning_rate": 9.096588591425322e-06, + "loss": 0.8559, + "step": 50912 + }, + { + "epoch": 3.28, + "learning_rate": 9.09637390240237e-06, + "loss": 0.7986, + "step": 50913 + }, + { + "epoch": 3.28, + "learning_rate": 9.096159213379421e-06, + "loss": 0.917, + "step": 50914 + }, + { + "epoch": 3.28, + "learning_rate": 9.09594452435647e-06, + "loss": 0.7899, + "step": 50915 + }, + { + "epoch": 3.28, + "learning_rate": 9.095729835333519e-06, + "loss": 0.8439, + "step": 50916 + }, + { + "epoch": 3.28, + "learning_rate": 9.09551514631057e-06, + "loss": 0.8807, + "step": 50917 + }, + { + "epoch": 3.28, + "learning_rate": 9.095300457287619e-06, + "loss": 0.7978, + "step": 50918 + }, + { + "epoch": 3.28, + "learning_rate": 9.09508576826467e-06, + "loss": 0.8952, + "step": 50919 + }, + { + "epoch": 3.28, + "learning_rate": 9.09487107924172e-06, + "loss": 0.7996, + "step": 50920 + }, + { + "epoch": 3.28, + "learning_rate": 9.094656390218769e-06, + "loss": 0.8656, + "step": 50921 + }, + { + "epoch": 3.28, + "learning_rate": 9.094441701195818e-06, + "loss": 0.842, + "step": 50922 + }, + { + "epoch": 3.28, + "learning_rate": 9.094227012172869e-06, + "loss": 0.8418, + "step": 50923 + }, + { + "epoch": 3.28, + "learning_rate": 9.094012323149918e-06, + "loss": 0.806, + "step": 50924 + }, + { + "epoch": 3.28, + "learning_rate": 9.093797634126967e-06, + "loss": 0.8532, + "step": 50925 + }, + { + "epoch": 3.28, + "learning_rate": 9.093582945104017e-06, + "loss": 0.8541, + "step": 50926 + }, + { + "epoch": 3.28, + "learning_rate": 9.093368256081068e-06, + "loss": 0.8159, + "step": 50927 + }, + { + "epoch": 3.28, + "learning_rate": 9.093153567058117e-06, + "loss": 0.7977, + "step": 50928 + }, + { + "epoch": 3.28, + "learning_rate": 9.092938878035168e-06, + "loss": 0.8079, + "step": 50929 + }, + { + "epoch": 3.28, + "learning_rate": 9.092724189012217e-06, + "loss": 0.8252, + "step": 50930 + }, + { + "epoch": 3.28, + "learning_rate": 9.092509499989265e-06, + "loss": 0.8688, + "step": 50931 + }, + { + "epoch": 3.28, + "learning_rate": 9.092294810966316e-06, + "loss": 0.8756, + "step": 50932 + }, + { + "epoch": 3.28, + "learning_rate": 9.092080121943367e-06, + "loss": 0.786, + "step": 50933 + }, + { + "epoch": 3.28, + "learning_rate": 9.091865432920416e-06, + "loss": 0.8725, + "step": 50934 + }, + { + "epoch": 3.28, + "learning_rate": 9.091650743897465e-06, + "loss": 0.7978, + "step": 50935 + }, + { + "epoch": 3.28, + "learning_rate": 9.091436054874515e-06, + "loss": 0.8241, + "step": 50936 + }, + { + "epoch": 3.28, + "learning_rate": 9.091221365851564e-06, + "loss": 0.8197, + "step": 50937 + }, + { + "epoch": 3.28, + "learning_rate": 9.091006676828615e-06, + "loss": 0.7825, + "step": 50938 + }, + { + "epoch": 3.28, + "learning_rate": 9.090791987805664e-06, + "loss": 0.8609, + "step": 50939 + }, + { + "epoch": 3.28, + "learning_rate": 9.090577298782713e-06, + "loss": 0.8134, + "step": 50940 + }, + { + "epoch": 3.28, + "learning_rate": 9.090362609759764e-06, + "loss": 0.7997, + "step": 50941 + }, + { + "epoch": 3.28, + "learning_rate": 9.090147920736814e-06, + "loss": 0.826, + "step": 50942 + }, + { + "epoch": 3.28, + "learning_rate": 9.089933231713863e-06, + "loss": 0.893, + "step": 50943 + }, + { + "epoch": 3.28, + "learning_rate": 9.089718542690912e-06, + "loss": 0.872, + "step": 50944 + }, + { + "epoch": 3.28, + "learning_rate": 9.089503853667963e-06, + "loss": 0.8245, + "step": 50945 + }, + { + "epoch": 3.28, + "learning_rate": 9.089289164645012e-06, + "loss": 0.7882, + "step": 50946 + }, + { + "epoch": 3.28, + "learning_rate": 9.089074475622062e-06, + "loss": 0.8692, + "step": 50947 + }, + { + "epoch": 3.28, + "learning_rate": 9.088859786599113e-06, + "loss": 0.8668, + "step": 50948 + }, + { + "epoch": 3.28, + "learning_rate": 9.088645097576162e-06, + "loss": 0.7838, + "step": 50949 + }, + { + "epoch": 3.28, + "learning_rate": 9.088430408553211e-06, + "loss": 0.8169, + "step": 50950 + }, + { + "epoch": 3.28, + "learning_rate": 9.088215719530262e-06, + "loss": 0.8628, + "step": 50951 + }, + { + "epoch": 3.28, + "learning_rate": 9.08800103050731e-06, + "loss": 0.8549, + "step": 50952 + }, + { + "epoch": 3.28, + "learning_rate": 9.08778634148436e-06, + "loss": 0.815, + "step": 50953 + }, + { + "epoch": 3.28, + "learning_rate": 9.08757165246141e-06, + "loss": 0.8291, + "step": 50954 + }, + { + "epoch": 3.28, + "learning_rate": 9.087356963438461e-06, + "loss": 0.7561, + "step": 50955 + }, + { + "epoch": 3.28, + "learning_rate": 9.08714227441551e-06, + "loss": 0.856, + "step": 50956 + }, + { + "epoch": 3.28, + "learning_rate": 9.08692758539256e-06, + "loss": 0.8153, + "step": 50957 + }, + { + "epoch": 3.28, + "learning_rate": 9.08671289636961e-06, + "loss": 0.8509, + "step": 50958 + }, + { + "epoch": 3.28, + "learning_rate": 9.086498207346659e-06, + "loss": 0.7839, + "step": 50959 + }, + { + "epoch": 3.28, + "learning_rate": 9.08628351832371e-06, + "loss": 0.8375, + "step": 50960 + }, + { + "epoch": 3.28, + "learning_rate": 9.086068829300758e-06, + "loss": 0.8045, + "step": 50961 + }, + { + "epoch": 3.28, + "learning_rate": 9.085854140277809e-06, + "loss": 0.8164, + "step": 50962 + }, + { + "epoch": 3.28, + "learning_rate": 9.085639451254858e-06, + "loss": 0.8391, + "step": 50963 + }, + { + "epoch": 3.28, + "learning_rate": 9.085424762231908e-06, + "loss": 0.7903, + "step": 50964 + }, + { + "epoch": 3.28, + "learning_rate": 9.085210073208957e-06, + "loss": 0.7983, + "step": 50965 + }, + { + "epoch": 3.28, + "learning_rate": 9.084995384186008e-06, + "loss": 0.8301, + "step": 50966 + }, + { + "epoch": 3.28, + "learning_rate": 9.084780695163057e-06, + "loss": 0.8299, + "step": 50967 + }, + { + "epoch": 3.28, + "learning_rate": 9.084566006140106e-06, + "loss": 0.9067, + "step": 50968 + }, + { + "epoch": 3.28, + "learning_rate": 9.084351317117157e-06, + "loss": 0.8715, + "step": 50969 + }, + { + "epoch": 3.28, + "learning_rate": 9.084136628094207e-06, + "loss": 0.8142, + "step": 50970 + }, + { + "epoch": 3.28, + "learning_rate": 9.083921939071256e-06, + "loss": 0.7542, + "step": 50971 + }, + { + "epoch": 3.28, + "learning_rate": 9.083707250048305e-06, + "loss": 0.8453, + "step": 50972 + }, + { + "epoch": 3.28, + "learning_rate": 9.083492561025356e-06, + "loss": 0.8115, + "step": 50973 + }, + { + "epoch": 3.28, + "learning_rate": 9.083277872002405e-06, + "loss": 0.8535, + "step": 50974 + }, + { + "epoch": 3.28, + "learning_rate": 9.083063182979456e-06, + "loss": 0.8663, + "step": 50975 + }, + { + "epoch": 3.28, + "learning_rate": 9.082848493956504e-06, + "loss": 0.7419, + "step": 50976 + }, + { + "epoch": 3.28, + "learning_rate": 9.082633804933555e-06, + "loss": 0.7841, + "step": 50977 + }, + { + "epoch": 3.28, + "learning_rate": 9.082419115910604e-06, + "loss": 0.8329, + "step": 50978 + }, + { + "epoch": 3.28, + "learning_rate": 9.082204426887655e-06, + "loss": 0.8592, + "step": 50979 + }, + { + "epoch": 3.28, + "learning_rate": 9.081989737864704e-06, + "loss": 0.8049, + "step": 50980 + }, + { + "epoch": 3.28, + "learning_rate": 9.081775048841753e-06, + "loss": 0.8426, + "step": 50981 + }, + { + "epoch": 3.28, + "learning_rate": 9.081560359818803e-06, + "loss": 0.802, + "step": 50982 + }, + { + "epoch": 3.28, + "learning_rate": 9.081345670795852e-06, + "loss": 0.856, + "step": 50983 + }, + { + "epoch": 3.28, + "learning_rate": 9.081130981772903e-06, + "loss": 0.8222, + "step": 50984 + }, + { + "epoch": 3.28, + "learning_rate": 9.080916292749954e-06, + "loss": 0.7513, + "step": 50985 + }, + { + "epoch": 3.28, + "learning_rate": 9.080701603727003e-06, + "loss": 0.8197, + "step": 50986 + }, + { + "epoch": 3.28, + "learning_rate": 9.080486914704052e-06, + "loss": 0.7961, + "step": 50987 + }, + { + "epoch": 3.28, + "learning_rate": 9.080272225681102e-06, + "loss": 0.8242, + "step": 50988 + }, + { + "epoch": 3.28, + "learning_rate": 9.080057536658151e-06, + "loss": 0.7829, + "step": 50989 + }, + { + "epoch": 3.28, + "learning_rate": 9.0798428476352e-06, + "loss": 0.8348, + "step": 50990 + }, + { + "epoch": 3.28, + "learning_rate": 9.07962815861225e-06, + "loss": 0.8132, + "step": 50991 + }, + { + "epoch": 3.28, + "learning_rate": 9.079413469589301e-06, + "loss": 0.8588, + "step": 50992 + }, + { + "epoch": 3.28, + "learning_rate": 9.07919878056635e-06, + "loss": 0.8621, + "step": 50993 + }, + { + "epoch": 3.28, + "learning_rate": 9.0789840915434e-06, + "loss": 0.7938, + "step": 50994 + }, + { + "epoch": 3.28, + "learning_rate": 9.07876940252045e-06, + "loss": 0.8366, + "step": 50995 + }, + { + "epoch": 3.28, + "learning_rate": 9.078554713497499e-06, + "loss": 0.8513, + "step": 50996 + }, + { + "epoch": 3.28, + "learning_rate": 9.07834002447455e-06, + "loss": 0.8524, + "step": 50997 + }, + { + "epoch": 3.28, + "learning_rate": 9.078125335451599e-06, + "loss": 0.7471, + "step": 50998 + }, + { + "epoch": 3.28, + "learning_rate": 9.07791064642865e-06, + "loss": 0.7836, + "step": 50999 + }, + { + "epoch": 3.28, + "learning_rate": 9.077695957405698e-06, + "loss": 0.8209, + "step": 51000 + }, + { + "epoch": 3.28, + "eval_laion/cider": 0.5772483171723494, + "eval_laion/meteor": 0.3959704103911782, + "eval_laion/pred_num_words": 37.103515625, + "eval_laion/sacrebleu": 15.950138207452015, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0537407398223877, + "eval_runtime": 96.0365, + "eval_samples_per_second": 5.196, + "eval_steps_per_second": 0.333, + "step": 51000 + }, + { + "epoch": 3.28, + "learning_rate": 9.077481268382749e-06, + "loss": 0.8506, + "step": 51001 + }, + { + "epoch": 3.28, + "learning_rate": 9.077266579359798e-06, + "loss": 0.8283, + "step": 51002 + }, + { + "epoch": 3.28, + "learning_rate": 9.077051890336847e-06, + "loss": 0.8618, + "step": 51003 + }, + { + "epoch": 3.28, + "learning_rate": 9.076837201313898e-06, + "loss": 0.8224, + "step": 51004 + }, + { + "epoch": 3.28, + "learning_rate": 9.076622512290946e-06, + "loss": 0.7656, + "step": 51005 + }, + { + "epoch": 3.28, + "learning_rate": 9.076407823267997e-06, + "loss": 0.8188, + "step": 51006 + }, + { + "epoch": 3.28, + "learning_rate": 9.076193134245048e-06, + "loss": 0.8179, + "step": 51007 + }, + { + "epoch": 3.28, + "learning_rate": 9.075978445222097e-06, + "loss": 0.8327, + "step": 51008 + }, + { + "epoch": 3.28, + "learning_rate": 9.075763756199146e-06, + "loss": 0.7675, + "step": 51009 + }, + { + "epoch": 3.28, + "learning_rate": 9.075549067176196e-06, + "loss": 0.8289, + "step": 51010 + }, + { + "epoch": 3.28, + "learning_rate": 9.075334378153245e-06, + "loss": 0.7832, + "step": 51011 + }, + { + "epoch": 3.28, + "learning_rate": 9.075119689130294e-06, + "loss": 0.8513, + "step": 51012 + }, + { + "epoch": 3.28, + "learning_rate": 9.074905000107345e-06, + "loss": 0.8863, + "step": 51013 + }, + { + "epoch": 3.28, + "learning_rate": 9.074690311084396e-06, + "loss": 0.8012, + "step": 51014 + }, + { + "epoch": 3.28, + "learning_rate": 9.074475622061445e-06, + "loss": 0.8161, + "step": 51015 + }, + { + "epoch": 3.28, + "learning_rate": 9.074260933038495e-06, + "loss": 0.8266, + "step": 51016 + }, + { + "epoch": 3.28, + "learning_rate": 9.074046244015544e-06, + "loss": 0.8379, + "step": 51017 + }, + { + "epoch": 3.28, + "learning_rate": 9.073831554992593e-06, + "loss": 0.854, + "step": 51018 + }, + { + "epoch": 3.28, + "learning_rate": 9.073616865969644e-06, + "loss": 0.8031, + "step": 51019 + }, + { + "epoch": 3.28, + "learning_rate": 9.073402176946695e-06, + "loss": 0.7789, + "step": 51020 + }, + { + "epoch": 3.28, + "learning_rate": 9.073187487923743e-06, + "loss": 0.8602, + "step": 51021 + }, + { + "epoch": 3.28, + "learning_rate": 9.072972798900792e-06, + "loss": 0.7862, + "step": 51022 + }, + { + "epoch": 3.28, + "learning_rate": 9.072758109877843e-06, + "loss": 0.8342, + "step": 51023 + }, + { + "epoch": 3.28, + "learning_rate": 9.072543420854892e-06, + "loss": 0.8362, + "step": 51024 + }, + { + "epoch": 3.28, + "learning_rate": 9.072328731831943e-06, + "loss": 0.7881, + "step": 51025 + }, + { + "epoch": 3.28, + "learning_rate": 9.072114042808992e-06, + "loss": 0.8089, + "step": 51026 + }, + { + "epoch": 3.28, + "learning_rate": 9.07189935378604e-06, + "loss": 0.8575, + "step": 51027 + }, + { + "epoch": 3.28, + "learning_rate": 9.071684664763091e-06, + "loss": 0.866, + "step": 51028 + }, + { + "epoch": 3.28, + "learning_rate": 9.071469975740142e-06, + "loss": 0.7985, + "step": 51029 + }, + { + "epoch": 3.28, + "learning_rate": 9.071255286717191e-06, + "loss": 0.8325, + "step": 51030 + }, + { + "epoch": 3.28, + "learning_rate": 9.07104059769424e-06, + "loss": 0.8138, + "step": 51031 + }, + { + "epoch": 3.28, + "learning_rate": 9.07082590867129e-06, + "loss": 0.7806, + "step": 51032 + }, + { + "epoch": 3.28, + "learning_rate": 9.07061121964834e-06, + "loss": 0.8696, + "step": 51033 + }, + { + "epoch": 3.28, + "learning_rate": 9.07039653062539e-06, + "loss": 0.8519, + "step": 51034 + }, + { + "epoch": 3.28, + "learning_rate": 9.07018184160244e-06, + "loss": 0.8208, + "step": 51035 + }, + { + "epoch": 3.28, + "learning_rate": 9.06996715257949e-06, + "loss": 0.9289, + "step": 51036 + }, + { + "epoch": 3.28, + "learning_rate": 9.069752463556539e-06, + "loss": 0.8514, + "step": 51037 + }, + { + "epoch": 3.28, + "learning_rate": 9.06953777453359e-06, + "loss": 0.7867, + "step": 51038 + }, + { + "epoch": 3.28, + "learning_rate": 9.069323085510638e-06, + "loss": 0.7847, + "step": 51039 + }, + { + "epoch": 3.28, + "learning_rate": 9.069108396487687e-06, + "loss": 0.7877, + "step": 51040 + }, + { + "epoch": 3.28, + "learning_rate": 9.068893707464738e-06, + "loss": 0.8778, + "step": 51041 + }, + { + "epoch": 3.28, + "learning_rate": 9.068679018441789e-06, + "loss": 0.81, + "step": 51042 + }, + { + "epoch": 3.28, + "learning_rate": 9.068464329418838e-06, + "loss": 0.8335, + "step": 51043 + }, + { + "epoch": 3.28, + "learning_rate": 9.068249640395888e-06, + "loss": 0.8082, + "step": 51044 + }, + { + "epoch": 3.28, + "learning_rate": 9.068034951372937e-06, + "loss": 0.8015, + "step": 51045 + }, + { + "epoch": 3.28, + "learning_rate": 9.067820262349986e-06, + "loss": 0.9253, + "step": 51046 + }, + { + "epoch": 3.28, + "learning_rate": 9.067605573327037e-06, + "loss": 0.8341, + "step": 51047 + }, + { + "epoch": 3.28, + "learning_rate": 9.067390884304086e-06, + "loss": 0.8069, + "step": 51048 + }, + { + "epoch": 3.28, + "learning_rate": 9.067176195281135e-06, + "loss": 0.8706, + "step": 51049 + }, + { + "epoch": 3.28, + "learning_rate": 9.066961506258185e-06, + "loss": 0.8083, + "step": 51050 + }, + { + "epoch": 3.28, + "learning_rate": 9.066746817235236e-06, + "loss": 0.8605, + "step": 51051 + }, + { + "epoch": 3.28, + "learning_rate": 9.066532128212285e-06, + "loss": 0.8013, + "step": 51052 + }, + { + "epoch": 3.28, + "learning_rate": 9.066317439189336e-06, + "loss": 0.8004, + "step": 51053 + }, + { + "epoch": 3.28, + "learning_rate": 9.066102750166385e-06, + "loss": 0.8155, + "step": 51054 + }, + { + "epoch": 3.28, + "learning_rate": 9.065888061143434e-06, + "loss": 0.7707, + "step": 51055 + }, + { + "epoch": 3.28, + "learning_rate": 9.065673372120484e-06, + "loss": 0.8716, + "step": 51056 + }, + { + "epoch": 3.28, + "learning_rate": 9.065458683097535e-06, + "loss": 0.8087, + "step": 51057 + }, + { + "epoch": 3.28, + "learning_rate": 9.065243994074584e-06, + "loss": 0.8, + "step": 51058 + }, + { + "epoch": 3.29, + "learning_rate": 9.065029305051633e-06, + "loss": 0.7719, + "step": 51059 + }, + { + "epoch": 3.29, + "learning_rate": 9.064814616028684e-06, + "loss": 0.8608, + "step": 51060 + }, + { + "epoch": 3.29, + "learning_rate": 9.064599927005733e-06, + "loss": 0.7876, + "step": 51061 + }, + { + "epoch": 3.29, + "learning_rate": 9.064385237982782e-06, + "loss": 0.7691, + "step": 51062 + }, + { + "epoch": 3.29, + "learning_rate": 9.064170548959832e-06, + "loss": 0.823, + "step": 51063 + }, + { + "epoch": 3.29, + "learning_rate": 9.063955859936883e-06, + "loss": 0.8019, + "step": 51064 + }, + { + "epoch": 3.29, + "learning_rate": 9.063741170913932e-06, + "loss": 0.7372, + "step": 51065 + }, + { + "epoch": 3.29, + "learning_rate": 9.063526481890982e-06, + "loss": 0.7837, + "step": 51066 + }, + { + "epoch": 3.29, + "learning_rate": 9.063311792868031e-06, + "loss": 0.8934, + "step": 51067 + }, + { + "epoch": 3.29, + "learning_rate": 9.06309710384508e-06, + "loss": 0.8703, + "step": 51068 + }, + { + "epoch": 3.29, + "learning_rate": 9.062882414822131e-06, + "loss": 0.8521, + "step": 51069 + }, + { + "epoch": 3.29, + "learning_rate": 9.06266772579918e-06, + "loss": 0.868, + "step": 51070 + }, + { + "epoch": 3.29, + "learning_rate": 9.06245303677623e-06, + "loss": 0.8129, + "step": 51071 + }, + { + "epoch": 3.29, + "learning_rate": 9.06223834775328e-06, + "loss": 0.8033, + "step": 51072 + }, + { + "epoch": 3.29, + "learning_rate": 9.06202365873033e-06, + "loss": 0.892, + "step": 51073 + }, + { + "epoch": 3.29, + "learning_rate": 9.06180896970738e-06, + "loss": 0.8504, + "step": 51074 + }, + { + "epoch": 3.29, + "learning_rate": 9.06159428068443e-06, + "loss": 0.8195, + "step": 51075 + }, + { + "epoch": 3.29, + "learning_rate": 9.061379591661479e-06, + "loss": 0.854, + "step": 51076 + }, + { + "epoch": 3.29, + "learning_rate": 9.061164902638528e-06, + "loss": 0.779, + "step": 51077 + }, + { + "epoch": 3.29, + "learning_rate": 9.060950213615579e-06, + "loss": 0.8272, + "step": 51078 + }, + { + "epoch": 3.29, + "learning_rate": 9.06073552459263e-06, + "loss": 0.8409, + "step": 51079 + }, + { + "epoch": 3.29, + "learning_rate": 9.060520835569678e-06, + "loss": 0.852, + "step": 51080 + }, + { + "epoch": 3.29, + "learning_rate": 9.060306146546727e-06, + "loss": 0.8596, + "step": 51081 + }, + { + "epoch": 3.29, + "learning_rate": 9.060091457523778e-06, + "loss": 0.7532, + "step": 51082 + }, + { + "epoch": 3.29, + "learning_rate": 9.059876768500827e-06, + "loss": 0.9136, + "step": 51083 + }, + { + "epoch": 3.29, + "learning_rate": 9.059662079477877e-06, + "loss": 0.85, + "step": 51084 + }, + { + "epoch": 3.29, + "learning_rate": 9.059447390454926e-06, + "loss": 0.8205, + "step": 51085 + }, + { + "epoch": 3.29, + "learning_rate": 9.059232701431977e-06, + "loss": 0.8419, + "step": 51086 + }, + { + "epoch": 3.29, + "learning_rate": 9.059018012409026e-06, + "loss": 0.815, + "step": 51087 + }, + { + "epoch": 3.29, + "learning_rate": 9.058803323386077e-06, + "loss": 0.876, + "step": 51088 + }, + { + "epoch": 3.29, + "learning_rate": 9.058588634363126e-06, + "loss": 0.8496, + "step": 51089 + }, + { + "epoch": 3.29, + "learning_rate": 9.058373945340175e-06, + "loss": 0.817, + "step": 51090 + }, + { + "epoch": 3.29, + "learning_rate": 9.058159256317225e-06, + "loss": 0.7851, + "step": 51091 + }, + { + "epoch": 3.29, + "learning_rate": 9.057944567294274e-06, + "loss": 0.7624, + "step": 51092 + }, + { + "epoch": 3.29, + "learning_rate": 9.057729878271325e-06, + "loss": 0.7718, + "step": 51093 + }, + { + "epoch": 3.29, + "learning_rate": 9.057515189248376e-06, + "loss": 0.8603, + "step": 51094 + }, + { + "epoch": 3.29, + "learning_rate": 9.057300500225424e-06, + "loss": 0.8015, + "step": 51095 + }, + { + "epoch": 3.29, + "learning_rate": 9.057085811202473e-06, + "loss": 0.762, + "step": 51096 + }, + { + "epoch": 3.29, + "learning_rate": 9.056871122179524e-06, + "loss": 0.8722, + "step": 51097 + }, + { + "epoch": 3.29, + "learning_rate": 9.056656433156573e-06, + "loss": 0.8822, + "step": 51098 + }, + { + "epoch": 3.29, + "learning_rate": 9.056441744133622e-06, + "loss": 0.7916, + "step": 51099 + }, + { + "epoch": 3.29, + "learning_rate": 9.056227055110673e-06, + "loss": 0.8314, + "step": 51100 + }, + { + "epoch": 3.29, + "learning_rate": 9.056012366087723e-06, + "loss": 0.8152, + "step": 51101 + }, + { + "epoch": 3.29, + "learning_rate": 9.055797677064772e-06, + "loss": 0.8109, + "step": 51102 + }, + { + "epoch": 3.29, + "learning_rate": 9.055582988041823e-06, + "loss": 0.8594, + "step": 51103 + }, + { + "epoch": 3.29, + "learning_rate": 9.055368299018872e-06, + "loss": 0.8103, + "step": 51104 + }, + { + "epoch": 3.29, + "learning_rate": 9.055153609995921e-06, + "loss": 0.8031, + "step": 51105 + }, + { + "epoch": 3.29, + "learning_rate": 9.054938920972972e-06, + "loss": 0.9162, + "step": 51106 + }, + { + "epoch": 3.29, + "learning_rate": 9.05472423195002e-06, + "loss": 0.8785, + "step": 51107 + }, + { + "epoch": 3.29, + "learning_rate": 9.054509542927071e-06, + "loss": 0.8514, + "step": 51108 + }, + { + "epoch": 3.29, + "learning_rate": 9.05429485390412e-06, + "loss": 0.8586, + "step": 51109 + }, + { + "epoch": 3.29, + "learning_rate": 9.05408016488117e-06, + "loss": 0.7883, + "step": 51110 + }, + { + "epoch": 3.29, + "learning_rate": 9.05386547585822e-06, + "loss": 0.8637, + "step": 51111 + }, + { + "epoch": 3.29, + "learning_rate": 9.05365078683527e-06, + "loss": 0.8172, + "step": 51112 + }, + { + "epoch": 3.29, + "learning_rate": 9.05343609781232e-06, + "loss": 0.8043, + "step": 51113 + }, + { + "epoch": 3.29, + "learning_rate": 9.053221408789368e-06, + "loss": 0.8316, + "step": 51114 + }, + { + "epoch": 3.29, + "learning_rate": 9.053006719766419e-06, + "loss": 0.773, + "step": 51115 + }, + { + "epoch": 3.29, + "learning_rate": 9.05279203074347e-06, + "loss": 0.8181, + "step": 51116 + }, + { + "epoch": 3.29, + "learning_rate": 9.052577341720519e-06, + "loss": 0.8223, + "step": 51117 + }, + { + "epoch": 3.29, + "learning_rate": 9.052362652697568e-06, + "loss": 0.8278, + "step": 51118 + }, + { + "epoch": 3.29, + "learning_rate": 9.052147963674618e-06, + "loss": 0.8085, + "step": 51119 + }, + { + "epoch": 3.29, + "learning_rate": 9.051933274651667e-06, + "loss": 0.8277, + "step": 51120 + }, + { + "epoch": 3.29, + "learning_rate": 9.051718585628718e-06, + "loss": 0.7981, + "step": 51121 + }, + { + "epoch": 3.29, + "learning_rate": 9.051503896605769e-06, + "loss": 0.8325, + "step": 51122 + }, + { + "epoch": 3.29, + "learning_rate": 9.051289207582818e-06, + "loss": 0.8679, + "step": 51123 + }, + { + "epoch": 3.29, + "learning_rate": 9.051074518559866e-06, + "loss": 0.8236, + "step": 51124 + }, + { + "epoch": 3.29, + "learning_rate": 9.050859829536917e-06, + "loss": 0.7632, + "step": 51125 + }, + { + "epoch": 3.29, + "learning_rate": 9.050645140513966e-06, + "loss": 0.8045, + "step": 51126 + }, + { + "epoch": 3.29, + "learning_rate": 9.050430451491015e-06, + "loss": 0.8328, + "step": 51127 + }, + { + "epoch": 3.29, + "learning_rate": 9.050215762468066e-06, + "loss": 0.8412, + "step": 51128 + }, + { + "epoch": 3.29, + "learning_rate": 9.050001073445116e-06, + "loss": 0.7839, + "step": 51129 + }, + { + "epoch": 3.29, + "learning_rate": 9.049786384422165e-06, + "loss": 0.833, + "step": 51130 + }, + { + "epoch": 3.29, + "learning_rate": 9.049571695399216e-06, + "loss": 0.7967, + "step": 51131 + }, + { + "epoch": 3.29, + "learning_rate": 9.049357006376265e-06, + "loss": 0.8506, + "step": 51132 + }, + { + "epoch": 3.29, + "learning_rate": 9.049142317353314e-06, + "loss": 0.8782, + "step": 51133 + }, + { + "epoch": 3.29, + "learning_rate": 9.048927628330365e-06, + "loss": 0.8665, + "step": 51134 + }, + { + "epoch": 3.29, + "learning_rate": 9.048712939307414e-06, + "loss": 0.8771, + "step": 51135 + }, + { + "epoch": 3.29, + "learning_rate": 9.048498250284463e-06, + "loss": 0.859, + "step": 51136 + }, + { + "epoch": 3.29, + "learning_rate": 9.048283561261513e-06, + "loss": 0.7892, + "step": 51137 + }, + { + "epoch": 3.29, + "learning_rate": 9.048068872238564e-06, + "loss": 0.8676, + "step": 51138 + }, + { + "epoch": 3.29, + "learning_rate": 9.047854183215613e-06, + "loss": 0.8665, + "step": 51139 + }, + { + "epoch": 3.29, + "learning_rate": 9.047639494192663e-06, + "loss": 0.8495, + "step": 51140 + }, + { + "epoch": 3.29, + "learning_rate": 9.047424805169712e-06, + "loss": 0.8131, + "step": 51141 + }, + { + "epoch": 3.29, + "learning_rate": 9.047210116146761e-06, + "loss": 0.8376, + "step": 51142 + }, + { + "epoch": 3.29, + "learning_rate": 9.046995427123812e-06, + "loss": 0.8602, + "step": 51143 + }, + { + "epoch": 3.29, + "learning_rate": 9.046780738100863e-06, + "loss": 0.784, + "step": 51144 + }, + { + "epoch": 3.29, + "learning_rate": 9.046566049077912e-06, + "loss": 0.7848, + "step": 51145 + }, + { + "epoch": 3.29, + "learning_rate": 9.04635136005496e-06, + "loss": 0.7677, + "step": 51146 + }, + { + "epoch": 3.29, + "learning_rate": 9.046136671032011e-06, + "loss": 0.7596, + "step": 51147 + }, + { + "epoch": 3.29, + "learning_rate": 9.04592198200906e-06, + "loss": 0.8148, + "step": 51148 + }, + { + "epoch": 3.29, + "learning_rate": 9.04570729298611e-06, + "loss": 0.7788, + "step": 51149 + }, + { + "epoch": 3.29, + "learning_rate": 9.04549260396316e-06, + "loss": 0.8087, + "step": 51150 + }, + { + "epoch": 3.29, + "learning_rate": 9.04527791494021e-06, + "loss": 0.8577, + "step": 51151 + }, + { + "epoch": 3.29, + "learning_rate": 9.04506322591726e-06, + "loss": 0.8042, + "step": 51152 + }, + { + "epoch": 3.29, + "learning_rate": 9.04484853689431e-06, + "loss": 0.8195, + "step": 51153 + }, + { + "epoch": 3.29, + "learning_rate": 9.044633847871359e-06, + "loss": 0.8401, + "step": 51154 + }, + { + "epoch": 3.29, + "learning_rate": 9.044419158848408e-06, + "loss": 0.8407, + "step": 51155 + }, + { + "epoch": 3.29, + "learning_rate": 9.044204469825459e-06, + "loss": 0.7992, + "step": 51156 + }, + { + "epoch": 3.29, + "learning_rate": 9.043989780802508e-06, + "loss": 0.7473, + "step": 51157 + }, + { + "epoch": 3.29, + "learning_rate": 9.043775091779558e-06, + "loss": 0.8272, + "step": 51158 + }, + { + "epoch": 3.29, + "learning_rate": 9.043560402756607e-06, + "loss": 0.7662, + "step": 51159 + }, + { + "epoch": 3.29, + "learning_rate": 9.043345713733658e-06, + "loss": 0.8336, + "step": 51160 + }, + { + "epoch": 3.29, + "learning_rate": 9.043131024710707e-06, + "loss": 0.7838, + "step": 51161 + }, + { + "epoch": 3.29, + "learning_rate": 9.042916335687758e-06, + "loss": 0.8301, + "step": 51162 + }, + { + "epoch": 3.29, + "learning_rate": 9.042701646664807e-06, + "loss": 0.8244, + "step": 51163 + }, + { + "epoch": 3.29, + "learning_rate": 9.042486957641856e-06, + "loss": 0.8281, + "step": 51164 + }, + { + "epoch": 3.29, + "learning_rate": 9.042272268618906e-06, + "loss": 0.804, + "step": 51165 + }, + { + "epoch": 3.29, + "learning_rate": 9.042057579595957e-06, + "loss": 0.8863, + "step": 51166 + }, + { + "epoch": 3.29, + "learning_rate": 9.041842890573006e-06, + "loss": 0.8203, + "step": 51167 + }, + { + "epoch": 3.29, + "learning_rate": 9.041628201550055e-06, + "loss": 0.8166, + "step": 51168 + }, + { + "epoch": 3.29, + "learning_rate": 9.041413512527105e-06, + "loss": 0.8321, + "step": 51169 + }, + { + "epoch": 3.29, + "learning_rate": 9.041198823504154e-06, + "loss": 0.8325, + "step": 51170 + }, + { + "epoch": 3.29, + "learning_rate": 9.040984134481205e-06, + "loss": 0.8293, + "step": 51171 + }, + { + "epoch": 3.29, + "learning_rate": 9.040769445458254e-06, + "loss": 0.8738, + "step": 51172 + }, + { + "epoch": 3.29, + "learning_rate": 9.040554756435305e-06, + "loss": 0.7877, + "step": 51173 + }, + { + "epoch": 3.29, + "learning_rate": 9.040340067412354e-06, + "loss": 0.7246, + "step": 51174 + }, + { + "epoch": 3.29, + "learning_rate": 9.040125378389404e-06, + "loss": 0.8169, + "step": 51175 + }, + { + "epoch": 3.29, + "learning_rate": 9.039910689366453e-06, + "loss": 0.8291, + "step": 51176 + }, + { + "epoch": 3.29, + "learning_rate": 9.039696000343502e-06, + "loss": 0.7987, + "step": 51177 + }, + { + "epoch": 3.29, + "learning_rate": 9.039481311320553e-06, + "loss": 0.8376, + "step": 51178 + }, + { + "epoch": 3.29, + "learning_rate": 9.039266622297602e-06, + "loss": 0.7973, + "step": 51179 + }, + { + "epoch": 3.29, + "learning_rate": 9.039051933274653e-06, + "loss": 0.7988, + "step": 51180 + }, + { + "epoch": 3.29, + "learning_rate": 9.038837244251703e-06, + "loss": 0.8834, + "step": 51181 + }, + { + "epoch": 3.29, + "learning_rate": 9.038622555228752e-06, + "loss": 0.7792, + "step": 51182 + }, + { + "epoch": 3.29, + "learning_rate": 9.038407866205801e-06, + "loss": 0.788, + "step": 51183 + }, + { + "epoch": 3.29, + "learning_rate": 9.038193177182852e-06, + "loss": 0.8161, + "step": 51184 + }, + { + "epoch": 3.29, + "learning_rate": 9.0379784881599e-06, + "loss": 0.8119, + "step": 51185 + }, + { + "epoch": 3.29, + "learning_rate": 9.03776379913695e-06, + "loss": 0.8317, + "step": 51186 + }, + { + "epoch": 3.29, + "learning_rate": 9.037549110114e-06, + "loss": 0.7594, + "step": 51187 + }, + { + "epoch": 3.29, + "learning_rate": 9.037334421091051e-06, + "loss": 0.7963, + "step": 51188 + }, + { + "epoch": 3.29, + "learning_rate": 9.0371197320681e-06, + "loss": 0.8617, + "step": 51189 + }, + { + "epoch": 3.29, + "learning_rate": 9.03690504304515e-06, + "loss": 0.8068, + "step": 51190 + }, + { + "epoch": 3.29, + "learning_rate": 9.0366903540222e-06, + "loss": 0.7575, + "step": 51191 + }, + { + "epoch": 3.29, + "learning_rate": 9.036475664999249e-06, + "loss": 0.8803, + "step": 51192 + }, + { + "epoch": 3.29, + "learning_rate": 9.0362609759763e-06, + "loss": 0.8026, + "step": 51193 + }, + { + "epoch": 3.29, + "learning_rate": 9.036046286953348e-06, + "loss": 0.8356, + "step": 51194 + }, + { + "epoch": 3.29, + "learning_rate": 9.035831597930399e-06, + "loss": 0.834, + "step": 51195 + }, + { + "epoch": 3.29, + "learning_rate": 9.035616908907448e-06, + "loss": 0.767, + "step": 51196 + }, + { + "epoch": 3.29, + "learning_rate": 9.035402219884499e-06, + "loss": 0.8337, + "step": 51197 + }, + { + "epoch": 3.29, + "learning_rate": 9.035187530861547e-06, + "loss": 0.7756, + "step": 51198 + }, + { + "epoch": 3.29, + "learning_rate": 9.034972841838598e-06, + "loss": 0.7931, + "step": 51199 + }, + { + "epoch": 3.29, + "learning_rate": 9.034758152815647e-06, + "loss": 0.8546, + "step": 51200 + }, + { + "epoch": 3.29, + "learning_rate": 9.034543463792696e-06, + "loss": 0.8227, + "step": 51201 + }, + { + "epoch": 3.29, + "learning_rate": 9.034328774769747e-06, + "loss": 0.855, + "step": 51202 + }, + { + "epoch": 3.29, + "learning_rate": 9.034114085746797e-06, + "loss": 0.863, + "step": 51203 + }, + { + "epoch": 3.29, + "learning_rate": 9.033899396723846e-06, + "loss": 0.8056, + "step": 51204 + }, + { + "epoch": 3.29, + "learning_rate": 9.033684707700895e-06, + "loss": 0.865, + "step": 51205 + }, + { + "epoch": 3.29, + "learning_rate": 9.033470018677946e-06, + "loss": 0.7278, + "step": 51206 + }, + { + "epoch": 3.29, + "learning_rate": 9.033255329654995e-06, + "loss": 0.8648, + "step": 51207 + }, + { + "epoch": 3.29, + "learning_rate": 9.033040640632046e-06, + "loss": 0.8093, + "step": 51208 + }, + { + "epoch": 3.29, + "learning_rate": 9.032825951609096e-06, + "loss": 0.902, + "step": 51209 + }, + { + "epoch": 3.29, + "learning_rate": 9.032611262586145e-06, + "loss": 0.8086, + "step": 51210 + }, + { + "epoch": 3.29, + "learning_rate": 9.032396573563194e-06, + "loss": 0.8485, + "step": 51211 + }, + { + "epoch": 3.29, + "learning_rate": 9.032181884540245e-06, + "loss": 0.8448, + "step": 51212 + }, + { + "epoch": 3.29, + "learning_rate": 9.031967195517294e-06, + "loss": 0.8584, + "step": 51213 + }, + { + "epoch": 3.29, + "learning_rate": 9.031752506494343e-06, + "loss": 0.8185, + "step": 51214 + }, + { + "epoch": 3.3, + "learning_rate": 9.031537817471393e-06, + "loss": 0.8415, + "step": 51215 + }, + { + "epoch": 3.3, + "learning_rate": 9.031323128448444e-06, + "loss": 0.7503, + "step": 51216 + }, + { + "epoch": 3.3, + "learning_rate": 9.031108439425493e-06, + "loss": 0.7737, + "step": 51217 + }, + { + "epoch": 3.3, + "learning_rate": 9.030893750402544e-06, + "loss": 0.8035, + "step": 51218 + }, + { + "epoch": 3.3, + "learning_rate": 9.030679061379593e-06, + "loss": 0.8696, + "step": 51219 + }, + { + "epoch": 3.3, + "learning_rate": 9.030464372356642e-06, + "loss": 0.7931, + "step": 51220 + }, + { + "epoch": 3.3, + "learning_rate": 9.030249683333692e-06, + "loss": 0.8196, + "step": 51221 + }, + { + "epoch": 3.3, + "learning_rate": 9.030034994310741e-06, + "loss": 0.8401, + "step": 51222 + }, + { + "epoch": 3.3, + "learning_rate": 9.02982030528779e-06, + "loss": 0.7949, + "step": 51223 + }, + { + "epoch": 3.3, + "learning_rate": 9.029605616264841e-06, + "loss": 0.8734, + "step": 51224 + }, + { + "epoch": 3.3, + "learning_rate": 9.029390927241892e-06, + "loss": 0.7827, + "step": 51225 + }, + { + "epoch": 3.3, + "learning_rate": 9.02917623821894e-06, + "loss": 0.849, + "step": 51226 + }, + { + "epoch": 3.3, + "learning_rate": 9.02896154919599e-06, + "loss": 0.8062, + "step": 51227 + }, + { + "epoch": 3.3, + "learning_rate": 9.02874686017304e-06, + "loss": 0.7784, + "step": 51228 + }, + { + "epoch": 3.3, + "learning_rate": 9.028532171150089e-06, + "loss": 0.8158, + "step": 51229 + }, + { + "epoch": 3.3, + "learning_rate": 9.02831748212714e-06, + "loss": 0.7763, + "step": 51230 + }, + { + "epoch": 3.3, + "learning_rate": 9.02810279310419e-06, + "loss": 0.8306, + "step": 51231 + }, + { + "epoch": 3.3, + "learning_rate": 9.02788810408124e-06, + "loss": 0.8398, + "step": 51232 + }, + { + "epoch": 3.3, + "learning_rate": 9.027673415058288e-06, + "loss": 0.7951, + "step": 51233 + }, + { + "epoch": 3.3, + "learning_rate": 9.027458726035339e-06, + "loss": 0.7941, + "step": 51234 + }, + { + "epoch": 3.3, + "learning_rate": 9.027244037012388e-06, + "loss": 0.8612, + "step": 51235 + }, + { + "epoch": 3.3, + "learning_rate": 9.027029347989437e-06, + "loss": 0.8152, + "step": 51236 + }, + { + "epoch": 3.3, + "learning_rate": 9.026814658966488e-06, + "loss": 0.8068, + "step": 51237 + }, + { + "epoch": 3.3, + "learning_rate": 9.026599969943538e-06, + "loss": 0.7589, + "step": 51238 + }, + { + "epoch": 3.3, + "learning_rate": 9.026385280920587e-06, + "loss": 0.7871, + "step": 51239 + }, + { + "epoch": 3.3, + "learning_rate": 9.026170591897638e-06, + "loss": 0.8346, + "step": 51240 + }, + { + "epoch": 3.3, + "learning_rate": 9.025955902874687e-06, + "loss": 0.795, + "step": 51241 + }, + { + "epoch": 3.3, + "learning_rate": 9.025741213851736e-06, + "loss": 0.7641, + "step": 51242 + }, + { + "epoch": 3.3, + "learning_rate": 9.025526524828786e-06, + "loss": 0.8349, + "step": 51243 + }, + { + "epoch": 3.3, + "learning_rate": 9.025311835805835e-06, + "loss": 0.8375, + "step": 51244 + }, + { + "epoch": 3.3, + "learning_rate": 9.025097146782884e-06, + "loss": 0.8194, + "step": 51245 + }, + { + "epoch": 3.3, + "learning_rate": 9.024882457759935e-06, + "loss": 0.8189, + "step": 51246 + }, + { + "epoch": 3.3, + "learning_rate": 9.024667768736986e-06, + "loss": 0.764, + "step": 51247 + }, + { + "epoch": 3.3, + "learning_rate": 9.024453079714035e-06, + "loss": 0.8534, + "step": 51248 + }, + { + "epoch": 3.3, + "learning_rate": 9.024238390691085e-06, + "loss": 0.8249, + "step": 51249 + }, + { + "epoch": 3.3, + "learning_rate": 9.024023701668134e-06, + "loss": 0.7579, + "step": 51250 + }, + { + "epoch": 3.3, + "learning_rate": 9.023809012645183e-06, + "loss": 0.8099, + "step": 51251 + }, + { + "epoch": 3.3, + "learning_rate": 9.023594323622234e-06, + "loss": 0.8159, + "step": 51252 + }, + { + "epoch": 3.3, + "learning_rate": 9.023379634599285e-06, + "loss": 0.8542, + "step": 51253 + }, + { + "epoch": 3.3, + "learning_rate": 9.023164945576334e-06, + "loss": 0.8282, + "step": 51254 + }, + { + "epoch": 3.3, + "learning_rate": 9.022950256553383e-06, + "loss": 0.8372, + "step": 51255 + }, + { + "epoch": 3.3, + "learning_rate": 9.022735567530433e-06, + "loss": 0.8896, + "step": 51256 + }, + { + "epoch": 3.3, + "learning_rate": 9.022520878507482e-06, + "loss": 0.8455, + "step": 51257 + }, + { + "epoch": 3.3, + "learning_rate": 9.022306189484533e-06, + "loss": 0.8735, + "step": 51258 + }, + { + "epoch": 3.3, + "learning_rate": 9.022091500461582e-06, + "loss": 0.7755, + "step": 51259 + }, + { + "epoch": 3.3, + "learning_rate": 9.021876811438632e-06, + "loss": 0.8099, + "step": 51260 + }, + { + "epoch": 3.3, + "learning_rate": 9.021662122415681e-06, + "loss": 0.7706, + "step": 51261 + }, + { + "epoch": 3.3, + "learning_rate": 9.021447433392732e-06, + "loss": 0.735, + "step": 51262 + }, + { + "epoch": 3.3, + "learning_rate": 9.021232744369781e-06, + "loss": 0.8321, + "step": 51263 + }, + { + "epoch": 3.3, + "learning_rate": 9.02101805534683e-06, + "loss": 0.8865, + "step": 51264 + }, + { + "epoch": 3.3, + "learning_rate": 9.02080336632388e-06, + "loss": 0.82, + "step": 51265 + }, + { + "epoch": 3.3, + "learning_rate": 9.02058867730093e-06, + "loss": 0.8195, + "step": 51266 + }, + { + "epoch": 3.3, + "learning_rate": 9.02037398827798e-06, + "loss": 0.8267, + "step": 51267 + }, + { + "epoch": 3.3, + "learning_rate": 9.020159299255031e-06, + "loss": 0.7967, + "step": 51268 + }, + { + "epoch": 3.3, + "learning_rate": 9.01994461023208e-06, + "loss": 0.817, + "step": 51269 + }, + { + "epoch": 3.3, + "learning_rate": 9.019729921209129e-06, + "loss": 0.8775, + "step": 51270 + }, + { + "epoch": 3.3, + "learning_rate": 9.01951523218618e-06, + "loss": 0.8176, + "step": 51271 + }, + { + "epoch": 3.3, + "learning_rate": 9.019300543163228e-06, + "loss": 0.9037, + "step": 51272 + }, + { + "epoch": 3.3, + "learning_rate": 9.019085854140277e-06, + "loss": 0.8633, + "step": 51273 + }, + { + "epoch": 3.3, + "learning_rate": 9.018871165117328e-06, + "loss": 0.861, + "step": 51274 + }, + { + "epoch": 3.3, + "learning_rate": 9.018656476094379e-06, + "loss": 0.8267, + "step": 51275 + }, + { + "epoch": 3.3, + "learning_rate": 9.018441787071428e-06, + "loss": 0.8354, + "step": 51276 + }, + { + "epoch": 3.3, + "learning_rate": 9.018227098048478e-06, + "loss": 0.8881, + "step": 51277 + }, + { + "epoch": 3.3, + "learning_rate": 9.018012409025527e-06, + "loss": 0.8752, + "step": 51278 + }, + { + "epoch": 3.3, + "learning_rate": 9.017797720002576e-06, + "loss": 0.735, + "step": 51279 + }, + { + "epoch": 3.3, + "learning_rate": 9.017583030979627e-06, + "loss": 0.8527, + "step": 51280 + }, + { + "epoch": 3.3, + "learning_rate": 9.017368341956676e-06, + "loss": 0.8793, + "step": 51281 + }, + { + "epoch": 3.3, + "learning_rate": 9.017153652933727e-06, + "loss": 0.7741, + "step": 51282 + }, + { + "epoch": 3.3, + "learning_rate": 9.016938963910776e-06, + "loss": 0.8111, + "step": 51283 + }, + { + "epoch": 3.3, + "learning_rate": 9.016724274887826e-06, + "loss": 0.8623, + "step": 51284 + }, + { + "epoch": 3.3, + "learning_rate": 9.016509585864875e-06, + "loss": 0.8235, + "step": 51285 + }, + { + "epoch": 3.3, + "learning_rate": 9.016294896841926e-06, + "loss": 0.8584, + "step": 51286 + }, + { + "epoch": 3.3, + "learning_rate": 9.016080207818975e-06, + "loss": 0.8113, + "step": 51287 + }, + { + "epoch": 3.3, + "learning_rate": 9.015865518796024e-06, + "loss": 0.8006, + "step": 51288 + }, + { + "epoch": 3.3, + "learning_rate": 9.015650829773074e-06, + "loss": 0.848, + "step": 51289 + }, + { + "epoch": 3.3, + "learning_rate": 9.015436140750125e-06, + "loss": 0.7774, + "step": 51290 + }, + { + "epoch": 3.3, + "learning_rate": 9.015221451727174e-06, + "loss": 0.8064, + "step": 51291 + }, + { + "epoch": 3.3, + "learning_rate": 9.015006762704223e-06, + "loss": 0.775, + "step": 51292 + }, + { + "epoch": 3.3, + "learning_rate": 9.014792073681274e-06, + "loss": 0.8265, + "step": 51293 + }, + { + "epoch": 3.3, + "learning_rate": 9.014577384658323e-06, + "loss": 0.8199, + "step": 51294 + }, + { + "epoch": 3.3, + "learning_rate": 9.014362695635373e-06, + "loss": 0.7883, + "step": 51295 + }, + { + "epoch": 3.3, + "learning_rate": 9.014148006612424e-06, + "loss": 0.7793, + "step": 51296 + }, + { + "epoch": 3.3, + "learning_rate": 