diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,7036 @@
+{
+  "best_metric": 0.26086243987083435,
+  "best_model_checkpoint": "./results/checkpoint-5500",
+  "epoch": 2.9880478087649402,
+  "global_step": 9000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 0.6922,
+      "step": 10
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 0.6938,
+      "step": 20
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3e-06,
+      "loss": 0.694,
+      "step": 30
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.6926,
+      "step": 40
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5e-06,
+      "loss": 0.6929,
+      "step": 50
+    },
+    {
+      "epoch": 0.02,
+      "eval_accuracy": 0.48244374990463257,
+      "eval_loss": 0.6933985948562622,
+      "eval_runtime": 2988.6075,
+      "eval_samples_per_second": 4.031,
+      "eval_steps_per_second": 0.063,
+      "step": 50
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6e-06,
+      "loss": 0.6935,
+      "step": 60
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.000000000000001e-06,
+      "loss": 0.6929,
+      "step": 70
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.6935,
+      "step": 80
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9e-06,
+      "loss": 0.6934,
+      "step": 90
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1e-05,
+      "loss": 0.6927,
+      "step": 100
+    },
+    {
+      "epoch": 0.03,
+      "eval_accuracy": 0.48244374990463257,
+      "eval_loss": 0.6929991245269775,
+      "eval_runtime": 3009.547,
+      "eval_samples_per_second": 4.003,
+      "eval_steps_per_second": 0.063,
+      "step": 100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.1000000000000001e-05,
+      "loss": 0.6933,
+      "step": 110
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2e-05,
+      "loss": 0.6919,
+      "step": 120
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3000000000000001e-05,
+      "loss": 0.6893,
+      "step": 130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.4000000000000001e-05,
+      "loss": 0.6794,
+      "step": 140
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5e-05,
+      "loss": 0.6701,
+      "step": 150
+    },
+    {
+      "epoch": 0.05,
+      "eval_accuracy": 0.671287477016449,
+      "eval_loss": 0.6748385429382324,
+      "eval_runtime": 3039.4115,
+      "eval_samples_per_second": 3.964,
+      "eval_steps_per_second": 0.062,
+      "step": 150
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.6603,
+      "step": 160
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.7000000000000003e-05,
+      "loss": 0.6428,
+      "step": 170
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8e-05,
+      "loss": 0.643,
+      "step": 180
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9e-05,
+      "loss": 0.6138,
+      "step": 190
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2e-05,
+      "loss": 0.5888,
+      "step": 200
+    },
+    {
+      "epoch": 0.07,
+      "eval_accuracy": 0.815970778465271,
+      "eval_loss": 0.5752072334289551,
+      "eval_runtime": 3007.0357,
+      "eval_samples_per_second": 4.006,
+      "eval_steps_per_second": 0.063,
+      "step": 200
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.1e-05,
+      "loss": 0.5843,
+      "step": 210
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.2000000000000003e-05,
+      "loss": 0.6005,
+      "step": 220
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.3000000000000003e-05,
+      "loss": 0.5417,
+      "step": 230
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.4e-05,
+      "loss": 0.5546,
+      "step": 240
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.5e-05,
+      "loss": 0.5493,
+      "step": 250
+    },
+    {
+      "epoch": 0.08,
+      "eval_accuracy": 0.829750120639801,
+      "eval_loss": 0.507931649684906,
+      "eval_runtime": 2890.9969,
+      "eval_samples_per_second": 4.167,
+      "eval_steps_per_second": 0.065,
+      "step": 250
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.6000000000000002e-05,
+      "loss": 0.519,
+      "step": 260
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.7000000000000002e-05,
+      "loss": 0.4943,
+      "step": 270
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.5312,
+      "step": 280
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9e-05,
+      "loss": 0.5532,
+      "step": 290
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3e-05,
+      "loss": 0.5315,
+      "step": 300
+    },
+    {
+      "epoch": 0.1,
+      "eval_accuracy": 0.7565369009971619,
+      "eval_loss": 0.5359323024749756,
+      "eval_runtime": 3029.4417,
+      "eval_samples_per_second": 3.977,
+      "eval_steps_per_second": 0.062,
+      "step": 300
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.1e-05,
+      "loss": 0.4794,
+      "step": 310
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.4717,
+      "step": 320
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.3e-05,
+      "loss": 0.5199,
+      "step": 330
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.4000000000000007e-05,
+      "loss": 0.4743,
+      "step": 340
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.5e-05,
+      "loss": 0.4576,
+      "step": 350
+    },
+    {
+      "epoch": 0.12,
+      "eval_accuracy": 0.8144766092300415,
+      "eval_loss": 0.4594773054122925,
+      "eval_runtime": 2944.0266,
+      "eval_samples_per_second": 4.092,
+      "eval_steps_per_second": 0.064,
+      "step": 350
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.6e-05,
+      "loss": 0.3828,
+      "step": 360
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.7e-05,
+      "loss": 0.4435,
+      "step": 370
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.8e-05,
+      "loss": 0.391,
+      "step": 380
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.9000000000000006e-05,
+      "loss": 0.5049,
+      "step": 390
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4e-05,
+      "loss": 0.4562,
+      "step": 400
+    },
+    {
+      "epoch": 0.13,
+      "eval_accuracy": 0.8325724005699158,
+      "eval_loss": 0.4149327874183655,
+      "eval_runtime": 2923.6572,
+      "eval_samples_per_second": 4.121,
+      "eval_steps_per_second": 0.065,
+      "step": 400
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.1e-05,
+      "loss": 0.4448,
+      "step": 410
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.2e-05,
+      "loss": 0.5251,
+      "step": 420
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3e-05,
+      "loss": 0.4321,
+      "step": 430
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.5235,
+      "step": 440
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5e-05,
+      "loss": 0.4184,
+      "step": 450
+    },
+    {
+      "epoch": 0.15,
+      "eval_accuracy": 0.8212003111839294,
+      "eval_loss": 0.45684748888015747,
+      "eval_runtime": 2914.0923,
+      "eval_samples_per_second": 4.134,
+      "eval_steps_per_second": 0.065,
+      "step": 450
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.600000000000001e-05,
+      "loss": 0.5847,
+      "step": 460
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7e-05,
+      "loss": 0.512,
+      "step": 470
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.8e-05,
+      "loss": 0.5393,
+      "step": 480
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.9e-05,
+      "loss": 0.5452,
+      "step": 490
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5e-05,
+      "loss": 0.4554,
+      "step": 500
+    },
+    {
+      "epoch": 0.17,
+      "eval_accuracy": 0.8339005708694458,
+      "eval_loss": 0.48061037063598633,
+      "eval_runtime": 2915.8808,
+      "eval_samples_per_second": 4.132,
+      "eval_steps_per_second": 0.065,
+      "step": 500
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.994142455482662e-05,
+      "loss": 0.4601,
+      "step": 510
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.9882849109653237e-05,
+      "loss": 0.5016,
+      "step": 520
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.9824273664479854e-05,
+      "loss": 0.4519,
+      "step": 530
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.9765698219306464e-05,
+      "loss": 0.5132,
+      "step": 540
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.970712277413309e-05,
+      "loss": 0.4444,
+      "step": 550
+    },
+    {
+      "epoch": 0.18,
+      "eval_accuracy": 0.8052627444267273,
+      "eval_loss": 0.4847351312637329,
+      "eval_runtime": 2834.0076,
+      "eval_samples_per_second": 4.251,
+      "eval_steps_per_second": 0.067,
+      "step": 550
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.9648547328959705e-05,
+      "loss": 0.4314,
+      "step": 560
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.9589971883786315e-05,
+      "loss": 0.5419,
+      "step": 570
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.953139643861294e-05,
+      "loss": 0.4737,
+      "step": 580
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.947282099343955e-05,
+      "loss": 0.5337,
+      "step": 590
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.9414245548266166e-05,
+      "loss": 0.538,
+      "step": 600
+    },
+    {
+      "epoch": 0.2,
+      "eval_accuracy": 0.8014443516731262,
+      "eval_loss": 0.45312491059303284,
+      "eval_runtime": 2917.8384,
+      "eval_samples_per_second": 4.129,
+      "eval_steps_per_second": 0.065,
+      "step": 600
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.935567010309279e-05,
+      "loss": 0.381,
+      "step": 610
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.92970946579194e-05,
+      "loss": 0.448,
+      "step": 620
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.923851921274602e-05,
+      "loss": 0.5391,
+      "step": 630
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.9179943767572635e-05,
+      "loss": 0.4214,
+      "step": 640
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.912136832239925e-05,
+      "loss": 0.4842,
+      "step": 650
+    },
+    {
+      "epoch": 0.22,
+      "eval_accuracy": 0.8117373585700989,
+      "eval_loss": 0.45403727889060974,
+      "eval_runtime": 2926.2641,
+      "eval_samples_per_second": 4.117,
+      "eval_steps_per_second": 0.065,
+      "step": 650
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.906279287722587e-05,
+      "loss": 0.456,
+      "step": 660
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.9004217432052486e-05,
+      "loss": 0.479,
+      "step": 670
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.89456419868791e-05,
+      "loss": 0.404,
+      "step": 680
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.888706654170572e-05,
+      "loss": 0.4451,
+      "step": 690
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.882849109653234e-05,
+      "loss": 0.3923,
+      "step": 700
+    },
+    {
+      "epoch": 0.23,
+      "eval_accuracy": 0.8407072424888611,
+      "eval_loss": 0.4229690134525299,
+      "eval_runtime": 2931.6632,
+      "eval_samples_per_second": 4.109,
+      "eval_steps_per_second": 0.064,
+      "step": 700
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.8769915651358954e-05,
+      "loss": 0.4551,
+      "step": 710
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.871134020618557e-05,
+      "loss": 0.4239,
+      "step": 720
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.865276476101219e-05,
+      "loss": 0.4673,
+      "step": 730
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.85941893158388e-05,
+      "loss": 0.4224,
+      "step": 740
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.853561387066542e-05,
+      "loss": 0.4277,
+      "step": 750
+    },
+    {
+      "epoch": 0.25,
+      "eval_accuracy": 0.821449339389801,
+      "eval_loss": 0.45256727933883667,
+      "eval_runtime": 2978.3304,
+      "eval_samples_per_second": 4.045,
+      "eval_steps_per_second": 0.063,
+      "step": 750
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.847703842549204e-05,
+      "loss": 0.4115,
+      "step": 760
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.841846298031865e-05,
+      "loss": 0.3928,
+      "step": 770
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.8359887535145274e-05,
+      "loss": 0.3658,
+      "step": 780
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.8301312089971884e-05,
+      "loss": 0.408,
+      "step": 790
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.82427366447985e-05,
+      "loss": 0.3838,
+      "step": 800
+    },
+    {
+      "epoch": 0.27,
+      "eval_accuracy": 0.8405411839485168,
+      "eval_loss": 0.421342670917511,
+      "eval_runtime": 3022.7481,
+      "eval_samples_per_second": 3.985,
+      "eval_steps_per_second": 0.063,
+      "step": 800
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.818416119962512e-05,
+      "loss": 0.4917,
+      "step": 810
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.8125585754451736e-05,
+      "loss": 0.5618,
+      "step": 820
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.806701030927835e-05,
+      "loss": 0.4129,
+      "step": 830
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.800843486410497e-05,
+      "loss": 0.3701,
+      "step": 840
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.794985941893159e-05,
+      "loss": 0.4728,
+      "step": 850
+    },
+    {
+      "epoch": 0.28,
+      "eval_accuracy": 0.8525773882865906,
+      "eval_loss": 0.36839208006858826,
+      "eval_runtime": 3083.8189,
+      "eval_samples_per_second": 3.907,
+      "eval_steps_per_second": 0.061,
+      "step": 850
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.7891283973758204e-05,
+      "loss": 0.4434,
+      "step": 860
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.783270852858482e-05,
+      "loss": 0.3602,
+      "step": 870
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.777413308341144e-05,
+      "loss": 0.3227,
+      "step": 880
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.771555763823805e-05,
+      "loss": 0.5942,
+      "step": 890
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.765698219306467e-05,
+      "loss": 0.5339,
+      "step": 900
+    },
+    {
+      "epoch": 0.3,
+      "eval_accuracy": 0.7447497248649597,
+      "eval_loss": 0.5473734736442566,
+      "eval_runtime": 3004.5491,
+      "eval_samples_per_second": 4.01,
+      "eval_steps_per_second": 0.063,
+      "step": 900
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.759840674789129e-05,
+      "loss": 0.65,
+      "step": 910
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.75398313027179e-05,
+      "loss": 0.6799,
+      "step": 920
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.7481255857544524e-05,
+      "loss": 0.6257,
+      "step": 930
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.7422680412371134e-05,
+      "loss": 0.501,
+      "step": 940
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.736410496719775e-05,
+      "loss": 0.4643,
+      "step": 950
+    },
+    {
+      "epoch": 0.32,
+      "eval_accuracy": 0.8384659886360168,
+      "eval_loss": 0.4451840817928314,
+      "eval_runtime": 3013.7247,
+      "eval_samples_per_second": 3.997,
+      "eval_steps_per_second": 0.063,
+      "step": 950
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.7305529522024375e-05,
+      "loss": 0.4393,
+      "step": 960
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.7246954076850985e-05,
+      "loss": 0.4277,
+      "step": 970
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.71883786316776e-05,
+      "loss": 0.4803,
+      "step": 980
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.712980318650422e-05,
+      "loss": 0.454,
+      "step": 990
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.7071227741330836e-05,
+      "loss": 0.5764,
+      "step": 1000
+    },
+    {
+      "epoch": 0.33,
+      "eval_accuracy": 0.8483439683914185,
+      "eval_loss": 0.4207330644130707,
+      "eval_runtime": 2999.2124,
+      "eval_samples_per_second": 4.017,
+      "eval_steps_per_second": 0.063,
+      "step": 1000
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.7012652296157454e-05,
+      "loss": 0.4157,
+      "step": 1010
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.695407685098407e-05,
+      "loss": 0.4516,
+      "step": 1020
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.689550140581069e-05,
+      "loss": 0.4845,
+      "step": 1030
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.68369259606373e-05,
+      "loss": 0.4002,
+      "step": 1040
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.677835051546392e-05,
+      "loss": 0.4336,
+      "step": 1050
+    },
+    {
+      "epoch": 0.35,
+      "eval_accuracy": 0.8451066613197327,
+      "eval_loss": 0.41939765214920044,
+      "eval_runtime": 3009.9889,
+      "eval_samples_per_second": 4.002,
+      "eval_steps_per_second": 0.063,
+      "step": 1050
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.671977507029054e-05,
+      "loss": 0.5973,
+      "step": 1060
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.666119962511715e-05,
+      "loss": 0.4049,
+      "step": 1070
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.660262417994377e-05,
+      "loss": 0.4494,
+      "step": 1080
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.6544048734770383e-05,
+      "loss": 0.4352,
+      "step": 1090
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.6485473289597e-05,
+      "loss": 0.3085,
+      "step": 1100
+    },
+    {
+      "epoch": 0.37,
+      "eval_accuracy": 0.8608782291412354,
+      "eval_loss": 0.36667364835739136,
+      "eval_runtime": 2940.4464,
+      "eval_samples_per_second": 4.097,
+      "eval_steps_per_second": 0.064,
+      "step": 1100
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.6426897844423624e-05,
+      "loss": 0.4418,
+      "step": 1110
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.6368322399250235e-05,
+      "loss": 0.2533,
+      "step": 1120
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.630974695407685e-05,
+      "loss": 0.4812,
+      "step": 1130
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.625117150890347e-05,
+      "loss": 0.3775,
+      "step": 1140
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.6192596063730086e-05,
+      "loss": 0.4591,
+      "step": 1150
+    },
+    {
+      "epoch": 0.38,
+      "eval_accuracy": 0.8424503803253174,
+      "eval_loss": 0.4464161694049835,
+      "eval_runtime": 2932.0604,
+      "eval_samples_per_second": 4.109,
+      "eval_steps_per_second": 0.064,
+      "step": 1150
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 4.61340206185567e-05,
+      "loss": 0.5435,
+      "step": 1160
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 4.607544517338332e-05,
+      "loss": 0.4174,
+      "step": 1170
