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": 0.8461538461538461,
+  "best_model_checkpoint": "MAE-CT-M1N0-M12_v8_split3_v3/checkpoint-3710",
+  "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": 1.7746223211288452,
+      "learning_rate": 9.661835748792271e-08,
+      "loss": 0.7034,
+      "step": 10
+    },
+    {
+      "epoch": 0.001932367149758454,
+      "grad_norm": 3.510434865951538,
+      "learning_rate": 1.9323671497584542e-07,
+      "loss": 0.7006,
+      "step": 20
+    },
+    {
+      "epoch": 0.002898550724637681,
+      "grad_norm": 5.10714054107666,
+      "learning_rate": 2.8985507246376816e-07,
+      "loss": 0.7,
+      "step": 30
+    },
+    {
+      "epoch": 0.003864734299516908,
+      "grad_norm": 1.8410993814468384,
+      "learning_rate": 3.8647342995169085e-07,
+      "loss": 0.6961,
+      "step": 40
+    },
+    {
+      "epoch": 0.004830917874396135,
+      "grad_norm": 1.225154161453247,
+      "learning_rate": 4.830917874396135e-07,
+      "loss": 0.6918,
+      "step": 50
+    },
+    {
+      "epoch": 0.005797101449275362,
+      "grad_norm": 2.331960678100586,
+      "learning_rate": 5.797101449275363e-07,
+      "loss": 0.683,
+      "step": 60
+    },
+    {
+      "epoch": 0.00676328502415459,
+      "grad_norm": 5.630166053771973,
+      "learning_rate": 6.763285024154589e-07,
+      "loss": 0.6676,
+      "step": 70
+    },
+    {
+      "epoch": 0.00676328502415459,
+      "eval_accuracy": 0.6538461538461539,
+      "eval_loss": 0.6765444874763489,
+      "eval_runtime": 18.2677,
+      "eval_samples_per_second": 4.27,
+      "eval_steps_per_second": 1.095,
+      "step": 70
+    },
+    {
+      "epoch": 1.0009661835748793,
+      "grad_norm": 1.8150583505630493,
+      "learning_rate": 7.729468599033817e-07,
+      "loss": 0.6831,
+      "step": 80
+    },
+    {
+      "epoch": 1.0019323671497584,
+      "grad_norm": 3.461575984954834,
+      "learning_rate": 8.695652173913044e-07,
+      "loss": 0.7022,
+      "step": 90
+    },
+    {
+      "epoch": 1.0028985507246377,
+      "grad_norm": 3.506470203399658,
+      "learning_rate": 9.66183574879227e-07,
+      "loss": 0.6608,
+      "step": 100
+    },
+    {
+      "epoch": 1.0038647342995168,
+      "grad_norm": 1.7554961442947388,
+      "learning_rate": 1.0628019323671499e-06,
+      "loss": 0.6582,
+      "step": 110
+    },
+    {
+      "epoch": 1.0048309178743962,
+      "grad_norm": 2.083617687225342,
+      "learning_rate": 1.1594202898550726e-06,
+      "loss": 0.6522,
+      "step": 120
+    },
+    {
+      "epoch": 1.0057971014492753,
+      "grad_norm": 1.8603112697601318,
+      "learning_rate": 1.2560386473429952e-06,
+      "loss": 0.6149,
+      "step": 130
+    },
+    {
+      "epoch": 1.0067632850241546,
+      "grad_norm": 8.43808364868164,
+      "learning_rate": 1.3526570048309178e-06,
+      "loss": 0.6735,
+      "step": 140
+    },
+    {
+      "epoch": 1.0067632850241546,
+      "eval_accuracy": 0.6666666666666666,
+      "eval_loss": 0.6379338502883911,
+      "eval_runtime": 17.518,
+      "eval_samples_per_second": 4.453,
+      "eval_steps_per_second": 1.142,
+      "step": 140
+    },
+    {
+      "epoch": 2.0009661835748793,
+      "grad_norm": 3.509326457977295,
+      "learning_rate": 1.4492753623188408e-06,
+      "loss": 0.6462,
+      "step": 150
+    },
+    {
+      "epoch": 2.0019323671497586,
+      "grad_norm": 2.7582528591156006,
+      "learning_rate": 1.5458937198067634e-06,
+      "loss": 0.6394,
+      "step": 160
+    },
+    {
+      "epoch": 2.0028985507246375,
+      "grad_norm": 4.160943508148193,
+      "learning_rate": 1.6425120772946862e-06,
+      "loss": 0.5903,
+      "step": 170
+    },
+    {
+      "epoch": 2.003864734299517,
+      "grad_norm": 4.066112995147705,
+      "learning_rate": 1.7391304347826088e-06,
+      "loss": 0.6221,
+      "step": 180
+    },
+    {
+      "epoch": 2.004830917874396,
+      "grad_norm": 7.548180103302002,
+      "learning_rate": 1.8357487922705318e-06,
+      "loss": 0.7383,
+      "step": 190
+    },
+    {
+      "epoch": 2.0057971014492755,
+      "grad_norm": 2.1525609493255615,
+      "learning_rate": 1.932367149758454e-06,
+      "loss": 0.6885,
+      "step": 200
+    },
+    {
+      "epoch": 2.006763285024155,
+      "grad_norm": 7.099206447601318,
+      "learning_rate": 2.028985507246377e-06,
+      "loss": 0.6391,
+      "step": 210
+    },
+    {
+      "epoch": 2.006763285024155,
+      "eval_accuracy": 0.6666666666666666,
+      "eval_loss": 0.6261193752288818,
+      "eval_runtime": 17.3999,
+      "eval_samples_per_second": 4.483,
+      "eval_steps_per_second": 1.149,
+      "step": 210
+    },
+    {
+      "epoch": 3.0009661835748793,
+      "grad_norm": 2.9164934158325195,
+      "learning_rate": 2.1256038647342997e-06,
+      "loss": 0.5678,
+      "step": 220
+    },
+    {
+      "epoch": 3.0019323671497586,
+      "grad_norm": 4.371841907501221,
+      "learning_rate": 2.222222222222222e-06,
+      "loss": 0.6158,
+      "step": 230
+    },
+    {
+      "epoch": 3.0028985507246375,
+      "grad_norm": 5.164270401000977,
+      "learning_rate": 2.3188405797101453e-06,
+      "loss": 0.6795,
+      "step": 240
+    },
+    {
+      "epoch": 3.003864734299517,
+      "grad_norm": 5.403680324554443,
+      "learning_rate": 2.4154589371980677e-06,
+      "loss": 0.5925,
+      "step": 250
+    },
+    {
+      "epoch": 3.004830917874396,
+      "grad_norm": 4.002570629119873,
+      "learning_rate": 2.5120772946859904e-06,
+      "loss": 0.621,
+      "step": 260
+    },
+    {
+      "epoch": 3.0057971014492755,
+      "grad_norm": 16.2905216217041,
+      "learning_rate": 2.6086956521739132e-06,
+      "loss": 0.5738,
+      "step": 270
+    },
+    {
+      "epoch": 3.006763285024155,
+      "grad_norm": 31.137666702270508,
+      "learning_rate": 2.7053140096618356e-06,
+      "loss": 0.8085,
+      "step": 280
+    },
+    {
+      "epoch": 3.006763285024155,
+      "eval_accuracy": 0.6666666666666666,
+      "eval_loss": 0.6050156950950623,
+      "eval_runtime": 17.572,
+      "eval_samples_per_second": 4.439,
+      "eval_steps_per_second": 1.138,
+      "step": 280
+    },
+    {
+      "epoch": 4.000966183574879,
+      "grad_norm": 4.449572563171387,
+      "learning_rate": 2.801932367149759e-06,
+      "loss": 0.6442,
+      "step": 290
+    },
+    {
+      "epoch": 4.001932367149759,
+      "grad_norm": 6.249354839324951,
+      "learning_rate": 2.8985507246376816e-06,
+      "loss": 0.574,
+      "step": 300
+    },
+    {
+      "epoch": 4.0028985507246375,
+      "grad_norm": 8.182052612304688,
+      "learning_rate": 2.995169082125604e-06,
+      "loss": 0.5859,
+      "step": 310
+    },
+    {
+      "epoch": 4.003864734299517,
+      "grad_norm": 9.79288387298584,
+      "learning_rate": 3.0917874396135268e-06,
+      "loss": 0.5761,
+      "step": 320
+    },
+    {
+      "epoch": 4.004830917874396,
+      "grad_norm": 9.698325157165527,
+      "learning_rate": 3.188405797101449e-06,
+      "loss": 0.5726,
+      "step": 330
+    },
+    {
+      "epoch": 4.005797101449275,
+      "grad_norm": 15.086708068847656,
+      "learning_rate": 3.2850241545893724e-06,
+      "loss": 0.6452,
+      "step": 340
+    },
+    {
+      "epoch": 4.006763285024155,
+      "grad_norm": 41.014373779296875,
+      "learning_rate": 3.381642512077295e-06,
+      "loss": 0.4774,
+      "step": 350
+    },
+    {
+      "epoch": 4.006763285024155,
+      "eval_accuracy": 0.6538461538461539,
+      "eval_loss": 0.5725159049034119,
+      "eval_runtime": 16.8333,
+      "eval_samples_per_second": 4.634,
+      "eval_steps_per_second": 1.188,
+      "step": 350
+    },
+    {
+      "epoch": 5.000966183574879,
+      "grad_norm": 9.569558143615723,
+      "learning_rate": 3.4782608695652175e-06,
+      "loss": 0.6024,
+      "step": 360
+    },
+    {
+      "epoch": 5.001932367149759,
+      "grad_norm": 12.862885475158691,
+      "learning_rate": 3.5748792270531403e-06,
+      "loss": 0.5685,
+      "step": 370
+    },
+    {
+      "epoch": 5.0028985507246375,
+      "grad_norm": 12.0186128616333,
+      "learning_rate": 3.6714975845410635e-06,
+      "loss": 0.519,
+      "step": 380
+    },
+    {
+      "epoch": 5.003864734299517,
+      "grad_norm": 15.317484855651855,
+      "learning_rate": 3.768115942028986e-06,
+      "loss": 0.5404,
+      "step": 390
+    },
+    {
+      "epoch": 5.004830917874396,
+      "grad_norm": 13.592208862304688,
+      "learning_rate": 3.864734299516908e-06,
+      "loss": 0.5157,
+      "step": 400
+    },
+    {
+      "epoch": 5.005797101449275,
+      "grad_norm": 7.7680253982543945,
+      "learning_rate": 3.961352657004831e-06,
+      "loss": 0.477,
+      "step": 410
+    },
+    {
+      "epoch": 5.006763285024155,
+      "grad_norm": 5.047906398773193,
+      "learning_rate": 4.057971014492754e-06,
+      "loss": 0.5668,
+      "step": 420
+    },
+    {
+      "epoch": 5.006763285024155,
+      "eval_accuracy": 0.717948717948718,
+      "eval_loss": 0.5513753294944763,
+      "eval_runtime": 15.872,
+      "eval_samples_per_second": 4.914,
+      "eval_steps_per_second": 1.26,
+      "step": 420
+    },
+    {
+      "epoch": 6.000966183574879,
+      "grad_norm": 7.734065055847168,
+      "learning_rate": 4.154589371980677e-06,
+      "loss": 0.398,
+      "step": 430
+    },
+    {
+      "epoch": 6.001932367149759,
+      "grad_norm": 31.831592559814453,
+      "learning_rate": 4.251207729468599e-06,
+      "loss": 0.5832,
+      "step": 440
+    },
+    {
+      "epoch": 6.0028985507246375,
+      "grad_norm": 33.391658782958984,
+      "learning_rate": 4.347826086956522e-06,
+      "loss": 0.4523,
+      "step": 450
+    },
+    {
+      "epoch": 6.003864734299517,
+      "grad_norm": 49.021305084228516,
+      "learning_rate": 4.444444444444444e-06,
+      "loss": 0.3297,
+      "step": 460
+    },
+    {
+      "epoch": 6.004830917874396,
+      "grad_norm": 16.681032180786133,
+      "learning_rate": 4.541062801932368e-06,
+      "loss": 0.5626,
+      "step": 470
+    },
+    {
+      "epoch": 6.005797101449275,
+      "grad_norm": 43.30965805053711,
+      "learning_rate": 4.637681159420291e-06,
+      "loss": 0.4457,
+      "step": 480
+    },
+    {
+      "epoch": 6.006763285024155,
+      "grad_norm": 130.68243408203125,
+      "learning_rate": 4.7342995169082125e-06,
+      "loss": 0.6607,
+      "step": 490
+    },
+    {
+      "epoch": 6.006763285024155,
+      "eval_accuracy": 0.7564102564102564,
+      "eval_loss": 0.4834001064300537,
+      "eval_runtime": 16.065,
+      "eval_samples_per_second": 4.855,
+      "eval_steps_per_second": 1.245,
+      "step": 490
+    },
+    {
+      "epoch": 7.000966183574879,
+      "grad_norm": 20.249547958374023,
+      "learning_rate": 4.830917874396135e-06,
+      "loss": 0.3441,
+      "step": 500
+    },
+    {
+      "epoch": 7.001932367149759,
+      "grad_norm": 4.507364273071289,
+      "learning_rate": 4.927536231884059e-06,
+      "loss": 0.4055,
+      "step": 510
+    },
+    {
+      "epoch": 7.0028985507246375,
+      "grad_norm": 32.87883377075195,
+      "learning_rate": 5.024154589371981e-06,
+      "loss": 0.347,
+      "step": 520
+    },
+    {
+      "epoch": 7.003864734299517,
+      "grad_norm": 51.900856018066406,
+      "learning_rate": 5.1207729468599045e-06,
+      "loss": 0.4185,
+      "step": 530
+    },
+    {
+      "epoch": 7.004830917874396,
+      "grad_norm": 59.27176284790039,
+      "learning_rate": 5.2173913043478265e-06,
+      "loss": 0.5276,
+      "step": 540
+    },
+    {
+      "epoch": 7.005797101449275,
+      "grad_norm": 49.032371520996094,
+      "learning_rate": 5.314009661835749e-06,
+      "loss": 0.4577,
+      "step": 550
+    },
+    {
+      "epoch": 7.006763285024155,
+      "grad_norm": 4.786675930023193,
+      "learning_rate": 5.410628019323671e-06,
+      "loss": 1.0769,
+      "step": 560
+    },
+    {
+      "epoch": 7.006763285024155,
+      "eval_accuracy": 0.6666666666666666,
+      "eval_loss": 0.7449601888656616,
+      "eval_runtime": 17.1147,
+      "eval_samples_per_second": 4.557,
+      "eval_steps_per_second": 1.169,
+      "step": 560
+    },
+    {
+      "epoch": 8.00096618357488,
+      "grad_norm": 8.022248268127441,
+      "learning_rate": 5.507246376811595e-06,
+      "loss": 0.6193,
+      "step": 570
+    },
+    {
+      "epoch": 8.001932367149758,
+      "grad_norm": 11.327919960021973,
+      "learning_rate": 5.603864734299518e-06,
+      "loss": 0.5499,
+      "step": 580
+    },
+    {
+      "epoch": 8.002898550724638,
+      "grad_norm": 21.584156036376953,
+      "learning_rate": 5.70048309178744e-06,
+      "loss": 0.4251,
+      "step": 590
+    },
+    {
+      "epoch": 8.003864734299517,
+      "grad_norm": 42.50540542602539,
+      "learning_rate": 5.797101449275363e-06,
+      "loss": 0.5429,
+      "step": 600
+    },
+    {
+      "epoch": 8.004830917874395,
+      "grad_norm": 77.20704650878906,
+      "learning_rate": 5.893719806763285e-06,
+      "loss": 0.3544,
+      "step": 610
+    },
+    {
+      "epoch": 8.005797101449275,
+      "grad_norm": 93.73331451416016,
+      "learning_rate": 5.990338164251208e-06,
+      "loss": 0.5156,
+      "step": 620
+    },
+    {
+      "epoch": 8.006763285024155,
+      "grad_norm": 2.17142915725708,
+      "learning_rate": 6.086956521739132e-06,
+      "loss": 0.5425,
+      "step": 630
+    },
+    {
+      "epoch": 8.006763285024155,
+      "eval_accuracy": 0.7948717948717948,
+      "eval_loss": 0.46234259009361267,
+      "eval_runtime": 17.202,
+      "eval_samples_per_second": 4.534,
+      "eval_steps_per_second": 1.163,
+      "step": 630
+    },
+    {
+      "epoch": 9.00096618357488,
+      "grad_norm": 57.35111999511719,
+      "learning_rate": 6.1835748792270535e-06,
+      "loss": 0.304,
+      "step": 640
+    },
+    {
+      "epoch": 9.001932367149758,
+      "grad_norm": 12.979510307312012,
+      "learning_rate": 6.280193236714976e-06,
+      "loss": 0.3022,
+      "step": 650
+    },
+    {
+      "epoch": 9.002898550724638,
+      "grad_norm": 26.668115615844727,
+      "learning_rate": 6.376811594202898e-06,
+      "loss": 0.7653,
+      "step": 660
+    },
+    {
+      "epoch": 9.003864734299517,
+      "grad_norm": 0.5207400918006897,
+      "learning_rate": 6.473429951690822e-06,
+      "loss": 0.3052,
+      "step": 670
+    },
+    {
+      "epoch": 9.004830917874395,
+      "grad_norm": 29.032527923583984,
+      "learning_rate": 6.570048309178745e-06,
+      "loss": 1.0119,
+      "step": 680
+    },
+    {
+      "epoch": 9.005797101449275,
+      "grad_norm": 15.286548614501953,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 0.6187,
+      "step": 690
+    },
+    {
+      "epoch": 9.006763285024155,
+      "grad_norm": 6.56998872756958,
+      "learning_rate": 6.76328502415459e-06,
+      "loss": 0.2959,
+      "step": 700
+    },
+    {
+      "epoch": 9.006763285024155,
+      "eval_accuracy": 0.7564102564102564,
+      "eval_loss": 0.511256217956543,
+      "eval_runtime": 17.0114,
+      "eval_samples_per_second": 4.585,
+      "eval_steps_per_second": 1.176,
+      "step": 700
+    },
+    {
+      "epoch": 10.00096618357488,
+      "grad_norm": 4.055560111999512,
+      "learning_rate": 6.859903381642513e-06,
+      "loss": 0.5726,
+      "step": 710
+    },
+    {
+      "epoch": 10.001932367149758,
+      "grad_norm": 100.66243743896484,
+      "learning_rate": 6.956521739130435e-06,
+      "loss": 0.2548,
+      "step": 720
+    },
+    {
+      "epoch": 10.002898550724638,
+      "grad_norm": 23.80630111694336,
+      "learning_rate": 7.053140096618359e-06,
+      "loss": 0.2119,
+      "step": 730
+    },
+    {
+      "epoch": 10.003864734299517,
+      "grad_norm": 103.37820434570312,
+      "learning_rate": 7.149758454106281e-06,
+      "loss": 0.4181,
+      "step": 740
+    },
+    {
+      "epoch": 10.004830917874395,
+      "grad_norm": 44.21693420410156,
+      "learning_rate": 7.246376811594203e-06,
+      "loss": 1.2421,
+      "step": 750
+    },
+    {
+      "epoch": 10.005797101449275,
+      "grad_norm": 13.183788299560547,
+      "learning_rate": 7.342995169082127e-06,
+      "loss": 0.4219,
+      "step": 760
+    },
+    {
+      "epoch": 10.006763285024155,
+      "grad_norm": 95.68376159667969,
+      "learning_rate": 7.439613526570049e-06,
+      "loss": 0.7674,
+      "step": 770
+    },
+    {
+      "epoch": 10.006763285024155,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 0.5330198407173157,
+      "eval_runtime": 16.5489,
+      "eval_samples_per_second": 4.713,
+      "eval_steps_per_second": 1.209,
+      "step": 770
+    },
+    {
+      "epoch": 11.00096618357488,
+      "grad_norm": 60.94060516357422,
+      "learning_rate": 7.536231884057972e-06,
+      "loss": 0.5811,
+      "step": 780
+    },
+    {
+      "epoch": 11.001932367149758,
+      "grad_norm": 33.39159393310547,
+      "learning_rate": 7.632850241545895e-06,
+      "loss": 0.6433,
+      "step": 790
+    },
+    {
+      "epoch": 11.002898550724638,
+      "grad_norm": 25.03746223449707,
+      "learning_rate": 7.729468599033817e-06,
+      "loss": 0.4988,
+      "step": 800
+    },
+    {
+      "epoch": 11.003864734299517,
+      "grad_norm": 17.853822708129883,
+      "learning_rate": 7.82608695652174e-06,
+      "loss": 0.4904,
+      "step": 810
+    },
+    {
+      "epoch": 11.004830917874395,
+      "grad_norm": 20.189794540405273,
+      "learning_rate": 7.922705314009662e-06,
+      "loss": 0.3451,
+      "step": 820
+    },
+    {
+      "epoch": 11.005797101449275,
+      "grad_norm": 35.152103424072266,
+      "learning_rate": 8.019323671497586e-06,
+      "loss": 0.6849,
+      "step": 830
+    },
+    {
+      "epoch": 11.006763285024155,
+      "grad_norm": 128.79225158691406,
+      "learning_rate": 8.115942028985508e-06,
+      "loss": 0.3015,
+      "step": 840
+    },
+    {
+      "epoch": 11.006763285024155,
+      "eval_accuracy": 0.7435897435897436,
+      "eval_loss": 0.6790037155151367,
+      "eval_runtime": 17.2359,
+      "eval_samples_per_second": 4.525,
+      "eval_steps_per_second": 1.16,
+      "step": 840
+    },
+    {
+      "epoch": 12.00096618357488,
+      "grad_norm": 35.15237808227539,
+      "learning_rate": 8.212560386473431e-06,
+      "loss": 0.4053,
+      "step": 850
+    },
+    {
+      "epoch": 12.001932367149758,
+      "grad_norm": 5.28515625,
+      "learning_rate": 8.309178743961353e-06,
+      "loss": 0.5247,
+      "step": 860
+    },
+    {
+      "epoch": 12.002898550724638,
+      "grad_norm": 10.3850736618042,
+      "learning_rate": 8.405797101449275e-06,
+      "loss": 0.216,
+      "step": 870
+    },
+    {
+      "epoch": 12.003864734299517,
+      "grad_norm": 19.821121215820312,
+      "learning_rate": 8.502415458937199e-06,
+      "loss": 0.2827,
+      "step": 880
+    },
+    {
+      "epoch": 12.004830917874395,
+      "grad_norm": 1.2210925817489624,
+      "learning_rate": 8.599033816425122e-06,
+      "loss": 0.2345,
+      "step": 890
+    },
+    {
+      "epoch": 12.005797101449275,
+      "grad_norm": 1.1479548215866089,
+      "learning_rate": 8.695652173913044e-06,
+      "loss": 0.5561,
+      "step": 900
+    },
+    {
+      "epoch": 12.006763285024155,
+      "grad_norm": 193.2356719970703,
+      "learning_rate": 8.792270531400966e-06,
+      "loss": 0.6252,
+      "step": 910
+    },
+    {
+      "epoch": 12.006763285024155,
+      "eval_accuracy": 0.6666666666666666,
+      "eval_loss": 1.6536943912506104,
+      "eval_runtime": 18.3082,
+      "eval_samples_per_second": 4.26,
+      "eval_steps_per_second": 1.092,
+      "step": 910
+    },
+    {
+      "epoch": 13.00096618357488,
+      "grad_norm": 13.043883323669434,
+      "learning_rate": 8.888888888888888e-06,
+      "loss": 1.3737,
+      "step": 920
+    },
+    {
+      "epoch": 13.001932367149758,
+      "grad_norm": 10.54146957397461,
+      "learning_rate": 8.985507246376812e-06,
+      "loss": 0.5883,
+      "step": 930
+    },
+    {
+      "epoch": 13.002898550724638,
+      "grad_norm": 53.698272705078125,
+      "learning_rate": 9.082125603864736e-06,
+      "loss": 0.2905,
+      "step": 940
+    },
+    {
+      "epoch": 13.003864734299517,
+      "grad_norm": 58.599098205566406,
+      "learning_rate": 9.178743961352658e-06,
+      "loss": 0.7171,
+      "step": 950
+    },
+    {
+      "epoch": 13.004830917874395,
+      "grad_norm": 46.291561126708984,
+      "learning_rate": 9.275362318840581e-06,
+      "loss": 0.2068,
+      "step": 960
+    },
+    {
+      "epoch": 13.005797101449275,
+      "grad_norm": 0.8758206367492676,
+      "learning_rate": 9.371980676328503e-06,
+      "loss": 0.3303,
+      "step": 970
+    },
+    {
+      "epoch": 13.006763285024155,
+      "grad_norm": 0.09083396941423416,
+      "learning_rate": 9.468599033816425e-06,
+      "loss": 0.4554,
+      "step": 980
+    },
+    {
+      "epoch": 13.006763285024155,
+      "eval_accuracy": 0.7564102564102564,
+      "eval_loss": 0.8944117426872253,
+      "eval_runtime": 17.9825,
+      "eval_samples_per_second": 4.338,
+      "eval_steps_per_second": 1.112,
+      "step": 980
+    },
+    {
+      "epoch": 14.00096618357488,
+      "grad_norm": 106.69209289550781,
+      "learning_rate": 9.565217391304349e-06,
+      "loss": 0.3241,
+      "step": 990
+    },
+    {
+      "epoch": 14.001932367149758,
+      "grad_norm": 86.01146697998047,
+      "learning_rate": 9.66183574879227e-06,
+      "loss": 0.4302,
+      "step": 1000
+    },
+    {
+      "epoch": 14.002898550724638,
+      "grad_norm": 93.19570922851562,
+      "learning_rate": 9.758454106280194e-06,
+      "loss": 0.692,
+      "step": 1010
+    },
+    {
+      "epoch": 14.003864734299517,
+      "grad_norm": 93.05182647705078,
+      "learning_rate": 9.855072463768118e-06,
+      "loss": 0.7247,
+      "step": 1020
+    },
+    {
+      "epoch": 14.004830917874395,
+      "grad_norm": 2.0111024379730225,
+      "learning_rate": 9.95169082125604e-06,
+      "loss": 0.3524,
+      "step": 1030
+    },
+    {
+      "epoch": 14.005797101449275,
+      "grad_norm": 48.2626838684082,
+      "learning_rate": 9.994632313472894e-06,
+      "loss": 0.3212,
+      "step": 1040
+    },
+    {
+      "epoch": 14.006763285024155,
+      "grad_norm": 0.2520197033882141,
+      "learning_rate": 9.98389694041868e-06,
+      "loss": 0.364,
+      "step": 1050
+    },
+    {
+      "epoch": 14.006763285024155,
+      "eval_accuracy": 0.717948717948718,
+      "eval_loss": 0.8103033900260925,
+      "eval_runtime": 18.9745,
+      "eval_samples_per_second": 4.111,
+      "eval_steps_per_second": 1.054,
+      "step": 1050
+    },
+    {
+      "epoch": 15.00096618357488,
+      "grad_norm": 0.2682894170284271,
+      "learning_rate": 9.973161567364467e-06,
+      "loss": 0.124,
+      "step": 1060
+    },
+    {
+      "epoch": 15.001932367149758,
+      "grad_norm": 31.360898971557617,
+      "learning_rate": 9.962426194310253e-06,
+      "loss": 0.3927,
+      "step": 1070
+    },
+    {
+      "epoch": 15.002898550724638,
+      "grad_norm": 36.74116134643555,
+      "learning_rate": 9.95169082125604e-06,
+      "loss": 0.4221,
+      "step": 1080
+    },
+    {
+      "epoch": 15.003864734299517,
+      "grad_norm": 196.0001220703125,
+      "learning_rate": 9.940955448201826e-06,
+      "loss": 0.1994,
+      "step": 1090
+    },
+    {
+      "epoch": 15.004830917874395,
+      "grad_norm": 18.42249298095703,
+      "learning_rate": 9.930220075147611e-06,
+      "loss": 0.4539,
+      "step": 1100
+    },
+    {
+      "epoch": 15.005797101449275,
+      "grad_norm": 19.64423179626465,
+      "learning_rate": 9.919484702093398e-06,
+      "loss": 0.2981,
+      "step": 1110
+    },
+    {
+      "epoch": 15.006763285024155,
+      "grad_norm": 13.708057403564453,
+      "learning_rate": 9.908749329039184e-06,
+      "loss": 0.444,
+      "step": 1120
+    },
+    {
+      "epoch": 15.006763285024155,
+      "eval_accuracy": 0.782051282051282,
+      "eval_loss": 0.710326611995697,
+      "eval_runtime": 21.7172,
+      "eval_samples_per_second": 3.592,
+      "eval_steps_per_second": 0.921,
+      "step": 1120
+    },
+    {
+      "epoch": 16.000966183574878,
+      "grad_norm": 25.7137451171875,
+      "learning_rate": 9.89801395598497e-06,
+      "loss": 0.3589,
+      "step": 1130
+    },
+    {
+      "epoch": 16.00193236714976,
+      "grad_norm": 180.5447540283203,
+      "learning_rate": 9.887278582930757e-06,
+      "loss": 0.3119,
+      "step": 1140
+    },
+    {
+      "epoch": 16.002898550724638,
+      "grad_norm": 1.0929948091506958,
+      "learning_rate": 9.876543209876543e-06,
+      "loss": 0.3492,
+      "step": 1150
+    },
+    {
+      "epoch": 16.003864734299516,
+      "grad_norm": 33.48530960083008,
+      "learning_rate": 9.865807836822331e-06,
+      "loss": 0.6954,
+      "step": 1160
+    },
+    {
+      "epoch": 16.004830917874397,
+      "grad_norm": 1.5131254196166992,
+      "learning_rate": 9.855072463768118e-06,
+      "loss": 0.5318,
+      "step": 1170
+    },
+    {
+      "epoch": 16.005797101449275,
+      "grad_norm": 1.1009883880615234,
+      "learning_rate": 9.844337090713904e-06,
+      "loss": 0.5071,
+      "step": 1180
+    },
+    {
+      "epoch": 16.006763285024153,
+      "grad_norm": 0.46484410762786865,
+      "learning_rate": 9.833601717659689e-06,
+      "loss": 0.104,
+      "step": 1190
+    },
+    {
+      "epoch": 16.006763285024153,
+      "eval_accuracy": 0.717948717948718,
+      "eval_loss": 0.8999806642532349,
+      "eval_runtime": 18.5229,
+      "eval_samples_per_second": 4.211,
+      "eval_steps_per_second": 1.08,
+      "step": 1190
+    },
+    {
+      "epoch": 17.000966183574878,
+      "grad_norm": 0.40560030937194824,
+      "learning_rate": 9.822866344605476e-06,
+      "loss": 0.4563,
+      "step": 1200
+    },
+    {
+      "epoch": 17.00193236714976,
+      "grad_norm": 124.55720520019531,
+      "learning_rate": 9.812130971551262e-06,
+      "loss": 0.5613,
+      "step": 1210
+    },
+    {
+      "epoch": 17.002898550724638,
+      "grad_norm": 4.143601417541504,
+      "learning_rate": 9.801395598497048e-06,
+      "loss": 0.3869,
+      "step": 1220
+    },
+    {
+      "epoch": 17.003864734299516,
+      "grad_norm": 0.04259469732642174,
+      "learning_rate": 9.790660225442835e-06,
+      "loss": 0.0657,
+      "step": 1230
+    },
+    {
+      "epoch": 17.004830917874397,
+      "grad_norm": 81.4675064086914,
+      "learning_rate": 9.779924852388621e-06,
+      "loss": 0.239,
+      "step": 1240
+    },
+    {
+      "epoch": 17.005797101449275,
+      "grad_norm": 0.18485671281814575,
+      "learning_rate": 9.769189479334408e-06,
+      "loss": 0.0619,
+      "step": 1250
+    },
+    {
+      "epoch": 17.006763285024153,
+      "grad_norm": 0.026976848021149635,
+      "learning_rate": 9.758454106280194e-06,
+      "loss": 0.5647,
+      "step": 1260
+    },
+    {
+      "epoch": 17.006763285024153,
+      "eval_accuracy": 0.7051282051282052,
+      "eval_loss": 1.4782413244247437,
+      "eval_runtime": 18.6906,
+      "eval_samples_per_second": 4.173,
+      "eval_steps_per_second": 1.07,
+      "step": 1260
+    },
+    {
+      "epoch": 18.000966183574878,
+      "grad_norm": 54.548683166503906,
+      "learning_rate": 9.74771873322598e-06,
+      "loss": 0.183,
+      "step": 1270
+    },
+    {
+      "epoch": 18.00193236714976,
+      "grad_norm": 12.151978492736816,
+      "learning_rate": 9.736983360171767e-06,
+      "loss": 0.5206,
+      "step": 1280
+    },
+    {
+      "epoch": 18.002898550724638,
+      "grad_norm": 275.7248840332031,
+      "learning_rate": 9.726247987117554e-06,
+      "loss": 0.3181,
+      "step": 1290
+    },
+    {
+      "epoch": 18.003864734299516,
+      "grad_norm": 0.09706564992666245,
+      "learning_rate": 9.71551261406334e-06,
+      "loss": 0.1154,
+      "step": 1300
+    },
+    {
+      "epoch": 18.004830917874397,
+      "grad_norm": 0.7155212163925171,
+      "learning_rate": 9.704777241009125e-06,
+      "loss": 0.2555,
+      "step": 1310
+    },
+    {
+      "epoch": 18.005797101449275,
+      "grad_norm": 97.17144012451172,