9.013933317589473e-06, + "loss": 0.8356, + "step": 51297 + }, + { + "epoch": 3.3, + "learning_rate": 9.013718628566522e-06, + "loss": 0.8286, + "step": 51298 + }, + { + "epoch": 3.3, + "learning_rate": 9.013503939543573e-06, + "loss": 0.7562, + "step": 51299 + }, + { + "epoch": 3.3, + "learning_rate": 9.013289250520622e-06, + "loss": 0.8261, + "step": 51300 + }, + { + "epoch": 3.3, + "learning_rate": 9.01307456149767e-06, + "loss": 0.8866, + "step": 51301 + }, + { + "epoch": 3.3, + "learning_rate": 9.012859872474721e-06, + "loss": 0.7997, + "step": 51302 + }, + { + "epoch": 3.3, + "learning_rate": 9.01264518345177e-06, + "loss": 0.9035, + "step": 51303 + }, + { + "epoch": 3.3, + "learning_rate": 9.01243049442882e-06, + "loss": 0.7922, + "step": 51304 + }, + { + "epoch": 3.3, + "learning_rate": 9.012215805405871e-06, + "loss": 0.8824, + "step": 51305 + }, + { + "epoch": 3.3, + "learning_rate": 9.01200111638292e-06, + "loss": 0.8579, + "step": 51306 + }, + { + "epoch": 3.3, + "learning_rate": 9.01178642735997e-06, + "loss": 0.8717, + "step": 51307 + }, + { + "epoch": 3.3, + "learning_rate": 9.01157173833702e-06, + "loss": 0.8051, + "step": 51308 + }, + { + "epoch": 3.3, + "learning_rate": 9.011357049314069e-06, + "loss": 0.8458, + "step": 51309 + }, + { + "epoch": 3.3, + "learning_rate": 9.011142360291118e-06, + "loss": 0.7983, + "step": 51310 + }, + { + "epoch": 3.3, + "learning_rate": 9.010927671268169e-06, + "loss": 0.8435, + "step": 51311 + }, + { + "epoch": 3.3, + "learning_rate": 9.01071298224522e-06, + "loss": 0.8519, + "step": 51312 + }, + { + "epoch": 3.3, + "learning_rate": 9.010498293222268e-06, + "loss": 0.8001, + "step": 51313 + }, + { + "epoch": 3.3, + "learning_rate": 9.010283604199317e-06, + "loss": 0.7808, + "step": 51314 + }, + { + "epoch": 3.3, + "learning_rate": 9.010068915176368e-06, + "loss": 0.8418, + "step": 51315 + }, + { + "epoch": 3.3, + "learning_rate": 9.009854226153417e-06, + "loss": 0.8611, + "step": 51316 + }, + { + "epoch": 3.3, + "learning_rate": 9.009639537130467e-06, + "loss": 0.8082, + "step": 51317 + }, + { + "epoch": 3.3, + "learning_rate": 9.009424848107518e-06, + "loss": 0.8196, + "step": 51318 + }, + { + "epoch": 3.3, + "learning_rate": 9.009210159084567e-06, + "loss": 0.8056, + "step": 51319 + }, + { + "epoch": 3.3, + "learning_rate": 9.008995470061616e-06, + "loss": 0.7972, + "step": 51320 + }, + { + "epoch": 3.3, + "learning_rate": 9.008780781038667e-06, + "loss": 0.8076, + "step": 51321 + }, + { + "epoch": 3.3, + "learning_rate": 9.008566092015716e-06, + "loss": 0.7935, + "step": 51322 + }, + { + "epoch": 3.3, + "learning_rate": 9.008351402992765e-06, + "loss": 0.8205, + "step": 51323 + }, + { + "epoch": 3.3, + "learning_rate": 9.008136713969815e-06, + "loss": 0.8459, + "step": 51324 + }, + { + "epoch": 3.3, + "learning_rate": 9.007922024946866e-06, + "loss": 0.8323, + "step": 51325 + }, + { + "epoch": 3.3, + "learning_rate": 9.007707335923915e-06, + "loss": 0.8097, + "step": 51326 + }, + { + "epoch": 3.3, + "learning_rate": 9.007492646900966e-06, + "loss": 0.7801, + "step": 51327 + }, + { + "epoch": 3.3, + "learning_rate": 9.007277957878015e-06, + "loss": 0.8474, + "step": 51328 + }, + { + "epoch": 3.3, + "learning_rate": 9.007063268855064e-06, + "loss": 0.8895, + "step": 51329 + }, + { + "epoch": 3.3, + "learning_rate": 9.006848579832114e-06, + "loss": 0.8411, + "step": 51330 + }, + { + "epoch": 3.3, + "learning_rate": 9.006633890809163e-06, + "loss": 0.8172, + "step": 51331 + }, + { + "epoch": 3.3, + "learning_rate": 9.006419201786212e-06, + "loss": 0.8702, + "step": 51332 + }, + { + "epoch": 3.3, + "learning_rate": 9.006204512763263e-06, + "loss": 0.8185, + "step": 51333 + }, + { + "epoch": 3.3, + "learning_rate": 9.005989823740313e-06, + "loss": 0.8151, + "step": 51334 + }, + { + "epoch": 3.3, + "learning_rate": 9.005775134717362e-06, + "loss": 0.818, + "step": 51335 + }, + { + "epoch": 3.3, + "learning_rate": 9.005560445694413e-06, + "loss": 0.7881, + "step": 51336 + }, + { + "epoch": 3.3, + "learning_rate": 9.005345756671462e-06, + "loss": 0.8864, + "step": 51337 + }, + { + "epoch": 3.3, + "learning_rate": 9.005131067648511e-06, + "loss": 0.9373, + "step": 51338 + }, + { + "epoch": 3.3, + "learning_rate": 9.004916378625562e-06, + "loss": 0.8583, + "step": 51339 + }, + { + "epoch": 3.3, + "learning_rate": 9.004701689602612e-06, + "loss": 0.7877, + "step": 51340 + }, + { + "epoch": 3.3, + "learning_rate": 9.004487000579661e-06, + "loss": 0.8574, + "step": 51341 + }, + { + "epoch": 3.3, + "learning_rate": 9.00427231155671e-06, + "loss": 0.8167, + "step": 51342 + }, + { + "epoch": 3.3, + "learning_rate": 9.004057622533761e-06, + "loss": 0.8734, + "step": 51343 + }, + { + "epoch": 3.3, + "learning_rate": 9.00384293351081e-06, + "loss": 0.8364, + "step": 51344 + }, + { + "epoch": 3.3, + "learning_rate": 9.00362824448786e-06, + "loss": 0.7763, + "step": 51345 + }, + { + "epoch": 3.3, + "learning_rate": 9.00341355546491e-06, + "loss": 0.8105, + "step": 51346 + }, + { + "epoch": 3.3, + "learning_rate": 9.00319886644196e-06, + "loss": 0.7507, + "step": 51347 + }, + { + "epoch": 3.3, + "learning_rate": 9.002984177419009e-06, + "loss": 0.8918, + "step": 51348 + }, + { + "epoch": 3.3, + "learning_rate": 9.00276948839606e-06, + "loss": 0.8221, + "step": 51349 + }, + { + "epoch": 3.3, + "learning_rate": 9.002554799373109e-06, + "loss": 0.8619, + "step": 51350 + }, + { + "epoch": 3.3, + "learning_rate": 9.002340110350158e-06, + "loss": 0.7879, + "step": 51351 + }, + { + "epoch": 3.3, + "learning_rate": 9.002125421327208e-06, + "loss": 0.8389, + "step": 51352 + }, + { + "epoch": 3.3, + "learning_rate": 9.001910732304257e-06, + "loss": 0.8257, + "step": 51353 + }, + { + "epoch": 3.3, + "learning_rate": 9.001696043281308e-06, + "loss": 0.7796, + "step": 51354 + }, + { + "epoch": 3.3, + "learning_rate": 9.001481354258359e-06, + "loss": 0.8397, + "step": 51355 + }, + { + "epoch": 3.3, + "learning_rate": 9.001266665235408e-06, + "loss": 0.8266, + "step": 51356 + }, + { + "epoch": 3.3, + "learning_rate": 9.001051976212457e-06, + "loss": 0.8342, + "step": 51357 + }, + { + "epoch": 3.3, + "learning_rate": 9.000837287189507e-06, + "loss": 0.7819, + "step": 51358 + }, + { + "epoch": 3.3, + "learning_rate": 9.000622598166556e-06, + "loss": 0.8125, + "step": 51359 + }, + { + "epoch": 3.3, + "learning_rate": 9.000407909143605e-06, + "loss": 0.7933, + "step": 51360 + }, + { + "epoch": 3.3, + "learning_rate": 9.000193220120656e-06, + "loss": 0.7554, + "step": 51361 + }, + { + "epoch": 3.3, + "learning_rate": 8.999978531097706e-06, + "loss": 0.7614, + "step": 51362 + }, + { + "epoch": 3.3, + "learning_rate": 8.999763842074755e-06, + "loss": 0.869, + "step": 51363 + }, + { + "epoch": 3.3, + "learning_rate": 8.999549153051806e-06, + "loss": 0.8707, + "step": 51364 + }, + { + "epoch": 3.3, + "learning_rate": 8.999334464028855e-06, + "loss": 0.8623, + "step": 51365 + }, + { + "epoch": 3.3, + "learning_rate": 8.999119775005904e-06, + "loss": 0.889, + "step": 51366 + }, + { + "epoch": 3.3, + "learning_rate": 8.998905085982955e-06, + "loss": 0.8611, + "step": 51367 + }, + { + "epoch": 3.3, + "learning_rate": 8.998690396960004e-06, + "loss": 0.8806, + "step": 51368 + }, + { + "epoch": 3.3, + "learning_rate": 8.998475707937054e-06, + "loss": 0.8584, + "step": 51369 + }, + { + "epoch": 3.31, + "learning_rate": 8.998261018914103e-06, + "loss": 0.8221, + "step": 51370 + }, + { + "epoch": 3.31, + "learning_rate": 8.998046329891154e-06, + "loss": 0.7489, + "step": 51371 + }, + { + "epoch": 3.31, + "learning_rate": 8.997831640868203e-06, + "loss": 0.8838, + "step": 51372 + }, + { + "epoch": 3.31, + "learning_rate": 8.997616951845254e-06, + "loss": 0.7959, + "step": 51373 + }, + { + "epoch": 3.31, + "learning_rate": 8.997402262822303e-06, + "loss": 0.8881, + "step": 51374 + }, + { + "epoch": 3.31, + "learning_rate": 8.997187573799351e-06, + "loss": 0.821, + "step": 51375 + }, + { + "epoch": 3.31, + "learning_rate": 8.996972884776402e-06, + "loss": 0.8366, + "step": 51376 + }, + { + "epoch": 3.31, + "learning_rate": 8.996758195753453e-06, + "loss": 0.7936, + "step": 51377 + }, + { + "epoch": 3.31, + "learning_rate": 8.996543506730502e-06, + "loss": 0.8188, + "step": 51378 + }, + { + "epoch": 3.31, + "learning_rate": 8.99632881770755e-06, + "loss": 0.8221, + "step": 51379 + }, + { + "epoch": 3.31, + "learning_rate": 8.996114128684601e-06, + "loss": 0.7929, + "step": 51380 + }, + { + "epoch": 3.31, + "learning_rate": 8.99589943966165e-06, + "loss": 0.847, + "step": 51381 + }, + { + "epoch": 3.31, + "learning_rate": 8.9956847506387e-06, + "loss": 0.8126, + "step": 51382 + }, + { + "epoch": 3.31, + "learning_rate": 8.995470061615752e-06, + "loss": 0.8246, + "step": 51383 + }, + { + "epoch": 3.31, + "learning_rate": 8.9952553725928e-06, + "loss": 0.8144, + "step": 51384 + }, + { + "epoch": 3.31, + "learning_rate": 8.99504068356985e-06, + "loss": 0.8492, + "step": 51385 + }, + { + "epoch": 3.31, + "learning_rate": 8.9948259945469e-06, + "loss": 0.8393, + "step": 51386 + }, + { + "epoch": 3.31, + "learning_rate": 8.99461130552395e-06, + "loss": 0.8191, + "step": 51387 + }, + { + "epoch": 3.31, + "learning_rate": 8.994396616500998e-06, + "loss": 0.7987, + "step": 51388 + }, + { + "epoch": 3.31, + "learning_rate": 8.994181927478049e-06, + "loss": 0.8624, + "step": 51389 + }, + { + "epoch": 3.31, + "learning_rate": 8.993967238455098e-06, + "loss": 0.8214, + "step": 51390 + }, + { + "epoch": 3.31, + "learning_rate": 8.993752549432148e-06, + "loss": 0.8527, + "step": 51391 + }, + { + "epoch": 3.31, + "learning_rate": 8.993537860409197e-06, + "loss": 0.8439, + "step": 51392 + }, + { + "epoch": 3.31, + "learning_rate": 8.993323171386248e-06, + "loss": 0.8635, + "step": 51393 + }, + { + "epoch": 3.31, + "learning_rate": 8.993108482363297e-06, + "loss": 0.7654, + "step": 51394 + }, + { + "epoch": 3.31, + "learning_rate": 8.992893793340348e-06, + "loss": 0.7823, + "step": 51395 + }, + { + "epoch": 3.31, + "learning_rate": 8.992679104317397e-06, + "loss": 0.8074, + "step": 51396 + }, + { + "epoch": 3.31, + "learning_rate": 8.992464415294446e-06, + "loss": 0.7821, + "step": 51397 + }, + { + "epoch": 3.31, + "learning_rate": 8.992249726271496e-06, + "loss": 0.8195, + "step": 51398 + }, + { + "epoch": 3.31, + "learning_rate": 8.992035037248547e-06, + "loss": 0.8232, + "step": 51399 + }, + { + "epoch": 3.31, + "learning_rate": 8.991820348225596e-06, + "loss": 0.7832, + "step": 51400 + }, + { + "epoch": 3.31, + "learning_rate": 8.991605659202645e-06, + "loss": 0.8708, + "step": 51401 + }, + { + "epoch": 3.31, + "learning_rate": 8.991390970179696e-06, + "loss": 0.7811, + "step": 51402 + }, + { + "epoch": 3.31, + "learning_rate": 8.991176281156745e-06, + "loss": 0.7948, + "step": 51403 + }, + { + "epoch": 3.31, + "learning_rate": 8.990961592133795e-06, + "loss": 0.8096, + "step": 51404 + }, + { + "epoch": 3.31, + "learning_rate": 8.990746903110846e-06, + "loss": 0.845, + "step": 51405 + }, + { + "epoch": 3.31, + "learning_rate": 8.990532214087895e-06, + "loss": 0.8097, + "step": 51406 + }, + { + "epoch": 3.31, + "learning_rate": 8.990317525064944e-06, + "loss": 0.928, + "step": 51407 + }, + { + "epoch": 3.31, + "learning_rate": 8.990102836041994e-06, + "loss": 0.851, + "step": 51408 + }, + { + "epoch": 3.31, + "learning_rate": 8.989888147019043e-06, + "loss": 0.7552, + "step": 51409 + }, + { + "epoch": 3.31, + "learning_rate": 8.989673457996092e-06, + "loss": 0.8534, + "step": 51410 + }, + { + "epoch": 3.31, + "learning_rate": 8.989458768973143e-06, + "loss": 0.8555, + "step": 51411 + }, + { + "epoch": 3.31, + "learning_rate": 8.989244079950194e-06, + "loss": 0.8224, + "step": 51412 + }, + { + "epoch": 3.31, + "learning_rate": 8.989029390927243e-06, + "loss": 0.7993, + "step": 51413 + }, + { + "epoch": 3.31, + "learning_rate": 8.988814701904293e-06, + "loss": 0.8596, + "step": 51414 + }, + { + "epoch": 3.31, + "learning_rate": 8.988600012881342e-06, + "loss": 0.8322, + "step": 51415 + }, + { + "epoch": 3.31, + "learning_rate": 8.988385323858391e-06, + "loss": 0.8387, + "step": 51416 + }, + { + "epoch": 3.31, + "learning_rate": 8.988170634835442e-06, + "loss": 0.8181, + "step": 51417 + }, + { + "epoch": 3.31, + "learning_rate": 8.987955945812491e-06, + "loss": 0.8123, + "step": 51418 + }, + { + "epoch": 3.31, + "learning_rate": 8.98774125678954e-06, + "loss": 0.8373, + "step": 51419 + }, + { + "epoch": 3.31, + "learning_rate": 8.98752656776659e-06, + "loss": 0.7811, + "step": 51420 + }, + { + "epoch": 3.31, + "learning_rate": 8.987311878743641e-06, + "loss": 0.8451, + "step": 51421 + }, + { + "epoch": 3.31, + "learning_rate": 8.98709718972069e-06, + "loss": 0.8471, + "step": 51422 + }, + { + "epoch": 3.31, + "learning_rate": 8.98688250069774e-06, + "loss": 0.8348, + "step": 51423 + }, + { + "epoch": 3.31, + "learning_rate": 8.98666781167479e-06, + "loss": 0.8608, + "step": 51424 + }, + { + "epoch": 3.31, + "learning_rate": 8.986453122651839e-06, + "loss": 0.9148, + "step": 51425 + }, + { + "epoch": 3.31, + "learning_rate": 8.98623843362889e-06, + "loss": 0.7967, + "step": 51426 + }, + { + "epoch": 3.31, + "learning_rate": 8.98602374460594e-06, + "loss": 0.7919, + "step": 51427 + }, + { + "epoch": 3.31, + "learning_rate": 8.985809055582989e-06, + "loss": 0.8165, + "step": 51428 + }, + { + "epoch": 3.31, + "learning_rate": 8.985594366560038e-06, + "loss": 0.8252, + "step": 51429 + }, + { + "epoch": 3.31, + "learning_rate": 8.985379677537089e-06, + "loss": 0.8299, + "step": 51430 + }, + { + "epoch": 3.31, + "learning_rate": 8.985164988514138e-06, + "loss": 0.7937, + "step": 51431 + }, + { + "epoch": 3.31, + "learning_rate": 8.984950299491188e-06, + "loss": 0.8207, + "step": 51432 + }, + { + "epoch": 3.31, + "learning_rate": 8.984735610468237e-06, + "loss": 0.8502, + "step": 51433 + }, + { + "epoch": 3.31, + "learning_rate": 8.984520921445288e-06, + "loss": 0.871, + "step": 51434 + }, + { + "epoch": 3.31, + "learning_rate": 8.984306232422337e-06, + "loss": 0.8812, + "step": 51435 + }, + { + "epoch": 3.31, + "learning_rate": 8.984091543399387e-06, + "loss": 0.8214, + "step": 51436 + }, + { + "epoch": 3.31, + "learning_rate": 8.983876854376436e-06, + "loss": 0.8411, + "step": 51437 + }, + { + "epoch": 3.31, + "learning_rate": 8.983662165353485e-06, + "loss": 0.8574, + "step": 51438 + }, + { + "epoch": 3.31, + "learning_rate": 8.983447476330536e-06, + "loss": 0.8197, + "step": 51439 + }, + { + "epoch": 3.31, + "learning_rate": 8.983232787307585e-06, + "loss": 0.7995, + "step": 51440 + }, + { + "epoch": 3.31, + "learning_rate": 8.983018098284636e-06, + "loss": 0.8058, + "step": 51441 + }, + { + "epoch": 3.31, + "learning_rate": 8.982803409261686e-06, + "loss": 0.7905, + "step": 51442 + }, + { + "epoch": 3.31, + "learning_rate": 8.982588720238735e-06, + "loss": 0.8803, + "step": 51443 + }, + { + "epoch": 3.31, + "learning_rate": 8.982374031215784e-06, + "loss": 0.8509, + "step": 51444 + }, + { + "epoch": 3.31, + "learning_rate": 8.982159342192835e-06, + "loss": 0.8648, + "step": 51445 + }, + { + "epoch": 3.31, + "learning_rate": 8.981944653169884e-06, + "loss": 0.7625, + "step": 51446 + }, + { + "epoch": 3.31, + "learning_rate": 8.981729964146933e-06, + "loss": 0.852, + "step": 51447 + }, + { + "epoch": 3.31, + "learning_rate": 8.981515275123984e-06, + "loss": 0.7729, + "step": 51448 + }, + { + "epoch": 3.31, + "learning_rate": 8.981300586101034e-06, + "loss": 0.8386, + "step": 51449 + }, + { + "epoch": 3.31, + "learning_rate": 8.981085897078083e-06, + "loss": 0.8325, + "step": 51450 + }, + { + "epoch": 3.31, + "learning_rate": 8.980871208055134e-06, + "loss": 0.8396, + "step": 51451 + }, + { + "epoch": 3.31, + "learning_rate": 8.980656519032183e-06, + "loss": 0.8199, + "step": 51452 + }, + { + "epoch": 3.31, + "learning_rate": 8.980441830009232e-06, + "loss": 0.8612, + "step": 51453 + }, + { + "epoch": 3.31, + "learning_rate": 8.980227140986282e-06, + "loss": 0.8054, + "step": 51454 + }, + { + "epoch": 3.31, + "learning_rate": 8.980012451963331e-06, + "loss": 0.8022, + "step": 51455 + }, + { + "epoch": 3.31, + "learning_rate": 8.979797762940382e-06, + "loss": 0.8706, + "step": 51456 + }, + { + "epoch": 3.31, + "learning_rate": 8.979583073917431e-06, + "loss": 0.853, + "step": 51457 + }, + { + "epoch": 3.31, + "learning_rate": 8.979368384894482e-06, + "loss": 0.833, + "step": 51458 + }, + { + "epoch": 3.31, + "learning_rate": 8.97915369587153e-06, + "loss": 0.8448, + "step": 51459 + }, + { + "epoch": 3.31, + "learning_rate": 8.978939006848581e-06, + "loss": 0.8073, + "step": 51460 + }, + { + "epoch": 3.31, + "learning_rate": 8.97872431782563e-06, + "loss": 0.8173, + "step": 51461 + }, + { + "epoch": 3.31, + "learning_rate": 8.97850962880268e-06, + "loss": 0.8434, + "step": 51462 + }, + { + "epoch": 3.31, + "learning_rate": 8.97829493977973e-06, + "loss": 0.8351, + "step": 51463 + }, + { + "epoch": 3.31, + "learning_rate": 8.97808025075678e-06, + "loss": 0.8278, + "step": 51464 + }, + { + "epoch": 3.31, + "learning_rate": 8.97786556173383e-06, + "loss": 0.8124, + "step": 51465 + }, + { + "epoch": 3.31, + "learning_rate": 8.977650872710878e-06, + "loss": 0.8901, + "step": 51466 + }, + { + "epoch": 3.31, + "learning_rate": 8.977436183687929e-06, + "loss": 0.8726, + "step": 51467 + }, + { + "epoch": 3.31, + "learning_rate": 8.977221494664978e-06, + "loss": 0.7558, + "step": 51468 + }, + { + "epoch": 3.31, + "learning_rate": 8.977006805642027e-06, + "loss": 0.8197, + "step": 51469 + }, + { + "epoch": 3.31, + "learning_rate": 8.976792116619078e-06, + "loss": 0.8771, + "step": 51470 + }, + { + "epoch": 3.31, + "learning_rate": 8.976577427596128e-06, + "loss": 0.8334, + "step": 51471 + }, + { + "epoch": 3.31, + "learning_rate": 8.976362738573177e-06, + "loss": 0.8086, + "step": 51472 + }, + { + "epoch": 3.31, + "learning_rate": 8.976148049550228e-06, + "loss": 0.8633, + "step": 51473 + }, + { + "epoch": 3.31, + "learning_rate": 8.975933360527277e-06, + "loss": 0.8871, + "step": 51474 + }, + { + "epoch": 3.31, + "learning_rate": 8.975718671504326e-06, + "loss": 0.7522, + "step": 51475 + }, + { + "epoch": 3.31, + "learning_rate": 8.975503982481377e-06, + "loss": 0.8045, + "step": 51476 + }, + { + "epoch": 3.31, + "learning_rate": 8.975289293458426e-06, + "loss": 0.8574, + "step": 51477 + }, + { + "epoch": 3.31, + "learning_rate": 8.975074604435476e-06, + "loss": 0.8563, + "step": 51478 + }, + { + "epoch": 3.31, + "learning_rate": 8.974859915412525e-06, + "loss": 0.7949, + "step": 51479 + }, + { + "epoch": 3.31, + "learning_rate": 8.974645226389576e-06, + "loss": 0.8008, + "step": 51480 + }, + { + "epoch": 3.31, + "learning_rate": 8.974430537366625e-06, + "loss": 0.8258, + "step": 51481 + }, + { + "epoch": 3.31, + "learning_rate": 8.974215848343675e-06, + "loss": 0.7335, + "step": 51482 + }, + { + "epoch": 3.31, + "learning_rate": 8.974001159320724e-06, + "loss": 0.8369, + "step": 51483 + }, + { + "epoch": 3.31, + "learning_rate": 8.973786470297773e-06, + "loss": 0.8146, + "step": 51484 + }, + { + "epoch": 3.31, + "learning_rate": 8.973571781274824e-06, + "loss": 0.8615, + "step": 51485 + }, + { + "epoch": 3.31, + "learning_rate": 8.973357092251875e-06, + "loss": 0.7614, + "step": 51486 + }, + { + "epoch": 3.31, + "learning_rate": 8.973142403228924e-06, + "loss": 0.8105, + "step": 51487 + }, + { + "epoch": 3.31, + "learning_rate": 8.972927714205973e-06, + "loss": 0.8674, + "step": 51488 + }, + { + "epoch": 3.31, + "learning_rate": 8.972713025183023e-06, + "loss": 0.8046, + "step": 51489 + }, + { + "epoch": 3.31, + "learning_rate": 8.972498336160072e-06, + "loss": 0.8279, + "step": 51490 + }, + { + "epoch": 3.31, + "learning_rate": 8.972283647137123e-06, + "loss": 0.7892, + "step": 51491 + }, + { + "epoch": 3.31, + "learning_rate": 8.972068958114174e-06, + "loss": 0.8023, + "step": 51492 + }, + { + "epoch": 3.31, + "learning_rate": 8.971854269091223e-06, + "loss": 0.8584, + "step": 51493 + }, + { + "epoch": 3.31, + "learning_rate": 8.971639580068271e-06, + "loss": 0.8205, + "step": 51494 + }, + { + "epoch": 3.31, + "learning_rate": 8.971424891045322e-06, + "loss": 0.8313, + "step": 51495 + }, + { + "epoch": 3.31, + "learning_rate": 8.971210202022371e-06, + "loss": 0.8342, + "step": 51496 + }, + { + "epoch": 3.31, + "learning_rate": 8.97099551299942e-06, + "loss": 0.8026, + "step": 51497 + }, + { + "epoch": 3.31, + "learning_rate": 8.97078082397647e-06, + "loss": 0.8289, + "step": 51498 + }, + { + "epoch": 3.31, + "learning_rate": 8.97056613495352e-06, + "loss": 0.8223, + "step": 51499 + }, + { + "epoch": 3.31, + "learning_rate": 8.97035144593057e-06, + "loss": 0.7648, + "step": 51500 + }, + { + "epoch": 3.31, + "learning_rate": 8.970136756907621e-06, + "loss": 0.7988, + "step": 51501 + }, + { + "epoch": 3.31, + "learning_rate": 8.96992206788467e-06, + "loss": 0.7315, + "step": 51502 + }, + { + "epoch": 3.31, + "learning_rate": 8.969707378861719e-06, + "loss": 0.8542, + "step": 51503 + }, + { + "epoch": 3.31, + "learning_rate": 8.96949268983877e-06, + "loss": 0.8733, + "step": 51504 + }, + { + "epoch": 3.31, + "learning_rate": 8.969278000815819e-06, + "loss": 0.8263, + "step": 51505 + }, + { + "epoch": 3.31, + "learning_rate": 8.969063311792868e-06, + "loss": 0.8806, + "step": 51506 + }, + { + "epoch": 3.31, + "learning_rate": 8.968848622769918e-06, + "loss": 0.8424, + "step": 51507 + }, + { + "epoch": 3.31, + "learning_rate": 8.968633933746969e-06, + "loss": 0.908, + "step": 51508 + }, + { + "epoch": 3.31, + "learning_rate": 8.968419244724018e-06, + "loss": 0.8067, + "step": 51509 + }, + { + "epoch": 3.31, + "learning_rate": 8.968204555701068e-06, + "loss": 0.873, + "step": 51510 + }, + { + "epoch": 3.31, + "learning_rate": 8.967989866678117e-06, + "loss": 0.8461, + "step": 51511 + }, + { + "epoch": 3.31, + "learning_rate": 8.967775177655166e-06, + "loss": 0.892, + "step": 51512 + }, + { + "epoch": 3.31, + "learning_rate": 8.967560488632217e-06, + "loss": 0.7487, + "step": 51513 + }, + { + "epoch": 3.31, + "learning_rate": 8.967345799609268e-06, + "loss": 0.7953, + "step": 51514 + }, + { + "epoch": 3.31, + "learning_rate": 8.967131110586317e-06, + "loss": 0.8606, + "step": 51515 + }, + { + "epoch": 3.31, + "learning_rate": 8.966916421563366e-06, + "loss": 0.8783, + "step": 51516 + }, + { + "epoch": 3.31, + "learning_rate": 8.966701732540416e-06, + "loss": 0.7713, + "step": 51517 + }, + { + "epoch": 3.31, + "learning_rate": 8.966487043517465e-06, + "loss": 0.794, + "step": 51518 + }, + { + "epoch": 3.31, + "learning_rate": 8.966272354494516e-06, + "loss": 0.8266, + "step": 51519 + }, + { + "epoch": 3.31, + "learning_rate": 8.966057665471565e-06, + "loss": 0.8259, + "step": 51520 + }, + { + "epoch": 3.31, + "learning_rate": 8.965842976448616e-06, + "loss": 0.8865, + "step": 51521 + }, + { + "epoch": 3.31, + "learning_rate": 8.965628287425665e-06, + "loss": 0.7987, + "step": 51522 + }, + { + "epoch": 3.31, + "learning_rate": 8.965413598402715e-06, + "loss": 0.8178, + "step": 51523 + }, + { + "epoch": 3.31, + "learning_rate": 8.965198909379764e-06, + "loss": 0.8154, + "step": 51524 + }, + { + "epoch": 3.31, + "learning_rate": 8.964984220356813e-06, + "loss": 0.8424, + "step": 51525 + }, + { + "epoch": 3.32, + "learning_rate": 8.964769531333864e-06, + "loss": 0.8195, + "step": 51526 + }, + { + "epoch": 3.32, + "learning_rate": 8.964554842310913e-06, + "loss": 0.8748, + "step": 51527 + }, + { + "epoch": 3.32, + "learning_rate": 8.964340153287963e-06, + "loss": 0.847, + "step": 51528 + }, + { + "epoch": 3.32, + "learning_rate": 8.964125464265014e-06, + "loss": 0.8873, + "step": 51529 + }, + { + "epoch": 3.32, + "learning_rate": 8.963910775242063e-06, + "loss": 0.7748, + "step": 51530 + }, + { + "epoch": 3.32, + "learning_rate": 8.963696086219112e-06, + "loss": 0.8316, + "step": 51531 + }, + { + "epoch": 3.32, + "learning_rate": 8.963481397196163e-06, + "loss": 0.7926, + "step": 51532 + }, + { + "epoch": 3.32, + "learning_rate": 8.963266708173212e-06, + "loss": 0.7724, + "step": 51533 + }, + { + "epoch": 3.32, + "learning_rate": 8.96305201915026e-06, + "loss": 0.7941, + "step": 51534 + }, + { + "epoch": 3.32, + "learning_rate": 8.962837330127311e-06, + "loss": 0.8463, + "step": 51535 + }, + { + "epoch": 3.32, + "learning_rate": 8.962622641104362e-06, + "loss": 0.7843, + "step": 51536 + }, + { + "epoch": 3.32, + "learning_rate": 8.962407952081411e-06, + "loss": 0.9264, + "step": 51537 + }, + { + "epoch": 3.32, + "learning_rate": 8.962193263058462e-06, + "loss": 0.8037, + "step": 51538 + }, + { + "epoch": 3.32, + "learning_rate": 8.96197857403551e-06, + "loss": 0.8796, + "step": 51539 + }, + { + "epoch": 3.32, + "learning_rate": 8.96176388501256e-06, + "loss": 0.7851, + "step": 51540 + }, + { + "epoch": 3.32, + "learning_rate": 8.96154919598961e-06, + "loss": 0.8311, + "step": 51541 + }, + { + "epoch": 3.32, + "learning_rate": 8.961334506966659e-06, + "loss": 0.8496, + "step": 51542 + }, + { + "epoch": 3.32, + "learning_rate": 8.96111981794371e-06, + "loss": 0.8163, + "step": 51543 + }, + { + "epoch": 3.32, + "learning_rate": 8.960905128920759e-06, + "loss": 0.8716, + "step": 51544 + }, + { + "epoch": 3.32, + "learning_rate": 8.96069043989781e-06, + "loss": 0.7838, + "step": 51545 + }, + { + "epoch": 3.32, + "learning_rate": 8.960475750874858e-06, + "loss": 0.8358, + "step": 51546 + }, + { + "epoch": 3.32, + "learning_rate": 8.960261061851907e-06, + "loss": 0.7742, + "step": 51547 + }, + { + "epoch": 3.32, + "learning_rate": 8.960046372828958e-06, + "loss": 0.7732, + "step": 51548 + }, + { + "epoch": 3.32, + "learning_rate": 8.959831683806007e-06, + "loss": 0.821, + "step": 51549 + }, + { + "epoch": 3.32, + "learning_rate": 8.959616994783058e-06, + "loss": 0.8334, + "step": 51550 + }, + { + "epoch": 3.32, + "learning_rate": 8.959402305760108e-06, + "loss": 0.8195, + "step": 51551 + }, + { + "epoch": 3.32, + "learning_rate": 8.959187616737157e-06, + "loss": 0.8316, + "step": 51552 + }, + { + "epoch": 3.32, + "learning_rate": 8.958972927714206e-06, + "loss": 0.8666, + "step": 51553 + }, + { + "epoch": 3.32, + "learning_rate": 8.958758238691257e-06, + "loss": 0.773, + "step": 51554 + }, + { + "epoch": 3.32, + "learning_rate": 8.958543549668306e-06, + "loss": 0.7957, + "step": 51555 + }, + { + "epoch": 3.32, + "learning_rate": 8.958328860645355e-06, + "loss": 0.8335, + "step": 51556 + }, + { + "epoch": 3.32, + "learning_rate": 8.958114171622405e-06, + "loss": 0.829, + "step": 51557 + }, + { + "epoch": 3.32, + "learning_rate": 8.957899482599456e-06, + "loss": 0.7973, + "step": 51558 + }, + { + "epoch": 3.32, + "learning_rate": 8.957684793576505e-06, + "loss": 0.8324, + "step": 51559 + }, + { + "epoch": 3.32, + "learning_rate": 8.957470104553556e-06, + "loss": 0.8141, + "step": 51560 + }, + { + "epoch": 3.32, + "learning_rate": 8.957255415530605e-06, + "loss": 0.8144, + "step": 51561 + }, + { + "epoch": 3.32, + "learning_rate": 8.957040726507654e-06, + "loss": 0.8798, + "step": 51562 + }, + { + "epoch": 3.32, + "learning_rate": 8.956826037484704e-06, + "loss": 0.8575, + "step": 51563 + }, + { + "epoch": 3.32, + "learning_rate": 8.956611348461753e-06, + "loss": 0.8088, + "step": 51564 + }, + { + "epoch": 3.32, + "learning_rate": 8.956396659438804e-06, + "loss": 0.8298, + "step": 51565 + }, + { + "epoch": 3.32, + "learning_rate": 8.956181970415853e-06, + "loss": 0.7176, + "step": 51566 + }, + { + "epoch": 3.32, + "learning_rate": 8.955967281392904e-06, + "loss": 0.7591, + "step": 51567 + }, + { + "epoch": 3.32, + "learning_rate": 8.955752592369952e-06, + "loss": 0.8376, + "step": 51568 + }, + { + "epoch": 3.32, + "learning_rate": 8.955537903347003e-06, + "loss": 0.799, + "step": 51569 + }, + { + "epoch": 3.32, + "learning_rate": 8.955323214324052e-06, + "loss": 0.7734, + "step": 51570 + }, + { + "epoch": 3.32, + "learning_rate": 8.955108525301101e-06, + "loss": 0.8607, + "step": 51571 + }, + { + "epoch": 3.32, + "learning_rate": 8.954893836278152e-06, + "loss": 0.8021, + "step": 51572 + }, + { + "epoch": 3.32, + "learning_rate": 8.954679147255202e-06, + "loss": 0.8495, + "step": 51573 + }, + { + "epoch": 3.32, + "learning_rate": 8.954464458232251e-06, + "loss": 0.8272, + "step": 51574 + }, + { + "epoch": 3.32, + "learning_rate": 8.9542497692093e-06, + "loss": 0.7935, + "step": 51575 + }, + { + "epoch": 3.32, + "learning_rate": 8.954035080186351e-06, + "loss": 0.8627, + "step": 51576 + }, + { + "epoch": 3.32, + "learning_rate": 8.9538203911634e-06, + "loss": 0.8341, + "step": 51577 + }, + { + "epoch": 3.32, + "learning_rate": 8.95360570214045e-06, + "loss": 0.8007, + "step": 51578 + }, + { + "epoch": 3.32, + "learning_rate": 8.953391013117501e-06, + "loss": 0.8476, + "step": 51579 + }, + { + "epoch": 3.32, + "learning_rate": 8.95317632409455e-06, + "loss": 0.8262, + "step": 51580 + }, + { + "epoch": 3.32, + "learning_rate": 8.9529616350716e-06, + "loss": 0.7848, + "step": 51581 + }, + { + "epoch": 3.32, + "learning_rate": 8.95274694604865e-06, + "loss": 0.826, + "step": 51582 + }, + { + "epoch": 3.32, + "learning_rate": 8.952532257025699e-06, + "loss": 0.7911, + "step": 51583 + }, + { + "epoch": 3.32, + "learning_rate": 8.952317568002748e-06, + "loss": 0.8095, + "step": 51584 + }, + { + "epoch": 3.32, + "learning_rate": 8.952102878979798e-06, + "loss": 0.8025, + "step": 51585 + }, + { + "epoch": 3.32, + "learning_rate": 8.951888189956847e-06, + "loss": 0.8128, + "step": 51586 + }, + { + "epoch": 3.32, + "learning_rate": 8.951673500933898e-06, + "loss": 0.7838, + "step": 51587 + }, + { + "epoch": 3.32, + "learning_rate": 8.951458811910949e-06, + "loss": 0.8212, + "step": 51588 + }, + { + "epoch": 3.32, + "learning_rate": 8.951244122887998e-06, + "loss": 0.855, + "step": 51589 + }, + { + "epoch": 3.32, + "learning_rate": 8.951029433865047e-06, + "loss": 0.8043, + "step": 51590 + }, + { + "epoch": 3.32, + "learning_rate": 8.950814744842097e-06, + "loss": 0.8635, + "step": 51591 + }, + { + "epoch": 3.32, + "learning_rate": 8.950600055819146e-06, + "loss": 0.856, + "step": 51592 + }, + { + "epoch": 3.32, + "learning_rate": 8.950385366796195e-06, + "loss": 0.762, + "step": 51593 + }, + { + "epoch": 3.32, + "learning_rate": 8.950170677773246e-06, + "loss": 0.7977, + "step": 51594 + }, + { + "epoch": 3.32, + "learning_rate": 8.949955988750297e-06, + "loss": 0.8259, + "step": 51595 + }, + { + "epoch": 3.32, + "learning_rate": 8.949741299727346e-06, + "loss": 0.8553, + "step": 51596 + }, + { + "epoch": 3.32, + "learning_rate": 8.949526610704396e-06, + "loss": 0.8193, + "step": 51597 + }, + { + "epoch": 3.32, + "learning_rate": 8.949311921681445e-06, + "loss": 0.8679, + "step": 51598 + }, + { + "epoch": 3.32, + "learning_rate": 8.949097232658494e-06, + "loss": 0.8501, + "step": 51599 + }, + { + "epoch": 3.32, + "learning_rate": 8.948882543635545e-06, + "loss": 0.8896, + "step": 51600 + }, + { + "epoch": 3.32, + "learning_rate": 8.948667854612595e-06, + "loss": 0.8515, + "step": 51601 + }, + { + "epoch": 3.32, + "learning_rate": 8.948453165589644e-06, + "loss": 0.8805, + "step": 51602 + }, + { + "epoch": 3.32, + "learning_rate": 8.948238476566693e-06, + "loss": 0.8285, + "step": 51603 + }, + { + "epoch": 3.32, + "learning_rate": 8.948023787543744e-06, + "loss": 0.8341, + "step": 51604 + }, + { + "epoch": 3.32, + "learning_rate": 8.947809098520793e-06, + "loss": 0.8644, + "step": 51605 + }, + { + "epoch": 3.32, + "learning_rate": 8.947594409497844e-06, + "loss": 0.7993, + "step": 51606 + }, + { + "epoch": 3.32, + "learning_rate": 8.947379720474893e-06, + "loss": 0.8253, + "step": 51607 + }, + { + "epoch": 3.32, + "learning_rate": 8.947165031451942e-06, + "loss": 0.8417, + "step": 51608 + }, + { + "epoch": 3.32, + "learning_rate": 8.946950342428992e-06, + "loss": 0.8615, + "step": 51609 + }, + { + "epoch": 3.32, + "learning_rate": 8.946735653406043e-06, + "loss": 0.829, + "step": 51610 + }, + { + "epoch": 3.32, + "learning_rate": 8.946520964383092e-06, + "loss": 0.7858, + "step": 51611 + }, + { + "epoch": 3.32, + "learning_rate": 8.94630627536014e-06, + "loss": 0.861, + "step": 51612 + }, + { + "epoch": 3.32, + "learning_rate": 8.946091586337191e-06, + "loss": 0.8477, + "step": 51613 + }, + { + "epoch": 3.32, + "learning_rate": 8.94587689731424e-06, + "loss": 0.8465, + "step": 51614 + }, + { + "epoch": 3.32, + "learning_rate": 8.945662208291291e-06, + "loss": 0.8293, + "step": 51615 + }, + { + "epoch": 3.32, + "learning_rate": 8.945447519268342e-06, + "loss": 0.7973, + "step": 51616 + }, + { + "epoch": 3.32, + "learning_rate": 8.94523283024539e-06, + "loss": 0.8126, + "step": 51617 + }, + { + "epoch": 3.32, + "learning_rate": 8.94501814122244e-06, + "loss": 0.8964, + "step": 51618 + }, + { + "epoch": 3.32, + "learning_rate": 8.94480345219949e-06, + "loss": 0.7553, + "step": 51619 + }, + { + "epoch": 3.32, + "learning_rate": 8.94458876317654e-06, + "loss": 0.8246, + "step": 51620 + }, + { + "epoch": 3.32, + "learning_rate": 8.944374074153588e-06, + "loss": 0.7841, + "step": 51621 + }, + { + "epoch": 3.32, + "learning_rate": 8.944159385130639e-06, + "loss": 0.9, + "step": 51622 + }, + { + "epoch": 3.32, + "learning_rate": 8.94394469610769e-06, + "loss": 0.8546, + "step": 51623 + }, + { + "epoch": 3.32, + "learning_rate": 8.943730007084739e-06, + "loss": 0.9309, + "step": 51624 + }, + { + "epoch": 3.32, + "learning_rate": 8.94351531806179e-06, + "loss": 0.7953, + "step": 51625 + }, + { + "epoch": 3.32, + "learning_rate": 8.943300629038838e-06, + "loss": 0.8598, + "step": 51626 + }, + { + "epoch": 3.32, + "learning_rate": 8.943085940015887e-06, + "loss": 0.8152, + "step": 51627 + }, + { + "epoch": 3.32, + "learning_rate": 8.942871250992938e-06, + "loss": 0.8316, + "step": 51628 + }, + { + "epoch": 3.32, + "learning_rate": 8.942656561969987e-06, + "loss": 0.8508, + "step": 51629 + }, + { + "epoch": 3.32, + "learning_rate": 8.942441872947037e-06, + "loss": 0.8359, + "step": 51630 + }, + { + "epoch": 3.32, + "learning_rate": 8.942227183924086e-06, + "loss": 0.8271, + "step": 51631 + }, + { + "epoch": 3.32, + "learning_rate": 8.942012494901137e-06, + "loss": 0.8357, + "step": 51632 + }, + { + "epoch": 3.32, + "learning_rate": 8.941797805878186e-06, + "loss": 0.8121, + "step": 51633 + }, + { + "epoch": 3.32, + "learning_rate": 8.941583116855235e-06, + "loss": 0.8388, + "step": 51634 + }, + { + "epoch": 3.32, + "learning_rate": 8.941368427832286e-06, + "loss": 0.84, + "step": 51635 + }, + { + "epoch": 3.32, + "learning_rate": 8.941153738809335e-06, + "loss": 0.798, + "step": 51636 + }, + { + "epoch": 3.32, + "learning_rate": 8.940939049786385e-06, + "loss": 0.8902, + "step": 51637 + }, + { + "epoch": 3.32, + "learning_rate": 8.940724360763436e-06, + "loss": 0.7879, + "step": 51638 + }, + { + "epoch": 3.32, + "learning_rate": 8.940509671740485e-06, + "loss": 0.8389, + "step": 51639 + }, + { + "epoch": 3.32, + "learning_rate": 8.940294982717534e-06, + "loss": 0.7908, + "step": 51640 + }, + { + "epoch": 3.32, + "learning_rate": 8.940080293694585e-06, + "loss": 0.7725, + "step": 51641 + }, + { + "epoch": 3.32, + "learning_rate": 8.939865604671634e-06, + "loss": 0.9026, + "step": 51642 + }, + { + "epoch": 3.32, + "learning_rate": 8.939650915648682e-06, + "loss": 0.8439, + "step": 51643 + }, + { + "epoch": 3.32, + "learning_rate": 8.939436226625733e-06, + "loss": 0.7551, + "step": 51644 + }, + { + "epoch": 3.32, + "learning_rate": 8.939221537602784e-06, + "loss": 0.8889, + "step": 51645 + }, + { + "epoch": 3.32, + "learning_rate": 8.939006848579833e-06, + "loss": 0.8104, + "step": 51646 + }, + { + "epoch": 3.32, + "learning_rate": 8.938792159556883e-06, + "loss": 0.8244, + "step": 51647 + }, + { + "epoch": 3.32, + "learning_rate": 8.938577470533932e-06, + "loss": 0.8577, + "step": 51648 + }, + { + "epoch": 3.32, + "learning_rate": 8.938362781510981e-06, + "loss": 0.8264, + "step": 51649 + }, + { + "epoch": 3.32, + "learning_rate": 8.938148092488032e-06, + "loss": 0.8237, + "step": 51650 + }, + { + "epoch": 3.32, + "learning_rate": 8.937933403465081e-06, + "loss": 0.8999, + "step": 51651 + }, + { + "epoch": 3.32, + "learning_rate": 8.937718714442132e-06, + "loss": 0.7929, + "step": 51652 + }, + { + "epoch": 3.32, + "learning_rate": 8.93750402541918e-06, + "loss": 0.8135, + "step": 51653 + }, + { + "epoch": 3.32, + "learning_rate": 8.937289336396231e-06, + "loss": 0.8586, + "step": 51654 + }, + { + "epoch": 3.32, + "learning_rate": 8.93707464737328e-06, + "loss": 0.8466, + "step": 51655 + }, + { + "epoch": 3.32, + "learning_rate": 8.936859958350331e-06, + "loss": 0.8262, + "step": 51656 + }, + { + "epoch": 3.32, + "learning_rate": 8.93664526932738e-06, + "loss": 0.7501, + "step": 51657 + }, + { + "epoch": 3.32, + "learning_rate": 8.936430580304429e-06, + "loss": 0.8, + "step": 51658 + }, + { + "epoch": 3.32, + "learning_rate": 8.93621589128148e-06, + "loss": 0.7994, + "step": 51659 + }, + { + "epoch": 3.32, + "learning_rate": 8.93600120225853e-06, + "loss": 0.8336, + "step": 51660 + }, + { + "epoch": 3.32, + "learning_rate": 8.935786513235579e-06, + "loss": 0.7706, + "step": 51661 + }, + { + "epoch": 3.32, + "learning_rate": 8.935571824212628e-06, + "loss": 0.7026, + "step": 51662 + }, + { + "epoch": 3.32, + "learning_rate": 8.935357135189679e-06, + "loss": 0.8604, + "step": 51663 + }, + { + "epoch": 3.32, + "learning_rate": 8.935142446166728e-06, + "loss": 0.8059, + "step": 51664 + }, + { + "epoch": 3.32, + "learning_rate": 8.934927757143778e-06, + "loss": 0.8251, + "step": 51665 + }, + { + "epoch": 3.32, + "learning_rate": 8.934713068120827e-06, + "loss": 0.8184, + "step": 51666 + }, + { + "epoch": 3.32, + "learning_rate": 8.934498379097878e-06, + "loss": 0.7731, + "step": 51667 + }, + { + "epoch": 3.32, + "learning_rate": 8.934283690074927e-06, + "loss": 0.804, + "step": 51668 + }, + { + "epoch": 3.32, + "learning_rate": 8.934069001051978e-06, + "loss": 0.7862, + "step": 51669 + }, + { + "epoch": 3.32, + "learning_rate": 8.933854312029027e-06, + "loss": 0.7685, + "step": 51670 + }, + { + "epoch": 3.32, + "learning_rate": 8.933639623006076e-06, + "loss": 0.79, + "step": 51671 + }, + { + "epoch": 3.32, + "learning_rate": 8.933424933983126e-06, + "loss": 0.8408, + "step": 51672 + }, + { + "epoch": 3.32, + "learning_rate": 8.933210244960175e-06, + "loss": 0.7341, + "step": 51673 + }, + { + "epoch": 3.32, + "learning_rate": 8.932995555937226e-06, + "loss": 0.862, + "step": 51674 + }, + { + "epoch": 3.32, + "learning_rate": 8.932780866914276e-06, + "loss": 0.798, + "step": 51675 + }, + { + "epoch": 3.32, + "learning_rate": 8.932566177891325e-06, + "loss": 0.8407, + "step": 51676 + }, + { + "epoch": 3.32, + "learning_rate": 8.932351488868374e-06, + "loss": 0.8346, + "step": 51677 + }, + { + "epoch": 3.32, + "learning_rate": 8.932136799845425e-06, + "loss": 0.7882, + "step": 51678 + }, + { + "epoch": 3.32, + "learning_rate": 8.931922110822474e-06, + "loss": 0.8115, + "step": 51679 + }, + { + "epoch": 3.32, + "learning_rate": 8.931707421799523e-06, + "loss": 0.7747, + "step": 51680 + }, + { + "epoch": 3.33, + "learning_rate": 8.931492732776574e-06, + "loss": 0.8012, + "step": 51681 + }, + { + "epoch": 3.33, + "learning_rate": 8.931278043753624e-06, + "loss": 0.809, + "step": 51682 + }, + { + "epoch": 3.33, + "learning_rate": 8.931063354730673e-06, + "loss": 0.8213, + "step": 51683 + }, + { + "epoch": 3.33, + "learning_rate": 8.930848665707724e-06, + "loss": 0.8254, + "step": 51684 + }, + { + "epoch": 3.33, + "learning_rate": 8.930633976684773e-06, + "loss": 0.8065, + "step": 