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 4.601686972820994e-05,
+      "loss": 0.3645,
+      "step": 1180
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.5958294283036554e-05,
+      "loss": 0.4174,
+      "step": 1190
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.589971883786317e-05,
+      "loss": 0.4081,
+      "step": 1200
+    },
+    {
+      "epoch": 0.4,
+      "eval_accuracy": 0.8350626826286316,
+      "eval_loss": 0.41294723749160767,
+      "eval_runtime": 2933.542,
+      "eval_samples_per_second": 4.107,
+      "eval_steps_per_second": 0.064,
+      "step": 1200
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.584114339268979e-05,
+      "loss": 0.469,
+      "step": 1210
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 4.5782567947516406e-05,
+      "loss": 0.2987,
+      "step": 1220
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 4.572399250234302e-05,
+      "loss": 0.4368,
+      "step": 1230
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 4.566541705716963e-05,
+      "loss": 0.4313,
+      "step": 1240
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 4.560684161199626e-05,
+      "loss": 0.3453,
+      "step": 1250
+    },
+    {
+      "epoch": 0.42,
+      "eval_accuracy": 0.8547356128692627,
+      "eval_loss": 0.36404505372047424,
+      "eval_runtime": 2949.5837,
+      "eval_samples_per_second": 4.084,
+      "eval_steps_per_second": 0.064,
+      "step": 1250
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 4.5548266166822874e-05,
+      "loss": 0.3765,
+      "step": 1260
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 4.5489690721649484e-05,
+      "loss": 0.4036,
+      "step": 1270
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 4.543111527647611e-05,
+      "loss": 0.4914,
+      "step": 1280
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 4.537253983130272e-05,
+      "loss": 0.399,
+      "step": 1290
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 4.5313964386129336e-05,
+      "loss": 0.3865,
+      "step": 1300
+    },
+    {
+      "epoch": 0.43,
+      "eval_accuracy": 0.8540715575218201,
+      "eval_loss": 0.3634878098964691,
+      "eval_runtime": 2974.1469,
+      "eval_samples_per_second": 4.051,
+      "eval_steps_per_second": 0.064,
+      "step": 1300
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 4.525538894095596e-05,
+      "loss": 0.4852,
+      "step": 1310
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 4.519681349578257e-05,
+      "loss": 0.4521,
+      "step": 1320
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 4.513823805060919e-05,
+      "loss": 0.4128,
+      "step": 1330
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 4.5079662605435804e-05,
+      "loss": 0.4878,
+      "step": 1340
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 4.502108716026242e-05,
+      "loss": 0.3853,
+      "step": 1350
+    },
+    {
+      "epoch": 0.45,
+      "eval_accuracy": 0.8607951998710632,
+      "eval_loss": 0.39044466614723206,
+      "eval_runtime": 3018.7831,
+      "eval_samples_per_second": 3.991,
+      "eval_steps_per_second": 0.063,
+      "step": 1350
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 4.496251171508904e-05,
+      "loss": 0.3551,
+      "step": 1360
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 4.4903936269915655e-05,
+      "loss": 0.4035,
+      "step": 1370
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 4.484536082474227e-05,
+      "loss": 0.4232,
+      "step": 1380
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 4.478678537956888e-05,
+      "loss": 0.456,
+      "step": 1390
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 4.4728209934395506e-05,
+      "loss": 0.5111,
+      "step": 1400
+    },
+    {
+      "epoch": 0.46,
+      "eval_accuracy": 0.84651780128479,
+      "eval_loss": 0.3800100088119507,
+      "eval_runtime": 3045.1951,
+      "eval_samples_per_second": 3.956,
+      "eval_steps_per_second": 0.062,
+      "step": 1400
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 4.466963448922212e-05,
+      "loss": 0.3289,
+      "step": 1410
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 4.4611059044048734e-05,
+      "loss": 0.4073,
+      "step": 1420
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 4.455248359887536e-05,
+      "loss": 0.3507,
+      "step": 1430
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 4.449390815370197e-05,
+      "loss": 0.4125,
+      "step": 1440
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 4.4435332708528585e-05,
+      "loss": 0.401,
+      "step": 1450
+    },
+    {
+      "epoch": 0.48,
+      "eval_accuracy": 0.8611272573471069,
+      "eval_loss": 0.37042567133903503,
+      "eval_runtime": 2973.5921,
+      "eval_samples_per_second": 4.051,
+      "eval_steps_per_second": 0.064,
+      "step": 1450
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 4.43767572633552e-05,
+      "loss": 0.4538,
+      "step": 1460
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 4.431818181818182e-05,
+      "loss": 0.3909,
+      "step": 1470
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 4.4259606373008436e-05,
+      "loss": 0.4123,
+      "step": 1480
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 4.4201030927835053e-05,
+      "loss": 0.4419,
+      "step": 1490
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.414245548266167e-05,
+      "loss": 0.3334,
+      "step": 1500
+    },
+    {
+      "epoch": 0.5,
+      "eval_accuracy": 0.8576409220695496,
+      "eval_loss": 0.38589969277381897,
+      "eval_runtime": 2945.6098,
+      "eval_samples_per_second": 4.09,
+      "eval_steps_per_second": 0.064,
+      "step": 1500
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.408388003748829e-05,
+      "loss": 0.28,
+      "step": 1510
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.4025304592314905e-05,
+      "loss": 0.3848,
+      "step": 1520
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.396672914714152e-05,
+      "loss": 0.3276,
+      "step": 1530
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.390815370196814e-05,
+      "loss": 0.3681,
+      "step": 1540
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.3849578256794756e-05,
+      "loss": 0.2806,
+      "step": 1550
+    },
+    {
+      "epoch": 0.51,
+      "eval_accuracy": 0.8264298439025879,
+      "eval_loss": 0.41383180022239685,
+      "eval_runtime": 2893.0129,
+      "eval_samples_per_second": 4.164,
+      "eval_steps_per_second": 0.065,
+      "step": 1550
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.3791002811621366e-05,
+      "loss": 0.3561,
+      "step": 1560
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.373242736644798e-05,
+      "loss": 0.3713,
+      "step": 1570
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.367385192127461e-05,
+      "loss": 0.3353,
+      "step": 1580
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.361527647610122e-05,
+      "loss": 0.3803,
+      "step": 1590
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.3556701030927835e-05,
+      "loss": 0.4595,
+      "step": 1600
+    },
+    {
+      "epoch": 0.53,
+      "eval_accuracy": 0.8476799130439758,
+      "eval_loss": 0.3633853793144226,
+      "eval_runtime": 2921.8289,
+      "eval_samples_per_second": 4.123,
+      "eval_steps_per_second": 0.065,
+      "step": 1600
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.349812558575445e-05,
+      "loss": 0.3614,
+      "step": 1610
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.343955014058107e-05,
+      "loss": 0.3779,
+      "step": 1620
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3380974695407686e-05,
+      "loss": 0.4611,
+      "step": 1630
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.33223992502343e-05,
+      "loss": 0.2935,
+      "step": 1640
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.326382380506092e-05,
+      "loss": 0.3706,
+      "step": 1650
+    },
+    {
+      "epoch": 0.55,
+      "eval_accuracy": 0.8651946783065796,
+      "eval_loss": 0.3683921694755554,
+      "eval_runtime": 2922.8189,
+      "eval_samples_per_second": 4.122,
+      "eval_steps_per_second": 0.065,
+      "step": 1650
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.320524835988754e-05,
+      "loss": 0.2781,
+      "step": 1660
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3146672914714154e-05,
+      "loss": 0.4412,
+      "step": 1670
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.308809746954077e-05,
+      "loss": 0.3983,
+      "step": 1680
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.302952202436739e-05,
+      "loss": 0.5213,
+      "step": 1690
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2970946579194005e-05,
+      "loss": 0.3744,
+      "step": 1700
+    },
+    {
+      "epoch": 0.56,
+      "eval_accuracy": 0.8672698736190796,
+      "eval_loss": 0.33778074383735657,
+      "eval_runtime": 2900.2504,
+      "eval_samples_per_second": 4.154,
+      "eval_steps_per_second": 0.065,
+      "step": 1700
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.2912371134020616e-05,
+      "loss": 0.267,
+      "step": 1710
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.285379568884724e-05,
+      "loss": 0.3318,
+      "step": 1720
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.279522024367386e-05,
+      "loss": 0.4457,
+      "step": 1730
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.273664479850047e-05,
+      "loss": 0.3101,
+      "step": 1740
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.267806935332709e-05,
+      "loss": 0.3573,
+      "step": 1750
+    },
+    {
+      "epoch": 0.58,
+      "eval_accuracy": 0.874906599521637,
+      "eval_loss": 0.3185945451259613,
+      "eval_runtime": 2921.437,
+      "eval_samples_per_second": 4.124,
+      "eval_steps_per_second": 0.065,
+      "step": 1750
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.26194939081537e-05,
+      "loss": 0.3407,
+      "step": 1760
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.256091846298032e-05,
+      "loss": 0.2661,
+      "step": 1770
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.250234301780694e-05,
+      "loss": 0.4254,
+      "step": 1780
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.244376757263355e-05,
+      "loss": 0.3771,
+      "step": 1790
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.238519212746017e-05,
+      "loss": 0.3605,
+      "step": 1800
+    },
+    {
+      "epoch": 0.6,
+      "eval_accuracy": 0.8687639832496643,
+      "eval_loss": 0.36973297595977783,
+      "eval_runtime": 2923.8943,
+      "eval_samples_per_second": 4.12,
+      "eval_steps_per_second": 0.065,
+      "step": 1800
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.232661668228679e-05,
+      "loss": 0.427,
+      "step": 1810
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.2268041237113404e-05,
+      "loss": 0.4451,
+      "step": 1820
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.220946579194002e-05,
+      "loss": 0.3673,
+      "step": 1830
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.215089034676664e-05,
+      "loss": 0.3658,
+      "step": 1840
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.2092314901593255e-05,
+      "loss": 0.4364,
+      "step": 1850
+    },
+    {
+      "epoch": 0.61,
+      "eval_accuracy": 0.8597161173820496,
+      "eval_loss": 0.3295934498310089,
+      "eval_runtime": 2931.5089,
+      "eval_samples_per_second": 4.109,
+      "eval_steps_per_second": 0.064,
+      "step": 1850
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 4.2033739456419865e-05,
+      "loss": 0.3393,
+      "step": 1860
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 4.197516401124649e-05,
+      "loss": 0.3867,
+      "step": 1870
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 4.1916588566073106e-05,
+      "loss": 0.2674,
+      "step": 1880
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 4.1858013120899717e-05,
+      "loss": 0.4421,
+      "step": 1890
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 4.179943767572634e-05,
+      "loss": 0.3998,
+      "step": 1900
+    },
+    {
+      "epoch": 0.63,
+      "eval_accuracy": 0.8741595149040222,
+      "eval_loss": 0.3256959021091461,
+      "eval_runtime": 2930.6044,
+      "eval_samples_per_second": 4.111,
+      "eval_steps_per_second": 0.064,
+      "step": 1900
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 4.174086223055295e-05,
+      "loss": 0.4353,
+      "step": 1910
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.168228678537957e-05,
+      "loss": 0.4334,
+      "step": 1920
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.162371134020619e-05,
+      "loss": 0.3184,
+      "step": 1930
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.15651358950328e-05,
+      "loss": 0.3467,
+      "step": 1940
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 4.150656044985942e-05,
+      "loss": 0.422,
+      "step": 1950
+    },
+    {
+      "epoch": 0.65,
+      "eval_accuracy": 0.8738275170326233,
+      "eval_loss": 0.32080963253974915,
+      "eval_runtime": 2931.2514,
+      "eval_samples_per_second": 4.11,
+      "eval_steps_per_second": 0.064,
+      "step": 1950
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 4.1447985004686036e-05,
+      "loss": 0.3593,
+      "step": 1960
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 4.138940955951265e-05,
+      "loss": 0.2707,
+      "step": 1970
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 4.133083411433927e-05,
+      "loss": 0.2846,
+      "step": 1980
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 4.127225866916589e-05,
+      "loss": 0.3786,
+      "step": 1990
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 4.1213683223992505e-05,
+      "loss": 0.3744,
+      "step": 2000
+    },
+    {
+      "epoch": 0.66,
+      "eval_accuracy": 0.8722503781318665,
+      "eval_loss": 0.32994359731674194,
+      "eval_runtime": 2967.2199,
+      "eval_samples_per_second": 4.06,
+      "eval_steps_per_second": 0.064,
+      "step": 2000
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.115510777881912e-05,
+      "loss": 0.3574,
+      "step": 2010
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.109653233364574e-05,
+      "loss": 0.3398,
+      "step": 2020
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.1037956888472356e-05,
+      "loss": 0.4685,
+      "step": 2030
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.097938144329897e-05,
+      "loss": 0.3352,
+      "step": 2040
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.092080599812559e-05,
+      "loss": 0.3303,
+      "step": 2050
+    },
+    {
+      "epoch": 0.68,
+      "eval_accuracy": 0.8800531029701233,
+      "eval_loss": 0.30461445450782776,
+      "eval_runtime": 2971.8504,
+      "eval_samples_per_second": 4.054,
+      "eval_steps_per_second": 0.064,
+      "step": 2050
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.08622305529522e-05,
+      "loss": 0.2883,
+      "step": 2060
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.0803655107778824e-05,
+      "loss": 0.3605,
+      "step": 2070
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.074507966260544e-05,
+      "loss": 0.3245,
+      "step": 2080
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.068650421743205e-05,
+      "loss": 0.3353,
+      "step": 2090
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.0627928772258675e-05,
+      "loss": 0.3163,
+      "step": 2100
+    },
+    {
+      "epoch": 0.7,
+      "eval_accuracy": 0.871171236038208,
+      "eval_loss": 0.3203169107437134,
+      "eval_runtime": 2994.2709,
+      "eval_samples_per_second": 4.023,
+      "eval_steps_per_second": 0.063,
+      "step": 2100
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.0569353327085286e-05,
+      "loss": 0.3334,
+      "step": 2110
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.05107778819119e-05,
+      "loss": 0.3228,
+      "step": 2120
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.045220243673852e-05,
+      "loss": 0.2895,
+      "step": 2130
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.039362699156514e-05,
+      "loss": 0.3289,
+      "step": 2140
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0335051546391754e-05,
+      "loss": 0.3407,
+      "step": 2150
+    },
+    {
+      "epoch": 0.71,
+      "eval_accuracy": 0.8810492157936096,
+      "eval_loss": 0.30052676796913147,
+      "eval_runtime": 2994.4783,
+      "eval_samples_per_second": 4.023,
+      "eval_steps_per_second": 0.063,
+      "step": 2150
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.027647610121837e-05,
+      "loss": 0.2437,
+      "step": 2160
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.021790065604499e-05,
+      "loss": 0.3425,
+      "step": 2170
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.0159325210871605e-05,
+      "loss": 0.4596,
+      "step": 2180
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.010074976569822e-05,
+      "loss": 0.3299,
+      "step": 2190
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.004217432052484e-05,
+      "loss": 0.3216,
+      "step": 2200
+    },
+    {
+      "epoch": 0.73,
+      "eval_accuracy": 0.8680169582366943,
+      "eval_loss": 0.3160869777202606,
+      "eval_runtime": 3007.0708,
+      "eval_samples_per_second": 4.006,
+      "eval_steps_per_second": 0.063,
+      "step": 2200
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.998359887535145e-05,
+      "loss": 0.3645,
+      "step": 2210
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.9925023430178074e-05,
+      "loss": 0.3015,
+      "step": 2220
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.986644798500469e-05,
+      "loss": 0.2759,
+      "step": 2230
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.98078725398313e-05,
+      "loss": 0.3127,
+      "step": 2240
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.9749297094657925e-05,
+      "loss": 0.298,
+      "step": 2250
+    },
+    {
+      "epoch": 0.75,
+      "eval_accuracy": 0.880385160446167,
+      "eval_loss": 0.3090185225009918,
+      "eval_runtime": 2969.1687,
+      "eval_samples_per_second": 4.057,
+      "eval_steps_per_second": 0.064,
+      "step": 2250
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.9690721649484535e-05,
+      "loss": 0.2428,
+      "step": 2260
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.963214620431115e-05,
+      "loss": 0.4462,
+      "step": 2270
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9573570759137776e-05,
+      "loss": 0.3003,
+      "step": 2280
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9514995313964387e-05,
+      "loss": 0.4282,
+      "step": 2290
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9456419868791004e-05,
+      "loss": 0.3384,
+      "step": 2300
+    },
+    {
+      "epoch": 0.76,
+      "eval_accuracy": 0.8790570497512817,
