{
  "best_metric": 0.922,
  "best_model_checkpoint": "checkpoint/vit-base/checkpoint-13653",
  "epoch": 100.0,
  "eval_steps": 500,
  "global_step": 33300,
  "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
    },
    {
      "epoch": 41.02,
      "grad_norm": 3.750443458557129,
      "learning_rate": 5.897897897897898e-06,
      "loss": 0.4269,
      "step": 13660
    },
    {
      "epoch": 41.05,
      "grad_norm": 4.158146858215332,
      "learning_rate": 5.894894894894896e-06,
      "loss": 0.3923,
      "step": 13670
    },
    {
      "epoch": 41.08,
      "grad_norm": 5.192984580993652,
      "learning_rate": 5.8918918918918924e-06,
      "loss": 0.4529,
      "step": 13680
    },
    {
      "epoch": 41.11,
      "grad_norm": 4.528656005859375,
      "learning_rate": 5.88888888888889e-06,
      "loss": 0.4201,
      "step": 13690
    },
    {
      "epoch": 41.14,
      "grad_norm": 4.887214183807373,
      "learning_rate": 5.885885885885886e-06,
      "loss": 0.4627,
      "step": 13700
    },
    {
      "epoch": 41.17,
      "grad_norm": 3.72375750541687,
      "learning_rate": 5.882882882882883e-06,
      "loss": 0.482,
      "step": 13710
    },
    {
      "epoch": 41.2,
      "grad_norm": 3.622239112854004,
      "learning_rate": 5.879879879879881e-06,
      "loss": 0.4287,
      "step": 13720
    },
    {
      "epoch": 41.23,
      "grad_norm": 5.498259544372559,
      "learning_rate": 5.876876876876877e-06,
      "loss": 0.436,
      "step": 13730
    },
    {
      "epoch": 41.26,
      "grad_norm": 3.3701064586639404,
      "learning_rate": 5.873873873873874e-06,
      "loss": 0.4958,
      "step": 13740
    },
    {
      "epoch": 41.29,
      "grad_norm": 3.281550407409668,
      "learning_rate": 5.870870870870871e-06,
      "loss": 0.4823,
      "step": 13750
    },
    {
      "epoch": 41.32,
      "grad_norm": 3.811711072921753,
      "learning_rate": 5.867867867867868e-06,
      "loss": 0.4667,
      "step": 13760
    },
    {
      "epoch": 41.35,
      "grad_norm": 24.785980224609375,
      "learning_rate": 5.8648648648648655e-06,
      "loss": 0.425,
      "step": 13770
    },
    {
      "epoch": 41.38,
      "grad_norm": 4.8750691413879395,
      "learning_rate": 5.861861861861863e-06,
      "loss": 0.548,
      "step": 13780
    },
    {
      "epoch": 41.41,
      "grad_norm": 3.9983041286468506,
      "learning_rate": 5.858858858858859e-06,
      "loss": 0.4954,
      "step": 13790
    },
    {
      "epoch": 41.44,
      "grad_norm": 3.3903467655181885,
      "learning_rate": 5.855855855855856e-06,
      "loss": 0.4156,
      "step": 13800
    },
    {
      "epoch": 41.47,
      "grad_norm": 5.495173454284668,
      "learning_rate": 5.852852852852854e-06,
      "loss": 0.483,
      "step": 13810
    },
    {
      "epoch": 41.5,
      "grad_norm": 3.4536101818084717,
      "learning_rate": 5.84984984984985e-06,
      "loss": 0.4221,
      "step": 13820
    },
    {
      "epoch": 41.53,
      "grad_norm": 4.945369720458984,
      "learning_rate": 5.846846846846848e-06,
      "loss": 0.4934,
      "step": 13830
    },
    {
      "epoch": 41.56,
      "grad_norm": 4.636402130126953,
      "learning_rate": 5.843843843843844e-06,
      "loss": 0.4532,
      "step": 13840
    },
    {
      "epoch": 41.59,
      "grad_norm": 3.990445137023926,
      "learning_rate": 5.840840840840841e-06,
      "loss": 0.4533,
      "step": 13850
    },
    {
      "epoch": 41.62,
      "grad_norm": 3.7954633235931396,
      "learning_rate": 5.837837837837839e-06,
      "loss": 0.4346,
      "step": 13860
    },
    {
      "epoch": 41.65,
      "grad_norm": 4.679460048675537,
      "learning_rate": 5.834834834834835e-06,
      "loss": 0.4742,
      "step": 13870
    },
    {
      "epoch": 41.68,
      "grad_norm": 5.723526954650879,
      "learning_rate": 5.831831831831833e-06,
      "loss": 0.4501,
      "step": 13880
    },
    {
      "epoch": 41.71,
      "grad_norm": 4.583334445953369,
      "learning_rate": 5.828828828828829e-06,
      "loss": 0.4669,
      "step": 13890
    },
    {
      "epoch": 41.74,
      "grad_norm": 5.014919757843018,
      "learning_rate": 5.825825825825826e-06,
      "loss": 0.4774,
      "step": 13900
    },
    {
      "epoch": 41.77,
      "grad_norm": 6.053770065307617,
      "learning_rate": 5.8228228228228234e-06,
      "loss": 0.4655,
      "step": 13910
    },
    {
      "epoch": 41.8,
      "grad_norm": 6.2766642570495605,
      "learning_rate": 5.819819819819821e-06,
      "loss": 0.4714,
      "step": 13920
    },
    {
      "epoch": 41.83,
      "grad_norm": 2.392648458480835,
      "learning_rate": 5.816816816816817e-06,
      "loss": 0.4387,
      "step": 13930
    },
    {
      "epoch": 41.86,
      "grad_norm": 3.895156145095825,
      "learning_rate": 5.813813813813814e-06,
      "loss": 0.4326,
      "step": 13940
    },
    {
      "epoch": 41.89,
      "grad_norm": 4.343978404998779,
      "learning_rate": 5.810810810810811e-06,
      "loss": 0.4026,
      "step": 13950
    },
    {
      "epoch": 41.92,
      "grad_norm": 4.4288740158081055,
      "learning_rate": 5.807807807807808e-06,
      "loss": 0.4284,
      "step": 13960
    },
    {
      "epoch": 41.95,
      "grad_norm": 6.651974678039551,
      "learning_rate": 5.804804804804806e-06,
      "loss": 0.4528,
      "step": 13970
    },
    {
      "epoch": 41.98,
      "grad_norm": 3.602437734603882,
      "learning_rate": 5.8018018018018015e-06,
      "loss": 0.4888,
      "step": 13980
    },
    {
      "epoch": 42.0,
      "eval_accuracy": 0.921,
      "eval_loss": 0.3057817220687866,
      "eval_runtime": 26.4186,
      "eval_samples_per_second": 378.521,
      "eval_steps_per_second": 1.514,
      "step": 13986
    },
    {
      "epoch": 42.01,
      "grad_norm": 13.993064880371094,
      "learning_rate": 5.798798798798799e-06,
      "loss": 0.4065,
      "step": 13990
    },
    {
      "epoch": 42.04,
      "grad_norm": 6.0162272453308105,
      "learning_rate": 5.7957957957957965e-06,
      "loss": 0.4811,
      "step": 14000
    },
    {
      "epoch": 42.07,
      "grad_norm": 3.667475938796997,
      "learning_rate": 5.792792792792793e-06,
      "loss": 0.5225,
      "step": 14010
    },
    {
      "epoch": 42.1,
      "grad_norm": 6.183557510375977,
      "learning_rate": 5.789789789789791e-06,
      "loss": 0.448,
      "step": 14020
    },
    {
      "epoch": 42.13,
      "grad_norm": 4.144411087036133,
      "learning_rate": 5.786786786786788e-06,
      "loss": 0.4558,
      "step": 14030
    },
    {
      "epoch": 42.16,
      "grad_norm": 5.252355575561523,
      "learning_rate": 5.783783783783784e-06,
      "loss": 0.4254,
      "step": 14040
    },
    {
      "epoch": 42.19,
      "grad_norm": 4.259465217590332,
      "learning_rate": 5.780780780780781e-06,
      "loss": 0.492,
      "step": 14050
    },
    {
      "epoch": 42.22,
      "grad_norm": 5.228707790374756,
      "learning_rate": 5.777777777777778e-06,
      "loss": 0.4544,
      "step": 14060
    },
    {
      "epoch": 42.25,
      "grad_norm": 3.4775843620300293,
      "learning_rate": 5.7747747747747755e-06,
      "loss": 0.4598,
      "step": 14070
    },
    {
      "epoch": 42.28,
      "grad_norm": 4.7394232749938965,
      "learning_rate": 5.771771771771772e-06,
      "loss": 0.4743,
      "step": 14080
    },
    {
      "epoch": 42.31,
      "grad_norm": 4.009560585021973,
      "learning_rate": 5.768768768768769e-06,
      "loss": 0.4594,
      "step": 14090
    },
    {
      "epoch": 42.34,
      "grad_norm": 5.766927242279053,
      "learning_rate": 5.765765765765766e-06,
      "loss": 0.4974,
      "step": 14100
    },
    {
      "epoch": 42.37,
      "grad_norm": 4.237868785858154,
      "learning_rate": 5.762762762762764e-06,
      "loss": 0.4477,
      "step": 14110
    },
    {
      "epoch": 42.4,
      "grad_norm": 4.423481464385986,
      "learning_rate": 5.7597597597597595e-06,
      "loss": 0.5139,
      "step": 14120
    },
    {
      "epoch": 42.43,
      "grad_norm": 3.779484748840332,
      "learning_rate": 5.756756756756757e-06,
      "loss": 0.4033,
      "step": 14130
    },
    {
      "epoch": 42.46,
      "grad_norm": 6.86984395980835,
      "learning_rate": 5.7537537537537544e-06,
      "loss": 0.4439,
      "step": 14140
    },
    {
      "epoch": 42.49,
      "grad_norm": 6.490490436553955,
      "learning_rate": 5.750750750750751e-06,
      "loss": 0.4068,
      "step": 14150
    },
    {
      "epoch": 42.52,
      "grad_norm": 3.871152400970459,
      "learning_rate": 5.7477477477477485e-06,
      "loss": 0.4863,
      "step": 14160
    },
    {
      "epoch": 42.55,
      "grad_norm": 5.139131546020508,
      "learning_rate": 5.744744744744746e-06,
      "loss": 0.4955,
      "step": 14170
    },
    {
      "epoch": 42.58,
      "grad_norm": 4.959539413452148,
      "learning_rate": 5.741741741741742e-06,
      "loss": 0.4349,
      "step": 14180
    },
    {
      "epoch": 42.61,
      "grad_norm": 4.078954219818115,
      "learning_rate": 5.738738738738739e-06,
      "loss": 0.4117,
      "step": 14190
    },
    {
      "epoch": 42.64,
      "grad_norm": 3.5708203315734863,
      "learning_rate": 5.735735735735736e-06,
      "loss": 0.4548,
      "step": 14200
    },
    {
      "epoch": 42.67,
      "grad_norm": 4.164892673492432,
      "learning_rate": 5.732732732732733e-06,
      "loss": 0.5256,
      "step": 14210
    },
    {
      "epoch": 42.7,
      "grad_norm": 4.725040435791016,
      "learning_rate": 5.729729729729731e-06,
      "loss": 0.4671,
      "step": 14220
    },
    {
      "epoch": 42.73,
      "grad_norm": 4.0342302322387695,
      "learning_rate": 5.726726726726727e-06,
      "loss": 0.4122,
      "step": 14230
    },
    {
      "epoch": 42.76,
      "grad_norm": 7.624879837036133,
      "learning_rate": 5.723723723723724e-06,
      "loss": 0.4625,
      "step": 14240
    },
    {
      "epoch": 42.79,
      "grad_norm": 5.141872882843018,
      "learning_rate": 5.720720720720722e-06,
      "loss": 0.4592,
      "step": 14250
    },
    {
      "epoch": 42.82,
      "grad_norm": 4.4452924728393555,
      "learning_rate": 5.717717717717718e-06,
      "loss": 0.4747,
      "step": 14260
    },
    {
      "epoch": 42.85,
      "grad_norm": 3.169069290161133,
      "learning_rate": 5.714714714714715e-06,
      "loss": 0.4078,
      "step": 14270
    },
    {
      "epoch": 42.88,
      "grad_norm": 4.386061668395996,
      "learning_rate": 5.711711711711712e-06,
      "loss": 0.3896,
      "step": 14280
    },
    {
      "epoch": 42.91,
      "grad_norm": 5.12047004699707,
      "learning_rate": 5.708708708708709e-06,
      "loss": 0.5097,
      "step": 14290
    },
    {
      "epoch": 42.94,
      "grad_norm": 5.010780334472656,
      "learning_rate": 5.7057057057057065e-06,
      "loss": 0.4646,
      "step": 14300
    },
    {
      "epoch": 42.97,
      "grad_norm": 6.267986297607422,
      "learning_rate": 5.702702702702702e-06,
      "loss": 0.4032,
      "step": 14310
    },
    {
      "epoch": 43.0,
      "eval_accuracy": 0.9211,
      "eval_loss": 0.3024618625640869,
      "eval_runtime": 26.5067,
      "eval_samples_per_second": 377.263,
      "eval_steps_per_second": 1.509,
      "step": 14319
    },
    {
      "epoch": 43.0,
      "grad_norm": 3.2177257537841797,
      "learning_rate": 5.6996996996997e-06,
      "loss": 0.4196,
      "step": 14320
    },
    {
      "epoch": 43.03,
      "grad_norm": 4.594657897949219,
      "learning_rate": 5.696696696696697e-06,
      "loss": 0.4307,
      "step": 14330
    },
    {
      "epoch": 43.06,
      "grad_norm": 4.050478458404541,
      "learning_rate": 5.693693693693694e-06,
      "loss": 0.5181,
      "step": 14340
    },
    {
      "epoch": 43.09,
      "grad_norm": 5.415587902069092,
      "learning_rate": 5.690690690690691e-06,
      "loss": 0.5067,
      "step": 14350
    },
    {
      "epoch": 43.12,
      "grad_norm": 4.965930461883545,
      "learning_rate": 5.687687687687689e-06,
      "loss": 0.4163,
      "step": 14360
    },
    {
      "epoch": 43.15,
      "grad_norm": 5.244529724121094,
      "learning_rate": 5.6846846846846846e-06,
      "loss": 0.4474,
      "step": 14370
    },
    {
      "epoch": 43.18,
      "grad_norm": 5.189445495605469,
      "learning_rate": 5.681681681681682e-06,
      "loss": 0.421,
      "step": 14380
    },
    {
      "epoch": 43.21,
      "grad_norm": 11.601569175720215,
      "learning_rate": 5.6786786786786795e-06,
      "loss": 0.4559,
      "step": 14390
    },
    {
      "epoch": 43.24,
      "grad_norm": 4.7187299728393555,
      "learning_rate": 5.675675675675676e-06,
      "loss": 0.3827,
      "step": 14400
    },
    {
      "epoch": 43.27,
      "grad_norm": 6.4540886878967285,
      "learning_rate": 5.672672672672674e-06,
      "loss": 0.4695,
      "step": 14410
    },
    {
      "epoch": 43.3,
      "grad_norm": 4.562963485717773,
      "learning_rate": 5.6696696696696694e-06,
      "loss": 0.3862,
      "step": 14420
    },
    {
      "epoch": 43.33,
      "grad_norm": 5.647899150848389,
      "learning_rate": 5.666666666666667e-06,
      "loss": 0.4686,
      "step": 14430
    },
    {
      "epoch": 43.36,
      "grad_norm": 4.951442241668701,
      "learning_rate": 5.663663663663664e-06,
      "loss": 0.4707,
      "step": 14440
    },
    {
      "epoch": 43.39,
      "grad_norm": 4.635676383972168,
      "learning_rate": 5.660660660660661e-06,
      "loss": 0.4718,
      "step": 14450
    },
    {
      "epoch": 43.42,
      "grad_norm": 4.30505895614624,
      "learning_rate": 5.657657657657658e-06,
      "loss": 0.4776,
      "step": 14460
    },
    {
      "epoch": 43.45,
      "grad_norm": 5.9097371101379395,
      "learning_rate": 5.654654654654655e-06,
      "loss": 0.4149,
      "step": 14470
    },
    {
      "epoch": 43.48,
      "grad_norm": 4.801514148712158,
      "learning_rate": 5.651651651651652e-06,
      "loss": 0.4721,
      "step": 14480
    },
    {
      "epoch": 43.51,
      "grad_norm": 8.069869041442871,
      "learning_rate": 5.648648648648649e-06,
      "loss": 0.3911,
      "step": 14490
    },
    {
      "epoch": 43.54,
      "grad_norm": 3.869701623916626,
      "learning_rate": 5.645645645645647e-06,
      "loss": 0.4602,
      "step": 14500
    },
    {
      "epoch": 43.57,
      "grad_norm": 9.471203804016113,
      "learning_rate": 5.6426426426426425e-06,
      "loss": 0.4536,
      "step": 14510
    },
    {
      "epoch": 43.6,
      "grad_norm": 4.9999823570251465,
      "learning_rate": 5.63963963963964e-06,
      "loss": 0.3952,
      "step": 14520
    },
    {
      "epoch": 43.63,
      "grad_norm": 4.7708024978637695,
      "learning_rate": 5.6366366366366375e-06,
      "loss": 0.4564,
      "step": 14530
    },
    {
      "epoch": 43.66,
      "grad_norm": 3.6051952838897705,
      "learning_rate": 5.633633633633634e-06,
      "loss": 0.4331,
      "step": 14540
    },
    {
      "epoch": 43.69,
      "grad_norm": 4.822908878326416,
      "learning_rate": 5.6306306306306316e-06,
      "loss": 0.4771,
      "step": 14550
    },
    {
      "epoch": 43.72,
      "grad_norm": 3.750990152359009,
      "learning_rate": 5.627627627627627e-06,
      "loss": 0.3927,
      "step": 14560
    },
    {
      "epoch": 43.75,
      "grad_norm": 4.896886825561523,
      "learning_rate": 5.624624624624625e-06,
      "loss": 0.4715,
      "step": 14570
    },
    {
      "epoch": 43.78,
      "grad_norm": 3.7200682163238525,
      "learning_rate": 5.621621621621622e-06,
      "loss": 0.4634,
      "step": 14580
    },
    {
      "epoch": 43.81,
      "grad_norm": 5.491897106170654,
      "learning_rate": 5.618618618618619e-06,
      "loss": 0.4406,
      "step": 14590
    },
    {
      "epoch": 43.84,
      "grad_norm": 3.980638265609741,
      "learning_rate": 5.615615615615616e-06,
      "loss": 0.4658,
      "step": 14600
    },
    {
      "epoch": 43.87,
      "grad_norm": 3.9713034629821777,
      "learning_rate": 5.612612612612614e-06,
      "loss": 0.3919,
      "step": 14610
    },
    {
      "epoch": 43.9,
      "grad_norm": 3.666268825531006,
      "learning_rate": 5.60960960960961e-06,
      "loss": 0.4289,
      "step": 14620
    },
    {
      "epoch": 43.93,
      "grad_norm": 5.256652355194092,
      "learning_rate": 5.606606606606607e-06,
      "loss": 0.4202,
      "step": 14630
    },
    {
      "epoch": 43.96,
      "grad_norm": 6.012149810791016,
      "learning_rate": 5.603603603603605e-06,
      "loss": 0.4391,
      "step": 14640
    },
    {
      "epoch": 43.99,
      "grad_norm": 4.95004940032959,
      "learning_rate": 5.600600600600601e-06,
      "loss": 0.4731,
      "step": 14650
    },
    {
      "epoch": 44.0,
      "eval_accuracy": 0.9202,
      "eval_loss": 0.3062850534915924,
      "eval_runtime": 26.3893,
      "eval_samples_per_second": 378.942,
      "eval_steps_per_second": 1.516,
      "step": 14652
    },
    {
      "epoch": 44.02,
      "grad_norm": 4.2291059494018555,
      "learning_rate": 5.597597597597598e-06,
      "loss": 0.3955,
      "step": 14660
    },
    {
      "epoch": 44.05,
      "grad_norm": 3.639636278152466,
      "learning_rate": 5.5945945945945945e-06,
      "loss": 0.4344,
      "step": 14670
    },
    {
      "epoch": 44.08,
      "grad_norm": 4.62522029876709,
      "learning_rate": 5.591591591591592e-06,
      "loss": 0.4274,
      "step": 14680
    },
    {
      "epoch": 44.11,
      "grad_norm": 3.2531635761260986,
      "learning_rate": 5.5885885885885895e-06,
      "loss": 0.5003,
      "step": 14690
    },
    {
      "epoch": 44.14,
      "grad_norm": 5.318567752838135,
      "learning_rate": 5.585585585585585e-06,
      "loss": 0.4339,
      "step": 14700
    },
    {
      "epoch": 44.17,
      "grad_norm": 4.41514253616333,
      "learning_rate": 5.582582582582583e-06,
      "loss": 0.4749,
      "step": 14710
    },
    {
      "epoch": 44.2,
      "grad_norm": 4.653870105743408,
      "learning_rate": 5.57957957957958e-06,
      "loss": 0.4055,
      "step": 14720
    },
    {
      "epoch": 44.23,
      "grad_norm": 4.7358784675598145,
      "learning_rate": 5.576576576576577e-06,
      "loss": 0.442,
      "step": 14730
    },
    {
      "epoch": 44.26,
      "grad_norm": 4.107892036437988,
      "learning_rate": 5.573573573573574e-06,
      "loss": 0.4764,
      "step": 14740
    },
    {
      "epoch": 44.29,
      "grad_norm": 3.939893960952759,
      "learning_rate": 5.570570570570572e-06,
      "loss": 0.4277,
      "step": 14750
    },
    {
      "epoch": 44.32,
      "grad_norm": 3.3102755546569824,
      "learning_rate": 5.567567567567568e-06,
      "loss": 0.4825,
      "step": 14760
    },
    {
      "epoch": 44.35,
      "grad_norm": 4.271834373474121,
      "learning_rate": 5.564564564564565e-06,
      "loss": 0.3984,
      "step": 14770
    },
    {
      "epoch": 44.38,
      "grad_norm": 3.7231245040893555,
      "learning_rate": 5.561561561561562e-06,
      "loss": 0.4255,
      "step": 14780
    },
    {
      "epoch": 44.41,
      "grad_norm": 3.7274367809295654,
      "learning_rate": 5.558558558558559e-06,
      "loss": 0.4108,
      "step": 14790
    },
    {
      "epoch": 44.44,
      "grad_norm": 14.244795799255371,
      "learning_rate": 5.555555555555557e-06,
      "loss": 0.4911,
      "step": 14800
    },
    {
      "epoch": 44.47,
      "grad_norm": 4.97599983215332,
      "learning_rate": 5.5525525525525525e-06,
      "loss": 0.4686,
      "step": 14810
    },
    {
      "epoch": 44.5,
      "grad_norm": 4.133742809295654,
      "learning_rate": 5.54954954954955e-06,
      "loss": 0.3819,
      "step": 14820
    },
    {
      "epoch": 44.53,
      "grad_norm": 3.7616562843322754,
      "learning_rate": 5.546546546546547e-06,
      "loss": 0.4278,
      "step": 14830
    },
    {
      "epoch": 44.56,
      "grad_norm": 6.118655204772949,
      "learning_rate": 5.543543543543544e-06,
      "loss": 0.4509,
      "step": 14840
    },
    {
      "epoch": 44.59,
      "grad_norm": 4.463027477264404,
      "learning_rate": 5.540540540540541e-06,
      "loss": 0.4567,
      "step": 14850
    },
    {
      "epoch": 44.62,
      "grad_norm": 3.562678575515747,
      "learning_rate": 5.537537537537538e-06,
      "loss": 0.3826,
      "step": 14860
    },
    {
      "epoch": 44.65,
      "grad_norm": 5.520012855529785,
      "learning_rate": 5.534534534534535e-06,
      "loss": 0.4064,
      "step": 14870
    },
    {
      "epoch": 44.68,
      "grad_norm": 5.109015464782715,
      "learning_rate": 5.531531531531532e-06,
      "loss": 0.4955,
      "step": 14880
    },
    {
      "epoch": 44.71,
      "grad_norm": 4.065291881561279,
      "learning_rate": 5.52852852852853e-06,
      "loss": 0.4558,
      "step": 14890
    },
    {
      "epoch": 44.74,
      "grad_norm": 4.1563544273376465,
      "learning_rate": 5.5255255255255255e-06,
      "loss": 0.4718,
      "step": 14900
    },
    {
      "epoch": 44.77,
      "grad_norm": 5.4338154792785645,
      "learning_rate": 5.522522522522523e-06,
      "loss": 0.4635,
      "step": 14910
    },
    {
      "epoch": 44.8,
      "grad_norm": 3.8264029026031494,
      "learning_rate": 5.51951951951952e-06,
      "loss": 0.4162,
      "step": 14920
    },
    {
      "epoch": 44.83,
      "grad_norm": 6.0958781242370605,
      "learning_rate": 5.516516516516517e-06,
      "loss": 0.4214,
      "step": 14930
    },
    {
      "epoch": 44.86,
      "grad_norm": 4.5139079093933105,
      "learning_rate": 5.513513513513515e-06,
      "loss": 0.4727,
      "step": 14940
    },
    {
      "epoch": 44.89,
      "grad_norm": 4.27504825592041,
      "learning_rate": 5.51051051051051e-06,
      "loss": 0.3931,
      "step": 14950
    },
    {
      "epoch": 44.92,
      "grad_norm": 3.450084924697876,
      "learning_rate": 5.507507507507508e-06,
      "loss": 0.4588,
      "step": 14960
    },
    {
      "epoch": 44.95,
      "grad_norm": 5.20892858505249,
      "learning_rate": 5.504504504504505e-06,
      "loss": 0.5073,
      "step": 14970
    },
    {
      "epoch": 44.98,
      "grad_norm": 4.821224689483643,
      "learning_rate": 5.501501501501502e-06,
      "loss": 0.4574,
      "step": 14980
    },
    {
      "epoch": 45.0,
      "eval_accuracy": 0.92,
      "eval_loss": 0.30516234040260315,
      "eval_runtime": 26.1703,
      "eval_samples_per_second": 382.113,
      "eval_steps_per_second": 1.528,
      "step": 14985
    },
    {
      "epoch": 45.02,
      "grad_norm": 5.444986343383789,
      "learning_rate": 5.4984984984984994e-06,
      "loss": 0.492,
      "step": 14990
    },
    {
      "epoch": 45.05,
      "grad_norm": 3.9214537143707275,
      "learning_rate": 5.495495495495496e-06,
      "loss": 0.4465,
      "step": 15000
    },
    {
      "epoch": 45.08,
      "grad_norm": 3.388601064682007,
      "learning_rate": 5.492492492492493e-06,
      "loss": 0.4182,
      "step": 15010
    },
    {
      "epoch": 45.11,
      "grad_norm": 4.917235851287842,
      "learning_rate": 5.48948948948949e-06,
      "loss": 0.4429,
      "step": 15020
    },
    {
      "epoch": 45.14,
      "grad_norm": 5.455203533172607,
      "learning_rate": 5.486486486486487e-06,
      "loss": 0.4209,
      "step": 15030
    },
    {
      "epoch": 45.17,
      "grad_norm": 5.006196975708008,
      "learning_rate": 5.4834834834834834e-06,
      "loss": 0.4441,
      "step": 15040
    },
    {
      "epoch": 45.2,
      "grad_norm": 6.538152694702148,
      "learning_rate": 5.480480480480481e-06,
      "loss": 0.4537,
      "step": 15050
    },
    {
      "epoch": 45.23,
      "grad_norm": 4.17431116104126,
      "learning_rate": 5.4774774774774776e-06,
      "loss": 0.3942,
      "step": 15060
    },
    {
      "epoch": 45.26,
      "grad_norm": 4.04706335067749,
      "learning_rate": 5.474474474474475e-06,
      "loss": 0.4089,
      "step": 15070
    },
    {
      "epoch": 45.29,
      "grad_norm": 4.886227130889893,
      "learning_rate": 5.4714714714714725e-06,
      "loss": 0.4583,
      "step": 15080
    },
    {
      "epoch": 45.32,
      "grad_norm": 4.194042205810547,
      "learning_rate": 5.468468468468468e-06,
      "loss": 0.4646,
      "step": 15090
    },
    {
      "epoch": 45.35,
      "grad_norm": 4.252357482910156,
      "learning_rate": 5.465465465465466e-06,
      "loss": 0.3726,
      "step": 15100
    },
    {
      "epoch": 45.38,
      "grad_norm": 3.572819471359253,
      "learning_rate": 5.462462462462463e-06,
      "loss": 0.4779,
      "step": 15110
    },
    {
      "epoch": 45.41,
      "grad_norm": 4.422137260437012,
      "learning_rate": 5.45945945945946e-06,
      "loss": 0.457,
      "step": 15120
    },
    {
      "epoch": 45.44,
      "grad_norm": 3.9973361492156982,
      "learning_rate": 5.456456456456457e-06,
      "loss": 0.4093,
      "step": 15130
    },
    {
      "epoch": 45.47,
      "grad_norm": 4.588779449462891,
      "learning_rate": 5.453453453453455e-06,
      "loss": 0.4376,
      "step": 15140
    },
    {
      "epoch": 45.5,
      "grad_norm": 4.632138729095459,
      "learning_rate": 5.450450450450451e-06,
      "loss": 0.4636,
      "step": 15150
    },
    {
      "epoch": 45.53,
      "grad_norm": 4.749342441558838,
      "learning_rate": 5.447447447447448e-06,
      "loss": 0.4239,
      "step": 15160
    },
    {
      "epoch": 45.56,
      "grad_norm": 4.0485029220581055,
      "learning_rate": 5.444444444444445e-06,
      "loss": 0.4573,
      "step": 15170
    },
    {
      "epoch": 45.59,
      "grad_norm": 4.666876792907715,
      "learning_rate": 5.441441441441442e-06,
      "loss": 0.3759,
      "step": 15180
    },
    {
      "epoch": 45.62,
      "grad_norm": 3.2414870262145996,
      "learning_rate": 5.438438438438439e-06,
      "loss": 0.4488,
      "step": 15190
    },
    {
      "epoch": 45.65,
      "grad_norm": 5.149371147155762,
      "learning_rate": 5.4354354354354355e-06,
      "loss": 0.414,
      "step": 15200
    },
    {
      "epoch": 45.68,
      "grad_norm": 5.600489616394043,
      "learning_rate": 5.432432432432433e-06,
      "loss": 0.4625,
      "step": 15210
    },
    {
      "epoch": 45.71,
      "grad_norm": 6.363697052001953,
      "learning_rate": 5.4294294294294304e-06,
      "loss": 0.3641,
      "step": 15220
    },
    {
      "epoch": 45.74,
      "grad_norm": 7.087682247161865,
      "learning_rate": 5.426426426426426e-06,
      "loss": 0.3988,
      "step": 15230
    },
    {
      "epoch": 45.77,
      "grad_norm": 4.25172758102417,
      "learning_rate": 5.423423423423424e-06,
      "loss": 0.4045,
      "step": 15240
    },
    {
      "epoch": 45.8,
      "grad_norm": 6.026050567626953,
      "learning_rate": 5.420420420420421e-06,
      "loss": 0.4341,
      "step": 15250
    },
    {
      "epoch": 45.83,
      "grad_norm": 4.505641937255859,
      "learning_rate": 5.417417417417418e-06,
      "loss": 0.4042,
      "step": 15260
    },
    {
      "epoch": 45.86,
      "grad_norm": 4.4395952224731445,
      "learning_rate": 5.414414414414415e-06,
      "loss": 0.4115,
      "step": 15270
    },
    {
      "epoch": 45.89,
      "grad_norm": 4.4234747886657715,
      "learning_rate": 5.411411411411411e-06,
      "loss": 0.4821,
      "step": 15280
    },
    {
      "epoch": 45.92,
      "grad_norm": 6.2002129554748535,
      "learning_rate": 5.4084084084084086e-06,
      "loss": 0.4418,
      "step": 15290
    },
    {
      "epoch": 45.95,
      "grad_norm": 3.783170461654663,
      "learning_rate": 5.405405405405406e-06,
      "loss": 0.4295,
      "step": 15300
    },
    {
      "epoch": 45.98,
      "grad_norm": 4.145613193511963,
      "learning_rate": 5.402402402402403e-06,
      "loss": 0.3993,
      "step": 15310
    },
    {
      "epoch": 46.0,
      "eval_accuracy": 0.9215,
      "eval_loss": 0.30980437994003296,
      "eval_runtime": 26.215,
      "eval_samples_per_second": 381.461,
      "eval_steps_per_second": 1.526,
      "step": 15318
    },
    {
      "epoch": 46.01,
      "grad_norm": 6.694674491882324,
      "learning_rate": 5.3993993993994e-06,
      "loss": 0.4,
      "step": 15320
    },
    {
      "epoch": 46.04,
      "grad_norm": 5.312897205352783,
      "learning_rate": 5.396396396396398e-06,
      "loss": 0.4072,
      "step": 15330
    },
    {
      "epoch": 46.07,
      "grad_norm": 3.136221408843994,
      "learning_rate": 5.393393393393393e-06,
