diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,3838 @@
+{
+  "best_metric": 0.7447552447552448,
+  "best_model_checkpoint": "wav2vec2-5Class-train-test-finetune-Medium/checkpoint-2640",
+  "epoch": 237.28813559322035,
+  "eval_steps": 500,
+  "global_step": 3500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.95,
+      "grad_norm": 0.8927629590034485,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 1.6852,
+      "step": 14
+    },
+    {
+      "epoch": 0.95,
+      "eval_accuracy": 0.34265734265734266,
+      "eval_loss": 1.5986738204956055,
+      "eval_runtime": 4.6543,
+      "eval_samples_per_second": 61.448,
+      "eval_steps_per_second": 3.867,
+      "step": 14
+    },
+    {
+      "epoch": 1.97,
+      "grad_norm": 0.6501776576042175,
+      "learning_rate": 2.4857142857142858e-06,
+      "loss": 1.5721,
+      "step": 29
+    },
+    {
+      "epoch": 1.97,
+      "eval_accuracy": 0.34265734265734266,
+      "eval_loss": 1.5975812673568726,
+      "eval_runtime": 4.4155,
+      "eval_samples_per_second": 64.772,
+      "eval_steps_per_second": 4.077,
+      "step": 29
+    },
+    {
+      "epoch": 2.98,
+      "grad_norm": 0.6221457719802856,
+      "learning_rate": 3.771428571428572e-06,
+      "loss": 1.5696,
+      "step": 44
+    },
+    {
+      "epoch": 2.98,
+      "eval_accuracy": 0.34265734265734266,
+      "eval_loss": 1.5957201719284058,
+      "eval_runtime": 4.5928,
+      "eval_samples_per_second": 62.272,
+      "eval_steps_per_second": 3.919,
+      "step": 44
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 1.563844919204712,
+      "learning_rate": 5.057142857142857e-06,
+      "loss": 1.5671,
+      "step": 59
+    },
+    {
+      "epoch": 4.0,
+      "eval_accuracy": 0.3356643356643357,
+      "eval_loss": 1.5932137966156006,
+      "eval_runtime": 4.5483,
+      "eval_samples_per_second": 62.881,
+      "eval_steps_per_second": 3.958,
+      "step": 59
+    },
+    {
+      "epoch": 4.95,
+      "grad_norm": 0.7324579358100891,
+      "learning_rate": 6.257142857142858e-06,
+      "loss": 1.6757,
+      "step": 73
+    },
+    {
+      "epoch": 4.95,
+      "eval_accuracy": 0.32517482517482516,
+      "eval_loss": 1.5901715755462646,
+      "eval_runtime": 4.9986,
+      "eval_samples_per_second": 57.216,
+      "eval_steps_per_second": 3.601,
+      "step": 73
+    },
+    {
+      "epoch": 5.97,
+      "grad_norm": 0.878238320350647,
+      "learning_rate": 7.542857142857144e-06,
+      "loss": 1.5595,
+      "step": 88
+    },
+    {
+      "epoch": 5.97,
+      "eval_accuracy": 0.32167832167832167,
+      "eval_loss": 1.5863642692565918,
+      "eval_runtime": 4.9711,
+      "eval_samples_per_second": 57.532,
+      "eval_steps_per_second": 3.621,
+      "step": 88
+    },
+    {
+      "epoch": 6.98,
+      "grad_norm": 1.4210667610168457,
+      "learning_rate": 8.828571428571429e-06,
+      "loss": 1.5536,
+      "step": 103
+    },
+    {
+      "epoch": 6.98,
+      "eval_accuracy": 0.3181818181818182,
+      "eval_loss": 1.5817956924438477,
+      "eval_runtime": 5.5676,
+      "eval_samples_per_second": 51.368,
+      "eval_steps_per_second": 3.233,
+      "step": 103
+    },
+    {
+      "epoch": 8.0,
+      "grad_norm": 0.49261531233787537,
+      "learning_rate": 1.0114285714285715e-05,
+      "loss": 1.5484,
+      "step": 118
+    },
+    {
+      "epoch": 8.0,
+      "eval_accuracy": 0.3111888111888112,
+      "eval_loss": 1.576475739479065,
+      "eval_runtime": 5.5008,
+      "eval_samples_per_second": 51.992,
+      "eval_steps_per_second": 3.272,
+      "step": 118
+    },
+    {
+      "epoch": 8.95,
+      "grad_norm": 0.9810589551925659,
+      "learning_rate": 1.1314285714285715e-05,
+      "loss": 1.6506,
+      "step": 132
+    },
+    {
+      "epoch": 8.95,
+      "eval_accuracy": 0.3076923076923077,
+      "eval_loss": 1.570804238319397,
+      "eval_runtime": 4.0121,
+      "eval_samples_per_second": 71.285,
+      "eval_steps_per_second": 4.486,
+      "step": 132
+    },
+    {
+      "epoch": 9.97,
+      "grad_norm": 0.6308433413505554,
+      "learning_rate": 1.26e-05,
+      "loss": 1.5317,
+      "step": 147
+    },
+    {
+      "epoch": 9.97,
+      "eval_accuracy": 0.3006993006993007,
+      "eval_loss": 1.5640877485275269,
+      "eval_runtime": 5.5443,
+      "eval_samples_per_second": 51.585,
+      "eval_steps_per_second": 3.247,
+      "step": 147
+    },
+    {
+      "epoch": 10.98,
+      "grad_norm": 1.381785273551941,
+      "learning_rate": 1.3885714285714286e-05,
+      "loss": 1.5226,
+      "step": 162
+    },
+    {
+      "epoch": 10.98,
+      "eval_accuracy": 0.2867132867132867,
+      "eval_loss": 1.5564289093017578,
+      "eval_runtime": 5.0151,
+      "eval_samples_per_second": 57.027,
+      "eval_steps_per_second": 3.589,
+      "step": 162
+    },
+    {
+      "epoch": 12.0,
+      "grad_norm": 0.552884578704834,
+      "learning_rate": 1.517142857142857e-05,
+      "loss": 1.5116,
+      "step": 177
+    },
+    {
+      "epoch": 12.0,
+      "eval_accuracy": 0.2692307692307692,
+      "eval_loss": 1.5478310585021973,
+      "eval_runtime": 5.4183,
+      "eval_samples_per_second": 52.784,
+      "eval_steps_per_second": 3.322,
+      "step": 177
+    },
+    {
+      "epoch": 12.95,
+      "grad_norm": 0.8657445907592773,
+      "learning_rate": 1.6371428571428572e-05,
+      "loss": 1.6046,
+      "step": 191
+    },
+    {
+      "epoch": 12.95,
+      "eval_accuracy": 0.26223776223776224,
+      "eval_loss": 1.5384888648986816,
+      "eval_runtime": 5.1185,
+      "eval_samples_per_second": 55.875,
+      "eval_steps_per_second": 3.517,
+      "step": 191
+    },
+    {
+      "epoch": 13.97,
+      "grad_norm": 0.8375154733657837,
+      "learning_rate": 1.7657142857142857e-05,
+      "loss": 1.4822,
+      "step": 206
+    },
+    {
+      "epoch": 13.97,
+      "eval_accuracy": 0.25524475524475526,
+      "eval_loss": 1.5284570455551147,
+      "eval_runtime": 5.5939,
+      "eval_samples_per_second": 51.127,
+      "eval_steps_per_second": 3.218,
+      "step": 206
+    },
+    {
+      "epoch": 14.98,
+      "grad_norm": 0.7776028513908386,
+      "learning_rate": 1.8942857142857145e-05,
+      "loss": 1.4614,
+      "step": 221
+    },
+    {
+      "epoch": 14.98,
+      "eval_accuracy": 0.25524475524475526,
+      "eval_loss": 1.5183566808700562,
+      "eval_runtime": 5.2849,
+      "eval_samples_per_second": 54.116,
+      "eval_steps_per_second": 3.406,
+      "step": 221
+    },
+    {
+      "epoch": 16.0,
+      "grad_norm": 0.3529152572154999,
+      "learning_rate": 2.022857142857143e-05,
+      "loss": 1.4396,
+      "step": 236
+    },
+    {
+      "epoch": 16.0,
+      "eval_accuracy": 0.2517482517482518,
+      "eval_loss": 1.510148048400879,
+      "eval_runtime": 6.3848,
+      "eval_samples_per_second": 44.794,
+      "eval_steps_per_second": 2.819,
+      "step": 236
+    },
+    {
+      "epoch": 16.95,
+      "grad_norm": 0.5623555779457092,
+      "learning_rate": 2.1428571428571428e-05,
+      "loss": 1.5047,
+      "step": 250
+    },
+    {
+      "epoch": 16.95,
+      "eval_accuracy": 0.24475524475524477,
+      "eval_loss": 1.510407567024231,
+      "eval_runtime": 5.762,
+      "eval_samples_per_second": 49.635,
+      "eval_steps_per_second": 3.124,
+      "step": 250
+    },
+    {
+      "epoch": 17.97,
+      "grad_norm": 0.889561116695404,
+      "learning_rate": 2.2714285714285713e-05,
+      "loss": 1.3741,
+      "step": 265
+    },
+    {
+      "epoch": 17.97,
+      "eval_accuracy": 0.2517482517482518,
+      "eval_loss": 1.524996280670166,
+      "eval_runtime": 5.3825,
+      "eval_samples_per_second": 53.135,
+      "eval_steps_per_second": 3.344,
+      "step": 265
+    },
+    {
+      "epoch": 18.98,
+      "grad_norm": 0.837506890296936,
+      "learning_rate": 2.4e-05,
+      "loss": 1.3512,
+      "step": 280
+    },
+    {
+      "epoch": 18.98,
+      "eval_accuracy": 0.26573426573426573,
+      "eval_loss": 1.5328779220581055,
+      "eval_runtime": 4.8496,
+      "eval_samples_per_second": 58.974,
+      "eval_steps_per_second": 3.712,
+      "step": 280
+    },
+    {
+      "epoch": 20.0,
+      "grad_norm": 0.6774041056632996,
+      "learning_rate": 2.5285714285714285e-05,
+      "loss": 1.3286,
+      "step": 295
+    },
+    {
+      "epoch": 20.0,
+      "eval_accuracy": 0.32517482517482516,
+      "eval_loss": 1.509504795074463,
+      "eval_runtime": 4.7142,
+      "eval_samples_per_second": 60.667,
+      "eval_steps_per_second": 3.818,
+      "step": 295
+    },
+    {
+      "epoch": 20.95,
+      "grad_norm": 0.5629450678825378,
+      "learning_rate": 2.6485714285714287e-05,
+      "loss": 1.3967,
+      "step": 309
+    },
+    {
+      "epoch": 20.95,
+      "eval_accuracy": 0.34965034965034963,
+      "eval_loss": 1.482858419418335,
+      "eval_runtime": 5.0574,
+      "eval_samples_per_second": 56.551,
+      "eval_steps_per_second": 3.559,
+      "step": 309
+    },
+    {
+      "epoch": 21.97,
+      "grad_norm": 0.5265232920646667,
+      "learning_rate": 2.7771428571428572e-05,
+      "loss": 1.2779,
+      "step": 324
+    },
+    {
+      "epoch": 21.97,
+      "eval_accuracy": 0.38461538461538464,
+      "eval_loss": 1.4598273038864136,
+      "eval_runtime": 5.2206,
+      "eval_samples_per_second": 54.783,
+      "eval_steps_per_second": 3.448,
+      "step": 324
+    },
+    {
+      "epoch": 22.98,
+      "grad_norm": 0.4762378931045532,
+      "learning_rate": 2.9057142857142856e-05,
+      "loss": 1.2449,
+      "step": 339
+    },
+    {
+      "epoch": 22.98,
+      "eval_accuracy": 0.4160839160839161,
+      "eval_loss": 1.4284018278121948,
+      "eval_runtime": 5.3164,
+      "eval_samples_per_second": 53.796,
+      "eval_steps_per_second": 3.386,
+      "step": 339
+    },
+    {
+      "epoch": 24.0,
+      "grad_norm": 0.598441481590271,
+      "learning_rate": 2.996190476190476e-05,
+      "loss": 1.2118,
+      "step": 354
+    },
+    {
+      "epoch": 24.0,
+      "eval_accuracy": 0.4230769230769231,
+      "eval_loss": 1.4161577224731445,
+      "eval_runtime": 4.9587,
+      "eval_samples_per_second": 57.677,
+      "eval_steps_per_second": 3.63,
+      "step": 354
+    },
+    {
+      "epoch": 24.95,
+      "grad_norm": 1.1748360395431519,
+      "learning_rate": 2.982857142857143e-05,
+      "loss": 1.2521,
+      "step": 368
+    },
+    {
+      "epoch": 24.95,
+      "eval_accuracy": 0.44755244755244755,
+      "eval_loss": 1.3797944784164429,
+      "eval_runtime": 7.057,
+      "eval_samples_per_second": 40.527,
+      "eval_steps_per_second": 2.551,
+      "step": 368
+    },
+    {
+      "epoch": 25.97,
+      "grad_norm": 0.7063133716583252,
+      "learning_rate": 2.9685714285714284e-05,
+      "loss": 1.1183,
+      "step": 383
+    },
+    {
+      "epoch": 25.97,
+      "eval_accuracy": 0.479020979020979,
+      "eval_loss": 1.324568271636963,
+      "eval_runtime": 5.5598,
+      "eval_samples_per_second": 51.44,
+      "eval_steps_per_second": 3.238,
+      "step": 383
+    },
+    {
+      "epoch": 26.98,
+      "grad_norm": 0.91399085521698,
+      "learning_rate": 2.9542857142857142e-05,
+      "loss": 1.0778,
+      "step": 398
+    },
+    {
+      "epoch": 26.98,
+      "eval_accuracy": 0.493006993006993,
+      "eval_loss": 1.276139259338379,
+      "eval_runtime": 4.8059,
+      "eval_samples_per_second": 59.51,
+      "eval_steps_per_second": 3.745,
+      "step": 398
+    },
+    {
+      "epoch": 28.0,
+      "grad_norm": 1.1466203927993774,
+      "learning_rate": 2.94e-05,
+      "loss": 1.0306,
+      "step": 413
+    },
+    {
+      "epoch": 28.0,
+      "eval_accuracy": 0.5104895104895105,
+      "eval_loss": 1.2299922704696655,
+      "eval_runtime": 6.0199,
+      "eval_samples_per_second": 47.509,
+      "eval_steps_per_second": 2.99,
+      "step": 413
+    },
+    {
+      "epoch": 28.95,
+      "grad_norm": 1.1782015562057495,
+      "learning_rate": 2.9266666666666665e-05,
+      "loss": 1.0808,
+      "step": 427
+    },
+    {
+      "epoch": 28.95,
+      "eval_accuracy": 0.513986013986014,
+      "eval_loss": 1.2064085006713867,
+      "eval_runtime": 6.0946,
+      "eval_samples_per_second": 46.927,
+      "eval_steps_per_second": 2.953,
+      "step": 427
+    },
+    {
+      "epoch": 29.97,
+      "grad_norm": 0.9410634636878967,
+      "learning_rate": 2.9123809523809523e-05,
+      "loss": 0.9767,
+      "step": 442
+    },
+    {
+      "epoch": 29.97,
