diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,5062 @@
+{
+  "best_metric": 0.4553995728492737,
+  "best_model_checkpoint": "./beans_outputs/checkpoint-69",
+  "epoch": 200.0,
+  "eval_steps": 500,
+  "global_step": 4600,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.43478260869565216,
+      "grad_norm": 9.284521102905273,
+      "learning_rate": 1.9956521739130435e-05,
+      "loss": 0.6169,
+      "step": 10
+    },
+    {
+      "epoch": 0.8695652173913043,
+      "grad_norm": 6.047142028808594,
+      "learning_rate": 1.9913043478260872e-05,
+      "loss": 0.5242,
+      "step": 20
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.7559055118110236,
+      "eval_loss": 0.4960778057575226,
+      "eval_runtime": 5.6891,
+      "eval_samples_per_second": 44.647,
+      "eval_steps_per_second": 0.703,
+      "step": 23
+    },
+    {
+      "epoch": 1.3043478260869565,
+      "grad_norm": 27.046829223632812,
+      "learning_rate": 1.9869565217391305e-05,
+      "loss": 0.5153,
+      "step": 30
+    },
+    {
+      "epoch": 1.7391304347826086,
+      "grad_norm": 9.23794937133789,
+      "learning_rate": 1.9826086956521742e-05,
+      "loss": 0.459,
+      "step": 40
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.7637795275590551,
+      "eval_loss": 0.5001373887062073,
+      "eval_runtime": 5.3498,
+      "eval_samples_per_second": 47.478,
+      "eval_steps_per_second": 0.748,
+      "step": 46
+    },
+    {
+      "epoch": 2.1739130434782608,
+      "grad_norm": 13.100432395935059,
+      "learning_rate": 1.9782608695652176e-05,
+      "loss": 0.4273,
+      "step": 50
+    },
+    {
+      "epoch": 2.608695652173913,
+      "grad_norm": 8.970976829528809,
+      "learning_rate": 1.973913043478261e-05,
+      "loss": 0.4429,
+      "step": 60
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.7874015748031497,
+      "eval_loss": 0.4553995728492737,
+      "eval_runtime": 5.921,
+      "eval_samples_per_second": 42.899,
+      "eval_steps_per_second": 0.676,
+      "step": 69
+    },
+    {
+      "epoch": 3.0434782608695654,
+      "grad_norm": 18.094818115234375,
+      "learning_rate": 1.9695652173913046e-05,
+      "loss": 0.4676,
+      "step": 70
+    },
+    {
+      "epoch": 3.4782608695652173,
+      "grad_norm": 23.408973693847656,
+      "learning_rate": 1.965217391304348e-05,
+      "loss": 0.4298,
+      "step": 80
+    },
+    {
+      "epoch": 3.9130434782608696,
+      "grad_norm": 8.188302993774414,
+      "learning_rate": 1.9608695652173916e-05,
+      "loss": 0.4308,
+      "step": 90
+    },
+    {
+      "epoch": 4.0,
+      "eval_accuracy": 0.7637795275590551,
+      "eval_loss": 0.49241384863853455,
+      "eval_runtime": 5.0789,
+      "eval_samples_per_second": 50.011,
+      "eval_steps_per_second": 0.788,
+      "step": 92
+    },
+    {
+      "epoch": 4.3478260869565215,
+      "grad_norm": 24.4296932220459,
+      "learning_rate": 1.956521739130435e-05,
+      "loss": 0.4158,
+      "step": 100
+    },
+    {
+      "epoch": 4.782608695652174,
+      "grad_norm": 6.616312503814697,
+      "learning_rate": 1.9521739130434786e-05,
+      "loss": 0.4319,
+      "step": 110
+    },
+    {
+      "epoch": 5.0,
+      "eval_accuracy": 0.7874015748031497,
+      "eval_loss": 0.46732431650161743,
+      "eval_runtime": 5.2746,
+      "eval_samples_per_second": 48.155,
+      "eval_steps_per_second": 0.758,
+      "step": 115
+    },
+    {
+      "epoch": 5.217391304347826,
+      "grad_norm": 11.640789031982422,
+      "learning_rate": 1.947826086956522e-05,
+      "loss": 0.4422,
+      "step": 120
+    },
+    {
+      "epoch": 5.6521739130434785,
+      "grad_norm": 14.8034029006958,
+      "learning_rate": 1.9434782608695653e-05,
+      "loss": 0.4047,
+      "step": 130
+    },
+    {
+      "epoch": 6.0,
+      "eval_accuracy": 0.7755905511811023,
+      "eval_loss": 0.4929516315460205,
+      "eval_runtime": 5.4996,
+      "eval_samples_per_second": 46.185,
+      "eval_steps_per_second": 0.727,
+      "step": 138
+    },
+    {
+      "epoch": 6.086956521739131,
+      "grad_norm": 17.933513641357422,
+      "learning_rate": 1.9391304347826087e-05,
+      "loss": 0.419,
+      "step": 140
+    },
+    {
+      "epoch": 6.521739130434782,
+      "grad_norm": 11.223569869995117,
+      "learning_rate": 1.9347826086956523e-05,
+      "loss": 0.4235,
+      "step": 150
+    },
+    {
+      "epoch": 6.956521739130435,
+      "grad_norm": 9.359394073486328,
+      "learning_rate": 1.9304347826086957e-05,
+      "loss": 0.425,
+      "step": 160
+    },
+    {
+      "epoch": 7.0,
+      "eval_accuracy": 0.7795275590551181,
+      "eval_loss": 0.4738819897174835,
+      "eval_runtime": 5.686,
+      "eval_samples_per_second": 44.671,
+      "eval_steps_per_second": 0.703,
+      "step": 161
+    },
+    {
+      "epoch": 7.391304347826087,
+      "grad_norm": 13.405791282653809,
+      "learning_rate": 1.9260869565217394e-05,
+      "loss": 0.4076,
+      "step": 170
+    },
+    {
+      "epoch": 7.826086956521739,
+      "grad_norm": 7.356156826019287,
+      "learning_rate": 1.9217391304347827e-05,
+      "loss": 0.4102,
+      "step": 180
+    },
+    {
+      "epoch": 8.0,
+      "eval_accuracy": 0.7598425196850394,
+      "eval_loss": 0.5118146538734436,
+      "eval_runtime": 5.9498,
+      "eval_samples_per_second": 42.691,
+      "eval_steps_per_second": 0.672,
+      "step": 184
+    },
+    {
+      "epoch": 8.26086956521739,
+      "grad_norm": 9.282608032226562,
+      "learning_rate": 1.9173913043478264e-05,
+      "loss": 0.3981,
+      "step": 190
+    },
+    {
+      "epoch": 8.695652173913043,
+      "grad_norm": 8.834388732910156,
+      "learning_rate": 1.9130434782608697e-05,
+      "loss": 0.3959,
+      "step": 200
+    },
+    {
+      "epoch": 9.0,
+      "eval_accuracy": 0.7480314960629921,
+      "eval_loss": 0.5490437746047974,
+      "eval_runtime": 6.1005,
+      "eval_samples_per_second": 41.636,
+      "eval_steps_per_second": 0.656,
+      "step": 207
+    },
+    {
+      "epoch": 9.130434782608695,
+      "grad_norm": 6.02927303314209,
+      "learning_rate": 1.9086956521739134e-05,
+      "loss": 0.4189,
+      "step": 210
+    },
+    {
+      "epoch": 9.565217391304348,
+      "grad_norm": 11.13778305053711,
+      "learning_rate": 1.9043478260869568e-05,
+      "loss": 0.3864,
+      "step": 220
+    },
+    {
+      "epoch": 10.0,
+      "grad_norm": 18.004470825195312,
+      "learning_rate": 1.9e-05,
+      "loss": 0.365,
+      "step": 230
+    },
+    {
+      "epoch": 10.0,
+      "eval_accuracy": 0.7637795275590551,
+      "eval_loss": 0.5260568261146545,
+      "eval_runtime": 6.2538,
+      "eval_samples_per_second": 40.615,
+      "eval_steps_per_second": 0.64,
+      "step": 230
+    },
+    {
+      "epoch": 10.434782608695652,
+      "grad_norm": 12.497597694396973,
+      "learning_rate": 1.8956521739130434e-05,
+      "loss": 0.3614,
+      "step": 240
+    },
+    {
+      "epoch": 10.869565217391305,
+      "grad_norm": 17.154403686523438,
+      "learning_rate": 1.891304347826087e-05,
+      "loss": 0.4214,
+      "step": 250
+    },
+    {
+      "epoch": 11.0,
+      "eval_accuracy": 0.7795275590551181,
+      "eval_loss": 0.5089420676231384,
+      "eval_runtime": 5.9748,
+      "eval_samples_per_second": 42.512,
+      "eval_steps_per_second": 0.669,
+      "step": 253
+    },
+    {
+      "epoch": 11.304347826086957,
+      "grad_norm": 11.148750305175781,
+      "learning_rate": 1.8869565217391305e-05,
+      "loss": 0.3476,
+      "step": 260
+    },
+    {
+      "epoch": 11.73913043478261,
+      "grad_norm": 18.083158493041992,
+      "learning_rate": 1.882608695652174e-05,
+      "loss": 0.3798,
+      "step": 270
+    },
+    {
+      "epoch": 12.0,
+      "eval_accuracy": 0.7992125984251969,
+      "eval_loss": 0.47105538845062256,
+      "eval_runtime": 6.2283,
+      "eval_samples_per_second": 40.782,
+      "eval_steps_per_second": 0.642,
+      "step": 276
+    },
+    {
+      "epoch": 12.173913043478262,
+      "grad_norm": 19.7681941986084,
+      "learning_rate": 1.8782608695652175e-05,
+      "loss": 0.3819,
+      "step": 280
+    },
+    {
+      "epoch": 12.608695652173914,
+      "grad_norm": 16.852922439575195,
+      "learning_rate": 1.8739130434782612e-05,
+      "loss": 0.3906,
+      "step": 290
+    },
+    {
+      "epoch": 13.0,
+      "eval_accuracy": 0.7913385826771654,
+      "eval_loss": 0.5035075545310974,
+      "eval_runtime": 6.1171,
+      "eval_samples_per_second": 41.523,
+      "eval_steps_per_second": 0.654,
+      "step": 299
+    },
+    {
+      "epoch": 13.043478260869565,
+      "grad_norm": 11.871081352233887,
+      "learning_rate": 1.8695652173913045e-05,
+      "loss": 0.3421,
+      "step": 300
+    },
+    {
+      "epoch": 13.478260869565217,
+      "grad_norm": 12.804304122924805,
+      "learning_rate": 1.865217391304348e-05,
+      "loss": 0.3452,
+      "step": 310
+    },
+    {
+      "epoch": 13.91304347826087,
+      "grad_norm": 13.583141326904297,
+      "learning_rate": 1.8608695652173912e-05,
+      "loss": 0.3706,
+      "step": 320
+    },
+    {
+      "epoch": 14.0,
+      "eval_accuracy": 0.7952755905511811,
+      "eval_loss": 0.49332255125045776,
+      "eval_runtime": 6.0462,
+      "eval_samples_per_second": 42.01,
+      "eval_steps_per_second": 0.662,
+      "step": 322
+    },
+    {
+      "epoch": 14.347826086956522,
+      "grad_norm": 10.307147979736328,
+      "learning_rate": 1.856521739130435e-05,
+      "loss": 0.3242,
+      "step": 330
+    },
+    {
+      "epoch": 14.782608695652174,
+      "grad_norm": 19.720266342163086,
+      "learning_rate": 1.8521739130434782e-05,
+      "loss": 0.3766,
+      "step": 340
+    },
+    {
+      "epoch": 15.0,
+      "eval_accuracy": 0.7992125984251969,
+      "eval_loss": 0.4972522258758545,
+      "eval_runtime": 5.4928,
+      "eval_samples_per_second": 46.242,
+      "eval_steps_per_second": 0.728,
+      "step": 345
+    },
+    {
+      "epoch": 15.217391304347826,
+      "grad_norm": 9.444562911987305,
+      "learning_rate": 1.847826086956522e-05,
+      "loss": 0.3569,
+      "step": 350
+    },
+    {
+      "epoch": 15.652173913043478,
+      "grad_norm": 14.776025772094727,
+      "learning_rate": 1.8434782608695653e-05,
+      "loss": 0.3213,
+      "step": 360
+    },
+    {
+      "epoch": 16.0,
+      "eval_accuracy": 0.7874015748031497,
+      "eval_loss": 0.5220504403114319,
+      "eval_runtime": 5.825,
+      "eval_samples_per_second": 43.605,
+      "eval_steps_per_second": 0.687,
+      "step": 368
+    },
+    {
+      "epoch": 16.08695652173913,
+      "grad_norm": 35.97469711303711,
+      "learning_rate": 1.839130434782609e-05,
+      "loss": 0.3719,
+      "step": 370
+    },
+    {
+      "epoch": 16.52173913043478,
+      "grad_norm": 15.621981620788574,
+      "learning_rate": 1.8347826086956523e-05,
+      "loss": 0.3303,
+      "step": 380
+    },
+    {
+      "epoch": 16.956521739130434,
+      "grad_norm": 12.130711555480957,
+      "learning_rate": 1.830434782608696e-05,
+      "loss": 0.329,
+      "step": 390
+    },
+    {
+      "epoch": 17.0,
+      "eval_accuracy": 0.7834645669291339,
+      "eval_loss": 0.5400053262710571,
+      "eval_runtime": 6.058,
+      "eval_samples_per_second": 41.928,
+      "eval_steps_per_second": 0.66,
+      "step": 391
+    },
+    {
+      "epoch": 17.391304347826086,
+      "grad_norm": 28.593765258789062,
+      "learning_rate": 1.8260869565217393e-05,
+      "loss": 0.3275,
+      "step": 400
+    },
+    {
+      "epoch": 17.82608695652174,
+      "grad_norm": 10.756831169128418,
+      "learning_rate": 1.8217391304347827e-05,
+      "loss": 0.3427,
+      "step": 410
+    },
+    {
+      "epoch": 18.0,
+      "eval_accuracy": 0.7913385826771654,
+      "eval_loss": 0.5252194404602051,
+      "eval_runtime": 5.781,
+      "eval_samples_per_second": 43.937,
+      "eval_steps_per_second": 0.692,
+      "step": 414
+    },
+    {
+      "epoch": 18.26086956521739,
+      "grad_norm": 9.596463203430176,
+      "learning_rate": 1.8173913043478263e-05,
+      "loss": 0.3363,
+      "step": 420
+    },
+    {
+      "epoch": 18.695652173913043,
+      "grad_norm": 12.830471992492676,
+      "learning_rate": 1.8130434782608697e-05,
+      "loss": 0.3472,
+      "step": 430
+    },
+    {
+      "epoch": 19.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 0.6207993626594543,
+      "eval_runtime": 5.8687,
+      "eval_samples_per_second": 43.281,
+      "eval_steps_per_second": 0.682,
+      "step": 437
+    },
+    {
+      "epoch": 19.130434782608695,
+      "grad_norm": 11.963924407958984,
+      "learning_rate": 1.808695652173913e-05,
+      "loss": 0.3509,
+      "step": 440
+    },
+    {
+      "epoch": 19.565217391304348,
+      "grad_norm": 9.393329620361328,
+      "learning_rate": 1.8043478260869567e-05,
+      "loss": 0.3149,
+      "step": 450
+    },
+    {
+      "epoch": 20.0,
+      "grad_norm": 13.855999946594238,
+      "learning_rate": 1.8e-05,
+      "loss": 0.3424,
+      "step": 460
+    },
+    {
+      "epoch": 20.0,
+      "eval_accuracy": 0.7795275590551181,
+      "eval_loss": 0.5319867730140686,
+      "eval_runtime": 5.8322,
+      "eval_samples_per_second": 43.551,
+      "eval_steps_per_second": 0.686,
+      "step": 460
+    },
+    {
+      "epoch": 20.434782608695652,
+      "grad_norm": 12.548827171325684,
+      "learning_rate": 1.7956521739130437e-05,
+      "loss": 0.322,
+      "step": 470
+    },
+    {
+      "epoch": 20.869565217391305,
+      "grad_norm": 12.916594505310059,
+      "learning_rate": 1.791304347826087e-05,
+      "loss": 0.3016,
+      "step": 480
+    },
+    {
+      "epoch": 21.0,
+      "eval_accuracy": 0.7795275590551181,
+      "eval_loss": 0.5488443970680237,
+      "eval_runtime": 6.0683,
+      "eval_samples_per_second": 41.857,
+      "eval_steps_per_second": 0.659,
+      "step": 483
+    },
+    {
+      "epoch": 21.304347826086957,
+      "grad_norm": 10.366270065307617,
+      "learning_rate": 1.7869565217391304e-05,
+      "loss": 0.2733,
+      "step": 490
+    },
+    {
+      "epoch": 21.73913043478261,
+      "grad_norm": 11.919926643371582,
+      "learning_rate": 1.782608695652174e-05,
+      "loss": 0.3033,
+      "step": 500
+    },
+    {
+      "epoch": 22.0,
+      "eval_accuracy": 0.7480314960629921,
+      "eval_loss": 0.5889012813568115,
+      "eval_runtime": 6.0978,
+      "eval_samples_per_second": 41.654,
+      "eval_steps_per_second": 0.656,
+      "step": 506
+    },
+    {
+      "epoch": 22.17391304347826,
+      "grad_norm": 13.522591590881348,
+      "learning_rate": 1.7782608695652174e-05,
+      "loss": 0.26,
+      "step": 510
+    },
+    {
+      "epoch": 22.608695652173914,
+      "grad_norm": 15.789691925048828,
+      "learning_rate": 1.773913043478261e-05,
+      "loss": 0.3083,
+      "step": 520
+    },
+    {
+      "epoch": 23.0,
+      "eval_accuracy": 0.7637795275590551,
+      "eval_loss": 0.6107717156410217,
+      "eval_runtime": 5.8735,
+      "eval_samples_per_second": 43.245,
+      "eval_steps_per_second": 0.681,
+      "step": 529
+    },
+    {
+      "epoch": 23.043478260869566,
+      "grad_norm": 13.747518539428711,
+      "learning_rate": 1.7695652173913045e-05,
+      "loss": 0.2794,
+      "step": 530
+    },
+    {
+      "epoch": 23.47826086956522,
+      "grad_norm": 14.642496109008789,
+      "learning_rate": 1.765217391304348e-05,
+      "loss": 0.2862,
+      "step": 540
+    },
+    {
+      "epoch": 23.91304347826087,
+      "grad_norm": 16.773733139038086,
+      "learning_rate": 1.7608695652173915e-05,
+      "loss": 0.2772,
+      "step": 550
+    },
+    {
+      "epoch": 24.0,
+      "eval_accuracy": 0.7480314960629921,
+      "eval_loss": 0.5844786763191223,
+      "eval_runtime": 6.1158,
+      "eval_samples_per_second": 41.532,
+      "eval_steps_per_second": 0.654,
+      "step": 552
+    },
+    {
+      "epoch": 24.347826086956523,
+      "grad_norm": 11.864821434020996,
+      "learning_rate": 1.756521739130435e-05,
+      "loss": 0.2693,
+      "step": 560
+    },
+    {
+      "epoch": 24.782608695652176,
+      "grad_norm": 19.129663467407227,
+      "learning_rate": 1.7521739130434785e-05,
+      "loss": 0.287,
+      "step": 570
+    },
+    {
+      "epoch": 25.0,
+      "eval_accuracy": 0.8070866141732284,
+      "eval_loss": 0.5241626501083374,
+      "eval_runtime": 6.0464,
+      "eval_samples_per_second": 42.008,
+      "eval_steps_per_second": 0.662,