      "loss": 0.4263,
      "step": 15340
    },
    {
      "epoch": 46.1,
      "grad_norm": 5.447866439819336,
      "learning_rate": 5.390390390390391e-06,
      "loss": 0.4103,
      "step": 15350
    },
    {
      "epoch": 46.13,
      "grad_norm": 6.334971904754639,
      "learning_rate": 5.387387387387388e-06,
      "loss": 0.44,
      "step": 15360
    },
    {
      "epoch": 46.16,
      "grad_norm": 4.249880790710449,
      "learning_rate": 5.384384384384385e-06,
      "loss": 0.4582,
      "step": 15370
    },
    {
      "epoch": 46.19,
      "grad_norm": 4.718948841094971,
      "learning_rate": 5.381381381381382e-06,
      "loss": 0.4222,
      "step": 15380
    },
    {
      "epoch": 46.22,
      "grad_norm": 4.423706531524658,
      "learning_rate": 5.378378378378378e-06,
      "loss": 0.3989,
      "step": 15390
    },
    {
      "epoch": 46.25,
      "grad_norm": 5.1298346519470215,
      "learning_rate": 5.375375375375376e-06,
      "loss": 0.4691,
      "step": 15400
    },
    {
      "epoch": 46.28,
      "grad_norm": 21.546573638916016,
      "learning_rate": 5.372372372372373e-06,
      "loss": 0.4423,
      "step": 15410
    },
    {
      "epoch": 46.31,
      "grad_norm": 4.353137969970703,
      "learning_rate": 5.369369369369369e-06,
      "loss": 0.4653,
      "step": 15420
    },
    {
      "epoch": 46.34,
      "grad_norm": 6.355914115905762,
      "learning_rate": 5.3663663663663665e-06,
      "loss": 0.4717,
      "step": 15430
    },
    {
      "epoch": 46.37,
      "grad_norm": 4.56928014755249,
      "learning_rate": 5.363363363363364e-06,
      "loss": 0.4398,
      "step": 15440
    },
    {
      "epoch": 46.4,
      "grad_norm": 4.304073810577393,
      "learning_rate": 5.360360360360361e-06,
      "loss": 0.3947,
      "step": 15450
    },
    {
      "epoch": 46.43,
      "grad_norm": 3.3754987716674805,
      "learning_rate": 5.357357357357358e-06,
      "loss": 0.4213,
      "step": 15460
    },
    {
      "epoch": 46.46,
      "grad_norm": 4.2586669921875,
      "learning_rate": 5.3543543543543555e-06,
      "loss": 0.4235,
      "step": 15470
    },
    {
      "epoch": 46.49,
      "grad_norm": 4.066396236419678,
      "learning_rate": 5.351351351351351e-06,
      "loss": 0.4516,
      "step": 15480
    },
    {
      "epoch": 46.52,
      "grad_norm": 4.8126912117004395,
      "learning_rate": 5.348348348348349e-06,
      "loss": 0.4187,
      "step": 15490
    },
    {
      "epoch": 46.55,
      "grad_norm": 4.339916706085205,
      "learning_rate": 5.345345345345346e-06,
      "loss": 0.448,
      "step": 15500
    },
    {
      "epoch": 46.58,
      "grad_norm": 4.151970386505127,
      "learning_rate": 5.342342342342343e-06,
      "loss": 0.4186,
      "step": 15510
    },
    {
      "epoch": 46.61,
      "grad_norm": 5.992697238922119,
      "learning_rate": 5.33933933933934e-06,
      "loss": 0.4257,
      "step": 15520
    },
    {
      "epoch": 46.64,
      "grad_norm": 4.657546520233154,
      "learning_rate": 5.336336336336336e-06,
      "loss": 0.4248,
      "step": 15530
    },
    {
      "epoch": 46.67,
      "grad_norm": 4.251131534576416,
      "learning_rate": 5.333333333333334e-06,
      "loss": 0.4232,
      "step": 15540
    },
    {
      "epoch": 46.7,
      "grad_norm": 4.150912761688232,
      "learning_rate": 5.330330330330331e-06,
      "loss": 0.4492,
      "step": 15550
    },
    {
      "epoch": 46.73,
      "grad_norm": 5.373692512512207,
      "learning_rate": 5.327327327327328e-06,
      "loss": 0.3993,
      "step": 15560
    },
    {
      "epoch": 46.76,
      "grad_norm": 4.430792331695557,
      "learning_rate": 5.324324324324324e-06,
      "loss": 0.4881,
      "step": 15570
    },
    {
      "epoch": 46.79,
      "grad_norm": 5.458256244659424,
      "learning_rate": 5.321321321321322e-06,
      "loss": 0.4322,
      "step": 15580
    },
    {
      "epoch": 46.82,
      "grad_norm": 3.6152031421661377,
      "learning_rate": 5.3183183183183185e-06,
      "loss": 0.4189,
      "step": 15590
    },
    {
      "epoch": 46.85,
      "grad_norm": 7.21606969833374,
      "learning_rate": 5.315315315315316e-06,
      "loss": 0.4829,
      "step": 15600
    },
    {
      "epoch": 46.88,
      "grad_norm": 5.945431709289551,
      "learning_rate": 5.3123123123123135e-06,
      "loss": 0.4498,
      "step": 15610
    },
    {
      "epoch": 46.91,
      "grad_norm": 5.194002628326416,
      "learning_rate": 5.309309309309309e-06,
      "loss": 0.4221,
      "step": 15620
    },
    {
      "epoch": 46.94,
      "grad_norm": 4.392026901245117,
      "learning_rate": 5.306306306306307e-06,
      "loss": 0.4223,
      "step": 15630
    },
    {
      "epoch": 46.97,
      "grad_norm": 4.706876277923584,
      "learning_rate": 5.303303303303303e-06,
      "loss": 0.4059,
      "step": 15640
    },
    {
      "epoch": 47.0,
      "grad_norm": 5.420499801635742,
      "learning_rate": 5.300300300300301e-06,
      "loss": 0.4631,
      "step": 15650
    },
    {
      "epoch": 47.0,
      "eval_accuracy": 0.9201,
      "eval_loss": 0.3078014850616455,
      "eval_runtime": 26.3498,
      "eval_samples_per_second": 379.509,
      "eval_steps_per_second": 1.518,
      "step": 15651
    },
    {
      "epoch": 47.03,
      "grad_norm": 6.087631702423096,
      "learning_rate": 5.297297297297298e-06,
      "loss": 0.4724,
      "step": 15660
    },
    {
      "epoch": 47.06,
      "grad_norm": 3.046431303024292,
      "learning_rate": 5.294294294294294e-06,
      "loss": 0.426,
      "step": 15670
    },
    {
      "epoch": 47.09,
      "grad_norm": 3.5722997188568115,
      "learning_rate": 5.291291291291292e-06,
      "loss": 0.4122,
      "step": 15680
    },
    {
      "epoch": 47.12,
      "grad_norm": 2.8889665603637695,
      "learning_rate": 5.288288288288289e-06,
      "loss": 0.4069,
      "step": 15690
    },
    {
      "epoch": 47.15,
      "grad_norm": 5.914718151092529,
      "learning_rate": 5.285285285285286e-06,
      "loss": 0.4734,
      "step": 15700
    },
    {
      "epoch": 47.18,
      "grad_norm": 3.742183208465576,
      "learning_rate": 5.282282282282283e-06,
      "loss": 0.4171,
      "step": 15710
    },
    {
      "epoch": 47.21,
      "grad_norm": 6.282610893249512,
      "learning_rate": 5.27927927927928e-06,
      "loss": 0.4389,
      "step": 15720
    },
    {
      "epoch": 47.24,
      "grad_norm": 5.001829147338867,
      "learning_rate": 5.2762762762762764e-06,
      "loss": 0.4688,
      "step": 15730
    },
    {
      "epoch": 47.27,
      "grad_norm": 4.715219497680664,
      "learning_rate": 5.273273273273274e-06,
      "loss": 0.4393,
      "step": 15740
    },
    {
      "epoch": 47.3,
      "grad_norm": 4.111049652099609,
      "learning_rate": 5.2702702702702705e-06,
      "loss": 0.3983,
      "step": 15750
    },
    {
      "epoch": 47.33,
      "grad_norm": 4.793153762817383,
      "learning_rate": 5.267267267267267e-06,
      "loss": 0.442,
      "step": 15760
    },
    {
      "epoch": 47.36,
      "grad_norm": 3.508852958679199,
      "learning_rate": 5.264264264264265e-06,
      "loss": 0.5187,
      "step": 15770
    },
    {
      "epoch": 47.39,
      "grad_norm": 4.40966272354126,
      "learning_rate": 5.261261261261261e-06,
      "loss": 0.4558,
      "step": 15780
    },
    {
      "epoch": 47.42,
      "grad_norm": 4.837982177734375,
      "learning_rate": 5.258258258258259e-06,
      "loss": 0.4811,
      "step": 15790
    },
    {
      "epoch": 47.45,
      "grad_norm": 5.225708961486816,
      "learning_rate": 5.255255255255256e-06,
      "loss": 0.4282,
      "step": 15800
    },
    {
      "epoch": 47.48,
      "grad_norm": 3.833953380584717,
      "learning_rate": 5.252252252252252e-06,
      "loss": 0.4804,
      "step": 15810
    },
    {
      "epoch": 47.51,
      "grad_norm": 4.869350910186768,
      "learning_rate": 5.2492492492492495e-06,
      "loss": 0.4296,
      "step": 15820
    },
    {
      "epoch": 47.54,
      "grad_norm": 2.9856760501861572,
      "learning_rate": 5.246246246246247e-06,
      "loss": 0.5092,
      "step": 15830
    },
    {
      "epoch": 47.57,
      "grad_norm": 4.569281578063965,
      "learning_rate": 5.243243243243244e-06,
      "loss": 0.3851,
      "step": 15840
    },
    {
      "epoch": 47.6,
      "grad_norm": 3.7475552558898926,
      "learning_rate": 5.240240240240241e-06,
      "loss": 0.4178,
      "step": 15850
    },
    {
      "epoch": 47.63,
      "grad_norm": 3.8802859783172607,
      "learning_rate": 5.2372372372372386e-06,
      "loss": 0.3964,
      "step": 15860
    },
    {
      "epoch": 47.66,
      "grad_norm": 6.050530910491943,
      "learning_rate": 5.234234234234234e-06,
      "loss": 0.4456,
      "step": 15870
    },
    {
      "epoch": 47.69,
      "grad_norm": 4.625086784362793,
      "learning_rate": 5.231231231231232e-06,
      "loss": 0.4611,
      "step": 15880
    },
    {
      "epoch": 47.72,
      "grad_norm": 4.375530242919922,
      "learning_rate": 5.2282282282282285e-06,
      "loss": 0.4168,
      "step": 15890
    },
    {
      "epoch": 47.75,
      "grad_norm": 3.6590945720672607,
      "learning_rate": 5.225225225225226e-06,
      "loss": 0.3775,
      "step": 15900
    },
    {
      "epoch": 47.78,
      "grad_norm": 2.1501262187957764,
      "learning_rate": 5.2222222222222226e-06,
      "loss": 0.3965,
      "step": 15910
    },
    {
      "epoch": 47.81,
      "grad_norm": 3.3284122943878174,
      "learning_rate": 5.219219219219219e-06,
      "loss": 0.4267,
      "step": 15920
    },
    {
      "epoch": 47.84,
      "grad_norm": 5.243638038635254,
      "learning_rate": 5.216216216216217e-06,
      "loss": 0.4432,
      "step": 15930
    },
    {
      "epoch": 47.87,
      "grad_norm": 3.4084110260009766,
      "learning_rate": 5.213213213213214e-06,
      "loss": 0.3933,
      "step": 15940
    },
    {
      "epoch": 47.9,
      "grad_norm": 4.671626567840576,
      "learning_rate": 5.21021021021021e-06,
      "loss": 0.4413,
      "step": 15950
    },
    {
      "epoch": 47.93,
      "grad_norm": 3.6707332134246826,
      "learning_rate": 5.2072072072072074e-06,
      "loss": 0.4342,
      "step": 15960
    },
    {
      "epoch": 47.96,
      "grad_norm": 4.894876956939697,
      "learning_rate": 5.204204204204205e-06,
      "loss": 0.414,
      "step": 15970
    },
    {
      "epoch": 47.99,
      "grad_norm": 4.534253120422363,
      "learning_rate": 5.2012012012012015e-06,
      "loss": 0.409,
      "step": 15980
    },
    {
      "epoch": 48.0,
      "eval_accuracy": 0.9197,
      "eval_loss": 0.3055519163608551,
      "eval_runtime": 26.9486,
      "eval_samples_per_second": 371.077,
      "eval_steps_per_second": 1.484,
      "step": 15984
    },
    {
      "epoch": 48.02,
      "grad_norm": 5.029295921325684,
      "learning_rate": 5.198198198198199e-06,
      "loss": 0.3628,
      "step": 15990
    },
    {
      "epoch": 48.05,
      "grad_norm": 4.425240993499756,
      "learning_rate": 5.195195195195195e-06,
      "loss": 0.474,
      "step": 16000
    },
    {
      "epoch": 48.08,
      "grad_norm": 5.647435665130615,
      "learning_rate": 5.192192192192192e-06,
      "loss": 0.4418,
      "step": 16010
    },
    {
      "epoch": 48.11,
      "grad_norm": 3.117643356323242,
      "learning_rate": 5.18918918918919e-06,
      "loss": 0.4075,
      "step": 16020
    },
    {
      "epoch": 48.14,
      "grad_norm": 5.890537261962891,
      "learning_rate": 5.186186186186186e-06,
      "loss": 0.4176,
      "step": 16030
    },
    {
      "epoch": 48.17,
      "grad_norm": 4.257110595703125,
      "learning_rate": 5.183183183183184e-06,
      "loss": 0.4638,
      "step": 16040
    },
    {
      "epoch": 48.2,
      "grad_norm": 4.908621788024902,
      "learning_rate": 5.180180180180181e-06,
      "loss": 0.4543,
      "step": 16050
    },
    {
      "epoch": 48.23,
      "grad_norm": 5.068427085876465,
      "learning_rate": 5.177177177177177e-06,
      "loss": 0.4235,
      "step": 16060
    },
    {
      "epoch": 48.26,
      "grad_norm": 4.841426372528076,
      "learning_rate": 5.174174174174175e-06,
      "loss": 0.448,
      "step": 16070
    },
    {
      "epoch": 48.29,
      "grad_norm": 7.153292179107666,
      "learning_rate": 5.171171171171172e-06,
      "loss": 0.4482,
      "step": 16080
    },
    {
      "epoch": 48.32,
      "grad_norm": 3.8400135040283203,
      "learning_rate": 5.168168168168169e-06,
      "loss": 0.4551,
      "step": 16090
    },
    {
      "epoch": 48.35,
      "grad_norm": 3.466789722442627,
      "learning_rate": 5.165165165165165e-06,
      "loss": 0.415,
      "step": 16100
    },
    {
      "epoch": 48.38,
      "grad_norm": 4.516876220703125,
      "learning_rate": 5.162162162162162e-06,
      "loss": 0.3723,
      "step": 16110
    },
    {
      "epoch": 48.41,
      "grad_norm": 4.280828952789307,
      "learning_rate": 5.1591591591591595e-06,
      "loss": 0.4562,
      "step": 16120
    },
    {
      "epoch": 48.44,
      "grad_norm": 5.910614967346191,
      "learning_rate": 5.156156156156157e-06,
      "loss": 0.4302,
      "step": 16130
    },
    {
      "epoch": 48.47,
      "grad_norm": 4.0810089111328125,
      "learning_rate": 5.153153153153153e-06,
      "loss": 0.435,
      "step": 16140
    },
    {
      "epoch": 48.5,
      "grad_norm": 6.312894821166992,
      "learning_rate": 5.15015015015015e-06,
      "loss": 0.3247,
      "step": 16150
    },
    {
      "epoch": 48.53,
      "grad_norm": 4.084491729736328,
      "learning_rate": 5.147147147147148e-06,
      "loss": 0.4592,
      "step": 16160
    },
    {
      "epoch": 48.56,
      "grad_norm": 4.468381404876709,
      "learning_rate": 5.144144144144144e-06,
      "loss": 0.4171,
      "step": 16170
    },
    {
      "epoch": 48.59,
      "grad_norm": 5.118911266326904,
      "learning_rate": 5.141141141141142e-06,
      "loss": 0.5041,
      "step": 16180
    },
    {
      "epoch": 48.62,
      "grad_norm": 4.910642623901367,
      "learning_rate": 5.138138138138139e-06,
      "loss": 0.3848,
      "step": 16190
    },
    {
      "epoch": 48.65,
      "grad_norm": 4.208536624908447,
      "learning_rate": 5.135135135135135e-06,
      "loss": 0.4017,
      "step": 16200
    },
    {
      "epoch": 48.68,
      "grad_norm": 5.494334697723389,
      "learning_rate": 5.1321321321321325e-06,
      "loss": 0.396,
      "step": 16210
    },
    {
      "epoch": 48.71,
      "grad_norm": 4.2409186363220215,
      "learning_rate": 5.12912912912913e-06,
      "loss": 0.3805,
      "step": 16220
    },
    {
      "epoch": 48.74,
      "grad_norm": 3.8621647357940674,
      "learning_rate": 5.126126126126127e-06,
      "loss": 0.3948,
      "step": 16230
    },
    {
      "epoch": 48.77,
      "grad_norm": 3.3955607414245605,
      "learning_rate": 5.123123123123124e-06,
      "loss": 0.414,
      "step": 16240
    },
    {
      "epoch": 48.8,
      "grad_norm": 4.547874450683594,
      "learning_rate": 5.12012012012012e-06,
      "loss": 0.4343,
      "step": 16250
    },
    {
      "epoch": 48.83,
      "grad_norm": 3.3248467445373535,
      "learning_rate": 5.117117117117117e-06,
      "loss": 0.3404,
      "step": 16260
    },
    {
      "epoch": 48.86,
      "grad_norm": 8.42970085144043,
      "learning_rate": 5.114114114114115e-06,
      "loss": 0.4503,
      "step": 16270
    },
    {
      "epoch": 48.89,
      "grad_norm": 2.210439920425415,
      "learning_rate": 5.1111111111111115e-06,
      "loss": 0.3731,
      "step": 16280
    },
    {
      "epoch": 48.92,
      "grad_norm": 5.0297956466674805,
      "learning_rate": 5.108108108108108e-06,
      "loss": 0.4678,
      "step": 16290
    },
    {
      "epoch": 48.95,
      "grad_norm": 4.120416641235352,
      "learning_rate": 5.105105105105106e-06,
      "loss": 0.4245,
      "step": 16300
    },
    {
      "epoch": 48.98,
      "grad_norm": 4.170934200286865,
      "learning_rate": 5.102102102102102e-06,
      "loss": 0.4584,
      "step": 16310
    },
    {
      "epoch": 49.0,
      "eval_accuracy": 0.9208,
      "eval_loss": 0.30599337816238403,
      "eval_runtime": 26.3649,
      "eval_samples_per_second": 379.292,
      "eval_steps_per_second": 1.517,
      "step": 16317
    },
    {
      "epoch": 49.01,
      "grad_norm": 4.681585788726807,
      "learning_rate": 5.0990990990991e-06,
      "loss": 0.5272,
      "step": 16320
    },
    {
      "epoch": 49.04,
      "grad_norm": 3.9807257652282715,
      "learning_rate": 5.096096096096097e-06,
      "loss": 0.4536,
      "step": 16330
    },
    {
      "epoch": 49.07,
      "grad_norm": 4.989492416381836,
      "learning_rate": 5.093093093093093e-06,
      "loss": 0.4223,
      "step": 16340
    },
    {
      "epoch": 49.1,
      "grad_norm": 5.326772212982178,
      "learning_rate": 5.0900900900900905e-06,
      "loss": 0.4164,
      "step": 16350
    },
    {
      "epoch": 49.13,
      "grad_norm": 4.621530532836914,
      "learning_rate": 5.087087087087087e-06,
      "loss": 0.4478,
      "step": 16360
    },
    {
      "epoch": 49.16,
      "grad_norm": 5.0887603759765625,
      "learning_rate": 5.0840840840840846e-06,
      "loss": 0.4483,
      "step": 16370
    },
    {
      "epoch": 49.19,
      "grad_norm": 6.91994047164917,
      "learning_rate": 5.081081081081082e-06,
      "loss": 0.4134,
      "step": 16380
    },
    {
      "epoch": 49.22,
      "grad_norm": 5.83650016784668,
      "learning_rate": 5.078078078078078e-06,
      "loss": 0.4765,
      "step": 16390
    },
    {
      "epoch": 49.25,
      "grad_norm": 5.485452175140381,
      "learning_rate": 5.075075075075075e-06,
      "loss": 0.4308,
      "step": 16400
    },
    {
      "epoch": 49.28,
      "grad_norm": 5.179877281188965,
      "learning_rate": 5.072072072072073e-06,
      "loss": 0.4332,
      "step": 16410
    },
    {
      "epoch": 49.31,
      "grad_norm": 4.3198323249816895,
      "learning_rate": 5.069069069069069e-06,
      "loss": 0.4175,
      "step": 16420
    },
    {
      "epoch": 49.34,
      "grad_norm": 10.478413581848145,
      "learning_rate": 5.066066066066067e-06,
      "loss": 0.4253,
      "step": 16430
    },
    {
      "epoch": 49.37,
      "grad_norm": 5.890543460845947,
      "learning_rate": 5.0630630630630635e-06,
      "loss": 0.4293,
      "step": 16440
    },
    {
      "epoch": 49.4,
      "grad_norm": 3.8914778232574463,
      "learning_rate": 5.06006006006006e-06,
      "loss": 0.4474,
      "step": 16450
    },
    {
      "epoch": 49.43,
      "grad_norm": 8.994999885559082,
      "learning_rate": 5.057057057057058e-06,
      "loss": 0.4371,
      "step": 16460
    },
    {
      "epoch": 49.46,
      "grad_norm": 4.208297252655029,
      "learning_rate": 5.054054054054054e-06,
      "loss": 0.4767,
      "step": 16470
    },
    {
      "epoch": 49.49,
      "grad_norm": 4.868051528930664,
      "learning_rate": 5.051051051051051e-06,
      "loss": 0.4013,
      "step": 16480
    },
    {
      "epoch": 49.52,
      "grad_norm": 4.0128912925720215,
      "learning_rate": 5.048048048048048e-06,
      "loss": 0.4758,
      "step": 16490
    },
    {
      "epoch": 49.55,
      "grad_norm": 5.092573642730713,
      "learning_rate": 5.045045045045045e-06,
      "loss": 0.3904,
      "step": 16500
    },
    {
      "epoch": 49.58,
      "grad_norm": 5.283992767333984,
      "learning_rate": 5.0420420420420425e-06,
      "loss": 0.3905,
      "step": 16510
    },
    {
      "epoch": 49.61,
      "grad_norm": 4.520591735839844,
      "learning_rate": 5.03903903903904e-06,
      "loss": 0.4926,
      "step": 16520
    },
    {
      "epoch": 49.64,
      "grad_norm": 6.357444763183594,
      "learning_rate": 5.036036036036036e-06,
      "loss": 0.4212,
      "step": 16530
    },
    {
      "epoch": 49.67,
      "grad_norm": 4.281061172485352,
      "learning_rate": 5.033033033033033e-06,
      "loss": 0.4481,
      "step": 16540
    },
    {
      "epoch": 49.7,
      "grad_norm": 5.377974510192871,
      "learning_rate": 5.030030030030031e-06,
      "loss": 0.4651,
      "step": 16550
    },
    {
      "epoch": 49.73,
      "grad_norm": 5.319176197052002,
      "learning_rate": 5.027027027027027e-06,
      "loss": 0.4264,
      "step": 16560
    },
    {
      "epoch": 49.76,
      "grad_norm": 4.083383560180664,
      "learning_rate": 5.024024024024025e-06,
      "loss": 0.4302,
      "step": 16570
    },
    {
      "epoch": 49.79,
      "grad_norm": 3.761942148208618,
      "learning_rate": 5.021021021021022e-06,
      "loss": 0.3966,
      "step": 16580
    },
    {
      "epoch": 49.82,
      "grad_norm": 4.500251770019531,
      "learning_rate": 5.018018018018018e-06,
      "loss": 0.4374,
      "step": 16590
    },
    {
      "epoch": 49.85,
      "grad_norm": 4.931020259857178,
      "learning_rate": 5.0150150150150156e-06,
      "loss": 0.3813,
      "step": 16600
    },
    {
      "epoch": 49.88,
      "grad_norm": 5.4990458488464355,
      "learning_rate": 5.012012012012012e-06,
      "loss": 0.3938,
      "step": 16610
    },
    {
      "epoch": 49.91,
      "grad_norm": 7.012874603271484,
      "learning_rate": 5.00900900900901e-06,
      "loss": 0.4253,
      "step": 16620
    },
    {
      "epoch": 49.94,
      "grad_norm": 5.192298889160156,
      "learning_rate": 5.006006006006007e-06,
      "loss": 0.4443,
      "step": 16630
    },
    {
      "epoch": 49.97,
      "grad_norm": 10.042019844055176,
      "learning_rate": 5.003003003003003e-06,
      "loss": 0.3836,
      "step": 16640
    },
    {
      "epoch": 50.0,
      "grad_norm": 0.9387697577476501,
      "learning_rate": 5e-06,
      "loss": 0.3853,
      "step": 16650
    },
    {
      "epoch": 50.0,
      "eval_accuracy": 0.9208,
      "eval_loss": 0.3061496615409851,
      "eval_runtime": 25.9592,
      "eval_samples_per_second": 385.22,
      "eval_steps_per_second": 1.541,
      "step": 16650
    },
    {
      "epoch": 50.03,
      "grad_norm": 5.4510016441345215,
      "learning_rate": 4.996996996996997e-06,
      "loss": 0.3733,
      "step": 16660
    },
    {
      "epoch": 50.06,
      "grad_norm": 5.434107780456543,
      "learning_rate": 4.9939939939939945e-06,
      "loss": 0.4513,
      "step": 16670
    },
    {
      "epoch": 50.09,
      "grad_norm": 5.186567306518555,
      "learning_rate": 4.990990990990991e-06,
      "loss": 0.4519,
      "step": 16680
    },
    {
      "epoch": 50.12,
      "grad_norm": 5.895889759063721,
      "learning_rate": 4.987987987987989e-06,
      "loss": 0.407,
      "step": 16690
    },
    {
      "epoch": 50.15,
      "grad_norm": 4.924978733062744,
      "learning_rate": 4.984984984984985e-06,
      "loss": 0.4261,
      "step": 16700
    },
    {
      "epoch": 50.18,
      "grad_norm": 3.0152699947357178,
      "learning_rate": 4.981981981981983e-06,
      "loss": 0.4327,
      "step": 16710
    },
    {
      "epoch": 50.21,
      "grad_norm": 6.40473747253418,
      "learning_rate": 4.978978978978979e-06,
      "loss": 0.4175,
      "step": 16720
    },
    {
      "epoch": 50.24,
      "grad_norm": 6.783638954162598,
      "learning_rate": 4.975975975975976e-06,
      "loss": 0.3976,
      "step": 16730
    },
    {
      "epoch": 50.27,
      "grad_norm": 6.256586074829102,
      "learning_rate": 4.9729729729729735e-06,
      "loss": 0.3698,
      "step": 16740
    },
    {
      "epoch": 50.3,
      "grad_norm": 5.1926751136779785,
      "learning_rate": 4.96996996996997e-06,
      "loss": 0.3883,
      "step": 16750
    },
    {
      "epoch": 50.33,
      "grad_norm": 3.9796361923217773,
      "learning_rate": 4.966966966966968e-06,
      "loss": 0.455,
      "step": 16760
    },
    {
      "epoch": 50.36,
      "grad_norm": 6.166399955749512,
      "learning_rate": 4.963963963963964e-06,
      "loss": 0.4889,
      "step": 16770
    },
    {
      "epoch": 50.39,
      "grad_norm": 4.4162187576293945,
      "learning_rate": 4.960960960960962e-06,
      "loss": 0.4706,
      "step": 16780
    },
    {
      "epoch": 50.42,
      "grad_norm": 3.389835834503174,
      "learning_rate": 4.957957957957958e-06,
      "loss": 0.3748,
      "step": 16790
    },
    {
      "epoch": 50.45,
      "grad_norm": 5.2870941162109375,
      "learning_rate": 4.954954954954955e-06,
      "loss": 0.4038,
      "step": 16800
    },
    {
      "epoch": 50.48,
      "grad_norm": 4.940831661224365,
      "learning_rate": 4.9519519519519524e-06,
      "loss": 0.4241,
      "step": 16810
    },
    {
      "epoch": 50.51,
      "grad_norm": 3.5071663856506348,
      "learning_rate": 4.94894894894895e-06,
      "loss": 0.403,
      "step": 16820
    },
    {
      "epoch": 50.54,
      "grad_norm": 4.041534423828125,
      "learning_rate": 4.9459459459459466e-06,
      "loss": 0.3906,
      "step": 16830
    },
    {
      "epoch": 50.57,
      "grad_norm": 3.8851935863494873,
      "learning_rate": 4.942942942942943e-06,
      "loss": 0.3595,
      "step": 16840
    },
    {
      "epoch": 50.6,
      "grad_norm": 4.452398300170898,
      "learning_rate": 4.939939939939941e-06,
      "loss": 0.415,
      "step": 16850
    },
    {
      "epoch": 50.63,
      "grad_norm": 4.765621185302734,
      "learning_rate": 4.936936936936937e-06,
      "loss": 0.3983,
      "step": 16860
    },
    {
      "epoch": 50.66,
      "grad_norm": 6.130590438842773,
      "learning_rate": 4.933933933933934e-06,
      "loss": 0.4321,
      "step": 16870
    },
    {
      "epoch": 50.69,
      "grad_norm": 3.1836891174316406,
      "learning_rate": 4.930930930930931e-06,
      "loss": 0.4368,
      "step": 16880
    },
    {
      "epoch": 50.72,
      "grad_norm": 4.539780139923096,
      "learning_rate": 4.927927927927929e-06,
      "loss": 0.3948,
      "step": 16890
    },
    {
      "epoch": 50.75,
      "grad_norm": 4.259836673736572,
      "learning_rate": 4.9249249249249255e-06,
      "loss": 0.4485,
      "step": 16900
    },
    {
      "epoch": 50.78,
      "grad_norm": 4.491191387176514,
      "learning_rate": 4.921921921921922e-06,
      "loss": 0.4061,
      "step": 16910
    },
    {
      "epoch": 50.81,
      "grad_norm": 5.331355571746826,
      "learning_rate": 4.91891891891892e-06,
      "loss": 0.4043,
      "step": 16920
    },
    {
      "epoch": 50.84,
      "grad_norm": 4.070823669433594,
      "learning_rate": 4.915915915915916e-06,
      "loss": 0.389,
      "step": 16930
    },
    {
      "epoch": 50.87,
      "grad_norm": 4.639046669006348,
      "learning_rate": 4.912912912912913e-06,
      "loss": 0.4568,
      "step": 16940
    },
    {
      "epoch": 50.9,
      "grad_norm": 4.152167797088623,
      "learning_rate": 4.90990990990991e-06,
      "loss": 0.3948,
      "step": 16950
    },
    {
      "epoch": 50.93,
      "grad_norm": 5.040772914886475,
      "learning_rate": 4.906906906906908e-06,
      "loss": 0.3882,
      "step": 16960
    },
    {
      "epoch": 50.96,
      "grad_norm": 4.293344020843506,
      "learning_rate": 4.9039039039039045e-06,
      "loss": 0.4667,
      "step": 16970
    },
    {
      "epoch": 50.99,
      "grad_norm": 3.4267101287841797,
      "learning_rate": 4.900900900900901e-06,
      "loss": 0.3836,
      "step": 16980
    },
    {
      "epoch": 51.0,
      "eval_accuracy": 0.9216,
      "eval_loss": 0.30722689628601074,
      "eval_runtime": 26.3799,
      "eval_samples_per_second": 379.076,
      "eval_steps_per_second": 1.516,
      "step": 16983
    },
    {
      "epoch": 51.02,
      "grad_norm": 4.319604396820068,
      "learning_rate": 4.897897897897898e-06,
      "loss": 0.3602,
      "step": 16990
    },
    {
      "epoch": 51.05,
      "grad_norm": 4.814965724945068,
      "learning_rate": 4.894894894894895e-06,
      "loss": 0.4202,
      "step": 17000
    },
    {
      "epoch": 51.08,
      "grad_norm": 3.458998203277588,
      "learning_rate": 4.891891891891893e-06,
      "loss": 0.3595,
      "step": 17010
    },
    {
      "epoch": 51.11,
      "grad_norm": 5.099784851074219,
      "learning_rate": 4.888888888888889e-06,
      "loss": 0.4633,
      "step": 17020
    },
    {
      "epoch": 51.14,
      "grad_norm": 4.56049108505249,
      "learning_rate": 4.885885885885887e-06,
      "loss": 0.4272,
      "step": 17030
    },
    {
      "epoch": 51.17,
      "grad_norm": 8.498510360717773,
      "learning_rate": 4.8828828828828834e-06,