51685 + }, + { + "epoch": 3.33, + "learning_rate": 8.930419287661822e-06, + "loss": 0.7869, + "step": 51686 + }, + { + "epoch": 3.33, + "learning_rate": 8.930204598638872e-06, + "loss": 0.842, + "step": 51687 + }, + { + "epoch": 3.33, + "learning_rate": 8.929989909615923e-06, + "loss": 0.8454, + "step": 51688 + }, + { + "epoch": 3.33, + "learning_rate": 8.929775220592972e-06, + "loss": 0.7771, + "step": 51689 + }, + { + "epoch": 3.33, + "learning_rate": 8.929560531570021e-06, + "loss": 0.8049, + "step": 51690 + }, + { + "epoch": 3.33, + "learning_rate": 8.929345842547072e-06, + "loss": 0.8158, + "step": 51691 + }, + { + "epoch": 3.33, + "learning_rate": 8.92913115352412e-06, + "loss": 0.8944, + "step": 51692 + }, + { + "epoch": 3.33, + "learning_rate": 8.928916464501171e-06, + "loss": 0.9114, + "step": 51693 + }, + { + "epoch": 3.33, + "learning_rate": 8.92870177547822e-06, + "loss": 0.8174, + "step": 51694 + }, + { + "epoch": 3.33, + "learning_rate": 8.92848708645527e-06, + "loss": 0.7819, + "step": 51695 + }, + { + "epoch": 3.33, + "learning_rate": 8.92827239743232e-06, + "loss": 0.8422, + "step": 51696 + }, + { + "epoch": 3.33, + "learning_rate": 8.92805770840937e-06, + "loss": 0.8512, + "step": 51697 + }, + { + "epoch": 3.33, + "learning_rate": 8.92784301938642e-06, + "loss": 0.8052, + "step": 51698 + }, + { + "epoch": 3.33, + "learning_rate": 8.927628330363469e-06, + "loss": 0.7941, + "step": 51699 + }, + { + "epoch": 3.33, + "learning_rate": 8.92741364134052e-06, + "loss": 0.851, + "step": 51700 + }, + { + "epoch": 3.33, + "learning_rate": 8.927198952317568e-06, + "loss": 0.76, + "step": 51701 + }, + { + "epoch": 3.33, + "learning_rate": 8.926984263294617e-06, + "loss": 0.8242, + "step": 51702 + }, + { + "epoch": 3.33, + "learning_rate": 8.92676957427167e-06, + "loss": 0.8012, + "step": 51703 + }, + { + "epoch": 3.33, + "learning_rate": 8.926554885248718e-06, + "loss": 0.7921, + "step": 51704 + }, + { + "epoch": 3.33, + "learning_rate": 8.926340196225767e-06, + "loss": 0.873, + "step": 51705 + }, + { + "epoch": 3.33, + "learning_rate": 8.926125507202818e-06, + "loss": 0.7869, + "step": 51706 + }, + { + "epoch": 3.33, + "learning_rate": 8.925910818179867e-06, + "loss": 0.8262, + "step": 51707 + }, + { + "epoch": 3.33, + "learning_rate": 8.925696129156916e-06, + "loss": 0.798, + "step": 51708 + }, + { + "epoch": 3.33, + "learning_rate": 8.925481440133967e-06, + "loss": 0.8451, + "step": 51709 + }, + { + "epoch": 3.33, + "learning_rate": 8.925266751111017e-06, + "loss": 0.8388, + "step": 51710 + }, + { + "epoch": 3.33, + "learning_rate": 8.925052062088066e-06, + "loss": 0.8445, + "step": 51711 + }, + { + "epoch": 3.33, + "learning_rate": 8.924837373065117e-06, + "loss": 0.8402, + "step": 51712 + }, + { + "epoch": 3.33, + "learning_rate": 8.924622684042166e-06, + "loss": 0.7891, + "step": 51713 + }, + { + "epoch": 3.33, + "learning_rate": 8.924407995019215e-06, + "loss": 0.7821, + "step": 51714 + }, + { + "epoch": 3.33, + "learning_rate": 8.924193305996266e-06, + "loss": 0.7985, + "step": 51715 + }, + { + "epoch": 3.33, + "learning_rate": 8.923978616973315e-06, + "loss": 0.8966, + "step": 51716 + }, + { + "epoch": 3.33, + "learning_rate": 8.923763927950365e-06, + "loss": 0.8405, + "step": 51717 + }, + { + "epoch": 3.33, + "learning_rate": 8.923549238927414e-06, + "loss": 0.8554, + "step": 51718 + }, + { + "epoch": 3.33, + "learning_rate": 8.923334549904465e-06, + "loss": 0.8856, + "step": 51719 + }, + { + "epoch": 3.33, + "learning_rate": 8.923119860881514e-06, + "loss": 0.8233, + "step": 51720 + }, + { + "epoch": 3.33, + "learning_rate": 8.922905171858563e-06, + "loss": 0.8072, + "step": 51721 + }, + { + "epoch": 3.33, + "learning_rate": 8.922690482835613e-06, + "loss": 0.8038, + "step": 51722 + }, + { + "epoch": 3.33, + "learning_rate": 8.922475793812662e-06, + "loss": 0.8031, + "step": 51723 + }, + { + "epoch": 3.33, + "learning_rate": 8.922261104789713e-06, + "loss": 0.8897, + "step": 51724 + }, + { + "epoch": 3.33, + "learning_rate": 8.922046415766764e-06, + "loss": 0.8719, + "step": 51725 + }, + { + "epoch": 3.33, + "learning_rate": 8.921831726743813e-06, + "loss": 0.903, + "step": 51726 + }, + { + "epoch": 3.33, + "learning_rate": 8.921617037720862e-06, + "loss": 0.7856, + "step": 51727 + }, + { + "epoch": 3.33, + "learning_rate": 8.921402348697912e-06, + "loss": 0.7888, + "step": 51728 + }, + { + "epoch": 3.33, + "learning_rate": 8.921187659674961e-06, + "loss": 0.7915, + "step": 51729 + }, + { + "epoch": 3.33, + "learning_rate": 8.92097297065201e-06, + "loss": 0.8333, + "step": 51730 + }, + { + "epoch": 3.33, + "learning_rate": 8.92075828162906e-06, + "loss": 0.8239, + "step": 51731 + }, + { + "epoch": 3.33, + "learning_rate": 8.920543592606111e-06, + "loss": 0.8298, + "step": 51732 + }, + { + "epoch": 3.33, + "learning_rate": 8.92032890358316e-06, + "loss": 0.8709, + "step": 51733 + }, + { + "epoch": 3.33, + "learning_rate": 8.920114214560211e-06, + "loss": 0.8711, + "step": 51734 + }, + { + "epoch": 3.33, + "learning_rate": 8.91989952553726e-06, + "loss": 0.8345, + "step": 51735 + }, + { + "epoch": 3.33, + "learning_rate": 8.919684836514309e-06, + "loss": 0.805, + "step": 51736 + }, + { + "epoch": 3.33, + "learning_rate": 8.91947014749136e-06, + "loss": 0.7808, + "step": 51737 + }, + { + "epoch": 3.33, + "learning_rate": 8.919255458468409e-06, + "loss": 0.8255, + "step": 51738 + }, + { + "epoch": 3.33, + "learning_rate": 8.91904076944546e-06, + "loss": 0.816, + "step": 51739 + }, + { + "epoch": 3.33, + "learning_rate": 8.918826080422508e-06, + "loss": 0.8181, + "step": 51740 + }, + { + "epoch": 3.33, + "learning_rate": 8.918611391399559e-06, + "loss": 0.8, + "step": 51741 + }, + { + "epoch": 3.33, + "learning_rate": 8.918396702376608e-06, + "loss": 0.7971, + "step": 51742 + }, + { + "epoch": 3.33, + "learning_rate": 8.918182013353659e-06, + "loss": 0.8604, + "step": 51743 + }, + { + "epoch": 3.33, + "learning_rate": 8.917967324330708e-06, + "loss": 0.8455, + "step": 51744 + }, + { + "epoch": 3.33, + "learning_rate": 8.917752635307757e-06, + "loss": 0.8276, + "step": 51745 + }, + { + "epoch": 3.33, + "learning_rate": 8.917537946284807e-06, + "loss": 0.8014, + "step": 51746 + }, + { + "epoch": 3.33, + "learning_rate": 8.917323257261858e-06, + "loss": 0.8232, + "step": 51747 + }, + { + "epoch": 3.33, + "learning_rate": 8.917108568238907e-06, + "loss": 0.8729, + "step": 51748 + }, + { + "epoch": 3.33, + "learning_rate": 8.916893879215956e-06, + "loss": 0.8722, + "step": 51749 + }, + { + "epoch": 3.33, + "learning_rate": 8.916679190193006e-06, + "loss": 0.8293, + "step": 51750 + }, + { + "epoch": 3.33, + "learning_rate": 8.916464501170055e-06, + "loss": 0.9073, + "step": 51751 + }, + { + "epoch": 3.33, + "learning_rate": 8.916249812147106e-06, + "loss": 0.8532, + "step": 51752 + }, + { + "epoch": 3.33, + "learning_rate": 8.916035123124155e-06, + "loss": 0.8071, + "step": 51753 + }, + { + "epoch": 3.33, + "learning_rate": 8.915820434101206e-06, + "loss": 0.8188, + "step": 51754 + }, + { + "epoch": 3.33, + "learning_rate": 8.915605745078255e-06, + "loss": 0.7923, + "step": 51755 + }, + { + "epoch": 3.33, + "learning_rate": 8.915391056055305e-06, + "loss": 0.7985, + "step": 51756 + }, + { + "epoch": 3.33, + "learning_rate": 8.915176367032354e-06, + "loss": 0.7994, + "step": 51757 + }, + { + "epoch": 3.33, + "learning_rate": 8.914961678009403e-06, + "loss": 0.8211, + "step": 51758 + }, + { + "epoch": 3.33, + "learning_rate": 8.914746988986454e-06, + "loss": 0.8303, + "step": 51759 + }, + { + "epoch": 3.33, + "learning_rate": 8.914532299963503e-06, + "loss": 0.7599, + "step": 51760 + }, + { + "epoch": 3.33, + "learning_rate": 8.914317610940554e-06, + "loss": 0.8004, + "step": 51761 + }, + { + "epoch": 3.33, + "learning_rate": 8.914102921917604e-06, + "loss": 0.8096, + "step": 51762 + }, + { + "epoch": 3.33, + "learning_rate": 8.913888232894653e-06, + "loss": 0.8591, + "step": 51763 + }, + { + "epoch": 3.33, + "learning_rate": 8.913673543871702e-06, + "loss": 0.8519, + "step": 51764 + }, + { + "epoch": 3.33, + "learning_rate": 8.913458854848753e-06, + "loss": 0.856, + "step": 51765 + }, + { + "epoch": 3.33, + "learning_rate": 8.913244165825802e-06, + "loss": 0.8294, + "step": 51766 + }, + { + "epoch": 3.33, + "learning_rate": 8.91302947680285e-06, + "loss": 0.8481, + "step": 51767 + }, + { + "epoch": 3.33, + "learning_rate": 8.912814787779901e-06, + "loss": 0.7534, + "step": 51768 + }, + { + "epoch": 3.33, + "learning_rate": 8.912600098756952e-06, + "loss": 0.8219, + "step": 51769 + }, + { + "epoch": 3.33, + "learning_rate": 8.912385409734001e-06, + "loss": 0.9141, + "step": 51770 + }, + { + "epoch": 3.33, + "learning_rate": 8.912170720711052e-06, + "loss": 0.7685, + "step": 51771 + }, + { + "epoch": 3.33, + "learning_rate": 8.9119560316881e-06, + "loss": 0.8696, + "step": 51772 + }, + { + "epoch": 3.33, + "learning_rate": 8.91174134266515e-06, + "loss": 0.916, + "step": 51773 + }, + { + "epoch": 3.33, + "learning_rate": 8.9115266536422e-06, + "loss": 0.8406, + "step": 51774 + }, + { + "epoch": 3.33, + "learning_rate": 8.911311964619251e-06, + "loss": 0.7773, + "step": 51775 + }, + { + "epoch": 3.33, + "learning_rate": 8.9110972755963e-06, + "loss": 0.8422, + "step": 51776 + }, + { + "epoch": 3.33, + "learning_rate": 8.910882586573349e-06, + "loss": 0.8818, + "step": 51777 + }, + { + "epoch": 3.33, + "learning_rate": 8.9106678975504e-06, + "loss": 0.8788, + "step": 51778 + }, + { + "epoch": 3.33, + "learning_rate": 8.910453208527448e-06, + "loss": 0.7799, + "step": 51779 + }, + { + "epoch": 3.33, + "learning_rate": 8.910238519504499e-06, + "loss": 0.7856, + "step": 51780 + }, + { + "epoch": 3.33, + "learning_rate": 8.910023830481548e-06, + "loss": 0.8533, + "step": 51781 + }, + { + "epoch": 3.33, + "learning_rate": 8.909809141458597e-06, + "loss": 0.8518, + "step": 51782 + }, + { + "epoch": 3.33, + "learning_rate": 8.909594452435648e-06, + "loss": 0.7958, + "step": 51783 + }, + { + "epoch": 3.33, + "learning_rate": 8.909379763412698e-06, + "loss": 0.7712, + "step": 51784 + }, + { + "epoch": 3.33, + "learning_rate": 8.909165074389747e-06, + "loss": 0.7825, + "step": 51785 + }, + { + "epoch": 3.33, + "learning_rate": 8.908950385366796e-06, + "loss": 0.8581, + "step": 51786 + }, + { + "epoch": 3.33, + "learning_rate": 8.908735696343847e-06, + "loss": 0.8003, + "step": 51787 + }, + { + "epoch": 3.33, + "learning_rate": 8.908521007320896e-06, + "loss": 0.8038, + "step": 51788 + }, + { + "epoch": 3.33, + "learning_rate": 8.908306318297945e-06, + "loss": 0.8563, + "step": 51789 + }, + { + "epoch": 3.33, + "learning_rate": 8.908091629274997e-06, + "loss": 0.8037, + "step": 51790 + }, + { + "epoch": 3.33, + "learning_rate": 8.907876940252046e-06, + "loss": 0.7877, + "step": 51791 + }, + { + "epoch": 3.33, + "learning_rate": 8.907662251229095e-06, + "loss": 0.8252, + "step": 51792 + }, + { + "epoch": 3.33, + "learning_rate": 8.907447562206146e-06, + "loss": 0.7818, + "step": 51793 + }, + { + "epoch": 3.33, + "learning_rate": 8.907232873183195e-06, + "loss": 0.8214, + "step": 51794 + }, + { + "epoch": 3.33, + "learning_rate": 8.907018184160244e-06, + "loss": 0.8466, + "step": 51795 + }, + { + "epoch": 3.33, + "learning_rate": 8.906803495137294e-06, + "loss": 0.7951, + "step": 51796 + }, + { + "epoch": 3.33, + "learning_rate": 8.906588806114345e-06, + "loss": 0.7872, + "step": 51797 + }, + { + "epoch": 3.33, + "learning_rate": 8.906374117091394e-06, + "loss": 0.8328, + "step": 51798 + }, + { + "epoch": 3.33, + "learning_rate": 8.906159428068443e-06, + "loss": 0.796, + "step": 51799 + }, + { + "epoch": 3.33, + "learning_rate": 8.905944739045494e-06, + "loss": 0.7986, + "step": 51800 + }, + { + "epoch": 3.33, + "learning_rate": 8.905730050022543e-06, + "loss": 0.8932, + "step": 51801 + }, + { + "epoch": 3.33, + "learning_rate": 8.905515360999593e-06, + "loss": 0.81, + "step": 51802 + }, + { + "epoch": 3.33, + "learning_rate": 8.905300671976642e-06, + "loss": 0.7923, + "step": 51803 + }, + { + "epoch": 3.33, + "learning_rate": 8.905085982953691e-06, + "loss": 0.8205, + "step": 51804 + }, + { + "epoch": 3.33, + "learning_rate": 8.904871293930742e-06, + "loss": 0.8466, + "step": 51805 + }, + { + "epoch": 3.33, + "learning_rate": 8.904656604907793e-06, + "loss": 0.8125, + "step": 51806 + }, + { + "epoch": 3.33, + "learning_rate": 8.904441915884841e-06, + "loss": 0.7845, + "step": 51807 + }, + { + "epoch": 3.33, + "learning_rate": 8.90422722686189e-06, + "loss": 0.7839, + "step": 51808 + }, + { + "epoch": 3.33, + "learning_rate": 8.904012537838941e-06, + "loss": 0.8739, + "step": 51809 + }, + { + "epoch": 3.33, + "learning_rate": 8.90379784881599e-06, + "loss": 0.8499, + "step": 51810 + }, + { + "epoch": 3.33, + "learning_rate": 8.90358315979304e-06, + "loss": 0.8667, + "step": 51811 + }, + { + "epoch": 3.33, + "learning_rate": 8.903368470770091e-06, + "loss": 0.8039, + "step": 51812 + }, + { + "epoch": 3.33, + "learning_rate": 8.90315378174714e-06, + "loss": 0.7744, + "step": 51813 + }, + { + "epoch": 3.33, + "learning_rate": 8.90293909272419e-06, + "loss": 0.8415, + "step": 51814 + }, + { + "epoch": 3.33, + "learning_rate": 8.90272440370124e-06, + "loss": 0.7585, + "step": 51815 + }, + { + "epoch": 3.33, + "learning_rate": 8.902509714678289e-06, + "loss": 0.8261, + "step": 51816 + }, + { + "epoch": 3.33, + "learning_rate": 8.902295025655338e-06, + "loss": 0.7952, + "step": 51817 + }, + { + "epoch": 3.33, + "learning_rate": 8.902080336632389e-06, + "loss": 0.8122, + "step": 51818 + }, + { + "epoch": 3.33, + "learning_rate": 8.90186564760944e-06, + "loss": 0.8034, + "step": 51819 + }, + { + "epoch": 3.33, + "learning_rate": 8.901650958586488e-06, + "loss": 0.8384, + "step": 51820 + }, + { + "epoch": 3.33, + "learning_rate": 8.901436269563539e-06, + "loss": 0.899, + "step": 51821 + }, + { + "epoch": 3.33, + "learning_rate": 8.901221580540588e-06, + "loss": 0.8459, + "step": 51822 + }, + { + "epoch": 3.33, + "learning_rate": 8.901006891517637e-06, + "loss": 0.8357, + "step": 51823 + }, + { + "epoch": 3.33, + "learning_rate": 8.900792202494687e-06, + "loss": 0.8143, + "step": 51824 + }, + { + "epoch": 3.33, + "learning_rate": 8.900577513471736e-06, + "loss": 0.8002, + "step": 51825 + }, + { + "epoch": 3.33, + "learning_rate": 8.900362824448787e-06, + "loss": 0.7695, + "step": 51826 + }, + { + "epoch": 3.33, + "learning_rate": 8.900148135425836e-06, + "loss": 0.7437, + "step": 51827 + }, + { + "epoch": 3.33, + "learning_rate": 8.899933446402887e-06, + "loss": 0.8298, + "step": 51828 + }, + { + "epoch": 3.33, + "learning_rate": 8.899718757379936e-06, + "loss": 0.8585, + "step": 51829 + }, + { + "epoch": 3.33, + "learning_rate": 8.899504068356986e-06, + "loss": 0.8432, + "step": 51830 + }, + { + "epoch": 3.33, + "learning_rate": 8.899289379334035e-06, + "loss": 0.7747, + "step": 51831 + }, + { + "epoch": 3.33, + "learning_rate": 8.899074690311084e-06, + "loss": 0.8192, + "step": 51832 + }, + { + "epoch": 3.33, + "learning_rate": 8.898860001288135e-06, + "loss": 0.8329, + "step": 51833 + }, + { + "epoch": 3.33, + "learning_rate": 8.898645312265186e-06, + "loss": 0.8196, + "step": 51834 + }, + { + "epoch": 3.33, + "learning_rate": 8.898430623242235e-06, + "loss": 0.8795, + "step": 51835 + }, + { + "epoch": 3.34, + "learning_rate": 8.898215934219283e-06, + "loss": 0.8401, + "step": 51836 + }, + { + "epoch": 3.34, + "learning_rate": 8.898001245196334e-06, + "loss": 0.848, + "step": 51837 + }, + { + "epoch": 3.34, + "learning_rate": 8.897786556173383e-06, + "loss": 0.814, + "step": 51838 + }, + { + "epoch": 3.34, + "learning_rate": 8.897571867150434e-06, + "loss": 0.7334, + "step": 51839 + }, + { + "epoch": 3.34, + "learning_rate": 8.897357178127483e-06, + "loss": 0.8573, + "step": 51840 + }, + { + "epoch": 3.34, + "learning_rate": 8.897142489104533e-06, + "loss": 0.8863, + "step": 51841 + }, + { + "epoch": 3.34, + "learning_rate": 8.896927800081582e-06, + "loss": 0.8376, + "step": 51842 + }, + { + "epoch": 3.34, + "learning_rate": 8.896713111058633e-06, + "loss": 0.8385, + "step": 51843 + }, + { + "epoch": 3.34, + "learning_rate": 8.896498422035682e-06, + "loss": 0.8509, + "step": 51844 + }, + { + "epoch": 3.34, + "learning_rate": 8.896283733012731e-06, + "loss": 0.8396, + "step": 51845 + }, + { + "epoch": 3.34, + "learning_rate": 8.896069043989782e-06, + "loss": 0.8744, + "step": 51846 + }, + { + "epoch": 3.34, + "learning_rate": 8.89585435496683e-06, + "loss": 0.9259, + "step": 51847 + }, + { + "epoch": 3.34, + "learning_rate": 8.895639665943881e-06, + "loss": 0.8184, + "step": 51848 + }, + { + "epoch": 3.34, + "learning_rate": 8.895424976920932e-06, + "loss": 0.8555, + "step": 51849 + }, + { + "epoch": 3.34, + "learning_rate": 8.89521028789798e-06, + "loss": 0.7636, + "step": 51850 + }, + { + "epoch": 3.34, + "learning_rate": 8.89499559887503e-06, + "loss": 0.8259, + "step": 51851 + }, + { + "epoch": 3.34, + "learning_rate": 8.89478090985208e-06, + "loss": 0.7871, + "step": 51852 + }, + { + "epoch": 3.34, + "learning_rate": 8.89456622082913e-06, + "loss": 0.7685, + "step": 51853 + }, + { + "epoch": 3.34, + "learning_rate": 8.894351531806178e-06, + "loss": 0.8045, + "step": 51854 + }, + { + "epoch": 3.34, + "learning_rate": 8.894136842783229e-06, + "loss": 0.8176, + "step": 51855 + }, + { + "epoch": 3.34, + "learning_rate": 8.89392215376028e-06, + "loss": 0.8711, + "step": 51856 + }, + { + "epoch": 3.34, + "learning_rate": 8.893707464737329e-06, + "loss": 0.8866, + "step": 51857 + }, + { + "epoch": 3.34, + "learning_rate": 8.89349277571438e-06, + "loss": 0.8116, + "step": 51858 + }, + { + "epoch": 3.34, + "learning_rate": 8.893278086691428e-06, + "loss": 0.7803, + "step": 51859 + }, + { + "epoch": 3.34, + "learning_rate": 8.893063397668477e-06, + "loss": 0.7155, + "step": 51860 + }, + { + "epoch": 3.34, + "learning_rate": 8.892848708645528e-06, + "loss": 0.7848, + "step": 51861 + }, + { + "epoch": 3.34, + "learning_rate": 8.892634019622577e-06, + "loss": 0.7938, + "step": 51862 + }, + { + "epoch": 3.34, + "learning_rate": 8.892419330599628e-06, + "loss": 0.8229, + "step": 51863 + }, + { + "epoch": 3.34, + "learning_rate": 8.892204641576677e-06, + "loss": 0.8817, + "step": 51864 + }, + { + "epoch": 3.34, + "learning_rate": 8.891989952553727e-06, + "loss": 0.8439, + "step": 51865 + }, + { + "epoch": 3.34, + "learning_rate": 8.891775263530776e-06, + "loss": 0.8287, + "step": 51866 + }, + { + "epoch": 3.34, + "learning_rate": 8.891560574507827e-06, + "loss": 0.8551, + "step": 51867 + }, + { + "epoch": 3.34, + "learning_rate": 8.891345885484876e-06, + "loss": 0.8933, + "step": 51868 + }, + { + "epoch": 3.34, + "learning_rate": 8.891131196461925e-06, + "loss": 0.7628, + "step": 51869 + }, + { + "epoch": 3.34, + "learning_rate": 8.890916507438975e-06, + "loss": 0.8532, + "step": 51870 + }, + { + "epoch": 3.34, + "learning_rate": 8.890701818416026e-06, + "loss": 0.8224, + "step": 51871 + }, + { + "epoch": 3.34, + "learning_rate": 8.890487129393075e-06, + "loss": 0.8968, + "step": 51872 + }, + { + "epoch": 3.34, + "learning_rate": 8.890272440370124e-06, + "loss": 0.7738, + "step": 51873 + }, + { + "epoch": 3.34, + "learning_rate": 8.890057751347175e-06, + "loss": 0.7885, + "step": 51874 + }, + { + "epoch": 3.34, + "learning_rate": 8.889843062324224e-06, + "loss": 0.7528, + "step": 51875 + }, + { + "epoch": 3.34, + "learning_rate": 8.889628373301273e-06, + "loss": 0.8891, + "step": 51876 + }, + { + "epoch": 3.34, + "learning_rate": 8.889413684278325e-06, + "loss": 0.8098, + "step": 51877 + }, + { + "epoch": 3.34, + "learning_rate": 8.889198995255374e-06, + "loss": 0.7885, + "step": 51878 + }, + { + "epoch": 3.34, + "learning_rate": 8.888984306232423e-06, + "loss": 0.9245, + "step": 51879 + }, + { + "epoch": 3.34, + "learning_rate": 8.888769617209474e-06, + "loss": 0.7656, + "step": 51880 + }, + { + "epoch": 3.34, + "learning_rate": 8.888554928186522e-06, + "loss": 0.825, + "step": 51881 + }, + { + "epoch": 3.34, + "learning_rate": 8.888340239163571e-06, + "loss": 0.8669, + "step": 51882 + }, + { + "epoch": 3.34, + "learning_rate": 8.888125550140622e-06, + "loss": 0.8733, + "step": 51883 + }, + { + "epoch": 3.34, + "learning_rate": 8.887910861117673e-06, + "loss": 0.8037, + "step": 51884 + }, + { + "epoch": 3.34, + "learning_rate": 8.887696172094722e-06, + "loss": 0.8347, + "step": 51885 + }, + { + "epoch": 3.34, + "learning_rate": 8.88748148307177e-06, + "loss": 0.7958, + "step": 51886 + }, + { + "epoch": 3.34, + "learning_rate": 8.887266794048821e-06, + "loss": 0.8026, + "step": 51887 + }, + { + "epoch": 3.34, + "learning_rate": 8.88705210502587e-06, + "loss": 0.8456, + "step": 51888 + }, + { + "epoch": 3.34, + "learning_rate": 8.886837416002921e-06, + "loss": 0.808, + "step": 51889 + }, + { + "epoch": 3.34, + "learning_rate": 8.88662272697997e-06, + "loss": 0.8623, + "step": 51890 + }, + { + "epoch": 3.34, + "learning_rate": 8.886408037957019e-06, + "loss": 0.8711, + "step": 51891 + }, + { + "epoch": 3.34, + "learning_rate": 8.88619334893407e-06, + "loss": 0.802, + "step": 51892 + }, + { + "epoch": 3.34, + "learning_rate": 8.88597865991112e-06, + "loss": 0.8184, + "step": 51893 + }, + { + "epoch": 3.34, + "learning_rate": 8.88576397088817e-06, + "loss": 0.8411, + "step": 51894 + }, + { + "epoch": 3.34, + "learning_rate": 8.885549281865218e-06, + "loss": 0.8477, + "step": 51895 + }, + { + "epoch": 3.34, + "learning_rate": 8.885334592842269e-06, + "loss": 0.8719, + "step": 51896 + }, + { + "epoch": 3.34, + "learning_rate": 8.885119903819318e-06, + "loss": 0.898, + "step": 51897 + }, + { + "epoch": 3.34, + "learning_rate": 8.884905214796368e-06, + "loss": 0.8741, + "step": 51898 + }, + { + "epoch": 3.34, + "learning_rate": 8.884690525773419e-06, + "loss": 0.8257, + "step": 51899 + }, + { + "epoch": 3.34, + "learning_rate": 8.884475836750468e-06, + "loss": 0.8599, + "step": 51900 + }, + { + "epoch": 3.34, + "learning_rate": 8.884261147727517e-06, + "loss": 0.7737, + "step": 51901 + }, + { + "epoch": 3.34, + "learning_rate": 8.884046458704568e-06, + "loss": 0.7801, + "step": 51902 + }, + { + "epoch": 3.34, + "learning_rate": 8.883831769681617e-06, + "loss": 0.8011, + "step": 51903 + }, + { + "epoch": 3.34, + "learning_rate": 8.883617080658666e-06, + "loss": 0.7987, + "step": 51904 + }, + { + "epoch": 3.34, + "learning_rate": 8.883402391635716e-06, + "loss": 0.8353, + "step": 51905 + }, + { + "epoch": 3.34, + "learning_rate": 8.883187702612767e-06, + "loss": 0.8589, + "step": 51906 + }, + { + "epoch": 3.34, + "learning_rate": 8.882973013589816e-06, + "loss": 0.8329, + "step": 51907 + }, + { + "epoch": 3.34, + "learning_rate": 8.882758324566867e-06, + "loss": 0.8088, + "step": 51908 + }, + { + "epoch": 3.34, + "learning_rate": 8.882543635543916e-06, + "loss": 0.824, + "step": 51909 + }, + { + "epoch": 3.34, + "learning_rate": 8.882328946520964e-06, + "loss": 0.8353, + "step": 51910 + }, + { + "epoch": 3.34, + "learning_rate": 8.882114257498015e-06, + "loss": 0.8806, + "step": 51911 + }, + { + "epoch": 3.34, + "learning_rate": 8.881899568475064e-06, + "loss": 0.8221, + "step": 51912 + }, + { + "epoch": 3.34, + "learning_rate": 8.881684879452115e-06, + "loss": 0.8071, + "step": 51913 + }, + { + "epoch": 3.34, + "learning_rate": 8.881470190429164e-06, + "loss": 0.8517, + "step": 51914 + }, + { + "epoch": 3.34, + "learning_rate": 8.881255501406214e-06, + "loss": 0.958, + "step": 51915 + }, + { + "epoch": 3.34, + "learning_rate": 8.881040812383263e-06, + "loss": 0.8835, + "step": 51916 + }, + { + "epoch": 3.34, + "learning_rate": 8.880826123360314e-06, + "loss": 0.8643, + "step": 51917 + }, + { + "epoch": 3.34, + "learning_rate": 8.880611434337363e-06, + "loss": 0.8161, + "step": 51918 + }, + { + "epoch": 3.34, + "learning_rate": 8.880396745314412e-06, + "loss": 0.8783, + "step": 51919 + }, + { + "epoch": 3.34, + "learning_rate": 8.880182056291463e-06, + "loss": 0.861, + "step": 51920 + }, + { + "epoch": 3.34, + "learning_rate": 8.879967367268513e-06, + "loss": 0.8296, + "step": 51921 + }, + { + "epoch": 3.34, + "learning_rate": 8.879752678245562e-06, + "loss": 0.7953, + "step": 51922 + }, + { + "epoch": 3.34, + "learning_rate": 8.879537989222611e-06, + "loss": 0.8209, + "step": 51923 + }, + { + "epoch": 3.34, + "learning_rate": 8.879323300199662e-06, + "loss": 0.8525, + "step": 51924 + }, + { + "epoch": 3.34, + "learning_rate": 8.87910861117671e-06, + "loss": 0.8331, + "step": 51925 + }, + { + "epoch": 3.34, + "learning_rate": 8.878893922153761e-06, + "loss": 0.8315, + "step": 51926 + }, + { + "epoch": 3.34, + "learning_rate": 8.87867923313081e-06, + "loss": 0.8776, + "step": 51927 + }, + { + "epoch": 3.34, + "learning_rate": 8.878464544107861e-06, + "loss": 0.83, + "step": 51928 + }, + { + "epoch": 3.34, + "learning_rate": 8.87824985508491e-06, + "loss": 0.8887, + "step": 51929 + }, + { + "epoch": 3.34, + "learning_rate": 8.87803516606196e-06, + "loss": 0.872, + "step": 51930 + }, + { + "epoch": 3.34, + "learning_rate": 8.87782047703901e-06, + "loss": 0.8274, + "step": 51931 + }, + { + "epoch": 3.34, + "learning_rate": 8.877605788016059e-06, + "loss": 0.8518, + "step": 51932 + }, + { + "epoch": 3.34, + "learning_rate": 8.87739109899311e-06, + "loss": 0.8848, + "step": 51933 + }, + { + "epoch": 3.34, + "learning_rate": 8.877176409970158e-06, + "loss": 0.8203, + "step": 51934 + }, + { + "epoch": 3.34, + "learning_rate": 8.876961720947209e-06, + "loss": 0.7952, + "step": 51935 + }, + { + "epoch": 3.34, + "learning_rate": 8.87674703192426e-06, + "loss": 0.8023, + "step": 51936 + }, + { + "epoch": 3.34, + "learning_rate": 8.876532342901309e-06, + "loss": 0.7365, + "step": 51937 + }, + { + "epoch": 3.34, + "learning_rate": 8.876317653878358e-06, + "loss": 0.801, + "step": 51938 + }, + { + "epoch": 3.34, + "learning_rate": 8.876102964855408e-06, + "loss": 0.8066, + "step": 51939 + }, + { + "epoch": 3.34, + "learning_rate": 8.875888275832457e-06, + "loss": 0.8066, + "step": 51940 + }, + { + "epoch": 3.34, + "learning_rate": 8.875673586809506e-06, + "loss": 0.8072, + "step": 51941 + }, + { + "epoch": 3.34, + "learning_rate": 8.875458897786557e-06, + "loss": 0.8569, + "step": 51942 + }, + { + "epoch": 3.34, + "learning_rate": 8.875244208763607e-06, + "loss": 0.8245, + "step": 51943 + }, + { + "epoch": 3.34, + "learning_rate": 8.875029519740656e-06, + "loss": 0.851, + "step": 51944 + }, + { + "epoch": 3.34, + "learning_rate": 8.874814830717707e-06, + "loss": 0.8595, + "step": 51945 + }, + { + "epoch": 3.34, + "learning_rate": 8.874600141694756e-06, + "loss": 0.8658, + "step": 51946 + }, + { + "epoch": 3.34, + "learning_rate": 8.874385452671805e-06, + "loss": 0.8613, + "step": 51947 + }, + { + "epoch": 3.34, + "learning_rate": 8.874170763648856e-06, + "loss": 0.8123, + "step": 51948 + }, + { + "epoch": 3.34, + "learning_rate": 8.873956074625905e-06, + "loss": 0.8849, + "step": 51949 + }, + { + "epoch": 3.34, + "learning_rate": 8.873741385602955e-06, + "loss": 0.848, + "step": 51950 + }, + { + "epoch": 3.34, + "learning_rate": 8.873526696580004e-06, + "loss": 0.7829, + "step": 51951 + }, + { + "epoch": 3.34, + "learning_rate": 8.873312007557055e-06, + "loss": 0.8489, + "step": 51952 + }, + { + "epoch": 3.34, + "learning_rate": 8.873097318534104e-06, + "loss": 0.8012, + "step": 51953 + }, + { + "epoch": 3.34, + "learning_rate": 8.872882629511153e-06, + "loss": 0.8051, + "step": 51954 + }, + { + "epoch": 3.34, + "learning_rate": 8.872667940488203e-06, + "loss": 0.863, + "step": 51955 + }, + { + "epoch": 3.34, + "learning_rate": 8.872453251465252e-06, + "loss": 0.815, + "step": 51956 + }, + { + "epoch": 3.34, + "learning_rate": 8.872238562442303e-06, + "loss": 0.7924, + "step": 51957 + }, + { + "epoch": 3.34, + "learning_rate": 8.872023873419354e-06, + "loss": 0.8182, + "step": 51958 + }, + { + "epoch": 3.34, + "learning_rate": 8.871809184396403e-06, + "loss": 0.8329, + "step": 51959 + }, + { + "epoch": 3.34, + "learning_rate": 8.871594495373452e-06, + "loss": 0.8409, + "step": 51960 + }, + { + "epoch": 3.34, + "learning_rate": 8.871379806350502e-06, + "loss": 0.8762, + "step": 51961 + }, + { + "epoch": 3.34, + "learning_rate": 8.871165117327551e-06, + "loss": 0.8076, + "step": 51962 + }, + { + "epoch": 3.34, + "learning_rate": 8.8709504283046e-06, + "loss": 0.8303, + "step": 51963 + }, + { + "epoch": 3.34, + "learning_rate": 8.870735739281651e-06, + "loss": 0.8019, + "step": 51964 + }, + { + "epoch": 3.34, + "learning_rate": 8.870521050258702e-06, + "loss": 0.8389, + "step": 51965 + }, + { + "epoch": 3.34, + "learning_rate": 8.87030636123575e-06, + "loss": 0.8969, + "step": 51966 + }, + { + "epoch": 3.34, + "learning_rate": 8.870091672212801e-06, + "loss": 0.7626, + "step": 51967 + }, + { + "epoch": 3.34, + "learning_rate": 8.86987698318985e-06, + "loss": 0.8577, + "step": 51968 + }, + { + "epoch": 3.34, + "learning_rate": 8.869662294166899e-06, + "loss": 0.883, + "step": 51969 + }, + { + "epoch": 3.34, + "learning_rate": 8.86944760514395e-06, + "loss": 0.7842, + "step": 51970 + }, + { + "epoch": 3.34, + "learning_rate": 8.869232916121e-06, + "loss": 0.8606, + "step": 51971 + }, + { + "epoch": 3.34, + "learning_rate": 8.86901822709805e-06, + "loss": 0.8637, + "step": 51972 + }, + { + "epoch": 3.34, + "learning_rate": 8.868803538075098e-06, + "loss": 0.7464, + "step": 51973 + }, + { + "epoch": 3.34, + "learning_rate": 8.868588849052149e-06, + "loss": 0.8218, + "step": 51974 + }, + { + "epoch": 3.34, + "learning_rate": 8.868374160029198e-06, + "loss": 0.8274, + "step": 51975 + }, + { + "epoch": 3.34, + "learning_rate": 8.868159471006249e-06, + "loss": 0.8673, + "step": 51976 + }, + { + "epoch": 3.34, + "learning_rate": 8.867944781983298e-06, + "loss": 0.8688, + "step": 51977 + }, + { + "epoch": 3.34, + "learning_rate": 8.867730092960347e-06, + "loss": 0.8585, + "step": 51978 + }, + { + "epoch": 3.34, + "learning_rate": 8.867515403937397e-06, + "loss": 0.8273, + "step": 51979 + }, + { + "epoch": 3.34, + "learning_rate": 8.867300714914448e-06, + "loss": 0.8528, + "step": 51980 + }, + { + "epoch": 3.34, + "learning_rate": 8.867086025891497e-06, + "loss": 0.7407, + "step": 51981 + }, + { + "epoch": 3.34, + "learning_rate": 8.866871336868546e-06, + "loss": 0.8578, + "step": 51982 + }, + { + "epoch": 3.34, + "learning_rate": 8.866656647845597e-06, + "loss": 0.7889, + "step": 51983 + }, + { + "epoch": 3.34, + "learning_rate": 8.866441958822645e-06, + "loss": 0.7941, + "step": 51984 + }, + { + "epoch": 3.34, + "learning_rate": 8.866227269799696e-06, + "loss": 0.9053, + "step": 51985 + }, + { + "epoch": 3.34, + "learning_rate": 8.866012580776747e-06, + "loss": 0.7923, + "step": 51986 + }, + { + "epoch": 3.34, + "learning_rate": 8.865797891753796e-06, + "loss": 0.8351, + "step": 51987 + }, + { + "epoch": 3.34, + "learning_rate": 8.865583202730845e-06, + "loss": 0.8101, + "step": 51988 + }, + { + "epoch": 3.34, + "learning_rate": 8.865368513707895e-06, + "loss": 0.8065, + "step": 51989 + }, + { + "epoch": 3.34, + "learning_rate": 8.865153824684944e-06, + "loss": 0.8336, + "step": 51990 + }, + { + "epoch": 3.34, + "learning_rate": 8.864939135661993e-06, + "loss": 0.9077, + "step": 51991 + }, + { + "epoch": 3.35, + "learning_rate": 8.864724446639044e-06, + "loss": 0.8098, + "step": 51992 + }, + { + "epoch": 3.35, + "learning_rate": 8.864509757616095e-06, + "loss": 0.8189, + "step": 51993 + }, + { + "epoch": 3.35, + "learning_rate": 8.864295068593144e-06, + "loss": 0.8439, + "step": 51994 + }, + { + "epoch": 3.35, + "learning_rate": 8.864080379570194e-06, + "loss": 0.8075, + "step": 51995 + }, + { + "epoch": 3.35, + "learning_rate": 8.863865690547243e-06, + "loss": 0.8204, + "step": 51996 + }, + { + "epoch": 3.35, + "learning_rate": 8.863651001524292e-06, + "loss": 0.8011, + "step": 51997 + }, + { + "epoch": 3.35, + "learning_rate": 8.863436312501343e-06, + "loss": 0.7761, + "step": 51998 + }, + { + "epoch": 3.35, + "learning_rate": 8.863221623478392e-06, + "loss": 0.8318, + "step": 51999 + }, + { + "epoch": 3.35, + "learning_rate": 8.86300693445544e-06, + "loss": 0.8359, + "step": 52000 + }, + { + "epoch": 3.35, + "eval_laion/cider": 0.5114318618076501, + "eval_laion/meteor": 0.3984655177429753, + "eval_laion/pred_num_words": 38.130859375, + "eval_laion/sacrebleu": 16.382924607209333, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.0545427799224854, + "eval_runtime": 99.8182, + "eval_samples_per_second": 4.999, + "eval_steps_per_second": 0.321, + "step": 52000 + }, + { + "epoch": 3.35, + "learning_rate": 8.862792245432491e-06, + "loss": 0.7806, + "step": 52001 + }, + { + "epoch": 3.35, + "learning_rate": 8.862577556409542e-06, + "loss": 0.8266, + "step": 52002 + }, + { + "epoch": 3.35, + "learning_rate": 8.862362867386591e-06, + "loss": 0.845, + "step": 52003 + }, + { + "epoch": 3.35, + "learning_rate": 8.862148178363642e-06, + "loss": 0.8491, + "step": 52004 + }, + { + "epoch": 3.35, + "learning_rate": 8.86193348934069e-06, + "loss": 0.8161, + "step": 52005 + }, + { + "epoch": 3.35, + "learning_rate": 8.86171880031774e-06, + "loss": 0.9092, + "step": 52006 + }, + { + "epoch": 3.35, + "learning_rate": 8.86150411129479e-06, + "loss": 0.8245, + "step": 52007 + }, + { + "epoch": 3.35, + "learning_rate": 8.861289422271841e-06, + "loss": 0.8562, + "step": 52008 + }, + { + "epoch": 3.35, + "learning_rate": 8.86107473324889e-06, + "loss": 0.8603, + "step": 52009 + }, + { + "epoch": 3.35, + "learning_rate": 8.860860044225939e-06, + "loss": 0.827, + "step": 52010 + }, + { + "epoch": 3.35, + "learning_rate": 8.86064535520299e-06, + "loss": 0.8092, + "step": 52011 + }, + { + "epoch": 3.35, + "learning_rate": 8.860430666180039e-06, + "loss": 0.804, + "step": 52012 + }, + { + "epoch": 3.35, + "learning_rate": 8.86021597715709e-06, + "loss": 0.8471, + "step": 52013 + }, + { + "epoch": 3.35, + "learning_rate": 8.860001288134138e-06, + "loss": 0.8807, + "step": 52014 + }, + { + "epoch": 3.35, + "learning_rate": 8.859786599111189e-06, + "loss": 0.8717, + "step": 52015 + }, + { + "epoch": 3.35, + "learning_rate": 8.859571910088238e-06, + "loss": 0.8473, + "step": 52016 + }, + { + "epoch": 3.35, + "learning_rate": 8.859357221065288e-06, + "loss": 0.8769, + "step": 52017 + }, + { + "epoch": 3.35, + "learning_rate": 8.859142532042337e-06, + "loss": 0.8569, + "step": 52018 + }, + { + "epoch": 3.35, + "learning_rate": 8.858927843019386e-06, + "loss": 0.8243, + "step": 52019 + }, + { + "epoch": 3.35, + "learning_rate": 8.858713153996437e-06, + "loss": 0.8527, + "step": 52020 + }, + { + "epoch": 3.35, + "learning_rate": 8.858498464973486e-06, + "loss": 0.8035, + "step": 52021 + }, + { + "epoch": 3.35, + "learning_rate": 8.858283775950537e-06, + "loss": 0.8579, + "step": 52022 + }, + { + "epoch": 3.35, + "learning_rate": 8.858069086927587e-06, + "loss": 0.7979, + "step": 52023 + }, + { + "epoch": 3.35, + "learning_rate": 8.857854397904636e-06, + "loss": 0.8332, + "step": 52024 + }, + { + "epoch": 3.35, + "learning_rate": 8.857639708881685e-06, + "loss": 0.8319, + "step": 52025 + }, + { + "epoch": 3.35, + "learning_rate": 8.857425019858736e-06, + "loss": 0.8335, + "step": 52026 + }, + { + "epoch": 3.35, + "learning_rate": 8.857210330835785e-06, + "loss": 0.7856, + "step": 52027 + }, + { + "epoch": 3.35, + "learning_rate": 8.856995641812834e-06, + "loss": 0.8376, + "step": 52028 + }, + { + "epoch": 3.35, + "learning_rate": 8.856780952789884e-06, + "loss": 0.8053, + "step": 52029 + }, + { + "epoch": 3.35, + "learning_rate": 8.856566263766935e-06, + "loss": 0.7581, + "step": 52030 + }, + { + "epoch": 3.35, + "learning_rate": 8.856351574743984e-06, + "loss": 0.8291, + "step": 52031 + }, + { + "epoch": 3.35, + "learning_rate": 8.856136885721035e-06, + "loss": 0.7913, + "step": 52032 + }, + { + "epoch": 3.35, + "learning_rate": 8.855922196698084e-06, + "loss": 0.8102, + "step": 52033 + }, + { + "epoch": 3.35, + "learning_rate": 8.855707507675133e-06, + "loss": 0.8512, + "step": 52034 + }, + { + "epoch": 3.35, + "learning_rate": 8.855492818652183e-06, + "loss": 0.8532, + "step": 52035 + }, + { + "epoch": 3.35, + "learning_rate": 8.855278129629232e-06, + "loss": 0.8195, + "step": 52036 + }, + { + "epoch": 3.35, + "learning_rate": 8.855063440606283e-06, + "loss": 0.7646, + "step": 52037 + }, + { + "epoch": 3.35, + "learning_rate": 8.854848751583332e-06, + "loss": 0.7838, + "step": 52038 + }, + { + "epoch": 3.35, + "learning_rate": 8.854634062560383e-06, + "loss": 0.8745, + "step": 52039 + }, + { + "epoch": 3.35, + "learning_rate": 8.854419373537432e-06, + "loss": 0.7983, + "step": 52040 + }, + { + "epoch": 3.35, + "learning_rate": 8.85420468451448e-06, + "loss": 0.8548, + "step": 52041 + }, + { + "epoch": 3.35, + "learning_rate": 8.853989995491531e-06, + "loss": 0.8034, + "step": 52042 + }, + { + "epoch": 3.35, + "learning_rate": 8.85377530646858e-06, + "loss": 0.7984, + "step": 52043 + }, + { + "epoch": 3.35, + "learning_rate": 8.85356061744563e-06, + "loss": 0.9221, + "step": 52044 + }, + { + "epoch": 3.35, + "learning_rate": 8.853345928422681e-06, + "loss": 0.8301, + "step": 52045 + }, + { + "epoch": 3.35, + "learning_rate": 8.85313123939973e-06, + "loss": 0.8448, + "step": 52046 + }, + { + "epoch": 3.35, + "learning_rate": 8.85291655037678e-06, + "loss": 0.8585, + "step": 52047 + }, + { + "epoch": 3.35, + "learning_rate": 8.85270186135383e-06, + "loss": 0.881, + "step": 52048 + }, + { + "epoch": 3.35, + "learning_rate": 8.852487172330879e-06, + "loss": 0.8068, + "step": 52049 + }, + { + "epoch": 3.35, + "learning_rate": 8.852272483307928e-06, + "loss": 0.8122, + "step": 52050 + }, + { + "epoch": 3.35, + "learning_rate": 8.852057794284979e-06, + "loss": 0.8269, + "step": 52051 + }, + { + "epoch": 3.35, + "learning_rate": 8.85184310526203e-06, + "loss": 0.7872, + "step": 52052 + }, + { + "epoch": 3.35, + "learning_rate": 8.851628416239078e-06, + "loss": 0.8295, + "step": 52053 + }, + { + "epoch": 3.35, + "learning_rate": 8.851413727216129e-06, + "loss": 0.8569, + "step": 52054 + }, + { + "epoch": 3.35, + "learning_rate": 8.851199038193178e-06, + "loss": 0.8319, + "step": 52055 + }, + { + "epoch": 3.35, + "learning_rate": 8.850984349170227e-06, + "loss": 0.865, + "step": 52056 + }, + { + "epoch": 3.35, + "learning_rate": 8.850769660147278e-06, + "loss": 0.7346, + "step": 52057 + }, + { + "epoch": 3.35, + "learning_rate": 8.850554971124326e-06, + "loss": 0.8084, + "step": 52058 + }, + { + "epoch": 3.35, + "learning_rate": 8.850340282101377e-06, + "loss": 0.7569, + "step": 52059 + }, + { + "epoch": 3.35, + "learning_rate": 8.850125593078426e-06, + "loss": 0.8112, + "step": 52060 + }, + { + "epoch": 3.35, + "learning_rate": 8.849910904055477e-06, + "loss": 0.8162, + "step": 52061 + }, + { + "epoch": 3.35, + "learning_rate": 8.849696215032526e-06, + "loss": 0.7789, + "step": 52062 + }, + { + "epoch": 3.35, + "learning_rate": 8.849481526009576e-06, + "loss": 0.8482, + "step": 52063 + }, + { + "epoch": 3.35, + "learning_rate": 8.849266836986625e-06, + "loss": 0.8338, + "step": 52064 + }, + { + "epoch": 3.35, + "learning_rate": 8.849052147963674e-06, + "loss": 0.7931, + "step": 52065 + }, + { + "epoch": 3.35, + "learning_rate": 8.848837458940725e-06, + "loss": 0.8075, + "step": 52066 + }, + { + "epoch": 3.35, + "learning_rate": 8.848622769917776e-06, + "loss": 0.8472, + "step": 52067 + }, + { + "epoch": 3.35, + "learning_rate": 8.848408080894825e-06, + "loss": 0.8438, + "step": 52068 + }, + { + "epoch": 3.35, + "learning_rate": 8.848193391871874e-06, + "loss": 0.7519, + "step": 52069 + }, + { + "epoch": 3.35, + "learning_rate": 8.847978702848924e-06, + "loss": 0.8178, + "step": 52070 + }, + { + "epoch": 3.35, + "learning_rate": 8.847764013825973e-06, + "loss": 0.8033, + "step": 52071 + }, + { + "epoch": 3.35, + "learning_rate": 8.847549324803024e-06, + "loss": 0.828, + "step": 52072 + }, + { + "epoch": 3.35, + "learning_rate": 8.847334635780075e-06, + "loss": 0.8548, + "step": 52073 + }, + { + "epoch": 3.35, + "learning_rate": 8.847119946757123e-06, + "loss": 0.7806, + "step": 52074 + }, + { + "epoch": 3.35, + "learning_rate": 8.846905257734172e-06, + "loss": 0.8396, + "step": 52075 + }, + { + "epoch": 3.35, + "learning_rate": 8.846690568711223e-06, + "loss": 0.8095, + "step": 52076 + }, + { + "epoch": 3.35, + "learning_rate": 8.846475879688272e-06, + "loss": 0.8288, + "step": 52077 + }, + { + "epoch": 3.35, + "learning_rate": 8.846261190665321e-06, + "loss": 0.8675, + "step": 52078 + }, + { + "epoch": 3.35, + "learning_rate": 8.846046501642372e-06, + "loss": 0.8366, + "step": 52079 + }, + { + "epoch": 3.35, + "learning_rate": 8.845831812619422e-06, + "loss": 0.8601, + "step": 52080 + }, + { + "epoch": 3.35, + "learning_rate": 8.845617123596471e-06, + "loss": 0.7998, + "step": 52081 + }, + { + "epoch": 3.35, + "learning_rate": 8.845402434573522e-06, + "loss": 0.8402, + "step": 52082 + }, + { + "epoch": 3.35, + "learning_rate": 8.845187745550571e-06, + "loss": 0.816, + "step": 52083 + }, + { + "epoch": 3.35, + "learning_rate": 8.84497305652762e-06, + "loss": 0.8005, + "step": 52084 + }, + { + "epoch": 3.35, + "learning_rate": 8.84475836750467e-06, + "loss": 0.8379, + "step": 52085 + }, + { + "epoch": 3.35, + "learning_rate": 8.84454367848172e-06, + "loss": 0.8165, + "step": 52086 + }, + { + "epoch": 3.35, + "learning_rate": 8.844328989458768e-06, + "loss": 0.768, + "step": 52087 + }, + { + "epoch": 3.35, + "learning_rate": 8.844114300435819e-06, + "loss": 0.8418, + "step": 52088 + }, + { + "epoch": 3.35, + "learning_rate": 8.84389961141287e-06, + "loss": 0.8454, + "step": 52089 + }, + { + "epoch": 3.35, + "learning_rate": 8.843684922389919e-06, + "loss": 0.8688, + "step": 52090 + }, + { + "epoch": 3.35, + "learning_rate": 8.84347023336697e-06, + "loss": 0.8265, + "step": 52091 + }, + { + "epoch": 3.35, + "learning_rate": 8.843255544344018e-06, + "loss": 0.7911, + "step": 52092 + }, + { + "epoch": 3.35, + "learning_rate": 8.843040855321067e-06, + "loss": 0.7919, + "step": 52093 + }, + { + "epoch": 3.35, + "learning_rate": 8.842826166298118e-06, + "loss": 0.8195, + "step": 52094 + }, + { + "epoch": 3.35, + "learning_rate": 8.842611477275169e-06, + "loss": 0.8033, + "step": 52095 + }, + { + "epoch": 3.35, + "learning_rate": 8.842396788252218e-06, + "loss": 0.8002, + "step": 52096 + }, + { + "epoch": 3.35, + "learning_rate": 8.842182099229267e-06, + "loss": 0.8263, + "step": 52097 + }, + { + "epoch": 3.35, + "learning_rate": 8.841967410206317e-06, + "loss": 0.8608, + "step": 52098 + }, + { + "epoch": 3.35, + "learning_rate": 8.841752721183366e-06, + "loss": 0.8291, + "step": 52099 + }, + { + "epoch": 3.35, + "learning_rate": 8.841538032160417e-06, + "loss": 0.8484, + "step": 52100 + }, + { + "epoch": 3.35, + "learning_rate": 8.841323343137466e-06, + "loss": 0.8248, + "step": 52101 + }, + { + "epoch": 3.35, + "learning_rate": 8.841108654114517e-06, + "loss": 0.7661, + "step": 52102 + }, + { + "epoch": 3.35, + "learning_rate": 8.840893965091565e-06, + "loss": 0.7551, + "step": 52103 + }, + { + "epoch": 3.35, + "learning_rate": 8.840679276068616e-06, + "loss": 0.8106, + "step": 52104 + }, + { + "epoch": 3.35, + "learning_rate": 8.840464587045665e-06, + "loss": 0.8369, + "step": 52105 + }, + { + "epoch": 3.35, + "learning_rate": 8.840249898022714e-06, + "loss": 0.8303, + "step": 52106 + }, + { + "epoch": 3.35, + "learning_rate": 8.840035208999765e-06, + "loss": 0.781, + "step": 52107 + }, + { + "epoch": 3.35, + "learning_rate": 8.839820519976814e-06, + "loss": 0.7836, + "step": 52108 + }, + { + "epoch": 3.35, + "learning_rate": 8.839605830953864e-06, + "loss": 0.7661, + "step": 52109 + }, + { + "epoch": 3.35, + "learning_rate": 8.839391141930915e-06, + "loss": 0.8196, + "step": 52110 + }, + { + "epoch": 3.35, + "learning_rate": 8.839176452907964e-06, + "loss": 0.8414, + "step": 52111 + }, + { + "epoch": 3.35, + "learning_rate": 8.838961763885013e-06, + "loss": 0.8371, + "step": 52112 + }, + { + "epoch": 3.35, + "learning_rate": 8.838747074862064e-06, + "loss": 0.7999, + "step": 52113 + }, + { + "epoch": 3.35, + "learning_rate": 8.838532385839113e-06, + "loss": 0.8078, + "step": 52114 + }, + { + "epoch": 3.35, + "learning_rate": 8.838317696816162e-06, + "loss": 0.8018, + "step": 52115 + }, + { + "epoch": 3.35, + "learning_rate": 8.838103007793212e-06, + "loss": 0.8219, + "step": 52116 + }, + { + "epoch": 3.35, + "learning_rate": 8.837888318770263e-06, + "loss": 0.8021, + "step": 52117 + }, + { + "epoch": 3.35, + "learning_rate": 8.837673629747312e-06, + "loss": 0.7916, + "step": 52118 + }, + { + "epoch": 3.35, + "learning_rate": 8.83745894072436e-06, + "loss": 0.8001, + "step": 52119 + }, + { + "epoch": 3.35, + "learning_rate": 8.837244251701411e-06, + "loss": 0.7692, + "step": 52120 + }, + { + "epoch": 3.35, + "learning_rate": 8.83702956267846e-06, + "loss": 0.7929, + "step": 52121 + }, + { + "epoch": 3.35, + "learning_rate": 8.836814873655511e-06, + "loss": 0.8042, + "step": 52122 + }, + { + "epoch": 3.35, + "learning_rate": 8.83660018463256e-06, + "loss": 0.8097, + "step": 52123 + }, + { + "epoch": 3.35, + "learning_rate": 8.83638549560961e-06, + "loss": 0.8597, + "step": 52124 + }, + { + "epoch": 3.35, + "learning_rate": 8.83617080658666e-06, + "loss": 0.8652, + "step": 52125 + }, + { + "epoch": 3.35, + "learning_rate": 8.83595611756371e-06, + "loss": 0.8266, + "step": 52126 + }, + { + "epoch": 3.35, + "learning_rate": 8.83574142854076e-06, + "loss": 0.8077, + "step": 52127 + }, + { + "epoch": 3.35, + "learning_rate": 8.835526739517808e-06, + "loss": 0.8356, + "step": 52128 + }, + { + "epoch": 3.35, + "learning_rate": 8.835312050494859e-06, + "loss": 0.8467, + "step": 52129 + }, + { + "epoch": 3.35, + "learning_rate": 8.835097361471908e-06, + "loss": 0.8894, + "step": 52130 + }, + { + "epoch": 3.35, + "learning_rate": 8.834882672448959e-06, + "loss": 0.7656, + "step": 52131 + }, + { + "epoch": 3.35, + "learning_rate": 8.83466798342601e-06, + "loss": 0.8368, + "step": 52132 + }, + { + "epoch": 3.35, + "learning_rate": 8.834453294403058e-06, + "loss": 0.8947, + "step": 52133 + }, + { + "epoch": 3.35, + "learning_rate": 8.834238605380107e-06, + "loss": 0.778, + "step": 52134 + }, + { + "epoch": 3.35, + "learning_rate": 8.834023916357158e-06, + "loss": 0.8324, + "step": 52135 + }, + { + "epoch": 3.35, + "learning_rate": 8.833809227334207e-06, + "loss": 0.791, + "step": 52136 + }, + { + "epoch": 3.35, + "learning_rate": 8.833594538311256e-06, + "loss": 0.9062, + "step": 52137 + }, + { + "epoch": 3.35, + "learning_rate": 8.833379849288306e-06, + "loss": 0.7818, + "step": 52138 + }, + { + "epoch": 3.35, + "learning_rate": 8.833165160265357e-06, + "loss": 0.8154, + "step": 52139 + }, + { + "epoch": 3.35, + "learning_rate": 8.832950471242406e-06, + "loss": 0.8175, + "step": 52140 + }, + { + "epoch": 3.35, + "learning_rate": 8.832735782219457e-06, + "loss": 0.8774, + "step": 52141 + }, + { + "epoch": 3.35, + "learning_rate": 8.832521093196506e-06, + "loss": 0.8182, + "step": 52142 + }, + { + "epoch": 3.35, + "learning_rate": 8.832306404173555e-06, + "loss": 0.84, + "step": 52143 + }, + { + "epoch": 3.35, + "learning_rate": 8.832091715150605e-06, + "loss": 0.8301, + "step": 52144 + }, + { + "epoch": 3.35, + "learning_rate": 8.831877026127654e-06, + "loss": 0.8132, + "step": 52145 + }, + { + "epoch": 3.35, + "learning_rate": 8.831662337104705e-06, + "loss": 0.8039, + "step": 52146 + }, + { + "epoch": 3.36, + "learning_rate": 8.831447648081754e-06, + "loss": 0.8031, + "step": 52147 + }, + { + "epoch": 3.36, + "learning_rate": 8.831232959058804e-06, + "loss": 0.8107, + "step": 52148 + }, + { + "epoch": 3.36, + "learning_rate": 8.831018270035853e-06, + "loss": 0.8136, + "step": 52149 + }, + { + "epoch": 3.36, + "learning_rate": 8.830803581012904e-06, + "loss": 0.809, + "step": 52150 + }, + { + "epoch": 3.36, + "learning_rate": 8.830588891989953e-06, + "loss": 0.7707, + "step": 52151 + }, + { + "epoch": 3.36, + "learning_rate": 8.830374202967002e-06, + "loss": 0.8821, + "step": 52152 + }, + { + "epoch": 3.36, + "learning_rate": 8.830159513944053e-06, + "loss": 0.8044, + "step": 52153 + }, + { + "epoch": 3.36, + "learning_rate": 8.829944824921103e-06, + "loss": 0.8628, + "step": 52154 + }, + { + "epoch": 3.36, + "learning_rate": 8.829730135898152e-06, + "loss": 0.7954, + "step": 52155 + }, + { + "epoch": 3.36, + "learning_rate": 8.829515446875201e-06, + "loss": 0.8238, + "step": 52156 + }, + { + "epoch": 3.36, + "learning_rate": 8.829300757852252e-06, + "loss": 0.7466, + "step": 52157 + }, + { + "epoch": 3.36, + "learning_rate": 8.829086068829301e-06, + "loss": 0.8639, + "step": 52158 + }, + { + "epoch": 3.36, + "learning_rate": 8.828871379806352e-06, + "loss": 0.7915, + "step": 52159 + }, + { + "epoch": 3.36, + "learning_rate": 8.828656690783402e-06, + "loss": 0.8659, + "step": 52160 + }, + { + "epoch": 3.36, + "learning_rate": 8.828442001760451e-06, + "loss": 0.8511, + "step": 52161 + }, + { + "epoch": 3.36, + "learning_rate": 8.8282273127375e-06, + "loss": 0.7785, + "step": 52162 + }, + { + "epoch": 3.36, + "learning_rate": 8.82801262371455e-06, + "loss": 0.8228, + "step": 52163 + }, + { + "epoch": 3.36, + "learning_rate": 8.8277979346916e-06, + "loss": 0.8238, + "step": 52164 + }, + { + "epoch": 3.36, + "learning_rate": 8.827583245668649e-06, + "loss": 0.8197, + "step": 52165 + }, + { + "epoch": 3.36, + "learning_rate": 8.8273685566457e-06, + "loss": 0.7806, + "step": 52166 + }, + { + "epoch": 3.36, + "learning_rate": 8.827153867622748e-06, + "loss": 0.8183, + "step": 52167 + }, + { + "epoch": 3.36, + "learning_rate": 8.826939178599799e-06, + "loss": 0.7756, + "step": 52168 + }, + { + "epoch": 3.36, + "learning_rate": 8.82672448957685e-06, + "loss": 0.8117, + "step": 52169 + }, + { + "epoch": 3.36, + "learning_rate": 8.826509800553899e-06, + "loss": 0.8766, + "step": 52170 + }, + { + "epoch": 3.36, + "learning_rate": 8.826295111530948e-06, + "loss": 0.8204, + "step": 52171 + }, + { + "epoch": 3.36, + "learning_rate": 8.826080422507998e-06, + "loss": 0.8216, + "step": 52172 + }, + { + "epoch": 3.36, + "learning_rate": 8.825865733485047e-06, + "loss": 0.8148, + "step": 52173 + }, + { + "epoch": 3.36, + "learning_rate": 8.825651044462096e-06, + "loss": 0.8697, + "step": 52174 + }, + { + "epoch": 3.36, + "learning_rate": 8.825436355439147e-06, + "loss": 0.8475, + "step": 52175 + }, + { + "epoch": 3.36, + "learning_rate": 8.825221666416198e-06, + "loss": 0.7879, + "step": 52176 + }, + { + "epoch": 3.36, + "learning_rate": 8.825006977393246e-06, + "loss": 0.8259, + "step": 52177 + }, + { + "epoch": 3.36, + "learning_rate": 8.824792288370297e-06, + "loss": 0.8596, + "step": 52178 + }, + { + "epoch": 3.36, + "learning_rate": 8.824577599347346e-06, + "loss": 0.7401, + "step": 52179 + }, + { + "epoch": 3.36, + "learning_rate": 8.824362910324395e-06, + "loss": 0.8664, + "step": 52180 + }, + { + "epoch": 3.36, + "learning_rate": 8.824148221301446e-06, + "loss": 0.7835, + "step": 52181 + }, + { + "epoch": 3.36, + "learning_rate": 8.823933532278496e-06, + "loss": 0.7969, + "step": 52182 + }, + { + "epoch": 3.36, + "learning_rate": 8.823718843255545e-06, + "loss": 0.8473, + "step": 52183 + }, + { + "epoch": 3.36, + "learning_rate": 8.823504154232594e-06, + "loss": 0.8225, + "step": 52184 + }, + { + "epoch": 3.36, + "learning_rate": 8.823289465209645e-06, + "loss": 0.8327, + "step": 52185 + }, + { + "epoch": 3.36, + "learning_rate": 8.823074776186694e-06, + "loss": 0.8528, + "step": 52186 + }, + { + "epoch": 3.36, + "learning_rate": 8.822860087163745e-06, + "loss": 0.8307, + "step": 52187 + }, + { + "epoch": 3.36, + "learning_rate": 8.822645398140794e-06, + "loss": 0.8784, + "step": 52188 + }, + { + "epoch": 3.36, + "learning_rate": 8.822430709117844e-06, + "loss": 0.7945, + "step": 52189 + }, + { + "epoch": 3.36, + "learning_rate": 8.822216020094893e-06, + "loss": 0.7944, + "step": 52190 + }, + { + "epoch": 3.36, + "learning_rate": 8.822001331071944e-06, + "loss": 0.8173, + "step": 52191 + }, + { + "epoch": 3.36, + "learning_rate": 8.821786642048993e-06, + "loss": 0.8665, + "step": 52192 + }, + { + "epoch": 3.36, + "learning_rate": 8.821571953026042e-06, + "loss": 0.8231, + "step": 52193 + }, + { + "epoch": 3.36, + "learning_rate": 8.821357264003092e-06, + "loss": 0.8157, + "step": 52194 + }, + { + "epoch": 3.36, + "learning_rate": 8.821142574980141e-06, + "loss": 0.8686, + "step": 52195 + }, + { + "epoch": 3.36, + "learning_rate": 8.82092788595719e-06, + "loss": 0.9064, + "step": 52196 + }, + { + "epoch": 3.36, + "learning_rate": 8.820713196934243e-06, + "loss": 0.8798, + "step": 52197 + }, + { + "epoch": 3.36, + "learning_rate": 8.820498507911292e-06, + "loss": 0.8206, + "step": 52198 + }, + { + "epoch": 3.36, + "learning_rate": 8.82028381888834e-06, + "loss": 0.8325, + "step": 52199 + }, + { + "epoch": 3.36, + "learning_rate": 8.820069129865391e-06, + "loss": 0.8275, + "step": 52200 + }, + { + "epoch": 3.36, + "learning_rate": 8.81985444084244e-06, + "loss": 0.8156, + "step": 52201 + }, + { + "epoch": 3.36, + "learning_rate": 8.81963975181949e-06, + "loss": 0.8138, + "step": 52202 + }, + { + "epoch": 3.36, + "learning_rate": 8.81942506279654e-06, + "loss": 0.8295, + "step": 52203 + }, + { + "epoch": 3.36, + "learning_rate": 8.81921037377359e-06, + "loss": 0.7975, + "step": 52204 + }, + { + "epoch": 3.36, + "learning_rate": 8.81899568475064e-06, + "loss": 0.853, + "step": 52205 + }, + { + "epoch": 3.36, + "learning_rate": 8.818780995727688e-06, + "loss": 0.7905, + "step": 52206 + }, + { + "epoch": 3.36, + "learning_rate": 8.818566306704739e-06, + "loss": 0.8735, + "step": 52207 + }, + { + "epoch": 3.36, + "learning_rate": 8.818351617681788e-06, + "loss": 0.8444, + "step": 52208 + }, + { + "epoch": 3.36, + "learning_rate": 8.818136928658839e-06, + "loss": 0.8277, + "step": 52209 + }, + { + "epoch": 3.36, + "learning_rate": 8.817922239635888e-06, + "loss": 0.8899, + "step": 52210 + }, + { + "epoch": 3.36, + "learning_rate": 8.817707550612938e-06, + "loss": 0.8448, + "step": 52211 + }, + { + "epoch": 3.36, + "learning_rate": 8.817492861589987e-06, + "loss": 0.7795, + "step": 52212 + }, + { + "epoch": 3.36, + "learning_rate": 8.817278172567038e-06, + "loss": 0.8762, + "step": 52213 + }, + { + "epoch": 3.36, + "learning_rate": 8.817063483544087e-06, + "loss": 0.7944, + "step": 52214 + }, + { + "epoch": 3.36, + "learning_rate": 8.816848794521136e-06, + "loss": 0.7811, + "step": 52215 + }, + { + "epoch": 3.36, + "learning_rate": 8.816634105498187e-06, + "loss": 0.8584, + "step": 52216 + }, + { + "epoch": 3.36, + "learning_rate": 8.816419416475236e-06, + "loss": 0.8589, + "step": 52217 + }, + { + "epoch": 3.36, + "learning_rate": 8.816204727452286e-06, + "loss": 0.848, + "step": 52218 + }, + { + "epoch": 3.36, + "learning_rate": 8.815990038429337e-06, + "loss": 0.7931, + "step": 52219 + }, + { + "epoch": 3.36, + "learning_rate": 8.815775349406386e-06, + "loss": 0.7919, + "step": 52220 + }, + { + "epoch": 3.36, + "learning_rate": 8.815560660383435e-06, + "loss": 0.838, + "step": 52221 + }, + { + "epoch": 3.36, + "learning_rate": 8.815345971360485e-06, + "loss": 0.8215, + "step": 52222 + }, + { + "epoch": 3.36, + "learning_rate": 8.815131282337534e-06, + "loss": 0.767, + "step": 52223 + }, + { + "epoch": 3.36, + "learning_rate": 8.814916593314583e-06, + "loss": 0.7946, + "step": 52224 + }, + { + "epoch": 3.36, + "learning_rate": 8.814701904291634e-06, + "loss": 0.8145, + "step": 52225 + }, + { + "epoch": 3.36, + "learning_rate": 8.814487215268685e-06, + "loss": 0.8161, + "step": 52226 + }, + { + "epoch": 3.36, + "learning_rate": 8.814272526245734e-06, + "loss": 0.826, + "step": 52227 + }, + { + "epoch": 3.36, + "learning_rate": 8.814057837222784e-06, + "loss": 0.836, + "step": 52228 + }, + { + "epoch": 3.36, + "learning_rate": 8.813843148199833e-06, + "loss": 0.794, + "step": 52229 + }, + { + "epoch": 3.36, + "learning_rate": 8.813628459176882e-06, + "loss": 0.8079, + "step": 52230 + }, + { + "epoch": 3.36, + "learning_rate": 8.813413770153933e-06, + "loss": 0.8313, + "step": 52231 + }, + { + "epoch": 3.36, + "learning_rate": 8.813199081130982e-06, + "loss": 0.7666, + "step": 52232 + }, + { + "epoch": 3.36, + "learning_rate": 8.812984392108033e-06, + "loss": 0.8238, + "step": 52233 + }, + { + "epoch": 3.36, + "learning_rate": 8.812769703085082e-06, + "loss": 0.8229, + "step": 52234 + }, + { + "epoch": 3.36, + "learning_rate": 8.812555014062132e-06, + "loss": 0.8207, + "step": 52235 + }, + { + "epoch": 3.36, + "learning_rate": 8.812340325039181e-06, + "loss": 0.8483, + "step": 52236 + }, + { + "epoch": 3.36, + "learning_rate": 8.812125636016232e-06, + "loss": 0.8214, + "step": 52237 + }, + { + "epoch": 3.36, + "learning_rate": 8.81191094699328e-06, + "loss": 0.8111, + "step": 52238 + }, + { + "epoch": 3.36, + "learning_rate": 8.81169625797033e-06, + "loss": 0.8525, + "step": 52239 + }, + { + "epoch": 3.36, + "learning_rate": 8.81148156894738e-06, + "loss": 0.8165, + "step": 52240 + }, + { + "epoch": 3.36, + "learning_rate": 8.811266879924431e-06, + "loss": 0.8751, + "step": 52241 + }, + { + "epoch": 3.36, + "learning_rate": 8.81105219090148e-06, + "loss": 0.8566, + "step": 52242 + }, + { + "epoch": 3.36, + "learning_rate": 8.810837501878529e-06, + "loss": 0.7674, + "step": 52243 + }, + { + "epoch": 3.36, + "learning_rate": 8.81062281285558e-06, + "loss": 0.8728, + "step": 52244 + }, + { + "epoch": 3.36, + "learning_rate": 8.810408123832629e-06, + "loss": 0.7959, + "step": 52245 + }, + { + "epoch": 3.36, + "learning_rate": 8.81019343480968e-06, + "loss": 0.8775, + "step": 52246 + }, + { + "epoch": 3.36, + "learning_rate": 8.80997874578673e-06, + "loss": 0.8326, + "step": 52247 + }, + { + "epoch": 3.36, + "learning_rate": 8.809764056763779e-06, + "loss": 0.915, + "step": 52248 + }, + { + "epoch": 3.36, + "learning_rate": 8.809549367740828e-06, + "loss": 0.7918, + "step": 52249 + }, + { + "epoch": 3.36, + "learning_rate": 8.809334678717879e-06, + "loss": 0.8067, + "step": 52250 + }, + { + "epoch": 3.36, + "learning_rate": 8.809119989694927e-06, + "loss": 0.8253, + "step": 52251 + }, + { + "epoch": 3.36, + "learning_rate": 8.808905300671976e-06, + "loss": 0.7521, + "step": 52252 + }, + { + "epoch": 3.36, + "learning_rate": 8.808690611649027e-06, + "loss": 0.8168, + "step": 52253 + }, + { + "epoch": 3.36, + "learning_rate": 8.808475922626076e-06, + "loss": 0.8779, + "step": 52254 + }, + { + "epoch": 3.36, + "learning_rate": 8.808261233603127e-06, + "loss": 0.8656, + "step": 52255 + }, + { + "epoch": 3.36, + "learning_rate": 8.808046544580177e-06, + "loss": 0.8155, + "step": 52256 + }, + { + "epoch": 3.36, + "learning_rate": 8.807831855557226e-06, + "loss": 0.8246, + "step": 52257 + }, + { + "epoch": 3.36, + "learning_rate": 8.807617166534275e-06, + "loss": 0.8693, + "step": 52258 + }, + { + "epoch": 3.36, + "learning_rate": 8.807402477511326e-06, + "loss": 0.8507, + "step": 52259 + }, + { + "epoch": 3.36, + "learning_rate": 8.807187788488375e-06, + "loss": 0.8107, + "step": 52260 + }, + { + "epoch": 3.36, + "learning_rate": 8.806973099465424e-06, + "loss": 0.8649, + "step": 52261 + }, + { + "epoch": 3.36, + "learning_rate": 8.806758410442475e-06, + "loss": 0.8058, + "step": 52262 + }, + { + "epoch": 3.36, + "learning_rate": 8.806543721419525e-06, + "loss": 0.7501, + "step": 52263 + }, + { + "epoch": 3.36, + "learning_rate": 8.806329032396574e-06, + "loss": 0.8105, + "step": 52264 + }, + { + "epoch": 3.36, + "learning_rate": 8.806114343373625e-06, + "loss": 0.8364, + "step": 52265 + }, + { + "epoch": 3.36, + "learning_rate": 8.805899654350674e-06, + "loss": 0.8761, + "step": 52266 + }, + { + "epoch": 3.36, + "learning_rate": 8.805684965327723e-06, + "loss": 0.8125, + "step": 52267 + }, + { + "epoch": 3.36, + "learning_rate": 8.805470276304773e-06, + "loss": 0.8797, + "step": 52268 + }, + { + "epoch": 3.36, + "learning_rate": 8.805255587281824e-06, + "loss": 0.8122, + "step": 52269 + }, + { + "epoch": 3.36, + "learning_rate": 8.805040898258873e-06, + "loss": 0.7964, + "step": 52270 + }, + { + "epoch": 3.36, + "learning_rate": 8.804826209235922e-06, + "loss": 0.8572, + "step": 52271 + }, + { + "epoch": 3.36, + "learning_rate": 8.804611520212973e-06, + "loss": 0.7827, + "step": 52272 + }, + { + "epoch": 3.36, + "learning_rate": 8.804396831190022e-06, + "loss": 0.8213, + "step": 52273 + }, + { + "epoch": 3.36, + "learning_rate": 8.80418214216707e-06, + "loss": 0.8705, + "step": 52274 + }, + { + "epoch": 3.36, + "learning_rate": 8.803967453144121e-06, + "loss": 0.786, + "step": 52275 + }, + { + "epoch": 3.36, + "learning_rate": 8.803752764121172e-06, + "loss": 0.8645, + "step": 52276 + }, + { + "epoch": 3.36, + "learning_rate": 8.803538075098221e-06, + "loss": 0.85, + "step": 52277 + }, + { + "epoch": 3.36, + "learning_rate": 8.803323386075272e-06, + "loss": 0.8586, + "step": 52278 + }, + { + "epoch": 3.36, + "learning_rate": 8.80310869705232e-06, + "loss": 0.9021, + "step": 52279 + }, + { + "epoch": 3.36, + "learning_rate": 8.80289400802937e-06, + "loss": 0.8383, + "step": 52280 + }, + { + "epoch": 3.36, + "learning_rate": 8.80267931900642e-06, + "loss": 0.854, + "step": 52281 + }, + { + "epoch": 3.36, + "learning_rate": 8.802464629983469e-06, + "loss": 0.8056, + "step": 52282 + }, + { + "epoch": 3.36, + "learning_rate": 8.802249940960518e-06, + "loss": 0.8661, + "step": 52283 + }, + { + "epoch": 3.36, + "learning_rate": 8.802035251937569e-06, + "loss": 0.8586, + "step": 52284 + }, + { + "epoch": 3.36, + "learning_rate": 8.80182056291462e-06, + "loss": 0.8737, + "step": 52285 + }, + { + "epoch": 3.36, + "learning_rate": 8.801605873891668e-06, + "loss": 0.803, + "step": 52286 + }, + { + "epoch": 3.36, + "learning_rate": 8.801391184868719e-06, + "loss": 0.7891, + "step": 52287 + }, + { + "epoch": 3.36, + "learning_rate": 8.801176495845768e-06, + "loss": 0.7852, + "step": 52288 + }, + { + "epoch": 3.36, + "learning_rate": 8.800961806822817e-06, + "loss": 0.8934, + "step": 52289 + }, + { + "epoch": 3.36, + "learning_rate": 8.800747117799868e-06, + "loss": 0.7898, + "step": 52290 + }, + { + "epoch": 3.36, + "learning_rate": 8.800532428776918e-06, + "loss": 0.8286, + "step": 52291 + }, + { + "epoch": 3.36, + "learning_rate": 8.800317739753967e-06, + "loss": 0.7674, + "step": 52292 + }, + { + "epoch": 3.36, + "learning_rate": 8.800103050731016e-06, + "loss": 0.8645, + "step": 52293 + }, + { + "epoch": 3.36, + "learning_rate": 8.799888361708067e-06, + "loss": 0.7913, + "step": 52294 + }, + { + "epoch": 3.36, + "learning_rate": 8.799673672685116e-06, + "loss": 0.8691, + "step": 52295 + }, + { + "epoch": 3.36, + "learning_rate": 8.799458983662166e-06, + "loss": 0.7617, + "step": 52296 + }, + { + "epoch": 3.36, + "learning_rate": 8.799244294639215e-06, + "loss": 0.7526, + "step": 52297 + }, + { + "epoch": 3.36, + "learning_rate": 8.799029605616266e-06, + "loss": 0.8196, + "step": 52298 + }, + { + "epoch": 3.36, + "learning_rate": 8.798814916593315e-06, + "loss": 0.8643, + "step": 52299 + }, + { + "epoch": 3.36, + "learning_rate": 8.798600227570366e-06, + "loss": 0.849, + "step": 52300 + }, + { + "epoch": 3.36, + "learning_rate": 8.798385538547415e-06, + "loss": 0.7635, + "step": 52301 + }, + { + "epoch": 3.36, + "learning_rate": 8.798170849524464e-06, + "loss": 0.8641, + "step": 52302 + }, + { + "epoch": 3.37, + "learning_rate": 8.797956160501514e-06, + "loss": 0.8208, + "step": 52303 + }, + { + "epoch": 3.37, + "learning_rate": 8.797741471478563e-06, + "loss": 0.8347, + "step": 52304 + }, + { + "epoch": 3.37, + "learning_rate": 8.797526782455614e-06, + "loss": 0.8364, + "step": 52305 + }, + { + "epoch": 3.37, + "learning_rate": 8.797312093432665e-06, + "loss": 0.7978, + "step": 52306 + }, + { + "epoch": 3.37, + "learning_rate": 8.797097404409714e-06, + "loss": 0.8109, + "step": 52307 + }, + { + "epoch": 3.37, + "learning_rate": 8.796882715386763e-06, + "loss": 0.8308, + "step": 52308 + }, + { + "epoch": 3.37, + "learning_rate": 8.796668026363813e-06, + "loss": 0.8201, + "step": 52309 + }, + { + "epoch": 3.37, + "learning_rate": 8.796453337340862e-06, + "loss": 0.8366, + "step": 52310 + }, + { + "epoch": 3.37, + "learning_rate": 8.796238648317911e-06, + "loss": 0.7674, + "step": 52311 + }, + { + "epoch": 3.37, + "learning_rate": 8.796023959294962e-06, + "loss": 0.8118, + "step": 52312 + }, + { + "epoch": 3.37, + "learning_rate": 8.795809270272012e-06, + "loss": 0.7891, + "step": 52313 + }, + { + "epoch": 3.37, + "learning_rate": 8.795594581249061e-06, + "loss": 0.8864, + "step": 52314 + }, + { + "epoch": 3.37, + "learning_rate": 8.795379892226112e-06, + "loss": 0.8353, + "step": 52315 + }, + { + "epoch": 3.37, + "learning_rate": 8.795165203203161e-06, + "loss": 0.8877, + "step": 52316 + }, + { + "epoch": 3.37, + "learning_rate": 8.79495051418021e-06, + "loss": 0.8557, + "step": 52317 + }, + { + "epoch": 3.37, + "learning_rate": 8.79473582515726e-06, + "loss": 0.7986, + "step": 52318 + }, + { + "epoch": 3.37, + "learning_rate": 8.79452113613431e-06, + "loss": 0.7892, + "step": 52319 + }, + { + "epoch": 3.37, + "learning_rate": 8.79430644711136e-06, + "loss": 0.858, + "step": 52320 + }, + { + "epoch": 3.37, + "learning_rate": 8.79409175808841e-06, + "loss": 0.8327, + "step": 52321 + }, + { + "epoch": 3.37, + "learning_rate": 8.79387706906546e-06, + "loss": 0.8188, + "step": 52322 + }, + { + "epoch": 3.37, + "learning_rate": 8.793662380042509e-06, + "loss": 0.8369, + "step": 52323 + }, + { + "epoch": 3.37, + "learning_rate": 8.79344769101956e-06, + "loss": 0.8091, + "step": 52324 + }, + { + "epoch": 3.37, + "learning_rate": 8.793233001996608e-06, + "loss": 0.7387, + "step": 52325 + }, + { + "epoch": 3.37, + "learning_rate": 8.793018312973657e-06, + "loss": 0.9149, + "step": 52326 + }, + { + "epoch": 3.37, + "learning_rate": 8.792803623950708e-06, + "loss": 0.867, + "step": 52327 + }, + { + "epoch": 3.37, + "learning_rate": 8.792588934927759e-06, + "loss": 0.8089, + "step": 52328 + }, + { + "epoch": 3.37, + "learning_rate": 8.792374245904808e-06, + "loss": 0.7972, + "step": 52329 + }, + { + "epoch": 3.37, + "learning_rate": 8.792159556881857e-06, + "loss": 0.7682, + "step": 52330 + }, + { + "epoch": 3.37, + "learning_rate": 8.791944867858907e-06, + "loss": 0.8267, + "step": 52331 + }, + { + "epoch": 3.37, + "learning_rate": 8.791730178835956e-06, + "loss": 0.8358, + "step": 52332 + }, + { + "epoch": 3.37, + "learning_rate": 8.791515489813007e-06, + "loss": 0.8884, + "step": 52333 + }, + { + "epoch": 3.37, + "learning_rate": 8.791300800790058e-06, + "loss": 0.8037, + "step": 52334 + }, + { + "epoch": 3.37, + "learning_rate": 8.791086111767107e-06, + "loss": 0.7966, + "step": 52335 + }, + { + "epoch": 3.37, + "learning_rate": 8.790871422744156e-06, + "loss": 0.8095, + "step": 52336 + }, + { + "epoch": 3.37, + "learning_rate": 8.790656733721206e-06, + "loss": 0.7425, + "step": 52337 + }, + { + "epoch": 3.37, + "learning_rate": 8.790442044698255e-06, + "loss": 0.7948, + "step": 52338 + }, + { + "epoch": 3.37, + "learning_rate": 8.790227355675304e-06, + "loss": 0.8586, + "step": 52339 + }, + { + "epoch": 3.37, + "learning_rate": 8.790012666652355e-06, + "loss": 0.8238, + "step": 52340 + }, + { + "epoch": 3.37, + "learning_rate": 8.789797977629404e-06, + "loss": 0.83, + "step": 52341 + }, + { + "epoch": 3.37, + "learning_rate": 8.789583288606454e-06, + "loss": 0.8102, + "step": 52342 + }, + { + "epoch": 3.37, + "learning_rate": 8.789368599583505e-06, + "loss": 0.8408, + "step": 52343 + }, + { + "epoch": 3.37, + "learning_rate": 8.789153910560554e-06, + "loss": 0.8663, + "step": 52344 + }, + { + "epoch": 3.37, + "learning_rate": 8.788939221537603e-06, + "loss": 0.8833, + "step": 52345 + }, + { + "epoch": 3.37, + "learning_rate": 8.788724532514654e-06, + "loss": 0.8348, + "step": 52346 + }, + { + "epoch": 3.37, + "learning_rate": 8.788509843491703e-06, + "loss": 0.8108, + "step": 52347 + }, + { + "epoch": 3.37, + "learning_rate": 8.788295154468752e-06, + "loss": 0.8386, + "step": 52348 + }, + { + "epoch": 3.37, + "learning_rate": 8.788080465445802e-06, + "loss": 0.8247, + "step": 52349 + }, + { + "epoch": 3.37, + "learning_rate": 8.787865776422853e-06, + "loss": 0.7894, + "step": 52350 + }, + { + "epoch": 3.37, + "learning_rate": 8.787651087399902e-06, + "loss": 0.797, + "step": 52351 + }, + { + "epoch": 3.37, + "learning_rate": 8.787436398376953e-06, + "loss": 0.8673, + "step": 52352 + }, + { + "epoch": 3.37, + "learning_rate": 8.787221709354002e-06, + "loss": 0.8282, + "step": 52353 + }, + { + "epoch": 3.37, + "learning_rate": 8.78700702033105e-06, + "loss": 0.7978, + "step": 52354 + }, + { + "epoch": 3.37, + "learning_rate": 8.786792331308101e-06, + "loss": 0.8656, + "step": 52355 + }, + { + "epoch": 3.37, + "learning_rate": 8.786577642285152e-06, + "loss": 0.8402, + "step": 52356 + }, + { + "epoch": 3.37, + "learning_rate": 8.7863629532622e-06, + "loss": 0.8034, + "step": 52357 + }, + { + "epoch": 3.37, + "learning_rate": 8.78614826423925e-06, + "loss": 0.8472, + "step": 52358 + }, + { + "epoch": 3.37, + "learning_rate": 8.7859335752163e-06, + "loss": 0.791, + "step": 52359 + }, + { + "epoch": 3.37, + "learning_rate": 8.78571888619335e-06, + "loss": 0.8535, + "step": 52360 + }, + { + "epoch": 3.37, + "learning_rate": 8.785504197170398e-06, + "loss": 0.7625, + "step": 52361 + }, + { + "epoch": 3.37, + "learning_rate": 8.785289508147449e-06, + "loss": 0.8154, + "step": 52362 + }, + { + "epoch": 3.37, + "learning_rate": 8.785074819124498e-06, + "loss": 0.8364, + "step": 52363 + }, + { + "epoch": 3.37, + "learning_rate": 8.784860130101549e-06, + "loss": 0.8805, + "step": 52364 + }, + { + "epoch": 3.37, + "learning_rate": 8.7846454410786e-06, + "loss": 0.8114, + "step": 52365 + }, + { + "epoch": 3.37, + "learning_rate": 8.784430752055648e-06, + "loss": 0.8289, + "step": 52366 + }, + { + "epoch": 3.37, + "learning_rate": 8.784216063032697e-06, + "loss": 0.7961, + "step": 52367 + }, + { + "epoch": 3.37, + "learning_rate": 8.784001374009748e-06, + "loss": 0.8233, + "step": 52368 + }, + { + "epoch": 3.37, + "learning_rate": 8.783786684986797e-06, + "loss": 0.9315, + "step": 52369 + }, + { + "epoch": 3.37, + "learning_rate": 8.783571995963846e-06, + "loss": 0.869, + "step": 52370 + }, + { + "epoch": 3.37, + "learning_rate": 8.783357306940896e-06, + "loss": 0.8262, + "step": 52371 + }, + { + "epoch": 3.37, + "learning_rate": 8.783142617917947e-06, + "loss": 0.8337, + "step": 52372 + }, + { + "epoch": 3.37, + "learning_rate": 8.782927928894996e-06, + "loss": 0.7897, + "step": 52373 + }, + { + "epoch": 3.37, + "learning_rate": 8.782713239872047e-06, + "loss": 0.8219, + "step": 52374 + }, + { + "epoch": 3.37, + "learning_rate": 8.782498550849096e-06, + "loss": 0.769, + "step": 52375 + }, + { + "epoch": 3.37, + "learning_rate": 8.782283861826145e-06, + "loss": 0.8524, + "step": 52376 + }, + { + "epoch": 3.37, + "learning_rate": 8.782069172803195e-06, + "loss": 0.7502, + "step": 52377 + }, + { + "epoch": 3.37, + "learning_rate": 8.781854483780246e-06, + "loss": 0.8836, + "step": 52378 + }, + { + "epoch": 3.37, + "learning_rate": 8.781639794757295e-06, + "loss": 0.775, + "step": 52379 + }, + { + "epoch": 3.37, + "learning_rate": 8.781425105734344e-06, + "loss": 0.8669, + "step": 52380 + }, + { + "epoch": 3.37, + "learning_rate": 8.781210416711395e-06, + "loss": 0.8769, + "step": 52381 + }, + { + "epoch": 3.37, + "learning_rate": 8.780995727688444e-06, + "loss": 0.8797, + "step": 52382 + }, + { + "epoch": 3.37, + "learning_rate": 8.780781038665494e-06, + "loss": 0.9006, + "step": 52383 + }, + { + "epoch": 3.37, + "learning_rate": 8.780566349642543e-06, + "loss": 0.8267, + "step": 52384 + }, + { + "epoch": 3.37, + "learning_rate": 8.780351660619594e-06, + "loss": 0.8501, + "step": 52385 + }, + { + "epoch": 3.37, + "learning_rate": 8.780136971596643e-06, + "loss": 0.8148, + "step": 52386 + }, + { + "epoch": 3.37, + "learning_rate": 8.779922282573693e-06, + "loss": 0.8232, + "step": 52387 + }, + { + "epoch": 3.37, + "learning_rate": 8.779707593550742e-06, + "loss": 0.8312, + "step": 52388 + }, + { + "epoch": 3.37, + "learning_rate": 8.779492904527791e-06, + "loss": 0.835, + "step": 52389 + }, + { + "epoch": 3.37, + "learning_rate": 8.779278215504842e-06, + "loss": 0.8147, + "step": 52390 + }, + { + "epoch": 3.37, + "learning_rate": 8.779063526481891e-06, + "loss": 0.8361, + "step": 52391 + }, + { + "epoch": 3.37, + "learning_rate": 8.778848837458942e-06, + "loss": 0.8408, + "step": 52392 + }, + { + "epoch": 3.37, + "learning_rate": 8.778634148435992e-06, + "loss": 0.8171, + "step": 52393 + }, + { + "epoch": 3.37, + "learning_rate": 8.778419459413041e-06, + "loss": 0.816, + "step": 52394 + }, + { + "epoch": 3.37, + "learning_rate": 8.77820477039009e-06, + "loss": 0.8107, + "step": 52395 + }, + { + "epoch": 3.37, + "learning_rate": 8.777990081367141e-06, + "loss": 0.8036, + "step": 52396 + }, + { + "epoch": 3.37, + "learning_rate": 8.77777539234419e-06, + "loss": 0.8929, + "step": 52397 + }, + { + "epoch": 3.37, + "learning_rate": 8.777560703321239e-06, + "loss": 0.8367, + "step": 52398 + }, + { + "epoch": 3.37, + "learning_rate": 8.77734601429829e-06, + "loss": 0.7729, + "step": 52399 + }, + { + "epoch": 3.37, + "learning_rate": 8.77713132527534e-06, + "loss": 0.8241, + "step": 52400 + }, + { + "epoch": 3.37, + "learning_rate": 8.776916636252389e-06, + "loss": 0.827, + "step": 52401 + }, + { + "epoch": 3.37, + "learning_rate": 8.77670194722944e-06, + "loss": 0.8051, + "step": 52402 + }, + { + "epoch": 3.37, + "learning_rate": 8.776487258206489e-06, + "loss": 0.8305, + "step": 52403 + }, + { + "epoch": 3.37, + "learning_rate": 8.776272569183538e-06, + "loss": 0.8687, + "step": 52404 + }, + { + "epoch": 3.37, + "learning_rate": 8.776057880160588e-06, + "loss": 0.8025, + "step": 52405 + }, + { + "epoch": 3.37, + "learning_rate": 8.775843191137637e-06, + "loss": 0.7989, + "step": 52406 + }, + { + "epoch": 3.37, + "learning_rate": 8.775628502114688e-06, + "loss": 0.8089, + "step": 52407 + }, + { + "epoch": 3.37, + "learning_rate": 8.775413813091737e-06, + "loss": 0.8614, + "step": 52408 + }, + { + "epoch": 3.37, + "learning_rate": 8.775199124068788e-06, + "loss": 0.7441, + "step": 52409 + }, + { + "epoch": 3.37, + "learning_rate": 8.774984435045837e-06, + "loss": 0.7653, + "step": 52410 + }, + { + "epoch": 3.37, + "learning_rate": 8.774769746022887e-06, + "loss": 0.8153, + "step": 52411 + }, + { + "epoch": 3.37, + "learning_rate": 8.774555056999936e-06, + "loss": 0.7934, + "step": 52412 + }, + { + "epoch": 3.37, + "learning_rate": 8.774340367976985e-06, + "loss": 0.9051, + "step": 52413 + }, + { + "epoch": 3.37, + "learning_rate": 8.774125678954036e-06, + "loss": 0.8742, + "step": 52414 + }, + { + "epoch": 3.37, + "learning_rate": 8.773910989931086e-06, + "loss": 0.8446, + "step": 52415 + }, + { + "epoch": 3.37, + "learning_rate": 8.773696300908135e-06, + "loss": 0.829, + "step": 52416 + }, + { + "epoch": 3.37, + "learning_rate": 8.773481611885184e-06, + "loss": 0.7954, + "step": 52417 + }, + { + "epoch": 3.37, + "learning_rate": 8.773266922862235e-06, + "loss": 0.8292, + "step": 52418 + }, + { + "epoch": 3.37, + "learning_rate": 8.773052233839284e-06, + "loss": 0.8276, + "step": 52419 + }, + { + "epoch": 3.37, + "learning_rate": 8.772837544816335e-06, + "loss": 0.7806, + "step": 52420 + }, + { + "epoch": 3.37, + "learning_rate": 8.772622855793384e-06, + "loss": 0.8256, + "step": 52421 + }, + { + "epoch": 3.37, + "learning_rate": 8.772408166770434e-06, + "loss": 0.8089, + "step": 52422 + }, + { + "epoch": 3.37, + "learning_rate": 8.772193477747483e-06, + "loss": 0.8182, + "step": 52423 + }, + { + "epoch": 3.37, + "learning_rate": 8.771978788724534e-06, + "loss": 0.775, + "step": 52424 + }, + { + "epoch": 3.37, + "learning_rate": 8.771764099701583e-06, + "loss": 0.8244, + "step": 52425 + }, + { + "epoch": 3.37, + "learning_rate": 8.771549410678632e-06, + "loss": 0.85, + "step": 52426 + }, + { + "epoch": 3.37, + "learning_rate": 8.771334721655683e-06, + "loss": 0.8845, + "step": 52427 + }, + { + "epoch": 3.37, + "learning_rate": 8.771120032632731e-06, + "loss": 0.8714, + "step": 52428 + }, + { + "epoch": 3.37, + "learning_rate": 8.770905343609782e-06, + "loss": 0.7844, + "step": 52429 + }, + { + "epoch": 3.37, + "learning_rate": 8.770690654586833e-06, + "loss": 0.812, + "step": 52430 + }, + { + "epoch": 3.37, + "learning_rate": 8.770475965563882e-06, + "loss": 0.8144, + "step": 52431 + }, + { + "epoch": 3.37, + "learning_rate": 8.77026127654093e-06, + "loss": 0.7947, + "step": 52432 + }, + { + "epoch": 3.37, + "learning_rate": 8.770046587517981e-06, + "loss": 0.8625, + "step": 52433 + }, + { + "epoch": 3.37, + "learning_rate": 8.76983189849503e-06, + "loss": 0.8722, + "step": 52434 + }, + { + "epoch": 3.37, + "learning_rate": 8.76961720947208e-06, + "loss": 0.8544, + "step": 52435 + }, + { + "epoch": 3.37, + "learning_rate": 8.76940252044913e-06, + "loss": 0.8588, + "step": 52436 + }, + { + "epoch": 3.37, + "learning_rate": 8.76918783142618e-06, + "loss": 0.8647, + "step": 52437 + }, + { + "epoch": 3.37, + "learning_rate": 8.76897314240323e-06, + "loss": 0.871, + "step": 52438 + }, + { + "epoch": 3.37, + "learning_rate": 8.768758453380279e-06, + "loss": 0.8361, + "step": 52439 + }, + { + "epoch": 3.37, + "learning_rate": 8.76854376435733e-06, + "loss": 0.8267, + "step": 52440 + }, + { + "epoch": 3.37, + "learning_rate": 8.768329075334378e-06, + "loss": 0.8242, + "step": 52441 + }, + { + "epoch": 3.37, + "learning_rate": 8.768114386311429e-06, + "loss": 0.8842, + "step": 52442 + }, + { + "epoch": 3.37, + "learning_rate": 8.76789969728848e-06, + "loss": 0.7872, + "step": 52443 + }, + { + "epoch": 3.37, + "learning_rate": 8.767685008265528e-06, + "loss": 0.7802, + "step": 52444 + }, + { + "epoch": 3.37, + "learning_rate": 8.767470319242577e-06, + "loss": 0.7851, + "step": 52445 + }, + { + "epoch": 3.37, + "learning_rate": 8.767255630219628e-06, + "loss": 0.8463, + "step": 52446 + }, + { + "epoch": 3.37, + "learning_rate": 8.767040941196677e-06, + "loss": 0.8198, + "step": 52447 + }, + { + "epoch": 3.37, + "learning_rate": 8.766826252173726e-06, + "loss": 0.7575, + "step": 52448 + }, + { + "epoch": 3.37, + "learning_rate": 8.766611563150777e-06, + "loss": 0.7692, + "step": 52449 + }, + { + "epoch": 3.37, + "learning_rate": 8.766396874127826e-06, + "loss": 0.7737, + "step": 52450 + }, + { + "epoch": 3.37, + "learning_rate": 8.766182185104876e-06, + "loss": 0.8381, + "step": 52451 + }, + { + "epoch": 3.37, + "learning_rate": 8.765967496081927e-06, + "loss": 0.8426, + "step": 52452 + }, + { + "epoch": 3.37, + "learning_rate": 8.765752807058976e-06, + "loss": 0.8343, + "step": 52453 + }, + { + "epoch": 3.37, + "learning_rate": 8.765538118036025e-06, + "loss": 0.8674, + "step": 52454 + }, + { + "epoch": 3.37, + "learning_rate": 8.765323429013076e-06, + "loss": 0.856, + "step": 52455 + }, + { + "epoch": 3.37, + "learning_rate": 8.765108739990125e-06, + "loss": 0.7964, + "step": 52456 + }, + { + "epoch": 3.37, + "learning_rate": 8.764894050967174e-06, + "loss": 0.8341, + "step": 