+      "step": 575
+    },
+    {
+      "epoch": 25.217391304347824,
+      "grad_norm": 9.217809677124023,
+      "learning_rate": 1.747826086956522e-05,
+      "loss": 0.309,
+      "step": 580
+    },
+    {
+      "epoch": 25.652173913043477,
+      "grad_norm": 20.24026870727539,
+      "learning_rate": 1.7434782608695652e-05,
+      "loss": 0.2651,
+      "step": 590
+    },
+    {
+      "epoch": 26.0,
+      "eval_accuracy": 0.7598425196850394,
+      "eval_loss": 0.6275522112846375,
+      "eval_runtime": 5.2184,
+      "eval_samples_per_second": 48.674,
+      "eval_steps_per_second": 0.767,
+      "step": 598
+    },
+    {
+      "epoch": 26.08695652173913,
+      "grad_norm": 14.04963207244873,
+      "learning_rate": 1.739130434782609e-05,
+      "loss": 0.2763,
+      "step": 600
+    },
+    {
+      "epoch": 26.52173913043478,
+      "grad_norm": 18.499040603637695,
+      "learning_rate": 1.7347826086956522e-05,
+      "loss": 0.2944,
+      "step": 610
+    },
+    {
+      "epoch": 26.956521739130434,
+      "grad_norm": 14.989178657531738,
+      "learning_rate": 1.730434782608696e-05,
+      "loss": 0.2696,
+      "step": 620
+    },
+    {
+      "epoch": 27.0,
+      "eval_accuracy": 0.7834645669291339,
+      "eval_loss": 0.5648530125617981,
+      "eval_runtime": 5.345,
+      "eval_samples_per_second": 47.521,
+      "eval_steps_per_second": 0.748,
+      "step": 621
+    },
+    {
+      "epoch": 27.391304347826086,
+      "grad_norm": 24.154279708862305,
+      "learning_rate": 1.7260869565217393e-05,
+      "loss": 0.2638,
+      "step": 630
+    },
+    {
+      "epoch": 27.82608695652174,
+      "grad_norm": 21.489530563354492,
+      "learning_rate": 1.721739130434783e-05,
+      "loss": 0.2701,
+      "step": 640
+    },
+    {
+      "epoch": 28.0,
+      "eval_accuracy": 0.7755905511811023,
+      "eval_loss": 0.6103212237358093,
+      "eval_runtime": 5.2378,
+      "eval_samples_per_second": 48.494,
+      "eval_steps_per_second": 0.764,
+      "step": 644
+    },
+    {
+      "epoch": 28.26086956521739,
+      "grad_norm": 7.044363975524902,
+      "learning_rate": 1.7173913043478263e-05,
+      "loss": 0.2929,
+      "step": 650
+    },
+    {
+      "epoch": 28.695652173913043,
+      "grad_norm": 8.331864356994629,
+      "learning_rate": 1.71304347826087e-05,
+      "loss": 0.2451,
+      "step": 660
+    },
+    {
+      "epoch": 29.0,
+      "eval_accuracy": 0.7637795275590551,
+      "eval_loss": 0.6206949949264526,
+      "eval_runtime": 5.2583,
+      "eval_samples_per_second": 48.305,
+      "eval_steps_per_second": 0.761,
+      "step": 667
+    },
+    {
+      "epoch": 29.130434782608695,
+      "grad_norm": 14.339310646057129,
+      "learning_rate": 1.708695652173913e-05,
+      "loss": 0.238,
+      "step": 670
+    },
+    {
+      "epoch": 29.565217391304348,
+      "grad_norm": 17.397079467773438,
+      "learning_rate": 1.7043478260869566e-05,
+      "loss": 0.2621,
+      "step": 680
+    },
+    {
+      "epoch": 30.0,
+      "grad_norm": 49.991676330566406,
+      "learning_rate": 1.7e-05,
+      "loss": 0.2705,
+      "step": 690
+    },
+    {
+      "epoch": 30.0,
+      "eval_accuracy": 0.7755905511811023,
+      "eval_loss": 0.5990407466888428,
+      "eval_runtime": 5.3203,
+      "eval_samples_per_second": 47.741,
+      "eval_steps_per_second": 0.752,
+      "step": 690
+    },
+    {
+      "epoch": 30.434782608695652,
+      "grad_norm": 10.006631851196289,
+      "learning_rate": 1.6956521739130437e-05,
+      "loss": 0.2313,
+      "step": 700
+    },
+    {
+      "epoch": 30.869565217391305,
+      "grad_norm": 19.18036651611328,
+      "learning_rate": 1.691304347826087e-05,
+      "loss": 0.2553,
+      "step": 710
+    },
+    {
+      "epoch": 31.0,
+      "eval_accuracy": 0.7834645669291339,
+      "eval_loss": 0.5962309241294861,
+      "eval_runtime": 5.2875,
+      "eval_samples_per_second": 48.038,
+      "eval_steps_per_second": 0.757,
+      "step": 713
+    },
+    {
+      "epoch": 31.304347826086957,
+      "grad_norm": 11.844828605651855,
+      "learning_rate": 1.6869565217391307e-05,
+      "loss": 0.2392,
+      "step": 720
+    },
+    {
+      "epoch": 31.73913043478261,
+      "grad_norm": 28.340103149414062,
+      "learning_rate": 1.682608695652174e-05,
+      "loss": 0.2559,
+      "step": 730
+    },
+    {
+      "epoch": 32.0,
+      "eval_accuracy": 0.7716535433070866,
+      "eval_loss": 0.6681433916091919,
+      "eval_runtime": 5.5265,
+      "eval_samples_per_second": 45.961,
+      "eval_steps_per_second": 0.724,
+      "step": 736
+    },
+    {
+      "epoch": 32.17391304347826,
+      "grad_norm": 19.246511459350586,
+      "learning_rate": 1.6782608695652177e-05,
+      "loss": 0.2441,
+      "step": 740
+    },
+    {
+      "epoch": 32.608695652173914,
+      "grad_norm": 14.579903602600098,
+      "learning_rate": 1.673913043478261e-05,
+      "loss": 0.2405,
+      "step": 750
+    },
+    {
+      "epoch": 33.0,
+      "eval_accuracy": 0.7637795275590551,
+      "eval_loss": 0.5917338132858276,
+      "eval_runtime": 5.8004,
+      "eval_samples_per_second": 43.79,
+      "eval_steps_per_second": 0.69,
+      "step": 759
+    },
+    {
+      "epoch": 33.04347826086956,
+      "grad_norm": 18.550579071044922,
+      "learning_rate": 1.6695652173913044e-05,
+      "loss": 0.2983,
+      "step": 760
+    },
+    {
+      "epoch": 33.47826086956522,
+      "grad_norm": 12.869200706481934,
+      "learning_rate": 1.6652173913043477e-05,
+      "loss": 0.2653,
+      "step": 770
+    },
+    {
+      "epoch": 33.91304347826087,
+      "grad_norm": 21.49362564086914,
+      "learning_rate": 1.6608695652173914e-05,
+      "loss": 0.2707,
+      "step": 780
+    },
+    {
+      "epoch": 34.0,
+      "eval_accuracy": 0.7637795275590551,
+      "eval_loss": 0.5905888080596924,
+      "eval_runtime": 6.1194,
+      "eval_samples_per_second": 41.508,
+      "eval_steps_per_second": 0.654,
+      "step": 782
+    },
+    {
+      "epoch": 34.34782608695652,
+      "grad_norm": 20.466447830200195,
+      "learning_rate": 1.6565217391304348e-05,
+      "loss": 0.2314,
+      "step": 790
+    },
+    {
+      "epoch": 34.78260869565217,
+      "grad_norm": 14.096879005432129,
+      "learning_rate": 1.6521739130434785e-05,
+      "loss": 0.3004,
+      "step": 800
+    },
+    {
+      "epoch": 35.0,
+      "eval_accuracy": 0.7874015748031497,
+      "eval_loss": 0.5905265212059021,
+      "eval_runtime": 5.7562,
+      "eval_samples_per_second": 44.126,
+      "eval_steps_per_second": 0.695,
+      "step": 805
+    },
+    {
+      "epoch": 35.21739130434783,
+      "grad_norm": 12.822495460510254,
+      "learning_rate": 1.6478260869565218e-05,
+      "loss": 0.2325,
+      "step": 810
+    },
+    {
+      "epoch": 35.65217391304348,
+      "grad_norm": 15.958830833435059,
+      "learning_rate": 1.6434782608695655e-05,
+      "loss": 0.2404,
+      "step": 820
+    },
+    {
+      "epoch": 36.0,
+      "eval_accuracy": 0.7677165354330708,
+      "eval_loss": 0.5914427042007446,
+      "eval_runtime": 5.657,
+      "eval_samples_per_second": 44.9,
+      "eval_steps_per_second": 0.707,
+      "step": 828
+    },
+    {
+      "epoch": 36.08695652173913,
+      "grad_norm": 14.542938232421875,
+      "learning_rate": 1.6391304347826088e-05,
+      "loss": 0.235,
+      "step": 830
+    },
+    {
+      "epoch": 36.52173913043478,
+      "grad_norm": 36.17627716064453,
+      "learning_rate": 1.6347826086956525e-05,
+      "loss": 0.258,
+      "step": 840
+    },
+    {
+      "epoch": 36.95652173913044,
+      "grad_norm": 16.92486572265625,
+      "learning_rate": 1.630434782608696e-05,
+      "loss": 0.242,
+      "step": 850
+    },
+    {
+      "epoch": 37.0,
+      "eval_accuracy": 0.7637795275590551,
+      "eval_loss": 0.7637115716934204,
+      "eval_runtime": 5.8801,
+      "eval_samples_per_second": 43.196,
+      "eval_steps_per_second": 0.68,
+      "step": 851
+    },
+    {
+      "epoch": 37.391304347826086,
+      "grad_norm": 16.343120574951172,
+      "learning_rate": 1.6260869565217392e-05,
+      "loss": 0.2075,
+      "step": 860
+    },
+    {
+      "epoch": 37.82608695652174,
+      "grad_norm": 26.078245162963867,
+      "learning_rate": 1.621739130434783e-05,
+      "loss": 0.2221,
+      "step": 870
+    },
+    {
+      "epoch": 38.0,
+      "eval_accuracy": 0.7598425196850394,
+      "eval_loss": 0.7117211222648621,
+      "eval_runtime": 5.1769,
+      "eval_samples_per_second": 49.064,
+      "eval_steps_per_second": 0.773,
+      "step": 874
+    },
+    {
+      "epoch": 38.26086956521739,
+      "grad_norm": 9.526216506958008,
+      "learning_rate": 1.6173913043478262e-05,
+      "loss": 0.2285,
+      "step": 880
+    },
+    {
+      "epoch": 38.69565217391305,
+      "grad_norm": 14.653572082519531,
+      "learning_rate": 1.6130434782608696e-05,
+      "loss": 0.2196,
+      "step": 890
+    },
+    {
+      "epoch": 39.0,
+      "eval_accuracy": 0.7834645669291339,
+      "eval_loss": 0.6442176699638367,
+      "eval_runtime": 6.1843,
+      "eval_samples_per_second": 41.072,
+      "eval_steps_per_second": 0.647,
+      "step": 897
+    },
+    {
+      "epoch": 39.130434782608695,
+      "grad_norm": 6.992471694946289,
+      "learning_rate": 1.6086956521739132e-05,
+      "loss": 0.2301,
+      "step": 900
+    },
+    {
+      "epoch": 39.56521739130435,
+      "grad_norm": 15.566644668579102,
+      "learning_rate": 1.6043478260869566e-05,
+      "loss": 0.234,
+      "step": 910
+    },
+    {
+      "epoch": 40.0,
+      "grad_norm": 21.968212127685547,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.23,
+      "step": 920
+    },
+    {
+      "epoch": 40.0,
+      "eval_accuracy": 0.7716535433070866,
+      "eval_loss": 0.7011120915412903,
+      "eval_runtime": 6.1863,
+      "eval_samples_per_second": 41.059,
+      "eval_steps_per_second": 0.647,
+      "step": 920
+    },
+    {
+      "epoch": 40.43478260869565,
+      "grad_norm": 7.890230655670166,
+      "learning_rate": 1.5956521739130436e-05,
+      "loss": 0.2143,
+      "step": 930
+    },
+    {
+      "epoch": 40.869565217391305,
+      "grad_norm": 13.847317695617676,
+      "learning_rate": 1.591304347826087e-05,
+      "loss": 0.2045,
+      "step": 940
+    },
+    {
+      "epoch": 41.0,
+      "eval_accuracy": 0.7598425196850394,
+      "eval_loss": 0.7821894288063049,
+      "eval_runtime": 6.1565,
+      "eval_samples_per_second": 41.257,
+      "eval_steps_per_second": 0.65,
+      "step": 943
+    },
+    {
+      "epoch": 41.30434782608695,
+      "grad_norm": 27.132835388183594,
+      "learning_rate": 1.5869565217391306e-05,
+      "loss": 0.2094,
+      "step": 950
+    },
+    {
+      "epoch": 41.73913043478261,
+      "grad_norm": 33.14794921875,
+      "learning_rate": 1.582608695652174e-05,
+      "loss": 0.2043,
+      "step": 960
+    },
+    {
+      "epoch": 42.0,
+      "eval_accuracy": 0.7519685039370079,
+      "eval_loss": 0.7339490056037903,
+      "eval_runtime": 6.1427,
+      "eval_samples_per_second": 41.35,
+      "eval_steps_per_second": 0.651,
+      "step": 966
+    },
+    {
+      "epoch": 42.17391304347826,
+      "grad_norm": 9.211811065673828,
+      "learning_rate": 1.5782608695652177e-05,
+      "loss": 0.2468,
+      "step": 970
+    },
+    {
+      "epoch": 42.608695652173914,
+      "grad_norm": 18.167051315307617,
+      "learning_rate": 1.573913043478261e-05,
+      "loss": 0.2413,
+      "step": 980
+    },
+    {
+      "epoch": 43.0,
+      "eval_accuracy": 0.7677165354330708,
+      "eval_loss": 0.6917413473129272,
+      "eval_runtime": 6.1303,
+      "eval_samples_per_second": 41.434,
+      "eval_steps_per_second": 0.652,
+      "step": 989
+    },
+    {
+      "epoch": 43.04347826086956,
+      "grad_norm": 16.584321975708008,
+      "learning_rate": 1.5695652173913047e-05,
+      "loss": 0.2255,
+      "step": 990
+    },
+    {
+      "epoch": 43.47826086956522,
+      "grad_norm": 13.140791893005371,
+      "learning_rate": 1.565217391304348e-05,
+      "loss": 0.216,
+      "step": 1000
+    },
+    {
+      "epoch": 43.91304347826087,
+      "grad_norm": 16.6044921875,
+      "learning_rate": 1.5608695652173914e-05,
+      "loss": 0.2135,
+      "step": 1010
+    },
+    {
+      "epoch": 44.0,
+      "eval_accuracy": 0.7716535433070866,
+      "eval_loss": 0.695396363735199,
+      "eval_runtime": 6.1325,
+      "eval_samples_per_second": 41.419,
+      "eval_steps_per_second": 0.652,
+      "step": 1012
+    },
+    {
+      "epoch": 44.34782608695652,
+      "grad_norm": 7.823153495788574,
+      "learning_rate": 1.5565217391304347e-05,
+      "loss": 0.2036,
+      "step": 1020
+    },
+    {
+      "epoch": 44.78260869565217,
+      "grad_norm": 26.390613555908203,
+      "learning_rate": 1.5521739130434784e-05,
+      "loss": 0.2194,
+      "step": 1030
+    },
+    {
+      "epoch": 45.0,
+      "eval_accuracy": 0.7795275590551181,
+      "eval_loss": 0.6728869080543518,
+      "eval_runtime": 6.0671,
+      "eval_samples_per_second": 41.865,
+      "eval_steps_per_second": 0.659,
+      "step": 1035
+    },
+    {
+      "epoch": 45.21739130434783,
+      "grad_norm": 22.04627227783203,
+      "learning_rate": 1.5478260869565217e-05,
+      "loss": 0.2219,
+      "step": 1040
+    },
+    {
+      "epoch": 45.65217391304348,
+      "grad_norm": 9.552602767944336,
+      "learning_rate": 1.5434782608695654e-05,
+      "loss": 0.211,
+      "step": 1050
+    },
+    {
+      "epoch": 46.0,
+      "eval_accuracy": 0.7834645669291339,
+      "eval_loss": 0.6840518712997437,
+      "eval_runtime": 6.1887,
+      "eval_samples_per_second": 41.043,
+      "eval_steps_per_second": 0.646,
+      "step": 1058
+    },
+    {
+      "epoch": 46.08695652173913,
+      "grad_norm": 17.527860641479492,
+      "learning_rate": 1.5391304347826088e-05,
+      "loss": 0.1984,
+      "step": 1060
+    },
+    {
+      "epoch": 46.52173913043478,
+      "grad_norm": 8.503457069396973,
+      "learning_rate": 1.5347826086956524e-05,
+      "loss": 0.1849,
+      "step": 1070
+    },
+    {
+      "epoch": 46.95652173913044,
+      "grad_norm": 27.90496826171875,
+      "learning_rate": 1.5304347826086958e-05,
+      "loss": 0.2155,
+      "step": 1080
+    },
+    {
+      "epoch": 47.0,
+      "eval_accuracy": 0.7677165354330708,
+      "eval_loss": 0.7107857465744019,
+      "eval_runtime": 7.3514,
+      "eval_samples_per_second": 34.551,
+      "eval_steps_per_second": 0.544,
+      "step": 1081
+    },
+    {
+      "epoch": 47.391304347826086,
+      "grad_norm": 16.541719436645508,
+      "learning_rate": 1.5260869565217395e-05,
+      "loss": 0.2054,
+      "step": 1090
+    },
+    {
+      "epoch": 47.82608695652174,
+      "grad_norm": 10.804960250854492,
+      "learning_rate": 1.5217391304347828e-05,
+      "loss": 0.2231,
+      "step": 1100
+    },
+    {
+      "epoch": 48.0,
+      "eval_accuracy": 0.7677165354330708,
+      "eval_loss": 0.6758210062980652,
+      "eval_runtime": 5.136,
+      "eval_samples_per_second": 49.455,
+      "eval_steps_per_second": 0.779,
+      "step": 1104
+    },
+    {
+      "epoch": 48.26086956521739,
+      "grad_norm": 13.976753234863281,
+      "learning_rate": 1.5173913043478262e-05,
+      "loss": 0.2362,
+      "step": 1110
+    },
+    {
+      "epoch": 48.69565217391305,
+      "grad_norm": 37.785911560058594,
+      "learning_rate": 1.5130434782608697e-05,
+      "loss": 0.2364,
+      "step": 1120
+    },
+    {
+      "epoch": 49.0,
+      "eval_accuracy": 0.7519685039370079,
+      "eval_loss": 0.7746602892875671,
+      "eval_runtime": 5.5686,
+      "eval_samples_per_second": 45.613,
+      "eval_steps_per_second": 0.718,
+      "step": 1127
+    },
+    {
+      "epoch": 49.130434782608695,
+      "grad_norm": 17.750612258911133,
+      "learning_rate": 1.5086956521739132e-05,
+      "loss": 0.2044,
+      "step": 1130
+    },
+    {
+      "epoch": 49.56521739130435,
+      "grad_norm": 29.463176727294922,
+      "learning_rate": 1.5043478260869567e-05,
+      "loss": 0.2272,
+      "step": 1140
+    },
+    {
+      "epoch": 50.0,
+      "grad_norm": 21.283435821533203,
+      "learning_rate": 1.5000000000000002e-05,
+      "loss": 0.222,
+      "step": 1150
+    },
+    {
+      "epoch": 50.0,
+      "eval_accuracy": 0.7637795275590551,
+      "eval_loss": 0.7104395627975464,
+      "eval_runtime": 5.8427,
+      "eval_samples_per_second": 43.473,
+      "eval_steps_per_second": 0.685,