      "loss": 0.4209,
      "step": 17040
    },
    {
      "epoch": 51.2,
      "grad_norm": 4.623964309692383,
      "learning_rate": 4.87987987987988e-06,
      "loss": 0.4208,
      "step": 17050
    },
    {
      "epoch": 51.23,
      "grad_norm": 4.963705062866211,
      "learning_rate": 4.876876876876877e-06,
      "loss": 0.4623,
      "step": 17060
    },
    {
      "epoch": 51.26,
      "grad_norm": 4.43102502822876,
      "learning_rate": 4.873873873873874e-06,
      "loss": 0.4298,
      "step": 17070
    },
    {
      "epoch": 51.29,
      "grad_norm": 2.940281629562378,
      "learning_rate": 4.870870870870872e-06,
      "loss": 0.3864,
      "step": 17080
    },
    {
      "epoch": 51.32,
      "grad_norm": 3.591057538986206,
      "learning_rate": 4.867867867867868e-06,
      "loss": 0.4337,
      "step": 17090
    },
    {
      "epoch": 51.35,
      "grad_norm": 5.616214275360107,
      "learning_rate": 4.864864864864866e-06,
      "loss": 0.4048,
      "step": 17100
    },
    {
      "epoch": 51.38,
      "grad_norm": 2.9872093200683594,
      "learning_rate": 4.861861861861862e-06,
      "loss": 0.3945,
      "step": 17110
    },
    {
      "epoch": 51.41,
      "grad_norm": 5.086316108703613,
      "learning_rate": 4.858858858858859e-06,
      "loss": 0.3869,
      "step": 17120
    },
    {
      "epoch": 51.44,
      "grad_norm": 4.262561798095703,
      "learning_rate": 4.855855855855856e-06,
      "loss": 0.3798,
      "step": 17130
    },
    {
      "epoch": 51.47,
      "grad_norm": 8.7372407913208,
      "learning_rate": 4.852852852852853e-06,
      "loss": 0.4284,
      "step": 17140
    },
    {
      "epoch": 51.5,
      "grad_norm": 4.295604705810547,
      "learning_rate": 4.849849849849851e-06,
      "loss": 0.3915,
      "step": 17150
    },
    {
      "epoch": 51.53,
      "grad_norm": 3.548208475112915,
      "learning_rate": 4.846846846846847e-06,
      "loss": 0.4528,
      "step": 17160
    },
    {
      "epoch": 51.56,
      "grad_norm": 12.3329496383667,
      "learning_rate": 4.843843843843844e-06,
      "loss": 0.3653,
      "step": 17170
    },
    {
      "epoch": 51.59,
      "grad_norm": 3.281158208847046,
      "learning_rate": 4.840840840840841e-06,
      "loss": 0.378,
      "step": 17180
    },
    {
      "epoch": 51.62,
      "grad_norm": 5.405097484588623,
      "learning_rate": 4.837837837837838e-06,
      "loss": 0.4599,
      "step": 17190
    },
    {
      "epoch": 51.65,
      "grad_norm": 4.687839508056641,
      "learning_rate": 4.8348348348348355e-06,
      "loss": 0.3912,
      "step": 17200
    },
    {
      "epoch": 51.68,
      "grad_norm": 3.2139227390289307,
      "learning_rate": 4.831831831831832e-06,
      "loss": 0.4149,
      "step": 17210
    },
    {
      "epoch": 51.71,
      "grad_norm": 17.51236915588379,
      "learning_rate": 4.82882882882883e-06,
      "loss": 0.4206,
      "step": 17220
    },
    {
      "epoch": 51.74,
      "grad_norm": 5.346531391143799,
      "learning_rate": 4.825825825825826e-06,
      "loss": 0.4187,
      "step": 17230
    },
    {
      "epoch": 51.77,
      "grad_norm": 5.157400608062744,
      "learning_rate": 4.822822822822823e-06,
      "loss": 0.3952,
      "step": 17240
    },
    {
      "epoch": 51.8,
      "grad_norm": 3.920642375946045,
      "learning_rate": 4.81981981981982e-06,
      "loss": 0.4378,
      "step": 17250
    },
    {
      "epoch": 51.83,
      "grad_norm": 4.922093868255615,
      "learning_rate": 4.816816816816817e-06,
      "loss": 0.4313,
      "step": 17260
    },
    {
      "epoch": 51.86,
      "grad_norm": 4.4549560546875,
      "learning_rate": 4.8138138138138144e-06,
      "loss": 0.3843,
      "step": 17270
    },
    {
      "epoch": 51.89,
      "grad_norm": 4.402873992919922,
      "learning_rate": 4.810810810810811e-06,
      "loss": 0.392,
      "step": 17280
    },
    {
      "epoch": 51.92,
      "grad_norm": 4.427642822265625,
      "learning_rate": 4.8078078078078085e-06,
      "loss": 0.4112,
      "step": 17290
    },
    {
      "epoch": 51.95,
      "grad_norm": 3.921144962310791,
      "learning_rate": 4.804804804804805e-06,
      "loss": 0.4597,
      "step": 17300
    },
    {
      "epoch": 51.98,
      "grad_norm": 5.853477954864502,
      "learning_rate": 4.801801801801802e-06,
      "loss": 0.3969,
      "step": 17310
    },
    {
      "epoch": 52.0,
      "eval_accuracy": 0.9197,
      "eval_loss": 0.30704566836357117,
      "eval_runtime": 26.6047,
      "eval_samples_per_second": 375.874,
      "eval_steps_per_second": 1.503,
      "step": 17316
    },
    {
      "epoch": 52.01,
      "grad_norm": 6.230672836303711,
      "learning_rate": 4.798798798798799e-06,
      "loss": 0.4327,
      "step": 17320
    },
    {
      "epoch": 52.04,
      "grad_norm": 5.472045421600342,
      "learning_rate": 4.795795795795796e-06,
      "loss": 0.4371,
      "step": 17330
    },
    {
      "epoch": 52.07,
      "grad_norm": 8.238033294677734,
      "learning_rate": 4.792792792792793e-06,
      "loss": 0.4823,
      "step": 17340
    },
    {
      "epoch": 52.1,
      "grad_norm": 4.268326282501221,
      "learning_rate": 4.78978978978979e-06,
      "loss": 0.3711,
      "step": 17350
    },
    {
      "epoch": 52.13,
      "grad_norm": 3.3628671169281006,
      "learning_rate": 4.7867867867867875e-06,
      "loss": 0.3951,
      "step": 17360
    },
    {
      "epoch": 52.16,
      "grad_norm": 6.018637180328369,
      "learning_rate": 4.783783783783784e-06,
      "loss": 0.4208,
      "step": 17370
    },
    {
      "epoch": 52.19,
      "grad_norm": 3.919670820236206,
      "learning_rate": 4.780780780780781e-06,
      "loss": 0.4162,
      "step": 17380
    },
    {
      "epoch": 52.22,
      "grad_norm": 4.199753761291504,
      "learning_rate": 4.777777777777778e-06,
      "loss": 0.4225,
      "step": 17390
    },
    {
      "epoch": 52.25,
      "grad_norm": 5.612927436828613,
      "learning_rate": 4.774774774774775e-06,
      "loss": 0.4056,
      "step": 17400
    },
    {
      "epoch": 52.28,
      "grad_norm": 3.8043949604034424,
      "learning_rate": 4.771771771771772e-06,
      "loss": 0.44,
      "step": 17410
    },
    {
      "epoch": 52.31,
      "grad_norm": 5.359033107757568,
      "learning_rate": 4.768768768768769e-06,
      "loss": 0.4074,
      "step": 17420
    },
    {
      "epoch": 52.34,
      "grad_norm": 4.500823020935059,
      "learning_rate": 4.7657657657657665e-06,
      "loss": 0.4545,
      "step": 17430
    },
    {
      "epoch": 52.37,
      "grad_norm": 2.894533157348633,
      "learning_rate": 4.762762762762763e-06,
      "loss": 0.4077,
      "step": 17440
    },
    {
      "epoch": 52.4,
      "grad_norm": 3.913682222366333,
      "learning_rate": 4.75975975975976e-06,
      "loss": 0.4062,
      "step": 17450
    },
    {
      "epoch": 52.43,
      "grad_norm": 4.386532306671143,
      "learning_rate": 4.756756756756757e-06,
      "loss": 0.4128,
      "step": 17460
    },
    {
      "epoch": 52.46,
      "grad_norm": 3.7199032306671143,
      "learning_rate": 4.753753753753754e-06,
      "loss": 0.3808,
      "step": 17470
    },
    {
      "epoch": 52.49,
      "grad_norm": 4.408360004425049,
      "learning_rate": 4.750750750750751e-06,
      "loss": 0.3901,
      "step": 17480
    },
    {
      "epoch": 52.52,
      "grad_norm": 3.3102588653564453,
      "learning_rate": 4.747747747747748e-06,
      "loss": 0.4471,
      "step": 17490
    },
    {
      "epoch": 52.55,
      "grad_norm": 4.0283942222595215,
      "learning_rate": 4.7447447447447454e-06,
      "loss": 0.4025,
      "step": 17500
    },
    {
      "epoch": 52.58,
      "grad_norm": 5.01807975769043,
      "learning_rate": 4.741741741741742e-06,
      "loss": 0.4314,
      "step": 17510
    },
    {
      "epoch": 52.61,
      "grad_norm": 5.700190544128418,
      "learning_rate": 4.738738738738739e-06,
      "loss": 0.3893,
      "step": 17520
    },
    {
      "epoch": 52.64,
      "grad_norm": 4.133469104766846,
      "learning_rate": 4.735735735735736e-06,
      "loss": 0.4437,
      "step": 17530
    },
    {
      "epoch": 52.67,
      "grad_norm": 4.3023834228515625,
      "learning_rate": 4.732732732732734e-06,
      "loss": 0.4063,
      "step": 17540
    },
    {
      "epoch": 52.7,
      "grad_norm": 5.060369491577148,
      "learning_rate": 4.72972972972973e-06,
      "loss": 0.4281,
      "step": 17550
    },
    {
      "epoch": 52.73,
      "grad_norm": 4.311286449432373,
      "learning_rate": 4.726726726726727e-06,
      "loss": 0.4023,
      "step": 17560
    },
    {
      "epoch": 52.76,
      "grad_norm": 5.8158159255981445,
      "learning_rate": 4.723723723723724e-06,
      "loss": 0.3718,
      "step": 17570
    },
    {
      "epoch": 52.79,
      "grad_norm": 4.747592926025391,
      "learning_rate": 4.720720720720721e-06,
      "loss": 0.4315,
      "step": 17580
    },
    {
      "epoch": 52.82,
      "grad_norm": 4.223723411560059,
      "learning_rate": 4.717717717717718e-06,
      "loss": 0.4123,
      "step": 17590
    },
    {
      "epoch": 52.85,
      "grad_norm": 5.022419452667236,
      "learning_rate": 4.714714714714715e-06,
      "loss": 0.4278,
      "step": 17600
    },
    {
      "epoch": 52.88,
      "grad_norm": 5.707874774932861,
      "learning_rate": 4.711711711711713e-06,
      "loss": 0.416,
      "step": 17610
    },
    {
      "epoch": 52.91,
      "grad_norm": 4.26486349105835,
      "learning_rate": 4.708708708708709e-06,
      "loss": 0.4183,
      "step": 17620
    },
    {
      "epoch": 52.94,
      "grad_norm": 5.666004657745361,
      "learning_rate": 4.705705705705706e-06,
      "loss": 0.4443,
      "step": 17630
    },
    {
      "epoch": 52.97,
      "grad_norm": 5.0734639167785645,
      "learning_rate": 4.702702702702703e-06,
      "loss": 0.453,
      "step": 17640
    },
    {
      "epoch": 53.0,
      "eval_accuracy": 0.9188,
      "eval_loss": 0.3059512674808502,
      "eval_runtime": 26.0948,
      "eval_samples_per_second": 383.218,
      "eval_steps_per_second": 1.533,
      "step": 17649
    },
    {
      "epoch": 53.0,
      "grad_norm": 3.7070770263671875,
      "learning_rate": 4.6996996996997e-06,
      "loss": 0.4156,
      "step": 17650
    },
    {
      "epoch": 53.03,
      "grad_norm": 3.19081711769104,
      "learning_rate": 4.696696696696697e-06,
      "loss": 0.3777,
      "step": 17660
    },
    {
      "epoch": 53.06,
      "grad_norm": 3.4403932094573975,
      "learning_rate": 4.693693693693694e-06,
      "loss": 0.4468,
      "step": 17670
    },
    {
      "epoch": 53.09,
      "grad_norm": 7.437161445617676,
      "learning_rate": 4.6906906906906916e-06,
      "loss": 0.4246,
      "step": 17680
    },
    {
      "epoch": 53.12,
      "grad_norm": 5.397223949432373,
      "learning_rate": 4.687687687687688e-06,
      "loss": 0.354,
      "step": 17690
    },
    {
      "epoch": 53.15,
      "grad_norm": 4.8160014152526855,
      "learning_rate": 4.684684684684685e-06,
      "loss": 0.3791,
      "step": 17700
    },
    {
      "epoch": 53.18,
      "grad_norm": 5.656338691711426,
      "learning_rate": 4.6816816816816815e-06,
      "loss": 0.4331,
      "step": 17710
    },
    {
      "epoch": 53.21,
      "grad_norm": 3.160717725753784,
      "learning_rate": 4.678678678678679e-06,
      "loss": 0.4241,
      "step": 17720
    },
    {
      "epoch": 53.24,
      "grad_norm": 3.6100687980651855,
      "learning_rate": 4.675675675675676e-06,
      "loss": 0.4037,
      "step": 17730
    },
    {
      "epoch": 53.27,
      "grad_norm": 4.517573356628418,
      "learning_rate": 4.672672672672673e-06,
      "loss": 0.4153,
      "step": 17740
    },
    {
      "epoch": 53.3,
      "grad_norm": 3.7838268280029297,
      "learning_rate": 4.6696696696696705e-06,
      "loss": 0.4043,
      "step": 17750
    },
    {
      "epoch": 53.33,
      "grad_norm": 5.175099849700928,
      "learning_rate": 4.666666666666667e-06,
      "loss": 0.42,
      "step": 17760
    },
    {
      "epoch": 53.36,
      "grad_norm": 4.6000542640686035,
      "learning_rate": 4.663663663663664e-06,
      "loss": 0.4343,
      "step": 17770
    },
    {
      "epoch": 53.39,
      "grad_norm": 5.372957706451416,
      "learning_rate": 4.6606606606606604e-06,
      "loss": 0.4032,
      "step": 17780
    },
    {
      "epoch": 53.42,
      "grad_norm": 7.259326457977295,
      "learning_rate": 4.657657657657658e-06,
      "loss": 0.4354,
      "step": 17790
    },
    {
      "epoch": 53.45,
      "grad_norm": 5.637961387634277,
      "learning_rate": 4.654654654654655e-06,
      "loss": 0.4492,
      "step": 17800
    },
    {
      "epoch": 53.48,
      "grad_norm": 4.004184722900391,
      "learning_rate": 4.651651651651652e-06,
      "loss": 0.435,
      "step": 17810
    },
    {
      "epoch": 53.51,
      "grad_norm": 3.5187957286834717,
      "learning_rate": 4.6486486486486495e-06,
      "loss": 0.428,
      "step": 17820
    },
    {
      "epoch": 53.54,
      "grad_norm": 5.119517803192139,
      "learning_rate": 4.645645645645646e-06,
      "loss": 0.4058,
      "step": 17830
    },
    {
      "epoch": 53.57,
      "grad_norm": 4.222315788269043,
      "learning_rate": 4.642642642642643e-06,
      "loss": 0.4306,
      "step": 17840
    },
    {
      "epoch": 53.6,
      "grad_norm": 5.206671714782715,
      "learning_rate": 4.63963963963964e-06,
      "loss": 0.4338,
      "step": 17850
    },
    {
      "epoch": 53.63,
      "grad_norm": 6.530751705169678,
      "learning_rate": 4.636636636636637e-06,
      "loss": 0.3725,
      "step": 17860
    },
    {
      "epoch": 53.66,
      "grad_norm": 5.021688461303711,
      "learning_rate": 4.633633633633634e-06,
      "loss": 0.4255,
      "step": 17870
    },
    {
      "epoch": 53.69,
      "grad_norm": 6.446868896484375,
      "learning_rate": 4.630630630630631e-06,
      "loss": 0.4812,
      "step": 17880
    },
    {
      "epoch": 53.72,
      "grad_norm": 4.0734148025512695,
      "learning_rate": 4.627627627627628e-06,
      "loss": 0.4121,
      "step": 17890
    },
    {
      "epoch": 53.75,
      "grad_norm": 4.178910255432129,
      "learning_rate": 4.624624624624625e-06,
      "loss": 0.3602,
      "step": 17900
    },
    {
      "epoch": 53.78,
      "grad_norm": 4.327620506286621,
      "learning_rate": 4.621621621621622e-06,
      "loss": 0.4196,
      "step": 17910
    },
    {
      "epoch": 53.81,
      "grad_norm": 6.962294101715088,
      "learning_rate": 4.618618618618619e-06,
      "loss": 0.3978,
      "step": 17920
    },
    {
      "epoch": 53.84,
      "grad_norm": 2.773702621459961,
      "learning_rate": 4.615615615615616e-06,
      "loss": 0.4055,
      "step": 17930
    },
    {
      "epoch": 53.87,
      "grad_norm": 4.743677616119385,
      "learning_rate": 4.612612612612613e-06,
      "loss": 0.3859,
      "step": 17940
    },
    {
      "epoch": 53.9,
      "grad_norm": 4.772582530975342,
      "learning_rate": 4.60960960960961e-06,
      "loss": 0.3891,
      "step": 17950
    },
    {
      "epoch": 53.93,
      "grad_norm": 9.191411972045898,
      "learning_rate": 4.6066066066066066e-06,
      "loss": 0.4126,
      "step": 17960
    },
    {
      "epoch": 53.96,
      "grad_norm": 4.2372331619262695,
      "learning_rate": 4.603603603603604e-06,
      "loss": 0.4416,
      "step": 17970
    },
    {
      "epoch": 53.99,
      "grad_norm": 3.6780498027801514,
      "learning_rate": 4.600600600600601e-06,
      "loss": 0.3802,
      "step": 17980
    },
    {
      "epoch": 54.0,
      "eval_accuracy": 0.9204,
      "eval_loss": 0.3046325147151947,
      "eval_runtime": 26.4568,
      "eval_samples_per_second": 377.975,
      "eval_steps_per_second": 1.512,
      "step": 17982
    },
    {
      "epoch": 54.02,
      "grad_norm": 3.395017147064209,
      "learning_rate": 4.597597597597598e-06,
      "loss": 0.3073,
      "step": 17990
    },
    {
      "epoch": 54.05,
      "grad_norm": 4.218894004821777,
      "learning_rate": 4.594594594594596e-06,
      "loss": 0.4382,
      "step": 18000
    },
    {
      "epoch": 54.08,
      "grad_norm": 4.327070713043213,
      "learning_rate": 4.591591591591592e-06,
      "loss": 0.4511,
      "step": 18010
    },
    {
      "epoch": 54.11,
      "grad_norm": 3.97296404838562,
      "learning_rate": 4.588588588588589e-06,
      "loss": 0.3913,
      "step": 18020
    },
    {
      "epoch": 54.14,
      "grad_norm": 4.62940788269043,
      "learning_rate": 4.5855855855855855e-06,
      "loss": 0.3654,
      "step": 18030
    },
    {
      "epoch": 54.17,
      "grad_norm": 4.676243305206299,
      "learning_rate": 4.582582582582583e-06,
      "loss": 0.4074,
      "step": 18040
    },
    {
      "epoch": 54.2,
      "grad_norm": 4.673247814178467,
      "learning_rate": 4.57957957957958e-06,
      "loss": 0.4076,
      "step": 18050
    },
    {
      "epoch": 54.23,
      "grad_norm": 4.308940887451172,
      "learning_rate": 4.576576576576577e-06,
      "loss": 0.4358,
      "step": 18060
    },
    {
      "epoch": 54.26,
      "grad_norm": 11.122332572937012,
      "learning_rate": 4.573573573573574e-06,
      "loss": 0.366,
      "step": 18070
    },
    {
      "epoch": 54.29,
      "grad_norm": 3.8195879459381104,
      "learning_rate": 4.570570570570571e-06,
      "loss": 0.4293,
      "step": 18080
    },
    {
      "epoch": 54.32,
      "grad_norm": 4.0169830322265625,
      "learning_rate": 4.567567567567568e-06,
      "loss": 0.4028,
      "step": 18090
    },
    {
      "epoch": 54.35,
      "grad_norm": 8.181219100952148,
      "learning_rate": 4.5645645645645645e-06,
      "loss": 0.4517,
      "step": 18100
    },
    {
      "epoch": 54.38,
      "grad_norm": 4.1873016357421875,
      "learning_rate": 4.561561561561562e-06,
      "loss": 0.4084,
      "step": 18110
    },
    {
      "epoch": 54.41,
      "grad_norm": 4.441383361816406,
      "learning_rate": 4.558558558558559e-06,
      "loss": 0.4029,
      "step": 18120
    },
    {
      "epoch": 54.44,
      "grad_norm": 4.467209815979004,
      "learning_rate": 4.555555555555556e-06,
      "loss": 0.4168,
      "step": 18130
    },
    {
      "epoch": 54.47,
      "grad_norm": 4.031890392303467,
      "learning_rate": 4.552552552552553e-06,
      "loss": 0.4716,
      "step": 18140
    },
    {
      "epoch": 54.5,
      "grad_norm": 4.344508171081543,
      "learning_rate": 4.54954954954955e-06,
      "loss": 0.3914,
      "step": 18150
    },
    {
      "epoch": 54.53,
      "grad_norm": 4.170722484588623,
      "learning_rate": 4.546546546546547e-06,
      "loss": 0.3636,
      "step": 18160
    },
    {
      "epoch": 54.56,
      "grad_norm": 3.257127046585083,
      "learning_rate": 4.5435435435435435e-06,
      "loss": 0.4121,
      "step": 18170
    },
    {
      "epoch": 54.59,
      "grad_norm": 7.440723896026611,
      "learning_rate": 4.540540540540541e-06,
      "loss": 0.3936,
      "step": 18180
    },
    {
      "epoch": 54.62,
      "grad_norm": 5.843325138092041,
      "learning_rate": 4.537537537537538e-06,
      "loss": 0.3665,
      "step": 18190
    },
    {
      "epoch": 54.65,
      "grad_norm": 4.213897705078125,
      "learning_rate": 4.534534534534535e-06,
      "loss": 0.3795,
      "step": 18200
    },
    {
      "epoch": 54.68,
      "grad_norm": 7.303389072418213,
      "learning_rate": 4.531531531531532e-06,
      "loss": 0.4491,
      "step": 18210
    },
    {
      "epoch": 54.71,
      "grad_norm": 4.540135383605957,
      "learning_rate": 4.528528528528529e-06,
      "loss": 0.4167,
      "step": 18220
    },
    {
      "epoch": 54.74,
      "grad_norm": 4.910680294036865,
      "learning_rate": 4.525525525525526e-06,
      "loss": 0.4855,
      "step": 18230
    },
    {
      "epoch": 54.77,
      "grad_norm": 4.308034896850586,
      "learning_rate": 4.522522522522522e-06,
      "loss": 0.379,
      "step": 18240
    },
    {
      "epoch": 54.8,
      "grad_norm": 4.407633304595947,
      "learning_rate": 4.51951951951952e-06,
      "loss": 0.3795,
      "step": 18250
    },
    {
      "epoch": 54.83,
      "grad_norm": 3.276524305343628,
      "learning_rate": 4.516516516516517e-06,
      "loss": 0.3852,
      "step": 18260
    },
    {
      "epoch": 54.86,
      "grad_norm": 4.447195053100586,
      "learning_rate": 4.513513513513514e-06,
      "loss": 0.3591,
      "step": 18270
    },
    {
      "epoch": 54.89,
      "grad_norm": 5.28013801574707,
      "learning_rate": 4.510510510510511e-06,
      "loss": 0.4406,
      "step": 18280
    },
    {
      "epoch": 54.92,
      "grad_norm": 5.964004039764404,
      "learning_rate": 4.507507507507508e-06,
      "loss": 0.4264,
      "step": 18290
    },
    {
      "epoch": 54.95,
      "grad_norm": 4.449484348297119,
      "learning_rate": 4.504504504504505e-06,
      "loss": 0.413,
      "step": 18300
    },
    {
      "epoch": 54.98,
      "grad_norm": 3.826570749282837,
      "learning_rate": 4.501501501501501e-06,
      "loss": 0.4191,
      "step": 18310
    },
    {
      "epoch": 55.0,
      "eval_accuracy": 0.9208,
      "eval_loss": 0.3075259327888489,
      "eval_runtime": 26.7638,
      "eval_samples_per_second": 373.639,
      "eval_steps_per_second": 1.495,
      "step": 18315
    },
    {
      "epoch": 55.02,
      "grad_norm": 4.403299808502197,
      "learning_rate": 4.498498498498499e-06,
      "loss": 0.3552,
      "step": 18320
    },
    {
      "epoch": 55.05,
      "grad_norm": 8.051570892333984,
      "learning_rate": 4.495495495495496e-06,
      "loss": 0.4302,
      "step": 18330
    },
    {
      "epoch": 55.08,
      "grad_norm": 3.9226596355438232,
      "learning_rate": 4.492492492492493e-06,
      "loss": 0.4126,
      "step": 18340
    },
    {
      "epoch": 55.11,
      "grad_norm": 5.262730598449707,
      "learning_rate": 4.48948948948949e-06,
      "loss": 0.436,
      "step": 18350
    },
    {
      "epoch": 55.14,
      "grad_norm": 3.15976881980896,
      "learning_rate": 4.486486486486487e-06,
      "loss": 0.4179,
      "step": 18360
    },
    {
      "epoch": 55.17,
      "grad_norm": 2.6332197189331055,
      "learning_rate": 4.483483483483484e-06,
      "loss": 0.4359,
      "step": 18370
    },
    {
      "epoch": 55.2,
      "grad_norm": 6.135165691375732,
      "learning_rate": 4.480480480480481e-06,
      "loss": 0.3587,
      "step": 18380
    },
    {
      "epoch": 55.23,
      "grad_norm": 3.679643392562866,
      "learning_rate": 4.477477477477478e-06,
      "loss": 0.41,
      "step": 18390
    },
    {
      "epoch": 55.26,
      "grad_norm": 7.925889492034912,
      "learning_rate": 4.474474474474475e-06,
      "loss": 0.3844,
      "step": 18400
    },
    {
      "epoch": 55.29,
      "grad_norm": 6.3421831130981445,
      "learning_rate": 4.471471471471472e-06,
      "loss": 0.3874,
      "step": 18410
    },
    {
      "epoch": 55.32,
      "grad_norm": 4.955140113830566,
      "learning_rate": 4.4684684684684686e-06,
      "loss": 0.3495,
      "step": 18420
    },
    {
      "epoch": 55.35,
      "grad_norm": 4.214719295501709,
      "learning_rate": 4.465465465465465e-06,
      "loss": 0.4247,
      "step": 18430
    },
    {
      "epoch": 55.38,
      "grad_norm": 4.887331008911133,
      "learning_rate": 4.462462462462463e-06,
      "loss": 0.4179,
      "step": 18440
    },
    {
      "epoch": 55.41,
      "grad_norm": 4.696966171264648,
      "learning_rate": 4.45945945945946e-06,
      "loss": 0.4352,
      "step": 18450
    },
    {
      "epoch": 55.44,
      "grad_norm": 4.691727161407471,
      "learning_rate": 4.456456456456457e-06,
      "loss": 0.3705,
      "step": 18460
    },
    {
      "epoch": 55.47,
      "grad_norm": 5.364315509796143,
      "learning_rate": 4.453453453453454e-06,
      "loss": 0.3185,
      "step": 18470
    },
    {
      "epoch": 55.5,
      "grad_norm": 2.1761608123779297,
      "learning_rate": 4.450450450450451e-06,
      "loss": 0.3757,
      "step": 18480
    },
    {
      "epoch": 55.53,
      "grad_norm": 4.898319721221924,
      "learning_rate": 4.4474474474474475e-06,
      "loss": 0.4101,
      "step": 18490
    },
    {
      "epoch": 55.56,
      "grad_norm": 4.770262718200684,
      "learning_rate": 4.444444444444444e-06,
      "loss": 0.4004,
      "step": 18500
    },
    {
      "epoch": 55.59,
      "grad_norm": 3.4409995079040527,
      "learning_rate": 4.441441441441442e-06,
      "loss": 0.3734,
      "step": 18510
    },
    {
      "epoch": 55.62,
      "grad_norm": 3.843642234802246,
      "learning_rate": 4.438438438438439e-06,
      "loss": 0.3648,
      "step": 18520
    },
    {
      "epoch": 55.65,
      "grad_norm": 4.793038368225098,
      "learning_rate": 4.435435435435436e-06,
      "loss": 0.4352,
      "step": 18530
    },
    {
      "epoch": 55.68,
      "grad_norm": 4.099161624908447,
      "learning_rate": 4.432432432432433e-06,
      "loss": 0.3932,
      "step": 18540
    },
    {
      "epoch": 55.71,
      "grad_norm": 4.231556415557861,
      "learning_rate": 4.42942942942943e-06,
      "loss": 0.4103,
      "step": 18550
    },
    {
      "epoch": 55.74,
      "grad_norm": 3.6762075424194336,
      "learning_rate": 4.4264264264264265e-06,
      "loss": 0.4349,
      "step": 18560
    },
    {
      "epoch": 55.77,
      "grad_norm": 5.304360866546631,
      "learning_rate": 4.423423423423424e-06,
      "loss": 0.4004,
      "step": 18570
    },
    {
      "epoch": 55.8,
      "grad_norm": 3.83604097366333,
      "learning_rate": 4.420420420420421e-06,
      "loss": 0.369,
      "step": 18580
    },
    {
      "epoch": 55.83,
      "grad_norm": 3.783693313598633,
      "learning_rate": 4.417417417417418e-06,
      "loss": 0.3964,
      "step": 18590
    },
    {
      "epoch": 55.86,
      "grad_norm": 3.9015114307403564,
      "learning_rate": 4.414414414414415e-06,
      "loss": 0.3562,
      "step": 18600
    },
    {
      "epoch": 55.89,
      "grad_norm": 3.605548143386841,
      "learning_rate": 4.411411411411412e-06,
      "loss": 0.4105,
      "step": 18610
    },
    {
      "epoch": 55.92,
      "grad_norm": 4.916491985321045,
      "learning_rate": 4.408408408408409e-06,
      "loss": 0.3875,
      "step": 18620
    },
    {
      "epoch": 55.95,
      "grad_norm": 3.5518367290496826,
      "learning_rate": 4.4054054054054054e-06,
      "loss": 0.4323,
      "step": 18630
    },
    {
      "epoch": 55.98,
      "grad_norm": 6.865616798400879,
      "learning_rate": 4.402402402402403e-06,
      "loss": 0.4245,
      "step": 18640
    },
    {
      "epoch": 56.0,
      "eval_accuracy": 0.9205,
      "eval_loss": 0.3018031120300293,
      "eval_runtime": 26.3866,
      "eval_samples_per_second": 378.98,
      "eval_steps_per_second": 1.516,
      "step": 18648
    },
    {
      "epoch": 56.01,
      "grad_norm": 4.466795444488525,
      "learning_rate": 4.3993993993993996e-06,
      "loss": 0.4064,
      "step": 18650
    },
    {
      "epoch": 56.04,
      "grad_norm": 5.5253753662109375,
      "learning_rate": 4.396396396396397e-06,
      "loss": 0.4065,
      "step": 18660
    },
    {
      "epoch": 56.07,
      "grad_norm": 6.008886337280273,
      "learning_rate": 4.393393393393394e-06,
      "loss": 0.3712,
      "step": 18670
    },
    {
      "epoch": 56.1,
      "grad_norm": 4.467586517333984,
      "learning_rate": 4.39039039039039e-06,
      "loss": 0.3658,
      "step": 18680
    },
    {
      "epoch": 56.13,
      "grad_norm": 4.632496356964111,
      "learning_rate": 4.387387387387388e-06,
      "loss": 0.406,
      "step": 18690
    },
    {
      "epoch": 56.16,
      "grad_norm": 4.244455814361572,
      "learning_rate": 4.384384384384384e-06,
      "loss": 0.4466,
      "step": 18700
    },
    {
      "epoch": 56.19,
      "grad_norm": 4.02488899230957,
      "learning_rate": 4.381381381381382e-06,
      "loss": 0.387,
      "step": 18710
    },
    {
      "epoch": 56.22,
      "grad_norm": 7.864763259887695,
      "learning_rate": 4.378378378378379e-06,
      "loss": 0.442,
      "step": 18720
    },
    {
      "epoch": 56.25,
      "grad_norm": 5.261714935302734,
      "learning_rate": 4.375375375375376e-06,
      "loss": 0.402,
      "step": 18730
    },
    {
      "epoch": 56.28,
      "grad_norm": 6.869850158691406,
      "learning_rate": 4.372372372372373e-06,