52457 + }, + { + "epoch": 3.38, + "learning_rate": 8.764679361944224e-06, + "loss": 0.8051, + "step": 52458 + }, + { + "epoch": 3.38, + "learning_rate": 8.764464672921275e-06, + "loss": 0.8234, + "step": 52459 + }, + { + "epoch": 3.38, + "learning_rate": 8.764249983898324e-06, + "loss": 0.8979, + "step": 52460 + }, + { + "epoch": 3.38, + "learning_rate": 8.764035294875374e-06, + "loss": 0.8148, + "step": 52461 + }, + { + "epoch": 3.38, + "learning_rate": 8.763820605852423e-06, + "loss": 0.814, + "step": 52462 + }, + { + "epoch": 3.38, + "learning_rate": 8.763605916829472e-06, + "loss": 0.7892, + "step": 52463 + }, + { + "epoch": 3.38, + "learning_rate": 8.763391227806523e-06, + "loss": 0.8131, + "step": 52464 + }, + { + "epoch": 3.38, + "learning_rate": 8.763176538783574e-06, + "loss": 0.7666, + "step": 52465 + }, + { + "epoch": 3.38, + "learning_rate": 8.762961849760623e-06, + "loss": 0.8247, + "step": 52466 + }, + { + "epoch": 3.38, + "learning_rate": 8.762747160737672e-06, + "loss": 0.8034, + "step": 52467 + }, + { + "epoch": 3.38, + "learning_rate": 8.762532471714722e-06, + "loss": 0.8095, + "step": 52468 + }, + { + "epoch": 3.38, + "learning_rate": 8.762317782691771e-06, + "loss": 0.7887, + "step": 52469 + }, + { + "epoch": 3.38, + "learning_rate": 8.762103093668822e-06, + "loss": 0.806, + "step": 52470 + }, + { + "epoch": 3.38, + "learning_rate": 8.761888404645871e-06, + "loss": 0.8774, + "step": 52471 + }, + { + "epoch": 3.38, + "learning_rate": 8.761673715622922e-06, + "loss": 0.851, + "step": 52472 + }, + { + "epoch": 3.38, + "learning_rate": 8.76145902659997e-06, + "loss": 0.8457, + "step": 52473 + }, + { + "epoch": 3.38, + "learning_rate": 8.761244337577021e-06, + "loss": 0.8232, + "step": 52474 + }, + { + "epoch": 3.38, + "learning_rate": 8.76102964855407e-06, + "loss": 0.8596, + "step": 52475 + }, + { + "epoch": 3.38, + "learning_rate": 8.760814959531119e-06, + "loss": 0.859, + "step": 52476 + }, + { + "epoch": 3.38, + "learning_rate": 8.76060027050817e-06, + "loss": 0.8, + "step": 52477 + }, + { + "epoch": 3.38, + "learning_rate": 8.760385581485219e-06, + "loss": 0.7737, + "step": 52478 + }, + { + "epoch": 3.38, + "learning_rate": 8.76017089246227e-06, + "loss": 0.8881, + "step": 52479 + }, + { + "epoch": 3.38, + "learning_rate": 8.75995620343932e-06, + "loss": 0.8431, + "step": 52480 + }, + { + "epoch": 3.38, + "learning_rate": 8.759741514416369e-06, + "loss": 0.8175, + "step": 52481 + }, + { + "epoch": 3.38, + "learning_rate": 8.759526825393418e-06, + "loss": 0.8311, + "step": 52482 + }, + { + "epoch": 3.38, + "learning_rate": 8.759312136370469e-06, + "loss": 0.871, + "step": 52483 + }, + { + "epoch": 3.38, + "learning_rate": 8.759097447347518e-06, + "loss": 0.8469, + "step": 52484 + }, + { + "epoch": 3.38, + "learning_rate": 8.758882758324567e-06, + "loss": 0.8093, + "step": 52485 + }, + { + "epoch": 3.38, + "learning_rate": 8.758668069301617e-06, + "loss": 0.8285, + "step": 52486 + }, + { + "epoch": 3.38, + "learning_rate": 8.758453380278668e-06, + "loss": 0.8431, + "step": 52487 + }, + { + "epoch": 3.38, + "learning_rate": 8.758238691255717e-06, + "loss": 0.7765, + "step": 52488 + }, + { + "epoch": 3.38, + "learning_rate": 8.758024002232767e-06, + "loss": 0.8036, + "step": 52489 + }, + { + "epoch": 3.38, + "learning_rate": 8.757809313209816e-06, + "loss": 0.8352, + "step": 52490 + }, + { + "epoch": 3.38, + "learning_rate": 8.757594624186865e-06, + "loss": 0.824, + "step": 52491 + }, + { + "epoch": 3.38, + "learning_rate": 8.757379935163916e-06, + "loss": 0.8278, + "step": 52492 + }, + { + "epoch": 3.38, + "learning_rate": 8.757165246140965e-06, + "loss": 0.7983, + "step": 52493 + }, + { + "epoch": 3.38, + "learning_rate": 8.756950557118016e-06, + "loss": 0.8663, + "step": 52494 + }, + { + "epoch": 3.38, + "learning_rate": 8.756735868095065e-06, + "loss": 0.7982, + "step": 52495 + }, + { + "epoch": 3.38, + "learning_rate": 8.756521179072115e-06, + "loss": 0.811, + "step": 52496 + }, + { + "epoch": 3.38, + "learning_rate": 8.756306490049164e-06, + "loss": 0.7853, + "step": 52497 + }, + { + "epoch": 3.38, + "learning_rate": 8.756091801026215e-06, + "loss": 0.8328, + "step": 52498 + }, + { + "epoch": 3.38, + "learning_rate": 8.755877112003264e-06, + "loss": 0.806, + "step": 52499 + }, + { + "epoch": 3.38, + "learning_rate": 8.755662422980313e-06, + "loss": 0.8482, + "step": 52500 + }, + { + "epoch": 3.38, + "learning_rate": 8.755447733957364e-06, + "loss": 0.8273, + "step": 52501 + }, + { + "epoch": 3.38, + "learning_rate": 8.755233044934414e-06, + "loss": 0.811, + "step": 52502 + }, + { + "epoch": 3.38, + "learning_rate": 8.755018355911463e-06, + "loss": 0.8578, + "step": 52503 + }, + { + "epoch": 3.38, + "learning_rate": 8.754803666888512e-06, + "loss": 0.8502, + "step": 52504 + }, + { + "epoch": 3.38, + "learning_rate": 8.754588977865563e-06, + "loss": 0.8131, + "step": 52505 + }, + { + "epoch": 3.38, + "learning_rate": 8.754374288842612e-06, + "loss": 0.8444, + "step": 52506 + }, + { + "epoch": 3.38, + "learning_rate": 8.754159599819662e-06, + "loss": 0.8706, + "step": 52507 + }, + { + "epoch": 3.38, + "learning_rate": 8.753944910796711e-06, + "loss": 0.8051, + "step": 52508 + }, + { + "epoch": 3.38, + "learning_rate": 8.753730221773762e-06, + "loss": 0.7983, + "step": 52509 + }, + { + "epoch": 3.38, + "learning_rate": 8.753515532750811e-06, + "loss": 0.8236, + "step": 52510 + }, + { + "epoch": 3.38, + "learning_rate": 8.753300843727862e-06, + "loss": 0.7964, + "step": 52511 + }, + { + "epoch": 3.38, + "learning_rate": 8.75308615470491e-06, + "loss": 0.7979, + "step": 52512 + }, + { + "epoch": 3.38, + "learning_rate": 8.75287146568196e-06, + "loss": 0.8244, + "step": 52513 + }, + { + "epoch": 3.38, + "learning_rate": 8.75265677665901e-06, + "loss": 0.8801, + "step": 52514 + }, + { + "epoch": 3.38, + "learning_rate": 8.75244208763606e-06, + "loss": 0.829, + "step": 52515 + }, + { + "epoch": 3.38, + "learning_rate": 8.75222739861311e-06, + "loss": 0.8276, + "step": 52516 + }, + { + "epoch": 3.38, + "learning_rate": 8.75201270959016e-06, + "loss": 0.8046, + "step": 52517 + }, + { + "epoch": 3.38, + "learning_rate": 8.75179802056721e-06, + "loss": 0.7979, + "step": 52518 + }, + { + "epoch": 3.38, + "learning_rate": 8.751583331544258e-06, + "loss": 0.8656, + "step": 52519 + }, + { + "epoch": 3.38, + "learning_rate": 8.751368642521309e-06, + "loss": 0.8204, + "step": 52520 + }, + { + "epoch": 3.38, + "learning_rate": 8.751153953498358e-06, + "loss": 0.7931, + "step": 52521 + }, + { + "epoch": 3.38, + "learning_rate": 8.750939264475407e-06, + "loss": 0.7717, + "step": 52522 + }, + { + "epoch": 3.38, + "learning_rate": 8.750724575452458e-06, + "loss": 0.7997, + "step": 52523 + }, + { + "epoch": 3.38, + "learning_rate": 8.750509886429508e-06, + "loss": 0.8407, + "step": 52524 + }, + { + "epoch": 3.38, + "learning_rate": 8.750295197406557e-06, + "loss": 0.8869, + "step": 52525 + }, + { + "epoch": 3.38, + "learning_rate": 8.750080508383606e-06, + "loss": 0.8264, + "step": 52526 + }, + { + "epoch": 3.38, + "learning_rate": 8.749865819360657e-06, + "loss": 0.8949, + "step": 52527 + }, + { + "epoch": 3.38, + "learning_rate": 8.749651130337706e-06, + "loss": 0.7829, + "step": 52528 + }, + { + "epoch": 3.38, + "learning_rate": 8.749436441314757e-06, + "loss": 0.8277, + "step": 52529 + }, + { + "epoch": 3.38, + "learning_rate": 8.749221752291806e-06, + "loss": 0.8499, + "step": 52530 + }, + { + "epoch": 3.38, + "learning_rate": 8.749007063268856e-06, + "loss": 0.8481, + "step": 52531 + }, + { + "epoch": 3.38, + "learning_rate": 8.748792374245905e-06, + "loss": 0.8817, + "step": 52532 + }, + { + "epoch": 3.38, + "learning_rate": 8.748577685222956e-06, + "loss": 0.8155, + "step": 52533 + }, + { + "epoch": 3.38, + "learning_rate": 8.748362996200005e-06, + "loss": 0.7842, + "step": 52534 + }, + { + "epoch": 3.38, + "learning_rate": 8.748148307177054e-06, + "loss": 0.892, + "step": 52535 + }, + { + "epoch": 3.38, + "learning_rate": 8.747933618154104e-06, + "loss": 0.8389, + "step": 52536 + }, + { + "epoch": 3.38, + "learning_rate": 8.747718929131153e-06, + "loss": 0.9363, + "step": 52537 + }, + { + "epoch": 3.38, + "learning_rate": 8.747504240108204e-06, + "loss": 0.7949, + "step": 52538 + }, + { + "epoch": 3.38, + "learning_rate": 8.747289551085255e-06, + "loss": 0.8257, + "step": 52539 + }, + { + "epoch": 3.38, + "learning_rate": 8.747074862062304e-06, + "loss": 0.8567, + "step": 52540 + }, + { + "epoch": 3.38, + "learning_rate": 8.746860173039353e-06, + "loss": 0.8119, + "step": 52541 + }, + { + "epoch": 3.38, + "learning_rate": 8.746645484016403e-06, + "loss": 0.7926, + "step": 52542 + }, + { + "epoch": 3.38, + "learning_rate": 8.746430794993452e-06, + "loss": 0.8128, + "step": 52543 + }, + { + "epoch": 3.38, + "learning_rate": 8.746216105970501e-06, + "loss": 0.8399, + "step": 52544 + }, + { + "epoch": 3.38, + "learning_rate": 8.746001416947552e-06, + "loss": 0.7813, + "step": 52545 + }, + { + "epoch": 3.38, + "learning_rate": 8.745786727924603e-06, + "loss": 0.8441, + "step": 52546 + }, + { + "epoch": 3.38, + "learning_rate": 8.745572038901651e-06, + "loss": 0.7587, + "step": 52547 + }, + { + "epoch": 3.38, + "learning_rate": 8.745357349878702e-06, + "loss": 0.8486, + "step": 52548 + }, + { + "epoch": 3.38, + "learning_rate": 8.745142660855751e-06, + "loss": 0.8386, + "step": 52549 + }, + { + "epoch": 3.38, + "learning_rate": 8.7449279718328e-06, + "loss": 0.8667, + "step": 52550 + }, + { + "epoch": 3.38, + "learning_rate": 8.74471328280985e-06, + "loss": 0.844, + "step": 52551 + }, + { + "epoch": 3.38, + "learning_rate": 8.744498593786901e-06, + "loss": 0.8439, + "step": 52552 + }, + { + "epoch": 3.38, + "learning_rate": 8.74428390476395e-06, + "loss": 0.791, + "step": 52553 + }, + { + "epoch": 3.38, + "learning_rate": 8.744069215741e-06, + "loss": 0.8297, + "step": 52554 + }, + { + "epoch": 3.38, + "learning_rate": 8.74385452671805e-06, + "loss": 0.871, + "step": 52555 + }, + { + "epoch": 3.38, + "learning_rate": 8.743639837695099e-06, + "loss": 0.8216, + "step": 52556 + }, + { + "epoch": 3.38, + "learning_rate": 8.74342514867215e-06, + "loss": 0.7802, + "step": 52557 + }, + { + "epoch": 3.38, + "learning_rate": 8.743210459649199e-06, + "loss": 0.8068, + "step": 52558 + }, + { + "epoch": 3.38, + "learning_rate": 8.742995770626248e-06, + "loss": 0.8067, + "step": 52559 + }, + { + "epoch": 3.38, + "learning_rate": 8.742781081603298e-06, + "loss": 0.8768, + "step": 52560 + }, + { + "epoch": 3.38, + "learning_rate": 8.742566392580349e-06, + "loss": 0.7939, + "step": 52561 + }, + { + "epoch": 3.38, + "learning_rate": 8.742351703557398e-06, + "loss": 0.8431, + "step": 52562 + }, + { + "epoch": 3.38, + "learning_rate": 8.742137014534447e-06, + "loss": 0.8176, + "step": 52563 + }, + { + "epoch": 3.38, + "learning_rate": 8.741922325511497e-06, + "loss": 0.858, + "step": 52564 + }, + { + "epoch": 3.38, + "learning_rate": 8.741707636488546e-06, + "loss": 0.863, + "step": 52565 + }, + { + "epoch": 3.38, + "learning_rate": 8.741492947465597e-06, + "loss": 0.8381, + "step": 52566 + }, + { + "epoch": 3.38, + "learning_rate": 8.741278258442648e-06, + "loss": 0.7946, + "step": 52567 + }, + { + "epoch": 3.38, + "learning_rate": 8.741063569419697e-06, + "loss": 0.8543, + "step": 52568 + }, + { + "epoch": 3.38, + "learning_rate": 8.740848880396746e-06, + "loss": 0.8524, + "step": 52569 + }, + { + "epoch": 3.38, + "learning_rate": 8.740634191373796e-06, + "loss": 0.8572, + "step": 52570 + }, + { + "epoch": 3.38, + "learning_rate": 8.740419502350845e-06, + "loss": 0.8105, + "step": 52571 + }, + { + "epoch": 3.38, + "learning_rate": 8.740204813327894e-06, + "loss": 0.8761, + "step": 52572 + }, + { + "epoch": 3.38, + "learning_rate": 8.739990124304945e-06, + "loss": 0.8109, + "step": 52573 + }, + { + "epoch": 3.38, + "learning_rate": 8.739775435281996e-06, + "loss": 0.8258, + "step": 52574 + }, + { + "epoch": 3.38, + "learning_rate": 8.739560746259045e-06, + "loss": 0.8382, + "step": 52575 + }, + { + "epoch": 3.38, + "learning_rate": 8.739346057236095e-06, + "loss": 0.7824, + "step": 52576 + }, + { + "epoch": 3.38, + "learning_rate": 8.739131368213144e-06, + "loss": 0.8267, + "step": 52577 + }, + { + "epoch": 3.38, + "learning_rate": 8.738916679190193e-06, + "loss": 0.8244, + "step": 52578 + }, + { + "epoch": 3.38, + "learning_rate": 8.738701990167244e-06, + "loss": 0.7608, + "step": 52579 + }, + { + "epoch": 3.38, + "learning_rate": 8.738487301144293e-06, + "loss": 0.8023, + "step": 52580 + }, + { + "epoch": 3.38, + "learning_rate": 8.738272612121343e-06, + "loss": 0.8202, + "step": 52581 + }, + { + "epoch": 3.38, + "learning_rate": 8.738057923098392e-06, + "loss": 0.9059, + "step": 52582 + }, + { + "epoch": 3.38, + "learning_rate": 8.737843234075443e-06, + "loss": 0.8545, + "step": 52583 + }, + { + "epoch": 3.38, + "learning_rate": 8.737628545052492e-06, + "loss": 0.8458, + "step": 52584 + }, + { + "epoch": 3.38, + "learning_rate": 8.737413856029543e-06, + "loss": 0.8232, + "step": 52585 + }, + { + "epoch": 3.38, + "learning_rate": 8.737199167006592e-06, + "loss": 0.7834, + "step": 52586 + }, + { + "epoch": 3.38, + "learning_rate": 8.73698447798364e-06, + "loss": 0.8055, + "step": 52587 + }, + { + "epoch": 3.38, + "learning_rate": 8.736769788960691e-06, + "loss": 0.793, + "step": 52588 + }, + { + "epoch": 3.38, + "learning_rate": 8.736555099937742e-06, + "loss": 0.8237, + "step": 52589 + }, + { + "epoch": 3.38, + "learning_rate": 8.736340410914791e-06, + "loss": 0.8067, + "step": 52590 + }, + { + "epoch": 3.38, + "learning_rate": 8.73612572189184e-06, + "loss": 0.8071, + "step": 52591 + }, + { + "epoch": 3.38, + "learning_rate": 8.73591103286889e-06, + "loss": 0.7615, + "step": 52592 + }, + { + "epoch": 3.38, + "learning_rate": 8.73569634384594e-06, + "loss": 0.8565, + "step": 52593 + }, + { + "epoch": 3.38, + "learning_rate": 8.735481654822988e-06, + "loss": 0.8165, + "step": 52594 + }, + { + "epoch": 3.38, + "learning_rate": 8.735266965800039e-06, + "loss": 0.8061, + "step": 52595 + }, + { + "epoch": 3.38, + "learning_rate": 8.73505227677709e-06, + "loss": 0.8034, + "step": 52596 + }, + { + "epoch": 3.38, + "learning_rate": 8.734837587754139e-06, + "loss": 0.8394, + "step": 52597 + }, + { + "epoch": 3.38, + "learning_rate": 8.73462289873119e-06, + "loss": 0.8356, + "step": 52598 + }, + { + "epoch": 3.38, + "learning_rate": 8.734408209708238e-06, + "loss": 0.8077, + "step": 52599 + }, + { + "epoch": 3.38, + "learning_rate": 8.734193520685287e-06, + "loss": 0.7707, + "step": 52600 + }, + { + "epoch": 3.38, + "learning_rate": 8.733978831662338e-06, + "loss": 0.791, + "step": 52601 + }, + { + "epoch": 3.38, + "learning_rate": 8.733764142639387e-06, + "loss": 0.7667, + "step": 52602 + }, + { + "epoch": 3.38, + "learning_rate": 8.733549453616438e-06, + "loss": 0.7966, + "step": 52603 + }, + { + "epoch": 3.38, + "learning_rate": 8.733334764593487e-06, + "loss": 0.8434, + "step": 52604 + }, + { + "epoch": 3.38, + "learning_rate": 8.733120075570537e-06, + "loss": 0.8786, + "step": 52605 + }, + { + "epoch": 3.38, + "learning_rate": 8.732905386547586e-06, + "loss": 0.9062, + "step": 52606 + }, + { + "epoch": 3.38, + "learning_rate": 8.732690697524637e-06, + "loss": 0.863, + "step": 52607 + }, + { + "epoch": 3.38, + "learning_rate": 8.732476008501686e-06, + "loss": 0.7767, + "step": 52608 + }, + { + "epoch": 3.38, + "learning_rate": 8.732261319478735e-06, + "loss": 0.7831, + "step": 52609 + }, + { + "epoch": 3.38, + "learning_rate": 8.732046630455785e-06, + "loss": 0.8403, + "step": 52610 + }, + { + "epoch": 3.38, + "learning_rate": 8.731831941432836e-06, + "loss": 0.8877, + "step": 52611 + }, + { + "epoch": 3.38, + "learning_rate": 8.731617252409885e-06, + "loss": 0.7536, + "step": 52612 + }, + { + "epoch": 3.38, + "learning_rate": 8.731402563386934e-06, + "loss": 0.8517, + "step": 52613 + }, + { + "epoch": 3.39, + "learning_rate": 8.731187874363985e-06, + "loss": 0.8543, + "step": 52614 + }, + { + "epoch": 3.39, + "learning_rate": 8.730973185341034e-06, + "loss": 0.803, + "step": 52615 + }, + { + "epoch": 3.39, + "learning_rate": 8.730758496318084e-06, + "loss": 0.8357, + "step": 52616 + }, + { + "epoch": 3.39, + "learning_rate": 8.730543807295133e-06, + "loss": 0.8526, + "step": 52617 + }, + { + "epoch": 3.39, + "learning_rate": 8.730329118272184e-06, + "loss": 0.7669, + "step": 52618 + }, + { + "epoch": 3.39, + "learning_rate": 8.730114429249233e-06, + "loss": 0.8285, + "step": 52619 + }, + { + "epoch": 3.39, + "learning_rate": 8.729899740226284e-06, + "loss": 0.8064, + "step": 52620 + }, + { + "epoch": 3.39, + "learning_rate": 8.729685051203333e-06, + "loss": 0.8513, + "step": 52621 + }, + { + "epoch": 3.39, + "learning_rate": 8.729470362180381e-06, + "loss": 0.7562, + "step": 52622 + }, + { + "epoch": 3.39, + "learning_rate": 8.729255673157432e-06, + "loss": 0.7282, + "step": 52623 + }, + { + "epoch": 3.39, + "learning_rate": 8.729040984134481e-06, + "loss": 0.7679, + "step": 52624 + }, + { + "epoch": 3.39, + "learning_rate": 8.728826295111532e-06, + "loss": 0.7361, + "step": 52625 + }, + { + "epoch": 3.39, + "learning_rate": 8.728611606088582e-06, + "loss": 0.7921, + "step": 52626 + }, + { + "epoch": 3.39, + "learning_rate": 8.728396917065631e-06, + "loss": 0.799, + "step": 52627 + }, + { + "epoch": 3.39, + "learning_rate": 8.72818222804268e-06, + "loss": 0.8445, + "step": 52628 + }, + { + "epoch": 3.39, + "learning_rate": 8.727967539019731e-06, + "loss": 0.8136, + "step": 52629 + }, + { + "epoch": 3.39, + "learning_rate": 8.72775284999678e-06, + "loss": 0.7995, + "step": 52630 + }, + { + "epoch": 3.39, + "learning_rate": 8.727538160973829e-06, + "loss": 0.8098, + "step": 52631 + }, + { + "epoch": 3.39, + "learning_rate": 8.72732347195088e-06, + "loss": 0.8328, + "step": 52632 + }, + { + "epoch": 3.39, + "learning_rate": 8.72710878292793e-06, + "loss": 0.8109, + "step": 52633 + }, + { + "epoch": 3.39, + "learning_rate": 8.72689409390498e-06, + "loss": 0.896, + "step": 52634 + }, + { + "epoch": 3.39, + "learning_rate": 8.72667940488203e-06, + "loss": 0.8328, + "step": 52635 + }, + { + "epoch": 3.39, + "learning_rate": 8.726464715859079e-06, + "loss": 0.8255, + "step": 52636 + }, + { + "epoch": 3.39, + "learning_rate": 8.726250026836128e-06, + "loss": 0.8315, + "step": 52637 + }, + { + "epoch": 3.39, + "learning_rate": 8.726035337813178e-06, + "loss": 0.8479, + "step": 52638 + }, + { + "epoch": 3.39, + "learning_rate": 8.725820648790229e-06, + "loss": 0.9132, + "step": 52639 + }, + { + "epoch": 3.39, + "learning_rate": 8.725605959767278e-06, + "loss": 0.8964, + "step": 52640 + }, + { + "epoch": 3.39, + "learning_rate": 8.725391270744327e-06, + "loss": 0.8583, + "step": 52641 + }, + { + "epoch": 3.39, + "learning_rate": 8.725176581721378e-06, + "loss": 0.787, + "step": 52642 + }, + { + "epoch": 3.39, + "learning_rate": 8.724961892698427e-06, + "loss": 0.8474, + "step": 52643 + }, + { + "epoch": 3.39, + "learning_rate": 8.724747203675477e-06, + "loss": 0.8793, + "step": 52644 + }, + { + "epoch": 3.39, + "learning_rate": 8.724532514652526e-06, + "loss": 0.8197, + "step": 52645 + }, + { + "epoch": 3.39, + "learning_rate": 8.724317825629575e-06, + "loss": 0.8303, + "step": 52646 + }, + { + "epoch": 3.39, + "learning_rate": 8.724103136606626e-06, + "loss": 0.8355, + "step": 52647 + }, + { + "epoch": 3.39, + "learning_rate": 8.723888447583677e-06, + "loss": 0.8686, + "step": 52648 + }, + { + "epoch": 3.39, + "learning_rate": 8.723673758560726e-06, + "loss": 0.8386, + "step": 52649 + }, + { + "epoch": 3.39, + "learning_rate": 8.723459069537775e-06, + "loss": 0.8681, + "step": 52650 + }, + { + "epoch": 3.39, + "learning_rate": 8.723244380514825e-06, + "loss": 0.8416, + "step": 52651 + }, + { + "epoch": 3.39, + "learning_rate": 8.723029691491874e-06, + "loss": 0.8266, + "step": 52652 + }, + { + "epoch": 3.39, + "learning_rate": 8.722815002468925e-06, + "loss": 0.9102, + "step": 52653 + }, + { + "epoch": 3.39, + "learning_rate": 8.722600313445975e-06, + "loss": 0.8562, + "step": 52654 + }, + { + "epoch": 3.39, + "learning_rate": 8.722385624423024e-06, + "loss": 0.8561, + "step": 52655 + }, + { + "epoch": 3.39, + "learning_rate": 8.722170935400073e-06, + "loss": 0.8504, + "step": 52656 + }, + { + "epoch": 3.39, + "learning_rate": 8.721956246377124e-06, + "loss": 0.8882, + "step": 52657 + }, + { + "epoch": 3.39, + "learning_rate": 8.721741557354173e-06, + "loss": 0.8131, + "step": 52658 + }, + { + "epoch": 3.39, + "learning_rate": 8.721741557354173e-06, + "loss": 0.8168, + "step": 52659 + }, + { + "epoch": 3.39, + "learning_rate": 8.721526868331222e-06, + "loss": 0.8539, + "step": 52660 + }, + { + "epoch": 3.39, + "learning_rate": 8.721312179308273e-06, + "loss": 0.7981, + "step": 52661 + }, + { + "epoch": 3.39, + "learning_rate": 8.721097490285323e-06, + "loss": 0.8183, + "step": 52662 + }, + { + "epoch": 3.39, + "learning_rate": 8.720882801262372e-06, + "loss": 0.8049, + "step": 52663 + }, + { + "epoch": 3.39, + "learning_rate": 8.720668112239423e-06, + "loss": 0.8105, + "step": 52664 + }, + { + "epoch": 3.39, + "learning_rate": 8.720453423216472e-06, + "loss": 0.828, + "step": 52665 + }, + { + "epoch": 3.39, + "learning_rate": 8.72023873419352e-06, + "loss": 0.8555, + "step": 52666 + }, + { + "epoch": 3.39, + "learning_rate": 8.720024045170572e-06, + "loss": 0.7761, + "step": 52667 + }, + { + "epoch": 3.39, + "learning_rate": 8.71980935614762e-06, + "loss": 0.8238, + "step": 52668 + }, + { + "epoch": 3.39, + "learning_rate": 8.719594667124671e-06, + "loss": 0.8048, + "step": 52669 + }, + { + "epoch": 3.39, + "learning_rate": 8.71937997810172e-06, + "loss": 0.8198, + "step": 52670 + }, + { + "epoch": 3.39, + "learning_rate": 8.71916528907877e-06, + "loss": 0.913, + "step": 52671 + }, + { + "epoch": 3.39, + "learning_rate": 8.71895060005582e-06, + "loss": 0.8921, + "step": 52672 + }, + { + "epoch": 3.39, + "learning_rate": 8.71873591103287e-06, + "loss": 0.8129, + "step": 52673 + }, + { + "epoch": 3.39, + "learning_rate": 8.71852122200992e-06, + "loss": 0.8302, + "step": 52674 + }, + { + "epoch": 3.39, + "learning_rate": 8.718306532986968e-06, + "loss": 0.7704, + "step": 52675 + }, + { + "epoch": 3.39, + "learning_rate": 8.718091843964019e-06, + "loss": 0.8198, + "step": 52676 + }, + { + "epoch": 3.39, + "learning_rate": 8.71787715494107e-06, + "loss": 0.794, + "step": 52677 + }, + { + "epoch": 3.39, + "learning_rate": 8.717662465918119e-06, + "loss": 0.8962, + "step": 52678 + }, + { + "epoch": 3.39, + "learning_rate": 8.717447776895168e-06, + "loss": 0.8197, + "step": 52679 + }, + { + "epoch": 3.39, + "learning_rate": 8.717233087872218e-06, + "loss": 0.8634, + "step": 52680 + }, + { + "epoch": 3.39, + "learning_rate": 8.717018398849267e-06, + "loss": 0.7761, + "step": 52681 + }, + { + "epoch": 3.39, + "learning_rate": 8.716803709826316e-06, + "loss": 0.8001, + "step": 52682 + }, + { + "epoch": 3.39, + "learning_rate": 8.716589020803367e-06, + "loss": 0.7934, + "step": 52683 + }, + { + "epoch": 3.39, + "learning_rate": 8.716374331780417e-06, + "loss": 0.8138, + "step": 52684 + }, + { + "epoch": 3.39, + "learning_rate": 8.716159642757466e-06, + "loss": 0.7691, + "step": 52685 + }, + { + "epoch": 3.39, + "learning_rate": 8.715944953734517e-06, + "loss": 0.8278, + "step": 52686 + }, + { + "epoch": 3.39, + "learning_rate": 8.715730264711566e-06, + "loss": 0.7656, + "step": 52687 + }, + { + "epoch": 3.39, + "learning_rate": 8.715515575688615e-06, + "loss": 0.8529, + "step": 52688 + }, + { + "epoch": 3.39, + "learning_rate": 8.715300886665666e-06, + "loss": 0.8288, + "step": 52689 + }, + { + "epoch": 3.39, + "learning_rate": 8.715086197642715e-06, + "loss": 0.8278, + "step": 52690 + }, + { + "epoch": 3.39, + "learning_rate": 8.714871508619765e-06, + "loss": 0.8439, + "step": 52691 + }, + { + "epoch": 3.39, + "learning_rate": 8.714656819596814e-06, + "loss": 0.8074, + "step": 52692 + }, + { + "epoch": 3.39, + "learning_rate": 8.714442130573865e-06, + "loss": 0.8666, + "step": 52693 + }, + { + "epoch": 3.39, + "learning_rate": 8.714227441550914e-06, + "loss": 0.8409, + "step": 52694 + }, + { + "epoch": 3.39, + "learning_rate": 8.714012752527965e-06, + "loss": 0.86, + "step": 52695 + }, + { + "epoch": 3.39, + "learning_rate": 8.713798063505014e-06, + "loss": 0.8435, + "step": 52696 + }, + { + "epoch": 3.39, + "learning_rate": 8.713583374482062e-06, + "loss": 0.8069, + "step": 52697 + }, + { + "epoch": 3.39, + "learning_rate": 8.713368685459113e-06, + "loss": 0.8037, + "step": 52698 + }, + { + "epoch": 3.39, + "learning_rate": 8.713153996436164e-06, + "loss": 0.927, + "step": 52699 + }, + { + "epoch": 3.39, + "learning_rate": 8.712939307413213e-06, + "loss": 0.835, + "step": 52700 + }, + { + "epoch": 3.39, + "learning_rate": 8.712724618390262e-06, + "loss": 0.7955, + "step": 52701 + }, + { + "epoch": 3.39, + "learning_rate": 8.712509929367312e-06, + "loss": 0.8401, + "step": 52702 + }, + { + "epoch": 3.39, + "learning_rate": 8.712295240344361e-06, + "loss": 0.8758, + "step": 52703 + }, + { + "epoch": 3.39, + "learning_rate": 8.712080551321412e-06, + "loss": 0.8163, + "step": 52704 + }, + { + "epoch": 3.39, + "learning_rate": 8.711865862298461e-06, + "loss": 0.7584, + "step": 52705 + }, + { + "epoch": 3.39, + "learning_rate": 8.711651173275512e-06, + "loss": 0.8175, + "step": 52706 + }, + { + "epoch": 3.39, + "learning_rate": 8.71143648425256e-06, + "loss": 0.8116, + "step": 52707 + }, + { + "epoch": 3.39, + "learning_rate": 8.711221795229611e-06, + "loss": 0.7994, + "step": 52708 + }, + { + "epoch": 3.39, + "learning_rate": 8.71100710620666e-06, + "loss": 0.8286, + "step": 52709 + }, + { + "epoch": 3.39, + "learning_rate": 8.71079241718371e-06, + "loss": 0.8315, + "step": 52710 + }, + { + "epoch": 3.39, + "learning_rate": 8.71057772816076e-06, + "loss": 0.8063, + "step": 52711 + }, + { + "epoch": 3.39, + "learning_rate": 8.710363039137809e-06, + "loss": 0.8496, + "step": 52712 + }, + { + "epoch": 3.39, + "learning_rate": 8.71014835011486e-06, + "loss": 0.8559, + "step": 52713 + }, + { + "epoch": 3.39, + "learning_rate": 8.70993366109191e-06, + "loss": 0.7743, + "step": 52714 + }, + { + "epoch": 3.39, + "learning_rate": 8.709718972068959e-06, + "loss": 0.8424, + "step": 52715 + }, + { + "epoch": 3.39, + "learning_rate": 8.709504283046008e-06, + "loss": 0.8253, + "step": 52716 + }, + { + "epoch": 3.39, + "learning_rate": 8.709289594023059e-06, + "loss": 0.7772, + "step": 52717 + }, + { + "epoch": 3.39, + "learning_rate": 8.709074905000108e-06, + "loss": 0.8213, + "step": 52718 + }, + { + "epoch": 3.39, + "learning_rate": 8.708860215977157e-06, + "loss": 0.7786, + "step": 52719 + }, + { + "epoch": 3.39, + "learning_rate": 8.708645526954207e-06, + "loss": 0.8308, + "step": 52720 + }, + { + "epoch": 3.39, + "learning_rate": 8.708430837931258e-06, + "loss": 0.8019, + "step": 52721 + }, + { + "epoch": 3.39, + "learning_rate": 8.708216148908307e-06, + "loss": 0.8896, + "step": 52722 + }, + { + "epoch": 3.39, + "learning_rate": 8.708001459885358e-06, + "loss": 0.8079, + "step": 52723 + }, + { + "epoch": 3.39, + "learning_rate": 8.707786770862407e-06, + "loss": 0.8046, + "step": 52724 + }, + { + "epoch": 3.39, + "learning_rate": 8.707572081839456e-06, + "loss": 0.7937, + "step": 52725 + }, + { + "epoch": 3.39, + "learning_rate": 8.707357392816506e-06, + "loss": 0.8472, + "step": 52726 + }, + { + "epoch": 3.39, + "learning_rate": 8.707142703793555e-06, + "loss": 0.8594, + "step": 52727 + }, + { + "epoch": 3.39, + "learning_rate": 8.706928014770606e-06, + "loss": 0.7302, + "step": 52728 + }, + { + "epoch": 3.39, + "learning_rate": 8.706713325747655e-06, + "loss": 0.747, + "step": 52729 + }, + { + "epoch": 3.39, + "learning_rate": 8.706498636724705e-06, + "loss": 0.7924, + "step": 52730 + }, + { + "epoch": 3.39, + "learning_rate": 8.706283947701754e-06, + "loss": 0.8306, + "step": 52731 + }, + { + "epoch": 3.39, + "learning_rate": 8.706069258678805e-06, + "loss": 0.8403, + "step": 52732 + }, + { + "epoch": 3.39, + "learning_rate": 8.705854569655854e-06, + "loss": 0.8358, + "step": 52733 + }, + { + "epoch": 3.39, + "learning_rate": 8.705639880632903e-06, + "loss": 0.863, + "step": 52734 + }, + { + "epoch": 3.39, + "learning_rate": 8.705425191609954e-06, + "loss": 0.8011, + "step": 52735 + }, + { + "epoch": 3.39, + "learning_rate": 8.705210502587004e-06, + "loss": 0.8046, + "step": 52736 + }, + { + "epoch": 3.39, + "learning_rate": 8.704995813564053e-06, + "loss": 0.819, + "step": 52737 + }, + { + "epoch": 3.39, + "learning_rate": 8.704781124541102e-06, + "loss": 0.7984, + "step": 52738 + }, + { + "epoch": 3.39, + "learning_rate": 8.704566435518153e-06, + "loss": 0.8994, + "step": 52739 + }, + { + "epoch": 3.39, + "learning_rate": 8.704351746495202e-06, + "loss": 0.8278, + "step": 52740 + }, + { + "epoch": 3.39, + "learning_rate": 8.704137057472253e-06, + "loss": 0.8196, + "step": 52741 + }, + { + "epoch": 3.39, + "learning_rate": 8.703922368449303e-06, + "loss": 0.8302, + "step": 52742 + }, + { + "epoch": 3.39, + "learning_rate": 8.703707679426352e-06, + "loss": 0.8723, + "step": 52743 + }, + { + "epoch": 3.39, + "learning_rate": 8.703492990403401e-06, + "loss": 0.7933, + "step": 52744 + }, + { + "epoch": 3.39, + "learning_rate": 8.703278301380452e-06, + "loss": 0.835, + "step": 52745 + }, + { + "epoch": 3.39, + "learning_rate": 8.7030636123575e-06, + "loss": 0.8572, + "step": 52746 + }, + { + "epoch": 3.39, + "learning_rate": 8.70284892333455e-06, + "loss": 0.8413, + "step": 52747 + }, + { + "epoch": 3.39, + "learning_rate": 8.7026342343116e-06, + "loss": 0.8079, + "step": 52748 + }, + { + "epoch": 3.39, + "learning_rate": 8.702419545288651e-06, + "loss": 0.911, + "step": 52749 + }, + { + "epoch": 3.39, + "learning_rate": 8.7022048562657e-06, + "loss": 0.917, + "step": 52750 + }, + { + "epoch": 3.39, + "learning_rate": 8.70199016724275e-06, + "loss": 0.8746, + "step": 52751 + }, + { + "epoch": 3.39, + "learning_rate": 8.7017754782198e-06, + "loss": 0.7957, + "step": 52752 + }, + { + "epoch": 3.39, + "learning_rate": 8.701560789196849e-06, + "loss": 0.864, + "step": 52753 + }, + { + "epoch": 3.39, + "learning_rate": 8.7013461001739e-06, + "loss": 0.8096, + "step": 52754 + }, + { + "epoch": 3.39, + "learning_rate": 8.701131411150948e-06, + "loss": 0.8573, + "step": 52755 + }, + { + "epoch": 3.39, + "learning_rate": 8.700916722127997e-06, + "loss": 0.8064, + "step": 52756 + }, + { + "epoch": 3.39, + "learning_rate": 8.700702033105048e-06, + "loss": 0.8311, + "step": 52757 + }, + { + "epoch": 3.39, + "learning_rate": 8.700487344082098e-06, + "loss": 0.7992, + "step": 52758 + }, + { + "epoch": 3.39, + "learning_rate": 8.700272655059147e-06, + "loss": 0.8765, + "step": 52759 + }, + { + "epoch": 3.39, + "learning_rate": 8.700057966036196e-06, + "loss": 0.8593, + "step": 52760 + }, + { + "epoch": 3.39, + "learning_rate": 8.699843277013247e-06, + "loss": 0.7342, + "step": 52761 + }, + { + "epoch": 3.39, + "learning_rate": 8.699628587990296e-06, + "loss": 0.8453, + "step": 52762 + }, + { + "epoch": 3.39, + "learning_rate": 8.699413898967347e-06, + "loss": 0.8375, + "step": 52763 + }, + { + "epoch": 3.39, + "learning_rate": 8.699199209944397e-06, + "loss": 0.8319, + "step": 52764 + }, + { + "epoch": 3.39, + "learning_rate": 8.698984520921446e-06, + "loss": 0.8054, + "step": 52765 + }, + { + "epoch": 3.39, + "learning_rate": 8.698769831898495e-06, + "loss": 0.8519, + "step": 52766 + }, + { + "epoch": 3.39, + "learning_rate": 8.698555142875546e-06, + "loss": 0.8251, + "step": 52767 + }, + { + "epoch": 3.39, + "learning_rate": 8.698340453852595e-06, + "loss": 0.7627, + "step": 52768 + }, + { + "epoch": 3.4, + "learning_rate": 8.698125764829644e-06, + "loss": 0.8235, + "step": 52769 + }, + { + "epoch": 3.4, + "learning_rate": 8.697911075806695e-06, + "loss": 0.805, + "step": 52770 + }, + { + "epoch": 3.4, + "learning_rate": 8.697696386783745e-06, + "loss": 0.8358, + "step": 52771 + }, + { + "epoch": 3.4, + "learning_rate": 8.697481697760794e-06, + "loss": 0.8911, + "step": 52772 + }, + { + "epoch": 3.4, + "learning_rate": 8.697267008737845e-06, + "loss": 0.8185, + "step": 52773 + }, + { + "epoch": 3.4, + "learning_rate": 8.697052319714894e-06, + "loss": 0.8165, + "step": 52774 + }, + { + "epoch": 3.4, + "learning_rate": 8.696837630691943e-06, + "loss": 0.8768, + "step": 52775 + }, + { + "epoch": 3.4, + "learning_rate": 8.696622941668993e-06, + "loss": 0.8438, + "step": 52776 + }, + { + "epoch": 3.4, + "learning_rate": 8.696408252646042e-06, + "loss": 0.791, + "step": 52777 + }, + { + "epoch": 3.4, + "learning_rate": 8.696193563623093e-06, + "loss": 0.8616, + "step": 52778 + }, + { + "epoch": 3.4, + "learning_rate": 8.695978874600142e-06, + "loss": 0.7911, + "step": 52779 + }, + { + "epoch": 3.4, + "learning_rate": 8.695764185577193e-06, + "loss": 0.7776, + "step": 52780 + }, + { + "epoch": 3.4, + "learning_rate": 8.695549496554242e-06, + "loss": 0.8222, + "step": 52781 + }, + { + "epoch": 3.4, + "learning_rate": 8.695334807531292e-06, + "loss": 0.81, + "step": 52782 + }, + { + "epoch": 3.4, + "learning_rate": 8.695120118508341e-06, + "loss": 0.83, + "step": 52783 + }, + { + "epoch": 3.4, + "learning_rate": 8.69490542948539e-06, + "loss": 0.7686, + "step": 52784 + }, + { + "epoch": 3.4, + "learning_rate": 8.69469074046244e-06, + "loss": 0.8733, + "step": 52785 + }, + { + "epoch": 3.4, + "learning_rate": 8.694476051439492e-06, + "loss": 0.7889, + "step": 52786 + }, + { + "epoch": 3.4, + "learning_rate": 8.69426136241654e-06, + "loss": 0.8052, + "step": 52787 + }, + { + "epoch": 3.4, + "learning_rate": 8.69404667339359e-06, + "loss": 0.76, + "step": 52788 + }, + { + "epoch": 3.4, + "learning_rate": 8.69383198437064e-06, + "loss": 0.8587, + "step": 52789 + }, + { + "epoch": 3.4, + "learning_rate": 8.693617295347689e-06, + "loss": 0.8574, + "step": 52790 + }, + { + "epoch": 3.4, + "learning_rate": 8.69340260632474e-06, + "loss": 0.8281, + "step": 52791 + }, + { + "epoch": 3.4, + "learning_rate": 8.693187917301789e-06, + "loss": 0.7964, + "step": 52792 + }, + { + "epoch": 3.4, + "learning_rate": 8.69297322827884e-06, + "loss": 0.8359, + "step": 52793 + }, + { + "epoch": 3.4, + "learning_rate": 8.692758539255888e-06, + "loss": 0.8424, + "step": 52794 + }, + { + "epoch": 3.4, + "learning_rate": 8.692543850232939e-06, + "loss": 0.8315, + "step": 52795 + }, + { + "epoch": 3.4, + "learning_rate": 8.692329161209988e-06, + "loss": 0.8476, + "step": 52796 + }, + { + "epoch": 3.4, + "learning_rate": 8.692114472187037e-06, + "loss": 0.8315, + "step": 52797 + }, + { + "epoch": 3.4, + "learning_rate": 8.691899783164088e-06, + "loss": 0.8759, + "step": 52798 + }, + { + "epoch": 3.4, + "learning_rate": 8.691685094141137e-06, + "loss": 0.869, + "step": 52799 + }, + { + "epoch": 3.4, + "learning_rate": 8.691470405118187e-06, + "loss": 0.7585, + "step": 52800 + }, + { + "epoch": 3.4, + "learning_rate": 8.691255716095238e-06, + "loss": 0.7888, + "step": 52801 + }, + { + "epoch": 3.4, + "learning_rate": 8.691041027072287e-06, + "loss": 0.8039, + "step": 52802 + }, + { + "epoch": 3.4, + "learning_rate": 8.690826338049336e-06, + "loss": 0.8164, + "step": 52803 + }, + { + "epoch": 3.4, + "learning_rate": 8.690611649026386e-06, + "loss": 0.7917, + "step": 52804 + }, + { + "epoch": 3.4, + "learning_rate": 8.690396960003435e-06, + "loss": 0.862, + "step": 52805 + }, + { + "epoch": 3.4, + "learning_rate": 8.690182270980484e-06, + "loss": 0.8155, + "step": 52806 + }, + { + "epoch": 3.4, + "learning_rate": 8.689967581957535e-06, + "loss": 0.7691, + "step": 52807 + }, + { + "epoch": 3.4, + "learning_rate": 8.689752892934586e-06, + "loss": 0.8508, + "step": 52808 + }, + { + "epoch": 3.4, + "learning_rate": 8.689538203911635e-06, + "loss": 0.824, + "step": 52809 + }, + { + "epoch": 3.4, + "learning_rate": 8.689323514888685e-06, + "loss": 0.8287, + "step": 52810 + }, + { + "epoch": 3.4, + "learning_rate": 8.689108825865734e-06, + "loss": 0.8189, + "step": 52811 + }, + { + "epoch": 3.4, + "learning_rate": 8.688894136842783e-06, + "loss": 0.7875, + "step": 52812 + }, + { + "epoch": 3.4, + "learning_rate": 8.688679447819834e-06, + "loss": 0.8867, + "step": 52813 + }, + { + "epoch": 3.4, + "learning_rate": 8.688464758796883e-06, + "loss": 0.8981, + "step": 52814 + }, + { + "epoch": 3.4, + "learning_rate": 8.688250069773934e-06, + "loss": 0.8188, + "step": 52815 + }, + { + "epoch": 3.4, + "learning_rate": 8.688035380750982e-06, + "loss": 0.8701, + "step": 52816 + }, + { + "epoch": 3.4, + "learning_rate": 8.687820691728033e-06, + "loss": 0.807, + "step": 52817 + }, + { + "epoch": 3.4, + "learning_rate": 8.687606002705082e-06, + "loss": 0.7897, + "step": 52818 + }, + { + "epoch": 3.4, + "learning_rate": 8.687391313682133e-06, + "loss": 0.8547, + "step": 52819 + }, + { + "epoch": 3.4, + "learning_rate": 8.687176624659182e-06, + "loss": 0.7776, + "step": 52820 + }, + { + "epoch": 3.4, + "learning_rate": 8.68696193563623e-06, + "loss": 0.8408, + "step": 52821 + }, + { + "epoch": 3.4, + "learning_rate": 8.686747246613281e-06, + "loss": 0.8196, + "step": 52822 + }, + { + "epoch": 3.4, + "learning_rate": 8.686532557590332e-06, + "loss": 0.8627, + "step": 52823 + }, + { + "epoch": 3.4, + "learning_rate": 8.686317868567381e-06, + "loss": 0.8504, + "step": 52824 + }, + { + "epoch": 3.4, + "learning_rate": 8.68610317954443e-06, + "loss": 0.8388, + "step": 52825 + }, + { + "epoch": 3.4, + "learning_rate": 8.68588849052148e-06, + "loss": 0.8901, + "step": 52826 + }, + { + "epoch": 3.4, + "learning_rate": 8.68567380149853e-06, + "loss": 0.9148, + "step": 52827 + }, + { + "epoch": 3.4, + "learning_rate": 8.68545911247558e-06, + "loss": 0.8518, + "step": 52828 + }, + { + "epoch": 3.4, + "learning_rate": 8.685244423452631e-06, + "loss": 0.8317, + "step": 52829 + }, + { + "epoch": 3.4, + "learning_rate": 8.68502973442968e-06, + "loss": 0.8049, + "step": 52830 + }, + { + "epoch": 3.4, + "learning_rate": 8.684815045406729e-06, + "loss": 0.7985, + "step": 52831 + }, + { + "epoch": 3.4, + "learning_rate": 8.68460035638378e-06, + "loss": 0.7857, + "step": 52832 + }, + { + "epoch": 3.4, + "learning_rate": 8.684385667360828e-06, + "loss": 0.8382, + "step": 52833 + }, + { + "epoch": 3.4, + "learning_rate": 8.684170978337877e-06, + "loss": 0.81, + "step": 52834 + }, + { + "epoch": 3.4, + "learning_rate": 8.683956289314928e-06, + "loss": 0.8043, + "step": 52835 + }, + { + "epoch": 3.4, + "learning_rate": 8.683741600291979e-06, + "loss": 0.7847, + "step": 52836 + }, + { + "epoch": 3.4, + "learning_rate": 8.683526911269028e-06, + "loss": 0.7786, + "step": 52837 + }, + { + "epoch": 3.4, + "learning_rate": 8.683312222246078e-06, + "loss": 0.8817, + "step": 52838 + }, + { + "epoch": 3.4, + "learning_rate": 8.683097533223127e-06, + "loss": 0.821, + "step": 52839 + }, + { + "epoch": 3.4, + "learning_rate": 8.682882844200176e-06, + "loss": 0.8075, + "step": 52840 + }, + { + "epoch": 3.4, + "learning_rate": 8.682668155177227e-06, + "loss": 0.7846, + "step": 52841 + }, + { + "epoch": 3.4, + "learning_rate": 8.682453466154276e-06, + "loss": 0.8726, + "step": 52842 + }, + { + "epoch": 3.4, + "learning_rate": 8.682238777131325e-06, + "loss": 0.8546, + "step": 52843 + }, + { + "epoch": 3.4, + "learning_rate": 8.682024088108376e-06, + "loss": 0.8871, + "step": 52844 + }, + { + "epoch": 3.4, + "learning_rate": 8.681809399085426e-06, + "loss": 0.8417, + "step": 52845 + }, + { + "epoch": 3.4, + "learning_rate": 8.681594710062475e-06, + "loss": 