+      "step": 1150
+    },
+    {
+      "epoch": 50.43478260869565,
+      "grad_norm": 13.56606674194336,
+      "learning_rate": 1.4956521739130436e-05,
+      "loss": 0.2,
+      "step": 1160
+    },
+    {
+      "epoch": 50.869565217391305,
+      "grad_norm": 19.158952713012695,
+      "learning_rate": 1.491304347826087e-05,
+      "loss": 0.2018,
+      "step": 1170
+    },
+    {
+      "epoch": 51.0,
+      "eval_accuracy": 0.7952755905511811,
+      "eval_loss": 0.6884562969207764,
+      "eval_runtime": 6.9894,
+      "eval_samples_per_second": 36.341,
+      "eval_steps_per_second": 0.572,
+      "step": 1173
+    },
+    {
+      "epoch": 51.30434782608695,
+      "grad_norm": 15.846772193908691,
+      "learning_rate": 1.4869565217391306e-05,
+      "loss": 0.2231,
+      "step": 1180
+    },
+    {
+      "epoch": 51.73913043478261,
+      "grad_norm": 19.628162384033203,
+      "learning_rate": 1.4826086956521741e-05,
+      "loss": 0.219,
+      "step": 1190
+    },
+    {
+      "epoch": 52.0,
+      "eval_accuracy": 0.7519685039370079,
+      "eval_loss": 0.7609480023384094,
+      "eval_runtime": 5.6933,
+      "eval_samples_per_second": 44.614,
+      "eval_steps_per_second": 0.703,
+      "step": 1196
+    },
+    {
+      "epoch": 52.17391304347826,
+      "grad_norm": 13.08530044555664,
+      "learning_rate": 1.4782608695652174e-05,
+      "loss": 0.1924,
+      "step": 1200
+    },
+    {
+      "epoch": 52.608695652173914,
+      "grad_norm": 18.541603088378906,
+      "learning_rate": 1.473913043478261e-05,
+      "loss": 0.1916,
+      "step": 1210
+    },
+    {
+      "epoch": 53.0,
+      "eval_accuracy": 0.7677165354330708,
+      "eval_loss": 0.83939129114151,
+      "eval_runtime": 5.8715,
+      "eval_samples_per_second": 43.26,
+      "eval_steps_per_second": 0.681,
+      "step": 1219
+    },
+    {
+      "epoch": 53.04347826086956,
+      "grad_norm": 25.077634811401367,
+      "learning_rate": 1.4695652173913045e-05,
+      "loss": 0.2072,
+      "step": 1220
+    },
+    {
+      "epoch": 53.47826086956522,
+      "grad_norm": 13.322381973266602,
+      "learning_rate": 1.465217391304348e-05,
+      "loss": 0.1981,
+      "step": 1230
+    },
+    {
+      "epoch": 53.91304347826087,
+      "grad_norm": 37.27448654174805,
+      "learning_rate": 1.4608695652173915e-05,
+      "loss": 0.1767,
+      "step": 1240
+    },
+    {
+      "epoch": 54.0,
+      "eval_accuracy": 0.7716535433070866,
+      "eval_loss": 0.7909868955612183,
+      "eval_runtime": 5.8932,
+      "eval_samples_per_second": 43.101,
+      "eval_steps_per_second": 0.679,
+      "step": 1242
+    },
+    {
+      "epoch": 54.34782608695652,
+      "grad_norm": 7.898425579071045,
+      "learning_rate": 1.456521739130435e-05,
+      "loss": 0.1524,
+      "step": 1250
+    },
+    {
+      "epoch": 54.78260869565217,
+      "grad_norm": 23.582290649414062,
+      "learning_rate": 1.4521739130434785e-05,
+      "loss": 0.236,
+      "step": 1260
+    },
+    {
+      "epoch": 55.0,
+      "eval_accuracy": 0.7755905511811023,
+      "eval_loss": 0.7601104974746704,
+      "eval_runtime": 5.1002,
+      "eval_samples_per_second": 49.802,
+      "eval_steps_per_second": 0.784,
+      "step": 1265
+    },
+    {
+      "epoch": 55.21739130434783,
+      "grad_norm": 18.55949592590332,
+      "learning_rate": 1.447826086956522e-05,
+      "loss": 0.2053,
+      "step": 1270
+    },
+    {
+      "epoch": 55.65217391304348,
+      "grad_norm": 11.066632270812988,
+      "learning_rate": 1.4434782608695654e-05,
+      "loss": 0.1898,
+      "step": 1280
+    },
+    {
+      "epoch": 56.0,
+      "eval_accuracy": 0.7716535433070866,
+      "eval_loss": 0.7501403093338013,
+      "eval_runtime": 6.1846,
+      "eval_samples_per_second": 41.07,
+      "eval_steps_per_second": 0.647,
+      "step": 1288
+    },
+    {
+      "epoch": 56.08695652173913,
+      "grad_norm": 13.702197074890137,
+      "learning_rate": 1.4391304347826087e-05,
+      "loss": 0.1817,
+      "step": 1290
+    },
+    {
+      "epoch": 56.52173913043478,
+      "grad_norm": 12.48442268371582,
+      "learning_rate": 1.4347826086956522e-05,
+      "loss": 0.2184,
+      "step": 1300
+    },
+    {
+      "epoch": 56.95652173913044,
+      "grad_norm": 12.403263092041016,
+      "learning_rate": 1.4304347826086957e-05,
+      "loss": 0.1876,
+      "step": 1310
+    },
+    {
+      "epoch": 57.0,
+      "eval_accuracy": 0.7755905511811023,
+      "eval_loss": 0.7492194175720215,
+      "eval_runtime": 5.4019,
+      "eval_samples_per_second": 47.02,
+      "eval_steps_per_second": 0.74,
+      "step": 1311
+    },
+    {
+      "epoch": 57.391304347826086,
+      "grad_norm": 46.26993179321289,
+      "learning_rate": 1.4260869565217392e-05,
+      "loss": 0.2015,
+      "step": 1320
+    },
+    {
+      "epoch": 57.82608695652174,
+      "grad_norm": 18.550594329833984,
+      "learning_rate": 1.4217391304347828e-05,
+      "loss": 0.1592,
+      "step": 1330
+    },
+    {
+      "epoch": 58.0,
+      "eval_accuracy": 0.7637795275590551,
+      "eval_loss": 0.7905071377754211,
+      "eval_runtime": 5.5787,
+      "eval_samples_per_second": 45.531,
+      "eval_steps_per_second": 0.717,
+      "step": 1334
+    },
+    {
+      "epoch": 58.26086956521739,
+      "grad_norm": 5.595022201538086,
+      "learning_rate": 1.4173913043478263e-05,
+      "loss": 0.1851,
+      "step": 1340
+    },
+    {
+      "epoch": 58.69565217391305,
+      "grad_norm": 12.843214988708496,
+      "learning_rate": 1.4130434782608698e-05,
+      "loss": 0.1772,
+      "step": 1350
+    },
+    {
+      "epoch": 59.0,
+      "eval_accuracy": 0.7716535433070866,
+      "eval_loss": 0.7410995364189148,
+      "eval_runtime": 5.5315,
+      "eval_samples_per_second": 45.919,
+      "eval_steps_per_second": 0.723,
+      "step": 1357
+    },
+    {
+      "epoch": 59.130434782608695,
+      "grad_norm": 15.449782371520996,
+      "learning_rate": 1.4086956521739133e-05,
+      "loss": 0.2286,
+      "step": 1360
+    },
+    {
+      "epoch": 59.56521739130435,
+      "grad_norm": 13.496415138244629,
+      "learning_rate": 1.4043478260869568e-05,
+      "loss": 0.1659,
+      "step": 1370
+    },
+    {
+      "epoch": 60.0,
+      "grad_norm": 18.721599578857422,
+      "learning_rate": 1.4e-05,
+      "loss": 0.1787,
+      "step": 1380
+    },
+    {
+      "epoch": 60.0,
+      "eval_accuracy": 0.7795275590551181,
+      "eval_loss": 0.8145360946655273,
+      "eval_runtime": 5.7016,
+      "eval_samples_per_second": 44.549,
+      "eval_steps_per_second": 0.702,
+      "step": 1380
+    },
+    {
+      "epoch": 60.43478260869565,
+      "grad_norm": 10.89456558227539,
+      "learning_rate": 1.3956521739130435e-05,
+      "loss": 0.1598,
+      "step": 1390
+    },
+    {
+      "epoch": 60.869565217391305,
+      "grad_norm": 13.629638671875,
+      "learning_rate": 1.391304347826087e-05,
+      "loss": 0.1782,
+      "step": 1400
+    },
+    {
+      "epoch": 61.0,
+      "eval_accuracy": 0.7795275590551181,
+      "eval_loss": 0.7721127271652222,
+      "eval_runtime": 5.8504,
+      "eval_samples_per_second": 43.416,
+      "eval_steps_per_second": 0.684,
+      "step": 1403
+    },
+    {
+      "epoch": 61.30434782608695,
+      "grad_norm": 17.057193756103516,
+      "learning_rate": 1.3869565217391305e-05,
+      "loss": 0.1892,
+      "step": 1410
+    },
+    {
+      "epoch": 61.73913043478261,
+      "grad_norm": 18.26725196838379,
+      "learning_rate": 1.382608695652174e-05,
+      "loss": 0.1781,
+      "step": 1420
+    },
+    {
+      "epoch": 62.0,
+      "eval_accuracy": 0.7834645669291339,
+      "eval_loss": 0.8022087216377258,
+      "eval_runtime": 5.9987,
+      "eval_samples_per_second": 42.342,
+      "eval_steps_per_second": 0.667,
+      "step": 1426
+    },
+    {
+      "epoch": 62.17391304347826,
+      "grad_norm": 23.26697540283203,
+      "learning_rate": 1.3782608695652175e-05,
+      "loss": 0.1886,
+      "step": 1430
+    },
+    {
+      "epoch": 62.608695652173914,
+      "grad_norm": 16.03859519958496,
+      "learning_rate": 1.373913043478261e-05,
+      "loss": 0.1884,
+      "step": 1440
+    },
+    {
+      "epoch": 63.0,
+      "eval_accuracy": 0.7755905511811023,
+      "eval_loss": 0.8629626035690308,
+      "eval_runtime": 5.976,
+      "eval_samples_per_second": 42.504,
+      "eval_steps_per_second": 0.669,
+      "step": 1449
+    },
+    {
+      "epoch": 63.04347826086956,
+      "grad_norm": 14.661111831665039,
+      "learning_rate": 1.3695652173913046e-05,
+      "loss": 0.1762,
+      "step": 1450
+    },
+    {
+      "epoch": 63.47826086956522,
+      "grad_norm": 11.116896629333496,
+      "learning_rate": 1.3652173913043479e-05,
+      "loss": 0.162,
+      "step": 1460
+    },
+    {
+      "epoch": 63.91304347826087,
+      "grad_norm": 17.106279373168945,
+      "learning_rate": 1.3608695652173913e-05,
+      "loss": 0.1905,
+      "step": 1470
+    },
+    {
+      "epoch": 64.0,
+      "eval_accuracy": 0.7952755905511811,
+      "eval_loss": 0.7472063899040222,
+      "eval_runtime": 5.0119,
+      "eval_samples_per_second": 50.68,
+      "eval_steps_per_second": 0.798,
+      "step": 1472
+    },
+    {
+      "epoch": 64.34782608695652,
+      "grad_norm": 17.491132736206055,
+      "learning_rate": 1.3565217391304348e-05,
+      "loss": 0.1824,
+      "step": 1480
+    },
+    {
+      "epoch": 64.78260869565217,
+      "grad_norm": 14.771021842956543,
+      "learning_rate": 1.3521739130434783e-05,
+      "loss": 0.16,
+      "step": 1490
+    },
+    {
+      "epoch": 65.0,
+      "eval_accuracy": 0.7874015748031497,
+      "eval_loss": 0.7761121988296509,
+      "eval_runtime": 6.264,
+      "eval_samples_per_second": 40.549,
+      "eval_steps_per_second": 0.639,
+      "step": 1495
+    },
+    {
+      "epoch": 65.21739130434783,
+      "grad_norm": 11.443580627441406,
+      "learning_rate": 1.3478260869565218e-05,
+      "loss": 0.1691,
+      "step": 1500
+    },
+    {
+      "epoch": 65.65217391304348,
+      "grad_norm": 9.551299095153809,
+      "learning_rate": 1.3434782608695653e-05,
+      "loss": 0.1619,
+      "step": 1510
+    },
+    {
+      "epoch": 66.0,
+      "eval_accuracy": 0.7795275590551181,
+      "eval_loss": 0.858647882938385,
+      "eval_runtime": 5.0805,
+      "eval_samples_per_second": 49.995,
+      "eval_steps_per_second": 0.787,
+      "step": 1518
+    },
+    {
+      "epoch": 66.08695652173913,
+      "grad_norm": 6.219148635864258,
+      "learning_rate": 1.3391304347826088e-05,
+      "loss": 0.1517,
+      "step": 1520
+    },
+    {
+      "epoch": 66.52173913043478,
+      "grad_norm": 18.751920700073242,
+      "learning_rate": 1.3347826086956523e-05,
+      "loss": 0.1635,
+      "step": 1530
+    },
+    {
+      "epoch": 66.95652173913044,
+      "grad_norm": 8.722993850708008,
+      "learning_rate": 1.3304347826086958e-05,
+      "loss": 0.1768,
+      "step": 1540
+    },
+    {
+      "epoch": 67.0,
+      "eval_accuracy": 0.7834645669291339,
+      "eval_loss": 0.7699764966964722,
+      "eval_runtime": 5.8289,
+      "eval_samples_per_second": 43.576,
+      "eval_steps_per_second": 0.686,
+      "step": 1541
+    },
+    {
+      "epoch": 67.3913043478261,
+      "grad_norm": 16.816171646118164,
+      "learning_rate": 1.3260869565217392e-05,
+      "loss": 0.184,
+      "step": 1550
+    },
+    {
+      "epoch": 67.82608695652173,
+      "grad_norm": 10.644329071044922,
+      "learning_rate": 1.3217391304347827e-05,
+      "loss": 0.1395,
+      "step": 1560
+    },
+    {
+      "epoch": 68.0,
+      "eval_accuracy": 0.7716535433070866,
+      "eval_loss": 0.8326212763786316,
+      "eval_runtime": 6.0684,
+      "eval_samples_per_second": 41.856,
+      "eval_steps_per_second": 0.659,
+      "step": 1564
+    },
+    {
+      "epoch": 68.26086956521739,
+      "grad_norm": 13.280193328857422,
+      "learning_rate": 1.3173913043478262e-05,
+      "loss": 0.1954,
+      "step": 1570
+    },
+    {
+      "epoch": 68.69565217391305,
+      "grad_norm": 15.933572769165039,
+      "learning_rate": 1.3130434782608697e-05,
+      "loss": 0.1536,
+      "step": 1580
+    },
+    {
+      "epoch": 69.0,
+      "eval_accuracy": 0.7755905511811023,
+      "eval_loss": 0.8442131280899048,
+      "eval_runtime": 5.1134,
+      "eval_samples_per_second": 49.673,
+      "eval_steps_per_second": 0.782,
+      "step": 1587
+    },
+    {
+      "epoch": 69.1304347826087,
+      "grad_norm": 8.563169479370117,
+      "learning_rate": 1.308695652173913e-05,
+      "loss": 0.1859,
+      "step": 1590
+    },
+    {
+      "epoch": 69.56521739130434,
+      "grad_norm": 52.772857666015625,
+      "learning_rate": 1.3043478260869566e-05,
+      "loss": 0.1987,
+      "step": 1600
+    },
+    {
+      "epoch": 70.0,
+      "grad_norm": 40.454124450683594,
+      "learning_rate": 1.3000000000000001e-05,
+      "loss": 0.208,
+      "step": 1610
+    },
+    {
+      "epoch": 70.0,
+      "eval_accuracy": 0.7677165354330708,
+      "eval_loss": 0.9288918972015381,
+      "eval_runtime": 6.0792,
+      "eval_samples_per_second": 41.782,
+      "eval_steps_per_second": 0.658,
+      "step": 1610
+    },
+    {
+      "epoch": 70.43478260869566,
+      "grad_norm": 6.047575950622559,
+      "learning_rate": 1.2956521739130436e-05,
+      "loss": 0.1494,
+      "step": 1620
+    },
+    {
+      "epoch": 70.8695652173913,
+      "grad_norm": 24.65239143371582,
+      "learning_rate": 1.2913043478260871e-05,
+      "loss": 0.1783,
+      "step": 1630
+    },
+    {
+      "epoch": 71.0,
+      "eval_accuracy": 0.7637795275590551,
+      "eval_loss": 0.9022064805030823,
+      "eval_runtime": 5.9906,
+      "eval_samples_per_second": 42.4,
+      "eval_steps_per_second": 0.668,
+      "step": 1633
+    },
+    {
+      "epoch": 71.30434782608695,
+      "grad_norm": 10.55053424835205,
+      "learning_rate": 1.2869565217391305e-05,
+      "loss": 0.1737,
+      "step": 1640
+    },
+    {
+      "epoch": 71.73913043478261,
+      "grad_norm": 10.525084495544434,
+      "learning_rate": 1.282608695652174e-05,
+      "loss": 0.1572,
+      "step": 1650
+    },
+    {
+      "epoch": 72.0,
+      "eval_accuracy": 0.7677165354330708,
+      "eval_loss": 0.8510112166404724,
+      "eval_runtime": 5.7557,
+      "eval_samples_per_second": 44.13,
+      "eval_steps_per_second": 0.695,
+      "step": 1656
+    },
+    {
+      "epoch": 72.17391304347827,
+      "grad_norm": 8.981342315673828,
+      "learning_rate": 1.2782608695652175e-05,
+      "loss": 0.1407,
+      "step": 1660
+    },
+    {
+      "epoch": 72.6086956521739,
+      "grad_norm": 10.919659614562988,
+      "learning_rate": 1.273913043478261e-05,
+      "loss": 0.1349,
+      "step": 1670
+    },
+    {
+      "epoch": 73.0,
+      "eval_accuracy": 0.7677165354330708,
+      "eval_loss": 0.7962220907211304,
+      "eval_runtime": 6.421,
+      "eval_samples_per_second": 39.558,
+      "eval_steps_per_second": 0.623,
+      "step": 1679
+    },
+    {
+      "epoch": 73.04347826086956,
+      "grad_norm": 16.983463287353516,
+      "learning_rate": 1.2695652173913045e-05,
+      "loss": 0.1981,
+      "step": 1680
+    },
+    {
+      "epoch": 73.47826086956522,
+      "grad_norm": 4.827481269836426,
+      "learning_rate": 1.265217391304348e-05,
+      "loss": 0.1354,
+      "step": 1690
+    },
+    {
+      "epoch": 73.91304347826087,
+      "grad_norm": 7.88926887512207,
+      "learning_rate": 1.2608695652173915e-05,
+      "loss": 0.148,
+      "step": 1700
+    },
+    {
+      "epoch": 74.0,
+      "eval_accuracy": 0.7755905511811023,
+      "eval_loss": 0.8641188144683838,
+      "eval_runtime": 6.5823,
+      "eval_samples_per_second": 38.588,
+      "eval_steps_per_second": 0.608,
+      "step": 1702
+    },
+    {
+      "epoch": 74.34782608695652,
+      "grad_norm": 15.213589668273926,
+      "learning_rate": 1.2565217391304349e-05,
+      "loss": 0.1763,
+      "step": 1710
+    },
+    {
+      "epoch": 74.78260869565217,
+      "grad_norm": 10.891641616821289,
+      "learning_rate": 1.2521739130434784e-05,
+      "loss": 0.1768,
+      "step": 1720
+    },
+    {
+      "epoch": 75.0,
+      "eval_accuracy": 0.7677165354330708,
+      "eval_loss": 0.9277058839797974,
+      "eval_runtime": 5.6008,
+      "eval_samples_per_second": 45.351,
+      "eval_steps_per_second": 0.714,
+      "step": 1725
+    },
+    {
+      "epoch": 75.21739130434783,
+      "grad_norm": 17.890514373779297,