+      "eval_accuracy": 0.5524475524475524,
+      "eval_loss": 1.1669567823410034,
+      "eval_runtime": 4.8236,
+      "eval_samples_per_second": 59.292,
+      "eval_steps_per_second": 3.732,
+      "step": 442
+    },
+    {
+      "epoch": 30.98,
+      "grad_norm": 1.2911593914031982,
+      "learning_rate": 2.898095238095238e-05,
+      "loss": 0.9589,
+      "step": 457
+    },
+    {
+      "epoch": 30.98,
+      "eval_accuracy": 0.5734265734265734,
+      "eval_loss": 1.126379132270813,
+      "eval_runtime": 5.0702,
+      "eval_samples_per_second": 56.408,
+      "eval_steps_per_second": 3.55,
+      "step": 457
+    },
+    {
+      "epoch": 32.0,
+      "grad_norm": 1.2354660034179688,
+      "learning_rate": 2.883809523809524e-05,
+      "loss": 0.9193,
+      "step": 472
+    },
+    {
+      "epoch": 32.0,
+      "eval_accuracy": 0.5874125874125874,
+      "eval_loss": 1.1012728214263916,
+      "eval_runtime": 5.9275,
+      "eval_samples_per_second": 48.249,
+      "eval_steps_per_second": 3.037,
+      "step": 472
+    },
+    {
+      "epoch": 32.95,
+      "grad_norm": 0.8961493968963623,
+      "learning_rate": 2.8704761904761905e-05,
+      "loss": 0.9462,
+      "step": 486
+    },
+    {
+      "epoch": 32.95,
+      "eval_accuracy": 0.6048951048951049,
+      "eval_loss": 1.0736197233200073,
+      "eval_runtime": 5.7691,
+      "eval_samples_per_second": 49.575,
+      "eval_steps_per_second": 3.12,
+      "step": 486
+    },
+    {
+      "epoch": 33.97,
+      "grad_norm": 0.9503061175346375,
+      "learning_rate": 2.8561904761904763e-05,
+      "loss": 0.85,
+      "step": 501
+    },
+    {
+      "epoch": 33.97,
+      "eval_accuracy": 0.6048951048951049,
+      "eval_loss": 1.0628284215927124,
+      "eval_runtime": 6.1054,
+      "eval_samples_per_second": 46.844,
+      "eval_steps_per_second": 2.948,
+      "step": 501
+    },
+    {
+      "epoch": 34.98,
+      "grad_norm": 1.0706520080566406,
+      "learning_rate": 2.841904761904762e-05,
+      "loss": 0.8294,
+      "step": 516
+    },
+    {
+      "epoch": 34.98,
+      "eval_accuracy": 0.6188811188811189,
+      "eval_loss": 1.0472813844680786,
+      "eval_runtime": 4.7734,
+      "eval_samples_per_second": 59.916,
+      "eval_steps_per_second": 3.771,
+      "step": 516
+    },
+    {
+      "epoch": 36.0,
+      "grad_norm": 0.9784532189369202,
+      "learning_rate": 2.827619047619048e-05,
+      "loss": 0.8025,
+      "step": 531
+    },
+    {
+      "epoch": 36.0,
+      "eval_accuracy": 0.6328671328671329,
+      "eval_loss": 1.0030184984207153,
+      "eval_runtime": 5.5291,
+      "eval_samples_per_second": 51.726,
+      "eval_steps_per_second": 3.255,
+      "step": 531
+    },
+    {
+      "epoch": 36.95,
+      "grad_norm": 0.8138810396194458,
+      "learning_rate": 2.8142857142857145e-05,
+      "loss": 0.8206,
+      "step": 545
+    },
+    {
+      "epoch": 36.95,
+      "eval_accuracy": 0.6398601398601399,
+      "eval_loss": 0.9963561296463013,
+      "eval_runtime": 5.8221,
+      "eval_samples_per_second": 49.123,
+      "eval_steps_per_second": 3.092,
+      "step": 545
+    },
+    {
+      "epoch": 37.97,
+      "grad_norm": 0.8114917278289795,
+      "learning_rate": 2.8e-05,
+      "loss": 0.7541,
+      "step": 560
+    },
+    {
+      "epoch": 37.97,
+      "eval_accuracy": 0.6608391608391608,
+      "eval_loss": 0.9604987502098083,
+      "eval_runtime": 4.9473,
+      "eval_samples_per_second": 57.809,
+      "eval_steps_per_second": 3.638,
+      "step": 560
+    },
+    {
+      "epoch": 38.98,
+      "grad_norm": 0.9221667051315308,
+      "learning_rate": 2.7857142857142858e-05,
+      "loss": 0.7413,
+      "step": 575
+    },
+    {
+      "epoch": 38.98,
+      "eval_accuracy": 0.6643356643356644,
+      "eval_loss": 0.946722686290741,
+      "eval_runtime": 4.7955,
+      "eval_samples_per_second": 59.64,
+      "eval_steps_per_second": 3.754,
+      "step": 575
+    },
+    {
+      "epoch": 40.0,
+      "grad_norm": 1.2665307521820068,
+      "learning_rate": 2.7714285714285716e-05,
+      "loss": 0.709,
+      "step": 590
+    },
+    {
+      "epoch": 40.0,
+      "eval_accuracy": 0.6678321678321678,
+      "eval_loss": 0.9348079562187195,
+      "eval_runtime": 5.3252,
+      "eval_samples_per_second": 53.707,
+      "eval_steps_per_second": 3.38,
+      "step": 590
+    },
+    {
+      "epoch": 40.95,
+      "grad_norm": 1.0893254280090332,
+      "learning_rate": 2.758095238095238e-05,
+      "loss": 0.7817,
+      "step": 604
+    },
+    {
+      "epoch": 40.95,
+      "eval_accuracy": 0.6678321678321678,
+      "eval_loss": 0.9366168975830078,
+      "eval_runtime": 6.4862,
+      "eval_samples_per_second": 44.094,
+      "eval_steps_per_second": 2.775,
+      "step": 604
+    },
+    {
+      "epoch": 41.97,
+      "grad_norm": 0.7136903405189514,
+      "learning_rate": 2.743809523809524e-05,
+      "loss": 0.7034,
+      "step": 619
+    },
+    {
+      "epoch": 41.97,
+      "eval_accuracy": 0.6818181818181818,
+      "eval_loss": 0.9109411835670471,
+      "eval_runtime": 6.1724,
+      "eval_samples_per_second": 46.335,
+      "eval_steps_per_second": 2.916,
+      "step": 619
+    },
+    {
+      "epoch": 42.98,
+      "grad_norm": 1.142288327217102,
+      "learning_rate": 2.7295238095238097e-05,
+      "loss": 0.6856,
+      "step": 634
+    },
+    {
+      "epoch": 42.98,
+      "eval_accuracy": 0.6573426573426573,
+      "eval_loss": 0.9276965856552124,
+      "eval_runtime": 4.6805,
+      "eval_samples_per_second": 61.105,
+      "eval_steps_per_second": 3.846,
+      "step": 634
+    },
+    {
+      "epoch": 44.0,
+      "grad_norm": 1.3661341667175293,
+      "learning_rate": 2.7152380952380952e-05,
+      "loss": 0.6625,
+      "step": 649
+    },
+    {
+      "epoch": 44.0,
+      "eval_accuracy": 0.6783216783216783,
+      "eval_loss": 0.8979520201683044,
+      "eval_runtime": 5.8084,
+      "eval_samples_per_second": 49.239,
+      "eval_steps_per_second": 3.099,
+      "step": 649
+    },
+    {
+      "epoch": 44.95,
+      "grad_norm": 0.7071289420127869,
+      "learning_rate": 2.701904761904762e-05,
+      "loss": 0.7207,
+      "step": 663
+    },
+    {
+      "epoch": 44.95,
+      "eval_accuracy": 0.6713286713286714,
+      "eval_loss": 0.9049975275993347,
+      "eval_runtime": 5.5071,
+      "eval_samples_per_second": 51.933,
+      "eval_steps_per_second": 3.268,
+      "step": 663
+    },
+    {
+      "epoch": 45.97,
+      "grad_norm": 0.9930222034454346,
+      "learning_rate": 2.6876190476190476e-05,
+      "loss": 0.6684,
+      "step": 678
+    },
+    {
+      "epoch": 45.97,
+      "eval_accuracy": 0.6748251748251748,
+      "eval_loss": 0.8973459005355835,
+      "eval_runtime": 4.8744,
+      "eval_samples_per_second": 58.674,
+      "eval_steps_per_second": 3.693,
+      "step": 678
+    },
+    {
+      "epoch": 46.98,
+      "grad_norm": 1.2905802726745605,
+      "learning_rate": 2.6733333333333334e-05,
+      "loss": 0.6651,
+      "step": 693
+    },
+    {
+      "epoch": 46.98,
+      "eval_accuracy": 0.6783216783216783,
+      "eval_loss": 0.8934686779975891,
+      "eval_runtime": 4.9699,
+      "eval_samples_per_second": 57.546,
+      "eval_steps_per_second": 3.622,
+      "step": 693
+    },
+    {
+      "epoch": 48.0,
+      "grad_norm": 1.5243291854858398,
+      "learning_rate": 2.6590476190476192e-05,
+      "loss": 0.6451,
+      "step": 708
+    },
+    {
+      "epoch": 48.0,
+      "eval_accuracy": 0.6748251748251748,
+      "eval_loss": 0.8944538831710815,
+      "eval_runtime": 4.9181,
+      "eval_samples_per_second": 58.153,
+      "eval_steps_per_second": 3.66,
+      "step": 708
+    },
+    {
+      "epoch": 48.95,
+      "grad_norm": 1.4421076774597168,
+      "learning_rate": 2.6457142857142857e-05,
+      "loss": 0.6774,
+      "step": 722
+    },
+    {
+      "epoch": 48.95,
+      "eval_accuracy": 0.6818181818181818,
+      "eval_loss": 0.8879284858703613,
+      "eval_runtime": 5.1997,
+      "eval_samples_per_second": 55.003,
+      "eval_steps_per_second": 3.462,
+      "step": 722
+    },
+    {
+      "epoch": 49.97,
+      "grad_norm": 1.3786752223968506,
+      "learning_rate": 2.6314285714285715e-05,
+      "loss": 0.6308,
+      "step": 737
+    },
+    {
+      "epoch": 49.97,
+      "eval_accuracy": 0.6818181818181818,
+      "eval_loss": 0.8882840275764465,
+      "eval_runtime": 5.1707,
+      "eval_samples_per_second": 55.312,
+      "eval_steps_per_second": 3.481,
+      "step": 737
+    },
+    {
+      "epoch": 50.98,
+      "grad_norm": 1.9291437864303589,
+      "learning_rate": 2.617142857142857e-05,
+      "loss": 0.6199,
+      "step": 752
+    },
+    {
+      "epoch": 50.98,
+      "eval_accuracy": 0.6818181818181818,
+      "eval_loss": 0.8825680017471313,
+      "eval_runtime": 4.9861,
+      "eval_samples_per_second": 57.359,
+      "eval_steps_per_second": 3.61,
+      "step": 752
+    },
+    {
+      "epoch": 52.0,
+      "grad_norm": 1.2054837942123413,
+      "learning_rate": 2.6028571428571428e-05,
+      "loss": 0.6379,
+      "step": 767
+    },
+    {
+      "epoch": 52.0,
+      "eval_accuracy": 0.6923076923076923,
+      "eval_loss": 0.8582118153572083,
+      "eval_runtime": 5.287,
+      "eval_samples_per_second": 54.095,
+      "eval_steps_per_second": 3.405,
+      "step": 767
+    },
+    {
+      "epoch": 52.95,
+      "grad_norm": 1.3063265085220337,
+      "learning_rate": 2.5895238095238094e-05,
+      "loss": 0.6588,
+      "step": 781
+    },
+    {
+      "epoch": 52.95,
+      "eval_accuracy": 0.6818181818181818,
+      "eval_loss": 0.8825483918190002,
+      "eval_runtime": 4.6024,
+      "eval_samples_per_second": 62.141,
+      "eval_steps_per_second": 3.911,
+      "step": 781
+    },
+    {
+      "epoch": 53.97,
+      "grad_norm": 1.2957918643951416,
+      "learning_rate": 2.5752380952380952e-05,
+      "loss": 0.5857,
+      "step": 796
+    },
+    {
+      "epoch": 53.97,
+      "eval_accuracy": 0.6748251748251748,
+      "eval_loss": 0.8808070421218872,
+      "eval_runtime": 5.2003,
+      "eval_samples_per_second": 54.997,
+      "eval_steps_per_second": 3.461,
+      "step": 796
+    },
+    {
+      "epoch": 54.98,
+      "grad_norm": 0.9337932467460632,
+      "learning_rate": 2.560952380952381e-05,
+      "loss": 0.6076,
+      "step": 811
+    },
+    {
+      "epoch": 54.98,
+      "eval_accuracy": 0.6958041958041958,
+      "eval_loss": 0.8554547429084778,
+      "eval_runtime": 5.108,
+      "eval_samples_per_second": 55.991,
+      "eval_steps_per_second": 3.524,
+      "step": 811
+    },
+    {
+      "epoch": 56.0,
+      "grad_norm": 1.4797887802124023,
+      "learning_rate": 2.5466666666666668e-05,
+      "loss": 0.5934,
+      "step": 826
+    },
+    {
+      "epoch": 56.0,
+      "eval_accuracy": 0.6888111888111889,
+      "eval_loss": 0.8653700351715088,
+      "eval_runtime": 4.9201,
+      "eval_samples_per_second": 58.128,
+      "eval_steps_per_second": 3.658,
+      "step": 826
+    },
+    {
+      "epoch": 56.95,
+      "grad_norm": 0.9749770164489746,
+      "learning_rate": 2.5333333333333334e-05,
+      "loss": 0.6427,
+      "step": 840
+    },
+    {
+      "epoch": 56.95,
+      "eval_accuracy": 0.6853146853146853,
+      "eval_loss": 0.861566424369812,
+      "eval_runtime": 5.5361,
+      "eval_samples_per_second": 51.661,
+      "eval_steps_per_second": 3.251,
+      "step": 840
+    },
+    {
+      "epoch": 57.97,
+      "grad_norm": 1.1588115692138672,
+      "learning_rate": 2.519047619047619e-05,
+      "loss": 0.5782,
+      "step": 855
+    },
+    {
+      "epoch": 57.97,
+      "eval_accuracy": 0.6678321678321678,
+      "eval_loss": 0.8711130619049072,
+      "eval_runtime": 5.1513,
+      "eval_samples_per_second": 55.52,
+      "eval_steps_per_second": 3.494,
+      "step": 855
+    },
+    {
+      "epoch": 58.98,
+      "grad_norm": 2.290128707885742,
+      "learning_rate": 2.504761904761905e-05,
+      "loss": 0.5819,
+      "step": 870
+    },
+    {
+      "epoch": 58.98,
+      "eval_accuracy": 0.6748251748251748,
+      "eval_loss": 0.8689377903938293,
+      "eval_runtime": 5.3476,
+      "eval_samples_per_second": 53.482,
+      "eval_steps_per_second": 3.366,
+      "step": 870
+    },
+    {
+      "epoch": 60.0,
+      "grad_norm": 1.8434489965438843,
+      "learning_rate": 2.4904761904761908e-05,
+      "loss": 0.5918,
+      "step": 885
+    },
+    {
+      "epoch": 60.0,
+      "eval_accuracy": 0.6923076923076923,