0.7603, + "step": 52846 + }, + { + "epoch": 3.4, + "learning_rate": 8.681380021039524e-06, + "loss": 0.8183, + "step": 52847 + }, + { + "epoch": 3.4, + "learning_rate": 8.681165332016575e-06, + "loss": 0.7499, + "step": 52848 + }, + { + "epoch": 3.4, + "learning_rate": 8.680950642993624e-06, + "loss": 0.831, + "step": 52849 + }, + { + "epoch": 3.4, + "learning_rate": 8.680735953970674e-06, + "loss": 0.8405, + "step": 52850 + }, + { + "epoch": 3.4, + "learning_rate": 8.680521264947725e-06, + "loss": 0.7586, + "step": 52851 + }, + { + "epoch": 3.4, + "learning_rate": 8.680306575924774e-06, + "loss": 0.7747, + "step": 52852 + }, + { + "epoch": 3.4, + "learning_rate": 8.680091886901823e-06, + "loss": 0.8286, + "step": 52853 + }, + { + "epoch": 3.4, + "learning_rate": 8.679877197878874e-06, + "loss": 0.7727, + "step": 52854 + }, + { + "epoch": 3.4, + "learning_rate": 8.679662508855923e-06, + "loss": 0.8036, + "step": 52855 + }, + { + "epoch": 3.4, + "learning_rate": 8.679447819832972e-06, + "loss": 0.8463, + "step": 52856 + }, + { + "epoch": 3.4, + "learning_rate": 8.679233130810022e-06, + "loss": 0.8548, + "step": 52857 + }, + { + "epoch": 3.4, + "learning_rate": 8.679018441787073e-06, + "loss": 0.8308, + "step": 52858 + }, + { + "epoch": 3.4, + "learning_rate": 8.678803752764122e-06, + "loss": 0.8367, + "step": 52859 + }, + { + "epoch": 3.4, + "learning_rate": 8.678589063741173e-06, + "loss": 0.8622, + "step": 52860 + }, + { + "epoch": 3.4, + "learning_rate": 8.678374374718221e-06, + "loss": 0.7999, + "step": 52861 + }, + { + "epoch": 3.4, + "learning_rate": 8.67815968569527e-06, + "loss": 0.8432, + "step": 52862 + }, + { + "epoch": 3.4, + "learning_rate": 8.677944996672321e-06, + "loss": 0.8072, + "step": 52863 + }, + { + "epoch": 3.4, + "learning_rate": 8.67773030764937e-06, + "loss": 0.8246, + "step": 52864 + }, + { + "epoch": 3.4, + "learning_rate": 8.67751561862642e-06, + "loss": 0.8367, + "step": 52865 + }, + { + "epoch": 3.4, + "learning_rate": 8.67730092960347e-06, + "loss": 0.8018, + "step": 52866 + }, + { + "epoch": 3.4, + "learning_rate": 8.67708624058052e-06, + "loss": 0.8371, + "step": 52867 + }, + { + "epoch": 3.4, + "learning_rate": 8.67687155155757e-06, + "loss": 0.8298, + "step": 52868 + }, + { + "epoch": 3.4, + "learning_rate": 8.67665686253462e-06, + "loss": 0.8596, + "step": 52869 + }, + { + "epoch": 3.4, + "learning_rate": 8.676442173511669e-06, + "loss": 0.8288, + "step": 52870 + }, + { + "epoch": 3.4, + "learning_rate": 8.676227484488718e-06, + "loss": 0.7744, + "step": 52871 + }, + { + "epoch": 3.4, + "learning_rate": 8.676012795465769e-06, + "loss": 0.8523, + "step": 52872 + }, + { + "epoch": 3.4, + "learning_rate": 8.67579810644282e-06, + "loss": 0.8906, + "step": 52873 + }, + { + "epoch": 3.4, + "learning_rate": 8.675583417419868e-06, + "loss": 0.84, + "step": 52874 + }, + { + "epoch": 3.4, + "learning_rate": 8.675368728396917e-06, + "loss": 0.8261, + "step": 52875 + }, + { + "epoch": 3.4, + "learning_rate": 8.675154039373968e-06, + "loss": 0.8358, + "step": 52876 + }, + { + "epoch": 3.4, + "learning_rate": 8.674939350351017e-06, + "loss": 0.8011, + "step": 52877 + }, + { + "epoch": 3.4, + "learning_rate": 8.674724661328067e-06, + "loss": 0.8279, + "step": 52878 + }, + { + "epoch": 3.4, + "learning_rate": 8.674509972305116e-06, + "loss": 0.7929, + "step": 52879 + }, + { + "epoch": 3.4, + "learning_rate": 8.674295283282167e-06, + "loss": 0.8695, + "step": 52880 + }, + { + "epoch": 3.4, + "learning_rate": 8.674080594259216e-06, + "loss": 0.8344, + "step": 52881 + }, + { + "epoch": 3.4, + "learning_rate": 8.673865905236267e-06, + "loss": 0.8288, + "step": 52882 + }, + { + "epoch": 3.4, + "learning_rate": 8.673651216213316e-06, + "loss": 0.7665, + "step": 52883 + }, + { + "epoch": 3.4, + "learning_rate": 8.673436527190365e-06, + "loss": 0.7368, + "step": 52884 + }, + { + "epoch": 3.4, + "learning_rate": 8.673221838167415e-06, + "loss": 0.8368, + "step": 52885 + }, + { + "epoch": 3.4, + "learning_rate": 8.673007149144464e-06, + "loss": 0.8093, + "step": 52886 + }, + { + "epoch": 3.4, + "learning_rate": 8.672792460121515e-06, + "loss": 0.8038, + "step": 52887 + }, + { + "epoch": 3.4, + "learning_rate": 8.672577771098566e-06, + "loss": 0.8766, + "step": 52888 + }, + { + "epoch": 3.4, + "learning_rate": 8.672363082075615e-06, + "loss": 0.8401, + "step": 52889 + }, + { + "epoch": 3.4, + "learning_rate": 8.672148393052663e-06, + "loss": 0.815, + "step": 52890 + }, + { + "epoch": 3.4, + "learning_rate": 8.671933704029714e-06, + "loss": 0.8298, + "step": 52891 + }, + { + "epoch": 3.4, + "learning_rate": 8.671719015006763e-06, + "loss": 0.8849, + "step": 52892 + }, + { + "epoch": 3.4, + "learning_rate": 8.671504325983812e-06, + "loss": 0.8304, + "step": 52893 + }, + { + "epoch": 3.4, + "learning_rate": 8.671289636960863e-06, + "loss": 0.8768, + "step": 52894 + }, + { + "epoch": 3.4, + "learning_rate": 8.671074947937913e-06, + "loss": 0.8092, + "step": 52895 + }, + { + "epoch": 3.4, + "learning_rate": 8.670860258914962e-06, + "loss": 0.8403, + "step": 52896 + }, + { + "epoch": 3.4, + "learning_rate": 8.670645569892013e-06, + "loss": 0.8783, + "step": 52897 + }, + { + "epoch": 3.4, + "learning_rate": 8.670430880869062e-06, + "loss": 0.829, + "step": 52898 + }, + { + "epoch": 3.4, + "learning_rate": 8.670216191846111e-06, + "loss": 0.8867, + "step": 52899 + }, + { + "epoch": 3.4, + "learning_rate": 8.670001502823162e-06, + "loss": 0.7857, + "step": 52900 + }, + { + "epoch": 3.4, + "learning_rate": 8.66978681380021e-06, + "loss": 0.8527, + "step": 52901 + }, + { + "epoch": 3.4, + "learning_rate": 8.669572124777261e-06, + "loss": 0.8215, + "step": 52902 + }, + { + "epoch": 3.4, + "learning_rate": 8.66935743575431e-06, + "loss": 0.8164, + "step": 52903 + }, + { + "epoch": 3.4, + "learning_rate": 8.66914274673136e-06, + "loss": 0.8844, + "step": 52904 + }, + { + "epoch": 3.4, + "learning_rate": 8.66892805770841e-06, + "loss": 0.9253, + "step": 52905 + }, + { + "epoch": 3.4, + "learning_rate": 8.66871336868546e-06, + "loss": 0.7432, + "step": 52906 + }, + { + "epoch": 3.4, + "learning_rate": 8.66849867966251e-06, + "loss": 0.8018, + "step": 52907 + }, + { + "epoch": 3.4, + "learning_rate": 8.668283990639558e-06, + "loss": 0.8175, + "step": 52908 + }, + { + "epoch": 3.4, + "learning_rate": 8.668069301616609e-06, + "loss": 0.8369, + "step": 52909 + }, + { + "epoch": 3.4, + "learning_rate": 8.66785461259366e-06, + "loss": 0.8026, + "step": 52910 + }, + { + "epoch": 3.4, + "learning_rate": 8.667639923570709e-06, + "loss": 0.8114, + "step": 52911 + }, + { + "epoch": 3.4, + "learning_rate": 8.667425234547758e-06, + "loss": 0.8094, + "step": 52912 + }, + { + "epoch": 3.4, + "learning_rate": 8.667210545524808e-06, + "loss": 0.8667, + "step": 52913 + }, + { + "epoch": 3.4, + "learning_rate": 8.666995856501857e-06, + "loss": 0.8468, + "step": 52914 + }, + { + "epoch": 3.4, + "learning_rate": 8.666781167478906e-06, + "loss": 0.7774, + "step": 52915 + }, + { + "epoch": 3.4, + "learning_rate": 8.666566478455959e-06, + "loss": 0.8097, + "step": 52916 + }, + { + "epoch": 3.4, + "learning_rate": 8.666351789433008e-06, + "loss": 0.8881, + "step": 52917 + }, + { + "epoch": 3.4, + "learning_rate": 8.666137100410057e-06, + "loss": 0.7901, + "step": 52918 + }, + { + "epoch": 3.4, + "learning_rate": 8.665922411387107e-06, + "loss": 0.7921, + "step": 52919 + }, + { + "epoch": 3.4, + "learning_rate": 8.665707722364156e-06, + "loss": 0.879, + "step": 52920 + }, + { + "epoch": 3.4, + "learning_rate": 8.665493033341205e-06, + "loss": 0.7553, + "step": 52921 + }, + { + "epoch": 3.4, + "learning_rate": 8.665278344318256e-06, + "loss": 0.8474, + "step": 52922 + }, + { + "epoch": 3.4, + "learning_rate": 8.665063655295305e-06, + "loss": 0.8326, + "step": 52923 + }, + { + "epoch": 3.41, + "learning_rate": 8.664848966272355e-06, + "loss": 0.8517, + "step": 52924 + }, + { + "epoch": 3.41, + "learning_rate": 8.664634277249406e-06, + "loss": 0.8353, + "step": 52925 + }, + { + "epoch": 3.41, + "learning_rate": 8.664419588226455e-06, + "loss": 0.7889, + "step": 52926 + }, + { + "epoch": 3.41, + "learning_rate": 8.664204899203504e-06, + "loss": 0.836, + "step": 52927 + }, + { + "epoch": 3.41, + "learning_rate": 8.663990210180555e-06, + "loss": 0.8965, + "step": 52928 + }, + { + "epoch": 3.41, + "learning_rate": 8.663775521157604e-06, + "loss": 0.8261, + "step": 52929 + }, + { + "epoch": 3.41, + "learning_rate": 8.663560832134653e-06, + "loss": 0.773, + "step": 52930 + }, + { + "epoch": 3.41, + "learning_rate": 8.663346143111703e-06, + "loss": 0.8484, + "step": 52931 + }, + { + "epoch": 3.41, + "learning_rate": 8.663131454088754e-06, + "loss": 0.8594, + "step": 52932 + }, + { + "epoch": 3.41, + "learning_rate": 8.662916765065803e-06, + "loss": 0.8354, + "step": 52933 + }, + { + "epoch": 3.41, + "learning_rate": 8.662702076042852e-06, + "loss": 0.8243, + "step": 52934 + }, + { + "epoch": 3.41, + "learning_rate": 8.662487387019902e-06, + "loss": 0.8638, + "step": 52935 + }, + { + "epoch": 3.41, + "learning_rate": 8.662272697996951e-06, + "loss": 0.8445, + "step": 52936 + }, + { + "epoch": 3.41, + "learning_rate": 8.662058008974002e-06, + "loss": 0.8157, + "step": 52937 + }, + { + "epoch": 3.41, + "learning_rate": 8.661843319951053e-06, + "loss": 0.8246, + "step": 52938 + }, + { + "epoch": 3.41, + "learning_rate": 8.661628630928102e-06, + "loss": 0.8007, + "step": 52939 + }, + { + "epoch": 3.41, + "learning_rate": 8.66141394190515e-06, + "loss": 0.8561, + "step": 52940 + }, + { + "epoch": 3.41, + "learning_rate": 8.661199252882201e-06, + "loss": 0.8587, + "step": 52941 + }, + { + "epoch": 3.41, + "learning_rate": 8.66098456385925e-06, + "loss": 0.7895, + "step": 52942 + }, + { + "epoch": 3.41, + "learning_rate": 8.6607698748363e-06, + "loss": 0.8576, + "step": 52943 + }, + { + "epoch": 3.41, + "learning_rate": 8.66055518581335e-06, + "loss": 0.8573, + "step": 52944 + }, + { + "epoch": 3.41, + "learning_rate": 8.6603404967904e-06, + "loss": 0.8652, + "step": 52945 + }, + { + "epoch": 3.41, + "learning_rate": 8.66012580776745e-06, + "loss": 0.8411, + "step": 52946 + }, + { + "epoch": 3.41, + "learning_rate": 8.6599111187445e-06, + "loss": 0.8132, + "step": 52947 + }, + { + "epoch": 3.41, + "learning_rate": 8.65969642972155e-06, + "loss": 0.852, + "step": 52948 + }, + { + "epoch": 3.41, + "learning_rate": 8.659481740698598e-06, + "loss": 0.8086, + "step": 52949 + }, + { + "epoch": 3.41, + "learning_rate": 8.659267051675649e-06, + "loss": 0.7897, + "step": 52950 + }, + { + "epoch": 3.41, + "learning_rate": 8.659052362652698e-06, + "loss": 0.8031, + "step": 52951 + }, + { + "epoch": 3.41, + "learning_rate": 8.658837673629747e-06, + "loss": 0.847, + "step": 52952 + }, + { + "epoch": 3.41, + "learning_rate": 8.658622984606797e-06, + "loss": 0.8208, + "step": 52953 + }, + { + "epoch": 3.41, + "learning_rate": 8.658408295583848e-06, + "loss": 0.8535, + "step": 52954 + }, + { + "epoch": 3.41, + "learning_rate": 8.658193606560897e-06, + "loss": 0.8163, + "step": 52955 + }, + { + "epoch": 3.41, + "learning_rate": 8.657978917537948e-06, + "loss": 0.8654, + "step": 52956 + }, + { + "epoch": 3.41, + "learning_rate": 8.657764228514997e-06, + "loss": 0.8539, + "step": 52957 + }, + { + "epoch": 3.41, + "learning_rate": 8.657549539492046e-06, + "loss": 0.7816, + "step": 52958 + }, + { + "epoch": 3.41, + "learning_rate": 8.657334850469096e-06, + "loss": 0.8725, + "step": 52959 + }, + { + "epoch": 3.41, + "learning_rate": 8.657120161446147e-06, + "loss": 0.7481, + "step": 52960 + }, + { + "epoch": 3.41, + "learning_rate": 8.656905472423196e-06, + "loss": 0.7917, + "step": 52961 + }, + { + "epoch": 3.41, + "learning_rate": 8.656690783400245e-06, + "loss": 0.8251, + "step": 52962 + }, + { + "epoch": 3.41, + "learning_rate": 8.656476094377296e-06, + "loss": 0.7598, + "step": 52963 + }, + { + "epoch": 3.41, + "learning_rate": 8.656261405354344e-06, + "loss": 0.8906, + "step": 52964 + }, + { + "epoch": 3.41, + "learning_rate": 8.656046716331395e-06, + "loss": 0.8434, + "step": 52965 + }, + { + "epoch": 3.41, + "learning_rate": 8.655832027308444e-06, + "loss": 0.7676, + "step": 52966 + }, + { + "epoch": 3.41, + "learning_rate": 8.655617338285495e-06, + "loss": 0.8846, + "step": 52967 + }, + { + "epoch": 3.41, + "learning_rate": 8.655402649262544e-06, + "loss": 0.768, + "step": 52968 + }, + { + "epoch": 3.41, + "learning_rate": 8.655187960239594e-06, + "loss": 0.8035, + "step": 52969 + }, + { + "epoch": 3.41, + "learning_rate": 8.654973271216643e-06, + "loss": 0.8324, + "step": 52970 + }, + { + "epoch": 3.41, + "learning_rate": 8.654758582193692e-06, + "loss": 0.7429, + "step": 52971 + }, + { + "epoch": 3.41, + "learning_rate": 8.654543893170743e-06, + "loss": 0.7575, + "step": 52972 + }, + { + "epoch": 3.41, + "learning_rate": 8.654329204147792e-06, + "loss": 0.8055, + "step": 52973 + }, + { + "epoch": 3.41, + "learning_rate": 8.654114515124843e-06, + "loss": 0.8, + "step": 52974 + }, + { + "epoch": 3.41, + "learning_rate": 8.653899826101893e-06, + "loss": 0.8535, + "step": 52975 + }, + { + "epoch": 3.41, + "learning_rate": 8.653685137078942e-06, + "loss": 0.7259, + "step": 52976 + }, + { + "epoch": 3.41, + "learning_rate": 8.653470448055991e-06, + "loss": 0.8124, + "step": 52977 + }, + { + "epoch": 3.41, + "learning_rate": 8.653255759033042e-06, + "loss": 0.8058, + "step": 52978 + }, + { + "epoch": 3.41, + "learning_rate": 8.65304107001009e-06, + "loss": 0.8791, + "step": 52979 + }, + { + "epoch": 3.41, + "learning_rate": 8.65282638098714e-06, + "loss": 0.8342, + "step": 52980 + }, + { + "epoch": 3.41, + "learning_rate": 8.65261169196419e-06, + "loss": 0.8112, + "step": 52981 + }, + { + "epoch": 3.41, + "learning_rate": 8.652397002941241e-06, + "loss": 0.8074, + "step": 52982 + }, + { + "epoch": 3.41, + "learning_rate": 8.65218231391829e-06, + "loss": 0.8197, + "step": 52983 + }, + { + "epoch": 3.41, + "learning_rate": 8.65196762489534e-06, + "loss": 0.8033, + "step": 52984 + }, + { + "epoch": 3.41, + "learning_rate": 8.65175293587239e-06, + "loss": 0.84, + "step": 52985 + }, + { + "epoch": 3.41, + "learning_rate": 8.651538246849439e-06, + "loss": 0.7645, + "step": 52986 + }, + { + "epoch": 3.41, + "learning_rate": 8.65132355782649e-06, + "loss": 0.7902, + "step": 52987 + }, + { + "epoch": 3.41, + "learning_rate": 8.651108868803538e-06, + "loss": 0.8719, + "step": 52988 + }, + { + "epoch": 3.41, + "learning_rate": 8.650894179780589e-06, + "loss": 0.8235, + "step": 52989 + }, + { + "epoch": 3.41, + "learning_rate": 8.650679490757638e-06, + "loss": 0.7993, + "step": 52990 + }, + { + "epoch": 3.41, + "learning_rate": 8.650464801734689e-06, + "loss": 0.7871, + "step": 52991 + }, + { + "epoch": 3.41, + "learning_rate": 8.650250112711738e-06, + "loss": 0.7965, + "step": 52992 + }, + { + "epoch": 3.41, + "learning_rate": 8.650035423688788e-06, + "loss": 0.8487, + "step": 52993 + }, + { + "epoch": 3.41, + "learning_rate": 8.649820734665837e-06, + "loss": 0.8271, + "step": 52994 + }, + { + "epoch": 3.41, + "learning_rate": 8.649606045642886e-06, + "loss": 0.7948, + "step": 52995 + }, + { + "epoch": 3.41, + "learning_rate": 8.649391356619937e-06, + "loss": 0.8278, + "step": 52996 + }, + { + "epoch": 3.41, + "learning_rate": 8.649176667596987e-06, + "loss": 0.7692, + "step": 52997 + }, + { + "epoch": 3.41, + "learning_rate": 8.648961978574036e-06, + "loss": 0.8942, + "step": 52998 + }, + { + "epoch": 3.41, + "learning_rate": 8.648747289551085e-06, + "loss": 0.8606, + "step": 52999 + }, + { + "epoch": 3.41, + "learning_rate": 8.648532600528136e-06, + "loss": 0.8282, + "step": 53000 + }, + { + "epoch": 3.41, + "eval_laion/cider": 0.5908470502598023, + "eval_laion/meteor": 0.39980469878172975, + "eval_laion/pred_num_words": 37.265625, + "eval_laion/sacrebleu": 16.249522651585558, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.055523157119751, + "eval_runtime": 98.6714, + "eval_samples_per_second": 5.057, + "eval_steps_per_second": 0.324, + "step": 53000 + }, + { + "epoch": 3.41, + "learning_rate": 8.648317911505185e-06, + "loss": 0.8386, + "step": 53001 + }, + { + "epoch": 3.41, + "learning_rate": 8.648103222482234e-06, + "loss": 0.8259, + "step": 53002 + }, + { + "epoch": 3.41, + "learning_rate": 8.647888533459286e-06, + "loss": 0.8426, + "step": 53003 + }, + { + "epoch": 3.41, + "learning_rate": 8.647673844436335e-06, + "loss": 0.8206, + "step": 53004 + }, + { + "epoch": 3.41, + "learning_rate": 8.647459155413384e-06, + "loss": 0.7958, + "step": 53005 + }, + { + "epoch": 3.41, + "learning_rate": 8.647244466390435e-06, + "loss": 0.8261, + "step": 53006 + }, + { + "epoch": 3.41, + "learning_rate": 8.647029777367484e-06, + "loss": 0.8064, + "step": 53007 + }, + { + "epoch": 3.41, + "learning_rate": 8.646815088344533e-06, + "loss": 0.8826, + "step": 53008 + }, + { + "epoch": 3.41, + "learning_rate": 8.646600399321583e-06, + "loss": 0.8646, + "step": 53009 + }, + { + "epoch": 3.41, + "learning_rate": 8.646385710298632e-06, + "loss": 0.799, + "step": 53010 + }, + { + "epoch": 3.41, + "learning_rate": 8.646171021275683e-06, + "loss": 0.8654, + "step": 53011 + }, + { + "epoch": 3.41, + "learning_rate": 8.645956332252732e-06, + "loss": 0.7749, + "step": 53012 + }, + { + "epoch": 3.41, + "learning_rate": 8.645741643229783e-06, + "loss": 0.8426, + "step": 53013 + }, + { + "epoch": 3.41, + "learning_rate": 8.645526954206832e-06, + "loss": 0.7828, + "step": 53014 + }, + { + "epoch": 3.41, + "learning_rate": 8.645312265183882e-06, + "loss": 0.786, + "step": 53015 + }, + { + "epoch": 3.41, + "learning_rate": 8.645097576160931e-06, + "loss": 0.8282, + "step": 53016 + }, + { + "epoch": 3.41, + "learning_rate": 8.64488288713798e-06, + "loss": 0.876, + "step": 53017 + }, + { + "epoch": 3.41, + "learning_rate": 8.644668198115031e-06, + "loss": 0.8331, + "step": 53018 + }, + { + "epoch": 3.41, + "learning_rate": 8.644453509092082e-06, + "loss": 0.8244, + "step": 53019 + }, + { + "epoch": 3.41, + "learning_rate": 8.64423882006913e-06, + "loss": 0.7998, + "step": 53020 + }, + { + "epoch": 3.41, + "learning_rate": 8.64402413104618e-06, + "loss": 0.7795, + "step": 53021 + }, + { + "epoch": 3.41, + "learning_rate": 8.64380944202323e-06, + "loss": 0.8145, + "step": 53022 + }, + { + "epoch": 3.41, + "learning_rate": 8.643594753000279e-06, + "loss": 0.8592, + "step": 53023 + }, + { + "epoch": 3.41, + "learning_rate": 8.64338006397733e-06, + "loss": 0.8852, + "step": 53024 + }, + { + "epoch": 3.41, + "learning_rate": 8.64316537495438e-06, + "loss": 0.8314, + "step": 53025 + }, + { + "epoch": 3.41, + "learning_rate": 8.64295068593143e-06, + "loss": 0.8623, + "step": 53026 + }, + { + "epoch": 3.41, + "learning_rate": 8.642735996908478e-06, + "loss": 0.8291, + "step": 53027 + }, + { + "epoch": 3.41, + "learning_rate": 8.642521307885529e-06, + "loss": 0.7615, + "step": 53028 + }, + { + "epoch": 3.41, + "learning_rate": 8.642306618862578e-06, + "loss": 0.7999, + "step": 53029 + }, + { + "epoch": 3.41, + "learning_rate": 8.642091929839627e-06, + "loss": 0.8542, + "step": 53030 + }, + { + "epoch": 3.41, + "learning_rate": 8.641877240816678e-06, + "loss": 0.7896, + "step": 53031 + }, + { + "epoch": 3.41, + "learning_rate": 8.641662551793728e-06, + "loss": 0.7849, + "step": 53032 + }, + { + "epoch": 3.41, + "learning_rate": 8.641447862770777e-06, + "loss": 0.8119, + "step": 53033 + }, + { + "epoch": 3.41, + "learning_rate": 8.641233173747828e-06, + "loss": 0.8308, + "step": 53034 + }, + { + "epoch": 3.41, + "learning_rate": 8.641018484724877e-06, + "loss": 0.801, + "step": 53035 + }, + { + "epoch": 3.41, + "learning_rate": 8.640803795701926e-06, + "loss": 0.795, + "step": 53036 + }, + { + "epoch": 3.41, + "learning_rate": 8.640589106678977e-06, + "loss": 0.8416, + "step": 53037 + }, + { + "epoch": 3.41, + "learning_rate": 8.640374417656025e-06, + "loss": 0.8189, + "step": 53038 + }, + { + "epoch": 3.41, + "learning_rate": 8.640159728633074e-06, + "loss": 0.8544, + "step": 53039 + }, + { + "epoch": 3.41, + "learning_rate": 8.639945039610125e-06, + "loss": 0.8463, + "step": 53040 + }, + { + "epoch": 3.41, + "learning_rate": 8.639730350587176e-06, + "loss": 0.8041, + "step": 53041 + }, + { + "epoch": 3.41, + "learning_rate": 8.639515661564225e-06, + "loss": 0.8339, + "step": 53042 + }, + { + "epoch": 3.41, + "learning_rate": 8.639300972541275e-06, + "loss": 0.8591, + "step": 53043 + }, + { + "epoch": 3.41, + "learning_rate": 8.639086283518324e-06, + "loss": 0.8827, + "step": 53044 + }, + { + "epoch": 3.41, + "learning_rate": 8.638871594495373e-06, + "loss": 0.7978, + "step": 53045 + }, + { + "epoch": 3.41, + "learning_rate": 8.638656905472424e-06, + "loss": 0.8252, + "step": 53046 + }, + { + "epoch": 3.41, + "learning_rate": 8.638442216449475e-06, + "loss": 0.7828, + "step": 53047 + }, + { + "epoch": 3.41, + "learning_rate": 8.638227527426524e-06, + "loss": 0.8798, + "step": 53048 + }, + { + "epoch": 3.41, + "learning_rate": 8.638012838403573e-06, + "loss": 0.8593, + "step": 53049 + }, + { + "epoch": 3.41, + "learning_rate": 8.637798149380623e-06, + "loss": 0.8479, + "step": 53050 + }, + { + "epoch": 3.41, + "learning_rate": 8.637583460357672e-06, + "loss": 0.8015, + "step": 53051 + }, + { + "epoch": 3.41, + "learning_rate": 8.637368771334723e-06, + "loss": 0.7939, + "step": 53052 + }, + { + "epoch": 3.41, + "learning_rate": 8.637154082311772e-06, + "loss": 0.8375, + "step": 53053 + }, + { + "epoch": 3.41, + "learning_rate": 8.636939393288822e-06, + "loss": 0.8281, + "step": 53054 + }, + { + "epoch": 3.41, + "learning_rate": 8.636724704265871e-06, + "loss": 0.8708, + "step": 53055 + }, + { + "epoch": 3.41, + "learning_rate": 8.636510015242922e-06, + "loss": 0.8298, + "step": 53056 + }, + { + "epoch": 3.41, + "learning_rate": 8.636295326219971e-06, + "loss": 0.8244, + "step": 53057 + }, + { + "epoch": 3.41, + "learning_rate": 8.63608063719702e-06, + "loss": 0.8603, + "step": 53058 + }, + { + "epoch": 3.41, + "learning_rate": 8.63586594817407e-06, + "loss": 0.7885, + "step": 53059 + }, + { + "epoch": 3.41, + "learning_rate": 8.63565125915112e-06, + "loss": 0.833, + "step": 53060 + }, + { + "epoch": 3.41, + "learning_rate": 8.63543657012817e-06, + "loss": 0.8426, + "step": 53061 + }, + { + "epoch": 3.41, + "learning_rate": 8.635221881105221e-06, + "loss": 0.8031, + "step": 53062 + }, + { + "epoch": 3.41, + "learning_rate": 8.63500719208227e-06, + "loss": 0.8326, + "step": 53063 + }, + { + "epoch": 3.41, + "learning_rate": 8.634792503059319e-06, + "loss": 0.8822, + "step": 53064 + }, + { + "epoch": 3.41, + "learning_rate": 8.63457781403637e-06, + "loss": 0.8004, + "step": 53065 + }, + { + "epoch": 3.41, + "learning_rate": 8.634363125013419e-06, + "loss": 0.8055, + "step": 53066 + }, + { + "epoch": 3.41, + "learning_rate": 8.634148435990467e-06, + "loss": 0.8537, + "step": 53067 + }, + { + "epoch": 3.41, + "learning_rate": 8.633933746967518e-06, + "loss": 0.8117, + "step": 53068 + }, + { + "epoch": 3.41, + "learning_rate": 8.633719057944569e-06, + "loss": 0.8422, + "step": 53069 + }, + { + "epoch": 3.41, + "learning_rate": 8.633504368921618e-06, + "loss": 0.7902, + "step": 53070 + }, + { + "epoch": 3.41, + "learning_rate": 8.633289679898668e-06, + "loss": 0.7819, + "step": 53071 + }, + { + "epoch": 3.41, + "learning_rate": 8.633074990875717e-06, + "loss": 0.8194, + "step": 53072 + }, + { + "epoch": 3.41, + "learning_rate": 8.632860301852766e-06, + "loss": 0.7938, + "step": 53073 + }, + { + "epoch": 3.41, + "learning_rate": 8.632645612829817e-06, + "loss": 0.8533, + "step": 53074 + }, + { + "epoch": 3.41, + "learning_rate": 8.632430923806866e-06, + "loss": 0.857, + "step": 53075 + }, + { + "epoch": 3.41, + "learning_rate": 8.632216234783917e-06, + "loss": 0.8008, + "step": 53076 + }, + { + "epoch": 3.41, + "learning_rate": 8.632001545760966e-06, + "loss": 0.8119, + "step": 53077 + }, + { + "epoch": 3.41, + "learning_rate": 8.631786856738016e-06, + "loss": 0.8272, + "step": 53078 + }, + { + "epoch": 3.41, + "learning_rate": 8.631572167715065e-06, + "loss": 0.8111, + "step": 53079 + }, + { + "epoch": 3.42, + "learning_rate": 8.631357478692114e-06, + "loss": 0.8153, + "step": 53080 + }, + { + "epoch": 3.42, + "learning_rate": 8.631142789669165e-06, + "loss": 0.7826, + "step": 53081 + }, + { + "epoch": 3.42, + "learning_rate": 8.630928100646214e-06, + "loss": 0.7931, + "step": 53082 + }, + { + "epoch": 3.42, + "learning_rate": 8.630713411623264e-06, + "loss": 0.8468, + "step": 53083 + }, + { + "epoch": 3.42, + "learning_rate": 8.630498722600315e-06, + "loss": 0.8566, + "step": 53084 + }, + { + "epoch": 3.42, + "learning_rate": 8.630284033577364e-06, + "loss": 0.788, + "step": 53085 + }, + { + "epoch": 3.42, + "learning_rate": 8.630069344554413e-06, + "loss": 0.8138, + "step": 53086 + }, + { + "epoch": 3.42, + "learning_rate": 8.629854655531464e-06, + "loss": 0.8341, + "step": 53087 + }, + { + "epoch": 3.42, + "learning_rate": 8.629639966508513e-06, + "loss": 0.82, + "step": 53088 + }, + { + "epoch": 3.42, + "learning_rate": 8.629425277485562e-06, + "loss": 0.8509, + "step": 53089 + }, + { + "epoch": 3.42, + "learning_rate": 8.629210588462612e-06, + "loss": 0.7448, + "step": 53090 + }, + { + "epoch": 3.42, + "learning_rate": 8.628995899439663e-06, + "loss": 0.8273, + "step": 53091 + }, + { + "epoch": 3.42, + "learning_rate": 8.628781210416712e-06, + "loss": 0.8579, + "step": 53092 + }, + { + "epoch": 3.42, + "learning_rate": 8.628566521393763e-06, + "loss": 0.7789, + "step": 53093 + }, + { + "epoch": 3.42, + "learning_rate": 8.628351832370812e-06, + "loss": 0.7996, + "step": 53094 + }, + { + "epoch": 3.42, + "learning_rate": 8.62813714334786e-06, + "loss": 0.8493, + "step": 53095 + }, + { + "epoch": 3.42, + "learning_rate": 8.627922454324911e-06, + "loss": 0.8012, + "step": 53096 + }, + { + "epoch": 3.42, + "learning_rate": 8.62770776530196e-06, + "loss": 0.8205, + "step": 53097 + }, + { + "epoch": 3.42, + "learning_rate": 8.62749307627901e-06, + "loss": 0.8403, + "step": 53098 + }, + { + "epoch": 3.42, + "learning_rate": 8.62727838725606e-06, + "loss": 0.8486, + "step": 53099 + }, + { + "epoch": 3.42, + "learning_rate": 8.62706369823311e-06, + "loss": 0.8384, + "step": 53100 + }, + { + "epoch": 3.42, + "learning_rate": 8.62684900921016e-06, + "loss": 0.7936, + "step": 53101 + }, + { + "epoch": 3.42, + "learning_rate": 8.62663432018721e-06, + "loss": 0.8236, + "step": 53102 + }, + { + "epoch": 3.42, + "learning_rate": 8.626419631164259e-06, + "loss": 0.7482, + "step": 53103 + }, + { + "epoch": 3.42, + "learning_rate": 8.626204942141308e-06, + "loss": 0.7883, + "step": 53104 + }, + { + "epoch": 3.42, + "learning_rate": 8.625990253118359e-06, + "loss": 0.8586, + "step": 53105 + }, + { + "epoch": 3.42, + "learning_rate": 8.62577556409541e-06, + "loss": 0.7966, + "step": 53106 + }, + { + "epoch": 3.42, + "learning_rate": 8.625560875072458e-06, + "loss": 0.9235, + "step": 53107 + }, + { + "epoch": 3.42, + "learning_rate": 8.625346186049507e-06, + "loss": 0.8391, + "step": 53108 + }, + { + "epoch": 3.42, + "learning_rate": 8.625131497026558e-06, + "loss": 0.851, + "step": 53109 + }, + { + "epoch": 3.42, + "learning_rate": 8.624916808003607e-06, + "loss": 0.8408, + "step": 53110 + }, + { + "epoch": 3.42, + "learning_rate": 8.624702118980658e-06, + "loss": 0.8329, + "step": 53111 + }, + { + "epoch": 3.42, + "learning_rate": 8.624487429957708e-06, + "loss": 0.7968, + "step": 53112 + }, + { + "epoch": 3.42, + "learning_rate": 8.624272740934757e-06, + "loss": 0.7438, + "step": 53113 + }, + { + "epoch": 3.42, + "learning_rate": 8.624058051911806e-06, + "loss": 0.8352, + "step": 53114 + }, + { + "epoch": 3.42, + "learning_rate": 8.623843362888857e-06, + "loss": 0.7725, + "step": 53115 + }, + { + "epoch": 3.42, + "learning_rate": 8.623628673865906e-06, + "loss": 0.8711, + "step": 53116 + }, + { + "epoch": 3.42, + "learning_rate": 8.623413984842955e-06, + "loss": 0.8439, + "step": 53117 + }, + { + "epoch": 3.42, + "learning_rate": 8.623199295820005e-06, + "loss": 0.8521, + "step": 53118 + }, + { + "epoch": 3.42, + "learning_rate": 8.622984606797054e-06, + "loss": 0.807, + "step": 53119 + }, + { + "epoch": 3.42, + "learning_rate": 8.622769917774105e-06, + "loss": 0.8413, + "step": 53120 + }, + { + "epoch": 3.42, + "learning_rate": 8.622555228751156e-06, + "loss": 0.8169, + "step": 53121 + }, + { + "epoch": 3.42, + "learning_rate": 8.622340539728205e-06, + "loss": 0.8135, + "step": 53122 + }, + { + "epoch": 3.42, + "learning_rate": 8.622125850705254e-06, + "loss": 0.8347, + "step": 53123 + }, + { + "epoch": 3.42, + "learning_rate": 8.621911161682304e-06, + "loss": 0.8555, + "step": 53124 + }, + { + "epoch": 3.42, + "learning_rate": 8.621696472659353e-06, + "loss": 0.7933, + "step": 53125 + }, + { + "epoch": 3.42, + "learning_rate": 8.621481783636402e-06, + "loss": 0.8381, + "step": 53126 + }, + { + "epoch": 3.42, + "learning_rate": 8.621267094613453e-06, + "loss": 0.8035, + "step": 53127 + }, + { + "epoch": 3.42, + "learning_rate": 8.621052405590503e-06, + "loss": 0.7383, + "step": 53128 + }, + { + "epoch": 3.42, + "learning_rate": 8.620837716567552e-06, + "loss": 0.7812, + "step": 53129 + }, + { + "epoch": 3.42, + "learning_rate": 8.620623027544603e-06, + "loss": 0.8222, + "step": 53130 + }, + { + "epoch": 3.42, + "learning_rate": 8.620408338521652e-06, + "loss": 0.9039, + "step": 53131 + }, + { + "epoch": 3.42, + "learning_rate": 8.620193649498701e-06, + "loss": 0.8136, + "step": 53132 + }, + { + "epoch": 3.42, + "learning_rate": 8.619978960475752e-06, + "loss": 0.9016, + "step": 53133 + }, + { + "epoch": 3.42, + "learning_rate": 8.619764271452802e-06, + "loss": 0.8147, + "step": 53134 + }, + { + "epoch": 3.42, + "learning_rate": 8.619549582429851e-06, + "loss": 0.8018, + "step": 53135 + }, + { + "epoch": 3.42, + "learning_rate": 8.6193348934069e-06, + "loss": 0.8984, + "step": 53136 + }, + { + "epoch": 3.42, + "learning_rate": 8.619120204383951e-06, + "loss": 0.7349, + "step": 53137 + }, + { + "epoch": 3.42, + "learning_rate": 8.618905515361e-06, + "loss": 0.8165, + "step": 53138 + }, + { + "epoch": 3.42, + "learning_rate": 8.61869082633805e-06, + "loss": 0.8783, + "step": 53139 + }, + { + "epoch": 3.42, + "learning_rate": 8.6184761373151e-06, + "loss": 0.8692, + "step": 53140 + }, + { + "epoch": 3.42, + "learning_rate": 8.61826144829215e-06, + "loss": 0.7995, + "step": 53141 + }, + { + "epoch": 3.42, + "learning_rate": 8.618046759269199e-06, + "loss": 0.797, + "step": 53142 + }, + { + "epoch": 3.42, + "learning_rate": 8.61783207024625e-06, + "loss": 0.8311, + "step": 53143 + }, + { + "epoch": 3.42, + "learning_rate": 8.617617381223299e-06, + "loss": 0.8604, + "step": 53144 + }, + { + "epoch": 3.42, + "learning_rate": 8.617402692200348e-06, + "loss": 0.7573, + "step": 53145 + }, + { + "epoch": 3.42, + "learning_rate": 8.617188003177398e-06, + "loss": 0.8681, + "step": 53146 + }, + { + "epoch": 3.42, + "learning_rate": 8.616973314154447e-06, + "loss": 0.871, + "step": 53147 + }, + { + "epoch": 3.42, + "learning_rate": 8.616758625131498e-06, + "loss": 0.8107, + "step": 53148 + }, + { + "epoch": 3.42, + "learning_rate": 8.616543936108549e-06, + "loss": 0.8638, + "step": 53149 + }, + { + "epoch": 3.42, + "learning_rate": 8.616329247085598e-06, + "loss": 0.8166, + "step": 53150 + }, + { + "epoch": 3.42, + "learning_rate": 8.616114558062647e-06, + "loss": 0.8463, + "step": 53151 + }, + { + "epoch": 3.42, + "learning_rate": 8.615899869039697e-06, + "loss": 0.8064, + "step": 53152 + }, + { + "epoch": 3.42, + "learning_rate": 8.615685180016746e-06, + "loss": 0.8742, + "step": 53153 + }, + { + "epoch": 3.42, + "learning_rate": 8.615470490993795e-06, + "loss": 0.8062, + "step": 53154 + }, + { + "epoch": 3.42, + "learning_rate": 8.615255801970846e-06, + "loss": 0.7788, + "step": 53155 + }, + { + "epoch": 3.42, + "learning_rate": 8.615041112947897e-06, + "loss": 0.8099, + "step": 53156 + }, + { + "epoch": 3.42, + "learning_rate": 8.614826423924945e-06, + "loss": 0.8343, + "step": 53157 + }, + { + "epoch": 3.42, + "learning_rate": 8.614611734901996e-06, + "loss": 0.8351, + "step": 53158 + }, + { + "epoch": 3.42, + "learning_rate": 8.614397045879045e-06, + "loss": 0.8153, + "step": 53159 + }, + { + "epoch": 3.42, + "learning_rate": 8.614182356856094e-06, + "loss": 0.8009, + "step": 53160 + }, + { + "epoch": 3.42, + "learning_rate": 8.613967667833145e-06, + "loss": 0.8188, + "step": 53161 + }, + { + "epoch": 3.42, + "learning_rate": 8.613752978810194e-06, + "loss": 0.8644, + "step": 53162 + }, + { + "epoch": 3.42, + "learning_rate": 8.613538289787244e-06, + "loss": 0.8166, + "step": 53163 + }, + { + "epoch": 3.42, + "learning_rate": 8.613323600764293e-06, + "loss": 0.8217, + "step": 53164 + }, + { + "epoch": 3.42, + "learning_rate": 8.613108911741344e-06, + "loss": 0.7931, + "step": 53165 + }, + { + "epoch": 3.42, + "learning_rate": 8.612894222718393e-06, + "loss": 0.7336, + "step": 53166 + }, + { + "epoch": 3.42, + "learning_rate": 8.612679533695442e-06, + "loss": 0.7759, + "step": 53167 + }, + { + "epoch": 3.42, + "learning_rate": 8.612464844672493e-06, + "loss": 0.8937, + "step": 53168 + }, + { + "epoch": 3.42, + "learning_rate": 8.612250155649542e-06, + "loss": 0.8777, + "step": 53169 + }, + { + "epoch": 3.42, + "learning_rate": 8.612035466626592e-06, + "loss": 0.7999, + "step": 53170 + }, + { + "epoch": 3.42, + "learning_rate": 8.611820777603643e-06, + "loss": 0.8297, + "step": 53171 + }, + { + "epoch": 3.42, + "learning_rate": 8.611606088580692e-06, + "loss": 0.8635, + "step": 53172 + }, + { + "epoch": 3.42, + "learning_rate": 8.61139139955774e-06, + "loss": 0.8446, + "step": 53173 + }, + { + "epoch": 3.42, + "learning_rate": 8.611176710534791e-06, + "loss": 0.8173, + "step": 53174 + }, + { + "epoch": 3.42, + "learning_rate": 8.61096202151184e-06, + "loss": 0.8545, + "step": 53175 + }, + { + "epoch": 3.42, + "learning_rate": 8.61074733248889e-06, + "loss": 0.8264, + "step": 53176 + }, + { + "epoch": 3.42, + "learning_rate": 8.61053264346594e-06, + "loss": 0.779, + "step": 53177 + }, + { + "epoch": 3.42, + "learning_rate": 8.61031795444299e-06, + "loss": 0.8376, + "step": 53178 + }, + { + "epoch": 3.42, + "learning_rate": 8.61010326542004e-06, + "loss": 0.8241, + "step": 53179 + }, + { + "epoch": 3.42, + "learning_rate": 8.60988857639709e-06, + "loss": 0.8229, + "step": 53180 + }, + { + "epoch": 3.42, + "learning_rate": 8.60967388737414e-06, + "loss": 0.8245, + "step": 53181 + }, + { + "epoch": 3.42, + "learning_rate": 8.609459198351188e-06, + "loss": 0.7709, + "step": 53182 + }, + { + "epoch": 3.42, + "learning_rate": 8.609244509328239e-06, + "loss": 0.7926, + "step": 53183 + }, + { + "epoch": 3.42, + "learning_rate": 8.609029820305288e-06, + "loss": 0.8258, + "step": 53184 + }, + { + "epoch": 3.42, + "learning_rate": 8.608815131282339e-06, + "loss": 0.8137, + "step": 53185 + }, + { + "epoch": 3.42, + "learning_rate": 8.608600442259387e-06, + "loss": 0.8186, + "step": 53186 + }, + { + "epoch": 3.42, + "learning_rate": 8.608385753236438e-06, + "loss": 0.8114, + "step": 53187 + }, + { + "epoch": 3.42, + "learning_rate": 8.608171064213487e-06, + "loss": 0.8447, + "step": 53188 + }, + { + "epoch": 3.42, + "learning_rate": 8.607956375190538e-06, + "loss": 0.8187, + "step": 53189 + }, + { + "epoch": 3.42, + "learning_rate": 8.607741686167587e-06, + "loss": 0.8184, + "step": 53190 + }, + { + "epoch": 3.42, + "learning_rate": 8.607526997144636e-06, + "loss": 0.8517, + "step": 53191 + }, + { + "epoch": 3.42, + "learning_rate": 8.607312308121686e-06, + "loss": 0.8079, + "step": 53192 + }, + { + "epoch": 3.42, + "learning_rate": 8.607097619098737e-06, + "loss": 0.8775, + "step": 53193 + }, + { + "epoch": 3.42, + "learning_rate": 8.606882930075786e-06, + "loss": 0.8045, + "step": 53194 + }, + { + "epoch": 3.42, + "learning_rate": 8.606668241052835e-06, + "loss": 0.8022, + "step": 53195 + }, + { + "epoch": 3.42, + "learning_rate": 8.606453552029886e-06, + "loss": 0.8078, + "step": 53196 + }, + { + "epoch": 3.42, + "learning_rate": 8.606238863006935e-06, + "loss": 0.8362, + "step": 53197 + }, + { + "epoch": 3.42, + "learning_rate": 8.606024173983985e-06, + "loss": 0.8158, + "step": 53198 + }, + { + "epoch": 3.42, + "learning_rate": 8.605809484961036e-06, + "loss": 0.7988, + "step": 53199 + }, + { + "epoch": 3.42, + "learning_rate": 8.605594795938085e-06, + "loss": 0.8506, + "step": 53200 + }, + { + "epoch": 3.42, + "learning_rate": 8.605380106915134e-06, + "loss": 0.8427, + "step": 53201 + }, + { + "epoch": 3.42, + "learning_rate": 8.605165417892184e-06, + "loss": 0.8071, + "step": 53202 + }, + { + "epoch": 3.42, + "learning_rate": 8.604950728869233e-06, + "loss": 0.7919, + "step": 53203 + }, + { + "epoch": 3.42, + "learning_rate": 8.604736039846282e-06, + "loss": 0.8875, + "step": 53204 + }, + { + "epoch": 3.42, + "learning_rate": 8.604521350823333e-06, + "loss": 0.9068, + "step": 53205 + }, + { + "epoch": 3.42, + "learning_rate": 8.604306661800382e-06, + "loss": 0.9043, + "step": 53206 + }, + { + "epoch": 3.42, + "learning_rate": 8.604091972777433e-06, + "loss": 0.8519, + "step": 53207 + }, + { + "epoch": 3.42, + "learning_rate": 8.603877283754483e-06, + "loss": 0.8298, + "step": 53208 + }, + { + "epoch": 3.42, + "learning_rate": 8.603662594731532e-06, + "loss": 0.8122, + "step": 53209 + }, + { + "epoch": 3.42, + "learning_rate": 8.603447905708581e-06, + "loss": 0.8914, + "step": 53210 + }, + { + "epoch": 3.42, + "learning_rate": 8.603233216685632e-06, + "loss": 0.8361, + "step": 53211 + }, + { + "epoch": 3.42, + "learning_rate": 8.603018527662681e-06, + "loss": 0.8705, + "step": 53212 + }, + { + "epoch": 3.42, + "learning_rate": 8.60280383863973e-06, + "loss": 0.7951, + "step": 53213 + }, + { + "epoch": 3.42, + "learning_rate": 8.60258914961678e-06, + "loss": 0.9102, + "step": 53214 + }, + { + "epoch": 3.42, + "learning_rate": 8.602374460593831e-06, + "loss": 0.7693, + "step": 53215 + }, + { + "epoch": 3.42, + "learning_rate": 8.60215977157088e-06, + "loss": 0.8297, + "step": 53216 + }, + { + "epoch": 3.42, + "learning_rate": 8.60194508254793e-06, + "loss": 0.8095, + "step": 53217 + }, + { + "epoch": 3.42, + "learning_rate": 8.60173039352498e-06, + "loss": 0.825, + "step": 53218 + }, + { + "epoch": 3.42, + "learning_rate": 8.601515704502029e-06, + "loss": 0.79, + "step": 53219 + }, + { + "epoch": 3.42, + "learning_rate": 8.60130101547908e-06, + "loss": 0.84, + "step": 53220 + }, + { + "epoch": 3.42, + "learning_rate": 8.60108632645613e-06, + "loss": 0.883, + "step": 53221 + }, + { + "epoch": 3.42, + "learning_rate": 8.600871637433179e-06, + "loss": 0.8292, + "step": 53222 + }, + { + "epoch": 3.42, + "learning_rate": 8.600656948410228e-06, + "loss": 0.8165, + "step": 53223 + }, + { + "epoch": 3.42, + "learning_rate": 8.600442259387279e-06, + "loss": 0.8872, + "step": 53224 + }, + { + "epoch": 3.42, + "learning_rate": 8.600227570364328e-06, + "loss": 0.7561, + "step": 53225 + }, + { + "epoch": 3.42, + "learning_rate": 8.600012881341378e-06, + "loss": 0.7725, + "step": 53226 + }, + { + "epoch": 3.42, + "learning_rate": 8.599798192318427e-06, + "loss": 0.8457, + "step": 53227 + }, + { + "epoch": 3.42, + "learning_rate": 8.599583503295478e-06, + "loss": 0.8083, + "step": 53228 + }, + { + "epoch": 3.42, + "learning_rate": 8.599368814272527e-06, + "loss": 0.8099, + "step": 53229 + }, + { + "epoch": 3.42, + "learning_rate": 8.599154125249578e-06, + "loss": 0.7893, + "step": 53230 + }, + { + "epoch": 3.42, + "learning_rate": 