diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,8637 @@
+{
+  "best_metric": 0.8701298701298701,
+  "best_model_checkpoint": "MAE-CT-M1N0-M12_v8_split5_v3/checkpoint-4760",
+  "epoch": 147.00579710144928,
+  "eval_steps": 500,
+  "global_step": 10350,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.000966183574879227,
+      "grad_norm": 5.63077974319458,
+      "learning_rate": 9.661835748792271e-08,
+      "loss": 0.7586,
+      "step": 10
+    },
+    {
+      "epoch": 0.001932367149758454,
+      "grad_norm": 2.0964934825897217,
+      "learning_rate": 1.9323671497584542e-07,
+      "loss": 0.7254,
+      "step": 20
+    },
+    {
+      "epoch": 0.002898550724637681,
+      "grad_norm": 2.0449180603027344,
+      "learning_rate": 2.8985507246376816e-07,
+      "loss": 0.6984,
+      "step": 30
+    },
+    {
+      "epoch": 0.003864734299516908,
+      "grad_norm": 1.9611525535583496,
+      "learning_rate": 3.8647342995169085e-07,
+      "loss": 0.7315,
+      "step": 40
+    },
+    {
+      "epoch": 0.004830917874396135,
+      "grad_norm": 3.5923993587493896,
+      "learning_rate": 4.830917874396135e-07,
+      "loss": 0.7284,
+      "step": 50
+    },
+    {
+      "epoch": 0.005797101449275362,
+      "grad_norm": 2.7955286502838135,
+      "learning_rate": 5.797101449275363e-07,
+      "loss": 0.6907,
+      "step": 60
+    },
+    {
+      "epoch": 0.00676328502415459,
+      "grad_norm": 4.491306304931641,
+      "learning_rate": 6.763285024154589e-07,
+      "loss": 0.685,
+      "step": 70
+    },
+    {
+      "epoch": 0.00676328502415459,
+      "eval_accuracy": 0.7792207792207793,
+      "eval_loss": 0.6756892800331116,
+      "eval_runtime": 18.2897,
+      "eval_samples_per_second": 4.21,
+      "eval_steps_per_second": 1.094,
+      "step": 70
+    },
+    {
+      "epoch": 1.0009661835748793,
+      "grad_norm": 3.633984088897705,
+      "learning_rate": 7.729468599033817e-07,
+      "loss": 0.6743,
+      "step": 80
+    },
+    {
+      "epoch": 1.0019323671497584,
+      "grad_norm": 4.844719409942627,
+      "learning_rate": 8.695652173913044e-07,
+      "loss": 0.65,
+      "step": 90
+    },
+    {
+      "epoch": 1.0028985507246377,
+      "grad_norm": 20.677204132080078,
+      "learning_rate": 9.66183574879227e-07,
+      "loss": 0.638,
+      "step": 100
+    },
+    {
+      "epoch": 1.0038647342995168,
+      "grad_norm": 4.349536895751953,
+      "learning_rate": 1.0628019323671499e-06,
+      "loss": 0.6479,
+      "step": 110
+    },
+    {
+      "epoch": 1.0048309178743962,
+      "grad_norm": 8.440875053405762,
+      "learning_rate": 1.1594202898550726e-06,
+      "loss": 0.5937,
+      "step": 120
+    },
+    {
+      "epoch": 1.0057971014492753,
+      "grad_norm": 9.938006401062012,
+      "learning_rate": 1.2560386473429952e-06,
+      "loss": 0.7384,
+      "step": 130
+    },
+    {
+      "epoch": 1.0067632850241546,
+      "grad_norm": 10.255725860595703,
+      "learning_rate": 1.3526570048309178e-06,
+      "loss": 0.5601,
+      "step": 140
+    },
+    {
+      "epoch": 1.0067632850241546,
+      "eval_accuracy": 0.6233766233766234,
+      "eval_loss": 0.6217549443244934,
+      "eval_runtime": 16.3642,
+      "eval_samples_per_second": 4.705,
+      "eval_steps_per_second": 1.222,
+      "step": 140
+    },
+    {
+      "epoch": 2.0009661835748793,
+      "grad_norm": 7.985101699829102,
+      "learning_rate": 1.4492753623188408e-06,
+      "loss": 0.7056,
+      "step": 150
+    },
+    {
+      "epoch": 2.0019323671497586,
+      "grad_norm": 7.1389007568359375,
+      "learning_rate": 1.5458937198067634e-06,
+      "loss": 0.5766,
+      "step": 160
+    },
+    {
+      "epoch": 2.0028985507246375,
+      "grad_norm": 12.669121742248535,
+      "learning_rate": 1.6425120772946862e-06,
+      "loss": 0.6746,
+      "step": 170
+    },
+    {
+      "epoch": 2.003864734299517,
+      "grad_norm": 6.364377498626709,
+      "learning_rate": 1.7391304347826088e-06,
+      "loss": 0.6363,
+      "step": 180
+    },
+    {
+      "epoch": 2.004830917874396,
+      "grad_norm": 9.995187759399414,
+      "learning_rate": 1.8357487922705318e-06,
+      "loss": 0.5837,
+      "step": 190
+    },
+    {
+      "epoch": 2.0057971014492755,
+      "grad_norm": 8.345070838928223,
+      "learning_rate": 1.932367149758454e-06,
+      "loss": 0.5432,
+      "step": 200
+    },
+    {
+      "epoch": 2.006763285024155,
+      "grad_norm": 37.9766845703125,
+      "learning_rate": 2.028985507246377e-06,
+      "loss": 0.6632,
+      "step": 210
+    },
+    {
+      "epoch": 2.006763285024155,
+      "eval_accuracy": 0.6233766233766234,
+      "eval_loss": 0.6156818866729736,
+      "eval_runtime": 16.4154,
+      "eval_samples_per_second": 4.691,
+      "eval_steps_per_second": 1.218,
+      "step": 210
+    },
+    {
+      "epoch": 3.0009661835748793,
+      "grad_norm": 12.400394439697266,
+      "learning_rate": 2.1256038647342997e-06,
+      "loss": 0.6558,
+      "step": 220
+    },
+    {
+      "epoch": 3.0019323671497586,
+      "grad_norm": 10.40713119506836,
+      "learning_rate": 2.222222222222222e-06,
+      "loss": 0.5752,
+      "step": 230
+    },
+    {
+      "epoch": 3.0028985507246375,
+      "grad_norm": 12.890250205993652,
+      "learning_rate": 2.3188405797101453e-06,
+      "loss": 0.528,
+      "step": 240
+    },
+    {
+      "epoch": 3.003864734299517,
+      "grad_norm": 10.023517608642578,
+      "learning_rate": 2.4154589371980677e-06,
+      "loss": 0.5814,
+      "step": 250
+    },
+    {
+      "epoch": 3.004830917874396,
+      "grad_norm": 10.220820426940918,
+      "learning_rate": 2.5120772946859904e-06,
+      "loss": 0.6507,
+      "step": 260
+    },
+    {
+      "epoch": 3.0057971014492755,
+      "grad_norm": 19.018766403198242,
+      "learning_rate": 2.6086956521739132e-06,
+      "loss": 0.5541,
+      "step": 270
+    },
+    {
+      "epoch": 3.006763285024155,
+      "grad_norm": 7.0940680503845215,
+      "learning_rate": 2.7053140096618356e-06,
+      "loss": 0.5153,
+      "step": 280
+    },
+    {
+      "epoch": 3.006763285024155,
+      "eval_accuracy": 0.6363636363636364,
+      "eval_loss": 0.565974771976471,
+      "eval_runtime": 16.5709,
+      "eval_samples_per_second": 4.647,
+      "eval_steps_per_second": 1.207,
+      "step": 280
+    },
+    {
+      "epoch": 4.000966183574879,
+      "grad_norm": 13.853311538696289,
+      "learning_rate": 2.801932367149759e-06,
+      "loss": 0.5608,
+      "step": 290
+    },
+    {
+      "epoch": 4.001932367149759,
+      "grad_norm": 24.638504028320312,
+      "learning_rate": 2.8985507246376816e-06,
+      "loss": 0.5793,
+      "step": 300
+    },
+    {
+      "epoch": 4.0028985507246375,
+      "grad_norm": 22.3065128326416,
+      "learning_rate": 2.995169082125604e-06,
+      "loss": 0.4284,
+      "step": 310
+    },
+    {
+      "epoch": 4.003864734299517,
+      "grad_norm": 30.956016540527344,
+      "learning_rate": 3.0917874396135268e-06,
+      "loss": 0.7425,
+      "step": 320
+    },
+    {
+      "epoch": 4.004830917874396,
+      "grad_norm": 12.773093223571777,
+      "learning_rate": 3.188405797101449e-06,
+      "loss": 0.5068,
+      "step": 330
+    },
+    {
+      "epoch": 4.005797101449275,
+      "grad_norm": 21.808074951171875,
+      "learning_rate": 3.2850241545893724e-06,
+      "loss": 0.4752,
+      "step": 340
+    },
+    {
+      "epoch": 4.006763285024155,
+      "grad_norm": 24.596450805664062,
+      "learning_rate": 3.381642512077295e-06,
+      "loss": 0.5008,
+      "step": 350
+    },
+    {
+      "epoch": 4.006763285024155,
+      "eval_accuracy": 0.7662337662337663,
+      "eval_loss": 0.5237749814987183,
+      "eval_runtime": 16.9105,
+      "eval_samples_per_second": 4.553,
+      "eval_steps_per_second": 1.183,
+      "step": 350
+    },
+    {
+      "epoch": 5.000966183574879,
+      "grad_norm": 41.83505630493164,
+      "learning_rate": 3.4782608695652175e-06,
+      "loss": 0.415,
+      "step": 360
+    },
+    {
+      "epoch": 5.001932367149759,
+      "grad_norm": 37.28093719482422,
+      "learning_rate": 3.5748792270531403e-06,
+      "loss": 0.5244,
+      "step": 370
+    },
+    {
+      "epoch": 5.0028985507246375,
+      "grad_norm": 14.932278633117676,
+      "learning_rate": 3.6714975845410635e-06,
+      "loss": 0.4871,
+      "step": 380
+    },
+    {
+      "epoch": 5.003864734299517,
+      "grad_norm": 35.8604736328125,
+      "learning_rate": 3.768115942028986e-06,
+      "loss": 0.5247,
+      "step": 390
+    },
+    {
+      "epoch": 5.004830917874396,
+      "grad_norm": 31.0769100189209,
+      "learning_rate": 3.864734299516908e-06,
+      "loss": 0.4985,
+      "step": 400
+    },
+    {
+      "epoch": 5.005797101449275,
+      "grad_norm": 16.91299057006836,
+      "learning_rate": 3.961352657004831e-06,
+      "loss": 0.5541,
+      "step": 410
+    },
+    {
+      "epoch": 5.006763285024155,
+      "grad_norm": 60.42302703857422,
+      "learning_rate": 4.057971014492754e-06,
+      "loss": 0.4879,
+      "step": 420
+    },
+    {
+      "epoch": 5.006763285024155,
+      "eval_accuracy": 0.7792207792207793,
+      "eval_loss": 0.5012220144271851,
+      "eval_runtime": 16.1267,
+      "eval_samples_per_second": 4.775,
+      "eval_steps_per_second": 1.24,
+      "step": 420
+    },
+    {
+      "epoch": 6.000966183574879,
+      "grad_norm": 12.599077224731445,
+      "learning_rate": 4.154589371980677e-06,
+      "loss": 0.4077,
+      "step": 430
+    },
+    {
+      "epoch": 6.001932367149759,
+      "grad_norm": 13.679814338684082,
+      "learning_rate": 4.251207729468599e-06,
+      "loss": 0.4849,
+      "step": 440
+    },
+    {
+      "epoch": 6.0028985507246375,
+      "grad_norm": 58.93899917602539,
+      "learning_rate": 4.347826086956522e-06,
+      "loss": 0.4561,
+      "step": 450
+    },
+    {
+      "epoch": 6.003864734299517,
+      "grad_norm": 46.164459228515625,
+      "learning_rate": 4.444444444444444e-06,
+      "loss": 0.4785,
+      "step": 460
+    },
+    {
+      "epoch": 6.004830917874396,
+      "grad_norm": 6.964693546295166,
+      "learning_rate": 4.541062801932368e-06,
+      "loss": 0.3693,
+      "step": 470
+    },
+    {
+      "epoch": 6.005797101449275,
+      "grad_norm": 65.39038848876953,
+      "learning_rate": 4.637681159420291e-06,
+      "loss": 0.6067,
+      "step": 480
+    },
+    {
+      "epoch": 6.006763285024155,
+      "grad_norm": 24.236276626586914,
+      "learning_rate": 4.7342995169082125e-06,
+      "loss": 0.3636,
+      "step": 490
+    },
+    {
+      "epoch": 6.006763285024155,
+      "eval_accuracy": 0.7012987012987013,
+      "eval_loss": 0.5639926195144653,
+      "eval_runtime": 16.4373,
+      "eval_samples_per_second": 4.684,
+      "eval_steps_per_second": 1.217,
+      "step": 490
+    },
+    {
+      "epoch": 7.000966183574879,
+      "grad_norm": 71.80524444580078,
+      "learning_rate": 4.830917874396135e-06,
+      "loss": 0.5243,
+      "step": 500
+    },
+    {
+      "epoch": 7.001932367149759,
+      "grad_norm": 60.86780548095703,
+      "learning_rate": 4.927536231884059e-06,
+      "loss": 0.4487,
+      "step": 510
+    },
+    {
+      "epoch": 7.0028985507246375,
+      "grad_norm": 18.15448760986328,
+      "learning_rate": 5.024154589371981e-06,
+      "loss": 0.4776,
+      "step": 520
+    },
+    {
+      "epoch": 7.003864734299517,
+      "grad_norm": 25.025348663330078,
+      "learning_rate": 5.1207729468599045e-06,
+      "loss": 0.3617,
+      "step": 530
+    },
+    {
+      "epoch": 7.004830917874396,
+      "grad_norm": 8.851715087890625,
+      "learning_rate": 5.2173913043478265e-06,
+      "loss": 0.3614,
+      "step": 540
+    },
+    {
+      "epoch": 7.005797101449275,
+      "grad_norm": 72.60469818115234,
+      "learning_rate": 5.314009661835749e-06,
+      "loss": 0.7049,
+      "step": 550
+    },
+    {
+      "epoch": 7.006763285024155,
+      "grad_norm": 8.072798728942871,
+      "learning_rate": 5.410628019323671e-06,
+      "loss": 0.7238,
+      "step": 560
+    },
+    {
+      "epoch": 7.006763285024155,
+      "eval_accuracy": 0.7012987012987013,
+      "eval_loss": 0.575639545917511,
+      "eval_runtime": 16.2503,
+      "eval_samples_per_second": 4.738,
+      "eval_steps_per_second": 1.231,
+      "step": 560
+    },
+    {
+      "epoch": 8.00096618357488,
+      "grad_norm": 26.914819717407227,
+      "learning_rate": 5.507246376811595e-06,
+      "loss": 0.4224,
+      "step": 570
+    },
+    {
+      "epoch": 8.001932367149758,
+      "grad_norm": 12.590118408203125,
+      "learning_rate": 5.603864734299518e-06,
+      "loss": 1.1756,
+      "step": 580
+    },
+    {
+      "epoch": 8.002898550724638,
+      "grad_norm": 31.019224166870117,
+      "learning_rate": 5.70048309178744e-06,
+      "loss": 0.5193,
+      "step": 590
+    },
+    {
+      "epoch": 8.003864734299517,
+      "grad_norm": 25.586570739746094,
+      "learning_rate": 5.797101449275363e-06,
+      "loss": 0.5632,
+      "step": 600
+    },
+    {
+      "epoch": 8.004830917874395,
+      "grad_norm": 7.837695121765137,
+      "learning_rate": 5.893719806763285e-06,
+      "loss": 0.2409,
+      "step": 610
+    },
+    {
+      "epoch": 8.005797101449275,
+      "grad_norm": 45.919281005859375,
+      "learning_rate": 5.990338164251208e-06,
+      "loss": 0.4052,
+      "step": 620
+    },
+    {
+      "epoch": 8.006763285024155,
+      "grad_norm": 9.32940673828125,
+      "learning_rate": 6.086956521739132e-06,
+      "loss": 0.3339,
+      "step": 630
+    },
+    {
+      "epoch": 8.006763285024155,
+      "eval_accuracy": 0.6883116883116883,
+      "eval_loss": 0.9894827604293823,
+      "eval_runtime": 15.6835,
+      "eval_samples_per_second": 4.91,
+      "eval_steps_per_second": 1.275,
+      "step": 630
+    },
+    {
+      "epoch": 9.00096618357488,
+      "grad_norm": 4.154600620269775,
+      "learning_rate": 6.1835748792270535e-06,
+      "loss": 0.4827,
+      "step": 640
+    },
+    {
+      "epoch": 9.001932367149758,
+      "grad_norm": 43.86359786987305,
+      "learning_rate": 6.280193236714976e-06,
+      "loss": 0.4219,
+      "step": 650
+    },
+    {
+      "epoch": 9.002898550724638,
+      "grad_norm": 87.51858520507812,
+      "learning_rate": 6.376811594202898e-06,
+      "loss": 0.2489,
+      "step": 660
+    },
+    {
+      "epoch": 9.003864734299517,
+      "grad_norm": 11.23896598815918,
+      "learning_rate": 6.473429951690822e-06,
+      "loss": 0.5007,
+      "step": 670
+    },
+    {
+      "epoch": 9.004830917874395,
+      "grad_norm": 27.375701904296875,
+      "learning_rate": 6.570048309178745e-06,
+      "loss": 0.2891,
+      "step": 680
+    },
+    {
+      "epoch": 9.005797101449275,
+      "grad_norm": 51.312923431396484,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 0.4243,
+      "step": 690
+    },
+    {
+      "epoch": 9.006763285024155,
+      "grad_norm": 29.63850975036621,
+      "learning_rate": 6.76328502415459e-06,
+      "loss": 0.4152,
+      "step": 700
+    },
+    {
+      "epoch": 9.006763285024155,
+      "eval_accuracy": 0.8181818181818182,
+      "eval_loss": 0.5031455159187317,
+      "eval_runtime": 15.7949,
+      "eval_samples_per_second": 4.875,
+      "eval_steps_per_second": 1.266,
+      "step": 700
+    },
+    {
+      "epoch": 10.00096618357488,
+      "grad_norm": 7.607165813446045,
+      "learning_rate": 6.859903381642513e-06,
+      "loss": 0.5239,
+      "step": 710
+    },
+    {
+      "epoch": 10.001932367149758,
+      "grad_norm": 6.940621852874756,
+      "learning_rate": 6.956521739130435e-06,
+      "loss": 0.2281,
+      "step": 720
+    },
+    {
+      "epoch": 10.002898550724638,
+      "grad_norm": 57.28858947753906,
+      "learning_rate": 7.053140096618359e-06,
+      "loss": 0.4428,
+      "step": 730
+    },
+    {
+      "epoch": 10.003864734299517,
+      "grad_norm": 90.76024627685547,
+      "learning_rate": 7.149758454106281e-06,
+      "loss": 0.6327,
+      "step": 740
+    },
+    {
+      "epoch": 10.004830917874395,
+      "grad_norm": 59.252777099609375,
+      "learning_rate": 7.246376811594203e-06,
+      "loss": 0.7428,
+      "step": 750
+    },
+    {
+      "epoch": 10.005797101449275,
+      "grad_norm": 24.238967895507812,
+      "learning_rate": 7.342995169082127e-06,
+      "loss": 0.4681,
+      "step": 760
+    },
+    {
+      "epoch": 10.006763285024155,
+      "grad_norm": 0.46667662262916565,
+      "learning_rate": 7.439613526570049e-06,
+      "loss": 0.3126,
+      "step": 770
+    },
+    {
+      "epoch": 10.006763285024155,
+      "eval_accuracy": 0.7272727272727273,
+      "eval_loss": 0.5349549055099487,
+      "eval_runtime": 15.9144,
+      "eval_samples_per_second": 4.838,
+      "eval_steps_per_second": 1.257,
+      "step": 770
+    },
+    {
+      "epoch": 11.00096618357488,
+      "grad_norm": 3.171098470687866,
+      "learning_rate": 7.536231884057972e-06,
+      "loss": 0.3612,
+      "step": 780
+    },
+    {
+      "epoch": 11.001932367149758,
+      "grad_norm": 51.79991912841797,
+      "learning_rate": 7.632850241545895e-06,
+      "loss": 0.2405,
+      "step": 790
+    },
+    {
+      "epoch": 11.002898550724638,
+      "grad_norm": 15.745019912719727,
+      "learning_rate": 7.729468599033817e-06,
+      "loss": 0.3578,
+      "step": 800
+    },
+    {
+      "epoch": 11.003864734299517,
+      "grad_norm": 46.960411071777344,
+      "learning_rate": 7.82608695652174e-06,
+      "loss": 0.4183,
+      "step": 810
+    },
+    {
+      "epoch": 11.004830917874395,
+      "grad_norm": 19.980567932128906,
+      "learning_rate": 7.922705314009662e-06,
+      "loss": 1.0693,
+      "step": 820
+    },
+    {
+      "epoch": 11.005797101449275,
+      "grad_norm": 7.018507480621338,
+      "learning_rate": 8.019323671497586e-06,
+      "loss": 0.4977,
+      "step": 830
+    },
+    {
+      "epoch": 11.006763285024155,
+      "grad_norm": 42.957664489746094,
+      "learning_rate": 8.115942028985508e-06,
+      "loss": 0.4479,
+      "step": 840
+    },
+    {
+      "epoch": 11.006763285024155,
+      "eval_accuracy": 0.8311688311688312,
+      "eval_loss": 0.42781147360801697,
+      "eval_runtime": 17.5652,
+      "eval_samples_per_second": 4.384,
+      "eval_steps_per_second": 1.139,
+      "step": 840
+    },
+    {
+      "epoch": 12.00096618357488,
+      "grad_norm": 30.186668395996094,
+      "learning_rate": 8.212560386473431e-06,
+      "loss": 0.3227,
+      "step": 850
+    },
+    {
+      "epoch": 12.001932367149758,
+      "grad_norm": 95.9303207397461,
+      "learning_rate": 8.309178743961353e-06,
+      "loss": 0.3342,
+      "step": 860
+    },
+    {
+      "epoch": 12.002898550724638,
+      "grad_norm": 72.12458038330078,
+      "learning_rate": 8.405797101449275e-06,
+      "loss": 0.3496,
+      "step": 870
+    },
+    {
+      "epoch": 12.003864734299517,
+      "grad_norm": 12.492197036743164,
+      "learning_rate": 8.502415458937199e-06,
+      "loss": 0.2547,
+      "step": 880
+    },
+    {
+      "epoch": 12.004830917874395,
+      "grad_norm": 31.763296127319336,
+      "learning_rate": 8.599033816425122e-06,
+      "loss": 0.4288,
+      "step": 890
+    },
+    {
+      "epoch": 12.005797101449275,
+      "grad_norm": 46.830039978027344,
+      "learning_rate": 8.695652173913044e-06,
+      "loss": 0.4244,
+      "step": 900
+    },
+    {
+      "epoch": 12.006763285024155,
+      "grad_norm": 66.69847869873047,
+      "learning_rate": 8.792270531400966e-06,
+      "loss": 0.5548,
+      "step": 910
+    },
+    {
+      "epoch": 12.006763285024155,
+      "eval_accuracy": 0.7012987012987013,
+      "eval_loss": 0.6865193843841553,
+      "eval_runtime": 16.7685,
+      "eval_samples_per_second": 4.592,
+      "eval_steps_per_second": 1.193,
+      "step": 910
+    },
+    {
+      "epoch": 13.00096618357488,
+      "grad_norm": 1.0070035457611084,
+      "learning_rate": 8.888888888888888e-06,
+      "loss": 0.1739,
+      "step": 920
+    },
+    {
+      "epoch": 13.001932367149758,
+      "grad_norm": 27.44700813293457,
+      "learning_rate": 8.985507246376812e-06,
+      "loss": 0.3626,
+      "step": 930
+    },
+    {
+      "epoch": 13.002898550724638,
+      "grad_norm": 83.7531967163086,
+      "learning_rate": 9.082125603864736e-06,
+      "loss": 0.4617,
+      "step": 940
+    },
+    {
+      "epoch": 13.003864734299517,
+      "grad_norm": 29.952964782714844,
+      "learning_rate": 9.178743961352658e-06,
+      "loss": 0.3059,
+      "step": 950
+    },
+    {
+      "epoch": 13.004830917874395,
+      "grad_norm": 19.046611785888672,
+      "learning_rate": 9.275362318840581e-06,
+      "loss": 0.3571,
+      "step": 960
+    },
+    {
+      "epoch": 13.005797101449275,
+      "grad_norm": 4.5947980880737305,
+      "learning_rate": 9.371980676328503e-06,
+      "loss": 0.5925,
+      "step": 970
+    },
+    {
+      "epoch": 13.006763285024155,
+      "grad_norm": 0.1116618886590004,
+      "learning_rate": 9.468599033816425e-06,
+      "loss": 0.1509,
+      "step": 980
+    },
+    {
+      "epoch": 13.006763285024155,
+      "eval_accuracy": 0.7142857142857143,
+      "eval_loss": 0.8144263029098511,
+      "eval_runtime": 16.8839,
+      "eval_samples_per_second": 4.561,
+      "eval_steps_per_second": 1.185,
+      "step": 980
+    },
+    {
+      "epoch": 14.00096618357488,
+      "grad_norm": 7.001115798950195,
+      "learning_rate": 9.565217391304349e-06,
+      "loss": 0.291,
+      "step": 990
+    },
+    {
+      "epoch": 14.001932367149758,
+      "grad_norm": 81.9195327758789,
+      "learning_rate": 9.66183574879227e-06,
+      "loss": 0.4609,
+      "step": 1000
+    },
+    {
+      "epoch": 14.002898550724638,
+      "grad_norm": 72.53138732910156,
+      "learning_rate": 9.758454106280194e-06,
+      "loss": 0.2614,
+      "step": 1010
+    },
+    {
+      "epoch": 14.003864734299517,
+      "grad_norm": 88.42354583740234,
+      "learning_rate": 9.855072463768118e-06,
+      "loss": 0.5673,
+      "step": 1020
+    },
+    {
+      "epoch": 14.004830917874395,
+      "grad_norm": 182.70054626464844,
+      "learning_rate": 9.95169082125604e-06,
+      "loss": 0.3395,
+      "step": 1030
+    },
+    {
+      "epoch": 14.005797101449275,
+      "grad_norm": 75.77576446533203,
+      "learning_rate": 9.994632313472894e-06,
+      "loss": 0.4042,
+      "step": 1040
+    },
+    {
+      "epoch": 14.006763285024155,
+      "grad_norm": 0.33039167523384094,
+      "learning_rate": 9.98389694041868e-06,
+      "loss": 0.4038,
+      "step": 1050
+    },
+    {
+      "epoch": 14.006763285024155,
+      "eval_accuracy": 0.7922077922077922,
+      "eval_loss": 0.6039356589317322,
+      "eval_runtime": 17.6515,
+      "eval_samples_per_second": 4.362,
+      "eval_steps_per_second": 1.133,
+      "step": 1050
+    },
+    {
+      "epoch": 15.00096618357488,
+      "grad_norm": 0.12073725461959839,
+      "learning_rate": 9.973161567364467e-06,
+      "loss": 0.2022,
+      "step": 1060
+    },
+    {
+      "epoch": 15.001932367149758,
+      "grad_norm": 30.217771530151367,
+      "learning_rate": 9.962426194310253e-06,
+      "loss": 0.2995,
+      "step": 1070
+    },
+    {
+      "epoch": 15.002898550724638,
+      "grad_norm": 59.161224365234375,
+      "learning_rate": 9.95169082125604e-06,
+      "loss": 0.287,
+      "step": 1080
+    },
+    {
+      "epoch": 15.003864734299517,
+      "grad_norm": 51.48099136352539,
+      "learning_rate": 9.940955448201826e-06,
+      "loss": 0.3779,
+      "step": 1090
+    },
+    {
+      "epoch": 15.004830917874395,
+      "grad_norm": 67.52349853515625,
+      "learning_rate": 9.930220075147611e-06,
+      "loss": 0.1764,
+      "step": 1100
+    },
+    {
+      "epoch": 15.005797101449275,
+      "grad_norm": 1.3956732749938965,
+      "learning_rate": 9.919484702093398e-06,
+      "loss": 0.4947,
+      "step": 1110
+    },
+    {
+      "epoch": 15.006763285024155,
+      "grad_norm": 0.3445068299770355,
+      "learning_rate": 9.908749329039184e-06,
+      "loss": 0.2748,
+      "step": 1120
+    },
+    {
+      "epoch": 15.006763285024155,
+      "eval_accuracy": 0.7662337662337663,
+      "eval_loss": 1.183449149131775,
+      "eval_runtime": 15.8908,
+      "eval_samples_per_second": 4.846,
+      "eval_steps_per_second": 1.259,
+      "step": 1120
+    },
+    {
+      "epoch": 16.000966183574878,
+      "grad_norm": 25.889986038208008,
+      "learning_rate": 9.89801395598497e-06,
+      "loss": 0.4726,
+      "step": 1130
+    },
+    {
+      "epoch": 16.00193236714976,
+      "grad_norm": 96.12169647216797,
+      "learning_rate": 9.887278582930757e-06,
+      "loss": 0.7904,
+      "step": 1140
+    },
+    {
+      "epoch": 16.002898550724638,
+      "grad_norm": 36.80066680908203,
+      "learning_rate": 9.876543209876543e-06,
+      "loss": 0.4876,
+      "step": 1150
+    },
+    {
+      "epoch": 16.003864734299516,
+      "grad_norm": 0.03129780665040016,
+      "learning_rate": 9.865807836822331e-06,
+      "loss": 0.0427,
+      "step": 1160
+    },
+    {
+      "epoch": 16.004830917874397,
+      "grad_norm": 11.805863380432129,
+      "learning_rate": 9.855072463768118e-06,
+      "loss": 0.5398,
+      "step": 1170
+    },
+    {
+      "epoch": 16.005797101449275,
+      "grad_norm": 14.385322570800781,
+      "learning_rate": 9.844337090713904e-06,
+      "loss": 0.6715,
+      "step": 1180
+    },
+    {
+      "epoch": 16.006763285024153,
+      "grad_norm": 0.18029144406318665,
+      "learning_rate": 9.833601717659689e-06,
+      "loss": 0.4552,
+      "step": 1190
+    },
+    {
+      "epoch": 16.006763285024153,
+      "eval_accuracy": 0.7532467532467533,
+      "eval_loss": 0.7594266533851624,
+      "eval_runtime": 15.3345,
+      "eval_samples_per_second": 5.021,
+      "eval_steps_per_second": 1.304,
+      "step": 1190
+    },
+    {
+      "epoch": 17.000966183574878,
+      "grad_norm": 0.4168717563152313,
+      "learning_rate": 9.822866344605476e-06,
+      "loss": 0.1674,
+      "step": 1200
+    },
+    {
+      "epoch": 17.00193236714976,
+      "grad_norm": 0.9499201774597168,
+      "learning_rate": 9.812130971551262e-06,
+      "loss": 0.2045,
+      "step": 1210
+    },
+    {
+      "epoch": 17.002898550724638,
+      "grad_norm": 162.95167541503906,
+      "learning_rate": 9.801395598497048e-06,
+      "loss": 0.7506,
+      "step": 1220
+    },
+    {
+      "epoch": 17.003864734299516,
+      "grad_norm": 1.6326782703399658,
+      "learning_rate": 9.790660225442835e-06,
+      "loss": 0.398,
+      "step": 1230
+    },
+    {
+      "epoch": 17.004830917874397,
+      "grad_norm": 15.130646705627441,
+      "learning_rate": 9.779924852388621e-06,
+      "loss": 0.3404,
+      "step": 1240
+    },
+    {
+      "epoch": 17.005797101449275,
+      "grad_norm": 0.08184617757797241,
+      "learning_rate": 9.769189479334408e-06,
+      "loss": 0.1294,
+      "step": 1250
+    },
+    {
+      "epoch": 17.006763285024153,
+      "grad_norm": 316.427001953125,
+      "learning_rate": 9.758454106280194e-06,
+      "loss": 0.5584,
+      "step": 1260
+    },
+    {
+      "epoch": 17.006763285024153,
+      "eval_accuracy": 0.7922077922077922,
+      "eval_loss": 0.9481449127197266,
+      "eval_runtime": 15.5575,
+      "eval_samples_per_second": 4.949,
+      "eval_steps_per_second": 1.286,
+      "step": 1260
+    },
+    {
+      "epoch": 18.000966183574878,
+      "grad_norm": 0.19292013347148895,
+      "learning_rate": 9.74771873322598e-06,
+      "loss": 0.2401,
+      "step": 1270
+    },
+    {
+      "epoch": 18.00193236714976,
+      "grad_norm": 0.07158540934324265,
+      "learning_rate": 9.736983360171767e-06,
+      "loss": 0.2077,
+      "step": 1280
+    },
+    {
+      "epoch": 18.002898550724638,
+      "grad_norm": 10.159412384033203,
+      "learning_rate": 9.726247987117554e-06,
+      "loss": 0.0463,
+      "step": 1290
+    },
+    {
+      "epoch": 18.003864734299516,
+      "grad_norm": 0.16102541983127594,
+      "learning_rate": 9.71551261406334e-06,
+      "loss": 0.114,
+      "step": 1300
+    },
+    {
+      "epoch": 18.004830917874397,
+      "grad_norm": 0.08238888531923294,
+      "learning_rate": 9.704777241009125e-06,
+      "loss": 0.2091,