+      "learning_rate": 1.2478260869565217e-05,
+      "loss": 0.1467,
+      "step": 1730
+    },
+    {
+      "epoch": 75.65217391304348,
+      "grad_norm": 19.329357147216797,
+      "learning_rate": 1.2434782608695652e-05,
+      "loss": 0.1833,
+      "step": 1740
+    },
+    {
+      "epoch": 76.0,
+      "eval_accuracy": 0.7637795275590551,
+      "eval_loss": 0.8663110136985779,
+      "eval_runtime": 5.6029,
+      "eval_samples_per_second": 45.334,
+      "eval_steps_per_second": 0.714,
+      "step": 1748
+    },
+    {
+      "epoch": 76.08695652173913,
+      "grad_norm": 21.218746185302734,
+      "learning_rate": 1.2391304347826088e-05,
+      "loss": 0.1451,
+      "step": 1750
+    },
+    {
+      "epoch": 76.52173913043478,
+      "grad_norm": 14.454313278198242,
+      "learning_rate": 1.2347826086956523e-05,
+      "loss": 0.1568,
+      "step": 1760
+    },
+    {
+      "epoch": 76.95652173913044,
+      "grad_norm": 17.544715881347656,
+      "learning_rate": 1.2304347826086958e-05,
+      "loss": 0.1696,
+      "step": 1770
+    },
+    {
+      "epoch": 77.0,
+      "eval_accuracy": 0.7755905511811023,
+      "eval_loss": 0.8302040696144104,
+      "eval_runtime": 5.0767,
+      "eval_samples_per_second": 50.032,
+      "eval_steps_per_second": 0.788,
+      "step": 1771
+    },
+    {
+      "epoch": 77.3913043478261,
+      "grad_norm": 23.46592903137207,
+      "learning_rate": 1.2260869565217393e-05,
+      "loss": 0.1504,
+      "step": 1780
+    },
+    {
+      "epoch": 77.82608695652173,
+      "grad_norm": 10.731809616088867,
+      "learning_rate": 1.2217391304347828e-05,
+      "loss": 0.1577,
+      "step": 1790
+    },
+    {
+      "epoch": 78.0,
+      "eval_accuracy": 0.7637795275590551,
+      "eval_loss": 0.8575807213783264,
+      "eval_runtime": 4.9965,
+      "eval_samples_per_second": 50.836,
+      "eval_steps_per_second": 0.801,
+      "step": 1794
+    },
+    {
+      "epoch": 78.26086956521739,
+      "grad_norm": 13.635187149047852,
+      "learning_rate": 1.2173913043478263e-05,
+      "loss": 0.1316,
+      "step": 1800
+    },
+    {
+      "epoch": 78.69565217391305,
+      "grad_norm": 13.30216121673584,
+      "learning_rate": 1.2130434782608698e-05,
+      "loss": 0.1724,
+      "step": 1810
+    },
+    {
+      "epoch": 79.0,
+      "eval_accuracy": 0.7598425196850394,
+      "eval_loss": 0.8651734590530396,
+      "eval_runtime": 5.1165,
+      "eval_samples_per_second": 49.643,
+      "eval_steps_per_second": 0.782,
+      "step": 1817
+    },
+    {
+      "epoch": 79.1304347826087,
+      "grad_norm": 12.134591102600098,
+      "learning_rate": 1.208695652173913e-05,
+      "loss": 0.1608,
+      "step": 1820
+    },
+    {
+      "epoch": 79.56521739130434,
+      "grad_norm": 11.177396774291992,
+      "learning_rate": 1.2043478260869565e-05,
+      "loss": 0.1615,
+      "step": 1830
+    },
+    {
+      "epoch": 80.0,
+      "grad_norm": 9.49387264251709,
+      "learning_rate": 1.2e-05,
+      "loss": 0.1525,
+      "step": 1840
+    },
+    {
+      "epoch": 80.0,
+      "eval_accuracy": 0.7716535433070866,
+      "eval_loss": 0.8567054271697998,
+      "eval_runtime": 5.5937,
+      "eval_samples_per_second": 45.408,
+      "eval_steps_per_second": 0.715,
+      "step": 1840
+    },
+    {
+      "epoch": 80.43478260869566,
+      "grad_norm": 7.956330299377441,
+      "learning_rate": 1.1956521739130435e-05,
+      "loss": 0.1478,
+      "step": 1850
+    },
+    {
+      "epoch": 80.8695652173913,
+      "grad_norm": 14.717869758605957,
+      "learning_rate": 1.191304347826087e-05,
+      "loss": 0.158,
+      "step": 1860
+    },
+    {
+      "epoch": 81.0,
+      "eval_accuracy": 0.7598425196850394,
+      "eval_loss": 0.9139416217803955,
+      "eval_runtime": 5.5012,
+      "eval_samples_per_second": 46.172,
+      "eval_steps_per_second": 0.727,
+      "step": 1863
+    },
+    {
+      "epoch": 81.30434782608695,
+      "grad_norm": 9.091207504272461,
+      "learning_rate": 1.1869565217391306e-05,
+      "loss": 0.1738,
+      "step": 1870
+    },
+    {
+      "epoch": 81.73913043478261,
+      "grad_norm": 13.691263198852539,
+      "learning_rate": 1.182608695652174e-05,
+      "loss": 0.1639,
+      "step": 1880
+    },
+    {
+      "epoch": 82.0,
+      "eval_accuracy": 0.7519685039370079,
+      "eval_loss": 0.9688799977302551,
+      "eval_runtime": 5.0902,
+      "eval_samples_per_second": 49.9,
+      "eval_steps_per_second": 0.786,
+      "step": 1886
+    },
+    {
+      "epoch": 82.17391304347827,
+      "grad_norm": 17.235044479370117,
+      "learning_rate": 1.1782608695652176e-05,
+      "loss": 0.1849,
+      "step": 1890
+    },
+    {
+      "epoch": 82.6086956521739,
+      "grad_norm": 22.74346923828125,
+      "learning_rate": 1.1739130434782611e-05,
+      "loss": 0.1424,
+      "step": 1900
+    },
+    {
+      "epoch": 83.0,
+      "eval_accuracy": 0.7637795275590551,
+      "eval_loss": 0.9698485136032104,
+      "eval_runtime": 4.9865,
+      "eval_samples_per_second": 50.938,
+      "eval_steps_per_second": 0.802,
+      "step": 1909
+    },
+    {
+      "epoch": 83.04347826086956,
+      "grad_norm": 16.867551803588867,
+      "learning_rate": 1.1695652173913043e-05,
+      "loss": 0.1463,
+      "step": 1910
+    },
+    {
+      "epoch": 83.47826086956522,
+      "grad_norm": 5.856212615966797,
+      "learning_rate": 1.1652173913043478e-05,
+      "loss": 0.1382,
+      "step": 1920
+    },
+    {
+      "epoch": 83.91304347826087,
+      "grad_norm": 14.235991477966309,
+      "learning_rate": 1.1608695652173913e-05,
+      "loss": 0.1224,
+      "step": 1930
+    },
+    {
+      "epoch": 84.0,
+      "eval_accuracy": 0.7716535433070866,
+      "eval_loss": 1.0238784551620483,
+      "eval_runtime": 5.1148,
+      "eval_samples_per_second": 49.66,
+      "eval_steps_per_second": 0.782,
+      "step": 1932
+    },
+    {
+      "epoch": 84.34782608695652,
+      "grad_norm": 5.4066925048828125,
+      "learning_rate": 1.1565217391304348e-05,
+      "loss": 0.1361,
+      "step": 1940
+    },
+    {
+      "epoch": 84.78260869565217,
+      "grad_norm": 9.836471557617188,
+      "learning_rate": 1.1521739130434783e-05,
+      "loss": 0.1765,
+      "step": 1950
+    },
+    {
+      "epoch": 85.0,
+      "eval_accuracy": 0.7795275590551181,
+      "eval_loss": 0.9071526527404785,
+      "eval_runtime": 6.2721,
+      "eval_samples_per_second": 40.497,
+      "eval_steps_per_second": 0.638,
+      "step": 1955
+    },
+    {
+      "epoch": 85.21739130434783,
+      "grad_norm": 17.78875160217285,
+      "learning_rate": 1.1478260869565218e-05,
+      "loss": 0.1696,
+      "step": 1960
+    },
+    {
+      "epoch": 85.65217391304348,
+      "grad_norm": 12.22424602508545,
+      "learning_rate": 1.1434782608695654e-05,
+      "loss": 0.1726,
+      "step": 1970
+    },
+    {
+      "epoch": 86.0,
+      "eval_accuracy": 0.7519685039370079,
+      "eval_loss": 0.9436205625534058,
+      "eval_runtime": 5.3198,
+      "eval_samples_per_second": 47.746,
+      "eval_steps_per_second": 0.752,
+      "step": 1978
+    },
+    {
+      "epoch": 86.08695652173913,
+      "grad_norm": 18.984779357910156,
+      "learning_rate": 1.1391304347826089e-05,
+      "loss": 0.1716,
+      "step": 1980
+    },
+    {
+      "epoch": 86.52173913043478,
+      "grad_norm": 12.640647888183594,
+      "learning_rate": 1.1347826086956524e-05,
+      "loss": 0.153,
+      "step": 1990
+    },
+    {
+      "epoch": 86.95652173913044,
+      "grad_norm": 20.94854736328125,
+      "learning_rate": 1.1304347826086957e-05,
+      "loss": 0.1584,
+      "step": 2000
+    },
+    {
+      "epoch": 87.0,
+      "eval_accuracy": 0.7637795275590551,
+      "eval_loss": 0.8775471448898315,
+      "eval_runtime": 5.2147,
+      "eval_samples_per_second": 48.708,
+      "eval_steps_per_second": 0.767,
+      "step": 2001
+    },
+    {
+      "epoch": 87.3913043478261,
+      "grad_norm": 16.81257438659668,
+      "learning_rate": 1.1260869565217392e-05,
+      "loss": 0.1945,
+      "step": 2010
+    },
+    {
+      "epoch": 87.82608695652173,
+      "grad_norm": 7.593163013458252,
+      "learning_rate": 1.1217391304347827e-05,
+      "loss": 0.164,
+      "step": 2020
+    },
+    {
+      "epoch": 88.0,
+      "eval_accuracy": 0.7716535433070866,
+      "eval_loss": 0.8591587543487549,
+      "eval_runtime": 6.3339,
+      "eval_samples_per_second": 40.102,
+      "eval_steps_per_second": 0.632,
+      "step": 2024
+    },
+    {
+      "epoch": 88.26086956521739,
+      "grad_norm": 27.010921478271484,
+      "learning_rate": 1.1173913043478261e-05,
+      "loss": 0.1514,
+      "step": 2030
+    },
+    {
+      "epoch": 88.69565217391305,
+      "grad_norm": 11.01710033416748,
+      "learning_rate": 1.1130434782608696e-05,
+      "loss": 0.1682,
+      "step": 2040
+    },
+    {
+      "epoch": 89.0,
+      "eval_accuracy": 0.7637795275590551,
+      "eval_loss": 0.9051375389099121,
+      "eval_runtime": 6.2969,
+      "eval_samples_per_second": 40.337,
+      "eval_steps_per_second": 0.635,
+      "step": 2047
+    },
+    {
+      "epoch": 89.1304347826087,
+      "grad_norm": 7.84626579284668,
+      "learning_rate": 1.1086956521739131e-05,
+      "loss": 0.1364,
+      "step": 2050
+    },
+    {
+      "epoch": 89.56521739130434,
+      "grad_norm": 10.766637802124023,
+      "learning_rate": 1.1043478260869566e-05,
+      "loss": 0.2003,
+      "step": 2060
+    },
+    {
+      "epoch": 90.0,
+      "grad_norm": 16.38211441040039,
+      "learning_rate": 1.1000000000000001e-05,
+      "loss": 0.1455,
+      "step": 2070
+    },
+    {
+      "epoch": 90.0,
+      "eval_accuracy": 0.7716535433070866,
+      "eval_loss": 1.0020496845245361,
+      "eval_runtime": 6.3353,
+      "eval_samples_per_second": 40.093,
+      "eval_steps_per_second": 0.631,
+      "step": 2070
+    },
+    {
+      "epoch": 90.43478260869566,
+      "grad_norm": 8.20938491821289,
+      "learning_rate": 1.0956521739130435e-05,
+      "loss": 0.1372,
+      "step": 2080
+    },
+    {
+      "epoch": 90.8695652173913,
+      "grad_norm": 29.13675880432129,
+      "learning_rate": 1.091304347826087e-05,
+      "loss": 0.1596,
+      "step": 2090
+    },
+    {
+      "epoch": 91.0,
+      "eval_accuracy": 0.7677165354330708,
+      "eval_loss": 0.9422626495361328,
+      "eval_runtime": 5.3345,
+      "eval_samples_per_second": 47.614,
+      "eval_steps_per_second": 0.75,
+      "step": 2093
+    },
+    {
+      "epoch": 91.30434782608695,
+      "grad_norm": 16.11428451538086,
+      "learning_rate": 1.0869565217391305e-05,
+      "loss": 0.1704,
+      "step": 2100
+    },
+    {
+      "epoch": 91.73913043478261,
+      "grad_norm": 11.314476013183594,
+      "learning_rate": 1.082608695652174e-05,
+      "loss": 0.1667,
+      "step": 2110
+    },
+    {
+      "epoch": 92.0,
+      "eval_accuracy": 0.7637795275590551,
+      "eval_loss": 0.9586116671562195,
+      "eval_runtime": 6.2772,
+      "eval_samples_per_second": 40.464,
+      "eval_steps_per_second": 0.637,
+      "step": 2116
+    },
+    {
+      "epoch": 92.17391304347827,
+      "grad_norm": 24.264659881591797,
+      "learning_rate": 1.0782608695652175e-05,
+      "loss": 0.1439,
+      "step": 2120
+    },
+    {
+      "epoch": 92.6086956521739,
+      "grad_norm": 13.49873161315918,
+      "learning_rate": 1.073913043478261e-05,
+      "loss": 0.132,
+      "step": 2130
+    },
+    {
+      "epoch": 93.0,
+      "eval_accuracy": 0.7637795275590551,
+      "eval_loss": 0.9890474081039429,
+      "eval_runtime": 6.376,
+      "eval_samples_per_second": 39.837,
+      "eval_steps_per_second": 0.627,
+      "step": 2139
+    },
+    {
+      "epoch": 93.04347826086956,
+      "grad_norm": 14.934507369995117,
+      "learning_rate": 1.0695652173913046e-05,
+      "loss": 0.1715,
+      "step": 2140
+    },
+    {
+      "epoch": 93.47826086956522,
+      "grad_norm": 7.096430778503418,
+      "learning_rate": 1.0652173913043479e-05,
+      "loss": 0.1454,
+      "step": 2150
+    },
+    {
+      "epoch": 93.91304347826087,
+      "grad_norm": 15.230493545532227,
+      "learning_rate": 1.0608695652173914e-05,
+      "loss": 0.1335,
+      "step": 2160
+    },
+    {
+      "epoch": 94.0,
+      "eval_accuracy": 0.7716535433070866,
+      "eval_loss": 0.9922319054603577,
+      "eval_runtime": 5.2755,
+      "eval_samples_per_second": 48.147,
+      "eval_steps_per_second": 0.758,
+      "step": 2162
+    },
+    {
+      "epoch": 94.34782608695652,
+      "grad_norm": 17.714771270751953,
+      "learning_rate": 1.0565217391304348e-05,
+      "loss": 0.1593,
+      "step": 2170
+    },
+    {
+      "epoch": 94.78260869565217,
+      "grad_norm": 33.39093780517578,
+      "learning_rate": 1.0521739130434783e-05,
+      "loss": 0.1538,
+      "step": 2180
+    },
+    {
+      "epoch": 95.0,
+      "eval_accuracy": 0.7519685039370079,
+      "eval_loss": 0.9534251093864441,
+      "eval_runtime": 5.5144,
+      "eval_samples_per_second": 46.061,
+      "eval_steps_per_second": 0.725,
+      "step": 2185
+    },
+    {
+      "epoch": 95.21739130434783,
+      "grad_norm": 17.226415634155273,
+      "learning_rate": 1.0478260869565218e-05,
+      "loss": 0.1422,
+      "step": 2190
+    },
+    {
+      "epoch": 95.65217391304348,
+      "grad_norm": 17.667068481445312,
+      "learning_rate": 1.0434782608695653e-05,
+      "loss": 0.1288,
+      "step": 2200
+    },
+    {
+      "epoch": 96.0,
+      "eval_accuracy": 0.7480314960629921,
+      "eval_loss": 1.0713515281677246,
+      "eval_runtime": 5.5478,
+      "eval_samples_per_second": 45.784,
+      "eval_steps_per_second": 0.721,
+      "step": 2208
+    },
+    {
+      "epoch": 96.08695652173913,
+      "grad_norm": 10.762154579162598,
+      "learning_rate": 1.0391304347826088e-05,
+      "loss": 0.1204,
+      "step": 2210
+    },
+    {
+      "epoch": 96.52173913043478,
+      "grad_norm": 31.42228126525879,
+      "learning_rate": 1.0347826086956523e-05,
+      "loss": 0.1401,
+      "step": 2220
+    },
+    {
+      "epoch": 96.95652173913044,
+      "grad_norm": 10.302103996276855,
+      "learning_rate": 1.0304347826086958e-05,
+      "loss": 0.1661,
+      "step": 2230
+    },
+    {
+      "epoch": 97.0,
+      "eval_accuracy": 0.7598425196850394,
+      "eval_loss": 0.9949864745140076,
+      "eval_runtime": 6.453,
+      "eval_samples_per_second": 39.361,
+      "eval_steps_per_second": 0.62,
+      "step": 2231
+    },
+    {
+      "epoch": 97.3913043478261,
+      "grad_norm": 13.885024070739746,
+      "learning_rate": 1.0260869565217393e-05,
+      "loss": 0.171,
+      "step": 2240
+    },
+    {
+      "epoch": 97.82608695652173,
+      "grad_norm": 20.33379364013672,
+      "learning_rate": 1.0217391304347829e-05,
+      "loss": 0.1392,
+      "step": 2250
+    },
+    {
+      "epoch": 98.0,
+      "eval_accuracy": 0.7519685039370079,
+      "eval_loss": 0.9865681529045105,
+      "eval_runtime": 5.2827,
+      "eval_samples_per_second": 48.081,
+      "eval_steps_per_second": 0.757,
+      "step": 2254
+    },
+    {
+      "epoch": 98.26086956521739,
+      "grad_norm": 22.363279342651367,
+      "learning_rate": 1.017391304347826e-05,
+      "loss": 0.1669,
+      "step": 2260
+    },
+    {
+      "epoch": 98.69565217391305,
+      "grad_norm": 27.860618591308594,
+      "learning_rate": 1.0130434782608695e-05,
+      "loss": 0.1413,
+      "step": 2270
+    },
+    {
+      "epoch": 99.0,
+      "eval_accuracy": 0.7598425196850394,
+      "eval_loss": 1.0637905597686768,
+      "eval_runtime": 5.2324,
+      "eval_samples_per_second": 48.544,
+      "eval_steps_per_second": 0.764,
+      "step": 2277
+    },
+    {
+      "epoch": 99.1304347826087,
+      "grad_norm": 15.204286575317383,
+      "learning_rate": 1.008695652173913e-05,
+      "loss": 0.1153,
+      "step": 2280
+    },
+    {
+      "epoch": 99.56521739130434,
+      "grad_norm": 14.230497360229492,
+      "learning_rate": 1.0043478260869566e-05,
+      "loss": 0.1309,
+      "step": 2290
+    },
+    {
+      "epoch": 100.0,
+      "grad_norm": 14.642973899841309,
+      "learning_rate": 1e-05,
+      "loss": 0.1619,
+      "step": 2300
+    },
+    {
+      "epoch": 100.0,
+      "eval_accuracy": 0.7598425196850394,
+      "eval_loss": 1.0178251266479492,
+      "eval_runtime": 5.2064,
+      "eval_samples_per_second": 48.787,
+      "eval_steps_per_second": 0.768,
+      "step": 2300
+    },
+    {
