{
  "best_metric": 0.922,
  "best_model_checkpoint": "checkpoint/vit-base/checkpoint-13653",
  "epoch": 41.0,
  "eval_steps": 500,
  "global_step": 13653,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.03,
      "grad_norm": 0.6713149547576904,
      "learning_rate": 9.996996996996998e-06,
      "loss": 4.6131,
      "step": 10
    },
    {
      "epoch": 0.06,
      "grad_norm": 0.6888127326965332,
      "learning_rate": 9.993993993993994e-06,
      "loss": 4.6033,
      "step": 20
    },
    {
      "epoch": 0.09,
      "grad_norm": 0.7794463634490967,
      "learning_rate": 9.990990990990992e-06,
      "loss": 4.6019,
      "step": 30
    },
    {
      "epoch": 0.12,
      "grad_norm": 0.7187056541442871,
      "learning_rate": 9.987987987987989e-06,
      "loss": 4.5877,
      "step": 40
    },
    {
      "epoch": 0.15,
      "grad_norm": 0.6908490657806396,
      "learning_rate": 9.984984984984985e-06,
      "loss": 4.578,
      "step": 50
    },
    {
      "epoch": 0.18,
      "grad_norm": 0.6914451718330383,
      "learning_rate": 9.981981981981982e-06,
      "loss": 4.5734,
      "step": 60
    },
    {
      "epoch": 0.21,
      "grad_norm": 0.7061388492584229,
      "learning_rate": 9.97897897897898e-06,
      "loss": 4.5547,
      "step": 70
    },
    {
      "epoch": 0.24,
      "grad_norm": 0.7158679366111755,
      "learning_rate": 9.975975975975977e-06,
      "loss": 4.544,
      "step": 80
    },
    {
      "epoch": 0.27,
      "grad_norm": 0.7526296973228455,
      "learning_rate": 9.972972972972975e-06,
      "loss": 4.5366,
      "step": 90
    },
    {
      "epoch": 0.3,
      "grad_norm": 0.7619982957839966,
      "learning_rate": 9.96996996996997e-06,
      "loss": 4.5252,
      "step": 100
    },
    {
      "epoch": 0.33,
      "grad_norm": 0.7292161583900452,
      "learning_rate": 9.966966966966968e-06,
      "loss": 4.511,
      "step": 110
    },
    {
      "epoch": 0.36,
      "grad_norm": 0.7648303508758545,
      "learning_rate": 9.963963963963965e-06,
      "loss": 4.505,
      "step": 120
    },
    {
      "epoch": 0.39,
      "grad_norm": 0.7616661190986633,
      "learning_rate": 9.960960960960961e-06,
      "loss": 4.4781,
      "step": 130
    },
    {
      "epoch": 0.42,
      "grad_norm": 0.792500913143158,
      "learning_rate": 9.957957957957959e-06,
      "loss": 4.467,
      "step": 140
    },
    {
      "epoch": 0.45,
      "grad_norm": 1.064399003982544,
      "learning_rate": 9.954954954954956e-06,
      "loss": 4.4612,
      "step": 150
    },
    {
      "epoch": 0.48,
      "grad_norm": 0.8003212809562683,
      "learning_rate": 9.951951951951952e-06,
      "loss": 4.4334,
      "step": 160
    },
    {
      "epoch": 0.51,
      "grad_norm": 0.8235049843788147,
      "learning_rate": 9.94894894894895e-06,
      "loss": 4.4173,
      "step": 170
    },
    {
      "epoch": 0.54,
      "grad_norm": 0.8271332383155823,
      "learning_rate": 9.945945945945947e-06,
      "loss": 4.3952,
      "step": 180
    },
    {
      "epoch": 0.57,
      "grad_norm": 0.8518035411834717,
      "learning_rate": 9.942942942942944e-06,
      "loss": 4.3846,
      "step": 190
    },
    {
      "epoch": 0.6,
      "grad_norm": 0.8331427574157715,
      "learning_rate": 9.93993993993994e-06,
      "loss": 4.3631,
      "step": 200
    },
    {
      "epoch": 0.63,
      "grad_norm": 0.8386998176574707,
      "learning_rate": 9.936936936936938e-06,
      "loss": 4.3418,
      "step": 210
    },
    {
      "epoch": 0.66,
      "grad_norm": 0.9001342058181763,
      "learning_rate": 9.933933933933935e-06,
      "loss": 4.3226,
      "step": 220
    },
    {
      "epoch": 0.69,
      "grad_norm": 0.9490275382995605,
      "learning_rate": 9.930930930930933e-06,
      "loss": 4.2988,
      "step": 230
    },
    {
      "epoch": 0.72,
      "grad_norm": 0.9027977585792542,
      "learning_rate": 9.927927927927928e-06,
      "loss": 4.2937,
      "step": 240
    },
    {
      "epoch": 0.75,
      "grad_norm": 0.8498959541320801,
      "learning_rate": 9.924924924924926e-06,
      "loss": 4.2621,
      "step": 250
    },
    {
      "epoch": 0.78,
      "grad_norm": 0.9752369523048401,
      "learning_rate": 9.921921921921923e-06,
      "loss": 4.2379,
      "step": 260
    },
    {
      "epoch": 0.81,
      "grad_norm": 0.8778105974197388,
      "learning_rate": 9.91891891891892e-06,
      "loss": 4.2288,
      "step": 270
    },
    {
      "epoch": 0.84,
      "grad_norm": 0.9475248456001282,
      "learning_rate": 9.915915915915917e-06,
      "loss": 4.2066,
      "step": 280
    },
    {
      "epoch": 0.87,
      "grad_norm": 1.0335466861724854,
      "learning_rate": 9.912912912912914e-06,
      "loss": 4.1814,
      "step": 290
    },
    {
      "epoch": 0.9,
      "grad_norm": 1.091416597366333,
      "learning_rate": 9.90990990990991e-06,
      "loss": 4.1663,
      "step": 300
    },
    {
      "epoch": 0.93,
      "grad_norm": 0.9959795475006104,
      "learning_rate": 9.906906906906907e-06,
      "loss": 4.145,
      "step": 310
    },
    {
      "epoch": 0.96,
      "grad_norm": 0.9685789942741394,
      "learning_rate": 9.903903903903905e-06,
      "loss": 4.1289,
      "step": 320
    },
    {
      "epoch": 0.99,
      "grad_norm": 1.0388822555541992,
      "learning_rate": 9.900900900900902e-06,
      "loss": 4.0904,
      "step": 330
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.6663,
      "eval_loss": 4.0142011642456055,
      "eval_runtime": 26.8951,
      "eval_samples_per_second": 371.815,
      "eval_steps_per_second": 1.487,
      "step": 333
    },
    {
      "epoch": 1.02,
      "grad_norm": 1.0019899606704712,
      "learning_rate": 9.8978978978979e-06,
      "loss": 4.0766,
      "step": 340
    },
    {
      "epoch": 1.05,
      "grad_norm": 1.0191450119018555,
      "learning_rate": 9.894894894894896e-06,
      "loss": 4.0525,
      "step": 350
    },
    {
      "epoch": 1.08,
      "grad_norm": 0.9841620326042175,
      "learning_rate": 9.891891891891893e-06,
      "loss": 4.0274,
      "step": 360
    },
    {
      "epoch": 1.11,
      "grad_norm": 1.0833967924118042,
      "learning_rate": 9.88888888888889e-06,
      "loss": 4.0079,
      "step": 370
    },
    {
      "epoch": 1.14,
      "grad_norm": 1.0421665906906128,
      "learning_rate": 9.885885885885886e-06,
      "loss": 3.9865,
      "step": 380
    },
    {
      "epoch": 1.17,
      "grad_norm": 1.1187000274658203,
      "learning_rate": 9.882882882882884e-06,
      "loss": 3.9617,
      "step": 390
    },
    {
      "epoch": 1.2,
      "grad_norm": 1.0615652799606323,
      "learning_rate": 9.879879879879881e-06,
      "loss": 3.9559,
      "step": 400
    },
    {
      "epoch": 1.23,
      "grad_norm": 1.0917859077453613,
      "learning_rate": 9.876876876876877e-06,
      "loss": 3.9249,
      "step": 410
    },
    {
      "epoch": 1.26,
      "grad_norm": 1.2941133975982666,
      "learning_rate": 9.873873873873875e-06,
      "loss": 3.9088,
      "step": 420
    },
    {
      "epoch": 1.29,
      "grad_norm": 1.2532916069030762,
      "learning_rate": 9.87087087087087e-06,
      "loss": 3.883,
      "step": 430
    },
    {
      "epoch": 1.32,
      "grad_norm": 1.1082998514175415,
      "learning_rate": 9.867867867867868e-06,
      "loss": 3.8705,
      "step": 440
    },
    {
      "epoch": 1.35,
      "grad_norm": 1.1208393573760986,
      "learning_rate": 9.864864864864865e-06,
      "loss": 3.8581,
      "step": 450
    },
    {
      "epoch": 1.38,
      "grad_norm": 1.2406553030014038,
      "learning_rate": 9.861861861861863e-06,
      "loss": 3.8187,
      "step": 460
    },
    {
      "epoch": 1.41,
      "grad_norm": 1.198391318321228,
      "learning_rate": 9.85885885885886e-06,
      "loss": 3.7917,
      "step": 470
    },
    {
      "epoch": 1.44,
      "grad_norm": 1.045142412185669,
      "learning_rate": 9.855855855855858e-06,
      "loss": 3.7865,
      "step": 480
    },
    {
      "epoch": 1.47,
      "grad_norm": 1.4195159673690796,
      "learning_rate": 9.852852852852854e-06,
      "loss": 3.7659,
      "step": 490
    },
    {
      "epoch": 1.5,
      "grad_norm": 1.2999833822250366,
      "learning_rate": 9.849849849849851e-06,
      "loss": 3.7566,
      "step": 500
    },
    {
      "epoch": 1.53,
      "grad_norm": 1.128721833229065,
      "learning_rate": 9.846846846846849e-06,
      "loss": 3.7394,
      "step": 510
    },
    {
      "epoch": 1.56,
      "grad_norm": 1.24117910861969,
      "learning_rate": 9.843843843843844e-06,
      "loss": 3.7364,
      "step": 520
    },
    {
      "epoch": 1.59,
      "grad_norm": 1.3166449069976807,
      "learning_rate": 9.840840840840842e-06,
      "loss": 3.6919,
      "step": 530
    },
    {
      "epoch": 1.62,
      "grad_norm": 1.3979932069778442,
      "learning_rate": 9.83783783783784e-06,
      "loss": 3.6865,
      "step": 540
    },
    {
      "epoch": 1.65,
      "grad_norm": 1.4577465057373047,
      "learning_rate": 9.834834834834835e-06,
      "loss": 3.67,
      "step": 550
    },
    {
      "epoch": 1.68,
      "grad_norm": 1.3105660676956177,
      "learning_rate": 9.831831831831833e-06,
      "loss": 3.6448,
      "step": 560
    },
    {
      "epoch": 1.71,
      "grad_norm": 1.2558164596557617,
      "learning_rate": 9.82882882882883e-06,
      "loss": 3.6515,
      "step": 570
    },
    {
      "epoch": 1.74,
      "grad_norm": 1.2949180603027344,
      "learning_rate": 9.825825825825826e-06,
      "loss": 3.637,
      "step": 580
    },
    {
      "epoch": 1.77,
      "grad_norm": 1.2413955926895142,
      "learning_rate": 9.822822822822823e-06,
      "loss": 3.5952,
      "step": 590
    },
    {
      "epoch": 1.8,
      "grad_norm": 1.7929736375808716,
      "learning_rate": 9.81981981981982e-06,
      "loss": 3.5952,
      "step": 600
    },
    {
      "epoch": 1.83,
      "grad_norm": 1.2751438617706299,
      "learning_rate": 9.816816816816818e-06,
      "loss": 3.5877,
      "step": 610
    },
    {
      "epoch": 1.86,
      "grad_norm": 1.4566322565078735,
      "learning_rate": 9.813813813813816e-06,
      "loss": 3.5815,
      "step": 620
    },
    {
      "epoch": 1.89,
      "grad_norm": 1.3014384508132935,
      "learning_rate": 9.810810810810811e-06,
      "loss": 3.5361,
      "step": 630
    },
    {
      "epoch": 1.92,
      "grad_norm": 1.343555212020874,
      "learning_rate": 9.807807807807809e-06,
      "loss": 3.5176,
      "step": 640
    },
    {
      "epoch": 1.95,
      "grad_norm": 1.2463815212249756,
      "learning_rate": 9.804804804804806e-06,
      "loss": 3.522,
      "step": 650
    },
    {
      "epoch": 1.98,
      "grad_norm": 1.288573145866394,
      "learning_rate": 9.801801801801802e-06,
      "loss": 3.5092,
      "step": 660
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.7659,
      "eval_loss": 3.3562543392181396,
      "eval_runtime": 26.0758,
      "eval_samples_per_second": 383.497,
      "eval_steps_per_second": 1.534,
      "step": 666
    },
    {
      "epoch": 2.01,
      "grad_norm": 1.3275058269500732,
      "learning_rate": 9.7987987987988e-06,
      "loss": 3.4905,
      "step": 670
    },
    {
      "epoch": 2.04,
      "grad_norm": 2.9843428134918213,
      "learning_rate": 9.795795795795795e-06,
      "loss": 3.4694,
      "step": 680
    },
    {
      "epoch": 2.07,
      "grad_norm": 1.4618003368377686,
      "learning_rate": 9.792792792792793e-06,
      "loss": 3.448,
      "step": 690
    },
    {
      "epoch": 2.1,
      "grad_norm": 1.5394887924194336,
      "learning_rate": 9.78978978978979e-06,
      "loss": 3.4395,
      "step": 700
    },
    {
      "epoch": 2.13,
      "grad_norm": 1.4299417734146118,
      "learning_rate": 9.786786786786788e-06,
      "loss": 3.4062,
      "step": 710
    },
    {
      "epoch": 2.16,
      "grad_norm": 1.4161120653152466,
      "learning_rate": 9.783783783783785e-06,
      "loss": 3.3925,
      "step": 720
    },
    {
      "epoch": 2.19,
      "grad_norm": 1.5492268800735474,
      "learning_rate": 9.780780780780781e-06,
      "loss": 3.3897,
      "step": 730
    },
    {
      "epoch": 2.22,
      "grad_norm": 2.0198769569396973,
      "learning_rate": 9.777777777777779e-06,
      "loss": 3.3676,
      "step": 740
    },
    {
      "epoch": 2.25,
      "grad_norm": 1.5388668775558472,
      "learning_rate": 9.774774774774776e-06,
      "loss": 3.358,
      "step": 750
    },
    {
      "epoch": 2.28,
      "grad_norm": 1.6485122442245483,
      "learning_rate": 9.771771771771774e-06,
      "loss": 3.3713,
      "step": 760
    },
    {
      "epoch": 2.31,
      "grad_norm": 1.4683903455734253,
      "learning_rate": 9.76876876876877e-06,
      "loss": 3.3385,
      "step": 770
    },
    {
      "epoch": 2.34,
      "grad_norm": 2.024306058883667,
      "learning_rate": 9.765765765765767e-06,
      "loss": 3.3183,
      "step": 780
    },
    {
      "epoch": 2.37,
      "grad_norm": 1.6113808155059814,
      "learning_rate": 9.762762762762763e-06,
      "loss": 3.2949,
      "step": 790
    },
    {
      "epoch": 2.4,
      "grad_norm": 1.5689467191696167,
      "learning_rate": 9.75975975975976e-06,
      "loss": 3.3244,
      "step": 800
    },
    {
      "epoch": 2.43,
      "grad_norm": 1.6313070058822632,
      "learning_rate": 9.756756756756758e-06,
      "loss": 3.3233,
      "step": 810
    },
    {
      "epoch": 2.46,
      "grad_norm": 1.8879677057266235,
      "learning_rate": 9.753753753753753e-06,
      "loss": 3.2708,
      "step": 820
    },
    {
      "epoch": 2.49,
      "grad_norm": 1.6785987615585327,
      "learning_rate": 9.750750750750751e-06,
      "loss": 3.2796,
      "step": 830
    },
    {
      "epoch": 2.52,
      "grad_norm": 1.5085214376449585,
      "learning_rate": 9.747747747747748e-06,
      "loss": 3.2767,
      "step": 840
    },
    {
      "epoch": 2.55,
      "grad_norm": 1.9259861707687378,
      "learning_rate": 9.744744744744746e-06,
      "loss": 3.2647,
      "step": 850
    },
    {
      "epoch": 2.58,
      "grad_norm": 1.6628590822219849,
      "learning_rate": 9.741741741741743e-06,
      "loss": 3.21,
      "step": 860
    },
    {
      "epoch": 2.61,
      "grad_norm": 1.4452886581420898,
      "learning_rate": 9.73873873873874e-06,
      "loss": 3.2561,
      "step": 870
    },
    {
      "epoch": 2.64,
      "grad_norm": 1.4690823554992676,
      "learning_rate": 9.735735735735737e-06,
      "loss": 3.2207,
      "step": 880
    },
    {
      "epoch": 2.67,
      "grad_norm": 1.831654667854309,
      "learning_rate": 9.732732732732734e-06,
      "loss": 3.1828,
      "step": 890
    },
    {
      "epoch": 2.7,
      "grad_norm": 1.5297893285751343,
      "learning_rate": 9.729729729729732e-06,
      "loss": 3.2005,
      "step": 900
    },
    {
      "epoch": 2.73,
      "grad_norm": 3.0812528133392334,
      "learning_rate": 9.726726726726727e-06,
      "loss": 3.2103,
      "step": 910
    },
    {
      "epoch": 2.76,
      "grad_norm": 1.4284906387329102,
      "learning_rate": 9.723723723723725e-06,
      "loss": 3.1591,
      "step": 920
    },
    {
      "epoch": 2.79,
      "grad_norm": 1.5879852771759033,
      "learning_rate": 9.72072072072072e-06,
      "loss": 3.1647,
      "step": 930
    },
    {
      "epoch": 2.82,
      "grad_norm": 2.1024491786956787,
      "learning_rate": 9.717717717717718e-06,
      "loss": 3.1297,
      "step": 940
    },
    {
      "epoch": 2.85,
      "grad_norm": 1.941714882850647,
      "learning_rate": 9.714714714714716e-06,
      "loss": 3.1306,
      "step": 950
    },
    {
      "epoch": 2.88,
      "grad_norm": 1.642816424369812,
      "learning_rate": 9.711711711711711e-06,
      "loss": 3.1393,
      "step": 960
    },
    {
      "epoch": 2.91,
      "grad_norm": 1.7811369895935059,
      "learning_rate": 9.708708708708709e-06,
      "loss": 3.1237,
      "step": 970
    },
    {
      "epoch": 2.94,
      "grad_norm": 1.7237188816070557,
      "learning_rate": 9.705705705705706e-06,
      "loss": 3.0829,
      "step": 980
    },
    {
      "epoch": 2.97,
      "grad_norm": 2.070939064025879,
      "learning_rate": 9.702702702702704e-06,
      "loss": 3.0987,
      "step": 990
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8043,
      "eval_loss": 2.906019926071167,
      "eval_runtime": 25.8478,
      "eval_samples_per_second": 386.881,
      "eval_steps_per_second": 1.548,
      "step": 999
    },
    {
      "epoch": 3.0,
      "grad_norm": 2.190770149230957,
      "learning_rate": 9.699699699699701e-06,
      "loss": 3.0663,
      "step": 1000
    },
    {
      "epoch": 3.03,
      "grad_norm": 1.7228102684020996,
      "learning_rate": 9.696696696696699e-06,
      "loss": 3.0587,
      "step": 1010
    },
    {
      "epoch": 3.06,
      "grad_norm": 1.7540990114212036,
      "learning_rate": 9.693693693693694e-06,
      "loss": 3.0384,
      "step": 1020
    },
    {
      "epoch": 3.09,
      "grad_norm": 1.5200778245925903,
      "learning_rate": 9.690690690690692e-06,
      "loss": 3.0352,
      "step": 1030
    },
    {
      "epoch": 3.12,
      "grad_norm": 1.7722601890563965,
      "learning_rate": 9.687687687687688e-06,
      "loss": 2.9858,
      "step": 1040
    },
    {
      "epoch": 3.15,
      "grad_norm": 1.7385673522949219,
      "learning_rate": 9.684684684684685e-06,
      "loss": 3.0233,
      "step": 1050
    },
    {
      "epoch": 3.18,
      "grad_norm": 2.0537796020507812,
      "learning_rate": 9.681681681681683e-06,
      "loss": 3.0097,
      "step": 1060
    },
    {
      "epoch": 3.21,
      "grad_norm": 1.9146968126296997,
      "learning_rate": 9.678678678678679e-06,
      "loss": 2.9755,
      "step": 1070
    },
    {
      "epoch": 3.24,
      "grad_norm": 1.7920832633972168,
      "learning_rate": 9.675675675675676e-06,
      "loss": 2.9815,
      "step": 1080
    },
    {
      "epoch": 3.27,
      "grad_norm": 2.126950979232788,
      "learning_rate": 9.672672672672673e-06,
      "loss": 2.991,
      "step": 1090
    },
    {
      "epoch": 3.3,
      "grad_norm": 1.6951396465301514,
      "learning_rate": 9.669669669669671e-06,
      "loss": 2.9565,
      "step": 1100
    },
    {
      "epoch": 3.33,
      "grad_norm": 1.8546791076660156,
      "learning_rate": 9.666666666666667e-06,
      "loss": 2.9878,
      "step": 1110
    },
    {
      "epoch": 3.36,
      "grad_norm": 1.9961843490600586,
      "learning_rate": 9.663663663663664e-06,
      "loss": 2.9606,
      "step": 1120
    },
    {
      "epoch": 3.39,
      "grad_norm": 2.422366142272949,
      "learning_rate": 9.660660660660662e-06,
      "loss": 2.9569,
      "step": 1130
    },
    {
      "epoch": 3.42,
      "grad_norm": 1.9534465074539185,
      "learning_rate": 9.65765765765766e-06,
      "loss": 2.9615,
      "step": 1140
    },
    {
      "epoch": 3.45,
      "grad_norm": 1.9808764457702637,
      "learning_rate": 9.654654654654655e-06,
      "loss": 2.9488,
      "step": 1150
    },
    {
      "epoch": 3.48,
      "grad_norm": 1.8011443614959717,
      "learning_rate": 9.651651651651652e-06,
      "loss": 2.9403,
      "step": 1160
    },
    {
      "epoch": 3.51,
      "grad_norm": 1.944164514541626,
      "learning_rate": 9.64864864864865e-06,
      "loss": 2.901,
      "step": 1170
    },
    {
      "epoch": 3.54,
      "grad_norm": 1.8396971225738525,
      "learning_rate": 9.645645645645646e-06,
      "loss": 2.9152,
      "step": 1180
    },
    {
      "epoch": 3.57,
      "grad_norm": 2.1650946140289307,
      "learning_rate": 9.642642642642643e-06,
      "loss": 2.856,
      "step": 1190
    },
    {
      "epoch": 3.6,
      "grad_norm": 1.9538077116012573,
      "learning_rate": 9.63963963963964e-06,
      "loss": 2.8426,
      "step": 1200
    },
    {
      "epoch": 3.63,
      "grad_norm": 2.1741650104522705,
      "learning_rate": 9.636636636636636e-06,
      "loss": 2.8884,
      "step": 1210
    },
    {
      "epoch": 3.66,
      "grad_norm": 2.2596497535705566,
      "learning_rate": 9.633633633633634e-06,
      "loss": 2.8665,
      "step": 1220
    },
    {
      "epoch": 3.69,
      "grad_norm": 2.1494603157043457,
      "learning_rate": 9.630630630630631e-06,
      "loss": 2.855,
      "step": 1230
    },
    {
      "epoch": 3.72,
      "grad_norm": 1.4631502628326416,
      "learning_rate": 9.627627627627629e-06,
      "loss": 2.8295,
      "step": 1240
    },
    {
      "epoch": 3.75,
      "grad_norm": 2.025465250015259,
      "learning_rate": 9.624624624624626e-06,
      "loss": 2.807,
      "step": 1250
    },
    {
      "epoch": 3.78,
      "grad_norm": 1.8180211782455444,
      "learning_rate": 9.621621621621622e-06,
      "loss": 2.8451,
      "step": 1260
    },
    {
      "epoch": 3.81,
      "grad_norm": 2.007255792617798,
      "learning_rate": 9.61861861861862e-06,
      "loss": 2.7825,
      "step": 1270
    },
    {
      "epoch": 3.84,
      "grad_norm": 1.749640703201294,
      "learning_rate": 9.615615615615617e-06,
      "loss": 2.7795,
      "step": 1280
    },
    {
      "epoch": 3.87,
      "grad_norm": 2.4561126232147217,
      "learning_rate": 9.612612612612613e-06,
      "loss": 2.8017,
      "step": 1290
    },
    {
      "epoch": 3.9,
      "grad_norm": 2.0643820762634277,
      "learning_rate": 9.60960960960961e-06,
      "loss": 2.7468,
      "step": 1300
    },
    {
      "epoch": 3.93,
      "grad_norm": 1.8208361864089966,
      "learning_rate": 9.606606606606608e-06,
      "loss": 2.7544,
      "step": 1310
    },
    {
      "epoch": 3.96,
      "grad_norm": 2.3445229530334473,
      "learning_rate": 9.603603603603604e-06,
      "loss": 2.7689,
      "step": 1320
    },
    {
      "epoch": 3.99,
      "grad_norm": 2.2914915084838867,
      "learning_rate": 9.600600600600601e-06,
      "loss": 2.7858,
      "step": 1330
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.827,
      "eval_loss": 2.5428378582000732,
      "eval_runtime": 25.8397,
      "eval_samples_per_second": 387.001,
      "eval_steps_per_second": 1.548,
      "step": 1332
    },
    {
      "epoch": 4.02,
      "grad_norm": 2.1701862812042236,
      "learning_rate": 9.597597597597599e-06,
      "loss": 2.7277,
      "step": 1340
    },
    {
      "epoch": 4.05,
      "grad_norm": 2.483489990234375,
      "learning_rate": 9.594594594594594e-06,
      "loss": 2.6969,
      "step": 1350
    },
    {
      "epoch": 4.08,
      "grad_norm": 2.0590124130249023,
      "learning_rate": 9.591591591591592e-06,
      "loss": 2.7261,
      "step": 1360
    },
    {
      "epoch": 4.11,
      "grad_norm": 2.3316245079040527,
      "learning_rate": 9.58858858858859e-06,
      "loss": 2.699,
      "step": 1370
    },
    {
      "epoch": 4.14,
      "grad_norm": 1.8884086608886719,
      "learning_rate": 9.585585585585587e-06,
      "loss": 2.6947,
      "step": 1380
    },
    {
      "epoch": 4.17,
      "grad_norm": 1.9525305032730103,
      "learning_rate": 9.582582582582584e-06,
      "loss": 2.66,
      "step": 1390
    },
    {
      "epoch": 4.2,
      "grad_norm": 2.617844581604004,
      "learning_rate": 9.57957957957958e-06,
      "loss": 2.6716,
      "step": 1400
    },
    {
      "epoch": 4.23,
      "grad_norm": 2.178778648376465,
      "learning_rate": 9.576576576576578e-06,
      "loss": 2.7055,
      "step": 1410
    },
    {
      "epoch": 4.26,
      "grad_norm": 2.4651923179626465,
      "learning_rate": 9.573573573573575e-06,
      "loss": 2.6734,
      "step": 1420
    },
    {
      "epoch": 4.29,
      "grad_norm": 2.1029953956604004,
      "learning_rate": 9.57057057057057e-06,
      "loss": 2.65,
      "step": 1430
    },
    {
      "epoch": 4.32,
      "grad_norm": 2.09940767288208,
      "learning_rate": 9.567567567567568e-06,
      "loss": 2.6153,
      "step": 1440
    },
    {
      "epoch": 4.35,
      "grad_norm": 2.0970473289489746,
      "learning_rate": 9.564564564564566e-06,
      "loss": 2.6249,
      "step": 1450
    },
    {
      "epoch": 4.38,
      "grad_norm": 2.4626379013061523,
      "learning_rate": 9.561561561561562e-06,
      "loss": 2.6275,
      "step": 1460
    },
    {
      "epoch": 4.41,
      "grad_norm": 2.286036968231201,
      "learning_rate": 9.558558558558559e-06,
      "loss": 2.6403,
      "step": 1470
    },
    {
      "epoch": 4.44,
      "grad_norm": 1.8655325174331665,
      "learning_rate": 9.555555555555556e-06,
      "loss": 2.5838,
      "step": 1480
    },
    {
      "epoch": 4.47,
      "grad_norm": 2.145075798034668,
      "learning_rate": 9.552552552552552e-06,
      "loss": 2.5676,
      "step": 1490
    },
    {
      "epoch": 4.5,
      "grad_norm": 2.7190279960632324,
      "learning_rate": 9.54954954954955e-06,
      "loss": 2.6144,
      "step": 1500
    },
    {
      "epoch": 4.53,
      "grad_norm": 2.168372869491577,
      "learning_rate": 9.546546546546547e-06,
      "loss": 2.6037,
      "step": 1510
    },
    {
      "epoch": 4.56,
      "grad_norm": 2.2117419242858887,
      "learning_rate": 9.543543543543545e-06,
      "loss": 2.5687,
      "step": 1520
    },
    {
      "epoch": 4.59,
      "grad_norm": 2.153658866882324,
      "learning_rate": 9.540540540540542e-06,
      "loss": 2.5987,
      "step": 1530
    },
    {
      "epoch": 4.62,
      "grad_norm": 2.9541666507720947,
      "learning_rate": 9.537537537537538e-06,
      "loss": 2.5457,
      "step": 1540
    },
    {
      "epoch": 4.65,
      "grad_norm": 2.2983791828155518,
      "learning_rate": 9.534534534534535e-06,
      "loss": 2.5875,
      "step": 1550
    },
    {
      "epoch": 4.68,
      "grad_norm": 2.269871711730957,
      "learning_rate": 9.531531531531533e-06,
      "loss": 2.5002,
      "step": 1560
    },
    {
      "epoch": 4.71,
      "grad_norm": 2.495638370513916,
      "learning_rate": 9.528528528528529e-06,
      "loss": 2.541,
      "step": 1570
    },
    {
      "epoch": 4.74,
      "grad_norm": 2.479295253753662,
      "learning_rate": 9.525525525525526e-06,
      "loss": 2.531,
      "step": 1580
    },
    {
      "epoch": 4.77,
      "grad_norm": 2.627950668334961,
      "learning_rate": 9.522522522522524e-06,
      "loss": 2.5567,
      "step": 1590
    },
    {
      "epoch": 4.8,
      "grad_norm": 2.0522141456604004,
      "learning_rate": 9.51951951951952e-06,
      "loss": 2.5528,
      "step": 1600
    },
    {
      "epoch": 4.83,
      "grad_norm": 2.1786510944366455,
      "learning_rate": 9.516516516516517e-06,
      "loss": 2.4905,
      "step": 1610
    },
    {
      "epoch": 4.86,
      "grad_norm": 1.8431909084320068,
      "learning_rate": 9.513513513513514e-06,
      "loss": 2.5162,
      "step": 1620
    },
    {
      "epoch": 4.89,
      "grad_norm": 3.211226224899292,
      "learning_rate": 9.510510510510512e-06,
      "loss": 2.4975,
      "step": 1630
    },
    {
      "epoch": 4.92,
      "grad_norm": 2.483825922012329,
      "learning_rate": 9.507507507507508e-06,
      "loss": 2.495,
      "step": 1640
    },
    {
      "epoch": 4.95,
      "grad_norm": 2.4815008640289307,
      "learning_rate": 9.504504504504505e-06,
      "loss": 2.5158,
      "step": 1650
    },
    {
      "epoch": 4.98,
      "grad_norm": 2.0666556358337402,
      "learning_rate": 9.501501501501503e-06,
      "loss": 2.4372,
      "step": 1660
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.8428,
      "eval_loss": 2.2353405952453613,
      "eval_runtime": 26.4297,
      "eval_samples_per_second": 378.362,
      "eval_steps_per_second": 1.513,
      "step": 1665
    },
    {
      "epoch": 5.02,
      "grad_norm": 2.162529468536377,
      "learning_rate": 9.4984984984985e-06,
      "loss": 2.3732,
      "step": 1670
    },
    {
      "epoch": 5.05,
      "grad_norm": 2.6510798931121826,
      "learning_rate": 9.495495495495496e-06,
      "loss": 2.4363,
      "step": 1680
    },
    {
      "epoch": 5.08,
      "grad_norm": 2.430441379547119,
      "learning_rate": 9.492492492492493e-06,
      "loss": 2.4362,
      "step": 1690
    },
    {
      "epoch": 5.11,
      "grad_norm": 2.468442678451538,
      "learning_rate": 9.489489489489491e-06,
      "loss": 2.4535,
      "step": 1700
    },
    {
      "epoch": 5.14,
      "grad_norm": 2.7464473247528076,
      "learning_rate": 9.486486486486487e-06,
      "loss": 2.3854,
      "step": 1710
    },
    {
      "epoch": 5.17,
      "grad_norm": 2.607032537460327,