+      "step": 1310
+    },
+    {
+      "epoch": 18.005797101449275,
+      "grad_norm": 0.6891394257545471,
+      "learning_rate": 9.694041867954911e-06,
+      "loss": 0.133,
+      "step": 1320
+    },
+    {
+      "epoch": 18.006763285024153,
+      "grad_norm": 0.02565833181142807,
+      "learning_rate": 9.683306494900698e-06,
+      "loss": 0.0919,
+      "step": 1330
+    },
+    {
+      "epoch": 18.006763285024153,
+      "eval_accuracy": 0.7662337662337663,
+      "eval_loss": 1.0079569816589355,
+      "eval_runtime": 17.4793,
+      "eval_samples_per_second": 4.405,
+      "eval_steps_per_second": 1.144,
+      "step": 1330
+    },
+    {
+      "epoch": 19.000966183574878,
+      "grad_norm": 0.20424237847328186,
+      "learning_rate": 9.672571121846484e-06,
+      "loss": 0.0852,
+      "step": 1340
+    },
+    {
+      "epoch": 19.00193236714976,
+      "grad_norm": 0.056213993579149246,
+      "learning_rate": 9.66183574879227e-06,
+      "loss": 0.0014,
+      "step": 1350
+    },
+    {
+      "epoch": 19.002898550724638,
+      "grad_norm": 66.11773681640625,
+      "learning_rate": 9.651100375738057e-06,
+      "loss": 0.0981,
+      "step": 1360
+    },
+    {
+      "epoch": 19.003864734299516,
+      "grad_norm": 0.020251819863915443,
+      "learning_rate": 9.640365002683844e-06,
+      "loss": 0.1629,
+      "step": 1370
+    },
+    {
+      "epoch": 19.004830917874397,
+      "grad_norm": 0.08519980311393738,
+      "learning_rate": 9.62962962962963e-06,
+      "loss": 0.1436,
+      "step": 1380
+    },
+    {
+      "epoch": 19.005797101449275,
+      "grad_norm": 0.04679597541689873,
+      "learning_rate": 9.618894256575418e-06,
+      "loss": 0.4893,
+      "step": 1390
+    },
+    {
+      "epoch": 19.006763285024153,
+      "grad_norm": 0.04914074391126633,
+      "learning_rate": 9.608158883521203e-06,
+      "loss": 0.2309,
+      "step": 1400
+    },
+    {
+      "epoch": 19.006763285024153,
+      "eval_accuracy": 0.8181818181818182,
+      "eval_loss": 0.8453315496444702,
+      "eval_runtime": 16.6003,
+      "eval_samples_per_second": 4.638,
+      "eval_steps_per_second": 1.205,
+      "step": 1400
+    },
+    {
+      "epoch": 20.000966183574878,
+      "grad_norm": 0.1529330313205719,
+      "learning_rate": 9.59742351046699e-06,
+      "loss": 0.2619,
+      "step": 1410
+    },
+    {
+      "epoch": 20.00193236714976,
+      "grad_norm": 0.02123161591589451,
+      "learning_rate": 9.586688137412776e-06,
+      "loss": 0.0256,
+      "step": 1420
+    },
+    {
+      "epoch": 20.002898550724638,
+      "grad_norm": 0.014065372757613659,
+      "learning_rate": 9.575952764358562e-06,
+      "loss": 0.1996,
+      "step": 1430
+    },
+    {
+      "epoch": 20.003864734299516,
+      "grad_norm": 0.7381812930107117,
+      "learning_rate": 9.565217391304349e-06,
+      "loss": 0.0424,
+      "step": 1440
+    },
+    {
+      "epoch": 20.004830917874397,
+      "grad_norm": 0.06800421327352524,
+      "learning_rate": 9.554482018250135e-06,
+      "loss": 0.21,
+      "step": 1450
+    },
+    {
+      "epoch": 20.005797101449275,
+      "grad_norm": 80.73758697509766,
+      "learning_rate": 9.543746645195922e-06,
+      "loss": 0.0242,
+      "step": 1460
+    },
+    {
+      "epoch": 20.006763285024153,
+      "grad_norm": 0.0316414013504982,
+      "learning_rate": 9.533011272141708e-06,
+      "loss": 0.191,
+      "step": 1470
+    },
+    {
+      "epoch": 20.006763285024153,
+      "eval_accuracy": 0.7662337662337663,
+      "eval_loss": 1.069457769393921,
+      "eval_runtime": 147.8544,
+      "eval_samples_per_second": 0.521,
+      "eval_steps_per_second": 0.135,
+      "step": 1470
+    },
+    {
+      "epoch": 21.000966183574878,
+      "grad_norm": 0.012025800533592701,
+      "learning_rate": 9.522275899087494e-06,
+      "loss": 0.0317,
+      "step": 1480
+    },
+    {
+      "epoch": 21.00193236714976,
+      "grad_norm": 115.9197769165039,
+      "learning_rate": 9.511540526033281e-06,
+      "loss": 0.2591,
+      "step": 1490
+    },
+    {
+      "epoch": 21.002898550724638,
+      "grad_norm": 0.7661476135253906,
+      "learning_rate": 9.500805152979067e-06,
+      "loss": 0.2929,
+      "step": 1500
+    },
+    {
+      "epoch": 21.003864734299516,
+      "grad_norm": 89.17171478271484,
+      "learning_rate": 9.490069779924854e-06,
+      "loss": 0.501,
+      "step": 1510
+    },
+    {
+      "epoch": 21.004830917874397,
+      "grad_norm": 0.056331489235162735,
+      "learning_rate": 9.479334406870639e-06,
+      "loss": 0.2755,
+      "step": 1520
+    },
+    {
+      "epoch": 21.005797101449275,
+      "grad_norm": 0.035313863307237625,
+      "learning_rate": 9.468599033816425e-06,
+      "loss": 0.315,
+      "step": 1530
+    },
+    {
+      "epoch": 21.006763285024153,
+      "grad_norm": 0.06920187175273895,
+      "learning_rate": 9.457863660762211e-06,
+      "loss": 0.2013,
+      "step": 1540
+    },
+    {
+      "epoch": 21.006763285024153,
+      "eval_accuracy": 0.7402597402597403,
+      "eval_loss": 1.465735673904419,
+      "eval_runtime": 17.3838,
+      "eval_samples_per_second": 4.429,
+      "eval_steps_per_second": 1.15,
+      "step": 1540
+    },
+    {
+      "epoch": 22.000966183574878,
+      "grad_norm": 3.0384302139282227,
+      "learning_rate": 9.447128287707998e-06,
+      "loss": 0.2171,
+      "step": 1550
+    },
+    {
+      "epoch": 22.00193236714976,
+      "grad_norm": 0.022978052496910095,
+      "learning_rate": 9.436392914653784e-06,
+      "loss": 0.2218,
+      "step": 1560
+    },
+    {
+      "epoch": 22.002898550724638,
+      "grad_norm": 0.2309640645980835,
+      "learning_rate": 9.42565754159957e-06,
+      "loss": 0.1913,
+      "step": 1570
+    },
+    {
+      "epoch": 22.003864734299516,
+      "grad_norm": 0.027831105515360832,
+      "learning_rate": 9.414922168545357e-06,
+      "loss": 0.0069,
+      "step": 1580
+    },
+    {
+      "epoch": 22.004830917874397,
+      "grad_norm": 0.01753852143883705,
+      "learning_rate": 9.404186795491144e-06,
+      "loss": 0.0014,
+      "step": 1590
+    },
+    {
+      "epoch": 22.005797101449275,
+      "grad_norm": 0.051825933158397675,
+      "learning_rate": 9.39345142243693e-06,
+      "loss": 0.0012,
+      "step": 1600
+    },
+    {
+      "epoch": 22.006763285024153,
+      "grad_norm": 23.33223533630371,
+      "learning_rate": 9.382716049382717e-06,
+      "loss": 0.6645,
+      "step": 1610
+    },
+    {
+      "epoch": 22.006763285024153,
+      "eval_accuracy": 0.8051948051948052,
+      "eval_loss": 1.0601636171340942,
+      "eval_runtime": 18.3493,
+      "eval_samples_per_second": 4.196,
+      "eval_steps_per_second": 1.09,
+      "step": 1610
+    },
+    {
+      "epoch": 23.000966183574878,
+      "grad_norm": 42.5572395324707,
+      "learning_rate": 9.371980676328503e-06,
+      "loss": 0.1616,
+      "step": 1620
+    },
+    {
+      "epoch": 23.00193236714976,
+      "grad_norm": 0.06940260529518127,
+      "learning_rate": 9.36124530327429e-06,
+      "loss": 0.0026,
+      "step": 1630
+    },
+    {
+      "epoch": 23.002898550724638,
+      "grad_norm": 24.54999351501465,
+      "learning_rate": 9.350509930220076e-06,
+      "loss": 0.1298,
+      "step": 1640
+    },
+    {
+      "epoch": 23.003864734299516,
+      "grad_norm": 0.014982779510319233,
+      "learning_rate": 9.339774557165862e-06,
+      "loss": 0.0021,
+      "step": 1650
+    },
+    {
+      "epoch": 23.004830917874397,
+      "grad_norm": 0.022154109552502632,
+      "learning_rate": 9.329039184111649e-06,
+      "loss": 0.1601,
+      "step": 1660
+    },
+    {
+      "epoch": 23.005797101449275,
+      "grad_norm": 0.1332763135433197,
+      "learning_rate": 9.318303811057435e-06,
+      "loss": 0.1839,
+      "step": 1670
+    },
+    {
+      "epoch": 23.006763285024153,
+      "grad_norm": 0.0714530497789383,
+      "learning_rate": 9.307568438003222e-06,
+      "loss": 0.1083,
+      "step": 1680
+    },
+    {
+      "epoch": 23.006763285024153,
+      "eval_accuracy": 0.7532467532467533,
+      "eval_loss": 1.2147879600524902,
+      "eval_runtime": 42.3924,
+      "eval_samples_per_second": 1.816,
+      "eval_steps_per_second": 0.472,
+      "step": 1680
+    },
+    {
+      "epoch": 24.000966183574878,
+      "grad_norm": 10.763915061950684,
+      "learning_rate": 9.296833064949008e-06,
+      "loss": 0.0153,
+      "step": 1690
+    },
+    {
+      "epoch": 24.00193236714976,
+      "grad_norm": 0.34083086252212524,
+      "learning_rate": 9.286097691894795e-06,
+      "loss": 0.3552,
+      "step": 1700
+    },
+    {
+      "epoch": 24.002898550724638,
+      "grad_norm": 0.09669429808855057,
+      "learning_rate": 9.275362318840581e-06,
+      "loss": 0.1833,
+      "step": 1710
+    },
+    {
+      "epoch": 24.003864734299516,
+      "grad_norm": 3.468942403793335,
+      "learning_rate": 9.264626945786368e-06,
+      "loss": 0.0984,
+      "step": 1720
+    },
+    {
+      "epoch": 24.004830917874397,
+      "grad_norm": 105.82479095458984,
+      "learning_rate": 9.253891572732154e-06,
+      "loss": 0.1335,
+      "step": 1730
+    },
+    {
+      "epoch": 24.005797101449275,
+      "grad_norm": 0.03978487849235535,
+      "learning_rate": 9.243156199677939e-06,
+      "loss": 0.0364,
+      "step": 1740
+    },
+    {
+      "epoch": 24.006763285024153,
+      "grad_norm": 0.05299947410821915,
+      "learning_rate": 9.232420826623725e-06,
+      "loss": 0.0885,
+      "step": 1750
+    },
+    {
+      "epoch": 24.006763285024153,
+      "eval_accuracy": 0.7792207792207793,
+      "eval_loss": 1.200799822807312,
+      "eval_runtime": 18.7931,
+      "eval_samples_per_second": 4.097,
+      "eval_steps_per_second": 1.064,
+      "step": 1750
+    },
+    {
+      "epoch": 25.000966183574878,
+      "grad_norm": 0.07938944548368454,
+      "learning_rate": 9.221685453569512e-06,
+      "loss": 0.1654,
+      "step": 1760
+    },
+    {
+      "epoch": 25.00193236714976,
+      "grad_norm": 0.017891909927129745,
+      "learning_rate": 9.210950080515298e-06,
+      "loss": 0.1959,
+      "step": 1770
+    },
+    {
+      "epoch": 25.002898550724638,
+      "grad_norm": 0.03742482513189316,
+      "learning_rate": 9.200214707461085e-06,
+      "loss": 0.2129,
+      "step": 1780
+    },
+    {
+      "epoch": 25.003864734299516,
+      "grad_norm": 0.018285313621163368,
+      "learning_rate": 9.189479334406871e-06,
+      "loss": 0.0005,
+      "step": 1790
+    },
+    {
+      "epoch": 25.004830917874397,
+      "grad_norm": 0.012646480463445187,
+      "learning_rate": 9.178743961352658e-06,
+      "loss": 0.2598,
+      "step": 1800
+    },
+    {
+      "epoch": 25.005797101449275,
+      "grad_norm": 0.1877349466085434,
+      "learning_rate": 9.168008588298444e-06,
+      "loss": 0.3652,
+      "step": 1810
+    },
+    {
+      "epoch": 25.006763285024153,
+      "grad_norm": 0.04589417949318886,
+      "learning_rate": 9.15727321524423e-06,
+      "loss": 0.0015,
+      "step": 1820
+    },
+    {
+      "epoch": 25.006763285024153,
+      "eval_accuracy": 0.7532467532467533,
+      "eval_loss": 1.298748254776001,
+      "eval_runtime": 18.6643,
+      "eval_samples_per_second": 4.126,
+      "eval_steps_per_second": 1.072,
+      "step": 1820
+    },
+    {
+      "epoch": 26.000966183574878,
+      "grad_norm": 0.007107927929610014,
+      "learning_rate": 9.146537842190017e-06,
+      "loss": 0.0241,
+      "step": 1830
+    },
+    {
+      "epoch": 26.00193236714976,
+      "grad_norm": 0.2932748794555664,
+      "learning_rate": 9.135802469135803e-06,
+      "loss": 0.1536,
+      "step": 1840
+    },
+    {
+      "epoch": 26.002898550724638,
+      "grad_norm": 0.005459180101752281,
+      "learning_rate": 9.12506709608159e-06,
+      "loss": 0.0019,
+      "step": 1850
+    },
+    {
+      "epoch": 26.003864734299516,
+      "grad_norm": 0.05199627950787544,
+      "learning_rate": 9.114331723027376e-06,
+      "loss": 0.0547,
+      "step": 1860
+    },
+    {
+      "epoch": 26.004830917874397,
+      "grad_norm": 0.0027262712828814983,
+      "learning_rate": 9.103596349973163e-06,
+      "loss": 0.0003,
+      "step": 1870
+    },
+    {
+      "epoch": 26.005797101449275,
+      "grad_norm": 0.005350043997168541,
+      "learning_rate": 9.092860976918949e-06,
+      "loss": 0.24,
+      "step": 1880
+    },
+    {
+      "epoch": 26.006763285024153,
+      "grad_norm": 0.003055410925298929,
+      "learning_rate": 9.082125603864736e-06,
+      "loss": 0.2372,
+      "step": 1890
+    },
+    {
+      "epoch": 26.006763285024153,
+      "eval_accuracy": 0.7532467532467533,
+      "eval_loss": 1.6224983930587769,
+      "eval_runtime": 19.2376,
+      "eval_samples_per_second": 4.003,
+      "eval_steps_per_second": 1.04,
+      "step": 1890
+    },
+    {
+      "epoch": 27.000966183574878,
+      "grad_norm": 0.013318424113094807,
+      "learning_rate": 9.071390230810522e-06,
+      "loss": 0.0192,
+      "step": 1900
+    },
+    {
+      "epoch": 27.00193236714976,
+      "grad_norm": 0.01380719244480133,
+      "learning_rate": 9.060654857756308e-06,
+      "loss": 0.0365,
+      "step": 1910
+    },
+    {
+      "epoch": 27.002898550724638,
+      "grad_norm": 12.373915672302246,
+      "learning_rate": 9.049919484702095e-06,
+      "loss": 0.6307,
+      "step": 1920
+    },
+    {
+      "epoch": 27.003864734299516,
+      "grad_norm": 186.6871337890625,
+      "learning_rate": 9.039184111647881e-06,
+      "loss": 0.5701,
+      "step": 1930
+    },
+    {
+      "epoch": 27.004830917874397,
+      "grad_norm": 0.41726887226104736,
+      "learning_rate": 9.028448738593668e-06,
+      "loss": 0.2799,
+      "step": 1940
+    },
+    {
+      "epoch": 27.005797101449275,
+      "grad_norm": 0.0075998492538928986,
+      "learning_rate": 9.017713365539453e-06,
+      "loss": 0.0908,
+      "step": 1950
+    },
+    {
+      "epoch": 27.006763285024153,
+      "grad_norm": 0.0030398606322705746,
+      "learning_rate": 9.006977992485239e-06,
+      "loss": 0.001,
+      "step": 1960
+    },
+    {
+      "epoch": 27.006763285024153,
+      "eval_accuracy": 0.7662337662337663,
+      "eval_loss": 1.1689021587371826,
+      "eval_runtime": 15.3074,
+      "eval_samples_per_second": 5.03,
+      "eval_steps_per_second": 1.307,
+      "step": 1960
+    },
+    {
+      "epoch": 28.000966183574878,
+      "grad_norm": 0.004823345225304365,
+      "learning_rate": 8.996242619431025e-06,
+      "loss": 0.002,
+      "step": 1970
+    },
+    {
+      "epoch": 28.00193236714976,
+      "grad_norm": 37.93857955932617,
+      "learning_rate": 8.985507246376812e-06,
+      "loss": 0.0136,
+      "step": 1980
+    },
+    {
+      "epoch": 28.002898550724638,
+      "grad_norm": 0.008819162845611572,
+      "learning_rate": 8.974771873322598e-06,
+      "loss": 0.0003,
+      "step": 1990
+    },
+    {
+      "epoch": 28.003864734299516,
+      "grad_norm": 0.02176712639629841,
+      "learning_rate": 8.964036500268385e-06,
+      "loss": 0.0005,
+      "step": 2000
+    },
+    {
+      "epoch": 28.004830917874397,
+      "grad_norm": 2.0372092723846436,
+      "learning_rate": 8.953301127214171e-06,
+      "loss": 0.0883,
+      "step": 2010
+    },
+    {
+      "epoch": 28.005797101449275,
+      "grad_norm": 0.0033761654049158096,
+      "learning_rate": 8.942565754159958e-06,
+      "loss": 0.0002,
+      "step": 2020
+    },
+    {
+      "epoch": 28.006763285024153,
+      "grad_norm": 0.015467183664441109,
+      "learning_rate": 8.931830381105744e-06,
+      "loss": 0.0006,
+      "step": 2030
+    },
+    {
+      "epoch": 28.006763285024153,
+      "eval_accuracy": 0.7532467532467533,
+      "eval_loss": 1.3817362785339355,
+      "eval_runtime": 14.6935,
+      "eval_samples_per_second": 5.24,
+      "eval_steps_per_second": 1.361,
+      "step": 2030
+    },
+    {
+      "epoch": 29.000966183574878,
+      "grad_norm": 0.00563892163336277,
+      "learning_rate": 8.92109500805153e-06,
+      "loss": 0.0002,
+      "step": 2040
+    },
+    {
+      "epoch": 29.00193236714976,
+      "grad_norm": 0.008719475008547306,
+      "learning_rate": 8.910359634997317e-06,
+      "loss": 0.0002,
+      "step": 2050
+    },
+    {
+      "epoch": 29.002898550724638,
+      "grad_norm": 0.012711520306766033,
+      "learning_rate": 8.899624261943104e-06,
+      "loss": 0.0021,
+      "step": 2060
+    },
+    {
+      "epoch": 29.003864734299516,
+      "grad_norm": 0.0029001745861023664,
+      "learning_rate": 8.888888888888888e-06,
+      "loss": 0.0012,
+      "step": 2070
+    },
+    {
+      "epoch": 29.004830917874397,
+      "grad_norm": 0.0027202004566788673,
+      "learning_rate": 8.878153515834675e-06,
+      "loss": 0.0259,
+      "step": 2080
+    },
+    {
+      "epoch": 29.005797101449275,
+      "grad_norm": 0.0022866763174533844,
+      "learning_rate": 8.867418142780463e-06,
+      "loss": 0.001,
+      "step": 2090
+    },
+    {
+      "epoch": 29.006763285024153,
+      "grad_norm": 0.009289148263633251,
+      "learning_rate": 8.85668276972625e-06,
+      "loss": 0.0002,
+      "step": 2100
+    },
+    {
+      "epoch": 29.006763285024153,
+      "eval_accuracy": 0.7272727272727273,
+      "eval_loss": 1.714272379875183,
+      "eval_runtime": 14.4439,
+      "eval_samples_per_second": 5.331,
+      "eval_steps_per_second": 1.385,
+      "step": 2100
+    },
+    {
+      "epoch": 30.000966183574878,
+      "grad_norm": 0.009948095306754112,
+      "learning_rate": 8.845947396672036e-06,
+      "loss": 0.1673,
+      "step": 2110
+    },
+    {
+      "epoch": 30.00193236714976,
+      "grad_norm": 0.0249008946120739,
+      "learning_rate": 8.835212023617822e-06,
+      "loss": 0.1536,
+      "step": 2120
+    },
+    {
+      "epoch": 30.002898550724638,
+      "grad_norm": 0.12144894897937775,
+      "learning_rate": 8.824476650563609e-06,
+      "loss": 0.1607,
+      "step": 2130
+    },
+    {
+      "epoch": 30.003864734299516,
+      "grad_norm": 7.651823997497559,
+      "learning_rate": 8.813741277509395e-06,
+      "loss": 0.1802,
+      "step": 2140
+    },
+    {
+      "epoch": 30.004830917874397,
+      "grad_norm": 0.005258599761873484,
+      "learning_rate": 8.803005904455182e-06,
+      "loss": 0.0002,
+      "step": 2150
+    },
+    {
+      "epoch": 30.005797101449275,
+      "grad_norm": 0.004621783271431923,
+      "learning_rate": 8.792270531400966e-06,
+      "loss": 0.0222,
+      "step": 2160
+    },
+    {
+      "epoch": 30.006763285024153,
+      "grad_norm": 0.012738402932882309,
+      "learning_rate": 8.781535158346753e-06,
+      "loss": 0.0012,
+      "step": 2170
+    },
+    {
+      "epoch": 30.006763285024153,
+      "eval_accuracy": 0.7272727272727273,
+      "eval_loss": 1.8865219354629517,
+      "eval_runtime": 14.9876,
+      "eval_samples_per_second": 5.138,
+      "eval_steps_per_second": 1.334,
+      "step": 2170
+    },
+    {
+      "epoch": 31.000966183574878,
+      "grad_norm": 0.004862803500145674,
+      "learning_rate": 8.77079978529254e-06,
+      "loss": 0.0002,
+      "step": 2180
+    },
+    {
+      "epoch": 31.00193236714976,
+      "grad_norm": 0.008277139626443386,
+      "learning_rate": 8.760064412238326e-06,
+      "loss": 0.0002,
+      "step": 2190
+    },
+    {
+      "epoch": 31.002898550724638,
+      "grad_norm": 0.011266889050602913,
+      "learning_rate": 8.749329039184112e-06,
+      "loss": 0.0256,
+      "step": 2200
+    },
+    {
+      "epoch": 31.003864734299516,
+      "grad_norm": 0.037181366235017776,
+      "learning_rate": 8.738593666129899e-06,
+      "loss": 0.0389,
+      "step": 2210
+    },
+    {
+      "epoch": 31.004830917874397,
+      "grad_norm": 0.011379989795386791,
+      "learning_rate": 8.727858293075685e-06,
+      "loss": 0.0703,
+      "step": 2220
+    },
+    {
+      "epoch": 31.005797101449275,
+      "grad_norm": 0.0035820791963487864,
+      "learning_rate": 8.717122920021472e-06,
+      "loss": 0.0123,
+      "step": 2230
+    },
+    {
+      "epoch": 31.006763285024153,
+      "grad_norm": 0.11101510375738144,
+      "learning_rate": 8.706387546967258e-06,
+      "loss": 0.153,
+      "step": 2240
+    },
+    {
+      "epoch": 31.006763285024153,
+      "eval_accuracy": 0.6623376623376623,
+      "eval_loss": 2.4574289321899414,
+      "eval_runtime": 14.6437,
+      "eval_samples_per_second": 5.258,
+      "eval_steps_per_second": 1.366,
+      "step": 2240
+    },
+    {
+      "epoch": 32.00096618357488,
+      "grad_norm": 0.007295526098459959,
+      "learning_rate": 8.695652173913044e-06,
+      "loss": 0.2227,
+      "step": 2250
+    },
+    {
+      "epoch": 32.001932367149756,
+      "grad_norm": 0.005096717271953821,
+      "learning_rate": 8.684916800858831e-06,
+      "loss": 0.1528,
+      "step": 2260
+    },
+    {
+      "epoch": 32.00289855072464,
+      "grad_norm": 0.007106207311153412,
+      "learning_rate": 8.674181427804617e-06,
+      "loss": 0.0649,
+      "step": 2270
+    },
+    {
+      "epoch": 32.00386473429952,
+      "grad_norm": 16.95619773864746,
+      "learning_rate": 8.663446054750402e-06,
+      "loss": 0.0751,
+      "step": 2280
+    },
+    {
+      "epoch": 32.00483091787439,
+      "grad_norm": 0.00349339353851974,
+      "learning_rate": 8.652710681696189e-06,
+      "loss": 0.0014,
+      "step": 2290
+    },
+    {
+      "epoch": 32.005797101449275,
+      "grad_norm": 0.010887386277318,
+      "learning_rate": 8.641975308641975e-06,
+      "loss": 0.0001,
+      "step": 2300
+    },
+    {
+      "epoch": 32.00676328502416,
+      "grad_norm": 0.011478613130748272,
+      "learning_rate": 8.631239935587761e-06,
+      "loss": 0.1308,
+      "step": 2310
+    },
+    {
+      "epoch": 32.00676328502416,
+      "eval_accuracy": 0.8051948051948052,
+      "eval_loss": 1.1799964904785156,
+      "eval_runtime": 15.238,
+      "eval_samples_per_second": 5.053,
+      "eval_steps_per_second": 1.313,
+      "step": 2310
+    },
+    {
+      "epoch": 33.00096618357488,
+      "grad_norm": 0.0030596402939409018,
+      "learning_rate": 8.62050456253355e-06,
+      "loss": 0.0001,
+      "step": 2320
+    },
+    {
+      "epoch": 33.001932367149756,
+      "grad_norm": 0.008988602086901665,
+      "learning_rate": 8.609769189479336e-06,
+      "loss": 0.1962,
+      "step": 2330
+    },
+    {
+      "epoch": 33.00289855072464,
+      "grad_norm": 0.11388076096773148,
+      "learning_rate": 8.599033816425122e-06,
+      "loss": 0.0001,
+      "step": 2340
+    },
+    {
+      "epoch": 33.00386473429952,
+      "grad_norm": 0.012941384688019753,
+      "learning_rate": 8.588298443370909e-06,
+      "loss": 0.0001,
+      "step": 2350
+    },
+    {
+      "epoch": 33.00483091787439,
+      "grad_norm": 117.23433685302734,
+      "learning_rate": 8.577563070316695e-06,
+      "loss": 0.2454,
+      "step": 2360
+    },
+    {
+      "epoch": 33.005797101449275,
+      "grad_norm": 0.0073850443586707115,
+      "learning_rate": 8.56682769726248e-06,
+      "loss": 0.0001,
+      "step": 2370
+    },
+    {
+      "epoch": 33.00676328502416,
+      "grad_norm": 0.038754310458898544,
+      "learning_rate": 8.556092324208267e-06,
+      "loss": 0.0002,
+      "step": 2380
+    },
+    {
+      "epoch": 33.00676328502416,
+      "eval_accuracy": 0.7792207792207793,
+      "eval_loss": 1.2817308902740479,
+      "eval_runtime": 14.7367,
+      "eval_samples_per_second": 5.225,
+      "eval_steps_per_second": 1.357,
+      "step": 2380
+    },
+    {
+      "epoch": 34.00096618357488,
+      "grad_norm": 0.006986475549638271,
+      "learning_rate": 8.545356951154053e-06,
+      "loss": 0.0001,
+      "step": 2390
+    },
+    {
+      "epoch": 34.001932367149756,
+      "grad_norm": 0.08525365591049194,
+      "learning_rate": 8.53462157809984e-06,
+      "loss": 0.1159,
+      "step": 2400
+    },
+    {
+      "epoch": 34.00289855072464,
+      "grad_norm": 0.005141557659953833,
+      "learning_rate": 8.523886205045626e-06,
+      "loss": 0.0209,
+      "step": 2410
+    },
+    {
+      "epoch": 34.00386473429952,
+      "grad_norm": 0.002347426488995552,
+      "learning_rate": 8.513150831991412e-06,
+      "loss": 0.0035,
+      "step": 2420
+    },
+    {
+      "epoch": 34.00483091787439,
+      "grad_norm": 0.021269038319587708,
+      "learning_rate": 8.502415458937199e-06,
+      "loss": 0.0002,
+      "step": 2430
+    },
+    {
+      "epoch": 34.005797101449275,
+      "grad_norm": 0.0015353817725554109,
+      "learning_rate": 8.491680085882985e-06,
+      "loss": 0.0001,
+      "step": 2440
+    },
+    {
+      "epoch": 34.00676328502416,
+      "grad_norm": 0.002692780690267682,
+      "learning_rate": 8.480944712828772e-06,
+      "loss": 0.0001,
+      "step": 2450
+    },
+    {
+      "epoch": 34.00676328502416,
+      "eval_accuracy": 0.7792207792207793,
+      "eval_loss": 1.2770259380340576,
+      "eval_runtime": 14.7085,
+      "eval_samples_per_second": 5.235,
+      "eval_steps_per_second": 1.36,
+      "step": 2450
+    },
+    {
+      "epoch": 35.00096618357488,
+      "grad_norm": 0.0036268895491957664,
+      "learning_rate": 8.470209339774558e-06,
+      "loss": 0.0001,
+      "step": 2460
+    },
+    {
+      "epoch": 35.001932367149756,
+      "grad_norm": 0.02713877335190773,
+      "learning_rate": 8.459473966720345e-06,
+      "loss": 0.0001,
+      "step": 2470
+    },
+    {
+      "epoch": 35.00289855072464,
+      "grad_norm": 0.004495933186262846,
+      "learning_rate": 8.448738593666131e-06,
+      "loss": 0.0001,
+      "step": 2480
+    },
+    {
+      "epoch": 35.00386473429952,
+      "grad_norm": 0.0023365523666143417,
+      "learning_rate": 8.438003220611916e-06,
+      "loss": 0.0001,
+      "step": 2490
+    },
+    {
+      "epoch": 35.00483091787439,
+      "grad_norm": 0.006139947567135096,
+      "learning_rate": 8.427267847557702e-06,
+      "loss": 0.0001,
+      "step": 2500
+    },
+    {
+      "epoch": 35.005797101449275,
+      "grad_norm": 0.3539928197860718,
+      "learning_rate": 8.416532474503489e-06,
+      "loss": 0.0001,
+      "step": 2510
+    },
+    {
+      "epoch": 35.00676328502416,
+      "grad_norm": 0.006096704863011837,
+      "learning_rate": 8.405797101449275e-06,
+      "loss": 0.0001,
+      "step": 2520
+    },
+    {
+      "epoch": 35.00676328502416,
+      "eval_accuracy": 0.7922077922077922,
+      "eval_loss": 1.2778986692428589,
+      "eval_runtime": 14.4303,
+      "eval_samples_per_second": 5.336,
+      "eval_steps_per_second": 1.386,
+      "step": 2520
+    },
+    {
+      "epoch": 36.00096618357488,
+      "grad_norm": 0.0016143172979354858,
+      "learning_rate": 8.395061728395062e-06,
+      "loss": 0.0001,
+      "step": 2530
+    },
+    {
+      "epoch": 36.001932367149756,
+      "grad_norm": 0.004879065323621035,
+      "learning_rate": 8.384326355340848e-06,
+      "loss": 0.0001,
+      "step": 2540
+    },
+    {
+      "epoch": 36.00289855072464,
+      "grad_norm": 0.002586913527920842,
+      "learning_rate": 8.373590982286636e-06,
+      "loss": 0.0001,
+      "step": 2550
+    },
+    {
+      "epoch": 36.00386473429952,
+      "grad_norm": 0.011908585205674171,
+      "learning_rate": 8.362855609232423e-06,
+      "loss": 0.0001,
+      "step": 2560
+    },
+    {
+      "epoch": 36.00483091787439,
+      "grad_norm": 0.002101797377690673,
+      "learning_rate": 8.352120236178209e-06,
+      "loss": 0.0005,
+      "step": 2570
+    },
+    {
+      "epoch": 36.005797101449275,
+      "grad_norm": 0.33965930342674255,
+      "learning_rate": 8.341384863123994e-06,
+      "loss": 0.1018,
+      "step": 2580
+    },
+    {
+      "epoch": 36.00676328502416,
+      "grad_norm": 0.0029156110249459743,
+      "learning_rate": 8.33064949006978e-06,
+      "loss": 0.0001,
+      "step": 2590
+    },
+    {
+      "epoch": 36.00676328502416,
+      "eval_accuracy": 0.7792207792207793,
+      "eval_loss": 1.3971278667449951,
+      "eval_runtime": 14.5016,
+      "eval_samples_per_second": 5.31,
+      "eval_steps_per_second": 1.379,
+      "step": 2590
+    },
+    {
+      "epoch": 37.00096618357488,
+      "grad_norm": 160.6613311767578,
+      "learning_rate": 8.319914117015567e-06,
+      "loss": 0.0107,
+      "step": 2600