+      "eval_loss": 0.3029511868953705,
+      "eval_runtime": 2941.0141,
+      "eval_samples_per_second": 4.096,
+      "eval_steps_per_second": 0.064,
+      "step": 2300
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.939784442361762e-05,
+      "loss": 0.3612,
+      "step": 2310
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.933926897844424e-05,
+      "loss": 0.3815,
+      "step": 2320
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.9280693533270855e-05,
+      "loss": 0.3859,
+      "step": 2330
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.922211808809747e-05,
+      "loss": 0.2775,
+      "step": 2340
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.916354264292409e-05,
+      "loss": 0.2791,
+      "step": 2350
+    },
+    {
+      "epoch": 0.78,
+      "eval_accuracy": 0.8856146931648254,
+      "eval_loss": 0.30784520506858826,
+      "eval_runtime": 2943.0764,
+      "eval_samples_per_second": 4.093,
+      "eval_steps_per_second": 0.064,
+      "step": 2350
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.91049671977507e-05,
+      "loss": 0.3565,
+      "step": 2360
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.904639175257732e-05,
+      "loss": 0.2971,
+      "step": 2370
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.898781630740394e-05,
+      "loss": 0.3991,
+      "step": 2380
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.892924086223055e-05,
+      "loss": 0.3312,
+      "step": 2390
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.8870665417057174e-05,
+      "loss": 0.365,
+      "step": 2400
+    },
+    {
+      "epoch": 0.8,
+      "eval_accuracy": 0.8855316638946533,
+      "eval_loss": 0.29047778248786926,
+      "eval_runtime": 3008.1086,
+      "eval_samples_per_second": 4.005,
+      "eval_steps_per_second": 0.063,
+      "step": 2400
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.8812089971883785e-05,
+      "loss": 0.2835,
+      "step": 2410
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.87535145267104e-05,
+      "loss": 0.3858,
+      "step": 2420
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.8694939081537026e-05,
+      "loss": 0.2081,
+      "step": 2430
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.8636363636363636e-05,
+      "loss": 0.4151,
+      "step": 2440
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.857778819119025e-05,
+      "loss": 0.273,
+      "step": 2450
+    },
+    {
+      "epoch": 0.81,
+      "eval_accuracy": 0.887523889541626,
+      "eval_loss": 0.3015758991241455,
+      "eval_runtime": 3006.6054,
+      "eval_samples_per_second": 4.007,
+      "eval_steps_per_second": 0.063,
+      "step": 2450
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.851921274601687e-05,
+      "loss": 0.2739,
+      "step": 2460
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.846063730084349e-05,
+      "loss": 0.3282,
+      "step": 2470
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.8402061855670104e-05,
+      "loss": 0.346,
+      "step": 2480
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.834348641049672e-05,
+      "loss": 0.3466,
+      "step": 2490
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.828491096532334e-05,
+      "loss": 0.3192,
+      "step": 2500
+    },
+    {
+      "epoch": 0.83,
+      "eval_accuracy": 0.8812152147293091,
+      "eval_loss": 0.29833853244781494,
+      "eval_runtime": 3010.2051,
+      "eval_samples_per_second": 4.002,
+      "eval_steps_per_second": 0.063,
+      "step": 2500
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.8226335520149956e-05,
+      "loss": 0.2924,
+      "step": 2510
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.816776007497657e-05,
+      "loss": 0.189,
+      "step": 2520
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.810918462980319e-05,
+      "loss": 0.3818,
+      "step": 2530
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.805060918462981e-05,
+      "loss": 0.3296,
+      "step": 2540
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.7992033739456424e-05,
+      "loss": 0.3126,
+      "step": 2550
+    },
+    {
+      "epoch": 0.85,
+      "eval_accuracy": 0.8775628805160522,
+      "eval_loss": 0.33056697249412537,
+      "eval_runtime": 2975.7054,
+      "eval_samples_per_second": 4.048,
+      "eval_steps_per_second": 0.064,
+      "step": 2550
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.7933458294283034e-05,
+      "loss": 0.4875,
+      "step": 2560
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.787488284910966e-05,
+      "loss": 0.2335,
+      "step": 2570
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.7816307403936275e-05,
+      "loss": 0.4201,
+      "step": 2580
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.7757731958762886e-05,
+      "loss": 0.2644,
+      "step": 2590
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.769915651358951e-05,
+      "loss": 0.282,
+      "step": 2600
+    },
+    {
+      "epoch": 0.86,
+      "eval_accuracy": 0.8815472722053528,
+      "eval_loss": 0.3129451870918274,
+      "eval_runtime": 2979.7422,
+      "eval_samples_per_second": 4.043,
+      "eval_steps_per_second": 0.063,
+      "step": 2600
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 3.764058106841612e-05,
+      "loss": 0.2824,
+      "step": 2610
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 3.758200562324274e-05,
+      "loss": 0.3408,
+      "step": 2620
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 3.752343017806936e-05,
+      "loss": 0.3771,
+      "step": 2630
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.746485473289597e-05,
+      "loss": 0.2865,
+      "step": 2640
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.740627928772259e-05,
+      "loss": 0.3097,
+      "step": 2650
+    },
+    {
+      "epoch": 0.88,
+      "eval_accuracy": 0.8850336074829102,
+      "eval_loss": 0.2975265085697174,
+      "eval_runtime": 2933.946,
+      "eval_samples_per_second": 4.106,
+      "eval_steps_per_second": 0.064,
+      "step": 2650
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7347703842549205e-05,
+      "loss": 0.302,
+      "step": 2660
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.728912839737582e-05,
+      "loss": 0.2886,
+      "step": 2670
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.723055295220244e-05,
+      "loss": 0.3093,
+      "step": 2680
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.7171977507029056e-05,
+      "loss": 0.3629,
+      "step": 2690
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 3.7113402061855674e-05,
+      "loss": 0.3492,
+      "step": 2700
+    },
+    {
+      "epoch": 0.9,
+      "eval_accuracy": 0.8750726580619812,
+      "eval_loss": 0.3024357259273529,
+      "eval_runtime": 2930.5608,
+      "eval_samples_per_second": 4.111,
+      "eval_steps_per_second": 0.064,
+      "step": 2700
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 3.7054826616682284e-05,
+      "loss": 0.2758,
+      "step": 2710
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 3.699625117150891e-05,
+      "loss": 0.3081,
+      "step": 2720
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.6937675726335525e-05,
+      "loss": 0.2572,
+      "step": 2730
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.6879100281162135e-05,
+      "loss": 0.4114,
+      "step": 2740
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.682052483598876e-05,
+      "loss": 0.3099,
+      "step": 2750
+    },
+    {
+      "epoch": 0.91,
+      "eval_accuracy": 0.867020845413208,
+      "eval_loss": 0.3481622338294983,
+      "eval_runtime": 2905.5736,
+      "eval_samples_per_second": 4.146,
+      "eval_steps_per_second": 0.065,
+      "step": 2750
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.676194939081537e-05,
+      "loss": 0.3577,
+      "step": 2760
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6703373945641986e-05,
+      "loss": 0.238,
+      "step": 2770
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.664479850046861e-05,
+      "loss": 0.3833,
+      "step": 2780
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.658622305529522e-05,
+      "loss": 0.3692,
+      "step": 2790
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.652764761012184e-05,
+      "loss": 0.3225,
+      "step": 2800
+    },
+    {
+      "epoch": 0.93,
+      "eval_accuracy": 0.8865277767181396,
+      "eval_loss": 0.3081686198711395,
+      "eval_runtime": 2912.8093,
+      "eval_samples_per_second": 4.136,
+      "eval_steps_per_second": 0.065,
+      "step": 2800
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6469072164948455e-05,
+      "loss": 0.3342,
+      "step": 2810
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.641049671977507e-05,
+      "loss": 0.3183,
+      "step": 2820
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.635192127460169e-05,
+      "loss": 0.3336,
+      "step": 2830
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.6293345829428306e-05,
+      "loss": 0.333,
+      "step": 2840
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.623477038425492e-05,
+      "loss": 0.3037,
+      "step": 2850
+    },
+    {
+      "epoch": 0.95,
+      "eval_accuracy": 0.8706731796264648,
+      "eval_loss": 0.34793028235435486,
+      "eval_runtime": 2922.6449,
+      "eval_samples_per_second": 4.122,
+      "eval_steps_per_second": 0.065,
+      "step": 2850
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.617619493908154e-05,
+      "loss": 0.3514,
+      "step": 2860
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.611761949390816e-05,
+      "loss": 0.3722,
+      "step": 2870
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6059044048734774e-05,
+      "loss": 0.3864,
+      "step": 2880
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.600046860356139e-05,
+      "loss": 0.3237,
+      "step": 2890
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.594189315838801e-05,
+      "loss": 0.3177,
+      "step": 2900
+    },
+    {
+      "epoch": 0.96,
+      "eval_accuracy": 0.8834564685821533,
+      "eval_loss": 0.31710025668144226,
+      "eval_runtime": 2918.9433,
+      "eval_samples_per_second": 4.127,
+      "eval_steps_per_second": 0.065,
+      "step": 2900
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.588331771321462e-05,
+      "loss": 0.3124,
+      "step": 2910
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5824742268041236e-05,
+      "loss": 0.373,
+      "step": 2920
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.576616682286786e-05,
+      "loss": 0.2667,
+      "step": 2930
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.570759137769447e-05,
+      "loss": 0.3555,
+      "step": 2940
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.564901593252109e-05,
+      "loss": 0.3124,
+      "step": 2950
+    },
+    {
+      "epoch": 0.98,
+      "eval_accuracy": 0.8821283578872681,
+      "eval_loss": 0.30336007475852966,
+      "eval_runtime": 2908.0609,
+      "eval_samples_per_second": 4.143,
+      "eval_steps_per_second": 0.065,
+      "step": 2950
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.5590440487347704e-05,
+      "loss": 0.3011,
+      "step": 2960
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.553186504217432e-05,
+      "loss": 0.3634,
+      "step": 2970
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.547328959700094e-05,
+      "loss": 0.3035,
+      "step": 2980
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.5414714151827556e-05,
+      "loss": 0.3098,
+      "step": 2990
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.535613870665417e-05,
+      "loss": 0.2924,
+      "step": 3000
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.8852826356887817,
+      "eval_loss": 0.29924488067626953,
+      "eval_runtime": 2973.317,
+      "eval_samples_per_second": 4.052,
+      "eval_steps_per_second": 0.064,
+      "step": 3000
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.529756326148079e-05,
+      "loss": 0.2763,
+      "step": 3010
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.523898781630741e-05,
+      "loss": 0.3134,
+      "step": 3020
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 3.5180412371134024e-05,
+      "loss": 0.1996,
+      "step": 3030
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 3.512183692596064e-05,
+      "loss": 0.2786,
+      "step": 3040
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 3.506326148078726e-05,
+      "loss": 0.2568,
+      "step": 3050
+    },
+    {
+      "epoch": 1.01,
+      "eval_accuracy": 0.8895160555839539,
+      "eval_loss": 0.3091294467449188,
+      "eval_runtime": 2930.6001,
+      "eval_samples_per_second": 4.111,
+      "eval_steps_per_second": 0.064,
+      "step": 3050
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 3.500468603561387e-05,
+      "loss": 0.2333,
+      "step": 3060
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 3.494611059044049e-05,
+      "loss": 0.342,
+      "step": 3070
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 3.488753514526711e-05,
+      "loss": 0.2609,
+      "step": 3080
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 3.482895970009372e-05,
+      "loss": 0.2487,
+      "step": 3090
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 3.4770384254920344e-05,
+      "loss": 0.3105,
+      "step": 3100
+    },
+    {
+      "epoch": 1.03,
+      "eval_accuracy": 0.8702581524848938,
+      "eval_loss": 0.3597377836704254,
+      "eval_runtime": 2919.3463,
+      "eval_samples_per_second": 4.127,
+      "eval_steps_per_second": 0.065,
+      "step": 3100
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 3.4711808809746954e-05,
+      "loss": 0.2523,
+      "step": 3110
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 3.465323336457357e-05,
+      "loss": 0.289,
+      "step": 3120
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 3.4594657919400195e-05,
+      "loss": 0.1788,
+      "step": 3130
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 3.4536082474226805e-05,
+      "loss": 0.2311,
+      "step": 3140
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 3.447750702905342e-05,
+      "loss": 0.2572,
+      "step": 3150
+    },
+    {
+      "epoch": 1.05,
+      "eval_accuracy": 0.8855316638946533,
+      "eval_loss": 0.2860332131385803,
+      "eval_runtime": 2916.4885,
+      "eval_samples_per_second": 4.131,
+      "eval_steps_per_second": 0.065,
+      "step": 3150
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 3.441893158388004e-05,
+      "loss": 0.1942,
+      "step": 3160
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 3.4360356138706656e-05,
+      "loss": 0.257,
+      "step": 3170
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 3.430178069353327e-05,
+      "loss": 0.2219,
+      "step": 3180
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 3.424320524835989e-05,
+      "loss": 0.2896,
+      "step": 3190
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 3.418462980318651e-05,
+      "loss": 0.2675,
+      "step": 3200
+    },
+    {
+      "epoch": 1.06,
+      "eval_accuracy": 0.8843695521354675,
+      "eval_loss": 0.33106133341789246,
+      "eval_runtime": 2926.6818,
+      "eval_samples_per_second": 4.116,
+      "eval_steps_per_second": 0.065,
+      "step": 3200
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 3.412605435801312e-05,
+      "loss": 0.1938,
+      "step": 3210
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 3.406747891283974e-05,
+      "loss": 0.1344,
+      "step": 3220
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 3.400890346766635e-05,
+      "loss": 0.3441,
+      "step": 3230
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 3.395032802249297e-05,
+      "loss": 0.2335,
+      "step": 3240
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 3.389175257731959e-05,
+      "loss": 0.2932,
+      "step": 3250
+    },
+    {
+      "epoch": 1.08,
+      "eval_accuracy": 0.8847845792770386,
+      "eval_loss": 0.35532134771347046,
+      "eval_runtime": 2915.1623,
+      "eval_samples_per_second": 4.133,
+      "eval_steps_per_second": 0.065,
+      "step": 3250
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 3.38331771321462e-05,
+      "loss": 0.3233,
+      "step": 3260
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 3.377460168697282e-05,
+      "loss": 0.1836,
+      "step": 3270
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 3.371602624179944e-05,
+      "loss": 0.1725,
+      "step": 3280
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 3.3657450796626055e-05,
+      "loss": 0.2483,
+      "step": 3290
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 3.359887535145267e-05,
+      "loss": 0.3317,
+      "step": 3300
+    },
+    {
+      "epoch": 1.1,
+      "eval_accuracy": 0.8862787485122681,
+      "eval_loss": 0.3190576136112213,
+      "eval_runtime": 2912.1453,
+      "eval_samples_per_second": 4.137,
+      "eval_steps_per_second": 0.065,
+      "step": 3300
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 3.354029990627929e-05,
+      "loss": 0.2593,
+      "step": 3310
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 3.3481724461105906e-05,
+      "loss": 0.2421,
+      "step": 3320
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 3.342314901593252e-05,
+      "loss": 0.2945,
+      "step": 3330
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 3.336457357075914e-05,
+      "loss": 0.311,
+      "step": 3340
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 3.330599812558576e-05,
+      "loss": 0.3351,
+      "step": 3350
+    },
+    {
+      "epoch": 1.11,
+      "eval_accuracy": 0.8708392381668091,
+      "eval_loss": 0.4064333736896515,
+      "eval_runtime": 3015.0212,
+      "eval_samples_per_second": 3.996,
+      "eval_steps_per_second": 0.063,
+      "step": 3350
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 3.3247422680412374e-05,
+      "loss": 0.2762,
+      "step": 3360
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 3.318884723523899e-05,
+      "loss": 0.3162,
+      "step": 3370
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 3.31302717900656e-05,
+      "loss": 0.3115,
+      "step": 3380
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 3.3071696344892225e-05,
+      "loss": 0.2265,
+      "step": 3390
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 3.301312089971884e-05,
+      "loss": 0.1996,
+      "step": 3400
+    },
+    {
+      "epoch": 1.13,
+      "eval_accuracy": 0.8877729177474976,
+      "eval_loss": 0.30873024463653564,
+      "eval_runtime": 2978.9416,
+      "eval_samples_per_second": 4.044,
+      "eval_steps_per_second": 0.063,
+      "step": 3400
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 3.295454545454545e-05,
+      "loss": 0.2439,
+      "step": 3410
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 3.289597000937208e-05,
