{ "best_metric": null, "best_model_checkpoint": null, "epoch": 1.0, "eval_steps": 500, "global_step": 9375, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 3.546099290780142e-08, "loss": 2.3158, "step": 1 }, { "epoch": 0.0, "learning_rate": 7.092198581560284e-08, "loss": 2.2055, "step": 2 }, { "epoch": 0.0, "learning_rate": 1.0638297872340426e-07, "loss": 2.3871, "step": 3 }, { "epoch": 0.0, "learning_rate": 1.4184397163120568e-07, "loss": 2.6017, "step": 4 }, { "epoch": 0.0, "learning_rate": 1.7730496453900713e-07, "loss": 2.2229, "step": 5 }, { "epoch": 0.0, "learning_rate": 2.1276595744680852e-07, "loss": 2.3659, "step": 6 }, { "epoch": 0.0, "learning_rate": 2.4822695035460997e-07, "loss": 2.2991, "step": 7 }, { "epoch": 0.0, "learning_rate": 2.8368794326241136e-07, "loss": 2.3525, "step": 8 }, { "epoch": 0.0, "learning_rate": 3.1914893617021275e-07, "loss": 2.4287, "step": 9 }, { "epoch": 0.0, "learning_rate": 3.5460992907801425e-07, "loss": 2.4118, "step": 10 }, { "epoch": 0.0, "learning_rate": 3.9007092198581565e-07, "loss": 2.6199, "step": 11 }, { "epoch": 0.0, "learning_rate": 4.2553191489361704e-07, "loss": 2.8125, "step": 12 }, { "epoch": 0.0, "learning_rate": 4.6099290780141843e-07, "loss": 2.3915, "step": 13 }, { "epoch": 0.0, "learning_rate": 4.964539007092199e-07, "loss": 2.3286, "step": 14 }, { "epoch": 0.0, "learning_rate": 5.319148936170213e-07, "loss": 2.4814, "step": 15 }, { "epoch": 0.0, "learning_rate": 5.673758865248227e-07, "loss": 2.2738, "step": 16 }, { "epoch": 0.0, "learning_rate": 6.028368794326241e-07, "loss": 2.1354, "step": 17 }, { "epoch": 0.0, "learning_rate": 6.382978723404255e-07, "loss": 2.3302, "step": 18 }, { "epoch": 0.0, "learning_rate": 6.73758865248227e-07, "loss": 2.3007, "step": 19 }, { "epoch": 0.0, "learning_rate": 7.092198581560285e-07, "loss": 2.5913, "step": 20 }, { "epoch": 0.0, "learning_rate": 7.446808510638298e-07, "loss": 2.3403, "step": 21 }, { "epoch": 0.0, "learning_rate": 7.801418439716313e-07, "loss": 2.0736, "step": 22 }, { "epoch": 0.0, "learning_rate": 8.156028368794328e-07, "loss": 1.978, "step": 23 }, { "epoch": 0.0, "learning_rate": 8.510638297872341e-07, "loss": 2.2008, "step": 24 }, { "epoch": 0.0, "learning_rate": 8.865248226950356e-07, "loss": 2.5363, "step": 25 }, { "epoch": 0.0, "learning_rate": 9.219858156028369e-07, "loss": 2.213, "step": 26 }, { "epoch": 0.0, "learning_rate": 9.574468085106384e-07, "loss": 2.109, "step": 27 }, { "epoch": 0.0, "learning_rate": 9.929078014184399e-07, "loss": 2.5121, "step": 28 }, { "epoch": 0.0, "learning_rate": 1.0283687943262412e-06, "loss": 2.0351, "step": 29 }, { "epoch": 0.0, "learning_rate": 1.0638297872340427e-06, "loss": 2.0635, "step": 30 }, { "epoch": 0.0, "learning_rate": 1.0992907801418442e-06, "loss": 2.2137, "step": 31 }, { "epoch": 0.0, "learning_rate": 1.1347517730496454e-06, "loss": 1.9304, "step": 32 }, { "epoch": 0.0, "learning_rate": 1.170212765957447e-06, "loss": 1.9055, "step": 33 }, { "epoch": 0.0, "learning_rate": 1.2056737588652482e-06, "loss": 2.1687, "step": 34 }, { "epoch": 0.0, "learning_rate": 1.2411347517730497e-06, "loss": 2.2072, "step": 35 }, { "epoch": 0.0, "learning_rate": 1.276595744680851e-06, "loss": 2.0783, "step": 36 }, { "epoch": 0.0, "learning_rate": 1.3120567375886525e-06, "loss": 2.0374, "step": 37 }, { "epoch": 0.0, "learning_rate": 1.347517730496454e-06, "loss": 2.0336, "step": 38 }, { "epoch": 0.0, "learning_rate": 1.3829787234042555e-06, "loss": 2.0079, "step": 39 }, { "epoch": 0.0, "learning_rate": 1.418439716312057e-06, "loss": 1.9667, "step": 40 }, { "epoch": 0.0, "learning_rate": 1.453900709219858e-06, "loss": 1.8078, "step": 41 }, { "epoch": 0.0, "learning_rate": 1.4893617021276596e-06, "loss": 1.6119, "step": 42 }, { "epoch": 0.0, "learning_rate": 1.524822695035461e-06, "loss": 1.9438, "step": 43 }, { "epoch": 0.0, "learning_rate": 1.5602836879432626e-06, "loss": 1.5472, "step": 44 }, { "epoch": 0.0, "learning_rate": 1.595744680851064e-06, "loss": 2.0744, "step": 45 }, { "epoch": 0.0, "learning_rate": 1.6312056737588656e-06, "loss": 1.7662, "step": 46 }, { "epoch": 0.01, "learning_rate": 1.6666666666666667e-06, "loss": 1.9349, "step": 47 }, { "epoch": 0.01, "learning_rate": 1.7021276595744682e-06, "loss": 1.9197, "step": 48 }, { "epoch": 0.01, "learning_rate": 1.7375886524822697e-06, "loss": 2.1485, "step": 49 }, { "epoch": 0.01, "learning_rate": 1.7730496453900712e-06, "loss": 2.0791, "step": 50 }, { "epoch": 0.01, "learning_rate": 1.8085106382978727e-06, "loss": 1.6205, "step": 51 }, { "epoch": 0.01, "learning_rate": 1.8439716312056737e-06, "loss": 1.7532, "step": 52 }, { "epoch": 0.01, "learning_rate": 1.8794326241134752e-06, "loss": 1.8713, "step": 53 }, { "epoch": 0.01, "learning_rate": 1.9148936170212767e-06, "loss": 2.0214, "step": 54 }, { "epoch": 0.01, "learning_rate": 1.9503546099290782e-06, "loss": 1.4967, "step": 55 }, { "epoch": 0.01, "learning_rate": 1.9858156028368797e-06, "loss": 1.8247, "step": 56 }, { "epoch": 0.01, "learning_rate": 2.021276595744681e-06, "loss": 1.5498, "step": 57 }, { "epoch": 0.01, "learning_rate": 2.0567375886524823e-06, "loss": 1.9241, "step": 58 }, { "epoch": 0.01, "learning_rate": 2.092198581560284e-06, "loss": 1.7679, "step": 59 }, { "epoch": 0.01, "learning_rate": 2.1276595744680853e-06, "loss": 1.7669, "step": 60 }, { "epoch": 0.01, "learning_rate": 2.163120567375887e-06, "loss": 1.7345, "step": 61 }, { "epoch": 0.01, "learning_rate": 2.1985815602836883e-06, "loss": 1.7026, "step": 62 }, { "epoch": 0.01, "learning_rate": 2.2340425531914894e-06, "loss": 1.7274, "step": 63 }, { "epoch": 0.01, "learning_rate": 2.269503546099291e-06, "loss": 1.8726, "step": 64 }, { "epoch": 0.01, "learning_rate": 2.3049645390070924e-06, "loss": 1.9631, "step": 65 }, { "epoch": 0.01, "learning_rate": 2.340425531914894e-06, "loss": 1.5626, "step": 66 }, { "epoch": 0.01, "learning_rate": 2.3758865248226954e-06, "loss": 1.5681, "step": 67 }, { "epoch": 0.01, "learning_rate": 2.4113475177304965e-06, "loss": 1.9644, "step": 68 }, { "epoch": 0.01, "learning_rate": 2.446808510638298e-06, "loss": 1.6553, "step": 69 }, { "epoch": 0.01, "learning_rate": 2.4822695035460995e-06, "loss": 1.9931, "step": 70 }, { "epoch": 0.01, "learning_rate": 2.5177304964539005e-06, "loss": 1.7749, "step": 71 }, { "epoch": 0.01, "learning_rate": 2.553191489361702e-06, "loss": 1.7932, "step": 72 }, { "epoch": 0.01, "learning_rate": 2.5886524822695035e-06, "loss": 1.9286, "step": 73 }, { "epoch": 0.01, "learning_rate": 2.624113475177305e-06, "loss": 1.8326, "step": 74 }, { "epoch": 0.01, "learning_rate": 2.6595744680851065e-06, "loss": 1.9796, "step": 75 }, { "epoch": 0.01, "learning_rate": 2.695035460992908e-06, "loss": 1.7205, "step": 76 }, { "epoch": 0.01, "learning_rate": 2.7304964539007095e-06, "loss": 1.5599, "step": 77 }, { "epoch": 0.01, "learning_rate": 2.765957446808511e-06, "loss": 1.698, "step": 78 }, { "epoch": 0.01, "learning_rate": 2.8014184397163125e-06, "loss": 1.8654, "step": 79 }, { "epoch": 0.01, "learning_rate": 2.836879432624114e-06, "loss": 1.8744, "step": 80 }, { "epoch": 0.01, "learning_rate": 2.8723404255319155e-06, "loss": 1.5733, "step": 81 }, { "epoch": 0.01, "learning_rate": 2.907801418439716e-06, "loss": 2.1111, "step": 82 }, { "epoch": 0.01, "learning_rate": 2.9432624113475177e-06, "loss": 1.6334, "step": 83 }, { "epoch": 0.01, "learning_rate": 2.978723404255319e-06, "loss": 1.9889, "step": 84 }, { "epoch": 0.01, "learning_rate": 3.0141843971631207e-06, "loss": 1.5935, "step": 85 }, { "epoch": 0.01, "learning_rate": 3.049645390070922e-06, "loss": 1.8515, "step": 86 }, { "epoch": 0.01, "learning_rate": 3.0851063829787237e-06, "loss": 1.676, "step": 87 }, { "epoch": 0.01, "learning_rate": 3.120567375886525e-06, "loss": 1.7081, "step": 88 }, { "epoch": 0.01, "learning_rate": 3.1560283687943267e-06, "loss": 1.6722, "step": 89 }, { "epoch": 0.01, "learning_rate": 3.191489361702128e-06, "loss": 1.7497, "step": 90 }, { "epoch": 0.01, "learning_rate": 3.2269503546099297e-06, "loss": 1.5862, "step": 91 }, { "epoch": 0.01, "learning_rate": 3.262411347517731e-06, "loss": 1.734, "step": 92 }, { "epoch": 0.01, "learning_rate": 3.297872340425532e-06, "loss": 1.7384, "step": 93 }, { "epoch": 0.01, "learning_rate": 3.3333333333333333e-06, "loss": 1.8243, "step": 94 }, { "epoch": 0.01, "learning_rate": 3.368794326241135e-06, "loss": 1.9562, "step": 95 }, { "epoch": 0.01, "learning_rate": 3.4042553191489363e-06, "loss": 1.6348, "step": 96 }, { "epoch": 0.01, "learning_rate": 3.439716312056738e-06, "loss": 1.9281, "step": 97 }, { "epoch": 0.01, "learning_rate": 3.4751773049645393e-06, "loss": 1.5509, "step": 98 }, { "epoch": 0.01, "learning_rate": 3.510638297872341e-06, "loss": 1.8801, "step": 99 }, { "epoch": 0.01, "learning_rate": 3.5460992907801423e-06, "loss": 1.6359, "step": 100 }, { "epoch": 0.01, "learning_rate": 3.581560283687944e-06, "loss": 1.6596, "step": 101 }, { "epoch": 0.01, "learning_rate": 3.6170212765957453e-06, "loss": 1.6551, "step": 102 }, { "epoch": 0.01, "learning_rate": 3.652482269503547e-06, "loss": 1.8687, "step": 103 }, { "epoch": 0.01, "learning_rate": 3.6879432624113475e-06, "loss": 1.8392, "step": 104 }, { "epoch": 0.01, "learning_rate": 3.723404255319149e-06, "loss": 1.5599, "step": 105 }, { "epoch": 0.01, "learning_rate": 3.7588652482269505e-06, "loss": 1.4459, "step": 106 }, { "epoch": 0.01, "learning_rate": 3.794326241134752e-06, "loss": 1.6055, "step": 107 }, { "epoch": 0.01, "learning_rate": 3.8297872340425535e-06, "loss": 1.6792, "step": 108 }, { "epoch": 0.01, "learning_rate": 3.865248226950355e-06, "loss": 2.04, "step": 109 }, { "epoch": 0.01, "learning_rate": 3.9007092198581565e-06, "loss": 1.8408, "step": 110 }, { "epoch": 0.01, "learning_rate": 3.936170212765958e-06, "loss": 1.718, "step": 111 }, { "epoch": 0.01, "learning_rate": 3.9716312056737595e-06, "loss": 1.5136, "step": 112 }, { "epoch": 0.01, "learning_rate": 4.007092198581561e-06, "loss": 1.5976, "step": 113 }, { "epoch": 0.01, "learning_rate": 4.042553191489362e-06, "loss": 1.7739, "step": 114 }, { "epoch": 0.01, "learning_rate": 4.078014184397163e-06, "loss": 1.3512, "step": 115 }, { "epoch": 0.01, "learning_rate": 4.113475177304965e-06, "loss": 1.7644, "step": 116 }, { "epoch": 0.01, "learning_rate": 4.148936170212766e-06, "loss": 1.5732, "step": 117 }, { "epoch": 0.01, "learning_rate": 4.184397163120568e-06, "loss": 1.7343, "step": 118 }, { "epoch": 0.01, "learning_rate": 4.219858156028369e-06, "loss": 1.6525, "step": 119 }, { "epoch": 0.01, "learning_rate": 4.255319148936171e-06, "loss": 1.6848, "step": 120 }, { "epoch": 0.01, "learning_rate": 4.290780141843972e-06, "loss": 1.698, "step": 121 }, { "epoch": 0.01, "learning_rate": 4.326241134751774e-06, "loss": 1.7206, "step": 122 }, { "epoch": 0.01, "learning_rate": 4.361702127659575e-06, "loss": 1.7891, "step": 123 }, { "epoch": 0.01, "learning_rate": 4.397163120567377e-06, "loss": 1.6416, "step": 124 }, { "epoch": 0.01, "learning_rate": 4.432624113475177e-06, "loss": 1.7396, "step": 125 }, { "epoch": 0.01, "learning_rate": 4.468085106382979e-06, "loss": 1.7948, "step": 126 }, { "epoch": 0.01, "learning_rate": 4.50354609929078e-06, "loss": 1.6542, "step": 127 }, { "epoch": 0.01, "learning_rate": 4.539007092198582e-06, "loss": 1.6514, "step": 128 }, { "epoch": 0.01, "learning_rate": 4.574468085106383e-06, "loss": 1.6807, "step": 129 }, { "epoch": 0.01, "learning_rate": 4.609929078014185e-06, "loss": 1.5656, "step": 130 }, { "epoch": 0.01, "learning_rate": 4.645390070921986e-06, "loss": 1.4168, "step": 131 }, { "epoch": 0.01, "learning_rate": 4.680851063829788e-06, "loss": 1.7839, "step": 132 }, { "epoch": 0.01, "learning_rate": 4.716312056737589e-06, "loss": 1.5946, "step": 133 }, { "epoch": 0.01, "learning_rate": 4.751773049645391e-06, "loss": 1.5225, "step": 134 }, { "epoch": 0.01, "learning_rate": 4.787234042553192e-06, "loss": 2.0029, "step": 135 }, { "epoch": 0.01, "learning_rate": 4.822695035460993e-06, "loss": 1.5007, "step": 136 }, { "epoch": 0.01, "learning_rate": 4.858156028368794e-06, "loss": 1.5473, "step": 137 }, { "epoch": 0.01, "learning_rate": 4.893617021276596e-06, "loss": 1.7001, "step": 138 }, { "epoch": 0.01, "learning_rate": 4.929078014184397e-06, "loss": 1.7766, "step": 139 }, { "epoch": 0.01, "learning_rate": 4.964539007092199e-06, "loss": 1.7107, "step": 140 }, { "epoch": 0.02, "learning_rate": 5e-06, "loss": 1.6595, "step": 141 }, { "epoch": 0.02, "learning_rate": 5.035460992907801e-06, "loss": 1.5207, "step": 142 }, { "epoch": 0.02, "learning_rate": 5.070921985815603e-06, "loss": 1.7887, "step": 143 }, { "epoch": 0.02, "learning_rate": 5.106382978723404e-06, "loss": 1.459, "step": 144 }, { "epoch": 0.02, "learning_rate": 5.141843971631206e-06, "loss": 1.6618, "step": 145 }, { "epoch": 0.02, "learning_rate": 5.177304964539007e-06, "loss": 1.928, "step": 146 }, { "epoch": 0.02, "learning_rate": 5.212765957446809e-06, "loss": 1.554, "step": 147 }, { "epoch": 0.02, "learning_rate": 5.24822695035461e-06, "loss": 1.6945, "step": 148 }, { "epoch": 0.02, "learning_rate": 5.283687943262412e-06, "loss": 1.6265, "step": 149 }, { "epoch": 0.02, "learning_rate": 5.319148936170213e-06, "loss": 1.5792, "step": 150 }, { "epoch": 0.02, "learning_rate": 5.354609929078015e-06, "loss": 1.445, "step": 151 }, { "epoch": 0.02, "learning_rate": 5.390070921985816e-06, "loss": 1.5104, "step": 152 }, { "epoch": 0.02, "learning_rate": 5.425531914893617e-06, "loss": 1.7946, "step": 153 }, { "epoch": 0.02, "learning_rate": 5.460992907801419e-06, "loss": 1.4712, "step": 154 }, { "epoch": 0.02, "learning_rate": 5.49645390070922e-06, "loss": 1.6309, "step": 155 }, { "epoch": 0.02, "learning_rate": 5.531914893617022e-06, "loss": 1.6381, "step": 156 }, { "epoch": 0.02, "learning_rate": 5.567375886524823e-06, "loss": 1.6228, "step": 157 }, { "epoch": 0.02, "learning_rate": 5.602836879432625e-06, "loss": 1.7836, "step": 158 }, { "epoch": 0.02, "learning_rate": 5.638297872340426e-06, "loss": 1.5796, "step": 159 }, { "epoch": 0.02, "learning_rate": 5.673758865248228e-06, "loss": 1.6917, "step": 160 }, { "epoch": 0.02, "learning_rate": 5.709219858156029e-06, "loss": 1.5937, "step": 161 }, { "epoch": 0.02, "learning_rate": 5.744680851063831e-06, "loss": 2.0016, "step": 162 }, { "epoch": 0.02, "learning_rate": 5.780141843971632e-06, "loss": 1.8732, "step": 163 }, { "epoch": 0.02, "learning_rate": 5.815602836879432e-06, "loss": 1.5748, "step": 164 }, { "epoch": 0.02, "learning_rate": 5.851063829787235e-06, "loss": 1.5238, "step": 165 }, { "epoch": 0.02, "learning_rate": 5.886524822695035e-06, "loss": 2.1618, "step": 166 }, { "epoch": 0.02, "learning_rate": 5.921985815602838e-06, "loss": 1.6294, "step": 167 }, { "epoch": 0.02, "learning_rate": 5.957446808510638e-06, "loss": 1.5611, "step": 168 }, { "epoch": 0.02, "learning_rate": 5.992907801418441e-06, "loss": 1.8085, "step": 169 }, { "epoch": 0.02, "learning_rate": 6.028368794326241e-06, "loss": 1.6225, "step": 170 }, { "epoch": 0.02, "learning_rate": 6.063829787234044e-06, "loss": 1.586, "step": 171 }, { "epoch": 0.02, "learning_rate": 6.099290780141844e-06, "loss": 1.7883, "step": 172 }, { "epoch": 0.02, "learning_rate": 6.134751773049647e-06, "loss": 1.7999, "step": 173 }, { "epoch": 0.02, "learning_rate": 6.170212765957447e-06, "loss": 1.5684, "step": 174 }, { "epoch": 0.02, "learning_rate": 6.205673758865248e-06, "loss": 1.7389, "step": 175 }, { "epoch": 0.02, "learning_rate": 6.24113475177305e-06, "loss": 1.5522, "step": 176 }, { "epoch": 0.02, "learning_rate": 6.276595744680851e-06, "loss": 1.806, "step": 177 }, { "epoch": 0.02, "learning_rate": 6.312056737588653e-06, "loss": 1.706, "step": 178 }, { "epoch": 0.02, "learning_rate": 6.347517730496454e-06, "loss": 1.6249, "step": 179 }, { "epoch": 0.02, "learning_rate": 6.382978723404256e-06, "loss": 1.8193, "step": 180 }, { "epoch": 0.02, "learning_rate": 6.418439716312057e-06, "loss": 1.579, "step": 181 }, { "epoch": 0.02, "learning_rate": 6.453900709219859e-06, "loss": 1.7793, "step": 182 }, { "epoch": 0.02, "learning_rate": 6.48936170212766e-06, "loss": 1.8394, "step": 183 }, { "epoch": 0.02, "learning_rate": 6.524822695035462e-06, "loss": 1.536, "step": 184 }, { "epoch": 0.02, "learning_rate": 6.560283687943263e-06, "loss": 1.5064, "step": 185 }, { "epoch": 0.02, "learning_rate": 6.595744680851064e-06, "loss": 1.7425, "step": 186 }, { "epoch": 0.02, "learning_rate": 6.631205673758866e-06, "loss": 1.4376, "step": 187 }, { "epoch": 0.02, "learning_rate": 6.666666666666667e-06, "loss": 1.5943, "step": 188 }, { "epoch": 0.02, "learning_rate": 6.702127659574469e-06, "loss": 1.5448, "step": 189 }, { "epoch": 0.02, "learning_rate": 6.73758865248227e-06, "loss": 2.1547, "step": 190 }, { "epoch": 0.02, "learning_rate": 6.773049645390072e-06, "loss": 1.6576, "step": 191 }, { "epoch": 0.02, "learning_rate": 6.808510638297873e-06, "loss": 1.5362, "step": 192 }, { "epoch": 0.02, "learning_rate": 6.843971631205675e-06, "loss": 1.5358, "step": 193 }, { "epoch": 0.02, "learning_rate": 6.879432624113476e-06, "loss": 1.5615, "step": 194 }, { "epoch": 0.02, "learning_rate": 6.914893617021278e-06, "loss": 1.5747, "step": 195 }, { "epoch": 0.02, "learning_rate": 6.950354609929079e-06, "loss": 1.8223, "step": 196 }, { "epoch": 0.02, "learning_rate": 6.985815602836879e-06, "loss": 1.6142, "step": 197 }, { "epoch": 0.02, "learning_rate": 7.021276595744682e-06, "loss": 1.5285, "step": 198 }, { "epoch": 0.02, "learning_rate": 7.056737588652482e-06, "loss": 1.3855, "step": 199 }, { "epoch": 0.02, "learning_rate": 7.092198581560285e-06, "loss": 1.5881, "step": 200 }, { "epoch": 0.02, "learning_rate": 7.127659574468085e-06, "loss": 1.9831, "step": 201 }, { "epoch": 0.02, "learning_rate": 7.163120567375888e-06, "loss": 1.8067, "step": 202 }, { "epoch": 0.02, "learning_rate": 7.198581560283688e-06, "loss": 1.4027, "step": 203 }, { "epoch": 0.02, "learning_rate": 7.234042553191491e-06, "loss": 1.6902, "step": 204 }, { "epoch": 0.02, "learning_rate": 7.269503546099291e-06, "loss": 1.4386, "step": 205 }, { "epoch": 0.02, "learning_rate": 7.304964539007094e-06, "loss": 1.6657, "step": 206 }, { "epoch": 0.02, "learning_rate": 7.340425531914894e-06, "loss": 1.6231, "step": 207 }, { "epoch": 0.02, "learning_rate": 7.375886524822695e-06, "loss": 1.5444, "step": 208 }, { "epoch": 0.02, "learning_rate": 7.411347517730497e-06, "loss": 1.924, "step": 209 }, { "epoch": 0.02, "learning_rate": 7.446808510638298e-06, "loss": 1.4463, "step": 210 }, { "epoch": 0.02, "learning_rate": 7.4822695035461e-06, "loss": 1.8598, "step": 211 }, { "epoch": 0.02, "learning_rate": 7.517730496453901e-06, "loss": 1.7418, "step": 212 }, { "epoch": 0.02, "learning_rate": 7.553191489361703e-06, "loss": 1.5673, "step": 213 }, { "epoch": 0.02, "learning_rate": 7.588652482269504e-06, "loss": 1.6071, "step": 214 }, { "epoch": 0.02, "learning_rate": 7.624113475177306e-06, "loss": 1.5775, "step": 215 }, { "epoch": 0.02, "learning_rate": 7.659574468085107e-06, "loss": 1.6697, "step": 216 }, { "epoch": 0.02, "learning_rate": 7.695035460992908e-06, "loss": 1.4901, "step": 217 }, { "epoch": 0.02, "learning_rate": 7.73049645390071e-06, "loss": 1.7522, "step": 218 }, { "epoch": 0.02, "learning_rate": 7.765957446808511e-06, "loss": 1.9224, "step": 219 }, { "epoch": 0.02, "learning_rate": 7.801418439716313e-06, "loss": 1.6055, "step": 220 }, { "epoch": 0.02, "learning_rate": 7.836879432624114e-06, "loss": 1.9617, "step": 221 }, { "epoch": 0.02, "learning_rate": 7.872340425531916e-06, "loss": 1.6708, "step": 222 }, { "epoch": 0.02, "learning_rate": 7.907801418439717e-06, "loss": 1.6602, "step": 223 }, { "epoch": 0.02, "learning_rate": 7.943262411347519e-06, "loss": 1.5906, "step": 224 }, { "epoch": 0.02, "learning_rate": 7.97872340425532e-06, "loss": 1.7244, "step": 225 }, { "epoch": 0.02, "learning_rate": 8.014184397163122e-06, "loss": 1.7634, "step": 226 }, { "epoch": 0.02, "learning_rate": 8.049645390070923e-06, "loss": 1.7657, "step": 227 }, { "epoch": 0.02, "learning_rate": 8.085106382978723e-06, "loss": 1.573, "step": 228 }, { "epoch": 0.02, "learning_rate": 8.120567375886525e-06, "loss": 1.5305, "step": 229 }, { "epoch": 0.02, "learning_rate": 8.156028368794326e-06, "loss": 1.587, "step": 230 }, { "epoch": 0.02, "learning_rate": 8.191489361702128e-06, "loss": 1.8304, "step": 231 }, { "epoch": 0.02, "learning_rate": 8.22695035460993e-06, "loss": 1.6283, "step": 232 }, { "epoch": 0.02, "learning_rate": 8.26241134751773e-06, "loss": 1.6331, "step": 233 }, { "epoch": 0.02, "learning_rate": 8.297872340425532e-06, "loss": 1.6509, "step": 234 }, { "epoch": 0.03, "learning_rate": 8.333333333333334e-06, "loss": 1.8751, "step": 235 }, { "epoch": 0.03, "learning_rate": 8.368794326241135e-06, "loss": 1.3581, "step": 236 }, { "epoch": 0.03, "learning_rate": 8.404255319148937e-06, "loss": 1.7169, "step": 237 }, { "epoch": 0.03, "learning_rate": 8.439716312056738e-06, "loss": 1.7349, "step": 238 }, { "epoch": 0.03, "learning_rate": 8.47517730496454e-06, "loss": 1.5962, "step": 239 }, { "epoch": 0.03, "learning_rate": 8.510638297872341e-06, "loss": 1.8424, "step": 240 }, { "epoch": 0.03, "learning_rate": 8.546099290780143e-06, "loss": 1.7879, "step": 241 }, { "epoch": 0.03, "learning_rate": 8.581560283687944e-06, "loss": 1.694, "step": 242 }, { "epoch": 0.03, "learning_rate": 8.617021276595746e-06, "loss": 1.6502, "step": 243 }, { "epoch": 0.03, "learning_rate": 8.652482269503547e-06, "loss": 1.5875, "step": 244 }, { "epoch": 0.03, "learning_rate": 8.687943262411349e-06, "loss": 1.7397, "step": 245 }, { "epoch": 0.03, "learning_rate": 8.72340425531915e-06, "loss": 1.6929, "step": 246 }, { "epoch": 0.03, "learning_rate": 8.758865248226952e-06, "loss": 1.4377, "step": 247 }, { "epoch": 0.03, "learning_rate": 8.794326241134753e-06, "loss": 1.5138, "step": 248 }, { "epoch": 0.03, "learning_rate": 8.829787234042555e-06, "loss": 1.6943, "step": 249 }, { "epoch": 0.03, "learning_rate": 8.865248226950355e-06, "loss": 1.8504, "step": 250 }, { "epoch": 0.03, "learning_rate": 8.900709219858156e-06, "loss": 1.3152, "step": 251 }, { "epoch": 0.03, "learning_rate": 8.936170212765958e-06, "loss": 1.5302, "step": 252 }, { "epoch": 0.03, "learning_rate": 8.971631205673759e-06, "loss": 1.6475, "step": 253 }, { "epoch": 0.03, "learning_rate": 9.00709219858156e-06, "loss": 1.6883, "step": 254 }, { "epoch": 0.03, "learning_rate": 9.042553191489362e-06, "loss": 1.8447, "step": 255 }, { "epoch": 0.03, "learning_rate": 9.078014184397164e-06, "loss": 1.4614, "step": 256 }, { "epoch": 0.03, "learning_rate": 9.113475177304965e-06, "loss": 1.9387, "step": 257 }, { "epoch": 0.03, "learning_rate": 9.148936170212767e-06, "loss": 1.5442, "step": 258 }, { "epoch": 0.03, "learning_rate": 9.184397163120568e-06, "loss": 1.629, "step": 259 }, { "epoch": 0.03, "learning_rate": 9.21985815602837e-06, "loss": 1.805, "step": 260 }, { "epoch": 0.03, "learning_rate": 9.255319148936171e-06, "loss": 1.2079, "step": 261 }, { "epoch": 0.03, "learning_rate": 9.290780141843973e-06, "loss": 1.4917, "step": 262 }, { "epoch": 0.03, "learning_rate": 9.326241134751774e-06, "loss": 1.4185, "step": 263 }, { "epoch": 0.03, "learning_rate": 9.361702127659576e-06, "loss": 1.6113, "step": 264 }, { "epoch": 0.03, "learning_rate": 9.397163120567377e-06, "loss": 1.6069, "step": 265 }, { "epoch": 0.03, "learning_rate": 9.432624113475179e-06, "loss": 1.6226, "step": 266 }, { "epoch": 0.03, "learning_rate": 9.46808510638298e-06, "loss": 1.5813, "step": 267 }, { "epoch": 0.03, "learning_rate": 9.503546099290782e-06, "loss": 1.7806, "step": 268 }, { "epoch": 0.03, "learning_rate": 9.539007092198583e-06, "loss": 1.4238, "step": 269 }, { "epoch": 0.03, "learning_rate": 9.574468085106385e-06, "loss": 1.4798, "step": 270 }, { "epoch": 0.03, "learning_rate": 9.609929078014186e-06, "loss": 1.7061, "step": 271 }, { "epoch": 0.03, "learning_rate": 9.645390070921986e-06, "loss": 1.7165, "step": 272 }, { "epoch": 0.03, "learning_rate": 9.680851063829787e-06, "loss": 1.614, "step": 273 }, { "epoch": 0.03, "learning_rate": 9.716312056737589e-06, "loss": 1.6066, "step": 274 }, { "epoch": 0.03, "learning_rate": 9.75177304964539e-06, "loss": 1.5454, "step": 275 }, { "epoch": 0.03, "learning_rate": 9.787234042553192e-06, "loss": 1.5636, "step": 276 }, { "epoch": 0.03, "learning_rate": 9.822695035460993e-06, "loss": 1.6946, "step": 277 }, { "epoch": 0.03, "learning_rate": 9.858156028368795e-06, "loss": 1.7989, "step": 278 }, { "epoch": 0.03, "learning_rate": 9.893617021276596e-06, "loss": 1.658, "step": 279 }, { "epoch": 0.03, "learning_rate": 9.929078014184398e-06, "loss": 1.656, "step": 280 }, { "epoch": 0.03, "learning_rate": 9.9645390070922e-06, "loss": 1.7294, "step": 281 }, { "epoch": 0.03, "learning_rate": 1e-05, "loss": 1.5139, "step": 282 }, { "epoch": 0.03, "learning_rate": 9.999999701581759e-06, "loss": 1.4776, "step": 283 }, { "epoch": 0.03, "learning_rate": 9.99999880632707e-06, "loss": 1.6574, "step": 284 }, { "epoch": 0.03, "learning_rate": 9.999997314236036e-06, "loss": 1.7381, "step": 285 }, { "epoch": 0.03, "learning_rate": 9.999995225308842e-06, "loss": 1.33, "step": 286 }, { "epoch": 0.03, "learning_rate": 9.999992539545732e-06, "loss": 1.5946, "step": 287 }, { "epoch": 0.03, "learning_rate": 9.999989256947029e-06, "loss": 1.6753, "step": 288 }, { "epoch": 0.03, "learning_rate": 9.999985377513126e-06, "loss": 1.6363, "step": 289 }, { "epoch": 0.03, "learning_rate": 9.999980901244483e-06, "loss": 1.7293, "step": 290 }, { "epoch": 0.03, "learning_rate": 9.999975828141635e-06, "loss": 1.57, "step": 291 }, { "epoch": 0.03, "learning_rate": 9.99997015820519e-06, "loss": 1.6875, "step": 292 }, { "epoch": 0.03, "learning_rate": 9.999963891435822e-06, "loss": 1.8595, "step": 293 }, { "epoch": 0.03, "learning_rate": 9.999957027834282e-06, "loss": 1.6232, "step": 294 }, { "epoch": 0.03, "learning_rate": 9.999949567401386e-06, "loss": 1.4337, "step": 295 }, { "epoch": 0.03, "learning_rate": 9.999941510138025e-06, "loss": 1.7758, "step": 296 }, { "epoch": 0.03, "learning_rate": 9.999932856045164e-06, "loss": 1.7183, "step": 297 }, { "epoch": 0.03, "learning_rate": 9.999923605123833e-06, "loss": 1.7933, "step": 298 }, { "epoch": 0.03, "learning_rate": 9.999913757375138e-06, "loss": 1.6982, "step": 299 }, { "epoch": 0.03, "learning_rate": 9.99990331280025e-06, "loss": 1.6447, "step": 300 }, { "epoch": 0.03, "learning_rate": 9.999892271400424e-06, "loss": 1.4807, "step": 301 }, { "epoch": 0.03, "learning_rate": 9.99988063317697e-06, "loss": 1.6015, "step": 302 }, { "epoch": 0.03, "learning_rate": 9.999868398131282e-06, "loss": 1.4899, "step": 303 }, { "epoch": 0.03, "learning_rate": 9.999855566264818e-06, "loss": 1.5795, "step": 304 }, { "epoch": 0.03, "learning_rate": 9.999842137579112e-06, "loss": 1.648, "step": 305 }, { "epoch": 0.03, "learning_rate": 9.999828112075764e-06, "loss": 1.365, "step": 306 }, { "epoch": 0.03, "learning_rate": 9.999813489756452e-06, "loss": 1.5681, "step": 307 }, { "epoch": 0.03, "learning_rate": 9.999798270622918e-06, "loss": 1.5068, "step": 308 }, { "epoch": 0.03, "learning_rate": 9.99978245467698e-06, "loss": 1.4256, "step": 309 }, { "epoch": 0.03, "learning_rate": 9.999766041920525e-06, "loss": 1.7858, "step": 310 }, { "epoch": 0.03, "learning_rate": 9.999749032355514e-06, "loss": 1.704, "step": 311 }, { "epoch": 0.03, "learning_rate": 9.999731425983975e-06, "loss": 1.7687, "step": 312 }, { "epoch": 0.03, "learning_rate": 9.999713222808014e-06, "loss": 1.7859, "step": 313 }, { "epoch": 0.03, "learning_rate": 9.999694422829798e-06, "loss": 1.5655, "step": 314 }, { "epoch": 0.03, "learning_rate": 9.999675026051576e-06, "loss": 1.6694, "step": 315 }, { "epoch": 0.03, "learning_rate": 9.99965503247566e-06, "loss": 1.6347, "step": 316 }, { "epoch": 0.03, "learning_rate": 9.999634442104438e-06, "loss": 1.5104, "step": 317 }, { "epoch": 0.03, "learning_rate": 9.999613254940368e-06, "loss": 1.6778, "step": 318 }, { "epoch": 0.03, "learning_rate": 9.999591470985979e-06, "loss": 1.2955, "step": 319 }, { "epoch": 0.03, "learning_rate": 9.99956909024387e-06, "loss": 1.5058, "step": 320 }, { "epoch": 0.03, "learning_rate": 9.999546112716715e-06, "loss": 1.6738, "step": 321 }, { "epoch": 0.03, "learning_rate": 9.999522538407253e-06, "loss": 1.4654, "step": 322 }, { "epoch": 0.03, "learning_rate": 9.999498367318303e-06, "loss": 1.4621, "step": 323 }, { "epoch": 0.03, "learning_rate": 9.999473599452746e-06, "loss": 1.7891, "step": 324 }, { "epoch": 0.03, "learning_rate": 9.999448234813541e-06, "loss": 1.446, "step": 325 }, { "epoch": 0.03, "learning_rate": 9.999422273403714e-06, "loss": 1.5356, "step": 326 }, { "epoch": 0.03, "learning_rate": 9.999395715226365e-06, "loss": 1.587, "step": 327 }, { "epoch": 0.03, "learning_rate": 9.999368560284663e-06, "loss": 1.281, "step": 328 }, { "epoch": 0.04, "learning_rate": 9.999340808581851e-06, "loss": 1.3646, "step": 329 }, { "epoch": 0.04, "learning_rate": 9.999312460121242e-06, "loss": 1.7839, "step": 330 }, { "epoch": 0.04, "learning_rate": 9.999283514906217e-06, "loss": 1.7553, "step": 331 }, { "epoch": 0.04, "learning_rate": 9.999253972940233e-06, "loss": 1.5545, "step": 332 }, { "epoch": 0.04, "learning_rate": 9.999223834226817e-06, "loss": 1.6097, "step": 333 }, { "epoch": 0.04, "learning_rate": 9.999193098769566e-06, "loss": 1.6592, "step": 334 }, { "epoch": 0.04, "learning_rate": 9.999161766572148e-06, "loss": 1.748, "step": 335 }, { "epoch": 0.04, "learning_rate": 9.999129837638303e-06, "loss": 1.7108, "step": 336 }, { "epoch": 0.04, "learning_rate": 9.999097311971846e-06, "loss": 1.8875, "step": 337 }, { "epoch": 0.04, "learning_rate": 9.999064189576653e-06, "loss": 1.6008, "step": 338 }, { "epoch": 0.04, "learning_rate": 9.999030470456684e-06, "loss": 1.8096, "step": 339 }, { "epoch": 0.04, "learning_rate": 9.998996154615959e-06, "loss": 1.5314, "step": 340 }, { "epoch": 0.04, "learning_rate": 9.998961242058578e-06, "loss": 1.8362, "step": 341 }, { "epoch": 0.04, "learning_rate": 9.998925732788706e-06, "loss": 1.2893, "step": 342 }, { "epoch": 0.04, "learning_rate": 9.998889626810581e-06, "loss": 1.6006, "step": 343 }, { "epoch": 0.04, "learning_rate": 9.998852924128518e-06, "loss": 1.7668, "step": 344 }, { "epoch": 0.04, "learning_rate": 9.99881562474689e-06, "loss": 1.4401, "step": 345 }, { "epoch": 0.04, "learning_rate": 9.998777728670156e-06, "loss": 1.6495, "step": 346 }, { "epoch": 0.04, "learning_rate": 9.998739235902836e-06, "loss": 1.8522, "step": 347 }, { "epoch": 0.04, "learning_rate": 9.998700146449528e-06, "loss": 1.547, "step": 348 }, { "epoch": 0.04, "learning_rate": 9.998660460314895e-06, "loss": 1.6714, "step": 349 }, { "epoch": 0.04, "learning_rate": 9.998620177503675e-06, "loss": 1.8747, "step": 350 }, { "epoch": 0.04, "learning_rate": 9.998579298020676e-06, "loss": 1.412, "step": 351 }, { "epoch": 0.04, "learning_rate": 9.99853782187078e-06, "loss": 1.6838, "step": 352 }, { "epoch": 0.04, "learning_rate": 9.998495749058935e-06, "loss": 1.6276, "step": 353 }, { "epoch": 0.04, "learning_rate": 9.998453079590167e-06, "loss": 1.5771, "step": 354 }, { "epoch": 0.04, "learning_rate": 9.998409813469563e-06, "loss": 1.7676, "step": 355 }, { "epoch": 0.04, "learning_rate": 9.998365950702294e-06, "loss": 1.529, "step": 356 }, { "epoch": 0.04, "learning_rate": 9.998321491293592e-06, "loss": 1.3431, "step": 357 }, { "epoch": 0.04, "learning_rate": 9.998276435248766e-06, "loss": 1.7557, "step": 358 }, { "epoch": 0.04, "learning_rate": 9.998230782573192e-06, "loss": 1.3257, "step": 359 }, { "epoch": 0.04, "learning_rate": 9.998184533272321e-06, "loss": 1.6638, "step": 360 }, { "epoch": 0.04, "learning_rate": 9.998137687351675e-06, "loss": 1.4728, "step": 361 }, { "epoch": 0.04, "learning_rate": 9.998090244816841e-06, "loss": 1.7045, "step": 362 }, { "epoch": 0.04, "learning_rate": 9.998042205673489e-06, "loss": 1.6437, "step": 363 }, { "epoch": 0.04, "learning_rate": 9.997993569927347e-06, "loss": 1.4374, "step": 364 }, { "epoch": 0.04, "learning_rate": 9.997944337584224e-06, "loss": 1.5488, "step": 365 }, { "epoch": 0.04, "learning_rate": 9.997894508649995e-06, "loss": 1.7224, "step": 366 }, { "epoch": 0.04, "learning_rate": 9.99784408313061e-06, "loss": 1.5511, "step": 367 }, { "epoch": 0.04, "learning_rate": 9.997793061032087e-06, "loss": 1.53, "step": 368 }, { "epoch": 0.04, "learning_rate": 9.997741442360515e-06, "loss": 1.842, "step": 369 }, { "epoch": 0.04, "learning_rate": 9.99768922712206e-06, "loss": 1.7144, "step": 370 }, { "epoch": 0.04, "learning_rate": 9.997636415322949e-06, "loss": 1.659, "step": 371 }, { "epoch": 0.04, "learning_rate": 9.99758300696949e-06, "loss": 1.7376, "step": 372 }, { "epoch": 0.04, "learning_rate": 9.997529002068056e-06, "loss": 1.4239, "step": 373 }, { "epoch": 0.04, "learning_rate": 9.997474400625096e-06, "loss": 1.5163, "step": 374 }, { "epoch": 0.04, "learning_rate": 9.997419202647124e-06, "loss": 1.8189, "step": 375 }, { "epoch": 0.04, "learning_rate": 9.997363408140732e-06, "loss": 1.6341, "step": 376 }, { "epoch": 0.04, "learning_rate": 9.997307017112579e-06, "loss": 1.5752, "step": 377 }, { "epoch": 0.04, "learning_rate": 9.997250029569395e-06, "loss": 1.7538, "step": 378 }, { "epoch": 0.04, "learning_rate": 9.997192445517985e-06, "loss": 1.6402, "step": 379 }, { "epoch": 0.04, "learning_rate": 9.99713426496522e-06, "loss": 1.4608, "step": 380 }, { "epoch": 0.04, "learning_rate": 9.997075487918047e-06, "loss": 1.477, "step": 381 }, { "epoch": 0.04, "learning_rate": 9.99701611438348e-06, "loss": 1.5604, "step": 382 }, { "epoch": 0.04, "learning_rate": 9.996956144368608e-06, "loss": 1.4379, "step": 383 }, { "epoch": 0.04, "learning_rate": 9.99689557788059e-06, "loss": 1.4854, "step": 384 }, { "epoch": 0.04, "learning_rate": 9.996834414926653e-06, "loss": 1.2965, "step": 385 }, { "epoch": 0.04, "learning_rate": 9.9967726555141e-06, "loss": 1.4524, "step": 386 }, { "epoch": 0.04, "learning_rate": 9.996710299650302e-06, "loss": 1.5107, "step": 387 }, { "epoch": 0.04, "learning_rate": 9.996647347342703e-06, "loss": 1.8207, "step": 388 }, { "epoch": 0.04, "learning_rate": 9.996583798598816e-06, "loss": 1.7247, "step": 389 }, { "epoch": 0.04, "learning_rate": 9.996519653426229e-06, "loss": 1.5735, "step": 390 }, { "epoch": 0.04, "learning_rate": 9.996454911832598e-06, "loss": 1.7618, "step": 391 }, { "epoch": 0.04, "learning_rate": 9.99638957382565e-06, "loss": 1.716, "step": 392 }, { "epoch": 0.04, "learning_rate": 9.996323639413185e-06, "loss": 1.6323, "step": 393 }, { "epoch": 0.04, "learning_rate": 9.996257108603073e-06, "loss": 1.5616, "step": 394 }, { "epoch": 0.04, "learning_rate": 9.996189981403255e-06, "loss": 1.5782, "step": 395 }, { "epoch": 0.04, "learning_rate": 9.996122257821746e-06, "loss": 1.5073, "step": 396 }, { "epoch": 0.04, "learning_rate": 9.996053937866628e-06, "loss": 1.5863, "step": 397 }, { "epoch": 0.04, "learning_rate": 9.995985021546058e-06, "loss": 1.4876, "step": 398 }, { "epoch": 0.04, "learning_rate": 9.99591550886826e-06, "loss": 1.5348, "step": 399 }, { "epoch": 0.04, "learning_rate": 9.995845399841533e-06, "loss": 1.5085, "step": 400 }, { "epoch": 0.04, "learning_rate": 9.995774694474245e-06, "loss": 1.3299, "step": 401 }, { "epoch": 0.04, "learning_rate": 9.995703392774836e-06, "loss": 1.4868, "step": 402 }, { "epoch": 0.04, "learning_rate": 9.995631494751819e-06, "loss": 1.5141, "step": 403 }, { "epoch": 0.04, "learning_rate": 9.995559000413773e-06, "loss": 1.5681, "step": 404 }, { "epoch": 0.04, "learning_rate": 9.995485909769354e-06, "loss": 1.5408, "step": 405 }, { "epoch": 0.04, "learning_rate": 9.995412222827286e-06, "loss": 1.6121, "step": 406 }, { "epoch": 0.04, "learning_rate": 9.995337939596364e-06, "loss": 1.4727, "step": 407 }, { "epoch": 0.04, "learning_rate": 9.995263060085456e-06, "loss": 1.6355, "step": 408 }, { "epoch": 0.04, "learning_rate": 9.995187584303498e-06, "loss": 1.5394, "step": 409 }, { "epoch": 0.04, "learning_rate": 9.995111512259503e-06, "loss": 1.7119, "step": 410 }, { "epoch": 0.04, "learning_rate": 9.99503484396255e-06, "loss": 1.3791, "step": 411 }, { "epoch": 0.04, "learning_rate": 9.994957579421789e-06, "loss": 1.3889, "step": 412 }, { "epoch": 0.04, "learning_rate": 9.994879718646445e-06, "loss": 1.7386, "step": 413 }, { "epoch": 0.04, "learning_rate": 9.99480126164581e-06, "loss": 1.8057, "step": 414 }, { "epoch": 0.04, "learning_rate": 9.994722208429251e-06, "loss": 1.5301, "step": 415 }, { "epoch": 0.04, "learning_rate": 9.994642559006204e-06, "loss": 1.4533, "step": 416 }, { "epoch": 0.04, "learning_rate": 9.994562313386177e-06, "loss": 1.8972, "step": 417 }, { "epoch": 0.04, "learning_rate": 9.994481471578748e-06, "loss": 1.8653, "step": 418 }, { "epoch": 0.04, "learning_rate": 9.994400033593566e-06, "loss": 1.536, "step": 419 }, { "epoch": 0.04, "learning_rate": 9.994317999440351e-06, "loss": 1.6031, "step": 420 }, { "epoch": 0.04, "learning_rate": 9.994235369128899e-06, "loss": 1.3316, "step": 421 }, { "epoch": 0.05, "learning_rate": 9.994152142669073e-06, "loss": 1.5406, "step": 422 }, { "epoch": 0.05, "learning_rate": 9.994068320070805e-06, "loss": 1.5241, "step": 423 }, { "epoch": 0.05, "learning_rate": 9.9939839013441e-06, "loss": 1.6501, "step": 424 }, { "epoch": 0.05, "learning_rate": 9.993898886499037e-06, "loss": 1.4812, "step": 425 }, { "epoch": 0.05, "learning_rate": 9.993813275545764e-06, "loss": 1.5441, "step": 426 }, { "epoch": 0.05, "learning_rate": 9.9937270684945e-06, "loss": 1.7472, "step": 427 }, { "epoch": 0.05, "learning_rate": 9.993640265355534e-06, "loss": 1.513, "step": 428 }, { "epoch": 0.05, "learning_rate": 9.99355286613923e-06, "loss": 1.6729, "step": 429 }, { "epoch": 0.05, "learning_rate": 9.993464870856017e-06, "loss": 1.6811, "step": 430 }, { "epoch": 0.05, "learning_rate": 9.9933762795164e-06, "loss": 1.5194, "step": 431 }, { "epoch": 0.05, "learning_rate": 9.993287092130956e-06, "loss": 1.6977, "step": 432 }, { "epoch": 0.05, "learning_rate": 9.993197308710332e-06, "loss": 1.7845, "step": 433 }, { "epoch": 0.05, "learning_rate": 9.99310692926524e-06, "loss": 1.5468, "step": 434 }, { "epoch": 0.05, "learning_rate": 9.993015953806472e-06, "loss": 1.6235, "step": 435 }, { "epoch": 0.05, "learning_rate": 9.992924382344887e-06, "loss": 1.4874, "step": 436 }, { "epoch": 0.05, "learning_rate": 9.992832214891415e-06, "loss": 1.6792, "step": 437 }, { "epoch": 0.05, "learning_rate": 9.992739451457058e-06, "loss": 1.6648, "step": 438 }, { "epoch": 0.05, "learning_rate": 9.992646092052888e-06, "loss": 1.6362, "step": 439 }, { "epoch": 0.05, "learning_rate": 9.992552136690052e-06, "loss": 1.7583, "step": 440 }, { "epoch": 0.05, "learning_rate": 9.992457585379764e-06, "loss": 1.7187, "step": 441 }, { "epoch": 0.05, "learning_rate": 9.992362438133307e-06, "loss": 1.4503, "step": 442 }, { "epoch": 0.05, "learning_rate": 9.992266694962044e-06, "loss": 1.6773, "step": 443 }, { "epoch": 0.05, "learning_rate": 9.992170355877398e-06, "loss": 1.5089, "step": 444 }, { "epoch": 0.05, "learning_rate": 9.992073420890873e-06, "loss": 1.5087, "step": 445 }, { "epoch": 0.05, "learning_rate": 9.991975890014039e-06, "loss": 1.5504, "step": 446 }, { "epoch": 0.05, "learning_rate": 9.991877763258538e-06, "loss": 1.5269, "step": 447 }, { "epoch": 0.05, "learning_rate": 9.99177904063608e-06, "loss": 1.2847, "step": 448 }, { "epoch": 0.05, "learning_rate": 9.991679722158451e-06, "loss": 1.7873, "step": 449 }, { "epoch": 0.05, "learning_rate": 9.991579807837511e-06, "loss": 1.6319, "step": 450 }, { "epoch": 0.05, "learning_rate": 9.99147929768518e-06, "loss": 1.7495, "step": 451 }, { "epoch": 0.05, "learning_rate": 9.991378191713458e-06, "loss": 1.3, "step": 452 }, { "epoch": 0.05, "learning_rate": 9.991276489934416e-06, "loss": 1.7717, "step": 453 }, { "epoch": 0.05, "learning_rate": 9.991174192360189e-06, "loss": 1.6141, "step": 454 }, { "epoch": 0.05, "learning_rate": 9.991071299002992e-06, "loss": 1.5339, "step": 455 }, { "epoch": 0.05, "learning_rate": 9.990967809875107e-06, "loss": 1.6499, "step": 456 }, { "epoch": 0.05, "learning_rate": 9.990863724988886e-06, "loss": 1.5526, "step": 457 }, { "epoch": 0.05, "learning_rate": 9.990759044356753e-06, "loss": 1.8673, "step": 458 }, { "epoch": 0.05, "learning_rate": 9.990653767991203e-06, "loss": 1.3274, "step": 459 }, { "epoch": 0.05, "learning_rate": 9.990547895904806e-06, "loss": 1.6918, "step": 460 }, { "epoch": 0.05, "learning_rate": 9.990441428110193e-06, "loss": 1.6305, "step": 461 }, { "epoch": 0.05, "learning_rate": 9.99033436462008e-06, "loss": 1.4862, "step": 462 }, { "epoch": 0.05, "learning_rate": 9.990226705447244e-06, "loss": 1.3468, "step": 463 }, { "epoch": 0.05, "learning_rate": 9.990118450604535e-06, "loss": 1.6821, "step": 464 }, { "epoch": 0.05, "learning_rate": 9.990009600104875e-06, "loss": 1.5297, "step": 465 }, { "epoch": 0.05, "learning_rate": 9.98990015396126e-06, "loss": 1.8783, "step": 466 }, { "epoch": 0.05, "learning_rate": 9.98979011218675e-06, "loss": 1.5218, "step": 467 }, { "epoch": 0.05, "learning_rate": 9.989679474794484e-06, "loss": 1.3904, "step": 468 }, { "epoch": 0.05, "learning_rate": 9.989568241797667e-06, "loss": 1.682, "step": 469 }, { "epoch": 0.05, "learning_rate": 9.989456413209578e-06, "loss": 1.4476, "step": 470 }, { "epoch": 0.05, "learning_rate": 9.989343989043563e-06, "loss": 1.8368, "step": 471 }, { "epoch": 0.05, "learning_rate": 9.989230969313044e-06, "loss": 1.52, "step": 472 }, { "epoch": 0.05, "learning_rate": 9.98911735403151e-06, "loss": 1.6247, "step": 473 }, { "epoch": 0.05, "learning_rate": 9.989003143212526e-06, "loss": 1.4333, "step": 474 }, { "epoch": 0.05, "learning_rate": 9.988888336869722e-06, "loss": 1.5057, "step": 475 }, { "epoch": 0.05, "learning_rate": 9.988772935016802e-06, "loss": 1.6075, "step": 476 }, { "epoch": 0.05, "learning_rate": 9.988656937667544e-06, "loss": 1.6488, "step": 477 }, { "epoch": 0.05, "learning_rate": 9.988540344835794e-06, "loss": 1.3106, "step": 478 }, { "epoch": 0.05, "learning_rate": 9.988423156535465e-06, "loss": 1.4183, "step": 479 }, { "epoch": 0.05, "learning_rate": 9.98830537278055e-06, "loss": 1.5497, "step": 480 }, { "epoch": 0.05, "learning_rate": 9.988186993585108e-06, "loss": 1.604, "step": 481 }, { "epoch": 0.05, "learning_rate": 9.988068018963268e-06, "loss": 1.3238, "step": 482 }, { "epoch": 0.05, "learning_rate": 9.987948448929232e-06, "loss": 1.5117, "step": 483 }, { "epoch": 0.05, "learning_rate": 9.987828283497272e-06, "loss": 1.5438, "step": 484 }, { "epoch": 0.05, "learning_rate": 9.987707522681735e-06, "loss": 1.6177, "step": 485 }, { "epoch": 0.05, "learning_rate": 9.987586166497032e-06, "loss": 1.7438, "step": 486 }, { "epoch": 0.05, "learning_rate": 9.987464214957652e-06, "loss": 1.7012, "step": 487 }, { "epoch": 0.05, "learning_rate": 9.98734166807815e-06, "loss": 1.617, "step": 488 }, { "epoch": 0.05, "learning_rate": 9.987218525873155e-06, "loss": 1.7793, "step": 489 }, { "epoch": 0.05, "learning_rate": 9.987094788357367e-06, "loss": 1.4342, "step": 490 }, { "epoch": 0.05, "learning_rate": 9.986970455545555e-06, "loss": 1.7657, "step": 491 }, { "epoch": 0.05, "learning_rate": 9.98684552745256e-06, "loss": 1.3478, "step": 492 }, { "epoch": 0.05, "learning_rate": 9.986720004093295e-06, "loss": 1.3339, "step": 493 }, { "epoch": 0.05, "learning_rate": 9.986593885482744e-06, "loss": 1.8348, "step": 494 }, { "epoch": 0.05, "learning_rate": 9.98646717163596e-06, "loss": 1.5259, "step": 495 }, { "epoch": 0.05, "learning_rate": 9.986339862568068e-06, "loss": 1.4533, "step": 496 }, { "epoch": 0.05, "learning_rate": 9.986211958294267e-06, "loss": 1.6568, "step": 497 }, { "epoch": 0.05, "learning_rate": 9.986083458829824e-06, "loss": 1.6381, "step": 498 }, { "epoch": 0.05, "learning_rate": 9.985954364190076e-06, "loss": 1.578, "step": 499 }, { "epoch": 0.05, "learning_rate": 9.985824674390434e-06, "loss": 1.5069, "step": 500 }, { "epoch": 0.05, "learning_rate": 9.985694389446378e-06, "loss": 1.5728, "step": 501 }, { "epoch": 0.05, "learning_rate": 9.98556350937346e-06, "loss": 1.5289, "step": 502 }, { "epoch": 0.05, "learning_rate": 9.985432034187304e-06, "loss": 1.6698, "step": 503 }, { "epoch": 0.05, "learning_rate": 9.9852999639036e-06, "loss": 1.3799, "step": 504 }, { "epoch": 0.05, "learning_rate": 9.98516729853812e-06, "loss": 1.4658, "step": 505 }, { "epoch": 0.05, "learning_rate": 9.98503403810669e-06, "loss": 1.3574, "step": 506 }, { "epoch": 0.05, "learning_rate": 9.984900182625226e-06, "loss": 1.8139, "step": 507 }, { "epoch": 0.05, "learning_rate": 9.984765732109703e-06, "loss": 1.6837, "step": 508 }, { "epoch": 0.05, "learning_rate": 9.984630686576167e-06, "loss": 1.7652, "step": 509 }, { "epoch": 0.05, "learning_rate": 9.98449504604074e-06, "loss": 1.2556, "step": 510 }, { "epoch": 0.05, "learning_rate": 9.984358810519616e-06, "loss": 1.4903, "step": 511 }, { "epoch": 0.05, "learning_rate": 9.984221980029054e-06, "loss": 1.5937, "step": 512 }, { "epoch": 0.05, "learning_rate": 9.984084554585387e-06, "loss": 1.6716, "step": 513 }, { "epoch": 0.05, "learning_rate": 9.98394653420502e-06, "loss": 1.596, "step": 514 }, { "epoch": 0.05, "learning_rate": 9.983807918904428e-06, "loss": 1.3769, "step": 515 }, { "epoch": 0.06, "learning_rate": 9.983668708700156e-06, "loss": 1.5019, "step": 516 }, { "epoch": 0.06, "learning_rate": 9.983528903608824e-06, "loss": 1.871, "step": 517 }, { "epoch": 0.06, "learning_rate": 9.983388503647117e-06, "loss": 1.4318, "step": 518 }, { "epoch": 0.06, "learning_rate": 9.983247508831795e-06, "loss": 1.5364, "step": 519 }, { "epoch": 0.06, "learning_rate": 9.98310591917969e-06, "loss": 1.5812, "step": 520 }, { "epoch": 0.06, "learning_rate": 9.982963734707701e-06, "loss": 1.6494, "step": 521 }, { "epoch": 0.06, "learning_rate": 9.9828209554328e-06, "loss": 1.729, "step": 522 }, { "epoch": 0.06, "learning_rate": 9.982677581372033e-06, "loss": 1.4152, "step": 523 }, { "epoch": 0.06, "learning_rate": 9.982533612542511e-06, "loss": 1.6263, "step": 524 }, { "epoch": 0.06, "learning_rate": 9.982389048961421e-06, "loss": 1.5955, "step": 525 }, { "epoch": 0.06, "learning_rate": 9.982243890646018e-06, "loss": 1.7978, "step": 526 }, { "epoch": 0.06, "learning_rate": 9.982098137613631e-06, "loss": 1.481, "step": 527 }, { "epoch": 0.06, "learning_rate": 9.981951789881657e-06, "loss": 1.2855, "step": 528 }, { "epoch": 0.06, "learning_rate": 9.981804847467564e-06, "loss": 1.5996, "step": 529 }, { "epoch": 0.06, "learning_rate": 9.981657310388893e-06, "loss": 1.6496, "step": 530 }, { "epoch": 0.06, "learning_rate": 9.981509178663256e-06, "loss": 1.5886, "step": 531 }, { "epoch": 0.06, "learning_rate": 9.981360452308334e-06, "loss": 1.6745, "step": 532 }, { "epoch": 0.06, "learning_rate": 9.981211131341881e-06, "loss": 1.3847, "step": 533 }, { "epoch": 0.06, "learning_rate": 9.98106121578172e-06, "loss": 1.7116, "step": 534 }, { "epoch": 0.06, "learning_rate": 9.980910705645747e-06, "loss": 1.5175, "step": 535 }, { "epoch": 0.06, "learning_rate": 9.980759600951926e-06, "loss": 1.728, "step": 536 }, { "epoch": 0.06, "learning_rate": 9.980607901718297e-06, "loss": 1.6749, "step": 537 }, { "epoch": 0.06, "learning_rate": 9.980455607962967e-06, "loss": 1.5016, "step": 538 }, { "epoch": 0.06, "learning_rate": 9.980302719704112e-06, "loss": 1.4494, "step": 539 }, { "epoch": 0.06, "learning_rate": 9.980149236959986e-06, "loss": 1.478, "step": 540 }, { "epoch": 0.06, "learning_rate": 9.979995159748907e-06, "loss": 1.4095, "step": 541 }, { "epoch": 0.06, "learning_rate": 9.979840488089268e-06, "loss": 1.4168, "step": 542 }, { "epoch": 0.06, "learning_rate": 9.979685221999532e-06, "loss": 1.5798, "step": 543 }, { "epoch": 0.06, "learning_rate": 9.979529361498233e-06, "loss": 1.3894, "step": 544 }, { "epoch": 0.06, "learning_rate": 9.979372906603973e-06, "loss": 1.5908, "step": 545 }, { "epoch": 0.06, "learning_rate": 9.97921585733543e-06, "loss": 1.4294, "step": 546 }, { "epoch": 0.06, "learning_rate": 9.97905821371135e-06, "loss": 1.5282, "step": 547 }, { "epoch": 0.06, "learning_rate": 9.978899975750548e-06, "loss": 1.5595, "step": 548 }, { "epoch": 0.06, "learning_rate": 9.97874114347192e-06, "loss": 1.5422, "step": 549 }, { "epoch": 0.06, "learning_rate": 9.978581716894417e-06, "loss": 1.6049, "step": 550 }, { "epoch": 0.06, "learning_rate": 9.978421696037073e-06, "loss": 1.3755, "step": 551 }, { "epoch": 0.06, "learning_rate": 9.978261080918988e-06, "loss": 1.6059, "step": 552 }, { "epoch": 0.06, "learning_rate": 9.978099871559338e-06, "loss": 1.6913, "step": 553 }, { "epoch": 0.06, "learning_rate": 9.977938067977359e-06, "loss": 1.5444, "step": 554 }, { "epoch": 0.06, "learning_rate": 9.977775670192373e-06, "loss": 1.7369, "step": 555 }, { "epoch": 0.06, "learning_rate": 9.977612678223759e-06, "loss": 1.3262, "step": 556 }, { "epoch": 0.06, "learning_rate": 9.977449092090977e-06, "loss": 1.5184, "step": 557 }, { "epoch": 0.06, "learning_rate": 9.977284911813549e-06, "loss": 1.6238, "step": 558 }, { "epoch": 0.06, "learning_rate": 9.977120137411076e-06, "loss": 1.4353, "step": 559 }, { "epoch": 0.06, "learning_rate": 9.976954768903228e-06, "loss": 1.3934, "step": 560 }, { "epoch": 0.06, "learning_rate": 9.976788806309742e-06, "loss": 1.6471, "step": 561 }, { "epoch": 0.06, "learning_rate": 9.976622249650432e-06, "loss": 1.4133, "step": 562 }, { "epoch": 0.06, "learning_rate": 9.976455098945175e-06, "loss": 1.6401, "step": 563 }, { "epoch": 0.06, "learning_rate": 9.976287354213924e-06, "loss": 1.3134, "step": 564 }, { "epoch": 0.06, "learning_rate": 9.976119015476706e-06, "loss": 1.482, "step": 565 }, { "epoch": 0.06, "learning_rate": 9.975950082753612e-06, "loss": 1.6827, "step": 566 }, { "epoch": 0.06, "learning_rate": 9.975780556064806e-06, "loss": 1.8014, "step": 567 }, { "epoch": 0.06, "learning_rate": 9.975610435430528e-06, "loss": 1.7475, "step": 568 }, { "epoch": 0.06, "learning_rate": 9.975439720871079e-06, "loss": 1.6008, "step": 569 }, { "epoch": 0.06, "learning_rate": 9.975268412406842e-06, "loss": 1.5643, "step": 570 }, { "epoch": 0.06, "learning_rate": 9.975096510058265e-06, "loss": 1.5909, "step": 571 }, { "epoch": 0.06, "learning_rate": 9.974924013845865e-06, "loss": 1.6284, "step": 572 }, { "epoch": 0.06, "learning_rate": 9.974750923790234e-06, "loss": 1.5873, "step": 573 }, { "epoch": 0.06, "learning_rate": 9.974577239912033e-06, "loss": 1.5657, "step": 574 }, { "epoch": 0.06, "learning_rate": 9.974402962231994e-06, "loss": 1.679, "step": 575 }, { "epoch": 0.06, "learning_rate": 9.97422809077092e-06, "loss": 1.7367, "step": 576 }, { "epoch": 0.06, "learning_rate": 9.974052625549687e-06, "loss": 1.4505, "step": 577 }, { "epoch": 0.06, "learning_rate": 9.973876566589236e-06, "loss": 1.6456, "step": 578 }, { "epoch": 0.06, "learning_rate": 9.973699913910584e-06, "loss": 1.528, "step": 579 }, { "epoch": 0.06, "learning_rate": 9.97352266753482e-06, "loss": 1.5277, "step": 580 }, { "epoch": 0.06, "learning_rate": 9.973344827483098e-06, "loss": 1.494, "step": 581 }, { "epoch": 0.06, "learning_rate": 9.97316639377665e-06, "loss": 1.4883, "step": 582 }, { "epoch": 0.06, "learning_rate": 9.972987366436772e-06, "loss": 1.3931, "step": 583 }, { "epoch": 0.06, "learning_rate": 9.972807745484833e-06, "loss": 1.2453, "step": 584 }, { "epoch": 0.06, "learning_rate": 9.97262753094228e-06, "loss": 1.6694, "step": 585 }, { "epoch": 0.06, "learning_rate": 9.972446722830618e-06, "loss": 1.6397, "step": 586 }, { "epoch": 0.06, "learning_rate": 9.972265321171433e-06, "loss": 1.8862, "step": 587 }, { "epoch": 0.06, "learning_rate": 9.972083325986377e-06, "loss": 1.4745, "step": 588 }, { "epoch": 0.06, "learning_rate": 9.971900737297176e-06, "loss": 1.8266, "step": 589 }, { "epoch": 0.06, "learning_rate": 9.971717555125623e-06, "loss": 1.5593, "step": 590 }, { "epoch": 0.06, "learning_rate": 9.971533779493586e-06, "loss": 1.452, "step": 591 }, { "epoch": 0.06, "learning_rate": 9.971349410423e-06, "loss": 1.9045, "step": 592 }, { "epoch": 0.06, "learning_rate": 9.971164447935875e-06, "loss": 1.583, "step": 593 }, { "epoch": 0.06, "learning_rate": 9.970978892054286e-06, "loss": 1.7951, "step": 594 }, { "epoch": 0.06, "learning_rate": 9.970792742800386e-06, "loss": 1.62, "step": 595 }, { "epoch": 0.06, "learning_rate": 9.970606000196392e-06, "loss": 1.6406, "step": 596 }, { "epoch": 0.06, "learning_rate": 9.970418664264596e-06, "loss": 1.6128, "step": 597 }, { "epoch": 0.06, "learning_rate": 9.970230735027362e-06, "loss": 1.8169, "step": 598 }, { "epoch": 0.06, "learning_rate": 9.970042212507118e-06, "loss": 1.5835, "step": 599 }, { "epoch": 0.06, "learning_rate": 9.969853096726372e-06, "loss": 1.4486, "step": 600 }, { "epoch": 0.06, "learning_rate": 9.969663387707696e-06, "loss": 1.4393, "step": 601 }, { "epoch": 0.06, "learning_rate": 9.969473085473735e-06, "loss": 1.3558, "step": 602 }, { "epoch": 0.06, "learning_rate": 9.969282190047207e-06, "loss": 1.7325, "step": 603 }, { "epoch": 0.06, "learning_rate": 9.969090701450896e-06, "loss": 1.9142, "step": 604 }, { "epoch": 0.06, "learning_rate": 9.96889861970766e-06, "loss": 1.4041, "step": 605 }, { "epoch": 0.06, "learning_rate": 9.968705944840428e-06, "loss": 1.4745, "step": 606 }, { "epoch": 0.06, "learning_rate": 9.9685126768722e-06, "loss": 1.6269, "step": 607 }, { "epoch": 0.06, "learning_rate": 9.968318815826041e-06, "loss": 1.8297, "step": 608 }, { "epoch": 0.06, "learning_rate": 9.968124361725098e-06, "loss": 1.7873, "step": 609 }, { "epoch": 0.07, "learning_rate": 9.96792931459258e-06, "loss": 1.6616, "step": 610 }, { "epoch": 0.07, "learning_rate": 9.96773367445177e-06, "loss": 1.7235, "step": 611 }, { "epoch": 0.07, "learning_rate": 9.967537441326018e-06, "loss": 1.6812, "step": 612 }, { "epoch": 0.07, "learning_rate": 9.96734061523875e-06, "loss": 1.5347, "step": 613 }, { "epoch": 0.07, "learning_rate": 9.96714319621346e-06, "loss": 1.5627, "step": 614 }, { "epoch": 0.07, "learning_rate": 9.966945184273716e-06, "loss": 1.6748, "step": 615 }, { "epoch": 0.07, "learning_rate": 9.966746579443152e-06, "loss": 1.647, "step": 616 }, { "epoch": 0.07, "learning_rate": 9.966547381745473e-06, "loss": 1.4385, "step": 617 }, { "epoch": 0.07, "learning_rate": 9.966347591204459e-06, "loss": 1.4881, "step": 618 }, { "epoch": 0.07, "learning_rate": 9.96614720784396e-06, "loss": 1.666, "step": 619 }, { "epoch": 0.07, "learning_rate": 9.96594623168789e-06, "loss": 1.6493, "step": 620 }, { "epoch": 0.07, "learning_rate": 9.965744662760246e-06, "loss": 1.5207, "step": 621 }, { "epoch": 0.07, "learning_rate": 9.965542501085082e-06, "loss": 1.3391, "step": 622 }, { "epoch": 0.07, "learning_rate": 9.965339746686536e-06, "loss": 1.4419, "step": 623 }, { "epoch": 0.07, "learning_rate": 9.965136399588803e-06, "loss": 1.4628, "step": 624 }, { "epoch": 0.07, "learning_rate": 9.964932459816161e-06, "loss": 1.3786, "step": 625 }, { "epoch": 0.07, "learning_rate": 9.964727927392954e-06, "loss": 1.5852, "step": 626 }, { "epoch": 0.07, "learning_rate": 9.964522802343593e-06, "loss": 1.7696, "step": 627 }, { "epoch": 0.07, "learning_rate": 9.964317084692568e-06, "loss": 1.5509, "step": 628 }, { "epoch": 0.07, "learning_rate": 9.964110774464429e-06, "loss": 1.4914, "step": 629 }, { "epoch": 0.07, "learning_rate": 9.963903871683806e-06, "loss": 1.7029, "step": 630 }, { "epoch": 0.07, "learning_rate": 9.963696376375399e-06, "loss": 1.3677, "step": 631 }, { "epoch": 0.07, "learning_rate": 9.963488288563972e-06, "loss": 1.4934, "step": 632 }, { "epoch": 0.07, "learning_rate": 9.963279608274364e-06, "loss": 1.4003, "step": 633 }, { "epoch": 0.07, "learning_rate": 9.963070335531488e-06, "loss": 1.4592, "step": 634 }, { "epoch": 0.07, "learning_rate": 9.962860470360321e-06, "loss": 1.7099, "step": 635 }, { "epoch": 0.07, "learning_rate": 9.962650012785917e-06, "loss": 1.5212, "step": 636 }, { "epoch": 0.07, "learning_rate": 9.962438962833393e-06, "loss": 1.5054, "step": 637 }, { "epoch": 0.07, "learning_rate": 9.962227320527946e-06, "loss": 1.5587, "step": 638 }, { "epoch": 0.07, "learning_rate": 9.962015085894838e-06, "loss": 1.473, "step": 639 }, { "epoch": 0.07, "learning_rate": 9.961802258959402e-06, "loss": 1.4512, "step": 640 }, { "epoch": 0.07, "learning_rate": 9.961588839747044e-06, "loss": 1.8017, "step": 641 }, { "epoch": 0.07, "learning_rate": 9.961374828283239e-06, "loss": 1.3407, "step": 642 }, { "epoch": 0.07, "learning_rate": 9.961160224593531e-06, "loss": 1.4556, "step": 643 }, { "epoch": 0.07, "learning_rate": 9.960945028703539e-06, "loss": 1.7725, "step": 644 }, { "epoch": 0.07, "learning_rate": 9.960729240638947e-06, "loss": 1.7478, "step": 645 }, { "epoch": 0.07, "learning_rate": 9.960512860425517e-06, "loss": 1.6754, "step": 646 }, { "epoch": 0.07, "learning_rate": 9.960295888089078e-06, "loss": 1.493, "step": 647 }, { "epoch": 0.07, "learning_rate": 9.960078323655524e-06, "loss": 1.6572, "step": 648 }, { "epoch": 0.07, "learning_rate": 9.959860167150832e-06, "loss": 1.2692, "step": 649 }, { "epoch": 0.07, "learning_rate": 9.959641418601037e-06, "loss": 1.7195, "step": 650 }, { "epoch": 0.07, "learning_rate": 9.959422078032253e-06, "loss": 1.6932, "step": 651 }, { "epoch": 0.07, "learning_rate": 9.959202145470663e-06, "loss": 1.8293, "step": 652 }, { "epoch": 0.07, "learning_rate": 9.958981620942519e-06, "loss": 1.4656, "step": 653 }, { "epoch": 0.07, "learning_rate": 9.958760504474144e-06, "loss": 1.595, "step": 654 }, { "epoch": 0.07, "learning_rate": 9.95853879609193e-06, "loss": 1.4272, "step": 655 }, { "epoch": 0.07, "learning_rate": 9.958316495822345e-06, "loss": 1.4598, "step": 656 }, { "epoch": 0.07, "learning_rate": 9.958093603691923e-06, "loss": 1.7935, "step": 657 }, { "epoch": 0.07, "learning_rate": 9.957870119727271e-06, "loss": 1.4502, "step": 658 }, { "epoch": 0.07, "learning_rate": 9.957646043955066e-06, "loss": 1.5746, "step": 659 }, { "epoch": 0.07, "learning_rate": 9.957421376402053e-06, "loss": 1.2831, "step": 660 }, { "epoch": 0.07, "learning_rate": 9.957196117095048e-06, "loss": 1.4115, "step": 661 }, { "epoch": 0.07, "learning_rate": 9.956970266060947e-06, "loss": 1.6881, "step": 662 }, { "epoch": 0.07, "learning_rate": 9.956743823326704e-06, "loss": 1.6986, "step": 663 }, { "epoch": 0.07, "learning_rate": 9.95651678891935e-06, "loss": 1.781, "step": 664 }, { "epoch": 0.07, "learning_rate": 9.956289162865987e-06, "loss": 1.6389, "step": 665 }, { "epoch": 0.07, "learning_rate": 9.956060945193781e-06, "loss": 1.3822, "step": 666 }, { "epoch": 0.07, "learning_rate": 9.955832135929978e-06, "loss": 1.5833, "step": 667 }, { "epoch": 0.07, "learning_rate": 9.955602735101892e-06, "loss": 1.8868, "step": 668 }, { "epoch": 0.07, "learning_rate": 9.955372742736903e-06, "loss": 1.8119, "step": 669 }, { "epoch": 0.07, "learning_rate": 9.955142158862463e-06, "loss": 1.5566, "step": 670 }, { "epoch": 0.07, "learning_rate": 9.9549109835061e-06, "loss": 1.4952, "step": 671 }, { "epoch": 0.07, "learning_rate": 9.954679216695406e-06, "loss": 1.4176, "step": 672 }, { "epoch": 0.07, "learning_rate": 9.95444685845805e-06, "loss": 1.7626, "step": 673 }, { "epoch": 0.07, "learning_rate": 9.954213908821762e-06, "loss": 1.6954, "step": 674 }, { "epoch": 0.07, "learning_rate": 9.953980367814354e-06, "loss": 1.7038, "step": 675 }, { "epoch": 0.07, "learning_rate": 9.953746235463699e-06, "loss": 1.4703, "step": 676 }, { "epoch": 0.07, "learning_rate": 9.95351151179775e-06, "loss": 1.5729, "step": 677 }, { "epoch": 0.07, "learning_rate": 9.953276196844519e-06, "loss": 1.8003, "step": 678 }, { "epoch": 0.07, "learning_rate": 9.9530402906321e-06, "loss": 1.5756, "step": 679 }, { "epoch": 0.07, "learning_rate": 9.95280379318865e-06, "loss": 1.4183, "step": 680 }, { "epoch": 0.07, "learning_rate": 9.9525667045424e-06, "loss": 1.6153, "step": 681 }, { "epoch": 0.07, "learning_rate": 9.95232902472165e-06, "loss": 1.359, "step": 682 }, { "epoch": 0.07, "learning_rate": 9.952090753754772e-06, "loss": 1.7143, "step": 683 }, { "epoch": 0.07, "learning_rate": 9.951851891670206e-06, "loss": 1.6061, "step": 684 }, { "epoch": 0.07, "learning_rate": 9.951612438496467e-06, "loss": 1.6962, "step": 685 }, { "epoch": 0.07, "learning_rate": 9.951372394262135e-06, "loss": 1.6735, "step": 686 }, { "epoch": 0.07, "learning_rate": 9.951131758995866e-06, "loss": 1.8219, "step": 687 }, { "epoch": 0.07, "learning_rate": 9.950890532726382e-06, "loss": 1.5118, "step": 688 }, { "epoch": 0.07, "learning_rate": 9.95064871548248e-06, "loss": 1.4755, "step": 689 }, { "epoch": 0.07, "learning_rate": 9.950406307293023e-06, "loss": 1.3991, "step": 690 }, { "epoch": 0.07, "learning_rate": 9.950163308186946e-06, "loss": 1.4538, "step": 691 }, { "epoch": 0.07, "learning_rate": 9.949919718193257e-06, "loss": 1.8093, "step": 692 }, { "epoch": 0.07, "learning_rate": 9.949675537341031e-06, "loss": 1.407, "step": 693 }, { "epoch": 0.07, "learning_rate": 9.949430765659417e-06, "loss": 1.7883, "step": 694 }, { "epoch": 0.07, "learning_rate": 9.949185403177632e-06, "loss": 1.537, "step": 695 }, { "epoch": 0.07, "learning_rate": 9.948939449924964e-06, "loss": 1.4456, "step": 696 }, { "epoch": 0.07, "learning_rate": 9.94869290593077e-06, "loss": 1.4764, "step": 697 }, { "epoch": 0.07, "learning_rate": 9.948445771224484e-06, "loss": 1.4147, "step": 698 }, { "epoch": 0.07, "learning_rate": 9.948198045835601e-06, "loss": 1.543, "step": 699 }, { "epoch": 0.07, "learning_rate": 9.947949729793693e-06, "loss": 1.5961, "step": 700 }, { "epoch": 0.07, "learning_rate": 9.947700823128403e-06, "loss": 1.6912, "step": 701 }, { "epoch": 0.07, "learning_rate": 9.94745132586944e-06, "loss": 1.3565, "step": 702 }, { "epoch": 0.07, "learning_rate": 9.947201238046585e-06, "loss": 1.3405, "step": 703 }, { "epoch": 0.08, "learning_rate": 9.946950559689691e-06, "loss": 1.7659, "step": 704 }, { "epoch": 0.08, "learning_rate": 9.946699290828683e-06, "loss": 1.4584, "step": 705 }, { "epoch": 0.08, "learning_rate": 9.946447431493553e-06, "loss": 1.6998, "step": 706 }, { "epoch": 0.08, "learning_rate": 9.946194981714364e-06, "loss": 1.4359, "step": 707 }, { "epoch": 0.08, "learning_rate": 9.94594194152125e-06, "loss": 1.6846, "step": 708 }, { "epoch": 0.08, "learning_rate": 9.945688310944415e-06, "loss": 1.559, "step": 709 }, { "epoch": 0.08, "learning_rate": 9.945434090014136e-06, "loss": 1.4531, "step": 710 }, { "epoch": 0.08, "learning_rate": 9.945179278760759e-06, "loss": 1.5819, "step": 711 }, { "epoch": 0.08, "learning_rate": 9.944923877214701e-06, "loss": 1.5219, "step": 712 }, { "epoch": 0.08, "learning_rate": 9.944667885406445e-06, "loss": 1.9019, "step": 713 }, { "epoch": 0.08, "learning_rate": 9.94441130336655e-06, "loss": 1.2682, "step": 714 }, { "epoch": 0.08, "learning_rate": 9.944154131125643e-06, "loss": 1.4954, "step": 715 }, { "epoch": 0.08, "learning_rate": 9.943896368714423e-06, "loss": 1.7321, "step": 716 }, { "epoch": 0.08, "learning_rate": 9.943638016163658e-06, "loss": 1.7305, "step": 717 }, { "epoch": 0.08, "learning_rate": 9.943379073504187e-06, "loss": 1.2885, "step": 718 }, { "epoch": 0.08, "learning_rate": 9.94311954076692e-06, "loss": 1.7663, "step": 719 }, { "epoch": 0.08, "learning_rate": 9.942859417982833e-06, "loss": 1.4326, "step": 720 }, { "epoch": 0.08, "learning_rate": 9.94259870518298e-06, "loss": 1.4033, "step": 721 }, { "epoch": 0.08, "learning_rate": 9.942337402398481e-06, "loss": 1.5034, "step": 722 }, { "epoch": 0.08, "learning_rate": 9.942075509660527e-06, "loss": 1.6213, "step": 723 }, { "epoch": 0.08, "learning_rate": 9.941813027000377e-06, "loss": 1.4841, "step": 724 }, { "epoch": 0.08, "learning_rate": 9.941549954449365e-06, "loss": 1.5709, "step": 725 }, { "epoch": 0.08, "learning_rate": 9.941286292038894e-06, "loss": 1.709, "step": 726 }, { "epoch": 0.08, "learning_rate": 9.941022039800437e-06, "loss": 1.5338, "step": 727 }, { "epoch": 0.08, "learning_rate": 9.940757197765533e-06, "loss": 1.6397, "step": 728 }, { "epoch": 0.08, "learning_rate": 9.940491765965798e-06, "loss": 1.742, "step": 729 }, { "epoch": 0.08, "learning_rate": 9.940225744432919e-06, "loss": 1.5769, "step": 730 }, { "epoch": 0.08, "learning_rate": 9.939959133198644e-06, "loss": 1.7751, "step": 731 }, { "epoch": 0.08, "learning_rate": 9.939691932294804e-06, "loss": 1.3854, "step": 732 }, { "epoch": 0.08, "learning_rate": 9.939424141753289e-06, "loss": 1.367, "step": 733 }, { "epoch": 0.08, "learning_rate": 9.939155761606066e-06, "loss": 1.5177, "step": 734 }, { "epoch": 0.08, "learning_rate": 9.938886791885172e-06, "loss": 1.6087, "step": 735 }, { "epoch": 0.08, "learning_rate": 9.938617232622713e-06, "loss": 1.6028, "step": 736 }, { "epoch": 0.08, "learning_rate": 9.938347083850866e-06, "loss": 1.6283, "step": 737 }, { "epoch": 0.08, "learning_rate": 9.938076345601875e-06, "loss": 1.2905, "step": 738 }, { "epoch": 0.08, "learning_rate": 9.93780501790806e-06, "loss": 1.7774, "step": 739 }, { "epoch": 0.08, "learning_rate": 9.937533100801808e-06, "loss": 1.5581, "step": 740 }, { "epoch": 0.08, "learning_rate": 9.937260594315578e-06, "loss": 1.7149, "step": 741 }, { "epoch": 0.08, "learning_rate": 9.936987498481896e-06, "loss": 1.6671, "step": 742 }, { "epoch": 0.08, "learning_rate": 9.936713813333362e-06, "loss": 1.7404, "step": 743 }, { "epoch": 0.08, "learning_rate": 9.936439538902644e-06, "loss": 1.3825, "step": 744 }, { "epoch": 0.08, "learning_rate": 9.936164675222485e-06, "loss": 1.4524, "step": 745 }, { "epoch": 0.08, "learning_rate": 9.93588922232569e-06, "loss": 1.4706, "step": 746 }, { "epoch": 0.08, "learning_rate": 9.935613180245143e-06, "loss": 1.7976, "step": 747 }, { "epoch": 0.08, "learning_rate": 9.935336549013791e-06, "loss": 1.7231, "step": 748 }, { "epoch": 0.08, "learning_rate": 9.935059328664657e-06, "loss": 1.5259, "step": 749 }, { "epoch": 0.08, "learning_rate": 9.934781519230832e-06, "loss": 1.5276, "step": 750 }, { "epoch": 0.08, "learning_rate": 9.934503120745476e-06, "loss": 1.5658, "step": 751 }, { "epoch": 0.08, "learning_rate": 9.934224133241823e-06, "loss": 1.4569, "step": 752 }, { "epoch": 0.08, "learning_rate": 9.933944556753173e-06, "loss": 1.4143, "step": 753 }, { "epoch": 0.08, "learning_rate": 9.933664391312897e-06, "loss": 1.3946, "step": 754 }, { "epoch": 0.08, "learning_rate": 9.93338363695444e-06, "loss": 1.5715, "step": 755 }, { "epoch": 0.08, "learning_rate": 9.933102293711314e-06, "loss": 1.6343, "step": 756 }, { "epoch": 0.08, "learning_rate": 9.932820361617104e-06, "loss": 1.184, "step": 757 }, { "epoch": 0.08, "learning_rate": 9.93253784070546e-06, "loss": 1.5569, "step": 758 }, { "epoch": 0.08, "learning_rate": 9.932254731010108e-06, "loss": 1.6529, "step": 759 }, { "epoch": 0.08, "learning_rate": 9.931971032564842e-06, "loss": 1.7953, "step": 760 }, { "epoch": 0.08, "learning_rate": 9.931686745403527e-06, "loss": 1.5865, "step": 761 }, { "epoch": 0.08, "learning_rate": 9.931401869560096e-06, "loss": 1.5324, "step": 762 }, { "epoch": 0.08, "learning_rate": 9.931116405068554e-06, "loss": 1.4654, "step": 763 }, { "epoch": 0.08, "learning_rate": 9.930830351962976e-06, "loss": 1.7972, "step": 764 }, { "epoch": 0.08, "learning_rate": 9.93054371027751e-06, "loss": 1.5653, "step": 765 }, { "epoch": 0.08, "learning_rate": 9.930256480046367e-06, "loss": 1.4124, "step": 766 }, { "epoch": 0.08, "learning_rate": 9.929968661303837e-06, "loss": 1.726, "step": 767 }, { "epoch": 0.08, "learning_rate": 9.929680254084273e-06, "loss": 1.6784, "step": 768 }, { "epoch": 0.08, "learning_rate": 9.929391258422106e-06, "loss": 1.4448, "step": 769 }, { "epoch": 0.08, "learning_rate": 9.929101674351827e-06, "loss": 1.6811, "step": 770 }, { "epoch": 0.08, "learning_rate": 9.928811501908006e-06, "loss": 1.4887, "step": 771 }, { "epoch": 0.08, "learning_rate": 9.92852074112528e-06, "loss": 1.6747, "step": 772 }, { "epoch": 0.08, "learning_rate": 9.928229392038356e-06, "loss": 1.6808, "step": 773 }, { "epoch": 0.08, "learning_rate": 9.92793745468201e-06, "loss": 1.6501, "step": 774 }, { "epoch": 0.08, "learning_rate": 9.927644929091094e-06, "loss": 1.7202, "step": 775 }, { "epoch": 0.08, "learning_rate": 9.927351815300522e-06, "loss": 1.4084, "step": 776 }, { "epoch": 0.08, "learning_rate": 9.927058113345282e-06, "loss": 1.7015, "step": 777 }, { "epoch": 0.08, "learning_rate": 9.926763823260437e-06, "loss": 1.6263, "step": 778 }, { "epoch": 0.08, "learning_rate": 9.926468945081109e-06, "loss": 1.4755, "step": 779 }, { "epoch": 0.08, "learning_rate": 9.926173478842502e-06, "loss": 1.6926, "step": 780 }, { "epoch": 0.08, "learning_rate": 9.925877424579884e-06, "loss": 1.6775, "step": 781 }, { "epoch": 0.08, "learning_rate": 9.925580782328592e-06, "loss": 1.5295, "step": 782 }, { "epoch": 0.08, "learning_rate": 9.925283552124039e-06, "loss": 1.4923, "step": 783 }, { "epoch": 0.08, "learning_rate": 9.9249857340017e-06, "loss": 1.3244, "step": 784 }, { "epoch": 0.08, "learning_rate": 9.924687327997128e-06, "loss": 1.6566, "step": 785 }, { "epoch": 0.08, "learning_rate": 9.924388334145943e-06, "loss": 1.5704, "step": 786 }, { "epoch": 0.08, "learning_rate": 9.924088752483834e-06, "loss": 1.4235, "step": 787 }, { "epoch": 0.08, "learning_rate": 9.923788583046561e-06, "loss": 1.5339, "step": 788 }, { "epoch": 0.08, "learning_rate": 9.923487825869955e-06, "loss": 1.5355, "step": 789 }, { "epoch": 0.08, "learning_rate": 9.923186480989916e-06, "loss": 1.2581, "step": 790 }, { "epoch": 0.08, "learning_rate": 9.922884548442416e-06, "loss": 1.6942, "step": 791 }, { "epoch": 0.08, "learning_rate": 9.922582028263495e-06, "loss": 1.5286, "step": 792 }, { "epoch": 0.08, "learning_rate": 9.922278920489262e-06, "loss": 1.477, "step": 793 }, { "epoch": 0.08, "learning_rate": 9.921975225155902e-06, "loss": 1.7184, "step": 794 }, { "epoch": 0.08, "learning_rate": 9.921670942299664e-06, "loss": 1.5819, "step": 795 }, { "epoch": 0.08, "learning_rate": 9.92136607195687e-06, "loss": 1.5482, "step": 796 }, { "epoch": 0.09, "learning_rate": 9.921060614163911e-06, "loss": 1.5667, "step": 797 }, { "epoch": 0.09, "learning_rate": 9.92075456895725e-06, "loss": 1.7756, "step": 798 }, { "epoch": 0.09, "learning_rate": 9.920447936373418e-06, "loss": 1.368, "step": 799 }, { "epoch": 0.09, "learning_rate": 9.920140716449016e-06, "loss": 1.5411, "step": 800 }, { "epoch": 0.09, "learning_rate": 9.919832909220717e-06, "loss": 1.322, "step": 801 }, { "epoch": 0.09, "learning_rate": 9.919524514725262e-06, "loss": 1.7059, "step": 802 }, { "epoch": 0.09, "learning_rate": 9.919215532999467e-06, "loss": 1.632, "step": 803 }, { "epoch": 0.09, "learning_rate": 9.91890596408021e-06, "loss": 1.197, "step": 804 }, { "epoch": 0.09, "learning_rate": 9.918595808004444e-06, "loss": 1.4875, "step": 805 }, { "epoch": 0.09, "learning_rate": 9.918285064809193e-06, "loss": 1.551, "step": 806 }, { "epoch": 0.09, "learning_rate": 9.917973734531549e-06, "loss": 1.3721, "step": 807 }, { "epoch": 0.09, "learning_rate": 9.917661817208675e-06, "loss": 1.7838, "step": 808 }, { "epoch": 0.09, "learning_rate": 9.917349312877802e-06, "loss": 1.374, "step": 809 }, { "epoch": 0.09, "learning_rate": 9.917036221576235e-06, "loss": 1.2005, "step": 810 }, { "epoch": 0.09, "learning_rate": 9.916722543341345e-06, "loss": 1.5434, "step": 811 }, { "epoch": 0.09, "learning_rate": 9.91640827821058e-06, "loss": 1.7451, "step": 812 }, { "epoch": 0.09, "learning_rate": 9.916093426221445e-06, "loss": 1.2927, "step": 813 }, { "epoch": 0.09, "learning_rate": 9.915777987411527e-06, "loss": 1.3975, "step": 814 }, { "epoch": 0.09, "learning_rate": 9.91546196181848e-06, "loss": 1.4734, "step": 815 }, { "epoch": 0.09, "learning_rate": 9.915145349480027e-06, "loss": 1.6287, "step": 816 }, { "epoch": 0.09, "learning_rate": 9.914828150433958e-06, "loss": 1.4432, "step": 817 }, { "epoch": 0.09, "learning_rate": 9.91451036471814e-06, "loss": 1.5681, "step": 818 }, { "epoch": 0.09, "learning_rate": 9.914191992370504e-06, "loss": 1.6349, "step": 819 }, { "epoch": 0.09, "learning_rate": 9.913873033429054e-06, "loss": 1.3528, "step": 820 }, { "epoch": 0.09, "learning_rate": 9.913553487931865e-06, "loss": 1.6091, "step": 821 }, { "epoch": 0.09, "learning_rate": 9.913233355917075e-06, "loss": 1.472, "step": 822 }, { "epoch": 0.09, "learning_rate": 9.912912637422905e-06, "loss": 1.5136, "step": 823 }, { "epoch": 0.09, "learning_rate": 9.912591332487631e-06, "loss": 1.6331, "step": 824 }, { "epoch": 0.09, "learning_rate": 9.91226944114961e-06, "loss": 1.7585, "step": 825 }, { "epoch": 0.09, "learning_rate": 9.911946963447265e-06, "loss": 1.687, "step": 826 }, { "epoch": 0.09, "learning_rate": 9.911623899419089e-06, "loss": 1.6048, "step": 827 }, { "epoch": 0.09, "learning_rate": 9.911300249103646e-06, "loss": 1.3621, "step": 828 }, { "epoch": 0.09, "learning_rate": 9.910976012539567e-06, "loss": 1.5964, "step": 829 }, { "epoch": 0.09, "learning_rate": 9.910651189765557e-06, "loss": 1.4258, "step": 830 }, { "epoch": 0.09, "learning_rate": 9.910325780820391e-06, "loss": 1.403, "step": 831 }, { "epoch": 0.09, "learning_rate": 9.909999785742908e-06, "loss": 1.6609, "step": 832 }, { "epoch": 0.09, "learning_rate": 9.909673204572023e-06, "loss": 1.5011, "step": 833 }, { "epoch": 0.09, "learning_rate": 9.90934603734672e-06, "loss": 1.6214, "step": 834 }, { "epoch": 0.09, "learning_rate": 9.909018284106054e-06, "loss": 1.3925, "step": 835 }, { "epoch": 0.09, "learning_rate": 9.908689944889143e-06, "loss": 1.5517, "step": 836 }, { "epoch": 0.09, "learning_rate": 9.908361019735181e-06, "loss": 1.513, "step": 837 }, { "epoch": 0.09, "learning_rate": 9.908031508683436e-06, "loss": 1.7635, "step": 838 }, { "epoch": 0.09, "learning_rate": 9.907701411773234e-06, "loss": 1.568, "step": 839 }, { "epoch": 0.09, "learning_rate": 9.907370729043984e-06, "loss": 1.4766, "step": 840 }, { "epoch": 0.09, "learning_rate": 9.907039460535153e-06, "loss": 1.2861, "step": 841 }, { "epoch": 0.09, "learning_rate": 9.906707606286287e-06, "loss": 1.5734, "step": 842 }, { "epoch": 0.09, "learning_rate": 9.906375166336998e-06, "loss": 1.4368, "step": 843 }, { "epoch": 0.09, "learning_rate": 9.90604214072697e-06, "loss": 1.5146, "step": 844 }, { "epoch": 0.09, "learning_rate": 9.905708529495953e-06, "loss": 1.7439, "step": 845 }, { "epoch": 0.09, "learning_rate": 9.905374332683768e-06, "loss": 1.3389, "step": 846 }, { "epoch": 0.09, "learning_rate": 9.90503955033031e-06, "loss": 1.7195, "step": 847 }, { "epoch": 0.09, "learning_rate": 9.904704182475542e-06, "loss": 1.6664, "step": 848 }, { "epoch": 0.09, "learning_rate": 9.904368229159494e-06, "loss": 1.6279, "step": 849 }, { "epoch": 0.09, "learning_rate": 9.904031690422266e-06, "loss": 1.3575, "step": 850 }, { "epoch": 0.09, "learning_rate": 9.903694566304031e-06, "loss": 1.706, "step": 851 }, { "epoch": 0.09, "learning_rate": 9.903356856845035e-06, "loss": 1.6124, "step": 852 }, { "epoch": 0.09, "learning_rate": 9.903018562085583e-06, "loss": 1.6807, "step": 853 }, { "epoch": 0.09, "learning_rate": 9.902679682066059e-06, "loss": 1.5372, "step": 854 }, { "epoch": 0.09, "learning_rate": 9.902340216826915e-06, "loss": 1.7125, "step": 855 }, { "epoch": 0.09, "learning_rate": 9.902000166408672e-06, "loss": 1.6261, "step": 856 }, { "epoch": 0.09, "learning_rate": 9.90165953085192e-06, "loss": 1.2165, "step": 857 }, { "epoch": 0.09, "learning_rate": 9.90131831019732e-06, "loss": 1.6056, "step": 858 }, { "epoch": 0.09, "learning_rate": 9.900976504485601e-06, "loss": 1.5965, "step": 859 }, { "epoch": 0.09, "learning_rate": 9.900634113757568e-06, "loss": 1.5316, "step": 860 }, { "epoch": 0.09, "learning_rate": 9.900291138054086e-06, "loss": 1.6883, "step": 861 }, { "epoch": 0.09, "learning_rate": 9.899947577416097e-06, "loss": 1.3782, "step": 862 }, { "epoch": 0.09, "learning_rate": 9.899603431884613e-06, "loss": 1.551, "step": 863 }, { "epoch": 0.09, "learning_rate": 9.899258701500712e-06, "loss": 1.7299, "step": 864 }, { "epoch": 0.09, "learning_rate": 9.898913386305542e-06, "loss": 1.7279, "step": 865 }, { "epoch": 0.09, "learning_rate": 9.898567486340325e-06, "loss": 1.6582, "step": 866 }, { "epoch": 0.09, "learning_rate": 9.89822100164635e-06, "loss": 1.503, "step": 867 }, { "epoch": 0.09, "learning_rate": 9.897873932264972e-06, "loss": 1.7705, "step": 868 }, { "epoch": 0.09, "learning_rate": 9.897526278237624e-06, "loss": 1.4934, "step": 869 }, { "epoch": 0.09, "learning_rate": 9.897178039605803e-06, "loss": 1.4269, "step": 870 }, { "epoch": 0.09, "learning_rate": 9.896829216411076e-06, "loss": 1.4922, "step": 871 }, { "epoch": 0.09, "learning_rate": 9.896479808695086e-06, "loss": 1.4434, "step": 872 }, { "epoch": 0.09, "learning_rate": 9.896129816499535e-06, "loss": 1.4031, "step": 873 }, { "epoch": 0.09, "learning_rate": 9.895779239866204e-06, "loss": 1.4319, "step": 874 }, { "epoch": 0.09, "learning_rate": 9.89542807883694e-06, "loss": 1.3222, "step": 875 }, { "epoch": 0.09, "learning_rate": 9.89507633345366e-06, "loss": 1.6923, "step": 876 }, { "epoch": 0.09, "learning_rate": 9.894724003758349e-06, "loss": 1.5617, "step": 877 }, { "epoch": 0.09, "learning_rate": 9.894371089793065e-06, "loss": 1.5404, "step": 878 }, { "epoch": 0.09, "learning_rate": 9.894017591599934e-06, "loss": 1.5225, "step": 879 }, { "epoch": 0.09, "learning_rate": 9.893663509221155e-06, "loss": 1.2448, "step": 880 }, { "epoch": 0.09, "learning_rate": 9.89330884269899e-06, "loss": 1.4143, "step": 881 }, { "epoch": 0.09, "learning_rate": 9.892953592075776e-06, "loss": 1.2362, "step": 882 }, { "epoch": 0.09, "learning_rate": 9.892597757393918e-06, "loss": 1.7226, "step": 883 }, { "epoch": 0.09, "learning_rate": 9.892241338695892e-06, "loss": 1.7645, "step": 884 }, { "epoch": 0.09, "learning_rate": 9.891884336024242e-06, "loss": 1.556, "step": 885 }, { "epoch": 0.09, "learning_rate": 9.891526749421583e-06, "loss": 1.6665, "step": 886 }, { "epoch": 0.09, "learning_rate": 9.891168578930597e-06, "loss": 1.832, "step": 887 }, { "epoch": 0.09, "learning_rate": 9.890809824594041e-06, "loss": 1.5574, "step": 888 }, { "epoch": 0.09, "learning_rate": 9.890450486454736e-06, "loss": 1.3075, "step": 889 }, { "epoch": 0.09, "learning_rate": 9.890090564555579e-06, "loss": 1.3686, "step": 890 }, { "epoch": 0.1, "learning_rate": 9.889730058939529e-06, "loss": 1.6001, "step": 891 }, { "epoch": 0.1, "learning_rate": 9.88936896964962e-06, "loss": 1.6415, "step": 892 }, { "epoch": 0.1, "learning_rate": 9.889007296728952e-06, "loss": 1.4954, "step": 893 }, { "epoch": 0.1, "learning_rate": 9.8886450402207e-06, "loss": 1.5809, "step": 894 }, { "epoch": 0.1, "learning_rate": 9.888282200168106e-06, "loss": 1.4674, "step": 895 }, { "epoch": 0.1, "learning_rate": 9.887918776614479e-06, "loss": 1.5484, "step": 896 }, { "epoch": 0.1, "learning_rate": 9.8875547696032e-06, "loss": 1.5477, "step": 897 }, { "epoch": 0.1, "learning_rate": 9.887190179177721e-06, "loss": 1.6058, "step": 898 }, { "epoch": 0.1, "learning_rate": 9.886825005381561e-06, "loss": 1.3197, "step": 899 }, { "epoch": 0.1, "learning_rate": 9.88645924825831e-06, "loss": 1.6469, "step": 900 }, { "epoch": 0.1, "learning_rate": 9.886092907851627e-06, "loss": 1.5409, "step": 901 }, { "epoch": 0.1, "learning_rate": 9.885725984205243e-06, "loss": 1.5474, "step": 902 }, { "epoch": 0.1, "learning_rate": 9.885358477362956e-06, "loss": 1.8208, "step": 903 }, { "epoch": 0.1, "learning_rate": 9.884990387368633e-06, "loss": 1.8261, "step": 904 }, { "epoch": 0.1, "learning_rate": 9.884621714266212e-06, "loss": 1.5157, "step": 905 }, { "epoch": 0.1, "learning_rate": 9.8842524580997e-06, "loss": 1.5028, "step": 906 }, { "epoch": 0.1, "learning_rate": 9.883882618913178e-06, "loss": 1.4364, "step": 907 }, { "epoch": 0.1, "learning_rate": 9.883512196750789e-06, "loss": 1.5583, "step": 908 }, { "epoch": 0.1, "learning_rate": 9.883141191656748e-06, "loss": 1.3261, "step": 909 }, { "epoch": 0.1, "learning_rate": 9.882769603675347e-06, "loss": 1.433, "step": 910 }, { "epoch": 0.1, "learning_rate": 9.882397432850935e-06, "loss": 1.5171, "step": 911 }, { "epoch": 0.1, "learning_rate": 9.88202467922794e-06, "loss": 1.2821, "step": 912 }, { "epoch": 0.1, "learning_rate": 9.881651342850856e-06, "loss": 1.6344, "step": 913 }, { "epoch": 0.1, "learning_rate": 9.881277423764246e-06, "loss": 1.5811, "step": 914 }, { "epoch": 0.1, "learning_rate": 9.880902922012747e-06, "loss": 1.2699, "step": 915 }, { "epoch": 0.1, "learning_rate": 9.88052783764106e-06, "loss": 1.7807, "step": 916 }, { "epoch": 0.1, "learning_rate": 9.880152170693957e-06, "loss": 1.4452, "step": 917 }, { "epoch": 0.1, "learning_rate": 9.879775921216284e-06, "loss": 1.5418, "step": 918 }, { "epoch": 0.1, "learning_rate": 9.879399089252947e-06, "loss": 1.58, "step": 919 }, { "epoch": 0.1, "learning_rate": 9.879021674848932e-06, "loss": 1.4377, "step": 920 }, { "epoch": 0.1, "learning_rate": 9.87864367804929e-06, "loss": 1.4353, "step": 921 }, { "epoch": 0.1, "learning_rate": 9.87826509889914e-06, "loss": 1.7167, "step": 922 }, { "epoch": 0.1, "learning_rate": 9.87788593744367e-06, "loss": 1.4849, "step": 923 }, { "epoch": 0.1, "learning_rate": 9.877506193728144e-06, "loss": 1.2443, "step": 924 }, { "epoch": 0.1, "learning_rate": 9.877125867797886e-06, "loss": 1.5532, "step": 925 }, { "epoch": 0.1, "learning_rate": 9.876744959698299e-06, "loss": 1.5213, "step": 926 }, { "epoch": 0.1, "learning_rate": 9.876363469474848e-06, "loss": 1.4388, "step": 927 }, { "epoch": 0.1, "learning_rate": 9.875981397173071e-06, "loss": 1.5789, "step": 928 }, { "epoch": 0.1, "learning_rate": 9.875598742838578e-06, "loss": 1.5887, "step": 929 }, { "epoch": 0.1, "learning_rate": 9.87521550651704e-06, "loss": 1.3239, "step": 930 }, { "epoch": 0.1, "learning_rate": 9.874831688254208e-06, "loss": 1.5111, "step": 931 }, { "epoch": 0.1, "learning_rate": 9.874447288095894e-06, "loss": 1.6793, "step": 932 }, { "epoch": 0.1, "learning_rate": 9.874062306087983e-06, "loss": 1.6046, "step": 933 }, { "epoch": 0.1, "learning_rate": 9.87367674227643e-06, "loss": 1.5828, "step": 934 }, { "epoch": 0.1, "learning_rate": 9.87329059670726e-06, "loss": 1.4788, "step": 935 }, { "epoch": 0.1, "learning_rate": 9.872903869426564e-06, "loss": 1.6457, "step": 936 }, { "epoch": 0.1, "learning_rate": 9.872516560480508e-06, "loss": 1.4053, "step": 937 }, { "epoch": 0.1, "learning_rate": 9.872128669915319e-06, "loss": 1.6271, "step": 938 }, { "epoch": 0.1, "learning_rate": 9.8717401977773e-06, "loss": 1.3505, "step": 939 }, { "epoch": 0.1, "learning_rate": 9.871351144112826e-06, "loss": 1.5512, "step": 940 }, { "epoch": 0.1, "learning_rate": 9.870961508968333e-06, "loss": 1.7935, "step": 941 }, { "epoch": 0.1, "learning_rate": 9.870571292390331e-06, "loss": 1.2952, "step": 942 }, { "epoch": 0.1, "learning_rate": 9.8701804944254e-06, "loss": 1.4274, "step": 943 }, { "epoch": 0.1, "learning_rate": 9.86978911512019e-06, "loss": 1.4955, "step": 944 }, { "epoch": 0.1, "learning_rate": 9.869397154521418e-06, "loss": 1.5935, "step": 945 }, { "epoch": 0.1, "learning_rate": 9.869004612675867e-06, "loss": 1.7871, "step": 946 }, { "epoch": 0.1, "learning_rate": 9.868611489630401e-06, "loss": 1.4748, "step": 947 }, { "epoch": 0.1, "learning_rate": 9.868217785431942e-06, "loss": 1.4944, "step": 948 }, { "epoch": 0.1, "learning_rate": 9.867823500127483e-06, "loss": 1.5665, "step": 949 }, { "epoch": 0.1, "learning_rate": 9.867428633764093e-06, "loss": 1.4588, "step": 950 }, { "epoch": 0.1, "learning_rate": 9.867033186388906e-06, "loss": 1.5993, "step": 951 }, { "epoch": 0.1, "learning_rate": 9.866637158049122e-06, "loss": 1.626, "step": 952 }, { "epoch": 0.1, "learning_rate": 9.866240548792018e-06, "loss": 1.7158, "step": 953 }, { "epoch": 0.1, "learning_rate": 9.865843358664933e-06, "loss": 1.5025, "step": 954 }, { "epoch": 0.1, "learning_rate": 9.86544558771528e-06, "loss": 1.3747, "step": 955 }, { "epoch": 0.1, "learning_rate": 9.86504723599054e-06, "loss": 1.4865, "step": 956 }, { "epoch": 0.1, "learning_rate": 9.86464830353826e-06, "loss": 1.6294, "step": 957 }, { "epoch": 0.1, "learning_rate": 9.864248790406063e-06, "loss": 1.587, "step": 958 }, { "epoch": 0.1, "learning_rate": 9.86384869664164e-06, "loss": 1.9286, "step": 959 }, { "epoch": 0.1, "learning_rate": 9.863448022292742e-06, "loss": 1.8419, "step": 960 }, { "epoch": 0.1, "learning_rate": 9.863046767407205e-06, "loss": 1.478, "step": 961 }, { "epoch": 0.1, "learning_rate": 9.862644932032918e-06, "loss": 1.7963, "step": 962 }, { "epoch": 0.1, "learning_rate": 9.86224251621785e-06, "loss": 1.4494, "step": 963 }, { "epoch": 0.1, "learning_rate": 9.861839520010038e-06, "loss": 1.4116, "step": 964 }, { "epoch": 0.1, "learning_rate": 9.861435943457585e-06, "loss": 1.5773, "step": 965 }, { "epoch": 0.1, "learning_rate": 9.861031786608663e-06, "loss": 1.738, "step": 966 }, { "epoch": 0.1, "learning_rate": 9.86062704951152e-06, "loss": 1.4718, "step": 967 }, { "epoch": 0.1, "learning_rate": 9.860221732214463e-06, "loss": 1.4762, "step": 968 }, { "epoch": 0.1, "learning_rate": 9.859815834765875e-06, "loss": 1.3964, "step": 969 }, { "epoch": 0.1, "learning_rate": 9.859409357214211e-06, "loss": 1.6363, "step": 970 }, { "epoch": 0.1, "learning_rate": 9.859002299607987e-06, "loss": 1.6485, "step": 971 }, { "epoch": 0.1, "learning_rate": 9.858594661995792e-06, "loss": 1.3257, "step": 972 }, { "epoch": 0.1, "learning_rate": 9.858186444426288e-06, "loss": 1.503, "step": 973 }, { "epoch": 0.1, "learning_rate": 9.8577776469482e-06, "loss": 1.6653, "step": 974 }, { "epoch": 0.1, "learning_rate": 9.857368269610325e-06, "loss": 1.676, "step": 975 }, { "epoch": 0.1, "learning_rate": 9.85695831246153e-06, "loss": 1.423, "step": 976 }, { "epoch": 0.1, "learning_rate": 9.856547775550752e-06, "loss": 1.2511, "step": 977 }, { "epoch": 0.1, "learning_rate": 9.856136658926993e-06, "loss": 1.3716, "step": 978 }, { "epoch": 0.1, "learning_rate": 9.85572496263933e-06, "loss": 1.5408, "step": 979 }, { "epoch": 0.1, "learning_rate": 9.855312686736902e-06, "loss": 1.3912, "step": 980 }, { "epoch": 0.1, "learning_rate": 9.854899831268926e-06, "loss": 1.5387, "step": 981 }, { "epoch": 0.1, "learning_rate": 9.854486396284678e-06, "loss": 1.6345, "step": 982 }, { "epoch": 0.1, "learning_rate": 9.854072381833512e-06, "loss": 1.3157, "step": 983 }, { "epoch": 0.1, "learning_rate": 9.85365778796485e-06, "loss": 1.3511, "step": 984 }, { "epoch": 0.11, "learning_rate": 9.853242614728175e-06, "loss": 1.9818, "step": 985 }, { "epoch": 0.11, "learning_rate": 9.852826862173052e-06, "loss": 1.587, "step": 986 }, { "epoch": 0.11, "learning_rate": 9.852410530349102e-06, "loss": 1.3118, "step": 987 }, { "epoch": 0.11, "learning_rate": 9.851993619306024e-06, "loss": 1.7487, "step": 988 }, { "epoch": 0.11, "learning_rate": 9.851576129093584e-06, "loss": 1.7461, "step": 989 }, { "epoch": 0.11, "learning_rate": 9.851158059761617e-06, "loss": 1.6462, "step": 990 }, { "epoch": 0.11, "learning_rate": 9.850739411360027e-06, "loss": 1.3172, "step": 991 }, { "epoch": 0.11, "learning_rate": 9.850320183938784e-06, "loss": 1.5284, "step": 992 }, { "epoch": 0.11, "learning_rate": 9.849900377547933e-06, "loss": 1.5618, "step": 993 }, { "epoch": 0.11, "learning_rate": 9.849479992237583e-06, "loss": 1.7357, "step": 994 }, { "epoch": 0.11, "learning_rate": 9.849059028057917e-06, "loss": 1.7013, "step": 995 }, { "epoch": 0.11, "learning_rate": 9.848637485059183e-06, "loss": 1.3564, "step": 996 }, { "epoch": 0.11, "learning_rate": 9.848215363291697e-06, "loss": 1.3724, "step": 997 }, { "epoch": 0.11, "learning_rate": 9.84779266280585e-06, "loss": 1.6383, "step": 998 }, { "epoch": 0.11, "learning_rate": 9.8473693836521e-06, "loss": 1.6367, "step": 999 }, { "epoch": 0.11, "learning_rate": 9.846945525880966e-06, "loss": 1.4584, "step": 1000 }, { "epoch": 0.11, "learning_rate": 9.84652108954305e-06, "loss": 1.5571, "step": 1001 }, { "epoch": 0.11, "learning_rate": 9.846096074689012e-06, "loss": 1.6042, "step": 1002 }, { "epoch": 0.11, "learning_rate": 9.845670481369585e-06, "loss": 1.486, "step": 1003 }, { "epoch": 0.11, "learning_rate": 9.845244309635571e-06, "loss": 1.4116, "step": 1004 }, { "epoch": 0.11, "learning_rate": 9.844817559537841e-06, "loss": 1.4912, "step": 1005 }, { "epoch": 0.11, "learning_rate": 9.844390231127337e-06, "loss": 1.5378, "step": 1006 }, { "epoch": 0.11, "learning_rate": 9.843962324455064e-06, "loss": 1.33, "step": 1007 }, { "epoch": 0.11, "learning_rate": 9.843533839572105e-06, "loss": 1.6593, "step": 1008 }, { "epoch": 0.11, "learning_rate": 9.843104776529606e-06, "loss": 1.8621, "step": 1009 }, { "epoch": 0.11, "learning_rate": 9.842675135378779e-06, "loss": 1.462, "step": 1010 }, { "epoch": 0.11, "learning_rate": 9.842244916170913e-06, "loss": 1.5831, "step": 1011 }, { "epoch": 0.11, "learning_rate": 9.84181411895736e-06, "loss": 1.4095, "step": 1012 }, { "epoch": 0.11, "learning_rate": 9.841382743789544e-06, "loss": 1.7082, "step": 1013 }, { "epoch": 0.11, "learning_rate": 9.840950790718959e-06, "loss": 1.405, "step": 1014 }, { "epoch": 0.11, "learning_rate": 9.840518259797163e-06, "loss": 1.8219, "step": 1015 }, { "epoch": 0.11, "learning_rate": 9.840085151075787e-06, "loss": 1.3648, "step": 1016 }, { "epoch": 0.11, "learning_rate": 9.83965146460653e-06, "loss": 1.4104, "step": 1017 }, { "epoch": 0.11, "learning_rate": 9.83921720044116e-06, "loss": 1.4081, "step": 1018 }, { "epoch": 0.11, "learning_rate": 9.838782358631516e-06, "loss": 1.6887, "step": 1019 }, { "epoch": 0.11, "learning_rate": 9.838346939229501e-06, "loss": 1.4117, "step": 1020 }, { "epoch": 0.11, "learning_rate": 9.837910942287091e-06, "loss": 1.2456, "step": 1021 }, { "epoch": 0.11, "learning_rate": 9.83747436785633e-06, "loss": 1.4869, "step": 1022 }, { "epoch": 0.11, "learning_rate": 9.83703721598933e-06, "loss": 1.7421, "step": 1023 }, { "epoch": 0.11, "learning_rate": 9.836599486738271e-06, "loss": 1.2656, "step": 1024 }, { "epoch": 0.11, "learning_rate": 9.83616118015541e-06, "loss": 1.3702, "step": 1025 }, { "epoch": 0.11, "learning_rate": 9.835722296293058e-06, "loss": 1.7946, "step": 1026 }, { "epoch": 0.11, "learning_rate": 9.83528283520361e-06, "loss": 1.5679, "step": 1027 }, { "epoch": 0.11, "learning_rate": 9.83484279693952e-06, "loss": 1.6035, "step": 1028 }, { "epoch": 0.11, "learning_rate": 9.834402181553314e-06, "loss": 1.4071, "step": 1029 }, { "epoch": 0.11, "learning_rate": 9.83396098909759e-06, "loss": 1.501, "step": 1030 }, { "epoch": 0.11, "learning_rate": 9.833519219625008e-06, "loss": 1.5163, "step": 1031 }, { "epoch": 0.11, "learning_rate": 9.833076873188303e-06, "loss": 1.5632, "step": 1032 }, { "epoch": 0.11, "learning_rate": 9.832633949840277e-06, "loss": 1.8133, "step": 1033 }, { "epoch": 0.11, "learning_rate": 9.832190449633801e-06, "loss": 1.6983, "step": 1034 }, { "epoch": 0.11, "learning_rate": 9.831746372621811e-06, "loss": 1.5021, "step": 1035 }, { "epoch": 0.11, "learning_rate": 9.83130171885732e-06, "loss": 1.6188, "step": 1036 }, { "epoch": 0.11, "learning_rate": 9.830856488393401e-06, "loss": 1.7335, "step": 1037 }, { "epoch": 0.11, "learning_rate": 9.830410681283203e-06, "loss": 1.6512, "step": 1038 }, { "epoch": 0.11, "learning_rate": 9.82996429757994e-06, "loss": 1.7334, "step": 1039 }, { "epoch": 0.11, "learning_rate": 9.829517337336893e-06, "loss": 1.5233, "step": 1040 }, { "epoch": 0.11, "learning_rate": 9.829069800607418e-06, "loss": 1.5783, "step": 1041 }, { "epoch": 0.11, "learning_rate": 9.828621687444935e-06, "loss": 1.5447, "step": 1042 }, { "epoch": 0.11, "learning_rate": 9.828172997902934e-06, "loss": 1.6315, "step": 1043 }, { "epoch": 0.11, "learning_rate": 9.827723732034972e-06, "loss": 1.7712, "step": 1044 }, { "epoch": 0.11, "learning_rate": 9.82727388989468e-06, "loss": 1.4402, "step": 1045 }, { "epoch": 0.11, "learning_rate": 9.826823471535754e-06, "loss": 1.5074, "step": 1046 }, { "epoch": 0.11, "learning_rate": 9.826372477011956e-06, "loss": 1.4142, "step": 1047 }, { "epoch": 0.11, "learning_rate": 9.825920906377124e-06, "loss": 1.6354, "step": 1048 }, { "epoch": 0.11, "learning_rate": 9.825468759685157e-06, "loss": 1.4593, "step": 1049 }, { "epoch": 0.11, "learning_rate": 9.825016036990029e-06, "loss": 1.2347, "step": 1050 }, { "epoch": 0.11, "learning_rate": 9.824562738345781e-06, "loss": 1.4779, "step": 1051 }, { "epoch": 0.11, "learning_rate": 9.82410886380652e-06, "loss": 1.6353, "step": 1052 }, { "epoch": 0.11, "learning_rate": 9.823654413426424e-06, "loss": 1.5191, "step": 1053 }, { "epoch": 0.11, "learning_rate": 9.82319938725974e-06, "loss": 1.6205, "step": 1054 }, { "epoch": 0.11, "learning_rate": 9.822743785360783e-06, "loss": 1.5898, "step": 1055 }, { "epoch": 0.11, "learning_rate": 9.822287607783938e-06, "loss": 1.7052, "step": 1056 }, { "epoch": 0.11, "learning_rate": 9.821830854583657e-06, "loss": 1.5637, "step": 1057 }, { "epoch": 0.11, "learning_rate": 9.82137352581446e-06, "loss": 1.485, "step": 1058 }, { "epoch": 0.11, "learning_rate": 9.820915621530939e-06, "loss": 1.6526, "step": 1059 }, { "epoch": 0.11, "learning_rate": 9.820457141787753e-06, "loss": 1.2195, "step": 1060 }, { "epoch": 0.11, "learning_rate": 9.819998086639628e-06, "loss": 1.4496, "step": 1061 }, { "epoch": 0.11, "learning_rate": 9.81953845614136e-06, "loss": 1.7976, "step": 1062 }, { "epoch": 0.11, "learning_rate": 9.819078250347817e-06, "loss": 1.5921, "step": 1063 }, { "epoch": 0.11, "learning_rate": 9.81861746931393e-06, "loss": 1.4404, "step": 1064 }, { "epoch": 0.11, "learning_rate": 9.818156113094699e-06, "loss": 1.7096, "step": 1065 }, { "epoch": 0.11, "learning_rate": 9.8176941817452e-06, "loss": 1.5023, "step": 1066 }, { "epoch": 0.11, "learning_rate": 9.817231675320566e-06, "loss": 1.5968, "step": 1067 }, { "epoch": 0.11, "learning_rate": 9.816768593876012e-06, "loss": 1.6217, "step": 1068 }, { "epoch": 0.11, "learning_rate": 9.81630493746681e-06, "loss": 1.6108, "step": 1069 }, { "epoch": 0.11, "learning_rate": 9.815840706148308e-06, "loss": 1.693, "step": 1070 }, { "epoch": 0.11, "learning_rate": 9.81537589997592e-06, "loss": 1.6316, "step": 1071 }, { "epoch": 0.11, "learning_rate": 9.814910519005126e-06, "loss": 1.4328, "step": 1072 }, { "epoch": 0.11, "learning_rate": 9.814444563291478e-06, "loss": 1.6463, "step": 1073 }, { "epoch": 0.11, "learning_rate": 9.8139780328906e-06, "loss": 1.3296, "step": 1074 }, { "epoch": 0.11, "learning_rate": 9.813510927858177e-06, "loss": 1.8508, "step": 1075 }, { "epoch": 0.11, "learning_rate": 9.813043248249965e-06, "loss": 1.499, "step": 1076 }, { "epoch": 0.11, "learning_rate": 9.812574994121791e-06, "loss": 1.7064, "step": 1077 }, { "epoch": 0.11, "learning_rate": 9.81210616552955e-06, "loss": 1.7453, "step": 1078 }, { "epoch": 0.12, "learning_rate": 9.811636762529205e-06, "loss": 1.6748, "step": 1079 }, { "epoch": 0.12, "learning_rate": 9.811166785176785e-06, "loss": 1.4982, "step": 1080 }, { "epoch": 0.12, "learning_rate": 9.810696233528391e-06, "loss": 1.31, "step": 1081 }, { "epoch": 0.12, "learning_rate": 9.810225107640195e-06, "loss": 1.4754, "step": 1082 }, { "epoch": 0.12, "learning_rate": 9.809753407568427e-06, "loss": 1.8735, "step": 1083 }, { "epoch": 0.12, "learning_rate": 9.809281133369399e-06, "loss": 1.6927, "step": 1084 }, { "epoch": 0.12, "learning_rate": 9.808808285099483e-06, "loss": 1.3425, "step": 1085 }, { "epoch": 0.12, "learning_rate": 9.80833486281512e-06, "loss": 1.6511, "step": 1086 }, { "epoch": 0.12, "learning_rate": 9.807860866572822e-06, "loss": 1.524, "step": 1087 }, { "epoch": 0.12, "learning_rate": 9.807386296429168e-06, "loss": 1.5515, "step": 1088 }, { "epoch": 0.12, "learning_rate": 9.80691115244081e-06, "loss": 1.5877, "step": 1089 }, { "epoch": 0.12, "learning_rate": 9.806435434664461e-06, "loss": 1.6653, "step": 1090 }, { "epoch": 0.12, "learning_rate": 9.805959143156905e-06, "loss": 1.5368, "step": 1091 }, { "epoch": 0.12, "learning_rate": 9.805482277974999e-06, "loss": 1.4348, "step": 1092 }, { "epoch": 0.12, "learning_rate": 9.805004839175664e-06, "loss": 1.677, "step": 1093 }, { "epoch": 0.12, "learning_rate": 9.80452682681589e-06, "loss": 1.5777, "step": 1094 }, { "epoch": 0.12, "learning_rate": 9.804048240952736e-06, "loss": 1.57, "step": 1095 }, { "epoch": 0.12, "learning_rate": 9.80356908164333e-06, "loss": 1.5979, "step": 1096 }, { "epoch": 0.12, "learning_rate": 9.803089348944868e-06, "loss": 1.7019, "step": 1097 }, { "epoch": 0.12, "learning_rate": 9.802609042914614e-06, "loss": 1.865, "step": 1098 }, { "epoch": 0.12, "learning_rate": 9.802128163609901e-06, "loss": 1.6943, "step": 1099 }, { "epoch": 0.12, "learning_rate": 9.80164671108813e-06, "loss": 1.6835, "step": 1100 }, { "epoch": 0.12, "learning_rate": 9.80116468540677e-06, "loss": 1.666, "step": 1101 }, { "epoch": 0.12, "learning_rate": 9.800682086623363e-06, "loss": 1.4122, "step": 1102 }, { "epoch": 0.12, "learning_rate": 9.80019891479551e-06, "loss": 1.5566, "step": 1103 }, { "epoch": 0.12, "learning_rate": 9.79971516998089e-06, "loss": 1.7001, "step": 1104 }, { "epoch": 0.12, "learning_rate": 9.799230852237243e-06, "loss": 1.4241, "step": 1105 }, { "epoch": 0.12, "learning_rate": 9.798745961622383e-06, "loss": 1.6297, "step": 1106 }, { "epoch": 0.12, "learning_rate": 9.79826049819419e-06, "loss": 1.4225, "step": 1107 }, { "epoch": 0.12, "learning_rate": 9.797774462010611e-06, "loss": 1.3983, "step": 1108 }, { "epoch": 0.12, "learning_rate": 9.797287853129666e-06, "loss": 1.5537, "step": 1109 }, { "epoch": 0.12, "learning_rate": 9.796800671609436e-06, "loss": 1.5413, "step": 1110 }, { "epoch": 0.12, "learning_rate": 9.796312917508078e-06, "loss": 1.631, "step": 1111 }, { "epoch": 0.12, "learning_rate": 9.795824590883812e-06, "loss": 1.3628, "step": 1112 }, { "epoch": 0.12, "learning_rate": 9.795335691794929e-06, "loss": 1.5239, "step": 1113 }, { "epoch": 0.12, "learning_rate": 9.794846220299787e-06, "loss": 1.5177, "step": 1114 }, { "epoch": 0.12, "learning_rate": 9.794356176456813e-06, "loss": 1.6588, "step": 1115 }, { "epoch": 0.12, "learning_rate": 9.793865560324503e-06, "loss": 1.5369, "step": 1116 }, { "epoch": 0.12, "learning_rate": 9.793374371961418e-06, "loss": 1.6676, "step": 1117 }, { "epoch": 0.12, "learning_rate": 9.792882611426193e-06, "loss": 1.4977, "step": 1118 }, { "epoch": 0.12, "learning_rate": 9.792390278777527e-06, "loss": 1.6037, "step": 1119 }, { "epoch": 0.12, "learning_rate": 9.791897374074188e-06, "loss": 1.6793, "step": 1120 }, { "epoch": 0.12, "learning_rate": 9.791403897375013e-06, "loss": 1.7076, "step": 1121 }, { "epoch": 0.12, "learning_rate": 9.790909848738907e-06, "loss": 1.4538, "step": 1122 }, { "epoch": 0.12, "learning_rate": 9.790415228224843e-06, "loss": 1.6519, "step": 1123 }, { "epoch": 0.12, "learning_rate": 9.789920035891863e-06, "loss": 1.6149, "step": 1124 }, { "epoch": 0.12, "learning_rate": 9.789424271799075e-06, "loss": 1.724, "step": 1125 }, { "epoch": 0.12, "learning_rate": 9.78892793600566e-06, "loss": 1.694, "step": 1126 }, { "epoch": 0.12, "learning_rate": 9.788431028570861e-06, "loss": 1.4678, "step": 1127 }, { "epoch": 0.12, "learning_rate": 9.787933549553996e-06, "loss": 1.5773, "step": 1128 }, { "epoch": 0.12, "learning_rate": 9.787435499014446e-06, "loss": 1.4462, "step": 1129 }, { "epoch": 0.12, "learning_rate": 9.786936877011662e-06, "loss": 1.7461, "step": 1130 }, { "epoch": 0.12, "learning_rate": 9.786437683605161e-06, "loss": 1.6068, "step": 1131 }, { "epoch": 0.12, "learning_rate": 9.785937918854536e-06, "loss": 1.2814, "step": 1132 }, { "epoch": 0.12, "learning_rate": 9.785437582819436e-06, "loss": 1.6243, "step": 1133 }, { "epoch": 0.12, "learning_rate": 9.78493667555959e-06, "loss": 1.6756, "step": 1134 }, { "epoch": 0.12, "learning_rate": 9.784435197134785e-06, "loss": 1.647, "step": 1135 }, { "epoch": 0.12, "learning_rate": 9.783933147604885e-06, "loss": 1.9122, "step": 1136 }, { "epoch": 0.12, "learning_rate": 9.783430527029818e-06, "loss": 1.5544, "step": 1137 }, { "epoch": 0.12, "learning_rate": 9.782927335469579e-06, "loss": 1.5141, "step": 1138 }, { "epoch": 0.12, "learning_rate": 9.782423572984234e-06, "loss": 1.7291, "step": 1139 }, { "epoch": 0.12, "learning_rate": 9.781919239633912e-06, "loss": 1.8511, "step": 1140 }, { "epoch": 0.12, "learning_rate": 9.781414335478821e-06, "loss": 1.3221, "step": 1141 }, { "epoch": 0.12, "learning_rate": 9.780908860579223e-06, "loss": 1.4027, "step": 1142 }, { "epoch": 0.12, "learning_rate": 9.780402814995458e-06, "loss": 1.4411, "step": 1143 }, { "epoch": 0.12, "learning_rate": 9.779896198787933e-06, "loss": 1.6798, "step": 1144 }, { "epoch": 0.12, "learning_rate": 9.77938901201712e-06, "loss": 1.5651, "step": 1145 }, { "epoch": 0.12, "learning_rate": 9.77888125474356e-06, "loss": 1.4732, "step": 1146 }, { "epoch": 0.12, "learning_rate": 9.778372927027861e-06, "loss": 1.9788, "step": 1147 }, { "epoch": 0.12, "learning_rate": 9.777864028930705e-06, "loss": 1.617, "step": 1148 }, { "epoch": 0.12, "learning_rate": 9.777354560512835e-06, "loss": 1.6138, "step": 1149 }, { "epoch": 0.12, "learning_rate": 9.776844521835064e-06, "loss": 1.5935, "step": 1150 }, { "epoch": 0.12, "learning_rate": 9.776333912958276e-06, "loss": 1.4657, "step": 1151 }, { "epoch": 0.12, "learning_rate": 9.77582273394342e-06, "loss": 1.4813, "step": 1152 }, { "epoch": 0.12, "learning_rate": 9.775310984851513e-06, "loss": 1.5062, "step": 1153 }, { "epoch": 0.12, "learning_rate": 9.774798665743646e-06, "loss": 1.3368, "step": 1154 }, { "epoch": 0.12, "learning_rate": 9.774285776680967e-06, "loss": 1.5665, "step": 1155 }, { "epoch": 0.12, "learning_rate": 9.7737723177247e-06, "loss": 1.5256, "step": 1156 }, { "epoch": 0.12, "learning_rate": 9.773258288936139e-06, "loss": 1.7638, "step": 1157 }, { "epoch": 0.12, "learning_rate": 9.772743690376636e-06, "loss": 1.3751, "step": 1158 }, { "epoch": 0.12, "learning_rate": 9.77222852210762e-06, "loss": 1.3326, "step": 1159 }, { "epoch": 0.12, "learning_rate": 9.771712784190588e-06, "loss": 1.7034, "step": 1160 }, { "epoch": 0.12, "learning_rate": 9.7711964766871e-06, "loss": 1.5013, "step": 1161 }, { "epoch": 0.12, "learning_rate": 9.770679599658786e-06, "loss": 1.776, "step": 1162 }, { "epoch": 0.12, "learning_rate": 9.770162153167343e-06, "loss": 1.4468, "step": 1163 }, { "epoch": 0.12, "learning_rate": 9.76964413727454e-06, "loss": 1.9054, "step": 1164 }, { "epoch": 0.12, "learning_rate": 9.769125552042207e-06, "loss": 1.6743, "step": 1165 }, { "epoch": 0.12, "learning_rate": 9.76860639753225e-06, "loss": 1.6033, "step": 1166 }, { "epoch": 0.12, "learning_rate": 9.768086673806638e-06, "loss": 1.5318, "step": 1167 }, { "epoch": 0.12, "learning_rate": 9.76756638092741e-06, "loss": 1.4437, "step": 1168 }, { "epoch": 0.12, "learning_rate": 9.767045518956671e-06, "loss": 1.5011, "step": 1169 }, { "epoch": 0.12, "learning_rate": 9.766524087956592e-06, "loss": 1.5024, "step": 1170 }, { "epoch": 0.12, "learning_rate": 9.76600208798942e-06, "loss": 1.3771, "step": 1171 }, { "epoch": 0.13, "learning_rate": 9.765479519117461e-06, "loss": 1.6863, "step": 1172 }, { "epoch": 0.13, "learning_rate": 9.764956381403095e-06, "loss": 1.418, "step": 1173 }, { "epoch": 0.13, "learning_rate": 9.764432674908766e-06, "loss": 1.4432, "step": 1174 }, { "epoch": 0.13, "learning_rate": 9.763908399696986e-06, "loss": 1.6386, "step": 1175 }, { "epoch": 0.13, "learning_rate": 9.76338355583034e-06, "loss": 1.175, "step": 1176 }, { "epoch": 0.13, "learning_rate": 9.762858143371476e-06, "loss": 1.6609, "step": 1177 }, { "epoch": 0.13, "learning_rate": 9.76233216238311e-06, "loss": 1.6583, "step": 1178 }, { "epoch": 0.13, "learning_rate": 9.761805612928025e-06, "loss": 1.5003, "step": 1179 }, { "epoch": 0.13, "learning_rate": 9.76127849506908e-06, "loss": 1.3507, "step": 1180 }, { "epoch": 0.13, "learning_rate": 9.760750808869188e-06, "loss": 1.4676, "step": 1181 }, { "epoch": 0.13, "learning_rate": 9.760222554391343e-06, "loss": 1.4871, "step": 1182 }, { "epoch": 0.13, "learning_rate": 9.7596937316986e-06, "loss": 1.4694, "step": 1183 }, { "epoch": 0.13, "learning_rate": 9.759164340854082e-06, "loss": 1.3431, "step": 1184 }, { "epoch": 0.13, "learning_rate": 9.758634381920982e-06, "loss": 1.3513, "step": 1185 }, { "epoch": 0.13, "learning_rate": 9.75810385496256e-06, "loss": 1.6224, "step": 1186 }, { "epoch": 0.13, "learning_rate": 