      "loss": 0.3671,
      "step": 18740
    },
    {
      "epoch": 56.31,
      "grad_norm": 4.165491580963135,
      "learning_rate": 4.369369369369369e-06,
      "loss": 0.3984,
      "step": 18750
    },
    {
      "epoch": 56.34,
      "grad_norm": 4.118197441101074,
      "learning_rate": 4.366366366366367e-06,
      "loss": 0.3832,
      "step": 18760
    },
    {
      "epoch": 56.37,
      "grad_norm": 3.9226462841033936,
      "learning_rate": 4.363363363363363e-06,
      "loss": 0.4047,
      "step": 18770
    },
    {
      "epoch": 56.4,
      "grad_norm": 3.613295793533325,
      "learning_rate": 4.360360360360361e-06,
      "loss": 0.3899,
      "step": 18780
    },
    {
      "epoch": 56.43,
      "grad_norm": 3.760272979736328,
      "learning_rate": 4.357357357357358e-06,
      "loss": 0.3946,
      "step": 18790
    },
    {
      "epoch": 56.46,
      "grad_norm": 5.6173906326293945,
      "learning_rate": 4.354354354354355e-06,
      "loss": 0.4291,
      "step": 18800
    },
    {
      "epoch": 56.49,
      "grad_norm": 5.3471269607543945,
      "learning_rate": 4.351351351351352e-06,
      "loss": 0.3556,
      "step": 18810
    },
    {
      "epoch": 56.52,
      "grad_norm": 4.534874439239502,
      "learning_rate": 4.348348348348348e-06,
      "loss": 0.3929,
      "step": 18820
    },
    {
      "epoch": 56.55,
      "grad_norm": 4.405369758605957,
      "learning_rate": 4.345345345345346e-06,
      "loss": 0.3755,
      "step": 18830
    },
    {
      "epoch": 56.58,
      "grad_norm": 5.081787109375,
      "learning_rate": 4.342342342342343e-06,
      "loss": 0.3465,
      "step": 18840
    },
    {
      "epoch": 56.61,
      "grad_norm": 4.178293228149414,
      "learning_rate": 4.33933933933934e-06,
      "loss": 0.3888,
      "step": 18850
    },
    {
      "epoch": 56.64,
      "grad_norm": 5.368361949920654,
      "learning_rate": 4.3363363363363364e-06,
      "loss": 0.3992,
      "step": 18860
    },
    {
      "epoch": 56.67,
      "grad_norm": 4.474520206451416,
      "learning_rate": 4.333333333333334e-06,
      "loss": 0.4155,
      "step": 18870
    },
    {
      "epoch": 56.7,
      "grad_norm": 10.38321304321289,
      "learning_rate": 4.3303303303303305e-06,
      "loss": 0.3626,
      "step": 18880
    },
    {
      "epoch": 56.73,
      "grad_norm": 6.310691833496094,
      "learning_rate": 4.327327327327327e-06,
      "loss": 0.4271,
      "step": 18890
    },
    {
      "epoch": 56.76,
      "grad_norm": 3.04647159576416,
      "learning_rate": 4.324324324324325e-06,
      "loss": 0.3403,
      "step": 18900
    },
    {
      "epoch": 56.79,
      "grad_norm": 4.11697244644165,
      "learning_rate": 4.321321321321322e-06,
      "loss": 0.3962,
      "step": 18910
    },
    {
      "epoch": 56.82,
      "grad_norm": 4.304756164550781,
      "learning_rate": 4.318318318318319e-06,
      "loss": 0.3943,
      "step": 18920
    },
    {
      "epoch": 56.85,
      "grad_norm": 2.9814178943634033,
      "learning_rate": 4.315315315315315e-06,
      "loss": 0.3761,
      "step": 18930
    },
    {
      "epoch": 56.88,
      "grad_norm": 6.85025691986084,
      "learning_rate": 4.312312312312313e-06,
      "loss": 0.369,
      "step": 18940
    },
    {
      "epoch": 56.91,
      "grad_norm": 2.926481008529663,
      "learning_rate": 4.3093093093093095e-06,
      "loss": 0.3891,
      "step": 18950
    },
    {
      "epoch": 56.94,
      "grad_norm": 4.791072845458984,
      "learning_rate": 4.306306306306306e-06,
      "loss": 0.4335,
      "step": 18960
    },
    {
      "epoch": 56.97,
      "grad_norm": 3.743326425552368,
      "learning_rate": 4.303303303303304e-06,
      "loss": 0.406,
      "step": 18970
    },
    {
      "epoch": 57.0,
      "grad_norm": 4.9799485206604,
      "learning_rate": 4.300300300300301e-06,
      "loss": 0.4356,
      "step": 18980
    },
    {
      "epoch": 57.0,
      "eval_accuracy": 0.9214,
      "eval_loss": 0.3033043444156647,
      "eval_runtime": 26.4504,
      "eval_samples_per_second": 378.065,
      "eval_steps_per_second": 1.512,
      "step": 18981
    },
    {
      "epoch": 57.03,
      "grad_norm": 3.222034215927124,
      "learning_rate": 4.297297297297298e-06,
      "loss": 0.4944,
      "step": 18990
    },
    {
      "epoch": 57.06,
      "grad_norm": 5.237534999847412,
      "learning_rate": 4.294294294294294e-06,
      "loss": 0.427,
      "step": 19000
    },
    {
      "epoch": 57.09,
      "grad_norm": 5.138498306274414,
      "learning_rate": 4.291291291291292e-06,
      "loss": 0.4181,
      "step": 19010
    },
    {
      "epoch": 57.12,
      "grad_norm": 4.010339260101318,
      "learning_rate": 4.2882882882882885e-06,
      "loss": 0.407,
      "step": 19020
    },
    {
      "epoch": 57.15,
      "grad_norm": 4.0047101974487305,
      "learning_rate": 4.285285285285286e-06,
      "loss": 0.377,
      "step": 19030
    },
    {
      "epoch": 57.18,
      "grad_norm": 3.8277862071990967,
      "learning_rate": 4.282282282282283e-06,
      "loss": 0.4136,
      "step": 19040
    },
    {
      "epoch": 57.21,
      "grad_norm": 3.9843053817749023,
      "learning_rate": 4.27927927927928e-06,
      "loss": 0.3595,
      "step": 19050
    },
    {
      "epoch": 57.24,
      "grad_norm": 4.864630222320557,
      "learning_rate": 4.276276276276277e-06,
      "loss": 0.3986,
      "step": 19060
    },
    {
      "epoch": 57.27,
      "grad_norm": 6.472599506378174,
      "learning_rate": 4.273273273273273e-06,
      "loss": 0.4146,
      "step": 19070
    },
    {
      "epoch": 57.3,
      "grad_norm": 4.172643184661865,
      "learning_rate": 4.270270270270271e-06,
      "loss": 0.381,
      "step": 19080
    },
    {
      "epoch": 57.33,
      "grad_norm": 3.9899628162384033,
      "learning_rate": 4.2672672672672674e-06,
      "loss": 0.4342,
      "step": 19090
    },
    {
      "epoch": 57.36,
      "grad_norm": 3.7790169715881348,
      "learning_rate": 4.264264264264265e-06,
      "loss": 0.4047,
      "step": 19100
    },
    {
      "epoch": 57.39,
      "grad_norm": 4.043527603149414,
      "learning_rate": 4.2612612612612615e-06,
      "loss": 0.3708,
      "step": 19110
    },
    {
      "epoch": 57.42,
      "grad_norm": 3.6510426998138428,
      "learning_rate": 4.258258258258259e-06,
      "loss": 0.3739,
      "step": 19120
    },
    {
      "epoch": 57.45,
      "grad_norm": 4.161368370056152,
      "learning_rate": 4.255255255255256e-06,
      "loss": 0.3728,
      "step": 19130
    },
    {
      "epoch": 57.48,
      "grad_norm": 4.108691692352295,
      "learning_rate": 4.252252252252252e-06,
      "loss": 0.3979,
      "step": 19140
    },
    {
      "epoch": 57.51,
      "grad_norm": 3.611224412918091,
      "learning_rate": 4.24924924924925e-06,
      "loss": 0.3142,
      "step": 19150
    },
    {
      "epoch": 57.54,
      "grad_norm": 4.055614948272705,
      "learning_rate": 4.246246246246246e-06,
      "loss": 0.4365,
      "step": 19160
    },
    {
      "epoch": 57.57,
      "grad_norm": 4.025052070617676,
      "learning_rate": 4.243243243243244e-06,
      "loss": 0.4351,
      "step": 19170
    },
    {
      "epoch": 57.6,
      "grad_norm": 3.274895429611206,
      "learning_rate": 4.2402402402402405e-06,
      "loss": 0.3932,
      "step": 19180
    },
    {
      "epoch": 57.63,
      "grad_norm": 4.449277877807617,
      "learning_rate": 4.237237237237238e-06,
      "loss": 0.4448,
      "step": 19190
    },
    {
      "epoch": 57.66,
      "grad_norm": 3.8885586261749268,
      "learning_rate": 4.234234234234235e-06,
      "loss": 0.4139,
      "step": 19200
    },
    {
      "epoch": 57.69,
      "grad_norm": 2.935800552368164,
      "learning_rate": 4.231231231231231e-06,
      "loss": 0.4289,
      "step": 19210
    },
    {
      "epoch": 57.72,
      "grad_norm": 3.8989992141723633,
      "learning_rate": 4.228228228228229e-06,
      "loss": 0.4139,
      "step": 19220
    },
    {
      "epoch": 57.75,
      "grad_norm": 5.179030418395996,
      "learning_rate": 4.225225225225225e-06,
      "loss": 0.4202,
      "step": 19230
    },
    {
      "epoch": 57.78,
      "grad_norm": 4.942713260650635,
      "learning_rate": 4.222222222222223e-06,
      "loss": 0.3589,
      "step": 19240
    },
    {
      "epoch": 57.81,
      "grad_norm": 8.229090690612793,
      "learning_rate": 4.2192192192192195e-06,
      "loss": 0.4527,
      "step": 19250
    },
    {
      "epoch": 57.84,
      "grad_norm": 5.07041072845459,
      "learning_rate": 4.216216216216217e-06,
      "loss": 0.3857,
      "step": 19260
    },
    {
      "epoch": 57.87,
      "grad_norm": 4.575139999389648,
      "learning_rate": 4.2132132132132136e-06,
      "loss": 0.3958,
      "step": 19270
    },
    {
      "epoch": 57.9,
      "grad_norm": 4.14766788482666,
      "learning_rate": 4.21021021021021e-06,
      "loss": 0.4161,
      "step": 19280
    },
    {
      "epoch": 57.93,
      "grad_norm": 3.927588939666748,
      "learning_rate": 4.207207207207208e-06,
      "loss": 0.386,
      "step": 19290
    },
    {
      "epoch": 57.96,
      "grad_norm": 4.351321220397949,
      "learning_rate": 4.204204204204204e-06,
      "loss": 0.3483,
      "step": 19300
    },
    {
      "epoch": 57.99,
      "grad_norm": 4.320892333984375,
      "learning_rate": 4.201201201201202e-06,
      "loss": 0.348,
      "step": 19310
    },
    {
      "epoch": 58.0,
      "eval_accuracy": 0.9208,
      "eval_loss": 0.308114230632782,
      "eval_runtime": 26.3632,
      "eval_samples_per_second": 379.316,
      "eval_steps_per_second": 1.517,
      "step": 19314
    },
    {
      "epoch": 58.02,
      "grad_norm": 3.2385594844818115,
      "learning_rate": 4.1981981981981984e-06,
      "loss": 0.4134,
      "step": 19320
    },
    {
      "epoch": 58.05,
      "grad_norm": 5.3496904373168945,
      "learning_rate": 4.195195195195196e-06,
      "loss": 0.3957,
      "step": 19330
    },
    {
      "epoch": 58.08,
      "grad_norm": 6.157802581787109,
      "learning_rate": 4.1921921921921925e-06,
      "loss": 0.4253,
      "step": 19340
    },
    {
      "epoch": 58.11,
      "grad_norm": 3.8979568481445312,
      "learning_rate": 4.189189189189189e-06,
      "loss": 0.4491,
      "step": 19350
    },
    {
      "epoch": 58.14,
      "grad_norm": 3.409968852996826,
      "learning_rate": 4.186186186186187e-06,
      "loss": 0.3825,
      "step": 19360
    },
    {
      "epoch": 58.17,
      "grad_norm": 5.369346618652344,
      "learning_rate": 4.183183183183184e-06,
      "loss": 0.3711,
      "step": 19370
    },
    {
      "epoch": 58.2,
      "grad_norm": 4.344148635864258,
      "learning_rate": 4.180180180180181e-06,
      "loss": 0.4401,
      "step": 19380
    },
    {
      "epoch": 58.23,
      "grad_norm": 3.75290846824646,
      "learning_rate": 4.177177177177177e-06,
      "loss": 0.4487,
      "step": 19390
    },
    {
      "epoch": 58.26,
      "grad_norm": 3.2389259338378906,
      "learning_rate": 4.174174174174174e-06,
      "loss": 0.3968,
      "step": 19400
    },
    {
      "epoch": 58.29,
      "grad_norm": 4.891212463378906,
      "learning_rate": 4.1711711711711715e-06,
      "loss": 0.4031,
      "step": 19410
    },
    {
      "epoch": 58.32,
      "grad_norm": 5.938599109649658,
      "learning_rate": 4.168168168168168e-06,
      "loss": 0.3893,
      "step": 19420
    },
    {
      "epoch": 58.35,
      "grad_norm": 3.844050645828247,
      "learning_rate": 4.165165165165166e-06,
      "loss": 0.418,
      "step": 19430
    },
    {
      "epoch": 58.38,
      "grad_norm": 4.659663677215576,
      "learning_rate": 4.162162162162163e-06,
      "loss": 0.3834,
      "step": 19440
    },
    {
      "epoch": 58.41,
      "grad_norm": 3.98610782623291,
      "learning_rate": 4.15915915915916e-06,
      "loss": 0.4415,
      "step": 19450
    },
    {
      "epoch": 58.44,
      "grad_norm": 2.837587356567383,
      "learning_rate": 4.156156156156156e-06,
      "loss": 0.3949,
      "step": 19460
    },
    {
      "epoch": 58.47,
      "grad_norm": 4.824269771575928,
      "learning_rate": 4.153153153153153e-06,
      "loss": 0.3935,
      "step": 19470
    },
    {
      "epoch": 58.5,
      "grad_norm": 4.342452526092529,
      "learning_rate": 4.1501501501501505e-06,
      "loss": 0.3967,
      "step": 19480
    },
    {
      "epoch": 58.53,
      "grad_norm": 4.652122974395752,
      "learning_rate": 4.147147147147147e-06,
      "loss": 0.409,
      "step": 19490
    },
    {
      "epoch": 58.56,
      "grad_norm": 4.575690269470215,
      "learning_rate": 4.1441441441441446e-06,
      "loss": 0.394,
      "step": 19500
    },
    {
      "epoch": 58.59,
      "grad_norm": 3.9230434894561768,
      "learning_rate": 4.141141141141142e-06,
      "loss": 0.3965,
      "step": 19510
    },
    {
      "epoch": 58.62,
      "grad_norm": 4.956518173217773,
      "learning_rate": 4.138138138138139e-06,
      "loss": 0.37,
      "step": 19520
    },
    {
      "epoch": 58.65,
      "grad_norm": 5.198852062225342,
      "learning_rate": 4.135135135135135e-06,
      "loss": 0.3995,
      "step": 19530
    },
    {
      "epoch": 58.68,
      "grad_norm": 3.277341604232788,
      "learning_rate": 4.132132132132132e-06,
      "loss": 0.3914,
      "step": 19540
    },
    {
      "epoch": 58.71,
      "grad_norm": 4.838464736938477,
      "learning_rate": 4.129129129129129e-06,
      "loss": 0.4143,
      "step": 19550
    },
    {
      "epoch": 58.74,
      "grad_norm": 4.8980536460876465,
      "learning_rate": 4.126126126126127e-06,
      "loss": 0.3785,
      "step": 19560
    },
    {
      "epoch": 58.77,
      "grad_norm": 4.953840732574463,
      "learning_rate": 4.1231231231231235e-06,
      "loss": 0.4218,
      "step": 19570
    },
    {
      "epoch": 58.8,
      "grad_norm": 3.5224502086639404,
      "learning_rate": 4.12012012012012e-06,
      "loss": 0.4469,
      "step": 19580
    },
    {
      "epoch": 58.83,
      "grad_norm": 3.90040922164917,
      "learning_rate": 4.117117117117118e-06,
      "loss": 0.4238,
      "step": 19590
    },
    {
      "epoch": 58.86,
      "grad_norm": 5.3171868324279785,
      "learning_rate": 4.114114114114114e-06,
      "loss": 0.4166,
      "step": 19600
    },
    {
      "epoch": 58.89,
      "grad_norm": 6.108798503875732,
      "learning_rate": 4.111111111111111e-06,
      "loss": 0.368,
      "step": 19610
    },
    {
      "epoch": 58.92,
      "grad_norm": 3.7460858821868896,
      "learning_rate": 4.108108108108108e-06,
      "loss": 0.4005,
      "step": 19620
    },
    {
      "epoch": 58.95,
      "grad_norm": 4.233922481536865,
      "learning_rate": 4.105105105105106e-06,
      "loss": 0.3399,
      "step": 19630
    },
    {
      "epoch": 58.98,
      "grad_norm": 7.330603122711182,
      "learning_rate": 4.1021021021021025e-06,
      "loss": 0.4232,
      "step": 19640
    },
    {
      "epoch": 59.0,
      "eval_accuracy": 0.9198,
      "eval_loss": 0.30576327443122864,
      "eval_runtime": 26.0155,
      "eval_samples_per_second": 384.387,
      "eval_steps_per_second": 1.538,
      "step": 19647
    },
    {
      "epoch": 59.01,
      "grad_norm": 3.4500293731689453,
      "learning_rate": 4.099099099099099e-06,
      "loss": 0.422,
      "step": 19650
    },
    {
      "epoch": 59.04,
      "grad_norm": 3.902200937271118,
      "learning_rate": 4.096096096096097e-06,
      "loss": 0.3879,
      "step": 19660
    },
    {
      "epoch": 59.07,
      "grad_norm": 4.767515659332275,
      "learning_rate": 4.093093093093093e-06,
      "loss": 0.4194,
      "step": 19670
    },
    {
      "epoch": 59.1,
      "grad_norm": 4.851142883300781,
      "learning_rate": 4.09009009009009e-06,
      "loss": 0.3939,
      "step": 19680
    },
    {
      "epoch": 59.13,
      "grad_norm": 5.619836807250977,
      "learning_rate": 4.087087087087087e-06,
      "loss": 0.4063,
      "step": 19690
    },
    {
      "epoch": 59.16,
      "grad_norm": 4.552589416503906,
      "learning_rate": 4.084084084084085e-06,
      "loss": 0.3148,
      "step": 19700
    },
    {
      "epoch": 59.19,
      "grad_norm": 5.441330432891846,
      "learning_rate": 4.0810810810810815e-06,
      "loss": 0.3806,
      "step": 19710
    },
    {
      "epoch": 59.22,
      "grad_norm": 5.744096279144287,
      "learning_rate": 4.078078078078078e-06,
      "loss": 0.3804,
      "step": 19720
    },
    {
      "epoch": 59.25,
      "grad_norm": 3.8247387409210205,
      "learning_rate": 4.0750750750750756e-06,
      "loss": 0.3926,
      "step": 19730
    },
    {
      "epoch": 59.28,
      "grad_norm": 5.349419116973877,
      "learning_rate": 4.072072072072072e-06,
      "loss": 0.3489,
      "step": 19740
    },
    {
      "epoch": 59.31,
      "grad_norm": 4.880401611328125,
      "learning_rate": 4.06906906906907e-06,
      "loss": 0.388,
      "step": 19750
    },
    {
      "epoch": 59.34,
      "grad_norm": 5.770273685455322,
      "learning_rate": 4.066066066066066e-06,
      "loss": 0.3571,
      "step": 19760
    },
    {
      "epoch": 59.37,
      "grad_norm": 3.5081915855407715,
      "learning_rate": 4.063063063063064e-06,
      "loss": 0.4065,
      "step": 19770
    },
    {
      "epoch": 59.4,
      "grad_norm": 8.270219802856445,
      "learning_rate": 4.06006006006006e-06,
      "loss": 0.4297,
      "step": 19780
    },
    {
      "epoch": 59.43,
      "grad_norm": 5.676990032196045,
      "learning_rate": 4.057057057057057e-06,
      "loss": 0.4034,
      "step": 19790
    },
    {
      "epoch": 59.46,
      "grad_norm": 7.0398359298706055,
      "learning_rate": 4.0540540540540545e-06,
      "loss": 0.4457,
      "step": 19800
    },
    {
      "epoch": 59.49,
      "grad_norm": 4.01641845703125,
      "learning_rate": 4.051051051051051e-06,
      "loss": 0.3884,
      "step": 19810
    },
    {
      "epoch": 59.52,
      "grad_norm": 3.347503900527954,
      "learning_rate": 4.048048048048049e-06,
      "loss": 0.3658,
      "step": 19820
    },
    {
      "epoch": 59.55,
      "grad_norm": 7.208119869232178,
      "learning_rate": 4.045045045045045e-06,
      "loss": 0.4067,
      "step": 19830
    },
    {
      "epoch": 59.58,
      "grad_norm": 3.601804733276367,
      "learning_rate": 4.042042042042043e-06,
      "loss": 0.3243,
      "step": 19840
    },
    {
      "epoch": 59.61,
      "grad_norm": 5.499717712402344,
      "learning_rate": 4.039039039039039e-06,
      "loss": 0.306,
      "step": 19850
    },
    {
      "epoch": 59.64,
      "grad_norm": 5.648370742797852,
      "learning_rate": 4.036036036036036e-06,
      "loss": 0.3644,
      "step": 19860
    },
    {
      "epoch": 59.67,
      "grad_norm": 3.934347629547119,
      "learning_rate": 4.0330330330330335e-06,
      "loss": 0.3658,
      "step": 19870
    },
    {
      "epoch": 59.7,
      "grad_norm": 6.51852560043335,
      "learning_rate": 4.03003003003003e-06,
      "loss": 0.4437,
      "step": 19880
    },
    {
      "epoch": 59.73,
      "grad_norm": 3.925356149673462,
      "learning_rate": 4.027027027027028e-06,
      "loss": 0.3592,
      "step": 19890
    },
    {
      "epoch": 59.76,
      "grad_norm": 15.933494567871094,
      "learning_rate": 4.024024024024024e-06,
      "loss": 0.3978,
      "step": 19900
    },
    {
      "epoch": 59.79,
      "grad_norm": 3.710503339767456,
      "learning_rate": 4.021021021021022e-06,
      "loss": 0.4313,
      "step": 19910
    },
    {
      "epoch": 59.82,
      "grad_norm": 4.64842414855957,
      "learning_rate": 4.018018018018018e-06,
      "loss": 0.4129,
      "step": 19920
    },
    {
      "epoch": 59.85,
      "grad_norm": 5.790719032287598,
      "learning_rate": 4.015015015015015e-06,
      "loss": 0.4371,
      "step": 19930
    },
    {
      "epoch": 59.88,
      "grad_norm": 4.395689487457275,
      "learning_rate": 4.0120120120120124e-06,
      "loss": 0.3483,
      "step": 19940
    },
    {
      "epoch": 59.91,
      "grad_norm": 3.8104326725006104,
      "learning_rate": 4.009009009009009e-06,
      "loss": 0.4138,
      "step": 19950
    },
    {
      "epoch": 59.94,
      "grad_norm": 3.2261905670166016,
      "learning_rate": 4.0060060060060066e-06,
      "loss": 0.3289,
      "step": 19960
    },
    {
      "epoch": 59.97,
      "grad_norm": 9.400279998779297,
      "learning_rate": 4.003003003003003e-06,
      "loss": 0.458,
      "step": 19970
    },
    {
      "epoch": 60.0,
      "grad_norm": 3.1749424934387207,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.3363,
      "step": 19980
    },
    {
      "epoch": 60.0,
      "eval_accuracy": 0.9195,
      "eval_loss": 0.3065536320209503,
      "eval_runtime": 26.3747,
      "eval_samples_per_second": 379.151,
      "eval_steps_per_second": 1.517,
      "step": 19980
    },
    {
      "epoch": 60.03,
      "grad_norm": 3.8435778617858887,
      "learning_rate": 3.996996996996997e-06,
      "loss": 0.3756,
      "step": 19990
    },
    {
      "epoch": 60.06,
      "grad_norm": 4.47481632232666,
      "learning_rate": 3.993993993993994e-06,
      "loss": 0.4192,
      "step": 20000
    },
    {
      "epoch": 60.09,
      "grad_norm": 4.498948097229004,
      "learning_rate": 3.990990990990991e-06,
      "loss": 0.3929,
      "step": 20010
    },
    {
      "epoch": 60.12,
      "grad_norm": 6.061063766479492,
      "learning_rate": 3.987987987987989e-06,
      "loss": 0.4538,
      "step": 20020
    },
    {
      "epoch": 60.15,
      "grad_norm": 4.179337024688721,
      "learning_rate": 3.9849849849849855e-06,
      "loss": 0.3682,
      "step": 20030
    },
    {
      "epoch": 60.18,
      "grad_norm": 5.216439723968506,
      "learning_rate": 3.981981981981982e-06,
      "loss": 0.3773,
      "step": 20040
    },
    {
      "epoch": 60.21,
      "grad_norm": 4.404881000518799,
      "learning_rate": 3.97897897897898e-06,
      "loss": 0.4092,
      "step": 20050
    },
    {
      "epoch": 60.24,
      "grad_norm": 4.393664360046387,
      "learning_rate": 3.975975975975976e-06,
      "loss": 0.3895,
      "step": 20060
    },
    {
      "epoch": 60.27,
      "grad_norm": 4.570151329040527,
      "learning_rate": 3.972972972972973e-06,
      "loss": 0.4166,
      "step": 20070
    },
    {
      "epoch": 60.3,
      "grad_norm": 5.764723300933838,
      "learning_rate": 3.96996996996997e-06,
      "loss": 0.3978,
      "step": 20080
    },
    {
      "epoch": 60.33,
      "grad_norm": 4.762311935424805,
      "learning_rate": 3.966966966966968e-06,
      "loss": 0.3973,
      "step": 20090
    },
    {
      "epoch": 60.36,
      "grad_norm": 4.624218463897705,
      "learning_rate": 3.9639639639639645e-06,
      "loss": 0.4147,
      "step": 20100
    },
    {
      "epoch": 60.39,
      "grad_norm": 4.683891296386719,
      "learning_rate": 3.960960960960961e-06,
      "loss": 0.3851,
      "step": 20110
    },
    {
      "epoch": 60.42,
      "grad_norm": 4.707220554351807,
      "learning_rate": 3.957957957957958e-06,
      "loss": 0.3844,
      "step": 20120
    },
    {
      "epoch": 60.45,
      "grad_norm": 5.220789909362793,
      "learning_rate": 3.954954954954955e-06,
      "loss": 0.3686,
      "step": 20130
    },
    {
      "epoch": 60.48,
      "grad_norm": 3.2193665504455566,
      "learning_rate": 3.951951951951952e-06,
      "loss": 0.3647,
      "step": 20140
    },
    {
      "epoch": 60.51,
      "grad_norm": 3.45483136177063,
      "learning_rate": 3.948948948948949e-06,
      "loss": 0.3567,
      "step": 20150
    },
    {
      "epoch": 60.54,
      "grad_norm": 4.213616371154785,
      "learning_rate": 3.945945945945947e-06,
      "loss": 0.3612,
      "step": 20160
    },
    {
      "epoch": 60.57,
      "grad_norm": 5.336034297943115,
      "learning_rate": 3.9429429429429434e-06,
      "loss": 0.3807,
      "step": 20170
    },
    {
      "epoch": 60.6,
      "grad_norm": 5.562955379486084,
      "learning_rate": 3.93993993993994e-06,
      "loss": 0.3947,
      "step": 20180
    },
    {
      "epoch": 60.63,
      "grad_norm": 6.122558116912842,
      "learning_rate": 3.936936936936937e-06,
      "loss": 0.3826,
      "step": 20190
    },
    {
      "epoch": 60.66,
      "grad_norm": 5.086697101593018,
      "learning_rate": 3.933933933933934e-06,
      "loss": 0.3894,
      "step": 20200
    },
    {
      "epoch": 60.69,
      "grad_norm": 5.080352783203125,
      "learning_rate": 3.930930930930932e-06,
      "loss": 0.3563,
      "step": 20210
    },
    {
      "epoch": 60.72,
      "grad_norm": 4.434232234954834,
      "learning_rate": 3.927927927927928e-06,
      "loss": 0.4014,
      "step": 20220
    },
    {
      "epoch": 60.75,
      "grad_norm": 4.557462215423584,
      "learning_rate": 3.924924924924926e-06,
      "loss": 0.3769,
      "step": 20230
    },
    {
      "epoch": 60.78,
      "grad_norm": 5.012494087219238,
      "learning_rate": 3.921921921921922e-06,
      "loss": 0.41,
      "step": 20240
    },
    {
      "epoch": 60.81,
      "grad_norm": 5.18105411529541,
      "learning_rate": 3.918918918918919e-06,
      "loss": 0.3893,
      "step": 20250
    },
    {
      "epoch": 60.84,
      "grad_norm": 4.821757793426514,
      "learning_rate": 3.915915915915916e-06,
      "loss": 0.4062,
      "step": 20260
    },
    {
      "epoch": 60.87,
      "grad_norm": 3.6783230304718018,
      "learning_rate": 3.912912912912913e-06,
      "loss": 0.3185,
      "step": 20270
    },
    {
      "epoch": 60.9,
      "grad_norm": 4.306304454803467,
      "learning_rate": 3.909909909909911e-06,
      "loss": 0.4082,
      "step": 20280
    },
    {
      "epoch": 60.93,
      "grad_norm": 5.269168376922607,
      "learning_rate": 3.906906906906907e-06,
      "loss": 0.3897,
      "step": 20290
    },
    {
      "epoch": 60.96,
      "grad_norm": 3.9647092819213867,
      "learning_rate": 3.903903903903904e-06,
      "loss": 0.3614,
      "step": 20300
    },
    {
      "epoch": 60.99,
      "grad_norm": 4.930453300476074,
      "learning_rate": 3.900900900900901e-06,
      "loss": 0.3537,
      "step": 20310
    },
    {
      "epoch": 61.0,
      "eval_accuracy": 0.9197,
      "eval_loss": 0.3066848814487457,
      "eval_runtime": 25.9533,
      "eval_samples_per_second": 385.308,
      "eval_steps_per_second": 1.541,
      "step": 20313
    },
    {
      "epoch": 61.02,
      "grad_norm": 4.945713520050049,
      "learning_rate": 3.897897897897898e-06,
      "loss": 0.3833,
      "step": 20320
    },
    {
      "epoch": 61.05,
      "grad_norm": 4.452799320220947,
      "learning_rate": 3.894894894894895e-06,
      "loss": 0.3783,
      "step": 20330
    },
    {
      "epoch": 61.08,
      "grad_norm": 4.242116928100586,
      "learning_rate": 3.891891891891892e-06,
      "loss": 0.4126,
      "step": 20340
    },
    {
      "epoch": 61.11,
      "grad_norm": 4.080423831939697,
      "learning_rate": 3.88888888888889e-06,
      "loss": 0.3872,
      "step": 20350
    },
    {
      "epoch": 61.14,
      "grad_norm": 4.973381042480469,
      "learning_rate": 3.885885885885886e-06,
      "loss": 0.3229,
      "step": 20360
    },
    {
      "epoch": 61.17,
      "grad_norm": 7.072704792022705,
      "learning_rate": 3.882882882882883e-06,
      "loss": 0.4061,
      "step": 20370
    },
    {
      "epoch": 61.2,
      "grad_norm": 5.432439804077148,
      "learning_rate": 3.87987987987988e-06,
      "loss": 0.383,
      "step": 20380
    },
    {
      "epoch": 61.23,
      "grad_norm": 3.5048320293426514,
      "learning_rate": 3.876876876876877e-06,
      "loss": 0.4088,
      "step": 20390
    },
    {
      "epoch": 61.26,
      "grad_norm": 5.393626689910889,
      "learning_rate": 3.8738738738738744e-06,
      "loss": 0.374,
      "step": 20400
    },
    {
      "epoch": 61.29,
      "grad_norm": 5.329527378082275,
      "learning_rate": 3.870870870870871e-06,
      "loss": 0.4234,
      "step": 20410
    },
    {
      "epoch": 61.32,
      "grad_norm": 5.24342679977417,
      "learning_rate": 3.8678678678678685e-06,
      "loss": 0.3811,
      "step": 20420
    },
    {
      "epoch": 61.35,
      "grad_norm": 4.858025550842285,
      "learning_rate": 3.864864864864865e-06,
      "loss": 0.4721,
      "step": 20430
    },
    {
      "epoch": 61.38,
      "grad_norm": 4.607154846191406,
      "learning_rate": 3.861861861861862e-06,
      "loss": 0.3892,