+      "eval_loss": 0.8602246642112732,
+      "eval_runtime": 5.1182,
+      "eval_samples_per_second": 55.879,
+      "eval_steps_per_second": 3.517,
+      "step": 885
+    },
+    {
+      "epoch": 60.95,
+      "grad_norm": 1.1310744285583496,
+      "learning_rate": 2.4771428571428573e-05,
+      "loss": 0.5845,
+      "step": 899
+    },
+    {
+      "epoch": 60.95,
+      "eval_accuracy": 0.6993006993006993,
+      "eval_loss": 0.8458691835403442,
+      "eval_runtime": 4.8028,
+      "eval_samples_per_second": 59.549,
+      "eval_steps_per_second": 3.748,
+      "step": 899
+    },
+    {
+      "epoch": 61.97,
+      "grad_norm": 1.319384217262268,
+      "learning_rate": 2.462857142857143e-05,
+      "loss": 0.5667,
+      "step": 914
+    },
+    {
+      "epoch": 61.97,
+      "eval_accuracy": 0.7027972027972028,
+      "eval_loss": 0.8466892838478088,
+      "eval_runtime": 4.9074,
+      "eval_samples_per_second": 58.279,
+      "eval_steps_per_second": 3.668,
+      "step": 914
+    },
+    {
+      "epoch": 62.98,
+      "grad_norm": 1.2371602058410645,
+      "learning_rate": 2.448571428571429e-05,
+      "loss": 0.5327,
+      "step": 929
+    },
+    {
+      "epoch": 62.98,
+      "eval_accuracy": 0.6923076923076923,
+      "eval_loss": 0.8539786338806152,
+      "eval_runtime": 5.32,
+      "eval_samples_per_second": 53.759,
+      "eval_steps_per_second": 3.383,
+      "step": 929
+    },
+    {
+      "epoch": 64.0,
+      "grad_norm": 2.1778481006622314,
+      "learning_rate": 2.434285714285714e-05,
+      "loss": 0.523,
+      "step": 944
+    },
+    {
+      "epoch": 64.0,
+      "eval_accuracy": 0.7062937062937062,
+      "eval_loss": 0.8323072791099548,
+      "eval_runtime": 5.1152,
+      "eval_samples_per_second": 55.912,
+      "eval_steps_per_second": 3.519,
+      "step": 944
+    },
+    {
+      "epoch": 64.95,
+      "grad_norm": 0.8219490051269531,
+      "learning_rate": 2.420952380952381e-05,
+      "loss": 0.548,
+      "step": 958
+    },
+    {
+      "epoch": 64.95,
+      "eval_accuracy": 0.6993006993006993,
+      "eval_loss": 0.8406782746315002,
+      "eval_runtime": 5.325,
+      "eval_samples_per_second": 53.709,
+      "eval_steps_per_second": 3.38,
+      "step": 958
+    },
+    {
+      "epoch": 65.97,
+      "grad_norm": 2.0205609798431396,
+      "learning_rate": 2.4066666666666664e-05,
+      "loss": 0.5399,
+      "step": 973
+    },
+    {
+      "epoch": 65.97,
+      "eval_accuracy": 0.6993006993006993,
+      "eval_loss": 0.837882936000824,
+      "eval_runtime": 5.6808,
+      "eval_samples_per_second": 50.345,
+      "eval_steps_per_second": 3.169,
+      "step": 973
+    },
+    {
+      "epoch": 66.98,
+      "grad_norm": 1.527024507522583,
+      "learning_rate": 2.3923809523809522e-05,
+      "loss": 0.5324,
+      "step": 988
+    },
+    {
+      "epoch": 66.98,
+      "eval_accuracy": 0.7027972027972028,
+      "eval_loss": 0.8118743896484375,
+      "eval_runtime": 5.543,
+      "eval_samples_per_second": 51.596,
+      "eval_steps_per_second": 3.247,
+      "step": 988
+    },
+    {
+      "epoch": 68.0,
+      "grad_norm": 1.3294800519943237,
+      "learning_rate": 2.378095238095238e-05,
+      "loss": 0.5171,
+      "step": 1003
+    },
+    {
+      "epoch": 68.0,
+      "eval_accuracy": 0.6923076923076923,
+      "eval_loss": 0.8444966673851013,
+      "eval_runtime": 5.5571,
+      "eval_samples_per_second": 51.465,
+      "eval_steps_per_second": 3.239,
+      "step": 1003
+    },
+    {
+      "epoch": 68.95,
+      "grad_norm": 1.6269547939300537,
+      "learning_rate": 2.3647619047619046e-05,
+      "loss": 0.538,
+      "step": 1017
+    },
+    {
+      "epoch": 68.95,
+      "eval_accuracy": 0.7097902097902098,
+      "eval_loss": 0.8196296095848083,
+      "eval_runtime": 4.8636,
+      "eval_samples_per_second": 58.805,
+      "eval_steps_per_second": 3.701,
+      "step": 1017
+    },
+    {
+      "epoch": 69.97,
+      "grad_norm": 1.0726577043533325,
+      "learning_rate": 2.3504761904761904e-05,
+      "loss": 0.5312,
+      "step": 1032
+    },
+    {
+      "epoch": 69.97,
+      "eval_accuracy": 0.6853146853146853,
+      "eval_loss": 0.8414965271949768,
+      "eval_runtime": 5.2322,
+      "eval_samples_per_second": 54.661,
+      "eval_steps_per_second": 3.44,
+      "step": 1032
+    },
+    {
+      "epoch": 70.98,
+      "grad_norm": 1.0595276355743408,
+      "learning_rate": 2.3361904761904762e-05,
+      "loss": 0.4914,
+      "step": 1047
+    },
+    {
+      "epoch": 70.98,
+      "eval_accuracy": 0.6958041958041958,
+      "eval_loss": 0.8184240460395813,
+      "eval_runtime": 4.6109,
+      "eval_samples_per_second": 62.027,
+      "eval_steps_per_second": 3.904,
+      "step": 1047
+    },
+    {
+      "epoch": 72.0,
+      "grad_norm": 1.3313194513320923,
+      "learning_rate": 2.321904761904762e-05,
+      "loss": 0.5055,
+      "step": 1062
+    },
+    {
+      "epoch": 72.0,
+      "eval_accuracy": 0.6923076923076923,
+      "eval_loss": 0.8218348622322083,
+      "eval_runtime": 5.5535,
+      "eval_samples_per_second": 51.499,
+      "eval_steps_per_second": 3.241,
+      "step": 1062
+    },
+    {
+      "epoch": 72.95,
+      "grad_norm": 1.9882231950759888,
+      "learning_rate": 2.3085714285714286e-05,
+      "loss": 0.5401,
+      "step": 1076
+    },
+    {
+      "epoch": 72.95,
+      "eval_accuracy": 0.7027972027972028,
+      "eval_loss": 0.815979540348053,
+      "eval_runtime": 6.0512,
+      "eval_samples_per_second": 47.264,
+      "eval_steps_per_second": 2.975,
+      "step": 1076
+    },
+    {
+      "epoch": 73.97,
+      "grad_norm": 1.8936933279037476,
+      "learning_rate": 2.2942857142857144e-05,
+      "loss": 0.4966,
+      "step": 1091
+    },
+    {
+      "epoch": 73.97,
+      "eval_accuracy": 0.6888111888111889,
+      "eval_loss": 0.8237566351890564,
+      "eval_runtime": 4.6622,
+      "eval_samples_per_second": 61.344,
+      "eval_steps_per_second": 3.861,
+      "step": 1091
+    },
+    {
+      "epoch": 74.98,
+      "grad_norm": 2.0124337673187256,
+      "learning_rate": 2.2800000000000002e-05,
+      "loss": 0.4768,
+      "step": 1106
+    },
+    {
+      "epoch": 74.98,
+      "eval_accuracy": 0.6993006993006993,
+      "eval_loss": 0.8185241222381592,
+      "eval_runtime": 4.2442,
+      "eval_samples_per_second": 67.386,
+      "eval_steps_per_second": 4.241,
+      "step": 1106
+    },
+    {
+      "epoch": 76.0,
+      "grad_norm": 1.7522495985031128,
+      "learning_rate": 2.265714285714286e-05,
+      "loss": 0.4789,
+      "step": 1121
+    },
+    {
+      "epoch": 76.0,
+      "eval_accuracy": 0.7027972027972028,
+      "eval_loss": 0.8260769844055176,
+      "eval_runtime": 6.128,
+      "eval_samples_per_second": 46.671,
+      "eval_steps_per_second": 2.937,
+      "step": 1121
+    },
+    {
+      "epoch": 76.95,
+      "grad_norm": 1.8254278898239136,
+      "learning_rate": 2.2523809523809526e-05,
+      "loss": 0.5176,
+      "step": 1135
+    },
+    {
+      "epoch": 76.95,
+      "eval_accuracy": 0.7097902097902098,
+      "eval_loss": 0.8109525442123413,
+      "eval_runtime": 6.0147,
+      "eval_samples_per_second": 47.55,
+      "eval_steps_per_second": 2.993,
+      "step": 1135
+    },
+    {
+      "epoch": 77.97,
+      "grad_norm": 1.4586073160171509,
+      "learning_rate": 2.238095238095238e-05,
+      "loss": 0.466,
+      "step": 1150
+    },
+    {
+      "epoch": 77.97,
+      "eval_accuracy": 0.6993006993006993,
+      "eval_loss": 0.814057469367981,
+      "eval_runtime": 4.782,
+      "eval_samples_per_second": 59.808,
+      "eval_steps_per_second": 3.764,
+      "step": 1150
+    },
+    {
+      "epoch": 78.98,
+      "grad_norm": 1.7543085813522339,
+      "learning_rate": 2.223809523809524e-05,
+      "loss": 0.4736,
+      "step": 1165
+    },
+    {
+      "epoch": 78.98,
+      "eval_accuracy": 0.7167832167832168,
+      "eval_loss": 0.7970029711723328,
+      "eval_runtime": 5.8124,
+      "eval_samples_per_second": 49.205,
+      "eval_steps_per_second": 3.097,
+      "step": 1165
+    },
+    {
+      "epoch": 80.0,
+      "grad_norm": 1.6676491498947144,
+      "learning_rate": 2.2095238095238096e-05,
+      "loss": 0.4785,
+      "step": 1180
+    },
+    {
+      "epoch": 80.0,
+      "eval_accuracy": 0.7097902097902098,
+      "eval_loss": 0.8062326312065125,
+      "eval_runtime": 4.5174,
+      "eval_samples_per_second": 63.31,
+      "eval_steps_per_second": 3.985,
+      "step": 1180
+    },
+    {
+      "epoch": 80.95,
+      "grad_norm": 1.827528476715088,
+      "learning_rate": 2.1961904761904762e-05,
+      "loss": 0.5309,
+      "step": 1194
+    },
+    {
+      "epoch": 80.95,
+      "eval_accuracy": 0.6958041958041958,
+      "eval_loss": 0.8051398396492004,
+      "eval_runtime": 4.9397,
+      "eval_samples_per_second": 57.899,
+      "eval_steps_per_second": 3.644,
+      "step": 1194
+    },
+    {
+      "epoch": 81.97,
+      "grad_norm": 2.3700757026672363,
+      "learning_rate": 2.181904761904762e-05,
+      "loss": 0.4571,
+      "step": 1209
+    },
+    {
+      "epoch": 81.97,
+      "eval_accuracy": 0.7097902097902098,
+      "eval_loss": 0.8024002909660339,
+      "eval_runtime": 6.3277,
+      "eval_samples_per_second": 45.198,
+      "eval_steps_per_second": 2.845,
+      "step": 1209
+    },
+    {
+      "epoch": 82.98,
+      "grad_norm": 1.5626410245895386,
+      "learning_rate": 2.1676190476190478e-05,
+      "loss": 0.47,
+      "step": 1224
+    },
+    {
+      "epoch": 82.98,
+      "eval_accuracy": 0.7167832167832168,
+      "eval_loss": 0.8030509948730469,
+      "eval_runtime": 5.4575,
+      "eval_samples_per_second": 52.405,
+      "eval_steps_per_second": 3.298,
+      "step": 1224
+    },
+    {
+      "epoch": 84.0,
+      "grad_norm": 1.4918785095214844,
+      "learning_rate": 2.1533333333333333e-05,
+      "loss": 0.4525,
+      "step": 1239
+    },
+    {
+      "epoch": 84.0,
+      "eval_accuracy": 0.7132867132867133,
+      "eval_loss": 0.7910680174827576,
+      "eval_runtime": 4.6183,
+      "eval_samples_per_second": 61.928,
+      "eval_steps_per_second": 3.898,
+      "step": 1239
+    },
+    {
+      "epoch": 84.95,
+      "grad_norm": 1.4631191492080688,
+      "learning_rate": 2.1400000000000002e-05,
+      "loss": 0.5058,
+      "step": 1253
+    },
+    {
+      "epoch": 84.95,
+      "eval_accuracy": 0.7132867132867133,
+      "eval_loss": 0.7877373695373535,
+      "eval_runtime": 4.6086,
+      "eval_samples_per_second": 62.058,
+      "eval_steps_per_second": 3.906,
+      "step": 1253
+    },
+    {
+      "epoch": 85.97,
+      "grad_norm": 1.463442087173462,
+      "learning_rate": 2.1257142857142856e-05,
+      "loss": 0.4627,
+      "step": 1268
+    },
+    {
+      "epoch": 85.97,
+      "eval_accuracy": 0.7062937062937062,
+      "eval_loss": 0.7917687296867371,
+      "eval_runtime": 6.4645,
+      "eval_samples_per_second": 44.241,
+      "eval_steps_per_second": 2.784,
+      "step": 1268
+    },
+    {
+      "epoch": 86.98,
+      "grad_norm": 1.5172946453094482,
+      "learning_rate": 2.1114285714285714e-05,
+      "loss": 0.4343,
+      "step": 1283
+    },
+    {
+      "epoch": 86.98,
+      "eval_accuracy": 0.7167832167832168,
+      "eval_loss": 0.78819739818573,
+      "eval_runtime": 5.3121,
+      "eval_samples_per_second": 53.839,
+      "eval_steps_per_second": 3.388,
+      "step": 1283
+    },
+    {
+      "epoch": 88.0,
+      "grad_norm": 1.8876270055770874,
+      "learning_rate": 2.0971428571428572e-05,
+      "loss": 0.4442,
+      "step": 1298
+    },
+    {
+      "epoch": 88.0,
+      "eval_accuracy": 0.7132867132867133,
+      "eval_loss": 0.805809736251831,
+      "eval_runtime": 5.4992,
+      "eval_samples_per_second": 52.007,
+      "eval_steps_per_second": 3.273,
+      "step": 1298
+    },
+    {
+      "epoch": 88.95,
+      "grad_norm": 1.4128847122192383,
+      "learning_rate": 2.0838095238095238e-05,
+      "loss": 0.4745,
+      "step": 1312
+    },
+    {
+      "epoch": 88.95,
+      "eval_accuracy": 0.7237762237762237,
+      "eval_loss": 0.7810028791427612,
+      "eval_runtime": 5.1967,
+      "eval_samples_per_second": 55.035,
+      "eval_steps_per_second": 3.464,
+      "step": 1312
+    },
+    {
+      "epoch": 89.97,
+      "grad_norm": 1.3974703550338745,
+      "learning_rate": 2.0695238095238096e-05,
+      "loss": 0.4282,
+      "step": 1327
+    },
+    {
+      "epoch": 89.97,
+      "eval_accuracy": 0.7097902097902098,