8.598939436226626e-06, + "loss": 0.8622, + "step": 53231 + }, + { + "epoch": 3.42, + "learning_rate": 8.598724747203675e-06, + "loss": 0.865, + "step": 53232 + }, + { + "epoch": 3.42, + "learning_rate": 8.598510058180726e-06, + "loss": 0.857, + "step": 53233 + }, + { + "epoch": 3.42, + "learning_rate": 8.598295369157775e-06, + "loss": 0.8585, + "step": 53234 + }, + { + "epoch": 3.43, + "learning_rate": 8.598080680134824e-06, + "loss": 0.8554, + "step": 53235 + }, + { + "epoch": 3.43, + "learning_rate": 8.597865991111876e-06, + "loss": 0.8609, + "step": 53236 + }, + { + "epoch": 3.43, + "learning_rate": 8.597651302088925e-06, + "loss": 0.8063, + "step": 53237 + }, + { + "epoch": 3.43, + "learning_rate": 8.597436613065974e-06, + "loss": 0.7711, + "step": 53238 + }, + { + "epoch": 3.43, + "learning_rate": 8.597221924043025e-06, + "loss": 0.7797, + "step": 53239 + }, + { + "epoch": 3.43, + "learning_rate": 8.597007235020074e-06, + "loss": 0.8427, + "step": 53240 + }, + { + "epoch": 3.43, + "learning_rate": 8.596792545997123e-06, + "loss": 0.8067, + "step": 53241 + }, + { + "epoch": 3.43, + "learning_rate": 8.596577856974174e-06, + "loss": 0.7657, + "step": 53242 + }, + { + "epoch": 3.43, + "learning_rate": 8.596363167951224e-06, + "loss": 0.8125, + "step": 53243 + }, + { + "epoch": 3.43, + "learning_rate": 8.596148478928273e-06, + "loss": 0.8239, + "step": 53244 + }, + { + "epoch": 3.43, + "learning_rate": 8.595933789905324e-06, + "loss": 0.8379, + "step": 53245 + }, + { + "epoch": 3.43, + "learning_rate": 8.595719100882373e-06, + "loss": 0.7202, + "step": 53246 + }, + { + "epoch": 3.43, + "learning_rate": 8.595504411859422e-06, + "loss": 0.8417, + "step": 53247 + }, + { + "epoch": 3.43, + "learning_rate": 8.595289722836472e-06, + "loss": 0.8643, + "step": 53248 + }, + { + "epoch": 3.43, + "learning_rate": 8.595075033813521e-06, + "loss": 0.817, + "step": 53249 + }, + { + "epoch": 3.43, + "learning_rate": 8.594860344790572e-06, + "loss": 0.8751, + "step": 53250 + }, + { + "epoch": 3.43, + "learning_rate": 8.594645655767621e-06, + "loss": 0.7832, + "step": 53251 + }, + { + "epoch": 3.43, + "learning_rate": 8.594430966744672e-06, + "loss": 0.8651, + "step": 53252 + }, + { + "epoch": 3.43, + "learning_rate": 8.59421627772172e-06, + "loss": 0.7803, + "step": 53253 + }, + { + "epoch": 3.43, + "learning_rate": 8.59400158869877e-06, + "loss": 0.8732, + "step": 53254 + }, + { + "epoch": 3.43, + "learning_rate": 8.59378689967582e-06, + "loss": 0.8366, + "step": 53255 + }, + { + "epoch": 3.43, + "learning_rate": 8.59357221065287e-06, + "loss": 0.8145, + "step": 53256 + }, + { + "epoch": 3.43, + "learning_rate": 8.59335752162992e-06, + "loss": 0.8465, + "step": 53257 + }, + { + "epoch": 3.43, + "learning_rate": 8.59314283260697e-06, + "loss": 0.7151, + "step": 53258 + }, + { + "epoch": 3.43, + "learning_rate": 8.59292814358402e-06, + "loss": 0.8679, + "step": 53259 + }, + { + "epoch": 3.43, + "learning_rate": 8.592713454561068e-06, + "loss": 0.7598, + "step": 53260 + }, + { + "epoch": 3.43, + "learning_rate": 8.592498765538119e-06, + "loss": 0.8652, + "step": 53261 + }, + { + "epoch": 3.43, + "learning_rate": 8.592284076515168e-06, + "loss": 0.8126, + "step": 53262 + }, + { + "epoch": 3.43, + "learning_rate": 8.592069387492217e-06, + "loss": 0.8286, + "step": 53263 + }, + { + "epoch": 3.43, + "learning_rate": 8.591854698469268e-06, + "loss": 0.8255, + "step": 53264 + }, + { + "epoch": 3.43, + "learning_rate": 8.591640009446318e-06, + "loss": 0.7986, + "step": 53265 + }, + { + "epoch": 3.43, + "learning_rate": 8.591425320423367e-06, + "loss": 0.8278, + "step": 53266 + }, + { + "epoch": 3.43, + "learning_rate": 8.591210631400418e-06, + "loss": 0.7766, + "step": 53267 + }, + { + "epoch": 3.43, + "learning_rate": 8.590995942377467e-06, + "loss": 0.8438, + "step": 53268 + }, + { + "epoch": 3.43, + "learning_rate": 8.590781253354516e-06, + "loss": 0.8638, + "step": 53269 + }, + { + "epoch": 3.43, + "learning_rate": 8.590566564331567e-06, + "loss": 0.9032, + "step": 53270 + }, + { + "epoch": 3.43, + "learning_rate": 8.590351875308616e-06, + "loss": 0.8528, + "step": 53271 + }, + { + "epoch": 3.43, + "learning_rate": 8.590137186285666e-06, + "loss": 0.8475, + "step": 53272 + }, + { + "epoch": 3.43, + "learning_rate": 8.589922497262715e-06, + "loss": 0.8335, + "step": 53273 + }, + { + "epoch": 3.43, + "learning_rate": 8.589707808239766e-06, + "loss": 0.8588, + "step": 53274 + }, + { + "epoch": 3.43, + "learning_rate": 8.589493119216815e-06, + "loss": 0.8451, + "step": 53275 + }, + { + "epoch": 3.43, + "learning_rate": 8.589278430193865e-06, + "loss": 0.8044, + "step": 53276 + }, + { + "epoch": 3.43, + "learning_rate": 8.589063741170914e-06, + "loss": 0.8451, + "step": 53277 + }, + { + "epoch": 3.43, + "learning_rate": 8.588849052147963e-06, + "loss": 0.8654, + "step": 53278 + }, + { + "epoch": 3.43, + "learning_rate": 8.588634363125014e-06, + "loss": 0.8321, + "step": 53279 + }, + { + "epoch": 3.43, + "learning_rate": 8.588419674102065e-06, + "loss": 0.807, + "step": 53280 + }, + { + "epoch": 3.43, + "learning_rate": 8.588204985079114e-06, + "loss": 0.8043, + "step": 53281 + }, + { + "epoch": 3.43, + "learning_rate": 8.587990296056163e-06, + "loss": 0.8948, + "step": 53282 + }, + { + "epoch": 3.43, + "learning_rate": 8.587775607033213e-06, + "loss": 0.8323, + "step": 53283 + }, + { + "epoch": 3.43, + "learning_rate": 8.587560918010262e-06, + "loss": 0.8267, + "step": 53284 + }, + { + "epoch": 3.43, + "learning_rate": 8.587346228987313e-06, + "loss": 0.8222, + "step": 53285 + }, + { + "epoch": 3.43, + "learning_rate": 8.587131539964362e-06, + "loss": 0.8378, + "step": 53286 + }, + { + "epoch": 3.43, + "learning_rate": 8.586916850941413e-06, + "loss": 0.836, + "step": 53287 + }, + { + "epoch": 3.43, + "learning_rate": 8.586702161918462e-06, + "loss": 0.7966, + "step": 53288 + }, + { + "epoch": 3.43, + "learning_rate": 8.586487472895512e-06, + "loss": 0.794, + "step": 53289 + }, + { + "epoch": 3.43, + "learning_rate": 8.586272783872561e-06, + "loss": 0.7891, + "step": 53290 + }, + { + "epoch": 3.43, + "learning_rate": 8.58605809484961e-06, + "loss": 0.7904, + "step": 53291 + }, + { + "epoch": 3.43, + "learning_rate": 8.58584340582666e-06, + "loss": 0.8668, + "step": 53292 + }, + { + "epoch": 3.43, + "learning_rate": 8.58562871680371e-06, + "loss": 0.8628, + "step": 53293 + }, + { + "epoch": 3.43, + "learning_rate": 8.58541402778076e-06, + "loss": 0.8302, + "step": 53294 + }, + { + "epoch": 3.43, + "learning_rate": 8.585199338757811e-06, + "loss": 0.8234, + "step": 53295 + }, + { + "epoch": 3.43, + "learning_rate": 8.58498464973486e-06, + "loss": 0.8422, + "step": 53296 + }, + { + "epoch": 3.43, + "learning_rate": 8.584769960711909e-06, + "loss": 0.8371, + "step": 53297 + }, + { + "epoch": 3.43, + "learning_rate": 8.58455527168896e-06, + "loss": 0.9136, + "step": 53298 + }, + { + "epoch": 3.43, + "learning_rate": 8.584340582666009e-06, + "loss": 0.7772, + "step": 53299 + }, + { + "epoch": 3.43, + "learning_rate": 8.584125893643058e-06, + "loss": 0.8399, + "step": 53300 + }, + { + "epoch": 3.43, + "learning_rate": 8.583911204620108e-06, + "loss": 0.8288, + "step": 53301 + }, + { + "epoch": 3.43, + "learning_rate": 8.583696515597159e-06, + "loss": 0.7899, + "step": 53302 + }, + { + "epoch": 3.43, + "learning_rate": 8.583481826574208e-06, + "loss": 0.8034, + "step": 53303 + }, + { + "epoch": 3.43, + "learning_rate": 8.583267137551259e-06, + "loss": 0.8411, + "step": 53304 + }, + { + "epoch": 3.43, + "learning_rate": 8.583052448528307e-06, + "loss": 0.7795, + "step": 53305 + }, + { + "epoch": 3.43, + "learning_rate": 8.582837759505356e-06, + "loss": 0.8269, + "step": 53306 + }, + { + "epoch": 3.43, + "learning_rate": 8.582623070482407e-06, + "loss": 0.7795, + "step": 53307 + }, + { + "epoch": 3.43, + "learning_rate": 8.582408381459458e-06, + "loss": 0.8685, + "step": 53308 + }, + { + "epoch": 3.43, + "learning_rate": 8.582193692436507e-06, + "loss": 0.808, + "step": 53309 + }, + { + "epoch": 3.43, + "learning_rate": 8.581979003413556e-06, + "loss": 0.7909, + "step": 53310 + }, + { + "epoch": 3.43, + "learning_rate": 8.581764314390606e-06, + "loss": 0.8454, + "step": 53311 + }, + { + "epoch": 3.43, + "learning_rate": 8.581549625367655e-06, + "loss": 0.8277, + "step": 53312 + }, + { + "epoch": 3.43, + "learning_rate": 8.581334936344706e-06, + "loss": 0.7885, + "step": 53313 + }, + { + "epoch": 3.43, + "learning_rate": 8.581120247321755e-06, + "loss": 0.8367, + "step": 53314 + }, + { + "epoch": 3.43, + "learning_rate": 8.580905558298804e-06, + "loss": 0.8283, + "step": 53315 + }, + { + "epoch": 3.43, + "learning_rate": 8.580690869275855e-06, + "loss": 0.7872, + "step": 53316 + }, + { + "epoch": 3.43, + "learning_rate": 8.580476180252905e-06, + "loss": 0.8612, + "step": 53317 + }, + { + "epoch": 3.43, + "learning_rate": 8.580261491229954e-06, + "loss": 0.9227, + "step": 53318 + }, + { + "epoch": 3.43, + "learning_rate": 8.580046802207003e-06, + "loss": 0.7913, + "step": 53319 + }, + { + "epoch": 3.43, + "learning_rate": 8.579832113184054e-06, + "loss": 0.8174, + "step": 53320 + }, + { + "epoch": 3.43, + "learning_rate": 8.579617424161103e-06, + "loss": 0.8236, + "step": 53321 + }, + { + "epoch": 3.43, + "learning_rate": 8.579402735138152e-06, + "loss": 0.7779, + "step": 53322 + }, + { + "epoch": 3.43, + "learning_rate": 8.579188046115204e-06, + "loss": 0.8208, + "step": 53323 + }, + { + "epoch": 3.43, + "learning_rate": 8.578973357092253e-06, + "loss": 0.7682, + "step": 53324 + }, + { + "epoch": 3.43, + "learning_rate": 8.578758668069302e-06, + "loss": 0.7512, + "step": 53325 + }, + { + "epoch": 3.43, + "learning_rate": 8.578543979046353e-06, + "loss": 0.852, + "step": 53326 + }, + { + "epoch": 3.43, + "learning_rate": 8.578329290023402e-06, + "loss": 0.8066, + "step": 53327 + }, + { + "epoch": 3.43, + "learning_rate": 8.57811460100045e-06, + "loss": 0.8446, + "step": 53328 + }, + { + "epoch": 3.43, + "learning_rate": 8.577899911977501e-06, + "loss": 0.7848, + "step": 53329 + }, + { + "epoch": 3.43, + "learning_rate": 8.577685222954552e-06, + "loss": 0.8768, + "step": 53330 + }, + { + "epoch": 3.43, + "learning_rate": 8.577470533931601e-06, + "loss": 0.7707, + "step": 53331 + }, + { + "epoch": 3.43, + "learning_rate": 8.57725584490865e-06, + "loss": 0.8333, + "step": 53332 + }, + { + "epoch": 3.43, + "learning_rate": 8.5770411558857e-06, + "loss": 0.8268, + "step": 53333 + }, + { + "epoch": 3.43, + "learning_rate": 8.57682646686275e-06, + "loss": 0.8269, + "step": 53334 + }, + { + "epoch": 3.43, + "learning_rate": 8.5766117778398e-06, + "loss": 0.8333, + "step": 53335 + }, + { + "epoch": 3.43, + "learning_rate": 8.576397088816849e-06, + "loss": 0.8197, + "step": 53336 + }, + { + "epoch": 3.43, + "learning_rate": 8.5761823997939e-06, + "loss": 0.8166, + "step": 53337 + }, + { + "epoch": 3.43, + "learning_rate": 8.575967710770949e-06, + "loss": 0.8324, + "step": 53338 + }, + { + "epoch": 3.43, + "learning_rate": 8.575753021748e-06, + "loss": 0.779, + "step": 53339 + }, + { + "epoch": 3.43, + "learning_rate": 8.575538332725048e-06, + "loss": 0.7538, + "step": 53340 + }, + { + "epoch": 3.43, + "learning_rate": 8.575323643702097e-06, + "loss": 0.911, + "step": 53341 + }, + { + "epoch": 3.43, + "learning_rate": 8.575108954679148e-06, + "loss": 0.8742, + "step": 53342 + }, + { + "epoch": 3.43, + "learning_rate": 8.574894265656197e-06, + "loss": 0.7994, + "step": 53343 + }, + { + "epoch": 3.43, + "learning_rate": 8.574679576633248e-06, + "loss": 0.8551, + "step": 53344 + }, + { + "epoch": 3.43, + "learning_rate": 8.574464887610298e-06, + "loss": 0.8487, + "step": 53345 + }, + { + "epoch": 3.43, + "learning_rate": 8.574250198587347e-06, + "loss": 0.8153, + "step": 53346 + }, + { + "epoch": 3.43, + "learning_rate": 8.574035509564396e-06, + "loss": 0.8317, + "step": 53347 + }, + { + "epoch": 3.43, + "learning_rate": 8.573820820541447e-06, + "loss": 0.7878, + "step": 53348 + }, + { + "epoch": 3.43, + "learning_rate": 8.573606131518496e-06, + "loss": 0.8532, + "step": 53349 + }, + { + "epoch": 3.43, + "learning_rate": 8.573391442495545e-06, + "loss": 0.7826, + "step": 53350 + }, + { + "epoch": 3.43, + "learning_rate": 8.573176753472595e-06, + "loss": 0.9004, + "step": 53351 + }, + { + "epoch": 3.43, + "learning_rate": 8.572962064449646e-06, + "loss": 0.8426, + "step": 53352 + }, + { + "epoch": 3.43, + "learning_rate": 8.572747375426695e-06, + "loss": 0.8295, + "step": 53353 + }, + { + "epoch": 3.43, + "learning_rate": 8.572532686403746e-06, + "loss": 0.8373, + "step": 53354 + }, + { + "epoch": 3.43, + "learning_rate": 8.572317997380795e-06, + "loss": 0.8117, + "step": 53355 + }, + { + "epoch": 3.43, + "learning_rate": 8.572103308357844e-06, + "loss": 0.7398, + "step": 53356 + }, + { + "epoch": 3.43, + "learning_rate": 8.571888619334894e-06, + "loss": 0.8062, + "step": 53357 + }, + { + "epoch": 3.43, + "learning_rate": 8.571673930311943e-06, + "loss": 0.835, + "step": 53358 + }, + { + "epoch": 3.43, + "learning_rate": 8.571459241288994e-06, + "loss": 0.8026, + "step": 53359 + }, + { + "epoch": 3.43, + "learning_rate": 8.571244552266043e-06, + "loss": 0.8376, + "step": 53360 + }, + { + "epoch": 3.43, + "learning_rate": 8.571029863243094e-06, + "loss": 0.7868, + "step": 53361 + }, + { + "epoch": 3.43, + "learning_rate": 8.570815174220143e-06, + "loss": 0.8737, + "step": 53362 + }, + { + "epoch": 3.43, + "learning_rate": 8.570600485197193e-06, + "loss": 0.8032, + "step": 53363 + }, + { + "epoch": 3.43, + "learning_rate": 8.570385796174242e-06, + "loss": 0.829, + "step": 53364 + }, + { + "epoch": 3.43, + "learning_rate": 8.570171107151291e-06, + "loss": 0.8679, + "step": 53365 + }, + { + "epoch": 3.43, + "learning_rate": 8.569956418128342e-06, + "loss": 0.8021, + "step": 53366 + }, + { + "epoch": 3.43, + "learning_rate": 8.569741729105392e-06, + "loss": 0.82, + "step": 53367 + }, + { + "epoch": 3.43, + "learning_rate": 8.569527040082441e-06, + "loss": 0.8764, + "step": 53368 + }, + { + "epoch": 3.43, + "learning_rate": 8.56931235105949e-06, + "loss": 0.7769, + "step": 53369 + }, + { + "epoch": 3.43, + "learning_rate": 8.569097662036541e-06, + "loss": 0.7897, + "step": 53370 + }, + { + "epoch": 3.43, + "learning_rate": 8.56888297301359e-06, + "loss": 0.7665, + "step": 53371 + }, + { + "epoch": 3.43, + "learning_rate": 8.56866828399064e-06, + "loss": 0.8112, + "step": 53372 + }, + { + "epoch": 3.43, + "learning_rate": 8.56845359496769e-06, + "loss": 0.8264, + "step": 53373 + }, + { + "epoch": 3.43, + "learning_rate": 8.56823890594474e-06, + "loss": 0.8586, + "step": 53374 + }, + { + "epoch": 3.43, + "learning_rate": 8.56802421692179e-06, + "loss": 0.8463, + "step": 53375 + }, + { + "epoch": 3.43, + "learning_rate": 8.56780952789884e-06, + "loss": 0.7848, + "step": 53376 + }, + { + "epoch": 3.43, + "learning_rate": 8.567594838875889e-06, + "loss": 0.8255, + "step": 53377 + }, + { + "epoch": 3.43, + "learning_rate": 8.567380149852938e-06, + "loss": 0.8359, + "step": 53378 + }, + { + "epoch": 3.43, + "learning_rate": 8.567165460829988e-06, + "loss": 0.8188, + "step": 53379 + }, + { + "epoch": 3.43, + "learning_rate": 8.566950771807037e-06, + "loss": 0.8655, + "step": 53380 + }, + { + "epoch": 3.43, + "learning_rate": 8.566736082784088e-06, + "loss": 0.8486, + "step": 53381 + }, + { + "epoch": 3.43, + "learning_rate": 8.566521393761139e-06, + "loss": 0.8212, + "step": 53382 + }, + { + "epoch": 3.43, + "learning_rate": 8.566306704738188e-06, + "loss": 0.816, + "step": 53383 + }, + { + "epoch": 3.43, + "learning_rate": 8.566092015715237e-06, + "loss": 0.8367, + "step": 53384 + }, + { + "epoch": 3.43, + "learning_rate": 8.565877326692287e-06, + "loss": 0.8072, + "step": 53385 + }, + { + "epoch": 3.43, + "learning_rate": 8.565662637669336e-06, + "loss": 0.8035, + "step": 53386 + }, + { + "epoch": 3.43, + "learning_rate": 8.565447948646385e-06, + "loss": 0.7496, + "step": 53387 + }, + { + "epoch": 3.43, + "learning_rate": 8.565233259623436e-06, + "loss": 0.8958, + "step": 53388 + }, + { + "epoch": 3.43, + "learning_rate": 8.565018570600487e-06, + "loss": 0.8185, + "step": 53389 + }, + { + "epoch": 3.43, + "learning_rate": 8.564803881577536e-06, + "loss": 0.7877, + "step": 53390 + }, + { + "epoch": 3.44, + "learning_rate": 8.564589192554586e-06, + "loss": 0.8511, + "step": 53391 + }, + { + "epoch": 3.44, + "learning_rate": 8.564374503531635e-06, + "loss": 0.8093, + "step": 53392 + }, + { + "epoch": 3.44, + "learning_rate": 8.564159814508684e-06, + "loss": 0.871, + "step": 53393 + }, + { + "epoch": 3.44, + "learning_rate": 8.563945125485735e-06, + "loss": 0.8617, + "step": 53394 + }, + { + "epoch": 3.44, + "learning_rate": 8.563730436462785e-06, + "loss": 0.7721, + "step": 53395 + }, + { + "epoch": 3.44, + "learning_rate": 8.563515747439834e-06, + "loss": 0.7628, + "step": 53396 + }, + { + "epoch": 3.44, + "learning_rate": 8.563301058416883e-06, + "loss": 0.7905, + "step": 53397 + }, + { + "epoch": 3.44, + "learning_rate": 8.563086369393934e-06, + "loss": 0.8191, + "step": 53398 + }, + { + "epoch": 3.44, + "learning_rate": 8.562871680370983e-06, + "loss": 0.9019, + "step": 53399 + }, + { + "epoch": 3.44, + "learning_rate": 8.562656991348034e-06, + "loss": 0.8588, + "step": 53400 + }, + { + "epoch": 3.44, + "learning_rate": 8.562442302325083e-06, + "loss": 0.8766, + "step": 53401 + }, + { + "epoch": 3.44, + "learning_rate": 8.562227613302132e-06, + "loss": 0.8125, + "step": 53402 + }, + { + "epoch": 3.44, + "learning_rate": 8.562012924279182e-06, + "loss": 0.842, + "step": 53403 + }, + { + "epoch": 3.44, + "learning_rate": 8.561798235256233e-06, + "loss": 0.8647, + "step": 53404 + }, + { + "epoch": 3.44, + "learning_rate": 8.561583546233282e-06, + "loss": 0.8402, + "step": 53405 + }, + { + "epoch": 3.44, + "learning_rate": 8.561368857210331e-06, + "loss": 0.869, + "step": 53406 + }, + { + "epoch": 3.44, + "learning_rate": 8.561154168187382e-06, + "loss": 0.873, + "step": 53407 + }, + { + "epoch": 3.44, + "learning_rate": 8.56093947916443e-06, + "loss": 0.7955, + "step": 53408 + }, + { + "epoch": 3.44, + "learning_rate": 8.56072479014148e-06, + "loss": 0.7835, + "step": 53409 + }, + { + "epoch": 3.44, + "learning_rate": 8.560510101118532e-06, + "loss": 0.7605, + "step": 53410 + }, + { + "epoch": 3.44, + "learning_rate": 8.56029541209558e-06, + "loss": 0.8572, + "step": 53411 + }, + { + "epoch": 3.44, + "learning_rate": 8.56008072307263e-06, + "loss": 0.8665, + "step": 53412 + }, + { + "epoch": 3.44, + "learning_rate": 8.55986603404968e-06, + "loss": 0.7944, + "step": 53413 + }, + { + "epoch": 3.44, + "learning_rate": 8.55965134502673e-06, + "loss": 0.7949, + "step": 53414 + }, + { + "epoch": 3.44, + "learning_rate": 8.559436656003778e-06, + "loss": 0.8815, + "step": 53415 + }, + { + "epoch": 3.44, + "learning_rate": 8.559221966980829e-06, + "loss": 0.8024, + "step": 53416 + }, + { + "epoch": 3.44, + "learning_rate": 8.55900727795788e-06, + "loss": 0.8634, + "step": 53417 + }, + { + "epoch": 3.44, + "learning_rate": 8.558792588934929e-06, + "loss": 0.817, + "step": 53418 + }, + { + "epoch": 3.44, + "learning_rate": 8.558577899911978e-06, + "loss": 0.7982, + "step": 53419 + }, + { + "epoch": 3.44, + "learning_rate": 8.558363210889028e-06, + "loss": 0.856, + "step": 53420 + }, + { + "epoch": 3.44, + "learning_rate": 8.558148521866077e-06, + "loss": 0.8251, + "step": 53421 + }, + { + "epoch": 3.44, + "learning_rate": 8.557933832843128e-06, + "loss": 0.8795, + "step": 53422 + }, + { + "epoch": 3.44, + "learning_rate": 8.557719143820177e-06, + "loss": 0.8373, + "step": 53423 + }, + { + "epoch": 3.44, + "learning_rate": 8.557504454797227e-06, + "loss": 0.849, + "step": 53424 + }, + { + "epoch": 3.44, + "learning_rate": 8.557289765774276e-06, + "loss": 0.8051, + "step": 53425 + }, + { + "epoch": 3.44, + "learning_rate": 8.557075076751327e-06, + "loss": 0.8652, + "step": 53426 + }, + { + "epoch": 3.44, + "learning_rate": 8.556860387728376e-06, + "loss": 0.8051, + "step": 53427 + }, + { + "epoch": 3.44, + "learning_rate": 8.556645698705425e-06, + "loss": 0.7497, + "step": 53428 + }, + { + "epoch": 3.44, + "learning_rate": 8.556431009682476e-06, + "loss": 0.8288, + "step": 53429 + }, + { + "epoch": 3.44, + "learning_rate": 8.556216320659525e-06, + "loss": 0.8411, + "step": 53430 + }, + { + "epoch": 3.44, + "learning_rate": 8.556001631636575e-06, + "loss": 0.8525, + "step": 53431 + }, + { + "epoch": 3.44, + "learning_rate": 8.555786942613626e-06, + "loss": 0.7903, + "step": 53432 + }, + { + "epoch": 3.44, + "learning_rate": 8.555572253590675e-06, + "loss": 0.7995, + "step": 53433 + }, + { + "epoch": 3.44, + "learning_rate": 8.555357564567724e-06, + "loss": 0.7917, + "step": 53434 + }, + { + "epoch": 3.44, + "learning_rate": 8.555142875544775e-06, + "loss": 0.8658, + "step": 53435 + }, + { + "epoch": 3.44, + "learning_rate": 8.554928186521824e-06, + "loss": 0.8521, + "step": 53436 + }, + { + "epoch": 3.44, + "learning_rate": 8.554713497498873e-06, + "loss": 0.7549, + "step": 53437 + }, + { + "epoch": 3.44, + "learning_rate": 8.554498808475923e-06, + "loss": 0.8144, + "step": 53438 + }, + { + "epoch": 3.44, + "learning_rate": 8.554284119452974e-06, + "loss": 0.8333, + "step": 53439 + }, + { + "epoch": 3.44, + "learning_rate": 8.554069430430023e-06, + "loss": 0.8691, + "step": 53440 + }, + { + "epoch": 3.44, + "learning_rate": 8.553854741407073e-06, + "loss": 0.8166, + "step": 53441 + }, + { + "epoch": 3.44, + "learning_rate": 8.553640052384122e-06, + "loss": 0.8545, + "step": 53442 + }, + { + "epoch": 3.44, + "learning_rate": 8.553425363361171e-06, + "loss": 0.7836, + "step": 53443 + }, + { + "epoch": 3.44, + "learning_rate": 8.553210674338222e-06, + "loss": 0.8369, + "step": 53444 + }, + { + "epoch": 3.44, + "learning_rate": 8.552995985315271e-06, + "loss": 0.8078, + "step": 53445 + }, + { + "epoch": 3.44, + "learning_rate": 8.552781296292322e-06, + "loss": 0.8397, + "step": 53446 + }, + { + "epoch": 3.44, + "learning_rate": 8.55256660726937e-06, + "loss": 0.8698, + "step": 53447 + }, + { + "epoch": 3.44, + "learning_rate": 8.552351918246421e-06, + "loss": 0.8565, + "step": 53448 + }, + { + "epoch": 3.44, + "learning_rate": 8.55213722922347e-06, + "loss": 0.8577, + "step": 53449 + }, + { + "epoch": 3.44, + "learning_rate": 8.551922540200521e-06, + "loss": 0.7521, + "step": 53450 + }, + { + "epoch": 3.44, + "learning_rate": 8.55170785117757e-06, + "loss": 0.837, + "step": 53451 + }, + { + "epoch": 3.44, + "learning_rate": 8.551493162154619e-06, + "loss": 0.8219, + "step": 53452 + }, + { + "epoch": 3.44, + "learning_rate": 8.55127847313167e-06, + "loss": 0.8133, + "step": 53453 + }, + { + "epoch": 3.44, + "learning_rate": 8.55106378410872e-06, + "loss": 0.7972, + "step": 53454 + }, + { + "epoch": 3.44, + "learning_rate": 8.550849095085769e-06, + "loss": 0.7888, + "step": 53455 + }, + { + "epoch": 3.44, + "learning_rate": 8.550634406062818e-06, + "loss": 0.7768, + "step": 53456 + }, + { + "epoch": 3.44, + "learning_rate": 8.550419717039869e-06, + "loss": 0.8038, + "step": 53457 + }, + { + "epoch": 3.44, + "learning_rate": 8.550205028016918e-06, + "loss": 0.8834, + "step": 53458 + }, + { + "epoch": 3.44, + "learning_rate": 8.549990338993968e-06, + "loss": 0.7975, + "step": 53459 + }, + { + "epoch": 3.44, + "learning_rate": 8.549775649971017e-06, + "loss": 0.7985, + "step": 53460 + }, + { + "epoch": 3.44, + "learning_rate": 8.549560960948068e-06, + "loss": 0.7811, + "step": 53461 + }, + { + "epoch": 3.44, + "learning_rate": 8.549346271925117e-06, + "loss": 0.8881, + "step": 53462 + }, + { + "epoch": 3.44, + "learning_rate": 8.549131582902168e-06, + "loss": 0.8809, + "step": 53463 + }, + { + "epoch": 3.44, + "learning_rate": 8.548916893879217e-06, + "loss": 0.8349, + "step": 53464 + }, + { + "epoch": 3.44, + "learning_rate": 8.548702204856266e-06, + "loss": 0.8229, + "step": 53465 + }, + { + "epoch": 3.44, + "learning_rate": 8.548487515833316e-06, + "loss": 0.8782, + "step": 53466 + }, + { + "epoch": 3.44, + "learning_rate": 8.548272826810365e-06, + "loss": 0.8804, + "step": 53467 + }, + { + "epoch": 3.44, + "learning_rate": 8.548058137787416e-06, + "loss": 0.8398, + "step": 53468 + }, + { + "epoch": 3.44, + "learning_rate": 8.547843448764466e-06, + "loss": 0.8633, + "step": 53469 + }, + { + "epoch": 3.44, + "learning_rate": 8.547628759741515e-06, + "loss": 0.8115, + "step": 53470 + }, + { + "epoch": 3.44, + "learning_rate": 8.547414070718564e-06, + "loss": 0.8225, + "step": 53471 + }, + { + "epoch": 3.44, + "learning_rate": 8.547199381695615e-06, + "loss": 0.8066, + "step": 53472 + }, + { + "epoch": 3.44, + "learning_rate": 8.546984692672664e-06, + "loss": 0.7889, + "step": 53473 + }, + { + "epoch": 3.44, + "learning_rate": 8.546770003649713e-06, + "loss": 0.8157, + "step": 53474 + }, + { + "epoch": 3.44, + "learning_rate": 8.546555314626764e-06, + "loss": 0.8172, + "step": 53475 + }, + { + "epoch": 3.44, + "learning_rate": 8.546340625603814e-06, + "loss": 0.9191, + "step": 53476 + }, + { + "epoch": 3.44, + "learning_rate": 8.546125936580863e-06, + "loss": 0.831, + "step": 53477 + }, + { + "epoch": 3.44, + "learning_rate": 8.545911247557914e-06, + "loss": 0.8129, + "step": 53478 + }, + { + "epoch": 3.44, + "learning_rate": 8.545696558534963e-06, + "loss": 0.8693, + "step": 53479 + }, + { + "epoch": 3.44, + "learning_rate": 8.545481869512012e-06, + "loss": 0.8156, + "step": 53480 + }, + { + "epoch": 3.44, + "learning_rate": 8.545267180489063e-06, + "loss": 0.8442, + "step": 53481 + }, + { + "epoch": 3.44, + "learning_rate": 8.545052491466112e-06, + "loss": 0.8099, + "step": 53482 + }, + { + "epoch": 3.44, + "learning_rate": 8.544837802443162e-06, + "loss": 0.7735, + "step": 53483 + }, + { + "epoch": 3.44, + "learning_rate": 8.544623113420211e-06, + "loss": 0.8129, + "step": 53484 + }, + { + "epoch": 3.44, + "learning_rate": 8.544408424397262e-06, + "loss": 0.8955, + "step": 53485 + }, + { + "epoch": 3.44, + "learning_rate": 8.54419373537431e-06, + "loss": 0.8494, + "step": 53486 + }, + { + "epoch": 3.44, + "learning_rate": 8.54397904635136e-06, + "loss": 0.8345, + "step": 53487 + }, + { + "epoch": 3.44, + "learning_rate": 8.54376435732841e-06, + "loss": 0.8385, + "step": 53488 + }, + { + "epoch": 3.44, + "learning_rate": 8.54354966830546e-06, + "loss": 0.8413, + "step": 53489 + }, + { + "epoch": 3.44, + "learning_rate": 8.54333497928251e-06, + "loss": 0.8455, + "step": 53490 + }, + { + "epoch": 3.44, + "learning_rate": 8.54312029025956e-06, + "loss": 0.8511, + "step": 53491 + }, + { + "epoch": 3.44, + "learning_rate": 8.54290560123661e-06, + "loss": 0.8162, + "step": 53492 + }, + { + "epoch": 3.44, + "learning_rate": 8.542690912213659e-06, + "loss": 0.8604, + "step": 53493 + }, + { + "epoch": 3.44, + "learning_rate": 8.54247622319071e-06, + "loss": 0.8423, + "step": 53494 + }, + { + "epoch": 3.44, + "learning_rate": 8.542261534167758e-06, + "loss": 0.8009, + "step": 53495 + }, + { + "epoch": 3.44, + "learning_rate": 8.542046845144807e-06, + "loss": 0.7971, + "step": 53496 + }, + { + "epoch": 3.44, + "learning_rate": 8.541832156121858e-06, + "loss": 0.81, + "step": 53497 + }, + { + "epoch": 3.44, + "learning_rate": 8.541617467098908e-06, + "loss": 0.8201, + "step": 53498 + }, + { + "epoch": 3.44, + "learning_rate": 8.541402778075957e-06, + "loss": 0.8175, + "step": 53499 + }, + { + "epoch": 3.44, + "learning_rate": 8.541188089053008e-06, + "loss": 0.8514, + "step": 53500 + }, + { + "epoch": 3.44, + "learning_rate": 8.540973400030057e-06, + "loss": 0.8391, + "step": 53501 + }, + { + "epoch": 3.44, + "learning_rate": 8.540758711007106e-06, + "loss": 0.8446, + "step": 53502 + }, + { + "epoch": 3.44, + "learning_rate": 8.540544021984157e-06, + "loss": 0.8156, + "step": 53503 + }, + { + "epoch": 3.44, + "learning_rate": 8.540329332961207e-06, + "loss": 0.849, + "step": 53504 + }, + { + "epoch": 3.44, + "learning_rate": 8.540114643938256e-06, + "loss": 0.7734, + "step": 53505 + }, + { + "epoch": 3.44, + "learning_rate": 8.539899954915305e-06, + "loss": 0.7805, + "step": 53506 + }, + { + "epoch": 3.44, + "learning_rate": 8.539685265892356e-06, + "loss": 0.8628, + "step": 53507 + }, + { + "epoch": 3.44, + "learning_rate": 8.539470576869405e-06, + "loss": 0.7997, + "step": 53508 + }, + { + "epoch": 3.44, + "learning_rate": 8.539255887846456e-06, + "loss": 0.8474, + "step": 53509 + }, + { + "epoch": 3.44, + "learning_rate": 8.539041198823505e-06, + "loss": 0.8759, + "step": 53510 + }, + { + "epoch": 3.44, + "learning_rate": 8.538826509800554e-06, + "loss": 0.8418, + "step": 53511 + }, + { + "epoch": 3.44, + "learning_rate": 8.538611820777604e-06, + "loss": 0.8149, + "step": 53512 + }, + { + "epoch": 3.44, + "learning_rate": 8.538397131754655e-06, + "loss": 0.7625, + "step": 53513 + }, + { + "epoch": 3.44, + "learning_rate": 8.538182442731704e-06, + "loss": 0.8115, + "step": 53514 + }, + { + "epoch": 3.44, + "learning_rate": 8.537967753708753e-06, + "loss": 0.8582, + "step": 53515 + }, + { + "epoch": 3.44, + "learning_rate": 8.537753064685803e-06, + "loss": 0.8596, + "step": 53516 + }, + { + "epoch": 3.44, + "learning_rate": 8.537538375662852e-06, + "loss": 0.8472, + "step": 53517 + }, + { + "epoch": 3.44, + "learning_rate": 8.537323686639903e-06, + "loss": 0.8101, + "step": 53518 + }, + { + "epoch": 3.44, + "learning_rate": 8.537108997616954e-06, + "loss": 0.7749, + "step": 53519 + }, + { + "epoch": 3.44, + "learning_rate": 8.536894308594003e-06, + "loss": 0.832, + "step": 53520 + }, + { + "epoch": 3.44, + "learning_rate": 8.536679619571052e-06, + "loss": 0.8119, + "step": 53521 + }, + { + "epoch": 3.44, + "learning_rate": 8.536464930548102e-06, + "loss": 0.7944, + "step": 53522 + }, + { + "epoch": 3.44, + "learning_rate": 8.536250241525151e-06, + "loss": 0.806, + "step": 53523 + }, + { + "epoch": 3.44, + "learning_rate": 8.5360355525022e-06, + "loss": 0.7787, + "step": 53524 + }, + { + "epoch": 3.44, + "learning_rate": 8.535820863479251e-06, + "loss": 0.8657, + "step": 53525 + }, + { + "epoch": 3.44, + "learning_rate": 8.535606174456302e-06, + "loss": 0.8963, + "step": 53526 + }, + { + "epoch": 3.44, + "learning_rate": 8.53539148543335e-06, + "loss": 0.7597, + "step": 53527 + }, + { + "epoch": 3.44, + "learning_rate": 8.535176796410401e-06, + "loss": 0.8116, + "step": 53528 + }, + { + "epoch": 3.44, + "learning_rate": 8.53496210738745e-06, + "loss": 0.8233, + "step": 53529 + }, + { + "epoch": 3.44, + "learning_rate": 8.534747418364499e-06, + "loss": 0.7561, + "step": 53530 + }, + { + "epoch": 3.44, + "learning_rate": 8.53453272934155e-06, + "loss": 0.8546, + "step": 53531 + }, + { + "epoch": 3.44, + "learning_rate": 8.534318040318599e-06, + "loss": 0.8414, + "step": 53532 + }, + { + "epoch": 3.44, + "learning_rate": 8.53410335129565e-06, + "loss": 0.8162, + "step": 53533 + }, + { + "epoch": 3.44, + "learning_rate": 8.533888662272698e-06, + "loss": 0.8448, + "step": 53534 + }, + { + "epoch": 3.44, + "learning_rate": 8.533673973249749e-06, + "loss": 0.8344, + "step": 53535 + }, + { + "epoch": 3.44, + "learning_rate": 8.533459284226798e-06, + "loss": 0.8444, + "step": 53536 + }, + { + "epoch": 3.44, + "learning_rate": 8.533244595203849e-06, + "loss": 0.7906, + "step": 53537 + }, + { + "epoch": 3.44, + "learning_rate": 8.533029906180898e-06, + "loss": 0.7966, + "step": 53538 + }, + { + "epoch": 3.44, + "learning_rate": 8.532815217157947e-06, + "loss": 0.839, + "step": 53539 + }, + { + "epoch": 3.44, + "learning_rate": 8.532600528134997e-06, + "loss": 0.84, + "step": 53540 + }, + { + "epoch": 3.44, + "learning_rate": 8.532385839112048e-06, + "loss": 0.8378, + "step": 53541 + }, + { + "epoch": 3.44, + "learning_rate": 8.532171150089097e-06, + "loss": 0.828, + "step": 53542 + }, + { + "epoch": 3.44, + "learning_rate": 8.531956461066146e-06, + "loss": 0.8677, + "step": 53543 + }, + { + "epoch": 3.44, + "learning_rate": 8.531741772043196e-06, + "loss": 0.8058, + "step": 53544 + }, + { + "epoch": 3.44, + "learning_rate": 8.531527083020245e-06, + "loss": 0.8499, + "step": 53545 + }, + { + "epoch": 3.45, + "learning_rate": 8.531312393997296e-06, + "loss": 0.7997, + "step": 53546 + }, + { + "epoch": 3.45, + "learning_rate": 8.531097704974345e-06, + "loss": 0.898, + "step": 53547 + }, + { + "epoch": 3.45, + "learning_rate": 8.530883015951396e-06, + "loss": 0.8509, + "step": 53548 + }, + { + "epoch": 3.45, + "learning_rate": 8.530668326928445e-06, + "loss": 0.7876, + "step": 53549 + }, + { + "epoch": 3.45, + "learning_rate": 8.530453637905495e-06, + "loss": 0.8241, + "step": 53550 + }, + { + "epoch": 3.45, + "learning_rate": 8.530238948882544e-06, + "loss": 0.7832, + "step": 53551 + }, + { + "epoch": 3.45, + "learning_rate": 8.530024259859593e-06, + "loss": 0.7869, + "step": 53552 + }, + { + "epoch": 3.45, + "learning_rate": 8.529809570836644e-06, + "loss": 0.8075, + "step": 53553 + }, + { + "epoch": 3.45, + "learning_rate": 8.529594881813693e-06, + "loss": 0.8663, + "step": 53554 + }, + { + "epoch": 3.45, + "learning_rate": 8.529380192790744e-06, + "loss": 0.8352, + "step": 53555 + }, + { + "epoch": 3.45, + "learning_rate": 8.529165503767794e-06, + "loss": 0.8137, + "step": 53556 + }, + { + "epoch": 3.45, + "learning_rate": 8.528950814744843e-06, + "loss": 0.7781, + "step": 53557 + }, + { + "epoch": 3.45, + "learning_rate": 8.528736125721892e-06, + "loss": 0.8136, + "step": 53558 + }, + { + "epoch": 3.45, + "learning_rate": 8.528521436698943e-06, + "loss": 0.8094, + "step": 53559 + }, + { + "epoch": 3.45, + "learning_rate": 8.528306747675992e-06, + "loss": 0.8316, + "step": 53560 + }, + { + "epoch": 3.45, + "learning_rate": 8.52809205865304e-06, + "loss": 0.7849, + "step": 53561 + }, + { + "epoch": 3.45, + "learning_rate": 8.527877369630091e-06, + "loss": 0.8711, + "step": 53562 + }, + { + "epoch": 3.45, + "learning_rate": 8.527662680607142e-06, + "loss": 0.7837, + "step": 53563 + }, + { + "epoch": 3.45, + "learning_rate": 8.527447991584191e-06, + "loss": 0.8136, + "step": 53564 + }, + { + "epoch": 3.45, + "learning_rate": 8.527233302561242e-06, + "loss": 0.7801, + "step": 53565 + }, + { + "epoch": 3.45, + "learning_rate": 8.52701861353829e-06, + "loss": 0.8765, + "step": 53566 + }, + { + "epoch": 3.45, + "learning_rate": 8.52680392451534e-06, + "loss": 0.7649, + "step": 53567 + }, + { + "epoch": 3.45, + "learning_rate": 8.52658923549239e-06, + "loss": 0.8656, + "step": 53568 + }, + { + "epoch": 3.45, + "learning_rate": 8.52637454646944e-06, + "loss": 0.8445, + "step": 53569 + }, + { + "epoch": 3.45, + "learning_rate": 8.52615985744649e-06, + "loss": 0.7861, + "step": 53570 + }, + { + "epoch": 3.45, + "learning_rate": 8.525945168423539e-06, + "loss": 0.7833, + "step": 53571 + }, + { + "epoch": 3.45, + "learning_rate": 8.52573047940059e-06, + "loss": 0.8227, + "step": 53572 + }, + { + "epoch": 3.45, + "learning_rate": 8.525515790377638e-06, + "loss": 0.8736, + "step": 53573 + }, + { + "epoch": 3.45, + "learning_rate": 8.525301101354687e-06, + "loss": 0.8512, + "step": 53574 + }, + { + "epoch": 3.45, + "learning_rate": 8.525086412331738e-06, + "loss": 0.8472, + "step": 53575 + }, + { + "epoch": 3.45, + "learning_rate": 8.524871723308787e-06, + "loss": 0.8283, + "step": 53576 + }, + { + "epoch": 3.45, + "learning_rate": 8.524657034285838e-06, + "loss": 0.8203, + "step": 53577 + }, + { + "epoch": 3.45, + "learning_rate": 8.524442345262888e-06, + "loss": 0.7892, + "step": 53578 + }, + { + "epoch": 3.45, + "learning_rate": 8.524227656239937e-06, + "loss": 0.81, + "step": 53579 + }, + { + "epoch": 3.45, + "learning_rate": 8.524012967216986e-06, + "loss": 0.7419, + "step": 53580 + }, + { + "epoch": 3.45, + "learning_rate": 8.523798278194037e-06, + "loss": 0.8522, + "step": 53581 + }, + { + "epoch": 3.45, + "learning_rate": 8.523583589171086e-06, + "loss": 0.8216, + "step": 53582 + }, + { + "epoch": 3.45, + "learning_rate": 8.523368900148135e-06, + "loss": 0.8063, + "step": 53583 + }, + { + "epoch": 3.45, + "learning_rate": 8.523154211125186e-06, + "loss": 0.8078, + "step": 53584 + }, + { + "epoch": 3.45, + "learning_rate": 8.522939522102236e-06, + "loss": 0.8217, + "step": 53585 + }, + { + "epoch": 3.45, + "learning_rate": 8.522724833079285e-06, + "loss": 0.8469, + "step": 53586 + }, + { + "epoch": 3.45, + "learning_rate": 8.522510144056336e-06, + "loss": 0.7976, + "step": 53587 + }, + { + "epoch": 3.45, + "learning_rate": 8.522295455033385e-06, + "loss": 0.8366, + "step": 53588 + }, + { + "epoch": 3.45, + "learning_rate": 8.522080766010434e-06, + "loss": 0.7724, + "step": 53589 + }, + { + "epoch": 3.45, + "learning_rate": 8.521866076987484e-06, + "loss": 0.797, + "step": 53590 + }, + { + "epoch": 3.45, + "learning_rate": 8.521651387964535e-06, + "loss": 0.8386, + "step": 53591 + }, + { + "epoch": 3.45, + "learning_rate": 8.521436698941584e-06, + "loss": 0.861, + "step": 53592 + }, + { + "epoch": 3.45, + "learning_rate": 8.521222009918633e-06, + "loss": 0.8675, + "step": 53593 + }, + { + "epoch": 3.45, + "learning_rate": 8.521007320895684e-06, + "loss": 0.8659, + "step": 53594 + }, + { + "epoch": 3.45, + "learning_rate": 8.520792631872733e-06, + "loss": 0.8035, + "step": 53595 + }, + { + "epoch": 3.45, + "learning_rate": 8.520577942849783e-06, + "loss": 0.8205, + "step": 53596 + }, + { + "epoch": 3.45, + "learning_rate": 8.520363253826832e-06, + "loss": 0.8058, + "step": 53597 + }, + { + "epoch": 3.45, + "learning_rate": 8.520148564803881e-06, + "loss": 0.8325, + "step": 53598 + }, + { + "epoch": 3.45, + "learning_rate": 8.519933875780932e-06, + "loss": 0.8562, + "step": 53599 + }, + { + "epoch": 3.45, + "learning_rate": 8.519719186757983e-06, + "loss": 0.7992, + "step": 53600 + }, + { + "epoch": 3.45, + "learning_rate": 8.519504497735032e-06, + "loss": 0.8207, + "step": 53601 + }, + { + "epoch": 3.45, + "learning_rate": 8.51928980871208e-06, + "loss": 0.8568, + "step": 53602 + }, + { + "epoch": 3.45, + "learning_rate": 8.519075119689131e-06, + "loss": 0.8523, + "step": 53603 + }, + { + "epoch": 3.45, + "learning_rate": 8.51886043066618e-06, + "loss": 0.8423, + "step": 53604 + }, + { + "epoch": 3.45, + "learning_rate": 8.51864574164323e-06, + "loss": 0.8521, + "step": 53605 + }, + { + "epoch": 3.45, + "learning_rate": 8.518431052620281e-06, + "loss": 0.8253, + "step": 53606 + }, + { + "epoch": 3.45, + "learning_rate": 8.51821636359733e-06, + "loss": 0.8401, + "step": 53607 + }, + { + "epoch": 3.45, + "learning_rate": 8.51800167457438e-06, + "loss": 0.891, + "step": 53608 + }, + { + "epoch": 3.45, + "learning_rate": 8.51778698555143e-06, + "loss": 0.8113, + "step": 53609 + }, + { + "epoch": 3.45, + "learning_rate": 8.517572296528479e-06, + "loss": 0.8391, + "step": 53610 + }, + { + "epoch": 3.45, + "learning_rate": 8.517357607505528e-06, + "loss": 0.8115, + "step": 53611 + }, + { + "epoch": 3.45, + "learning_rate": 8.517142918482579e-06, + "loss": 0.8865, + "step": 53612 + }, + { + "epoch": 3.45, + "learning_rate": 8.51692822945963e-06, + "loss": 0.8091, + "step": 53613 + }, + { + "epoch": 3.45, + "learning_rate": 8.516713540436678e-06, + "loss": 0.8441, + "step": 53614 + }, + { + "epoch": 3.45, + "learning_rate": 8.516498851413729e-06, + "loss": 0.8774, + "step": 53615 + }, + { + "epoch": 3.45, + "learning_rate": 8.516284162390778e-06, + "loss": 0.8008, + "step": 53616 + }, + { + "epoch": 3.45, + "learning_rate": 8.516069473367827e-06, + "loss": 0.8163, + "step": 53617 + }, + { + "epoch": 3.45, + "learning_rate": 8.515854784344877e-06, + "loss": 0.8606, + "step": 53618 + }, + { + "epoch": 3.45, + "learning_rate": 