+      "epoch": 100.43478260869566,
+      "grad_norm": 64.10249328613281,
+      "learning_rate": 9.956521739130436e-06,
+      "loss": 0.1249,
+      "step": 2310
+    },
+    {
+      "epoch": 100.8695652173913,
+      "grad_norm": 8.547202110290527,
+      "learning_rate": 9.913043478260871e-06,
+      "loss": 0.1537,
+      "step": 2320
+    },
+    {
+      "epoch": 101.0,
+      "eval_accuracy": 0.7637795275590551,
+      "eval_loss": 0.9891794919967651,
+      "eval_runtime": 5.2872,
+      "eval_samples_per_second": 48.04,
+      "eval_steps_per_second": 0.757,
+      "step": 2323
+    },
+    {
+      "epoch": 101.30434782608695,
+      "grad_norm": 21.414203643798828,
+      "learning_rate": 9.869565217391304e-06,
+      "loss": 0.1626,
+      "step": 2330
+    },
+    {
+      "epoch": 101.73913043478261,
+      "grad_norm": 7.053102970123291,
+      "learning_rate": 9.82608695652174e-06,
+      "loss": 0.137,
+      "step": 2340
+    },
+    {
+      "epoch": 102.0,
+      "eval_accuracy": 0.7559055118110236,
+      "eval_loss": 0.9523509740829468,
+      "eval_runtime": 5.9665,
+      "eval_samples_per_second": 42.571,
+      "eval_steps_per_second": 0.67,
+      "step": 2346
+    },
+    {
+      "epoch": 102.17391304347827,
+      "grad_norm": 18.426395416259766,
+      "learning_rate": 9.782608695652175e-06,
+      "loss": 0.1623,
+      "step": 2350
+    },
+    {
+      "epoch": 102.6086956521739,
+      "grad_norm": 7.689826011657715,
+      "learning_rate": 9.73913043478261e-06,
+      "loss": 0.1416,
+      "step": 2360
+    },
+    {
+      "epoch": 103.0,
+      "eval_accuracy": 0.7401574803149606,
+      "eval_loss": 1.0539218187332153,
+      "eval_runtime": 6.2316,
+      "eval_samples_per_second": 40.76,
+      "eval_steps_per_second": 0.642,
+      "step": 2369
+    },
+    {
+      "epoch": 103.04347826086956,
+      "grad_norm": 7.151523590087891,
+      "learning_rate": 9.695652173913043e-06,
+      "loss": 0.1245,
+      "step": 2370
+    },
+    {
+      "epoch": 103.47826086956522,
+      "grad_norm": 15.233275413513184,
+      "learning_rate": 9.652173913043478e-06,
+      "loss": 0.1347,
+      "step": 2380
+    },
+    {
+      "epoch": 103.91304347826087,
+      "grad_norm": 6.604703903198242,
+      "learning_rate": 9.608695652173914e-06,
+      "loss": 0.1477,
+      "step": 2390
+    },
+    {
+      "epoch": 104.0,
+      "eval_accuracy": 0.7283464566929134,
+      "eval_loss": 1.0824700593948364,
+      "eval_runtime": 5.2926,
+      "eval_samples_per_second": 47.992,
+      "eval_steps_per_second": 0.756,
+      "step": 2392
+    },
+    {
+      "epoch": 104.34782608695652,
+      "grad_norm": 16.267160415649414,
+      "learning_rate": 9.565217391304349e-06,
+      "loss": 0.1353,
+      "step": 2400
+    },
+    {
+      "epoch": 104.78260869565217,
+      "grad_norm": 14.780786514282227,
+      "learning_rate": 9.521739130434784e-06,
+      "loss": 0.1283,
+      "step": 2410
+    },
+    {
+      "epoch": 105.0,
+      "eval_accuracy": 0.7519685039370079,
+      "eval_loss": 1.0008113384246826,
+      "eval_runtime": 5.335,
+      "eval_samples_per_second": 47.61,
+      "eval_steps_per_second": 0.75,
+      "step": 2415
+    },
+    {
+      "epoch": 105.21739130434783,
+      "grad_norm": 18.90896224975586,
+      "learning_rate": 9.478260869565217e-06,
+      "loss": 0.1235,
+      "step": 2420
+    },
+    {
+      "epoch": 105.65217391304348,
+      "grad_norm": 13.357507705688477,
+      "learning_rate": 9.434782608695652e-06,
+      "loss": 0.1498,
+      "step": 2430
+    },
+    {
+      "epoch": 106.0,
+      "eval_accuracy": 0.7637795275590551,
+      "eval_loss": 0.9702416658401489,
+      "eval_runtime": 5.3141,
+      "eval_samples_per_second": 47.797,
+      "eval_steps_per_second": 0.753,
+      "step": 2438
+    },
+    {
+      "epoch": 106.08695652173913,
+      "grad_norm": 15.780552864074707,
+      "learning_rate": 9.391304347826087e-06,
+      "loss": 0.1228,
+      "step": 2440
+    },
+    {
+      "epoch": 106.52173913043478,
+      "grad_norm": 17.9113826751709,
+      "learning_rate": 9.347826086956523e-06,
+      "loss": 0.1394,
+      "step": 2450
+    },
+    {
+      "epoch": 106.95652173913044,
+      "grad_norm": 9.112041473388672,
+      "learning_rate": 9.304347826086956e-06,
+      "loss": 0.1576,
+      "step": 2460
+    },
+    {
+      "epoch": 107.0,
+      "eval_accuracy": 0.7677165354330708,
+      "eval_loss": 1.0144392251968384,
+      "eval_runtime": 5.3399,
+      "eval_samples_per_second": 47.567,
+      "eval_steps_per_second": 0.749,
+      "step": 2461
+    },
+    {
+      "epoch": 107.3913043478261,
+      "grad_norm": 19.214317321777344,
+      "learning_rate": 9.260869565217391e-06,
+      "loss": 0.1488,
+      "step": 2470
+    },
+    {
+      "epoch": 107.82608695652173,
+      "grad_norm": 12.886323928833008,
+      "learning_rate": 9.217391304347826e-06,
+      "loss": 0.1433,
+      "step": 2480
+    },
+    {
+      "epoch": 108.0,
+      "eval_accuracy": 0.7637795275590551,
+      "eval_loss": 0.945662260055542,
+      "eval_runtime": 5.4883,
+      "eval_samples_per_second": 46.28,
+      "eval_steps_per_second": 0.729,
+      "step": 2484
+    },
+    {
+      "epoch": 108.26086956521739,
+      "grad_norm": 6.37844181060791,
+      "learning_rate": 9.173913043478261e-06,
+      "loss": 0.1411,
+      "step": 2490
+    },
+    {
+      "epoch": 108.69565217391305,
+      "grad_norm": 8.28097152709961,
+      "learning_rate": 9.130434782608697e-06,
+      "loss": 0.1377,
+      "step": 2500
+    },
+    {
+      "epoch": 109.0,
+      "eval_accuracy": 0.7677165354330708,
+      "eval_loss": 0.9770110249519348,
+      "eval_runtime": 5.6903,
+      "eval_samples_per_second": 44.638,
+      "eval_steps_per_second": 0.703,
+      "step": 2507
+    },
+    {
+      "epoch": 109.1304347826087,
+      "grad_norm": 14.171891212463379,
+      "learning_rate": 9.086956521739132e-06,
+      "loss": 0.1449,
+      "step": 2510
+    },
+    {
+      "epoch": 109.56521739130434,
+      "grad_norm": 7.651449203491211,
+      "learning_rate": 9.043478260869565e-06,
+      "loss": 0.1068,
+      "step": 2520
+    },
+    {
+      "epoch": 110.0,
+      "grad_norm": 8.761197090148926,
+      "learning_rate": 9e-06,
+      "loss": 0.1163,
+      "step": 2530
+    },
+    {
+      "epoch": 110.0,
+      "eval_accuracy": 0.7559055118110236,
+      "eval_loss": 1.1386067867279053,
+      "eval_runtime": 5.7019,
+      "eval_samples_per_second": 44.546,
+      "eval_steps_per_second": 0.702,
+      "step": 2530
+    },
+    {
+      "epoch": 110.43478260869566,
+      "grad_norm": 15.50783634185791,
+      "learning_rate": 8.956521739130435e-06,
+      "loss": 0.1379,
+      "step": 2540
+    },
+    {
+      "epoch": 110.8695652173913,
+      "grad_norm": 19.807313919067383,
+      "learning_rate": 8.91304347826087e-06,
+      "loss": 0.1449,
+      "step": 2550
+    },
+    {
+      "epoch": 111.0,
+      "eval_accuracy": 0.7559055118110236,
+      "eval_loss": 1.0589202642440796,
+      "eval_runtime": 6.1192,
+      "eval_samples_per_second": 41.509,
+      "eval_steps_per_second": 0.654,
+      "step": 2553
+    },
+    {
+      "epoch": 111.30434782608695,
+      "grad_norm": 15.006197929382324,
+      "learning_rate": 8.869565217391306e-06,
+      "loss": 0.1397,
+      "step": 2560
+    },
+    {
+      "epoch": 111.73913043478261,
+      "grad_norm": 24.991825103759766,
+      "learning_rate": 8.82608695652174e-06,
+      "loss": 0.1475,
+      "step": 2570
+    },
+    {
+      "epoch": 112.0,
+      "eval_accuracy": 0.7480314960629921,
+      "eval_loss": 1.0109583139419556,
+      "eval_runtime": 6.251,
+      "eval_samples_per_second": 40.633,
+      "eval_steps_per_second": 0.64,
+      "step": 2576
+    },
+    {
+      "epoch": 112.17391304347827,
+      "grad_norm": 15.339753150939941,
+      "learning_rate": 8.782608695652174e-06,
+      "loss": 0.1204,
+      "step": 2580
+    },
+    {
+      "epoch": 112.6086956521739,
+      "grad_norm": 17.208599090576172,
+      "learning_rate": 8.73913043478261e-06,
+      "loss": 0.1582,
+      "step": 2590
+    },
+    {
+      "epoch": 113.0,
+      "eval_accuracy": 0.7677165354330708,
+      "eval_loss": 0.9657435417175293,
+      "eval_runtime": 5.3503,
+      "eval_samples_per_second": 47.474,
+      "eval_steps_per_second": 0.748,
+      "step": 2599
+    },
+    {
+      "epoch": 113.04347826086956,
+      "grad_norm": 16.048084259033203,
+      "learning_rate": 8.695652173913044e-06,
+      "loss": 0.1543,
+      "step": 2600
+    },
+    {
+      "epoch": 113.47826086956522,
+      "grad_norm": 10.532240867614746,
+      "learning_rate": 8.65217391304348e-06,
+      "loss": 0.1468,
+      "step": 2610
+    },
+    {
+      "epoch": 113.91304347826087,
+      "grad_norm": 10.69536018371582,
+      "learning_rate": 8.608695652173915e-06,
+      "loss": 0.1291,
+      "step": 2620
+    },
+    {
+      "epoch": 114.0,
+      "eval_accuracy": 0.7755905511811023,
+      "eval_loss": 0.9563351273536682,
+      "eval_runtime": 5.5866,
+      "eval_samples_per_second": 45.466,
+      "eval_steps_per_second": 0.716,
+      "step": 2622
+    },
+    {
+      "epoch": 114.34782608695652,
+      "grad_norm": 14.370439529418945,
+      "learning_rate": 8.56521739130435e-06,
+      "loss": 0.1491,
+      "step": 2630
+    },
+    {
+      "epoch": 114.78260869565217,
+      "grad_norm": 6.434089660644531,
+      "learning_rate": 8.521739130434783e-06,
+      "loss": 0.1106,
+      "step": 2640
+    },
+    {
+      "epoch": 115.0,
+      "eval_accuracy": 0.7480314960629921,
+      "eval_loss": 1.1004494428634644,
+      "eval_runtime": 5.299,
+      "eval_samples_per_second": 47.934,
+      "eval_steps_per_second": 0.755,
+      "step": 2645
+    },
+    {
+      "epoch": 115.21739130434783,
+      "grad_norm": 16.293176651000977,
+      "learning_rate": 8.478260869565218e-06,
+      "loss": 0.1276,
+      "step": 2650
+    },
+    {
+      "epoch": 115.65217391304348,
+      "grad_norm": 6.88930082321167,
+      "learning_rate": 8.434782608695653e-06,
+      "loss": 0.1339,
+      "step": 2660
+    },
+    {
+      "epoch": 116.0,
+      "eval_accuracy": 0.7519685039370079,
+      "eval_loss": 1.0326673984527588,
+      "eval_runtime": 5.6971,
+      "eval_samples_per_second": 44.584,
+      "eval_steps_per_second": 0.702,
+      "step": 2668
+    },
+    {
+      "epoch": 116.08695652173913,
+      "grad_norm": 13.512252807617188,
+      "learning_rate": 8.391304347826089e-06,
+      "loss": 0.1317,
+      "step": 2670
+    },
+    {
+      "epoch": 116.52173913043478,
+      "grad_norm": 6.733714580535889,
+      "learning_rate": 8.347826086956522e-06,
+      "loss": 0.1209,
+      "step": 2680
+    },
+    {
+      "epoch": 116.95652173913044,
+      "grad_norm": 15.30413818359375,
+      "learning_rate": 8.304347826086957e-06,
+      "loss": 0.1344,
+      "step": 2690
+    },
+    {
+      "epoch": 117.0,
+      "eval_accuracy": 0.7519685039370079,
+      "eval_loss": 1.0160647630691528,
+      "eval_runtime": 5.8565,
+      "eval_samples_per_second": 43.371,
+      "eval_steps_per_second": 0.683,
+      "step": 2691
+    },
+    {
+      "epoch": 117.3913043478261,
+      "grad_norm": 15.041736602783203,
+      "learning_rate": 8.260869565217392e-06,
+      "loss": 0.1674,
+      "step": 2700
+    },
+    {
+      "epoch": 117.82608695652173,
+      "grad_norm": 23.318103790283203,
+      "learning_rate": 8.217391304347827e-06,
+      "loss": 0.1433,
+      "step": 2710
+    },
+    {
+      "epoch": 118.0,
+      "eval_accuracy": 0.7559055118110236,
+      "eval_loss": 1.0311795473098755,
+      "eval_runtime": 5.6699,
+      "eval_samples_per_second": 44.798,
+      "eval_steps_per_second": 0.705,
+      "step": 2714
+    },
+    {
+      "epoch": 118.26086956521739,
+      "grad_norm": 9.980913162231445,
+      "learning_rate": 8.173913043478263e-06,
+      "loss": 0.1347,
+      "step": 2720
+    },
+    {
+      "epoch": 118.69565217391305,
+      "grad_norm": 7.0901408195495605,
+      "learning_rate": 8.130434782608696e-06,
+      "loss": 0.1271,
+      "step": 2730
+    },
+    {
+      "epoch": 119.0,
+      "eval_accuracy": 0.7598425196850394,
+      "eval_loss": 1.0266056060791016,
+      "eval_runtime": 5.2706,
+      "eval_samples_per_second": 48.192,
+      "eval_steps_per_second": 0.759,
+      "step": 2737
+    },
+    {
+      "epoch": 119.1304347826087,
+      "grad_norm": 43.82113265991211,
+      "learning_rate": 8.086956521739131e-06,
+      "loss": 0.1068,
+      "step": 2740
+    },
+    {
+      "epoch": 119.56521739130434,
+      "grad_norm": 7.058877468109131,
+      "learning_rate": 8.043478260869566e-06,
+      "loss": 0.1144,
+      "step": 2750
+    },
+    {
+      "epoch": 120.0,
+      "grad_norm": 4.748641014099121,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.1222,
+      "step": 2760
+    },
+    {
+      "epoch": 120.0,
+      "eval_accuracy": 0.7637795275590551,
+      "eval_loss": 1.0119496583938599,
+      "eval_runtime": 6.2061,
+      "eval_samples_per_second": 40.928,
+      "eval_steps_per_second": 0.645,
+      "step": 2760
+    },
+    {
+      "epoch": 120.43478260869566,
+      "grad_norm": 15.84010124206543,
+      "learning_rate": 7.956521739130435e-06,
+      "loss": 0.1193,
+      "step": 2770
+    },
+    {
+      "epoch": 120.8695652173913,
+      "grad_norm": 13.664319038391113,
+      "learning_rate": 7.91304347826087e-06,
+      "loss": 0.1235,
+      "step": 2780
+    },
+    {
+      "epoch": 121.0,
+      "eval_accuracy": 0.7519685039370079,
+      "eval_loss": 1.0808361768722534,
+      "eval_runtime": 5.23,
+      "eval_samples_per_second": 48.566,
+      "eval_steps_per_second": 0.765,
+      "step": 2783
+    },
+    {
+      "epoch": 121.30434782608695,
+      "grad_norm": 12.779178619384766,
+      "learning_rate": 7.869565217391305e-06,
+      "loss": 0.1322,
+      "step": 2790
+    },
+    {
+      "epoch": 121.73913043478261,
+      "grad_norm": 13.491189002990723,
+      "learning_rate": 7.82608695652174e-06,
+      "loss": 0.1311,
+      "step": 2800
+    },
+    {
+      "epoch": 122.0,
+      "eval_accuracy": 0.7519685039370079,
+      "eval_loss": 1.0612245798110962,
+      "eval_runtime": 5.2527,
+      "eval_samples_per_second": 48.356,
+      "eval_steps_per_second": 0.762,
+      "step": 2806
+    },
+    {
+      "epoch": 122.17391304347827,
+      "grad_norm": 12.711108207702637,
+      "learning_rate": 7.782608695652174e-06,
+      "loss": 0.1415,
+      "step": 2810
+    },
+    {
+      "epoch": 122.6086956521739,
+      "grad_norm": 21.42779541015625,
+      "learning_rate": 7.739130434782609e-06,
+      "loss": 0.1219,
+      "step": 2820
+    },
+    {
+      "epoch": 123.0,
+      "eval_accuracy": 0.7519685039370079,
+      "eval_loss": 1.1412242650985718,
+      "eval_runtime": 5.1306,
+      "eval_samples_per_second": 49.507,
+      "eval_steps_per_second": 0.78,
+      "step": 2829
+    },
+    {
+      "epoch": 123.04347826086956,
+      "grad_norm": 23.719886779785156,
+      "learning_rate": 7.695652173913044e-06,
+      "loss": 0.138,
+      "step": 2830
+    },
+    {
+      "epoch": 123.47826086956522,
+      "grad_norm": 9.191445350646973,
+      "learning_rate": 7.652173913043479e-06,
+      "loss": 0.111,
+      "step": 2840
+    },
+    {
+      "epoch": 123.91304347826087,
+      "grad_norm": 29.234426498413086,
+      "learning_rate": 7.608695652173914e-06,
+      "loss": 0.148,
+      "step": 2850
+    },
+    {
+      "epoch": 124.0,
+      "eval_accuracy": 0.7401574803149606,
+      "eval_loss": 1.0836260318756104,
+      "eval_runtime": 5.2645,
+      "eval_samples_per_second": 48.248,
+      "eval_steps_per_second": 0.76,
+      "step": 2852
+    },
+    {
+      "epoch": 124.34782608695652,
+      "grad_norm": 11.18213939666748,
+      "learning_rate": 7.565217391304348e-06,
+      "loss": 0.13,
+      "step": 2860
+    },
+    {
+      "epoch": 124.78260869565217,
+      "grad_norm": 15.413270950317383,
+      "learning_rate": 7.5217391304347835e-06,
+      "loss": 0.1076,
+      "step": 2870
+    },
+    {
+      "epoch": 125.0,
+      "eval_accuracy": 0.7559055118110236,
+      "eval_loss": 1.0629363059997559,
+      "eval_runtime": 5.8955,
+      "eval_samples_per_second": 43.084,
+      "eval_steps_per_second": 0.678,
+      "step": 2875
+    },
+    {
+      "epoch": 125.21739130434783,
+      "grad_norm": 8.161211013793945,
+      "learning_rate": 7.478260869565218e-06,