+      "learning_rate": 9.694041867954911e-06,
+      "loss": 0.5985,
+      "step": 1320
+    },
+    {
+      "epoch": 18.006763285024153,
+      "grad_norm": 0.2159235030412674,
+      "learning_rate": 9.683306494900698e-06,
+      "loss": 0.783,
+      "step": 1330
+    },
+    {
+      "epoch": 18.006763285024153,
+      "eval_accuracy": 0.8205128205128205,
+      "eval_loss": 0.8539489507675171,
+      "eval_runtime": 19.5687,
+      "eval_samples_per_second": 3.986,
+      "eval_steps_per_second": 1.022,
+      "step": 1330
+    },
+    {
+      "epoch": 19.000966183574878,
+      "grad_norm": 117.28670501708984,
+      "learning_rate": 9.672571121846484e-06,
+      "loss": 0.3806,
+      "step": 1340
+    },
+    {
+      "epoch": 19.00193236714976,
+      "grad_norm": 0.10745132714509964,
+      "learning_rate": 9.66183574879227e-06,
+      "loss": 0.2618,
+      "step": 1350
+    },
+    {
+      "epoch": 19.002898550724638,
+      "grad_norm": 0.10657572746276855,
+      "learning_rate": 9.651100375738057e-06,
+      "loss": 0.1682,
+      "step": 1360
+    },
+    {
+      "epoch": 19.003864734299516,
+      "grad_norm": 17.445707321166992,
+      "learning_rate": 9.640365002683844e-06,
+      "loss": 0.0466,
+      "step": 1370
+    },
+    {
+      "epoch": 19.004830917874397,
+      "grad_norm": 0.01827850751578808,
+      "learning_rate": 9.62962962962963e-06,
+      "loss": 0.3148,
+      "step": 1380
+    },
+    {
+      "epoch": 19.005797101449275,
+      "grad_norm": 67.58177947998047,
+      "learning_rate": 9.618894256575418e-06,
+      "loss": 0.5658,
+      "step": 1390
+    },
+    {
+      "epoch": 19.006763285024153,
+      "grad_norm": 0.4448859393596649,
+      "learning_rate": 9.608158883521203e-06,
+      "loss": 0.5938,
+      "step": 1400
+    },
+    {
+      "epoch": 19.006763285024153,
+      "eval_accuracy": 0.7564102564102564,
+      "eval_loss": 0.8426063656806946,
+      "eval_runtime": 23.6146,
+      "eval_samples_per_second": 3.303,
+      "eval_steps_per_second": 0.847,
+      "step": 1400
+    },
+    {
+      "epoch": 20.000966183574878,
+      "grad_norm": 119.15202331542969,
+      "learning_rate": 9.59742351046699e-06,
+      "loss": 0.2652,
+      "step": 1410
+    },
+    {
+      "epoch": 20.00193236714976,
+      "grad_norm": 52.361080169677734,
+      "learning_rate": 9.586688137412776e-06,
+      "loss": 0.5006,
+      "step": 1420
+    },
+    {
+      "epoch": 20.002898550724638,
+      "grad_norm": 33.284122467041016,
+      "learning_rate": 9.575952764358562e-06,
+      "loss": 0.3182,
+      "step": 1430
+    },
+    {
+      "epoch": 20.003864734299516,
+      "grad_norm": 0.39933377504348755,
+      "learning_rate": 9.565217391304349e-06,
+      "loss": 0.0969,
+      "step": 1440
+    },
+    {
+      "epoch": 20.004830917874397,
+      "grad_norm": 374.1847839355469,
+      "learning_rate": 9.554482018250135e-06,
+      "loss": 0.0703,
+      "step": 1450
+    },
+    {
+      "epoch": 20.005797101449275,
+      "grad_norm": 0.1522367000579834,
+      "learning_rate": 9.543746645195922e-06,
+      "loss": 0.2595,
+      "step": 1460
+    },
+    {
+      "epoch": 20.006763285024153,
+      "grad_norm": 0.15178386867046356,
+      "learning_rate": 9.533011272141708e-06,
+      "loss": 0.5341,
+      "step": 1470
+    },
+    {
+      "epoch": 20.006763285024153,
+      "eval_accuracy": 0.7948717948717948,
+      "eval_loss": 0.9861828684806824,
+      "eval_runtime": 20.329,
+      "eval_samples_per_second": 3.837,
+      "eval_steps_per_second": 0.984,
+      "step": 1470
+    },
+    {
+      "epoch": 21.000966183574878,
+      "grad_norm": 244.99745178222656,
+      "learning_rate": 9.522275899087494e-06,
+      "loss": 0.2293,
+      "step": 1480
+    },
+    {
+      "epoch": 21.00193236714976,
+      "grad_norm": 31.514474868774414,
+      "learning_rate": 9.511540526033281e-06,
+      "loss": 0.0871,
+      "step": 1490
+    },
+    {
+      "epoch": 21.002898550724638,
+      "grad_norm": 0.014621300622820854,
+      "learning_rate": 9.500805152979067e-06,
+      "loss": 0.2701,
+      "step": 1500
+    },
+    {
+      "epoch": 21.003864734299516,
+      "grad_norm": 0.14359234273433685,
+      "learning_rate": 9.490069779924854e-06,
+      "loss": 0.2155,
+      "step": 1510
+    },
+    {
+      "epoch": 21.004830917874397,
+      "grad_norm": 17.387298583984375,
+      "learning_rate": 9.479334406870639e-06,
+      "loss": 0.6972,
+      "step": 1520
+    },
+    {
+      "epoch": 21.005797101449275,
+      "grad_norm": 263.881591796875,
+      "learning_rate": 9.468599033816425e-06,
+      "loss": 0.495,
+      "step": 1530
+    },
+    {
+      "epoch": 21.006763285024153,
+      "grad_norm": 0.6225783228874207,
+      "learning_rate": 9.457863660762211e-06,
+      "loss": 0.3391,
+      "step": 1540
+    },
+    {
+      "epoch": 21.006763285024153,
+      "eval_accuracy": 0.7692307692307693,
+      "eval_loss": 1.1159329414367676,
+      "eval_runtime": 20.3876,
+      "eval_samples_per_second": 3.826,
+      "eval_steps_per_second": 0.981,
+      "step": 1540
+    },
+    {
+      "epoch": 22.000966183574878,
+      "grad_norm": 2.346330404281616,
+      "learning_rate": 9.447128287707998e-06,
+      "loss": 0.0129,
+      "step": 1550
+    },
+    {
+      "epoch": 22.00193236714976,
+      "grad_norm": 0.19982995092868805,
+      "learning_rate": 9.436392914653784e-06,
+      "loss": 0.2005,
+      "step": 1560
+    },
+    {
+      "epoch": 22.002898550724638,
+      "grad_norm": 0.0399252213537693,
+      "learning_rate": 9.42565754159957e-06,
+      "loss": 0.0019,
+      "step": 1570
+    },
+    {
+      "epoch": 22.003864734299516,
+      "grad_norm": 0.027430692687630653,
+      "learning_rate": 9.414922168545357e-06,
+      "loss": 0.0012,
+      "step": 1580
+    },
+    {
+      "epoch": 22.004830917874397,
+      "grad_norm": 0.05206064507365227,
+      "learning_rate": 9.404186795491144e-06,
+      "loss": 0.3566,
+      "step": 1590
+    },
+    {
+      "epoch": 22.005797101449275,
+      "grad_norm": 0.07161853462457657,
+      "learning_rate": 9.39345142243693e-06,
+      "loss": 0.3149,
+      "step": 1600
+    },
+    {
+      "epoch": 22.006763285024153,
+      "grad_norm": 0.1686069220304489,
+      "learning_rate": 9.382716049382717e-06,
+      "loss": 0.2071,
+      "step": 1610
+    },
+    {
+      "epoch": 22.006763285024153,
+      "eval_accuracy": 0.7692307692307693,
+      "eval_loss": 1.5833228826522827,
+      "eval_runtime": 20.3409,
+      "eval_samples_per_second": 3.835,
+      "eval_steps_per_second": 0.983,
+      "step": 1610
+    },
+    {
+      "epoch": 23.000966183574878,
+      "grad_norm": 0.11971097439527512,
+      "learning_rate": 9.371980676328503e-06,
+      "loss": 0.1224,
+      "step": 1620
+    },
+    {
+      "epoch": 23.00193236714976,
+      "grad_norm": 0.04427675902843475,
+      "learning_rate": 9.36124530327429e-06,
+      "loss": 0.2063,
+      "step": 1630
+    },
+    {
+      "epoch": 23.002898550724638,
+      "grad_norm": 0.017134735360741615,
+      "learning_rate": 9.350509930220076e-06,
+      "loss": 0.0009,
+      "step": 1640
+    },
+    {
+      "epoch": 23.003864734299516,
+      "grad_norm": 1.9943066835403442,
+      "learning_rate": 9.339774557165862e-06,
+      "loss": 0.2292,
+      "step": 1650
+    },
+    {
+      "epoch": 23.004830917874397,
+      "grad_norm": 53.962642669677734,
+      "learning_rate": 9.329039184111649e-06,
+      "loss": 0.3195,
+      "step": 1660
+    },
+    {
+      "epoch": 23.005797101449275,
+      "grad_norm": 0.11377642303705215,
+      "learning_rate": 9.318303811057435e-06,
+      "loss": 0.1909,
+      "step": 1670
+    },
+    {
+      "epoch": 23.006763285024153,
+      "grad_norm": 0.021974513307213783,
+      "learning_rate": 9.307568438003222e-06,
+      "loss": 0.1159,
+      "step": 1680
+    },
+    {
+      "epoch": 23.006763285024153,
+      "eval_accuracy": 0.8205128205128205,
+      "eval_loss": 1.0204980373382568,
+      "eval_runtime": 20.0597,
+      "eval_samples_per_second": 3.888,
+      "eval_steps_per_second": 0.997,
+      "step": 1680
+    },
+    {
+      "epoch": 24.000966183574878,
+      "grad_norm": 0.022601498290896416,
+      "learning_rate": 9.296833064949008e-06,
+      "loss": 0.2401,
+      "step": 1690
+    },
+    {
+      "epoch": 24.00193236714976,
+      "grad_norm": 0.023662317544221878,
+      "learning_rate": 9.286097691894795e-06,
+      "loss": 0.003,
+      "step": 1700
+    },
+    {
+      "epoch": 24.002898550724638,
+      "grad_norm": 0.17592191696166992,
+      "learning_rate": 9.275362318840581e-06,
+      "loss": 0.1163,
+      "step": 1710
+    },
+    {
+      "epoch": 24.003864734299516,
+      "grad_norm": 0.014995891600847244,
+      "learning_rate": 9.264626945786368e-06,
+      "loss": 0.0144,
+      "step": 1720
+    },
+    {
+      "epoch": 24.004830917874397,
+      "grad_norm": 0.16420097649097443,
+      "learning_rate": 9.253891572732154e-06,
+      "loss": 0.1508,
+      "step": 1730
+    },
+    {
+      "epoch": 24.005797101449275,
+      "grad_norm": 11.045714378356934,
+      "learning_rate": 9.243156199677939e-06,
+      "loss": 0.6033,
+      "step": 1740
+    },
+    {
+      "epoch": 24.006763285024153,
+      "grad_norm": 0.010024874471127987,
+      "learning_rate": 9.232420826623725e-06,
+      "loss": 0.1579,
+      "step": 1750
+    },
+    {
+      "epoch": 24.006763285024153,
+      "eval_accuracy": 0.7692307692307693,
+      "eval_loss": 1.4633172750473022,
+      "eval_runtime": 21.196,
+      "eval_samples_per_second": 3.68,
+      "eval_steps_per_second": 0.944,
+      "step": 1750
+    },
+    {
+      "epoch": 25.000966183574878,
+      "grad_norm": 3.15612530708313,
+      "learning_rate": 9.221685453569512e-06,
+      "loss": 0.1799,
+      "step": 1760
+    },
+    {
+      "epoch": 25.00193236714976,
+      "grad_norm": 0.048897918313741684,
+      "learning_rate": 9.210950080515298e-06,
+      "loss": 0.2754,
+      "step": 1770
+    },
+    {
+      "epoch": 25.002898550724638,
+      "grad_norm": 0.013033032417297363,
+      "learning_rate": 9.200214707461085e-06,
+      "loss": 0.078,
+      "step": 1780
+    },
+    {
+      "epoch": 25.003864734299516,
+      "grad_norm": 228.9864044189453,
+      "learning_rate": 9.189479334406871e-06,
+      "loss": 0.0223,
+      "step": 1790
+    },
+    {
+      "epoch": 25.004830917874397,
+      "grad_norm": 0.38404083251953125,
+      "learning_rate": 9.178743961352658e-06,
+      "loss": 0.09,
+      "step": 1800
+    },
+    {
+      "epoch": 25.005797101449275,
+      "grad_norm": 9.970251083374023,
+      "learning_rate": 9.168008588298444e-06,
+      "loss": 0.0889,
+      "step": 1810
+    },
+    {
+      "epoch": 25.006763285024153,
+      "grad_norm": 0.005024947226047516,
+      "learning_rate": 9.15727321524423e-06,
+      "loss": 0.1042,
+      "step": 1820
+    },
+    {
+      "epoch": 25.006763285024153,
+      "eval_accuracy": 0.7564102564102564,
+      "eval_loss": 1.5863817930221558,
+      "eval_runtime": 16.6058,
+      "eval_samples_per_second": 4.697,
+      "eval_steps_per_second": 1.204,
+      "step": 1820
+    },
+    {
+      "epoch": 26.000966183574878,
+      "grad_norm": 34.71379089355469,
+      "learning_rate": 9.146537842190017e-06,
+      "loss": 0.4249,
+      "step": 1830
+    },
+    {
+      "epoch": 26.00193236714976,
+      "grad_norm": 0.02387184463441372,
+      "learning_rate": 9.135802469135803e-06,
+      "loss": 0.1022,
+      "step": 1840
+    },
+    {
+      "epoch": 26.002898550724638,
+      "grad_norm": 456.8200988769531,
+      "learning_rate": 9.12506709608159e-06,
+      "loss": 0.3789,
+      "step": 1850
+    },
+    {
+      "epoch": 26.003864734299516,
+      "grad_norm": 0.012509716674685478,
+      "learning_rate": 9.114331723027376e-06,
+      "loss": 0.1754,
+      "step": 1860
+    },
+    {
+      "epoch": 26.004830917874397,
+      "grad_norm": 0.04585114121437073,
+      "learning_rate": 9.103596349973163e-06,
+      "loss": 0.0927,
+      "step": 1870
+    },
+    {
+      "epoch": 26.005797101449275,
+      "grad_norm": 0.029922403395175934,
+      "learning_rate": 9.092860976918949e-06,
+      "loss": 0.3417,
+      "step": 1880
+    },
+    {
+      "epoch": 26.006763285024153,
+      "grad_norm": 0.011729502119123936,
+      "learning_rate": 9.082125603864736e-06,
+      "loss": 0.1466,
+      "step": 1890
+    },
+    {
+      "epoch": 26.006763285024153,
+      "eval_accuracy": 0.7692307692307693,
+      "eval_loss": 1.2990210056304932,
+      "eval_runtime": 16.4386,
+      "eval_samples_per_second": 4.745,
+      "eval_steps_per_second": 1.217,
+      "step": 1890
+    },
+    {
+      "epoch": 27.000966183574878,
+      "grad_norm": 0.03241152688860893,
+      "learning_rate": 9.071390230810522e-06,
+      "loss": 0.0027,
+      "step": 1900
+    },
+    {
+      "epoch": 27.00193236714976,
+      "grad_norm": 19.074296951293945,
+      "learning_rate": 9.060654857756308e-06,
+      "loss": 0.1867,
+      "step": 1910
+    },
+    {
+      "epoch": 27.002898550724638,
+      "grad_norm": 0.04132108390331268,
+      "learning_rate": 9.049919484702095e-06,
+      "loss": 0.0007,
+      "step": 1920
+    },
+    {
+      "epoch": 27.003864734299516,
+      "grad_norm": 0.014680879190564156,
+      "learning_rate": 9.039184111647881e-06,
+      "loss": 0.0004,
+      "step": 1930
+    },
+    {
+      "epoch": 27.004830917874397,
+      "grad_norm": 0.01582922227680683,
+      "learning_rate": 9.028448738593668e-06,
+      "loss": 0.001,
+      "step": 1940
+    },
+    {
+      "epoch": 27.005797101449275,
+      "grad_norm": 0.02607502043247223,
+      "learning_rate": 9.017713365539453e-06,
+      "loss": 0.0036,
+      "step": 1950
+    },
+    {
+      "epoch": 27.006763285024153,
+      "grad_norm": 0.021080270409584045,
+      "learning_rate": 9.006977992485239e-06,
+      "loss": 0.0006,
+      "step": 1960
+    },
+    {
+      "epoch": 27.006763285024153,
+      "eval_accuracy": 0.7564102564102564,
+      "eval_loss": 1.5596784353256226,
+      "eval_runtime": 16.1439,
+      "eval_samples_per_second": 4.832,
+      "eval_steps_per_second": 1.239,
+      "step": 1960
+    },
+    {
+      "epoch": 28.000966183574878,
+      "grad_norm": 0.00953815970569849,
+      "learning_rate": 8.996242619431025e-06,
+      "loss": 0.1386,
+      "step": 1970
+    },
+    {
+      "epoch": 28.00193236714976,
+      "grad_norm": 148.93067932128906,
+      "learning_rate": 8.985507246376812e-06,
+      "loss": 0.2969,
+      "step": 1980
+    },
+    {
+      "epoch": 28.002898550724638,
+      "grad_norm": 0.05550064891576767,
+      "learning_rate": 8.974771873322598e-06,
+      "loss": 0.0064,
+      "step": 1990
+    },
+    {
+      "epoch": 28.003864734299516,
+      "grad_norm": 0.03587929904460907,
+      "learning_rate": 8.964036500268385e-06,
+      "loss": 0.0021,
+      "step": 2000
+    },
+    {
+      "epoch": 28.004830917874397,
+      "grad_norm": 0.009905187413096428,
+      "learning_rate": 8.953301127214171e-06,
+      "loss": 0.0016,
+      "step": 2010
+    },
+    {
+      "epoch": 28.005797101449275,
+      "grad_norm": 0.016557922586798668,
+      "learning_rate": 8.942565754159958e-06,
+      "loss": 0.332,
+      "step": 2020
+    },
+    {
+      "epoch": 28.006763285024153,
+      "grad_norm": 0.0018835013033822179,
+      "learning_rate": 8.931830381105744e-06,
+      "loss": 0.001,
+      "step": 2030
+    },
+    {
+      "epoch": 28.006763285024153,
+      "eval_accuracy": 0.7564102564102564,
+      "eval_loss": 1.6433550119400024,
+      "eval_runtime": 16.9604,
+      "eval_samples_per_second": 4.599,
+      "eval_steps_per_second": 1.179,
+      "step": 2030
+    },
+    {
+      "epoch": 29.000966183574878,
+      "grad_norm": 0.011498975567519665,
+      "learning_rate": 8.92109500805153e-06,
+      "loss": 0.0006,
+      "step": 2040
+    },
+    {
+      "epoch": 29.00193236714976,
+      "grad_norm": 0.022594112902879715,
+      "learning_rate": 8.910359634997317e-06,
+      "loss": 0.0019,
+      "step": 2050
+    },
+    {
+      "epoch": 29.002898550724638,
+      "grad_norm": 108.30265045166016,
+      "learning_rate": 8.899624261943104e-06,
+      "loss": 0.2073,
+      "step": 2060
+    },
+    {
+      "epoch": 29.003864734299516,
+      "grad_norm": 0.018151551485061646,
+      "learning_rate": 8.888888888888888e-06,
+      "loss": 0.0019,
+      "step": 2070
+    },
+    {
+      "epoch": 29.004830917874397,
+      "grad_norm": 171.54039001464844,
+      "learning_rate": 8.878153515834675e-06,
+      "loss": 0.5704,
+      "step": 2080
+    },
+    {
+      "epoch": 29.005797101449275,
+      "grad_norm": 0.12190333008766174,
+      "learning_rate": 8.867418142780463e-06,
+      "loss": 0.2071,
+      "step": 2090
+    },
+    {
+      "epoch": 29.006763285024153,
+      "grad_norm": 0.1234963983297348,
+      "learning_rate": 8.85668276972625e-06,
+      "loss": 0.0096,
+      "step": 2100
+    },
+    {
+      "epoch": 29.006763285024153,
+      "eval_accuracy": 0.782051282051282,
+      "eval_loss": 1.4161337614059448,
+      "eval_runtime": 75.4835,
+      "eval_samples_per_second": 1.033,
+      "eval_steps_per_second": 0.265,
+      "step": 2100
+    },
+    {
+      "epoch": 30.000966183574878,
+      "grad_norm": 0.02181861735880375,
+      "learning_rate": 8.845947396672036e-06,
+      "loss": 0.072,
+      "step": 2110
+    },
+    {
+      "epoch": 30.00193236714976,
+      "grad_norm": 0.012783943675458431,
+      "learning_rate": 8.835212023617822e-06,
+      "loss": 0.0005,
+      "step": 2120
+    },
+    {
+      "epoch": 30.002898550724638,
+      "grad_norm": 0.009668432176113129,
+      "learning_rate": 8.824476650563609e-06,
+      "loss": 0.1864,
+      "step": 2130
+    },
+    {
+      "epoch": 30.003864734299516,
+      "grad_norm": 171.04098510742188,
+      "learning_rate": 8.813741277509395e-06,
+      "loss": 0.0289,
+      "step": 2140
+    },
+    {
+      "epoch": 30.004830917874397,
+      "grad_norm": 2.430178642272949,
+      "learning_rate": 8.803005904455182e-06,
+      "loss": 0.2184,
+      "step": 2150
+    },
+    {
+      "epoch": 30.005797101449275,
+      "grad_norm": 0.37614426016807556,
+      "learning_rate": 8.792270531400966e-06,
+      "loss": 0.1647,
+      "step": 2160
+    },
+    {
+      "epoch": 30.006763285024153,
+      "grad_norm": 0.03991688787937164,
+      "learning_rate": 8.781535158346753e-06,
+      "loss": 0.0015,
+      "step": 2170
+    },
+    {
+      "epoch": 30.006763285024153,
+      "eval_accuracy": 0.7948717948717948,
+      "eval_loss": 1.3385039567947388,
+      "eval_runtime": 22.7136,
+      "eval_samples_per_second": 3.434,
+      "eval_steps_per_second": 0.881,
+      "step": 2170
+    },
+    {
+      "epoch": 31.000966183574878,
+      "grad_norm": 0.011698647402226925,
+      "learning_rate": 8.77079978529254e-06,
+      "loss": 0.1503,
+      "step": 2180
+    },
+    {
+      "epoch": 31.00193236714976,
+      "grad_norm": 0.005670130252838135,
+      "learning_rate": 8.760064412238326e-06,
+      "loss": 0.0006,
+      "step": 2190
+    },
+    {
+      "epoch": 31.002898550724638,
+      "grad_norm": 0.05591421201825142,
+      "learning_rate": 8.749329039184112e-06,
+      "loss": 0.0745,
+      "step": 2200
+    },
+    {
+      "epoch": 31.003864734299516,
+      "grad_norm": 3.357697010040283,
+      "learning_rate": 8.738593666129899e-06,
+      "loss": 0.0008,
+      "step": 2210
+    },
+    {
+      "epoch": 31.004830917874397,
+      "grad_norm": 51.18307876586914,
+      "learning_rate": 8.727858293075685e-06,
+      "loss": 0.167,
+      "step": 2220
+    },
+    {
+      "epoch": 31.005797101449275,
+      "grad_norm": 73.81510925292969,
+      "learning_rate": 8.717122920021472e-06,
+      "loss": 0.099,
+      "step": 2230
+    },
+    {
+      "epoch": 31.006763285024153,
+      "grad_norm": 0.019633440300822258,
+      "learning_rate": 8.706387546967258e-06,
+      "loss": 0.3627,
+      "step": 2240
+    },
+    {
+      "epoch": 31.006763285024153,
+      "eval_accuracy": 0.7435897435897436,
+      "eval_loss": 1.7863702774047852,
+      "eval_runtime": 22.6796,
+      "eval_samples_per_second": 3.439,
+      "eval_steps_per_second": 0.882,
+      "step": 2240
+    },
+    {
+      "epoch": 32.00096618357488,
+      "grad_norm": 0.0671672448515892,
+      "learning_rate": 8.695652173913044e-06,
+      "loss": 0.0005,
+      "step": 2250
+    },
+    {
+      "epoch": 32.001932367149756,
+      "grad_norm": 0.35884493589401245,
+      "learning_rate": 8.684916800858831e-06,
+      "loss": 0.0008,
+      "step": 2260
+    },
+    {
+      "epoch": 32.00289855072464,
+      "grad_norm": 0.008876653388142586,
+      "learning_rate": 8.674181427804617e-06,
+      "loss": 0.0002,
+      "step": 2270
+    },
+    {
+      "epoch": 32.00386473429952,
+      "grad_norm": 0.0027054892852902412,
+      "learning_rate": 8.663446054750402e-06,
+      "loss": 0.0003,
+      "step": 2280
+    },
+    {
+      "epoch": 32.00483091787439,
+      "grad_norm": 0.00598740391433239,
+      "learning_rate": 8.652710681696189e-06,
+      "loss": 0.0006,
+      "step": 2290
+    },
+    {
+      "epoch": 32.005797101449275,
+      "grad_norm": 0.008918968960642815,
+      "learning_rate": 8.641975308641975e-06,
+      "loss": 0.0182,
+      "step": 2300
+    },
+    {
+      "epoch": 32.00676328502416,
+      "grad_norm": 0.005757227074354887,
+      "learning_rate": 8.631239935587761e-06,
+      "loss": 0.1541,
+      "step": 2310
+    },
+    {
+      "epoch": 32.00676328502416,
+      "eval_accuracy": 0.782051282051282,
+      "eval_loss": 1.5617612600326538,
+      "eval_runtime": 24.1692,
+      "eval_samples_per_second": 3.227,
+      "eval_steps_per_second": 0.827,
+      "step": 2310
+    },
+    {
+      "epoch": 33.00096618357488,
+      "grad_norm": 0.006783771328628063,
+      "learning_rate": 8.62050456253355e-06,
+      "loss": 0.0002,
+      "step": 2320
+    },
+    {
+      "epoch": 33.001932367149756,
+      "grad_norm": 0.037278104573488235,
+      "learning_rate": 8.609769189479336e-06,
+      "loss": 0.0176,
+      "step": 2330
+    },
+    {
+      "epoch": 33.00289855072464,
+      "grad_norm": 427.9874267578125,
+      "learning_rate": 8.599033816425122e-06,
+      "loss": 0.0923,
+      "step": 2340
+    },
+    {
+      "epoch": 33.00386473429952,
+      "grad_norm": 0.005532737821340561,
+      "learning_rate": 8.588298443370909e-06,
+      "loss": 0.0549,
+      "step": 2350
+    },
+    {
+      "epoch": 33.00483091787439,
+      "grad_norm": 0.0045069679617881775,
+      "learning_rate": 8.577563070316695e-06,
+      "loss": 0.0134,
+      "step": 2360
+    },
+    {
+      "epoch": 33.005797101449275,
+      "grad_norm": 231.7147979736328,
+      "learning_rate": 8.56682769726248e-06,
+      "loss": 0.1822,
+      "step": 2370
+    },
+    {
+      "epoch": 33.00676328502416,
+      "grad_norm": 0.005993337836116552,
+      "learning_rate": 8.556092324208267e-06,
+      "loss": 0.1285,
+      "step": 2380
+    },
+    {
+      "epoch": 33.00676328502416,
+      "eval_accuracy": 0.7948717948717948,
+      "eval_loss": 1.30621337890625,
+      "eval_runtime": 20.254,
+      "eval_samples_per_second": 3.851,
+      "eval_steps_per_second": 0.987,
+      "step": 2380
+    },
+    {
+      "epoch": 34.00096618357488,
+      "grad_norm": 0.010524412617087364,
+      "learning_rate": 8.545356951154053e-06,
+      "loss": 0.0002,
+      "step": 2390
+    },
+    {
+      "epoch": 34.001932367149756,
+      "grad_norm": 0.005317069590091705,
+      "learning_rate": 8.53462157809984e-06,
+      "loss": 0.1733,
+      "step": 2400
+    },
+    {
+      "epoch": 34.00289855072464,
+      "grad_norm": 0.007308751344680786,
+      "learning_rate": 8.523886205045626e-06,
+      "loss": 0.1447,
+      "step": 2410
+    },
+    {
+      "epoch": 34.00386473429952,
+      "grad_norm": 0.6433327794075012,
+      "learning_rate": 8.513150831991412e-06,
+      "loss": 0.0151,
+      "step": 2420
+    },
+    {
+      "epoch": 34.00483091787439,
+      "grad_norm": 0.013438474386930466,
+      "learning_rate": 8.502415458937199e-06,
+      "loss": 0.0002,
+      "step": 2430
+    },
+    {
+      "epoch": 34.005797101449275,
+      "grad_norm": 0.006189200561493635,
+      "learning_rate": 8.491680085882985e-06,
+      "loss": 0.0764,
+      "step": 2440
+    },
+    {
+      "epoch": 34.00676328502416,
+      "grad_norm": 0.011971415020525455,
+      "learning_rate": 8.480944712828772e-06,
+      "loss": 0.2193,
+      "step": 2450
+    },
+    {
+      "epoch": 34.00676328502416,
+      "eval_accuracy": 0.782051282051282,
+      "eval_loss": 1.555395483970642,
+      "eval_runtime": 20.8944,
+      "eval_samples_per_second": 3.733,
+      "eval_steps_per_second": 0.957,
+      "step": 2450
+    },
+    {
+      "epoch": 35.00096618357488,
+      "grad_norm": 0.007673191372305155,
+      "learning_rate": 8.470209339774558e-06,
+      "loss": 0.0002,
+      "step": 2460
+    },
+    {
+      "epoch": 35.001932367149756,
+      "grad_norm": 1.346144437789917,
+      "learning_rate": 8.459473966720345e-06,
+      "loss": 0.1336,
+      "step": 2470
+    },
+    {
+      "epoch": 35.00289855072464,
+      "grad_norm": 0.05735534802079201,
+      "learning_rate": 8.448738593666131e-06,
+      "loss": 0.0006,
+      "step": 2480
+    },
+    {
+      "epoch": 35.00386473429952,
+      "grad_norm": 0.0024482065346091986,
+      "learning_rate": 8.438003220611916e-06,
+      "loss": 0.0002,
+      "step": 2490
+    },
+    {
+      "epoch": 35.00483091787439,
+      "grad_norm": 0.24069902300834656,
+      "learning_rate": 8.427267847557702e-06,
+      "loss": 0.0247,
+      "step": 2500
+    },
+    {
+      "epoch": 35.005797101449275,
+      "grad_norm": 0.0022392012178897858,
+      "learning_rate": 8.416532474503489e-06,
+      "loss": 0.0005,
+      "step": 2510
+    },
+    {
+      "epoch": 35.00676328502416,
+      "grad_norm": 0.0014107138849794865,
+      "learning_rate": 8.405797101449275e-06,
+      "loss": 0.0002,
+      "step": 2520
+    },
+    {
+      "epoch": 35.00676328502416,
+      "eval_accuracy": 0.7948717948717948,
+      "eval_loss": 1.5444982051849365,
+      "eval_runtime": 20.3236,
+      "eval_samples_per_second": 3.838,
+      "eval_steps_per_second": 0.984,
+      "step": 2520
+    },
+    {
+      "epoch": 36.00096618357488,
+      "grad_norm": 0.001967865275219083,
+      "learning_rate": 8.395061728395062e-06,
+      "loss": 0.0033,
+      "step": 2530
+    },
+    {
+      "epoch": 36.001932367149756,
+      "grad_norm": 403.927490234375,
+      "learning_rate": 8.384326355340848e-06,
+      "loss": 0.0848,
+      "step": 2540
+    },
+    {
+      "epoch": 36.00289855072464,
+      "grad_norm": 0.012290756218135357,
+      "learning_rate": 8.373590982286636e-06,
+      "loss": 0.002,
+      "step": 2550
+    },
+    {
+      "epoch": 36.00386473429952,
+      "grad_norm": 0.007809770293533802,
+      "learning_rate": 8.362855609232423e-06,
+      "loss": 0.0001,
+      "step": 2560
+    },
+    {
+      "epoch": 36.00483091787439,
+      "grad_norm": 167.53094482421875,
+      "learning_rate": 8.352120236178209e-06,
+      "loss": 0.1313,
+      "step": 2570
+    },
+    {
+      "epoch": 36.005797101449275,
+      "grad_norm": 0.005764774978160858,
+      "learning_rate": 8.341384863123994e-06,
+      "loss": 0.0001,
+      "step": 2580
+    },
+    {
+      "epoch": 36.00676328502416,
+      "grad_norm": 0.006284950766712427,
+      "learning_rate": 8.33064949006978e-06,
+      "loss": 0.0003,
+      "step": 2590
+    },
+    {
+      "epoch": 36.00676328502416,
+      "eval_accuracy": 0.782051282051282,
+      "eval_loss": 1.718872308731079,
+      "eval_runtime": 20.0134,
+      "eval_samples_per_second": 3.897,
+      "eval_steps_per_second": 0.999,
+      "step": 2590
+    },
+    {
+      "epoch": 37.00096618357488,
+      "grad_norm": 0.002441326156258583,
+      "learning_rate": 8.319914117015567e-06,
+      "loss": 0.0001,
+      "step": 2600
+    },
+    {
+      "epoch": 37.001932367149756,
+      "grad_norm": 0.005556989461183548,
+      "learning_rate": 8.309178743961353e-06,