      "learning_rate": 9.483483483483484e-06,
      "loss": 2.3926,
      "step": 1720
    },
    {
      "epoch": 5.2,
      "grad_norm": 2.357286214828491,
      "learning_rate": 9.480480480480482e-06,
      "loss": 2.433,
      "step": 1730
    },
    {
      "epoch": 5.23,
      "grad_norm": 2.7163023948669434,
      "learning_rate": 9.477477477477477e-06,
      "loss": 2.4164,
      "step": 1740
    },
    {
      "epoch": 5.26,
      "grad_norm": 2.316063404083252,
      "learning_rate": 9.474474474474475e-06,
      "loss": 2.3618,
      "step": 1750
    },
    {
      "epoch": 5.29,
      "grad_norm": 2.23887300491333,
      "learning_rate": 9.471471471471472e-06,
      "loss": 2.3878,
      "step": 1760
    },
    {
      "epoch": 5.32,
      "grad_norm": 2.347841262817383,
      "learning_rate": 9.46846846846847e-06,
      "loss": 2.3571,
      "step": 1770
    },
    {
      "epoch": 5.35,
      "grad_norm": 2.708570718765259,
      "learning_rate": 9.465465465465467e-06,
      "loss": 2.3421,
      "step": 1780
    },
    {
      "epoch": 5.38,
      "grad_norm": 2.5647084712982178,
      "learning_rate": 9.462462462462463e-06,
      "loss": 2.347,
      "step": 1790
    },
    {
      "epoch": 5.41,
      "grad_norm": 2.5164411067962646,
      "learning_rate": 9.45945945945946e-06,
      "loss": 2.3584,
      "step": 1800
    },
    {
      "epoch": 5.44,
      "grad_norm": 4.1986541748046875,
      "learning_rate": 9.456456456456458e-06,
      "loss": 2.3429,
      "step": 1810
    },
    {
      "epoch": 5.47,
      "grad_norm": 2.8616836071014404,
      "learning_rate": 9.453453453453454e-06,
      "loss": 2.3678,
      "step": 1820
    },
    {
      "epoch": 5.5,
      "grad_norm": 3.228412628173828,
      "learning_rate": 9.450450450450451e-06,
      "loss": 2.2858,
      "step": 1830
    },
    {
      "epoch": 5.53,
      "grad_norm": 2.5712051391601562,
      "learning_rate": 9.447447447447449e-06,
      "loss": 2.2745,
      "step": 1840
    },
    {
      "epoch": 5.56,
      "grad_norm": 2.5423223972320557,
      "learning_rate": 9.444444444444445e-06,
      "loss": 2.323,
      "step": 1850
    },
    {
      "epoch": 5.59,
      "grad_norm": 2.4743051528930664,
      "learning_rate": 9.441441441441442e-06,
      "loss": 2.2725,
      "step": 1860
    },
    {
      "epoch": 5.62,
      "grad_norm": 3.4513115882873535,
      "learning_rate": 9.43843843843844e-06,
      "loss": 2.3433,
      "step": 1870
    },
    {
      "epoch": 5.65,
      "grad_norm": 3.3393611907958984,
      "learning_rate": 9.435435435435435e-06,
      "loss": 2.3089,
      "step": 1880
    },
    {
      "epoch": 5.68,
      "grad_norm": 2.2898778915405273,
      "learning_rate": 9.432432432432433e-06,
      "loss": 2.2618,
      "step": 1890
    },
    {
      "epoch": 5.71,
      "grad_norm": 3.142050266265869,
      "learning_rate": 9.42942942942943e-06,
      "loss": 2.3027,
      "step": 1900
    },
    {
      "epoch": 5.74,
      "grad_norm": 2.540493965148926,
      "learning_rate": 9.426426426426428e-06,
      "loss": 2.2619,
      "step": 1910
    },
    {
      "epoch": 5.77,
      "grad_norm": 1.9641002416610718,
      "learning_rate": 9.423423423423425e-06,
      "loss": 2.2468,
      "step": 1920
    },
    {
      "epoch": 5.8,
      "grad_norm": 3.1919631958007812,
      "learning_rate": 9.420420420420421e-06,
      "loss": 2.3087,
      "step": 1930
    },
    {
      "epoch": 5.83,
      "grad_norm": 2.1615185737609863,
      "learning_rate": 9.417417417417418e-06,
      "loss": 2.2576,
      "step": 1940
    },
    {
      "epoch": 5.86,
      "grad_norm": 2.5164833068847656,
      "learning_rate": 9.414414414414416e-06,
      "loss": 2.2505,
      "step": 1950
    },
    {
      "epoch": 5.89,
      "grad_norm": 3.334547281265259,
      "learning_rate": 9.411411411411412e-06,
      "loss": 2.2617,
      "step": 1960
    },
    {
      "epoch": 5.92,
      "grad_norm": 2.6057589054107666,
      "learning_rate": 9.40840840840841e-06,
      "loss": 2.2157,
      "step": 1970
    },
    {
      "epoch": 5.95,
      "grad_norm": 2.547347068786621,
      "learning_rate": 9.405405405405407e-06,
      "loss": 2.2166,
      "step": 1980
    },
    {
      "epoch": 5.98,
      "grad_norm": 2.4648900032043457,
      "learning_rate": 9.402402402402402e-06,
      "loss": 2.2157,
      "step": 1990
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.8568,
      "eval_loss": 1.9597220420837402,
      "eval_runtime": 25.9049,
      "eval_samples_per_second": 386.028,
      "eval_steps_per_second": 1.544,
      "step": 1998
    },
    {
      "epoch": 6.01,
      "grad_norm": 2.6400487422943115,
      "learning_rate": 9.3993993993994e-06,
      "loss": 2.275,
      "step": 2000
    },
    {
      "epoch": 6.04,
      "grad_norm": 3.2325823307037354,
      "learning_rate": 9.396396396396397e-06,
      "loss": 2.2424,
      "step": 2010
    },
    {
      "epoch": 6.07,
      "grad_norm": 2.2359609603881836,
      "learning_rate": 9.393393393393393e-06,
      "loss": 2.1585,
      "step": 2020
    },
    {
      "epoch": 6.1,
      "grad_norm": 2.844468116760254,
      "learning_rate": 9.39039039039039e-06,
      "loss": 2.1369,
      "step": 2030
    },
    {
      "epoch": 6.13,
      "grad_norm": 2.3116044998168945,
      "learning_rate": 9.387387387387388e-06,
      "loss": 2.1565,
      "step": 2040
    },
    {
      "epoch": 6.16,
      "grad_norm": 4.145083904266357,
      "learning_rate": 9.384384384384386e-06,
      "loss": 2.1671,
      "step": 2050
    },
    {
      "epoch": 6.19,
      "grad_norm": 2.8794326782226562,
      "learning_rate": 9.381381381381383e-06,
      "loss": 2.1543,
      "step": 2060
    },
    {
      "epoch": 6.22,
      "grad_norm": 3.18340802192688,
      "learning_rate": 9.378378378378379e-06,
      "loss": 2.1705,
      "step": 2070
    },
    {
      "epoch": 6.25,
      "grad_norm": 2.4386444091796875,
      "learning_rate": 9.375375375375376e-06,
      "loss": 2.1628,
      "step": 2080
    },
    {
      "epoch": 6.28,
      "grad_norm": 2.837346315383911,
      "learning_rate": 9.372372372372374e-06,
      "loss": 2.1315,
      "step": 2090
    },
    {
      "epoch": 6.31,
      "grad_norm": 2.193432092666626,
      "learning_rate": 9.36936936936937e-06,
      "loss": 2.1403,
      "step": 2100
    },
    {
      "epoch": 6.34,
      "grad_norm": 3.1398682594299316,
      "learning_rate": 9.366366366366367e-06,
      "loss": 2.1258,
      "step": 2110
    },
    {
      "epoch": 6.37,
      "grad_norm": 2.416149377822876,
      "learning_rate": 9.363363363363363e-06,
      "loss": 2.1114,
      "step": 2120
    },
    {
      "epoch": 6.4,
      "grad_norm": 3.2373199462890625,
      "learning_rate": 9.36036036036036e-06,
      "loss": 2.1422,
      "step": 2130
    },
    {
      "epoch": 6.43,
      "grad_norm": 2.6956839561462402,
      "learning_rate": 9.357357357357358e-06,
      "loss": 2.0692,
      "step": 2140
    },
    {
      "epoch": 6.46,
      "grad_norm": 3.576822280883789,
      "learning_rate": 9.354354354354355e-06,
      "loss": 2.136,
      "step": 2150
    },
    {
      "epoch": 6.49,
      "grad_norm": 2.1885507106781006,
      "learning_rate": 9.351351351351353e-06,
      "loss": 2.1256,
      "step": 2160
    },
    {
      "epoch": 6.52,
      "grad_norm": 3.042860507965088,
      "learning_rate": 9.34834834834835e-06,
      "loss": 2.1085,
      "step": 2170
    },
    {
      "epoch": 6.55,
      "grad_norm": 3.4762566089630127,
      "learning_rate": 9.345345345345346e-06,
      "loss": 2.0583,
      "step": 2180
    },
    {
      "epoch": 6.58,
      "grad_norm": 3.6814165115356445,
      "learning_rate": 9.342342342342344e-06,
      "loss": 2.1014,
      "step": 2190
    },
    {
      "epoch": 6.61,
      "grad_norm": 2.7193524837493896,
      "learning_rate": 9.339339339339341e-06,
      "loss": 2.0804,
      "step": 2200
    },
    {
      "epoch": 6.64,
      "grad_norm": 2.850829601287842,
      "learning_rate": 9.336336336336337e-06,
      "loss": 2.0757,
      "step": 2210
    },
    {
      "epoch": 6.67,
      "grad_norm": 2.2540390491485596,
      "learning_rate": 9.333333333333334e-06,
      "loss": 2.0607,
      "step": 2220
    },
    {
      "epoch": 6.7,
      "grad_norm": 2.819641351699829,
      "learning_rate": 9.330330330330332e-06,
      "loss": 2.0624,
      "step": 2230
    },
    {
      "epoch": 6.73,
      "grad_norm": 2.1947712898254395,
      "learning_rate": 9.327327327327328e-06,
      "loss": 2.0198,
      "step": 2240
    },
    {
      "epoch": 6.76,
      "grad_norm": 2.8388874530792236,
      "learning_rate": 9.324324324324325e-06,
      "loss": 1.9486,
      "step": 2250
    },
    {
      "epoch": 6.79,
      "grad_norm": 2.8357861042022705,
      "learning_rate": 9.321321321321321e-06,
      "loss": 2.02,
      "step": 2260
    },
    {
      "epoch": 6.82,
      "grad_norm": 2.7252373695373535,
      "learning_rate": 9.318318318318318e-06,
      "loss": 1.9955,
      "step": 2270
    },
    {
      "epoch": 6.85,
      "grad_norm": 3.4677445888519287,
      "learning_rate": 9.315315315315316e-06,
      "loss": 2.0506,
      "step": 2280
    },
    {
      "epoch": 6.88,
      "grad_norm": 3.0668785572052,
      "learning_rate": 9.312312312312313e-06,
      "loss": 1.9839,
      "step": 2290
    },
    {
      "epoch": 6.91,
      "grad_norm": 2.9034082889556885,
      "learning_rate": 9.30930930930931e-06,
      "loss": 1.9957,
      "step": 2300
    },
    {
      "epoch": 6.94,
      "grad_norm": 2.9479753971099854,
      "learning_rate": 9.306306306306308e-06,
      "loss": 2.0329,
      "step": 2310
    },
    {
      "epoch": 6.97,
      "grad_norm": 3.0393028259277344,
      "learning_rate": 9.303303303303304e-06,
      "loss": 2.0185,
      "step": 2320
    },
    {
      "epoch": 7.0,
      "grad_norm": 3.204620361328125,
      "learning_rate": 9.300300300300302e-06,
      "loss": 1.9573,
      "step": 2330
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.8685,
      "eval_loss": 1.7240691184997559,
      "eval_runtime": 26.3924,
      "eval_samples_per_second": 378.897,
      "eval_steps_per_second": 1.516,
      "step": 2331
    },
    {
      "epoch": 7.03,
      "grad_norm": 3.030730962753296,
      "learning_rate": 9.297297297297299e-06,
      "loss": 1.9975,
      "step": 2340
    },
    {
      "epoch": 7.06,
      "grad_norm": 2.9937710762023926,
      "learning_rate": 9.294294294294295e-06,
      "loss": 1.9828,
      "step": 2350
    },
    {
      "epoch": 7.09,
      "grad_norm": 2.864546775817871,
      "learning_rate": 9.291291291291292e-06,
      "loss": 2.0085,
      "step": 2360
    },
    {
      "epoch": 7.12,
      "grad_norm": 2.979231357574463,
      "learning_rate": 9.288288288288288e-06,
      "loss": 1.891,
      "step": 2370
    },
    {
      "epoch": 7.15,
      "grad_norm": 2.9546091556549072,
      "learning_rate": 9.285285285285286e-06,
      "loss": 1.9396,
      "step": 2380
    },
    {
      "epoch": 7.18,
      "grad_norm": 2.627347469329834,
      "learning_rate": 9.282282282282283e-06,
      "loss": 1.9539,
      "step": 2390
    },
    {
      "epoch": 7.21,
      "grad_norm": 3.009409189224243,
      "learning_rate": 9.27927927927928e-06,
      "loss": 1.992,
      "step": 2400
    },
    {
      "epoch": 7.24,
      "grad_norm": 2.4301156997680664,
      "learning_rate": 9.276276276276276e-06,
      "loss": 1.8915,
      "step": 2410
    },
    {
      "epoch": 7.27,
      "grad_norm": 2.8977396488189697,
      "learning_rate": 9.273273273273274e-06,
      "loss": 1.9001,
      "step": 2420
    },
    {
      "epoch": 7.3,
      "grad_norm": 3.451801300048828,
      "learning_rate": 9.270270270270271e-06,
      "loss": 1.8914,
      "step": 2430
    },
    {
      "epoch": 7.33,
      "grad_norm": 3.6057212352752686,
      "learning_rate": 9.267267267267269e-06,
      "loss": 1.8743,
      "step": 2440
    },
    {
      "epoch": 7.36,
      "grad_norm": 3.5656979084014893,
      "learning_rate": 9.264264264264266e-06,
      "loss": 1.9253,
      "step": 2450
    },
    {
      "epoch": 7.39,
      "grad_norm": 2.9827640056610107,
      "learning_rate": 9.261261261261262e-06,
      "loss": 1.8978,
      "step": 2460
    },
    {
      "epoch": 7.42,
      "grad_norm": 3.191035032272339,
      "learning_rate": 9.25825825825826e-06,
      "loss": 1.8851,
      "step": 2470
    },
    {
      "epoch": 7.45,
      "grad_norm": 3.8623464107513428,
      "learning_rate": 9.255255255255255e-06,
      "loss": 1.8979,
      "step": 2480
    },
    {
      "epoch": 7.48,
      "grad_norm": 2.779446840286255,
      "learning_rate": 9.252252252252253e-06,
      "loss": 1.867,
      "step": 2490
    },
    {
      "epoch": 7.51,
      "grad_norm": 4.140806674957275,
      "learning_rate": 9.24924924924925e-06,
      "loss": 1.9244,
      "step": 2500
    },
    {
      "epoch": 7.54,
      "grad_norm": 3.2873432636260986,
      "learning_rate": 9.246246246246246e-06,
      "loss": 1.8786,
      "step": 2510
    },
    {
      "epoch": 7.57,
      "grad_norm": 3.417663812637329,
      "learning_rate": 9.243243243243243e-06,
      "loss": 1.8828,
      "step": 2520
    },
    {
      "epoch": 7.6,
      "grad_norm": 3.7559661865234375,
      "learning_rate": 9.240240240240241e-06,
      "loss": 1.8255,
      "step": 2530
    },
    {
      "epoch": 7.63,
      "grad_norm": 3.8991100788116455,
      "learning_rate": 9.237237237237238e-06,
      "loss": 1.8616,
      "step": 2540
    },
    {
      "epoch": 7.66,
      "grad_norm": 2.5654900074005127,
      "learning_rate": 9.234234234234236e-06,
      "loss": 1.8223,
      "step": 2550
    },
    {
      "epoch": 7.69,
      "grad_norm": 3.494316577911377,
      "learning_rate": 9.231231231231232e-06,
      "loss": 1.818,
      "step": 2560
    },
    {
      "epoch": 7.72,
      "grad_norm": 3.2533693313598633,
      "learning_rate": 9.228228228228229e-06,
      "loss": 1.8748,
      "step": 2570
    },
    {
      "epoch": 7.75,
      "grad_norm": 3.4433581829071045,
      "learning_rate": 9.225225225225227e-06,
      "loss": 1.8446,
      "step": 2580
    },
    {
      "epoch": 7.78,
      "grad_norm": 3.0548644065856934,
      "learning_rate": 9.222222222222224e-06,
      "loss": 1.8367,
      "step": 2590
    },
    {
      "epoch": 7.81,
      "grad_norm": 3.341350793838501,
      "learning_rate": 9.21921921921922e-06,
      "loss": 1.8764,
      "step": 2600
    },
    {
      "epoch": 7.84,
      "grad_norm": 3.763580560684204,
      "learning_rate": 9.216216216216217e-06,
      "loss": 1.8317,
      "step": 2610
    },
    {
      "epoch": 7.87,
      "grad_norm": 3.323333263397217,
      "learning_rate": 9.213213213213213e-06,
      "loss": 1.8128,
      "step": 2620
    },
    {
      "epoch": 7.9,
      "grad_norm": 2.7645206451416016,
      "learning_rate": 9.21021021021021e-06,
      "loss": 1.8034,
      "step": 2630
    },
    {
      "epoch": 7.93,
      "grad_norm": 4.599252223968506,
      "learning_rate": 9.207207207207208e-06,
      "loss": 1.7933,
      "step": 2640
    },
    {
      "epoch": 7.96,
      "grad_norm": 2.671976089477539,
      "learning_rate": 9.204204204204204e-06,
      "loss": 1.7799,
      "step": 2650
    },
    {
      "epoch": 7.99,
      "grad_norm": 4.781398296356201,
      "learning_rate": 9.201201201201201e-06,
      "loss": 1.841,
      "step": 2660
    },
    {
      "epoch": 8.0,
      "eval_accuracy": 0.8736,
      "eval_loss": 1.5209704637527466,
      "eval_runtime": 26.2803,
      "eval_samples_per_second": 380.513,
      "eval_steps_per_second": 1.522,
      "step": 2664
    },
    {
      "epoch": 8.02,
      "grad_norm": 2.8974978923797607,
      "learning_rate": 9.198198198198199e-06,
      "loss": 1.7335,
      "step": 2670
    },
    {
      "epoch": 8.05,
      "grad_norm": 3.6113295555114746,
      "learning_rate": 9.195195195195196e-06,
      "loss": 1.7897,
      "step": 2680
    },
    {
      "epoch": 8.08,
      "grad_norm": 3.8488831520080566,
      "learning_rate": 9.192192192192194e-06,
      "loss": 1.7668,
      "step": 2690
    },
    {
      "epoch": 8.11,
      "grad_norm": 2.893803834915161,
      "learning_rate": 9.189189189189191e-06,
      "loss": 1.7553,
      "step": 2700
    },
    {
      "epoch": 8.14,
      "grad_norm": 3.7040750980377197,
      "learning_rate": 9.186186186186187e-06,
      "loss": 1.7481,
      "step": 2710
    },
    {
      "epoch": 8.17,
      "grad_norm": 3.1729960441589355,
      "learning_rate": 9.183183183183185e-06,
      "loss": 1.7584,
      "step": 2720
    },
    {
      "epoch": 8.2,
      "grad_norm": 2.709282875061035,
      "learning_rate": 9.18018018018018e-06,
      "loss": 1.7882,
      "step": 2730
    },
    {
      "epoch": 8.23,
      "grad_norm": 3.1258764266967773,
      "learning_rate": 9.177177177177178e-06,
      "loss": 1.7394,
      "step": 2740
    },
    {
      "epoch": 8.26,
      "grad_norm": 3.4164059162139893,
      "learning_rate": 9.174174174174175e-06,
      "loss": 1.7657,
      "step": 2750
    },
    {
      "epoch": 8.29,
      "grad_norm": 4.909099578857422,
      "learning_rate": 9.171171171171171e-06,
      "loss": 1.7344,
      "step": 2760
    },
    {
      "epoch": 8.32,
      "grad_norm": 3.1216490268707275,
      "learning_rate": 9.168168168168169e-06,
      "loss": 1.714,
      "step": 2770
    },
    {
      "epoch": 8.35,
      "grad_norm": 2.712817430496216,
      "learning_rate": 9.165165165165166e-06,
      "loss": 1.7288,
      "step": 2780
    },
    {
      "epoch": 8.38,
      "grad_norm": 3.531217098236084,
      "learning_rate": 9.162162162162162e-06,
      "loss": 1.7329,
      "step": 2790
    },
    {
      "epoch": 8.41,
      "grad_norm": 3.097011089324951,
      "learning_rate": 9.15915915915916e-06,
      "loss": 1.7194,
      "step": 2800
    },
    {
      "epoch": 8.44,
      "grad_norm": 3.811913251876831,
      "learning_rate": 9.156156156156157e-06,
      "loss": 1.7242,
      "step": 2810
    },
    {
      "epoch": 8.47,
      "grad_norm": 3.676722288131714,
      "learning_rate": 9.153153153153154e-06,
      "loss": 1.6821,
      "step": 2820
    },
    {
      "epoch": 8.5,
      "grad_norm": 3.473322868347168,
      "learning_rate": 9.150150150150152e-06,
      "loss": 1.6764,
      "step": 2830
    },
    {
      "epoch": 8.53,
      "grad_norm": 3.613896369934082,
      "learning_rate": 9.147147147147147e-06,
      "loss": 1.6737,
      "step": 2840
    },
    {
      "epoch": 8.56,
      "grad_norm": 3.478484869003296,
      "learning_rate": 9.144144144144145e-06,
      "loss": 1.6648,
      "step": 2850
    },
    {
      "epoch": 8.59,
      "grad_norm": 3.295175313949585,
      "learning_rate": 9.141141141141142e-06,
      "loss": 1.6682,
      "step": 2860
    },
    {
      "epoch": 8.62,
      "grad_norm": 3.504593849182129,
      "learning_rate": 9.138138138138138e-06,
      "loss": 1.641,
      "step": 2870
    },
    {
      "epoch": 8.65,
      "grad_norm": 3.8884923458099365,
      "learning_rate": 9.135135135135136e-06,
      "loss": 1.6504,
      "step": 2880
    },
    {
      "epoch": 8.68,
      "grad_norm": 3.4164955615997314,
      "learning_rate": 9.132132132132133e-06,
      "loss": 1.6809,
      "step": 2890
    },
    {
      "epoch": 8.71,
      "grad_norm": 2.9317514896392822,
      "learning_rate": 9.129129129129129e-06,
      "loss": 1.6526,
      "step": 2900
    },
    {
      "epoch": 8.74,
      "grad_norm": 3.4553208351135254,
      "learning_rate": 9.126126126126126e-06,
      "loss": 1.6936,
      "step": 2910
    },
    {
      "epoch": 8.77,
      "grad_norm": 3.3463358879089355,
      "learning_rate": 9.123123123123124e-06,
      "loss": 1.6249,
      "step": 2920
    },
    {
      "epoch": 8.8,
      "grad_norm": 2.9135587215423584,
      "learning_rate": 9.120120120120121e-06,
      "loss": 1.6142,
      "step": 2930
    },
    {
      "epoch": 8.83,
      "grad_norm": 3.325805902481079,
      "learning_rate": 9.117117117117117e-06,
      "loss": 1.6483,
      "step": 2940
    },
    {
      "epoch": 8.86,
      "grad_norm": 3.7595407962799072,
      "learning_rate": 9.114114114114115e-06,
      "loss": 1.5844,
      "step": 2950
    },
    {
      "epoch": 8.89,
      "grad_norm": 3.486954927444458,
      "learning_rate": 9.111111111111112e-06,
      "loss": 1.6402,
      "step": 2960
    },
    {
      "epoch": 8.92,
      "grad_norm": 3.0209648609161377,
      "learning_rate": 9.10810810810811e-06,
      "loss": 1.6557,
      "step": 2970
    },
    {
      "epoch": 8.95,
      "grad_norm": 3.117098093032837,
      "learning_rate": 9.105105105105105e-06,
      "loss": 1.613,
      "step": 2980
    },
    {
      "epoch": 8.98,
      "grad_norm": 4.14963436126709,
      "learning_rate": 9.102102102102103e-06,
      "loss": 1.6085,
      "step": 2990
    },
    {
      "epoch": 9.0,
      "eval_accuracy": 0.8832,
      "eval_loss": 1.3363327980041504,
      "eval_runtime": 25.8706,
      "eval_samples_per_second": 386.54,
      "eval_steps_per_second": 1.546,
      "step": 2997
    },
    {
      "epoch": 9.01,
      "grad_norm": 3.161076784133911,
      "learning_rate": 9.0990990990991e-06,
      "loss": 1.7363,
      "step": 3000
    },
    {
      "epoch": 9.04,
      "grad_norm": 3.7530171871185303,
      "learning_rate": 9.096096096096096e-06,
      "loss": 1.6056,
      "step": 3010
    },
    {
      "epoch": 9.07,
      "grad_norm": 3.0861854553222656,
      "learning_rate": 9.093093093093094e-06,
      "loss": 1.6208,
      "step": 3020
    },
    {
      "epoch": 9.1,
      "grad_norm": 4.3751702308654785,
      "learning_rate": 9.090090090090091e-06,
      "loss": 1.582,
      "step": 3030
    },
    {
      "epoch": 9.13,
      "grad_norm": 2.86613130569458,
      "learning_rate": 9.087087087087087e-06,
      "loss": 1.5261,
      "step": 3040
    },
    {
      "epoch": 9.16,
      "grad_norm": 2.8464765548706055,
      "learning_rate": 9.084084084084084e-06,
      "loss": 1.5857,
      "step": 3050
    },
    {
      "epoch": 9.19,
      "grad_norm": 2.6522128582000732,
      "learning_rate": 9.081081081081082e-06,
      "loss": 1.66,
      "step": 3060
    },
    {
      "epoch": 9.22,
      "grad_norm": 3.4797539710998535,
      "learning_rate": 9.07807807807808e-06,
      "loss": 1.587,
      "step": 3070
    },
    {
      "epoch": 9.25,
      "grad_norm": 3.0189497470855713,
      "learning_rate": 9.075075075075077e-06,
      "loss": 1.5915,
      "step": 3080
    },
    {
      "epoch": 9.28,
      "grad_norm": 5.351146221160889,
      "learning_rate": 9.072072072072073e-06,
      "loss": 1.5165,
      "step": 3090
    },
    {
      "epoch": 9.31,
      "grad_norm": 3.7793538570404053,
      "learning_rate": 9.06906906906907e-06,
      "loss": 1.5507,
      "step": 3100
    },
    {
      "epoch": 9.34,
      "grad_norm": 3.4388442039489746,
      "learning_rate": 9.066066066066068e-06,
      "loss": 1.5718,
      "step": 3110
    },
    {
      "epoch": 9.37,
      "grad_norm": 3.262924909591675,
      "learning_rate": 9.063063063063063e-06,
      "loss": 1.612,
      "step": 3120
    },
    {
      "epoch": 9.4,
      "grad_norm": 5.9477386474609375,
      "learning_rate": 9.06006006006006e-06,
      "loss": 1.5597,
      "step": 3130
    },
    {
      "epoch": 9.43,
      "grad_norm": 3.848048448562622,
      "learning_rate": 9.057057057057058e-06,
      "loss": 1.5803,
      "step": 3140
    },
    {
      "epoch": 9.46,
      "grad_norm": 4.263827800750732,
      "learning_rate": 9.054054054054054e-06,
      "loss": 1.5315,
      "step": 3150
    },
    {
      "epoch": 9.49,
      "grad_norm": 5.525787830352783,
      "learning_rate": 9.051051051051052e-06,
      "loss": 1.5075,
      "step": 3160
    },
    {
      "epoch": 9.52,
      "grad_norm": 2.918323516845703,
      "learning_rate": 9.048048048048049e-06,
      "loss": 1.4725,
      "step": 3170
    },
    {
      "epoch": 9.55,
      "grad_norm": 6.338906764984131,
      "learning_rate": 9.045045045045045e-06,
      "loss": 1.594,
      "step": 3180
    },
    {
      "epoch": 9.58,
      "grad_norm": 3.055828809738159,
      "learning_rate": 9.042042042042042e-06,
      "loss": 1.5254,
      "step": 3190
    },
    {
      "epoch": 9.61,
      "grad_norm": 3.1687676906585693,
      "learning_rate": 9.03903903903904e-06,
      "loss": 1.5004,
      "step": 3200
    },
    {
      "epoch": 9.64,
      "grad_norm": 2.2005269527435303,
      "learning_rate": 9.036036036036037e-06,
      "loss": 1.502,
      "step": 3210
    },
    {
      "epoch": 9.67,
      "grad_norm": 3.7019267082214355,
      "learning_rate": 9.033033033033035e-06,
      "loss": 1.5771,
      "step": 3220
    },
    {
      "epoch": 9.7,
      "grad_norm": 4.179267883300781,
      "learning_rate": 9.03003003003003e-06,
      "loss": 1.506,
      "step": 3230
    },
    {
      "epoch": 9.73,
      "grad_norm": 4.042637825012207,
      "learning_rate": 9.027027027027028e-06,
      "loss": 1.5327,
      "step": 3240
    },
    {
      "epoch": 9.76,
      "grad_norm": 5.137342929840088,
      "learning_rate": 9.024024024024025e-06,
      "loss": 1.5486,
      "step": 3250
    },
    {
      "epoch": 9.79,
      "grad_norm": 3.571098566055298,
      "learning_rate": 9.021021021021021e-06,
      "loss": 1.4718,
      "step": 3260
    },
    {
      "epoch": 9.82,
      "grad_norm": 3.28196382522583,
      "learning_rate": 9.018018018018019e-06,
      "loss": 1.5306,
      "step": 3270
    },
    {
      "epoch": 9.85,
      "grad_norm": 2.9630990028381348,
      "learning_rate": 9.015015015015016e-06,
      "loss": 1.4988,
      "step": 3280
    },
    {
      "epoch": 9.88,
      "grad_norm": 4.726624965667725,
      "learning_rate": 9.012012012012012e-06,
      "loss": 1.4633,
      "step": 3290
    },
    {
      "epoch": 9.91,
      "grad_norm": 3.4337375164031982,
      "learning_rate": 9.00900900900901e-06,
      "loss": 1.4439,
      "step": 3300
    },
    {
      "epoch": 9.94,
      "grad_norm": 4.365289688110352,
      "learning_rate": 9.006006006006007e-06,
      "loss": 1.498,
      "step": 3310
    },
    {
      "epoch": 9.97,
      "grad_norm": 3.175220012664795,
      "learning_rate": 9.003003003003003e-06,
      "loss": 1.4573,
      "step": 3320
    },
    {
      "epoch": 10.0,
      "grad_norm": 3.100367307662964,
      "learning_rate": 9e-06,
      "loss": 1.4188,
      "step": 3330
    },
    {
      "epoch": 10.0,
      "eval_accuracy": 0.8861,
      "eval_loss": 1.1857481002807617,
      "eval_runtime": 25.9956,
      "eval_samples_per_second": 384.681,
      "eval_steps_per_second": 1.539,
      "step": 3330
    },
    {
      "epoch": 10.03,
      "grad_norm": 3.940972328186035,
      "learning_rate": 8.996996996996998e-06,
      "loss": 1.4333,
      "step": 3340
    },
    {
      "epoch": 10.06,
      "grad_norm": 4.0231499671936035,
      "learning_rate": 8.993993993993995e-06,
      "loss": 1.4701,
      "step": 3350
    },
    {
      "epoch": 10.09,
      "grad_norm": 3.5897438526153564,
      "learning_rate": 8.990990990990993e-06,
      "loss": 1.4368,
      "step": 3360
    },
    {
      "epoch": 10.12,
      "grad_norm": 4.010433673858643,
      "learning_rate": 8.987987987987988e-06,
      "loss": 1.4501,
      "step": 3370
    },
    {
      "epoch": 10.15,
      "grad_norm": 3.788407325744629,
      "learning_rate": 8.984984984984986e-06,
      "loss": 1.4382,
      "step": 3380
    },
    {
      "epoch": 10.18,
      "grad_norm": 3.2629690170288086,
      "learning_rate": 8.981981981981983e-06,
      "loss": 1.4122,
      "step": 3390
    },
    {
      "epoch": 10.21,
      "grad_norm": 4.393489360809326,
      "learning_rate": 8.97897897897898e-06,
      "loss": 1.4517,
      "step": 3400
    },
    {
      "epoch": 10.24,
      "grad_norm": 3.116422653198242,
      "learning_rate": 8.975975975975977e-06,
      "loss": 1.4486,
      "step": 3410
    },
    {
      "epoch": 10.27,
      "grad_norm": 3.059415340423584,
      "learning_rate": 8.972972972972974e-06,
      "loss": 1.4273,
      "step": 3420
    },
    {
      "epoch": 10.3,
      "grad_norm": 3.2235398292541504,
      "learning_rate": 8.96996996996997e-06,
      "loss": 1.4208,
      "step": 3430
    },
    {
      "epoch": 10.33,
      "grad_norm": 3.3876900672912598,