+      "eval_loss": 0.7951435446739197,
+      "eval_runtime": 5.9497,
+      "eval_samples_per_second": 48.07,
+      "eval_steps_per_second": 3.025,
+      "step": 1327
+    },
+    {
+      "epoch": 90.98,
+      "grad_norm": 1.8035740852355957,
+      "learning_rate": 2.055238095238095e-05,
+      "loss": 0.4307,
+      "step": 1342
+    },
+    {
+      "epoch": 90.98,
+      "eval_accuracy": 0.7167832167832168,
+      "eval_loss": 0.7739275097846985,
+      "eval_runtime": 5.0539,
+      "eval_samples_per_second": 56.59,
+      "eval_steps_per_second": 3.562,
+      "step": 1342
+    },
+    {
+      "epoch": 92.0,
+      "grad_norm": 2.104257822036743,
+      "learning_rate": 2.040952380952381e-05,
+      "loss": 0.4403,
+      "step": 1357
+    },
+    {
+      "epoch": 92.0,
+      "eval_accuracy": 0.7202797202797203,
+      "eval_loss": 0.7788061499595642,
+      "eval_runtime": 4.9104,
+      "eval_samples_per_second": 58.244,
+      "eval_steps_per_second": 3.666,
+      "step": 1357
+    },
+    {
+      "epoch": 92.95,
+      "grad_norm": 1.6670126914978027,
+      "learning_rate": 2.0276190476190475e-05,
+      "loss": 0.4567,
+      "step": 1371
+    },
+    {
+      "epoch": 92.95,
+      "eval_accuracy": 0.7167832167832168,
+      "eval_loss": 0.7926862835884094,
+      "eval_runtime": 5.6039,
+      "eval_samples_per_second": 51.036,
+      "eval_steps_per_second": 3.212,
+      "step": 1371
+    },
+    {
+      "epoch": 93.97,
+      "grad_norm": 0.9627218842506409,
+      "learning_rate": 2.0133333333333333e-05,
+      "loss": 0.4233,
+      "step": 1386
+    },
+    {
+      "epoch": 93.97,
+      "eval_accuracy": 0.7202797202797203,
+      "eval_loss": 0.7885140180587769,
+      "eval_runtime": 4.3051,
+      "eval_samples_per_second": 66.433,
+      "eval_steps_per_second": 4.181,
+      "step": 1386
+    },
+    {
+      "epoch": 94.98,
+      "grad_norm": 1.5669583082199097,
+      "learning_rate": 1.999047619047619e-05,
+      "loss": 0.4347,
+      "step": 1401
+    },
+    {
+      "epoch": 94.98,
+      "eval_accuracy": 0.7202797202797203,
+      "eval_loss": 0.7849389314651489,
+      "eval_runtime": 4.7453,
+      "eval_samples_per_second": 60.27,
+      "eval_steps_per_second": 3.793,
+      "step": 1401
+    },
+    {
+      "epoch": 96.0,
+      "grad_norm": 1.5593161582946777,
+      "learning_rate": 1.984761904761905e-05,
+      "loss": 0.4167,
+      "step": 1416
+    },
+    {
+      "epoch": 96.0,
+      "eval_accuracy": 0.7237762237762237,
+      "eval_loss": 0.7880135774612427,
+      "eval_runtime": 5.155,
+      "eval_samples_per_second": 55.48,
+      "eval_steps_per_second": 3.492,
+      "step": 1416
+    },
+    {
+      "epoch": 96.95,
+      "grad_norm": 1.337389349937439,
+      "learning_rate": 1.9714285714285714e-05,
+      "loss": 0.4394,
+      "step": 1430
+    },
+    {
+      "epoch": 96.95,
+      "eval_accuracy": 0.7202797202797203,
+      "eval_loss": 0.7888504862785339,
+      "eval_runtime": 4.6698,
+      "eval_samples_per_second": 61.244,
+      "eval_steps_per_second": 3.855,
+      "step": 1430
+    },
+    {
+      "epoch": 97.97,
+      "grad_norm": 1.8546737432479858,
+      "learning_rate": 1.9571428571428572e-05,
+      "loss": 0.4359,
+      "step": 1445
+    },
+    {
+      "epoch": 97.97,
+      "eval_accuracy": 0.7202797202797203,
+      "eval_loss": 0.778459370136261,
+      "eval_runtime": 4.7578,
+      "eval_samples_per_second": 60.112,
+      "eval_steps_per_second": 3.783,
+      "step": 1445
+    },
+    {
+      "epoch": 98.98,
+      "grad_norm": 1.3580291271209717,
+      "learning_rate": 1.942857142857143e-05,
+      "loss": 0.4085,
+      "step": 1460
+    },
+    {
+      "epoch": 98.98,
+      "eval_accuracy": 0.7132867132867133,
+      "eval_loss": 0.7851797342300415,
+      "eval_runtime": 5.8892,
+      "eval_samples_per_second": 48.564,
+      "eval_steps_per_second": 3.056,
+      "step": 1460
+    },
+    {
+      "epoch": 100.0,
+      "grad_norm": 1.4461547136306763,
+      "learning_rate": 1.928571428571429e-05,
+      "loss": 0.3965,
+      "step": 1475
+    },
+    {
+      "epoch": 100.0,
+      "eval_accuracy": 0.7272727272727273,
+      "eval_loss": 0.7784654498100281,
+      "eval_runtime": 5.3557,
+      "eval_samples_per_second": 53.401,
+      "eval_steps_per_second": 3.361,
+      "step": 1475
+    },
+    {
+      "epoch": 100.95,
+      "grad_norm": 1.765655755996704,
+      "learning_rate": 1.9152380952380954e-05,
+      "loss": 0.445,
+      "step": 1489
+    },
+    {
+      "epoch": 100.95,
+      "eval_accuracy": 0.7202797202797203,
+      "eval_loss": 0.7826104760169983,
+      "eval_runtime": 5.628,
+      "eval_samples_per_second": 50.818,
+      "eval_steps_per_second": 3.198,
+      "step": 1489
+    },
+    {
+      "epoch": 101.97,
+      "grad_norm": 1.7069743871688843,
+      "learning_rate": 1.9009523809523812e-05,
+      "loss": 0.3988,
+      "step": 1504
+    },
+    {
+      "epoch": 101.97,
+      "eval_accuracy": 0.7097902097902098,
+      "eval_loss": 0.8045200109481812,
+      "eval_runtime": 5.2204,
+      "eval_samples_per_second": 54.785,
+      "eval_steps_per_second": 3.448,
+      "step": 1504
+    },
+    {
+      "epoch": 102.98,
+      "grad_norm": 1.2846873998641968,
+      "learning_rate": 1.886666666666667e-05,
+      "loss": 0.4129,
+      "step": 1519
+    },
+    {
+      "epoch": 102.98,
+      "eval_accuracy": 0.7272727272727273,
+      "eval_loss": 0.7685949206352234,
+      "eval_runtime": 5.3375,
+      "eval_samples_per_second": 53.583,
+      "eval_steps_per_second": 3.372,
+      "step": 1519
+    },
+    {
+      "epoch": 104.0,
+      "grad_norm": 1.9040664434432983,
+      "learning_rate": 1.872380952380952e-05,
+      "loss": 0.3937,
+      "step": 1534
+    },
+    {
+      "epoch": 104.0,
+      "eval_accuracy": 0.7132867132867133,
+      "eval_loss": 0.7911521196365356,
+      "eval_runtime": 5.5547,
+      "eval_samples_per_second": 51.488,
+      "eval_steps_per_second": 3.24,
+      "step": 1534
+    },
+    {
+      "epoch": 104.95,
+      "grad_norm": 1.998403549194336,
+      "learning_rate": 1.8590476190476194e-05,
+      "loss": 0.4356,
+      "step": 1548
+    },
+    {
+      "epoch": 104.95,
+      "eval_accuracy": 0.7132867132867133,
+      "eval_loss": 0.7922284007072449,
+      "eval_runtime": 5.247,
+      "eval_samples_per_second": 54.507,
+      "eval_steps_per_second": 3.431,
+      "step": 1548
+    },
+    {
+      "epoch": 105.97,
+      "grad_norm": 1.3670754432678223,
+      "learning_rate": 1.8447619047619045e-05,
+      "loss": 0.3969,
+      "step": 1563
+    },
+    {
+      "epoch": 105.97,
+      "eval_accuracy": 0.7202797202797203,
+      "eval_loss": 0.7751501202583313,
+      "eval_runtime": 4.5814,
+      "eval_samples_per_second": 62.427,
+      "eval_steps_per_second": 3.929,
+      "step": 1563
+    },
+    {
+      "epoch": 106.98,
+      "grad_norm": 1.0057995319366455,
+      "learning_rate": 1.8304761904761903e-05,
+      "loss": 0.4051,
+      "step": 1578
+    },
+    {
+      "epoch": 106.98,
+      "eval_accuracy": 0.7132867132867133,
+      "eval_loss": 0.7917311787605286,
+      "eval_runtime": 5.1261,
+      "eval_samples_per_second": 55.793,
+      "eval_steps_per_second": 3.511,
+      "step": 1578
+    },
+    {
+      "epoch": 108.0,
+      "grad_norm": 1.7176306247711182,
+      "learning_rate": 1.816190476190476e-05,
+      "loss": 0.3982,
+      "step": 1593
+    },
+    {
+      "epoch": 108.0,
+      "eval_accuracy": 0.7097902097902098,
+      "eval_loss": 0.7916986346244812,
+      "eval_runtime": 6.2925,
+      "eval_samples_per_second": 45.451,
+      "eval_steps_per_second": 2.861,
+      "step": 1593
+    },
+    {
+      "epoch": 108.95,
+      "grad_norm": 1.6695342063903809,
+      "learning_rate": 1.8028571428571427e-05,
+      "loss": 0.4117,
+      "step": 1607
+    },
+    {
+      "epoch": 108.95,
+      "eval_accuracy": 0.7062937062937062,
+      "eval_loss": 0.8070589303970337,
+      "eval_runtime": 4.8482,
+      "eval_samples_per_second": 58.99,
+      "eval_steps_per_second": 3.713,
+      "step": 1607
+    },
+    {
+      "epoch": 109.97,
+      "grad_norm": 1.1871509552001953,
+      "learning_rate": 1.7885714285714285e-05,
+      "loss": 0.3666,
+      "step": 1622
+    },
+    {
+      "epoch": 109.97,
+      "eval_accuracy": 0.7202797202797203,
+      "eval_loss": 0.7839590907096863,
+      "eval_runtime": 5.4971,
+      "eval_samples_per_second": 52.027,
+      "eval_steps_per_second": 3.274,
+      "step": 1622
+    },
+    {
+      "epoch": 110.98,
+      "grad_norm": 2.196869134902954,
+      "learning_rate": 1.7742857142857143e-05,
+      "loss": 0.3894,
+      "step": 1637
+    },
+    {
+      "epoch": 110.98,
+      "eval_accuracy": 0.7237762237762237,
+      "eval_loss": 0.7790002226829529,
+      "eval_runtime": 6.5509,
+      "eval_samples_per_second": 43.658,
+      "eval_steps_per_second": 2.748,
+      "step": 1637
+    },
+    {
+      "epoch": 112.0,
+      "grad_norm": 2.341435432434082,
+      "learning_rate": 1.76e-05,
+      "loss": 0.3858,
+      "step": 1652
+    },
+    {
+      "epoch": 112.0,
+      "eval_accuracy": 0.7097902097902098,
+      "eval_loss": 0.7961041927337646,
+      "eval_runtime": 4.8263,
+      "eval_samples_per_second": 59.258,
+      "eval_steps_per_second": 3.73,
+      "step": 1652
+    },
+    {
+      "epoch": 112.95,
+      "grad_norm": 1.5795267820358276,
+      "learning_rate": 1.7466666666666667e-05,
+      "loss": 0.4037,
+      "step": 1666
+    },
+    {
+      "epoch": 112.95,
+      "eval_accuracy": 0.7202797202797203,
+      "eval_loss": 0.7822238802909851,
+      "eval_runtime": 5.6008,
+      "eval_samples_per_second": 51.064,
+      "eval_steps_per_second": 3.214,
+      "step": 1666
+    },
+    {
+      "epoch": 113.97,
+      "grad_norm": 1.5341583490371704,
+      "learning_rate": 1.7333333333333332e-05,
+      "loss": 0.3886,
+      "step": 1681
+    },
+    {
+      "epoch": 113.97,
+      "eval_accuracy": 0.7237762237762237,
+      "eval_loss": 0.7748440504074097,
+      "eval_runtime": 4.7211,
+      "eval_samples_per_second": 60.579,
+      "eval_steps_per_second": 3.813,
+      "step": 1681
+    },
+    {
+      "epoch": 114.98,
+      "grad_norm": 1.056776523590088,
+      "learning_rate": 1.719047619047619e-05,
+      "loss": 0.3762,
+      "step": 1696
+    },
+    {
+      "epoch": 114.98,
+      "eval_accuracy": 0.7167832167832168,
+      "eval_loss": 0.7781729102134705,
+      "eval_runtime": 6.5039,
+      "eval_samples_per_second": 43.974,
+      "eval_steps_per_second": 2.768,
+      "step": 1696
+    },
+    {
+      "epoch": 116.0,
+      "grad_norm": 2.7518413066864014,
+      "learning_rate": 1.704761904761905e-05,
+      "loss": 0.3444,
+      "step": 1711
+    },
+    {
+      "epoch": 116.0,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 0.7745847702026367,
+      "eval_runtime": 6.1348,
+      "eval_samples_per_second": 46.619,
+      "eval_steps_per_second": 2.934,
+      "step": 1711
+    },
+    {
+      "epoch": 116.95,
+      "grad_norm": 1.8518308401107788,
+      "learning_rate": 1.6914285714285714e-05,
+      "loss": 0.3961,
+      "step": 1725
+    },
+    {
+      "epoch": 116.95,
+      "eval_accuracy": 0.7202797202797203,
+      "eval_loss": 0.7842003703117371,
+      "eval_runtime": 4.8863,
+      "eval_samples_per_second": 58.531,
+      "eval_steps_per_second": 3.684,
+      "step": 1725
+    },
+    {
+      "epoch": 117.97,
+      "grad_norm": 1.487108588218689,
+      "learning_rate": 1.6771428571428572e-05,
+      "loss": 0.3578,
+      "step": 1740
+    },
+    {
+      "epoch": 117.97,
+      "eval_accuracy": 0.7202797202797203,
+      "eval_loss": 0.7819164991378784,
+      "eval_runtime": 5.6099,
+      "eval_samples_per_second": 50.981,
+      "eval_steps_per_second": 3.209,
+      "step": 1740
+    },
+    {
+      "epoch": 118.98,
+      "grad_norm": 1.7816615104675293,
+      "learning_rate": 1.662857142857143e-05,
+      "loss": 0.3578,
+      "step": 1755
+    },
+    {
+      "epoch": 118.98,
+      "eval_accuracy": 0.7202797202797203,
+      "eval_loss": 0.7805832624435425,
+      "eval_runtime": 5.5595,
+      "eval_samples_per_second": 51.443,
+      "eval_steps_per_second": 3.238,
+      "step": 1755
+    },
+    {
+      "epoch": 120.0,
+      "grad_norm": 2.4715042114257812,
+      "learning_rate": 1.6485714285714285e-05,
+      "loss": 0.3489,
+      "step": 1770
+    },