+      "loss": 0.0004,
+      "step": 2610
+    },
+    {
+      "epoch": 37.00289855072464,
+      "grad_norm": 0.0032976714428514242,
+      "learning_rate": 8.29844337090714e-06,
+      "loss": 0.0001,
+      "step": 2620
+    },
+    {
+      "epoch": 37.00386473429952,
+      "grad_norm": 0.006368944887071848,
+      "learning_rate": 8.287707997852926e-06,
+      "loss": 0.0001,
+      "step": 2630
+    },
+    {
+      "epoch": 37.00483091787439,
+      "grad_norm": 0.005885376129299402,
+      "learning_rate": 8.276972624798713e-06,
+      "loss": 0.0001,
+      "step": 2640
+    },
+    {
+      "epoch": 37.005797101449275,
+      "grad_norm": 0.01179414801299572,
+      "learning_rate": 8.266237251744499e-06,
+      "loss": 0.2815,
+      "step": 2650
+    },
+    {
+      "epoch": 37.00676328502416,
+      "grad_norm": 0.0038128988817334175,
+      "learning_rate": 8.255501878690286e-06,
+      "loss": 0.092,
+      "step": 2660
+    },
+    {
+      "epoch": 37.00676328502416,
+      "eval_accuracy": 0.7692307692307693,
+      "eval_loss": 1.4979729652404785,
+      "eval_runtime": 14.9225,
+      "eval_samples_per_second": 5.227,
+      "eval_steps_per_second": 1.34,
+      "step": 2660
+    },
+    {
+      "epoch": 38.00096618357488,
+      "grad_norm": 1.8284555673599243,
+      "learning_rate": 8.244766505636072e-06,
+      "loss": 0.0003,
+      "step": 2670
+    },
+    {
+      "epoch": 38.001932367149756,
+      "grad_norm": 0.010567315854132175,
+      "learning_rate": 8.234031132581858e-06,
+      "loss": 0.0006,
+      "step": 2680
+    },
+    {
+      "epoch": 38.00289855072464,
+      "grad_norm": 0.0020523052662611008,
+      "learning_rate": 8.223295759527645e-06,
+      "loss": 0.0001,
+      "step": 2690
+    },
+    {
+      "epoch": 38.00386473429952,
+      "grad_norm": 0.0025786529295146465,
+      "learning_rate": 8.212560386473431e-06,
+      "loss": 0.188,
+      "step": 2700
+    },
+    {
+      "epoch": 38.00483091787439,
+      "grad_norm": 0.005663599353283644,
+      "learning_rate": 8.201825013419216e-06,
+      "loss": 0.2146,
+      "step": 2710
+    },
+    {
+      "epoch": 38.005797101449275,
+      "grad_norm": 0.028957316651940346,
+      "learning_rate": 8.191089640365003e-06,
+      "loss": 0.0002,
+      "step": 2720
+    },
+    {
+      "epoch": 38.00676328502416,
+      "grad_norm": 0.007074129767715931,
+      "learning_rate": 8.180354267310789e-06,
+      "loss": 0.2403,
+      "step": 2730
+    },
+    {
+      "epoch": 38.00676328502416,
+      "eval_accuracy": 0.7435897435897436,
+      "eval_loss": 2.099257230758667,
+      "eval_runtime": 15.1057,
+      "eval_samples_per_second": 5.164,
+      "eval_steps_per_second": 1.324,
+      "step": 2730
+    },
+    {
+      "epoch": 39.00096618357488,
+      "grad_norm": 0.0016518068732693791,
+      "learning_rate": 8.169618894256575e-06,
+      "loss": 0.0003,
+      "step": 2740
+    },
+    {
+      "epoch": 39.001932367149756,
+      "grad_norm": 0.008634793572127819,
+      "learning_rate": 8.158883521202362e-06,
+      "loss": 0.0034,
+      "step": 2750
+    },
+    {
+      "epoch": 39.00289855072464,
+      "grad_norm": 0.004315139725804329,
+      "learning_rate": 8.148148148148148e-06,
+      "loss": 0.1378,
+      "step": 2760
+    },
+    {
+      "epoch": 39.00386473429952,
+      "grad_norm": 0.015739772468805313,
+      "learning_rate": 8.137412775093935e-06,
+      "loss": 0.1428,
+      "step": 2770
+    },
+    {
+      "epoch": 39.00483091787439,
+      "grad_norm": 0.21879592537879944,
+      "learning_rate": 8.126677402039721e-06,
+      "loss": 0.0106,
+      "step": 2780
+    },
+    {
+      "epoch": 39.005797101449275,
+      "grad_norm": 1.373992919921875,
+      "learning_rate": 8.115942028985508e-06,
+      "loss": 0.0003,
+      "step": 2790
+    },
+    {
+      "epoch": 39.00676328502416,
+      "grad_norm": 0.008464109152555466,
+      "learning_rate": 8.105206655931294e-06,
+      "loss": 0.0001,
+      "step": 2800
+    },
+    {
+      "epoch": 39.00676328502416,
+      "eval_accuracy": 0.7435897435897436,
+      "eval_loss": 1.8645843267440796,
+      "eval_runtime": 15.7999,
+      "eval_samples_per_second": 4.937,
+      "eval_steps_per_second": 1.266,
+      "step": 2800
+    },
+    {
+      "epoch": 40.00096618357488,
+      "grad_norm": 0.01016903854906559,
+      "learning_rate": 8.09447128287708e-06,
+      "loss": 0.0001,
+      "step": 2810
+    },
+    {
+      "epoch": 40.001932367149756,
+      "grad_norm": 0.006220806855708361,
+      "learning_rate": 8.083735909822867e-06,
+      "loss": 0.0011,
+      "step": 2820
+    },
+    {
+      "epoch": 40.00289855072464,
+      "grad_norm": 0.36463844776153564,
+      "learning_rate": 8.073000536768653e-06,
+      "loss": 0.002,
+      "step": 2830
+    },
+    {
+      "epoch": 40.00386473429952,
+      "grad_norm": 0.0067412094213068485,
+      "learning_rate": 8.06226516371444e-06,
+      "loss": 0.0001,
+      "step": 2840
+    },
+    {
+      "epoch": 40.00483091787439,
+      "grad_norm": 0.002454320201650262,
+      "learning_rate": 8.051529790660226e-06,
+      "loss": 0.0001,
+      "step": 2850
+    },
+    {
+      "epoch": 40.005797101449275,
+      "grad_norm": 0.019683407619595528,
+      "learning_rate": 8.040794417606013e-06,
+      "loss": 0.0005,
+      "step": 2860
+    },
+    {
+      "epoch": 40.00676328502416,
+      "grad_norm": 0.012780338525772095,
+      "learning_rate": 8.0300590445518e-06,
+      "loss": 0.0001,
+      "step": 2870
+    },
+    {
+      "epoch": 40.00676328502416,
+      "eval_accuracy": 0.7435897435897436,
+      "eval_loss": 1.6642485857009888,
+      "eval_runtime": 17.6893,
+      "eval_samples_per_second": 4.409,
+      "eval_steps_per_second": 1.131,
+      "step": 2870
+    },
+    {
+      "epoch": 41.00096618357488,
+      "grad_norm": 0.0050564175471663475,
+      "learning_rate": 8.019323671497586e-06,
+      "loss": 0.3199,
+      "step": 2880
+    },
+    {
+      "epoch": 41.001932367149756,
+      "grad_norm": 0.09579924494028091,
+      "learning_rate": 8.008588298443372e-06,
+      "loss": 0.0002,
+      "step": 2890
+    },
+    {
+      "epoch": 41.00289855072464,
+      "grad_norm": 0.008467406034469604,
+      "learning_rate": 7.997852925389159e-06,
+      "loss": 0.0002,
+      "step": 2900
+    },
+    {
+      "epoch": 41.00386473429952,
+      "grad_norm": 0.004457333590835333,
+      "learning_rate": 7.987117552334945e-06,
+      "loss": 0.0006,
+      "step": 2910
+    },
+    {
+      "epoch": 41.00483091787439,
+      "grad_norm": 0.0018141858745366335,
+      "learning_rate": 7.97638217928073e-06,
+      "loss": 0.0001,
+      "step": 2920
+    },
+    {
+      "epoch": 41.005797101449275,
+      "grad_norm": 0.021920841187238693,
+      "learning_rate": 7.965646806226516e-06,
+      "loss": 0.0147,
+      "step": 2930
+    },
+    {
+      "epoch": 41.00676328502416,
+      "grad_norm": 0.006291843950748444,
+      "learning_rate": 7.954911433172303e-06,
+      "loss": 0.2108,
+      "step": 2940
+    },
+    {
+      "epoch": 41.00676328502416,
+      "eval_accuracy": 0.7948717948717948,
+      "eval_loss": 1.4565685987472534,
+      "eval_runtime": 23.0447,
+      "eval_samples_per_second": 3.385,
+      "eval_steps_per_second": 0.868,
+      "step": 2940
+    },
+    {
+      "epoch": 42.00096618357488,
+      "grad_norm": 0.0030977341812103987,
+      "learning_rate": 7.94417606011809e-06,
+      "loss": 0.1498,
+      "step": 2950
+    },
+    {
+      "epoch": 42.001932367149756,
+      "grad_norm": 0.002562603447586298,
+      "learning_rate": 7.933440687063876e-06,
+      "loss": 0.0006,
+      "step": 2960
+    },
+    {
+      "epoch": 42.00289855072464,
+      "grad_norm": 48.39828872680664,
+      "learning_rate": 7.922705314009662e-06,
+      "loss": 0.2303,
+      "step": 2970
+    },
+    {
+      "epoch": 42.00386473429952,
+      "grad_norm": 0.058131780475378036,
+      "learning_rate": 7.911969940955449e-06,
+      "loss": 0.0002,
+      "step": 2980
+    },
+    {
+      "epoch": 42.00483091787439,
+      "grad_norm": 0.656736433506012,
+      "learning_rate": 7.901234567901235e-06,
+      "loss": 0.083,
+      "step": 2990
+    },
+    {
+      "epoch": 42.005797101449275,
+      "grad_norm": 0.0014849025756120682,
+      "learning_rate": 7.890499194847021e-06,
+      "loss": 0.0963,
+      "step": 3000
+    },
+    {
+      "epoch": 42.00676328502416,
+      "grad_norm": 0.0057001574896276,
+      "learning_rate": 7.879763821792808e-06,
+      "loss": 0.1085,
+      "step": 3010
+    },
+    {
+      "epoch": 42.00676328502416,
+      "eval_accuracy": 0.782051282051282,
+      "eval_loss": 1.291454792022705,
+      "eval_runtime": 20.8555,
+      "eval_samples_per_second": 3.74,
+      "eval_steps_per_second": 0.959,
+      "step": 3010
+    },
+    {
+      "epoch": 43.00096618357488,
+      "grad_norm": 0.0038009320851415396,
+      "learning_rate": 7.869028448738594e-06,
+      "loss": 0.0035,
+      "step": 3020
+    },
+    {
+      "epoch": 43.001932367149756,
+      "grad_norm": 0.049530353397130966,
+      "learning_rate": 7.85829307568438e-06,
+      "loss": 0.0001,
+      "step": 3030
+    },
+    {
+      "epoch": 43.00289855072464,
+      "grad_norm": 0.007268795743584633,
+      "learning_rate": 7.847557702630167e-06,
+      "loss": 0.0615,
+      "step": 3040
+    },
+    {
+      "epoch": 43.00386473429952,
+      "grad_norm": 0.022801999002695084,
+      "learning_rate": 7.836822329575954e-06,
+      "loss": 0.1555,
+      "step": 3050
+    },
+    {
+      "epoch": 43.00483091787439,
+      "grad_norm": 0.004046064801514149,
+      "learning_rate": 7.82608695652174e-06,
+      "loss": 0.0001,
+      "step": 3060
+    },
+    {
+      "epoch": 43.005797101449275,
+      "grad_norm": 0.006020054221153259,
+      "learning_rate": 7.815351583467527e-06,
+      "loss": 0.0006,
+      "step": 3070
+    },
+    {
+      "epoch": 43.00676328502416,
+      "grad_norm": 0.0009516979334875941,
+      "learning_rate": 7.804616210413313e-06,
+      "loss": 0.0019,
+      "step": 3080
+    },
+    {
+      "epoch": 43.00676328502416,
+      "eval_accuracy": 0.8333333333333334,
+      "eval_loss": 1.2690619230270386,
+      "eval_runtime": 21.1783,
+      "eval_samples_per_second": 3.683,
+      "eval_steps_per_second": 0.944,
+      "step": 3080
+    },
+    {
+      "epoch": 44.00096618357488,
+      "grad_norm": 0.004416549112647772,
+      "learning_rate": 7.7938808373591e-06,
+      "loss": 0.0008,
+      "step": 3090
+    },
+    {
+      "epoch": 44.001932367149756,
+      "grad_norm": 0.002933249343186617,
+      "learning_rate": 7.783145464304886e-06,
+      "loss": 0.0001,
+      "step": 3100
+    },
+    {
+      "epoch": 44.00289855072464,
+      "grad_norm": 0.005305302329361439,
+      "learning_rate": 7.772410091250672e-06,
+      "loss": 0.0001,
+      "step": 3110
+    },
+    {
+      "epoch": 44.00386473429952,
+      "grad_norm": 0.15504007041454315,
+      "learning_rate": 7.761674718196459e-06,
+      "loss": 0.0014,
+      "step": 3120
+    },
+    {
+      "epoch": 44.00483091787439,
+      "grad_norm": 0.00801057554781437,
+      "learning_rate": 7.750939345142244e-06,
+      "loss": 0.1411,
+      "step": 3130
+    },
+    {
+      "epoch": 44.005797101449275,
+      "grad_norm": 0.013626204803586006,
+      "learning_rate": 7.74020397208803e-06,
+      "loss": 0.0293,
+      "step": 3140
+    },
+    {
+      "epoch": 44.00676328502416,
+      "grad_norm": 0.0010916618630290031,
+      "learning_rate": 7.729468599033817e-06,
+      "loss": 0.0072,
+      "step": 3150
+    },
+    {
+      "epoch": 44.00676328502416,
+      "eval_accuracy": 0.717948717948718,
+      "eval_loss": 2.270263195037842,
+      "eval_runtime": 22.5558,
+      "eval_samples_per_second": 3.458,
+      "eval_steps_per_second": 0.887,
+      "step": 3150
+    },
+    {
+      "epoch": 45.00096618357488,
+      "grad_norm": 357.7193908691406,
+      "learning_rate": 7.718733225979603e-06,
+      "loss": 0.4039,
+      "step": 3160
+    },
+    {
+      "epoch": 45.001932367149756,
+      "grad_norm": 94.0329360961914,
+      "learning_rate": 7.70799785292539e-06,
+      "loss": 0.2366,
+      "step": 3170
+    },
+    {
+      "epoch": 45.00289855072464,
+      "grad_norm": 0.08033226430416107,
+      "learning_rate": 7.697262479871176e-06,
+      "loss": 0.0739,
+      "step": 3180
+    },
+    {
+      "epoch": 45.00386473429952,
+      "grad_norm": 0.017788292840123177,
+      "learning_rate": 7.686527106816962e-06,
+      "loss": 0.1424,
+      "step": 3190
+    },
+    {
+      "epoch": 45.00483091787439,
+      "grad_norm": 0.009477581828832626,
+      "learning_rate": 7.675791733762749e-06,
+      "loss": 0.0002,
+      "step": 3200
+    },
+    {
+      "epoch": 45.005797101449275,
+      "grad_norm": 0.007442084141075611,
+      "learning_rate": 7.665056360708535e-06,
+      "loss": 0.0002,
+      "step": 3210
+    },
+    {
+      "epoch": 45.00676328502416,
+      "grad_norm": 0.031041273847222328,
+      "learning_rate": 7.654320987654322e-06,
+      "loss": 0.0002,
+      "step": 3220
+    },
+    {
+      "epoch": 45.00676328502416,
+      "eval_accuracy": 0.7564102564102564,
+      "eval_loss": 1.515126347541809,
+      "eval_runtime": 23.2938,
+      "eval_samples_per_second": 3.349,
+      "eval_steps_per_second": 0.859,
+      "step": 3220
+    },
+    {
+      "epoch": 46.00096618357488,
+      "grad_norm": 0.0019831734243780375,
+      "learning_rate": 7.643585614600108e-06,
+      "loss": 0.0004,
+      "step": 3230
+    },
+    {
+      "epoch": 46.001932367149756,
+      "grad_norm": 0.006759632378816605,
+      "learning_rate": 7.632850241545895e-06,
+      "loss": 0.0002,
+      "step": 3240
+    },
+    {
+      "epoch": 46.00289855072464,
+      "grad_norm": 84.79361724853516,
+      "learning_rate": 7.622114868491681e-06,
+      "loss": 0.3816,
+      "step": 3250
+    },
+    {
+      "epoch": 46.00386473429952,
+      "grad_norm": 0.003224059008061886,
+      "learning_rate": 7.6113794954374675e-06,
+      "loss": 0.3455,
+      "step": 3260
+    },
+    {
+      "epoch": 46.00483091787439,
+      "grad_norm": 0.007698831148445606,
+      "learning_rate": 7.600644122383254e-06,
+      "loss": 0.6513,
+      "step": 3270
+    },
+    {
+      "epoch": 46.005797101449275,
+      "grad_norm": 0.10313475131988525,
+      "learning_rate": 7.58990874932904e-06,
+      "loss": 0.2645,
+      "step": 3280
+    },
+    {
+      "epoch": 46.00676328502416,
+      "grad_norm": 0.015344126150012016,
+      "learning_rate": 7.579173376274827e-06,
+      "loss": 0.4057,
+      "step": 3290
+    },
+    {
+      "epoch": 46.00676328502416,
+      "eval_accuracy": 0.7564102564102564,
+      "eval_loss": 1.417032241821289,
+      "eval_runtime": 20.3874,
+      "eval_samples_per_second": 3.826,
+      "eval_steps_per_second": 0.981,
+      "step": 3290
+    },
+    {
+      "epoch": 47.00096618357488,
+      "grad_norm": 0.01232170406728983,
+      "learning_rate": 7.568438003220613e-06,
+      "loss": 0.0008,
+      "step": 3300
+    },
+    {
+      "epoch": 47.001932367149756,
+      "grad_norm": 0.0009604791994206607,
+      "learning_rate": 7.557702630166399e-06,
+      "loss": 0.1195,
+      "step": 3310
+    },
+    {
+      "epoch": 47.00289855072464,
+      "grad_norm": 0.06583663076162338,
+      "learning_rate": 7.546967257112185e-06,
+      "loss": 0.4292,
+      "step": 3320
+    },
+    {
+      "epoch": 47.00386473429952,
+      "grad_norm": 1.0096828937530518,
+      "learning_rate": 7.536231884057972e-06,
+      "loss": 0.0005,
+      "step": 3330
+    },
+    {
+      "epoch": 47.00483091787439,
+      "grad_norm": 18.510780334472656,
+      "learning_rate": 7.525496511003758e-06,
+      "loss": 0.0017,
+      "step": 3340
+    },
+    {
+      "epoch": 47.005797101449275,
+      "grad_norm": 0.027283035218715668,
+      "learning_rate": 7.514761137949545e-06,
+      "loss": 0.0019,
+      "step": 3350
+    },
+    {
+      "epoch": 47.00676328502416,
+      "grad_norm": 0.24246840178966522,
+      "learning_rate": 7.504025764895331e-06,
+      "loss": 0.0862,
+      "step": 3360
+    },
+    {
+      "epoch": 47.00676328502416,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 2.1071906089782715,
+      "eval_runtime": 17.1389,
+      "eval_samples_per_second": 4.551,
+      "eval_steps_per_second": 1.167,
+      "step": 3360
+    },
+    {
+      "epoch": 48.00096618357488,
+      "grad_norm": 0.007916704751551151,
+      "learning_rate": 7.493290391841117e-06,
+      "loss": 0.0023,
+      "step": 3370
+    },
+    {
+      "epoch": 48.001932367149756,
+      "grad_norm": 58.70967483520508,
+      "learning_rate": 7.482555018786903e-06,
+      "loss": 0.0026,
+      "step": 3380
+    },
+    {
+      "epoch": 48.00289855072464,
+      "grad_norm": 0.005796114448457956,
+      "learning_rate": 7.47181964573269e-06,
+      "loss": 0.0003,
+      "step": 3390
+    },
+    {
+      "epoch": 48.00386473429952,
+      "grad_norm": 0.0461343489587307,
+      "learning_rate": 7.461084272678476e-06,
+      "loss": 0.0002,
+      "step": 3400
+    },
+    {
+      "epoch": 48.00483091787439,
+      "grad_norm": 0.0035491653252393007,
+      "learning_rate": 7.4503488996242625e-06,
+      "loss": 0.0092,
+      "step": 3410
+    },
+    {
+      "epoch": 48.005797101449275,
+      "grad_norm": 274.37054443359375,
+      "learning_rate": 7.439613526570049e-06,
+      "loss": 0.0569,
+      "step": 3420
+    },
+    {
+      "epoch": 48.00676328502416,
+      "grad_norm": 0.0009124857024289668,
+      "learning_rate": 7.428878153515835e-06,
+      "loss": 0.0478,
+      "step": 3430
+    },
+    {
+      "epoch": 48.00676328502416,
+      "eval_accuracy": 0.7948717948717948,
+      "eval_loss": 1.6825991868972778,
+      "eval_runtime": 21.1396,
+      "eval_samples_per_second": 3.69,
+      "eval_steps_per_second": 0.946,
+      "step": 3430
+    },
+    {
+      "epoch": 49.00096618357488,
+      "grad_norm": 0.020062750205397606,
+      "learning_rate": 7.418142780461621e-06,
+      "loss": 0.775,
+      "step": 3440
+    },
+    {
+      "epoch": 49.001932367149756,
+      "grad_norm": 0.017441576346755028,
+      "learning_rate": 7.4074074074074075e-06,
+      "loss": 0.1064,
+      "step": 3450
+    },
+    {
+      "epoch": 49.00289855072464,
+      "grad_norm": 0.022105790674686432,
+      "learning_rate": 7.396672034353194e-06,
+      "loss": 0.0002,
+      "step": 3460
+    },
+    {
+      "epoch": 49.00386473429952,
+      "grad_norm": 0.006332984659820795,
+      "learning_rate": 7.38593666129898e-06,
+      "loss": 0.003,
+      "step": 3470
+    },
+    {
+      "epoch": 49.00483091787439,
+      "grad_norm": 0.023586558178067207,
+      "learning_rate": 7.375201288244767e-06,
+      "loss": 0.0002,
+      "step": 3480
+    },
+    {
+      "epoch": 49.005797101449275,
+      "grad_norm": 0.002561570843681693,
+      "learning_rate": 7.364465915190554e-06,
+      "loss": 0.287,
+      "step": 3490
+    },
+    {
+      "epoch": 49.00676328502416,
+      "grad_norm": 0.02558230608701706,
+      "learning_rate": 7.353730542136341e-06,
+      "loss": 0.001,
+      "step": 3500
+    },
+    {
+      "epoch": 49.00676328502416,
+      "eval_accuracy": 0.7564102564102564,
+      "eval_loss": 1.4348629713058472,
+      "eval_runtime": 17.0482,
+      "eval_samples_per_second": 4.575,
+      "eval_steps_per_second": 1.173,
+      "step": 3500
+    },
+    {
+      "epoch": 50.00096618357488,
+      "grad_norm": 0.006134858354926109,
+      "learning_rate": 7.342995169082127e-06,
+      "loss": 0.0001,
+      "step": 3510
+    },
+    {
+      "epoch": 50.001932367149756,
+      "grad_norm": 0.0029954123310744762,
+      "learning_rate": 7.332259796027913e-06,
+      "loss": 0.0001,
+      "step": 3520
+    },
+    {
+      "epoch": 50.00289855072464,
+      "grad_norm": 0.0036094398237764835,
+      "learning_rate": 7.321524422973699e-06,
+      "loss": 0.157,
+      "step": 3530
+    },
+    {
+      "epoch": 50.00386473429952,
+      "grad_norm": 0.0021507777273654938,
+      "learning_rate": 7.3107890499194855e-06,
+      "loss": 0.0001,
+      "step": 3540
+    },
+    {
+      "epoch": 50.00483091787439,
+      "grad_norm": 0.003133822698146105,
+      "learning_rate": 7.300053676865272e-06,
+      "loss": 0.0001,
+      "step": 3550
+    },
+    {
+      "epoch": 50.005797101449275,
+      "grad_norm": 0.0019445763900876045,
+      "learning_rate": 7.2893183038110584e-06,
+      "loss": 0.0001,
+      "step": 3560
+    },
+    {
+      "epoch": 50.00676328502416,
+      "grad_norm": 0.013452505692839622,
+      "learning_rate": 7.278582930756845e-06,
+      "loss": 0.0001,
+      "step": 3570
+    },
+    {
+      "epoch": 50.00676328502416,
+      "eval_accuracy": 0.7948717948717948,
+      "eval_loss": 1.442294955253601,
+      "eval_runtime": 17.0525,
+      "eval_samples_per_second": 4.574,
+      "eval_steps_per_second": 1.173,
+      "step": 3570
+    },
+    {
+      "epoch": 51.00096618357488,
+      "grad_norm": 0.004946423228830099,
+      "learning_rate": 7.2678475577026305e-06,
+      "loss": 0.0001,
+      "step": 3580
+    },
+    {
+      "epoch": 51.001932367149756,
+      "grad_norm": 0.003967686556279659,
+      "learning_rate": 7.257112184648417e-06,
+      "loss": 0.1159,
+      "step": 3590
+    },
+    {
+      "epoch": 51.00289855072464,
+      "grad_norm": 0.0038182190619409084,
+      "learning_rate": 7.246376811594203e-06,
+      "loss": 0.0001,
+      "step": 3600
+    },
+    {
+      "epoch": 51.00386473429952,
+      "grad_norm": 0.026833100244402885,
+      "learning_rate": 7.23564143853999e-06,
+      "loss": 0.0002,
+      "step": 3610
+    },
+    {
+      "epoch": 51.00483091787439,
+      "grad_norm": 0.06868147104978561,
+      "learning_rate": 7.224906065485776e-06,
+      "loss": 0.0056,
+      "step": 3620
+    },
+    {
+      "epoch": 51.005797101449275,
+      "grad_norm": 0.0036590364761650562,
+      "learning_rate": 7.214170692431563e-06,
+      "loss": 0.0001,
+      "step": 3630
+    },
+    {
+      "epoch": 51.00676328502416,
+      "grad_norm": 0.007523122243583202,
+      "learning_rate": 7.203435319377348e-06,
+      "loss": 0.0056,
+      "step": 3640
+    },
+    {
+      "epoch": 51.00676328502416,
+      "eval_accuracy": 0.7051282051282052,
+      "eval_loss": 2.2833075523376465,
+      "eval_runtime": 17.3876,
+      "eval_samples_per_second": 4.486,
+      "eval_steps_per_second": 1.15,
+      "step": 3640
+    },
+    {
+      "epoch": 52.00096618357488,
+      "grad_norm": 0.012034900486469269,
+      "learning_rate": 7.192699946323135e-06,
+      "loss": 0.4707,
+      "step": 3650
+    },
+    {
+      "epoch": 52.001932367149756,
+      "grad_norm": 0.011260480619966984,
+      "learning_rate": 7.181964573268921e-06,
+      "loss": 0.421,
+      "step": 3660
+    },
+    {
+      "epoch": 52.00289855072464,
+      "grad_norm": 0.043340399861335754,
+      "learning_rate": 7.171229200214708e-06,
+      "loss": 0.1122,
+      "step": 3670
+    },
+    {
+      "epoch": 52.00386473429952,
+      "grad_norm": 3.0821592807769775,
+      "learning_rate": 7.160493827160494e-06,
+      "loss": 0.0005,
+      "step": 3680
+    },
+    {
+      "epoch": 52.00483091787439,
+      "grad_norm": 0.01043855119496584,
+      "learning_rate": 7.149758454106281e-06,
+      "loss": 0.067,
+      "step": 3690
+    },
+    {
+      "epoch": 52.005797101449275,
+      "grad_norm": 0.0030516660772264004,
+      "learning_rate": 7.139023081052067e-06,
+      "loss": 0.0305,
+      "step": 3700
+    },
+    {
+      "epoch": 52.00676328502416,
+      "grad_norm": 0.0027135934215039015,
+      "learning_rate": 7.128287707997853e-06,
+      "loss": 0.0004,
+      "step": 3710
+    },
+    {
+      "epoch": 52.00676328502416,
+      "eval_accuracy": 0.8461538461538461,
+      "eval_loss": 1.0547696352005005,
+      "eval_runtime": 16.8224,
+      "eval_samples_per_second": 4.637,
+      "eval_steps_per_second": 1.189,
+      "step": 3710
+    },
+    {
+      "epoch": 53.00096618357488,
+      "grad_norm": 0.0029095993377268314,
+      "learning_rate": 7.117552334943641e-06,
+      "loss": 0.0001,
+      "step": 3720
+    },
+    {
+      "epoch": 53.001932367149756,
+      "grad_norm": 0.08320170640945435,
+      "learning_rate": 7.106816961889426e-06,
+      "loss": 0.0003,
+      "step": 3730
+    },
+    {
+      "epoch": 53.00289855072464,
+      "grad_norm": 0.004328227136284113,
+      "learning_rate": 7.096081588835213e-06,
+      "loss": 0.0002,
+      "step": 3740
+    },
+    {
+      "epoch": 53.00386473429952,
+      "grad_norm": 0.005063340999186039,
+      "learning_rate": 7.085346215780999e-06,
+      "loss": 0.0171,
+      "step": 3750
+    },
+    {
+      "epoch": 53.00483091787439,
+      "grad_norm": 0.005174571648240089,
+      "learning_rate": 7.074610842726786e-06,
+      "loss": 0.0001,
+      "step": 3760
+    },
+    {
+      "epoch": 53.005797101449275,
+      "grad_norm": 0.0030827228911221027,
+      "learning_rate": 7.063875469672572e-06,
+      "loss": 0.0001,
+      "step": 3770
+    },
+    {
+      "epoch": 53.00676328502416,
+      "grad_norm": 0.01317480206489563,
+      "learning_rate": 7.053140096618359e-06,
+      "loss": 0.1768,
+      "step": 3780
+    },
+    {
+      "epoch": 53.00676328502416,
+      "eval_accuracy": 0.8076923076923077,
+      "eval_loss": 1.2974127531051636,
+      "eval_runtime": 20.4623,
+      "eval_samples_per_second": 3.812,
+      "eval_steps_per_second": 0.977,
+      "step": 3780
+    },
+    {
+      "epoch": 54.00096618357488,
+      "grad_norm": 0.00562130706384778,
+      "learning_rate": 7.042404723564144e-06,
+      "loss": 0.0001,
+      "step": 3790
+    },
+    {
+      "epoch": 54.001932367149756,
+      "grad_norm": 0.009645653888583183,
+      "learning_rate": 7.031669350509931e-06,
+      "loss": 0.0001,
+      "step": 3800
+    },
+    {
+      "epoch": 54.00289855072464,
+      "grad_norm": 0.006110002752393484,
+      "learning_rate": 7.020933977455717e-06,
+      "loss": 0.0001,
+      "step": 3810
+    },
+    {
+      "epoch": 54.00386473429952,
+      "grad_norm": 0.002363593550398946,
+      "learning_rate": 7.010198604401504e-06,
+      "loss": 0.0174,
+      "step": 3820
+    },
+    {
+      "epoch": 54.00483091787439,
+      "grad_norm": 0.0013595304917544127,
+      "learning_rate": 6.99946323134729e-06,
+      "loss": 0.017,
+      "step": 3830
+    },
+    {
+      "epoch": 54.005797101449275,
+      "grad_norm": 0.003808892797678709,
+      "learning_rate": 6.9887278582930765e-06,
+      "loss": 0.0001,
+      "step": 3840
+    },
+    {
+      "epoch": 54.00676328502416,
+      "grad_norm": 0.006192604545503855,
+      "learning_rate": 6.977992485238862e-06,
+      "loss": 0.0001,
+      "step": 3850
+    },
+    {
+      "epoch": 54.00676328502416,
+      "eval_accuracy": 0.8333333333333334,
+      "eval_loss": 1.2796709537506104,
+      "eval_runtime": 14.9163,
+      "eval_samples_per_second": 5.229,
+      "eval_steps_per_second": 1.341,
+      "step": 3850
+    },
+    {
+      "epoch": 55.00096618357488,
+      "grad_norm": 0.0006895597325637937,
+      "learning_rate": 6.9672571121846486e-06,
+      "loss": 0.0001,
+      "step": 3860
+    },
+    {
+      "epoch": 55.001932367149756,
+      "grad_norm": 0.0019923022482544184,
+      "learning_rate": 6.956521739130435e-06,
+      "loss": 0.0001,
+      "step": 3870
+    },
+    {
+      "epoch": 55.00289855072464,
+      "grad_norm": 0.006965520326048136,
+      "learning_rate": 6.9457863660762215e-06,
+      "loss": 0.0002,
+      "step": 3880
+    },
+    {
+      "epoch": 55.00386473429952,
+      "grad_norm": 0.0015802200650796294,
+      "learning_rate": 6.935050993022008e-06,
+      "loss": 0.1119,
+      "step": 3890
+    },
+    {
+      "epoch": 55.00483091787439,
+      "grad_norm": 0.006184786558151245,
+      "learning_rate": 6.924315619967794e-06,
+      "loss": 0.0001,
+      "step": 3900
+    },
+    {
+      "epoch": 55.005797101449275,
+      "grad_norm": 0.01038430817425251,
+      "learning_rate": 6.913580246913581e-06,
+      "loss": 0.1995,
+      "step": 3910