      "learning_rate": 8.966966966966967e-06,
      "loss": 1.4676,
      "step": 3440
    },
    {
      "epoch": 10.36,
      "grad_norm": 4.044349193572998,
      "learning_rate": 8.963963963963965e-06,
      "loss": 1.4552,
      "step": 3450
    },
    {
      "epoch": 10.39,
      "grad_norm": 3.1774368286132812,
      "learning_rate": 8.960960960960962e-06,
      "loss": 1.3933,
      "step": 3460
    },
    {
      "epoch": 10.42,
      "grad_norm": 3.186774253845215,
      "learning_rate": 8.957957957957958e-06,
      "loss": 1.4075,
      "step": 3470
    },
    {
      "epoch": 10.45,
      "grad_norm": 4.519390106201172,
      "learning_rate": 8.954954954954956e-06,
      "loss": 1.4225,
      "step": 3480
    },
    {
      "epoch": 10.48,
      "grad_norm": 3.620778799057007,
      "learning_rate": 8.951951951951953e-06,
      "loss": 1.4318,
      "step": 3490
    },
    {
      "epoch": 10.51,
      "grad_norm": 3.7709479331970215,
      "learning_rate": 8.94894894894895e-06,
      "loss": 1.4314,
      "step": 3500
    },
    {
      "epoch": 10.54,
      "grad_norm": 4.269083023071289,
      "learning_rate": 8.945945945945946e-06,
      "loss": 1.3506,
      "step": 3510
    },
    {
      "epoch": 10.57,
      "grad_norm": 3.909588098526001,
      "learning_rate": 8.942942942942944e-06,
      "loss": 1.3592,
      "step": 3520
    },
    {
      "epoch": 10.6,
      "grad_norm": 7.832637310028076,
      "learning_rate": 8.939939939939941e-06,
      "loss": 1.4296,
      "step": 3530
    },
    {
      "epoch": 10.63,
      "grad_norm": 4.4831719398498535,
      "learning_rate": 8.936936936936937e-06,
      "loss": 1.3888,
      "step": 3540
    },
    {
      "epoch": 10.66,
      "grad_norm": 3.919921398162842,
      "learning_rate": 8.933933933933935e-06,
      "loss": 1.4129,
      "step": 3550
    },
    {
      "epoch": 10.69,
      "grad_norm": 3.411519765853882,
      "learning_rate": 8.93093093093093e-06,
      "loss": 1.3775,
      "step": 3560
    },
    {
      "epoch": 10.72,
      "grad_norm": 4.0573506355285645,
      "learning_rate": 8.927927927927928e-06,
      "loss": 1.3593,
      "step": 3570
    },
    {
      "epoch": 10.75,
      "grad_norm": 5.557109832763672,
      "learning_rate": 8.924924924924925e-06,
      "loss": 1.3374,
      "step": 3580
    },
    {
      "epoch": 10.78,
      "grad_norm": 4.917846202850342,
      "learning_rate": 8.921921921921923e-06,
      "loss": 1.3134,
      "step": 3590
    },
    {
      "epoch": 10.81,
      "grad_norm": 3.7236461639404297,
      "learning_rate": 8.91891891891892e-06,
      "loss": 1.3706,
      "step": 3600
    },
    {
      "epoch": 10.84,
      "grad_norm": 4.710603713989258,
      "learning_rate": 8.915915915915918e-06,
      "loss": 1.3007,
      "step": 3610
    },
    {
      "epoch": 10.87,
      "grad_norm": 3.778068780899048,
      "learning_rate": 8.912912912912914e-06,
      "loss": 1.3659,
      "step": 3620
    },
    {
      "epoch": 10.9,
      "grad_norm": 4.073859214782715,
      "learning_rate": 8.909909909909911e-06,
      "loss": 1.3516,
      "step": 3630
    },
    {
      "epoch": 10.93,
      "grad_norm": 3.431516408920288,
      "learning_rate": 8.906906906906909e-06,
      "loss": 1.3217,
      "step": 3640
    },
    {
      "epoch": 10.96,
      "grad_norm": 4.807550430297852,
      "learning_rate": 8.903903903903904e-06,
      "loss": 1.3234,
      "step": 3650
    },
    {
      "epoch": 10.99,
      "grad_norm": 2.7062273025512695,
      "learning_rate": 8.900900900900902e-06,
      "loss": 1.3424,
      "step": 3660
    },
    {
      "epoch": 11.0,
      "eval_accuracy": 0.8923,
      "eval_loss": 1.0521777868270874,
      "eval_runtime": 26.5374,
      "eval_samples_per_second": 376.827,
      "eval_steps_per_second": 1.507,
      "step": 3663
    },
    {
      "epoch": 11.02,
      "grad_norm": 3.720083236694336,
      "learning_rate": 8.8978978978979e-06,
      "loss": 1.4965,
      "step": 3670
    },
    {
      "epoch": 11.05,
      "grad_norm": 9.219380378723145,
      "learning_rate": 8.894894894894895e-06,
      "loss": 1.3053,
      "step": 3680
    },
    {
      "epoch": 11.08,
      "grad_norm": 4.394057273864746,
      "learning_rate": 8.891891891891893e-06,
      "loss": 1.3092,
      "step": 3690
    },
    {
      "epoch": 11.11,
      "grad_norm": 4.5011887550354,
      "learning_rate": 8.888888888888888e-06,
      "loss": 1.3167,
      "step": 3700
    },
    {
      "epoch": 11.14,
      "grad_norm": 3.506056070327759,
      "learning_rate": 8.885885885885886e-06,
      "loss": 1.285,
      "step": 3710
    },
    {
      "epoch": 11.17,
      "grad_norm": 2.9524905681610107,
      "learning_rate": 8.882882882882883e-06,
      "loss": 1.2807,
      "step": 3720
    },
    {
      "epoch": 11.2,
      "grad_norm": 4.951297760009766,
      "learning_rate": 8.87987987987988e-06,
      "loss": 1.3685,
      "step": 3730
    },
    {
      "epoch": 11.23,
      "grad_norm": 4.545038223266602,
      "learning_rate": 8.876876876876878e-06,
      "loss": 1.2753,
      "step": 3740
    },
    {
      "epoch": 11.26,
      "grad_norm": 3.9689831733703613,
      "learning_rate": 8.873873873873876e-06,
      "loss": 1.2925,
      "step": 3750
    },
    {
      "epoch": 11.29,
      "grad_norm": 3.231689929962158,
      "learning_rate": 8.870870870870871e-06,
      "loss": 1.3154,
      "step": 3760
    },
    {
      "epoch": 11.32,
      "grad_norm": 4.227672100067139,
      "learning_rate": 8.867867867867869e-06,
      "loss": 1.3276,
      "step": 3770
    },
    {
      "epoch": 11.35,
      "grad_norm": 3.2093513011932373,
      "learning_rate": 8.864864864864866e-06,
      "loss": 1.3571,
      "step": 3780
    },
    {
      "epoch": 11.38,
      "grad_norm": 3.391765832901001,
      "learning_rate": 8.861861861861862e-06,
      "loss": 1.3214,
      "step": 3790
    },
    {
      "epoch": 11.41,
      "grad_norm": 2.6600348949432373,
      "learning_rate": 8.85885885885886e-06,
      "loss": 1.3242,
      "step": 3800
    },
    {
      "epoch": 11.44,
      "grad_norm": 3.3445725440979004,
      "learning_rate": 8.855855855855855e-06,
      "loss": 1.2893,
      "step": 3810
    },
    {
      "epoch": 11.47,
      "grad_norm": 5.254781246185303,
      "learning_rate": 8.852852852852853e-06,
      "loss": 1.2612,
      "step": 3820
    },
    {
      "epoch": 11.5,
      "grad_norm": 3.754554271697998,
      "learning_rate": 8.84984984984985e-06,
      "loss": 1.2711,
      "step": 3830
    },
    {
      "epoch": 11.53,
      "grad_norm": 4.283574104309082,
      "learning_rate": 8.846846846846848e-06,
      "loss": 1.2976,
      "step": 3840
    },
    {
      "epoch": 11.56,
      "grad_norm": 5.312755107879639,
      "learning_rate": 8.843843843843844e-06,
      "loss": 1.2624,
      "step": 3850
    },
    {
      "epoch": 11.59,
      "grad_norm": 3.4620511531829834,
      "learning_rate": 8.840840840840841e-06,
      "loss": 1.3101,
      "step": 3860
    },
    {
      "epoch": 11.62,
      "grad_norm": 3.3053994178771973,
      "learning_rate": 8.837837837837839e-06,
      "loss": 1.292,
      "step": 3870
    },
    {
      "epoch": 11.65,
      "grad_norm": 3.8711774349212646,
      "learning_rate": 8.834834834834836e-06,
      "loss": 1.2923,
      "step": 3880
    },
    {
      "epoch": 11.68,
      "grad_norm": 3.261052131652832,
      "learning_rate": 8.831831831831834e-06,
      "loss": 1.2687,
      "step": 3890
    },
    {
      "epoch": 11.71,
      "grad_norm": 5.33121919631958,
      "learning_rate": 8.82882882882883e-06,
      "loss": 1.3165,
      "step": 3900
    },
    {
      "epoch": 11.74,
      "grad_norm": 4.399143218994141,
      "learning_rate": 8.825825825825827e-06,
      "loss": 1.2678,
      "step": 3910
    },
    {
      "epoch": 11.77,
      "grad_norm": 4.487939357757568,
      "learning_rate": 8.822822822822824e-06,
      "loss": 1.227,
      "step": 3920
    },
    {
      "epoch": 11.8,
      "grad_norm": 4.024313449859619,
      "learning_rate": 8.81981981981982e-06,
      "loss": 1.2681,
      "step": 3930
    },
    {
      "epoch": 11.83,
      "grad_norm": 4.159395694732666,
      "learning_rate": 8.816816816816818e-06,
      "loss": 1.2048,
      "step": 3940
    },
    {
      "epoch": 11.86,
      "grad_norm": 3.9923174381256104,
      "learning_rate": 8.813813813813813e-06,
      "loss": 1.2211,
      "step": 3950
    },
    {
      "epoch": 11.89,
      "grad_norm": 4.025728225708008,
      "learning_rate": 8.810810810810811e-06,
      "loss": 1.2415,
      "step": 3960
    },
    {
      "epoch": 11.92,
      "grad_norm": 3.927666664123535,
      "learning_rate": 8.807807807807808e-06,
      "loss": 1.2254,
      "step": 3970
    },
    {
      "epoch": 11.95,
      "grad_norm": 5.660053253173828,
      "learning_rate": 8.804804804804806e-06,
      "loss": 1.2388,
      "step": 3980
    },
    {
      "epoch": 11.98,
      "grad_norm": 3.2951526641845703,
      "learning_rate": 8.801801801801803e-06,
      "loss": 1.1924,
      "step": 3990
    },
    {
      "epoch": 12.0,
      "eval_accuracy": 0.8983,
      "eval_loss": 0.9380095601081848,
      "eval_runtime": 26.1486,
      "eval_samples_per_second": 382.429,
      "eval_steps_per_second": 1.53,
      "step": 3996
    },
    {
      "epoch": 12.01,
      "grad_norm": 4.126723289489746,
      "learning_rate": 8.798798798798799e-06,
      "loss": 1.3484,
      "step": 4000
    },
    {
      "epoch": 12.04,
      "grad_norm": 4.03225564956665,
      "learning_rate": 8.795795795795797e-06,
      "loss": 1.2141,
      "step": 4010
    },
    {
      "epoch": 12.07,
      "grad_norm": 4.17138671875,
      "learning_rate": 8.792792792792794e-06,
      "loss": 1.1898,
      "step": 4020
    },
    {
      "epoch": 12.1,
      "grad_norm": 2.8586363792419434,
      "learning_rate": 8.789789789789792e-06,
      "loss": 1.1838,
      "step": 4030
    },
    {
      "epoch": 12.13,
      "grad_norm": 4.225535869598389,
      "learning_rate": 8.786786786786787e-06,
      "loss": 1.1863,
      "step": 4040
    },
    {
      "epoch": 12.16,
      "grad_norm": 4.490497589111328,
      "learning_rate": 8.783783783783785e-06,
      "loss": 1.218,
      "step": 4050
    },
    {
      "epoch": 12.19,
      "grad_norm": 5.3085408210754395,
      "learning_rate": 8.78078078078078e-06,
      "loss": 1.189,
      "step": 4060
    },
    {
      "epoch": 12.22,
      "grad_norm": 4.248027801513672,
      "learning_rate": 8.777777777777778e-06,
      "loss": 1.1828,
      "step": 4070
    },
    {
      "epoch": 12.25,
      "grad_norm": 4.555266380310059,
      "learning_rate": 8.774774774774776e-06,
      "loss": 1.2203,
      "step": 4080
    },
    {
      "epoch": 12.28,
      "grad_norm": 4.271499156951904,
      "learning_rate": 8.771771771771771e-06,
      "loss": 1.2298,
      "step": 4090
    },
    {
      "epoch": 12.31,
      "grad_norm": 4.123808860778809,
      "learning_rate": 8.768768768768769e-06,
      "loss": 1.1945,
      "step": 4100
    },
    {
      "epoch": 12.34,
      "grad_norm": 4.744080066680908,
      "learning_rate": 8.765765765765766e-06,
      "loss": 1.1765,
      "step": 4110
    },
    {
      "epoch": 12.37,
      "grad_norm": 5.163188934326172,
      "learning_rate": 8.762762762762764e-06,
      "loss": 1.2019,
      "step": 4120
    },
    {
      "epoch": 12.4,
      "grad_norm": 4.672330379486084,
      "learning_rate": 8.759759759759761e-06,
      "loss": 1.2162,
      "step": 4130
    },
    {
      "epoch": 12.43,
      "grad_norm": 3.3714725971221924,
      "learning_rate": 8.756756756756759e-06,
      "loss": 1.2412,
      "step": 4140
    },
    {
      "epoch": 12.46,
      "grad_norm": 4.335172653198242,
      "learning_rate": 8.753753753753755e-06,
      "loss": 1.1686,
      "step": 4150
    },
    {
      "epoch": 12.49,
      "grad_norm": 3.776484251022339,
      "learning_rate": 8.750750750750752e-06,
      "loss": 1.1799,
      "step": 4160
    },
    {
      "epoch": 12.52,
      "grad_norm": 4.604945659637451,
      "learning_rate": 8.747747747747748e-06,
      "loss": 1.1592,
      "step": 4170
    },
    {
      "epoch": 12.55,
      "grad_norm": 3.4792492389678955,
      "learning_rate": 8.744744744744745e-06,
      "loss": 1.1912,
      "step": 4180
    },
    {
      "epoch": 12.58,
      "grad_norm": 5.771131992340088,
      "learning_rate": 8.741741741741743e-06,
      "loss": 1.1675,
      "step": 4190
    },
    {
      "epoch": 12.61,
      "grad_norm": 3.7320287227630615,
      "learning_rate": 8.738738738738739e-06,
      "loss": 1.1597,
      "step": 4200
    },
    {
      "epoch": 12.64,
      "grad_norm": 5.218122959136963,
      "learning_rate": 8.735735735735736e-06,
      "loss": 1.1764,
      "step": 4210
    },
    {
      "epoch": 12.67,
      "grad_norm": 3.9445109367370605,
      "learning_rate": 8.732732732732733e-06,
      "loss": 1.1086,
      "step": 4220
    },
    {
      "epoch": 12.7,
      "grad_norm": 4.147899150848389,
      "learning_rate": 8.72972972972973e-06,
      "loss": 1.1232,
      "step": 4230
    },
    {
      "epoch": 12.73,
      "grad_norm": 5.463786602020264,
      "learning_rate": 8.726726726726727e-06,
      "loss": 1.1803,
      "step": 4240
    },
    {
      "epoch": 12.76,
      "grad_norm": 3.1907215118408203,
      "learning_rate": 8.723723723723724e-06,
      "loss": 1.0963,
      "step": 4250
    },
    {
      "epoch": 12.79,
      "grad_norm": 4.256779193878174,
      "learning_rate": 8.720720720720722e-06,
      "loss": 1.1338,
      "step": 4260
    },
    {
      "epoch": 12.82,
      "grad_norm": 3.1030843257904053,
      "learning_rate": 8.71771771771772e-06,
      "loss": 1.1643,
      "step": 4270
    },
    {
      "epoch": 12.85,
      "grad_norm": 3.6219348907470703,
      "learning_rate": 8.714714714714717e-06,
      "loss": 1.1564,
      "step": 4280
    },
    {
      "epoch": 12.88,
      "grad_norm": 4.349027633666992,
      "learning_rate": 8.711711711711712e-06,
      "loss": 1.0917,
      "step": 4290
    },
    {
      "epoch": 12.91,
      "grad_norm": 3.0795490741729736,
      "learning_rate": 8.70870870870871e-06,
      "loss": 1.1245,
      "step": 4300
    },
    {
      "epoch": 12.94,
      "grad_norm": 3.5159428119659424,
      "learning_rate": 8.705705705705706e-06,
      "loss": 1.1587,
      "step": 4310
    },
    {
      "epoch": 12.97,
      "grad_norm": 3.8004889488220215,
      "learning_rate": 8.702702702702703e-06,
      "loss": 1.1764,
      "step": 4320
    },
    {
      "epoch": 13.0,
      "eval_accuracy": 0.8999,
      "eval_loss": 0.8404828310012817,
      "eval_runtime": 26.4667,
      "eval_samples_per_second": 377.833,
      "eval_steps_per_second": 1.511,
      "step": 4329
    },
    {
      "epoch": 13.0,
      "grad_norm": 5.436849594116211,
      "learning_rate": 8.6996996996997e-06,
      "loss": 1.1035,
      "step": 4330
    },
    {
      "epoch": 13.03,
      "grad_norm": 4.124845504760742,
      "learning_rate": 8.696696696696696e-06,
      "loss": 1.0423,
      "step": 4340
    },
    {
      "epoch": 13.06,
      "grad_norm": 7.9258036613464355,
      "learning_rate": 8.693693693693694e-06,
      "loss": 1.096,
      "step": 4350
    },
    {
      "epoch": 13.09,
      "grad_norm": 4.796426296234131,
      "learning_rate": 8.690690690690691e-06,
      "loss": 1.1952,
      "step": 4360
    },
    {
      "epoch": 13.12,
      "grad_norm": 4.917626857757568,
      "learning_rate": 8.687687687687689e-06,
      "loss": 1.1613,
      "step": 4370
    },
    {
      "epoch": 13.15,
      "grad_norm": 3.8574671745300293,
      "learning_rate": 8.684684684684686e-06,
      "loss": 1.1469,
      "step": 4380
    },
    {
      "epoch": 13.18,
      "grad_norm": 3.908940315246582,
      "learning_rate": 8.681681681681682e-06,
      "loss": 1.1275,
      "step": 4390
    },
    {
      "epoch": 13.21,
      "grad_norm": 3.9402410984039307,
      "learning_rate": 8.67867867867868e-06,
      "loss": 1.141,
      "step": 4400
    },
    {
      "epoch": 13.24,
      "grad_norm": 2.884037971496582,
      "learning_rate": 8.675675675675677e-06,
      "loss": 1.0775,
      "step": 4410
    },
    {
      "epoch": 13.27,
      "grad_norm": 5.2757439613342285,
      "learning_rate": 8.672672672672673e-06,
      "loss": 1.1441,
      "step": 4420
    },
    {
      "epoch": 13.3,
      "grad_norm": 5.043304920196533,
      "learning_rate": 8.66966966966967e-06,
      "loss": 1.1053,
      "step": 4430
    },
    {
      "epoch": 13.33,
      "grad_norm": 3.6216320991516113,
      "learning_rate": 8.666666666666668e-06,
      "loss": 1.0889,
      "step": 4440
    },
    {
      "epoch": 13.36,
      "grad_norm": 3.5936968326568604,
      "learning_rate": 8.663663663663664e-06,
      "loss": 1.0993,
      "step": 4450
    },
    {
      "epoch": 13.39,
      "grad_norm": 4.1380743980407715,
      "learning_rate": 8.660660660660661e-06,
      "loss": 1.1261,
      "step": 4460
    },
    {
      "epoch": 13.42,
      "grad_norm": 3.888875722885132,
      "learning_rate": 8.657657657657659e-06,
      "loss": 1.1264,
      "step": 4470
    },
    {
      "epoch": 13.45,
      "grad_norm": 4.108713150024414,
      "learning_rate": 8.654654654654654e-06,
      "loss": 1.1277,
      "step": 4480
    },
    {
      "epoch": 13.48,
      "grad_norm": 4.775360107421875,
      "learning_rate": 8.651651651651652e-06,
      "loss": 1.0757,
      "step": 4490
    },
    {
      "epoch": 13.51,
      "grad_norm": 4.780664920806885,
      "learning_rate": 8.64864864864865e-06,
      "loss": 1.0774,
      "step": 4500
    },
    {
      "epoch": 13.54,
      "grad_norm": 4.424095630645752,
      "learning_rate": 8.645645645645647e-06,
      "loss": 1.1321,
      "step": 4510
    },
    {
      "epoch": 13.57,
      "grad_norm": 5.375193119049072,
      "learning_rate": 8.642642642642644e-06,
      "loss": 1.0712,
      "step": 4520
    },
    {
      "epoch": 13.6,
      "grad_norm": 4.035747051239014,
      "learning_rate": 8.63963963963964e-06,
      "loss": 1.1009,
      "step": 4530
    },
    {
      "epoch": 13.63,
      "grad_norm": 5.080639362335205,
      "learning_rate": 8.636636636636638e-06,
      "loss": 1.1534,
      "step": 4540
    },
    {
      "epoch": 13.66,
      "grad_norm": 3.3346140384674072,
      "learning_rate": 8.633633633633635e-06,
      "loss": 1.0454,
      "step": 4550
    },
    {
      "epoch": 13.69,
      "grad_norm": 3.6367032527923584,
      "learning_rate": 8.63063063063063e-06,
      "loss": 1.1244,
      "step": 4560
    },
    {
      "epoch": 13.72,
      "grad_norm": 6.010049819946289,
      "learning_rate": 8.627627627627628e-06,
      "loss": 1.0784,
      "step": 4570
    },
    {
      "epoch": 13.75,
      "grad_norm": 3.2728335857391357,
      "learning_rate": 8.624624624624626e-06,
      "loss": 1.0053,
      "step": 4580
    },
    {
      "epoch": 13.78,
      "grad_norm": 4.756971836090088,
      "learning_rate": 8.621621621621622e-06,
      "loss": 1.0427,
      "step": 4590
    },
    {
      "epoch": 13.81,
      "grad_norm": 3.9494152069091797,
      "learning_rate": 8.618618618618619e-06,
      "loss": 1.0729,
      "step": 4600
    },
    {
      "epoch": 13.84,
      "grad_norm": 4.257194995880127,
      "learning_rate": 8.615615615615616e-06,
      "loss": 1.0893,
      "step": 4610
    },
    {
      "epoch": 13.87,
      "grad_norm": 4.306645393371582,
      "learning_rate": 8.612612612612612e-06,
      "loss": 1.0025,
      "step": 4620
    },
    {
      "epoch": 13.9,
      "grad_norm": 4.929202556610107,
      "learning_rate": 8.60960960960961e-06,
      "loss": 1.1036,
      "step": 4630
    },
    {
      "epoch": 13.93,
      "grad_norm": 4.744569301605225,
      "learning_rate": 8.606606606606607e-06,
      "loss": 1.0783,
      "step": 4640
    },
    {
      "epoch": 13.96,
      "grad_norm": 5.201243877410889,
      "learning_rate": 8.603603603603605e-06,
      "loss": 1.0203,
      "step": 4650
    },
    {
      "epoch": 13.99,
      "grad_norm": 2.8255743980407715,
      "learning_rate": 8.600600600600602e-06,
      "loss": 1.0548,
      "step": 4660
    },
    {
      "epoch": 14.0,
      "eval_accuracy": 0.9024,
      "eval_loss": 0.7641175389289856,
      "eval_runtime": 26.4626,
      "eval_samples_per_second": 377.892,
      "eval_steps_per_second": 1.512,
      "step": 4662
    },
    {
      "epoch": 14.02,
      "grad_norm": 4.445658206939697,
      "learning_rate": 8.597597597597598e-06,
      "loss": 1.0731,
      "step": 4670
    },
    {
      "epoch": 14.05,
      "grad_norm": 4.949143409729004,
      "learning_rate": 8.594594594594595e-06,
      "loss": 1.0116,
      "step": 4680
    },
    {
      "epoch": 14.08,
      "grad_norm": 3.9592392444610596,
      "learning_rate": 8.591591591591593e-06,
      "loss": 1.0098,
      "step": 4690
    },
    {
      "epoch": 14.11,
      "grad_norm": 4.729857444763184,
      "learning_rate": 8.588588588588589e-06,
      "loss": 1.0176,
      "step": 4700
    },
    {
      "epoch": 14.14,
      "grad_norm": 4.354032039642334,
      "learning_rate": 8.585585585585586e-06,
      "loss": 1.0034,
      "step": 4710
    },
    {
      "epoch": 14.17,
      "grad_norm": 5.001422882080078,
      "learning_rate": 8.582582582582584e-06,
      "loss": 1.0534,
      "step": 4720
    },
    {
      "epoch": 14.2,
      "grad_norm": 4.593340873718262,
      "learning_rate": 8.57957957957958e-06,
      "loss": 1.1093,
      "step": 4730
    },
    {
      "epoch": 14.23,
      "grad_norm": 4.776356220245361,
      "learning_rate": 8.576576576576577e-06,
      "loss": 1.1008,
      "step": 4740
    },
    {
      "epoch": 14.26,
      "grad_norm": 4.727325439453125,
      "learning_rate": 8.573573573573574e-06,
      "loss": 1.0395,
      "step": 4750
    },
    {
      "epoch": 14.29,
      "grad_norm": 5.222812652587891,
      "learning_rate": 8.570570570570572e-06,
      "loss": 1.0424,
      "step": 4760
    },
    {
      "epoch": 14.32,
      "grad_norm": 4.294810771942139,
      "learning_rate": 8.567567567567568e-06,
      "loss": 0.9906,
      "step": 4770
    },
    {
      "epoch": 14.35,
      "grad_norm": 3.130147933959961,
      "learning_rate": 8.564564564564565e-06,
      "loss": 1.0348,
      "step": 4780
    },
    {
      "epoch": 14.38,
      "grad_norm": 4.623157501220703,
      "learning_rate": 8.561561561561563e-06,
      "loss": 1.0151,
      "step": 4790
    },
    {
      "epoch": 14.41,
      "grad_norm": 7.708536148071289,
      "learning_rate": 8.55855855855856e-06,
      "loss": 1.0329,
      "step": 4800
    },
    {
      "epoch": 14.44,
      "grad_norm": 2.8359479904174805,
      "learning_rate": 8.555555555555556e-06,
      "loss": 1.0228,
      "step": 4810
    },
    {
      "epoch": 14.47,
      "grad_norm": 4.290976524353027,
      "learning_rate": 8.552552552552553e-06,
      "loss": 1.0335,
      "step": 4820
    },
    {
      "epoch": 14.5,
      "grad_norm": 4.422244548797607,
      "learning_rate": 8.549549549549551e-06,
      "loss": 1.0498,
      "step": 4830
    },
    {
      "epoch": 14.53,
      "grad_norm": 4.271435260772705,
      "learning_rate": 8.546546546546547e-06,
      "loss": 1.0276,
      "step": 4840
    },
    {
      "epoch": 14.56,
      "grad_norm": 3.2531402111053467,
      "learning_rate": 8.543543543543544e-06,
      "loss": 1.0197,
      "step": 4850
    },
    {
      "epoch": 14.59,
      "grad_norm": 5.6989946365356445,
      "learning_rate": 8.540540540540542e-06,
      "loss": 0.9598,
      "step": 4860
    },
    {
      "epoch": 14.62,
      "grad_norm": 5.341368198394775,
      "learning_rate": 8.537537537537537e-06,
      "loss": 1.021,
      "step": 4870
    },
    {
      "epoch": 14.65,
      "grad_norm": 2.799990653991699,
      "learning_rate": 8.534534534534535e-06,
      "loss": 0.9969,
      "step": 4880
    },
    {
      "epoch": 14.68,
      "grad_norm": 5.680851936340332,
      "learning_rate": 8.531531531531532e-06,
      "loss": 1.0417,
      "step": 4890
    },
    {
      "epoch": 14.71,
      "grad_norm": 4.588523864746094,
      "learning_rate": 8.52852852852853e-06,
      "loss": 0.9967,
      "step": 4900
    },
    {
      "epoch": 14.74,
      "grad_norm": 3.542405843734741,
      "learning_rate": 8.525525525525527e-06,
      "loss": 1.006,
      "step": 4910
    },
    {
      "epoch": 14.77,
      "grad_norm": 5.6309123039245605,
      "learning_rate": 8.522522522522523e-06,
      "loss": 0.981,
      "step": 4920
    },
    {
      "epoch": 14.8,
      "grad_norm": 4.7694196701049805,
      "learning_rate": 8.51951951951952e-06,
      "loss": 1.0281,
      "step": 4930
    },
    {
      "epoch": 14.83,
      "grad_norm": 4.105522632598877,
      "learning_rate": 8.516516516516518e-06,
      "loss": 1.0287,
      "step": 4940
    },
    {
      "epoch": 14.86,
      "grad_norm": 6.505147933959961,
      "learning_rate": 8.513513513513514e-06,
      "loss": 1.0281,
      "step": 4950
    },
    {
      "epoch": 14.89,
      "grad_norm": 4.380159854888916,
      "learning_rate": 8.510510510510511e-06,
      "loss": 0.9798,
      "step": 4960
    },
    {
      "epoch": 14.92,
      "grad_norm": 4.89952278137207,
      "learning_rate": 8.507507507507509e-06,
      "loss": 1.0214,
      "step": 4970
    },
    {
      "epoch": 14.95,
      "grad_norm": 4.74617862701416,
      "learning_rate": 8.504504504504505e-06,
      "loss": 1.047,
      "step": 4980
    },
    {
      "epoch": 14.98,
      "grad_norm": 3.973592519760132,
      "learning_rate": 8.501501501501502e-06,
      "loss": 0.9714,
      "step": 4990
    },
    {
      "epoch": 15.0,
      "eval_accuracy": 0.9069,
      "eval_loss": 0.6897292137145996,
      "eval_runtime": 26.404,
      "eval_samples_per_second": 378.73,
      "eval_steps_per_second": 1.515,
      "step": 4995
    },
    {
      "epoch": 15.02,
      "grad_norm": 3.2270395755767822,
      "learning_rate": 8.4984984984985e-06,
      "loss": 0.8831,
      "step": 5000
    },
    {
      "epoch": 15.05,
      "grad_norm": 3.8310062885284424,
      "learning_rate": 8.495495495495495e-06,
      "loss": 0.944,
      "step": 5010
    },
    {
      "epoch": 15.08,
      "grad_norm": 4.674674987792969,
      "learning_rate": 8.492492492492493e-06,
      "loss": 0.9558,
      "step": 5020
    },
    {
      "epoch": 15.11,
      "grad_norm": 5.555057048797607,
      "learning_rate": 8.48948948948949e-06,
      "loss": 1.0067,
      "step": 5030
    },
    {
      "epoch": 15.14,
      "grad_norm": 4.9669389724731445,
      "learning_rate": 8.486486486486488e-06,
      "loss": 0.9521,
      "step": 5040
    },
    {
      "epoch": 15.17,
      "grad_norm": 5.623544692993164,
      "learning_rate": 8.483483483483485e-06,
      "loss": 0.967,
      "step": 5050
    },
    {
      "epoch": 15.2,
      "grad_norm": 3.8444600105285645,
      "learning_rate": 8.480480480480481e-06,
      "loss": 0.9233,
      "step": 5060
    },
    {
      "epoch": 15.23,
      "grad_norm": 2.460385322570801,
      "learning_rate": 8.477477477477478e-06,
      "loss": 1.038,
      "step": 5070
    },
    {
      "epoch": 15.26,
      "grad_norm": 3.19730806350708,
      "learning_rate": 8.474474474474476e-06,
      "loss": 0.982,
      "step": 5080
    },
    {
      "epoch": 15.29,
      "grad_norm": 4.365762233734131,
      "learning_rate": 8.471471471471472e-06,
      "loss": 0.9707,
      "step": 5090
    },
    {
      "epoch": 15.32,
      "grad_norm": 3.1159462928771973,
      "learning_rate": 8.46846846846847e-06,
      "loss": 0.9719,
      "step": 5100
    },
    {
      "epoch": 15.35,
      "grad_norm": 4.497450828552246,
      "learning_rate": 8.465465465465467e-06,
      "loss": 0.905,
      "step": 5110
    },
    {
      "epoch": 15.38,
      "grad_norm": 4.356529712677002,
      "learning_rate": 8.462462462462462e-06,
      "loss": 0.9398,
      "step": 5120
    },
    {
      "epoch": 15.41,
      "grad_norm": 5.023562431335449,
      "learning_rate": 8.45945945945946e-06,
      "loss": 0.96,
      "step": 5130
    },
    {
      "epoch": 15.44,
      "grad_norm": 4.207388877868652,
      "learning_rate": 8.456456456456457e-06,
      "loss": 0.9279,
      "step": 5140
    },
    {
      "epoch": 15.47,
      "grad_norm": 5.606916904449463,
      "learning_rate": 8.453453453453453e-06,