+      "loss": 0.1656,
+      "step": 3420
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 3.283739456419869e-05,
+      "loss": 0.2516,
+      "step": 3430
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 3.2778819119025304e-05,
+      "loss": 0.3645,
+      "step": 3440
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 3.272024367385193e-05,
+      "loss": 0.2047,
+      "step": 3450
+    },
+    {
+      "epoch": 1.15,
+      "eval_accuracy": 0.8866937756538391,
+      "eval_loss": 0.34220683574676514,
+      "eval_runtime": 3014.7588,
+      "eval_samples_per_second": 3.996,
+      "eval_steps_per_second": 0.063,
+      "step": 3450
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 3.266166822867854e-05,
+      "loss": 0.3218,
+      "step": 3460
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 3.2603092783505155e-05,
+      "loss": 0.2646,
+      "step": 3470
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 3.254451733833177e-05,
+      "loss": 0.3432,
+      "step": 3480
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 3.248594189315839e-05,
+      "loss": 0.2153,
+      "step": 3490
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 3.242736644798501e-05,
+      "loss": 0.2515,
+      "step": 3500
+    },
+    {
+      "epoch": 1.16,
+      "eval_accuracy": 0.8876898884773254,
+      "eval_loss": 0.3146258592605591,
+      "eval_runtime": 2970.8234,
+      "eval_samples_per_second": 4.055,
+      "eval_steps_per_second": 0.064,
+      "step": 3500
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 3.2368791002811624e-05,
+      "loss": 0.251,
+      "step": 3510
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 3.231021555763824e-05,
+      "loss": 0.3896,
+      "step": 3520
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 3.225164011246485e-05,
+      "loss": 0.3057,
+      "step": 3530
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 3.2193064667291475e-05,
+      "loss": 0.2114,
+      "step": 3540
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 3.213448922211809e-05,
+      "loss": 0.2652,
+      "step": 3550
+    },
+    {
+      "epoch": 1.18,
+      "eval_accuracy": 0.8839545249938965,
+      "eval_loss": 0.3532884418964386,
+      "eval_runtime": 2967.08,
+      "eval_samples_per_second": 4.06,
+      "eval_steps_per_second": 0.064,
+      "step": 3550
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 3.20759137769447e-05,
+      "loss": 0.2321,
+      "step": 3560
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 3.2017338331771326e-05,
+      "loss": 0.311,
+      "step": 3570
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 3.1958762886597937e-05,
+      "loss": 0.3379,
+      "step": 3580
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 3.1900187441424554e-05,
+      "loss": 0.1771,
+      "step": 3590
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 3.184161199625118e-05,
+      "loss": 0.3286,
+      "step": 3600
+    },
+    {
+      "epoch": 1.2,
+      "eval_accuracy": 0.8923383355140686,
+      "eval_loss": 0.30157408118247986,
+      "eval_runtime": 3002.6026,
+      "eval_samples_per_second": 4.012,
+      "eval_steps_per_second": 0.063,
+      "step": 3600
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 3.178303655107779e-05,
+      "loss": 0.3123,
+      "step": 3610
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 3.1724461105904405e-05,
+      "loss": 0.2986,
+      "step": 3620
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.166588566073102e-05,
+      "loss": 0.2499,
+      "step": 3630
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.160731021555764e-05,
+      "loss": 0.2738,
+      "step": 3640
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.1548734770384256e-05,
+      "loss": 0.2658,
+      "step": 3650
+    },
+    {
+      "epoch": 1.21,
+      "eval_accuracy": 0.8765667676925659,
+      "eval_loss": 0.35946688055992126,
+      "eval_runtime": 2957.309,
+      "eval_samples_per_second": 4.074,
+      "eval_steps_per_second": 0.064,
+      "step": 3650
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.149015932521087e-05,
+      "loss": 0.2732,
+      "step": 3660
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.143158388003749e-05,
+      "loss": 0.3002,
+      "step": 3670
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.13730084348641e-05,
+      "loss": 0.3389,
+      "step": 3680
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.1314432989690725e-05,
+      "loss": 0.3097,
+      "step": 3690
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.125585754451734e-05,
+      "loss": 0.2652,
+      "step": 3700
+    },
+    {
+      "epoch": 1.23,
+      "eval_accuracy": 0.8722503781318665,
+      "eval_loss": 0.30808377265930176,
+      "eval_runtime": 2963.9764,
+      "eval_samples_per_second": 4.064,
+      "eval_steps_per_second": 0.064,
+      "step": 3700
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.119728209934395e-05,
+      "loss": 0.2672,
+      "step": 3710
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.1138706654170576e-05,
+      "loss": 0.3262,
+      "step": 3720
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.1080131208997186e-05,
+      "loss": 0.2852,
+      "step": 3730
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.10215557638238e-05,
+      "loss": 0.3505,
+      "step": 3740
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.096298031865043e-05,
+      "loss": 0.2781,
+      "step": 3750
+    },
+    {
+      "epoch": 1.25,
+      "eval_accuracy": 0.8886029720306396,
+      "eval_loss": 0.30135467648506165,
+      "eval_runtime": 2939.792,
+      "eval_samples_per_second": 4.098,
+      "eval_steps_per_second": 0.064,
+      "step": 3750
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.090440487347704e-05,
+      "loss": 0.3526,
+      "step": 3760
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.0845829428303654e-05,
+      "loss": 0.2841,
+      "step": 3770
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.078725398313027e-05,
+      "loss": 0.2578,
+      "step": 3780
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.072867853795689e-05,
+      "loss": 0.2886,
+      "step": 3790
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.0670103092783506e-05,
+      "loss": 0.2308,
+      "step": 3800
+    },
+    {
+      "epoch": 1.26,
+      "eval_accuracy": 0.8944135308265686,
+      "eval_loss": 0.3107089102268219,
+      "eval_runtime": 2925.7198,
+      "eval_samples_per_second": 4.118,
+      "eval_steps_per_second": 0.065,
+      "step": 3800
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.061152764761012e-05,
+      "loss": 0.3303,
+      "step": 3810
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.055295220243674e-05,
+      "loss": 0.2801,
+      "step": 3820
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.0494376757263354e-05,
+      "loss": 0.2308,
+      "step": 3830
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.0435801312089974e-05,
+      "loss": 0.2462,
+      "step": 3840
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.037722586691659e-05,
+      "loss": 0.2896,
+      "step": 3850
+    },
+    {
+      "epoch": 1.28,
+      "eval_accuracy": 0.8955756425857544,
+      "eval_loss": 0.2939203977584839,
+      "eval_runtime": 2925.3139,
+      "eval_samples_per_second": 4.118,
+      "eval_steps_per_second": 0.065,
+      "step": 3850
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.0318650421743205e-05,
+      "loss": 0.2736,
+      "step": 3860
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.0260074976569825e-05,
+      "loss": 0.2567,
+      "step": 3870
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.020149953139644e-05,
+      "loss": 0.2009,
+      "step": 3880
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.0142924086223056e-05,
+      "loss": 0.2943,
+      "step": 3890
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.0084348641049677e-05,
+      "loss": 0.339,
+      "step": 3900
+    },
+    {
+      "epoch": 1.29,
+      "eval_accuracy": 0.8927533626556396,
+      "eval_loss": 0.29802992939949036,
+      "eval_runtime": 2930.0892,
+      "eval_samples_per_second": 4.111,
+      "eval_steps_per_second": 0.065,
+      "step": 3900
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.002577319587629e-05,
+      "loss": 0.2817,
+      "step": 3910
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.9967197750702907e-05,
+      "loss": 0.3471,
+      "step": 3920
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.990862230552952e-05,
+      "loss": 0.3022,
+      "step": 3930
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.985004686035614e-05,
+      "loss": 0.1821,
+      "step": 3940
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.979147141518276e-05,
+      "loss": 0.3788,
+      "step": 3950
+    },
+    {
+      "epoch": 1.31,
+      "eval_accuracy": 0.8848676085472107,
+      "eval_loss": 0.335862398147583,
+      "eval_runtime": 2890.2656,
+      "eval_samples_per_second": 4.168,
+      "eval_steps_per_second": 0.065,
+      "step": 3950
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.9732895970009372e-05,
+      "loss": 0.2704,
+      "step": 3960
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.9674320524835993e-05,
+      "loss": 0.2548,
+      "step": 3970
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.9615745079662603e-05,
+      "loss": 0.2334,
+      "step": 3980
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.9557169634489224e-05,
+      "loss": 0.2045,
+      "step": 3990
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.949859418931584e-05,
+      "loss": 0.1653,
+      "step": 4000
+    },
+    {
+      "epoch": 1.33,
+      "eval_accuracy": 0.8860297203063965,
+      "eval_loss": 0.3242672383785248,
+      "eval_runtime": 2928.0303,
+      "eval_samples_per_second": 4.114,
+      "eval_steps_per_second": 0.065,
+      "step": 4000
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.9440018744142454e-05,
+      "loss": 0.2394,
+      "step": 4010
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.9381443298969075e-05,
+      "loss": 0.3278,
+      "step": 4020
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.932286785379569e-05,
+      "loss": 0.1747,
+      "step": 4030
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.9264292408622306e-05,
+      "loss": 0.2607,
+      "step": 4040
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.9205716963448926e-05,
+      "loss": 0.3088,
+      "step": 4050
+    },
+    {
+      "epoch": 1.34,
+      "eval_accuracy": 0.8871088027954102,
+      "eval_loss": 0.3149142861366272,
+      "eval_runtime": 2882.8389,
+      "eval_samples_per_second": 4.179,
+      "eval_steps_per_second": 0.066,
+      "step": 4050
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.914714151827554e-05,
+      "loss": 0.2744,
+      "step": 4060
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.9088566073102157e-05,
+      "loss": 0.3001,
+      "step": 4070
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.902999062792877e-05,
+      "loss": 0.2545,
+      "step": 4080
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.897141518275539e-05,
+      "loss": 0.1678,
+      "step": 4090
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.8912839737582008e-05,
+      "loss": 0.4402,
+      "step": 4100
+    },
+    {
+      "epoch": 1.36,
+      "eval_accuracy": 0.8904291391372681,
+      "eval_loss": 0.32059910893440247,
+      "eval_runtime": 2925.63,
+      "eval_samples_per_second": 4.118,
+      "eval_steps_per_second": 0.065,
+      "step": 4100
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.8854264292408622e-05,
+      "loss": 0.3024,
+      "step": 4110
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.8795688847235242e-05,
+      "loss": 0.1949,
+      "step": 4120
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.8737113402061856e-05,
+      "loss": 0.2697,
+      "step": 4130
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.8678537956888473e-05,
+      "loss": 0.1598,
+      "step": 4140
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.8619962511715094e-05,
+      "loss": 0.2302,
+      "step": 4150
+    },
+    {
+      "epoch": 1.38,
+      "eval_accuracy": 0.8888520002365112,
+      "eval_loss": 0.34775903820991516,
+      "eval_runtime": 2941.5064,
+      "eval_samples_per_second": 4.096,
+      "eval_steps_per_second": 0.064,
+      "step": 4150
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.8561387066541707e-05,
+      "loss": 0.2443,
+      "step": 4160
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.8502811621368324e-05,
+      "loss": 0.2784,
+      "step": 4170
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.8444236176194938e-05,
+      "loss": 0.2139,
+      "step": 4180
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.838566073102156e-05,
+      "loss": 0.1431,
+      "step": 4190
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.8327085285848176e-05,
+      "loss": 0.3066,
+      "step": 4200
+    },
+    {
+      "epoch": 1.39,
+      "eval_accuracy": 0.8925043344497681,
+      "eval_loss": 0.3472656011581421,
+      "eval_runtime": 2995.6082,
+      "eval_samples_per_second": 4.022,
+      "eval_steps_per_second": 0.063,
+      "step": 4200
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.826850984067479e-05,
+      "loss": 0.2992,
+      "step": 4210
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.820993439550141e-05,
+      "loss": 0.4127,
+      "step": 4220
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.815135895032802e-05,
+      "loss": 0.1718,
+      "step": 4230
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.809278350515464e-05,
+      "loss": 0.2215,
+      "step": 4240
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.803420805998126e-05,
+      "loss": 0.2434,
+      "step": 4250
+    },
+    {
+      "epoch": 1.41,
+      "eval_accuracy": 0.8939985036849976,
+      "eval_loss": 0.32785314321517944,
+      "eval_runtime": 2930.7504,
+      "eval_samples_per_second": 4.111,
+      "eval_steps_per_second": 0.064,
+      "step": 4250
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.797563261480787e-05,
+      "loss": 0.2215,
+      "step": 4260
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.7917057169634492e-05,
+      "loss": 0.3578,
+      "step": 4270
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.7858481724461106e-05,
+      "loss": 0.2208,
+      "step": 4280
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.7799906279287723e-05,
+      "loss": 0.2271,
+      "step": 4290
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.7741330834114343e-05,
+      "loss": 0.2801,
+      "step": 4300
+    },
+    {
+      "epoch": 1.43,
+      "eval_accuracy": 0.8939154744148254,
+      "eval_loss": 0.32341817021369934,
+      "eval_runtime": 2930.7491,
+      "eval_samples_per_second": 4.111,
+      "eval_steps_per_second": 0.064,
+      "step": 4300
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.7682755388940957e-05,
+      "loss": 0.2672,
+      "step": 4310
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.7624179943767574e-05,
+      "loss": 0.2383,
+      "step": 4320
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.7565604498594188e-05,
+      "loss": 0.2971,
+      "step": 4330
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.7507029053420808e-05,
+      "loss": 0.2431,
+      "step": 4340
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.7448453608247425e-05,
+      "loss": 0.3461,
+      "step": 4350
+    },
+    {
+      "epoch": 1.44,
+      "eval_accuracy": 0.8964887261390686,
+      "eval_loss": 0.2783967852592468,
+      "eval_runtime": 2926.1616,
+      "eval_samples_per_second": 4.117,
+      "eval_steps_per_second": 0.065,
+      "step": 4350
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.738987816307404e-05,
+      "loss": 0.2098,
+      "step": 4360
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.733130271790066e-05,
+      "loss": 0.244,
+      "step": 4370
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.7272727272727273e-05,
+      "loss": 0.2899,
+      "step": 4380
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.721415182755389e-05,
+      "loss": 0.1871,
+      "step": 4390
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.715557638238051e-05,
+      "loss": 0.3233,
+      "step": 4400
+    },
+    {
+      "epoch": 1.46,
+      "eval_accuracy": 0.8915082812309265,
+      "eval_loss": 0.32419413328170776,
+      "eval_runtime": 2925.7579,
+      "eval_samples_per_second": 4.118,
+      "eval_steps_per_second": 0.065,
+      "step": 4400
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.7097000937207124e-05,
+      "loss": 0.2698,
+      "step": 4410
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.703842549203374e-05,
+      "loss": 0.2418,
+      "step": 4420
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.6979850046860355e-05,
+      "loss": 0.312,
+      "step": 4430
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.6921274601686976e-05,
+      "loss": 0.2791,
+      "step": 4440
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.6862699156513593e-05,
+      "loss": 0.1694,
+      "step": 4450
+    },
+    {
+      "epoch": 1.48,
+      "eval_accuracy": 0.8930854201316833,
+      "eval_loss": 0.3108774423599243,
+      "eval_runtime": 2939.9189,
+      "eval_samples_per_second": 4.098,
+      "eval_steps_per_second": 0.064,
+      "step": 4450
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.6804123711340206e-05,
+      "loss": 0.2592,
+      "step": 4460
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.6745548266166827e-05,
+      "loss": 0.1924,
+      "step": 4470
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.6686972820993437e-05,
+      "loss": 0.2456,
+      "step": 4480
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.6628397375820058e-05,
+      "loss": 0.1603,
+      "step": 4490
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.6569821930646678e-05,
+      "loss": 0.2608,
+      "step": 4500
+    },
+    {
+      "epoch": 1.49,
+      "eval_accuracy": 0.8968207836151123,
+      "eval_loss": 0.3301639258861542,
+      "eval_runtime": 2947.6766,
+      "eval_samples_per_second": 4.087,
+      "eval_steps_per_second": 0.064,
+      "step": 4500
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.651124648547329e-05,
+      "loss": 0.2059,
+      "step": 4510
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.645267104029991e-05,
+      "loss": 0.3259,
+      "step": 4520
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.6394095595126523e-05,
+      "loss": 0.2387,
+      "step": 4530
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.633552014995314e-05,
+      "loss": 0.2449,
+      "step": 4540
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.627694470477976e-05,
+      "loss": 0.2651,