+    },
+    {
+      "epoch": 55.00676328502416,
+      "grad_norm": 0.003573243273422122,
+      "learning_rate": 6.9028448738593664e-06,
+      "loss": 0.1027,
+      "step": 3920
+    },
+    {
+      "epoch": 55.00676328502416,
+      "eval_accuracy": 0.8076923076923077,
+      "eval_loss": 1.515874981880188,
+      "eval_runtime": 17.0594,
+      "eval_samples_per_second": 4.572,
+      "eval_steps_per_second": 1.172,
+      "step": 3920
+    },
+    {
+      "epoch": 56.00096618357488,
+      "grad_norm": 0.008314134553074837,
+      "learning_rate": 6.892109500805153e-06,
+      "loss": 0.0001,
+      "step": 3930
+    },
+    {
+      "epoch": 56.001932367149756,
+      "grad_norm": 0.03252334147691727,
+      "learning_rate": 6.881374127750939e-06,
+      "loss": 0.0002,
+      "step": 3940
+    },
+    {
+      "epoch": 56.00289855072464,
+      "grad_norm": 0.0016289795748889446,
+      "learning_rate": 6.870638754696727e-06,
+      "loss": 0.0158,
+      "step": 3950
+    },
+    {
+      "epoch": 56.00386473429952,
+      "grad_norm": 0.0037534793373197317,
+      "learning_rate": 6.859903381642513e-06,
+      "loss": 0.0023,
+      "step": 3960
+    },
+    {
+      "epoch": 56.00483091787439,
+      "grad_norm": 0.025467460975050926,
+      "learning_rate": 6.8491680085882995e-06,
+      "loss": 0.178,
+      "step": 3970
+    },
+    {
+      "epoch": 56.005797101449275,
+      "grad_norm": 44.050533294677734,
+      "learning_rate": 6.838432635534086e-06,
+      "loss": 0.3517,
+      "step": 3980
+    },
+    {
+      "epoch": 56.00676328502416,
+      "grad_norm": 0.0007292564259842038,
+      "learning_rate": 6.8276972624798724e-06,
+      "loss": 0.1638,
+      "step": 3990
+    },
+    {
+      "epoch": 56.00676328502416,
+      "eval_accuracy": 0.7564102564102564,
+      "eval_loss": 1.9402605295181274,
+      "eval_runtime": 16.8977,
+      "eval_samples_per_second": 4.616,
+      "eval_steps_per_second": 1.184,
+      "step": 3990
+    },
+    {
+      "epoch": 57.00096618357488,
+      "grad_norm": 0.0029645762406289577,
+      "learning_rate": 6.816961889425658e-06,
+      "loss": 0.0016,
+      "step": 4000
+    },
+    {
+      "epoch": 57.001932367149756,
+      "grad_norm": 0.002086229156702757,
+      "learning_rate": 6.8062265163714445e-06,
+      "loss": 0.2358,
+      "step": 4010
+    },
+    {
+      "epoch": 57.00289855072464,
+      "grad_norm": 0.004452128428965807,
+      "learning_rate": 6.795491143317231e-06,
+      "loss": 0.0001,
+      "step": 4020
+    },
+    {
+      "epoch": 57.00386473429952,
+      "grad_norm": 0.004634437616914511,
+      "learning_rate": 6.784755770263017e-06,
+      "loss": 0.0004,
+      "step": 4030
+    },
+    {
+      "epoch": 57.00483091787439,
+      "grad_norm": 0.01934254914522171,
+      "learning_rate": 6.774020397208804e-06,
+      "loss": 0.0001,
+      "step": 4040
+    },
+    {
+      "epoch": 57.005797101449275,
+      "grad_norm": 0.023365547880530357,
+      "learning_rate": 6.76328502415459e-06,
+      "loss": 0.3372,
+      "step": 4050
+    },
+    {
+      "epoch": 57.00676328502416,
+      "grad_norm": 0.002822543028742075,
+      "learning_rate": 6.752549651100376e-06,
+      "loss": 0.0001,
+      "step": 4060
+    },
+    {
+      "epoch": 57.00676328502416,
+      "eval_accuracy": 0.8076923076923077,
+      "eval_loss": 1.5075114965438843,
+      "eval_runtime": 20.119,
+      "eval_samples_per_second": 3.877,
+      "eval_steps_per_second": 0.994,
+      "step": 4060
+    },
+    {
+      "epoch": 58.00096618357488,
+      "grad_norm": 0.07366270571947098,
+      "learning_rate": 6.741814278046162e-06,
+      "loss": 0.0001,
+      "step": 4070
+    },
+    {
+      "epoch": 58.001932367149756,
+      "grad_norm": 0.0015621267957612872,
+      "learning_rate": 6.731078904991949e-06,
+      "loss": 0.0001,
+      "step": 4080
+    },
+    {
+      "epoch": 58.00289855072464,
+      "grad_norm": 0.0036368737928569317,
+      "learning_rate": 6.720343531937735e-06,
+      "loss": 0.0002,
+      "step": 4090
+    },
+    {
+      "epoch": 58.00386473429952,
+      "grad_norm": 0.037495680153369904,
+      "learning_rate": 6.709608158883522e-06,
+      "loss": 0.0001,
+      "step": 4100
+    },
+    {
+      "epoch": 58.00483091787439,
+      "grad_norm": 0.0035316827706992626,
+      "learning_rate": 6.698872785829308e-06,
+      "loss": 0.0001,
+      "step": 4110
+    },
+    {
+      "epoch": 58.005797101449275,
+      "grad_norm": 0.0035044823307543993,
+      "learning_rate": 6.688137412775095e-06,
+      "loss": 0.2844,
+      "step": 4120
+    },
+    {
+      "epoch": 58.00676328502416,
+      "grad_norm": 0.00098858040291816,
+      "learning_rate": 6.67740203972088e-06,
+      "loss": 0.0003,
+      "step": 4130
+    },
+    {
+      "epoch": 58.00676328502416,
+      "eval_accuracy": 0.7435897435897436,
+      "eval_loss": 2.1290876865386963,
+      "eval_runtime": 20.9101,
+      "eval_samples_per_second": 3.73,
+      "eval_steps_per_second": 0.956,
+      "step": 4130
+    },
+    {
+      "epoch": 59.00096618357488,
+      "grad_norm": 0.02900398150086403,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 0.0014,
+      "step": 4140
+    },
+    {
+      "epoch": 59.001932367149756,
+      "grad_norm": 0.004477238282561302,
+      "learning_rate": 6.655931293612453e-06,
+      "loss": 0.0001,
+      "step": 4150
+    },
+    {
+      "epoch": 59.00289855072464,
+      "grad_norm": 0.6924675107002258,
+      "learning_rate": 6.6451959205582395e-06,
+      "loss": 0.0003,
+      "step": 4160
+    },
+    {
+      "epoch": 59.00386473429952,
+      "grad_norm": 0.008186038583517075,
+      "learning_rate": 6.634460547504026e-06,
+      "loss": 0.0008,
+      "step": 4170
+    },
+    {
+      "epoch": 59.00483091787439,
+      "grad_norm": 0.0031193087343126535,
+      "learning_rate": 6.623725174449813e-06,
+      "loss": 0.1875,
+      "step": 4180
+    },
+    {
+      "epoch": 59.005797101449275,
+      "grad_norm": 0.0018959605367854238,
+      "learning_rate": 6.6129898013956e-06,
+      "loss": 0.0001,
+      "step": 4190
+    },
+    {
+      "epoch": 59.00676328502416,
+      "grad_norm": 0.0005922632990404963,
+      "learning_rate": 6.602254428341386e-06,
+      "loss": 0.004,
+      "step": 4200
+    },
+    {
+      "epoch": 59.00676328502416,
+      "eval_accuracy": 0.7948717948717948,
+      "eval_loss": 1.6104003190994263,
+      "eval_runtime": 20.6604,
+      "eval_samples_per_second": 3.775,
+      "eval_steps_per_second": 0.968,
+      "step": 4200
+    },
+    {
+      "epoch": 60.00096618357488,
+      "grad_norm": 0.0038133368361741304,
+      "learning_rate": 6.591519055287172e-06,
+      "loss": 0.0001,
+      "step": 4210
+    },
+    {
+      "epoch": 60.001932367149756,
+      "grad_norm": 0.0029480988159775734,
+      "learning_rate": 6.580783682232958e-06,
+      "loss": 0.0001,
+      "step": 4220
+    },
+    {
+      "epoch": 60.00289855072464,
+      "grad_norm": 0.005030333995819092,
+      "learning_rate": 6.570048309178745e-06,
+      "loss": 0.0001,
+      "step": 4230
+    },
+    {
+      "epoch": 60.00386473429952,
+      "grad_norm": 0.0027628433890640736,
+      "learning_rate": 6.559312936124531e-06,
+      "loss": 0.0001,
+      "step": 4240
+    },
+    {
+      "epoch": 60.00483091787439,
+      "grad_norm": 0.002624318702146411,
+      "learning_rate": 6.548577563070318e-06,
+      "loss": 0.0001,
+      "step": 4250
+    },
+    {
+      "epoch": 60.005797101449275,
+      "grad_norm": 0.003871429478749633,
+      "learning_rate": 6.537842190016104e-06,
+      "loss": 0.0069,
+      "step": 4260
+    },
+    {
+      "epoch": 60.00676328502416,
+      "grad_norm": 0.00036783877294510603,
+      "learning_rate": 6.52710681696189e-06,
+      "loss": 0.0214,
+      "step": 4270
+    },
+    {
+      "epoch": 60.00676328502416,
+      "eval_accuracy": 0.7948717948717948,
+      "eval_loss": 1.7016513347625732,
+      "eval_runtime": 21.0149,
+      "eval_samples_per_second": 3.712,
+      "eval_steps_per_second": 0.952,
+      "step": 4270
+    },
+    {
+      "epoch": 61.00096618357488,
+      "grad_norm": 0.0021181628108024597,
+      "learning_rate": 6.516371443907676e-06,
+      "loss": 0.0001,
+      "step": 4280
+    },
+    {
+      "epoch": 61.001932367149756,
+      "grad_norm": 0.001137953600846231,
+      "learning_rate": 6.5056360708534626e-06,
+      "loss": 0.0001,
+      "step": 4290
+    },
+    {
+      "epoch": 61.00289855072464,
+      "grad_norm": 0.0026911040768027306,
+      "learning_rate": 6.494900697799249e-06,
+      "loss": 0.0,
+      "step": 4300
+    },
+    {
+      "epoch": 61.00386473429952,
+      "grad_norm": 0.0015700599178671837,
+      "learning_rate": 6.4841653247450355e-06,
+      "loss": 0.0001,
+      "step": 4310
+    },
+    {
+      "epoch": 61.00483091787439,
+      "grad_norm": 0.0005428745062090456,
+      "learning_rate": 6.473429951690822e-06,
+      "loss": 0.0,
+      "step": 4320
+    },
+    {
+      "epoch": 61.005797101449275,
+      "grad_norm": 0.0032843085937201977,
+      "learning_rate": 6.462694578636608e-06,
+      "loss": 0.0001,
+      "step": 4330
+    },
+    {
+      "epoch": 61.00676328502416,
+      "grad_norm": 0.00028793522506020963,
+      "learning_rate": 6.451959205582394e-06,
+      "loss": 0.0,
+      "step": 4340
+    },
+    {
+      "epoch": 61.00676328502416,
+      "eval_accuracy": 0.8205128205128205,
+      "eval_loss": 1.64847993850708,
+      "eval_runtime": 24.004,
+      "eval_samples_per_second": 3.249,
+      "eval_steps_per_second": 0.833,
+      "step": 4340
+    },
+    {
+      "epoch": 62.00096618357488,
+      "grad_norm": 0.0031514782458543777,
+      "learning_rate": 6.44122383252818e-06,
+      "loss": 0.0,
+      "step": 4350
+    },
+    {
+      "epoch": 62.001932367149756,
+      "grad_norm": 0.002257132437080145,
+      "learning_rate": 6.430488459473967e-06,
+      "loss": 0.0,
+      "step": 4360
+    },
+    {
+      "epoch": 62.00289855072464,
+      "grad_norm": 0.0012779063545167446,
+      "learning_rate": 6.419753086419753e-06,
+      "loss": 0.0001,
+      "step": 4370
+    },
+    {
+      "epoch": 62.00386473429952,
+      "grad_norm": 0.0008087829919531941,
+      "learning_rate": 6.40901771336554e-06,
+      "loss": 0.0,
+      "step": 4380
+    },
+    {
+      "epoch": 62.00483091787439,
+      "grad_norm": 0.004048364236950874,
+      "learning_rate": 6.398282340311326e-06,
+      "loss": 0.0,
+      "step": 4390
+    },
+    {
+      "epoch": 62.005797101449275,
+      "grad_norm": 0.0023077051155269146,
+      "learning_rate": 6.387546967257112e-06,
+      "loss": 0.0001,
+      "step": 4400
+    },
+    {
+      "epoch": 62.00676328502416,
+      "grad_norm": 0.0035665419418364763,
+      "learning_rate": 6.376811594202898e-06,
+      "loss": 0.0,
+      "step": 4410
+    },
+    {
+      "epoch": 62.00676328502416,
+      "eval_accuracy": 0.8076923076923077,
+      "eval_loss": 1.6667520999908447,
+      "eval_runtime": 24.1821,
+      "eval_samples_per_second": 3.226,
+      "eval_steps_per_second": 0.827,
+      "step": 4410
+    },
+    {
+      "epoch": 63.00096618357488,
+      "grad_norm": 0.0010491169523447752,
+      "learning_rate": 6.3660762211486856e-06,
+      "loss": 0.0,
+      "step": 4420
+    },
+    {
+      "epoch": 63.001932367149756,
+      "grad_norm": 0.00069583480944857,
+      "learning_rate": 6.355340848094472e-06,
+      "loss": 0.0225,
+      "step": 4430
+    },
+    {
+      "epoch": 63.00289855072464,
+      "grad_norm": 0.016136689111590385,
+      "learning_rate": 6.3446054750402585e-06,
+      "loss": 0.0185,
+      "step": 4440
+    },
+    {
+      "epoch": 63.00386473429952,
+      "grad_norm": 0.002296636812388897,
+      "learning_rate": 6.333870101986045e-06,
+      "loss": 0.0001,
+      "step": 4450
+    },
+    {
+      "epoch": 63.00483091787439,
+      "grad_norm": 0.0012935524573549628,
+      "learning_rate": 6.323134728931831e-06,
+      "loss": 0.0,
+      "step": 4460
+    },
+    {
+      "epoch": 63.005797101449275,
+      "grad_norm": 0.00099344237241894,
+      "learning_rate": 6.312399355877618e-06,
+      "loss": 0.0,
+      "step": 4470
+    },
+    {
+      "epoch": 63.00676328502416,
+      "grad_norm": 0.0032291836105287075,
+      "learning_rate": 6.301663982823404e-06,
+      "loss": 0.1604,
+      "step": 4480
+    },
+    {
+      "epoch": 63.00676328502416,
+      "eval_accuracy": 0.7948717948717948,
+      "eval_loss": 1.7437233924865723,
+      "eval_runtime": 20.6625,
+      "eval_samples_per_second": 3.775,
+      "eval_steps_per_second": 0.968,
+      "step": 4480
+    },
+    {
+      "epoch": 64.00096618357487,
+      "grad_norm": 0.006006534211337566,
+      "learning_rate": 6.29092860976919e-06,
+      "loss": 0.0001,
+      "step": 4490
+    },
+    {
+      "epoch": 64.00193236714976,
+      "grad_norm": 0.0023545522708445787,
+      "learning_rate": 6.280193236714976e-06,
+      "loss": 0.0002,
+      "step": 4500
+    },
+    {
+      "epoch": 64.00289855072464,
+      "grad_norm": 0.2263268679380417,
+      "learning_rate": 6.269457863660763e-06,
+      "loss": 0.1125,
+      "step": 4510
+    },
+    {
+      "epoch": 64.00386473429951,
+      "grad_norm": 0.0006757387891411781,
+      "learning_rate": 6.258722490606549e-06,
+      "loss": 0.0403,
+      "step": 4520
+    },
+    {
+      "epoch": 64.0048309178744,
+      "grad_norm": 0.001137339510023594,
+      "learning_rate": 6.247987117552336e-06,
+      "loss": 0.1675,
+      "step": 4530
+    },
+    {
+      "epoch": 64.00579710144928,
+      "grad_norm": 0.0034416404087096453,
+      "learning_rate": 6.237251744498122e-06,
+      "loss": 0.0001,
+      "step": 4540
+    },
+    {
+      "epoch": 64.00676328502415,
+      "grad_norm": 0.00035281089367344975,
+      "learning_rate": 6.226516371443908e-06,
+      "loss": 0.0002,
+      "step": 4550
+    },
+    {
+      "epoch": 64.00676328502415,
+      "eval_accuracy": 0.782051282051282,
+      "eval_loss": 1.676971197128296,
+      "eval_runtime": 21.1031,
+      "eval_samples_per_second": 3.696,
+      "eval_steps_per_second": 0.948,
+      "step": 4550
+    },
+    {
+      "epoch": 65.00096618357487,
+      "grad_norm": 0.13484346866607666,
+      "learning_rate": 6.215780998389694e-06,
+      "loss": 0.0001,
+      "step": 4560
+    },
+    {
+      "epoch": 65.00193236714976,
+      "grad_norm": 0.024939000606536865,
+      "learning_rate": 6.205045625335481e-06,
+      "loss": 0.0001,
+      "step": 4570
+    },
+    {
+      "epoch": 65.00289855072464,
+      "grad_norm": 0.0018361130496487021,
+      "learning_rate": 6.194310252281267e-06,
+      "loss": 0.0001,
+      "step": 4580
+    },
+    {
+      "epoch": 65.00386473429951,
+      "grad_norm": 0.0014188734348863363,
+      "learning_rate": 6.1835748792270535e-06,
+      "loss": 0.0001,
+      "step": 4590
+    },
+    {
+      "epoch": 65.0048309178744,
+      "grad_norm": 0.0007777014980092645,
+      "learning_rate": 6.17283950617284e-06,
+      "loss": 0.0,
+      "step": 4600
+    },
+    {
+      "epoch": 65.00579710144928,
+      "grad_norm": 0.0012215470196679235,
+      "learning_rate": 6.162104133118626e-06,
+      "loss": 0.0175,
+      "step": 4610
+    },
+    {
+      "epoch": 65.00676328502415,
+      "grad_norm": 0.002018085913732648,
+      "learning_rate": 6.151368760064412e-06,
+      "loss": 0.0,
+      "step": 4620
+    },
+    {
+      "epoch": 65.00676328502415,
+      "eval_accuracy": 0.782051282051282,
+      "eval_loss": 1.7766458988189697,
+      "eval_runtime": 20.7581,
+      "eval_samples_per_second": 3.758,
+      "eval_steps_per_second": 0.963,
+      "step": 4620
+    },
+    {
+      "epoch": 66.00096618357487,
+      "grad_norm": 0.005290956702083349,
+      "learning_rate": 6.1406333870101985e-06,
+      "loss": 0.0001,
+      "step": 4630
+    },
+    {
+      "epoch": 66.00193236714976,
+      "grad_norm": 0.0006458330899477005,
+      "learning_rate": 6.129898013955985e-06,
+      "loss": 0.0329,
+      "step": 4640
+    },
+    {
+      "epoch": 66.00289855072464,
+      "grad_norm": 0.0008798078051768243,
+      "learning_rate": 6.119162640901772e-06,
+      "loss": 0.0,
+      "step": 4650
+    },
+    {
+      "epoch": 66.00386473429951,
+      "grad_norm": 0.000433132256148383,
+      "learning_rate": 6.108427267847559e-06,
+      "loss": 0.0,
+      "step": 4660
+    },
+    {
+      "epoch": 66.0048309178744,
+      "grad_norm": 0.0006990334368310869,
+      "learning_rate": 6.097691894793345e-06,
+      "loss": 0.0001,
+      "step": 4670
+    },
+    {
+      "epoch": 66.00579710144928,
+      "grad_norm": 0.001073139370419085,
+      "learning_rate": 6.086956521739132e-06,
+      "loss": 0.2237,
+      "step": 4680
+    },
+    {
+      "epoch": 66.00676328502415,
+      "grad_norm": 0.001728722476400435,
+      "learning_rate": 6.076221148684918e-06,
+      "loss": 0.0214,
+      "step": 4690
+    },
+    {
+      "epoch": 66.00676328502415,
+      "eval_accuracy": 0.782051282051282,
+      "eval_loss": 1.635103464126587,
+      "eval_runtime": 20.6814,
+      "eval_samples_per_second": 3.771,
+      "eval_steps_per_second": 0.967,
+      "step": 4690
+    },
+    {
+      "epoch": 67.00096618357487,
+      "grad_norm": 0.001974219921976328,
+      "learning_rate": 6.065485775630704e-06,
+      "loss": 0.0,
+      "step": 4700
+    },
+    {
+      "epoch": 67.00193236714976,
+      "grad_norm": 0.007904019206762314,
+      "learning_rate": 6.05475040257649e-06,
+      "loss": 0.0,
+      "step": 4710
+    },
+    {
+      "epoch": 67.00289855072464,
+      "grad_norm": 0.007640121970325708,
+      "learning_rate": 6.0440150295222766e-06,
+      "loss": 0.0002,
+      "step": 4720
+    },
+    {
+      "epoch": 67.00386473429951,
+      "grad_norm": 0.0008578920387662947,
+      "learning_rate": 6.033279656468063e-06,
+      "loss": 0.1054,
+      "step": 4730
+    },
+    {
+      "epoch": 67.0048309178744,
+      "grad_norm": 0.0038633705116808414,
+      "learning_rate": 6.0225442834138495e-06,
+      "loss": 0.0,
+      "step": 4740
+    },
+    {
+      "epoch": 67.00579710144928,
+      "grad_norm": 0.000722988392226398,
+      "learning_rate": 6.011808910359636e-06,
+      "loss": 0.1052,
+      "step": 4750
+    },
+    {
+      "epoch": 67.00676328502415,
+      "grad_norm": 0.0019577303901314735,
+      "learning_rate": 6.0010735373054215e-06,
+      "loss": 0.0,
+      "step": 4760
+    },
+    {
+      "epoch": 67.00676328502415,
+      "eval_accuracy": 0.782051282051282,
+      "eval_loss": 1.687774896621704,
+      "eval_runtime": 21.4599,
+      "eval_samples_per_second": 3.635,
+      "eval_steps_per_second": 0.932,
+      "step": 4760
+    },
+    {
+      "epoch": 68.00096618357487,
+      "grad_norm": 0.0018796432996168733,
+      "learning_rate": 5.990338164251208e-06,
+      "loss": 0.0,
+      "step": 4770
+    },
+    {
+      "epoch": 68.00193236714976,
+      "grad_norm": 0.04547438025474548,
+      "learning_rate": 5.979602791196994e-06,
+      "loss": 0.0001,
+      "step": 4780
+    },
+    {
+      "epoch": 68.00289855072464,
+      "grad_norm": 0.0011448647128418088,
+      "learning_rate": 5.968867418142781e-06,
+      "loss": 0.0054,
+      "step": 4790
+    },
+    {
+      "epoch": 68.00386473429951,
+      "grad_norm": 0.0004199968825560063,
+      "learning_rate": 5.958132045088567e-06,
+      "loss": 0.0001,
+      "step": 4800
+    },
+    {
+      "epoch": 68.0048309178744,
+      "grad_norm": 0.0010871714912354946,
+      "learning_rate": 5.947396672034354e-06,
+      "loss": 0.0001,
+      "step": 4810
+    },
+    {
+      "epoch": 68.00579710144928,
+      "grad_norm": 0.000813886639662087,
+      "learning_rate": 5.936661298980139e-06,
+      "loss": 0.0,
+      "step": 4820
+    },
+    {
+      "epoch": 68.00676328502415,
+      "grad_norm": 0.0005087672616355121,
+      "learning_rate": 5.925925925925926e-06,
+      "loss": 0.0,
+      "step": 4830
+    },
+    {
+      "epoch": 68.00676328502415,
+      "eval_accuracy": 0.7564102564102564,
+      "eval_loss": 1.8763519525527954,
+      "eval_runtime": 20.8251,
+      "eval_samples_per_second": 3.745,
+      "eval_steps_per_second": 0.96,
+      "step": 4830
+    },
+    {
+      "epoch": 69.00096618357487,
+      "grad_norm": 0.0011922065168619156,
+      "learning_rate": 5.915190552871712e-06,
+      "loss": 0.0,
+      "step": 4840
+    },
+    {
+      "epoch": 69.00193236714976,
+      "grad_norm": 0.0004207005840726197,
+      "learning_rate": 5.904455179817499e-06,
+      "loss": 0.1623,
+      "step": 4850
+    },
+    {
+      "epoch": 69.00289855072464,
+      "grad_norm": 0.00039689309778623283,
+      "learning_rate": 5.893719806763285e-06,
+      "loss": 0.0,
+      "step": 4860
+    },
+    {
+      "epoch": 69.00386473429951,
+      "grad_norm": 0.0007422741036862135,
+      "learning_rate": 5.882984433709072e-06,
+      "loss": 0.073,
+      "step": 4870
+    },
+    {
+      "epoch": 69.0048309178744,
+      "grad_norm": 0.0005210656672716141,
+      "learning_rate": 5.872249060654859e-06,
+      "loss": 0.0,
+      "step": 4880
+    },
+    {
+      "epoch": 69.00579710144928,
+      "grad_norm": 0.0009733252227306366,
+      "learning_rate": 5.861513687600645e-06,
+      "loss": 0.0003,
+      "step": 4890
+    },
+    {
+      "epoch": 69.00676328502415,
+      "grad_norm": 0.0003511267132125795,
+      "learning_rate": 5.850778314546432e-06,
+      "loss": 0.2082,
+      "step": 4900
+    },
+    {
+      "epoch": 69.00676328502415,
+      "eval_accuracy": 0.7564102564102564,
+      "eval_loss": 1.779943585395813,
+      "eval_runtime": 21.1124,
+      "eval_samples_per_second": 3.695,
+      "eval_steps_per_second": 0.947,
+      "step": 4900
+    },
+    {
+      "epoch": 70.00096618357487,
+      "grad_norm": 0.0011452921899035573,
+      "learning_rate": 5.840042941492217e-06,
+      "loss": 0.0,
+      "step": 4910
+    },
+    {
+      "epoch": 70.00193236714976,
+      "grad_norm": 0.004901723936200142,
+      "learning_rate": 5.829307568438004e-06,
+      "loss": 0.0024,
+      "step": 4920
+    },
+    {
+      "epoch": 70.00289855072464,
+      "grad_norm": 0.006064664572477341,
+      "learning_rate": 5.81857219538379e-06,
+      "loss": 0.0,
+      "step": 4930
+    },
+    {
+      "epoch": 70.00386473429951,
+      "grad_norm": 0.0014497352531179786,
+      "learning_rate": 5.807836822329577e-06,
+      "loss": 0.0065,
+      "step": 4940
+    },
+    {
+      "epoch": 70.0048309178744,
+      "grad_norm": 0.0013388212537392974,
+      "learning_rate": 5.797101449275363e-06,
+      "loss": 0.0001,
+      "step": 4950
+    },
+    {
+      "epoch": 70.00579710144928,
+      "grad_norm": 0.0011337369214743376,
+      "learning_rate": 5.78636607622115e-06,
+      "loss": 0.0,
+      "step": 4960
+    },
+    {
+      "epoch": 70.00676328502415,
+      "grad_norm": 0.000552912475541234,
+      "learning_rate": 5.775630703166935e-06,
+      "loss": 0.0,
+      "step": 4970
+    },
+    {
+      "epoch": 70.00676328502415,
+      "eval_accuracy": 0.7948717948717948,
+      "eval_loss": 1.7388193607330322,
+      "eval_runtime": 24.3328,
+      "eval_samples_per_second": 3.206,
+      "eval_steps_per_second": 0.822,
+      "step": 4970
+    },
+    {
+      "epoch": 71.00096618357487,
+      "grad_norm": 0.0012520031305029988,
+      "learning_rate": 5.764895330112722e-06,
+      "loss": 0.0,
+      "step": 4980
+    },
+    {
+      "epoch": 71.00193236714976,
+      "grad_norm": 0.0008118404657579958,
+      "learning_rate": 5.754159957058508e-06,
+      "loss": 0.0,
+      "step": 4990
+    },
+    {
+      "epoch": 71.00289855072464,
+      "grad_norm": 0.0011605226900428534,
+      "learning_rate": 5.743424584004295e-06,
+      "loss": 0.0,
+      "step": 5000
+    },
+    {
+      "epoch": 71.00386473429951,
+      "grad_norm": 0.0014740672195330262,
+      "learning_rate": 5.732689210950081e-06,
+      "loss": 0.0,
+      "step": 5010
+    },
+    {
+      "epoch": 71.0048309178744,
+      "grad_norm": 0.0008719124598428607,
+      "learning_rate": 5.7219538378958675e-06,
+      "loss": 0.0001,
+      "step": 5020
+    },
+    {
+      "epoch": 71.00579710144928,
+      "grad_norm": 0.0011694729328155518,
+      "learning_rate": 5.711218464841653e-06,
+      "loss": 0.0,
+      "step": 5030
+    },
+    {
+      "epoch": 71.00676328502415,
+      "grad_norm": 0.0014133198419585824,
+      "learning_rate": 5.70048309178744e-06,
+      "loss": 0.0,
+      "step": 5040
+    },
+    {
+      "epoch": 71.00676328502415,
+      "eval_accuracy": 0.7948717948717948,
+      "eval_loss": 1.6719121932983398,
+      "eval_runtime": 15.9285,
+      "eval_samples_per_second": 4.897,
+      "eval_steps_per_second": 1.256,
+      "step": 5040
+    },
+    {
+      "epoch": 72.00096618357487,
+      "grad_norm": 0.0005829191650263965,
+      "learning_rate": 5.689747718733226e-06,
+      "loss": 0.0,
+      "step": 5050
+    },
+    {
+      "epoch": 72.00193236714976,
+      "grad_norm": 0.0113700395449996,
+      "learning_rate": 5.6790123456790125e-06,
+      "loss": 0.0,
+      "step": 5060
+    },
+    {
+      "epoch": 72.00289855072464,
+      "grad_norm": 0.001089294906705618,
+      "learning_rate": 5.668276972624799e-06,
+      "loss": 0.0455,
+      "step": 5070
+    },
+    {
+      "epoch": 72.00386473429951,
+      "grad_norm": 0.00036309906863607466,
+      "learning_rate": 5.657541599570585e-06,
+      "loss": 0.1185,
+      "step": 5080
+    },
+    {
+      "epoch": 72.0048309178744,
+      "grad_norm": 0.0007753897807560861,
+      "learning_rate": 5.646806226516372e-06,
+      "loss": 0.0,
+      "step": 5090
+    },
+    {
+      "epoch": 72.00579710144928,
+      "grad_norm": 0.0005843242397531867,
+      "learning_rate": 5.6360708534621574e-06,
+      "loss": 0.0,
+      "step": 5100
+    },
+    {
+      "epoch": 72.00676328502415,
+      "grad_norm": 0.007903228513896465,
+      "learning_rate": 5.625335480407944e-06,
+      "loss": 0.0001,
+      "step": 5110
+    },
+    {
+      "epoch": 72.00676328502415,
+      "eval_accuracy": 0.7948717948717948,
+      "eval_loss": 1.6066139936447144,
+      "eval_runtime": 15.2419,
+      "eval_samples_per_second": 5.117,
+      "eval_steps_per_second": 1.312,
+      "step": 5110
+    },
+    {
+      "epoch": 73.00096618357487,
+      "grad_norm": 0.09604029357433319,
+      "learning_rate": 5.614600107353731e-06,
+      "loss": 0.0037,
+      "step": 5120
+    },
+    {
+      "epoch": 73.00193236714976,
+      "grad_norm": 0.00048322087968699634,
+      "learning_rate": 5.603864734299518e-06,
+      "loss": 0.0,
+      "step": 5130
+    },
+    {
+      "epoch": 73.00289855072464,
+      "grad_norm": 0.0017982334829866886,
+      "learning_rate": 5.593129361245304e-06,
+      "loss": 0.0,
+      "step": 5140
+    },
+    {
+      "epoch": 73.00386473429951,
+      "grad_norm": 0.02436600998044014,
+      "learning_rate": 5.5823939881910905e-06,
+      "loss": 0.0951,
+      "step": 5150
+    },
+    {
+      "epoch": 73.0048309178744,
+      "grad_norm": 0.0008387943380512297,
+      "learning_rate": 5.571658615136877e-06,
+      "loss": 0.0,
+      "step": 5160
+    },
+    {
+      "epoch": 73.00579710144928,
+      "grad_norm": 0.0010867329547181726,
+      "learning_rate": 5.5609232420826634e-06,
+      "loss": 0.0049,
+      "step": 5170
+    },
+    {
+      "epoch": 73.00676328502415,
+      "grad_norm": 0.002065508160740137,
+      "learning_rate": 5.550187869028449e-06,
+      "loss": 0.0001,
+      "step": 5180
+    },
+    {
+      "epoch": 73.00676328502415,
+      "eval_accuracy": 0.7564102564102564,
+      "eval_loss": 2.1180622577667236,
+      "eval_runtime": 14.9441,
+      "eval_samples_per_second": 5.219,
+      "eval_steps_per_second": 1.338,
+      "step": 5180
+    },
+    {
+      "epoch": 74.00096618357487,
+      "grad_norm": 0.001679830951616168,
+      "learning_rate": 5.5394524959742355e-06,
+      "loss": 0.0,
+      "step": 5190
+    },
+    {
+      "epoch": 74.00193236714976,
+      "grad_norm": 0.0013725162716582417,
+      "learning_rate": 5.528717122920022e-06,
+      "loss": 0.2049,
+      "step": 5200
+    },
+    {