      "loss": 0.9059,
      "step": 5150
    },
    {
      "epoch": 15.5,
      "grad_norm": 3.986994743347168,
      "learning_rate": 8.45045045045045e-06,
      "loss": 0.9314,
      "step": 5160
    },
    {
      "epoch": 15.53,
      "grad_norm": 3.543445587158203,
      "learning_rate": 8.447447447447448e-06,
      "loss": 0.9505,
      "step": 5170
    },
    {
      "epoch": 15.56,
      "grad_norm": 4.311727046966553,
      "learning_rate": 8.444444444444446e-06,
      "loss": 0.9104,
      "step": 5180
    },
    {
      "epoch": 15.59,
      "grad_norm": 4.540422439575195,
      "learning_rate": 8.441441441441443e-06,
      "loss": 0.9093,
      "step": 5190
    },
    {
      "epoch": 15.62,
      "grad_norm": 4.569993495941162,
      "learning_rate": 8.438438438438439e-06,
      "loss": 0.919,
      "step": 5200
    },
    {
      "epoch": 15.65,
      "grad_norm": 4.752470016479492,
      "learning_rate": 8.435435435435436e-06,
      "loss": 0.8959,
      "step": 5210
    },
    {
      "epoch": 15.68,
      "grad_norm": 3.8412997722625732,
      "learning_rate": 8.432432432432434e-06,
      "loss": 0.91,
      "step": 5220
    },
    {
      "epoch": 15.71,
      "grad_norm": 4.719942092895508,
      "learning_rate": 8.42942942942943e-06,
      "loss": 0.9158,
      "step": 5230
    },
    {
      "epoch": 15.74,
      "grad_norm": 4.649778366088867,
      "learning_rate": 8.426426426426427e-06,
      "loss": 0.8962,
      "step": 5240
    },
    {
      "epoch": 15.77,
      "grad_norm": 6.160231113433838,
      "learning_rate": 8.423423423423423e-06,
      "loss": 0.9212,
      "step": 5250
    },
    {
      "epoch": 15.8,
      "grad_norm": 4.66862678527832,
      "learning_rate": 8.42042042042042e-06,
      "loss": 0.9635,
      "step": 5260
    },
    {
      "epoch": 15.83,
      "grad_norm": 8.437322616577148,
      "learning_rate": 8.417417417417418e-06,
      "loss": 0.8503,
      "step": 5270
    },
    {
      "epoch": 15.86,
      "grad_norm": 6.350149631500244,
      "learning_rate": 8.414414414414415e-06,
      "loss": 0.8894,
      "step": 5280
    },
    {
      "epoch": 15.89,
      "grad_norm": 4.726019382476807,
      "learning_rate": 8.411411411411413e-06,
      "loss": 0.9115,
      "step": 5290
    },
    {
      "epoch": 15.92,
      "grad_norm": 3.8601720333099365,
      "learning_rate": 8.408408408408409e-06,
      "loss": 0.9795,
      "step": 5300
    },
    {
      "epoch": 15.95,
      "grad_norm": 3.991065502166748,
      "learning_rate": 8.405405405405406e-06,
      "loss": 0.9379,
      "step": 5310
    },
    {
      "epoch": 15.98,
      "grad_norm": 4.6855316162109375,
      "learning_rate": 8.402402402402404e-06,
      "loss": 0.9141,
      "step": 5320
    },
    {
      "epoch": 16.0,
      "eval_accuracy": 0.9047,
      "eval_loss": 0.6327019333839417,
      "eval_runtime": 26.6918,
      "eval_samples_per_second": 374.647,
      "eval_steps_per_second": 1.499,
      "step": 5328
    },
    {
      "epoch": 16.01,
      "grad_norm": 6.888444423675537,
      "learning_rate": 8.399399399399401e-06,
      "loss": 0.9849,
      "step": 5330
    },
    {
      "epoch": 16.04,
      "grad_norm": 4.066012382507324,
      "learning_rate": 8.396396396396397e-06,
      "loss": 0.9008,
      "step": 5340
    },
    {
      "epoch": 16.07,
      "grad_norm": 3.7079455852508545,
      "learning_rate": 8.393393393393394e-06,
      "loss": 0.9026,
      "step": 5350
    },
    {
      "epoch": 16.1,
      "grad_norm": 4.255161285400391,
      "learning_rate": 8.390390390390392e-06,
      "loss": 0.9208,
      "step": 5360
    },
    {
      "epoch": 16.13,
      "grad_norm": 4.1778645515441895,
      "learning_rate": 8.387387387387388e-06,
      "loss": 0.9302,
      "step": 5370
    },
    {
      "epoch": 16.16,
      "grad_norm": 3.891451835632324,
      "learning_rate": 8.384384384384385e-06,
      "loss": 0.932,
      "step": 5380
    },
    {
      "epoch": 16.19,
      "grad_norm": 5.1825947761535645,
      "learning_rate": 8.381381381381381e-06,
      "loss": 0.8951,
      "step": 5390
    },
    {
      "epoch": 16.22,
      "grad_norm": 5.37294864654541,
      "learning_rate": 8.378378378378378e-06,
      "loss": 0.8875,
      "step": 5400
    },
    {
      "epoch": 16.25,
      "grad_norm": 6.806169509887695,
      "learning_rate": 8.375375375375376e-06,
      "loss": 0.9221,
      "step": 5410
    },
    {
      "epoch": 16.28,
      "grad_norm": 5.307746887207031,
      "learning_rate": 8.372372372372373e-06,
      "loss": 0.9109,
      "step": 5420
    },
    {
      "epoch": 16.31,
      "grad_norm": 5.282186985015869,
      "learning_rate": 8.36936936936937e-06,
      "loss": 0.8565,
      "step": 5430
    },
    {
      "epoch": 16.34,
      "grad_norm": 4.689208030700684,
      "learning_rate": 8.366366366366368e-06,
      "loss": 0.9608,
      "step": 5440
    },
    {
      "epoch": 16.37,
      "grad_norm": 6.57575798034668,
      "learning_rate": 8.363363363363364e-06,
      "loss": 0.8458,
      "step": 5450
    },
    {
      "epoch": 16.4,
      "grad_norm": 7.146213531494141,
      "learning_rate": 8.360360360360362e-06,
      "loss": 0.9741,
      "step": 5460
    },
    {
      "epoch": 16.43,
      "grad_norm": 5.553043842315674,
      "learning_rate": 8.357357357357359e-06,
      "loss": 0.9529,
      "step": 5470
    },
    {
      "epoch": 16.46,
      "grad_norm": 5.228372097015381,
      "learning_rate": 8.354354354354355e-06,
      "loss": 0.8963,
      "step": 5480
    },
    {
      "epoch": 16.49,
      "grad_norm": 3.6776034832000732,
      "learning_rate": 8.351351351351352e-06,
      "loss": 0.9088,
      "step": 5490
    },
    {
      "epoch": 16.52,
      "grad_norm": 5.969200134277344,
      "learning_rate": 8.348348348348348e-06,
      "loss": 0.8746,
      "step": 5500
    },
    {
      "epoch": 16.55,
      "grad_norm": 4.356595039367676,
      "learning_rate": 8.345345345345346e-06,
      "loss": 0.882,
      "step": 5510
    },
    {
      "epoch": 16.58,
      "grad_norm": 4.669258117675781,
      "learning_rate": 8.342342342342343e-06,
      "loss": 0.8578,
      "step": 5520
    },
    {
      "epoch": 16.61,
      "grad_norm": 5.297147750854492,
      "learning_rate": 8.339339339339339e-06,
      "loss": 0.8958,
      "step": 5530
    },
    {
      "epoch": 16.64,
      "grad_norm": 4.6578216552734375,
      "learning_rate": 8.336336336336336e-06,
      "loss": 0.8829,
      "step": 5540
    },
    {
      "epoch": 16.67,
      "grad_norm": 2.4359843730926514,
      "learning_rate": 8.333333333333334e-06,
      "loss": 0.8524,
      "step": 5550
    },
    {
      "epoch": 16.7,
      "grad_norm": 5.646344184875488,
      "learning_rate": 8.330330330330331e-06,
      "loss": 0.8514,
      "step": 5560
    },
    {
      "epoch": 16.73,
      "grad_norm": 3.243112564086914,
      "learning_rate": 8.327327327327329e-06,
      "loss": 0.8425,
      "step": 5570
    },
    {
      "epoch": 16.76,
      "grad_norm": 4.088650703430176,
      "learning_rate": 8.324324324324326e-06,
      "loss": 0.8675,
      "step": 5580
    },
    {
      "epoch": 16.79,
      "grad_norm": 4.98700475692749,
      "learning_rate": 8.321321321321322e-06,
      "loss": 0.8815,
      "step": 5590
    },
    {
      "epoch": 16.82,
      "grad_norm": 4.231012344360352,
      "learning_rate": 8.31831831831832e-06,
      "loss": 0.8928,
      "step": 5600
    },
    {
      "epoch": 16.85,
      "grad_norm": 5.899864196777344,
      "learning_rate": 8.315315315315317e-06,
      "loss": 0.8743,
      "step": 5610
    },
    {
      "epoch": 16.88,
      "grad_norm": 6.303285598754883,
      "learning_rate": 8.312312312312313e-06,
      "loss": 0.8809,
      "step": 5620
    },
    {
      "epoch": 16.91,
      "grad_norm": 6.3348164558410645,
      "learning_rate": 8.30930930930931e-06,
      "loss": 0.9054,
      "step": 5630
    },
    {
      "epoch": 16.94,
      "grad_norm": 5.820091724395752,
      "learning_rate": 8.306306306306306e-06,
      "loss": 0.8742,
      "step": 5640
    },
    {
      "epoch": 16.97,
      "grad_norm": 4.7649383544921875,
      "learning_rate": 8.303303303303303e-06,
      "loss": 0.9408,
      "step": 5650
    },
    {
      "epoch": 17.0,
      "grad_norm": 5.467824935913086,
      "learning_rate": 8.300300300300301e-06,
      "loss": 0.8937,
      "step": 5660
    },
    {
      "epoch": 17.0,
      "eval_accuracy": 0.9065,
      "eval_loss": 0.5862383842468262,
      "eval_runtime": 26.5146,
      "eval_samples_per_second": 377.151,
      "eval_steps_per_second": 1.509,
      "step": 5661
    },
    {
      "epoch": 17.03,
      "grad_norm": 34.90660858154297,
      "learning_rate": 8.297297297297298e-06,
      "loss": 0.9659,
      "step": 5670
    },
    {
      "epoch": 17.06,
      "grad_norm": 4.263108730316162,
      "learning_rate": 8.294294294294294e-06,
      "loss": 0.8664,
      "step": 5680
    },
    {
      "epoch": 17.09,
      "grad_norm": 5.2203803062438965,
      "learning_rate": 8.291291291291292e-06,
      "loss": 0.8954,
      "step": 5690
    },
    {
      "epoch": 17.12,
      "grad_norm": 5.210129261016846,
      "learning_rate": 8.288288288288289e-06,
      "loss": 0.843,
      "step": 5700
    },
    {
      "epoch": 17.15,
      "grad_norm": 5.5564446449279785,
      "learning_rate": 8.285285285285287e-06,
      "loss": 0.8385,
      "step": 5710
    },
    {
      "epoch": 17.18,
      "grad_norm": 4.366509437561035,
      "learning_rate": 8.282282282282284e-06,
      "loss": 0.8602,
      "step": 5720
    },
    {
      "epoch": 17.21,
      "grad_norm": 4.130090236663818,
      "learning_rate": 8.27927927927928e-06,
      "loss": 0.8346,
      "step": 5730
    },
    {
      "epoch": 17.24,
      "grad_norm": 5.142440319061279,
      "learning_rate": 8.276276276276277e-06,
      "loss": 0.8611,
      "step": 5740
    },
    {
      "epoch": 17.27,
      "grad_norm": 4.728271961212158,
      "learning_rate": 8.273273273273273e-06,
      "loss": 0.8361,
      "step": 5750
    },
    {
      "epoch": 17.3,
      "grad_norm": 5.093695640563965,
      "learning_rate": 8.27027027027027e-06,
      "loss": 0.8924,
      "step": 5760
    },
    {
      "epoch": 17.33,
      "grad_norm": 3.6658284664154053,
      "learning_rate": 8.267267267267268e-06,
      "loss": 0.9114,
      "step": 5770
    },
    {
      "epoch": 17.36,
      "grad_norm": 5.7932891845703125,
      "learning_rate": 8.264264264264264e-06,
      "loss": 0.8824,
      "step": 5780
    },
    {
      "epoch": 17.39,
      "grad_norm": 4.216789722442627,
      "learning_rate": 8.261261261261261e-06,
      "loss": 0.8199,
      "step": 5790
    },
    {
      "epoch": 17.42,
      "grad_norm": 6.029932022094727,
      "learning_rate": 8.258258258258259e-06,
      "loss": 0.9022,
      "step": 5800
    },
    {
      "epoch": 17.45,
      "grad_norm": 5.9107985496521,
      "learning_rate": 8.255255255255256e-06,
      "loss": 0.8622,
      "step": 5810
    },
    {
      "epoch": 17.48,
      "grad_norm": 3.0279664993286133,
      "learning_rate": 8.252252252252254e-06,
      "loss": 0.7735,
      "step": 5820
    },
    {
      "epoch": 17.51,
      "grad_norm": 4.147484302520752,
      "learning_rate": 8.24924924924925e-06,
      "loss": 0.8706,
      "step": 5830
    },
    {
      "epoch": 17.54,
      "grad_norm": 4.788569450378418,
      "learning_rate": 8.246246246246247e-06,
      "loss": 0.8346,
      "step": 5840
    },
    {
      "epoch": 17.57,
      "grad_norm": 5.4136738777160645,
      "learning_rate": 8.243243243243245e-06,
      "loss": 0.8235,
      "step": 5850
    },
    {
      "epoch": 17.6,
      "grad_norm": 5.190497875213623,
      "learning_rate": 8.24024024024024e-06,
      "loss": 0.8227,
      "step": 5860
    },
    {
      "epoch": 17.63,
      "grad_norm": 3.4695847034454346,
      "learning_rate": 8.237237237237238e-06,
      "loss": 0.8652,
      "step": 5870
    },
    {
      "epoch": 17.66,
      "grad_norm": 4.851005554199219,
      "learning_rate": 8.234234234234235e-06,
      "loss": 0.8731,
      "step": 5880
    },
    {
      "epoch": 17.69,
      "grad_norm": 4.962414264678955,
      "learning_rate": 8.231231231231231e-06,
      "loss": 0.842,
      "step": 5890
    },
    {
      "epoch": 17.72,
      "grad_norm": 5.408501148223877,
      "learning_rate": 8.228228228228229e-06,
      "loss": 0.8513,
      "step": 5900
    },
    {
      "epoch": 17.75,
      "grad_norm": 4.542473316192627,
      "learning_rate": 8.225225225225226e-06,
      "loss": 0.8533,
      "step": 5910
    },
    {
      "epoch": 17.78,
      "grad_norm": 4.716324806213379,
      "learning_rate": 8.222222222222222e-06,
      "loss": 0.832,
      "step": 5920
    },
    {
      "epoch": 17.81,
      "grad_norm": 2.916837692260742,
      "learning_rate": 8.21921921921922e-06,
      "loss": 0.8694,
      "step": 5930
    },
    {
      "epoch": 17.84,
      "grad_norm": 5.19083309173584,
      "learning_rate": 8.216216216216217e-06,
      "loss": 0.8215,
      "step": 5940
    },
    {
      "epoch": 17.87,
      "grad_norm": 4.947284698486328,
      "learning_rate": 8.213213213213214e-06,
      "loss": 0.7846,
      "step": 5950
    },
    {
      "epoch": 17.9,
      "grad_norm": 4.968457221984863,
      "learning_rate": 8.210210210210212e-06,
      "loss": 0.7975,
      "step": 5960
    },
    {
      "epoch": 17.93,
      "grad_norm": 3.813506841659546,
      "learning_rate": 8.20720720720721e-06,
      "loss": 0.8211,
      "step": 5970
    },
    {
      "epoch": 17.96,
      "grad_norm": 4.17324161529541,
      "learning_rate": 8.204204204204205e-06,
      "loss": 0.8244,
      "step": 5980
    },
    {
      "epoch": 17.99,
      "grad_norm": 7.182973861694336,
      "learning_rate": 8.201201201201202e-06,
      "loss": 0.79,
      "step": 5990
    },
    {
      "epoch": 18.0,
      "eval_accuracy": 0.9104,
      "eval_loss": 0.5389306545257568,
      "eval_runtime": 26.2611,
      "eval_samples_per_second": 380.792,
      "eval_steps_per_second": 1.523,
      "step": 5994
    },
    {
      "epoch": 18.02,
      "grad_norm": 4.469026565551758,
      "learning_rate": 8.198198198198198e-06,
      "loss": 0.7846,
      "step": 6000
    },
    {
      "epoch": 18.05,
      "grad_norm": 5.259685039520264,
      "learning_rate": 8.195195195195196e-06,
      "loss": 0.7855,
      "step": 6010
    },
    {
      "epoch": 18.08,
      "grad_norm": 4.313201904296875,
      "learning_rate": 8.192192192192193e-06,
      "loss": 0.829,
      "step": 6020
    },
    {
      "epoch": 18.11,
      "grad_norm": 4.317193508148193,
      "learning_rate": 8.189189189189189e-06,
      "loss": 0.7573,
      "step": 6030
    },
    {
      "epoch": 18.14,
      "grad_norm": 5.022390842437744,
      "learning_rate": 8.186186186186186e-06,
      "loss": 0.7977,
      "step": 6040
    },
    {
      "epoch": 18.17,
      "grad_norm": 4.285593032836914,
      "learning_rate": 8.183183183183184e-06,
      "loss": 0.8062,
      "step": 6050
    },
    {
      "epoch": 18.2,
      "grad_norm": 5.628978729248047,
      "learning_rate": 8.18018018018018e-06,
      "loss": 0.7889,
      "step": 6060
    },
    {
      "epoch": 18.23,
      "grad_norm": 3.628568172454834,
      "learning_rate": 8.177177177177177e-06,
      "loss": 0.827,
      "step": 6070
    },
    {
      "epoch": 18.26,
      "grad_norm": 5.467437744140625,
      "learning_rate": 8.174174174174175e-06,
      "loss": 0.8763,
      "step": 6080
    },
    {
      "epoch": 18.29,
      "grad_norm": 4.808385372161865,
      "learning_rate": 8.171171171171172e-06,
      "loss": 0.8299,
      "step": 6090
    },
    {
      "epoch": 18.32,
      "grad_norm": 7.098764896392822,
      "learning_rate": 8.16816816816817e-06,
      "loss": 0.8324,
      "step": 6100
    },
    {
      "epoch": 18.35,
      "grad_norm": 4.591148853302002,
      "learning_rate": 8.165165165165165e-06,
      "loss": 0.7832,
      "step": 6110
    },
    {
      "epoch": 18.38,
      "grad_norm": 4.7516913414001465,
      "learning_rate": 8.162162162162163e-06,
      "loss": 0.7704,
      "step": 6120
    },
    {
      "epoch": 18.41,
      "grad_norm": 4.104935169219971,
      "learning_rate": 8.15915915915916e-06,
      "loss": 0.8418,
      "step": 6130
    },
    {
      "epoch": 18.44,
      "grad_norm": 4.477002143859863,
      "learning_rate": 8.156156156156156e-06,
      "loss": 0.7522,
      "step": 6140
    },
    {
      "epoch": 18.47,
      "grad_norm": 4.729291915893555,
      "learning_rate": 8.153153153153154e-06,
      "loss": 0.7396,
      "step": 6150
    },
    {
      "epoch": 18.5,
      "grad_norm": 6.148501396179199,
      "learning_rate": 8.150150150150151e-06,
      "loss": 0.7802,
      "step": 6160
    },
    {
      "epoch": 18.53,
      "grad_norm": 4.43754768371582,
      "learning_rate": 8.147147147147147e-06,
      "loss": 0.7856,
      "step": 6170
    },
    {
      "epoch": 18.56,
      "grad_norm": 5.1505231857299805,
      "learning_rate": 8.144144144144144e-06,
      "loss": 0.7998,
      "step": 6180
    },
    {
      "epoch": 18.59,
      "grad_norm": 4.7718305587768555,
      "learning_rate": 8.141141141141142e-06,
      "loss": 0.835,
      "step": 6190
    },
    {
      "epoch": 18.62,
      "grad_norm": 3.4573163986206055,
      "learning_rate": 8.13813813813814e-06,
      "loss": 0.7804,
      "step": 6200
    },
    {
      "epoch": 18.65,
      "grad_norm": 4.4927802085876465,
      "learning_rate": 8.135135135135137e-06,
      "loss": 0.7563,
      "step": 6210
    },
    {
      "epoch": 18.68,
      "grad_norm": 5.646451473236084,
      "learning_rate": 8.132132132132133e-06,
      "loss": 0.8245,
      "step": 6220
    },
    {
      "epoch": 18.71,
      "grad_norm": 4.9558892250061035,
      "learning_rate": 8.12912912912913e-06,
      "loss": 0.7667,
      "step": 6230
    },
    {
      "epoch": 18.74,
      "grad_norm": 3.646571159362793,
      "learning_rate": 8.126126126126128e-06,
      "loss": 0.8132,
      "step": 6240
    },
    {
      "epoch": 18.77,
      "grad_norm": 3.691779851913452,
      "learning_rate": 8.123123123123123e-06,
      "loss": 0.8171,
      "step": 6250
    },
    {
      "epoch": 18.8,
      "grad_norm": 4.477748394012451,
      "learning_rate": 8.12012012012012e-06,
      "loss": 0.8198,
      "step": 6260
    },
    {
      "epoch": 18.83,
      "grad_norm": 4.975278377532959,
      "learning_rate": 8.117117117117118e-06,
      "loss": 0.8123,
      "step": 6270
    },
    {
      "epoch": 18.86,
      "grad_norm": 4.925590515136719,
      "learning_rate": 8.114114114114114e-06,
      "loss": 0.8034,
      "step": 6280
    },
    {
      "epoch": 18.89,
      "grad_norm": 3.9182472229003906,
      "learning_rate": 8.111111111111112e-06,
      "loss": 0.7115,
      "step": 6290
    },
    {
      "epoch": 18.92,
      "grad_norm": 10.641616821289062,
      "learning_rate": 8.108108108108109e-06,
      "loss": 0.794,
      "step": 6300
    },
    {
      "epoch": 18.95,
      "grad_norm": 4.6885786056518555,
      "learning_rate": 8.105105105105105e-06,
      "loss": 0.8127,
      "step": 6310
    },
    {
      "epoch": 18.98,
      "grad_norm": 5.1985578536987305,
      "learning_rate": 8.102102102102102e-06,
      "loss": 0.6761,
      "step": 6320
    },
    {
      "epoch": 19.0,
      "eval_accuracy": 0.9075,
      "eval_loss": 0.5091767907142639,
      "eval_runtime": 26.335,
      "eval_samples_per_second": 379.723,
      "eval_steps_per_second": 1.519,
      "step": 6327
    },
    {
      "epoch": 19.01,
      "grad_norm": 3.5245776176452637,
      "learning_rate": 8.0990990990991e-06,
      "loss": 0.8029,
      "step": 6330
    },
    {
      "epoch": 19.04,
      "grad_norm": 4.780681133270264,
      "learning_rate": 8.096096096096097e-06,
      "loss": 0.8413,
      "step": 6340
    },
    {
      "epoch": 19.07,
      "grad_norm": 6.333914279937744,
      "learning_rate": 8.093093093093095e-06,
      "loss": 0.7798,
      "step": 6350
    },
    {
      "epoch": 19.1,
      "grad_norm": 6.965935707092285,
      "learning_rate": 8.09009009009009e-06,
      "loss": 0.7757,
      "step": 6360
    },
    {
      "epoch": 19.13,
      "grad_norm": 6.49807071685791,
      "learning_rate": 8.087087087087088e-06,
      "loss": 0.7936,
      "step": 6370
    },
    {
      "epoch": 19.16,
      "grad_norm": 3.88280987739563,
      "learning_rate": 8.084084084084085e-06,
      "loss": 0.6997,
      "step": 6380
    },
    {
      "epoch": 19.19,
      "grad_norm": 5.297958850860596,
      "learning_rate": 8.081081081081081e-06,
      "loss": 0.7622,
      "step": 6390
    },
    {
      "epoch": 19.22,
      "grad_norm": 4.647511959075928,
      "learning_rate": 8.078078078078079e-06,
      "loss": 0.7805,
      "step": 6400
    },
    {
      "epoch": 19.25,
      "grad_norm": 4.701191425323486,
      "learning_rate": 8.075075075075076e-06,
      "loss": 0.7882,
      "step": 6410
    },
    {
      "epoch": 19.28,
      "grad_norm": 5.120426177978516,
      "learning_rate": 8.072072072072072e-06,
      "loss": 0.767,
      "step": 6420
    },
    {
      "epoch": 19.31,
      "grad_norm": 4.553008079528809,
      "learning_rate": 8.06906906906907e-06,
      "loss": 0.7741,
      "step": 6430
    },
    {
      "epoch": 19.34,
      "grad_norm": 5.475098133087158,
      "learning_rate": 8.066066066066067e-06,
      "loss": 0.7638,
      "step": 6440
    },
    {
      "epoch": 19.37,
      "grad_norm": 3.622540235519409,
      "learning_rate": 8.063063063063063e-06,
      "loss": 0.7792,
      "step": 6450
    },
    {
      "epoch": 19.4,
      "grad_norm": 5.9133076667785645,
      "learning_rate": 8.06006006006006e-06,
      "loss": 0.7559,
      "step": 6460
    },
    {
      "epoch": 19.43,
      "grad_norm": 4.649819850921631,
      "learning_rate": 8.057057057057058e-06,
      "loss": 0.7506,
      "step": 6470
    },
    {
      "epoch": 19.46,
      "grad_norm": 4.202900409698486,
      "learning_rate": 8.054054054054055e-06,
      "loss": 0.714,
      "step": 6480
    },
    {
      "epoch": 19.49,
      "grad_norm": 6.11083459854126,
      "learning_rate": 8.051051051051053e-06,
      "loss": 0.8234,
      "step": 6490
    },
    {
      "epoch": 19.52,
      "grad_norm": 5.627078533172607,
      "learning_rate": 8.048048048048048e-06,
      "loss": 0.7255,
      "step": 6500
    },
    {
      "epoch": 19.55,
      "grad_norm": 3.8992910385131836,
      "learning_rate": 8.045045045045046e-06,
      "loss": 0.8122,
      "step": 6510
    },
    {
      "epoch": 19.58,
      "grad_norm": 5.471251010894775,
      "learning_rate": 8.042042042042043e-06,
      "loss": 0.7491,
      "step": 6520
    },
    {
      "epoch": 19.61,
      "grad_norm": 4.810721397399902,
      "learning_rate": 8.03903903903904e-06,
      "loss": 0.7327,
      "step": 6530
    },
    {
      "epoch": 19.64,
      "grad_norm": 4.824802875518799,
      "learning_rate": 8.036036036036037e-06,
      "loss": 0.7356,
      "step": 6540
    },
    {
      "epoch": 19.67,
      "grad_norm": 4.189336776733398,
      "learning_rate": 8.033033033033034e-06,
      "loss": 0.7634,
      "step": 6550
    },
    {
      "epoch": 19.7,
      "grad_norm": 6.21077299118042,
      "learning_rate": 8.03003003003003e-06,
      "loss": 0.7516,
      "step": 6560
    },
    {
      "epoch": 19.73,
      "grad_norm": 4.769148349761963,
      "learning_rate": 8.027027027027027e-06,
      "loss": 0.7596,
      "step": 6570
    },
    {
      "epoch": 19.76,
      "grad_norm": 4.551455020904541,
      "learning_rate": 8.024024024024025e-06,
      "loss": 0.7268,
      "step": 6580
    },
    {
      "epoch": 19.79,
      "grad_norm": 5.299874782562256,
      "learning_rate": 8.021021021021022e-06,
      "loss": 0.699,
      "step": 6590
    },
    {
      "epoch": 19.82,
      "grad_norm": 4.511897563934326,
      "learning_rate": 8.018018018018018e-06,
      "loss": 0.8107,
      "step": 6600
    },
    {
      "epoch": 19.85,
      "grad_norm": 3.8506810665130615,
      "learning_rate": 8.015015015015016e-06,
      "loss": 0.8022,
      "step": 6610
    },
    {
      "epoch": 19.88,
      "grad_norm": 4.714992523193359,
      "learning_rate": 8.012012012012013e-06,
      "loss": 0.7327,
      "step": 6620
    },
    {
      "epoch": 19.91,
      "grad_norm": 5.256269931793213,
      "learning_rate": 8.00900900900901e-06,
      "loss": 0.7623,
      "step": 6630
    },
    {
      "epoch": 19.94,
      "grad_norm": 4.982200622558594,
      "learning_rate": 8.006006006006006e-06,
      "loss": 0.7532,
      "step": 6640
    },
    {
      "epoch": 19.97,
      "grad_norm": 4.6494574546813965,
      "learning_rate": 8.003003003003004e-06,
      "loss": 0.7429,
      "step": 6650
    },
    {
      "epoch": 20.0,
      "grad_norm": 21.53489875793457,
      "learning_rate": 8.000000000000001e-06,
      "loss": 0.7064,
      "step": 6660
    },
    {
      "epoch": 20.0,
      "eval_accuracy": 0.9162,
      "eval_loss": 0.4759812355041504,
      "eval_runtime": 26.709,
      "eval_samples_per_second": 374.405,
      "eval_steps_per_second": 1.498,
      "step": 6660
    },
    {
      "epoch": 20.03,
      "grad_norm": 4.216804504394531,
      "learning_rate": 7.996996996996997e-06,
      "loss": 0.75,
      "step": 6670
    },
    {
      "epoch": 20.06,
      "grad_norm": 3.8654181957244873,
      "learning_rate": 7.993993993993995e-06,
      "loss": 0.7137,
      "step": 6680
    },
    {
      "epoch": 20.09,
      "grad_norm": 6.066682815551758,
      "learning_rate": 7.990990990990992e-06,
      "loss": 0.7759,
      "step": 6690
    },
    {
      "epoch": 20.12,
      "grad_norm": 3.9181618690490723,
      "learning_rate": 7.987987987987988e-06,
      "loss": 0.7022,
      "step": 6700
    },
    {
      "epoch": 20.15,
      "grad_norm": 4.61527156829834,
      "learning_rate": 7.984984984984985e-06,
      "loss": 0.6314,
      "step": 6710
    },
    {
      "epoch": 20.18,
      "grad_norm": 5.603513717651367,
      "learning_rate": 7.981981981981983e-06,
      "loss": 0.7045,
      "step": 6720
    },
    {
      "epoch": 20.21,
      "grad_norm": 3.54504132270813,
      "learning_rate": 7.97897897897898e-06,
      "loss": 0.7426,
      "step": 6730
    },
    {
      "epoch": 20.24,
      "grad_norm": 6.494993686676025,
      "learning_rate": 7.975975975975978e-06,
      "loss": 0.7779,
      "step": 6740
    },
    {
      "epoch": 20.27,
      "grad_norm": 4.856626510620117,
      "learning_rate": 7.972972972972974e-06,
      "loss": 0.7332,
      "step": 6750
    },
    {
      "epoch": 20.3,
      "grad_norm": 4.99015474319458,
      "learning_rate": 7.969969969969971e-06,
      "loss": 0.7016,
      "step": 6760
    },
    {
      "epoch": 20.33,
      "grad_norm": 4.442424297332764,
      "learning_rate": 7.966966966966969e-06,
      "loss": 0.7601,
      "step": 6770
    },
    {
      "epoch": 20.36,
      "grad_norm": 4.10810661315918,
      "learning_rate": 7.963963963963964e-06,
      "loss": 0.8095,
      "step": 6780
    },
    {
      "epoch": 20.39,
      "grad_norm": 5.386140823364258,
      "learning_rate": 7.960960960960962e-06,
      "loss": 0.7368,
      "step": 6790
    },
    {
      "epoch": 20.42,
      "grad_norm": 3.9678328037261963,
      "learning_rate": 7.95795795795796e-06,
      "loss": 0.7755,
      "step": 6800
    },
    {
      "epoch": 20.45,
      "grad_norm": 4.667354583740234,
      "learning_rate": 7.954954954954955e-06,
      "loss": 0.754,
      "step": 6810
    },
    {
      "epoch": 20.48,
      "grad_norm": 4.3042683601379395,
      "learning_rate": 7.951951951951953e-06,
      "loss": 0.7509,
      "step": 6820
    },
    {
      "epoch": 20.51,
      "grad_norm": 4.8756585121154785,
      "learning_rate": 7.948948948948948e-06,
      "loss": 0.7243,
      "step": 6830
    },
    {
      "epoch": 20.54,
      "grad_norm": 4.626062870025635,
      "learning_rate": 7.945945945945946e-06,
      "loss": 0.7072,
      "step": 6840
    },
    {
      "epoch": 20.57,
      "grad_norm": 7.135468482971191,
      "learning_rate": 7.942942942942943e-06,
      "loss": 0.6947,
      "step": 6850
    },
    {
      "epoch": 20.6,
      "grad_norm": 6.38662576675415,
      "learning_rate": 7.93993993993994e-06,
      "loss": 0.7602,