+      "loss": 0.1678,
+      "step": 2880
+    },
+    {
+      "epoch": 125.65217391304348,
+      "grad_norm": 8.079183578491211,
+      "learning_rate": 7.434782608695653e-06,
+      "loss": 0.1306,
+      "step": 2890
+    },
+    {
+      "epoch": 126.0,
+      "eval_accuracy": 0.7362204724409449,
+      "eval_loss": 1.0791066884994507,
+      "eval_runtime": 5.794,
+      "eval_samples_per_second": 43.839,
+      "eval_steps_per_second": 0.69,
+      "step": 2898
+    },
+    {
+      "epoch": 126.08695652173913,
+      "grad_norm": 20.435319900512695,
+      "learning_rate": 7.391304347826087e-06,
+      "loss": 0.1227,
+      "step": 2900
+    },
+    {
+      "epoch": 126.52173913043478,
+      "grad_norm": 5.470503807067871,
+      "learning_rate": 7.347826086956522e-06,
+      "loss": 0.1258,
+      "step": 2910
+    },
+    {
+      "epoch": 126.95652173913044,
+      "grad_norm": 21.71735191345215,
+      "learning_rate": 7.304347826086957e-06,
+      "loss": 0.1153,
+      "step": 2920
+    },
+    {
+      "epoch": 127.0,
+      "eval_accuracy": 0.7401574803149606,
+      "eval_loss": 1.149482250213623,
+      "eval_runtime": 5.2212,
+      "eval_samples_per_second": 48.648,
+      "eval_steps_per_second": 0.766,
+      "step": 2921
+    },
+    {
+      "epoch": 127.3913043478261,
+      "grad_norm": 7.169867038726807,
+      "learning_rate": 7.2608695652173925e-06,
+      "loss": 0.1023,
+      "step": 2930
+    },
+    {
+      "epoch": 127.82608695652173,
+      "grad_norm": 5.137673377990723,
+      "learning_rate": 7.217391304347827e-06,
+      "loss": 0.1239,
+      "step": 2940
+    },
+    {
+      "epoch": 128.0,
+      "eval_accuracy": 0.7519685039370079,
+      "eval_loss": 1.1446442604064941,
+      "eval_runtime": 5.8267,
+      "eval_samples_per_second": 43.592,
+      "eval_steps_per_second": 0.686,
+      "step": 2944
+    },
+    {
+      "epoch": 128.2608695652174,
+      "grad_norm": 46.98186492919922,
+      "learning_rate": 7.173913043478261e-06,
+      "loss": 0.1256,
+      "step": 2950
+    },
+    {
+      "epoch": 128.69565217391303,
+      "grad_norm": 18.51282501220703,
+      "learning_rate": 7.130434782608696e-06,
+      "loss": 0.1533,
+      "step": 2960
+    },
+    {
+      "epoch": 129.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.0817540884017944,
+      "eval_runtime": 5.9648,
+      "eval_samples_per_second": 42.583,
+      "eval_steps_per_second": 0.671,
+      "step": 2967
+    },
+    {
+      "epoch": 129.1304347826087,
+      "grad_norm": 30.390338897705078,
+      "learning_rate": 7.086956521739131e-06,
+      "loss": 0.1534,
+      "step": 2970
+    },
+    {
+      "epoch": 129.56521739130434,
+      "grad_norm": 15.933720588684082,
+      "learning_rate": 7.0434782608695665e-06,
+      "loss": 0.1542,
+      "step": 2980
+    },
+    {
+      "epoch": 130.0,
+      "grad_norm": 28.77271270751953,
+      "learning_rate": 7e-06,
+      "loss": 0.136,
+      "step": 2990
+    },
+    {
+      "epoch": 130.0,
+      "eval_accuracy": 0.7519685039370079,
+      "eval_loss": 1.0557951927185059,
+      "eval_runtime": 5.2305,
+      "eval_samples_per_second": 48.561,
+      "eval_steps_per_second": 0.765,
+      "step": 2990
+    },
+    {
+      "epoch": 130.43478260869566,
+      "grad_norm": 13.821724891662598,
+      "learning_rate": 6.956521739130435e-06,
+      "loss": 0.1428,
+      "step": 3000
+    },
+    {
+      "epoch": 130.8695652173913,
+      "grad_norm": 12.773445129394531,
+      "learning_rate": 6.91304347826087e-06,
+      "loss": 0.1189,
+      "step": 3010
+    },
+    {
+      "epoch": 131.0,
+      "eval_accuracy": 0.7519685039370079,
+      "eval_loss": 1.0423293113708496,
+      "eval_runtime": 5.2446,
+      "eval_samples_per_second": 48.431,
+      "eval_steps_per_second": 0.763,
+      "step": 3013
+    },
+    {
+      "epoch": 131.30434782608697,
+      "grad_norm": 12.808055877685547,
+      "learning_rate": 6.869565217391305e-06,
+      "loss": 0.1204,
+      "step": 3020
+    },
+    {
+      "epoch": 131.7391304347826,
+      "grad_norm": 12.246822357177734,
+      "learning_rate": 6.8260869565217395e-06,
+      "loss": 0.1247,
+      "step": 3030
+    },
+    {
+      "epoch": 132.0,
+      "eval_accuracy": 0.7637795275590551,
+      "eval_loss": 1.0581423044204712,
+      "eval_runtime": 5.6958,
+      "eval_samples_per_second": 44.594,
+      "eval_steps_per_second": 0.702,
+      "step": 3036
+    },
+    {
+      "epoch": 132.17391304347825,
+      "grad_norm": 8.1224946975708,
+      "learning_rate": 6.782608695652174e-06,
+      "loss": 0.1259,
+      "step": 3040
+    },
+    {
+      "epoch": 132.6086956521739,
+      "grad_norm": 9.494956016540527,
+      "learning_rate": 6.739130434782609e-06,
+      "loss": 0.1136,
+      "step": 3050
+    },
+    {
+      "epoch": 133.0,
+      "eval_accuracy": 0.7716535433070866,
+      "eval_loss": 1.0132337808609009,
+      "eval_runtime": 5.341,
+      "eval_samples_per_second": 47.556,
+      "eval_steps_per_second": 0.749,
+      "step": 3059
+    },
+    {
+      "epoch": 133.04347826086956,
+      "grad_norm": 6.921084880828857,
+      "learning_rate": 6.695652173913044e-06,
+      "loss": 0.1295,
+      "step": 3060
+    },
+    {
+      "epoch": 133.47826086956522,
+      "grad_norm": 11.8038330078125,
+      "learning_rate": 6.652173913043479e-06,
+      "loss": 0.094,
+      "step": 3070
+    },
+    {
+      "epoch": 133.91304347826087,
+      "grad_norm": 14.180817604064941,
+      "learning_rate": 6.6086956521739135e-06,
+      "loss": 0.1492,
+      "step": 3080
+    },
+    {
+      "epoch": 134.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.1127182245254517,
+      "eval_runtime": 5.2916,
+      "eval_samples_per_second": 48.0,
+      "eval_steps_per_second": 0.756,
+      "step": 3082
+    },
+    {
+      "epoch": 134.34782608695653,
+      "grad_norm": 35.78520202636719,
+      "learning_rate": 6.565217391304349e-06,
+      "loss": 0.1522,
+      "step": 3090
+    },
+    {
+      "epoch": 134.7826086956522,
+      "grad_norm": 16.837764739990234,
+      "learning_rate": 6.521739130434783e-06,
+      "loss": 0.1184,
+      "step": 3100
+    },
+    {
+      "epoch": 135.0,
+      "eval_accuracy": 0.7401574803149606,
+      "eval_loss": 1.1449700593948364,
+      "eval_runtime": 5.1179,
+      "eval_samples_per_second": 49.629,
+      "eval_steps_per_second": 0.782,
+      "step": 3105
+    },
+    {
+      "epoch": 135.2173913043478,
+      "grad_norm": 18.732372283935547,
+      "learning_rate": 6.478260869565218e-06,
+      "loss": 0.1284,
+      "step": 3110
+    },
+    {
+      "epoch": 135.65217391304347,
+      "grad_norm": 47.60773468017578,
+      "learning_rate": 6.434782608695652e-06,
+      "loss": 0.1122,
+      "step": 3120
+    },
+    {
+      "epoch": 136.0,
+      "eval_accuracy": 0.7519685039370079,
+      "eval_loss": 1.1063224077224731,
+      "eval_runtime": 5.1054,
+      "eval_samples_per_second": 49.751,
+      "eval_steps_per_second": 0.783,
+      "step": 3128
+    },
+    {
+      "epoch": 136.08695652173913,
+      "grad_norm": 13.677179336547852,
+      "learning_rate": 6.391304347826087e-06,
+      "loss": 0.1027,
+      "step": 3130
+    },
+    {
+      "epoch": 136.52173913043478,
+      "grad_norm": 25.119609832763672,
+      "learning_rate": 6.3478260869565225e-06,
+      "loss": 0.1485,
+      "step": 3140
+    },
+    {
+      "epoch": 136.95652173913044,
+      "grad_norm": 7.572585105895996,
+      "learning_rate": 6.304347826086958e-06,
+      "loss": 0.1047,
+      "step": 3150
+    },
+    {
+      "epoch": 137.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.1029216051101685,
+      "eval_runtime": 5.1837,
+      "eval_samples_per_second": 48.999,
+      "eval_steps_per_second": 0.772,
+      "step": 3151
+    },
+    {
+      "epoch": 137.3913043478261,
+      "grad_norm": 15.767362594604492,
+      "learning_rate": 6.260869565217392e-06,
+      "loss": 0.1319,
+      "step": 3160
+    },
+    {
+      "epoch": 137.82608695652175,
+      "grad_norm": 7.367244243621826,
+      "learning_rate": 6.217391304347826e-06,
+      "loss": 0.1285,
+      "step": 3170
+    },
+    {
+      "epoch": 138.0,
+      "eval_accuracy": 0.7401574803149606,
+      "eval_loss": 1.156273603439331,
+      "eval_runtime": 6.0904,
+      "eval_samples_per_second": 41.705,
+      "eval_steps_per_second": 0.657,
+      "step": 3174
+    },
+    {
+      "epoch": 138.2608695652174,
+      "grad_norm": 17.1787166595459,
+      "learning_rate": 6.173913043478261e-06,
+      "loss": 0.1134,
+      "step": 3180
+    },
+    {
+      "epoch": 138.69565217391303,
+      "grad_norm": 8.421507835388184,
+      "learning_rate": 6.1304347826086965e-06,
+      "loss": 0.1004,
+      "step": 3190
+    },
+    {
+      "epoch": 139.0,
+      "eval_accuracy": 0.7362204724409449,
+      "eval_loss": 1.1551874876022339,
+      "eval_runtime": 5.6544,
+      "eval_samples_per_second": 44.921,
+      "eval_steps_per_second": 0.707,
+      "step": 3197
+    },
+    {
+      "epoch": 139.1304347826087,
+      "grad_norm": 13.920239448547363,
+      "learning_rate": 6.086956521739132e-06,
+      "loss": 0.1237,
+      "step": 3200
+    },
+    {
+      "epoch": 139.56521739130434,
+      "grad_norm": 11.163459777832031,
+      "learning_rate": 6.043478260869565e-06,
+      "loss": 0.1278,
+      "step": 3210
+    },
+    {
+      "epoch": 140.0,
+      "grad_norm": 38.89981460571289,
+      "learning_rate": 6e-06,
+      "loss": 0.1285,
+      "step": 3220
+    },
+    {
+      "epoch": 140.0,
+      "eval_accuracy": 0.7480314960629921,
+      "eval_loss": 1.109745979309082,
+      "eval_runtime": 5.999,
+      "eval_samples_per_second": 42.34,
+      "eval_steps_per_second": 0.667,
+      "step": 3220
+    },
+    {
+      "epoch": 140.43478260869566,
+      "grad_norm": 7.645209312438965,
+      "learning_rate": 5.956521739130435e-06,
+      "loss": 0.1422,
+      "step": 3230
+    },
+    {
+      "epoch": 140.8695652173913,
+      "grad_norm": 20.615463256835938,
+      "learning_rate": 5.91304347826087e-06,
+      "loss": 0.1257,
+      "step": 3240
+    },
+    {
+      "epoch": 141.0,
+      "eval_accuracy": 0.7401574803149606,
+      "eval_loss": 1.1602002382278442,
+      "eval_runtime": 6.0238,
+      "eval_samples_per_second": 42.166,
+      "eval_steps_per_second": 0.664,
+      "step": 3243
+    },
+    {
+      "epoch": 141.30434782608697,
+      "grad_norm": 13.762594223022461,
+      "learning_rate": 5.8695652173913055e-06,
+      "loss": 0.0992,
+      "step": 3250
+    },
+    {
+      "epoch": 141.7391304347826,
+      "grad_norm": 5.704482555389404,
+      "learning_rate": 5.826086956521739e-06,
+      "loss": 0.1075,
+      "step": 3260
+    },
+    {
+      "epoch": 142.0,
+      "eval_accuracy": 0.7559055118110236,
+      "eval_loss": 1.1911606788635254,
+      "eval_runtime": 6.0583,
+      "eval_samples_per_second": 41.926,
+      "eval_steps_per_second": 0.66,
+      "step": 3266
+    },
+    {
+      "epoch": 142.17391304347825,
+      "grad_norm": 18.972148895263672,
+      "learning_rate": 5.782608695652174e-06,
+      "loss": 0.1112,
+      "step": 3270
+    },
+    {
+      "epoch": 142.6086956521739,
+      "grad_norm": 21.352266311645508,
+      "learning_rate": 5.739130434782609e-06,
+      "loss": 0.1098,
+      "step": 3280
+    },
+    {
+      "epoch": 143.0,
+      "eval_accuracy": 0.7519685039370079,
+      "eval_loss": 1.1894166469573975,
+      "eval_runtime": 5.6001,
+      "eval_samples_per_second": 45.357,
+      "eval_steps_per_second": 0.714,
+      "step": 3289
+    },
+    {
+      "epoch": 143.04347826086956,
+      "grad_norm": 12.431015014648438,
+      "learning_rate": 5.695652173913044e-06,
+      "loss": 0.0992,
+      "step": 3290
+    },
+    {
+      "epoch": 143.47826086956522,
+      "grad_norm": 17.39335823059082,
+      "learning_rate": 5.652173913043479e-06,
+      "loss": 0.1301,
+      "step": 3300
+    },
+    {
+      "epoch": 143.91304347826087,
+      "grad_norm": 3.1455869674682617,
+      "learning_rate": 5.608695652173914e-06,
+      "loss": 0.1148,
+      "step": 3310
+    },
+    {
+      "epoch": 144.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.155055046081543,
+      "eval_runtime": 6.0411,
+      "eval_samples_per_second": 42.045,
+      "eval_steps_per_second": 0.662,
+      "step": 3312
+    },
+    {
+      "epoch": 144.34782608695653,
+      "grad_norm": 19.077762603759766,
+      "learning_rate": 5.565217391304348e-06,
+      "loss": 0.1051,
+      "step": 3320
+    },
+    {
+      "epoch": 144.7826086956522,
+      "grad_norm": 22.914405822753906,
+      "learning_rate": 5.521739130434783e-06,
+      "loss": 0.1489,
+      "step": 3330
+    },
+    {
+      "epoch": 145.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.137885332107544,
+      "eval_runtime": 5.3417,
+      "eval_samples_per_second": 47.55,
+      "eval_steps_per_second": 0.749,
+      "step": 3335
+    },
+    {
+      "epoch": 145.2173913043478,
+      "grad_norm": 33.728397369384766,
+      "learning_rate": 5.478260869565217e-06,
+      "loss": 0.1242,
+      "step": 3340
+    },
+    {
+      "epoch": 145.65217391304347,
+      "grad_norm": 22.441125869750977,
+      "learning_rate": 5.4347826086956525e-06,
+      "loss": 0.1461,
+      "step": 3350
+    },
+    {
+      "epoch": 146.0,
+      "eval_accuracy": 0.7480314960629921,
+      "eval_loss": 1.1726444959640503,
+      "eval_runtime": 5.1056,
+      "eval_samples_per_second": 49.749,
+      "eval_steps_per_second": 0.783,
+      "step": 3358
+    },
+    {
+      "epoch": 146.08695652173913,
+      "grad_norm": 23.851354598999023,
+      "learning_rate": 5.391304347826088e-06,
+      "loss": 0.1157,
+      "step": 3360
+    },
+    {
+      "epoch": 146.52173913043478,
+      "grad_norm": 11.18212604522705,
+      "learning_rate": 5.347826086956523e-06,
+      "loss": 0.1238,
+      "step": 3370
+    },
+    {
+      "epoch": 146.95652173913044,
+      "grad_norm": 10.293829917907715,
+      "learning_rate": 5.304347826086957e-06,
+      "loss": 0.1171,
+      "step": 3380
+    },
+    {
+      "epoch": 147.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.1190757751464844,
+      "eval_runtime": 5.1299,
+      "eval_samples_per_second": 49.513,
+      "eval_steps_per_second": 0.78,
+      "step": 3381
+    },
+    {
+      "epoch": 147.3913043478261,
+      "grad_norm": 12.49338436126709,
+      "learning_rate": 5.260869565217391e-06,
+      "loss": 0.1363,
+      "step": 3390
+    },
+    {
+      "epoch": 147.82608695652175,
+      "grad_norm": 16.56803321838379,
+      "learning_rate": 5.2173913043478265e-06,
+      "loss": 0.1262,
+      "step": 3400
+    },
+    {
+      "epoch": 148.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.1661558151245117,
+      "eval_runtime": 5.1432,
+      "eval_samples_per_second": 49.386,
+      "eval_steps_per_second": 0.778,
+      "step": 3404
+    },
+    {
+      "epoch": 148.2608695652174,
+      "grad_norm": 13.47645092010498,
+      "learning_rate": 5.173913043478262e-06,
+      "loss": 0.1236,
+      "step": 3410
+    },
+    {
+      "epoch": 148.69565217391303,
+      "grad_norm": 6.461868762969971,
+      "learning_rate": 5.130434782608697e-06,
+      "loss": 0.1137,
+      "step": 3420
+    },
+    {
+      "epoch": 149.0,
+      "eval_accuracy": 0.7480314960629921,
+      "eval_loss": 1.1282668113708496,
+      "eval_runtime": 5.2173,
+      "eval_samples_per_second": 48.684,
+      "eval_steps_per_second": 0.767,
+      "step": 3427
+    },
+    {
+      "epoch": 149.1304347826087,
+      "grad_norm": 9.386457443237305,
+      "learning_rate": 5.08695652173913e-06,
+      "loss": 0.0944,
+      "step": 3430
+    },
+    {
+      "epoch": 149.56521739130434,
+      "grad_norm": 17.175273895263672,
+      "learning_rate": 5.043478260869565e-06,
+      "loss": 0.1179,
+      "step": 3440
+    },
+    {
+      "epoch": 150.0,
+      "grad_norm": 3.21197247505188,
+      "learning_rate": 5e-06,
+      "loss": 0.1118,
+      "step": 3450
+    },
+    {
+      "epoch": 150.0,
+      "eval_accuracy": 0.7480314960629921,
+      "eval_loss": 1.1388078927993774,
+      "eval_runtime": 5.2199,
+      "eval_samples_per_second": 48.66,
+      "eval_steps_per_second": 0.766,
+      "step": 3450
+    },
+    {
+      "epoch": 150.43478260869566,
+      "grad_norm": 13.216156005859375,
+      "learning_rate": 4.9565217391304355e-06,
+      "loss": 0.132,
+      "step": 3460