+      "epoch": 74.00289855072464,
+      "grad_norm": 0.0021361317485570908,
+      "learning_rate": 5.517981749865808e-06,
+      "loss": 0.0,
+      "step": 5210
+    },
+    {
+      "epoch": 74.00386473429951,
+      "grad_norm": 0.006481006741523743,
+      "learning_rate": 5.507246376811595e-06,
+      "loss": 0.0,
+      "step": 5220
+    },
+    {
+      "epoch": 74.0048309178744,
+      "grad_norm": 0.037138450890779495,
+      "learning_rate": 5.496511003757381e-06,
+      "loss": 0.0,
+      "step": 5230
+    },
+    {
+      "epoch": 74.00579710144928,
+      "grad_norm": 0.001368780038319528,
+      "learning_rate": 5.485775630703167e-06,
+      "loss": 0.1058,
+      "step": 5240
+    },
+    {
+      "epoch": 74.00676328502415,
+      "grad_norm": 0.0019940112251788378,
+      "learning_rate": 5.475040257648953e-06,
+      "loss": 0.0,
+      "step": 5250
+    },
+    {
+      "epoch": 74.00676328502415,
+      "eval_accuracy": 0.7564102564102564,
+      "eval_loss": 2.1772854328155518,
+      "eval_runtime": 17.1681,
+      "eval_samples_per_second": 4.543,
+      "eval_steps_per_second": 1.165,
+      "step": 5250
+    },
+    {
+      "epoch": 75.00096618357487,
+      "grad_norm": 0.0009713310282677412,
+      "learning_rate": 5.46430488459474e-06,
+      "loss": 0.0,
+      "step": 5260
+    },
+    {
+      "epoch": 75.00193236714976,
+      "grad_norm": 0.0003569677355699241,
+      "learning_rate": 5.453569511540526e-06,
+      "loss": 0.2552,
+      "step": 5270
+    },
+    {
+      "epoch": 75.00289855072464,
+      "grad_norm": 0.0033191258553415537,
+      "learning_rate": 5.442834138486313e-06,
+      "loss": 0.0,
+      "step": 5280
+    },
+    {
+      "epoch": 75.00386473429951,
+      "grad_norm": 0.019612202420830727,
+      "learning_rate": 5.432098765432099e-06,
+      "loss": 0.0001,
+      "step": 5290
+    },
+    {
+      "epoch": 75.0048309178744,
+      "grad_norm": 0.0007422531489282846,
+      "learning_rate": 5.421363392377886e-06,
+      "loss": 0.0001,
+      "step": 5300
+    },
+    {
+      "epoch": 75.00579710144928,
+      "grad_norm": 0.0014051502803340554,
+      "learning_rate": 5.410628019323671e-06,
+      "loss": 0.0053,
+      "step": 5310
+    },
+    {
+      "epoch": 75.00676328502415,
+      "grad_norm": 0.0010339098516851664,
+      "learning_rate": 5.399892646269458e-06,
+      "loss": 0.0,
+      "step": 5320
+    },
+    {
+      "epoch": 75.00676328502415,
+      "eval_accuracy": 0.717948717948718,
+      "eval_loss": 2.5632450580596924,
+      "eval_runtime": 17.5653,
+      "eval_samples_per_second": 4.441,
+      "eval_steps_per_second": 1.139,
+      "step": 5320
+    },
+    {
+      "epoch": 76.00096618357487,
+      "grad_norm": 0.0006680580554530025,
+      "learning_rate": 5.389157273215244e-06,
+      "loss": 0.1044,
+      "step": 5330
+    },
+    {
+      "epoch": 76.00193236714976,
+      "grad_norm": 0.0012755277566611767,
+      "learning_rate": 5.3784219001610306e-06,
+      "loss": 0.2949,
+      "step": 5340
+    },
+    {
+      "epoch": 76.00289855072464,
+      "grad_norm": 0.0015628045657649636,
+      "learning_rate": 5.367686527106818e-06,
+      "loss": 0.0003,
+      "step": 5350
+    },
+    {
+      "epoch": 76.00386473429951,
+      "grad_norm": 0.0014394833706319332,
+      "learning_rate": 5.356951154052604e-06,
+      "loss": 0.0271,
+      "step": 5360
+    },
+    {
+      "epoch": 76.0048309178744,
+      "grad_norm": 0.006877475883811712,
+      "learning_rate": 5.346215780998391e-06,
+      "loss": 0.0,
+      "step": 5370
+    },
+    {
+      "epoch": 76.00579710144928,
+      "grad_norm": 0.0015750023303553462,
+      "learning_rate": 5.335480407944177e-06,
+      "loss": 0.0,
+      "step": 5380
+    },
+    {
+      "epoch": 76.00676328502415,
+      "grad_norm": 0.0003570659027900547,
+      "learning_rate": 5.324745034889963e-06,
+      "loss": 0.0,
+      "step": 5390
+    },
+    {
+      "epoch": 76.00676328502415,
+      "eval_accuracy": 0.8076923076923077,
+      "eval_loss": 1.1549041271209717,
+      "eval_runtime": 15.4003,
+      "eval_samples_per_second": 5.065,
+      "eval_steps_per_second": 1.299,
+      "step": 5390
+    },
+    {
+      "epoch": 77.00096618357487,
+      "grad_norm": 0.0005772082367911935,
+      "learning_rate": 5.314009661835749e-06,
+      "loss": 0.0,
+      "step": 5400
+    },
+    {
+      "epoch": 77.00193236714976,
+      "grad_norm": 0.0010372302494943142,
+      "learning_rate": 5.303274288781536e-06,
+      "loss": 0.127,
+      "step": 5410
+    },
+    {
+      "epoch": 77.00289855072464,
+      "grad_norm": 0.0018043516902253032,
+      "learning_rate": 5.292538915727322e-06,
+      "loss": 0.0,
+      "step": 5420
+    },
+    {
+      "epoch": 77.00386473429951,
+      "grad_norm": 0.0005784116219729185,
+      "learning_rate": 5.281803542673109e-06,
+      "loss": 0.2015,
+      "step": 5430
+    },
+    {
+      "epoch": 77.0048309178744,
+      "grad_norm": 0.006113346666097641,
+      "learning_rate": 5.271068169618895e-06,
+      "loss": 0.0,
+      "step": 5440
+    },
+    {
+      "epoch": 77.00579710144928,
+      "grad_norm": 0.002624011831358075,
+      "learning_rate": 5.2603327965646815e-06,
+      "loss": 0.0003,
+      "step": 5450
+    },
+    {
+      "epoch": 77.00676328502415,
+      "grad_norm": 0.0012411015341058373,
+      "learning_rate": 5.249597423510467e-06,
+      "loss": 0.0006,
+      "step": 5460
+    },
+    {
+      "epoch": 77.00676328502415,
+      "eval_accuracy": 0.7435897435897436,
+      "eval_loss": 2.229558229446411,
+      "eval_runtime": 17.9739,
+      "eval_samples_per_second": 4.34,
+      "eval_steps_per_second": 1.113,
+      "step": 5460
+    },
+    {
+      "epoch": 78.00096618357487,
+      "grad_norm": 0.0013451834674924612,
+      "learning_rate": 5.2388620504562536e-06,
+      "loss": 0.0023,
+      "step": 5470
+    },
+    {
+      "epoch": 78.00193236714976,
+      "grad_norm": 0.0013466801028698683,
+      "learning_rate": 5.22812667740204e-06,
+      "loss": 0.0005,
+      "step": 5480
+    },
+    {
+      "epoch": 78.00289855072464,
+      "grad_norm": 0.2549854815006256,
+      "learning_rate": 5.2173913043478265e-06,
+      "loss": 0.1312,
+      "step": 5490
+    },
+    {
+      "epoch": 78.00386473429951,
+      "grad_norm": 0.011888462118804455,
+      "learning_rate": 5.206655931293613e-06,
+      "loss": 0.2297,
+      "step": 5500
+    },
+    {
+      "epoch": 78.0048309178744,
+      "grad_norm": 0.006290052086114883,
+      "learning_rate": 5.195920558239399e-06,
+      "loss": 0.0001,
+      "step": 5510
+    },
+    {
+      "epoch": 78.00579710144928,
+      "grad_norm": 0.0024707193952053785,
+      "learning_rate": 5.185185185185185e-06,
+      "loss": 0.0196,
+      "step": 5520
+    },
+    {
+      "epoch": 78.00676328502415,
+      "grad_norm": 0.0009130489779636264,
+      "learning_rate": 5.1744498121309714e-06,
+      "loss": 0.0,
+      "step": 5530
+    },
+    {
+      "epoch": 78.00676328502415,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 2.5381572246551514,
+      "eval_runtime": 15.5391,
+      "eval_samples_per_second": 5.02,
+      "eval_steps_per_second": 1.287,
+      "step": 5530
+    },
+    {
+      "epoch": 79.00096618357487,
+      "grad_norm": 0.0007505626999773085,
+      "learning_rate": 5.163714439076758e-06,
+      "loss": 0.0064,
+      "step": 5540
+    },
+    {
+      "epoch": 79.00193236714976,
+      "grad_norm": 0.0007742494344711304,
+      "learning_rate": 5.152979066022544e-06,
+      "loss": 0.0,
+      "step": 5550
+    },
+    {
+      "epoch": 79.00289855072464,
+      "grad_norm": 0.0012070463271811604,
+      "learning_rate": 5.142243692968331e-06,
+      "loss": 0.1978,
+      "step": 5560
+    },
+    {
+      "epoch": 79.00386473429951,
+      "grad_norm": 0.0034262288827449083,
+      "learning_rate": 5.131508319914117e-06,
+      "loss": 0.0,
+      "step": 5570
+    },
+    {
+      "epoch": 79.0048309178744,
+      "grad_norm": 0.0023414520546793938,
+      "learning_rate": 5.1207729468599045e-06,
+      "loss": 0.0003,
+      "step": 5580
+    },
+    {
+      "epoch": 79.00579710144928,
+      "grad_norm": 0.0022120054345577955,
+      "learning_rate": 5.110037573805691e-06,
+      "loss": 0.1615,
+      "step": 5590
+    },
+    {
+      "epoch": 79.00676328502415,
+      "grad_norm": 0.0005375827313400805,
+      "learning_rate": 5.099302200751477e-06,
+      "loss": 0.0001,
+      "step": 5600
+    },
+    {
+      "epoch": 79.00676328502415,
+      "eval_accuracy": 0.782051282051282,
+      "eval_loss": 1.5726310014724731,
+      "eval_runtime": 15.7655,
+      "eval_samples_per_second": 4.948,
+      "eval_steps_per_second": 1.269,
+      "step": 5600
+    },
+    {
+      "epoch": 80.00096618357487,
+      "grad_norm": 0.0010982960229739547,
+      "learning_rate": 5.088566827697263e-06,
+      "loss": 0.0001,
+      "step": 5610
+    },
+    {
+      "epoch": 80.00193236714976,
+      "grad_norm": 0.0007016141898930073,
+      "learning_rate": 5.0778314546430495e-06,
+      "loss": 0.0,
+      "step": 5620
+    },
+    {
+      "epoch": 80.00289855072464,
+      "grad_norm": 0.001342971809208393,
+      "learning_rate": 5.067096081588836e-06,
+      "loss": 0.0,
+      "step": 5630
+    },
+    {
+      "epoch": 80.00386473429951,
+      "grad_norm": 0.019569087773561478,
+      "learning_rate": 5.056360708534622e-06,
+      "loss": 0.0,
+      "step": 5640
+    },
+    {
+      "epoch": 80.0048309178744,
+      "grad_norm": 0.00047855902812443674,
+      "learning_rate": 5.045625335480409e-06,
+      "loss": 0.0,
+      "step": 5650
+    },
+    {
+      "epoch": 80.00579710144928,
+      "grad_norm": 0.0011246672365814447,
+      "learning_rate": 5.034889962426195e-06,
+      "loss": 0.0,
+      "step": 5660
+    },
+    {
+      "epoch": 80.00676328502415,
+      "grad_norm": 0.0016472653951495886,
+      "learning_rate": 5.024154589371981e-06,
+      "loss": 0.0002,
+      "step": 5670
+    },
+    {
+      "epoch": 80.00676328502415,
+      "eval_accuracy": 0.8076923076923077,
+      "eval_loss": 1.57036292552948,
+      "eval_runtime": 14.8715,
+      "eval_samples_per_second": 5.245,
+      "eval_steps_per_second": 1.345,
+      "step": 5670
+    },
+    {
+      "epoch": 81.00096618357487,
+      "grad_norm": 0.01656234823167324,
+      "learning_rate": 5.013419216317767e-06,
+      "loss": 0.0001,
+      "step": 5680
+    },
+    {
+      "epoch": 81.00193236714976,
+      "grad_norm": 0.002044126158580184,
+      "learning_rate": 5.002683843263554e-06,
+      "loss": 0.0,
+      "step": 5690
+    },
+    {
+      "epoch": 81.00289855072464,
+      "grad_norm": 0.0057681528851389885,
+      "learning_rate": 4.99194847020934e-06,
+      "loss": 0.0,
+      "step": 5700
+    },
+    {
+      "epoch": 81.00386473429951,
+      "grad_norm": 0.006356089375913143,
+      "learning_rate": 4.981213097155127e-06,
+      "loss": 0.001,
+      "step": 5710
+    },
+    {
+      "epoch": 81.0048309178744,
+      "grad_norm": 0.0005163434543646872,
+      "learning_rate": 4.970477724100913e-06,
+      "loss": 0.198,
+      "step": 5720
+    },
+    {
+      "epoch": 81.00579710144928,
+      "grad_norm": 0.0019573138561099768,
+      "learning_rate": 4.959742351046699e-06,
+      "loss": 0.0,
+      "step": 5730
+    },
+    {
+      "epoch": 81.00676328502415,
+      "grad_norm": 0.0007160149980336428,
+      "learning_rate": 4.949006977992485e-06,
+      "loss": 0.0199,
+      "step": 5740
+    },
+    {
+      "epoch": 81.00676328502415,
+      "eval_accuracy": 0.8076923076923077,
+      "eval_loss": 1.5503082275390625,
+      "eval_runtime": 14.9894,
+      "eval_samples_per_second": 5.204,
+      "eval_steps_per_second": 1.334,
+      "step": 5740
+    },
+    {
+      "epoch": 82.00096618357487,
+      "grad_norm": 0.000732768268790096,
+      "learning_rate": 4.938271604938272e-06,
+      "loss": 0.0001,
+      "step": 5750
+    },
+    {
+      "epoch": 82.00193236714976,
+      "grad_norm": 0.0015916397096589208,
+      "learning_rate": 4.927536231884059e-06,
+      "loss": 0.0,
+      "step": 5760
+    },
+    {
+      "epoch": 82.00289855072464,
+      "grad_norm": 0.0016159799415618181,
+      "learning_rate": 4.9168008588298446e-06,
+      "loss": 0.0006,
+      "step": 5770
+    },
+    {
+      "epoch": 82.00386473429951,
+      "grad_norm": 0.003708163509145379,
+      "learning_rate": 4.906065485775631e-06,
+      "loss": 0.0,
+      "step": 5780
+    },
+    {
+      "epoch": 82.0048309178744,
+      "grad_norm": 0.0009543473133817315,
+      "learning_rate": 4.8953301127214175e-06,
+      "loss": 0.2218,
+      "step": 5790
+    },
+    {
+      "epoch": 82.00579710144928,
+      "grad_norm": 0.0033412850461900234,
+      "learning_rate": 4.884594739667204e-06,
+      "loss": 0.0,
+      "step": 5800
+    },
+    {
+      "epoch": 82.00676328502415,
+      "grad_norm": 0.003886610036715865,
+      "learning_rate": 4.87385936661299e-06,
+      "loss": 0.0001,
+      "step": 5810
+    },
+    {
+      "epoch": 82.00676328502415,
+      "eval_accuracy": 0.8076923076923077,
+      "eval_loss": 1.365352749824524,
+      "eval_runtime": 15.0916,
+      "eval_samples_per_second": 5.168,
+      "eval_steps_per_second": 1.325,
+      "step": 5810
+    },
+    {
+      "epoch": 83.00096618357487,
+      "grad_norm": 0.01847156696021557,
+      "learning_rate": 4.863123993558777e-06,
+      "loss": 0.0855,
+      "step": 5820
+    },
+    {
+      "epoch": 83.00193236714976,
+      "grad_norm": 0.0023459733929485083,
+      "learning_rate": 4.852388620504562e-06,
+      "loss": 0.0009,
+      "step": 5830
+    },
+    {
+      "epoch": 83.00289855072464,
+      "grad_norm": 0.0007858324679546058,
+      "learning_rate": 4.841653247450349e-06,
+      "loss": 0.0,
+      "step": 5840
+    },
+    {
+      "epoch": 83.00386473429951,
+      "grad_norm": 0.0015262184897437692,
+      "learning_rate": 4.830917874396135e-06,
+      "loss": 0.0,
+      "step": 5850
+    },
+    {
+      "epoch": 83.0048309178744,
+      "grad_norm": 0.0011216209968551993,
+      "learning_rate": 4.820182501341922e-06,
+      "loss": 0.0,
+      "step": 5860
+    },
+    {
+      "epoch": 83.00579710144928,
+      "grad_norm": 0.010345948860049248,
+      "learning_rate": 4.809447128287709e-06,
+      "loss": 0.0,
+      "step": 5870
+    },
+    {
+      "epoch": 83.00676328502415,
+      "grad_norm": 0.00381542113609612,
+      "learning_rate": 4.798711755233495e-06,
+      "loss": 0.0,
+      "step": 5880
+    },
+    {
+      "epoch": 83.00676328502415,
+      "eval_accuracy": 0.8076923076923077,
+      "eval_loss": 1.4726909399032593,
+      "eval_runtime": 14.7359,
+      "eval_samples_per_second": 5.293,
+      "eval_steps_per_second": 1.357,
+      "step": 5880
+    },
+    {
+      "epoch": 84.00096618357487,
+      "grad_norm": 0.001281603006646037,
+      "learning_rate": 4.787976382179281e-06,
+      "loss": 0.0004,
+      "step": 5890
+    },
+    {
+      "epoch": 84.00193236714976,
+      "grad_norm": 0.0006520389579236507,
+      "learning_rate": 4.7772410091250676e-06,
+      "loss": 0.0,
+      "step": 5900
+    },
+    {
+      "epoch": 84.00289855072464,
+      "grad_norm": 0.0008780735661275685,
+      "learning_rate": 4.766505636070854e-06,
+      "loss": 0.0,
+      "step": 5910
+    },
+    {
+      "epoch": 84.00386473429951,
+      "grad_norm": 0.00047110416926443577,
+      "learning_rate": 4.7557702630166405e-06,
+      "loss": 0.1301,
+      "step": 5920
+    },
+    {
+      "epoch": 84.0048309178744,
+      "grad_norm": 0.0020175003446638584,
+      "learning_rate": 4.745034889962427e-06,
+      "loss": 0.0,
+      "step": 5930
+    },
+    {
+      "epoch": 84.00579710144928,
+      "grad_norm": 0.001587681588716805,
+      "learning_rate": 4.7342995169082125e-06,
+      "loss": 0.0585,
+      "step": 5940
+    },
+    {
+      "epoch": 84.00676328502415,
+      "grad_norm": 0.006288263946771622,
+      "learning_rate": 4.723564143853999e-06,
+      "loss": 0.0001,
+      "step": 5950
+    },
+    {
+      "epoch": 84.00676328502415,
+      "eval_accuracy": 0.717948717948718,
+      "eval_loss": 2.2248222827911377,
+      "eval_runtime": 15.8238,
+      "eval_samples_per_second": 4.929,
+      "eval_steps_per_second": 1.264,
+      "step": 5950
+    },
+    {
+      "epoch": 85.00096618357487,
+      "grad_norm": 0.0007544896216131747,
+      "learning_rate": 4.712828770799785e-06,
+      "loss": 0.1264,
+      "step": 5960
+    },
+    {
+      "epoch": 85.00193236714976,
+      "grad_norm": 0.015535094775259495,
+      "learning_rate": 4.702093397745572e-06,
+      "loss": 0.0,
+      "step": 5970
+    },
+    {
+      "epoch": 85.00289855072464,
+      "grad_norm": 0.0009989738464355469,
+      "learning_rate": 4.691358024691358e-06,
+      "loss": 0.0,
+      "step": 5980
+    },
+    {
+      "epoch": 85.00386473429951,
+      "grad_norm": 0.0020626296754926443,
+      "learning_rate": 4.680622651637145e-06,
+      "loss": 0.0,
+      "step": 5990
+    },
+    {
+      "epoch": 85.0048309178744,
+      "grad_norm": 0.0007889845292083919,
+      "learning_rate": 4.669887278582931e-06,
+      "loss": 0.0,
+      "step": 6000
+    },
+    {
+      "epoch": 85.00579710144928,
+      "grad_norm": 0.0008909267489798367,
+      "learning_rate": 4.659151905528718e-06,
+      "loss": 0.0001,
+      "step": 6010
+    },
+    {
+      "epoch": 85.00676328502415,
+      "grad_norm": 0.0013790903612971306,
+      "learning_rate": 4.648416532474504e-06,
+      "loss": 0.0,
+      "step": 6020
+    },
+    {
+      "epoch": 85.00676328502415,
+      "eval_accuracy": 0.782051282051282,
+      "eval_loss": 1.7187656164169312,
+      "eval_runtime": 15.1982,
+      "eval_samples_per_second": 5.132,
+      "eval_steps_per_second": 1.316,
+      "step": 6020
+    },
+    {
+      "epoch": 86.00096618357487,
+      "grad_norm": 0.0003605932288337499,
+      "learning_rate": 4.637681159420291e-06,
+      "loss": 0.2248,
+      "step": 6030
+    },
+    {
+      "epoch": 86.00193236714976,
+      "grad_norm": 0.002730882028117776,
+      "learning_rate": 4.626945786366077e-06,
+      "loss": 0.001,
+      "step": 6040
+    },
+    {
+      "epoch": 86.00289855072464,
+      "grad_norm": 0.0010079998755827546,
+      "learning_rate": 4.616210413311863e-06,
+      "loss": 0.1221,
+      "step": 6050
+    },
+    {
+      "epoch": 86.00386473429951,
+      "grad_norm": 0.0005993593367747962,
+      "learning_rate": 4.605475040257649e-06,
+      "loss": 0.0,
+      "step": 6060
+    },
+    {
+      "epoch": 86.0048309178744,
+      "grad_norm": 0.0010675563244149089,
+      "learning_rate": 4.5947396672034355e-06,
+      "loss": 0.0,
+      "step": 6070
+    },
+    {
+      "epoch": 86.00579710144928,
+      "grad_norm": 0.0026733041740953922,
+      "learning_rate": 4.584004294149222e-06,
+      "loss": 0.0,
+      "step": 6080
+    },
+    {
+      "epoch": 86.00676328502415,
+      "grad_norm": 0.0020553949289023876,
+      "learning_rate": 4.5732689210950084e-06,
+      "loss": 0.0101,
+      "step": 6090
+    },
+    {
+      "epoch": 86.00676328502415,
+      "eval_accuracy": 0.7435897435897436,
+      "eval_loss": 1.927074909210205,
+      "eval_runtime": 14.9313,
+      "eval_samples_per_second": 5.224,
+      "eval_steps_per_second": 1.339,
+      "step": 6090
+    },
+    {
+      "epoch": 87.00096618357487,
+      "grad_norm": 0.0008128826157189906,
+      "learning_rate": 4.562533548040795e-06,
+      "loss": 0.0024,
+      "step": 6100
+    },
+    {
+      "epoch": 87.00193236714976,
+      "grad_norm": 0.0006628745468333364,
+      "learning_rate": 4.551798174986581e-06,
+      "loss": 0.0,
+      "step": 6110
+    },
+    {
+      "epoch": 87.00289855072464,
+      "grad_norm": 0.04763895273208618,
+      "learning_rate": 4.541062801932368e-06,
+      "loss": 0.0,
+      "step": 6120
+    },
+    {
+      "epoch": 87.00386473429951,
+      "grad_norm": 0.0007302842568606138,
+      "learning_rate": 4.530327428878154e-06,
+      "loss": 0.0001,
+      "step": 6130
+    },
+    {
+      "epoch": 87.0048309178744,
+      "grad_norm": 0.0005195592530071735,
+      "learning_rate": 4.519592055823941e-06,
+      "loss": 0.0,
+      "step": 6140
+    },
+    {
+      "epoch": 87.00579710144928,
+      "grad_norm": 0.001573888468556106,
+      "learning_rate": 4.508856682769726e-06,
+      "loss": 0.0,
+      "step": 6150
+    },
+    {
+      "epoch": 87.00676328502415,
+      "grad_norm": 0.0013048958498984575,
+      "learning_rate": 4.498121309715513e-06,
+      "loss": 0.0001,
+      "step": 6160
+    },
+    {
+      "epoch": 87.00676328502415,
+      "eval_accuracy": 0.8333333333333334,
+      "eval_loss": 1.5355632305145264,
+      "eval_runtime": 14.8356,
+      "eval_samples_per_second": 5.258,
+      "eval_steps_per_second": 1.348,
+      "step": 6160
+    },
+    {
+      "epoch": 88.00096618357487,
+      "grad_norm": 0.0004708681080956012,
+      "learning_rate": 4.487385936661299e-06,
+      "loss": 0.0,
+      "step": 6170
+    },
+    {
+      "epoch": 88.00193236714976,
+      "grad_norm": 0.00878297071903944,
+      "learning_rate": 4.476650563607086e-06,
+      "loss": 0.0,
+      "step": 6180
+    },
+    {
+      "epoch": 88.00289855072464,
+      "grad_norm": 0.0006720071542076766,
+      "learning_rate": 4.465915190552872e-06,
+      "loss": 0.0,
+      "step": 6190
+    },
+    {
+      "epoch": 88.00386473429951,
+      "grad_norm": 0.0007516877958551049,
+      "learning_rate": 4.4551798174986585e-06,
+      "loss": 0.051,
+      "step": 6200
+    },
+    {
+      "epoch": 88.0048309178744,
+      "grad_norm": 0.001620340975932777,
+      "learning_rate": 4.444444444444444e-06,
+      "loss": 0.0,
+      "step": 6210
+    },
+    {
+      "epoch": 88.00579710144928,
+      "grad_norm": 0.0005804314860142767,
+      "learning_rate": 4.4337090713902314e-06,
+      "loss": 0.0005,
+      "step": 6220
+    },
+    {
+      "epoch": 88.00676328502415,
+      "grad_norm": 0.0008902663248591125,
+      "learning_rate": 4.422973698336018e-06,
+      "loss": 0.0,
+      "step": 6230
+    },
+    {
+      "epoch": 88.00676328502415,
+      "eval_accuracy": 0.8076923076923077,
+      "eval_loss": 1.6682919263839722,
+      "eval_runtime": 14.9271,
+      "eval_samples_per_second": 5.225,
+      "eval_steps_per_second": 1.34,
+      "step": 6230
+    },
+    {
+      "epoch": 89.00096618357487,
+      "grad_norm": 0.0005711275152862072,
+      "learning_rate": 4.412238325281804e-06,
+      "loss": 0.0,
+      "step": 6240
+    },
+    {
+      "epoch": 89.00193236714976,
+      "grad_norm": 0.0005052529741078615,
+      "learning_rate": 4.401502952227591e-06,
+      "loss": 0.0,
+      "step": 6250
+    },
+    {
+      "epoch": 89.00289855072464,
+      "grad_norm": 0.0013751588994637132,
+      "learning_rate": 4.390767579173376e-06,
+      "loss": 0.0,
+      "step": 6260
+    },
+    {
+      "epoch": 89.00386473429951,
+      "grad_norm": 0.0006107139051891863,
+      "learning_rate": 4.380032206119163e-06,
+      "loss": 0.0,
+      "step": 6270
+    },
+    {
+      "epoch": 89.0048309178744,
+      "grad_norm": 0.000693678914103657,
+      "learning_rate": 4.369296833064949e-06,
+      "loss": 0.0,
+      "step": 6280
+    },
+    {
+      "epoch": 89.00579710144928,
+      "grad_norm": 0.0005502082640305161,
+      "learning_rate": 4.358561460010736e-06,
+      "loss": 0.2555,
+      "step": 6290
+    },
+    {
+      "epoch": 89.00676328502415,
+      "grad_norm": 0.0003575296141207218,
+      "learning_rate": 4.347826086956522e-06,
+      "loss": 0.0,
+      "step": 6300
+    },
+    {
+      "epoch": 89.00676328502415,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 2.5565624237060547,
+      "eval_runtime": 16.7423,
+      "eval_samples_per_second": 4.659,
+      "eval_steps_per_second": 1.195,
+      "step": 6300
+    },
+    {
+      "epoch": 90.00096618357487,
+      "grad_norm": 0.0003587829996831715,
+      "learning_rate": 4.337090713902309e-06,
+      "loss": 0.0006,
+      "step": 6310
+    },
+    {
+      "epoch": 90.00193236714976,
+      "grad_norm": 0.001737708691507578,
+      "learning_rate": 4.326355340848094e-06,
+      "loss": 0.0,
+      "step": 6320
+    },
+    {
+      "epoch": 90.00289855072464,
+      "grad_norm": 43.55590057373047,
+      "learning_rate": 4.315619967793881e-06,
+      "loss": 0.0018,
+      "step": 6330
+    },
+    {
+      "epoch": 90.00386473429951,
+      "grad_norm": 4.491352081298828,
+      "learning_rate": 4.304884594739668e-06,
+      "loss": 0.0003,
+      "step": 6340
+    },
+    {
+      "epoch": 90.0048309178744,
+      "grad_norm": 0.0008951724739745259,
+      "learning_rate": 4.2941492216854545e-06,
+      "loss": 0.0,
+      "step": 6350
+    },
+    {
+      "epoch": 90.00579710144928,
+      "grad_norm": 0.008534472435712814,
+      "learning_rate": 4.28341384863124e-06,
+      "loss": 0.0,
+      "step": 6360
+    },
+    {
+      "epoch": 90.00676328502415,
+      "grad_norm": 0.00027089714421890676,
+      "learning_rate": 4.2726784755770265e-06,
+      "loss": 0.0,
+      "step": 6370
+    },
+    {
+      "epoch": 90.00676328502415,
+      "eval_accuracy": 0.8076923076923077,
+      "eval_loss": 1.7776321172714233,
+      "eval_runtime": 17.3577,
+      "eval_samples_per_second": 4.494,
+      "eval_steps_per_second": 1.152,
+      "step": 6370
+    },
+    {
+      "epoch": 91.00096618357487,
+      "grad_norm": 0.0012238288763910532,
+      "learning_rate": 4.261943102522813e-06,
+      "loss": 0.0,
+      "step": 6380
+    },
+    {
+      "epoch": 91.00193236714976,
+      "grad_norm": 0.0015267673879861832,
+      "learning_rate": 4.251207729468599e-06,
+      "loss": 0.0,
+      "step": 6390
+    },
+    {
+      "epoch": 91.00289855072464,
+      "grad_norm": 0.0006313137710094452,
+      "learning_rate": 4.240472356414386e-06,
+      "loss": 0.0,
+      "step": 6400
+    },
+    {
+      "epoch": 91.00386473429951,
+      "grad_norm": 0.00024374033091589808,
+      "learning_rate": 4.229736983360172e-06,
+      "loss": 0.0,
+      "step": 6410
+    },
+    {
+      "epoch": 91.0048309178744,
+      "grad_norm": 0.000769898877479136,
+      "learning_rate": 4.219001610305958e-06,
+      "loss": 0.0,
+      "step": 6420
+    },
+    {
+      "epoch": 91.00579710144928,
+      "grad_norm": 0.0005884930142201483,
+      "learning_rate": 4.208266237251744e-06,
+      "loss": 0.0,
+      "step": 6430
+    },
+    {
+      "epoch": 91.00676328502415,
+      "grad_norm": 0.0004250435740686953,
+      "learning_rate": 4.197530864197531e-06,
+      "loss": 0.0028,
+      "step": 6440
+    },
+    {
+      "epoch": 91.00676328502415,
+      "eval_accuracy": 0.7948717948717948,
+      "eval_loss": 1.749398112297058,
+      "eval_runtime": 15.1757,
+      "eval_samples_per_second": 5.14,
+      "eval_steps_per_second": 1.318,
+      "step": 6440
+    },
+    {
+      "epoch": 92.00096618357487,
+      "grad_norm": 0.0007963450625538826,
+      "learning_rate": 4.186795491143318e-06,
+      "loss": 0.0,
+      "step": 6450
+    },
+    {
+      "epoch": 92.00193236714976,
+      "grad_norm": 0.000419673859141767,
+      "learning_rate": 4.1760601180891046e-06,
+      "loss": 0.0,
+      "step": 6460
+    },
+    {
+      "epoch": 92.00289855072464,
+      "grad_norm": 0.0020634315442293882,
+      "learning_rate": 4.16532474503489e-06,
+      "loss": 0.0,
+      "step": 6470
+    },
+    {
+      "epoch": 92.00386473429951,
+      "grad_norm": 0.002074671443551779,
+      "learning_rate": 4.154589371980677e-06,
+      "loss": 0.0,
+      "step": 6480
+    },
+    {
+      "epoch": 92.0048309178744,
+      "grad_norm": 0.0013138147769495845,
+      "learning_rate": 4.143853998926463e-06,
+      "loss": 0.0,
+      "step": 6490
+    },
+    {
+      "epoch": 92.00579710144928,
+      "grad_norm": 0.0006493937107734382,
+      "learning_rate": 4.1331186258722495e-06,
+      "loss": 0.0,
+      "step": 6500
+    },
+    {
+      "epoch": 92.00676328502415,
+      "grad_norm": 0.00048424158012494445,
+      "learning_rate": 4.122383252818036e-06,
+      "loss": 0.0,
+      "step": 6510
+    },
+    {
+      "epoch": 92.00676328502415,