      "step": 6860
    },
    {
      "epoch": 20.63,
      "grad_norm": 5.721635818481445,
      "learning_rate": 7.936936936936938e-06,
      "loss": 0.668,
      "step": 6870
    },
    {
      "epoch": 20.66,
      "grad_norm": 4.916971683502197,
      "learning_rate": 7.933933933933936e-06,
      "loss": 0.6956,
      "step": 6880
    },
    {
      "epoch": 20.69,
      "grad_norm": 3.4424617290496826,
      "learning_rate": 7.930930930930931e-06,
      "loss": 0.7335,
      "step": 6890
    },
    {
      "epoch": 20.72,
      "grad_norm": 4.152014255523682,
      "learning_rate": 7.927927927927929e-06,
      "loss": 0.6986,
      "step": 6900
    },
    {
      "epoch": 20.75,
      "grad_norm": 5.503407001495361,
      "learning_rate": 7.924924924924926e-06,
      "loss": 0.7834,
      "step": 6910
    },
    {
      "epoch": 20.78,
      "grad_norm": 4.274868011474609,
      "learning_rate": 7.921921921921922e-06,
      "loss": 0.6885,
      "step": 6920
    },
    {
      "epoch": 20.81,
      "grad_norm": 4.806682109832764,
      "learning_rate": 7.91891891891892e-06,
      "loss": 0.6944,
      "step": 6930
    },
    {
      "epoch": 20.84,
      "grad_norm": 10.31600284576416,
      "learning_rate": 7.915915915915915e-06,
      "loss": 0.7379,
      "step": 6940
    },
    {
      "epoch": 20.87,
      "grad_norm": 5.682564735412598,
      "learning_rate": 7.912912912912913e-06,
      "loss": 0.7691,
      "step": 6950
    },
    {
      "epoch": 20.9,
      "grad_norm": 6.624948978424072,
      "learning_rate": 7.90990990990991e-06,
      "loss": 0.7127,
      "step": 6960
    },
    {
      "epoch": 20.93,
      "grad_norm": 4.5933146476745605,
      "learning_rate": 7.906906906906908e-06,
      "loss": 0.7696,
      "step": 6970
    },
    {
      "epoch": 20.96,
      "grad_norm": 5.328336238861084,
      "learning_rate": 7.903903903903904e-06,
      "loss": 0.7376,
      "step": 6980
    },
    {
      "epoch": 20.99,
      "grad_norm": 5.580850601196289,
      "learning_rate": 7.900900900900901e-06,
      "loss": 0.7224,
      "step": 6990
    },
    {
      "epoch": 21.0,
      "eval_accuracy": 0.9127,
      "eval_loss": 0.45015576481819153,
      "eval_runtime": 26.4818,
      "eval_samples_per_second": 377.619,
      "eval_steps_per_second": 1.51,
      "step": 6993
    },
    {
      "epoch": 21.02,
      "grad_norm": 4.94426965713501,
      "learning_rate": 7.897897897897899e-06,
      "loss": 0.89,
      "step": 7000
    },
    {
      "epoch": 21.05,
      "grad_norm": 4.805978298187256,
      "learning_rate": 7.894894894894896e-06,
      "loss": 0.7465,
      "step": 7010
    },
    {
      "epoch": 21.08,
      "grad_norm": 2.917875051498413,
      "learning_rate": 7.891891891891894e-06,
      "loss": 0.6913,
      "step": 7020
    },
    {
      "epoch": 21.11,
      "grad_norm": 5.422092914581299,
      "learning_rate": 7.88888888888889e-06,
      "loss": 0.6617,
      "step": 7030
    },
    {
      "epoch": 21.14,
      "grad_norm": 4.472092628479004,
      "learning_rate": 7.885885885885887e-06,
      "loss": 0.6749,
      "step": 7040
    },
    {
      "epoch": 21.17,
      "grad_norm": 4.079427242279053,
      "learning_rate": 7.882882882882884e-06,
      "loss": 0.6885,
      "step": 7050
    },
    {
      "epoch": 21.2,
      "grad_norm": 5.0683746337890625,
      "learning_rate": 7.87987987987988e-06,
      "loss": 0.7106,
      "step": 7060
    },
    {
      "epoch": 21.23,
      "grad_norm": 3.5775668621063232,
      "learning_rate": 7.876876876876878e-06,
      "loss": 0.7099,
      "step": 7070
    },
    {
      "epoch": 21.26,
      "grad_norm": 4.4518609046936035,
      "learning_rate": 7.873873873873873e-06,
      "loss": 0.6964,
      "step": 7080
    },
    {
      "epoch": 21.29,
      "grad_norm": 5.447687149047852,
      "learning_rate": 7.870870870870871e-06,
      "loss": 0.7655,
      "step": 7090
    },
    {
      "epoch": 21.32,
      "grad_norm": 5.607600212097168,
      "learning_rate": 7.867867867867868e-06,
      "loss": 0.6468,
      "step": 7100
    },
    {
      "epoch": 21.35,
      "grad_norm": 4.0597405433654785,
      "learning_rate": 7.864864864864866e-06,
      "loss": 0.7164,
      "step": 7110
    },
    {
      "epoch": 21.38,
      "grad_norm": 3.5956780910491943,
      "learning_rate": 7.861861861861863e-06,
      "loss": 0.6696,
      "step": 7120
    },
    {
      "epoch": 21.41,
      "grad_norm": 5.072645664215088,
      "learning_rate": 7.858858858858859e-06,
      "loss": 0.7251,
      "step": 7130
    },
    {
      "epoch": 21.44,
      "grad_norm": 5.149384021759033,
      "learning_rate": 7.855855855855857e-06,
      "loss": 0.7253,
      "step": 7140
    },
    {
      "epoch": 21.47,
      "grad_norm": 4.9628729820251465,
      "learning_rate": 7.852852852852854e-06,
      "loss": 0.6974,
      "step": 7150
    },
    {
      "epoch": 21.5,
      "grad_norm": 5.860617637634277,
      "learning_rate": 7.849849849849852e-06,
      "loss": 0.6898,
      "step": 7160
    },
    {
      "epoch": 21.53,
      "grad_norm": 3.291947364807129,
      "learning_rate": 7.846846846846847e-06,
      "loss": 0.7617,
      "step": 7170
    },
    {
      "epoch": 21.56,
      "grad_norm": 4.9691643714904785,
      "learning_rate": 7.843843843843845e-06,
      "loss": 0.6942,
      "step": 7180
    },
    {
      "epoch": 21.59,
      "grad_norm": 3.975054979324341,
      "learning_rate": 7.84084084084084e-06,
      "loss": 0.7012,
      "step": 7190
    },
    {
      "epoch": 21.62,
      "grad_norm": 5.745656967163086,
      "learning_rate": 7.837837837837838e-06,
      "loss": 0.6321,
      "step": 7200
    },
    {
      "epoch": 21.65,
      "grad_norm": 4.987421989440918,
      "learning_rate": 7.834834834834836e-06,
      "loss": 0.7249,
      "step": 7210
    },
    {
      "epoch": 21.68,
      "grad_norm": 3.740715980529785,
      "learning_rate": 7.831831831831831e-06,
      "loss": 0.7071,
      "step": 7220
    },
    {
      "epoch": 21.71,
      "grad_norm": 3.934730052947998,
      "learning_rate": 7.828828828828829e-06,
      "loss": 0.6625,
      "step": 7230
    },
    {
      "epoch": 21.74,
      "grad_norm": 4.433104991912842,
      "learning_rate": 7.825825825825826e-06,
      "loss": 0.7322,
      "step": 7240
    },
    {
      "epoch": 21.77,
      "grad_norm": 5.523029327392578,
      "learning_rate": 7.822822822822824e-06,
      "loss": 0.7235,
      "step": 7250
    },
    {
      "epoch": 21.8,
      "grad_norm": 5.185804843902588,
      "learning_rate": 7.819819819819821e-06,
      "loss": 0.7276,
      "step": 7260
    },
    {
      "epoch": 21.83,
      "grad_norm": 3.194908857345581,
      "learning_rate": 7.816816816816819e-06,
      "loss": 0.6634,
      "step": 7270
    },
    {
      "epoch": 21.86,
      "grad_norm": 4.772044658660889,
      "learning_rate": 7.813813813813815e-06,
      "loss": 0.694,
      "step": 7280
    },
    {
      "epoch": 21.89,
      "grad_norm": 8.646495819091797,
      "learning_rate": 7.810810810810812e-06,
      "loss": 0.6685,
      "step": 7290
    },
    {
      "epoch": 21.92,
      "grad_norm": 5.639026165008545,
      "learning_rate": 7.807807807807808e-06,
      "loss": 0.6577,
      "step": 7300
    },
    {
      "epoch": 21.95,
      "grad_norm": 3.9437367916107178,
      "learning_rate": 7.804804804804805e-06,
      "loss": 0.7355,
      "step": 7310
    },
    {
      "epoch": 21.98,
      "grad_norm": 7.671295166015625,
      "learning_rate": 7.801801801801803e-06,
      "loss": 0.712,
      "step": 7320
    },
    {
      "epoch": 22.0,
      "eval_accuracy": 0.913,
      "eval_loss": 0.42887207865715027,
      "eval_runtime": 25.909,
      "eval_samples_per_second": 385.966,
      "eval_steps_per_second": 1.544,
      "step": 7326
    },
    {
      "epoch": 22.01,
      "grad_norm": 5.855823993682861,
      "learning_rate": 7.798798798798799e-06,
      "loss": 0.7903,
      "step": 7330
    },
    {
      "epoch": 22.04,
      "grad_norm": 4.0135931968688965,
      "learning_rate": 7.795795795795796e-06,
      "loss": 0.6977,
      "step": 7340
    },
    {
      "epoch": 22.07,
      "grad_norm": 5.877246856689453,
      "learning_rate": 7.792792792792793e-06,
      "loss": 0.698,
      "step": 7350
    },
    {
      "epoch": 22.1,
      "grad_norm": 4.86218786239624,
      "learning_rate": 7.78978978978979e-06,
      "loss": 0.6627,
      "step": 7360
    },
    {
      "epoch": 22.13,
      "grad_norm": 4.43895149230957,
      "learning_rate": 7.786786786786787e-06,
      "loss": 0.6337,
      "step": 7370
    },
    {
      "epoch": 22.16,
      "grad_norm": 6.09828519821167,
      "learning_rate": 7.783783783783784e-06,
      "loss": 0.6661,
      "step": 7380
    },
    {
      "epoch": 22.19,
      "grad_norm": 3.708390235900879,
      "learning_rate": 7.780780780780782e-06,
      "loss": 0.6576,
      "step": 7390
    },
    {
      "epoch": 22.22,
      "grad_norm": 3.6503195762634277,
      "learning_rate": 7.77777777777778e-06,
      "loss": 0.665,
      "step": 7400
    },
    {
      "epoch": 22.25,
      "grad_norm": 5.96017599105835,
      "learning_rate": 7.774774774774777e-06,
      "loss": 0.6926,
      "step": 7410
    },
    {
      "epoch": 22.28,
      "grad_norm": 4.6361823081970215,
      "learning_rate": 7.771771771771772e-06,
      "loss": 0.6541,
      "step": 7420
    },
    {
      "epoch": 22.31,
      "grad_norm": 6.2728657722473145,
      "learning_rate": 7.76876876876877e-06,
      "loss": 0.6683,
      "step": 7430
    },
    {
      "epoch": 22.34,
      "grad_norm": 4.679554462432861,
      "learning_rate": 7.765765765765766e-06,
      "loss": 0.6363,
      "step": 7440
    },
    {
      "epoch": 22.37,
      "grad_norm": 5.297094821929932,
      "learning_rate": 7.762762762762763e-06,
      "loss": 0.6763,
      "step": 7450
    },
    {
      "epoch": 22.4,
      "grad_norm": 5.149691581726074,
      "learning_rate": 7.75975975975976e-06,
      "loss": 0.6713,
      "step": 7460
    },
    {
      "epoch": 22.43,
      "grad_norm": 4.853106498718262,
      "learning_rate": 7.756756756756756e-06,
      "loss": 0.6743,
      "step": 7470
    },
    {
      "epoch": 22.46,
      "grad_norm": 4.042623996734619,
      "learning_rate": 7.753753753753754e-06,
      "loss": 0.7152,
      "step": 7480
    },
    {
      "epoch": 22.49,
      "grad_norm": 4.078772068023682,
      "learning_rate": 7.750750750750751e-06,
      "loss": 0.705,
      "step": 7490
    },
    {
      "epoch": 22.52,
      "grad_norm": 6.992520332336426,
      "learning_rate": 7.747747747747749e-06,
      "loss": 0.6737,
      "step": 7500
    },
    {
      "epoch": 22.55,
      "grad_norm": 4.892293930053711,
      "learning_rate": 7.744744744744745e-06,
      "loss": 0.6685,
      "step": 7510
    },
    {
      "epoch": 22.58,
      "grad_norm": 3.3463852405548096,
      "learning_rate": 7.741741741741742e-06,
      "loss": 0.6417,
      "step": 7520
    },
    {
      "epoch": 22.61,
      "grad_norm": 4.571963310241699,
      "learning_rate": 7.73873873873874e-06,
      "loss": 0.678,
      "step": 7530
    },
    {
      "epoch": 22.64,
      "grad_norm": 4.723329544067383,
      "learning_rate": 7.735735735735737e-06,
      "loss": 0.6616,
      "step": 7540
    },
    {
      "epoch": 22.67,
      "grad_norm": 5.058360576629639,
      "learning_rate": 7.732732732732733e-06,
      "loss": 0.6625,
      "step": 7550
    },
    {
      "epoch": 22.7,
      "grad_norm": 4.79093599319458,
      "learning_rate": 7.72972972972973e-06,
      "loss": 0.6757,
      "step": 7560
    },
    {
      "epoch": 22.73,
      "grad_norm": 4.190638065338135,
      "learning_rate": 7.726726726726728e-06,
      "loss": 0.6625,
      "step": 7570
    },
    {
      "epoch": 22.76,
      "grad_norm": 4.085198879241943,
      "learning_rate": 7.723723723723724e-06,
      "loss": 0.7429,
      "step": 7580
    },
    {
      "epoch": 22.79,
      "grad_norm": 5.105234146118164,
      "learning_rate": 7.720720720720721e-06,
      "loss": 0.6598,
      "step": 7590
    },
    {
      "epoch": 22.82,
      "grad_norm": 5.374044895172119,
      "learning_rate": 7.717717717717719e-06,
      "loss": 0.7003,
      "step": 7600
    },
    {
      "epoch": 22.85,
      "grad_norm": 6.302853107452393,
      "learning_rate": 7.714714714714714e-06,
      "loss": 0.6475,
      "step": 7610
    },
    {
      "epoch": 22.88,
      "grad_norm": 4.560659885406494,
      "learning_rate": 7.711711711711712e-06,
      "loss": 0.695,
      "step": 7620
    },
    {
      "epoch": 22.91,
      "grad_norm": 4.321108818054199,
      "learning_rate": 7.70870870870871e-06,
      "loss": 0.6393,
      "step": 7630
    },
    {
      "epoch": 22.94,
      "grad_norm": 4.939061164855957,
      "learning_rate": 7.705705705705707e-06,
      "loss": 0.6672,
      "step": 7640
    },
    {
      "epoch": 22.97,
      "grad_norm": 3.5717639923095703,
      "learning_rate": 7.702702702702704e-06,
      "loss": 0.6541,
      "step": 7650
    },
    {
      "epoch": 23.0,
      "eval_accuracy": 0.913,
      "eval_loss": 0.4087793827056885,
      "eval_runtime": 26.1621,
      "eval_samples_per_second": 382.232,
      "eval_steps_per_second": 1.529,
      "step": 7659
    },
    {
      "epoch": 23.0,
      "grad_norm": 6.035078048706055,
      "learning_rate": 7.6996996996997e-06,
      "loss": 0.6338,
      "step": 7660
    },
    {
      "epoch": 23.03,
      "grad_norm": 6.63695764541626,
      "learning_rate": 7.696696696696698e-06,
      "loss": 0.6343,
      "step": 7670
    },
    {
      "epoch": 23.06,
      "grad_norm": 4.683477878570557,
      "learning_rate": 7.693693693693695e-06,
      "loss": 0.6791,
      "step": 7680
    },
    {
      "epoch": 23.09,
      "grad_norm": 4.074741363525391,
      "learning_rate": 7.69069069069069e-06,
      "loss": 0.6373,
      "step": 7690
    },
    {
      "epoch": 23.12,
      "grad_norm": 3.735003709793091,
      "learning_rate": 7.687687687687688e-06,
      "loss": 0.5999,
      "step": 7700
    },
    {
      "epoch": 23.15,
      "grad_norm": 4.406135082244873,
      "learning_rate": 7.684684684684686e-06,
      "loss": 0.6841,
      "step": 7710
    },
    {
      "epoch": 23.18,
      "grad_norm": 5.173962116241455,
      "learning_rate": 7.681681681681682e-06,
      "loss": 0.6907,
      "step": 7720
    },
    {
      "epoch": 23.21,
      "grad_norm": 5.311886310577393,
      "learning_rate": 7.678678678678679e-06,
      "loss": 0.6303,
      "step": 7730
    },
    {
      "epoch": 23.24,
      "grad_norm": 8.055031776428223,
      "learning_rate": 7.675675675675676e-06,
      "loss": 0.6364,
      "step": 7740
    },
    {
      "epoch": 23.27,
      "grad_norm": 4.631924152374268,
      "learning_rate": 7.672672672672672e-06,
      "loss": 0.6847,
      "step": 7750
    },
    {
      "epoch": 23.3,
      "grad_norm": 3.524216413497925,
      "learning_rate": 7.66966966966967e-06,
      "loss": 0.6307,
      "step": 7760
    },
    {
      "epoch": 23.33,
      "grad_norm": 6.029527187347412,
      "learning_rate": 7.666666666666667e-06,
      "loss": 0.6106,
      "step": 7770
    },
    {
      "epoch": 23.36,
      "grad_norm": 4.7900543212890625,
      "learning_rate": 7.663663663663665e-06,
      "loss": 0.6498,
      "step": 7780
    },
    {
      "epoch": 23.39,
      "grad_norm": 5.736597537994385,
      "learning_rate": 7.660660660660662e-06,
      "loss": 0.6356,
      "step": 7790
    },
    {
      "epoch": 23.42,
      "grad_norm": 4.166688919067383,
      "learning_rate": 7.657657657657658e-06,
      "loss": 0.657,
      "step": 7800
    },
    {
      "epoch": 23.45,
      "grad_norm": 5.274576187133789,
      "learning_rate": 7.654654654654655e-06,
      "loss": 0.6014,
      "step": 7810
    },
    {
      "epoch": 23.48,
      "grad_norm": 6.013502597808838,
      "learning_rate": 7.651651651651653e-06,
      "loss": 0.6372,
      "step": 7820
    },
    {
      "epoch": 23.51,
      "grad_norm": 5.510165214538574,
      "learning_rate": 7.648648648648649e-06,
      "loss": 0.6537,
      "step": 7830
    },
    {
      "epoch": 23.54,
      "grad_norm": 4.288418769836426,
      "learning_rate": 7.645645645645646e-06,
      "loss": 0.6389,
      "step": 7840
    },
    {
      "epoch": 23.57,
      "grad_norm": 4.00223970413208,
      "learning_rate": 7.642642642642644e-06,
      "loss": 0.7018,
      "step": 7850
    },
    {
      "epoch": 23.6,
      "grad_norm": 3.205495595932007,
      "learning_rate": 7.63963963963964e-06,
      "loss": 0.5617,
      "step": 7860
    },
    {
      "epoch": 23.63,
      "grad_norm": 5.546922206878662,
      "learning_rate": 7.636636636636637e-06,
      "loss": 0.6275,
      "step": 7870
    },
    {
      "epoch": 23.66,
      "grad_norm": 4.7760114669799805,
      "learning_rate": 7.633633633633634e-06,
      "loss": 0.6531,
      "step": 7880
    },
    {
      "epoch": 23.69,
      "grad_norm": 4.954517841339111,
      "learning_rate": 7.63063063063063e-06,
      "loss": 0.6735,
      "step": 7890
    },
    {
      "epoch": 23.72,
      "grad_norm": 3.2795095443725586,
      "learning_rate": 7.6276276276276285e-06,
      "loss": 0.639,
      "step": 7900
    },
    {
      "epoch": 23.75,
      "grad_norm": 4.749567031860352,
      "learning_rate": 7.624624624624624e-06,
      "loss": 0.6141,
      "step": 7910
    },
    {
      "epoch": 23.78,
      "grad_norm": 4.061180114746094,
      "learning_rate": 7.621621621621622e-06,
      "loss": 0.6722,
      "step": 7920
    },
    {
      "epoch": 23.81,
      "grad_norm": 5.958366870880127,
      "learning_rate": 7.618618618618619e-06,
      "loss": 0.6463,
      "step": 7930
    },
    {
      "epoch": 23.84,
      "grad_norm": 4.867851257324219,
      "learning_rate": 7.615615615615616e-06,
      "loss": 0.6448,
      "step": 7940
    },
    {
      "epoch": 23.87,
      "grad_norm": 6.845958232879639,
      "learning_rate": 7.612612612612613e-06,
      "loss": 0.6182,
      "step": 7950
    },
    {
      "epoch": 23.9,
      "grad_norm": 4.476239204406738,
      "learning_rate": 7.609609609609611e-06,
      "loss": 0.5616,
      "step": 7960
    },
    {
      "epoch": 23.93,
      "grad_norm": 3.8598573207855225,
      "learning_rate": 7.606606606606607e-06,
      "loss": 0.6865,
      "step": 7970
    },
    {
      "epoch": 23.96,
      "grad_norm": 4.300548076629639,
      "learning_rate": 7.603603603603604e-06,
      "loss": 0.6685,
      "step": 7980
    },
    {
      "epoch": 23.99,
      "grad_norm": 5.2045512199401855,
      "learning_rate": 7.600600600600602e-06,
      "loss": 0.6338,
      "step": 7990
    },
    {
      "epoch": 24.0,
      "eval_accuracy": 0.9172,
      "eval_loss": 0.3914088010787964,
      "eval_runtime": 26.5071,
      "eval_samples_per_second": 377.258,
      "eval_steps_per_second": 1.509,
      "step": 7992
    },
    {
      "epoch": 24.02,
      "grad_norm": 4.160776138305664,
      "learning_rate": 7.597597597597598e-06,
      "loss": 0.6086,
      "step": 8000
    },
    {
      "epoch": 24.05,
      "grad_norm": 5.0602850914001465,
      "learning_rate": 7.594594594594596e-06,
      "loss": 0.6292,
      "step": 8010
    },
    {
      "epoch": 24.08,
      "grad_norm": 5.049746990203857,
      "learning_rate": 7.591591591591592e-06,
      "loss": 0.6355,
      "step": 8020
    },
    {
      "epoch": 24.11,
      "grad_norm": 2.8283441066741943,
      "learning_rate": 7.588588588588589e-06,
      "loss": 0.5961,
      "step": 8030
    },
    {
      "epoch": 24.14,
      "grad_norm": 4.665635585784912,
      "learning_rate": 7.5855855855855865e-06,
      "loss": 0.6135,
      "step": 8040
    },
    {
      "epoch": 24.17,
      "grad_norm": 5.210675239562988,
      "learning_rate": 7.582582582582583e-06,
      "loss": 0.6286,
      "step": 8050
    },
    {
      "epoch": 24.2,
      "grad_norm": 5.123960971832275,
      "learning_rate": 7.57957957957958e-06,
      "loss": 0.6597,
      "step": 8060
    },
    {
      "epoch": 24.23,
      "grad_norm": 4.584262847900391,
      "learning_rate": 7.576576576576577e-06,
      "loss": 0.6415,
      "step": 8070
    },
    {
      "epoch": 24.26,
      "grad_norm": 6.948488712310791,
      "learning_rate": 7.573573573573574e-06,
      "loss": 0.602,
      "step": 8080
    },
    {
      "epoch": 24.29,
      "grad_norm": 5.196147441864014,
      "learning_rate": 7.570570570570571e-06,
      "loss": 0.6441,
      "step": 8090
    },
    {
      "epoch": 24.32,
      "grad_norm": 4.633663177490234,
      "learning_rate": 7.567567567567569e-06,
      "loss": 0.6568,
      "step": 8100
    },
    {
      "epoch": 24.35,
      "grad_norm": 4.8977484703063965,
      "learning_rate": 7.5645645645645646e-06,
      "loss": 0.6384,
      "step": 8110
    },
    {
      "epoch": 24.38,
      "grad_norm": 4.151755332946777,
      "learning_rate": 7.561561561561562e-06,
      "loss": 0.6328,
      "step": 8120
    },
    {
      "epoch": 24.41,
      "grad_norm": 4.1300883293151855,
      "learning_rate": 7.5585585585585595e-06,
      "loss": 0.6744,
      "step": 8130
    },
    {
      "epoch": 24.44,
      "grad_norm": 5.872833728790283,
      "learning_rate": 7.555555555555556e-06,
      "loss": 0.6554,
      "step": 8140
    },
    {
      "epoch": 24.47,
      "grad_norm": 4.006340503692627,
      "learning_rate": 7.552552552552554e-06,
      "loss": 0.63,
      "step": 8150
    },
    {
      "epoch": 24.5,
      "grad_norm": 5.045310974121094,
      "learning_rate": 7.549549549549549e-06,
      "loss": 0.6258,
      "step": 8160
    },
    {
      "epoch": 24.53,
      "grad_norm": 4.250802040100098,
      "learning_rate": 7.546546546546547e-06,
      "loss": 0.6523,
      "step": 8170
    },
    {
      "epoch": 24.56,
      "grad_norm": 4.446349620819092,
      "learning_rate": 7.543543543543544e-06,
      "loss": 0.61,
      "step": 8180
    },
    {
      "epoch": 24.59,
      "grad_norm": 5.197299957275391,
      "learning_rate": 7.540540540540541e-06,
      "loss": 0.619,
      "step": 8190
    },
    {
      "epoch": 24.62,
      "grad_norm": 4.9567484855651855,
      "learning_rate": 7.5375375375375385e-06,
      "loss": 0.5888,
      "step": 8200
    },
    {
      "epoch": 24.65,
      "grad_norm": 4.429915428161621,
      "learning_rate": 7.534534534534535e-06,
      "loss": 0.6331,
      "step": 8210
    },
    {
      "epoch": 24.68,
      "grad_norm": 5.928262233734131,
      "learning_rate": 7.531531531531532e-06,
      "loss": 0.6002,
      "step": 8220
    },
    {
      "epoch": 24.71,
      "grad_norm": 3.7171638011932373,
      "learning_rate": 7.528528528528529e-06,
      "loss": 0.6216,
      "step": 8230
    },
    {
      "epoch": 24.74,
      "grad_norm": 3.967848300933838,
      "learning_rate": 7.525525525525527e-06,
      "loss": 0.5985,
      "step": 8240
    },
    {
      "epoch": 24.77,
      "grad_norm": 4.106054306030273,
      "learning_rate": 7.5225225225225225e-06,
      "loss": 0.6488,
      "step": 8250
    },
    {
      "epoch": 24.8,
      "grad_norm": 4.115420341491699,
      "learning_rate": 7.51951951951952e-06,
      "loss": 0.6393,
      "step": 8260
    },
    {
      "epoch": 24.83,
      "grad_norm": 5.088920593261719,
      "learning_rate": 7.516516516516517e-06,
      "loss": 0.6386,
      "step": 8270
    },
    {
      "epoch": 24.86,
      "grad_norm": 4.073686122894287,
      "learning_rate": 7.513513513513514e-06,
      "loss": 0.5533,
      "step": 8280
    },
    {
      "epoch": 24.89,
      "grad_norm": 4.537994861602783,
      "learning_rate": 7.5105105105105116e-06,
      "loss": 0.6044,
      "step": 8290
    },
    {
      "epoch": 24.92,
      "grad_norm": 4.048253059387207,
      "learning_rate": 7.507507507507507e-06,
      "loss": 0.5597,
      "step": 8300
    },
    {
      "epoch": 24.95,
      "grad_norm": 6.220718860626221,
      "learning_rate": 7.504504504504505e-06,
      "loss": 0.6354,
      "step": 8310
    },
    {
      "epoch": 24.98,
      "grad_norm": 4.656418800354004,
      "learning_rate": 7.501501501501502e-06,
      "loss": 0.6097,
      "step": 8320
    },
    {
      "epoch": 25.0,
      "eval_accuracy": 0.9182,
      "eval_loss": 0.37756532430648804,
      "eval_runtime": 26.2998,
      "eval_samples_per_second": 380.232,
      "eval_steps_per_second": 1.521,
      "step": 8325
    },
    {
      "epoch": 25.02,
      "grad_norm": 4.741297721862793,
      "learning_rate": 7.498498498498499e-06,
      "loss": 0.6371,
      "step": 8330
    },
    {
      "epoch": 25.05,
      "grad_norm": 7.324318885803223,
      "learning_rate": 7.495495495495496e-06,
      "loss": 0.6157,
      "step": 8340
    },
    {
      "epoch": 25.08,
      "grad_norm": 5.345495700836182,
      "learning_rate": 7.492492492492494e-06,
      "loss": 0.6325,
      "step": 8350
    },
    {
      "epoch": 25.11,
      "grad_norm": 5.9051690101623535,
      "learning_rate": 7.48948948948949e-06,
      "loss": 0.5899,
      "step": 8360
    },
    {
      "epoch": 25.14,
      "grad_norm": 6.787704944610596,
      "learning_rate": 7.486486486486487e-06,
      "loss": 0.6416,
      "step": 8370
    },
    {
      "epoch": 25.17,
      "grad_norm": 4.713385581970215,
      "learning_rate": 7.483483483483485e-06,
      "loss": 0.6718,
      "step": 8380
    },
    {
      "epoch": 25.2,
      "grad_norm": 5.411506652832031,
      "learning_rate": 7.480480480480481e-06,
      "loss": 0.6059,
      "step": 8390
    },
    {
      "epoch": 25.23,
      "grad_norm": 4.1856207847595215,
      "learning_rate": 7.477477477477479e-06,
      "loss": 0.5297,
      "step": 8400
    },
    {
      "epoch": 25.26,
      "grad_norm": 3.9937260150909424,
      "learning_rate": 7.4744744744744745e-06,
      "loss": 0.607,
      "step": 8410
    },
    {
      "epoch": 25.29,
      "grad_norm": 3.861163854598999,
      "learning_rate": 7.471471471471472e-06,
      "loss": 0.5885,
      "step": 8420
    },
    {
      "epoch": 25.32,
      "grad_norm": 5.328300952911377,
      "learning_rate": 7.4684684684684695e-06,
      "loss": 0.695,
      "step": 8430
    },
    {
      "epoch": 25.35,
      "grad_norm": 4.425115585327148,
      "learning_rate": 7.465465465465466e-06,
      "loss": 0.5812,
      "step": 8440
    },
    {
      "epoch": 25.38,
      "grad_norm": 5.137642860412598,
      "learning_rate": 7.462462462462463e-06,
      "loss": 0.6438,
      "step": 8450
    },
    {
      "epoch": 25.41,
      "grad_norm": 4.726527690887451,
      "learning_rate": 7.45945945945946e-06,
      "loss": 0.621,
      "step": 8460
    },
    {
      "epoch": 25.44,
      "grad_norm": 4.208578109741211,
      "learning_rate": 7.456456456456457e-06,
      "loss": 0.6252,
      "step": 8470
    },
    {
      "epoch": 25.47,
      "grad_norm": 4.630558013916016,
      "learning_rate": 7.453453453453454e-06,
      "loss": 0.5678,
      "step": 8480
    },
    {
      "epoch": 25.5,
      "grad_norm": 5.314132213592529,
      "learning_rate": 7.450450450450452e-06,
      "loss": 0.6658,
      "step": 8490
    },
    {
      "epoch": 25.53,
      "grad_norm": 5.784879207611084,
      "learning_rate": 7.447447447447448e-06,
      "loss": 0.622,
      "step": 8500
    },
    {
      "epoch": 25.56,
      "grad_norm": 5.003414630889893,
      "learning_rate": 7.444444444444445e-06,
      "loss": 0.6246,
      "step": 8510
    },
    {
      "epoch": 25.59,
      "grad_norm": 3.977597236633301,
      "learning_rate": 7.441441441441442e-06,
      "loss": 0.6009,
      "step": 8520
    },
    {
      "epoch": 25.62,
      "grad_norm": 5.353960037231445,
      "learning_rate": 7.438438438438439e-06,
      "loss": 0.6211,
      "step": 8530
    },
    {
      "epoch": 25.65,
      "grad_norm": 5.512233257293701,
      "learning_rate": 7.435435435435437e-06,
      "loss": 0.6511,
      "step": 8540
    },
    {
      "epoch": 25.68,
      "grad_norm": 3.425570011138916,
      "learning_rate": 7.4324324324324324e-06,
      "loss": 0.6653,
      "step": 8550
    },
    {
      "epoch": 25.71,
      "grad_norm": 5.093016147613525,
      "learning_rate": 7.42942942942943e-06,
      "loss": 0.649,
      "step": 8560
    },
    {
      "epoch": 25.74,
      "grad_norm": 4.121246814727783,
      "learning_rate": 7.426426426426427e-06,
      "loss": 0.6374,
      "step": 8570