8.515640095321926e-06, + "loss": 0.7934, + "step": 53619 + }, + { + "epoch": 3.45, + "learning_rate": 8.515425406298977e-06, + "loss": 0.8266, + "step": 53620 + }, + { + "epoch": 3.45, + "learning_rate": 8.515210717276026e-06, + "loss": 0.8062, + "step": 53621 + }, + { + "epoch": 3.45, + "learning_rate": 8.514996028253077e-06, + "loss": 0.7854, + "step": 53622 + }, + { + "epoch": 3.45, + "learning_rate": 8.514781339230126e-06, + "loss": 0.7724, + "step": 53623 + }, + { + "epoch": 3.45, + "learning_rate": 8.514566650207176e-06, + "loss": 0.8149, + "step": 53624 + }, + { + "epoch": 3.45, + "learning_rate": 8.514351961184225e-06, + "loss": 0.7694, + "step": 53625 + }, + { + "epoch": 3.45, + "learning_rate": 8.514137272161274e-06, + "loss": 0.8755, + "step": 53626 + }, + { + "epoch": 3.45, + "learning_rate": 8.513922583138325e-06, + "loss": 0.8393, + "step": 53627 + }, + { + "epoch": 3.45, + "learning_rate": 8.513707894115376e-06, + "loss": 0.8219, + "step": 53628 + }, + { + "epoch": 3.45, + "learning_rate": 8.513493205092425e-06, + "loss": 0.7517, + "step": 53629 + }, + { + "epoch": 3.45, + "learning_rate": 8.513278516069474e-06, + "loss": 0.8075, + "step": 53630 + }, + { + "epoch": 3.45, + "learning_rate": 8.513063827046524e-06, + "loss": 0.8507, + "step": 53631 + }, + { + "epoch": 3.45, + "learning_rate": 8.512849138023573e-06, + "loss": 0.833, + "step": 53632 + }, + { + "epoch": 3.45, + "learning_rate": 8.512634449000624e-06, + "loss": 0.7949, + "step": 53633 + }, + { + "epoch": 3.45, + "learning_rate": 8.512419759977673e-06, + "loss": 0.7976, + "step": 53634 + }, + { + "epoch": 3.45, + "learning_rate": 8.512205070954723e-06, + "loss": 0.7954, + "step": 53635 + }, + { + "epoch": 3.45, + "learning_rate": 8.511990381931772e-06, + "loss": 0.8312, + "step": 53636 + }, + { + "epoch": 3.45, + "learning_rate": 8.511775692908823e-06, + "loss": 0.7837, + "step": 53637 + }, + { + "epoch": 3.45, + "learning_rate": 8.511561003885872e-06, + "loss": 0.8188, + "step": 53638 + }, + { + "epoch": 3.45, + "learning_rate": 8.511346314862921e-06, + "loss": 0.8598, + "step": 53639 + }, + { + "epoch": 3.45, + "learning_rate": 8.511131625839972e-06, + "loss": 0.8514, + "step": 53640 + }, + { + "epoch": 3.45, + "learning_rate": 8.51091693681702e-06, + "loss": 0.8088, + "step": 53641 + }, + { + "epoch": 3.45, + "learning_rate": 8.510702247794071e-06, + "loss": 0.7819, + "step": 53642 + }, + { + "epoch": 3.45, + "learning_rate": 8.510487558771122e-06, + "loss": 0.8766, + "step": 53643 + }, + { + "epoch": 3.45, + "learning_rate": 8.510272869748171e-06, + "loss": 0.9239, + "step": 53644 + }, + { + "epoch": 3.45, + "learning_rate": 8.51005818072522e-06, + "loss": 0.8664, + "step": 53645 + }, + { + "epoch": 3.45, + "learning_rate": 8.50984349170227e-06, + "loss": 0.861, + "step": 53646 + }, + { + "epoch": 3.45, + "learning_rate": 8.50962880267932e-06, + "loss": 0.7695, + "step": 53647 + }, + { + "epoch": 3.45, + "learning_rate": 8.509414113656368e-06, + "loss": 0.8775, + "step": 53648 + }, + { + "epoch": 3.45, + "learning_rate": 8.509199424633419e-06, + "loss": 0.8742, + "step": 53649 + }, + { + "epoch": 3.45, + "learning_rate": 8.50898473561047e-06, + "loss": 0.8145, + "step": 53650 + }, + { + "epoch": 3.45, + "learning_rate": 8.508770046587519e-06, + "loss": 0.8903, + "step": 53651 + }, + { + "epoch": 3.45, + "learning_rate": 8.508555357564568e-06, + "loss": 0.7875, + "step": 53652 + }, + { + "epoch": 3.45, + "learning_rate": 8.508340668541618e-06, + "loss": 0.8098, + "step": 53653 + }, + { + "epoch": 3.45, + "learning_rate": 8.508125979518667e-06, + "loss": 0.8586, + "step": 53654 + }, + { + "epoch": 3.45, + "learning_rate": 8.507911290495718e-06, + "loss": 0.8262, + "step": 53655 + }, + { + "epoch": 3.45, + "learning_rate": 8.507696601472767e-06, + "loss": 0.7875, + "step": 53656 + }, + { + "epoch": 3.45, + "learning_rate": 8.507481912449818e-06, + "loss": 0.8582, + "step": 53657 + }, + { + "epoch": 3.45, + "learning_rate": 8.507267223426867e-06, + "loss": 0.7889, + "step": 53658 + }, + { + "epoch": 3.45, + "learning_rate": 8.507052534403917e-06, + "loss": 0.8795, + "step": 53659 + }, + { + "epoch": 3.45, + "learning_rate": 8.506837845380966e-06, + "loss": 0.8512, + "step": 53660 + }, + { + "epoch": 3.45, + "learning_rate": 8.506623156358015e-06, + "loss": 0.8315, + "step": 53661 + }, + { + "epoch": 3.45, + "learning_rate": 8.506408467335066e-06, + "loss": 0.8568, + "step": 53662 + }, + { + "epoch": 3.45, + "learning_rate": 8.506193778312115e-06, + "loss": 0.7986, + "step": 53663 + }, + { + "epoch": 3.45, + "learning_rate": 8.505979089289165e-06, + "loss": 0.8057, + "step": 53664 + }, + { + "epoch": 3.45, + "learning_rate": 8.505764400266216e-06, + "loss": 0.8896, + "step": 53665 + }, + { + "epoch": 3.45, + "learning_rate": 8.505549711243265e-06, + "loss": 0.8833, + "step": 53666 + }, + { + "epoch": 3.45, + "learning_rate": 8.505335022220314e-06, + "loss": 0.8481, + "step": 53667 + }, + { + "epoch": 3.45, + "learning_rate": 8.505120333197365e-06, + "loss": 0.7832, + "step": 53668 + }, + { + "epoch": 3.45, + "learning_rate": 8.504905644174414e-06, + "loss": 0.8285, + "step": 53669 + }, + { + "epoch": 3.45, + "learning_rate": 8.504690955151463e-06, + "loss": 0.8965, + "step": 53670 + }, + { + "epoch": 3.45, + "learning_rate": 8.504476266128513e-06, + "loss": 0.8388, + "step": 53671 + }, + { + "epoch": 3.45, + "learning_rate": 8.504261577105564e-06, + "loss": 0.8001, + "step": 53672 + }, + { + "epoch": 3.45, + "learning_rate": 8.504046888082613e-06, + "loss": 0.8778, + "step": 53673 + }, + { + "epoch": 3.45, + "learning_rate": 8.503832199059664e-06, + "loss": 0.7729, + "step": 53674 + }, + { + "epoch": 3.45, + "learning_rate": 8.503617510036713e-06, + "loss": 0.781, + "step": 53675 + }, + { + "epoch": 3.45, + "learning_rate": 8.503402821013761e-06, + "loss": 0.8446, + "step": 53676 + }, + { + "epoch": 3.45, + "learning_rate": 8.503188131990812e-06, + "loss": 0.8637, + "step": 53677 + }, + { + "epoch": 3.45, + "learning_rate": 8.502973442967861e-06, + "loss": 0.8076, + "step": 53678 + }, + { + "epoch": 3.45, + "learning_rate": 8.502758753944912e-06, + "loss": 0.7675, + "step": 53679 + }, + { + "epoch": 3.45, + "learning_rate": 8.50254406492196e-06, + "loss": 0.8751, + "step": 53680 + }, + { + "epoch": 3.45, + "learning_rate": 8.502329375899011e-06, + "loss": 0.8113, + "step": 53681 + }, + { + "epoch": 3.45, + "learning_rate": 8.50211468687606e-06, + "loss": 0.8356, + "step": 53682 + }, + { + "epoch": 3.45, + "learning_rate": 8.501899997853111e-06, + "loss": 0.8313, + "step": 53683 + }, + { + "epoch": 3.45, + "learning_rate": 8.50168530883016e-06, + "loss": 0.8125, + "step": 53684 + }, + { + "epoch": 3.45, + "learning_rate": 8.501470619807209e-06, + "loss": 0.8022, + "step": 53685 + }, + { + "epoch": 3.45, + "learning_rate": 8.50125593078426e-06, + "loss": 0.8638, + "step": 53686 + }, + { + "epoch": 3.45, + "learning_rate": 8.50104124176131e-06, + "loss": 0.8695, + "step": 53687 + }, + { + "epoch": 3.45, + "learning_rate": 8.50082655273836e-06, + "loss": 0.8258, + "step": 53688 + }, + { + "epoch": 3.45, + "learning_rate": 8.500611863715408e-06, + "loss": 0.884, + "step": 53689 + }, + { + "epoch": 3.45, + "learning_rate": 8.500397174692459e-06, + "loss": 0.8472, + "step": 53690 + }, + { + "epoch": 3.45, + "learning_rate": 8.500182485669508e-06, + "loss": 0.8289, + "step": 53691 + }, + { + "epoch": 3.45, + "learning_rate": 8.499967796646558e-06, + "loss": 0.8563, + "step": 53692 + }, + { + "epoch": 3.45, + "learning_rate": 8.499753107623609e-06, + "loss": 0.8481, + "step": 53693 + }, + { + "epoch": 3.45, + "learning_rate": 8.499538418600658e-06, + "loss": 0.794, + "step": 53694 + }, + { + "epoch": 3.45, + "learning_rate": 8.499323729577707e-06, + "loss": 0.8025, + "step": 53695 + }, + { + "epoch": 3.45, + "learning_rate": 8.499109040554758e-06, + "loss": 0.8208, + "step": 53696 + }, + { + "epoch": 3.45, + "learning_rate": 8.498894351531807e-06, + "loss": 0.8709, + "step": 53697 + }, + { + "epoch": 3.45, + "learning_rate": 8.498679662508856e-06, + "loss": 0.8284, + "step": 53698 + }, + { + "epoch": 3.45, + "learning_rate": 8.498464973485906e-06, + "loss": 0.85, + "step": 53699 + }, + { + "epoch": 3.45, + "learning_rate": 8.498250284462957e-06, + "loss": 0.837, + "step": 53700 + }, + { + "epoch": 3.45, + "learning_rate": 8.498035595440006e-06, + "loss": 0.8521, + "step": 53701 + }, + { + "epoch": 3.46, + "learning_rate": 8.497820906417057e-06, + "loss": 0.8448, + "step": 53702 + }, + { + "epoch": 3.46, + "learning_rate": 8.497606217394106e-06, + "loss": 0.9285, + "step": 53703 + }, + { + "epoch": 3.46, + "learning_rate": 8.497391528371155e-06, + "loss": 0.8235, + "step": 53704 + }, + { + "epoch": 3.46, + "learning_rate": 8.497176839348205e-06, + "loss": 0.8597, + "step": 53705 + }, + { + "epoch": 3.46, + "learning_rate": 8.496962150325254e-06, + "loss": 0.8247, + "step": 53706 + }, + { + "epoch": 3.46, + "learning_rate": 8.496747461302303e-06, + "loss": 0.8209, + "step": 53707 + }, + { + "epoch": 3.46, + "learning_rate": 8.496532772279354e-06, + "loss": 0.8244, + "step": 53708 + }, + { + "epoch": 3.46, + "learning_rate": 8.496318083256404e-06, + "loss": 0.8198, + "step": 53709 + }, + { + "epoch": 3.46, + "learning_rate": 8.496103394233453e-06, + "loss": 0.8962, + "step": 53710 + }, + { + "epoch": 3.46, + "learning_rate": 8.495888705210504e-06, + "loss": 0.8466, + "step": 53711 + }, + { + "epoch": 3.46, + "learning_rate": 8.495674016187553e-06, + "loss": 0.8105, + "step": 53712 + }, + { + "epoch": 3.46, + "learning_rate": 8.495459327164602e-06, + "loss": 0.8591, + "step": 53713 + }, + { + "epoch": 3.46, + "learning_rate": 8.495244638141653e-06, + "loss": 0.7961, + "step": 53714 + }, + { + "epoch": 3.46, + "learning_rate": 8.495029949118703e-06, + "loss": 0.8512, + "step": 53715 + }, + { + "epoch": 3.46, + "learning_rate": 8.494815260095752e-06, + "loss": 0.8959, + "step": 53716 + }, + { + "epoch": 3.46, + "learning_rate": 8.494600571072801e-06, + "loss": 0.862, + "step": 53717 + }, + { + "epoch": 3.46, + "learning_rate": 8.494385882049852e-06, + "loss": 0.806, + "step": 53718 + }, + { + "epoch": 3.46, + "learning_rate": 8.4941711930269e-06, + "loss": 0.8214, + "step": 53719 + }, + { + "epoch": 3.46, + "learning_rate": 8.493956504003952e-06, + "loss": 0.8376, + "step": 53720 + }, + { + "epoch": 3.46, + "learning_rate": 8.493741814981e-06, + "loss": 0.828, + "step": 53721 + }, + { + "epoch": 3.46, + "learning_rate": 8.493527125958051e-06, + "loss": 0.8674, + "step": 53722 + }, + { + "epoch": 3.46, + "learning_rate": 8.4933124369351e-06, + "loss": 0.7972, + "step": 53723 + }, + { + "epoch": 3.46, + "learning_rate": 8.49309774791215e-06, + "loss": 0.8857, + "step": 53724 + }, + { + "epoch": 3.46, + "learning_rate": 8.4928830588892e-06, + "loss": 0.8261, + "step": 53725 + }, + { + "epoch": 3.46, + "learning_rate": 8.492668369866249e-06, + "loss": 0.8162, + "step": 53726 + }, + { + "epoch": 3.46, + "learning_rate": 8.4924536808433e-06, + "loss": 0.8258, + "step": 53727 + }, + { + "epoch": 3.46, + "learning_rate": 8.492238991820348e-06, + "loss": 0.795, + "step": 53728 + }, + { + "epoch": 3.46, + "learning_rate": 8.492024302797399e-06, + "loss": 0.8284, + "step": 53729 + }, + { + "epoch": 3.46, + "learning_rate": 8.49180961377445e-06, + "loss": 0.8047, + "step": 53730 + }, + { + "epoch": 3.46, + "learning_rate": 8.491594924751499e-06, + "loss": 0.8681, + "step": 53731 + }, + { + "epoch": 3.46, + "learning_rate": 8.491380235728548e-06, + "loss": 0.8306, + "step": 53732 + }, + { + "epoch": 3.46, + "learning_rate": 8.491165546705598e-06, + "loss": 0.8369, + "step": 53733 + }, + { + "epoch": 3.46, + "learning_rate": 8.490950857682647e-06, + "loss": 0.8241, + "step": 53734 + }, + { + "epoch": 3.46, + "learning_rate": 8.490736168659696e-06, + "loss": 0.8576, + "step": 53735 + }, + { + "epoch": 3.46, + "learning_rate": 8.490521479636747e-06, + "loss": 0.8414, + "step": 53736 + }, + { + "epoch": 3.46, + "learning_rate": 8.490306790613797e-06, + "loss": 0.8313, + "step": 53737 + }, + { + "epoch": 3.46, + "learning_rate": 8.490092101590846e-06, + "loss": 0.7851, + "step": 53738 + }, + { + "epoch": 3.46, + "learning_rate": 8.489877412567895e-06, + "loss": 0.853, + "step": 53739 + }, + { + "epoch": 3.46, + "learning_rate": 8.489662723544946e-06, + "loss": 0.8133, + "step": 53740 + }, + { + "epoch": 3.46, + "learning_rate": 8.489448034521995e-06, + "loss": 0.8705, + "step": 53741 + }, + { + "epoch": 3.46, + "learning_rate": 8.489233345499046e-06, + "loss": 0.7918, + "step": 53742 + }, + { + "epoch": 3.46, + "learning_rate": 8.489018656476095e-06, + "loss": 0.7843, + "step": 53743 + }, + { + "epoch": 3.46, + "learning_rate": 8.488803967453145e-06, + "loss": 0.8455, + "step": 53744 + }, + { + "epoch": 3.46, + "learning_rate": 8.488589278430194e-06, + "loss": 0.8218, + "step": 53745 + }, + { + "epoch": 3.46, + "learning_rate": 8.488374589407245e-06, + "loss": 0.8798, + "step": 53746 + }, + { + "epoch": 3.46, + "learning_rate": 8.488159900384294e-06, + "loss": 0.7912, + "step": 53747 + }, + { + "epoch": 3.46, + "learning_rate": 8.487945211361343e-06, + "loss": 0.7957, + "step": 53748 + }, + { + "epoch": 3.46, + "learning_rate": 8.487730522338394e-06, + "loss": 0.7794, + "step": 53749 + }, + { + "epoch": 3.46, + "learning_rate": 8.487515833315442e-06, + "loss": 0.7902, + "step": 53750 + }, + { + "epoch": 3.46, + "learning_rate": 8.487301144292493e-06, + "loss": 0.8124, + "step": 53751 + }, + { + "epoch": 3.46, + "learning_rate": 8.487086455269544e-06, + "loss": 0.7396, + "step": 53752 + }, + { + "epoch": 3.46, + "learning_rate": 8.486871766246593e-06, + "loss": 0.7997, + "step": 53753 + }, + { + "epoch": 3.46, + "learning_rate": 8.486657077223642e-06, + "loss": 0.7886, + "step": 53754 + }, + { + "epoch": 3.46, + "learning_rate": 8.486442388200692e-06, + "loss": 0.7895, + "step": 53755 + }, + { + "epoch": 3.46, + "learning_rate": 8.486227699177741e-06, + "loss": 0.7869, + "step": 53756 + }, + { + "epoch": 3.46, + "learning_rate": 8.48601301015479e-06, + "loss": 0.8944, + "step": 53757 + }, + { + "epoch": 3.46, + "learning_rate": 8.485798321131841e-06, + "loss": 0.818, + "step": 53758 + }, + { + "epoch": 3.46, + "learning_rate": 8.485583632108892e-06, + "loss": 0.8233, + "step": 53759 + }, + { + "epoch": 3.46, + "learning_rate": 8.48536894308594e-06, + "loss": 0.8784, + "step": 53760 + }, + { + "epoch": 3.46, + "learning_rate": 8.485154254062991e-06, + "loss": 0.7803, + "step": 53761 + }, + { + "epoch": 3.46, + "learning_rate": 8.48493956504004e-06, + "loss": 0.8358, + "step": 53762 + }, + { + "epoch": 3.46, + "learning_rate": 8.48472487601709e-06, + "loss": 0.8096, + "step": 53763 + }, + { + "epoch": 3.46, + "learning_rate": 8.48451018699414e-06, + "loss": 0.7963, + "step": 53764 + }, + { + "epoch": 3.46, + "learning_rate": 8.484295497971189e-06, + "loss": 0.8931, + "step": 53765 + }, + { + "epoch": 3.46, + "learning_rate": 8.48408080894824e-06, + "loss": 0.7682, + "step": 53766 + }, + { + "epoch": 3.46, + "learning_rate": 8.483866119925288e-06, + "loss": 0.8004, + "step": 53767 + }, + { + "epoch": 3.46, + "learning_rate": 8.483651430902339e-06, + "loss": 0.8381, + "step": 53768 + }, + { + "epoch": 3.46, + "learning_rate": 8.483436741879388e-06, + "loss": 0.8221, + "step": 53769 + }, + { + "epoch": 3.46, + "learning_rate": 8.483222052856439e-06, + "loss": 0.8835, + "step": 53770 + }, + { + "epoch": 3.46, + "learning_rate": 8.483007363833488e-06, + "loss": 0.7876, + "step": 53771 + }, + { + "epoch": 3.46, + "learning_rate": 8.482792674810537e-06, + "loss": 0.8494, + "step": 53772 + }, + { + "epoch": 3.46, + "learning_rate": 8.482577985787587e-06, + "loss": 0.8348, + "step": 53773 + }, + { + "epoch": 3.46, + "learning_rate": 8.482363296764638e-06, + "loss": 0.7731, + "step": 53774 + }, + { + "epoch": 3.46, + "learning_rate": 8.482148607741687e-06, + "loss": 0.7745, + "step": 53775 + }, + { + "epoch": 3.46, + "learning_rate": 8.481933918718736e-06, + "loss": 0.8589, + "step": 53776 + }, + { + "epoch": 3.46, + "learning_rate": 8.481719229695787e-06, + "loss": 0.7993, + "step": 53777 + }, + { + "epoch": 3.46, + "learning_rate": 8.481504540672836e-06, + "loss": 0.7585, + "step": 53778 + }, + { + "epoch": 3.46, + "learning_rate": 8.481289851649886e-06, + "loss": 0.8105, + "step": 53779 + }, + { + "epoch": 3.46, + "learning_rate": 8.481075162626937e-06, + "loss": 0.8401, + "step": 53780 + }, + { + "epoch": 3.46, + "learning_rate": 8.480860473603986e-06, + "loss": 0.8302, + "step": 53781 + }, + { + "epoch": 3.46, + "learning_rate": 8.480645784581035e-06, + "loss": 0.8195, + "step": 53782 + }, + { + "epoch": 3.46, + "learning_rate": 8.480431095558085e-06, + "loss": 0.9434, + "step": 53783 + }, + { + "epoch": 3.46, + "learning_rate": 8.480216406535134e-06, + "loss": 0.8663, + "step": 53784 + }, + { + "epoch": 3.46, + "learning_rate": 8.480001717512183e-06, + "loss": 0.9041, + "step": 53785 + }, + { + "epoch": 3.46, + "learning_rate": 8.479787028489234e-06, + "loss": 0.8303, + "step": 53786 + }, + { + "epoch": 3.46, + "learning_rate": 8.479572339466285e-06, + "loss": 0.81, + "step": 53787 + }, + { + "epoch": 3.46, + "learning_rate": 8.479357650443334e-06, + "loss": 0.8579, + "step": 53788 + }, + { + "epoch": 3.46, + "learning_rate": 8.479142961420384e-06, + "loss": 0.8419, + "step": 53789 + }, + { + "epoch": 3.46, + "learning_rate": 8.478928272397433e-06, + "loss": 0.8254, + "step": 53790 + }, + { + "epoch": 3.46, + "learning_rate": 8.478713583374482e-06, + "loss": 0.7926, + "step": 53791 + }, + { + "epoch": 3.46, + "learning_rate": 8.478498894351533e-06, + "loss": 0.8412, + "step": 53792 + }, + { + "epoch": 3.46, + "learning_rate": 8.478284205328582e-06, + "loss": 0.807, + "step": 53793 + }, + { + "epoch": 3.46, + "learning_rate": 8.47806951630563e-06, + "loss": 0.8122, + "step": 53794 + }, + { + "epoch": 3.46, + "learning_rate": 8.477854827282681e-06, + "loss": 0.8036, + "step": 53795 + }, + { + "epoch": 3.46, + "learning_rate": 8.477640138259732e-06, + "loss": 0.8115, + "step": 53796 + }, + { + "epoch": 3.46, + "learning_rate": 8.477425449236781e-06, + "loss": 0.8419, + "step": 53797 + }, + { + "epoch": 3.46, + "learning_rate": 8.477210760213832e-06, + "loss": 0.8474, + "step": 53798 + }, + { + "epoch": 3.46, + "learning_rate": 8.47699607119088e-06, + "loss": 0.7831, + "step": 53799 + }, + { + "epoch": 3.46, + "learning_rate": 8.47678138216793e-06, + "loss": 0.832, + "step": 53800 + }, + { + "epoch": 3.46, + "learning_rate": 8.47656669314498e-06, + "loss": 0.8058, + "step": 53801 + }, + { + "epoch": 3.46, + "learning_rate": 8.476352004122031e-06, + "loss": 0.8314, + "step": 53802 + }, + { + "epoch": 3.46, + "learning_rate": 8.47613731509908e-06, + "loss": 0.8115, + "step": 53803 + }, + { + "epoch": 3.46, + "learning_rate": 8.475922626076129e-06, + "loss": 0.7998, + "step": 53804 + }, + { + "epoch": 3.46, + "learning_rate": 8.47570793705318e-06, + "loss": 0.7941, + "step": 53805 + }, + { + "epoch": 3.46, + "learning_rate": 8.475493248030229e-06, + "loss": 0.8551, + "step": 53806 + }, + { + "epoch": 3.46, + "learning_rate": 8.475278559007278e-06, + "loss": 0.7412, + "step": 53807 + }, + { + "epoch": 3.46, + "learning_rate": 8.475063869984328e-06, + "loss": 0.8332, + "step": 53808 + }, + { + "epoch": 3.46, + "learning_rate": 8.474849180961379e-06, + "loss": 0.7716, + "step": 53809 + }, + { + "epoch": 3.46, + "learning_rate": 8.474634491938428e-06, + "loss": 0.8788, + "step": 53810 + }, + { + "epoch": 3.46, + "learning_rate": 8.474419802915478e-06, + "loss": 0.82, + "step": 53811 + }, + { + "epoch": 3.46, + "learning_rate": 8.474205113892527e-06, + "loss": 0.7907, + "step": 53812 + }, + { + "epoch": 3.46, + "learning_rate": 8.473990424869576e-06, + "loss": 0.8715, + "step": 53813 + }, + { + "epoch": 3.46, + "learning_rate": 8.473775735846627e-06, + "loss": 0.8295, + "step": 53814 + }, + { + "epoch": 3.46, + "learning_rate": 8.473561046823676e-06, + "loss": 0.8685, + "step": 53815 + }, + { + "epoch": 3.46, + "learning_rate": 8.473346357800727e-06, + "loss": 0.8331, + "step": 53816 + }, + { + "epoch": 3.46, + "learning_rate": 8.473131668777776e-06, + "loss": 0.7843, + "step": 53817 + }, + { + "epoch": 3.46, + "learning_rate": 8.472916979754826e-06, + "loss": 0.8263, + "step": 53818 + }, + { + "epoch": 3.46, + "learning_rate": 8.472702290731875e-06, + "loss": 0.7798, + "step": 53819 + }, + { + "epoch": 3.46, + "learning_rate": 8.472487601708926e-06, + "loss": 0.8483, + "step": 53820 + }, + { + "epoch": 3.46, + "learning_rate": 8.472272912685975e-06, + "loss": 0.7526, + "step": 53821 + }, + { + "epoch": 3.46, + "learning_rate": 8.472058223663024e-06, + "loss": 0.8739, + "step": 53822 + }, + { + "epoch": 3.46, + "learning_rate": 8.471843534640075e-06, + "loss": 0.794, + "step": 53823 + }, + { + "epoch": 3.46, + "learning_rate": 8.471628845617125e-06, + "loss": 0.7596, + "step": 53824 + }, + { + "epoch": 3.46, + "learning_rate": 8.471414156594174e-06, + "loss": 0.8327, + "step": 53825 + }, + { + "epoch": 3.46, + "learning_rate": 8.471199467571223e-06, + "loss": 0.7912, + "step": 53826 + }, + { + "epoch": 3.46, + "learning_rate": 8.470984778548274e-06, + "loss": 0.9225, + "step": 53827 + }, + { + "epoch": 3.46, + "learning_rate": 8.470770089525323e-06, + "loss": 0.8028, + "step": 53828 + }, + { + "epoch": 3.46, + "learning_rate": 8.470555400502373e-06, + "loss": 0.8329, + "step": 53829 + }, + { + "epoch": 3.46, + "learning_rate": 8.470340711479422e-06, + "loss": 0.774, + "step": 53830 + }, + { + "epoch": 3.46, + "learning_rate": 8.470126022456473e-06, + "loss": 0.855, + "step": 53831 + }, + { + "epoch": 3.46, + "learning_rate": 8.469911333433522e-06, + "loss": 0.7432, + "step": 53832 + }, + { + "epoch": 3.46, + "learning_rate": 8.469696644410573e-06, + "loss": 0.8795, + "step": 53833 + }, + { + "epoch": 3.46, + "learning_rate": 8.469481955387622e-06, + "loss": 0.8012, + "step": 53834 + }, + { + "epoch": 3.46, + "learning_rate": 8.46926726636467e-06, + "loss": 0.8244, + "step": 53835 + }, + { + "epoch": 3.46, + "learning_rate": 8.469052577341721e-06, + "loss": 0.8369, + "step": 53836 + }, + { + "epoch": 3.46, + "learning_rate": 8.46883788831877e-06, + "loss": 0.7842, + "step": 53837 + }, + { + "epoch": 3.46, + "learning_rate": 8.46862319929582e-06, + "loss": 0.8955, + "step": 53838 + }, + { + "epoch": 3.46, + "learning_rate": 8.468408510272872e-06, + "loss": 0.8111, + "step": 53839 + }, + { + "epoch": 3.46, + "learning_rate": 8.46819382124992e-06, + "loss": 0.7931, + "step": 53840 + }, + { + "epoch": 3.46, + "learning_rate": 8.46797913222697e-06, + "loss": 0.8451, + "step": 53841 + }, + { + "epoch": 3.46, + "learning_rate": 8.46776444320402e-06, + "loss": 0.8239, + "step": 53842 + }, + { + "epoch": 3.46, + "learning_rate": 8.467549754181069e-06, + "loss": 0.8429, + "step": 53843 + }, + { + "epoch": 3.46, + "learning_rate": 8.467335065158118e-06, + "loss": 0.8234, + "step": 53844 + }, + { + "epoch": 3.46, + "learning_rate": 8.467120376135169e-06, + "loss": 0.872, + "step": 53845 + }, + { + "epoch": 3.46, + "learning_rate": 8.46690568711222e-06, + "loss": 0.7841, + "step": 53846 + }, + { + "epoch": 3.46, + "learning_rate": 8.466690998089268e-06, + "loss": 0.8411, + "step": 53847 + }, + { + "epoch": 3.46, + "learning_rate": 8.466476309066319e-06, + "loss": 0.8112, + "step": 53848 + }, + { + "epoch": 3.46, + "learning_rate": 8.466261620043368e-06, + "loss": 0.8166, + "step": 53849 + }, + { + "epoch": 3.46, + "learning_rate": 8.466046931020417e-06, + "loss": 0.8882, + "step": 53850 + }, + { + "epoch": 3.46, + "learning_rate": 8.465832241997468e-06, + "loss": 0.8308, + "step": 53851 + }, + { + "epoch": 3.46, + "learning_rate": 8.465617552974517e-06, + "loss": 0.8195, + "step": 53852 + }, + { + "epoch": 3.46, + "learning_rate": 8.465402863951567e-06, + "loss": 0.8406, + "step": 53853 + }, + { + "epoch": 3.46, + "learning_rate": 8.465188174928616e-06, + "loss": 0.8841, + "step": 53854 + }, + { + "epoch": 3.46, + "learning_rate": 8.464973485905667e-06, + "loss": 0.815, + "step": 53855 + }, + { + "epoch": 3.46, + "learning_rate": 8.464758796882716e-06, + "loss": 0.8324, + "step": 53856 + }, + { + "epoch": 3.47, + "learning_rate": 8.464544107859766e-06, + "loss": 0.8397, + "step": 53857 + }, + { + "epoch": 3.47, + "learning_rate": 8.464329418836815e-06, + "loss": 0.8299, + "step": 53858 + }, + { + "epoch": 3.47, + "learning_rate": 8.464114729813864e-06, + "loss": 0.8603, + "step": 53859 + }, + { + "epoch": 3.47, + "learning_rate": 8.463900040790915e-06, + "loss": 0.8126, + "step": 53860 + }, + { + "epoch": 3.47, + "learning_rate": 8.463685351767966e-06, + "loss": 0.756, + "step": 53861 + }, + { + "epoch": 3.47, + "learning_rate": 8.463470662745015e-06, + "loss": 0.803, + "step": 53862 + }, + { + "epoch": 3.47, + "learning_rate": 8.463255973722064e-06, + "loss": 0.8229, + "step": 53863 + }, + { + "epoch": 3.47, + "learning_rate": 8.463041284699114e-06, + "loss": 0.846, + "step": 53864 + }, + { + "epoch": 3.47, + "learning_rate": 8.462826595676163e-06, + "loss": 0.7536, + "step": 53865 + }, + { + "epoch": 3.47, + "learning_rate": 8.462611906653214e-06, + "loss": 0.8271, + "step": 53866 + }, + { + "epoch": 3.47, + "learning_rate": 8.462397217630265e-06, + "loss": 0.7796, + "step": 53867 + }, + { + "epoch": 3.47, + "learning_rate": 8.462182528607314e-06, + "loss": 0.8289, + "step": 53868 + }, + { + "epoch": 3.47, + "learning_rate": 8.461967839584362e-06, + "loss": 0.7481, + "step": 53869 + }, + { + "epoch": 3.47, + "learning_rate": 8.461753150561413e-06, + "loss": 0.898, + "step": 53870 + }, + { + "epoch": 3.47, + "learning_rate": 8.461538461538462e-06, + "loss": 0.7771, + "step": 53871 + }, + { + "epoch": 3.47, + "learning_rate": 8.461323772515511e-06, + "loss": 0.8375, + "step": 53872 + }, + { + "epoch": 3.47, + "learning_rate": 8.461109083492562e-06, + "loss": 0.8672, + "step": 53873 + }, + { + "epoch": 3.47, + "learning_rate": 8.46089439446961e-06, + "loss": 0.8293, + "step": 53874 + }, + { + "epoch": 3.47, + "learning_rate": 8.460679705446661e-06, + "loss": 0.87, + "step": 53875 + }, + { + "epoch": 3.47, + "learning_rate": 8.460465016423712e-06, + "loss": 0.7825, + "step": 53876 + }, + { + "epoch": 3.47, + "learning_rate": 8.460250327400761e-06, + "loss": 0.8482, + "step": 53877 + }, + { + "epoch": 3.47, + "learning_rate": 8.46003563837781e-06, + "loss": 0.8686, + "step": 53878 + }, + { + "epoch": 3.47, + "learning_rate": 8.45982094935486e-06, + "loss": 0.855, + "step": 53879 + }, + { + "epoch": 3.47, + "learning_rate": 8.45960626033191e-06, + "loss": 0.7558, + "step": 53880 + }, + { + "epoch": 3.47, + "learning_rate": 8.459391571308959e-06, + "loss": 0.7716, + "step": 53881 + }, + { + "epoch": 3.47, + "learning_rate": 8.45917688228601e-06, + "loss": 0.8474, + "step": 53882 + }, + { + "epoch": 3.47, + "learning_rate": 8.45896219326306e-06, + "loss": 0.8884, + "step": 53883 + }, + { + "epoch": 3.47, + "learning_rate": 8.458747504240109e-06, + "loss": 0.8767, + "step": 53884 + }, + { + "epoch": 3.47, + "learning_rate": 8.45853281521716e-06, + "loss": 0.8401, + "step": 53885 + }, + { + "epoch": 3.47, + "learning_rate": 8.458318126194208e-06, + "loss": 0.8107, + "step": 53886 + }, + { + "epoch": 3.47, + "learning_rate": 8.458103437171257e-06, + "loss": 0.8351, + "step": 53887 + }, + { + "epoch": 3.47, + "learning_rate": 8.457888748148308e-06, + "loss": 0.8672, + "step": 53888 + }, + { + "epoch": 3.47, + "learning_rate": 8.457674059125359e-06, + "loss": 0.8108, + "step": 53889 + }, + { + "epoch": 3.47, + "learning_rate": 8.457459370102408e-06, + "loss": 0.8382, + "step": 53890 + }, + { + "epoch": 3.47, + "learning_rate": 8.457244681079457e-06, + "loss": 0.8324, + "step": 53891 + }, + { + "epoch": 3.47, + "learning_rate": 8.457029992056507e-06, + "loss": 0.7818, + "step": 53892 + }, + { + "epoch": 3.47, + "learning_rate": 8.456815303033556e-06, + "loss": 0.8159, + "step": 53893 + }, + { + "epoch": 3.47, + "learning_rate": 8.456600614010605e-06, + "loss": 0.8166, + "step": 53894 + }, + { + "epoch": 3.47, + "learning_rate": 8.456385924987656e-06, + "loss": 0.9128, + "step": 53895 + }, + { + "epoch": 3.47, + "learning_rate": 8.456171235964707e-06, + "loss": 0.8794, + "step": 53896 + }, + { + "epoch": 3.47, + "learning_rate": 8.455956546941756e-06, + "loss": 0.8386, + "step": 53897 + }, + { + "epoch": 3.47, + "learning_rate": 8.455741857918806e-06, + "loss": 0.8596, + "step": 53898 + }, + { + "epoch": 3.47, + "learning_rate": 8.455527168895855e-06, + "loss": 0.8368, + "step": 53899 + }, + { + "epoch": 3.47, + "learning_rate": 8.455312479872904e-06, + "loss": 0.8055, + "step": 53900 + }, + { + "epoch": 3.47, + "learning_rate": 8.455097790849955e-06, + "loss": 0.7921, + "step": 53901 + }, + { + "epoch": 3.47, + "learning_rate": 8.454883101827004e-06, + "loss": 0.8489, + "step": 53902 + }, + { + "epoch": 3.47, + "learning_rate": 8.454668412804053e-06, + "loss": 0.8616, + "step": 53903 + }, + { + "epoch": 3.47, + "learning_rate": 8.454453723781103e-06, + "loss": 0.8461, + "step": 53904 + }, + { + "epoch": 3.47, + "learning_rate": 8.454239034758154e-06, + "loss": 0.8841, + "step": 53905 + }, + { + "epoch": 3.47, + "learning_rate": 8.454024345735203e-06, + "loss": 0.8085, + "step": 53906 + }, + { + "epoch": 3.47, + "learning_rate": 8.453809656712254e-06, + "loss": 0.782, + "step": 53907 + }, + { + "epoch": 3.47, + "learning_rate": 8.453594967689303e-06, + "loss": 0.8182, + "step": 53908 + }, + { + "epoch": 3.47, + "learning_rate": 8.453380278666352e-06, + "loss": 0.7568, + "step": 53909 + }, + { + "epoch": 3.47, + "learning_rate": 8.453165589643402e-06, + "loss": 0.814, + "step": 53910 + }, + { + "epoch": 3.47, + "learning_rate": 8.452950900620453e-06, + "loss": 0.803, + "step": 53911 + }, + { + "epoch": 3.47, + "learning_rate": 8.452736211597502e-06, + "loss": 0.8083, + "step": 53912 + }, + { + "epoch": 3.47, + "learning_rate": 8.45252152257455e-06, + "loss": 0.7912, + "step": 53913 + }, + { + "epoch": 3.47, + "learning_rate": 8.452306833551601e-06, + "loss": 0.7968, + "step": 53914 + }, + { + "epoch": 3.47, + "learning_rate": 8.45209214452865e-06, + "loss": 0.8192, + "step": 53915 + }, + { + "epoch": 3.47, + "learning_rate": 8.451877455505701e-06, + "loss": 0.8608, + "step": 53916 + }, + { + "epoch": 3.47, + "learning_rate": 8.45166276648275e-06, + "loss": 0.8208, + "step": 53917 + }, + { + "epoch": 3.47, + "learning_rate": 8.4514480774598e-06, + "loss": 0.8279, + "step": 53918 + }, + { + "epoch": 3.47, + "learning_rate": 8.45123338843685e-06, + "loss": 0.855, + "step": 53919 + }, + { + "epoch": 3.47, + "learning_rate": 8.4510186994139e-06, + "loss": 0.8736, + "step": 53920 + }, + { + "epoch": 3.47, + "learning_rate": 8.45080401039095e-06, + "loss": 0.7739, + "step": 53921 + }, + { + "epoch": 3.47, + "learning_rate": 8.450589321367998e-06, + "loss": 0.7505, + "step": 53922 + }, + { + "epoch": 3.47, + "learning_rate": 8.450374632345049e-06, + "loss": 0.8445, + "step": 53923 + }, + { + "epoch": 3.47, + "learning_rate": 8.450159943322098e-06, + "loss": 0.8123, + "step": 53924 + }, + { + "epoch": 3.47, + "learning_rate": 8.449945254299149e-06, + "loss": 0.7984, + "step": 53925 + }, + { + "epoch": 3.47, + "learning_rate": 8.4497305652762e-06, + "loss": 0.8307, + "step": 53926 + }, + { + "epoch": 3.47, + "learning_rate": 8.449515876253248e-06, + "loss": 0.8758, + "step": 53927 + }, + { + "epoch": 3.47, + "learning_rate": 8.449301187230297e-06, + "loss": 0.8863, + "step": 53928 + }, + { + "epoch": 3.47, + "learning_rate": 8.449086498207348e-06, + "loss": 0.8121, + "step": 53929 + }, + { + "epoch": 3.47, + "learning_rate": 8.448871809184397e-06, + "loss": 0.8211, + "step": 53930 + }, + { + "epoch": 3.47, + "learning_rate": 8.448657120161446e-06, + "loss": 0.8347, + "step": 53931 + }, + { + "epoch": 3.47, + "learning_rate": 8.448442431138496e-06, + "loss": 0.8444, + "step": 53932 + }, + { + "epoch": 3.47, + "learning_rate": 8.448227742115547e-06, + "loss": 0.7662, + "step": 53933 + }, + { + "epoch": 3.47, + "learning_rate": 8.448013053092596e-06, + "loss": 0.7852, + "step": 53934 + }, + { + "epoch": 3.47, + "learning_rate": 8.447798364069647e-06, + "loss": 0.7814, + "step": 53935 + }, + { + "epoch": 3.47, + "learning_rate": 8.447583675046696e-06, + "loss": 0.8075, + "step": 53936 + }, + { + "epoch": 3.47, + "learning_rate": 8.447368986023745e-06, + "loss": 0.8093, + "step": 53937 + }, + { + "epoch": 3.47, + "learning_rate": 8.447154297000795e-06, + "loss": 0.8203, + "step": 53938 + }, + { + "epoch": 3.47, + "learning_rate": 8.446939607977844e-06, + "loss": 0.7721, + "step": 53939 + }, + { + "epoch": 3.47, + "learning_rate": 8.446724918954895e-06, + "loss": 0.8643, + "step": 53940 + }, + { + "epoch": 3.47, + "learning_rate": 8.446510229931944e-06, + "loss": 0.8765, + "step": 53941 + }, + { + "epoch": 3.47, + "learning_rate": 8.446295540908995e-06, + "loss": 0.8191, + "step": 53942 + }, + { + "epoch": 3.47, + "learning_rate": 8.446080851886043e-06, + "loss": 0.8455, + "step": 53943 + }, + { + "epoch": 3.47, + "learning_rate": 8.445866162863094e-06, + "loss": 0.8617, + "step": 53944 + }, + { + "epoch": 3.47, + "learning_rate": 8.445651473840143e-06, + "loss": 0.7899, + "step": 53945 + }, + { + "epoch": 3.47, + "learning_rate": 8.445436784817192e-06, + "loss": 0.7945, + "step": 53946 + }, + { + "epoch": 3.47, + "learning_rate": 8.445222095794243e-06, + "loss": 0.8952, + "step": 53947 + }, + { + "epoch": 3.47, + "learning_rate": 8.445007406771293e-06, + "loss": 0.8374, + "step": 53948 + }, + { + "epoch": 3.47, + "learning_rate": 8.444792717748342e-06, + "loss": 0.8506, + "step": 53949 + }, + { + "epoch": 3.47, + "learning_rate": 8.444578028725391e-06, + "loss": 0.9121, + "step": 53950 + }, + { + "epoch": 3.47, + "learning_rate": 8.444363339702442e-06, + "loss": 0.9225, + "step": 53951 + }, + { + "epoch": 3.47, + "learning_rate": 8.444148650679491e-06, + "loss": 0.8002, + "step": 53952 + }, + { + "epoch": 3.47, + "learning_rate": 8.443933961656542e-06, + "loss": 0.862, + "step": 53953 + }, + { + "epoch": 3.47, + "learning_rate": 8.443719272633592e-06, + "loss": 0.8129, + "step": 53954 + }, + { + "epoch": 3.47, + "learning_rate": 8.443504583610641e-06, + "loss": 0.8629, + "step": 53955 + }, + { + "epoch": 3.47, + "learning_rate": 8.44328989458769e-06, + "loss": 0.8536, + "step": 53956 + }, + { + "epoch": 3.47, + "learning_rate": 8.443075205564741e-06, + "loss": 0.8239, + "step": 53957 + }, + { + "epoch": 3.47, + "learning_rate": 8.44286051654179e-06, + "loss": 0.8888, + "step": 53958 + }, + { + "epoch": 3.47, + "learning_rate": 8.442645827518839e-06, + "loss": 0.8392, + "step": 53959 + }, + { + "epoch": 3.47, + "learning_rate": 8.44243113849589e-06, + "loss": 0.8808, + "step": 53960 + }, + { + "epoch": 3.47, + "learning_rate": 8.442216449472938e-06, + "loss": 0.8233, + "step": 53961 + }, + { + "epoch": 3.47, + "learning_rate": 8.442001760449989e-06, + "loss": 0.8663, + "step": 53962 + }, + { + "epoch": 3.47, + "learning_rate": 8.44178707142704e-06, + "loss": 0.7783, + "step": 53963 + }, + { + "epoch": 3.47, + "learning_rate": 8.441572382404089e-06, + "loss": 0.8253, + "step": 53964 + }, + { + "epoch": 3.47, + "learning_rate": 8.441357693381138e-06, + "loss": 0.9428, + "step": 53965 + }, + { + "epoch": 3.47, + "learning_rate": 8.441143004358188e-06, + "loss": 0.8324, + "step": 53966 + }, + { + "epoch": 3.47, + "learning_rate": 8.440928315335237e-06, + "loss": 0.8633, + "step": 53967 + }, + { + "epoch": 3.47, + "learning_rate": 8.440713626312286e-06, + "loss": 0.8313, + "step": 53968 + }, + { + "epoch": 3.47, + "learning_rate": 8.440498937289337e-06, + "loss": 0.8173, + "step": 53969 + }, + { + "epoch": 3.47, + "learning_rate": 8.440284248266388e-06, + "loss": 0.8595, + "step": 53970 + }, + { + "epoch": 3.47, + "learning_rate": 8.440069559243437e-06, + "loss": 0.8215, + "step": 53971 + }, + { + "epoch": 3.47, + "learning_rate": 8.439854870220485e-06, + "loss": 0.8463, + "step": 53972 + }, + { + "epoch": 3.47, + "learning_rate": 8.439640181197536e-06, + "loss": 0.807, + "step": 53973 + }, + { + "epoch": 3.47, + "learning_rate": 8.439425492174585e-06, + "loss": 0.7997, + "step": 53974 + }, + { + "epoch": 3.47, + "learning_rate": 8.439210803151636e-06, + "loss": 0.8092, + "step": 53975 + }, + { + "epoch": 3.47, + "learning_rate": 8.438996114128686e-06, + "loss": 0.7613, + "step": 53976 + }, + { + "epoch": 3.47, + "learning_rate": 8.438781425105735e-06, + "loss": 0.851, + "step": 53977 + }, + { + "epoch": 3.47, + "learning_rate": 8.438566736082784e-06, + "loss": 0.7773, + "step": 53978 + }, + { + "epoch": 3.47, + "learning_rate": 8.438352047059835e-06, + "loss": 0.9021, + "step": 53979 + }, + { + "epoch": 3.47, + "learning_rate": 8.438137358036884e-06, + "loss": 0.8325, + "step": 53980 + }, + { + "epoch": 3.47, + "learning_rate": 8.437922669013933e-06, + "loss": 0.8567, + "step": 53981 + }, + { + "epoch": 3.47, + "learning_rate": 8.437707979990984e-06, + "loss": 0.8965, + "step": 53982 + }, + { + "epoch": 3.47, + "learning_rate": 8.437493290968034e-06, + "loss": 0.8677, + "step": 53983 + }, + { + "epoch": 3.47, + "learning_rate": 8.437278601945083e-06, + "loss": 0.8337, + "step": 53984 + }, + { + "epoch": 3.47, + "learning_rate": 8.437063912922134e-06, + "loss": 0.7904, + "step": 53985 + }, + { + "epoch": 3.47, + "learning_rate": 8.436849223899183e-06, + "loss": 0.8273, + "step": 53986 + }, + { + "epoch": 3.47, + "learning_rate": 8.436634534876232e-06, + "loss": 0.8576, + "step": 53987 + }, + { + "epoch": 3.47, + "learning_rate": 8.436419845853282e-06, + "loss": 0.7894, + "step": 53988 + }, + { + "epoch": 3.47, + "learning_rate": 8.436205156830331e-06, + "loss": 0.8181, + "step": 53989 + }, + { + "epoch": 3.47, + "learning_rate": 8.43599046780738e-06, + "loss": 0.8539, + "step": 53990 + }, + { + "epoch": 3.47, + "learning_rate": 8.435775778784431e-06, + "loss": 0.8198, + "step": 53991 + }, + { + "epoch": 3.47, + "learning_rate": 8.435561089761482e-06, + "loss": 0.8691, + "step": 53992 + }, + { + "epoch": 3.47, + "learning_rate": 8.43534640073853e-06, + "loss": 0.7967, + "step": 53993 + }, + { + "epoch": 3.47, + "learning_rate": 8.435131711715581e-06, + "loss": 0.7647, + "step": 53994 + }, + { + "epoch": 3.47, + "learning_rate": 8.43491702269263e-06, + "loss": 0.8164, + "step": 53995 + }, + { + "epoch": 3.47, + "learning_rate": 8.43470233366968e-06, + "loss": 0.8415, + "step": 53996 + }, + { + "epoch": 3.47, + "learning_rate": 8.43448764464673e-06, + "loss": 0.8486, + "step": 53997 + }, + { + "epoch": 3.47, + "learning_rate": 8.43427295562378e-06, + "loss": 0.7782, + "step": 53998 + }, + { + "epoch": 3.47, + "learning_rate": 8.43405826660083e-06, + "loss": 0.8453, + "step": 53999 + }, + { + "epoch": 3.47, + "learning_rate": 8.433843577577879e-06, + "loss": 0.8238, + "step": 54000 + }, + { + "epoch": 3.47, + "eval_laion/cider": 0.5086972211589291, + "eval_laion/meteor": 0.39559805585827085, + "eval_laion/pred_num_words": 37.46875, + "eval_laion/sacrebleu": 16.021026587935047, + "eval_laion/true_num_words": 41.380859375, + "eval_loss": 1.051003336906433, + "eval_runtime": 104.0477, + "eval_samples_per_second": 4.796, + "eval_steps_per_second": 0.308, + "step": 54000 + } + ], + "logging_steps": 1, + "max_steps": 93258, + "num_input_tokens_seen": 0, + "num_train_epochs": 6, + "save_steps": 1000, + "total_flos": 4.9867556382572544e+20, + "train_batch_size": 8, + "trial_name": null, + "trial_params": null +}