      "step": 20440
    },
    {
      "epoch": 61.41,
      "grad_norm": 5.811942100524902,
      "learning_rate": 3.858858858858859e-06,
      "loss": 0.3953,
      "step": 20450
    },
    {
      "epoch": 61.44,
      "grad_norm": 4.223262786865234,
      "learning_rate": 3.855855855855856e-06,
      "loss": 0.3688,
      "step": 20460
    },
    {
      "epoch": 61.47,
      "grad_norm": 2.956552267074585,
      "learning_rate": 3.852852852852853e-06,
      "loss": 0.3564,
      "step": 20470
    },
    {
      "epoch": 61.5,
      "grad_norm": 3.6181747913360596,
      "learning_rate": 3.84984984984985e-06,
      "loss": 0.4303,
      "step": 20480
    },
    {
      "epoch": 61.53,
      "grad_norm": 5.11487340927124,
      "learning_rate": 3.8468468468468475e-06,
      "loss": 0.4152,
      "step": 20490
    },
    {
      "epoch": 61.56,
      "grad_norm": 3.2137460708618164,
      "learning_rate": 3.843843843843844e-06,
      "loss": 0.3635,
      "step": 20500
    },
    {
      "epoch": 61.59,
      "grad_norm": 4.978244304656982,
      "learning_rate": 3.840840840840841e-06,
      "loss": 0.4166,
      "step": 20510
    },
    {
      "epoch": 61.62,
      "grad_norm": 4.7142181396484375,
      "learning_rate": 3.837837837837838e-06,
      "loss": 0.4087,
      "step": 20520
    },
    {
      "epoch": 61.65,
      "grad_norm": 5.6237640380859375,
      "learning_rate": 3.834834834834835e-06,
      "loss": 0.4064,
      "step": 20530
    },
    {
      "epoch": 61.68,
      "grad_norm": 5.961127758026123,
      "learning_rate": 3.831831831831832e-06,
      "loss": 0.3795,
      "step": 20540
    },
    {
      "epoch": 61.71,
      "grad_norm": 4.050546169281006,
      "learning_rate": 3.828828828828829e-06,
      "loss": 0.3936,
      "step": 20550
    },
    {
      "epoch": 61.74,
      "grad_norm": 3.263345956802368,
      "learning_rate": 3.8258258258258265e-06,
      "loss": 0.3528,
      "step": 20560
    },
    {
      "epoch": 61.77,
      "grad_norm": 4.271215438842773,
      "learning_rate": 3.822822822822823e-06,
      "loss": 0.3509,
      "step": 20570
    },
    {
      "epoch": 61.8,
      "grad_norm": 2.8282196521759033,
      "learning_rate": 3.81981981981982e-06,
      "loss": 0.3777,
      "step": 20580
    },
    {
      "epoch": 61.83,
      "grad_norm": 5.88720178604126,
      "learning_rate": 3.816816816816817e-06,
      "loss": 0.4158,
      "step": 20590
    },
    {
      "epoch": 61.86,
      "grad_norm": 4.451913356781006,
      "learning_rate": 3.8138138138138143e-06,
      "loss": 0.3835,
      "step": 20600
    },
    {
      "epoch": 61.89,
      "grad_norm": 5.0170578956604,
      "learning_rate": 3.810810810810811e-06,
      "loss": 0.3739,
      "step": 20610
    },
    {
      "epoch": 61.92,
      "grad_norm": 4.939899444580078,
      "learning_rate": 3.807807807807808e-06,
      "loss": 0.3287,
      "step": 20620
    },
    {
      "epoch": 61.95,
      "grad_norm": 6.6928935050964355,
      "learning_rate": 3.8048048048048054e-06,
      "loss": 0.3789,
      "step": 20630
    },
    {
      "epoch": 61.98,
      "grad_norm": 3.528116464614868,
      "learning_rate": 3.801801801801802e-06,
      "loss": 0.3613,
      "step": 20640
    },
    {
      "epoch": 62.0,
      "eval_accuracy": 0.9192,
      "eval_loss": 0.3064844608306885,
      "eval_runtime": 26.08,
      "eval_samples_per_second": 383.436,
      "eval_steps_per_second": 1.534,
      "step": 20646
    },
    {
      "epoch": 62.01,
      "grad_norm": 4.800917625427246,
      "learning_rate": 3.798798798798799e-06,
      "loss": 0.3249,
      "step": 20650
    },
    {
      "epoch": 62.04,
      "grad_norm": 3.9344377517700195,
      "learning_rate": 3.795795795795796e-06,
      "loss": 0.3733,
      "step": 20660
    },
    {
      "epoch": 62.07,
      "grad_norm": 4.527621746063232,
      "learning_rate": 3.7927927927927932e-06,
      "loss": 0.3921,
      "step": 20670
    },
    {
      "epoch": 62.1,
      "grad_norm": 6.294924736022949,
      "learning_rate": 3.78978978978979e-06,
      "loss": 0.3951,
      "step": 20680
    },
    {
      "epoch": 62.13,
      "grad_norm": 4.456297397613525,
      "learning_rate": 3.786786786786787e-06,
      "loss": 0.3988,
      "step": 20690
    },
    {
      "epoch": 62.16,
      "grad_norm": 4.620143413543701,
      "learning_rate": 3.7837837837837844e-06,
      "loss": 0.4046,
      "step": 20700
    },
    {
      "epoch": 62.19,
      "grad_norm": 4.305004119873047,
      "learning_rate": 3.780780780780781e-06,
      "loss": 0.3839,
      "step": 20710
    },
    {
      "epoch": 62.22,
      "grad_norm": 5.435354709625244,
      "learning_rate": 3.777777777777778e-06,
      "loss": 0.3582,
      "step": 20720
    },
    {
      "epoch": 62.25,
      "grad_norm": 5.776912689208984,
      "learning_rate": 3.7747747747747747e-06,
      "loss": 0.3987,
      "step": 20730
    },
    {
      "epoch": 62.28,
      "grad_norm": 6.167961597442627,
      "learning_rate": 3.771771771771772e-06,
      "loss": 0.4124,
      "step": 20740
    },
    {
      "epoch": 62.31,
      "grad_norm": 4.485419750213623,
      "learning_rate": 3.7687687687687692e-06,
      "loss": 0.3736,
      "step": 20750
    },
    {
      "epoch": 62.34,
      "grad_norm": 4.205562114715576,
      "learning_rate": 3.765765765765766e-06,
      "loss": 0.4019,
      "step": 20760
    },
    {
      "epoch": 62.37,
      "grad_norm": 5.3255157470703125,
      "learning_rate": 3.7627627627627634e-06,
      "loss": 0.3175,
      "step": 20770
    },
    {
      "epoch": 62.4,
      "grad_norm": 5.313967704772949,
      "learning_rate": 3.75975975975976e-06,
      "loss": 0.3539,
      "step": 20780
    },
    {
      "epoch": 62.43,
      "grad_norm": 4.386773109436035,
      "learning_rate": 3.756756756756757e-06,
      "loss": 0.3508,
      "step": 20790
    },
    {
      "epoch": 62.46,
      "grad_norm": 9.731864929199219,
      "learning_rate": 3.7537537537537537e-06,
      "loss": 0.435,
      "step": 20800
    },
    {
      "epoch": 62.49,
      "grad_norm": 3.1956357955932617,
      "learning_rate": 3.750750750750751e-06,
      "loss": 0.3929,
      "step": 20810
    },
    {
      "epoch": 62.52,
      "grad_norm": 4.68299674987793,
      "learning_rate": 3.747747747747748e-06,
      "loss": 0.3873,
      "step": 20820
    },
    {
      "epoch": 62.55,
      "grad_norm": 7.765761375427246,
      "learning_rate": 3.744744744744745e-06,
      "loss": 0.4335,
      "step": 20830
    },
    {
      "epoch": 62.58,
      "grad_norm": 3.818394660949707,
      "learning_rate": 3.7417417417417423e-06,
      "loss": 0.4259,
      "step": 20840
    },
    {
      "epoch": 62.61,
      "grad_norm": 3.908114194869995,
      "learning_rate": 3.7387387387387394e-06,
      "loss": 0.4126,
      "step": 20850
    },
    {
      "epoch": 62.64,
      "grad_norm": 5.528392791748047,
      "learning_rate": 3.735735735735736e-06,
      "loss": 0.3949,
      "step": 20860
    },
    {
      "epoch": 62.67,
      "grad_norm": 4.3458967208862305,
      "learning_rate": 3.732732732732733e-06,
      "loss": 0.402,
      "step": 20870
    },
    {
      "epoch": 62.7,
      "grad_norm": 4.576776027679443,
      "learning_rate": 3.72972972972973e-06,
      "loss": 0.3944,
      "step": 20880
    },
    {
      "epoch": 62.73,
      "grad_norm": 4.224873065948486,
      "learning_rate": 3.726726726726727e-06,
      "loss": 0.3504,
      "step": 20890
    },
    {
      "epoch": 62.76,
      "grad_norm": 5.1518096923828125,
      "learning_rate": 3.723723723723724e-06,
      "loss": 0.3751,
      "step": 20900
    },
    {
      "epoch": 62.79,
      "grad_norm": 5.555248260498047,
      "learning_rate": 3.720720720720721e-06,
      "loss": 0.4528,
      "step": 20910
    },
    {
      "epoch": 62.82,
      "grad_norm": 6.734523296356201,
      "learning_rate": 3.7177177177177183e-06,
      "loss": 0.3675,
      "step": 20920
    },
    {
      "epoch": 62.85,
      "grad_norm": 7.117071151733398,
      "learning_rate": 3.714714714714715e-06,
      "loss": 0.4167,
      "step": 20930
    },
    {
      "epoch": 62.88,
      "grad_norm": 3.421349048614502,
      "learning_rate": 3.711711711711712e-06,
      "loss": 0.396,
      "step": 20940
    },
    {
      "epoch": 62.91,
      "grad_norm": 3.2929751873016357,
      "learning_rate": 3.708708708708709e-06,
      "loss": 0.3755,
      "step": 20950
    },
    {
      "epoch": 62.94,
      "grad_norm": 6.547852993011475,
      "learning_rate": 3.705705705705706e-06,
      "loss": 0.3673,
      "step": 20960
    },
    {
      "epoch": 62.97,
      "grad_norm": 5.214502811431885,
      "learning_rate": 3.7027027027027028e-06,
      "loss": 0.4121,
      "step": 20970
    },
    {
      "epoch": 63.0,
      "eval_accuracy": 0.9211,
      "eval_loss": 0.30859071016311646,
      "eval_runtime": 26.189,
      "eval_samples_per_second": 381.839,
      "eval_steps_per_second": 1.527,
      "step": 20979
    },
    {
      "epoch": 63.0,
      "grad_norm": 3.956397533416748,
      "learning_rate": 3.6996996996997e-06,
      "loss": 0.3036,
      "step": 20980
    },
    {
      "epoch": 63.03,
      "grad_norm": 5.618265628814697,
      "learning_rate": 3.6966966966966973e-06,
      "loss": 0.4,
      "step": 20990
    },
    {
      "epoch": 63.06,
      "grad_norm": 2.949518918991089,
      "learning_rate": 3.693693693693694e-06,
      "loss": 0.3892,
      "step": 21000
    },
    {
      "epoch": 63.09,
      "grad_norm": 3.7895524501800537,
      "learning_rate": 3.690690690690691e-06,
      "loss": 0.3739,
      "step": 21010
    },
    {
      "epoch": 63.12,
      "grad_norm": 5.325167655944824,
      "learning_rate": 3.6876876876876885e-06,
      "loss": 0.3692,
      "step": 21020
    },
    {
      "epoch": 63.15,
      "grad_norm": 4.152174472808838,
      "learning_rate": 3.684684684684685e-06,
      "loss": 0.4118,
      "step": 21030
    },
    {
      "epoch": 63.18,
      "grad_norm": 3.1754322052001953,
      "learning_rate": 3.681681681681682e-06,
      "loss": 0.4452,
      "step": 21040
    },
    {
      "epoch": 63.21,
      "grad_norm": 13.56212043762207,
      "learning_rate": 3.6786786786786788e-06,
      "loss": 0.3399,
      "step": 21050
    },
    {
      "epoch": 63.24,
      "grad_norm": 6.920172214508057,
      "learning_rate": 3.6756756756756763e-06,
      "loss": 0.3607,
      "step": 21060
    },
    {
      "epoch": 63.27,
      "grad_norm": 5.599009990692139,
      "learning_rate": 3.672672672672673e-06,
      "loss": 0.3711,
      "step": 21070
    },
    {
      "epoch": 63.3,
      "grad_norm": 5.135616302490234,
      "learning_rate": 3.66966966966967e-06,
      "loss": 0.328,
      "step": 21080
    },
    {
      "epoch": 63.33,
      "grad_norm": 4.890881538391113,
      "learning_rate": 3.6666666666666666e-06,
      "loss": 0.3791,
      "step": 21090
    },
    {
      "epoch": 63.36,
      "grad_norm": 3.6989517211914062,
      "learning_rate": 3.663663663663664e-06,
      "loss": 0.3914,
      "step": 21100
    },
    {
      "epoch": 63.39,
      "grad_norm": 4.700137138366699,
      "learning_rate": 3.660660660660661e-06,
      "loss": 0.4341,
      "step": 21110
    },
    {
      "epoch": 63.42,
      "grad_norm": 4.795035362243652,
      "learning_rate": 3.6576576576576577e-06,
      "loss": 0.4069,
      "step": 21120
    },
    {
      "epoch": 63.45,
      "grad_norm": 4.954772472381592,
      "learning_rate": 3.6546546546546552e-06,
      "loss": 0.384,
      "step": 21130
    },
    {
      "epoch": 63.48,
      "grad_norm": 5.6409220695495605,
      "learning_rate": 3.651651651651652e-06,
      "loss": 0.4592,
      "step": 21140
    },
    {
      "epoch": 63.51,
      "grad_norm": 4.693268775939941,
      "learning_rate": 3.648648648648649e-06,
      "loss": 0.382,
      "step": 21150
    },
    {
      "epoch": 63.54,
      "grad_norm": 5.199280738830566,
      "learning_rate": 3.6456456456456455e-06,
      "loss": 0.3433,
      "step": 21160
    },
    {
      "epoch": 63.57,
      "grad_norm": 5.5577921867370605,
      "learning_rate": 3.642642642642643e-06,
      "loss": 0.3656,
      "step": 21170
    },
    {
      "epoch": 63.6,
      "grad_norm": 4.887319564819336,
      "learning_rate": 3.63963963963964e-06,
      "loss": 0.3866,
      "step": 21180
    },
    {
      "epoch": 63.63,
      "grad_norm": 5.078019142150879,
      "learning_rate": 3.6366366366366367e-06,
      "loss": 0.37,
      "step": 21190
    },
    {
      "epoch": 63.66,
      "grad_norm": 5.7338056564331055,
      "learning_rate": 3.633633633633634e-06,
      "loss": 0.3537,
      "step": 21200
    },
    {
      "epoch": 63.69,
      "grad_norm": 4.43660831451416,
      "learning_rate": 3.6306306306306312e-06,
      "loss": 0.4301,
      "step": 21210
    },
    {
      "epoch": 63.72,
      "grad_norm": 3.753462314605713,
      "learning_rate": 3.627627627627628e-06,
      "loss": 0.3823,
      "step": 21220
    },
    {
      "epoch": 63.75,
      "grad_norm": 5.3210015296936035,
      "learning_rate": 3.624624624624625e-06,
      "loss": 0.324,
      "step": 21230
    },
    {
      "epoch": 63.78,
      "grad_norm": 2.9679131507873535,
      "learning_rate": 3.621621621621622e-06,
      "loss": 0.3646,
      "step": 21240
    },
    {
      "epoch": 63.81,
      "grad_norm": 5.5320329666137695,
      "learning_rate": 3.618618618618619e-06,
      "loss": 0.3609,
      "step": 21250
    },
    {
      "epoch": 63.84,
      "grad_norm": 3.8269875049591064,
      "learning_rate": 3.6156156156156157e-06,
      "loss": 0.3616,
      "step": 21260
    },
    {
      "epoch": 63.87,
      "grad_norm": 6.501304626464844,
      "learning_rate": 3.6126126126126127e-06,
      "loss": 0.3967,
      "step": 21270
    },
    {
      "epoch": 63.9,
      "grad_norm": 4.398569107055664,
      "learning_rate": 3.60960960960961e-06,
      "loss": 0.3422,
      "step": 21280
    },
    {
      "epoch": 63.93,
      "grad_norm": 4.101565361022949,
      "learning_rate": 3.606606606606607e-06,
      "loss": 0.397,
      "step": 21290
    },
    {
      "epoch": 63.96,
      "grad_norm": 6.4529619216918945,
      "learning_rate": 3.603603603603604e-06,
      "loss": 0.3943,
      "step": 21300
    },
    {
      "epoch": 63.99,
      "grad_norm": 4.06441593170166,
      "learning_rate": 3.600600600600601e-06,
      "loss": 0.3939,
      "step": 21310
    },
    {
      "epoch": 64.0,
      "eval_accuracy": 0.9207,
      "eval_loss": 0.30954065918922424,
      "eval_runtime": 26.2905,
      "eval_samples_per_second": 380.365,
      "eval_steps_per_second": 1.521,
      "step": 21312
    },
    {
      "epoch": 64.02,
      "grad_norm": 4.953286170959473,
      "learning_rate": 3.597597597597598e-06,
      "loss": 0.3615,
      "step": 21320
    },
    {
      "epoch": 64.05,
      "grad_norm": 5.115283489227295,
      "learning_rate": 3.5945945945945946e-06,
      "loss": 0.3785,
      "step": 21330
    },
    {
      "epoch": 64.08,
      "grad_norm": 3.517745018005371,
      "learning_rate": 3.5915915915915917e-06,
      "loss": 0.3923,
      "step": 21340
    },
    {
      "epoch": 64.11,
      "grad_norm": 3.9651553630828857,
      "learning_rate": 3.588588588588589e-06,
      "loss": 0.3428,
      "step": 21350
    },
    {
      "epoch": 64.14,
      "grad_norm": 4.92030668258667,
      "learning_rate": 3.5855855855855858e-06,
      "loss": 0.3862,
      "step": 21360
    },
    {
      "epoch": 64.17,
      "grad_norm": 4.2861785888671875,
      "learning_rate": 3.582582582582583e-06,
      "loss": 0.4044,
      "step": 21370
    },
    {
      "epoch": 64.2,
      "grad_norm": 5.567362308502197,
      "learning_rate": 3.5795795795795803e-06,
      "loss": 0.381,
      "step": 21380
    },
    {
      "epoch": 64.23,
      "grad_norm": 4.110671520233154,
      "learning_rate": 3.576576576576577e-06,
      "loss": 0.4121,
      "step": 21390
    },
    {
      "epoch": 64.26,
      "grad_norm": 6.83501672744751,
      "learning_rate": 3.573573573573574e-06,
      "loss": 0.3873,
      "step": 21400
    },
    {
      "epoch": 64.29,
      "grad_norm": 3.0805516242980957,
      "learning_rate": 3.5705705705705706e-06,
      "loss": 0.3095,
      "step": 21410
    },
    {
      "epoch": 64.32,
      "grad_norm": 5.887385368347168,
      "learning_rate": 3.567567567567568e-06,
      "loss": 0.3781,
      "step": 21420
    },
    {
      "epoch": 64.35,
      "grad_norm": 5.104264736175537,
      "learning_rate": 3.5645645645645647e-06,
      "loss": 0.375,
      "step": 21430
    },
    {
      "epoch": 64.38,
      "grad_norm": 6.251524448394775,
      "learning_rate": 3.561561561561562e-06,
      "loss": 0.3793,
      "step": 21440
    },
    {
      "epoch": 64.41,
      "grad_norm": 3.7513630390167236,
      "learning_rate": 3.5585585585585584e-06,
      "loss": 0.408,
      "step": 21450
    },
    {
      "epoch": 64.44,
      "grad_norm": 5.234687328338623,
      "learning_rate": 3.555555555555556e-06,
      "loss": 0.3879,
      "step": 21460
    },
    {
      "epoch": 64.47,
      "grad_norm": 4.965696334838867,
      "learning_rate": 3.552552552552553e-06,
      "loss": 0.3681,
      "step": 21470
    },
    {
      "epoch": 64.5,
      "grad_norm": 3.486311435699463,
      "learning_rate": 3.5495495495495496e-06,
      "loss": 0.3424,
      "step": 21480
    },
    {
      "epoch": 64.53,
      "grad_norm": 3.9898505210876465,
      "learning_rate": 3.546546546546547e-06,
      "loss": 0.3365,
      "step": 21490
    },
    {
      "epoch": 64.56,
      "grad_norm": 3.61309552192688,
      "learning_rate": 3.5435435435435437e-06,
      "loss": 0.4277,
      "step": 21500
    },
    {
      "epoch": 64.59,
      "grad_norm": 4.627030849456787,
      "learning_rate": 3.5405405405405408e-06,
      "loss": 0.4315,
      "step": 21510
    },
    {
      "epoch": 64.62,
      "grad_norm": 5.69821310043335,
      "learning_rate": 3.5375375375375374e-06,
      "loss": 0.3505,
      "step": 21520
    },
    {
      "epoch": 64.65,
      "grad_norm": 4.802726745605469,
      "learning_rate": 3.534534534534535e-06,
      "loss": 0.4128,
      "step": 21530
    },
    {
      "epoch": 64.68,
      "grad_norm": 5.268336772918701,
      "learning_rate": 3.531531531531532e-06,
      "loss": 0.3591,
      "step": 21540
    },
    {
      "epoch": 64.71,
      "grad_norm": 4.43294620513916,
      "learning_rate": 3.5285285285285286e-06,
      "loss": 0.384,
      "step": 21550
    },
    {
      "epoch": 64.74,
      "grad_norm": 3.6053569316864014,
      "learning_rate": 3.525525525525526e-06,
      "loss": 0.3426,
      "step": 21560
    },
    {
      "epoch": 64.77,
      "grad_norm": 4.746013164520264,
      "learning_rate": 3.522522522522523e-06,
      "loss": 0.3665,
      "step": 21570
    },
    {
      "epoch": 64.8,
      "grad_norm": 6.0806050300598145,
      "learning_rate": 3.5195195195195197e-06,
      "loss": 0.3687,
      "step": 21580
    },
    {
      "epoch": 64.83,
      "grad_norm": 5.150928497314453,
      "learning_rate": 3.5165165165165168e-06,
      "loss": 0.4005,
      "step": 21590
    },
    {
      "epoch": 64.86,
      "grad_norm": 4.460043907165527,
      "learning_rate": 3.513513513513514e-06,
      "loss": 0.4371,
      "step": 21600
    },
    {
      "epoch": 64.89,
      "grad_norm": 4.84251594543457,
      "learning_rate": 3.510510510510511e-06,
      "loss": 0.4031,
      "step": 21610
    },
    {
      "epoch": 64.92,
      "grad_norm": 4.184168338775635,
      "learning_rate": 3.5075075075075075e-06,
      "loss": 0.3984,
      "step": 21620
    },
    {
      "epoch": 64.95,
      "grad_norm": 4.626136779785156,
      "learning_rate": 3.5045045045045046e-06,
      "loss": 0.3556,
      "step": 21630
    },
    {
      "epoch": 64.98,
      "grad_norm": 4.924071311950684,
      "learning_rate": 3.501501501501502e-06,
      "loss": 0.3616,
      "step": 21640
    },
    {
      "epoch": 65.0,
      "eval_accuracy": 0.9215,
      "eval_loss": 0.3061123490333557,
      "eval_runtime": 26.5534,
      "eval_samples_per_second": 376.6,
      "eval_steps_per_second": 1.506,
      "step": 21645
    },
    {
      "epoch": 65.02,
      "grad_norm": 4.482754707336426,
      "learning_rate": 3.4984984984984987e-06,
      "loss": 0.4084,
      "step": 21650
    },
    {
      "epoch": 65.05,
      "grad_norm": 4.970823287963867,
      "learning_rate": 3.4954954954954957e-06,
      "loss": 0.3466,
      "step": 21660
    },
    {
      "epoch": 65.08,
      "grad_norm": 10.77085018157959,
      "learning_rate": 3.492492492492493e-06,
      "loss": 0.3682,
      "step": 21670
    },
    {
      "epoch": 65.11,
      "grad_norm": 3.991971254348755,
      "learning_rate": 3.48948948948949e-06,
      "loss": 0.3643,
      "step": 21680
    },
    {
      "epoch": 65.14,
      "grad_norm": 3.6868784427642822,
      "learning_rate": 3.4864864864864865e-06,
      "loss": 0.3756,
      "step": 21690
    },
    {
      "epoch": 65.17,
      "grad_norm": 3.726219892501831,
      "learning_rate": 3.4834834834834835e-06,
      "loss": 0.3795,
      "step": 21700
    },
    {
      "epoch": 65.2,
      "grad_norm": 5.108314037322998,
      "learning_rate": 3.480480480480481e-06,
      "loss": 0.3777,
      "step": 21710
    },
    {
      "epoch": 65.23,
      "grad_norm": 5.024679660797119,
      "learning_rate": 3.4774774774774776e-06,
      "loss": 0.3858,
      "step": 21720
    },
    {
      "epoch": 65.26,
      "grad_norm": 3.717787981033325,
      "learning_rate": 3.4744744744744747e-06,
      "loss": 0.3916,
      "step": 21730
    },
    {
      "epoch": 65.29,
      "grad_norm": 5.6035332679748535,
      "learning_rate": 3.471471471471472e-06,
      "loss": 0.3911,
      "step": 21740
    },
    {
      "epoch": 65.32,
      "grad_norm": 4.82771635055542,
      "learning_rate": 3.468468468468469e-06,
      "loss": 0.3635,
      "step": 21750
    },
    {
      "epoch": 65.35,
      "grad_norm": 3.7722058296203613,
      "learning_rate": 3.465465465465466e-06,
      "loss": 0.412,
      "step": 21760
    },
    {
      "epoch": 65.38,
      "grad_norm": 4.262643337249756,
      "learning_rate": 3.4624624624624625e-06,
      "loss": 0.3689,
      "step": 21770
    },
    {
      "epoch": 65.41,
      "grad_norm": 3.9743552207946777,
      "learning_rate": 3.45945945945946e-06,
      "loss": 0.3779,
      "step": 21780
    },
    {
      "epoch": 65.44,
      "grad_norm": 5.614193916320801,
      "learning_rate": 3.4564564564564566e-06,
      "loss": 0.3699,
      "step": 21790
    },
    {
      "epoch": 65.47,
      "grad_norm": 3.662853956222534,
      "learning_rate": 3.4534534534534537e-06,
      "loss": 0.314,
      "step": 21800
    },
    {
      "epoch": 65.5,
      "grad_norm": 6.2071638107299805,
      "learning_rate": 3.4504504504504503e-06,
      "loss": 0.382,
      "step": 21810
    },
    {
      "epoch": 65.53,
      "grad_norm": 4.553779602050781,
      "learning_rate": 3.4474474474474478e-06,
      "loss": 0.3925,
      "step": 21820
    },
    {
      "epoch": 65.56,
      "grad_norm": 4.637483596801758,
      "learning_rate": 3.444444444444445e-06,
      "loss": 0.3528,
      "step": 21830
    },
    {
      "epoch": 65.59,
      "grad_norm": 5.174117088317871,
      "learning_rate": 3.4414414414414415e-06,
      "loss": 0.3892,
      "step": 21840
    },
    {
      "epoch": 65.62,
      "grad_norm": 4.458976745605469,
      "learning_rate": 3.438438438438439e-06,
      "loss": 0.3604,
      "step": 21850
    },
    {
      "epoch": 65.65,
      "grad_norm": 4.422422885894775,
      "learning_rate": 3.435435435435436e-06,
      "loss": 0.3961,
      "step": 21860
    },
    {
      "epoch": 65.68,
      "grad_norm": 5.07048225402832,
      "learning_rate": 3.4324324324324326e-06,
      "loss": 0.3577,
      "step": 21870
    },
    {
      "epoch": 65.71,
      "grad_norm": 4.713285446166992,
      "learning_rate": 3.4294294294294293e-06,
      "loss": 0.3906,
      "step": 21880
    },
    {
      "epoch": 65.74,
      "grad_norm": 3.4346678256988525,
      "learning_rate": 3.4264264264264267e-06,
      "loss": 0.3224,
      "step": 21890
    },
    {
      "epoch": 65.77,
      "grad_norm": 4.42486047744751,
      "learning_rate": 3.423423423423424e-06,
      "loss": 0.3887,
      "step": 21900
    },
    {
      "epoch": 65.8,
      "grad_norm": 6.2922587394714355,
      "learning_rate": 3.4204204204204204e-06,
      "loss": 0.3875,
      "step": 21910
    },
    {
      "epoch": 65.83,
      "grad_norm": 4.207531929016113,
      "learning_rate": 3.417417417417418e-06,
      "loss": 0.3701,
      "step": 21920
    },
    {
      "epoch": 65.86,
      "grad_norm": 5.1708664894104,
      "learning_rate": 3.414414414414415e-06,
      "loss": 0.4074,
      "step": 21930
    },
    {
      "epoch": 65.89,
      "grad_norm": 4.039127826690674,
      "learning_rate": 3.4114114114114116e-06,
      "loss": 0.4095,
      "step": 21940
    },
    {
      "epoch": 65.92,
      "grad_norm": 3.3113348484039307,
      "learning_rate": 3.4084084084084086e-06,
      "loss": 0.395,
      "step": 21950
    },
    {
      "epoch": 65.95,
      "grad_norm": 3.6494743824005127,
      "learning_rate": 3.4054054054054057e-06,
      "loss": 0.4135,
      "step": 21960
    },
    {
      "epoch": 65.98,
      "grad_norm": 4.562663555145264,
      "learning_rate": 3.4024024024024028e-06,
      "loss": 0.3645,
      "step": 21970
    },
    {
      "epoch": 66.0,
      "eval_accuracy": 0.9197,
      "eval_loss": 0.3084704875946045,
      "eval_runtime": 26.4819,
      "eval_samples_per_second": 377.616,
      "eval_steps_per_second": 1.51,
      "step": 21978
    },
    {
      "epoch": 66.01,
      "grad_norm": 3.9480884075164795,
      "learning_rate": 3.3993993993993994e-06,
      "loss": 0.4167,
      "step": 21980
    },
    {
      "epoch": 66.04,
      "grad_norm": 5.431272983551025,
      "learning_rate": 3.3963963963963964e-06,
      "loss": 0.3488,
      "step": 21990
    },
    {
      "epoch": 66.07,
      "grad_norm": 4.069661617279053,
      "learning_rate": 3.393393393393394e-06,
      "loss": 0.4026,
      "step": 22000
    },
    {
      "epoch": 66.1,
      "grad_norm": 4.347344398498535,
      "learning_rate": 3.3903903903903905e-06,
      "loss": 0.378,
      "step": 22010
    },
    {
      "epoch": 66.13,
      "grad_norm": 3.6565163135528564,
      "learning_rate": 3.3873873873873876e-06,
      "loss": 0.3489,
      "step": 22020
    },
    {
      "epoch": 66.16,
      "grad_norm": 4.22859525680542,
      "learning_rate": 3.384384384384385e-06,
      "loss": 0.3592,
      "step": 22030
    },
    {
      "epoch": 66.19,
      "grad_norm": 5.074247360229492,
      "learning_rate": 3.3813813813813817e-06,
      "loss": 0.3658,
      "step": 22040
    },
    {
      "epoch": 66.22,
      "grad_norm": 5.796971321105957,
      "learning_rate": 3.3783783783783788e-06,
      "loss": 0.3482,
      "step": 22050
    },
    {
      "epoch": 66.25,
      "grad_norm": 3.799100399017334,
      "learning_rate": 3.3753753753753754e-06,
      "loss": 0.381,
      "step": 22060
    },
    {
      "epoch": 66.28,
      "grad_norm": 5.22889518737793,
      "learning_rate": 3.372372372372373e-06,
      "loss": 0.42,
      "step": 22070
    },
    {
      "epoch": 66.31,
      "grad_norm": 4.133579254150391,
      "learning_rate": 3.3693693693693695e-06,
      "loss": 0.4066,
      "step": 22080
    },
    {
      "epoch": 66.34,
      "grad_norm": 5.095855236053467,
      "learning_rate": 3.3663663663663666e-06,
      "loss": 0.3943,
      "step": 22090
    },
    {
      "epoch": 66.37,
      "grad_norm": 5.371054172515869,
      "learning_rate": 3.363363363363364e-06,
      "loss": 0.4097,
      "step": 22100
    },
    {
      "epoch": 66.4,
      "grad_norm": 4.6746649742126465,
      "learning_rate": 3.3603603603603607e-06,
      "loss": 0.4029,
      "step": 22110
    },
    {
      "epoch": 66.43,
      "grad_norm": 4.835991382598877,
      "learning_rate": 3.3573573573573577e-06,
      "loss": 0.3844,
      "step": 22120
    },
    {
      "epoch": 66.46,
      "grad_norm": 5.3595781326293945,
      "learning_rate": 3.3543543543543544e-06,
      "loss": 0.4025,
      "step": 22130
    },
    {
      "epoch": 66.49,
      "grad_norm": 4.14584493637085,
      "learning_rate": 3.351351351351352e-06,