+    },
+    {
+      "epoch": 150.8695652173913,
+      "grad_norm": 17.5484561920166,
+      "learning_rate": 4.91304347826087e-06,
+      "loss": 0.1169,
+      "step": 3470
+    },
+    {
+      "epoch": 151.0,
+      "eval_accuracy": 0.7519685039370079,
+      "eval_loss": 1.162711501121521,
+      "eval_runtime": 6.0191,
+      "eval_samples_per_second": 42.199,
+      "eval_steps_per_second": 0.665,
+      "step": 3473
+    },
+    {
+      "epoch": 151.30434782608697,
+      "grad_norm": 12.96456527709961,
+      "learning_rate": 4.869565217391305e-06,
+      "loss": 0.0991,
+      "step": 3480
+    },
+    {
+      "epoch": 151.7391304347826,
+      "grad_norm": 16.27593994140625,
+      "learning_rate": 4.826086956521739e-06,
+      "loss": 0.1021,
+      "step": 3490
+    },
+    {
+      "epoch": 152.0,
+      "eval_accuracy": 0.7322834645669292,
+      "eval_loss": 1.182125210762024,
+      "eval_runtime": 6.074,
+      "eval_samples_per_second": 41.817,
+      "eval_steps_per_second": 0.659,
+      "step": 3496
+    },
+    {
+      "epoch": 152.17391304347825,
+      "grad_norm": 17.580163955688477,
+      "learning_rate": 4.782608695652174e-06,
+      "loss": 0.1196,
+      "step": 3500
+    },
+    {
+      "epoch": 152.6086956521739,
+      "grad_norm": 16.174192428588867,
+      "learning_rate": 4.739130434782609e-06,
+      "loss": 0.1392,
+      "step": 3510
+    },
+    {
+      "epoch": 153.0,
+      "eval_accuracy": 0.7322834645669292,
+      "eval_loss": 1.1671814918518066,
+      "eval_runtime": 5.2176,
+      "eval_samples_per_second": 48.681,
+      "eval_steps_per_second": 0.767,
+      "step": 3519
+    },
+    {
+      "epoch": 153.04347826086956,
+      "grad_norm": 34.79978561401367,
+      "learning_rate": 4.695652173913044e-06,
+      "loss": 0.1162,
+      "step": 3520
+    },
+    {
+      "epoch": 153.47826086956522,
+      "grad_norm": 17.405305862426758,
+      "learning_rate": 4.652173913043478e-06,
+      "loss": 0.1022,
+      "step": 3530
+    },
+    {
+      "epoch": 153.91304347826087,
+      "grad_norm": 10.909713745117188,
+      "learning_rate": 4.608695652173913e-06,
+      "loss": 0.1111,
+      "step": 3540
+    },
+    {
+      "epoch": 154.0,
+      "eval_accuracy": 0.7401574803149606,
+      "eval_loss": 1.2136441469192505,
+      "eval_runtime": 5.568,
+      "eval_samples_per_second": 45.618,
+      "eval_steps_per_second": 0.718,
+      "step": 3542
+    },
+    {
+      "epoch": 154.34782608695653,
+      "grad_norm": 21.353759765625,
+      "learning_rate": 4.565217391304348e-06,
+      "loss": 0.143,
+      "step": 3550
+    },
+    {
+      "epoch": 154.7826086956522,
+      "grad_norm": 17.216047286987305,
+      "learning_rate": 4.5217391304347826e-06,
+      "loss": 0.1298,
+      "step": 3560
+    },
+    {
+      "epoch": 155.0,
+      "eval_accuracy": 0.7401574803149606,
+      "eval_loss": 1.1966124773025513,
+      "eval_runtime": 6.0928,
+      "eval_samples_per_second": 41.689,
+      "eval_steps_per_second": 0.657,
+      "step": 3565
+    },
+    {
+      "epoch": 155.2173913043478,
+      "grad_norm": 8.084007263183594,
+      "learning_rate": 4.478260869565218e-06,
+      "loss": 0.0944,
+      "step": 3570
+    },
+    {
+      "epoch": 155.65217391304347,
+      "grad_norm": 13.08774471282959,
+      "learning_rate": 4.434782608695653e-06,
+      "loss": 0.1114,
+      "step": 3580
+    },
+    {
+      "epoch": 156.0,
+      "eval_accuracy": 0.7362204724409449,
+      "eval_loss": 1.138235092163086,
+      "eval_runtime": 5.2162,
+      "eval_samples_per_second": 48.694,
+      "eval_steps_per_second": 0.767,
+      "step": 3588
+    },
+    {
+      "epoch": 156.08695652173913,
+      "grad_norm": 9.64503002166748,
+      "learning_rate": 4.391304347826087e-06,
+      "loss": 0.1216,
+      "step": 3590
+    },
+    {
+      "epoch": 156.52173913043478,
+      "grad_norm": 20.011825561523438,
+      "learning_rate": 4.347826086956522e-06,
+      "loss": 0.1352,
+      "step": 3600
+    },
+    {
+      "epoch": 156.95652173913044,
+      "grad_norm": 9.31452751159668,
+      "learning_rate": 4.304347826086957e-06,
+      "loss": 0.09,
+      "step": 3610
+    },
+    {
+      "epoch": 157.0,
+      "eval_accuracy": 0.7322834645669292,
+      "eval_loss": 1.145975947380066,
+      "eval_runtime": 6.2926,
+      "eval_samples_per_second": 40.365,
+      "eval_steps_per_second": 0.636,
+      "step": 3611
+    },
+    {
+      "epoch": 157.3913043478261,
+      "grad_norm": 22.55306625366211,
+      "learning_rate": 4.260869565217392e-06,
+      "loss": 0.0952,
+      "step": 3620
+    },
+    {
+      "epoch": 157.82608695652175,
+      "grad_norm": 11.786699295043945,
+      "learning_rate": 4.217391304347827e-06,
+      "loss": 0.1294,
+      "step": 3630
+    },
+    {
+      "epoch": 158.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.1612186431884766,
+      "eval_runtime": 5.5409,
+      "eval_samples_per_second": 45.841,
+      "eval_steps_per_second": 0.722,
+      "step": 3634
+    },
+    {
+      "epoch": 158.2608695652174,
+      "grad_norm": 10.745357513427734,
+      "learning_rate": 4.173913043478261e-06,
+      "loss": 0.0985,
+      "step": 3640
+    },
+    {
+      "epoch": 158.69565217391303,
+      "grad_norm": 12.405003547668457,
+      "learning_rate": 4.130434782608696e-06,
+      "loss": 0.1186,
+      "step": 3650
+    },
+    {
+      "epoch": 159.0,
+      "eval_accuracy": 0.7401574803149606,
+      "eval_loss": 1.2204416990280151,
+      "eval_runtime": 5.9711,
+      "eval_samples_per_second": 42.538,
+      "eval_steps_per_second": 0.67,
+      "step": 3657
+    },
+    {
+      "epoch": 159.1304347826087,
+      "grad_norm": 20.51667022705078,
+      "learning_rate": 4.086956521739131e-06,
+      "loss": 0.1335,
+      "step": 3660
+    },
+    {
+      "epoch": 159.56521739130434,
+      "grad_norm": 12.348188400268555,
+      "learning_rate": 4.0434782608695655e-06,
+      "loss": 0.1106,
+      "step": 3670
+    },
+    {
+      "epoch": 160.0,
+      "grad_norm": 24.83527183532715,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.1096,
+      "step": 3680
+    },
+    {
+      "epoch": 160.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.2095911502838135,
+      "eval_runtime": 6.1017,
+      "eval_samples_per_second": 41.628,
+      "eval_steps_per_second": 0.656,
+      "step": 3680
+    },
+    {
+      "epoch": 160.43478260869566,
+      "grad_norm": 21.34354019165039,
+      "learning_rate": 3.956521739130435e-06,
+      "loss": 0.0921,
+      "step": 3690
+    },
+    {
+      "epoch": 160.8695652173913,
+      "grad_norm": 9.070474624633789,
+      "learning_rate": 3.91304347826087e-06,
+      "loss": 0.1107,
+      "step": 3700
+    },
+    {
+      "epoch": 161.0,
+      "eval_accuracy": 0.7480314960629921,
+      "eval_loss": 1.182215690612793,
+      "eval_runtime": 6.0899,
+      "eval_samples_per_second": 41.709,
+      "eval_steps_per_second": 0.657,
+      "step": 3703
+    },
+    {
+      "epoch": 161.30434782608697,
+      "grad_norm": 17.007827758789062,
+      "learning_rate": 3.869565217391304e-06,
+      "loss": 0.1148,
+      "step": 3710
+    },
+    {
+      "epoch": 161.7391304347826,
+      "grad_norm": 15.50957202911377,
+      "learning_rate": 3.8260869565217395e-06,
+      "loss": 0.1094,
+      "step": 3720
+    },
+    {
+      "epoch": 162.0,
+      "eval_accuracy": 0.7480314960629921,
+      "eval_loss": 1.1907768249511719,
+      "eval_runtime": 6.0337,
+      "eval_samples_per_second": 42.097,
+      "eval_steps_per_second": 0.663,
+      "step": 3726
+    },
+    {
+      "epoch": 162.17391304347825,
+      "grad_norm": 22.703414916992188,
+      "learning_rate": 3.782608695652174e-06,
+      "loss": 0.1123,
+      "step": 3730
+    },
+    {
+      "epoch": 162.6086956521739,
+      "grad_norm": 11.196767807006836,
+      "learning_rate": 3.739130434782609e-06,
+      "loss": 0.1112,
+      "step": 3740
+    },
+    {
+      "epoch": 163.0,
+      "eval_accuracy": 0.7401574803149606,
+      "eval_loss": 1.164740800857544,
+      "eval_runtime": 5.9944,
+      "eval_samples_per_second": 42.373,
+      "eval_steps_per_second": 0.667,
+      "step": 3749
+    },
+    {
+      "epoch": 163.04347826086956,
+      "grad_norm": 8.430397987365723,
+      "learning_rate": 3.6956521739130436e-06,
+      "loss": 0.1113,
+      "step": 3750
+    },
+    {
+      "epoch": 163.47826086956522,
+      "grad_norm": 4.709465980529785,
+      "learning_rate": 3.6521739130434787e-06,
+      "loss": 0.1078,
+      "step": 3760
+    },
+    {
+      "epoch": 163.91304347826087,
+      "grad_norm": 13.146780014038086,
+      "learning_rate": 3.6086956521739134e-06,
+      "loss": 0.1042,
+      "step": 3770
+    },
+    {
+      "epoch": 164.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.252306342124939,
+      "eval_runtime": 5.2297,
+      "eval_samples_per_second": 48.569,
+      "eval_steps_per_second": 0.765,
+      "step": 3772
+    },
+    {
+      "epoch": 164.34782608695653,
+      "grad_norm": 10.450764656066895,
+      "learning_rate": 3.565217391304348e-06,
+      "loss": 0.1549,
+      "step": 3780
+    },
+    {
+      "epoch": 164.7826086956522,
+      "grad_norm": 3.6140904426574707,
+      "learning_rate": 3.5217391304347832e-06,
+      "loss": 0.0993,
+      "step": 3790
+    },
+    {
+      "epoch": 165.0,
+      "eval_accuracy": 0.7401574803149606,
+      "eval_loss": 1.2039833068847656,
+      "eval_runtime": 6.3786,
+      "eval_samples_per_second": 39.821,
+      "eval_steps_per_second": 0.627,
+      "step": 3795
+    },
+    {
+      "epoch": 165.2173913043478,
+      "grad_norm": 12.08646297454834,
+      "learning_rate": 3.4782608695652175e-06,
+      "loss": 0.1,
+      "step": 3800
+    },
+    {
+      "epoch": 165.65217391304347,
+      "grad_norm": 19.647462844848633,
+      "learning_rate": 3.4347826086956526e-06,
+      "loss": 0.105,
+      "step": 3810
+    },
+    {
+      "epoch": 166.0,
+      "eval_accuracy": 0.7401574803149606,
+      "eval_loss": 1.2296301126480103,
+      "eval_runtime": 6.2193,
+      "eval_samples_per_second": 40.84,
+      "eval_steps_per_second": 0.643,
+      "step": 3818
+    },
+    {
+      "epoch": 166.08695652173913,
+      "grad_norm": 14.596138000488281,
+      "learning_rate": 3.391304347826087e-06,
+      "loss": 0.0947,
+      "step": 3820
+    },
+    {
+      "epoch": 166.52173913043478,
+      "grad_norm": 12.743565559387207,
+      "learning_rate": 3.347826086956522e-06,
+      "loss": 0.1144,
+      "step": 3830
+    },
+    {
+      "epoch": 166.95652173913044,
+      "grad_norm": 2.257753610610962,
+      "learning_rate": 3.3043478260869567e-06,
+      "loss": 0.1071,
+      "step": 3840
+    },
+    {
+      "epoch": 167.0,
+      "eval_accuracy": 0.7480314960629921,
+      "eval_loss": 1.286328673362732,
+      "eval_runtime": 6.2909,
+      "eval_samples_per_second": 40.376,
+      "eval_steps_per_second": 0.636,
+      "step": 3841
+    },
+    {
+      "epoch": 167.3913043478261,
+      "grad_norm": 9.273666381835938,
+      "learning_rate": 3.2608695652173914e-06,
+      "loss": 0.11,
+      "step": 3850
+    },
+    {
+      "epoch": 167.82608695652175,
+      "grad_norm": 9.474180221557617,
+      "learning_rate": 3.217391304347826e-06,
+      "loss": 0.108,
+      "step": 3860
+    },
+    {
+      "epoch": 168.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.2372475862503052,
+      "eval_runtime": 6.1145,
+      "eval_samples_per_second": 41.541,
+      "eval_steps_per_second": 0.654,
+      "step": 3864
+    },
+    {
+      "epoch": 168.2608695652174,
+      "grad_norm": 9.455063819885254,
+      "learning_rate": 3.1739130434782613e-06,
+      "loss": 0.0999,
+      "step": 3870
+    },
+    {
+      "epoch": 168.69565217391303,
+      "grad_norm": 15.079169273376465,
+      "learning_rate": 3.130434782608696e-06,
+      "loss": 0.1076,
+      "step": 3880
+    },
+    {
+      "epoch": 169.0,
+      "eval_accuracy": 0.7480314960629921,
+      "eval_loss": 1.1871860027313232,
+      "eval_runtime": 5.0604,
+      "eval_samples_per_second": 50.193,
+      "eval_steps_per_second": 0.79,
+      "step": 3887
+    },
+    {
+      "epoch": 169.1304347826087,
+      "grad_norm": 12.766210556030273,
+      "learning_rate": 3.0869565217391307e-06,
+      "loss": 0.1003,
+      "step": 3890
+    },
+    {
+      "epoch": 169.56521739130434,
+      "grad_norm": 7.451639175415039,
+      "learning_rate": 3.043478260869566e-06,
+      "loss": 0.1141,
+      "step": 3900
+    },
+    {
+      "epoch": 170.0,
+      "grad_norm": 23.210697174072266,
+      "learning_rate": 3e-06,
+      "loss": 0.1107,
+      "step": 3910
+    },
+    {
+      "epoch": 170.0,
+      "eval_accuracy": 0.7322834645669292,
+      "eval_loss": 1.2354222536087036,
+      "eval_runtime": 6.0785,
+      "eval_samples_per_second": 41.787,
+      "eval_steps_per_second": 0.658,
+      "step": 3910
+    },
+    {
+      "epoch": 170.43478260869566,
+      "grad_norm": 13.231768608093262,
+      "learning_rate": 2.956521739130435e-06,
+      "loss": 0.1264,
+      "step": 3920
+    },
+    {
+      "epoch": 170.8695652173913,
+      "grad_norm": 8.206528663635254,
+      "learning_rate": 2.9130434782608695e-06,
+      "loss": 0.1012,
+      "step": 3930
+    },
+    {
+      "epoch": 171.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.2104681730270386,
+      "eval_runtime": 6.0246,
+      "eval_samples_per_second": 42.161,
+      "eval_steps_per_second": 0.664,
+      "step": 3933
+    },
+    {
+      "epoch": 171.30434782608697,
+      "grad_norm": 5.853540420532227,
+      "learning_rate": 2.8695652173913046e-06,
+      "loss": 0.0994,
+      "step": 3940
+    },
+    {
+      "epoch": 171.7391304347826,
+      "grad_norm": 23.807804107666016,
+      "learning_rate": 2.8260869565217393e-06,
+      "loss": 0.0918,
+      "step": 3950
+    },
+    {
+      "epoch": 172.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.2026124000549316,
+      "eval_runtime": 6.0268,
+      "eval_samples_per_second": 42.145,
+      "eval_steps_per_second": 0.664,
+      "step": 3956
+    },
+    {
+      "epoch": 172.17391304347825,
+      "grad_norm": 21.178815841674805,
+      "learning_rate": 2.782608695652174e-06,
+      "loss": 0.1558,
+      "step": 3960
+    },
+    {
+      "epoch": 172.6086956521739,
+      "grad_norm": 16.011613845825195,
+      "learning_rate": 2.7391304347826087e-06,
+      "loss": 0.1043,
+      "step": 3970
+    },
+    {
+      "epoch": 173.0,
+      "eval_accuracy": 0.7559055118110236,
+      "eval_loss": 1.2925167083740234,
+      "eval_runtime": 6.0458,
+      "eval_samples_per_second": 42.012,
+      "eval_steps_per_second": 0.662,
+      "step": 3979
+    },
+    {
+      "epoch": 173.04347826086956,
+      "grad_norm": 11.591164588928223,
+      "learning_rate": 2.695652173913044e-06,
+      "loss": 0.1244,
+      "step": 3980
+    },
+    {
+      "epoch": 173.47826086956522,
+      "grad_norm": 4.560230255126953,
+      "learning_rate": 2.6521739130434785e-06,
+      "loss": 0.1179,
+      "step": 3990
+    },
+    {
+      "epoch": 173.91304347826087,
+      "grad_norm": 4.175049304962158,
+      "learning_rate": 2.6086956521739132e-06,
+      "loss": 0.1035,
+      "step": 4000
+    },
+    {
+      "epoch": 174.0,
+      "eval_accuracy": 0.7401574803149606,
+      "eval_loss": 1.2313941717147827,
+      "eval_runtime": 5.123,
+      "eval_samples_per_second": 49.58,
+      "eval_steps_per_second": 0.781,
+      "step": 4002
+    },
+    {
+      "epoch": 174.34782608695653,
+      "grad_norm": 23.1351375579834,
+      "learning_rate": 2.5652173913043484e-06,
+      "loss": 0.1087,
+      "step": 4010
+    },
+    {
+      "epoch": 174.7826086956522,
+      "grad_norm": 16.3070068359375,
+      "learning_rate": 2.5217391304347826e-06,
+      "loss": 0.1101,
+      "step": 4020
+    },
+    {
+      "epoch": 175.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.194298267364502,
+      "eval_runtime": 5.4034,
+      "eval_samples_per_second": 47.008,
+      "eval_steps_per_second": 0.74,
+      "step": 4025
+    },
+    {
+      "epoch": 175.2173913043478,
+      "grad_norm": 24.940147399902344,
+      "learning_rate": 2.4782608695652178e-06,
+      "loss": 0.1229,
+      "step": 4030
+    },
+    {
+      "epoch": 175.65217391304347,
+      "grad_norm": 12.233097076416016,
+      "learning_rate": 2.4347826086956525e-06,
+      "loss": 0.1084,
+      "step": 4040
+    },
+    {