+      "step": 4550
+    },
+    {
+      "epoch": 1.51,
+      "eval_accuracy": 0.8920063376426697,
+      "eval_loss": 0.30787232518196106,
+      "eval_runtime": 2968.9355,
+      "eval_samples_per_second": 4.058,
+      "eval_steps_per_second": 0.064,
+      "step": 4550
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.6218369259606374e-05,
+      "loss": 0.2294,
+      "step": 4560
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.615979381443299e-05,
+      "loss": 0.3269,
+      "step": 4570
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.6101218369259605e-05,
+      "loss": 0.2596,
+      "step": 4580
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.6042642924086225e-05,
+      "loss": 0.2729,
+      "step": 4590
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5984067478912842e-05,
+      "loss": 0.2724,
+      "step": 4600
+    },
+    {
+      "epoch": 1.53,
+      "eval_accuracy": 0.8871918320655823,
+      "eval_loss": 0.3105985224246979,
+      "eval_runtime": 2944.8958,
+      "eval_samples_per_second": 4.091,
+      "eval_steps_per_second": 0.064,
+      "step": 4600
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5925492033739456e-05,
+      "loss": 0.2914,
+      "step": 4610
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5866916588566076e-05,
+      "loss": 0.1969,
+      "step": 4620
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.580834114339269e-05,
+      "loss": 0.2184,
+      "step": 4630
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5749765698219307e-05,
+      "loss": 0.3381,
+      "step": 4640
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5691190253045928e-05,
+      "loss": 0.2006,
+      "step": 4650
+    },
+    {
+      "epoch": 1.54,
+      "eval_accuracy": 0.8908441662788391,
+      "eval_loss": 0.2958328127861023,
+      "eval_runtime": 2984.5836,
+      "eval_samples_per_second": 4.036,
+      "eval_steps_per_second": 0.063,
+      "step": 4650
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.563261480787254e-05,
+      "loss": 0.2845,
+      "step": 4660
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.557403936269916e-05,
+      "loss": 0.2758,
+      "step": 4670
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5515463917525772e-05,
+      "loss": 0.2844,
+      "step": 4680
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5456888472352393e-05,
+      "loss": 0.2664,
+      "step": 4690
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.539831302717901e-05,
+      "loss": 0.2121,
+      "step": 4700
+    },
+    {
+      "epoch": 1.56,
+      "eval_accuracy": 0.8965717554092407,
+      "eval_loss": 0.2965669333934784,
+      "eval_runtime": 2939.0813,
+      "eval_samples_per_second": 4.099,
+      "eval_steps_per_second": 0.064,
+      "step": 4700
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5339737582005623e-05,
+      "loss": 0.2384,
+      "step": 4710
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5281162136832244e-05,
+      "loss": 0.3088,
+      "step": 4720
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5222586691658858e-05,
+      "loss": 0.2495,
+      "step": 4730
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5164011246485475e-05,
+      "loss": 0.2194,
+      "step": 4740
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5105435801312095e-05,
+      "loss": 0.2619,
+      "step": 4750
+    },
+    {
+      "epoch": 1.58,
+      "eval_accuracy": 0.8969038128852844,
+      "eval_loss": 0.28781718015670776,
+      "eval_runtime": 2990.3887,
+      "eval_samples_per_second": 4.029,
+      "eval_steps_per_second": 0.063,
+      "step": 4750
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5046860356138705e-05,
+      "loss": 0.2988,
+      "step": 4760
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.4988284910965323e-05,
+      "loss": 0.1826,
+      "step": 4770
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.4929709465791943e-05,
+      "loss": 0.2966,
+      "step": 4780
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.4871134020618557e-05,
+      "loss": 0.2689,
+      "step": 4790
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.4812558575445174e-05,
+      "loss": 0.2,
+      "step": 4800
+    },
+    {
+      "epoch": 1.59,
+      "eval_accuracy": 0.8967377543449402,
+      "eval_loss": 0.2746070623397827,
+      "eval_runtime": 3059.3436,
+      "eval_samples_per_second": 3.938,
+      "eval_steps_per_second": 0.062,
+      "step": 4800
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.475398313027179e-05,
+      "loss": 0.2848,
+      "step": 4810
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.4695407685098408e-05,
+      "loss": 0.2542,
+      "step": 4820
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.4636832239925025e-05,
+      "loss": 0.1774,
+      "step": 4830
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.4578256794751642e-05,
+      "loss": 0.2458,
+      "step": 4840
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.451968134957826e-05,
+      "loss": 0.2847,
+      "step": 4850
+    },
+    {
+      "epoch": 1.61,
+      "eval_accuracy": 0.8977338671684265,
+      "eval_loss": 0.3023954927921295,
+      "eval_runtime": 3022.6022,
+      "eval_samples_per_second": 3.986,
+      "eval_steps_per_second": 0.063,
+      "step": 4850
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.4461105904404873e-05,
+      "loss": 0.2877,
+      "step": 4860
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.440253045923149e-05,
+      "loss": 0.2538,
+      "step": 4870
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.434395501405811e-05,
+      "loss": 0.2158,
+      "step": 4880
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.4285379568884724e-05,
+      "loss": 0.2689,
+      "step": 4890
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.422680412371134e-05,
+      "loss": 0.1995,
+      "step": 4900
+    },
+    {
+      "epoch": 1.63,
+      "eval_accuracy": 0.8991450071334839,
+      "eval_loss": 0.2890797555446625,
+      "eval_runtime": 2941.9008,
+      "eval_samples_per_second": 4.095,
+      "eval_steps_per_second": 0.064,
+      "step": 4900
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.416822867853796e-05,
+      "loss": 0.1642,
+      "step": 4910
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.4109653233364572e-05,
+      "loss": 0.2836,
+      "step": 4920
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.4051077788191193e-05,
+      "loss": 0.203,
+      "step": 4930
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.399250234301781e-05,
+      "loss": 0.2255,
+      "step": 4940
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.3933926897844423e-05,
+      "loss": 0.1943,
+      "step": 4950
+    },
+    {
+      "epoch": 1.64,
+      "eval_accuracy": 0.8883539438247681,
+      "eval_loss": 0.36784863471984863,
+      "eval_runtime": 2935.7382,
+      "eval_samples_per_second": 4.104,
+      "eval_steps_per_second": 0.064,
+      "step": 4950
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.387535145267104e-05,
+      "loss": 0.2346,
+      "step": 4960
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.3816776007497658e-05,
+      "loss": 0.3136,
+      "step": 4970
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.3758200562324275e-05,
+      "loss": 0.2796,
+      "step": 4980
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.3699625117150892e-05,
+      "loss": 0.2094,
+      "step": 4990
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.364104967197751e-05,
+      "loss": 0.2818,
+      "step": 5000
+    },
+    {
+      "epoch": 1.66,
+      "eval_accuracy": 0.897152841091156,
+      "eval_loss": 0.3128949999809265,
+      "eval_runtime": 2934.3113,
+      "eval_samples_per_second": 4.106,
+      "eval_steps_per_second": 0.064,
+      "step": 5000
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.3582474226804126e-05,
+      "loss": 0.3021,
+      "step": 5010
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.352389878163074e-05,
+      "loss": 0.1889,
+      "step": 5020
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.346532333645736e-05,
+      "loss": 0.176,
+      "step": 5030
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.3406747891283977e-05,
+      "loss": 0.1692,
+      "step": 5040
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.334817244611059e-05,
+      "loss": 0.2598,
+      "step": 5050
+    },
+    {
+      "epoch": 1.68,
+      "eval_accuracy": 0.8980659246444702,
+      "eval_loss": 0.3228696286678314,
+      "eval_runtime": 2867.793,
+      "eval_samples_per_second": 4.201,
+      "eval_steps_per_second": 0.066,
+      "step": 5050
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.3289597000937208e-05,
+      "loss": 0.3448,
+      "step": 5060
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.3231021555763825e-05,
+      "loss": 0.2667,
+      "step": 5070
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.3172446110590442e-05,
+      "loss": 0.2407,
+      "step": 5080
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.311387066541706e-05,
+      "loss": 0.2999,
+      "step": 5090
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.3055295220243676e-05,
+      "loss": 0.3133,
+      "step": 5100
+    },
+    {
+      "epoch": 1.69,
+      "eval_accuracy": 0.8896820545196533,
+      "eval_loss": 0.3091619908809662,
+      "eval_runtime": 2938.1739,
+      "eval_samples_per_second": 4.1,
+      "eval_steps_per_second": 0.064,
+      "step": 5100
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.299671977507029e-05,
+      "loss": 0.1843,
+      "step": 5110
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.2938144329896907e-05,
+      "loss": 0.3259,
+      "step": 5120
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.2879568884723528e-05,
+      "loss": 0.294,
+      "step": 5130
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.282099343955014e-05,
+      "loss": 0.2551,
+      "step": 5140
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.276241799437676e-05,
+      "loss": 0.1876,
+      "step": 5150
+    },
+    {
+      "epoch": 1.71,
+      "eval_accuracy": 0.8994770646095276,
+      "eval_loss": 0.26903077960014343,
+      "eval_runtime": 2957.9725,
+      "eval_samples_per_second": 4.073,
+      "eval_steps_per_second": 0.064,
+      "step": 5150
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.2703842549203375e-05,
+      "loss": 0.3053,
+      "step": 5160
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.264526710402999e-05,
+      "loss": 0.1407,
+      "step": 5170
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.258669165885661e-05,
+      "loss": 0.2645,
+      "step": 5180
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.2528116213683227e-05,
+      "loss": 0.2395,
+      "step": 5190
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.246954076850984e-05,
+      "loss": 0.431,
+      "step": 5200
+    },
+    {
+      "epoch": 1.73,
+      "eval_accuracy": 0.8953266143798828,
+      "eval_loss": 0.324379026889801,
+      "eval_runtime": 2916.0243,
+      "eval_samples_per_second": 4.131,
+      "eval_steps_per_second": 0.065,
+      "step": 5200
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.2410965323336457e-05,
+      "loss": 0.2358,
+      "step": 5210
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.2352389878163075e-05,
+      "loss": 0.2162,
+      "step": 5220
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.229381443298969e-05,
+      "loss": 0.2704,
+      "step": 5230
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.223523898781631e-05,
+      "loss": 0.2596,
+      "step": 5240
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.2176663542642926e-05,
+      "loss": 0.2305,
+      "step": 5250
+    },
+    {
+      "epoch": 1.74,
+      "eval_accuracy": 0.9003901481628418,
+      "eval_loss": 0.27731630206108093,
+      "eval_runtime": 2979.377,
+      "eval_samples_per_second": 4.043,
+      "eval_steps_per_second": 0.063,
+      "step": 5250
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.2118088097469543e-05,
+      "loss": 0.2945,
+      "step": 5260
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.2059512652296157e-05,
+      "loss": 0.3199,
+      "step": 5270
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.2000937207122777e-05,
+      "loss": 0.2279,
+      "step": 5280
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.1942361761949394e-05,
+      "loss": 0.2088,
+      "step": 5290
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.1883786316776008e-05,
+      "loss": 0.32,
+      "step": 5300
+    },
+    {
+      "epoch": 1.76,
+      "eval_accuracy": 0.8993940353393555,
+      "eval_loss": 0.30973345041275024,
+      "eval_runtime": 2962.0345,
+      "eval_samples_per_second": 4.067,
+      "eval_steps_per_second": 0.064,
+      "step": 5300
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.1825210871602625e-05,
+      "loss": 0.2853,
+      "step": 5310
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.1766635426429242e-05,
+      "loss": 0.1371,
+      "step": 5320
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.170805998125586e-05,
+      "loss": 0.2466,
+      "step": 5330
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.1649484536082476e-05,
+      "loss": 0.2126,
+      "step": 5340
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.1590909090909093e-05,
+      "loss": 0.2632,
+      "step": 5350
+    },
+    {
+      "epoch": 1.78,
+      "eval_accuracy": 0.901303231716156,
+      "eval_loss": 0.28299954533576965,
+      "eval_runtime": 2929.0636,
+      "eval_samples_per_second": 4.113,
+      "eval_steps_per_second": 0.065,
+      "step": 5350
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.1532333645735707e-05,
+      "loss": 0.3151,
+      "step": 5360
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.1473758200562324e-05,
+      "loss": 0.2333,
+      "step": 5370
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1415182755388945e-05,
+      "loss": 0.1944,
+      "step": 5380
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1356607310215558e-05,
+      "loss": 0.3294,
+      "step": 5390
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1298031865042175e-05,
+      "loss": 0.2296,
+      "step": 5400
+    },
+    {
+      "epoch": 1.79,
+      "eval_accuracy": 0.8985639810562134,
+      "eval_loss": 0.2716383934020996,
+      "eval_runtime": 2933.414,
+      "eval_samples_per_second": 4.107,
+      "eval_steps_per_second": 0.064,
+      "step": 5400
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.1239456419868792e-05,
+      "loss": 0.2017,
+      "step": 5410
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.118088097469541e-05,
+      "loss": 0.2426,
+      "step": 5420
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.1122305529522023e-05,
+      "loss": 0.2222,
+      "step": 5430
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.1063730084348644e-05,
+      "loss": 0.2208,
+      "step": 5440
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.1005154639175257e-05,
+      "loss": 0.2276,
+      "step": 5450
+    },
+    {
+      "epoch": 1.81,
+      "eval_accuracy": 0.8921723365783691,
+      "eval_loss": 0.31200098991394043,
+      "eval_runtime": 2927.7023,
+      "eval_samples_per_second": 4.115,
+      "eval_steps_per_second": 0.065,
+      "step": 5450
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.0946579194001875e-05,
+      "loss": 0.1967,
+      "step": 5460
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.088800374882849e-05,
+      "loss": 0.2521,
+      "step": 5470
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.082942830365511e-05,
+      "loss": 0.2899,
+      "step": 5480
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.0770852858481726e-05,
+      "loss": 0.3003,
+      "step": 5490
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.0712277413308343e-05,
+      "loss": 0.237,
+      "step": 5500
+    },
+    {
+      "epoch": 1.83,
+      "eval_accuracy": 0.9008051753044128,
+      "eval_loss": 0.26086243987083435,
+      "eval_runtime": 2933.7335,
+      "eval_samples_per_second": 4.106,
+      "eval_steps_per_second": 0.064,
+      "step": 5500
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.065370196813496e-05,
+      "loss": 0.2551,
+      "step": 5510
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.0595126522961574e-05,
+      "loss": 0.1407,
+      "step": 5520
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.053655107778819e-05,
+      "loss": 0.2094,
+      "step": 5530
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.047797563261481e-05,
+      "loss": 0.2599,
+      "step": 5540
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.0419400187441425e-05,
+      "loss": 0.2505,
+      "step": 5550
+    },
+    {
+      "epoch": 1.84,
+      "eval_accuracy": 0.9030464291572571,
+      "eval_loss": 0.254546195268631,
+      "eval_runtime": 2895.0779,
+      "eval_samples_per_second": 4.161,
+      "eval_steps_per_second": 0.065,
+      "step": 5550
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.0360824742268042e-05,
+      "loss": 0.2148,
+      "step": 5560
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.030224929709466e-05,
+      "loss": 0.1769,
+      "step": 5570
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.0243673851921273e-05,
+      "loss": 0.231,
+      "step": 5580
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.0185098406747893e-05,
+      "loss": 0.1985,
+      "step": 5590
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.012652296157451e-05,
+      "loss": 0.3394,
+      "step": 5600
+    },
+    {
+      "epoch": 1.86,
+      "eval_accuracy": 0.9037104845046997,
+      "eval_loss": 0.29787400364875793,
+      "eval_runtime": 2929.8682,
+      "eval_samples_per_second": 4.112,
+      "eval_steps_per_second": 0.065,
+      "step": 5600
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.0067947516401124e-05,
+      "loss": 0.2696,
+      "step": 5610
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.000937207122774e-05,
+      "loss": 0.2014,
+      "step": 5620
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.9950796626054358e-05,
+      "loss": 0.2366,
+      "step": 5630
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.9892221180880975e-05,
+      "loss": 0.2282,
+      "step": 5640
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.9833645735707592e-05,
+      "loss": 0.2466,
+      "step": 5650
+    },
+    {
+      "epoch": 1.88,
+      "eval_accuracy": 0.9034614562988281,
+      "eval_loss": 0.28928226232528687,
+      "eval_runtime": 2838.2613,
+      "eval_samples_per_second": 4.244,
+      "eval_steps_per_second": 0.067,
+      "step": 5650
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.977507029053421e-05,
+      "loss": 0.2576,
+      "step": 5660