+      "eval_accuracy": 0.8076923076923077,
+      "eval_loss": 1.614148736000061,
+      "eval_runtime": 16.0064,
+      "eval_samples_per_second": 4.873,
+      "eval_steps_per_second": 1.249,
+      "step": 6510
+    },
+    {
+      "epoch": 93.00096618357487,
+      "grad_norm": 0.0007483679219149053,
+      "learning_rate": 4.1116478797638224e-06,
+      "loss": 0.0,
+      "step": 6520
+    },
+    {
+      "epoch": 93.00193236714976,
+      "grad_norm": 0.0005844564293511212,
+      "learning_rate": 4.100912506709608e-06,
+      "loss": 0.0,
+      "step": 6530
+    },
+    {
+      "epoch": 93.00289855072464,
+      "grad_norm": 0.0004827914817724377,
+      "learning_rate": 4.0901771336553945e-06,
+      "loss": 0.0,
+      "step": 6540
+    },
+    {
+      "epoch": 93.00386473429951,
+      "grad_norm": 0.0021162095945328474,
+      "learning_rate": 4.079441760601181e-06,
+      "loss": 0.0619,
+      "step": 6550
+    },
+    {
+      "epoch": 93.0048309178744,
+      "grad_norm": 0.007160414941608906,
+      "learning_rate": 4.068706387546967e-06,
+      "loss": 0.0,
+      "step": 6560
+    },
+    {
+      "epoch": 93.00579710144928,
+      "grad_norm": 0.0008341994835063815,
+      "learning_rate": 4.057971014492754e-06,
+      "loss": 0.1666,
+      "step": 6570
+    },
+    {
+      "epoch": 93.00676328502415,
+      "grad_norm": 0.0007460052147507668,
+      "learning_rate": 4.04723564143854e-06,
+      "loss": 0.0,
+      "step": 6580
+    },
+    {
+      "epoch": 93.00676328502415,
+      "eval_accuracy": 0.717948717948718,
+      "eval_loss": 2.82431697845459,
+      "eval_runtime": 15.0775,
+      "eval_samples_per_second": 5.173,
+      "eval_steps_per_second": 1.326,
+      "step": 6580
+    },
+    {
+      "epoch": 94.00096618357487,
+      "grad_norm": 0.000210269630770199,
+      "learning_rate": 4.036500268384327e-06,
+      "loss": 0.0015,
+      "step": 6590
+    },
+    {
+      "epoch": 94.00193236714976,
+      "grad_norm": 0.00048051681369543076,
+      "learning_rate": 4.025764895330113e-06,
+      "loss": 0.1711,
+      "step": 6600
+    },
+    {
+      "epoch": 94.00289855072464,
+      "grad_norm": 148.72152709960938,
+      "learning_rate": 4.0150295222759e-06,
+      "loss": 0.0117,
+      "step": 6610
+    },
+    {
+      "epoch": 94.00386473429951,
+      "grad_norm": 0.0029819430783391,
+      "learning_rate": 4.004294149221686e-06,
+      "loss": 0.0,
+      "step": 6620
+    },
+    {
+      "epoch": 94.0048309178744,
+      "grad_norm": 0.0169463362544775,
+      "learning_rate": 3.9935587761674725e-06,
+      "loss": 0.0001,
+      "step": 6630
+    },
+    {
+      "epoch": 94.00579710144928,
+      "grad_norm": 0.0007479405030608177,
+      "learning_rate": 3.982823403113258e-06,
+      "loss": 0.2424,
+      "step": 6640
+    },
+    {
+      "epoch": 94.00676328502415,
+      "grad_norm": 0.0004274906241334975,
+      "learning_rate": 3.972088030059045e-06,
+      "loss": 0.0,
+      "step": 6650
+    },
+    {
+      "epoch": 94.00676328502415,
+      "eval_accuracy": 0.7051282051282052,
+      "eval_loss": 1.9025695323944092,
+      "eval_runtime": 14.9527,
+      "eval_samples_per_second": 5.216,
+      "eval_steps_per_second": 1.338,
+      "step": 6650
+    },
+    {
+      "epoch": 95.00096618357487,
+      "grad_norm": 0.002755245892331004,
+      "learning_rate": 3.961352657004831e-06,
+      "loss": 0.0,
+      "step": 6660
+    },
+    {
+      "epoch": 95.00193236714976,
+      "grad_norm": 0.0005277282907627523,
+      "learning_rate": 3.9506172839506175e-06,
+      "loss": 0.0,
+      "step": 6670
+    },
+    {
+      "epoch": 95.00289855072464,
+      "grad_norm": 0.0004030791751574725,
+      "learning_rate": 3.939881910896404e-06,
+      "loss": 0.0,
+      "step": 6680
+    },
+    {
+      "epoch": 95.00386473429951,
+      "grad_norm": 0.0007481653010472655,
+      "learning_rate": 3.92914653784219e-06,
+      "loss": 0.0,
+      "step": 6690
+    },
+    {
+      "epoch": 95.0048309178744,
+      "grad_norm": 0.003996316809207201,
+      "learning_rate": 3.918411164787977e-06,
+      "loss": 0.0,
+      "step": 6700
+    },
+    {
+      "epoch": 95.00579710144928,
+      "grad_norm": 0.0004657425743062049,
+      "learning_rate": 3.907675791733763e-06,
+      "loss": 0.0357,
+      "step": 6710
+    },
+    {
+      "epoch": 95.00676328502415,
+      "grad_norm": 0.0005334661109372973,
+      "learning_rate": 3.89694041867955e-06,
+      "loss": 0.0002,
+      "step": 6720
+    },
+    {
+      "epoch": 95.00676328502415,
+      "eval_accuracy": 0.6666666666666666,
+      "eval_loss": 2.236795663833618,
+      "eval_runtime": 14.9418,
+      "eval_samples_per_second": 5.22,
+      "eval_steps_per_second": 1.339,
+      "step": 6720
+    },
+    {
+      "epoch": 96.00096618357487,
+      "grad_norm": 0.000555448408704251,
+      "learning_rate": 3.886205045625336e-06,
+      "loss": 0.0,
+      "step": 6730
+    },
+    {
+      "epoch": 96.00193236714976,
+      "grad_norm": 0.22761359810829163,
+      "learning_rate": 3.875469672571122e-06,
+      "loss": 0.0009,
+      "step": 6740
+    },
+    {
+      "epoch": 96.00289855072464,
+      "grad_norm": 0.007025455124676228,
+      "learning_rate": 3.864734299516908e-06,
+      "loss": 0.0,
+      "step": 6750
+    },
+    {
+      "epoch": 96.00386473429951,
+      "grad_norm": 0.00039805268170312047,
+      "learning_rate": 3.853998926462695e-06,
+      "loss": 0.0,
+      "step": 6760
+    },
+    {
+      "epoch": 96.0048309178744,
+      "grad_norm": 0.00040781483403407037,
+      "learning_rate": 3.843263553408481e-06,
+      "loss": 0.0,
+      "step": 6770
+    },
+    {
+      "epoch": 96.00579710144928,
+      "grad_norm": 0.0007238932885229588,
+      "learning_rate": 3.832528180354268e-06,
+      "loss": 0.0,
+      "step": 6780
+    },
+    {
+      "epoch": 96.00676328502415,
+      "grad_norm": 0.0008674617856740952,
+      "learning_rate": 3.821792807300054e-06,
+      "loss": 0.0998,
+      "step": 6790
+    },
+    {
+      "epoch": 96.00676328502415,
+      "eval_accuracy": 0.717948717948718,
+      "eval_loss": 2.4984593391418457,
+      "eval_runtime": 16.4073,
+      "eval_samples_per_second": 4.754,
+      "eval_steps_per_second": 1.219,
+      "step": 6790
+    },
+    {
+      "epoch": 97.00096618357487,
+      "grad_norm": 0.0004405768122524023,
+      "learning_rate": 3.8110574342458405e-06,
+      "loss": 0.0,
+      "step": 6800
+    },
+    {
+      "epoch": 97.00193236714976,
+      "grad_norm": 0.001103090588003397,
+      "learning_rate": 3.800322061191627e-06,
+      "loss": 0.0,
+      "step": 6810
+    },
+    {
+      "epoch": 97.00289855072464,
+      "grad_norm": 0.0003414273087400943,
+      "learning_rate": 3.7895866881374134e-06,
+      "loss": 0.0,
+      "step": 6820
+    },
+    {
+      "epoch": 97.00386473429951,
+      "grad_norm": 0.00032865177490748465,
+      "learning_rate": 3.7788513150831994e-06,
+      "loss": 0.0,
+      "step": 6830
+    },
+    {
+      "epoch": 97.0048309178744,
+      "grad_norm": 0.0020416765473783016,
+      "learning_rate": 3.768115942028986e-06,
+      "loss": 0.0,
+      "step": 6840
+    },
+    {
+      "epoch": 97.00579710144928,
+      "grad_norm": 0.0008840957889333367,
+      "learning_rate": 3.7573805689747723e-06,
+      "loss": 0.0,
+      "step": 6850
+    },
+    {
+      "epoch": 97.00676328502415,
+      "grad_norm": 0.000713666551746428,
+      "learning_rate": 3.7466451959205584e-06,
+      "loss": 0.0,
+      "step": 6860
+    },
+    {
+      "epoch": 97.00676328502415,
+      "eval_accuracy": 0.7435897435897436,
+      "eval_loss": 2.0685932636260986,
+      "eval_runtime": 17.7738,
+      "eval_samples_per_second": 4.388,
+      "eval_steps_per_second": 1.125,
+      "step": 6860
+    },
+    {
+      "epoch": 98.00096618357487,
+      "grad_norm": 0.00042999107972718775,
+      "learning_rate": 3.735909822866345e-06,
+      "loss": 0.0,
+      "step": 6870
+    },
+    {
+      "epoch": 98.00193236714976,
+      "grad_norm": 0.0008358382037840784,
+      "learning_rate": 3.7251744498121313e-06,
+      "loss": 0.0,
+      "step": 6880
+    },
+    {
+      "epoch": 98.00289855072464,
+      "grad_norm": 0.0005103634321130812,
+      "learning_rate": 3.7144390767579173e-06,
+      "loss": 0.0,
+      "step": 6890
+    },
+    {
+      "epoch": 98.00386473429951,
+      "grad_norm": 0.0003126839583273977,
+      "learning_rate": 3.7037037037037037e-06,
+      "loss": 0.0,
+      "step": 6900
+    },
+    {
+      "epoch": 98.0048309178744,
+      "grad_norm": 0.00033128069480881095,
+      "learning_rate": 3.69296833064949e-06,
+      "loss": 0.0,
+      "step": 6910
+    },
+    {
+      "epoch": 98.00579710144928,
+      "grad_norm": 0.0004542351234704256,
+      "learning_rate": 3.682232957595277e-06,
+      "loss": 0.0002,
+      "step": 6920
+    },
+    {
+      "epoch": 98.00676328502415,
+      "grad_norm": 0.00022706468007527292,
+      "learning_rate": 3.6714975845410635e-06,
+      "loss": 0.0074,
+      "step": 6930
+    },
+    {
+      "epoch": 98.00676328502415,
+      "eval_accuracy": 0.7051282051282052,
+      "eval_loss": 2.7098388671875,
+      "eval_runtime": 22.2504,
+      "eval_samples_per_second": 3.506,
+      "eval_steps_per_second": 0.899,
+      "step": 6930
+    },
+    {
+      "epoch": 99.00096618357487,
+      "grad_norm": 0.0004110004520043731,
+      "learning_rate": 3.6607622114868495e-06,
+      "loss": 0.0,
+      "step": 6940
+    },
+    {
+      "epoch": 99.00193236714976,
+      "grad_norm": 0.0007272631046362221,
+      "learning_rate": 3.650026838432636e-06,
+      "loss": 0.0,
+      "step": 6950
+    },
+    {
+      "epoch": 99.00289855072464,
+      "grad_norm": 0.0016168636502698064,
+      "learning_rate": 3.6392914653784224e-06,
+      "loss": 0.0,
+      "step": 6960
+    },
+    {
+      "epoch": 99.00386473429951,
+      "grad_norm": 0.0007112635066732764,
+      "learning_rate": 3.6285560923242085e-06,
+      "loss": 0.0,
+      "step": 6970
+    },
+    {
+      "epoch": 99.0048309178744,
+      "grad_norm": 0.00035714227124117315,
+      "learning_rate": 3.617820719269995e-06,
+      "loss": 0.0,
+      "step": 6980
+    },
+    {
+      "epoch": 99.00579710144928,
+      "grad_norm": 0.0005834894836880267,
+      "learning_rate": 3.6070853462157814e-06,
+      "loss": 0.0,
+      "step": 6990
+    },
+    {
+      "epoch": 99.00676328502415,
+      "grad_norm": 0.0008813188760541379,
+      "learning_rate": 3.5963499731615674e-06,
+      "loss": 0.0,
+      "step": 7000
+    },
+    {
+      "epoch": 99.00676328502415,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 2.2764525413513184,
+      "eval_runtime": 20.108,
+      "eval_samples_per_second": 3.879,
+      "eval_steps_per_second": 0.995,
+      "step": 7000
+    },
+    {
+      "epoch": 100.00096618357487,
+      "grad_norm": 0.00035179706173948944,
+      "learning_rate": 3.585614600107354e-06,
+      "loss": 0.0,
+      "step": 7010
+    },
+    {
+      "epoch": 100.00193236714976,
+      "grad_norm": 0.0001833884307416156,
+      "learning_rate": 3.5748792270531403e-06,
+      "loss": 0.0,
+      "step": 7020
+    },
+    {
+      "epoch": 100.00289855072464,
+      "grad_norm": 0.0007354604313150048,
+      "learning_rate": 3.5641438539989263e-06,
+      "loss": 0.0,
+      "step": 7030
+    },
+    {
+      "epoch": 100.00386473429951,
+      "grad_norm": 0.0019487851532176137,
+      "learning_rate": 3.553408480944713e-06,
+      "loss": 0.0,
+      "step": 7040
+    },
+    {
+      "epoch": 100.0048309178744,
+      "grad_norm": 0.0004975392948836088,
+      "learning_rate": 3.5426731078904997e-06,
+      "loss": 0.0,
+      "step": 7050
+    },
+    {
+      "epoch": 100.00579710144928,
+      "grad_norm": 0.00027223333017900586,
+      "learning_rate": 3.531937734836286e-06,
+      "loss": 0.0,
+      "step": 7060
+    },
+    {
+      "epoch": 100.00676328502415,
+      "grad_norm": 0.0006046319613233209,
+      "learning_rate": 3.521202361782072e-06,
+      "loss": 0.0,
+      "step": 7070
+    },
+    {
+      "epoch": 100.00676328502415,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 2.2792937755584717,
+      "eval_runtime": 21.3442,
+      "eval_samples_per_second": 3.654,
+      "eval_steps_per_second": 0.937,
+      "step": 7070
+    },
+    {
+      "epoch": 101.00096618357487,
+      "grad_norm": 0.0004367881338112056,
+      "learning_rate": 3.5104669887278586e-06,
+      "loss": 0.0,
+      "step": 7080
+    },
+    {
+      "epoch": 101.00193236714976,
+      "grad_norm": 0.00014857234782539308,
+      "learning_rate": 3.499731615673645e-06,
+      "loss": 0.0,
+      "step": 7090
+    },
+    {
+      "epoch": 101.00289855072464,
+      "grad_norm": 0.000281750806607306,
+      "learning_rate": 3.488996242619431e-06,
+      "loss": 0.0,
+      "step": 7100
+    },
+    {
+      "epoch": 101.00386473429951,
+      "grad_norm": 0.0006434857496060431,
+      "learning_rate": 3.4782608695652175e-06,
+      "loss": 0.0,
+      "step": 7110
+    },
+    {
+      "epoch": 101.0048309178744,
+      "grad_norm": 0.0004700597492046654,
+      "learning_rate": 3.467525496511004e-06,
+      "loss": 0.0,
+      "step": 7120
+    },
+    {
+      "epoch": 101.00579710144928,
+      "grad_norm": 0.0005693004932254553,
+      "learning_rate": 3.4567901234567904e-06,
+      "loss": 0.0,
+      "step": 7130
+    },
+    {
+      "epoch": 101.00676328502415,
+      "grad_norm": 0.003832819638773799,
+      "learning_rate": 3.4460547504025764e-06,
+      "loss": 0.0,
+      "step": 7140
+    },
+    {
+      "epoch": 101.00676328502415,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 2.2027432918548584,
+      "eval_runtime": 19.1337,
+      "eval_samples_per_second": 4.077,
+      "eval_steps_per_second": 1.045,
+      "step": 7140
+    },
+    {
+      "epoch": 102.00096618357487,
+      "grad_norm": 0.0006276327185332775,
+      "learning_rate": 3.4353193773483633e-06,
+      "loss": 0.0,
+      "step": 7150
+    },
+    {
+      "epoch": 102.00193236714976,
+      "grad_norm": 0.0005548395565710962,
+      "learning_rate": 3.4245840042941498e-06,
+      "loss": 0.0,
+      "step": 7160
+    },
+    {
+      "epoch": 102.00289855072464,
+      "grad_norm": 0.0007787638460285962,
+      "learning_rate": 3.4138486312399362e-06,
+      "loss": 0.0,
+      "step": 7170
+    },
+    {
+      "epoch": 102.00386473429951,
+      "grad_norm": 0.0005334399756975472,
+      "learning_rate": 3.4031132581857222e-06,
+      "loss": 0.0,
+      "step": 7180
+    },
+    {
+      "epoch": 102.0048309178744,
+      "grad_norm": 0.0004569404118228704,
+      "learning_rate": 3.3923778851315087e-06,
+      "loss": 0.0,
+      "step": 7190
+    },
+    {
+      "epoch": 102.00579710144928,
+      "grad_norm": 0.00034970356500707567,
+      "learning_rate": 3.381642512077295e-06,
+      "loss": 0.0,
+      "step": 7200
+    },
+    {
+      "epoch": 102.00676328502415,
+      "grad_norm": 0.0002953490475192666,
+      "learning_rate": 3.370907139023081e-06,
+      "loss": 0.0,
+      "step": 7210
+    },
+    {
+      "epoch": 102.00676328502415,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 2.2387261390686035,
+      "eval_runtime": 19.3569,
+      "eval_samples_per_second": 4.03,
+      "eval_steps_per_second": 1.033,
+      "step": 7210
+    },
+    {
+      "epoch": 103.00096618357487,
+      "grad_norm": 0.0010686102323234081,
+      "learning_rate": 3.3601717659688676e-06,
+      "loss": 0.0,
+      "step": 7220
+    },
+    {
+      "epoch": 103.00193236714976,
+      "grad_norm": 0.00021685549290850759,
+      "learning_rate": 3.349436392914654e-06,
+      "loss": 0.0,
+      "step": 7230
+    },
+    {
+      "epoch": 103.00289855072464,
+      "grad_norm": 0.0007019720505923033,
+      "learning_rate": 3.33870101986044e-06,
+      "loss": 0.0815,
+      "step": 7240
+    },
+    {
+      "epoch": 103.00386473429951,
+      "grad_norm": 0.030306629836559296,
+      "learning_rate": 3.3279656468062265e-06,
+      "loss": 0.0,
+      "step": 7250
+    },
+    {
+      "epoch": 103.0048309178744,
+      "grad_norm": 0.0014363530790433288,
+      "learning_rate": 3.317230273752013e-06,
+      "loss": 0.1962,
+      "step": 7260
+    },
+    {
+      "epoch": 103.00579710144928,
+      "grad_norm": 0.00043242896208539605,
+      "learning_rate": 3.3064949006978e-06,
+      "loss": 0.0003,
+      "step": 7270
+    },
+    {
+      "epoch": 103.00676328502415,
+      "grad_norm": 0.0012811111519113183,
+      "learning_rate": 3.295759527643586e-06,
+      "loss": 0.0,
+      "step": 7280
+    },
+    {
+      "epoch": 103.00676328502415,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 2.197131395339966,
+      "eval_runtime": 72.0117,
+      "eval_samples_per_second": 1.083,
+      "eval_steps_per_second": 0.278,
+      "step": 7280
+    },
+    {
+      "epoch": 104.00096618357487,
+      "grad_norm": 0.000573929282836616,
+      "learning_rate": 3.2850241545893724e-06,
+      "loss": 0.0,
+      "step": 7290
+    },
+    {
+      "epoch": 104.00193236714976,
+      "grad_norm": 0.01300587598234415,
+      "learning_rate": 3.274288781535159e-06,
+      "loss": 0.0,
+      "step": 7300
+    },
+    {
+      "epoch": 104.00289855072464,
+      "grad_norm": 0.0005633372347801924,
+      "learning_rate": 3.263553408480945e-06,
+      "loss": 0.0,
+      "step": 7310
+    },
+    {
+      "epoch": 104.00386473429951,
+      "grad_norm": 0.0004255858657415956,
+      "learning_rate": 3.2528180354267313e-06,
+      "loss": 0.0,
+      "step": 7320
+    },
+    {
+      "epoch": 104.0048309178744,
+      "grad_norm": 0.000905704393517226,
+      "learning_rate": 3.2420826623725177e-06,
+      "loss": 0.0,
+      "step": 7330
+    },
+    {
+      "epoch": 104.00579710144928,
+      "grad_norm": 0.00022334480308927596,
+      "learning_rate": 3.231347289318304e-06,
+      "loss": 0.0,
+      "step": 7340
+    },
+    {
+      "epoch": 104.00676328502415,
+      "grad_norm": 0.0007362953037954867,
+      "learning_rate": 3.22061191626409e-06,
+      "loss": 0.0,
+      "step": 7350
+    },
+    {
+      "epoch": 104.00676328502415,
+      "eval_accuracy": 0.717948717948718,
+      "eval_loss": 2.324617862701416,
+      "eval_runtime": 21.2746,
+      "eval_samples_per_second": 3.666,
+      "eval_steps_per_second": 0.94,
+      "step": 7350
+    },
+    {
+      "epoch": 105.00096618357487,
+      "grad_norm": 0.0006282405229285359,
+      "learning_rate": 3.2098765432098767e-06,
+      "loss": 0.0,
+      "step": 7360
+    },
+    {
+      "epoch": 105.00193236714976,
+      "grad_norm": 0.00037624052492901683,
+      "learning_rate": 3.199141170155663e-06,
+      "loss": 0.0,
+      "step": 7370
+    },
+    {
+      "epoch": 105.00289855072464,
+      "grad_norm": 194.47848510742188,
+      "learning_rate": 3.188405797101449e-06,
+      "loss": 0.1884,
+      "step": 7380
+    },
+    {
+      "epoch": 105.00386473429951,
+      "grad_norm": 0.0003303911071270704,
+      "learning_rate": 3.177670424047236e-06,
+      "loss": 0.0002,
+      "step": 7390
+    },
+    {
+      "epoch": 105.0048309178744,
+      "grad_norm": 0.00046643143286928535,
+      "learning_rate": 3.1669350509930225e-06,
+      "loss": 0.0,
+      "step": 7400
+    },
+    {
+      "epoch": 105.00579710144928,
+      "grad_norm": 0.0017922611441463232,
+      "learning_rate": 3.156199677938809e-06,
+      "loss": 0.0054,
+      "step": 7410
+    },
+    {
+      "epoch": 105.00676328502415,
+      "grad_norm": 0.0007599503733217716,
+      "learning_rate": 3.145464304884595e-06,
+      "loss": 0.0,
+      "step": 7420
+    },
+    {
+      "epoch": 105.00676328502415,
+      "eval_accuracy": 0.8076923076923077,
+      "eval_loss": 1.5935262441635132,
+      "eval_runtime": 17.2874,
+      "eval_samples_per_second": 4.512,
+      "eval_steps_per_second": 1.157,
+      "step": 7420
+    },
+    {
+      "epoch": 106.00096618357487,
+      "grad_norm": 0.0019118065247312188,
+      "learning_rate": 3.1347289318303814e-06,
+      "loss": 0.0011,
+      "step": 7430
+    },
+    {
+      "epoch": 106.00193236714976,
+      "grad_norm": 0.002028007060289383,
+      "learning_rate": 3.123993558776168e-06,
+      "loss": 0.2234,
+      "step": 7440
+    },
+    {
+      "epoch": 106.00289855072464,
+      "grad_norm": 0.002168497536331415,
+      "learning_rate": 3.113258185721954e-06,
+      "loss": 0.0,
+      "step": 7450
+    },
+    {
+      "epoch": 106.00386473429951,
+      "grad_norm": 0.0005404739058576524,
+      "learning_rate": 3.1025228126677403e-06,
+      "loss": 0.0,
+      "step": 7460
+    },
+    {
+      "epoch": 106.0048309178744,
+      "grad_norm": 0.0015547404764220119,
+      "learning_rate": 3.0917874396135268e-06,
+      "loss": 0.0,
+      "step": 7470
+    },
+    {
+      "epoch": 106.00579710144928,
+      "grad_norm": 0.0010462005157023668,
+      "learning_rate": 3.081052066559313e-06,
+      "loss": 0.0,
+      "step": 7480
+    },
+    {
+      "epoch": 106.00676328502415,
+      "grad_norm": 0.001083563780412078,
+      "learning_rate": 3.0703166935050992e-06,
+      "loss": 0.0,
+      "step": 7490
+    },
+    {
+      "epoch": 106.00676328502415,
+      "eval_accuracy": 0.8076923076923077,
+      "eval_loss": 1.4796079397201538,
+      "eval_runtime": 20.0162,
+      "eval_samples_per_second": 3.897,
+      "eval_steps_per_second": 0.999,
+      "step": 7490
+    },
+    {
+      "epoch": 107.00096618357487,
+      "grad_norm": 0.0008712081471458077,
+      "learning_rate": 3.059581320450886e-06,
+      "loss": 0.0,
+      "step": 7500
+    },
+    {
+      "epoch": 107.00193236714976,
+      "grad_norm": 0.0006524740601889789,
+      "learning_rate": 3.0488459473966726e-06,
+      "loss": 0.0,
+      "step": 7510
+    },
+    {
+      "epoch": 107.00289855072464,
+      "grad_norm": 0.001063215546309948,
+      "learning_rate": 3.038110574342459e-06,
+      "loss": 0.0,
+      "step": 7520
+    },
+    {
+      "epoch": 107.00386473429951,
+      "grad_norm": 0.0003702947578858584,
+      "learning_rate": 3.027375201288245e-06,
+      "loss": 0.0,
+      "step": 7530
+    },
+    {
+      "epoch": 107.0048309178744,
+      "grad_norm": 0.001226250664331019,
+      "learning_rate": 3.0166398282340315e-06,
+      "loss": 0.0043,
+      "step": 7540
+    },
+    {
+      "epoch": 107.00579710144928,
+      "grad_norm": 0.0021951072849333286,
+      "learning_rate": 3.005904455179818e-06,
+      "loss": 0.0,
+      "step": 7550
+    },
+    {
+      "epoch": 107.00676328502415,
+      "grad_norm": 0.0006592991994693875,
+      "learning_rate": 2.995169082125604e-06,
+      "loss": 0.0001,
+      "step": 7560
+    },
+    {
+      "epoch": 107.00676328502415,
+      "eval_accuracy": 0.782051282051282,
+      "eval_loss": 1.7051782608032227,
+      "eval_runtime": 16.7621,
+      "eval_samples_per_second": 4.653,
+      "eval_steps_per_second": 1.193,
+      "step": 7560
+    },
+    {
+      "epoch": 108.00096618357487,
+      "grad_norm": 0.00048738584155216813,
+      "learning_rate": 2.9844337090713904e-06,
+      "loss": 0.0001,
+      "step": 7570
+    },
+    {
+      "epoch": 108.00193236714976,
+      "grad_norm": 0.0021665666718035936,
+      "learning_rate": 2.973698336017177e-06,
+      "loss": 0.0,
+      "step": 7580
+    },
+    {
+      "epoch": 108.00289855072464,
+      "grad_norm": 0.0005433819605968893,
+      "learning_rate": 2.962962962962963e-06,
+      "loss": 0.0,
+      "step": 7590
+    },
+    {
+      "epoch": 108.00386473429951,
+      "grad_norm": 0.0008435167255811393,
+      "learning_rate": 2.9522275899087494e-06,
+      "loss": 0.0,
+      "step": 7600
+    },
+    {
+      "epoch": 108.0048309178744,
+      "grad_norm": 0.00032446393743157387,
+      "learning_rate": 2.941492216854536e-06,
+      "loss": 0.0,
+      "step": 7610
+    },
+    {
+      "epoch": 108.00579710144928,
+      "grad_norm": 0.00036059069680050015,
+      "learning_rate": 2.9307568438003227e-06,
+      "loss": 0.0,
+      "step": 7620
+    },
+    {
+      "epoch": 108.00676328502415,
+      "grad_norm": 0.0009860590798780322,
+      "learning_rate": 2.9200214707461087e-06,
+      "loss": 0.0,
+      "step": 7630
+    },
+    {
+      "epoch": 108.00676328502415,
+      "eval_accuracy": 0.8076923076923077,
+      "eval_loss": 1.6021957397460938,
+      "eval_runtime": 21.0408,
+      "eval_samples_per_second": 3.707,
+      "eval_steps_per_second": 0.951,
+      "step": 7630
+    },
+    {
+      "epoch": 109.00096618357487,
+      "grad_norm": 0.0004186592123005539,
+      "learning_rate": 2.909286097691895e-06,
+      "loss": 0.0,
+      "step": 7640
+    },
+    {
+      "epoch": 109.00193236714976,
+      "grad_norm": 0.0013808052754029632,
+      "learning_rate": 2.8985507246376816e-06,
+      "loss": 0.0,
+      "step": 7650
+    },
+    {
+      "epoch": 109.00289855072464,
+      "grad_norm": 0.000451247877208516,
+      "learning_rate": 2.8878153515834676e-06,
+      "loss": 0.0,
+      "step": 7660
+    },
+    {
+      "epoch": 109.00386473429951,
+      "grad_norm": 0.0003690690209623426,
+      "learning_rate": 2.877079978529254e-06,
+      "loss": 0.0,
+      "step": 7670
+    },
+    {
+      "epoch": 109.0048309178744,
+      "grad_norm": 0.002147768158465624,
+      "learning_rate": 2.8663446054750405e-06,
+      "loss": 0.0,
+      "step": 7680
+    },
+    {
+      "epoch": 109.00579710144928,
+      "grad_norm": 0.00025449952227063477,
+      "learning_rate": 2.8556092324208266e-06,
+      "loss": 0.0,
+      "step": 7690
+    },
+    {
+      "epoch": 109.00676328502415,
+      "grad_norm": 2.826117753982544,
+      "learning_rate": 2.844873859366613e-06,
+      "loss": 0.0002,
+      "step": 7700
+    },
+    {
+      "epoch": 109.00676328502415,
+      "eval_accuracy": 0.8076923076923077,
+      "eval_loss": 1.674877643585205,
+      "eval_runtime": 16.7713,
+      "eval_samples_per_second": 4.651,
+      "eval_steps_per_second": 1.193,
+      "step": 7700
+    },
+    {
+      "epoch": 110.00096618357487,
+      "grad_norm": 0.0015585910296067595,
+      "learning_rate": 2.8341384863123995e-06,
+      "loss": 0.0,
+      "step": 7710
+    },
+    {
+      "epoch": 110.00193236714976,
+      "grad_norm": 0.001015316229313612,
+      "learning_rate": 2.823403113258186e-06,
+      "loss": 0.0,
+      "step": 7720
+    },
+    {
+      "epoch": 110.00289855072464,
+      "grad_norm": 0.00042512797517701983,
+      "learning_rate": 2.812667740203972e-06,
+      "loss": 0.0,
+      "step": 7730
+    },
+    {
+      "epoch": 110.00386473429951,
+      "grad_norm": 0.0018657109467312694,
+      "learning_rate": 2.801932367149759e-06,
+      "loss": 0.0,
+      "step": 7740
+    },
+    {
+      "epoch": 110.0048309178744,
+      "grad_norm": 0.0018489286303520203,
+      "learning_rate": 2.7911969940955453e-06,
+      "loss": 0.0,
+      "step": 7750
+    },
+    {
+      "epoch": 110.00579710144928,
+      "grad_norm": 0.0005694982828572392,
+      "learning_rate": 2.7804616210413317e-06,
+      "loss": 0.0,
+      "step": 7760
+    },
+    {
+      "epoch": 110.00676328502415,
+      "grad_norm": 0.0007545383996330202,
+      "learning_rate": 2.7697262479871177e-06,
+      "loss": 0.0,
+      "step": 7770
+    },
+    {
+      "epoch": 110.00676328502415,
+      "eval_accuracy": 0.7435897435897436,
+      "eval_loss": 1.7948025465011597,
+      "eval_runtime": 21.1507,
+      "eval_samples_per_second": 3.688,
+      "eval_steps_per_second": 0.946,
+      "step": 7770
+    },
+    {
+      "epoch": 111.00096618357487,
+      "grad_norm": 0.0008683151681907475,
+      "learning_rate": 2.758990874932904e-06,
+      "loss": 0.0,
+      "step": 7780
+    },
+    {
+      "epoch": 111.00193236714976,
+      "grad_norm": 0.0006877990090288222,
+      "learning_rate": 2.7482555018786907e-06,
+      "loss": 0.0,
+      "step": 7790
+    },
+    {
+      "epoch": 111.00289855072464,
+      "grad_norm": 0.000607641413807869,
+      "learning_rate": 2.7375201288244767e-06,
+      "loss": 0.0,
+      "step": 7800
+    },
+    {
+      "epoch": 111.00386473429951,
+      "grad_norm": 0.0011498964158818126,
+      "learning_rate": 2.726784755770263e-06,
+      "loss": 0.0,