+    },
+    {
+      "epoch": 37.001932367149756,
+      "grad_norm": 0.001831389730796218,
+      "learning_rate": 8.309178743961353e-06,
+      "loss": 0.0021,
+      "step": 2610
+    },
+    {
+      "epoch": 37.00289855072464,
+      "grad_norm": 0.0028115217573940754,
+      "learning_rate": 8.29844337090714e-06,
+      "loss": 0.0002,
+      "step": 2620
+    },
+    {
+      "epoch": 37.00386473429952,
+      "grad_norm": 100.3589859008789,
+      "learning_rate": 8.287707997852926e-06,
+      "loss": 0.1948,
+      "step": 2630
+    },
+    {
+      "epoch": 37.00483091787439,
+      "grad_norm": 0.0055212159641087055,
+      "learning_rate": 8.276972624798713e-06,
+      "loss": 0.0035,
+      "step": 2640
+    },
+    {
+      "epoch": 37.005797101449275,
+      "grad_norm": 0.0011701483745127916,
+      "learning_rate": 8.266237251744499e-06,
+      "loss": 0.0058,
+      "step": 2650
+    },
+    {
+      "epoch": 37.00676328502416,
+      "grad_norm": 0.0021236445754766464,
+      "learning_rate": 8.255501878690286e-06,
+      "loss": 0.0001,
+      "step": 2660
+    },
+    {
+      "epoch": 37.00676328502416,
+      "eval_accuracy": 0.8181818181818182,
+      "eval_loss": 1.126293420791626,
+      "eval_runtime": 14.5035,
+      "eval_samples_per_second": 5.309,
+      "eval_steps_per_second": 1.379,
+      "step": 2660
+    },
+    {
+      "epoch": 38.00096618357488,
+      "grad_norm": 0.002591678872704506,
+      "learning_rate": 8.244766505636072e-06,
+      "loss": 0.0001,
+      "step": 2670
+    },
+    {
+      "epoch": 38.001932367149756,
+      "grad_norm": 0.002082381397485733,
+      "learning_rate": 8.234031132581858e-06,
+      "loss": 0.0001,
+      "step": 2680
+    },
+    {
+      "epoch": 38.00289855072464,
+      "grad_norm": 0.0038508835714310408,
+      "learning_rate": 8.223295759527645e-06,
+      "loss": 0.1384,
+      "step": 2690
+    },
+    {
+      "epoch": 38.00386473429952,
+      "grad_norm": 0.0028780594002455473,
+      "learning_rate": 8.212560386473431e-06,
+      "loss": 0.0001,
+      "step": 2700
+    },
+    {
+      "epoch": 38.00483091787439,
+      "grad_norm": 0.0069783455692231655,
+      "learning_rate": 8.201825013419216e-06,
+      "loss": 0.0194,
+      "step": 2710
+    },
+    {
+      "epoch": 38.005797101449275,
+      "grad_norm": 0.0021866625174880028,
+      "learning_rate": 8.191089640365003e-06,
+      "loss": 0.0001,
+      "step": 2720
+    },
+    {
+      "epoch": 38.00676328502416,
+      "grad_norm": 0.002578013576567173,
+      "learning_rate": 8.180354267310789e-06,
+      "loss": 0.0001,
+      "step": 2730
+    },
+    {
+      "epoch": 38.00676328502416,
+      "eval_accuracy": 0.8181818181818182,
+      "eval_loss": 1.1232645511627197,
+      "eval_runtime": 14.7179,
+      "eval_samples_per_second": 5.232,
+      "eval_steps_per_second": 1.359,
+      "step": 2730
+    },
+    {
+      "epoch": 39.00096618357488,
+      "grad_norm": 0.003141900757327676,
+      "learning_rate": 8.169618894256575e-06,
+      "loss": 0.0001,
+      "step": 2740
+    },
+    {
+      "epoch": 39.001932367149756,
+      "grad_norm": 0.004278377164155245,
+      "learning_rate": 8.158883521202362e-06,
+      "loss": 0.1379,
+      "step": 2750
+    },
+    {
+      "epoch": 39.00289855072464,
+      "grad_norm": 0.0016402786131948233,
+      "learning_rate": 8.148148148148148e-06,
+      "loss": 0.1475,
+      "step": 2760
+    },
+    {
+      "epoch": 39.00386473429952,
+      "grad_norm": 0.1566852182149887,
+      "learning_rate": 8.137412775093935e-06,
+      "loss": 0.3523,
+      "step": 2770
+    },
+    {
+      "epoch": 39.00483091787439,
+      "grad_norm": 0.0037191323935985565,
+      "learning_rate": 8.126677402039721e-06,
+      "loss": 0.2542,
+      "step": 2780
+    },
+    {
+      "epoch": 39.005797101449275,
+      "grad_norm": 0.0030009618494659662,
+      "learning_rate": 8.115942028985508e-06,
+      "loss": 0.0003,
+      "step": 2790
+    },
+    {
+      "epoch": 39.00676328502416,
+      "grad_norm": 0.0024690297432243824,
+      "learning_rate": 8.105206655931294e-06,
+      "loss": 0.0675,
+      "step": 2800
+    },
+    {
+      "epoch": 39.00676328502416,
+      "eval_accuracy": 0.7662337662337663,
+      "eval_loss": 1.4884586334228516,
+      "eval_runtime": 15.941,
+      "eval_samples_per_second": 4.83,
+      "eval_steps_per_second": 1.255,
+      "step": 2800
+    },
+    {
+      "epoch": 40.00096618357488,
+      "grad_norm": 0.00310198194347322,
+      "learning_rate": 8.09447128287708e-06,
+      "loss": 0.0029,
+      "step": 2810
+    },
+    {
+      "epoch": 40.001932367149756,
+      "grad_norm": 0.02491198293864727,
+      "learning_rate": 8.083735909822867e-06,
+      "loss": 0.0004,
+      "step": 2820
+    },
+    {
+      "epoch": 40.00289855072464,
+      "grad_norm": 0.0024280883371829987,
+      "learning_rate": 8.073000536768653e-06,
+      "loss": 0.0016,
+      "step": 2830
+    },
+    {
+      "epoch": 40.00386473429952,
+      "grad_norm": 39.02526092529297,
+      "learning_rate": 8.06226516371444e-06,
+      "loss": 0.2267,
+      "step": 2840
+    },
+    {
+      "epoch": 40.00483091787439,
+      "grad_norm": 0.007119787856936455,
+      "learning_rate": 8.051529790660226e-06,
+      "loss": 0.1246,
+      "step": 2850
+    },
+    {
+      "epoch": 40.005797101449275,
+      "grad_norm": 0.0054905591532588005,
+      "learning_rate": 8.040794417606013e-06,
+      "loss": 0.0002,
+      "step": 2860
+    },
+    {
+      "epoch": 40.00676328502416,
+      "grad_norm": 0.003688159631565213,
+      "learning_rate": 8.0300590445518e-06,
+      "loss": 0.0002,
+      "step": 2870
+    },
+    {
+      "epoch": 40.00676328502416,
+      "eval_accuracy": 0.7012987012987013,
+      "eval_loss": 1.8405940532684326,
+      "eval_runtime": 38.4778,
+      "eval_samples_per_second": 2.001,
+      "eval_steps_per_second": 0.52,
+      "step": 2870
+    },
+    {
+      "epoch": 41.00096618357488,
+      "grad_norm": 0.0019798579160124063,
+      "learning_rate": 8.019323671497586e-06,
+      "loss": 0.0001,
+      "step": 2880
+    },
+    {
+      "epoch": 41.001932367149756,
+      "grad_norm": 0.011942965909838676,
+      "learning_rate": 8.008588298443372e-06,
+      "loss": 0.0002,
+      "step": 2890
+    },
+    {
+      "epoch": 41.00289855072464,
+      "grad_norm": 0.0018374843057245016,
+      "learning_rate": 7.997852925389159e-06,
+      "loss": 0.0003,
+      "step": 2900
+    },
+    {
+      "epoch": 41.00386473429952,
+      "grad_norm": 0.0034626899287104607,
+      "learning_rate": 7.987117552334945e-06,
+      "loss": 0.0001,
+      "step": 2910
+    },
+    {
+      "epoch": 41.00483091787439,
+      "grad_norm": 0.0038331467658281326,
+      "learning_rate": 7.97638217928073e-06,
+      "loss": 0.0001,
+      "step": 2920
+    },
+    {
+      "epoch": 41.005797101449275,
+      "grad_norm": 0.0017002819804474711,
+      "learning_rate": 7.965646806226516e-06,
+      "loss": 0.0002,
+      "step": 2930
+    },
+    {
+      "epoch": 41.00676328502416,
+      "grad_norm": 0.0059528648853302,
+      "learning_rate": 7.954911433172303e-06,
+      "loss": 0.0001,
+      "step": 2940
+    },
+    {
+      "epoch": 41.00676328502416,
+      "eval_accuracy": 0.7532467532467533,
+      "eval_loss": 1.9084670543670654,
+      "eval_runtime": 15.0568,
+      "eval_samples_per_second": 5.114,
+      "eval_steps_per_second": 1.328,
+      "step": 2940
+    },
+    {
+      "epoch": 42.00096618357488,
+      "grad_norm": 0.0026344398502260447,
+      "learning_rate": 7.94417606011809e-06,
+      "loss": 0.0001,
+      "step": 2950
+    },
+    {
+      "epoch": 42.001932367149756,
+      "grad_norm": 0.0017737130401656032,
+      "learning_rate": 7.933440687063876e-06,
+      "loss": 0.0001,
+      "step": 2960
+    },
+    {
+      "epoch": 42.00289855072464,
+      "grad_norm": 222.6859588623047,
+      "learning_rate": 7.922705314009662e-06,
+      "loss": 0.179,
+      "step": 2970
+    },
+    {
+      "epoch": 42.00386473429952,
+      "grad_norm": 0.022273195907473564,
+      "learning_rate": 7.911969940955449e-06,
+      "loss": 0.1046,
+      "step": 2980
+    },
+    {
+      "epoch": 42.00483091787439,
+      "grad_norm": 0.02884945087134838,
+      "learning_rate": 7.901234567901235e-06,
+      "loss": 0.0018,
+      "step": 2990
+    },
+    {
+      "epoch": 42.005797101449275,
+      "grad_norm": 1.2516549825668335,
+      "learning_rate": 7.890499194847021e-06,
+      "loss": 0.6459,
+      "step": 3000
+    },
+    {
+      "epoch": 42.00676328502416,
+      "grad_norm": 0.2831589877605438,
+      "learning_rate": 7.879763821792808e-06,
+      "loss": 0.0005,
+      "step": 3010
+    },
+    {
+      "epoch": 42.00676328502416,
+      "eval_accuracy": 0.7142857142857143,
+      "eval_loss": 1.9380286931991577,
+      "eval_runtime": 15.4466,
+      "eval_samples_per_second": 4.985,
+      "eval_steps_per_second": 1.295,
+      "step": 3010
+    },
+    {
+      "epoch": 43.00096618357488,
+      "grad_norm": 0.00937322061508894,
+      "learning_rate": 7.869028448738594e-06,
+      "loss": 0.4209,
+      "step": 3020
+    },
+    {
+      "epoch": 43.001932367149756,
+      "grad_norm": 5.275388717651367,
+      "learning_rate": 7.85829307568438e-06,
+      "loss": 0.1863,
+      "step": 3030
+    },
+    {
+      "epoch": 43.00289855072464,
+      "grad_norm": 0.0025148040149360895,
+      "learning_rate": 7.847557702630167e-06,
+      "loss": 0.0002,
+      "step": 3040
+    },
+    {
+      "epoch": 43.00386473429952,
+      "grad_norm": 0.011758288368582726,
+      "learning_rate": 7.836822329575954e-06,
+      "loss": 0.0002,
+      "step": 3050
+    },
+    {
+      "epoch": 43.00483091787439,
+      "grad_norm": 145.53335571289062,
+      "learning_rate": 7.82608695652174e-06,
+      "loss": 0.2425,
+      "step": 3060
+    },
+    {
+      "epoch": 43.005797101449275,
+      "grad_norm": 0.005929389502853155,
+      "learning_rate": 7.815351583467527e-06,
+      "loss": 0.179,
+      "step": 3070
+    },
+    {
+      "epoch": 43.00676328502416,
+      "grad_norm": 0.00786946527659893,
+      "learning_rate": 7.804616210413313e-06,
+      "loss": 0.1589,
+      "step": 3080
+    },
+    {
+      "epoch": 43.00676328502416,
+      "eval_accuracy": 0.8311688311688312,
+      "eval_loss": 0.9673866033554077,
+      "eval_runtime": 16.8427,
+      "eval_samples_per_second": 4.572,
+      "eval_steps_per_second": 1.187,
+      "step": 3080
+    },
+    {
+      "epoch": 44.00096618357488,
+      "grad_norm": 0.005430703517049551,
+      "learning_rate": 7.7938808373591e-06,
+      "loss": 0.0002,
+      "step": 3090
+    },
+    {
+      "epoch": 44.001932367149756,
+      "grad_norm": 4.347066879272461,
+      "learning_rate": 7.783145464304886e-06,
+      "loss": 0.0009,
+      "step": 3100
+    },
+    {
+      "epoch": 44.00289855072464,
+      "grad_norm": 0.014723542146384716,
+      "learning_rate": 7.772410091250672e-06,
+      "loss": 0.0006,
+      "step": 3110
+    },
+    {
+      "epoch": 44.00386473429952,
+      "grad_norm": 289.0671691894531,
+      "learning_rate": 7.761674718196459e-06,
+      "loss": 0.2869,
+      "step": 3120
+    },
+    {
+      "epoch": 44.00483091787439,
+      "grad_norm": 0.0035440954379737377,
+      "learning_rate": 7.750939345142244e-06,
+      "loss": 0.0001,
+      "step": 3130
+    },
+    {
+      "epoch": 44.005797101449275,
+      "grad_norm": 0.0034120541531592607,
+      "learning_rate": 7.74020397208803e-06,
+      "loss": 0.2225,
+      "step": 3140
+    },
+    {
+      "epoch": 44.00676328502416,
+      "grad_norm": 0.007564049679785967,
+      "learning_rate": 7.729468599033817e-06,
+      "loss": 0.0001,
+      "step": 3150
+    },
+    {
+      "epoch": 44.00676328502416,
+      "eval_accuracy": 0.7402597402597403,
+      "eval_loss": 1.5574208498001099,
+      "eval_runtime": 15.8522,
+      "eval_samples_per_second": 4.857,
+      "eval_steps_per_second": 1.262,
+      "step": 3150
+    },
+    {
+      "epoch": 45.00096618357488,
+      "grad_norm": 0.013103250414133072,
+      "learning_rate": 7.718733225979603e-06,
+      "loss": 0.0063,
+      "step": 3160
+    },
+    {
+      "epoch": 45.001932367149756,
+      "grad_norm": 0.010216855444014072,
+      "learning_rate": 7.70799785292539e-06,
+      "loss": 0.2025,
+      "step": 3170
+    },
+    {
+      "epoch": 45.00289855072464,
+      "grad_norm": 0.004929398186504841,
+      "learning_rate": 7.697262479871176e-06,
+      "loss": 0.4529,
+      "step": 3180
+    },
+    {
+      "epoch": 45.00386473429952,
+      "grad_norm": 48.59390640258789,
+      "learning_rate": 7.686527106816962e-06,
+      "loss": 0.1994,
+      "step": 3190
+    },
+    {
+      "epoch": 45.00483091787439,
+      "grad_norm": 0.003639970440417528,
+      "learning_rate": 7.675791733762749e-06,
+      "loss": 0.0002,
+      "step": 3200
+    },
+    {
+      "epoch": 45.005797101449275,
+      "grad_norm": 0.007403469644486904,
+      "learning_rate": 7.665056360708535e-06,
+      "loss": 0.0121,
+      "step": 3210
+    },
+    {
+      "epoch": 45.00676328502416,
+      "grad_norm": 0.0292381439357996,
+      "learning_rate": 7.654320987654322e-06,
+      "loss": 0.0353,
+      "step": 3220
+    },
+    {
+      "epoch": 45.00676328502416,
+      "eval_accuracy": 0.8311688311688312,
+      "eval_loss": 1.1688188314437866,
+      "eval_runtime": 16.7449,
+      "eval_samples_per_second": 4.598,
+      "eval_steps_per_second": 1.194,
+      "step": 3220
+    },
+    {
+      "epoch": 46.00096618357488,
+      "grad_norm": 0.4787992238998413,
+      "learning_rate": 7.643585614600108e-06,
+      "loss": 0.0008,
+      "step": 3230
+    },
+    {
+      "epoch": 46.001932367149756,
+      "grad_norm": 0.006452680099755526,
+      "learning_rate": 7.632850241545895e-06,
+      "loss": 0.2781,
+      "step": 3240
+    },
+    {
+      "epoch": 46.00289855072464,
+      "grad_norm": 0.003960897680372,
+      "learning_rate": 7.622114868491681e-06,
+      "loss": 0.0366,
+      "step": 3250
+    },
+    {
+      "epoch": 46.00386473429952,
+      "grad_norm": 0.0023400746285915375,
+      "learning_rate": 7.6113794954374675e-06,
+      "loss": 0.0001,
+      "step": 3260
+    },
+    {
+      "epoch": 46.00483091787439,
+      "grad_norm": 0.015635930001735687,
+      "learning_rate": 7.600644122383254e-06,
+      "loss": 0.0001,
+      "step": 3270
+    },
+    {
+      "epoch": 46.005797101449275,
+      "grad_norm": 0.002309344243258238,
+      "learning_rate": 7.58990874932904e-06,
+      "loss": 0.0001,
+      "step": 3280
+    },
+    {
+      "epoch": 46.00676328502416,
+      "grad_norm": 0.0033812953624874353,
+      "learning_rate": 7.579173376274827e-06,
+      "loss": 0.0001,
+      "step": 3290
+    },
+    {
+      "epoch": 46.00676328502416,
+      "eval_accuracy": 0.7142857142857143,
+      "eval_loss": 1.7684170007705688,
+      "eval_runtime": 16.1588,
+      "eval_samples_per_second": 4.765,
+      "eval_steps_per_second": 1.238,
+      "step": 3290
+    },
+    {
+      "epoch": 47.00096618357488,
+      "grad_norm": 0.004736083559691906,
+      "learning_rate": 7.568438003220613e-06,
+      "loss": 0.0001,
+      "step": 3300
+    },
+    {
+      "epoch": 47.001932367149756,
+      "grad_norm": 0.0027902736328542233,
+      "learning_rate": 7.557702630166399e-06,
+      "loss": 0.0001,
+      "step": 3310
+    },
+    {
+      "epoch": 47.00289855072464,
+      "grad_norm": 0.0036582525353878736,
+      "learning_rate": 7.546967257112185e-06,
+      "loss": 0.0001,
+      "step": 3320
+    },
+    {
+      "epoch": 47.00386473429952,
+      "grad_norm": 0.002314971061423421,
+      "learning_rate": 7.536231884057972e-06,
+      "loss": 0.0001,
+      "step": 3330
+    },
+    {
+      "epoch": 47.00483091787439,
+      "grad_norm": 231.95797729492188,
+      "learning_rate": 7.525496511003758e-06,
+      "loss": 0.0304,
+      "step": 3340
+    },
+    {
+      "epoch": 47.005797101449275,
+      "grad_norm": 9.556648254394531,
+      "learning_rate": 7.514761137949545e-06,
+      "loss": 0.1891,
+      "step": 3350
+    },
+    {
+      "epoch": 47.00676328502416,
+      "grad_norm": 0.08842357248067856,
+      "learning_rate": 7.504025764895331e-06,
+      "loss": 0.0002,
+      "step": 3360
+    },
+    {
+      "epoch": 47.00676328502416,
+      "eval_accuracy": 0.7792207792207793,
+      "eval_loss": 1.3363182544708252,
+      "eval_runtime": 14.464,
+      "eval_samples_per_second": 5.324,
+      "eval_steps_per_second": 1.383,
+      "step": 3360
+    },
+    {
+      "epoch": 48.00096618357488,
+      "grad_norm": 0.0023621630389243364,
+      "learning_rate": 7.493290391841117e-06,
+      "loss": 0.0051,
+      "step": 3370
+    },
+    {
+      "epoch": 48.001932367149756,
+      "grad_norm": 0.0059842816554009914,
+      "learning_rate": 7.482555018786903e-06,
+      "loss": 0.0001,
+      "step": 3380
+    },
+    {
+      "epoch": 48.00289855072464,
+      "grad_norm": 0.016945144161581993,
+      "learning_rate": 7.47181964573269e-06,
+      "loss": 0.0014,
+      "step": 3390
+    },
+    {
+      "epoch": 48.00386473429952,
+      "grad_norm": 0.014096066355705261,
+      "learning_rate": 7.461084272678476e-06,
+      "loss": 0.214,
+      "step": 3400
+    },
+    {
+      "epoch": 48.00483091787439,
+      "grad_norm": 0.0021912576630711555,
+      "learning_rate": 7.4503488996242625e-06,
+      "loss": 0.0001,
+      "step": 3410
+    },
+    {
+      "epoch": 48.005797101449275,
+      "grad_norm": 0.00600380590185523,
+      "learning_rate": 7.439613526570049e-06,
+      "loss": 0.0001,
+      "step": 3420
+    },
+    {
+      "epoch": 48.00676328502416,
+      "grad_norm": 0.013205859810113907,
+      "learning_rate": 7.428878153515835e-06,
+      "loss": 0.1237,
+      "step": 3430
+    },
+    {
+      "epoch": 48.00676328502416,
+      "eval_accuracy": 0.7922077922077922,
+      "eval_loss": 1.2230392694473267,
+      "eval_runtime": 14.5301,
+      "eval_samples_per_second": 5.299,
+      "eval_steps_per_second": 1.376,
+      "step": 3430
+    },
+    {
+      "epoch": 49.00096618357488,
+      "grad_norm": 0.002923307241871953,
+      "learning_rate": 7.418142780461621e-06,
+      "loss": 0.0003,
+      "step": 3440
+    },
+    {
+      "epoch": 49.001932367149756,
+      "grad_norm": 0.07183127850294113,
+      "learning_rate": 7.4074074074074075e-06,
+      "loss": 0.0009,
+      "step": 3450
+    },
+    {
+      "epoch": 49.00289855072464,
+      "grad_norm": 0.01003690529614687,
+      "learning_rate": 7.396672034353194e-06,
+      "loss": 0.0001,
+      "step": 3460
+    },
+    {
+      "epoch": 49.00386473429952,
+      "grad_norm": 0.003056786023080349,
+      "learning_rate": 7.38593666129898e-06,
+      "loss": 0.0081,
+      "step": 3470
+    },
+    {
+      "epoch": 49.00483091787439,
+      "grad_norm": 0.0015192265855148435,
+      "learning_rate": 7.375201288244767e-06,
+      "loss": 0.0001,
+      "step": 3480
+    },
+    {
+      "epoch": 49.005797101449275,
+      "grad_norm": 0.005432800389826298,
+      "learning_rate": 7.364465915190554e-06,
+      "loss": 0.0001,
+      "step": 3490
+    },
+    {
+      "epoch": 49.00676328502416,
+      "grad_norm": 0.002280445536598563,
+      "learning_rate": 7.353730542136341e-06,
+      "loss": 0.0001,
+      "step": 3500
+    },
+    {
+      "epoch": 49.00676328502416,
+      "eval_accuracy": 0.7792207792207793,
+      "eval_loss": 1.4665330648422241,
+      "eval_runtime": 14.8215,
+      "eval_samples_per_second": 5.195,
+      "eval_steps_per_second": 1.349,
+      "step": 3500
+    },
+    {
+      "epoch": 50.00096618357488,
+      "grad_norm": 0.0026954906061291695,
+      "learning_rate": 7.342995169082127e-06,
+      "loss": 0.0001,
+      "step": 3510
+    },
+    {
+      "epoch": 50.001932367149756,
+      "grad_norm": 0.007547179237008095,
+      "learning_rate": 7.332259796027913e-06,
+      "loss": 0.0001,
+      "step": 3520
+    },
+    {
+      "epoch": 50.00289855072464,
+      "grad_norm": 0.0018638098845258355,
+      "learning_rate": 7.321524422973699e-06,
+      "loss": 0.0001,
+      "step": 3530
+    },
+    {
+      "epoch": 50.00386473429952,
+      "grad_norm": 0.00512617826461792,
+      "learning_rate": 7.3107890499194855e-06,
+      "loss": 0.0001,
+      "step": 3540
+    },
+    {
+      "epoch": 50.00483091787439,
+      "grad_norm": 0.001102678943425417,
+      "learning_rate": 7.300053676865272e-06,
+      "loss": 0.0001,
+      "step": 3550
+    },
+    {
+      "epoch": 50.005797101449275,
+      "grad_norm": 0.008461062796413898,
+      "learning_rate": 7.2893183038110584e-06,
+      "loss": 0.0001,
+      "step": 3560
+    },
+    {
+      "epoch": 50.00676328502416,
+      "grad_norm": 0.006839872803539038,
+      "learning_rate": 7.278582930756845e-06,
+      "loss": 0.0,
+      "step": 3570
+    },
+    {
+      "epoch": 50.00676328502416,
+      "eval_accuracy": 0.7662337662337663,
+      "eval_loss": 1.5471916198730469,
+      "eval_runtime": 14.9136,
+      "eval_samples_per_second": 5.163,
+      "eval_steps_per_second": 1.341,
+      "step": 3570
+    },
+    {
+      "epoch": 51.00096618357488,
+      "grad_norm": 0.0022453684359788895,
+      "learning_rate": 7.2678475577026305e-06,
+      "loss": 0.0001,
+      "step": 3580
+    },
+    {
+      "epoch": 51.001932367149756,
+      "grad_norm": 0.004750104621052742,
+      "learning_rate": 7.257112184648417e-06,
+      "loss": 0.0001,
+      "step": 3590
+    },
+    {
+      "epoch": 51.00289855072464,
+      "grad_norm": 0.001577241811901331,
+      "learning_rate": 7.246376811594203e-06,
+      "loss": 0.0001,
+      "step": 3600
+    },
+    {
+      "epoch": 51.00386473429952,
+      "grad_norm": 0.001378105953335762,
+      "learning_rate": 7.23564143853999e-06,
+      "loss": 0.0002,
+      "step": 3610
+    },
+    {
+      "epoch": 51.00483091787439,
+      "grad_norm": 0.007165323477238417,
+      "learning_rate": 7.224906065485776e-06,
+      "loss": 0.0001,
+      "step": 3620
+    },
+    {
+      "epoch": 51.005797101449275,
+      "grad_norm": 0.002125757047906518,
+      "learning_rate": 7.214170692431563e-06,
+      "loss": 0.1188,
+      "step": 3630
+    },
+    {
+      "epoch": 51.00676328502416,
+      "grad_norm": 1.5844531059265137,
+      "learning_rate": 7.203435319377348e-06,
+      "loss": 0.1479,
+      "step": 3640
+    },
+    {
+      "epoch": 51.00676328502416,
+      "eval_accuracy": 0.7272727272727273,
+      "eval_loss": 2.3368756771087646,
+      "eval_runtime": 14.6055,
+      "eval_samples_per_second": 5.272,
+      "eval_steps_per_second": 1.369,
+      "step": 3640
+    },
+    {
+      "epoch": 52.00096618357488,
+      "grad_norm": 0.0030117840506136417,
+      "learning_rate": 7.192699946323135e-06,
+      "loss": 0.0001,
+      "step": 3650
+    },
+    {
+      "epoch": 52.001932367149756,
+      "grad_norm": 0.0017524833092465997,
+      "learning_rate": 7.181964573268921e-06,
+      "loss": 0.0001,
+      "step": 3660
+    },
+    {
+      "epoch": 52.00289855072464,
+      "grad_norm": 0.002778457012027502,
+      "learning_rate": 7.171229200214708e-06,
+      "loss": 0.0001,
+      "step": 3670
+    },
+    {
+      "epoch": 52.00386473429952,
+      "grad_norm": 333.92315673828125,
+      "learning_rate": 7.160493827160494e-06,
+      "loss": 0.0826,
+      "step": 3680
+    },
+    {
+      "epoch": 52.00483091787439,
+      "grad_norm": 0.0024222638458013535,
+      "learning_rate": 7.149758454106281e-06,
+      "loss": 0.0297,
+      "step": 3690
+    },
+    {
+      "epoch": 52.005797101449275,
+      "grad_norm": 0.0016652902122586966,
+      "learning_rate": 7.139023081052067e-06,
+      "loss": 0.0707,
+      "step": 3700
+    },
+    {
+      "epoch": 52.00676328502416,
+      "grad_norm": 0.00755368173122406,
+      "learning_rate": 7.128287707997853e-06,
+      "loss": 0.0001,
+      "step": 3710
+    },
+    {
+      "epoch": 52.00676328502416,
+      "eval_accuracy": 0.6753246753246753,
+      "eval_loss": 2.2529079914093018,
+      "eval_runtime": 14.6501,
+      "eval_samples_per_second": 5.256,
+      "eval_steps_per_second": 1.365,
+      "step": 3710
+    },
+    {
+      "epoch": 53.00096618357488,
+      "grad_norm": 0.0013488616095855832,
+      "learning_rate": 7.117552334943641e-06,
+      "loss": 0.0001,
+      "step": 3720
+    },
+    {
+      "epoch": 53.001932367149756,
+      "grad_norm": 0.0027625716757029295,
+      "learning_rate": 7.106816961889426e-06,
+      "loss": 0.0001,
+      "step": 3730
+    },
+    {
+      "epoch": 53.00289855072464,
+      "grad_norm": 0.0010352464159950614,
+      "learning_rate": 7.096081588835213e-06,
+      "loss": 0.0,
+      "step": 3740
+    },
+    {
+      "epoch": 53.00386473429952,
+      "grad_norm": 0.002267819829285145,
+      "learning_rate": 7.085346215780999e-06,
+      "loss": 0.2287,
+      "step": 3750
+    },
+    {
+      "epoch": 53.00483091787439,
+      "grad_norm": 0.0032848615664988756,
+      "learning_rate": 7.074610842726786e-06,
+      "loss": 0.0001,
+      "step": 3760
+    },
+    {
+      "epoch": 53.005797101449275,
+      "grad_norm": 0.0024115026462823153,
+      "learning_rate": 7.063875469672572e-06,
+      "loss": 0.0001,
+      "step": 3770
+    },
+    {
+      "epoch": 53.00676328502416,
+      "grad_norm": 0.0190240778028965,
+      "learning_rate": 7.053140096618359e-06,
+      "loss": 0.1081,
+      "step": 3780
+    },
+    {
+      "epoch": 53.00676328502416,
+      "eval_accuracy": 0.7272727272727273,
+      "eval_loss": 1.4744716882705688,
+      "eval_runtime": 14.7059,
+      "eval_samples_per_second": 5.236,
+      "eval_steps_per_second": 1.36,
+      "step": 3780
+    },
+    {
+      "epoch": 54.00096618357488,
+      "grad_norm": 0.011739949695765972,
+      "learning_rate": 7.042404723564144e-06,
+      "loss": 0.0001,
+      "step": 3790
+    },
+    {
+      "epoch": 54.001932367149756,
+      "grad_norm": 0.001719196210615337,
+      "learning_rate": 7.031669350509931e-06,
+      "loss": 0.0004,
+      "step": 3800
+    },
+    {
+      "epoch": 54.00289855072464,
+      "grad_norm": 379.4192810058594,
+      "learning_rate": 7.020933977455717e-06,
+      "loss": 0.1521,
+      "step": 3810
+    },
+    {
+      "epoch": 54.00386473429952,
+      "grad_norm": 0.003941697999835014,
+      "learning_rate": 7.010198604401504e-06,
+      "loss": 0.0015,
+      "step": 3820
+    },
+    {
+      "epoch": 54.00483091787439,
+      "grad_norm": 0.002875625155866146,
+      "learning_rate": 6.99946323134729e-06,
+      "loss": 0.0561,
+      "step": 3830
+    },
+    {
+      "epoch": 54.005797101449275,
+      "grad_norm": 0.0037783801089972258,
+      "learning_rate": 6.9887278582930765e-06,
+      "loss": 0.0001,
+      "step": 3840
+    },
+    {
+      "epoch": 54.00676328502416,
+      "grad_norm": 0.0007094612810760736,
+      "learning_rate": 6.977992485238862e-06,
+      "loss": 0.0002,
+      "step": 3850
+    },
+    {
+      "epoch": 54.00676328502416,
+      "eval_accuracy": 0.7402597402597403,
+      "eval_loss": 1.581336498260498,
+      "eval_runtime": 14.5929,
+      "eval_samples_per_second": 5.277,
+      "eval_steps_per_second": 1.371,
+      "step": 3850
+    },
+    {
+      "epoch": 55.00096618357488,
+      "grad_norm": 0.0009815659141167998,
+      "learning_rate": 6.9672571121846486e-06,
+      "loss": 0.0,
+      "step": 3860
+    },
+    {
+      "epoch": 55.001932367149756,
+      "grad_norm": 0.0013574979966506362,
+      "learning_rate": 6.956521739130435e-06,
+      "loss": 0.0001,
+      "step": 3870
+    },
+    {
+      "epoch": 55.00289855072464,
+      "grad_norm": 335.1041259765625,
+      "learning_rate": 6.9457863660762215e-06,
+      "loss": 0.1636,
+      "step": 3880
+    },
+    {
+      "epoch": 55.00386473429952,
+      "grad_norm": 0.010944723151624203,
+      "learning_rate": 6.935050993022008e-06,
+      "loss": 0.0001,
+      "step": 3890
+    },
+    {
+      "epoch": 55.00483091787439,
+      "grad_norm": 0.0034687011502683163,
+      "learning_rate": 6.924315619967794e-06,
+      "loss": 0.092,
+      "step": 3900
+    },
+    {
+      "epoch": 55.005797101449275,
+      "grad_norm": 0.0015824460424482822,