+    {
+      "epoch": 120.0,
+      "eval_accuracy": 0.7237762237762237,
+      "eval_loss": 0.7809211015701294,
+      "eval_runtime": 6.1205,
+      "eval_samples_per_second": 46.728,
+      "eval_steps_per_second": 2.941,
+      "step": 1770
+    },
+    {
+      "epoch": 120.95,
+      "grad_norm": 1.1066619157791138,
+      "learning_rate": 1.635238095238095e-05,
+      "loss": 0.3622,
+      "step": 1784
+    },
+    {
+      "epoch": 120.95,
+      "eval_accuracy": 0.7097902097902098,
+      "eval_loss": 0.7947035431861877,
+      "eval_runtime": 5.0722,
+      "eval_samples_per_second": 56.386,
+      "eval_steps_per_second": 3.549,
+      "step": 1784
+    },
+    {
+      "epoch": 121.97,
+      "grad_norm": 1.3460161685943604,
+      "learning_rate": 1.620952380952381e-05,
+      "loss": 0.3545,
+      "step": 1799
+    },
+    {
+      "epoch": 121.97,
+      "eval_accuracy": 0.7167832167832168,
+      "eval_loss": 0.7877638339996338,
+      "eval_runtime": 5.4395,
+      "eval_samples_per_second": 52.578,
+      "eval_steps_per_second": 3.309,
+      "step": 1799
+    },
+    {
+      "epoch": 122.98,
+      "grad_norm": 1.7303364276885986,
+      "learning_rate": 1.6066666666666666e-05,
+      "loss": 0.3361,
+      "step": 1814
+    },
+    {
+      "epoch": 122.98,
+      "eval_accuracy": 0.7167832167832168,
+      "eval_loss": 0.7854802012443542,
+      "eval_runtime": 4.9544,
+      "eval_samples_per_second": 57.726,
+      "eval_steps_per_second": 3.633,
+      "step": 1814
+    },
+    {
+      "epoch": 124.0,
+      "grad_norm": 1.5440623760223389,
+      "learning_rate": 1.5923809523809524e-05,
+      "loss": 0.3618,
+      "step": 1829
+    },
+    {
+      "epoch": 124.0,
+      "eval_accuracy": 0.7132867132867133,
+      "eval_loss": 0.7890083193778992,
+      "eval_runtime": 4.6202,
+      "eval_samples_per_second": 61.902,
+      "eval_steps_per_second": 3.896,
+      "step": 1829
+    },
+    {
+      "epoch": 124.95,
+      "grad_norm": 1.2985795736312866,
+      "learning_rate": 1.579047619047619e-05,
+      "loss": 0.3472,
+      "step": 1843
+    },
+    {
+      "epoch": 124.95,
+      "eval_accuracy": 0.7167832167832168,
+      "eval_loss": 0.7809839844703674,
+      "eval_runtime": 4.5623,
+      "eval_samples_per_second": 62.687,
+      "eval_steps_per_second": 3.945,
+      "step": 1843
+    },
+    {
+      "epoch": 125.97,
+      "grad_norm": 1.3930552005767822,
+      "learning_rate": 1.5647619047619048e-05,
+      "loss": 0.3511,
+      "step": 1858
+    },
+    {
+      "epoch": 125.97,
+      "eval_accuracy": 0.7132867132867133,
+      "eval_loss": 0.7897189259529114,
+      "eval_runtime": 5.1961,
+      "eval_samples_per_second": 55.041,
+      "eval_steps_per_second": 3.464,
+      "step": 1858
+    },
+    {
+      "epoch": 126.98,
+      "grad_norm": 2.0258781909942627,
+      "learning_rate": 1.5504761904761906e-05,
+      "loss": 0.3389,
+      "step": 1873
+    },
+    {
+      "epoch": 126.98,
+      "eval_accuracy": 0.7132867132867133,
+      "eval_loss": 0.7922906875610352,
+      "eval_runtime": 5.0606,
+      "eval_samples_per_second": 56.515,
+      "eval_steps_per_second": 3.557,
+      "step": 1873
+    },
+    {
+      "epoch": 128.0,
+      "grad_norm": 1.0623548030853271,
+      "learning_rate": 1.5361904761904764e-05,
+      "loss": 0.3391,
+      "step": 1888
+    },
+    {
+      "epoch": 128.0,
+      "eval_accuracy": 0.7272727272727273,
+      "eval_loss": 0.7781790494918823,
+      "eval_runtime": 4.5848,
+      "eval_samples_per_second": 62.38,
+      "eval_steps_per_second": 3.926,
+      "step": 1888
+    },
+    {
+      "epoch": 128.95,
+      "grad_norm": 1.2839761972427368,
+      "learning_rate": 1.522857142857143e-05,
+      "loss": 0.3746,
+      "step": 1902
+    },
+    {
+      "epoch": 128.95,
+      "eval_accuracy": 0.7202797202797203,
+      "eval_loss": 0.7838397026062012,
+      "eval_runtime": 6.371,
+      "eval_samples_per_second": 44.891,
+      "eval_steps_per_second": 2.825,
+      "step": 1902
+    },
+    {
+      "epoch": 129.97,
+      "grad_norm": 1.7424538135528564,
+      "learning_rate": 1.5085714285714288e-05,
+      "loss": 0.3238,
+      "step": 1917
+    },
+    {
+      "epoch": 129.97,
+      "eval_accuracy": 0.7167832167832168,
+      "eval_loss": 0.7943305969238281,
+      "eval_runtime": 5.0078,
+      "eval_samples_per_second": 57.111,
+      "eval_steps_per_second": 3.594,
+      "step": 1917
+    },
+    {
+      "epoch": 130.98,
+      "grad_norm": 1.9641584157943726,
+      "learning_rate": 1.4942857142857143e-05,
+      "loss": 0.3601,
+      "step": 1932
+    },
+    {
+      "epoch": 130.98,
+      "eval_accuracy": 0.7167832167832168,
+      "eval_loss": 0.786338746547699,
+      "eval_runtime": 5.7221,
+      "eval_samples_per_second": 49.982,
+      "eval_steps_per_second": 3.146,
+      "step": 1932
+    },
+    {
+      "epoch": 132.0,
+      "grad_norm": 1.6700012683868408,
+      "learning_rate": 1.48e-05,
+      "loss": 0.3339,
+      "step": 1947
+    },
+    {
+      "epoch": 132.0,
+      "eval_accuracy": 0.7132867132867133,
+      "eval_loss": 0.7948570847511292,
+      "eval_runtime": 4.5956,
+      "eval_samples_per_second": 62.234,
+      "eval_steps_per_second": 3.917,
+      "step": 1947
+    },
+    {
+      "epoch": 132.95,
+      "grad_norm": 1.4314067363739014,
+      "learning_rate": 1.4666666666666666e-05,
+      "loss": 0.3805,
+      "step": 1961
+    },
+    {
+      "epoch": 132.95,
+      "eval_accuracy": 0.7237762237762237,
+      "eval_loss": 0.7823219299316406,
+      "eval_runtime": 5.3022,
+      "eval_samples_per_second": 53.94,
+      "eval_steps_per_second": 3.395,
+      "step": 1961
+    },
+    {
+      "epoch": 133.97,
+      "grad_norm": 1.9961072206497192,
+      "learning_rate": 1.4523809523809524e-05,
+      "loss": 0.3524,
+      "step": 1976
+    },
+    {
+      "epoch": 133.97,
+      "eval_accuracy": 0.7097902097902098,
+      "eval_loss": 0.8052372932434082,
+      "eval_runtime": 4.8993,
+      "eval_samples_per_second": 58.375,
+      "eval_steps_per_second": 3.674,
+      "step": 1976
+    },
+    {
+      "epoch": 134.98,
+      "grad_norm": 1.1320672035217285,
+      "learning_rate": 1.4380952380952382e-05,
+      "loss": 0.3103,
+      "step": 1991
+    },
+    {
+      "epoch": 134.98,
+      "eval_accuracy": 0.7237762237762237,
+      "eval_loss": 0.7809023261070251,
+      "eval_runtime": 5.8892,
+      "eval_samples_per_second": 48.563,
+      "eval_steps_per_second": 3.056,
+      "step": 1991
+    },
+    {
+      "epoch": 136.0,
+      "grad_norm": 2.298990249633789,
+      "learning_rate": 1.4238095238095239e-05,
+      "loss": 0.3484,
+      "step": 2006
+    },
+    {
+      "epoch": 136.0,
+      "eval_accuracy": 0.7202797202797203,
+      "eval_loss": 0.787878692150116,
+      "eval_runtime": 5.2761,
+      "eval_samples_per_second": 54.207,
+      "eval_steps_per_second": 3.412,
+      "step": 2006
+    },
+    {
+      "epoch": 136.95,
+      "grad_norm": 2.3017499446868896,
+      "learning_rate": 1.4104761904761906e-05,
+      "loss": 0.3424,
+      "step": 2020
+    },
+    {
+      "epoch": 136.95,
+      "eval_accuracy": 0.7272727272727273,
+      "eval_loss": 0.7875267267227173,
+      "eval_runtime": 4.7804,
+      "eval_samples_per_second": 59.828,
+      "eval_steps_per_second": 3.765,
+      "step": 2020
+    },
+    {
+      "epoch": 137.97,
+      "grad_norm": 1.4992213249206543,
+      "learning_rate": 1.3961904761904762e-05,
+      "loss": 0.316,
+      "step": 2035
+    },
+    {
+      "epoch": 137.97,
+      "eval_accuracy": 0.7272727272727273,
+      "eval_loss": 0.7829337120056152,
+      "eval_runtime": 4.9522,
+      "eval_samples_per_second": 57.752,
+      "eval_steps_per_second": 3.635,
+      "step": 2035
+    },
+    {
+      "epoch": 138.98,
+      "grad_norm": 1.634419560432434,
+      "learning_rate": 1.3819047619047619e-05,
+      "loss": 0.3171,
+      "step": 2050
+    },
+    {
+      "epoch": 138.98,
+      "eval_accuracy": 0.7202797202797203,
+      "eval_loss": 0.7882408499717712,
+      "eval_runtime": 4.8523,
+      "eval_samples_per_second": 58.941,
+      "eval_steps_per_second": 3.71,
+      "step": 2050
+    },
+    {
+      "epoch": 140.0,
+      "grad_norm": 1.2980549335479736,
+      "learning_rate": 1.3676190476190477e-05,
+      "loss": 0.3155,
+      "step": 2065
+    },
+    {
+      "epoch": 140.0,
+      "eval_accuracy": 0.7167832167832168,
+      "eval_loss": 0.7830457091331482,
+      "eval_runtime": 4.9611,
+      "eval_samples_per_second": 57.649,
+      "eval_steps_per_second": 3.628,
+      "step": 2065
+    },
+    {
+      "epoch": 140.95,
+      "grad_norm": 1.5925828218460083,
+      "learning_rate": 1.3542857142857142e-05,
+      "loss": 0.3382,
+      "step": 2079
+    },
+    {
+      "epoch": 140.95,
+      "eval_accuracy": 0.7272727272727273,
+      "eval_loss": 0.7825762033462524,
+      "eval_runtime": 4.6558,
+      "eval_samples_per_second": 61.428,
+      "eval_steps_per_second": 3.866,
+      "step": 2079
+    },
+    {
+      "epoch": 141.97,
+      "grad_norm": 1.9007666110992432,
+      "learning_rate": 1.34e-05,
+      "loss": 0.3175,
+      "step": 2094
+    },
+    {
+      "epoch": 141.97,
+      "eval_accuracy": 0.7202797202797203,
+      "eval_loss": 0.7963916659355164,
+      "eval_runtime": 5.2803,
+      "eval_samples_per_second": 54.164,
+      "eval_steps_per_second": 3.409,
+      "step": 2094
+    },
+    {
+      "epoch": 142.98,
+      "grad_norm": 1.8862413167953491,
+      "learning_rate": 1.3257142857142858e-05,
+      "loss": 0.3444,
+      "step": 2109
+    },
+    {
+      "epoch": 142.98,
+      "eval_accuracy": 0.7237762237762237,
+      "eval_loss": 0.7858501672744751,
+      "eval_runtime": 5.4501,
+      "eval_samples_per_second": 52.476,
+      "eval_steps_per_second": 3.303,
+      "step": 2109
+    },
+    {
+      "epoch": 144.0,
+      "grad_norm": 2.4451475143432617,
+      "learning_rate": 1.3114285714285715e-05,
+      "loss": 0.3208,
+      "step": 2124
+    },
+    {
+      "epoch": 144.0,
+      "eval_accuracy": 0.7272727272727273,
+      "eval_loss": 0.7859659194946289,
+      "eval_runtime": 4.5157,
+      "eval_samples_per_second": 63.334,
+      "eval_steps_per_second": 3.986,
+      "step": 2124
+    },
+    {
+      "epoch": 144.95,
+      "grad_norm": 1.6911462545394897,
+      "learning_rate": 1.298095238095238e-05,
+      "loss": 0.3286,
+      "step": 2138
+    },
+    {
+      "epoch": 144.95,
+      "eval_accuracy": 0.7272727272727273,
+      "eval_loss": 0.7869133353233337,
+      "eval_runtime": 5.6207,
+      "eval_samples_per_second": 50.883,
+      "eval_steps_per_second": 3.202,
+      "step": 2138
+    },
+    {
+      "epoch": 145.97,
+      "grad_norm": 1.9792907238006592,
+      "learning_rate": 1.2838095238095239e-05,
+      "loss": 0.3319,
+      "step": 2153
+    },
+    {
+      "epoch": 145.97,
+      "eval_accuracy": 0.7167832167832168,
+      "eval_loss": 0.7916193604469299,
+      "eval_runtime": 5.434,
+      "eval_samples_per_second": 52.631,
+      "eval_steps_per_second": 3.312,
+      "step": 2153
+    },
+    {
+      "epoch": 146.98,
+      "grad_norm": 1.3403879404067993,
+      "learning_rate": 1.2695238095238095e-05,
+      "loss": 0.2954,
+      "step": 2168
+    },
+    {
+      "epoch": 146.98,
+      "eval_accuracy": 0.7237762237762237,
+      "eval_loss": 0.7937904000282288,
+      "eval_runtime": 4.5934,
+      "eval_samples_per_second": 62.264,
+      "eval_steps_per_second": 3.919,
+      "step": 2168
+    },
+    {
+      "epoch": 148.0,
+      "grad_norm": 2.6414599418640137,
+      "learning_rate": 1.2552380952380953e-05,
+      "loss": 0.3283,
+      "step": 2183
+    },
+    {
+      "epoch": 148.0,
+      "eval_accuracy": 0.7167832167832168,
+      "eval_loss": 0.7974053025245667,
+      "eval_runtime": 4.6622,
+      "eval_samples_per_second": 61.344,
+      "eval_steps_per_second": 3.861,
+      "step": 2183
+    },
+    {
+      "epoch": 148.95,
+      "grad_norm": 1.7404941320419312,
+      "learning_rate": 1.241904761904762e-05,
+      "loss": 0.3306,
+      "step": 2197
+    },
+    {
+      "epoch": 148.95,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 0.7795438766479492,
+      "eval_runtime": 4.8202,
+      "eval_samples_per_second": 59.334,
+      "eval_steps_per_second": 3.734,
+      "step": 2197
+    },
+    {
+      "epoch": 149.97,
+      "grad_norm": 1.3558377027511597,