      "loss": 0.38,
      "step": 22140
    },
    {
      "epoch": 66.52,
      "grad_norm": 4.5364179611206055,
      "learning_rate": 3.3483483483483485e-06,
      "loss": 0.3873,
      "step": 22150
    },
    {
      "epoch": 66.55,
      "grad_norm": 4.699276924133301,
      "learning_rate": 3.3453453453453455e-06,
      "loss": 0.3859,
      "step": 22160
    },
    {
      "epoch": 66.58,
      "grad_norm": 5.478759765625,
      "learning_rate": 3.342342342342342e-06,
      "loss": 0.3977,
      "step": 22170
    },
    {
      "epoch": 66.61,
      "grad_norm": 6.01693058013916,
      "learning_rate": 3.3393393393393396e-06,
      "loss": 0.343,
      "step": 22180
    },
    {
      "epoch": 66.64,
      "grad_norm": 4.839807510375977,
      "learning_rate": 3.3363363363363367e-06,
      "loss": 0.4181,
      "step": 22190
    },
    {
      "epoch": 66.67,
      "grad_norm": 6.35122013092041,
      "learning_rate": 3.3333333333333333e-06,
      "loss": 0.3949,
      "step": 22200
    },
    {
      "epoch": 66.7,
      "grad_norm": 5.245551586151123,
      "learning_rate": 3.330330330330331e-06,
      "loss": 0.3878,
      "step": 22210
    },
    {
      "epoch": 66.73,
      "grad_norm": 4.4067511558532715,
      "learning_rate": 3.327327327327328e-06,
      "loss": 0.3834,
      "step": 22220
    },
    {
      "epoch": 66.76,
      "grad_norm": 4.225093364715576,
      "learning_rate": 3.3243243243243245e-06,
      "loss": 0.3421,
      "step": 22230
    },
    {
      "epoch": 66.79,
      "grad_norm": 5.001962184906006,
      "learning_rate": 3.3213213213213215e-06,
      "loss": 0.366,
      "step": 22240
    },
    {
      "epoch": 66.82,
      "grad_norm": 3.8100368976593018,
      "learning_rate": 3.3183183183183186e-06,
      "loss": 0.3203,
      "step": 22250
    },
    {
      "epoch": 66.85,
      "grad_norm": 4.9912919998168945,
      "learning_rate": 3.3153153153153157e-06,
      "loss": 0.4215,
      "step": 22260
    },
    {
      "epoch": 66.88,
      "grad_norm": 4.240078449249268,
      "learning_rate": 3.3123123123123123e-06,
      "loss": 0.3949,
      "step": 22270
    },
    {
      "epoch": 66.91,
      "grad_norm": 4.508683204650879,
      "learning_rate": 3.3093093093093098e-06,
      "loss": 0.3263,
      "step": 22280
    },
    {
      "epoch": 66.94,
      "grad_norm": 2.2749359607696533,
      "learning_rate": 3.306306306306307e-06,
      "loss": 0.3652,
      "step": 22290
    },
    {
      "epoch": 66.97,
      "grad_norm": 5.107765197753906,
      "learning_rate": 3.3033033033033035e-06,
      "loss": 0.3526,
      "step": 22300
    },
    {
      "epoch": 67.0,
      "grad_norm": 4.587709903717041,
      "learning_rate": 3.3003003003003005e-06,
      "loss": 0.42,
      "step": 22310
    },
    {
      "epoch": 67.0,
      "eval_accuracy": 0.9191,
      "eval_loss": 0.30880001187324524,
      "eval_runtime": 26.3181,
      "eval_samples_per_second": 379.967,
      "eval_steps_per_second": 1.52,
      "step": 22311
    },
    {
      "epoch": 67.03,
      "grad_norm": 3.8284177780151367,
      "learning_rate": 3.2972972972972976e-06,
      "loss": 0.3997,
      "step": 22320
    },
    {
      "epoch": 67.06,
      "grad_norm": 3.72904896736145,
      "learning_rate": 3.2942942942942946e-06,
      "loss": 0.3192,
      "step": 22330
    },
    {
      "epoch": 67.09,
      "grad_norm": 5.0836334228515625,
      "learning_rate": 3.2912912912912912e-06,
      "loss": 0.3623,
      "step": 22340
    },
    {
      "epoch": 67.12,
      "grad_norm": 5.408961772918701,
      "learning_rate": 3.2882882882882887e-06,
      "loss": 0.3777,
      "step": 22350
    },
    {
      "epoch": 67.15,
      "grad_norm": 4.439548492431641,
      "learning_rate": 3.2852852852852858e-06,
      "loss": 0.4206,
      "step": 22360
    },
    {
      "epoch": 67.18,
      "grad_norm": 3.496751308441162,
      "learning_rate": 3.2822822822822824e-06,
      "loss": 0.352,
      "step": 22370
    },
    {
      "epoch": 67.21,
      "grad_norm": 21.46847915649414,
      "learning_rate": 3.2792792792792795e-06,
      "loss": 0.4106,
      "step": 22380
    },
    {
      "epoch": 67.24,
      "grad_norm": 3.786468982696533,
      "learning_rate": 3.276276276276277e-06,
      "loss": 0.3492,
      "step": 22390
    },
    {
      "epoch": 67.27,
      "grad_norm": 5.429421901702881,
      "learning_rate": 3.2732732732732736e-06,
      "loss": 0.4072,
      "step": 22400
    },
    {
      "epoch": 67.3,
      "grad_norm": 2.6024420261383057,
      "learning_rate": 3.2702702702702706e-06,
      "loss": 0.4079,
      "step": 22410
    },
    {
      "epoch": 67.33,
      "grad_norm": 4.384423732757568,
      "learning_rate": 3.2672672672672673e-06,
      "loss": 0.3915,
      "step": 22420
    },
    {
      "epoch": 67.36,
      "grad_norm": 7.983978271484375,
      "learning_rate": 3.2642642642642647e-06,
      "loss": 0.3531,
      "step": 22430
    },
    {
      "epoch": 67.39,
      "grad_norm": 3.854764699935913,
      "learning_rate": 3.2612612612612614e-06,
      "loss": 0.4003,
      "step": 22440
    },
    {
      "epoch": 67.42,
      "grad_norm": 3.7188901901245117,
      "learning_rate": 3.2582582582582584e-06,
      "loss": 0.3669,
      "step": 22450
    },
    {
      "epoch": 67.45,
      "grad_norm": 6.388355255126953,
      "learning_rate": 3.255255255255256e-06,
      "loss": 0.3817,
      "step": 22460
    },
    {
      "epoch": 67.48,
      "grad_norm": 4.381053447723389,
      "learning_rate": 3.2522522522522525e-06,
      "loss": 0.3313,
      "step": 22470
    },
    {
      "epoch": 67.51,
      "grad_norm": 4.686676502227783,
      "learning_rate": 3.2492492492492496e-06,
      "loss": 0.3811,
      "step": 22480
    },
    {
      "epoch": 67.54,
      "grad_norm": 4.62145471572876,
      "learning_rate": 3.2462462462462462e-06,
      "loss": 0.3865,
      "step": 22490
    },
    {
      "epoch": 67.57,
      "grad_norm": 5.3339643478393555,
      "learning_rate": 3.2432432432432437e-06,
      "loss": 0.3823,
      "step": 22500
    },
    {
      "epoch": 67.6,
      "grad_norm": 4.197082042694092,
      "learning_rate": 3.2402402402402403e-06,
      "loss": 0.3621,
      "step": 22510
    },
    {
      "epoch": 67.63,
      "grad_norm": 5.226799488067627,
      "learning_rate": 3.2372372372372374e-06,
      "loss": 0.4038,
      "step": 22520
    },
    {
      "epoch": 67.66,
      "grad_norm": 3.9460418224334717,
      "learning_rate": 3.234234234234235e-06,
      "loss": 0.3644,
      "step": 22530
    },
    {
      "epoch": 67.69,
      "grad_norm": 2.902524948120117,
      "learning_rate": 3.2312312312312315e-06,
      "loss": 0.336,
      "step": 22540
    },
    {
      "epoch": 67.72,
      "grad_norm": 4.78118371963501,
      "learning_rate": 3.2282282282282286e-06,
      "loss": 0.359,
      "step": 22550
    },
    {
      "epoch": 67.75,
      "grad_norm": 3.091318368911743,
      "learning_rate": 3.225225225225225e-06,
      "loss": 0.3378,
      "step": 22560
    },
    {
      "epoch": 67.78,
      "grad_norm": 6.15193510055542,
      "learning_rate": 3.2222222222222227e-06,
      "loss": 0.3691,
      "step": 22570
    },
    {
      "epoch": 67.81,
      "grad_norm": 5.554571628570557,
      "learning_rate": 3.2192192192192197e-06,
      "loss": 0.2938,
      "step": 22580
    },
    {
      "epoch": 67.84,
      "grad_norm": 3.568535566329956,
      "learning_rate": 3.2162162162162164e-06,
      "loss": 0.4071,
      "step": 22590
    },
    {
      "epoch": 67.87,
      "grad_norm": 5.819300651550293,
      "learning_rate": 3.2132132132132134e-06,
      "loss": 0.3614,
      "step": 22600
    },
    {
      "epoch": 67.9,
      "grad_norm": 6.227848052978516,
      "learning_rate": 3.2102102102102105e-06,
      "loss": 0.3567,
      "step": 22610
    },
    {
      "epoch": 67.93,
      "grad_norm": 5.12805700302124,
      "learning_rate": 3.2072072072072075e-06,
      "loss": 0.3909,
      "step": 22620
    },
    {
      "epoch": 67.96,
      "grad_norm": 4.625154495239258,
      "learning_rate": 3.204204204204204e-06,
      "loss": 0.3771,
      "step": 22630
    },
    {
      "epoch": 67.99,
      "grad_norm": 2.732945442199707,
      "learning_rate": 3.2012012012012016e-06,
      "loss": 0.3862,
      "step": 22640
    },
    {
      "epoch": 68.0,
      "eval_accuracy": 0.9193,
      "eval_loss": 0.3083004355430603,
      "eval_runtime": 26.2751,
      "eval_samples_per_second": 380.588,
      "eval_steps_per_second": 1.522,
      "step": 22644
    },
    {
      "epoch": 68.02,
      "grad_norm": 5.564293384552002,
      "learning_rate": 3.1981981981981987e-06,
      "loss": 0.4034,
      "step": 22650
    },
    {
      "epoch": 68.05,
      "grad_norm": 4.967087745666504,
      "learning_rate": 3.1951951951951953e-06,
      "loss": 0.3893,
      "step": 22660
    },
    {
      "epoch": 68.08,
      "grad_norm": 3.683514356613159,
      "learning_rate": 3.1921921921921924e-06,
      "loss": 0.3786,
      "step": 22670
    },
    {
      "epoch": 68.11,
      "grad_norm": 4.66683292388916,
      "learning_rate": 3.1891891891891894e-06,
      "loss": 0.3645,
      "step": 22680
    },
    {
      "epoch": 68.14,
      "grad_norm": 4.635051727294922,
      "learning_rate": 3.1861861861861865e-06,
      "loss": 0.3877,
      "step": 22690
    },
    {
      "epoch": 68.17,
      "grad_norm": 14.731842994689941,
      "learning_rate": 3.183183183183183e-06,
      "loss": 0.3877,
      "step": 22700
    },
    {
      "epoch": 68.2,
      "grad_norm": 3.941525936126709,
      "learning_rate": 3.1801801801801806e-06,
      "loss": 0.3906,
      "step": 22710
    },
    {
      "epoch": 68.23,
      "grad_norm": 5.458199977874756,
      "learning_rate": 3.1771771771771776e-06,
      "loss": 0.3909,
      "step": 22720
    },
    {
      "epoch": 68.26,
      "grad_norm": 5.221258163452148,
      "learning_rate": 3.1741741741741743e-06,
      "loss": 0.3795,
      "step": 22730
    },
    {
      "epoch": 68.29,
      "grad_norm": 5.936361789703369,
      "learning_rate": 3.1711711711711713e-06,
      "loss": 0.3356,
      "step": 22740
    },
    {
      "epoch": 68.32,
      "grad_norm": 4.711806297302246,
      "learning_rate": 3.168168168168169e-06,
      "loss": 0.3908,
      "step": 22750
    },
    {
      "epoch": 68.35,
      "grad_norm": 4.556100845336914,
      "learning_rate": 3.1651651651651654e-06,
      "loss": 0.397,
      "step": 22760
    },
    {
      "epoch": 68.38,
      "grad_norm": 6.554943084716797,
      "learning_rate": 3.1621621621621625e-06,
      "loss": 0.3827,
      "step": 22770
    },
    {
      "epoch": 68.41,
      "grad_norm": 6.468562602996826,
      "learning_rate": 3.159159159159159e-06,
      "loss": 0.3972,
      "step": 22780
    },
    {
      "epoch": 68.44,
      "grad_norm": 6.863269805908203,
      "learning_rate": 3.1561561561561566e-06,
      "loss": 0.3542,
      "step": 22790
    },
    {
      "epoch": 68.47,
      "grad_norm": 5.451284408569336,
      "learning_rate": 3.1531531531531532e-06,
      "loss": 0.3513,
      "step": 22800
    },
    {
      "epoch": 68.5,
      "grad_norm": 3.8810582160949707,
      "learning_rate": 3.1501501501501503e-06,
      "loss": 0.3579,
      "step": 22810
    },
    {
      "epoch": 68.53,
      "grad_norm": 4.597529888153076,
      "learning_rate": 3.1471471471471478e-06,
      "loss": 0.4145,
      "step": 22820
    },
    {
      "epoch": 68.56,
      "grad_norm": 5.767913341522217,
      "learning_rate": 3.1441441441441444e-06,
      "loss": 0.371,
      "step": 22830
    },
    {
      "epoch": 68.59,
      "grad_norm": 4.868569850921631,
      "learning_rate": 3.1411411411411415e-06,
      "loss": 0.3706,
      "step": 22840
    },
    {
      "epoch": 68.62,
      "grad_norm": 3.964306354522705,
      "learning_rate": 3.138138138138138e-06,
      "loss": 0.4271,
      "step": 22850
    },
    {
      "epoch": 68.65,
      "grad_norm": 4.96278190612793,
      "learning_rate": 3.1351351351351356e-06,
      "loss": 0.3609,
      "step": 22860
    },
    {
      "epoch": 68.68,
      "grad_norm": 3.673582077026367,
      "learning_rate": 3.132132132132132e-06,
      "loss": 0.356,
      "step": 22870
    },
    {
      "epoch": 68.71,
      "grad_norm": 9.806784629821777,
      "learning_rate": 3.1291291291291293e-06,
      "loss": 0.3857,
      "step": 22880
    },
    {
      "epoch": 68.74,
      "grad_norm": 4.269336223602295,
      "learning_rate": 3.1261261261261267e-06,
      "loss": 0.3291,
      "step": 22890
    },
    {
      "epoch": 68.77,
      "grad_norm": 3.036133050918579,
      "learning_rate": 3.1231231231231234e-06,
      "loss": 0.3236,
      "step": 22900
    },
    {
      "epoch": 68.8,
      "grad_norm": 5.000271797180176,
      "learning_rate": 3.1201201201201204e-06,
      "loss": 0.3881,
      "step": 22910
    },
    {
      "epoch": 68.83,
      "grad_norm": 5.859011650085449,
      "learning_rate": 3.117117117117117e-06,
      "loss": 0.3749,
      "step": 22920
    },
    {
      "epoch": 68.86,
      "grad_norm": 2.9163527488708496,
      "learning_rate": 3.1141141141141145e-06,
      "loss": 0.3456,
      "step": 22930
    },
    {
      "epoch": 68.89,
      "grad_norm": 4.472874164581299,
      "learning_rate": 3.1111111111111116e-06,
      "loss": 0.3342,
      "step": 22940
    },
    {
      "epoch": 68.92,
      "grad_norm": 5.105175971984863,
      "learning_rate": 3.1081081081081082e-06,
      "loss": 0.3566,
      "step": 22950
    },
    {
      "epoch": 68.95,
      "grad_norm": 3.8452229499816895,
      "learning_rate": 3.1051051051051053e-06,
      "loss": 0.3531,
      "step": 22960
    },
    {
      "epoch": 68.98,
      "grad_norm": 5.9138875007629395,
      "learning_rate": 3.1021021021021023e-06,
      "loss": 0.3519,
      "step": 22970
    },
    {
      "epoch": 69.0,
      "eval_accuracy": 0.9187,
      "eval_loss": 0.31031349301338196,
      "eval_runtime": 26.3789,
      "eval_samples_per_second": 379.091,
      "eval_steps_per_second": 1.516,
      "step": 22977
    },
    {
      "epoch": 69.01,
      "grad_norm": 3.973846197128296,
      "learning_rate": 3.0990990990990994e-06,
      "loss": 0.3762,
      "step": 22980
    },
    {
      "epoch": 69.04,
      "grad_norm": 5.474660873413086,
      "learning_rate": 3.096096096096096e-06,
      "loss": 0.3483,
      "step": 22990
    },
    {
      "epoch": 69.07,
      "grad_norm": 6.07138204574585,
      "learning_rate": 3.0930930930930935e-06,
      "loss": 0.3672,
      "step": 23000
    },
    {
      "epoch": 69.1,
      "grad_norm": 4.198864459991455,
      "learning_rate": 3.0900900900900905e-06,
      "loss": 0.3642,
      "step": 23010
    },
    {
      "epoch": 69.13,
      "grad_norm": 4.192826747894287,
      "learning_rate": 3.087087087087087e-06,
      "loss": 0.3275,
      "step": 23020
    },
    {
      "epoch": 69.16,
      "grad_norm": 3.3911166191101074,
      "learning_rate": 3.0840840840840842e-06,
      "loss": 0.429,
      "step": 23030
    },
    {
      "epoch": 69.19,
      "grad_norm": 3.8359243869781494,
      "learning_rate": 3.0810810810810817e-06,
      "loss": 0.3476,
      "step": 23040
    },
    {
      "epoch": 69.22,
      "grad_norm": 5.238132953643799,
      "learning_rate": 3.0780780780780783e-06,
      "loss": 0.416,
      "step": 23050
    },
    {
      "epoch": 69.25,
      "grad_norm": 4.762311935424805,
      "learning_rate": 3.0750750750750754e-06,
      "loss": 0.3151,
      "step": 23060
    },
    {
      "epoch": 69.28,
      "grad_norm": 4.5570969581604,
      "learning_rate": 3.0720720720720725e-06,
      "loss": 0.399,
      "step": 23070
    },
    {
      "epoch": 69.31,
      "grad_norm": 4.111698627471924,
      "learning_rate": 3.0690690690690695e-06,
      "loss": 0.3638,
      "step": 23080
    },
    {
      "epoch": 69.34,
      "grad_norm": 4.29660701751709,
      "learning_rate": 3.066066066066066e-06,
      "loss": 0.3516,
      "step": 23090
    },
    {
      "epoch": 69.37,
      "grad_norm": 5.011861801147461,
      "learning_rate": 3.063063063063063e-06,
      "loss": 0.3725,
      "step": 23100
    },
    {
      "epoch": 69.4,
      "grad_norm": 4.56400728225708,
      "learning_rate": 3.0600600600600607e-06,
      "loss": 0.3831,
      "step": 23110
    },
    {
      "epoch": 69.43,
      "grad_norm": 4.996874809265137,
      "learning_rate": 3.0570570570570573e-06,
      "loss": 0.3369,
      "step": 23120
    },
    {
      "epoch": 69.46,
      "grad_norm": 3.691049814224243,
      "learning_rate": 3.0540540540540544e-06,
      "loss": 0.3672,
      "step": 23130
    },
    {
      "epoch": 69.49,
      "grad_norm": 5.456169605255127,
      "learning_rate": 3.051051051051051e-06,
      "loss": 0.3882,
      "step": 23140
    },
    {
      "epoch": 69.52,
      "grad_norm": 2.7542948722839355,
      "learning_rate": 3.0480480480480485e-06,
      "loss": 0.3509,
      "step": 23150
    },
    {
      "epoch": 69.55,
      "grad_norm": 3.245677947998047,
      "learning_rate": 3.045045045045045e-06,
      "loss": 0.3047,
      "step": 23160
    },
    {
      "epoch": 69.58,
      "grad_norm": 4.396929740905762,
      "learning_rate": 3.042042042042042e-06,
      "loss": 0.3183,
      "step": 23170
    },
    {
      "epoch": 69.61,
      "grad_norm": 4.586196422576904,
      "learning_rate": 3.0390390390390396e-06,
      "loss": 0.3541,
      "step": 23180
    },
    {
      "epoch": 69.64,
      "grad_norm": 3.4663100242614746,
      "learning_rate": 3.0360360360360363e-06,
      "loss": 0.374,
      "step": 23190
    },
    {
      "epoch": 69.67,
      "grad_norm": 3.6912474632263184,
      "learning_rate": 3.0330330330330333e-06,
      "loss": 0.3673,
      "step": 23200
    },
    {
      "epoch": 69.7,
      "grad_norm": 4.935665607452393,
      "learning_rate": 3.03003003003003e-06,
      "loss": 0.3639,
      "step": 23210
    },
    {
      "epoch": 69.73,
      "grad_norm": 6.210367202758789,
      "learning_rate": 3.0270270270270274e-06,
      "loss": 0.3665,
      "step": 23220
    },
    {
      "epoch": 69.76,
      "grad_norm": 3.692854881286621,
      "learning_rate": 3.0240240240240245e-06,
      "loss": 0.3311,
      "step": 23230
    },
    {
      "epoch": 69.79,
      "grad_norm": 9.0880765914917,
      "learning_rate": 3.021021021021021e-06,
      "loss": 0.3744,
      "step": 23240
    },
    {
      "epoch": 69.82,
      "grad_norm": 3.73091721534729,
      "learning_rate": 3.0180180180180186e-06,
      "loss": 0.3882,
      "step": 23250
    },
    {
      "epoch": 69.85,
      "grad_norm": 5.430549621582031,
      "learning_rate": 3.0150150150150152e-06,
      "loss": 0.3772,
      "step": 23260
    },
    {
      "epoch": 69.88,
      "grad_norm": 4.998918056488037,
      "learning_rate": 3.0120120120120123e-06,
      "loss": 0.3558,
      "step": 23270
    },
    {
      "epoch": 69.91,
      "grad_norm": 4.7240190505981445,
      "learning_rate": 3.009009009009009e-06,
      "loss": 0.4005,
      "step": 23280
    },
    {
      "epoch": 69.94,
      "grad_norm": 2.821683406829834,
      "learning_rate": 3.0060060060060064e-06,
      "loss": 0.3896,
      "step": 23290
    },
    {
      "epoch": 69.97,
      "grad_norm": 3.8931691646575928,
      "learning_rate": 3.0030030030030034e-06,
      "loss": 0.376,
      "step": 23300
    },
    {
      "epoch": 70.0,
      "grad_norm": 40.32636260986328,
      "learning_rate": 3e-06,
      "loss": 0.4464,
      "step": 23310
    },
    {
      "epoch": 70.0,
      "eval_accuracy": 0.9192,
      "eval_loss": 0.3111260235309601,
      "eval_runtime": 26.4679,
      "eval_samples_per_second": 377.817,
      "eval_steps_per_second": 1.511,
      "step": 23310
    },
    {
      "epoch": 70.03,
      "grad_norm": 3.9731593132019043,
      "learning_rate": 2.996996996996997e-06,
      "loss": 0.3878,
      "step": 23320
    },
    {
      "epoch": 70.06,
      "grad_norm": 7.838066577911377,
      "learning_rate": 2.993993993993994e-06,
      "loss": 0.3697,
      "step": 23330
    },
    {
      "epoch": 70.09,
      "grad_norm": 4.132843017578125,
      "learning_rate": 2.9909909909909912e-06,
      "loss": 0.3411,
      "step": 23340
    },
    {
      "epoch": 70.12,
      "grad_norm": 5.701239109039307,
      "learning_rate": 2.987987987987988e-06,
      "loss": 0.33,
      "step": 23350
    },
    {
      "epoch": 70.15,
      "grad_norm": 4.622178554534912,
      "learning_rate": 2.9849849849849854e-06,
      "loss": 0.372,
      "step": 23360
    },
    {
      "epoch": 70.18,
      "grad_norm": 4.944065570831299,
      "learning_rate": 2.9819819819819824e-06,
      "loss": 0.3609,
      "step": 23370
    },
    {
      "epoch": 70.21,
      "grad_norm": 3.9269816875457764,
      "learning_rate": 2.978978978978979e-06,
      "loss": 0.3408,
      "step": 23380
    },
    {
      "epoch": 70.24,
      "grad_norm": 3.354602813720703,
      "learning_rate": 2.975975975975976e-06,
      "loss": 0.3953,
      "step": 23390
    },
    {
      "epoch": 70.27,
      "grad_norm": 6.247049331665039,
      "learning_rate": 2.9729729729729736e-06,
      "loss": 0.3691,
      "step": 23400
    },
    {
      "epoch": 70.3,
      "grad_norm": 4.222929954528809,
      "learning_rate": 2.96996996996997e-06,
      "loss": 0.3884,
      "step": 23410
    },
    {
      "epoch": 70.33,
      "grad_norm": 4.59262228012085,
      "learning_rate": 2.9669669669669673e-06,
      "loss": 0.3514,
      "step": 23420
    },
    {
      "epoch": 70.36,
      "grad_norm": 4.966153144836426,
      "learning_rate": 2.9639639639639643e-06,
      "loss": 0.3235,
      "step": 23430
    },
    {
      "epoch": 70.39,
      "grad_norm": 3.7056405544281006,
      "learning_rate": 2.9609609609609614e-06,
      "loss": 0.3658,
      "step": 23440
    },
    {
      "epoch": 70.42,
      "grad_norm": 6.478189468383789,
      "learning_rate": 2.957957957957958e-06,
      "loss": 0.3677,
      "step": 23450
    },
    {
      "epoch": 70.45,
      "grad_norm": 3.6061854362487793,
      "learning_rate": 2.954954954954955e-06,
      "loss": 0.3186,
      "step": 23460
    },
    {
      "epoch": 70.48,
      "grad_norm": 6.0526251792907715,
      "learning_rate": 2.9519519519519525e-06,
      "loss": 0.3494,
      "step": 23470
    },
    {
      "epoch": 70.51,
      "grad_norm": 5.661704063415527,
      "learning_rate": 2.948948948948949e-06,
      "loss": 0.3664,
      "step": 23480
    },
    {
      "epoch": 70.54,
      "grad_norm": 4.902756214141846,
      "learning_rate": 2.9459459459459462e-06,
      "loss": 0.4169,
      "step": 23490
    },
    {
      "epoch": 70.57,
      "grad_norm": 3.9421026706695557,
      "learning_rate": 2.942942942942943e-06,
      "loss": 0.3171,
      "step": 23500
    },
    {
      "epoch": 70.6,
      "grad_norm": 4.2920122146606445,
      "learning_rate": 2.9399399399399403e-06,
      "loss": 0.3761,
      "step": 23510
    },
    {
      "epoch": 70.63,
      "grad_norm": 5.310624599456787,
      "learning_rate": 2.936936936936937e-06,
      "loss": 0.3636,
      "step": 23520
    },
    {
      "epoch": 70.66,
      "grad_norm": 4.217187881469727,
      "learning_rate": 2.933933933933934e-06,
      "loss": 0.3548,
      "step": 23530
    },
    {
      "epoch": 70.69,
      "grad_norm": 5.018453598022461,
      "learning_rate": 2.9309309309309315e-06,
      "loss": 0.3869,
      "step": 23540
    },
    {
      "epoch": 70.72,
      "grad_norm": 4.380861282348633,
      "learning_rate": 2.927927927927928e-06,
      "loss": 0.3467,
      "step": 23550
    },
    {
      "epoch": 70.75,
      "grad_norm": 5.190391540527344,
      "learning_rate": 2.924924924924925e-06,
      "loss": 0.2929,
      "step": 23560
    },
    {
      "epoch": 70.78,
      "grad_norm": 2.5751378536224365,
      "learning_rate": 2.921921921921922e-06,
      "loss": 0.3417,
      "step": 23570
    },
    {
      "epoch": 70.81,
      "grad_norm": 2.795762777328491,
      "learning_rate": 2.9189189189189193e-06,
      "loss": 0.3418,
      "step": 23580
    },
    {
      "epoch": 70.84,
      "grad_norm": 3.9604082107543945,
      "learning_rate": 2.9159159159159163e-06,
      "loss": 0.3822,
      "step": 23590
    },
    {
      "epoch": 70.87,
      "grad_norm": 3.7368993759155273,
      "learning_rate": 2.912912912912913e-06,
      "loss": 0.327,
      "step": 23600
    },
    {
      "epoch": 70.9,
      "grad_norm": 3.057781457901001,
      "learning_rate": 2.9099099099099105e-06,
      "loss": 0.3426,
      "step": 23610
    },
    {
      "epoch": 70.93,
      "grad_norm": 5.832949638366699,
      "learning_rate": 2.906906906906907e-06,
      "loss": 0.4557,
      "step": 23620
    },
    {
      "epoch": 70.96,
      "grad_norm": 4.144787311553955,
      "learning_rate": 2.903903903903904e-06,
      "loss": 0.3128,
      "step": 23630
    },
    {
      "epoch": 70.99,
      "grad_norm": 4.221506595611572,
      "learning_rate": 2.9009009009009008e-06,
      "loss": 0.3852,
      "step": 23640
    },
    {
      "epoch": 71.0,
      "eval_accuracy": 0.919,
      "eval_loss": 0.31162288784980774,
      "eval_runtime": 26.0438,
      "eval_samples_per_second": 383.969,
      "eval_steps_per_second": 1.536,
      "step": 23643
    },
    {
      "epoch": 71.02,
      "grad_norm": 2.365196466445923,
      "learning_rate": 2.8978978978978983e-06,
      "loss": 0.4374,
      "step": 23650
    },
    {
      "epoch": 71.05,
      "grad_norm": 6.07986307144165,
      "learning_rate": 2.8948948948948953e-06,
      "loss": 0.3822,
      "step": 23660
    },
    {
      "epoch": 71.08,
      "grad_norm": 4.160985946655273,
      "learning_rate": 2.891891891891892e-06,
      "loss": 0.3916,
      "step": 23670
    },
    {
      "epoch": 71.11,
      "grad_norm": 3.393815040588379,
      "learning_rate": 2.888888888888889e-06,
      "loss": 0.3929,
      "step": 23680
    },
    {
      "epoch": 71.14,
      "grad_norm": 5.477105140686035,
      "learning_rate": 2.885885885885886e-06,
      "loss": 0.3759,
      "step": 23690
    },
    {
      "epoch": 71.17,
      "grad_norm": 8.547496795654297,
      "learning_rate": 2.882882882882883e-06,
      "loss": 0.3652,
      "step": 23700
    },
    {
      "epoch": 71.2,
      "grad_norm": 5.620762348175049,
      "learning_rate": 2.8798798798798797e-06,
      "loss": 0.3309,
      "step": 23710
    },
    {
      "epoch": 71.23,
      "grad_norm": 4.632009506225586,
      "learning_rate": 2.8768768768768772e-06,
      "loss": 0.3703,
      "step": 23720
    },
    {
      "epoch": 71.26,
      "grad_norm": 3.8131065368652344,
      "learning_rate": 2.8738738738738743e-06,
      "loss": 0.3579,
      "step": 23730
    },
    {
      "epoch": 71.29,
      "grad_norm": 4.285787105560303,
      "learning_rate": 2.870870870870871e-06,
      "loss": 0.4276,
      "step": 23740
    },
    {
      "epoch": 71.32,
      "grad_norm": 4.171909809112549,
      "learning_rate": 2.867867867867868e-06,
      "loss": 0.3143,
      "step": 23750
    },
    {
      "epoch": 71.35,
      "grad_norm": 2.8653926849365234,
      "learning_rate": 2.8648648648648654e-06,
      "loss": 0.3509,
      "step": 23760
    },
    {
      "epoch": 71.38,
      "grad_norm": 4.054377555847168,
      "learning_rate": 2.861861861861862e-06,
      "loss": 0.3636,
      "step": 23770
    },
    {
      "epoch": 71.41,
      "grad_norm": 4.700258731842041,
      "learning_rate": 2.858858858858859e-06,
      "loss": 0.3424,
      "step": 23780
    },
    {
      "epoch": 71.44,
      "grad_norm": 6.453307628631592,
      "learning_rate": 2.855855855855856e-06,
      "loss": 0.3354,
      "step": 23790
    },
    {
      "epoch": 71.47,
      "grad_norm": 6.066850662231445,
      "learning_rate": 2.8528528528528532e-06,
      "loss": 0.3579,
      "step": 23800
    },
    {
      "epoch": 71.5,
      "grad_norm": 5.375420093536377,
      "learning_rate": 2.84984984984985e-06,
      "loss": 0.4144,
      "step": 23810
    },
    {
      "epoch": 71.53,
      "grad_norm": 4.179471492767334,
      "learning_rate": 2.846846846846847e-06,
      "loss": 0.365,
      "step": 23820
    },
    {
      "epoch": 71.56,
      "grad_norm": 8.616321563720703,
      "learning_rate": 2.8438438438438444e-06,
      "loss": 0.4177,
      "step": 23830
    },
    {
      "epoch": 71.59,