+      "learning_rate": 6.913580246913581e-06,
+      "loss": 0.0001,
+      "step": 3910
+    },
+    {
+      "epoch": 55.00676328502416,
+      "grad_norm": 0.001366764772683382,
+      "learning_rate": 6.9028448738593664e-06,
+      "loss": 0.0119,
+      "step": 3920
+    },
+    {
+      "epoch": 55.00676328502416,
+      "eval_accuracy": 0.7662337662337663,
+      "eval_loss": 1.6006883382797241,
+      "eval_runtime": 14.4536,
+      "eval_samples_per_second": 5.327,
+      "eval_steps_per_second": 1.384,
+      "step": 3920
+    },
+    {
+      "epoch": 56.00096618357488,
+      "grad_norm": 0.0016056658932939172,
+      "learning_rate": 6.892109500805153e-06,
+      "loss": 0.0647,
+      "step": 3930
+    },
+    {
+      "epoch": 56.001932367149756,
+      "grad_norm": 0.0010164413833990693,
+      "learning_rate": 6.881374127750939e-06,
+      "loss": 0.0018,
+      "step": 3940
+    },
+    {
+      "epoch": 56.00289855072464,
+      "grad_norm": 0.0008678404265083373,
+      "learning_rate": 6.870638754696727e-06,
+      "loss": 0.0,
+      "step": 3950
+    },
+    {
+      "epoch": 56.00386473429952,
+      "grad_norm": 0.0008416047203354537,
+      "learning_rate": 6.859903381642513e-06,
+      "loss": 0.1129,
+      "step": 3960
+    },
+    {
+      "epoch": 56.00483091787439,
+      "grad_norm": 0.010661150328814983,
+      "learning_rate": 6.8491680085882995e-06,
+      "loss": 0.0004,
+      "step": 3970
+    },
+    {
+      "epoch": 56.005797101449275,
+      "grad_norm": 0.0018914261600002646,
+      "learning_rate": 6.838432635534086e-06,
+      "loss": 0.0442,
+      "step": 3980
+    },
+    {
+      "epoch": 56.00676328502416,
+      "grad_norm": 0.007342246826738119,
+      "learning_rate": 6.8276972624798724e-06,
+      "loss": 0.1478,
+      "step": 3990
+    },
+    {
+      "epoch": 56.00676328502416,
+      "eval_accuracy": 0.7142857142857143,
+      "eval_loss": 2.3310229778289795,
+      "eval_runtime": 14.3306,
+      "eval_samples_per_second": 5.373,
+      "eval_steps_per_second": 1.396,
+      "step": 3990
+    },
+    {
+      "epoch": 57.00096618357488,
+      "grad_norm": 0.003021875163540244,
+      "learning_rate": 6.816961889425658e-06,
+      "loss": 0.0082,
+      "step": 4000
+    },
+    {
+      "epoch": 57.001932367149756,
+      "grad_norm": 0.0016037713503465056,
+      "learning_rate": 6.8062265163714445e-06,
+      "loss": 0.0011,
+      "step": 4010
+    },
+    {
+      "epoch": 57.00289855072464,
+      "grad_norm": 0.0027177485171705484,
+      "learning_rate": 6.795491143317231e-06,
+      "loss": 0.2467,
+      "step": 4020
+    },
+    {
+      "epoch": 57.00386473429952,
+      "grad_norm": 0.0009143136558122933,
+      "learning_rate": 6.784755770263017e-06,
+      "loss": 0.0537,
+      "step": 4030
+    },
+    {
+      "epoch": 57.00483091787439,
+      "grad_norm": 0.07625051587820053,
+      "learning_rate": 6.774020397208804e-06,
+      "loss": 0.2509,
+      "step": 4040
+    },
+    {
+      "epoch": 57.005797101449275,
+      "grad_norm": 0.0012727385619655252,
+      "learning_rate": 6.76328502415459e-06,
+      "loss": 0.001,
+      "step": 4050
+    },
+    {
+      "epoch": 57.00676328502416,
+      "grad_norm": 0.009039295837283134,
+      "learning_rate": 6.752549651100376e-06,
+      "loss": 0.0001,
+      "step": 4060
+    },
+    {
+      "epoch": 57.00676328502416,
+      "eval_accuracy": 0.8051948051948052,
+      "eval_loss": 1.4787706136703491,
+      "eval_runtime": 14.4009,
+      "eval_samples_per_second": 5.347,
+      "eval_steps_per_second": 1.389,
+      "step": 4060
+    },
+    {
+      "epoch": 58.00096618357488,
+      "grad_norm": 0.001038399524986744,
+      "learning_rate": 6.741814278046162e-06,
+      "loss": 0.0001,
+      "step": 4070
+    },
+    {
+      "epoch": 58.001932367149756,
+      "grad_norm": 0.0029470643494278193,
+      "learning_rate": 6.731078904991949e-06,
+      "loss": 0.1885,
+      "step": 4080
+    },
+    {
+      "epoch": 58.00289855072464,
+      "grad_norm": 0.19380953907966614,
+      "learning_rate": 6.720343531937735e-06,
+      "loss": 0.0001,
+      "step": 4090
+    },
+    {
+      "epoch": 58.00386473429952,
+      "grad_norm": 0.0030477980617433786,
+      "learning_rate": 6.709608158883522e-06,
+      "loss": 0.0002,
+      "step": 4100
+    },
+    {
+      "epoch": 58.00483091787439,
+      "grad_norm": 0.0017886882415041327,
+      "learning_rate": 6.698872785829308e-06,
+      "loss": 0.0001,
+      "step": 4110
+    },
+    {
+      "epoch": 58.005797101449275,
+      "grad_norm": 0.004812142346054316,
+      "learning_rate": 6.688137412775095e-06,
+      "loss": 0.0001,
+      "step": 4120
+    },
+    {
+      "epoch": 58.00676328502416,
+      "grad_norm": 0.0009715890046209097,
+      "learning_rate": 6.67740203972088e-06,
+      "loss": 0.0001,
+      "step": 4130
+    },
+    {
+      "epoch": 58.00676328502416,
+      "eval_accuracy": 0.8441558441558441,
+      "eval_loss": 1.185050129890442,
+      "eval_runtime": 14.7015,
+      "eval_samples_per_second": 5.238,
+      "eval_steps_per_second": 1.36,
+      "step": 4130
+    },
+    {
+      "epoch": 59.00096618357488,
+      "grad_norm": 0.0017103978898376226,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 0.0,
+      "step": 4140
+    },
+    {
+      "epoch": 59.001932367149756,
+      "grad_norm": 0.002025015652179718,
+      "learning_rate": 6.655931293612453e-06,
+      "loss": 0.0,
+      "step": 4150
+    },
+    {
+      "epoch": 59.00289855072464,
+      "grad_norm": 0.01431421097368002,
+      "learning_rate": 6.6451959205582395e-06,
+      "loss": 0.0001,
+      "step": 4160
+    },
+    {
+      "epoch": 59.00386473429952,
+      "grad_norm": 0.0034724946599453688,
+      "learning_rate": 6.634460547504026e-06,
+      "loss": 0.0001,
+      "step": 4170
+    },
+    {
+      "epoch": 59.00483091787439,
+      "grad_norm": 0.0009903705213218927,
+      "learning_rate": 6.623725174449813e-06,
+      "loss": 0.0,
+      "step": 4180
+    },
+    {
+      "epoch": 59.005797101449275,
+      "grad_norm": 0.001704953727312386,
+      "learning_rate": 6.6129898013956e-06,
+      "loss": 0.0,
+      "step": 4190
+    },
+    {
+      "epoch": 59.00676328502416,
+      "grad_norm": 0.001701401430182159,
+      "learning_rate": 6.602254428341386e-06,
+      "loss": 0.0001,
+      "step": 4200
+    },
+    {
+      "epoch": 59.00676328502416,
+      "eval_accuracy": 0.8571428571428571,
+      "eval_loss": 1.1919652223587036,
+      "eval_runtime": 15.4843,
+      "eval_samples_per_second": 4.973,
+      "eval_steps_per_second": 1.292,
+      "step": 4200
+    },
+    {
+      "epoch": 60.00096618357488,
+      "grad_norm": 0.0020329791586846113,
+      "learning_rate": 6.591519055287172e-06,
+      "loss": 0.0,
+      "step": 4210
+    },
+    {
+      "epoch": 60.001932367149756,
+      "grad_norm": 0.002356892451643944,
+      "learning_rate": 6.580783682232958e-06,
+      "loss": 0.0,
+      "step": 4220
+    },
+    {
+      "epoch": 60.00289855072464,
+      "grad_norm": 0.0017729549435898662,
+      "learning_rate": 6.570048309178745e-06,
+      "loss": 0.0,
+      "step": 4230
+    },
+    {
+      "epoch": 60.00386473429952,
+      "grad_norm": 0.006136258598417044,
+      "learning_rate": 6.559312936124531e-06,
+      "loss": 0.0,
+      "step": 4240
+    },
+    {
+      "epoch": 60.00483091787439,
+      "grad_norm": 0.0009840623242780566,
+      "learning_rate": 6.548577563070318e-06,
+      "loss": 0.0,
+      "step": 4250
+    },
+    {
+      "epoch": 60.005797101449275,
+      "grad_norm": 0.014522840268909931,
+      "learning_rate": 6.537842190016104e-06,
+      "loss": 0.0,
+      "step": 4260
+    },
+    {
+      "epoch": 60.00676328502416,
+      "grad_norm": 0.0010055291932076216,
+      "learning_rate": 6.52710681696189e-06,
+      "loss": 0.0904,
+      "step": 4270
+    },
+    {
+      "epoch": 60.00676328502416,
+      "eval_accuracy": 0.8311688311688312,
+      "eval_loss": 1.1857966184616089,
+      "eval_runtime": 15.6797,
+      "eval_samples_per_second": 4.911,
+      "eval_steps_per_second": 1.276,
+      "step": 4270
+    },
+    {
+      "epoch": 61.00096618357488,
+      "grad_norm": 0.0076795658096671104,
+      "learning_rate": 6.516371443907676e-06,
+      "loss": 0.0931,
+      "step": 4280
+    },
+    {
+      "epoch": 61.001932367149756,
+      "grad_norm": 0.0016425478970631957,
+      "learning_rate": 6.5056360708534626e-06,
+      "loss": 0.0,
+      "step": 4290
+    },
+    {
+      "epoch": 61.00289855072464,
+      "grad_norm": 0.0013257049722597003,
+      "learning_rate": 6.494900697799249e-06,
+      "loss": 0.0002,
+      "step": 4300
+    },
+    {
+      "epoch": 61.00386473429952,
+      "grad_norm": 0.0032681403681635857,
+      "learning_rate": 6.4841653247450355e-06,
+      "loss": 0.0004,
+      "step": 4310
+    },
+    {
+      "epoch": 61.00483091787439,
+      "grad_norm": 355.2313232421875,
+      "learning_rate": 6.473429951690822e-06,
+      "loss": 0.1449,
+      "step": 4320
+    },
+    {
+      "epoch": 61.005797101449275,
+      "grad_norm": 0.0017356682801619172,
+      "learning_rate": 6.462694578636608e-06,
+      "loss": 0.0002,
+      "step": 4330
+    },
+    {
+      "epoch": 61.00676328502416,
+      "grad_norm": 0.007396780885756016,
+      "learning_rate": 6.451959205582394e-06,
+      "loss": 0.0001,
+      "step": 4340
+    },
+    {
+      "epoch": 61.00676328502416,
+      "eval_accuracy": 0.7662337662337663,
+      "eval_loss": 1.4534144401550293,
+      "eval_runtime": 15.1151,
+      "eval_samples_per_second": 5.094,
+      "eval_steps_per_second": 1.323,
+      "step": 4340
+    },
+    {
+      "epoch": 62.00096618357488,
+      "grad_norm": 0.0020279129967093468,
+      "learning_rate": 6.44122383252818e-06,
+      "loss": 0.0001,
+      "step": 4350
+    },
+    {
+      "epoch": 62.001932367149756,
+      "grad_norm": 0.0022937143221497536,
+      "learning_rate": 6.430488459473967e-06,
+      "loss": 0.0,
+      "step": 4360
+    },
+    {
+      "epoch": 62.00289855072464,
+      "grad_norm": 0.0018241552170366049,
+      "learning_rate": 6.419753086419753e-06,
+      "loss": 0.0001,
+      "step": 4370
+    },
+    {
+      "epoch": 62.00386473429952,
+      "grad_norm": 0.0027427878230810165,
+      "learning_rate": 6.40901771336554e-06,
+      "loss": 0.0008,
+      "step": 4380
+    },
+    {
+      "epoch": 62.00483091787439,
+      "grad_norm": 159.2539520263672,
+      "learning_rate": 6.398282340311326e-06,
+      "loss": 0.2086,
+      "step": 4390
+    },
+    {
+      "epoch": 62.005797101449275,
+      "grad_norm": 0.0029428687412291765,
+      "learning_rate": 6.387546967257112e-06,
+      "loss": 0.0693,
+      "step": 4400
+    },
+    {
+      "epoch": 62.00676328502416,
+      "grad_norm": 0.0011164512252435088,
+      "learning_rate": 6.376811594202898e-06,
+      "loss": 0.0017,
+      "step": 4410
+    },
+    {
+      "epoch": 62.00676328502416,
+      "eval_accuracy": 0.7792207792207793,
+      "eval_loss": 1.6715654134750366,
+      "eval_runtime": 14.8229,
+      "eval_samples_per_second": 5.195,
+      "eval_steps_per_second": 1.349,
+      "step": 4410
+    },
+    {
+      "epoch": 63.00096618357488,
+      "grad_norm": 0.36577680706977844,
+      "learning_rate": 6.3660762211486856e-06,
+      "loss": 0.0004,
+      "step": 4420
+    },
+    {
+      "epoch": 63.001932367149756,
+      "grad_norm": 0.0026161600835621357,
+      "learning_rate": 6.355340848094472e-06,
+      "loss": 0.4853,
+      "step": 4430
+    },
+    {
+      "epoch": 63.00289855072464,
+      "grad_norm": 0.0019047785317525268,
+      "learning_rate": 6.3446054750402585e-06,
+      "loss": 0.04,
+      "step": 4440
+    },
+    {
+      "epoch": 63.00386473429952,
+      "grad_norm": 0.00939449667930603,
+      "learning_rate": 6.333870101986045e-06,
+      "loss": 0.0151,
+      "step": 4450
+    },
+    {
+      "epoch": 63.00483091787439,
+      "grad_norm": 0.0031830815132707357,
+      "learning_rate": 6.323134728931831e-06,
+      "loss": 0.0005,
+      "step": 4460
+    },
+    {
+      "epoch": 63.005797101449275,
+      "grad_norm": 0.0016728241462260485,
+      "learning_rate": 6.312399355877618e-06,
+      "loss": 0.0151,
+      "step": 4470
+    },
+    {
+      "epoch": 63.00676328502416,
+      "grad_norm": 0.03096197545528412,
+      "learning_rate": 6.301663982823404e-06,
+      "loss": 0.0001,
+      "step": 4480
+    },
+    {
+      "epoch": 63.00676328502416,
+      "eval_accuracy": 0.6883116883116883,
+      "eval_loss": 2.201735019683838,
+      "eval_runtime": 15.4263,
+      "eval_samples_per_second": 4.991,
+      "eval_steps_per_second": 1.296,
+      "step": 4480
+    },
+    {
+      "epoch": 64.00096618357487,
+      "grad_norm": 0.005010953638702631,
+      "learning_rate": 6.29092860976919e-06,
+      "loss": 0.192,
+      "step": 4490
+    },
+    {
+      "epoch": 64.00193236714976,
+      "grad_norm": 0.0011758300242945552,
+      "learning_rate": 6.280193236714976e-06,
+      "loss": 0.0159,
+      "step": 4500
+    },
+    {
+      "epoch": 64.00289855072464,
+      "grad_norm": 0.010257155634462833,
+      "learning_rate": 6.269457863660763e-06,
+      "loss": 0.2044,
+      "step": 4510
+    },
+    {
+      "epoch": 64.00386473429951,
+      "grad_norm": 0.0021658441983163357,
+      "learning_rate": 6.258722490606549e-06,
+      "loss": 0.0001,
+      "step": 4520
+    },
+    {
+      "epoch": 64.0048309178744,
+      "grad_norm": 0.03462962433695793,
+      "learning_rate": 6.247987117552336e-06,
+      "loss": 0.0682,
+      "step": 4530
+    },
+    {
+      "epoch": 64.00579710144928,
+      "grad_norm": 0.020629707723855972,
+      "learning_rate": 6.237251744498122e-06,
+      "loss": 0.3837,
+      "step": 4540
+    },
+    {
+      "epoch": 64.00676328502415,
+      "grad_norm": 0.1369558870792389,
+      "learning_rate": 6.226516371443908e-06,
+      "loss": 0.3407,
+      "step": 4550
+    },
+    {
+      "epoch": 64.00676328502415,
+      "eval_accuracy": 0.8051948051948052,
+      "eval_loss": 1.2423864603042603,
+      "eval_runtime": 15.4244,
+      "eval_samples_per_second": 4.992,
+      "eval_steps_per_second": 1.297,
+      "step": 4550
+    },
+    {
+      "epoch": 65.00096618357487,
+      "grad_norm": 90.10851287841797,
+      "learning_rate": 6.215780998389694e-06,
+      "loss": 0.016,
+      "step": 4560
+    },
+    {
+      "epoch": 65.00193236714976,
+      "grad_norm": 0.003571063280105591,
+      "learning_rate": 6.205045625335481e-06,
+      "loss": 0.0371,
+      "step": 4570
+    },
+    {
+      "epoch": 65.00289855072464,
+      "grad_norm": 0.00869289506226778,
+      "learning_rate": 6.194310252281267e-06,
+      "loss": 0.0001,
+      "step": 4580
+    },
+    {
+      "epoch": 65.00386473429951,
+      "grad_norm": 0.042137209326028824,
+      "learning_rate": 6.1835748792270535e-06,
+      "loss": 0.0098,
+      "step": 4590
+    },
+    {
+      "epoch": 65.0048309178744,
+      "grad_norm": 1.3429404497146606,
+      "learning_rate": 6.17283950617284e-06,
+      "loss": 0.0002,
+      "step": 4600
+    },
+    {
+      "epoch": 65.00579710144928,
+      "grad_norm": 0.009308219887316227,
+      "learning_rate": 6.162104133118626e-06,
+      "loss": 0.2129,
+      "step": 4610
+    },
+    {
+      "epoch": 65.00676328502415,
+      "grad_norm": 0.010884756222367287,
+      "learning_rate": 6.151368760064412e-06,
+      "loss": 0.0001,
+      "step": 4620
+    },
+    {
+      "epoch": 65.00676328502415,
+      "eval_accuracy": 0.7792207792207793,
+      "eval_loss": 1.5786070823669434,
+      "eval_runtime": 14.9171,
+      "eval_samples_per_second": 5.162,
+      "eval_steps_per_second": 1.341,
+      "step": 4620
+    },
+    {
+      "epoch": 66.00096618357487,
+      "grad_norm": 0.001252433517947793,
+      "learning_rate": 6.1406333870101985e-06,
+      "loss": 0.0031,
+      "step": 4630
+    },
+    {
+      "epoch": 66.00193236714976,
+      "grad_norm": 0.0008501612464897335,
+      "learning_rate": 6.129898013955985e-06,
+      "loss": 0.0051,
+      "step": 4640
+    },
+    {
+      "epoch": 66.00289855072464,
+      "grad_norm": 0.0021611701231449842,
+      "learning_rate": 6.119162640901772e-06,
+      "loss": 0.0001,
+      "step": 4650
+    },
+    {
+      "epoch": 66.00386473429951,
+      "grad_norm": 0.001628038240596652,
+      "learning_rate": 6.108427267847559e-06,
+      "loss": 0.0226,
+      "step": 4660
+    },
+    {
+      "epoch": 66.0048309178744,
+      "grad_norm": 0.003019214142113924,
+      "learning_rate": 6.097691894793345e-06,
+      "loss": 0.0001,
+      "step": 4670
+    },
+    {
+      "epoch": 66.00579710144928,
+      "grad_norm": 0.006033017765730619,
+      "learning_rate": 6.086956521739132e-06,
+      "loss": 0.0001,
+      "step": 4680
+    },
+    {
+      "epoch": 66.00676328502415,
+      "grad_norm": 0.0029068191070109606,
+      "learning_rate": 6.076221148684918e-06,
+      "loss": 0.0002,
+      "step": 4690
+    },
+    {
+      "epoch": 66.00676328502415,
+      "eval_accuracy": 0.8181818181818182,
+      "eval_loss": 1.3379225730895996,
+      "eval_runtime": 15.4001,
+      "eval_samples_per_second": 5.0,
+      "eval_steps_per_second": 1.299,
+      "step": 4690
+    },
+    {
+      "epoch": 67.00096618357487,
+      "grad_norm": 0.0017993523506447673,
+      "learning_rate": 6.065485775630704e-06,
+      "loss": 0.0,
+      "step": 4700
+    },
+    {
+      "epoch": 67.00193236714976,
+      "grad_norm": 0.020115720108151436,
+      "learning_rate": 6.05475040257649e-06,
+      "loss": 0.0001,
+      "step": 4710
+    },
+    {
+      "epoch": 67.00289855072464,
+      "grad_norm": 1.1940608024597168,
+      "learning_rate": 6.0440150295222766e-06,
+      "loss": 0.0001,
+      "step": 4720
+    },
+    {
+      "epoch": 67.00386473429951,
+      "grad_norm": 0.0011647121282294393,
+      "learning_rate": 6.033279656468063e-06,
+      "loss": 0.0,
+      "step": 4730
+    },
+    {
+      "epoch": 67.0048309178744,
+      "grad_norm": 0.004334421828389168,
+      "learning_rate": 6.0225442834138495e-06,
+      "loss": 0.0,
+      "step": 4740
+    },
+    {
+      "epoch": 67.00579710144928,
+      "grad_norm": 0.0025098167825490236,
+      "learning_rate": 6.011808910359636e-06,
+      "loss": 0.0013,
+      "step": 4750
+    },
+    {
+      "epoch": 67.00676328502415,
+      "grad_norm": 0.0017039639642462134,
+      "learning_rate": 6.0010735373054215e-06,
+      "loss": 0.0005,
+      "step": 4760
+    },
+    {
+      "epoch": 67.00676328502415,
+      "eval_accuracy": 0.8701298701298701,
+      "eval_loss": 1.1517488956451416,
+      "eval_runtime": 14.5324,
+      "eval_samples_per_second": 5.299,
+      "eval_steps_per_second": 1.376,
+      "step": 4760
+    },
+    {
+      "epoch": 68.00096618357487,
+      "grad_norm": 0.0010669537587091327,
+      "learning_rate": 5.990338164251208e-06,
+      "loss": 0.0001,
+      "step": 4770
+    },
+    {
+      "epoch": 68.00193236714976,
+      "grad_norm": 0.002076763892546296,
+      "learning_rate": 5.979602791196994e-06,
+      "loss": 0.0,
+      "step": 4780
+    },
+    {
+      "epoch": 68.00289855072464,
+      "grad_norm": 0.0011606409680098295,
+      "learning_rate": 5.968867418142781e-06,
+      "loss": 0.0,
+      "step": 4790
+    },
+    {
+      "epoch": 68.00386473429951,
+      "grad_norm": 0.0014143381267786026,
+      "learning_rate": 5.958132045088567e-06,
+      "loss": 0.0,
+      "step": 4800
+    },
+    {
+      "epoch": 68.0048309178744,
+      "grad_norm": 0.0015609278343617916,
+      "learning_rate": 5.947396672034354e-06,
+      "loss": 0.0001,
+      "step": 4810
+    },
+    {
+      "epoch": 68.00579710144928,
+      "grad_norm": 0.002568858675658703,
+      "learning_rate": 5.936661298980139e-06,
+      "loss": 0.0,
+      "step": 4820
+    },
+    {
+      "epoch": 68.00676328502415,
+      "grad_norm": 0.0010036162566393614,
+      "learning_rate": 5.925925925925926e-06,
+      "loss": 0.0,
+      "step": 4830
+    },
+    {
+      "epoch": 68.00676328502415,
+      "eval_accuracy": 0.7792207792207793,
+      "eval_loss": 1.529435396194458,
+      "eval_runtime": 14.6615,
+      "eval_samples_per_second": 5.252,
+      "eval_steps_per_second": 1.364,
+      "step": 4830
+    },
+    {
+      "epoch": 69.00096618357487,
+      "grad_norm": 0.0011304821819067001,
+      "learning_rate": 5.915190552871712e-06,
+      "loss": 0.0,
+      "step": 4840
+    },
+    {
+      "epoch": 69.00193236714976,
+      "grad_norm": 0.001226439024321735,
+      "learning_rate": 5.904455179817499e-06,
+      "loss": 0.0,
+      "step": 4850
+    },
+    {
+      "epoch": 69.00289855072464,
+      "grad_norm": 0.000802580441813916,
+      "learning_rate": 5.893719806763285e-06,
+      "loss": 0.0,
+      "step": 4860
+    },
+    {
+      "epoch": 69.00386473429951,
+      "grad_norm": 0.0015111935790628195,
+      "learning_rate": 5.882984433709072e-06,
+      "loss": 0.0701,
+      "step": 4870
+    },
+    {
+      "epoch": 69.0048309178744,
+      "grad_norm": 0.0005540599231608212,
+      "learning_rate": 5.872249060654859e-06,
+      "loss": 0.0,
+      "step": 4880
+    },
+    {
+      "epoch": 69.00579710144928,
+      "grad_norm": 0.000656483112834394,
+      "learning_rate": 5.861513687600645e-06,
+      "loss": 0.2159,
+      "step": 4890
+    },
+    {
+      "epoch": 69.00676328502415,
+      "grad_norm": 0.0013909523840993643,
+      "learning_rate": 5.850778314546432e-06,
+      "loss": 0.0,
+      "step": 4900
+    },
+    {
+      "epoch": 69.00676328502415,
+      "eval_accuracy": 0.6883116883116883,
+      "eval_loss": 2.438077449798584,
+      "eval_runtime": 14.6754,
+      "eval_samples_per_second": 5.247,
+      "eval_steps_per_second": 1.363,
+      "step": 4900
+    },
+    {
+      "epoch": 70.00096618357487,
+      "grad_norm": 0.0017317779129371047,
+      "learning_rate": 5.840042941492217e-06,
+      "loss": 0.094,
+      "step": 4910
+    },
+    {
+      "epoch": 70.00193236714976,
+      "grad_norm": 0.0013916000025346875,
+      "learning_rate": 5.829307568438004e-06,
+      "loss": 0.0001,
+      "step": 4920
+    },
+    {
+      "epoch": 70.00289855072464,
+      "grad_norm": 0.027323855087161064,
+      "learning_rate": 5.81857219538379e-06,
+      "loss": 0.0001,
+      "step": 4930
+    },
+    {
+      "epoch": 70.00386473429951,
+      "grad_norm": 0.0019869667012244463,
+      "learning_rate": 5.807836822329577e-06,
+      "loss": 0.0001,
+      "step": 4940
+    },
+    {
+      "epoch": 70.0048309178744,
+      "grad_norm": 0.0005596183473244309,
+      "learning_rate": 5.797101449275363e-06,
+      "loss": 0.0002,
+      "step": 4950
+    },
+    {
+      "epoch": 70.00579710144928,
+      "grad_norm": 0.0012641664361581206,
+      "learning_rate": 5.78636607622115e-06,
+      "loss": 0.0,
+      "step": 4960
+    },
+    {
+      "epoch": 70.00676328502415,
+      "grad_norm": 0.0033598807640373707,
+      "learning_rate": 5.775630703166935e-06,
+      "loss": 0.0032,
+      "step": 4970
+    },
+    {
+      "epoch": 70.00676328502415,
+      "eval_accuracy": 0.7532467532467533,
+      "eval_loss": 1.7951840162277222,
+      "eval_runtime": 17.871,
+      "eval_samples_per_second": 4.309,
+      "eval_steps_per_second": 1.119,
+      "step": 4970
+    },
+    {
+      "epoch": 71.00096618357487,
+      "grad_norm": 0.0025404111947864294,
+      "learning_rate": 5.764895330112722e-06,
+      "loss": 0.0,
+      "step": 4980
+    },
+    {
+      "epoch": 71.00193236714976,
+      "grad_norm": 0.0016201400430873036,
+      "learning_rate": 5.754159957058508e-06,
+      "loss": 0.0,
+      "step": 4990
+    },
+    {
+      "epoch": 71.00289855072464,
+      "grad_norm": 0.000684636237565428,
+      "learning_rate": 5.743424584004295e-06,
+      "loss": 0.0,
+      "step": 5000
+    },
+    {
+      "epoch": 71.00386473429951,
+      "grad_norm": 0.0006000622524879873,
+      "learning_rate": 5.732689210950081e-06,
+      "loss": 0.0,
+      "step": 5010
+    },
+    {
+      "epoch": 71.0048309178744,
+      "grad_norm": 0.000872518983669579,
+      "learning_rate": 5.7219538378958675e-06,
+      "loss": 0.0,
+      "step": 5020
+    },
+    {
+      "epoch": 71.00579710144928,
+      "grad_norm": 102.48182678222656,
+      "learning_rate": 5.711218464841653e-06,
+      "loss": 0.0035,
+      "step": 5030
+    },
+    {
+      "epoch": 71.00676328502415,
+      "grad_norm": 0.0009674491011537611,
+      "learning_rate": 5.70048309178744e-06,
+      "loss": 0.0,
+      "step": 5040
+    },
+    {
+      "epoch": 71.00676328502415,
+      "eval_accuracy": 0.6753246753246753,
+      "eval_loss": 3.0253195762634277,
+      "eval_runtime": 30.56,
+      "eval_samples_per_second": 2.52,
+      "eval_steps_per_second": 0.654,
+      "step": 5040
+    },
+    {
+      "epoch": 72.00096618357487,
+      "grad_norm": 641.0316162109375,
+      "learning_rate": 5.689747718733226e-06,
+      "loss": 0.14,
+      "step": 5050
+    },
+    {
+      "epoch": 72.00193236714976,
+      "grad_norm": 0.0011702714255079627,
+      "learning_rate": 5.6790123456790125e-06,
+      "loss": 0.0,
+      "step": 5060
+    },
+    {
+      "epoch": 72.00289855072464,
+      "grad_norm": 0.0009515401907265186,
+      "learning_rate": 5.668276972624799e-06,
+      "loss": 0.0,
+      "step": 5070
+    },
+    {
+      "epoch": 72.00386473429951,
+      "grad_norm": 0.0014254461275413632,
+      "learning_rate": 5.657541599570585e-06,
+      "loss": 0.2006,
+      "step": 5080
+    },
+    {
+      "epoch": 72.0048309178744,
+      "grad_norm": 0.0018207349348813295,
+      "learning_rate": 5.646806226516372e-06,
+      "loss": 0.0,
+      "step": 5090
+    },
+    {
+      "epoch": 72.00579710144928,
+      "grad_norm": 0.01077065709978342,
+      "learning_rate": 5.6360708534621574e-06,
+      "loss": 0.0001,
+      "step": 5100
+    },
+    {
+      "epoch": 72.00676328502415,
+      "grad_norm": 0.0008518044487573206,
+      "learning_rate": 5.625335480407944e-06,
+      "loss": 0.214,
+      "step": 5110
+    },
+    {
+      "epoch": 72.00676328502415,
+      "eval_accuracy": 0.7142857142857143,
+      "eval_loss": 1.932701826095581,
+      "eval_runtime": 14.6424,
+      "eval_samples_per_second": 5.259,
+      "eval_steps_per_second": 1.366,
+      "step": 5110
+    },
+    {
+      "epoch": 73.00096618357487,
+      "grad_norm": 0.002738331910222769,
+      "learning_rate": 5.614600107353731e-06,
+      "loss": 0.0,
+      "step": 5120
+    },
+    {
+      "epoch": 73.00193236714976,
+      "grad_norm": 0.0025839614681899548,
+      "learning_rate": 5.603864734299518e-06,
+      "loss": 0.0,
+      "step": 5130
+    },
+    {
+      "epoch": 73.00289855072464,
+      "grad_norm": 0.0006668622372671962,
+      "learning_rate": 5.593129361245304e-06,
+      "loss": 0.0745,
+      "step": 5140
+    },
+    {
+      "epoch": 73.00386473429951,
+      "grad_norm": 0.0026406822726130486,
+      "learning_rate": 5.5823939881910905e-06,
+      "loss": 0.0001,
+      "step": 5150
+    },
+    {
+      "epoch": 73.0048309178744,
+      "grad_norm": 0.01901336945593357,
+      "learning_rate": 5.571658615136877e-06,
+      "loss": 0.0001,
+      "step": 5160
+    },
+    {
+      "epoch": 73.00579710144928,
+      "grad_norm": 0.04490555077791214,
+      "learning_rate": 5.5609232420826634e-06,
+      "loss": 0.0001,
+      "step": 5170
+    },
+    {
+      "epoch": 73.00676328502415,
+      "grad_norm": 0.0009415218373760581,
+      "learning_rate": 5.550187869028449e-06,
+      "loss": 0.0,
+      "step": 5180
+    },
+    {
+      "epoch": 73.00676328502415,
+      "eval_accuracy": 0.7532467532467533,
+      "eval_loss": 2.0236287117004395,
+      "eval_runtime": 15.6052,
+      "eval_samples_per_second": 4.934,
+      "eval_steps_per_second": 1.282,
+      "step": 5180
+    },
+    {
+      "epoch": 74.00096618357487,
+      "grad_norm": 0.0010939829517155886,
+      "learning_rate": 5.5394524959742355e-06,
+      "loss": 0.0,
+      "step": 5190
+    },
+    {
+      "epoch": 74.00193236714976,
+      "grad_norm": 0.000889449380338192,
+      "learning_rate": 5.528717122920022e-06,
+      "loss": 0.0,
+      "step": 5200
+    },
+    {