    },
    {
      "epoch": 25.77,
      "grad_norm": 4.50221586227417,
      "learning_rate": 7.423423423423424e-06,
      "loss": 0.5874,
      "step": 8580
    },
    {
      "epoch": 25.8,
      "grad_norm": 5.9485063552856445,
      "learning_rate": 7.4204204204204215e-06,
      "loss": 0.6417,
      "step": 8590
    },
    {
      "epoch": 25.83,
      "grad_norm": 4.642158508300781,
      "learning_rate": 7.417417417417418e-06,
      "loss": 0.5855,
      "step": 8600
    },
    {
      "epoch": 25.86,
      "grad_norm": 5.189647674560547,
      "learning_rate": 7.414414414414415e-06,
      "loss": 0.5878,
      "step": 8610
    },
    {
      "epoch": 25.89,
      "grad_norm": 4.6904191970825195,
      "learning_rate": 7.411411411411412e-06,
      "loss": 0.637,
      "step": 8620
    },
    {
      "epoch": 25.92,
      "grad_norm": 3.443445920944214,
      "learning_rate": 7.408408408408409e-06,
      "loss": 0.607,
      "step": 8630
    },
    {
      "epoch": 25.95,
      "grad_norm": 4.997288703918457,
      "learning_rate": 7.4054054054054055e-06,
      "loss": 0.602,
      "step": 8640
    },
    {
      "epoch": 25.98,
      "grad_norm": 4.938547134399414,
      "learning_rate": 7.402402402402403e-06,
      "loss": 0.6369,
      "step": 8650
    },
    {
      "epoch": 26.0,
      "eval_accuracy": 0.9155,
      "eval_loss": 0.3676045536994934,
      "eval_runtime": 26.6726,
      "eval_samples_per_second": 374.916,
      "eval_steps_per_second": 1.5,
      "step": 8658
    },
    {
      "epoch": 26.01,
      "grad_norm": 5.075904846191406,
      "learning_rate": 7.3993993993994e-06,
      "loss": 0.5574,
      "step": 8660
    },
    {
      "epoch": 26.04,
      "grad_norm": 4.672639846801758,
      "learning_rate": 7.396396396396397e-06,
      "loss": 0.5918,
      "step": 8670
    },
    {
      "epoch": 26.07,
      "grad_norm": 4.947727203369141,
      "learning_rate": 7.393393393393395e-06,
      "loss": 0.6122,
      "step": 8680
    },
    {
      "epoch": 26.1,
      "grad_norm": 3.5392355918884277,
      "learning_rate": 7.39039039039039e-06,
      "loss": 0.5744,
      "step": 8690
    },
    {
      "epoch": 26.13,
      "grad_norm": 6.616043567657471,
      "learning_rate": 7.387387387387388e-06,
      "loss": 0.5518,
      "step": 8700
    },
    {
      "epoch": 26.16,
      "grad_norm": 3.9501121044158936,
      "learning_rate": 7.384384384384385e-06,
      "loss": 0.5757,
      "step": 8710
    },
    {
      "epoch": 26.19,
      "grad_norm": 6.87493896484375,
      "learning_rate": 7.381381381381382e-06,
      "loss": 0.6087,
      "step": 8720
    },
    {
      "epoch": 26.22,
      "grad_norm": 3.392703056335449,
      "learning_rate": 7.3783783783783794e-06,
      "loss": 0.6107,
      "step": 8730
    },
    {
      "epoch": 26.25,
      "grad_norm": 5.120011806488037,
      "learning_rate": 7.375375375375377e-06,
      "loss": 0.614,
      "step": 8740
    },
    {
      "epoch": 26.28,
      "grad_norm": 4.2964348793029785,
      "learning_rate": 7.372372372372373e-06,
      "loss": 0.5861,
      "step": 8750
    },
    {
      "epoch": 26.31,
      "grad_norm": 4.34578275680542,
      "learning_rate": 7.36936936936937e-06,
      "loss": 0.664,
      "step": 8760
    },
    {
      "epoch": 26.34,
      "grad_norm": 4.854133129119873,
      "learning_rate": 7.366366366366367e-06,
      "loss": 0.5872,
      "step": 8770
    },
    {
      "epoch": 26.37,
      "grad_norm": 5.136003494262695,
      "learning_rate": 7.363363363363364e-06,
      "loss": 0.6231,
      "step": 8780
    },
    {
      "epoch": 26.4,
      "grad_norm": 6.438120365142822,
      "learning_rate": 7.360360360360361e-06,
      "loss": 0.64,
      "step": 8790
    },
    {
      "epoch": 26.43,
      "grad_norm": 5.689127445220947,
      "learning_rate": 7.3573573573573575e-06,
      "loss": 0.7184,
      "step": 8800
    },
    {
      "epoch": 26.46,
      "grad_norm": 4.496994495391846,
      "learning_rate": 7.354354354354355e-06,
      "loss": 0.633,
      "step": 8810
    },
    {
      "epoch": 26.49,
      "grad_norm": 4.275875568389893,
      "learning_rate": 7.3513513513513525e-06,
      "loss": 0.6198,
      "step": 8820
    },
    {
      "epoch": 26.52,
      "grad_norm": 5.224747180938721,
      "learning_rate": 7.348348348348348e-06,
      "loss": 0.5837,
      "step": 8830
    },
    {
      "epoch": 26.55,
      "grad_norm": 5.464904308319092,
      "learning_rate": 7.345345345345346e-06,
      "loss": 0.5972,
      "step": 8840
    },
    {
      "epoch": 26.58,
      "grad_norm": 15.829516410827637,
      "learning_rate": 7.342342342342343e-06,
      "loss": 0.6058,
      "step": 8850
    },
    {
      "epoch": 26.61,
      "grad_norm": 5.066461563110352,
      "learning_rate": 7.33933933933934e-06,
      "loss": 0.6161,
      "step": 8860
    },
    {
      "epoch": 26.64,
      "grad_norm": 6.489962577819824,
      "learning_rate": 7.336336336336337e-06,
      "loss": 0.6586,
      "step": 8870
    },
    {
      "epoch": 26.67,
      "grad_norm": 4.583408832550049,
      "learning_rate": 7.333333333333333e-06,
      "loss": 0.6123,
      "step": 8880
    },
    {
      "epoch": 26.7,
      "grad_norm": 4.31550931930542,
      "learning_rate": 7.330330330330331e-06,
      "loss": 0.5898,
      "step": 8890
    },
    {
      "epoch": 26.73,
      "grad_norm": 3.4311511516571045,
      "learning_rate": 7.327327327327328e-06,
      "loss": 0.5722,
      "step": 8900
    },
    {
      "epoch": 26.76,
      "grad_norm": 4.0307464599609375,
      "learning_rate": 7.324324324324325e-06,
      "loss": 0.5315,
      "step": 8910
    },
    {
      "epoch": 26.79,
      "grad_norm": 4.3041911125183105,
      "learning_rate": 7.321321321321322e-06,
      "loss": 0.5483,
      "step": 8920
    },
    {
      "epoch": 26.82,
      "grad_norm": 5.518500804901123,
      "learning_rate": 7.31831831831832e-06,
      "loss": 0.6019,
      "step": 8930
    },
    {
      "epoch": 26.85,
      "grad_norm": 4.37266206741333,
      "learning_rate": 7.3153153153153155e-06,
      "loss": 0.5912,
      "step": 8940
    },
    {
      "epoch": 26.88,
      "grad_norm": 6.165101051330566,
      "learning_rate": 7.312312312312313e-06,
      "loss": 0.5828,
      "step": 8950
    },
    {
      "epoch": 26.91,
      "grad_norm": 4.071359157562256,
      "learning_rate": 7.3093093093093104e-06,
      "loss": 0.5642,
      "step": 8960
    },
    {
      "epoch": 26.94,
      "grad_norm": 7.619660377502441,
      "learning_rate": 7.306306306306307e-06,
      "loss": 0.608,
      "step": 8970
    },
    {
      "epoch": 26.97,
      "grad_norm": 5.814097881317139,
      "learning_rate": 7.303303303303304e-06,
      "loss": 0.5829,
      "step": 8980
    },
    {
      "epoch": 27.0,
      "grad_norm": 3.6844120025634766,
      "learning_rate": 7.3003003003003e-06,
      "loss": 0.6007,
      "step": 8990
    },
    {
      "epoch": 27.0,
      "eval_accuracy": 0.9149,
      "eval_loss": 0.3631533086299896,
      "eval_runtime": 26.1151,
      "eval_samples_per_second": 382.92,
      "eval_steps_per_second": 1.532,
      "step": 8991
    },
    {
      "epoch": 27.03,
      "grad_norm": 3.1218578815460205,
      "learning_rate": 7.297297297297298e-06,
      "loss": 0.5263,
      "step": 9000
    },
    {
      "epoch": 27.06,
      "grad_norm": 31.088542938232422,
      "learning_rate": 7.294294294294295e-06,
      "loss": 0.5519,
      "step": 9010
    },
    {
      "epoch": 27.09,
      "grad_norm": 5.128907203674316,
      "learning_rate": 7.291291291291291e-06,
      "loss": 0.5794,
      "step": 9020
    },
    {
      "epoch": 27.12,
      "grad_norm": 5.448819160461426,
      "learning_rate": 7.2882882882882885e-06,
      "loss": 0.5923,
      "step": 9030
    },
    {
      "epoch": 27.15,
      "grad_norm": 5.304131031036377,
      "learning_rate": 7.285285285285286e-06,
      "loss": 0.6297,
      "step": 9040
    },
    {
      "epoch": 27.18,
      "grad_norm": 3.740703582763672,
      "learning_rate": 7.282282282282283e-06,
      "loss": 0.5815,
      "step": 9050
    },
    {
      "epoch": 27.21,
      "grad_norm": 4.914813995361328,
      "learning_rate": 7.27927927927928e-06,
      "loss": 0.5864,
      "step": 9060
    },
    {
      "epoch": 27.24,
      "grad_norm": 4.869528770446777,
      "learning_rate": 7.276276276276278e-06,
      "loss": 0.5698,
      "step": 9070
    },
    {
      "epoch": 27.27,
      "grad_norm": 4.342233180999756,
      "learning_rate": 7.273273273273273e-06,
      "loss": 0.5462,
      "step": 9080
    },
    {
      "epoch": 27.3,
      "grad_norm": 4.266785144805908,
      "learning_rate": 7.270270270270271e-06,
      "loss": 0.5573,
      "step": 9090
    },
    {
      "epoch": 27.33,
      "grad_norm": 3.1235883235931396,
      "learning_rate": 7.267267267267268e-06,
      "loss": 0.5371,
      "step": 9100
    },
    {
      "epoch": 27.36,
      "grad_norm": 5.869687080383301,
      "learning_rate": 7.264264264264265e-06,
      "loss": 0.613,
      "step": 9110
    },
    {
      "epoch": 27.39,
      "grad_norm": 4.293118476867676,
      "learning_rate": 7.2612612612612625e-06,
      "loss": 0.6284,
      "step": 9120
    },
    {
      "epoch": 27.42,
      "grad_norm": 6.922875881195068,
      "learning_rate": 7.258258258258258e-06,
      "loss": 0.611,
      "step": 9130
    },
    {
      "epoch": 27.45,
      "grad_norm": 3.4923338890075684,
      "learning_rate": 7.255255255255256e-06,
      "loss": 0.5713,
      "step": 9140
    },
    {
      "epoch": 27.48,
      "grad_norm": 4.0018696784973145,
      "learning_rate": 7.252252252252253e-06,
      "loss": 0.6327,
      "step": 9150
    },
    {
      "epoch": 27.51,
      "grad_norm": 6.069986820220947,
      "learning_rate": 7.24924924924925e-06,
      "loss": 0.6136,
      "step": 9160
    },
    {
      "epoch": 27.54,
      "grad_norm": 4.478414058685303,
      "learning_rate": 7.2462462462462465e-06,
      "loss": 0.5468,
      "step": 9170
    },
    {
      "epoch": 27.57,
      "grad_norm": 6.923449993133545,
      "learning_rate": 7.243243243243244e-06,
      "loss": 0.5554,
      "step": 9180
    },
    {
      "epoch": 27.6,
      "grad_norm": 3.9428858757019043,
      "learning_rate": 7.240240240240241e-06,
      "loss": 0.5201,
      "step": 9190
    },
    {
      "epoch": 27.63,
      "grad_norm": 5.154752254486084,
      "learning_rate": 7.237237237237238e-06,
      "loss": 0.6153,
      "step": 9200
    },
    {
      "epoch": 27.66,
      "grad_norm": 7.000457286834717,
      "learning_rate": 7.2342342342342355e-06,
      "loss": 0.5862,
      "step": 9210
    },
    {
      "epoch": 27.69,
      "grad_norm": 4.479945182800293,
      "learning_rate": 7.231231231231231e-06,
      "loss": 0.5617,
      "step": 9220
    },
    {
      "epoch": 27.72,
      "grad_norm": 3.5248751640319824,
      "learning_rate": 7.228228228228229e-06,
      "loss": 0.6648,
      "step": 9230
    },
    {
      "epoch": 27.75,
      "grad_norm": 4.968978404998779,
      "learning_rate": 7.2252252252252254e-06,
      "loss": 0.6252,
      "step": 9240
    },
    {
      "epoch": 27.78,
      "grad_norm": 5.074804306030273,
      "learning_rate": 7.222222222222223e-06,
      "loss": 0.6298,
      "step": 9250
    },
    {
      "epoch": 27.81,
      "grad_norm": 6.238649368286133,
      "learning_rate": 7.21921921921922e-06,
      "loss": 0.5815,
      "step": 9260
    },
    {
      "epoch": 27.84,
      "grad_norm": 5.709503650665283,
      "learning_rate": 7.216216216216216e-06,
      "loss": 0.6067,
      "step": 9270
    },
    {
      "epoch": 27.87,
      "grad_norm": 2.9142568111419678,
      "learning_rate": 7.213213213213214e-06,
      "loss": 0.5691,
      "step": 9280
    },
    {
      "epoch": 27.9,
      "grad_norm": 4.102430820465088,
      "learning_rate": 7.210210210210211e-06,
      "loss": 0.5858,
      "step": 9290
    },
    {
      "epoch": 27.93,
      "grad_norm": 5.3256120681762695,
      "learning_rate": 7.207207207207208e-06,
      "loss": 0.5633,
      "step": 9300
    },
    {
      "epoch": 27.96,
      "grad_norm": 5.945285797119141,
      "learning_rate": 7.204204204204205e-06,
      "loss": 0.6175,
      "step": 9310
    },
    {
      "epoch": 27.99,
      "grad_norm": 4.57643461227417,
      "learning_rate": 7.201201201201202e-06,
      "loss": 0.6059,
      "step": 9320
    },
    {
      "epoch": 28.0,
      "eval_accuracy": 0.9187,
      "eval_loss": 0.3552107512950897,
      "eval_runtime": 26.4619,
      "eval_samples_per_second": 377.902,
      "eval_steps_per_second": 1.512,
      "step": 9324
    },
    {
      "epoch": 28.02,
      "grad_norm": 3.537818193435669,
      "learning_rate": 7.1981981981981985e-06,
      "loss": 0.555,
      "step": 9330
    },
    {
      "epoch": 28.05,
      "grad_norm": 5.589657783508301,
      "learning_rate": 7.195195195195196e-06,
      "loss": 0.555,
      "step": 9340
    },
    {
      "epoch": 28.08,
      "grad_norm": 5.387350559234619,
      "learning_rate": 7.1921921921921935e-06,
      "loss": 0.6027,
      "step": 9350
    },
    {
      "epoch": 28.11,
      "grad_norm": 9.464296340942383,
      "learning_rate": 7.189189189189189e-06,
      "loss": 0.6326,
      "step": 9360
    },
    {
      "epoch": 28.14,
      "grad_norm": 5.724453926086426,
      "learning_rate": 7.186186186186187e-06,
      "loss": 0.567,
      "step": 9370
    },
    {
      "epoch": 28.17,
      "grad_norm": 4.872731685638428,
      "learning_rate": 7.183183183183183e-06,
      "loss": 0.5613,
      "step": 9380
    },
    {
      "epoch": 28.2,
      "grad_norm": 8.261558532714844,
      "learning_rate": 7.180180180180181e-06,
      "loss": 0.5409,
      "step": 9390
    },
    {
      "epoch": 28.23,
      "grad_norm": 4.817095756530762,
      "learning_rate": 7.177177177177178e-06,
      "loss": 0.5777,
      "step": 9400
    },
    {
      "epoch": 28.26,
      "grad_norm": 8.762271881103516,
      "learning_rate": 7.174174174174174e-06,
      "loss": 0.5176,
      "step": 9410
    },
    {
      "epoch": 28.29,
      "grad_norm": 4.3167405128479,
      "learning_rate": 7.1711711711711716e-06,
      "loss": 0.5607,
      "step": 9420
    },
    {
      "epoch": 28.32,
      "grad_norm": 4.858583450317383,
      "learning_rate": 7.168168168168169e-06,
      "loss": 0.5385,
      "step": 9430
    },
    {
      "epoch": 28.35,
      "grad_norm": 3.918891429901123,
      "learning_rate": 7.165165165165166e-06,
      "loss": 0.6074,
      "step": 9440
    },
    {
      "epoch": 28.38,
      "grad_norm": 6.432498455047607,
      "learning_rate": 7.162162162162163e-06,
      "loss": 0.5603,
      "step": 9450
    },
    {
      "epoch": 28.41,
      "grad_norm": 3.913588762283325,
      "learning_rate": 7.159159159159161e-06,
      "loss": 0.5224,
      "step": 9460
    },
    {
      "epoch": 28.44,
      "grad_norm": 5.474986553192139,
      "learning_rate": 7.156156156156156e-06,
      "loss": 0.57,
      "step": 9470
    },
    {
      "epoch": 28.47,
      "grad_norm": 3.7697129249572754,
      "learning_rate": 7.153153153153154e-06,
      "loss": 0.5842,
      "step": 9480
    },
    {
      "epoch": 28.5,
      "grad_norm": 3.3923611640930176,
      "learning_rate": 7.1501501501501505e-06,
      "loss": 0.5817,
      "step": 9490
    },
    {
      "epoch": 28.53,
      "grad_norm": 4.007638931274414,
      "learning_rate": 7.147147147147148e-06,
      "loss": 0.5397,
      "step": 9500
    },
    {
      "epoch": 28.56,
      "grad_norm": 5.719387531280518,
      "learning_rate": 7.144144144144145e-06,
      "loss": 0.578,
      "step": 9510
    },
    {
      "epoch": 28.59,
      "grad_norm": 3.7497189044952393,
      "learning_rate": 7.141141141141141e-06,
      "loss": 0.561,
      "step": 9520
    },
    {
      "epoch": 28.62,
      "grad_norm": 6.521295070648193,
      "learning_rate": 7.138138138138139e-06,
      "loss": 0.6263,
      "step": 9530
    },
    {
      "epoch": 28.65,
      "grad_norm": 6.38608980178833,
      "learning_rate": 7.135135135135136e-06,
      "loss": 0.5146,
      "step": 9540
    },
    {
      "epoch": 28.68,
      "grad_norm": 4.536646366119385,
      "learning_rate": 7.132132132132132e-06,
      "loss": 0.5913,
      "step": 9550
    },
    {
      "epoch": 28.71,
      "grad_norm": 5.932296276092529,
      "learning_rate": 7.1291291291291295e-06,
      "loss": 0.6529,
      "step": 9560
    },
    {
      "epoch": 28.74,
      "grad_norm": 4.608974456787109,
      "learning_rate": 7.126126126126127e-06,
      "loss": 0.5663,
      "step": 9570
    },
    {
      "epoch": 28.77,
      "grad_norm": 5.467312335968018,
      "learning_rate": 7.123123123123124e-06,
      "loss": 0.5624,
      "step": 9580
    },
    {
      "epoch": 28.8,
      "grad_norm": 4.99942684173584,
      "learning_rate": 7.120120120120121e-06,
      "loss": 0.6335,
      "step": 9590
    },
    {
      "epoch": 28.83,
      "grad_norm": 5.651036262512207,
      "learning_rate": 7.117117117117117e-06,
      "loss": 0.5798,
      "step": 9600
    },
    {
      "epoch": 28.86,
      "grad_norm": 3.7196743488311768,
      "learning_rate": 7.114114114114114e-06,
      "loss": 0.5707,
      "step": 9610
    },
    {
      "epoch": 28.89,
      "grad_norm": 4.347270965576172,
      "learning_rate": 7.111111111111112e-06,
      "loss": 0.5273,
      "step": 9620
    },
    {
      "epoch": 28.92,
      "grad_norm": 4.918589115142822,
      "learning_rate": 7.1081081081081085e-06,
      "loss": 0.5301,
      "step": 9630
    },
    {
      "epoch": 28.95,
      "grad_norm": 5.1846923828125,
      "learning_rate": 7.105105105105106e-06,
      "loss": 0.6091,
      "step": 9640
    },
    {
      "epoch": 28.98,
      "grad_norm": 5.065564155578613,
      "learning_rate": 7.102102102102103e-06,
      "loss": 0.5227,
      "step": 9650
    },
    {
      "epoch": 29.0,
      "eval_accuracy": 0.9178,
      "eval_loss": 0.345379114151001,
      "eval_runtime": 26.7038,
      "eval_samples_per_second": 374.479,
      "eval_steps_per_second": 1.498,
      "step": 9657
    },
    {
      "epoch": 29.01,
      "grad_norm": 4.072543621063232,
      "learning_rate": 7.099099099099099e-06,
      "loss": 0.581,
      "step": 9660
    },
    {
      "epoch": 29.04,
      "grad_norm": 5.343716621398926,
      "learning_rate": 7.096096096096097e-06,
      "loss": 0.6008,
      "step": 9670
    },
    {
      "epoch": 29.07,
      "grad_norm": 4.13471794128418,
      "learning_rate": 7.093093093093094e-06,
      "loss": 0.5271,
      "step": 9680
    },
    {
      "epoch": 29.1,
      "grad_norm": 4.203214168548584,
      "learning_rate": 7.090090090090091e-06,
      "loss": 0.5814,
      "step": 9690
    },
    {
      "epoch": 29.13,
      "grad_norm": 5.564302444458008,
      "learning_rate": 7.087087087087087e-06,
      "loss": 0.5533,
      "step": 9700
    },
    {
      "epoch": 29.16,
      "grad_norm": 4.4252519607543945,
      "learning_rate": 7.084084084084085e-06,
      "loss": 0.5629,
      "step": 9710
    },
    {
      "epoch": 29.19,
      "grad_norm": 5.020333290100098,
      "learning_rate": 7.0810810810810815e-06,
      "loss": 0.5193,
      "step": 9720
    },
    {
      "epoch": 29.22,
      "grad_norm": 4.683363914489746,
      "learning_rate": 7.078078078078079e-06,
      "loss": 0.578,
      "step": 9730
    },
    {
      "epoch": 29.25,
      "grad_norm": 4.328330993652344,
      "learning_rate": 7.075075075075075e-06,
      "loss": 0.5853,
      "step": 9740
    },
    {
      "epoch": 29.28,
      "grad_norm": 4.674912452697754,
      "learning_rate": 7.072072072072072e-06,
      "loss": 0.5226,
      "step": 9750
    },
    {
      "epoch": 29.31,
      "grad_norm": 5.221179008483887,
      "learning_rate": 7.06906906906907e-06,
      "loss": 0.5026,
      "step": 9760
    },
    {
      "epoch": 29.34,
      "grad_norm": 5.488534450531006,
      "learning_rate": 7.066066066066066e-06,
      "loss": 0.5842,
      "step": 9770
    },
    {
      "epoch": 29.37,
      "grad_norm": 4.482065200805664,
      "learning_rate": 7.063063063063064e-06,
      "loss": 0.5211,
      "step": 9780
    },
    {
      "epoch": 29.4,
      "grad_norm": 4.155788898468018,
      "learning_rate": 7.060060060060061e-06,
      "loss": 0.5769,
      "step": 9790
    },
    {
      "epoch": 29.43,
      "grad_norm": 3.558490037918091,
      "learning_rate": 7.057057057057057e-06,
      "loss": 0.5288,
      "step": 9800
    },
    {
      "epoch": 29.46,
      "grad_norm": 4.873110294342041,
      "learning_rate": 7.054054054054055e-06,
      "loss": 0.5363,
      "step": 9810
    },
    {
      "epoch": 29.49,
      "grad_norm": 4.669723033905029,
      "learning_rate": 7.051051051051052e-06,
      "loss": 0.5283,
      "step": 9820
    },
    {
      "epoch": 29.52,
      "grad_norm": 6.239320278167725,
      "learning_rate": 7.048048048048049e-06,
      "loss": 0.544,
      "step": 9830
    },
    {
      "epoch": 29.55,
      "grad_norm": 5.154757976531982,
      "learning_rate": 7.045045045045046e-06,
      "loss": 0.6143,
      "step": 9840
    },
    {
      "epoch": 29.58,
      "grad_norm": 5.274998188018799,
      "learning_rate": 7.042042042042042e-06,
      "loss": 0.5441,
      "step": 9850
    },
    {
      "epoch": 29.61,
      "grad_norm": 3.258530616760254,
      "learning_rate": 7.0390390390390395e-06,
      "loss": 0.5639,
      "step": 9860
    },
    {
      "epoch": 29.64,
      "grad_norm": 5.662111282348633,
      "learning_rate": 7.036036036036037e-06,
      "loss": 0.5195,
      "step": 9870
    },
    {
      "epoch": 29.67,
      "grad_norm": 4.919576168060303,
      "learning_rate": 7.0330330330330336e-06,
      "loss": 0.5178,
      "step": 9880
    },
    {
      "epoch": 29.7,
      "grad_norm": 6.108078479766846,
      "learning_rate": 7.03003003003003e-06,
      "loss": 0.5967,
      "step": 9890
    },
    {
      "epoch": 29.73,
      "grad_norm": 3.6324610710144043,
      "learning_rate": 7.027027027027028e-06,
      "loss": 0.5153,
      "step": 9900
    },
    {
      "epoch": 29.76,
      "grad_norm": 5.744423866271973,
      "learning_rate": 7.024024024024024e-06,
      "loss": 0.5474,
      "step": 9910
    },
    {
      "epoch": 29.79,
      "grad_norm": 4.71481990814209,
      "learning_rate": 7.021021021021022e-06,
      "loss": 0.5775,
      "step": 9920
    },
    {
      "epoch": 29.82,
      "grad_norm": 6.34856653213501,
      "learning_rate": 7.018018018018019e-06,
      "loss": 0.5263,
      "step": 9930
    },
    {
      "epoch": 29.85,
      "grad_norm": 4.096996784210205,
      "learning_rate": 7.015015015015015e-06,
      "loss": 0.6233,
      "step": 9940
    },
    {
      "epoch": 29.88,
      "grad_norm": 7.632816791534424,
      "learning_rate": 7.0120120120120125e-06,
      "loss": 0.5495,
      "step": 9950
    },
    {
      "epoch": 29.91,
      "grad_norm": 4.823956489562988,
      "learning_rate": 7.009009009009009e-06,
      "loss": 0.5571,
      "step": 9960
    },
    {
      "epoch": 29.94,
      "grad_norm": 3.5478086471557617,
      "learning_rate": 7.006006006006007e-06,
      "loss": 0.6026,
      "step": 9970
    },
    {
      "epoch": 29.97,
      "grad_norm": 5.499380111694336,
      "learning_rate": 7.003003003003004e-06,
      "loss": 0.6102,
      "step": 9980
    },
    {
      "epoch": 30.0,
      "grad_norm": 52.46392822265625,
      "learning_rate": 7e-06,
      "loss": 0.6712,
      "step": 9990
    },
    {
      "epoch": 30.0,
      "eval_accuracy": 0.9183,
      "eval_loss": 0.3375239968299866,
      "eval_runtime": 26.3797,
      "eval_samples_per_second": 379.079,
      "eval_steps_per_second": 1.516,
      "step": 9990
    },
    {
      "epoch": 30.03,
      "grad_norm": 6.403550624847412,
      "learning_rate": 6.996996996996997e-06,
      "loss": 0.5387,
      "step": 10000
    },
    {
      "epoch": 30.06,
      "grad_norm": 6.658787250518799,
      "learning_rate": 6.993993993993995e-06,
      "loss": 0.6126,
      "step": 10010
    },
    {
      "epoch": 30.09,
      "grad_norm": 3.6636009216308594,
      "learning_rate": 6.9909909909909915e-06,
      "loss": 0.5327,
      "step": 10020
    },
    {
      "epoch": 30.12,
      "grad_norm": 6.220738410949707,
      "learning_rate": 6.987987987987989e-06,
      "loss": 0.5723,
      "step": 10030
    },
    {
      "epoch": 30.15,
      "grad_norm": 5.872793674468994,
      "learning_rate": 6.984984984984986e-06,
      "loss": 0.5094,
      "step": 10040
    },
    {
      "epoch": 30.18,
      "grad_norm": 5.606639385223389,
      "learning_rate": 6.981981981981982e-06,
      "loss": 0.5274,
      "step": 10050
    },
    {
      "epoch": 30.21,
      "grad_norm": 4.649966716766357,
      "learning_rate": 6.97897897897898e-06,
      "loss": 0.5077,
      "step": 10060
    },
    {
      "epoch": 30.24,
      "grad_norm": 3.147263765335083,
      "learning_rate": 6.975975975975977e-06,
      "loss": 0.5242,
      "step": 10070
    },
    {
      "epoch": 30.27,
      "grad_norm": 4.2813310623168945,
      "learning_rate": 6.972972972972973e-06,
      "loss": 0.6066,
      "step": 10080
    },
    {
      "epoch": 30.3,
      "grad_norm": 4.643487930297852,
      "learning_rate": 6.9699699699699704e-06,
      "loss": 0.5249,
      "step": 10090
    },
    {
      "epoch": 30.33,
      "grad_norm": 4.673160552978516,
      "learning_rate": 6.966966966966967e-06,
      "loss": 0.5169,
      "step": 10100
    },
    {
      "epoch": 30.36,
      "grad_norm": 5.770195007324219,
      "learning_rate": 6.9639639639639646e-06,
      "loss": 0.5177,
      "step": 10110
    },
    {
      "epoch": 30.39,
      "grad_norm": 6.050931930541992,
      "learning_rate": 6.960960960960962e-06,
      "loss": 0.5477,
      "step": 10120
    },
    {
      "epoch": 30.42,
      "grad_norm": 4.330102443695068,
      "learning_rate": 6.957957957957958e-06,
      "loss": 0.533,
      "step": 10130
    },
    {
      "epoch": 30.45,
      "grad_norm": 5.163640022277832,
      "learning_rate": 6.954954954954955e-06,
      "loss": 0.531,
      "step": 10140
    },
    {
      "epoch": 30.48,
      "grad_norm": 6.337249755859375,
      "learning_rate": 6.951951951951953e-06,
      "loss": 0.5424,
      "step": 10150
    },
    {
      "epoch": 30.51,
      "grad_norm": 4.544989585876465,
      "learning_rate": 6.948948948948949e-06,
      "loss": 0.5534,
      "step": 10160
    },
    {
      "epoch": 30.54,
      "grad_norm": 3.9418797492980957,
      "learning_rate": 6.945945945945947e-06,
      "loss": 0.5304,
      "step": 10170
    },
    {
      "epoch": 30.57,
      "grad_norm": 3.7905588150024414,
      "learning_rate": 6.942942942942944e-06,
      "loss": 0.4708,
      "step": 10180
    },
    {
      "epoch": 30.6,
      "grad_norm": 4.000282287597656,
      "learning_rate": 6.93993993993994e-06,
      "loss": 0.4982,
      "step": 10190
    },
    {
      "epoch": 30.63,
      "grad_norm": 4.103896141052246,
      "learning_rate": 6.936936936936938e-06,
      "loss": 0.5633,
      "step": 10200
    },
    {
      "epoch": 30.66,
      "grad_norm": 3.971468925476074,
      "learning_rate": 6.933933933933934e-06,
      "loss": 0.4992,
      "step": 10210
    },
    {
      "epoch": 30.69,
      "grad_norm": 4.747451305389404,
      "learning_rate": 6.930930930930932e-06,
      "loss": 0.548,
      "step": 10220
    },
    {
      "epoch": 30.72,
      "grad_norm": 2.6908531188964844,
      "learning_rate": 6.927927927927928e-06,
      "loss": 0.5545,
      "step": 10230
    },
    {
      "epoch": 30.75,
      "grad_norm": 4.251894474029541,
      "learning_rate": 6.924924924924925e-06,
      "loss": 0.4754,
      "step": 10240
    },
    {
      "epoch": 30.78,
      "grad_norm": 4.229438781738281,
      "learning_rate": 6.9219219219219225e-06,
      "loss": 0.5479,
      "step": 10250
    },
    {
      "epoch": 30.81,
      "grad_norm": 3.511378288269043,
      "learning_rate": 6.91891891891892e-06,
      "loss": 0.576,
      "step": 10260
    },
    {
      "epoch": 30.84,
      "grad_norm": 4.298564910888672,
      "learning_rate": 6.915915915915916e-06,
      "loss": 0.5447,
      "step": 10270
    },
    {
      "epoch": 30.87,
      "grad_norm": 3.6350417137145996,
      "learning_rate": 6.912912912912913e-06,
      "loss": 0.5073,
      "step": 10280