9.757572760042141e-06, "loss": 1.3375, "step": 1187 }, { "epoch": 0.13, "learning_rate": 9.757041097223123e-06, "loss": 1.8991, "step": 1188 }, { "epoch": 0.13, "learning_rate": 9.75650886656897e-06, "loss": 1.2338, "step": 1189 }, { "epoch": 0.13, "learning_rate": 9.75597606814321e-06, "loss": 1.5093, "step": 1190 }, { "epoch": 0.13, "learning_rate": 9.755442702009443e-06, "loss": 1.4194, "step": 1191 }, { "epoch": 0.13, "learning_rate": 9.754908768231337e-06, "loss": 1.6269, "step": 1192 }, { "epoch": 0.13, "learning_rate": 9.754374266872624e-06, "loss": 1.7046, "step": 1193 }, { "epoch": 0.13, "learning_rate": 9.753839197997105e-06, "loss": 1.2588, "step": 1194 }, { "epoch": 0.13, "learning_rate": 9.753303561668654e-06, "loss": 1.5248, "step": 1195 }, { "epoch": 0.13, "learning_rate": 9.752767357951206e-06, "loss": 1.4379, "step": 1196 }, { "epoch": 0.13, "learning_rate": 9.752230586908767e-06, "loss": 1.5639, "step": 1197 }, { "epoch": 0.13, "learning_rate": 9.751693248605406e-06, "loss": 1.5892, "step": 1198 }, { "epoch": 0.13, "learning_rate": 9.751155343105269e-06, "loss": 1.3139, "step": 1199 }, { "epoch": 0.13, "learning_rate": 9.75061687047256e-06, "loss": 1.4903, "step": 1200 }, { "epoch": 0.13, "learning_rate": 9.75007783077156e-06, "loss": 1.4946, "step": 1201 }, { "epoch": 0.13, "learning_rate": 9.749538224066607e-06, "loss": 1.6313, "step": 1202 }, { "epoch": 0.13, "learning_rate": 9.748998050422117e-06, "loss": 1.3187, "step": 1203 }, { "epoch": 0.13, "learning_rate": 9.748457309902566e-06, "loss": 1.5483, "step": 1204 }, { "epoch": 0.13, "learning_rate": 9.747916002572502e-06, "loss": 1.3913, "step": 1205 }, { "epoch": 0.13, "learning_rate": 9.747374128496541e-06, "loss": 1.5954, "step": 1206 }, { "epoch": 0.13, "learning_rate": 9.746831687739361e-06, "loss": 1.3951, "step": 1207 }, { "epoch": 0.13, "learning_rate": 9.746288680365716e-06, "loss": 1.393, "step": 1208 }, { "epoch": 0.13, "learning_rate": 9.745745106440422e-06, "loss": 1.476, "step": 1209 }, { "epoch": 0.13, "learning_rate": 9.745200966028362e-06, "loss": 1.4056, "step": 1210 }, { "epoch": 0.13, "learning_rate": 9.74465625919449e-06, "loss": 1.5048, "step": 1211 }, { "epoch": 0.13, "learning_rate": 9.744110986003826e-06, "loss": 1.5687, "step": 1212 }, { "epoch": 0.13, "learning_rate": 9.743565146521459e-06, "loss": 1.6521, "step": 1213 }, { "epoch": 0.13, "learning_rate": 9.743018740812541e-06, "loss": 1.4184, "step": 1214 }, { "epoch": 0.13, "learning_rate": 9.742471768942299e-06, "loss": 1.6139, "step": 1215 }, { "epoch": 0.13, "learning_rate": 9.741924230976023e-06, "loss": 1.4703, "step": 1216 }, { "epoch": 0.13, "learning_rate": 9.741376126979069e-06, "loss": 1.5833, "step": 1217 }, { "epoch": 0.13, "learning_rate": 9.740827457016863e-06, "loss": 1.6699, "step": 1218 }, { "epoch": 0.13, "learning_rate": 9.740278221154899e-06, "loss": 1.3659, "step": 1219 }, { "epoch": 0.13, "learning_rate": 9.739728419458738e-06, "loss": 1.4001, "step": 1220 }, { "epoch": 0.13, "learning_rate": 9.739178051994008e-06, "loss": 1.6821, "step": 1221 }, { "epoch": 0.13, "learning_rate": 9.738627118826404e-06, "loss": 1.6674, "step": 1222 }, { "epoch": 0.13, "learning_rate": 9.738075620021691e-06, "loss": 1.4935, "step": 1223 }, { "epoch": 0.13, "learning_rate": 9.7375235556457e-06, "loss": 1.5973, "step": 1224 }, { "epoch": 0.13, "learning_rate": 9.736970925764326e-06, "loss": 1.558, "step": 1225 }, { "epoch": 0.13, "learning_rate": 9.73641773044354e-06, "loss": 1.54, "step": 1226 }, { "epoch": 0.13, "learning_rate": 9.735863969749373e-06, "loss": 1.7119, "step": 1227 }, { "epoch": 0.13, "learning_rate": 9.735309643747926e-06, "loss": 1.7234, "step": 1228 }, { "epoch": 0.13, "learning_rate": 9.734754752505366e-06, "loss": 1.4574, "step": 1229 }, { "epoch": 0.13, "learning_rate": 9.734199296087932e-06, "loss": 1.2406, "step": 1230 }, { "epoch": 0.13, "learning_rate": 9.733643274561924e-06, "loss": 1.6091, "step": 1231 }, { "epoch": 0.13, "learning_rate": 9.733086687993714e-06, "loss": 1.9197, "step": 1232 }, { "epoch": 0.13, "learning_rate": 9.732529536449741e-06, "loss": 1.7005, "step": 1233 }, { "epoch": 0.13, "learning_rate": 9.731971819996513e-06, "loss": 1.5906, "step": 1234 }, { "epoch": 0.13, "learning_rate": 9.731413538700597e-06, "loss": 1.3876, "step": 1235 }, { "epoch": 0.13, "learning_rate": 9.730854692628637e-06, "loss": 1.6237, "step": 1236 }, { "epoch": 0.13, "learning_rate": 9.730295281847342e-06, "loss": 1.4871, "step": 1237 }, { "epoch": 0.13, "learning_rate": 9.729735306423486e-06, "loss": 1.3864, "step": 1238 }, { "epoch": 0.13, "learning_rate": 9.729174766423912e-06, "loss": 1.4551, "step": 1239 }, { "epoch": 0.13, "learning_rate": 9.72861366191553e-06, "loss": 1.331, "step": 1240 }, { "epoch": 0.13, "learning_rate": 9.728051992965316e-06, "loss": 1.5923, "step": 1241 }, { "epoch": 0.13, "learning_rate": 9.72748975964032e-06, "loss": 1.421, "step": 1242 }, { "epoch": 0.13, "learning_rate": 9.726926962007647e-06, "loss": 1.5634, "step": 1243 }, { "epoch": 0.13, "learning_rate": 9.726363600134482e-06, "loss": 1.3268, "step": 1244 }, { "epoch": 0.13, "learning_rate": 9.725799674088072e-06, "loss": 1.5547, "step": 1245 }, { "epoch": 0.13, "learning_rate": 9.725235183935729e-06, "loss": 1.6014, "step": 1246 }, { "epoch": 0.13, "learning_rate": 9.724670129744834e-06, "loss": 1.386, "step": 1247 }, { "epoch": 0.13, "learning_rate": 9.724104511582838e-06, "loss": 1.4995, "step": 1248 }, { "epoch": 0.13, "learning_rate": 9.723538329517257e-06, "loss": 1.5724, "step": 1249 }, { "epoch": 0.13, "learning_rate": 9.722971583615674e-06, "loss": 1.4993, "step": 1250 }, { "epoch": 0.13, "learning_rate": 9.72240427394574e-06, "loss": 1.4156, "step": 1251 }, { "epoch": 0.13, "learning_rate": 9.721836400575173e-06, "loss": 1.3992, "step": 1252 }, { "epoch": 0.13, "learning_rate": 9.72126796357176e-06, "loss": 1.7882, "step": 1253 }, { "epoch": 0.13, "learning_rate": 9.720698963003351e-06, "loss": 1.6251, "step": 1254 }, { "epoch": 0.13, "learning_rate": 9.720129398937871e-06, "loss": 1.3852, "step": 1255 }, { "epoch": 0.13, "learning_rate": 9.719559271443303e-06, "loss": 1.6301, "step": 1256 }, { "epoch": 0.13, "learning_rate": 9.7189885805877e-06, "loss": 1.4554, "step": 1257 }, { "epoch": 0.13, "learning_rate": 9.71841732643919e-06, "loss": 1.6947, "step": 1258 }, { "epoch": 0.13, "learning_rate": 9.717845509065958e-06, "loss": 1.4979, "step": 1259 }, { "epoch": 0.13, "learning_rate": 9.717273128536259e-06, "loss": 1.3169, "step": 1260 }, { "epoch": 0.13, "learning_rate": 9.716700184918419e-06, "loss": 1.5889, "step": 1261 }, { "epoch": 0.13, "learning_rate": 9.716126678280829e-06, "loss": 1.844, "step": 1262 }, { "epoch": 0.13, "learning_rate": 9.715552608691944e-06, "loss": 1.4674, "step": 1263 }, { "epoch": 0.13, "learning_rate": 9.714977976220295e-06, "loss": 1.654, "step": 1264 }, { "epoch": 0.13, "learning_rate": 9.714402780934467e-06, "loss": 1.6888, "step": 1265 }, { "epoch": 0.14, "learning_rate": 9.713827022903124e-06, "loss": 1.6122, "step": 1266 }, { "epoch": 0.14, "learning_rate": 9.713250702194993e-06, "loss": 1.5645, "step": 1267 }, { "epoch": 0.14, "learning_rate": 9.712673818878867e-06, "loss": 1.8917, "step": 1268 }, { "epoch": 0.14, "learning_rate": 9.712096373023603e-06, "loss": 1.5528, "step": 1269 }, { "epoch": 0.14, "learning_rate": 9.711518364698136e-06, "loss": 1.805, "step": 1270 }, { "epoch": 0.14, "learning_rate": 9.710939793971456e-06, "loss": 1.5662, "step": 1271 }, { "epoch": 0.14, "learning_rate": 9.710360660912629e-06, "loss": 1.5435, "step": 1272 }, { "epoch": 0.14, "learning_rate": 9.709780965590782e-06, "loss": 1.4428, "step": 1273 }, { "epoch": 0.14, "learning_rate": 9.709200708075114e-06, "loss": 1.4652, "step": 1274 }, { "epoch": 0.14, "learning_rate": 9.708619888434887e-06, "loss": 1.5998, "step": 1275 }, { "epoch": 0.14, "learning_rate": 9.70803850673943e-06, "loss": 1.6374, "step": 1276 }, { "epoch": 0.14, "learning_rate": 9.707456563058146e-06, "loss": 1.3437, "step": 1277 }, { "epoch": 0.14, "learning_rate": 9.706874057460497e-06, "loss": 1.4756, "step": 1278 }, { "epoch": 0.14, "learning_rate": 9.706290990016013e-06, "loss": 1.4615, "step": 1279 }, { "epoch": 0.14, "learning_rate": 9.705707360794299e-06, "loss": 1.5123, "step": 1280 }, { "epoch": 0.14, "learning_rate": 9.705123169865016e-06, "loss": 1.3048, "step": 1281 }, { "epoch": 0.14, "learning_rate": 9.704538417297899e-06, "loss": 1.3965, "step": 1282 }, { "epoch": 0.14, "learning_rate": 9.703953103162748e-06, "loss": 1.4856, "step": 1283 }, { "epoch": 0.14, "learning_rate": 9.703367227529432e-06, "loss": 1.5561, "step": 1284 }, { "epoch": 0.14, "learning_rate": 9.702780790467884e-06, "loss": 1.7939, "step": 1285 }, { "epoch": 0.14, "learning_rate": 9.702193792048104e-06, "loss": 1.4319, "step": 1286 }, { "epoch": 0.14, "learning_rate": 9.701606232340165e-06, "loss": 1.4015, "step": 1287 }, { "epoch": 0.14, "learning_rate": 9.701018111414198e-06, "loss": 1.4337, "step": 1288 }, { "epoch": 0.14, "learning_rate": 9.700429429340407e-06, "loss": 1.5142, "step": 1289 }, { "epoch": 0.14, "learning_rate": 9.699840186189061e-06, "loss": 1.5036, "step": 1290 }, { "epoch": 0.14, "learning_rate": 9.699250382030496e-06, "loss": 1.4903, "step": 1291 }, { "epoch": 0.14, "learning_rate": 9.698660016935117e-06, "loss": 1.3797, "step": 1292 }, { "epoch": 0.14, "learning_rate": 9.698069090973391e-06, "loss": 1.5697, "step": 1293 }, { "epoch": 0.14, "learning_rate": 9.69747760421586e-06, "loss": 1.6384, "step": 1294 }, { "epoch": 0.14, "learning_rate": 9.696885556733126e-06, "loss": 1.6074, "step": 1295 }, { "epoch": 0.14, "learning_rate": 9.696292948595857e-06, "loss": 1.523, "step": 1296 }, { "epoch": 0.14, "learning_rate": 9.695699779874796e-06, "loss": 1.9136, "step": 1297 }, { "epoch": 0.14, "learning_rate": 9.695106050640745e-06, "loss": 1.446, "step": 1298 }, { "epoch": 0.14, "learning_rate": 9.694511760964578e-06, "loss": 1.5702, "step": 1299 }, { "epoch": 0.14, "learning_rate": 9.69391691091723e-06, "loss": 1.5894, "step": 1300 }, { "epoch": 0.14, "learning_rate": 9.69332150056971e-06, "loss": 1.5456, "step": 1301 }, { "epoch": 0.14, "learning_rate": 9.69272552999309e-06, "loss": 1.597, "step": 1302 }, { "epoch": 0.14, "learning_rate": 9.692128999258508e-06, "loss": 1.6992, "step": 1303 }, { "epoch": 0.14, "learning_rate": 9.691531908437171e-06, "loss": 1.4871, "step": 1304 }, { "epoch": 0.14, "learning_rate": 9.690934257600353e-06, "loss": 1.2551, "step": 1305 }, { "epoch": 0.14, "learning_rate": 9.690336046819393e-06, "loss": 1.4431, "step": 1306 }, { "epoch": 0.14, "learning_rate": 9.689737276165698e-06, "loss": 1.3173, "step": 1307 }, { "epoch": 0.14, "learning_rate": 9.689137945710742e-06, "loss": 1.1122, "step": 1308 }, { "epoch": 0.14, "learning_rate": 9.688538055526065e-06, "loss": 1.5224, "step": 1309 }, { "epoch": 0.14, "learning_rate": 9.687937605683274e-06, "loss": 1.485, "step": 1310 }, { "epoch": 0.14, "learning_rate": 9.687336596254045e-06, "loss": 1.6672, "step": 1311 }, { "epoch": 0.14, "learning_rate": 9.686735027310115e-06, "loss": 1.3189, "step": 1312 }, { "epoch": 0.14, "learning_rate": 9.686132898923295e-06, "loss": 1.7376, "step": 1313 }, { "epoch": 0.14, "learning_rate": 9.685530211165459e-06, "loss": 1.6984, "step": 1314 }, { "epoch": 0.14, "learning_rate": 9.684926964108546e-06, "loss": 1.5419, "step": 1315 }, { "epoch": 0.14, "learning_rate": 9.684323157824567e-06, "loss": 1.3849, "step": 1316 }, { "epoch": 0.14, "learning_rate": 9.683718792385595e-06, "loss": 1.4312, "step": 1317 }, { "epoch": 0.14, "learning_rate": 9.683113867863772e-06, "loss": 1.3583, "step": 1318 }, { "epoch": 0.14, "learning_rate": 9.682508384331306e-06, "loss": 1.5168, "step": 1319 }, { "epoch": 0.14, "learning_rate": 9.681902341860471e-06, "loss": 1.5091, "step": 1320 }, { "epoch": 0.14, "learning_rate": 9.68129574052361e-06, "loss": 1.4708, "step": 1321 }, { "epoch": 0.14, "learning_rate": 9.680688580393133e-06, "loss": 1.582, "step": 1322 }, { "epoch": 0.14, "learning_rate": 9.680080861541511e-06, "loss": 1.3791, "step": 1323 }, { "epoch": 0.14, "learning_rate": 9.679472584041289e-06, "loss": 1.746, "step": 1324 }, { "epoch": 0.14, "learning_rate": 9.678863747965073e-06, "loss": 1.3815, "step": 1325 }, { "epoch": 0.14, "learning_rate": 9.678254353385538e-06, "loss": 1.5475, "step": 1326 }, { "epoch": 0.14, "learning_rate": 9.67764440037543e-06, "loss": 1.4327, "step": 1327 }, { "epoch": 0.14, "learning_rate": 9.677033889007554e-06, "loss": 1.6448, "step": 1328 }, { "epoch": 0.14, "learning_rate": 9.676422819354785e-06, "loss": 1.3399, "step": 1329 }, { "epoch": 0.14, "learning_rate": 9.675811191490065e-06, "loss": 1.439, "step": 1330 }, { "epoch": 0.14, "learning_rate": 9.675199005486404e-06, "loss": 1.9367, "step": 1331 }, { "epoch": 0.14, "learning_rate": 9.674586261416874e-06, "loss": 1.4022, "step": 1332 }, { "epoch": 0.14, "learning_rate": 9.673972959354621e-06, "loss": 1.5833, "step": 1333 }, { "epoch": 0.14, "learning_rate": 9.673359099372847e-06, "loss": 1.7201, "step": 1334 }, { "epoch": 0.14, "learning_rate": 9.672744681544834e-06, "loss": 1.3385, "step": 1335 }, { "epoch": 0.14, "learning_rate": 9.672129705943917e-06, "loss": 1.4352, "step": 1336 }, { "epoch": 0.14, "learning_rate": 9.671514172643508e-06, "loss": 1.5541, "step": 1337 }, { "epoch": 0.14, "learning_rate": 9.670898081717079e-06, "loss": 1.5359, "step": 1338 }, { "epoch": 0.14, "learning_rate": 9.670281433238173e-06, "loss": 1.5185, "step": 1339 }, { "epoch": 0.14, "learning_rate": 9.669664227280398e-06, "loss": 1.7435, "step": 1340 }, { "epoch": 0.14, "learning_rate": 9.669046463917427e-06, "loss": 1.4912, "step": 1341 }, { "epoch": 0.14, "learning_rate": 9.668428143223e-06, "loss": 1.7452, "step": 1342 }, { "epoch": 0.14, "learning_rate": 9.667809265270926e-06, "loss": 1.7006, "step": 1343 }, { "epoch": 0.14, "learning_rate": 9.667189830135078e-06, "loss": 1.6192, "step": 1344 }, { "epoch": 0.14, "learning_rate": 9.666569837889396e-06, "loss": 1.3411, "step": 1345 }, { "epoch": 0.14, "learning_rate": 9.665949288607889e-06, "loss": 1.4982, "step": 1346 }, { "epoch": 0.14, "learning_rate": 9.665328182364627e-06, "loss": 1.4408, "step": 1347 }, { "epoch": 0.14, "learning_rate": 9.66470651923375e-06, "loss": 1.6062, "step": 1348 }, { "epoch": 0.14, "learning_rate": 9.664084299289467e-06, "loss": 1.4598, "step": 1349 }, { "epoch": 0.14, "learning_rate": 9.663461522606049e-06, "loss": 1.5333, "step": 1350 }, { "epoch": 0.14, "learning_rate": 9.662838189257836e-06, "loss": 1.5885, "step": 1351 }, { "epoch": 0.14, "learning_rate": 9.662214299319231e-06, "loss": 1.6275, "step": 1352 }, { "epoch": 0.14, "learning_rate": 9.66158985286471e-06, "loss": 1.7458, "step": 1353 }, { "epoch": 0.14, "learning_rate": 9.660964849968809e-06, "loss": 1.6024, "step": 1354 }, { "epoch": 0.14, "learning_rate": 9.660339290706132e-06, "loss": 1.7529, "step": 1355 }, { "epoch": 0.14, "learning_rate": 9.659713175151352e-06, "loss": 1.4001, "step": 1356 }, { "epoch": 0.14, "learning_rate": 9.659086503379208e-06, "loss": 1.662, "step": 1357 }, { "epoch": 0.14, "learning_rate": 9.658459275464502e-06, "loss": 1.4187, "step": 1358 }, { "epoch": 0.14, "learning_rate": 9.657831491482103e-06, "loss": 1.4821, "step": 1359 }, { "epoch": 0.15, "learning_rate": 9.657203151506953e-06, "loss": 1.3874, "step": 1360 }, { "epoch": 0.15, "learning_rate": 9.656574255614051e-06, "loss": 1.5636, "step": 1361 }, { "epoch": 0.15, "learning_rate": 9.655944803878467e-06, "loss": 1.4814, "step": 1362 }, { "epoch": 0.15, "learning_rate": 9.65531479637534e-06, "loss": 1.4269, "step": 1363 }, { "epoch": 0.15, "learning_rate": 9.654684233179867e-06, "loss": 1.7134, "step": 1364 }, { "epoch": 0.15, "learning_rate": 9.654053114367321e-06, "loss": 1.4676, "step": 1365 }, { "epoch": 0.15, "learning_rate": 9.653421440013037e-06, "loss": 1.6604, "step": 1366 }, { "epoch": 0.15, "learning_rate": 9.652789210192412e-06, "loss": 1.3766, "step": 1367 }, { "epoch": 0.15, "learning_rate": 9.65215642498092e-06, "loss": 1.5068, "step": 1368 }, { "epoch": 0.15, "learning_rate": 9.65152308445409e-06, "loss": 1.478, "step": 1369 }, { "epoch": 0.15, "learning_rate": 9.650889188687522e-06, "loss": 1.3197, "step": 1370 }, { "epoch": 0.15, "learning_rate": 9.650254737756883e-06, "loss": 1.5998, "step": 1371 }, { "epoch": 0.15, "learning_rate": 9.64961973173791e-06, "loss": 1.3549, "step": 1372 }, { "epoch": 0.15, "learning_rate": 9.648984170706395e-06, "loss": 1.407, "step": 1373 }, { "epoch": 0.15, "learning_rate": 9.648348054738208e-06, "loss": 1.5751, "step": 1374 }, { "epoch": 0.15, "learning_rate": 9.64771138390928e-06, "loss": 1.4365, "step": 1375 }, { "epoch": 0.15, "learning_rate": 9.647074158295608e-06, "loss": 1.6382, "step": 1376 }, { "epoch": 0.15, "learning_rate": 9.646436377973253e-06, "loss": 1.2082, "step": 1377 }, { "epoch": 0.15, "learning_rate": 9.645798043018352e-06, "loss": 1.5196, "step": 1378 }, { "epoch": 0.15, "learning_rate": 9.645159153507095e-06, "loss": 1.3409, "step": 1379 }, { "epoch": 0.15, "learning_rate": 9.644519709515746e-06, "loss": 1.266, "step": 1380 }, { "epoch": 0.15, "learning_rate": 9.643879711120636e-06, "loss": 1.6374, "step": 1381 }, { "epoch": 0.15, "learning_rate": 9.643239158398157e-06, "loss": 1.6447, "step": 1382 }, { "epoch": 0.15, "learning_rate": 9.642598051424772e-06, "loss": 1.4628, "step": 1383 }, { "epoch": 0.15, "learning_rate": 9.641956390277007e-06, "loss": 1.5265, "step": 1384 }, { "epoch": 0.15, "learning_rate": 9.641314175031456e-06, "loss": 1.5726, "step": 1385 }, { "epoch": 0.15, "learning_rate": 9.640671405764777e-06, "loss": 1.5076, "step": 1386 }, { "epoch": 0.15, "learning_rate": 9.640028082553699e-06, "loss": 1.6641, "step": 1387 }, { "epoch": 0.15, "learning_rate": 9.639384205475012e-06, "loss": 1.5063, "step": 1388 }, { "epoch": 0.15, "learning_rate": 9.638739774605572e-06, "loss": 1.4541, "step": 1389 }, { "epoch": 0.15, "learning_rate": 9.638094790022306e-06, "loss": 1.7834, "step": 1390 }, { "epoch": 0.15, "learning_rate": 9.6374492518022e-06, "loss": 1.4088, "step": 1391 }, { "epoch": 0.15, "learning_rate": 9.636803160022314e-06, "loss": 1.5707, "step": 1392 }, { "epoch": 0.15, "learning_rate": 9.636156514759771e-06, "loss": 1.8404, "step": 1393 }, { "epoch": 0.15, "learning_rate": 9.635509316091756e-06, "loss": 1.2106, "step": 1394 }, { "epoch": 0.15, "learning_rate": 9.634861564095525e-06, "loss": 1.0685, "step": 1395 }, { "epoch": 0.15, "learning_rate": 9.634213258848397e-06, "loss": 1.5497, "step": 1396 }, { "epoch": 0.15, "learning_rate": 9.63356440042776e-06, "loss": 1.2951, "step": 1397 }, { "epoch": 0.15, "learning_rate": 9.632914988911066e-06, "loss": 1.3132, "step": 1398 }, { "epoch": 0.15, "learning_rate": 9.632265024375835e-06, "loss": 1.4499, "step": 1399 }, { "epoch": 0.15, "learning_rate": 9.631614506899648e-06, "loss": 1.4142, "step": 1400 }, { "epoch": 0.15, "learning_rate": 9.63096343656016e-06, "loss": 1.4208, "step": 1401 }, { "epoch": 0.15, "learning_rate": 9.630311813435082e-06, "loss": 1.5422, "step": 1402 }, { "epoch": 0.15, "learning_rate": 9.6296596376022e-06, "loss": 1.7375, "step": 1403 }, { "epoch": 0.15, "learning_rate": 9.629006909139363e-06, "loss": 1.4968, "step": 1404 }, { "epoch": 0.15, "learning_rate": 9.628353628124484e-06, "loss": 1.706, "step": 1405 }, { "epoch": 0.15, "learning_rate": 9.627699794635545e-06, "loss": 1.5057, "step": 1406 }, { "epoch": 0.15, "learning_rate": 9.62704540875059e-06, "loss": 1.5976, "step": 1407 }, { "epoch": 0.15, "learning_rate": 9.626390470547733e-06, "loss": 1.5852, "step": 1408 }, { "epoch": 0.15, "learning_rate": 9.62573498010515e-06, "loss": 1.7537, "step": 1409 }, { "epoch": 0.15, "learning_rate": 9.625078937501089e-06, "loss": 1.442, "step": 1410 }, { "epoch": 0.15, "learning_rate": 9.624422342813857e-06, "loss": 1.4474, "step": 1411 }, { "epoch": 0.15, "learning_rate": 9.62376519612183e-06, "loss": 1.6584, "step": 1412 }, { "epoch": 0.15, "learning_rate": 9.62310749750345e-06, "loss": 1.4793, "step": 1413 }, { "epoch": 0.15, "learning_rate": 9.622449247037227e-06, "loss": 1.4049, "step": 1414 }, { "epoch": 0.15, "learning_rate": 9.621790444801732e-06, "loss": 1.5971, "step": 1415 }, { "epoch": 0.15, "learning_rate": 9.621131090875603e-06, "loss": 1.6978, "step": 1416 }, { "epoch": 0.15, "learning_rate": 9.620471185337552e-06, "loss": 1.5668, "step": 1417 }, { "epoch": 0.15, "learning_rate": 9.619810728266344e-06, "loss": 1.2772, "step": 1418 }, { "epoch": 0.15, "learning_rate": 9.619149719740817e-06, "loss": 1.5959, "step": 1419 }, { "epoch": 0.15, "learning_rate": 9.618488159839874e-06, "loss": 1.5008, "step": 1420 }, { "epoch": 0.15, "learning_rate": 9.617826048642484e-06, "loss": 1.4712, "step": 1421 }, { "epoch": 0.15, "learning_rate": 9.617163386227683e-06, "loss": 1.4722, "step": 1422 }, { "epoch": 0.15, "learning_rate": 9.616500172674568e-06, "loss": 2.0091, "step": 1423 }, { "epoch": 0.15, "learning_rate": 9.615836408062307e-06, "loss": 1.5566, "step": 1424 }, { "epoch": 0.15, "learning_rate": 9.615172092470134e-06, "loss": 1.537, "step": 1425 }, { "epoch": 0.15, "learning_rate": 9.614507225977342e-06, "loss": 1.5262, "step": 1426 }, { "epoch": 0.15, "learning_rate": 9.613841808663296e-06, "loss": 1.6764, "step": 1427 }, { "epoch": 0.15, "learning_rate": 9.613175840607428e-06, "loss": 1.6334, "step": 1428 }, { "epoch": 0.15, "learning_rate": 9.612509321889228e-06, "loss": 1.4772, "step": 1429 }, { "epoch": 0.15, "learning_rate": 9.611842252588259e-06, "loss": 1.2798, "step": 1430 }, { "epoch": 0.15, "learning_rate": 9.611174632784147e-06, "loss": 1.3227, "step": 1431 }, { "epoch": 0.15, "learning_rate": 9.610506462556583e-06, "loss": 1.5763, "step": 1432 }, { "epoch": 0.15, "learning_rate": 9.60983774198533e-06, "loss": 1.6, "step": 1433 }, { "epoch": 0.15, "learning_rate": 9.609168471150202e-06, "loss": 1.3513, "step": 1434 }, { "epoch": 0.15, "learning_rate": 9.608498650131095e-06, "loss": 1.6236, "step": 1435 }, { "epoch": 0.15, "learning_rate": 9.607828279007962e-06, "loss": 1.6392, "step": 1436 }, { "epoch": 0.15, "learning_rate": 9.607157357860823e-06, "loss": 1.4585, "step": 1437 }, { "epoch": 0.15, "learning_rate": 9.606485886769766e-06, "loss": 1.6678, "step": 1438 }, { "epoch": 0.15, "learning_rate": 9.60581386581494e-06, "loss": 1.3172, "step": 1439 }, { "epoch": 0.15, "learning_rate": 9.605141295076561e-06, "loss": 1.5385, "step": 1440 }, { "epoch": 0.15, "learning_rate": 9.604468174634917e-06, "loss": 1.7557, "step": 1441 }, { "epoch": 0.15, "learning_rate": 9.603794504570352e-06, "loss": 1.5939, "step": 1442 }, { "epoch": 0.15, "learning_rate": 9.603120284963284e-06, "loss": 1.6819, "step": 1443 }, { "epoch": 0.15, "learning_rate": 9.60244551589419e-06, "loss": 1.5806, "step": 1444 }, { "epoch": 0.15, "learning_rate": 9.601770197443615e-06, "loss": 1.3141, "step": 1445 }, { "epoch": 0.15, "learning_rate": 9.601094329692173e-06, "loss": 1.4467, "step": 1446 }, { "epoch": 0.15, "learning_rate": 9.600417912720537e-06, "loss": 1.5422, "step": 1447 }, { "epoch": 0.15, "learning_rate": 9.599740946609451e-06, "loss": 1.7134, "step": 1448 }, { "epoch": 0.15, "learning_rate": 9.599063431439721e-06, "loss": 1.1911, "step": 1449 }, { "epoch": 0.15, "learning_rate": 9.598385367292224e-06, "loss": 1.662, "step": 1450 }, { "epoch": 0.15, "learning_rate": 9.597706754247895e-06, "loss": 1.3375, "step": 1451 }, { "epoch": 0.15, "learning_rate": 9.597027592387739e-06, "loss": 1.5488, "step": 1452 }, { "epoch": 0.15, "learning_rate": 9.596347881792827e-06, "loss": 1.6211, "step": 1453 }, { "epoch": 0.16, "learning_rate": 9.595667622544291e-06, "loss": 1.2572, "step": 1454 }, { "epoch": 0.16, "learning_rate": 9.594986814723335e-06, "loss": 1.7921, "step": 1455 }, { "epoch": 0.16, "learning_rate": 9.594305458411225e-06, "loss": 1.4244, "step": 1456 }, { "epoch": 0.16, "learning_rate": 9.59362355368929e-06, "loss": 1.5101, "step": 1457 }, { "epoch": 0.16, "learning_rate": 9.59294110063893e-06, "loss": 1.2617, "step": 1458 }, { "epoch": 0.16, "learning_rate": 9.592258099341608e-06, "loss": 1.5741, "step": 1459 }, { "epoch": 0.16, "learning_rate": 9.591574549878848e-06, "loss": 1.3902, "step": 1460 }, { "epoch": 0.16, "learning_rate": 9.590890452332249e-06, "loss": 1.4596, "step": 1461 }, { "epoch": 0.16, "learning_rate": 9.590205806783463e-06, "loss": 1.4919, "step": 1462 }, { "epoch": 0.16, "learning_rate": 9.58952061331422e-06, "loss": 1.2328, "step": 1463 }, { "epoch": 0.16, "learning_rate": 9.588834872006308e-06, "loss": 1.5525, "step": 1464 }, { "epoch": 0.16, "learning_rate": 9.588148582941583e-06, "loss": 1.2841, "step": 1465 }, { "epoch": 0.16, "learning_rate": 9.587461746201963e-06, "loss": 1.3347, "step": 1466 }, { "epoch": 0.16, "learning_rate": 9.586774361869436e-06, "loss": 1.8875, "step": 1467 }, { "epoch": 0.16, "learning_rate": 9.58608643002605e-06, "loss": 1.5408, "step": 1468 }, { "epoch": 0.16, "learning_rate": 9.585397950753926e-06, "loss": 1.3174, "step": 1469 }, { "epoch": 0.16, "learning_rate": 9.584708924135245e-06, "loss": 1.4216, "step": 1470 }, { "epoch": 0.16, "learning_rate": 9.58401935025225e-06, "loss": 1.6315, "step": 1471 }, { "epoch": 0.16, "learning_rate": 9.583329229187259e-06, "loss": 1.468, "step": 1472 }, { "epoch": 0.16, "learning_rate": 9.582638561022646e-06, "loss": 1.2098, "step": 1473 }, { "epoch": 0.16, "learning_rate": 9.581947345840858e-06, "loss": 1.6396, "step": 1474 }, { "epoch": 0.16, "learning_rate": 9.5812555837244e-06, "loss": 1.1979, "step": 1475 }, { "epoch": 0.16, "learning_rate": 9.580563274755848e-06, "loss": 1.6944, "step": 1476 }, { "epoch": 0.16, "learning_rate": 9.579870419017838e-06, "loss": 1.7977, "step": 1477 }, { "epoch": 0.16, "learning_rate": 9.579177016593077e-06, "loss": 1.3296, "step": 1478 }, { "epoch": 0.16, "learning_rate": 9.578483067564335e-06, "loss": 1.3867, "step": 1479 }, { "epoch": 0.16, "learning_rate": 9.577788572014447e-06, "loss": 1.2854, "step": 1480 }, { "epoch": 0.16, "learning_rate": 9.577093530026309e-06, "loss": 1.7073, "step": 1481 }, { "epoch": 0.16, "learning_rate": 9.576397941682891e-06, "loss": 1.4895, "step": 1482 }, { "epoch": 0.16, "learning_rate": 9.57570180706722e-06, "loss": 1.3329, "step": 1483 }, { "epoch": 0.16, "learning_rate": 9.575005126262395e-06, "loss": 1.3153, "step": 1484 }, { "epoch": 0.16, "learning_rate": 9.574307899351574e-06, "loss": 1.6391, "step": 1485 }, { "epoch": 0.16, "learning_rate": 9.573610126417985e-06, "loss": 1.3376, "step": 1486 }, { "epoch": 0.16, "learning_rate": 9.57291180754492e-06, "loss": 1.4582, "step": 1487 }, { "epoch": 0.16, "learning_rate": 9.572212942815734e-06, "loss": 1.6361, "step": 1488 }, { "epoch": 0.16, "learning_rate": 9.571513532313844e-06, "loss": 1.4764, "step": 1489 }, { "epoch": 0.16, "learning_rate": 9.570813576122746e-06, "loss": 1.716, "step": 1490 }, { "epoch": 0.16, "learning_rate": 9.570113074325986e-06, "loss": 1.5901, "step": 1491 }, { "epoch": 0.16, "learning_rate": 9.569412027007183e-06, "loss": 1.4481, "step": 1492 }, { "epoch": 0.16, "learning_rate": 9.568710434250017e-06, "loss": 1.4505, "step": 1493 }, { "epoch": 0.16, "learning_rate": 9.568008296138238e-06, "loss": 1.3535, "step": 1494 }, { "epoch": 0.16, "learning_rate": 9.567305612755654e-06, "loss": 1.6873, "step": 1495 }, { "epoch": 0.16, "learning_rate": 9.56660238418615e-06, "loss": 1.4176, "step": 1496 }, { "epoch": 0.16, "learning_rate": 9.565898610513661e-06, "loss": 1.4805, "step": 1497 }, { "epoch": 0.16, "learning_rate": 9.565194291822198e-06, "loss": 1.4552, "step": 1498 }, { "epoch": 0.16, "learning_rate": 9.564489428195834e-06, "loss": 1.2519, "step": 1499 }, { "epoch": 0.16, "learning_rate": 9.563784019718704e-06, "loss": 1.5291, "step": 1500 }, { "epoch": 0.16, "learning_rate": 9.563078066475012e-06, "loss": 1.7616, "step": 1501 }, { "epoch": 0.16, "learning_rate": 9.562371568549027e-06, "loss": 1.7717, "step": 1502 }, { "epoch": 0.16, "learning_rate": 9.561664526025082e-06, "loss": 1.6346, "step": 1503 }, { "epoch": 0.16, "learning_rate": 9.560956938987573e-06, "loss": 1.4986, "step": 1504 }, { "epoch": 0.16, "learning_rate": 9.560248807520963e-06, "loss": 1.3172, "step": 1505 }, { "epoch": 0.16, "learning_rate": 9.55954013170978e-06, "loss": 1.4528, "step": 1506 }, { "epoch": 0.16, "learning_rate": 9.558830911638616e-06, "loss": 1.6411, "step": 1507 }, { "epoch": 0.16, "learning_rate": 9.558121147392132e-06, "loss": 1.3662, "step": 1508 }, { "epoch": 0.16, "learning_rate": 9.557410839055047e-06, "loss": 1.6489, "step": 1509 }, { "epoch": 0.16, "learning_rate": 9.55669998671215e-06, "loss": 1.5034, "step": 1510 }, { "epoch": 0.16, "learning_rate": 9.555988590448292e-06, "loss": 1.5618, "step": 1511 }, { "epoch": 0.16, "learning_rate": 9.555276650348393e-06, "loss": 1.5063, "step": 1512 }, { "epoch": 0.16, "learning_rate": 9.554564166497434e-06, "loss": 1.4692, "step": 1513 }, { "epoch": 0.16, "learning_rate": 9.553851138980462e-06, "loss": 1.2582, "step": 1514 }, { "epoch": 0.16, "learning_rate": 9.55313756788259e-06, "loss": 1.2956, "step": 1515 }, { "epoch": 0.16, "learning_rate": 9.552423453288995e-06, "loss": 1.6612, "step": 1516 }, { "epoch": 0.16, "learning_rate": 9.551708795284917e-06, "loss": 1.2788, "step": 1517 }, { "epoch": 0.16, "learning_rate": 9.550993593955665e-06, "loss": 1.6689, "step": 1518 }, { "epoch": 0.16, "learning_rate": 9.55027784938661e-06, "loss": 1.398, "step": 1519 }, { "epoch": 0.16, "learning_rate": 9.54956156166319e-06, "loss": 1.4049, "step": 1520 }, { "epoch": 0.16, "learning_rate": 9.548844730870903e-06, "loss": 1.5694, "step": 1521 }, { "epoch": 0.16, "learning_rate": 9.548127357095316e-06, "loss": 1.3873, "step": 1522 }, { "epoch": 0.16, "learning_rate": 9.547409440422061e-06, "loss": 1.5448, "step": 1523 }, { "epoch": 0.16, "learning_rate": 9.546690980936836e-06, "loss": 1.5388, "step": 1524 }, { "epoch": 0.16, "learning_rate": 9.545971978725399e-06, "loss": 1.6191, "step": 1525 }, { "epoch": 0.16, "learning_rate": 9.545252433873574e-06, "loss": 1.4571, "step": 1526 }, { "epoch": 0.16, "learning_rate": 9.544532346467254e-06, "loss": 1.8593, "step": 1527 }, { "epoch": 0.16, "learning_rate": 9.543811716592391e-06, "loss": 1.6225, "step": 1528 }, { "epoch": 0.16, "learning_rate": 9.543090544335008e-06, "loss": 1.58, "step": 1529 }, { "epoch": 0.16, "learning_rate": 9.542368829781186e-06, "loss": 1.4566, "step": 1530 }, { "epoch": 0.16, "learning_rate": 9.541646573017077e-06, "loss": 1.6764, "step": 1531 }, { "epoch": 0.16, "learning_rate": 9.540923774128893e-06, "loss": 1.6531, "step": 1532 }, { "epoch": 0.16, "learning_rate": 9.540200433202913e-06, "loss": 1.4697, "step": 1533 }, { "epoch": 0.16, "learning_rate": 9.539476550325481e-06, "loss": 1.5715, "step": 1534 }, { "epoch": 0.16, "learning_rate": 9.538752125583003e-06, "loss": 1.3609, "step": 1535 }, { "epoch": 0.16, "learning_rate": 9.538027159061955e-06, "loss": 2.126, "step": 1536 }, { "epoch": 0.16, "learning_rate": 9.537301650848872e-06, "loss": 1.5198, "step": 1537 }, { "epoch": 0.16, "learning_rate": 9.536575601030355e-06, "loss": 1.2805, "step": 1538 }, { "epoch": 0.16, "learning_rate": 9.535849009693072e-06, "loss": 1.6009, "step": 1539 }, { "epoch": 0.16, "learning_rate": 9.535121876923754e-06, "loss": 1.4681, "step": 1540 }, { "epoch": 0.16, "learning_rate": 9.534394202809198e-06, "loss": 1.5031, "step": 1541 }, { "epoch": 0.16, "learning_rate": 9.533665987436262e-06, "loss": 1.3236, "step": 1542 }, { "epoch": 0.16, "learning_rate": 9.532937230891874e-06, "loss": 1.3728, "step": 1543 }, { "epoch": 0.16, "learning_rate": 9.53220793326302e-06, "loss": 1.6736, "step": 1544 }, { "epoch": 0.16, "learning_rate": 9.531478094636758e-06, "loss": 1.283, "step": 1545 }, { "epoch": 0.16, "learning_rate": 9.530747715100205e-06, "loss": 1.5924, "step": 1546 }, { "epoch": 0.17, "learning_rate": 9.530016794740546e-06, "loss": 1.733, "step": 1547 }, { "epoch": 0.17, "learning_rate": 9.529285333645027e-06, "loss": 1.5765, "step": 1548 }, { "epoch": 0.17, "learning_rate": 9.52855333190096e-06, "loss": 1.2816, "step": 1549 }, { "epoch": 0.17, "learning_rate": 9.527820789595725e-06, "loss": 1.6859, "step": 1550 }, { "epoch": 0.17, "learning_rate": 9.527087706816762e-06, "loss": 1.7311, "step": 1551 }, { "epoch": 0.17, "learning_rate": 9.526354083651576e-06, "loss": 1.5555, "step": 1552 }, { "epoch": 0.17, "learning_rate": 9.52561992018774e-06, "loss": 1.6024, "step": 1553 }, { "epoch": 0.17, "learning_rate": 9.524885216512887e-06, "loss": 1.5175, "step": 1554 }, { "epoch": 0.17, "learning_rate": 9.524149972714717e-06, "loss": 1.5606, "step": 1555 }, { "epoch": 0.17, "learning_rate": 9.523414188880994e-06, "loss": 1.705, "step": 1556 }, { "epoch": 0.17, "learning_rate": 9.522677865099548e-06, "loss": 1.194, "step": 1557 }, { "epoch": 0.17, "learning_rate": 9.521941001458272e-06, "loss": 1.7964, "step": 1558 }, { "epoch": 0.17, "learning_rate": 9.521203598045122e-06, "loss": 1.3566, "step": 1559 }, { "epoch": 0.17, "learning_rate": 9.520465654948119e-06, "loss": 1.2925, "step": 1560 }, { "epoch": 0.17, "learning_rate": 9.519727172255351e-06, "loss": 1.426, "step": 1561 }, { "epoch": 0.17, "learning_rate": 9.518988150054968e-06, "loss": 1.6032, "step": 1562 }, { "epoch": 0.17, "learning_rate": 9.518248588435185e-06, "loss": 1.303, "step": 1563 }, { "epoch": 0.17, "learning_rate": 9.517508487484283e-06, "loss": 1.6585, "step": 1564 }, { "epoch": 0.17, "learning_rate": 9.516767847290603e-06, "loss": 1.6213, "step": 1565 }, { "epoch": 0.17, "learning_rate": 9.516026667942557e-06, "loss": 1.3136, "step": 1566 }, { "epoch": 0.17, "learning_rate": 9.515284949528614e-06, "loss": 1.6103, "step": 1567 }, { "epoch": 0.17, "learning_rate": 9.514542692137311e-06, "loss": 1.6547, "step": 1568 }, { "epoch": 0.17, "learning_rate": 9.513799895857252e-06, "loss": 1.5618, "step": 1569 }, { "epoch": 0.17, "learning_rate": 9.513056560777102e-06, "loss": 1.5355, "step": 1570 }, { "epoch": 0.17, "learning_rate": 9.512312686985589e-06, "loss": 1.4711, "step": 1571 }, { "epoch": 0.17, "learning_rate": 9.511568274571508e-06, "loss": 1.4639, "step": 1572 }, { "epoch": 0.17, "learning_rate": 9.510823323623718e-06, "loss": 1.5498, "step": 1573 }, { "epoch": 0.17, "learning_rate": 9.510077834231141e-06, "loss": 1.3109, "step": 1574 }, { "epoch": 0.17, "learning_rate": 9.509331806482767e-06, "loss": 1.5003, "step": 1575 }, { "epoch": 0.17, "learning_rate": 9.508585240467642e-06, "loss": 1.3861, "step": 1576 }, { "epoch": 0.17, "learning_rate": 9.507838136274887e-06, "loss": 1.7588, "step": 1577 }, { "epoch": 0.17, "learning_rate": 9.507090493993677e-06, "loss": 1.6696, "step": 1578 }, { "epoch": 0.17, "learning_rate": 9.50634231371326e-06, "loss": 1.5883, "step": 1579 }, { "epoch": 0.17, "learning_rate": 9.505593595522941e-06, "loss": 1.5924, "step": 1580 }, { "epoch": 0.17, "learning_rate": 9.504844339512096e-06, "loss": 1.5284, "step": 1581 }, { "epoch": 0.17, "learning_rate": 9.50409454577016e-06, "loss": 1.1143, "step": 1582 }, { "epoch": 0.17, "learning_rate": 9.503344214386632e-06, "loss": 1.6791, "step": 1583 }, { "epoch": 0.17, "learning_rate": 9.502593345451078e-06, "loss": 1.471, "step": 1584 }, { "epoch": 0.17, "learning_rate": 9.50184193905313e-06, "loss": 1.4303, "step": 1585 }, { "epoch": 0.17, "learning_rate": 9.501089995282478e-06, "loss": 1.4389, "step": 1586 }, { "epoch": 0.17, "learning_rate": 9.500337514228878e-06, "loss": 1.5871, "step": 1587 }, { "epoch": 0.17, "learning_rate": 9.499584495982157e-06, "loss": 1.4424, "step": 1588 }, { "epoch": 0.17, "learning_rate": 9.498830940632199e-06, "loss": 1.6751, "step": 1589 }, { "epoch": 0.17, "learning_rate": 9.49807684826895e-06, "loss": 1.6138, "step": 1590 }, { "epoch": 0.17, "learning_rate": 9.49732221898243e-06, "loss": 1.4394, "step": 1591 }, { "epoch": 0.17, "learning_rate": 9.496567052862713e-06, "loss": 1.4358, "step": 1592 }, { "epoch": 0.17, "learning_rate": 9.495811349999941e-06, "loss": 1.5728, "step": 1593 }, { "epoch": 0.17, "learning_rate": 9.495055110484322e-06, "loss": 1.6027, "step": 1594 }, { "epoch": 0.17, "learning_rate": 9.494298334406125e-06, "loss": 1.4951, "step": 1595 }, { "epoch": 0.17, "learning_rate": 9.493541021855685e-06, "loss": 1.624, "step": 1596 }, { "epoch": 0.17, "learning_rate": 9.4927831729234e-06, "loss": 1.561, "step": 1597 }, { "epoch": 0.17, "learning_rate": 9.492024787699733e-06, "loss": 1.525, "step": 1598 }, { "epoch": 0.17, "learning_rate": 9.49126586627521e-06, "loss": 1.2852, "step": 1599 }, { "epoch": 0.17, "learning_rate": 9.49050640874042e-06, "loss": 1.595, "step": 1600 }, { "epoch": 0.17, "learning_rate": 9.48974641518602e-06, "loss": 1.6351, "step": 1601 }, { "epoch": 0.17, "learning_rate": 9.488985885702728e-06, "loss": 1.4962, "step": 1602 }, { "epoch": 0.17, "learning_rate": 9.488224820381324e-06, "loss": 1.5944, "step": 1603 }, { "epoch": 0.17, "learning_rate": 9.487463219312657e-06, "loss": 1.2207, "step": 1604 }, { "epoch": 0.17, "learning_rate": 9.486701082587635e-06, "loss": 1.4897, "step": 1605 }, { "epoch": 0.17, "learning_rate": 9.485938410297233e-06, "loss": 1.9196, "step": 1606 }, { "epoch": 0.17, "learning_rate": 9.48517520253249e-06, "loss": 1.4132, "step": 1607 }, { "epoch": 0.17, "learning_rate": 9.484411459384508e-06, "loss": 1.7142, "step": 1608 }, { "epoch": 0.17, "learning_rate": 9.48364718094445e-06, "loss": 1.4015, "step": 1609 }, { "epoch": 0.17, "learning_rate": 9.482882367303552e-06, "loss": 1.4199, "step": 1610 }, { "epoch": 0.17, "learning_rate": 9.482117018553101e-06, "loss": 1.3889, "step": 1611 }, { "epoch": 0.17, "learning_rate": 9.481351134784458e-06, "loss": 1.5608, "step": 1612 }, { "epoch": 0.17, "learning_rate": 9.480584716089045e-06, "loss": 1.2908, "step": 1613 }, { "epoch": 0.17, "learning_rate": 9.479817762558345e-06, "loss": 1.5561, "step": 1614 }, { "epoch": 0.17, "learning_rate": 9.47905027428391e-06, "loss": 1.386, "step": 1615 }, { "epoch": 0.17, "learning_rate": 9.478282251357352e-06, "loss": 1.7139, "step": 1616 }, { "epoch": 0.17, "learning_rate": 9.477513693870347e-06, "loss": 1.6739, "step": 1617 }, { "epoch": 0.17, "learning_rate": 9.476744601914634e-06, "loss": 1.4318, "step": 1618 }, { "epoch": 0.17, "learning_rate": 9.475974975582021e-06, "loss": 1.6615, "step": 1619 }, { "epoch": 0.17, "learning_rate": 9.475204814964374e-06, "loss": 1.6151, "step": 1620 }, { "epoch": 0.17, "learning_rate": 9.474434120153626e-06, "loss": 1.533, "step": 1621 }, { "epoch": 0.17, "learning_rate": 9.473662891241773e-06, "loss": 1.5625, "step": 1622 }, { "epoch": 0.17, "learning_rate": 9.472891128320874e-06, "loss": 1.731, "step": 1623 }, { "epoch": 0.17, "learning_rate": 9.472118831483052e-06, "loss": 1.709, "step": 1624 }, { "epoch": 0.17, "learning_rate": 9.471346000820493e-06, "loss": 1.7115, "step": 1625 }, { "epoch": 0.17, "learning_rate": 9.470572636425451e-06, "loss": 1.4129, "step": 1626 }, { "epoch": 0.17, "learning_rate": 9.469798738390236e-06, "loss": 1.3771, "step": 1627 }, { "epoch": 0.17, "learning_rate": 9.469024306807229e-06, "loss": 1.8019, "step": 1628 }, { "epoch": 0.17, "learning_rate": 9.46824934176887e-06, "loss": 1.5105, "step": 1629 }, { "epoch": 0.17, "learning_rate": 9.467473843367668e-06, "loss": 1.4225, "step": 1630 }, { "epoch": 0.17, "learning_rate": 9.46669781169619e-06, "loss": 1.6469, "step": 1631 }, { "epoch": 0.17, "learning_rate": 9.465921246847067e-06, "loss": 1.5295, "step": 1632 }, { "epoch": 0.17, "learning_rate": 9.465144148912997e-06, "loss": 1.4158, "step": 1633 }, { "epoch": 0.17, "learning_rate": 9.46436651798674e-06, "loss": 1.5594, "step": 1634 }, { "epoch": 0.17, "learning_rate": 9.463588354161122e-06, "loss": 1.5421, "step": 1635 }, { "epoch": 0.17, "learning_rate": 9.462809657529025e-06, "loss": 1.4976, "step": 1636 }, { "epoch": 0.17, "learning_rate": 9.462030428183406e-06, "loss": 1.3744, "step": 1637 }, { "epoch": 0.17, "learning_rate": 9.461250666217277e-06, "loss": 1.5017, "step": 1638 }, { "epoch": 0.17, "learning_rate": 9.460470371723714e-06, "loss": 1.5741, "step": 1639 }, { "epoch": 0.17, "learning_rate": 9.459689544795859e-06, "loss": 1.3518, "step": 1640 }, { "epoch": 0.18, "learning_rate": 9.458908185526921e-06, "loss": 1.3491, "step": 1641 }, { "epoch": 0.18, "learning_rate": 9.458126294010164e-06, "loss": 1.592, "step": 1642 }, { "epoch": 0.18, "learning_rate": 9.457343870338924e-06, "loss": 1.4799, "step": 1643 }, { "epoch": 0.18, "learning_rate": 9.456560914606594e-06, "loss": 1.4772, "step": 1644 }, { "epoch": 0.18, "learning_rate": 9.455777426906635e-06, "loss": 1.6976, "step": 1645 }, { "epoch": 0.18, "learning_rate": 9.454993407332572e-06, "loss": 1.5195, "step": 1646 }, { "epoch": 0.18, "learning_rate": 9.454208855977986e-06, "loss": 1.7215, "step": 1647 }, { "epoch": 0.18, "learning_rate": 9.453423772936529e-06, "loss": 1.6184, "step": 1648 }, { "epoch": 0.18, "learning_rate": 9.452638158301917e-06, "loss": 1.657, "step": 1649 }, { "epoch": 0.18, "learning_rate": 9.451852012167924e-06, "loss": 1.5092, "step": 1650 }, { "epoch": 0.18, "learning_rate": 9.45106533462839e-06, "loss": 1.3698, "step": 1651 }, { "epoch": 0.18, "learning_rate": 9.45027812577722e-06, "loss": 1.7427, "step": 1652 }, { "epoch": 0.18, "learning_rate": 9.449490385708378e-06, "loss": 1.4124, "step": 1653 }, { "epoch": 0.18, "learning_rate": 9.448702114515897e-06, "loss": 1.4621, "step": 1654 }, { "epoch": 0.18, "learning_rate": 9.447913312293872e-06, "loss": 1.4656, "step": 1655 }, { "epoch": 0.18, "learning_rate": 9.447123979136457e-06, "loss": 1.7423, "step": 1656 }, { "epoch": 0.18, "learning_rate": 9.446334115137876e-06, "loss": 1.4905, "step": 1657 }, { "epoch": 0.18, "learning_rate": 9.44554372039241e-06, "loss": 1.9252, "step": 1658 }, { "epoch": 0.18, "learning_rate": 9.444752794994408e-06, "loss": 1.5538, "step": 1659 }, { "epoch": 0.18, "learning_rate": 9.443961339038278e-06, "loss": 1.5363, "step": 1660 }, { "epoch": 0.18, "learning_rate": 9.443169352618498e-06, "loss": 1.5803, "step": 1661 }, { "epoch": 0.18, "learning_rate": 9.4423768358296e-06, "loss": 1.6334, "step": 1662 }, { "epoch": 0.18, "learning_rate": 9.441583788766191e-06, "loss": 1.342, "step": 1663 }, { "epoch": 0.18, "learning_rate": 9.44079021152293e-06, "loss": 1.5754, "step": 1664 }, { "epoch": 0.18, "learning_rate": 9.439996104194546e-06, "loss": 1.6448, "step": 1665 }, { "epoch": 0.18, "learning_rate": 9.439201466875831e-06, "loss": 1.5103, "step": 1666 }, { "epoch": 0.18, "learning_rate": 9.438406299661636e-06, "loss": 1.5284, "step": 1667 }, { "epoch": 0.18, "learning_rate": 9.437610602646878e-06, "loss": 1.4237, "step": 1668 }, { "epoch": 0.18, "learning_rate": 9.43681437592654e-06, "loss": 1.3407, "step": 1669 }, { "epoch": 0.18, "learning_rate": 9.43601761959566e-06, "loss": 1.4518, "step": 1670 }, { "epoch": 0.18, "learning_rate": 9.43522033374935e-06, "loss": 1.5229, "step": 1671 }, { "epoch": 0.18, "learning_rate": 9.43442251848278e-06, "loss": 1.2809, "step": 1672 }, { "epoch": 0.18, "learning_rate": 9.433624173891181e-06, "loss": 1.4715, "step": 1673 }, { "epoch": 0.18, "learning_rate": 9.432825300069848e-06, "loss": 1.6856, "step": 1674 }, { "epoch": 0.18, "learning_rate": 9.43202589711414e-06, "loss": 1.319, "step": 1675 }, { "epoch": 0.18, "learning_rate": 9.431225965119483e-06, "loss": 1.5227, "step": 1676 }, { "epoch": 0.18, "learning_rate": 9.430425504181361e-06, "loss": 1.6195, "step": 1677 }, { "epoch": 0.18, "learning_rate": 9.429624514395324e-06, "loss": 1.5597, "step": 1678 }, { "epoch": 0.18, "learning_rate": 9.428822995856984e-06, "loss": 1.5555, "step": 1679 }, { "epoch": 0.18, "learning_rate": 9.428020948662012e-06, "loss": 1.7465, "step": 1680 }, { "epoch": 0.18, "learning_rate": 9.427218372906151e-06, "loss": 1.5906, "step": 1681 }, { "epoch": 0.18, "learning_rate": 9.426415268685198e-06, "loss": 1.6037, "step": 1682 }, { "epoch": 0.18, "learning_rate": 9.425611636095023e-06, "loss": 1.8378, "step": 1683 }, { "epoch": 0.18, "learning_rate": 9.424807475231548e-06, "loss": 1.6669, "step": 1684 }, { "epoch": 0.18, "learning_rate": 9.424002786190768e-06, "loss": 1.8881, "step": 1685 }, { "epoch": 0.18, "learning_rate": 9.423197569068733e-06, "loss": 1.5005, "step": 1686 }, { "epoch": 0.18, "learning_rate": 9.42239182396156e-06, "loss": 1.4681, "step": 1687 }, { "epoch": 0.18, "learning_rate": 9.421585550965432e-06, "loss": 1.4912, "step": 1688 }, { "epoch": 0.18, "learning_rate": 9.420778750176588e-06, "loss": 1.4232, "step": 1689 }, { "epoch": 0.18, "learning_rate": 9.419971421691335e-06, "loss": 1.4001, "step": 1690 }, { "epoch": 0.18, "learning_rate": 9.419163565606042e-06, "loss": 1.4911, "step": 1691 }, { "epoch": 0.18, "learning_rate": 9.418355182017138e-06, "loss": 1.7982, "step": 1692 }, { "epoch": 0.18, "learning_rate": 9.417546271021122e-06, "loss": 1.2706, "step": 1693 }, { "epoch": 0.18, "learning_rate": 9.416736832714547e-06, "loss": 1.4113, "step": 1694 }, { "epoch": 0.18, "learning_rate": 9.41592686719404e-06, "loss": 1.5193, "step": 1695 }, { "epoch": 0.18, "learning_rate": 9.415116374556276e-06, "loss": 1.5917, "step": 1696 }, { "epoch": 0.18, "learning_rate": 9.414305354898005e-06, "loss": 1.8404, "step": 1697 }, { "epoch": 0.18, "learning_rate": 9.413493808316038e-06, "loss": 1.3228, "step": 1698 }, { "epoch": 0.18, "learning_rate": 9.412681734907246e-06, "loss": 1.6035, "step": 1699 }, { "epoch": 0.18, "learning_rate": 9.411869134768563e-06, "loss": 1.5244, "step": 1700 }, { "epoch": 0.18, "learning_rate": 9.411056007996989e-06, "loss": 1.3237, "step": 1701 }, { "epoch": 0.18, "learning_rate": 9.410242354689582e-06, "loss": 1.4953, "step": 1702 }, { "epoch": 0.18, "learning_rate": 9.409428174943468e-06, "loss": 1.8084, "step": 1703 }, { "epoch": 0.18, "learning_rate": 9.408613468855829e-06, "loss": 1.5149, "step": 1704 }, { "epoch": 0.18, "learning_rate": 9.407798236523921e-06, "loss": 1.4674, "step": 1705 }, { "epoch": 0.18, "learning_rate": 9.406982478045052e-06, "loss": 1.4532, "step": 1706 }, { "epoch": 0.18, "learning_rate": 9.406166193516596e-06, "loss": 1.2757, "step": 1707 }, { "epoch": 0.18, "learning_rate": 9.405349383035995e-06, "loss": 1.5163, "step": 1708 }, { "epoch": 0.18, "learning_rate": 9.404532046700745e-06, "loss": 1.3791, "step": 1709 }, { "epoch": 0.18, "learning_rate": 9.403714184608411e-06, "loss": 1.4084, "step": 1710 }, { "epoch": 0.18, "learning_rate": 9.402895796856619e-06, "loss": 1.4229, "step": 1711 }, { "epoch": 0.18, "learning_rate": 9.40207688354306e-06, "loss": 1.4376, "step": 1712 }, { "epoch": 0.18, "learning_rate": 9.40125744476548e-06, "loss": 1.884, "step": 1713 }, { "epoch": 0.18, "learning_rate": 9.400437480621697e-06, "loss": 1.6439, "step": 1714 }, { "epoch": 0.18, "learning_rate": 9.399616991209587e-06, "loss": 1.3386, "step": 1715 }, { "epoch": 0.18, "learning_rate": 9.398795976627091e-06, "loss": 1.5231, "step": 1716 }, { "epoch": 0.18, "learning_rate": 9.397974436972208e-06, "loss": 1.5058, "step": 1717 }, { "epoch": 0.18, "learning_rate": 9.397152372343008e-06, "loss": 1.5469, "step": 1718 }, { "epoch": 0.18, "learning_rate": 9.396329782837614e-06, "loss": 1.4199, "step": 1719 }, { "epoch": 0.18, "learning_rate": 9.395506668554218e-06, "loss": 1.5142, "step": 1720 }, { "epoch": 0.18, "learning_rate": 9.394683029591074e-06, "loss": 1.6112, "step": 1721 }, { "epoch": 0.18, "learning_rate": 9.393858866046494e-06, "loss": 1.456, "step": 1722 }, { "epoch": 0.18, "learning_rate": 9.39303417801886e-06, "loss": 1.3813, "step": 1723 }, { "epoch": 0.18, "learning_rate": 9.392208965606613e-06, "loss": 1.6393, "step": 1724 }, { "epoch": 0.18, "learning_rate": 9.391383228908253e-06, "loss": 1.2128, "step": 1725 }, { "epoch": 0.18, "learning_rate": 9.390556968022348e-06, "loss": 1.4058, "step": 1726 }, { "epoch": 0.18, "learning_rate": 9.389730183047528e-06, "loss": 1.919, "step": 1727 }, { "epoch": 0.18, "learning_rate": 9.388902874082482e-06, "loss": 1.7193, "step": 1728 }, { "epoch": 0.18, "learning_rate": 9.388075041225962e-06, "loss": 1.5015, "step": 1729 }, { "epoch": 0.18, "learning_rate": 9.387246684576788e-06, "loss": 1.5506, "step": 1730 }, { "epoch": 0.18, "learning_rate": 9.386417804233836e-06, "loss": 1.8349, "step": 1731 }, { "epoch": 0.18, "learning_rate": 9.385588400296049e-06, "loss": 1.7414, "step": 1732 }, { "epoch": 0.18, "learning_rate": 9.384758472862428e-06, "loss": 1.5273, "step": 1733 }, { "epoch": 0.18, "learning_rate": 9.383928022032044e-06, "loss": 1.5417, "step": 1734 }, { "epoch": 0.19, "learning_rate": 9.38309704790402e-06, "loss": 1.6997, "step": 1735 }, { "epoch": 0.19, "learning_rate": 9.382265550577549e-06, "loss": 1.4466, "step": 1736 }, { "epoch": 0.19, "learning_rate": 9.381433530151887e-06, "loss": 1.3498, "step": 1737 }, { "epoch": 0.19, "learning_rate": 9.380600986726349e-06, "loss": 1.737, "step": 1738 }, { "epoch": 0.19, "learning_rate": 9.379767920400313e-06, "loss": 1.4975, "step": 1739 }, { "epoch": 0.19, "learning_rate": 9.37893433127322e-06, "loss": 1.5079, "step": 1740 }, { "epoch": 0.19, "learning_rate": 9.37810021944457e-06, "loss": 1.2409, "step": 1741 }, { "epoch": 0.19, "learning_rate": 9.377265585013934e-06, "loss": 1.6313, "step": 1742 }, { "epoch": 0.19, "learning_rate": 9.376430428080939e-06, "loss": 1.3931, "step": 1743 }, { "epoch": 0.19, "learning_rate": 9.375594748745271e-06, "loss": 1.4728, "step": 1744 }, { "epoch": 0.19, "learning_rate": 9.374758547106689e-06, "loss": 1.463, "step": 1745 }, { "epoch": 0.19, "learning_rate": 9.373921823265004e-06, "loss": 1.4332, "step": 1746 }, { "epoch": 0.19, "learning_rate": 9.373084577320092e-06, "loss": 1.4985, "step": 1747 }, { "epoch": 0.19, "learning_rate": 9.372246809371898e-06, "loss": 1.6084, "step": 1748 }, { "epoch": 0.19, "learning_rate": 9.371408519520421e-06, "loss": 1.4091, "step": 1749 }, { "epoch": 0.19, "learning_rate": 9.370569707865727e-06, "loss": 1.2691, "step": 1750 }, { "epoch": 0.19, "learning_rate": 9.369730374507942e-06, "loss": 1.5982, "step": 1751 }, { "epoch": 0.19, "learning_rate": 9.36889051954725e-06, "loss": 1.37, "step": 1752 }, { "epoch": 0.19, "learning_rate": 9.368050143083912e-06, "loss": 1.345, "step": 1753 }, { "epoch": 0.19, "learning_rate": 9.367209245218234e-06, "loss": 1.6378, "step": 1754 }, { "epoch": 0.19, "learning_rate": 9.366367826050593e-06, "loss": 1.7364, "step": 1755 }, { "epoch": 0.19, "learning_rate": 9.36552588568143e-06, "loss": 1.5771, "step": 1756 }, { "epoch": 0.19, "learning_rate": 9.364683424211241e-06, "loss": 1.3913, "step": 1757 }, { "epoch": 0.19, "learning_rate": 9.36384044174059e-06, "loss": 1.4624, "step": 1758 }, { "epoch": 0.19, "learning_rate": 9.362996938370103e-06, "loss": 1.3849, "step": 1759 }, { "epoch": 0.19, "learning_rate": 9.362152914200465e-06, "loss": 1.7151, "step": 1760 }, { "epoch": 0.19, "learning_rate": 9.361308369332426e-06, "loss": 1.4235, "step": 1761 }, { "epoch": 0.19, "learning_rate": 9.360463303866795e-06, "loss": 1.8509, "step": 1762 }, { "epoch": 0.19, "learning_rate": 9.359617717904447e-06, "loss": 1.6221, "step": 1763 }, { "epoch": 0.19, "learning_rate": 9.358771611546319e-06, "loss": 1.6042, "step": 1764 }, { "epoch": 0.19, "learning_rate": 9.357924984893405e-06, "loss": 1.5774, "step": 1765 }, { "epoch": 0.19, "learning_rate": 9.357077838046766e-06, "loss": 1.3037, "step": 1766 }, { "epoch": 0.19, "learning_rate": 9.356230171107524e-06, "loss": 1.3218, "step": 1767 }, { "epoch": 0.19, "learning_rate": 9.355381984176861e-06, "loss": 1.414, "step": 1768 }, { "epoch": 0.19, "learning_rate": 9.354533277356026e-06, "loss": 1.5971, "step": 1769 }, { "epoch": 0.19, "learning_rate": 9.353684050746323e-06, "loss": 1.8391, "step": 1770 }, { "epoch": 0.19, "learning_rate": 9.352834304449124e-06, "loss": 1.4944, "step": 1771 }, { "epoch": 0.19, "learning_rate": 9.351984038565862e-06, "loss": 1.759, "step": 1772 }, { "epoch": 0.19, "learning_rate": 9.351133253198027e-06, "loss": 1.7103, "step": 1773 }, { "epoch": 0.19, "learning_rate": 9.350281948447181e-06, "loss": 1.6672, "step": 1774 }, { "epoch": 0.19, "learning_rate": 9.349430124414936e-06, "loss": 1.34, "step": 1775 }, { "epoch": 0.19, "learning_rate": 9.348577781202976e-06, "loss": 1.4682, "step": 1776 }, { "epoch": 0.19, "learning_rate": 9.34772491891304e-06, "loss": 1.3187, "step": 1777 }, { "epoch": 0.19, "learning_rate": 9.346871537646934e-06, "loss": 1.5308, "step": 1778 }, { "epoch": 0.19, "learning_rate": 9.346017637506523e-06, "loss": 1.6612, "step": 1779 }, { "epoch": 0.19, "learning_rate": 9.345163218593735e-06, "loss": 1.6954, "step": 1780 }, { "epoch": 0.19, "learning_rate": 9.34430828101056e-06, "loss": 1.7109, "step": 1781 }, { "epoch": 0.19, "learning_rate": 9.343452824859048e-06, "loss": 1.5577, "step": 1782 }, { "epoch": 0.19, "learning_rate": 9.342596850241313e-06, "loss": 1.56, "step": 1783 }, { "epoch": 0.19, "learning_rate": 9.341740357259532e-06, "loss": 1.4241, "step": 1784 }, { "epoch": 0.19, "learning_rate": 9.340883346015941e-06, "loss": 1.4336, "step": 1785 }, { "epoch": 0.19, "learning_rate": 9.340025816612838e-06, "loss": 1.7394, "step": 1786 }, { "epoch": 0.19, "learning_rate": 9.339167769152588e-06, "loss": 1.5501, "step": 1787 }, { "epoch": 0.19, "learning_rate": 9.338309203737609e-06, "loss": 1.3398, "step": 1788 }, { "epoch": 0.19, "learning_rate": 9.337450120470389e-06, "loss": 1.6255, "step": 1789 }, { "epoch": 0.19, "learning_rate": 9.336590519453472e-06, "loss": 1.7058, "step": 1790 }, { "epoch": 0.19, "learning_rate": 9.335730400789466e-06, "loss": 1.7711, "step": 1791 }, { "epoch": 0.19, "learning_rate": 9.334869764581045e-06, "loss": 1.5941, "step": 1792 }, { "epoch": 0.19, "learning_rate": 9.334008610930939e-06, "loss": 1.5847, "step": 1793 }, { "epoch": 0.19, "learning_rate": 9.333146939941938e-06, "loss": 1.6176, "step": 1794 }, { "epoch": 0.19, "learning_rate": 9.3322847517169e-06, "loss": 1.5001, "step": 1795 }, { "epoch": 0.19, "learning_rate": 9.331422046358745e-06, "loss": 1.801, "step": 1796 }, { "epoch": 0.19, "learning_rate": 9.330558823970448e-06, "loss": 1.349, "step": 1797 }, { "epoch": 0.19, "learning_rate": 9.32969508465505e-06, "loss": 1.3841, "step": 1798 }, { "epoch": 0.19, "learning_rate": 9.328830828515653e-06, "loss": 1.2838, "step": 1799 }, { "epoch": 0.19, "learning_rate": 9.327966055655424e-06, "loss": 1.2387, "step": 1800 }, { "epoch": 0.19, "learning_rate": 9.327100766177585e-06, "loss": 1.4902, "step": 1801 }, { "epoch": 0.19, "learning_rate": 9.326234960185424e-06, "loss": 1.4069, "step": 1802 }, { "epoch": 0.19, "learning_rate": 9.325368637782292e-06, "loss": 1.4877, "step": 1803 }, { "epoch": 0.19, "learning_rate": 9.324501799071597e-06, "loss": 1.4172, "step": 1804 }, { "epoch": 0.19, "learning_rate": 9.323634444156813e-06, "loss": 1.4725, "step": 1805 }, { "epoch": 0.19, "learning_rate": 9.322766573141473e-06, "loss": 1.5818, "step": 1806 }, { "epoch": 0.19, "learning_rate": 9.321898186129172e-06, "loss": 1.6253, "step": 1807 }, { "epoch": 0.19, "learning_rate": 9.321029283223567e-06, "loss": 1.6483, "step": 1808 }, { "epoch": 0.19, "learning_rate": 9.320159864528378e-06, "loss": 1.4629, "step": 1809 }, { "epoch": 0.19, "learning_rate": 9.319289930147383e-06, "loss": 1.6033, "step": 1810 }, { "epoch": 0.19, "learning_rate": 9.318419480184427e-06, "loss": 1.4291, "step": 1811 }, { "epoch": 0.19, "learning_rate": 9.31754851474341e-06, "loss": 1.5927, "step": 1812 }, { "epoch": 0.19, "learning_rate": 9.316677033928298e-06, "loss": 1.6195, "step": 1813 }, { "epoch": 0.19, "learning_rate": 9.315805037843118e-06, "loss": 1.5628, "step": 1814 }, { "epoch": 0.19, "learning_rate": 9.314932526591956e-06, "loss": 1.791, "step": 1815 }, { "epoch": 0.19, "learning_rate": 9.314059500278962e-06, "loss": 1.4376, "step": 1816 }, { "epoch": 0.19, "learning_rate": 9.31318595900835e-06, "loss": 1.6722, "step": 1817 }, { "epoch": 0.19, "learning_rate": 9.312311902884388e-06, "loss": 1.4915, "step": 1818 }, { "epoch": 0.19, "learning_rate": 9.311437332011411e-06, "loss": 1.4871, "step": 1819 }, { "epoch": 0.19, "learning_rate": 9.310562246493812e-06, "loss": 1.5692, "step": 1820 }, { "epoch": 0.19, "learning_rate": 9.309686646436053e-06, "loss": 1.2398, "step": 1821 }, { "epoch": 0.19, "learning_rate": 9.30881053194265e-06, "loss": 1.6596, "step": 1822 }, { "epoch": 0.19, "learning_rate": 9.30793390311818e-06, "loss": 1.6558, "step": 1823 }, { "epoch": 0.19, "learning_rate": 9.307056760067284e-06, "loss": 1.5882, "step": 1824 }, { "epoch": 0.19, "learning_rate": 9.306179102894667e-06, "loss": 1.4704, "step": 1825 }, { "epoch": 0.19, "learning_rate": 9.30530093170509e-06, "loss": 1.2524, "step": 1826 }, { "epoch": 0.19, "learning_rate": 9.30442224660338e-06, "loss": 1.4704, "step": 1827 }, { "epoch": 0.19, "learning_rate": 9.30354304769442e-06, "loss": 1.354, "step": 1828 }, { "epoch": 0.2, "learning_rate": 9.302663335083161e-06, "loss": 1.5299, "step": 1829 }, { "epoch": 0.2, "learning_rate": 9.301783108874611e-06, "loss": 1.4848, "step": 1830 }, { "epoch": 0.2, "learning_rate": 9.300902369173841e-06, "loss": 1.6713, "step": 1831 }, { "epoch": 0.2, "learning_rate": 9.30002111608598e-06, "loss": 1.6087, "step": 1832 }, { "epoch": 0.2, "learning_rate": 9.299139349716221e-06, "loss": 1.4341, "step": 1833 }, { "epoch": 0.2, "learning_rate": 9.298257070169822e-06, "loss": 1.3799, "step": 1834 }, { "epoch": 0.2, "learning_rate": 9.297374277552094e-06, "loss": 1.3346, "step": 1835 }, { "epoch": 0.2, "learning_rate": 9.296490971968416e-06, "loss": 1.6966, "step": 1836 }, { "epoch": 0.2, "learning_rate": 9.295607153524224e-06, "loss": 1.3302, "step": 1837 }, { "epoch": 0.2, "learning_rate": 9.29472282232502e-06, "loss": 1.339, "step": 1838 }, { "epoch": 0.2, "learning_rate": 9.293837978476359e-06, "loss": 1.4446, "step": 1839 }, { "epoch": 0.2, "learning_rate": 9.292952622083867e-06, "loss": 1.6251, "step": 1840 }, { "epoch": 0.2, "learning_rate": 9.292066753253226e-06, "loss": 1.4267, "step": 1841 }, { "epoch": 0.2, "learning_rate": 9.291180372090178e-06, "loss": 1.7346, "step": 1842 }, { "epoch": 0.2, "learning_rate": 9.29029347870053e-06, "loss": 1.3882, "step": 1843 }, { "epoch": 0.2, "learning_rate": 9.289406073190146e-06, "loss": 1.3257, "step": 1844 }, { "epoch": 0.2, "learning_rate": 9.288518155664956e-06, "loss": 1.4599, "step": 1845 }, { "epoch": 0.2, "learning_rate": 9.287629726230945e-06, "loss": 1.5345, "step": 1846 }, { "epoch": 0.2, "learning_rate": 9.286740784994164e-06, "loss": 1.2952, "step": 1847 }, { "epoch": 0.2, "learning_rate": 9.285851332060722e-06, "loss": 1.4058, "step": 1848 }, { "epoch": 0.2, "learning_rate": 9.284961367536795e-06, "loss": 1.2377, "step": 1849 }, { "epoch": 0.2, "learning_rate": 9.28407089152861e-06, "loss": 1.7734, "step": 1850 }, { "epoch": 0.2, "learning_rate": 9.283179904142465e-06, "loss": 1.617, "step": 1851 }, { "epoch": 0.2, "learning_rate": 9.282288405484712e-06, "loss": 1.6365, "step": 1852 }, { "epoch": 0.2, "learning_rate": 9.281396395661771e-06, "loss": 1.4746, "step": 1853 }, { "epoch": 0.2, "learning_rate": 9.280503874780112e-06, "loss": 1.5297, "step": 1854 }, { "epoch": 0.2, "learning_rate": 9.279610842946278e-06, "loss": 1.6545, "step": 1855 }, { "epoch": 0.2, "learning_rate": 9.278717300266866e-06, "loss": 1.7182, "step": 1856 }, { "epoch": 0.2, "learning_rate": 9.277823246848537e-06, "loss": 1.7672, "step": 1857 }, { "epoch": 0.2, "learning_rate": 9.27692868279801e-06, "loss": 1.2983, "step": 1858 }, { "epoch": 0.2, "learning_rate": 9.276033608222068e-06, "loss": 1.3573, "step": 1859 }, { "epoch": 0.2, "learning_rate": 9.275138023227555e-06, "loss": 1.2659, "step": 1860 }, { "epoch": 0.2, "learning_rate": 9.27424192792137e-06, "loss": 1.6648, "step": 1861 }, { "epoch": 0.2, "learning_rate": 9.27334532241048e-06, "loss": 1.5149, "step": 1862 }, { "epoch": 0.2, "learning_rate": 9.272448206801912e-06, "loss": 1.3951, "step": 1863 }, { "epoch": 0.2, "learning_rate": 9.27155058120275e-06, "loss": 1.4055, "step": 1864 }, { "epoch": 0.2, "learning_rate": 9.270652445720143e-06, "loss": 1.7201, "step": 1865 }, { "epoch": 0.2, "learning_rate": 9.269753800461299e-06, "loss": 1.5468, "step": 1866 }, { "epoch": 0.2, "learning_rate": 9.268854645533483e-06, "loss": 1.4347, "step": 1867 }, { "epoch": 0.2, "learning_rate": 9.26795498104403e-06, "loss": 1.563, "step": 1868 }, { "epoch": 0.2, "learning_rate": 9.267054807100327e-06, "loss": 1.7417, "step": 1869 }, { "epoch": 0.2, "learning_rate": 9.266154123809825e-06, "loss": 1.3142, "step": 1870 }, { "epoch": 0.2, "learning_rate": 9.26525293128004e-06, "loss": 1.6165, "step": 1871 }, { "epoch": 0.2, "learning_rate": 9.264351229618541e-06, "loss": 1.5735, "step": 1872 }, { "epoch": 0.2, "learning_rate": 9.263449018932964e-06, "loss": 1.669, "step": 1873 }, { "epoch": 0.2, "learning_rate": 9.262546299331e-06, "loss": 1.306, "step": 1874 }, { "epoch": 0.2, "learning_rate": 9.261643070920409e-06, "loss": 1.7331, "step": 1875 }, { "epoch": 0.2, "learning_rate": 9.260739333809006e-06, "loss": 1.4793, "step": 1876 }, { "epoch": 0.2, "learning_rate": 9.259835088104665e-06, "loss": 1.5012, "step": 1877 }, { "epoch": 0.2, "learning_rate": 9.258930333915325e-06, "loss": 1.384, "step": 1878 }, { "epoch": 0.2, "learning_rate": 9.258025071348984e-06, "loss": 1.6816, "step": 1879 }, { "epoch": 0.2, "learning_rate": 9.2571193005137e-06, "loss": 1.3034, "step": 1880 }, { "epoch": 0.2, "learning_rate": 9.256213021517593e-06, "loss": 1.3709, "step": 1881 }, { "epoch": 0.2, "learning_rate": 9.255306234468844e-06, "loss": 1.4469, "step": 1882 }, { "epoch": 0.2, "learning_rate": 9.25439893947569e-06, "loss": 1.5843, "step": 1883 }, { "epoch": 0.2, "learning_rate": 9.253491136646437e-06, "loss": 1.6804, "step": 1884 }, { "epoch": 0.2, "learning_rate": 9.252582826089447e-06, "loss": 1.2899, "step": 1885 }, { "epoch": 0.2, "learning_rate": 9.251674007913138e-06, "loss": 1.6265, "step": 1886 }, { "epoch": 0.2, "learning_rate": 9.250764682225997e-06, "loss": 1.648, "step": 1887 }, { "epoch": 0.2, "learning_rate": 9.249854849136566e-06, "loss": 1.2679, "step": 1888 }, { "epoch": 0.2, "learning_rate": 9.24894450875345e-06, "loss": 1.5363, "step": 1889 }, { "epoch": 0.2, "learning_rate": 9.248033661185313e-06, "loss": 1.5161, "step": 1890 }, { "epoch": 0.2, "learning_rate": 9.247122306540882e-06, "loss": 1.7021, "step": 1891 }, { "epoch": 0.2, "learning_rate": 9.246210444928942e-06, "loss": 1.5235, "step": 1892 }, { "epoch": 0.2, "learning_rate": 9.24529807645834e-06, "loss": 1.5167, "step": 1893 }, { "epoch": 0.2, "learning_rate": 9.244385201237982e-06, "loss": 1.5627, "step": 1894 }, { "epoch": 0.2, "learning_rate": 9.243471819376837e-06, "loss": 1.3728, "step": 1895 }, { "epoch": 0.2, "learning_rate": 9.24255793098393e-06, "loss": 1.244, "step": 1896 }, { "epoch": 0.2, "learning_rate": 9.241643536168351e-06, "loss": 1.7508, "step": 1897 }, { "epoch": 0.2, "learning_rate": 9.240728635039252e-06, "loss": 1.5745, "step": 1898 }, { "epoch": 0.2, "learning_rate": 9.23981322770584e-06, "loss": 1.5927, "step": 1899 }, { "epoch": 0.2, "learning_rate": 9.23889731427738e-06, "loss": 1.3942, "step": 1900 }, { "epoch": 0.2, "learning_rate": 9.237980894863208e-06, "loss": 1.4488, "step": 1901 }, { "epoch": 0.2, "learning_rate": 9.237063969572713e-06, "loss": 1.5548, "step": 1902 }, { "epoch": 0.2, "learning_rate": 9.236146538515345e-06, "loss": 1.3197, "step": 1903 }, { "epoch": 0.2, "learning_rate": 9.235228601800616e-06, "loss": 1.4884, "step": 1904 }, { "epoch": 0.2, "learning_rate": 9.2343101595381e-06, "loss": 1.4642, "step": 1905 }, { "epoch": 0.2, "learning_rate": 9.233391211837423e-06, "loss": 1.532, "step": 1906 }, { "epoch": 0.2, "learning_rate": 9.232471758808282e-06, "loss": 1.493, "step": 1907 }, { "epoch": 0.2, "learning_rate": 9.23155180056043e-06, "loss": 1.5074, "step": 1908 }, { "epoch": 0.2, "learning_rate": 9.230631337203675e-06, "loss": 1.6479, "step": 1909 }, { "epoch": 0.2, "learning_rate": 9.229710368847896e-06, "loss": 1.512, "step": 1910 }, { "epoch": 0.2, "learning_rate": 9.228788895603024e-06, "loss": 1.5565, "step": 1911 }, { "epoch": 0.2, "learning_rate": 9.22786691757905e-06, "loss": 1.2345, "step": 1912 }, { "epoch": 0.2, "learning_rate": 9.226944434886034e-06, "loss": 1.358, "step": 1913 }, { "epoch": 0.2, "learning_rate": 9.226021447634085e-06, "loss": 1.4803, "step": 1914 }, { "epoch": 0.2, "learning_rate": 9.225097955933382e-06, "loss": 1.5036, "step": 1915 }, { "epoch": 0.2, "learning_rate": 9.224173959894157e-06, "loss": 1.5025, "step": 1916 }, { "epoch": 0.2, "learning_rate": 9.223249459626704e-06, "loss": 1.311, "step": 1917 }, { "epoch": 0.2, "learning_rate": 9.22232445524138e-06, "loss": 1.2955, "step": 1918 }, { "epoch": 0.2, "learning_rate": 9.221398946848598e-06, "loss": 1.5269, "step": 1919 }, { "epoch": 0.2, "learning_rate": 9.220472934558838e-06, "loss": 1.3515, "step": 1920 }, { "epoch": 0.2, "learning_rate": 9.21954641848263e-06, "loss": 1.3295, "step": 1921 }, { "epoch": 0.21, "learning_rate": 9.218619398730572e-06, "loss": 1.3655, "step": 1922 }, { "epoch": 0.21, "learning_rate": 9.217691875413323e-06, "loss": 1.321, "step": 1923 }, { "epoch": 0.21, "learning_rate": 9.216763848641595e-06, "loss": 1.4948, "step": 1924 }, { "epoch": 0.21, "learning_rate": 9.215835318526164e-06, "loss": 1.4953, "step": 1925 }, { "epoch": 0.21, "learning_rate": 9.214906285177867e-06, "loss": 1.5617, "step": 1926 }, { "epoch": 0.21, "learning_rate": 9.213976748707602e-06, "loss": 1.2829, "step": 1927 }, { "epoch": 0.21, "learning_rate": 9.213046709226323e-06, "loss": 1.3759, "step": 1928 }, { "epoch": 0.21, "learning_rate": 9.212116166845048e-06, "loss": 1.7027, "step": 1929 }, { "epoch": 0.21, "learning_rate": 9.211185121674851e-06, "loss": 1.6979, "step": 1930 }, { "epoch": 0.21, "learning_rate": 9.21025357382687e-06, "loss": 1.5128, "step": 1931 }, { "epoch": 0.21, "learning_rate": 9.209321523412303e-06, "loss": 1.6887, "step": 1932 }, { "epoch": 0.21, "learning_rate": 9.208388970542401e-06, "loss": 1.6978, "step": 1933 }, { "epoch": 0.21, "learning_rate": 9.207455915328487e-06, "loss": 1.7236, "step": 1934 }, { "epoch": 0.21, "learning_rate": 9.206522357881931e-06, "loss": 1.5627, "step": 1935 }, { "epoch": 0.21, "learning_rate": 9.205588298314175e-06, "loss": 1.6764, "step": 1936 }, { "epoch": 0.21, "learning_rate": 9.204653736736711e-06, "loss": 1.7665, "step": 1937 }, { "epoch": 0.21, "learning_rate": 9.203718673261098e-06, "loss": 1.3771, "step": 1938 }, { "epoch": 0.21, "learning_rate": 9.202783107998947e-06, "loss": 1.3684, "step": 1939 }, { "epoch": 0.21, "learning_rate": 9.20184704106194e-06, "loss": 1.5233, "step": 1940 }, { "epoch": 0.21, "learning_rate": 9.20091047256181e-06, "loss": 1.8691, "step": 1941 }, { "epoch": 0.21, "learning_rate": 9.199973402610351e-06, "loss": 1.6403, "step": 1942 }, { "epoch": 0.21, "learning_rate": 9.199035831319422e-06, "loss": 1.8792, "step": 1943 }, { "epoch": 0.21, "learning_rate": 9.198097758800938e-06, "loss": 1.6043, "step": 1944 }, { "epoch": 0.21, "learning_rate": 9.197159185166871e-06, "loss": 1.448, "step": 1945 }, { "epoch": 0.21, "learning_rate": 9.19622011052926e-06, "loss": 1.4674, "step": 1946 }, { "epoch": 0.21, "learning_rate": 9.195280535000196e-06, "loss": 1.5641, "step": 1947 }, { "epoch": 0.21, "learning_rate": 9.194340458691833e-06, "loss": 1.4025, "step": 1948 }, { "epoch": 0.21, "learning_rate": 9.19339988171639e-06, "loss": 1.2988, "step": 1949 }, { "epoch": 0.21, "learning_rate": 9.19245880418614e-06, "loss": 1.5248, "step": 1950 }, { "epoch": 0.21, "learning_rate": 9.191517226213413e-06, "loss": 1.3648, "step": 1951 }, { "epoch": 0.21, "learning_rate": 9.190575147910606e-06, "loss": 1.4951, "step": 1952 }, { "epoch": 0.21, "learning_rate": 9.189632569390172e-06, "loss": 1.4757, "step": 1953 }, { "epoch": 0.21, "learning_rate": 9.188689490764624e-06, "loss": 1.4179, "step": 1954 }, { "epoch": 0.21, "learning_rate": 9.187745912146535e-06, "loss": 1.6026, "step": 1955 }, { "epoch": 0.21, "learning_rate": 9.186801833648535e-06, "loss": 1.3961, "step": 1956 }, { "epoch": 0.21, "learning_rate": 9.185857255383319e-06, "loss": 1.2057, "step": 1957 }, { "epoch": 0.21, "learning_rate": 9.184912177463637e-06, "loss": 1.2159, "step": 1958 }, { "epoch": 0.21, "learning_rate": 9.183966600002301e-06, "loss": 1.5255, "step": 1959 }, { "epoch": 0.21, "learning_rate": 9.183020523112183e-06, "loss": 1.6445, "step": 1960 }, { "epoch": 0.21, "learning_rate": 9.182073946906212e-06, "loss": 1.4147, "step": 1961 }, { "epoch": 0.21, "learning_rate": 9.181126871497378e-06, "loss": 1.5556, "step": 1962 }, { "epoch": 0.21, "learning_rate": 9.180179296998733e-06, "loss": 1.5594, "step": 1963 }, { "epoch": 0.21, "learning_rate": 9.179231223523385e-06, "loss": 1.4115, "step": 1964 }, { "epoch": 0.21, "learning_rate": 9.178282651184506e-06, "loss": 1.6397, "step": 1965 }, { "epoch": 0.21, "learning_rate": 9.177333580095318e-06, "loss": 1.7444, "step": 1966 }, { "epoch": 0.21, "learning_rate": 9.176384010369113e-06, "loss": 1.5014, "step": 1967 }, { "epoch": 0.21, "learning_rate": 9.175433942119238e-06, "loss": 1.4228, "step": 1968 }, { "epoch": 0.21, "learning_rate": 9.174483375459102e-06, "loss": 1.4272, "step": 1969 }, { "epoch": 0.21, "learning_rate": 9.173532310502169e-06, "loss": 1.338, "step": 1970 }, { "epoch": 0.21, "learning_rate": 9.172580747361968e-06, "loss": 1.4283, "step": 1971 }, { "epoch": 0.21, "learning_rate": 9.17162868615208e-06, "loss": 1.3997, "step": 1972 }, { "epoch": 0.21, "learning_rate": 9.170676126986154e-06, "loss": 1.3832, "step": 1973 }, { "epoch": 0.21, "learning_rate": 9.169723069977892e-06, "loss": 1.3671, "step": 1974 }, { "epoch": 0.21, "learning_rate": 9.16876951524106e-06, "loss": 1.6028, "step": 1975 }, { "epoch": 0.21, "learning_rate": 9.167815462889477e-06, "loss": 1.4742, "step": 1976 }, { "epoch": 0.21, "learning_rate": 9.166860913037032e-06, "loss": 1.5909, "step": 1977 }, { "epoch": 0.21, "learning_rate": 9.165905865797661e-06, "loss": 1.6241, "step": 1978 }, { "epoch": 0.21, "learning_rate": 9.164950321285371e-06, "loss": 1.6564, "step": 1979 }, { "epoch": 0.21, "learning_rate": 9.163994279614218e-06, "loss": 1.5938, "step": 1980 }, { "epoch": 0.21, "learning_rate": 9.163037740898324e-06, "loss": 1.648, "step": 1981 }, { "epoch": 0.21, "learning_rate": 9.16208070525187e-06, "loss": 1.6044, "step": 1982 }, { "epoch": 0.21, "learning_rate": 9.161123172789091e-06, "loss": 1.3955, "step": 1983 }, { "epoch": 0.21, "learning_rate": 9.160165143624289e-06, "loss": 1.5837, "step": 1984 }, { "epoch": 0.21, "learning_rate": 9.159206617871819e-06, "loss": 1.6568, "step": 1985 }, { "epoch": 0.21, "learning_rate": 9.158247595646098e-06, "loss": 1.3284, "step": 1986 }, { "epoch": 0.21, "learning_rate": 9.1572880770616e-06, "loss": 1.6656, "step": 1987 }, { "epoch": 0.21, "learning_rate": 9.156328062232867e-06, "loss": 1.3562, "step": 1988 }, { "epoch": 0.21, "learning_rate": 9.155367551274485e-06, "loss": 1.349, "step": 1989 }, { "epoch": 0.21, "learning_rate": 9.154406544301113e-06, "loss": 1.3918, "step": 1990 }, { "epoch": 0.21, "learning_rate": 9.15344504142746e-06, "loss": 1.3725, "step": 1991 }, { "epoch": 0.21, "learning_rate": 9.152483042768302e-06, "loss": 1.4764, "step": 1992 }, { "epoch": 0.21, "learning_rate": 9.151520548438468e-06, "loss": 1.4958, "step": 1993 }, { "epoch": 0.21, "learning_rate": 9.150557558552847e-06, "loss": 1.3047, "step": 1994 }, { "epoch": 0.21, "learning_rate": 9.149594073226391e-06, "loss": 1.421, "step": 1995 }, { "epoch": 0.21, "learning_rate": 9.148630092574109e-06, "loss": 1.3741, "step": 1996 }, { "epoch": 0.21, "learning_rate": 9.147665616711065e-06, "loss": 1.6438, "step": 1997 }, { "epoch": 0.21, "learning_rate": 9.14670064575239e-06, "loss": 1.2666, "step": 1998 }, { "epoch": 0.21, "learning_rate": 9.145735179813269e-06, "loss": 1.2066, "step": 1999 }, { "epoch": 0.21, "learning_rate": 9.144769219008945e-06, "loss": 1.4423, "step": 2000 }, { "epoch": 0.21, "learning_rate": 9.143802763454723e-06, "loss": 1.4914, "step": 2001 }, { "epoch": 0.21, "learning_rate": 9.142835813265966e-06, "loss": 1.4626, "step": 2002 }, { "epoch": 0.21, "learning_rate": 9.1418683685581e-06, "loss": 1.4786, "step": 2003 }, { "epoch": 0.21, "learning_rate": 9.140900429446601e-06, "loss": 1.5594, "step": 2004 }, { "epoch": 0.21, "learning_rate": 9.139931996047012e-06, "loss": 1.8095, "step": 2005 }, { "epoch": 0.21, "learning_rate": 9.13896306847493e-06, "loss": 1.3602, "step": 2006 }, { "epoch": 0.21, "learning_rate": 9.137993646846018e-06, "loss": 1.3634, "step": 2007 }, { "epoch": 0.21, "learning_rate": 9.137023731275988e-06, "loss": 1.4693, "step": 2008 }, { "epoch": 0.21, "learning_rate": 9.13605332188062e-06, "loss": 1.3013, "step": 2009 }, { "epoch": 0.21, "learning_rate": 9.135082418775746e-06, "loss": 1.4538, "step": 2010 }, { "epoch": 0.21, "learning_rate": 9.134111022077263e-06, "loss": 1.4048, "step": 2011 }, { "epoch": 0.21, "learning_rate": 9.133139131901123e-06, "loss": 1.5946, "step": 2012 }, { "epoch": 0.21, "learning_rate": 9.132166748363335e-06, "loss": 1.5736, "step": 2013 }, { "epoch": 0.21, "learning_rate": 9.131193871579975e-06, "loss": 1.3468, "step": 2014 }, { "epoch": 0.21, "learning_rate": 9.130220501667168e-06, "loss": 1.4065, "step": 2015 }, { "epoch": 0.22, "learning_rate": 9.129246638741108e-06, "loss": 1.3223, "step": 2016 }, { "epoch": 0.22, "learning_rate": 9.128272282918036e-06, "loss": 1.4715, "step": 2017 }, { "epoch": 0.22, "learning_rate": 9.127297434314262e-06, "loss": 1.6067, "step": 2018 }, { "epoch": 0.22, "learning_rate": 9.126322093046149e-06, "loss": 1.5259, "step": 2019 }, { "epoch": 0.22, "learning_rate": 9.125346259230123e-06, "loss": 1.5028, "step": 2020 }, { "epoch": 0.22, "learning_rate": 9.124369932982667e-06, "loss": 1.3892, "step": 2021 }, { "epoch": 0.22, "learning_rate": 9.123393114420318e-06, "loss": 1.631, "step": 2022 }, { "epoch": 0.22, "learning_rate": 9.12241580365968e-06, "loss": 1.3613, "step": 2023 }, { "epoch": 0.22, "learning_rate": 9.121438000817413e-06, "loss": 1.5029, "step": 2024 }, { "epoch": 0.22, "learning_rate": 9.120459706010233e-06, "loss": 1.5162, "step": 2025 }, { "epoch": 0.22, "learning_rate": 9.119480919354916e-06, "loss": 1.4869, "step": 2026 }, { "epoch": 0.22, "learning_rate": 9.118501640968295e-06, "loss": 1.4352, "step": 2027 }, { "epoch": 0.22, "learning_rate": 9.11752187096727e-06, "loss": 1.4511, "step": 2028 }, { "epoch": 0.22, "learning_rate": 9.116541609468788e-06, "loss": 1.5377, "step": 2029 }, { "epoch": 0.22, "learning_rate": 9.115560856589863e-06, "loss": 1.3945, "step": 2030 }, { "epoch": 0.22, "learning_rate": 9.114579612447562e-06, "loss": 1.4603, "step": 2031 }, { "epoch": 0.22, "learning_rate": 9.113597877159014e-06, "loss": 1.5513, "step": 2032 }, { "epoch": 0.22, "learning_rate": 9.112615650841412e-06, "loss": 1.3196, "step": 2033 }, { "epoch": 0.22, "learning_rate": 9.111632933611993e-06, "loss": 1.6372, "step": 2034 }, { "epoch": 0.22, "learning_rate": 9.110649725588067e-06, "loss": 1.2836, "step": 2035 }, { "epoch": 0.22, "learning_rate": 9.109666026886995e-06, "loss": 1.4386, "step": 2036 }, { "epoch": 0.22, "learning_rate": 9.108681837626199e-06, "loss": 1.5084, "step": 2037 }, { "epoch": 0.22, "learning_rate": 9.107697157923156e-06, "loss": 1.7213, "step": 2038 }, { "epoch": 0.22, "learning_rate": 9.106711987895411e-06, "loss": 1.3987, "step": 2039 }, { "epoch": 0.22, "learning_rate": 9.105726327660556e-06, "loss": 1.6095, "step": 2040 }, { "epoch": 0.22, "learning_rate": 9.104740177336246e-06, "loss": 1.479, "step": 2041 }, { "epoch": 0.22, "learning_rate": 9.103753537040199e-06, "loss": 1.5356, "step": 2042 }, { "epoch": 0.22, "learning_rate": 9.102766406890185e-06, "loss": 1.5671, "step": 2043 }, { "epoch": 0.22, "learning_rate": 9.101778787004037e-06, "loss": 1.3237, "step": 2044 }, { "epoch": 0.22, "learning_rate": 9.10079067749964e-06, "loss": 1.3165, "step": 2045 }, { "epoch": 0.22, "learning_rate": 9.099802078494947e-06, "loss": 1.4425, "step": 2046 }, { "epoch": 0.22, "learning_rate": 9.098812990107964e-06, "loss": 1.819, "step": 2047 }, { "epoch": 0.22, "learning_rate": 9.097823412456752e-06, "loss": 1.5722, "step": 2048 }, { "epoch": 0.22, "learning_rate": 9.096833345659437e-06, "loss": 1.8692, "step": 2049 }, { "epoch": 0.22, "learning_rate": 9.0958427898342e-06, "loss": 1.4527, "step": 2050 }, { "epoch": 0.22, "learning_rate": 9.094851745099282e-06, "loss": 1.4189, "step": 2051 }, { "epoch": 0.22, "learning_rate": 9.09386021157298e-06, "loss": 1.6325, "step": 2052 }, { "epoch": 0.22, "learning_rate": 9.092868189373651e-06, "loss": 1.6374, "step": 2053 }, { "epoch": 0.22, "learning_rate": 9.09187567861971e-06, "loss": 1.527, "step": 2054 }, { "epoch": 0.22, "learning_rate": 9.09088267942963e-06, "loss": 1.307, "step": 2055 }, { "epoch": 0.22, "learning_rate": 9.089889191921944e-06, "loss": 1.7394, "step": 2056 }, { "epoch": 0.22, "learning_rate": 9.088895216215241e-06, "loss": 1.3755, "step": 2057 }, { "epoch": 0.22, "learning_rate": 9.087900752428168e-06, "loss": 1.4717, "step": 2058 }, { "epoch": 0.22, "learning_rate": 9.086905800679433e-06, "loss": 1.6955, "step": 2059 }, { "epoch": 0.22, "learning_rate": 9.085910361087802e-06, "loss": 1.5735, "step": 2060 }, { "epoch": 0.22, "learning_rate": 9.084914433772094e-06, "loss": 1.4119, "step": 2061 }, { "epoch": 0.22, "learning_rate": 9.083918018851193e-06, "loss": 1.4362, "step": 2062 }, { "epoch": 0.22, "learning_rate": 9.082921116444038e-06, "loss": 1.5534, "step": 2063 }, { "epoch": 0.22, "learning_rate": 9.081923726669626e-06, "loss": 1.4665, "step": 2064 }, { "epoch": 0.22, "learning_rate": 9.080925849647014e-06, "loss": 1.6517, "step": 2065 }, { "epoch": 0.22, "learning_rate": 9.079927485495314e-06, "loss": 1.4826, "step": 2066 }, { "epoch": 0.22, "learning_rate": 9.0789286343337e-06, "loss": 1.3057, "step": 2067 }, { "epoch": 0.22, "learning_rate": 9.0779292962814e-06, "loss": 1.5605, "step": 2068 }, { "epoch": 0.22, "learning_rate": 9.076929471457704e-06, "loss": 1.2754, "step": 2069 }, { "epoch": 0.22, "learning_rate": 9.075929159981957e-06, "loss": 1.5692, "step": 2070 }, { "epoch": 0.22, "learning_rate": 9.074928361973565e-06, "loss": 1.6587, "step": 2071 }, { "epoch": 0.22, "learning_rate": 9.073927077551989e-06, "loss": 1.5595, "step": 2072 }, { "epoch": 0.22, "learning_rate": 9.072925306836751e-06, "loss": 1.4073, "step": 2073 }, { "epoch": 0.22, "learning_rate": 9.071923049947429e-06, "loss": 1.6732, "step": 2074 }, { "epoch": 0.22, "learning_rate": 9.07092030700366e-06, "loss": 1.4921, "step": 2075 }, { "epoch": 0.22, "learning_rate": 9.06991707812514e-06, "loss": 1.5617, "step": 2076 }, { "epoch": 0.22, "learning_rate": 9.068913363431617e-06, "loss": 1.5621, "step": 2077 }, { "epoch": 0.22, "learning_rate": 9.067909163042907e-06, "loss": 1.2487, "step": 2078 }, { "epoch": 0.22, "learning_rate": 9.066904477078875e-06, "loss": 1.4082, "step": 2079 }, { "epoch": 0.22, "learning_rate": 9.065899305659452e-06, "loss": 1.3902, "step": 2080 }, { "epoch": 0.22, "learning_rate": 9.064893648904617e-06, "loss": 1.5927, "step": 2081 }, { "epoch": 0.22, "learning_rate": 9.063887506934417e-06, "loss": 1.3336, "step": 2082 }, { "epoch": 0.22, "learning_rate": 9.06288087986895e-06, "loss": 1.6543, "step": 2083 }, { "epoch": 0.22, "learning_rate": 9.061873767828375e-06, "loss": 1.2651, "step": 2084 }, { "epoch": 0.22, "learning_rate": 9.06086617093291e-06, "loss": 1.6874, "step": 2085 }, { "epoch": 0.22, "learning_rate": 9.059858089302825e-06, "loss": 1.7514, "step": 2086 }, { "epoch": 0.22, "learning_rate": 9.058849523058457e-06, "loss": 1.6401, "step": 2087 }, { "epoch": 0.22, "learning_rate": 9.057840472320192e-06, "loss": 1.2314, "step": 2088 }, { "epoch": 0.22, "learning_rate": 9.056830937208478e-06, "loss": 1.701, "step": 2089 }, { "epoch": 0.22, "learning_rate": 9.055820917843825e-06, "loss": 1.6884, "step": 2090 }, { "epoch": 0.22, "learning_rate": 9.054810414346789e-06, "loss": 1.4312, "step": 2091 }, { "epoch": 0.22, "learning_rate": 9.053799426837997e-06, "loss": 1.2809, "step": 2092 }, { "epoch": 0.22, "learning_rate": 9.052787955438125e-06, "loss": 1.1908, "step": 2093 }, { "epoch": 0.22, "learning_rate": 9.05177600026791e-06, "loss": 1.5952, "step": 2094 }, { "epoch": 0.22, "learning_rate": 9.050763561448147e-06, "loss": 1.3267, "step": 2095 }, { "epoch": 0.22, "learning_rate": 9.049750639099689e-06, "loss": 1.4007, "step": 2096 }, { "epoch": 0.22, "learning_rate": 9.048737233343442e-06, "loss": 1.6816, "step": 2097 }, { "epoch": 0.22, "learning_rate": 9.047723344300376e-06, "loss": 1.3606, "step": 2098 }, { "epoch": 0.22, "learning_rate": 9.046708972091519e-06, "loss": 1.8285, "step": 2099 }, { "epoch": 0.22, "learning_rate": 9.045694116837948e-06, "loss": 1.388, "step": 2100 }, { "epoch": 0.22, "learning_rate": 9.044678778660808e-06, "loss": 1.5639, "step": 2101 }, { "epoch": 0.22, "learning_rate": 9.043662957681297e-06, "loss": 1.5511, "step": 2102 }, { "epoch": 0.22, "learning_rate": 9.042646654020667e-06, "loss": 1.5509, "step": 2103 }, { "epoch": 0.22, "learning_rate": 9.041629867800236e-06, "loss": 1.404, "step": 2104 }, { "epoch": 0.22, "learning_rate": 9.040612599141375e-06, "loss": 1.3636, "step": 2105 }, { "epoch": 0.22, "learning_rate": 9.039594848165507e-06, "loss": 1.472, "step": 2106 }, { "epoch": 0.22, "learning_rate": 9.038576614994124e-06, "loss": 1.5877, "step": 2107 }, { "epoch": 0.22, "learning_rate": 9.037557899748765e-06, "loss": 1.6624, "step": 2108 }, { "epoch": 0.22, "learning_rate": 9.036538702551037e-06, "loss": 1.4413, "step": 2109 }, { "epoch": 0.23, "learning_rate": 9.035519023522592e-06, "loss": 1.6925, "step": 2110 }, { "epoch": 0.23, "learning_rate": 9.034498862785152e-06, "loss": 1.5694, "step": 2111 }, { "epoch": 0.23, "learning_rate": 9.03347822046049e-06, "loss": 1.2848, "step": 2112 }, { "epoch": 0.23, "learning_rate": 9.032457096670435e-06, "loss": 1.5716, "step": 2113 }, { "epoch": 0.23, "learning_rate": 9.031435491536877e-06, "loss": 1.6232, "step": 2114 }, { "epoch": 0.23, "learning_rate": 9.03041340518176e-06, "loss": 1.4126, "step": 2115 }, { "epoch": 0.23, "learning_rate": 9.029390837727094e-06, "loss": 1.5124, "step": 2116 }, { "epoch": 0.23, "learning_rate": 9.028367789294934e-06, "loss": 1.5688, "step": 2117 }, { "epoch": 0.23, "learning_rate": 9.027344260007401e-06, "loss": 1.5221, "step": 2118 }, { "epoch": 0.23, "learning_rate": 9.02632024998667e-06, "loss": 1.5825, "step": 2119 }, { "epoch": 0.23, "learning_rate": 9.025295759354976e-06, "loss": 1.3551, "step": 2120 }, { "epoch": 0.23, "learning_rate": 9.024270788234606e-06, "loss": 1.2921, "step": 2121 }, { "epoch": 0.23, "learning_rate": 9.023245336747911e-06, "loss": 1.4206, "step": 2122 }, { "epoch": 0.23, "learning_rate": 9.022219405017296e-06, "loss": 1.9157, "step": 2123 }, { "epoch": 0.23, "learning_rate": 9.021192993165224e-06, "loss": 1.4583, "step": 2124 }, { "epoch": 0.23, "learning_rate": 9.020166101314215e-06, "loss": 1.6348, "step": 2125 }, { "epoch": 0.23, "learning_rate": 9.019138729586846e-06, "loss": 1.548, "step": 2126 }, { "epoch": 0.23, "learning_rate": 9.01811087810575e-06, "loss": 1.4633, "step": 2127 }, { "epoch": 0.23, "learning_rate": 9.017082546993621e-06, "loss": 1.5508, "step": 2128 }, { "epoch": 0.23, "learning_rate": 9.016053736373207e-06, "loss": 1.4753, "step": 2129 }, { "epoch": 0.23, "learning_rate": 9.015024446367315e-06, "loss": 1.4642, "step": 2130 }, { "epoch": 0.23, "learning_rate": 9.013994677098808e-06, "loss": 1.2966, "step": 2131 }, { "epoch": 0.23, "learning_rate": 9.012964428690606e-06, "loss": 1.6632, "step": 2132 }, { "epoch": 0.23, "learning_rate": 9.01193370126569e-06, "loss": 1.1469, "step": 2133 }, { "epoch": 0.23, "learning_rate": 9.010902494947093e-06, "loss": 1.6484, "step": 2134 }, { "epoch": 0.23, "learning_rate": 9.009870809857907e-06, "loss": 1.5005, "step": 2135 }, { "epoch": 0.23, "learning_rate": 9.008838646121282e-06, "loss": 1.6651, "step": 2136 }, { "epoch": 0.23, "learning_rate": 9.007806003860424e-06, "loss": 1.4415, "step": 2137 }, { "epoch": 0.23, "learning_rate": 9.006772883198598e-06, "loss": 1.7197, "step": 2138 }, { "epoch": 0.23, "learning_rate": 9.005739284259123e-06, "loss": 1.712, "step": 2139 }, { "epoch": 0.23, "learning_rate": 9.00470520716538e-06, "loss": 1.7338, "step": 2140 }, { "epoch": 0.23, "learning_rate": 9.0036706520408e-06, "loss": 1.5832, "step": 2141 }, { "epoch": 0.23, "learning_rate": 9.002635619008877e-06, "loss": 1.6735, "step": 2142 }, { "epoch": 0.23, "learning_rate": 9.001600108193162e-06, "loss": 1.5115, "step": 2143 }, { "epoch": 0.23, "learning_rate": 9.000564119717256e-06, "loss": 1.8044, "step": 2144 }, { "epoch": 0.23, "learning_rate": 8.999527653704829e-06, "loss": 1.5597, "step": 2145 }, { "epoch": 0.23, "learning_rate": 8.998490710279596e-06, "loss": 1.5641, "step": 2146 }, { "epoch": 0.23, "learning_rate": 8.997453289565336e-06, "loss": 1.4884, "step": 2147 }, { "epoch": 0.23, "learning_rate": 8.996415391685882e-06, "loss": 1.6125, "step": 2148 }, { "epoch": 0.23, "learning_rate": 8.995377016765126e-06, "loss": 1.5766, "step": 2149 }, { "epoch": 0.23, "learning_rate": 8.994338164927015e-06, "loss": 1.5584, "step": 2150 }, { "epoch": 0.23, "learning_rate": 8.993298836295556e-06, "loss": 1.4716, "step": 2151 }, { "epoch": 0.23, "learning_rate": 8.99225903099481e-06, "loss": 1.3045, "step": 2152 }, { "epoch": 0.23, "learning_rate": 8.991218749148894e-06, "loss": 1.4384, "step": 2153 }, { "epoch": 0.23, "learning_rate": 8.990177990881986e-06, "loss": 1.3702, "step": 2154 }, { "epoch": 0.23, "learning_rate": 8.989136756318317e-06, "loss": 1.5195, "step": 2155 }, { "epoch": 0.23, "learning_rate": 8.988095045582178e-06, "loss": 1.5626, "step": 2156 }, { "epoch": 0.23, "learning_rate": 8.987052858797914e-06, "loss": 1.3879, "step": 2157 }, { "epoch": 0.23, "learning_rate": 8.986010196089925e-06, "loss": 1.403, "step": 2158 }, { "epoch": 0.23, "learning_rate": 8.984967057582676e-06, "loss": 1.4128, "step": 2159 }, { "epoch": 0.23, "learning_rate": 8.983923443400682e-06, "loss": 1.4145, "step": 2160 }, { "epoch": 0.23, "learning_rate": 8.982879353668516e-06, "loss": 1.4093, "step": 2161 }, { "epoch": 0.23, "learning_rate": 8.981834788510808e-06, "loss": 1.3178, "step": 2162 }, { "epoch": 0.23, "learning_rate": 8.980789748052245e-06, "loss": 1.9266, "step": 2163 }, { "epoch": 0.23, "learning_rate": 8.97974423241757e-06, "loss": 1.411, "step": 2164 }, { "epoch": 0.23, "learning_rate": 8.978698241731586e-06, "loss": 1.4909, "step": 2165 }, { "epoch": 0.23, "learning_rate": 8.977651776119145e-06, "loss": 1.365, "step": 2166 }, { "epoch": 0.23, "learning_rate": 8.976604835705165e-06, "loss": 1.3012, "step": 2167 }, { "epoch": 0.23, "learning_rate": 8.975557420614615e-06, "loss": 1.7229, "step": 2168 }, { "epoch": 0.23, "learning_rate": 8.974509530972523e-06, "loss": 1.4607, "step": 2169 }, { "epoch": 0.23, "learning_rate": 8.973461166903974e-06, "loss": 1.4575, "step": 2170 }, { "epoch": 0.23, "learning_rate": 8.972412328534104e-06, "loss": 1.5808, "step": 2171 }, { "epoch": 0.23, "learning_rate": 8.971363015988115e-06, "loss": 1.4897, "step": 2172 }, { "epoch": 0.23, "learning_rate": 8.970313229391257e-06, "loss": 1.3813, "step": 2173 }, { "epoch": 0.23, "learning_rate": 8.96926296886884e-06, "loss": 1.6161, "step": 2174 }, { "epoch": 0.23, "learning_rate": 8.968212234546235e-06, "loss": 1.4083, "step": 2175 }, { "epoch": 0.23, "learning_rate": 8.967161026548858e-06, "loss": 1.2795, "step": 2176 }, { "epoch": 0.23, "learning_rate": 8.966109345002196e-06, "loss": 1.3568, "step": 2177 }, { "epoch": 0.23, "learning_rate": 8.965057190031785e-06, "loss": 1.5128, "step": 2178 }, { "epoch": 0.23, "learning_rate": 8.964004561763213e-06, "loss": 1.3007, "step": 2179 }, { "epoch": 0.23, "learning_rate": 8.962951460322132e-06, "loss": 1.4829, "step": 2180 }, { "epoch": 0.23, "learning_rate": 8.961897885834247e-06, "loss": 1.849, "step": 2181 }, { "epoch": 0.23, "learning_rate": 8.960843838425324e-06, "loss": 1.3369, "step": 2182 }, { "epoch": 0.23, "learning_rate": 8.959789318221178e-06, "loss": 1.311, "step": 2183 }, { "epoch": 0.23, "learning_rate": 8.958734325347684e-06, "loss": 1.6416, "step": 2184 }, { "epoch": 0.23, "learning_rate": 8.957678859930774e-06, "loss": 1.4502, "step": 2185 }, { "epoch": 0.23, "learning_rate": 8.956622922096438e-06, "loss": 1.4278, "step": 2186 }, { "epoch": 0.23, "learning_rate": 8.955566511970721e-06, "loss": 1.4358, "step": 2187 }, { "epoch": 0.23, "learning_rate": 8.95450962967972e-06, "loss": 1.6587, "step": 2188 }, { "epoch": 0.23, "learning_rate": 8.953452275349596e-06, "loss": 1.5203, "step": 2189 }, { "epoch": 0.23, "learning_rate": 8.95239444910656e-06, "loss": 1.4689, "step": 2190 }, { "epoch": 0.23, "learning_rate": 8.951336151076883e-06, "loss": 1.7181, "step": 2191 }, { "epoch": 0.23, "learning_rate": 8.95027738138689e-06, "loss": 1.3725, "step": 2192 }, { "epoch": 0.23, "learning_rate": 8.949218140162965e-06, "loss": 1.3543, "step": 2193 }, { "epoch": 0.23, "learning_rate": 8.948158427531547e-06, "loss": 1.7561, "step": 2194 }, { "epoch": 0.23, "learning_rate": 8.94709824361913e-06, "loss": 1.5414, "step": 2195 }, { "epoch": 0.23, "learning_rate": 8.946037588552266e-06, "loss": 1.5492, "step": 2196 }, { "epoch": 0.23, "learning_rate": 8.944976462457559e-06, "loss": 1.6919, "step": 2197 }, { "epoch": 0.23, "learning_rate": 8.943914865461678e-06, "loss": 1.5484, "step": 2198 }, { "epoch": 0.23, "learning_rate": 8.94285279769134e-06, "loss": 1.5671, "step": 2199 }, { "epoch": 0.23, "learning_rate": 8.941790259273325e-06, "loss": 1.5016, "step": 2200 }, { "epoch": 0.23, "learning_rate": 8.940727250334458e-06, "loss": 1.4868, "step": 2201 }, { "epoch": 0.23, "learning_rate": 8.939663771001632e-06, "loss": 1.8852, "step": 2202 }, { "epoch": 0.23, "learning_rate": 8.938599821401795e-06, "loss": 1.361, "step": 2203 }, { "epoch": 0.24, "learning_rate": 8.93753540166194e-06, "loss": 1.5314, "step": 2204 }, { "epoch": 0.24, "learning_rate": 8.93647051190913e-06, "loss": 1.4536, "step": 2205 }, { "epoch": 0.24, "learning_rate": 8.935405152270477e-06, "loss": 1.2256, "step": 2206 }, { "epoch": 0.24, "learning_rate": 8.934339322873149e-06, "loss": 1.6271, "step": 2207 }, { "epoch": 0.24, "learning_rate": 8.93327302384437e-06, "loss": 1.4063, "step": 2208 }, { "epoch": 0.24, "learning_rate": 8.932206255311421e-06, "loss": 1.4189, "step": 2209 }, { "epoch": 0.24, "learning_rate": 8.931139017401645e-06, "loss": 1.5163, "step": 2210 }, { "epoch": 0.24, "learning_rate": 8.930071310242429e-06, "loss": 1.3635, "step": 2211 }, { "epoch": 0.24, "learning_rate": 8.929003133961225e-06, "loss": 1.5465, "step": 2212 }, { "epoch": 0.24, "learning_rate": 8.927934488685536e-06, "loss": 1.5509, "step": 2213 }, { "epoch": 0.24, "learning_rate": 8.926865374542928e-06, "loss": 1.6539, "step": 2214 }, { "epoch": 0.24, "learning_rate": 8.925795791661014e-06, "loss": 1.6185, "step": 2215 }, { "epoch": 0.24, "learning_rate": 8.924725740167469e-06, "loss": 1.497, "step": 2216 }, { "epoch": 0.24, "learning_rate": 8.92365522019002e-06, "loss": 1.6908, "step": 2217 }, { "epoch": 0.24, "learning_rate": 8.922584231856454e-06, "loss": 1.4488, "step": 2218 }, { "epoch": 0.24, "learning_rate": 8.921512775294614e-06, "loss": 1.7098, "step": 2219 }, { "epoch": 0.24, "learning_rate": 8.920440850632395e-06, "loss": 1.5904, "step": 2220 }, { "epoch": 0.24, "learning_rate": 8.919368457997747e-06, "loss": 1.6062, "step": 2221 }, { "epoch": 0.24, "learning_rate": 8.918295597518683e-06, "loss": 1.6976, "step": 2222 }, { "epoch": 0.24, "learning_rate": 8.917222269323263e-06, "loss": 1.5648, "step": 2223 }, { "epoch": 0.24, "learning_rate": 8.916148473539613e-06, "loss": 1.5393, "step": 2224 }, { "epoch": 0.24, "learning_rate": 8.915074210295903e-06, "loss": 1.4146, "step": 2225 }, { "epoch": 0.24, "learning_rate": 8.91399947972037e-06, "loss": 1.566, "step": 2226 }, { "epoch": 0.24, "learning_rate": 8.912924281941298e-06, "loss": 1.4031, "step": 2227 }, { "epoch": 0.24, "learning_rate": 8.91184861708703e-06, "loss": 1.6371, "step": 2228 }, { "epoch": 0.24, "learning_rate": 8.91077248528597e-06, "loss": 1.5464, "step": 2229 }, { "epoch": 0.24, "learning_rate": 8.90969588666657e-06, "loss": 1.3223, "step": 2230 }, { "epoch": 0.24, "learning_rate": 8.90861882135734e-06, "loss": 1.2939, "step": 2231 }, { "epoch": 0.24, "learning_rate": 8.907541289486847e-06, "loss": 1.2756, "step": 2232 }, { "epoch": 0.24, "learning_rate": 8.906463291183714e-06, "loss": 1.6908, "step": 2233 }, { "epoch": 0.24, "learning_rate": 8.905384826576618e-06, "loss": 1.3811, "step": 2234 }, { "epoch": 0.24, "learning_rate": 8.904305895794292e-06, "loss": 1.4675, "step": 2235 }, { "epoch": 0.24, "learning_rate": 8.903226498965524e-06, "loss": 1.7511, "step": 2236 }, { "epoch": 0.24, "learning_rate": 8.90214663621916e-06, "loss": 1.3226, "step": 2237 }, { "epoch": 0.24, "learning_rate": 8.901066307684102e-06, "loss": 1.3827, "step": 2238 }, { "epoch": 0.24, "learning_rate": 8.899985513489304e-06, "loss": 1.4465, "step": 2239 }, { "epoch": 0.24, "learning_rate": 8.898904253763778e-06, "loss": 1.6043, "step": 2240 }, { "epoch": 0.24, "learning_rate": 8.89782252863659e-06, "loss": 1.6159, "step": 2241 }, { "epoch": 0.24, "learning_rate": 8.896740338236863e-06, "loss": 1.6616, "step": 2242 }, { "epoch": 0.24, "learning_rate": 8.895657682693774e-06, "loss": 1.4802, "step": 2243 }, { "epoch": 0.24, "learning_rate": 8.894574562136561e-06, "loss": 1.7515, "step": 2244 }, { "epoch": 0.24, "learning_rate": 8.89349097669451e-06, "loss": 1.5326, "step": 2245 }, { "epoch": 0.24, "learning_rate": 8.892406926496967e-06, "loss": 1.5436, "step": 2246 }, { "epoch": 0.24, "learning_rate": 8.89132241167333e-06, "loss": 1.3342, "step": 2247 }, { "epoch": 0.24, "learning_rate": 8.890237432353055e-06, "loss": 1.159, "step": 2248 }, { "epoch": 0.24, "learning_rate": 8.889151988665654e-06, "loss": 1.2761, "step": 2249 }, { "epoch": 0.24, "learning_rate": 8.888066080740692e-06, "loss": 1.2468, "step": 2250 }, { "epoch": 0.24, "learning_rate": 8.886979708707795e-06, "loss": 1.4111, "step": 2251 }, { "epoch": 0.24, "learning_rate": 8.885892872696635e-06, "loss": 1.5644, "step": 2252 }, { "epoch": 0.24, "learning_rate": 8.88480557283695e-06, "loss": 1.6832, "step": 2253 }, { "epoch": 0.24, "learning_rate": 8.883717809258523e-06, "loss": 1.9074, "step": 2254 }, { "epoch": 0.24, "learning_rate": 8.8826295820912e-06, "loss": 1.5262, "step": 2255 }, { "epoch": 0.24, "learning_rate": 8.88154089146488e-06, "loss": 1.4415, "step": 2256 }, { "epoch": 0.24, "learning_rate": 8.880451737509514e-06, "loss": 1.4572, "step": 2257 }, { "epoch": 0.24, "learning_rate": 8.879362120355118e-06, "loss": 1.4278, "step": 2258 }, { "epoch": 0.24, "learning_rate": 8.878272040131748e-06, "loss": 1.6118, "step": 2259 }, { "epoch": 0.24, "learning_rate": 8.877181496969531e-06, "loss": 1.4406, "step": 2260 }, { "epoch": 0.24, "learning_rate": 8.876090490998639e-06, "loss": 1.7783, "step": 2261 }, { "epoch": 0.24, "learning_rate": 8.874999022349303e-06, "loss": 1.6336, "step": 2262 }, { "epoch": 0.24, "learning_rate": 8.873907091151808e-06, "loss": 1.6656, "step": 2263 }, { "epoch": 0.24, "learning_rate": 8.872814697536497e-06, "loss": 1.2605, "step": 2264 }, { "epoch": 0.24, "learning_rate": 8.871721841633762e-06, "loss": 1.6252, "step": 2265 }, { "epoch": 0.24, "learning_rate": 8.87062852357406e-06, "loss": 1.5416, "step": 2266 }, { "epoch": 0.24, "learning_rate": 8.869534743487893e-06, "loss": 1.52, "step": 2267 }, { "epoch": 0.24, "learning_rate": 8.868440501505822e-06, "loss": 1.6229, "step": 2268 }, { "epoch": 0.24, "learning_rate": 8.867345797758468e-06, "loss": 1.4073, "step": 2269 }, { "epoch": 0.24, "learning_rate": 8.866250632376499e-06, "loss": 1.5916, "step": 2270 }, { "epoch": 0.24, "learning_rate": 8.865155005490643e-06, "loss": 1.3697, "step": 2271 }, { "epoch": 0.24, "learning_rate": 8.864058917231684e-06, "loss": 1.5186, "step": 2272 }, { "epoch": 0.24, "learning_rate": 8.862962367730456e-06, "loss": 1.7717, "step": 2273 }, { "epoch": 0.24, "learning_rate": 8.861865357117852e-06, "loss": 1.6378, "step": 2274 }, { "epoch": 0.24, "learning_rate": 8.860767885524821e-06, "loss": 0.9929, "step": 2275 }, { "epoch": 0.24, "learning_rate": 8.859669953082364e-06, "loss": 1.4558, "step": 2276 }, { "epoch": 0.24, "learning_rate": 8.858571559921539e-06, "loss": 1.4687, "step": 2277 }, { "epoch": 0.24, "learning_rate": 8.857472706173455e-06, "loss": 1.3614, "step": 2278 }, { "epoch": 0.24, "learning_rate": 8.856373391969282e-06, "loss": 1.7588, "step": 2279 }, { "epoch": 0.24, "learning_rate": 8.855273617440243e-06, "loss": 1.6558, "step": 2280 }, { "epoch": 0.24, "learning_rate": 8.854173382717612e-06, "loss": 1.5388, "step": 2281 }, { "epoch": 0.24, "learning_rate": 8.853072687932724e-06, "loss": 1.3593, "step": 2282 }, { "epoch": 0.24, "learning_rate": 8.851971533216968e-06, "loss": 1.3965, "step": 2283 }, { "epoch": 0.24, "learning_rate": 8.85086991870178e-06, "loss": 1.421, "step": 2284 }, { "epoch": 0.24, "learning_rate": 8.849767844518658e-06, "loss": 1.4128, "step": 2285 }, { "epoch": 0.24, "learning_rate": 8.848665310799156e-06, "loss": 1.5226, "step": 2286 }, { "epoch": 0.24, "learning_rate": 8.84756231767488e-06, "loss": 1.475, "step": 2287 }, { "epoch": 0.24, "learning_rate": 8.846458865277492e-06, "loss": 1.3913, "step": 2288 }, { "epoch": 0.24, "learning_rate": 8.845354953738706e-06, "loss": 1.3403, "step": 2289 }, { "epoch": 0.24, "learning_rate": 8.844250583190295e-06, "loss": 1.5028, "step": 2290 }, { "epoch": 0.24, "learning_rate": 8.843145753764083e-06, "loss": 1.6158, "step": 2291 }, { "epoch": 0.24, "learning_rate": 8.84204046559195e-06, "loss": 1.8228, "step": 2292 }, { "epoch": 0.24, "learning_rate": 8.840934718805832e-06, "loss": 1.4949, "step": 2293 }, { "epoch": 0.24, "learning_rate": 8.839828513537722e-06, "loss": 1.8491, "step": 2294 }, { "epoch": 0.24, "learning_rate": 8.83872184991966e-06, "loss": 1.7809, "step": 2295 }, { "epoch": 0.24, "learning_rate": 8.83761472808375e-06, "loss": 1.3952, "step": 2296 }, { "epoch": 0.25, "learning_rate": 8.836507148162141e-06, "loss": 1.3511, "step": 2297 }, { "epoch": 0.25, "learning_rate": 8.835399110287046e-06, "loss": 1.5517, "step": 2298 }, { "epoch": 0.25, "learning_rate": 8.834290614590729e-06, "loss": 1.5282, "step": 2299 }, { "epoch": 0.25, "learning_rate": 8.833181661205503e-06, "loss": 1.5799, "step": 2300 }, { "epoch": 0.25, "learning_rate": 8.832072250263746e-06, "loss": 1.5701, "step": 2301 }, { "epoch": 0.25, "learning_rate": 8.830962381897882e-06, "loss": 1.8039, "step": 2302 }, { "epoch": 0.25, "learning_rate": 8.829852056240396e-06, "loss": 1.3643, "step": 2303 }, { "epoch": 0.25, "learning_rate": 8.82874127342382e-06, "loss": 1.1701, "step": 2304 }, { "epoch": 0.25, "learning_rate": 8.827630033580752e-06, "loss": 1.6346, "step": 2305 }, { "epoch": 0.25, "learning_rate": 8.826518336843831e-06, "loss": 1.4187, "step": 2306 }, { "epoch": 0.25, "learning_rate": 8.82540618334576e-06, "loss": 1.6351, "step": 2307 }, { "epoch": 0.25, "learning_rate": 8.824293573219295e-06, "loss": 1.5318, "step": 2308 }, { "epoch": 0.25, "learning_rate": 8.823180506597244e-06, "loss": 1.5986, "step": 2309 }, { "epoch": 0.25, "learning_rate": 8.82206698361247e-06, "loss": 1.4676, "step": 2310 }, { "epoch": 0.25, "learning_rate": 8.820953004397892e-06, "loss": 1.5848, "step": 2311 }, { "epoch": 0.25, "learning_rate": 8.819838569086482e-06, "loss": 1.5449, "step": 2312 }, { "epoch": 0.25, "learning_rate": 8.818723677811269e-06, "loss": 1.654, "step": 2313 }, { "epoch": 0.25, "learning_rate": 8.817608330705334e-06, "loss": 1.5741, "step": 2314 }, { "epoch": 0.25, "learning_rate": 8.816492527901811e-06, "loss": 1.1968, "step": 2315 }, { "epoch": 0.25, "learning_rate": 8.815376269533893e-06, "loss": 1.5648, "step": 2316 }, { "epoch": 0.25, "learning_rate": 8.814259555734823e-06, "loss": 1.3112, "step": 2317 }, { "epoch": 0.25, "learning_rate": 8.8131423866379e-06, "loss": 1.412, "step": 2318 }, { "epoch": 0.25, "learning_rate": 8.812024762376477e-06, "loss": 1.4322, "step": 2319 }, { "epoch": 0.25, "learning_rate": 8.810906683083964e-06, "loss": 1.2356, "step": 2320 }, { "epoch": 0.25, "learning_rate": 8.809788148893824e-06, "loss": 1.6139, "step": 2321 }, { "epoch": 0.25, "learning_rate": 8.808669159939568e-06, "loss": 1.3794, "step": 2322 }, { "epoch": 0.25, "learning_rate": 8.807549716354772e-06, "loss": 1.7093, "step": 2323 }, { "epoch": 0.25, "learning_rate": 8.806429818273057e-06, "loss": 1.5361, "step": 2324 }, { "epoch": 0.25, "learning_rate": 8.805309465828105e-06, "loss": 1.3936, "step": 2325 }, { "epoch": 0.25, "learning_rate": 8.80418865915365e-06, "loss": 1.4926, "step": 2326 }, { "epoch": 0.25, "learning_rate": 8.803067398383477e-06, "loss": 1.4292, "step": 2327 }, { "epoch": 0.25, "learning_rate": 8.80194568365143e-06, "loss": 1.4349, "step": 2328 }, { "epoch": 0.25, "learning_rate": 8.800823515091403e-06, "loss": 1.3269, "step": 2329 }, { "epoch": 0.25, "learning_rate": 8.799700892837348e-06, "loss": 1.6153, "step": 2330 }, { "epoch": 0.25, "learning_rate": 8.798577817023269e-06, "loss": 1.4781, "step": 2331 }, { "epoch": 0.25, "learning_rate": 8.797454287783224e-06, "loss": 1.5682, "step": 2332 }, { "epoch": 0.25, "learning_rate": 8.796330305251326e-06, "loss": 1.5297, "step": 2333 }, { "epoch": 0.25, "learning_rate": 8.795205869561742e-06, "loss": 1.3888, "step": 2334 }, { "epoch": 0.25, "learning_rate": 8.79408098084869e-06, "loss": 1.6153, "step": 2335 }, { "epoch": 0.25, "learning_rate": 8.792955639246452e-06, "loss": 1.369, "step": 2336 }, { "epoch": 0.25, "learning_rate": 8.79182984488935e-06, "loss": 1.7548, "step": 2337 }, { "epoch": 0.25, "learning_rate": 8.790703597911769e-06, "loss": 1.465, "step": 2338 }, { "epoch": 0.25, "learning_rate": 8.78957689844815e-06, "loss": 1.443, "step": 2339 }, { "epoch": 0.25, "learning_rate": 8.788449746632976e-06, "loss": 1.3239, "step": 2340 }, { "epoch": 0.25, "learning_rate": 8.787322142600799e-06, "loss": 1.6441, "step": 2341 }, { "epoch": 0.25, "learning_rate": 8.786194086486215e-06, "loss": 1.442, "step": 2342 }, { "epoch": 0.25, "learning_rate": 8.78506557842388e-06, "loss": 1.5379, "step": 2343 }, { "epoch": 0.25, "learning_rate": 8.783936618548496e-06, "loss": 1.5418, "step": 2344 }, { "epoch": 0.25, "learning_rate": 8.782807206994826e-06, "loss": 1.3357, "step": 2345 }, { "epoch": 0.25, "learning_rate": 8.781677343897687e-06, "loss": 1.5714, "step": 2346 }, { "epoch": 0.25, "learning_rate": 8.780547029391947e-06, "loss": 1.7369, "step": 2347 }, { "epoch": 0.25, "learning_rate": 8.779416263612526e-06, "loss": 1.4623, "step": 2348 }, { "epoch": 0.25, "learning_rate": 8.778285046694403e-06, "loss": 1.3865, "step": 2349 }, { "epoch": 0.25, "learning_rate": 8.777153378772608e-06, "loss": 1.5342, "step": 2350 }, { "epoch": 0.25, "learning_rate": 8.776021259982224e-06, "loss": 1.4062, "step": 2351 }, { "epoch": 0.25, "learning_rate": 8.77488869045839e-06, "loss": 1.6923, "step": 2352 }, { "epoch": 0.25, "learning_rate": 8.773755670336297e-06, "loss": 1.5625, "step": 2353 }, { "epoch": 0.25, "learning_rate": 8.77262219975119e-06, "loss": 1.5773, "step": 2354 }, { "epoch": 0.25, "learning_rate": 8.771488278838368e-06, "loss": 1.7063, "step": 2355 }, { "epoch": 0.25, "learning_rate": 8.77035390773319e-06, "loss": 1.5317, "step": 2356 }, { "epoch": 0.25, "learning_rate": 8.769219086571054e-06, "loss": 1.6113, "step": 2357 }, { "epoch": 0.25, "learning_rate": 8.768083815487428e-06, "loss": 1.4828, "step": 2358 }, { "epoch": 0.25, "learning_rate": 8.766948094617819e-06, "loss": 1.401, "step": 2359 }, { "epoch": 0.25, "learning_rate": 8.7658119240978e-06, "loss": 1.5424, "step": 2360 }, { "epoch": 0.25, "learning_rate": 8.764675304062992e-06, "loss": 1.4564, "step": 2361 }, { "epoch": 0.25, "learning_rate": 