+      "epoch": 74.00289855072464,
+      "grad_norm": 0.0009812179487198591,
+      "learning_rate": 5.517981749865808e-06,
+      "loss": 0.0,
+      "step": 5210
+    },
+    {
+      "epoch": 74.00386473429951,
+      "grad_norm": 0.000881813233718276,
+      "learning_rate": 5.507246376811595e-06,
+      "loss": 0.0,
+      "step": 5220
+    },
+    {
+      "epoch": 74.0048309178744,
+      "grad_norm": 0.0004946072003804147,
+      "learning_rate": 5.496511003757381e-06,
+      "loss": 0.0,
+      "step": 5230
+    },
+    {
+      "epoch": 74.00579710144928,
+      "grad_norm": 0.000799790199380368,
+      "learning_rate": 5.485775630703167e-06,
+      "loss": 0.0,
+      "step": 5240
+    },
+    {
+      "epoch": 74.00676328502415,
+      "grad_norm": 0.002813389990478754,
+      "learning_rate": 5.475040257648953e-06,
+      "loss": 0.0,
+      "step": 5250
+    },
+    {
+      "epoch": 74.00676328502415,
+      "eval_accuracy": 0.7662337662337663,
+      "eval_loss": 1.907597303390503,
+      "eval_runtime": 15.6164,
+      "eval_samples_per_second": 4.931,
+      "eval_steps_per_second": 1.281,
+      "step": 5250
+    },
+    {
+      "epoch": 75.00096618357487,
+      "grad_norm": 0.0011628314387053251,
+      "learning_rate": 5.46430488459474e-06,
+      "loss": 0.0,
+      "step": 5260
+    },
+    {
+      "epoch": 75.00193236714976,
+      "grad_norm": 0.037489522248506546,
+      "learning_rate": 5.453569511540526e-06,
+      "loss": 0.0,
+      "step": 5270
+    },
+    {
+      "epoch": 75.00289855072464,
+      "grad_norm": 0.0011903907870873809,
+      "learning_rate": 5.442834138486313e-06,
+      "loss": 0.0,
+      "step": 5280
+    },
+    {
+      "epoch": 75.00386473429951,
+      "grad_norm": 0.0006152648711577058,
+      "learning_rate": 5.432098765432099e-06,
+      "loss": 0.0,
+      "step": 5290
+    },
+    {
+      "epoch": 75.0048309178744,
+      "grad_norm": 0.00442465441301465,
+      "learning_rate": 5.421363392377886e-06,
+      "loss": 0.0,
+      "step": 5300
+    },
+    {
+      "epoch": 75.00579710144928,
+      "grad_norm": 0.0028682094998657703,
+      "learning_rate": 5.410628019323671e-06,
+      "loss": 0.0003,
+      "step": 5310
+    },
+    {
+      "epoch": 75.00676328502415,
+      "grad_norm": 0.0016861088806763291,
+      "learning_rate": 5.399892646269458e-06,
+      "loss": 0.0,
+      "step": 5320
+    },
+    {
+      "epoch": 75.00676328502415,
+      "eval_accuracy": 0.8051948051948052,
+      "eval_loss": 1.707045316696167,
+      "eval_runtime": 14.9212,
+      "eval_samples_per_second": 5.16,
+      "eval_steps_per_second": 1.34,
+      "step": 5320
+    },
+    {
+      "epoch": 76.00096618357487,
+      "grad_norm": 0.0007070906576700509,
+      "learning_rate": 5.389157273215244e-06,
+      "loss": 0.0002,
+      "step": 5330
+    },
+    {
+      "epoch": 76.00193236714976,
+      "grad_norm": 0.0004906499525532126,
+      "learning_rate": 5.3784219001610306e-06,
+      "loss": 0.0,
+      "step": 5340
+    },
+    {
+      "epoch": 76.00289855072464,
+      "grad_norm": 0.0011014473857358098,
+      "learning_rate": 5.367686527106818e-06,
+      "loss": 0.0,
+      "step": 5350
+    },
+    {
+      "epoch": 76.00386473429951,
+      "grad_norm": 0.0008452738984487951,
+      "learning_rate": 5.356951154052604e-06,
+      "loss": 0.0002,
+      "step": 5360
+    },
+    {
+      "epoch": 76.0048309178744,
+      "grad_norm": 0.0014290370745584369,
+      "learning_rate": 5.346215780998391e-06,
+      "loss": 0.0,
+      "step": 5370
+    },
+    {
+      "epoch": 76.00579710144928,
+      "grad_norm": 0.00044006307143718004,
+      "learning_rate": 5.335480407944177e-06,
+      "loss": 0.0,
+      "step": 5380
+    },
+    {
+      "epoch": 76.00676328502415,
+      "grad_norm": 0.0015216703759506345,
+      "learning_rate": 5.324745034889963e-06,
+      "loss": 0.0003,
+      "step": 5390
+    },
+    {
+      "epoch": 76.00676328502415,
+      "eval_accuracy": 0.7532467532467533,
+      "eval_loss": 1.8620779514312744,
+      "eval_runtime": 14.639,
+      "eval_samples_per_second": 5.26,
+      "eval_steps_per_second": 1.366,
+      "step": 5390
+    },
+    {
+      "epoch": 77.00096618357487,
+      "grad_norm": 0.006501473020762205,
+      "learning_rate": 5.314009661835749e-06,
+      "loss": 0.0,
+      "step": 5400
+    },
+    {
+      "epoch": 77.00193236714976,
+      "grad_norm": 0.006775288842618465,
+      "learning_rate": 5.303274288781536e-06,
+      "loss": 0.0,
+      "step": 5410
+    },
+    {
+      "epoch": 77.00289855072464,
+      "grad_norm": 0.002048175549134612,
+      "learning_rate": 5.292538915727322e-06,
+      "loss": 0.0,
+      "step": 5420
+    },
+    {
+      "epoch": 77.00386473429951,
+      "grad_norm": 0.0005493107601068914,
+      "learning_rate": 5.281803542673109e-06,
+      "loss": 0.0,
+      "step": 5430
+    },
+    {
+      "epoch": 77.0048309178744,
+      "grad_norm": 0.004135518334805965,
+      "learning_rate": 5.271068169618895e-06,
+      "loss": 0.0,
+      "step": 5440
+    },
+    {
+      "epoch": 77.00579710144928,
+      "grad_norm": 0.0032074879854917526,
+      "learning_rate": 5.2603327965646815e-06,
+      "loss": 0.0,
+      "step": 5450
+    },
+    {
+      "epoch": 77.00676328502415,
+      "grad_norm": 0.0005827154964208603,
+      "learning_rate": 5.249597423510467e-06,
+      "loss": 0.0,
+      "step": 5460
+    },
+    {
+      "epoch": 77.00676328502415,
+      "eval_accuracy": 0.7662337662337663,
+      "eval_loss": 1.8846988677978516,
+      "eval_runtime": 14.2136,
+      "eval_samples_per_second": 5.417,
+      "eval_steps_per_second": 1.407,
+      "step": 5460
+    },
+    {
+      "epoch": 78.00096618357487,
+      "grad_norm": 0.0004196877998765558,
+      "learning_rate": 5.2388620504562536e-06,
+      "loss": 0.0,
+      "step": 5470
+    },
+    {
+      "epoch": 78.00193236714976,
+      "grad_norm": 0.00211125984787941,
+      "learning_rate": 5.22812667740204e-06,
+      "loss": 0.0,
+      "step": 5480
+    },
+    {
+      "epoch": 78.00289855072464,
+      "grad_norm": 0.00045795412734150887,
+      "learning_rate": 5.2173913043478265e-06,
+      "loss": 0.0,
+      "step": 5490
+    },
+    {
+      "epoch": 78.00386473429951,
+      "grad_norm": 0.0019368949579074979,
+      "learning_rate": 5.206655931293613e-06,
+      "loss": 0.0,
+      "step": 5500
+    },
+    {
+      "epoch": 78.0048309178744,
+      "grad_norm": 0.0030042598955333233,
+      "learning_rate": 5.195920558239399e-06,
+      "loss": 0.0,
+      "step": 5510
+    },
+    {
+      "epoch": 78.00579710144928,
+      "grad_norm": 0.017981361597776413,
+      "learning_rate": 5.185185185185185e-06,
+      "loss": 0.0,
+      "step": 5520
+    },
+    {
+      "epoch": 78.00676328502415,
+      "grad_norm": 0.0014057859079912305,
+      "learning_rate": 5.1744498121309714e-06,
+      "loss": 0.0,
+      "step": 5530
+    },
+    {
+      "epoch": 78.00676328502415,
+      "eval_accuracy": 0.7662337662337663,
+      "eval_loss": 1.8880295753479004,
+      "eval_runtime": 14.1273,
+      "eval_samples_per_second": 5.45,
+      "eval_steps_per_second": 1.416,
+      "step": 5530
+    },
+    {
+      "epoch": 79.00096618357487,
+      "grad_norm": 0.0005187008646316826,
+      "learning_rate": 5.163714439076758e-06,
+      "loss": 0.0,
+      "step": 5540
+    },
+    {
+      "epoch": 79.00193236714976,
+      "grad_norm": 0.005755449179559946,
+      "learning_rate": 5.152979066022544e-06,
+      "loss": 0.0,
+      "step": 5550
+    },
+    {
+      "epoch": 79.00289855072464,
+      "grad_norm": 0.0008575405227020383,
+      "learning_rate": 5.142243692968331e-06,
+      "loss": 0.0,
+      "step": 5560
+    },
+    {
+      "epoch": 79.00386473429951,
+      "grad_norm": 0.0010344963520765305,
+      "learning_rate": 5.131508319914117e-06,
+      "loss": 0.0,
+      "step": 5570
+    },
+    {
+      "epoch": 79.0048309178744,
+      "grad_norm": 0.0012300886446610093,
+      "learning_rate": 5.1207729468599045e-06,
+      "loss": 0.0,
+      "step": 5580
+    },
+    {
+      "epoch": 79.00579710144928,
+      "grad_norm": 0.0005992982187308371,
+      "learning_rate": 5.110037573805691e-06,
+      "loss": 0.0,
+      "step": 5590
+    },
+    {
+      "epoch": 79.00676328502415,
+      "grad_norm": 0.0005942785064689815,
+      "learning_rate": 5.099302200751477e-06,
+      "loss": 0.0001,
+      "step": 5600
+    },
+    {
+      "epoch": 79.00676328502415,
+      "eval_accuracy": 0.7792207792207793,
+      "eval_loss": 1.8182388544082642,
+      "eval_runtime": 14.6812,
+      "eval_samples_per_second": 5.245,
+      "eval_steps_per_second": 1.362,
+      "step": 5600
+    },
+    {
+      "epoch": 80.00096618357487,
+      "grad_norm": 0.016651395708322525,
+      "learning_rate": 5.088566827697263e-06,
+      "loss": 0.0,
+      "step": 5610
+    },
+    {
+      "epoch": 80.00193236714976,
+      "grad_norm": 0.0006846770993433893,
+      "learning_rate": 5.0778314546430495e-06,
+      "loss": 0.0,
+      "step": 5620
+    },
+    {
+      "epoch": 80.00289855072464,
+      "grad_norm": 0.0005728365504182875,
+      "learning_rate": 5.067096081588836e-06,
+      "loss": 0.0,
+      "step": 5630
+    },
+    {
+      "epoch": 80.00386473429951,
+      "grad_norm": 0.0016862893244251609,
+      "learning_rate": 5.056360708534622e-06,
+      "loss": 0.002,
+      "step": 5640
+    },
+    {
+      "epoch": 80.0048309178744,
+      "grad_norm": 0.0006224879762157798,
+      "learning_rate": 5.045625335480409e-06,
+      "loss": 0.0,
+      "step": 5650
+    },
+    {
+      "epoch": 80.00579710144928,
+      "grad_norm": 0.0005681588081642985,
+      "learning_rate": 5.034889962426195e-06,
+      "loss": 0.0118,
+      "step": 5660
+    },
+    {
+      "epoch": 80.00676328502415,
+      "grad_norm": 0.00028208878939040005,
+      "learning_rate": 5.024154589371981e-06,
+      "loss": 0.0,
+      "step": 5670
+    },
+    {
+      "epoch": 80.00676328502415,
+      "eval_accuracy": 0.8051948051948052,
+      "eval_loss": 1.7965093851089478,
+      "eval_runtime": 14.7927,
+      "eval_samples_per_second": 5.205,
+      "eval_steps_per_second": 1.352,
+      "step": 5670
+    },
+    {
+      "epoch": 81.00096618357487,
+      "grad_norm": 0.014044140465557575,
+      "learning_rate": 5.013419216317767e-06,
+      "loss": 0.091,
+      "step": 5680
+    },
+    {
+      "epoch": 81.00193236714976,
+      "grad_norm": 0.0007891967543400824,
+      "learning_rate": 5.002683843263554e-06,
+      "loss": 0.1254,
+      "step": 5690
+    },
+    {
+      "epoch": 81.00289855072464,
+      "grad_norm": 0.0007473126170225441,
+      "learning_rate": 4.99194847020934e-06,
+      "loss": 0.0,
+      "step": 5700
+    },
+    {
+      "epoch": 81.00386473429951,
+      "grad_norm": 0.001082300441339612,
+      "learning_rate": 4.981213097155127e-06,
+      "loss": 0.0,
+      "step": 5710
+    },
+    {
+      "epoch": 81.0048309178744,
+      "grad_norm": 0.0004553549806587398,
+      "learning_rate": 4.970477724100913e-06,
+      "loss": 0.0,
+      "step": 5720
+    },
+    {
+      "epoch": 81.00579710144928,
+      "grad_norm": 0.00046312747872434556,
+      "learning_rate": 4.959742351046699e-06,
+      "loss": 0.0001,
+      "step": 5730
+    },
+    {
+      "epoch": 81.00676328502415,
+      "grad_norm": 0.0009012920781970024,
+      "learning_rate": 4.949006977992485e-06,
+      "loss": 0.0001,
+      "step": 5740
+    },
+    {
+      "epoch": 81.00676328502415,
+      "eval_accuracy": 0.6753246753246753,
+      "eval_loss": 3.0535547733306885,
+      "eval_runtime": 14.6889,
+      "eval_samples_per_second": 5.242,
+      "eval_steps_per_second": 1.362,
+      "step": 5740
+    },
+    {
+      "epoch": 82.00096618357487,
+      "grad_norm": 0.000618190853856504,
+      "learning_rate": 4.938271604938272e-06,
+      "loss": 0.0018,
+      "step": 5750
+    },
+    {
+      "epoch": 82.00193236714976,
+      "grad_norm": 0.0016643240815028548,
+      "learning_rate": 4.927536231884059e-06,
+      "loss": 0.1313,
+      "step": 5760
+    },
+    {
+      "epoch": 82.00289855072464,
+      "grad_norm": 0.10652245581150055,
+      "learning_rate": 4.9168008588298446e-06,
+      "loss": 0.6473,
+      "step": 5770
+    },
+    {
+      "epoch": 82.00386473429951,
+      "grad_norm": 0.0007297329721041024,
+      "learning_rate": 4.906065485775631e-06,
+      "loss": 0.3155,
+      "step": 5780
+    },
+    {
+      "epoch": 82.0048309178744,
+      "grad_norm": 0.0021761939860880375,
+      "learning_rate": 4.8953301127214175e-06,
+      "loss": 0.0016,
+      "step": 5790
+    },
+    {
+      "epoch": 82.00579710144928,
+      "grad_norm": 0.0007605966529808939,
+      "learning_rate": 4.884594739667204e-06,
+      "loss": 0.0,
+      "step": 5800
+    },
+    {
+      "epoch": 82.00676328502415,
+      "grad_norm": 0.0007947117555886507,
+      "learning_rate": 4.87385936661299e-06,
+      "loss": 0.0005,
+      "step": 5810
+    },
+    {
+      "epoch": 82.00676328502415,
+      "eval_accuracy": 0.7922077922077922,
+      "eval_loss": 1.5426697731018066,
+      "eval_runtime": 14.4326,
+      "eval_samples_per_second": 5.335,
+      "eval_steps_per_second": 1.386,
+      "step": 5810
+    },
+    {
+      "epoch": 83.00096618357487,
+      "grad_norm": 0.016618404537439346,
+      "learning_rate": 4.863123993558777e-06,
+      "loss": 0.0,
+      "step": 5820
+    },
+    {
+      "epoch": 83.00193236714976,
+      "grad_norm": 0.0010025931987911463,
+      "learning_rate": 4.852388620504562e-06,
+      "loss": 0.0,
+      "step": 5830
+    },
+    {
+      "epoch": 83.00289855072464,
+      "grad_norm": 0.003782440209761262,
+      "learning_rate": 4.841653247450349e-06,
+      "loss": 0.0871,
+      "step": 5840
+    },
+    {
+      "epoch": 83.00386473429951,
+      "grad_norm": 0.03455811366438866,
+      "learning_rate": 4.830917874396135e-06,
+      "loss": 0.0,
+      "step": 5850
+    },
+    {
+      "epoch": 83.0048309178744,
+      "grad_norm": 0.0011453116312623024,
+      "learning_rate": 4.820182501341922e-06,
+      "loss": 0.0,
+      "step": 5860
+    },
+    {
+      "epoch": 83.00579710144928,
+      "grad_norm": 0.0074026999063789845,
+      "learning_rate": 4.809447128287709e-06,
+      "loss": 0.0,
+      "step": 5870
+    },
+    {
+      "epoch": 83.00676328502415,
+      "grad_norm": 0.0005923007847741246,
+      "learning_rate": 4.798711755233495e-06,
+      "loss": 0.0006,
+      "step": 5880
+    },
+    {
+      "epoch": 83.00676328502415,
+      "eval_accuracy": 0.7402597402597403,
+      "eval_loss": 1.8892239332199097,
+      "eval_runtime": 14.285,
+      "eval_samples_per_second": 5.39,
+      "eval_steps_per_second": 1.4,
+      "step": 5880
+    },
+    {
+      "epoch": 84.00096618357487,
+      "grad_norm": 0.0004263128503225744,
+      "learning_rate": 4.787976382179281e-06,
+      "loss": 0.0,
+      "step": 5890
+    },
+    {
+      "epoch": 84.00193236714976,
+      "grad_norm": 0.0024923263117671013,
+      "learning_rate": 4.7772410091250676e-06,
+      "loss": 0.1811,
+      "step": 5900
+    },
+    {
+      "epoch": 84.00289855072464,
+      "grad_norm": 0.0008309787954203784,
+      "learning_rate": 4.766505636070854e-06,
+      "loss": 0.0001,
+      "step": 5910
+    },
+    {
+      "epoch": 84.00386473429951,
+      "grad_norm": 0.0008232109830714762,
+      "learning_rate": 4.7557702630166405e-06,
+      "loss": 0.0001,
+      "step": 5920
+    },
+    {
+      "epoch": 84.0048309178744,
+      "grad_norm": 0.0004610401520039886,
+      "learning_rate": 4.745034889962427e-06,
+      "loss": 0.0,
+      "step": 5930
+    },
+    {
+      "epoch": 84.00579710144928,
+      "grad_norm": 0.0005724320071749389,
+      "learning_rate": 4.7342995169082125e-06,
+      "loss": 0.2355,
+      "step": 5940
+    },
+    {
+      "epoch": 84.00676328502415,
+      "grad_norm": 0.0007560947560705245,
+      "learning_rate": 4.723564143853999e-06,
+      "loss": 0.0001,
+      "step": 5950
+    },
+    {
+      "epoch": 84.00676328502415,
+      "eval_accuracy": 0.7402597402597403,
+      "eval_loss": 1.9647512435913086,
+      "eval_runtime": 14.5317,
+      "eval_samples_per_second": 5.299,
+      "eval_steps_per_second": 1.376,
+      "step": 5950
+    },
+    {
+      "epoch": 85.00096618357487,
+      "grad_norm": 0.0017191518563777208,
+      "learning_rate": 4.712828770799785e-06,
+      "loss": 0.0,
+      "step": 5960
+    },
+    {
+      "epoch": 85.00193236714976,
+      "grad_norm": 0.004426192957907915,
+      "learning_rate": 4.702093397745572e-06,
+      "loss": 0.0,
+      "step": 5970
+    },
+    {
+      "epoch": 85.00289855072464,
+      "grad_norm": 0.0018137526931241155,
+      "learning_rate": 4.691358024691358e-06,
+      "loss": 0.1938,
+      "step": 5980
+    },
+    {
+      "epoch": 85.00386473429951,
+      "grad_norm": 0.0018216772004961967,
+      "learning_rate": 4.680622651637145e-06,
+      "loss": 0.0137,
+      "step": 5990
+    },
+    {
+      "epoch": 85.0048309178744,
+      "grad_norm": 0.0010804088087752461,
+      "learning_rate": 4.669887278582931e-06,
+      "loss": 0.2,
+      "step": 6000
+    },
+    {
+      "epoch": 85.00579710144928,
+      "grad_norm": 0.0011872736504301429,
+      "learning_rate": 4.659151905528718e-06,
+      "loss": 0.0,
+      "step": 6010
+    },
+    {
+      "epoch": 85.00676328502415,
+      "grad_norm": 0.0015459536807611585,
+      "learning_rate": 4.648416532474504e-06,
+      "loss": 0.0,
+      "step": 6020
+    },
+    {
+      "epoch": 85.00676328502415,
+      "eval_accuracy": 0.7532467532467533,
+      "eval_loss": 1.7624553442001343,
+      "eval_runtime": 14.3078,
+      "eval_samples_per_second": 5.382,
+      "eval_steps_per_second": 1.398,
+      "step": 6020
+    },
+    {
+      "epoch": 86.00096618357487,
+      "grad_norm": 0.0007234421791508794,
+      "learning_rate": 4.637681159420291e-06,
+      "loss": 0.0,
+      "step": 6030
+    },
+    {
+      "epoch": 86.00193236714976,
+      "grad_norm": 0.001056589768268168,
+      "learning_rate": 4.626945786366077e-06,
+      "loss": 0.0,
+      "step": 6040
+    },
+    {
+      "epoch": 86.00289855072464,
+      "grad_norm": 0.0015586380613967776,
+      "learning_rate": 4.616210413311863e-06,
+      "loss": 0.0,
+      "step": 6050
+    },
+    {
+      "epoch": 86.00386473429951,
+      "grad_norm": 0.0019250113982707262,
+      "learning_rate": 4.605475040257649e-06,
+      "loss": 0.0,
+      "step": 6060
+    },
+    {
+      "epoch": 86.0048309178744,
+      "grad_norm": 0.4438691735267639,
+      "learning_rate": 4.5947396672034355e-06,
+      "loss": 0.0001,
+      "step": 6070
+    },
+    {
+      "epoch": 86.00579710144928,
+      "grad_norm": 0.00045592489186674356,
+      "learning_rate": 4.584004294149222e-06,
+      "loss": 0.0,
+      "step": 6080
+    },
+    {
+      "epoch": 86.00676328502415,
+      "grad_norm": 0.0016952749574556947,
+      "learning_rate": 4.5732689210950084e-06,
+      "loss": 0.1655,
+      "step": 6090
+    },
+    {
+      "epoch": 86.00676328502415,
+      "eval_accuracy": 0.7662337662337663,
+      "eval_loss": 1.6750868558883667,
+      "eval_runtime": 14.1595,
+      "eval_samples_per_second": 5.438,
+      "eval_steps_per_second": 1.412,
+      "step": 6090
+    },
+    {
+      "epoch": 87.00096618357487,
+      "grad_norm": 0.00048763351514935493,
+      "learning_rate": 4.562533548040795e-06,
+      "loss": 0.0001,
+      "step": 6100
+    },
+    {
+      "epoch": 87.00193236714976,
+      "grad_norm": 0.0004426055820658803,
+      "learning_rate": 4.551798174986581e-06,
+      "loss": 0.0,
+      "step": 6110
+    },
+    {
+      "epoch": 87.00289855072464,
+      "grad_norm": 0.00048457886441610754,
+      "learning_rate": 4.541062801932368e-06,
+      "loss": 0.0,
+      "step": 6120
+    },
+    {
+      "epoch": 87.00386473429951,
+      "grad_norm": 0.0007973539759404957,
+      "learning_rate": 4.530327428878154e-06,
+      "loss": 0.0,
+      "step": 6130
+    },
+    {
+      "epoch": 87.0048309178744,
+      "grad_norm": 0.004327856469899416,
+      "learning_rate": 4.519592055823941e-06,
+      "loss": 0.0,
+      "step": 6140
+    },
+    {
+      "epoch": 87.00579710144928,
+      "grad_norm": 0.0011828347342088819,
+      "learning_rate": 4.508856682769726e-06,
+      "loss": 0.0,
+      "step": 6150
+    },
+    {
+      "epoch": 87.00676328502415,
+      "grad_norm": 0.0006070521776564419,
+      "learning_rate": 4.498121309715513e-06,
+      "loss": 0.0,
+      "step": 6160
+    },
+    {
+      "epoch": 87.00676328502415,
+      "eval_accuracy": 0.7402597402597403,
+      "eval_loss": 1.8558989763259888,
+      "eval_runtime": 17.515,
+      "eval_samples_per_second": 4.396,
+      "eval_steps_per_second": 1.142,
+      "step": 6160
+    },
+    {
+      "epoch": 88.00096618357487,
+      "grad_norm": 0.010230960324406624,
+      "learning_rate": 4.487385936661299e-06,
+      "loss": 0.0,
+      "step": 6170
+    },
+    {
+      "epoch": 88.00193236714976,
+      "grad_norm": 0.0005492049385793507,
+      "learning_rate": 4.476650563607086e-06,
+      "loss": 0.0,
+      "step": 6180
+    },
+    {
+      "epoch": 88.00289855072464,
+      "grad_norm": 0.0004779798910021782,
+      "learning_rate": 4.465915190552872e-06,
+      "loss": 0.0,
+      "step": 6190
+    },
+    {
+      "epoch": 88.00386473429951,
+      "grad_norm": 0.0003807265020441264,
+      "learning_rate": 4.4551798174986585e-06,
+      "loss": 0.0,
+      "step": 6200
+    },
+    {
+      "epoch": 88.0048309178744,
+      "grad_norm": 0.0005069471662864089,
+      "learning_rate": 4.444444444444444e-06,
+      "loss": 0.0,
+      "step": 6210
+    },
+    {
+      "epoch": 88.00579710144928,
+      "grad_norm": 0.0007103086682036519,
+      "learning_rate": 4.4337090713902314e-06,
+      "loss": 0.0,
+      "step": 6220
+    },
+    {
+      "epoch": 88.00676328502415,
+      "grad_norm": 0.0003308405866846442,
+      "learning_rate": 4.422973698336018e-06,
+      "loss": 0.0,
+      "step": 6230
+    },
+    {
+      "epoch": 88.00676328502415,
+      "eval_accuracy": 0.7532467532467533,
+      "eval_loss": 1.8885866403579712,
+      "eval_runtime": 14.812,
+      "eval_samples_per_second": 5.198,
+      "eval_steps_per_second": 1.35,
+      "step": 6230
+    },
+    {
+      "epoch": 89.00096618357487,
+      "grad_norm": 0.001938937697559595,
+      "learning_rate": 4.412238325281804e-06,
+      "loss": 0.0,
+      "step": 6240
+    },
+    {
+      "epoch": 89.00193236714976,
+      "grad_norm": 0.0003955487336497754,
+      "learning_rate": 4.401502952227591e-06,
+      "loss": 0.0,
+      "step": 6250
+    },
+    {
+      "epoch": 89.00289855072464,
+      "grad_norm": 0.0007904915837571025,
+      "learning_rate": 4.390767579173376e-06,
+      "loss": 0.0,
+      "step": 6260
+    },
+    {
+      "epoch": 89.00386473429951,
+      "grad_norm": 0.0005482785636559129,
+      "learning_rate": 4.380032206119163e-06,
+      "loss": 0.0,
+      "step": 6270
+    },
+    {
+      "epoch": 89.0048309178744,
+      "grad_norm": 0.0005851537571288645,
+      "learning_rate": 4.369296833064949e-06,
+      "loss": 0.0,
+      "step": 6280
+    },
+    {
+      "epoch": 89.00579710144928,
+      "grad_norm": 0.0003069818776566535,
+      "learning_rate": 4.358561460010736e-06,
+      "loss": 0.0,
+      "step": 6290
+    },
+    {
+      "epoch": 89.00676328502415,
+      "grad_norm": 0.019225388765335083,
+      "learning_rate": 4.347826086956522e-06,
+      "loss": 0.0,
+      "step": 6300
+    },
+    {
+      "epoch": 89.00676328502415,
+      "eval_accuracy": 0.7532467532467533,
+      "eval_loss": 1.8957127332687378,
+      "eval_runtime": 17.6162,
+      "eval_samples_per_second": 4.371,
+      "eval_steps_per_second": 1.135,
+      "step": 6300
+    },
+    {
+      "epoch": 90.00096618357487,
+      "grad_norm": 0.0005754005396738648,
+      "learning_rate": 4.337090713902309e-06,
+      "loss": 0.0,
+      "step": 6310
+    },
+    {
+      "epoch": 90.00193236714976,
+      "grad_norm": 0.0008540857234038413,
+      "learning_rate": 4.326355340848094e-06,
+      "loss": 0.0,
+      "step": 6320
+    },
+    {
+      "epoch": 90.00289855072464,
+      "grad_norm": 0.0004658667021431029,
+      "learning_rate": 4.315619967793881e-06,
+      "loss": 0.0,
+      "step": 6330
+    },
+    {
+      "epoch": 90.00386473429951,
+      "grad_norm": 0.0007123078685253859,
+      "learning_rate": 4.304884594739668e-06,
+      "loss": 0.0,
+      "step": 6340
+    },
+    {
+      "epoch": 90.0048309178744,
+      "grad_norm": 0.0006745746941305697,
+      "learning_rate": 4.2941492216854545e-06,
+      "loss": 0.0,
+      "step": 6350
+    },
+    {
+      "epoch": 90.00579710144928,
+      "grad_norm": 0.0003169019764754921,
+      "learning_rate": 4.28341384863124e-06,
+      "loss": 0.0,
+      "step": 6360
+    },
+    {
+      "epoch": 90.00676328502415,
+      "grad_norm": 0.0005867715808562934,
+      "learning_rate": 4.2726784755770265e-06,
+      "loss": 0.0,
+      "step": 6370
+    },
+    {
+      "epoch": 90.00676328502415,
+      "eval_accuracy": 0.7662337662337663,
+      "eval_loss": 1.8181320428848267,
+      "eval_runtime": 17.5597,
+      "eval_samples_per_second": 4.385,
+      "eval_steps_per_second": 1.139,
+      "step": 6370
+    },
+    {
+      "epoch": 91.00096618357487,
+      "grad_norm": 0.0008447110303677619,
+      "learning_rate": 4.261943102522813e-06,
+      "loss": 0.0,
+      "step": 6380
+    },
+    {
+      "epoch": 91.00193236714976,
+      "grad_norm": 0.00033777134376578033,
+      "learning_rate": 4.251207729468599e-06,
+      "loss": 0.0,
+      "step": 6390
+    },
+    {
+      "epoch": 91.00289855072464,
+      "grad_norm": 0.0013186403084546328,
+      "learning_rate": 4.240472356414386e-06,
+      "loss": 0.0,
+      "step": 6400
+    },
+    {
+      "epoch": 91.00386473429951,
+      "grad_norm": 0.0007186243892647326,
+      "learning_rate": 4.229736983360172e-06,
+      "loss": 0.0,
+      "step": 6410
+    },
+    {
+      "epoch": 91.0048309178744,
+      "grad_norm": 0.0007361398893408477,
+      "learning_rate": 4.219001610305958e-06,
+      "loss": 0.0,
+      "step": 6420
+    },
+    {
+      "epoch": 91.00579710144928,
+      "grad_norm": 0.0004776048008352518,
+      "learning_rate": 4.208266237251744e-06,
+      "loss": 0.0,
+      "step": 6430
+    },
+    {
+      "epoch": 91.00676328502415,
+      "grad_norm": 0.0003450379881542176,
+      "learning_rate": 4.197530864197531e-06,
+      "loss": 0.0,
+      "step": 6440
+    },
+    {
+      "epoch": 91.00676328502415,
+      "eval_accuracy": 0.7532467532467533,
+      "eval_loss": 1.829852819442749,
+      "eval_runtime": 17.0231,
+      "eval_samples_per_second": 4.523,
+      "eval_steps_per_second": 1.175,
+      "step": 6440
+    },
+    {
+      "epoch": 92.00096618357487,
+      "grad_norm": 0.001014379202388227,
+      "learning_rate": 4.186795491143318e-06,
+      "loss": 0.0,
+      "step": 6450
+    },
+    {
+      "epoch": 92.00193236714976,
+      "grad_norm": 0.006536729633808136,
+      "learning_rate": 4.1760601180891046e-06,
+      "loss": 0.2397,
+      "step": 6460
+    },
+    {
+      "epoch": 92.00289855072464,
+      "grad_norm": 0.0005509128677658737,
+      "learning_rate": 4.16532474503489e-06,
+      "loss": 0.0,
+      "step": 6470
+    },
+    {
+      "epoch": 92.00386473429951,
+      "grad_norm": 0.0008106320165097713,
+      "learning_rate": 4.154589371980677e-06,
+      "loss": 0.0,
+      "step": 6480
+    },
+    {
+      "epoch": 92.0048309178744,
+      "grad_norm": 0.0016772482777014375,
+      "learning_rate": 4.143853998926463e-06,
+      "loss": 0.0,
+      "step": 6490
+    },
+    {
+      "epoch": 92.00579710144928,
+      "grad_norm": 0.0008628064533695579,
+      "learning_rate": 4.1331186258722495e-06,
+      "loss": 0.0,
+      "step": 6500
+    },
+    {
+      "epoch": 92.00676328502415,
+      "grad_norm": 0.00043893515248782933,
+      "learning_rate": 4.122383252818036e-06,
+      "loss": 0.0,
+      "step": 6510
+    },
+    {
+      "epoch": 92.00676328502415,
+      "eval_accuracy": 0.8181818181818182,
+      "eval_loss": 1.518646478652954,