+      "step": 7810
+    },
+    {
+      "epoch": 111.0048309178744,
+      "grad_norm": 0.0019288529874756932,
+      "learning_rate": 2.7160493827160496e-06,
+      "loss": 0.0,
+      "step": 7820
+    },
+    {
+      "epoch": 111.00579710144928,
+      "grad_norm": 0.0004927438567392528,
+      "learning_rate": 2.7053140096618356e-06,
+      "loss": 0.0,
+      "step": 7830
+    },
+    {
+      "epoch": 111.00676328502415,
+      "grad_norm": 0.0006650699651800096,
+      "learning_rate": 2.694578636607622e-06,
+      "loss": 0.0,
+      "step": 7840
+    },
+    {
+      "epoch": 111.00676328502415,
+      "eval_accuracy": 0.7948717948717948,
+      "eval_loss": 1.8455309867858887,
+      "eval_runtime": 20.0998,
+      "eval_samples_per_second": 3.881,
+      "eval_steps_per_second": 0.995,
+      "step": 7840
+    },
+    {
+      "epoch": 112.00096618357487,
+      "grad_norm": 0.0009932676330208778,
+      "learning_rate": 2.683843263553409e-06,
+      "loss": 0.0,
+      "step": 7850
+    },
+    {
+      "epoch": 112.00193236714976,
+      "grad_norm": 0.0004688102926593274,
+      "learning_rate": 2.6731078904991954e-06,
+      "loss": 0.0,
+      "step": 7860
+    },
+    {
+      "epoch": 112.00289855072464,
+      "grad_norm": 0.0006550211692228913,
+      "learning_rate": 2.6623725174449814e-06,
+      "loss": 0.0,
+      "step": 7870
+    },
+    {
+      "epoch": 112.00386473429951,
+      "grad_norm": 0.0024278289638459682,
+      "learning_rate": 2.651637144390768e-06,
+      "loss": 0.0,
+      "step": 7880
+    },
+    {
+      "epoch": 112.0048309178744,
+      "grad_norm": 0.0003039553703274578,
+      "learning_rate": 2.6409017713365543e-06,
+      "loss": 0.0,
+      "step": 7890
+    },
+    {
+      "epoch": 112.00579710144928,
+      "grad_norm": 0.0003849596541840583,
+      "learning_rate": 2.6301663982823408e-06,
+      "loss": 0.0,
+      "step": 7900
+    },
+    {
+      "epoch": 112.00676328502415,
+      "grad_norm": 0.002140798605978489,
+      "learning_rate": 2.6194310252281268e-06,
+      "loss": 0.0,
+      "step": 7910
+    },
+    {
+      "epoch": 112.00676328502415,
+      "eval_accuracy": 0.7948717948717948,
+      "eval_loss": 1.8600115776062012,
+      "eval_runtime": 16.4137,
+      "eval_samples_per_second": 4.752,
+      "eval_steps_per_second": 1.218,
+      "step": 7910
+    },
+    {
+      "epoch": 113.00096618357487,
+      "grad_norm": 0.0019010236719623208,
+      "learning_rate": 2.6086956521739132e-06,
+      "loss": 0.0,
+      "step": 7920
+    },
+    {
+      "epoch": 113.00193236714976,
+      "grad_norm": 0.00038240986759774387,
+      "learning_rate": 2.5979602791196997e-06,
+      "loss": 0.0,
+      "step": 7930
+    },
+    {
+      "epoch": 113.00289855072464,
+      "grad_norm": 0.0014828953426331282,
+      "learning_rate": 2.5872249060654857e-06,
+      "loss": 0.0,
+      "step": 7940
+    },
+    {
+      "epoch": 113.00386473429951,
+      "grad_norm": 0.00037424080073833466,
+      "learning_rate": 2.576489533011272e-06,
+      "loss": 0.0,
+      "step": 7950
+    },
+    {
+      "epoch": 113.0048309178744,
+      "grad_norm": 0.0005094591761007905,
+      "learning_rate": 2.5657541599570586e-06,
+      "loss": 0.0,
+      "step": 7960
+    },
+    {
+      "epoch": 113.00579710144928,
+      "grad_norm": 0.0018624026561155915,
+      "learning_rate": 2.5550187869028455e-06,
+      "loss": 0.0,
+      "step": 7970
+    },
+    {
+      "epoch": 113.00676328502415,
+      "grad_norm": 0.00043899298179894686,
+      "learning_rate": 2.5442834138486315e-06,
+      "loss": 0.0,
+      "step": 7980
+    },
+    {
+      "epoch": 113.00676328502415,
+      "eval_accuracy": 0.7948717948717948,
+      "eval_loss": 1.8182809352874756,
+      "eval_runtime": 16.1873,
+      "eval_samples_per_second": 4.819,
+      "eval_steps_per_second": 1.236,
+      "step": 7980
+    },
+    {
+      "epoch": 114.00096618357487,
+      "grad_norm": 0.0022914118599146605,
+      "learning_rate": 2.533548040794418e-06,
+      "loss": 0.0,
+      "step": 7990
+    },
+    {
+      "epoch": 114.00193236714976,
+      "grad_norm": 0.0003928690275643021,
+      "learning_rate": 2.5228126677402044e-06,
+      "loss": 0.0,
+      "step": 8000
+    },
+    {
+      "epoch": 114.00289855072464,
+      "grad_norm": 0.0008931594784371555,
+      "learning_rate": 2.5120772946859904e-06,
+      "loss": 0.0,
+      "step": 8010
+    },
+    {
+      "epoch": 114.00386473429951,
+      "grad_norm": 0.00023587950272485614,
+      "learning_rate": 2.501341921631777e-06,
+      "loss": 0.0,
+      "step": 8020
+    },
+    {
+      "epoch": 114.0048309178744,
+      "grad_norm": 0.0005247633089311421,
+      "learning_rate": 2.4906065485775633e-06,
+      "loss": 0.0,
+      "step": 8030
+    },
+    {
+      "epoch": 114.00579710144928,
+      "grad_norm": 0.00018518311844673008,
+      "learning_rate": 2.4798711755233494e-06,
+      "loss": 0.0,
+      "step": 8040
+    },
+    {
+      "epoch": 114.00676328502415,
+      "grad_norm": 0.0007796736899763346,
+      "learning_rate": 2.469135802469136e-06,
+      "loss": 0.0,
+      "step": 8050
+    },
+    {
+      "epoch": 114.00676328502415,
+      "eval_accuracy": 0.7948717948717948,
+      "eval_loss": 1.7861602306365967,
+      "eval_runtime": 16.6133,
+      "eval_samples_per_second": 4.695,
+      "eval_steps_per_second": 1.204,
+      "step": 8050
+    },
+    {
+      "epoch": 115.00096618357487,
+      "grad_norm": 0.00030822984990663826,
+      "learning_rate": 2.4584004294149223e-06,
+      "loss": 0.0,
+      "step": 8060
+    },
+    {
+      "epoch": 115.00193236714976,
+      "grad_norm": 3.4960391521453857,
+      "learning_rate": 2.4476650563607087e-06,
+      "loss": 0.0001,
+      "step": 8070
+    },
+    {
+      "epoch": 115.00289855072464,
+      "grad_norm": 0.00040909554809331894,
+      "learning_rate": 2.436929683306495e-06,
+      "loss": 0.2293,
+      "step": 8080
+    },
+    {
+      "epoch": 115.00386473429951,
+      "grad_norm": 0.0004301641311030835,
+      "learning_rate": 2.426194310252281e-06,
+      "loss": 0.0,
+      "step": 8090
+    },
+    {
+      "epoch": 115.0048309178744,
+      "grad_norm": 0.00037556581082753837,
+      "learning_rate": 2.4154589371980677e-06,
+      "loss": 0.0,
+      "step": 8100
+    },
+    {
+      "epoch": 115.00579710144928,
+      "grad_norm": 0.00048973155207932,
+      "learning_rate": 2.4047235641438545e-06,
+      "loss": 0.0,
+      "step": 8110
+    },
+    {
+      "epoch": 115.00676328502415,
+      "grad_norm": 0.00013631054025609046,
+      "learning_rate": 2.3939881910896406e-06,
+      "loss": 0.0,
+      "step": 8120
+    },
+    {
+      "epoch": 115.00676328502415,
+      "eval_accuracy": 0.782051282051282,
+      "eval_loss": 1.8597110509872437,
+      "eval_runtime": 16.1665,
+      "eval_samples_per_second": 4.825,
+      "eval_steps_per_second": 1.237,
+      "step": 8120
+    },
+    {
+      "epoch": 116.00096618357487,
+      "grad_norm": 0.00032639241544529796,
+      "learning_rate": 2.383252818035427e-06,
+      "loss": 0.0,
+      "step": 8130
+    },
+    {
+      "epoch": 116.00193236714976,
+      "grad_norm": 0.000251809018664062,
+      "learning_rate": 2.3725174449812135e-06,
+      "loss": 0.0,
+      "step": 8140
+    },
+    {
+      "epoch": 116.00289855072464,
+      "grad_norm": 0.00035566542646847665,
+      "learning_rate": 2.3617820719269995e-06,
+      "loss": 0.0,
+      "step": 8150
+    },
+    {
+      "epoch": 116.00386473429951,
+      "grad_norm": 0.00013090217544231564,
+      "learning_rate": 2.351046698872786e-06,
+      "loss": 0.0,
+      "step": 8160
+    },
+    {
+      "epoch": 116.0048309178744,
+      "grad_norm": 0.0003571859560906887,
+      "learning_rate": 2.3403113258185724e-06,
+      "loss": 0.0,
+      "step": 8170
+    },
+    {
+      "epoch": 116.00579710144928,
+      "grad_norm": 0.0007861918420530856,
+      "learning_rate": 2.329575952764359e-06,
+      "loss": 0.0,
+      "step": 8180
+    },
+    {
+      "epoch": 116.00676328502415,
+      "grad_norm": 0.00013054760347586125,
+      "learning_rate": 2.3188405797101453e-06,
+      "loss": 0.0,
+      "step": 8190
+    },
+    {
+      "epoch": 116.00676328502415,
+      "eval_accuracy": 0.782051282051282,
+      "eval_loss": 1.820298433303833,
+      "eval_runtime": 19.7868,
+      "eval_samples_per_second": 3.942,
+      "eval_steps_per_second": 1.011,
+      "step": 8190
+    },
+    {
+      "epoch": 117.00096618357487,
+      "grad_norm": 0.0005784878740087152,
+      "learning_rate": 2.3081052066559313e-06,
+      "loss": 0.0,
+      "step": 8200
+    },
+    {
+      "epoch": 117.00193236714976,
+      "grad_norm": 0.0004840072651859373,
+      "learning_rate": 2.2973698336017178e-06,
+      "loss": 0.0,
+      "step": 8210
+    },
+    {
+      "epoch": 117.00289855072464,
+      "grad_norm": 0.0005657835863530636,
+      "learning_rate": 2.2866344605475042e-06,
+      "loss": 0.0,
+      "step": 8220
+    },
+    {
+      "epoch": 117.00386473429951,
+      "grad_norm": 0.0003718466032296419,
+      "learning_rate": 2.2758990874932907e-06,
+      "loss": 0.0,
+      "step": 8230
+    },
+    {
+      "epoch": 117.0048309178744,
+      "grad_norm": 0.0004277382104191929,
+      "learning_rate": 2.265163714439077e-06,
+      "loss": 0.0,
+      "step": 8240
+    },
+    {
+      "epoch": 117.00579710144928,
+      "grad_norm": 0.005644855089485645,
+      "learning_rate": 2.254428341384863e-06,
+      "loss": 0.0,
+      "step": 8250
+    },
+    {
+      "epoch": 117.00676328502415,
+      "grad_norm": 0.00029681739397346973,
+      "learning_rate": 2.2436929683306496e-06,
+      "loss": 0.0,
+      "step": 8260
+    },
+    {
+      "epoch": 117.00676328502415,
+      "eval_accuracy": 0.7948717948717948,
+      "eval_loss": 1.8342883586883545,
+      "eval_runtime": 17.1194,
+      "eval_samples_per_second": 4.556,
+      "eval_steps_per_second": 1.168,
+      "step": 8260
+    },
+    {
+      "epoch": 118.00096618357487,
+      "grad_norm": 0.00037026609061285853,
+      "learning_rate": 2.232957595276436e-06,
+      "loss": 0.0,
+      "step": 8270
+    },
+    {
+      "epoch": 118.00193236714976,
+      "grad_norm": 0.00031471162219531834,
+      "learning_rate": 2.222222222222222e-06,
+      "loss": 0.0,
+      "step": 8280
+    },
+    {
+      "epoch": 118.00289855072464,
+      "grad_norm": 0.00044593116035684943,
+      "learning_rate": 2.211486849168009e-06,
+      "loss": 0.0,
+      "step": 8290
+    },
+    {
+      "epoch": 118.00386473429951,
+      "grad_norm": 0.001200909842737019,
+      "learning_rate": 2.2007514761137954e-06,
+      "loss": 0.0,
+      "step": 8300
+    },
+    {
+      "epoch": 118.0048309178744,
+      "grad_norm": 0.00038560116081498563,
+      "learning_rate": 2.1900161030595814e-06,
+      "loss": 0.0,
+      "step": 8310
+    },
+    {
+      "epoch": 118.00579710144928,
+      "grad_norm": 0.0014041089452803135,
+      "learning_rate": 2.179280730005368e-06,
+      "loss": 0.0,
+      "step": 8320
+    },
+    {
+      "epoch": 118.00676328502415,
+      "grad_norm": 0.0004680801066569984,
+      "learning_rate": 2.1685453569511543e-06,
+      "loss": 0.0,
+      "step": 8330
+    },
+    {
+      "epoch": 118.00676328502415,
+      "eval_accuracy": 0.7948717948717948,
+      "eval_loss": 1.8416794538497925,
+      "eval_runtime": 16.87,
+      "eval_samples_per_second": 4.624,
+      "eval_steps_per_second": 1.186,
+      "step": 8330
+    },
+    {
+      "epoch": 119.00096618357487,
+      "grad_norm": 0.0004902587970718741,
+      "learning_rate": 2.1578099838969404e-06,
+      "loss": 0.0,
+      "step": 8340
+    },
+    {
+      "epoch": 119.00193236714976,
+      "grad_norm": 91.31414794921875,
+      "learning_rate": 2.1470746108427272e-06,
+      "loss": 0.2185,
+      "step": 8350
+    },
+    {
+      "epoch": 119.00289855072464,
+      "grad_norm": 0.00014069379540160298,
+      "learning_rate": 2.1363392377885133e-06,
+      "loss": 0.0,
+      "step": 8360
+    },
+    {
+      "epoch": 119.00386473429951,
+      "grad_norm": 0.000278379739029333,
+      "learning_rate": 2.1256038647342997e-06,
+      "loss": 0.0,
+      "step": 8370
+    },
+    {
+      "epoch": 119.0048309178744,
+      "grad_norm": 0.019783247262239456,
+      "learning_rate": 2.114868491680086e-06,
+      "loss": 0.0,
+      "step": 8380
+    },
+    {
+      "epoch": 119.00579710144928,
+      "grad_norm": 0.002190082333981991,
+      "learning_rate": 2.104133118625872e-06,
+      "loss": 0.0617,
+      "step": 8390
+    },
+    {
+      "epoch": 119.00676328502415,
+      "grad_norm": 0.0004942225059494376,
+      "learning_rate": 2.093397745571659e-06,
+      "loss": 0.0,
+      "step": 8400
+    },
+    {
+      "epoch": 119.00676328502415,
+      "eval_accuracy": 0.782051282051282,
+      "eval_loss": 1.7663315534591675,
+      "eval_runtime": 20.9495,
+      "eval_samples_per_second": 3.723,
+      "eval_steps_per_second": 0.955,
+      "step": 8400
+    },
+    {
+      "epoch": 120.00096618357487,
+      "grad_norm": 0.0003238871868234128,
+      "learning_rate": 2.082662372517445e-06,
+      "loss": 0.0,
+      "step": 8410
+    },
+    {
+      "epoch": 120.00193236714976,
+      "grad_norm": 0.00031754918745718896,
+      "learning_rate": 2.0719269994632315e-06,
+      "loss": 0.0,
+      "step": 8420
+    },
+    {
+      "epoch": 120.00289855072464,
+      "grad_norm": 0.0003926875942852348,
+      "learning_rate": 2.061191626409018e-06,
+      "loss": 0.0,
+      "step": 8430
+    },
+    {
+      "epoch": 120.00386473429951,
+      "grad_norm": 0.0003822804428637028,
+      "learning_rate": 2.050456253354804e-06,
+      "loss": 0.0,
+      "step": 8440
+    },
+    {
+      "epoch": 120.0048309178744,
+      "grad_norm": 0.0006702033570036292,
+      "learning_rate": 2.0397208803005905e-06,
+      "loss": 0.0,
+      "step": 8450
+    },
+    {
+      "epoch": 120.00579710144928,
+      "grad_norm": 0.0023265674244612455,
+      "learning_rate": 2.028985507246377e-06,
+      "loss": 0.0,
+      "step": 8460
+    },
+    {
+      "epoch": 120.00676328502415,
+      "grad_norm": 0.0004378359590191394,
+      "learning_rate": 2.0182501341921634e-06,
+      "loss": 0.0,
+      "step": 8470
+    },
+    {
+      "epoch": 120.00676328502415,
+      "eval_accuracy": 0.782051282051282,
+      "eval_loss": 1.9610668420791626,
+      "eval_runtime": 23.002,
+      "eval_samples_per_second": 3.391,
+      "eval_steps_per_second": 0.869,
+      "step": 8470
+    },
+    {
+      "epoch": 121.00096618357487,
+      "grad_norm": 0.0014331090496852994,
+      "learning_rate": 2.00751476113795e-06,
+      "loss": 0.0,
+      "step": 8480
+    },
+    {
+      "epoch": 121.00193236714976,
+      "grad_norm": 0.0001426087401341647,
+      "learning_rate": 1.9967793880837363e-06,
+      "loss": 0.0,
+      "step": 8490
+    },
+    {
+      "epoch": 121.00289855072464,
+      "grad_norm": 0.00037875105044804513,
+      "learning_rate": 1.9860440150295223e-06,
+      "loss": 0.0,
+      "step": 8500
+    },
+    {
+      "epoch": 121.00386473429951,
+      "grad_norm": 0.0009291688329540193,
+      "learning_rate": 1.9753086419753087e-06,
+      "loss": 0.0,
+      "step": 8510
+    },
+    {
+      "epoch": 121.0048309178744,
+      "grad_norm": 0.0003902267199009657,
+      "learning_rate": 1.964573268921095e-06,
+      "loss": 0.0,
+      "step": 8520
+    },
+    {
+      "epoch": 121.00579710144928,
+      "grad_norm": 0.00017865381960291415,
+      "learning_rate": 1.9538378958668816e-06,
+      "loss": 0.0,
+      "step": 8530
+    },
+    {
+      "epoch": 121.00676328502415,
+      "grad_norm": 0.0015924603212624788,
+      "learning_rate": 1.943102522812668e-06,
+      "loss": 0.0,
+      "step": 8540
+    },
+    {
+      "epoch": 121.00676328502415,
+      "eval_accuracy": 0.7948717948717948,
+      "eval_loss": 1.9584029912948608,
+      "eval_runtime": 22.508,
+      "eval_samples_per_second": 3.465,
+      "eval_steps_per_second": 0.889,
+      "step": 8540
+    },
+    {
+      "epoch": 122.00096618357487,
+      "grad_norm": 0.0007341349264606833,
+      "learning_rate": 1.932367149758454e-06,
+      "loss": 0.0,
+      "step": 8550
+    },
+    {
+      "epoch": 122.00193236714976,
+      "grad_norm": 0.0006165707600302994,
+      "learning_rate": 1.9216317767042406e-06,
+      "loss": 0.0,
+      "step": 8560
+    },
+    {
+      "epoch": 122.00289855072464,
+      "grad_norm": 0.0003926469071302563,
+      "learning_rate": 1.910896403650027e-06,
+      "loss": 0.0015,
+      "step": 8570
+    },
+    {
+      "epoch": 122.00386473429951,
+      "grad_norm": 0.0003498373262118548,
+      "learning_rate": 1.9001610305958135e-06,
+      "loss": 0.0,
+      "step": 8580
+    },
+    {
+      "epoch": 122.0048309178744,
+      "grad_norm": 0.0002549053751863539,
+      "learning_rate": 1.8894256575415997e-06,
+      "loss": 0.0,
+      "step": 8590
+    },
+    {
+      "epoch": 122.00579710144928,
+      "grad_norm": 0.008167664520442486,
+      "learning_rate": 1.8786902844873862e-06,
+      "loss": 0.0,
+      "step": 8600
+    },
+    {
+      "epoch": 122.00676328502415,
+      "grad_norm": 0.00036383105907589197,
+      "learning_rate": 1.8679549114331724e-06,
+      "loss": 0.0,
+      "step": 8610
+    },
+    {
+      "epoch": 122.00676328502415,
+      "eval_accuracy": 0.8205128205128205,
+      "eval_loss": 1.5670628547668457,
+      "eval_runtime": 20.8607,
+      "eval_samples_per_second": 3.739,
+      "eval_steps_per_second": 0.959,
+      "step": 8610
+    },
+    {
+      "epoch": 123.00096618357487,
+      "grad_norm": 0.0004942956147715449,
+      "learning_rate": 1.8572195383789586e-06,
+      "loss": 0.0,
+      "step": 8620
+    },
+    {
+      "epoch": 123.00193236714976,
+      "grad_norm": 0.0004234071238897741,
+      "learning_rate": 1.846484165324745e-06,
+      "loss": 0.0004,
+      "step": 8630
+    },
+    {
+      "epoch": 123.00289855072464,
+      "grad_norm": 0.0005698536988347769,
+      "learning_rate": 1.8357487922705318e-06,
+      "loss": 0.0,
+      "step": 8640
+    },
+    {
+      "epoch": 123.00386473429951,
+      "grad_norm": 0.0009004671010188758,
+      "learning_rate": 1.825013419216318e-06,
+      "loss": 0.0,
+      "step": 8650
+    },
+    {
+      "epoch": 123.0048309178744,
+      "grad_norm": 0.000181853465619497,
+      "learning_rate": 1.8142780461621042e-06,
+      "loss": 0.0,
+      "step": 8660
+    },
+    {
+      "epoch": 123.00579710144928,
+      "grad_norm": 0.0006418362027034163,
+      "learning_rate": 1.8035426731078907e-06,
+      "loss": 0.0,
+      "step": 8670
+    },
+    {
+      "epoch": 123.00676328502415,
+      "grad_norm": 0.00031043903436511755,
+      "learning_rate": 1.792807300053677e-06,
+      "loss": 0.0,
+      "step": 8680
+    },
+    {
+      "epoch": 123.00676328502415,
+      "eval_accuracy": 0.7564102564102564,
+      "eval_loss": 2.345607042312622,
+      "eval_runtime": 24.4869,
+      "eval_samples_per_second": 3.185,
+      "eval_steps_per_second": 0.817,
+      "step": 8680
+    },
+    {
+      "epoch": 124.00096618357487,
+      "grad_norm": 0.00024151946126949042,
+      "learning_rate": 1.7820719269994632e-06,
+      "loss": 0.0,
+      "step": 8690
+    },
+    {
+      "epoch": 124.00193236714976,
+      "grad_norm": 0.00013747204502578825,
+      "learning_rate": 1.7713365539452498e-06,
+      "loss": 0.0,
+      "step": 8700
+    },
+    {
+      "epoch": 124.00289855072464,
+      "grad_norm": 0.0007334667025133967,
+      "learning_rate": 1.760601180891036e-06,
+      "loss": 0.0,
+      "step": 8710
+    },
+    {
+      "epoch": 124.00386473429951,
+      "grad_norm": 0.0001992643519770354,
+      "learning_rate": 1.7498658078368225e-06,
+      "loss": 0.0,
+      "step": 8720
+    },
+    {
+      "epoch": 124.0048309178744,
+      "grad_norm": 0.007898527197539806,
+      "learning_rate": 1.7391304347826088e-06,
+      "loss": 0.0,
+      "step": 8730
+    },
+    {
+      "epoch": 124.00579710144928,
+      "grad_norm": 0.00034817663254216313,
+      "learning_rate": 1.7283950617283952e-06,
+      "loss": 0.0,
+      "step": 8740
+    },
+    {
+      "epoch": 124.00676328502415,
+      "grad_norm": 0.0022646754514425993,
+      "learning_rate": 1.7176596886741817e-06,
+      "loss": 0.0,
+      "step": 8750
+    },
+    {
+      "epoch": 124.00676328502415,
+      "eval_accuracy": 0.7564102564102564,
+      "eval_loss": 2.34527587890625,
+      "eval_runtime": 17.9708,
+      "eval_samples_per_second": 4.34,
+      "eval_steps_per_second": 1.113,
+      "step": 8750
+    },
+    {
+      "epoch": 125.00096618357487,
+      "grad_norm": 0.0003508002555463463,
+      "learning_rate": 1.7069243156199681e-06,
+      "loss": 0.0,
+      "step": 8760
+    },
+    {
+      "epoch": 125.00193236714976,
+      "grad_norm": 0.00030854917713440955,
+      "learning_rate": 1.6961889425657543e-06,
+      "loss": 0.0,
+      "step": 8770
+    },
+    {
+      "epoch": 125.00289855072464,
+      "grad_norm": 0.0007584840641357005,
+      "learning_rate": 1.6854535695115406e-06,
+      "loss": 0.0,
+      "step": 8780
+    },
+    {
+      "epoch": 125.00386473429951,
+      "grad_norm": 0.0006423608283512294,
+      "learning_rate": 1.674718196457327e-06,
+      "loss": 0.0,
+      "step": 8790
+    },
+    {
+      "epoch": 125.0048309178744,
+      "grad_norm": 0.00041385908843949437,
+      "learning_rate": 1.6639828234031133e-06,
+      "loss": 0.0,
+      "step": 8800
+    },
+    {
+      "epoch": 125.00579710144928,
+      "grad_norm": 0.0002265427610836923,
+      "learning_rate": 1.6532474503489e-06,
+      "loss": 0.0,
+      "step": 8810
+    },
+    {
+      "epoch": 125.00676328502415,
+      "grad_norm": 0.00011521170381456614,
+      "learning_rate": 1.6425120772946862e-06,
+      "loss": 0.0,
+      "step": 8820
+    },
+    {
+      "epoch": 125.00676328502415,
+      "eval_accuracy": 0.7435897435897436,
+      "eval_loss": 2.4119598865509033,
+      "eval_runtime": 19.241,
+      "eval_samples_per_second": 4.054,
+      "eval_steps_per_second": 1.039,
+      "step": 8820
+    },
+    {
+      "epoch": 126.00096618357487,
+      "grad_norm": 0.00031076266895979643,
+      "learning_rate": 1.6317767042404724e-06,
+      "loss": 0.0,
+      "step": 8830
+    },
+    {
+      "epoch": 126.00193236714976,
+      "grad_norm": 0.0004146753635723144,
+      "learning_rate": 1.6210413311862589e-06,
+      "loss": 0.0,
+      "step": 8840
+    },
+    {
+      "epoch": 126.00289855072464,
+      "grad_norm": 0.00010951059084618464,
+      "learning_rate": 1.610305958132045e-06,
+      "loss": 0.0,
+      "step": 8850
+    },
+    {
+      "epoch": 126.00386473429951,
+      "grad_norm": 0.0006361532141454518,
+      "learning_rate": 1.5995705850778316e-06,
+      "loss": 0.0,
+      "step": 8860
+    },
+    {
+      "epoch": 126.0048309178744,
+      "grad_norm": 0.00019192650506738573,
+      "learning_rate": 1.588835212023618e-06,
+      "loss": 0.0,
+      "step": 8870
+    },
+    {
+      "epoch": 126.00579710144928,
+      "grad_norm": 0.00010023318463936448,
+      "learning_rate": 1.5780998389694045e-06,
+      "loss": 0.0,
+      "step": 8880
+    },
+    {
+      "epoch": 126.00676328502415,
+      "grad_norm": 0.00033185991924256086,
+      "learning_rate": 1.5673644659151907e-06,
+      "loss": 0.0,
+      "step": 8890
+    },
+    {
+      "epoch": 126.00676328502415,
+      "eval_accuracy": 0.7435897435897436,
+      "eval_loss": 2.3774123191833496,
+      "eval_runtime": 20.0988,
+      "eval_samples_per_second": 3.881,
+      "eval_steps_per_second": 0.995,
+      "step": 8890
+    },
+    {
+      "epoch": 127.00096618357487,
+      "grad_norm": 0.00048786698607727885,
+      "learning_rate": 1.556629092860977e-06,
+      "loss": 0.0,
+      "step": 8900
+    },
+    {
+      "epoch": 127.00193236714976,
+      "grad_norm": 0.0003051810781471431,
+      "learning_rate": 1.5458937198067634e-06,
+      "loss": 0.0,
+      "step": 8910
+    },
+    {
+      "epoch": 127.00289855072464,
+      "grad_norm": 0.0003112946287728846,
+      "learning_rate": 1.5351583467525496e-06,
+      "loss": 0.0,
+      "step": 8920
+    },
+    {
+      "epoch": 127.00386473429951,
+      "grad_norm": 0.00031099646002985537,
+      "learning_rate": 1.5244229736983363e-06,
+      "loss": 0.0,
+      "step": 8930
+    },
+    {
+      "epoch": 127.0048309178744,
+      "grad_norm": 0.0002164303296012804,
+      "learning_rate": 1.5136876006441225e-06,
+      "loss": 0.0,
+      "step": 8940
+    },
+    {
+      "epoch": 127.00579710144928,
+      "grad_norm": 0.0003160216729156673,
+      "learning_rate": 1.502952227589909e-06,
+      "loss": 0.0,
+      "step": 8950
+    },
+    {
+      "epoch": 127.00676328502415,
+      "grad_norm": 0.00017373952141497284,
+      "learning_rate": 1.4922168545356952e-06,
+      "loss": 0.0,
+      "step": 8960
+    },
+    {
+      "epoch": 127.00676328502415,
+      "eval_accuracy": 0.7435897435897436,
+      "eval_loss": 2.3609466552734375,
+      "eval_runtime": 23.3283,
+      "eval_samples_per_second": 3.344,
+      "eval_steps_per_second": 0.857,
+      "step": 8960
+    },
+    {
+      "epoch": 128.0009661835749,
+      "grad_norm": 0.0002286910021211952,
+      "learning_rate": 1.4814814814814815e-06,
+      "loss": 0.0,
+      "step": 8970
+    },
+    {
+      "epoch": 128.00193236714975,
+      "grad_norm": 0.0015229767886921763,
+      "learning_rate": 1.470746108427268e-06,
+      "loss": 0.0,
+      "step": 8980
+    },
+    {
+      "epoch": 128.00289855072464,
+      "grad_norm": 0.00010285177268087864,
+      "learning_rate": 1.4600107353730544e-06,
+      "loss": 0.0,
+      "step": 8990
+    },
+    {
+      "epoch": 128.00386473429953,
+      "grad_norm": 0.00026322112535126507,
+      "learning_rate": 1.4492753623188408e-06,
+      "loss": 0.0,
+      "step": 9000
+    },
+    {
+      "epoch": 128.0048309178744,
+      "grad_norm": 0.00018546039063949138,
+      "learning_rate": 1.438539989264627e-06,
+      "loss": 0.0,
+      "step": 9010
+    },
+    {
+      "epoch": 128.00579710144928,
+      "grad_norm": 0.00019054964650422335,
+      "learning_rate": 1.4278046162104133e-06,
+      "loss": 0.0,
+      "step": 9020
+    },
+    {
+      "epoch": 128.00676328502416,
+      "grad_norm": 0.0009014433017000556,
+      "learning_rate": 1.4170692431561997e-06,
+      "loss": 0.0,
+      "step": 9030
+    },
+    {
+      "epoch": 128.00676328502416,
+      "eval_accuracy": 0.7564102564102564,
+      "eval_loss": 2.353100061416626,
+      "eval_runtime": 19.8354,
+      "eval_samples_per_second": 3.932,
+      "eval_steps_per_second": 1.008,
+      "step": 9030
+    },
+    {
+      "epoch": 129.0009661835749,
+      "grad_norm": 0.0002870812895707786,
+      "learning_rate": 1.406333870101986e-06,
+      "loss": 0.0,
+      "step": 9040
+    },
+    {
+      "epoch": 129.00193236714975,
+      "grad_norm": 0.00018334065680392087,
+      "learning_rate": 1.3955984970477726e-06,
+      "loss": 0.0,
+      "step": 9050
+    },
+    {
+      "epoch": 129.00289855072464,
+      "grad_norm": 0.00016877887537702918,
+      "learning_rate": 1.3848631239935589e-06,
+      "loss": 0.0,
+      "step": 9060
+    },
+    {
+      "epoch": 129.00386473429953,
+      "grad_norm": 0.0004532422171905637,
+      "learning_rate": 1.3741277509393453e-06,
+      "loss": 0.0,
+      "step": 9070
+    },
+    {
+      "epoch": 129.0048309178744,
+      "grad_norm": 0.0008466316503472626,
+      "learning_rate": 1.3633923778851316e-06,
+      "loss": 0.0,
+      "step": 9080
+    },
+    {
+      "epoch": 129.00579710144928,
+      "grad_norm": 0.00024063742603175342,
+      "learning_rate": 1.3526570048309178e-06,
+      "loss": 0.0,
+      "step": 9090
+    },
+    {
+      "epoch": 129.00676328502416,
+      "grad_norm": 0.0019017226295545697,
+      "learning_rate": 1.3419216317767045e-06,
+      "loss": 0.0,
+      "step": 9100
+    },
+    {
+      "epoch": 129.00676328502416,
+      "eval_accuracy": 0.782051282051282,
+      "eval_loss": 1.991046667098999,
+      "eval_runtime": 19.7334,
+      "eval_samples_per_second": 3.953,