8.763538234649069e-06, "loss": 1.561, "step": 2362 }, { "epoch": 0.25, "learning_rate": 8.76240071599176e-06, "loss": 1.4917, "step": 2363 }, { "epoch": 0.25, "learning_rate": 8.76126274822685e-06, "loss": 1.4423, "step": 2364 }, { "epoch": 0.25, "learning_rate": 8.760124331490174e-06, "loss": 1.682, "step": 2365 }, { "epoch": 0.25, "learning_rate": 8.75898546591762e-06, "loss": 1.5433, "step": 2366 }, { "epoch": 0.25, "learning_rate": 8.75784615164513e-06, "loss": 1.5228, "step": 2367 }, { "epoch": 0.25, "learning_rate": 8.756706388808704e-06, "loss": 1.4258, "step": 2368 }, { "epoch": 0.25, "learning_rate": 8.755566177544392e-06, "loss": 1.5923, "step": 2369 }, { "epoch": 0.25, "learning_rate": 8.754425517988298e-06, "loss": 1.6666, "step": 2370 }, { "epoch": 0.25, "learning_rate": 8.753284410276579e-06, "loss": 1.4384, "step": 2371 }, { "epoch": 0.25, "learning_rate": 8.752142854545447e-06, "loss": 1.5245, "step": 2372 }, { "epoch": 0.25, "learning_rate": 8.751000850931162e-06, "loss": 1.4713, "step": 2373 }, { "epoch": 0.25, "learning_rate": 8.749858399570047e-06, "loss": 1.5571, "step": 2374 }, { "epoch": 0.25, "learning_rate": 8.748715500598472e-06, "loss": 1.5684, "step": 2375 }, { "epoch": 0.25, "learning_rate": 8.74757215415286e-06, "loss": 1.3053, "step": 2376 }, { "epoch": 0.25, "learning_rate": 8.746428360369691e-06, "loss": 1.2637, "step": 2377 }, { "epoch": 0.25, "learning_rate": 8.745284119385496e-06, "loss": 1.6832, "step": 2378 }, { "epoch": 0.25, "learning_rate": 8.74413943133686e-06, "loss": 1.4814, "step": 2379 }, { "epoch": 0.25, "learning_rate": 8.74299429636042e-06, "loss": 1.4407, "step": 2380 }, { "epoch": 0.25, "learning_rate": 8.74184871459287e-06, "loss": 1.4894, "step": 2381 }, { "epoch": 0.25, "learning_rate": 8.740702686170955e-06, "loss": 1.6075, "step": 2382 }, { "epoch": 0.25, "learning_rate": 8.739556211231472e-06, "loss": 1.3982, "step": 2383 }, { "epoch": 0.25, "learning_rate": 8.738409289911272e-06, "loss": 1.4051, "step": 2384 }, { "epoch": 0.25, "learning_rate": 8.73726192234726e-06, "loss": 1.271, "step": 2385 }, { "epoch": 0.25, "learning_rate": 8.736114108676396e-06, "loss": 1.2926, "step": 2386 }, { "epoch": 0.25, "learning_rate": 8.73496584903569e-06, "loss": 1.6079, "step": 2387 }, { "epoch": 0.25, "learning_rate": 8.733817143562207e-06, "loss": 1.4606, "step": 2388 }, { "epoch": 0.25, "learning_rate": 8.732667992393064e-06, "loss": 1.368, "step": 2389 }, { "epoch": 0.25, "learning_rate": 8.731518395665434e-06, "loss": 1.5218, "step": 2390 }, { "epoch": 0.26, "learning_rate": 8.73036835351654e-06, "loss": 1.6267, "step": 2391 }, { "epoch": 0.26, "learning_rate": 8.729217866083661e-06, "loss": 1.3303, "step": 2392 }, { "epoch": 0.26, "learning_rate": 8.728066933504124e-06, "loss": 1.7202, "step": 2393 }, { "epoch": 0.26, "learning_rate": 8.726915555915317e-06, "loss": 1.5142, "step": 2394 }, { "epoch": 0.26, "learning_rate": 8.725763733454673e-06, "loss": 1.474, "step": 2395 }, { "epoch": 0.26, "learning_rate": 8.724611466259682e-06, "loss": 1.8309, "step": 2396 }, { "epoch": 0.26, "learning_rate": 8.723458754467893e-06, "loss": 1.481, "step": 2397 }, { "epoch": 0.26, "learning_rate": 8.722305598216895e-06, "loss": 1.316, "step": 2398 }, { "epoch": 0.26, "learning_rate": 8.72115199764434e-06, "loss": 1.6726, "step": 2399 }, { "epoch": 0.26, "learning_rate": 8.719997952887932e-06, "loss": 1.2775, "step": 2400 }, { "epoch": 0.26, "learning_rate": 8.71884346408542e-06, "loss": 1.4965, "step": 2401 }, { "epoch": 0.26, "learning_rate": 8.71768853137462e-06, "loss": 1.2482, "step": 2402 }, { "epoch": 0.26, "learning_rate": 8.71653315489339e-06, "loss": 1.6995, "step": 2403 }, { "epoch": 0.26, "learning_rate": 8.715377334779642e-06, "loss": 1.6384, "step": 2404 }, { "epoch": 0.26, "learning_rate": 8.714221071171345e-06, "loss": 1.4889, "step": 2405 }, { "epoch": 0.26, "learning_rate": 8.71306436420652e-06, "loss": 1.4758, "step": 2406 }, { "epoch": 0.26, "learning_rate": 8.711907214023238e-06, "loss": 1.4017, "step": 2407 }, { "epoch": 0.26, "learning_rate": 8.710749620759627e-06, "loss": 1.4431, "step": 2408 }, { "epoch": 0.26, "learning_rate": 8.709591584553865e-06, "loss": 1.5136, "step": 2409 }, { "epoch": 0.26, "learning_rate": 8.708433105544183e-06, "loss": 1.5064, "step": 2410 }, { "epoch": 0.26, "learning_rate": 8.707274183868865e-06, "loss": 1.3437, "step": 2411 }, { "epoch": 0.26, "learning_rate": 8.706114819666249e-06, "loss": 1.3984, "step": 2412 }, { "epoch": 0.26, "learning_rate": 8.704955013074727e-06, "loss": 1.5829, "step": 2413 }, { "epoch": 0.26, "learning_rate": 8.703794764232739e-06, "loss": 1.3473, "step": 2414 }, { "epoch": 0.26, "learning_rate": 8.702634073278784e-06, "loss": 1.2607, "step": 2415 }, { "epoch": 0.26, "learning_rate": 8.701472940351407e-06, "loss": 1.3919, "step": 2416 }, { "epoch": 0.26, "learning_rate": 8.70031136558921e-06, "loss": 1.3919, "step": 2417 }, { "epoch": 0.26, "learning_rate": 8.699149349130848e-06, "loss": 1.5508, "step": 2418 }, { "epoch": 0.26, "learning_rate": 8.697986891115031e-06, "loss": 1.6727, "step": 2419 }, { "epoch": 0.26, "learning_rate": 8.696823991680514e-06, "loss": 1.474, "step": 2420 }, { "epoch": 0.26, "learning_rate": 8.695660650966109e-06, "loss": 1.4262, "step": 2421 }, { "epoch": 0.26, "learning_rate": 8.694496869110682e-06, "loss": 1.5676, "step": 2422 }, { "epoch": 0.26, "learning_rate": 8.693332646253151e-06, "loss": 1.6295, "step": 2423 }, { "epoch": 0.26, "learning_rate": 8.692167982532487e-06, "loss": 1.429, "step": 2424 }, { "epoch": 0.26, "learning_rate": 8.69100287808771e-06, "loss": 1.3909, "step": 2425 }, { "epoch": 0.26, "learning_rate": 8.689837333057899e-06, "loss": 1.4117, "step": 2426 }, { "epoch": 0.26, "learning_rate": 8.688671347582178e-06, "loss": 1.2896, "step": 2427 }, { "epoch": 0.26, "learning_rate": 8.687504921799731e-06, "loss": 1.4367, "step": 2428 }, { "epoch": 0.26, "learning_rate": 8.686338055849789e-06, "loss": 1.4476, "step": 2429 }, { "epoch": 0.26, "learning_rate": 8.685170749871638e-06, "loss": 1.3921, "step": 2430 }, { "epoch": 0.26, "learning_rate": 8.684003004004618e-06, "loss": 1.5011, "step": 2431 }, { "epoch": 0.26, "learning_rate": 8.682834818388115e-06, "loss": 1.453, "step": 2432 }, { "epoch": 0.26, "learning_rate": 8.681666193161578e-06, "loss": 1.529, "step": 2433 }, { "epoch": 0.26, "learning_rate": 8.680497128464498e-06, "loss": 1.3358, "step": 2434 }, { "epoch": 0.26, "learning_rate": 8.679327624436425e-06, "loss": 1.4374, "step": 2435 }, { "epoch": 0.26, "learning_rate": 8.67815768121696e-06, "loss": 1.467, "step": 2436 }, { "epoch": 0.26, "learning_rate": 8.676987298945757e-06, "loss": 1.2456, "step": 2437 }, { "epoch": 0.26, "learning_rate": 8.675816477762516e-06, "loss": 1.402, "step": 2438 }, { "epoch": 0.26, "learning_rate": 8.674645217807e-06, "loss": 1.5172, "step": 2439 }, { "epoch": 0.26, "learning_rate": 8.673473519219018e-06, "loss": 1.4322, "step": 2440 }, { "epoch": 0.26, "learning_rate": 8.672301382138432e-06, "loss": 1.3924, "step": 2441 }, { "epoch": 0.26, "learning_rate": 8.671128806705159e-06, "loss": 1.557, "step": 2442 }, { "epoch": 0.26, "learning_rate": 8.669955793059163e-06, "loss": 1.3732, "step": 2443 }, { "epoch": 0.26, "learning_rate": 8.668782341340464e-06, "loss": 1.6446, "step": 2444 }, { "epoch": 0.26, "learning_rate": 8.667608451689135e-06, "loss": 1.4932, "step": 2445 }, { "epoch": 0.26, "learning_rate": 8.666434124245298e-06, "loss": 1.6175, "step": 2446 }, { "epoch": 0.26, "learning_rate": 8.665259359149132e-06, "loss": 1.607, "step": 2447 }, { "epoch": 0.26, "learning_rate": 8.664084156540864e-06, "loss": 1.4933, "step": 2448 }, { "epoch": 0.26, "learning_rate": 8.662908516560774e-06, "loss": 1.2888, "step": 2449 }, { "epoch": 0.26, "learning_rate": 8.661732439349198e-06, "loss": 1.4966, "step": 2450 }, { "epoch": 0.26, "learning_rate": 8.660555925046518e-06, "loss": 1.2831, "step": 2451 }, { "epoch": 0.26, "learning_rate": 8.659378973793173e-06, "loss": 1.6917, "step": 2452 }, { "epoch": 0.26, "learning_rate": 8.658201585729652e-06, "loss": 1.3836, "step": 2453 }, { "epoch": 0.26, "learning_rate": 8.657023760996497e-06, "loss": 1.57, "step": 2454 }, { "epoch": 0.26, "learning_rate": 8.655845499734298e-06, "loss": 1.4441, "step": 2455 }, { "epoch": 0.26, "learning_rate": 8.654666802083708e-06, "loss": 1.7064, "step": 2456 }, { "epoch": 0.26, "learning_rate": 8.653487668185419e-06, "loss": 1.5249, "step": 2457 }, { "epoch": 0.26, "learning_rate": 8.652308098180186e-06, "loss": 1.4221, "step": 2458 }, { "epoch": 0.26, "learning_rate": 8.651128092208805e-06, "loss": 1.6241, "step": 2459 }, { "epoch": 0.26, "learning_rate": 8.649947650412135e-06, "loss": 1.4676, "step": 2460 }, { "epoch": 0.26, "learning_rate": 8.648766772931081e-06, "loss": 1.4005, "step": 2461 }, { "epoch": 0.26, "learning_rate": 8.647585459906599e-06, "loss": 1.5997, "step": 2462 }, { "epoch": 0.26, "learning_rate": 8.646403711479702e-06, "loss": 1.354, "step": 2463 }, { "epoch": 0.26, "learning_rate": 8.64522152779145e-06, "loss": 1.5304, "step": 2464 }, { "epoch": 0.26, "learning_rate": 8.64403890898296e-06, "loss": 1.4376, "step": 2465 }, { "epoch": 0.26, "learning_rate": 8.642855855195394e-06, "loss": 1.3684, "step": 2466 }, { "epoch": 0.26, "learning_rate": 8.641672366569972e-06, "loss": 1.3497, "step": 2467 }, { "epoch": 0.26, "learning_rate": 8.640488443247966e-06, "loss": 1.585, "step": 2468 }, { "epoch": 0.26, "learning_rate": 8.639304085370692e-06, "loss": 1.5421, "step": 2469 }, { "epoch": 0.26, "learning_rate": 8.63811929307953e-06, "loss": 1.577, "step": 2470 }, { "epoch": 0.26, "learning_rate": 8.636934066515901e-06, "loss": 1.3271, "step": 2471 }, { "epoch": 0.26, "learning_rate": 8.635748405821285e-06, "loss": 1.7547, "step": 2472 }, { "epoch": 0.26, "learning_rate": 8.634562311137209e-06, "loss": 1.5438, "step": 2473 }, { "epoch": 0.26, "learning_rate": 8.633375782605256e-06, "loss": 1.4792, "step": 2474 }, { "epoch": 0.26, "learning_rate": 8.632188820367056e-06, "loss": 1.4901, "step": 2475 }, { "epoch": 0.26, "learning_rate": 8.631001424564298e-06, "loss": 1.3392, "step": 2476 }, { "epoch": 0.26, "learning_rate": 8.629813595338712e-06, "loss": 1.4605, "step": 2477 }, { "epoch": 0.26, "learning_rate": 8.62862533283209e-06, "loss": 1.4829, "step": 2478 }, { "epoch": 0.26, "learning_rate": 8.627436637186272e-06, "loss": 1.4175, "step": 2479 }, { "epoch": 0.26, "learning_rate": 8.62624750854315e-06, "loss": 1.5689, "step": 2480 }, { "epoch": 0.26, "learning_rate": 8.625057947044662e-06, "loss": 1.3582, "step": 2481 }, { "epoch": 0.26, "learning_rate": 8.623867952832806e-06, "loss": 1.4557, "step": 2482 }, { "epoch": 0.26, "learning_rate": 8.622677526049629e-06, "loss": 1.3121, "step": 2483 }, { "epoch": 0.26, "learning_rate": 8.62148666683723e-06, "loss": 1.4326, "step": 2484 }, { "epoch": 0.27, "learning_rate": 8.620295375337757e-06, "loss": 1.3781, "step": 2485 }, { "epoch": 0.27, "learning_rate": 8.619103651693413e-06, "loss": 1.368, "step": 2486 }, { "epoch": 0.27, "learning_rate": 8.617911496046446e-06, "loss": 1.4884, "step": 2487 }, { "epoch": 0.27, "learning_rate": 8.616718908539165e-06, "loss": 1.5519, "step": 2488 }, { "epoch": 0.27, "learning_rate": 8.615525889313924e-06, "loss": 1.7034, "step": 2489 }, { "epoch": 0.27, "learning_rate": 8.614332438513132e-06, "loss": 1.6404, "step": 2490 }, { "epoch": 0.27, "learning_rate": 8.613138556279248e-06, "loss": 1.5346, "step": 2491 }, { "epoch": 0.27, "learning_rate": 8.611944242754779e-06, "loss": 1.5784, "step": 2492 }, { "epoch": 0.27, "learning_rate": 8.610749498082291e-06, "loss": 1.6913, "step": 2493 }, { "epoch": 0.27, "learning_rate": 8.609554322404396e-06, "loss": 1.331, "step": 2494 }, { "epoch": 0.27, "learning_rate": 8.60835871586376e-06, "loss": 1.4537, "step": 2495 }, { "epoch": 0.27, "learning_rate": 8.607162678603097e-06, "loss": 1.5354, "step": 2496 }, { "epoch": 0.27, "learning_rate": 8.605966210765176e-06, "loss": 1.6449, "step": 2497 }, { "epoch": 0.27, "learning_rate": 8.604769312492817e-06, "loss": 1.4446, "step": 2498 }, { "epoch": 0.27, "learning_rate": 8.603571983928888e-06, "loss": 1.7899, "step": 2499 }, { "epoch": 0.27, "learning_rate": 8.602374225216313e-06, "loss": 1.5815, "step": 2500 }, { "epoch": 0.27, "learning_rate": 8.601176036498066e-06, "loss": 1.4943, "step": 2501 }, { "epoch": 0.27, "learning_rate": 8.599977417917169e-06, "loss": 1.6921, "step": 2502 }, { "epoch": 0.27, "learning_rate": 8.598778369616699e-06, "loss": 1.5878, "step": 2503 }, { "epoch": 0.27, "learning_rate": 8.597578891739784e-06, "loss": 1.6652, "step": 2504 }, { "epoch": 0.27, "learning_rate": 8.5963789844296e-06, "loss": 1.2782, "step": 2505 }, { "epoch": 0.27, "learning_rate": 8.59517864782938e-06, "loss": 1.4765, "step": 2506 }, { "epoch": 0.27, "learning_rate": 8.593977882082403e-06, "loss": 1.4597, "step": 2507 }, { "epoch": 0.27, "learning_rate": 8.592776687332003e-06, "loss": 1.674, "step": 2508 }, { "epoch": 0.27, "learning_rate": 8.59157506372156e-06, "loss": 1.5067, "step": 2509 }, { "epoch": 0.27, "learning_rate": 8.59037301139451e-06, "loss": 1.5013, "step": 2510 }, { "epoch": 0.27, "learning_rate": 8.58917053049434e-06, "loss": 1.191, "step": 2511 }, { "epoch": 0.27, "learning_rate": 8.587967621164586e-06, "loss": 1.4426, "step": 2512 }, { "epoch": 0.27, "learning_rate": 8.586764283548837e-06, "loss": 1.7788, "step": 2513 }, { "epoch": 0.27, "learning_rate": 8.58556051779073e-06, "loss": 1.4664, "step": 2514 }, { "epoch": 0.27, "learning_rate": 8.584356324033955e-06, "loss": 1.6566, "step": 2515 }, { "epoch": 0.27, "learning_rate": 8.583151702422258e-06, "loss": 1.5132, "step": 2516 }, { "epoch": 0.27, "learning_rate": 8.581946653099427e-06, "loss": 1.481, "step": 2517 }, { "epoch": 0.27, "learning_rate": 8.580741176209307e-06, "loss": 1.5114, "step": 2518 }, { "epoch": 0.27, "learning_rate": 8.579535271895791e-06, "loss": 1.5069, "step": 2519 }, { "epoch": 0.27, "learning_rate": 8.578328940302827e-06, "loss": 1.6123, "step": 2520 }, { "epoch": 0.27, "learning_rate": 8.57712218157441e-06, "loss": 1.5811, "step": 2521 }, { "epoch": 0.27, "learning_rate": 8.575914995854588e-06, "loss": 1.5964, "step": 2522 }, { "epoch": 0.27, "learning_rate": 8.574707383287459e-06, "loss": 1.3029, "step": 2523 }, { "epoch": 0.27, "learning_rate": 8.573499344017174e-06, "loss": 1.5234, "step": 2524 }, { "epoch": 0.27, "learning_rate": 8.57229087818793e-06, "loss": 1.6879, "step": 2525 }, { "epoch": 0.27, "learning_rate": 8.571081985943984e-06, "loss": 1.5075, "step": 2526 }, { "epoch": 0.27, "learning_rate": 8.56987266742963e-06, "loss": 1.4613, "step": 2527 }, { "epoch": 0.27, "learning_rate": 8.568662922789229e-06, "loss": 1.425, "step": 2528 }, { "epoch": 0.27, "learning_rate": 8.567452752167183e-06, "loss": 1.6853, "step": 2529 }, { "epoch": 0.27, "learning_rate": 8.566242155707942e-06, "loss": 1.9553, "step": 2530 }, { "epoch": 0.27, "learning_rate": 8.565031133556016e-06, "loss": 1.2958, "step": 2531 }, { "epoch": 0.27, "learning_rate": 8.563819685855963e-06, "loss": 1.7408, "step": 2532 }, { "epoch": 0.27, "learning_rate": 8.562607812752386e-06, "loss": 1.594, "step": 2533 }, { "epoch": 0.27, "learning_rate": 8.561395514389945e-06, "loss": 1.3345, "step": 2534 }, { "epoch": 0.27, "learning_rate": 8.560182790913349e-06, "loss": 1.5158, "step": 2535 }, { "epoch": 0.27, "learning_rate": 8.558969642467356e-06, "loss": 1.5912, "step": 2536 }, { "epoch": 0.27, "learning_rate": 8.557756069196779e-06, "loss": 1.5635, "step": 2537 }, { "epoch": 0.27, "learning_rate": 8.556542071246476e-06, "loss": 1.4962, "step": 2538 }, { "epoch": 0.27, "learning_rate": 8.555327648761362e-06, "loss": 1.5343, "step": 2539 }, { "epoch": 0.27, "learning_rate": 8.554112801886396e-06, "loss": 1.2939, "step": 2540 }, { "epoch": 0.27, "learning_rate": 8.552897530766592e-06, "loss": 1.5988, "step": 2541 }, { "epoch": 0.27, "learning_rate": 8.551681835547014e-06, "loss": 1.4554, "step": 2542 }, { "epoch": 0.27, "learning_rate": 8.550465716372777e-06, "loss": 1.4022, "step": 2543 }, { "epoch": 0.27, "learning_rate": 8.549249173389045e-06, "loss": 1.4586, "step": 2544 }, { "epoch": 0.27, "learning_rate": 8.548032206741033e-06, "loss": 1.4543, "step": 2545 }, { "epoch": 0.27, "learning_rate": 8.546814816574008e-06, "loss": 1.4837, "step": 2546 }, { "epoch": 0.27, "learning_rate": 8.545597003033286e-06, "loss": 1.6405, "step": 2547 }, { "epoch": 0.27, "learning_rate": 8.544378766264231e-06, "loss": 1.3842, "step": 2548 }, { "epoch": 0.27, "learning_rate": 8.54316010641227e-06, "loss": 1.4569, "step": 2549 }, { "epoch": 0.27, "learning_rate": 8.54194102362286e-06, "loss": 1.5816, "step": 2550 }, { "epoch": 0.27, "learning_rate": 8.540721518041527e-06, "loss": 1.5792, "step": 2551 }, { "epoch": 0.27, "learning_rate": 8.539501589813837e-06, "loss": 1.3823, "step": 2552 }, { "epoch": 0.27, "learning_rate": 8.538281239085411e-06, "loss": 1.5534, "step": 2553 }, { "epoch": 0.27, "learning_rate": 8.537060466001917e-06, "loss": 1.6564, "step": 2554 }, { "epoch": 0.27, "learning_rate": 8.535839270709076e-06, "loss": 1.6398, "step": 2555 }, { "epoch": 0.27, "learning_rate": 8.534617653352661e-06, "loss": 1.4152, "step": 2556 }, { "epoch": 0.27, "learning_rate": 8.533395614078492e-06, "loss": 1.3974, "step": 2557 }, { "epoch": 0.27, "learning_rate": 8.532173153032439e-06, "loss": 1.5092, "step": 2558 }, { "epoch": 0.27, "learning_rate": 8.530950270360425e-06, "loss": 1.4684, "step": 2559 }, { "epoch": 0.27, "learning_rate": 8.529726966208423e-06, "loss": 1.8696, "step": 2560 }, { "epoch": 0.27, "learning_rate": 8.528503240722452e-06, "loss": 1.5267, "step": 2561 }, { "epoch": 0.27, "learning_rate": 8.52727909404859e-06, "loss": 1.779, "step": 2562 }, { "epoch": 0.27, "learning_rate": 8.526054526332957e-06, "loss": 1.6054, "step": 2563 }, { "epoch": 0.27, "learning_rate": 8.524829537721725e-06, "loss": 1.4881, "step": 2564 }, { "epoch": 0.27, "learning_rate": 8.523604128361123e-06, "loss": 1.5801, "step": 2565 }, { "epoch": 0.27, "learning_rate": 8.522378298397418e-06, "loss": 1.5398, "step": 2566 }, { "epoch": 0.27, "learning_rate": 8.52115204797694e-06, "loss": 1.452, "step": 2567 }, { "epoch": 0.27, "learning_rate": 8.519925377246057e-06, "loss": 1.4058, "step": 2568 }, { "epoch": 0.27, "learning_rate": 8.518698286351199e-06, "loss": 1.2503, "step": 2569 }, { "epoch": 0.27, "learning_rate": 8.517470775438838e-06, "loss": 1.6976, "step": 2570 }, { "epoch": 0.27, "learning_rate": 8.516242844655498e-06, "loss": 1.4507, "step": 2571 }, { "epoch": 0.27, "learning_rate": 8.515014494147758e-06, "loss": 1.7099, "step": 2572 }, { "epoch": 0.27, "learning_rate": 8.513785724062236e-06, "loss": 1.545, "step": 2573 }, { "epoch": 0.27, "learning_rate": 8.512556534545612e-06, "loss": 1.4689, "step": 2574 }, { "epoch": 0.27, "learning_rate": 8.51132692574461e-06, "loss": 1.5074, "step": 2575 }, { "epoch": 0.27, "learning_rate": 8.510096897806004e-06, "loss": 1.3605, "step": 2576 }, { "epoch": 0.27, "learning_rate": 8.50886645087662e-06, "loss": 1.7367, "step": 2577 }, { "epoch": 0.27, "learning_rate": 8.507635585103333e-06, "loss": 1.6025, "step": 2578 }, { "epoch": 0.28, "learning_rate": 8.506404300633069e-06, "loss": 1.6691, "step": 2579 }, { "epoch": 0.28, "learning_rate": 8.5051725976128e-06, "loss": 1.6623, "step": 2580 }, { "epoch": 0.28, "learning_rate": 8.503940476189556e-06, "loss": 1.4692, "step": 2581 }, { "epoch": 0.28, "learning_rate": 8.502707936510406e-06, "loss": 1.407, "step": 2582 }, { "epoch": 0.28, "learning_rate": 8.50147497872248e-06, "loss": 1.3465, "step": 2583 }, { "epoch": 0.28, "learning_rate": 8.500241602972952e-06, "loss": 1.427, "step": 2584 }, { "epoch": 0.28, "learning_rate": 8.499007809409043e-06, "loss": 1.6551, "step": 2585 }, { "epoch": 0.28, "learning_rate": 8.497773598178033e-06, "loss": 1.8058, "step": 2586 }, { "epoch": 0.28, "learning_rate": 8.496538969427243e-06, "loss": 1.4947, "step": 2587 }, { "epoch": 0.28, "learning_rate": 8.495303923304047e-06, "loss": 1.6203, "step": 2588 }, { "epoch": 0.28, "learning_rate": 8.494068459955871e-06, "loss": 1.3416, "step": 2589 }, { "epoch": 0.28, "learning_rate": 8.492832579530188e-06, "loss": 1.387, "step": 2590 }, { "epoch": 0.28, "learning_rate": 8.491596282174523e-06, "loss": 1.5521, "step": 2591 }, { "epoch": 0.28, "learning_rate": 8.490359568036446e-06, "loss": 1.6685, "step": 2592 }, { "epoch": 0.28, "learning_rate": 8.489122437263585e-06, "loss": 1.5092, "step": 2593 }, { "epoch": 0.28, "learning_rate": 8.48788489000361e-06, "loss": 1.4673, "step": 2594 }, { "epoch": 0.28, "learning_rate": 8.486646926404243e-06, "loss": 1.4745, "step": 2595 }, { "epoch": 0.28, "learning_rate": 8.485408546613258e-06, "loss": 1.5882, "step": 2596 }, { "epoch": 0.28, "learning_rate": 8.484169750778476e-06, "loss": 1.5266, "step": 2597 }, { "epoch": 0.28, "learning_rate": 8.48293053904777e-06, "loss": 1.675, "step": 2598 }, { "epoch": 0.28, "learning_rate": 8.48169091156906e-06, "loss": 1.3177, "step": 2599 }, { "epoch": 0.28, "learning_rate": 8.480450868490317e-06, "loss": 1.3554, "step": 2600 }, { "epoch": 0.28, "learning_rate": 8.479210409959565e-06, "loss": 1.2361, "step": 2601 }, { "epoch": 0.28, "learning_rate": 8.477969536124868e-06, "loss": 1.5132, "step": 2602 }, { "epoch": 0.28, "learning_rate": 8.476728247134352e-06, "loss": 1.5235, "step": 2603 }, { "epoch": 0.28, "learning_rate": 8.475486543136181e-06, "loss": 1.4709, "step": 2604 }, { "epoch": 0.28, "learning_rate": 8.474244424278578e-06, "loss": 1.6788, "step": 2605 }, { "epoch": 0.28, "learning_rate": 8.47300189070981e-06, "loss": 1.5054, "step": 2606 }, { "epoch": 0.28, "learning_rate": 8.471758942578193e-06, "loss": 1.5886, "step": 2607 }, { "epoch": 0.28, "learning_rate": 8.470515580032096e-06, "loss": 1.4344, "step": 2608 }, { "epoch": 0.28, "learning_rate": 8.469271803219939e-06, "loss": 1.2252, "step": 2609 }, { "epoch": 0.28, "learning_rate": 8.46802761229018e-06, "loss": 1.529, "step": 2610 }, { "epoch": 0.28, "learning_rate": 8.466783007391342e-06, "loss": 1.6565, "step": 2611 }, { "epoch": 0.28, "learning_rate": 8.465537988671987e-06, "loss": 1.646, "step": 2612 }, { "epoch": 0.28, "learning_rate": 8.464292556280734e-06, "loss": 1.5626, "step": 2613 }, { "epoch": 0.28, "learning_rate": 8.463046710366239e-06, "loss": 1.636, "step": 2614 }, { "epoch": 0.28, "learning_rate": 8.46180045107722e-06, "loss": 1.3083, "step": 2615 }, { "epoch": 0.28, "learning_rate": 8.46055377856244e-06, "loss": 1.605, "step": 2616 }, { "epoch": 0.28, "learning_rate": 8.459306692970712e-06, "loss": 1.3115, "step": 2617 }, { "epoch": 0.28, "learning_rate": 8.458059194450895e-06, "loss": 1.1818, "step": 2618 }, { "epoch": 0.28, "learning_rate": 8.456811283151896e-06, "loss": 1.5044, "step": 2619 }, { "epoch": 0.28, "learning_rate": 8.455562959222682e-06, "loss": 1.6259, "step": 2620 }, { "epoch": 0.28, "learning_rate": 8.454314222812257e-06, "loss": 1.4225, "step": 2621 }, { "epoch": 0.28, "learning_rate": 8.453065074069682e-06, "loss": 1.272, "step": 2622 }, { "epoch": 0.28, "learning_rate": 8.451815513144062e-06, "loss": 1.2127, "step": 2623 }, { "epoch": 0.28, "learning_rate": 8.450565540184557e-06, "loss": 1.4612, "step": 2624 }, { "epoch": 0.28, "learning_rate": 8.449315155340369e-06, "loss": 1.3925, "step": 2625 }, { "epoch": 0.28, "learning_rate": 8.448064358760757e-06, "loss": 1.4634, "step": 2626 }, { "epoch": 0.28, "learning_rate": 8.446813150595022e-06, "loss": 1.5229, "step": 2627 }, { "epoch": 0.28, "learning_rate": 8.44556153099252e-06, "loss": 1.2855, "step": 2628 }, { "epoch": 0.28, "learning_rate": 8.444309500102651e-06, "loss": 1.3474, "step": 2629 }, { "epoch": 0.28, "learning_rate": 8.443057058074869e-06, "loss": 1.4976, "step": 2630 }, { "epoch": 0.28, "learning_rate": 8.441804205058672e-06, "loss": 1.1262, "step": 2631 }, { "epoch": 0.28, "learning_rate": 8.44055094120361e-06, "loss": 1.6133, "step": 2632 }, { "epoch": 0.28, "learning_rate": 8.439297266659285e-06, "loss": 1.6402, "step": 2633 }, { "epoch": 0.28, "learning_rate": 8.43804318157534e-06, "loss": 1.1545, "step": 2634 }, { "epoch": 0.28, "learning_rate": 8.436788686101475e-06, "loss": 1.4556, "step": 2635 }, { "epoch": 0.28, "learning_rate": 8.435533780387436e-06, "loss": 1.5833, "step": 2636 }, { "epoch": 0.28, "learning_rate": 8.434278464583018e-06, "loss": 1.7065, "step": 2637 }, { "epoch": 0.28, "learning_rate": 8.43302273883806e-06, "loss": 1.4065, "step": 2638 }, { "epoch": 0.28, "learning_rate": 8.431766603302459e-06, "loss": 1.6966, "step": 2639 }, { "epoch": 0.28, "learning_rate": 8.430510058126156e-06, "loss": 1.2194, "step": 2640 }, { "epoch": 0.28, "learning_rate": 8.429253103459139e-06, "loss": 1.7047, "step": 2641 }, { "epoch": 0.28, "learning_rate": 8.427995739451451e-06, "loss": 1.5483, "step": 2642 }, { "epoch": 0.28, "learning_rate": 8.426737966253176e-06, "loss": 1.5225, "step": 2643 }, { "epoch": 0.28, "learning_rate": 8.425479784014455e-06, "loss": 1.4946, "step": 2644 }, { "epoch": 0.28, "learning_rate": 8.42422119288547e-06, "loss": 1.7481, "step": 2645 }, { "epoch": 0.28, "learning_rate": 8.422962193016459e-06, "loss": 1.3259, "step": 2646 }, { "epoch": 0.28, "learning_rate": 8.421702784557704e-06, "loss": 1.5604, "step": 2647 }, { "epoch": 0.28, "learning_rate": 8.42044296765954e-06, "loss": 1.6996, "step": 2648 }, { "epoch": 0.28, "learning_rate": 8.41918274247234e-06, "loss": 1.058, "step": 2649 }, { "epoch": 0.28, "learning_rate": 8.417922109146541e-06, "loss": 1.5949, "step": 2650 }, { "epoch": 0.28, "learning_rate": 8.41666106783262e-06, "loss": 1.5111, "step": 2651 }, { "epoch": 0.28, "learning_rate": 8.415399618681101e-06, "loss": 1.6061, "step": 2652 }, { "epoch": 0.28, "learning_rate": 8.414137761842565e-06, "loss": 1.4543, "step": 2653 }, { "epoch": 0.28, "learning_rate": 8.412875497467632e-06, "loss": 1.4088, "step": 2654 }, { "epoch": 0.28, "learning_rate": 8.411612825706976e-06, "loss": 1.3901, "step": 2655 }, { "epoch": 0.28, "learning_rate": 8.41034974671132e-06, "loss": 1.4884, "step": 2656 }, { "epoch": 0.28, "learning_rate": 8.409086260631434e-06, "loss": 1.7239, "step": 2657 }, { "epoch": 0.28, "learning_rate": 8.407822367618135e-06, "loss": 1.1329, "step": 2658 }, { "epoch": 0.28, "learning_rate": 8.406558067822294e-06, "loss": 1.6366, "step": 2659 }, { "epoch": 0.28, "learning_rate": 8.405293361394825e-06, "loss": 1.4008, "step": 2660 }, { "epoch": 0.28, "learning_rate": 8.40402824848669e-06, "loss": 1.4969, "step": 2661 }, { "epoch": 0.28, "learning_rate": 8.402762729248907e-06, "loss": 1.4658, "step": 2662 }, { "epoch": 0.28, "learning_rate": 8.401496803832534e-06, "loss": 1.5138, "step": 2663 }, { "epoch": 0.28, "learning_rate": 8.400230472388684e-06, "loss": 1.6184, "step": 2664 }, { "epoch": 0.28, "learning_rate": 8.398963735068515e-06, "loss": 1.6271, "step": 2665 }, { "epoch": 0.28, "learning_rate": 8.39769659202323e-06, "loss": 1.7605, "step": 2666 }, { "epoch": 0.28, "learning_rate": 8.396429043404088e-06, "loss": 1.4991, "step": 2667 }, { "epoch": 0.28, "learning_rate": 8.395161089362393e-06, "loss": 1.5108, "step": 2668 }, { "epoch": 0.28, "learning_rate": 8.393892730049497e-06, "loss": 1.5894, "step": 2669 }, { "epoch": 0.28, "learning_rate": 8.3926239656168e-06, "loss": 1.3433, "step": 2670 }, { "epoch": 0.28, "learning_rate": 8.391354796215751e-06, "loss": 1.3737, "step": 2671 }, { "epoch": 0.29, "learning_rate": 8.390085221997849e-06, "loss": 1.5909, "step": 2672 }, { "epoch": 0.29, "learning_rate": 8.388815243114637e-06, "loss": 1.409, "step": 2673 }, { "epoch": 0.29, "learning_rate": 8.387544859717709e-06, "loss": 1.3945, "step": 2674 }, { "epoch": 0.29, "learning_rate": 8.38627407195871e-06, "loss": 1.4654, "step": 2675 }, { "epoch": 0.29, "learning_rate": 8.385002879989328e-06, "loss": 1.422, "step": 2676 }, { "epoch": 0.29, "learning_rate": 8.383731283961303e-06, "loss": 1.6927, "step": 2677 }, { "epoch": 0.29, "learning_rate": 8.382459284026421e-06, "loss": 1.3692, "step": 2678 }, { "epoch": 0.29, "learning_rate": 8.381186880336518e-06, "loss": 1.5005, "step": 2679 }, { "epoch": 0.29, "learning_rate": 8.379914073043477e-06, "loss": 1.4843, "step": 2680 }, { "epoch": 0.29, "learning_rate": 8.37864086229923e-06, "loss": 1.3837, "step": 2681 }, { "epoch": 0.29, "learning_rate": 8.377367248255757e-06, "loss": 1.3855, "step": 2682 }, { "epoch": 0.29, "learning_rate": 8.376093231065084e-06, "loss": 1.3726, "step": 2683 }, { "epoch": 0.29, "learning_rate": 8.374818810879288e-06, "loss": 1.267, "step": 2684 }, { "epoch": 0.29, "learning_rate": 8.373543987850494e-06, "loss": 1.7588, "step": 2685 }, { "epoch": 0.29, "learning_rate": 8.372268762130874e-06, "loss": 1.3311, "step": 2686 }, { "epoch": 0.29, "learning_rate": 8.370993133872648e-06, "loss": 1.474, "step": 2687 }, { "epoch": 0.29, "learning_rate": 8.369717103228084e-06, "loss": 1.7962, "step": 2688 }, { "epoch": 0.29, "learning_rate": 8.368440670349498e-06, "loss": 1.4757, "step": 2689 }, { "epoch": 0.29, "learning_rate": 8.367163835389253e-06, "loss": 1.4639, "step": 2690 }, { "epoch": 0.29, "learning_rate": 8.365886598499766e-06, "loss": 1.2704, "step": 2691 }, { "epoch": 0.29, "learning_rate": 8.364608959833495e-06, "loss": 1.4044, "step": 2692 }, { "epoch": 0.29, "learning_rate": 8.363330919542944e-06, "loss": 1.2734, "step": 2693 }, { "epoch": 0.29, "learning_rate": 8.362052477780677e-06, "loss": 1.743, "step": 2694 }, { "epoch": 0.29, "learning_rate": 8.360773634699291e-06, "loss": 1.5013, "step": 2695 }, { "epoch": 0.29, "learning_rate": 8.359494390451442e-06, "loss": 1.6473, "step": 2696 }, { "epoch": 0.29, "learning_rate": 8.35821474518983e-06, "loss": 1.4256, "step": 2697 }, { "epoch": 0.29, "learning_rate": 8.3569346990672e-06, "loss": 1.35, "step": 2698 }, { "epoch": 0.29, "learning_rate": 8.355654252236352e-06, "loss": 1.5909, "step": 2699 }, { "epoch": 0.29, "learning_rate": 8.354373404850124e-06, "loss": 1.5576, "step": 2700 }, { "epoch": 0.29, "learning_rate": 8.353092157061412e-06, "loss": 1.5585, "step": 2701 }, { "epoch": 0.29, "learning_rate": 8.351810509023153e-06, "loss": 1.4072, "step": 2702 }, { "epoch": 0.29, "learning_rate": 8.350528460888334e-06, "loss": 1.4644, "step": 2703 }, { "epoch": 0.29, "learning_rate": 8.349246012809991e-06, "loss": 1.2796, "step": 2704 }, { "epoch": 0.29, "learning_rate": 8.347963164941204e-06, "loss": 1.477, "step": 2705 }, { "epoch": 0.29, "learning_rate": 8.346679917435104e-06, "loss": 1.3049, "step": 2706 }, { "epoch": 0.29, "learning_rate": 8.345396270444868e-06, "loss": 1.4991, "step": 2707 }, { "epoch": 0.29, "learning_rate": 8.344112224123723e-06, "loss": 1.5413, "step": 2708 }, { "epoch": 0.29, "learning_rate": 8.342827778624943e-06, "loss": 1.4831, "step": 2709 }, { "epoch": 0.29, "learning_rate": 8.341542934101847e-06, "loss": 1.0932, "step": 2710 }, { "epoch": 0.29, "learning_rate": 8.340257690707805e-06, "loss": 1.2835, "step": 2711 }, { "epoch": 0.29, "learning_rate": 8.33897204859623e-06, "loss": 1.1611, "step": 2712 }, { "epoch": 0.29, "learning_rate": 8.337686007920586e-06, "loss": 1.4113, "step": 2713 }, { "epoch": 0.29, "learning_rate": 8.33639956883439e-06, "loss": 1.7278, "step": 2714 }, { "epoch": 0.29, "learning_rate": 8.335112731491192e-06, "loss": 1.5285, "step": 2715 }, { "epoch": 0.29, "learning_rate": 8.333825496044605e-06, "loss": 1.4848, "step": 2716 }, { "epoch": 0.29, "learning_rate": 8.33253786264828e-06, "loss": 1.6016, "step": 2717 }, { "epoch": 0.29, "learning_rate": 8.331249831455921e-06, "loss": 1.5319, "step": 2718 }, { "epoch": 0.29, "learning_rate": 8.329961402621274e-06, "loss": 1.6509, "step": 2719 }, { "epoch": 0.29, "learning_rate": 8.328672576298133e-06, "loss": 1.6728, "step": 2720 }, { "epoch": 0.29, "learning_rate": 8.327383352640347e-06, "loss": 1.6201, "step": 2721 }, { "epoch": 0.29, "learning_rate": 8.326093731801805e-06, "loss": 1.3198, "step": 2722 }, { "epoch": 0.29, "learning_rate": 8.324803713936445e-06, "loss": 1.3678, "step": 2723 }, { "epoch": 0.29, "learning_rate": 8.323513299198252e-06, "loss": 1.5809, "step": 2724 }, { "epoch": 0.29, "learning_rate": 8.322222487741261e-06, "loss": 1.469, "step": 2725 }, { "epoch": 0.29, "learning_rate": 8.320931279719553e-06, "loss": 1.5128, "step": 2726 }, { "epoch": 0.29, "learning_rate": 8.319639675287255e-06, "loss": 1.4415, "step": 2727 }, { "epoch": 0.29, "learning_rate": 8.318347674598544e-06, "loss": 1.5403, "step": 2728 }, { "epoch": 0.29, "learning_rate": 8.317055277807639e-06, "loss": 1.7256, "step": 2729 }, { "epoch": 0.29, "learning_rate": 8.315762485068815e-06, "loss": 1.6879, "step": 2730 }, { "epoch": 0.29, "learning_rate": 8.314469296536384e-06, "loss": 1.4348, "step": 2731 }, { "epoch": 0.29, "learning_rate": 8.313175712364712e-06, "loss": 1.4485, "step": 2732 }, { "epoch": 0.29, "learning_rate": 8.311881732708213e-06, "loss": 1.391, "step": 2733 }, { "epoch": 0.29, "learning_rate": 8.310587357721345e-06, "loss": 1.3973, "step": 2734 }, { "epoch": 0.29, "learning_rate": 8.309292587558612e-06, "loss": 1.4528, "step": 2735 }, { "epoch": 0.29, "learning_rate": 8.307997422374569e-06, "loss": 1.4429, "step": 2736 }, { "epoch": 0.29, "learning_rate": 8.306701862323815e-06, "loss": 1.4248, "step": 2737 }, { "epoch": 0.29, "learning_rate": 8.305405907560999e-06, "loss": 1.49, "step": 2738 }, { "epoch": 0.29, "learning_rate": 8.304109558240817e-06, "loss": 1.6259, "step": 2739 }, { "epoch": 0.29, "learning_rate": 8.302812814518006e-06, "loss": 1.6051, "step": 2740 }, { "epoch": 0.29, "learning_rate": 8.301515676547358e-06, "loss": 1.4154, "step": 2741 }, { "epoch": 0.29, "learning_rate": 8.300218144483709e-06, "loss": 1.453, "step": 2742 }, { "epoch": 0.29, "learning_rate": 8.298920218481941e-06, "loss": 1.4381, "step": 2743 }, { "epoch": 0.29, "learning_rate": 8.297621898696984e-06, "loss": 1.5336, "step": 2744 }, { "epoch": 0.29, "learning_rate": 8.296323185283816e-06, "loss": 1.6495, "step": 2745 }, { "epoch": 0.29, "learning_rate": 8.29502407839746e-06, "loss": 1.7824, "step": 2746 }, { "epoch": 0.29, "learning_rate": 8.293724578192986e-06, "loss": 1.2129, "step": 2747 }, { "epoch": 0.29, "learning_rate": 8.292424684825514e-06, "loss": 1.5404, "step": 2748 }, { "epoch": 0.29, "learning_rate": 8.291124398450204e-06, "loss": 1.5294, "step": 2749 }, { "epoch": 0.29, "learning_rate": 8.289823719222275e-06, "loss": 1.422, "step": 2750 }, { "epoch": 0.29, "learning_rate": 8.28852264729698e-06, "loss": 1.4528, "step": 2751 }, { "epoch": 0.29, "learning_rate": 8.287221182829626e-06, "loss": 1.5937, "step": 2752 }, { "epoch": 0.29, "learning_rate": 8.285919325975566e-06, "loss": 1.4997, "step": 2753 }, { "epoch": 0.29, "learning_rate": 8.284617076890199e-06, "loss": 1.6036, "step": 2754 }, { "epoch": 0.29, "learning_rate": 8.283314435728968e-06, "loss": 1.4967, "step": 2755 }, { "epoch": 0.29, "learning_rate": 8.28201140264737e-06, "loss": 1.399, "step": 2756 }, { "epoch": 0.29, "learning_rate": 8.280707977800944e-06, "loss": 1.5316, "step": 2757 }, { "epoch": 0.29, "learning_rate": 8.279404161345275e-06, "loss": 1.4984, "step": 2758 }, { "epoch": 0.29, "learning_rate": 8.278099953435995e-06, "loss": 1.6149, "step": 2759 }, { "epoch": 0.29, "learning_rate": 8.276795354228785e-06, "loss": 1.5915, "step": 2760 }, { "epoch": 0.29, "learning_rate": 8.275490363879372e-06, "loss": 1.5916, "step": 2761 }, { "epoch": 0.29, "learning_rate": 8.274184982543527e-06, "loss": 1.286, "step": 2762 }, { "epoch": 0.29, "learning_rate": 8.272879210377074e-06, "loss": 1.6972, "step": 2763 }, { "epoch": 0.29, "learning_rate": 8.271573047535875e-06, "loss": 1.4881, "step": 2764 }, { "epoch": 0.29, "learning_rate": 8.270266494175847e-06, "loss": 1.3245, "step": 2765 }, { "epoch": 0.3, "learning_rate": 8.268959550452946e-06, "loss": 1.4961, "step": 2766 }, { "epoch": 0.3, "learning_rate": 8.26765221652318e-06, "loss": 1.5004, "step": 2767 }, { "epoch": 0.3, "learning_rate": 8.266344492542603e-06, "loss": 1.3025, "step": 2768 }, { "epoch": 0.3, "learning_rate": 8.265036378667312e-06, "loss": 1.5993, "step": 2769 }, { "epoch": 0.3, "learning_rate": 8.263727875053457e-06, "loss": 1.6322, "step": 2770 }, { "epoch": 0.3, "learning_rate": 8.262418981857226e-06, "loss": 1.3569, "step": 2771 }, { "epoch": 0.3, "learning_rate": 8.261109699234862e-06, "loss": 1.6072, "step": 2772 }, { "epoch": 0.3, "learning_rate": 8.259800027342645e-06, "loss": 1.5739, "step": 2773 }, { "epoch": 0.3, "learning_rate": 8.258489966336915e-06, "loss": 1.5439, "step": 2774 }, { "epoch": 0.3, "learning_rate": 8.257179516374045e-06, "loss": 1.3928, "step": 2775 }, { "epoch": 0.3, "learning_rate": 8.25586867761046e-06, "loss": 1.2103, "step": 2776 }, { "epoch": 0.3, "learning_rate": 8.25455745020263e-06, "loss": 1.4604, "step": 2777 }, { "epoch": 0.3, "learning_rate": 8.253245834307079e-06, "loss": 1.3013, "step": 2778 }, { "epoch": 0.3, "learning_rate": 8.251933830080365e-06, "loss": 1.3526, "step": 2779 }, { "epoch": 0.3, "learning_rate": 8.250621437679103e-06, "loss": 1.6695, "step": 2780 }, { "epoch": 0.3, "learning_rate": 8.249308657259943e-06, "loss": 1.5739, "step": 2781 }, { "epoch": 0.3, "learning_rate": 8.247995488979594e-06, "loss": 1.2944, "step": 2782 }, { "epoch": 0.3, "learning_rate": 8.246681932994804e-06, "loss": 1.5764, "step": 2783 }, { "epoch": 0.3, "learning_rate": 8.245367989462368e-06, "loss": 1.4452, "step": 2784 }, { "epoch": 0.3, "learning_rate": 8.244053658539127e-06, "loss": 1.4012, "step": 2785 }, { "epoch": 0.3, "learning_rate": 8.24273894038197e-06, "loss": 1.7239, "step": 2786 }, { "epoch": 0.3, "learning_rate": 8.241423835147833e-06, "loss": 1.4163, "step": 2787 }, { "epoch": 0.3, "learning_rate": 8.240108342993694e-06, "loss": 1.5105, "step": 2788 }, { "epoch": 0.3, "learning_rate": 8.238792464076582e-06, "loss": 1.342, "step": 2789 }, { "epoch": 0.3, "learning_rate": 8.237476198553567e-06, "loss": 1.4726, "step": 2790 }, { "epoch": 0.3, "learning_rate": 8.23615954658177e-06, "loss": 1.2393, "step": 2791 }, { "epoch": 0.3, "learning_rate": 8.234842508318357e-06, "loss": 1.5054, "step": 2792 }, { "epoch": 0.3, "learning_rate": 8.233525083920536e-06, "loss": 1.6372, "step": 2793 }, { "epoch": 0.3, "learning_rate": 8.23220727354557e-06, "loss": 1.5304, "step": 2794 }, { "epoch": 0.3, "learning_rate": 8.230889077350755e-06, "loss": 1.5833, "step": 2795 }, { "epoch": 0.3, "learning_rate": 8.229570495493447e-06, "loss": 1.5915, "step": 2796 }, { "epoch": 0.3, "learning_rate": 8.22825152813104e-06, "loss": 1.3702, "step": 2797 }, { "epoch": 0.3, "learning_rate": 8.226932175420972e-06, "loss": 1.4104, "step": 2798 }, { "epoch": 0.3, "learning_rate": 8.225612437520736e-06, "loss": 1.6106, "step": 2799 }, { "epoch": 0.3, "learning_rate": 8.224292314587862e-06, "loss": 1.5453, "step": 2800 }, { "epoch": 0.3, "learning_rate": 8.222971806779929e-06, "loss": 1.3086, "step": 2801 }, { "epoch": 0.3, "learning_rate": 8.221650914254566e-06, "loss": 1.5909, "step": 2802 }, { "epoch": 0.3, "learning_rate": 8.220329637169441e-06, "loss": 1.5312, "step": 2803 }, { "epoch": 0.3, "learning_rate": 8.219007975682273e-06, "loss": 1.6347, "step": 2804 }, { "epoch": 0.3, "learning_rate": 8.217685929950823e-06, "loss": 1.5921, "step": 2805 }, { "epoch": 0.3, "learning_rate": 8.216363500132903e-06, "loss": 1.4834, "step": 2806 }, { "epoch": 0.3, "learning_rate": 8.215040686386367e-06, "loss": 1.5103, "step": 2807 }, { "epoch": 0.3, "learning_rate": 8.213717488869113e-06, "loss": 1.5775, "step": 2808 }, { "epoch": 0.3, "learning_rate": 8.21239390773909e-06, "loss": 1.3376, "step": 2809 }, { "epoch": 0.3, "learning_rate": 8.211069943154292e-06, "loss": 1.3961, "step": 2810 }, { "epoch": 0.3, "learning_rate": 8.209745595272755e-06, "loss": 1.4275, "step": 2811 }, { "epoch": 0.3, "learning_rate": 8.208420864252562e-06, "loss": 1.7415, "step": 2812 }, { "epoch": 0.3, "learning_rate": 8.207095750251843e-06, "loss": 1.5838, "step": 2813 }, { "epoch": 0.3, "learning_rate": 8.205770253428775e-06, "loss": 1.7271, "step": 2814 }, { "epoch": 0.3, "learning_rate": 8.204444373941576e-06, "loss": 1.5546, "step": 2815 }, { "epoch": 0.3, "learning_rate": 8.203118111948516e-06, "loss": 1.6647, "step": 2816 }, { "epoch": 0.3, "learning_rate": 8.201791467607905e-06, "loss": 1.5384, "step": 2817 }, { "epoch": 0.3, "learning_rate": 8.2004644410781e-06, "loss": 1.5677, "step": 2818 }, { "epoch": 0.3, "learning_rate": 8.199137032517507e-06, "loss": 1.5666, "step": 2819 }, { "epoch": 0.3, "learning_rate": 8.197809242084575e-06, "loss": 1.4127, "step": 2820 }, { "epoch": 0.3, "learning_rate": 8.1964810699378e-06, "loss": 1.5529, "step": 2821 }, { "epoch": 0.3, "learning_rate": 8.195152516235718e-06, "loss": 1.7485, "step": 2822 }, { "epoch": 0.3, "learning_rate": 8.193823581136919e-06, "loss": 1.7303, "step": 2823 }, { "epoch": 0.3, "learning_rate": 8.19249426480003e-06, "loss": 1.5313, "step": 2824 }, { "epoch": 0.3, "learning_rate": 8.191164567383733e-06, "loss": 1.28, "step": 2825 }, { "epoch": 0.3, "learning_rate": 8.189834489046746e-06, "loss": 1.3196, "step": 2826 }, { "epoch": 0.3, "learning_rate": 8.188504029947841e-06, "loss": 1.2696, "step": 2827 }, { "epoch": 0.3, "learning_rate": 8.187173190245827e-06, "loss": 1.6155, "step": 2828 }, { "epoch": 0.3, "learning_rate": 8.185841970099566e-06, "loss": 1.6154, "step": 2829 }, { "epoch": 0.3, "learning_rate": 8.184510369667962e-06, "loss": 1.6004, "step": 2830 }, { "epoch": 0.3, "learning_rate": 8.183178389109963e-06, "loss": 1.6455, "step": 2831 }, { "epoch": 0.3, "learning_rate": 8.181846028584563e-06, "loss": 1.3387, "step": 2832 }, { "epoch": 0.3, "learning_rate": 8.180513288250804e-06, "loss": 1.2384, "step": 2833 }, { "epoch": 0.3, "learning_rate": 8.179180168267772e-06, "loss": 1.8114, "step": 2834 }, { "epoch": 0.3, "learning_rate": 8.177846668794598e-06, "loss": 1.3449, "step": 2835 }, { "epoch": 0.3, "learning_rate": 8.176512789990457e-06, "loss": 1.5357, "step": 2836 }, { "epoch": 0.3, "learning_rate": 8.175178532014571e-06, "loss": 1.6103, "step": 2837 }, { "epoch": 0.3, "learning_rate": 8.173843895026207e-06, "loss": 1.2611, "step": 2838 }, { "epoch": 0.3, "learning_rate": 8.172508879184675e-06, "loss": 1.7331, "step": 2839 }, { "epoch": 0.3, "learning_rate": 8.171173484649337e-06, "loss": 1.4845, "step": 2840 }, { "epoch": 0.3, "learning_rate": 8.169837711579591e-06, "loss": 1.3435, "step": 2841 }, { "epoch": 0.3, "learning_rate": 8.168501560134886e-06, "loss": 1.5846, "step": 2842 }, { "epoch": 0.3, "learning_rate": 8.167165030474715e-06, "loss": 1.5922, "step": 2843 }, { "epoch": 0.3, "learning_rate": 8.165828122758615e-06, "loss": 1.5836, "step": 2844 }, { "epoch": 0.3, "learning_rate": 8.164490837146173e-06, "loss": 1.6711, "step": 2845 }, { "epoch": 0.3, "learning_rate": 8.16315317379701e-06, "loss": 1.2937, "step": 2846 }, { "epoch": 0.3, "learning_rate": 8.161815132870806e-06, "loss": 1.5646, "step": 2847 }, { "epoch": 0.3, "learning_rate": 8.160476714527274e-06, "loss": 1.6486, "step": 2848 }, { "epoch": 0.3, "learning_rate": 8.159137918926182e-06, "loss": 1.5808, "step": 2849 }, { "epoch": 0.3, "learning_rate": 8.157798746227337e-06, "loss": 1.4979, "step": 2850 }, { "epoch": 0.3, "learning_rate": 8.15645919659059e-06, "loss": 1.4466, "step": 2851 }, { "epoch": 0.3, "learning_rate": 8.155119270175842e-06, "loss": 1.5086, "step": 2852 }, { "epoch": 0.3, "learning_rate": 8.153778967143035e-06, "loss": 1.4281, "step": 2853 }, { "epoch": 0.3, "learning_rate": 8.152438287652161e-06, "loss": 1.6304, "step": 2854 }, { "epoch": 0.3, "learning_rate": 8.151097231863247e-06, "loss": 1.5475, "step": 2855 }, { "epoch": 0.3, "learning_rate": 8.149755799936377e-06, "loss": 1.5296, "step": 2856 }, { "epoch": 0.3, "learning_rate": 8.14841399203167e-06, "loss": 1.2817, "step": 2857 }, { "epoch": 0.3, "learning_rate": 8.147071808309295e-06, "loss": 1.3598, "step": 2858 }, { "epoch": 0.3, "learning_rate": 8.145729248929466e-06, "loss": 1.3921, "step": 2859 }, { "epoch": 0.31, "learning_rate": 8.14438631405244e-06, "loss": 1.4562, "step": 2860 }, { "epoch": 0.31, "learning_rate": 8.14304300383852e-06, "loss": 1.5565, "step": 2861 }, { "epoch": 0.31, "learning_rate": 8.141699318448053e-06, "loss": 1.6406, "step": 2862 }, { "epoch": 0.31, "learning_rate": 8.140355258041431e-06, "loss": 1.6074, "step": 2863 }, { "epoch": 0.31, "learning_rate": 8.13901082277909e-06, "loss": 1.3919, "step": 2864 }, { "epoch": 0.31, "learning_rate": 8.137666012821514e-06, "loss": 1.6377, "step": 2865 }, { "epoch": 0.31, "learning_rate": 8.136320828329227e-06, "loss": 1.6495, "step": 2866 }, { "epoch": 0.31, "learning_rate": 8.134975269462801e-06, "loss": 1.3707, "step": 2867 }, { "epoch": 0.31, "learning_rate": 8.13362933638285e-06, "loss": 1.4956, "step": 2868 }, { "epoch": 0.31, "learning_rate": 8.132283029250038e-06, "loss": 1.5405, "step": 2869 }, { "epoch": 0.31, "learning_rate": 8.130936348225069e-06, "loss": 1.48, "step": 2870 }, { "epoch": 0.31, "learning_rate": 8.129589293468689e-06, "loss": 1.7637, "step": 2871 }, { "epoch": 0.31, "learning_rate": 8.128241865141697e-06, "loss": 1.4131, "step": 2872 }, { "epoch": 0.31, "learning_rate": 8.12689406340493e-06, "loss": 1.4515, "step": 2873 }, { "epoch": 0.31, "learning_rate": 8.125545888419269e-06, "loss": 1.3284, "step": 2874 }, { "epoch": 0.31, "learning_rate": 8.124197340345646e-06, "loss": 1.4373, "step": 2875 }, { "epoch": 0.31, "learning_rate": 8.12284841934503e-06, "loss": 1.4073, "step": 2876 }, { "epoch": 0.31, "learning_rate": 8.12149912557844e-06, "loss": 1.5577, "step": 2877 }, { "epoch": 0.31, "learning_rate": 8.120149459206942e-06, "loss": 1.4036, "step": 2878 }, { "epoch": 0.31, "learning_rate": 8.118799420391632e-06, "loss": 1.5025, "step": 2879 }, { "epoch": 0.31, "learning_rate": 8.117449009293668e-06, "loss": 1.7281, "step": 2880 }, { "epoch": 0.31, "learning_rate": 8.116098226074244e-06, "loss": 1.4658, "step": 2881 }, { "epoch": 0.31, "learning_rate": 8.114747070894597e-06, "loss": 1.3952, "step": 2882 }, { "epoch": 0.31, "learning_rate": 8.113395543916012e-06, "loss": 1.6108, "step": 2883 }, { "epoch": 0.31, "learning_rate": 8.112043645299817e-06, "loss": 1.497, "step": 2884 }, { "epoch": 0.31, "learning_rate": 8.110691375207385e-06, "loss": 1.3522, "step": 2885 }, { "epoch": 0.31, "learning_rate": 8.109338733800132e-06, "loss": 1.5436, "step": 2886 }, { "epoch": 0.31, "learning_rate": 8.10798572123952e-06, "loss": 1.2696, "step": 2887 }, { "epoch": 0.31, "learning_rate": 8.106632337687052e-06, "loss": 1.5915, "step": 2888 }, { "epoch": 0.31, "learning_rate": 8.10527858330428e-06, "loss": 1.4296, "step": 2889 }, { "epoch": 0.31, "learning_rate": 8.103924458252801e-06, "loss": 1.4669, "step": 2890 }, { "epoch": 0.31, "learning_rate": 8.102569962694247e-06, "loss": 1.4143, "step": 2891 }, { "epoch": 0.31, "learning_rate": 8.101215096790305e-06, "loss": 1.4963, "step": 2892 }, { "epoch": 0.31, "learning_rate": 8.099859860702698e-06, "loss": 1.4497, "step": 2893 }, { "epoch": 0.31, "learning_rate": 8.098504254593203e-06, "loss": 1.4647, "step": 2894 }, { "epoch": 0.31, "learning_rate": 8.097148278623628e-06, "loss": 1.8112, "step": 2895 }, { "epoch": 0.31, "learning_rate": 8.095791932955836e-06, "loss": 1.5674, "step": 2896 }, { "epoch": 0.31, "learning_rate": 8.09443521775173e-06, "loss": 1.5964, "step": 2897 }, { "epoch": 0.31, "learning_rate": 8.093078133173256e-06, "loss": 1.3803, "step": 2898 }, { "epoch": 0.31, "learning_rate": 8.091720679382407e-06, "loss": 1.5679, "step": 2899 }, { "epoch": 0.31, "learning_rate": 8.090362856541218e-06, "loss": 1.3479, "step": 2900 }, { "epoch": 0.31, "learning_rate": 8.089004664811767e-06, "loss": 0.9907, "step": 2901 }, { "epoch": 0.31, "learning_rate": 8.087646104356181e-06, "loss": 1.6856, "step": 2902 }, { "epoch": 0.31, "learning_rate": 8.086287175336625e-06, "loss": 1.38, "step": 2903 }, { "epoch": 0.31, "learning_rate": 8.084927877915314e-06, "loss": 1.4735, "step": 2904 }, { "epoch": 0.31, "learning_rate": 8.0835682122545e-06, "loss": 1.6422, "step": 2905 }, { "epoch": 0.31, "learning_rate": 8.082208178516484e-06, "loss": 1.4869, "step": 2906 }, { "epoch": 0.31, "learning_rate": 8.080847776863609e-06, "loss": 1.7651, "step": 2907 }, { "epoch": 0.31, "learning_rate": 8.079487007458265e-06, "loss": 1.4948, "step": 2908 }, { "epoch": 0.31, "learning_rate": 8.078125870462878e-06, "loss": 1.4226, "step": 2909 }, { "epoch": 0.31, "learning_rate": 8.07676436603993e-06, "loss": 1.4411, "step": 2910 }, { "epoch": 0.31, "learning_rate": 8.075402494351936e-06, "loss": 1.3351, "step": 2911 }, { "epoch": 0.31, "learning_rate": 8.07404025556146e-06, "loss": 1.4705, "step": 2912 }, { "epoch": 0.31, "learning_rate": 8.072677649831107e-06, "loss": 1.3845, "step": 2913 }, { "epoch": 0.31, "learning_rate": 8.07131467732353e-06, "loss": 1.6093, "step": 2914 }, { "epoch": 0.31, "learning_rate": 8.069951338201421e-06, "loss": 1.7447, "step": 2915 }, { "epoch": 0.31, "learning_rate": 8.068587632627521e-06, "loss": 1.4403, "step": 2916 }, { "epoch": 0.31, "learning_rate": 8.06722356076461e-06, "loss": 1.1632, "step": 2917 }, { "epoch": 0.31, "learning_rate": 8.065859122775513e-06, "loss": 1.4105, "step": 2918 }, { "epoch": 0.31, "learning_rate": 8.064494318823102e-06, "loss": 1.6647, "step": 2919 }, { "epoch": 0.31, "learning_rate": 8.063129149070286e-06, "loss": 1.5435, "step": 2920 }, { "epoch": 0.31, "learning_rate": 8.061763613680024e-06, "loss": 1.7761, "step": 2921 }, { "epoch": 0.31, "learning_rate": 8.060397712815318e-06, "loss": 1.1457, "step": 2922 }, { "epoch": 0.31, "learning_rate": 8.059031446639208e-06, "loss": 1.3184, "step": 2923 }, { "epoch": 0.31, "learning_rate": 8.057664815314784e-06, "loss": 1.3882, "step": 2924 }, { "epoch": 0.31, "learning_rate": 8.056297819005177e-06, "loss": 1.6243, "step": 2925 }, { "epoch": 0.31, "learning_rate": 8.05493045787356e-06, "loss": 1.4461, "step": 2926 }, { "epoch": 0.31, "learning_rate": 8.053562732083153e-06, "loss": 1.3506, "step": 2927 }, { "epoch": 0.31, "learning_rate": 8.052194641797217e-06, "loss": 1.3077, "step": 2928 }, { "epoch": 0.31, "learning_rate": 8.050826187179059e-06, "loss": 1.3571, "step": 2929 }, { "epoch": 0.31, "learning_rate": 8.049457368392024e-06, "loss": 1.4714, "step": 2930 }, { "epoch": 0.31, "learning_rate": 8.048088185599507e-06, "loss": 1.5094, "step": 2931 }, { "epoch": 0.31, "learning_rate": 8.046718638964943e-06, "loss": 1.3925, "step": 2932 }, { "epoch": 0.31, "learning_rate": 8.04534872865181e-06, "loss": 1.4919, "step": 2933 }, { "epoch": 0.31, "learning_rate": 8.043978454823632e-06, "loss": 1.4331, "step": 2934 }, { "epoch": 0.31, "learning_rate": 8.042607817643974e-06, "loss": 1.5483, "step": 2935 }, { "epoch": 0.31, "learning_rate": 8.041236817276446e-06, "loss": 1.4617, "step": 2936 }, { "epoch": 0.31, "learning_rate": 8.0398654538847e-06, "loss": 1.4022, "step": 2937 }, { "epoch": 0.31, "learning_rate": 8.038493727632432e-06, "loss": 1.3262, "step": 2938 }, { "epoch": 0.31, "learning_rate": 8.037121638683383e-06, "loss": 1.4851, "step": 2939 }, { "epoch": 0.31, "learning_rate": 8.035749187201333e-06, "loss": 1.5041, "step": 2940 }, { "epoch": 0.31, "learning_rate": 8.034376373350109e-06, "loss": 1.3767, "step": 2941 }, { "epoch": 0.31, "learning_rate": 8.033003197293578e-06, "loss": 1.3753, "step": 2942 }, { "epoch": 0.31, "learning_rate": 8.031629659195657e-06, "loss": 1.3287, "step": 2943 }, { "epoch": 0.31, "learning_rate": 8.030255759220296e-06, "loss": 1.6525, "step": 2944 }, { "epoch": 0.31, "learning_rate": 8.028881497531498e-06, "loss": 1.5818, "step": 2945 }, { "epoch": 0.31, "learning_rate": 8.027506874293304e-06, "loss": 1.5813, "step": 2946 }, { "epoch": 0.31, "learning_rate": 8.026131889669796e-06, "loss": 1.2657, "step": 2947 }, { "epoch": 0.31, "learning_rate": 8.024756543825108e-06, "loss": 1.2227, "step": 2948 }, { "epoch": 0.31, "learning_rate": 8.023380836923404e-06, "loss": 1.3807, "step": 2949 }, { "epoch": 0.31, "learning_rate": 8.022004769128904e-06, "loss": 1.4809, "step": 2950 }, { "epoch": 0.31, "learning_rate": 8.020628340605866e-06, "loss": 1.5504, "step": 2951 }, { "epoch": 0.31, "learning_rate": 8.019251551518585e-06, "loss": 1.4541, "step": 2952 }, { "epoch": 0.31, "learning_rate": 8.017874402031409e-06, "loss": 1.3365, "step": 2953 }, { "epoch": 0.32, "learning_rate": 8.016496892308724e-06, "loss": 1.8772, "step": 2954 }, { "epoch": 0.32, "learning_rate": 8.015119022514958e-06, "loss": 1.5474, "step": 2955 }, { "epoch": 0.32, "learning_rate": 8.013740792814589e-06, "loss": 1.6178, "step": 2956 }, { "epoch": 0.32, "learning_rate": 8.012362203372124e-06, "loss": 1.618, "step": 2957 }, { "epoch": 0.32, "learning_rate": 8.010983254352127e-06, "loss": 1.5858, "step": 2958 }, { "epoch": 0.32, "learning_rate": 8.009603945919197e-06, "loss": 1.5823, "step": 2959 }, { "epoch": 0.32, "learning_rate": 8.008224278237982e-06, "loss": 1.5889, "step": 2960 }, { "epoch": 0.32, "learning_rate": 8.006844251473165e-06, "loss": 1.4198, "step": 2961 }, { "epoch": 0.32, "learning_rate": 8.005463865789477e-06, "loss": 1.614, "step": 2962 }, { "epoch": 0.32, "learning_rate": 8.004083121351695e-06, "loss": 1.4895, "step": 2963 }, { "epoch": 0.32, "learning_rate": 8.002702018324629e-06, "loss": 1.3366, "step": 2964 }, { "epoch": 0.32, "learning_rate": 8.00132055687314e-06, "loss": 1.4688, "step": 2965 }, { "epoch": 0.32, "learning_rate": 7.999938737162132e-06, "loss": 1.5007, "step": 2966 }, { "epoch": 0.32, "learning_rate": 7.998556559356543e-06, "loss": 1.3276, "step": 2967 }, { "epoch": 0.32, "learning_rate": 7.997174023621364e-06, "loss": 1.3787, "step": 2968 }, { "epoch": 0.32, "learning_rate": 7.995791130121624e-06, "loss": 1.3927, "step": 2969 }, { "epoch": 0.32, "learning_rate": 7.994407879022397e-06, "loss": 1.3278, "step": 2970 }, { "epoch": 0.32, "learning_rate": 7.993024270488794e-06, "loss": 1.2738, "step": 2971 }, { "epoch": 0.32, "learning_rate": 7.991640304685975e-06, "loss": 1.5096, "step": 2972 }, { "epoch": 0.32, "learning_rate": 7.990255981779139e-06, "loss": 1.6252, "step": 2973 }, { "epoch": 0.32, "learning_rate": 7.98887130193353e-06, "loss": 1.3864, "step": 2974 }, { "epoch": 0.32, "learning_rate": 7.987486265314435e-06, "loss": 1.527, "step": 2975 }, { "epoch": 0.32, "learning_rate": 7.986100872087177e-06, "loss": 1.443, "step": 2976 }, { "epoch": 0.32, "learning_rate": 7.984715122417133e-06, "loss": 1.6884, "step": 2977 }, { "epoch": 0.32, "learning_rate": 7.983329016469712e-06, "loss": 1.5373, "step": 2978 }, { "epoch": 0.32, "learning_rate": 7.981942554410371e-06, "loss": 1.5428, "step": 2979 }, { "epoch": 0.32, "learning_rate": 7.980555736404608e-06, "loss": 1.5276, "step": 2980 }, { "epoch": 0.32, "learning_rate": 7.979168562617965e-06, "loss": 1.2873, "step": 2981 }, { "epoch": 0.32, "learning_rate": 7.97778103321602e-06, "loss": 1.5265, "step": 2982 }, { "epoch": 0.32, "learning_rate": 7.976393148364407e-06, "loss": 1.4744, "step": 2983 }, { "epoch": 0.32, "learning_rate": 7.975004908228787e-06, "loss": 1.3631, "step": 2984 }, { "epoch": 0.32, "learning_rate": 7.973616312974876e-06, "loss": 1.5862, "step": 2985 }, { "epoch": 0.32, "learning_rate": 7.972227362768421e-06, "loss": 1.4068, "step": 2986 }, { "epoch": 0.32, "learning_rate": 7.970838057775222e-06, "loss": 1.3778, "step": 2987 }, { "epoch": 0.32, "learning_rate": 7.969448398161115e-06, "loss": 1.5547, "step": 2988 }, { "epoch": 0.32, "learning_rate": 7.968058384091978e-06, "loss": 1.645, "step": 2989 }, { "epoch": 0.32, "learning_rate": 7.966668015733736e-06, "loss": 1.7017, "step": 2990 }, { "epoch": 0.32, "learning_rate": 7.965277293252354e-06, "loss": 1.635, "step": 2991 }, { "epoch": 0.32, "learning_rate": 7.963886216813836e-06, "loss": 1.2357, "step": 2992 }, { "epoch": 0.32, "learning_rate": 7.962494786584232e-06, "loss": 1.4969, "step": 2993 }, { "epoch": 0.32, "learning_rate": 7.961103002729634e-06, "loss": 1.6702, "step": 2994 }, { "epoch": 0.32, "learning_rate": 7.959710865416173e-06, "loss": 1.3258, "step": 2995 }, { "epoch": 0.32, "learning_rate": 7.958318374810028e-06, "loss": 1.6911, "step": 2996 }, { "epoch": 0.32, "learning_rate": 7.956925531077417e-06, "loss": 1.4459, "step": 2997 }, { "epoch": 0.32, "learning_rate": 7.955532334384597e-06, "loss": 1.6115, "step": 2998 }, { "epoch": 0.32, "learning_rate": 7.954138784897873e-06, "loss": 1.3947, "step": 2999 }, { "epoch": 0.32, "learning_rate": 7.952744882783587e-06, "loss": 1.406, "step": 3000 }, { "epoch": 0.32, "learning_rate": 7.951350628208126e-06, "loss": 1.5857, "step": 3001 }, { "epoch": 0.32, "learning_rate": 7.949956021337918e-06, "loss": 1.4756, "step": 3002 }, { "epoch": 0.32, "learning_rate": 7.948561062339435e-06, "loss": 1.4674, "step": 3003 }, { "epoch": 0.32, "learning_rate": 7.947165751379189e-06, "loss": 1.4123, "step": 3004 }, { "epoch": 0.32, "learning_rate": 7.945770088623735e-06, "loss": 1.487, "step": 3005 }, { "epoch": 0.32, "learning_rate": 7.944374074239665e-06, "loss": 1.3721, "step": 3006 }, { "epoch": 0.32, "learning_rate": 7.942977708393624e-06, "loss": 1.21, "step": 3007 }, { "epoch": 0.32, "learning_rate": 7.941580991252288e-06, "loss": 1.3135, "step": 3008 }, { "epoch": 0.32, "learning_rate": 7.940183922982381e-06, "loss": 1.4325, "step": 3009 }, { "epoch": 0.32, "learning_rate": 7.938786503750668e-06, "loss": 1.5571, "step": 3010 }, { "epoch": 0.32, "learning_rate": 7.937388733723952e-06, "loss": 1.3764, "step": 3011 }, { "epoch": 0.32, "learning_rate": 7.935990613069087e-06, "loss": 1.4421, "step": 3012 }, { "epoch": 0.32, "learning_rate": 7.934592141952954e-06, "loss": 1.5017, "step": 3013 }, { "epoch": 0.32, "learning_rate": 7.933193320542493e-06, "loss": 1.3424, "step": 3014 }, { "epoch": 0.32, "learning_rate": 7.931794149004675e-06, "loss": 1.5122, "step": 3015 }, { "epoch": 0.32, "learning_rate": 7.930394627506513e-06, "loss": 1.3597, "step": 3016 }, { "epoch": 0.32, "learning_rate": 7.928994756215067e-06, "loss": 1.6185, "step": 3017 }, { "epoch": 0.32, "learning_rate": 7.927594535297433e-06, "loss": 1.6006, "step": 3018 }, { "epoch": 0.32, "learning_rate": 7.926193964920756e-06, "loss": 1.6357, "step": 3019 }, { "epoch": 0.32, "learning_rate": 7.924793045252213e-06, "loss": 1.8983, "step": 3020 }, { "epoch": 0.32, "learning_rate": 7.923391776459031e-06, "loss": 1.5314, "step": 3021 }, { "epoch": 0.32, "learning_rate": 7.921990158708478e-06, "loss": 1.2658, "step": 3022 }, { "epoch": 0.32, "learning_rate": 7.920588192167854e-06, "loss": 1.5588, "step": 3023 }, { "epoch": 0.32, "learning_rate": 7.919185877004515e-06, "loss": 1.4902, "step": 3024 }, { "epoch": 0.32, "learning_rate": 7.917783213385849e-06, "loss": 1.4346, "step": 3025 }, { "epoch": 0.32, "learning_rate": 7.916380201479287e-06, "loss": 1.497, "step": 3026 }, { "epoch": 0.32, "learning_rate": 7.914976841452304e-06, "loss": 1.4156, "step": 3027 }, { "epoch": 0.32, "learning_rate": 7.913573133472417e-06, "loss": 1.4811, "step": 3028 }, { "epoch": 0.32, "learning_rate": 7.912169077707179e-06, "loss": 1.0605, "step": 3029 }, { "epoch": 0.32, "learning_rate": 7.91076467432419e-06, "loss": 1.677, "step": 3030 }, { "epoch": 0.32, "learning_rate": 7.909359923491092e-06, "loss": 1.3713, "step": 3031 }, { "epoch": 0.32, "learning_rate": 7.907954825375564e-06, "loss": 1.7046, "step": 3032 }, { "epoch": 0.32, "learning_rate": 7.90654938014533e-06, "loss": 1.5698, "step": 3033 }, { "epoch": 0.32, "learning_rate": 7.905143587968153e-06, "loss": 1.4171, "step": 3034 }, { "epoch": 0.32, "learning_rate": 7.903737449011838e-06, "loss": 1.3423, "step": 3035 }, { "epoch": 0.32, "learning_rate": 7.902330963444234e-06, "loss": 1.5506, "step": 3036 }, { "epoch": 0.32, "learning_rate": 7.90092413143323e-06, "loss": 1.7316, "step": 3037 }, { "epoch": 0.32, "learning_rate": 7.899516953146751e-06, "loss": 1.3989, "step": 3038 }, { "epoch": 0.32, "learning_rate": 7.898109428752773e-06, "loss": 1.3398, "step": 3039 }, { "epoch": 0.32, "learning_rate": 7.896701558419306e-06, "loss": 1.3428, "step": 3040 }, { "epoch": 0.32, "learning_rate": 7.895293342314406e-06, "loss": 1.3617, "step": 3041 }, { "epoch": 0.32, "learning_rate": 7.893884780606164e-06, "loss": 1.4531, "step": 3042 }, { "epoch": 0.32, "learning_rate": 7.89247587346272e-06, "loss": 1.5536, "step": 3043 }, { "epoch": 0.32, "learning_rate": 7.891066621052251e-06, "loss": 1.4562, "step": 3044 }, { "epoch": 0.32, "learning_rate": 7.889657023542973e-06, "loss": 1.2443, "step": 3045 }, { "epoch": 0.32, "learning_rate": 7.88824708110315e-06, "loss": 1.3626, "step": 3046 }, { "epoch": 0.33, "learning_rate": 7.886836793901077e-06, "loss": 1.1504, "step": 3047 }, { "epoch": 0.33, "learning_rate": 7.885426162105101e-06, "loss": 1.2852, "step": 3048 }, { "epoch": 0.33, "learning_rate": 7.884015185883602e-06, "loss": 1.32, "step": 3049 }, { "epoch": 0.33, "learning_rate": 7.882603865405008e-06, "loss": 1.376, "step": 3050 }, { "epoch": 0.33, "learning_rate": 7.881192200837785e-06, "loss": 1.5679, "step": 3051 }, { "epoch": 0.33, "learning_rate": 7.879780192350434e-06, "loss": 1.5923, "step": 3052 }, { "epoch": 0.33, "learning_rate": 7.878367840111508e-06, "loss": 1.3742, "step": 3053 }, { "epoch": 0.33, "learning_rate": 7.876955144289594e-06, "loss": 1.5398, "step": 3054 }, { "epoch": 0.33, "learning_rate": 7.87554210505332e-06, "loss": 1.3651, "step": 3055 }, { "epoch": 0.33, "learning_rate": 7.874128722571359e-06, "loss": 1.4194, "step": 3056 }, { "epoch": 0.33, "learning_rate": 7.872714997012421e-06, "loss": 1.4559, "step": 3057 }, { "epoch": 0.33, "learning_rate": 7.87130092854526e-06, "loss": 1.2533, "step": 3058 }, { "epoch": 0.33, "learning_rate": 7.86988651733867e-06, "loss": 1.3553, "step": 3059 }, { "epoch": 0.33, "learning_rate": 7.868471763561482e-06, "loss": 1.3652, "step": 3060 }, { "epoch": 0.33, "learning_rate": 7.867056667382576e-06, "loss": 1.3577, "step": 3061 }, { "epoch": 0.33, "learning_rate": 7.865641228970865e-06, "loss": 1.4439, "step": 3062 }, { "epoch": 0.33, "learning_rate": 7.864225448495304e-06, "loss": 1.4915, "step": 3063 }, { "epoch": 0.33, "learning_rate": 7.862809326124896e-06, "loss": 1.4513, "step": 3064 }, { "epoch": 0.33, "learning_rate": 7.861392862028678e-06, "loss": 1.5291, "step": 3065 }, { "epoch": 0.33, "learning_rate": 7.85997605637573e-06, "loss": 1.7202, "step": 3066 }, { "epoch": 0.33, "learning_rate": 7.85855890933517e-06, "loss": 1.6165, "step": 3067 }, { "epoch": 0.33, "learning_rate": 7.85714142107616e-06, "loss": 1.5292, "step": 3068 }, { "epoch": 0.33, "learning_rate": 7.855723591767903e-06, "loss": 1.6404, "step": 3069 }, { "epoch": 0.33, "learning_rate": 7.85430542157964e-06, "loss": 1.4796, "step": 3070 }, { "epoch": 0.33, "learning_rate": 7.852886910680654e-06, "loss": 1.3566, "step": 3071 }, { "epoch": 0.33, "learning_rate": 7.85146805924027e-06, "loss": 1.5151, "step": 3072 }, { "epoch": 0.33, "learning_rate": 7.850048867427851e-06, "loss": 1.3719, "step": 3073 }, { "epoch": 0.33, "learning_rate": 7.848629335412803e-06, "loss": 1.5804, "step": 3074 }, { "epoch": 0.33, "learning_rate": 7.847209463364574e-06, "loss": 1.5993, "step": 3075 }, { "epoch": 0.33, "learning_rate": 7.845789251452646e-06, "loss": 1.3709, "step": 3076 }, { "epoch": 0.33, "learning_rate": 7.844368699846547e-06, "loss": 1.451, "step": 3077 }, { "epoch": 0.33, "learning_rate": 7.842947808715848e-06, "loss": 1.3588, "step": 3078 }, { "epoch": 0.33, "learning_rate": 7.84152657823015e-06, "loss": 1.4445, "step": 3079 }, { "epoch": 0.33, "learning_rate": 7.840105008559109e-06, "loss": 1.6225, "step": 3080 }, { "epoch": 0.33, "learning_rate": 7.83868309987241e-06, "loss": 1.2619, "step": 3081 }, { "epoch": 0.33, "learning_rate": 7.837260852339782e-06, "loss": 1.614, "step": 3082 }, { "epoch": 0.33, "learning_rate": 7.835838266130997e-06, "loss": 1.4935, "step": 3083 }, { "epoch": 0.33, "learning_rate": 7.834415341415862e-06, "loss": 1.5361, "step": 3084 }, { "epoch": 0.33, "learning_rate": 7.832992078364231e-06, "loss": 1.931, "step": 3085 }, { "epoch": 0.33, "learning_rate": 7.831568477145994e-06, "loss": 1.379, "step": 3086 }, { "epoch": 0.33, "learning_rate": 7.830144537931082e-06, "loss": 1.3089, "step": 3087 }, { "epoch": 0.33, "learning_rate": 7.828720260889468e-06, "loss": 1.5463, "step": 3088 }, { "epoch": 0.33, "learning_rate": 7.827295646191161e-06, "loss": 1.403, "step": 3089 }, { "epoch": 0.33, "learning_rate": 7.825870694006217e-06, "loss": 1.4932, "step": 3090 }, { "epoch": 0.33, "learning_rate": 7.824445404504727e-06, "loss": 1.3697, "step": 3091 }, { "epoch": 0.33, "learning_rate": 7.823019777856823e-06, "loss": 1.7315, "step": 3092 }, { "epoch": 0.33, "learning_rate": 7.82159381423268e-06, "loss": 1.5751, "step": 3093 }, { "epoch": 0.33, "learning_rate": 7.82016751380251e-06, "loss": 1.583, "step": 3094 }, { "epoch": 0.33, "learning_rate": 7.818740876736568e-06, "loss": 1.5523, "step": 3095 }, { "epoch": 0.33, "learning_rate": 7.817313903205148e-06, "loss": 1.4309, "step": 3096 }, { "epoch": 0.33, "learning_rate": 7.815886593378583e-06, "loss": 1.6806, "step": 3097 }, { "epoch": 0.33, "learning_rate": 7.814458947427246e-06, "loss": 1.4336, "step": 3098 }, { "epoch": 0.33, "learning_rate": 7.813030965521554e-06, "loss": 1.4196, "step": 3099 }, { "epoch": 0.33, "learning_rate": 7.811602647831959e-06, "loss": 1.1679, "step": 3100 }, { "epoch": 0.33, "learning_rate": 7.810173994528956e-06, "loss": 1.4321, "step": 3101 }, { "epoch": 0.33, "learning_rate": 7.80874500578308e-06, "loss": 1.5438, "step": 3102 }, { "epoch": 0.33, "learning_rate": 7.807315681764907e-06, "loss": 1.3775, "step": 3103 }, { "epoch": 0.33, "learning_rate": 7.805886022645046e-06, "loss": 1.552, "step": 3104 }, { "epoch": 0.33, "learning_rate": 7.804456028594158e-06, "loss": 1.3036, "step": 3105 }, { "epoch": 0.33, "learning_rate": 7.803025699782935e-06, "loss": 1.5939, "step": 3106 }, { "epoch": 0.33, "learning_rate": 7.80159503638211e-06, "loss": 1.4319, "step": 3107 }, { "epoch": 0.33, "learning_rate": 7.80016403856246e-06, "loss": 1.4687, "step": 3108 }, { "epoch": 0.33, "learning_rate": 7.798732706494799e-06, "loss": 1.7107, "step": 3109 }, { "epoch": 0.33, "learning_rate": 7.797301040349978e-06, "loss": 1.1221, "step": 3110 }, { "epoch": 0.33, "learning_rate": 7.795869040298895e-06, "loss": 1.445, "step": 3111 }, { "epoch": 0.33, "learning_rate": 7.794436706512483e-06, "loss": 1.4103, "step": 3112 }, { "epoch": 0.33, "learning_rate": 7.793004039161714e-06, "loss": 1.1987, "step": 3113 }, { "epoch": 0.33, "learning_rate": 7.791571038417602e-06, "loss": 1.4081, "step": 3114 }, { "epoch": 0.33, "learning_rate": 7.790137704451203e-06, "loss": 1.5182, "step": 3115 }, { "epoch": 0.33, "learning_rate": 7.788704037433608e-06, "loss": 1.2784, "step": 3116 }, { "epoch": 0.33, "learning_rate": 7.78727003753595e-06, "loss": 1.6821, "step": 3117 }, { "epoch": 0.33, "learning_rate": 7.785835704929403e-06, "loss": 1.552, "step": 3118 }, { "epoch": 0.33, "learning_rate": 7.784401039785179e-06, "loss": 1.4231, "step": 3119 }, { "epoch": 0.33, "learning_rate": 7.782966042274529e-06, "loss": 1.746, "step": 3120 }, { "epoch": 0.33, "learning_rate": 7.781530712568746e-06, "loss": 1.4907, "step": 3121 }, { "epoch": 0.33, "learning_rate": 7.78009505083916e-06, "loss": 1.3925, "step": 3122 }, { "epoch": 0.33, "learning_rate": 7.778659057257144e-06, "loss": 1.569, "step": 3123 }, { "epoch": 0.33, "learning_rate": 7.777222731994107e-06, "loss": 1.4825, "step": 3124 }, { "epoch": 0.33, "learning_rate": 7.7757860752215e-06, "loss": 1.4482, "step": 3125 }, { "epoch": 0.33, "learning_rate": 7.774349087110813e-06, "loss": 1.2114, "step": 3126 }, { "epoch": 0.33, "learning_rate": 7.772911767833576e-06, "loss": 1.3529, "step": 3127 }, { "epoch": 0.33, "learning_rate": 7.771474117561353e-06, "loss": 1.5582, "step": 3128 }, { "epoch": 0.33, "learning_rate": 7.77003613646576e-06, "loss": 1.5854, "step": 3129 }, { "epoch": 0.33, "learning_rate": 7.76859782471844e-06, "loss": 1.4174, "step": 3130 }, { "epoch": 0.33, "learning_rate": 7.767159182491084e-06, "loss": 1.4514, "step": 3131 }, { "epoch": 0.33, "learning_rate": 7.765720209955414e-06, "loss": 1.5784, "step": 3132 }, { "epoch": 0.33, "learning_rate": 7.764280907283198e-06, "loss": 1.7642, "step": 3133 }, { "epoch": 0.33, "learning_rate": 7.762841274646247e-06, "loss": 1.2109, "step": 3134 }, { "epoch": 0.33, "learning_rate": 7.761401312216398e-06, "loss": 1.5999, "step": 3135 }, { "epoch": 0.33, "learning_rate": 7.75996102016554e-06, "loss": 1.6018, "step": 3136 }, { "epoch": 0.33, "learning_rate": 7.758520398665596e-06, "loss": 1.7179, "step": 3137 }, { "epoch": 0.33, "learning_rate": 7.757079447888529e-06, "loss": 1.3579, "step": 3138 }, { "epoch": 0.33, "learning_rate": 7.75563816800634e-06, "loss": 1.4721, "step": 3139 }, { "epoch": 0.33, "learning_rate": 7.754196559191072e-06, "loss": 1.3162, "step": 3140 }, { "epoch": 0.34, "learning_rate": 7.752754621614807e-06, "loss": 1.6429, "step": 3141 }, { "epoch": 0.34, "learning_rate": 7.751312355449664e-06, "loss": 1.4233, "step": 3142 }, { "epoch": 0.34, "learning_rate": 7.749869760867803e-06, "loss": 1.439, "step": 3143 }, { "epoch": 0.34, "learning_rate": 7.748426838041421e-06, "loss": 1.6989, "step": 3144 }, { "epoch": 0.34, "learning_rate": 7.746983587142757e-06, "loss": 1.5361, "step": 3145 }, { "epoch": 0.34, "learning_rate": 7.745540008344088e-06, "loss": 1.4704, "step": 3146 }, { "epoch": 0.34, "learning_rate": 7.744096101817731e-06, "loss": 1.6864, "step": 3147 }, { "epoch": 0.34, "learning_rate": 7.742651867736037e-06, "loss": 1.5412, "step": 3148 }, { "epoch": 0.34, "learning_rate": 7.741207306271409e-06, "loss": 1.4139, "step": 3149 }, { "epoch": 0.34, "learning_rate": 7.73976241759627e-06, "loss": 1.9307, "step": 3150 }, { "epoch": 0.34, "learning_rate": 7.738317201883098e-06, "loss": 1.6023, "step": 3151 }, { "epoch": 0.34, "learning_rate": 7.736871659304404e-06, "loss": 1.3299, "step": 3152 }, { "epoch": 0.34, "learning_rate": 7.73542579003274e-06, "loss": 1.5052, "step": 3153 }, { "epoch": 0.34, "learning_rate": 7.733979594240694e-06, "loss": 1.2454, "step": 3154 }, { "epoch": 0.34, "learning_rate": 7.732533072100892e-06, "loss": 1.2276, "step": 3155 }, { "epoch": 0.34, "learning_rate": 7.731086223786006e-06, "loss": 1.2264, "step": 3156 }, { "epoch": 0.34, "learning_rate": 7.72963904946874e-06, "loss": 1.3346, "step": 3157 }, { "epoch": 0.34, "learning_rate": 7.728191549321837e-06, "loss": 1.1761, "step": 3158 }, { "epoch": 0.34, "learning_rate": 7.726743723518087e-06, "loss": 1.3265, "step": 3159 }, { "epoch": 0.34, "learning_rate": 7.725295572230307e-06, "loss": 1.6672, "step": 3160 }, { "epoch": 0.34, "learning_rate": 7.723847095631362e-06, "loss": 1.7443, "step": 3161 }, { "epoch": 0.34, "learning_rate": 7.722398293894153e-06, "loss": 1.5189, "step": 3162 }, { "epoch": 0.34, "learning_rate": 7.72094916719162e-06, "loss": 1.6434, "step": 3163 }, { "epoch": 0.34, "learning_rate": 7.719499715696738e-06, "loss": 1.4765, "step": 3164 }, { "epoch": 0.34, "learning_rate": 7.718049939582529e-06, "loss": 1.2686, "step": 3165 }, { "epoch": 0.34, "learning_rate": 7.716599839022044e-06, "loss": 1.5666, "step": 3166 }, { "epoch": 0.34, "learning_rate": 7.71514941418838e-06, "loss": 1.6217, "step": 3167 }, { "epoch": 0.34, "learning_rate": 7.713698665254669e-06, "loss": 1.6862, "step": 3168 }, { "epoch": 0.34, "learning_rate": 7.712247592394087e-06, "loss": 1.5, "step": 3169 }, { "epoch": 0.34, "learning_rate": 7.71079619577984e-06, "loss": 1.4302, "step": 3170 }, { "epoch": 0.34, "learning_rate": 7.70934447558518e-06, "loss": 1.6232, "step": 3171 }, { "epoch": 0.34, "learning_rate": 7.70789243198339e-06, "loss": 1.4531, "step": 3172 }, { "epoch": 0.34, "learning_rate": 7.706440065147805e-06, "loss": 1.4876, "step": 3173 }, { "epoch": 0.34, "learning_rate": 7.704987375251782e-06, "loss": 1.3623, "step": 3174 }, { "epoch": 0.34, "learning_rate": 7.70353436246873e-06, "loss": 1.3316, "step": 3175 }, { "epoch": 0.34, "learning_rate": 7.70208102697209e-06, "loss": 1.443, "step": 3176 }, { "epoch": 0.34, "learning_rate": 7.70062736893534e-06, "loss": 1.5088, "step": 3177 }, { "epoch": 0.34, "learning_rate": 7.699173388532004e-06, "loss": 1.433, "step": 3178 }, { "epoch": 0.34, "learning_rate": 7.697719085935634e-06, "loss": 1.573, "step": 3179 }, { "epoch": 0.34, "learning_rate": 7.696264461319831e-06, "loss": 1.2611, "step": 3180 }, { "epoch": 0.34, "learning_rate": 7.694809514858227e-06, "loss": 1.4652, "step": 3181 }, { "epoch": 0.34, "learning_rate": 7.693354246724496e-06, "loss": 1.5343, "step": 3182 }, { "epoch": 0.34, "learning_rate": 7.69189865709235e-06, "loss": 1.329, "step": 3183 }, { "epoch": 0.34, "learning_rate": 7.690442746135538e-06, "loss": 1.2978, "step": 3184 }, { "epoch": 0.34, "learning_rate": 7.688986514027848e-06, "loss": 1.2323, "step": 3185 }, { "epoch": 0.34, "learning_rate": 7.687529960943107e-06, "loss": 1.4498, "step": 3186 }, { "epoch": 0.34, "learning_rate": 7.686073087055179e-06, "loss": 1.6062, "step": 3187 }, { "epoch": 0.34, "learning_rate": 7.684615892537968e-06, "loss": 1.5483, "step": 3188 }, { "epoch": 0.34, "learning_rate": 7.683158377565415e-06, "loss": 1.5366, "step": 3189 }, { "epoch": 0.34, "learning_rate": 7.681700542311499e-06, "loss": 1.5312, "step": 3190 }, { "epoch": 0.34, "learning_rate": 7.680242386950239e-06, "loss": 1.5615, "step": 3191 }, { "epoch": 0.34, "learning_rate": 7.678783911655691e-06, "loss": 1.4465, "step": 3192 }, { "epoch": 0.34, "learning_rate": 7.677325116601948e-06, "loss": 1.4115, "step": 3193 }, { "epoch": 0.34, "learning_rate": 7.675866001963144e-06, "loss": 1.5066, "step": 3194 }, { "epoch": 0.34, "learning_rate": 7.674406567913447e-06, "loss": 1.526, "step": 3195 }, { "epoch": 0.34, "learning_rate": 7.672946814627069e-06, "loss": 1.3401, "step": 3196 }, { "epoch": 0.34, "learning_rate": 7.671486742278254e-06, "loss": 1.6119, "step": 3197 }, { "epoch": 0.34, "learning_rate": 7.67002635104129e-06, "loss": 1.4755, "step": 3198 }, { "epoch": 0.34, "learning_rate": 7.668565641090495e-06, "loss": 1.3421, "step": 3199 }, { "epoch": 0.34, "learning_rate": 7.667104612600236e-06, "loss": 1.7398, "step": 3200 }, { "epoch": 0.34, "learning_rate": 7.66564326574491e-06, "loss": 1.6282, "step": 3201 }, { "epoch": 0.34, "learning_rate": 7.66418160069895e-06, "loss": 1.5878, "step": 3202 }, { "epoch": 0.34, "learning_rate": 7.662719617636836e-06, "loss": 1.509, "step": 3203 }, { "epoch": 0.34, "learning_rate": 7.661257316733078e-06, "loss": 1.4885, "step": 3204 }, { "epoch": 0.34, "learning_rate": 7.659794698162229e-06, "loss": 1.3122, "step": 3205 }, { "epoch": 0.34, "learning_rate": 7.658331762098876e-06, "loss": 1.5315, "step": 3206 }, { "epoch": 0.34, "learning_rate": 7.656868508717648e-06, "loss": 1.4654, "step": 3207 }, { "epoch": 0.34, "learning_rate": 7.655404938193207e-06, "loss": 1.4327, "step": 3208 }, { "epoch": 0.34, "learning_rate": 7.653941050700257e-06, "loss": 1.4893, "step": 3209 }, { "epoch": 0.34, "learning_rate": 7.652476846413537e-06, "loss": 1.453, "step": 3210 }, { "epoch": 0.34, "learning_rate": 7.651012325507827e-06, "loss": 1.3376, "step": 3211 }, { "epoch": 0.34, "learning_rate": 7.649547488157943e-06, "loss": 1.688, "step": 3212 }, { "epoch": 0.34, "learning_rate": 7.648082334538735e-06, "loss": 1.4583, "step": 3213 }, { "epoch": 0.34, "learning_rate": 7.646616864825099e-06, "loss": 1.5024, "step": 3214 }, { "epoch": 0.34, "learning_rate": 7.645151079191962e-06, "loss": 1.5354, "step": 3215 }, { "epoch": 0.34, "learning_rate": 7.64368497781429e-06, "loss": 1.5507, "step": 3216 }, { "epoch": 0.34, "learning_rate": 7.64221856086709e-06, "loss": 1.6281, "step": 3217 }, { "epoch": 0.34, "learning_rate": 7.640751828525402e-06, "loss": 1.5449, "step": 3218 }, { "epoch": 0.34, "learning_rate": 7.639284780964307e-06, "loss": 1.4807, "step": 3219 }, { "epoch": 0.34, "learning_rate": 7.637817418358922e-06, "loss": 1.4963, "step": 3220 }, { "epoch": 0.34, "learning_rate": 7.636349740884402e-06, "loss": 1.3868, "step": 3221 }, { "epoch": 0.34, "learning_rate": 7.634881748715941e-06, "loss": 1.4677, "step": 3222 }, { "epoch": 0.34, "learning_rate": 7.633413442028768e-06, "loss": 1.4864, "step": 3223 }, { "epoch": 0.34, "learning_rate": 7.63194482099815e-06, "loss": 1.7466, "step": 3224 }, { "epoch": 0.34, "learning_rate": 7.630475885799395e-06, "loss": 1.2701, "step": 3225 }, { "epoch": 0.34, "learning_rate": 7.629006636607843e-06, "loss": 1.3012, "step": 3226 }, { "epoch": 0.34, "learning_rate": 7.627537073598876e-06, "loss": 1.5114, "step": 3227 }, { "epoch": 0.34, "learning_rate": 7.626067196947913e-06, "loss": 1.5968, "step": 3228 }, { "epoch": 0.34, "learning_rate": 7.624597006830405e-06, "loss": 1.4821, "step": 3229 }, { "epoch": 0.34, "learning_rate": 7.623126503421848e-06, "loss": 1.3297, "step": 3230 }, { "epoch": 0.34, "learning_rate": 7.621655686897771e-06, "loss": 1.448, "step": 3231 }, { "epoch": 0.34, "learning_rate": 7.620184557433742e-06, "loss": 1.4455, "step": 3232 }, { "epoch": 0.34, "learning_rate": 7.618713115205364e-06, "loss": 1.6143, "step": 3233 }, { "epoch": 0.34, "learning_rate": 7.617241360388282e-06, "loss": 1.4397, "step": 3234 }, { "epoch": 0.35, "learning_rate": 7.615769293158173e-06, "loss": 1.4853, "step": 3235 }, { "epoch": 0.35, "learning_rate": 7.614296913690756e-06, "loss": 1.5189, "step": 3236 }, { "epoch": 0.35, "learning_rate": 7.612824222161781e-06, "loss": 1.3793, "step": 3237 }, { "epoch": 0.35, "learning_rate": 7.61135121874704e-06, "loss": 1.4405, "step": 3238 }, { "epoch": 0.35, "learning_rate": 7.609877903622368e-06, "loss": 1.5477, "step": 3239 }, { "epoch": 0.35, "learning_rate": 7.608404276963623e-06, "loss": 1.4607, "step": 3240 }, { "epoch": 0.35, "learning_rate": 7.60693033894671e-06, "loss": 1.4577, "step": 3241 }, { "epoch": 0.35, "learning_rate": 7.6054560897475715e-06, "loss": 1.4493, "step": 3242 }, { "epoch": 0.35, "learning_rate": 7.60398152954218e-06, "loss": 1.4219, "step": 3243 }, { "epoch": 0.35, "learning_rate": 7.602506658506554e-06, "loss": 1.5473, "step": 3244 }, { "epoch": 0.35, "learning_rate": 7.601031476816742e-06, "loss": 1.4808, "step": 3245 }, { "epoch": 0.35, "learning_rate": 7.599555984648836e-06, "loss": 1.6078, "step": 3246 }, { "epoch": 0.35, "learning_rate": 7.598080182178957e-06, "loss": 1.3433, "step": 3247 }, { "epoch": 0.35, "learning_rate": 7.5966040695832725e-06, "loss": 1.8063, "step": 3248 }, { "epoch": 0.35, "learning_rate": 7.595127647037976e-06, "loss": 1.6228, "step": 3249 }, { "epoch": 0.35, "learning_rate": 7.593650914719311e-06, "loss": 1.5225, "step": 3250 }, { "epoch": 0.35, "learning_rate": 7.5921738728035454e-06, "loss": 1.5049, "step": 3251 }, { "epoch": 0.35, "learning_rate": 7.590696521466992e-06, "loss": 1.4472, "step": 3252 }, { "epoch": 0.35, "learning_rate": 7.5892188608859985e-06, "loss": 1.3415, "step": 3253 }, { "epoch": 0.35, "learning_rate": 7.587740891236948e-06, "loss": 1.2874, "step": 3254 }, { "epoch": 0.35, "learning_rate": 7.586262612696263e-06, "loss": 1.5921, "step": 3255 }, { "epoch": 0.35, "learning_rate": 7.5847840254404e-06, "loss": 1.566, "step": 3256 }, { "epoch": 0.35, "learning_rate": 7.583305129645857e-06, "loss": 1.4654, "step": 3257 }, { "epoch": 0.35, "learning_rate": 7.5818259254891614e-06, "loss": 1.7092, "step": 3258 }, { "epoch": 0.35, "learning_rate": 7.580346413146886e-06, "loss": 1.4262, "step": 3259 }, { "epoch": 0.35, "learning_rate": 7.578866592795635e-06, "loss": 1.451, "step": 3260 }, { "epoch": 0.35, "learning_rate": 7.577386464612049e-06, "loss": 1.5209, "step": 3261 }, { "epoch": 0.35, "learning_rate": 7.575906028772808e-06, "loss": 1.5469, "step": 3262 }, { "epoch": 0.35, "learning_rate": 7.574425285454627e-06, "loss": 1.5056, "step": 3263 }, { "epoch": 0.35, "learning_rate": 7.572944234834261e-06, "loss": 1.4535, "step": 3264 }, { "epoch": 0.35, "learning_rate": 7.571462877088496e-06, "loss": 1.5525, "step": 3265 }, { "epoch": 0.35, "learning_rate": 7.569981212394157e-06, "loss": 1.4649, "step": 3266 }, { "epoch": 0.35, "learning_rate": 7.568499240928109e-06, "loss": 1.3387, "step": 3267 }, { "epoch": 0.35, "learning_rate": 7.56701696286725e-06, "loss": 1.5482, "step": 3268 }, { "epoch": 0.35, "learning_rate": 7.565534378388515e-06, "loss": 1.3899, "step": 3269 }, { "epoch": 0.35, "learning_rate": 7.5640514876688765e-06, "loss": 1.29, "step": 3270 }, { "epoch": 0.35, "learning_rate": 7.562568290885344e-06, "loss": 1.5793, "step": 3271 }, { "epoch": 0.35, "learning_rate": 7.561084788214958e-06, "loss": 1.4961, "step": 3272 }, { "epoch": 0.35, "learning_rate": 7.559600979834809e-06, "loss": 1.2953, "step": 3273 }, { "epoch": 0.35, "learning_rate": 7.558116865922007e-06, "loss": 1.3607, "step": 3274 }, { "epoch": 0.35, "learning_rate": 7.556632446653711e-06, "loss": 1.5303, "step": 3275 }, { "epoch": 0.35, "learning_rate": 7.555147722207111e-06, "loss": 1.5974, "step": 3276 }, { "epoch": 0.35, "learning_rate": 7.553662692759434e-06, "loss": 1.3604, "step": 3277 }, { "epoch": 0.35, "learning_rate": 7.552177358487944e-06, "loss": 1.227, "step": 3278 }, { "epoch": 0.35, "learning_rate": 7.550691719569944e-06, "loss": 1.6309, "step": 3279 }, { "epoch": 0.35, "learning_rate": 7.549205776182766e-06, "loss": 1.4643, "step": 3280 }, { "epoch": 0.35, "learning_rate": 7.5477195285037875e-06, "loss": 1.528, "step": 3281 }, { "epoch": 0.35, "learning_rate": 7.546232976710413e-06, "loss": 1.4813, "step": 3282 }, { "epoch": 0.35, "learning_rate": 7.544746120980093e-06, "loss": 1.2501, "step": 3283 }, { "epoch": 0.35, "learning_rate": 7.543258961490309e-06, "loss": 1.3952, "step": 3284 }, { "epoch": 0.35, "learning_rate": 7.541771498418575e-06, "loss": 1.3418, "step": 3285 }, { "epoch": 0.35, "learning_rate": 7.5402837319424495e-06, "loss": 1.1307, "step": 3286 }, { "epoch": 0.35, "learning_rate": 7.538795662239522e-06, "loss": 1.4171, "step": 3287 }, { "epoch": 0.35, "learning_rate": 7.537307289487419e-06, "loss": 1.2349, "step": 3288 }, { "epoch": 0.35, "learning_rate": 7.535818613863804e-06, "loss": 1.6823, "step": 3289 }, { "epoch": 0.35, "learning_rate": 7.534329635546374e-06, "loss": 1.3818, "step": 3290 }, { "epoch": 0.35, "learning_rate": 7.532840354712868e-06, "loss": 1.6681, "step": 3291 }, { "epoch": 0.35, "learning_rate": 7.531350771541055e-06, "loss": 1.5467, "step": 3292 }, { "epoch": 0.35, "learning_rate": 7.529860886208744e-06, "loss": 1.6225, "step": 3293 }, { "epoch": 0.35, "learning_rate": 7.5283706988937765e-06, "loss": 1.7217, "step": 3294 }, { "epoch": 0.35, "learning_rate": 7.526880209774034e-06, "loss": 1.4529, "step": 3295 }, { "epoch": 0.35, "learning_rate": 7.5253894190274335e-06, "loss": 1.4002, "step": 3296 }, { "epoch": 0.35, "learning_rate": 7.523898326831921e-06, "loss": 1.4832, "step": 3297 }, { "epoch": 0.35, "learning_rate": 7.522406933365491e-06, "loss": 1.4447, "step": 3298 }, { "epoch": 0.35, "learning_rate": 7.520915238806161e-06, "loss": 1.5641, "step": 3299 }, { "epoch": 0.35, "learning_rate": 7.5194232433319955e-06, "loss": 1.4316, "step": 3300 }, { "epoch": 0.35, "learning_rate": 7.517930947121088e-06, "loss": 1.5473, "step": 3301 }, { "epoch": 0.35, "learning_rate": 7.51643835035157e-06, "loss": 1.6385, "step": 3302 }, { "epoch": 0.35, "learning_rate": 7.514945453201608e-06, "loss": 1.5383, "step": 3303 }, { "epoch": 0.35, "learning_rate": 7.513452255849405e-06, "loss": 1.7155, "step": 3304 }, { "epoch": 0.35, "learning_rate": 7.511958758473201e-06, "loss": 1.5924, "step": 3305 }, { "epoch": 0.35, "learning_rate": 7.510464961251271e-06, "loss": 1.5507, "step": 3306 }, { "epoch": 0.35, "learning_rate": 7.508970864361924e-06, "loss": 1.4266, "step": 3307 }, { "epoch": 0.35, "learning_rate": 7.507476467983508e-06, "loss": 1.4132, "step": 3308 }, { "epoch": 0.35, "learning_rate": 7.505981772294404e-06, "loss": 1.3391, "step": 3309 }, { "epoch": 0.35, "learning_rate": 7.50448677747303e-06, "loss": 1.7161, "step": 3310 }, { "epoch": 0.35, "learning_rate": 7.502991483697839e-06, "loss": 1.4529, "step": 3311 }, { "epoch": 0.35, "learning_rate": 7.501495891147322e-06, "loss": 1.7114, "step": 3312 }, { "epoch": 0.35, "learning_rate": 7.500000000000001e-06, "loss": 1.178, "step": 3313 }, { "epoch": 0.35, "learning_rate": 7.498503810434438e-06, "loss": 1.1925, "step": 3314 }, { "epoch": 0.35, "learning_rate": 7.497007322629231e-06, "loss": 1.928, "step": 3315 }, { "epoch": 0.35, "learning_rate": 7.495510536763008e-06, "loss": 1.4229, "step": 3316 }, { "epoch": 0.35, "learning_rate": 7.494013453014438e-06, "loss": 1.5223, "step": 3317 }, { "epoch": 0.35, "learning_rate": 7.492516071562226e-06, "loss": 1.5042, "step": 3318 }, { "epoch": 0.35, "learning_rate": 7.4910183925851055e-06, "loss": 1.5927, "step": 3319 }, { "epoch": 0.35, "learning_rate": 7.489520416261855e-06, "loss": 1.4837, "step": 3320 }, { "epoch": 0.35, "learning_rate": 7.488022142771282e-06, "loss": 1.232, "step": 3321 }, { "epoch": 0.35, "learning_rate": 7.486523572292231e-06, "loss": 1.639, "step": 3322 }, { "epoch": 0.35, "learning_rate": 7.485024705003583e-06, "loss": 1.2201, "step": 3323 }, { "epoch": 0.35, "learning_rate": 7.483525541084253e-06, "loss": 1.5319, "step": 3324 }, { "epoch": 0.35, "learning_rate": 7.4820260807131935e-06, "loss": 1.4265, "step": 3325 }, { "epoch": 0.35, "learning_rate": 7.480526324069391e-06, "loss": 1.4094, "step": 3326 }, { "epoch": 0.35, "learning_rate": 7.479026271331864e-06, "loss": 1.5187, "step": 3327 }, { "epoch": 0.35, "learning_rate": 7.477525922679673e-06, "loss": 1.6863, "step": 3328 }, { "epoch": 0.36, "learning_rate": 7.476025278291912e-06, "loss": 1.6243, "step": 3329 }, { "epoch": 0.36, "learning_rate": 7.4745243383477055e-06, "loss": 1.3256, "step": 3330 }, { "epoch": 0.36, "learning_rate": 7.4730231030262164e-06, "loss": 1.3587, "step": 3331 }, { "epoch": 0.36, "learning_rate": 7.471521572506647e-06, "loss": 1.5336, "step": 3332 }, { "epoch": 0.36, "learning_rate": 7.470019746968226e-06, "loss": 1.3432, "step": 3333 }, { "epoch": 0.36, "learning_rate": 7.468517626590229e-06, "loss": 1.5988, "step": 3334 }, { "epoch": 0.36, "learning_rate": 7.467015211551951e-06, "loss": 1.7854, "step": 3335 }, { "epoch": 0.36, "learning_rate": 7.4655125020327376e-06, "loss": 1.5588, "step": 3336 }, { "epoch": 0.36, "learning_rate": 7.464009498211964e-06, "loss": 1.3195, "step": 3337 }, { "epoch": 0.36, "learning_rate": 7.462506200269034e-06, "loss": 1.6761, "step": 3338 }, { "epoch": 0.36, "learning_rate": 7.461002608383396e-06, "loss": 1.5148, "step": 3339 }, { "epoch": 0.36, "learning_rate": 7.4594987227345314e-06, "loss": 1.4122, "step": 3340 }, { "epoch": 0.36, "learning_rate": 7.457994543501951e-06, "loss": 1.4632, "step": 3341 }, { "epoch": 0.36, "learning_rate": 7.456490070865206e-06, "loss": 1.5733, "step": 3342 }, { "epoch": 0.36, "learning_rate": 7.454985305003884e-06, "loss": 1.4075, "step": 3343 }, { "epoch": 0.36, "learning_rate": 7.453480246097601e-06, "loss": 1.645, "step": 3344 }, { "epoch": 0.36, "learning_rate": 7.4519748943260126e-06, "loss": 1.4959, "step": 3345 }, { "epoch": 0.36, "learning_rate": 7.45046924986881e-06, "loss": 1.7734, "step": 3346 }, { "epoch": 0.36, "learning_rate": 7.448963312905717e-06, "loss": 1.7277, "step": 3347 }, { "epoch": 0.36, "learning_rate": 7.447457083616494e-06, "loss": 1.4837, "step": 3348 }, { "epoch": 0.36, "learning_rate": 7.445950562180935e-06, "loss": 1.4656, "step": 3349 }, { "epoch": 0.36, "learning_rate": 7.444443748778869e-06, "loss": 1.7313, "step": 3350 }, { "epoch": 0.36, "learning_rate": 7.44293664359016e-06, "loss": 1.2318, "step": 3351 }, { "epoch": 0.36, "learning_rate": 7.441429246794708e-06, "loss": 1.3608, "step": 3352 }, { "epoch": 0.36, "learning_rate": 7.4399215585724445e-06, "loss": 1.5625, "step": 3353 }, { "epoch": 0.36, "learning_rate": 7.438413579103344e-06, "loss": 1.3487, "step": 3354 }, { "epoch": 0.36, "learning_rate": 7.436905308567404e-06, "loss": 1.2101, "step": 3355 }, { "epoch": 0.36, "learning_rate": 7.435396747144664e-06, "loss": 1.3131, "step": 3356 }, { "epoch": 0.36, "learning_rate": 7.433887895015199e-06, "loss": 1.438, "step": 3357 }, { "epoch": 0.36, "learning_rate": 7.432378752359115e-06, "loss": 1.6353, "step": 3358 }, { "epoch": 0.36, "learning_rate": 7.430869319356554e-06, "loss": 1.1174, "step": 3359 }, { "epoch": 0.36, "learning_rate": 7.429359596187694e-06, "loss": 1.4971, "step": 3360 }, { "epoch": 0.36, "learning_rate": 7.4278495830327465e-06, "loss": 1.6485, "step": 3361 }, { "epoch": 0.36, "learning_rate": 7.426339280071957e-06, "loss": 1.4449, "step": 3362 }, { "epoch": 0.36, "learning_rate": 7.424828687485606e-06, "loss": 1.0336, "step": 3363 }, { "epoch": 0.36, "learning_rate": 7.42331780545401e-06, "loss": 1.495, "step": 3364 }, { "epoch": 0.36, "learning_rate": 7.421806634157518e-06, "loss": 1.5774, "step": 3365 }, { "epoch": 0.36, "learning_rate": 7.420295173776515e-06, "loss": 1.3983, "step": 3366 }, { "epoch": 0.36, "learning_rate": 7.418783424491418e-06, "loss": 1.6429, "step": 3367 }, { "epoch": 0.36, "learning_rate": 7.417271386482684e-06, "loss": 1.4104, "step": 3368 }, { "epoch": 0.36, "learning_rate": 7.415759059930799e-06, "loss": 1.3453, "step": 3369 }, { "epoch": 0.36, "learning_rate": 7.414246445016283e-06, "loss": 1.5234, "step": 3370 }, { "epoch": 0.36, "learning_rate": 7.412733541919699e-06, "loss": 1.406, "step": 3371 }, { "epoch": 0.36, "learning_rate": 7.411220350821631e-06, "loss": 1.3137, "step": 3372 }, { "epoch": 0.36, "learning_rate": 7.409706871902709e-06, "loss": 1.3387, "step": 3373 }, { "epoch": 0.36, "learning_rate": 7.408193105343592e-06, "loss": 1.1254, "step": 3374 }, { "epoch": 0.36, "learning_rate": 7.406679051324972e-06, "loss": 1.4661, "step": 3375 }, { "epoch": 0.36, "learning_rate": 7.40516471002758e-06, "loss": 1.3307, "step": 3376 }, { "epoch": 0.36, "learning_rate": 7.403650081632178e-06, "loss": 1.6183, "step": 3377 }, { "epoch": 0.36, "learning_rate": 7.402135166319567e-06, "loss": 1.6119, "step": 3378 }, { "epoch": 0.36, "learning_rate": 7.400619964270571e-06, "loss": 1.4824, "step": 3379 }, { "epoch": 0.36, "learning_rate": 7.399104475666059e-06, "loss": 1.4522, "step": 3380 }, { "epoch": 0.36, "learning_rate": 7.397588700686933e-06, "loss": 1.3512, "step": 3381 }, { "epoch": 0.36, "learning_rate": 7.396072639514123e-06, "loss": 1.3042, "step": 3382 }, { "epoch": 0.36, "learning_rate": 7.394556292328601e-06, "loss": 1.5114, "step": 3383 }, { "epoch": 0.36, "learning_rate": 7.393039659311366e-06, "loss": 1.4021, "step": 3384 }, { "epoch": 0.36, "learning_rate": 7.391522740643456e-06, "loss": 1.3644, "step": 3385 }, { "epoch": 0.36, "learning_rate": 7.390005536505942e-06, "loss": 1.4014, "step": 3386 }, { "epoch": 0.36, "learning_rate": 7.388488047079927e-06, "loss": 1.5945, "step": 3387 }, { "epoch": 0.36, "learning_rate": 7.386970272546551e-06, "loss": 1.5337, "step": 3388 }, { "epoch": 0.36, "learning_rate": 7.385452213086985e-06, "loss": 1.455, "step": 3389 }, { "epoch": 0.36, "learning_rate": 7.383933868882438e-06, "loss": 1.3361, "step": 3390 }, { "epoch": 0.36, "learning_rate": 7.382415240114149e-06, "loss": 1.5049, "step": 3391 }, { "epoch": 0.36, "learning_rate": 7.380896326963392e-06, "loss": 1.2469, "step": 3392 }, { "epoch": 0.36, "learning_rate": 7.379377129611478e-06, "loss": 1.5137, "step": 3393 }, { "epoch": 0.36, "learning_rate": 7.377857648239748e-06, "loss": 1.6737, "step": 3394 }, { "epoch": 0.36, "learning_rate": 7.376337883029576e-06, "loss": 1.5525, "step": 3395 }, { "epoch": 0.36, "learning_rate": 7.374817834162378e-06, "loss": 1.1984, "step": 3396 }, { "epoch": 0.36, "learning_rate": 7.373297501819591e-06, "loss": 1.4345, "step": 3397 }, { "epoch": 0.36, "learning_rate": 7.371776886182699e-06, "loss": 1.5044, "step": 3398 }, { "epoch": 0.36, "learning_rate": 7.3702559874332125e-06, "loss": 1.4613, "step": 3399 }, { "epoch": 0.36, "learning_rate": 7.368734805752673e-06, "loss": 1.7155, "step": 3400 }, { "epoch": 0.36, "learning_rate": 7.367213341322666e-06, "loss": 1.3387, "step": 3401 }, { "epoch": 0.36, "learning_rate": 7.3656915943247984e-06, "loss": 1.4207, "step": 3402 }, { "epoch": 0.36, "learning_rate": 7.364169564940722e-06, "loss": 1.5288, "step": 3403 }, { "epoch": 0.36, "learning_rate": 7.362647253352116e-06, "loss": 1.5044, "step": 3404 }, { "epoch": 0.36, "learning_rate": 7.3611246597406925e-06, "loss": 1.4212, "step": 3405 }, { "epoch": 0.36, "learning_rate": 7.3596017842882025e-06, "loss": 1.5418, "step": 3406 }, { "epoch": 0.36, "learning_rate": 7.358078627176426e-06, "loss": 1.3773, "step": 3407 }, { "epoch": 0.36, "learning_rate": 7.356555188587178e-06, "loss": 1.5008, "step": 3408 }, { "epoch": 0.36, "learning_rate": 7.355031468702305e-06, "loss": 1.7606, "step": 3409 }, { "epoch": 0.36, "learning_rate": 7.3535074677036956e-06, "loss": 1.3062, "step": 3410 }, { "epoch": 0.36, "learning_rate": 7.351983185773259e-06, "loss": 1.456, "step": 3411 }, { "epoch": 0.36, "learning_rate": 7.350458623092947e-06, "loss": 1.3699, "step": 3412 }, { "epoch": 0.36, "learning_rate": 7.348933779844746e-06, "loss": 1.4665, "step": 3413 }, { "epoch": 0.36, "learning_rate": 7.347408656210666e-06, "loss": 1.6852, "step": 3414 }, { "epoch": 0.36, "learning_rate": 7.3458832523727615e-06, "loss": 1.4776, "step": 3415 }, { "epoch": 0.36, "learning_rate": 7.344357568513116e-06, "loss": 1.399, "step": 3416 }, { "epoch": 0.36, "learning_rate": 7.342831604813844e-06, "loss": 1.5831, "step": 3417 }, { "epoch": 0.36, "learning_rate": 7.341305361457096e-06, "loss": 1.4665, "step": 3418 }, { "epoch": 0.36, "learning_rate": 7.339778838625055e-06, "loss": 1.4643, "step": 3419 }, { "epoch": 0.36, "learning_rate": 7.338252036499941e-06, "loss": 1.4732, "step": 3420 }, { "epoch": 0.36, "learning_rate": 7.336724955264001e-06, "loss": 1.6802, "step": 3421 }, { "epoch": 0.37, "learning_rate": 7.33519759509952e-06, "loss": 1.5465, "step": 3422 }, { "epoch": 0.37, "learning_rate": 7.333669956188815e-06, "loss": 1.5121, "step": 3423 }, { "epoch": 0.37, "learning_rate": 7.332142038714236e-06, "loss": 1.3728, "step": 3424 }, { "epoch": 0.37, "learning_rate": 7.330613842858165e-06, "loss": 1.2733, "step": 3425 }, { "epoch": 0.37, "learning_rate": 7.3290853688030196e-06, "loss": 1.3786, "step": 3426 }, { "epoch": 0.37, "learning_rate": 7.3275566167312504e-06, "loss": 1.1326, "step": 3427 }, { "epoch": 0.37, "learning_rate": 7.32602758682534e-06, "loss": 1.6996, "step": 3428 }, { "epoch": 0.37, "learning_rate": 7.324498279267803e-06, "loss": 1.5747, "step": 3429 }, { "epoch": 0.37, "learning_rate": 7.322968694241193e-06, "loss": 1.5662, "step": 3430 }, { "epoch": 0.37, "learning_rate": 7.3214388319280845e-06, "loss": 1.507, "step": 3431 }, { "epoch": 0.37, "learning_rate": 7.319908692511103e-06, "loss": 1.3634, "step": 3432 }, { "epoch": 0.37, "learning_rate": 7.318378276172889e-06, "loss": 1.6076, "step": 3433 }, { "epoch": 0.37, "learning_rate": 7.316847583096128e-06, "loss": 1.6545, "step": 3434 }, { "epoch": 0.37, "learning_rate": 7.315316613463535e-06, "loss": 1.6015, "step": 3435 }, { "epoch": 0.37, "learning_rate": 7.313785367457855e-06, "loss": 1.5039, "step": 3436 }, { "epoch": 0.37, "learning_rate": 7.312253845261871e-06, "loss": 1.6723, "step": 3437 }, { "epoch": 0.37, "learning_rate": 7.310722047058396e-06, "loss": 1.5639, "step": 3438 }, { "epoch": 0.37, "learning_rate": 7.3091899730302765e-06, "loss": 1.4114, "step": 3439 }, { "epoch": 0.37, "learning_rate": 7.307657623360393e-06, "loss": 1.5033, "step": 3440 }, { "epoch": 0.37, "learning_rate": 7.306124998231655e-06, "loss": 1.5365, "step": 3441 }, { "epoch": 0.37, "learning_rate": 7.3045920978270116e-06, "loss": 1.3594, "step": 3442 }, { "epoch": 0.37, "learning_rate": 7.303058922329439e-06, "loss": 1.2397, "step": 3443 }, { "epoch": 0.37, "learning_rate": 7.301525471921949e-06, "loss": 1.4982, "step": 3444 }, { "epoch": 0.37, "learning_rate": 7.299991746787584e-06, "loss": 1.3638, "step": 3445 }, { "epoch": 0.37, "learning_rate": 7.298457747109421e-06, "loss": 1.385, "step": 3446 }, { "epoch": 0.37, "learning_rate": 7.296923473070571e-06, "loss": 1.697, "step": 3447 }, { "epoch": 0.37, "learning_rate": 7.295388924854174e-06, "loss": 1.5864, "step": 3448 }, { "epoch": 0.37, "learning_rate": 7.293854102643407e-06, "loss": 1.7166, "step": 3449 }, { "epoch": 0.37, "learning_rate": 7.292319006621477e-06, "loss": 1.2218, "step": 3450 }, { "epoch": 0.37, "learning_rate": 7.290783636971622e-06, "loss": 1.4318, "step": 3451 }, { "epoch": 0.37, "learning_rate": 7.289247993877119e-06, "loss": 1.3318, "step": 3452 }, { "epoch": 0.37, "learning_rate": 7.2877120775212685e-06, "loss": 1.6934, "step": 3453 }, { "epoch": 0.37, "learning_rate": 7.286175888087413e-06, "loss": 1.5981, "step": 3454 }, { "epoch": 0.37, "learning_rate": 7.284639425758924e-06, "loss": 1.4725, "step": 3455 }, { "epoch": 0.37, "learning_rate": 7.283102690719198e-06, "loss": 1.6759, "step": 3456 }, { "epoch": 0.37, "learning_rate": 7.281565683151679e-06, "loss": 1.3111, "step": 3457 }, { "epoch": 0.37, "learning_rate": 7.28002840323983e-06, "loss": 1.6172, "step": 3458 }, { "epoch": 0.37, "learning_rate": 7.278490851167155e-06, "loss": 1.5052, "step": 3459 }, { "epoch": 0.37, "learning_rate": 7.276953027117186e-06, "loss": 1.6996, "step": 3460 }, { "epoch": 0.37, "learning_rate": 7.275414931273489e-06, "loss": 1.3867, "step": 3461 }, { "epoch": 0.37, "learning_rate": 7.2738765638196625e-06, "loss": 1.3614, "step": 3462 }, { "epoch": 0.37, "learning_rate": 7.272337924939338e-06, "loss": 1.7057, "step": 3463 }, { "epoch": 0.37, "learning_rate": 7.270799014816178e-06, "loss": 1.2332, "step": 3464 }, { "epoch": 0.37, "learning_rate": 7.269259833633877e-06, "loss": 1.3216, "step": 3465 }, { "epoch": 0.37, "learning_rate": 7.267720381576165e-06, "loss": 1.5184, "step": 3466 }, { "epoch": 0.37, "learning_rate": 7.2661806588268015e-06, "loss": 1.5802, "step": 3467 }, { "epoch": 0.37, "learning_rate": 7.264640665569577e-06, "loss": 1.4191, "step": 3468 }, { "epoch": 0.37, "learning_rate": 7.263100401988319e-06, "loss": 1.258, "step": 3469 }, { "epoch": 0.37, "learning_rate": 7.261559868266883e-06, "loss": 1.432, "step": 3470 }, { "epoch": 0.37, "learning_rate": 7.26001906458916e-06, "loss": 1.4223, "step": 3471 }, { "epoch": 0.37, "learning_rate": 7.258477991139071e-06, "loss": 1.463, "step": 3472 }, { "epoch": 0.37, "learning_rate": 7.256936648100567e-06, "loss": 1.3076, "step": 3473 }, { "epoch": 0.37, "learning_rate": 7.255395035657639e-06, "loss": 1.5512, "step": 3474 }, { "epoch": 0.37, "learning_rate": 7.253853153994301e-06, "loss": 1.4187, "step": 3475 }, { "epoch": 0.37, "learning_rate": 7.252311003294605e-06, "loss": 1.5716, "step": 3476 }, { "epoch": 0.37, "learning_rate": 7.250768583742634e-06, "loss": 1.4815, "step": 3477 }, { "epoch": 0.37, "learning_rate": 7.2492258955224995e-06, "loss": 1.7142, "step": 3478 }, { "epoch": 0.37, "learning_rate": 7.24768293881835e-06, "loss": 1.4911, "step": 3479 }, { "epoch": 0.37, "learning_rate": 7.246139713814365e-06, "loss": 1.2816, "step": 3480 }, { "epoch": 0.37, "learning_rate": 7.244596220694754e-06, "loss": 1.6503, "step": 3481 }, { "epoch": 0.37, "learning_rate": 7.24305245964376e-06, "loss": 1.5052, "step": 3482 }, { "epoch": 0.37, "learning_rate": 7.241508430845656e-06, "loss": 1.4946, "step": 3483 }, { "epoch": 0.37, "learning_rate": 7.239964134484752e-06, "loss": 1.4084, "step": 3484 }, { "epoch": 0.37, "learning_rate": 7.238419570745383e-06, "loss": 1.5152, "step": 3485 }, { "epoch": 0.37, "learning_rate": 7.236874739811921e-06, "loss": 1.2646, "step": 3486 }, { "epoch": 0.37, "learning_rate": 7.235329641868768e-06, "loss": 1.1159, "step": 3487 }, { "epoch": 0.37, "learning_rate": 7.233784277100359e-06, "loss": 1.3822, "step": 3488 }, { "epoch": 0.37, "learning_rate": 7.232238645691157e-06, "loss": 1.4683, "step": 3489 }, { "epoch": 0.37, "learning_rate": 7.2306927478256624e-06, "loss": 1.5594, "step": 3490 }, { "epoch": 0.37, "learning_rate": 7.2291465836884075e-06, "loss": 1.5436, "step": 3491 }, { "epoch": 0.37, "learning_rate": 7.227600153463947e-06, "loss": 1.1826, "step": 3492 }, { "epoch": 0.37, "learning_rate": 7.226053457336881e-06, "loss": 1.4753, "step": 3493 }, { "epoch": 0.37, "learning_rate": 7.224506495491831e-06, "loss": 1.6333, "step": 3494 }, { "epoch": 0.37, "learning_rate": 7.222959268113452e-06, "loss": 1.5676, "step": 3495 }, { "epoch": 0.37, "learning_rate": 7.221411775386435e-06, "loss": 1.4135, "step": 3496 }, { "epoch": 0.37, "learning_rate": 7.219864017495499e-06, "loss": 1.4802, "step": 3497 }, { "epoch": 0.37, "learning_rate": 7.218315994625397e-06, "loss": 1.5415, "step": 3498 }, { "epoch": 0.37, "learning_rate": 7.216767706960911e-06, "loss": 1.2626, "step": 3499 }, { "epoch": 0.37, "learning_rate": 7.215219154686855e-06, "loss": 1.4473, "step": 3500 }, { "epoch": 0.37, "learning_rate": 7.213670337988079e-06, "loss": 1.6646, "step": 3501 }, { "epoch": 0.37, "learning_rate": 7.212121257049457e-06, "loss": 1.5413, "step": 3502 }, { "epoch": 0.37, "learning_rate": 7.2105719120559014e-06, "loss": 1.251, "step": 3503 }, { "epoch": 0.37, "learning_rate": 7.209022303192351e-06, "loss": 1.5399, "step": 3504 }, { "epoch": 0.37, "learning_rate": 7.207472430643781e-06, "loss": 1.3256, "step": 3505 }, { "epoch": 0.37, "learning_rate": 7.205922294595194e-06, "loss": 1.2364, "step": 3506 }, { "epoch": 0.37, "learning_rate": 7.204371895231623e-06, "loss": 1.672, "step": 3507 }, { "epoch": 0.37, "learning_rate": 7.202821232738142e-06, "loss": 1.3102, "step": 3508 }, { "epoch": 0.37, "learning_rate": 7.201270307299844e-06, "loss": 1.4693, "step": 3509 }, { "epoch": 0.37, "learning_rate": 7.199719119101858e-06, "loss": 1.6651, "step": 3510 }, { "epoch": 0.37, "learning_rate": 7.198167668329348e-06, "loss": 1.3417, "step": 3511 }, { "epoch": 0.37, "learning_rate": 7.196615955167505e-06, "loss": 1.9345, "step": 3512 }, { "epoch": 0.37, "learning_rate": 7.195063979801554e-06, "loss": 1.5417, "step": 3513 }, { "epoch": 0.37, "learning_rate": 7.19351174241675e-06, "loss": 1.6674, "step": 3514 }, { "epoch": 0.37, "learning_rate": 7.191959243198379e-06, "loss": 1.388, "step": 3515 }, { "epoch": 0.38, "learning_rate": 7.190406482331757e-06, "loss": 1.7249, "step": 3516 }, { "epoch": 0.38, "learning_rate": 7.188853460002235e-06, "loss": 1.6179, "step": 3517 }, { "epoch": 0.38, "learning_rate": 7.187300176395192e-06, "loss": 1.3585, "step": 3518 }, { "epoch": 0.38, "learning_rate": 7.18574663169604e-06, "loss": 1.6182, "step": 3519 }, { "epoch": 0.38, "learning_rate": 7.1841928260902215e-06, "loss": 1.7371, "step": 3520 }, { "epoch": 0.38, "learning_rate": 7.18263875976321e-06, "loss": 1.6313, "step": 3521 }, { "epoch": 0.38, "learning_rate": 7.1810844329005095e-06, "loss": 1.3279, "step": 3522 }, { "epoch": 0.38, "learning_rate": 7.179529845687656e-06, "loss": 1.3825, "step": 3523 }, { "epoch": 0.38, "learning_rate": 7.1779749983102155e-06, "loss": 1.4448, "step": 3524 }, { "epoch": 0.38, "learning_rate": 7.176419890953788e-06, "loss": 1.6076, "step": 3525 }, { "epoch": 0.38, "learning_rate": 7.174864523804002e-06, "loss": 1.7198, "step": 3526 }, { "epoch": 0.38, "learning_rate": 7.173308897046516e-06, "loss": 1.3562, "step": 3527 }, { "epoch": 0.38, "learning_rate": 7.171753010867023e-06, "loss": 1.4325, "step": 3528 }, { "epoch": 0.38, "learning_rate": 7.170196865451242e-06, "loss": 1.5684, "step": 3529 }, { "epoch": 0.38, "learning_rate": 7.168640460984929e-06, "loss": 1.2316, "step": 3530 }, { "epoch": 0.38, "learning_rate": 7.167083797653866e-06, "loss": 1.6824, "step": 3531 }, { "epoch": 0.38, "learning_rate": 7.1655268756438675e-06, "loss": 1.6569, "step": 3532 }, { "epoch": 0.38, "learning_rate": 7.163969695140782e-06, "loss": 1.5355, "step": 3533 }, { "epoch": 0.38, "learning_rate": 7.162412256330481e-06, "loss": 1.5009, "step": 3534 }, { "epoch": 0.38, "learning_rate": 7.160854559398876e-06, "loss": 1.2874, "step": 3535 }, { "epoch": 0.38, "learning_rate": 