+      "eval_runtime": 14.5281,
+      "eval_samples_per_second": 5.3,
+      "eval_steps_per_second": 1.377,
+      "step": 6510
+    },
+    {
+      "epoch": 93.00096618357487,
+      "grad_norm": 0.00041715340921655297,
+      "learning_rate": 4.1116478797638224e-06,
+      "loss": 0.0,
+      "step": 6520
+    },
+    {
+      "epoch": 93.00193236714976,
+      "grad_norm": 0.0004664282314479351,
+      "learning_rate": 4.100912506709608e-06,
+      "loss": 0.0,
+      "step": 6530
+    },
+    {
+      "epoch": 93.00289855072464,
+      "grad_norm": 0.000606657937169075,
+      "learning_rate": 4.0901771336553945e-06,
+      "loss": 0.0,
+      "step": 6540
+    },
+    {
+      "epoch": 93.00386473429951,
+      "grad_norm": 0.00048786430852487683,
+      "learning_rate": 4.079441760601181e-06,
+      "loss": 0.0,
+      "step": 6550
+    },
+    {
+      "epoch": 93.0048309178744,
+      "grad_norm": 0.00024992445833049715,
+      "learning_rate": 4.068706387546967e-06,
+      "loss": 0.0,
+      "step": 6560
+    },
+    {
+      "epoch": 93.00579710144928,
+      "grad_norm": 0.0005548360059037805,
+      "learning_rate": 4.057971014492754e-06,
+      "loss": 0.0,
+      "step": 6570
+    },
+    {
+      "epoch": 93.00676328502415,
+      "grad_norm": 0.0003060603339690715,
+      "learning_rate": 4.04723564143854e-06,
+      "loss": 0.0393,
+      "step": 6580
+    },
+    {
+      "epoch": 93.00676328502415,
+      "eval_accuracy": 0.7792207792207793,
+      "eval_loss": 1.9233503341674805,
+      "eval_runtime": 14.3806,
+      "eval_samples_per_second": 5.354,
+      "eval_steps_per_second": 1.391,
+      "step": 6580
+    },
+    {
+      "epoch": 94.00096618357487,
+      "grad_norm": 0.0003109258250333369,
+      "learning_rate": 4.036500268384327e-06,
+      "loss": 0.0031,
+      "step": 6590
+    },
+    {
+      "epoch": 94.00193236714976,
+      "grad_norm": 0.0008100513368844986,
+      "learning_rate": 4.025764895330113e-06,
+      "loss": 0.0,
+      "step": 6600
+    },
+    {
+      "epoch": 94.00289855072464,
+      "grad_norm": 0.007947184145450592,
+      "learning_rate": 4.0150295222759e-06,
+      "loss": 0.0,
+      "step": 6610
+    },
+    {
+      "epoch": 94.00386473429951,
+      "grad_norm": 0.0008597071864642203,
+      "learning_rate": 4.004294149221686e-06,
+      "loss": 0.0,
+      "step": 6620
+    },
+    {
+      "epoch": 94.0048309178744,
+      "grad_norm": 0.0004177772789262235,
+      "learning_rate": 3.9935587761674725e-06,
+      "loss": 0.0,
+      "step": 6630
+    },
+    {
+      "epoch": 94.00579710144928,
+      "grad_norm": 0.0028885523788630962,
+      "learning_rate": 3.982823403113258e-06,
+      "loss": 0.0,
+      "step": 6640
+    },
+    {
+      "epoch": 94.00676328502415,
+      "grad_norm": 0.00042609605588950217,
+      "learning_rate": 3.972088030059045e-06,
+      "loss": 0.0,
+      "step": 6650
+    },
+    {
+      "epoch": 94.00676328502415,
+      "eval_accuracy": 0.7272727272727273,
+      "eval_loss": 2.1198887825012207,
+      "eval_runtime": 16.0169,
+      "eval_samples_per_second": 4.807,
+      "eval_steps_per_second": 1.249,
+      "step": 6650
+    },
+    {
+      "epoch": 95.00096618357487,
+      "grad_norm": 0.0003437358245719224,
+      "learning_rate": 3.961352657004831e-06,
+      "loss": 0.0,
+      "step": 6660
+    },
+    {
+      "epoch": 95.00193236714976,
+      "grad_norm": 0.0002173613611375913,
+      "learning_rate": 3.9506172839506175e-06,
+      "loss": 0.0,
+      "step": 6670
+    },
+    {
+      "epoch": 95.00289855072464,
+      "grad_norm": 0.0003850398352369666,
+      "learning_rate": 3.939881910896404e-06,
+      "loss": 0.0,
+      "step": 6680
+    },
+    {
+      "epoch": 95.00386473429951,
+      "grad_norm": 0.00038904434768483043,
+      "learning_rate": 3.92914653784219e-06,
+      "loss": 0.0,
+      "step": 6690
+    },
+    {
+      "epoch": 95.0048309178744,
+      "grad_norm": 0.0010339280124753714,
+      "learning_rate": 3.918411164787977e-06,
+      "loss": 0.0,
+      "step": 6700
+    },
+    {
+      "epoch": 95.00579710144928,
+      "grad_norm": 0.0012966475915163755,
+      "learning_rate": 3.907675791733763e-06,
+      "loss": 0.0,
+      "step": 6710
+    },
+    {
+      "epoch": 95.00676328502415,
+      "grad_norm": 0.16356423497200012,
+      "learning_rate": 3.89694041867955e-06,
+      "loss": 0.0,
+      "step": 6720
+    },
+    {
+      "epoch": 95.00676328502415,
+      "eval_accuracy": 0.7402597402597403,
+      "eval_loss": 2.1308841705322266,
+      "eval_runtime": 18.5425,
+      "eval_samples_per_second": 4.153,
+      "eval_steps_per_second": 1.079,
+      "step": 6720
+    },
+    {
+      "epoch": 96.00096618357487,
+      "grad_norm": 0.000308900693198666,
+      "learning_rate": 3.886205045625336e-06,
+      "loss": 0.0,
+      "step": 6730
+    },
+    {
+      "epoch": 96.00193236714976,
+      "grad_norm": 0.008645855821669102,
+      "learning_rate": 3.875469672571122e-06,
+      "loss": 0.0,
+      "step": 6740
+    },
+    {
+      "epoch": 96.00289855072464,
+      "grad_norm": 0.001571167609654367,
+      "learning_rate": 3.864734299516908e-06,
+      "loss": 0.1613,
+      "step": 6750
+    },
+    {
+      "epoch": 96.00386473429951,
+      "grad_norm": 0.000592782802414149,
+      "learning_rate": 3.853998926462695e-06,
+      "loss": 0.1357,
+      "step": 6760
+    },
+    {
+      "epoch": 96.0048309178744,
+      "grad_norm": 0.0014255845453590155,
+      "learning_rate": 3.843263553408481e-06,
+      "loss": 0.0,
+      "step": 6770
+    },
+    {
+      "epoch": 96.00579710144928,
+      "grad_norm": 0.0014380579814314842,
+      "learning_rate": 3.832528180354268e-06,
+      "loss": 0.0,
+      "step": 6780
+    },
+    {
+      "epoch": 96.00676328502415,
+      "grad_norm": 0.00042985283653251827,
+      "learning_rate": 3.821792807300054e-06,
+      "loss": 0.0009,
+      "step": 6790
+    },
+    {
+      "epoch": 96.00676328502415,
+      "eval_accuracy": 0.7532467532467533,
+      "eval_loss": 1.931089162826538,
+      "eval_runtime": 17.8388,
+      "eval_samples_per_second": 4.316,
+      "eval_steps_per_second": 1.121,
+      "step": 6790
+    },
+    {
+      "epoch": 97.00096618357487,
+      "grad_norm": 0.03030485101044178,
+      "learning_rate": 3.8110574342458405e-06,
+      "loss": 0.0,
+      "step": 6800
+    },
+    {
+      "epoch": 97.00193236714976,
+      "grad_norm": 0.0003676621417980641,
+      "learning_rate": 3.800322061191627e-06,
+      "loss": 0.0015,
+      "step": 6810
+    },
+    {
+      "epoch": 97.00289855072464,
+      "grad_norm": 0.0006548650562763214,
+      "learning_rate": 3.7895866881374134e-06,
+      "loss": 0.0,
+      "step": 6820
+    },
+    {
+      "epoch": 97.00386473429951,
+      "grad_norm": 0.00035164476139470935,
+      "learning_rate": 3.7788513150831994e-06,
+      "loss": 0.0,
+      "step": 6830
+    },
+    {
+      "epoch": 97.0048309178744,
+      "grad_norm": 0.14484809339046478,
+      "learning_rate": 3.768115942028986e-06,
+      "loss": 0.0,
+      "step": 6840
+    },
+    {
+      "epoch": 97.00579710144928,
+      "grad_norm": 0.00030128343496471643,
+      "learning_rate": 3.7573805689747723e-06,
+      "loss": 0.0003,
+      "step": 6850
+    },
+    {
+      "epoch": 97.00676328502415,
+      "grad_norm": 0.0002703612844925374,
+      "learning_rate": 3.7466451959205584e-06,
+      "loss": 0.0001,
+      "step": 6860
+    },
+    {
+      "epoch": 97.00676328502415,
+      "eval_accuracy": 0.7792207792207793,
+      "eval_loss": 1.7857838869094849,
+      "eval_runtime": 17.5204,
+      "eval_samples_per_second": 4.395,
+      "eval_steps_per_second": 1.142,
+      "step": 6860
+    },
+    {
+      "epoch": 98.00096618357487,
+      "grad_norm": 0.0002425261918688193,
+      "learning_rate": 3.735909822866345e-06,
+      "loss": 0.0007,
+      "step": 6870
+    },
+    {
+      "epoch": 98.00193236714976,
+      "grad_norm": 0.0003744879795704037,
+      "learning_rate": 3.7251744498121313e-06,
+      "loss": 0.0,
+      "step": 6880
+    },
+    {
+      "epoch": 98.00289855072464,
+      "grad_norm": 0.0009580018813721836,
+      "learning_rate": 3.7144390767579173e-06,
+      "loss": 0.0,
+      "step": 6890
+    },
+    {
+      "epoch": 98.00386473429951,
+      "grad_norm": 0.00039155595004558563,
+      "learning_rate": 3.7037037037037037e-06,
+      "loss": 0.2244,
+      "step": 6900
+    },
+    {
+      "epoch": 98.0048309178744,
+      "grad_norm": 0.001988824224099517,
+      "learning_rate": 3.69296833064949e-06,
+      "loss": 0.0,
+      "step": 6910
+    },
+    {
+      "epoch": 98.00579710144928,
+      "grad_norm": 0.00045414830674417317,
+      "learning_rate": 3.682232957595277e-06,
+      "loss": 0.0294,
+      "step": 6920
+    },
+    {
+      "epoch": 98.00676328502415,
+      "grad_norm": 0.0010452043497934937,
+      "learning_rate": 3.6714975845410635e-06,
+      "loss": 0.0894,
+      "step": 6930
+    },
+    {
+      "epoch": 98.00676328502415,
+      "eval_accuracy": 0.8051948051948052,
+      "eval_loss": 1.5577243566513062,
+      "eval_runtime": 18.0262,
+      "eval_samples_per_second": 4.272,
+      "eval_steps_per_second": 1.109,
+      "step": 6930
+    },
+    {
+      "epoch": 99.00096618357487,
+      "grad_norm": 0.010575399734079838,
+      "learning_rate": 3.6607622114868495e-06,
+      "loss": 0.0002,
+      "step": 6940
+    },
+    {
+      "epoch": 99.00193236714976,
+      "grad_norm": 0.010563873685896397,
+      "learning_rate": 3.650026838432636e-06,
+      "loss": 0.0,
+      "step": 6950
+    },
+    {
+      "epoch": 99.00289855072464,
+      "grad_norm": 0.00037548152613453567,
+      "learning_rate": 3.6392914653784224e-06,
+      "loss": 0.0,
+      "step": 6960
+    },
+    {
+      "epoch": 99.00386473429951,
+      "grad_norm": 0.0008382857777178288,
+      "learning_rate": 3.6285560923242085e-06,
+      "loss": 0.0,
+      "step": 6970
+    },
+    {
+      "epoch": 99.0048309178744,
+      "grad_norm": 0.000387381121981889,
+      "learning_rate": 3.617820719269995e-06,
+      "loss": 0.0,
+      "step": 6980
+    },
+    {
+      "epoch": 99.00579710144928,
+      "grad_norm": 0.0008243945194408298,
+      "learning_rate": 3.6070853462157814e-06,
+      "loss": 0.0001,
+      "step": 6990
+    },
+    {
+      "epoch": 99.00676328502415,
+      "grad_norm": 0.0006387074827216566,
+      "learning_rate": 3.5963499731615674e-06,
+      "loss": 0.0,
+      "step": 7000
+    },
+    {
+      "epoch": 99.00676328502415,
+      "eval_accuracy": 0.7792207792207793,
+      "eval_loss": 1.8137568235397339,
+      "eval_runtime": 17.7771,
+      "eval_samples_per_second": 4.331,
+      "eval_steps_per_second": 1.125,
+      "step": 7000
+    },
+    {
+      "epoch": 100.00096618357487,
+      "grad_norm": 0.0014033769257366657,
+      "learning_rate": 3.585614600107354e-06,
+      "loss": 0.105,
+      "step": 7010
+    },
+    {
+      "epoch": 100.00193236714976,
+      "grad_norm": 0.00042622030014172196,
+      "learning_rate": 3.5748792270531403e-06,
+      "loss": 0.0,
+      "step": 7020
+    },
+    {
+      "epoch": 100.00289855072464,
+      "grad_norm": 0.0019426278304308653,
+      "learning_rate": 3.5641438539989263e-06,
+      "loss": 0.0,
+      "step": 7030
+    },
+    {
+      "epoch": 100.00386473429951,
+      "grad_norm": 0.0008854815387167037,
+      "learning_rate": 3.553408480944713e-06,
+      "loss": 0.0,
+      "step": 7040
+    },
+    {
+      "epoch": 100.0048309178744,
+      "grad_norm": 0.0004940590006299317,
+      "learning_rate": 3.5426731078904997e-06,
+      "loss": 0.0,
+      "step": 7050
+    },
+    {
+      "epoch": 100.00579710144928,
+      "grad_norm": 0.0007130287704057992,
+      "learning_rate": 3.531937734836286e-06,
+      "loss": 0.0,
+      "step": 7060
+    },
+    {
+      "epoch": 100.00676328502415,
+      "grad_norm": 0.0007455471204593778,
+      "learning_rate": 3.521202361782072e-06,
+      "loss": 0.0,
+      "step": 7070
+    },
+    {
+      "epoch": 100.00676328502415,
+      "eval_accuracy": 0.7532467532467533,
+      "eval_loss": 2.0068461894989014,
+      "eval_runtime": 15.1443,
+      "eval_samples_per_second": 5.084,
+      "eval_steps_per_second": 1.321,
+      "step": 7070
+    },
+    {
+      "epoch": 101.00096618357487,
+      "grad_norm": 0.0003840252757072449,
+      "learning_rate": 3.5104669887278586e-06,
+      "loss": 0.0,
+      "step": 7080
+    },
+    {
+      "epoch": 101.00193236714976,
+      "grad_norm": 0.00036437122616916895,
+      "learning_rate": 3.499731615673645e-06,
+      "loss": 0.0006,
+      "step": 7090
+    },
+    {
+      "epoch": 101.00289855072464,
+      "grad_norm": 0.001055071479640901,
+      "learning_rate": 3.488996242619431e-06,
+      "loss": 0.0,
+      "step": 7100
+    },
+    {
+      "epoch": 101.00386473429951,
+      "grad_norm": 0.04166965186595917,
+      "learning_rate": 3.4782608695652175e-06,
+      "loss": 0.174,
+      "step": 7110
+    },
+    {
+      "epoch": 101.0048309178744,
+      "grad_norm": 0.0005369763821363449,
+      "learning_rate": 3.467525496511004e-06,
+      "loss": 0.0004,
+      "step": 7120
+    },
+    {
+      "epoch": 101.00579710144928,
+      "grad_norm": 0.0004044979577884078,
+      "learning_rate": 3.4567901234567904e-06,
+      "loss": 0.0,
+      "step": 7130
+    },
+    {
+      "epoch": 101.00676328502415,
+      "grad_norm": 0.0003100241592619568,
+      "learning_rate": 3.4460547504025764e-06,
+      "loss": 0.0163,
+      "step": 7140
+    },
+    {
+      "epoch": 101.00676328502415,
+      "eval_accuracy": 0.7922077922077922,
+      "eval_loss": 1.8340067863464355,
+      "eval_runtime": 17.1815,
+      "eval_samples_per_second": 4.482,
+      "eval_steps_per_second": 1.164,
+      "step": 7140
+    },
+    {
+      "epoch": 102.00096618357487,
+      "grad_norm": 0.0005807825946249068,
+      "learning_rate": 3.4353193773483633e-06,
+      "loss": 0.0015,
+      "step": 7150
+    },
+    {
+      "epoch": 102.00193236714976,
+      "grad_norm": 0.0004728983039967716,
+      "learning_rate": 3.4245840042941498e-06,
+      "loss": 0.0,
+      "step": 7160
+    },
+    {
+      "epoch": 102.00289855072464,
+      "grad_norm": 0.00022850323875900358,
+      "learning_rate": 3.4138486312399362e-06,
+      "loss": 0.0,
+      "step": 7170
+    },
+    {
+      "epoch": 102.00386473429951,
+      "grad_norm": 0.006223591510206461,
+      "learning_rate": 3.4031132581857222e-06,
+      "loss": 0.0,
+      "step": 7180
+    },
+    {
+      "epoch": 102.0048309178744,
+      "grad_norm": 0.00024900989956222475,
+      "learning_rate": 3.3923778851315087e-06,
+      "loss": 0.0,
+      "step": 7190
+    },
+    {
+      "epoch": 102.00579710144928,
+      "grad_norm": 0.0019101315410807729,
+      "learning_rate": 3.381642512077295e-06,
+      "loss": 0.0,
+      "step": 7200
+    },
+    {
+      "epoch": 102.00676328502415,
+      "grad_norm": 0.0003227620618417859,
+      "learning_rate": 3.370907139023081e-06,
+      "loss": 0.0,
+      "step": 7210
+    },
+    {
+      "epoch": 102.00676328502415,
+      "eval_accuracy": 0.8311688311688312,
+      "eval_loss": 1.3226364850997925,
+      "eval_runtime": 18.0086,
+      "eval_samples_per_second": 4.276,
+      "eval_steps_per_second": 1.111,
+      "step": 7210
+    },
+    {
+      "epoch": 103.00096618357487,
+      "grad_norm": 0.0004981742240488529,
+      "learning_rate": 3.3601717659688676e-06,
+      "loss": 0.0,
+      "step": 7220
+    },
+    {
+      "epoch": 103.00193236714976,
+      "grad_norm": 0.000562319066375494,
+      "learning_rate": 3.349436392914654e-06,
+      "loss": 0.0,
+      "step": 7230
+    },
+    {
+      "epoch": 103.00289855072464,
+      "grad_norm": 0.0005576030816882849,
+      "learning_rate": 3.33870101986044e-06,
+      "loss": 0.0,
+      "step": 7240
+    },
+    {
+      "epoch": 103.00386473429951,
+      "grad_norm": 0.0009905042825266719,
+      "learning_rate": 3.3279656468062265e-06,
+      "loss": 0.0,
+      "step": 7250
+    },
+    {
+      "epoch": 103.0048309178744,
+      "grad_norm": 0.00032818218460306525,
+      "learning_rate": 3.317230273752013e-06,
+      "loss": 0.0833,
+      "step": 7260
+    },
+    {
+      "epoch": 103.00579710144928,
+      "grad_norm": 0.0029643636662513018,
+      "learning_rate": 3.3064949006978e-06,
+      "loss": 0.0,
+      "step": 7270
+    },
+    {
+      "epoch": 103.00676328502415,
+      "grad_norm": 0.0008541871211491525,
+      "learning_rate": 3.295759527643586e-06,
+      "loss": 0.0,
+      "step": 7280
+    },
+    {
+      "epoch": 103.00676328502415,
+      "eval_accuracy": 0.7532467532467533,
+      "eval_loss": 2.460693597793579,
+      "eval_runtime": 17.2217,
+      "eval_samples_per_second": 4.471,
+      "eval_steps_per_second": 1.161,
+      "step": 7280
+    },
+    {
+      "epoch": 104.00096618357487,
+      "grad_norm": 0.0004660151316784322,
+      "learning_rate": 3.2850241545893724e-06,
+      "loss": 0.2667,
+      "step": 7290
+    },
+    {
+      "epoch": 104.00193236714976,
+      "grad_norm": 0.006872973870486021,
+      "learning_rate": 3.274288781535159e-06,
+      "loss": 0.0,
+      "step": 7300
+    },
+    {
+      "epoch": 104.00289855072464,
+      "grad_norm": 0.0034734373912215233,
+      "learning_rate": 3.263553408480945e-06,
+      "loss": 0.0,
+      "step": 7310
+    },
+    {
+      "epoch": 104.00386473429951,
+      "grad_norm": 0.002416298259049654,
+      "learning_rate": 3.2528180354267313e-06,
+      "loss": 0.0952,
+      "step": 7320
+    },
+    {
+      "epoch": 104.0048309178744,
+      "grad_norm": 0.00047030055429786444,
+      "learning_rate": 3.2420826623725177e-06,
+      "loss": 0.0,
+      "step": 7330
+    },
+    {
+      "epoch": 104.00579710144928,
+      "grad_norm": 0.002508195349946618,
+      "learning_rate": 3.231347289318304e-06,
+      "loss": 0.0,
+      "step": 7340
+    },
+    {
+      "epoch": 104.00676328502415,
+      "grad_norm": 0.006559464149177074,
+      "learning_rate": 3.22061191626409e-06,
+      "loss": 0.0683,
+      "step": 7350
+    },
+    {
+      "epoch": 104.00676328502415,
+      "eval_accuracy": 0.7922077922077922,
+      "eval_loss": 1.754974126815796,
+      "eval_runtime": 14.6171,
+      "eval_samples_per_second": 5.268,
+      "eval_steps_per_second": 1.368,
+      "step": 7350
+    },
+    {
+      "epoch": 105.00096618357487,
+      "grad_norm": 0.0030552211683243513,
+      "learning_rate": 3.2098765432098767e-06,
+      "loss": 0.0,
+      "step": 7360
+    },
+    {
+      "epoch": 105.00193236714976,
+      "grad_norm": 0.000478706177091226,
+      "learning_rate": 3.199141170155663e-06,
+      "loss": 0.0002,
+      "step": 7370
+    },
+    {
+      "epoch": 105.00289855072464,
+      "grad_norm": 0.0003611068823374808,
+      "learning_rate": 3.188405797101449e-06,
+      "loss": 0.0,
+      "step": 7380
+    },
+    {
+      "epoch": 105.00386473429951,
+      "grad_norm": 0.000608165399171412,
+      "learning_rate": 3.177670424047236e-06,
+      "loss": 0.0653,
+      "step": 7390
+    },
+    {
+      "epoch": 105.0048309178744,
+      "grad_norm": 0.0009645195677876472,
+      "learning_rate": 3.1669350509930225e-06,
+      "loss": 0.0,
+      "step": 7400
+    },
+    {
+      "epoch": 105.00579710144928,
+      "grad_norm": 0.0010169371962547302,
+      "learning_rate": 3.156199677938809e-06,
+      "loss": 0.0,
+      "step": 7410
+    },
+    {
+      "epoch": 105.00676328502415,
+      "grad_norm": 0.004264978691935539,
+      "learning_rate": 3.145464304884595e-06,
+      "loss": 0.0,
+      "step": 7420
+    },
+    {
+      "epoch": 105.00676328502415,
+      "eval_accuracy": 0.8311688311688312,
+      "eval_loss": 1.4899604320526123,
+      "eval_runtime": 14.5365,
+      "eval_samples_per_second": 5.297,
+      "eval_steps_per_second": 1.376,
+      "step": 7420
+    },
+    {
+      "epoch": 106.00096618357487,
+      "grad_norm": 0.0018262077355757356,
+      "learning_rate": 3.1347289318303814e-06,
+      "loss": 0.0,
+      "step": 7430
+    },
+    {
+      "epoch": 106.00193236714976,
+      "grad_norm": 0.000662806851323694,
+      "learning_rate": 3.123993558776168e-06,
+      "loss": 0.0,
+      "step": 7440
+    },
+    {
+      "epoch": 106.00289855072464,
+      "grad_norm": 0.0005241535254754126,
+      "learning_rate": 3.113258185721954e-06,
+      "loss": 0.198,
+      "step": 7450
+    },
+    {
+      "epoch": 106.00386473429951,
+      "grad_norm": 0.0009639605414122343,
+      "learning_rate": 3.1025228126677403e-06,
+      "loss": 0.0,
+      "step": 7460
+    },
+    {
+      "epoch": 106.0048309178744,
+      "grad_norm": 0.0003563823411241174,
+      "learning_rate": 3.0917874396135268e-06,
+      "loss": 0.0008,
+      "step": 7470
+    },
+    {
+      "epoch": 106.00579710144928,
+      "grad_norm": 0.00036870595067739487,
+      "learning_rate": 3.081052066559313e-06,
+      "loss": 0.0,
+      "step": 7480
+    },
+    {
+      "epoch": 106.00676328502415,
+      "grad_norm": 0.00026721181347966194,
+      "learning_rate": 3.0703166935050992e-06,
+      "loss": 0.0,
+      "step": 7490
+    },
+    {
+      "epoch": 106.00676328502415,
+      "eval_accuracy": 0.7662337662337663,
+      "eval_loss": 1.5684095621109009,
+      "eval_runtime": 16.9217,
+      "eval_samples_per_second": 4.55,
+      "eval_steps_per_second": 1.182,
+      "step": 7490
+    },
+    {
+      "epoch": 107.00096618357487,
+      "grad_norm": 624.6681518554688,
+      "learning_rate": 3.059581320450886e-06,
+      "loss": 0.0527,
+      "step": 7500
+    },
+    {
+      "epoch": 107.00193236714976,
+      "grad_norm": 0.0003674910985864699,
+      "learning_rate": 3.0488459473966726e-06,
+      "loss": 0.0,
+      "step": 7510
+    },
+    {
+      "epoch": 107.00289855072464,
+      "grad_norm": 0.0003636605979409069,
+      "learning_rate": 3.038110574342459e-06,
+      "loss": 0.0,
+      "step": 7520
+    },
+    {
+      "epoch": 107.00386473429951,
+      "grad_norm": 0.0015479899011552334,
+      "learning_rate": 3.027375201288245e-06,
+      "loss": 0.0,
+      "step": 7530
+    },
+    {
+      "epoch": 107.0048309178744,
+      "grad_norm": 0.0007740533910691738,
+      "learning_rate": 3.0166398282340315e-06,
+      "loss": 0.0,
+      "step": 7540
+    },
+    {
+      "epoch": 107.00579710144928,
+      "grad_norm": 0.0003727949515450746,
+      "learning_rate": 3.005904455179818e-06,
+      "loss": 0.0001,
+      "step": 7550
+    },
+    {
+      "epoch": 107.00676328502415,
+      "grad_norm": 0.00039142067544162273,
+      "learning_rate": 2.995169082125604e-06,
+      "loss": 0.0,
+      "step": 7560
+    },
+    {
+      "epoch": 107.00676328502415,
+      "eval_accuracy": 0.8051948051948052,
+      "eval_loss": 1.7333017587661743,
+      "eval_runtime": 16.6874,
+      "eval_samples_per_second": 4.614,
+      "eval_steps_per_second": 1.199,
+      "step": 7560
+    },
+    {
+      "epoch": 108.00096618357487,
+      "grad_norm": 0.0002302713255630806,
+      "learning_rate": 2.9844337090713904e-06,
+      "loss": 0.0,
+      "step": 7570
+    },
+    {
+      "epoch": 108.00193236714976,
+      "grad_norm": 0.00029453245224431157,
+      "learning_rate": 2.973698336017177e-06,
+      "loss": 0.0001,
+      "step": 7580
+    },
+    {
+      "epoch": 108.00289855072464,
+      "grad_norm": 0.0043038977310061455,
+      "learning_rate": 2.962962962962963e-06,
+      "loss": 0.0,
+      "step": 7590
+    },
+    {
+      "epoch": 108.00386473429951,
+      "grad_norm": 292.66595458984375,
+      "learning_rate": 2.9522275899087494e-06,
+      "loss": 0.0253,
+      "step": 7600
+    },
+    {
+      "epoch": 108.0048309178744,
+      "grad_norm": 0.0004467540711630136,
+      "learning_rate": 2.941492216854536e-06,
+      "loss": 0.0,
+      "step": 7610
+    },
+    {
+      "epoch": 108.00579710144928,
+      "grad_norm": 0.0002925453591160476,
+      "learning_rate": 2.9307568438003227e-06,
+      "loss": 0.3068,
+      "step": 7620
+    },
+    {
+      "epoch": 108.00676328502415,
+      "grad_norm": 0.0020885635167360306,
+      "learning_rate": 2.9200214707461087e-06,
+      "loss": 0.0,
+      "step": 7630
+    },
+    {
+      "epoch": 108.00676328502415,
+      "eval_accuracy": 0.7922077922077922,
+      "eval_loss": 1.423266887664795,
+      "eval_runtime": 14.9029,
+      "eval_samples_per_second": 5.167,
+      "eval_steps_per_second": 1.342,
+      "step": 7630
+    },
+    {
+      "epoch": 109.00096618357487,
+      "grad_norm": 0.0003770174225792289,
+      "learning_rate": 2.909286097691895e-06,
+      "loss": 0.0,
+      "step": 7640
+    },
+    {
+      "epoch": 109.00193236714976,
+      "grad_norm": 319.81683349609375,
+      "learning_rate": 2.8985507246376816e-06,
+      "loss": 0.2953,
+      "step": 7650
+    },
+    {
+      "epoch": 109.00289855072464,
+      "grad_norm": 0.000277810642728582,
+      "learning_rate": 2.8878153515834676e-06,
+      "loss": 0.0,
+      "step": 7660
+    },
+    {
+      "epoch": 109.00386473429951,
+      "grad_norm": 0.0017440535593777895,
+      "learning_rate": 2.877079978529254e-06,
+      "loss": 0.0,
+      "step": 7670
+    },
+    {
+      "epoch": 109.0048309178744,
+      "grad_norm": 0.004614552017301321,
+      "learning_rate": 2.8663446054750405e-06,
+      "loss": 0.0,
+      "step": 7680
+    },
+    {
+      "epoch": 109.00579710144928,
+      "grad_norm": 0.0003517542500048876,
+      "learning_rate": 2.8556092324208266e-06,
+      "loss": 0.0,
+      "step": 7690
+    },
+    {
+      "epoch": 109.00676328502415,
+      "grad_norm": 0.0006136648589745164,
+      "learning_rate": 2.844873859366613e-06,
+      "loss": 0.0001,
+      "step": 7700
+    },
+    {
+      "epoch": 109.00676328502415,
+      "eval_accuracy": 0.7792207792207793,
+      "eval_loss": 1.7541829347610474,
+      "eval_runtime": 14.6741,
+      "eval_samples_per_second": 5.247,
+      "eval_steps_per_second": 1.363,
+      "step": 7700
+    },
+    {
+      "epoch": 110.00096618357487,
+      "grad_norm": 0.003632753388956189,
+      "learning_rate": 2.8341384863123995e-06,
+      "loss": 0.0,
+      "step": 7710
+    },
+    {
+      "epoch": 110.00193236714976,
+      "grad_norm": 0.0011343782534822822,
+      "learning_rate": 2.823403113258186e-06,
+      "loss": 0.0,
+      "step": 7720
+    },
+    {
+      "epoch": 110.00289855072464,
+      "grad_norm": 0.00032457453198730946,
+      "learning_rate": 2.812667740203972e-06,
+      "loss": 0.0172,
+      "step": 7730
+    },
+    {
+      "epoch": 110.00386473429951,
+      "grad_norm": 0.0008219513692893088,
+      "learning_rate": 2.801932367149759e-06,
+      "loss": 0.0,
+      "step": 7740
+    },
+    {
+      "epoch": 110.0048309178744,
+      "grad_norm": 0.00046765225124545395,
+      "learning_rate": 2.7911969940955453e-06,
+      "loss": 0.0,
+      "step": 7750
+    },
+    {
+      "epoch": 110.00579710144928,
+      "grad_norm": 0.006037019193172455,
+      "learning_rate": 2.7804616210413317e-06,
+      "loss": 0.0,
+      "step": 7760
+    },
+    {
+      "epoch": 110.00676328502415,
+      "grad_norm": 0.00038607208989560604,
+      "learning_rate": 2.7697262479871177e-06,
+      "loss": 0.0,
+      "step": 7770
+    },
+    {
+      "epoch": 110.00676328502415,
+      "eval_accuracy": 0.8051948051948052,
+      "eval_loss": 1.4554383754730225,
+      "eval_runtime": 18.1834,
+      "eval_samples_per_second": 4.235,
+      "eval_steps_per_second": 1.1,
+      "step": 7770
+    },
+    {
+      "epoch": 111.00096618357487,
+      "grad_norm": 0.003369076643139124,
+      "learning_rate": 2.758990874932904e-06,
+      "loss": 0.0,
+      "step": 7780
+    },
+    {
+      "epoch": 111.00193236714976,
+      "grad_norm": 0.00028180310619063675,
+      "learning_rate": 2.7482555018786907e-06,
+      "loss": 0.0,
+      "step": 7790
+    },
+    {
+      "epoch": 111.00289855072464,
+      "grad_norm": 0.00040791460196487606,
+      "learning_rate": 2.7375201288244767e-06,
+      "loss": 0.0,
+      "step": 7800
+    },
+    {
+      "epoch": 111.00386473429951,
+      "grad_norm": 0.2944580912590027,
+      "learning_rate": 2.726784755770263e-06,
+      "loss": 0.0001,
+      "step": 7810
+    },
+    {