+      "eval_steps_per_second": 1.014,
+      "step": 9100
+    },
+    {
+      "epoch": 130.0009661835749,
+      "grad_norm": 0.0004907360416837037,
+      "learning_rate": 1.3311862587224907e-06,
+      "loss": 0.0,
+      "step": 9110
+    },
+    {
+      "epoch": 130.00193236714975,
+      "grad_norm": 0.00020176426914986223,
+      "learning_rate": 1.3204508856682772e-06,
+      "loss": 0.0,
+      "step": 9120
+    },
+    {
+      "epoch": 130.00289855072464,
+      "grad_norm": 0.0003256254130974412,
+      "learning_rate": 1.3097155126140634e-06,
+      "loss": 0.0,
+      "step": 9130
+    },
+    {
+      "epoch": 130.00386473429953,
+      "grad_norm": 0.0005137431435286999,
+      "learning_rate": 1.2989801395598498e-06,
+      "loss": 0.0,
+      "step": 9140
+    },
+    {
+      "epoch": 130.0048309178744,
+      "grad_norm": 0.00027728540590032935,
+      "learning_rate": 1.288244766505636e-06,
+      "loss": 0.0,
+      "step": 9150
+    },
+    {
+      "epoch": 130.00579710144928,
+      "grad_norm": 0.00032328334054909647,
+      "learning_rate": 1.2775093934514227e-06,
+      "loss": 0.0,
+      "step": 9160
+    },
+    {
+      "epoch": 130.00676328502416,
+      "grad_norm": 0.0008982921717688441,
+      "learning_rate": 1.266774020397209e-06,
+      "loss": 0.0,
+      "step": 9170
+    },
+    {
+      "epoch": 130.00676328502416,
+      "eval_accuracy": 0.782051282051282,
+      "eval_loss": 2.003160238265991,
+      "eval_runtime": 20.5795,
+      "eval_samples_per_second": 3.79,
+      "eval_steps_per_second": 0.972,
+      "step": 9170
+    },
+    {
+      "epoch": 131.0009661835749,
+      "grad_norm": 0.0008950904011726379,
+      "learning_rate": 1.2560386473429952e-06,
+      "loss": 0.0,
+      "step": 9180
+    },
+    {
+      "epoch": 131.00193236714975,
+      "grad_norm": 0.00045019862591288984,
+      "learning_rate": 1.2453032742887817e-06,
+      "loss": 0.0,
+      "step": 9190
+    },
+    {
+      "epoch": 131.00289855072464,
+      "grad_norm": 0.00017776140884961933,
+      "learning_rate": 1.234567901234568e-06,
+      "loss": 0.0,
+      "step": 9200
+    },
+    {
+      "epoch": 131.00386473429953,
+      "grad_norm": 0.0013761166483163834,
+      "learning_rate": 1.2238325281803544e-06,
+      "loss": 0.0,
+      "step": 9210
+    },
+    {
+      "epoch": 131.0048309178744,
+      "grad_norm": 0.00021241333161015064,
+      "learning_rate": 1.2130971551261406e-06,
+      "loss": 0.0,
+      "step": 9220
+    },
+    {
+      "epoch": 131.00579710144928,
+      "grad_norm": 0.025066981092095375,
+      "learning_rate": 1.2023617820719273e-06,
+      "loss": 0.0,
+      "step": 9230
+    },
+    {
+      "epoch": 131.00676328502416,
+      "grad_norm": 0.0006026471965014935,
+      "learning_rate": 1.1916264090177135e-06,
+      "loss": 0.0,
+      "step": 9240
+    },
+    {
+      "epoch": 131.00676328502416,
+      "eval_accuracy": 0.7692307692307693,
+      "eval_loss": 2.0645220279693604,
+      "eval_runtime": 20.8831,
+      "eval_samples_per_second": 3.735,
+      "eval_steps_per_second": 0.958,
+      "step": 9240
+    },
+    {
+      "epoch": 132.0009661835749,
+      "grad_norm": 0.00033965182956308126,
+      "learning_rate": 1.1808910359634997e-06,
+      "loss": 0.0,
+      "step": 9250
+    },
+    {
+      "epoch": 132.00193236714975,
+      "grad_norm": 0.00014069851022213697,
+      "learning_rate": 1.1701556629092862e-06,
+      "loss": 0.0,
+      "step": 9260
+    },
+    {
+      "epoch": 132.00289855072464,
+      "grad_norm": 0.00030967529164627194,
+      "learning_rate": 1.1594202898550726e-06,
+      "loss": 0.0,
+      "step": 9270
+    },
+    {
+      "epoch": 132.00386473429953,
+      "grad_norm": 0.000856581493280828,
+      "learning_rate": 1.1486849168008589e-06,
+      "loss": 0.0,
+      "step": 9280
+    },
+    {
+      "epoch": 132.0048309178744,
+      "grad_norm": 0.0006151138222776353,
+      "learning_rate": 1.1379495437466453e-06,
+      "loss": 0.0,
+      "step": 9290
+    },
+    {
+      "epoch": 132.00579710144928,
+      "grad_norm": 0.00021990279492456466,
+      "learning_rate": 1.1272141706924316e-06,
+      "loss": 0.0,
+      "step": 9300
+    },
+    {
+      "epoch": 132.00676328502416,
+      "grad_norm": 0.0009203171939589083,
+      "learning_rate": 1.116478797638218e-06,
+      "loss": 0.0,
+      "step": 9310
+    },
+    {
+      "epoch": 132.00676328502416,
+      "eval_accuracy": 0.7692307692307693,
+      "eval_loss": 2.0597715377807617,
+      "eval_runtime": 14.3229,
+      "eval_samples_per_second": 5.446,
+      "eval_steps_per_second": 1.396,
+      "step": 9310
+    },
+    {
+      "epoch": 133.0009661835749,
+      "grad_norm": 0.0005204367334954441,
+      "learning_rate": 1.1057434245840045e-06,
+      "loss": 0.0,
+      "step": 9320
+    },
+    {
+      "epoch": 133.00193236714975,
+      "grad_norm": 0.00021785238641314209,
+      "learning_rate": 1.0950080515297907e-06,
+      "loss": 0.0,
+      "step": 9330
+    },
+    {
+      "epoch": 133.00289855072464,
+      "grad_norm": 0.000165929232025519,
+      "learning_rate": 1.0842726784755772e-06,
+      "loss": 0.0,
+      "step": 9340
+    },
+    {
+      "epoch": 133.00386473429953,
+      "grad_norm": 0.00023569545010104775,
+      "learning_rate": 1.0735373054213636e-06,
+      "loss": 0.0,
+      "step": 9350
+    },
+    {
+      "epoch": 133.0048309178744,
+      "grad_norm": 0.00020854534523095936,
+      "learning_rate": 1.0628019323671499e-06,
+      "loss": 0.0,
+      "step": 9360
+    },
+    {
+      "epoch": 133.00579710144928,
+      "grad_norm": 0.000279417319688946,
+      "learning_rate": 1.052066559312936e-06,
+      "loss": 0.0,
+      "step": 9370
+    },
+    {
+      "epoch": 133.00676328502416,
+      "grad_norm": 0.0008123007137328386,
+      "learning_rate": 1.0413311862587225e-06,
+      "loss": 0.0,
+      "step": 9380
+    },
+    {
+      "epoch": 133.00676328502416,
+      "eval_accuracy": 0.7692307692307693,
+      "eval_loss": 2.05936598777771,
+      "eval_runtime": 15.7378,
+      "eval_samples_per_second": 4.956,
+      "eval_steps_per_second": 1.271,
+      "step": 9380
+    },
+    {
+      "epoch": 134.0009661835749,
+      "grad_norm": 0.0008277930901385844,
+      "learning_rate": 1.030595813204509e-06,
+      "loss": 0.0,
+      "step": 9390
+    },
+    {
+      "epoch": 134.00193236714975,
+      "grad_norm": 0.00015786287258379161,
+      "learning_rate": 1.0198604401502952e-06,
+      "loss": 0.0,
+      "step": 9400
+    },
+    {
+      "epoch": 134.00289855072464,
+      "grad_norm": 0.00024702053633518517,
+      "learning_rate": 1.0091250670960817e-06,
+      "loss": 0.0,
+      "step": 9410
+    },
+    {
+      "epoch": 134.00386473429953,
+      "grad_norm": 0.0004332618846092373,
+      "learning_rate": 9.983896940418681e-07,
+      "loss": 0.0,
+      "step": 9420
+    },
+    {
+      "epoch": 134.0048309178744,
+      "grad_norm": 0.00013526807015296072,
+      "learning_rate": 9.876543209876544e-07,
+      "loss": 0.0,
+      "step": 9430
+    },
+    {
+      "epoch": 134.00579710144928,
+      "grad_norm": 0.0002207817160524428,
+      "learning_rate": 9.769189479334408e-07,
+      "loss": 0.0,
+      "step": 9440
+    },
+    {
+      "epoch": 134.00676328502416,
+      "grad_norm": 0.0002987273037433624,
+      "learning_rate": 9.66183574879227e-07,
+      "loss": 0.0,
+      "step": 9450
+    },
+    {
+      "epoch": 134.00676328502416,
+      "eval_accuracy": 0.7692307692307693,
+      "eval_loss": 2.0567851066589355,
+      "eval_runtime": 18.6289,
+      "eval_samples_per_second": 4.187,
+      "eval_steps_per_second": 1.074,
+      "step": 9450
+    },
+    {
+      "epoch": 135.0009661835749,
+      "grad_norm": 0.00019443879136815667,
+      "learning_rate": 9.554482018250135e-07,
+      "loss": 0.0,
+      "step": 9460
+    },
+    {
+      "epoch": 135.00193236714975,
+      "grad_norm": 0.0002544256276451051,
+      "learning_rate": 9.447128287707999e-07,
+      "loss": 0.0,
+      "step": 9470
+    },
+    {
+      "epoch": 135.00289855072464,
+      "grad_norm": 0.008219041861593723,
+      "learning_rate": 9.339774557165862e-07,
+      "loss": 0.0,
+      "step": 9480
+    },
+    {
+      "epoch": 135.00386473429953,
+      "grad_norm": 0.00030363400583155453,
+      "learning_rate": 9.232420826623725e-07,
+      "loss": 0.0,
+      "step": 9490
+    },
+    {
+      "epoch": 135.0048309178744,
+      "grad_norm": 0.00029681427986361086,
+      "learning_rate": 9.12506709608159e-07,
+      "loss": 0.0,
+      "step": 9500
+    },
+    {
+      "epoch": 135.00579710144928,
+      "grad_norm": 0.00039895492955110967,
+      "learning_rate": 9.017713365539453e-07,
+      "loss": 0.0,
+      "step": 9510
+    },
+    {
+      "epoch": 135.00676328502416,
+      "grad_norm": 0.0014713755808770657,
+      "learning_rate": 8.910359634997316e-07,
+      "loss": 0.0,
+      "step": 9520
+    },
+    {
+      "epoch": 135.00676328502416,
+      "eval_accuracy": 0.782051282051282,
+      "eval_loss": 2.0521583557128906,
+      "eval_runtime": 16.9486,
+      "eval_samples_per_second": 4.602,
+      "eval_steps_per_second": 1.18,
+      "step": 9520
+    },
+    {
+      "epoch": 136.0009661835749,
+      "grad_norm": 0.0003343697462696582,
+      "learning_rate": 8.80300590445518e-07,
+      "loss": 0.0,
+      "step": 9530
+    },
+    {
+      "epoch": 136.00193236714975,
+      "grad_norm": 0.00029433504096232355,
+      "learning_rate": 8.695652173913044e-07,
+      "loss": 0.0,
+      "step": 9540
+    },
+    {
+      "epoch": 136.00289855072464,
+      "grad_norm": 0.0001634250656934455,
+      "learning_rate": 8.588298443370908e-07,
+      "loss": 0.0,
+      "step": 9550
+    },
+    {
+      "epoch": 136.00386473429953,
+      "grad_norm": 0.00036819910746999085,
+      "learning_rate": 8.480944712828772e-07,
+      "loss": 0.0,
+      "step": 9560
+    },
+    {
+      "epoch": 136.0048309178744,
+      "grad_norm": 0.0008551109349355102,
+      "learning_rate": 8.373590982286635e-07,
+      "loss": 0.0,
+      "step": 9570
+    },
+    {
+      "epoch": 136.00579710144928,
+      "grad_norm": 0.00022303566220216453,
+      "learning_rate": 8.2662372517445e-07,
+      "loss": 0.0,
+      "step": 9580
+    },
+    {
+      "epoch": 136.00676328502416,
+      "grad_norm": 0.0008858810761012137,
+      "learning_rate": 8.158883521202362e-07,
+      "loss": 0.0,
+      "step": 9590
+    },
+    {
+      "epoch": 136.00676328502416,
+      "eval_accuracy": 0.7564102564102564,
+      "eval_loss": 1.9970703125,
+      "eval_runtime": 16.5991,
+      "eval_samples_per_second": 4.699,
+      "eval_steps_per_second": 1.205,
+      "step": 9590
+    },
+    {
+      "epoch": 137.0009661835749,
+      "grad_norm": 0.0007048854604363441,
+      "learning_rate": 8.051529790660226e-07,
+      "loss": 0.0,
+      "step": 9600
+    },
+    {
+      "epoch": 137.00193236714975,
+      "grad_norm": 0.00012740909005515277,
+      "learning_rate": 7.94417606011809e-07,
+      "loss": 0.0,
+      "step": 9610
+    },
+    {
+      "epoch": 137.00289855072464,
+      "grad_norm": 0.0002288664982188493,
+      "learning_rate": 7.836822329575953e-07,
+      "loss": 0.0,
+      "step": 9620
+    },
+    {
+      "epoch": 137.00386473429953,
+      "grad_norm": 0.00017713684064801782,
+      "learning_rate": 7.729468599033817e-07,
+      "loss": 0.0,
+      "step": 9630
+    },
+    {
+      "epoch": 137.0048309178744,
+      "grad_norm": 0.0006587646203115582,
+      "learning_rate": 7.622114868491681e-07,
+      "loss": 0.0,
+      "step": 9640
+    },
+    {
+      "epoch": 137.00579710144928,
+      "grad_norm": 9.106951620196924e-05,
+      "learning_rate": 7.514761137949545e-07,
+      "loss": 0.0,
+      "step": 9650
+    },
+    {
+      "epoch": 137.00676328502416,
+      "grad_norm": 0.00013192169717513025,
+      "learning_rate": 7.407407407407407e-07,
+      "loss": 0.0,
+      "step": 9660
+    },
+    {
+      "epoch": 137.00676328502416,
+      "eval_accuracy": 0.7564102564102564,
+      "eval_loss": 1.997692346572876,
+      "eval_runtime": 17.822,
+      "eval_samples_per_second": 4.377,
+      "eval_steps_per_second": 1.122,
+      "step": 9660
+    },
+    {
+      "epoch": 138.0009661835749,
+      "grad_norm": 0.0002446017460897565,
+      "learning_rate": 7.300053676865272e-07,
+      "loss": 0.0,
+      "step": 9670
+    },
+    {
+      "epoch": 138.00193236714975,
+      "grad_norm": 0.001299283467233181,
+      "learning_rate": 7.192699946323135e-07,
+      "loss": 0.0,
+      "step": 9680
+    },
+    {
+      "epoch": 138.00289855072464,
+      "grad_norm": 0.0001351604878436774,
+      "learning_rate": 7.085346215780999e-07,
+      "loss": 0.0,
+      "step": 9690
+    },
+    {
+      "epoch": 138.00386473429953,
+      "grad_norm": 0.0011268679518252611,
+      "learning_rate": 6.977992485238863e-07,
+      "loss": 0.0,
+      "step": 9700
+    },
+    {
+      "epoch": 138.0048309178744,
+      "grad_norm": 0.0016840985044836998,
+      "learning_rate": 6.870638754696727e-07,
+      "loss": 0.0,
+      "step": 9710
+    },
+    {
+      "epoch": 138.00579710144928,
+      "grad_norm": 0.0004186121514067054,
+      "learning_rate": 6.763285024154589e-07,
+      "loss": 0.0,
+      "step": 9720
+    },
+    {
+      "epoch": 138.00676328502416,
+      "grad_norm": 9.602143109077588e-05,
+      "learning_rate": 6.655931293612454e-07,
+      "loss": 0.0,
+      "step": 9730
+    },
+    {
+      "epoch": 138.00676328502416,
+      "eval_accuracy": 0.7692307692307693,
+      "eval_loss": 2.089564800262451,
+      "eval_runtime": 14.6925,
+      "eval_samples_per_second": 5.309,
+      "eval_steps_per_second": 1.361,
+      "step": 9730
+    },
+    {
+      "epoch": 139.0009661835749,
+      "grad_norm": 0.0003402529691811651,
+      "learning_rate": 6.548577563070317e-07,
+      "loss": 0.0,
+      "step": 9740
+    },
+    {
+      "epoch": 139.00193236714975,
+      "grad_norm": 0.0008135391981340945,
+      "learning_rate": 6.44122383252818e-07,
+      "loss": 0.0,
+      "step": 9750
+    },
+    {
+      "epoch": 139.00289855072464,
+      "grad_norm": 0.00022346228070091456,
+      "learning_rate": 6.333870101986045e-07,
+      "loss": 0.0,
+      "step": 9760
+    },
+    {
+      "epoch": 139.00386473429953,
+      "grad_norm": 0.0002428313164273277,
+      "learning_rate": 6.226516371443908e-07,
+      "loss": 0.0,
+      "step": 9770
+    },
+    {
+      "epoch": 139.0048309178744,
+      "grad_norm": 0.00042313747690059245,
+      "learning_rate": 6.119162640901772e-07,
+      "loss": 0.0,
+      "step": 9780
+    },
+    {
+      "epoch": 139.00579710144928,
+      "grad_norm": 0.00024634020519442856,
+      "learning_rate": 6.011808910359636e-07,
+      "loss": 0.0,
+      "step": 9790
+    },
+    {
+      "epoch": 139.00676328502416,
+      "grad_norm": 0.0003057706926483661,
+      "learning_rate": 5.904455179817499e-07,
+      "loss": 0.0,
+      "step": 9800
+    },
+    {
+      "epoch": 139.00676328502416,
+      "eval_accuracy": 0.7692307692307693,
+      "eval_loss": 2.1549952030181885,
+      "eval_runtime": 14.6543,
+      "eval_samples_per_second": 5.323,
+      "eval_steps_per_second": 1.365,
+      "step": 9800
+    },
+    {
+      "epoch": 140.0009661835749,
+      "grad_norm": 0.00018014867964666337,
+      "learning_rate": 5.797101449275363e-07,
+      "loss": 0.0,
+      "step": 9810
+    },
+    {
+      "epoch": 140.00193236714975,
+      "grad_norm": 0.0005324236699379981,
+      "learning_rate": 5.689747718733227e-07,
+      "loss": 0.0014,
+      "step": 9820
+    },
+    {
+      "epoch": 140.00289855072464,
+      "grad_norm": 0.00020519823010545224,
+      "learning_rate": 5.58239398819109e-07,
+      "loss": 0.0,
+      "step": 9830
+    },
+    {
+      "epoch": 140.00386473429953,
+      "grad_norm": 0.00020373713050503284,
+      "learning_rate": 5.475040257648954e-07,
+      "loss": 0.0,
+      "step": 9840
+    },
+    {
+      "epoch": 140.0048309178744,
+      "grad_norm": 0.00020074652275070548,
+      "learning_rate": 5.367686527106818e-07,
+      "loss": 0.0,
+      "step": 9850
+    },
+    {
+      "epoch": 140.00579710144928,
+      "grad_norm": 0.0001167950322269462,
+      "learning_rate": 5.26033279656468e-07,
+      "loss": 0.0,
+      "step": 9860
+    },
+    {
+      "epoch": 140.00676328502416,
+      "grad_norm": 0.0004204206052236259,
+      "learning_rate": 5.152979066022545e-07,
+      "loss": 0.0,
+      "step": 9870
+    },
+    {
+      "epoch": 140.00676328502416,
+      "eval_accuracy": 0.7692307692307693,
+      "eval_loss": 2.1875460147857666,
+      "eval_runtime": 13.9179,
+      "eval_samples_per_second": 5.604,
+      "eval_steps_per_second": 1.437,
+      "step": 9870
+    },
+    {
+      "epoch": 141.0009661835749,
+      "grad_norm": 0.0008608843781985343,
+      "learning_rate": 5.045625335480408e-07,
+      "loss": 0.0,
+      "step": 9880
+    },
+    {
+      "epoch": 141.00193236714975,
+      "grad_norm": 0.0035127594601362944,
+      "learning_rate": 4.938271604938272e-07,
+      "loss": 0.0,
+      "step": 9890
+    },
+    {
+      "epoch": 141.00289855072464,
+      "grad_norm": 0.00016304482414852828,
+      "learning_rate": 4.830917874396135e-07,
+      "loss": 0.0,
+      "step": 9900
+    },
+    {
+      "epoch": 141.00386473429953,
+      "grad_norm": 0.00017550381016917527,
+      "learning_rate": 4.7235641438539993e-07,
+      "loss": 0.0,
+      "step": 9910
+    },
+    {
+      "epoch": 141.0048309178744,
+      "grad_norm": 0.00027925425092689693,
+      "learning_rate": 4.616210413311863e-07,
+      "loss": 0.0,
+      "step": 9920
+    },
+    {
+      "epoch": 141.00579710144928,
+      "grad_norm": 0.00017096209921874106,
+      "learning_rate": 4.5088566827697267e-07,
+      "loss": 0.0,
+      "step": 9930
+    },
+    {
+      "epoch": 141.00676328502416,
+      "grad_norm": 0.0006372460629791021,
+      "learning_rate": 4.40150295222759e-07,
+      "loss": 0.0,
+      "step": 9940
+    },
+    {
+      "epoch": 141.00676328502416,
+      "eval_accuracy": 0.7692307692307693,
+      "eval_loss": 2.1873533725738525,
+      "eval_runtime": 13.9433,
+      "eval_samples_per_second": 5.594,
+      "eval_steps_per_second": 1.434,
+      "step": 9940
+    },
+    {
+      "epoch": 142.0009661835749,
+      "grad_norm": 0.0002361346414545551,
+      "learning_rate": 4.294149221685454e-07,
+      "loss": 0.0,
+      "step": 9950
+    },
+    {
+      "epoch": 142.00193236714975,
+      "grad_norm": 0.0004928818088956177,
+      "learning_rate": 4.1867954911433176e-07,
+      "loss": 0.0,
+      "step": 9960
+    },
+    {
+      "epoch": 142.00289855072464,
+      "grad_norm": 0.00033226932282559574,
+      "learning_rate": 4.079441760601181e-07,
+      "loss": 0.0,
+      "step": 9970
+    },
+    {
+      "epoch": 142.00386473429953,
+      "grad_norm": 0.00016688242612872273,
+      "learning_rate": 3.972088030059045e-07,
+      "loss": 0.0,
+      "step": 9980
+    },
+    {
+      "epoch": 142.0048309178744,
+      "grad_norm": 0.0001479385537095368,
+      "learning_rate": 3.8647342995169085e-07,
+      "loss": 0.0,
+      "step": 9990
+    },
+    {
+      "epoch": 142.00579710144928,
+      "grad_norm": 0.00019251066260039806,
+      "learning_rate": 3.7573805689747724e-07,
+      "loss": 0.0,
+      "step": 10000
+    },
+    {
+      "epoch": 142.00676328502416,
+      "grad_norm": 0.000183121781446971,
+      "learning_rate": 3.650026838432636e-07,
+      "loss": 0.0,
+      "step": 10010
+    },
+    {
+      "epoch": 142.00676328502416,
+      "eval_accuracy": 0.7692307692307693,
+      "eval_loss": 2.1822385787963867,
+      "eval_runtime": 14.308,
+      "eval_samples_per_second": 5.452,
+      "eval_steps_per_second": 1.398,
+      "step": 10010
+    },
+    {
+      "epoch": 143.0009661835749,
+      "grad_norm": 0.00021158168965484947,
+      "learning_rate": 3.5426731078904993e-07,
+      "loss": 0.0,
+      "step": 10020
+    },
+    {
+      "epoch": 143.00193236714975,
+      "grad_norm": 0.0001671000791247934,
+      "learning_rate": 3.4353193773483633e-07,
+      "loss": 0.0,
+      "step": 10030
+    },
+    {
+      "epoch": 143.00289855072464,
+      "grad_norm": 0.00016769721696618944,
+      "learning_rate": 3.327965646806227e-07,
+      "loss": 0.0,
+      "step": 10040
+    },
+    {
+      "epoch": 143.00386473429953,
+      "grad_norm": 0.00015122311015147716,
+      "learning_rate": 3.22061191626409e-07,
+      "loss": 0.0,
+      "step": 10050
+    },
+    {
+      "epoch": 143.0048309178744,
+      "grad_norm": 0.00033184216590598226,
+      "learning_rate": 3.113258185721954e-07,
+      "loss": 0.0,
+      "step": 10060
+    },
+    {
+      "epoch": 143.00579710144928,
+      "grad_norm": 0.0002991911314893514,
+      "learning_rate": 3.005904455179818e-07,
+      "loss": 0.0,
+      "step": 10070
+    },
+    {
+      "epoch": 143.00676328502416,
+      "grad_norm": 0.00013400911120697856,
+      "learning_rate": 2.8985507246376816e-07,
+      "loss": 0.0,
+      "step": 10080
+    },
+    {
+      "epoch": 143.00676328502416,
+      "eval_accuracy": 0.7692307692307693,
+      "eval_loss": 2.181835412979126,
+      "eval_runtime": 14.6353,
+      "eval_samples_per_second": 5.33,
+      "eval_steps_per_second": 1.367,
+      "step": 10080
+    },
+    {
+      "epoch": 144.0009661835749,
+      "grad_norm": 0.0001004399309749715,
+      "learning_rate": 2.791196994095545e-07,
+      "loss": 0.0,
+      "step": 10090
+    },
+    {
+      "epoch": 144.00193236714975,
+      "grad_norm": 0.00020495509670581669,
+      "learning_rate": 2.683843263553409e-07,
+      "loss": 0.0,
+      "step": 10100
+    },
+    {
+      "epoch": 144.00289855072464,
+      "grad_norm": 0.00013104511890560389,
+      "learning_rate": 2.5764895330112725e-07,
+      "loss": 0.0,
+      "step": 10110
+    },
+    {
+      "epoch": 144.00386473429953,
+      "grad_norm": 0.0002047176967607811,
+      "learning_rate": 2.469135802469136e-07,
+      "loss": 0.0,
+      "step": 10120
+    },
+    {
+      "epoch": 144.0048309178744,
+      "grad_norm": 0.000391674431739375,
+      "learning_rate": 2.3617820719269996e-07,
+      "loss": 0.0,
+      "step": 10130
+    },
+    {
+      "epoch": 144.00579710144928,
+      "grad_norm": 0.0001891203864943236,
+      "learning_rate": 2.2544283413848634e-07,
+      "loss": 0.0,
+      "step": 10140
+    },
+    {
+      "epoch": 144.00676328502416,
+      "grad_norm": 0.00012977355800103396,
+      "learning_rate": 2.147074610842727e-07,
+      "loss": 0.0,
+      "step": 10150
+    },
+    {
+      "epoch": 144.00676328502416,
+      "eval_accuracy": 0.7692307692307693,
+      "eval_loss": 2.1805622577667236,
+      "eval_runtime": 13.9887,
+      "eval_samples_per_second": 5.576,
+      "eval_steps_per_second": 1.43,
+      "step": 10150
+    },
+    {
+      "epoch": 145.0009661835749,
+      "grad_norm": 0.0001362557668471709,
+      "learning_rate": 2.0397208803005905e-07,
+      "loss": 0.0,
+      "step": 10160
+    },
+    {
+      "epoch": 145.00193236714975,
+      "grad_norm": 0.0005237432778812945,
+      "learning_rate": 1.9323671497584542e-07,
+      "loss": 0.0,
+      "step": 10170
+    },
+    {
+      "epoch": 145.00289855072464,
+      "grad_norm": 0.0002967139007523656,
+      "learning_rate": 1.825013419216318e-07,
+      "loss": 0.0,
+      "step": 10180
+    },
+    {
+      "epoch": 145.00386473429953,
+      "grad_norm": 0.00015602614439558238,
+      "learning_rate": 1.7176596886741817e-07,
+      "loss": 0.0,
+      "step": 10190
+    },
+    {
+      "epoch": 145.0048309178744,
+      "grad_norm": 0.0002610948868095875,
+      "learning_rate": 1.610305958132045e-07,
+      "loss": 0.0,
+      "step": 10200
+    },
+    {
+      "epoch": 145.00579710144928,
+      "grad_norm": 0.00014871751773171127,
+      "learning_rate": 1.502952227589909e-07,
+      "loss": 0.0,
+      "step": 10210
+    },
+    {
+      "epoch": 145.00676328502416,
+      "grad_norm": 0.00018696517508942634,
+      "learning_rate": 1.3955984970477725e-07,
+      "loss": 0.0,
+      "step": 10220
+    },
+    {
+      "epoch": 145.00676328502416,
+      "eval_accuracy": 0.7692307692307693,
+      "eval_loss": 2.1803364753723145,
+      "eval_runtime": 14.0396,
+      "eval_samples_per_second": 5.556,
+      "eval_steps_per_second": 1.425,
+      "step": 10220
+    },
+    {
+      "epoch": 146.0009661835749,
+      "grad_norm": 0.0006746940198354423,
+      "learning_rate": 1.2882447665056362e-07,
+      "loss": 0.0,
+      "step": 10230
+    },
+    {
+      "epoch": 146.00193236714975,
+      "grad_norm": 0.0010222846176475286,
+      "learning_rate": 1.1808910359634998e-07,
+      "loss": 0.0,
+      "step": 10240
+    },
+    {
+      "epoch": 146.00289855072464,
+      "grad_norm": 0.00016020600742194802,
+      "learning_rate": 1.0735373054213635e-07,
+      "loss": 0.0,
+      "step": 10250
+    },
+    {
+      "epoch": 146.00386473429953,
+      "grad_norm": 0.00013268800103105605,
+      "learning_rate": 9.661835748792271e-08,
+      "loss": 0.0,
+      "step": 10260
+    },
+    {
+      "epoch": 146.0048309178744,
+      "grad_norm": 0.0003070248058065772,
+      "learning_rate": 8.588298443370908e-08,
+      "loss": 0.0,
+      "step": 10270
+    },
+    {
+      "epoch": 146.00579710144928,
+      "grad_norm": 0.00037805610918439925,
+      "learning_rate": 7.514761137949545e-08,
+      "loss": 0.0,
+      "step": 10280
+    },
+    {
+      "epoch": 146.00676328502416,
+      "grad_norm": 0.0001392628182657063,
+      "learning_rate": 6.441223832528181e-08,
+      "loss": 0.0,
+      "step": 10290
+    },
+    {
+      "epoch": 146.00676328502416,
+      "eval_accuracy": 0.7692307692307693,
+      "eval_loss": 2.180433988571167,
+      "eval_runtime": 14.6632,
+      "eval_samples_per_second": 5.319,
+      "eval_steps_per_second": 1.364,
+      "step": 10290
+    },
+    {
+      "epoch": 147.0009661835749,
+      "grad_norm": 0.0004906918038614094,
+      "learning_rate": 5.367686527106818e-08,
+      "loss": 0.0,
+      "step": 10300
+    },
+    {
+      "epoch": 147.00193236714975,
+      "grad_norm": 0.0002476818044669926,
+      "learning_rate": 4.294149221685454e-08,
+      "loss": 0.0,
+      "step": 10310
+    },
+    {
+      "epoch": 147.00289855072464,
+      "grad_norm": 0.0003119547909591347,
+      "learning_rate": 3.2206119162640906e-08,
+      "loss": 0.0,
+      "step": 10320
+    },
+    {
+      "epoch": 147.00386473429953,
+      "grad_norm": 0.00017419341020286083,
+      "learning_rate": 2.147074610842727e-08,
+      "loss": 0.0,
+      "step": 10330
+    },
+    {
+      "epoch": 147.0048309178744,
+      "grad_norm": 0.00013093346206005663,
+      "learning_rate": 1.0735373054213635e-08,
+      "loss": 0.0,
+      "step": 10340
+    },
+    {
+      "epoch": 147.00579710144928,
+      "grad_norm": 0.0008992272196337581,
+      "learning_rate": 0.0,
+      "loss": 0.0,
+      "step": 10350
+    },
+    {
+      "epoch": 147.00579710144928,
+      "eval_accuracy": 0.7692307692307693,
+      "eval_loss": 2.1802780628204346,
+      "eval_runtime": 14.9692,
+      "eval_samples_per_second": 5.211,
+      "eval_steps_per_second": 1.336,
+      "step": 10350
+    },
+    {
+      "epoch": 147.00579710144928,
+      "step": 10350,
+      "total_flos": 1.7985204684824032e+20,
+      "train_loss": 0.10165577027017988,
+      "train_runtime": 21165.7287,
+      "train_samples_per_second": 1.956,
+      "train_steps_per_second": 0.489
+    },
+    {
+      "epoch": 147.00579710144928,
+      "eval_accuracy": 0.8461538461538461,
+      "eval_loss": 1.0547696352005005,
+      "eval_runtime": 15.5882,
+      "eval_samples_per_second": 5.004,
+      "eval_steps_per_second": 1.283,
+      "step": 10350
+    },
+    {
+      "epoch": 147.00579710144928,
+      "eval_accuracy": 0.8461538461538461,
+      "eval_loss": 1.054769515991211,
+      "eval_runtime": 14.4609,
+      "eval_samples_per_second": 5.394,
+      "eval_steps_per_second": 1.383,
+      "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": 1.7985204684824032e+20,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}