7.159296604531902e-06, "loss": 1.6906, "step": 3536 }, { "epoch": 0.38, "learning_rate": 7.157738391915531e-06, "loss": 1.3697, "step": 3537 }, { "epoch": 0.38, "learning_rate": 7.15617992173576e-06, "loss": 1.8056, "step": 3538 }, { "epoch": 0.38, "learning_rate": 7.15462119417862e-06, "loss": 1.711, "step": 3539 }, { "epoch": 0.38, "learning_rate": 7.153062209430174e-06, "loss": 1.5805, "step": 3540 }, { "epoch": 0.38, "learning_rate": 7.151502967676511e-06, "loss": 1.4864, "step": 3541 }, { "epoch": 0.38, "learning_rate": 7.149943469103755e-06, "loss": 1.2188, "step": 3542 }, { "epoch": 0.38, "learning_rate": 7.148383713898058e-06, "loss": 1.6299, "step": 3543 }, { "epoch": 0.38, "learning_rate": 7.146823702245606e-06, "loss": 1.8165, "step": 3544 }, { "epoch": 0.38, "learning_rate": 7.145263434332611e-06, "loss": 1.3876, "step": 3545 }, { "epoch": 0.38, "learning_rate": 7.143702910345318e-06, "loss": 1.209, "step": 3546 }, { "epoch": 0.38, "learning_rate": 7.142142130470007e-06, "loss": 1.4517, "step": 3547 }, { "epoch": 0.38, "learning_rate": 7.140581094892978e-06, "loss": 1.4018, "step": 3548 }, { "epoch": 0.38, "learning_rate": 7.139019803800569e-06, "loss": 1.5034, "step": 3549 }, { "epoch": 0.38, "learning_rate": 7.137458257379151e-06, "loss": 1.2275, "step": 3550 }, { "epoch": 0.38, "learning_rate": 7.135896455815117e-06, "loss": 1.5827, "step": 3551 }, { "epoch": 0.38, "learning_rate": 7.134334399294897e-06, "loss": 1.4666, "step": 3552 }, { "epoch": 0.38, "learning_rate": 7.132772088004951e-06, "loss": 1.348, "step": 3553 }, { "epoch": 0.38, "learning_rate": 7.131209522131764e-06, "loss": 1.3856, "step": 3554 }, { "epoch": 0.38, "learning_rate": 7.129646701861858e-06, "loss": 1.4814, "step": 3555 }, { "epoch": 0.38, "learning_rate": 7.128083627381782e-06, "loss": 1.1785, "step": 3556 }, { "epoch": 0.38, "learning_rate": 7.1265202988781165e-06, "loss": 1.419, "step": 3557 }, { "epoch": 0.38, "learning_rate": 7.124956716537471e-06, "loss": 1.6238, "step": 3558 }, { "epoch": 0.38, "learning_rate": 7.123392880546487e-06, "loss": 1.6578, "step": 3559 }, { "epoch": 0.38, "learning_rate": 7.121828791091834e-06, "loss": 1.6462, "step": 3560 }, { "epoch": 0.38, "learning_rate": 7.120264448360214e-06, "loss": 1.2362, "step": 3561 }, { "epoch": 0.38, "learning_rate": 7.11869985253836e-06, "loss": 1.4477, "step": 3562 }, { "epoch": 0.38, "learning_rate": 7.11713500381303e-06, "loss": 1.2921, "step": 3563 }, { "epoch": 0.38, "learning_rate": 7.115569902371018e-06, "loss": 1.489, "step": 3564 }, { "epoch": 0.38, "learning_rate": 7.114004548399146e-06, "loss": 1.1227, "step": 3565 }, { "epoch": 0.38, "learning_rate": 7.112438942084264e-06, "loss": 1.2971, "step": 3566 }, { "epoch": 0.38, "learning_rate": 7.110873083613259e-06, "loss": 1.3452, "step": 3567 }, { "epoch": 0.38, "learning_rate": 7.109306973173038e-06, "loss": 1.3663, "step": 3568 }, { "epoch": 0.38, "learning_rate": 7.107740610950546e-06, "loss": 1.4822, "step": 3569 }, { "epoch": 0.38, "learning_rate": 7.106173997132755e-06, "loss": 1.645, "step": 3570 }, { "epoch": 0.38, "learning_rate": 7.104607131906667e-06, "loss": 1.2158, "step": 3571 }, { "epoch": 0.38, "learning_rate": 7.103040015459315e-06, "loss": 1.3841, "step": 3572 }, { "epoch": 0.38, "learning_rate": 7.101472647977761e-06, "loss": 1.4216, "step": 3573 }, { "epoch": 0.38, "learning_rate": 7.099905029649098e-06, "loss": 1.395, "step": 3574 }, { "epoch": 0.38, "learning_rate": 7.098337160660449e-06, "loss": 1.4472, "step": 3575 }, { "epoch": 0.38, "learning_rate": 7.096769041198964e-06, "loss": 1.699, "step": 3576 }, { "epoch": 0.38, "learning_rate": 7.0952006714518275e-06, "loss": 1.6221, "step": 3577 }, { "epoch": 0.38, "learning_rate": 7.09363205160625e-06, "loss": 1.395, "step": 3578 }, { "epoch": 0.38, "learning_rate": 7.0920631818494745e-06, "loss": 1.1652, "step": 3579 }, { "epoch": 0.38, "learning_rate": 7.090494062368771e-06, "loss": 1.36, "step": 3580 }, { "epoch": 0.38, "learning_rate": 7.088924693351445e-06, "loss": 1.569, "step": 3581 }, { "epoch": 0.38, "learning_rate": 7.087355074984823e-06, "loss": 1.4081, "step": 3582 }, { "epoch": 0.38, "learning_rate": 7.085785207456268e-06, "loss": 1.494, "step": 3583 }, { "epoch": 0.38, "learning_rate": 7.084215090953172e-06, "loss": 1.2928, "step": 3584 }, { "epoch": 0.38, "learning_rate": 7.082644725662954e-06, "loss": 1.393, "step": 3585 }, { "epoch": 0.38, "learning_rate": 7.081074111773066e-06, "loss": 1.4263, "step": 3586 }, { "epoch": 0.38, "learning_rate": 7.079503249470987e-06, "loss": 1.3188, "step": 3587 }, { "epoch": 0.38, "learning_rate": 7.077932138944225e-06, "loss": 1.4022, "step": 3588 }, { "epoch": 0.38, "learning_rate": 7.076360780380321e-06, "loss": 1.5786, "step": 3589 }, { "epoch": 0.38, "learning_rate": 7.074789173966843e-06, "loss": 1.5079, "step": 3590 }, { "epoch": 0.38, "learning_rate": 7.073217319891391e-06, "loss": 1.3815, "step": 3591 }, { "epoch": 0.38, "learning_rate": 7.071645218341593e-06, "loss": 1.3179, "step": 3592 }, { "epoch": 0.38, "learning_rate": 7.070072869505103e-06, "loss": 1.3824, "step": 3593 }, { "epoch": 0.38, "learning_rate": 7.068500273569612e-06, "loss": 1.2439, "step": 3594 }, { "epoch": 0.38, "learning_rate": 7.066927430722833e-06, "loss": 1.351, "step": 3595 }, { "epoch": 0.38, "learning_rate": 7.065354341152515e-06, "loss": 1.5084, "step": 3596 }, { "epoch": 0.38, "learning_rate": 7.063781005046433e-06, "loss": 1.5352, "step": 3597 }, { "epoch": 0.38, "learning_rate": 7.062207422592391e-06, "loss": 1.3033, "step": 3598 }, { "epoch": 0.38, "learning_rate": 7.060633593978223e-06, "loss": 1.6151, "step": 3599 }, { "epoch": 0.38, "learning_rate": 7.059059519391794e-06, "loss": 1.3958, "step": 3600 }, { "epoch": 0.38, "learning_rate": 7.057485199020996e-06, "loss": 1.4819, "step": 3601 }, { "epoch": 0.38, "learning_rate": 7.055910633053751e-06, "loss": 1.2985, "step": 3602 }, { "epoch": 0.38, "learning_rate": 7.054335821678012e-06, "loss": 1.5358, "step": 3603 }, { "epoch": 0.38, "learning_rate": 7.0527607650817585e-06, "loss": 1.4702, "step": 3604 }, { "epoch": 0.38, "learning_rate": 7.051185463453001e-06, "loss": 1.2095, "step": 3605 }, { "epoch": 0.38, "learning_rate": 7.049609916979782e-06, "loss": 1.3563, "step": 3606 }, { "epoch": 0.38, "learning_rate": 7.048034125850165e-06, "loss": 1.3943, "step": 3607 }, { "epoch": 0.38, "learning_rate": 7.046458090252252e-06, "loss": 1.7749, "step": 3608 }, { "epoch": 0.38, "learning_rate": 7.044881810374169e-06, "loss": 1.6261, "step": 3609 }, { "epoch": 0.39, "learning_rate": 7.043305286404071e-06, "loss": 1.382, "step": 3610 }, { "epoch": 0.39, "learning_rate": 7.0417285185301445e-06, "loss": 1.6237, "step": 3611 }, { "epoch": 0.39, "learning_rate": 7.040151506940605e-06, "loss": 1.5972, "step": 3612 }, { "epoch": 0.39, "learning_rate": 7.038574251823694e-06, "loss": 1.612, "step": 3613 }, { "epoch": 0.39, "learning_rate": 7.036996753367686e-06, "loss": 1.3659, "step": 3614 }, { "epoch": 0.39, "learning_rate": 7.035419011760882e-06, "loss": 1.4565, "step": 3615 }, { "epoch": 0.39, "learning_rate": 7.0338410271916126e-06, "loss": 1.5164, "step": 3616 }, { "epoch": 0.39, "learning_rate": 7.032262799848237e-06, "loss": 1.2823, "step": 3617 }, { "epoch": 0.39, "learning_rate": 7.0306843299191465e-06, "loss": 1.4969, "step": 3618 }, { "epoch": 0.39, "learning_rate": 7.029105617592755e-06, "loss": 1.4066, "step": 3619 }, { "epoch": 0.39, "learning_rate": 7.027526663057511e-06, "loss": 1.3886, "step": 3620 }, { "epoch": 0.39, "learning_rate": 7.0259474665018915e-06, "loss": 1.4178, "step": 3621 }, { "epoch": 0.39, "learning_rate": 7.024368028114397e-06, "loss": 1.7289, "step": 3622 }, { "epoch": 0.39, "learning_rate": 7.022788348083567e-06, "loss": 1.6488, "step": 3623 }, { "epoch": 0.39, "learning_rate": 7.0212084265979575e-06, "loss": 1.3582, "step": 3624 }, { "epoch": 0.39, "learning_rate": 7.0196282638461624e-06, "loss": 1.5528, "step": 3625 }, { "epoch": 0.39, "learning_rate": 7.0180478600168e-06, "loss": 1.3048, "step": 3626 }, { "epoch": 0.39, "learning_rate": 7.016467215298519e-06, "loss": 1.477, "step": 3627 }, { "epoch": 0.39, "learning_rate": 7.0148863298800005e-06, "loss": 1.4041, "step": 3628 }, { "epoch": 0.39, "learning_rate": 7.013305203949945e-06, "loss": 1.2712, "step": 3629 }, { "epoch": 0.39, "learning_rate": 7.011723837697091e-06, "loss": 1.4147, "step": 3630 }, { "epoch": 0.39, "learning_rate": 7.010142231310201e-06, "loss": 1.3827, "step": 3631 }, { "epoch": 0.39, "learning_rate": 7.008560384978065e-06, "loss": 1.7972, "step": 3632 }, { "epoch": 0.39, "learning_rate": 7.0069782988895056e-06, "loss": 1.4165, "step": 3633 }, { "epoch": 0.39, "learning_rate": 7.005395973233372e-06, "loss": 1.3722, "step": 3634 }, { "epoch": 0.39, "learning_rate": 7.003813408198543e-06, "loss": 1.7652, "step": 3635 }, { "epoch": 0.39, "learning_rate": 7.002230603973924e-06, "loss": 1.5566, "step": 3636 }, { "epoch": 0.39, "learning_rate": 7.00064756074845e-06, "loss": 1.3932, "step": 3637 }, { "epoch": 0.39, "learning_rate": 6.999064278711085e-06, "loss": 1.4527, "step": 3638 }, { "epoch": 0.39, "learning_rate": 6.9974807580508205e-06, "loss": 1.3321, "step": 3639 }, { "epoch": 0.39, "learning_rate": 6.995896998956677e-06, "loss": 1.8368, "step": 3640 }, { "epoch": 0.39, "learning_rate": 6.994313001617705e-06, "loss": 1.2517, "step": 3641 }, { "epoch": 0.39, "learning_rate": 6.992728766222982e-06, "loss": 1.4502, "step": 3642 }, { "epoch": 0.39, "learning_rate": 6.991144292961611e-06, "loss": 1.1344, "step": 3643 }, { "epoch": 0.39, "learning_rate": 6.989559582022728e-06, "loss": 1.5068, "step": 3644 }, { "epoch": 0.39, "learning_rate": 6.987974633595498e-06, "loss": 1.5999, "step": 3645 }, { "epoch": 0.39, "learning_rate": 6.986389447869108e-06, "loss": 1.3372, "step": 3646 }, { "epoch": 0.39, "learning_rate": 6.984804025032781e-06, "loss": 1.3601, "step": 3647 }, { "epoch": 0.39, "learning_rate": 6.9832183652757625e-06, "loss": 1.4168, "step": 3648 }, { "epoch": 0.39, "learning_rate": 6.981632468787327e-06, "loss": 1.577, "step": 3649 }, { "epoch": 0.39, "learning_rate": 6.980046335756782e-06, "loss": 1.3971, "step": 3650 }, { "epoch": 0.39, "learning_rate": 6.978459966373458e-06, "loss": 1.5309, "step": 3651 }, { "epoch": 0.39, "learning_rate": 6.976873360826716e-06, "loss": 1.6902, "step": 3652 }, { "epoch": 0.39, "learning_rate": 6.975286519305944e-06, "loss": 1.2776, "step": 3653 }, { "epoch": 0.39, "learning_rate": 6.973699442000561e-06, "loss": 1.7062, "step": 3654 }, { "epoch": 0.39, "learning_rate": 6.972112129100011e-06, "loss": 1.3688, "step": 3655 }, { "epoch": 0.39, "learning_rate": 6.970524580793766e-06, "loss": 1.4769, "step": 3656 }, { "epoch": 0.39, "learning_rate": 6.96893679727133e-06, "loss": 1.4867, "step": 3657 }, { "epoch": 0.39, "learning_rate": 6.967348778722229e-06, "loss": 1.6046, "step": 3658 }, { "epoch": 0.39, "learning_rate": 6.965760525336025e-06, "loss": 1.6272, "step": 3659 }, { "epoch": 0.39, "learning_rate": 6.9641720373022996e-06, "loss": 1.2602, "step": 3660 }, { "epoch": 0.39, "learning_rate": 6.962583314810666e-06, "loss": 1.6735, "step": 3661 }, { "epoch": 0.39, "learning_rate": 6.96099435805077e-06, "loss": 1.4477, "step": 3662 }, { "epoch": 0.39, "learning_rate": 6.959405167212278e-06, "loss": 1.3945, "step": 3663 }, { "epoch": 0.39, "learning_rate": 6.957815742484886e-06, "loss": 1.4411, "step": 3664 }, { "epoch": 0.39, "learning_rate": 6.956226084058324e-06, "loss": 1.5231, "step": 3665 }, { "epoch": 0.39, "learning_rate": 6.954636192122339e-06, "loss": 1.6542, "step": 3666 }, { "epoch": 0.39, "learning_rate": 6.953046066866717e-06, "loss": 1.3771, "step": 3667 }, { "epoch": 0.39, "learning_rate": 6.951455708481268e-06, "loss": 1.5593, "step": 3668 }, { "epoch": 0.39, "learning_rate": 6.949865117155823e-06, "loss": 1.4984, "step": 3669 }, { "epoch": 0.39, "learning_rate": 6.948274293080252e-06, "loss": 1.2196, "step": 3670 }, { "epoch": 0.39, "learning_rate": 6.946683236444444e-06, "loss": 1.7656, "step": 3671 }, { "epoch": 0.39, "learning_rate": 6.94509194743832e-06, "loss": 1.5573, "step": 3672 }, { "epoch": 0.39, "learning_rate": 6.94350042625183e-06, "loss": 1.143, "step": 3673 }, { "epoch": 0.39, "learning_rate": 6.941908673074946e-06, "loss": 1.3804, "step": 3674 }, { "epoch": 0.39, "learning_rate": 6.940316688097675e-06, "loss": 1.7416, "step": 3675 }, { "epoch": 0.39, "learning_rate": 6.938724471510043e-06, "loss": 1.4541, "step": 3676 }, { "epoch": 0.39, "learning_rate": 6.937132023502114e-06, "loss": 1.4798, "step": 3677 }, { "epoch": 0.39, "learning_rate": 6.935539344263971e-06, "loss": 1.4708, "step": 3678 }, { "epoch": 0.39, "learning_rate": 6.9339464339857295e-06, "loss": 1.541, "step": 3679 }, { "epoch": 0.39, "learning_rate": 6.93235329285753e-06, "loss": 1.6456, "step": 3680 }, { "epoch": 0.39, "learning_rate": 6.93075992106954e-06, "loss": 1.7645, "step": 3681 }, { "epoch": 0.39, "learning_rate": 6.929166318811959e-06, "loss": 1.5164, "step": 3682 }, { "epoch": 0.39, "learning_rate": 6.927572486275007e-06, "loss": 1.6538, "step": 3683 }, { "epoch": 0.39, "learning_rate": 6.925978423648941e-06, "loss": 1.4849, "step": 3684 }, { "epoch": 0.39, "learning_rate": 6.9243841311240356e-06, "loss": 1.691, "step": 3685 }, { "epoch": 0.39, "learning_rate": 6.922789608890598e-06, "loss": 1.204, "step": 3686 }, { "epoch": 0.39, "learning_rate": 6.921194857138963e-06, "loss": 1.4061, "step": 3687 }, { "epoch": 0.39, "learning_rate": 6.91959987605949e-06, "loss": 1.4964, "step": 3688 }, { "epoch": 0.39, "learning_rate": 6.918004665842571e-06, "loss": 1.4973, "step": 3689 }, { "epoch": 0.39, "learning_rate": 6.91640922667862e-06, "loss": 1.5003, "step": 3690 }, { "epoch": 0.39, "learning_rate": 6.914813558758078e-06, "loss": 1.2469, "step": 3691 }, { "epoch": 0.39, "learning_rate": 6.913217662271419e-06, "loss": 1.3997, "step": 3692 }, { "epoch": 0.39, "learning_rate": 6.911621537409139e-06, "loss": 1.4288, "step": 3693 }, { "epoch": 0.39, "learning_rate": 6.910025184361764e-06, "loss": 1.3142, "step": 3694 }, { "epoch": 0.39, "learning_rate": 6.908428603319844e-06, "loss": 1.1912, "step": 3695 }, { "epoch": 0.39, "learning_rate": 6.906831794473963e-06, "loss": 1.6401, "step": 3696 }, { "epoch": 0.39, "learning_rate": 6.905234758014724e-06, "loss": 1.3447, "step": 3697 }, { "epoch": 0.39, "learning_rate": 6.903637494132762e-06, "loss": 1.4688, "step": 3698 }, { "epoch": 0.39, "learning_rate": 6.9020400030187394e-06, "loss": 1.5234, "step": 3699 }, { "epoch": 0.39, "learning_rate": 6.900442284863342e-06, "loss": 1.2714, "step": 3700 }, { "epoch": 0.39, "learning_rate": 6.898844339857288e-06, "loss": 1.4973, "step": 3701 }, { "epoch": 0.39, "learning_rate": 6.897246168191317e-06, "loss": 1.4452, "step": 3702 }, { "epoch": 0.39, "learning_rate": 6.895647770056199e-06, "loss": 1.2912, "step": 3703 }, { "epoch": 0.4, "learning_rate": 6.894049145642731e-06, "loss": 1.5678, "step": 3704 }, { "epoch": 0.4, "learning_rate": 6.892450295141737e-06, "loss": 1.5427, "step": 3705 }, { "epoch": 0.4, "learning_rate": 6.890851218744068e-06, "loss": 1.6085, "step": 3706 }, { "epoch": 0.4, "learning_rate": 6.8892519166406e-06, "loss": 1.77, "step": 3707 }, { "epoch": 0.4, "learning_rate": 6.887652389022236e-06, "loss": 1.5181, "step": 3708 }, { "epoch": 0.4, "learning_rate": 6.8860526360799105e-06, "loss": 1.3471, "step": 3709 }, { "epoch": 0.4, "learning_rate": 6.88445265800458e-06, "loss": 1.5856, "step": 3710 }, { "epoch": 0.4, "learning_rate": 6.88285245498723e-06, "loss": 1.5877, "step": 3711 }, { "epoch": 0.4, "learning_rate": 6.881252027218872e-06, "loss": 1.3006, "step": 3712 }, { "epoch": 0.4, "learning_rate": 6.8796513748905446e-06, "loss": 1.3799, "step": 3713 }, { "epoch": 0.4, "learning_rate": 6.878050498193314e-06, "loss": 1.371, "step": 3714 }, { "epoch": 0.4, "learning_rate": 6.876449397318272e-06, "loss": 1.4407, "step": 3715 }, { "epoch": 0.4, "learning_rate": 6.874848072456538e-06, "loss": 1.5929, "step": 3716 }, { "epoch": 0.4, "learning_rate": 6.873246523799256e-06, "loss": 1.4835, "step": 3717 }, { "epoch": 0.4, "learning_rate": 6.871644751537604e-06, "loss": 1.3048, "step": 3718 }, { "epoch": 0.4, "learning_rate": 6.870042755862775e-06, "loss": 1.3869, "step": 3719 }, { "epoch": 0.4, "learning_rate": 6.868440536965997e-06, "loss": 1.5077, "step": 3720 }, { "epoch": 0.4, "learning_rate": 6.866838095038524e-06, "loss": 1.5016, "step": 3721 }, { "epoch": 0.4, "learning_rate": 6.865235430271633e-06, "loss": 1.2851, "step": 3722 }, { "epoch": 0.4, "learning_rate": 6.863632542856632e-06, "loss": 1.3415, "step": 3723 }, { "epoch": 0.4, "learning_rate": 6.862029432984852e-06, "loss": 1.6563, "step": 3724 }, { "epoch": 0.4, "learning_rate": 6.86042610084765e-06, "loss": 1.398, "step": 3725 }, { "epoch": 0.4, "learning_rate": 6.858822546636417e-06, "loss": 1.3805, "step": 3726 }, { "epoch": 0.4, "learning_rate": 6.857218770542557e-06, "loss": 1.7074, "step": 3727 }, { "epoch": 0.4, "learning_rate": 6.855614772757516e-06, "loss": 1.4729, "step": 3728 }, { "epoch": 0.4, "learning_rate": 6.854010553472757e-06, "loss": 1.5108, "step": 3729 }, { "epoch": 0.4, "learning_rate": 6.852406112879766e-06, "loss": 1.5019, "step": 3730 }, { "epoch": 0.4, "learning_rate": 6.850801451170066e-06, "loss": 1.3076, "step": 3731 }, { "epoch": 0.4, "learning_rate": 6.849196568535201e-06, "loss": 1.2858, "step": 3732 }, { "epoch": 0.4, "learning_rate": 6.847591465166741e-06, "loss": 1.4592, "step": 3733 }, { "epoch": 0.4, "learning_rate": 6.84598614125628e-06, "loss": 1.6653, "step": 3734 }, { "epoch": 0.4, "learning_rate": 6.8443805969954445e-06, "loss": 1.5177, "step": 3735 }, { "epoch": 0.4, "learning_rate": 6.8427748325758834e-06, "loss": 1.405, "step": 3736 }, { "epoch": 0.4, "learning_rate": 6.8411688481892715e-06, "loss": 1.4137, "step": 3737 }, { "epoch": 0.4, "learning_rate": 6.839562644027311e-06, "loss": 1.4641, "step": 3738 }, { "epoch": 0.4, "learning_rate": 6.8379562202817295e-06, "loss": 1.5019, "step": 3739 }, { "epoch": 0.4, "learning_rate": 6.836349577144284e-06, "loss": 1.482, "step": 3740 }, { "epoch": 0.4, "learning_rate": 6.834742714806754e-06, "loss": 1.6523, "step": 3741 }, { "epoch": 0.4, "learning_rate": 6.833135633460943e-06, "loss": 1.4897, "step": 3742 }, { "epoch": 0.4, "learning_rate": 6.83152833329869e-06, "loss": 1.711, "step": 3743 }, { "epoch": 0.4, "learning_rate": 6.8299208145118475e-06, "loss": 1.4686, "step": 3744 }, { "epoch": 0.4, "learning_rate": 6.828313077292307e-06, "loss": 1.3929, "step": 3745 }, { "epoch": 0.4, "learning_rate": 6.8267051218319766e-06, "loss": 1.4933, "step": 3746 }, { "epoch": 0.4, "learning_rate": 6.825096948322791e-06, "loss": 1.4401, "step": 3747 }, { "epoch": 0.4, "learning_rate": 6.823488556956719e-06, "loss": 1.4916, "step": 3748 }, { "epoch": 0.4, "learning_rate": 6.821879947925747e-06, "loss": 1.4395, "step": 3749 }, { "epoch": 0.4, "learning_rate": 6.820271121421889e-06, "loss": 1.5469, "step": 3750 }, { "epoch": 0.4, "learning_rate": 6.818662077637189e-06, "loss": 1.5758, "step": 3751 }, { "epoch": 0.4, "learning_rate": 6.817052816763713e-06, "loss": 1.261, "step": 3752 }, { "epoch": 0.4, "learning_rate": 6.815443338993554e-06, "loss": 1.3422, "step": 3753 }, { "epoch": 0.4, "learning_rate": 6.81383364451883e-06, "loss": 1.6309, "step": 3754 }, { "epoch": 0.4, "learning_rate": 6.812223733531689e-06, "loss": 1.6703, "step": 3755 }, { "epoch": 0.4, "learning_rate": 6.810613606224299e-06, "loss": 1.317, "step": 3756 }, { "epoch": 0.4, "learning_rate": 6.809003262788857e-06, "loss": 1.3265, "step": 3757 }, { "epoch": 0.4, "learning_rate": 6.807392703417586e-06, "loss": 1.568, "step": 3758 }, { "epoch": 0.4, "learning_rate": 6.805781928302732e-06, "loss": 1.2927, "step": 3759 }, { "epoch": 0.4, "learning_rate": 6.804170937636573e-06, "loss": 1.2507, "step": 3760 }, { "epoch": 0.4, "learning_rate": 6.802559731611404e-06, "loss": 1.5513, "step": 3761 }, { "epoch": 0.4, "learning_rate": 6.800948310419554e-06, "loss": 1.7034, "step": 3762 }, { "epoch": 0.4, "learning_rate": 6.799336674253373e-06, "loss": 1.1815, "step": 3763 }, { "epoch": 0.4, "learning_rate": 6.797724823305236e-06, "loss": 1.3332, "step": 3764 }, { "epoch": 0.4, "learning_rate": 6.796112757767547e-06, "loss": 1.4674, "step": 3765 }, { "epoch": 0.4, "learning_rate": 6.794500477832733e-06, "loss": 1.4173, "step": 3766 }, { "epoch": 0.4, "learning_rate": 6.792887983693248e-06, "loss": 1.6283, "step": 3767 }, { "epoch": 0.4, "learning_rate": 6.7912752755415716e-06, "loss": 1.5425, "step": 3768 }, { "epoch": 0.4, "learning_rate": 6.789662353570207e-06, "loss": 1.3961, "step": 3769 }, { "epoch": 0.4, "learning_rate": 6.788049217971685e-06, "loss": 1.3314, "step": 3770 }, { "epoch": 0.4, "learning_rate": 6.786435868938561e-06, "loss": 1.439, "step": 3771 }, { "epoch": 0.4, "learning_rate": 6.7848223066634174e-06, "loss": 1.682, "step": 3772 }, { "epoch": 0.4, "learning_rate": 6.783208531338861e-06, "loss": 1.4099, "step": 3773 }, { "epoch": 0.4, "learning_rate": 6.78159454315752e-06, "loss": 1.3462, "step": 3774 }, { "epoch": 0.4, "learning_rate": 6.779980342312056e-06, "loss": 1.3748, "step": 3775 }, { "epoch": 0.4, "learning_rate": 6.77836592899515e-06, "loss": 1.379, "step": 3776 }, { "epoch": 0.4, "learning_rate": 6.776751303399509e-06, "loss": 1.3309, "step": 3777 }, { "epoch": 0.4, "learning_rate": 6.775136465717868e-06, "loss": 1.553, "step": 3778 }, { "epoch": 0.4, "learning_rate": 6.773521416142987e-06, "loss": 1.1418, "step": 3779 }, { "epoch": 0.4, "learning_rate": 6.771906154867649e-06, "loss": 1.3783, "step": 3780 }, { "epoch": 0.4, "learning_rate": 6.770290682084661e-06, "loss": 1.5138, "step": 3781 }, { "epoch": 0.4, "learning_rate": 6.768674997986863e-06, "loss": 1.5991, "step": 3782 }, { "epoch": 0.4, "learning_rate": 6.767059102767109e-06, "loss": 1.5176, "step": 3783 }, { "epoch": 0.4, "learning_rate": 6.765442996618287e-06, "loss": 1.8988, "step": 3784 }, { "epoch": 0.4, "learning_rate": 6.763826679733307e-06, "loss": 1.532, "step": 3785 }, { "epoch": 0.4, "learning_rate": 6.7622101523051045e-06, "loss": 1.7708, "step": 3786 }, { "epoch": 0.4, "learning_rate": 6.7605934145266395e-06, "loss": 1.5586, "step": 3787 }, { "epoch": 0.4, "learning_rate": 6.758976466590898e-06, "loss": 1.6347, "step": 3788 }, { "epoch": 0.4, "learning_rate": 6.757359308690889e-06, "loss": 1.6987, "step": 3789 }, { "epoch": 0.4, "learning_rate": 6.755741941019652e-06, "loss": 1.354, "step": 3790 }, { "epoch": 0.4, "learning_rate": 6.754124363770246e-06, "loss": 1.2854, "step": 3791 }, { "epoch": 0.4, "learning_rate": 6.7525065771357546e-06, "loss": 1.3853, "step": 3792 }, { "epoch": 0.4, "learning_rate": 6.7508885813092915e-06, "loss": 1.3245, "step": 3793 }, { "epoch": 0.4, "learning_rate": 6.749270376483992e-06, "loss": 1.3746, "step": 3794 }, { "epoch": 0.4, "learning_rate": 6.7476519628530145e-06, "loss": 1.4608, "step": 3795 }, { "epoch": 0.4, "learning_rate": 6.746033340609548e-06, "loss": 1.4717, "step": 3796 }, { "epoch": 0.41, "learning_rate": 6.744414509946802e-06, "loss": 1.4798, "step": 3797 }, { "epoch": 0.41, "learning_rate": 6.742795471058009e-06, "loss": 1.1475, "step": 3798 }, { "epoch": 0.41, "learning_rate": 6.741176224136434e-06, "loss": 1.3791, "step": 3799 }, { "epoch": 0.41, "learning_rate": 6.739556769375359e-06, "loss": 1.3196, "step": 3800 }, { "epoch": 0.41, "learning_rate": 6.737937106968094e-06, "loss": 1.5542, "step": 3801 }, { "epoch": 0.41, "learning_rate": 6.736317237107976e-06, "loss": 1.4257, "step": 3802 }, { "epoch": 0.41, "learning_rate": 6.734697159988362e-06, "loss": 1.3766, "step": 3803 }, { "epoch": 0.41, "learning_rate": 6.7330768758026374e-06, "loss": 1.4381, "step": 3804 }, { "epoch": 0.41, "learning_rate": 6.731456384744212e-06, "loss": 1.5918, "step": 3805 }, { "epoch": 0.41, "learning_rate": 6.729835687006519e-06, "loss": 1.3693, "step": 3806 }, { "epoch": 0.41, "learning_rate": 6.728214782783013e-06, "loss": 1.3782, "step": 3807 }, { "epoch": 0.41, "learning_rate": 6.726593672267183e-06, "loss": 1.6385, "step": 3808 }, { "epoch": 0.41, "learning_rate": 6.724972355652533e-06, "loss": 1.4667, "step": 3809 }, { "epoch": 0.41, "learning_rate": 6.723350833132596e-06, "loss": 1.5882, "step": 3810 }, { "epoch": 0.41, "learning_rate": 6.721729104900927e-06, "loss": 1.8381, "step": 3811 }, { "epoch": 0.41, "learning_rate": 6.720107171151112e-06, "loss": 1.5503, "step": 3812 }, { "epoch": 0.41, "learning_rate": 6.7184850320767505e-06, "loss": 1.4693, "step": 3813 }, { "epoch": 0.41, "learning_rate": 6.716862687871477e-06, "loss": 1.5728, "step": 3814 }, { "epoch": 0.41, "learning_rate": 6.715240138728946e-06, "loss": 1.5987, "step": 3815 }, { "epoch": 0.41, "learning_rate": 6.7136173848428375e-06, "loss": 1.5869, "step": 3816 }, { "epoch": 0.41, "learning_rate": 6.711994426406853e-06, "loss": 1.4868, "step": 3817 }, { "epoch": 0.41, "learning_rate": 6.71037126361472e-06, "loss": 1.2526, "step": 3818 }, { "epoch": 0.41, "learning_rate": 6.708747896660196e-06, "loss": 1.7676, "step": 3819 }, { "epoch": 0.41, "learning_rate": 6.707124325737051e-06, "loss": 1.5442, "step": 3820 }, { "epoch": 0.41, "learning_rate": 6.7055005510390935e-06, "loss": 1.5436, "step": 3821 }, { "epoch": 0.41, "learning_rate": 6.703876572760144e-06, "loss": 1.4978, "step": 3822 }, { "epoch": 0.41, "learning_rate": 6.702252391094053e-06, "loss": 1.4663, "step": 3823 }, { "epoch": 0.41, "learning_rate": 6.700628006234698e-06, "loss": 1.6181, "step": 3824 }, { "epoch": 0.41, "learning_rate": 6.6990034183759726e-06, "loss": 1.5942, "step": 3825 }, { "epoch": 0.41, "learning_rate": 6.6973786277118026e-06, "loss": 1.4495, "step": 3826 }, { "epoch": 0.41, "learning_rate": 6.695753634436134e-06, "loss": 1.2507, "step": 3827 }, { "epoch": 0.41, "learning_rate": 6.694128438742939e-06, "loss": 1.5539, "step": 3828 }, { "epoch": 0.41, "learning_rate": 6.692503040826211e-06, "loss": 1.3886, "step": 3829 }, { "epoch": 0.41, "learning_rate": 6.690877440879971e-06, "loss": 1.4462, "step": 3830 }, { "epoch": 0.41, "learning_rate": 6.689251639098261e-06, "loss": 1.1469, "step": 3831 }, { "epoch": 0.41, "learning_rate": 6.68762563567515e-06, "loss": 1.141, "step": 3832 }, { "epoch": 0.41, "learning_rate": 6.685999430804729e-06, "loss": 1.3181, "step": 3833 }, { "epoch": 0.41, "learning_rate": 6.684373024681112e-06, "loss": 1.5545, "step": 3834 }, { "epoch": 0.41, "learning_rate": 6.6827464174984405e-06, "loss": 1.2799, "step": 3835 }, { "epoch": 0.41, "learning_rate": 6.681119609450879e-06, "loss": 1.6436, "step": 3836 }, { "epoch": 0.41, "learning_rate": 6.679492600732614e-06, "loss": 1.6489, "step": 3837 }, { "epoch": 0.41, "learning_rate": 6.677865391537857e-06, "loss": 1.1673, "step": 3838 }, { "epoch": 0.41, "learning_rate": 6.676237982060844e-06, "loss": 1.4151, "step": 3839 }, { "epoch": 0.41, "learning_rate": 6.674610372495832e-06, "loss": 1.5092, "step": 3840 }, { "epoch": 0.41, "learning_rate": 6.67298256303711e-06, "loss": 1.4257, "step": 3841 }, { "epoch": 0.41, "learning_rate": 6.67135455387898e-06, "loss": 1.7112, "step": 3842 }, { "epoch": 0.41, "learning_rate": 6.669726345215776e-06, "loss": 1.4294, "step": 3843 }, { "epoch": 0.41, "learning_rate": 6.668097937241852e-06, "loss": 1.4499, "step": 3844 }, { "epoch": 0.41, "learning_rate": 6.666469330151585e-06, "loss": 1.6623, "step": 3845 }, { "epoch": 0.41, "learning_rate": 6.66484052413938e-06, "loss": 1.4615, "step": 3846 }, { "epoch": 0.41, "learning_rate": 6.663211519399661e-06, "loss": 1.4505, "step": 3847 }, { "epoch": 0.41, "learning_rate": 6.66158231612688e-06, "loss": 1.2527, "step": 3848 }, { "epoch": 0.41, "learning_rate": 6.659952914515508e-06, "loss": 1.2176, "step": 3849 }, { "epoch": 0.41, "learning_rate": 6.658323314760046e-06, "loss": 1.4148, "step": 3850 }, { "epoch": 0.41, "learning_rate": 6.656693517055011e-06, "loss": 1.4804, "step": 3851 }, { "epoch": 0.41, "learning_rate": 6.65506352159495e-06, "loss": 1.5269, "step": 3852 }, { "epoch": 0.41, "learning_rate": 6.65343332857443e-06, "loss": 1.322, "step": 3853 }, { "epoch": 0.41, "learning_rate": 6.651802938188044e-06, "loss": 1.6185, "step": 3854 }, { "epoch": 0.41, "learning_rate": 6.650172350630406e-06, "loss": 1.5144, "step": 3855 }, { "epoch": 0.41, "learning_rate": 6.648541566096155e-06, "loss": 1.529, "step": 3856 }, { "epoch": 0.41, "learning_rate": 6.646910584779953e-06, "loss": 1.3795, "step": 3857 }, { "epoch": 0.41, "learning_rate": 6.645279406876488e-06, "loss": 1.3539, "step": 3858 }, { "epoch": 0.41, "learning_rate": 6.643648032580466e-06, "loss": 1.4709, "step": 3859 }, { "epoch": 0.41, "learning_rate": 6.6420164620866226e-06, "loss": 1.2874, "step": 3860 }, { "epoch": 0.41, "learning_rate": 6.640384695589714e-06, "loss": 1.606, "step": 3861 }, { "epoch": 0.41, "learning_rate": 6.638752733284514e-06, "loss": 1.6803, "step": 3862 }, { "epoch": 0.41, "learning_rate": 6.637120575365834e-06, "loss": 1.3686, "step": 3863 }, { "epoch": 0.41, "learning_rate": 6.635488222028497e-06, "loss": 1.5782, "step": 3864 }, { "epoch": 0.41, "learning_rate": 6.63385567346735e-06, "loss": 1.4529, "step": 3865 }, { "epoch": 0.41, "learning_rate": 6.632222929877268e-06, "loss": 1.4373, "step": 3866 }, { "epoch": 0.41, "learning_rate": 6.630589991453148e-06, "loss": 1.4767, "step": 3867 }, { "epoch": 0.41, "learning_rate": 6.6289568583899074e-06, "loss": 1.4968, "step": 3868 }, { "epoch": 0.41, "learning_rate": 6.627323530882491e-06, "loss": 1.4529, "step": 3869 }, { "epoch": 0.41, "learning_rate": 6.6256900091258644e-06, "loss": 1.4891, "step": 3870 }, { "epoch": 0.41, "learning_rate": 6.624056293315015e-06, "loss": 1.6678, "step": 3871 }, { "epoch": 0.41, "learning_rate": 6.622422383644956e-06, "loss": 1.5731, "step": 3872 }, { "epoch": 0.41, "learning_rate": 6.620788280310722e-06, "loss": 1.3768, "step": 3873 }, { "epoch": 0.41, "learning_rate": 6.619153983507373e-06, "loss": 1.439, "step": 3874 }, { "epoch": 0.41, "learning_rate": 6.617519493429991e-06, "loss": 1.4167, "step": 3875 }, { "epoch": 0.41, "learning_rate": 6.615884810273678e-06, "loss": 1.4693, "step": 3876 }, { "epoch": 0.41, "learning_rate": 6.614249934233565e-06, "loss": 1.6713, "step": 3877 }, { "epoch": 0.41, "learning_rate": 6.6126148655048e-06, "loss": 1.6665, "step": 3878 }, { "epoch": 0.41, "learning_rate": 6.610979604282557e-06, "loss": 1.3981, "step": 3879 }, { "epoch": 0.41, "learning_rate": 6.609344150762035e-06, "loss": 1.2944, "step": 3880 }, { "epoch": 0.41, "learning_rate": 6.60770850513845e-06, "loss": 1.2086, "step": 3881 }, { "epoch": 0.41, "learning_rate": 6.606072667607048e-06, "loss": 1.3776, "step": 3882 }, { "epoch": 0.41, "learning_rate": 6.604436638363094e-06, "loss": 1.4254, "step": 3883 }, { "epoch": 0.41, "learning_rate": 6.602800417601874e-06, "loss": 1.2296, "step": 3884 }, { "epoch": 0.41, "learning_rate": 6.601164005518702e-06, "loss": 1.3288, "step": 3885 }, { "epoch": 0.41, "learning_rate": 6.5995274023089105e-06, "loss": 1.5402, "step": 3886 }, { "epoch": 0.41, "learning_rate": 6.597890608167856e-06, "loss": 1.3485, "step": 3887 }, { "epoch": 0.41, "learning_rate": 6.59625362329092e-06, "loss": 1.3959, "step": 3888 }, { "epoch": 0.41, "learning_rate": 6.594616447873506e-06, "loss": 1.4547, "step": 3889 }, { "epoch": 0.41, "learning_rate": 6.592979082111034e-06, "loss": 1.4754, "step": 3890 }, { "epoch": 0.42, "learning_rate": 6.591341526198955e-06, "loss": 1.3259, "step": 3891 }, { "epoch": 0.42, "learning_rate": 6.589703780332742e-06, "loss": 1.2665, "step": 3892 }, { "epoch": 0.42, "learning_rate": 6.588065844707885e-06, "loss": 1.2757, "step": 3893 }, { "epoch": 0.42, "learning_rate": 6.586427719519901e-06, "loss": 1.2416, "step": 3894 }, { "epoch": 0.42, "learning_rate": 6.584789404964329e-06, "loss": 1.4294, "step": 3895 }, { "epoch": 0.42, "learning_rate": 6.5831509012367275e-06, "loss": 1.5613, "step": 3896 }, { "epoch": 0.42, "learning_rate": 6.581512208532685e-06, "loss": 1.5218, "step": 3897 }, { "epoch": 0.42, "learning_rate": 6.579873327047805e-06, "loss": 1.5054, "step": 3898 }, { "epoch": 0.42, "learning_rate": 6.5782342569777155e-06, "loss": 1.4733, "step": 3899 }, { "epoch": 0.42, "learning_rate": 6.576594998518071e-06, "loss": 1.6112, "step": 3900 }, { "epoch": 0.42, "learning_rate": 6.574955551864541e-06, "loss": 1.2172, "step": 3901 }, { "epoch": 0.42, "learning_rate": 6.573315917212826e-06, "loss": 1.5064, "step": 3902 }, { "epoch": 0.42, "learning_rate": 6.5716760947586425e-06, "loss": 1.4414, "step": 3903 }, { "epoch": 0.42, "learning_rate": 6.570036084697732e-06, "loss": 1.6261, "step": 3904 }, { "epoch": 0.42, "learning_rate": 6.568395887225858e-06, "loss": 1.3852, "step": 3905 }, { "epoch": 0.42, "learning_rate": 6.566755502538806e-06, "loss": 1.4365, "step": 3906 }, { "epoch": 0.42, "learning_rate": 6.565114930832387e-06, "loss": 1.1956, "step": 3907 }, { "epoch": 0.42, "learning_rate": 6.563474172302429e-06, "loss": 1.5269, "step": 3908 }, { "epoch": 0.42, "learning_rate": 6.561833227144784e-06, "loss": 1.4636, "step": 3909 }, { "epoch": 0.42, "learning_rate": 6.5601920955553285e-06, "loss": 1.5516, "step": 3910 }, { "epoch": 0.42, "learning_rate": 6.558550777729962e-06, "loss": 1.338, "step": 3911 }, { "epoch": 0.42, "learning_rate": 6.556909273864601e-06, "loss": 1.3183, "step": 3912 }, { "epoch": 0.42, "learning_rate": 6.555267584155188e-06, "loss": 1.4838, "step": 3913 }, { "epoch": 0.42, "learning_rate": 6.55362570879769e-06, "loss": 1.366, "step": 3914 }, { "epoch": 0.42, "learning_rate": 6.551983647988089e-06, "loss": 1.4815, "step": 3915 }, { "epoch": 0.42, "learning_rate": 6.550341401922395e-06, "loss": 1.3344, "step": 3916 }, { "epoch": 0.42, "learning_rate": 6.548698970796641e-06, "loss": 1.3892, "step": 3917 }, { "epoch": 0.42, "learning_rate": 6.547056354806874e-06, "loss": 1.206, "step": 3918 }, { "epoch": 0.42, "learning_rate": 6.545413554149175e-06, "loss": 1.2673, "step": 3919 }, { "epoch": 0.42, "learning_rate": 6.543770569019635e-06, "loss": 1.3924, "step": 3920 }, { "epoch": 0.42, "learning_rate": 6.542127399614376e-06, "loss": 1.4176, "step": 3921 }, { "epoch": 0.42, "learning_rate": 6.54048404612954e-06, "loss": 1.5844, "step": 3922 }, { "epoch": 0.42, "learning_rate": 6.538840508761286e-06, "loss": 1.5148, "step": 3923 }, { "epoch": 0.42, "learning_rate": 6.5371967877058e-06, "loss": 1.5854, "step": 3924 }, { "epoch": 0.42, "learning_rate": 6.53555288315929e-06, "loss": 1.6146, "step": 3925 }, { "epoch": 0.42, "learning_rate": 6.533908795317982e-06, "loss": 1.3026, "step": 3926 }, { "epoch": 0.42, "learning_rate": 6.532264524378128e-06, "loss": 1.4553, "step": 3927 }, { "epoch": 0.42, "learning_rate": 6.530620070536e-06, "loss": 1.229, "step": 3928 }, { "epoch": 0.42, "learning_rate": 6.528975433987892e-06, "loss": 1.4986, "step": 3929 }, { "epoch": 0.42, "learning_rate": 6.52733061493012e-06, "loss": 1.6122, "step": 3930 }, { "epoch": 0.42, "learning_rate": 6.52568561355902e-06, "loss": 1.2967, "step": 3931 }, { "epoch": 0.42, "learning_rate": 6.524040430070953e-06, "loss": 1.3561, "step": 3932 }, { "epoch": 0.42, "learning_rate": 6.522395064662299e-06, "loss": 1.469, "step": 3933 }, { "epoch": 0.42, "learning_rate": 6.520749517529463e-06, "loss": 1.501, "step": 3934 }, { "epoch": 0.42, "learning_rate": 6.519103788868866e-06, "loss": 1.3954, "step": 3935 }, { "epoch": 0.42, "learning_rate": 6.517457878876958e-06, "loss": 1.307, "step": 3936 }, { "epoch": 0.42, "learning_rate": 6.515811787750204e-06, "loss": 1.2166, "step": 3937 }, { "epoch": 0.42, "learning_rate": 6.5141655156850936e-06, "loss": 1.6027, "step": 3938 }, { "epoch": 0.42, "learning_rate": 6.512519062878142e-06, "loss": 1.4576, "step": 3939 }, { "epoch": 0.42, "learning_rate": 6.510872429525875e-06, "loss": 1.2588, "step": 3940 }, { "epoch": 0.42, "learning_rate": 6.509225615824851e-06, "loss": 1.4609, "step": 3941 }, { "epoch": 0.42, "learning_rate": 6.507578621971646e-06, "loss": 1.2519, "step": 3942 }, { "epoch": 0.42, "learning_rate": 6.505931448162857e-06, "loss": 1.4664, "step": 3943 }, { "epoch": 0.42, "learning_rate": 6.504284094595099e-06, "loss": 1.5311, "step": 3944 }, { "epoch": 0.42, "learning_rate": 6.502636561465018e-06, "loss": 1.3908, "step": 3945 }, { "epoch": 0.42, "learning_rate": 6.500988848969271e-06, "loss": 1.3859, "step": 3946 }, { "epoch": 0.42, "learning_rate": 6.499340957304543e-06, "loss": 1.4496, "step": 3947 }, { "epoch": 0.42, "learning_rate": 6.497692886667537e-06, "loss": 1.5528, "step": 3948 }, { "epoch": 0.42, "learning_rate": 6.496044637254982e-06, "loss": 1.5637, "step": 3949 }, { "epoch": 0.42, "learning_rate": 6.4943962092636205e-06, "loss": 1.3796, "step": 3950 }, { "epoch": 0.42, "learning_rate": 6.492747602890223e-06, "loss": 1.3885, "step": 3951 }, { "epoch": 0.42, "learning_rate": 6.4910988183315805e-06, "loss": 1.6607, "step": 3952 }, { "epoch": 0.42, "learning_rate": 6.4894498557845035e-06, "loss": 1.5629, "step": 3953 }, { "epoch": 0.42, "learning_rate": 6.487800715445822e-06, "loss": 1.5875, "step": 3954 }, { "epoch": 0.42, "learning_rate": 6.486151397512391e-06, "loss": 1.4794, "step": 3955 }, { "epoch": 0.42, "learning_rate": 6.484501902181086e-06, "loss": 1.3628, "step": 3956 }, { "epoch": 0.42, "learning_rate": 6.4828522296488014e-06, "loss": 1.4952, "step": 3957 }, { "epoch": 0.42, "learning_rate": 6.481202380112454e-06, "loss": 1.3456, "step": 3958 }, { "epoch": 0.42, "learning_rate": 6.479552353768986e-06, "loss": 1.4307, "step": 3959 }, { "epoch": 0.42, "learning_rate": 6.477902150815347e-06, "loss": 1.2149, "step": 3960 }, { "epoch": 0.42, "learning_rate": 6.4762517714485275e-06, "loss": 1.3251, "step": 3961 }, { "epoch": 0.42, "learning_rate": 6.474601215865523e-06, "loss": 1.3709, "step": 3962 }, { "epoch": 0.42, "learning_rate": 6.472950484263359e-06, "loss": 1.7505, "step": 3963 }, { "epoch": 0.42, "learning_rate": 6.471299576839076e-06, "loss": 1.6589, "step": 3964 }, { "epoch": 0.42, "learning_rate": 6.469648493789741e-06, "loss": 1.177, "step": 3965 }, { "epoch": 0.42, "learning_rate": 6.467997235312437e-06, "loss": 1.5307, "step": 3966 }, { "epoch": 0.42, "learning_rate": 6.466345801604272e-06, "loss": 1.409, "step": 3967 }, { "epoch": 0.42, "learning_rate": 6.464694192862372e-06, "loss": 1.5388, "step": 3968 }, { "epoch": 0.42, "learning_rate": 6.463042409283885e-06, "loss": 1.5174, "step": 3969 }, { "epoch": 0.42, "learning_rate": 6.461390451065982e-06, "loss": 1.6847, "step": 3970 }, { "epoch": 0.42, "learning_rate": 6.45973831840585e-06, "loss": 1.5064, "step": 3971 }, { "epoch": 0.42, "learning_rate": 6.458086011500703e-06, "loss": 1.6455, "step": 3972 }, { "epoch": 0.42, "learning_rate": 6.456433530547768e-06, "loss": 1.6032, "step": 3973 }, { "epoch": 0.42, "learning_rate": 6.4547808757442995e-06, "loss": 1.453, "step": 3974 }, { "epoch": 0.42, "learning_rate": 6.453128047287573e-06, "loss": 1.3537, "step": 3975 }, { "epoch": 0.42, "learning_rate": 6.451475045374877e-06, "loss": 1.6403, "step": 3976 }, { "epoch": 0.42, "learning_rate": 6.449821870203529e-06, "loss": 1.4927, "step": 3977 }, { "epoch": 0.42, "learning_rate": 6.448168521970865e-06, "loss": 1.5647, "step": 3978 }, { "epoch": 0.42, "learning_rate": 6.4465150008742365e-06, "loss": 1.3558, "step": 3979 }, { "epoch": 0.42, "learning_rate": 6.444861307111023e-06, "loss": 1.505, "step": 3980 }, { "epoch": 0.42, "learning_rate": 6.443207440878624e-06, "loss": 1.731, "step": 3981 }, { "epoch": 0.42, "learning_rate": 6.441553402374451e-06, "loss": 1.6755, "step": 3982 }, { "epoch": 0.42, "learning_rate": 6.439899191795946e-06, "loss": 1.2069, "step": 3983 }, { "epoch": 0.42, "learning_rate": 6.438244809340568e-06, "loss": 1.5029, "step": 3984 }, { "epoch": 0.43, "learning_rate": 6.4365902552057945e-06, "loss": 1.477, "step": 3985 }, { "epoch": 0.43, "learning_rate": 6.4349355295891246e-06, "loss": 1.3265, "step": 3986 }, { "epoch": 0.43, "learning_rate": 6.43328063268808e-06, "loss": 1.3969, "step": 3987 }, { "epoch": 0.43, "learning_rate": 6.431625564700201e-06, "loss": 1.1782, "step": 3988 }, { "epoch": 0.43, "learning_rate": 6.4299703258230485e-06, "loss": 1.3825, "step": 3989 }, { "epoch": 0.43, "learning_rate": 6.428314916254203e-06, "loss": 1.6097, "step": 3990 }, { "epoch": 0.43, "learning_rate": 6.4266593361912665e-06, "loss": 1.6476, "step": 3991 }, { "epoch": 0.43, "learning_rate": 6.4250035858318635e-06, "loss": 1.2454, "step": 3992 }, { "epoch": 0.43, "learning_rate": 6.423347665373633e-06, "loss": 1.2962, "step": 3993 }, { "epoch": 0.43, "learning_rate": 6.42169157501424e-06, "loss": 1.4168, "step": 3994 }, { "epoch": 0.43, "learning_rate": 6.4200353149513674e-06, "loss": 1.2721, "step": 3995 }, { "epoch": 0.43, "learning_rate": 6.418378885382716e-06, "loss": 1.5403, "step": 3996 }, { "epoch": 0.43, "learning_rate": 6.416722286506013e-06, "loss": 1.4498, "step": 3997 }, { "epoch": 0.43, "learning_rate": 6.415065518519e-06, "loss": 1.3738, "step": 3998 }, { "epoch": 0.43, "learning_rate": 6.41340858161944e-06, "loss": 1.3094, "step": 3999 }, { "epoch": 0.43, "learning_rate": 6.411751476005119e-06, "loss": 1.3021, "step": 4000 }, { "epoch": 0.43, "learning_rate": 6.410094201873842e-06, "loss": 1.2991, "step": 4001 }, { "epoch": 0.43, "learning_rate": 6.408436759423431e-06, "loss": 1.5784, "step": 4002 }, { "epoch": 0.43, "learning_rate": 6.406779148851731e-06, "loss": 1.4466, "step": 4003 }, { "epoch": 0.43, "learning_rate": 6.405121370356606e-06, "loss": 1.388, "step": 4004 }, { "epoch": 0.43, "learning_rate": 6.403463424135943e-06, "loss": 1.3748, "step": 4005 }, { "epoch": 0.43, "learning_rate": 6.401805310387644e-06, "loss": 1.5453, "step": 4006 }, { "epoch": 0.43, "learning_rate": 6.400147029309635e-06, "loss": 1.3748, "step": 4007 }, { "epoch": 0.43, "learning_rate": 6.398488581099859e-06, "loss": 1.4551, "step": 4008 }, { "epoch": 0.43, "learning_rate": 6.396829965956283e-06, "loss": 1.5362, "step": 4009 }, { "epoch": 0.43, "learning_rate": 6.395171184076889e-06, "loss": 1.1408, "step": 4010 }, { "epoch": 0.43, "learning_rate": 6.393512235659681e-06, "loss": 1.4053, "step": 4011 }, { "epoch": 0.43, "learning_rate": 6.391853120902687e-06, "loss": 1.4794, "step": 4012 }, { "epoch": 0.43, "learning_rate": 6.390193840003945e-06, "loss": 1.6041, "step": 4013 }, { "epoch": 0.43, "learning_rate": 6.388534393161525e-06, "loss": 1.4896, "step": 4014 }, { "epoch": 0.43, "learning_rate": 6.3868747805735084e-06, "loss": 1.2779, "step": 4015 }, { "epoch": 0.43, "learning_rate": 6.385215002437997e-06, "loss": 1.4425, "step": 4016 }, { "epoch": 0.43, "learning_rate": 6.383555058953115e-06, "loss": 1.2672, "step": 4017 }, { "epoch": 0.43, "learning_rate": 6.381894950317006e-06, "loss": 1.5415, "step": 4018 }, { "epoch": 0.43, "learning_rate": 6.380234676727832e-06, "loss": 1.2942, "step": 4019 }, { "epoch": 0.43, "learning_rate": 6.378574238383776e-06, "loss": 1.5298, "step": 4020 }, { "epoch": 0.43, "learning_rate": 6.3769136354830395e-06, "loss": 1.5064, "step": 4021 }, { "epoch": 0.43, "learning_rate": 6.375252868223845e-06, "loss": 1.4571, "step": 4022 }, { "epoch": 0.43, "learning_rate": 6.373591936804433e-06, "loss": 1.5929, "step": 4023 }, { "epoch": 0.43, "learning_rate": 6.3719308414230645e-06, "loss": 1.512, "step": 4024 }, { "epoch": 0.43, "learning_rate": 6.3702695822780206e-06, "loss": 1.2492, "step": 4025 }, { "epoch": 0.43, "learning_rate": 6.3686081595676e-06, "loss": 1.5901, "step": 4026 }, { "epoch": 0.43, "learning_rate": 6.366946573490124e-06, "loss": 1.5823, "step": 4027 }, { "epoch": 0.43, "learning_rate": 6.365284824243929e-06, "loss": 1.4772, "step": 4028 }, { "epoch": 0.43, "learning_rate": 6.3636229120273766e-06, "loss": 1.1458, "step": 4029 }, { "epoch": 0.43, "learning_rate": 6.361960837038842e-06, "loss": 1.4031, "step": 4030 }, { "epoch": 0.43, "learning_rate": 6.360298599476727e-06, "loss": 1.5468, "step": 4031 }, { "epoch": 0.43, "learning_rate": 6.3586361995394415e-06, "loss": 1.5611, "step": 4032 }, { "epoch": 0.43, "learning_rate": 6.356973637425427e-06, "loss": 1.5196, "step": 4033 }, { "epoch": 0.43, "learning_rate": 6.355310913333139e-06, "loss": 1.3107, "step": 4034 }, { "epoch": 0.43, "learning_rate": 6.353648027461048e-06, "loss": 1.2874, "step": 4035 }, { "epoch": 0.43, "learning_rate": 6.351984980007653e-06, "loss": 1.3171, "step": 4036 }, { "epoch": 0.43, "learning_rate": 6.350321771171467e-06, "loss": 1.5465, "step": 4037 }, { "epoch": 0.43, "learning_rate": 6.348658401151018e-06, "loss": 1.2824, "step": 4038 }, { "epoch": 0.43, "learning_rate": 6.3469948701448644e-06, "loss": 1.3969, "step": 4039 }, { "epoch": 0.43, "learning_rate": 6.345331178351572e-06, "loss": 1.4068, "step": 4040 }, { "epoch": 0.43, "learning_rate": 6.343667325969736e-06, "loss": 1.3513, "step": 4041 }, { "epoch": 0.43, "learning_rate": 6.342003313197962e-06, "loss": 1.3935, "step": 4042 }, { "epoch": 0.43, "learning_rate": 6.340339140234882e-06, "loss": 1.2262, "step": 4043 }, { "epoch": 0.43, "learning_rate": 6.3386748072791395e-06, "loss": 1.1174, "step": 4044 }, { "epoch": 0.43, "learning_rate": 6.3370103145294045e-06, "loss": 1.4097, "step": 4045 }, { "epoch": 0.43, "learning_rate": 6.3353456621843635e-06, "loss": 1.5593, "step": 4046 }, { "epoch": 0.43, "learning_rate": 6.33368085044272e-06, "loss": 1.3995, "step": 4047 }, { "epoch": 0.43, "learning_rate": 6.332015879503198e-06, "loss": 1.164, "step": 4048 }, { "epoch": 0.43, "learning_rate": 6.330350749564542e-06, "loss": 1.6159, "step": 4049 }, { "epoch": 0.43, "learning_rate": 6.328685460825512e-06, "loss": 1.3788, "step": 4050 }, { "epoch": 0.43, "learning_rate": 6.327020013484893e-06, "loss": 1.478, "step": 4051 }, { "epoch": 0.43, "learning_rate": 6.32535440774148e-06, "loss": 1.4384, "step": 4052 }, { "epoch": 0.43, "learning_rate": 6.323688643794094e-06, "loss": 1.4012, "step": 4053 }, { "epoch": 0.43, "learning_rate": 6.322022721841574e-06, "loss": 1.6839, "step": 4054 }, { "epoch": 0.43, "learning_rate": 6.320356642082774e-06, "loss": 1.502, "step": 4055 }, { "epoch": 0.43, "learning_rate": 6.318690404716572e-06, "loss": 1.3035, "step": 4056 }, { "epoch": 0.43, "learning_rate": 6.31702400994186e-06, "loss": 1.5002, "step": 4057 }, { "epoch": 0.43, "learning_rate": 6.315357457957553e-06, "loss": 1.5911, "step": 4058 }, { "epoch": 0.43, "learning_rate": 6.313690748962582e-06, "loss": 1.4071, "step": 4059 }, { "epoch": 0.43, "learning_rate": 6.312023883155897e-06, "loss": 1.4036, "step": 4060 }, { "epoch": 0.43, "learning_rate": 6.310356860736468e-06, "loss": 1.3446, "step": 4061 }, { "epoch": 0.43, "learning_rate": 6.3086896819032814e-06, "loss": 1.7202, "step": 4062 }, { "epoch": 0.43, "learning_rate": 6.307022346855347e-06, "loss": 1.5003, "step": 4063 }, { "epoch": 0.43, "learning_rate": 6.305354855791688e-06, "loss": 1.4476, "step": 4064 }, { "epoch": 0.43, "learning_rate": 6.303687208911348e-06, "loss": 1.2579, "step": 4065 }, { "epoch": 0.43, "learning_rate": 6.30201940641339e-06, "loss": 1.5155, "step": 4066 }, { "epoch": 0.43, "learning_rate": 6.300351448496894e-06, "loss": 1.4121, "step": 4067 }, { "epoch": 0.43, "learning_rate": 6.298683335360962e-06, "loss": 1.2598, "step": 4068 }, { "epoch": 0.43, "learning_rate": 6.2970150672047115e-06, "loss": 1.2488, "step": 4069 }, { "epoch": 0.43, "learning_rate": 6.295346644227278e-06, "loss": 1.6346, "step": 4070 }, { "epoch": 0.43, "learning_rate": 6.293678066627816e-06, "loss": 1.4271, "step": 4071 }, { "epoch": 0.43, "learning_rate": 6.292009334605501e-06, "loss": 1.2467, "step": 4072 }, { "epoch": 0.43, "learning_rate": 6.290340448359526e-06, "loss": 1.455, "step": 4073 }, { "epoch": 0.43, "learning_rate": 6.288671408089098e-06, "loss": 1.5068, "step": 4074 }, { "epoch": 0.43, "learning_rate": 6.287002213993446e-06, "loss": 1.3651, "step": 4075 }, { "epoch": 0.43, "learning_rate": 6.2853328662718215e-06, "loss": 1.3644, "step": 4076 }, { "epoch": 0.43, "learning_rate": 6.283663365123486e-06, "loss": 1.4694, "step": 4077 }, { "epoch": 0.43, "learning_rate": 6.281993710747725e-06, "loss": 1.4074, "step": 4078 }, { "epoch": 0.44, "learning_rate": 6.280323903343839e-06, "loss": 1.56, "step": 4079 }, { "epoch": 0.44, "learning_rate": 6.278653943111152e-06, "loss": 1.3908, "step": 4080 }, { "epoch": 0.44, "learning_rate": 6.276983830248998e-06, "loss": 1.4032, "step": 4081 }, { "epoch": 0.44, "learning_rate": 6.275313564956738e-06, "loss": 1.5719, "step": 4082 }, { "epoch": 0.44, "learning_rate": 6.273643147433743e-06, "loss": 1.4142, "step": 4083 }, { "epoch": 0.44, "learning_rate": 6.27197257787941e-06, "loss": 1.3288, "step": 4084 }, { "epoch": 0.44, "learning_rate": 6.270301856493149e-06, "loss": 1.3632, "step": 4085 }, { "epoch": 0.44, "learning_rate": 6.268630983474388e-06, "loss": 1.1742, "step": 4086 }, { "epoch": 0.44, "learning_rate": 6.266959959022578e-06, "loss": 1.3191, "step": 4087 }, { "epoch": 0.44, "learning_rate": 6.265288783337181e-06, "loss": 1.2103, "step": 4088 }, { "epoch": 0.44, "learning_rate": 6.263617456617681e-06, "loss": 1.3129, "step": 4089 }, { "epoch": 0.44, "learning_rate": 6.2619459790635835e-06, "loss": 1.3632, "step": 4090 }, { "epoch": 0.44, "learning_rate": 6.260274350874406e-06, "loss": 1.4502, "step": 4091 }, { "epoch": 0.44, "learning_rate": 6.258602572249683e-06, "loss": 1.5115, "step": 4092 }, { "epoch": 0.44, "learning_rate": 6.256930643388977e-06, "loss": 1.4356, "step": 4093 }, { "epoch": 0.44, "learning_rate": 6.255258564491854e-06, "loss": 1.3269, "step": 4094 }, { "epoch": 0.44, "learning_rate": 6.2535863357579105e-06, "loss": 1.4849, "step": 4095 }, { "epoch": 0.44, "learning_rate": 6.251913957386757e-06, "loss": 1.6521, "step": 4096 }, { "epoch": 0.44, "learning_rate": 6.250241429578017e-06, "loss": 1.6175, "step": 4097 }, { "epoch": 0.44, "learning_rate": 6.248568752531337e-06, "loss": 1.4029, "step": 4098 }, { "epoch": 0.44, "learning_rate": 6.2468959264463805e-06, "loss": 1.4317, "step": 4099 }, { "epoch": 0.44, "learning_rate": 6.2452229515228276e-06, "loss": 1.2357, "step": 4100 }, { "epoch": 0.44, "learning_rate": 6.243549827960378e-06, "loss": 1.3996, "step": 4101 }, { "epoch": 0.44, "learning_rate": 6.241876555958746e-06, "loss": 1.6554, "step": 4102 }, { "epoch": 0.44, "learning_rate": 6.240203135717667e-06, "loss": 1.4179, "step": 4103 }, { "epoch": 0.44, "learning_rate": 6.238529567436892e-06, "loss": 1.5059, "step": 4104 }, { "epoch": 0.44, "learning_rate": 6.23685585131619e-06, "loss": 1.3854, "step": 4105 }, { "epoch": 0.44, "learning_rate": 6.235181987555349e-06, "loss": 1.5194, "step": 4106 }, { "epoch": 0.44, "learning_rate": 6.233507976354174e-06, "loss": 1.518, "step": 4107 }, { "epoch": 0.44, "learning_rate": 6.231833817912486e-06, "loss": 1.3245, "step": 4108 }, { "epoch": 0.44, "learning_rate": 6.230159512430124e-06, "loss": 1.3999, "step": 4109 }, { "epoch": 0.44, "learning_rate": 6.228485060106948e-06, "loss": 1.664, "step": 4110 }, { "epoch": 0.44, "learning_rate": 6.226810461142829e-06, "loss": 1.4092, "step": 4111 }, { "epoch": 0.44, "learning_rate": 6.225135715737662e-06, "loss": 1.3682, "step": 4112 }, { "epoch": 0.44, "learning_rate": 6.223460824091358e-06, "loss": 1.3937, "step": 4113 }, { "epoch": 0.44, "learning_rate": 6.221785786403843e-06, "loss": 1.5911, "step": 4114 }, { "epoch": 0.44, "learning_rate": 6.2201106028750605e-06, "loss": 1.8099, "step": 4115 }, { "epoch": 0.44, "learning_rate": 6.218435273704973e-06, "loss": 1.3205, "step": 4116 }, { "epoch": 0.44, "learning_rate": 6.216759799093561e-06, "loss": 1.609, "step": 4117 }, { "epoch": 0.44, "learning_rate": 6.215084179240821e-06, "loss": 1.5115, "step": 4118 }, { "epoch": 0.44, "learning_rate": 6.213408414346765e-06, "loss": 1.4672, "step": 4119 }, { "epoch": 0.44, "learning_rate": 6.211732504611429e-06, "loss": 1.2204, "step": 4120 }, { "epoch": 0.44, "learning_rate": 6.21005645023486e-06, "loss": 1.3066, "step": 4121 }, { "epoch": 0.44, "learning_rate": 6.208380251417122e-06, "loss": 1.253, "step": 4122 }, { "epoch": 0.44, "learning_rate": 6.206703908358301e-06, "loss": 1.7477, "step": 4123 }, { "epoch": 0.44, "learning_rate": 6.205027421258496e-06, "loss": 1.5021, "step": 4124 }, { "epoch": 0.44, "learning_rate": 6.203350790317825e-06, "loss": 1.3617, "step": 4125 }, { "epoch": 0.44, "learning_rate": 6.201674015736423e-06, "loss": 1.4475, "step": 4126 }, { "epoch": 0.44, "learning_rate": 6.199997097714443e-06, "loss": 1.7382, "step": 4127 }, { "epoch": 0.44, "learning_rate": 6.198320036452051e-06, "loss": 1.3363, "step": 4128 }, { "epoch": 0.44, "learning_rate": 6.196642832149439e-06, "loss": 1.6098, "step": 4129 }, { "epoch": 0.44, "learning_rate": 6.194965485006805e-06, "loss": 1.4245, "step": 4130 }, { "epoch": 0.44, "learning_rate": 6.193287995224371e-06, "loss": 1.5436, "step": 4131 }, { "epoch": 0.44, "learning_rate": 6.191610363002376e-06, "loss": 1.4439, "step": 4132 }, { "epoch": 0.44, "learning_rate": 6.189932588541071e-06, "loss": 1.4857, "step": 4133 }, { "epoch": 0.44, "learning_rate": 6.18825467204073e-06, "loss": 1.6356, "step": 4134 }, { "epoch": 0.44, "learning_rate": 6.186576613701644e-06, "loss": 1.3431, "step": 4135 }, { "epoch": 0.44, "learning_rate": 6.184898413724111e-06, "loss": 1.141, "step": 4136 }, { "epoch": 0.44, "learning_rate": 6.183220072308459e-06, "loss": 1.7555, "step": 4137 }, { "epoch": 0.44, "learning_rate": 6.181541589655026e-06, "loss": 1.835, "step": 4138 }, { "epoch": 0.44, "learning_rate": 6.1798629659641676e-06, "loss": 1.2851, "step": 4139 }, { "epoch": 0.44, "learning_rate": 6.178184201436256e-06, "loss": 1.5158, "step": 4140 }, { "epoch": 0.44, "learning_rate": 6.1765052962716805e-06, "loss": 1.5954, "step": 4141 }, { "epoch": 0.44, "learning_rate": 6.174826250670849e-06, "loss": 1.4521, "step": 4142 }, { "epoch": 0.44, "learning_rate": 6.173147064834183e-06, "loss": 1.4723, "step": 4143 }, { "epoch": 0.44, "learning_rate": 6.171467738962123e-06, "loss": 1.3779, "step": 4144 }, { "epoch": 0.44, "learning_rate": 6.169788273255126e-06, "loss": 1.2114, "step": 4145 }, { "epoch": 0.44, "learning_rate": 6.168108667913666e-06, "loss": 1.5309, "step": 4146 }, { "epoch": 0.44, "learning_rate": 6.1664289231382315e-06, "loss": 1.6634, "step": 4147 }, { "epoch": 0.44, "learning_rate": 6.164749039129328e-06, "loss": 1.5761, "step": 4148 }, { "epoch": 0.44, "learning_rate": 6.163069016087483e-06, "loss": 1.4506, "step": 4149 }, { "epoch": 0.44, "learning_rate": 6.161388854213231e-06, "loss": 1.3837, "step": 4150 }, { "epoch": 0.44, "learning_rate": 6.159708553707131e-06, "loss": 1.3215, "step": 4151 }, { "epoch": 0.44, "learning_rate": 6.158028114769758e-06, "loss": 1.6809, "step": 4152 }, { "epoch": 0.44, "learning_rate": 6.156347537601698e-06, "loss": 1.5322, "step": 4153 }, { "epoch": 0.44, "learning_rate": 6.154666822403558e-06, "loss": 1.187, "step": 4154 }, { "epoch": 0.44, "learning_rate": 6.152985969375962e-06, "loss": 1.641, "step": 4155 }, { "epoch": 0.44, "learning_rate": 6.151304978719546e-06, "loss": 1.7015, "step": 4156 }, { "epoch": 0.44, "learning_rate": 6.149623850634968e-06, "loss": 1.6283, "step": 4157 }, { "epoch": 0.44, "learning_rate": 6.147942585322898e-06, "loss": 1.2737, "step": 4158 }, { "epoch": 0.44, "learning_rate": 6.146261182984026e-06, "loss": 1.6835, "step": 4159 }, { "epoch": 0.44, "learning_rate": 6.144579643819053e-06, "loss": 1.3397, "step": 4160 }, { "epoch": 0.44, "learning_rate": 6.142897968028704e-06, "loss": 1.401, "step": 4161 }, { "epoch": 0.44, "learning_rate": 6.141216155813713e-06, "loss": 1.4175, "step": 4162 }, { "epoch": 0.44, "learning_rate": 6.1395342073748345e-06, "loss": 1.5681, "step": 4163 }, { "epoch": 0.44, "learning_rate": 6.137852122912839e-06, "loss": 1.2445, "step": 4164 }, { "epoch": 0.44, "learning_rate": 6.136169902628508e-06, "loss": 1.4104, "step": 4165 }, { "epoch": 0.44, "learning_rate": 6.134487546722652e-06, "loss": 1.5624, "step": 4166 }, { "epoch": 0.44, "learning_rate": 6.1328050553960804e-06, "loss": 1.5675, "step": 4167 }, { "epoch": 0.44, "learning_rate": 6.131122428849634e-06, "loss": 1.2415, "step": 4168 }, { "epoch": 0.44, "learning_rate": 6.12943966728416e-06, "loss": 1.2883, "step": 4169 }, { "epoch": 0.44, "learning_rate": 6.1277567709005245e-06, "loss": 1.3579, "step": 4170 }, { "epoch": 0.44, "learning_rate": 6.126073739899615e-06, "loss": 1.5798, "step": 4171 }, { "epoch": 0.45, "learning_rate": 6.124390574482324e-06, "loss": 1.3913, "step": 4172 }, { "epoch": 0.45, "learning_rate": 6.122707274849572e-06, "loss": 1.5158, "step": 4173 }, { "epoch": 0.45, "learning_rate": 6.1210238412022875e-06, "loss": 1.7402, "step": 4174 }, { "epoch": 0.45, "learning_rate": 6.1193402737414154e-06, "loss": 1.2429, "step": 4175 }, { "epoch": 0.45, "learning_rate": 6.117656572667921e-06, "loss": 1.3283, "step": 4176 }, { "epoch": 0.45, "learning_rate": 6.115972738182784e-06, "loss": 1.2271, "step": 4177 }, { "epoch": 0.45, "learning_rate": 6.114288770486998e-06, "loss": 1.2991, "step": 4178 }, { "epoch": 0.45, "learning_rate": 6.112604669781572e-06, "loss": 1.4835, "step": 4179 }, { "epoch": 0.45, "learning_rate": 6.110920436267536e-06, "loss": 1.1249, "step": 4180 }, { "epoch": 0.45, "learning_rate": 6.10923607014593e-06, "loss": 1.3017, "step": 4181 }, { "epoch": 0.45, "learning_rate": 6.107551571617813e-06, "loss": 1.6009, "step": 4182 }, { "epoch": 0.45, "learning_rate": 6.1058669408842586e-06, "loss": 1.428, "step": 4183 }, { "epoch": 0.45, "learning_rate": 6.1041821781463566e-06, "loss": 1.5524, "step": 4184 }, { "epoch": 0.45, "learning_rate": 6.1024972836052135e-06, "loss": 1.4545, "step": 4185 }, { "epoch": 0.45, "learning_rate": 6.100812257461949e-06, "loss": 1.4533, "step": 4186 }, { "epoch": 0.45, "learning_rate": 6.0991270999177e-06, "loss": 1.5882, "step": 4187 }, { "epoch": 0.45, "learning_rate": 6.0974418111736235e-06, "loss": 1.5501, "step": 4188 }, { "epoch": 0.45, "learning_rate": 6.0957563914308815e-06, "loss": 1.4212, "step": 4189 }, { "epoch": 0.45, "learning_rate": 6.094070840890663e-06, "loss": 1.5667, "step": 4190 }, { "epoch": 0.45, "learning_rate": 6.092385159754165e-06, "loss": 1.5933, "step": 4191 }, { "epoch": 0.45, "learning_rate": 6.0906993482226025e-06, "loss": 1.5207, "step": 4192 }, { "epoch": 0.45, "learning_rate": 6.0890134064972085e-06, "loss": 1.6153, "step": 4193 }, { "epoch": 0.45, "learning_rate": 6.0873273347792275e-06, "loss": 1.6415, "step": 4194 }, { "epoch": 0.45, "learning_rate": 6.085641133269923e-06, "loss": 1.4979, "step": 4195 }, { "epoch": 0.45, "learning_rate": 6.083954802170569e-06, "loss": 1.3259, "step": 4196 }, { "epoch": 0.45, "learning_rate": 6.0822683416824625e-06, "loss": 1.3506, "step": 4197 }, { "epoch": 0.45, "learning_rate": 6.080581752006909e-06, "loss": 1.7889, "step": 4198 }, { "epoch": 0.45, "learning_rate": 6.0788950333452324e-06, "loss": 1.4165, "step": 4199 }, { "epoch": 0.45, "learning_rate": 6.077208185898772e-06, "loss": 1.4728, "step": 4200 }, { "epoch": 0.45, "learning_rate": 6.075521209868882e-06, "loss": 1.5193, "step": 4201 }, { "epoch": 0.45, "learning_rate": 6.073834105456934e-06, "loss": 1.2443, "step": 4202 }, { "epoch": 0.45, "learning_rate": 6.07214687286431e-06, "loss": 1.3444, "step": 4203 }, { "epoch": 0.45, "learning_rate": 6.070459512292413e-06, "loss": 1.5384, "step": 4204 }, { "epoch": 0.45, "learning_rate": 6.068772023942659e-06, "loss": 1.8701, "step": 4205 }, { "epoch": 0.45, "learning_rate": 6.067084408016475e-06, "loss": 1.4574, "step": 4206 }, { "epoch": 0.45, "learning_rate": 6.065396664715311e-06, "loss": 1.5406, "step": 4207 }, { "epoch": 0.45, "learning_rate": 6.063708794240627e-06, "loss": 1.4036, "step": 4208 }, { "epoch": 0.45, "learning_rate": 6.0620207967939e-06, "loss": 1.5993, "step": 4209 }, { "epoch": 0.45, "learning_rate": 6.0603326725766195e-06, "loss": 1.469, "step": 4210 }, { "epoch": 0.45, "learning_rate": 6.058644421790296e-06, "loss": 1.3458, "step": 4211 }, { "epoch": 0.45, "learning_rate": 6.0569560446364495e-06, "loss": 1.3363, "step": 4212 }, { "epoch": 0.45, "learning_rate": 6.0552675413166185e-06, "loss": 1.6514, "step": 4213 }, { "epoch": 0.45, "learning_rate": 6.05357891203235e-06, "loss": 1.6292, "step": 4214 }, { "epoch": 0.45, "learning_rate": 6.051890156985217e-06, "loss": 1.281, "step": 4215 }, { "epoch": 0.45, "learning_rate": 6.0502012763768e-06, "loss": 1.5609, "step": 4216 }, { "epoch": 0.45, "learning_rate": 6.0485122704086955e-06, "loss": 1.6275, "step": 4217 }, { "epoch": 0.45, "learning_rate": 6.046823139282515e-06, "loss": 1.4646, "step": 4218 }, { "epoch": 0.45, "learning_rate": 6.0451338831998855e-06, "loss": 1.5601, "step": 4219 }, { "epoch": 0.45, "learning_rate": 6.04344450236245e-06, "loss": 1.7067, "step": 4220 }, { "epoch": 0.45, "learning_rate": 6.041754996971866e-06, "loss": 1.4458, "step": 4221 }, { "epoch": 0.45, "learning_rate": 6.040065367229803e-06, "loss": 1.4445, "step": 4222 }, { "epoch": 0.45, "learning_rate": 6.038375613337949e-06, "loss": 1.3923, "step": 4223 }, { "epoch": 0.45, "learning_rate": 6.036685735498004e-06, "loss": 1.3974, "step": 4224 }, { "epoch": 0.45, "learning_rate": 6.034995733911687e-06, "loss": 1.4064, "step": 4225 }, { "epoch": 0.45, "learning_rate": 6.033305608780725e-06, "loss": 1.7041, "step": 4226 }, { "epoch": 0.45, "learning_rate": 6.031615360306867e-06, "loss": 1.4077, "step": 4227 }, { "epoch": 0.45, "learning_rate": 6.029924988691872e-06, "loss": 1.085, "step": 4228 }, { "epoch": 0.45, "learning_rate": 6.028234494137515e-06, "loss": 1.6945, "step": 4229 }, { "epoch": 0.45, "learning_rate": 6.026543876845586e-06, "loss": 1.3226, "step": 4230 }, { "epoch": 0.45, "learning_rate": 6.024853137017889e-06, "loss": 1.5939, "step": 4231 }, { "epoch": 0.45, "learning_rate": 6.023162274856243e-06, "loss": 1.5579, "step": 4232 }, { "epoch": 0.45, "learning_rate": 6.021471290562484e-06, "loss": 1.2321, "step": 4233 }, { "epoch": 0.45, "learning_rate": 6.019780184338457e-06, "loss": 1.1898, "step": 4234 }, { "epoch": 0.45, "learning_rate": 6.018088956386025e-06, "loss": 1.2453, "step": 4235 }, { "epoch": 0.45, "learning_rate": 6.016397606907069e-06, "loss": 1.2959, "step": 4236 }, { "epoch": 0.45, "learning_rate": 6.0147061361034765e-06, "loss": 1.5103, "step": 4237 }, { "epoch": 0.45, "learning_rate": 6.013014544177156e-06, "loss": 1.3165, "step": 4238 }, { "epoch": 0.45, "learning_rate": 6.011322831330028e-06, "loss": 1.4682, "step": 4239 }, { "epoch": 0.45, "learning_rate": 6.009630997764027e-06, "loss": 1.508, "step": 4240 }, { "epoch": 0.45, "learning_rate": 6.0079390436811025e-06, "loss": 1.3231, "step": 4241 }, { "epoch": 0.45, "learning_rate": 6.0062469692832205e-06, "loss": 1.351, "step": 4242 }, { "epoch": 0.45, "learning_rate": 6.004554774772356e-06, "loss": 1.668, "step": 4243 }, { "epoch": 0.45, "learning_rate": 6.002862460350505e-06, "loss": 1.5685, "step": 4244 }, { "epoch": 0.45, "learning_rate": 6.001170026219673e-06, "loss": 1.4544, "step": 4245 }, { "epoch": 0.45, "learning_rate": 5.99947747258188e-06, "loss": 1.4221, "step": 4246 }, { "epoch": 0.45, "learning_rate": 5.9977847996391634e-06, "loss": 1.4396, "step": 4247 }, { "epoch": 0.45, "learning_rate": 5.996092007593572e-06, "loss": 1.294, "step": 4248 }, { "epoch": 0.45, "learning_rate": 5.994399096647172e-06, "loss": 1.5588, "step": 4249 }, { "epoch": 0.45, "learning_rate": 5.99270606700204e-06, "loss": 1.4252, "step": 4250 }, { "epoch": 0.45, "learning_rate": 5.9910129188602665e-06, "loss": 1.262, "step": 4251 }, { "epoch": 0.45, "learning_rate": 5.9893196524239604e-06, "loss": 1.5262, "step": 4252 }, { "epoch": 0.45, "learning_rate": 5.987626267895242e-06, "loss": 1.3326, "step": 4253 }, { "epoch": 0.45, "learning_rate": 5.985932765476246e-06, "loss": 1.4233, "step": 4254 }, { "epoch": 0.45, "learning_rate": 5.98423914536912e-06, "loss": 1.1737, "step": 4255 }, { "epoch": 0.45, "learning_rate": 5.9825454077760295e-06, "loss": 1.383, "step": 4256 }, { "epoch": 0.45, "learning_rate": 5.9808515528991486e-06, "loss": 1.3955, "step": 4257 }, { "epoch": 0.45, "learning_rate": 5.979157580940669e-06, "loss": 1.3977, "step": 4258 }, { "epoch": 0.45, "learning_rate": 5.977463492102796e-06, "loss": 1.615, "step": 4259 }, { "epoch": 0.45, "learning_rate": 5.975769286587747e-06, "loss": 1.3581, "step": 4260 }, { "epoch": 0.45, "learning_rate": 5.974074964597759e-06, "loss": 1.5536, "step": 4261 }, { "epoch": 0.45, "learning_rate": 5.972380526335074e-06, "loss": 1.505, "step": 4262 }, { "epoch": 0.45, "learning_rate": 5.970685972001953e-06, "loss": 1.4898, "step": 4263 }, { "epoch": 0.45, "learning_rate": 5.968991301800674e-06, "loss": 1.6373, "step": 4264 }, { "epoch": 0.45, "learning_rate": 5.967296515933519e-06, "loss": 1.5089, "step": 4265 }, { "epoch": 0.46, "learning_rate": 5.965601614602798e-06, "loss": 1.2991, "step": 4266 }, { "epoch": 0.46, "learning_rate": 5.96390659801082e-06, "loss": 1.2893, "step": 4267 }, { "epoch": 0.46, "learning_rate": 5.962211466359917e-06, "loss": 1.3965, "step": 4268 }, { "epoch": 0.46, "learning_rate": 5.960516219852433e-06, "loss": 1.4763, "step": 4269 }, { "epoch": 0.46, "learning_rate": 5.958820858690724e-06, "loss": 1.4157, "step": 4270 }, { "epoch": 0.46, "learning_rate": 5.95712538307716e-06, "loss": 1.3868, "step": 4271 }, { "epoch": 0.46, "learning_rate": 5.955429793214129e-06, "loss": 1.4911, "step": 4272 }, { "epoch": 0.46, "learning_rate": 5.9537340893040244e-06, "loss": 1.4247, "step": 4273 }, { "epoch": 0.46, "learning_rate": 5.95203827154926e-06, "loss": 1.4659, "step": 4274 }, { "epoch": 0.46, "learning_rate": 5.950342340152261e-06, "loss": 1.502, "step": 4275 }, { "epoch": 0.46, "learning_rate": 5.9486462953154656e-06, "loss": 1.3858, "step": 4276 }, { "epoch": 0.46, "learning_rate": 5.9469501372413255e-06, "loss": 1.4257, "step": 4277 }, { "epoch": 0.46, "learning_rate": 5.945253866132308e-06, "loss": 1.4241, "step": 4278 }, { "epoch": 0.46, "learning_rate": 5.9435574821908914e-06, "loss": 1.5034, "step": 4279 }, { "epoch": 0.46, "learning_rate": 5.94186098561957e-06, "loss": 1.3765, "step": 4280 }, { "epoch": 0.46, "learning_rate": 5.940164376620847e-06, "loss": 1.5006, "step": 4281 }, { "epoch": 0.46, "learning_rate": 5.938467655397244e-06, "loss": 1.5478, "step": 4282 }, { "epoch": 0.46, "learning_rate": 5.936770822151297e-06, "loss": 1.4405, "step": 4283 }, { "epoch": 0.46, "learning_rate": 5.935073877085546e-06, "loss": 1.5698, "step": 4284 }, { "epoch": 0.46, "learning_rate": 5.933376820402554e-06, "loss": 1.447, "step": 4285 }, { "epoch": 0.46, "learning_rate": 5.931679652304896e-06, "loss": 1.6779, "step": 4286 }, { "epoch": 0.46, "learning_rate": 5.9299823729951544e-06, "loss": 1.5232, "step": 4287 }, { "epoch": 0.46, "learning_rate": 5.928284982675931e-06, "loss": 1.4975, "step": 4288 }, { "epoch": 0.46, "learning_rate": 5.926587481549841e-06, "loss": 1.3425, "step": 4289 }, { "epoch": 0.46, "learning_rate": 5.9248898698195054e-06, "loss": 1.3697, "step": 4290 }, { "epoch": 0.46, "learning_rate": 5.923192147687568e-06, "loss": 1.6916, "step": 4291 }, { "epoch": 0.46, "learning_rate": 5.921494315356679e-06, "loss": 1.5464, "step": 4292 }, { "epoch": 0.46, "learning_rate": 5.919796373029504e-06, "loss": 1.4006, "step": 4293 }, { "epoch": 0.46, "learning_rate": 5.918098320908723e-06, "loss": 1.5414, "step": 4294 }, { "epoch": 0.46, "learning_rate": 5.916400159197027e-06, "loss": 1.6123, "step": 4295 }, { "epoch": 0.46, "learning_rate": 5.914701888097121e-06, "loss": 1.4944, "step": 4296 }, { "epoch": 0.46, "learning_rate": 5.913003507811723e-06, "loss": 1.4812, "step": 4297 }, { "epoch": 0.46, "learning_rate": 5.911305018543565e-06, "loss": 1.5128, "step": 4298 }, { "epoch": 0.46, "learning_rate": 5.90960642049539e-06, "loss": 1.2353, "step": 4299 }, { "epoch": 0.46, "learning_rate": 5.9079077138699555e-06, "loss": 1.6863, "step": 4300 }, { "epoch": 0.46, "learning_rate": 5.906208898870032e-06, "loss": 1.4343, "step": 4301 }, { "epoch": 0.46, "learning_rate": 5.904509975698399e-06, "loss": 1.6872, "step": 4302 }, { "epoch": 0.46, "learning_rate": 5.90281094455786e-06, "loss": 1.3263, "step": 4303 }, { "epoch": 0.46, "learning_rate": 5.9011118056512155e-06, "loss": 1.4315, "step": 4304 }, { "epoch": 0.46, "learning_rate": 5.8994125591812914e-06, "loss": 1.4131, "step": 4305 }, { "epoch": 0.46, "learning_rate": 5.897713205350923e-06, "loss": 1.5862, "step": 4306 }, { "epoch": 0.46, "learning_rate": 5.896013744362954e-06, "loss": 1.7011, "step": 4307 }, { "epoch": 0.46, "learning_rate": 5.894314176420247e-06, "loss": 1.7879, "step": 4308 }, { "epoch": 0.46, "learning_rate": 5.892614501725674e-06, "loss": 1.3176, "step": 4309 }, { "epoch": 0.46, "learning_rate": 5.890914720482122e-06, "loss": 1.4309, "step": 4310 }, { "epoch": 0.46, "learning_rate": 5.889214832892489e-06, "loss": 1.3496, "step": 4311 }, { "epoch": 0.46, "learning_rate": 5.887514839159683e-06, "loss": 1.2872, "step": 4312 }, { "epoch": 0.46, "learning_rate": 5.8858147394866296e-06, "loss": 1.513, "step": 4313 }, { "epoch": 0.46, "learning_rate": 5.8841145340762665e-06, "loss": 1.4331, "step": 4314 }, { "epoch": 0.46, "learning_rate": 5.88241422313154e-06, "loss": 1.4476, "step": 4315 }, { "epoch": 0.46, "learning_rate": 5.880713806855415e-06, "loss": 1.5925, "step": 4316 }, { "epoch": 0.46, "learning_rate": 5.879013285450863e-06, "loss": 1.3435, "step": 4317 }, { "epoch": 0.46, "learning_rate": 5.877312659120871e-06, "loss": 1.5577, "step": 4318 }, { "epoch": 0.46, "learning_rate": 5.875611928068439e-06, "loss": 1.3476, "step": 4319 }, { "epoch": 0.46, "learning_rate": 5.873911092496577e-06, "loss": 1.613, "step": 4320 }, { "epoch": 0.46, "learning_rate": 5.872210152608311e-06, "loss": 1.4778, "step": 4321 }, { "epoch": 0.46, "learning_rate": 5.870509108606677e-06, "loss": 1.3831, "step": 4322 }, { "epoch": 0.46, "learning_rate": 5.8688079606947226e-06, "loss": 1.2876, "step": 4323 }, { "epoch": 0.46, "learning_rate": 5.867106709075511e-06, "loss": 1.3641, "step": 4324 }, { "epoch": 0.46, "learning_rate": 5.865405353952116e-06, "loss": 1.5577, "step": 4325 }, { "epoch": 0.46, "learning_rate": 5.8637038955276225e-06, "loss": 1.3188, "step": 4326 }, { "epoch": 0.46, "learning_rate": 5.86200233400513e-06, "loss": 1.6521, "step": 4327 }, { "epoch": 0.46, "learning_rate": 5.8603006695877505e-06, "loss": 1.5401, "step": 4328 }, { "epoch": 0.46, "learning_rate": 5.858598902478604e-06, "loss": 1.6646, "step": 4329 }, { "epoch": 0.46, "learning_rate": 5.856897032880829e-06, "loss": 1.6117, "step": 4330 }, { "epoch": 0.46, "learning_rate": 5.855195060997571e-06, "loss": 1.4536, "step": 4331 }, { "epoch": 0.46, "learning_rate": 5.853492987031989e-06, "loss": 1.3703, "step": 4332 }, { "epoch": 0.46, "learning_rate": 5.851790811187258e-06, "loss": 1.7438, "step": 4333 }, { "epoch": 0.46, "learning_rate": 5.850088533666559e-06, "loss": 1.4245, "step": 4334 }, { "epoch": 0.46, "learning_rate": 5.8483861546730915e-06, "loss": 1.8189, "step": 4335 }, { "epoch": 0.46, "learning_rate": 5.846683674410061e-06, "loss": 1.4532, "step": 4336 }, { "epoch": 0.46, "learning_rate": 5.844981093080688e-06, "loss": 1.565, "step": 4337 }, { "epoch": 0.46, "learning_rate": 5.843278410888208e-06, "loss": 1.5133, "step": 4338 }, { "epoch": 0.46, "learning_rate": 5.841575628035863e-06, "loss": 1.5471, "step": 4339 }, { "epoch": 0.46, "learning_rate": 5.8398727447269084e-06, "loss": 1.0546, "step": 4340 }, { "epoch": 0.46, "learning_rate": 5.838169761164616e-06, "loss": 1.397, "step": 4341 }, { "epoch": 0.46, "learning_rate": 5.836466677552266e-06, "loss": 1.6761, "step": 4342 }, { "epoch": 0.46, "learning_rate": 5.8347634940931494e-06, "loss": 1.3815, "step": 4343 }, { "epoch": 0.46, "learning_rate": 5.83306021099057e-06, "loss": 1.1879, "step": 4344 }, { "epoch": 0.46, "learning_rate": 5.831356828447847e-06, "loss": 1.4413, "step": 4345 }, { "epoch": 0.46, "learning_rate": 5.829653346668305e-06, "loss": 1.4437, "step": 4346 }, { "epoch": 0.46, "learning_rate": 5.827949765855285e-06, "loss": 1.2398, "step": 4347 }, { "epoch": 0.46, "learning_rate": 5.826246086212142e-06, "loss": 1.3086, "step": 4348 }, { "epoch": 0.46, "learning_rate": 5.824542307942236e-06, "loss": 1.4445, "step": 4349 }, { "epoch": 0.46, "learning_rate": 5.822838431248943e-06, "loss": 1.5042, "step": 4350 }, { "epoch": 0.46, "learning_rate": 5.821134456335652e-06, "loss": 1.2058, "step": 4351 }, { "epoch": 0.46, "learning_rate": 5.81943038340576e-06, "loss": 1.3678, "step": 4352 }, { "epoch": 0.46, "learning_rate": 5.817726212662678e-06, "loss": 1.4311, "step": 4353 }, { "epoch": 0.46, "learning_rate": 5.816021944309828e-06, "loss": 1.434, "step": 4354 }, { "epoch": 0.46, "learning_rate": 5.8143175785506455e-06, "loss": 1.4774, "step": 4355 }, { "epoch": 0.46, "learning_rate": 5.812613115588575e-06, "loss": 1.5552, "step": 4356 }, { "epoch": 0.46, "learning_rate": 5.810908555627072e-06, "loss": 1.3914, "step": 4357 }, { "epoch": 0.46, "learning_rate": 5.809203898869608e-06, "loss": 1.498, "step": 4358 }, { "epoch": 0.46, "learning_rate": 5.807499145519663e-06, "loss": 1.4772, "step": 4359 }, { "epoch": 0.47, "learning_rate": 5.805794295780727e-06, "loss": 1.4927, "step": 4360 }, { "epoch": 0.47, "learning_rate": 5.804089349856302e-06, "loss": 1.3959, "step": 4361 }, { "epoch": 0.47, "learning_rate": 5.802384307949909e-06, "loss": 1.2127, "step": 4362 }, { "epoch": 0.47, "learning_rate": 5.800679170265067e-06, "loss": 1.425, "step": 4363 }, { "epoch": 0.47, "learning_rate": 5.798973937005319e-06, "loss": 1.4671, "step": 4364 }, { "epoch": 0.47, "learning_rate": 5.79726860837421e-06, "loss": 1.4448, "step": 4365 }, { "epoch": 0.47, "learning_rate": 5.7955631845753026e-06, "loss": 1.4105, "step": 4366 }, { "epoch": 0.47, "learning_rate": 5.7938576658121705e-06, "loss": 1.5378, "step": 4367 }, { "epoch": 0.47, "learning_rate": 5.792152052288391e-06, "loss": 1.3907, "step": 4368 }, { "epoch": 0.47, "learning_rate": 5.7904463442075655e-06, "loss": 1.4699, "step": 4369 }, { "epoch": 0.47, "learning_rate": 5.788740541773296e-06, "loss": 1.6813, "step": 4370 }, { "epoch": 0.47, "learning_rate": 5.787034645189199e-06, "loss": 1.5392, "step": 4371 }, { "epoch": 0.47, "learning_rate": 5.785328654658905e-06, "loss": 1.3495, "step": 4372 }, { "epoch": 0.47, "learning_rate": 5.7836225703860525e-06, "loss": 1.4794, "step": 4373 }, { "epoch": 0.47, "learning_rate": 5.7819163925742915e-06, "loss": 1.4214, "step": 4374 }, { "epoch": 0.47, "learning_rate": 5.780210121427286e-06, "loss": 1.2087, "step": 4375 }, { "epoch": 0.47, "learning_rate": 5.778503757148706e-06, "loss": 1.2471, "step": 4376 }, { "epoch": 0.47, "learning_rate": 5.776797299942236e-06, "loss": 1.4523, "step": 4377 }, { "epoch": 0.47, "learning_rate": 5.7750907500115735e-06, "loss": 1.4081, "step": 4378 }, { "epoch": 0.47, "learning_rate": 5.773384107560424e-06, "loss": 1.5929, "step": 4379 }, { "epoch": 0.47, "learning_rate": 5.771677372792502e-06, "loss": 1.4055, "step": 4380 }, { "epoch": 0.47, "learning_rate": 5.76997054591154e-06, "loss": 1.3933, "step": 4381 }, { "epoch": 0.47, "learning_rate": 5.7682636271212746e-06, "loss": 1.4822, "step": 4382 }, { "epoch": 0.47, "learning_rate": 5.766556616625456e-06, "loss": 1.798, "step": 4383 }, { "epoch": 0.47, "learning_rate": 5.764849514627848e-06, "loss": 1.4614, "step": 4384 }, { "epoch": 0.47, "learning_rate": 5.7631423213322195e-06, "loss": 1.5084, "step": 4385 }, { "epoch": 0.47, "learning_rate": 5.7614350369423555e-06, "loss": 1.5651, "step": 4386 }, { "epoch": 0.47, "learning_rate": 5.75972766166205e-06, "loss": 1.4058, "step": 4387 }, { "epoch": 0.47, "learning_rate": 5.758020195695107e-06, "loss": 1.3359, "step": 4388 }, { "epoch": 0.47, "learning_rate": 5.7563126392453415e-06, "loss": 1.4524, "step": 4389 }, { "epoch": 0.47, "learning_rate": 5.754604992516581e-06, "loss": 1.7966, "step": 4390 }, { "epoch": 0.47, "learning_rate": 5.7528972557126625e-06, "loss": 1.4107, "step": 4391 }, { "epoch": 0.47, "learning_rate": 5.751189429037435e-06, "loss": 1.3449, "step": 4392 }, { "epoch": 0.47, "learning_rate": 5.749481512694755e-06, "loss": 1.5954, "step": 4393 }, { "epoch": 0.47, "learning_rate": 5.747773506888492e-06, "loss": 1.235, "step": 4394 }, { "epoch": 0.47, "learning_rate": 5.746065411822528e-06, "loss": 1.5096, "step": 4395 }, { "epoch": 0.47, "learning_rate": 5.744357227700752e-06, "loss": 1.4368, "step": 4396 }, { "epoch": 0.47, "learning_rate": 5.742648954727066e-06, "loss": 1.5713, "step": 4397 }, { "epoch": 0.47, "learning_rate": 5.740940593105383e-06, "loss": 1.7975, "step": 4398 }, { "epoch": 0.47, "learning_rate": 5.739232143039623e-06, "loss": 1.5418, "step": 4399 }, { "epoch": 0.47, "learning_rate": 5.737523604733719e-06, "loss": 1.418, "step": 4400 }, { "epoch": 0.47, "learning_rate": 5.73581497839162e-06, "loss": 1.4099, "step": 4401 }, { "epoch": 0.47, "learning_rate": 5.734106264217272e-06, "loss": 1.5935, "step": 4402 }, { "epoch": 0.47, "learning_rate": 5.732397462414645e-06, "loss": 1.4, "step": 4403 }, { "epoch": 0.47, "learning_rate": 5.730688573187715e-06, "loss": 1.5359, "step": 4404 }, { "epoch": 0.47, "learning_rate": 5.7289795967404624e-06, "loss": 1.523, "step": 4405 }, { "epoch": 0.47, "learning_rate": 5.7272705332768865e-06, "loss": 1.5, "step": 4406 }, { "epoch": 0.47, "learning_rate": 5.725561383000994e-06, "loss": 1.5305, "step": 4407 }, { "epoch": 0.47, "learning_rate": 5.723852146116799e-06, "loss": 1.5446, "step": 4408 }, { "epoch": 0.47, "learning_rate": 5.7221428228283325e-06, "loss": 1.5483, "step": 4409 }, { "epoch": 0.47, "learning_rate": 5.720433413339627e-06, "loss": 1.4435, "step": 4410 }, { "epoch": 0.47, "learning_rate": 5.718723917854733e-06, "loss": 1.4527, "step": 4411 }, { "epoch": 0.47, "learning_rate": 5.717014336577709e-06, "loss": 1.408, "step": 4412 }, { "epoch": 0.47, "learning_rate": 5.71530466971262e-06, "loss": 1.6298, "step": 4413 }, { "epoch": 0.47, "learning_rate": 5.713594917463546e-06, "loss": 1.5078, "step": 4414 }, { "epoch": 0.47, "learning_rate": 5.711885080034578e-06, "loss": 1.5788, "step": 4415 }, { "epoch": 0.47, "learning_rate": 5.710175157629812e-06, "loss": 1.4642, "step": 4416 }, { "epoch": 0.47, "learning_rate": 5.708465150453356e-06, "loss": 1.4025, "step": 4417 }, { "epoch": 0.47, "learning_rate": 5.70675505870933e-06, "loss": 1.4618, "step": 4418 }, { "epoch": 0.47, "learning_rate": 5.705044882601862e-06, "loss": 1.2811, "step": 4419 }, { "epoch": 0.47, "learning_rate": 5.703334622335096e-06, "loss": 1.3409, "step": 4420 }, { "epoch": 0.47, "learning_rate": 5.7016242781131746e-06, "loss": 1.3705, "step": 4421 }, { "epoch": 0.47, "learning_rate": 5.69991385014026e-06, "loss": 1.3735, "step": 4422 }, { "epoch": 0.47, "learning_rate": 5.698203338620523e-06, "loss": 1.7548, "step": 4423 }, { "epoch": 0.47, "learning_rate": 5.696492743758138e-06, "loss": 1.4324, "step": 4424 }, { "epoch": 0.47, "learning_rate": 5.694782065757298e-06, "loss": 1.3922, "step": 4425 }, { "epoch": 0.47, "learning_rate": 5.693071304822203e-06, "loss": 1.5669, "step": 4426 }, { "epoch": 0.47, "learning_rate": 5.691360461157056e-06, "loss": 1.0793, "step": 4427 }, { "epoch": 0.47, "learning_rate": 5.689649534966083e-06, "loss": 1.3594, "step": 4428 }, { "epoch": 0.47, "learning_rate": 5.687938526453507e-06, "loss": 1.4187, "step": 4429 }, { "epoch": 0.47, "learning_rate": 5.6862274358235705e-06, "loss": 1.3583, "step": 4430 }, { "epoch": 0.47, "learning_rate": 5.684516263280519e-06, "loss": 1.6832, "step": 4431 }, { "epoch": 0.47, "learning_rate": 5.682805009028613e-06, "loss": 1.195, "step": 4432 }, { "epoch": 0.47, "learning_rate": 5.681093673272117e-06, "loss": 1.4196, "step": 4433 }, { "epoch": 0.47, "learning_rate": 5.679382256215311e-06, "loss": 1.4776, "step": 4434 }, { "epoch": 0.47, "learning_rate": 5.677670758062481e-06, "loss": 1.2877, "step": 4435 }, { "epoch": 0.47, "learning_rate": 5.675959179017925e-06, "loss": 1.4702, "step": 4436 }, { "epoch": 0.47, "learning_rate": 5.674247519285951e-06, "loss": 1.4976, "step": 4437 }, { "epoch": 0.47, "learning_rate": 5.67253577907087e-06, "loss": 1.6004, "step": 4438 }, { "epoch": 0.47, "learning_rate": 5.670823958577012e-06, "loss": 1.4, "step": 4439 }, { "epoch": 0.47, "learning_rate": 5.6691120580087126e-06, "loss": 1.3104, "step": 4440 }, { "epoch": 0.47, "learning_rate": 5.667400077570314e-06, "loss": 1.6077, "step": 4441 }, { "epoch": 0.47, "learning_rate": 5.665688017466173e-06, "loss": 1.4329, "step": 4442 }, { "epoch": 0.47, "learning_rate": 5.6639758779006535e-06, "loss": 1.3047, "step": 4443 }, { "epoch": 0.47, "learning_rate": 5.662263659078127e-06, "loss": 1.4561, "step": 4444 }, { "epoch": 0.47, "learning_rate": 5.660551361202979e-06, "loss": 1.4591, "step": 4445 }, { "epoch": 0.47, "learning_rate": 5.6588389844796e-06, "loss": 1.2909, "step": 4446 }, { "epoch": 0.47, "learning_rate": 5.657126529112393e-06, "loss": 1.4099, "step": 4447 }, { "epoch": 0.47, "learning_rate": 5.655413995305769e-06, "loss": 1.423, "step": 4448 }, { "epoch": 0.47, "learning_rate": 5.653701383264147e-06, "loss": 1.2937, "step": 4449 }, { "epoch": 0.47, "learning_rate": 5.651988693191958e-06, "loss": 1.5876, "step": 4450 }, { "epoch": 0.47, "learning_rate": 5.6502759252936415e-06, "loss": 1.3316, "step": 4451 }, { "epoch": 0.47, "learning_rate": 5.648563079773646e-06, "loss": 1.4904, "step": 4452 }, { "epoch": 0.47, "learning_rate": 5.646850156836427e-06, "loss": 1.2398, "step": 4453 }, { "epoch": 0.48, "learning_rate": 5.645137156686455e-06, "loss": 1.2571, "step": 4454 }, { "epoch": 0.48, "learning_rate": 5.6434240795282045e-06, "loss": 1.4861, "step": 4455 }, { "epoch": 0.48, "learning_rate": 5.64171092556616e-06, "loss": 1.4937, "step": 4456 }, { "epoch": 0.48, "learning_rate": 5.639997695004819e-06, "loss": 1.2918, "step": 4457 }, { "epoch": 0.48, "learning_rate": 5.63828438804868e-06, "loss": 1.4967, "step": 4458 }, { "epoch": 0.48, "learning_rate": 5.636571004902262e-06, "loss": 1.3846, "step": 4459 }, { "epoch": 0.48, "learning_rate": 5.634857545770083e-06, "loss": 1.459, "step": 4460 }, { "epoch": 0.48, "learning_rate": 5.6331440108566735e-06, "loss": 0.9854, "step": 4461 }, { "epoch": 0.48, "learning_rate": 5.631430400366577e-06, "loss": 1.301, "step": 4462 }, { "epoch": 0.48, "learning_rate": 5.6297167145043394e-06, "loss": 1.5564, "step": 4463 }, { "epoch": 0.48, "learning_rate": 5.628002953474521e-06, "loss": 1.6035, "step": 4464 }, { "epoch": 0.48, "learning_rate": 5.626289117481687e-06, "loss": 1.5003, "step": 4465 }, { "epoch": 0.48, "learning_rate": 5.624575206730413e-06, "loss": 1.3484, "step": 4466 }, { "epoch": 0.48, "learning_rate": 5.622861221425286e-06, "loss": 1.6227, "step": 4467 }, { "epoch": 0.48, "learning_rate": 5.621147161770898e-06, "loss": 1.5775, "step": 4468 }, { "epoch": 0.48, "learning_rate": 5.619433027971854e-06, "loss": 1.1999, "step": 4469 }, { "epoch": 0.48, "learning_rate": 5.617718820232762e-06, "loss": 1.2756, "step": 4470 }, { "epoch": 0.48, "learning_rate": 5.6160045387582455e-06, "loss": 1.4009, "step": 4471 }, { "epoch": 0.48, "learning_rate": 5.61429018375293e-06, "loss": 1.5656, "step": 4472 }, { "epoch": 0.48, "learning_rate": 5.612575755421459e-06, "loss": 1.4993, "step": 4473 }, { "epoch": 0.48, "learning_rate": 5.6108612539684745e-06, "loss": 1.4213, "step": 4474 }, { "epoch": 0.48, "learning_rate": 5.609146679598634e-06, "loss": 1.1275, "step": 4475 }, { "epoch": 0.48, "learning_rate": 5.607432032516601e-06, "loss": 1.3275, "step": 4476 }, { "epoch": 0.48, "learning_rate": 5.605717312927047e-06, "loss": 1.2379, "step": 4477 }, { "epoch": 0.48, "learning_rate": 5.604002521034655e-06, "loss": 1.5648, "step": 4478 }, { "epoch": 0.48, "learning_rate": 5.602287657044116e-06, "loss": 1.4272, "step": 4479 }, { "epoch": 0.48, "learning_rate": 5.600572721160127e-06, "loss": 1.363, "step": 4480 }, { "epoch": 0.48, "learning_rate": 5.5988577135873954e-06, "loss": 1.3838, "step": 4481 }, { "epoch": 0.48, "learning_rate": 5.597142634530639e-06, "loss": 1.1064, "step": 4482 }, { "epoch": 0.48, "learning_rate": 5.595427484194578e-06, "loss": 1.6733, "step": 4483 }, { "epoch": 0.48, "learning_rate": 5.5937122627839495e-06, "loss": 1.4427, "step": 4484 }, { "epoch": 0.48, "learning_rate": 5.5919969705034914e-06, "loss": 1.3956, "step": 4485 }, { "epoch": 0.48, "learning_rate": 5.590281607557957e-06, "loss": 1.3642, "step": 4486 }, { "epoch": 0.48, "learning_rate": 5.588566174152101e-06, "loss": 1.4511, "step": 4487 }, { "epoch": 0.48, "learning_rate": 5.586850670490694e-06, "loss": 1.4451, "step": 4488 }, { "epoch": 0.48, "learning_rate": 5.5851350967785075e-06, "loss": 1.4246, "step": 4489 }, { "epoch": 0.48, "learning_rate": 5.583419453220327e-06, "loss": 1.434, "step": 4490 }, { "epoch": 0.48, "learning_rate": 5.581703740020943e-06, "loss": 1.3436, "step": 4491 }, { "epoch": 0.48, "learning_rate": 5.579987957385155e-06, "loss": 1.5365, "step": 4492 }, { "epoch": 0.48, "learning_rate": 5.578272105517775e-06, "loss": 1.3887, "step": 4493 }, { "epoch": 0.48, "learning_rate": 5.576556184623615e-06, "loss": 1.3735, "step": 4494 }, { "epoch": 0.48, "learning_rate": 5.5748401949075e-06, "loss": 1.437, "step": 4495 }, { "epoch": 0.48, "learning_rate": 5.573124136574268e-06, "loss": 1.2375, "step": 4496 }, { "epoch": 0.48, "learning_rate": 5.571408009828757e-06, "loss": 1.3464, "step": 4497 }, { "epoch": 0.48, "learning_rate": 5.569691814875814e-06, "loss": 1.4012, "step": 4498 }, { "epoch": 0.48, "learning_rate": 5.567975551920302e-06, "loss": 1.5334, "step": 4499 }, { "epoch": 0.48, "learning_rate": 5.56625922116708e-06, "loss": 1.5738, "step": 4500 }, { "epoch": 0.48, "learning_rate": 5.5645428228210286e-06, "loss": 1.2204, "step": 4501 }, { "epoch": 0.48, "learning_rate": 5.562826357087028e-06, "loss": 1.3921, "step": 4502 }, { "epoch": 0.48, "learning_rate": 5.561109824169962e-06, "loss": 1.4265, "step": 4503 }, { "epoch": 0.48, "learning_rate": 5.5593932242747365e-06, "loss": 1.5411, "step": 4504 }, { "epoch": 0.48, "learning_rate": 5.557676557606252e-06, "loss": 1.4078, "step": 4505 }, { "epoch": 0.48, "learning_rate": 5.555959824369426e-06, "loss": 1.5111, "step": 4506 }, { "epoch": 0.48, "learning_rate": 5.554243024769177e-06, "loss": 1.5425, "step": 4507 }, { "epoch": 0.48, "learning_rate": 5.5525261590104375e-06, "loss": 1.4005, "step": 4508 }, { "epoch": 0.48, "learning_rate": 5.550809227298144e-06, "loss": 1.5386, "step": 4509 }, { "epoch": 0.48, "learning_rate": 5.549092229837242e-06, "loss": 1.2892, "step": 4510 }, { "epoch": 0.48, "learning_rate": 5.5473751668326846e-06, "loss": 1.3872, "step": 4511 }, { "epoch": 0.48, "learning_rate": 5.545658038489433e-06, "loss": 1.4822, "step": 4512 }, { "epoch": 0.48, "learning_rate": 5.543940845012457e-06, "loss": 1.474, "step": 4513 }, { "epoch": 0.48, "learning_rate": 5.542223586606732e-06, "loss": 1.7333, "step": 4514 }, { "epoch": 0.48, "learning_rate": 5.540506263477243e-06, "loss": 1.4477, "step": 4515 }, { "epoch": 0.48, "learning_rate": 5.538788875828983e-06, "loss": 1.5064, "step": 4516 }, { "epoch": 0.48, "learning_rate": 5.53707142386695e-06, "loss": 1.491, "step": 4517 }, { "epoch": 0.48, "learning_rate": 5.535353907796155e-06, "loss": 1.2473, "step": 4518 }, { "epoch": 0.48, "learning_rate": 5.533636327821611e-06, "loss": 1.2752, "step": 4519 }, { "epoch": 0.48, "learning_rate": 5.531918684148338e-06, "loss": 1.4937, "step": 4520 }, { "epoch": 0.48, "learning_rate": 5.530200976981375e-06, "loss": 1.3556, "step": 4521 }, { "epoch": 0.48, "learning_rate": 5.52848320652575e-06, "loss": 1.5663, "step": 4522 }, { "epoch": 0.48, "learning_rate": 5.526765372986514e-06, "loss": 1.5341, "step": 4523 }, { "epoch": 0.48, "learning_rate": 5.525047476568722e-06, "loss": 1.2585, "step": 4524 }, { "epoch": 0.48, "learning_rate": 5.52332951747743e-06, "loss": 1.6517, "step": 4525 }, { "epoch": 0.48, "learning_rate": 5.521611495917707e-06, "loss": 1.2859, "step": 4526 }, { "epoch": 0.48, "learning_rate": 5.519893412094631e-06, "loss": 1.5302, "step": 4527 }, { "epoch": 0.48, "learning_rate": 5.518175266213284e-06, "loss": 1.3181, "step": 4528 }, { "epoch": 0.48, "learning_rate": 5.516457058478757e-06, "loss": 1.4109, "step": 4529 }, { "epoch": 0.48, "learning_rate": 5.514738789096146e-06, "loss": 1.5692, "step": 4530 }, { "epoch": 0.48, "learning_rate": 5.5130204582705574e-06, "loss": 1.848, "step": 4531 }, { "epoch": 0.48, "learning_rate": 5.5113020662071035e-06, "loss": 1.5101, "step": 4532 }, { "epoch": 0.48, "learning_rate": 5.509583613110904e-06, "loss": 1.5071, "step": 4533 }, { "epoch": 0.48, "learning_rate": 5.507865099187085e-06, "loss": 1.509, "step": 4534 }, { "epoch": 0.48, "learning_rate": 5.506146524640784e-06, "loss": 1.5766, "step": 4535 }, { "epoch": 0.48, "learning_rate": 5.504427889677141e-06, "loss": 1.4479, "step": 4536 }, { "epoch": 0.48, "learning_rate": 5.502709194501301e-06, "loss": 1.5802, "step": 4537 }, { "epoch": 0.48, "learning_rate": 5.500990439318427e-06, "loss": 1.4769, "step": 4538 }, { "epoch": 0.48, "learning_rate": 5.499271624333676e-06, "loss": 1.5, "step": 4539 }, { "epoch": 0.48, "learning_rate": 5.497552749752222e-06, "loss": 1.2552, "step": 4540 }, { "epoch": 0.48, "learning_rate": 5.495833815779243e-06, "loss": 1.3217, "step": 4541 }, { "epoch": 0.48, "learning_rate": 5.494114822619918e-06, "loss": 1.3236, "step": 4542 }, { "epoch": 0.48, "learning_rate": 5.492395770479444e-06, "loss": 1.6035, "step": 4543 }, { "epoch": 0.48, "learning_rate": 5.490676659563018e-06, "loss": 1.4384, "step": 4544 }, { "epoch": 0.48, "learning_rate": 5.488957490075846e-06, "loss": 1.59, "step": 4545 }, { "epoch": 0.48, "learning_rate": 5.4872382622231405e-06, "loss": 1.2411, "step": 4546 }, { "epoch": 0.49, "learning_rate": 5.485518976210119e-06, "loss": 1.464, "step": 4547 }, { "epoch": 0.49, "learning_rate": 5.483799632242012e-06, "loss": 1.3791, "step": 4548 }, { "epoch": 0.49, "learning_rate": 5.482080230524048e-06, "loss": 1.471, "step": 4549 }, { "epoch": 0.49, "learning_rate": 5.4803607712614725e-06, "loss": 1.3825, "step": 4550 }, { "epoch": 0.49, "learning_rate": 5.478641254659528e-06, "loss": 1.4044, "step": 4551 }, { "epoch": 0.49, "learning_rate": 5.476921680923474e-06, "loss": 1.5987, "step": 4552 }, { "epoch": 0.49, "learning_rate": 5.475202050258565e-06, "loss": 1.516, "step": 4553 }, { "epoch": 0.49, "learning_rate": 5.473482362870073e-06, "loss": 1.7492, "step": 4554 }, { "epoch": 0.49, "learning_rate": 5.4717626189632725e-06, "loss": 1.5377, "step": 4555 }, { "epoch": 0.49, "learning_rate": 5.470042818743441e-06, "loss": 1.6008, "step": 4556 }, { "epoch": 0.49, "learning_rate": 5.468322962415871e-06, "loss": 1.6642, "step": 4557 }, { "epoch": 0.49, "learning_rate": 5.4666030501858544e-06, "loss": 1.5885, "step": 4558 }, { "epoch": 0.49, "learning_rate": 5.464883082258692e-06, "loss": 1.4863, "step": 4559 }, { "epoch": 0.49, "learning_rate": 5.463163058839694e-06, "loss": 1.3452, "step": 4560 }, { "epoch": 0.49, "learning_rate": 5.461442980134173e-06, "loss": 1.2972, "step": 4561 }, { "epoch": 0.49, "learning_rate": 5.459722846347451e-06, "loss": 1.2641, "step": 4562 }, { "epoch": 0.49, "learning_rate": 5.4580026576848565e-06, "loss": 1.3721, "step": 4563 }, { "epoch": 0.49, "learning_rate": 5.456282414351722e-06, "loss": 1.5732, "step": 4564 }, { "epoch": 0.49, "learning_rate": 5.454562116553388e-06, "loss": 1.3784, "step": 4565 }, { "epoch": 0.49, "learning_rate": 5.452841764495203e-06, "loss": 1.3224, "step": 4566 }, { "epoch": 0.49, "learning_rate": 5.451121358382523e-06, "loss": 1.3585, "step": 4567 }, { "epoch": 0.49, "learning_rate": 5.449400898420705e-06, "loss": 1.5012, "step": 4568 }, { "epoch": 0.49, "learning_rate": 5.4476803848151146e-06, "loss": 1.5045, "step": 4569 }, { "epoch": 0.49, "learning_rate": 5.445959817771129e-06, "loss": 1.2416, "step": 4570 }, { "epoch": 0.49, "learning_rate": 5.444239197494125e-06, "loss": 1.5775, "step": 4571 }, { "epoch": 0.49, "learning_rate": 5.442518524189489e-06, "loss": 1.2883, "step": 4572 }, { "epoch": 0.49, "learning_rate": 5.440797798062611e-06, "loss": 1.6292, "step": 4573 }, { "epoch": 0.49, "learning_rate": 5.439077019318894e-06, "loss": 1.427, "step": 4574 }, { "epoch": 0.49, "learning_rate": 5.4373561881637405e-06, "loss": 1.4428, "step": 4575 }, { "epoch": 0.49, "learning_rate": 5.435635304802558e-06, "loss": 1.7829, "step": 4576 }, { "epoch": 0.49, "learning_rate": 5.43391436944077e-06, "loss": 1.5973, "step": 4577 }, { "epoch": 0.49, "learning_rate": 5.432193382283794e-06, "loss": 1.3009, "step": 4578 }, { "epoch": 0.49, "learning_rate": 5.430472343537065e-06, "loss": 1.4852, "step": 4579 }, { "epoch": 0.49, "learning_rate": 5.428751253406015e-06, "loss": 1.4394, "step": 4580 }, { "epoch": 0.49, "learning_rate": 5.4270301120960856e-06, "loss": 1.502, "step": 4581 }, { "epoch": 0.49, "learning_rate": 5.425308919812726e-06, "loss": 1.3818, "step": 4582 }, { "epoch": 0.49, "learning_rate": 5.423587676761392e-06, "loss": 1.3445, "step": 4583 }, { "epoch": 0.49, "learning_rate": 5.421866383147541e-06, "loss": 1.4075, "step": 4584 }, { "epoch": 0.49, "learning_rate": 5.420145039176641e-06, "loss": 1.5535, "step": 4585 }, { "epoch": 0.49, "learning_rate": 5.4184236450541624e-06, "loss": 1.2677, "step": 4586 }, { "epoch": 0.49, "learning_rate": 5.416702200985585e-06, "loss": 1.5502, "step": 4587 }, { "epoch": 0.49, "learning_rate": 5.414980707176392e-06, "loss": 1.1454, "step": 4588 }, { "epoch": 0.49, "learning_rate": 5.413259163832074e-06, "loss": 1.3611, "step": 4589 }, { "epoch": 0.49, "learning_rate": 5.411537571158127e-06, "loss": 1.4565, "step": 4590 }, { "epoch": 0.49, "learning_rate": 5.4098159293600525e-06, "loss": 1.5676, "step": 4591 }, { "epoch": 0.49, "learning_rate": 5.408094238643359e-06, "loss": 1.4622, "step": 4592 }, { "epoch": 0.49, "learning_rate": 5.406372499213557e-06, "loss": 1.4008, "step": 4593 }, { "epoch": 0.49, "learning_rate": 5.4046507112761714e-06, "loss": 1.1246, "step": 4594 }, { "epoch": 0.49, "learning_rate": 5.402928875036722e-06, "loss": 1.4442, "step": 4595 }, { "epoch": 0.49, "learning_rate": 5.401206990700741e-06, "loss": 1.5028, "step": 4596 }, { "epoch": 0.49, "learning_rate": 5.3994850584737705e-06, "loss": 1.5065, "step": 4597 }, { "epoch": 0.49, "learning_rate": 5.397763078561344e-06, "loss": 1.4953, "step": 4598 }, { "epoch": 0.49, "learning_rate": 5.396041051169016e-06, "loss": 1.5446, "step": 4599 }, { "epoch": 0.49, "learning_rate": 5.3943189765023375e-06, "loss": 1.5347, "step": 4600 }, { "epoch": 0.49, "learning_rate": 5.392596854766869e-06, "loss": 1.2938, "step": 4601 }, { "epoch": 0.49, "learning_rate": 5.390874686168176e-06, "loss": 1.6566, "step": 4602 }, { "epoch": 0.49, "learning_rate": 5.389152470911828e-06, "loss": 1.3352, "step": 4603 }, { "epoch": 0.49, "learning_rate": 5.3874302092034015e-06, "loss": 1.3371, "step": 4604 }, { "epoch": 0.49, "learning_rate": 5.385707901248478e-06, "loss": 1.362, "step": 4605 }, { "epoch": 0.49, "learning_rate": 5.383985547252645e-06, "loss": 1.3084, "step": 4606 }, { "epoch": 0.49, "learning_rate": 5.382263147421495e-06, "loss": 1.3569, "step": 4607 }, { "epoch": 0.49, "learning_rate": 5.380540701960627e-06, "loss": 1.6642, "step": 4608 }, { "epoch": 0.49, "learning_rate": 5.378818211075644e-06, "loss": 1.4084, "step": 4609 }, { "epoch": 0.49, "learning_rate": 5.377095674972155e-06, "loss": 1.5082, "step": 4610 }, { "epoch": 0.49, "learning_rate": 5.375373093855774e-06, "loss": 1.6122, "step": 4611 }, { "epoch": 0.49, "learning_rate": 5.373650467932122e-06, "loss": 1.3771, "step": 4612 }, { "epoch": 0.49, "learning_rate": 5.371927797406824e-06, "loss": 1.6553, "step": 4613 }, { "epoch": 0.49, "learning_rate": 5.37020508248551e-06, "loss": 1.3906, "step": 4614 }, { "epoch": 0.49, "learning_rate": 5.368482323373815e-06, "loss": 1.3131, "step": 4615 }, { "epoch": 0.49, "learning_rate": 5.3667595202773835e-06, "loss": 1.5811, "step": 4616 }, { "epoch": 0.49, "learning_rate": 5.365036673401857e-06, "loss": 1.5014, "step": 4617 }, { "epoch": 0.49, "learning_rate": 5.36331378295289e-06, "loss": 1.2001, "step": 4618 }, { "epoch": 0.49, "learning_rate": 5.361590849136141e-06, "loss": 1.4875, "step": 4619 }, { "epoch": 0.49, "learning_rate": 5.359867872157267e-06, "loss": 1.5337, "step": 4620 }, { "epoch": 0.49, "learning_rate": 5.3581448522219405e-06, "loss": 1.5971, "step": 4621 }, { "epoch": 0.49, "learning_rate": 5.35642178953583e-06, "loss": 1.2846, "step": 4622 }, { "epoch": 0.49, "learning_rate": 5.354698684304613e-06, "loss": 1.6626, "step": 4623 }, { "epoch": 0.49, "learning_rate": 5.352975536733974e-06, "loss": 1.2944, "step": 4624 }, { "epoch": 0.49, "learning_rate": 5.3512523470295995e-06, "loss": 1.2513, "step": 4625 }, { "epoch": 0.49, "learning_rate": 5.3495291153971806e-06, "loss": 1.4385, "step": 4626 }, { "epoch": 0.49, "learning_rate": 5.347805842042417e-06, "loss": 1.4379, "step": 4627 }, { "epoch": 0.49, "learning_rate": 5.34608252717101e-06, "loss": 1.6149, "step": 4628 }, { "epoch": 0.49, "learning_rate": 5.344359170988668e-06, "loss": 1.4197, "step": 4629 }, { "epoch": 0.49, "learning_rate": 5.342635773701102e-06, "loss": 1.488, "step": 4630 }, { "epoch": 0.49, "learning_rate": 5.34091233551403e-06, "loss": 1.4208, "step": 4631 }, { "epoch": 0.49, "learning_rate": 5.339188856633173e-06, "loss": 1.4496, "step": 4632 }, { "epoch": 0.49, "learning_rate": 5.337465337264262e-06, "loss": 1.2952, "step": 4633 }, { "epoch": 0.49, "learning_rate": 5.335741777613023e-06, "loss": 1.2487, "step": 4634 }, { "epoch": 0.49, "learning_rate": 5.3340181778851954e-06, "loss": 1.3278, "step": 4635 }, { "epoch": 0.49, "learning_rate": 5.332294538286523e-06, "loss": 1.5681, "step": 4636 }, { "epoch": 0.49, "learning_rate": 5.330570859022748e-06, "loss": 1.5843, "step": 4637 }, { "epoch": 0.49, "learning_rate": 5.328847140299624e-06, "loss": 1.2359, "step": 4638 }, { "epoch": 0.49, "learning_rate": 5.3271233823229064e-06, "loss": 1.3935, "step": 4639 }, { "epoch": 0.49, "learning_rate": 5.325399585298354e-06, "loss": 1.5123, "step": 4640 }, { "epoch": 0.5, "learning_rate": 5.323675749431732e-06, "loss": 1.5969, "step": 4641 }, { "epoch": 0.5, "learning_rate": 5.321951874928812e-06, "loss": 1.5937, "step": 4642 }, { "epoch": 0.5, "learning_rate": 5.3202279619953675e-06, "loss": 1.2445, "step": 4643 }, { "epoch": 0.5, "learning_rate": 5.318504010837175e-06, "loss": 1.2732, "step": 4644 }, { "epoch": 0.5, "learning_rate": 5.3167800216600215e-06, "loss": 1.3418, "step": 4645 }, { "epoch": 0.5, "learning_rate": 5.3150559946696934e-06, "loss": 1.29, "step": 4646 }, { "epoch": 0.5, "learning_rate": 5.313331930071981e-06, "loss": 1.5282, "step": 4647 }, { "epoch": 0.5, "learning_rate": 5.311607828072685e-06, "loss": 1.4497, "step": 4648 }, { "epoch": 0.5, "learning_rate": 5.309883688877601e-06, "loss": 1.639, "step": 4649 }, { "epoch": 0.5, "learning_rate": 5.308159512692544e-06, "loss": 1.4362, "step": 4650 }, { "epoch": 0.5, "learning_rate": 5.306435299723317e-06, "loss": 1.7525, "step": 4651 }, { "epoch": 0.5, "learning_rate": 5.304711050175735e-06, "loss": 1.3674, "step": 4652 }, { "epoch": 0.5, "learning_rate": 5.302986764255621e-06, "loss": 1.5436, "step": 4653 }, { "epoch": 0.5, "learning_rate": 5.3012624421687945e-06, "loss": 1.4772, "step": 4654 }, { "epoch": 0.5, "learning_rate": 5.2995380841210845e-06, "loss": 1.294, "step": 4655 }, { "epoch": 0.5, "learning_rate": 5.297813690318325e-06, "loss": 1.3849, "step": 4656 }, { "epoch": 0.5, "learning_rate": 5.296089260966347e-06, "loss": 1.3662, "step": 4657 }, { "epoch": 0.5, "learning_rate": 5.2943647962709975e-06, "loss": 1.31, "step": 4658 }, { "epoch": 0.5, "learning_rate": 5.292640296438116e-06, "loss": 1.2653, "step": 4659 }, { "epoch": 0.5, "learning_rate": 5.290915761673554e-06, "loss": 1.6861, "step": 4660 }, { "epoch": 0.5, "learning_rate": 5.289191192183165e-06, "loss": 1.3562, "step": 4661 }, { "epoch": 0.5, "learning_rate": 5.287466588172804e-06, "loss": 1.5386, "step": 4662 }, { "epoch": 0.5, "learning_rate": 5.2857419498483345e-06, "loss": 1.4917, "step": 4663 }, { "epoch": 0.5, "learning_rate": 5.284017277415619e-06, "loss": 1.4329, "step": 4664 }, { "epoch": 0.5, "learning_rate": 5.2822925710805305e-06, "loss": 1.4181, "step": 4665 }, { "epoch": 0.5, "learning_rate": 5.2805678310489405e-06, "loss": 1.2875, "step": 4666 }, { "epoch": 0.5, "learning_rate": 5.2788430575267266e-06, "loss": 1.4601, "step": 4667 }, { "epoch": 0.5, "learning_rate": 5.27711825071977e-06, "loss": 1.6751, "step": 4668 }, { "epoch": 0.5, "learning_rate": 5.275393410833959e-06, "loss": 1.2538, "step": 4669 }, { "epoch": 0.5, "learning_rate": 5.273668538075179e-06, "loss": 1.3918, "step": 4670 }, { "epoch": 0.5, "learning_rate": 5.2719436326493255e-06, "loss": 1.3882, "step": 4671 }, { "epoch": 0.5, "learning_rate": 5.270218694762297e-06, "loss": 1.4891, "step": 4672 }, { "epoch": 0.5, "learning_rate": 5.268493724619993e-06, "loss": 1.4213, "step": 4673 }, { "epoch": 0.5, "learning_rate": 5.266768722428318e-06, "loss": 1.2989, "step": 4674 }, { "epoch": 0.5, "learning_rate": 5.265043688393183e-06, "loss": 1.4651, "step": 4675 }, { "epoch": 0.5, "learning_rate": 5.263318622720498e-06, "loss": 1.5526, "step": 4676 }, { "epoch": 0.5, "learning_rate": 5.261593525616181e-06, "loss": 1.8564, "step": 4677 }, { "epoch": 0.5, "learning_rate": 5.259868397286154e-06, "loss": 1.576, "step": 4678 }, { "epoch": 0.5, "learning_rate": 5.258143237936336e-06, "loss": 1.3168, "step": 4679 }, { "epoch": 0.5, "learning_rate": 5.256418047772659e-06, "loss": 1.3778, "step": 4680 }, { "epoch": 0.5, "learning_rate": 5.254692827001053e-06, "loss": 1.4503, "step": 4681 }, { "epoch": 0.5, "learning_rate": 5.252967575827452e-06, "loss": 1.5241, "step": 4682 }, { "epoch": 0.5, "learning_rate": 5.251242294457796e-06, "loss": 1.352, "step": 4683 }, { "epoch": 0.5, "learning_rate": 5.249516983098026e-06, "loss": 1.5435, "step": 4684 }, { "epoch": 0.5, "learning_rate": 5.247791641954089e-06, "loss": 1.7242, "step": 4685 }, { "epoch": 0.5, "learning_rate": 5.2460662712319335e-06, "loss": 1.5868, "step": 4686 }, { "epoch": 0.5, "learning_rate": 5.244340871137511e-06, "loss": 1.4614, "step": 4687 }, { "epoch": 0.5, "learning_rate": 5.242615441876779e-06, "loss": 1.4287, "step": 4688 }, { "epoch": 0.5, "learning_rate": 5.240889983655701e-06, "loss": 1.3065, "step": 4689 }, { "epoch": 0.5, "learning_rate": 5.239164496680233e-06, "loss": 1.3475, "step": 4690 }, { "epoch": 0.5, "learning_rate": 5.237438981156346e-06, "loss": 1.2121, "step": 4691 }, { "epoch": 0.5, "learning_rate": 5.235713437290012e-06, "loss": 1.5421, "step": 4692 }, { "epoch": 0.5, "learning_rate": 5.2339878652872e-06, "loss": 1.5476, "step": 4693 }, { "epoch": 0.5, "learning_rate": 5.232262265353889e-06, "loss": 1.5073, "step": 4694 }, { "epoch": 0.5, "learning_rate": 5.230536637696062e-06, "loss": 1.2978, "step": 4695 }, { "epoch": 0.5, "learning_rate": 5.228810982519696e-06, "loss": 1.5268, "step": 4696 }, { "epoch": 0.5, "learning_rate": 5.2270853000307845e-06, "loss": 1.4877, "step": 4697 }, { "epoch": 0.5, "learning_rate": 5.225359590435312e-06, "loss": 1.3657, "step": 4698 }, { "epoch": 0.5, "learning_rate": 5.223633853939276e-06, "loss": 1.4007, "step": 4699 }, { "epoch": 0.5, "learning_rate": 5.2219080907486715e-06, "loss": 1.3385, "step": 4700 }, { "epoch": 0.5, "learning_rate": 5.220182301069499e-06, "loss": 1.4003, "step": 4701 }, { "epoch": 0.5, "learning_rate": 5.2184564851077575e-06, "loss": 1.3084, "step": 4702 }, { "epoch": 0.5, "learning_rate": 5.216730643069458e-06, "loss": 1.5602, "step": 4703 }, { "epoch": 0.5, "learning_rate": 5.215004775160608e-06, "loss": 1.4169, "step": 4704 }, { "epoch": 0.5, "learning_rate": 5.213278881587217e-06, "loss": 1.4767, "step": 4705 }, { "epoch": 0.5, "learning_rate": 5.211552962555305e-06, "loss": 1.48, "step": 4706 }, { "epoch": 0.5, "learning_rate": 5.209827018270886e-06, "loss": 1.3514, "step": 4707 }, { "epoch": 0.5, "learning_rate": 5.208101048939984e-06, "loss": 1.3629, "step": 