+      "epoch": 111.0048309178744,
+      "grad_norm": 0.000506663927808404,
+      "learning_rate": 2.7160493827160496e-06,
+      "loss": 0.0,
+      "step": 7820
+    },
+    {
+      "epoch": 111.00579710144928,
+      "grad_norm": 0.0006107607623562217,
+      "learning_rate": 2.7053140096618356e-06,
+      "loss": 0.0,
+      "step": 7830
+    },
+    {
+      "epoch": 111.00676328502415,
+      "grad_norm": 0.0007616582443006337,
+      "learning_rate": 2.694578636607622e-06,
+      "loss": 0.0,
+      "step": 7840
+    },
+    {
+      "epoch": 111.00676328502415,
+      "eval_accuracy": 0.8571428571428571,
+      "eval_loss": 1.3537590503692627,
+      "eval_runtime": 25.5207,
+      "eval_samples_per_second": 3.017,
+      "eval_steps_per_second": 0.784,
+      "step": 7840
+    },
+    {
+      "epoch": 112.00096618357487,
+      "grad_norm": 0.00020077962835785002,
+      "learning_rate": 2.683843263553409e-06,
+      "loss": 0.0,
+      "step": 7850
+    },
+    {
+      "epoch": 112.00193236714976,
+      "grad_norm": 0.0007396721048280597,
+      "learning_rate": 2.6731078904991954e-06,
+      "loss": 0.0,
+      "step": 7860
+    },
+    {
+      "epoch": 112.00289855072464,
+      "grad_norm": 0.001002350589260459,
+      "learning_rate": 2.6623725174449814e-06,
+      "loss": 0.0,
+      "step": 7870
+    },
+    {
+      "epoch": 112.00386473429951,
+      "grad_norm": 0.0007556098862551153,
+      "learning_rate": 2.651637144390768e-06,
+      "loss": 0.0,
+      "step": 7880
+    },
+    {
+      "epoch": 112.0048309178744,
+      "grad_norm": 0.0019177094800397754,
+      "learning_rate": 2.6409017713365543e-06,
+      "loss": 0.0,
+      "step": 7890
+    },
+    {
+      "epoch": 112.00579710144928,
+      "grad_norm": 0.004157165996730328,
+      "learning_rate": 2.6301663982823408e-06,
+      "loss": 0.0,
+      "step": 7900
+    },
+    {
+      "epoch": 112.00676328502415,
+      "grad_norm": 0.0004525023396126926,
+      "learning_rate": 2.6194310252281268e-06,
+      "loss": 0.0,
+      "step": 7910
+    },
+    {
+      "epoch": 112.00676328502415,
+      "eval_accuracy": 0.8571428571428571,
+      "eval_loss": 1.4164801836013794,
+      "eval_runtime": 14.6287,
+      "eval_samples_per_second": 5.264,
+      "eval_steps_per_second": 1.367,
+      "step": 7910
+    },
+    {
+      "epoch": 113.00096618357487,
+      "grad_norm": 0.0003341760893817991,
+      "learning_rate": 2.6086956521739132e-06,
+      "loss": 0.0,
+      "step": 7920
+    },
+    {
+      "epoch": 113.00193236714976,
+      "grad_norm": 0.0007722051814198494,
+      "learning_rate": 2.5979602791196997e-06,
+      "loss": 0.0,
+      "step": 7930
+    },
+    {
+      "epoch": 113.00289855072464,
+      "grad_norm": 0.0003543299390003085,
+      "learning_rate": 2.5872249060654857e-06,
+      "loss": 0.0,
+      "step": 7940
+    },
+    {
+      "epoch": 113.00386473429951,
+      "grad_norm": 0.003534851362928748,
+      "learning_rate": 2.576489533011272e-06,
+      "loss": 0.0,
+      "step": 7950
+    },
+    {
+      "epoch": 113.0048309178744,
+      "grad_norm": 0.0003302747500129044,
+      "learning_rate": 2.5657541599570586e-06,
+      "loss": 0.0,
+      "step": 7960
+    },
+    {
+      "epoch": 113.00579710144928,
+      "grad_norm": 0.00041915668407455087,
+      "learning_rate": 2.5550187869028455e-06,
+      "loss": 0.0,
+      "step": 7970
+    },
+    {
+      "epoch": 113.00676328502415,
+      "grad_norm": 0.02611609548330307,
+      "learning_rate": 2.5442834138486315e-06,
+      "loss": 0.0,
+      "step": 7980
+    },
+    {
+      "epoch": 113.00676328502415,
+      "eval_accuracy": 0.8571428571428571,
+      "eval_loss": 1.4228990077972412,
+      "eval_runtime": 14.244,
+      "eval_samples_per_second": 5.406,
+      "eval_steps_per_second": 1.404,
+      "step": 7980
+    },
+    {
+      "epoch": 114.00096618357487,
+      "grad_norm": 0.0005513859214261174,
+      "learning_rate": 2.533548040794418e-06,
+      "loss": 0.0,
+      "step": 7990
+    },
+    {
+      "epoch": 114.00193236714976,
+      "grad_norm": 0.0006181516218930483,
+      "learning_rate": 2.5228126677402044e-06,
+      "loss": 0.0,
+      "step": 8000
+    },
+    {
+      "epoch": 114.00289855072464,
+      "grad_norm": 0.0005718721658922732,
+      "learning_rate": 2.5120772946859904e-06,
+      "loss": 0.0,
+      "step": 8010
+    },
+    {
+      "epoch": 114.00386473429951,
+      "grad_norm": 0.0015514239203184843,
+      "learning_rate": 2.501341921631777e-06,
+      "loss": 0.0,
+      "step": 8020
+    },
+    {
+      "epoch": 114.0048309178744,
+      "grad_norm": 0.0005155864637345076,
+      "learning_rate": 2.4906065485775633e-06,
+      "loss": 0.0,
+      "step": 8030
+    },
+    {
+      "epoch": 114.00579710144928,
+      "grad_norm": 0.0005761452484875917,
+      "learning_rate": 2.4798711755233494e-06,
+      "loss": 0.0,
+      "step": 8040
+    },
+    {
+      "epoch": 114.00676328502415,
+      "grad_norm": 0.0003508743247948587,
+      "learning_rate": 2.469135802469136e-06,
+      "loss": 0.0,
+      "step": 8050
+    },
+    {
+      "epoch": 114.00676328502415,
+      "eval_accuracy": 0.8571428571428571,
+      "eval_loss": 1.4191442728042603,
+      "eval_runtime": 14.3381,
+      "eval_samples_per_second": 5.37,
+      "eval_steps_per_second": 1.395,
+      "step": 8050
+    },
+    {
+      "epoch": 115.00096618357487,
+      "grad_norm": 0.00038109711022116244,
+      "learning_rate": 2.4584004294149223e-06,
+      "loss": 0.0,
+      "step": 8060
+    },
+    {
+      "epoch": 115.00193236714976,
+      "grad_norm": 0.0004090418224222958,
+      "learning_rate": 2.4476650563607087e-06,
+      "loss": 0.0,
+      "step": 8070
+    },
+    {
+      "epoch": 115.00289855072464,
+      "grad_norm": 0.00023632495140191168,
+      "learning_rate": 2.436929683306495e-06,
+      "loss": 0.0,
+      "step": 8080
+    },
+    {
+      "epoch": 115.00386473429951,
+      "grad_norm": 0.00041399948531761765,
+      "learning_rate": 2.426194310252281e-06,
+      "loss": 0.0,
+      "step": 8090
+    },
+    {
+      "epoch": 115.0048309178744,
+      "grad_norm": 0.0004869647673331201,
+      "learning_rate": 2.4154589371980677e-06,
+      "loss": 0.0,
+      "step": 8100
+    },
+    {
+      "epoch": 115.00579710144928,
+      "grad_norm": 0.0002684770443011075,
+      "learning_rate": 2.4047235641438545e-06,
+      "loss": 0.0,
+      "step": 8110
+    },
+    {
+      "epoch": 115.00676328502415,
+      "grad_norm": 0.005061628762632608,
+      "learning_rate": 2.3939881910896406e-06,
+      "loss": 0.0,
+      "step": 8120
+    },
+    {
+      "epoch": 115.00676328502415,
+      "eval_accuracy": 0.8571428571428571,
+      "eval_loss": 1.4363770484924316,
+      "eval_runtime": 14.3464,
+      "eval_samples_per_second": 5.367,
+      "eval_steps_per_second": 1.394,
+      "step": 8120
+    },
+    {
+      "epoch": 116.00096618357487,
+      "grad_norm": 0.00040192605229094625,
+      "learning_rate": 2.383252818035427e-06,
+      "loss": 0.0,
+      "step": 8130
+    },
+    {
+      "epoch": 116.00193236714976,
+      "grad_norm": 0.0004934113239869475,
+      "learning_rate": 2.3725174449812135e-06,
+      "loss": 0.0,
+      "step": 8140
+    },
+    {
+      "epoch": 116.00289855072464,
+      "grad_norm": 0.0008514419896528125,
+      "learning_rate": 2.3617820719269995e-06,
+      "loss": 0.0,
+      "step": 8150
+    },
+    {
+      "epoch": 116.00386473429951,
+      "grad_norm": 0.01015880610793829,
+      "learning_rate": 2.351046698872786e-06,
+      "loss": 0.0,
+      "step": 8160
+    },
+    {
+      "epoch": 116.0048309178744,
+      "grad_norm": 0.0004901864449493587,
+      "learning_rate": 2.3403113258185724e-06,
+      "loss": 0.0,
+      "step": 8170
+    },
+    {
+      "epoch": 116.00579710144928,
+      "grad_norm": 0.00022043399803806096,
+      "learning_rate": 2.329575952764359e-06,
+      "loss": 0.0,
+      "step": 8180
+    },
+    {
+      "epoch": 116.00676328502415,
+      "grad_norm": 0.00031955583835951984,
+      "learning_rate": 2.3188405797101453e-06,
+      "loss": 0.0,
+      "step": 8190
+    },
+    {
+      "epoch": 116.00676328502415,
+      "eval_accuracy": 0.8311688311688312,
+      "eval_loss": 1.457450032234192,
+      "eval_runtime": 16.0943,
+      "eval_samples_per_second": 4.784,
+      "eval_steps_per_second": 1.243,
+      "step": 8190
+    },
+    {
+      "epoch": 117.00096618357487,
+      "grad_norm": 0.0003374749212525785,
+      "learning_rate": 2.3081052066559313e-06,
+      "loss": 0.0,
+      "step": 8200
+    },
+    {
+      "epoch": 117.00193236714976,
+      "grad_norm": 0.0002894297649618238,
+      "learning_rate": 2.2973698336017178e-06,
+      "loss": 0.0,
+      "step": 8210
+    },
+    {
+      "epoch": 117.00289855072464,
+      "grad_norm": 0.00022455077851191163,
+      "learning_rate": 2.2866344605475042e-06,
+      "loss": 0.0,
+      "step": 8220
+    },
+    {
+      "epoch": 117.00386473429951,
+      "grad_norm": 0.00020910402236040682,
+      "learning_rate": 2.2758990874932907e-06,
+      "loss": 0.0,
+      "step": 8230
+    },
+    {
+      "epoch": 117.0048309178744,
+      "grad_norm": 0.00024087671772576869,
+      "learning_rate": 2.265163714439077e-06,
+      "loss": 0.0,
+      "step": 8240
+    },
+    {
+      "epoch": 117.00579710144928,
+      "grad_norm": 0.000636860728263855,
+      "learning_rate": 2.254428341384863e-06,
+      "loss": 0.0,
+      "step": 8250
+    },
+    {
+      "epoch": 117.00676328502415,
+      "grad_norm": 0.0007749819196760654,
+      "learning_rate": 2.2436929683306496e-06,
+      "loss": 0.0,
+      "step": 8260
+    },
+    {
+      "epoch": 117.00676328502415,
+      "eval_accuracy": 0.8311688311688312,
+      "eval_loss": 1.4639664888381958,
+      "eval_runtime": 15.7612,
+      "eval_samples_per_second": 4.885,
+      "eval_steps_per_second": 1.269,
+      "step": 8260
+    },
+    {
+      "epoch": 118.00096618357487,
+      "grad_norm": 0.0007536153425462544,
+      "learning_rate": 2.232957595276436e-06,
+      "loss": 0.0,
+      "step": 8270
+    },
+    {
+      "epoch": 118.00193236714976,
+      "grad_norm": 0.00032429906423203647,
+      "learning_rate": 2.222222222222222e-06,
+      "loss": 0.0,
+      "step": 8280
+    },
+    {
+      "epoch": 118.00289855072464,
+      "grad_norm": 0.0006189732230268419,
+      "learning_rate": 2.211486849168009e-06,
+      "loss": 0.0,
+      "step": 8290
+    },
+    {
+      "epoch": 118.00386473429951,
+      "grad_norm": 0.0018710630247369409,
+      "learning_rate": 2.2007514761137954e-06,
+      "loss": 0.0,
+      "step": 8300
+    },
+    {
+      "epoch": 118.0048309178744,
+      "grad_norm": 0.00020172024960629642,
+      "learning_rate": 2.1900161030595814e-06,
+      "loss": 0.0,
+      "step": 8310
+    },
+    {
+      "epoch": 118.00579710144928,
+      "grad_norm": 0.0002452440094202757,
+      "learning_rate": 2.179280730005368e-06,
+      "loss": 0.0,
+      "step": 8320
+    },
+    {
+      "epoch": 118.00676328502415,
+      "grad_norm": 0.0005612657987512648,
+      "learning_rate": 2.1685453569511543e-06,
+      "loss": 0.0,
+      "step": 8330
+    },
+    {
+      "epoch": 118.00676328502415,
+      "eval_accuracy": 0.8311688311688312,
+      "eval_loss": 1.480674386024475,
+      "eval_runtime": 16.8552,
+      "eval_samples_per_second": 4.568,
+      "eval_steps_per_second": 1.187,
+      "step": 8330
+    },
+    {
+      "epoch": 119.00096618357487,
+      "grad_norm": 0.004412544891238213,
+      "learning_rate": 2.1578099838969404e-06,
+      "loss": 0.0,
+      "step": 8340
+    },
+    {
+      "epoch": 119.00193236714976,
+      "grad_norm": 0.0001875559682957828,
+      "learning_rate": 2.1470746108427272e-06,
+      "loss": 0.0,
+      "step": 8350
+    },
+    {
+      "epoch": 119.00289855072464,
+      "grad_norm": 0.00025120144709944725,
+      "learning_rate": 2.1363392377885133e-06,
+      "loss": 0.0,
+      "step": 8360
+    },
+    {
+      "epoch": 119.00386473429951,
+      "grad_norm": 0.0002451414766255766,
+      "learning_rate": 2.1256038647342997e-06,
+      "loss": 0.0,
+      "step": 8370
+    },
+    {
+      "epoch": 119.0048309178744,
+      "grad_norm": 0.00071542157093063,
+      "learning_rate": 2.114868491680086e-06,
+      "loss": 0.0,
+      "step": 8380
+    },
+    {
+      "epoch": 119.00579710144928,
+      "grad_norm": 0.0005497967358678579,
+      "learning_rate": 2.104133118625872e-06,
+      "loss": 0.0,
+      "step": 8390
+    },
+    {
+      "epoch": 119.00676328502415,
+      "grad_norm": 0.00022419539163820446,
+      "learning_rate": 2.093397745571659e-06,
+      "loss": 0.0,
+      "step": 8400
+    },
+    {
+      "epoch": 119.00676328502415,
+      "eval_accuracy": 0.8311688311688312,
+      "eval_loss": 1.5029710531234741,
+      "eval_runtime": 16.8065,
+      "eval_samples_per_second": 4.582,
+      "eval_steps_per_second": 1.19,
+      "step": 8400
+    },
+    {
+      "epoch": 120.00096618357487,
+      "grad_norm": 0.00019987497944384813,
+      "learning_rate": 2.082662372517445e-06,
+      "loss": 0.0,
+      "step": 8410
+    },
+    {
+      "epoch": 120.00193236714976,
+      "grad_norm": 0.00037460497696883976,
+      "learning_rate": 2.0719269994632315e-06,
+      "loss": 0.0,
+      "step": 8420
+    },
+    {
+      "epoch": 120.00289855072464,
+      "grad_norm": 0.0003980303881689906,
+      "learning_rate": 2.061191626409018e-06,
+      "loss": 0.0,
+      "step": 8430
+    },
+    {
+      "epoch": 120.00386473429951,
+      "grad_norm": 0.00016923666407819837,
+      "learning_rate": 2.050456253354804e-06,
+      "loss": 0.0,
+      "step": 8440
+    },
+    {
+      "epoch": 120.0048309178744,
+      "grad_norm": 0.0003269107255619019,
+      "learning_rate": 2.0397208803005905e-06,
+      "loss": 0.0,
+      "step": 8450
+    },
+    {
+      "epoch": 120.00579710144928,
+      "grad_norm": 0.0008120771963149309,
+      "learning_rate": 2.028985507246377e-06,
+      "loss": 0.0,
+      "step": 8460
+    },
+    {
+      "epoch": 120.00676328502415,
+      "grad_norm": 0.0005975480889901519,
+      "learning_rate": 2.0182501341921634e-06,
+      "loss": 0.0,
+      "step": 8470
+    },
+    {
+      "epoch": 120.00676328502415,
+      "eval_accuracy": 0.8311688311688312,
+      "eval_loss": 1.518750786781311,
+      "eval_runtime": 17.4768,
+      "eval_samples_per_second": 4.406,
+      "eval_steps_per_second": 1.144,
+      "step": 8470
+    },
+    {
+      "epoch": 121.00096618357487,
+      "grad_norm": 0.00044110111775808036,
+      "learning_rate": 2.00751476113795e-06,
+      "loss": 0.0,
+      "step": 8480
+    },
+    {
+      "epoch": 121.00193236714976,
+      "grad_norm": 0.00020110349578317255,
+      "learning_rate": 1.9967793880837363e-06,
+      "loss": 0.0,
+      "step": 8490
+    },
+    {
+      "epoch": 121.00289855072464,
+      "grad_norm": 0.000270154036115855,
+      "learning_rate": 1.9860440150295223e-06,
+      "loss": 0.0,
+      "step": 8500
+    },
+    {
+      "epoch": 121.00386473429951,
+      "grad_norm": 0.0002893579949159175,
+      "learning_rate": 1.9753086419753087e-06,
+      "loss": 0.0,
+      "step": 8510
+    },
+    {
+      "epoch": 121.0048309178744,
+      "grad_norm": 0.001373679144307971,
+      "learning_rate": 1.964573268921095e-06,
+      "loss": 0.0,
+      "step": 8520
+    },
+    {
+      "epoch": 121.00579710144928,
+      "grad_norm": 0.0003092541010119021,
+      "learning_rate": 1.9538378958668816e-06,
+      "loss": 0.0,
+      "step": 8530
+    },
+    {
+      "epoch": 121.00676328502415,
+      "grad_norm": 0.00024617547751404345,
+      "learning_rate": 1.943102522812668e-06,
+      "loss": 0.0,
+      "step": 8540
+    },
+    {
+      "epoch": 121.00676328502415,
+      "eval_accuracy": 0.8181818181818182,
+      "eval_loss": 1.5642110109329224,
+      "eval_runtime": 16.6862,
+      "eval_samples_per_second": 4.615,
+      "eval_steps_per_second": 1.199,
+      "step": 8540
+    },
+    {
+      "epoch": 122.00096618357487,
+      "grad_norm": 0.00046582010691054165,
+      "learning_rate": 1.932367149758454e-06,
+      "loss": 0.0,
+      "step": 8550
+    },
+    {
+      "epoch": 122.00193236714976,
+      "grad_norm": 0.00022211925534065813,
+      "learning_rate": 1.9216317767042406e-06,
+      "loss": 0.0,
+      "step": 8560
+    },
+    {
+      "epoch": 122.00289855072464,
+      "grad_norm": 0.0002465721336193383,
+      "learning_rate": 1.910896403650027e-06,
+      "loss": 0.0,
+      "step": 8570
+    },
+    {
+      "epoch": 122.00386473429951,
+      "grad_norm": 0.0007177837542258203,
+      "learning_rate": 1.9001610305958135e-06,
+      "loss": 0.0,
+      "step": 8580
+    },
+    {
+      "epoch": 122.0048309178744,
+      "grad_norm": 0.0001443681976525113,
+      "learning_rate": 1.8894256575415997e-06,
+      "loss": 0.0,
+      "step": 8590
+    },
+    {
+      "epoch": 122.00579710144928,
+      "grad_norm": 0.00030018811230547726,
+      "learning_rate": 1.8786902844873862e-06,
+      "loss": 0.0,
+      "step": 8600
+    },
+    {
+      "epoch": 122.00676328502415,
+      "grad_norm": 0.00017983531870413572,
+      "learning_rate": 1.8679549114331724e-06,
+      "loss": 0.0,
+      "step": 8610
+    },
+    {
+      "epoch": 122.00676328502415,
+      "eval_accuracy": 0.8181818181818182,
+      "eval_loss": 1.5662660598754883,
+      "eval_runtime": 16.7576,
+      "eval_samples_per_second": 4.595,
+      "eval_steps_per_second": 1.193,
+      "step": 8610
+    },
+    {
+      "epoch": 123.00096618357487,
+      "grad_norm": 0.00021314997866284102,
+      "learning_rate": 1.8572195383789586e-06,
+      "loss": 0.0,
+      "step": 8620
+    },
+    {
+      "epoch": 123.00193236714976,
+      "grad_norm": 0.00018588421517051756,
+      "learning_rate": 1.846484165324745e-06,
+      "loss": 0.0,
+      "step": 8630
+    },
+    {
+      "epoch": 123.00289855072464,
+      "grad_norm": 0.0005342851509340107,
+      "learning_rate": 1.8357487922705318e-06,
+      "loss": 0.0,
+      "step": 8640
+    },
+    {
+      "epoch": 123.00386473429951,
+      "grad_norm": 0.00028879634919576347,
+      "learning_rate": 1.825013419216318e-06,
+      "loss": 0.0,
+      "step": 8650
+    },
+    {
+      "epoch": 123.0048309178744,
+      "grad_norm": 0.0004933462478220463,
+      "learning_rate": 1.8142780461621042e-06,
+      "loss": 0.0,
+      "step": 8660
+    },
+    {
+      "epoch": 123.00579710144928,
+      "grad_norm": 0.0003157542960252613,
+      "learning_rate": 1.8035426731078907e-06,
+      "loss": 0.0,
+      "step": 8670
+    },
+    {
+      "epoch": 123.00676328502415,
+      "grad_norm": 0.00021309992007445544,
+      "learning_rate": 1.792807300053677e-06,
+      "loss": 0.0,
+      "step": 8680
+    },
+    {
+      "epoch": 123.00676328502415,
+      "eval_accuracy": 0.8181818181818182,
+      "eval_loss": 1.5686160326004028,
+      "eval_runtime": 17.2629,
+      "eval_samples_per_second": 4.46,
+      "eval_steps_per_second": 1.159,
+      "step": 8680
+    },
+    {
+      "epoch": 124.00096618357487,
+      "grad_norm": 0.24094708263874054,
+      "learning_rate": 1.7820719269994632e-06,
+      "loss": 0.0,
+      "step": 8690
+    },
+    {
+      "epoch": 124.00193236714976,
+      "grad_norm": 0.0003956431755796075,
+      "learning_rate": 1.7713365539452498e-06,
+      "loss": 0.0,
+      "step": 8700
+    },
+    {
+      "epoch": 124.00289855072464,
+      "grad_norm": 0.00026967411395162344,
+      "learning_rate": 1.760601180891036e-06,
+      "loss": 0.0,
+      "step": 8710
+    },
+    {
+      "epoch": 124.00386473429951,
+      "grad_norm": 0.0006772116175852716,
+      "learning_rate": 1.7498658078368225e-06,
+      "loss": 0.0,
+      "step": 8720
+    },
+    {
+      "epoch": 124.0048309178744,
+      "grad_norm": 0.0004103895334992558,
+      "learning_rate": 1.7391304347826088e-06,
+      "loss": 0.0,
+      "step": 8730
+    },
+    {
+      "epoch": 124.00579710144928,
+      "grad_norm": 0.00021756005298811942,
+      "learning_rate": 1.7283950617283952e-06,
+      "loss": 0.0,
+      "step": 8740
+    },
+    {
+      "epoch": 124.00676328502415,
+      "grad_norm": 0.0003636313776951283,
+      "learning_rate": 1.7176596886741817e-06,
+      "loss": 0.0,
+      "step": 8750
+    },
+    {
+      "epoch": 124.00676328502415,
+      "eval_accuracy": 0.8571428571428571,
+      "eval_loss": 1.4284155368804932,
+      "eval_runtime": 14.6131,
+      "eval_samples_per_second": 5.269,
+      "eval_steps_per_second": 1.369,
+      "step": 8750
+    },
+    {
+      "epoch": 125.00096618357487,
+      "grad_norm": 0.0002465183206368238,
+      "learning_rate": 1.7069243156199681e-06,
+      "loss": 0.0,
+      "step": 8760
+    },
+    {
+      "epoch": 125.00193236714976,
+      "grad_norm": 0.0002711952547542751,
+      "learning_rate": 1.6961889425657543e-06,
+      "loss": 0.0,
+      "step": 8770
+    },
+    {
+      "epoch": 125.00289855072464,
+      "grad_norm": 0.0005906698643229902,
+      "learning_rate": 1.6854535695115406e-06,
+      "loss": 0.0,
+      "step": 8780
+    },
+    {
+      "epoch": 125.00386473429951,
+      "grad_norm": 0.0008631067466922104,
+      "learning_rate": 1.674718196457327e-06,
+      "loss": 0.0,
+      "step": 8790
+    },
+    {
+      "epoch": 125.0048309178744,
+      "grad_norm": 0.00018012075452134013,
+      "learning_rate": 1.6639828234031133e-06,
+      "loss": 0.0,
+      "step": 8800
+    },
+    {
+      "epoch": 125.00579710144928,
+      "grad_norm": 0.00032579831895418465,
+      "learning_rate": 1.6532474503489e-06,
+      "loss": 0.0,
+      "step": 8810
+    },
+    {
+      "epoch": 125.00676328502415,
+      "grad_norm": 0.00019309454364702106,
+      "learning_rate": 1.6425120772946862e-06,
+      "loss": 0.0,
+      "step": 8820
+    },
+    {
+      "epoch": 125.00676328502415,
+      "eval_accuracy": 0.8571428571428571,
+      "eval_loss": 1.4352055788040161,
+      "eval_runtime": 15.1216,
+      "eval_samples_per_second": 5.092,
+      "eval_steps_per_second": 1.323,
+      "step": 8820
+    },
+    {
+      "epoch": 126.00096618357487,
+      "grad_norm": 0.00021101922902744263,
+      "learning_rate": 1.6317767042404724e-06,
+      "loss": 0.0,
+      "step": 8830
+    },
+    {
+      "epoch": 126.00193236714976,
+      "grad_norm": 0.00024815279175527394,
+      "learning_rate": 1.6210413311862589e-06,
+      "loss": 0.0,
+      "step": 8840
+    },
+    {
+      "epoch": 126.00289855072464,
+      "grad_norm": 0.00025953681324608624,
+      "learning_rate": 1.610305958132045e-06,
+      "loss": 0.0,
+      "step": 8850
+    },
+    {
+      "epoch": 126.00386473429951,
+      "grad_norm": 0.00018137147708330303,
+      "learning_rate": 1.5995705850778316e-06,
+      "loss": 0.0,
+      "step": 8860
+    },
+    {
+      "epoch": 126.0048309178744,
+      "grad_norm": 0.0001805800711736083,
+      "learning_rate": 1.588835212023618e-06,
+      "loss": 0.0,
+      "step": 8870
+    },
+    {
+      "epoch": 126.00579710144928,
+      "grad_norm": 0.0001789192610885948,
+      "learning_rate": 1.5780998389694045e-06,
+      "loss": 0.0,
+      "step": 8880
+    },
+    {
+      "epoch": 126.00676328502415,
+      "grad_norm": 0.0029380114283412695,
+      "learning_rate": 1.5673644659151907e-06,
+      "loss": 0.0,
+      "step": 8890
+    },
+    {
+      "epoch": 126.00676328502415,
+      "eval_accuracy": 0.8571428571428571,
+      "eval_loss": 1.4391511678695679,
+      "eval_runtime": 14.8548,
+      "eval_samples_per_second": 5.184,
+      "eval_steps_per_second": 1.346,
+      "step": 8890
+    },
+    {
+      "epoch": 127.00096618357487,
+      "grad_norm": 0.00032747775549069047,
+      "learning_rate": 1.556629092860977e-06,
+      "loss": 0.0,
+      "step": 8900
+    },
+    {
+      "epoch": 127.00193236714976,
+      "grad_norm": 0.0001772013056324795,
+      "learning_rate": 1.5458937198067634e-06,
+      "loss": 0.0,
+      "step": 8910
+    },
+    {
+      "epoch": 127.00289855072464,
+      "grad_norm": 0.0002507749304641038,
+      "learning_rate": 1.5351583467525496e-06,
+      "loss": 0.0,
+      "step": 8920
+    },
+    {
+      "epoch": 127.00386473429951,
+      "grad_norm": 0.00020645925542339683,
+      "learning_rate": 1.5244229736983363e-06,
+      "loss": 0.0,
+      "step": 8930
+    },
+    {
+      "epoch": 127.0048309178744,
+      "grad_norm": 0.0002455931098666042,
+      "learning_rate": 1.5136876006441225e-06,
+      "loss": 0.0,
+      "step": 8940
+    },
+    {
+      "epoch": 127.00579710144928,
+      "grad_norm": 0.0001611367188161239,
+      "learning_rate": 1.502952227589909e-06,
+      "loss": 0.0,
+      "step": 8950
+    },
+    {
+      "epoch": 127.00676328502415,
+      "grad_norm": 0.00029621319845318794,
+      "learning_rate": 1.4922168545356952e-06,
+      "loss": 0.0,
+      "step": 8960
+    },
+    {
+      "epoch": 127.00676328502415,
+      "eval_accuracy": 0.8441558441558441,
+      "eval_loss": 1.5199832916259766,
+      "eval_runtime": 17.1653,
+      "eval_samples_per_second": 4.486,
+      "eval_steps_per_second": 1.165,
+      "step": 8960
+    },
+    {
+      "epoch": 128.0009661835749,
+      "grad_norm": 0.00026790890842676163,
+      "learning_rate": 1.4814814814814815e-06,
+      "loss": 0.0,
+      "step": 8970
+    },
+    {
+      "epoch": 128.00193236714975,
+      "grad_norm": 0.0002922885469160974,
+      "learning_rate": 1.470746108427268e-06,
+      "loss": 0.0,
+      "step": 8980
+    },
+    {
+      "epoch": 128.00289855072464,
+      "grad_norm": 0.00022152680321596563,
+      "learning_rate": 1.4600107353730544e-06,
+      "loss": 0.0,
+      "step": 8990
+    },
+    {
+      "epoch": 128.00386473429953,
+      "grad_norm": 0.0002079719997709617,
+      "learning_rate": 1.4492753623188408e-06,
+      "loss": 0.0,
+      "step": 9000
+    },
+    {
+      "epoch": 128.0048309178744,
+      "grad_norm": 0.0004370302485767752,
+      "learning_rate": 1.438539989264627e-06,
+      "loss": 0.0,
+      "step": 9010
+    },
+    {
+      "epoch": 128.00579710144928,
+      "grad_norm": 0.00025471995468251407,
+      "learning_rate": 1.4278046162104133e-06,
+      "loss": 0.0,
+      "step": 9020
+    },
+    {
+      "epoch": 128.00676328502416,
+      "grad_norm": 0.00036597729194909334,
+      "learning_rate": 1.4170692431561997e-06,
+      "loss": 0.0,
+      "step": 9030
+    },
+    {
+      "epoch": 128.00676328502416,
+      "eval_accuracy": 0.8441558441558441,
+      "eval_loss": 1.5243946313858032,
+      "eval_runtime": 19.1328,
+      "eval_samples_per_second": 4.025,
+      "eval_steps_per_second": 1.045,
+      "step": 9030
+    },
+    {
+      "epoch": 129.0009661835749,
+      "grad_norm": 0.0020494377240538597,
+      "learning_rate": 1.406333870101986e-06,
+      "loss": 0.0,
+      "step": 9040
+    },
+    {
+      "epoch": 129.00193236714975,
+      "grad_norm": 0.0005171003867872059,
+      "learning_rate": 1.3955984970477726e-06,
+      "loss": 0.0,
+      "step": 9050
+    },
+    {
+      "epoch": 129.00289855072464,
+      "grad_norm": 0.0006602657376788557,
+      "learning_rate": 1.3848631239935589e-06,
+      "loss": 0.0,
+      "step": 9060
+    },
+    {
+      "epoch": 129.00386473429953,
+      "grad_norm": 0.000498250825330615,
+      "learning_rate": 1.3741277509393453e-06,
+      "loss": 0.0,
+      "step": 9070
+    },
+    {
+      "epoch": 129.0048309178744,
+      "grad_norm": 0.00011794278543675318,
+      "learning_rate": 1.3633923778851316e-06,
+      "loss": 0.0,
+      "step": 9080
+    },
+    {
+      "epoch": 129.00579710144928,
+      "grad_norm": 0.00023995035735424608,
+      "learning_rate": 1.3526570048309178e-06,
+      "loss": 0.0,
+      "step": 9090
+    },
+    {
+      "epoch": 129.00676328502416,
+      "grad_norm": 0.00020481545652728528,
+      "learning_rate": 1.3419216317767045e-06,
+      "loss": 0.0,
+      "step": 9100
+    },
+    {
+      "epoch": 129.00676328502416,
+      "eval_accuracy": 0.8441558441558441,
+      "eval_loss": 1.5282416343688965,
+      "eval_runtime": 20.8777,
+      "eval_samples_per_second": 3.688,
+      "eval_steps_per_second": 0.958,
+      "step": 9100
+    },
+    {
+      "epoch": 130.0009661835749,