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.9716494845360827e-05,
+      "loss": 0.2305,
+      "step": 5670
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.965791940018744e-05,
+      "loss": 0.2171,
+      "step": 5680
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.959934395501406e-05,
+      "loss": 0.1958,
+      "step": 5690
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.9540768509840678e-05,
+      "loss": 0.1952,
+      "step": 5700
+    },
+    {
+      "epoch": 1.89,
+      "eval_accuracy": 0.9032953977584839,
+      "eval_loss": 0.2843443751335144,
+      "eval_runtime": 2916.8898,
+      "eval_samples_per_second": 4.13,
+      "eval_steps_per_second": 0.065,
+      "step": 5700
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.948219306466729e-05,
+      "loss": 0.173,
+      "step": 5710
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.942361761949391e-05,
+      "loss": 0.2321,
+      "step": 5720
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.9365042174320526e-05,
+      "loss": 0.2239,
+      "step": 5730
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.9306466729147143e-05,
+      "loss": 0.2248,
+      "step": 5740
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.924789128397376e-05,
+      "loss": 0.1808,
+      "step": 5750
+    },
+    {
+      "epoch": 1.91,
+      "eval_accuracy": 0.9003071188926697,
+      "eval_loss": 0.3049992620944977,
+      "eval_runtime": 2979.4474,
+      "eval_samples_per_second": 4.043,
+      "eval_steps_per_second": 0.063,
+      "step": 5750
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.9189315838800377e-05,
+      "loss": 0.2406,
+      "step": 5760
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.913074039362699e-05,
+      "loss": 0.324,
+      "step": 5770
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.9072164948453608e-05,
+      "loss": 0.1952,
+      "step": 5780
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.9013589503280228e-05,
+      "loss": 0.2374,
+      "step": 5790
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.8955014058106842e-05,
+      "loss": 0.1889,
+      "step": 5800
+    },
+    {
+      "epoch": 1.93,
+      "eval_accuracy": 0.8986469507217407,
+      "eval_loss": 0.29144901037216187,
+      "eval_runtime": 2939.5021,
+      "eval_samples_per_second": 4.098,
+      "eval_steps_per_second": 0.064,
+      "step": 5800
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.889643861293346e-05,
+      "loss": 0.1922,
+      "step": 5810
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.8837863167760076e-05,
+      "loss": 0.1673,
+      "step": 5820
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 1.877928772258669e-05,
+      "loss": 0.2481,
+      "step": 5830
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 1.872071227741331e-05,
+      "loss": 0.2624,
+      "step": 5840
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 1.8662136832239927e-05,
+      "loss": 0.267,
+      "step": 5850
+    },
+    {
+      "epoch": 1.94,
+      "eval_accuracy": 0.8944965600967407,
+      "eval_loss": 0.32017406821250916,
+      "eval_runtime": 2923.7537,
+      "eval_samples_per_second": 4.12,
+      "eval_steps_per_second": 0.065,
+      "step": 5850
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 1.860356138706654e-05,
+      "loss": 0.268,
+      "step": 5860
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 1.8544985941893158e-05,
+      "loss": 0.2846,
+      "step": 5870
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 1.8486410496719775e-05,
+      "loss": 0.1709,
+      "step": 5880
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.8427835051546392e-05,
+      "loss": 0.223,
+      "step": 5890
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.836925960637301e-05,
+      "loss": 0.2135,
+      "step": 5900
+    },
+    {
+      "epoch": 1.96,
+      "eval_accuracy": 0.901718258857727,
+      "eval_loss": 0.28145769238471985,
+      "eval_runtime": 3056.4339,
+      "eval_samples_per_second": 3.942,
+      "eval_steps_per_second": 0.062,
+      "step": 5900
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.8310684161199627e-05,
+      "loss": 0.3075,
+      "step": 5910
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.8252108716026244e-05,
+      "loss": 0.2394,
+      "step": 5920
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.8193533270852857e-05,
+      "loss": 0.2749,
+      "step": 5930
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.8134957825679478e-05,
+      "loss": 0.2142,
+      "step": 5940
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.8076382380506095e-05,
+      "loss": 0.1889,
+      "step": 5950
+    },
+    {
+      "epoch": 1.98,
+      "eval_accuracy": 0.9028803706169128,
+      "eval_loss": 0.2930201292037964,
+      "eval_runtime": 3040.6257,
+      "eval_samples_per_second": 3.962,
+      "eval_steps_per_second": 0.062,
+      "step": 5950
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.801780693533271e-05,
+      "loss": 0.2439,
+      "step": 5960
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.7959231490159326e-05,
+      "loss": 0.2177,
+      "step": 5970
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.7900656044985943e-05,
+      "loss": 0.2013,
+      "step": 5980
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.784208059981256e-05,
+      "loss": 0.2064,
+      "step": 5990
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.7783505154639177e-05,
+      "loss": 0.2761,
+      "step": 6000
+    },
+    {
+      "epoch": 1.99,
+      "eval_accuracy": 0.9062007069587708,
+      "eval_loss": 0.28722548484802246,
+      "eval_runtime": 3021.7697,
+      "eval_samples_per_second": 3.987,
+      "eval_steps_per_second": 0.063,
+      "step": 6000
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.7724929709465794e-05,
+      "loss": 0.2081,
+      "step": 6010
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.7666354264292408e-05,
+      "loss": 0.1955,
+      "step": 6020
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.7607778819119025e-05,
+      "loss": 0.2395,
+      "step": 6030
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 1.7549203373945645e-05,
+      "loss": 0.1571,
+      "step": 6040
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 1.749062792877226e-05,
+      "loss": 0.0578,
+      "step": 6050
+    },
+    {
+      "epoch": 2.01,
+      "eval_accuracy": 0.905453622341156,
+      "eval_loss": 0.32240164279937744,
+      "eval_runtime": 3073.9569,
+      "eval_samples_per_second": 3.919,
+      "eval_steps_per_second": 0.061,
+      "step": 6050
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 1.7432052483598876e-05,
+      "loss": 0.1371,
+      "step": 6060
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 1.7373477038425493e-05,
+      "loss": 0.1411,
+      "step": 6070
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 1.731490159325211e-05,
+      "loss": 0.3104,
+      "step": 6080
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 1.7256326148078727e-05,
+      "loss": 0.3536,
+      "step": 6090
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 1.7197750702905344e-05,
+      "loss": 0.2053,
+      "step": 6100
+    },
+    {
+      "epoch": 2.03,
+      "eval_accuracy": 0.905453622341156,
+      "eval_loss": 0.341753751039505,
+      "eval_runtime": 2956.8334,
+      "eval_samples_per_second": 4.074,
+      "eval_steps_per_second": 0.064,
+      "step": 6100
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 1.7139175257731958e-05,
+      "loss": 0.1298,
+      "step": 6110
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 1.7080599812558575e-05,
+      "loss": 0.1208,
+      "step": 6120
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 1.7022024367385192e-05,
+      "loss": 0.1429,
+      "step": 6130
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 1.696344892221181e-05,
+      "loss": 0.1677,
+      "step": 6140
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 1.6904873477038426e-05,
+      "loss": 0.146,
+      "step": 6150
+    },
+    {
+      "epoch": 2.04,
+      "eval_accuracy": 0.9042085409164429,
+      "eval_loss": 0.3489157259464264,
+      "eval_runtime": 2926.9095,
+      "eval_samples_per_second": 4.116,
+      "eval_steps_per_second": 0.065,
+      "step": 6150
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 1.6846298031865044e-05,
+      "loss": 0.215,
+      "step": 6160
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 1.678772258669166e-05,
+      "loss": 0.1922,
+      "step": 6170
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 1.6729147141518274e-05,
+      "loss": 0.2388,
+      "step": 6180
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 1.6670571696344895e-05,
+      "loss": 0.1186,
+      "step": 6190
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 1.6611996251171512e-05,
+      "loss": 0.2463,
+      "step": 6200
+    },
+    {
+      "epoch": 2.06,
+      "eval_accuracy": 0.9061176776885986,
+      "eval_loss": 0.32390135526657104,
+      "eval_runtime": 2930.7518,
+      "eval_samples_per_second": 4.111,
+      "eval_steps_per_second": 0.064,
+      "step": 6200
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 1.6553420805998126e-05,
+      "loss": 0.1714,
+      "step": 6210
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 1.6494845360824743e-05,
+      "loss": 0.1604,
+      "step": 6220
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 1.643626991565136e-05,
+      "loss": 0.0988,
+      "step": 6230
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 1.6377694470477977e-05,
+      "loss": 0.155,
+      "step": 6240
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 1.6319119025304594e-05,
+      "loss": 0.2154,
+      "step": 6250
+    },
+    {
+      "epoch": 2.08,
+      "eval_accuracy": 0.9044575691223145,
+      "eval_loss": 0.3351038992404938,
+      "eval_runtime": 2927.1842,
+      "eval_samples_per_second": 4.116,
+      "eval_steps_per_second": 0.065,
+      "step": 6250
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 1.626054358013121e-05,
+      "loss": 0.0297,
+      "step": 6260
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 1.6201968134957825e-05,
+      "loss": 0.1048,
+      "step": 6270
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 1.6143392689784442e-05,
+      "loss": 0.212,
+      "step": 6280
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 1.608481724461106e-05,
+      "loss": 0.2283,
+      "step": 6290
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 1.6026241799437676e-05,
+      "loss": 0.1476,
+      "step": 6300
+    },
+    {
+      "epoch": 2.09,
+      "eval_accuracy": 0.9061176776885986,
+      "eval_loss": 0.3789742588996887,
+      "eval_runtime": 2919.2007,
+      "eval_samples_per_second": 4.127,
+      "eval_steps_per_second": 0.065,
+      "step": 6300
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 1.5967666354264293e-05,
+      "loss": 0.2079,
+      "step": 6310
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 1.590909090909091e-05,
+      "loss": 0.2238,
+      "step": 6320
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 1.5850515463917527e-05,
+      "loss": 0.2421,
+      "step": 6330
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 1.579194001874414e-05,
+      "loss": 0.0967,
+      "step": 6340
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 1.573336457357076e-05,
+      "loss": 0.246,
+      "step": 6350
+    },
+    {
+      "epoch": 2.11,
+      "eval_accuracy": 0.9053706526756287,
+      "eval_loss": 0.3258315622806549,
+      "eval_runtime": 2928.8044,
+      "eval_samples_per_second": 4.113,
+      "eval_steps_per_second": 0.065,
+      "step": 6350
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 1.567478912839738e-05,
+      "loss": 0.2457,
+      "step": 6360
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 1.5616213683223992e-05,
+      "loss": 0.1559,
+      "step": 6370
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 1.555763823805061e-05,
+      "loss": 0.1161,
+      "step": 6380
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 1.5499062792877226e-05,
+      "loss": 0.2317,
+      "step": 6390
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 1.5440487347703843e-05,
+      "loss": 0.1515,
+      "step": 6400
+    },
+    {
+      "epoch": 2.12,
+      "eval_accuracy": 0.9027974009513855,
+      "eval_loss": 0.3141821324825287,
+      "eval_runtime": 2924.2509,
+      "eval_samples_per_second": 4.12,
+      "eval_steps_per_second": 0.065,
+      "step": 6400
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 1.538191190253046e-05,
+      "loss": 0.086,
+      "step": 6410
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 1.5323336457357078e-05,
+      "loss": 0.1537,
+      "step": 6420
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 1.526476101218369e-05,
+      "loss": 0.1267,
+      "step": 6430
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 1.5206185567010308e-05,
+      "loss": 0.0861,
+      "step": 6440
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 1.5147610121836927e-05,
+      "loss": 0.2135,
+      "step": 6450
+    },
+    {
+      "epoch": 2.14,
+      "eval_accuracy": 0.9001411199569702,
+      "eval_loss": 0.3778562545776367,
+      "eval_runtime": 2929.0135,
+      "eval_samples_per_second": 4.113,
+      "eval_steps_per_second": 0.065,
+      "step": 6450
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 1.5089034676663544e-05,
+      "loss": 0.1662,
+      "step": 6460
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 1.503045923149016e-05,
+      "loss": 0.0579,
+      "step": 6470
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 1.4971883786316777e-05,
+      "loss": 0.1741,
+      "step": 6480
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 1.4913308341143392e-05,
+      "loss": 0.1987,
+      "step": 6490
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 1.4854732895970011e-05,
+      "loss": 0.2403,
+      "step": 6500
+    },
+    {
+      "epoch": 2.16,
+      "eval_accuracy": 0.9042085409164429,
+      "eval_loss": 0.3711290657520294,
+      "eval_runtime": 2995.3297,
+      "eval_samples_per_second": 4.022,
+      "eval_steps_per_second": 0.063,
+      "step": 6500
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 1.4796157450796626e-05,
+      "loss": 0.1528,
+      "step": 6510
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 1.4737582005623243e-05,
+      "loss": 0.1333,
+      "step": 6520
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 1.4679006560449859e-05,
+      "loss": 0.2565,
+      "step": 6530
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 1.4620431115276476e-05,
+      "loss": 0.1425,
+      "step": 6540
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 1.4561855670103095e-05,
+      "loss": 0.1042,
+      "step": 6550
+    },
+    {
+      "epoch": 2.17,
+      "eval_accuracy": 0.9028803706169128,
+      "eval_loss": 0.3553823232650757,
+      "eval_runtime": 3025.6018,
+      "eval_samples_per_second": 3.982,
+      "eval_steps_per_second": 0.062,
+      "step": 6550
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 1.450328022492971e-05,
+      "loss": 0.1155,
+      "step": 6560
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 1.4444704779756327e-05,
+      "loss": 0.1392,
+      "step": 6570
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 1.4386129334582943e-05,
+      "loss": 0.1684,
+      "step": 6580
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 1.432755388940956e-05,
+      "loss": 0.1048,
+      "step": 6590
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 1.4268978444236178e-05,
+      "loss": 0.1894,
+      "step": 6600
+    },
+    {
+      "epoch": 2.19,
+      "eval_accuracy": 0.8909271955490112,
+      "eval_loss": 0.41722801327705383,
+      "eval_runtime": 3014.6942,
+      "eval_samples_per_second": 3.996,
+      "eval_steps_per_second": 0.063,
+      "step": 6600
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 1.4210402999062794e-05,
+      "loss": 0.201,
+      "step": 6610
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 1.4151827553889411e-05,
+      "loss": 0.1758,
+      "step": 6620
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 1.4093252108716026e-05,
+      "loss": 0.1584,
+      "step": 6630
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 1.4034676663542642e-05,
+      "loss": 0.149,
+      "step": 6640
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 1.397610121836926e-05,
+      "loss": 0.1719,
+      "step": 6650
+    },
+    {
+      "epoch": 2.21,
+      "eval_accuracy": 0.9047895669937134,
+      "eval_loss": 0.35078778862953186,
+      "eval_runtime": 2997.7202,
+      "eval_samples_per_second": 4.019,
+      "eval_steps_per_second": 0.063,
+      "step": 6650
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 1.3917525773195878e-05,
+      "loss": 0.2618,
+      "step": 6660
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 1.3858950328022493e-05,
+      "loss": 0.1631,
+      "step": 6670
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 1.380037488284911e-05,
+      "loss": 0.1594,
+      "step": 6680
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 1.3741799437675725e-05,
+      "loss": 0.1883,
+      "step": 6690
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 1.3683223992502344e-05,
+      "loss": 0.1667,
+      "step": 6700
+    },
+    {
+      "epoch": 2.22,
+      "eval_accuracy": 0.9044575691223145,
+      "eval_loss": 0.3378527760505676,
+      "eval_runtime": 3030.8668,
+      "eval_samples_per_second": 3.975,
+      "eval_steps_per_second": 0.062,
+      "step": 6700
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 1.3624648547328961e-05,
+      "loss": 0.1216,
+      "step": 6710
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 1.3566073102155577e-05,
+      "loss": 0.2331,
+      "step": 6720
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 1.3507497656982194e-05,
+      "loss": 0.1457,
+      "step": 6730
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 1.344892221180881e-05,
+      "loss": 0.2942,
+      "step": 6740
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 1.3390346766635428e-05,
+      "loss": 0.2224,
+      "step": 6750
+    },
+    {
+      "epoch": 2.24,
+      "eval_accuracy": 0.9038764834403992,
+      "eval_loss": 0.3271961510181427,
+      "eval_runtime": 3008.8745,
+      "eval_samples_per_second": 4.004,
+      "eval_steps_per_second": 0.063,
+      "step": 6750
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 1.3331771321462045e-05,
+      "loss": 0.1547,
+      "step": 6760
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 1.327319587628866e-05,
+      "loss": 0.3336,
+      "step": 6770
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 1.3214620431115276e-05,
+      "loss": 0.2491,
+      "step": 6780