4708 }, { "epoch": 0.5, "learning_rate": 5.2063750547686224e-06, "loss": 1.4685, "step": 4709 }, { "epoch": 0.5, "learning_rate": 5.204649035962825e-06, "loss": 1.6281, "step": 4710 }, { "epoch": 0.5, "learning_rate": 5.202922992728631e-06, "loss": 1.378, "step": 4711 }, { "epoch": 0.5, "learning_rate": 5.201196925272064e-06, "loss": 1.2221, "step": 4712 }, { "epoch": 0.5, "learning_rate": 5.199470833799164e-06, "loss": 1.4585, "step": 4713 }, { "epoch": 0.5, "learning_rate": 5.197744718515973e-06, "loss": 1.7642, "step": 4714 }, { "epoch": 0.5, "learning_rate": 5.196018579628526e-06, "loss": 1.2759, "step": 4715 }, { "epoch": 0.5, "learning_rate": 5.1942924173428725e-06, "loss": 1.3036, "step": 4716 }, { "epoch": 0.5, "learning_rate": 5.192566231865058e-06, "loss": 1.6001, "step": 4717 }, { "epoch": 0.5, "learning_rate": 5.190840023401132e-06, "loss": 1.5722, "step": 4718 }, { "epoch": 0.5, "learning_rate": 5.18911379215715e-06, "loss": 1.3172, "step": 4719 }, { "epoch": 0.5, "learning_rate": 5.1873875383391655e-06, "loss": 1.684, "step": 4720 }, { "epoch": 0.5, "learning_rate": 5.185661262153236e-06, "loss": 1.4706, "step": 4721 }, { "epoch": 0.5, "learning_rate": 5.1839349638054245e-06, "loss": 1.4558, "step": 4722 }, { "epoch": 0.5, "learning_rate": 5.1822086435017936e-06, "loss": 1.1381, "step": 4723 }, { "epoch": 0.5, "learning_rate": 5.18048230144841e-06, "loss": 1.5455, "step": 4724 }, { "epoch": 0.5, "learning_rate": 5.178755937851341e-06, "loss": 1.3818, "step": 4725 }, { "epoch": 0.5, "learning_rate": 5.177029552916659e-06, "loss": 1.2123, "step": 4726 }, { "epoch": 0.5, "learning_rate": 5.175303146850439e-06, "loss": 1.5312, "step": 4727 }, { "epoch": 0.5, "learning_rate": 5.173576719858755e-06, "loss": 1.6638, "step": 4728 }, { "epoch": 0.5, "learning_rate": 5.171850272147687e-06, "loss": 1.3012, "step": 4729 }, { "epoch": 0.5, "learning_rate": 5.170123803923316e-06, "loss": 1.4693, "step": 4730 }, { "epoch": 0.5, "learning_rate": 5.168397315391729e-06, "loss": 1.4444, "step": 4731 }, { "epoch": 0.5, "learning_rate": 5.166670806759007e-06, "loss": 1.7633, "step": 4732 }, { "epoch": 0.5, "learning_rate": 5.164944278231242e-06, "loss": 1.4804, "step": 4733 }, { "epoch": 0.5, "learning_rate": 5.1632177300145255e-06, "loss": 1.5416, "step": 4734 }, { "epoch": 0.51, "learning_rate": 5.161491162314949e-06, "loss": 1.3296, "step": 4735 }, { "epoch": 0.51, "learning_rate": 5.159764575338608e-06, "loss": 1.3347, "step": 4736 }, { "epoch": 0.51, "learning_rate": 5.1580379692916025e-06, "loss": 1.4405, "step": 4737 }, { "epoch": 0.51, "learning_rate": 5.156311344380033e-06, "loss": 1.0613, "step": 4738 }, { "epoch": 0.51, "learning_rate": 5.1545847008099995e-06, "loss": 1.4948, "step": 4739 }, { "epoch": 0.51, "learning_rate": 5.152858038787608e-06, "loss": 1.2682, "step": 4740 }, { "epoch": 0.51, "learning_rate": 5.151131358518966e-06, "loss": 1.3251, "step": 4741 }, { "epoch": 0.51, "learning_rate": 5.1494046602101825e-06, "loss": 1.4175, "step": 4742 }, { "epoch": 0.51, "learning_rate": 5.147677944067368e-06, "loss": 1.5346, "step": 4743 }, { "epoch": 0.51, "learning_rate": 5.145951210296637e-06, "loss": 1.5554, "step": 4744 }, { "epoch": 0.51, "learning_rate": 5.144224459104104e-06, "loss": 1.3666, "step": 4745 }, { "epoch": 0.51, "learning_rate": 5.142497690695888e-06, "loss": 1.372, "step": 4746 }, { "epoch": 0.51, "learning_rate": 5.140770905278106e-06, "loss": 1.4075, "step": 4747 }, { "epoch": 0.51, "learning_rate": 5.139044103056885e-06, "loss": 1.4138, "step": 4748 }, { "epoch": 0.51, "learning_rate": 5.137317284238344e-06, "loss": 1.5521, "step": 4749 }, { "epoch": 0.51, "learning_rate": 5.135590449028608e-06, "loss": 1.3541, "step": 4750 }, { "epoch": 0.51, "learning_rate": 5.133863597633809e-06, "loss": 1.6164, "step": 4751 }, { "epoch": 0.51, "learning_rate": 5.1321367302600726e-06, "loss": 1.57, "step": 4752 }, { "epoch": 0.51, "learning_rate": 5.130409847113532e-06, "loss": 1.4103, "step": 4753 }, { "epoch": 0.51, "learning_rate": 5.128682948400321e-06, "loss": 1.3269, "step": 4754 }, { "epoch": 0.51, "learning_rate": 5.126956034326573e-06, "loss": 1.2726, "step": 4755 }, { "epoch": 0.51, "learning_rate": 5.125229105098429e-06, "loss": 1.6546, "step": 4756 }, { "epoch": 0.51, "learning_rate": 5.123502160922024e-06, "loss": 1.5363, "step": 4757 }, { "epoch": 0.51, "learning_rate": 5.121775202003499e-06, "loss": 1.4185, "step": 4758 }, { "epoch": 0.51, "learning_rate": 5.120048228548999e-06, "loss": 1.1998, "step": 4759 }, { "epoch": 0.51, "learning_rate": 5.1183212407646655e-06, "loss": 1.2073, "step": 4760 }, { "epoch": 0.51, "learning_rate": 5.116594238856645e-06, "loss": 1.4795, "step": 4761 }, { "epoch": 0.51, "learning_rate": 5.114867223031086e-06, "loss": 1.3469, "step": 4762 }, { "epoch": 0.51, "learning_rate": 5.113140193494136e-06, "loss": 1.405, "step": 4763 }, { "epoch": 0.51, "learning_rate": 5.111413150451948e-06, "loss": 1.4193, "step": 4764 }, { "epoch": 0.51, "learning_rate": 5.109686094110674e-06, "loss": 1.2033, "step": 4765 }, { "epoch": 0.51, "learning_rate": 5.107959024676466e-06, "loss": 1.3551, "step": 4766 }, { "epoch": 0.51, "learning_rate": 5.1062319423554815e-06, "loss": 1.7691, "step": 4767 }, { "epoch": 0.51, "learning_rate": 5.104504847353878e-06, "loss": 1.4156, "step": 4768 }, { "epoch": 0.51, "learning_rate": 5.102777739877812e-06, "loss": 1.1243, "step": 4769 }, { "epoch": 0.51, "learning_rate": 5.101050620133447e-06, "loss": 1.6259, "step": 4770 }, { "epoch": 0.51, "learning_rate": 5.099323488326941e-06, "loss": 1.439, "step": 4771 }, { "epoch": 0.51, "learning_rate": 5.097596344664458e-06, "loss": 1.2113, "step": 4772 }, { "epoch": 0.51, "learning_rate": 5.095869189352166e-06, "loss": 1.5156, "step": 4773 }, { "epoch": 0.51, "learning_rate": 5.094142022596224e-06, "loss": 1.5406, "step": 4774 }, { "epoch": 0.51, "learning_rate": 5.092414844602807e-06, "loss": 1.39, "step": 4775 }, { "epoch": 0.51, "learning_rate": 5.090687655578078e-06, "loss": 1.512, "step": 4776 }, { "epoch": 0.51, "learning_rate": 5.088960455728209e-06, "loss": 1.476, "step": 4777 }, { "epoch": 0.51, "learning_rate": 5.087233245259371e-06, "loss": 1.4776, "step": 4778 }, { "epoch": 0.51, "learning_rate": 5.0855060243777366e-06, "loss": 1.3126, "step": 4779 }, { "epoch": 0.51, "learning_rate": 5.0837787932894786e-06, "loss": 1.2883, "step": 4780 }, { "epoch": 0.51, "learning_rate": 5.082051552200773e-06, "loss": 1.6291, "step": 4781 }, { "epoch": 0.51, "learning_rate": 5.080324301317795e-06, "loss": 1.3908, "step": 4782 }, { "epoch": 0.51, "learning_rate": 5.078597040846723e-06, "loss": 1.595, "step": 4783 }, { "epoch": 0.51, "learning_rate": 5.076869770993734e-06, "loss": 1.3572, "step": 4784 }, { "epoch": 0.51, "learning_rate": 5.0751424919650085e-06, "loss": 1.5489, "step": 4785 }, { "epoch": 0.51, "learning_rate": 5.073415203966726e-06, "loss": 1.3773, "step": 4786 }, { "epoch": 0.51, "learning_rate": 5.0716879072050704e-06, "loss": 1.4819, "step": 4787 }, { "epoch": 0.51, "learning_rate": 5.069960601886224e-06, "loss": 1.247, "step": 4788 }, { "epoch": 0.51, "learning_rate": 5.068233288216367e-06, "loss": 1.831, "step": 4789 }, { "epoch": 0.51, "learning_rate": 5.066505966401689e-06, "loss": 1.487, "step": 4790 }, { "epoch": 0.51, "learning_rate": 5.064778636648371e-06, "loss": 1.4475, "step": 4791 }, { "epoch": 0.51, "learning_rate": 5.063051299162604e-06, "loss": 1.4793, "step": 4792 }, { "epoch": 0.51, "learning_rate": 5.061323954150576e-06, "loss": 1.4535, "step": 4793 }, { "epoch": 0.51, "learning_rate": 5.05959660181847e-06, "loss": 1.5332, "step": 4794 }, { "epoch": 0.51, "learning_rate": 5.057869242372479e-06, "loss": 1.7096, "step": 4795 }, { "epoch": 0.51, "learning_rate": 5.0561418760187944e-06, "loss": 1.4532, "step": 4796 }, { "epoch": 0.51, "learning_rate": 5.054414502963605e-06, "loss": 1.4203, "step": 4797 }, { "epoch": 0.51, "learning_rate": 5.052687123413104e-06, "loss": 1.4505, "step": 4798 }, { "epoch": 0.51, "learning_rate": 5.050959737573483e-06, "loss": 1.2882, "step": 4799 }, { "epoch": 0.51, "learning_rate": 5.049232345650936e-06, "loss": 1.5384, "step": 4800 }, { "epoch": 0.51, "learning_rate": 5.047504947851657e-06, "loss": 1.4328, "step": 4801 }, { "epoch": 0.51, "learning_rate": 5.0457775443818415e-06, "loss": 1.374, "step": 4802 }, { "epoch": 0.51, "learning_rate": 5.044050135447682e-06, "loss": 1.3822, "step": 4803 }, { "epoch": 0.51, "learning_rate": 5.042322721255379e-06, "loss": 1.4737, "step": 4804 }, { "epoch": 0.51, "learning_rate": 5.040595302011127e-06, "loss": 1.5163, "step": 4805 }, { "epoch": 0.51, "learning_rate": 5.038867877921124e-06, "loss": 1.6706, "step": 4806 }, { "epoch": 0.51, "learning_rate": 5.037140449191567e-06, "loss": 1.5594, "step": 4807 }, { "epoch": 0.51, "learning_rate": 5.035413016028653e-06, "loss": 1.2145, "step": 4808 }, { "epoch": 0.51, "learning_rate": 5.033685578638586e-06, "loss": 1.5149, "step": 4809 }, { "epoch": 0.51, "learning_rate": 5.031958137227562e-06, "loss": 1.2334, "step": 4810 }, { "epoch": 0.51, "learning_rate": 5.030230692001779e-06, "loss": 1.582, "step": 4811 }, { "epoch": 0.51, "learning_rate": 5.028503243167443e-06, "loss": 1.2557, "step": 4812 }, { "epoch": 0.51, "learning_rate": 5.02677579093075e-06, "loss": 1.2249, "step": 4813 }, { "epoch": 0.51, "learning_rate": 5.0250483354979045e-06, "loss": 1.6035, "step": 4814 }, { "epoch": 0.51, "learning_rate": 5.023320877075107e-06, "loss": 1.5882, "step": 4815 }, { "epoch": 0.51, "learning_rate": 5.021593415868558e-06, "loss": 1.0941, "step": 4816 }, { "epoch": 0.51, "learning_rate": 5.019865952084462e-06, "loss": 1.1969, "step": 4817 }, { "epoch": 0.51, "learning_rate": 5.0181384859290215e-06, "loss": 1.4317, "step": 4818 }, { "epoch": 0.51, "learning_rate": 5.016411017608438e-06, "loss": 1.5349, "step": 4819 }, { "epoch": 0.51, "learning_rate": 5.014683547328917e-06, "loss": 1.2958, "step": 4820 }, { "epoch": 0.51, "learning_rate": 5.01295607529666e-06, "loss": 1.3078, "step": 4821 }, { "epoch": 0.51, "learning_rate": 5.011228601717871e-06, "loss": 1.7194, "step": 4822 }, { "epoch": 0.51, "learning_rate": 5.009501126798755e-06, "loss": 1.4246, "step": 4823 }, { "epoch": 0.51, "learning_rate": 5.007773650745514e-06, "loss": 1.3798, "step": 4824 }, { "epoch": 0.51, "learning_rate": 5.006046173764353e-06, "loss": 1.7076, "step": 4825 }, { "epoch": 0.51, "learning_rate": 5.004318696061479e-06, "loss": 1.3243, "step": 4826 }, { "epoch": 0.51, "learning_rate": 5.0025912178430925e-06, "loss": 1.2125, "step": 4827 }, { "epoch": 0.51, "learning_rate": 5.000863739315398e-06, "loss": 1.3812, "step": 4828 }, { "epoch": 0.52, "learning_rate": 4.9991362606846025e-06, "loss": 1.5404, "step": 4829 }, { "epoch": 0.52, "learning_rate": 4.997408782156909e-06, "loss": 1.6592, "step": 4830 }, { "epoch": 0.52, "learning_rate": 4.995681303938523e-06, "loss": 1.3983, "step": 4831 }, { "epoch": 0.52, "learning_rate": 4.993953826235649e-06, "loss": 1.5853, "step": 4832 }, { "epoch": 0.52, "learning_rate": 4.9922263492544885e-06, "loss": 1.3817, "step": 4833 }, { "epoch": 0.52, "learning_rate": 4.990498873201247e-06, "loss": 1.6525, "step": 4834 }, { "epoch": 0.52, "learning_rate": 4.988771398282131e-06, "loss": 1.3511, "step": 4835 }, { "epoch": 0.52, "learning_rate": 4.987043924703342e-06, "loss": 1.3153, "step": 4836 }, { "epoch": 0.52, "learning_rate": 4.985316452671084e-06, "loss": 1.3121, "step": 4837 }, { "epoch": 0.52, "learning_rate": 4.983588982391564e-06, "loss": 1.5157, "step": 4838 }, { "epoch": 0.52, "learning_rate": 4.981861514070979e-06, "loss": 1.3177, "step": 4839 }, { "epoch": 0.52, "learning_rate": 4.980134047915539e-06, "loss": 1.2856, "step": 4840 }, { "epoch": 0.52, "learning_rate": 4.978406584131443e-06, "loss": 1.5917, "step": 4841 }, { "epoch": 0.52, "learning_rate": 4.976679122924896e-06, "loss": 1.4166, "step": 4842 }, { "epoch": 0.52, "learning_rate": 4.974951664502097e-06, "loss": 1.5571, "step": 4843 }, { "epoch": 0.52, "learning_rate": 4.97322420906925e-06, "loss": 1.5531, "step": 4844 }, { "epoch": 0.52, "learning_rate": 4.971496756832557e-06, "loss": 1.5705, "step": 4845 }, { "epoch": 0.52, "learning_rate": 4.9697693079982215e-06, "loss": 1.5975, "step": 4846 }, { "epoch": 0.52, "learning_rate": 4.968041862772441e-06, "loss": 1.5529, "step": 4847 }, { "epoch": 0.52, "learning_rate": 4.966314421361416e-06, "loss": 1.5058, "step": 4848 }, { "epoch": 0.52, "learning_rate": 4.964586983971347e-06, "loss": 1.1599, "step": 4849 }, { "epoch": 0.52, "learning_rate": 4.962859550808435e-06, "loss": 1.3945, "step": 4850 }, { "epoch": 0.52, "learning_rate": 4.9611321220788775e-06, "loss": 1.1028, "step": 4851 }, { "epoch": 0.52, "learning_rate": 4.959404697988875e-06, "loss": 1.4249, "step": 4852 }, { "epoch": 0.52, "learning_rate": 4.957677278744621e-06, "loss": 1.4115, "step": 4853 }, { "epoch": 0.52, "learning_rate": 4.955949864552318e-06, "loss": 1.5169, "step": 4854 }, { "epoch": 0.52, "learning_rate": 4.954222455618161e-06, "loss": 1.4346, "step": 4855 }, { "epoch": 0.52, "learning_rate": 4.952495052148344e-06, "loss": 1.2917, "step": 4856 }, { "epoch": 0.52, "learning_rate": 4.950767654349067e-06, "loss": 1.3591, "step": 4857 }, { "epoch": 0.52, "learning_rate": 4.949040262426518e-06, "loss": 1.2794, "step": 4858 }, { "epoch": 0.52, "learning_rate": 4.947312876586898e-06, "loss": 1.2643, "step": 4859 }, { "epoch": 0.52, "learning_rate": 4.945585497036396e-06, "loss": 1.4564, "step": 4860 }, { "epoch": 0.52, "learning_rate": 4.943858123981208e-06, "loss": 1.2717, "step": 4861 }, { "epoch": 0.52, "learning_rate": 4.942130757627522e-06, "loss": 1.5001, "step": 4862 }, { "epoch": 0.52, "learning_rate": 4.940403398181531e-06, "loss": 1.6142, "step": 4863 }, { "epoch": 0.52, "learning_rate": 4.938676045849427e-06, "loss": 1.337, "step": 4864 }, { "epoch": 0.52, "learning_rate": 4.936948700837397e-06, "loss": 1.2553, "step": 4865 }, { "epoch": 0.52, "learning_rate": 4.935221363351631e-06, "loss": 1.4294, "step": 4866 }, { "epoch": 0.52, "learning_rate": 4.933494033598314e-06, "loss": 1.4088, "step": 4867 }, { "epoch": 0.52, "learning_rate": 4.931766711783633e-06, "loss": 1.5777, "step": 4868 }, { "epoch": 0.52, "learning_rate": 4.930039398113779e-06, "loss": 1.5665, "step": 4869 }, { "epoch": 0.52, "learning_rate": 4.92831209279493e-06, "loss": 1.4723, "step": 4870 }, { "epoch": 0.52, "learning_rate": 4.9265847960332756e-06, "loss": 1.5019, "step": 4871 }, { "epoch": 0.52, "learning_rate": 4.924857508034994e-06, "loss": 1.442, "step": 4872 }, { "epoch": 0.52, "learning_rate": 4.923130229006267e-06, "loss": 1.3717, "step": 4873 }, { "epoch": 0.52, "learning_rate": 4.9214029591532785e-06, "loss": 1.3486, "step": 4874 }, { "epoch": 0.52, "learning_rate": 4.919675698682206e-06, "loss": 1.1702, "step": 4875 }, { "epoch": 0.52, "learning_rate": 4.917948447799228e-06, "loss": 1.3039, "step": 4876 }, { "epoch": 0.52, "learning_rate": 4.916221206710524e-06, "loss": 1.5443, "step": 4877 }, { "epoch": 0.52, "learning_rate": 4.914493975622263e-06, "loss": 1.2944, "step": 4878 }, { "epoch": 0.52, "learning_rate": 4.91276675474063e-06, "loss": 1.4536, "step": 4879 }, { "epoch": 0.52, "learning_rate": 4.911039544271792e-06, "loss": 1.6742, "step": 4880 }, { "epoch": 0.52, "learning_rate": 4.909312344421923e-06, "loss": 1.5753, "step": 4881 }, { "epoch": 0.52, "learning_rate": 4.907585155397195e-06, "loss": 1.3838, "step": 4882 }, { "epoch": 0.52, "learning_rate": 4.905857977403775e-06, "loss": 1.2279, "step": 4883 }, { "epoch": 0.52, "learning_rate": 4.904130810647836e-06, "loss": 1.547, "step": 4884 }, { "epoch": 0.52, "learning_rate": 4.902403655335543e-06, "loss": 1.4704, "step": 4885 }, { "epoch": 0.52, "learning_rate": 4.900676511673061e-06, "loss": 1.2927, "step": 4886 }, { "epoch": 0.52, "learning_rate": 4.898949379866556e-06, "loss": 1.4531, "step": 4887 }, { "epoch": 0.52, "learning_rate": 4.897222260122189e-06, "loss": 1.4943, "step": 4888 }, { "epoch": 0.52, "learning_rate": 4.895495152646124e-06, "loss": 1.3158, "step": 4889 }, { "epoch": 0.52, "learning_rate": 4.893768057644519e-06, "loss": 1.3163, "step": 4890 }, { "epoch": 0.52, "learning_rate": 4.892040975323536e-06, "loss": 1.5489, "step": 4891 }, { "epoch": 0.52, "learning_rate": 4.890313905889329e-06, "loss": 1.45, "step": 4892 }, { "epoch": 0.52, "learning_rate": 4.888586849548053e-06, "loss": 1.4089, "step": 4893 }, { "epoch": 0.52, "learning_rate": 4.886859806505865e-06, "loss": 1.4021, "step": 4894 }, { "epoch": 0.52, "learning_rate": 4.885132776968915e-06, "loss": 1.3734, "step": 4895 }, { "epoch": 0.52, "learning_rate": 4.883405761143357e-06, "loss": 1.4385, "step": 4896 }, { "epoch": 0.52, "learning_rate": 4.881678759235335e-06, "loss": 1.5719, "step": 4897 }, { "epoch": 0.52, "learning_rate": 4.879951771451002e-06, "loss": 1.4981, "step": 4898 }, { "epoch": 0.52, "learning_rate": 4.878224797996502e-06, "loss": 1.5948, "step": 4899 }, { "epoch": 0.52, "learning_rate": 4.876497839077978e-06, "loss": 1.4456, "step": 4900 }, { "epoch": 0.52, "learning_rate": 4.8747708949015725e-06, "loss": 1.404, "step": 4901 }, { "epoch": 0.52, "learning_rate": 4.873043965673427e-06, "loss": 1.455, "step": 4902 }, { "epoch": 0.52, "learning_rate": 4.87131705159968e-06, "loss": 1.3492, "step": 4903 }, { "epoch": 0.52, "learning_rate": 4.86959015288647e-06, "loss": 1.5838, "step": 4904 }, { "epoch": 0.52, "learning_rate": 4.86786326973993e-06, "loss": 1.5582, "step": 4905 }, { "epoch": 0.52, "learning_rate": 4.866136402366194e-06, "loss": 1.4349, "step": 4906 }, { "epoch": 0.52, "learning_rate": 4.864409550971393e-06, "loss": 1.3117, "step": 4907 }, { "epoch": 0.52, "learning_rate": 4.862682715761658e-06, "loss": 1.4183, "step": 4908 }, { "epoch": 0.52, "learning_rate": 4.860955896943117e-06, "loss": 1.7456, "step": 4909 }, { "epoch": 0.52, "learning_rate": 4.859229094721895e-06, "loss": 1.3125, "step": 4910 }, { "epoch": 0.52, "learning_rate": 4.857502309304114e-06, "loss": 1.5363, "step": 4911 }, { "epoch": 0.52, "learning_rate": 4.8557755408958965e-06, "loss": 1.4431, "step": 4912 }, { "epoch": 0.52, "learning_rate": 4.854048789703364e-06, "loss": 1.4369, "step": 4913 }, { "epoch": 0.52, "learning_rate": 4.852322055932633e-06, "loss": 1.4609, "step": 4914 }, { "epoch": 0.52, "learning_rate": 4.850595339789819e-06, "loss": 1.5675, "step": 4915 }, { "epoch": 0.52, "learning_rate": 4.848868641481036e-06, "loss": 1.276, "step": 4916 }, { "epoch": 0.52, "learning_rate": 4.8471419612123925e-06, "loss": 1.4314, "step": 4917 }, { "epoch": 0.52, "learning_rate": 4.845415299190002e-06, "loss": 1.1646, "step": 4918 }, { "epoch": 0.52, "learning_rate": 4.843688655619969e-06, "loss": 1.3746, "step": 4919 }, { "epoch": 0.52, "learning_rate": 4.841962030708398e-06, "loss": 1.2057, "step": 4920 }, { "epoch": 0.52, "learning_rate": 4.840235424661393e-06, "loss": 1.4111, "step": 4921 }, { "epoch": 0.53, "learning_rate": 4.8385088376850516e-06, "loss": 1.367, "step": 4922 }, { "epoch": 0.53, "learning_rate": 4.836782269985475e-06, "loss": 1.4864, "step": 4923 }, { "epoch": 0.53, "learning_rate": 4.83505572176876e-06, "loss": 1.4426, "step": 4924 }, { "epoch": 0.53, "learning_rate": 4.833329193240995e-06, "loss": 1.2785, "step": 4925 }, { "epoch": 0.53, "learning_rate": 4.831602684608274e-06, "loss": 1.3577, "step": 4926 }, { "epoch": 0.53, "learning_rate": 4.829876196076684e-06, "loss": 1.2557, "step": 4927 }, { "epoch": 0.53, "learning_rate": 4.828149727852314e-06, "loss": 1.4041, "step": 4928 }, { "epoch": 0.53, "learning_rate": 4.826423280141247e-06, "loss": 1.2779, "step": 4929 }, { "epoch": 0.53, "learning_rate": 4.824696853149564e-06, "loss": 1.3834, "step": 4930 }, { "epoch": 0.53, "learning_rate": 4.822970447083343e-06, "loss": 1.527, "step": 4931 }, { "epoch": 0.53, "learning_rate": 4.82124406214866e-06, "loss": 1.2764, "step": 4932 }, { "epoch": 0.53, "learning_rate": 4.819517698551592e-06, "loss": 1.3286, "step": 4933 }, { "epoch": 0.53, "learning_rate": 4.817791356498207e-06, "loss": 1.4292, "step": 4934 }, { "epoch": 0.53, "learning_rate": 4.816065036194576e-06, "loss": 1.57, "step": 4935 }, { "epoch": 0.53, "learning_rate": 4.814338737846766e-06, "loss": 1.445, "step": 4936 }, { "epoch": 0.53, "learning_rate": 4.812612461660835e-06, "loss": 1.5247, "step": 4937 }, { "epoch": 0.53, "learning_rate": 4.810886207842852e-06, "loss": 1.1951, "step": 4938 }, { "epoch": 0.53, "learning_rate": 4.8091599765988685e-06, "loss": 1.4768, "step": 4939 }, { "epoch": 0.53, "learning_rate": 4.807433768134944e-06, "loss": 1.6408, "step": 4940 }, { "epoch": 0.53, "learning_rate": 4.80570758265713e-06, "loss": 1.7192, "step": 4941 }, { "epoch": 0.53, "learning_rate": 4.803981420371475e-06, "loss": 1.3927, "step": 4942 }, { "epoch": 0.53, "learning_rate": 4.802255281484029e-06, "loss": 1.1572, "step": 4943 }, { "epoch": 0.53, "learning_rate": 4.800529166200837e-06, "loss": 1.5386, "step": 4944 }, { "epoch": 0.53, "learning_rate": 4.798803074727938e-06, "loss": 1.4294, "step": 4945 }, { "epoch": 0.53, "learning_rate": 4.797077007271371e-06, "loss": 1.8245, "step": 4946 }, { "epoch": 0.53, "learning_rate": 4.795350964037174e-06, "loss": 1.197, "step": 4947 }, { "epoch": 0.53, "learning_rate": 4.79362494523138e-06, "loss": 1.6083, "step": 4948 }, { "epoch": 0.53, "learning_rate": 4.791898951060018e-06, "loss": 1.7216, "step": 4949 }, { "epoch": 0.53, "learning_rate": 4.790172981729116e-06, "loss": 1.4562, "step": 4950 }, { "epoch": 0.53, "learning_rate": 4.788447037444696e-06, "loss": 1.6034, "step": 4951 }, { "epoch": 0.53, "learning_rate": 4.786721118412784e-06, "loss": 1.2758, "step": 4952 }, { "epoch": 0.53, "learning_rate": 4.784995224839394e-06, "loss": 1.5601, "step": 4953 }, { "epoch": 0.53, "learning_rate": 4.783269356930543e-06, "loss": 1.531, "step": 4954 }, { "epoch": 0.53, "learning_rate": 4.781543514892244e-06, "loss": 1.3067, "step": 4955 }, { "epoch": 0.53, "learning_rate": 4.779817698930502e-06, "loss": 1.4117, "step": 4956 }, { "epoch": 0.53, "learning_rate": 4.778091909251329e-06, "loss": 1.378, "step": 4957 }, { "epoch": 0.53, "learning_rate": 4.776366146060725e-06, "loss": 1.495, "step": 4958 }, { "epoch": 0.53, "learning_rate": 4.774640409564688e-06, "loss": 1.5636, "step": 4959 }, { "epoch": 0.53, "learning_rate": 4.772914699969218e-06, "loss": 1.4505, "step": 4960 }, { "epoch": 0.53, "learning_rate": 4.771189017480304e-06, "loss": 1.4336, "step": 4961 }, { "epoch": 0.53, "learning_rate": 4.76946336230394e-06, "loss": 1.4597, "step": 4962 }, { "epoch": 0.53, "learning_rate": 4.767737734646112e-06, "loss": 1.4842, "step": 4963 }, { "epoch": 0.53, "learning_rate": 4.766012134712802e-06, "loss": 1.4457, "step": 4964 }, { "epoch": 0.53, "learning_rate": 4.76428656270999e-06, "loss": 1.3114, "step": 4965 }, { "epoch": 0.53, "learning_rate": 4.762561018843655e-06, "loss": 1.259, "step": 4966 }, { "epoch": 0.53, "learning_rate": 4.7608355033197684e-06, "loss": 1.5952, "step": 4967 }, { "epoch": 0.53, "learning_rate": 4.759110016344302e-06, "loss": 1.527, "step": 4968 }, { "epoch": 0.53, "learning_rate": 4.757384558123223e-06, "loss": 1.6028, "step": 4969 }, { "epoch": 0.53, "learning_rate": 4.755659128862491e-06, "loss": 1.3988, "step": 4970 }, { "epoch": 0.53, "learning_rate": 4.753933728768069e-06, "loss": 1.1879, "step": 4971 }, { "epoch": 0.53, "learning_rate": 4.752208358045913e-06, "loss": 1.4204, "step": 4972 }, { "epoch": 0.53, "learning_rate": 4.750483016901975e-06, "loss": 1.5315, "step": 4973 }, { "epoch": 0.53, "learning_rate": 4.748757705542205e-06, "loss": 1.4154, "step": 4974 }, { "epoch": 0.53, "learning_rate": 4.74703242417255e-06, "loss": 1.4615, "step": 4975 }, { "epoch": 0.53, "learning_rate": 4.745307172998948e-06, "loss": 1.2666, "step": 4976 }, { "epoch": 0.53, "learning_rate": 4.743581952227342e-06, "loss": 1.101, "step": 4977 }, { "epoch": 0.53, "learning_rate": 4.7418567620636655e-06, "loss": 1.4092, "step": 4978 }, { "epoch": 0.53, "learning_rate": 4.740131602713849e-06, "loss": 1.5462, "step": 4979 }, { "epoch": 0.53, "learning_rate": 4.73840647438382e-06, "loss": 1.5009, "step": 4980 }, { "epoch": 0.53, "learning_rate": 4.736681377279503e-06, "loss": 1.633, "step": 4981 }, { "epoch": 0.53, "learning_rate": 4.734956311606818e-06, "loss": 1.2371, "step": 4982 }, { "epoch": 0.53, "learning_rate": 4.733231277571683e-06, "loss": 1.2833, "step": 4983 }, { "epoch": 0.53, "learning_rate": 4.73150627538001e-06, "loss": 1.3262, "step": 4984 }, { "epoch": 0.53, "learning_rate": 4.729781305237705e-06, "loss": 1.4664, "step": 4985 }, { "epoch": 0.53, "learning_rate": 4.7280563673506745e-06, "loss": 1.3829, "step": 4986 }, { "epoch": 0.53, "learning_rate": 4.7263314619248215e-06, "loss": 1.3846, "step": 4987 }, { "epoch": 0.53, "learning_rate": 4.724606589166043e-06, "loss": 1.4086, "step": 4988 }, { "epoch": 0.53, "learning_rate": 4.722881749280232e-06, "loss": 1.5001, "step": 4989 }, { "epoch": 0.53, "learning_rate": 4.721156942473274e-06, "loss": 1.3478, "step": 4990 }, { "epoch": 0.53, "learning_rate": 4.719432168951061e-06, "loss": 1.4361, "step": 4991 }, { "epoch": 0.53, "learning_rate": 4.717707428919471e-06, "loss": 1.2896, "step": 4992 }, { "epoch": 0.53, "learning_rate": 4.715982722584382e-06, "loss": 1.4938, "step": 4993 }, { "epoch": 0.53, "learning_rate": 4.714258050151668e-06, "loss": 1.346, "step": 4994 }, { "epoch": 0.53, "learning_rate": 4.712533411827197e-06, "loss": 1.3961, "step": 4995 }, { "epoch": 0.53, "learning_rate": 4.7108088078168355e-06, "loss": 1.632, "step": 4996 }, { "epoch": 0.53, "learning_rate": 4.709084238326447e-06, "loss": 1.3745, "step": 4997 }, { "epoch": 0.53, "learning_rate": 4.707359703561885e-06, "loss": 1.4456, "step": 4998 }, { "epoch": 0.53, "learning_rate": 4.705635203729004e-06, "loss": 1.4524, "step": 4999 }, { "epoch": 0.53, "learning_rate": 4.703910739033653e-06, "loss": 1.4362, "step": 5000 }, { "epoch": 0.53, "learning_rate": 4.702186309681677e-06, "loss": 1.4729, "step": 5001 }, { "epoch": 0.53, "learning_rate": 4.700461915878917e-06, "loss": 1.4385, "step": 5002 }, { "epoch": 0.53, "learning_rate": 4.698737557831208e-06, "loss": 1.4506, "step": 5003 }, { "epoch": 0.53, "learning_rate": 4.697013235744382e-06, "loss": 1.4346, "step": 5004 }, { "epoch": 0.53, "learning_rate": 4.695288949824266e-06, "loss": 1.1971, "step": 5005 }, { "epoch": 0.53, "learning_rate": 4.693564700276685e-06, "loss": 1.3151, "step": 5006 }, { "epoch": 0.53, "learning_rate": 4.6918404873074574e-06, "loss": 1.3272, "step": 5007 }, { "epoch": 0.53, "learning_rate": 4.690116311122399e-06, "loss": 1.24, "step": 5008 }, { "epoch": 0.53, "learning_rate": 4.6883921719273185e-06, "loss": 1.5466, "step": 5009 }, { "epoch": 0.53, "learning_rate": 4.68666806992802e-06, "loss": 1.3748, "step": 5010 }, { "epoch": 0.53, "learning_rate": 4.684944005330308e-06, "loss": 1.2146, "step": 5011 }, { "epoch": 0.53, "learning_rate": 4.683219978339979e-06, "loss": 1.1558, "step": 5012 }, { "epoch": 0.53, "learning_rate": 4.681495989162826e-06, "loss": 1.5418, "step": 5013 }, { "epoch": 0.53, "learning_rate": 4.679772038004635e-06, "loss": 1.4205, "step": 5014 }, { "epoch": 0.53, "learning_rate": 4.678048125071188e-06, "loss": 1.3216, "step": 5015 }, { "epoch": 0.54, "learning_rate": 4.676324250568269e-06, "loss": 1.0816, "step": 5016 }, { "epoch": 0.54, "learning_rate": 4.6746004147016475e-06, "loss": 1.4399, "step": 5017 }, { "epoch": 0.54, "learning_rate": 4.672876617677096e-06, "loss": 1.4055, "step": 5018 }, { "epoch": 0.54, "learning_rate": 4.671152859700377e-06, "loss": 1.6396, "step": 5019 }, { "epoch": 0.54, "learning_rate": 4.669429140977252e-06, "loss": 1.5076, "step": 5020 }, { "epoch": 0.54, "learning_rate": 4.667705461713478e-06, "loss": 1.4201, "step": 5021 }, { "epoch": 0.54, "learning_rate": 4.665981822114805e-06, "loss": 1.5718, "step": 5022 }, { "epoch": 0.54, "learning_rate": 4.6642582223869795e-06, "loss": 1.3572, "step": 5023 }, { "epoch": 0.54, "learning_rate": 4.662534662735742e-06, "loss": 1.2918, "step": 5024 }, { "epoch": 0.54, "learning_rate": 4.660811143366828e-06, "loss": 1.3326, "step": 5025 }, { "epoch": 0.54, "learning_rate": 4.659087664485972e-06, "loss": 1.4716, "step": 5026 }, { "epoch": 0.54, "learning_rate": 4.657364226298899e-06, "loss": 1.4795, "step": 5027 }, { "epoch": 0.54, "learning_rate": 4.655640829011335e-06, "loss": 1.3577, "step": 5028 }, { "epoch": 0.54, "learning_rate": 4.653917472828992e-06, "loss": 1.3673, "step": 5029 }, { "epoch": 0.54, "learning_rate": 4.652194157957584e-06, "loss": 1.416, "step": 5030 }, { "epoch": 0.54, "learning_rate": 4.65047088460282e-06, "loss": 1.5558, "step": 5031 }, { "epoch": 0.54, "learning_rate": 4.648747652970402e-06, "loss": 1.4716, "step": 5032 }, { "epoch": 0.54, "learning_rate": 4.647024463266027e-06, "loss": 1.6909, "step": 5033 }, { "epoch": 0.54, "learning_rate": 4.645301315695387e-06, "loss": 1.2202, "step": 5034 }, { "epoch": 0.54, "learning_rate": 4.643578210464171e-06, "loss": 1.655, "step": 5035 }, { "epoch": 0.54, "learning_rate": 4.641855147778061e-06, "loss": 1.2787, "step": 5036 }, { "epoch": 0.54, "learning_rate": 4.6401321278427334e-06, "loss": 1.4223, "step": 5037 }, { "epoch": 0.54, "learning_rate": 4.638409150863861e-06, "loss": 1.3174, "step": 5038 }, { "epoch": 0.54, "learning_rate": 4.63668621704711e-06, "loss": 1.573, "step": 5039 }, { "epoch": 0.54, "learning_rate": 4.634963326598143e-06, "loss": 1.1753, "step": 5040 }, { "epoch": 0.54, "learning_rate": 4.633240479722618e-06, "loss": 1.4192, "step": 5041 }, { "epoch": 0.54, "learning_rate": 4.631517676626186e-06, "loss": 1.1141, "step": 5042 }, { "epoch": 0.54, "learning_rate": 4.629794917514492e-06, "loss": 1.3711, "step": 5043 }, { "epoch": 0.54, "learning_rate": 4.6280722025931776e-06, "loss": 1.2178, "step": 5044 }, { "epoch": 0.54, "learning_rate": 4.626349532067879e-06, "loss": 1.2557, "step": 5045 }, { "epoch": 0.54, "learning_rate": 4.624626906144227e-06, "loss": 1.3056, "step": 5046 }, { "epoch": 0.54, "learning_rate": 4.622904325027847e-06, "loss": 1.4809, "step": 5047 }, { "epoch": 0.54, "learning_rate": 4.621181788924359e-06, "loss": 1.5729, "step": 5048 }, { "epoch": 0.54, "learning_rate": 4.619459298039373e-06, "loss": 1.3817, "step": 5049 }, { "epoch": 0.54, "learning_rate": 4.617736852578506e-06, "loss": 1.3511, "step": 5050 }, { "epoch": 0.54, "learning_rate": 4.6160144527473565e-06, "loss": 1.5077, "step": 5051 }, { "epoch": 0.54, "learning_rate": 4.614292098751524e-06, "loss": 1.5809, "step": 5052 }, { "epoch": 0.54, "learning_rate": 4.612569790796601e-06, "loss": 1.5801, "step": 5053 }, { "epoch": 0.54, "learning_rate": 4.610847529088172e-06, "loss": 1.4496, "step": 5054 }, { "epoch": 0.54, "learning_rate": 4.609125313831826e-06, "loss": 1.2751, "step": 5055 }, { "epoch": 0.54, "learning_rate": 4.607403145233132e-06, "loss": 1.3399, "step": 5056 }, { "epoch": 0.54, "learning_rate": 4.605681023497663e-06, "loss": 1.467, "step": 5057 }, { "epoch": 0.54, "learning_rate": 4.603958948830985e-06, "loss": 1.4713, "step": 5058 }, { "epoch": 0.54, "learning_rate": 4.6022369214386565e-06, "loss": 1.1934, "step": 5059 }, { "epoch": 0.54, "learning_rate": 4.600514941526232e-06, "loss": 1.331, "step": 5060 }, { "epoch": 0.54, "learning_rate": 4.5987930092992596e-06, "loss": 1.8314, "step": 5061 }, { "epoch": 0.54, "learning_rate": 4.59707112496328e-06, "loss": 1.3745, "step": 5062 }, { "epoch": 0.54, "learning_rate": 4.595349288723832e-06, "loss": 1.3093, "step": 5063 }, { "epoch": 0.54, "learning_rate": 4.593627500786444e-06, "loss": 1.4381, "step": 5064 }, { "epoch": 0.54, "learning_rate": 4.591905761356643e-06, "loss": 1.6045, "step": 5065 }, { "epoch": 0.54, "learning_rate": 4.590184070639948e-06, "loss": 1.5797, "step": 5066 }, { "epoch": 0.54, "learning_rate": 4.588462428841875e-06, "loss": 1.2816, "step": 5067 }, { "epoch": 0.54, "learning_rate": 4.586740836167928e-06, "loss": 1.4165, "step": 5068 }, { "epoch": 0.54, "learning_rate": 4.585019292823609e-06, "loss": 1.3806, "step": 5069 }, { "epoch": 0.54, "learning_rate": 4.5832977990144165e-06, "loss": 1.4409, "step": 5070 }, { "epoch": 0.54, "learning_rate": 4.581576354945838e-06, "loss": 1.5762, "step": 5071 }, { "epoch": 0.54, "learning_rate": 4.579854960823361e-06, "loss": 1.3356, "step": 5072 }, { "epoch": 0.54, "learning_rate": 4.578133616852462e-06, "loss": 1.5472, "step": 5073 }, { "epoch": 0.54, "learning_rate": 4.576412323238609e-06, "loss": 1.3302, "step": 5074 }, { "epoch": 0.54, "learning_rate": 4.574691080187275e-06, "loss": 1.415, "step": 5075 }, { "epoch": 0.54, "learning_rate": 4.572969887903916e-06, "loss": 1.4406, "step": 5076 }, { "epoch": 0.54, "learning_rate": 4.571248746593988e-06, "loss": 1.5454, "step": 5077 }, { "epoch": 0.54, "learning_rate": 4.569527656462936e-06, "loss": 1.5421, "step": 5078 }, { "epoch": 0.54, "learning_rate": 4.5678066177162065e-06, "loss": 1.4588, "step": 5079 }, { "epoch": 0.54, "learning_rate": 4.5660856305592314e-06, "loss": 1.3403, "step": 5080 }, { "epoch": 0.54, "learning_rate": 4.564364695197443e-06, "loss": 1.2814, "step": 5081 }, { "epoch": 0.54, "learning_rate": 4.562643811836263e-06, "loss": 1.4797, "step": 5082 }, { "epoch": 0.54, "learning_rate": 4.5609229806811066e-06, "loss": 1.163, "step": 5083 }, { "epoch": 0.54, "learning_rate": 4.559202201937389e-06, "loss": 1.4454, "step": 5084 }, { "epoch": 0.54, "learning_rate": 4.557481475810512e-06, "loss": 1.4405, "step": 5085 }, { "epoch": 0.54, "learning_rate": 4.555760802505876e-06, "loss": 1.4835, "step": 5086 }, { "epoch": 0.54, "learning_rate": 4.554040182228874e-06, "loss": 1.2198, "step": 5087 }, { "epoch": 0.54, "learning_rate": 4.5523196151848846e-06, "loss": 1.4272, "step": 5088 }, { "epoch": 0.54, "learning_rate": 4.550599101579297e-06, "loss": 1.5441, "step": 5089 }, { "epoch": 0.54, "learning_rate": 4.548878641617479e-06, "loss": 1.4678, "step": 5090 }, { "epoch": 0.54, "learning_rate": 4.547158235504797e-06, "loss": 1.0211, "step": 5091 }, { "epoch": 0.54, "learning_rate": 4.545437883446613e-06, "loss": 1.5811, "step": 5092 }, { "epoch": 0.54, "learning_rate": 4.543717585648279e-06, "loss": 1.4305, "step": 5093 }, { "epoch": 0.54, "learning_rate": 4.541997342315145e-06, "loss": 1.3478, "step": 5094 }, { "epoch": 0.54, "learning_rate": 4.540277153652551e-06, "loss": 1.6587, "step": 5095 }, { "epoch": 0.54, "learning_rate": 4.538557019865829e-06, "loss": 1.4549, "step": 5096 }, { "epoch": 0.54, "learning_rate": 4.536836941160308e-06, "loss": 1.5932, "step": 5097 }, { "epoch": 0.54, "learning_rate": 4.535116917741308e-06, "loss": 1.4735, "step": 5098 }, { "epoch": 0.54, "learning_rate": 4.533396949814146e-06, "loss": 1.493, "step": 5099 }, { "epoch": 0.54, "learning_rate": 4.5316770375841315e-06, "loss": 1.5578, "step": 5100 }, { "epoch": 0.54, "learning_rate": 4.52995718125656e-06, "loss": 1.2828, "step": 5101 }, { "epoch": 0.54, "learning_rate": 4.52823738103673e-06, "loss": 1.3854, "step": 5102 }, { "epoch": 0.54, "learning_rate": 4.526517637129927e-06, "loss": 1.4564, "step": 5103 }, { "epoch": 0.54, "learning_rate": 4.524797949741435e-06, "loss": 1.4885, "step": 5104 }, { "epoch": 0.54, "learning_rate": 4.523078319076528e-06, "loss": 1.349, "step": 5105 }, { "epoch": 0.54, "learning_rate": 4.5213587453404736e-06, "loss": 1.4473, "step": 5106 }, { "epoch": 0.54, "learning_rate": 4.51963922873853e-06, "loss": 1.441, "step": 5107 }, { "epoch": 0.54, "learning_rate": 4.5179197694759525e-06, "loss": 1.2722, "step": 5108 }, { "epoch": 0.54, "learning_rate": 4.5162003677579905e-06, "loss": 1.4573, "step": 5109 }, { "epoch": 0.55, "learning_rate": 4.514481023789882e-06, "loss": 1.2335, "step": 5110 }, { "epoch": 0.55, "learning_rate": 4.512761737776862e-06, "loss": 1.5074, "step": 5111 }, { "epoch": 0.55, "learning_rate": 4.511042509924157e-06, "loss": 1.299, "step": 5112 }, { "epoch": 0.55, "learning_rate": 4.5093233404369815e-06, "loss": 1.4443, "step": 5113 }, { "epoch": 0.55, "learning_rate": 4.5076042295205575e-06, "loss": 1.4207, "step": 5114 }, { "epoch": 0.55, "learning_rate": 4.505885177380083e-06, "loss": 1.4915, "step": 5115 }, { "epoch": 0.55, "learning_rate": 4.504166184220761e-06, "loss": 1.343, "step": 5116 }, { "epoch": 0.55, "learning_rate": 4.5024472502477795e-06, "loss": 1.1042, "step": 5117 }, { "epoch": 0.55, "learning_rate": 4.5007283756663245e-06, "loss": 1.3267, "step": 5118 }, { "epoch": 0.55, "learning_rate": 4.499009560681574e-06, "loss": 1.3377, "step": 5119 }, { "epoch": 0.55, "learning_rate": 4.4972908054987005e-06, "loss": 1.5295, "step": 5120 }, { "epoch": 0.55, "learning_rate": 4.495572110322862e-06, "loss": 1.6012, "step": 5121 }, { "epoch": 0.55, "learning_rate": 4.493853475359217e-06, "loss": 1.3924, "step": 5122 }, { "epoch": 0.55, "learning_rate": 4.492134900812915e-06, "loss": 1.4198, "step": 5123 }, { "epoch": 0.55, "learning_rate": 4.490416386889097e-06, "loss": 1.3889, "step": 5124 }, { "epoch": 0.55, "learning_rate": 4.488697933792897e-06, "loss": 1.3334, "step": 5125 }, { "epoch": 0.55, "learning_rate": 4.486979541729445e-06, "loss": 1.4508, "step": 5126 }, { "epoch": 0.55, "learning_rate": 4.485261210903854e-06, "loss": 1.206, "step": 5127 }, { "epoch": 0.55, "learning_rate": 4.483542941521245e-06, "loss": 1.3429, "step": 5128 }, { "epoch": 0.55, "learning_rate": 4.481824733786716e-06, "loss": 1.4278, "step": 5129 }, { "epoch": 0.55, "learning_rate": 4.48010658790537e-06, "loss": 1.3585, "step": 5130 }, { "epoch": 0.55, "learning_rate": 4.4783885040822935e-06, "loss": 1.462, "step": 5131 }, { "epoch": 0.55, "learning_rate": 4.476670482522571e-06, "loss": 1.4693, "step": 5132 }, { "epoch": 0.55, "learning_rate": 4.47495252343128e-06, "loss": 1.5363, "step": 5133 }, { "epoch": 0.55, "learning_rate": 4.473234627013487e-06, "loss": 1.4894, "step": 5134 }, { "epoch": 0.55, "learning_rate": 4.471516793474252e-06, "loss": 1.4603, "step": 5135 }, { "epoch": 0.55, "learning_rate": 4.469799023018628e-06, "loss": 1.401, "step": 5136 }, { "epoch": 0.55, "learning_rate": 4.468081315851661e-06, "loss": 1.4032, "step": 5137 }, { "epoch": 0.55, "learning_rate": 4.466363672178391e-06, "loss": 1.5264, "step": 5138 }, { "epoch": 0.55, "learning_rate": 4.464646092203846e-06, "loss": 1.4845, "step": 5139 }, { "epoch": 0.55, "learning_rate": 4.4629285761330515e-06, "loss": 1.3937, "step": 5140 }, { "epoch": 0.55, "learning_rate": 4.46121112417102e-06, "loss": 1.3328, "step": 5141 }, { "epoch": 0.55, "learning_rate": 4.459493736522759e-06, "loss": 1.6684, "step": 5142 }, { "epoch": 0.55, "learning_rate": 4.4577764133932696e-06, "loss": 1.1904, "step": 5143 }, { "epoch": 0.55, "learning_rate": 4.456059154987545e-06, "loss": 1.3826, "step": 5144 }, { "epoch": 0.55, "learning_rate": 4.4543419615105685e-06, "loss": 1.6366, "step": 5145 }, { "epoch": 0.55, "learning_rate": 4.452624833167318e-06, "loss": 1.3338, "step": 5146 }, { "epoch": 0.55, "learning_rate": 4.450907770162758e-06, "loss": 1.5163, "step": 5147 }, { "epoch": 0.55, "learning_rate": 4.449190772701857e-06, "loss": 1.5471, "step": 5148 }, { "epoch": 0.55, "learning_rate": 4.447473840989563e-06, "loss": 1.7073, "step": 5149 }, { "epoch": 0.55, "learning_rate": 4.445756975230823e-06, "loss": 1.3736, "step": 5150 }, { "epoch": 0.55, "learning_rate": 4.444040175630577e-06, "loss": 1.4553, "step": 5151 }, { "epoch": 0.55, "learning_rate": 4.4423234423937474e-06, "loss": 1.6429, "step": 5152 }, { "epoch": 0.55, "learning_rate": 4.440606775725265e-06, "loss": 1.3975, "step": 5153 }, { "epoch": 0.55, "learning_rate": 4.438890175830039e-06, "loss": 1.396, "step": 5154 }, { "epoch": 0.55, "learning_rate": 4.437173642912976e-06, "loss": 1.4795, "step": 5155 }, { "epoch": 0.55, "learning_rate": 4.435457177178972e-06, "loss": 1.1267, "step": 5156 }, { "epoch": 0.55, "learning_rate": 4.433740778832919e-06, "loss": 1.592, "step": 5157 }, { "epoch": 0.55, "learning_rate": 4.432024448079699e-06, "loss": 1.5414, "step": 5158 }, { "epoch": 0.55, "learning_rate": 4.430308185124187e-06, "loss": 1.4417, "step": 5159 }, { "epoch": 0.55, "learning_rate": 4.428591990171246e-06, "loss": 1.3683, "step": 5160 }, { "epoch": 0.55, "learning_rate": 4.426875863425733e-06, "loss": 1.1966, "step": 5161 }, { "epoch": 0.55, "learning_rate": 4.4251598050925e-06, "loss": 1.1907, "step": 5162 }, { "epoch": 0.55, "learning_rate": 4.423443815376387e-06, "loss": 1.558, "step": 5163 }, { "epoch": 0.55, "learning_rate": 4.421727894482227e-06, "loss": 1.7696, "step": 5164 }, { "epoch": 0.55, "learning_rate": 4.420012042614847e-06, "loss": 1.2235, "step": 5165 }, { "epoch": 0.55, "learning_rate": 4.41829625997906e-06, "loss": 1.7217, "step": 5166 }, { "epoch": 0.55, "learning_rate": 4.416580546779675e-06, "loss": 1.346, "step": 5167 }, { "epoch": 0.55, "learning_rate": 4.414864903221493e-06, "loss": 1.3828, "step": 5168 }, { "epoch": 0.55, "learning_rate": 4.413149329509307e-06, "loss": 1.5543, "step": 5169 }, { "epoch": 0.55, "learning_rate": 4.4114338258478995e-06, "loss": 1.8447, "step": 5170 }, { "epoch": 0.55, "learning_rate": 4.409718392442044e-06, "loss": 1.5997, "step": 5171 }, { "epoch": 0.55, "learning_rate": 4.4080030294965085e-06, "loss": 1.6291, "step": 5172 }, { "epoch": 0.55, "learning_rate": 4.406287737216053e-06, "loss": 1.451, "step": 5173 }, { "epoch": 0.55, "learning_rate": 4.404572515805424e-06, "loss": 1.4357, "step": 5174 }, { "epoch": 0.55, "learning_rate": 4.402857365469364e-06, "loss": 1.4429, "step": 5175 }, { "epoch": 0.55, "learning_rate": 4.401142286412605e-06, "loss": 1.434, "step": 5176 }, { "epoch": 0.55, "learning_rate": 4.399427278839874e-06, "loss": 1.2134, "step": 5177 }, { "epoch": 0.55, "learning_rate": 4.397712342955885e-06, "loss": 1.1688, "step": 5178 }, { "epoch": 0.55, "learning_rate": 4.395997478965347e-06, "loss": 1.463, "step": 5179 }, { "epoch": 0.55, "learning_rate": 4.394282687072955e-06, "loss": 1.4298, "step": 5180 }, { "epoch": 0.55, "learning_rate": 4.392567967483401e-06, "loss": 1.4444, "step": 5181 }, { "epoch": 0.55, "learning_rate": 4.390853320401367e-06, "loss": 1.5445, "step": 5182 }, { "epoch": 0.55, "learning_rate": 4.389138746031526e-06, "loss": 1.4232, "step": 5183 }, { "epoch": 0.55, "learning_rate": 4.387424244578543e-06, "loss": 1.4357, "step": 5184 }, { "epoch": 0.55, "learning_rate": 4.385709816247071e-06, "loss": 1.4238, "step": 5185 }, { "epoch": 0.55, "learning_rate": 4.383995461241755e-06, "loss": 1.642, "step": 5186 }, { "epoch": 0.55, "learning_rate": 4.38228117976724e-06, "loss": 1.4619, "step": 5187 }, { "epoch": 0.55, "learning_rate": 4.380566972028149e-06, "loss": 1.489, "step": 5188 }, { "epoch": 0.55, "learning_rate": 4.378852838229104e-06, "loss": 1.7069, "step": 5189 }, { "epoch": 0.55, "learning_rate": 4.377138778574716e-06, "loss": 1.5645, "step": 5190 }, { "epoch": 0.55, "learning_rate": 4.375424793269588e-06, "loss": 1.3234, "step": 5191 }, { "epoch": 0.55, "learning_rate": 4.373710882518314e-06, "loss": 1.1702, "step": 5192 }, { "epoch": 0.55, "learning_rate": 4.371997046525481e-06, "loss": 1.4772, "step": 5193 }, { "epoch": 0.55, "learning_rate": 4.370283285495662e-06, "loss": 1.2666, "step": 5194 }, { "epoch": 0.55, "learning_rate": 4.368569599633425e-06, "loss": 1.5837, "step": 5195 }, { "epoch": 0.55, "learning_rate": 4.366855989143326e-06, "loss": 1.2282, "step": 5196 }, { "epoch": 0.55, "learning_rate": 4.365142454229918e-06, "loss": 1.5057, "step": 5197 }, { "epoch": 0.55, "learning_rate": 4.363428995097741e-06, "loss": 1.3744, "step": 5198 }, { "epoch": 0.55, "learning_rate": 4.3617156119513206e-06, "loss": 1.4095, "step": 5199 }, { "epoch": 0.55, "learning_rate": 4.360002304995184e-06, "loss": 1.2953, "step": 5200 }, { "epoch": 0.55, "learning_rate": 4.35828907443384e-06, "loss": 1.403, "step": 5201 }, { "epoch": 0.55, "learning_rate": 4.356575920471796e-06, "loss": 1.4883, "step": 5202 }, { "epoch": 0.55, "learning_rate": 4.354862843313547e-06, "loss": 1.4566, "step": 5203 }, { "epoch": 0.56, "learning_rate": 4.353149843163575e-06, "loss": 1.3741, "step": 5204 }, { "epoch": 0.56, "learning_rate": 4.351436920226357e-06, "loss": 1.2117, "step": 5205 }, { "epoch": 0.56, "learning_rate": 4.34972407470636e-06, "loss": 1.6238, "step": 5206 }, { "epoch": 0.56, "learning_rate": 4.348011306808044e-06, "loss": 1.4702, "step": 5207 }, { "epoch": 0.56, "learning_rate": 4.346298616735855e-06, "loss": 1.3692, "step": 5208 }, { "epoch": 0.56, "learning_rate": 4.344586004694234e-06, "loss": 1.3301, "step": 5209 }, { "epoch": 0.56, "learning_rate": 4.342873470887609e-06, "loss": 1.4502, "step": 5210 }, { "epoch": 0.56, "learning_rate": 4.3411610155204e-06, "loss": 1.3537, "step": 5211 }, { "epoch": 0.56, "learning_rate": 4.339448638797022e-06, "loss": 1.3375, "step": 5212 }, { "epoch": 0.56, "learning_rate": 4.337736340921874e-06, "loss": 1.4452, "step": 5213 }, { "epoch": 0.56, "learning_rate": 4.336024122099348e-06, "loss": 1.69, "step": 5214 }, { "epoch": 0.56, "learning_rate": 4.334311982533827e-06, "loss": 1.5949, "step": 5215 }, { "epoch": 0.56, "learning_rate": 4.3325999224296864e-06, "loss": 1.1354, "step": 5216 }, { "epoch": 0.56, "learning_rate": 4.330887941991288e-06, "loss": 1.4529, "step": 5217 }, { "epoch": 0.56, "learning_rate": 4.32917604142299e-06, "loss": 1.6563, "step": 5218 }, { "epoch": 0.56, "learning_rate": 4.327464220929132e-06, "loss": 1.6056, "step": 5219 }, { "epoch": 0.56, "learning_rate": 4.325752480714052e-06, "loss": 1.3719, "step": 5220 }, { "epoch": 0.56, "learning_rate": 4.324040820982076e-06, "loss": 1.5394, "step": 5221 }, { "epoch": 0.56, "learning_rate": 4.3223292419375196e-06, "loss": 1.5209, "step": 5222 }, { "epoch": 0.56, "learning_rate": 4.320617743784691e-06, "loss": 1.6632, "step": 5223 }, { "epoch": 0.56, "learning_rate": 4.318906326727886e-06, "loss": 1.336, "step": 5224 }, { "epoch": 0.56, "learning_rate": 4.317194990971388e-06, "loss": 1.4815, "step": 5225 }, { "epoch": 0.56, "learning_rate": 4.315483736719482e-06, "loss": 1.535, "step": 5226 }, { "epoch": 0.56, "learning_rate": 4.313772564176431e-06, "loss": 1.471, "step": 5227 }, { "epoch": 0.56, "learning_rate": 4.312061473546494e-06, "loss": 1.3742, "step": 5228 }, { "epoch": 0.56, "learning_rate": 4.310350465033919e-06, "loss": 1.2701, "step": 5229 }, { "epoch": 0.56, "learning_rate": 4.308639538842943e-06, "loss": 1.5777, "step": 5230 }, { "epoch": 0.56, "learning_rate": 4.306928695177799e-06, "loss": 1.3739, "step": 5231 }, { "epoch": 0.56, "learning_rate": 4.305217934242703e-06, "loss": 1.3802, "step": 5232 }, { "epoch": 0.56, "learning_rate": 4.303507256241863e-06, "loss": 1.5881, "step": 5233 }, { "epoch": 0.56, "learning_rate": 4.30179666137948e-06, "loss": 1.3937, "step": 5234 }, { "epoch": 0.56, "learning_rate": 4.30008614985974e-06, "loss": 1.4407, "step": 5235 }, { "epoch": 0.56, "learning_rate": 4.298375721886826e-06, "loss": 1.263, "step": 5236 }, { "epoch": 0.56, "learning_rate": 4.2966653776649056e-06, "loss": 1.4522, "step": 5237 }, { "epoch": 0.56, "learning_rate": 4.294955117398139e-06, "loss": 1.4638, "step": 5238 }, { "epoch": 0.56, "learning_rate": 4.293244941290672e-06, "loss": 1.1638, "step": 5239 }, { "epoch": 0.56, "learning_rate": 4.2915348495466454e-06, "loss": 1.439, "step": 5240 }, { "epoch": 0.56, "learning_rate": 4.28982484237019e-06, "loss": 1.5442, "step": 5241 }, { "epoch": 0.56, "learning_rate": 4.288114919965423e-06, "loss": 1.235, "step": 5242 }, { "epoch": 0.56, "learning_rate": 4.286405082536455e-06, "loss": 1.4252, "step": 5243 }, { "epoch": 0.56, "learning_rate": 4.284695330287383e-06, "loss": 1.7032, "step": 5244 }, { "epoch": 0.56, "learning_rate": 4.282985663422292e-06, "loss": 1.5978, "step": 5245 }, { "epoch": 0.56, "learning_rate": 4.281276082145268e-06, "loss": 1.4958, "step": 5246 }, { "epoch": 0.56, "learning_rate": 4.279566586660375e-06, "loss": 1.5708, "step": 5247 }, { "epoch": 0.56, "learning_rate": 4.27785717717167e-06, "loss": 1.4029, "step": 5248 }, { "epoch": 0.56, "learning_rate": 4.276147853883203e-06, "loss": 1.2491, "step": 5249 }, { "epoch": 0.56, "learning_rate": 4.274438616999007e-06, "loss": 1.5101, "step": 5250 }, { "epoch": 0.56, "learning_rate": 4.272729466723114e-06, "loss": 1.4426, "step": 5251 }, { "epoch": 0.56, "learning_rate": 4.271020403259539e-06, "loss": 1.3222, "step": 5252 }, { "epoch": 0.56, "learning_rate": 4.269311426812287e-06, "loss": 1.5007, "step": 5253 }, { "epoch": 0.56, "learning_rate": 4.2676025375853555e-06, "loss": 1.4792, "step": 5254 }, { "epoch": 0.56, "learning_rate": 4.265893735782729e-06, "loss": 1.3679, "step": 5255 }, { "epoch": 0.56, "learning_rate": 4.264185021608382e-06, "loss": 1.5098, "step": 5256 }, { "epoch": 0.56, "learning_rate": 4.262476395266282e-06, "loss": 1.5931, "step": 5257 }, { "epoch": 0.56, "learning_rate": 4.26076785696038e-06, "loss": 1.1999, "step": 5258 }, { "epoch": 0.56, "learning_rate": 4.259059406894619e-06, "loss": 1.3413, "step": 5259 }, { "epoch": 0.56, "learning_rate": 4.257351045272935e-06, "loss": 1.2856, "step": 5260 }, { "epoch": 0.56, "learning_rate": 4.255642772299249e-06, "loss": 1.3412, "step": 5261 }, { "epoch": 0.56, "learning_rate": 4.253934588177473e-06, "loss": 1.1257, "step": 5262 }, { "epoch": 0.56, "learning_rate": 4.25222649311151e-06, "loss": 1.4075, "step": 5263 }, { "epoch": 0.56, "learning_rate": 4.250518487305246e-06, "loss": 1.546, "step": 5264 }, { "epoch": 0.56, "learning_rate": 4.248810570962567e-06, "loss": 1.3431, "step": 5265 }, { "epoch": 0.56, "learning_rate": 4.247102744287338e-06, "loss": 1.5643, "step": 5266 }, { "epoch": 0.56, "learning_rate": 4.245395007483421e-06, "loss": 1.3593, "step": 5267 }, { "epoch": 0.56, "learning_rate": 4.24368736075466e-06, "loss": 1.5978, "step": 5268 }, { "epoch": 0.56, "learning_rate": 4.241979804304894e-06, "loss": 1.389, "step": 5269 }, { "epoch": 0.56, "learning_rate": 4.240272338337951e-06, "loss": 1.4153, "step": 5270 }, { "epoch": 0.56, "learning_rate": 4.238564963057646e-06, "loss": 1.3983, "step": 5271 }, { "epoch": 0.56, "learning_rate": 4.236857678667782e-06, "loss": 1.3415, "step": 5272 }, { "epoch": 0.56, "learning_rate": 4.2351504853721545e-06, "loss": 1.3359, "step": 5273 }, { "epoch": 0.56, "learning_rate": 4.233443383374545e-06, "loss": 1.461, "step": 5274 }, { "epoch": 0.56, "learning_rate": 4.231736372878726e-06, "loss": 1.2582, "step": 5275 }, { "epoch": 0.56, "learning_rate": 4.230029454088461e-06, "loss": 1.5544, "step": 5276 }, { "epoch": 0.56, "learning_rate": 4.228322627207499e-06, "loss": 1.4633, "step": 5277 }, { "epoch": 0.56, "learning_rate": 4.2266158924395796e-06, "loss": 1.3935, "step": 5278 }, { "epoch": 0.56, "learning_rate": 4.224909249988427e-06, "loss": 1.3819, "step": 5279 }, { "epoch": 0.56, "learning_rate": 4.223202700057765e-06, "loss": 1.82, "step": 5280 }, { "epoch": 0.56, "learning_rate": 4.221496242851296e-06, "loss": 1.5755, "step": 5281 }, { "epoch": 0.56, "learning_rate": 4.219789878572717e-06, "loss": 1.4332, "step": 5282 }, { "epoch": 0.56, "learning_rate": 4.21808360742571e-06, "loss": 1.1626, "step": 5283 }, { "epoch": 0.56, "learning_rate": 4.2163774296139474e-06, "loss": 1.3733, "step": 5284 }, { "epoch": 0.56, "learning_rate": 4.214671345341096e-06, "loss": 1.3055, "step": 5285 }, { "epoch": 0.56, "learning_rate": 4.212965354810802e-06, "loss": 1.5517, "step": 5286 }, { "epoch": 0.56, "learning_rate": 4.211259458226706e-06, "loss": 1.3845, "step": 5287 }, { "epoch": 0.56, "learning_rate": 4.209553655792436e-06, "loss": 1.2908, "step": 5288 }, { "epoch": 0.56, "learning_rate": 4.207847947711609e-06, "loss": 1.2959, "step": 5289 }, { "epoch": 0.56, "learning_rate": 4.206142334187832e-06, "loss": 1.3958, "step": 5290 }, { "epoch": 0.56, "learning_rate": 4.204436815424699e-06, "loss": 1.4169, "step": 5291 }, { "epoch": 0.56, "learning_rate": 4.202731391625793e-06, "loss": 1.5239, "step": 5292 }, { "epoch": 0.56, "learning_rate": 4.201026062994684e-06, "loss": 1.5823, "step": 5293 }, { "epoch": 0.56, "learning_rate": 4.199320829734934e-06, "loss": 1.4167, "step": 5294 }, { "epoch": 0.56, "learning_rate": 4.1976156920500935e-06, "loss": 1.3559, "step": 5295 }, { "epoch": 0.56, "learning_rate": 4.1959106501436985e-06, "loss": 1.5291, "step": 5296 }, { "epoch": 0.57, "learning_rate": 4.194205704219276e-06, "loss": 1.3674, "step": 5297 }, { "epoch": 0.57, "learning_rate": 4.19250085448034e-06, "loss": 1.5473, "step": 5298 }, { "epoch": 0.57, "learning_rate": 4.190796101130392e-06, "loss": 1.6885, "step": 5299 }, { "epoch": 0.57, "learning_rate": 4.189091444372928e-06, "loss": 1.7078, "step": 5300 }, { "epoch": 0.57, "learning_rate": 4.187386884411426e-06, "loss": 1.5562, "step": 5301 }, { "epoch": 0.57, "learning_rate": 4.185682421449356e-06, "loss": 1.35, "step": 5302 }, { "epoch": 0.57, "learning_rate": 4.1839780556901716e-06, "loss": 1.2794, "step": 5303 }, { "epoch": 0.57, "learning_rate": 4.182273787337323e-06, "loss": 1.6938, "step": 5304 }, { "epoch": 0.57, "learning_rate": 4.180569616594242e-06, "loss": 1.5518, "step": 5305 }, { "epoch": 0.57, "learning_rate": 4.17886554366435e-06, "loss": 1.4269, "step": 5306 }, { "epoch": 0.57, "learning_rate": 4.177161568751058e-06, "loss": 1.3522, "step": 5307 }, { "epoch": 0.57, "learning_rate": 4.175457692057765e-06, "loss": 1.3703, "step": 5308 }, { "epoch": 0.57, "learning_rate": 4.173753913787859e-06, "loss": 1.5881, "step": 5309 }, { "epoch": 0.57, "learning_rate": 4.172050234144716e-06, "loss": 1.4221, "step": 5310 }, { "epoch": 0.57, "learning_rate": 4.1703466533316975e-06, "loss": 1.1397, "step": 5311 }, { "epoch": 0.57, "learning_rate": 4.168643171552156e-06, "loss": 1.3395, "step": 5312 }, { "epoch": 0.57, "learning_rate": 4.16693978900943e-06, "loss": 1.4528, "step": 5313 }, { "epoch": 0.57, "learning_rate": 4.165236505906851e-06, "loss": 1.4925, "step": 5314 }, { "epoch": 0.57, "learning_rate": 4.163533322447734e-06, "loss": 1.5561, "step": 5315 }, { "epoch": 0.57, "learning_rate": 4.161830238835386e-06, "loss": 1.7238, "step": 5316 }, { "epoch": 0.57, "learning_rate": 4.160127255273093e-06, "loss": 1.4116, "step": 5317 }, { "epoch": 0.57, "learning_rate": 4.158424371964139e-06, "loss": 1.3724, "step": 5318 }, { "epoch": 0.57, "learning_rate": 4.156721589111794e-06, "loss": 1.3777, "step": 5319 }, { "epoch": 0.57, "learning_rate": 4.1550189069193134e-06, "loss": 1.4233, "step": 5320 }, { "epoch": 0.57, "learning_rate": 4.153316325589941e-06, "loss": 1.4245, "step": 5321 }, { "epoch": 0.57, "learning_rate": 4.151613845326912e-06, "loss": 1.4947, "step": 5322 }, { "epoch": 0.57, "learning_rate": 4.149911466333441e-06, "loss": 1.3812, "step": 5323 }, { "epoch": 0.57, "learning_rate": 4.1482091888127444e-06, "loss": 1.8155, "step": 5324 }, { "epoch": 0.57, "learning_rate": 4.146507012968013e-06, "loss": 1.4655, "step": 5325 }, { "epoch": 0.57, "learning_rate": 4.144804939002432e-06, "loss": 1.3001, "step": 5326 }, { "epoch": 0.57, "learning_rate": 4.143102967119173e-06, "loss": 1.4464, "step": 5327 }, { "epoch": 0.57, "learning_rate": 4.141401097521396e-06, "loss": 1.2734, "step": 5328 }, { "epoch": 0.57, "learning_rate": 4.13969933041225e-06, "loss": 1.3442, "step": 5329 }, { "epoch": 0.57, "learning_rate": 4.137997665994871e-06, "loss": 1.4784, "step": 5330 }, { "epoch": 0.57, "learning_rate": 4.136296104472378e-06, "loss": 1.9277, "step": 5331 }, { "epoch": 0.57, "learning_rate": 4.134594646047886e-06, "loss": 1.2029, "step": 5332 }, { "epoch": 0.57, "learning_rate": 4.13289329092449e-06, "loss": 1.3417, "step": 5333 }, { "epoch": 0.57, "learning_rate": 4.131192039305278e-06, "loss": 1.3182, "step": 5334 }, { "epoch": 0.57, "learning_rate": 4.129490891393325e-06, "loss": 1.2598, "step": 5335 }, { "epoch": 0.57, "learning_rate": 4.127789847391692e-06, "loss": 1.3636, "step": 5336 }, { "epoch": 0.57, "learning_rate": 4.1260889075034254e-06, "loss": 1.2786, "step": 5337 }, { "epoch": 0.57, "learning_rate": 4.124388071931563e-06, "loss": 1.4765, "step": 5338 }, { "epoch": 0.57, "learning_rate": 4.122687340879131e-06, "loss": 1.4031, "step": 5339 }, { "epoch": 0.57, "learning_rate": 4.120986714549139e-06, "loss": 1.5652, "step": 5340 }, { "epoch": 0.57, "learning_rate": 4.119286193144587e-06, "loss": 1.1463, "step": 5341 }, { "epoch": 0.57, "learning_rate": 4.1175857768684614e-06, "loss": 1.6842, "step": 5342 }, { "epoch": 0.57, "learning_rate": 4.115885465923734e-06, "loss": 1.3445, "step": 5343 }, { "epoch": 0.57, "learning_rate": 4.114185260513371e-06, "loss": 1.3095, "step": 5344 }, { "epoch": 0.57, "learning_rate": 4.11248516084032e-06, "loss": 1.2476, "step": 5345 }, { "epoch": 0.57, "learning_rate": 4.110785167107514e-06, "loss": 1.3838, "step": 5346 }, { "epoch": 0.57, "learning_rate": 4.109085279517878e-06, "loss": 1.4936, "step": 5347 }, { "epoch": 0.57, "learning_rate": 4.1073854982743256e-06, "loss": 1.365, "step": 5348 }, { "epoch": 0.57, "learning_rate": 4.1056858235797545e-06, "loss": 1.464, "step": 5349 }, { "epoch": 0.57, "learning_rate": 4.103986255637048e-06, "loss": 1.6058, "step": 5350 }, { "epoch": 0.57, "learning_rate": 4.1022867946490795e-06, "loss": 1.6217, "step": 5351 }, { "epoch": 0.57, "learning_rate": 4.100587440818709e-06, "loss": 1.2619, "step": 5352 }, { "epoch": 0.57, "learning_rate": 4.098888194348785e-06, "loss": 1.7236, "step": 5353 }, { "epoch": 0.57, "learning_rate": 4.097189055442142e-06, "loss": 1.5354, "step": 5354 }, { "epoch": 0.57, "learning_rate": 4.0954900243016016e-06, "loss": 1.2077, "step": 5355 }, { "epoch": 0.57, "learning_rate": 4.093791101129972e-06, "loss": 1.4217, "step": 5356 }, { "epoch": 0.57, "learning_rate": 4.092092286130046e-06, "loss": 1.262, "step": 5357 }, { "epoch": 0.57, "learning_rate": 4.090393579504612e-06, "loss": 1.5801, "step": 5358 }, { "epoch": 0.57, "learning_rate": 4.088694981456437e-06, "loss": 1.1341, "step": 5359 }, { "epoch": 0.57, "learning_rate": 4.086996492188279e-06, "loss": 1.3962, "step": 5360 }, { "epoch": 0.57, "learning_rate": 4.085298111902882e-06, "loss": 1.2868, "step": 5361 }, { "epoch": 0.57, "learning_rate": 4.083599840802974e-06, "loss": 1.697, "step": 5362 }, { "epoch": 0.57, "learning_rate": 4.081901679091279e-06, "loss": 1.4904, "step": 5363 }, { "epoch": 0.57, "learning_rate": 4.080203626970498e-06, "loss": 1.4485, "step": 5364 }, { "epoch": 0.57, "learning_rate": 4.078505684643323e-06, "loss": 1.6106, "step": 5365 }, { "epoch": 0.57, "learning_rate": 4.076807852312434e-06, "loss": 1.3133, "step": 5366 }, { "epoch": 0.57, "learning_rate": 4.0751101301804945e-06, "loss": 1.194, "step": 5367 }, { "epoch": 0.57, "learning_rate": 4.0734125184501605e-06, "loss": 1.4019, "step": 5368 }, { "epoch": 0.57, "learning_rate": 4.07171501732407e-06, "loss": 1.2827, "step": 5369 }, { "epoch": 0.57, "learning_rate": 4.070017627004847e-06, "loss": 1.4712, "step": 5370 }, { "epoch": 0.57, "learning_rate": 4.0683203476951065e-06, "loss": 1.478, "step": 5371 }, { "epoch": 0.57, "learning_rate": 4.066623179597447e-06, "loss": 1.5334, "step": 5372 }, { "epoch": 0.57, "learning_rate": 4.0649261229144554e-06, "loss": 1.4033, "step": 5373 }, { "epoch": 0.57, "learning_rate": 4.063229177848706e-06, "loss": 1.7017, "step": 5374 }, { "epoch": 0.57, "learning_rate": 4.061532344602757e-06, "loss": 1.1737, "step": 5375 }, { "epoch": 0.57, "learning_rate": 4.059835623379155e-06, "loss": 1.3455, "step": 5376 }, { "epoch": 0.57, "learning_rate": 4.058139014380432e-06, "loss": 1.2787, "step": 5377 }, { "epoch": 0.57, "learning_rate": 4.056442517809109e-06, "loss": 1.512, "step": 5378 }, { "epoch": 0.57, "learning_rate": 4.054746133867693e-06, "loss": 1.1636, "step": 5379 }, { "epoch": 0.57, "learning_rate": 4.053049862758675e-06, "loss": 1.2386, "step": 5380 }, { "epoch": 0.57, "learning_rate": 4.051353704684538e-06, "loss": 1.3969, "step": 5381 }, { "epoch": 0.57, "learning_rate": 4.0496576598477396e-06, "loss": 1.6143, "step": 5382 }, { "epoch": 0.57, "learning_rate": 4.047961728450742e-06, "loss": 1.4338, "step": 5383 }, { "epoch": 0.57, "learning_rate": 4.046265910695977e-06, "loss": 1.3813, "step": 5384 }, { "epoch": 0.57, "learning_rate": 4.044570206785874e-06, "loss": 1.537, "step": 5385 }, { "epoch": 0.57, "learning_rate": 4.042874616922841e-06, "loss": 1.3725, "step": 5386 }, { "epoch": 0.57, "learning_rate": 4.041179141309276e-06, "loss": 1.7014, "step": 5387 }, { "epoch": 0.57, "learning_rate": 4.039483780147568e-06, "loss": 1.3186, "step": 5388 }, { "epoch": 0.57, "learning_rate": 4.0377885336400845e-06, "loss": 1.5576, "step": 5389 }, { "epoch": 0.57, "learning_rate": 4.036093401989182e-06, "loss": 1.382, "step": 5390 }, { "epoch": 0.58, "learning_rate": 4.0343983853972045e-06, "loss": 1.2661, "step": 5391 }, { "epoch": 0.58, "learning_rate": 4.03270348406648e-06, "loss": 1.4846, "step": 5392 }, { "epoch": 0.58, "learning_rate": 4.031008698199328e-06, "loss": 1.5275, "step": 5393 }, { "epoch": 0.58, "learning_rate": 4.029314027998049e-06, "loss": 1.0808, "step": 5394 }, { "epoch": 0.58, "learning_rate": 4.0276194736649285e-06, "loss": 1.5211, "step": 5395 }, { "epoch": 0.58, "learning_rate": 4.025925035402242e-06, "loss": 1.6125, "step": 5396 }, { "epoch": 0.58, "learning_rate": 4.024230713412253e-06, "loss": 1.4245, "step": 5397 }, { "epoch": 0.58, "learning_rate": 4.022536507897206e-06, "loss": 1.432, "step": 5398 }, { "epoch": 0.58, "learning_rate": 4.0208424190593325e-06, "loss": 1.7057, "step": 5399 }, { "epoch": 0.58, "learning_rate": 4.019148447100855e-06, "loss": 1.5947, "step": 5400 }, { "epoch": 0.58, "learning_rate": 4.017454592223971e-06, "loss": 1.2688, "step": 5401 }, { "epoch": 0.58, "learning_rate": 4.015760854630881e-06, "loss": 1.322, "step": 5402 }, { "epoch": 0.58, "learning_rate": 4.014067234523756e-06, "loss": 1.4465, "step": 5403 }, { "epoch": 0.58, "learning_rate": 4.012373732104759e-06, "loss": 1.277, "step": 5404 }, { "epoch": 0.58, "learning_rate": 4.010680347576041e-06, "loss": 1.3088, "step": 5405 }, { "epoch": 0.58, "learning_rate": 4.008987081139734e-06, "loss": 1.5015, "step": 5406 }, { "epoch": 0.58, "learning_rate": 4.007293932997962e-06, "loss": 1.5779, "step": 5407 }, { "epoch": 0.58, "learning_rate": 4.00560090335283e-06, "loss": 1.2387, "step": 5408 }, { "epoch": 0.58, "learning_rate": 4.0039079924064285e-06, "loss": 1.3054, "step": 5409 }, { "epoch": 0.58, "learning_rate": 4.002215200360838e-06, "loss": 1.4814, "step": 5410 }, { "epoch": 0.58, "learning_rate": 4.000522527418121e-06, "loss": 1.4085, "step": 5411 }, { "epoch": 0.58, "learning_rate": 3.998829973780329e-06, "loss": 1.4132, "step": 5412 }, { "epoch": 0.58, "learning_rate": 3.9971375396494965e-06, "loss": 1.4879, "step": 5413 }, { "epoch": 0.58, "learning_rate": 3.995445225227646e-06, "loss": 1.5593, "step": 5414 }, { "epoch": 0.58, "learning_rate": 3.993753030716783e-06, "loss": 1.5841, "step": 5415 }, { "epoch": 0.58, "learning_rate": 3.992060956318898e-06, "loss": 1.4901, "step": 5416 }, { "epoch": 0.58, "learning_rate": 3.990369002235975e-06, "loss": 1.4773, "step": 5417 }, { "epoch": 0.58, "learning_rate": 3.988677168669974e-06, "loss": 1.4873, "step": 5418 }, { "epoch": 0.58, "learning_rate": 3.9869854558228455e-06, "loss": 1.2442, "step": 5419 }, { "epoch": 0.58, "learning_rate": 3.985293863896525e-06, "loss": 1.6792, "step": 5420 }, { "epoch": 0.58, "learning_rate": 3.983602393092931e-06, "loss": 1.2974, "step": 5421 }, { "epoch": 0.58, "learning_rate": 3.9819110436139755e-06, "loss": 1.4869, "step": 5422 }, { "epoch": 0.58, "learning_rate": 3.980219815661545e-06, "loss": 1.3137, "step": 5423 }, { "epoch": 0.58, "learning_rate": 3.978528709437518e-06, "loss": 1.4685, "step": 5424 }, { "epoch": 0.58, "learning_rate": 3.976837725143758e-06, "loss": 1.4568, "step": 5425 }, { "epoch": 0.58, "learning_rate": 3.975146862982111e-06, "loss": 1.3763, "step": 5426 }, { "epoch": 0.58, "learning_rate": 3.973456123154415e-06, "loss": 1.6383, "step": 5427 }, { "epoch": 0.58, "learning_rate": 3.9717655058624874e-06, "loss": 1.6246, "step": 5428 }, { "epoch": 0.58, "learning_rate": 3.97007501130813e-06, "loss": 1.6975, "step": 5429 }, { "epoch": 0.58, "learning_rate": 3.9683846396931345e-06, "loss": 1.5167, "step": 5430 }, { "epoch": 0.58, "learning_rate": 3.966694391219276e-06, "loss": 1.272, "step": 5431 }, { "epoch": 0.58, "learning_rate": 3.965004266088315e-06, "loss": 1.48, "step": 5432 }, { "epoch": 0.58, "learning_rate": 3.9633142645019965e-06, "loss": 1.3352, "step": 5433 }, { "epoch": 0.58, "learning_rate": 3.961624386662053e-06, "loss": 1.4118, "step": 5434 }, { "epoch": 0.58, "learning_rate": 3.9599346327702e-06, "loss": 1.1944, "step": 5435 }, { "epoch": 0.58, "learning_rate": 3.958245003028136e-06, "loss": 1.4121, "step": 5436 }, { "epoch": 0.58, "learning_rate": 3.956555497637551e-06, "loss": 1.2938, "step": 5437 }, { "epoch": 0.58, "learning_rate": 3.954866116800115e-06, "loss": 1.3463, "step": 5438 }, { "epoch": 0.58, "learning_rate": 3.953176860717488e-06, "loss": 1.4327, "step": 5439 }, { "epoch": 0.58, "learning_rate": 3.951487729591305e-06, "loss": 1.3006, "step": 5440 }, { "epoch": 0.58, "learning_rate": 3.949798723623201e-06, "loss": 1.3087, "step": 5441 }, { "epoch": 0.58, "learning_rate": 3.948109843014784e-06, "loss": 1.5335, "step": 5442 }, { "epoch": 0.58, "learning_rate": 3.94642108796765e-06, "loss": 1.3318, "step": 5443 }, { "epoch": 0.58, "learning_rate": 3.944732458683385e-06, "loss": 1.438, "step": 5444 }, { "epoch": 0.58, "learning_rate": 3.9430439553635504e-06, "loss": 1.3617, "step": 5445 }, { "epoch": 0.58, "learning_rate": 3.941355578209704e-06, "loss": 1.4018, "step": 5446 }, { "epoch": 0.58, "learning_rate": 3.939667327423381e-06, "loss": 1.2829, "step": 5447 }, { "epoch": 0.58, "learning_rate": 3.937979203206103e-06, "loss": 1.3424, "step": 5448 }, { "epoch": 0.58, "learning_rate": 3.936291205759375e-06, "loss": 1.244, "step": 5449 }, { "epoch": 0.58, "learning_rate": 3.93460333528469e-06, "loss": 1.5412, "step": 5450 }, { "epoch": 0.58, "learning_rate": 3.932915591983526e-06, "loss": 1.3352, "step": 5451 }, { "epoch": 0.58, "learning_rate": 3.931227976057344e-06, "loss": 1.2149, "step": 5452 }, { "epoch": 0.58, "learning_rate": 3.929540487707589e-06, "loss": 1.2811, "step": 5453 }, { "epoch": 0.58, "learning_rate": 3.927853127135692e-06, "loss": 1.3383, "step": 5454 }, { "epoch": 0.58, "learning_rate": 3.9261658945430675e-06, "loss": 1.3396, "step": 5455 }, { "epoch": 0.58, "learning_rate": 3.924478790131119e-06, "loss": 1.5754, "step": 5456 }, { "epoch": 0.58, "learning_rate": 3.92279181410123e-06, "loss": 1.0465, "step": 5457 }, { "epoch": 0.58, "learning_rate": 3.921104966654769e-06, "loss": 1.3811, "step": 5458 }, { "epoch": 0.58, "learning_rate": 3.919418247993094e-06, "loss": 1.2914, "step": 5459 }, { "epoch": 0.58, "learning_rate": 3.917731658317538e-06, "loss": 1.3094, "step": 5460 }, { "epoch": 0.58, "learning_rate": 3.916045197829431e-06, "loss": 1.648, "step": 5461 }, { "epoch": 0.58, "learning_rate": 3.9143588667300795e-06, "loss": 1.4571, "step": 5462 }, { "epoch": 0.58, "learning_rate": 3.912672665220773e-06, "loss": 1.2399, "step": 5463 }, { "epoch": 0.58, "learning_rate": 3.910986593502793e-06, "loss": 1.5639, "step": 5464 }, { "epoch": 0.58, "learning_rate": 3.909300651777398e-06, "loss": 1.3632, "step": 5465 }, { "epoch": 0.58, "learning_rate": 3.907614840245836e-06, "loss": 1.5545, "step": 5466 }, { "epoch": 0.58, "learning_rate": 3.90592915910934e-06, "loss": 1.5252, "step": 5467 }, { "epoch": 0.58, "learning_rate": 3.90424360856912e-06, "loss": 1.3597, "step": 5468 }, { "epoch": 0.58, "learning_rate": 3.90255818882638e-06, "loss": 1.3905, "step": 5469 }, { "epoch": 0.58, "learning_rate": 3.9008729000823e-06, "loss": 1.3577, "step": 5470 }, { "epoch": 0.58, "learning_rate": 3.899187742538053e-06, "loss": 1.6105, "step": 5471 }, { "epoch": 0.58, "learning_rate": 3.897502716394789e-06, "loss": 1.6, "step": 5472 }, { "epoch": 0.58, "learning_rate": 3.895817821853646e-06, "loss": 1.406, "step": 5473 }, { "epoch": 0.58, "learning_rate": 3.894133059115745e-06, "loss": 1.4116, "step": 5474 }, { "epoch": 0.58, "learning_rate": 3.892448428382189e-06, "loss": 1.3465, "step": 5475 }, { "epoch": 0.58, "learning_rate": 3.890763929854071e-06, "loss": 1.2031, "step": 5476 }, { "epoch": 0.58, "learning_rate": 3.889079563732465e-06, "loss": 1.3233, "step": 5477 }, { "epoch": 0.58, "learning_rate": 3.887395330218429e-06, "loss": 1.4517, "step": 5478 }, { "epoch": 0.58, "learning_rate": 3.885711229513005e-06, "loss": 1.4762, "step": 5479 }, { "epoch": 0.58, "learning_rate": 3.884027261817216e-06, "loss": 1.374, "step": 5480 }, { "epoch": 0.58, "learning_rate": 3.8823434273320794e-06, "loss": 1.5841, "step": 5481 }, { "epoch": 0.58, "learning_rate": 3.880659726258586e-06, "loss": 1.4215, "step": 5482 }, { "epoch": 0.58, "learning_rate": 3.878976158797715e-06, "loss": 1.3595, "step": 5483 }, { "epoch": 0.58, "learning_rate": 3.877292725150429e-06, "loss": 1.4739, "step": 5484 }, { "epoch": 0.59, "learning_rate": 3.875609425517676e-06, "loss": 1.3263, "step": 5485 }, { "epoch": 0.59, "learning_rate": 3.873926260100386e-06, "loss": 1.2407, "step": 5486 }, { "epoch": 0.59, "learning_rate": 3.872243229099476e-06, "loss": 1.3965, "step": 5487 }, { "epoch": 0.59, "learning_rate": 3.870560332715842e-06, "loss": 1.531, "step": 5488 }, { "epoch": 0.59, "learning_rate": 3.868877571150367e-06, "loss": 1.5437, "step": 5489 }, { "epoch": 0.59, "learning_rate": 3.86719494460392e-06, "loss": 1.506, "step": 5490 }, { "epoch": 0.59, "learning_rate": 3.865512453277351e-06, "loss": 1.3542, "step": 5491 }, { "epoch": 0.59, "learning_rate": 3.863830097371493e-06, "loss": 1.6443, "step": 5492 }, { "epoch": 0.59, "learning_rate": 3.8621478770871645e-06, "loss": 1.4704, "step": 5493 }, { "epoch": 0.59, "learning_rate": 3.860465792625166e-06, "loss": 1.6109, "step": 5494 }, { "epoch": 0.59, "learning_rate": 3.8587838441862884e-06, "loss": 1.5309, "step": 5495 }, { "epoch": 0.59, "learning_rate": 3.857102031971298e-06, "loss": 1.3146, "step": 5496 }, { "epoch": 0.59, "learning_rate": 3.8554203561809475e-06, "loss": 1.3498, "step": 5497 }, { "epoch": 0.59, "learning_rate": 3.853738817015977e-06, "loss": 1.439, "step": 5498 }, { "epoch": 0.59, "learning_rate": 3.852057414677102e-06, "loss": 1.3888, "step": 5499 }, { "epoch": 0.59, "learning_rate": 3.850376149365034e-06, "loss": 1.4431, "step": 5500 }, { "epoch": 0.59, "learning_rate": 3.848695021280455e-06, "loss": 1.4075, "step": 5501 }, { "epoch": 0.59, "learning_rate": 3.84701403062404e-06, "loss": 1.3337, "step": 5502 }, { "epoch": 0.59, "learning_rate": 3.845333177596443e-06, "loss": 1.5758, "step": 5503 }, { "epoch": 0.59, "learning_rate": 3.843652462398303e-06, "loss": 1.3575, "step": 5504 }, { "epoch": 0.59, "learning_rate": 3.841971885230243e-06, "loss": 1.3865, "step": 5505 }, { "epoch": 0.59, "learning_rate": 3.84029144629287e-06, "loss": 1.3827, "step": 5506 }, { "epoch": 0.59, "learning_rate": 3.838611145786771e-06, "loss": 1.5037, "step": 5507 }, { "epoch": 0.59, "learning_rate": 3.83693098391252e-06, "loss": 1.1848, "step": 5508 }, { "epoch": 0.59, "learning_rate": 3.835250960870672e-06, "loss": 1.3632, "step": 5509 }, { "epoch": 0.59, "learning_rate": 3.83357107686177e-06, "loss": 1.4433, "step": 5510 }, { "epoch": 0.59, "learning_rate": 3.8318913320863355e-06, "loss": 1.3123, "step": 5511 }, { "epoch": 0.59, "learning_rate": 3.8302117267448754e-06, "loss": 1.1928, "step": 5512 }, { "epoch": 0.59, "learning_rate": 3.828532261037879e-06, "loss": 1.3615, "step": 5513 }, { "epoch": 0.59, "learning_rate": 3.826852935165818e-06, "loss": 1.4018, "step": 5514 }, { "epoch": 0.59, "learning_rate": 3.825173749329153e-06, "loss": 1.5682, "step": 5515 }, { "epoch": 0.59, "learning_rate": 3.823494703728321e-06, "loss": 1.5909, "step": 5516 }, { "epoch": 0.59, "learning_rate": 3.8218157985637465e-06, "loss": 1.4183, "step": 5517 }, { "epoch": 0.59, "learning_rate": 3.820137034035835e-06, "loss": 1.5577, "step": 5518 }, { "epoch": 0.59, "learning_rate": 3.818458410344974e-06, "loss": 1.3832, "step": 5519 }, { "epoch": 0.59, "learning_rate": 3.816779927691542e-06, "loss": 1.6499, "step": 5520 }, { "epoch": 0.59, "learning_rate": 3.81510158627589e-06, "loss": 1.4854, "step": 5521 }, { "epoch": 0.59, "learning_rate": 3.8134233862983594e-06, "loss": 1.5106, "step": 5522 }, { "epoch": 0.59, "learning_rate": 3.811745327959271e-06, "loss": 1.3915, "step": 5523 }, { "epoch": 0.59, "learning_rate": 3.81006741145893e-06, "loss": 1.4176, "step": 5524 }, { "epoch": 0.59, "learning_rate": 3.808389636997626e-06, "loss": 1.5657, "step": 5525 }, { "epoch": 0.59, "learning_rate": 3.8067120047756313e-06, "loss": 1.5027, "step": 5526 }, { "epoch": 0.59, "learning_rate": 3.805034514993198e-06, "loss": 1.4167, "step": 5527 }, { "epoch": 0.59, "learning_rate": 3.8033571678505626e-06, "loss": 1.3848, "step": 5528 }, { "epoch": 0.59, "learning_rate": 3.801679963547949e-06, "loss": 1.2993, "step": 5529 }, { "epoch": 0.59, "learning_rate": 3.8000029022855587e-06, "loss": 1.3988, "step": 5530 }, { "epoch": 0.59, "learning_rate": 3.798325984263578e-06, "loss": 1.3139, "step": 5531 }, { "epoch": 0.59, "learning_rate": 3.7966492096821773e-06, "loss": 1.25, "step": 5532 }, { "epoch": 0.59, "learning_rate": 3.7949725787415044e-06, "loss": 1.2459, "step": 5533 }, { "epoch": 0.59, "learning_rate": 3.7932960916417004e-06, "loss": 1.4941, "step": 5534 }, { "epoch": 0.59, "learning_rate": 3.7916197485828793e-06, "loss": 1.6411, "step": 5535 }, { "epoch": 0.59, "learning_rate": 3.7899435497651416e-06, "loss": 1.4515, "step": 5536 }, { "epoch": 0.59, "learning_rate": 3.7882674953885736e-06, "loss": 0.9782, "step": 5537 }, { "epoch": 0.59, "learning_rate": 3.786591585653235e-06, "loss": 1.3982, "step": 5538 }, { "epoch": 0.59, "learning_rate": 3.7849158207591806e-06, "loss": 1.1327, "step": 5539 }, { "epoch": 0.59, "learning_rate": 3.783240200906442e-06, "loss": 1.5553, "step": 5540 }, { "epoch": 0.59, "learning_rate": 3.7815647262950293e-06, "loss": 1.4636, "step": 5541 }, { "epoch": 0.59, "learning_rate": 3.7798893971249424e-06, "loss": 1.5054, "step": 5542 }, { "epoch": 0.59, "learning_rate": 3.7782142135961586e-06, "loss": 1.4007, "step": 5543 }, { "epoch": 0.59, "learning_rate": 3.7765391759086424e-06, "loss": 1.5553, "step": 5544 }, { "epoch": 0.59, "learning_rate": 3.774864284262339e-06, "loss": 1.3842, "step": 5545 }, { "epoch": 0.59, "learning_rate": 3.7731895388571725e-06, "loss": 1.301, "step": 5546 }, { "epoch": 0.59, "learning_rate": 3.771514939893055e-06, "loss": 1.372, "step": 5547 }, { "epoch": 0.59, "learning_rate": 3.769840487569877e-06, "loss": 1.375, "step": 5548 }, { "epoch": 0.59, "learning_rate": 3.7681661820875156e-06, "loss": 1.1182, "step": 5549 }, { "epoch": 0.59, "learning_rate": 3.766492023645827e-06, "loss": 1.4435, "step": 5550 }, { "epoch": 0.59, "learning_rate": 3.764818012444652e-06, "loss": 1.3952, "step": 5551 }, { "epoch": 0.59, "learning_rate": 3.7631441486838115e-06, "loss": 1.5913, "step": 5552 }, { "epoch": 0.59, "learning_rate": 3.761470432563109e-06, "loss": 1.5647, "step": 5553 }, { "epoch": 0.59, "learning_rate": 3.7597968642823348e-06, "loss": 1.6247, "step": 5554 }, { "epoch": 0.59, "learning_rate": 3.758123444041255e-06, "loss": 1.0512, "step": 5555 }, { "epoch": 0.59, "learning_rate": 3.7564501720396242e-06, "loss": 1.2417, "step": 5556 }, { "epoch": 0.59, "learning_rate": 3.7547770484771745e-06, "loss": 1.5572, "step": 5557 }, { "epoch": 0.59, "learning_rate": 3.7531040735536194e-06, "loss": 1.4502, "step": 5558 }, { "epoch": 0.59, "learning_rate": 3.7514312474686643e-06, "loss": 1.4468, "step": 5559 }, { "epoch": 0.59, "learning_rate": 3.7497585704219845e-06, "loss": 1.5418, "step": 5560 }, { "epoch": 0.59, "learning_rate": 3.748086042613245e-06, "loss": 1.2713, "step": 5561 }, { "epoch": 0.59, "learning_rate": 3.74641366424209e-06, "loss": 1.5304, "step": 5562 }, { "epoch": 0.59, "learning_rate": 3.7447414355081465e-06, "loss": 1.3562, "step": 5563 }, { "epoch": 0.59, "learning_rate": 3.7430693566110254e-06, "loss": 1.4273, "step": 5564 }, { "epoch": 0.59, "learning_rate": 3.7413974277503183e-06, "loss": 1.4054, "step": 5565 }, { "epoch": 0.59, "learning_rate": 3.7397256491255973e-06, "loss": 1.2849, "step": 5566 }, { "epoch": 0.59, "learning_rate": 3.738054020936418e-06, "loss": 1.3992, "step": 5567 }, { "epoch": 0.59, "learning_rate": 3.7363825433823187e-06, "loss": 1.5629, "step": 5568 }, { "epoch": 0.59, "learning_rate": 3.7347112166628207e-06, "loss": 1.5156, "step": 5569 }, { "epoch": 0.59, "learning_rate": 3.7330400409774236e-06, "loss": 1.7792, "step": 5570 }, { "epoch": 0.59, "learning_rate": 3.7313690165256134e-06, "loss": 1.356, "step": 5571 }, { "epoch": 0.59, "learning_rate": 3.729698143506851e-06, "loss": 1.4846, "step": 5572 }, { "epoch": 0.59, "learning_rate": 3.7280274221205907e-06, "loss": 1.5262, "step": 5573 }, { "epoch": 0.59, "learning_rate": 3.7263568525662574e-06, "loss": 1.4247, "step": 5574 }, { "epoch": 0.59, "learning_rate": 3.7246864350432636e-06, "loss": 1.478, "step": 5575 }, { "epoch": 0.59, "learning_rate": 3.723016169751003e-06, "loss": 1.5006, "step": 5576 }, { "epoch": 0.59, "learning_rate": 3.7213460568888493e-06, "loss": 1.4908, "step": 5577 }, { "epoch": 0.59, "learning_rate": 3.7196760966561607e-06, "loss": 1.5234, "step": 5578 }, { "epoch": 0.6, "learning_rate": 3.7180062892522774e-06, "loss": 1.5832, "step": 5579 }, { "epoch": 0.6, "learning_rate": 3.716336634876516e-06, "loss": 1.5054, "step": 5580 }, { "epoch": 0.6, "learning_rate": 3.71466713372818e-06, "loss": 1.5664, "step": 5581 }, { "epoch": 0.6, "learning_rate": 3.712997786006554e-06, "loss": 1.3949, "step": 5582 }, { "epoch": 0.6, "learning_rate": 3.711328591910904e-06, "loss": 1.6258, "step": 5583 }, { "epoch": 0.6, "learning_rate": 3.709659551640476e-06, "loss": 1.477, "step": 5584 }, { "epoch": 0.6, "learning_rate": 3.7079906653945008e-06, "loss": 1.4482, "step": 5585 }, { "epoch": 0.6, "learning_rate": 3.7063219333721857e-06, "loss": 1.5162, "step": 5586 }, { "epoch": 0.6, "learning_rate": 3.7046533557727238e-06, "loss": 1.341, "step": 5587 }, { "epoch": 0.6, "learning_rate": 3.7029849327952897e-06, "loss": 1.4186, "step": 5588 }, { "epoch": 0.6, "learning_rate": 3.7013166646390384e-06, "loss": 1.5252, "step": 5589 }, { "epoch": 0.6, "learning_rate": 3.6996485515031077e-06, "loss": 1.2306, "step": 5590 }, { "epoch": 0.6, "learning_rate": 3.6979805935866125e-06, "loss": 1.4299, "step": 5591 }, { "epoch": 0.6, "learning_rate": 3.6963127910886526e-06, "loss": 1.6821, "step": 5592 }, { "epoch": 0.6, "learning_rate": 3.694645144208314e-06, "loss": 1.7653, "step": 5593 }, { "epoch": 0.6, "learning_rate": 3.6929776531446544e-06, "loss": 1.0517, "step": 5594 }, { "epoch": 0.6, "learning_rate": 3.691310318096719e-06, "loss": 1.5504, "step": 5595 }, { "epoch": 0.6, "learning_rate": 3.6896431392635345e-06, "loss": 1.5843, "step": 5596 }, { "epoch": 0.6, "learning_rate": 3.6879761168441038e-06, "loss": 1.4746, "step": 5597 }, { "epoch": 0.6, "learning_rate": 3.6863092510374198e-06, "loss": 1.6383, "step": 5598 }, { "epoch": 0.6, "learning_rate": 3.684642542042449e-06, "loss": 1.5019, "step": 5599 }, { "epoch": 0.6, "learning_rate": 3.6829759900581413e-06, "loss": 1.3477, "step": 5600 }, { "epoch": 0.6, "learning_rate": 3.68130959528343e-06, "loss": 1.347, "step": 5601 }, { "epoch": 0.6, "learning_rate": 3.6796433579172265e-06, "loss": 1.7497, "step": 5602 }, { "epoch": 0.6, "learning_rate": 3.677977278158427e-06, "loss": 1.272, "step": 5603 }, { "epoch": 0.6, "learning_rate": 3.6763113562059077e-06, "loss": 1.3053, "step": 5604 }, { "epoch": 0.6, "learning_rate": 3.6746455922585222e-06, "loss": 1.2755, "step": 5605 }, { "epoch": 0.6, "learning_rate": 3.6729799865151094e-06, "loss": 1.4125, "step": 5606 }, { "epoch": 0.6, "learning_rate": 3.6713145391744877e-06, "loss": 1.4164, "step": 5607 }, { "epoch": 0.6, "learning_rate": 3.6696492504354585e-06, "loss": 1.465, "step": 5608 }, { "epoch": 0.6, "learning_rate": 3.6679841204968025e-06, "loss": 1.418, "step": 5609 }, { "epoch": 0.6, "learning_rate": 3.6663191495572827e-06, "loss": 1.613, "step": 5610 }, { "epoch": 0.6, "learning_rate": 3.664654337815639e-06, "loss": 1.3122, "step": 5611 }, { "epoch": 0.6, "learning_rate": 3.6629896854705963e-06, "loss": 1.4064, "step": 5612 }, { "epoch": 0.6, "learning_rate": 3.661325192720862e-06, "loss": 1.3377, "step": 5613 }, { "epoch": 0.6, "learning_rate": 3.6596608597651205e-06, "loss": 1.2855, "step": 5614 }, { "epoch": 0.6, "learning_rate": 3.657996686802039e-06, "loss": 1.5096, "step": 5615 }, { "epoch": 0.6, "learning_rate": 3.6563326740302664e-06, "loss": 1.5908, "step": 5616 }, { "epoch": 0.6, "learning_rate": 3.6546688216484272e-06, "loss": 1.457, "step": 5617 }, { "epoch": 0.6, "learning_rate": 3.6530051298551372e-06, "loss": 1.5718, "step": 5618 }, { "epoch": 0.6, "learning_rate": 3.6513415988489824e-06, "loss": 1.4937, "step": 5619 }, { "epoch": 0.6, "learning_rate": 3.649678228828535e-06, "loss": 1.3499, "step": 5620 }, { "epoch": 0.6, "learning_rate": 3.648015019992347e-06, "loss": 1.4592, "step": 5621 }, { "epoch": 0.6, "learning_rate": 3.6463519725389516e-06, "loss": 1.498, "step": 5622 }, { "epoch": 0.6, "learning_rate": 3.6446890866668627e-06, "loss": 1.3735, "step": 5623 }, { "epoch": 0.6, "learning_rate": 3.6430263625745744e-06, "loss": 1.4124, "step": 5624 }, { "epoch": 0.6, "learning_rate": 3.64136380046056e-06, "loss": 1.5609, "step": 5625 }, { "epoch": 0.6, "learning_rate": 3.6397014005232754e-06, "loss": 1.2679, "step": 5626 }, { "epoch": 0.6, "learning_rate": 3.6380391629611585e-06, "loss": 1.3515, "step": 5627 }, { "epoch": 0.6, "learning_rate": 3.6363770879726247e-06, "loss": 1.3459, "step": 5628 }, { "epoch": 0.6, "learning_rate": 3.6347151757560716e-06, "loss": 1.4338, "step": 5629 }, { "epoch": 0.6, "learning_rate": 3.6330534265098793e-06, "loss": 1.4283, "step": 5630 }, { "epoch": 0.6, "learning_rate": 3.6313918404324e-06, "loss": 1.3797, "step": 5631 }, { "epoch": 0.6, "learning_rate": 3.629730417721981e-06, "loss": 1.2437, "step": 5632 }, { "epoch": 0.6, "learning_rate": 3.6280691585769368e-06, "loss": 1.3001, "step": 5633 }, { "epoch": 0.6, "learning_rate": 3.6264080631955683e-06, "loss": 1.298, "step": 5634 }, { "epoch": 0.6, "learning_rate": 3.6247471317761572e-06, "loss": 1.4499, "step": 5635 }, { "epoch": 0.6, "learning_rate": 3.62308636451696e-06, "loss": 1.4214, "step": 5636 }, { "epoch": 0.6, "learning_rate": 3.621425761616224e-06, "loss": 1.4846, "step": 5637 }, { "epoch": 0.6, "learning_rate": 3.6197653232721696e-06, "loss": 1.3465, "step": 5638 }, { "epoch": 0.6, "learning_rate": 3.6181050496829963e-06, "loss": 1.0858, "step": 5639 }, { "epoch": 0.6, "learning_rate": 3.616444941046887e-06, "loss": 1.3424, "step": 5640 }, { "epoch": 0.6, "learning_rate": 3.614784997562004e-06, "loss": 1.4934, "step": 5641 }, { "epoch": 0.6, "learning_rate": 3.6131252194264932e-06, "loss": 1.6203, "step": 5642 }, { "epoch": 0.6, "learning_rate": 3.6114656068384767e-06, "loss": 1.4319, "step": 5643 }, { "epoch": 0.6, "learning_rate": 3.609806159996056e-06, "loss": 1.4265, "step": 5644 }, { "epoch": 0.6, "learning_rate": 3.608146879097316e-06, "loss": 1.5545, "step": 5645 }, { "epoch": 0.6, "learning_rate": 3.6064877643403194e-06, "loss": 1.4479, "step": 5646 }, { "epoch": 0.6, "learning_rate": 3.604828815923113e-06, "loss": 1.6684, "step": 5647 }, { "epoch": 0.6, "learning_rate": 3.6031700340437188e-06, "loss": 1.4936, "step": 5648 }, { "epoch": 0.6, "learning_rate": 3.601511418900143e-06, "loss": 1.4783, "step": 5649 }, { "epoch": 0.6, "learning_rate": 3.5998529706903673e-06, "loss": 1.3915, "step": 5650 }, { "epoch": 0.6, "learning_rate": 3.5981946896123576e-06, "loss": 1.3564, "step": 5651 }, { "epoch": 0.6, "learning_rate": 3.5965365758640587e-06, "loss": 1.4454, "step": 5652 }, { "epoch": 0.6, "learning_rate": 3.5948786296433948e-06, "loss": 1.6396, "step": 5653 }, { "epoch": 0.6, "learning_rate": 3.5932208511482707e-06, "loss": 1.5369, "step": 5654 }, { "epoch": 0.6, "learning_rate": 3.591563240576572e-06, "loss": 1.4062, "step": 5655 }, { "epoch": 0.6, "learning_rate": 3.5899057981261586e-06, "loss": 1.2254, "step": 5656 }, { "epoch": 0.6, "learning_rate": 3.588248523994882e-06, "loss": 1.4343, "step": 5657 }, { "epoch": 0.6, "learning_rate": 3.5865914183805606e-06, "loss": 1.403, "step": 5658 }, { "epoch": 0.6, "learning_rate": 3.584934481481002e-06, "loss": 1.3834, "step": 5659 }, { "epoch": 0.6, "learning_rate": 3.5832777134939888e-06, "loss": 1.5472, "step": 5660 }, { "epoch": 0.6, "learning_rate": 3.581621114617284e-06, "loss": 1.4086, "step": 5661 }, { "epoch": 0.6, "learning_rate": 3.5799646850486334e-06, "loss": 1.3261, "step": 5662 }, { "epoch": 0.6, "learning_rate": 3.5783084249857615e-06, "loss": 1.515, "step": 5663 }, { "epoch": 0.6, "learning_rate": 3.5766523346263682e-06, "loss": 1.3881, "step": 5664 }, { "epoch": 0.6, "learning_rate": 3.574996414168137e-06, "loss": 1.844, "step": 5665 }, { "epoch": 0.6, "learning_rate": 3.573340663808733e-06, "loss": 1.3704, "step": 5666 }, { "epoch": 0.6, "learning_rate": 3.571685083745798e-06, "loss": 1.499, "step": 5667 }, { "epoch": 0.6, "learning_rate": 3.5700296741769524e-06, "loss": 1.1998, "step": 5668 }, { "epoch": 0.6, "learning_rate": 3.568374435299801e-06, "loss": 1.3823, "step": 5669 }, { "epoch": 0.6, "learning_rate": 3.56671936731192e-06, "loss": 1.2167, "step": 5670 }, { "epoch": 0.6, "learning_rate": 3.5650644704108767e-06, "loss": 1.6542, "step": 5671 }, { "epoch": 0.61, "learning_rate": 3.563409744794207e-06, "loss": 1.5161, "step": 5672 }, { "epoch": 0.61, "learning_rate": 3.561755190659434e-06, "loss": 1.612, "step": 5673 }, { "epoch": 0.61, "learning_rate": 3.5601008082040545e-06, "loss": 1.4244, "step": 5674 }, { "epoch": 0.61, "learning_rate": 3.5584465976255487e-06, "loss": 1.384, "step": 5675 }, { "epoch": 0.61, "learning_rate": 3.556792559121377e-06, "loss": 1.5908, "step": 5676 }, { "epoch": 0.61, "learning_rate": 3.5551386928889774e-06, "loss": 1.311, "step": 5677 }, { "epoch": 0.61, "learning_rate": 3.553484999125765e-06, "loss": 1.5426, "step": 5678 }, { "epoch": 0.61, "learning_rate": 3.5518314780291384e-06, "loss": 1.1465, "step": 5679 }, { "epoch": 0.61, "learning_rate": 3.550178129796472e-06, "loss": 1.0846, "step": 5680 }, { "epoch": 0.61, "learning_rate": 3.5485249546251244e-06, "loss": 1.514, "step": 5681 }, { "epoch": 0.61, "learning_rate": 3.5468719527124294e-06, "loss": 1.3398, "step": 5682 }, { "epoch": 0.61, "learning_rate": 3.545219124255702e-06, "loss": 1.3557, "step": 5683 }, { "epoch": 0.61, "learning_rate": 3.5435664694522343e-06, "loss": 1.698, "step": 5684 }, { "epoch": 0.61, "learning_rate": 3.541913988499299e-06, "loss": 1.5394, "step": 5685 }, { "epoch": 0.61, "learning_rate": 3.5402616815941504e-06, "loss": 1.387, "step": 5686 }, { "epoch": 0.61, "learning_rate": 3.5386095489340188e-06, "loss": 1.4828, "step": 5687 }, { "epoch": 0.61, "learning_rate": 3.5369575907161167e-06, "loss": 1.3632, "step": 5688 }, { "epoch": 0.61, "learning_rate": 3.53530580713763e-06, "loss": 1.529, "step": 5689 }, { "epoch": 0.61, "learning_rate": 3.5336541983957285e-06, "loss": 1.403, "step": 5690 }, { "epoch": 0.61, "learning_rate": 3.5320027646875643e-06, "loss": 1.2957, "step": 5691 }, { "epoch": 0.61, "learning_rate": 3.5303515062102604e-06, "loss": 1.3379, "step": 5692 }, { "epoch": 0.61, "learning_rate": 3.5287004231609245e-06, "loss": 1.4322, "step": 5693 }, { "epoch": 0.61, "learning_rate": 3.5270495157366434e-06, "loss": 1.3811, "step": 5694 }, { "epoch": 0.61, "learning_rate": 3.5253987841344766e-06, "loss": 1.4545, "step": 5695 }, { "epoch": 0.61, "learning_rate": 3.523748228551474e-06, "loss": 1.2325, "step": 5696 }, { "epoch": 0.61, "learning_rate": 3.5220978491846534e-06, "loss": 1.5822, "step": 5697 }, { "epoch": 0.61, "learning_rate": 3.5204476462310177e-06, "loss": 1.5175, "step": 5698 }, { "epoch": 0.61, "learning_rate": 3.5187976198875466e-06, "loss": 1.2221, "step": 5699 }, { "epoch": 0.61, "learning_rate": 3.517147770351199e-06, "loss": 1.5358, "step": 5700 }, { "epoch": 0.61, "learning_rate": 3.5154980978189147e-06, "loss": 1.377, "step": 5701 }, { "epoch": 0.61, "learning_rate": 3.5138486024876107e-06, "loss": 1.2411, "step": 5702 }, { "epoch": 0.61, "learning_rate": 3.5121992845541797e-06, "loss": 1.2951, "step": 5703 }, { "epoch": 0.61, "learning_rate": 3.5105501442154986e-06, "loss": 1.2798, "step": 5704 }, { "epoch": 0.61, "learning_rate": 3.5089011816684203e-06, "loss": 1.6907, "step": 5705 }, { "epoch": 0.61, "learning_rate": 3.507252397109777e-06, "loss": 1.2363, "step": 5706 }, { "epoch": 0.61, "learning_rate": 3.505603790736381e-06, "loss": 1.2257, "step": 5707 }, { "epoch": 0.61, "learning_rate": 3.5039553627450213e-06, "loss": 1.4106, "step": 5708 }, { "epoch": 0.61, "learning_rate": 3.5023071133324627e-06, "loss": 1.4121, "step": 5709 }, { "epoch": 0.61, "learning_rate": 3.500659042695459e-06, "loss": 1.5932, "step": 5710 }, { "epoch": 0.61, "learning_rate": 3.4990111510307305e-06, "loss": 1.3997, "step": 5711 }, { "epoch": 0.61, "learning_rate": 3.497363438534984e-06, "loss": 1.376, "step": 5712 }, { "epoch": 0.61, "learning_rate": 3.4957159054049015e-06, "loss": 1.2538, "step": 5713 }, { "epoch": 0.61, "learning_rate": 3.4940685518371444e-06, "loss": 1.4669, "step": 5714 }, { "epoch": 0.61, "learning_rate": 3.4924213780283545e-06, "loss": 1.4381, "step": 5715 }, { "epoch": 0.61, "learning_rate": 3.4907743841751494e-06, "loss": 1.479, "step": 5716 }, { "epoch": 0.61, "learning_rate": 3.4891275704741267e-06, "loss": 1.536, "step": 5717 }, { "epoch": 0.61, "learning_rate": 3.4874809371218608e-06, "loss": 1.341, "step": 5718 }, { "epoch": 0.61, "learning_rate": 3.485834484314906e-06, "loss": 1.2705, "step": 5719 }, { "epoch": 0.61, "learning_rate": 3.4841882122497973e-06, "loss": 1.4098, "step": 5720 }, { "epoch": 0.61, "learning_rate": 3.4825421211230437e-06, "loss": 1.3584, "step": 5721 }, { "epoch": 0.61, "learning_rate": 3.4808962111311354e-06, "loss": 1.4913, "step": 5722 }, { "epoch": 0.61, "learning_rate": 3.47925048247054e-06, "loss": 1.6055, "step": 5723 }, { "epoch": 0.61, "learning_rate": 3.4776049353377016e-06, "loss": 1.6242, "step": 5724 }, { "epoch": 0.61, "learning_rate": 3.4759595699290482e-06, "loss": 1.3933, "step": 5725 }, { "epoch": 0.61, "learning_rate": 3.4743143864409817e-06, "loss": 1.4275, "step": 5726 }, { "epoch": 0.61, "learning_rate": 3.4726693850698824e-06, "loss": 1.6664, "step": 5727 }, { "epoch": 0.61, "learning_rate": 3.4710245660121107e-06, "loss": 1.5188, "step": 5728 }, { "epoch": 0.61, "learning_rate": 3.469379929464e-06, "loss": 1.4369, "step": 5729 }, { "epoch": 0.61, "learning_rate": 3.467735475621873e-06, "loss": 1.5389, "step": 5730 }, { "epoch": 0.61, "learning_rate": 3.4660912046820195e-06, "loss": 1.4123, "step": 5731 }, { "epoch": 0.61, "learning_rate": 3.4644471168407124e-06, "loss": 1.5307, "step": 5732 }, { "epoch": 0.61, "learning_rate": 3.4628032122942024e-06, "loss": 1.6044, "step": 5733 }, { "epoch": 0.61, "learning_rate": 3.4611594912387148e-06, "loss": 1.6564, "step": 5734 }, { "epoch": 0.61, "learning_rate": 3.4595159538704613e-06, "loss": 1.1415, "step": 5735 }, { "epoch": 0.61, "learning_rate": 3.4578726003856245e-06, "loss": 1.3558, "step": 5736 }, { "epoch": 0.61, "learning_rate": 3.456229430980367e-06, "loss": 1.3997, "step": 5737 }, { "epoch": 0.61, "learning_rate": 3.454586445850828e-06, "loss": 1.5355, "step": 5738 }, { "epoch": 0.61, "learning_rate": 3.4529436451931263e-06, "loss": 1.2798, "step": 5739 }, { "epoch": 0.61, "learning_rate": 3.451301029203361e-06, "loss": 1.4977, "step": 5740 }, { "epoch": 0.61, "learning_rate": 3.4496585980776066e-06, "loss": 1.6985, "step": 5741 }, { "epoch": 0.61, "learning_rate": 3.448016352011914e-06, "loss": 1.2792, "step": 5742 }, { "epoch": 0.61, "learning_rate": 3.4463742912023127e-06, "loss": 1.3233, "step": 5743 }, { "epoch": 0.61, "learning_rate": 3.4447324158448126e-06, "loss": 1.7556, "step": 5744 }, { "epoch": 0.61, "learning_rate": 3.4430907261354e-06, "loss": 1.4186, "step": 5745 }, { "epoch": 0.61, "learning_rate": 3.4414492222700394e-06, "loss": 1.5592, "step": 5746 }, { "epoch": 0.61, "learning_rate": 3.439807904444673e-06, "loss": 1.3953, "step": 5747 }, { "epoch": 0.61, "learning_rate": 3.438166772855218e-06, "loss": 1.273, "step": 5748 }, { "epoch": 0.61, "learning_rate": 3.4365258276975734e-06, "loss": 1.6895, "step": 5749 }, { "epoch": 0.61, "learning_rate": 3.4348850691676146e-06, "loss": 1.5829, "step": 5750 }, { "epoch": 0.61, "learning_rate": 3.4332444974611946e-06, "loss": 1.3221, "step": 5751 }, { "epoch": 0.61, "learning_rate": 3.4316041127741428e-06, "loss": 1.5296, "step": 5752 }, { "epoch": 0.61, "learning_rate": 3.4299639153022686e-06, "loss": 1.3124, "step": 5753 }, { "epoch": 0.61, "learning_rate": 3.428323905241358e-06, "loss": 1.2755, "step": 5754 }, { "epoch": 0.61, "learning_rate": 3.4266840827871755e-06, "loss": 1.2103, "step": 5755 }, { "epoch": 0.61, "learning_rate": 3.42504444813546e-06, "loss": 1.4685, "step": 5756 }, { "epoch": 0.61, "learning_rate": 3.4234050014819308e-06, "loss": 1.6381, "step": 5757 }, { "epoch": 0.61, "learning_rate": 3.4217657430222845e-06, "loss": 1.4097, "step": 5758 }, { "epoch": 0.61, "learning_rate": 3.4201266729521958e-06, "loss": 1.4288, "step": 5759 }, { "epoch": 0.61, "learning_rate": 3.4184877914673155e-06, "loss": 1.2803, "step": 5760 }, { "epoch": 0.61, "learning_rate": 3.416849098763273e-06, "loss": 1.5026, "step": 5761 }, { "epoch": 0.61, "learning_rate": 3.415210595035674e-06, "loss": 1.5037, "step": 5762 }, { "epoch": 0.61, "learning_rate": 3.4135722804801004e-06, "loss": 1.4394, "step": 5763 }, { "epoch": 0.61, "learning_rate": 3.411934155292116e-06, "loss": 1.3622, "step": 5764 }, { "epoch": 0.61, "learning_rate": 3.4102962196672597e-06, "loss": 1.3417, "step": 5765 }, { "epoch": 0.62, "learning_rate": 3.4086584738010455e-06, "loss": 1.2612, "step": 5766 }, { "epoch": 0.62, "learning_rate": 3.4070209178889684e-06, "loss": 1.3632, "step": 5767 }, { "epoch": 0.62, "learning_rate": 3.4053835521264956e-06, "loss": 1.4896, "step": 5768 }, { "epoch": 0.62, "learning_rate": 3.4037463767090807e-06, "loss": 1.3416, "step": 5769 }, { "epoch": 0.62, "learning_rate": 3.4021093918321445e-06, "loss": 1.5102, "step": 5770 }, { "epoch": 0.62, "learning_rate": 3.400472597691091e-06, "loss": 1.5559, "step": 5771 }, { "epoch": 0.62, "learning_rate": 3.3988359944812997e-06, "loss": 1.5316, "step": 5772 }, { "epoch": 0.62, "learning_rate": 3.3971995823981265e-06, "loss": 1.6497, "step": 5773 }, { "epoch": 0.62, "learning_rate": 3.3955633616369066e-06, "loss": 1.2843, "step": 5774 }, { "epoch": 0.62, "learning_rate": 3.3939273323929533e-06, "loss": 1.554, "step": 5775 }, { "epoch": 0.62, "learning_rate": 3.3922914948615515e-06, "loss": 1.6127, "step": 5776 }, { "epoch": 0.62, "learning_rate": 3.390655849237967e-06, "loss": 1.7344, "step": 5777 }, { "epoch": 0.62, "learning_rate": 3.3890203957174437e-06, "loss": 1.6009, "step": 5778 }, { "epoch": 0.62, "learning_rate": 3.3873851344952013e-06, "loss": 1.4115, "step": 5779 }, { "epoch": 0.62, "learning_rate": 3.3857500657664368e-06, "loss": 1.3006, "step": 5780 }, { "epoch": 0.62, "learning_rate": 3.3841151897263234e-06, "loss": 1.5209, "step": 5781 }, { "epoch": 0.62, "learning_rate": 3.3824805065700113e-06, "loss": 1.4006, "step": 5782 }, { "epoch": 0.62, "learning_rate": 3.380846016492627e-06, "loss": 1.3288, "step": 5783 }, { "epoch": 0.62, "learning_rate": 3.379211719689278e-06, "loss": 1.3306, "step": 5784 }, { "epoch": 0.62, "learning_rate": 3.3775776163550455e-06, "loss": 1.4542, "step": 5785 }, { "epoch": 0.62, "learning_rate": 3.3759437066849876e-06, "loss": 1.518, "step": 5786 }, { "epoch": 0.62, "learning_rate": 3.3743099908741385e-06, "loss": 1.2808, "step": 5787 }, { "epoch": 0.62, "learning_rate": 3.3726764691175085e-06, "loss": 1.249, "step": 5788 }, { "epoch": 0.62, "learning_rate": 3.371043141610093e-06, "loss": 1.3986, "step": 5789 }, { "epoch": 0.62, "learning_rate": 3.3694100085468535e-06, "loss": 1.3326, "step": 5790 }, { "epoch": 0.62, "learning_rate": 3.367777070122733e-06, "loss": 1.1489, "step": 5791 }, { "epoch": 0.62, "learning_rate": 3.3661443265326528e-06, "loss": 1.4303, "step": 5792 }, { "epoch": 0.62, "learning_rate": 3.364511777971504e-06, "loss": 1.5285, "step": 5793 }, { "epoch": 0.62, "learning_rate": 3.3628794246341667e-06, "loss": 1.3503, "step": 5794 }, { "epoch": 0.62, "learning_rate": 3.361247266715486e-06, "loss": 1.4097, "step": 5795 }, { "epoch": 0.62, "learning_rate": 3.3596153044102897e-06, "loss": 1.4824, "step": 5796 }, { "epoch": 0.62, "learning_rate": 3.3579835379133787e-06, "loss": 1.5717, "step": 5797 }, { "epoch": 0.62, "learning_rate": 3.356351967419535e-06, "loss": 1.4904, "step": 5798 }, { "epoch": 0.62, "learning_rate": 3.354720593123514e-06, "loss": 1.5469, "step": 5799 }, { "epoch": 0.62, "learning_rate": 3.353089415220049e-06, "loss": 1.2504, "step": 5800 }, { "epoch": 0.62, "learning_rate": 3.3514584339038476e-06, "loss": 1.4573, "step": 5801 }, { "epoch": 0.62, "learning_rate": 3.349827649369596e-06, "loss": 1.7237, "step": 5802 }, { "epoch": 0.62, "learning_rate": 3.3481970618119575e-06, "loss": 1.5382, "step": 5803 }, { "epoch": 0.62, "learning_rate": 3.3465666714255707e-06, "loss": 1.4313, "step": 5804 }, { "epoch": 0.62, "learning_rate": 3.3449364784050515e-06, "loss": 1.4769, "step": 5805 }, { "epoch": 0.62, "learning_rate": 3.3433064829449913e-06, "loss": 1.0682, "step": 5806 }, { "epoch": 0.62, "learning_rate": 3.3416766852399553e-06, "loss": 1.5952, "step": 5807 }, { "epoch": 0.62, "learning_rate": 3.3400470854844925e-06, "loss": 1.2015, "step": 5808 }, { "epoch": 0.62, "learning_rate": 3.338417683873122e-06, "loss": 1.4398, "step": 5809 }, { "epoch": 0.62, "learning_rate": 3.3367884806003402e-06, "loss": 1.706, "step": 5810 }, { "epoch": 0.62, "learning_rate": 3.3351594758606222e-06, "loss": 1.3758, "step": 5811 }, { "epoch": 0.62, "learning_rate": 3.333530669848416e-06, "loss": 1.2243, "step": 5812 }, { "epoch": 0.62, "learning_rate": 3.3319020627581494e-06, "loss": 1.3489, "step": 5813 }, { "epoch": 0.62, "learning_rate": 3.3302736547842263e-06, "loss": 1.405, "step": 5814 }, { "epoch": 0.62, "learning_rate": 3.3286454461210216e-06, "loss": 1.4231, "step": 5815 }, { "epoch": 0.62, "learning_rate": 3.327017436962892e-06, "loss": 1.6344, "step": 5816 }, { "epoch": 0.62, "learning_rate": 3.3253896275041677e-06, "loss": 1.3263, "step": 5817 }, { "epoch": 0.62, "learning_rate": 3.3237620179391584e-06, "loss": 1.3027, "step": 5818 }, { "epoch": 0.62, "learning_rate": 3.3221346084621447e-06, "loss": 1.3812, "step": 5819 }, { "epoch": 0.62, "learning_rate": 3.3205073992673885e-06, "loss": 1.3645, "step": 5820 }, { "epoch": 0.62, "learning_rate": 3.3188803905491233e-06, "loss": 1.1663, "step": 5821 }, { "epoch": 0.62, "learning_rate": 3.3172535825015608e-06, "loss": 1.3435, "step": 5822 }, { "epoch": 0.62, "learning_rate": 3.3156269753188895e-06, "loss": 1.3652, "step": 5823 }, { "epoch": 0.62, "learning_rate": 3.3140005691952735e-06, "loss": 1.6248, "step": 5824 }, { "epoch": 0.62, "learning_rate": 3.312374364324852e-06, "loss": 1.4793, "step": 5825 }, { "epoch": 0.62, "learning_rate": 3.310748360901741e-06, "loss": 1.4619, "step": 5826 }, { "epoch": 0.62, "learning_rate": 3.3091225591200292e-06, "loss": 1.1914, "step": 5827 }, { "epoch": 0.62, "learning_rate": 3.3074969591737906e-06, "loss": 1.5063, "step": 5828 }, { "epoch": 0.62, "learning_rate": 3.3058715612570623e-06, "loss": 1.273, "step": 5829 }, { "epoch": 0.62, "learning_rate": 3.304246365563867e-06, "loss": 1.1436, "step": 5830 }, { "epoch": 0.62, "learning_rate": 3.3026213722882e-06, "loss": 1.6583, "step": 5831 }, { "epoch": 0.62, "learning_rate": 3.300996581624028e-06, "loss": 1.4462, "step": 5832 }, { "epoch": 0.62, "learning_rate": 3.2993719937653037e-06, "loss": 1.4947, "step": 5833 }, { "epoch": 0.62, "learning_rate": 3.2977476089059484e-06, "loss": 1.2256, "step": 5834 }, { "epoch": 0.62, "learning_rate": 3.2961234272398578e-06, "loss": 1.2919, "step": 5835 }, { "epoch": 0.62, "learning_rate": 3.294499448960909e-06, "loss": 1.5047, "step": 5836 }, { "epoch": 0.62, "learning_rate": 3.2928756742629486e-06, "loss": 1.3698, "step": 5837 }, { "epoch": 0.62, "learning_rate": 3.291252103339806e-06, "loss": 1.6637, "step": 5838 }, { "epoch": 0.62, "learning_rate": 3.289628736385281e-06, "loss": 1.3264, "step": 5839 }, { "epoch": 0.62, "learning_rate": 3.28800557359315e-06, "loss": 1.2823, "step": 5840 }, { "epoch": 0.62, "learning_rate": 3.2863826151571654e-06, "loss": 1.2867, "step": 5841 }, { "epoch": 0.62, "learning_rate": 3.2847598612710546e-06, "loss": 1.3803, "step": 5842 }, { "epoch": 0.62, "learning_rate": 3.283137312128524e-06, "loss": 1.2691, "step": 5843 }, { "epoch": 0.62, "learning_rate": 3.2815149679232507e-06, "loss": 1.4261, "step": 5844 }, { "epoch": 0.62, "learning_rate": 3.2798928288488917e-06, "loss": 1.6141, "step": 5845 }, { "epoch": 0.62, "learning_rate": 3.278270895099073e-06, "loss": 1.5555, "step": 5846 }, { "epoch": 0.62, "learning_rate": 3.276649166867406e-06, "loss": 1.2775, "step": 5847 }, { "epoch": 0.62, "learning_rate": 3.275027644347469e-06, "loss": 1.4151, "step": 5848 }, { "epoch": 0.62, "learning_rate": 3.2734063277328187e-06, "loss": 1.3812, "step": 5849 }, { "epoch": 0.62, "learning_rate": 3.271785217216987e-06, "loss": 1.6645, "step": 5850 }, { "epoch": 0.62, "learning_rate": 3.270164312993482e-06, "loss": 1.4257, "step": 5851 }, { "epoch": 0.62, "learning_rate": 3.268543615255788e-06, "loss": 1.1956, "step": 5852 }, { "epoch": 0.62, "learning_rate": 3.266923124197363e-06, "loss": 1.3047, "step": 5853 }, { "epoch": 0.62, "learning_rate": 3.2653028400116395e-06, "loss": 1.4226, "step": 5854 }, { "epoch": 0.62, "learning_rate": 3.2636827628920258e-06, "loss": 1.478, "step": 5855 }, { "epoch": 0.62, "learning_rate": 3.2620628930319065e-06, "loss": 1.4501, "step": 5856 }, { "epoch": 0.62, "learning_rate": 3.2604432306246424e-06, "loss": 1.0869, "step": 5857 }, { "epoch": 0.62, "learning_rate": 3.258823775863568e-06, "loss": 1.4938, "step": 5858 }, { "epoch": 0.62, "learning_rate": 3.257204528941993e-06, "loss": 1.4338, "step": 5859 }, { "epoch": 0.63, "learning_rate": 3.255585490053201e-06, "loss": 1.3873, "step": 5860 }, { "epoch": 0.63, "learning_rate": 3.2539666593904534e-06, "loss": 1.4179, "step": 5861 }, { "epoch": 0.63, "learning_rate": 3.2523480371469863e-06, "loss": 1.2511, "step": 5862 }, { "epoch": 0.63, "learning_rate": 3.25072962351601e-06, "loss": 1.2411, "step": 5863 }, { "epoch": 0.63, "learning_rate": 3.2491114186907097e-06, "loss": 1.4413, "step": 5864 }, { "epoch": 0.63, "learning_rate": 3.2474934228642475e-06, "loss": 1.5475, "step": 5865 }, { "epoch": 0.63, "learning_rate": 3.245875636229755e-06, "loss": 1.57, "step": 5866 }, { "epoch": 0.63, "learning_rate": 3.2442580589803484e-06, "loss": 1.3503, "step": 5867 }, { "epoch": 0.63, "learning_rate": 3.242640691309111e-06, "loss": 1.2419, "step": 5868 }, { "epoch": 0.63, "learning_rate": 3.241023533409104e-06, "loss": 1.3549, "step": 5869 }, { "epoch": 0.63, "learning_rate": 3.2394065854733626e-06, "loss": 1.4752, "step": 5870 }, { "epoch": 0.63, "learning_rate": 3.2377898476948964e-06, "loss": 1.3029, "step": 5871 }, { "epoch": 0.63, "learning_rate": 3.236173320266694e-06, "loss": 1.5949, "step": 5872 }, { "epoch": 0.63, "learning_rate": 3.2345570033817153e-06, "loss": 1.3758, "step": 5873 }, { "epoch": 0.63, "learning_rate": 3.2329408972328934e-06, "loss": 1.6943, "step": 5874 }, { "epoch": 0.63, "learning_rate": 3.23132500201314e-06, "loss": 1.2783, "step": 5875 }, { "epoch": 0.63, "learning_rate": 3.2297093179153394e-06, "loss": 1.3421, "step": 5876 }, { "epoch": 0.63, "learning_rate": 3.2280938451323524e-06, "loss": 1.5856, "step": 5877 }, { "epoch": 0.63, "learning_rate": 3.2264785838570134e-06, "loss": 1.5301, "step": 5878 }, { "epoch": 0.63, "learning_rate": 3.2248635342821334e-06, "loss": 1.3508, "step": 5879 }, { "epoch": 0.63, "learning_rate": 3.223248696600493e-06, "loss": 1.461, "step": 5880 }, { "epoch": 0.63, "learning_rate": 3.221634071004852e-06, "loss": 1.683, "step": 5881 }, { "epoch": 0.63, "learning_rate": 3.2200196576879463e-06, "loss": 1.5807, "step": 5882 }, { "epoch": 0.63, "learning_rate": 3.2184054568424817e-06, "loss": 1.3945, "step": 5883 }, { "epoch": 0.63, "learning_rate": 3.2167914686611425e-06, "loss": 1.6019, "step": 5884 }, { "epoch": 0.63, "learning_rate": 3.2151776933365842e-06, "loss": 1.3173, "step": 5885 }, { "epoch": 0.63, "learning_rate": 3.2135641310614383e-06, "loss": 1.1385, "step": 5886 }, { "epoch": 0.63, "learning_rate": 3.2119507820283158e-06, "loss": 1.4638, "step": 5887 }, { "epoch": 0.63, "learning_rate": 3.2103376464297942e-06, "loss": 1.4001, "step": 5888 }, { "epoch": 0.63, "learning_rate": 3.20872472445843e-06, "loss": 1.5774, "step": 5889 }, { "epoch": 0.63, "learning_rate": 3.2071120163067527e-06, "loss": 1.6348, "step": 5890 }, { "epoch": 0.63, "learning_rate": 3.2054995221672676e-06, "loss": 1.0648, "step": 5891 }, { "epoch": 0.63, "learning_rate": 3.203887242232455e-06, "loss": 1.3965, "step": 5892 }, { "epoch": 0.63, "learning_rate": 3.2022751766947656e-06, "loss": 1.2144, "step": 5893 }, { "epoch": 0.63, "learning_rate": 3.200663325746629e-06, "loss": 1.2389, "step": 5894 }, { "epoch": 0.63, "learning_rate": 3.1990516895804467e-06, "loss": 1.439, "step": 5895 }, { "epoch": 0.63, "learning_rate": 3.1974402683885963e-06, "loss": 1.4558, "step": 5896 }, { "epoch": 0.63, "learning_rate": 3.1958290623634285e-06, "loss": 1.3816, "step": 5897 }, { "epoch": 0.63, "learning_rate": 3.1942180716972698e-06, "loss": 1.2196, "step": 5898 }, { "epoch": 0.63, "learning_rate": 3.1926072965824172e-06, "loss": 1.5707, "step": 5899 }, { "epoch": 0.63, "learning_rate": 3.190996737211145e-06, "loss": 1.4339, "step": 5900 }, { "epoch": 0.63, "learning_rate": 3.189386393775703e-06, "loss": 1.3232, "step": 5901 }, { "epoch": 0.63, "learning_rate": 3.1877762664683125e-06, "loss": 1.364, "step": 5902 }, { "epoch": 0.63, "learning_rate": 3.1861663554811707e-06, "loss": 1.47, "step": 5903 }, { "epoch": 0.63, "learning_rate": 3.1845566610064487e-06, "loss": 1.8288, "step": 5904 }, { "epoch": 0.63, "learning_rate": 3.182947183236288e-06, "loss": 1.4617, "step": 5905 }, { "epoch": 0.63, "learning_rate": 3.1813379223628127e-06, "loss": 1.4496, "step": 5906 }, { "epoch": 0.63, "learning_rate": 3.179728878578112e-06, "loss": 1.5794, "step": 5907 }, { "epoch": 0.63, "learning_rate": 3.178120052074255e-06, "loss": 1.4723, "step": 5908 }, { "epoch": 0.63, "learning_rate": 3.1765114430432826e-06, "loss": 1.4502, "step": 5909 }, { "epoch": 0.63, "learning_rate": 3.1749030516772084e-06, "loss": 1.4015, "step": 5910 }, { "epoch": 0.63, "learning_rate": 3.173294878168025e-06, "loss": 1.3991, "step": 5911 }, { "epoch": 0.63, "learning_rate": 3.1716869227076946e-06, "loss": 1.2339, "step": 5912 }, { "epoch": 0.63, "learning_rate": 3.170079185488153e-06, "loss": 1.1388, "step": 5913 }, { "epoch": 0.63, "learning_rate": 3.1684716667013125e-06, "loss": 1.3107, "step": 5914 }, { "epoch": 0.63, "learning_rate": 3.166864366539057e-06, "loss": 1.1856, "step": 5915 }, { "epoch": 0.63, "learning_rate": 3.165257285193248e-06, "loss": 1.4234, "step": 5916 }, { "epoch": 0.63, "learning_rate": 3.163650422855717e-06, "loss": 1.2744, "step": 5917 }, { "epoch": 0.63, "learning_rate": 3.1620437797182717e-06, "loss": 1.4196, "step": 5918 }, { "epoch": 0.63, "learning_rate": 3.1604373559726915e-06, "loss": 1.245, "step": 5919 }, { "epoch": 0.63, "learning_rate": 3.15883115181073e-06, "loss": 1.1845, "step": 5920 }, { "epoch": 0.63, "learning_rate": 3.1572251674241182e-06, "loss": 1.3369, "step": 5921 }, { "epoch": 0.63, "learning_rate": 3.1556194030045563e-06, "loss": 1.5411, "step": 5922 }, { "epoch": 0.63, "learning_rate": 3.1540138587437206e-06, "loss": 1.2228, "step": 5923 }, { "epoch": 0.63, "learning_rate": 3.1524085348332622e-06, "loss": 1.3692, "step": 5924 }, { "epoch": 0.63, "learning_rate": 3.1508034314647994e-06, "loss": 1.6178, "step": 5925 }, { "epoch": 0.63, "learning_rate": 3.149198548829935e-06, "loss": 1.4435, "step": 5926 }, { "epoch": 0.63, "learning_rate": 3.1475938871202356e-06, "loss": 1.5101, "step": 5927 }, { "epoch": 0.63, "learning_rate": 3.1459894465272467e-06, "loss": 1.479, "step": 5928 }, { "epoch": 0.63, "learning_rate": 3.1443852272424863e-06, "loss": 1.5445, "step": 5929 }, { "epoch": 0.63, "learning_rate": 3.142781229457442e-06, "loss": 1.3843, "step": 5930 }, { "epoch": 0.63, "learning_rate": 3.1411774533635854e-06, "loss": 1.3092, "step": 5931 }, { "epoch": 0.63, "learning_rate": 3.139573899152351e-06, "loss": 1.4862, "step": 5932 }, { "epoch": 0.63, "learning_rate": 3.1379705670151504e-06, "loss": 1.4656, "step": 5933 }, { "epoch": 0.63, "learning_rate": 3.136367457143369e-06, "loss": 1.2681, "step": 5934 }, { "epoch": 0.63, "learning_rate": 3.134764569728368e-06, "loss": 1.5923, "step": 5935 }, { "epoch": 0.63, "learning_rate": 3.133161904961477e-06, "loss": 1.5493, "step": 5936 }, { "epoch": 0.63, "learning_rate": 3.1315594630340052e-06, "loss": 1.3799, "step": 5937 }, { "epoch": 0.63, "learning_rate": 3.1299572441372274e-06, "loss": 1.3731, "step": 5938 }, { "epoch": 0.63, "learning_rate": 3.128355248462397e-06, "loss": 1.6523, "step": 5939 }, { "epoch": 0.63, "learning_rate": 3.1267534762007435e-06, "loss": 1.4239, "step": 5940 }, { "epoch": 0.63, "learning_rate": 3.125151927543464e-06, "loss": 1.4888, "step": 5941 }, { "epoch": 0.63, "learning_rate": 3.1235506026817296e-06, "loss": 1.32, "step": 5942 }, { "epoch": 0.63, "learning_rate": 3.1219495018066888e-06, "loss": 1.4443, "step": 5943 }, { "epoch": 0.63, "learning_rate": 3.1203486251094562e-06, "loss": 1.5666, "step": 5944 }, { "epoch": 0.63, "learning_rate": 3.11874797278113e-06, "loss": 1.33, "step": 5945 }, { "epoch": 0.63, "learning_rate": 3.1171475450127717e-06, "loss": 1.3803, "step": 5946 }, { "epoch": 0.63, "learning_rate": 3.115547341995422e-06, "loss": 1.5224, "step": 5947 }, { "epoch": 0.63, "learning_rate": 3.113947363920091e-06, "loss": 1.3989, "step": 5948 }, { "epoch": 0.63, "learning_rate": 3.112347610977764e-06, "loss": 1.6458, "step": 5949 }, { "epoch": 0.63, "learning_rate": 3.1107480833594018e-06, "loss": 1.3634, "step": 5950 }, { "epoch": 0.63, "learning_rate": 3.109148781255934e-06, "loss": 1.4834, "step": 5951 }, { "epoch": 0.63, "learning_rate": 3.1075497048582635e-06, "loss": 1.0962, "step": 5952 }, { "epoch": 0.63, "learning_rate": 3.1059508543572703e-06, "loss": 1.597, "step": 5953 }, { "epoch": 0.64, "learning_rate": 3.1043522299438023e-06, "loss": 1.273, "step": 5954 }, { "epoch": 0.64, "learning_rate": 3.102753831808685e-06, "loss": 1.4028, "step": 5955 }, { "epoch": 0.64, "learning_rate": 3.1011556601427138e-06, "loss": 1.6193, "step": 5956 }, { "epoch": 0.64, "learning_rate": 3.0995577151366594e-06, "loss": 1.4058, "step": 5957 }, { "epoch": 0.64, "learning_rate": 3.097959996981263e-06, "loss": 1.5356, "step": 5958 }, { "epoch": 0.64, "learning_rate": 3.0963625058672384e-06, "loss": 1.5038, "step": 5959 }, { "epoch": 0.64, "learning_rate": 3.0947652419852768e-06, "loss": 1.5104, "step": 5960 }, { "epoch": 0.64, "learning_rate": 3.093168205526038e-06, "loss": 1.6407, "step": 5961 }, { "epoch": 0.64, "learning_rate": 3.091571396680156e-06, "loss": 1.2973, "step": 5962 }, { "epoch": 0.64, "learning_rate": 3.089974815638239e-06, "loss": 1.3983, "step": 5963 }, { "epoch": 0.64, "learning_rate": 3.0883784625908618e-06, "loss": 1.3202, "step": 5964 }, { "epoch": 0.64, "learning_rate": 3.0867823377285835e-06, "loss": 1.1724, "step": 5965 }, { "epoch": 0.64, "learning_rate": 3.0851864412419236e-06, "loss": 1.3393, "step": 5966 }, { "epoch": 0.64, "learning_rate": 3.083590773321383e-06, "loss": 1.3064, "step": 5967 }, { "epoch": 0.64, "learning_rate": 3.0819953341574305e-06, "loss": 1.2562, "step": 5968 }, { "epoch": 0.64, "learning_rate": 3.0804001239405097e-06, "loss": 1.3783, "step": 5969 }, { "epoch": 0.64, "learning_rate": 3.0788051428610377e-06, "loss": 1.3434, "step": 5970 }, { "epoch": 0.64, "learning_rate": 3.077210391109404e-06, "loss": 1.4595, "step": 5971 }, { "epoch": 0.64, "learning_rate": 3.075615868875967e-06, "loss": 1.2923, "step": 5972 }, { "epoch": 0.64, "learning_rate": 3.0740215763510617e-06, "loss": 1.4064, "step": 5973 }, { "epoch": 0.64, "learning_rate": 3.072427513724993e-06, "loss": 1.3834, "step": 5974 }, { "epoch": 0.64, "learning_rate": 3.070833681188043e-06, "loss": 1.4058, "step": 5975 }, { "epoch": 0.64, "learning_rate": 3.069240078930461e-06, "loss": 1.4241, "step": 5976 }, { "epoch": 0.64, "learning_rate": 3.0676467071424733e-06, "loss": 1.4622, "step": 5977 }, { "epoch": 0.64, "learning_rate": 3.066053566014271e-06, "loss": 1.3509, "step": 5978 }, { "epoch": 0.64, "learning_rate": 3.0644606557360303e-06, "loss": 1.2912, "step": 5979 }, { "epoch": 0.64, "learning_rate": 3.0628679764978875e-06, "loss": 1.2291, "step": 5980 }, { "epoch": 0.64, "learning_rate": 3.0612755284899577e-06, "loss": 1.3937, "step": 5981 }, { "epoch": 0.64, "learning_rate": 3.0596833119023283e-06, "loss": 1.21, "step": 5982 }, { "epoch": 0.64, "learning_rate": 3.0580913269250545e-06, "loss": 1.4875, "step": 5983 }, { "epoch": 0.64, "learning_rate": 3.0564995737481706e-06, "loss": 1.463, "step": 5984 }, { "epoch": 0.64, "learning_rate": 3.054908052561681e-06, "loss": 1.503, "step": 5985 }, { "epoch": 0.64, "learning_rate": 3.053316763555557e-06, "loss": 1.2752, "step": 5986 }, { "epoch": 0.64, "learning_rate": 3.0517257069197497e-06, "loss": 1.4444, "step": 5987 }, { "epoch": 0.64, "learning_rate": 3.0501348828441767e-06, "loss": 1.4892, "step": 5988 }, { "epoch": 0.64, "learning_rate": 3.0485442915187335e-06, "loss": 1.2733, "step": 5989 }, { "epoch": 0.64, "learning_rate": 3.046953933133283e-06, "loss": 1.3427, "step": 5990 }, { "epoch": 0.64, "learning_rate": 3.0453638078776614e-06, "loss": 1.1792, "step": 5991 }, { "epoch": 0.64, "learning_rate": 3.043773915941679e-06, "loss": 1.6177, "step": 5992 }, { "epoch": 0.64, "learning_rate": 3.042184257515114e-06, "loss": 1.5185, "step": 5993 }, { "epoch": 0.64, "learning_rate": 3.0405948327877233e-06, "loss": 1.6378, "step": 5994 }, { "epoch": 0.64, "learning_rate": 3.039005641949231e-06, "loss": 1.176, "step": 5995 }, { "epoch": 0.64, "learning_rate": 3.037416685189335e-06, "loss": 1.3812, "step": 5996 }, { "epoch": 0.64, "learning_rate": 3.0358279626977034e-06, "loss": 1.479, "step": 5997 }, { "epoch": 0.64, "learning_rate": 3.0342394746639768e-06, "loss": 1.5671, "step": 5998 }, { "epoch": 0.64, "learning_rate": 3.0326512212777715e-06, "loss": 1.3086, "step": 5999 }, { "epoch": 0.64, "learning_rate": 3.0310632027286717e-06, "loss": 1.4659, "step": 6000 }, { "epoch": 0.64, "learning_rate": 3.0294754192062346e-06, "loss": 1.44, "step": 6001 }, { "epoch": 0.64, "learning_rate": 3.0278878708999917e-06, "loss": 1.5564, "step": 6002 }, { "epoch": 0.64, "learning_rate": 3.026300557999439e-06, "loss": 1.347, "step": 6003 }, { "epoch": 0.64, "learning_rate": 3.024713480694057e-06, "loss": 1.488, "step": 6004 }, { "epoch": 0.64, "learning_rate": 3.023126639173286e-06, "loss": 1.3107, "step": 6005 }, { "epoch": 0.64, "learning_rate": 3.021540033626544e-06, "loss": 1.1935, "step": 6006 }, { "epoch": 0.64, "learning_rate": 3.01995366424322e-06, "loss": 1.4111, "step": 6007 }, { "epoch": 0.64, "learning_rate": 3.0183675312126737e-06, "loss": 1.3419, "step": 6008 }, { "epoch": 0.64, "learning_rate": 3.0167816347242396e-06, "loss": 1.4804, "step": 6009 }, { "epoch": 0.64, "learning_rate": 3.015195974967221e-06, "loss": 1.6667, "step": 6010 }, { "epoch": 0.64, "learning_rate": 3.0136105521308933e-06, "loss": 1.5435, "step": 6011 }, { "epoch": 0.64, "learning_rate": 3.012025366404504e-06, "loss": 1.2359, "step": 6012 }, { "epoch": 0.64, "learning_rate": 3.010440417977272e-06, "loss": 1.4147, "step": 6013 }, { "epoch": 0.64, "learning_rate": 3.00885570703839e-06, "loss": 1.5015, "step": 6014 }, { "epoch": 0.64, "learning_rate": 3.00727123377702e-06, "loss": 1.2556, "step": 6015 }, { "epoch": 0.64, "learning_rate": 3.005686998382297e-06, "loss": 1.5453, "step": 6016 }, { "epoch": 0.64, "learning_rate": 3.004103001043325e-06, "loss": 1.2501, "step": 6017 }, { "epoch": 0.64, "learning_rate": 3.002519241949181e-06, "loss": 1.3047, "step": 6018 }, { "epoch": 0.64, "learning_rate": 3.000935721288917e-06, "loss": 1.3653, "step": 6019 }, { "epoch": 0.64, "learning_rate": 2.999352439251552e-06, "loss": 1.4269, "step": 6020 }, { "epoch": 0.64, "learning_rate": 2.997769396026078e-06, "loss": 1.4977, "step": 6021 }, { "epoch": 0.64, "learning_rate": 2.9961865918014575e-06, "loss": 1.3352, "step": 6022 }, { "epoch": 0.64, "learning_rate": 2.9946040267666276e-06, "loss": 1.5285, "step": 6023 }, { "epoch": 0.64, "learning_rate": 2.9930217011104957e-06, "loss": 1.6037, "step": 6024 }, { "epoch": 0.64, "learning_rate": 2.991439615021937e-06, "loss": 1.3066, "step": 6025 }, { "epoch": 0.64, "learning_rate": 2.989857768689802e-06, "loss": 1.5107, "step": 6026 }, { "epoch": 0.64, "learning_rate": 2.98827616230291e-06, "loss": 1.2925, "step": 6027 }, { "epoch": 0.64, "learning_rate": 2.9866947960500557e-06, "loss": 1.4288, "step": 6028 }, { "epoch": 0.64, "learning_rate": 2.985113670120001e-06, "loss": 1.3073, "step": 6029 }, { "epoch": 0.64, "learning_rate": 2.9835327847014816e-06, "loss": 1.4484, "step": 6030 }, { "epoch": 0.64, "learning_rate": 2.9819521399832017e-06, "loss": 1.6033, "step": 6031 }, { "epoch": 0.64, "learning_rate": 2.9803717361538397e-06, "loss": 1.1826, "step": 6032 }, { "epoch": 0.64, "learning_rate": 2.9787915734020446e-06, "loss": 1.4788, "step": 6033 }, { "epoch": 0.64, "learning_rate": 2.977211651916435e-06, "loss": 1.2318, "step": 6034 }, { "epoch": 0.64, "learning_rate": 2.9756319718856043e-06, "loss": 1.5451, "step": 6035 }, { "epoch": 0.64, "learning_rate": 2.9740525334981105e-06, "loss": 1.2955, "step": 6036 }, { "epoch": 0.64, "learning_rate": 2.9724733369424887e-06, "loss": 1.4445, "step": 6037 }, { "epoch": 0.64, "learning_rate": 2.9708943824072466e-06, "loss": 1.4691, "step": 6038 }, { "epoch": 0.64, "learning_rate": 2.9693156700808556e-06, "loss": 1.3045, "step": 6039 }, { "epoch": 0.64, "learning_rate": 2.9677372001517636e-06, "loss": 1.4261, "step": 6040 }, { "epoch": 0.64, "learning_rate": 2.9661589728083895e-06, "loss": 1.4451, "step": 6041 }, { "epoch": 0.64, "learning_rate": 2.9645809882391187e-06, "loss": 1.4007, "step": 6042 }, { "epoch": 0.64, "learning_rate": 2.963003246632315e-06, "loss": 1.3062, "step": 6043 }, { "epoch": 0.64, "learning_rate": 2.9614257481763065e-06, "loss": 1.267, "step": 6044 }, { "epoch": 0.64, "learning_rate": 2.959848493059396e-06, "loss": 1.5155, "step": 6045 }, { "epoch": 0.64, "learning_rate": 2.9582714814698567e-06, "loss": 1.322, "step": 6046 }, { "epoch": 0.65, "learning_rate": 2.9566947135959294e-06, "loss": 1.1895, "step": 6047 }, { "epoch": 0.65, "learning_rate": 2.9551181896258317e-06, "loss": 1.5018, "step": 6048 }, { "epoch": 0.65, "learning_rate": 2.9535419097477497e-06, "loss": 1.3809, "step": 6049 }, { "epoch": 0.65, "learning_rate": 2.951965874149837e-06, "loss": 1.4867, "step": 6050 }, { "epoch": 0.65, "learning_rate": 2.9503900830202202e-06, "loss": 1.2926, "step": 6051 }, { "epoch": 0.65, "learning_rate": 2.948814536546999e-06, "loss": 1.471, "step": 6052 }, { "epoch": 0.65, "learning_rate": 2.9472392349182427e-06, "loss": 1.4381, "step": 6053 }, { "epoch": 0.65, "learning_rate": 2.9456641783219897e-06, "loss": 1.29, "step": 6054 }, { "epoch": 0.65, "learning_rate": 2.944089366946251e-06, "loss": 1.4662, "step": 6055 }, { "epoch": 0.65, "learning_rate": 2.9425148009790067e-06, "loss": 1.3966, "step": 6056 }, { "epoch": 0.65, "learning_rate": 2.9409404806082077e-06, "loss": 1.2915, "step": 6057 }, { "epoch": 0.65, "learning_rate": 2.939366406021778e-06, "loss": 1.342, "step": 6058 }, { "epoch": 0.65, "learning_rate": 2.9377925774076106e-06, "loss": 1.6312, "step": 6059 }, { "epoch": 0.65, "learning_rate": 2.936218994953568e-06, "loss": 1.2714, "step": 6060 }, { "epoch": 0.65, "learning_rate": 2.934645658847487e-06, "loss": 1.3904, "step": 6061 }, { "epoch": 0.65, "learning_rate": 2.933072569277167e-06, "loss": 1.2887, "step": 6062 }, { "epoch": 0.65, "learning_rate": 2.93149972643039e-06, "loss": 1.3267, "step": 6063 }, { "epoch": 0.65, "learning_rate": 2.9299271304948985e-06, "loss": 1.5283, "step": 6064 }, { "epoch": 0.65, "learning_rate": 2.9283547816584097e-06, "loss": 1.602, "step": 6065 }, { "epoch": 0.65, "learning_rate": 2.9267826801086103e-06, "loss": 1.2633, "step": 6066 }, { "epoch": 0.65, "learning_rate": 2.9252108260331567e-06, "loss": 1.4376, "step": 6067 }, { "epoch": 0.65, "learning_rate": 2.9236392196196795e-06, "loss": 1.4664, "step": 6068 }, { "epoch": 0.65, "learning_rate": 2.9220678610557773e-06, "loss": 1.2178, "step": 6069 }, { "epoch": 0.65, "learning_rate": 2.9204967505290162e-06, "loss": 1.143, "step": 6070 }, { "epoch": 0.65, "learning_rate": 2.918925888226935e-06, "loss": 1.2956, "step": 6071 }, { "epoch": 0.65, "learning_rate": 2.9173552743370454e-06, "loss": 1.4229, "step": 6072 }, { "epoch": 0.65, "learning_rate": 2.9157849090468294e-06, "loss": 1.4732, "step": 6073 }, { "epoch": 0.65, "learning_rate": 2.9142147925437336e-06, "loss": 1.1987, "step": 6074 }, { "epoch": 0.65, "learning_rate": 2.912644925015179e-06, "loss": 1.4024, "step": 6075 }, { "epoch": 0.65, "learning_rate": 2.9110753066485566e-06, "loss": 1.4198, "step": 6076 }, { "epoch": 0.65, "learning_rate": 2.9095059376312294e-06, "loss": 1.5651, "step": 6077 }, { "epoch": 0.65, "learning_rate": 2.9079368181505263e-06, "loss": 1.6197, "step": 6078 }, { "epoch": 0.65, "learning_rate": 2.9063679483937513e-06, "loss": 1.4589, "step": 6079 }, { "epoch": 0.65, "learning_rate": 2.904799328548175e-06, "loss": 1.3861, "step": 6080 }, { "epoch": 0.65, "learning_rate": 2.9032309588010372e-06, "loss": 1.4555, "step": 6081 }, { "epoch": 0.65, "learning_rate": 2.901662839339552e-06, "loss": 1.5192, "step": 6082 }, { "epoch": 0.65, "learning_rate": 2.900094970350903e-06, "loss": 1.5707, "step": 6083 }, { "epoch": 0.65, "learning_rate": 2.8985273520222414e-06, "loss": 1.3307, "step": 6084 }, { "epoch": 0.65, "learning_rate": 2.896959984540687e-06, "loss": 1.5292, "step": 6085 }, { "epoch": 0.65, "learning_rate": 2.895392868093334e-06, "loss": 1.4868, "step": 6086 }, { "epoch": 0.65, "learning_rate": 2.893826002867247e-06, "loss": 1.5461, "step": 6087 }, { "epoch": 0.65, "learning_rate": 2.892259389049455e-06, "loss": 1.5998, "step": 6088 }, { "epoch": 0.65, "learning_rate": 2.890693026826964e-06, "loss": 1.4589, "step": 6089 }, { "epoch": 0.65, "learning_rate": 2.889126916386744e-06, "loss": 1.5698, "step": 6090 }, { "epoch": 0.65, "learning_rate": 2.8875610579157347e-06, "loss": 1.1952, "step": 6091 }, { "epoch": 0.65, "learning_rate": 2.885995451600855e-06, "loss": 1.2907, "step": 6092 }, { "epoch": 0.65, "learning_rate": 2.884430097628984e-06, "loss": 1.1765, "step": 6093 }, { "epoch": 0.65, "learning_rate": 2.8828649961869716e-06, "loss": 1.5842, "step": 6094 }, { "epoch": 0.65, "learning_rate": 2.881300147461643e-06, "loss": 1.2251, "step": 6095 }, { "epoch": 0.65, "learning_rate": 2.879735551639787e-06, "loss": 1.2204, "step": 6096 }, { "epoch": 0.65, "learning_rate": 2.878171208908166e-06, "loss": 1.3905, "step": 6097 }, { "epoch": 0.65, "learning_rate": 2.876607119453515e-06, "loss": 1.606, "step": 6098 }, { "epoch": 0.65, "learning_rate": 2.8750432834625312e-06, "loss": 1.4476, "step": 6099 }, { "epoch": 0.65, "learning_rate": 2.873479701121885e-06, "loss": 1.5213, "step": 6100 }, { "epoch": 0.65, "learning_rate": 2.8719163726182185e-06, "loss": 1.3047, "step": 6101 }, { "epoch": 0.65, "learning_rate": 2.8703532981381437e-06, "loss": 1.3939, "step": 6102 }, { "epoch": 0.65, "learning_rate": 2.868790477868237e-06, "loss": 1.7078, "step": 6103 }, { "epoch": 0.65, "learning_rate": 2.867227911995052e-06, "loss": 1.4392, "step": 6104 }, { "epoch": 0.65, "learning_rate": 2.8656656007051055e-06, "loss": 1.4458, "step": 6105 }, { "epoch": 0.65, "learning_rate": 2.864103544184885e-06, "loss": 1.2032, "step": 6106 }, { "epoch": 0.65, "learning_rate": 2.8625417426208503e-06, "loss": 1.4414, "step": 6107 }, { "epoch": 0.65, "learning_rate": 2.860980196199432e-06, "loss": 1.2783, "step": 6108 }, { "epoch": 0.65, "learning_rate": 2.859418905107023e-06, "loss": 1.3584, "step": 6109 }, { "epoch": 0.65, "learning_rate": 2.857857869529996e-06, "loss": 1.4742, "step": 6110 }, { "epoch": 0.65, "learning_rate": 2.8562970896546815e-06, "loss": 1.5867, "step": 6111 }, { "epoch": 0.65, "learning_rate": 2.8547365656673907e-06, "loss": 1.296, "step": 6112 }, { "epoch": 0.65, "learning_rate": 2.8531762977543954e-06, "loss": 1.5484, "step": 6113 }, { "epoch": 0.65, "learning_rate": 2.8516162861019437e-06, "loss": 1.4276, "step": 6114 }, { "epoch": 0.65, "learning_rate": 2.850056530896247e-06, "loss": 1.2485, "step": 6115 }, { "epoch": 0.65, "learning_rate": 2.8484970323234897e-06, "loss": 1.386, "step": 6116 }, { "epoch": 0.65, "learning_rate": 2.846937790569828e-06, "loss": 1.1785, "step": 6117 }, { "epoch": 0.65, "learning_rate": 2.8453788058213814e-06, "loss": 1.5184, "step": 6118 }, { "epoch": 0.65, "learning_rate": 2.8438200782642416e-06, "loss": 1.527, "step": 6119 }, { "epoch": 0.65, "learning_rate": 2.84226160808447e-06, "loss": 1.523, "step": 6120 }, { "epoch": 0.65, "learning_rate": 2.8407033954680995e-06, "loss": 1.6614, "step": 6121 }, { "epoch": 0.65, "learning_rate": 2.8391454406011255e-06, "loss": 1.403, "step": 6122 }, { "epoch": 0.65, "learning_rate": 2.837587743669521e-06, "loss": 1.3392, "step": 6123 }, { "epoch": 0.65, "learning_rate": 2.836030304859222e-06, "loss": 1.437, "step": 6124 }, { "epoch": 0.65, "learning_rate": 2.8344731243561317e-06, "loss": 1.465, "step": 6125 }, { "epoch": 0.65, "learning_rate": 2.8329162023461355e-06, "loss": 1.3579, "step": 6126 }, { "epoch": 0.65, "learning_rate": 2.831359539015073e-06, "loss": 1.3651, "step": 6127 }, { "epoch": 0.65, "learning_rate": 2.8298031345487585e-06, "loss": 1.288, "step": 6128 }, { "epoch": 0.65, "learning_rate": 2.82824698913298e-06, "loss": 1.4809, "step": 6129 }, { "epoch": 0.65, "learning_rate": 2.8266911029534848e-06, "loss": 1.6119, "step": 6130 }, { "epoch": 0.65, "learning_rate": 2.825135476195998e-06, "loss": 1.3221, "step": 6131 }, { "epoch": 0.65, "learning_rate": 2.823580109046212e-06, "loss": 1.4832, "step": 6132 }, { "epoch": 0.65, "learning_rate": 2.822025001689786e-06, "loss": 1.5466, "step": 6133 }, { "epoch": 0.65, "learning_rate": 2.820470154312346e-06, "loss": 1.3197, "step": 6134 }, { "epoch": 0.65, "learning_rate": 2.8189155670994913e-06, "loss": 1.4173, "step": 6135 }, { "epoch": 0.65, "learning_rate": 2.817361240236791e-06, "loss": 1.476, "step": 6136 }, { "epoch": 0.65, "learning_rate": 2.81580717390978e-06, "loss": 1.4733, "step": 6137 }, { "epoch": 0.65, "learning_rate": 2.814253368303961e-06, "loss": 1.1697, "step": 6138 }, { "epoch": 0.65, "learning_rate": 2.8126998236048096e-06, "loss": 1.564, "step": 6139 }, { "epoch": 0.65, "learning_rate": 2.8111465399977667e-06, "loss": 1.4547, "step": 6140 }, { "epoch": 0.66, "learning_rate": 2.809593517668243e-06, "loss": 1.555, "step": 6141 }, { "epoch": 0.66, "learning_rate": 2.808040756801623e-06, "loss": 1.6143, "step": 6142 }, { "epoch": 0.66, "learning_rate": 2.806488257583253e-06, "loss": 1.617, "step": 6143 }, { "epoch": 0.66, "learning_rate": 2.804936020198447e-06, "loss": 1.229, "step": 6144 }, { "epoch": 0.66, "learning_rate": 2.8033840448324956e-06, "loss": 1.5983, "step": 6145 }, { "epoch": 0.66, "learning_rate": 2.801832331670654e-06, "loss": 1.6584, "step": 6146 }, { "epoch": 0.66, "learning_rate": 2.800280880898143e-06, "loss": 1.2391, "step": 6147 }, { "epoch": 0.66, "learning_rate": 2.7987296927001597e-06, "loss": 1.5608, "step": 6148 }, { "epoch": 0.66, "learning_rate": 2.7971787672618618e-06, "loss": 1.1942, "step": 6149 }, { "epoch": 0.66, "learning_rate": 2.795628104768376e-06, "loss": 1.445, "step": 6150 }, { "epoch": 0.66, "learning_rate": 2.7940777054048085e-06, "loss": 1.4735, "step": 6151 }, { "epoch": 0.66, "learning_rate": 2.7925275693562213e-06, "loss": 1.3812, "step": 6152 }, { "epoch": 0.66, "learning_rate": 2.79097769680765e-06, "loss": 1.3625, "step": 6153 }, { "epoch": 0.66, "learning_rate": 2.7894280879441015e-06, "loss": 1.0851, "step": 6154 }, { "epoch": 0.66, "learning_rate": 2.7878787429505444e-06, "loss": 1.4357, "step": 6155 }, { "epoch": 0.66, "learning_rate": 2.7863296620119217e-06, "loss": 1.5077, "step": 6156 }, { "epoch": 0.66, "learning_rate": 2.7847808453131454e-06, "loss": 1.4923, "step": 6157 }, { "epoch": 0.66, "learning_rate": 2.7832322930390914e-06, "loss": 1.5114, "step": 6158 }, { "epoch": 0.66, "learning_rate": 2.781684005374604e-06, "loss": 1.1839, "step": 6159 }, { "epoch": 0.66, "learning_rate": 2.780135982504501e-06, "loss": 1.2165, "step": 6160 }, { "epoch": 0.66, "learning_rate": 2.7785882246135665e-06, "loss": 1.2808, "step": 6161 }, { "epoch": 0.66, "learning_rate": 2.777040731886549e-06, "loss": 1.5274, "step": 6162 }, { "epoch": 0.66, "learning_rate": 2.7754935045081716e-06, "loss": 1.397, "step": 6163 }, { "epoch": 0.66, "learning_rate": 2.7739465426631195e-06, "loss": 1.5534, "step": 6164 }, { "epoch": 0.66, "learning_rate": 2.7723998465360537e-06, "loss": 1.3934, "step": 6165 }, { "epoch": 0.66, "learning_rate": 2.7708534163115942e-06, "loss": 1.4733, "step": 6166 }, { "epoch": 0.66, "learning_rate": 2.769307252174338e-06, "loss": 1.6524, "step": 6167 }, { "epoch": 0.66, "learning_rate": 2.7677613543088432e-06, "loss": 1.3727, "step": 6168 }, { "epoch": 0.66, "learning_rate": 2.766215722899642e-06, "loss": 1.3415, "step": 6169 }, { "epoch": 0.66, "learning_rate": 2.7646703581312336e-06, "loss": 1.6332, "step": 6170 }, { "epoch": 0.66, "learning_rate": 2.7631252601880816e-06, "loss": 1.5333, "step": 6171 }, { "epoch": 0.66, "learning_rate": 2.7615804292546185e-06, "loss": 1.339, "step": 6172 }, { "epoch": 0.66, "learning_rate": 2.7600358655152504e-06, "loss": 1.4145, "step": 6173 }, { "epoch": 0.66, "learning_rate": 2.7584915691543444e-06, "loss": 1.4143, "step": 6174 }, { "epoch": 0.66, "learning_rate": 2.7569475403562408e-06, "loss": 1.4355, "step": 6175 }, { "epoch": 0.66, "learning_rate": 2.7554037793052476e-06, "loss": 1.5082, "step": 6176 }, { "epoch": 0.66, "learning_rate": 2.753860286185637e-06, "loss": 1.4731, "step": 6177 }, { "epoch": 0.66, "learning_rate": 2.7523170611816507e-06, "loss": 1.447, "step": 6178 }, { "epoch": 0.66, "learning_rate": 2.7507741044775018e-06, "loss": 1.3271, "step": 6179 }, { "epoch": 0.66, "learning_rate": 2.7492314162573687e-06, "loss": 1.6248, "step": 6180 }, { "epoch": 0.66, "learning_rate": 2.7476889967053953e-06, "loss": 1.1986, "step": 6181 }, { "epoch": 0.66, "learning_rate": 2.7461468460057007e-06, "loss": 1.4171, "step": 6182 }, { "epoch": 0.66, "learning_rate": 2.744604964342364e-06, "loss": 1.4173, "step": 6183 }, { "epoch": 0.66, "learning_rate": 2.7430633518994314e-06, "loss": 1.1999, "step": 6184 }, { "epoch": 0.66, "learning_rate": 2.74152200886093e-06, "loss": 1.547, "step": 6185 }, { "epoch": 0.66, "learning_rate": 2.7399809354108415e-06, "loss": 1.4661, "step": 6186 }, { "epoch": 0.66, "learning_rate": 2.738440131733117e-06, "loss": 1.3435, "step": 6187 }, { "epoch": 0.66, "learning_rate": 2.736899598011683e-06, "loss": 1.3153, "step": 6188 }, { "epoch": 0.66, "learning_rate": 2.735359334430424e-06, "loss": 1.5554, "step": 6189 }, { "epoch": 0.66, "learning_rate": 2.733819341173202e-06, "loss": 1.52, "step": 6190 }, { "epoch": 0.66, "learning_rate": 2.7322796184238363e-06, "loss": 1.4267, "step": 6191 }, { "epoch": 0.66, "learning_rate": 2.7307401663661247e-06, "loss": 1.3618, "step": 6192 }, { "epoch": 0.66, "learning_rate": 2.7292009851838237e-06, "loss": 1.3611, "step": 6193 }, { "epoch": 0.66, "learning_rate": 2.727662075060662e-06, "loss": 1.384, "step": 6194 }, { "epoch": 0.66, "learning_rate": 2.7261234361803383e-06, "loss": 1.2147, "step": 6195 }, { "epoch": 0.66, "learning_rate": 2.724585068726513e-06, "loss": 1.1312, "step": 6196 }, { "epoch": 0.66, "learning_rate": 2.7230469728828156e-06, "loss": 1.4203, "step": 6197 }, { "epoch": 0.66, "learning_rate": 2.721509148832847e-06, "loss": 1.3657, "step": 6198 }, { "epoch": 0.66, "learning_rate": 2.7199715967601715e-06, "loss": 1.2431, "step": 6199 }, { "epoch": 0.66, "learning_rate": 2.7184343168483218e-06, "loss": 1.5129, "step": 6200 }, { "epoch": 0.66, "learning_rate": 2.7168973092808025e-06, "loss": 1.4619, "step": 6201 }, { "epoch": 0.66, "learning_rate": 2.7153605742410806e-06, "loss": 1.4389, "step": 6202 }, { "epoch": 0.66, "learning_rate": 2.713824111912586e-06, "loss": 1.3542, "step": 6203 }, { "epoch": 0.66, "learning_rate": 2.7122879224787315e-06, "loss": 1.2852, "step": 6204 }, { "epoch": 0.66, "learning_rate": 2.710752006122884e-06, "loss": 1.4544, "step": 6205 }, { "epoch": 0.66, "learning_rate": 2.7092163630283786e-06, "loss": 1.3487, "step": 6206 }, { "epoch": 0.66, "learning_rate": 2.7076809933785254e-06, "loss": 1.3093, "step": 6207 }, { "epoch": 0.66, "learning_rate": 2.7061458973565934e-06, "loss": 1.6312, "step": 6208 }, { "epoch": 0.66, "learning_rate": 2.7046110751458256e-06, "loss": 1.6978, "step": 6209 }, { "epoch": 0.66, "learning_rate": 2.70307652692943e-06, "loss": 1.8352, "step": 6210 }, { "epoch": 0.66, "learning_rate": 2.701542252890581e-06, "loss": 1.4998, "step": 6211 }, { "epoch": 0.66, "learning_rate": 2.7000082532124182e-06, "loss": 1.2836, "step": 6212 }, { "epoch": 0.66, "learning_rate": 2.6984745280780524e-06, "loss": 1.4445, "step": 6213 }, { "epoch": 0.66, "learning_rate": 2.6969410776705627e-06, "loss": 1.4981, "step": 6214 }, { "epoch": 0.66, "learning_rate": 2.695407902172989e-06, "loss": 1.3661, "step": 6215 }, { "epoch": 0.66, "learning_rate": 2.6938750017683457e-06, "loss": 1.4652, "step": 6216 }, { "epoch": 0.66, "learning_rate": 2.6923423766396107e-06, "loss": 1.1462, "step": 6217 }, { "epoch": 0.66, "learning_rate": 2.690810026969725e-06, "loss": 1.0379, "step": 6218 }, { "epoch": 0.66, "learning_rate": 2.6892779529416045e-06, "loss": 1.4247, "step": 6219 }, { "epoch": 0.66, "learning_rate": 2.687746154738131e-06, "loss": 1.1764, "step": 6220 }, { "epoch": 0.66, "learning_rate": 2.6862146325421455e-06, "loss": 1.2928, "step": 6221 }, { "epoch": 0.66, "learning_rate": 2.6846833865364674e-06, "loss": 1.4318, "step": 6222 }, { "epoch": 0.66, "learning_rate": 2.6831524169038724e-06, "loss": 1.5609, "step": 6223 }, { "epoch": 0.66, "learning_rate": 2.6816217238271124e-06, "loss": 1.2963, "step": 6224 }, { "epoch": 0.66, "learning_rate": 2.6800913074888984e-06, "loss": 1.2361, "step": 6225 }, { "epoch": 0.66, "learning_rate": 2.678561168071916e-06, "loss": 1.478, "step": 6226 }, { "epoch": 0.66, "learning_rate": 2.6770313057588114e-06, "loss": 1.4052, "step": 6227 }, { "epoch": 0.66, "learning_rate": 2.6755017207321964e-06, "loss": 1.4045, "step": 6228 }, { "epoch": 0.66, "learning_rate": 2.6739724131746615e-06, "loss": 1.5079, "step": 6229 }, { "epoch": 0.66, "learning_rate": 2.6724433832687512e-06, "loss": 1.2363, "step": 6230 }, { "epoch": 0.66, "learning_rate": 2.6709146311969813e-06, "loss": 1.6058, "step": 6231 }, { "epoch": 0.66, "learning_rate": 2.6693861571418372e-06, "loss": 1.5405, "step": 6232 }, { "epoch": 0.66, "learning_rate": 2.667857961285766e-06, "loss": 1.6756, "step": 6233 }, { "epoch": 0.66, "learning_rate": 2.666330043811185e-06, "loss": 1.3829, "step": 6234 }, { "epoch": 0.67, "learning_rate": 2.6648024049004804e-06, "loss": 1.2259, "step": 6235 }, { "epoch": 0.67, "learning_rate": 2.663275044736001e-06, "loss": 1.5668, "step": 6236 }, { "epoch": 0.67, "learning_rate": 2.66174796350006e-06, "loss": 1.399, "step": 6237 }, { "epoch": 0.67, "learning_rate": 2.6602211613749446e-06, "loss": 1.2244, "step": 6238 }, { "epoch": 0.67, "learning_rate": 2.6586946385429056e-06, "loss": 1.4345, "step": 6239 }, { "epoch": 0.67, "learning_rate": 2.657168395186157e-06, "loss": 1.6643, "step": 6240 }, { "epoch": 0.67, "learning_rate": 2.6556424314868854e-06, "loss": 1.3654, "step": 6241 }, { "epoch": 0.67, "learning_rate": 2.65411674762724e-06, "loss": 1.122, "step": 6242 }, { "epoch": 0.67, "learning_rate": 2.6525913437893346e-06, "loss": 1.2795, "step": 6243 }, { "epoch": 0.67, "learning_rate": 2.6510662201552554e-06, "loss": 1.1871, "step": 6244 }, { "epoch": 0.67, "learning_rate": 2.6495413769070537e-06, "loss": 1.0742, "step": 6245 }, { "epoch": 0.67, "learning_rate": 2.648016814226742e-06, "loss": 1.2453, "step": 6246 }, { "epoch": 0.67, "learning_rate": 2.6464925322963053e-06, "loss": 1.5762, "step": 6247 }, { "epoch": 0.67, "learning_rate": 2.6449685312976948e-06, "loss": 1.42, "step": 6248 }, { "epoch": 0.67, "learning_rate": 2.6434448114128252e-06, "loss": 1.5985, "step": 6249 }, { "epoch": 0.67, "learning_rate": 2.641921372823576e-06, "loss": 1.626, "step": 6250 }, { "epoch": 0.67, "learning_rate": 2.6403982157117996e-06, "loss": 1.301, "step": 6251 }, { "epoch": 0.67, "learning_rate": 2.6388753402593083e-06, "loss": 1.5671, "step": 6252 }, { "epoch": 0.67, "learning_rate": 2.6373527466478843e-06, "loss": 1.5712, "step": 6253 }, { "epoch": 0.67, "learning_rate": 2.635830435059279e-06, "loss": 1.6128, "step": 6254 }, { "epoch": 0.67, "learning_rate": 2.6343084056752032e-06, "loss": 1.6289, "step": 6255 }, { "epoch": 0.67, "learning_rate": 2.6327866586773364e-06, "loss": 1.2619, "step": 6256 }, { "epoch": 0.67, "learning_rate": 2.631265194247327e-06, "loss": 1.5148, "step": 6257 }, { "epoch": 0.67, "learning_rate": 2.6297440125667904e-06, "loss": 1.2219, "step": 6258 }, { "epoch": 0.67, "learning_rate": 2.628223113817302e-06, "loss": 1.2114, "step": 6259 }, { "epoch": 0.67, "learning_rate": 2.62670249818041e-06, "loss": 1.4327, "step": 6260 }, { "epoch": 0.67, "learning_rate": 2.6251821658376265e-06, "loss": 1.2807, "step": 6261 }, { "epoch": 0.67, "learning_rate": 2.6236621169704234e-06, "loss": 1.5555, "step": 6262 }, { "epoch": 0.67, "learning_rate": 2.622142351760254e-06, "loss": 1.3339, "step": 6263 }, { "epoch": 0.67, "learning_rate": 2.620622870388524e-06, "loss": 1.521, "step": 6264 }, { "epoch": 0.67, "learning_rate": 2.619103673036608e-06, "loss": 1.1703, "step": 6265 }, { "epoch": 0.67, "learning_rate": 2.617584759885853e-06, "loss": 1.1023, "step": 6266 }, { "epoch": 0.67, "learning_rate": 2.616066131117563e-06, "loss": 1.3003, "step": 6267 }, { "epoch": 0.67, "learning_rate": 2.6145477869130143e-06, "loss": 1.4458, "step": 6268 }, { "epoch": 0.67, "learning_rate": 2.61302972745345e-06, "loss": 1.4666, "step": 6269 }, { "epoch": 0.67, "learning_rate": 2.6115119529200748e-06, "loss": 1.4602, "step": 6270 }, { "epoch": 0.67, "learning_rate": 2.609994463494059e-06, "loss": 1.5195, "step": 6271 }, { "epoch": 0.67, "learning_rate": 2.608477259356543e-06, "loss": 1.6906, "step": 6272 }, { "epoch": 0.67, "learning_rate": 2.6069603406886347e-06, "loss": 1.5554, "step": 6273 }, { "epoch": 0.67, "learning_rate": 2.6054437076713997e-06, "loss": 1.3888, "step": 6274 }, { "epoch": 0.67, "learning_rate": 2.603927360485878e-06, "loss": 1.5316, "step": 6275 }, { "epoch": 0.67, "learning_rate": 2.60241129931307e-06, "loss": 1.4588, "step": 6276 }, { "epoch": 0.67, "learning_rate": 2.6008955243339417e-06, "loss": 1.1971, "step": 6277 }, { "epoch": 0.67, "learning_rate": 2.5993800357294298e-06, "loss": 1.1685, "step": 6278 }, { "epoch": 0.67, "learning_rate": 2.597864833680436e-06, "loss": 1.4165, "step": 6279 }, { "epoch": 0.67, "learning_rate": 2.5963499183678233e-06, "loss": 1.2724, "step": 6280 }, { "epoch": 0.67, "learning_rate": 2.5948352899724206e-06, "loss": 1.0605, "step": 6281 }, { "epoch": 0.67, "learning_rate": 2.593320948675029e-06, "loss": 1.4215, "step": 6282 }, { "epoch": 0.67, "learning_rate": 2.5918068946564113e-06, "loss": 1.6797, "step": 6283 }, { "epoch": 0.67, "learning_rate": 2.590293128097292e-06, "loss": 1.4036, "step": 6284 }, { "epoch": 0.67, "learning_rate": 2.588779649178371e-06, "loss": 1.5052, "step": 6285 }, { "epoch": 0.67, "learning_rate": 2.5872664580803054e-06, "loss": 1.3007, "step": 6286 }, { "epoch": 0.67, "learning_rate": 2.585753554983716e-06, "loss": 1.3422, "step": 6287 }, { "epoch": 0.67, "learning_rate": 2.5842409400692026e-06, "loss": 1.5258, "step": 6288 }, { "epoch": 0.67, "learning_rate": 2.582728613517318e-06, "loss": 1.4439, "step": 6289 }, { "epoch": 0.67, "learning_rate": 2.5812165755085828e-06, "loss": 1.6999, "step": 6290 }, { "epoch": 0.67, "learning_rate": 2.579704826223488e-06, "loss": 1.3753, "step": 6291 }, { "epoch": 0.67, "learning_rate": 2.578193365842484e-06, "loss": 1.391, "step": 6292 }, { "epoch": 0.67, "learning_rate": 2.576682194545991e-06, "loss": 1.2926, "step": 6293 }, { "epoch": 0.67, "learning_rate": 2.575171312514395e-06, "loss": 1.6157, "step": 6294 }, { "epoch": 0.67, "learning_rate": 2.5736607199280457e-06, "loss": 1.4462, "step": 6295 }, { "epoch": 0.67, "learning_rate": 2.572150416967255e-06, "loss": 1.3628, "step": 6296 }, { "epoch": 0.67, "learning_rate": 2.570640403812306e-06, "loss": 1.5342, "step": 6297 }, { "epoch": 0.67, "learning_rate": 2.569130680643447e-06, "loss": 1.4728, "step": 6298 }, { "epoch": 0.67, "learning_rate": 2.5676212476408858e-06, "loss": 1.3255, "step": 6299 }, { "epoch": 0.67, "learning_rate": 2.5661121049848026e-06, "loss": 1.5185, "step": 6300 }, { "epoch": 0.67, "learning_rate": 2.564603252855336e-06, "loss": 1.3556, "step": 6301 }, { "epoch": 0.67, "learning_rate": 2.5630946914325983e-06, "loss": 1.381, "step": 6302 }, { "epoch": 0.67, "learning_rate": 2.5615864208966573e-06, "loss": 1.1586, "step": 6303 }, { "epoch": 0.67, "learning_rate": 2.560078441427556e-06, "loss": 1.5026, "step": 6304 }, { "epoch": 0.67, "learning_rate": 2.5585707532052933e-06, "loss": 1.3412, "step": 6305 }, { "epoch": 0.67, "learning_rate": 2.55706335640984e-06, "loss": 1.4008, "step": 6306 }, { "epoch": 0.67, "learning_rate": 2.5555562512211327e-06, "loss": 1.3157, "step": 6307 }, { "epoch": 0.67, "learning_rate": 2.5540494378190674e-06, "loss": 1.4999, "step": 6308 }, { "epoch": 0.67, "learning_rate": 2.552542916383507e-06, "loss": 1.3432, "step": 6309 }, { "epoch": 0.67, "learning_rate": 2.551036687094285e-06, "loss": 1.5027, "step": 6310 }, { "epoch": 0.67, "learning_rate": 2.5495307501311904e-06, "loss": 1.4927, "step": 6311 }, { "epoch": 0.67, "learning_rate": 2.5480251056739874e-06, "loss": 1.439, "step": 6312 }, { "epoch": 0.67, "learning_rate": 2.5465197539024006e-06, "loss": 1.2253, "step": 6313 }, { "epoch": 0.67, "learning_rate": 2.5450146949961187e-06, "loss": 1.2055, "step": 6314 }, { "epoch": 0.67, "learning_rate": 2.543509929134794e-06, "loss": 1.5822, "step": 6315 }, { "epoch": 0.67, "learning_rate": 2.5420054564980497e-06, "loss": 1.5777, "step": 6316 }, { "epoch": 0.67, "learning_rate": 2.5405012772654702e-06, "loss": 1.2601, "step": 6317 }, { "epoch": 0.67, "learning_rate": 2.5389973916166037e-06, "loss": 1.3496, "step": 6318 }, { "epoch": 0.67, "learning_rate": 2.5374937997309677e-06, "loss": 1.9216, "step": 6319 }, { "epoch": 0.67, "learning_rate": 2.5359905017880406e-06, "loss": 1.3802, "step": 6320 }, { "epoch": 0.67, "learning_rate": 2.534487497967262e-06, "loss": 1.4567, "step": 6321 }, { "epoch": 0.67, "learning_rate": 2.5329847884480495e-06, "loss": 1.5371, "step": 6322 }, { "epoch": 0.67, "learning_rate": 2.5314823734097748e-06, "loss": 1.3682, "step": 6323 }, { "epoch": 0.67, "learning_rate": 2.529980253031774e-06, "loss": 1.2741, "step": 6324 }, { "epoch": 0.67, "learning_rate": 2.528478427493355e-06, "loss": 1.5027, "step": 6325 }, { "epoch": 0.67, "learning_rate": 2.526976896973784e-06, "loss": 1.3294, "step": 6326 }, { "epoch": 0.67, "learning_rate": 2.5254756616522953e-06, "loss": 1.2798, "step": 6327 }, { "epoch": 0.67, "learning_rate": 2.523974721708089e-06, "loss": 1.3238, "step": 6328 }, { "epoch": 0.68, "learning_rate": 2.5224740773203282e-06, "loss": 1.4631, "step": 6329 }, { "epoch": 0.68, "learning_rate": 2.5209737286681367e-06, "loss": 1.2867, "step": 6330 }, { "epoch": 0.68, "learning_rate": 2.5194736759306106e-06, "loss": 1.479, "step": 6331 }, { "epoch": 0.68, "learning_rate": 2.5179739192868073e-06, "loss": 1.4768, "step": 6332 }, { "epoch": 0.68, "learning_rate": 2.5164744589157488e-06, "loss": 1.4439, "step": 6333 }, { "epoch": 0.68, "learning_rate": 2.514975294996418e-06, "loss": 1.351, "step": 6334 }, { "epoch": 0.68, "learning_rate": 2.5134764277077713e-06, "loss": 1.3773, "step": 6335 }, { "epoch": 0.68, "learning_rate": 2.5119778572287195e-06, "loss": 1.2821, "step": 6336 }, { "epoch": 0.68, "learning_rate": 2.5104795837381457e-06, "loss": 1.2649, "step": 6337 }, { "epoch": 0.68, "learning_rate": 2.5089816074148953e-06, "loss": 1.3536, "step": 6338 }, { "epoch": 0.68, "learning_rate": 2.5074839284377774e-06, "loss": 1.3237, "step": 6339 }, { "epoch": 0.68, "learning_rate": 2.5059865469855617e-06, "loss": 1.5961, "step": 6340 }, { "epoch": 0.68, "learning_rate": 2.504489463236993e-06, "loss": 1.4283, "step": 6341 }, { "epoch": 0.68, "learning_rate": 2.5029926773707713e-06, "loss": 1.666, "step": 6342 }, { "epoch": 0.68, "learning_rate": 2.5014961895655628e-06, "loss": 1.6949, "step": 6343 }, { "epoch": 0.68, "learning_rate": 2.5000000000000015e-06, "loss": 1.3107, "step": 6344 }, { "epoch": 0.68, "learning_rate": 2.49850410885268e-06, "loss": 1.1607, "step": 6345 }, { "epoch": 0.68, "learning_rate": 2.497008516302161e-06, "loss": 1.3549, "step": 6346 }, { "epoch": 0.68, "learning_rate": 2.4955132225269718e-06, "loss": 1.3667, "step": 6347 }, { "epoch": 0.68, "learning_rate": 2.4940182277055987e-06, "loss": 1.6528, "step": 6348 }, { "epoch": 0.68, "learning_rate": 2.4925235320164935e-06, "loss": 1.4437, "step": 6349 }, { "epoch": 0.68, "learning_rate": 2.491029135638076e-06, "loss": 1.3678, "step": 6350 }, { "epoch": 0.68, "learning_rate": 2.4895350387487304e-06, "loss": 1.6247, "step": 6351 }, { "epoch": 0.68, "learning_rate": 2.488041241526799e-06, "loss": 1.2024, "step": 6352 }, { "epoch": 0.68, "learning_rate": 2.4865477441505963e-06, "loss": 1.6338, "step": 6353 }, { "epoch": 0.68, "learning_rate": 2.485054546798395e-06, "loss": 1.4651, "step": 6354 }, { "epoch": 0.68, "learning_rate": 2.483561649648432e-06, "loss": 1.3163, "step": 6355 }, { "epoch": 0.68, "learning_rate": 2.4820690528789126e-06, "loss": 1.0483, "step": 6356 }, { "epoch": 0.68, "learning_rate": 2.4805767566680057e-06, "loss": 1.5926, "step": 6357 }, { "epoch": 0.68, "learning_rate": 2.479084761193839e-06, "loss": 1.485, "step": 6358 }, { "epoch": 0.68, "learning_rate": 2.477593066634512e-06, "loss": 1.3788, "step": 6359 }, { "epoch": 0.68, "learning_rate": 2.4761016731680792e-06, "loss": 1.1943, "step": 6360 }, { "epoch": 0.68, "learning_rate": 2.4746105809725694e-06, "loss": 1.4048, "step": 6361 }, { "epoch": 0.68, "learning_rate": 2.4731197902259665e-06, "loss": 1.4682, "step": 6362 }, { "epoch": 0.68, "learning_rate": 2.4716293011062248e-06, "loss": 1.2396, "step": 6363 }, { "epoch": 0.68, "learning_rate": 2.4701391137912573e-06, "loss": 1.3222, "step": 6364 }, { "epoch": 0.68, "learning_rate": 2.4686492284589447e-06, "loss": 1.2466, "step": 6365 }, { "epoch": 0.68, "learning_rate": 2.467159645287133e-06, "loss": 1.545, "step": 6366 }, { "epoch": 0.68, "learning_rate": 2.4656703644536277e-06, "loss": 1.5514, "step": 6367 }, { "epoch": 0.68, "learning_rate": 2.464181386136198e-06, "loss": 1.499, "step": 6368 }, { "epoch": 0.68, "learning_rate": 2.4626927105125834e-06, "loss": 1.0319, "step": 6369 }, { "epoch": 0.68, "learning_rate": 2.4612043377604795e-06, "loss": 1.4809, "step": 6370 }, { "epoch": 0.68, "learning_rate": 2.459716268057551e-06, "loss": 1.301, "step": 6371 }, { "epoch": 0.68, "learning_rate": 2.4582285015814263e-06, "loss": 1.5725, "step": 6372 }, { "epoch": 0.68, "learning_rate": 2.456741038509694e-06, "loss": 1.3218, "step": 6373 }, { "epoch": 0.68, "learning_rate": 2.4552538790199075e-06, "loss": 1.469, "step": 6374 }, { "epoch": 0.68, "learning_rate": 2.4537670232895866e-06, "loss": 1.6244, "step": 6375 }, { "epoch": 0.68, "learning_rate": 2.452280471496215e-06, "loss": 1.2043, "step": 6376 }, { "epoch": 0.68, "learning_rate": 2.4507942238172346e-06, "loss": 1.3723, "step": 6377 }, { "epoch": 0.68, "learning_rate": 2.4493082804300585e-06, "loss": 1.414, "step": 6378 }, { "epoch": 0.68, "learning_rate": 2.447822641512058e-06, "loss": 1.35, "step": 6379 }, { "epoch": 0.68, "learning_rate": 2.4463373072405655e-06, "loss": 1.385, "step": 6380 }, { "epoch": 0.68, "learning_rate": 2.4448522777928903e-06, "loss": 1.3472, "step": 6381 }, { "epoch": 0.68, "learning_rate": 2.443367553346291e-06, "loss": 1.3136, "step": 6382 }, { "epoch": 0.68, "learning_rate": 2.4418831340779934e-06, "loss": 1.4453, "step": 6383 }, { "epoch": 0.68, "learning_rate": 2.4403990201651915e-06, "loss": 1.5193, "step": 6384 }, { "epoch": 0.68, "learning_rate": 2.438915211785041e-06, "loss": 1.4758, "step": 6385 }, { "epoch": 0.68, "learning_rate": 2.4374317091146593e-06, "loss": 1.5526, "step": 6386 }, { "epoch": 0.68, "learning_rate": 2.435948512331125e-06, "loss": 1.5897, "step": 6387 }, { "epoch": 0.68, "learning_rate": 2.4344656216114878e-06, "loss": 0.9369, "step": 6388 }, { "epoch": 0.68, "learning_rate": 2.4329830371327513e-06, "loss": 1.2372, "step": 6389 }, { "epoch": 0.68, "learning_rate": 2.4315007590718913e-06, "loss": 1.271, "step": 6390 }, { "epoch": 0.68, "learning_rate": 2.430018787605844e-06, "loss": 1.5154, "step": 6391 }, { "epoch": 0.68, "learning_rate": 2.4285371229115073e-06, "loss": 1.517, "step": 6392 }, { "epoch": 0.68, "learning_rate": 2.427055765165741e-06, "loss": 1.36, "step": 6393 }, { "epoch": 0.68, "learning_rate": 2.425574714545373e-06, "loss": 1.1727, "step": 6394 }, { "epoch": 0.68, "learning_rate": 2.4240939712271933e-06, "loss": 1.249, "step": 6395 }, { "epoch": 0.68, "learning_rate": 2.4226135353879516e-06, "loss": 1.219, "step": 6396 }, { "epoch": 0.68, "learning_rate": 2.4211334072043668e-06, "loss": 1.7015, "step": 6397 }, { "epoch": 0.68, "learning_rate": 2.419653586853116e-06, "loss": 1.4242, "step": 6398 }, { "epoch": 0.68, "learning_rate": 2.4181740745108377e-06, "loss": 1.5033, "step": 6399 }, { "epoch": 0.68, "learning_rate": 2.416694870354145e-06, "loss": 1.4552, "step": 6400 }, { "epoch": 0.68, "learning_rate": 2.4152159745596015e-06, "loss": 1.3706, "step": 6401 }, { "epoch": 0.68, "learning_rate": 2.413737387303739e-06, "loss": 1.3428, "step": 6402 }, { "epoch": 0.68, "learning_rate": 2.4122591087630548e-06, "loss": 1.152, "step": 6403 }, { "epoch": 0.68, "learning_rate": 2.4107811391140036e-06, "loss": 1.4076, "step": 6404 }, { "epoch": 0.68, "learning_rate": 2.4093034785330087e-06, "loss": 1.3645, "step": 6405 }, { "epoch": 0.68, "learning_rate": 2.4078261271964567e-06, "loss": 1.243, "step": 6406 }, { "epoch": 0.68, "learning_rate": 2.406349085280692e-06, "loss": 1.5644, "step": 6407 }, { "epoch": 0.68, "learning_rate": 2.4048723529620246e-06, "loss": 1.433, "step": 6408 }, { "epoch": 0.68, "learning_rate": 2.4033959304167288e-06, "loss": 1.3834, "step": 6409 }, { "epoch": 0.68, "learning_rate": 2.4019198178210435e-06, "loss": 1.4997, "step": 6410 }, { "epoch": 0.68, "learning_rate": 2.4004440153511642e-06, "loss": 1.3901, "step": 6411 }, { "epoch": 0.68, "learning_rate": 2.398968523183258e-06, "loss": 1.2791, "step": 6412 }, { "epoch": 0.68, "learning_rate": 2.397493341493448e-06, "loss": 1.239, "step": 6413 }, { "epoch": 0.68, "learning_rate": 2.396018470457821e-06, "loss": 1.6929, "step": 6414 }, { "epoch": 0.68, "learning_rate": 2.3945439102524306e-06, "loss": 1.3044, "step": 6415 }, { "epoch": 0.68, "learning_rate": 2.3930696610532915e-06, "loss": 1.2319, "step": 6416 }, { "epoch": 0.68, "learning_rate": 2.3915957230363783e-06, "loss": 1.6422, "step": 6417 }, { "epoch": 0.68, "learning_rate": 2.3901220963776343e-06, "loss": 1.5982, "step": 6418 }, { "epoch": 0.68, "learning_rate": 2.388648781252959e-06, "loss": 1.3761, "step": 6419 }, { "epoch": 0.68, "learning_rate": 2.3871757778382216e-06, "loss": 1.3927, "step": 6420 }, { "epoch": 0.68, "learning_rate": 2.385703086309247e-06, "loss": 1.4887, "step": 6421 }, { "epoch": 0.69, "learning_rate": 2.384230706841829e-06, "loss": 1.3648, "step": 6422 }, { "epoch": 0.69, "learning_rate": 2.3827586396117207e-06, "loss": 1.569, "step": 6423 }, { "epoch": 0.69, "learning_rate": 2.3812868847946357e-06, "loss": 1.2914, "step": 6424 }, { "epoch": 0.69, "learning_rate": 2.3798154425662597e-06, "loss": 1.3582, "step": 6425 }, { "epoch": 0.69, "learning_rate": 2.378344313102231e-06, "loss": 1.2558, "step": 6426 }, { "epoch": 0.69, "learning_rate": 2.376873496578153e-06, "loss": 1.4733, "step": 6427 }, { "epoch": 0.69, "learning_rate": 2.3754029931695954e-06, "loss": 1.2088, "step": 6428 }, { "epoch": 0.69, "learning_rate": 2.373932803052089e-06, "loss": 1.3795, "step": 6429 }, { "epoch": 0.69, "learning_rate": 2.372462926401124e-06, "loss": 1.4776, "step": 6430 }, { "epoch": 0.69, "learning_rate": 2.3709933633921576e-06, "loss": 1.1524, "step": 6431 }, { "epoch": 0.69, "learning_rate": 2.369524114200607e-06, "loss": 1.4373, "step": 6432 }, { "epoch": 0.69, "learning_rate": 2.3680551790018507e-06, "loss": 1.5517, "step": 6433 }, { "epoch": 0.69, "learning_rate": 2.3665865579712327e-06, "loss": 1.4348, "step": 6434 }, { "epoch": 0.69, "learning_rate": 2.3651182512840604e-06, "loss": 1.3698, "step": 6435 }, { "epoch": 0.69, "learning_rate": 2.363650259115598e-06, "loss": 1.3921, "step": 6436 }, { "epoch": 0.69, "learning_rate": 2.3621825816410797e-06, "loss": 1.2704, "step": 6437 }, { "epoch": 0.69, "learning_rate": 2.360715219035694e-06, "loss": 1.2525, "step": 6438 }, { "epoch": 0.69, "learning_rate": 2.3592481714745995e-06, "loss": 1.3556, "step": 6439 }, { "epoch": 0.69, "learning_rate": 2.357781439132911e-06, "loss": 1.284, "step": 6440 }, { "epoch": 0.69, "learning_rate": 2.356315022185712e-06, "loss": 1.6987, "step": 6441 }, { "epoch": 0.69, "learning_rate": 2.3548489208080392e-06, "loss": 1.4263, "step": 6442 }, { "epoch": 0.69, "learning_rate": 2.353383135174901e-06, "loss": 1.1475, "step": 6443 }, { "epoch": 0.69, "learning_rate": 2.3519176654612657e-06, "loss": 1.2295, "step": 6444 }, { "epoch": 0.69, "learning_rate": 2.35045251184206e-06, "loss": 1.54, "step": 6445 }, { "epoch": 0.69, "learning_rate": 2.3489876744921743e-06, "loss": 1.4147, "step": 6446 }, { "epoch": 0.69, "learning_rate": 2.3475231535864653e-06, "loss": 1.1742, "step": 6447 }, { "epoch": 0.69, "learning_rate": 2.346058949299745e-06, "loss": 1.2609, "step": 6448 }, { "epoch": 0.69, "learning_rate": 2.3445950618067935e-06, "loss": 1.7344, "step": 6449 }, { "epoch": 0.69, "learning_rate": 2.3431314912823543e-06, "loss": 1.6481, "step": 6450 }, { "epoch": 0.69, "learning_rate": 2.3416682379011264e-06, "loss": 1.604, "step": 6451 }, { "epoch": 0.69, "learning_rate": 2.340205301837773e-06, "loss": 1.2228, "step": 6452 }, { "epoch": 0.69, "learning_rate": 2.338742683266923e-06, "loss": 1.5486, "step": 6453 }, { "epoch": 0.69, "learning_rate": 2.337280382363166e-06, "loss": 1.5428, "step": 6454 }, { "epoch": 0.69, "learning_rate": 2.3358183993010513e-06, "loss": 1.2912, "step": 6455 }, { "epoch": 0.69, "learning_rate": 2.3343567342550933e-06, "loss": 1.4898, "step": 6456 }, { "epoch": 0.69, "learning_rate": 2.3328953873997666e-06, "loss": 1.3167, "step": 6457 }, { "epoch": 0.69, "learning_rate": 2.331434358909504e-06, "loss": 1.3494, "step": 6458 }, { "epoch": 0.69, "learning_rate": 2.329973648958712e-06, "loss": 1.4529, "step": 6459 }, { "epoch": 0.69, "learning_rate": 2.328513257721748e-06, "loss": 1.2892, "step": 6460 }, { "epoch": 0.69, "learning_rate": 2.327053185372933e-06, "loss": 1.4999, "step": 6461 }, { "epoch": 0.69, "learning_rate": 2.3255934320865555e-06, "loss": 1.4787, "step": 6462 }, { "epoch": 0.69, "learning_rate": 2.3241339980368584e-06, "loss": 1.3363, "step": 6463 }, { "epoch": 0.69, "learning_rate": 2.322674883398053e-06, "loss": 1.5125, "step": 6464 }, { "epoch": 0.69, "learning_rate": 2.3212160883443107e-06, "loss": 1.2071, "step": 6465 }, { "epoch": 0.69, "learning_rate": 2.319757613049763e-06, "loss": 1.5064, "step": 6466 }, { "epoch": 0.69, "learning_rate": 2.318299457688502e-06, "loss": 1.6659, "step": 6467 }, { "epoch": 0.69, "learning_rate": 2.316841622434586e-06, "loss": 1.458, "step": 6468 }, { "epoch": 0.69, "learning_rate": 2.315384107462033e-06, "loss": 1.2941, "step": 6469 }, { "epoch": 0.69, "learning_rate": 2.313926912944821e-06, "loss": 1.4797, "step": 6470 }, { "epoch": 0.69, "learning_rate": 2.3124700390568945e-06, "loss": 1.45, "step": 6471 }, { "epoch": 0.69, "learning_rate": 2.311013485972152e-06, "loss": 1.3529, "step": 6472 }, { "epoch": 0.69, "learning_rate": 2.309557253864463e-06, "loss": 1.3549, "step": 6473 }, { "epoch": 0.69, "learning_rate": 2.30810134290765e-06, "loss": 1.3846, "step": 6474 }, { "epoch": 0.69, "learning_rate": 2.306645753275504e-06, "loss": 1.4551, "step": 6475 }, { "epoch": 0.69, "learning_rate": 2.305190485141775e-06, "loss": 1.1761, "step": 6476 }, { "epoch": 0.69, "learning_rate": 2.3037355386801683e-06, "loss": 1.3317, "step": 6477 }, { "epoch": 0.69, "learning_rate": 2.3022809140643664e-06, "loss": 1.305, "step": 6478 }, { "epoch": 0.69, "learning_rate": 2.300826611467999e-06, "loss": 1.3682, "step": 6479 }, { "epoch": 0.69, "learning_rate": 2.2993726310646603e-06, "loss": 1.2453, "step": 6480 }, { "epoch": 0.69, "learning_rate": 2.297918973027913e-06, "loss": 1.3827, "step": 6481 }, { "epoch": 0.69, "learning_rate": 2.296465637531271e-06, "loss": 1.4207, "step": 6482 }, { "epoch": 0.69, "learning_rate": 2.2950126247482178e-06, "loss": 1.5056, "step": 6483 }, { "epoch": 0.69, "learning_rate": 2.2935599348521974e-06, "loss": 1.5324, "step": 6484 }, { "epoch": 0.69, "learning_rate": 2.2921075680166116e-06, "loss": 1.63, "step": 6485 }, { "epoch": 0.69, "learning_rate": 2.2906555244148233e-06, "loss": 1.3444, "step": 6486 }, { "epoch": 0.69, "learning_rate": 2.2892038042201615e-06, "loss": 1.2544, "step": 6487 }, { "epoch": 0.69, "learning_rate": 2.2877524076059155e-06, "loss": 1.4269, "step": 6488 }, { "epoch": 0.69, "learning_rate": 2.2863013347453305e-06, "loss": 1.5321, "step": 6489 }, { "epoch": 0.69, "learning_rate": 2.2848505858116216e-06, "loss": 1.3743, "step": 6490 }, { "epoch": 0.69, "learning_rate": 2.283400160977959e-06, "loss": 1.6724, "step": 6491 }, { "epoch": 0.69, "learning_rate": 2.2819500604174733e-06, "loss": 1.6063, "step": 6492 }, { "epoch": 0.69, "learning_rate": 2.280500284303262e-06, "loss": 1.3224, "step": 6493 }, { "epoch": 0.69, "learning_rate": 2.2790508328083823e-06, "loss": 1.3595, "step": 6494 }, { "epoch": 0.69, "learning_rate": 2.277601706105847e-06, "loss": 1.2998, "step": 6495 }, { "epoch": 0.69, "learning_rate": 2.2761529043686386e-06, "loss": 1.1975, "step": 6496 }, { "epoch": 0.69, "learning_rate": 2.2747044277696934e-06, "loss": 1.3978, "step": 6497 }, { "epoch": 0.69, "learning_rate": 2.2732562764819157e-06, "loss": 1.4048, "step": 6498 }, { "epoch": 0.69, "learning_rate": 2.271808450678163e-06, "loss": 1.3786, "step": 6499 }, { "epoch": 0.69, "learning_rate": 2.2703609505312628e-06, "loss": 1.4769, "step": 6500 }, { "epoch": 0.69, "learning_rate": 2.2689137762139952e-06, "loss": 1.4468, "step": 6501 }, { "epoch": 0.69, "learning_rate": 2.2674669278991076e-06, "loss": 1.6313, "step": 6502 }, { "epoch": 0.69, "learning_rate": 2.2660204057593076e-06, "loss": 1.4656, "step": 6503 }, { "epoch": 0.69, "learning_rate": 2.264574209967262e-06, "loss": 1.6881, "step": 6504 }, { "epoch": 0.69, "learning_rate": 2.263128340695596e-06, "loss": 1.6615, "step": 6505 }, { "epoch": 0.69, "learning_rate": 2.2616827981169036e-06, "loss": 1.3606, "step": 6506 }, { "epoch": 0.69, "learning_rate": 2.260237582403732e-06, "loss": 1.4882, "step": 6507 }, { "epoch": 0.69, "learning_rate": 2.2587926937285935e-06, "loss": 1.2442, "step": 6508 }, { "epoch": 0.69, "learning_rate": 2.2573481322639635e-06, "loss": 1.3059, "step": 6509 }, { "epoch": 0.69, "learning_rate": 2.2559038981822724e-06, "loss": 1.4014, "step": 6510 }, { "epoch": 0.69, "learning_rate": 2.2544599916559135e-06, "loss": 1.4446, "step": 6511 }, { "epoch": 0.69, "learning_rate": 2.253016412857244e-06, "loss": 1.561, "step": 6512 }, { "epoch": 0.69, "learning_rate": 2.2515731619585814e-06, "loss": 1.4439, "step": 6513 }, { "epoch": 0.69, "learning_rate": 2.2501302391321985e-06, "loss": 1.3873, "step": 6514 }, { "epoch": 0.69, "learning_rate": 2.2486876445503375e-06, "loss": 1.2875, "step": 6515 }, { "epoch": 0.7, "learning_rate": 2.247245378385195e-06, "loss": 1.2599, "step": 6516 }, { "epoch": 0.7, "learning_rate": 2.245803440808927e-06, "loss": 1.3216, "step": 6517 }, { "epoch": 0.7, "learning_rate": 2.2443618319936605e-06, "loss": 1.225, "step": 6518 }, { "epoch": 0.7, "learning_rate": 2.242920552111473e-06, "loss": 1.6458, "step": 6519 }, { "epoch": 0.7, "learning_rate": 2.2414796013344053e-06, "loss": 1.5536, "step": 6520 }, { "epoch": 0.7, "learning_rate": 2.2400389798344602e-06, "loss": 1.1525, "step": 6521 }, { "epoch": 0.7, "learning_rate": 2.238598687783603e-06, "loss": 1.3623, "step": 6522 }, { "epoch": 0.7, "learning_rate": 2.2371587253537536e-06, "loss": 1.2367, "step": 6523 }, { "epoch": 0.7, "learning_rate": 2.235719092716801e-06, "loss": 1.1852, "step": 6524 }, { "epoch": 0.7, "learning_rate": 2.234279790044588e-06, "loss": 1.4986, "step": 6525 }, { "epoch": 0.7, "learning_rate": 2.232840817508918e-06, "loss": 1.3254, "step": 6526 }, { "epoch": 0.7, "learning_rate": 2.2314021752815596e-06, "loss": 1.4505, "step": 6527 }, { "epoch": 0.7, "learning_rate": 2.229963863534241e-06, "loss": 1.4789, "step": 6528 }, { "epoch": 0.7, "learning_rate": 2.2285258824386487e-06, "loss": 1.5544, "step": 6529 }, { "epoch": 0.7, "learning_rate": 2.227088232166427e-06, "loss": 1.5091, "step": 6530 }, { "epoch": 0.7, "learning_rate": 2.225650912889188e-06, "loss": 1.4768, "step": 6531 }, { "epoch": 0.7, "learning_rate": 2.2242139247785016e-06, "loss": 1.341, "step": 6532 }, { "epoch": 0.7, "learning_rate": 2.222777268005894e-06, "loss": 1.4116, "step": 6533 }, { "epoch": 0.7, "learning_rate": 2.221340942742858e-06, "loss": 1.5664, "step": 6534 }, { "epoch": 0.7, "learning_rate": 2.219904949160842e-06, "loss": 1.2838, "step": 6535 }, { "epoch": 0.7, "learning_rate": 2.2184692874312542e-06, "loss": 1.5033, "step": 6536 }, { "epoch": 0.7, "learning_rate": 2.2170339577254714e-06, "loss": 1.4009, "step": 6537 }, { "epoch": 0.7, "learning_rate": 2.2155989602148224e-06, "loss": 1.3406, "step": 6538 }, { "epoch": 0.7, "learning_rate": 2.2141642950705973e-06, "loss": 1.3381, "step": 6539 }, { "epoch": 0.7, "learning_rate": 2.212729962464051e-06, "loss": 1.3993, "step": 6540 }, { "epoch": 0.7, "learning_rate": 2.2112959625663926e-06, "loss": 1.3401, "step": 6541 }, { "epoch": 0.7, "learning_rate": 2.209862295548797e-06, "loss": 1.3884, "step": 6542 }, { "epoch": 0.7, "learning_rate": 2.208428961582399e-06, "loss": 1.4216, "step": 6543 }, { "epoch": 0.7, "learning_rate": 2.206995960838289e-06, "loss": 1.1743, "step": 6544 }, { "epoch": 0.7, "learning_rate": 2.2055632934875194e-06, "loss": 1.4138, "step": 6545 }, { "epoch": 0.7, "learning_rate": 2.2041309597011057e-06, "loss": 1.2689, "step": 6546 }, { "epoch": 0.7, "learning_rate": 2.202698959650023e-06, "loss": 1.1034, "step": 6547 }, { "epoch": 0.7, "learning_rate": 2.2012672935052025e-06, "loss": 1.6019, "step": 6548 }, { "epoch": 0.7, "learning_rate": 2.1998359614375412e-06, "loss": 1.2585, "step": 6549 }, { "epoch": 0.7, "learning_rate": 2.198404963617892e-06, "loss": 1.3793, "step": 6550 }, { "epoch": 0.7, "learning_rate": 2.196974300217067e-06, "loss": 1.5814, "step": 6551 }, { "epoch": 0.7, "learning_rate": 2.1955439714058422e-06, "loss": 1.631, "step": 6552 }, { "epoch": 0.7, "learning_rate": 2.194113977354955e-06, "loss": 1.2699, "step": 6553 }, { "epoch": 0.7, "learning_rate": 2.1926843182350955e-06, "loss": 1.6247, "step": 6554 }, { "epoch": 0.7, "learning_rate": 2.191254994216922e-06, "loss": 1.4434, "step": 6555 }, { "epoch": 0.7, "learning_rate": 2.189826005471045e-06, "loss": 1.5375, "step": 6556 }, { "epoch": 0.7, "learning_rate": 2.1883973521680435e-06, "loss": 1.4759, "step": 6557 }, { "epoch": 0.7, "learning_rate": 2.186969034478448e-06, "loss": 1.4956, "step": 6558 }, { "epoch": 0.7, "learning_rate": 2.1855410525727556e-06, "loss": 1.5095, "step": 6559 }, { "epoch": 0.7, "learning_rate": 2.1841134066214186e-06, "loss": 1.3909, "step": 6560 }, { "epoch": 0.7, "learning_rate": 2.182686096794852e-06, "loss": 1.3089, "step": 6561 }, { "epoch": 0.7, "learning_rate": 2.181259123263432e-06, "loss": 1.5122, "step": 6562 }, { "epoch": 0.7, "learning_rate": 2.179832486197492e-06, "loss": 1.2631, "step": 6563 }, { "epoch": 0.7, "learning_rate": 2.1784061857673217e-06, "loss": 1.59, "step": 6564 }, { "epoch": 0.7, "learning_rate": 2.176980222143178e-06, "loss": 1.4734, "step": 6565 }, { "epoch": 0.7, "learning_rate": 2.1755545954952755e-06, "loss": 1.7285, "step": 6566 }, { "epoch": 0.7, "learning_rate": 2.174129305993784e-06, "loss": 1.7385, "step": 6567 }, { "epoch": 0.7, "learning_rate": 2.1727043538088406e-06, "loss": 1.398, "step": 6568 }, { "epoch": 0.7, "learning_rate": 2.1712797391105354e-06, "loss": 1.4212, "step": 6569 }, { "epoch": 0.7, "learning_rate": 2.1698554620689178e-06, "loss": 1.315, "step": 6570 }, { "epoch": 0.7, "learning_rate": 2.1684315228540066e-06, "loss": 1.3698, "step": 6571 }, { "epoch": 0.7, "learning_rate": 2.16700792163577e-06, "loss": 1.3504, "step": 6572 }, { "epoch": 0.7, "learning_rate": 2.165584658584138e-06, "loss": 1.3046, "step": 6573 }, { "epoch": 0.7, "learning_rate": 2.1641617338690054e-06, "loss": 1.3627, "step": 6574 }, { "epoch": 0.7, "learning_rate": 2.162739147660219e-06, "loss": 1.419, "step": 6575 }, { "epoch": 0.7, "learning_rate": 2.16131690012759e-06, "loss": 1.4205, "step": 6576 }, { "epoch": 0.7, "learning_rate": 2.1598949914408917e-06, "loss": 1.4102, "step": 6577 }, { "epoch": 0.7, "learning_rate": 2.1584734217698507e-06, "loss": 1.6465, "step": 6578 }, { "epoch": 0.7, "learning_rate": 2.157052191284154e-06, "loss": 1.3492, "step": 6579 }, { "epoch": 0.7, "learning_rate": 2.155631300153453e-06, "loss": 1.1644, "step": 6580 }, { "epoch": 0.7, "learning_rate": 2.1542107485473557e-06, "loss": 1.4366, "step": 6581 }, { "epoch": 0.7, "learning_rate": 2.1527905366354292e-06, "loss": 1.2237, "step": 6582 }, { "epoch": 0.7, "learning_rate": 2.1513706645871973e-06, "loss": 1.5121, "step": 6583 }, { "epoch": 0.7, "learning_rate": 2.149951132572151e-06, "loss": 1.5377, "step": 6584 }, { "epoch": 0.7, "learning_rate": 2.1485319407597315e-06, "loss": 1.2109, "step": 6585 }, { "epoch": 0.7, "learning_rate": 2.147113089319347e-06, "loss": 1.5836, "step": 6586 }, { "epoch": 0.7, "learning_rate": 2.145694578420362e-06, "loss": 1.4586, "step": 6587 }, { "epoch": 0.7, "learning_rate": 2.1442764082321e-06, "loss": 1.1299, "step": 6588 }, { "epoch": 0.7, "learning_rate": 2.1428585789238416e-06, "loss": 1.1259, "step": 6589 }, { "epoch": 0.7, "learning_rate": 2.1414410906648308e-06, "loss": 1.3984, "step": 6590 }, { "epoch": 0.7, "learning_rate": 2.140023943624272e-06, "loss": 1.5469, "step": 6591 }, { "epoch": 0.7, "learning_rate": 2.138607137971322e-06, "loss": 1.1508, "step": 6592 }, { "epoch": 0.7, "learning_rate": 2.137190673875105e-06, "loss": 1.1188, "step": 6593 }, { "epoch": 0.7, "learning_rate": 2.135774551504698e-06, "loss": 1.0909, "step": 6594 }, { "epoch": 0.7, "learning_rate": 2.134358771029137e-06, "loss": 1.3078, "step": 6595 }, { "epoch": 0.7, "learning_rate": 2.1329433326174265e-06, "loss": 1.4916, "step": 6596 }, { "epoch": 0.7, "learning_rate": 2.1315282364385197e-06, "loss": 1.2771, "step": 6597 }, { "epoch": 0.7, "learning_rate": 2.130113482661332e-06, "loss": 1.4802, "step": 6598 }, { "epoch": 0.7, "learning_rate": 2.1286990714547417e-06, "loss": 1.4815, "step": 6599 }, { "epoch": 0.7, "learning_rate": 2.1272850029875802e-06, "loss": 1.1294, "step": 6600 }, { "epoch": 0.7, "learning_rate": 2.1258712774286416e-06, "loss": 1.3939, "step": 6601 }, { "epoch": 0.7, "learning_rate": 2.1244578949466814e-06, "loss": 1.4137, "step": 6602 }, { "epoch": 0.7, "learning_rate": 2.1230448557104087e-06, "loss": 1.4463, "step": 6603 }, { "epoch": 0.7, "learning_rate": 2.1216321598884935e-06, "loss": 1.3318, "step": 6604 }, { "epoch": 0.7, "learning_rate": 2.1202198076495663e-06, "loss": 1.4043, "step": 6605 }, { "epoch": 0.7, "learning_rate": 2.1188077991622174e-06, "loss": 1.3623, "step": 6606 }, { "epoch": 0.7, "learning_rate": 2.1173961345949915e-06, "loss": 1.4042, "step": 6607 }, { "epoch": 0.7, "learning_rate": 2.1159848141163988e-06, "loss": 1.608, "step": 6608 }, { "epoch": 0.7, "learning_rate": 2.1145738378949004e-06, "loss": 1.6769, "step": 6609 }, { "epoch": 0.71, "learning_rate": 2.1131632060989255e-06, "loss": 1.2223, "step": 6610 }, { "epoch": 0.71, "learning_rate": 2.111752918896853e-06, "loss": 1.5368, "step": 6611 }, { "epoch": 0.71, "learning_rate": 2.110342976457029e-06, "loss": 1.403, "step": 6612 }, { "epoch": 0.71, "learning_rate": 2.1089333789477502e-06, "loss": 1.4872, "step": 6613 }, { "epoch": 0.71, "learning_rate": 2.107524126537279e-06, "loss": 1.3929, "step": 6614 }, { "epoch": 0.71, "learning_rate": 2.1061152193938355e-06, "loss": 1.4891, "step": 6615 }, { "epoch": 0.71, "learning_rate": 2.1047066576855964e-06, "loss": 1.2215, "step": 6616 }, { "epoch": 0.71, "learning_rate": 2.103298441580694e-06, "loss": 1.3455, "step": 6617 }, { "epoch": 0.71, "learning_rate": 2.1018905712472285e-06, "loss": 1.2607, "step": 6618 }, { "epoch": 0.71, "learning_rate": 2.10048304685325e-06, "loss": 1.4742, "step": 6619 }, { "epoch": 0.71, "learning_rate": 2.099075868566772e-06, "loss": 1.4065, "step": 6620 }, { "epoch": 0.71, "learning_rate": 2.0976690365557673e-06, "loss": 1.4301, "step": 6621 }, { "epoch": 0.71, "learning_rate": 2.096262550988164e-06, "loss": 1.2781, "step": 6622 }, { "epoch": 0.71, "learning_rate": 2.0948564120318488e-06, "loss": 1.3267, "step": 6623 }, { "epoch": 0.71, "learning_rate": 2.093450619854671e-06, "loss": 1.3638, "step": 6624 }, { "epoch": 0.71, "learning_rate": 2.0920451746244367e-06, "loss": 1.3316, "step": 6625 }, { "epoch": 0.71, "learning_rate": 2.090640076508908e-06, "loss": 1.3786, "step": 6626 }, { "epoch": 0.71, "learning_rate": 2.0892353256758107e-06, "loss": 1.3127, "step": 6627 }, { "epoch": 0.71, "learning_rate": 2.087830922292824e-06, "loss": 1.4265, "step": 6628 }, { "epoch": 0.71, "learning_rate": 2.0864268665275854e-06, "loss": 1.1046, "step": 6629 }, { "epoch": 0.71, "learning_rate": 2.0850231585476965e-06, "loss": 1.3842, "step": 6630 }, { "epoch": 0.71, "learning_rate": 2.083619798520715e-06, "loss": 1.3193, "step": 6631 }, { "epoch": 0.71, "learning_rate": 2.082216786614153e-06, "loss": 1.4251, "step": 6632 }, { "epoch": 0.71, "learning_rate": 2.0808141229954876e-06, "loss": 1.4227, "step": 6633 }, { "epoch": 0.71, "learning_rate": 2.079411807832147e-06, "loss": 1.3089, "step": 6634 }, { "epoch": 0.71, "learning_rate": 2.0780098412915257e-06, "loss": 1.5424, "step": 6635 }, { "epoch": 0.71, "learning_rate": 2.0766082235409695e-06, "loss": 1.3119, "step": 6636 }, { "epoch": 0.71, "learning_rate": 2.0752069547477887e-06, "loss": 1.4649, "step": 6637 }, { "epoch": 0.71, "learning_rate": 2.0738060350792454e-06, "loss": 1.1983, "step": 6638 }, { "epoch": 0.71, "learning_rate": 2.072405464702566e-06, "loss": 1.1625, "step": 6639 }, { "epoch": 0.71, "learning_rate": 2.0710052437849342e-06, "loss": 1.2889, "step": 6640 }, { "epoch": 0.71, "learning_rate": 2.0696053724934882e-06, "loss": 1.6584, "step": 6641 }, { "epoch": 0.71, "learning_rate": 2.068205850995326e-06, "loss": 1.4973, "step": 6642 }, { "epoch": 0.71, "learning_rate": 2.066806679457508e-06, "loss": 1.3564, "step": 6643 }, { "epoch": 0.71, "learning_rate": 2.065407858047046e-06, "loss": 1.5286, "step": 6644 }, { "epoch": 0.71, "learning_rate": 2.064009386930915e-06, "loss": 1.4681, "step": 6645 }, { "epoch": 0.71, "learning_rate": 2.062611266276049e-06, "loss": 1.2733, "step": 6646 }, { "epoch": 0.71, "learning_rate": 2.0612134962493353e-06, "loss": 1.4862, "step": 6647 }, { "epoch": 0.71, "learning_rate": 2.0598160770176208e-06, "loss": 1.2324, "step": 6648 }, { "epoch": 0.71, "learning_rate": 2.058419008747713e-06, "loss": 1.358, "step": 6649 }, { "epoch": 0.71, "learning_rate": 2.0570222916063782e-06, "loss": 1.3713, "step": 6650 }, { "epoch": 0.71, "learning_rate": 2.0556259257603355e-06, "loss": 1.5036, "step": 6651 }, { "epoch": 0.71, "learning_rate": 2.054229911376269e-06, "loss": 1.4613, "step": 6652 }, { "epoch": 0.71, "learning_rate": 2.0528342486208127e-06, "loss": 1.4915, "step": 6653 }, { "epoch": 0.71, "learning_rate": 2.0514389376605646e-06, "loss": 1.456, "step": 6654 }, { "epoch": 0.71, "learning_rate": 2.050043978662083e-06, "loss": 1.1633, "step": 6655 }, { "epoch": 0.71, "learning_rate": 2.0486493717918765e-06, "loss": 1.2556, "step": 6656 }, { "epoch": 0.71, "learning_rate": 2.0472551172164152e-06, "loss": 1.1118, "step": 6657 }, { "epoch": 0.71, "learning_rate": 2.045861215102128e-06, "loss": 1.5569, "step": 6658 }, { "epoch": 0.71, "learning_rate": 2.0444676656154037e-06, "loss": 1.3374, "step": 6659 }, { "epoch": 0.71, "learning_rate": 2.0430744689225833e-06, "loss": 1.4501, "step": 6660 }, { "epoch": 0.71, "learning_rate": 2.0416816251899726e-06, "loss": 1.3367, "step": 6661 }, { "epoch": 0.71, "learning_rate": 2.040289134583829e-06, "loss": 1.2118, "step": 6662 }, { "epoch": 0.71, "learning_rate": 2.0388969972703688e-06, "loss": 1.6304, "step": 6663 }, { "epoch": 0.71, "learning_rate": 2.0375052134157697e-06, "loss": 1.6487, "step": 6664 }, { "epoch": 0.71, "learning_rate": 2.0361137831861664e-06, "loss": 1.3806, "step": 6665 }, { "epoch": 0.71, "learning_rate": 2.0347227067476478e-06, "loss": 1.2336, "step": 6666 }, { "epoch": 0.71, "learning_rate": 2.033331984266265e-06, "loss": 1.2434, "step": 6667 }, { "epoch": 0.71, "learning_rate": 2.0319416159080217e-06, "loss": 1.4467, "step": 6668 }, { "epoch": 0.71, "learning_rate": 2.030551601838887e-06, "loss": 1.4347, "step": 6669 }, { "epoch": 0.71, "learning_rate": 2.0291619422247786e-06, "loss": 1.3526, "step": 6670 }, { "epoch": 0.71, "learning_rate": 2.0277726372315805e-06, "loss": 1.391, "step": 6671 }, { "epoch": 0.71, "learning_rate": 2.0263836870251277e-06, "loss": 1.1417, "step": 6672 }, { "epoch": 0.71, "learning_rate": 2.024995091771212e-06, "loss": 1.2092, "step": 6673 }, { "epoch": 0.71, "learning_rate": 2.023606851635594e-06, "loss": 1.4669, "step": 6674 }, { "epoch": 0.71, "learning_rate": 2.0222189667839805e-06, "loss": 1.6471, "step": 6675 }, { "epoch": 0.71, "learning_rate": 2.020831437382037e-06, "loss": 1.6995, "step": 6676 }, { "epoch": 0.71, "learning_rate": 2.019444263595394e-06, "loss": 1.2534, "step": 6677 }, { "epoch": 0.71, "learning_rate": 2.01805744558963e-06, "loss": 1.1573, "step": 6678 }, { "epoch": 0.71, "learning_rate": 2.0166709835302883e-06, "loss": 1.326, "step": 6679 }, { "epoch": 0.71, "learning_rate": 2.015284877582868e-06, "loss": 1.2165, "step": 6680 }, { "epoch": 0.71, "learning_rate": 2.013899127912824e-06, "loss": 1.2948, "step": 6681 }, { "epoch": 0.71, "learning_rate": 2.012513734685567e-06, "loss": 1.4153, "step": 6682 }, { "epoch": 0.71, "learning_rate": 2.0111286980664703e-06, "loss": 1.3609, "step": 6683 }, { "epoch": 0.71, "learning_rate": 2.009744018220863e-06, "loss": 1.19, "step": 6684 }, { "epoch": 0.71, "learning_rate": 2.0083596953140266e-06, "loss": 1.4935, "step": 6685 }, { "epoch": 0.71, "learning_rate": 2.006975729511208e-06, "loss": 1.3843, "step": 6686 }, { "epoch": 0.71, "learning_rate": 2.005592120977606e-06, "loss": 1.5281, "step": 6687 }, { "epoch": 0.71, "learning_rate": 2.0042088698783767e-06, "loss": 1.3452, "step": 6688 }, { "epoch": 0.71, "learning_rate": 2.0028259763786363e-06, "loss": 1.4238, "step": 6689 }, { "epoch": 0.71, "learning_rate": 2.0014434406434584e-06, "loss": 1.4724, "step": 6690 }, { "epoch": 0.71, "learning_rate": 2.00006126283787e-06, "loss": 1.5934, "step": 6691 }, { "epoch": 0.71, "learning_rate": 1.9986794431268606e-06, "loss": 1.2505, "step": 6692 }, { "epoch": 0.71, "learning_rate": 1.9972979816753717e-06, "loss": 1.448, "step": 6693 }, { "epoch": 0.71, "learning_rate": 1.9959168786483074e-06, "loss": 1.7023, "step": 6694 }, { "epoch": 0.71, "learning_rate": 1.9945361342105225e-06, "loss": 1.4766, "step": 6695 }, { "epoch": 0.71, "learning_rate": 1.9931557485268365e-06, "loss": 1.4389, "step": 6696 }, { "epoch": 0.71, "learning_rate": 1.9917757217620193e-06, "loss": 1.4813, "step": 6697 }, { "epoch": 0.71, "learning_rate": 1.9903960540808026e-06, "loss": 1.6737, "step": 6698 }, { "epoch": 0.71, "learning_rate": 1.9890167456478748e-06, "loss": 1.1128, "step": 6699 }, { "epoch": 0.71, "learning_rate": 1.9876377966278786e-06, "loss": 1.118, "step": 6700 }, { "epoch": 0.71, "learning_rate": 1.9862592071854137e-06, "loss": 1.412, "step": 6701 }, { "epoch": 0.71, "learning_rate": 1.984880977485041e-06, "loss": 1.4125, "step": 6702 }, { "epoch": 0.71, "learning_rate": 1.983503107691277e-06, "loss": 1.4693, "step": 6703 }, { "epoch": 0.72, "learning_rate": 1.982125597968591e-06, "loss": 1.3416, "step": 6704 }, { "epoch": 0.72, "learning_rate": 1.980748448481416e-06, "loss": 1.205, "step": 6705 }, { "epoch": 0.72, "learning_rate": 1.9793716593941374e-06, "loss": 1.1455, "step": 6706 }, { "epoch": 0.72, "learning_rate": 1.977995230871095e-06, "loss": 1.2725, "step": 6707 }, { "epoch": 0.72, "learning_rate": 1.9766191630765964e-06, "loss": 1.5062, "step": 6708 }, { "epoch": 0.72, "learning_rate": 1.975243456174895e-06, "loss": 1.3046, "step": 6709 }, { "epoch": 0.72, "learning_rate": 1.9738681103302044e-06, "loss": 1.4855, "step": 6710 }, { "epoch": 0.72, "learning_rate": 1.9724931257066988e-06, "loss": 1.2448, "step": 6711 }, { "epoch": 0.72, "learning_rate": 1.9711185024685024e-06, "loss": 1.3918, "step": 6712 }, { "epoch": 0.72, "learning_rate": 1.9697442407797035e-06, "loss": 1.3223, "step": 6713 }, { "epoch": 0.72, "learning_rate": 1.9683703408043447e-06, "loss": 1.5395, "step": 6714 }, { "epoch": 0.72, "learning_rate": 1.9669968027064234e-06, "loss": 1.4886, "step": 6715 }, { "epoch": 0.72, "learning_rate": 1.965623626649893e-06, "loss": 1.3218, "step": 6716 }, { "epoch": 0.72, "learning_rate": 1.9642508127986676e-06, "loss": 1.549, "step": 6717 }, { "epoch": 0.72, "learning_rate": 1.9628783613166184e-06, "loss": 1.3719, "step": 6718 }, { "epoch": 0.72, "learning_rate": 1.961506272367567e-06, "loss": 1.2671, "step": 6719 }, { "epoch": 0.72, "learning_rate": 1.9601345461153005e-06, "loss": 1.3432, "step": 6720 }, { "epoch": 0.72, "learning_rate": 1.9587631827235557e-06, "loss": 1.2544, "step": 6721 }, { "epoch": 0.72, "learning_rate": 1.9573921823560273e-06, "loss": 1.4093, "step": 6722 }, { "epoch": 0.72, "learning_rate": 1.9560215451763685e-06, "loss": 1.4582, "step": 6723 }, { "epoch": 0.72, "learning_rate": 1.954651271348192e-06, "loss": 1.6758, "step": 6724 }, { "epoch": 0.72, "learning_rate": 1.9532813610350604e-06, "loss": 1.3716, "step": 6725 }, { "epoch": 0.72, "learning_rate": 1.951911814400495e-06, "loss": 1.6488, "step": 6726 }, { "epoch": 0.72, "learning_rate": 1.950542631607977e-06, "loss": 1.2395, "step": 6727 }, { "epoch": 0.72, "learning_rate": 1.9491738128209437e-06, "loss": 1.3765, "step": 6728 }, { "epoch": 0.72, "learning_rate": 1.9478053582027826e-06, "loss": 1.4444, "step": 6729 }, { "epoch": 0.72, "learning_rate": 1.9464372679168483e-06, "loss": 1.5532, "step": 6730 }, { "epoch": 0.72, "learning_rate": 1.945069542126442e-06, "loss": 1.3581, "step": 6731 }, { "epoch": 0.72, "learning_rate": 1.9437021809948232e-06, "loss": 1.3967, "step": 6732 }, { "epoch": 0.72, "learning_rate": 1.942335184685217e-06, "loss": 1.5038, "step": 6733 }, { "epoch": 0.72, "learning_rate": 1.9409685533607936e-06, "loss": 1.5339, "step": 6734 }, { "epoch": 0.72, "learning_rate": 1.9396022871846836e-06, "loss": 1.6689, "step": 6735 }, { "epoch": 0.72, "learning_rate": 1.9382363863199773e-06, "loss": 1.2648, "step": 6736 }, { "epoch": 0.72, "learning_rate": 1.936870850929715e-06, "loss": 1.3646, "step": 6737 }, { "epoch": 0.72, "learning_rate": 1.935505681176899e-06, "loss": 1.4934, "step": 6738 }, { "epoch": 0.72, "learning_rate": 1.934140877224488e-06, "loss": 1.4351, "step": 6739 }, { "epoch": 0.72, "learning_rate": 1.9327764392353927e-06, "loss": 1.4241, "step": 6740 }, { "epoch": 0.72, "learning_rate": 1.9314123673724805e-06, "loss": 1.4633, "step": 6741 }, { "epoch": 0.72, "learning_rate": 1.9300486617985792e-06, "loss": 1.5012, "step": 6742 }, { "epoch": 0.72, "learning_rate": 1.9286853226764725e-06, "loss": 1.4353, "step": 6743 }, { "epoch": 0.72, "learning_rate": 1.9273223501688943e-06, "loss": 1.3608, "step": 6744 }, { "epoch": 0.72, "learning_rate": 1.925959744438543e-06, "loss": 1.5955, "step": 6745 }, { "epoch": 0.72, "learning_rate": 1.9245975056480652e-06, "loss": 1.4775, "step": 6746 }, { "epoch": 0.72, "learning_rate": 1.9232356339600717e-06, "loss": 1.4439, "step": 6747 }, { "epoch": 0.72, "learning_rate": 1.9218741295371216e-06, "loss": 1.2753, "step": 6748 }, { "epoch": 0.72, "learning_rate": 1.9205129925417375e-06, "loss": 1.7544, "step": 6749 }, { "epoch": 0.72, "learning_rate": 1.919152223136391e-06, "loss": 1.3048, "step": 6750 }, { "epoch": 0.72, "learning_rate": 1.917791821483516e-06, "loss": 1.1974, "step": 6751 }, { "epoch": 0.72, "learning_rate": 1.9164317877455014e-06, "loss": 1.3396, "step": 6752 }, { "epoch": 0.72, "learning_rate": 1.9150721220846884e-06, "loss": 1.3174, "step": 6753 }, { "epoch": 0.72, "learning_rate": 1.913712824663375e-06, "loss": 1.753, "step": 6754 }, { "epoch": 0.72, "learning_rate": 1.912353895643821e-06, "loss": 1.4172, "step": 6755 }, { "epoch": 0.72, "learning_rate": 1.910995335188234e-06, "loss": 1.165, "step": 6756 }, { "epoch": 0.72, "learning_rate": 1.9096371434587836e-06, "loss": 1.4474, "step": 6757 }, { "epoch": 0.72, "learning_rate": 1.9082793206175954e-06, "loss": 1.3621, "step": 6758 }, { "epoch": 0.72, "learning_rate": 1.906921866826747e-06, "loss": 1.5309, "step": 6759 }, { "epoch": 0.72, "learning_rate": 1.9055647822482725e-06, "loss": 1.3545, "step": 6760 }, { "epoch": 0.72, "learning_rate": 1.9042080670441648e-06, "loss": 1.4163, "step": 6761 }, { "epoch": 0.72, "learning_rate": 1.9028517213763737e-06, "loss": 1.4618, "step": 6762 }, { "epoch": 0.72, "learning_rate": 1.9014957454067984e-06, "loss": 1.5056, "step": 6763 }, { "epoch": 0.72, "learning_rate": 1.9001401392973018e-06, "loss": 1.2313, "step": 6764 }, { "epoch": 0.72, "learning_rate": 1.8987849032096973e-06, "loss": 1.4746, "step": 6765 }, { "epoch": 0.72, "learning_rate": 1.8974300373057518e-06, "loss": 1.4571, "step": 6766 }, { "epoch": 0.72, "learning_rate": 1.8960755417471998e-06, "loss": 1.3519, "step": 6767 }, { "epoch": 0.72, "learning_rate": 1.89472141669572e-06, "loss": 1.5417, "step": 6768 }, { "epoch": 0.72, "learning_rate": 1.8933676623129481e-06, "loss": 1.4387, "step": 6769 }, { "epoch": 0.72, "learning_rate": 1.8920142787604829e-06, "loss": 1.465, "step": 6770 }, { "epoch": 0.72, "learning_rate": 1.8906612661998698e-06, "loss": 1.2523, "step": 6771 }, { "epoch": 0.72, "learning_rate": 1.8893086247926157e-06, "loss": 1.3539, "step": 6772 }, { "epoch": 0.72, "learning_rate": 1.8879563547001844e-06, "loss": 1.6603, "step": 6773 }, { "epoch": 0.72, "learning_rate": 1.8866044560839902e-06, "loss": 1.3246, "step": 6774 }, { "epoch": 0.72, "learning_rate": 1.8852529291054046e-06, "loss": 1.8482, "step": 6775 }, { "epoch": 0.72, "learning_rate": 1.883901773925757e-06, "loss": 1.5602, "step": 6776 }, { "epoch": 0.72, "learning_rate": 1.8825509907063328e-06, "loss": 1.7339, "step": 6777 }, { "epoch": 0.72, "learning_rate": 1.88120057960837e-06, "loss": 1.3364, "step": 6778 }, { "epoch": 0.72, "learning_rate": 1.8798505407930612e-06, "loss": 1.7943, "step": 6779 }, { "epoch": 0.72, "learning_rate": 1.8785008744215606e-06, "loss": 1.6393, "step": 6780 }, { "epoch": 0.72, "learning_rate": 1.877151580654971e-06, "loss": 1.5412, "step": 6781 }, { "epoch": 0.72, "learning_rate": 1.8758026596543555e-06, "loss": 1.508, "step": 6782 }, { "epoch": 0.72, "learning_rate": 1.874454111580733e-06, "loss": 1.3918, "step": 6783 }, { "epoch": 0.72, "learning_rate": 1.8731059365950737e-06, "loss": 1.1217, "step": 6784 }, { "epoch": 0.72, "learning_rate": 1.8717581348583052e-06, "loss": 1.5202, "step": 6785 }, { "epoch": 0.72, "learning_rate": 1.8704107065313116e-06, "loss": 1.4115, "step": 6786 }, { "epoch": 0.72, "learning_rate": 1.8690636517749333e-06, "loss": 1.186, "step": 6787 }, { "epoch": 0.72, "learning_rate": 1.867716970749962e-06, "loss": 1.398, "step": 6788 }, { "epoch": 0.72, "learning_rate": 1.8663706636171503e-06, "loss": 1.436, "step": 6789 }, { "epoch": 0.72, "learning_rate": 1.8650247305371998e-06, "loss": 1.4752, "step": 6790 }, { "epoch": 0.72, "learning_rate": 1.863679171670773e-06, "loss": 1.4639, "step": 6791 }, { "epoch": 0.72, "learning_rate": 1.8623339871784869e-06, "loss": 1.4062, "step": 6792 }, { "epoch": 0.72, "learning_rate": 1.860989177220911e-06, "loss": 1.2669, "step": 6793 }, { "epoch": 0.72, "learning_rate": 1.85964474195857e-06, "loss": 1.3645, "step": 6794 }, { "epoch": 0.72, "learning_rate": 1.8583006815519473e-06, "loss": 1.431, "step": 6795 }, { "epoch": 0.72, "learning_rate": 1.856956996161481e-06, "loss": 1.3323, "step": 6796 }, { "epoch": 0.73, "learning_rate": 1.85561368594756e-06, "loss": 1.6654, "step": 6797 }, { "epoch": 0.73, "learning_rate": 1.8542707510705355e-06, "loss": 1.4534, "step": 6798 }, { "epoch": 0.73, "learning_rate": 1.852928191690707e-06, "loss": 1.3944, "step": 6799 }, { "epoch": 0.73, "learning_rate": 1.8515860079683322e-06, "loss": 1.3946, "step": 6800 }, { "epoch": 0.73, "learning_rate": 1.8502442000636246e-06, "loss": 1.4271, "step": 6801 }, { "epoch": 0.73, "learning_rate": 1.8489027681367544e-06, "loss": 1.3099, "step": 6802 }, { "epoch": 0.73, "learning_rate": 1.8475617123478406e-06, "loss": 1.5822, "step": 6803 }, { "epoch": 0.73, "learning_rate": 1.846221032856965e-06, "loss": 1.4232, "step": 6804 }, { "epoch": 0.73, "learning_rate": 1.8448807298241584e-06, "loss": 1.1965, "step": 6805 }, { "epoch": 0.73, "learning_rate": 1.8435408034094116e-06, "loss": 1.4105, "step": 6806 }, { "epoch": 0.73, "learning_rate": 1.8422012537726646e-06, "loss": 1.4985, "step": 6807 }, { "epoch": 0.73, "learning_rate": 1.8408620810738191e-06, "loss": 1.4225, "step": 6808 }, { "epoch": 0.73, "learning_rate": 1.8395232854727263e-06, "loss": 1.3446, "step": 6809 }, { "epoch": 0.73, "learning_rate": 1.8381848671291953e-06, "loss": 1.3548, "step": 6810 }, { "epoch": 0.73, "learning_rate": 1.836846826202991e-06, "loss": 1.2948, "step": 6811 }, { "epoch": 0.73, "learning_rate": 1.8355091628538308e-06, "loss": 1.4301, "step": 6812 }, { "epoch": 0.73, "learning_rate": 1.8341718772413852e-06, "loss": 1.2601, "step": 6813 }, { "epoch": 0.73, "learning_rate": 1.8328349695252872e-06, "loss": 1.428, "step": 6814 }, { "epoch": 0.73, "learning_rate": 1.8314984398651153e-06, "loss": 1.2134, "step": 6815 }, { "epoch": 0.73, "learning_rate": 1.8301622884204096e-06, "loss": 1.8912, "step": 6816 }, { "epoch": 0.73, "learning_rate": 1.8288265153506646e-06, "loss": 1.3855, "step": 6817 }, { "epoch": 0.73, "learning_rate": 1.8274911208153262e-06, "loss": 1.3822, "step": 6818 }, { "epoch": 0.73, "learning_rate": 1.8261561049737946e-06, "loss": 1.5072, "step": 6819 }, { "epoch": 0.73, "learning_rate": 1.8248214679854298e-06, "loss": 1.279, "step": 6820 }, { "epoch": 0.73, "learning_rate": 1.8234872100095445e-06, "loss": 1.5973, "step": 6821 }, { "epoch": 0.73, "learning_rate": 1.8221533312054024e-06, "loss": 1.3642, "step": 6822 }, { "epoch": 0.73, "learning_rate": 1.820819831732229e-06, "loss": 1.4926, "step": 6823 }, { "epoch": 0.73, "learning_rate": 1.8194867117491978e-06, "loss": 1.516, "step": 6824 }, { "epoch": 0.73, "learning_rate": 1.818153971415439e-06, "loss": 1.458, "step": 6825 }, { "epoch": 0.73, "learning_rate": 1.8168216108900389e-06, "loss": 1.3448, "step": 6826 }, { "epoch": 0.73, "learning_rate": 1.8154896303320402e-06, "loss": 1.2837, "step": 6827 }, { "epoch": 0.73, "learning_rate": 1.8141580299004342e-06, "loss": 1.3961, "step": 6828 }, { "epoch": 0.73, "learning_rate": 1.812826809754174e-06, "loss": 1.409, "step": 6829 }, { "epoch": 0.73, "learning_rate": 1.8114959700521606e-06, "loss": 1.3368, "step": 6830 }, { "epoch": 0.73, "learning_rate": 1.8101655109532552e-06, "loss": 1.419, "step": 6831 }, { "epoch": 0.73, "learning_rate": 1.8088354326162683e-06, "loss": 1.2007, "step": 6832 }, { "epoch": 0.73, "learning_rate": 1.8075057351999714e-06, "loss": 1.3972, "step": 6833 }, { "epoch": 0.73, "learning_rate": 1.8061764188630831e-06, "loss": 1.272, "step": 6834 }, { "epoch": 0.73, "learning_rate": 1.8048474837642821e-06, "loss": 1.2428, "step": 6835 }, { "epoch": 0.73, "learning_rate": 1.8035189300622019e-06, "loss": 1.4476, "step": 6836 }, { "epoch": 0.73, "learning_rate": 1.8021907579154257e-06, "loss": 1.4051, "step": 6837 }, { "epoch": 0.73, "learning_rate": 1.8008629674824928e-06, "loss": 1.3021, "step": 6838 }, { "epoch": 0.73, "learning_rate": 1.7995355589219e-06, "loss": 1.5188, "step": 6839 }, { "epoch": 0.73, "learning_rate": 1.7982085323920973e-06, "loss": 1.2618, "step": 6840 }, { "epoch": 0.73, "learning_rate": 1.7968818880514855e-06, "loss": 1.4604, "step": 6841 }, { "epoch": 0.73, "learning_rate": 1.7955556260584256e-06, "loss": 1.8058, "step": 6842 }, { "epoch": 0.73, "learning_rate": 1.7942297465712282e-06, "loss": 0.9693, "step": 6843 }, { "epoch": 0.73, "learning_rate": 1.792904249748157e-06, "loss": 1.5956, "step": 6844 }, { "epoch": 0.73, "learning_rate": 1.7915791357474394e-06, "loss": 1.294, "step": 6845 }, { "epoch": 0.73, "learning_rate": 1.7902544047272468e-06, "loss": 1.2903, "step": 6846 }, { "epoch": 0.73, "learning_rate": 1.7889300568457085e-06, "loss": 1.6581, "step": 6847 }, { "epoch": 0.73, "learning_rate": 1.78760609226091e-06, "loss": 1.4609, "step": 6848 }, { "epoch": 0.73, "learning_rate": 1.7862825111308873e-06, "loss": 1.2364, "step": 6849 }, { "epoch": 0.73, "learning_rate": 1.784959313613634e-06, "loss": 1.6876, "step": 6850 }, { "epoch": 0.73, "learning_rate": 1.783636499867098e-06, "loss": 1.4381, "step": 6851 }, { "epoch": 0.73, "learning_rate": 1.7823140700491786e-06, "loss": 1.5563, "step": 6852 }, { "epoch": 0.73, "learning_rate": 1.7809920243177287e-06, "loss": 1.3178, "step": 6853 }, { "epoch": 0.73, "learning_rate": 1.7796703628305596e-06, "loss": 1.6176, "step": 6854 }, { "epoch": 0.73, "learning_rate": 1.7783490857454354e-06, "loss": 1.4376, "step": 6855 }, { "epoch": 0.73, "learning_rate": 1.7770281932200706e-06, "loss": 1.5884, "step": 6856 }, { "epoch": 0.73, "learning_rate": 1.7757076854121396e-06, "loss": 1.3881, "step": 6857 }, { "epoch": 0.73, "learning_rate": 1.7743875624792662e-06, "loss": 1.5353, "step": 6858 }, { "epoch": 0.73, "learning_rate": 1.7730678245790283e-06, "loss": 1.4335, "step": 6859 }, { "epoch": 0.73, "learning_rate": 1.771748471868961e-06, "loss": 1.5233, "step": 6860 }, { "epoch": 0.73, "learning_rate": 1.770429504506554e-06, "loss": 1.2787, "step": 6861 }, { "epoch": 0.73, "learning_rate": 1.7691109226492448e-06, "loss": 1.3128, "step": 6862 }, { "epoch": 0.73, "learning_rate": 1.767792726454433e-06, "loss": 1.3426, "step": 6863 }, { "epoch": 0.73, "learning_rate": 1.7664749160794642e-06, "loss": 1.6759, "step": 6864 }, { "epoch": 0.73, "learning_rate": 1.765157491681646e-06, "loss": 1.3566, "step": 6865 }, { "epoch": 0.73, "learning_rate": 1.7638404534182313e-06, "loss": 1.4636, "step": 6866 }, { "epoch": 0.73, "learning_rate": 1.7625238014464358e-06, "loss": 1.4756, "step": 6867 }, { "epoch": 0.73, "learning_rate": 1.761207535923422e-06, "loss": 1.3305, "step": 6868 }, { "epoch": 0.73, "learning_rate": 1.7598916570063064e-06, "loss": 1.5733, "step": 6869 }, { "epoch": 0.73, "learning_rate": 1.7585761648521688e-06, "loss": 1.5087, "step": 6870 }, { "epoch": 0.73, "learning_rate": 1.7572610596180316e-06, "loss": 1.2412, "step": 6871 }, { "epoch": 0.73, "learning_rate": 1.7559463414608747e-06, "loss": 1.4458, "step": 6872 }, { "epoch": 0.73, "learning_rate": 1.7546320105376346e-06, "loss": 1.4523, "step": 6873 }, { "epoch": 0.73, "learning_rate": 1.7533180670051974e-06, "loss": 1.4105, "step": 6874 }, { "epoch": 0.73, "learning_rate": 1.752004511020406e-06, "loss": 1.2088, "step": 6875 }, { "epoch": 0.73, "learning_rate": 1.750691342740058e-06, "loss": 1.434, "step": 6876 }, { "epoch": 0.73, "learning_rate": 1.7493785623209009e-06, "loss": 1.2686, "step": 6877 }, { "epoch": 0.73, "learning_rate": 1.7480661699196356e-06, "loss": 1.4895, "step": 6878 }, { "epoch": 0.73, "learning_rate": 1.746754165692921e-06, "loss": 1.3874, "step": 6879 }, { "epoch": 0.73, "learning_rate": 1.7454425497973692e-06, "loss": 1.5251, "step": 6880 }, { "epoch": 0.73, "learning_rate": 1.7441313223895412e-06, "loss": 1.4745, "step": 6881 }, { "epoch": 0.73, "learning_rate": 1.742820483625957e-06, "loss": 1.7703, "step": 6882 }, { "epoch": 0.73, "learning_rate": 1.7415100336630858e-06, "loss": 1.2364, "step": 6883 }, { "epoch": 0.73, "learning_rate": 1.740199972657355e-06, "loss": 1.1244, "step": 6884 }, { "epoch": 0.73, "learning_rate": 1.7388903007651398e-06, "loss": 1.2404, "step": 6885 }, { "epoch": 0.73, "learning_rate": 1.7375810181427755e-06, "loss": 1.2685, "step": 6886 }, { "epoch": 0.73, "learning_rate": 1.7362721249465448e-06, "loss": 1.6111, "step": 6887 }, { "epoch": 0.73, "learning_rate": 1.7349636213326876e-06, "loss": 1.4167, "step": 6888 }, { "epoch": 0.73, "learning_rate": 1.7336555074573985e-06, "loss": 1.4168, "step": 6889 }, { "epoch": 0.73, "learning_rate": 1.732347783476822e-06, "loss": 1.285, "step": 6890 }, { "epoch": 0.74, "learning_rate": 1.7310404495470557e-06, "loss": 1.3441, "step": 6891 }, { "epoch": 0.74, "learning_rate": 1.729733505824156e-06, "loss": 1.3693, "step": 6892 }, { "epoch": 0.74, "learning_rate": 1.7284269524641262e-06, "loss": 1.1964, "step": 6893 }, { "epoch": 0.74, "learning_rate": 1.727120789622927e-06, "loss": 1.3767, "step": 6894 }, { "epoch": 0.74, "learning_rate": 1.7258150174564737e-06, "loss": 1.4671, "step": 6895 }, { "epoch": 0.74, "learning_rate": 1.7245096361206304e-06, "loss": 1.237, "step": 6896 }, { "epoch": 0.74, "learning_rate": 1.7232046457712164e-06, "loss": 1.5424, "step": 6897 }, { "epoch": 0.74, "learning_rate": 1.7219000465640057e-06, "loss": 1.4662, "step": 6898 }, { "epoch": 0.74, "learning_rate": 1.7205958386547273e-06, "loss": 1.3027, "step": 6899 }, { "epoch": 0.74, "learning_rate": 1.7192920221990566e-06, "loss": 1.1179, "step": 6900 }, { "epoch": 0.74, "learning_rate": 1.7179885973526305e-06, "loss": 1.5994, "step": 6901 }, { "epoch": 0.74, "learning_rate": 1.7166855642710333e-06, "loss": 1.5475, "step": 6902 }, { "epoch": 0.74, "learning_rate": 1.7153829231098018e-06, "loss": 1.6144, "step": 6903 }, { "epoch": 0.74, "learning_rate": 1.7140806740244354e-06, "loss": 1.5535, "step": 6904 }, { "epoch": 0.74, "learning_rate": 1.7127788171703757e-06, "loss": 1.1083, "step": 6905 }, { "epoch": 0.74, "learning_rate": 1.7114773527030215e-06, "loss": 1.3837, "step": 6906 }, { "epoch": 0.74, "learning_rate": 1.7101762807777277e-06, "loss": 1.3933, "step": 6907 }, { "epoch": 0.74, "learning_rate": 1.7088756015497966e-06, "loss": 1.3416, "step": 6908 }, { "epoch": 0.74, "learning_rate": 1.7075753151744885e-06, "loss": 1.4, "step": 6909 }, { "epoch": 0.74, "learning_rate": 1.706275421807016e-06, "loss": 1.6508, "step": 6910 }, { "epoch": 0.74, "learning_rate": 1.704975921602543e-06, "loss": 1.5149, "step": 6911 }, { "epoch": 0.74, "learning_rate": 1.7036768147161853e-06, "loss": 1.3721, "step": 6912 }, { "epoch": 0.74, "learning_rate": 1.7023781013030162e-06, "loss": 1.5128, "step": 6913 }, { "epoch": 0.74, "learning_rate": 1.70107978151806e-06, "loss": 1.6769, "step": 6914 }, { "epoch": 0.74, "learning_rate": 1.6997818555162915e-06, "loss": 1.6182, "step": 6915 }, { "epoch": 0.74, "learning_rate": 1.6984843234526433e-06, "loss": 1.5461, "step": 6916 }, { "epoch": 0.74, "learning_rate": 1.697187185481996e-06, "loss": 1.3861, "step": 6917 }, { "epoch": 0.74, "learning_rate": 1.6958904417591853e-06, "loss": 1.4213, "step": 6918 }, { "epoch": 0.74, "learning_rate": 1.6945940924390009e-06, "loss": 1.3683, "step": 6919 }, { "epoch": 0.74, "learning_rate": 1.6932981376761858e-06, "loss": 1.3479, "step": 6920 }, { "epoch": 0.74, "learning_rate": 1.6920025776254334e-06, "loss": 1.1463, "step": 6921 }, { "epoch": 0.74, "learning_rate": 1.690707412441388e-06, "loss": 1.267, "step": 6922 }, { "epoch": 0.74, "learning_rate": 1.689412642278656e-06, "loss": 1.2278, "step": 6923 }, { "epoch": 0.74, "learning_rate": 1.6881182672917879e-06, "loss": 1.2559, "step": 6924 }, { "epoch": 0.74, "learning_rate": 1.686824287635288e-06, "loss": 1.3709, "step": 6925 }, { "epoch": 0.74, "learning_rate": 1.685530703463618e-06, "loss": 1.1159, "step": 6926 }, { "epoch": 0.74, "learning_rate": 1.6842375149311868e-06, "loss": 1.4748, "step": 6927 }, { "epoch": 0.74, "learning_rate": 1.6829447221923606e-06, "loss": 1.7714, "step": 6928 }, { "epoch": 0.74, "learning_rate": 1.6816523254014577e-06, "loss": 1.584, "step": 6929 }, { "epoch": 0.74, "learning_rate": 1.680360324712746e-06, "loss": 1.3788, "step": 6930 }, { "epoch": 0.74, "learning_rate": 1.6790687202804473e-06, "loss": 1.1828, "step": 6931 }, { "epoch": 0.74, "learning_rate": 1.6777775122587387e-06, "loss": 1.3692, "step": 6932 }, { "epoch": 0.74, "learning_rate": 1.6764867008017493e-06, "loss": 1.6622, "step": 6933 }, { "epoch": 0.74, "learning_rate": 1.6751962860635562e-06, "loss": 1.4246, "step": 6934 }, { "epoch": 0.74, "learning_rate": 1.673906268198197e-06, "loss": 1.4411, "step": 6935 }, { "epoch": 0.74, "learning_rate": 1.672616647359655e-06, "loss": 1.0712, "step": 6936 }, { "epoch": 0.74, "learning_rate": 1.6713274237018679e-06, "loss": 1.3669, "step": 6937 }, { "epoch": 0.74, "learning_rate": 1.6700385973787282e-06, "loss": 1.3234, "step": 6938 }, { "epoch": 0.74, "learning_rate": 1.668750168544081e-06, "loss": 1.7028, "step": 6939 }, { "epoch": 0.74, "learning_rate": 1.667462137351719e-06, "loss": 1.3133, "step": 6940 }, { "epoch": 0.74, "learning_rate": 1.6661745039553955e-06, "loss": 1.3493, "step": 6941 }, { "epoch": 0.74, "learning_rate": 1.664887268508808e-06, "loss": 1.2844, "step": 6942 }, { "epoch": 0.74, "learning_rate": 1.663600431165613e-06, "loss": 1.5219, "step": 6943 }, { "epoch": 0.74, "learning_rate": 1.6623139920794135e-06, "loss": 1.3717, "step": 6944 }, { "epoch": 0.74, "learning_rate": 1.6610279514037725e-06, "loss": 1.2788, "step": 6945 }, { "epoch": 0.74, "learning_rate": 1.6597423092921972e-06, "loss": 1.2154, "step": 6946 }, { "epoch": 0.74, "learning_rate": 1.658457065898153e-06, "loss": 1.2587, "step": 6947 }, { "epoch": 0.74, "learning_rate": 1.657172221375058e-06, "loss": 1.5414, "step": 6948 }, { "epoch": 0.74, "learning_rate": 1.6558877758762782e-06, "loss": 1.3323, "step": 6949 }, { "epoch": 0.74, "learning_rate": 1.6546037295551332e-06, "loss": 1.3803, "step": 6950 }, { "epoch": 0.74, "learning_rate": 1.6533200825648993e-06, "loss": 1.4595, "step": 6951 }, { "epoch": 0.74, "learning_rate": 1.6520368350587984e-06, "loss": 1.3616, "step": 6952 }, { "epoch": 0.74, "learning_rate": 1.6507539871900109e-06, "loss": 1.1834, "step": 6953 }, { "epoch": 0.74, "learning_rate": 1.6494715391116671e-06, "loss": 1.4254, "step": 6954 }, { "epoch": 0.74, "learning_rate": 1.6481894909768492e-06, "loss": 1.5033, "step": 6955 }, { "epoch": 0.74, "learning_rate": 1.6469078429385893e-06, "loss": 1.411, "step": 6956 }, { "epoch": 0.74, "learning_rate": 1.6456265951498763e-06, "loss": 1.2679, "step": 6957 }, { "epoch": 0.74, "learning_rate": 1.6443457477636505e-06, "loss": 1.7648, "step": 6958 }, { "epoch": 0.74, "learning_rate": 1.6430653009328002e-06, "loss": 1.3581, "step": 6959 }, { "epoch": 0.74, "learning_rate": 1.641785254810172e-06, "loss": 1.5448, "step": 6960 }, { "epoch": 0.74, "learning_rate": 1.64050560954856e-06, "loss": 1.269, "step": 6961 }, { "epoch": 0.74, "learning_rate": 1.639226365300709e-06, "loss": 1.6637, "step": 6962 }, { "epoch": 0.74, "learning_rate": 1.6379475222193248e-06, "loss": 1.3893, "step": 6963 }, { "epoch": 0.74, "learning_rate": 1.6366690804570567e-06, "loss": 1.5416, "step": 6964 }, { "epoch": 0.74, "learning_rate": 1.635391040166507e-06, "loss": 1.5046, "step": 6965 }, { "epoch": 0.74, "learning_rate": 1.6341134015002352e-06, "loss": 1.3366, "step": 6966 }, { "epoch": 0.74, "learning_rate": 1.6328361646107465e-06, "loss": 1.1849, "step": 6967 }, { "epoch": 0.74, "learning_rate": 1.6315593296505039e-06, "loss": 1.2954, "step": 6968 }, { "epoch": 0.74, "learning_rate": 1.6302828967719175e-06, "loss": 1.401, "step": 6969 }, { "epoch": 0.74, "learning_rate": 1.6290068661273539e-06, "loss": 1.2008, "step": 6970 }, { "epoch": 0.74, "learning_rate": 1.6277312378691268e-06, "loss": 1.2845, "step": 6971 }, { "epoch": 0.74, "learning_rate": 1.626456012149506e-06, "loss": 1.3017, "step": 6972 }, { "epoch": 0.74, "learning_rate": 1.6251811891207132e-06, "loss": 1.5447, "step": 6973 }, { "epoch": 0.74, "learning_rate": 1.6239067689349186e-06, "loss": 1.2356, "step": 6974 }, { "epoch": 0.74, "learning_rate": 1.6226327517442453e-06, "loss": 1.4431, "step": 6975 }, { "epoch": 0.74, "learning_rate": 1.6213591377007708e-06, "loss": 1.4634, "step": 6976 }, { "epoch": 0.74, "learning_rate": 1.6200859269565244e-06, "loss": 1.1321, "step": 6977 }, { "epoch": 0.74, "learning_rate": 1.6188131196634827e-06, "loss": 1.28, "step": 6978 }, { "epoch": 0.74, "learning_rate": 1.6175407159735807e-06, "loss": 1.3813, "step": 6979 }, { "epoch": 0.74, "learning_rate": 1.6162687160386998e-06, "loss": 1.3646, "step": 6980 }, { "epoch": 0.74, "learning_rate": 1.6149971200106723e-06, "loss": 1.2814, "step": 6981 }, { "epoch": 0.74, "learning_rate": 1.6137259280412909e-06, "loss": 1.4752, "step": 6982 }, { "epoch": 0.74, "learning_rate": 1.6124551402822925e-06, "loss": 1.572, "step": 6983 }, { "epoch": 0.74, "learning_rate": 1.6111847568853645e-06, "loss": 1.6072, "step": 6984 }, { "epoch": 0.75, "learning_rate": 1.6099147780021534e-06, "loss": 1.3221, "step": 6985 }, { "epoch": 0.75, "learning_rate": 1.6086452037842493e-06, "loss": 1.524, "step": 6986 }, { "epoch": 0.75, "learning_rate": 1.6073760343831996e-06, "loss": 1.4439, "step": 6987 }, { "epoch": 0.75, "learning_rate": 1.6061072699505037e-06, "loss": 1.3567, "step": 6988 }, { "epoch": 0.75, "learning_rate": 1.6048389106376083e-06, "loss": 1.4668, "step": 6989 }, { "epoch": 0.75, "learning_rate": 1.603570956595913e-06, "loss": 1.2877, "step": 6990 }, { "epoch": 0.75, "learning_rate": 1.6023034079767707e-06, "loss": 1.2566, "step": 6991 }, { "epoch": 0.75, "learning_rate": 1.6010362649314881e-06, "loss": 1.4029, "step": 6992 }, { "epoch": 0.75, "learning_rate": 1.5997695276113168e-06, "loss": 1.576, "step": 6993 }, { "epoch": 0.75, "learning_rate": 1.5985031961674669e-06, "loss": 1.5096, "step": 6994 }, { "epoch": 0.75, "learning_rate": 1.597237270751096e-06, "loss": 1.5909, "step": 6995 }, { "epoch": 0.75, "learning_rate": 1.595971751513311e-06, "loss": 1.3731, "step": 6996 }, { "epoch": 0.75, "learning_rate": 1.5947066386051774e-06, "loss": 1.3487, "step": 6997 }, { "epoch": 0.75, "learning_rate": 1.5934419321777083e-06, "loss": 1.3762, "step": 6998 }, { "epoch": 0.75, "learning_rate": 1.5921776323818655e-06, "loss": 1.4951, "step": 6999 }, { "epoch": 0.75, "learning_rate": 1.590913739368568e-06, "loss": 1.3331, "step": 7000 }, { "epoch": 0.75, "learning_rate": 1.589650253288681e-06, "loss": 1.447, "step": 7001 }, { "epoch": 0.75, "learning_rate": 1.5883871742930257e-06, "loss": 1.4718, "step": 7002 }, { "epoch": 0.75, "learning_rate": 1.5871245025323695e-06, "loss": 1.4056, "step": 7003 }, { "epoch": 0.75, "learning_rate": 1.5858622381574373e-06, "loss": 1.3192, "step": 7004 }, { "epoch": 0.75, "learning_rate": 1.5846003813188993e-06, "loss": 1.3836, "step": 7005 }, { "epoch": 0.75, "learning_rate": 1.5833389321673808e-06, "loss": 1.5162, "step": 7006 }, { "epoch": 0.75, "learning_rate": 1.5820778908534595e-06, "loss": 1.5624, "step": 7007 }, { "epoch": 0.75, "learning_rate": 1.5808172575276615e-06, "loss": 1.2956, "step": 7008 }, { "epoch": 0.75, "learning_rate": 1.579557032340463e-06, "loss": 1.6152, "step": 7009 }, { "epoch": 0.75, "learning_rate": 1.578297215442297e-06, "loss": 1.4824, "step": 7010 }, { "epoch": 0.75, "learning_rate": 1.5770378069835412e-06, "loss": 1.3595, "step": 7011 }, { "epoch": 0.75, "learning_rate": 1.5757788071145291e-06, "loss": 1.4706, "step": 7012 }, { "epoch": 0.75, "learning_rate": 1.5745202159855466e-06, "loss": 1.2508, "step": 7013 }, { "epoch": 0.75, "learning_rate": 1.5732620337468258e-06, "loss": 1.3693, "step": 7014 }, { "epoch": 0.75, "learning_rate": 1.5720042605485497e-06, "loss": 1.5095, "step": 7015 }, { "epoch": 0.75, "learning_rate": 1.5707468965408618e-06, "loss": 1.4346, "step": 7016 }, { "epoch": 0.75, "learning_rate": 1.5694899418738462e-06, "loss": 1.3357, "step": 7017 }, { "epoch": 0.75, "learning_rate": 1.5682333966975421e-06, "loss": 1.6528, "step": 7018 }, { "epoch": 0.75, "learning_rate": 1.566977261161942e-06, "loss": 1.2368, "step": 7019 }, { "epoch": 0.75, "learning_rate": 1.5657215354169841e-06, "loss": 1.3007, "step": 7020 }, { "epoch": 0.75, "learning_rate": 1.5644662196125649e-06, "loss": 1.3174, "step": 7021 }, { "epoch": 0.75, "learning_rate": 1.5632113138985245e-06, "loss": 1.2008, "step": 7022 }, { "epoch": 0.75, "learning_rate": 1.561956818424661e-06, "loss": 1.4462, "step": 7023 }, { "epoch": 0.75, "learning_rate": 1.5607027333407165e-06, "loss": 1.6478, "step": 7024 }, { "epoch": 0.75, "learning_rate": 1.5594490587963896e-06, "loss": 1.5206, "step": 7025 }, { "epoch": 0.75, "learning_rate": 1.5581957949413295e-06, "loss": 1.4585, "step": 7026 }, { "epoch": 0.75, "learning_rate": 1.5569429419251337e-06, "loss": 1.413, "step": 7027 }, { "epoch": 0.75, "learning_rate": 1.5556904998973498e-06, "loss": 1.3889, "step": 7028 }, { "epoch": 0.75, "learning_rate": 1.554438469007482e-06, "loss": 1.345, "step": 7029 }, { "epoch": 0.75, "learning_rate": 1.553186849404979e-06, "loss": 1.3201, "step": 7030 }, { "epoch": 0.75, "learning_rate": 1.5519356412392433e-06, "loss": 1.4615, "step": 7031 }, { "epoch": 0.75, "learning_rate": 1.5506848446596317e-06, "loss": 1.322, "step": 7032 }, { "epoch": 0.75, "learning_rate": 1.549434459815446e-06, "loss": 1.3517, "step": 7033 }, { "epoch": 0.75, "learning_rate": 1.5481844868559392e-06, "loss": 1.4158, "step": 7034 }, { "epoch": 0.75, "learning_rate": 1.546934925930319e-06, "loss": 1.3131, "step": 7035 }, { "epoch": 0.75, "learning_rate": 1.5456857771877443e-06, "loss": 1.5768, "step": 7036 }, { "epoch": 0.75, "learning_rate": 1.544437040777319e-06, "loss": 1.5786, "step": 7037 }, { "epoch": 0.75, "learning_rate": 1.5431887168481051e-06, "loss": 1.463, "step": 7038 }, { "epoch": 0.75, "learning_rate": 1.5419408055491091e-06, "loss": 1.4338, "step": 7039 }, { "epoch": 0.75, "learning_rate": 1.5406933070292884e-06, "loss": 1.3111, "step": 7040 }, { "epoch": 0.75, "learning_rate": 1.5394462214375593e-06, "loss": 1.0997, "step": 7041 }, { "epoch": 0.75, "learning_rate": 1.5381995489227801e-06, "loss": 1.5206, "step": 7042 }, { "epoch": 0.75, "learning_rate": 1.5369532896337614e-06, "loss": 1.4868, "step": 7043 }, { "epoch": 