    },
    {
      "epoch": 30.9,
      "grad_norm": 4.037917613983154,
      "learning_rate": 6.909909909909911e-06,
      "loss": 0.5399,
      "step": 10290
    },
    {
      "epoch": 30.93,
      "grad_norm": 4.715947151184082,
      "learning_rate": 6.906906906906907e-06,
      "loss": 0.5516,
      "step": 10300
    },
    {
      "epoch": 30.96,
      "grad_norm": 7.880973815917969,
      "learning_rate": 6.903903903903905e-06,
      "loss": 0.5438,
      "step": 10310
    },
    {
      "epoch": 30.99,
      "grad_norm": 4.695303440093994,
      "learning_rate": 6.900900900900901e-06,
      "loss": 0.5053,
      "step": 10320
    },
    {
      "epoch": 31.0,
      "eval_accuracy": 0.9171,
      "eval_loss": 0.33546704053878784,
      "eval_runtime": 26.5691,
      "eval_samples_per_second": 376.376,
      "eval_steps_per_second": 1.506,
      "step": 10323
    },
    {
      "epoch": 31.02,
      "grad_norm": 4.5637712478637695,
      "learning_rate": 6.897897897897898e-06,
      "loss": 0.5222,
      "step": 10330
    },
    {
      "epoch": 31.05,
      "grad_norm": 4.128244400024414,
      "learning_rate": 6.8948948948948955e-06,
      "loss": 0.5947,
      "step": 10340
    },
    {
      "epoch": 31.08,
      "grad_norm": 4.192866802215576,
      "learning_rate": 6.891891891891892e-06,
      "loss": 0.5615,
      "step": 10350
    },
    {
      "epoch": 31.11,
      "grad_norm": 3.5042011737823486,
      "learning_rate": 6.88888888888889e-06,
      "loss": 0.4907,
      "step": 10360
    },
    {
      "epoch": 31.14,
      "grad_norm": 5.038252830505371,
      "learning_rate": 6.885885885885887e-06,
      "loss": 0.4821,
      "step": 10370
    },
    {
      "epoch": 31.17,
      "grad_norm": 4.238375663757324,
      "learning_rate": 6.882882882882883e-06,
      "loss": 0.4759,
      "step": 10380
    },
    {
      "epoch": 31.2,
      "grad_norm": 3.3415908813476562,
      "learning_rate": 6.87987987987988e-06,
      "loss": 0.5448,
      "step": 10390
    },
    {
      "epoch": 31.23,
      "grad_norm": 3.935137987136841,
      "learning_rate": 6.876876876876878e-06,
      "loss": 0.5556,
      "step": 10400
    },
    {
      "epoch": 31.26,
      "grad_norm": 5.417572021484375,
      "learning_rate": 6.8738738738738745e-06,
      "loss": 0.5054,
      "step": 10410
    },
    {
      "epoch": 31.29,
      "grad_norm": 3.5927047729492188,
      "learning_rate": 6.870870870870872e-06,
      "loss": 0.4932,
      "step": 10420
    },
    {
      "epoch": 31.32,
      "grad_norm": 4.9424147605896,
      "learning_rate": 6.867867867867869e-06,
      "loss": 0.5413,
      "step": 10430
    },
    {
      "epoch": 31.35,
      "grad_norm": 4.6282572746276855,
      "learning_rate": 6.864864864864865e-06,
      "loss": 0.5611,
      "step": 10440
    },
    {
      "epoch": 31.38,
      "grad_norm": 5.602158546447754,
      "learning_rate": 6.861861861861863e-06,
      "loss": 0.5624,
      "step": 10450
    },
    {
      "epoch": 31.41,
      "grad_norm": 4.025387763977051,
      "learning_rate": 6.8588588588588585e-06,
      "loss": 0.5034,
      "step": 10460
    },
    {
      "epoch": 31.44,
      "grad_norm": 4.116804599761963,
      "learning_rate": 6.855855855855856e-06,
      "loss": 0.51,
      "step": 10470
    },
    {
      "epoch": 31.47,
      "grad_norm": 5.079034805297852,
      "learning_rate": 6.8528528528528535e-06,
      "loss": 0.5951,
      "step": 10480
    },
    {
      "epoch": 31.5,
      "grad_norm": 5.287024021148682,
      "learning_rate": 6.84984984984985e-06,
      "loss": 0.5248,
      "step": 10490
    },
    {
      "epoch": 31.53,
      "grad_norm": 4.518507957458496,
      "learning_rate": 6.846846846846848e-06,
      "loss": 0.5718,
      "step": 10500
    },
    {
      "epoch": 31.56,
      "grad_norm": 4.743940830230713,
      "learning_rate": 6.843843843843845e-06,
      "loss": 0.5249,
      "step": 10510
    },
    {
      "epoch": 31.59,
      "grad_norm": 3.6553006172180176,
      "learning_rate": 6.840840840840841e-06,
      "loss": 0.4958,
      "step": 10520
    },
    {
      "epoch": 31.62,
      "grad_norm": 5.176079750061035,
      "learning_rate": 6.837837837837838e-06,
      "loss": 0.5568,
      "step": 10530
    },
    {
      "epoch": 31.65,
      "grad_norm": 7.153314590454102,
      "learning_rate": 6.834834834834836e-06,
      "loss": 0.4945,
      "step": 10540
    },
    {
      "epoch": 31.68,
      "grad_norm": 4.934965133666992,
      "learning_rate": 6.8318318318318324e-06,
      "loss": 0.5023,
      "step": 10550
    },
    {
      "epoch": 31.71,
      "grad_norm": 5.286891937255859,
      "learning_rate": 6.82882882882883e-06,
      "loss": 0.5313,
      "step": 10560
    },
    {
      "epoch": 31.74,
      "grad_norm": 6.007424831390381,
      "learning_rate": 6.825825825825826e-06,
      "loss": 0.5512,
      "step": 10570
    },
    {
      "epoch": 31.77,
      "grad_norm": 5.364544868469238,
      "learning_rate": 6.822822822822823e-06,
      "loss": 0.5405,
      "step": 10580
    },
    {
      "epoch": 31.8,
      "grad_norm": 4.781249046325684,
      "learning_rate": 6.819819819819821e-06,
      "loss": 0.5878,
      "step": 10590
    },
    {
      "epoch": 31.83,
      "grad_norm": 5.178708553314209,
      "learning_rate": 6.816816816816817e-06,
      "loss": 0.4952,
      "step": 10600
    },
    {
      "epoch": 31.86,
      "grad_norm": 4.609602928161621,
      "learning_rate": 6.813813813813815e-06,
      "loss": 0.5262,
      "step": 10610
    },
    {
      "epoch": 31.89,
      "grad_norm": 6.407437801361084,
      "learning_rate": 6.810810810810811e-06,
      "loss": 0.4991,
      "step": 10620
    },
    {
      "epoch": 31.92,
      "grad_norm": 3.346106767654419,
      "learning_rate": 6.807807807807808e-06,
      "loss": 0.5157,
      "step": 10630
    },
    {
      "epoch": 31.95,
      "grad_norm": 4.991146087646484,
      "learning_rate": 6.8048048048048055e-06,
      "loss": 0.4841,
      "step": 10640
    },
    {
      "epoch": 31.98,
      "grad_norm": 4.926365375518799,
      "learning_rate": 6.801801801801803e-06,
      "loss": 0.5432,
      "step": 10650
    },
    {
      "epoch": 32.0,
      "eval_accuracy": 0.917,
      "eval_loss": 0.3327946662902832,
      "eval_runtime": 26.3244,
      "eval_samples_per_second": 379.875,
      "eval_steps_per_second": 1.52,
      "step": 10656
    },
    {
      "epoch": 32.01,
      "grad_norm": 5.152103900909424,
      "learning_rate": 6.798798798798799e-06,
      "loss": 0.466,
      "step": 10660
    },
    {
      "epoch": 32.04,
      "grad_norm": 4.942965984344482,
      "learning_rate": 6.795795795795796e-06,
      "loss": 0.5277,
      "step": 10670
    },
    {
      "epoch": 32.07,
      "grad_norm": 3.914383888244629,
      "learning_rate": 6.792792792792793e-06,
      "loss": 0.5241,
      "step": 10680
    },
    {
      "epoch": 32.1,
      "grad_norm": 4.517918109893799,
      "learning_rate": 6.78978978978979e-06,
      "loss": 0.5013,
      "step": 10690
    },
    {
      "epoch": 32.13,
      "grad_norm": 4.828157901763916,
      "learning_rate": 6.786786786786788e-06,
      "loss": 0.5212,
      "step": 10700
    },
    {
      "epoch": 32.16,
      "grad_norm": 5.400358200073242,
      "learning_rate": 6.783783783783784e-06,
      "loss": 0.5706,
      "step": 10710
    },
    {
      "epoch": 32.19,
      "grad_norm": 5.0671067237854,
      "learning_rate": 6.780780780780781e-06,
      "loss": 0.5967,
      "step": 10720
    },
    {
      "epoch": 32.22,
      "grad_norm": 4.591043949127197,
      "learning_rate": 6.777777777777779e-06,
      "loss": 0.5163,
      "step": 10730
    },
    {
      "epoch": 32.25,
      "grad_norm": 5.528631210327148,
      "learning_rate": 6.774774774774775e-06,
      "loss": 0.5845,
      "step": 10740
    },
    {
      "epoch": 32.28,
      "grad_norm": 5.988685607910156,
      "learning_rate": 6.771771771771773e-06,
      "loss": 0.532,
      "step": 10750
    },
    {
      "epoch": 32.31,
      "grad_norm": 3.5375077724456787,
      "learning_rate": 6.76876876876877e-06,
      "loss": 0.4809,
      "step": 10760
    },
    {
      "epoch": 32.34,
      "grad_norm": 4.084622383117676,
      "learning_rate": 6.765765765765766e-06,
      "loss": 0.469,
      "step": 10770
    },
    {
      "epoch": 32.37,
      "grad_norm": 5.619451999664307,
      "learning_rate": 6.7627627627627634e-06,
      "loss": 0.5272,
      "step": 10780
    },
    {
      "epoch": 32.4,
      "grad_norm": 5.900362968444824,
      "learning_rate": 6.759759759759761e-06,
      "loss": 0.508,
      "step": 10790
    },
    {
      "epoch": 32.43,
      "grad_norm": 3.274456739425659,
      "learning_rate": 6.7567567567567575e-06,
      "loss": 0.5378,
      "step": 10800
    },
    {
      "epoch": 32.46,
      "grad_norm": 5.264253616333008,
      "learning_rate": 6.753753753753754e-06,
      "loss": 0.5458,
      "step": 10810
    },
    {
      "epoch": 32.49,
      "grad_norm": 5.250488758087158,
      "learning_rate": 6.750750750750751e-06,
      "loss": 0.4558,
      "step": 10820
    },
    {
      "epoch": 32.52,
      "grad_norm": 5.334564208984375,
      "learning_rate": 6.747747747747748e-06,
      "loss": 0.4916,
      "step": 10830
    },
    {
      "epoch": 32.55,
      "grad_norm": 4.669946193695068,
      "learning_rate": 6.744744744744746e-06,
      "loss": 0.5769,
      "step": 10840
    },
    {
      "epoch": 32.58,
      "grad_norm": 5.030605792999268,
      "learning_rate": 6.7417417417417415e-06,
      "loss": 0.5051,
      "step": 10850
    },
    {
      "epoch": 32.61,
      "grad_norm": 4.161187648773193,
      "learning_rate": 6.738738738738739e-06,
      "loss": 0.4946,
      "step": 10860
    },
    {
      "epoch": 32.64,
      "grad_norm": 4.491575241088867,
      "learning_rate": 6.7357357357357365e-06,
      "loss": 0.4936,
      "step": 10870
    },
    {
      "epoch": 32.67,
      "grad_norm": 4.474544048309326,
      "learning_rate": 6.732732732732733e-06,
      "loss": 0.4773,
      "step": 10880
    },
    {
      "epoch": 32.7,
      "grad_norm": 4.078099727630615,
      "learning_rate": 6.729729729729731e-06,
      "loss": 0.5129,
      "step": 10890
    },
    {
      "epoch": 32.73,
      "grad_norm": 4.027213096618652,
      "learning_rate": 6.726726726726728e-06,
      "loss": 0.5077,
      "step": 10900
    },
    {
      "epoch": 32.76,
      "grad_norm": 4.637345790863037,
      "learning_rate": 6.723723723723724e-06,
      "loss": 0.4782,
      "step": 10910
    },
    {
      "epoch": 32.79,
      "grad_norm": 4.181465148925781,
      "learning_rate": 6.720720720720721e-06,
      "loss": 0.503,
      "step": 10920
    },
    {
      "epoch": 32.82,
      "grad_norm": 3.633840560913086,
      "learning_rate": 6.717717717717718e-06,
      "loss": 0.4897,
      "step": 10930
    },
    {
      "epoch": 32.85,
      "grad_norm": 4.9459147453308105,
      "learning_rate": 6.7147147147147155e-06,
      "loss": 0.4924,
      "step": 10940
    },
    {
      "epoch": 32.88,
      "grad_norm": 5.306789398193359,
      "learning_rate": 6.711711711711713e-06,
      "loss": 0.4854,
      "step": 10950
    },
    {
      "epoch": 32.91,
      "grad_norm": 3.652146339416504,
      "learning_rate": 6.708708708708709e-06,
      "loss": 0.52,
      "step": 10960
    },
    {
      "epoch": 32.94,
      "grad_norm": 4.799497604370117,
      "learning_rate": 6.705705705705706e-06,
      "loss": 0.5273,
      "step": 10970
    },
    {
      "epoch": 32.97,
      "grad_norm": 4.157444477081299,
      "learning_rate": 6.702702702702704e-06,
      "loss": 0.4617,
      "step": 10980
    },
    {
      "epoch": 33.0,
      "eval_accuracy": 0.9191,
      "eval_loss": 0.32952603697776794,
      "eval_runtime": 26.412,
      "eval_samples_per_second": 378.616,
      "eval_steps_per_second": 1.514,
      "step": 10989
    },
    {
      "epoch": 33.0,
      "grad_norm": 4.564639568328857,
      "learning_rate": 6.6996996996997e-06,
      "loss": 0.432,
      "step": 10990
    },
    {
      "epoch": 33.03,
      "grad_norm": 6.127586841583252,
      "learning_rate": 6.696696696696697e-06,
      "loss": 0.5657,
      "step": 11000
    },
    {
      "epoch": 33.06,
      "grad_norm": 4.598598003387451,
      "learning_rate": 6.693693693693694e-06,
      "loss": 0.5202,
      "step": 11010
    },
    {
      "epoch": 33.09,
      "grad_norm": 4.336296081542969,
      "learning_rate": 6.690690690690691e-06,
      "loss": 0.5148,
      "step": 11020
    },
    {
      "epoch": 33.12,
      "grad_norm": 5.0793304443359375,
      "learning_rate": 6.6876876876876885e-06,
      "loss": 0.5735,
      "step": 11030
    },
    {
      "epoch": 33.15,
      "grad_norm": 5.527975082397461,
      "learning_rate": 6.684684684684684e-06,
      "loss": 0.4741,
      "step": 11040
    },
    {
      "epoch": 33.18,
      "grad_norm": 4.1530375480651855,
      "learning_rate": 6.681681681681682e-06,
      "loss": 0.4882,
      "step": 11050
    },
    {
      "epoch": 33.21,
      "grad_norm": 6.072080612182617,
      "learning_rate": 6.678678678678679e-06,
      "loss": 0.5918,
      "step": 11060
    },
    {
      "epoch": 33.24,
      "grad_norm": 4.9077863693237305,
      "learning_rate": 6.675675675675676e-06,
      "loss": 0.5253,
      "step": 11070
    },
    {
      "epoch": 33.27,
      "grad_norm": 3.628720998764038,
      "learning_rate": 6.672672672672673e-06,
      "loss": 0.4282,
      "step": 11080
    },
    {
      "epoch": 33.3,
      "grad_norm": 6.934624671936035,
      "learning_rate": 6.669669669669671e-06,
      "loss": 0.5032,
      "step": 11090
    },
    {
      "epoch": 33.33,
      "grad_norm": 5.978390693664551,
      "learning_rate": 6.666666666666667e-06,
      "loss": 0.5319,
      "step": 11100
    },
    {
      "epoch": 33.36,
      "grad_norm": 7.497355937957764,
      "learning_rate": 6.663663663663664e-06,
      "loss": 0.5025,
      "step": 11110
    },
    {
      "epoch": 33.39,
      "grad_norm": 4.441858291625977,
      "learning_rate": 6.660660660660662e-06,
      "loss": 0.4608,
      "step": 11120
    },
    {
      "epoch": 33.42,
      "grad_norm": 4.574911117553711,
      "learning_rate": 6.657657657657658e-06,
      "loss": 0.5327,
      "step": 11130
    },
    {
      "epoch": 33.45,
      "grad_norm": 4.213235855102539,
      "learning_rate": 6.654654654654656e-06,
      "loss": 0.4994,
      "step": 11140
    },
    {
      "epoch": 33.48,
      "grad_norm": 4.079199314117432,
      "learning_rate": 6.651651651651652e-06,
      "loss": 0.4945,
      "step": 11150
    },
    {
      "epoch": 33.51,
      "grad_norm": 4.505585193634033,
      "learning_rate": 6.648648648648649e-06,
      "loss": 0.5462,
      "step": 11160
    },
    {
      "epoch": 33.54,
      "grad_norm": 7.309465408325195,
      "learning_rate": 6.6456456456456465e-06,
      "loss": 0.4975,
      "step": 11170
    },
    {
      "epoch": 33.57,
      "grad_norm": 5.451502323150635,
      "learning_rate": 6.642642642642643e-06,
      "loss": 0.5328,
      "step": 11180
    },
    {
      "epoch": 33.6,
      "grad_norm": 5.075081825256348,
      "learning_rate": 6.63963963963964e-06,
      "loss": 0.4684,
      "step": 11190
    },
    {
      "epoch": 33.63,
      "grad_norm": 15.912128448486328,
      "learning_rate": 6.636636636636637e-06,
      "loss": 0.4503,
      "step": 11200
    },
    {
      "epoch": 33.66,
      "grad_norm": 4.8083295822143555,
      "learning_rate": 6.633633633633634e-06,
      "loss": 0.5233,
      "step": 11210
    },
    {
      "epoch": 33.69,
      "grad_norm": 6.54976224899292,
      "learning_rate": 6.630630630630631e-06,
      "loss": 0.564,
      "step": 11220
    },
    {
      "epoch": 33.72,
      "grad_norm": 5.533057689666748,
      "learning_rate": 6.627627627627629e-06,
      "loss": 0.5073,
      "step": 11230
    },
    {
      "epoch": 33.75,
      "grad_norm": 5.056384563446045,
      "learning_rate": 6.6246246246246246e-06,
      "loss": 0.5933,
      "step": 11240
    },
    {
      "epoch": 33.78,
      "grad_norm": 4.461758136749268,
      "learning_rate": 6.621621621621622e-06,
      "loss": 0.5609,
      "step": 11250
    },
    {
      "epoch": 33.81,
      "grad_norm": 4.772905349731445,
      "learning_rate": 6.6186186186186195e-06,
      "loss": 0.5211,
      "step": 11260
    },
    {
      "epoch": 33.84,
      "grad_norm": 4.28291654586792,
      "learning_rate": 6.615615615615616e-06,
      "loss": 0.5741,
      "step": 11270
    },
    {
      "epoch": 33.87,
      "grad_norm": 4.813282489776611,
      "learning_rate": 6.612612612612614e-06,
      "loss": 0.5078,
      "step": 11280
    },
    {
      "epoch": 33.9,
      "grad_norm": 6.550593852996826,
      "learning_rate": 6.609609609609609e-06,
      "loss": 0.5039,
      "step": 11290
    },
    {
      "epoch": 33.93,
      "grad_norm": 3.385647773742676,
      "learning_rate": 6.606606606606607e-06,
      "loss": 0.498,
      "step": 11300
    },
    {
      "epoch": 33.96,
      "grad_norm": 6.887855529785156,
      "learning_rate": 6.603603603603604e-06,
      "loss": 0.5166,
      "step": 11310
    },
    {
      "epoch": 33.99,
      "grad_norm": 4.921820163726807,
      "learning_rate": 6.600600600600601e-06,
      "loss": 0.4784,
      "step": 11320
    },
    {
      "epoch": 34.0,
      "eval_accuracy": 0.918,
      "eval_loss": 0.3249960243701935,
      "eval_runtime": 27.0964,
      "eval_samples_per_second": 369.052,
      "eval_steps_per_second": 1.476,
      "step": 11322
    },
    {
      "epoch": 34.02,
      "grad_norm": 8.019493103027344,
      "learning_rate": 6.5975975975975985e-06,
      "loss": 0.5349,
      "step": 11330
    },
    {
      "epoch": 34.05,
      "grad_norm": 5.084954261779785,
      "learning_rate": 6.594594594594595e-06,
      "loss": 0.4918,
      "step": 11340
    },
    {
      "epoch": 34.08,
      "grad_norm": 4.809759140014648,
      "learning_rate": 6.591591591591592e-06,
      "loss": 0.5505,
      "step": 11350
    },
    {
      "epoch": 34.11,
      "grad_norm": 4.049439430236816,
      "learning_rate": 6.588588588588589e-06,
      "loss": 0.4728,
      "step": 11360
    },
    {
      "epoch": 34.14,
      "grad_norm": 4.804196834564209,
      "learning_rate": 6.585585585585587e-06,
      "loss": 0.4778,
      "step": 11370
    },
    {
      "epoch": 34.17,
      "grad_norm": 4.499908924102783,
      "learning_rate": 6.5825825825825825e-06,
      "loss": 0.4996,
      "step": 11380
    },
    {
      "epoch": 34.2,
      "grad_norm": 5.686038970947266,
      "learning_rate": 6.57957957957958e-06,
      "loss": 0.5266,
      "step": 11390
    },
    {
      "epoch": 34.23,
      "grad_norm": 6.619571208953857,
      "learning_rate": 6.5765765765765775e-06,
      "loss": 0.537,
      "step": 11400
    },
    {
      "epoch": 34.26,
      "grad_norm": 4.216808319091797,
      "learning_rate": 6.573573573573574e-06,
      "loss": 0.4832,
      "step": 11410
    },
    {
      "epoch": 34.29,
      "grad_norm": 5.596823692321777,
      "learning_rate": 6.5705705705705716e-06,
      "loss": 0.5208,
      "step": 11420
    },
    {
      "epoch": 34.32,
      "grad_norm": 3.564544677734375,
      "learning_rate": 6.567567567567567e-06,
      "loss": 0.4957,
      "step": 11430
    },
    {
      "epoch": 34.35,
      "grad_norm": 2.9732589721679688,
      "learning_rate": 6.564564564564565e-06,
      "loss": 0.4692,
      "step": 11440
    },
    {
      "epoch": 34.38,
      "grad_norm": 4.947221755981445,
      "learning_rate": 6.561561561561562e-06,
      "loss": 0.5326,
      "step": 11450
    },
    {
      "epoch": 34.41,
      "grad_norm": 4.498212814331055,
      "learning_rate": 6.558558558558559e-06,
      "loss": 0.4989,
      "step": 11460
    },
    {
      "epoch": 34.44,
      "grad_norm": 6.685206413269043,
      "learning_rate": 6.555555555555556e-06,
      "loss": 0.5345,
      "step": 11470
    },
    {
      "epoch": 34.47,
      "grad_norm": 5.036656379699707,
      "learning_rate": 6.552552552552554e-06,
      "loss": 0.5189,
      "step": 11480
    },
    {
      "epoch": 34.5,
      "grad_norm": 5.761340141296387,
      "learning_rate": 6.54954954954955e-06,
      "loss": 0.4814,
      "step": 11490
    },
    {
      "epoch": 34.53,
      "grad_norm": 8.407565116882324,
      "learning_rate": 6.546546546546547e-06,
      "loss": 0.4944,
      "step": 11500
    },
    {
      "epoch": 34.56,
      "grad_norm": 4.169130802154541,
      "learning_rate": 6.543543543543545e-06,
      "loss": 0.5854,
      "step": 11510
    },
    {
      "epoch": 34.59,
      "grad_norm": 5.203417778015137,
      "learning_rate": 6.540540540540541e-06,
      "loss": 0.5219,
      "step": 11520
    },
    {
      "epoch": 34.62,
      "grad_norm": 6.78031063079834,
      "learning_rate": 6.537537537537538e-06,
      "loss": 0.5629,
      "step": 11530
    },
    {
      "epoch": 34.65,
      "grad_norm": 5.361577033996582,
      "learning_rate": 6.5345345345345345e-06,
      "loss": 0.4669,
      "step": 11540
    },
    {
      "epoch": 34.68,
      "grad_norm": 3.6456704139709473,
      "learning_rate": 6.531531531531532e-06,
      "loss": 0.5041,
      "step": 11550
    },
    {
      "epoch": 34.71,
      "grad_norm": 4.542550086975098,
      "learning_rate": 6.5285285285285295e-06,
      "loss": 0.525,
      "step": 11560
    },
    {
      "epoch": 34.74,
      "grad_norm": 5.005729675292969,
      "learning_rate": 6.525525525525525e-06,
      "loss": 0.4992,
      "step": 11570
    },
    {
      "epoch": 34.77,
      "grad_norm": 4.900482654571533,
      "learning_rate": 6.522522522522523e-06,
      "loss": 0.5136,
      "step": 11580
    },
    {
      "epoch": 34.8,
      "grad_norm": 5.906069755554199,
      "learning_rate": 6.51951951951952e-06,
      "loss": 0.5294,
      "step": 11590
    },
    {
      "epoch": 34.83,
      "grad_norm": 3.6246538162231445,
      "learning_rate": 6.516516516516517e-06,
      "loss": 0.4772,
      "step": 11600
    },
    {
      "epoch": 34.86,
      "grad_norm": 4.865238666534424,
      "learning_rate": 6.513513513513514e-06,
      "loss": 0.5077,
      "step": 11610
    },
    {
      "epoch": 34.89,
      "grad_norm": 4.1368021965026855,
      "learning_rate": 6.510510510510512e-06,
      "loss": 0.5087,
      "step": 11620
    },
    {
      "epoch": 34.92,
      "grad_norm": 4.264941215515137,
      "learning_rate": 6.507507507507508e-06,
      "loss": 0.5117,
      "step": 11630
    },
    {
      "epoch": 34.95,
      "grad_norm": 3.78835129737854,
      "learning_rate": 6.504504504504505e-06,
      "loss": 0.4964,
      "step": 11640
    },
    {
      "epoch": 34.98,
      "grad_norm": 5.336776256561279,
      "learning_rate": 6.501501501501502e-06,
      "loss": 0.5088,
      "step": 11650
    },
    {
      "epoch": 35.0,
      "eval_accuracy": 0.9195,
      "eval_loss": 0.31883320212364197,
      "eval_runtime": 26.4852,
      "eval_samples_per_second": 377.57,
      "eval_steps_per_second": 1.51,
      "step": 11655
    },
    {
      "epoch": 35.02,
      "grad_norm": 5.168522834777832,
      "learning_rate": 6.498498498498499e-06,
      "loss": 0.493,
      "step": 11660
    },
    {
      "epoch": 35.05,
      "grad_norm": 5.572307586669922,
      "learning_rate": 6.495495495495497e-06,
      "loss": 0.4604,
      "step": 11670
    },
    {
      "epoch": 35.08,
      "grad_norm": 4.209198474884033,
      "learning_rate": 6.4924924924924924e-06,
      "loss": 0.5438,
      "step": 11680
    },
    {
      "epoch": 35.11,
      "grad_norm": 4.411211013793945,
      "learning_rate": 6.48948948948949e-06,
      "loss": 0.4837,
      "step": 11690
    },
    {
      "epoch": 35.14,
      "grad_norm": 4.3489789962768555,
      "learning_rate": 6.486486486486487e-06,
      "loss": 0.5236,
      "step": 11700
    },
    {
      "epoch": 35.17,
      "grad_norm": 4.487356662750244,
      "learning_rate": 6.483483483483484e-06,
      "loss": 0.5342,
      "step": 11710
    },
    {
      "epoch": 35.2,
      "grad_norm": 4.561159133911133,
      "learning_rate": 6.480480480480481e-06,
      "loss": 0.4909,
      "step": 11720
    },
    {
      "epoch": 35.23,
      "grad_norm": 4.875969886779785,
      "learning_rate": 6.477477477477478e-06,
      "loss": 0.5449,
      "step": 11730
    },
    {
      "epoch": 35.26,
      "grad_norm": 4.210017204284668,
      "learning_rate": 6.474474474474475e-06,
      "loss": 0.5407,
      "step": 11740
    },
    {
      "epoch": 35.29,
      "grad_norm": 4.259859561920166,
      "learning_rate": 6.471471471471472e-06,
      "loss": 0.4857,
      "step": 11750
    },
    {
      "epoch": 35.32,
      "grad_norm": 4.824864864349365,
      "learning_rate": 6.46846846846847e-06,
      "loss": 0.5466,
      "step": 11760
    },
    {
      "epoch": 35.35,
      "grad_norm": 5.433022499084473,
      "learning_rate": 6.4654654654654655e-06,
      "loss": 0.4856,
      "step": 11770
    },
    {
      "epoch": 35.38,
      "grad_norm": 4.4445343017578125,
      "learning_rate": 6.462462462462463e-06,
      "loss": 0.5254,
      "step": 11780
    },
    {
      "epoch": 35.41,
      "grad_norm": 5.057773590087891,
      "learning_rate": 6.45945945945946e-06,
      "loss": 0.5527,
      "step": 11790
    },
    {
      "epoch": 35.44,
      "grad_norm": 3.377410650253296,
      "learning_rate": 6.456456456456457e-06,
      "loss": 0.4687,
      "step": 11800
    },
    {
      "epoch": 35.47,
      "grad_norm": 6.953456878662109,
      "learning_rate": 6.453453453453455e-06,
      "loss": 0.5279,
      "step": 11810
    },
    {
      "epoch": 35.5,
      "grad_norm": 4.230295181274414,
      "learning_rate": 6.45045045045045e-06,
      "loss": 0.5659,
      "step": 11820
    },
    {
      "epoch": 35.53,
      "grad_norm": 4.3861918449401855,
      "learning_rate": 6.447447447447448e-06,
      "loss": 0.5215,
      "step": 11830
    },
    {
      "epoch": 35.56,
      "grad_norm": 5.295562267303467,
      "learning_rate": 6.444444444444445e-06,
      "loss": 0.5083,
      "step": 11840
    },
    {
      "epoch": 35.59,
      "grad_norm": 5.285841941833496,
      "learning_rate": 6.441441441441442e-06,
      "loss": 0.5354,
      "step": 11850
    },
    {
      "epoch": 35.62,
      "grad_norm": 3.731440305709839,
      "learning_rate": 6.4384384384384394e-06,
      "loss": 0.4666,
      "step": 11860
    },
    {
      "epoch": 35.65,
      "grad_norm": 4.5777082443237305,
      "learning_rate": 6.435435435435436e-06,
      "loss": 0.537,
      "step": 11870
    },
    {
      "epoch": 35.68,
      "grad_norm": 5.057626724243164,
      "learning_rate": 6.432432432432433e-06,
      "loss": 0.5012,
      "step": 11880
    },
    {
      "epoch": 35.71,
      "grad_norm": 3.5204951763153076,
      "learning_rate": 6.42942942942943e-06,
      "loss": 0.5544,
      "step": 11890
    },
    {
      "epoch": 35.74,
      "grad_norm": 4.735846996307373,
      "learning_rate": 6.426426426426427e-06,
      "loss": 0.5025,
      "step": 11900
    },
    {
      "epoch": 35.77,
      "grad_norm": 4.379212379455566,
      "learning_rate": 6.4234234234234234e-06,
      "loss": 0.4677,
      "step": 11910
    },
    {
      "epoch": 35.8,
      "grad_norm": 4.914200782775879,
      "learning_rate": 6.420420420420421e-06,
      "loss": 0.5281,
      "step": 11920
    },
    {
      "epoch": 35.83,
      "grad_norm": 5.16389274597168,
      "learning_rate": 6.4174174174174176e-06,
      "loss": 0.467,
      "step": 11930
    },
    {
      "epoch": 35.86,
      "grad_norm": 4.597350597381592,
      "learning_rate": 6.414414414414415e-06,
      "loss": 0.5338,
      "step": 11940
    },
    {
      "epoch": 35.89,
      "grad_norm": 4.771205902099609,
      "learning_rate": 6.4114114114114125e-06,
      "loss": 0.4937,
      "step": 11950
    },
    {
      "epoch": 35.92,
      "grad_norm": 5.844981670379639,
      "learning_rate": 6.408408408408408e-06,
      "loss": 0.5128,
      "step": 11960
    },
    {
      "epoch": 35.95,
      "grad_norm": 4.290022373199463,
      "learning_rate": 6.405405405405406e-06,
      "loss": 0.4717,
      "step": 11970
    },
    {
      "epoch": 35.98,
      "grad_norm": 3.495290994644165,
      "learning_rate": 6.402402402402403e-06,
      "loss": 0.5121,
      "step": 11980