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 1.3156044985941893e-05,
+      "loss": 0.289,
+      "step": 6790
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 1.3097469540768512e-05,
+      "loss": 0.1617,
+      "step": 6800
+    },
+    {
+      "epoch": 2.26,
+      "eval_accuracy": 0.9041255116462708,
+      "eval_loss": 0.3173203468322754,
+      "eval_runtime": 2945.8188,
+      "eval_samples_per_second": 4.09,
+      "eval_steps_per_second": 0.064,
+      "step": 6800
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 1.3038894095595127e-05,
+      "loss": 0.1473,
+      "step": 6810
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 1.2980318650421744e-05,
+      "loss": 0.1916,
+      "step": 6820
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 1.292174320524836e-05,
+      "loss": 0.1805,
+      "step": 6830
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 1.2863167760074977e-05,
+      "loss": 0.1262,
+      "step": 6840
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 1.2804592314901595e-05,
+      "loss": 0.2179,
+      "step": 6850
+    },
+    {
+      "epoch": 2.27,
+      "eval_accuracy": 0.9020503163337708,
+      "eval_loss": 0.32655876874923706,
+      "eval_runtime": 2909.0561,
+      "eval_samples_per_second": 4.141,
+      "eval_steps_per_second": 0.065,
+      "step": 6850
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 1.2746016869728211e-05,
+      "loss": 0.2169,
+      "step": 6860
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 1.2687441424554828e-05,
+      "loss": 0.1838,
+      "step": 6870
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 1.2628865979381443e-05,
+      "loss": 0.0674,
+      "step": 6880
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 1.2570290534208059e-05,
+      "loss": 0.2081,
+      "step": 6890
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 1.251171508903468e-05,
+      "loss": 0.1948,
+      "step": 6900
+    },
+    {
+      "epoch": 2.29,
+      "eval_accuracy": 0.9030464291572571,
+      "eval_loss": 0.33025914430618286,
+      "eval_runtime": 2926.2553,
+      "eval_samples_per_second": 4.117,
+      "eval_steps_per_second": 0.065,
+      "step": 6900
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 1.2453139643861295e-05,
+      "loss": 0.1809,
+      "step": 6910
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 1.239456419868791e-05,
+      "loss": 0.1197,
+      "step": 6920
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 1.2335988753514527e-05,
+      "loss": 0.1115,
+      "step": 6930
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 1.2277413308341144e-05,
+      "loss": 0.2005,
+      "step": 6940
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 1.2218837863167761e-05,
+      "loss": 0.2305,
+      "step": 6950
+    },
+    {
+      "epoch": 2.31,
+      "eval_accuracy": 0.9047895669937134,
+      "eval_loss": 0.3400987684726715,
+      "eval_runtime": 2976.0412,
+      "eval_samples_per_second": 4.048,
+      "eval_steps_per_second": 0.064,
+      "step": 6950
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 1.2160262417994377e-05,
+      "loss": 0.1895,
+      "step": 6960
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 1.2101686972820994e-05,
+      "loss": 0.1343,
+      "step": 6970
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 1.204311152764761e-05,
+      "loss": 0.1792,
+      "step": 6980
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 1.1984536082474228e-05,
+      "loss": 0.1439,
+      "step": 6990
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 1.1925960637300843e-05,
+      "loss": 0.1625,
+      "step": 7000
+    },
+    {
+      "epoch": 2.32,
+      "eval_accuracy": 0.8987299799919128,
+      "eval_loss": 0.35047200322151184,
+      "eval_runtime": 2974.9568,
+      "eval_samples_per_second": 4.049,
+      "eval_steps_per_second": 0.064,
+      "step": 7000
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 1.186738519212746e-05,
+      "loss": 0.1275,
+      "step": 7010
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 1.1808809746954077e-05,
+      "loss": 0.1262,
+      "step": 7020
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 1.1750234301780695e-05,
+      "loss": 0.2174,
+      "step": 7030
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 1.1691658856607312e-05,
+      "loss": 0.1786,
+      "step": 7040
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 1.1633083411433927e-05,
+      "loss": 0.1483,
+      "step": 7050
+    },
+    {
+      "epoch": 2.34,
+      "eval_accuracy": 0.9062007069587708,
+      "eval_loss": 0.3373454511165619,
+      "eval_runtime": 3023.9939,
+      "eval_samples_per_second": 3.984,
+      "eval_steps_per_second": 0.063,
+      "step": 7050
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 1.1574507966260544e-05,
+      "loss": 0.0845,
+      "step": 7060
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 1.1515932521087161e-05,
+      "loss": 0.0877,
+      "step": 7070
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 1.1457357075913777e-05,
+      "loss": 0.3632,
+      "step": 7080
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 1.1398781630740395e-05,
+      "loss": 0.1592,
+      "step": 7090
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 1.134020618556701e-05,
+      "loss": 0.2008,
+      "step": 7100
+    },
+    {
+      "epoch": 2.36,
+      "eval_accuracy": 0.9058687090873718,
+      "eval_loss": 0.3449901044368744,
+      "eval_runtime": 2922.6824,
+      "eval_samples_per_second": 4.122,
+      "eval_steps_per_second": 0.065,
+      "step": 7100
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 1.1281630740393626e-05,
+      "loss": 0.1318,
+      "step": 7110
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 1.1223055295220245e-05,
+      "loss": 0.0926,
+      "step": 7120
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 1.116447985004686e-05,
+      "loss": 0.1249,
+      "step": 7130
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 1.1105904404873477e-05,
+      "loss": 0.1599,
+      "step": 7140
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 1.1047328959700095e-05,
+      "loss": 0.157,
+      "step": 7150
+    },
+    {
+      "epoch": 2.37,
+      "eval_accuracy": 0.9022993445396423,
+      "eval_loss": 0.36756283044815063,
+      "eval_runtime": 2931.9116,
+      "eval_samples_per_second": 4.109,
+      "eval_steps_per_second": 0.064,
+      "step": 7150
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 1.098875351452671e-05,
+      "loss": 0.2067,
+      "step": 7160
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 1.0930178069353329e-05,
+      "loss": 0.1858,
+      "step": 7170
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 1.0871602624179944e-05,
+      "loss": 0.1462,
+      "step": 7180
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.0813027179006561e-05,
+      "loss": 0.0693,
+      "step": 7190
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.0754451733833178e-05,
+      "loss": 0.1636,
+      "step": 7200
+    },
+    {
+      "epoch": 2.39,
+      "eval_accuracy": 0.9042915105819702,
+      "eval_loss": 0.3568061292171478,
+      "eval_runtime": 2926.2418,
+      "eval_samples_per_second": 4.117,
+      "eval_steps_per_second": 0.065,
+      "step": 7200
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.0695876288659794e-05,
+      "loss": 0.2989,
+      "step": 7210
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.063730084348641e-05,
+      "loss": 0.0984,
+      "step": 7220
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.0578725398313028e-05,
+      "loss": 0.2496,
+      "step": 7230
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.0520149953139645e-05,
+      "loss": 0.3308,
+      "step": 7240
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.046157450796626e-05,
+      "loss": 0.1587,
+      "step": 7250
+    },
+    {
+      "epoch": 2.41,
+      "eval_accuracy": 0.9057026505470276,
+      "eval_loss": 0.32082468271255493,
+      "eval_runtime": 2922.0631,
+      "eval_samples_per_second": 4.123,
+      "eval_steps_per_second": 0.065,
+      "step": 7250
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.0402999062792877e-05,
+      "loss": 0.1226,
+      "step": 7260
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.0344423617619495e-05,
+      "loss": 0.1719,
+      "step": 7270
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.0285848172446112e-05,
+      "loss": 0.2171,
+      "step": 7280
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.0227272727272729e-05,
+      "loss": 0.1593,
+      "step": 7290
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.0168697282099344e-05,
+      "loss": 0.1547,
+      "step": 7300
+    },
+    {
+      "epoch": 2.42,
+      "eval_accuracy": 0.9052045941352844,
+      "eval_loss": 0.331007719039917,
+      "eval_runtime": 2909.36,
+      "eval_samples_per_second": 4.141,
+      "eval_steps_per_second": 0.065,
+      "step": 7300
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.0110121836925961e-05,
+      "loss": 0.1314,
+      "step": 7310
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.0051546391752578e-05,
+      "loss": 0.0937,
+      "step": 7320
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 9.992970946579194e-06,
+      "loss": 0.2031,
+      "step": 7330
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 9.934395501405812e-06,
+      "loss": 0.2195,
+      "step": 7340
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 9.875820056232428e-06,
+      "loss": 0.1686,
+      "step": 7350
+    },
+    {
+      "epoch": 2.44,
+      "eval_accuracy": 0.9052876234054565,
+      "eval_loss": 0.33452099561691284,
+      "eval_runtime": 2935.6217,
+      "eval_samples_per_second": 4.104,
+      "eval_steps_per_second": 0.064,
+      "step": 7350
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 9.817244611059045e-06,
+      "loss": 0.0445,
+      "step": 7360
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 9.758669165885662e-06,
+      "loss": 0.0765,
+      "step": 7370
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 9.700093720712277e-06,
+      "loss": 0.2234,
+      "step": 7380
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 9.641518275538894e-06,
+      "loss": 0.1938,
+      "step": 7390
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 9.582942830365512e-06,
+      "loss": 0.0898,
+      "step": 7400
+    },
+    {
+      "epoch": 2.46,
+      "eval_accuracy": 0.9061176776885986,
+      "eval_loss": 0.3567478060722351,
+      "eval_runtime": 2922.0881,
+      "eval_samples_per_second": 4.123,
+      "eval_steps_per_second": 0.065,
+      "step": 7400
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 9.524367385192127e-06,
+      "loss": 0.1014,
+      "step": 7410
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 9.465791940018746e-06,
+      "loss": 0.214,
+      "step": 7420
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 9.407216494845361e-06,
+      "loss": 0.209,
+      "step": 7430
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 9.348641049671977e-06,
+      "loss": 0.1669,
+      "step": 7440
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 9.290065604498595e-06,
+      "loss": 0.1869,
+      "step": 7450
+    },
+    {
+      "epoch": 2.47,
+      "eval_accuracy": 0.9061176776885986,
+      "eval_loss": 0.3379247188568115,
+      "eval_runtime": 2935.7141,
+      "eval_samples_per_second": 4.104,
+      "eval_steps_per_second": 0.064,
+      "step": 7450
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 9.23149015932521e-06,
+      "loss": 0.147,
+      "step": 7460
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 9.172914714151828e-06,
+      "loss": 0.137,
+      "step": 7470
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 9.114339268978445e-06,
+      "loss": 0.302,
+      "step": 7480
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 9.05576382380506e-06,
+      "loss": 0.1999,
+      "step": 7490
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 8.997188378631679e-06,
+      "loss": 0.1468,
+      "step": 7500
+    },
+    {
+      "epoch": 2.49,
+      "eval_accuracy": 0.9056196808815002,
+      "eval_loss": 0.33066657185554504,
+      "eval_runtime": 2930.2607,
+      "eval_samples_per_second": 4.111,
+      "eval_steps_per_second": 0.064,
+      "step": 7500
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 8.938612933458294e-06,
+      "loss": 0.1849,
+      "step": 7510
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 8.880037488284912e-06,
+      "loss": 0.2767,
+      "step": 7520
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 8.821462043111529e-06,
+      "loss": 0.2727,
+      "step": 7530
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 8.762886597938144e-06,
+      "loss": 0.1813,
+      "step": 7540
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 8.704311152764761e-06,
+      "loss": 0.1361,
+      "step": 7550
+    },
+    {
+      "epoch": 2.51,
+      "eval_accuracy": 0.9052876234054565,
+      "eval_loss": 0.31330642104148865,
+      "eval_runtime": 2943.2985,
+      "eval_samples_per_second": 4.093,
+      "eval_steps_per_second": 0.064,
+      "step": 7550
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 8.645735707591378e-06,
+      "loss": 0.1222,
+      "step": 7560
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 8.587160262417995e-06,
+      "loss": 0.1212,
+      "step": 7570
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 8.52858481724461e-06,
+      "loss": 0.0897,
+      "step": 7580
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 8.470009372071228e-06,
+      "loss": 0.1459,
+      "step": 7590
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 8.411433926897845e-06,
+      "loss": 0.2228,
+      "step": 7600
+    },
+    {
+      "epoch": 2.52,
+      "eval_accuracy": 0.905453622341156,
+      "eval_loss": 0.3354151248931885,
+      "eval_runtime": 2935.7615,
+      "eval_samples_per_second": 4.104,
+      "eval_steps_per_second": 0.064,
+      "step": 7600
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 8.352858481724462e-06,
+      "loss": 0.1579,
+      "step": 7610
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 8.294283036551079e-06,
+      "loss": 0.2439,
+      "step": 7620
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 8.235707591377694e-06,
+      "loss": 0.1875,
+      "step": 7630
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 8.177132146204312e-06,
+      "loss": 0.1262,
+      "step": 7640
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 8.118556701030929e-06,
+      "loss": 0.0414,
+      "step": 7650
+    },
+    {
+      "epoch": 2.54,
+      "eval_accuracy": 0.9047065377235413,
+      "eval_loss": 0.3257265090942383,
+      "eval_runtime": 2945.036,
+      "eval_samples_per_second": 4.091,
+      "eval_steps_per_second": 0.064,
+      "step": 7650
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 8.059981255857544e-06,
+      "loss": 0.215,
+      "step": 7660
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 8.001405810684163e-06,
+      "loss": 0.135,
+      "step": 7670
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.942830365510778e-06,
+      "loss": 0.1628,
+      "step": 7680
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.884254920337395e-06,
+      "loss": 0.1233,
+      "step": 7690
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.825679475164012e-06,
+      "loss": 0.2026,
+      "step": 7700
+    },
+    {
+      "epoch": 2.56,
+      "eval_accuracy": 0.9019672870635986,
+      "eval_loss": 0.34675589203834534,
+      "eval_runtime": 2933.6699,
+      "eval_samples_per_second": 4.106,
+      "eval_steps_per_second": 0.064,
+      "step": 7700
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.767104029990628e-06,
+      "loss": 0.1428,
+      "step": 7710
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.708528584817245e-06,
+      "loss": 0.2086,
+      "step": 7720
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.649953139643862e-06,
+      "loss": 0.1112,
+      "step": 7730
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.591377694470478e-06,
+      "loss": 0.1809,
+      "step": 7740
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.532802249297095e-06,
+      "loss": 0.1686,
+      "step": 7750
+    },
+    {
+      "epoch": 2.57,
+      "eval_accuracy": 0.9063667058944702,
+      "eval_loss": 0.3348318338394165,
+      "eval_runtime": 2938.6603,
+      "eval_samples_per_second": 4.099,
+      "eval_steps_per_second": 0.064,
+      "step": 7750
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.4742268041237115e-06,
+      "loss": 0.227,
+      "step": 7760
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.415651358950329e-06,
+      "loss": 0.1678,
+      "step": 7770
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.357075913776945e-06,
+      "loss": 0.2293,
+      "step": 7780
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.298500468603561e-06,
+      "loss": 0.2007,
+      "step": 7790
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.239925023430179e-06,
+      "loss": 0.204,
+      "step": 7800
+    },
+    {
+      "epoch": 2.59,
+      "eval_accuracy": 0.9064497351646423,
+      "eval_loss": 0.3327842950820923,
+      "eval_runtime": 2934.137,
+      "eval_samples_per_second": 4.106,
+      "eval_steps_per_second": 0.064,
+      "step": 7800
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.181349578256795e-06,
+      "loss": 0.1153,
+      "step": 7810
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.122774133083412e-06,
+      "loss": 0.2967,
+      "step": 7820
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.0641986879100285e-06,
+      "loss": 0.1436,
+      "step": 7830
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.005623242736645e-06,
+      "loss": 0.1987,
+      "step": 7840
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 6.947047797563262e-06,
+      "loss": 0.1569,
+      "step": 7850
+    },
+    {
+      "epoch": 2.61,
+      "eval_accuracy": 0.9070308208465576,
+      "eval_loss": 0.335011750459671,
+      "eval_runtime": 3009.9385,
+      "eval_samples_per_second": 4.002,
+      "eval_steps_per_second": 0.063,
+      "step": 7850
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 6.888472352389878e-06,
+      "loss": 0.1791,
+      "step": 7860
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 6.829896907216494e-06,
+      "loss": 0.1888,
+      "step": 7870
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 6.771321462043112e-06,
+      "loss": 0.2497,
+      "step": 7880
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 6.7127460168697285e-06,
+      "loss": 0.1993,
+      "step": 7890
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 6.654170571696346e-06,
+      "loss": 0.0918,
+      "step": 7900
+    },
+    {
+      "epoch": 2.62,
+      "eval_accuracy": 0.9042085409164429,