      "grad_norm": 5.9217939376831055,
      "learning_rate": 2.840840840840841e-06,
      "loss": 0.3939,
      "step": 23840
    },
    {
      "epoch": 71.62,
      "grad_norm": 3.257387161254883,
      "learning_rate": 2.837837837837838e-06,
      "loss": 0.3258,
      "step": 23850
    },
    {
      "epoch": 71.65,
      "grad_norm": 5.62135648727417,
      "learning_rate": 2.8348348348348347e-06,
      "loss": 0.3521,
      "step": 23860
    },
    {
      "epoch": 71.68,
      "grad_norm": 3.2584187984466553,
      "learning_rate": 2.831831831831832e-06,
      "loss": 0.3409,
      "step": 23870
    },
    {
      "epoch": 71.71,
      "grad_norm": 4.696722030639648,
      "learning_rate": 2.828828828828829e-06,
      "loss": 0.3877,
      "step": 23880
    },
    {
      "epoch": 71.74,
      "grad_norm": 4.042671203613281,
      "learning_rate": 2.825825825825826e-06,
      "loss": 0.3315,
      "step": 23890
    },
    {
      "epoch": 71.77,
      "grad_norm": 4.876395225524902,
      "learning_rate": 2.8228228228228234e-06,
      "loss": 0.3971,
      "step": 23900
    },
    {
      "epoch": 71.8,
      "grad_norm": 5.139949321746826,
      "learning_rate": 2.81981981981982e-06,
      "loss": 0.3415,
      "step": 23910
    },
    {
      "epoch": 71.83,
      "grad_norm": 4.070745468139648,
      "learning_rate": 2.816816816816817e-06,
      "loss": 0.4153,
      "step": 23920
    },
    {
      "epoch": 71.86,
      "grad_norm": 5.952834129333496,
      "learning_rate": 2.8138138138138137e-06,
      "loss": 0.4075,
      "step": 23930
    },
    {
      "epoch": 71.89,
      "grad_norm": 4.8311567306518555,
      "learning_rate": 2.810810810810811e-06,
      "loss": 0.2916,
      "step": 23940
    },
    {
      "epoch": 71.92,
      "grad_norm": 3.5638105869293213,
      "learning_rate": 2.807807807807808e-06,
      "loss": 0.3902,
      "step": 23950
    },
    {
      "epoch": 71.95,
      "grad_norm": 4.386170864105225,
      "learning_rate": 2.804804804804805e-06,
      "loss": 0.3362,
      "step": 23960
    },
    {
      "epoch": 71.98,
      "grad_norm": 5.6449689865112305,
      "learning_rate": 2.8018018018018023e-06,
      "loss": 0.3406,
      "step": 23970
    },
    {
      "epoch": 72.0,
      "eval_accuracy": 0.9194,
      "eval_loss": 0.3081548511981964,
      "eval_runtime": 26.3193,
      "eval_samples_per_second": 379.949,
      "eval_steps_per_second": 1.52,
      "step": 23976
    },
    {
      "epoch": 72.01,
      "grad_norm": 4.679892063140869,
      "learning_rate": 2.798798798798799e-06,
      "loss": 0.4127,
      "step": 23980
    },
    {
      "epoch": 72.04,
      "grad_norm": 5.134451389312744,
      "learning_rate": 2.795795795795796e-06,
      "loss": 0.3872,
      "step": 23990
    },
    {
      "epoch": 72.07,
      "grad_norm": 6.398600101470947,
      "learning_rate": 2.7927927927927926e-06,
      "loss": 0.3261,
      "step": 24000
    },
    {
      "epoch": 72.1,
      "grad_norm": 16.747257232666016,
      "learning_rate": 2.78978978978979e-06,
      "loss": 0.3833,
      "step": 24010
    },
    {
      "epoch": 72.13,
      "grad_norm": 3.9407284259796143,
      "learning_rate": 2.786786786786787e-06,
      "loss": 0.3608,
      "step": 24020
    },
    {
      "epoch": 72.16,
      "grad_norm": 5.557214260101318,
      "learning_rate": 2.783783783783784e-06,
      "loss": 0.3558,
      "step": 24030
    },
    {
      "epoch": 72.19,
      "grad_norm": 3.671473979949951,
      "learning_rate": 2.780780780780781e-06,
      "loss": 0.379,
      "step": 24040
    },
    {
      "epoch": 72.22,
      "grad_norm": 3.5799529552459717,
      "learning_rate": 2.7777777777777783e-06,
      "loss": 0.3701,
      "step": 24050
    },
    {
      "epoch": 72.25,
      "grad_norm": 4.3985772132873535,
      "learning_rate": 2.774774774774775e-06,
      "loss": 0.3701,
      "step": 24060
    },
    {
      "epoch": 72.28,
      "grad_norm": 3.9684081077575684,
      "learning_rate": 2.771771771771772e-06,
      "loss": 0.3561,
      "step": 24070
    },
    {
      "epoch": 72.31,
      "grad_norm": 3.5902743339538574,
      "learning_rate": 2.768768768768769e-06,
      "loss": 0.3473,
      "step": 24080
    },
    {
      "epoch": 72.34,
      "grad_norm": 3.9501242637634277,
      "learning_rate": 2.765765765765766e-06,
      "loss": 0.3565,
      "step": 24090
    },
    {
      "epoch": 72.37,
      "grad_norm": 3.774057388305664,
      "learning_rate": 2.7627627627627628e-06,
      "loss": 0.3472,
      "step": 24100
    },
    {
      "epoch": 72.4,
      "grad_norm": 3.3087427616119385,
      "learning_rate": 2.75975975975976e-06,
      "loss": 0.4142,
      "step": 24110
    },
    {
      "epoch": 72.43,
      "grad_norm": 3.7041425704956055,
      "learning_rate": 2.7567567567567573e-06,
      "loss": 0.3656,
      "step": 24120
    },
    {
      "epoch": 72.46,
      "grad_norm": 4.670281410217285,
      "learning_rate": 2.753753753753754e-06,
      "loss": 0.401,
      "step": 24130
    },
    {
      "epoch": 72.49,
      "grad_norm": 4.1876983642578125,
      "learning_rate": 2.750750750750751e-06,
      "loss": 0.4207,
      "step": 24140
    },
    {
      "epoch": 72.52,
      "grad_norm": 4.032341003417969,
      "learning_rate": 2.747747747747748e-06,
      "loss": 0.3932,
      "step": 24150
    },
    {
      "epoch": 72.55,
      "grad_norm": 4.460512638092041,
      "learning_rate": 2.744744744744745e-06,
      "loss": 0.3817,
      "step": 24160
    },
    {
      "epoch": 72.58,
      "grad_norm": 3.9044270515441895,
      "learning_rate": 2.7417417417417417e-06,
      "loss": 0.3497,
      "step": 24170
    },
    {
      "epoch": 72.61,
      "grad_norm": 3.0837409496307373,
      "learning_rate": 2.7387387387387388e-06,
      "loss": 0.3796,
      "step": 24180
    },
    {
      "epoch": 72.64,
      "grad_norm": 4.316883087158203,
      "learning_rate": 2.7357357357357363e-06,
      "loss": 0.3986,
      "step": 24190
    },
    {
      "epoch": 72.67,
      "grad_norm": 3.4387145042419434,
      "learning_rate": 2.732732732732733e-06,
      "loss": 0.4095,
      "step": 24200
    },
    {
      "epoch": 72.7,
      "grad_norm": 3.7125887870788574,
      "learning_rate": 2.72972972972973e-06,
      "loss": 0.2805,
      "step": 24210
    },
    {
      "epoch": 72.73,
      "grad_norm": 4.373105049133301,
      "learning_rate": 2.7267267267267274e-06,
      "loss": 0.3993,
      "step": 24220
    },
    {
      "epoch": 72.76,
      "grad_norm": 4.775018692016602,
      "learning_rate": 2.723723723723724e-06,
      "loss": 0.376,
      "step": 24230
    },
    {
      "epoch": 72.79,
      "grad_norm": 4.634377479553223,
      "learning_rate": 2.720720720720721e-06,
      "loss": 0.3418,
      "step": 24240
    },
    {
      "epoch": 72.82,
      "grad_norm": 4.547117710113525,
      "learning_rate": 2.7177177177177177e-06,
      "loss": 0.3485,
      "step": 24250
    },
    {
      "epoch": 72.85,
      "grad_norm": 4.406383514404297,
      "learning_rate": 2.7147147147147152e-06,
      "loss": 0.3562,
      "step": 24260
    },
    {
      "epoch": 72.88,
      "grad_norm": 4.838225364685059,
      "learning_rate": 2.711711711711712e-06,
      "loss": 0.3763,
      "step": 24270
    },
    {
      "epoch": 72.91,
      "grad_norm": 3.7401673793792725,
      "learning_rate": 2.708708708708709e-06,
      "loss": 0.3459,
      "step": 24280
    },
    {
      "epoch": 72.94,
      "grad_norm": 4.0809502601623535,
      "learning_rate": 2.7057057057057055e-06,
      "loss": 0.3862,
      "step": 24290
    },
    {
      "epoch": 72.97,
      "grad_norm": 3.01916766166687,
      "learning_rate": 2.702702702702703e-06,
      "loss": 0.3785,
      "step": 24300
    },
    {
      "epoch": 73.0,
      "eval_accuracy": 0.9191,
      "eval_loss": 0.3070811927318573,
      "eval_runtime": 26.4006,
      "eval_samples_per_second": 378.779,
      "eval_steps_per_second": 1.515,
      "step": 24309
    },
    {
      "epoch": 73.0,
      "grad_norm": 3.6586384773254395,
      "learning_rate": 2.6996996996997e-06,
      "loss": 0.3352,
      "step": 24310
    },
    {
      "epoch": 73.03,
      "grad_norm": 5.579931259155273,
      "learning_rate": 2.6966966966966967e-06,
      "loss": 0.3595,
      "step": 24320
    },
    {
      "epoch": 73.06,
      "grad_norm": 4.494871616363525,
      "learning_rate": 2.693693693693694e-06,
      "loss": 0.4096,
      "step": 24330
    },
    {
      "epoch": 73.09,
      "grad_norm": 5.282838344573975,
      "learning_rate": 2.690690690690691e-06,
      "loss": 0.3295,
      "step": 24340
    },
    {
      "epoch": 73.12,
      "grad_norm": 4.90192985534668,
      "learning_rate": 2.687687687687688e-06,
      "loss": 0.3473,
      "step": 24350
    },
    {
      "epoch": 73.15,
      "grad_norm": 4.061066150665283,
      "learning_rate": 2.6846846846846845e-06,
      "loss": 0.3752,
      "step": 24360
    },
    {
      "epoch": 73.18,
      "grad_norm": 5.1341352462768555,
      "learning_rate": 2.681681681681682e-06,
      "loss": 0.3823,
      "step": 24370
    },
    {
      "epoch": 73.21,
      "grad_norm": 3.0099151134490967,
      "learning_rate": 2.678678678678679e-06,
      "loss": 0.3926,
      "step": 24380
    },
    {
      "epoch": 73.24,
      "grad_norm": 4.0567522048950195,
      "learning_rate": 2.6756756756756757e-06,
      "loss": 0.3605,
      "step": 24390
    },
    {
      "epoch": 73.27,
      "grad_norm": 4.276137828826904,
      "learning_rate": 2.672672672672673e-06,
      "loss": 0.3787,
      "step": 24400
    },
    {
      "epoch": 73.3,
      "grad_norm": 5.335641384124756,
      "learning_rate": 2.66966966966967e-06,
      "loss": 0.3677,
      "step": 24410
    },
    {
      "epoch": 73.33,
      "grad_norm": 5.405757427215576,
      "learning_rate": 2.666666666666667e-06,
      "loss": 0.3923,
      "step": 24420
    },
    {
      "epoch": 73.36,
      "grad_norm": 11.899343490600586,
      "learning_rate": 2.663663663663664e-06,
      "loss": 0.3983,
      "step": 24430
    },
    {
      "epoch": 73.39,
      "grad_norm": 8.977961540222168,
      "learning_rate": 2.660660660660661e-06,
      "loss": 0.3652,
      "step": 24440
    },
    {
      "epoch": 73.42,
      "grad_norm": 3.4779210090637207,
      "learning_rate": 2.657657657657658e-06,
      "loss": 0.3813,
      "step": 24450
    },
    {
      "epoch": 73.45,
      "grad_norm": 8.737194061279297,
      "learning_rate": 2.6546546546546546e-06,
      "loss": 0.4058,
      "step": 24460
    },
    {
      "epoch": 73.48,
      "grad_norm": 9.307788848876953,
      "learning_rate": 2.6516516516516517e-06,
      "loss": 0.3704,
      "step": 24470
    },
    {
      "epoch": 73.51,
      "grad_norm": 5.806252479553223,
      "learning_rate": 2.648648648648649e-06,
      "loss": 0.4308,
      "step": 24480
    },
    {
      "epoch": 73.54,
      "grad_norm": 2.473646879196167,
      "learning_rate": 2.645645645645646e-06,
      "loss": 0.35,
      "step": 24490
    },
    {
      "epoch": 73.57,
      "grad_norm": 4.046924591064453,
      "learning_rate": 2.642642642642643e-06,
      "loss": 0.3772,
      "step": 24500
    },
    {
      "epoch": 73.6,
      "grad_norm": 7.512773513793945,
      "learning_rate": 2.63963963963964e-06,
      "loss": 0.352,
      "step": 24510
    },
    {
      "epoch": 73.63,
      "grad_norm": 4.16456937789917,
      "learning_rate": 2.636636636636637e-06,
      "loss": 0.309,
      "step": 24520
    },
    {
      "epoch": 73.66,
      "grad_norm": 5.551898956298828,
      "learning_rate": 2.6336336336336336e-06,
      "loss": 0.3831,
      "step": 24530
    },
    {
      "epoch": 73.69,
      "grad_norm": 3.978297710418701,
      "learning_rate": 2.6306306306306306e-06,
      "loss": 0.3443,
      "step": 24540
    },
    {
      "epoch": 73.72,
      "grad_norm": 4.803989410400391,
      "learning_rate": 2.627627627627628e-06,
      "loss": 0.3654,
      "step": 24550
    },
    {
      "epoch": 73.75,
      "grad_norm": 3.644221305847168,
      "learning_rate": 2.6246246246246248e-06,
      "loss": 0.3877,
      "step": 24560
    },
    {
      "epoch": 73.78,
      "grad_norm": 4.657214641571045,
      "learning_rate": 2.621621621621622e-06,
      "loss": 0.359,
      "step": 24570
    },
    {
      "epoch": 73.81,
      "grad_norm": 3.704192638397217,
      "learning_rate": 2.6186186186186193e-06,
      "loss": 0.3892,
      "step": 24580
    },
    {
      "epoch": 73.84,
      "grad_norm": 7.370591163635254,
      "learning_rate": 2.615615615615616e-06,
      "loss": 0.3912,
      "step": 24590
    },
    {
      "epoch": 73.87,
      "grad_norm": 4.725093364715576,
      "learning_rate": 2.612612612612613e-06,
      "loss": 0.3768,
      "step": 24600
    },
    {
      "epoch": 73.9,
      "grad_norm": 4.5920305252075195,
      "learning_rate": 2.6096096096096096e-06,
      "loss": 0.3567,
      "step": 24610
    },
    {
      "epoch": 73.93,
      "grad_norm": 4.506215572357178,
      "learning_rate": 2.606606606606607e-06,
      "loss": 0.3646,
      "step": 24620
    },
    {
      "epoch": 73.96,
      "grad_norm": 4.7424798011779785,
      "learning_rate": 2.6036036036036037e-06,
      "loss": 0.3434,
      "step": 24630
    },
    {
      "epoch": 73.99,
      "grad_norm": 4.18931770324707,
      "learning_rate": 2.6006006006006008e-06,
      "loss": 0.3559,
      "step": 24640
    },
    {
      "epoch": 74.0,
      "eval_accuracy": 0.9194,
      "eval_loss": 0.3101232945919037,
      "eval_runtime": 26.227,
      "eval_samples_per_second": 381.287,
      "eval_steps_per_second": 1.525,
      "step": 24642
    },
    {
      "epoch": 74.02,
      "grad_norm": 4.140317916870117,
      "learning_rate": 2.5975975975975974e-06,
      "loss": 0.3608,
      "step": 24650
    },
    {
      "epoch": 74.05,
      "grad_norm": 4.223869800567627,
      "learning_rate": 2.594594594594595e-06,
      "loss": 0.3763,
      "step": 24660
    },
    {
      "epoch": 74.08,
      "grad_norm": 5.73573112487793,
      "learning_rate": 2.591591591591592e-06,
      "loss": 0.3788,
      "step": 24670
    },
    {
      "epoch": 74.11,
      "grad_norm": 4.337681293487549,
      "learning_rate": 2.5885885885885886e-06,
      "loss": 0.3851,
      "step": 24680
    },
    {
      "epoch": 74.14,
      "grad_norm": 6.438630104064941,
      "learning_rate": 2.585585585585586e-06,
      "loss": 0.3659,
      "step": 24690
    },
    {
      "epoch": 74.17,
      "grad_norm": 5.986374378204346,
      "learning_rate": 2.5825825825825827e-06,
      "loss": 0.3512,
      "step": 24700
    },
    {
      "epoch": 74.2,
      "grad_norm": 3.377227306365967,
      "learning_rate": 2.5795795795795797e-06,
      "loss": 0.3861,
      "step": 24710
    },
    {
      "epoch": 74.23,
      "grad_norm": 4.813278675079346,
      "learning_rate": 2.5765765765765764e-06,
      "loss": 0.3533,
      "step": 24720
    },
    {
      "epoch": 74.26,
      "grad_norm": 3.9393057823181152,
      "learning_rate": 2.573573573573574e-06,
      "loss": 0.367,
      "step": 24730
    },
    {
      "epoch": 74.29,
      "grad_norm": 3.5958847999572754,
      "learning_rate": 2.570570570570571e-06,
      "loss": 0.366,
      "step": 24740
    },
    {
      "epoch": 74.32,
      "grad_norm": 5.89169979095459,
      "learning_rate": 2.5675675675675675e-06,
      "loss": 0.3618,
      "step": 24750
    },
    {
      "epoch": 74.35,
      "grad_norm": 4.103945732116699,
      "learning_rate": 2.564564564564565e-06,
      "loss": 0.373,
      "step": 24760
    },
    {
      "epoch": 74.38,
      "grad_norm": 2.91805362701416,
      "learning_rate": 2.561561561561562e-06,
      "loss": 0.4113,
      "step": 24770
    },
    {
      "epoch": 74.41,
      "grad_norm": 6.933594703674316,
      "learning_rate": 2.5585585585585587e-06,
      "loss": 0.3538,
      "step": 24780
    },
    {
      "epoch": 74.44,
      "grad_norm": 4.638087272644043,
      "learning_rate": 2.5555555555555557e-06,
      "loss": 0.3739,
      "step": 24790
    },
    {
      "epoch": 74.47,
      "grad_norm": 3.1220591068267822,
      "learning_rate": 2.552552552552553e-06,
      "loss": 0.3446,
      "step": 24800
    },
    {
      "epoch": 74.5,
      "grad_norm": 4.613528251647949,
      "learning_rate": 2.54954954954955e-06,
      "loss": 0.3703,
      "step": 24810
    },
    {
      "epoch": 74.53,
      "grad_norm": 3.160144805908203,
      "learning_rate": 2.5465465465465465e-06,
      "loss": 0.3895,
      "step": 24820
    },
    {
      "epoch": 74.56,
      "grad_norm": 5.139934062957764,
      "learning_rate": 2.5435435435435435e-06,
      "loss": 0.3345,
      "step": 24830
    },
    {
      "epoch": 74.59,
      "grad_norm": 6.216730117797852,
      "learning_rate": 2.540540540540541e-06,
      "loss": 0.3778,
      "step": 24840
    },
    {
      "epoch": 74.62,
      "grad_norm": 5.177805423736572,
      "learning_rate": 2.5375375375375377e-06,
      "loss": 0.3259,
      "step": 24850
    },
    {
      "epoch": 74.65,
      "grad_norm": 3.9643661975860596,
      "learning_rate": 2.5345345345345347e-06,
      "loss": 0.3087,
      "step": 24860
    },
    {
      "epoch": 74.68,
      "grad_norm": 5.304629802703857,
      "learning_rate": 2.5315315315315318e-06,
      "loss": 0.3246,
      "step": 24870
    },
    {
      "epoch": 74.71,
      "grad_norm": 5.237442970275879,
      "learning_rate": 2.528528528528529e-06,
      "loss": 0.2898,
      "step": 24880
    },
    {
      "epoch": 74.74,
      "grad_norm": 5.509350776672363,
      "learning_rate": 2.5255255255255254e-06,
      "loss": 0.3692,
      "step": 24890
    },
    {
      "epoch": 74.77,
      "grad_norm": 5.299699783325195,
      "learning_rate": 2.5225225225225225e-06,
      "loss": 0.3625,
      "step": 24900
    },
    {
      "epoch": 74.8,
      "grad_norm": 5.23745584487915,
      "learning_rate": 2.51951951951952e-06,
      "loss": 0.4172,
      "step": 24910
    },
    {
      "epoch": 74.83,
      "grad_norm": 5.137388706207275,
      "learning_rate": 2.5165165165165166e-06,
      "loss": 0.3309,
      "step": 24920
    },
    {
      "epoch": 74.86,
      "grad_norm": 4.968721866607666,
      "learning_rate": 2.5135135135135137e-06,
      "loss": 0.3191,
      "step": 24930
    },
    {
      "epoch": 74.89,
      "grad_norm": 3.8109066486358643,
      "learning_rate": 2.510510510510511e-06,
      "loss": 0.3734,
      "step": 24940
    },
    {
      "epoch": 74.92,
      "grad_norm": 3.3641743659973145,
      "learning_rate": 2.5075075075075078e-06,
      "loss": 0.3432,
      "step": 24950
    },
    {
      "epoch": 74.95,
      "grad_norm": 4.264516830444336,
      "learning_rate": 2.504504504504505e-06,
      "loss": 0.3545,
      "step": 24960
    },
    {
      "epoch": 74.98,
      "grad_norm": 2.8389031887054443,
      "learning_rate": 2.5015015015015015e-06,
      "loss": 0.3298,
      "step": 24970
    },
    {
      "epoch": 75.0,
      "eval_accuracy": 0.9187,
      "eval_loss": 0.30985313653945923,
      "eval_runtime": 26.3809,
      "eval_samples_per_second": 379.062,
      "eval_steps_per_second": 1.516,
      "step": 24975
    },
    {
      "epoch": 75.02,
      "grad_norm": 2.647611618041992,
      "learning_rate": 2.4984984984984985e-06,
      "loss": 0.2919,
      "step": 24980
    },
    {
      "epoch": 75.05,
      "grad_norm": 3.7018468379974365,
      "learning_rate": 2.4954954954954956e-06,
      "loss": 0.2919,
      "step": 24990
    },
    {
      "epoch": 75.08,
      "grad_norm": 4.203715801239014,
      "learning_rate": 2.4924924924924926e-06,
      "loss": 0.3988,
      "step": 25000
    },
    {
      "epoch": 75.11,
      "grad_norm": 5.609640598297119,
      "learning_rate": 2.4894894894894897e-06,
      "loss": 0.3893,
      "step": 25010
    },
    {
      "epoch": 75.14,
      "grad_norm": 3.516289234161377,
      "learning_rate": 2.4864864864864867e-06,
      "loss": 0.3464,
      "step": 25020
    },
    {
      "epoch": 75.17,
      "grad_norm": 5.230327129364014,
      "learning_rate": 2.483483483483484e-06,
      "loss": 0.3704,
      "step": 25030
    },
    {
      "epoch": 75.2,
      "grad_norm": 4.416625499725342,
      "learning_rate": 2.480480480480481e-06,
      "loss": 0.3151,
      "step": 25040
    },
    {
      "epoch": 75.23,
      "grad_norm": 5.022232532501221,
      "learning_rate": 2.4774774774774775e-06,
      "loss": 0.3687,
      "step": 25050
    },
    {
      "epoch": 75.26,
      "grad_norm": 4.635036468505859,
      "learning_rate": 2.474474474474475e-06,
      "loss": 0.3505,
      "step": 25060
    },
    {
      "epoch": 75.29,
      "grad_norm": 3.1952192783355713,
      "learning_rate": 2.4714714714714716e-06,
      "loss": 0.3569,
      "step": 25070
    },
    {
      "epoch": 75.32,
      "grad_norm": 3.883758306503296,
      "learning_rate": 2.4684684684684686e-06,
      "loss": 0.3184,
      "step": 25080
    },
    {
      "epoch": 75.35,
      "grad_norm": 4.174134254455566,
      "learning_rate": 2.4654654654654657e-06,
      "loss": 0.4082,
      "step": 25090
    },
    {
      "epoch": 75.38,
      "grad_norm": 4.038928985595703,
      "learning_rate": 2.4624624624624628e-06,
      "loss": 0.3544,
      "step": 25100
    },
    {
      "epoch": 75.41,
      "grad_norm": 6.403235912322998,
      "learning_rate": 2.45945945945946e-06,
      "loss": 0.4098,
      "step": 25110
    },
    {
      "epoch": 75.44,
      "grad_norm": 4.0694146156311035,
      "learning_rate": 2.4564564564564564e-06,
      "loss": 0.3831,
      "step": 25120
    },
    {
      "epoch": 75.47,
      "grad_norm": 3.7605605125427246,
      "learning_rate": 2.453453453453454e-06,
      "loss": 0.4216,
      "step": 25130
    },
    {
      "epoch": 75.5,
      "grad_norm": 3.5430238246917725,
      "learning_rate": 2.4504504504504506e-06,
      "loss": 0.3614,
      "step": 25140
    },
    {
      "epoch": 75.53,
      "grad_norm": 5.834523677825928,
      "learning_rate": 2.4474474474474476e-06,
      "loss": 0.3778,
      "step": 25150
    },
    {
      "epoch": 75.56,
      "grad_norm": 3.6949920654296875,
      "learning_rate": 2.4444444444444447e-06,
      "loss": 0.3514,
      "step": 25160
    },
    {
      "epoch": 75.59,
      "grad_norm": 7.654335021972656,
      "learning_rate": 2.4414414414414417e-06,
      "loss": 0.3471,
      "step": 25170
    },
    {
      "epoch": 75.62,
      "grad_norm": 4.019693851470947,
      "learning_rate": 2.4384384384384383e-06,
      "loss": 0.3615,
      "step": 25180
    },
    {
      "epoch": 75.65,
      "grad_norm": 4.266978740692139,
      "learning_rate": 2.435435435435436e-06,
      "loss": 0.3626,
      "step": 25190
    },
    {
      "epoch": 75.68,
      "grad_norm": 5.3567352294921875,
      "learning_rate": 2.432432432432433e-06,
      "loss": 0.4082,
      "step": 25200
    },
    {
      "epoch": 75.71,
      "grad_norm": 3.784461736679077,
      "learning_rate": 2.4294294294294295e-06,
      "loss": 0.3143,
      "step": 25210
    },
    {
      "epoch": 75.74,
      "grad_norm": 3.9390547275543213,
      "learning_rate": 2.4264264264264266e-06,
      "loss": 0.3511,
      "step": 25220
    },
    {
      "epoch": 75.77,
      "grad_norm": 5.29757833480835,
      "learning_rate": 2.4234234234234236e-06,
      "loss": 0.3547,
      "step": 25230
    },
    {
      "epoch": 75.8,
      "grad_norm": 4.362313747406006,
      "learning_rate": 2.4204204204204207e-06,
      "loss": 0.3792,
      "step": 25240
    },
    {
      "epoch": 75.83,
      "grad_norm": 3.7680556774139404,
      "learning_rate": 2.4174174174174177e-06,
      "loss": 0.3459,
      "step": 25250
    },
    {
      "epoch": 75.86,
      "grad_norm": 5.173712730407715,
      "learning_rate": 2.414414414414415e-06,
      "loss": 0.3662,
      "step": 25260
    },
    {
      "epoch": 75.89,
      "grad_norm": 5.308409690856934,
      "learning_rate": 2.4114114114114114e-06,
      "loss": 0.3574,
      "step": 25270
    },
    {
      "epoch": 75.92,
      "grad_norm": 5.02219295501709,
      "learning_rate": 2.4084084084084085e-06,
      "loss": 0.3948,
      "step": 25280
    },
    {
      "epoch": 75.95,
      "grad_norm": 3.983201742172241,
      "learning_rate": 2.4054054054054055e-06,
      "loss": 0.3709,
      "step": 25290
    },
    {
      "epoch": 75.98,
      "grad_norm": 4.726694583892822,
      "learning_rate": 2.4024024024024026e-06,
      "loss": 0.3596,
      "step": 25300
    },
    {
      "epoch": 76.0,
      "eval_accuracy": 0.9208,
      "eval_loss": 0.3099121153354645,
      "eval_runtime": 26.4034,
      "eval_samples_per_second": 378.739,
      "eval_steps_per_second": 1.515,
      "step": 25308
    },
    {
      "epoch": 76.01,
      "grad_norm": 3.8848838806152344,
      "learning_rate": 2.3993993993993996e-06,
      "loss": 0.3128,
      "step": 25310
    },
    {
      "epoch": 76.04,
      "grad_norm": 3.564084768295288,
      "learning_rate": 2.3963963963963967e-06,
      "loss": 0.3352,
      "step": 25320
    },
    {
      "epoch": 76.07,
      "grad_norm": 4.917530059814453,
      "learning_rate": 2.3933933933933938e-06,
      "loss": 0.3744,
      "step": 25330
    },
    {
      "epoch": 76.1,
      "grad_norm": 4.377292156219482,
      "learning_rate": 2.3903903903903904e-06,
      "loss": 0.3555,
      "step": 25340
    },
    {
      "epoch": 76.13,
      "grad_norm": 2.951019048690796,
      "learning_rate": 2.3873873873873874e-06,
      "loss": 0.3387,
      "step": 25350
    },
    {
      "epoch": 76.16,
      "grad_norm": 5.150670051574707,
      "learning_rate": 2.3843843843843845e-06,
      "loss": 0.3459,
      "step": 25360
    },
    {
      "epoch": 76.19,
      "grad_norm": 4.062009334564209,
      "learning_rate": 2.3813813813813815e-06,
      "loss": 0.3786,
      "step": 25370
    },
    {
      "epoch": 76.22,
      "grad_norm": 3.29832124710083,
      "learning_rate": 2.3783783783783786e-06,
      "loss": 0.3526,
      "step": 25380
    },
    {
      "epoch": 76.25,
      "grad_norm": 4.91969633102417,
      "learning_rate": 2.3753753753753757e-06,
      "loss": 0.3577,
      "step": 25390
    },
    {
      "epoch": 76.28,
      "grad_norm": 6.996547222137451,
      "learning_rate": 2.3723723723723727e-06,
      "loss": 0.3891,
      "step": 25400
    },
    {
      "epoch": 76.31,
      "grad_norm": 4.091534614562988,
      "learning_rate": 2.3693693693693693e-06,
      "loss": 0.3566,
      "step": 25410
    },
    {
      "epoch": 76.34,
      "grad_norm": 5.036163806915283,
      "learning_rate": 2.366366366366367e-06,
      "loss": 0.3528,
      "step": 25420
    },
    {
      "epoch": 76.37,
      "grad_norm": 4.043753147125244,
      "learning_rate": 2.3633633633633635e-06,
      "loss": 0.3806,
      "step": 25430
    },
    {
      "epoch": 76.4,
      "grad_norm": 5.221184253692627,
      "learning_rate": 2.3603603603603605e-06,
      "loss": 0.3676,
      "step": 25440
    },
    {
      "epoch": 76.43,
      "grad_norm": 5.204960346221924,
      "learning_rate": 2.3573573573573576e-06,
      "loss": 0.3483,
      "step": 25450
    },
    {
      "epoch": 76.46,
      "grad_norm": 4.301537990570068,
      "learning_rate": 2.3543543543543546e-06,
      "loss": 0.3223,
      "step": 25460
    },
    {
      "epoch": 76.49,
      "grad_norm": 3.0659539699554443,
      "learning_rate": 2.3513513513513517e-06,
      "loss": 0.4291,
      "step": 25470
    },
    {
      "epoch": 76.52,
      "grad_norm": 4.311132907867432,
      "learning_rate": 2.3483483483483483e-06,
      "loss": 0.4123,
      "step": 25480
    },
    {
      "epoch": 76.55,
      "grad_norm": 3.837833881378174,
      "learning_rate": 2.3453453453453458e-06,
      "loss": 0.3527,
      "step": 25490
    },
    {
      "epoch": 76.58,
      "grad_norm": 3.254117250442505,
      "learning_rate": 2.3423423423423424e-06,
      "loss": 0.3318,
      "step": 25500
    },
    {
      "epoch": 76.61,
      "grad_norm": 3.416591167449951,
      "learning_rate": 2.3393393393393395e-06,
      "loss": 0.3041,
      "step": 25510
    },
    {
      "epoch": 76.64,
      "grad_norm": 3.9906957149505615,
      "learning_rate": 2.3363363363363365e-06,
      "loss": 0.3046,
      "step": 25520
    },
    {
      "epoch": 76.67,
      "grad_norm": 4.718999862670898,
      "learning_rate": 2.3333333333333336e-06,