+      "epoch": 176.0,
+      "eval_accuracy": 0.7362204724409449,
+      "eval_loss": 1.2069385051727295,
+      "eval_runtime": 5.5007,
+      "eval_samples_per_second": 46.176,
+      "eval_steps_per_second": 0.727,
+      "step": 4048
+    },
+    {
+      "epoch": 176.08695652173913,
+      "grad_norm": 2.12101149559021,
+      "learning_rate": 2.391304347826087e-06,
+      "loss": 0.1072,
+      "step": 4050
+    },
+    {
+      "epoch": 176.52173913043478,
+      "grad_norm": 6.768868446350098,
+      "learning_rate": 2.347826086956522e-06,
+      "loss": 0.1174,
+      "step": 4060
+    },
+    {
+      "epoch": 176.95652173913044,
+      "grad_norm": 15.918899536132812,
+      "learning_rate": 2.3043478260869566e-06,
+      "loss": 0.1247,
+      "step": 4070
+    },
+    {
+      "epoch": 177.0,
+      "eval_accuracy": 0.7519685039370079,
+      "eval_loss": 1.2303253412246704,
+      "eval_runtime": 5.7,
+      "eval_samples_per_second": 44.561,
+      "eval_steps_per_second": 0.702,
+      "step": 4071
+    },
+    {
+      "epoch": 177.3913043478261,
+      "grad_norm": 19.0693416595459,
+      "learning_rate": 2.2608695652173913e-06,
+      "loss": 0.1032,
+      "step": 4080
+    },
+    {
+      "epoch": 177.82608695652175,
+      "grad_norm": 21.971101760864258,
+      "learning_rate": 2.2173913043478264e-06,
+      "loss": 0.1278,
+      "step": 4090
+    },
+    {
+      "epoch": 178.0,
+      "eval_accuracy": 0.7480314960629921,
+      "eval_loss": 1.2118239402770996,
+      "eval_runtime": 5.8876,
+      "eval_samples_per_second": 43.142,
+      "eval_steps_per_second": 0.679,
+      "step": 4094
+    },
+    {
+      "epoch": 178.2608695652174,
+      "grad_norm": 9.593711853027344,
+      "learning_rate": 2.173913043478261e-06,
+      "loss": 0.1073,
+      "step": 4100
+    },
+    {
+      "epoch": 178.69565217391303,
+      "grad_norm": 17.482452392578125,
+      "learning_rate": 2.130434782608696e-06,
+      "loss": 0.1117,
+      "step": 4110
+    },
+    {
+      "epoch": 179.0,
+      "eval_accuracy": 0.7480314960629921,
+      "eval_loss": 1.2212954759597778,
+      "eval_runtime": 6.0371,
+      "eval_samples_per_second": 42.073,
+      "eval_steps_per_second": 0.663,
+      "step": 4117
+    },
+    {
+      "epoch": 179.1304347826087,
+      "grad_norm": 11.529203414916992,
+      "learning_rate": 2.0869565217391305e-06,
+      "loss": 0.0907,
+      "step": 4120
+    },
+    {
+      "epoch": 179.56521739130434,
+      "grad_norm": 14.801724433898926,
+      "learning_rate": 2.0434782608695656e-06,
+      "loss": 0.1028,
+      "step": 4130
+    },
+    {
+      "epoch": 180.0,
+      "grad_norm": 12.524520874023438,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 0.1123,
+      "step": 4140
+    },
+    {
+      "epoch": 180.0,
+      "eval_accuracy": 0.7480314960629921,
+      "eval_loss": 1.2403359413146973,
+      "eval_runtime": 5.0925,
+      "eval_samples_per_second": 49.877,
+      "eval_steps_per_second": 0.785,
+      "step": 4140
+    },
+    {
+      "epoch": 180.43478260869566,
+      "grad_norm": 16.31502914428711,
+      "learning_rate": 1.956521739130435e-06,
+      "loss": 0.1072,
+      "step": 4150
+    },
+    {
+      "epoch": 180.8695652173913,
+      "grad_norm": 10.127395629882812,
+      "learning_rate": 1.9130434782608697e-06,
+      "loss": 0.0918,
+      "step": 4160
+    },
+    {
+      "epoch": 181.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.1987248659133911,
+      "eval_runtime": 6.0905,
+      "eval_samples_per_second": 41.705,
+      "eval_steps_per_second": 0.657,
+      "step": 4163
+    },
+    {
+      "epoch": 181.30434782608697,
+      "grad_norm": 9.865519523620605,
+      "learning_rate": 1.8695652173913044e-06,
+      "loss": 0.0797,
+      "step": 4170
+    },
+    {
+      "epoch": 181.7391304347826,
+      "grad_norm": 13.656352043151855,
+      "learning_rate": 1.8260869565217394e-06,
+      "loss": 0.0827,
+      "step": 4180
+    },
+    {
+      "epoch": 182.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.2357696294784546,
+      "eval_runtime": 6.0801,
+      "eval_samples_per_second": 41.776,
+      "eval_steps_per_second": 0.658,
+      "step": 4186
+    },
+    {
+      "epoch": 182.17391304347825,
+      "grad_norm": 5.673104763031006,
+      "learning_rate": 1.782608695652174e-06,
+      "loss": 0.0948,
+      "step": 4190
+    },
+    {
+      "epoch": 182.6086956521739,
+      "grad_norm": 4.035088539123535,
+      "learning_rate": 1.7391304347826088e-06,
+      "loss": 0.0814,
+      "step": 4200
+    },
+    {
+      "epoch": 183.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.260762333869934,
+      "eval_runtime": 5.5795,
+      "eval_samples_per_second": 45.523,
+      "eval_steps_per_second": 0.717,
+      "step": 4209
+    },
+    {
+      "epoch": 183.04347826086956,
+      "grad_norm": 23.201988220214844,
+      "learning_rate": 1.6956521739130435e-06,
+      "loss": 0.1265,
+      "step": 4210
+    },
+    {
+      "epoch": 183.47826086956522,
+      "grad_norm": 16.737319946289062,
+      "learning_rate": 1.6521739130434784e-06,
+      "loss": 0.108,
+      "step": 4220
+    },
+    {
+      "epoch": 183.91304347826087,
+      "grad_norm": 10.299909591674805,
+      "learning_rate": 1.608695652173913e-06,
+      "loss": 0.0897,
+      "step": 4230
+    },
+    {
+      "epoch": 184.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.2369588613510132,
+      "eval_runtime": 5.5508,
+      "eval_samples_per_second": 45.759,
+      "eval_steps_per_second": 0.721,
+      "step": 4232
+    },
+    {
+      "epoch": 184.34782608695653,
+      "grad_norm": 7.6501641273498535,
+      "learning_rate": 1.565217391304348e-06,
+      "loss": 0.0951,
+      "step": 4240
+    },
+    {
+      "epoch": 184.7826086956522,
+      "grad_norm": 12.67705249786377,
+      "learning_rate": 1.521739130434783e-06,
+      "loss": 0.1321,
+      "step": 4250
+    },
+    {
+      "epoch": 185.0,
+      "eval_accuracy": 0.7480314960629921,
+      "eval_loss": 1.231702446937561,
+      "eval_runtime": 5.603,
+      "eval_samples_per_second": 45.333,
+      "eval_steps_per_second": 0.714,
+      "step": 4255
+    },
+    {
+      "epoch": 185.2173913043478,
+      "grad_norm": 13.189390182495117,
+      "learning_rate": 1.4782608695652176e-06,
+      "loss": 0.1084,
+      "step": 4260
+    },
+    {
+      "epoch": 185.65217391304347,
+      "grad_norm": 14.240042686462402,
+      "learning_rate": 1.4347826086956523e-06,
+      "loss": 0.1194,
+      "step": 4270
+    },
+    {
+      "epoch": 186.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.228926658630371,
+      "eval_runtime": 5.7824,
+      "eval_samples_per_second": 43.927,
+      "eval_steps_per_second": 0.692,
+      "step": 4278
+    },
+    {
+      "epoch": 186.08695652173913,
+      "grad_norm": 12.107095718383789,
+      "learning_rate": 1.391304347826087e-06,
+      "loss": 0.104,
+      "step": 4280
+    },
+    {
+      "epoch": 186.52173913043478,
+      "grad_norm": 11.538783073425293,
+      "learning_rate": 1.347826086956522e-06,
+      "loss": 0.1249,
+      "step": 4290
+    },
+    {
+      "epoch": 186.95652173913044,
+      "grad_norm": 12.4583740234375,
+      "learning_rate": 1.3043478260869566e-06,
+      "loss": 0.1154,
+      "step": 4300
+    },
+    {
+      "epoch": 187.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.1963870525360107,
+      "eval_runtime": 5.0691,
+      "eval_samples_per_second": 50.107,
+      "eval_steps_per_second": 0.789,
+      "step": 4301
+    },
+    {
+      "epoch": 187.3913043478261,
+      "grad_norm": 9.538721084594727,
+      "learning_rate": 1.2608695652173913e-06,
+      "loss": 0.1291,
+      "step": 4310
+    },
+    {
+      "epoch": 187.82608695652175,
+      "grad_norm": 12.623342514038086,
+      "learning_rate": 1.2173913043478262e-06,
+      "loss": 0.0964,
+      "step": 4320
+    },
+    {
+      "epoch": 188.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.200947642326355,
+      "eval_runtime": 6.199,
+      "eval_samples_per_second": 40.975,
+      "eval_steps_per_second": 0.645,
+      "step": 4324
+    },
+    {
+      "epoch": 188.2608695652174,
+      "grad_norm": 12.05130386352539,
+      "learning_rate": 1.173913043478261e-06,
+      "loss": 0.0965,
+      "step": 4330
+    },
+    {
+      "epoch": 188.69565217391303,
+      "grad_norm": 9.74976921081543,
+      "learning_rate": 1.1304347826086956e-06,
+      "loss": 0.0903,
+      "step": 4340
+    },
+    {
+      "epoch": 189.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.212328314781189,
+      "eval_runtime": 6.2152,
+      "eval_samples_per_second": 40.868,
+      "eval_steps_per_second": 0.644,
+      "step": 4347
+    },
+    {
+      "epoch": 189.1304347826087,
+      "grad_norm": 23.216026306152344,
+      "learning_rate": 1.0869565217391306e-06,
+      "loss": 0.114,
+      "step": 4350
+    },
+    {
+      "epoch": 189.56521739130434,
+      "grad_norm": 9.926918029785156,
+      "learning_rate": 1.0434782608695653e-06,
+      "loss": 0.0896,
+      "step": 4360
+    },
+    {
+      "epoch": 190.0,
+      "grad_norm": 21.126604080200195,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 0.1174,
+      "step": 4370
+    },
+    {
+      "epoch": 190.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.233533263206482,
+      "eval_runtime": 6.2079,
+      "eval_samples_per_second": 40.916,
+      "eval_steps_per_second": 0.644,
+      "step": 4370
+    },
+    {
+      "epoch": 190.43478260869566,
+      "grad_norm": 24.496295928955078,
+      "learning_rate": 9.565217391304349e-07,
+      "loss": 0.1433,
+      "step": 4380
+    },
+    {
+      "epoch": 190.8695652173913,
+      "grad_norm": 10.999076843261719,
+      "learning_rate": 9.130434782608697e-07,
+      "loss": 0.0846,
+      "step": 4390
+    },
+    {
+      "epoch": 191.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.2399306297302246,
+      "eval_runtime": 5.1624,
+      "eval_samples_per_second": 49.202,
+      "eval_steps_per_second": 0.775,
+      "step": 4393
+    },
+    {
+      "epoch": 191.30434782608697,
+      "grad_norm": 16.923410415649414,
+      "learning_rate": 8.695652173913044e-07,
+      "loss": 0.0901,
+      "step": 4400
+    },
+    {
+      "epoch": 191.7391304347826,
+      "grad_norm": 10.834136962890625,
+      "learning_rate": 8.260869565217392e-07,
+      "loss": 0.1073,
+      "step": 4410
+    },
+    {
+      "epoch": 192.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.243245244026184,
+      "eval_runtime": 5.0755,
+      "eval_samples_per_second": 50.044,
+      "eval_steps_per_second": 0.788,
+      "step": 4416
+    },
+    {
+      "epoch": 192.17391304347825,
+      "grad_norm": 9.724759101867676,
+      "learning_rate": 7.82608695652174e-07,
+      "loss": 0.126,
+      "step": 4420
+    },
+    {
+      "epoch": 192.6086956521739,
+      "grad_norm": 6.699430465698242,
+      "learning_rate": 7.391304347826088e-07,
+      "loss": 0.0892,
+      "step": 4430
+    },
+    {
+      "epoch": 193.0,
+      "eval_accuracy": 0.7480314960629921,
+      "eval_loss": 1.2603920698165894,
+      "eval_runtime": 5.3314,
+      "eval_samples_per_second": 47.643,
+      "eval_steps_per_second": 0.75,
+      "step": 4439
+    },
+    {
+      "epoch": 193.04347826086956,
+      "grad_norm": 14.250825881958008,
+      "learning_rate": 6.956521739130435e-07,
+      "loss": 0.1211,
+      "step": 4440
+    },
+    {
+      "epoch": 193.47826086956522,
+      "grad_norm": 9.508411407470703,
+      "learning_rate": 6.521739130434783e-07,
+      "loss": 0.0994,
+      "step": 4450
+    },
+    {
+      "epoch": 193.91304347826087,
+      "grad_norm": 13.035717010498047,
+      "learning_rate": 6.086956521739131e-07,
+      "loss": 0.1158,
+      "step": 4460
+    },
+    {
+      "epoch": 194.0,
+      "eval_accuracy": 0.7480314960629921,
+      "eval_loss": 1.2473300695419312,
+      "eval_runtime": 5.0663,
+      "eval_samples_per_second": 50.135,
+      "eval_steps_per_second": 0.79,
+      "step": 4462
+    },
+    {
+      "epoch": 194.34782608695653,
+      "grad_norm": 15.420454978942871,
+      "learning_rate": 5.652173913043478e-07,
+      "loss": 0.1159,
+      "step": 4470
+    },
+    {
+      "epoch": 194.7826086956522,
+      "grad_norm": 26.569854736328125,
+      "learning_rate": 5.217391304347826e-07,
+      "loss": 0.1153,
+      "step": 4480
+    },
+    {
+      "epoch": 195.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.2267494201660156,
+      "eval_runtime": 7.2318,
+      "eval_samples_per_second": 35.123,
+      "eval_steps_per_second": 0.553,
+      "step": 4485
+    },
+    {
+      "epoch": 195.2173913043478,
+      "grad_norm": 22.089872360229492,
+      "learning_rate": 4.782608695652174e-07,
+      "loss": 0.1009,
+      "step": 4490
+    },
+    {
+      "epoch": 195.65217391304347,
+      "grad_norm": 18.216550827026367,
+      "learning_rate": 4.347826086956522e-07,
+      "loss": 0.1208,
+      "step": 4500
+    },
+    {
+      "epoch": 196.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.2177777290344238,
+      "eval_runtime": 6.0473,
+      "eval_samples_per_second": 42.002,
+      "eval_steps_per_second": 0.661,
+      "step": 4508
+    },
+    {
+      "epoch": 196.08695652173913,
+      "grad_norm": 8.497892379760742,
+      "learning_rate": 3.91304347826087e-07,
+      "loss": 0.0903,
+      "step": 4510
+    },
+    {
+      "epoch": 196.52173913043478,
+      "grad_norm": 8.363734245300293,
+      "learning_rate": 3.4782608695652175e-07,
+      "loss": 0.0976,
+      "step": 4520
+    },
+    {
+      "epoch": 196.95652173913044,
+      "grad_norm": 11.241954803466797,
+      "learning_rate": 3.0434782608695656e-07,
+      "loss": 0.083,
+      "step": 4530
+    },
+    {
+      "epoch": 197.0,
+      "eval_accuracy": 0.7480314960629921,
+      "eval_loss": 1.2145416736602783,
+      "eval_runtime": 5.0766,
+      "eval_samples_per_second": 50.034,
+      "eval_steps_per_second": 0.788,
+      "step": 4531
+    },
+    {
+      "epoch": 197.3913043478261,
+      "grad_norm": 5.664416313171387,
+      "learning_rate": 2.608695652173913e-07,
+      "loss": 0.0773,
+      "step": 4540
+    },
+    {
+      "epoch": 197.82608695652175,
+      "grad_norm": 6.778021335601807,
+      "learning_rate": 2.173913043478261e-07,
+      "loss": 0.1331,
+      "step": 4550
+    },
+    {
+      "epoch": 198.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.2214672565460205,
+      "eval_runtime": 5.8006,
+      "eval_samples_per_second": 43.789,
+      "eval_steps_per_second": 0.69,
+      "step": 4554
+    },
+    {
+      "epoch": 198.2608695652174,
+      "grad_norm": 19.4536075592041,
+      "learning_rate": 1.7391304347826088e-07,
+      "loss": 0.1095,
+      "step": 4560
+    },
+    {
+      "epoch": 198.69565217391303,
+      "grad_norm": 11.348213195800781,
+      "learning_rate": 1.3043478260869566e-07,
+      "loss": 0.0943,
+      "step": 4570
+    },
+    {
+      "epoch": 199.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.2238408327102661,
+      "eval_runtime": 5.0996,
+      "eval_samples_per_second": 49.808,
+      "eval_steps_per_second": 0.784,
+      "step": 4577
+    },
+    {
+      "epoch": 199.1304347826087,
+      "grad_norm": 17.247846603393555,
+      "learning_rate": 8.695652173913044e-08,
+      "loss": 0.1049,
+      "step": 4580
+    },
+    {
+      "epoch": 199.56521739130434,
+      "grad_norm": 35.16756820678711,
+      "learning_rate": 4.347826086956522e-08,
+      "loss": 0.1164,
+      "step": 4590
+    },
+    {
+      "epoch": 200.0,
+      "grad_norm": 3.6230416297912598,
+      "learning_rate": 0.0,
+      "loss": 0.0926,
+      "step": 4600
+    },
+    {
+      "epoch": 200.0,
+      "eval_accuracy": 0.7440944881889764,
+      "eval_loss": 1.2236244678497314,
+      "eval_runtime": 6.0673,
+      "eval_samples_per_second": 41.864,
+      "eval_steps_per_second": 0.659,
+      "step": 4600
+    },
+    {
+      "epoch": 200.0,
+      "step": 4600,
+      "total_flos": 7.223651244601344e+18,
+      "train_loss": 0.17802202463150024,
+      "train_runtime": 11399.934,
+      "train_samples_per_second": 25.176,
+      "train_steps_per_second": 0.404
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 4600,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 200,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.223651244601344e+18,
+  "train_batch_size": 64,
+  "trial_name": null,
+  "trial_params": null
+}