+      "grad_norm": 0.00041437524487264454,
+      "learning_rate": 1.3311862587224907e-06,
+      "loss": 0.0,
+      "step": 9110
+    },
+    {
+      "epoch": 130.00193236714975,
+      "grad_norm": 0.00026965918368659914,
+      "learning_rate": 1.3204508856682772e-06,
+      "loss": 0.0,
+      "step": 9120
+    },
+    {
+      "epoch": 130.00289855072464,
+      "grad_norm": 0.00022876296134199947,
+      "learning_rate": 1.3097155126140634e-06,
+      "loss": 0.0,
+      "step": 9130
+    },
+    {
+      "epoch": 130.00386473429953,
+      "grad_norm": 0.0004432128043845296,
+      "learning_rate": 1.2989801395598498e-06,
+      "loss": 0.0,
+      "step": 9140
+    },
+    {
+      "epoch": 130.0048309178744,
+      "grad_norm": 0.00015545626229140908,
+      "learning_rate": 1.288244766505636e-06,
+      "loss": 0.0,
+      "step": 9150
+    },
+    {
+      "epoch": 130.00579710144928,
+      "grad_norm": 0.00017510508769191802,
+      "learning_rate": 1.2775093934514227e-06,
+      "loss": 0.0,
+      "step": 9160
+    },
+    {
+      "epoch": 130.00676328502416,
+      "grad_norm": 0.00014436942001339048,
+      "learning_rate": 1.266774020397209e-06,
+      "loss": 0.0,
+      "step": 9170
+    },
+    {
+      "epoch": 130.00676328502416,
+      "eval_accuracy": 0.8441558441558441,
+      "eval_loss": 1.5338480472564697,
+      "eval_runtime": 21.879,
+      "eval_samples_per_second": 3.519,
+      "eval_steps_per_second": 0.914,
+      "step": 9170
+    },
+    {
+      "epoch": 131.0009661835749,
+      "grad_norm": 0.0001512864837422967,
+      "learning_rate": 1.2560386473429952e-06,
+      "loss": 0.0,
+      "step": 9180
+    },
+    {
+      "epoch": 131.00193236714975,
+      "grad_norm": 0.0008248549420386553,
+      "learning_rate": 1.2453032742887817e-06,
+      "loss": 0.0,
+      "step": 9190
+    },
+    {
+      "epoch": 131.00289855072464,
+      "grad_norm": 0.00021860776178073138,
+      "learning_rate": 1.234567901234568e-06,
+      "loss": 0.0,
+      "step": 9200
+    },
+    {
+      "epoch": 131.00386473429953,
+      "grad_norm": 0.00015075539704412222,
+      "learning_rate": 1.2238325281803544e-06,
+      "loss": 0.0,
+      "step": 9210
+    },
+    {
+      "epoch": 131.0048309178744,
+      "grad_norm": 0.00016689079347997904,
+      "learning_rate": 1.2130971551261406e-06,
+      "loss": 0.0,
+      "step": 9220
+    },
+    {
+      "epoch": 131.00579710144928,
+      "grad_norm": 0.00032013689633458853,
+      "learning_rate": 1.2023617820719273e-06,
+      "loss": 0.0,
+      "step": 9230
+    },
+    {
+      "epoch": 131.00676328502416,
+      "grad_norm": 0.0002864132111426443,
+      "learning_rate": 1.1916264090177135e-06,
+      "loss": 0.0,
+      "step": 9240
+    },
+    {
+      "epoch": 131.00676328502416,
+      "eval_accuracy": 0.8441558441558441,
+      "eval_loss": 1.5488603115081787,
+      "eval_runtime": 33.1783,
+      "eval_samples_per_second": 2.321,
+      "eval_steps_per_second": 0.603,
+      "step": 9240
+    },
+    {
+      "epoch": 132.0009661835749,
+      "grad_norm": 0.00016558170318603516,
+      "learning_rate": 1.1808910359634997e-06,
+      "loss": 0.0,
+      "step": 9250
+    },
+    {
+      "epoch": 132.00193236714975,
+      "grad_norm": 0.0001805421052267775,
+      "learning_rate": 1.1701556629092862e-06,
+      "loss": 0.0,
+      "step": 9260
+    },
+    {
+      "epoch": 132.00289855072464,
+      "grad_norm": 0.00022684381110593677,
+      "learning_rate": 1.1594202898550726e-06,
+      "loss": 0.0,
+      "step": 9270
+    },
+    {
+      "epoch": 132.00386473429953,
+      "grad_norm": 0.00018865799938794225,
+      "learning_rate": 1.1486849168008589e-06,
+      "loss": 0.0,
+      "step": 9280
+    },
+    {
+      "epoch": 132.0048309178744,
+      "grad_norm": 0.0002286021481268108,
+      "learning_rate": 1.1379495437466453e-06,
+      "loss": 0.0,
+      "step": 9290
+    },
+    {
+      "epoch": 132.00579710144928,
+      "grad_norm": 0.0002894134959205985,
+      "learning_rate": 1.1272141706924316e-06,
+      "loss": 0.0,
+      "step": 9300
+    },
+    {
+      "epoch": 132.00676328502416,
+      "grad_norm": 0.00023053883342072368,
+      "learning_rate": 1.116478797638218e-06,
+      "loss": 0.0,
+      "step": 9310
+    },
+    {
+      "epoch": 132.00676328502416,
+      "eval_accuracy": 0.8441558441558441,
+      "eval_loss": 1.5529676675796509,
+      "eval_runtime": 22.5575,
+      "eval_samples_per_second": 3.413,
+      "eval_steps_per_second": 0.887,
+      "step": 9310
+    },
+    {
+      "epoch": 133.0009661835749,
+      "grad_norm": 0.00038470380241051316,
+      "learning_rate": 1.1057434245840045e-06,
+      "loss": 0.0,
+      "step": 9320
+    },
+    {
+      "epoch": 133.00193236714975,
+      "grad_norm": 0.00019789740326814353,
+      "learning_rate": 1.0950080515297907e-06,
+      "loss": 0.0,
+      "step": 9330
+    },
+    {
+      "epoch": 133.00289855072464,
+      "grad_norm": 0.00023541324480902404,
+      "learning_rate": 1.0842726784755772e-06,
+      "loss": 0.0,
+      "step": 9340
+    },
+    {
+      "epoch": 133.00386473429953,
+      "grad_norm": 0.0025287040043622255,
+      "learning_rate": 1.0735373054213636e-06,
+      "loss": 0.0,
+      "step": 9350
+    },
+    {
+      "epoch": 133.0048309178744,
+      "grad_norm": 0.0004253005899954587,
+      "learning_rate": 1.0628019323671499e-06,
+      "loss": 0.0,
+      "step": 9360
+    },
+    {
+      "epoch": 133.00579710144928,
+      "grad_norm": 0.0001564955891808495,
+      "learning_rate": 1.052066559312936e-06,
+      "loss": 0.0,
+      "step": 9370
+    },
+    {
+      "epoch": 133.00676328502416,
+      "grad_norm": 0.0002814004838000983,
+      "learning_rate": 1.0413311862587225e-06,
+      "loss": 0.0,
+      "step": 9380
+    },
+    {
+      "epoch": 133.00676328502416,
+      "eval_accuracy": 0.8441558441558441,
+      "eval_loss": 1.5586316585540771,
+      "eval_runtime": 21.6913,
+      "eval_samples_per_second": 3.55,
+      "eval_steps_per_second": 0.922,
+      "step": 9380
+    },
+    {
+      "epoch": 134.0009661835749,
+      "grad_norm": 0.00016957947809714824,
+      "learning_rate": 1.030595813204509e-06,
+      "loss": 0.0,
+      "step": 9390
+    },
+    {
+      "epoch": 134.00193236714975,
+      "grad_norm": 0.0001453941804356873,
+      "learning_rate": 1.0198604401502952e-06,
+      "loss": 0.0,
+      "step": 9400
+    },
+    {
+      "epoch": 134.00289855072464,
+      "grad_norm": 0.0002459472743794322,
+      "learning_rate": 1.0091250670960817e-06,
+      "loss": 0.0,
+      "step": 9410
+    },
+    {
+      "epoch": 134.00386473429953,
+      "grad_norm": 0.0001978687650989741,
+      "learning_rate": 9.983896940418681e-07,
+      "loss": 0.0,
+      "step": 9420
+    },
+    {
+      "epoch": 134.0048309178744,
+      "grad_norm": 0.0002101514401147142,
+      "learning_rate": 9.876543209876544e-07,
+      "loss": 0.0,
+      "step": 9430
+    },
+    {
+      "epoch": 134.00579710144928,
+      "grad_norm": 0.00020504761778283864,
+      "learning_rate": 9.769189479334408e-07,
+      "loss": 0.0,
+      "step": 9440
+    },
+    {
+      "epoch": 134.00676328502416,
+      "grad_norm": 0.00021766134887002409,
+      "learning_rate": 9.66183574879227e-07,
+      "loss": 0.0,
+      "step": 9450
+    },
+    {
+      "epoch": 134.00676328502416,
+      "eval_accuracy": 0.8441558441558441,
+      "eval_loss": 1.564234733581543,
+      "eval_runtime": 20.8788,
+      "eval_samples_per_second": 3.688,
+      "eval_steps_per_second": 0.958,
+      "step": 9450
+    },
+    {
+      "epoch": 135.0009661835749,
+      "grad_norm": 0.00020511502225417644,
+      "learning_rate": 9.554482018250135e-07,
+      "loss": 0.0,
+      "step": 9460
+    },
+    {
+      "epoch": 135.00193236714975,
+      "grad_norm": 0.001212193281389773,
+      "learning_rate": 9.447128287707999e-07,
+      "loss": 0.0,
+      "step": 9470
+    },
+    {
+      "epoch": 135.00289855072464,
+      "grad_norm": 0.0002920062397606671,
+      "learning_rate": 9.339774557165862e-07,
+      "loss": 0.0,
+      "step": 9480
+    },
+    {
+      "epoch": 135.00386473429953,
+      "grad_norm": 0.0001773640251485631,
+      "learning_rate": 9.232420826623725e-07,
+      "loss": 0.0,
+      "step": 9490
+    },
+    {
+      "epoch": 135.0048309178744,
+      "grad_norm": 0.0002293359866598621,
+      "learning_rate": 9.12506709608159e-07,
+      "loss": 0.0,
+      "step": 9500
+    },
+    {
+      "epoch": 135.00579710144928,
+      "grad_norm": 0.00026728856028057635,
+      "learning_rate": 9.017713365539453e-07,
+      "loss": 0.0,
+      "step": 9510
+    },
+    {
+      "epoch": 135.00676328502416,
+      "grad_norm": 0.00027180055622011423,
+      "learning_rate": 8.910359634997316e-07,
+      "loss": 0.0,
+      "step": 9520
+    },
+    {
+      "epoch": 135.00676328502416,
+      "eval_accuracy": 0.8441558441558441,
+      "eval_loss": 1.5596168041229248,
+      "eval_runtime": 21.4836,
+      "eval_samples_per_second": 3.584,
+      "eval_steps_per_second": 0.931,
+      "step": 9520
+    },
+    {
+      "epoch": 136.0009661835749,
+      "grad_norm": 0.0005629212246276438,
+      "learning_rate": 8.80300590445518e-07,
+      "loss": 0.0,
+      "step": 9530
+    },
+    {
+      "epoch": 136.00193236714975,
+      "grad_norm": 0.000151635889778845,
+      "learning_rate": 8.695652173913044e-07,
+      "loss": 0.0,
+      "step": 9540
+    },
+    {
+      "epoch": 136.00289855072464,
+      "grad_norm": 0.0002448921441100538,
+      "learning_rate": 8.588298443370908e-07,
+      "loss": 0.0,
+      "step": 9550
+    },
+    {
+      "epoch": 136.00386473429953,
+      "grad_norm": 0.008513822220265865,
+      "learning_rate": 8.480944712828772e-07,
+      "loss": 0.0,
+      "step": 9560
+    },
+    {
+      "epoch": 136.0048309178744,
+      "grad_norm": 0.00017195107648149133,
+      "learning_rate": 8.373590982286635e-07,
+      "loss": 0.0,
+      "step": 9570
+    },
+    {
+      "epoch": 136.00579710144928,
+      "grad_norm": 0.00014387645933311433,
+      "learning_rate": 8.2662372517445e-07,
+      "loss": 0.0,
+      "step": 9580
+    },
+    {
+      "epoch": 136.00676328502416,
+      "grad_norm": 0.0002819636429194361,
+      "learning_rate": 8.158883521202362e-07,
+      "loss": 0.0,
+      "step": 9590
+    },
+    {
+      "epoch": 136.00676328502416,
+      "eval_accuracy": 0.8441558441558441,
+      "eval_loss": 1.5680683851242065,
+      "eval_runtime": 19.9494,
+      "eval_samples_per_second": 3.86,
+      "eval_steps_per_second": 1.003,
+      "step": 9590
+    },
+    {
+      "epoch": 137.0009661835749,
+      "grad_norm": 0.00016729481285437942,
+      "learning_rate": 8.051529790660226e-07,
+      "loss": 0.0,
+      "step": 9600
+    },
+    {
+      "epoch": 137.00193236714975,
+      "grad_norm": 0.0003245656844228506,
+      "learning_rate": 7.94417606011809e-07,
+      "loss": 0.0,
+      "step": 9610
+    },
+    {
+      "epoch": 137.00289855072464,
+      "grad_norm": 0.00014796361210756004,
+      "learning_rate": 7.836822329575953e-07,
+      "loss": 0.0,
+      "step": 9620
+    },
+    {
+      "epoch": 137.00386473429953,
+      "grad_norm": 633.5186767578125,
+      "learning_rate": 7.729468599033817e-07,
+      "loss": 0.0272,
+      "step": 9630
+    },
+    {
+      "epoch": 137.0048309178744,
+      "grad_norm": 0.0002833566104527563,
+      "learning_rate": 7.622114868491681e-07,
+      "loss": 0.0,
+      "step": 9640
+    },
+    {
+      "epoch": 137.00579710144928,
+      "grad_norm": 0.00018804871069733053,
+      "learning_rate": 7.514761137949545e-07,
+      "loss": 0.0,
+      "step": 9650
+    },
+    {
+      "epoch": 137.00676328502416,
+      "grad_norm": 0.0010792305693030357,
+      "learning_rate": 7.407407407407407e-07,
+      "loss": 0.0,
+      "step": 9660
+    },
+    {
+      "epoch": 137.00676328502416,
+      "eval_accuracy": 0.8181818181818182,
+      "eval_loss": 1.449820637702942,
+      "eval_runtime": 20.7119,
+      "eval_samples_per_second": 3.718,
+      "eval_steps_per_second": 0.966,
+      "step": 9660
+    },
+    {
+      "epoch": 138.0009661835749,
+      "grad_norm": 0.00046943960478529334,
+      "learning_rate": 7.300053676865272e-07,
+      "loss": 0.0,
+      "step": 9670
+    },
+    {
+      "epoch": 138.00193236714975,
+      "grad_norm": 0.002076608594506979,
+      "learning_rate": 7.192699946323135e-07,
+      "loss": 0.0,
+      "step": 9680
+    },
+    {
+      "epoch": 138.00289855072464,
+      "grad_norm": 0.0034116620663553476,
+      "learning_rate": 7.085346215780999e-07,
+      "loss": 0.0001,
+      "step": 9690
+    },
+    {
+      "epoch": 138.00386473429953,
+      "grad_norm": 0.00017700823082122952,
+      "learning_rate": 6.977992485238863e-07,
+      "loss": 0.0,
+      "step": 9700
+    },
+    {
+      "epoch": 138.0048309178744,
+      "grad_norm": 0.0004598258819896728,
+      "learning_rate": 6.870638754696727e-07,
+      "loss": 0.0,
+      "step": 9710
+    },
+    {
+      "epoch": 138.00579710144928,
+      "grad_norm": 0.0019175253110006452,
+      "learning_rate": 6.763285024154589e-07,
+      "loss": 0.0,
+      "step": 9720
+    },
+    {
+      "epoch": 138.00676328502416,
+      "grad_norm": 0.00016673437494318932,
+      "learning_rate": 6.655931293612454e-07,
+      "loss": 0.0,
+      "step": 9730
+    },
+    {
+      "epoch": 138.00676328502416,
+      "eval_accuracy": 0.8311688311688312,
+      "eval_loss": 1.6159300804138184,
+      "eval_runtime": 19.9379,
+      "eval_samples_per_second": 3.862,
+      "eval_steps_per_second": 1.003,
+      "step": 9730
+    },
+    {
+      "epoch": 139.0009661835749,
+      "grad_norm": 0.00019458989845588803,
+      "learning_rate": 6.548577563070317e-07,
+      "loss": 0.0,
+      "step": 9740
+    },
+    {
+      "epoch": 139.00193236714975,
+      "grad_norm": 0.0003470699302852154,
+      "learning_rate": 6.44122383252818e-07,
+      "loss": 0.0,
+      "step": 9750
+    },
+    {
+      "epoch": 139.00289855072464,
+      "grad_norm": 0.00048812173190526664,
+      "learning_rate": 6.333870101986045e-07,
+      "loss": 0.0,
+      "step": 9760
+    },
+    {
+      "epoch": 139.00386473429953,
+      "grad_norm": 0.0002671309339348227,
+      "learning_rate": 6.226516371443908e-07,
+      "loss": 0.0,
+      "step": 9770
+    },
+    {
+      "epoch": 139.0048309178744,
+      "grad_norm": 0.0003757915983442217,
+      "learning_rate": 6.119162640901772e-07,
+      "loss": 0.0,
+      "step": 9780
+    },
+    {
+      "epoch": 139.00579710144928,
+      "grad_norm": 0.0009131430997513235,
+      "learning_rate": 6.011808910359636e-07,
+      "loss": 0.0,
+      "step": 9790
+    },
+    {
+      "epoch": 139.00676328502416,
+      "grad_norm": 0.0003928108490072191,
+      "learning_rate": 5.904455179817499e-07,
+      "loss": 0.0,
+      "step": 9800
+    },
+    {
+      "epoch": 139.00676328502416,
+      "eval_accuracy": 0.8181818181818182,
+      "eval_loss": 1.6949611902236938,
+      "eval_runtime": 20.4828,
+      "eval_samples_per_second": 3.759,
+      "eval_steps_per_second": 0.976,
+      "step": 9800
+    },
+    {
+      "epoch": 140.0009661835749,
+      "grad_norm": 0.00022178050130605698,
+      "learning_rate": 5.797101449275363e-07,
+      "loss": 0.0,
+      "step": 9810
+    },
+    {
+      "epoch": 140.00193236714975,
+      "grad_norm": 0.00025778423878364265,
+      "learning_rate": 5.689747718733227e-07,
+      "loss": 0.0,
+      "step": 9820
+    },
+    {
+      "epoch": 140.00289855072464,
+      "grad_norm": 0.0012736802455037832,
+      "learning_rate": 5.58239398819109e-07,
+      "loss": 0.0,
+      "step": 9830
+    },
+    {
+      "epoch": 140.00386473429953,
+      "grad_norm": 0.00027229098486714065,
+      "learning_rate": 5.475040257648954e-07,
+      "loss": 0.0,
+      "step": 9840
+    },
+    {
+      "epoch": 140.0048309178744,
+      "grad_norm": 0.00017557268438395113,
+      "learning_rate": 5.367686527106818e-07,
+      "loss": 0.0,
+      "step": 9850
+    },
+    {
+      "epoch": 140.00579710144928,
+      "grad_norm": 0.0002299802436027676,
+      "learning_rate": 5.26033279656468e-07,
+      "loss": 0.0,
+      "step": 9860
+    },
+    {
+      "epoch": 140.00676328502416,
+      "grad_norm": 0.00015179581532720476,
+      "learning_rate": 5.152979066022545e-07,
+      "loss": 0.0,
+      "step": 9870
+    },
+    {
+      "epoch": 140.00676328502416,
+      "eval_accuracy": 0.8181818181818182,
+      "eval_loss": 1.6978023052215576,
+      "eval_runtime": 17.5663,
+      "eval_samples_per_second": 4.383,
+      "eval_steps_per_second": 1.139,
+      "step": 9870
+    },
+    {
+      "epoch": 141.0009661835749,
+      "grad_norm": 0.0007337291608564556,
+      "learning_rate": 5.045625335480408e-07,
+      "loss": 0.0,
+      "step": 9880
+    },
+    {
+      "epoch": 141.00193236714975,
+      "grad_norm": 0.00023553297796752304,
+      "learning_rate": 4.938271604938272e-07,
+      "loss": 0.0,
+      "step": 9890
+    },
+    {
+      "epoch": 141.00289855072464,
+      "grad_norm": 0.00022139857173897326,
+      "learning_rate": 4.830917874396135e-07,
+      "loss": 0.0,
+      "step": 9900
+    },
+    {
+      "epoch": 141.00386473429953,
+      "grad_norm": 0.0001859077747212723,
+      "learning_rate": 4.7235641438539993e-07,
+      "loss": 0.0,
+      "step": 9910
+    },
+    {
+      "epoch": 141.0048309178744,
+      "grad_norm": 0.00024027664039749652,
+      "learning_rate": 4.616210413311863e-07,
+      "loss": 0.0,
+      "step": 9920
+    },
+    {
+      "epoch": 141.00579710144928,
+      "grad_norm": 0.00047888257540762424,
+      "learning_rate": 4.5088566827697267e-07,
+      "loss": 0.0,
+      "step": 9930
+    },
+    {
+      "epoch": 141.00676328502416,
+      "grad_norm": 0.00018929631914943457,
+      "learning_rate": 4.40150295222759e-07,
+      "loss": 0.0,
+      "step": 9940
+    },
+    {
+      "epoch": 141.00676328502416,
+      "eval_accuracy": 0.8181818181818182,
+      "eval_loss": 1.6985359191894531,
+      "eval_runtime": 14.7785,
+      "eval_samples_per_second": 5.21,
+      "eval_steps_per_second": 1.353,
+      "step": 9940
+    },
+    {
+      "epoch": 142.0009661835749,
+      "grad_norm": 0.0006224442622624338,
+      "learning_rate": 4.294149221685454e-07,
+      "loss": 0.0,
+      "step": 9950
+    },
+    {
+      "epoch": 142.00193236714975,
+      "grad_norm": 0.00015741850074846298,
+      "learning_rate": 4.1867954911433176e-07,
+      "loss": 0.0,
+      "step": 9960
+    },
+    {
+      "epoch": 142.00289855072464,
+      "grad_norm": 0.0006889476208016276,
+      "learning_rate": 4.079441760601181e-07,
+      "loss": 0.0,
+      "step": 9970
+    },
+    {
+      "epoch": 142.00386473429953,
+      "grad_norm": 0.0005071698105894029,
+      "learning_rate": 3.972088030059045e-07,
+      "loss": 0.0,
+      "step": 9980
+    },
+    {
+      "epoch": 142.0048309178744,
+      "grad_norm": 0.00013677505194209516,
+      "learning_rate": 3.8647342995169085e-07,
+      "loss": 0.0,
+      "step": 9990
+    },
+    {
+      "epoch": 142.00579710144928,
+      "grad_norm": 0.0006237445049919188,
+      "learning_rate": 3.7573805689747724e-07,
+      "loss": 0.0,
+      "step": 10000
+    },
+    {
+      "epoch": 142.00676328502416,
+      "grad_norm": 0.00034121310454793274,
+      "learning_rate": 3.650026838432636e-07,
+      "loss": 0.0,
+      "step": 10010
+    },
+    {
+      "epoch": 142.00676328502416,
+      "eval_accuracy": 0.8181818181818182,
+      "eval_loss": 1.6995044946670532,
+      "eval_runtime": 15.0633,
+      "eval_samples_per_second": 5.112,
+      "eval_steps_per_second": 1.328,
+      "step": 10010
+    },
+    {
+      "epoch": 143.0009661835749,
+      "grad_norm": 0.00015302658721338958,
+      "learning_rate": 3.5426731078904993e-07,
+      "loss": 0.0,
+      "step": 10020
+    },
+    {
+      "epoch": 143.00193236714975,
+      "grad_norm": 0.00034528967808000743,
+      "learning_rate": 3.4353193773483633e-07,
+      "loss": 0.0,
+      "step": 10030
+    },
+    {
+      "epoch": 143.00289855072464,
+      "grad_norm": 0.00026275747222825885,
+      "learning_rate": 3.327965646806227e-07,
+      "loss": 0.0,
+      "step": 10040
+    },
+    {
+      "epoch": 143.00386473429953,
+      "grad_norm": 0.0006553785642609,
+      "learning_rate": 3.22061191626409e-07,
+      "loss": 0.0,
+      "step": 10050
+    },
+    {
+      "epoch": 143.0048309178744,
+      "grad_norm": 0.0002892035699915141,
+      "learning_rate": 3.113258185721954e-07,
+      "loss": 0.0,
+      "step": 10060
+    },
+    {
+      "epoch": 143.00579710144928,
+      "grad_norm": 0.0003178605402354151,
+      "learning_rate": 3.005904455179818e-07,
+      "loss": 0.0,
+      "step": 10070
+    },
+    {
+      "epoch": 143.00676328502416,
+      "grad_norm": 0.00017081064288504422,
+      "learning_rate": 2.8985507246376816e-07,
+      "loss": 0.0,
+      "step": 10080
+    },
+    {
+      "epoch": 143.00676328502416,
+      "eval_accuracy": 0.8051948051948052,
+      "eval_loss": 1.7036974430084229,
+      "eval_runtime": 14.5168,
+      "eval_samples_per_second": 5.304,
+      "eval_steps_per_second": 1.378,
+      "step": 10080
+    },
+    {
+      "epoch": 144.0009661835749,
+      "grad_norm": 0.00016073655569925904,
+      "learning_rate": 2.791196994095545e-07,
+      "loss": 0.0,
+      "step": 10090
+    },
+    {
+      "epoch": 144.00193236714975,
+      "grad_norm": 0.0004204573924653232,
+      "learning_rate": 2.683843263553409e-07,
+      "loss": 0.0,
+      "step": 10100
+    },
+    {
+      "epoch": 144.00289855072464,
+      "grad_norm": 0.0003274252521805465,
+      "learning_rate": 2.5764895330112725e-07,
+      "loss": 0.0,
+      "step": 10110
+    },
+    {
+      "epoch": 144.00386473429953,
+      "grad_norm": 0.0001679103879723698,
+      "learning_rate": 2.469135802469136e-07,
+      "loss": 0.0,
+      "step": 10120
+    },
+    {
+      "epoch": 144.0048309178744,
+      "grad_norm": 0.00028845362248830497,
+      "learning_rate": 2.3617820719269996e-07,
+      "loss": 0.0,
+      "step": 10130
+    },
+    {
+      "epoch": 144.00579710144928,
+      "grad_norm": 0.00044526177225634456,
+      "learning_rate": 2.2544283413848634e-07,
+      "loss": 0.0,
+      "step": 10140
+    },
+    {
+      "epoch": 144.00676328502416,
+      "grad_norm": 0.0007641853298991919,
+      "learning_rate": 2.147074610842727e-07,
+      "loss": 0.0,
+      "step": 10150
+    },
+    {
+      "epoch": 144.00676328502416,
+      "eval_accuracy": 0.8051948051948052,
+      "eval_loss": 1.7055643796920776,
+      "eval_runtime": 16.0115,
+      "eval_samples_per_second": 4.809,
+      "eval_steps_per_second": 1.249,
+      "step": 10150
+    },
+    {
+      "epoch": 145.0009661835749,
+      "grad_norm": 0.0006160350749269128,
+      "learning_rate": 2.0397208803005905e-07,
+      "loss": 0.0,
+      "step": 10160
+    },
+    {
+      "epoch": 145.00193236714975,
+      "grad_norm": 0.0005327651742845774,
+      "learning_rate": 1.9323671497584542e-07,
+      "loss": 0.0,
+      "step": 10170
+    },
+    {
+      "epoch": 145.00289855072464,
+      "grad_norm": 0.00017193098028656095,
+      "learning_rate": 1.825013419216318e-07,
+      "loss": 0.0,
+      "step": 10180
+    },
+    {
+      "epoch": 145.00386473429953,
+      "grad_norm": 0.00017320859478786588,
+      "learning_rate": 1.7176596886741817e-07,
+      "loss": 0.0,
+      "step": 10190
+    },
+    {
+      "epoch": 145.0048309178744,
+      "grad_norm": 0.0001653715589782223,
+      "learning_rate": 1.610305958132045e-07,
+      "loss": 0.0,
+      "step": 10200
+    },
+    {
+      "epoch": 145.00579710144928,
+      "grad_norm": 0.0001977159408852458,
+      "learning_rate": 1.502952227589909e-07,
+      "loss": 0.0,
+      "step": 10210
+    },
+    {
+      "epoch": 145.00676328502416,
+      "grad_norm": 0.001278374344110489,
+      "learning_rate": 1.3955984970477725e-07,
+      "loss": 0.0,
+      "step": 10220
+    },
+    {
+      "epoch": 145.00676328502416,
+      "eval_accuracy": 0.8051948051948052,
+      "eval_loss": 1.7054346799850464,
+      "eval_runtime": 16.0538,
+      "eval_samples_per_second": 4.796,
+      "eval_steps_per_second": 1.246,
+      "step": 10220
+    },
+    {
+      "epoch": 146.0009661835749,
+      "grad_norm": 0.0002044195425696671,
+      "learning_rate": 1.2882447665056362e-07,
+      "loss": 0.0,
+      "step": 10230
+    },
+    {
+      "epoch": 146.00193236714975,
+      "grad_norm": 0.00018759335216600448,
+      "learning_rate": 1.1808910359634998e-07,
+      "loss": 0.0,
+      "step": 10240
+    },
+    {
+      "epoch": 146.00289855072464,
+      "grad_norm": 0.0003087896038778126,
+      "learning_rate": 1.0735373054213635e-07,
+      "loss": 0.0,
+      "step": 10250
+    },
+    {
+      "epoch": 146.00386473429953,
+      "grad_norm": 0.0002801946538966149,
+      "learning_rate": 9.661835748792271e-08,
+      "loss": 0.0,
+      "step": 10260
+    },
+    {
+      "epoch": 146.0048309178744,
+      "grad_norm": 0.00012826151214540005,
+      "learning_rate": 8.588298443370908e-08,
+      "loss": 0.0,
+      "step": 10270
+    },
+    {
+      "epoch": 146.00579710144928,
+      "grad_norm": 0.0002482128038536757,
+      "learning_rate": 7.514761137949545e-08,
+      "loss": 0.0,
+      "step": 10280
+    },
+    {
+      "epoch": 146.00676328502416,
+      "grad_norm": 0.0003177243925165385,
+      "learning_rate": 6.441223832528181e-08,
+      "loss": 0.0,
+      "step": 10290
+    },
+    {
+      "epoch": 146.00676328502416,
+      "eval_accuracy": 0.8051948051948052,
+      "eval_loss": 1.7054239511489868,
+      "eval_runtime": 18.0904,
+      "eval_samples_per_second": 4.256,
+      "eval_steps_per_second": 1.106,
+      "step": 10290
+    },
+    {
+      "epoch": 147.0009661835749,
+      "grad_norm": 0.0007799161248840392,
+      "learning_rate": 5.367686527106818e-08,
+      "loss": 0.0,
+      "step": 10300
+    },
+    {
+      "epoch": 147.00193236714975,
+      "grad_norm": 0.00712132453918457,
+      "learning_rate": 4.294149221685454e-08,
+      "loss": 0.0,
+      "step": 10310
+    },
+    {
+      "epoch": 147.00289855072464,
+      "grad_norm": 0.00016804441111162305,
+      "learning_rate": 3.2206119162640906e-08,
+      "loss": 0.0,
+      "step": 10320
+    },
+    {
+      "epoch": 147.00386473429953,
+      "grad_norm": 0.00015434020315296948,
+      "learning_rate": 2.147074610842727e-08,
+      "loss": 0.0,
+      "step": 10330
+    },
+    {
+      "epoch": 147.0048309178744,
+      "grad_norm": 0.4991607367992401,
+      "learning_rate": 1.0735373054213635e-08,
+      "loss": 0.0,
+      "step": 10340
+    },
+    {
+      "epoch": 147.00579710144928,
+      "grad_norm": 0.00018625461962074041,
+      "learning_rate": 0.0,
+      "loss": 0.0,
+      "step": 10350
+    },
+    {
+      "epoch": 147.00579710144928,
+      "eval_accuracy": 0.8051948051948052,
+      "eval_loss": 1.7040551900863647,
+      "eval_runtime": 18.6543,
+      "eval_samples_per_second": 4.128,
+      "eval_steps_per_second": 1.072,
+      "step": 10350
+    },
+    {
+      "epoch": 147.00579710144928,
+      "step": 10350,
+      "total_flos": 1.8049752771658895e+20,
+      "train_loss": 0.08939714265258908,
+      "train_runtime": 20289.1727,
+      "train_samples_per_second": 2.04,
+      "train_steps_per_second": 0.51
+    },
+    {
+      "epoch": 147.00579710144928,
+      "eval_accuracy": 0.8701298701298701,
+      "eval_loss": 1.1517490148544312,
+      "eval_runtime": 20.1404,
+      "eval_samples_per_second": 3.823,
+      "eval_steps_per_second": 0.993,
+      "step": 10350
+    },
+    {
+      "epoch": 147.00579710144928,
+      "eval_accuracy": 0.8701298701298701,
+      "eval_loss": 1.1517488956451416,
+      "eval_runtime": 18.4764,
+      "eval_samples_per_second": 4.167,
+      "eval_steps_per_second": 1.082,
+      "step": 10350
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 10350,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 9223372036854775807,
+  "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": 1.8049752771658895e+20,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}