+      "learning_rate": 1.2276190476190477e-05,
+      "loss": 0.3073,
+      "step": 2212
+    },
+    {
+      "epoch": 149.97,
+      "eval_accuracy": 0.7202797202797203,
+      "eval_loss": 0.7910019755363464,
+      "eval_runtime": 5.1476,
+      "eval_samples_per_second": 55.56,
+      "eval_steps_per_second": 3.497,
+      "step": 2212
+    },
+    {
+      "epoch": 150.98,
+      "grad_norm": 1.2379992008209229,
+      "learning_rate": 1.2133333333333333e-05,
+      "loss": 0.3089,
+      "step": 2227
+    },
+    {
+      "epoch": 150.98,
+      "eval_accuracy": 0.7202797202797203,
+      "eval_loss": 0.7942177653312683,
+      "eval_runtime": 5.4133,
+      "eval_samples_per_second": 52.833,
+      "eval_steps_per_second": 3.325,
+      "step": 2227
+    },
+    {
+      "epoch": 152.0,
+      "grad_norm": 1.3452534675598145,
+      "learning_rate": 1.1990476190476191e-05,
+      "loss": 0.2915,
+      "step": 2242
+    },
+    {
+      "epoch": 152.0,
+      "eval_accuracy": 0.7167832167832168,
+      "eval_loss": 0.7933531403541565,
+      "eval_runtime": 6.3553,
+      "eval_samples_per_second": 45.002,
+      "eval_steps_per_second": 2.832,
+      "step": 2242
+    },
+    {
+      "epoch": 152.95,
+      "grad_norm": 0.9324106574058533,
+      "learning_rate": 1.1857142857142857e-05,
+      "loss": 0.3286,
+      "step": 2256
+    },
+    {
+      "epoch": 152.95,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 0.7807846069335938,
+      "eval_runtime": 5.094,
+      "eval_samples_per_second": 56.144,
+      "eval_steps_per_second": 3.534,
+      "step": 2256
+    },
+    {
+      "epoch": 153.97,
+      "grad_norm": 1.5133529901504517,
+      "learning_rate": 1.1714285714285715e-05,
+      "loss": 0.2817,
+      "step": 2271
+    },
+    {
+      "epoch": 153.97,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 0.7788205742835999,
+      "eval_runtime": 5.0531,
+      "eval_samples_per_second": 56.599,
+      "eval_steps_per_second": 3.562,
+      "step": 2271
+    },
+    {
+      "epoch": 154.98,
+      "grad_norm": 2.8459701538085938,
+      "learning_rate": 1.1571428571428573e-05,
+      "loss": 0.3118,
+      "step": 2286
+    },
+    {
+      "epoch": 154.98,
+      "eval_accuracy": 0.7272727272727273,
+      "eval_loss": 0.7898407578468323,
+      "eval_runtime": 4.8393,
+      "eval_samples_per_second": 59.1,
+      "eval_steps_per_second": 3.72,
+      "step": 2286
+    },
+    {
+      "epoch": 156.0,
+      "grad_norm": 1.8526194095611572,
+      "learning_rate": 1.1428571428571429e-05,
+      "loss": 0.3155,
+      "step": 2301
+    },
+    {
+      "epoch": 156.0,
+      "eval_accuracy": 0.7202797202797203,
+      "eval_loss": 0.796588122844696,
+      "eval_runtime": 5.8687,
+      "eval_samples_per_second": 48.733,
+      "eval_steps_per_second": 3.067,
+      "step": 2301
+    },
+    {
+      "epoch": 156.95,
+      "grad_norm": 1.130321741104126,
+      "learning_rate": 1.1295238095238096e-05,
+      "loss": 0.3156,
+      "step": 2315
+    },
+    {
+      "epoch": 156.95,
+      "eval_accuracy": 0.7202797202797203,
+      "eval_loss": 0.7947112321853638,
+      "eval_runtime": 4.5714,
+      "eval_samples_per_second": 62.564,
+      "eval_steps_per_second": 3.938,
+      "step": 2315
+    },
+    {
+      "epoch": 157.97,
+      "grad_norm": 1.3919442892074585,
+      "learning_rate": 1.1152380952380953e-05,
+      "loss": 0.2936,
+      "step": 2330
+    },
+    {
+      "epoch": 157.97,
+      "eval_accuracy": 0.7167832167832168,
+      "eval_loss": 0.7916660904884338,
+      "eval_runtime": 4.554,
+      "eval_samples_per_second": 62.802,
+      "eval_steps_per_second": 3.953,
+      "step": 2330
+    },
+    {
+      "epoch": 158.98,
+      "grad_norm": 2.1628525257110596,
+      "learning_rate": 1.1009523809523809e-05,
+      "loss": 0.3049,
+      "step": 2345
+    },
+    {
+      "epoch": 158.98,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 0.7779849171638489,
+      "eval_runtime": 5.5631,
+      "eval_samples_per_second": 51.41,
+      "eval_steps_per_second": 3.236,
+      "step": 2345
+    },
+    {
+      "epoch": 160.0,
+      "grad_norm": 1.0748810768127441,
+      "learning_rate": 1.0866666666666667e-05,
+      "loss": 0.2896,
+      "step": 2360
+    },
+    {
+      "epoch": 160.0,
+      "eval_accuracy": 0.7272727272727273,
+      "eval_loss": 0.792616605758667,
+      "eval_runtime": 5.3985,
+      "eval_samples_per_second": 52.978,
+      "eval_steps_per_second": 3.334,
+      "step": 2360
+    },
+    {
+      "epoch": 160.95,
+      "grad_norm": 1.9327268600463867,
+      "learning_rate": 1.0733333333333333e-05,
+      "loss": 0.3194,
+      "step": 2374
+    },
+    {
+      "epoch": 160.95,
+      "eval_accuracy": 0.7202797202797203,
+      "eval_loss": 0.8022683262825012,
+      "eval_runtime": 5.5802,
+      "eval_samples_per_second": 51.253,
+      "eval_steps_per_second": 3.226,
+      "step": 2374
+    },
+    {
+      "epoch": 161.97,
+      "grad_norm": 1.610187292098999,
+      "learning_rate": 1.059047619047619e-05,
+      "loss": 0.2918,
+      "step": 2389
+    },
+    {
+      "epoch": 161.97,
+      "eval_accuracy": 0.7272727272727273,
+      "eval_loss": 0.7933264374732971,
+      "eval_runtime": 5.0333,
+      "eval_samples_per_second": 56.821,
+      "eval_steps_per_second": 3.576,
+      "step": 2389
+    },
+    {
+      "epoch": 162.98,
+      "grad_norm": 1.961053729057312,
+      "learning_rate": 1.0447619047619049e-05,
+      "loss": 0.2992,
+      "step": 2404
+    },
+    {
+      "epoch": 162.98,
+      "eval_accuracy": 0.7412587412587412,
+      "eval_loss": 0.7828559875488281,
+      "eval_runtime": 5.4812,
+      "eval_samples_per_second": 52.178,
+      "eval_steps_per_second": 3.284,
+      "step": 2404
+    },
+    {
+      "epoch": 164.0,
+      "grad_norm": 1.2120810747146606,
+      "learning_rate": 1.0304761904761905e-05,
+      "loss": 0.3,
+      "step": 2419
+    },
+    {
+      "epoch": 164.0,
+      "eval_accuracy": 0.7202797202797203,
+      "eval_loss": 0.7946493029594421,
+      "eval_runtime": 5.351,
+      "eval_samples_per_second": 53.448,
+      "eval_steps_per_second": 3.364,
+      "step": 2419
+    },
+    {
+      "epoch": 164.95,
+      "grad_norm": 1.112051010131836,
+      "learning_rate": 1.0171428571428571e-05,
+      "loss": 0.322,
+      "step": 2433
+    },
+    {
+      "epoch": 164.95,
+      "eval_accuracy": 0.7272727272727273,
+      "eval_loss": 0.79693204164505,
+      "eval_runtime": 6.2682,
+      "eval_samples_per_second": 45.627,
+      "eval_steps_per_second": 2.872,
+      "step": 2433
+    },
+    {
+      "epoch": 165.97,
+      "grad_norm": 1.2547581195831299,
+      "learning_rate": 1.0028571428571429e-05,
+      "loss": 0.2994,
+      "step": 2448
+    },
+    {
+      "epoch": 165.97,
+      "eval_accuracy": 0.7237762237762237,
+      "eval_loss": 0.8075766563415527,
+      "eval_runtime": 5.4676,
+      "eval_samples_per_second": 52.308,
+      "eval_steps_per_second": 3.292,
+      "step": 2448
+    },
+    {
+      "epoch": 166.98,
+      "grad_norm": 3.3027942180633545,
+      "learning_rate": 9.885714285714285e-06,
+      "loss": 0.2849,
+      "step": 2463
+    },
+    {
+      "epoch": 166.98,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 0.7950677275657654,
+      "eval_runtime": 5.6445,
+      "eval_samples_per_second": 50.668,
+      "eval_steps_per_second": 3.189,
+      "step": 2463
+    },
+    {
+      "epoch": 168.0,
+      "grad_norm": 3.2351794242858887,
+      "learning_rate": 9.742857142857143e-06,
+      "loss": 0.2745,
+      "step": 2478
+    },
+    {
+      "epoch": 168.0,
+      "eval_accuracy": 0.7342657342657343,
+      "eval_loss": 0.7892395853996277,
+      "eval_runtime": 6.6609,
+      "eval_samples_per_second": 42.937,
+      "eval_steps_per_second": 2.702,
+      "step": 2478
+    },
+    {
+      "epoch": 168.95,
+      "grad_norm": 2.2650949954986572,
+      "learning_rate": 9.60952380952381e-06,
+      "loss": 0.2974,
+      "step": 2492
+    },
+    {
+      "epoch": 168.95,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 0.791623055934906,
+      "eval_runtime": 5.8468,
+      "eval_samples_per_second": 48.916,
+      "eval_steps_per_second": 3.079,
+      "step": 2492
+    },
+    {
+      "epoch": 169.97,
+      "grad_norm": 1.3676173686981201,
+      "learning_rate": 9.466666666666667e-06,
+      "loss": 0.2656,
+      "step": 2507
+    },
+    {
+      "epoch": 169.97,
+      "eval_accuracy": 0.7342657342657343,
+      "eval_loss": 0.7994617223739624,
+      "eval_runtime": 4.7953,
+      "eval_samples_per_second": 59.642,
+      "eval_steps_per_second": 3.754,
+      "step": 2507
+    },
+    {
+      "epoch": 170.98,
+      "grad_norm": 1.5123904943466187,
+      "learning_rate": 9.323809523809523e-06,
+      "loss": 0.295,
+      "step": 2522
+    },
+    {
+      "epoch": 170.98,
+      "eval_accuracy": 0.7237762237762237,
+      "eval_loss": 0.802618682384491,
+      "eval_runtime": 6.3216,
+      "eval_samples_per_second": 45.241,
+      "eval_steps_per_second": 2.847,
+      "step": 2522
+    },
+    {
+      "epoch": 172.0,
+      "grad_norm": 1.5210902690887451,
+      "learning_rate": 9.180952380952381e-06,
+      "loss": 0.2791,
+      "step": 2537
+    },
+    {
+      "epoch": 172.0,
+      "eval_accuracy": 0.7342657342657343,
+      "eval_loss": 0.797250509262085,
+      "eval_runtime": 5.4447,
+      "eval_samples_per_second": 52.528,
+      "eval_steps_per_second": 3.306,
+      "step": 2537
+    },
+    {
+      "epoch": 172.95,
+      "grad_norm": 1.0823432207107544,
+      "learning_rate": 9.047619047619047e-06,
+      "loss": 0.2836,
+      "step": 2551
+    },
+    {
+      "epoch": 172.95,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 0.8022569417953491,
+      "eval_runtime": 5.0556,
+      "eval_samples_per_second": 56.571,
+      "eval_steps_per_second": 3.56,
+      "step": 2551
+    },
+    {
+      "epoch": 173.97,
+      "grad_norm": 1.4100669622421265,
+      "learning_rate": 8.904761904761905e-06,
+      "loss": 0.2806,
+      "step": 2566
+    },
+    {
+      "epoch": 173.97,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 0.8013490438461304,
+      "eval_runtime": 5.503,
+      "eval_samples_per_second": 51.972,
+      "eval_steps_per_second": 3.271,
+      "step": 2566
+    },
+    {
+      "epoch": 174.98,
+      "grad_norm": 1.389672875404358,
+      "learning_rate": 8.761904761904763e-06,
+      "loss": 0.2661,
+      "step": 2581
+    },
+    {
+      "epoch": 174.98,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 0.796493649482727,
+      "eval_runtime": 5.7105,
+      "eval_samples_per_second": 50.083,
+      "eval_steps_per_second": 3.152,
+      "step": 2581
+    },
+    {
+      "epoch": 176.0,
+      "grad_norm": 1.8702772855758667,
+      "learning_rate": 8.61904761904762e-06,
+      "loss": 0.2695,
+      "step": 2596
+    },
+    {
+      "epoch": 176.0,
+      "eval_accuracy": 0.7272727272727273,
+      "eval_loss": 0.8062567114830017,
+      "eval_runtime": 7.451,
+      "eval_samples_per_second": 38.384,
+      "eval_steps_per_second": 2.416,
+      "step": 2596
+    },
+    {
+      "epoch": 176.95,
+      "grad_norm": 3.131314992904663,
+      "learning_rate": 8.485714285714287e-06,
+      "loss": 0.286,
+      "step": 2610
+    },
+    {
+      "epoch": 176.95,
+      "eval_accuracy": 0.7237762237762237,
+      "eval_loss": 0.7963144779205322,
+      "eval_runtime": 4.431,
+      "eval_samples_per_second": 64.546,
+      "eval_steps_per_second": 4.062,
+      "step": 2610
+    },
+    {
+      "epoch": 177.97,
+      "grad_norm": 1.7147862911224365,
+      "learning_rate": 8.342857142857143e-06,
+      "loss": 0.2743,
+      "step": 2625
+    },
+    {
+      "epoch": 177.97,
+      "eval_accuracy": 0.7412587412587412,
+      "eval_loss": 0.7928534150123596,
+      "eval_runtime": 5.7357,
+      "eval_samples_per_second": 49.863,
+      "eval_steps_per_second": 3.138,
+      "step": 2625
+    },
+    {
+      "epoch": 178.98,
+      "grad_norm": 1.669243574142456,
+      "learning_rate": 8.2e-06,
+      "loss": 0.2775,
+      "step": 2640
+    },
+    {
+      "epoch": 178.98,
+      "eval_accuracy": 0.7447552447552448,
+      "eval_loss": 0.7855107188224792,
+      "eval_runtime": 5.7848,
+      "eval_samples_per_second": 49.44,
+      "eval_steps_per_second": 3.112,
+      "step": 2640
+    },
+    {
+      "epoch": 180.0,
+      "grad_norm": 2.0756781101226807,
+      "learning_rate": 8.057142857142857e-06,