    },
    {
      "epoch": 36.0,
      "eval_accuracy": 0.9172,
      "eval_loss": 0.31883522868156433,
      "eval_runtime": 26.15,
      "eval_samples_per_second": 382.409,
      "eval_steps_per_second": 1.53,
      "step": 11988
    },
    {
      "epoch": 36.01,
      "grad_norm": 4.688857555389404,
      "learning_rate": 6.3993993993994e-06,
      "loss": 0.4498,
      "step": 11990
    },
    {
      "epoch": 36.04,
      "grad_norm": 4.36939001083374,
      "learning_rate": 6.396396396396397e-06,
      "loss": 0.4247,
      "step": 12000
    },
    {
      "epoch": 36.07,
      "grad_norm": 3.401184558868408,
      "learning_rate": 6.393393393393393e-06,
      "loss": 0.5333,
      "step": 12010
    },
    {
      "epoch": 36.1,
      "grad_norm": 4.953362941741943,
      "learning_rate": 6.390390390390391e-06,
      "loss": 0.4527,
      "step": 12020
    },
    {
      "epoch": 36.13,
      "grad_norm": 3.899674892425537,
      "learning_rate": 6.387387387387388e-06,
      "loss": 0.5379,
      "step": 12030
    },
    {
      "epoch": 36.16,
      "grad_norm": 3.688502550125122,
      "learning_rate": 6.384384384384385e-06,
      "loss": 0.5328,
      "step": 12040
    },
    {
      "epoch": 36.19,
      "grad_norm": 3.257268190383911,
      "learning_rate": 6.381381381381382e-06,
      "loss": 0.4246,
      "step": 12050
    },
    {
      "epoch": 36.22,
      "grad_norm": 3.4354043006896973,
      "learning_rate": 6.378378378378379e-06,
      "loss": 0.4765,
      "step": 12060
    },
    {
      "epoch": 36.25,
      "grad_norm": 5.303877830505371,
      "learning_rate": 6.3753753753753755e-06,
      "loss": 0.431,
      "step": 12070
    },
    {
      "epoch": 36.28,
      "grad_norm": 4.975677490234375,
      "learning_rate": 6.372372372372373e-06,
      "loss": 0.4844,
      "step": 12080
    },
    {
      "epoch": 36.31,
      "grad_norm": 4.587069034576416,
      "learning_rate": 6.3693693693693704e-06,
      "loss": 0.4586,
      "step": 12090
    },
    {
      "epoch": 36.34,
      "grad_norm": 4.4542036056518555,
      "learning_rate": 6.366366366366366e-06,
      "loss": 0.4543,
      "step": 12100
    },
    {
      "epoch": 36.37,
      "grad_norm": 4.5289177894592285,
      "learning_rate": 6.363363363363364e-06,
      "loss": 0.4921,
      "step": 12110
    },
    {
      "epoch": 36.4,
      "grad_norm": 3.766392230987549,
      "learning_rate": 6.360360360360361e-06,
      "loss": 0.5229,
      "step": 12120
    },
    {
      "epoch": 36.43,
      "grad_norm": 4.781548023223877,
      "learning_rate": 6.357357357357358e-06,
      "loss": 0.5054,
      "step": 12130
    },
    {
      "epoch": 36.46,
      "grad_norm": 3.786529541015625,
      "learning_rate": 6.354354354354355e-06,
      "loss": 0.4808,
      "step": 12140
    },
    {
      "epoch": 36.49,
      "grad_norm": 4.112846374511719,
      "learning_rate": 6.351351351351351e-06,
      "loss": 0.4725,
      "step": 12150
    },
    {
      "epoch": 36.52,
      "grad_norm": 5.792331218719482,
      "learning_rate": 6.3483483483483485e-06,
      "loss": 0.495,
      "step": 12160
    },
    {
      "epoch": 36.55,
      "grad_norm": 11.136505126953125,
      "learning_rate": 6.345345345345346e-06,
      "loss": 0.5116,
      "step": 12170
    },
    {
      "epoch": 36.58,
      "grad_norm": 4.237509727478027,
      "learning_rate": 6.342342342342343e-06,
      "loss": 0.5132,
      "step": 12180
    },
    {
      "epoch": 36.61,
      "grad_norm": 5.541947841644287,
      "learning_rate": 6.33933933933934e-06,
      "loss": 0.4997,
      "step": 12190
    },
    {
      "epoch": 36.64,
      "grad_norm": 9.812138557434082,
      "learning_rate": 6.336336336336338e-06,
      "loss": 0.4906,
      "step": 12200
    },
    {
      "epoch": 36.67,
      "grad_norm": 5.890273094177246,
      "learning_rate": 6.333333333333333e-06,
      "loss": 0.4977,
      "step": 12210
    },
    {
      "epoch": 36.7,
      "grad_norm": 4.361057758331299,
      "learning_rate": 6.330330330330331e-06,
      "loss": 0.4694,
      "step": 12220
    },
    {
      "epoch": 36.73,
      "grad_norm": 6.098171234130859,
      "learning_rate": 6.327327327327328e-06,
      "loss": 0.495,
      "step": 12230
    },
    {
      "epoch": 36.76,
      "grad_norm": 3.5321567058563232,
      "learning_rate": 6.324324324324325e-06,
      "loss": 0.4492,
      "step": 12240
    },
    {
      "epoch": 36.79,
      "grad_norm": 5.207321643829346,
      "learning_rate": 6.321321321321322e-06,
      "loss": 0.4912,
      "step": 12250
    },
    {
      "epoch": 36.82,
      "grad_norm": 5.184374809265137,
      "learning_rate": 6.318318318318318e-06,
      "loss": 0.512,
      "step": 12260
    },
    {
      "epoch": 36.85,
      "grad_norm": 4.539032459259033,
      "learning_rate": 6.315315315315316e-06,
      "loss": 0.444,
      "step": 12270
    },
    {
      "epoch": 36.88,
      "grad_norm": 6.274582862854004,
      "learning_rate": 6.312312312312313e-06,
      "loss": 0.4954,
      "step": 12280
    },
    {
      "epoch": 36.91,
      "grad_norm": 4.470296859741211,
      "learning_rate": 6.309309309309309e-06,
      "loss": 0.4896,
      "step": 12290
    },
    {
      "epoch": 36.94,
      "grad_norm": 4.493797779083252,
      "learning_rate": 6.3063063063063065e-06,
      "loss": 0.4783,
      "step": 12300
    },
    {
      "epoch": 36.97,
      "grad_norm": 5.021604537963867,
      "learning_rate": 6.303303303303304e-06,
      "loss": 0.5179,
      "step": 12310
    },
    {
      "epoch": 37.0,
      "grad_norm": 4.532876014709473,
      "learning_rate": 6.300300300300301e-06,
      "loss": 0.4734,
      "step": 12320
    },
    {
      "epoch": 37.0,
      "eval_accuracy": 0.9193,
      "eval_loss": 0.3174118101596832,
      "eval_runtime": 25.9732,
      "eval_samples_per_second": 385.012,
      "eval_steps_per_second": 1.54,
      "step": 12321
    },
    {
      "epoch": 37.03,
      "grad_norm": 4.60337495803833,
      "learning_rate": 6.297297297297298e-06,
      "loss": 0.467,
      "step": 12330
    },
    {
      "epoch": 37.06,
      "grad_norm": 4.364501476287842,
      "learning_rate": 6.2942942942942955e-06,
      "loss": 0.457,
      "step": 12340
    },
    {
      "epoch": 37.09,
      "grad_norm": 6.581215858459473,
      "learning_rate": 6.291291291291291e-06,
      "loss": 0.483,
      "step": 12350
    },
    {
      "epoch": 37.12,
      "grad_norm": 5.2674150466918945,
      "learning_rate": 6.288288288288289e-06,
      "loss": 0.5288,
      "step": 12360
    },
    {
      "epoch": 37.15,
      "grad_norm": 5.03020715713501,
      "learning_rate": 6.2852852852852854e-06,
      "loss": 0.4659,
      "step": 12370
    },
    {
      "epoch": 37.18,
      "grad_norm": 4.408522605895996,
      "learning_rate": 6.282282282282283e-06,
      "loss": 0.4963,
      "step": 12380
    },
    {
      "epoch": 37.21,
      "grad_norm": 4.20792818069458,
      "learning_rate": 6.27927927927928e-06,
      "loss": 0.5078,
      "step": 12390
    },
    {
      "epoch": 37.24,
      "grad_norm": 3.7951042652130127,
      "learning_rate": 6.276276276276276e-06,
      "loss": 0.5024,
      "step": 12400
    },
    {
      "epoch": 37.27,
      "grad_norm": 5.275200366973877,
      "learning_rate": 6.273273273273274e-06,
      "loss": 0.4775,
      "step": 12410
    },
    {
      "epoch": 37.3,
      "grad_norm": 4.369055271148682,
      "learning_rate": 6.270270270270271e-06,
      "loss": 0.4902,
      "step": 12420
    },
    {
      "epoch": 37.33,
      "grad_norm": 6.0532331466674805,
      "learning_rate": 6.267267267267268e-06,
      "loss": 0.4808,
      "step": 12430
    },
    {
      "epoch": 37.36,
      "grad_norm": 5.8879594802856445,
      "learning_rate": 6.264264264264264e-06,
      "loss": 0.506,
      "step": 12440
    },
    {
      "epoch": 37.39,
      "grad_norm": 4.580647945404053,
      "learning_rate": 6.261261261261262e-06,
      "loss": 0.4653,
      "step": 12450
    },
    {
      "epoch": 37.42,
      "grad_norm": 4.638852119445801,
      "learning_rate": 6.2582582582582585e-06,
      "loss": 0.4214,
      "step": 12460
    },
    {
      "epoch": 37.45,
      "grad_norm": 3.8956124782562256,
      "learning_rate": 6.255255255255256e-06,
      "loss": 0.5519,
      "step": 12470
    },
    {
      "epoch": 37.48,
      "grad_norm": 6.023074150085449,
      "learning_rate": 6.2522522522522535e-06,
      "loss": 0.5331,
      "step": 12480
    },
    {
      "epoch": 37.51,
      "grad_norm": 5.346560001373291,
      "learning_rate": 6.249249249249249e-06,
      "loss": 0.4302,
      "step": 12490
    },
    {
      "epoch": 37.54,
      "grad_norm": 3.9033756256103516,
      "learning_rate": 6.246246246246247e-06,
      "loss": 0.4819,
      "step": 12500
    },
    {
      "epoch": 37.57,
      "grad_norm": 4.675529479980469,
      "learning_rate": 6.243243243243243e-06,
      "loss": 0.4787,
      "step": 12510
    },
    {
      "epoch": 37.6,
      "grad_norm": 4.389764308929443,
      "learning_rate": 6.240240240240241e-06,
      "loss": 0.5223,
      "step": 12520
    },
    {
      "epoch": 37.63,
      "grad_norm": 4.986979007720947,
      "learning_rate": 6.237237237237238e-06,
      "loss": 0.5087,
      "step": 12530
    },
    {
      "epoch": 37.66,
      "grad_norm": 5.554018020629883,
      "learning_rate": 6.234234234234234e-06,
      "loss": 0.4889,
      "step": 12540
    },
    {
      "epoch": 37.69,
      "grad_norm": 4.088570594787598,
      "learning_rate": 6.2312312312312316e-06,
      "loss": 0.4818,
      "step": 12550
    },
    {
      "epoch": 37.72,
      "grad_norm": 4.196356773376465,
      "learning_rate": 6.228228228228229e-06,
      "loss": 0.4197,
      "step": 12560
    },
    {
      "epoch": 37.75,
      "grad_norm": 4.250049591064453,
      "learning_rate": 6.225225225225226e-06,
      "loss": 0.45,
      "step": 12570
    },
    {
      "epoch": 37.78,
      "grad_norm": 3.9590983390808105,
      "learning_rate": 6.222222222222223e-06,
      "loss": 0.4352,
      "step": 12580
    },
    {
      "epoch": 37.81,
      "grad_norm": 4.1846442222595215,
      "learning_rate": 6.219219219219221e-06,
      "loss": 0.4968,
      "step": 12590
    },
    {
      "epoch": 37.84,
      "grad_norm": 4.758632183074951,
      "learning_rate": 6.2162162162162164e-06,
      "loss": 0.4516,
      "step": 12600
    },
    {
      "epoch": 37.87,
      "grad_norm": 4.275233745574951,
      "learning_rate": 6.213213213213214e-06,
      "loss": 0.4844,
      "step": 12610
    },
    {
      "epoch": 37.9,
      "grad_norm": 4.182071685791016,
      "learning_rate": 6.2102102102102105e-06,
      "loss": 0.4612,
      "step": 12620
    },
    {
      "epoch": 37.93,
      "grad_norm": 4.859713077545166,
      "learning_rate": 6.207207207207208e-06,
      "loss": 0.5172,
      "step": 12630
    },
    {
      "epoch": 37.96,
      "grad_norm": 4.407437801361084,
      "learning_rate": 6.204204204204205e-06,
      "loss": 0.5203,
      "step": 12640
    },
    {
      "epoch": 37.99,
      "grad_norm": 5.01877498626709,
      "learning_rate": 6.201201201201201e-06,
      "loss": 0.5554,
      "step": 12650
    },
    {
      "epoch": 38.0,
      "eval_accuracy": 0.9196,
      "eval_loss": 0.31083497405052185,
      "eval_runtime": 26.225,
      "eval_samples_per_second": 381.315,
      "eval_steps_per_second": 1.525,
      "step": 12654
    },
    {
      "epoch": 38.02,
      "grad_norm": 6.460936546325684,
      "learning_rate": 6.198198198198199e-06,
      "loss": 0.4583,
      "step": 12660
    },
    {
      "epoch": 38.05,
      "grad_norm": 5.002228260040283,
      "learning_rate": 6.195195195195196e-06,
      "loss": 0.5129,
      "step": 12670
    },
    {
      "epoch": 38.08,
      "grad_norm": 5.627586841583252,
      "learning_rate": 6.192192192192192e-06,
      "loss": 0.4622,
      "step": 12680
    },
    {
      "epoch": 38.11,
      "grad_norm": 3.3762929439544678,
      "learning_rate": 6.1891891891891895e-06,
      "loss": 0.4751,
      "step": 12690
    },
    {
      "epoch": 38.14,
      "grad_norm": 5.8351545333862305,
      "learning_rate": 6.186186186186187e-06,
      "loss": 0.4964,
      "step": 12700
    },
    {
      "epoch": 38.17,
      "grad_norm": 3.943376302719116,
      "learning_rate": 6.183183183183184e-06,
      "loss": 0.4533,
      "step": 12710
    },
    {
      "epoch": 38.2,
      "grad_norm": 5.2256269454956055,
      "learning_rate": 6.180180180180181e-06,
      "loss": 0.5179,
      "step": 12720
    },
    {
      "epoch": 38.23,
      "grad_norm": 4.85332727432251,
      "learning_rate": 6.177177177177177e-06,
      "loss": 0.4436,
      "step": 12730
    },
    {
      "epoch": 38.26,
      "grad_norm": 32.34022521972656,
      "learning_rate": 6.174174174174174e-06,
      "loss": 0.4833,
      "step": 12740
    },
    {
      "epoch": 38.29,
      "grad_norm": 5.060525417327881,
      "learning_rate": 6.171171171171172e-06,
      "loss": 0.4905,
      "step": 12750
    },
    {
      "epoch": 38.32,
      "grad_norm": 6.252566814422607,
      "learning_rate": 6.1681681681681685e-06,
      "loss": 0.4897,
      "step": 12760
    },
    {
      "epoch": 38.35,
      "grad_norm": 5.269845485687256,
      "learning_rate": 6.165165165165166e-06,
      "loss": 0.44,
      "step": 12770
    },
    {
      "epoch": 38.38,
      "grad_norm": 4.707264423370361,
      "learning_rate": 6.162162162162163e-06,
      "loss": 0.4289,
      "step": 12780
    },
    {
      "epoch": 38.41,
      "grad_norm": 4.847848892211914,
      "learning_rate": 6.159159159159159e-06,
      "loss": 0.4568,
      "step": 12790
    },
    {
      "epoch": 38.44,
      "grad_norm": 6.254144668579102,
      "learning_rate": 6.156156156156157e-06,
      "loss": 0.4907,
      "step": 12800
    },
    {
      "epoch": 38.47,
      "grad_norm": 3.5328307151794434,
      "learning_rate": 6.153153153153154e-06,
      "loss": 0.4832,
      "step": 12810
    },
    {
      "epoch": 38.5,
      "grad_norm": 3.0988495349884033,
      "learning_rate": 6.150150150150151e-06,
      "loss": 0.4956,
      "step": 12820
    },
    {
      "epoch": 38.53,
      "grad_norm": 4.955842971801758,
      "learning_rate": 6.147147147147147e-06,
      "loss": 0.46,
      "step": 12830
    },
    {
      "epoch": 38.56,
      "grad_norm": 6.138220310211182,
      "learning_rate": 6.144144144144145e-06,
      "loss": 0.494,
      "step": 12840
    },
    {
      "epoch": 38.59,
      "grad_norm": 5.635361194610596,
      "learning_rate": 6.1411411411411415e-06,
      "loss": 0.4608,
      "step": 12850
    },
    {
      "epoch": 38.62,
      "grad_norm": 4.4537458419799805,
      "learning_rate": 6.138138138138139e-06,
      "loss": 0.5117,
      "step": 12860
    },
    {
      "epoch": 38.65,
      "grad_norm": 5.3862481117248535,
      "learning_rate": 6.135135135135135e-06,
      "loss": 0.4349,
      "step": 12870
    },
    {
      "epoch": 38.68,
      "grad_norm": 4.768416404724121,
      "learning_rate": 6.132132132132132e-06,
      "loss": 0.4979,
      "step": 12880
    },
    {
      "epoch": 38.71,
      "grad_norm": 5.998769760131836,
      "learning_rate": 6.12912912912913e-06,
      "loss": 0.5081,
      "step": 12890
    },
    {
      "epoch": 38.74,
      "grad_norm": 5.756679534912109,
      "learning_rate": 6.126126126126126e-06,
      "loss": 0.4924,
      "step": 12900
    },
    {
      "epoch": 38.77,
      "grad_norm": 3.7122647762298584,
      "learning_rate": 6.123123123123124e-06,
      "loss": 0.4304,
      "step": 12910
    },
    {
      "epoch": 38.8,
      "grad_norm": 3.1792144775390625,
      "learning_rate": 6.120120120120121e-06,
      "loss": 0.4214,
      "step": 12920
    },
    {
      "epoch": 38.83,
      "grad_norm": 4.193253993988037,
      "learning_rate": 6.117117117117117e-06,
      "loss": 0.4842,
      "step": 12930
    },
    {
      "epoch": 38.86,
      "grad_norm": 5.543241024017334,
      "learning_rate": 6.114114114114115e-06,
      "loss": 0.4797,
      "step": 12940
    },
    {
      "epoch": 38.89,
      "grad_norm": 5.420256614685059,
      "learning_rate": 6.111111111111112e-06,
      "loss": 0.4779,
      "step": 12950
    },
    {
      "epoch": 38.92,
      "grad_norm": 4.960844993591309,
      "learning_rate": 6.108108108108109e-06,
      "loss": 0.4977,
      "step": 12960
    },
    {
      "epoch": 38.95,
      "grad_norm": 4.85521936416626,
      "learning_rate": 6.105105105105106e-06,
      "loss": 0.4911,
      "step": 12970
    },
    {
      "epoch": 38.98,
      "grad_norm": 3.604191780090332,
      "learning_rate": 6.102102102102102e-06,
      "loss": 0.4573,
      "step": 12980
    },
    {
      "epoch": 39.0,
      "eval_accuracy": 0.9203,
      "eval_loss": 0.3111189901828766,
      "eval_runtime": 26.4985,
      "eval_samples_per_second": 377.38,
      "eval_steps_per_second": 1.51,
      "step": 12987
    },
    {
      "epoch": 39.01,
      "grad_norm": 5.761185169219971,
      "learning_rate": 6.0990990990990995e-06,
      "loss": 0.4552,
      "step": 12990
    },
    {
      "epoch": 39.04,
      "grad_norm": 8.790044784545898,
      "learning_rate": 6.096096096096097e-06,
      "loss": 0.4732,
      "step": 13000
    },
    {
      "epoch": 39.07,
      "grad_norm": 4.661570072174072,
      "learning_rate": 6.0930930930930936e-06,
      "loss": 0.4442,
      "step": 13010
    },
    {
      "epoch": 39.1,
      "grad_norm": 4.484933376312256,
      "learning_rate": 6.09009009009009e-06,
      "loss": 0.5375,
      "step": 13020
    },
    {
      "epoch": 39.13,
      "grad_norm": 4.309755325317383,
      "learning_rate": 6.087087087087088e-06,
      "loss": 0.465,
      "step": 13030
    },
    {
      "epoch": 39.16,
      "grad_norm": 5.274332046508789,
      "learning_rate": 6.084084084084084e-06,
      "loss": 0.4746,
      "step": 13040
    },
    {
      "epoch": 39.19,
      "grad_norm": 6.931982040405273,
      "learning_rate": 6.081081081081082e-06,
      "loss": 0.4746,
      "step": 13050
    },
    {
      "epoch": 39.22,
      "grad_norm": 3.775691509246826,
      "learning_rate": 6.078078078078079e-06,
      "loss": 0.4113,
      "step": 13060
    },
    {
      "epoch": 39.25,
      "grad_norm": 6.639736652374268,
      "learning_rate": 6.075075075075075e-06,
      "loss": 0.4563,
      "step": 13070
    },
    {
      "epoch": 39.28,
      "grad_norm": 6.107689380645752,
      "learning_rate": 6.0720720720720725e-06,
      "loss": 0.4773,
      "step": 13080
    },
    {
      "epoch": 39.31,
      "grad_norm": 5.7024078369140625,
      "learning_rate": 6.06906906906907e-06,
      "loss": 0.5295,
      "step": 13090
    },
    {
      "epoch": 39.34,
      "grad_norm": 4.996414661407471,
      "learning_rate": 6.066066066066067e-06,
      "loss": 0.4752,
      "step": 13100
    },
    {
      "epoch": 39.37,
      "grad_norm": 3.619349956512451,
      "learning_rate": 6.063063063063064e-06,
      "loss": 0.5057,
      "step": 13110
    },
    {
      "epoch": 39.4,
      "grad_norm": 3.311629295349121,
      "learning_rate": 6.06006006006006e-06,
      "loss": 0.4678,
      "step": 13120
    },
    {
      "epoch": 39.43,
      "grad_norm": 4.446750640869141,
      "learning_rate": 6.057057057057057e-06,
      "loss": 0.5085,
      "step": 13130
    },
    {
      "epoch": 39.46,
      "grad_norm": 4.55859375,
      "learning_rate": 6.054054054054055e-06,
      "loss": 0.4852,
      "step": 13140
    },
    {
      "epoch": 39.49,
      "grad_norm": 4.04985237121582,
      "learning_rate": 6.0510510510510515e-06,
      "loss": 0.4818,
      "step": 13150
    },
    {
      "epoch": 39.52,
      "grad_norm": 5.457259178161621,
      "learning_rate": 6.048048048048049e-06,
      "loss": 0.3926,
      "step": 13160
    },
    {
      "epoch": 39.55,
      "grad_norm": 3.2358996868133545,
      "learning_rate": 6.045045045045046e-06,
      "loss": 0.5146,
      "step": 13170
    },
    {
      "epoch": 39.58,
      "grad_norm": 3.3737552165985107,
      "learning_rate": 6.042042042042042e-06,
      "loss": 0.4892,
      "step": 13180
    },
    {
      "epoch": 39.61,
      "grad_norm": 3.9016125202178955,
      "learning_rate": 6.03903903903904e-06,
      "loss": 0.5026,
      "step": 13190
    },
    {
      "epoch": 39.64,
      "grad_norm": 4.943258285522461,
      "learning_rate": 6.036036036036037e-06,
      "loss": 0.5245,
      "step": 13200
    },
    {
      "epoch": 39.67,
      "grad_norm": 5.694433689117432,
      "learning_rate": 6.033033033033033e-06,
      "loss": 0.5168,
      "step": 13210
    },
    {
      "epoch": 39.7,
      "grad_norm": 3.50801420211792,
      "learning_rate": 6.0300300300300304e-06,
      "loss": 0.5069,
      "step": 13220
    },
    {
      "epoch": 39.73,
      "grad_norm": 5.018118381500244,
      "learning_rate": 6.027027027027027e-06,
      "loss": 0.4794,
      "step": 13230
    },
    {
      "epoch": 39.76,
      "grad_norm": 3.6679513454437256,
      "learning_rate": 6.0240240240240246e-06,
      "loss": 0.4994,
      "step": 13240
    },
    {
      "epoch": 39.79,
      "grad_norm": 4.077887058258057,
      "learning_rate": 6.021021021021022e-06,
      "loss": 0.5254,
      "step": 13250
    },
    {
      "epoch": 39.82,
      "grad_norm": 5.2488274574279785,
      "learning_rate": 6.018018018018018e-06,
      "loss": 0.4703,
      "step": 13260
    },
    {
      "epoch": 39.85,
      "grad_norm": 4.226322174072266,
      "learning_rate": 6.015015015015015e-06,
      "loss": 0.5161,
      "step": 13270
    },
    {
      "epoch": 39.88,
      "grad_norm": 4.291458606719971,
      "learning_rate": 6.012012012012013e-06,
      "loss": 0.4265,
      "step": 13280
    },
    {
      "epoch": 39.91,
      "grad_norm": 4.763695240020752,
      "learning_rate": 6.009009009009009e-06,
      "loss": 0.4394,
      "step": 13290
    },
    {
      "epoch": 39.94,
      "grad_norm": 5.397831916809082,
      "learning_rate": 6.006006006006007e-06,
      "loss": 0.4785,
      "step": 13300
    },
    {
      "epoch": 39.97,
      "grad_norm": 3.8020317554473877,
      "learning_rate": 6.003003003003004e-06,
      "loss": 0.4506,
      "step": 13310
    },
    {
      "epoch": 40.0,
      "grad_norm": 0.3561044931411743,
      "learning_rate": 6e-06,
      "loss": 0.4692,
      "step": 13320
    },
    {
      "epoch": 40.0,
      "eval_accuracy": 0.9203,
      "eval_loss": 0.3074478507041931,
      "eval_runtime": 26.6396,
      "eval_samples_per_second": 375.381,
      "eval_steps_per_second": 1.502,
      "step": 13320
    },
    {
      "epoch": 40.03,
      "grad_norm": 4.700529098510742,
      "learning_rate": 5.996996996996998e-06,
      "loss": 0.5115,
      "step": 13330
    },
    {
      "epoch": 40.06,
      "grad_norm": 6.283649921417236,
      "learning_rate": 5.993993993993994e-06,
      "loss": 0.5058,
      "step": 13340
    },
    {
      "epoch": 40.09,
      "grad_norm": 4.070054054260254,
      "learning_rate": 5.990990990990992e-06,
      "loss": 0.4418,
      "step": 13350
    },
    {
      "epoch": 40.12,
      "grad_norm": 5.326721668243408,
      "learning_rate": 5.987987987987988e-06,
      "loss": 0.526,
      "step": 13360
    },
    {
      "epoch": 40.15,
      "grad_norm": 4.9221720695495605,
      "learning_rate": 5.984984984984985e-06,
      "loss": 0.4852,
      "step": 13370
    },
    {
      "epoch": 40.18,
      "grad_norm": 6.132660865783691,
      "learning_rate": 5.9819819819819825e-06,
      "loss": 0.5221,
      "step": 13380
    },
    {
      "epoch": 40.21,
      "grad_norm": 5.296045780181885,
      "learning_rate": 5.97897897897898e-06,
      "loss": 0.4917,
      "step": 13390
    },
    {
      "epoch": 40.24,
      "grad_norm": 5.326955318450928,
      "learning_rate": 5.975975975975976e-06,
      "loss": 0.4982,
      "step": 13400
    },
    {
      "epoch": 40.27,
      "grad_norm": 4.939719200134277,
      "learning_rate": 5.972972972972973e-06,
      "loss": 0.4549,
      "step": 13410
    },
    {
      "epoch": 40.3,
      "grad_norm": 5.04403829574585,
      "learning_rate": 5.969969969969971e-06,
      "loss": 0.4178,
      "step": 13420
    },
    {
      "epoch": 40.33,
      "grad_norm": 4.405360698699951,
      "learning_rate": 5.966966966966967e-06,
      "loss": 0.4578,
      "step": 13430
    },
    {
      "epoch": 40.36,
      "grad_norm": 4.024507999420166,
      "learning_rate": 5.963963963963965e-06,
      "loss": 0.4424,
      "step": 13440
    },
    {
      "epoch": 40.39,
      "grad_norm": 4.401228904724121,
      "learning_rate": 5.960960960960962e-06,
      "loss": 0.5585,
      "step": 13450
    },
    {
      "epoch": 40.42,
      "grad_norm": 3.543808698654175,
      "learning_rate": 5.957957957957958e-06,
      "loss": 0.4806,
      "step": 13460
    },
    {
      "epoch": 40.45,
      "grad_norm": 7.205944061279297,
      "learning_rate": 5.9549549549549556e-06,
      "loss": 0.4336,
      "step": 13470
    },
    {
      "epoch": 40.48,
      "grad_norm": 6.0269455909729,
      "learning_rate": 5.951951951951952e-06,
      "loss": 0.4664,
      "step": 13480
    },
    {
      "epoch": 40.51,
      "grad_norm": 4.739034175872803,
      "learning_rate": 5.94894894894895e-06,
      "loss": 0.4657,
      "step": 13490
    },
    {
      "epoch": 40.54,
      "grad_norm": 4.843464374542236,
      "learning_rate": 5.945945945945947e-06,
      "loss": 0.4424,
      "step": 13500
    },
    {
      "epoch": 40.57,
      "grad_norm": 3.6364638805389404,
      "learning_rate": 5.942942942942943e-06,
      "loss": 0.4873,
      "step": 13510
    },
    {
      "epoch": 40.6,
      "grad_norm": 3.925708293914795,
      "learning_rate": 5.93993993993994e-06,
      "loss": 0.4664,
      "step": 13520
    },
    {
      "epoch": 40.63,
      "grad_norm": 4.003354072570801,
      "learning_rate": 5.936936936936938e-06,
      "loss": 0.4738,
      "step": 13530
    },
    {
      "epoch": 40.66,
      "grad_norm": 3.8031704425811768,
      "learning_rate": 5.9339339339339345e-06,
      "loss": 0.4573,
      "step": 13540
    },
    {
      "epoch": 40.69,
      "grad_norm": 4.621785640716553,
      "learning_rate": 5.930930930930931e-06,
      "loss": 0.5061,
      "step": 13550
    },
    {
      "epoch": 40.72,
      "grad_norm": 6.505673408508301,
      "learning_rate": 5.927927927927929e-06,
      "loss": 0.4471,
      "step": 13560
    },
    {
      "epoch": 40.75,
      "grad_norm": 4.7050042152404785,
      "learning_rate": 5.924924924924925e-06,
      "loss": 0.4382,
      "step": 13570
    },
    {
      "epoch": 40.78,
      "grad_norm": 3.5394630432128906,
      "learning_rate": 5.921921921921923e-06,
      "loss": 0.4156,
      "step": 13580
    },
    {
      "epoch": 40.81,
      "grad_norm": 4.195276737213135,
      "learning_rate": 5.9189189189189185e-06,
      "loss": 0.4377,
      "step": 13590
    },
    {
      "epoch": 40.84,
      "grad_norm": 4.865070819854736,
      "learning_rate": 5.915915915915916e-06,
      "loss": 0.4552,
      "step": 13600
    },
    {
      "epoch": 40.87,
      "grad_norm": 4.878195762634277,
      "learning_rate": 5.9129129129129135e-06,
      "loss": 0.469,
      "step": 13610
    },
    {
      "epoch": 40.9,
      "grad_norm": 5.631471633911133,
      "learning_rate": 5.90990990990991e-06,
      "loss": 0.4635,
      "step": 13620
    },
    {
      "epoch": 40.93,
      "grad_norm": 3.38757586479187,
      "learning_rate": 5.906906906906908e-06,
      "loss": 0.4781,
      "step": 13630
    },
    {
      "epoch": 40.96,
      "grad_norm": 5.4056549072265625,
      "learning_rate": 5.903903903903905e-06,
      "loss": 0.4698,
      "step": 13640
    },
    {
      "epoch": 40.99,
      "grad_norm": 4.04258918762207,
      "learning_rate": 5.900900900900901e-06,
      "loss": 0.481,
      "step": 13650
    },
    {
      "epoch": 41.0,
      "eval_accuracy": 0.922,
      "eval_loss": 0.3042304813861847,
      "eval_runtime": 26.1006,
      "eval_samples_per_second": 383.133,
      "eval_steps_per_second": 1.533,
      "step": 13653
    }
  ],
  "logging_steps": 10,
  "max_steps": 33300,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 100,
  "save_steps": 500,
  "total_flos": 1.3514839437625344e+20,
  "train_batch_size": 128,
  "trial_name": null,
  "trial_params": null
}