+      "eval_loss": 0.32580670714378357,
+      "eval_runtime": 3008.2052,
+      "eval_samples_per_second": 4.005,
+      "eval_steps_per_second": 0.063,
+      "step": 7900
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 6.595595126522962e-06,
+      "loss": 0.1994,
+      "step": 7910
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 6.537019681349578e-06,
+      "loss": 0.2249,
+      "step": 7920
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 6.478444236176195e-06,
+      "loss": 0.1242,
+      "step": 7930
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.419868791002811e-06,
+      "loss": 0.1836,
+      "step": 7940
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.361293345829429e-06,
+      "loss": 0.1938,
+      "step": 7950
+    },
+    {
+      "epoch": 2.64,
+      "eval_accuracy": 0.9063667058944702,
+      "eval_loss": 0.32377755641937256,
+      "eval_runtime": 3036.3884,
+      "eval_samples_per_second": 3.968,
+      "eval_steps_per_second": 0.062,
+      "step": 7950
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.302717900656046e-06,
+      "loss": 0.1902,
+      "step": 7960
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.244142455482662e-06,
+      "loss": 0.1663,
+      "step": 7970
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.185567010309279e-06,
+      "loss": 0.1357,
+      "step": 7980
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.126991565135895e-06,
+      "loss": 0.1478,
+      "step": 7990
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.068416119962512e-06,
+      "loss": 0.2395,
+      "step": 8000
+    },
+    {
+      "epoch": 2.66,
+      "eval_accuracy": 0.9052045941352844,
+      "eval_loss": 0.3201657235622406,
+      "eval_runtime": 2976.8512,
+      "eval_samples_per_second": 4.047,
+      "eval_steps_per_second": 0.063,
+      "step": 8000
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.0098406747891285e-06,
+      "loss": 0.1911,
+      "step": 8010
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 5.951265229615746e-06,
+      "loss": 0.085,
+      "step": 8020
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.892689784442363e-06,
+      "loss": 0.1792,
+      "step": 8030
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.834114339268978e-06,
+      "loss": 0.2266,
+      "step": 8040
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.775538894095595e-06,
+      "loss": 0.0698,
+      "step": 8050
+    },
+    {
+      "epoch": 2.67,
+      "eval_accuracy": 0.9039595127105713,
+      "eval_loss": 0.3339182734489441,
+      "eval_runtime": 2973.4075,
+      "eval_samples_per_second": 4.052,
+      "eval_steps_per_second": 0.064,
+      "step": 8050
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.716963448922212e-06,
+      "loss": 0.0918,
+      "step": 8060
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.658388003748829e-06,
+      "loss": 0.1287,
+      "step": 8070
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.5998125585754456e-06,
+      "loss": 0.2152,
+      "step": 8080
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.541237113402062e-06,
+      "loss": 0.124,
+      "step": 8090
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.482661668228679e-06,
+      "loss": 0.1848,
+      "step": 8100
+    },
+    {
+      "epoch": 2.69,
+      "eval_accuracy": 0.9032124280929565,
+      "eval_loss": 0.35521620512008667,
+      "eval_runtime": 3043.2708,
+      "eval_samples_per_second": 3.959,
+      "eval_steps_per_second": 0.062,
+      "step": 8100
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.424086223055295e-06,
+      "loss": 0.1448,
+      "step": 8110
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.365510777881912e-06,
+      "loss": 0.2439,
+      "step": 8120
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.306935332708529e-06,
+      "loss": 0.236,
+      "step": 8130
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.2483598875351456e-06,
+      "loss": 0.2276,
+      "step": 8140
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 5.189784442361762e-06,
+      "loss": 0.2527,
+      "step": 8150
+    },
+    {
+      "epoch": 2.71,
+      "eval_accuracy": 0.9052045941352844,
+      "eval_loss": 0.33176401257514954,
+      "eval_runtime": 2970.8031,
+      "eval_samples_per_second": 4.055,
+      "eval_steps_per_second": 0.064,
+      "step": 8150
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 5.131208997188379e-06,
+      "loss": 0.1179,
+      "step": 8160
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 5.072633552014996e-06,
+      "loss": 0.0792,
+      "step": 8170
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 5.014058106841612e-06,
+      "loss": 0.1402,
+      "step": 8180
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.9554826616682284e-06,
+      "loss": 0.2519,
+      "step": 8190
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.8969072164948455e-06,
+      "loss": 0.1749,
+      "step": 8200
+    },
+    {
+      "epoch": 2.72,
+      "eval_accuracy": 0.903793454170227,
+      "eval_loss": 0.34064415097236633,
+      "eval_runtime": 3033.3872,
+      "eval_samples_per_second": 3.971,
+      "eval_steps_per_second": 0.062,
+      "step": 8200
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.838331771321463e-06,
+      "loss": 0.1474,
+      "step": 8210
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.779756326148079e-06,
+      "loss": 0.161,
+      "step": 8220
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.721180880974695e-06,
+      "loss": 0.161,
+      "step": 8230
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 4.662605435801312e-06,
+      "loss": 0.1951,
+      "step": 8240
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 4.604029990627929e-06,
+      "loss": 0.1782,
+      "step": 8250
+    },
+    {
+      "epoch": 2.74,
+      "eval_accuracy": 0.9065327644348145,
+      "eval_loss": 0.3299373388290405,
+      "eval_runtime": 2913.5612,
+      "eval_samples_per_second": 4.135,
+      "eval_steps_per_second": 0.065,
+      "step": 8250
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 4.5454545454545455e-06,
+      "loss": 0.1494,
+      "step": 8260
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 4.486879100281163e-06,
+      "loss": 0.1059,
+      "step": 8270
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 4.428303655107779e-06,
+      "loss": 0.1523,
+      "step": 8280
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 4.369728209934396e-06,
+      "loss": 0.1697,
+      "step": 8290
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 4.311152764761012e-06,
+      "loss": 0.1278,
+      "step": 8300
+    },
+    {
+      "epoch": 2.76,
+      "eval_accuracy": 0.9081099033355713,
+      "eval_loss": 0.3326606750488281,
+      "eval_runtime": 2939.1204,
+      "eval_samples_per_second": 4.099,
+      "eval_steps_per_second": 0.064,
+      "step": 8300
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 4.252577319587629e-06,
+      "loss": 0.2058,
+      "step": 8310
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 4.194001874414246e-06,
+      "loss": 0.1229,
+      "step": 8320
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 4.135426429240863e-06,
+      "loss": 0.1856,
+      "step": 8330
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 4.076850984067479e-06,
+      "loss": 0.14,
+      "step": 8340
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 4.018275538894096e-06,
+      "loss": 0.2235,
+      "step": 8350
+    },
+    {
+      "epoch": 2.77,
+      "eval_accuracy": 0.9040424823760986,
+      "eval_loss": 0.3395898938179016,
+      "eval_runtime": 2930.8362,
+      "eval_samples_per_second": 4.11,
+      "eval_steps_per_second": 0.064,
+      "step": 8350
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 3.959700093720713e-06,
+      "loss": 0.143,
+      "step": 8360
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 3.901124648547329e-06,
+      "loss": 0.1388,
+      "step": 8370
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 3.8425492033739455e-06,
+      "loss": 0.1586,
+      "step": 8380
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 3.7839737582005626e-06,
+      "loss": 0.1379,
+      "step": 8390
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 3.7253983130271792e-06,
+      "loss": 0.1995,
+      "step": 8400
+    },
+    {
+      "epoch": 2.79,
+      "eval_accuracy": 0.9082759022712708,
+      "eval_loss": 0.3357061445713043,
+      "eval_runtime": 2931.3182,
+      "eval_samples_per_second": 4.11,
+      "eval_steps_per_second": 0.064,
+      "step": 8400
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 3.666822867853796e-06,
+      "loss": 0.2437,
+      "step": 8410
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 3.608247422680412e-06,
+      "loss": 0.2284,
+      "step": 8420
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 3.5496719775070292e-06,
+      "loss": 0.2423,
+      "step": 8430
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 3.491096532333646e-06,
+      "loss": 0.2038,
+      "step": 8440
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 3.4325210871602626e-06,
+      "loss": 0.2077,
+      "step": 8450
+    },
+    {
+      "epoch": 2.81,
+      "eval_accuracy": 0.9079439043998718,
+      "eval_loss": 0.3241910934448242,
+      "eval_runtime": 2925.6294,
+      "eval_samples_per_second": 4.118,
+      "eval_steps_per_second": 0.065,
+      "step": 8450
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 3.3739456419868796e-06,
+      "loss": 0.0964,
+      "step": 8460
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 3.3153701968134955e-06,
+      "loss": 0.134,
+      "step": 8470
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 3.2567947516401125e-06,
+      "loss": 0.1819,
+      "step": 8480
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 3.1982193064667292e-06,
+      "loss": 0.1255,
+      "step": 8490
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 3.1396438612933463e-06,
+      "loss": 0.1616,
+      "step": 8500
+    },
+    {
+      "epoch": 2.82,
+      "eval_accuracy": 0.9079439043998718,
+      "eval_loss": 0.32991456985473633,
+      "eval_runtime": 2922.3321,
+      "eval_samples_per_second": 4.122,
+      "eval_steps_per_second": 0.065,
+      "step": 8500
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 3.0810684161199625e-06,
+      "loss": 0.2345,
+      "step": 8510
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 3.0224929709465796e-06,
+      "loss": 0.2142,
+      "step": 8520
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.963917525773196e-06,
+      "loss": 0.1439,
+      "step": 8530
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.9053420805998125e-06,
+      "loss": 0.1242,
+      "step": 8540
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.8467666354264296e-06,
+      "loss": 0.1563,
+      "step": 8550
+    },
+    {
+      "epoch": 2.84,
+      "eval_accuracy": 0.9080268740653992,
+      "eval_loss": 0.32703760266304016,
+      "eval_runtime": 2911.9201,
+      "eval_samples_per_second": 4.137,
+      "eval_steps_per_second": 0.065,
+      "step": 8550
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.788191190253046e-06,
+      "loss": 0.1439,
+      "step": 8560
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.729615745079663e-06,
+      "loss": 0.125,
+      "step": 8570
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.671040299906279e-06,
+      "loss": 0.2587,
+      "step": 8580
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.6124648547328963e-06,
+      "loss": 0.1411,
+      "step": 8590
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.5538894095595125e-06,
+      "loss": 0.152,
+      "step": 8600
+    },
+    {
+      "epoch": 2.86,
+      "eval_accuracy": 0.907528817653656,
+      "eval_loss": 0.3290516138076782,
+      "eval_runtime": 2912.3621,
+      "eval_samples_per_second": 4.137,
+      "eval_steps_per_second": 0.065,
+      "step": 8600
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.4953139643861296e-06,
+      "loss": 0.1454,
+      "step": 8610
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.4367385192127463e-06,
+      "loss": 0.1647,
+      "step": 8620
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.378163074039363e-06,
+      "loss": 0.1442,
+      "step": 8630
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.3195876288659796e-06,
+      "loss": 0.0622,
+      "step": 8640
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.2610121836925963e-06,
+      "loss": 0.2968,
+      "step": 8650
+    },
+    {
+      "epoch": 2.87,
+      "eval_accuracy": 0.9074458479881287,
+      "eval_loss": 0.3356473743915558,
+      "eval_runtime": 2926.168,
+      "eval_samples_per_second": 4.117,
+      "eval_steps_per_second": 0.065,
+      "step": 8650
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.202436738519213e-06,
+      "loss": 0.1639,
+      "step": 8660
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.143861293345829e-06,
+      "loss": 0.2248,
+      "step": 8670
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.0852858481724463e-06,
+      "loss": 0.2802,
+      "step": 8680
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.026710402999063e-06,
+      "loss": 0.1014,
+      "step": 8690
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 1.9681349578256796e-06,
+      "loss": 0.1805,
+      "step": 8700
+    },
+    {
+      "epoch": 2.89,
+      "eval_accuracy": 0.9051216244697571,
+      "eval_loss": 0.3323352634906769,
+      "eval_runtime": 2932.0605,
+      "eval_samples_per_second": 4.109,
+      "eval_steps_per_second": 0.064,
+      "step": 8700
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 1.9095595126522962e-06,
+      "loss": 0.1026,
+      "step": 8710
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 1.8509840674789127e-06,
+      "loss": 0.1766,
+      "step": 8720
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 1.7924086223055296e-06,
+      "loss": 0.1394,
+      "step": 8730
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 1.7338331771321464e-06,
+      "loss": 0.1899,
+      "step": 8740
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 1.675257731958763e-06,
+      "loss": 0.1436,
+      "step": 8750
+    },
+    {
+      "epoch": 2.91,
+      "eval_accuracy": 0.9080268740653992,
+      "eval_loss": 0.3264749348163605,
+      "eval_runtime": 2872.7031,
+      "eval_samples_per_second": 4.194,
+      "eval_steps_per_second": 0.066,
+      "step": 8750
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 1.6166822867853798e-06,
+      "loss": 0.1757,
+      "step": 8760
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 1.5581068416119964e-06,
+      "loss": 0.2008,
+      "step": 8770
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 1.499531396438613e-06,
+      "loss": 0.2105,
+      "step": 8780
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 1.4409559512652298e-06,
+      "loss": 0.1791,
+      "step": 8790
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 1.3823805060918464e-06,
+      "loss": 0.1618,
+      "step": 8800
+    },
+    {
+      "epoch": 2.92,
+      "eval_accuracy": 0.9077778458595276,
+      "eval_loss": 0.32518285512924194,
+      "eval_runtime": 3015.7711,
+      "eval_samples_per_second": 3.995,
+      "eval_steps_per_second": 0.063,
+      "step": 8800
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 1.323805060918463e-06,
+      "loss": 0.1754,
+      "step": 8810
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 1.2652296157450795e-06,
+      "loss": 0.1863,
+      "step": 8820
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 1.2066541705716964e-06,
+      "loss": 0.1553,
+      "step": 8830
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 1.148078725398313e-06,
+      "loss": 0.1781,
+      "step": 8840
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.0895032802249298e-06,
+      "loss": 0.1003,
+      "step": 8850
+    },
+    {
+      "epoch": 2.94,
+      "eval_accuracy": 0.9066987633705139,
+      "eval_loss": 0.32578587532043457,
+      "eval_runtime": 3032.7236,
+      "eval_samples_per_second": 3.972,
+      "eval_steps_per_second": 0.062,
+      "step": 8850
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.0309278350515464e-06,
+      "loss": 0.1603,
+      "step": 8860
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 9.72352389878163e-07,
+      "loss": 0.1526,
+      "step": 8870
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 9.137769447047797e-07,
+      "loss": 0.2109,
+      "step": 8880
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 8.552014995313965e-07,
+      "loss": 0.1855,
+      "step": 8890
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.966260543580132e-07,
+      "loss": 0.1913,
+      "step": 8900
+    },
+    {
+      "epoch": 2.95,
+      "eval_accuracy": 0.9065327644348145,
+      "eval_loss": 0.3263765275478363,
+      "eval_runtime": 2995.9468,
+      "eval_samples_per_second": 4.021,
+      "eval_steps_per_second": 0.063,
+      "step": 8900
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.380506091846298e-07,
+      "loss": 0.1779,
+      "step": 8910
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 6.794751640112465e-07,
+      "loss": 0.1558,
+      "step": 8920
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 6.208997188378633e-07,
+      "loss": 0.1279,
+      "step": 8930
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.623242736644799e-07,
+      "loss": 0.2031,
+      "step": 8940
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.037488284910965e-07,
+      "loss": 0.1735,
+      "step": 8950
+    },
+    {
+      "epoch": 2.97,
+      "eval_accuracy": 0.9068647623062134,
+      "eval_loss": 0.32649698853492737,
+      "eval_runtime": 3074.1842,
+      "eval_samples_per_second": 3.919,
+      "eval_steps_per_second": 0.061,
+      "step": 8950
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.4517338331771327e-07,
+      "loss": 0.1671,
+      "step": 8960
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.8659793814432993e-07,
+      "loss": 0.1658,
+      "step": 8970
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.280224929709466e-07,
+      "loss": 0.1953,
+      "step": 8980
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.6944704779756326e-07,
+      "loss": 0.1551,
+      "step": 8990
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1087160262417998e-07,
+      "loss": 0.1587,
+      "step": 9000
+    },
+    {
+      "epoch": 2.99,
+      "eval_accuracy": 0.9070308208465576,
+      "eval_loss": 0.3263450562953949,
+      "eval_runtime": 2966.6999,
+      "eval_samples_per_second": 4.061,
+      "eval_steps_per_second": 0.064,
+      "step": 9000
+    }
+  ],
+  "max_steps": 9036,
+  "num_train_epochs": 3,
+  "total_flos": 3.788430841706496e+16,
+  "trial_name": null,
+  "trial_params": null
+}