+      "loss": 0.2878,
+      "step": 2655
+    },
+    {
+      "epoch": 180.0,
+      "eval_accuracy": 0.7377622377622378,
+      "eval_loss": 0.7894182801246643,
+      "eval_runtime": 6.034,
+      "eval_samples_per_second": 47.398,
+      "eval_steps_per_second": 2.983,
+      "step": 2655
+    },
+    {
+      "epoch": 180.95,
+      "grad_norm": 1.7884511947631836,
+      "learning_rate": 7.923809523809525e-06,
+      "loss": 0.2757,
+      "step": 2669
+    },
+    {
+      "epoch": 180.95,
+      "eval_accuracy": 0.7272727272727273,
+      "eval_loss": 0.8012902736663818,
+      "eval_runtime": 5.2896,
+      "eval_samples_per_second": 54.068,
+      "eval_steps_per_second": 3.403,
+      "step": 2669
+    },
+    {
+      "epoch": 181.97,
+      "grad_norm": 1.1550542116165161,
+      "learning_rate": 7.780952380952381e-06,
+      "loss": 0.3067,
+      "step": 2684
+    },
+    {
+      "epoch": 181.97,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 0.8015040755271912,
+      "eval_runtime": 5.5658,
+      "eval_samples_per_second": 51.385,
+      "eval_steps_per_second": 3.234,
+      "step": 2684
+    },
+    {
+      "epoch": 182.98,
+      "grad_norm": 1.497018814086914,
+      "learning_rate": 7.63809523809524e-06,
+      "loss": 0.2412,
+      "step": 2699
+    },
+    {
+      "epoch": 182.98,
+      "eval_accuracy": 0.7272727272727273,
+      "eval_loss": 0.7975053191184998,
+      "eval_runtime": 4.7894,
+      "eval_samples_per_second": 59.715,
+      "eval_steps_per_second": 3.758,
+      "step": 2699
+    },
+    {
+      "epoch": 184.0,
+      "grad_norm": 1.293284296989441,
+      "learning_rate": 7.4952380952380955e-06,
+      "loss": 0.2686,
+      "step": 2714
+    },
+    {
+      "epoch": 184.0,
+      "eval_accuracy": 0.7237762237762237,
+      "eval_loss": 0.8036520481109619,
+      "eval_runtime": 5.5902,
+      "eval_samples_per_second": 51.161,
+      "eval_steps_per_second": 3.22,
+      "step": 2714
+    },
+    {
+      "epoch": 184.95,
+      "grad_norm": 1.8103567361831665,
+      "learning_rate": 7.361904761904762e-06,
+      "loss": 0.3176,
+      "step": 2728
+    },
+    {
+      "epoch": 184.95,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 0.8016763925552368,
+      "eval_runtime": 4.3848,
+      "eval_samples_per_second": 65.225,
+      "eval_steps_per_second": 4.105,
+      "step": 2728
+    },
+    {
+      "epoch": 185.97,
+      "grad_norm": 1.6021767854690552,
+      "learning_rate": 7.219047619047619e-06,
+      "loss": 0.269,
+      "step": 2743
+    },
+    {
+      "epoch": 185.97,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 0.8060269355773926,
+      "eval_runtime": 4.679,
+      "eval_samples_per_second": 61.125,
+      "eval_steps_per_second": 3.847,
+      "step": 2743
+    },
+    {
+      "epoch": 186.98,
+      "grad_norm": 1.6407794952392578,
+      "learning_rate": 7.076190476190476e-06,
+      "loss": 0.2668,
+      "step": 2758
+    },
+    {
+      "epoch": 186.98,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 0.8030104041099548,
+      "eval_runtime": 5.0657,
+      "eval_samples_per_second": 56.458,
+      "eval_steps_per_second": 3.553,
+      "step": 2758
+    },
+    {
+      "epoch": 188.0,
+      "grad_norm": 1.3405295610427856,
+      "learning_rate": 6.933333333333334e-06,
+      "loss": 0.2761,
+      "step": 2773
+    },
+    {
+      "epoch": 188.0,
+      "eval_accuracy": 0.7412587412587412,
+      "eval_loss": 0.800729513168335,
+      "eval_runtime": 5.2351,
+      "eval_samples_per_second": 54.631,
+      "eval_steps_per_second": 3.438,
+      "step": 2773
+    },
+    {
+      "epoch": 188.95,
+      "grad_norm": 1.5641790628433228,
+      "learning_rate": 6.8e-06,
+      "loss": 0.2731,
+      "step": 2787
+    },
+    {
+      "epoch": 188.95,
+      "eval_accuracy": 0.7342657342657343,
+      "eval_loss": 0.8038657903671265,
+      "eval_runtime": 5.309,
+      "eval_samples_per_second": 53.871,
+      "eval_steps_per_second": 3.39,
+      "step": 2787
+    },
+    {
+      "epoch": 189.97,
+      "grad_norm": 1.6037400960922241,
+      "learning_rate": 6.657142857142857e-06,
+      "loss": 0.2678,
+      "step": 2802
+    },
+    {
+      "epoch": 189.97,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 0.8054195642471313,
+      "eval_runtime": 5.3283,
+      "eval_samples_per_second": 53.675,
+      "eval_steps_per_second": 3.378,
+      "step": 2802
+    },
+    {
+      "epoch": 190.98,
+      "grad_norm": 1.289398193359375,
+      "learning_rate": 6.5142857142857145e-06,
+      "loss": 0.2686,
+      "step": 2817
+    },
+    {
+      "epoch": 190.98,
+      "eval_accuracy": 0.7342657342657343,
+      "eval_loss": 0.8041976094245911,
+      "eval_runtime": 4.8716,
+      "eval_samples_per_second": 58.708,
+      "eval_steps_per_second": 3.695,
+      "step": 2817
+    },
+    {
+      "epoch": 192.0,
+      "grad_norm": 2.173532009124756,
+      "learning_rate": 6.371428571428572e-06,
+      "loss": 0.2721,
+      "step": 2832
+    },
+    {
+      "epoch": 192.0,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 0.806255042552948,
+      "eval_runtime": 4.423,
+      "eval_samples_per_second": 64.662,
+      "eval_steps_per_second": 4.07,
+      "step": 2832
+    },
+    {
+      "epoch": 192.95,
+      "grad_norm": 1.6040128469467163,
+      "learning_rate": 6.238095238095238e-06,
+      "loss": 0.3111,
+      "step": 2846
+    },
+    {
+      "epoch": 192.95,
+      "eval_accuracy": 0.7377622377622378,
+      "eval_loss": 0.8075191378593445,
+      "eval_runtime": 5.3815,
+      "eval_samples_per_second": 53.145,
+      "eval_steps_per_second": 3.345,
+      "step": 2846
+    },
+    {
+      "epoch": 193.97,
+      "grad_norm": 1.7221040725708008,
+      "learning_rate": 6.095238095238095e-06,
+      "loss": 0.2525,
+      "step": 2861
+    },
+    {
+      "epoch": 193.97,
+      "eval_accuracy": 0.7272727272727273,
+      "eval_loss": 0.8114144802093506,
+      "eval_runtime": 5.0641,
+      "eval_samples_per_second": 56.476,
+      "eval_steps_per_second": 3.554,
+      "step": 2861
+    },
+    {
+      "epoch": 194.98,
+      "grad_norm": 1.8386750221252441,
+      "learning_rate": 5.9523809523809525e-06,
+      "loss": 0.2589,
+      "step": 2876
+    },
+    {
+      "epoch": 194.98,
+      "eval_accuracy": 0.7412587412587412,
+      "eval_loss": 0.807061493396759,
+      "eval_runtime": 4.8818,
+      "eval_samples_per_second": 58.585,
+      "eval_steps_per_second": 3.687,
+      "step": 2876
+    },
+    {
+      "epoch": 196.0,
+      "grad_norm": 0.7993654012680054,
+      "learning_rate": 5.80952380952381e-06,
+      "loss": 0.2597,
+      "step": 2891
+    },
+    {
+      "epoch": 196.0,
+      "eval_accuracy": 0.7377622377622378,
+      "eval_loss": 0.8161465525627136,
+      "eval_runtime": 5.7644,
+      "eval_samples_per_second": 49.615,
+      "eval_steps_per_second": 3.123,
+      "step": 2891
+    },
+    {
+      "epoch": 196.95,
+      "grad_norm": 1.822001338005066,
+      "learning_rate": 5.676190476190476e-06,
+      "loss": 0.3034,
+      "step": 2905
+    },
+    {
+      "epoch": 196.95,
+      "eval_accuracy": 0.7377622377622378,
+      "eval_loss": 0.8163018822669983,
+      "eval_runtime": 5.4614,
+      "eval_samples_per_second": 52.367,
+      "eval_steps_per_second": 3.296,
+      "step": 2905
+    },
+    {
+      "epoch": 197.97,
+      "grad_norm": 2.4189586639404297,
+      "learning_rate": 5.5333333333333334e-06,
+      "loss": 0.2721,
+      "step": 2920
+    },
+    {
+      "epoch": 197.97,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 0.8150739073753357,
+      "eval_runtime": 4.8549,
+      "eval_samples_per_second": 58.91,
+      "eval_steps_per_second": 3.708,
+      "step": 2920
+    },
+    {
+      "epoch": 198.98,
+      "grad_norm": 1.6175144910812378,
+      "learning_rate": 5.390476190476191e-06,
+      "loss": 0.2783,
+      "step": 2935
+    },
+    {
+      "epoch": 198.98,
+      "eval_accuracy": 0.7447552447552448,
+      "eval_loss": 0.8093083500862122,
+      "eval_runtime": 4.4141,
+      "eval_samples_per_second": 64.792,
+      "eval_steps_per_second": 4.078,
+      "step": 2935
+    },
+    {
+      "epoch": 200.0,
+      "grad_norm": 1.7226580381393433,
+      "learning_rate": 5.247619047619048e-06,
+      "loss": 0.259,
+      "step": 2950
+    },
+    {
+      "epoch": 200.0,
+      "eval_accuracy": 0.7412587412587412,
+      "eval_loss": 0.8025929927825928,
+      "eval_runtime": 5.4584,
+      "eval_samples_per_second": 52.397,
+      "eval_steps_per_second": 3.298,
+      "step": 2950
+    },
+    {
+      "epoch": 200.95,
+      "grad_norm": 2.5513391494750977,
+      "learning_rate": 5.114285714285714e-06,
+      "loss": 0.2896,
+      "step": 2964
+    },
+    {
+      "epoch": 200.95,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 0.8070325255393982,
+      "eval_runtime": 4.9068,
+      "eval_samples_per_second": 58.286,
+      "eval_steps_per_second": 3.668,
+      "step": 2964
+    },
+    {
+      "epoch": 201.97,
+      "grad_norm": 1.116995930671692,
+      "learning_rate": 4.9714285714285715e-06,
+      "loss": 0.2584,
+      "step": 2979
+    },
+    {
+      "epoch": 201.97,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 0.8112642765045166,
+      "eval_runtime": 5.2035,
+      "eval_samples_per_second": 54.963,
+      "eval_steps_per_second": 3.459,
+      "step": 2979
+    },
+    {
+      "epoch": 202.98,
+      "grad_norm": 1.8753693103790283,
+      "learning_rate": 4.828571428571429e-06,
+      "loss": 0.2672,
+      "step": 2994
+    },
+    {
+      "epoch": 202.98,
+      "eval_accuracy": 0.7342657342657343,
+      "eval_loss": 0.8096128106117249,
+      "eval_runtime": 4.9714,
+      "eval_samples_per_second": 57.529,
+      "eval_steps_per_second": 3.621,
+      "step": 2994
+    },
+    {
+      "epoch": 204.0,
+      "grad_norm": 1.2377465963363647,
+      "learning_rate": 4.685714285714286e-06,
+      "loss": 0.2622,
+      "step": 3009
+    },
+    {
+      "epoch": 204.0,
+      "eval_accuracy": 0.7412587412587412,
+      "eval_loss": 0.8059563040733337,
+      "eval_runtime": 5.4896,
+      "eval_samples_per_second": 52.098,
+      "eval_steps_per_second": 3.279,
+      "step": 3009
+    },
+    {
+      "epoch": 204.95,
+      "grad_norm": 1.5672334432601929,
+      "learning_rate": 4.552380952380952e-06,
+      "loss": 0.2677,
+      "step": 3023
+    },
+    {
+      "epoch": 204.95,
+      "eval_accuracy": 0.7342657342657343,
+      "eval_loss": 0.8084142804145813,
+      "eval_runtime": 5.7936,
+      "eval_samples_per_second": 49.365,
+      "eval_steps_per_second": 3.107,
+      "step": 3023
+    },
+    {
+      "epoch": 205.97,
+      "grad_norm": 0.9782791137695312,
+      "learning_rate": 4.4095238095238096e-06,
+      "loss": 0.263,
+      "step": 3038
+    },
+    {
+      "epoch": 205.97,
+      "eval_accuracy": 0.7377622377622378,
+      "eval_loss": 0.8009874820709229,
+      "eval_runtime": 4.9002,
+      "eval_samples_per_second": 58.364,
+      "eval_steps_per_second": 3.673,
+      "step": 3038
+    },
+    {
+      "epoch": 206.98,
+      "grad_norm": 2.033815622329712,
+      "learning_rate": 4.266666666666667e-06,
+      "loss": 0.2608,
+      "step": 3053
+    },
+    {
+      "epoch": 206.98,
+      "eval_accuracy": 0.7447552447552448,
+      "eval_loss": 0.7988951802253723,
+      "eval_runtime": 5.2669,
+      "eval_samples_per_second": 54.302,
+      "eval_steps_per_second": 3.418,
+      "step": 3053
+    },
+    {
+      "epoch": 208.0,
+      "grad_norm": 2.0094175338745117,
+      "learning_rate": 4.123809523809524e-06,
+      "loss": 0.2528,
+      "step": 3068
+    },
+    {
+      "epoch": 208.0,
+      "eval_accuracy": 0.7447552447552448,
+      "eval_loss": 0.7953740954399109,
+      "eval_runtime": 6.154,
+      "eval_samples_per_second": 46.474,
+      "eval_steps_per_second": 2.925,
+      "step": 3068
+    },
+    {
+      "epoch": 208.95,
+      "grad_norm": 1.639021635055542,
+      "learning_rate": 3.9904761904761905e-06,
+      "loss": 0.2553,
+      "step": 3082
+    },
+    {
+      "epoch": 208.95,
+      "eval_accuracy": 0.7412587412587412,
+      "eval_loss": 0.7965402603149414,
+      "eval_runtime": 5.6365,
+      "eval_samples_per_second": 50.741,
+      "eval_steps_per_second": 3.193,
+      "step": 3082
+    },
+    {
+      "epoch": 209.97,
+      "grad_norm": 2.0743513107299805,
+      "learning_rate": 3.847619047619048e-06,
+      "loss": 0.2652,