0.75, "learning_rate": 1.5357074437192688e-06, "loss": 1.4283, "step": 7044 }, { "epoch": 0.75, "learning_rate": 1.5344620113280123e-06, "loss": 1.2468, "step": 7045 }, { "epoch": 0.75, "learning_rate": 1.5332169926086576e-06, "loss": 1.373, "step": 7046 }, { "epoch": 0.75, "learning_rate": 1.5319723877098202e-06, "loss": 1.5382, "step": 7047 }, { "epoch": 0.75, "learning_rate": 1.5307281967800646e-06, "loss": 1.2787, "step": 7048 }, { "epoch": 0.75, "learning_rate": 1.5294844199679044e-06, "loss": 1.5792, "step": 7049 }, { "epoch": 0.75, "learning_rate": 1.5282410574218072e-06, "loss": 1.2624, "step": 7050 }, { "epoch": 0.75, "learning_rate": 1.526998109290192e-06, "loss": 1.1876, "step": 7051 }, { "epoch": 0.75, "learning_rate": 1.5257555757214222e-06, "loss": 1.3889, "step": 7052 }, { "epoch": 0.75, "learning_rate": 1.5245134568638197e-06, "loss": 1.5046, "step": 7053 }, { "epoch": 0.75, "learning_rate": 1.5232717528656504e-06, "loss": 1.4018, "step": 7054 }, { "epoch": 0.75, "learning_rate": 1.5220304638751327e-06, "loss": 1.4319, "step": 7055 }, { "epoch": 0.75, "learning_rate": 1.5207895900404363e-06, "loss": 1.267, "step": 7056 }, { "epoch": 0.75, "learning_rate": 1.5195491315096833e-06, "loss": 1.2151, "step": 7057 }, { "epoch": 0.75, "learning_rate": 1.518309088430941e-06, "loss": 1.4817, "step": 7058 }, { "epoch": 0.75, "learning_rate": 1.5170694609522306e-06, "loss": 1.731, "step": 7059 }, { "epoch": 0.75, "learning_rate": 1.5158302492215248e-06, "loss": 1.2223, "step": 7060 }, { "epoch": 0.75, "learning_rate": 1.5145914533867445e-06, "loss": 1.3082, "step": 7061 }, { "epoch": 0.75, "learning_rate": 1.5133530735957586e-06, "loss": 1.4598, "step": 7062 }, { "epoch": 0.75, "learning_rate": 1.5121151099963928e-06, "loss": 1.3982, "step": 7063 }, { "epoch": 0.75, "learning_rate": 1.5108775627364163e-06, "loss": 1.5048, "step": 7064 }, { "epoch": 0.75, "learning_rate": 1.5096404319635533e-06, "loss": 1.4016, "step": 7065 }, { "epoch": 0.75, "learning_rate": 1.5084037178254784e-06, "loss": 1.5785, "step": 7066 }, { "epoch": 0.75, "learning_rate": 1.5071674204698133e-06, "loss": 1.6368, "step": 7067 }, { "epoch": 0.75, "learning_rate": 1.50593154004413e-06, "loss": 1.4964, "step": 7068 }, { "epoch": 0.75, "learning_rate": 1.504696076695953e-06, "loss": 1.5766, "step": 7069 }, { "epoch": 0.75, "learning_rate": 1.5034610305727588e-06, "loss": 1.3472, "step": 7070 }, { "epoch": 0.75, "learning_rate": 1.502226401821968e-06, "loss": 1.2086, "step": 7071 }, { "epoch": 0.75, "learning_rate": 1.5009921905909575e-06, "loss": 1.502, "step": 7072 }, { "epoch": 0.75, "learning_rate": 1.4997583970270508e-06, "loss": 1.5115, "step": 7073 }, { "epoch": 0.75, "learning_rate": 1.498525021277521e-06, "loss": 1.3151, "step": 7074 }, { "epoch": 0.75, "learning_rate": 1.4972920634895943e-06, "loss": 1.2996, "step": 7075 }, { "epoch": 0.75, "learning_rate": 1.496059523810447e-06, "loss": 1.5511, "step": 7076 }, { "epoch": 0.75, "learning_rate": 1.4948274023872005e-06, "loss": 1.6557, "step": 7077 }, { "epoch": 0.75, "learning_rate": 1.4935956993669338e-06, "loss": 1.3111, "step": 7078 }, { "epoch": 0.76, "learning_rate": 1.4923644148966682e-06, "loss": 1.3486, "step": 7079 }, { "epoch": 0.76, "learning_rate": 1.4911335491233818e-06, "loss": 1.2254, "step": 7080 }, { "epoch": 0.76, "learning_rate": 1.4899031021939974e-06, "loss": 1.5264, "step": 7081 }, { "epoch": 0.76, "learning_rate": 1.4886730742553923e-06, "loss": 1.3419, "step": 7082 }, { "epoch": 0.76, "learning_rate": 1.487443465454389e-06, "loss": 1.614, "step": 7083 }, { "epoch": 0.76, "learning_rate": 1.4862142759377646e-06, "loss": 1.7206, "step": 7084 }, { "epoch": 0.76, "learning_rate": 1.4849855058522445e-06, "loss": 1.2561, "step": 7085 }, { "epoch": 0.76, "learning_rate": 1.483757155344503e-06, "loss": 1.3542, "step": 7086 }, { "epoch": 0.76, "learning_rate": 1.4825292245611633e-06, "loss": 1.1367, "step": 7087 }, { "epoch": 0.76, "learning_rate": 1.4813017136488028e-06, "loss": 1.5554, "step": 7088 }, { "epoch": 0.76, "learning_rate": 1.4800746227539437e-06, "loss": 1.3159, "step": 7089 }, { "epoch": 0.76, "learning_rate": 1.478847952023062e-06, "loss": 1.324, "step": 7090 }, { "epoch": 0.76, "learning_rate": 1.477621701602583e-06, "loss": 1.5047, "step": 7091 }, { "epoch": 0.76, "learning_rate": 1.4763958716388798e-06, "loss": 1.2676, "step": 7092 }, { "epoch": 0.76, "learning_rate": 1.4751704622782754e-06, "loss": 1.5759, "step": 7093 }, { "epoch": 0.76, "learning_rate": 1.4739454736670438e-06, "loss": 1.3234, "step": 7094 }, { "epoch": 0.76, "learning_rate": 1.4727209059514114e-06, "loss": 1.4604, "step": 7095 }, { "epoch": 0.76, "learning_rate": 1.4714967592775481e-06, "loss": 1.2422, "step": 7096 }, { "epoch": 0.76, "learning_rate": 1.4702730337915794e-06, "loss": 1.342, "step": 7097 }, { "epoch": 0.76, "learning_rate": 1.4690497296395773e-06, "loss": 1.3616, "step": 7098 }, { "epoch": 0.76, "learning_rate": 1.467826846967561e-06, "loss": 1.5072, "step": 7099 }, { "epoch": 0.76, "learning_rate": 1.466604385921509e-06, "loss": 1.3633, "step": 7100 }, { "epoch": 0.76, "learning_rate": 1.46538234664734e-06, "loss": 1.4408, "step": 7101 }, { "epoch": 0.76, "learning_rate": 1.4641607292909237e-06, "loss": 1.0792, "step": 7102 }, { "epoch": 0.76, "learning_rate": 1.4629395339980828e-06, "loss": 1.6054, "step": 7103 }, { "epoch": 0.76, "learning_rate": 1.4617187609145906e-06, "loss": 1.4736, "step": 7104 }, { "epoch": 0.76, "learning_rate": 1.460498410186163e-06, "loss": 1.1348, "step": 7105 }, { "epoch": 0.76, "learning_rate": 1.4592784819584742e-06, "loss": 1.5467, "step": 7106 }, { "epoch": 0.76, "learning_rate": 1.4580589763771413e-06, "loss": 1.4837, "step": 7107 }, { "epoch": 0.76, "learning_rate": 1.4568398935877326e-06, "loss": 1.5279, "step": 7108 }, { "epoch": 0.76, "learning_rate": 1.455621233735768e-06, "loss": 1.3319, "step": 7109 }, { "epoch": 0.76, "learning_rate": 1.4544029969667167e-06, "loss": 1.1936, "step": 7110 }, { "epoch": 0.76, "learning_rate": 1.4531851834259937e-06, "loss": 1.4391, "step": 7111 }, { "epoch": 0.76, "learning_rate": 1.4519677932589693e-06, "loss": 1.5955, "step": 7112 }, { "epoch": 0.76, "learning_rate": 1.4507508266109565e-06, "loss": 1.3148, "step": 7113 }, { "epoch": 0.76, "learning_rate": 1.4495342836272252e-06, "loss": 1.3599, "step": 7114 }, { "epoch": 0.76, "learning_rate": 1.448318164452987e-06, "loss": 1.5629, "step": 7115 }, { "epoch": 0.76, "learning_rate": 1.4471024692334101e-06, "loss": 1.4682, "step": 7116 }, { "epoch": 0.76, "learning_rate": 1.4458871981136074e-06, "loss": 1.6704, "step": 7117 }, { "epoch": 0.76, "learning_rate": 1.4446723512386391e-06, "loss": 1.2206, "step": 7118 }, { "epoch": 0.76, "learning_rate": 1.4434579287535244e-06, "loss": 1.3889, "step": 7119 }, { "epoch": 0.76, "learning_rate": 1.4422439308032228e-06, "loss": 1.4271, "step": 7120 }, { "epoch": 0.76, "learning_rate": 1.4410303575326446e-06, "loss": 1.2822, "step": 7121 }, { "epoch": 0.76, "learning_rate": 1.439817209086653e-06, "loss": 1.306, "step": 7122 }, { "epoch": 0.76, "learning_rate": 1.4386044856100562e-06, "loss": 1.4723, "step": 7123 }, { "epoch": 0.76, "learning_rate": 1.4373921872476143e-06, "loss": 1.3825, "step": 7124 }, { "epoch": 0.76, "learning_rate": 1.4361803141440384e-06, "loss": 1.4654, "step": 7125 }, { "epoch": 0.76, "learning_rate": 1.4349688664439848e-06, "loss": 1.5493, "step": 7126 }, { "epoch": 0.76, "learning_rate": 1.4337578442920592e-06, "loss": 1.2263, "step": 7127 }, { "epoch": 0.76, "learning_rate": 1.432547247832819e-06, "loss": 1.4916, "step": 7128 }, { "epoch": 0.76, "learning_rate": 1.4313370772107715e-06, "loss": 1.2647, "step": 7129 }, { "epoch": 0.76, "learning_rate": 1.430127332570369e-06, "loss": 1.5685, "step": 7130 }, { "epoch": 0.76, "learning_rate": 1.4289180140560189e-06, "loss": 1.4026, "step": 7131 }, { "epoch": 0.76, "learning_rate": 1.4277091218120715e-06, "loss": 1.3234, "step": 7132 }, { "epoch": 0.76, "learning_rate": 1.4265006559828282e-06, "loss": 1.3209, "step": 7133 }, { "epoch": 0.76, "learning_rate": 1.4252926167125413e-06, "loss": 1.4451, "step": 7134 }, { "epoch": 0.76, "learning_rate": 1.4240850041454136e-06, "loss": 1.3778, "step": 7135 }, { "epoch": 0.76, "learning_rate": 1.4228778184255908e-06, "loss": 1.6592, "step": 7136 }, { "epoch": 0.76, "learning_rate": 1.421671059697175e-06, "loss": 1.595, "step": 7137 }, { "epoch": 0.76, "learning_rate": 1.42046472810421e-06, "loss": 1.2715, "step": 7138 }, { "epoch": 0.76, "learning_rate": 1.4192588237906957e-06, "loss": 1.2745, "step": 7139 }, { "epoch": 0.76, "learning_rate": 1.418053346900574e-06, "loss": 1.4102, "step": 7140 }, { "epoch": 0.76, "learning_rate": 1.4168482975777441e-06, "loss": 1.5865, "step": 7141 }, { "epoch": 0.76, "learning_rate": 1.4156436759660447e-06, "loss": 1.3129, "step": 7142 }, { "epoch": 0.76, "learning_rate": 1.4144394822092712e-06, "loss": 1.2152, "step": 7143 }, { "epoch": 0.76, "learning_rate": 1.4132357164511652e-06, "loss": 1.3976, "step": 7144 }, { "epoch": 0.76, "learning_rate": 1.4120323788354156e-06, "loss": 1.4475, "step": 7145 }, { "epoch": 0.76, "learning_rate": 1.4108294695056606e-06, "loss": 1.4227, "step": 7146 }, { "epoch": 0.76, "learning_rate": 1.4096269886054904e-06, "loss": 1.1774, "step": 7147 }, { "epoch": 0.76, "learning_rate": 1.408424936278442e-06, "loss": 1.397, "step": 7148 }, { "epoch": 0.76, "learning_rate": 1.4072233126679985e-06, "loss": 1.5897, "step": 7149 }, { "epoch": 0.76, "learning_rate": 1.4060221179175976e-06, "loss": 1.671, "step": 7150 }, { "epoch": 0.76, "learning_rate": 1.404821352170621e-06, "loss": 1.553, "step": 7151 }, { "epoch": 0.76, "learning_rate": 1.4036210155703989e-06, "loss": 1.1592, "step": 7152 }, { "epoch": 0.76, "learning_rate": 1.4024211082602168e-06, "loss": 1.2323, "step": 7153 }, { "epoch": 0.76, "learning_rate": 1.401221630383302e-06, "loss": 1.5411, "step": 7154 }, { "epoch": 0.76, "learning_rate": 1.4000225820828317e-06, "loss": 1.3064, "step": 7155 }, { "epoch": 0.76, "learning_rate": 1.3988239635019357e-06, "loss": 1.0902, "step": 7156 }, { "epoch": 0.76, "learning_rate": 1.3976257747836875e-06, "loss": 1.3527, "step": 7157 }, { "epoch": 0.76, "learning_rate": 1.3964280160711119e-06, "loss": 1.367, "step": 7158 }, { "epoch": 0.76, "learning_rate": 1.3952306875071847e-06, "loss": 1.3931, "step": 7159 }, { "epoch": 0.76, "learning_rate": 1.3940337892348255e-06, "loss": 1.3679, "step": 7160 }, { "epoch": 0.76, "learning_rate": 1.3928373213969038e-06, "loss": 1.4045, "step": 7161 }, { "epoch": 0.76, "learning_rate": 1.3916412841362404e-06, "loss": 1.7236, "step": 7162 }, { "epoch": 0.76, "learning_rate": 1.3904456775956044e-06, "loss": 1.3674, "step": 7163 }, { "epoch": 0.76, "learning_rate": 1.38925050191771e-06, "loss": 1.4019, "step": 7164 }, { "epoch": 0.76, "learning_rate": 1.3880557572452213e-06, "loss": 1.276, "step": 7165 }, { "epoch": 0.76, "learning_rate": 1.3868614437207545e-06, "loss": 1.3921, "step": 7166 }, { "epoch": 0.76, "learning_rate": 1.3856675614868687e-06, "loss": 1.5226, "step": 7167 }, { "epoch": 0.76, "learning_rate": 1.3844741106860759e-06, "loss": 1.3531, "step": 7168 }, { "epoch": 0.76, "learning_rate": 1.3832810914608364e-06, "loss": 1.4618, "step": 7169 }, { "epoch": 0.76, "learning_rate": 1.3820885039535564e-06, "loss": 1.5413, "step": 7170 }, { "epoch": 0.76, "learning_rate": 1.3808963483065902e-06, "loss": 1.3197, "step": 7171 }, { "epoch": 0.77, "learning_rate": 1.3797046246622431e-06, "loss": 1.3742, "step": 7172 }, { "epoch": 0.77, "learning_rate": 1.378513333162771e-06, "loss": 1.4252, "step": 7173 }, { "epoch": 0.77, "learning_rate": 1.3773224739503704e-06, "loss": 1.3718, "step": 7174 }, { "epoch": 0.77, "learning_rate": 1.3761320471671951e-06, "loss": 1.5203, "step": 7175 }, { "epoch": 0.77, "learning_rate": 1.3749420529553414e-06, "loss": 1.4034, "step": 7176 }, { "epoch": 0.77, "learning_rate": 1.3737524914568523e-06, "loss": 1.4047, "step": 7177 }, { "epoch": 0.77, "learning_rate": 1.3725633628137292e-06, "loss": 1.3311, "step": 7178 }, { "epoch": 0.77, "learning_rate": 1.3713746671679112e-06, "loss": 1.5819, "step": 7179 }, { "epoch": 0.77, "learning_rate": 1.3701864046612888e-06, "loss": 1.3362, "step": 7180 }, { "epoch": 0.77, "learning_rate": 1.3689985754357054e-06, "loss": 1.1952, "step": 7181 }, { "epoch": 0.77, "learning_rate": 1.3678111796329446e-06, "loss": 1.377, "step": 7182 }, { "epoch": 0.77, "learning_rate": 1.3666242173947447e-06, "loss": 1.357, "step": 7183 }, { "epoch": 0.77, "learning_rate": 1.3654376888627918e-06, "loss": 1.1812, "step": 7184 }, { "epoch": 0.77, "learning_rate": 1.3642515941787171e-06, "loss": 1.3413, "step": 7185 }, { "epoch": 0.77, "learning_rate": 1.3630659334841002e-06, "loss": 1.3162, "step": 7186 }, { "epoch": 0.77, "learning_rate": 1.3618807069204708e-06, "loss": 1.2915, "step": 7187 }, { "epoch": 0.77, "learning_rate": 1.3606959146293086e-06, "loss": 1.2917, "step": 7188 }, { "epoch": 0.77, "learning_rate": 1.3595115567520361e-06, "loss": 1.5268, "step": 7189 }, { "epoch": 0.77, "learning_rate": 1.3583276334300295e-06, "loss": 1.6092, "step": 7190 }, { "epoch": 0.77, "learning_rate": 1.3571441448046086e-06, "loss": 1.4682, "step": 7191 }, { "epoch": 0.77, "learning_rate": 1.3559610910170423e-06, "loss": 1.3278, "step": 7192 }, { "epoch": 0.77, "learning_rate": 1.3547784722085505e-06, "loss": 1.3277, "step": 7193 }, { "epoch": 0.77, "learning_rate": 1.3535962885202997e-06, "loss": 1.5575, "step": 7194 }, { "epoch": 0.77, "learning_rate": 1.3524145400934019e-06, "loss": 1.4645, "step": 7195 }, { "epoch": 0.77, "learning_rate": 1.35123322706892e-06, "loss": 1.5452, "step": 7196 }, { "epoch": 0.77, "learning_rate": 1.350052349587866e-06, "loss": 1.4205, "step": 7197 }, { "epoch": 0.77, "learning_rate": 1.3488719077911965e-06, "loss": 1.3252, "step": 7198 }, { "epoch": 0.77, "learning_rate": 1.3476919018198159e-06, "loss": 1.5073, "step": 7199 }, { "epoch": 0.77, "learning_rate": 1.3465123318145817e-06, "loss": 1.2802, "step": 7200 }, { "epoch": 0.77, "learning_rate": 1.3453331979162932e-06, "loss": 1.2499, "step": 7201 }, { "epoch": 0.77, "learning_rate": 1.3441545002657008e-06, "loss": 1.5022, "step": 7202 }, { "epoch": 0.77, "learning_rate": 1.342976239003505e-06, "loss": 1.3503, "step": 7203 }, { "epoch": 0.77, "learning_rate": 1.34179841427035e-06, "loss": 1.3532, "step": 7204 }, { "epoch": 0.77, "learning_rate": 1.340621026206828e-06, "loss": 1.2303, "step": 7205 }, { "epoch": 0.77, "learning_rate": 1.339444074953482e-06, "loss": 1.2616, "step": 7206 }, { "epoch": 0.77, "learning_rate": 1.3382675606508028e-06, "loss": 1.5491, "step": 7207 }, { "epoch": 0.77, "learning_rate": 1.3370914834392251e-06, "loss": 1.5545, "step": 7208 }, { "epoch": 0.77, "learning_rate": 1.335915843459137e-06, "loss": 1.2991, "step": 7209 }, { "epoch": 0.77, "learning_rate": 1.3347406408508695e-06, "loss": 1.4959, "step": 7210 }, { "epoch": 0.77, "learning_rate": 1.3335658757547015e-06, "loss": 1.5948, "step": 7211 }, { "epoch": 0.77, "learning_rate": 1.3323915483108662e-06, "loss": 1.0607, "step": 7212 }, { "epoch": 0.77, "learning_rate": 1.3312176586595377e-06, "loss": 1.5474, "step": 7213 }, { "epoch": 0.77, "learning_rate": 1.330044206940838e-06, "loss": 1.3859, "step": 7214 }, { "epoch": 0.77, "learning_rate": 1.3288711932948427e-06, "loss": 1.2589, "step": 7215 }, { "epoch": 0.77, "learning_rate": 1.3276986178615676e-06, "loss": 1.339, "step": 7216 }, { "epoch": 0.77, "learning_rate": 1.3265264807809824e-06, "loss": 1.5427, "step": 7217 }, { "epoch": 0.77, "learning_rate": 1.3253547821930002e-06, "loss": 1.5595, "step": 7218 }, { "epoch": 0.77, "learning_rate": 1.3241835222374855e-06, "loss": 1.2725, "step": 7219 }, { "epoch": 0.77, "learning_rate": 1.323012701054246e-06, "loss": 1.4197, "step": 7220 }, { "epoch": 0.77, "learning_rate": 1.3218423187830409e-06, "loss": 1.213, "step": 7221 }, { "epoch": 0.77, "learning_rate": 1.3206723755635765e-06, "loss": 1.1495, "step": 7222 }, { "epoch": 0.77, "learning_rate": 1.3195028715355046e-06, "loss": 1.1948, "step": 7223 }, { "epoch": 0.77, "learning_rate": 1.3183338068384243e-06, "loss": 1.5642, "step": 7224 }, { "epoch": 0.77, "learning_rate": 1.317165181611887e-06, "loss": 1.4495, "step": 7225 }, { "epoch": 0.77, "learning_rate": 1.315996995995385e-06, "loss": 1.4082, "step": 7226 }, { "epoch": 0.77, "learning_rate": 1.3148292501283627e-06, "loss": 1.4138, "step": 7227 }, { "epoch": 0.77, "learning_rate": 1.3136619441502124e-06, "loss": 1.4696, "step": 7228 }, { "epoch": 0.77, "learning_rate": 1.312495078200271e-06, "loss": 1.3517, "step": 7229 }, { "epoch": 0.77, "learning_rate": 1.3113286524178232e-06, "loss": 1.3666, "step": 7230 }, { "epoch": 0.77, "learning_rate": 1.3101626669421019e-06, "loss": 1.5491, "step": 7231 }, { "epoch": 0.77, "learning_rate": 1.308997121912291e-06, "loss": 1.4657, "step": 7232 }, { "epoch": 0.77, "learning_rate": 1.3078320174675141e-06, "loss": 1.3831, "step": 7233 }, { "epoch": 0.77, "learning_rate": 1.3066673537468499e-06, "loss": 1.4165, "step": 7234 }, { "epoch": 0.77, "learning_rate": 1.3055031308893206e-06, "loss": 1.6294, "step": 7235 }, { "epoch": 0.77, "learning_rate": 1.3043393490338918e-06, "loss": 1.391, "step": 7236 }, { "epoch": 0.77, "learning_rate": 1.3031760083194884e-06, "loss": 1.1433, "step": 7237 }, { "epoch": 0.77, "learning_rate": 1.3020131088849714e-06, "loss": 1.1348, "step": 7238 }, { "epoch": 0.77, "learning_rate": 1.3008506508691516e-06, "loss": 1.4011, "step": 7239 }, { "epoch": 0.77, "learning_rate": 1.29968863441079e-06, "loss": 1.4755, "step": 7240 }, { "epoch": 0.77, "learning_rate": 1.2985270596485949e-06, "loss": 1.4912, "step": 7241 }, { "epoch": 0.77, "learning_rate": 1.2973659267212173e-06, "loss": 1.218, "step": 7242 }, { "epoch": 0.77, "learning_rate": 1.2962052357672617e-06, "loss": 1.1487, "step": 7243 }, { "epoch": 0.77, "learning_rate": 1.295044986925275e-06, "loss": 1.2284, "step": 7244 }, { "epoch": 0.77, "learning_rate": 1.2938851803337516e-06, "loss": 1.3313, "step": 7245 }, { "epoch": 0.77, "learning_rate": 1.2927258161311357e-06, "loss": 1.3691, "step": 7246 }, { "epoch": 0.77, "learning_rate": 1.2915668944558192e-06, "loss": 1.4914, "step": 7247 }, { "epoch": 0.77, "learning_rate": 1.290408415446136e-06, "loss": 1.3572, "step": 7248 }, { "epoch": 0.77, "learning_rate": 1.2892503792403737e-06, "loss": 1.4194, "step": 7249 }, { "epoch": 0.77, "learning_rate": 1.2880927859767617e-06, "loss": 1.2811, "step": 7250 }, { "epoch": 0.77, "learning_rate": 1.2869356357934815e-06, "loss": 1.1224, "step": 7251 }, { "epoch": 0.77, "learning_rate": 1.2857789288286553e-06, "loss": 1.3413, "step": 7252 }, { "epoch": 0.77, "learning_rate": 1.2846226652203602e-06, "loss": 1.423, "step": 7253 }, { "epoch": 0.77, "learning_rate": 1.2834668451066118e-06, "loss": 1.4493, "step": 7254 }, { "epoch": 0.77, "learning_rate": 1.2823114686253802e-06, "loss": 1.2236, "step": 7255 }, { "epoch": 0.77, "learning_rate": 1.28115653591458e-06, "loss": 1.4383, "step": 7256 }, { "epoch": 0.77, "learning_rate": 1.2800020471120717e-06, "loss": 1.504, "step": 7257 }, { "epoch": 0.77, "learning_rate": 1.2788480023556615e-06, "loss": 1.3313, "step": 7258 }, { "epoch": 0.77, "learning_rate": 1.2776944017831071e-06, "loss": 1.3366, "step": 7259 }, { "epoch": 0.77, "learning_rate": 1.276541245532109e-06, "loss": 1.5078, "step": 7260 }, { "epoch": 0.77, "learning_rate": 1.275388533740317e-06, "loss": 1.4794, "step": 7261 }, { "epoch": 0.77, "learning_rate": 1.274236266545329e-06, "loss": 1.3028, "step": 7262 }, { "epoch": 0.77, "learning_rate": 1.2730844440846862e-06, "loss": 1.3711, "step": 7263 }, { "epoch": 0.77, "learning_rate": 1.2719330664958767e-06, "loss": 1.4479, "step": 7264 }, { "epoch": 0.77, "learning_rate": 1.27078213391634e-06, "loss": 1.3519, "step": 7265 }, { "epoch": 0.78, "learning_rate": 1.2696316464834607e-06, "loss": 1.3942, "step": 7266 }, { "epoch": 0.78, "learning_rate": 1.268481604334566e-06, "loss": 1.5561, "step": 7267 }, { "epoch": 0.78, "learning_rate": 1.2673320076069363e-06, "loss": 1.2693, "step": 7268 }, { "epoch": 0.78, "learning_rate": 1.2661828564377948e-06, "loss": 1.4135, "step": 7269 }, { "epoch": 0.78, "learning_rate": 1.2650341509643116e-06, "loss": 1.2988, "step": 7270 }, { "epoch": 0.78, "learning_rate": 1.2638858913236045e-06, "loss": 1.3887, "step": 7271 }, { "epoch": 0.78, "learning_rate": 1.2627380776527415e-06, "loss": 1.4475, "step": 7272 }, { "epoch": 0.78, "learning_rate": 1.2615907100887298e-06, "loss": 1.2522, "step": 7273 }, { "epoch": 0.78, "learning_rate": 1.2604437887685306e-06, "loss": 1.396, "step": 7274 }, { "epoch": 0.78, "learning_rate": 1.259297313829046e-06, "loss": 1.4444, "step": 7275 }, { "epoch": 0.78, "learning_rate": 1.258151285407131e-06, "loss": 1.2982, "step": 7276 }, { "epoch": 0.78, "learning_rate": 1.25700570363958e-06, "loss": 1.3739, "step": 7277 }, { "epoch": 0.78, "learning_rate": 1.255860568663142e-06, "loss": 1.2506, "step": 7278 }, { "epoch": 0.78, "learning_rate": 1.2547158806145049e-06, "loss": 1.3449, "step": 7279 }, { "epoch": 0.78, "learning_rate": 1.2535716396303093e-06, "loss": 1.3544, "step": 7280 }, { "epoch": 0.78, "learning_rate": 1.2524278458471411e-06, "loss": 1.5568, "step": 7281 }, { "epoch": 0.78, "learning_rate": 1.2512844994015304e-06, "loss": 1.4811, "step": 7282 }, { "epoch": 0.78, "learning_rate": 1.2501416004299537e-06, "loss": 1.3399, "step": 7283 }, { "epoch": 0.78, "learning_rate": 1.248999149068838e-06, "loss": 1.2785, "step": 7284 }, { "epoch": 0.78, "learning_rate": 1.2478571454545558e-06, "loss": 1.3942, "step": 7285 }, { "epoch": 0.78, "learning_rate": 1.2467155897234212e-06, "loss": 1.5391, "step": 7286 }, { "epoch": 0.78, "learning_rate": 1.2455744820117028e-06, "loss": 1.5161, "step": 7287 }, { "epoch": 0.78, "learning_rate": 1.244433822455609e-06, "loss": 1.2182, "step": 7288 }, { "epoch": 0.78, "learning_rate": 1.2432936111912946e-06, "loss": 1.3632, "step": 7289 }, { "epoch": 0.78, "learning_rate": 1.2421538483548706e-06, "loss": 1.3329, "step": 7290 }, { "epoch": 0.78, "learning_rate": 1.2410145340823827e-06, "loss": 1.2385, "step": 7291 }, { "epoch": 0.78, "learning_rate": 1.2398756685098272e-06, "loss": 1.5685, "step": 7292 }, { "epoch": 0.78, "learning_rate": 1.2387372517731505e-06, "loss": 1.4261, "step": 7293 }, { "epoch": 0.78, "learning_rate": 1.237599284008239e-06, "loss": 1.3405, "step": 7294 }, { "epoch": 0.78, "learning_rate": 1.2364617653509302e-06, "loss": 1.494, "step": 7295 }, { "epoch": 0.78, "learning_rate": 1.2353246959370086e-06, "loss": 1.3613, "step": 7296 }, { "epoch": 0.78, "learning_rate": 1.2341880759022013e-06, "loss": 1.2652, "step": 7297 }, { "epoch": 0.78, "learning_rate": 1.2330519053821821e-06, "loss": 1.3244, "step": 7298 }, { "epoch": 0.78, "learning_rate": 1.2319161845125744e-06, "loss": 1.4133, "step": 7299 }, { "epoch": 0.78, "learning_rate": 1.2307809134289466e-06, "loss": 1.1452, "step": 7300 }, { "epoch": 0.78, "learning_rate": 1.2296460922668113e-06, "loss": 1.4148, "step": 7301 }, { "epoch": 0.78, "learning_rate": 1.228511721161631e-06, "loss": 1.3767, "step": 7302 }, { "epoch": 0.78, "learning_rate": 1.2273778002488117e-06, "loss": 1.542, "step": 7303 }, { "epoch": 0.78, "learning_rate": 1.226244329663705e-06, "loss": 1.1912, "step": 7304 }, { "epoch": 0.78, "learning_rate": 1.2251113095416113e-06, "loss": 1.3992, "step": 7305 }, { "epoch": 0.78, "learning_rate": 1.2239787400177771e-06, "loss": 1.6719, "step": 7306 }, { "epoch": 0.78, "learning_rate": 1.2228466212273926e-06, "loss": 1.4602, "step": 7307 }, { "epoch": 0.78, "learning_rate": 1.2217149533055976e-06, "loss": 1.3364, "step": 7308 }, { "epoch": 0.78, "learning_rate": 1.220583736387474e-06, "loss": 1.5662, "step": 7309 }, { "epoch": 0.78, "learning_rate": 1.2194529706080543e-06, "loss": 1.2359, "step": 7310 }, { "epoch": 0.78, "learning_rate": 1.2183226561023132e-06, "loss": 1.5818, "step": 7311 }, { "epoch": 0.78, "learning_rate": 1.2171927930051747e-06, "loss": 1.2475, "step": 7312 }, { "epoch": 0.78, "learning_rate": 1.2160633814515072e-06, "loss": 1.4042, "step": 7313 }, { "epoch": 0.78, "learning_rate": 1.2149344215761216e-06, "loss": 1.3507, "step": 7314 }, { "epoch": 0.78, "learning_rate": 1.213805913513786e-06, "loss": 1.4835, "step": 7315 }, { "epoch": 0.78, "learning_rate": 1.2126778573992031e-06, "loss": 1.3917, "step": 7316 }, { "epoch": 0.78, "learning_rate": 1.2115502533670253e-06, "loss": 1.1464, "step": 7317 }, { "epoch": 0.78, "learning_rate": 1.2104231015518535e-06, "loss": 1.4225, "step": 7318 }, { "epoch": 0.78, "learning_rate": 1.2092964020882315e-06, "loss": 1.2939, "step": 7319 }, { "epoch": 0.78, "learning_rate": 1.2081701551106506e-06, "loss": 1.1938, "step": 7320 }, { "epoch": 0.78, "learning_rate": 1.2070443607535498e-06, "loss": 1.3136, "step": 7321 }, { "epoch": 0.78, "learning_rate": 1.2059190191513104e-06, "loss": 1.3144, "step": 7322 }, { "epoch": 0.78, "learning_rate": 1.20479413043826e-06, "loss": 1.3006, "step": 7323 }, { "epoch": 0.78, "learning_rate": 1.2036696947486748e-06, "loss": 1.4062, "step": 7324 }, { "epoch": 0.78, "learning_rate": 1.2025457122167777e-06, "loss": 1.5496, "step": 7325 }, { "epoch": 0.78, "learning_rate": 1.201422182976732e-06, "loss": 1.4438, "step": 7326 }, { "epoch": 0.78, "learning_rate": 1.2002991071626534e-06, "loss": 1.3901, "step": 7327 }, { "epoch": 0.78, "learning_rate": 1.199176484908598e-06, "loss": 1.419, "step": 7328 }, { "epoch": 0.78, "learning_rate": 1.1980543163485726e-06, "loss": 1.3947, "step": 7329 }, { "epoch": 0.78, "learning_rate": 1.1969326016165239e-06, "loss": 1.3173, "step": 7330 }, { "epoch": 0.78, "learning_rate": 1.1958113408463518e-06, "loss": 1.5121, "step": 7331 }, { "epoch": 0.78, "learning_rate": 1.1946905341718951e-06, "loss": 1.3032, "step": 7332 }, { "epoch": 0.78, "learning_rate": 1.193570181726943e-06, "loss": 1.2708, "step": 7333 }, { "epoch": 0.78, "learning_rate": 1.1924502836452294e-06, "loss": 1.4513, "step": 7334 }, { "epoch": 0.78, "learning_rate": 1.1913308400604339e-06, "loss": 1.3829, "step": 7335 }, { "epoch": 0.78, "learning_rate": 1.1902118511061783e-06, "loss": 1.3728, "step": 7336 }, { "epoch": 0.78, "learning_rate": 1.1890933169160374e-06, "loss": 1.3891, "step": 7337 }, { "epoch": 0.78, "learning_rate": 1.1879752376235231e-06, "loss": 1.3853, "step": 7338 }, { "epoch": 0.78, "learning_rate": 1.1868576133621012e-06, "loss": 1.4941, "step": 7339 }, { "epoch": 0.78, "learning_rate": 1.1857404442651788e-06, "loss": 1.4715, "step": 7340 }, { "epoch": 0.78, "learning_rate": 1.1846237304661095e-06, "loss": 1.2019, "step": 7341 }, { "epoch": 0.78, "learning_rate": 1.18350747209819e-06, "loss": 1.3148, "step": 7342 }, { "epoch": 0.78, "learning_rate": 1.182391669294667e-06, "loss": 1.4443, "step": 7343 }, { "epoch": 0.78, "learning_rate": 1.181276322188732e-06, "loss": 1.4191, "step": 7344 }, { "epoch": 0.78, "learning_rate": 1.1801614309135178e-06, "loss": 1.2546, "step": 7345 }, { "epoch": 0.78, "learning_rate": 1.1790469956021095e-06, "loss": 1.5071, "step": 7346 }, { "epoch": 0.78, "learning_rate": 1.1779330163875325e-06, "loss": 1.3044, "step": 7347 }, { "epoch": 0.78, "learning_rate": 1.1768194934027565e-06, "loss": 1.5007, "step": 7348 }, { "epoch": 0.78, "learning_rate": 1.175706426780706e-06, "loss": 1.4418, "step": 7349 }, { "epoch": 0.78, "learning_rate": 1.1745938166542414e-06, "loss": 1.4368, "step": 7350 }, { "epoch": 0.78, "learning_rate": 1.1734816631561702e-06, "loss": 1.3107, "step": 7351 }, { "epoch": 0.78, "learning_rate": 1.1723699664192507e-06, "loss": 1.1226, "step": 7352 }, { "epoch": 0.78, "learning_rate": 1.1712587265761799e-06, "loss": 1.0565, "step": 7353 }, { "epoch": 0.78, "learning_rate": 1.1701479437596054e-06, "loss": 1.5284, "step": 7354 }, { "epoch": 0.78, "learning_rate": 1.169037618102119e-06, "loss": 1.2183, "step": 7355 }, { "epoch": 0.78, "learning_rate": 1.1679277497362563e-06, "loss": 1.363, "step": 7356 }, { "epoch": 0.78, "learning_rate": 1.166818338794498e-06, "loss": 1.2473, "step": 7357 }, { "epoch": 0.78, "learning_rate": 1.1657093854092727e-06, "loss": 1.5359, "step": 7358 }, { "epoch": 0.78, "learning_rate": 1.1646008897129546e-06, "loss": 1.3848, "step": 7359 }, { "epoch": 0.79, "learning_rate": 1.1634928518378602e-06, "loss": 1.22, "step": 7360 }, { "epoch": 0.79, "learning_rate": 1.162385271916252e-06, "loss": 1.5076, "step": 7361 }, { "epoch": 0.79, "learning_rate": 1.161278150080341e-06, "loss": 1.4695, "step": 7362 }, { "epoch": 0.79, "learning_rate": 1.16017148646228e-06, "loss": 1.2672, "step": 7363 }, { "epoch": 0.79, "learning_rate": 1.1590652811941677e-06, "loss": 1.4608, "step": 7364 }, { "epoch": 0.79, "learning_rate": 1.157959534408052e-06, "loss": 1.2711, "step": 7365 }, { "epoch": 0.79, "learning_rate": 1.1568542462359206e-06, "loss": 1.6344, "step": 7366 }, { "epoch": 0.79, "learning_rate": 1.1557494168097077e-06, "loss": 1.5818, "step": 7367 }, { "epoch": 0.79, "learning_rate": 1.1546450462612951e-06, "loss": 1.2924, "step": 7368 }, { "epoch": 0.79, "learning_rate": 1.1535411347225096e-06, "loss": 1.3968, "step": 7369 }, { "epoch": 0.79, "learning_rate": 1.15243768232512e-06, "loss": 1.3398, "step": 7370 }, { "epoch": 0.79, "learning_rate": 1.151334689200845e-06, "loss": 1.4784, "step": 7371 }, { "epoch": 0.79, "learning_rate": 1.1502321554813429e-06, "loss": 1.302, "step": 7372 }, { "epoch": 0.79, "learning_rate": 1.1491300812982216e-06, "loss": 1.4614, "step": 7373 }, { "epoch": 0.79, "learning_rate": 1.1480284667830343e-06, "loss": 1.3164, "step": 7374 }, { "epoch": 0.79, "learning_rate": 1.1469273120672764e-06, "loss": 1.3633, "step": 7375 }, { "epoch": 0.79, "learning_rate": 1.1458266172823879e-06, "loss": 0.9717, "step": 7376 }, { "epoch": 0.79, "learning_rate": 1.1447263825597577e-06, "loss": 1.4477, "step": 7377 }, { "epoch": 0.79, "learning_rate": 1.143626608030719e-06, "loss": 1.3922, "step": 7378 }, { "epoch": 0.79, "learning_rate": 1.1425272938265463e-06, "loss": 1.6161, "step": 7379 }, { "epoch": 0.79, "learning_rate": 1.1414284400784643e-06, "loss": 1.5491, "step": 7380 }, { "epoch": 0.79, "learning_rate": 1.1403300469176387e-06, "loss": 1.5007, "step": 7381 }, { "epoch": 0.79, "learning_rate": 1.1392321144751805e-06, "loss": 1.2581, "step": 7382 }, { "epoch": 0.79, "learning_rate": 1.1381346428821482e-06, "loss": 1.5486, "step": 7383 }, { "epoch": 0.79, "learning_rate": 1.1370376322695458e-06, "loss": 1.5595, "step": 7384 }, { "epoch": 0.79, "learning_rate": 1.1359410827683177e-06, "loss": 1.3326, "step": 7385 }, { "epoch": 0.79, "learning_rate": 1.134844994509358e-06, "loss": 1.481, "step": 7386 }, { "epoch": 0.79, "learning_rate": 1.1337493676235023e-06, "loss": 1.6415, "step": 7387 }, { "epoch": 0.79, "learning_rate": 1.1326542022415343e-06, "loss": 1.4027, "step": 7388 }, { "epoch": 0.79, "learning_rate": 1.1315594984941786e-06, "loss": 1.2851, "step": 7389 }, { "epoch": 0.79, "learning_rate": 1.13046525651211e-06, "loss": 1.2061, "step": 7390 }, { "epoch": 0.79, "learning_rate": 1.1293714764259418e-06, "loss": 1.5855, "step": 7391 }, { "epoch": 0.79, "learning_rate": 1.1282781583662372e-06, "loss": 1.5768, "step": 7392 }, { "epoch": 0.79, "learning_rate": 1.127185302463505e-06, "loss": 1.1335, "step": 7393 }, { "epoch": 0.79, "learning_rate": 1.1260929088481932e-06, "loss": 1.5512, "step": 7394 }, { "epoch": 0.79, "learning_rate": 1.1250009776506982e-06, "loss": 1.5495, "step": 7395 }, { "epoch": 0.79, "learning_rate": 1.123909509001363e-06, "loss": 1.3861, "step": 7396 }, { "epoch": 0.79, "learning_rate": 1.12281850303047e-06, "loss": 1.4448, "step": 7397 }, { "epoch": 0.79, "learning_rate": 1.1217279598682518e-06, "loss": 1.288, "step": 7398 }, { "epoch": 0.79, "learning_rate": 1.1206378796448847e-06, "loss": 1.5248, "step": 7399 }, { "epoch": 0.79, "learning_rate": 1.119548262490487e-06, "loss": 1.1837, "step": 7400 }, { "epoch": 0.79, "learning_rate": 1.118459108535122e-06, "loss": 1.5601, "step": 7401 }, { "epoch": 0.79, "learning_rate": 1.1173704179088007e-06, "loss": 1.3484, "step": 7402 }, { "epoch": 0.79, "learning_rate": 1.1162821907414789e-06, "loss": 1.4156, "step": 7403 }, { "epoch": 0.79, "learning_rate": 1.1151944271630517e-06, "loss": 1.384, "step": 7404 }, { "epoch": 0.79, "learning_rate": 1.114107127303366e-06, "loss": 1.2198, "step": 7405 }, { "epoch": 0.79, "learning_rate": 1.1130202912922077e-06, "loss": 1.4099, "step": 7406 }, { "epoch": 0.79, "learning_rate": 1.1119339192593077e-06, "loss": 1.2517, "step": 7407 }, { "epoch": 0.79, "learning_rate": 1.1108480113343478e-06, "loss": 1.4129, "step": 7408 }, { "epoch": 0.79, "learning_rate": 1.1097625676469475e-06, "loss": 1.5018, "step": 7409 }, { "epoch": 0.79, "learning_rate": 1.1086775883266725e-06, "loss": 1.3061, "step": 7410 }, { "epoch": 0.79, "learning_rate": 1.107593073503036e-06, "loss": 1.3548, "step": 7411 }, { "epoch": 0.79, "learning_rate": 1.1065090233054908e-06, "loss": 1.3529, "step": 7412 }, { "epoch": 0.79, "learning_rate": 1.1054254378634399e-06, "loss": 1.4106, "step": 7413 }, { "epoch": 0.79, "learning_rate": 1.1043423173062257e-06, "loss": 1.1094, "step": 7414 }, { "epoch": 0.79, "learning_rate": 1.1032596617631392e-06, "loss": 1.4139, "step": 7415 }, { "epoch": 0.79, "learning_rate": 1.102177471363412e-06, "loss": 1.4883, "step": 7416 }, { "epoch": 0.79, "learning_rate": 1.1010957462362233e-06, "loss": 1.4492, "step": 7417 }, { "epoch": 0.79, "learning_rate": 1.1000144865106976e-06, "loss": 1.5625, "step": 7418 }, { "epoch": 0.79, "learning_rate": 1.0989336923158999e-06, "loss": 1.2562, "step": 7419 }, { "epoch": 0.79, "learning_rate": 1.0978533637808403e-06, "loss": 1.2672, "step": 7420 }, { "epoch": 0.79, "learning_rate": 1.0967735010344765e-06, "loss": 1.356, "step": 7421 }, { "epoch": 0.79, "learning_rate": 1.0956941042057106e-06, "loss": 1.4714, "step": 7422 }, { "epoch": 0.79, "learning_rate": 1.0946151734233834e-06, "loss": 1.1923, "step": 7423 }, { "epoch": 0.79, "learning_rate": 1.0935367088162873e-06, "loss": 1.3199, "step": 7424 }, { "epoch": 0.79, "learning_rate": 1.0924587105131546e-06, "loss": 1.329, "step": 7425 }, { "epoch": 0.79, "learning_rate": 1.0913811786426597e-06, "loss": 1.393, "step": 7426 }, { "epoch": 0.79, "learning_rate": 1.0903041133334309e-06, "loss": 1.3913, "step": 7427 }, { "epoch": 0.79, "learning_rate": 1.0892275147140307e-06, "loss": 1.4756, "step": 7428 }, { "epoch": 0.79, "learning_rate": 1.0881513829129696e-06, "loss": 1.422, "step": 7429 }, { "epoch": 0.79, "learning_rate": 1.0870757180587044e-06, "loss": 1.4968, "step": 7430 }, { "epoch": 0.79, "learning_rate": 1.086000520279632e-06, "loss": 1.2369, "step": 7431 }, { "epoch": 0.79, "learning_rate": 1.084925789704097e-06, "loss": 1.2005, "step": 7432 }, { "epoch": 0.79, "learning_rate": 1.0838515264603893e-06, "loss": 1.2067, "step": 7433 }, { "epoch": 0.79, "learning_rate": 1.0827777306767384e-06, "loss": 1.3113, "step": 7434 }, { "epoch": 0.79, "learning_rate": 1.0817044024813189e-06, "loss": 1.1507, "step": 7435 }, { "epoch": 0.79, "learning_rate": 1.0806315420022535e-06, "loss": 1.3793, "step": 7436 }, { "epoch": 0.79, "learning_rate": 1.0795591493676072e-06, "loss": 1.2532, "step": 7437 }, { "epoch": 0.79, "learning_rate": 1.0784872247053863e-06, "loss": 1.465, "step": 7438 }, { "epoch": 0.79, "learning_rate": 1.0774157681435455e-06, "loss": 1.6045, "step": 7439 }, { "epoch": 0.79, "learning_rate": 1.0763447798099813e-06, "loss": 1.395, "step": 7440 }, { "epoch": 0.79, "learning_rate": 1.0752742598325332e-06, "loss": 1.435, "step": 7441 }, { "epoch": 0.79, "learning_rate": 1.074204208338987e-06, "loss": 1.4525, "step": 7442 }, { "epoch": 0.79, "learning_rate": 1.0731346254570735e-06, "loss": 1.5328, "step": 7443 }, { "epoch": 0.79, "learning_rate": 1.0720655113144635e-06, "loss": 1.1471, "step": 7444 }, { "epoch": 0.79, "learning_rate": 1.0709968660387766e-06, "loss": 1.2314, "step": 7445 }, { "epoch": 0.79, "learning_rate": 1.0699286897575718e-06, "loss": 1.3781, "step": 7446 }, { "epoch": 0.79, "learning_rate": 1.0688609825983565e-06, "loss": 1.566, "step": 7447 }, { "epoch": 0.79, "learning_rate": 1.0677937446885784e-06, "loss": 1.2916, "step": 7448 }, { "epoch": 0.79, "learning_rate": 1.066726976155632e-06, "loss": 1.6733, "step": 7449 }, { "epoch": 0.79, "learning_rate": 1.065660677126853e-06, "loss": 1.2725, "step": 7450 }, { "epoch": 0.79, "learning_rate": 1.0645948477295236e-06, "loss": 1.5669, "step": 7451 }, { "epoch": 0.79, "learning_rate": 1.0635294880908702e-06, "loss": 1.2108, "step": 7452 }, { "epoch": 0.79, "learning_rate": 1.0624645983380611e-06, "loss": 1.2213, "step": 7453 }, { "epoch": 0.8, "learning_rate": 1.061400178598207e-06, "loss": 1.1999, "step": 7454 }, { "epoch": 0.8, "learning_rate": 1.0603362289983687e-06, "loss": 1.4853, "step": 7455 }, { "epoch": 0.8, "learning_rate": 1.0592727496655431e-06, "loss": 1.2816, "step": 7456 }, { "epoch": 0.8, "learning_rate": 1.0582097407266772e-06, "loss": 1.4454, "step": 7457 }, { "epoch": 0.8, "learning_rate": 1.0571472023086604e-06, "loss": 1.4579, "step": 7458 }, { "epoch": 0.8, "learning_rate": 1.0560851345383233e-06, "loss": 1.2755, "step": 7459 }, { "epoch": 0.8, "learning_rate": 1.0550235375424416e-06, "loss": 1.5018, "step": 7460 }, { "epoch": 0.8, "learning_rate": 1.053962411447736e-06, "loss": 1.3009, "step": 7461 }, { "epoch": 0.8, "learning_rate": 1.0529017563808719e-06, "loss": 1.2842, "step": 7462 }, { "epoch": 0.8, "learning_rate": 1.051841572468454e-06, "loss": 1.1735, "step": 7463 }, { "epoch": 0.8, "learning_rate": 1.0507818598370355e-06, "loss": 1.3524, "step": 7464 }, { "epoch": 0.8, "learning_rate": 1.0497226186131104e-06, "loss": 1.2804, "step": 7465 }, { "epoch": 0.8, "learning_rate": 1.0486638489231187e-06, "loss": 1.4768, "step": 7466 }, { "epoch": 0.8, "learning_rate": 1.0476055508934408e-06, "loss": 1.364, "step": 7467 }, { "epoch": 0.8, "learning_rate": 1.0465477246504059e-06, "loss": 1.3604, "step": 7468 }, { "epoch": 0.8, "learning_rate": 1.04549037032028e-06, "loss": 1.0986, "step": 7469 }, { "epoch": 0.8, "learning_rate": 1.0444334880292794e-06, "loss": 1.1302, "step": 7470 }, { "epoch": 0.8, "learning_rate": 1.0433770779035618e-06, "loss": 1.4404, "step": 7471 }, { "epoch": 0.8, "learning_rate": 1.0423211400692274e-06, "loss": 1.2565, "step": 7472 }, { "epoch": 0.8, "learning_rate": 1.0412656746523182e-06, "loss": 1.373, "step": 7473 }, { "epoch": 0.8, "learning_rate": 1.0402106817788255e-06, "loss": 1.6301, "step": 7474 }, { "epoch": 0.8, "learning_rate": 1.0391561615746782e-06, "loss": 1.4615, "step": 7475 }, { "epoch": 0.8, "learning_rate": 1.0381021141657526e-06, "loss": 1.3127, "step": 7476 }, { "epoch": 0.8, "learning_rate": 1.0370485396778696e-06, "loss": 1.5178, "step": 7477 }, { "epoch": 0.8, "learning_rate": 1.0359954382367898e-06, "loss": 1.5589, "step": 7478 }, { "epoch": 0.8, "learning_rate": 1.0349428099682173e-06, "loss": 1.2479, "step": 7479 }, { "epoch": 0.8, "learning_rate": 1.0338906549978035e-06, "loss": 1.3759, "step": 7480 }, { "epoch": 0.8, "learning_rate": 1.0328389734511424e-06, "loss": 1.5396, "step": 7481 }, { "epoch": 0.8, "learning_rate": 1.0317877654537672e-06, "loss": 1.6518, "step": 7482 }, { "epoch": 0.8, "learning_rate": 1.0307370311311615e-06, "loss": 1.5586, "step": 7483 }, { "epoch": 0.8, "learning_rate": 1.029686770608746e-06, "loss": 1.3649, "step": 7484 }, { "epoch": 0.8, "learning_rate": 1.0286369840118859e-06, "loss": 1.5193, "step": 7485 }, { "epoch": 0.8, "learning_rate": 1.027587671465896e-06, "loss": 1.2229, "step": 7486 }, { "epoch": 0.8, "learning_rate": 1.026538833096028e-06, "loss": 1.3151, "step": 7487 }, { "epoch": 0.8, "learning_rate": 1.025490469027477e-06, "loss": 1.4058, "step": 7488 }, { "epoch": 0.8, "learning_rate": 1.0244425793853857e-06, "loss": 1.2372, "step": 7489 }, { "epoch": 0.8, "learning_rate": 1.023395164294836e-06, "loss": 1.5541, "step": 7490 }, { "epoch": 0.8, "learning_rate": 1.0223482238808557e-06, "loss": 1.2708, "step": 7491 }, { "epoch": 0.8, "learning_rate": 1.021301758268417e-06, "loss": 1.4386, "step": 7492 }, { "epoch": 0.8, "learning_rate": 1.020255767582432e-06, "loss": 1.218, "step": 7493 }, { "epoch": 0.8, "learning_rate": 1.0192102519477565e-06, "loss": 1.4384, "step": 7494 }, { "epoch": 0.8, "learning_rate": 1.0181652114891926e-06, "loss": 1.5557, "step": 7495 }, { "epoch": 0.8, "learning_rate": 1.0171206463314848e-06, "loss": 1.2676, "step": 7496 }, { "epoch": 0.8, "learning_rate": 1.016076556599318e-06, "loss": 1.4917, "step": 7497 }, { "epoch": 0.8, "learning_rate": 1.0150329424173243e-06, "loss": 1.5509, "step": 7498 }, { "epoch": 0.8, "learning_rate": 1.013989803910076e-06, "loss": 1.2257, "step": 7499 }, { "epoch": 0.8, "learning_rate": 1.0129471412020886e-06, "loss": 1.6364, "step": 7500 }, { "epoch": 0.8, "learning_rate": 1.0119049544178233e-06, "loss": 1.3686, "step": 7501 }, { "epoch": 0.8, "learning_rate": 1.010863243681684e-06, "loss": 1.2123, "step": 7502 }, { "epoch": 0.8, "learning_rate": 1.0098220091180145e-06, "loss": 1.3896, "step": 7503 }, { "epoch": 0.8, "learning_rate": 1.008781250851107e-06, "loss": 1.3606, "step": 7504 }, { "epoch": 0.8, "learning_rate": 1.007740969005191e-06, "loss": 1.3661, "step": 7505 }, { "epoch": 0.8, "learning_rate": 1.006701163704445e-06, "loss": 1.4868, "step": 7506 }, { "epoch": 0.8, "learning_rate": 1.0056618350729851e-06, "loss": 1.5448, "step": 7507 }, { "epoch": 0.8, "learning_rate": 1.0046229832348758e-06, "loss": 1.1938, "step": 7508 }, { "epoch": 0.8, "learning_rate": 1.0035846083141193e-06, "loss": 1.5578, "step": 7509 }, { "epoch": 0.8, "learning_rate": 1.0025467104346653e-06, "loss": 1.6459, "step": 7510 }, { "epoch": 0.8, "learning_rate": 1.0015092897204053e-06, "loss": 1.5154, "step": 7511 }, { "epoch": 0.8, "learning_rate": 1.0004723462951732e-06, "loss": 1.6771, "step": 7512 }, { "epoch": 0.8, "learning_rate": 9.994358802827437e-07, "loss": 1.4393, "step": 7513 }, { "epoch": 0.8, "learning_rate": 9.983998918068389e-07, "loss": 1.548, "step": 7514 }, { "epoch": 0.8, "learning_rate": 9.973643809911238e-07, "loss": 1.4471, "step": 7515 }, { "epoch": 0.8, "learning_rate": 9.963293479592006e-07, "loss": 1.2606, "step": 7516 }, { "epoch": 0.8, "learning_rate": 9.952947928346218e-07, "loss": 1.3148, "step": 7517 }, { "epoch": 0.8, "learning_rate": 9.942607157408784e-07, "loss": 1.2536, "step": 7518 }, { "epoch": 0.8, "learning_rate": 9.932271168014035e-07, "loss": 1.4777, "step": 7519 }, { "epoch": 0.8, "learning_rate": 9.921939961395767e-07, "loss": 1.2664, "step": 7520 }, { "epoch": 0.8, "learning_rate": 9.911613538787196e-07, "loss": 1.4567, "step": 7521 }, { "epoch": 0.8, "learning_rate": 9.901291901420939e-07, "loss": 1.1655, "step": 7522 }, { "epoch": 0.8, "learning_rate": 9.890975050529084e-07, "loss": 1.21, "step": 7523 }, { "epoch": 0.8, "learning_rate": 9.880662987343103e-07, "loss": 1.6856, "step": 7524 }, { "epoch": 0.8, "learning_rate": 9.870355713093942e-07, "loss": 1.4154, "step": 7525 }, { "epoch": 0.8, "learning_rate": 9.860053229011929e-07, "loss": 1.2403, "step": 7526 }, { "epoch": 0.8, "learning_rate": 9.849755536326866e-07, "loss": 1.3789, "step": 7527 }, { "epoch": 0.8, "learning_rate": 9.839462636267944e-07, "loss": 1.4786, "step": 7528 }, { "epoch": 0.8, "learning_rate": 9.829174530063795e-07, "loss": 1.318, "step": 7529 }, { "epoch": 0.8, "learning_rate": 9.818891218942511e-07, "loss": 1.4528, "step": 7530 }, { "epoch": 0.8, "learning_rate": 9.808612704131565e-07, "loss": 1.357, "step": 7531 }, { "epoch": 0.8, "learning_rate": 9.798338986857863e-07, "loss": 1.4645, "step": 7532 }, { "epoch": 0.8, "learning_rate": 9.78807006834777e-07, "loss": 1.5049, "step": 7533 }, { "epoch": 0.8, "learning_rate": 9.777805949827046e-07, "loss": 1.4592, "step": 7534 }, { "epoch": 0.8, "learning_rate": 9.767546632520892e-07, "loss": 1.501, "step": 7535 }, { "epoch": 0.8, "learning_rate": 9.757292117653955e-07, "loss": 1.4295, "step": 7536 }, { "epoch": 0.8, "learning_rate": 9.747042406450275e-07, "loss": 1.5505, "step": 7537 }, { "epoch": 0.8, "learning_rate": 9.736797500133316e-07, "loss": 1.3522, "step": 7538 }, { "epoch": 0.8, "learning_rate": 9.726557399925995e-07, "loss": 1.5634, "step": 7539 }, { "epoch": 0.8, "learning_rate": 9.71632210705067e-07, "loss": 1.4167, "step": 7540 }, { "epoch": 0.8, "learning_rate": 9.706091622729065e-07, "loss": 1.2558, "step": 7541 }, { "epoch": 0.8, "learning_rate": 9.695865948182392e-07, "loss": 1.4441, "step": 7542 }, { "epoch": 0.8, "learning_rate": 9.685645084631251e-07, "loss": 1.4971, "step": 7543 }, { "epoch": 0.8, "learning_rate": 9.675429033295652e-07, "loss": 1.3406, "step": 7544 }, { "epoch": 0.8, "learning_rate": 9.66521779539511e-07, "loss": 1.3151, "step": 7545 }, { "epoch": 0.8, "learning_rate": 9.655011372148487e-07, "loss": 1.6005, "step": 7546 }, { "epoch": 0.81, "learning_rate": 9.644809764774082e-07, "loss": 1.2517, "step": 7547 }, { "epoch": 0.81, "learning_rate": 9.63461297448966e-07, "loss": 1.594, "step": 7548 }, { "epoch": 0.81, "learning_rate": 9.624421002512358e-07, "loss": 1.4488, "step": 7549 }, { "epoch": 0.81, "learning_rate": 9.61423385005878e-07, "loss": 1.4627, "step": 7550 }, { "epoch": 0.81, "learning_rate": 9.604051518344948e-07, "loss": 1.4504, "step": 7551 }, { "epoch": 0.81, "learning_rate": 9.593874008586285e-07, "loss": 1.4408, "step": 7552 }, { "epoch": 0.81, "learning_rate": 9.583701321997645e-07, "loss": 1.1117, "step": 7553 }, { "epoch": 0.81, "learning_rate": 9.57353345979332e-07, "loss": 1.51, "step": 7554 }, { "epoch": 0.81, "learning_rate": 9.563370423187046e-07, "loss": 1.4533, "step": 7555 }, { "epoch": 0.81, "learning_rate": 9.553212213391933e-07, "loss": 1.2286, "step": 7556 }, { "epoch": 0.81, "learning_rate": 9.543058831620528e-07, "loss": 1.1804, "step": 7557 }, { "epoch": 0.81, "learning_rate": 9.532910279084822e-07, "loss": 1.4553, "step": 7558 }, { "epoch": 0.81, "learning_rate": 9.522766556996243e-07, "loss": 1.3527, "step": 7559 }, { "epoch": 0.81, "learning_rate": 9.512627666565588e-07, "loss": 1.2438, "step": 7560 }, { "epoch": 0.81, "learning_rate": 9.502493609003138e-07, "loss": 1.6504, "step": 7561 }, { "epoch": 0.81, "learning_rate": 9.492364385518554e-07, "loss": 1.487, "step": 7562 }, { "epoch": 0.81, "learning_rate": 9.482239997320903e-07, "loss": 1.2238, "step": 7563 }, { "epoch": 0.81, "learning_rate": 9.47212044561876e-07, "loss": 1.2334, "step": 7564 }, { "epoch": 0.81, "learning_rate": 9.46200573162005e-07, "loss": 1.2885, "step": 7565 }, { "epoch": 0.81, "learning_rate": 9.451895856532117e-07, "loss": 1.6477, "step": 7566 }, { "epoch": 0.81, "learning_rate": 9.441790821561781e-07, "loss": 1.4143, "step": 7567 }, { "epoch": 0.81, "learning_rate": 9.431690627915218e-07, "loss": 1.1778, "step": 7568 }, { "epoch": 0.81, "learning_rate": 9.421595276798084e-07, "loss": 1.2882, "step": 7569 }, { "epoch": 0.81, "learning_rate": 9.411504769415447e-07, "loss": 1.4446, "step": 7570 }, { "epoch": 0.81, "learning_rate": 9.401419106971765e-07, "loss": 1.3375, "step": 7571 }, { "epoch": 0.81, "learning_rate": 9.39133829067092e-07, "loss": 1.3006, "step": 7572 }, { "epoch": 0.81, "learning_rate": 9.381262321716255e-07, "loss": 1.2703, "step": 7573 }, { "epoch": 0.81, "learning_rate": 9.371191201310515e-07, "loss": 1.1738, "step": 7574 }, { "epoch": 0.81, "learning_rate": 9.361124930655841e-07, "loss": 1.4523, "step": 7575 }, { "epoch": 0.81, "learning_rate": 9.351063510953845e-07, "loss": 1.4566, "step": 7576 }, { "epoch": 0.81, "learning_rate": 9.341006943405511e-07, "loss": 1.8693, "step": 7577 }, { "epoch": 0.81, "learning_rate": 9.330955229211259e-07, "loss": 1.2228, "step": 7578 }, { "epoch": 0.81, "learning_rate": 9.320908369570941e-07, "loss": 1.1901, "step": 7579 }, { "epoch": 0.81, "learning_rate": 9.310866365683846e-07, "loss": 1.3844, "step": 7580 }, { "epoch": 0.81, "learning_rate": 9.300829218748625e-07, "loss": 1.3322, "step": 7581 }, { "epoch": 0.81, "learning_rate": 9.290796929963414e-07, "loss": 1.3836, "step": 7582 }, { "epoch": 0.81, "learning_rate": 9.280769500525716e-07, "loss": 1.4502, "step": 7583 }, { "epoch": 0.81, "learning_rate": 9.270746931632501e-07, "loss": 1.0487, "step": 7584 }, { "epoch": 0.81, "learning_rate": 9.260729224480114e-07, "loss": 1.5224, "step": 7585 }, { "epoch": 0.81, "learning_rate": 9.250716380264368e-07, "loss": 1.4761, "step": 7586 }, { "epoch": 0.81, "learning_rate": 9.240708400180437e-07, "loss": 1.4033, "step": 7587 }, { "epoch": 0.81, "learning_rate": 9.23070528542297e-07, "loss": 1.37, "step": 7588 }, { "epoch": 0.81, "learning_rate": 9.220707037186011e-07, "loss": 1.4235, "step": 7589 }, { "epoch": 0.81, "learning_rate": 9.210713656663023e-07, "loss": 1.1827, "step": 7590 }, { "epoch": 0.81, "learning_rate": 9.200725145046868e-07, "loss": 1.5531, "step": 7591 }, { "epoch": 0.81, "learning_rate": 9.190741503529876e-07, "loss": 1.3072, "step": 7592 }, { "epoch": 0.81, "learning_rate": 9.180762733303745e-07, "loss": 1.3892, "step": 7593 }, { "epoch": 0.81, "learning_rate": 9.170788835559624e-07, "loss": 1.3957, "step": 7594 }, { "epoch": 0.81, "learning_rate": 9.160819811488081e-07, "loss": 1.3314, "step": 7595 }, { "epoch": 0.81, "learning_rate": 9.150855662279079e-07, "loss": 1.5003, "step": 7596 }, { "epoch": 0.81, "learning_rate": 9.14089638912199e-07, "loss": 1.5315, "step": 7597 }, { "epoch": 0.81, "learning_rate": 9.130941993205672e-07, "loss": 1.4312, "step": 7598 }, { "epoch": 0.81, "learning_rate": 9.120992475718333e-07, "loss": 1.4702, "step": 7599 }, { "epoch": 0.81, "learning_rate": 9.111047837847603e-07, "loss": 1.4905, "step": 7600 }, { "epoch": 0.81, "learning_rate": 9.101108080780574e-07, "loss": 1.4853, "step": 7601 }, { "epoch": 0.81, "learning_rate": 9.091173205703708e-07, "loss": 1.3734, "step": 7602 }, { "epoch": 0.81, "learning_rate": 9.081243213802904e-07, "loss": 1.1115, "step": 7603 }, { "epoch": 0.81, "learning_rate": 9.071318106263499e-07, "loss": 1.5031, "step": 7604 }, { "epoch": 0.81, "learning_rate": 9.061397884270217e-07, "loss": 1.7396, "step": 7605 }, { "epoch": 0.81, "learning_rate": 9.05148254900719e-07, "loss": 1.3964, "step": 7606 }, { "epoch": 0.81, "learning_rate": 9.041572101657997e-07, "loss": 1.1447, "step": 7607 }, { "epoch": 0.81, "learning_rate": 9.031666543405637e-07, "loss": 1.2987, "step": 7608 }, { "epoch": 0.81, "learning_rate": 9.021765875432498e-07, "loss": 1.4165, "step": 7609 }, { "epoch": 0.81, "learning_rate": 9.011870098920378e-07, "loss": 1.2779, "step": 7610 }, { "epoch": 0.81, "learning_rate": 9.001979215050544e-07, "loss": 1.3453, "step": 7611 }, { "epoch": 0.81, "learning_rate": 8.992093225003606e-07, "loss": 1.342, "step": 7612 }, { "epoch": 0.81, "learning_rate": 8.98221212995965e-07, "loss": 1.1576, "step": 7613 }, { "epoch": 0.81, "learning_rate": 8.972335931098159e-07, "loss": 1.438, "step": 7614 }, { "epoch": 0.81, "learning_rate": 8.962464629598028e-07, "loss": 1.1909, "step": 7615 }, { "epoch": 0.81, "learning_rate": 8.952598226637549e-07, "loss": 1.3151, "step": 7616 }, { "epoch": 0.81, "learning_rate": 8.942736723394458e-07, "loss": 1.3823, "step": 7617 }, { "epoch": 0.81, "learning_rate": 8.932880121045911e-07, "loss": 1.4507, "step": 7618 }, { "epoch": 0.81, "learning_rate": 8.923028420768437e-07, "loss": 1.3978, "step": 7619 }, { "epoch": 0.81, "learning_rate": 8.913181623738032e-07, "loss": 1.3207, "step": 7620 }, { "epoch": 0.81, "learning_rate": 8.903339731130073e-07, "loss": 1.4779, "step": 7621 }, { "epoch": 0.81, "learning_rate": 8.893502744119337e-07, "loss": 1.0417, "step": 7622 }, { "epoch": 0.81, "learning_rate": 8.883670663880078e-07, "loss": 1.5106, "step": 7623 }, { "epoch": 0.81, "learning_rate": 8.87384349158591e-07, "loss": 1.4622, "step": 7624 }, { "epoch": 0.81, "learning_rate": 8.864021228409853e-07, "loss": 1.2597, "step": 7625 }, { "epoch": 0.81, "learning_rate": 8.854203875524403e-07, "loss": 1.4965, "step": 7626 }, { "epoch": 0.81, "learning_rate": 8.84439143410139e-07, "loss": 1.3439, "step": 7627 }, { "epoch": 0.81, "learning_rate": 8.834583905312127e-07, "loss": 1.3857, "step": 7628 }, { "epoch": 0.81, "learning_rate": 8.824781290327317e-07, "loss": 1.2385, "step": 7629 }, { "epoch": 0.81, "learning_rate": 8.814983590317055e-07, "loss": 1.2855, "step": 7630 }, { "epoch": 0.81, "learning_rate": 8.805190806450858e-07, "loss": 1.6649, "step": 7631 }, { "epoch": 0.81, "learning_rate": 8.795402939897679e-07, "loss": 1.5648, "step": 7632 }, { "epoch": 0.81, "learning_rate": 8.785619991825877e-07, "loss": 1.3533, "step": 7633 }, { "epoch": 0.81, "learning_rate": 8.775841963403198e-07, "loss": 1.4447, "step": 7634 }, { "epoch": 0.81, "learning_rate": 8.766068855796833e-07, "loss": 1.5617, "step": 7635 }, { "epoch": 0.81, "learning_rate": 8.756300670173368e-07, "loss": 1.4652, "step": 7636 }, { "epoch": 0.81, "learning_rate": 8.746537407698785e-07, "loss": 1.2261, "step": 7637 }, { "epoch": 0.81, "learning_rate": 8.736779069538521e-07, "loss": 1.3685, "step": 7638 }, { "epoch": 0.81, "learning_rate": 8.727025656857407e-07, "loss": 1.4224, "step": 7639 }, { "epoch": 0.81, "learning_rate": 8.717277170819655e-07, "loss": 1.4796, "step": 7640 }, { "epoch": 0.82, "learning_rate": 8.707533612588948e-07, "loss": 1.1982, "step": 7641 }, { "epoch": 0.82, "learning_rate": 8.697794983328317e-07, "loss": 1.212, "step": 7642 }, { "epoch": 0.82, "learning_rate": 8.688061284200266e-07, "loss": 1.3887, "step": 7643 }, { "epoch": 0.82, "learning_rate": 8.67833251636665e-07, "loss": 1.1641, "step": 7644 }, { "epoch": 0.82, "learning_rate": 8.668608680988794e-07, "loss": 1.5222, "step": 7645 }, { "epoch": 0.82, "learning_rate": 8.658889779227376e-07, "loss": 1.4968, "step": 7646 }, { "epoch": 0.82, "learning_rate": 8.649175812242532e-07, "loss": 1.2943, "step": 7647 }, { "epoch": 0.82, "learning_rate": 8.639466781193812e-07, "loss": 1.2874, "step": 7648 }, { "epoch": 0.82, "learning_rate": 8.629762687240129e-07, "loss": 1.3961, "step": 7649 }, { "epoch": 0.82, "learning_rate": 8.62006353153983e-07, "loss": 1.4499, "step": 7650 }, { "epoch": 0.82, "learning_rate": 8.610369315250694e-07, "loss": 1.5229, "step": 7651 }, { "epoch": 0.82, "learning_rate": 8.600680039529896e-07, "loss": 1.3039, "step": 7652 }, { "epoch": 0.82, "learning_rate": 8.590995705533994e-07, "loss": 1.5605, "step": 7653 }, { "epoch": 0.82, "learning_rate": 8.581316314419019e-07, "loss": 1.3625, "step": 7654 }, { "epoch": 0.82, "learning_rate": 8.571641867340346e-07, "loss": 1.5605, "step": 7655 }, { "epoch": 0.82, "learning_rate": 8.561972365452775e-07, "loss": 1.6256, "step": 7656 }, { "epoch": 0.82, "learning_rate": 8.552307809910565e-07, "loss": 1.3258, "step": 7657 }, { "epoch": 0.82, "learning_rate": 8.542648201867337e-07, "loss": 1.4621, "step": 7658 }, { "epoch": 0.82, "learning_rate": 8.532993542476108e-07, "loss": 1.4906, "step": 7659 }, { "epoch": 0.82, "learning_rate": 8.523343832889358e-07, "loss": 1.3723, "step": 7660 }, { "epoch": 0.82, "learning_rate": 8.513699074258924e-07, "loss": 1.5768, "step": 7661 }, { "epoch": 0.82, "learning_rate": 8.504059267736097e-07, "loss": 1.1447, "step": 7662 }, { "epoch": 0.82, "learning_rate": 8.494424414471536e-07, "loss": 1.1645, "step": 7663 }, { "epoch": 0.82, "learning_rate": 8.484794515615341e-07, "loss": 1.3002, "step": 7664 }, { "epoch": 0.82, "learning_rate": 8.475169572316988e-07, "loss": 1.4424, "step": 7665 }, { "epoch": 0.82, "learning_rate": 8.465549585725397e-07, "loss": 1.6113, "step": 7666 }, { "epoch": 0.82, "learning_rate": 8.455934556988888e-07, "loss": 1.2208, "step": 7667 }, { "epoch": 0.82, "learning_rate": 8.446324487255164e-07, "loss": 1.4158, "step": 7668 }, { "epoch": 0.82, "learning_rate": 8.436719377671349e-07, "loss": 1.3343, "step": 7669 }, { "epoch": 0.82, "learning_rate": 8.427119229384001e-07, "loss": 1.2348, "step": 7670 }, { "epoch": 0.82, "learning_rate": 8.417524043539038e-07, "loss": 1.5768, "step": 7671 }, { "epoch": 0.82, "learning_rate": 8.407933821281822e-07, "loss": 1.4803, "step": 7672 }, { "epoch": 0.82, "learning_rate": 8.398348563757125e-07, "loss": 1.5588, "step": 7673 }, { "epoch": 0.82, "learning_rate": 8.388768272109105e-07, "loss": 1.7392, "step": 7674 }, { "epoch": 0.82, "learning_rate": 8.379192947481319e-07, "loss": 1.2883, "step": 7675 }, { "epoch": 0.82, "learning_rate": 8.369622591016762e-07, "loss": 1.3563, "step": 7676 }, { "epoch": 0.82, "learning_rate": 8.36005720385783e-07, "loss": 1.2263, "step": 7677 }, { "epoch": 0.82, "learning_rate": 8.350496787146301e-07, "loss": 1.5512, "step": 7678 }, { "epoch": 0.82, "learning_rate": 8.34094134202339e-07, "loss": 1.2888, "step": 7679 }, { "epoch": 0.82, "learning_rate": 8.331390869629702e-07, "loss": 1.5875, "step": 7680 }, { "epoch": 0.82, "learning_rate": 8.321845371105225e-07, "loss": 1.6629, "step": 7681 }, { "epoch": 0.82, "learning_rate": 8.31230484758942e-07, "loss": 1.346, "step": 7682 }, { "epoch": 0.82, "learning_rate": 8.302769300221098e-07, "loss": 1.4335, "step": 7683 }, { "epoch": 0.82, "learning_rate": 8.293238730138475e-07, "loss": 1.2328, "step": 7684 }, { "epoch": 0.82, "learning_rate": 8.283713138479216e-07, "loss": 1.243, "step": 7685 }, { "epoch": 0.82, "learning_rate": 8.274192526380337e-07, "loss": 1.2563, "step": 7686 }, { "epoch": 0.82, "learning_rate": 8.264676894978307e-07, "loss": 1.546, "step": 7687 }, { "epoch": 0.82, "learning_rate": 8.255166245408985e-07, "loss": 1.2894, "step": 7688 }, { "epoch": 0.82, "learning_rate": 8.24566057880763e-07, "loss": 1.3855, "step": 7689 }, { "epoch": 0.82, "learning_rate": 8.236159896308882e-07, "loss": 1.5542, "step": 7690 }, { "epoch": 0.82, "learning_rate": 8.226664199046835e-07, "loss": 1.2879, "step": 7691 }, { "epoch": 0.82, "learning_rate": 8.217173488154972e-07, "loss": 1.2516, "step": 7692 }, { "epoch": 0.82, "learning_rate": 8.207687764766148e-07, "loss": 1.5631, "step": 7693 }, { "epoch": 0.82, "learning_rate": 8.198207030012678e-07, "loss": 1.657, "step": 7694 }, { "epoch": 0.82, "learning_rate": 8.188731285026219e-07, "loss": 1.5336, "step": 7695 }, { "epoch": 0.82, "learning_rate": 8.179260530937899e-07, "loss": 1.6495, "step": 7696 }, { "epoch": 0.82, "learning_rate": 8.169794768878186e-07, "loss": 1.4716, "step": 7697 }, { "epoch": 0.82, "learning_rate": 8.160333999977004e-07, "loss": 1.5425, "step": 7698 }, { "epoch": 0.82, "learning_rate": 8.150878225363656e-07, "loss": 1.3847, "step": 7699 }, { "epoch": 0.82, "learning_rate": 8.141427446166822e-07, "loss": 1.3444, "step": 7700 }, { "epoch": 0.82, "learning_rate": 8.131981663514665e-07, "loss": 1.4215, "step": 7701 }, { "epoch": 0.82, "learning_rate": 8.122540878534679e-07, "loss": 1.4208, "step": 7702 }, { "epoch": 0.82, "learning_rate": 8.113105092353774e-07, "loss": 1.3841, "step": 7703 }, { "epoch": 0.82, "learning_rate": 8.103674306098291e-07, "loss": 1.3154, "step": 7704 }, { "epoch": 0.82, "learning_rate": 8.094248520893949e-07, "loss": 1.449, "step": 7705 }, { "epoch": 0.82, "learning_rate": 8.084827737865875e-07, "loss": 1.3171, "step": 7706 }, { "epoch": 0.82, "learning_rate": 8.075411958138623e-07, "loss": 1.5417, "step": 7707 }, { "epoch": 0.82, "learning_rate": 8.066001182836114e-07, "loss": 1.2949, "step": 7708 }, { "epoch": 0.82, "learning_rate": 8.056595413081675e-07, "loss": 1.2446, "step": 7709 }, { "epoch": 0.82, "learning_rate": 8.047194649998063e-07, "loss": 1.5207, "step": 7710 }, { "epoch": 0.82, "learning_rate": 8.037798894707427e-07, "loss": 1.175, "step": 7711 }, { "epoch": 0.82, "learning_rate": 8.028408148331296e-07, "loss": 1.3132, "step": 7712 }, { "epoch": 0.82, "learning_rate": 8.019022411990634e-07, "loss": 1.2341, "step": 7713 }, { "epoch": 0.82, "learning_rate": 8.009641686805786e-07, "loss": 1.2594, "step": 7714 }, { "epoch": 0.82, "learning_rate": 8.000265973896487e-07, "loss": 1.2453, "step": 7715 }, { "epoch": 0.82, "learning_rate": 7.99089527438191e-07, "loss": 1.2811, "step": 7716 }, { "epoch": 0.82, "learning_rate": 7.98152958938061e-07, "loss": 1.5083, "step": 7717 }, { "epoch": 0.82, "learning_rate": 7.972168920010531e-07, "loss": 1.2971, "step": 7718 }, { "epoch": 0.82, "learning_rate": 7.962813267389052e-07, "loss": 1.3096, "step": 7719 }, { "epoch": 0.82, "learning_rate": 7.953462632632903e-07, "loss": 1.3153, "step": 7720 }, { "epoch": 0.82, "learning_rate": 7.944117016858272e-07, "loss": 1.6234, "step": 7721 }, { "epoch": 0.82, "learning_rate": 7.93477642118069e-07, "loss": 1.435, "step": 7722 }, { "epoch": 0.82, "learning_rate": 7.925440846715154e-07, "loss": 1.2816, "step": 7723 }, { "epoch": 0.82, "learning_rate": 7.916110294575996e-07, "loss": 1.3187, "step": 7724 }, { "epoch": 0.82, "learning_rate": 7.906784765876985e-07, "loss": 1.3795, "step": 7725 }, { "epoch": 0.82, "learning_rate": 7.897464261731308e-07, "loss": 1.4483, "step": 7726 }, { "epoch": 0.82, "learning_rate": 7.888148783251509e-07, "loss": 1.74, "step": 7727 }, { "epoch": 0.82, "learning_rate": 7.878838331549538e-07, "loss": 1.3497, "step": 7728 }, { "epoch": 0.82, "learning_rate": 7.869532907736787e-07, "loss": 1.5357, "step": 7729 }, { "epoch": 0.82, "learning_rate": 7.860232512923993e-07, "loss": 1.4845, "step": 7730 }, { "epoch": 0.82, "learning_rate": 7.850937148221332e-07, "loss": 1.3643, "step": 7731 }, { "epoch": 0.82, "learning_rate": 7.841646814738374e-07, "loss": 1.2516, "step": 7732 }, { "epoch": 0.82, "learning_rate": 7.83236151358408e-07, "loss": 1.2082, "step": 7733 }, { "epoch": 0.82, "learning_rate": 7.823081245866776e-07, "loss": 1.3724, "step": 7734 }, { "epoch": 0.83, "learning_rate": 7.813806012694275e-07, "loss": 1.411, "step": 7735 }, { "epoch": 0.83, "learning_rate": 7.804535815173714e-07, "loss": 1.2227, "step": 7736 }, { "epoch": 0.83, "learning_rate": 7.795270654411635e-07, "loss": 1.1977, "step": 7737 }, { "epoch": 0.83, "learning_rate": 7.786010531514026e-07, "loss": 1.5015, "step": 7738 }, { "epoch": 0.83, "learning_rate": 7.776755447586209e-07, "loss": 1.2915, "step": 7739 }, { "epoch": 0.83, "learning_rate": 7.767505403732961e-07, "loss": 1.7489, "step": 7740 }, { "epoch": 0.83, "learning_rate": 7.758260401058442e-07, "loss": 1.5377, "step": 7741 }, { "epoch": 0.83, "learning_rate": 7.749020440666188e-07, "loss": 1.4179, "step": 7742 }, { "epoch": 0.83, "learning_rate": 7.739785523659144e-07, "loss": 1.3684, "step": 7743 }, { "epoch": 0.83, "learning_rate": 7.73055565113966e-07, "loss": 1.3503, "step": 7744 }, { "epoch": 0.83, "learning_rate": 7.721330824209495e-07, "loss": 1.4384, "step": 7745 }, { "epoch": 0.83, "learning_rate": 7.712111043969772e-07, "loss": 1.2786, "step": 7746 }, { "epoch": 0.83, "learning_rate": 7.702896311521052e-07, "loss": 1.3058, "step": 7747 }, { "epoch": 0.83, "learning_rate": 7.69368662796326e-07, "loss": 1.6006, "step": 7748 }, { "epoch": 0.83, "learning_rate": 7.684481994395726e-07, "loss": 1.291, "step": 7749 }, { "epoch": 0.83, "learning_rate": 7.675282411917184e-07, "loss": 1.401, "step": 7750 }, { "epoch": 0.83, "learning_rate": 7.666087881625778e-07, "loss": 1.2998, "step": 7751 }, { "epoch": 0.83, "learning_rate": 7.656898404619029e-07, "loss": 1.3851, "step": 7752 }, { "epoch": 0.83, "learning_rate": 7.647713981993843e-07, "loss": 1.4229, "step": 7753 }, { "epoch": 0.83, "learning_rate": 7.638534614846548e-07, "loss": 1.3886, "step": 7754 }, { "epoch": 0.83, "learning_rate": 7.629360304272882e-07, "loss": 1.5631, "step": 7755 }, { "epoch": 0.83, "learning_rate": 7.620191051367926e-07, "loss": 1.1855, "step": 7756 }, { "epoch": 0.83, "learning_rate": 7.611026857226212e-07, "loss": 1.3259, "step": 7757 }, { "epoch": 0.83, "learning_rate": 7.601867722941642e-07, "loss": 1.6121, "step": 7758 }, { "epoch": 0.83, "learning_rate": 7.59271364960748e-07, "loss": 1.4246, "step": 7759 }, { "epoch": 0.83, "learning_rate": 7.583564638316481e-07, "loss": 1.4783, "step": 7760 }, { "epoch": 0.83, "learning_rate": 7.57442069016071e-07, "loss": 1.3511, "step": 7761 }, { "epoch": 0.83, "learning_rate": 7.565281806231645e-07, "loss": 1.3852, "step": 7762 }, { "epoch": 0.83, "learning_rate": 7.556147987620193e-07, "loss": 1.5098, "step": 7763 }, { "epoch": 0.83, "learning_rate": 7.547019235416609e-07, "loss": 1.0703, "step": 7764 }, { "epoch": 0.83, "learning_rate": 7.537895550710583e-07, "loss": 1.5739, "step": 7765 }, { "epoch": 0.83, "learning_rate": 7.528776934591186e-07, "loss": 1.7112, "step": 7766 }, { "epoch": 0.83, "learning_rate": 7.519663388146886e-07, "loss": 1.5542, "step": 7767 }, { "epoch": 0.83, "learning_rate": 7.510554912465517e-07, "loss": 1.426, "step": 7768 }, { "epoch": 0.83, "learning_rate": 7.50145150863435e-07, "loss": 1.6202, "step": 7769 }, { "epoch": 0.83, "learning_rate": 7.492353177740047e-07, "loss": 1.6426, "step": 7770 }, { "epoch": 0.83, "learning_rate": 7.48325992086863e-07, "loss": 1.3935, "step": 7771 }, { "epoch": 0.83, "learning_rate": 7.474171739105551e-07, "loss": 1.4936, "step": 7772 }, { "epoch": 0.83, "learning_rate": 7.465088633535639e-07, "loss": 1.2608, "step": 7773 }, { "epoch": 0.83, "learning_rate": 7.456010605243103e-07, "loss": 1.3212, "step": 7774 }, { "epoch": 0.83, "learning_rate": 7.446937655311576e-07, "loss": 1.1847, "step": 7775 }, { "epoch": 0.83, "learning_rate": 7.437869784824086e-07, "loss": 1.3896, "step": 7776 }, { "epoch": 0.83, "learning_rate": 7.428806994863014e-07, "loss": 1.3937, "step": 7777 }, { "epoch": 0.83, "learning_rate": 7.41974928651017e-07, "loss": 1.4701, "step": 7778 }, { "epoch": 0.83, "learning_rate": 7.410696660846761e-07, "loss": 1.5474, "step": 7779 }, { "epoch": 0.83, "learning_rate": 7.401649118953364e-07, "loss": 1.5577, "step": 7780 }, { "epoch": 0.83, "learning_rate": 7.392606661909946e-07, "loss": 1.5285, "step": 7781 }, { "epoch": 0.83, "learning_rate": 7.383569290795911e-07, "loss": 1.3861, "step": 7782 }, { "epoch": 0.83, "learning_rate": 7.374537006689991e-07, "loss": 1.3229, "step": 7783 }, { "epoch": 0.83, "learning_rate": 7.365509810670369e-07, "loss": 1.3894, "step": 7784 }, { "epoch": 0.83, "learning_rate": 7.356487703814602e-07, "loss": 1.3169, "step": 7785 }, { "epoch": 0.83, "learning_rate": 7.34747068719962e-07, "loss": 1.2974, "step": 7786 }, { "epoch": 0.83, "learning_rate": 7.338458761901757e-07, "loss": 1.3604, "step": 7787 }, { "epoch": 0.83, "learning_rate": 7.329451928996745e-07, "loss": 1.3721, "step": 7788 }, { "epoch": 0.83, "learning_rate": 7.320450189559719e-07, "loss": 1.6974, "step": 7789 }, { "epoch": 0.83, "learning_rate": 7.311453544665175e-07, "loss": 1.272, "step": 7790 }, { "epoch": 0.83, "learning_rate": 7.302461995387033e-07, "loss": 1.4601, "step": 7791 }, { "epoch": 0.83, "learning_rate": 7.293475542798589e-07, "loss": 1.4484, "step": 7792 }, { "epoch": 0.83, "learning_rate": 7.284494187972496e-07, "loss": 1.3438, "step": 7793 }, { "epoch": 0.83, "learning_rate": 7.275517931980886e-07, "loss": 1.385, "step": 7794 }, { "epoch": 0.83, "learning_rate": 7.266546775895212e-07, "loss": 1.23, "step": 7795 }, { "epoch": 0.83, "learning_rate": 7.257580720786317e-07, "loss": 1.3837, "step": 7796 }, { "epoch": 0.83, "learning_rate": 7.24861976772448e-07, "loss": 1.5837, "step": 7797 }, { "epoch": 0.83, "learning_rate": 7.239663917779327e-07, "loss": 1.3369, "step": 7798 }, { "epoch": 0.83, "learning_rate": 7.230713172019899e-07, "loss": 1.455, "step": 7799 }, { "epoch": 0.83, "learning_rate": 7.22176753151464e-07, "loss": 1.2587, "step": 7800 }, { "epoch": 0.83, "learning_rate": 7.212826997331352e-07, "loss": 1.4653, "step": 7801 }, { "epoch": 0.83, "learning_rate": 7.203891570537225e-07, "loss": 1.453, "step": 7802 }, { "epoch": 0.83, "learning_rate": 7.194961252198885e-07, "loss": 1.6991, "step": 7803 }, { "epoch": 0.83, "learning_rate": 7.186036043382311e-07, "loss": 1.1444, "step": 7804 }, { "epoch": 0.83, "learning_rate": 7.177115945152885e-07, "loss": 1.3751, "step": 7805 }, { "epoch": 0.83, "learning_rate": 7.168200958575361e-07, "loss": 1.2795, "step": 7806 }, { "epoch": 0.83, "learning_rate": 7.15929108471391e-07, "loss": 1.4008, "step": 7807 }, { "epoch": 0.83, "learning_rate": 7.150386324632069e-07, "loss": 1.5326, "step": 7808 }, { "epoch": 0.83, "learning_rate": 7.141486679392778e-07, "loss": 1.6145, "step": 7809 }, { "epoch": 0.83, "learning_rate": 7.132592150058382e-07, "loss": 1.3591, "step": 7810 }, { "epoch": 0.83, "learning_rate": 7.123702737690574e-07, "loss": 1.6471, "step": 7811 }, { "epoch": 0.83, "learning_rate": 7.114818443350463e-07, "loss": 1.2316, "step": 7812 }, { "epoch": 0.83, "learning_rate": 7.105939268098544e-07, "loss": 1.5706, "step": 7813 }, { "epoch": 0.83, "learning_rate": 7.097065212994714e-07, "loss": 1.4568, "step": 7814 }, { "epoch": 0.83, "learning_rate": 7.088196279098225e-07, "loss": 1.5636, "step": 7815 }, { "epoch": 0.83, "learning_rate": 7.079332467467753e-07, "loss": 1.3458, "step": 7816 }, { "epoch": 0.83, "learning_rate": 7.070473779161346e-07, "loss": 1.6102, "step": 7817 }, { "epoch": 0.83, "learning_rate": 7.061620215236415e-07, "loss": 1.4138, "step": 7818 }, { "epoch": 0.83, "learning_rate": 7.052771776749823e-07, "loss": 1.3568, "step": 7819 }, { "epoch": 0.83, "learning_rate": 7.043928464757776e-07, "loss": 1.2956, "step": 7820 }, { "epoch": 0.83, "learning_rate": 7.035090280315854e-07, "loss": 1.3001, "step": 7821 }, { "epoch": 0.83, "learning_rate": 7.026257224479067e-07, "loss": 1.4716, "step": 7822 }, { "epoch": 0.83, "learning_rate": 7.017429298301792e-07, "loss": 1.6535, "step": 7823 }, { "epoch": 0.83, "learning_rate": 7.008606502837784e-07, "loss": 1.5427, "step": 7824 }, { "epoch": 0.83, "learning_rate": 6.999788839140214e-07, "loss": 1.3677, "step": 7825 }, { "epoch": 0.83, "learning_rate": 6.99097630826161e-07, "loss": 1.1369, "step": 7826 }, { "epoch": 0.83, "learning_rate": 6.982168911253895e-07, "loss": 1.5804, "step": 7827 }, { "epoch": 0.83, "learning_rate": 6.973366649168389e-07, "loss": 1.1945, "step": 7828 }, { "epoch": 0.84, "learning_rate": 6.964569523055808e-07, "loss": 1.2601, "step": 7829 }, { "epoch": 0.84, "learning_rate": 6.955777533966212e-07, "loss": 1.4082, "step": 7830 }, { "epoch": 0.84, "learning_rate": 6.946990682949111e-07, "loss": 1.3533, "step": 7831 }, { "epoch": 0.84, "learning_rate": 6.938208971053345e-07, "loss": 1.4427, "step": 7832 }, { "epoch": 0.84, "learning_rate": 6.929432399327174e-07, "loss": 1.4071, "step": 7833 }, { "epoch": 0.84, "learning_rate": 6.920660968818216e-07, "loss": 1.4649, "step": 7834 }, { "epoch": 0.84, "learning_rate": 6.911894680573522e-07, "loss": 1.3179, "step": 7835 }, { "epoch": 0.84, "learning_rate": 6.903133535639467e-07, "loss": 1.4401, "step": 7836 }, { "epoch": 0.84, "learning_rate": 6.89437753506187e-07, "loss": 1.4896, "step": 7837 }, { "epoch": 0.84, "learning_rate": 6.88562667988591e-07, "loss": 1.3439, "step": 7838 }, { "epoch": 0.84, "learning_rate": 6.876880971156147e-07, "loss": 1.6745, "step": 7839 }, { "epoch": 0.84, "learning_rate": 6.868140409916518e-07, "loss": 1.5603, "step": 7840 }, { "epoch": 0.84, "learning_rate": 6.859404997210384e-07, "loss": 1.0298, "step": 7841 }, { "epoch": 0.84, "learning_rate": 6.850674734080454e-07, "loss": 1.3648, "step": 7842 }, { "epoch": 0.84, "learning_rate": 6.841949621568833e-07, "loss": 1.2747, "step": 7843 }, { "epoch": 0.84, "learning_rate": 6.833229660717033e-07, "loss": 1.5178, "step": 7844 }, { "epoch": 0.84, "learning_rate": 6.824514852565922e-07, "loss": 1.2873, "step": 7845 }, { "epoch": 0.84, "learning_rate": 6.815805198155745e-07, "loss": 1.6132, "step": 7846 }, { "epoch": 0.84, "learning_rate": 6.80710069852617e-07, "loss": 1.2509, "step": 7847 }, { "epoch": 0.84, "learning_rate": 6.798401354716233e-07, "loss": 1.3666, "step": 7848 }, { "epoch": 0.84, "learning_rate": 6.789707167764337e-07, "loss": 1.4269, "step": 7849 }, { "epoch": 0.84, "learning_rate": 6.7810181387083e-07, "loss": 1.443, "step": 7850 }, { "epoch": 0.84, "learning_rate": 6.772334268585296e-07, "loss": 1.4612, "step": 7851 }, { "epoch": 0.84, "learning_rate": 6.763655558431875e-07, "loss": 1.3322, "step": 7852 }, { "epoch": 0.84, "learning_rate": 6.754982009284039e-07, "loss": 1.3747, "step": 7853 }, { "epoch": 0.84, "learning_rate": 6.746313622177097e-07, "loss": 1.2537, "step": 7854 }, { "epoch": 0.84, "learning_rate": 6.737650398145762e-07, "loss": 1.4295, "step": 7855 }, { "epoch": 0.84, "learning_rate": 6.728992338224166e-07, "loss": 1.1816, "step": 7856 }, { "epoch": 0.84, "learning_rate": 6.720339443445772e-07, "loss": 1.4137, "step": 7857 }, { "epoch": 0.84, "learning_rate": 6.711691714843476e-07, "loss": 1.4762, "step": 7858 }, { "epoch": 0.84, "learning_rate": 6.703049153449509e-07, "loss": 1.3444, "step": 7859 }, { "epoch": 0.84, "learning_rate": 6.694411760295538e-07, "loss": 1.3638, "step": 7860 }, { "epoch": 0.84, "learning_rate": 6.685779536412563e-07, "loss": 1.3058, "step": 7861 }, { "epoch": 0.84, "learning_rate": 6.67715248283099e-07, "loss": 1.1421, "step": 7862 }, { "epoch": 0.84, "learning_rate": 6.66853060058063e-07, "loss": 1.5066, "step": 7863 }, { "epoch": 0.84, "learning_rate": 6.65991389069064e-07, "loss": 1.4639, "step": 7864 }, { "epoch": 0.84, "learning_rate": 6.651302354189559e-07, "loss": 1.4012, "step": 7865 }, { "epoch": 0.84, "learning_rate": 6.642695992105347e-07, "loss": 1.3436, "step": 7866 }, { "epoch": 0.84, "learning_rate": 6.634094805465296e-07, "loss": 1.3575, "step": 7867 }, { "epoch": 0.84, "learning_rate": 6.625498795296125e-07, "loss": 1.1942, "step": 7868 }, { "epoch": 0.84, "learning_rate": 6.61690796262392e-07, "loss": 1.2683, "step": 7869 }, { "epoch": 0.84, "learning_rate": 6.608322308474141e-07, "loss": 1.216, "step": 7870 }, { "epoch": 0.84, "learning_rate": 6.59974183387161e-07, "loss": 1.2105, "step": 7871 }, { "epoch": 0.84, "learning_rate": 6.591166539840599e-07, "loss": 1.0533, "step": 7872 }, { "epoch": 0.84, "learning_rate": 6.582596427404692e-07, "loss": 1.6035, "step": 7873 }, { "epoch": 0.84, "learning_rate": 6.574031497586874e-07, "loss": 1.6788, "step": 7874 }, { "epoch": 0.84, "learning_rate": 6.565471751409541e-07, "loss": 1.534, "step": 7875 }, { "epoch": 0.84, "learning_rate": 6.556917189894418e-07, "loss": 1.3701, "step": 7876 }, { "epoch": 0.84, "learning_rate": 6.548367814062656e-07, "loss": 1.5219, "step": 7877 }, { "epoch": 0.84, "learning_rate": 6.539823624934777e-07, "loss": 1.3089, "step": 7878 }, { "epoch": 0.84, "learning_rate": 6.53128462353067e-07, "loss": 1.276, "step": 7879 }, { "epoch": 0.84, "learning_rate": 6.522750810869604e-07, "loss": 1.4638, "step": 7880 }, { "epoch": 0.84, "learning_rate": 6.514222187970248e-07, "loss": 1.3315, "step": 7881 }, { "epoch": 0.84, "learning_rate": 6.505698755850648e-07, "loss": 0.991, "step": 7882 }, { "epoch": 0.84, "learning_rate": 6.497180515528201e-07, "loss": 1.5296, "step": 7883 }, { "epoch": 0.84, "learning_rate": 6.488667468019727e-07, "loss": 1.2733, "step": 7884 }, { "epoch": 0.84, "learning_rate": 6.480159614341402e-07, "loss": 1.4295, "step": 7885 }, { "epoch": 0.84, "learning_rate": 6.471656955508771e-07, "loss": 1.4412, "step": 7886 }, { "epoch": 0.84, "learning_rate": 6.46315949253678e-07, "loss": 1.5452, "step": 7887 }, { "epoch": 0.84, "learning_rate": 6.45466722643976e-07, "loss": 1.4117, "step": 7888 }, { "epoch": 0.84, "learning_rate": 6.446180158231397e-07, "loss": 1.5794, "step": 7889 }, { "epoch": 0.84, "learning_rate": 6.437698288924777e-07, "loss": 1.3224, "step": 7890 }, { "epoch": 0.84, "learning_rate": 6.429221619532349e-07, "loss": 1.4539, "step": 7891 }, { "epoch": 0.84, "learning_rate": 6.420750151065963e-07, "loss": 1.2331, "step": 7892 }, { "epoch": 0.84, "learning_rate": 6.412283884536818e-07, "loss": 1.7072, "step": 7893 }, { "epoch": 0.84, "learning_rate": 6.40382282095553e-07, "loss": 1.4169, "step": 7894 }, { "epoch": 0.84, "learning_rate": 6.395366961332061e-07, "loss": 1.4199, "step": 7895 }, { "epoch": 0.84, "learning_rate": 6.38691630667575e-07, "loss": 1.295, "step": 7896 }, { "epoch": 0.84, "learning_rate": 6.378470857995362e-07, "loss": 1.2481, "step": 7897 }, { "epoch": 0.84, "learning_rate": 6.370030616298989e-07, "loss": 1.3834, "step": 7898 }, { "epoch": 0.84, "learning_rate": 6.36159558259411e-07, "loss": 0.8548, "step": 7899 }, { "epoch": 0.84, "learning_rate": 6.353165757887614e-07, "loss": 1.2732, "step": 7900 }, { "epoch": 0.84, "learning_rate": 6.344741143185723e-07, "loss": 1.2615, "step": 7901 }, { "epoch": 0.84, "learning_rate": 6.336321739494072e-07, "loss": 1.3786, "step": 7902 }, { "epoch": 0.84, "learning_rate": 6.327907547817674e-07, "loss": 1.483, "step": 7903 }, { "epoch": 0.84, "learning_rate": 6.319498569160898e-07, "loss": 1.4324, "step": 7904 }, { "epoch": 0.84, "learning_rate": 6.31109480452749e-07, "loss": 1.3643, "step": 7905 }, { "epoch": 0.84, "learning_rate": 6.302696254920598e-07, "loss": 1.5137, "step": 7906 }, { "epoch": 0.84, "learning_rate": 6.294302921342737e-07, "loss": 1.3324, "step": 7907 }, { "epoch": 0.84, "learning_rate": 6.285914804795784e-07, "loss": 1.3431, "step": 7908 }, { "epoch": 0.84, "learning_rate": 6.27753190628102e-07, "loss": 1.3646, "step": 7909 }, { "epoch": 0.84, "learning_rate": 6.269154226799085e-07, "loss": 1.2391, "step": 7910 }, { "epoch": 0.84, "learning_rate": 6.260781767349983e-07, "loss": 1.629, "step": 7911 }, { "epoch": 0.84, "learning_rate": 6.252414528933126e-07, "loss": 1.3592, "step": 7912 }, { "epoch": 0.84, "learning_rate": 6.244052512547299e-07, "loss": 1.5283, "step": 7913 }, { "epoch": 0.84, "learning_rate": 6.235695719190632e-07, "loss": 1.4185, "step": 7914 }, { "epoch": 0.84, "learning_rate": 6.227344149860665e-07, "loss": 1.371, "step": 7915 }, { "epoch": 0.84, "learning_rate": 6.218997805554305e-07, "loss": 1.2838, "step": 7916 }, { "epoch": 0.84, "learning_rate": 6.210656687267835e-07, "loss": 1.2516, "step": 7917 }, { "epoch": 0.84, "learning_rate": 6.202320795996885e-07, "loss": 1.63, "step": 7918 }, { "epoch": 0.84, "learning_rate": 6.193990132736527e-07, "loss": 1.2759, "step": 7919 }, { "epoch": 0.84, "learning_rate": 6.185664698481137e-07, "loss": 1.1008, "step": 7920 }, { "epoch": 0.84, "learning_rate": 6.177344494224513e-07, "loss": 1.5444, "step": 7921 }, { "epoch": 0.85, "learning_rate": 6.169029520959824e-07, "loss": 1.5275, "step": 7922 }, { "epoch": 0.85, "learning_rate": 6.160719779679597e-07, "loss": 1.3769, "step": 7923 }, { "epoch": 0.85, "learning_rate": 6.152415271375733e-07, "loss": 1.3944, "step": 7924 }, { "epoch": 0.85, "learning_rate": 6.144115997039529e-07, "loss": 1.2887, "step": 7925 }, { "epoch": 0.85, "learning_rate": 6.135821957661658e-07, "loss": 1.1819, "step": 7926 }, { "epoch": 0.85, "learning_rate": 6.127533154232135e-07, "loss": 1.8257, "step": 7927 }, { "epoch": 0.85, "learning_rate": 6.119249587740395e-07, "loss": 1.219, "step": 7928 }, { "epoch": 0.85, "learning_rate": 6.110971259175208e-07, "loss": 1.4563, "step": 7929 }, { "epoch": 0.85, "learning_rate": 6.102698169524723e-07, "loss": 1.3728, "step": 7930 }, { "epoch": 0.85, "learning_rate": 6.094430319776517e-07, "loss": 1.493, "step": 7931 }, { "epoch": 0.85, "learning_rate": 6.086167710917479e-07, "loss": 1.5031, "step": 7932 }, { "epoch": 0.85, "learning_rate": 6.077910343933879e-07, "loss": 1.2103, "step": 7933 }, { "epoch": 0.85, "learning_rate": 6.069658219811403e-07, "loss": 1.5704, "step": 7934 }, { "epoch": 0.85, "learning_rate": 6.061411339535062e-07, "loss": 1.5179, "step": 7935 }, { "epoch": 0.85, "learning_rate": 6.053169704089274e-07, "loss": 1.2491, "step": 7936 }, { "epoch": 0.85, "learning_rate": 6.044933314457829e-07, "loss": 1.4622, "step": 7937 }, { "epoch": 0.85, "learning_rate": 6.036702171623876e-07, "loss": 1.4308, "step": 7938 }, { "epoch": 0.85, "learning_rate": 6.028476276569933e-07, "loss": 1.5227, "step": 7939 }, { "epoch": 0.85, "learning_rate": 6.020255630277916e-07, "loss": 1.3983, "step": 7940 }, { "epoch": 0.85, "learning_rate": 6.012040233729105e-07, "loss": 1.255, "step": 7941 }, { "epoch": 0.85, "learning_rate": 6.003830087904133e-07, "loss": 1.2878, "step": 7942 }, { "epoch": 0.85, "learning_rate": 5.995625193783044e-07, "loss": 1.4798, "step": 7943 }, { "epoch": 0.85, "learning_rate": 5.987425552345222e-07, "loss": 1.4572, "step": 7944 }, { "epoch": 0.85, "learning_rate": 5.979231164569426e-07, "loss": 1.5219, "step": 7945 }, { "epoch": 0.85, "learning_rate": 5.97104203143381e-07, "loss": 1.61, "step": 7946 }, { "epoch": 0.85, "learning_rate": 5.962858153915896e-07, "loss": 0.982, "step": 7947 }, { "epoch": 0.85, "learning_rate": 5.954679532992564e-07, "loss": 1.2455, "step": 7948 }, { "epoch": 0.85, "learning_rate": 5.946506169640065e-07, "loss": 1.4962, "step": 7949 }, { "epoch": 0.85, "learning_rate": 5.938338064834037e-07, "loss": 1.4283, "step": 7950 }, { "epoch": 0.85, "learning_rate": 5.930175219549494e-07, "loss": 1.1918, "step": 7951 }, { "epoch": 0.85, "learning_rate": 5.922017634760796e-07, "loss": 1.3339, "step": 7952 }, { "epoch": 0.85, "learning_rate": 5.913865311441714e-07, "loss": 1.3192, "step": 7953 }, { "epoch": 0.85, "learning_rate": 5.905718250565351e-07, "loss": 1.3855, "step": 7954 }, { "epoch": 0.85, "learning_rate": 5.897576453104187e-07, "loss": 1.4915, "step": 7955 }, { "epoch": 0.85, "learning_rate": 5.889439920030127e-07, "loss": 1.3545, "step": 7956 }, { "epoch": 0.85, "learning_rate": 5.881308652314376e-07, "loss": 1.3569, "step": 7957 }, { "epoch": 0.85, "learning_rate": 5.873182650927545e-07, "loss": 1.5921, "step": 7958 }, { "epoch": 0.85, "learning_rate": 5.865061916839615e-07, "loss": 1.4309, "step": 7959 }, { "epoch": 0.85, "learning_rate": 5.856946451019952e-07, "loss": 1.4205, "step": 7960 }, { "epoch": 0.85, "learning_rate": 5.848836254437251e-07, "loss": 1.0129, "step": 7961 }, { "epoch": 0.85, "learning_rate": 5.840731328059629e-07, "loss": 1.4253, "step": 7962 }, { "epoch": 0.85, "learning_rate": 5.832631672854533e-07, "loss": 1.3637, "step": 7963 }, { "epoch": 0.85, "learning_rate": 5.824537289788789e-07, "loss": 1.2481, "step": 7964 }, { "epoch": 0.85, "learning_rate": 5.816448179828616e-07, "loss": 1.4468, "step": 7965 }, { "epoch": 0.85, "learning_rate": 5.808364343939598e-07, "loss": 1.3667, "step": 7966 }, { "epoch": 0.85, "learning_rate": 5.800285783086657e-07, "loss": 1.4598, "step": 7967 }, { "epoch": 0.85, "learning_rate": 5.792212498234134e-07, "loss": 1.408, "step": 7968 }, { "epoch": 0.85, "learning_rate": 5.784144490345689e-07, "loss": 1.3147, "step": 7969 }, { "epoch": 0.85, "learning_rate": 5.776081760384406e-07, "loss": 1.4103, "step": 7970 }, { "epoch": 0.85, "learning_rate": 5.768024309312681e-07, "loss": 1.4299, "step": 7971 }, { "epoch": 0.85, "learning_rate": 5.75997213809234e-07, "loss": 1.3616, "step": 7972 }, { "epoch": 0.85, "learning_rate": 5.751925247684525e-07, "loss": 1.4177, "step": 7973 }, { "epoch": 0.85, "learning_rate": 5.74388363904978e-07, "loss": 1.2301, "step": 7974 }, { "epoch": 0.85, "learning_rate": 5.735847313148024e-07, "loss": 1.3209, "step": 7975 }, { "epoch": 0.85, "learning_rate": 5.727816270938514e-07, "loss": 1.3421, "step": 7976 }, { "epoch": 0.85, "learning_rate": 5.719790513379891e-07, "loss": 1.3151, "step": 7977 }, { "epoch": 0.85, "learning_rate": 5.711770041430187e-07, "loss": 1.1705, "step": 7978 }, { "epoch": 0.85, "learning_rate": 5.703754856046767e-07, "loss": 1.438, "step": 7979 }, { "epoch": 0.85, "learning_rate": 5.695744958186383e-07, "loss": 1.6827, "step": 7980 }, { "epoch": 0.85, "learning_rate": 5.687740348805171e-07, "loss": 1.5926, "step": 7981 }, { "epoch": 0.85, "learning_rate": 5.67974102885861e-07, "loss": 1.5482, "step": 7982 }, { "epoch": 0.85, "learning_rate": 5.671746999301542e-07, "loss": 1.2767, "step": 7983 }, { "epoch": 0.85, "learning_rate": 5.663758261088209e-07, "loss": 1.3354, "step": 7984 }, { "epoch": 0.85, "learning_rate": 5.655774815172211e-07, "loss": 1.3727, "step": 7985 }, { "epoch": 0.85, "learning_rate": 5.647796662506493e-07, "loss": 1.4162, "step": 7986 }, { "epoch": 0.85, "learning_rate": 5.639823804043403e-07, "loss": 1.4825, "step": 7987 }, { "epoch": 0.85, "learning_rate": 5.631856240734629e-07, "loss": 1.4542, "step": 7988 }, { "epoch": 0.85, "learning_rate": 5.623893973531225e-07, "loss": 1.1578, "step": 7989 }, { "epoch": 0.85, "learning_rate": 5.615937003383654e-07, "loss": 1.3208, "step": 7990 }, { "epoch": 0.85, "learning_rate": 5.607985331241705e-07, "loss": 1.2852, "step": 7991 }, { "epoch": 0.85, "learning_rate": 5.600038958054538e-07, "loss": 1.639, "step": 7992 }, { "epoch": 0.85, "learning_rate": 5.592097884770709e-07, "loss": 1.4045, "step": 7993 }, { "epoch": 0.85, "learning_rate": 5.584162112338099e-07, "loss": 1.5351, "step": 7994 }, { "epoch": 0.85, "learning_rate": 5.576231641703994e-07, "loss": 1.521, "step": 7995 }, { "epoch": 0.85, "learning_rate": 5.568306473815044e-07, "loss": 1.4005, "step": 7996 }, { "epoch": 0.85, "learning_rate": 5.560386609617235e-07, "loss": 1.4069, "step": 7997 }, { "epoch": 0.85, "learning_rate": 5.552472050055946e-07, "loss": 1.4936, "step": 7998 }, { "epoch": 0.85, "learning_rate": 5.544562796075909e-07, "loss": 1.4041, "step": 7999 }, { "epoch": 0.85, "learning_rate": 5.536658848621257e-07, "loss": 1.1911, "step": 8000 }, { "epoch": 0.85, "learning_rate": 5.528760208635436e-07, "loss": 1.4647, "step": 8001 }, { "epoch": 0.85, "learning_rate": 5.520866877061293e-07, "loss": 1.4254, "step": 8002 }, { "epoch": 0.85, "learning_rate": 5.512978854841028e-07, "loss": 1.5758, "step": 8003 }, { "epoch": 0.85, "learning_rate": 5.505096142916233e-07, "loss": 1.4284, "step": 8004 }, { "epoch": 0.85, "learning_rate": 5.497218742227817e-07, "loss": 1.2603, "step": 8005 }, { "epoch": 0.85, "learning_rate": 5.489346653716116e-07, "loss": 1.3915, "step": 8006 }, { "epoch": 0.85, "learning_rate": 5.481479878320784e-07, "loss": 1.5858, "step": 8007 }, { "epoch": 0.85, "learning_rate": 5.473618416980831e-07, "loss": 1.4359, "step": 8008 }, { "epoch": 0.85, "learning_rate": 5.465762270634705e-07, "loss": 1.4927, "step": 8009 }, { "epoch": 0.85, "learning_rate": 5.457911440220154e-07, "loss": 1.2993, "step": 8010 }, { "epoch": 0.85, "learning_rate": 5.450065926674297e-07, "loss": 1.484, "step": 8011 }, { "epoch": 0.85, "learning_rate": 5.442225730933653e-07, "loss": 1.5033, "step": 8012 }, { "epoch": 0.85, "learning_rate": 5.434390853934063e-07, "loss": 1.2022, "step": 8013 }, { "epoch": 0.85, "learning_rate": 5.426561296610766e-07, "loss": 1.4533, "step": 8014 }, { "epoch": 0.85, "learning_rate": 5.418737059898365e-07, "loss": 1.6027, "step": 8015 }, { "epoch": 0.86, "learning_rate": 5.410918144730815e-07, "loss": 1.4679, "step": 8016 }, { "epoch": 0.86, "learning_rate": 5.403104552041416e-07, "loss": 1.4183, "step": 8017 }, { "epoch": 0.86, "learning_rate": 5.39529628276288e-07, "loss": 1.507, "step": 8018 }, { "epoch": 0.86, "learning_rate": 5.387493337827254e-07, "loss": 1.3043, "step": 8019 }, { "epoch": 0.86, "learning_rate": 5.379695718165945e-07, "loss": 1.4185, "step": 8020 }, { "epoch": 0.86, "learning_rate": 5.371903424709746e-07, "loss": 1.4568, "step": 8021 }, { "epoch": 0.86, "learning_rate": 5.364116458388802e-07, "loss": 1.1147, "step": 8022 }, { "epoch": 0.86, "learning_rate": 5.356334820132602e-07, "loss": 1.3022, "step": 8023 }, { "epoch": 0.86, "learning_rate": 5.348558510870033e-07, "loss": 1.5513, "step": 8024 }, { "epoch": 0.86, "learning_rate": 5.340787531529346e-07, "loss": 1.5319, "step": 8025 }, { "epoch": 0.86, "learning_rate": 5.333021883038114e-07, "loss": 1.4152, "step": 8026 }, { "epoch": 0.86, "learning_rate": 5.325261566323331e-07, "loss": 1.3274, "step": 8027 }, { "epoch": 0.86, "learning_rate": 5.3175065823113e-07, "loss": 1.6428, "step": 8028 }, { "epoch": 0.86, "learning_rate": 5.309756931927728e-07, "loss": 1.3269, "step": 8029 }, { "epoch": 0.86, "learning_rate": 5.302012616097652e-07, "loss": 1.116, "step": 8030 }, { "epoch": 0.86, "learning_rate": 5.294273635745517e-07, "loss": 1.4644, "step": 8031 }, { "epoch": 0.86, "learning_rate": 5.286539991795081e-07, "loss": 1.4998, "step": 8032 }, { "epoch": 0.86, "learning_rate": 5.278811685169493e-07, "loss": 1.2886, "step": 8033 }, { "epoch": 0.86, "learning_rate": 5.271088716791273e-07, "loss": 1.3464, "step": 8034 }, { "epoch": 0.86, "learning_rate": 5.263371087582281e-07, "loss": 1.5011, "step": 8035 }, { "epoch": 0.86, "learning_rate": 5.255658798463742e-07, "loss": 1.5196, "step": 8036 }, { "epoch": 0.86, "learning_rate": 5.24795185035627e-07, "loss": 1.3962, "step": 8037 }, { "epoch": 0.86, "learning_rate": 5.240250244179801e-07, "loss": 1.5445, "step": 8038 }, { "epoch": 0.86, "learning_rate": 5.232553980853667e-07, "loss": 1.157, "step": 8039 }, { "epoch": 0.86, "learning_rate": 5.224863061296553e-07, "loss": 1.2623, "step": 8040 }, { "epoch": 0.86, "learning_rate": 5.217177486426506e-07, "loss": 1.6261, "step": 8041 }, { "epoch": 0.86, "learning_rate": 5.209497257160911e-07, "loss": 1.4715, "step": 8042 }, { "epoch": 0.86, "learning_rate": 5.201822374416549e-07, "loss": 1.5649, "step": 8043 }, { "epoch": 0.86, "learning_rate": 5.194152839109562e-07, "loss": 1.3181, "step": 8044 }, { "epoch": 0.86, "learning_rate": 5.186488652155425e-07, "loss": 1.5127, "step": 8045 }, { "epoch": 0.86, "learning_rate": 5.178829814469006e-07, "loss": 1.1379, "step": 8046 }, { "epoch": 0.86, "learning_rate": 5.171176326964511e-07, "loss": 1.4253, "step": 8047 }, { "epoch": 0.86, "learning_rate": 5.163528190555495e-07, "loss": 1.2526, "step": 8048 }, { "epoch": 0.86, "learning_rate": 5.155885406154937e-07, "loss": 1.2601, "step": 8049 }, { "epoch": 0.86, "learning_rate": 5.148247974675113e-07, "loss": 1.3186, "step": 8050 }, { "epoch": 0.86, "learning_rate": 5.140615897027679e-07, "loss": 1.3871, "step": 8051 }, { "epoch": 0.86, "learning_rate": 5.132989174123659e-07, "loss": 1.4759, "step": 8052 }, { "epoch": 0.86, "learning_rate": 5.125367806873449e-07, "loss": 1.2772, "step": 8053 }, { "epoch": 0.86, "learning_rate": 5.117751796186776e-07, "loss": 1.3742, "step": 8054 }, { "epoch": 0.86, "learning_rate": 5.110141142972735e-07, "loss": 1.4482, "step": 8055 }, { "epoch": 0.86, "learning_rate": 5.102535848139812e-07, "loss": 1.337, "step": 8056 }, { "epoch": 0.86, "learning_rate": 5.094935912595805e-07, "loss": 1.412, "step": 8057 }, { "epoch": 0.86, "learning_rate": 5.087341337247914e-07, "loss": 1.277, "step": 8058 }, { "epoch": 0.86, "learning_rate": 5.079752123002684e-07, "loss": 1.3898, "step": 8059 }, { "epoch": 0.86, "learning_rate": 5.07216827076602e-07, "loss": 1.298, "step": 8060 }, { "epoch": 0.86, "learning_rate": 5.064589781443163e-07, "loss": 1.38, "step": 8061 }, { "epoch": 0.86, "learning_rate": 5.057016655938763e-07, "loss": 1.2271, "step": 8062 }, { "epoch": 0.86, "learning_rate": 5.0494488951568e-07, "loss": 1.5202, "step": 8063 }, { "epoch": 0.86, "learning_rate": 5.041886500000603e-07, "loss": 1.4177, "step": 8064 }, { "epoch": 0.86, "learning_rate": 5.03432947137289e-07, "loss": 1.4556, "step": 8065 }, { "epoch": 0.86, "learning_rate": 5.026777810175721e-07, "loss": 1.5011, "step": 8066 }, { "epoch": 0.86, "learning_rate": 5.019231517310491e-07, "loss": 1.3785, "step": 8067 }, { "epoch": 0.86, "learning_rate": 5.01169059367802e-07, "loss": 1.3271, "step": 8068 }, { "epoch": 0.86, "learning_rate": 5.004155040178432e-07, "loss": 1.3514, "step": 8069 }, { "epoch": 0.86, "learning_rate": 4.996624857711219e-07, "loss": 1.4025, "step": 8070 }, { "epoch": 0.86, "learning_rate": 4.989100047175243e-07, "loss": 1.3511, "step": 8071 }, { "epoch": 0.86, "learning_rate": 4.981580609468717e-07, "loss": 1.2432, "step": 8072 }, { "epoch": 0.86, "learning_rate": 4.97406654548922e-07, "loss": 1.4356, "step": 8073 }, { "epoch": 0.86, "learning_rate": 4.966557856133697e-07, "loss": 1.4118, "step": 8074 }, { "epoch": 0.86, "learning_rate": 4.959054542298425e-07, "loss": 1.4527, "step": 8075 }, { "epoch": 0.86, "learning_rate": 4.951556604879049e-07, "loss": 1.1905, "step": 8076 }, { "epoch": 0.86, "learning_rate": 4.94406404477058e-07, "loss": 1.6078, "step": 8077 }, { "epoch": 0.86, "learning_rate": 4.936576862867409e-07, "loss": 1.2032, "step": 8078 }, { "epoch": 0.86, "learning_rate": 4.929095060063227e-07, "loss": 1.529, "step": 8079 }, { "epoch": 0.86, "learning_rate": 4.921618637251141e-07, "loss": 1.2919, "step": 8080 }, { "epoch": 0.86, "learning_rate": 4.914147595323581e-07, "loss": 1.4285, "step": 8081 }, { "epoch": 0.86, "learning_rate": 4.906681935172342e-07, "loss": 1.308, "step": 8082 }, { "epoch": 0.86, "learning_rate": 4.89922165768858e-07, "loss": 1.4315, "step": 8083 }, { "epoch": 0.86, "learning_rate": 4.891766763762823e-07, "loss": 1.4685, "step": 8084 }, { "epoch": 0.86, "learning_rate": 4.88431725428492e-07, "loss": 1.3076, "step": 8085 }, { "epoch": 0.86, "learning_rate": 4.876873130144122e-07, "loss": 1.3885, "step": 8086 }, { "epoch": 0.86, "learning_rate": 4.869434392228989e-07, "loss": 1.7199, "step": 8087 }, { "epoch": 0.86, "learning_rate": 4.862001041427488e-07, "loss": 1.4242, "step": 8088 }, { "epoch": 0.86, "learning_rate": 4.85457307862689e-07, "loss": 1.2891, "step": 8089 }, { "epoch": 0.86, "learning_rate": 4.847150504713882e-07, "loss": 1.511, "step": 8090 }, { "epoch": 0.86, "learning_rate": 4.839733320574457e-07, "loss": 1.4656, "step": 8091 }, { "epoch": 0.86, "learning_rate": 4.832321527093969e-07, "loss": 1.6169, "step": 8092 }, { "epoch": 0.86, "learning_rate": 4.824915125157187e-07, "loss": 1.3073, "step": 8093 }, { "epoch": 0.86, "learning_rate": 4.817514115648164e-07, "loss": 1.6397, "step": 8094 }, { "epoch": 0.86, "learning_rate": 4.810118499450333e-07, "loss": 1.3215, "step": 8095 }, { "epoch": 0.86, "learning_rate": 4.8027282774465e-07, "loss": 1.2174, "step": 8096 }, { "epoch": 0.86, "learning_rate": 4.795343450518825e-07, "loss": 1.2995, "step": 8097 }, { "epoch": 0.86, "learning_rate": 4.787964019548797e-07, "loss": 1.401, "step": 8098 }, { "epoch": 0.86, "learning_rate": 4.780589985417294e-07, "loss": 1.6289, "step": 8099 }, { "epoch": 0.86, "learning_rate": 4.773221349004531e-07, "loss": 1.406, "step": 8100 }, { "epoch": 0.86, "learning_rate": 4.765858111190053e-07, "loss": 1.4117, "step": 8101 }, { "epoch": 0.86, "learning_rate": 4.7585002728528393e-07, "loss": 1.4776, "step": 8102 }, { "epoch": 0.86, "learning_rate": 4.7511478348711447e-07, "loss": 1.4578, "step": 8103 }, { "epoch": 0.86, "learning_rate": 4.743800798122611e-07, "loss": 1.4791, "step": 8104 }, { "epoch": 0.86, "learning_rate": 4.736459163484253e-07, "loss": 1.3751, "step": 8105 }, { "epoch": 0.86, "learning_rate": 4.729122931832392e-07, "loss": 1.4585, "step": 8106 }, { "epoch": 0.86, "learning_rate": 4.7217921040427604e-07, "loss": 1.6182, "step": 8107 }, { "epoch": 0.86, "learning_rate": 4.7144666809903984e-07, "loss": 1.4921, "step": 8108 }, { "epoch": 0.86, "learning_rate": 4.707146663549744e-07, "loss": 1.2702, "step": 8109 }, { "epoch": 0.87, "learning_rate": 4.6998320525945494e-07, "loss": 1.3275, "step": 8110 }, { "epoch": 0.87, "learning_rate": 4.6925228489979434e-07, "loss": 1.3788, "step": 8111 }, { "epoch": 0.87, "learning_rate": 4.685219053632423e-07, "loss": 1.5293, "step": 8112 }, { "epoch": 0.87, "learning_rate": 4.677920667369806e-07, "loss": 1.3929, "step": 8113 }, { "epoch": 0.87, "learning_rate": 4.6706276910812755e-07, "loss": 1.3782, "step": 8114 }, { "epoch": 0.87, "learning_rate": 4.663340125637389e-07, "loss": 1.558, "step": 8115 }, { "epoch": 0.87, "learning_rate": 4.6560579719080343e-07, "loss": 1.2601, "step": 8116 }, { "epoch": 0.87, "learning_rate": 4.6487812307624613e-07, "loss": 1.1733, "step": 8117 }, { "epoch": 0.87, "learning_rate": 4.6415099030692914e-07, "loss": 1.3407, "step": 8118 }, { "epoch": 0.87, "learning_rate": 4.6342439896964686e-07, "loss": 1.2637, "step": 8119 }, { "epoch": 0.87, "learning_rate": 4.6269834915112945e-07, "loss": 1.5735, "step": 8120 }, { "epoch": 0.87, "learning_rate": 4.619728409380453e-07, "loss": 1.352, "step": 8121 }, { "epoch": 0.87, "learning_rate": 4.612478744169968e-07, "loss": 1.4466, "step": 8122 }, { "epoch": 0.87, "learning_rate": 4.6052344967451977e-07, "loss": 1.4763, "step": 8123 }, { "epoch": 0.87, "learning_rate": 4.597995667970878e-07, "loss": 1.3942, "step": 8124 }, { "epoch": 0.87, "learning_rate": 4.5907622587110845e-07, "loss": 1.4277, "step": 8125 }, { "epoch": 0.87, "learning_rate": 4.583534269829232e-07, "loss": 1.3848, "step": 8126 }, { "epoch": 0.87, "learning_rate": 4.5763117021881467e-07, "loss": 1.1503, "step": 8127 }, { "epoch": 0.87, "learning_rate": 4.569094556649939e-07, "loss": 1.6358, "step": 8128 }, { "epoch": 0.87, "learning_rate": 4.561882834076098e-07, "loss": 1.4091, "step": 8129 }, { "epoch": 0.87, "learning_rate": 4.5546765353274846e-07, "loss": 1.4745, "step": 8130 }, { "epoch": 0.87, "learning_rate": 4.547475661264272e-07, "loss": 1.4874, "step": 8131 }, { "epoch": 0.87, "learning_rate": 4.5402802127460233e-07, "loss": 1.2133, "step": 8132 }, { "epoch": 0.87, "learning_rate": 4.5330901906316506e-07, "loss": 1.332, "step": 8133 }, { "epoch": 0.87, "learning_rate": 4.525905595779395e-07, "loss": 1.2356, "step": 8134 }, { "epoch": 0.87, "learning_rate": 4.518726429046849e-07, "loss": 1.6567, "step": 8135 }, { "epoch": 0.87, "learning_rate": 4.511552691290988e-07, "loss": 1.1127, "step": 8136 }, { "epoch": 0.87, "learning_rate": 4.5043843833681265e-07, "loss": 1.3428, "step": 8137 }, { "epoch": 0.87, "learning_rate": 4.4972215061339087e-07, "loss": 1.4827, "step": 8138 }, { "epoch": 0.87, "learning_rate": 4.490064060443361e-07, "loss": 1.2242, "step": 8139 }, { "epoch": 0.87, "learning_rate": 4.4829120471508393e-07, "loss": 1.5636, "step": 8140 }, { "epoch": 0.87, "learning_rate": 4.4757654671100714e-07, "loss": 1.5054, "step": 8141 }, { "epoch": 0.87, "learning_rate": 4.468624321174109e-07, "loss": 1.5296, "step": 8142 }, { "epoch": 0.87, "learning_rate": 4.4614886101953915e-07, "loss": 1.3773, "step": 8143 }, { "epoch": 0.87, "learning_rate": 4.454358335025677e-07, "loss": 1.2078, "step": 8144 }, { "epoch": 0.87, "learning_rate": 4.4472334965160736e-07, "loss": 1.4021, "step": 8145 }, { "epoch": 0.87, "learning_rate": 4.440114095517084e-07, "loss": 1.1994, "step": 8146 }, { "epoch": 0.87, "learning_rate": 4.4330001328785176e-07, "loss": 1.4338, "step": 8147 }, { "epoch": 0.87, "learning_rate": 4.4258916094495394e-07, "loss": 1.2552, "step": 8148 }, { "epoch": 0.87, "learning_rate": 4.4187885260786933e-07, "loss": 1.2829, "step": 8149 }, { "epoch": 0.87, "learning_rate": 4.411690883613834e-07, "loss": 1.5792, "step": 8150 }, { "epoch": 0.87, "learning_rate": 4.4045986829022e-07, "loss": 1.3213, "step": 8151 }, { "epoch": 0.87, "learning_rate": 4.397511924790382e-07, "loss": 1.3207, "step": 8152 }, { "epoch": 0.87, "learning_rate": 4.3904306101242855e-07, "loss": 1.3857, "step": 8153 }, { "epoch": 0.87, "learning_rate": 4.38335473974919e-07, "loss": 1.3538, "step": 8154 }, { "epoch": 0.87, "learning_rate": 4.376284314509727e-07, "loss": 1.464, "step": 8155 }, { "epoch": 0.87, "learning_rate": 4.369219335249886e-07, "loss": 1.3829, "step": 8156 }, { "epoch": 0.87, "learning_rate": 4.362159802812971e-07, "loss": 1.6283, "step": 8157 }, { "epoch": 0.87, "learning_rate": 4.3551057180416855e-07, "loss": 1.4347, "step": 8158 }, { "epoch": 0.87, "learning_rate": 4.34805708177804e-07, "loss": 1.3652, "step": 8159 }, { "epoch": 0.87, "learning_rate": 4.341013894863405e-07, "loss": 1.2888, "step": 8160 }, { "epoch": 0.87, "learning_rate": 4.333976158138514e-07, "loss": 1.4604, "step": 8161 }, { "epoch": 0.87, "learning_rate": 4.3269438724434563e-07, "loss": 1.587, "step": 8162 }, { "epoch": 0.87, "learning_rate": 4.3199170386176325e-07, "loss": 1.654, "step": 8163 }, { "epoch": 0.87, "learning_rate": 4.3128956574998436e-07, "loss": 1.4356, "step": 8164 }, { "epoch": 0.87, "learning_rate": 4.3058797299281864e-07, "loss": 1.4861, "step": 8165 }, { "epoch": 0.87, "learning_rate": 4.2988692567401515e-07, "loss": 1.2709, "step": 8166 }, { "epoch": 0.87, "learning_rate": 4.2918642387725475e-07, "loss": 1.5333, "step": 8167 }, { "epoch": 0.87, "learning_rate": 4.284864676861561e-07, "loss": 1.304, "step": 8168 }, { "epoch": 0.87, "learning_rate": 4.2778705718426907e-07, "loss": 1.6367, "step": 8169 }, { "epoch": 0.87, "learning_rate": 4.2708819245508126e-07, "loss": 1.3758, "step": 8170 }, { "epoch": 0.87, "learning_rate": 4.2638987358201546e-07, "loss": 1.5132, "step": 8171 }, { "epoch": 0.87, "learning_rate": 4.2569210064842716e-07, "loss": 1.3726, "step": 8172 }, { "epoch": 0.87, "learning_rate": 4.249948737376064e-07, "loss": 1.4513, "step": 8173 }, { "epoch": 0.87, "learning_rate": 4.2429819293278107e-07, "loss": 1.4774, "step": 8174 }, { "epoch": 0.87, "learning_rate": 4.236020583171108e-07, "loss": 1.3919, "step": 8175 }, { "epoch": 0.87, "learning_rate": 4.2290646997369176e-07, "loss": 1.5476, "step": 8176 }, { "epoch": 0.87, "learning_rate": 4.2221142798555545e-07, "loss": 1.3511, "step": 8177 }, { "epoch": 0.87, "learning_rate": 4.215169324356666e-07, "loss": 1.2825, "step": 8178 }, { "epoch": 0.87, "learning_rate": 4.208229834069233e-07, "loss": 1.5183, "step": 8179 }, { "epoch": 0.87, "learning_rate": 4.201295809821626e-07, "loss": 1.3429, "step": 8180 }, { "epoch": 0.87, "learning_rate": 4.194367252441545e-07, "loss": 1.5091, "step": 8181 }, { "epoch": 0.87, "learning_rate": 4.1874441627560105e-07, "loss": 1.304, "step": 8182 }, { "epoch": 0.87, "learning_rate": 4.180526541591434e-07, "loss": 1.4274, "step": 8183 }, { "epoch": 0.87, "learning_rate": 4.1736143897735394e-07, "loss": 1.2571, "step": 8184 }, { "epoch": 0.87, "learning_rate": 4.1667077081274153e-07, "loss": 1.6679, "step": 8185 }, { "epoch": 0.87, "learning_rate": 4.1598064974775033e-07, "loss": 1.4825, "step": 8186 }, { "epoch": 0.87, "learning_rate": 4.152910758647577e-07, "loss": 1.3431, "step": 8187 }, { "epoch": 0.87, "learning_rate": 4.146020492460751e-07, "loss": 1.5854, "step": 8188 }, { "epoch": 0.87, "learning_rate": 4.1391356997394994e-07, "loss": 1.3315, "step": 8189 }, { "epoch": 0.87, "learning_rate": 4.1322563813056606e-07, "loss": 1.2401, "step": 8190 }, { "epoch": 0.87, "learning_rate": 4.1253825379803826e-07, "loss": 1.2598, "step": 8191 }, { "epoch": 0.87, "learning_rate": 4.118514170584187e-07, "loss": 1.4131, "step": 8192 }, { "epoch": 0.87, "learning_rate": 4.111651279936929e-07, "loss": 1.4495, "step": 8193 }, { "epoch": 0.87, "learning_rate": 4.104793866857804e-07, "loss": 1.4135, "step": 8194 }, { "epoch": 0.87, "learning_rate": 4.097941932165367e-07, "loss": 1.4815, "step": 8195 }, { "epoch": 0.87, "learning_rate": 4.091095476677531e-07, "loss": 1.321, "step": 8196 }, { "epoch": 0.87, "learning_rate": 4.0842545012115263e-07, "loss": 1.2547, "step": 8197 }, { "epoch": 0.87, "learning_rate": 4.0774190065839324e-07, "loss": 1.2615, "step": 8198 }, { "epoch": 0.87, "learning_rate": 4.070588993610697e-07, "loss": 1.1638, "step": 8199 }, { "epoch": 0.87, "learning_rate": 4.0637644631071004e-07, "loss": 1.5674, "step": 8200 }, { "epoch": 0.87, "learning_rate": 4.0569454158877587e-07, "loss": 1.4155, "step": 8201 }, { "epoch": 0.87, "learning_rate": 4.050131852766659e-07, "loss": 1.2691, "step": 8202 }, { "epoch": 0.87, "learning_rate": 4.043323774557106e-07, "loss": 1.3019, "step": 8203 }, { "epoch": 0.88, "learning_rate": 4.036521182071745e-07, "loss": 1.387, "step": 8204 }, { "epoch": 0.88, "learning_rate": 4.029724076122621e-07, "loss": 1.2957, "step": 8205 }, { "epoch": 0.88, "learning_rate": 4.022932457521067e-07, "loss": 1.4401, "step": 8206 }, { "epoch": 0.88, "learning_rate": 4.016146327077775e-07, "loss": 1.3668, "step": 8207 }, { "epoch": 0.88, "learning_rate": 4.009365685602795e-07, "loss": 1.7666, "step": 8208 }, { "epoch": 0.88, "learning_rate": 4.002590533905504e-07, "loss": 1.2907, "step": 8209 }, { "epoch": 0.88, "learning_rate": 3.9958208727946426e-07, "loss": 1.4506, "step": 8210 }, { "epoch": 0.88, "learning_rate": 3.989056703078292e-07, "loss": 1.2922, "step": 8211 }, { "epoch": 0.88, "learning_rate": 3.982298025563863e-07, "loss": 1.3854, "step": 8212 }, { "epoch": 0.88, "learning_rate": 3.975544841058121e-07, "loss": 1.4529, "step": 8213 }, { "epoch": 0.88, "learning_rate": 3.968797150367171e-07, "loss": 1.4813, "step": 8214 }, { "epoch": 0.88, "learning_rate": 3.962054954296485e-07, "loss": 1.3329, "step": 8215 }, { "epoch": 0.88, "learning_rate": 3.9553182536508417e-07, "loss": 1.3252, "step": 8216 }, { "epoch": 0.88, "learning_rate": 3.948587049234398e-07, "loss": 1.2513, "step": 8217 }, { "epoch": 0.88, "learning_rate": 3.941861341850634e-07, "loss": 1.5069, "step": 8218 }, { "epoch": 0.88, "learning_rate": 3.9351411323023613e-07, "loss": 1.324, "step": 8219 }, { "epoch": 0.88, "learning_rate": 3.928426421391773e-07, "loss": 1.3471, "step": 8220 }, { "epoch": 0.88, "learning_rate": 3.9217172099203947e-07, "loss": 1.2271, "step": 8221 }, { "epoch": 0.88, "learning_rate": 3.915013498689058e-07, "loss": 1.5912, "step": 8222 }, { "epoch": 0.88, "learning_rate": 3.9083152884979935e-07, "loss": 1.3119, "step": 8223 }, { "epoch": 0.88, "learning_rate": 3.9016225801467355e-07, "loss": 1.476, "step": 8224 }, { "epoch": 0.88, "learning_rate": 3.8949353744341777e-07, "loss": 1.2873, "step": 8225 }, { "epoch": 0.88, "learning_rate": 3.8882536721585486e-07, "loss": 1.2082, "step": 8226 }, { "epoch": 0.88, "learning_rate": 3.881577474117432e-07, "loss": 1.5507, "step": 8227 }, { "epoch": 0.88, "learning_rate": 3.87490678110774e-07, "loss": 1.5812, "step": 8228 }, { "epoch": 0.88, "learning_rate": 3.868241593925742e-07, "loss": 1.1787, "step": 8229 }, { "epoch": 0.88, "learning_rate": 3.861581913367046e-07, "loss": 1.255, "step": 8230 }, { "epoch": 0.88, "learning_rate": 3.8549277402265993e-07, "loss": 1.4056, "step": 8231 }, { "epoch": 0.88, "learning_rate": 3.848279075298678e-07, "loss": 1.4488, "step": 8232 }, { "epoch": 0.88, "learning_rate": 3.841635919376924e-07, "loss": 1.146, "step": 8233 }, { "epoch": 0.88, "learning_rate": 3.8349982732543257e-07, "loss": 1.4112, "step": 8234 }, { "epoch": 0.88, "learning_rate": 3.828366137723183e-07, "loss": 1.5019, "step": 8235 }, { "epoch": 0.88, "learning_rate": 3.821739513575168e-07, "loss": 1.1555, "step": 8236 }, { "epoch": 0.88, "learning_rate": 3.815118401601281e-07, "loss": 1.4881, "step": 8237 }, { "epoch": 0.88, "learning_rate": 3.80850280259184e-07, "loss": 1.3359, "step": 8238 }, { "epoch": 0.88, "learning_rate": 3.8018927173365803e-07, "loss": 1.3771, "step": 8239 }, { "epoch": 0.88, "learning_rate": 3.7952881466244985e-07, "loss": 1.1669, "step": 8240 }, { "epoch": 0.88, "learning_rate": 3.7886890912439633e-07, "loss": 1.333, "step": 8241 }, { "epoch": 0.88, "learning_rate": 3.782095551982695e-07, "loss": 1.3053, "step": 8242 }, { "epoch": 0.88, "learning_rate": 3.775507529627742e-07, "loss": 1.3402, "step": 8243 }, { "epoch": 0.88, "learning_rate": 3.768925024965503e-07, "loss": 1.3468, "step": 8244 }, { "epoch": 0.88, "learning_rate": 3.7623480387817155e-07, "loss": 1.3685, "step": 8245 }, { "epoch": 0.88, "learning_rate": 3.7557765718614523e-07, "loss": 0.9324, "step": 8246 }, { "epoch": 0.88, "learning_rate": 3.749210624989125e-07, "loss": 1.3537, "step": 8247 }, { "epoch": 0.88, "learning_rate": 3.7426501989485e-07, "loss": 1.4114, "step": 8248 }, { "epoch": 0.88, "learning_rate": 3.736095294522685e-07, "loss": 1.3727, "step": 8249 }, { "epoch": 0.88, "learning_rate": 3.729545912494115e-07, "loss": 1.2231, "step": 8250 }, { "epoch": 0.88, "learning_rate": 3.7230020536445643e-07, "loss": 1.3011, "step": 8251 }, { "epoch": 0.88, "learning_rate": 3.71646371875517e-07, "loss": 1.4362, "step": 8252 }, { "epoch": 0.88, "learning_rate": 3.7099309086063794e-07, "loss": 1.1987, "step": 8253 }, { "epoch": 0.88, "learning_rate": 3.7034036239780024e-07, "loss": 1.5327, "step": 8254 }, { "epoch": 0.88, "learning_rate": 3.696881865649193e-07, "loss": 1.3576, "step": 8255 }, { "epoch": 0.88, "learning_rate": 3.6903656343984293e-07, "loss": 1.1546, "step": 8256 }, { "epoch": 0.88, "learning_rate": 3.6838549310035275e-07, "loss": 1.2283, "step": 8257 }, { "epoch": 0.88, "learning_rate": 3.677349756241666e-07, "loss": 1.4838, "step": 8258 }, { "epoch": 0.88, "learning_rate": 3.670850110889346e-07, "loss": 1.3225, "step": 8259 }, { "epoch": 0.88, "learning_rate": 3.664355995722402e-07, "loss": 1.4307, "step": 8260 }, { "epoch": 0.88, "learning_rate": 3.6578674115160416e-07, "loss": 1.2489, "step": 8261 }, { "epoch": 0.88, "learning_rate": 3.651384359044774e-07, "loss": 1.4365, "step": 8262 }, { "epoch": 0.88, "learning_rate": 3.644906839082446e-07, "loss": 1.3416, "step": 8263 }, { "epoch": 0.88, "learning_rate": 3.6384348524023003e-07, "loss": 1.0839, "step": 8264 }, { "epoch": 0.88, "learning_rate": 3.631968399776864e-07, "loss": 1.2129, "step": 8265 }, { "epoch": 0.88, "learning_rate": 3.6255074819780023e-07, "loss": 1.471, "step": 8266 }, { "epoch": 0.88, "learning_rate": 3.619052099776965e-07, "loss": 1.3443, "step": 8267 }, { "epoch": 0.88, "learning_rate": 3.6126022539442975e-07, "loss": 1.2942, "step": 8268 }, { "epoch": 0.88, "learning_rate": 3.6061579452498996e-07, "loss": 1.4022, "step": 8269 }, { "epoch": 0.88, "learning_rate": 3.5997191744630233e-07, "loss": 1.3958, "step": 8270 }, { "epoch": 0.88, "learning_rate": 3.593285942352237e-07, "loss": 1.3389, "step": 8271 }, { "epoch": 0.88, "learning_rate": 3.58685824968546e-07, "loss": 1.5464, "step": 8272 }, { "epoch": 0.88, "learning_rate": 3.5804360972299446e-07, "loss": 1.4511, "step": 8273 }, { "epoch": 0.88, "learning_rate": 3.5740194857523e-07, "loss": 1.3798, "step": 8274 }, { "epoch": 0.88, "learning_rate": 3.5676084160184423e-07, "loss": 1.2599, "step": 8275 }, { "epoch": 0.88, "learning_rate": 3.5612028887936576e-07, "loss": 1.7082, "step": 8276 }, { "epoch": 0.88, "learning_rate": 3.554802904842547e-07, "loss": 1.4848, "step": 8277 }, { "epoch": 0.88, "learning_rate": 3.548408464929065e-07, "loss": 1.5432, "step": 8278 }, { "epoch": 0.88, "learning_rate": 3.5420195698164907e-07, "loss": 1.4626, "step": 8279 }, { "epoch": 0.88, "learning_rate": 3.5356362202674687e-07, "loss": 1.2935, "step": 8280 }, { "epoch": 0.88, "learning_rate": 3.5292584170439347e-07, "loss": 1.5069, "step": 8281 }, { "epoch": 0.88, "learning_rate": 3.522886160907202e-07, "loss": 1.3617, "step": 8282 }, { "epoch": 0.88, "learning_rate": 3.516519452617922e-07, "loss": 1.3302, "step": 8283 }, { "epoch": 0.88, "learning_rate": 3.510158292936061e-07, "loss": 1.4716, "step": 8284 }, { "epoch": 0.88, "learning_rate": 3.503802682620922e-07, "loss": 1.425, "step": 8285 }, { "epoch": 0.88, "learning_rate": 3.4974526224311744e-07, "loss": 1.2818, "step": 8286 }, { "epoch": 0.88, "learning_rate": 3.491108113124797e-07, "loss": 1.5333, "step": 8287 }, { "epoch": 0.88, "learning_rate": 3.4847691554591225e-07, "loss": 1.1438, "step": 8288 }, { "epoch": 0.88, "learning_rate": 3.478435750190817e-07, "loss": 1.0549, "step": 8289 }, { "epoch": 0.88, "learning_rate": 3.4721078980758826e-07, "loss": 1.3835, "step": 8290 }, { "epoch": 0.88, "learning_rate": 3.465785599869648e-07, "loss": 1.4237, "step": 8291 }, { "epoch": 0.88, "learning_rate": 3.459468856326792e-07, "loss": 1.4768, "step": 8292 }, { "epoch": 0.88, "learning_rate": 3.453157668201335e-07, "loss": 1.3704, "step": 8293 }, { "epoch": 0.88, "learning_rate": 3.446852036246617e-07, "loss": 1.4825, "step": 8294 }, { "epoch": 0.88, "learning_rate": 3.4405519612153326e-07, "loss": 1.3444, "step": 8295 }, { "epoch": 0.88, "learning_rate": 3.434257443859507e-07, "loss": 1.3019, "step": 8296 }, { "epoch": 0.89, "learning_rate": 3.4279684849304716e-07, "loss": 1.657, "step": 8297 }, { "epoch": 0.89, "learning_rate": 3.4216850851789663e-07, "loss": 1.5659, "step": 8298 }, { "epoch": 0.89, "learning_rate": 3.415407245354996e-07, "loss": 1.3908, "step": 8299 }, { "epoch": 0.89, "learning_rate": 3.4091349662079275e-07, "loss": 1.3763, "step": 8300 }, { "epoch": 0.89, "learning_rate": 3.402868248486485e-07, "loss": 1.4848, "step": 8301 }, { "epoch": 0.89, "learning_rate": 3.396607092938692e-07, "loss": 1.3941, "step": 8302 }, { "epoch": 0.89, "learning_rate": 3.3903515003119334e-07, "loss": 1.2055, "step": 8303 }, { "epoch": 0.89, "learning_rate": 3.3841014713529184e-07, "loss": 1.4726, "step": 8304 }, { "epoch": 0.89, "learning_rate": 3.3778570068077e-07, "loss": 1.3194, "step": 8305 }, { "epoch": 0.89, "learning_rate": 3.371618107421659e-07, "loss": 1.4337, "step": 8306 }, { "epoch": 0.89, "learning_rate": 3.3653847739395174e-07, "loss": 1.1948, "step": 8307 }, { "epoch": 0.89, "learning_rate": 3.3591570071053345e-07, "loss": 1.5022, "step": 8308 }, { "epoch": 0.89, "learning_rate": 3.3529348076625057e-07, "loss": 1.2745, "step": 8309 }, { "epoch": 0.89, "learning_rate": 3.346718176353747e-07, "loss": 1.3175, "step": 8310 }, { "epoch": 0.89, "learning_rate": 3.3405071139211266e-07, "loss": 1.4424, "step": 8311 }, { "epoch": 0.89, "learning_rate": 3.3343016211060397e-07, "loss": 1.3045, "step": 8312 }, { "epoch": 0.89, "learning_rate": 3.3281016986492165e-07, "loss": 1.4579, "step": 8313 }, { "epoch": 0.89, "learning_rate": 3.3219073472907427e-07, "loss": 1.379, "step": 8314 }, { "epoch": 0.89, "learning_rate": 3.3157185677700043e-07, "loss": 1.4983, "step": 8315 }, { "epoch": 0.89, "learning_rate": 3.3095353608257385e-07, "loss": 1.7012, "step": 8316 }, { "epoch": 0.89, "learning_rate": 3.3033577271960213e-07, "loss": 1.3091, "step": 8317 }, { "epoch": 0.89, "learning_rate": 3.297185667618269e-07, "loss": 1.4413, "step": 8318 }, { "epoch": 0.89, "learning_rate": 3.2910191828292083e-07, "loss": 1.503, "step": 8319 }, { "epoch": 0.89, "learning_rate": 3.2848582735649335e-07, "loss": 1.1478, "step": 8320 }, { "epoch": 0.89, "learning_rate": 3.2787029405608405e-07, "loss": 1.243, "step": 8321 }, { "epoch": 0.89, "learning_rate": 3.2725531845516744e-07, "loss": 1.4354, "step": 8322 }, { "epoch": 0.89, "learning_rate": 3.2664090062715315e-07, "loss": 1.2895, "step": 8323 }, { "epoch": 0.89, "learning_rate": 3.2602704064538137e-07, "loss": 1.4024, "step": 8324 }, { "epoch": 0.89, "learning_rate": 3.254137385831263e-07, "loss": 1.3857, "step": 8325 }, { "epoch": 0.89, "learning_rate": 3.2480099451359716e-07, "loss": 1.3529, "step": 8326 }, { "epoch": 0.89, "learning_rate": 3.241888085099354e-07, "loss": 1.3079, "step": 8327 }, { "epoch": 0.89, "learning_rate": 3.2357718064521594e-07, "loss": 1.3234, "step": 8328 }, { "epoch": 0.89, "learning_rate": 3.22966110992447e-07, "loss": 1.3743, "step": 8329 }, { "epoch": 0.89, "learning_rate": 3.2235559962457087e-07, "loss": 1.4659, "step": 8330 }, { "epoch": 0.89, "learning_rate": 3.217456466144614e-07, "loss": 1.3403, "step": 8331 }, { "epoch": 0.89, "learning_rate": 3.2113625203492813e-07, "loss": 1.5426, "step": 8332 }, { "epoch": 0.89, "learning_rate": 3.2052741595871294e-07, "loss": 1.5974, "step": 8333 }, { "epoch": 0.89, "learning_rate": 3.199191384584893e-07, "loss": 1.3667, "step": 8334 }, { "epoch": 0.89, "learning_rate": 3.1931141960686864e-07, "loss": 1.3543, "step": 8335 }, { "epoch": 0.89, "learning_rate": 3.187042594763895e-07, "loss": 1.4205, "step": 8336 }, { "epoch": 0.89, "learning_rate": 3.180976581395295e-07, "loss": 1.3256, "step": 8337 }, { "epoch": 0.89, "learning_rate": 3.1749161566869466e-07, "loss": 1.393, "step": 8338 }, { "epoch": 0.89, "learning_rate": 3.1688613213622876e-07, "loss": 1.3683, "step": 8339 }, { "epoch": 0.89, "learning_rate": 3.1628120761440616e-07, "loss": 1.4443, "step": 8340 }, { "epoch": 0.89, "learning_rate": 3.1567684217543303e-07, "loss": 1.3265, "step": 8341 }, { "epoch": 0.89, "learning_rate": 3.1507303589145445e-07, "loss": 1.2691, "step": 8342 }, { "epoch": 0.89, "learning_rate": 3.144697888345427e-07, "loss": 1.4971, "step": 8343 }, { "epoch": 0.89, "learning_rate": 3.1386710107670583e-07, "loss": 1.4673, "step": 8344 }, { "epoch": 0.89, "learning_rate": 3.1326497268988623e-07, "loss": 1.5543, "step": 8345 }, { "epoch": 0.89, "learning_rate": 3.1266340374595693e-07, "loss": 1.3165, "step": 8346 }, { "epoch": 0.89, "learning_rate": 3.120623943167267e-07, "loss": 1.2311, "step": 8347 }, { "epoch": 0.89, "learning_rate": 3.1146194447393587e-07, "loss": 1.4949, "step": 8348 }, { "epoch": 0.89, "learning_rate": 3.108620542892593e-07, "loss": 1.3546, "step": 8349 }, { "epoch": 0.89, "learning_rate": 3.102627238343026e-07, "loss": 1.3823, "step": 8350 }, { "epoch": 0.89, "learning_rate": 3.0966395318060725e-07, "loss": 1.4448, "step": 8351 }, { "epoch": 0.89, "learning_rate": 3.0906574239964795e-07, "loss": 1.4776, "step": 8352 }, { "epoch": 0.89, "learning_rate": 3.0846809156282906e-07, "loss": 1.487, "step": 8353 }, { "epoch": 0.89, "learning_rate": 3.078710007414931e-07, "loss": 1.5748, "step": 8354 }, { "epoch": 0.89, "learning_rate": 3.072744700069119e-07, "loss": 1.2498, "step": 8355 }, { "epoch": 0.89, "learning_rate": 3.066784994302907e-07, "loss": 1.3669, "step": 8356 }, { "epoch": 0.89, "learning_rate": 3.060830890827704e-07, "loss": 1.1005, "step": 8357 }, { "epoch": 0.89, "learning_rate": 3.054882390354241e-07, "loss": 1.5552, "step": 8358 }, { "epoch": 0.89, "learning_rate": 3.048939493592551e-07, "loss": 1.319, "step": 8359 }, { "epoch": 0.89, "learning_rate": 3.0430022012520486e-07, "loss": 1.4145, "step": 8360 }, { "epoch": 0.89, "learning_rate": 3.0370705140414293e-07, "loss": 1.3055, "step": 8361 }, { "epoch": 0.89, "learning_rate": 3.031144432668759e-07, "loss": 1.6281, "step": 8362 }, { "epoch": 0.89, "learning_rate": 3.0252239578414e-07, "loss": 1.4592, "step": 8363 }, { "epoch": 0.89, "learning_rate": 3.019309090266087e-07, "loss": 1.4717, "step": 8364 }, { "epoch": 0.89, "learning_rate": 3.0133998306488443e-07, "loss": 1.2179, "step": 8365 }, { "epoch": 0.89, "learning_rate": 3.007496179695046e-07, "loss": 1.537, "step": 8366 }, { "epoch": 0.89, "learning_rate": 3.0015981381094073e-07, "loss": 1.2673, "step": 8367 }, { "epoch": 0.89, "learning_rate": 2.995705706595947e-07, "loss": 1.4728, "step": 8368 }, { "epoch": 0.89, "learning_rate": 2.9898188858580313e-07, "loss": 1.4377, "step": 8369 }, { "epoch": 0.89, "learning_rate": 2.9839376765983583e-07, "loss": 1.379, "step": 8370 }, { "epoch": 0.89, "learning_rate": 2.978062079518956e-07, "loss": 1.3341, "step": 8371 }, { "epoch": 0.89, "learning_rate": 2.972192095321169e-07, "loss": 1.5009, "step": 8372 }, { "epoch": 0.89, "learning_rate": 2.9663277247056923e-07, "loss": 1.5333, "step": 8373 }, { "epoch": 0.89, "learning_rate": 2.960468968372532e-07, "loss": 1.2538, "step": 8374 }, { "epoch": 0.89, "learning_rate": 2.954615827021018e-07, "loss": 1.5472, "step": 8375 }, { "epoch": 0.89, "learning_rate": 2.9487683013498523e-07, "loss": 1.3972, "step": 8376 }, { "epoch": 0.89, "learning_rate": 2.942926392057027e-07, "loss": 1.2651, "step": 8377 }, { "epoch": 0.89, "learning_rate": 2.9370900998398675e-07, "loss": 1.2729, "step": 8378 }, { "epoch": 0.89, "learning_rate": 2.93125942539505e-07, "loss": 1.2997, "step": 8379 }, { "epoch": 0.89, "learning_rate": 2.925434369418545e-07, "loss": 1.2889, "step": 8380 }, { "epoch": 0.89, "learning_rate": 2.9196149326056967e-07, "loss": 1.4523, "step": 8381 }, { "epoch": 0.89, "learning_rate": 2.913801115651144e-07, "loss": 1.3633, "step": 8382 }, { "epoch": 0.89, "learning_rate": 2.9079929192488756e-07, "loss": 1.3772, "step": 8383 }, { "epoch": 0.89, "learning_rate": 2.902190344092182e-07, "loss": 1.1015, "step": 8384 }, { "epoch": 0.89, "learning_rate": 2.896393390873714e-07, "loss": 1.3323, "step": 8385 }, { "epoch": 0.89, "learning_rate": 2.8906020602854414e-07, "loss": 1.3422, "step": 8386 }, { "epoch": 0.89, "learning_rate": 2.884816353018649e-07, "loss": 1.4511, "step": 8387 }, { "epoch": 0.89, "learning_rate": 2.8790362697639685e-07, "loss": 1.3529, "step": 8388 }, { "epoch": 0.89, "learning_rate": 2.8732618112113586e-07, "loss": 1.327, "step": 8389 }, { "epoch": 0.89, "learning_rate": 2.867492978050079e-07, "loss": 1.1557, "step": 8390 }, { "epoch": 0.9, "learning_rate": 2.8617297709687577e-07, "loss": 1.3194, "step": 8391 }, { "epoch": 0.9, "learning_rate": 2.855972190655337e-07, "loss": 1.573, "step": 8392 }, { "epoch": 0.9, "learning_rate": 2.8502202377970746e-07, "loss": 1.6425, "step": 8393 }, { "epoch": 0.9, "learning_rate": 2.8444739130805587e-07, "loss": 1.5465, "step": 8394 }, { "epoch": 0.9, "learning_rate": 2.8387332171917247e-07, "loss": 1.6275, "step": 8395 }, { "epoch": 0.9, "learning_rate": 2.832998150815819e-07, "loss": 1.3541, "step": 8396 }, { "epoch": 0.9, "learning_rate": 2.827268714637421e-07, "loss": 1.2331, "step": 8397 }, { "epoch": 0.9, "learning_rate": 2.8215449093404457e-07, "loss": 1.3457, "step": 8398 }, { "epoch": 0.9, "learning_rate": 2.815826735608124e-07, "loss": 1.3403, "step": 8399 }, { "epoch": 0.9, "learning_rate": 2.810114194122998e-07, "loss": 1.4244, "step": 8400 }, { "epoch": 0.9, "learning_rate": 2.80440728556699e-07, "loss": 1.3171, "step": 8401 }, { "epoch": 0.9, "learning_rate": 2.798706010621305e-07, "loss": 1.233, "step": 8402 }, { "epoch": 0.9, "learning_rate": 2.793010369966487e-07, "loss": 1.4628, "step": 8403 }, { "epoch": 0.9, "learning_rate": 2.787320364282414e-07, "loss": 1.6461, "step": 8404 }, { "epoch": 0.9, "learning_rate": 2.781635994248272e-07, "loss": 1.5809, "step": 8405 }, { "epoch": 0.9, "learning_rate": 2.7759572605426057e-07, "loss": 1.6293, "step": 8406 }, { "epoch": 0.9, "learning_rate": 2.7702841638432734e-07, "loss": 1.3989, "step": 8407 }, { "epoch": 0.9, "learning_rate": 2.764616704827444e-07, "loss": 1.4712, "step": 8408 }, { "epoch": 0.9, "learning_rate": 2.7589548841716274e-07, "loss": 1.2701, "step": 8409 }, { "epoch": 0.9, "learning_rate": 2.7532987025516653e-07, "loss": 1.5024, "step": 8410 }, { "epoch": 0.9, "learning_rate": 2.747648160642724e-07, "loss": 1.3992, "step": 8411 }, { "epoch": 0.9, "learning_rate": 2.7420032591192856e-07, "loss": 1.5677, "step": 8412 }, { "epoch": 0.9, "learning_rate": 2.736363998655173e-07, "loss": 1.3676, "step": 8413 }, { "epoch": 0.9, "learning_rate": 2.730730379923524e-07, "loss": 1.1393, "step": 8414 }, { "epoch": 0.9, "learning_rate": 2.7251024035968134e-07, "loss": 1.1668, "step": 8415 }, { "epoch": 0.9, "learning_rate": 2.7194800703468305e-07, "loss": 1.3706, "step": 8416 }, { "epoch": 0.9, "learning_rate": 2.7138633808447057e-07, "loss": 1.4514, "step": 8417 }, { "epoch": 0.9, "learning_rate": 2.7082523357608856e-07, "loss": 1.2888, "step": 8418 }, { "epoch": 0.9, "learning_rate": 2.702646935765141e-07, "loss": 1.3883, "step": 8419 }, { "epoch": 0.9, "learning_rate": 2.69704718152658e-07, "loss": 1.2521, "step": 8420 }, { "epoch": 0.9, "learning_rate": 2.6914530737136346e-07, "loss": 1.3661, "step": 8421 }, { "epoch": 0.9, "learning_rate": 2.685864612994038e-07, "loss": 1.4078, "step": 8422 }, { "epoch": 0.9, "learning_rate": 2.6802818000348894e-07, "loss": 1.31, "step": 8423 }, { "epoch": 0.9, "learning_rate": 2.674704635502584e-07, "loss": 1.5144, "step": 8424 }, { "epoch": 0.9, "learning_rate": 2.6691331200628556e-07, "loss": 1.3211, "step": 8425 }, { "epoch": 0.9, "learning_rate": 2.663567254380772e-07, "loss": 1.3373, "step": 8426 }, { "epoch": 0.9, "learning_rate": 2.658007039120697e-07, "loss": 1.6648, "step": 8427 }, { "epoch": 0.9, "learning_rate": 2.652452474946349e-07, "loss": 1.5309, "step": 8428 }, { "epoch": 0.9, "learning_rate": 2.646903562520753e-07, "loss": 1.3676, "step": 8429 }, { "epoch": 0.9, "learning_rate": 2.64136030250628e-07, "loss": 1.3148, "step": 8430 }, { "epoch": 0.9, "learning_rate": 2.6358226955645994e-07, "loss": 1.4506, "step": 8431 }, { "epoch": 0.9, "learning_rate": 2.630290742356739e-07, "loss": 1.3669, "step": 8432 }, { "epoch": 0.9, "learning_rate": 2.6247644435430263e-07, "loss": 1.5935, "step": 8433 }, { "epoch": 0.9, "learning_rate": 2.6192437997830946e-07, "loss": 1.2266, "step": 8434 }, { "epoch": 0.9, "learning_rate": 2.6137288117359716e-07, "loss": 1.5701, "step": 8435 }, { "epoch": 0.9, "learning_rate": 2.6082194800599424e-07, "loss": 1.2702, "step": 8436 }, { "epoch": 0.9, "learning_rate": 2.602715805412637e-07, "loss": 1.6319, "step": 8437 }, { "epoch": 0.9, "learning_rate": 2.5972177884510287e-07, "loss": 1.412, "step": 8438 }, { "epoch": 0.9, "learning_rate": 2.591725429831382e-07, "loss": 1.4145, "step": 8439 }, { "epoch": 0.9, "learning_rate": 2.586238730209323e-07, "loss": 1.2911, "step": 8440 }, { "epoch": 0.9, "learning_rate": 2.5807576902397823e-07, "loss": 1.2323, "step": 8441 }, { "epoch": 0.9, "learning_rate": 2.57528231057701e-07, "loss": 1.5696, "step": 8442 }, { "epoch": 0.9, "learning_rate": 2.5698125918745874e-07, "loss": 1.3492, "step": 8443 }, { "epoch": 0.9, "learning_rate": 2.564348534785416e-07, "loss": 1.4087, "step": 8444 }, { "epoch": 0.9, "learning_rate": 2.558890139961745e-07, "loss": 1.437, "step": 8445 }, { "epoch": 0.9, "learning_rate": 2.5534374080551094e-07, "loss": 1.1695, "step": 8446 }, { "epoch": 0.9, "learning_rate": 2.5479903397163887e-07, "loss": 1.356, "step": 8447 }, { "epoch": 0.9, "learning_rate": 2.5425489355957956e-07, "loss": 1.461, "step": 8448 }, { "epoch": 0.9, "learning_rate": 2.537113196342844e-07, "loss": 1.2261, "step": 8449 }, { "epoch": 0.9, "learning_rate": 2.531683122606382e-07, "loss": 1.2027, "step": 8450 }, { "epoch": 0.9, "learning_rate": 2.526258715034602e-07, "loss": 1.4762, "step": 8451 }, { "epoch": 0.9, "learning_rate": 2.5208399742749855e-07, "loss": 1.395, "step": 8452 }, { "epoch": 0.9, "learning_rate": 2.5154269009743383e-07, "loss": 1.3583, "step": 8453 }, { "epoch": 0.9, "learning_rate": 2.510019495778837e-07, "loss": 1.15, "step": 8454 }, { "epoch": 0.9, "learning_rate": 2.5046177593339385e-07, "loss": 1.3777, "step": 8455 }, { "epoch": 0.9, "learning_rate": 2.499221692284415e-07, "loss": 1.2869, "step": 8456 }, { "epoch": 0.9, "learning_rate": 2.4938312952744016e-07, "loss": 1.3898, "step": 8457 }, { "epoch": 0.9, "learning_rate": 2.4884465689473223e-07, "loss": 1.2762, "step": 8458 }, { "epoch": 0.9, "learning_rate": 2.4830675139459416e-07, "loss": 1.3026, "step": 8459 }, { "epoch": 0.9, "learning_rate": 2.477694130912356e-07, "loss": 1.3484, "step": 8460 }, { "epoch": 0.9, "learning_rate": 2.4723264204879526e-07, "loss": 1.3516, "step": 8461 }, { "epoch": 0.9, "learning_rate": 2.4669643833134637e-07, "loss": 1.3964, "step": 8462 }, { "epoch": 0.9, "learning_rate": 2.461608020028944e-07, "loss": 1.213, "step": 8463 }, { "epoch": 0.9, "learning_rate": 2.456257331273776e-07, "loss": 1.4197, "step": 8464 }, { "epoch": 0.9, "learning_rate": 2.4509123176866376e-07, "loss": 1.1985, "step": 8465 }, { "epoch": 0.9, "learning_rate": 2.445572979905575e-07, "loss": 1.3929, "step": 8466 }, { "epoch": 0.9, "learning_rate": 2.440239318567911e-07, "loss": 1.2127, "step": 8467 }, { "epoch": 0.9, "learning_rate": 2.4349113343103135e-07, "loss": 1.399, "step": 8468 }, { "epoch": 0.9, "learning_rate": 2.4295890277687695e-07, "loss": 1.5276, "step": 8469 }, { "epoch": 0.9, "learning_rate": 2.4242723995785977e-07, "loss": 1.8195, "step": 8470 }, { "epoch": 0.9, "learning_rate": 2.4189614503744183e-07, "loss": 1.6843, "step": 8471 }, { "epoch": 0.9, "learning_rate": 2.4136561807901916e-07, "loss": 1.3982, "step": 8472 }, { "epoch": 0.9, "learning_rate": 2.408356591459188e-07, "loss": 1.4514, "step": 8473 }, { "epoch": 0.9, "learning_rate": 2.4030626830140124e-07, "loss": 1.3884, "step": 8474 }, { "epoch": 0.9, "learning_rate": 2.397774456086577e-07, "loss": 1.1403, "step": 8475 }, { "epoch": 0.9, "learning_rate": 2.392491911308126e-07, "loss": 1.418, "step": 8476 }, { "epoch": 0.9, "learning_rate": 2.3872150493092215e-07, "loss": 1.4511, "step": 8477 }, { "epoch": 0.9, "learning_rate": 2.3819438707197495e-07, "loss": 1.7195, "step": 8478 }, { "epoch": 0.9, "learning_rate": 2.376678376168917e-07, "loss": 1.2385, "step": 8479 }, { "epoch": 0.9, "learning_rate": 2.3714185662852553e-07, "loss": 1.4197, "step": 8480 }, { "epoch": 0.9, "learning_rate": 2.3661644416966057e-07, "loss": 1.5131, "step": 8481 }, { "epoch": 0.9, "learning_rate": 2.3609160030301447e-07, "loss": 1.4742, "step": 8482 }, { "epoch": 0.9, "learning_rate": 2.3556732509123603e-07, "loss": 1.3074, "step": 8483 }, { "epoch": 0.9, "learning_rate": 2.3504361859690628e-07, "loss": 1.4823, "step": 8484 }, { "epoch": 0.91, "learning_rate": 2.3452048088253965e-07, "loss": 1.3855, "step": 8485 }, { "epoch": 0.91, "learning_rate": 2.3399791201058174e-07, "loss": 1.2477, "step": 8486 }, { "epoch": 0.91, "learning_rate": 2.3347591204340881e-07, "loss": 1.4775, "step": 8487 }, { "epoch": 0.91, "learning_rate": 2.3295448104333096e-07, "loss": 1.6275, "step": 8488 }, { "epoch": 0.91, "learning_rate": 2.3243361907259122e-07, "loss": 1.4546, "step": 8489 }, { "epoch": 0.91, "learning_rate": 2.3191332619336204e-07, "loss": 1.3969, "step": 8490 }, { "epoch": 0.91, "learning_rate": 2.31393602467751e-07, "loss": 1.2263, "step": 8491 }, { "epoch": 0.91, "learning_rate": 2.3087444795779456e-07, "loss": 1.4391, "step": 8492 }, { "epoch": 0.91, "learning_rate": 2.3035586272546207e-07, "loss": 1.6032, "step": 8493 }, { "epoch": 0.91, "learning_rate": 2.2983784683265842e-07, "loss": 1.2179, "step": 8494 }, { "epoch": 0.91, "learning_rate": 2.2932040034121638e-07, "loss": 1.5568, "step": 8495 }, { "epoch": 0.91, "learning_rate": 2.2880352331290102e-07, "loss": 1.5496, "step": 8496 }, { "epoch": 0.91, "learning_rate": 2.2828721580941237e-07, "loss": 1.5854, "step": 8497 }, { "epoch": 0.91, "learning_rate": 2.277714778923795e-07, "loss": 1.4411, "step": 8498 }, { "epoch": 0.91, "learning_rate": 2.2725630962336542e-07, "loss": 1.4534, "step": 8499 }, { "epoch": 0.91, "learning_rate": 2.2674171106386312e-07, "loss": 1.2257, "step": 8500 }, { "epoch": 0.91, "learning_rate": 2.2622768227530068e-07, "loss": 1.3483, "step": 8501 }, { "epoch": 0.91, "learning_rate": 2.2571422331903458e-07, "loss": 1.5569, "step": 8502 }, { "epoch": 0.91, "learning_rate": 2.2520133425635526e-07, "loss": 1.2666, "step": 8503 }, { "epoch": 0.91, "learning_rate": 2.246890151484865e-07, "loss": 1.1984, "step": 8504 }, { "epoch": 0.91, "learning_rate": 2.2417726605658164e-07, "loss": 1.1384, "step": 8505 }, { "epoch": 0.91, "learning_rate": 2.2366608704172565e-07, "loss": 1.4849, "step": 8506 }, { "epoch": 0.91, "learning_rate": 2.2315547816493698e-07, "loss": 1.3959, "step": 8507 }, { "epoch": 0.91, "learning_rate": 2.226454394871669e-07, "loss": 1.706, "step": 8508 }, { "epoch": 0.91, "learning_rate": 2.2213597106929608e-07, "loss": 1.2191, "step": 8509 }, { "epoch": 0.91, "learning_rate": 2.2162707297213982e-07, "loss": 1.4688, "step": 8510 }, { "epoch": 0.91, "learning_rate": 2.2111874525644228e-07, "loss": 0.9964, "step": 8511 }, { "epoch": 0.91, "learning_rate": 2.2061098798288107e-07, "loss": 1.1907, "step": 8512 }, { "epoch": 0.91, "learning_rate": 2.2010380121206766e-07, "loss": 1.3935, "step": 8513 }, { "epoch": 0.91, "learning_rate": 2.1959718500454196e-07, "loss": 1.5117, "step": 8514 }, { "epoch": 0.91, "learning_rate": 2.1909113942077787e-07, "loss": 1.3685, "step": 8515 }, { "epoch": 0.91, "learning_rate": 2.1858566452118146e-07, "loss": 1.4522, "step": 8516 }, { "epoch": 0.91, "learning_rate": 2.1808076036608783e-07, "loss": 1.2808, "step": 8517 }, { "epoch": 0.91, "learning_rate": 2.1757642701576765e-07, "loss": 1.4541, "step": 8518 }, { "epoch": 0.91, "learning_rate": 2.170726645304222e-07, "loss": 1.3203, "step": 8519 }, { "epoch": 0.91, "learning_rate": 2.165694729701834e-07, "loss": 1.5137, "step": 8520 }, { "epoch": 0.91, "learning_rate": 2.1606685239511537e-07, "loss": 1.3844, "step": 8521 }, { "epoch": 0.91, "learning_rate": 2.155648028652152e-07, "loss": 1.4056, "step": 8522 }, { "epoch": 0.91, "learning_rate": 2.1506332444041212e-07, "loss": 1.4217, "step": 8523 }, { "epoch": 0.91, "learning_rate": 2.1456241718056436e-07, "loss": 1.3745, "step": 8524 }, { "epoch": 0.91, "learning_rate": 2.140620811454658e-07, "loss": 1.1511, "step": 8525 }, { "epoch": 0.91, "learning_rate": 2.1356231639483917e-07, "loss": 1.5187, "step": 8526 }, { "epoch": 0.91, "learning_rate": 2.1306312298833954e-07, "loss": 1.5586, "step": 8527 }, { "epoch": 0.91, "learning_rate": 2.1256450098555426e-07, "loss": 1.3309, "step": 8528 }, { "epoch": 0.91, "learning_rate": 2.1206645044600404e-07, "loss": 1.3382, "step": 8529 }, { "epoch": 0.91, "learning_rate": 2.1156897142913858e-07, "loss": 1.293, "step": 8530 }, { "epoch": 0.91, "learning_rate": 2.1107206399434144e-07, "loss": 1.2445, "step": 8531 }, { "epoch": 0.91, "learning_rate": 2.1057572820092576e-07, "loss": 1.3062, "step": 8532 }, { "epoch": 0.91, "learning_rate": 2.1007996410813912e-07, "loss": 1.4576, "step": 8533 }, { "epoch": 0.91, "learning_rate": 2.0958477177515812e-07, "loss": 1.3377, "step": 8534 }, { "epoch": 0.91, "learning_rate": 2.0909015126109488e-07, "loss": 1.3859, "step": 8535 }, { "epoch": 0.91, "learning_rate": 2.0859610262498887e-07, "loss": 1.347, "step": 8536 }, { "epoch": 0.91, "learning_rate": 2.0810262592581287e-07, "loss": 1.2554, "step": 8537 }, { "epoch": 0.91, "learning_rate": 2.0760972122247425e-07, "loss": 1.1606, "step": 8538 }, { "epoch": 0.91, "learning_rate": 2.0711738857380814e-07, "loss": 1.6228, "step": 8539 }, { "epoch": 0.91, "learning_rate": 2.0662562803858255e-07, "loss": 1.488, "step": 8540 }, { "epoch": 0.91, "learning_rate": 2.061344396754994e-07, "loss": 1.6548, "step": 8541 }, { "epoch": 0.91, "learning_rate": 2.0564382354318791e-07, "loss": 1.4099, "step": 8542 }, { "epoch": 0.91, "learning_rate": 2.05153779700214e-07, "loss": 1.2091, "step": 8543 }, { "epoch": 0.91, "learning_rate": 2.04664308205072e-07, "loss": 1.3626, "step": 8544 }, { "epoch": 0.91, "learning_rate": 2.0417540911618906e-07, "loss": 1.2526, "step": 8545 }, { "epoch": 0.91, "learning_rate": 2.0368708249192182e-07, "loss": 1.3503, "step": 8546 }, { "epoch": 0.91, "learning_rate": 2.0319932839056365e-07, "loss": 1.4346, "step": 8547 }, { "epoch": 0.91, "learning_rate": 2.027121468703347e-07, "loss": 1.3463, "step": 8548 }, { "epoch": 0.91, "learning_rate": 2.0222553798938836e-07, "loss": 1.269, "step": 8549 }, { "epoch": 0.91, "learning_rate": 2.0173950180581047e-07, "loss": 1.3722, "step": 8550 }, { "epoch": 0.91, "learning_rate": 2.0125403837761736e-07, "loss": 1.3474, "step": 8551 }, { "epoch": 0.91, "learning_rate": 2.0076914776275824e-07, "loss": 1.1708, "step": 8552 }, { "epoch": 0.91, "learning_rate": 2.002848300191118e-07, "loss": 1.2926, "step": 8553 }, { "epoch": 0.91, "learning_rate": 1.9980108520449127e-07, "loss": 1.2102, "step": 8554 }, { "epoch": 0.91, "learning_rate": 1.9931791337663875e-07, "loss": 1.692, "step": 8555 }, { "epoch": 0.91, "learning_rate": 1.988353145932298e-07, "loss": 1.2959, "step": 8556 }, { "epoch": 0.91, "learning_rate": 1.983532889118711e-07, "loss": 1.3021, "step": 8557 }, { "epoch": 0.91, "learning_rate": 1.9787183639010055e-07, "loss": 1.2339, "step": 8558 }, { "epoch": 0.91, "learning_rate": 1.9739095708538714e-07, "loss": 1.5834, "step": 8559 }, { "epoch": 0.91, "learning_rate": 1.9691065105513328e-07, "loss": 1.5021, "step": 8560 }, { "epoch": 0.91, "learning_rate": 1.9643091835667095e-07, "loss": 1.3646, "step": 8561 }, { "epoch": 0.91, "learning_rate": 1.9595175904726481e-07, "loss": 1.2226, "step": 8562 }, { "epoch": 0.91, "learning_rate": 1.954731731841114e-07, "loss": 1.3507, "step": 8563 }, { "epoch": 0.91, "learning_rate": 1.9499516082433723e-07, "loss": 1.0494, "step": 8564 }, { "epoch": 0.91, "learning_rate": 1.9451772202500163e-07, "loss": 1.5174, "step": 8565 }, { "epoch": 0.91, "learning_rate": 1.9404085684309516e-07, "loss": 1.4551, "step": 8566 }, { "epoch": 0.91, "learning_rate": 1.9356456533554113e-07, "loss": 1.1459, "step": 8567 }, { "epoch": 0.91, "learning_rate": 1.9308884755919132e-07, "loss": 1.3931, "step": 8568 }, { "epoch": 0.91, "learning_rate": 1.9261370357083198e-07, "loss": 1.1856, "step": 8569 }, { "epoch": 0.91, "learning_rate": 1.9213913342717995e-07, "loss": 1.6248, "step": 8570 }, { "epoch": 0.91, "learning_rate": 1.9166513718488155e-07, "loss": 1.3745, "step": 8571 }, { "epoch": 0.91, "learning_rate": 1.911917149005188e-07, "loss": 1.4101, "step": 8572 }, { "epoch": 0.91, "learning_rate": 1.9071886663060202e-07, "loss": 1.3643, "step": 8573 }, { "epoch": 0.91, "learning_rate": 1.902465924315733e-07, "loss": 1.4405, "step": 8574 }, { "epoch": 0.91, "learning_rate": 1.897748923598075e-07, "loss": 1.4256, "step": 8575 }, { "epoch": 0.91, "learning_rate": 1.8930376647160908e-07, "loss": 1.2069, "step": 8576 }, { "epoch": 0.91, "learning_rate": 1.8883321482321583e-07, "loss": 1.6755, "step": 8577 }, { "epoch": 0.91, "learning_rate": 1.8836323747079665e-07, "loss": 1.3741, "step": 8578 }, { "epoch": 0.92, "learning_rate": 1.8789383447045119e-07, "loss": 1.4709, "step": 8579 }, { "epoch": 0.92, "learning_rate": 1.8742500587820955e-07, "loss": 1.3297, "step": 8580 }, { "epoch": 0.92, "learning_rate": 1.8695675175003592e-07, "loss": 1.5233, "step": 8581 }, { "epoch": 0.92, "learning_rate": 1.86489072141825e-07, "loss": 1.3523, "step": 8582 }, { "epoch": 0.92, "learning_rate": 1.86021967109401e-07, "loss": 1.4334, "step": 8583 }, { "epoch": 0.92, "learning_rate": 1.855554367085216e-07, "loss": 1.3059, "step": 8584 }, { "epoch": 0.92, "learning_rate": 1.850894809948761e-07, "loss": 1.3237, "step": 8585 }, { "epoch": 0.92, "learning_rate": 1.8462410002408228e-07, "loss": 1.594, "step": 8586 }, { "epoch": 0.92, "learning_rate": 1.8415929385169285e-07, "loss": 1.34, "step": 8587 }, { "epoch": 0.92, "learning_rate": 1.8369506253319124e-07, "loss": 1.299, "step": 8588 }, { "epoch": 0.92, "learning_rate": 1.8323140612399038e-07, "loss": 1.3764, "step": 8589 }, { "epoch": 0.92, "learning_rate": 1.827683246794343e-07, "loss": 1.5408, "step": 8590 }, { "epoch": 0.92, "learning_rate": 1.8230581825480264e-07, "loss": 1.416, "step": 8591 }, { "epoch": 0.92, "learning_rate": 1.8184388690530242e-07, "loss": 1.1912, "step": 8592 }, { "epoch": 0.92, "learning_rate": 1.813825306860728e-07, "loss": 1.2597, "step": 8593 }, { "epoch": 0.92, "learning_rate": 1.809217496521848e-07, "loss": 1.2918, "step": 8594 }, { "epoch": 0.92, "learning_rate": 1.804615438586399e-07, "loss": 1.2518, "step": 8595 }, { "epoch": 0.92, "learning_rate": 1.8000191336037252e-07, "loss": 1.2962, "step": 8596 }, { "epoch": 0.92, "learning_rate": 1.7954285821224827e-07, "loss": 1.3796, "step": 8597 }, { "epoch": 0.92, "learning_rate": 1.7908437846906158e-07, "loss": 1.4852, "step": 8598 }, { "epoch": 0.92, "learning_rate": 1.786264741855409e-07, "loss": 1.5295, "step": 8599 }, { "epoch": 0.92, "learning_rate": 1.7816914541634423e-07, "loss": 1.2476, "step": 8600 }, { "epoch": 0.92, "learning_rate": 1.7771239221606285e-07, "loss": 1.194, "step": 8601 }, { "epoch": 0.92, "learning_rate": 1.7725621463921761e-07, "loss": 1.12, "step": 8602 }, { "epoch": 0.92, "learning_rate": 1.7680061274026105e-07, "loss": 1.4672, "step": 8603 }, { "epoch": 0.92, "learning_rate": 1.7634558657357748e-07, "loss": 1.4105, "step": 8604 }, { "epoch": 0.92, "learning_rate": 1.7589113619348174e-07, "loss": 1.393, "step": 8605 }, { "epoch": 0.92, "learning_rate": 1.7543726165421993e-07, "loss": 1.1493, "step": 8606 }, { "epoch": 0.92, "learning_rate": 1.7498396300997146e-07, "loss": 1.4231, "step": 8607 }, { "epoch": 0.92, "learning_rate": 1.7453124031484303e-07, "loss": 1.2131, "step": 8608 }, { "epoch": 0.92, "learning_rate": 1.7407909362287756e-07, "loss": 1.1944, "step": 8609 }, { "epoch": 0.92, "learning_rate": 1.736275229880441e-07, "loss": 1.5232, "step": 8610 }, { "epoch": 0.92, "learning_rate": 1.7317652846424727e-07, "loss": 1.298, "step": 8611 }, { "epoch": 0.92, "learning_rate": 1.7272611010532014e-07, "loss": 1.4044, "step": 8612 }, { "epoch": 0.92, "learning_rate": 1.7227626796502807e-07, "loss": 1.4303, "step": 8613 }, { "epoch": 0.92, "learning_rate": 1.718270020970675e-07, "loss": 1.6174, "step": 8614 }, { "epoch": 0.92, "learning_rate": 1.713783125550661e-07, "loss": 1.4872, "step": 8615 }, { "epoch": 0.92, "learning_rate": 1.7093019939258327e-07, "loss": 1.2464, "step": 8616 }, { "epoch": 0.92, "learning_rate": 1.7048266266310842e-07, "loss": 1.3115, "step": 8617 }, { "epoch": 0.92, "learning_rate": 1.7003570242006273e-07, "loss": 1.1479, "step": 8618 }, { "epoch": 0.92, "learning_rate": 1.6958931871679908e-07, "loss": 1.2859, "step": 8619 }, { "epoch": 0.92, "learning_rate": 1.6914351160660036e-07, "loss": 1.4372, "step": 8620 }, { "epoch": 0.92, "learning_rate": 1.6869828114268183e-07, "loss": 1.4054, "step": 8621 }, { "epoch": 0.92, "learning_rate": 1.6825362737818985e-07, "loss": 1.4549, "step": 8622 }, { "epoch": 0.92, "learning_rate": 1.6780955036620138e-07, "loss": 1.2158, "step": 8623 }, { "epoch": 0.92, "learning_rate": 1.6736605015972406e-07, "loss": 1.4742, "step": 8624 }, { "epoch": 0.92, "learning_rate": 1.6692312681169775e-07, "loss": 1.3162, "step": 8625 }, { "epoch": 0.92, "learning_rate": 1.664807803749935e-07, "loss": 1.0943, "step": 8626 }, { "epoch": 0.92, "learning_rate": 1.660390109024118e-07, "loss": 1.7391, "step": 8627 }, { "epoch": 0.92, "learning_rate": 1.6559781844668666e-07, "loss": 1.4707, "step": 8628 }, { "epoch": 0.92, "learning_rate": 1.65157203060482e-07, "loss": 1.5662, "step": 8629 }, { "epoch": 0.92, "learning_rate": 1.6471716479639077e-07, "loss": 1.4526, "step": 8630 }, { "epoch": 0.92, "learning_rate": 1.6427770370694208e-07, "loss": 1.3025, "step": 8631 }, { "epoch": 0.92, "learning_rate": 1.6383881984459227e-07, "loss": 1.339, "step": 8632 }, { "epoch": 0.92, "learning_rate": 1.6340051326172834e-07, "loss": 1.4817, "step": 8633 }, { "epoch": 0.92, "learning_rate": 1.6296278401067122e-07, "loss": 1.2536, "step": 8634 }, { "epoch": 0.92, "learning_rate": 1.6252563214367134e-07, "loss": 1.6477, "step": 8635 }, { "epoch": 0.92, "learning_rate": 1.6208905771290974e-07, "loss": 1.4361, "step": 8636 }, { "epoch": 0.92, "learning_rate": 1.6165306077049969e-07, "loss": 1.244, "step": 8637 }, { "epoch": 0.92, "learning_rate": 1.6121764136848517e-07, "loss": 1.3085, "step": 8638 }, { "epoch": 0.92, "learning_rate": 1.6078279955883957e-07, "loss": 1.7862, "step": 8639 }, { "epoch": 0.92, "learning_rate": 1.603485353934703e-07, "loss": 1.3355, "step": 8640 }, { "epoch": 0.92, "learning_rate": 1.5991484892421416e-07, "loss": 1.4466, "step": 8641 }, { "epoch": 0.92, "learning_rate": 1.594817402028387e-07, "loss": 1.5468, "step": 8642 }, { "epoch": 0.92, "learning_rate": 1.5904920928104196e-07, "loss": 1.2471, "step": 8643 }, { "epoch": 0.92, "learning_rate": 1.5861725621045542e-07, "loss": 1.3604, "step": 8644 }, { "epoch": 0.92, "learning_rate": 1.5818588104264065e-07, "loss": 1.4166, "step": 8645 }, { "epoch": 0.92, "learning_rate": 1.577550838290881e-07, "loss": 1.3562, "step": 8646 }, { "epoch": 0.92, "learning_rate": 1.5732486462122166e-07, "loss": 1.2729, "step": 8647 }, { "epoch": 0.92, "learning_rate": 1.5689522347039577e-07, "loss": 1.3482, "step": 8648 }, { "epoch": 0.92, "learning_rate": 1.564661604278944e-07, "loss": 1.3388, "step": 8649 }, { "epoch": 0.92, "learning_rate": 1.56037675544935e-07, "loss": 1.4884, "step": 8650 }, { "epoch": 0.92, "learning_rate": 1.5560976887266433e-07, "loss": 1.5479, "step": 8651 }, { "epoch": 0.92, "learning_rate": 1.5518244046215936e-07, "loss": 1.1607, "step": 8652 }, { "epoch": 0.92, "learning_rate": 1.547556903644304e-07, "loss": 1.4373, "step": 8653 }, { "epoch": 0.92, "learning_rate": 1.5432951863041666e-07, "loss": 1.4477, "step": 8654 }, { "epoch": 0.92, "learning_rate": 1.539039253109892e-07, "loss": 1.309, "step": 8655 }, { "epoch": 0.92, "learning_rate": 1.5347891045695118e-07, "loss": 1.6288, "step": 8656 }, { "epoch": 0.92, "learning_rate": 1.530544741190343e-07, "loss": 1.4341, "step": 8657 }, { "epoch": 0.92, "learning_rate": 1.526306163479019e-07, "loss": 1.2185, "step": 8658 }, { "epoch": 0.92, "learning_rate": 1.5220733719414904e-07, "loss": 1.4195, "step": 8659 }, { "epoch": 0.92, "learning_rate": 1.5178463670830312e-07, "loss": 1.4348, "step": 8660 }, { "epoch": 0.92, "learning_rate": 1.5136251494081822e-07, "loss": 1.6879, "step": 8661 }, { "epoch": 0.92, "learning_rate": 1.5094097194208402e-07, "loss": 1.4675, "step": 8662 }, { "epoch": 0.92, "learning_rate": 1.5052000776241748e-07, "loss": 1.3745, "step": 8663 }, { "epoch": 0.92, "learning_rate": 1.5009962245206845e-07, "loss": 0.979, "step": 8664 }, { "epoch": 0.92, "learning_rate": 1.4967981606121673e-07, "loss": 1.3647, "step": 8665 }, { "epoch": 0.92, "learning_rate": 1.492605886399745e-07, "loss": 1.4297, "step": 8666 }, { "epoch": 0.92, "learning_rate": 1.488419402383834e-07, "loss": 1.3869, "step": 8667 }, { "epoch": 0.92, "learning_rate": 1.484238709064162e-07, "loss": 1.3334, "step": 8668 }, { "epoch": 0.92, "learning_rate": 1.4800638069397633e-07, "loss": 1.4885, "step": 8669 }, { "epoch": 0.92, "learning_rate": 1.4758946965089894e-07, "loss": 1.5205, "step": 8670 }, { "epoch": 0.92, "learning_rate": 1.4717313782694975e-07, "loss": 1.4285, "step": 8671 }, { "epoch": 0.93, "learning_rate": 1.467573852718246e-07, "loss": 1.2825, "step": 8672 }, { "epoch": 0.93, "learning_rate": 1.4634221203515097e-07, "loss": 1.3566, "step": 8673 }, { "epoch": 0.93, "learning_rate": 1.459276181664876e-07, "loss": 1.4203, "step": 8674 }, { "epoch": 0.93, "learning_rate": 1.4551360371532264e-07, "loss": 1.3677, "step": 8675 }, { "epoch": 0.93, "learning_rate": 1.4510016873107657e-07, "loss": 1.5043, "step": 8676 }, { "epoch": 0.93, "learning_rate": 1.446873132630988e-07, "loss": 1.2135, "step": 8677 }, { "epoch": 0.93, "learning_rate": 1.4427503736067162e-07, "loss": 1.4821, "step": 8678 }, { "epoch": 0.93, "learning_rate": 1.4386334107300727e-07, "loss": 1.3651, "step": 8679 }, { "epoch": 0.93, "learning_rate": 1.4345222444924867e-07, "loss": 1.4349, "step": 8680 }, { "epoch": 0.93, "learning_rate": 1.4304168753847049e-07, "loss": 1.2797, "step": 8681 }, { "epoch": 0.93, "learning_rate": 1.4263173038967627e-07, "loss": 1.2989, "step": 8682 }, { "epoch": 0.93, "learning_rate": 1.4222235305180133e-07, "loss": 1.467, "step": 8683 }, { "epoch": 0.93, "learning_rate": 1.4181355557371323e-07, "loss": 1.3906, "step": 8684 }, { "epoch": 0.93, "learning_rate": 1.4140533800420853e-07, "loss": 1.3041, "step": 8685 }, { "epoch": 0.93, "learning_rate": 1.409977003920143e-07, "loss": 1.3303, "step": 8686 }, { "epoch": 0.93, "learning_rate": 1.4059064278579005e-07, "loss": 1.3378, "step": 8687 }, { "epoch": 0.93, "learning_rate": 1.401841652341246e-07, "loss": 1.4566, "step": 8688 }, { "epoch": 0.93, "learning_rate": 1.3977826778553805e-07, "loss": 1.3118, "step": 8689 }, { "epoch": 0.93, "learning_rate": 1.393729504884822e-07, "loss": 1.2778, "step": 8690 }, { "epoch": 0.93, "learning_rate": 1.389682133913378e-07, "loss": 1.299, "step": 8691 }, { "epoch": 0.93, "learning_rate": 1.3856405654241678e-07, "loss": 1.4236, "step": 8692 }, { "epoch": 0.93, "learning_rate": 1.3816047998996275e-07, "loss": 1.5024, "step": 8693 }, { "epoch": 0.93, "learning_rate": 1.3775748378215047e-07, "loss": 1.2675, "step": 8694 }, { "epoch": 0.93, "learning_rate": 1.3735506796708375e-07, "loss": 1.6326, "step": 8695 }, { "epoch": 0.93, "learning_rate": 1.3695323259279748e-07, "loss": 1.453, "step": 8696 }, { "epoch": 0.93, "learning_rate": 1.3655197770725826e-07, "loss": 1.0417, "step": 8697 }, { "epoch": 0.93, "learning_rate": 1.361513033583617e-07, "loss": 1.2888, "step": 8698 }, { "epoch": 0.93, "learning_rate": 1.3575120959393672e-07, "loss": 1.5103, "step": 8699 }, { "epoch": 0.93, "learning_rate": 1.3535169646174073e-07, "loss": 1.1501, "step": 8700 }, { "epoch": 0.93, "learning_rate": 1.3495276400946278e-07, "loss": 1.6014, "step": 8701 }, { "epoch": 0.93, "learning_rate": 1.3455441228472199e-07, "loss": 1.4332, "step": 8702 }, { "epoch": 0.93, "learning_rate": 1.3415664133506812e-07, "loss": 1.4697, "step": 8703 }, { "epoch": 0.93, "learning_rate": 1.3375945120798372e-07, "loss": 1.5734, "step": 8704 }, { "epoch": 0.93, "learning_rate": 1.333628419508781e-07, "loss": 1.2848, "step": 8705 }, { "epoch": 0.93, "learning_rate": 1.3296681361109564e-07, "loss": 1.5656, "step": 8706 }, { "epoch": 0.93, "learning_rate": 1.3257136623590795e-07, "loss": 1.3223, "step": 8707 }, { "epoch": 0.93, "learning_rate": 1.3217649987251725e-07, "loss": 1.0103, "step": 8708 }, { "epoch": 0.93, "learning_rate": 1.3178221456806028e-07, "loss": 1.2577, "step": 8709 }, { "epoch": 0.93, "learning_rate": 1.3138851036959998e-07, "loss": 1.3185, "step": 8710 }, { "epoch": 0.93, "learning_rate": 1.3099538732413263e-07, "loss": 1.4491, "step": 8711 }, { "epoch": 0.93, "learning_rate": 1.3060284547858403e-07, "loss": 1.2398, "step": 8712 }, { "epoch": 0.93, "learning_rate": 1.3021088487981058e-07, "loss": 1.2521, "step": 8713 }, { "epoch": 0.93, "learning_rate": 1.2981950557459932e-07, "loss": 1.4856, "step": 8714 }, { "epoch": 0.93, "learning_rate": 1.2942870760966952e-07, "loss": 1.3333, "step": 8715 }, { "epoch": 0.93, "learning_rate": 1.2903849103166831e-07, "loss": 1.3336, "step": 8716 }, { "epoch": 0.93, "learning_rate": 1.2864885588717512e-07, "loss": 1.3158, "step": 8717 }, { "epoch": 0.93, "learning_rate": 1.282598022226994e-07, "loss": 1.2713, "step": 8718 }, { "epoch": 0.93, "learning_rate": 1.2787133008468289e-07, "loss": 1.2737, "step": 8719 }, { "epoch": 0.93, "learning_rate": 1.2748343951949405e-07, "loss": 1.1977, "step": 8720 }, { "epoch": 0.93, "learning_rate": 1.270961305734364e-07, "loss": 1.3357, "step": 8721 }, { "epoch": 0.93, "learning_rate": 1.267094032927413e-07, "loss": 1.379, "step": 8722 }, { "epoch": 0.93, "learning_rate": 1.263232577235707e-07, "loss": 1.5189, "step": 8723 }, { "epoch": 0.93, "learning_rate": 1.2593769391201827e-07, "loss": 1.5173, "step": 8724 }, { "epoch": 0.93, "learning_rate": 1.2555271190410833e-07, "loss": 1.3859, "step": 8725 }, { "epoch": 0.93, "learning_rate": 1.2516831174579357e-07, "loss": 1.2847, "step": 8726 }, { "epoch": 0.93, "learning_rate": 1.247844934829606e-07, "loss": 1.2163, "step": 8727 }, { "epoch": 0.93, "learning_rate": 1.2440125716142392e-07, "loss": 1.331, "step": 8728 }, { "epoch": 0.93, "learning_rate": 1.2401860282692913e-07, "loss": 1.325, "step": 8729 }, { "epoch": 0.93, "learning_rate": 1.2363653052515302e-07, "loss": 1.2966, "step": 8730 }, { "epoch": 0.93, "learning_rate": 1.2325504030170243e-07, "loss": 1.4025, "step": 8731 }, { "epoch": 0.93, "learning_rate": 1.2287413220211487e-07, "loss": 1.2419, "step": 8732 }, { "epoch": 0.93, "learning_rate": 1.2249380627185781e-07, "loss": 1.2392, "step": 8733 }, { "epoch": 0.93, "learning_rate": 1.2211406255633107e-07, "loss": 1.4929, "step": 8734 }, { "epoch": 0.93, "learning_rate": 1.2173490110086228e-07, "loss": 1.6146, "step": 8735 }, { "epoch": 0.93, "learning_rate": 1.2135632195071133e-07, "loss": 1.3759, "step": 8736 }, { "epoch": 0.93, "learning_rate": 1.2097832515106767e-07, "loss": 1.6359, "step": 8737 }, { "epoch": 0.93, "learning_rate": 1.206009107470535e-07, "loss": 1.5248, "step": 8738 }, { "epoch": 0.93, "learning_rate": 1.202240787837178e-07, "loss": 1.5994, "step": 8739 }, { "epoch": 0.93, "learning_rate": 1.1984782930604345e-07, "loss": 1.3097, "step": 8740 }, { "epoch": 0.93, "learning_rate": 1.194721623589412e-07, "loss": 1.3436, "step": 8741 }, { "epoch": 0.93, "learning_rate": 1.1909707798725412e-07, "loss": 1.3108, "step": 8742 }, { "epoch": 0.93, "learning_rate": 1.1872257623575412e-07, "loss": 1.5485, "step": 8743 }, { "epoch": 0.93, "learning_rate": 1.1834865714914601e-07, "loss": 1.3597, "step": 8744 }, { "epoch": 0.93, "learning_rate": 1.1797532077206187e-07, "loss": 1.4241, "step": 8745 }, { "epoch": 0.93, "learning_rate": 1.1760256714906715e-07, "loss": 1.3303, "step": 8746 }, { "epoch": 0.93, "learning_rate": 1.1723039632465516e-07, "loss": 1.2074, "step": 8747 }, { "epoch": 0.93, "learning_rate": 1.1685880834325203e-07, "loss": 1.383, "step": 8748 }, { "epoch": 0.93, "learning_rate": 1.1648780324921283e-07, "loss": 1.2925, "step": 8749 }, { "epoch": 0.93, "learning_rate": 1.1611738108682325e-07, "loss": 1.3786, "step": 8750 }, { "epoch": 0.93, "learning_rate": 1.1574754190030014e-07, "loss": 1.2753, "step": 8751 }, { "epoch": 0.93, "learning_rate": 1.1537828573378929e-07, "loss": 1.2847, "step": 8752 }, { "epoch": 0.93, "learning_rate": 1.1500961263136878e-07, "loss": 1.5149, "step": 8753 }, { "epoch": 0.93, "learning_rate": 1.1464152263704565e-07, "loss": 1.3669, "step": 8754 }, { "epoch": 0.93, "learning_rate": 1.142740157947575e-07, "loss": 1.358, "step": 8755 }, { "epoch": 0.93, "learning_rate": 1.1390709214837314e-07, "loss": 1.4396, "step": 8756 }, { "epoch": 0.93, "learning_rate": 1.1354075174169088e-07, "loss": 1.6265, "step": 8757 }, { "epoch": 0.93, "learning_rate": 1.1317499461844017e-07, "loss": 1.5385, "step": 8758 }, { "epoch": 0.93, "learning_rate": 1.1280982082228054e-07, "loss": 1.4205, "step": 8759 }, { "epoch": 0.93, "learning_rate": 1.12445230396801e-07, "loss": 1.1151, "step": 8760 }, { "epoch": 0.93, "learning_rate": 1.1208122338552229e-07, "loss": 1.426, "step": 8761 }, { "epoch": 0.93, "learning_rate": 1.1171779983189523e-07, "loss": 1.3669, "step": 8762 }, { "epoch": 0.93, "learning_rate": 1.1135495977930011e-07, "loss": 1.5344, "step": 8763 }, { "epoch": 0.93, "learning_rate": 1.1099270327104839e-07, "loss": 1.2327, "step": 8764 }, { "epoch": 0.93, "learning_rate": 1.1063103035038214e-07, "loss": 1.4452, "step": 8765 }, { "epoch": 0.94, "learning_rate": 1.1026994106047296e-07, "loss": 1.3245, "step": 8766 }, { "epoch": 0.94, "learning_rate": 1.0990943544442189e-07, "loss": 1.1403, "step": 8767 }, { "epoch": 0.94, "learning_rate": 1.0954951354526344e-07, "loss": 1.1072, "step": 8768 }, { "epoch": 0.94, "learning_rate": 1.0919017540595933e-07, "loss": 1.3409, "step": 8769 }, { "epoch": 0.94, "learning_rate": 1.0883142106940303e-07, "loss": 1.5856, "step": 8770 }, { "epoch": 0.94, "learning_rate": 1.0847325057841862e-07, "loss": 1.3606, "step": 8771 }, { "epoch": 0.94, "learning_rate": 1.0811566397575912e-07, "loss": 1.2303, "step": 8772 }, { "epoch": 0.94, "learning_rate": 1.0775866130410928e-07, "loss": 1.5828, "step": 8773 }, { "epoch": 0.94, "learning_rate": 1.0740224260608334e-07, "loss": 1.6467, "step": 8774 }, { "epoch": 0.94, "learning_rate": 1.0704640792422616e-07, "loss": 1.3696, "step": 8775 }, { "epoch": 0.94, "learning_rate": 1.066911573010121e-07, "loss": 1.572, "step": 8776 }, { "epoch": 0.94, "learning_rate": 1.0633649077884667e-07, "loss": 1.4243, "step": 8777 }, { "epoch": 0.94, "learning_rate": 1.0598240840006658e-07, "loss": 1.5597, "step": 8778 }, { "epoch": 0.94, "learning_rate": 1.0562891020693633e-07, "loss": 1.2911, "step": 8779 }, { "epoch": 0.94, "learning_rate": 1.0527599624165275e-07, "loss": 1.28, "step": 8780 }, { "epoch": 0.94, "learning_rate": 1.0492366654634211e-07, "loss": 1.2115, "step": 8781 }, { "epoch": 0.94, "learning_rate": 1.0457192116306137e-07, "loss": 1.3667, "step": 8782 }, { "epoch": 0.94, "learning_rate": 1.0422076013379634e-07, "loss": 1.3415, "step": 8783 }, { "epoch": 0.94, "learning_rate": 1.0387018350046519e-07, "loss": 1.4814, "step": 8784 }, { "epoch": 0.94, "learning_rate": 1.0352019130491552e-07, "loss": 1.3113, "step": 8785 }, { "epoch": 0.94, "learning_rate": 1.0317078358892286e-07, "loss": 1.3102, "step": 8786 }, { "epoch": 0.94, "learning_rate": 1.0282196039419823e-07, "loss": 1.2546, "step": 8787 }, { "epoch": 0.94, "learning_rate": 1.024737217623778e-07, "loss": 1.4622, "step": 8788 }, { "epoch": 0.94, "learning_rate": 1.0212606773502943e-07, "loss": 1.5187, "step": 8789 }, { "epoch": 0.94, "learning_rate": 1.0177899835365323e-07, "loss": 1.3134, "step": 8790 }, { "epoch": 0.94, "learning_rate": 1.0143251365967666e-07, "loss": 0.9563, "step": 8791 }, { "epoch": 0.94, "learning_rate": 1.0108661369445882e-07, "loss": 1.6426, "step": 8792 }, { "epoch": 0.94, "learning_rate": 1.0074129849928948e-07, "loss": 1.2245, "step": 8793 }, { "epoch": 0.94, "learning_rate": 1.0039656811538789e-07, "loss": 1.4999, "step": 8794 }, { "epoch": 0.94, "learning_rate": 1.000524225839028e-07, "loss": 1.3299, "step": 8795 }, { "epoch": 0.94, "learning_rate": 9.970886194591467e-08, "loss": 1.3746, "step": 8796 }, { "epoch": 0.94, "learning_rate": 9.936588624243348e-08, "loss": 1.5167, "step": 8797 }, { "epoch": 0.94, "learning_rate": 9.902349551439871e-08, "loss": 1.413, "step": 8798 }, { "epoch": 0.94, "learning_rate": 9.8681689802681e-08, "loss": 1.2993, "step": 8799 }, { "epoch": 0.94, "learning_rate": 9.834046914808104e-08, "loss": 1.6829, "step": 8800 }, { "epoch": 0.94, "learning_rate": 9.799983359132848e-08, "loss": 1.1959, "step": 8801 }, { "epoch": 0.94, "learning_rate": 9.765978317308522e-08, "loss": 1.4619, "step": 8802 }, { "epoch": 0.94, "learning_rate": 9.732031793394158e-08, "loss": 1.37, "step": 8803 }, { "epoch": 0.94, "learning_rate": 9.698143791441794e-08, "loss": 1.2952, "step": 8804 }, { "epoch": 0.94, "learning_rate": 9.664314315496692e-08, "loss": 1.3129, "step": 8805 }, { "epoch": 0.94, "learning_rate": 9.630543369596846e-08, "loss": 1.3177, "step": 8806 }, { "epoch": 0.94, "learning_rate": 9.596830957773528e-08, "loss": 1.5249, "step": 8807 }, { "epoch": 0.94, "learning_rate": 9.5631770840508e-08, "loss": 1.5274, "step": 8808 }, { "epoch": 0.94, "learning_rate": 9.529581752445949e-08, "loss": 1.6146, "step": 8809 }, { "epoch": 0.94, "learning_rate": 9.496044966968987e-08, "loss": 1.5816, "step": 8810 }, { "epoch": 0.94, "learning_rate": 9.462566731623213e-08, "loss": 1.7283, "step": 8811 }, { "epoch": 0.94, "learning_rate": 9.429147050404874e-08, "loss": 1.5413, "step": 8812 }, { "epoch": 0.94, "learning_rate": 9.395785927303169e-08, "loss": 1.4782, "step": 8813 }, { "epoch": 0.94, "learning_rate": 9.36248336630019e-08, "loss": 1.5648, "step": 8814 }, { "epoch": 0.94, "learning_rate": 9.329239371371312e-08, "loss": 1.3853, "step": 8815 }, { "epoch": 0.94, "learning_rate": 9.296053946484807e-08, "loss": 1.1513, "step": 8816 }, { "epoch": 0.94, "learning_rate": 9.262927095601782e-08, "loss": 1.5273, "step": 8817 }, { "epoch": 0.94, "learning_rate": 9.22985882267663e-08, "loss": 1.3383, "step": 8818 }, { "epoch": 0.94, "learning_rate": 9.196849131656638e-08, "loss": 1.289, "step": 8819 }, { "epoch": 0.94, "learning_rate": 9.163898026481876e-08, "loss": 1.6691, "step": 8820 }, { "epoch": 0.94, "learning_rate": 9.131005511085867e-08, "loss": 1.4754, "step": 8821 }, { "epoch": 0.94, "learning_rate": 9.098171589394855e-08, "loss": 1.2347, "step": 8822 }, { "epoch": 0.94, "learning_rate": 9.065396265327986e-08, "loss": 1.6778, "step": 8823 }, { "epoch": 0.94, "learning_rate": 9.03267954279774e-08, "loss": 1.2031, "step": 8824 }, { "epoch": 0.94, "learning_rate": 9.000021425709326e-08, "loss": 1.4272, "step": 8825 }, { "epoch": 0.94, "learning_rate": 8.967421917961072e-08, "loss": 1.2881, "step": 8826 }, { "epoch": 0.94, "learning_rate": 8.934881023444365e-08, "loss": 1.5139, "step": 8827 }, { "epoch": 0.94, "learning_rate": 8.902398746043428e-08, "loss": 1.6813, "step": 8828 }, { "epoch": 0.94, "learning_rate": 8.869975089635552e-08, "loss": 1.1717, "step": 8829 }, { "epoch": 0.94, "learning_rate": 8.83761005809114e-08, "loss": 1.5662, "step": 8830 }, { "epoch": 0.94, "learning_rate": 8.805303655273601e-08, "loss": 1.2962, "step": 8831 }, { "epoch": 0.94, "learning_rate": 8.773055885039072e-08, "loss": 1.3594, "step": 8832 }, { "epoch": 0.94, "learning_rate": 8.740866751236976e-08, "loss": 1.475, "step": 8833 }, { "epoch": 0.94, "learning_rate": 8.708736257709738e-08, "loss": 1.2819, "step": 8834 }, { "epoch": 0.94, "learning_rate": 8.676664408292457e-08, "loss": 1.3122, "step": 8835 }, { "epoch": 0.94, "learning_rate": 8.644651206813625e-08, "loss": 1.1769, "step": 8836 }, { "epoch": 0.94, "learning_rate": 8.612696657094633e-08, "loss": 1.0845, "step": 8837 }, { "epoch": 0.94, "learning_rate": 8.580800762949704e-08, "loss": 1.6246, "step": 8838 }, { "epoch": 0.94, "learning_rate": 8.548963528186071e-08, "loss": 1.2936, "step": 8839 }, { "epoch": 0.94, "learning_rate": 8.517184956604252e-08, "loss": 1.2232, "step": 8840 }, { "epoch": 0.94, "learning_rate": 8.485465051997488e-08, "loss": 1.3643, "step": 8841 }, { "epoch": 0.94, "learning_rate": 8.453803818152029e-08, "loss": 1.441, "step": 8842 }, { "epoch": 0.94, "learning_rate": 8.422201258847351e-08, "loss": 1.2982, "step": 8843 }, { "epoch": 0.94, "learning_rate": 8.39065737785566e-08, "loss": 1.2201, "step": 8844 }, { "epoch": 0.94, "learning_rate": 8.359172178942165e-08, "loss": 1.1743, "step": 8845 }, { "epoch": 0.94, "learning_rate": 8.327745665865417e-08, "loss": 1.6148, "step": 8846 }, { "epoch": 0.94, "learning_rate": 8.296377842376524e-08, "loss": 1.338, "step": 8847 }, { "epoch": 0.94, "learning_rate": 8.265068712219826e-08, "loss": 1.3545, "step": 8848 }, { "epoch": 0.94, "learning_rate": 8.233818279132611e-08, "loss": 1.3997, "step": 8849 }, { "epoch": 0.94, "learning_rate": 8.202626546845172e-08, "loss": 1.405, "step": 8850 }, { "epoch": 0.94, "learning_rate": 8.171493519080753e-08, "loss": 1.2094, "step": 8851 }, { "epoch": 0.94, "learning_rate": 8.14041919955566e-08, "loss": 1.2829, "step": 8852 }, { "epoch": 0.94, "learning_rate": 8.109403591979148e-08, "loss": 1.4322, "step": 8853 }, { "epoch": 0.94, "learning_rate": 8.078446700053422e-08, "loss": 1.6735, "step": 8854 }, { "epoch": 0.94, "learning_rate": 8.047548527473748e-08, "loss": 1.5485, "step": 8855 }, { "epoch": 0.94, "learning_rate": 8.016709077928397e-08, "loss": 1.5014, "step": 8856 }, { "epoch": 0.94, "learning_rate": 7.985928355098483e-08, "loss": 1.1823, "step": 8857 }, { "epoch": 0.94, "learning_rate": 7.955206362658341e-08, "loss": 1.2514, "step": 8858 }, { "epoch": 0.94, "learning_rate": 7.924543104275095e-08, "loss": 1.5568, "step": 8859 }, { "epoch": 0.95, "learning_rate": 7.893938583608985e-08, "loss": 1.2522, "step": 8860 }, { "epoch": 0.95, "learning_rate": 7.863392804313086e-08, "loss": 1.3063, "step": 8861 }, { "epoch": 0.95, "learning_rate": 7.832905770033705e-08, "loss": 1.2086, "step": 8862 }, { "epoch": 0.95, "learning_rate": 7.802477484409931e-08, "loss": 1.5222, "step": 8863 }, { "epoch": 0.95, "learning_rate": 7.77210795107386e-08, "loss": 1.2668, "step": 8864 }, { "epoch": 0.95, "learning_rate": 7.7417971736507e-08, "loss": 1.1987, "step": 8865 }, { "epoch": 0.95, "learning_rate": 7.711545155758615e-08, "loss": 1.394, "step": 8866 }, { "epoch": 0.95, "learning_rate": 7.681351901008494e-08, "loss": 1.2987, "step": 8867 }, { "epoch": 0.95, "learning_rate": 7.651217413004674e-08, "loss": 1.3895, "step": 8868 }, { "epoch": 0.95, "learning_rate": 7.621141695344003e-08, "loss": 1.291, "step": 8869 }, { "epoch": 0.95, "learning_rate": 7.591124751616718e-08, "loss": 1.3878, "step": 8870 }, { "epoch": 0.95, "learning_rate": 7.561166585405789e-08, "loss": 1.3568, "step": 8871 }, { "epoch": 0.95, "learning_rate": 7.53126720028724e-08, "loss": 1.5116, "step": 8872 }, { "epoch": 0.95, "learning_rate": 7.501426599830053e-08, "loss": 1.2426, "step": 8873 }, { "epoch": 0.95, "learning_rate": 7.47164478759621e-08, "loss": 1.3835, "step": 8874 }, { "epoch": 0.95, "learning_rate": 7.441921767140814e-08, "loss": 1.3923, "step": 8875 }, { "epoch": 0.95, "learning_rate": 7.412257542011692e-08, "loss": 1.3765, "step": 8876 }, { "epoch": 0.95, "learning_rate": 7.382652115749789e-08, "loss": 1.5185, "step": 8877 }, { "epoch": 0.95, "learning_rate": 7.353105491889112e-08, "loss": 1.4568, "step": 8878 }, { "epoch": 0.95, "learning_rate": 7.323617673956452e-08, "loss": 1.2822, "step": 8879 }, { "epoch": 0.95, "learning_rate": 7.294188665471769e-08, "loss": 1.3218, "step": 8880 }, { "epoch": 0.95, "learning_rate": 7.264818469947921e-08, "loss": 1.3916, "step": 8881 }, { "epoch": 0.95, "learning_rate": 7.235507090890715e-08, "loss": 1.4823, "step": 8882 }, { "epoch": 0.95, "learning_rate": 7.206254531799018e-08, "loss": 1.3395, "step": 8883 }, { "epoch": 0.95, "learning_rate": 7.177060796164536e-08, "loss": 1.2727, "step": 8884 }, { "epoch": 0.95, "learning_rate": 7.147925887472096e-08, "loss": 1.4204, "step": 8885 }, { "epoch": 0.95, "learning_rate": 7.118849809199524e-08, "loss": 1.5905, "step": 8886 }, { "epoch": 0.95, "learning_rate": 7.089832564817434e-08, "loss": 1.5965, "step": 8887 }, { "epoch": 0.95, "learning_rate": 7.06087415778961e-08, "loss": 1.2936, "step": 8888 }, { "epoch": 0.95, "learning_rate": 7.031974591572732e-08, "loss": 1.3338, "step": 8889 }, { "epoch": 0.95, "learning_rate": 7.003133869616485e-08, "loss": 1.4022, "step": 8890 }, { "epoch": 0.95, "learning_rate": 6.974351995363449e-08, "loss": 1.4178, "step": 8891 }, { "epoch": 0.95, "learning_rate": 6.945628972249208e-08, "loss": 1.2381, "step": 8892 }, { "epoch": 0.95, "learning_rate": 6.91696480370252e-08, "loss": 1.4149, "step": 8893 }, { "epoch": 0.95, "learning_rate": 6.888359493144758e-08, "loss": 1.2815, "step": 8894 }, { "epoch": 0.95, "learning_rate": 6.859813043990526e-08, "loss": 1.4888, "step": 8895 }, { "epoch": 0.95, "learning_rate": 6.831325459647375e-08, "loss": 1.3657, "step": 8896 }, { "epoch": 0.95, "learning_rate": 6.802896743515807e-08, "loss": 1.3174, "step": 8897 }, { "epoch": 0.95, "learning_rate": 6.77452689898922e-08, "loss": 1.3995, "step": 8898 }, { "epoch": 0.95, "learning_rate": 6.746215929454014e-08, "loss": 1.2287, "step": 8899 }, { "epoch": 0.95, "learning_rate": 6.71796383828971e-08, "loss": 1.1186, "step": 8900 }, { "epoch": 0.95, "learning_rate": 6.689770628868609e-08, "loss": 1.3685, "step": 8901 }, { "epoch": 0.95, "learning_rate": 6.661636304556074e-08, "loss": 1.2055, "step": 8902 }, { "epoch": 0.95, "learning_rate": 6.63356086871042e-08, "loss": 1.4755, "step": 8903 }, { "epoch": 0.95, "learning_rate": 6.605544324682855e-08, "loss": 1.3507, "step": 8904 }, { "epoch": 0.95, "learning_rate": 6.577586675817816e-08, "loss": 1.4927, "step": 8905 }, { "epoch": 0.95, "learning_rate": 6.549687925452408e-08, "loss": 1.3943, "step": 8906 }, { "epoch": 0.95, "learning_rate": 6.521848076916859e-08, "loss": 1.4794, "step": 8907 }, { "epoch": 0.95, "learning_rate": 6.494067133534288e-08, "loss": 1.621, "step": 8908 }, { "epoch": 0.95, "learning_rate": 6.46634509862093e-08, "loss": 1.3153, "step": 8909 }, { "epoch": 0.95, "learning_rate": 6.438681975485805e-08, "loss": 1.5258, "step": 8910 }, { "epoch": 0.95, "learning_rate": 6.411077767431051e-08, "loss": 1.3557, "step": 8911 }, { "epoch": 0.95, "learning_rate": 6.383532477751641e-08, "loss": 1.5139, "step": 8912 }, { "epoch": 0.95, "learning_rate": 6.356046109735614e-08, "loss": 1.4941, "step": 8913 }, { "epoch": 0.95, "learning_rate": 6.328618666663899e-08, "loss": 1.3411, "step": 8914 }, { "epoch": 0.95, "learning_rate": 6.301250151810545e-08, "loss": 1.3175, "step": 8915 }, { "epoch": 0.95, "learning_rate": 6.273940568442327e-08, "loss": 1.5511, "step": 8916 }, { "epoch": 0.95, "learning_rate": 6.246689919819249e-08, "loss": 1.2962, "step": 8917 }, { "epoch": 0.95, "learning_rate": 6.219498209194042e-08, "loss": 1.7309, "step": 8918 }, { "epoch": 0.95, "learning_rate": 6.192365439812553e-08, "loss": 1.4217, "step": 8919 }, { "epoch": 0.95, "learning_rate": 6.165291614913527e-08, "loss": 1.4605, "step": 8920 }, { "epoch": 0.95, "learning_rate": 6.13827673772871e-08, "loss": 1.2455, "step": 8921 }, { "epoch": 0.95, "learning_rate": 6.111320811482802e-08, "loss": 1.4721, "step": 8922 }, { "epoch": 0.95, "learning_rate": 6.084423839393394e-08, "loss": 1.3037, "step": 8923 }, { "epoch": 0.95, "learning_rate": 6.057585824671252e-08, "loss": 1.3464, "step": 8924 }, { "epoch": 0.95, "learning_rate": 6.030806770519815e-08, "loss": 1.2094, "step": 8925 }, { "epoch": 0.95, "learning_rate": 6.004086680135634e-08, "loss": 1.6414, "step": 8926 }, { "epoch": 0.95, "learning_rate": 5.977425556708327e-08, "loss": 1.3386, "step": 8927 }, { "epoch": 0.95, "learning_rate": 5.9508234034202364e-08, "loss": 1.5575, "step": 8928 }, { "epoch": 0.95, "learning_rate": 5.924280223446821e-08, "loss": 1.5081, "step": 8929 }, { "epoch": 0.95, "learning_rate": 5.897796019956492e-08, "loss": 1.4257, "step": 8930 }, { "epoch": 0.95, "learning_rate": 5.871370796110665e-08, "loss": 1.4964, "step": 8931 }, { "epoch": 0.95, "learning_rate": 5.8450045550634826e-08, "loss": 1.1599, "step": 8932 }, { "epoch": 0.95, "learning_rate": 5.8186972999623724e-08, "loss": 1.5428, "step": 8933 }, { "epoch": 0.95, "learning_rate": 5.7924490339474335e-08, "loss": 1.4507, "step": 8934 }, { "epoch": 0.95, "learning_rate": 5.7662597601519374e-08, "loss": 1.2659, "step": 8935 }, { "epoch": 0.95, "learning_rate": 5.7401294817020505e-08, "loss": 1.4474, "step": 8936 }, { "epoch": 0.95, "learning_rate": 5.7140582017167764e-08, "loss": 1.2376, "step": 8937 }, { "epoch": 0.95, "learning_rate": 5.688045923308183e-08, "loss": 0.8775, "step": 8938 }, { "epoch": 0.95, "learning_rate": 5.662092649581341e-08, "loss": 1.3515, "step": 8939 }, { "epoch": 0.95, "learning_rate": 5.636198383634217e-08, "loss": 1.4288, "step": 8940 }, { "epoch": 0.95, "learning_rate": 5.610363128557727e-08, "loss": 1.4193, "step": 8941 }, { "epoch": 0.95, "learning_rate": 5.584586887435739e-08, "loss": 1.3154, "step": 8942 }, { "epoch": 0.95, "learning_rate": 5.558869663345123e-08, "loss": 1.488, "step": 8943 }, { "epoch": 0.95, "learning_rate": 5.533211459355648e-08, "loss": 1.4203, "step": 8944 }, { "epoch": 0.95, "learning_rate": 5.507612278530083e-08, "loss": 1.4833, "step": 8945 }, { "epoch": 0.95, "learning_rate": 5.482072123924098e-08, "loss": 1.3519, "step": 8946 }, { "epoch": 0.95, "learning_rate": 5.4565909985864196e-08, "loss": 1.3033, "step": 8947 }, { "epoch": 0.95, "learning_rate": 5.431168905558559e-08, "loss": 1.2307, "step": 8948 }, { "epoch": 0.95, "learning_rate": 5.405805847875256e-08, "loss": 1.3424, "step": 8949 }, { "epoch": 0.95, "learning_rate": 5.3805018285638664e-08, "loss": 1.1494, "step": 8950 }, { "epoch": 0.95, "learning_rate": 5.3552568506448633e-08, "loss": 1.1632, "step": 8951 }, { "epoch": 0.95, "learning_rate": 5.330070917131724e-08, "loss": 1.3724, "step": 8952 }, { "epoch": 0.95, "learning_rate": 5.304944031030879e-08, "loss": 1.4742, "step": 8953 }, { "epoch": 0.96, "learning_rate": 5.279876195341593e-08, "loss": 1.3089, "step": 8954 }, { "epoch": 0.96, "learning_rate": 5.2548674130561974e-08, "loss": 1.3174, "step": 8955 }, { "epoch": 0.96, "learning_rate": 5.229917687159802e-08, "loss": 1.3941, "step": 8956 }, { "epoch": 0.96, "learning_rate": 5.205027020630693e-08, "loss": 1.2458, "step": 8957 }, { "epoch": 0.96, "learning_rate": 5.1801954164399925e-08, "loss": 1.3483, "step": 8958 }, { "epoch": 0.96, "learning_rate": 5.155422877551774e-08, "loss": 1.6767, "step": 8959 }, { "epoch": 0.96, "learning_rate": 5.130709406923007e-08, "loss": 1.3704, "step": 8960 }, { "epoch": 0.96, "learning_rate": 5.106055007503774e-08, "loss": 1.4207, "step": 8961 }, { "epoch": 0.96, "learning_rate": 5.081459682236889e-08, "loss": 1.2073, "step": 8962 }, { "epoch": 0.96, "learning_rate": 5.056923434058336e-08, "loss": 1.4761, "step": 8963 }, { "epoch": 0.96, "learning_rate": 5.0324462658969395e-08, "loss": 1.3402, "step": 8964 }, { "epoch": 0.96, "learning_rate": 5.008028180674418e-08, "loss": 1.5345, "step": 8965 }, { "epoch": 0.96, "learning_rate": 4.983669181305495e-08, "loss": 1.4049, "step": 8966 }, { "epoch": 0.96, "learning_rate": 4.959369270697789e-08, "loss": 1.1831, "step": 8967 }, { "epoch": 0.96, "learning_rate": 4.935128451752091e-08, "loss": 1.456, "step": 8968 }, { "epoch": 0.96, "learning_rate": 4.910946727361754e-08, "loss": 1.7962, "step": 8969 }, { "epoch": 0.96, "learning_rate": 4.886824100413412e-08, "loss": 1.257, "step": 8970 }, { "epoch": 0.96, "learning_rate": 4.86276057378654e-08, "loss": 1.2452, "step": 8971 }, { "epoch": 0.96, "learning_rate": 4.838756150353396e-08, "loss": 1.6471, "step": 8972 }, { "epoch": 0.96, "learning_rate": 4.814810832979411e-08, "loss": 1.3911, "step": 8973 }, { "epoch": 0.96, "learning_rate": 4.790924624522908e-08, "loss": 1.4597, "step": 8974 }, { "epoch": 0.96, "learning_rate": 4.767097527835052e-08, "loss": 1.2837, "step": 8975 }, { "epoch": 0.96, "learning_rate": 4.743329545760122e-08, "loss": 1.2697, "step": 8976 }, { "epoch": 0.96, "learning_rate": 4.719620681135073e-08, "loss": 1.7502, "step": 8977 }, { "epoch": 0.96, "learning_rate": 4.695970936790084e-08, "loss": 1.5817, "step": 8978 }, { "epoch": 0.96, "learning_rate": 4.67238031554812e-08, "loss": 1.3404, "step": 8979 }, { "epoch": 0.96, "learning_rate": 4.648848820225205e-08, "loss": 1.1061, "step": 8980 }, { "epoch": 0.96, "learning_rate": 4.625376453630148e-08, "loss": 1.1681, "step": 8981 }, { "epoch": 0.96, "learning_rate": 4.6019632185647645e-08, "loss": 1.2175, "step": 8982 }, { "epoch": 0.96, "learning_rate": 4.578609117823873e-08, "loss": 1.235, "step": 8983 }, { "epoch": 0.96, "learning_rate": 4.555314154195245e-08, "loss": 1.3747, "step": 8984 }, { "epoch": 0.96, "learning_rate": 4.532078330459433e-08, "loss": 1.1182, "step": 8985 }, { "epoch": 0.96, "learning_rate": 4.508901649390107e-08, "loss": 1.5399, "step": 8986 }, { "epoch": 0.96, "learning_rate": 4.4857841137537216e-08, "loss": 1.5525, "step": 8987 }, { "epoch": 0.96, "learning_rate": 4.4627257263098465e-08, "loss": 1.3705, "step": 8988 }, { "epoch": 0.96, "learning_rate": 4.439726489810836e-08, "loss": 1.3619, "step": 8989 }, { "epoch": 0.96, "learning_rate": 4.4167864070021605e-08, "loss": 1.5745, "step": 8990 }, { "epoch": 0.96, "learning_rate": 4.393905480621907e-08, "loss": 1.4565, "step": 8991 }, { "epoch": 0.96, "learning_rate": 4.371083713401503e-08, "loss": 1.3558, "step": 8992 }, { "epoch": 0.96, "learning_rate": 4.348321108065046e-08, "loss": 1.2617, "step": 8993 }, { "epoch": 0.96, "learning_rate": 4.3256176673295846e-08, "loss": 1.4294, "step": 8994 }, { "epoch": 0.96, "learning_rate": 4.302973393905341e-08, "loss": 1.1243, "step": 8995 }, { "epoch": 0.96, "learning_rate": 4.280388290495097e-08, "loss": 1.2405, "step": 8996 }, { "epoch": 0.96, "learning_rate": 4.2578623597949174e-08, "loss": 1.6066, "step": 8997 }, { "epoch": 0.96, "learning_rate": 4.2353956044935974e-08, "loss": 1.4826, "step": 8998 }, { "epoch": 0.96, "learning_rate": 4.212988027272991e-08, "loss": 1.5446, "step": 8999 }, { "epoch": 0.96, "learning_rate": 4.1906396308077356e-08, "loss": 1.4578, "step": 9000 }, { "epoch": 0.96, "learning_rate": 4.1683504177655875e-08, "loss": 1.4146, "step": 9001 }, { "epoch": 0.96, "learning_rate": 4.146120390807085e-08, "loss": 1.3902, "step": 9002 }, { "epoch": 0.96, "learning_rate": 4.123949552585826e-08, "loss": 1.478, "step": 9003 }, { "epoch": 0.96, "learning_rate": 4.1018379057482517e-08, "loss": 1.1114, "step": 9004 }, { "epoch": 0.96, "learning_rate": 4.079785452933749e-08, "loss": 1.6889, "step": 9005 }, { "epoch": 0.96, "learning_rate": 4.0577921967747126e-08, "loss": 1.2363, "step": 9006 }, { "epoch": 0.96, "learning_rate": 4.035858139896376e-08, "loss": 1.3481, "step": 9007 }, { "epoch": 0.96, "learning_rate": 4.0139832849169224e-08, "loss": 1.5199, "step": 9008 }, { "epoch": 0.96, "learning_rate": 3.9921676344475966e-08, "loss": 1.4121, "step": 9009 }, { "epoch": 0.96, "learning_rate": 3.970411191092372e-08, "loss": 1.3683, "step": 9010 }, { "epoch": 0.96, "learning_rate": 3.94871395744828e-08, "loss": 1.563, "step": 9011 }, { "epoch": 0.96, "learning_rate": 3.927075936105307e-08, "loss": 1.2265, "step": 9012 }, { "epoch": 0.96, "learning_rate": 3.905497129646274e-08, "loss": 1.3121, "step": 9013 }, { "epoch": 0.96, "learning_rate": 3.88397754064701e-08, "loss": 1.4092, "step": 9014 }, { "epoch": 0.96, "learning_rate": 3.8625171716762385e-08, "loss": 1.2465, "step": 9015 }, { "epoch": 0.96, "learning_rate": 3.841116025295633e-08, "loss": 1.3894, "step": 9016 }, { "epoch": 0.96, "learning_rate": 3.819774104059759e-08, "loss": 1.4705, "step": 9017 }, { "epoch": 0.96, "learning_rate": 3.7984914105162474e-08, "loss": 1.44, "step": 9018 }, { "epoch": 0.96, "learning_rate": 3.7772679472053986e-08, "loss": 1.2658, "step": 9019 }, { "epoch": 0.96, "learning_rate": 3.756103716660742e-08, "loss": 1.5551, "step": 9020 }, { "epoch": 0.96, "learning_rate": 3.7349987214084784e-08, "loss": 1.3327, "step": 9021 }, { "epoch": 0.96, "learning_rate": 3.713952963967982e-08, "loss": 1.568, "step": 9022 }, { "epoch": 0.96, "learning_rate": 3.692966446851298e-08, "loss": 1.4397, "step": 9023 }, { "epoch": 0.96, "learning_rate": 3.672039172563646e-08, "loss": 1.441, "step": 9024 }, { "epoch": 0.96, "learning_rate": 3.651171143602972e-08, "loss": 1.5021, "step": 9025 }, { "epoch": 0.96, "learning_rate": 3.630362362460227e-08, "loss": 1.4881, "step": 9026 }, { "epoch": 0.96, "learning_rate": 3.609612831619369e-08, "loss": 1.464, "step": 9027 }, { "epoch": 0.96, "learning_rate": 3.5889225535571945e-08, "loss": 1.628, "step": 9028 }, { "epoch": 0.96, "learning_rate": 3.5682915307434505e-08, "loss": 1.4555, "step": 9029 }, { "epoch": 0.96, "learning_rate": 3.547719765640778e-08, "loss": 1.3098, "step": 9030 }, { "epoch": 0.96, "learning_rate": 3.527207260704713e-08, "loss": 1.569, "step": 9031 }, { "epoch": 0.96, "learning_rate": 3.5067540183839064e-08, "loss": 1.3183, "step": 9032 }, { "epoch": 0.96, "learning_rate": 3.4863600411197404e-08, "loss": 1.2142, "step": 9033 }, { "epoch": 0.96, "learning_rate": 3.4660253313466564e-08, "loss": 1.1748, "step": 9034 }, { "epoch": 0.96, "learning_rate": 3.445749891491823e-08, "loss": 1.3268, "step": 9035 }, { "epoch": 0.96, "learning_rate": 3.425533723975527e-08, "loss": 1.3599, "step": 9036 }, { "epoch": 0.96, "learning_rate": 3.405376831211005e-08, "loss": 1.3342, "step": 9037 }, { "epoch": 0.96, "learning_rate": 3.3852792156041646e-08, "loss": 1.3298, "step": 9038 }, { "epoch": 0.96, "learning_rate": 3.365240879554144e-08, "loss": 1.2857, "step": 9039 }, { "epoch": 0.96, "learning_rate": 3.345261825452806e-08, "loss": 1.181, "step": 9040 }, { "epoch": 0.96, "learning_rate": 3.325342055684966e-08, "loss": 1.5169, "step": 9041 }, { "epoch": 0.96, "learning_rate": 3.3054815726285e-08, "loss": 1.4162, "step": 9042 }, { "epoch": 0.96, "learning_rate": 3.2856803786540105e-08, "loss": 1.3216, "step": 9043 }, { "epoch": 0.96, "learning_rate": 3.265938476125108e-08, "loss": 1.5029, "step": 9044 }, { "epoch": 0.96, "learning_rate": 3.2462558673983516e-08, "loss": 1.5941, "step": 9045 }, { "epoch": 0.96, "learning_rate": 3.22663255482325e-08, "loss": 1.0404, "step": 9046 }, { "epoch": 0.97, "learning_rate": 3.2070685407420974e-08, "loss": 1.4798, "step": 9047 }, { "epoch": 0.97, "learning_rate": 3.1875638274902476e-08, "loss": 1.3082, "step": 9048 }, { "epoch": 0.97, "learning_rate": 3.168118417395949e-08, "loss": 1.3324, "step": 9049 }, { "epoch": 0.97, "learning_rate": 3.148732312780234e-08, "loss": 1.5639, "step": 9050 }, { "epoch": 0.97, "learning_rate": 3.129405515957307e-08, "loss": 1.3352, "step": 9051 }, { "epoch": 0.97, "learning_rate": 3.1101380292341e-08, "loss": 1.3098, "step": 9052 }, { "epoch": 0.97, "learning_rate": 3.090929854910552e-08, "loss": 1.3838, "step": 9053 }, { "epoch": 0.97, "learning_rate": 3.071780995279439e-08, "loss": 1.2374, "step": 9054 }, { "epoch": 0.97, "learning_rate": 3.0526914526264884e-08, "loss": 1.3668, "step": 9055 }, { "epoch": 0.97, "learning_rate": 3.033661229230434e-08, "loss": 1.3363, "step": 9056 }, { "epoch": 0.97, "learning_rate": 3.014690327362846e-08, "loss": 1.786, "step": 9057 }, { "epoch": 0.97, "learning_rate": 2.9957787492882474e-08, "loss": 1.1716, "step": 9058 }, { "epoch": 0.97, "learning_rate": 2.9769264972639435e-08, "loss": 1.5002, "step": 9059 }, { "epoch": 0.97, "learning_rate": 2.9581335735404672e-08, "loss": 1.441, "step": 9060 }, { "epoch": 0.97, "learning_rate": 2.939399980360913e-08, "loss": 1.2403, "step": 9061 }, { "epoch": 0.97, "learning_rate": 2.9207257199616033e-08, "loss": 1.3695, "step": 9062 }, { "epoch": 0.97, "learning_rate": 2.9021107945714777e-08, "loss": 1.5485, "step": 9063 }, { "epoch": 0.97, "learning_rate": 2.883555206412647e-08, "loss": 1.5614, "step": 9064 }, { "epoch": 0.97, "learning_rate": 2.8650589577000066e-08, "loss": 1.3298, "step": 9065 }, { "epoch": 0.97, "learning_rate": 2.8466220506414565e-08, "loss": 1.4935, "step": 9066 }, { "epoch": 0.97, "learning_rate": 2.828244487437737e-08, "loss": 1.2738, "step": 9067 }, { "epoch": 0.97, "learning_rate": 2.8099262702824814e-08, "loss": 1.0927, "step": 9068 }, { "epoch": 0.97, "learning_rate": 2.79166740136233e-08, "loss": 1.2966, "step": 9069 }, { "epoch": 0.97, "learning_rate": 2.773467882856762e-08, "loss": 1.6678, "step": 9070 }, { "epoch": 0.97, "learning_rate": 2.755327716938261e-08, "loss": 1.3991, "step": 9071 }, { "epoch": 0.97, "learning_rate": 2.7372469057721506e-08, "loss": 1.4783, "step": 9072 }, { "epoch": 0.97, "learning_rate": 2.7192254515166495e-08, "loss": 1.319, "step": 9073 }, { "epoch": 0.97, "learning_rate": 2.7012633563229808e-08, "loss": 1.4025, "step": 9074 }, { "epoch": 0.97, "learning_rate": 2.6833606223351515e-08, "loss": 1.5457, "step": 9075 }, { "epoch": 0.97, "learning_rate": 2.6655172516902304e-08, "loss": 1.1294, "step": 9076 }, { "epoch": 0.97, "learning_rate": 2.647733246518125e-08, "loss": 1.2827, "step": 9077 }, { "epoch": 0.97, "learning_rate": 2.6300086089416366e-08, "loss": 1.0901, "step": 9078 }, { "epoch": 0.97, "learning_rate": 2.612343341076573e-08, "loss": 1.5369, "step": 9079 }, { "epoch": 0.97, "learning_rate": 2.5947374450315254e-08, "loss": 1.4548, "step": 9080 }, { "epoch": 0.97, "learning_rate": 2.577190922908035e-08, "loss": 1.5349, "step": 9081 }, { "epoch": 0.97, "learning_rate": 2.559703776800704e-08, "loss": 1.3405, "step": 9082 }, { "epoch": 0.97, "learning_rate": 2.542276008796807e-08, "loss": 1.4427, "step": 9083 }, { "epoch": 0.97, "learning_rate": 2.5249076209767353e-08, "loss": 1.3905, "step": 9084 }, { "epoch": 0.97, "learning_rate": 2.507598615413609e-08, "loss": 1.4911, "step": 9085 }, { "epoch": 0.97, "learning_rate": 2.490348994173608e-08, "loss": 1.2979, "step": 9086 }, { "epoch": 0.97, "learning_rate": 2.473158759315808e-08, "loss": 1.4015, "step": 9087 }, { "epoch": 0.97, "learning_rate": 2.4560279128921226e-08, "loss": 1.219, "step": 9088 }, { "epoch": 0.97, "learning_rate": 2.4389564569474166e-08, "loss": 1.2596, "step": 9089 }, { "epoch": 0.97, "learning_rate": 2.421944393519504e-08, "loss": 1.5536, "step": 9090 }, { "epoch": 0.97, "learning_rate": 2.404991724638983e-08, "loss": 1.0309, "step": 9091 }, { "epoch": 0.97, "learning_rate": 2.3880984523295124e-08, "loss": 1.1716, "step": 9092 }, { "epoch": 0.97, "learning_rate": 2.3712645786075905e-08, "loss": 1.2065, "step": 9093 }, { "epoch": 0.97, "learning_rate": 2.35449010548261e-08, "loss": 1.5151, "step": 9094 }, { "epoch": 0.97, "learning_rate": 2.337775034956913e-08, "loss": 1.5415, "step": 9095 }, { "epoch": 0.97, "learning_rate": 2.3211193690257373e-08, "loss": 1.3564, "step": 9096 }, { "epoch": 0.97, "learning_rate": 2.304523109677159e-08, "loss": 1.3841, "step": 9097 }, { "epoch": 0.97, "learning_rate": 2.2879862588923717e-08, "loss": 1.1682, "step": 9098 }, { "epoch": 0.97, "learning_rate": 2.271508818645185e-08, "loss": 1.2699, "step": 9099 }, { "epoch": 0.97, "learning_rate": 2.2550907909025255e-08, "loss": 1.4234, "step": 9100 }, { "epoch": 0.97, "learning_rate": 2.2387321776242143e-08, "loss": 1.2461, "step": 9101 }, { "epoch": 0.97, "learning_rate": 2.222432980762912e-08, "loss": 1.3279, "step": 9102 }, { "epoch": 0.97, "learning_rate": 2.2061932022641175e-08, "loss": 1.5812, "step": 9103 }, { "epoch": 0.97, "learning_rate": 2.1900128440663913e-08, "loss": 1.4987, "step": 9104 }, { "epoch": 0.97, "learning_rate": 2.1738919081012446e-08, "loss": 1.3947, "step": 9105 }, { "epoch": 0.97, "learning_rate": 2.157830396292804e-08, "loss": 1.6971, "step": 9106 }, { "epoch": 0.97, "learning_rate": 2.1418283105584246e-08, "loss": 1.1771, "step": 9107 }, { "epoch": 0.97, "learning_rate": 2.1258856528081906e-08, "loss": 1.3039, "step": 9108 }, { "epoch": 0.97, "learning_rate": 2.1100024249451347e-08, "loss": 1.4902, "step": 9109 }, { "epoch": 0.97, "learning_rate": 2.094178628865129e-08, "loss": 1.6844, "step": 9110 }, { "epoch": 0.97, "learning_rate": 2.0784142664571626e-08, "loss": 1.5341, "step": 9111 }, { "epoch": 0.97, "learning_rate": 2.0627093396028418e-08, "loss": 1.4435, "step": 9112 }, { "epoch": 0.97, "learning_rate": 2.047063850176889e-08, "loss": 1.1045, "step": 9113 }, { "epoch": 0.97, "learning_rate": 2.031477800046866e-08, "loss": 1.2604, "step": 9114 }, { "epoch": 0.97, "learning_rate": 2.015951191073229e-08, "loss": 1.366, "step": 9115 }, { "epoch": 0.97, "learning_rate": 2.0004840251093284e-08, "loss": 1.5111, "step": 9116 }, { "epoch": 0.97, "learning_rate": 1.9850763040014654e-08, "loss": 1.4083, "step": 9117 }, { "epoch": 0.97, "learning_rate": 1.9697280295888356e-08, "loss": 1.4531, "step": 9118 }, { "epoch": 0.97, "learning_rate": 1.9544392037034175e-08, "loss": 1.4568, "step": 9119 }, { "epoch": 0.97, "learning_rate": 1.939209828170363e-08, "loss": 1.2309, "step": 9120 }, { "epoch": 0.97, "learning_rate": 1.9240399048074398e-08, "loss": 1.3273, "step": 9121 }, { "epoch": 0.97, "learning_rate": 1.9089294354254216e-08, "loss": 1.1792, "step": 9122 }, { "epoch": 0.97, "learning_rate": 1.8938784218281435e-08, "loss": 1.3714, "step": 9123 }, { "epoch": 0.97, "learning_rate": 1.8788868658120573e-08, "loss": 1.3915, "step": 9124 }, { "epoch": 0.97, "learning_rate": 1.863954769166676e-08, "loss": 1.4016, "step": 9125 }, { "epoch": 0.97, "learning_rate": 1.849082133674518e-08, "loss": 1.51, "step": 9126 }, { "epoch": 0.97, "learning_rate": 1.83426896111083e-08, "loss": 1.3837, "step": 9127 }, { "epoch": 0.97, "learning_rate": 1.8195152532437533e-08, "loss": 1.6338, "step": 9128 }, { "epoch": 0.97, "learning_rate": 1.80482101183449e-08, "loss": 1.0652, "step": 9129 }, { "epoch": 0.97, "learning_rate": 1.790186238637026e-08, "loss": 1.5472, "step": 9130 }, { "epoch": 0.97, "learning_rate": 1.775610935398242e-08, "loss": 1.4242, "step": 9131 }, { "epoch": 0.97, "learning_rate": 1.761095103858024e-08, "loss": 1.5787, "step": 9132 }, { "epoch": 0.97, "learning_rate": 1.7466387457489853e-08, "loss": 1.2489, "step": 9133 }, { "epoch": 0.97, "learning_rate": 1.7322418627968574e-08, "loss": 1.3344, "step": 9134 }, { "epoch": 0.97, "learning_rate": 1.717904456720043e-08, "loss": 1.6767, "step": 9135 }, { "epoch": 0.97, "learning_rate": 1.7036265292300624e-08, "loss": 1.2491, "step": 9136 }, { "epoch": 0.97, "learning_rate": 1.689408082031163e-08, "loss": 1.4602, "step": 9137 }, { "epoch": 0.97, "learning_rate": 1.675249116820543e-08, "loss": 1.4165, "step": 9138 }, { "epoch": 0.97, "learning_rate": 1.661149635288406e-08, "loss": 1.3123, "step": 9139 }, { "epoch": 0.97, "learning_rate": 1.6471096391177388e-08, "loss": 1.572, "step": 9140 }, { "epoch": 0.98, "learning_rate": 1.6331291299844233e-08, "loss": 1.1643, "step": 9141 }, { "epoch": 0.98, "learning_rate": 1.6192081095572908e-08, "loss": 1.4434, "step": 9142 }, { "epoch": 0.98, "learning_rate": 1.6053465794980682e-08, "loss": 1.4378, "step": 9143 }, { "epoch": 0.98, "learning_rate": 1.5915445414613208e-08, "loss": 1.5724, "step": 9144 }, { "epoch": 0.98, "learning_rate": 1.5778019970946744e-08, "loss": 1.4271, "step": 9145 }, { "epoch": 0.98, "learning_rate": 1.564118948038429e-08, "loss": 1.3544, "step": 9146 }, { "epoch": 0.98, "learning_rate": 1.550495395925944e-08, "loss": 1.5458, "step": 9147 }, { "epoch": 0.98, "learning_rate": 1.5369313423833632e-08, "loss": 1.3556, "step": 9148 }, { "epoch": 0.98, "learning_rate": 1.5234267890298915e-08, "loss": 1.5214, "step": 9149 }, { "epoch": 0.98, "learning_rate": 1.5099817374774615e-08, "loss": 1.524, "step": 9150 }, { "epoch": 0.98, "learning_rate": 1.496596189331012e-08, "loss": 1.3734, "step": 9151 }, { "epoch": 0.98, "learning_rate": 1.4832701461883203e-08, "loss": 1.022, "step": 9152 }, { "epoch": 0.98, "learning_rate": 1.4700036096400028e-08, "loss": 1.2113, "step": 9153 }, { "epoch": 0.98, "learning_rate": 1.4567965812697926e-08, "loss": 1.2123, "step": 9154 }, { "epoch": 0.98, "learning_rate": 1.4436490626540955e-08, "loss": 1.5361, "step": 9155 }, { "epoch": 0.98, "learning_rate": 1.4305610553623228e-08, "loss": 1.3395, "step": 9156 }, { "epoch": 0.98, "learning_rate": 1.4175325609567248e-08, "loss": 1.3191, "step": 9157 }, { "epoch": 0.98, "learning_rate": 1.4045635809925018e-08, "loss": 1.5476, "step": 9158 }, { "epoch": 0.98, "learning_rate": 1.3916541170176934e-08, "loss": 1.4102, "step": 9159 }, { "epoch": 0.98, "learning_rate": 1.378804170573289e-08, "loss": 1.4684, "step": 9160 }, { "epoch": 0.98, "learning_rate": 1.3660137431932286e-08, "loss": 1.2931, "step": 9161 }, { "epoch": 0.98, "learning_rate": 1.3532828364041239e-08, "loss": 1.1399, "step": 9162 }, { "epoch": 0.98, "learning_rate": 1.340611451725704e-08, "loss": 1.495, "step": 9163 }, { "epoch": 0.98, "learning_rate": 1.3279995906705367e-08, "loss": 1.2699, "step": 9164 }, { "epoch": 0.98, "learning_rate": 1.3154472547440289e-08, "loss": 1.2542, "step": 9165 }, { "epoch": 0.98, "learning_rate": 1.302954445444593e-08, "loss": 1.0721, "step": 9166 }, { "epoch": 0.98, "learning_rate": 1.2905211642633698e-08, "loss": 1.3388, "step": 9167 }, { "epoch": 0.98, "learning_rate": 1.2781474126845051e-08, "loss": 1.3725, "step": 9168 }, { "epoch": 0.98, "learning_rate": 1.2658331921850398e-08, "loss": 1.4712, "step": 9169 }, { "epoch": 0.98, "learning_rate": 1.2535785042349091e-08, "loss": 1.4452, "step": 9170 }, { "epoch": 0.98, "learning_rate": 1.241383350296832e-08, "loss": 1.4788, "step": 9171 }, { "epoch": 0.98, "learning_rate": 1.2292477318266438e-08, "loss": 1.2855, "step": 9172 }, { "epoch": 0.98, "learning_rate": 1.2171716502728525e-08, "loss": 1.4458, "step": 9173 }, { "epoch": 0.98, "learning_rate": 1.2051551070769719e-08, "loss": 1.6515, "step": 9174 }, { "epoch": 0.98, "learning_rate": 1.1931981036733542e-08, "loss": 1.4626, "step": 9175 }, { "epoch": 0.98, "learning_rate": 1.181300641489358e-08, "loss": 1.3448, "step": 9176 }, { "epoch": 0.98, "learning_rate": 1.1694627219450694e-08, "loss": 1.5305, "step": 9177 }, { "epoch": 0.98, "learning_rate": 1.1576843464535248e-08, "loss": 1.2669, "step": 9178 }, { "epoch": 0.98, "learning_rate": 1.1459655164208216e-08, "loss": 1.2728, "step": 9179 }, { "epoch": 0.98, "learning_rate": 1.134306233245619e-08, "loss": 1.2926, "step": 9180 }, { "epoch": 0.98, "learning_rate": 1.1227064983198032e-08, "loss": 1.25, "step": 9181 }, { "epoch": 0.98, "learning_rate": 1.1111663130279337e-08, "loss": 1.3807, "step": 9182 }, { "epoch": 0.98, "learning_rate": 1.0996856787475197e-08, "loss": 1.4145, "step": 9183 }, { "epoch": 0.98, "learning_rate": 1.0882645968490203e-08, "loss": 1.7118, "step": 9184 }, { "epoch": 0.98, "learning_rate": 1.0769030686956783e-08, "loss": 1.3093, "step": 9185 }, { "epoch": 0.98, "learning_rate": 1.0656010956437979e-08, "loss": 1.1026, "step": 9186 }, { "epoch": 0.98, "learning_rate": 1.0543586790423e-08, "loss": 1.4211, "step": 9187 }, { "epoch": 0.98, "learning_rate": 1.043175820233333e-08, "loss": 1.4458, "step": 9188 }, { "epoch": 0.98, "learning_rate": 1.0320525205516629e-08, "loss": 1.2792, "step": 9189 }, { "epoch": 0.98, "learning_rate": 1.0209887813250053e-08, "loss": 1.3347, "step": 9190 }, { "epoch": 0.98, "learning_rate": 1.0099846038741368e-08, "loss": 1.3362, "step": 9191 }, { "epoch": 0.98, "learning_rate": 9.990399895125624e-09, "loss": 1.3448, "step": 9192 }, { "epoch": 0.98, "learning_rate": 9.881549395466262e-09, "loss": 1.4762, "step": 9193 }, { "epoch": 0.98, "learning_rate": 9.773294552757328e-09, "loss": 1.1556, "step": 9194 }, { "epoch": 0.98, "learning_rate": 9.665635379920157e-09, "loss": 1.3989, "step": 9195 }, { "epoch": 0.98, "learning_rate": 9.558571889806689e-09, "loss": 1.7087, "step": 9196 }, { "epoch": 0.98, "learning_rate": 9.452104095196146e-09, "loss": 1.2244, "step": 9197 }, { "epoch": 0.98, "learning_rate": 9.346232008797252e-09, "loss": 1.3593, "step": 9198 }, { "epoch": 0.98, "learning_rate": 9.240955643248229e-09, "loss": 1.3324, "step": 9199 }, { "epoch": 0.98, "learning_rate": 9.136275011115137e-09, "loss": 1.541, "step": 9200 }, { "epoch": 0.98, "learning_rate": 9.032190124893536e-09, "loss": 1.3846, "step": 9201 }, { "epoch": 0.98, "learning_rate": 8.928700997007933e-09, "loss": 1.2097, "step": 9202 }, { "epoch": 0.98, "learning_rate": 8.825807639811224e-09, "loss": 1.3683, "step": 9203 }, { "epoch": 0.98, "learning_rate": 8.723510065585806e-09, "loss": 1.6869, "step": 9204 }, { "epoch": 0.98, "learning_rate": 8.621808286542466e-09, "loss": 1.6746, "step": 9205 }, { "epoch": 0.98, "learning_rate": 8.520702314821493e-09, "loss": 1.1058, "step": 9206 }, { "epoch": 0.98, "learning_rate": 8.42019216249046e-09, "loss": 1.2632, "step": 9207 }, { "epoch": 0.98, "learning_rate": 8.320277841548652e-09, "loss": 1.2778, "step": 9208 }, { "epoch": 0.98, "learning_rate": 8.220959363921533e-09, "loss": 1.3726, "step": 9209 }, { "epoch": 0.98, "learning_rate": 8.122236741464618e-09, "loss": 1.4839, "step": 9210 }, { "epoch": 0.98, "learning_rate": 8.024109985961815e-09, "loss": 1.365, "step": 9211 }, { "epoch": 0.98, "learning_rate": 7.926579109127086e-09, "loss": 1.1875, "step": 9212 }, { "epoch": 0.98, "learning_rate": 7.82964412260223e-09, "loss": 1.5032, "step": 9213 }, { "epoch": 0.98, "learning_rate": 7.733305037957439e-09, "loss": 1.2329, "step": 9214 }, { "epoch": 0.98, "learning_rate": 7.637561866692955e-09, "loss": 1.3164, "step": 9215 }, { "epoch": 0.98, "learning_rate": 7.542414620237414e-09, "loss": 1.091, "step": 9216 }, { "epoch": 0.98, "learning_rate": 7.4478633099484e-09, "loss": 1.3171, "step": 9217 }, { "epoch": 0.98, "learning_rate": 7.353907947111882e-09, "loss": 1.2387, "step": 9218 }, { "epoch": 0.98, "learning_rate": 7.260548542943335e-09, "loss": 1.2196, "step": 9219 }, { "epoch": 0.98, "learning_rate": 7.167785108586622e-09, "loss": 1.4129, "step": 9220 }, { "epoch": 0.98, "learning_rate": 7.0756176551145525e-09, "loss": 1.4207, "step": 9221 }, { "epoch": 0.98, "learning_rate": 6.984046193528881e-09, "loss": 1.3305, "step": 9222 }, { "epoch": 0.98, "learning_rate": 6.89307073476142e-09, "loss": 1.4444, "step": 9223 }, { "epoch": 0.98, "learning_rate": 6.802691289669594e-09, "loss": 1.6958, "step": 9224 }, { "epoch": 0.98, "learning_rate": 6.712907869043661e-09, "loss": 1.3345, "step": 9225 }, { "epoch": 0.98, "learning_rate": 6.623720483600049e-09, "loss": 1.4048, "step": 9226 }, { "epoch": 0.98, "learning_rate": 6.535129143984131e-09, "loss": 1.3287, "step": 9227 }, { "epoch": 0.98, "learning_rate": 6.447133860771893e-09, "loss": 1.5483, "step": 9228 }, { "epoch": 0.98, "learning_rate": 6.359734644467153e-09, "loss": 1.3488, "step": 9229 }, { "epoch": 0.98, "learning_rate": 6.272931505501567e-09, "loss": 1.4718, "step": 9230 }, { "epoch": 0.98, "learning_rate": 6.186724454236847e-09, "loss": 1.6016, "step": 9231 }, { "epoch": 0.98, "learning_rate": 6.1011135009636515e-09, "loss": 1.4669, "step": 9232 }, { "epoch": 0.98, "learning_rate": 6.016098655901026e-09, "loss": 1.3972, "step": 9233 }, { "epoch": 0.98, "learning_rate": 5.9316799291969654e-09, "loss": 1.4506, "step": 9234 }, { "epoch": 0.99, "learning_rate": 5.8478573309284085e-09, "loss": 1.288, "step": 9235 }, { "epoch": 0.99, "learning_rate": 5.76463087110124e-09, "loss": 1.4912, "step": 9236 }, { "epoch": 0.99, "learning_rate": 5.682000559649181e-09, "loss": 1.2828, "step": 9237 }, { "epoch": 0.99, "learning_rate": 5.599966406436008e-09, "loss": 1.1786, "step": 9238 }, { "epoch": 0.99, "learning_rate": 5.518528421253888e-09, "loss": 1.3911, "step": 9239 }, { "epoch": 0.99, "learning_rate": 5.437686613823934e-09, "loss": 1.4447, "step": 9240 }, { "epoch": 0.99, "learning_rate": 5.357440993796204e-09, "loss": 1.3603, "step": 9241 }, { "epoch": 0.99, "learning_rate": 5.2777915707491465e-09, "loss": 1.3255, "step": 9242 }, { "epoch": 0.99, "learning_rate": 5.198738354190158e-09, "loss": 1.4405, "step": 9243 }, { "epoch": 0.99, "learning_rate": 5.120281353556134e-09, "loss": 1.3351, "step": 9244 }, { "epoch": 0.99, "learning_rate": 5.042420578211804e-09, "loss": 1.159, "step": 9245 }, { "epoch": 0.99, "learning_rate": 4.9651560374514015e-09, "loss": 1.5984, "step": 9246 }, { "epoch": 0.99, "learning_rate": 4.88848774049755e-09, "loss": 1.2434, "step": 9247 }, { "epoch": 0.99, "learning_rate": 4.812415696502371e-09, "loss": 1.1307, "step": 9248 }, { "epoch": 0.99, "learning_rate": 4.736939914545824e-09, "loss": 1.4957, "step": 9249 }, { "epoch": 0.99, "learning_rate": 4.66206040363737e-09, "loss": 1.604, "step": 9250 }, { "epoch": 0.99, "learning_rate": 4.587777172715413e-09, "loss": 1.5535, "step": 9251 }, { "epoch": 0.99, "learning_rate": 4.514090230647305e-09, "loss": 1.5767, "step": 9252 }, { "epoch": 0.99, "learning_rate": 4.440999586228234e-09, "loss": 1.6683, "step": 9253 }, { "epoch": 0.99, "learning_rate": 4.3685052481828864e-09, "loss": 1.2925, "step": 9254 }, { "epoch": 0.99, "learning_rate": 4.296607225164895e-09, "loss": 1.2004, "step": 9255 }, { "epoch": 0.99, "learning_rate": 4.225305525756285e-09, "loss": 1.4221, "step": 9256 }, { "epoch": 0.99, "learning_rate": 4.1546001584685806e-09, "loss": 1.2956, "step": 9257 }, { "epoch": 0.99, "learning_rate": 4.084491131741697e-09, "loss": 1.6233, "step": 9258 }, { "epoch": 0.99, "learning_rate": 4.0149784539439405e-09, "loss": 1.2444, "step": 9259 }, { "epoch": 0.99, "learning_rate": 3.946062133372563e-09, "loss": 1.7246, "step": 9260 }, { "epoch": 0.99, "learning_rate": 3.877742178254873e-09, "loss": 1.3868, "step": 9261 }, { "epoch": 0.99, "learning_rate": 3.810018596745457e-09, "loss": 1.4343, "step": 9262 }, { "epoch": 0.99, "learning_rate": 3.7428913969284055e-09, "loss": 1.3795, "step": 9263 }, { "epoch": 0.99, "learning_rate": 3.6763605868167516e-09, "loss": 1.6322, "step": 9264 }, { "epoch": 0.99, "learning_rate": 3.610426174351367e-09, "loss": 1.3314, "step": 9265 }, { "epoch": 0.99, "learning_rate": 3.5450881674031766e-09, "loss": 1.4759, "step": 9266 }, { "epoch": 0.99, "learning_rate": 3.4803465737714983e-09, "loss": 1.3322, "step": 9267 }, { "epoch": 0.99, "learning_rate": 3.4162014011845932e-09, "loss": 1.2061, "step": 9268 }, { "epoch": 0.99, "learning_rate": 3.3526526572985607e-09, "loss": 1.2848, "step": 9269 }, { "epoch": 0.99, "learning_rate": 3.289700349698999e-09, "loss": 1.5528, "step": 9270 }, { "epoch": 0.99, "learning_rate": 3.2273444859015625e-09, "loss": 1.2049, "step": 9271 }, { "epoch": 0.99, "learning_rate": 3.1655850733480766e-09, "loss": 1.3827, "step": 9272 }, { "epoch": 0.99, "learning_rate": 3.104422119411532e-09, "loss": 1.5593, "step": 9273 }, { "epoch": 0.99, "learning_rate": 3.043855631392756e-09, "loss": 1.419, "step": 9274 }, { "epoch": 0.99, "learning_rate": 2.98388561652041e-09, "loss": 1.3438, "step": 9275 }, { "epoch": 0.99, "learning_rate": 2.9245120819543226e-09, "loss": 1.4822, "step": 9276 }, { "epoch": 0.99, "learning_rate": 2.8657350347810473e-09, "loss": 1.4156, "step": 9277 }, { "epoch": 0.99, "learning_rate": 2.807554482016639e-09, "loss": 1.2423, "step": 9278 }, { "epoch": 0.99, "learning_rate": 2.749970430605542e-09, "loss": 1.1964, "step": 9279 }, { "epoch": 0.99, "learning_rate": 2.6929828874222573e-09, "loss": 1.5161, "step": 9280 }, { "epoch": 0.99, "learning_rate": 2.636591859269122e-09, "loss": 1.3364, "step": 9281 }, { "epoch": 0.99, "learning_rate": 2.5807973528768626e-09, "loss": 1.3855, "step": 9282 }, { "epoch": 0.99, "learning_rate": 2.5255993749057075e-09, "loss": 1.159, "step": 9283 }, { "epoch": 0.99, "learning_rate": 2.4709979319442743e-09, "loss": 1.4135, "step": 9284 }, { "epoch": 0.99, "learning_rate": 2.416993030511239e-09, "loss": 1.3965, "step": 9285 }, { "epoch": 0.99, "learning_rate": 2.3635846770514448e-09, "loss": 1.439, "step": 9286 }, { "epoch": 0.99, "learning_rate": 2.3107728779414584e-09, "loss": 1.333, "step": 9287 }, { "epoch": 0.99, "learning_rate": 2.258557639484571e-09, "loss": 1.4386, "step": 9288 }, { "epoch": 0.99, "learning_rate": 2.2069389679135746e-09, "loss": 1.4067, "step": 9289 }, { "epoch": 0.99, "learning_rate": 2.1559168693902066e-09, "loss": 1.386, "step": 9290 }, { "epoch": 0.99, "learning_rate": 2.1054913500051512e-09, "loss": 1.3262, "step": 9291 }, { "epoch": 0.99, "learning_rate": 2.0556624157769265e-09, "loss": 1.2899, "step": 9292 }, { "epoch": 0.99, "learning_rate": 2.0064300726541084e-09, "loss": 1.4539, "step": 9293 }, { "epoch": 0.99, "learning_rate": 1.957794326513107e-09, "loss": 1.5578, "step": 9294 }, { "epoch": 0.99, "learning_rate": 1.9097551831592785e-09, "loss": 1.1297, "step": 9295 }, { "epoch": 0.99, "learning_rate": 1.862312648326925e-09, "loss": 1.376, "step": 9296 }, { "epoch": 0.99, "learning_rate": 1.8154667276798488e-09, "loss": 1.6084, "step": 9297 }, { "epoch": 0.99, "learning_rate": 1.7692174268091334e-09, "loss": 1.3133, "step": 9298 }, { "epoch": 0.99, "learning_rate": 1.7235647512353627e-09, "loss": 1.5864, "step": 9299 }, { "epoch": 0.99, "learning_rate": 1.6785087064086213e-09, "loss": 1.2903, "step": 9300 }, { "epoch": 0.99, "learning_rate": 1.63404929770683e-09, "loss": 1.473, "step": 9301 }, { "epoch": 0.99, "learning_rate": 1.5901865304368547e-09, "loss": 1.3253, "step": 9302 }, { "epoch": 0.99, "learning_rate": 1.546920409834507e-09, "loss": 1.2744, "step": 9303 }, { "epoch": 0.99, "learning_rate": 1.5042509410645445e-09, "loss": 1.3534, "step": 9304 }, { "epoch": 0.99, "learning_rate": 1.4621781292201155e-09, "loss": 1.2551, "step": 9305 }, { "epoch": 0.99, "learning_rate": 1.4207019793238686e-09, "loss": 1.3667, "step": 9306 }, { "epoch": 0.99, "learning_rate": 1.3798224963257335e-09, "loss": 1.3568, "step": 9307 }, { "epoch": 0.99, "learning_rate": 1.3395396851056951e-09, "loss": 1.4391, "step": 9308 }, { "epoch": 0.99, "learning_rate": 1.299853550472685e-09, "loss": 1.511, "step": 9309 }, { "epoch": 0.99, "learning_rate": 1.2607640971640245e-09, "loss": 1.234, "step": 9310 }, { "epoch": 0.99, "learning_rate": 1.2222713298448706e-09, "loss": 1.2836, "step": 9311 }, { "epoch": 0.99, "learning_rate": 1.1843752531104368e-09, "loss": 1.3096, "step": 9312 }, { "epoch": 0.99, "learning_rate": 1.1470758714843267e-09, "loss": 1.2975, "step": 9313 }, { "epoch": 0.99, "learning_rate": 1.1103731894190894e-09, "loss": 1.3891, "step": 9314 }, { "epoch": 0.99, "learning_rate": 1.0742672112951103e-09, "loss": 1.181, "step": 9315 }, { "epoch": 0.99, "learning_rate": 1.038757941423385e-09, "loss": 1.2315, "step": 9316 }, { "epoch": 0.99, "learning_rate": 1.0038453840416352e-09, "loss": 1.1365, "step": 9317 }, { "epoch": 0.99, "learning_rate": 9.695295433170826e-10, "loss": 1.5099, "step": 9318 }, { "epoch": 0.99, "learning_rate": 9.358104233470055e-10, "loss": 1.346, "step": 9319 }, { "epoch": 0.99, "learning_rate": 9.026880281554073e-10, "loss": 1.3913, "step": 9320 }, { "epoch": 0.99, "learning_rate": 8.701623616963472e-10, "loss": 1.3676, "step": 9321 }, { "epoch": 0.99, "learning_rate": 8.382334278528304e-10, "loss": 1.2588, "step": 9322 }, { "epoch": 0.99, "learning_rate": 8.069012304351421e-10, "loss": 1.564, "step": 9323 }, { "epoch": 0.99, "learning_rate": 7.761657731836236e-10, "loss": 1.4469, "step": 9324 }, { "epoch": 0.99, "learning_rate": 7.460270597675622e-10, "loss": 1.3887, "step": 9325 }, { "epoch": 0.99, "learning_rate": 7.164850937840806e-10, "loss": 1.3006, "step": 9326 }, { "epoch": 0.99, "learning_rate": 6.87539878759802e-10, "loss": 1.031, "step": 9327 }, { "epoch": 0.99, "learning_rate": 6.591914181497405e-10, "loss": 1.4455, "step": 9328 }, { "epoch": 1.0, "learning_rate": 6.314397153378559e-10, "loss": 1.3495, "step": 9329 }, { "epoch": 1.0, "learning_rate": 6.042847736364987e-10, "loss": 1.3944, "step": 9330 }, { "epoch": 1.0, "learning_rate": 5.7772659628752e-10, "loss": 1.5773, "step": 9331 }, { "epoch": 1.0, "learning_rate": 5.517651864606066e-10, "loss": 1.4717, "step": 9332 }, { "epoch": 1.0, "learning_rate": 5.264005472549461e-10, "loss": 1.0868, "step": 9333 }, { "epoch": 1.0, "learning_rate": 5.016326816981165e-10, "loss": 1.6405, "step": 9334 }, { "epoch": 1.0, "learning_rate": 4.774615927471971e-10, "loss": 1.3413, "step": 9335 }, { "epoch": 1.0, "learning_rate": 4.538872832865471e-10, "loss": 1.3493, "step": 9336 }, { "epoch": 1.0, "learning_rate": 4.3090975613058194e-10, "loss": 1.3358, "step": 9337 }, { "epoch": 1.0, "learning_rate": 4.0852901402210764e-10, "loss": 1.2114, "step": 9338 }, { "epoch": 1.0, "learning_rate": 3.867450596328759e-10, "loss": 1.4271, "step": 9339 }, { "epoch": 1.0, "learning_rate": 3.655578955624739e-10, "loss": 1.3626, "step": 9340 }, { "epoch": 1.0, "learning_rate": 3.449675243405448e-10, "loss": 1.2559, "step": 9341 }, { "epoch": 1.0, "learning_rate": 3.2497394842512244e-10, "loss": 1.2957, "step": 9342 }, { "epoch": 1.0, "learning_rate": 3.05577170202076e-10, "loss": 1.4928, "step": 9343 }, { "epoch": 1.0, "learning_rate": 2.8677719198677566e-10, "loss": 1.3999, "step": 9344 }, { "epoch": 1.0, "learning_rate": 2.685740160240924e-10, "loss": 1.5018, "step": 9345 }, { "epoch": 1.0, "learning_rate": 2.509676444861775e-10, "loss": 1.5719, "step": 9346 }, { "epoch": 1.0, "learning_rate": 2.339580794752383e-10, "loss": 1.297, "step": 9347 }, { "epoch": 1.0, "learning_rate": 2.1754532302076247e-10, "loss": 1.3284, "step": 9348 }, { "epoch": 1.0, "learning_rate": 2.017293770828488e-10, "loss": 1.3543, "step": 9349 }, { "epoch": 1.0, "learning_rate": 1.8651024354887637e-10, "loss": 1.7276, "step": 9350 }, { "epoch": 1.0, "learning_rate": 1.718879242357252e-10, "loss": 1.3625, "step": 9351 }, { "epoch": 1.0, "learning_rate": 1.578624208886659e-10, "loss": 1.6177, "step": 9352 }, { "epoch": 1.0, "learning_rate": 1.4443373518246984e-10, "loss": 1.5216, "step": 9353 }, { "epoch": 1.0, "learning_rate": 1.316018687191889e-10, "loss": 1.172, "step": 9354 }, { "epoch": 1.0, "learning_rate": 1.1936682303037573e-10, "loss": 1.1624, "step": 9355 }, { "epoch": 1.0, "learning_rate": 1.0772859957763892e-10, "loss": 1.1216, "step": 9356 }, { "epoch": 1.0, "learning_rate": 9.668719974986751e-11, "loss": 1.3731, "step": 9357 }, { "epoch": 1.0, "learning_rate": 8.62426248643411e-11, "loss": 1.4921, "step": 9358 }, { "epoch": 1.0, "learning_rate": 7.639487616839525e-11, "loss": 1.2956, "step": 9359 }, { "epoch": 1.0, "learning_rate": 6.714395483720105e-11, "loss": 1.3485, "step": 9360 }, { "epoch": 1.0, "learning_rate": 5.848986197487528e-11, "loss": 1.4707, "step": 9361 }, { "epoch": 1.0, "learning_rate": 5.043259861559069e-11, "loss": 1.599, "step": 9362 }, { "epoch": 1.0, "learning_rate": 4.297216571969021e-11, "loss": 1.3927, "step": 9363 }, { "epoch": 1.0, "learning_rate": 3.6108564178127804e-11, "loss": 1.4516, "step": 9364 }, { "epoch": 1.0, "learning_rate": 2.9841794810803185e-11, "loss": 1.4351, "step": 9365 }, { "epoch": 1.0, "learning_rate": 2.417185836545155e-11, "loss": 1.4206, "step": 9366 }, { "epoch": 1.0, "learning_rate": 1.909875551819873e-11, "loss": 1.5888, "step": 9367 }, { "epoch": 1.0, "learning_rate": 1.4622486875226494e-11, "loss": 1.124, "step": 9368 }, { "epoch": 1.0, "learning_rate": 1.0743052971107225e-11, "loss": 1.4875, "step": 9369 }, { "epoch": 1.0, "learning_rate": 7.46045426880393e-12, "loss": 1.5424, "step": 9370 }, { "epoch": 1.0, "learning_rate": 4.774691159115108e-12, "loss": 1.3367, "step": 9371 }, { "epoch": 1.0, "learning_rate": 2.6857639640054387e-12, "loss": 1.4345, "step": 9372 }, { "epoch": 1.0, "learning_rate": 1.19367293216488e-12, "loss": 1.3414, "step": 9373 }, { "epoch": 1.0, "learning_rate": 2.984182417842263e-13, "loss": 1.6349, "step": 9374 }, { "epoch": 1.0, "learning_rate": 0.0, "loss": 1.1603, "step": 9375 }, { "epoch": 1.0, "step": 9375, "total_flos": 8.015698364538552e+17, "train_loss": 1.4598437309837342, "train_runtime": 32220.9094, "train_samples_per_second": 9.311, "train_steps_per_second": 0.291 } ], "logging_steps": 1.0, "max_steps": 9375, "num_input_tokens_seen": 0, "num_train_epochs": 1, "save_steps": 1000, "total_flos": 8.015698364538552e+17, "train_batch_size": 4, "trial_name": null, "trial_params": null }