+      "step": 3097
+    },
+    {
+      "epoch": 209.97,
+      "eval_accuracy": 0.7412587412587412,
+      "eval_loss": 0.799461841583252,
+      "eval_runtime": 4.1526,
+      "eval_samples_per_second": 68.873,
+      "eval_steps_per_second": 4.335,
+      "step": 3097
+    },
+    {
+      "epoch": 210.98,
+      "grad_norm": 1.6969355344772339,
+      "learning_rate": 3.704761904761905e-06,
+      "loss": 0.246,
+      "step": 3112
+    },
+    {
+      "epoch": 210.98,
+      "eval_accuracy": 0.7377622377622378,
+      "eval_loss": 0.8026475310325623,
+      "eval_runtime": 4.4732,
+      "eval_samples_per_second": 63.936,
+      "eval_steps_per_second": 4.024,
+      "step": 3112
+    },
+    {
+      "epoch": 212.0,
+      "grad_norm": 2.7497739791870117,
+      "learning_rate": 3.561904761904762e-06,
+      "loss": 0.2665,
+      "step": 3127
+    },
+    {
+      "epoch": 212.0,
+      "eval_accuracy": 0.7377622377622378,
+      "eval_loss": 0.8049183487892151,
+      "eval_runtime": 5.7553,
+      "eval_samples_per_second": 49.693,
+      "eval_steps_per_second": 3.128,
+      "step": 3127
+    },
+    {
+      "epoch": 212.95,
+      "grad_norm": 1.5636447668075562,
+      "learning_rate": 3.4285714285714285e-06,
+      "loss": 0.2731,
+      "step": 3141
+    },
+    {
+      "epoch": 212.95,
+      "eval_accuracy": 0.7377622377622378,
+      "eval_loss": 0.8051545023918152,
+      "eval_runtime": 5.9082,
+      "eval_samples_per_second": 48.407,
+      "eval_steps_per_second": 3.047,
+      "step": 3141
+    },
+    {
+      "epoch": 213.97,
+      "grad_norm": 0.8648618459701538,
+      "learning_rate": 3.2857142857142857e-06,
+      "loss": 0.2539,
+      "step": 3156
+    },
+    {
+      "epoch": 213.97,
+      "eval_accuracy": 0.7377622377622378,
+      "eval_loss": 0.8088578581809998,
+      "eval_runtime": 5.3417,
+      "eval_samples_per_second": 53.541,
+      "eval_steps_per_second": 3.37,
+      "step": 3156
+    },
+    {
+      "epoch": 214.98,
+      "grad_norm": 1.2608678340911865,
+      "learning_rate": 3.142857142857143e-06,
+      "loss": 0.2376,
+      "step": 3171
+    },
+    {
+      "epoch": 214.98,
+      "eval_accuracy": 0.7377622377622378,
+      "eval_loss": 0.8115506768226624,
+      "eval_runtime": 5.2232,
+      "eval_samples_per_second": 54.756,
+      "eval_steps_per_second": 3.446,
+      "step": 3171
+    },
+    {
+      "epoch": 216.0,
+      "grad_norm": 1.721834421157837,
+      "learning_rate": 3e-06,
+      "loss": 0.2667,
+      "step": 3186
+    },
+    {
+      "epoch": 216.0,
+      "eval_accuracy": 0.7412587412587412,
+      "eval_loss": 0.80994713306427,
+      "eval_runtime": 6.1297,
+      "eval_samples_per_second": 46.658,
+      "eval_steps_per_second": 2.937,
+      "step": 3186
+    },
+    {
+      "epoch": 216.95,
+      "grad_norm": 2.3429391384124756,
+      "learning_rate": 2.866666666666667e-06,
+      "loss": 0.2768,
+      "step": 3200
+    },
+    {
+      "epoch": 216.95,
+      "eval_accuracy": 0.7412587412587412,
+      "eval_loss": 0.807944655418396,
+      "eval_runtime": 4.2958,
+      "eval_samples_per_second": 66.576,
+      "eval_steps_per_second": 4.19,
+      "step": 3200
+    },
+    {
+      "epoch": 217.97,
+      "grad_norm": 2.041189193725586,
+      "learning_rate": 2.7238095238095238e-06,
+      "loss": 0.2551,
+      "step": 3215
+    },
+    {
+      "epoch": 217.97,
+      "eval_accuracy": 0.7447552447552448,
+      "eval_loss": 0.8064839243888855,
+      "eval_runtime": 5.6836,
+      "eval_samples_per_second": 50.32,
+      "eval_steps_per_second": 3.167,
+      "step": 3215
+    },
+    {
+      "epoch": 218.98,
+      "grad_norm": 1.3890482187271118,
+      "learning_rate": 2.580952380952381e-06,
+      "loss": 0.256,
+      "step": 3230
+    },
+    {
+      "epoch": 218.98,
+      "eval_accuracy": 0.7412587412587412,
+      "eval_loss": 0.8079765439033508,
+      "eval_runtime": 5.1325,
+      "eval_samples_per_second": 55.724,
+      "eval_steps_per_second": 3.507,
+      "step": 3230
+    },
+    {
+      "epoch": 220.0,
+      "grad_norm": 1.2208611965179443,
+      "learning_rate": 2.438095238095238e-06,
+      "loss": 0.2387,
+      "step": 3245
+    },
+    {
+      "epoch": 220.0,
+      "eval_accuracy": 0.7447552447552448,
+      "eval_loss": 0.8076029419898987,
+      "eval_runtime": 5.8067,
+      "eval_samples_per_second": 49.254,
+      "eval_steps_per_second": 3.1,
+      "step": 3245
+    },
+    {
+      "epoch": 220.95,
+      "grad_norm": 1.1974842548370361,
+      "learning_rate": 2.3047619047619047e-06,
+      "loss": 0.2736,
+      "step": 3259
+    },
+    {
+      "epoch": 220.95,
+      "eval_accuracy": 0.7447552447552448,
+      "eval_loss": 0.8080862760543823,
+      "eval_runtime": 4.7889,
+      "eval_samples_per_second": 59.722,
+      "eval_steps_per_second": 3.759,
+      "step": 3259
+    },
+    {
+      "epoch": 221.97,
+      "grad_norm": 1.3414784669876099,
+      "learning_rate": 2.161904761904762e-06,
+      "loss": 0.2551,
+      "step": 3274
+    },
+    {
+      "epoch": 221.97,
+      "eval_accuracy": 0.7447552447552448,
+      "eval_loss": 0.809248149394989,
+      "eval_runtime": 4.967,
+      "eval_samples_per_second": 57.58,
+      "eval_steps_per_second": 3.624,
+      "step": 3274
+    },
+    {
+      "epoch": 222.98,
+      "grad_norm": 1.6082030534744263,
+      "learning_rate": 2.0190476190476195e-06,
+      "loss": 0.2482,
+      "step": 3289
+    },
+    {
+      "epoch": 222.98,
+      "eval_accuracy": 0.7447552447552448,
+      "eval_loss": 0.8084933757781982,
+      "eval_runtime": 4.7114,
+      "eval_samples_per_second": 60.704,
+      "eval_steps_per_second": 3.821,
+      "step": 3289
+    },
+    {
+      "epoch": 224.0,
+      "grad_norm": 1.8340036869049072,
+      "learning_rate": 1.876190476190476e-06,
+      "loss": 0.2432,
+      "step": 3304
+    },
+    {
+      "epoch": 224.0,
+      "eval_accuracy": 0.7447552447552448,
+      "eval_loss": 0.8085303902626038,
+      "eval_runtime": 5.6786,
+      "eval_samples_per_second": 50.365,
+      "eval_steps_per_second": 3.17,
+      "step": 3304
+    },
+    {
+      "epoch": 224.95,
+      "grad_norm": 1.9474186897277832,
+      "learning_rate": 1.742857142857143e-06,
+      "loss": 0.2857,
+      "step": 3318
+    },
+    {
+      "epoch": 224.95,
+      "eval_accuracy": 0.7447552447552448,
+      "eval_loss": 0.8078930974006653,
+      "eval_runtime": 7.5562,
+      "eval_samples_per_second": 37.85,
+      "eval_steps_per_second": 2.382,
+      "step": 3318
+    },
+    {
+      "epoch": 225.97,
+      "grad_norm": 1.7995171546936035,
+      "learning_rate": 1.6000000000000001e-06,
+      "loss": 0.2434,
+      "step": 3333
+    },
+    {
+      "epoch": 225.97,
+      "eval_accuracy": 0.7447552447552448,
+      "eval_loss": 0.8078944087028503,
+      "eval_runtime": 5.9694,
+      "eval_samples_per_second": 47.911,
+      "eval_steps_per_second": 3.015,
+      "step": 3333
+    },
+    {
+      "epoch": 226.98,
+      "grad_norm": 1.4122081995010376,
+      "learning_rate": 1.4571428571428571e-06,
+      "loss": 0.2531,
+      "step": 3348
+    },
+    {
+      "epoch": 226.98,
+      "eval_accuracy": 0.7447552447552448,
+      "eval_loss": 0.8078155517578125,
+      "eval_runtime": 4.7783,
+      "eval_samples_per_second": 59.853,
+      "eval_steps_per_second": 3.767,
+      "step": 3348
+    },
+    {
+      "epoch": 228.0,
+      "grad_norm": 1.6360557079315186,
+      "learning_rate": 1.3142857142857145e-06,
+      "loss": 0.222,
+      "step": 3363
+    },
+    {
+      "epoch": 228.0,
+      "eval_accuracy": 0.7447552447552448,
+      "eval_loss": 0.8092585206031799,
+      "eval_runtime": 5.1759,
+      "eval_samples_per_second": 55.256,
+      "eval_steps_per_second": 3.478,
+      "step": 3363
+    },
+    {
+      "epoch": 228.95,
+      "grad_norm": 1.262143611907959,
+      "learning_rate": 1.180952380952381e-06,
+      "loss": 0.2651,
+      "step": 3377
+    },
+    {
+      "epoch": 228.95,
+      "eval_accuracy": 0.7447552447552448,
+      "eval_loss": 0.8105089664459229,
+      "eval_runtime": 6.304,
+      "eval_samples_per_second": 45.368,
+      "eval_steps_per_second": 2.855,
+      "step": 3377
+    },
+    {
+      "epoch": 229.97,
+      "grad_norm": 1.6864830255508423,
+      "learning_rate": 1.038095238095238e-06,
+      "loss": 0.2886,
+      "step": 3392
+    },
+    {
+      "epoch": 229.97,
+      "eval_accuracy": 0.7447552447552448,
+      "eval_loss": 0.8115819096565247,
+      "eval_runtime": 4.9539,
+      "eval_samples_per_second": 57.732,
+      "eval_steps_per_second": 3.634,
+      "step": 3392
+    },
+    {
+      "epoch": 230.98,
+      "grad_norm": 2.204496383666992,
+      "learning_rate": 8.952380952380953e-07,
+      "loss": 0.2534,
+      "step": 3407
+    },
+    {
+      "epoch": 230.98,
+      "eval_accuracy": 0.7447552447552448,
+      "eval_loss": 0.8115825653076172,
+      "eval_runtime": 5.2788,
+      "eval_samples_per_second": 54.179,
+      "eval_steps_per_second": 3.41,
+      "step": 3407
+    },
+    {
+      "epoch": 232.0,
+      "grad_norm": 1.418483853340149,
+      "learning_rate": 7.523809523809525e-07,
+      "loss": 0.2483,
+      "step": 3422
+    },
+    {
+      "epoch": 232.0,
+      "eval_accuracy": 0.7447552447552448,
+      "eval_loss": 0.8116432428359985,
+      "eval_runtime": 6.7125,
+      "eval_samples_per_second": 42.607,
+      "eval_steps_per_second": 2.682,
+      "step": 3422
+    },
+    {
+      "epoch": 232.95,
+      "grad_norm": 1.5767290592193604,
+      "learning_rate": 6.190476190476191e-07,
+      "loss": 0.272,
+      "step": 3436
+    },
+    {
+      "epoch": 232.95,
+      "eval_accuracy": 0.7447552447552448,
+      "eval_loss": 0.8118025660514832,
+      "eval_runtime": 6.2184,
+      "eval_samples_per_second": 45.993,
+      "eval_steps_per_second": 2.895,
+      "step": 3436
+    },
+    {
+      "epoch": 233.97,
+      "grad_norm": 1.1176037788391113,
+      "learning_rate": 4.761904761904762e-07,
+      "loss": 0.2562,
+      "step": 3451
+    },
+    {
+      "epoch": 233.97,
+      "eval_accuracy": 0.7447552447552448,
+      "eval_loss": 0.8119560480117798,
+      "eval_runtime": 5.396,
+      "eval_samples_per_second": 53.002,
+      "eval_steps_per_second": 3.336,
+      "step": 3451
+    },
+    {
+      "epoch": 234.98,
+      "grad_norm": 2.2027063369750977,
+      "learning_rate": 3.3333333333333335e-07,
+      "loss": 0.2531,
+      "step": 3466
+    },
+    {
+      "epoch": 234.98,
+      "eval_accuracy": 0.7447552447552448,
+      "eval_loss": 0.8122441172599792,
+      "eval_runtime": 4.8695,
+      "eval_samples_per_second": 58.733,
+      "eval_steps_per_second": 3.697,
+      "step": 3466
+    },
+    {
+      "epoch": 236.0,
+      "grad_norm": 1.984872579574585,
+      "learning_rate": 1.9047619047619048e-07,
+      "loss": 0.2397,
+      "step": 3481
+    },
+    {
+      "epoch": 236.0,
+      "eval_accuracy": 0.7447552447552448,
+      "eval_loss": 0.8123078346252441,
+      "eval_runtime": 5.7128,
+      "eval_samples_per_second": 50.063,
+      "eval_steps_per_second": 3.151,
+      "step": 3481
+    },
+    {
+      "epoch": 236.95,
+      "grad_norm": 1.5013829469680786,
+      "learning_rate": 5.7142857142857144e-08,
+      "loss": 0.2573,
+      "step": 3495
+    },
+    {
+      "epoch": 236.95,
+      "eval_accuracy": 0.7447552447552448,
+      "eval_loss": 0.8123247027397156,
+      "eval_runtime": 5.9714,
+      "eval_samples_per_second": 47.895,
+      "eval_steps_per_second": 3.014,
+      "step": 3495
+    },
+    {
+      "epoch": 237.29,
+      "grad_norm": 1.3738845586776733,
+      "learning_rate": 9.523809523809524e-09,
+      "loss": 0.2023,
+      "step": 3500
+    },
+    {
+      "epoch": 237.29,
+      "eval_accuracy": 0.7447552447552448,
+      "eval_loss": 0.8122022151947021,
+      "eval_runtime": 5.4038,
+      "eval_samples_per_second": 52.926,
+      "eval_steps_per_second": 3.331,
+      "step": 3500
+    },
+    {
+      "epoch": 237.29,
+      "step": 3500,
+      "total_flos": 1.939107912046368e+18,
+      "train_loss": 0.5242921064240592,
+      "train_runtime": 3651.9597,
+      "train_samples_per_second": 63.87,
+      "train_steps_per_second": 0.958
+    }
+  ],
+  "logging_steps": 100.0,
+  "max_steps": 3500,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 250,
+  "save_steps": 500,
+  "total_flos": 1.939107912046368e+18,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}