      "loss": 0.3363,
      "step": 25530
    },
    {
      "epoch": 76.7,
      "grad_norm": 5.844274520874023,
      "learning_rate": 2.3303303303303302e-06,
      "loss": 0.3712,
      "step": 25540
    },
    {
      "epoch": 76.73,
      "grad_norm": 3.6091501712799072,
      "learning_rate": 2.3273273273273277e-06,
      "loss": 0.3765,
      "step": 25550
    },
    {
      "epoch": 76.76,
      "grad_norm": 3.9883275032043457,
      "learning_rate": 2.3243243243243247e-06,
      "loss": 0.3582,
      "step": 25560
    },
    {
      "epoch": 76.79,
      "grad_norm": 4.031697750091553,
      "learning_rate": 2.3213213213213214e-06,
      "loss": 0.3466,
      "step": 25570
    },
    {
      "epoch": 76.82,
      "grad_norm": 3.932941436767578,
      "learning_rate": 2.3183183183183184e-06,
      "loss": 0.3339,
      "step": 25580
    },
    {
      "epoch": 76.85,
      "grad_norm": 5.171661376953125,
      "learning_rate": 2.3153153153153155e-06,
      "loss": 0.3986,
      "step": 25590
    },
    {
      "epoch": 76.88,
      "grad_norm": 4.379799842834473,
      "learning_rate": 2.3123123123123125e-06,
      "loss": 0.3195,
      "step": 25600
    },
    {
      "epoch": 76.91,
      "grad_norm": 4.38922643661499,
      "learning_rate": 2.3093093093093096e-06,
      "loss": 0.3627,
      "step": 25610
    },
    {
      "epoch": 76.94,
      "grad_norm": 4.423058032989502,
      "learning_rate": 2.3063063063063067e-06,
      "loss": 0.405,
      "step": 25620
    },
    {
      "epoch": 76.97,
      "grad_norm": 4.799821853637695,
      "learning_rate": 2.3033033033033033e-06,
      "loss": 0.3353,
      "step": 25630
    },
    {
      "epoch": 77.0,
      "grad_norm": 4.43877649307251,
      "learning_rate": 2.3003003003003003e-06,
      "loss": 0.3419,
      "step": 25640
    },
    {
      "epoch": 77.0,
      "eval_accuracy": 0.9201,
      "eval_loss": 0.311969131231308,
      "eval_runtime": 26.3374,
      "eval_samples_per_second": 379.688,
      "eval_steps_per_second": 1.519,
      "step": 25641
    },
    {
      "epoch": 77.03,
      "grad_norm": 3.7246921062469482,
      "learning_rate": 2.297297297297298e-06,
      "loss": 0.317,
      "step": 25650
    },
    {
      "epoch": 77.06,
      "grad_norm": 6.915863990783691,
      "learning_rate": 2.2942942942942944e-06,
      "loss": 0.4262,
      "step": 25660
    },
    {
      "epoch": 77.09,
      "grad_norm": 4.640294551849365,
      "learning_rate": 2.2912912912912915e-06,
      "loss": 0.3666,
      "step": 25670
    },
    {
      "epoch": 77.12,
      "grad_norm": 3.5786235332489014,
      "learning_rate": 2.2882882882882886e-06,
      "loss": 0.4063,
      "step": 25680
    },
    {
      "epoch": 77.15,
      "grad_norm": 3.300715446472168,
      "learning_rate": 2.2852852852852856e-06,
      "loss": 0.3614,
      "step": 25690
    },
    {
      "epoch": 77.18,
      "grad_norm": 4.162263870239258,
      "learning_rate": 2.2822822822822822e-06,
      "loss": 0.3722,
      "step": 25700
    },
    {
      "epoch": 77.21,
      "grad_norm": 4.885679721832275,
      "learning_rate": 2.2792792792792793e-06,
      "loss": 0.3381,
      "step": 25710
    },
    {
      "epoch": 77.24,
      "grad_norm": 3.1327507495880127,
      "learning_rate": 2.2762762762762764e-06,
      "loss": 0.3698,
      "step": 25720
    },
    {
      "epoch": 77.27,
      "grad_norm": 3.9108171463012695,
      "learning_rate": 2.2732732732732734e-06,
      "loss": 0.3394,
      "step": 25730
    },
    {
      "epoch": 77.3,
      "grad_norm": 5.891586780548096,
      "learning_rate": 2.2702702702702705e-06,
      "loss": 0.3678,
      "step": 25740
    },
    {
      "epoch": 77.33,
      "grad_norm": 5.72971248626709,
      "learning_rate": 2.2672672672672675e-06,
      "loss": 0.3196,
      "step": 25750
    },
    {
      "epoch": 77.36,
      "grad_norm": 4.004449844360352,
      "learning_rate": 2.2642642642642646e-06,
      "loss": 0.3157,
      "step": 25760
    },
    {
      "epoch": 77.39,
      "grad_norm": 5.830347537994385,
      "learning_rate": 2.261261261261261e-06,
      "loss": 0.3499,
      "step": 25770
    },
    {
      "epoch": 77.42,
      "grad_norm": 3.3828866481781006,
      "learning_rate": 2.2582582582582587e-06,
      "loss": 0.3603,
      "step": 25780
    },
    {
      "epoch": 77.45,
      "grad_norm": 4.691468238830566,
      "learning_rate": 2.2552552552552553e-06,
      "loss": 0.3534,
      "step": 25790
    },
    {
      "epoch": 77.48,
      "grad_norm": 6.159532070159912,
      "learning_rate": 2.2522522522522524e-06,
      "loss": 0.3865,
      "step": 25800
    },
    {
      "epoch": 77.51,
      "grad_norm": 4.088760852813721,
      "learning_rate": 2.2492492492492494e-06,
      "loss": 0.2931,
      "step": 25810
    },
    {
      "epoch": 77.54,
      "grad_norm": 3.893993616104126,
      "learning_rate": 2.2462462462462465e-06,
      "loss": 0.2858,
      "step": 25820
    },
    {
      "epoch": 77.57,
      "grad_norm": 5.2851386070251465,
      "learning_rate": 2.2432432432432435e-06,
      "loss": 0.3693,
      "step": 25830
    },
    {
      "epoch": 77.6,
      "grad_norm": 3.5877768993377686,
      "learning_rate": 2.2402402402402406e-06,
      "loss": 0.3675,
      "step": 25840
    },
    {
      "epoch": 77.63,
      "grad_norm": 6.06298828125,
      "learning_rate": 2.2372372372372376e-06,
      "loss": 0.378,
      "step": 25850
    },
    {
      "epoch": 77.66,
      "grad_norm": 5.485525608062744,
      "learning_rate": 2.2342342342342343e-06,
      "loss": 0.3653,
      "step": 25860
    },
    {
      "epoch": 77.69,
      "grad_norm": 3.886467218399048,
      "learning_rate": 2.2312312312312313e-06,
      "loss": 0.416,
      "step": 25870
    },
    {
      "epoch": 77.72,
      "grad_norm": 3.7569162845611572,
      "learning_rate": 2.2282282282282284e-06,
      "loss": 0.3325,
      "step": 25880
    },
    {
      "epoch": 77.75,
      "grad_norm": 3.709028720855713,
      "learning_rate": 2.2252252252252254e-06,
      "loss": 0.3538,
      "step": 25890
    },
    {
      "epoch": 77.78,
      "grad_norm": 2.983882427215576,
      "learning_rate": 2.222222222222222e-06,
      "loss": 0.32,
      "step": 25900
    },
    {
      "epoch": 77.81,
      "grad_norm": 4.826866626739502,
      "learning_rate": 2.2192192192192196e-06,
      "loss": 0.3965,
      "step": 25910
    },
    {
      "epoch": 77.84,
      "grad_norm": 2.939681053161621,
      "learning_rate": 2.2162162162162166e-06,
      "loss": 0.3285,
      "step": 25920
    },
    {
      "epoch": 77.87,
      "grad_norm": 5.774621486663818,
      "learning_rate": 2.2132132132132132e-06,
      "loss": 0.3854,
      "step": 25930
    },
    {
      "epoch": 77.9,
      "grad_norm": 3.492159128189087,
      "learning_rate": 2.2102102102102103e-06,
      "loss": 0.335,
      "step": 25940
    },
    {
      "epoch": 77.93,
      "grad_norm": 5.0627217292785645,
      "learning_rate": 2.2072072072072073e-06,
      "loss": 0.3663,
      "step": 25950
    },
    {
      "epoch": 77.96,
      "grad_norm": 5.784512042999268,
      "learning_rate": 2.2042042042042044e-06,
      "loss": 0.3586,
      "step": 25960
    },
    {
      "epoch": 77.99,
      "grad_norm": 6.522767543792725,
      "learning_rate": 2.2012012012012015e-06,
      "loss": 0.3918,
      "step": 25970
    },
    {
      "epoch": 78.0,
      "eval_accuracy": 0.9201,
      "eval_loss": 0.30773985385894775,
      "eval_runtime": 26.273,
      "eval_samples_per_second": 380.618,
      "eval_steps_per_second": 1.522,
      "step": 25974
    },
    {
      "epoch": 78.02,
      "grad_norm": 4.673668384552002,
      "learning_rate": 2.1981981981981985e-06,
      "loss": 0.3657,
      "step": 25980
    },
    {
      "epoch": 78.05,
      "grad_norm": 4.2784342765808105,
      "learning_rate": 2.195195195195195e-06,
      "loss": 0.3926,
      "step": 25990
    },
    {
      "epoch": 78.08,
      "grad_norm": 3.7317323684692383,
      "learning_rate": 2.192192192192192e-06,
      "loss": 0.3491,
      "step": 26000
    },
    {
      "epoch": 78.11,
      "grad_norm": 3.3739635944366455,
      "learning_rate": 2.1891891891891897e-06,
      "loss": 0.3232,
      "step": 26010
    },
    {
      "epoch": 78.14,
      "grad_norm": 3.77476167678833,
      "learning_rate": 2.1861861861861863e-06,
      "loss": 0.3289,
      "step": 26020
    },
    {
      "epoch": 78.17,
      "grad_norm": 5.433454990386963,
      "learning_rate": 2.1831831831831834e-06,
      "loss": 0.3941,
      "step": 26030
    },
    {
      "epoch": 78.2,
      "grad_norm": 3.8521242141723633,
      "learning_rate": 2.1801801801801804e-06,
      "loss": 0.3692,
      "step": 26040
    },
    {
      "epoch": 78.23,
      "grad_norm": 3.314911127090454,
      "learning_rate": 2.1771771771771775e-06,
      "loss": 0.3223,
      "step": 26050
    },
    {
      "epoch": 78.26,
      "grad_norm": 3.950944662094116,
      "learning_rate": 2.174174174174174e-06,
      "loss": 0.3788,
      "step": 26060
    },
    {
      "epoch": 78.29,
      "grad_norm": 2.781872034072876,
      "learning_rate": 2.1711711711711716e-06,
      "loss": 0.3507,
      "step": 26070
    },
    {
      "epoch": 78.32,
      "grad_norm": 6.140153884887695,
      "learning_rate": 2.1681681681681682e-06,
      "loss": 0.3586,
      "step": 26080
    },
    {
      "epoch": 78.35,
      "grad_norm": 5.543150424957275,
      "learning_rate": 2.1651651651651653e-06,
      "loss": 0.3888,
      "step": 26090
    },
    {
      "epoch": 78.38,
      "grad_norm": 4.54295539855957,
      "learning_rate": 2.1621621621621623e-06,
      "loss": 0.3669,
      "step": 26100
    },
    {
      "epoch": 78.41,
      "grad_norm": 4.7346882820129395,
      "learning_rate": 2.1591591591591594e-06,
      "loss": 0.3872,
      "step": 26110
    },
    {
      "epoch": 78.44,
      "grad_norm": 4.123915672302246,
      "learning_rate": 2.1561561561561564e-06,
      "loss": 0.3481,
      "step": 26120
    },
    {
      "epoch": 78.47,
      "grad_norm": 3.8088018894195557,
      "learning_rate": 2.153153153153153e-06,
      "loss": 0.344,
      "step": 26130
    },
    {
      "epoch": 78.5,
      "grad_norm": 5.422818660736084,
      "learning_rate": 2.1501501501501505e-06,
      "loss": 0.3828,
      "step": 26140
    },
    {
      "epoch": 78.53,
      "grad_norm": 4.193113803863525,
      "learning_rate": 2.147147147147147e-06,
      "loss": 0.3927,
      "step": 26150
    },
    {
      "epoch": 78.56,
      "grad_norm": 4.59672212600708,
      "learning_rate": 2.1441441441441442e-06,
      "loss": 0.333,
      "step": 26160
    },
    {
      "epoch": 78.59,
      "grad_norm": 6.043910503387451,
      "learning_rate": 2.1411411411411413e-06,
      "loss": 0.3232,
      "step": 26170
    },
    {
      "epoch": 78.62,
      "grad_norm": 4.8449859619140625,
      "learning_rate": 2.1381381381381383e-06,
      "loss": 0.3663,
      "step": 26180
    },
    {
      "epoch": 78.65,
      "grad_norm": 4.4025959968566895,
      "learning_rate": 2.1351351351351354e-06,
      "loss": 0.3265,
      "step": 26190
    },
    {
      "epoch": 78.68,
      "grad_norm": 4.092075347900391,
      "learning_rate": 2.1321321321321325e-06,
      "loss": 0.3299,
      "step": 26200
    },
    {
      "epoch": 78.71,
      "grad_norm": 3.840696334838867,
      "learning_rate": 2.1291291291291295e-06,
      "loss": 0.3206,
      "step": 26210
    },
    {
      "epoch": 78.74,
      "grad_norm": 7.876099109649658,
      "learning_rate": 2.126126126126126e-06,
      "loss": 0.3845,
      "step": 26220
    },
    {
      "epoch": 78.77,
      "grad_norm": 4.138664245605469,
      "learning_rate": 2.123123123123123e-06,
      "loss": 0.3896,
      "step": 26230
    },
    {
      "epoch": 78.8,
      "grad_norm": 4.147843360900879,
      "learning_rate": 2.1201201201201203e-06,
      "loss": 0.3792,
      "step": 26240
    },
    {
      "epoch": 78.83,
      "grad_norm": 3.6445870399475098,
      "learning_rate": 2.1171171171171173e-06,
      "loss": 0.3529,
      "step": 26250
    },
    {
      "epoch": 78.86,
      "grad_norm": 4.092494487762451,
      "learning_rate": 2.1141141141141144e-06,
      "loss": 0.396,
      "step": 26260
    },
    {
      "epoch": 78.89,
      "grad_norm": 4.143327236175537,
      "learning_rate": 2.1111111111111114e-06,
      "loss": 0.3662,
      "step": 26270
    },
    {
      "epoch": 78.92,
      "grad_norm": 3.4769155979156494,
      "learning_rate": 2.1081081081081085e-06,
      "loss": 0.3758,
      "step": 26280
    },
    {
      "epoch": 78.95,
      "grad_norm": 7.702763080596924,
      "learning_rate": 2.105105105105105e-06,
      "loss": 0.3364,
      "step": 26290
    },
    {
      "epoch": 78.98,
      "grad_norm": 4.711182117462158,
      "learning_rate": 2.102102102102102e-06,
      "loss": 0.3571,
      "step": 26300
    },
    {
      "epoch": 79.0,
      "eval_accuracy": 0.9195,
      "eval_loss": 0.3119402229785919,
      "eval_runtime": 26.415,
      "eval_samples_per_second": 378.573,
      "eval_steps_per_second": 1.514,
      "step": 26307
    },
    {
      "epoch": 79.01,
      "grad_norm": 4.214149475097656,
      "learning_rate": 2.0990990990990992e-06,
      "loss": 0.2827,
      "step": 26310
    },
    {
      "epoch": 79.04,
      "grad_norm": 10.097718238830566,
      "learning_rate": 2.0960960960960963e-06,
      "loss": 0.3513,
      "step": 26320
    },
    {
      "epoch": 79.07,
      "grad_norm": 4.275505065917969,
      "learning_rate": 2.0930930930930933e-06,
      "loss": 0.3628,
      "step": 26330
    },
    {
      "epoch": 79.1,
      "grad_norm": 5.002494812011719,
      "learning_rate": 2.0900900900900904e-06,
      "loss": 0.3495,
      "step": 26340
    },
    {
      "epoch": 79.13,
      "grad_norm": 4.64259147644043,
      "learning_rate": 2.087087087087087e-06,
      "loss": 0.3227,
      "step": 26350
    },
    {
      "epoch": 79.16,
      "grad_norm": 3.504368305206299,
      "learning_rate": 2.084084084084084e-06,
      "loss": 0.4008,
      "step": 26360
    },
    {
      "epoch": 79.19,
      "grad_norm": 5.1687822341918945,
      "learning_rate": 2.0810810810810815e-06,
      "loss": 0.3032,
      "step": 26370
    },
    {
      "epoch": 79.22,
      "grad_norm": 3.8527510166168213,
      "learning_rate": 2.078078078078078e-06,
      "loss": 0.315,
      "step": 26380
    },
    {
      "epoch": 79.25,
      "grad_norm": 4.061619758605957,
      "learning_rate": 2.0750750750750752e-06,
      "loss": 0.3413,
      "step": 26390
    },
    {
      "epoch": 79.28,
      "grad_norm": 8.303925514221191,
      "learning_rate": 2.0720720720720723e-06,
      "loss": 0.3979,
      "step": 26400
    },
    {
      "epoch": 79.31,
      "grad_norm": 6.806997776031494,
      "learning_rate": 2.0690690690690693e-06,
      "loss": 0.3352,
      "step": 26410
    },
    {
      "epoch": 79.34,
      "grad_norm": 9.438483238220215,
      "learning_rate": 2.066066066066066e-06,
      "loss": 0.3711,
      "step": 26420
    },
    {
      "epoch": 79.37,
      "grad_norm": 4.446671009063721,
      "learning_rate": 2.0630630630630634e-06,
      "loss": 0.3644,
      "step": 26430
    },
    {
      "epoch": 79.4,
      "grad_norm": 4.411134719848633,
      "learning_rate": 2.06006006006006e-06,
      "loss": 0.3675,
      "step": 26440
    },
    {
      "epoch": 79.43,
      "grad_norm": 5.646937370300293,
      "learning_rate": 2.057057057057057e-06,
      "loss": 0.3417,
      "step": 26450
    },
    {
      "epoch": 79.46,
      "grad_norm": 4.043561935424805,
      "learning_rate": 2.054054054054054e-06,
      "loss": 0.3337,
      "step": 26460
    },
    {
      "epoch": 79.49,
      "grad_norm": 4.135385036468506,
      "learning_rate": 2.0510510510510512e-06,
      "loss": 0.3739,
      "step": 26470
    },
    {
      "epoch": 79.52,
      "grad_norm": 4.485921859741211,
      "learning_rate": 2.0480480480480483e-06,
      "loss": 0.3413,
      "step": 26480
    },
    {
      "epoch": 79.55,
      "grad_norm": 3.9926671981811523,
      "learning_rate": 2.045045045045045e-06,
      "loss": 0.3469,
      "step": 26490
    },
    {
      "epoch": 79.58,
      "grad_norm": 3.3427035808563232,
      "learning_rate": 2.0420420420420424e-06,
      "loss": 0.3467,
      "step": 26500
    },
    {
      "epoch": 79.61,
      "grad_norm": 4.4577765464782715,
      "learning_rate": 2.039039039039039e-06,
      "loss": 0.3425,
      "step": 26510
    },
    {
      "epoch": 79.64,
      "grad_norm": 4.792611598968506,
      "learning_rate": 2.036036036036036e-06,
      "loss": 0.3788,
      "step": 26520
    },
    {
      "epoch": 79.67,
      "grad_norm": 4.125549793243408,
      "learning_rate": 2.033033033033033e-06,
      "loss": 0.3331,
      "step": 26530
    },
    {
      "epoch": 79.7,
      "grad_norm": 5.205685615539551,
      "learning_rate": 2.03003003003003e-06,
      "loss": 0.3776,
      "step": 26540
    },
    {
      "epoch": 79.73,
      "grad_norm": 4.373479843139648,
      "learning_rate": 2.0270270270270273e-06,
      "loss": 0.3244,
      "step": 26550
    },
    {
      "epoch": 79.76,
      "grad_norm": 4.503603458404541,
      "learning_rate": 2.0240240240240243e-06,
      "loss": 0.3381,
      "step": 26560
    },
    {
      "epoch": 79.79,
      "grad_norm": 3.910428524017334,
      "learning_rate": 2.0210210210210214e-06,
      "loss": 0.3197,
      "step": 26570
    },
    {
      "epoch": 79.82,
      "grad_norm": 8.611750602722168,
      "learning_rate": 2.018018018018018e-06,
      "loss": 0.4043,
      "step": 26580
    },
    {
      "epoch": 79.85,
      "grad_norm": 4.655651569366455,
      "learning_rate": 2.015015015015015e-06,
      "loss": 0.362,
      "step": 26590
    },
    {
      "epoch": 79.88,
      "grad_norm": 3.3452017307281494,
      "learning_rate": 2.012012012012012e-06,
      "loss": 0.3695,
      "step": 26600
    },
    {
      "epoch": 79.91,
      "grad_norm": 4.068109512329102,
      "learning_rate": 2.009009009009009e-06,
      "loss": 0.3378,
      "step": 26610
    },
    {
      "epoch": 79.94,
      "grad_norm": 3.963881731033325,
      "learning_rate": 2.0060060060060062e-06,
      "loss": 0.3133,
      "step": 26620
    },
    {
      "epoch": 79.97,
      "grad_norm": 5.272686958312988,
      "learning_rate": 2.0030030030030033e-06,
      "loss": 0.3489,
      "step": 26630
    },
    {
      "epoch": 80.0,
      "grad_norm": 4.051087379455566,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 0.3609,
      "step": 26640
    },
    {
      "epoch": 80.0,
      "eval_accuracy": 0.9195,
      "eval_loss": 0.3120231032371521,
      "eval_runtime": 26.2413,
      "eval_samples_per_second": 381.078,
      "eval_steps_per_second": 1.524,
      "step": 26640
    },
    {
      "epoch": 80.03,
      "grad_norm": 3.8018898963928223,
      "learning_rate": 1.996996996996997e-06,
      "loss": 0.3505,
      "step": 26650
    },
    {
      "epoch": 80.06,
      "grad_norm": 3.8653721809387207,
      "learning_rate": 1.9939939939939944e-06,
      "loss": 0.3421,
      "step": 26660
    },
    {
      "epoch": 80.09,
      "grad_norm": 3.6045594215393066,
      "learning_rate": 1.990990990990991e-06,
      "loss": 0.3618,
      "step": 26670
    },
    {
      "epoch": 80.12,
      "grad_norm": 5.400451183319092,
      "learning_rate": 1.987987987987988e-06,
      "loss": 0.3452,
      "step": 26680
    },
    {
      "epoch": 80.15,
      "grad_norm": 4.897947788238525,
      "learning_rate": 1.984984984984985e-06,
      "loss": 0.3522,
      "step": 26690
    },
    {
      "epoch": 80.18,
      "grad_norm": 4.104059219360352,
      "learning_rate": 1.9819819819819822e-06,
      "loss": 0.3745,
      "step": 26700
    },
    {
      "epoch": 80.21,
      "grad_norm": 4.459397792816162,
      "learning_rate": 1.978978978978979e-06,
      "loss": 0.3934,
      "step": 26710
    },
    {
      "epoch": 80.24,
      "grad_norm": 4.023631572723389,
      "learning_rate": 1.975975975975976e-06,
      "loss": 0.33,
      "step": 26720
    },
    {
      "epoch": 80.27,
      "grad_norm": 5.148887634277344,
      "learning_rate": 1.9729729729729734e-06,
      "loss": 0.3799,
      "step": 26730
    },
    {
      "epoch": 80.3,
      "grad_norm": 4.939486980438232,
      "learning_rate": 1.96996996996997e-06,
      "loss": 0.3552,
      "step": 26740
    },
    {
      "epoch": 80.33,
      "grad_norm": 4.360585689544678,
      "learning_rate": 1.966966966966967e-06,
      "loss": 0.3355,
      "step": 26750
    },
    {
      "epoch": 80.36,
      "grad_norm": 3.46449613571167,
      "learning_rate": 1.963963963963964e-06,
      "loss": 0.3763,
      "step": 26760
    },
    {
      "epoch": 80.39,
      "grad_norm": 6.351463794708252,
      "learning_rate": 1.960960960960961e-06,
      "loss": 0.327,
      "step": 26770
    },
    {
      "epoch": 80.42,
      "grad_norm": 4.960982799530029,
      "learning_rate": 1.957957957957958e-06,
      "loss": 0.3522,
      "step": 26780
    },
    {
      "epoch": 80.45,
      "grad_norm": 4.413058280944824,
      "learning_rate": 1.9549549549549553e-06,
      "loss": 0.365,
      "step": 26790
    },
    {
      "epoch": 80.48,
      "grad_norm": 3.134146213531494,
      "learning_rate": 1.951951951951952e-06,
      "loss": 0.3284,
      "step": 26800
    },
    {
      "epoch": 80.51,
      "grad_norm": 4.047032356262207,
      "learning_rate": 1.948948948948949e-06,
      "loss": 0.341,
      "step": 26810
    },
    {
      "epoch": 80.54,
      "grad_norm": 4.29499626159668,
      "learning_rate": 1.945945945945946e-06,
      "loss": 0.3837,
      "step": 26820
    },
    {
      "epoch": 80.57,
      "grad_norm": 4.46774435043335,
      "learning_rate": 1.942942942942943e-06,
      "loss": 0.3568,
      "step": 26830
    },
    {
      "epoch": 80.6,
      "grad_norm": 5.553523540496826,
      "learning_rate": 1.93993993993994e-06,
      "loss": 0.3558,
      "step": 26840
    },
    {
      "epoch": 80.63,
      "grad_norm": 4.414850234985352,
      "learning_rate": 1.9369369369369372e-06,
      "loss": 0.3321,
      "step": 26850
    },
    {
      "epoch": 80.66,
      "grad_norm": 16.202510833740234,
      "learning_rate": 1.9339339339339343e-06,
      "loss": 0.409,
      "step": 26860
    },
    {
      "epoch": 80.69,
      "grad_norm": 5.092227458953857,
      "learning_rate": 1.930930930930931e-06,
      "loss": 0.3808,
      "step": 26870
    },
    {
      "epoch": 80.72,
      "grad_norm": 3.3495876789093018,
      "learning_rate": 1.927927927927928e-06,
      "loss": 0.3865,
      "step": 26880
    },
    {
      "epoch": 80.75,
      "grad_norm": 2.8189918994903564,
      "learning_rate": 1.924924924924925e-06,
      "loss": 0.302,
      "step": 26890
    },
    {
      "epoch": 80.78,
      "grad_norm": 5.149996757507324,
      "learning_rate": 1.921921921921922e-06,
      "loss": 0.3664,
      "step": 26900
    },
    {
      "epoch": 80.81,
      "grad_norm": 3.803631544113159,
      "learning_rate": 1.918918918918919e-06,
      "loss": 0.3583,
      "step": 26910
    },
    {
      "epoch": 80.84,
      "grad_norm": 4.1913838386535645,
      "learning_rate": 1.915915915915916e-06,
      "loss": 0.3409,
      "step": 26920
    },
    {
      "epoch": 80.87,
      "grad_norm": 5.084752082824707,
      "learning_rate": 1.9129129129129132e-06,
      "loss": 0.3226,
      "step": 26930
    },
    {
      "epoch": 80.9,
      "grad_norm": 5.19244384765625,
      "learning_rate": 1.90990990990991e-06,
      "loss": 0.3839,
      "step": 26940
    },
    {
      "epoch": 80.93,
      "grad_norm": 3.713278293609619,
      "learning_rate": 1.9069069069069071e-06,
      "loss": 0.3436,
      "step": 26950
    },
    {
      "epoch": 80.96,
      "grad_norm": 3.6275076866149902,
      "learning_rate": 1.903903903903904e-06,
      "loss": 0.3007,
      "step": 26960
    },
    {
      "epoch": 80.99,
      "grad_norm": 3.242652654647827,
      "learning_rate": 1.900900900900901e-06,
      "loss": 0.3324,
      "step": 26970
    },
    {
      "epoch": 81.0,
      "eval_accuracy": 0.9194,
      "eval_loss": 0.3119645118713379,
      "eval_runtime": 26.2986,
      "eval_samples_per_second": 380.248,
      "eval_steps_per_second": 1.521,
      "step": 26973
    },
    {
      "epoch": 81.02,
      "grad_norm": 3.0082156658172607,
      "learning_rate": 1.897897897897898e-06,
      "loss": 0.3377,
      "step": 26980
    },
    {
      "epoch": 81.05,
      "grad_norm": 4.562652587890625,
      "learning_rate": 1.894894894894895e-06,
      "loss": 0.3139,
      "step": 26990
    },
    {
      "epoch": 81.08,
      "grad_norm": 5.133596897125244,
      "learning_rate": 1.8918918918918922e-06,
      "loss": 0.3508,
      "step": 27000
    },
    {
      "epoch": 81.11,
      "grad_norm": 3.8124804496765137,
      "learning_rate": 1.888888888888889e-06,
      "loss": 0.3368,
      "step": 27010
    },
    {
      "epoch": 81.14,
      "grad_norm": 5.156115531921387,
      "learning_rate": 1.885885885885886e-06,
      "loss": 0.3939,
      "step": 27020
    },
    {
      "epoch": 81.17,
      "grad_norm": 4.8171706199646,
      "learning_rate": 1.882882882882883e-06,
      "loss": 0.3786,
      "step": 27030
    },
    {
      "epoch": 81.2,
      "grad_norm": 3.5435335636138916,
      "learning_rate": 1.87987987987988e-06,
      "loss": 0.3313,
      "step": 27040
    },
    {
      "epoch": 81.23,
      "grad_norm": 6.529219627380371,
      "learning_rate": 1.8768768768768768e-06,
      "loss": 0.3772,
      "step": 27050
    },
    {
      "epoch": 81.26,
      "grad_norm": 4.352439880371094,
      "learning_rate": 1.873873873873874e-06,
      "loss": 0.3511,
      "step": 27060
    },
    {
      "epoch": 81.29,
      "grad_norm": 3.884099006652832,
      "learning_rate": 1.8708708708708712e-06,
      "loss": 0.3412,
      "step": 27070
    },
    {
      "epoch": 81.32,
      "grad_norm": 2.405987501144409,
      "learning_rate": 1.867867867867868e-06,
      "loss": 0.348,
      "step": 27080
    },
    {
      "epoch": 81.35,
      "grad_norm": 4.360373497009277,
      "learning_rate": 1.864864864864865e-06,
      "loss": 0.3276,
      "step": 27090
    },
    {
      "epoch": 81.38,
      "grad_norm": 5.636603355407715,
      "learning_rate": 1.861861861861862e-06,
      "loss": 0.3073,
      "step": 27100
    },
    {
      "epoch": 81.41,
      "grad_norm": 4.376233100891113,
      "learning_rate": 1.8588588588588592e-06,
      "loss": 0.326,
      "step": 27110
    },
    {
      "epoch": 81.44,
      "grad_norm": 3.830524206161499,
      "learning_rate": 1.855855855855856e-06,
      "loss": 0.3294,
      "step": 27120
    },
    {
      "epoch": 81.47,
      "grad_norm": 3.781371831893921,
      "learning_rate": 1.852852852852853e-06,
      "loss": 0.3598,
      "step": 27130
    },
    {
      "epoch": 81.5,
      "grad_norm": 4.620359420776367,
      "learning_rate": 1.84984984984985e-06,
      "loss": 0.3536,
      "step": 27140
    },
    {
      "epoch": 81.53,
      "grad_norm": 5.590673923492432,
      "learning_rate": 1.846846846846847e-06,
      "loss": 0.3173,
      "step": 27150
    },
    {
      "epoch": 81.56,
      "grad_norm": 6.378412246704102,
      "learning_rate": 1.8438438438438442e-06,
      "loss": 0.3568,
      "step": 27160
    },
    {
      "epoch": 81.59,
      "grad_norm": 6.295036315917969,
      "learning_rate": 1.840840840840841e-06,
      "loss": 0.4,
      "step": 27170
    },
    {
      "epoch": 81.62,
      "grad_norm": 5.026501178741455,
      "learning_rate": 1.8378378378378381e-06,
      "loss": 0.3473,
      "step": 27180
    },
    {
      "epoch": 81.65,
      "grad_norm": 4.446352958679199,
      "learning_rate": 1.834834834834835e-06,
      "loss": 0.329,
      "step": 27190
    },
    {
      "epoch": 81.68,
      "grad_norm": 3.3507978916168213,
      "learning_rate": 1.831831831831832e-06,
      "loss": 0.338,
      "step": 27200
    },
    {
      "epoch": 81.71,
      "grad_norm": 4.917937755584717,
      "learning_rate": 1.8288288288288289e-06,
      "loss": 0.3566,
      "step": 27210
    },
    {
      "epoch": 81.74,
      "grad_norm": 3.061997652053833,
      "learning_rate": 1.825825825825826e-06,
      "loss": 0.3419,
      "step": 27220
    },
    {
      "epoch": 81.77,
      "grad_norm": 5.524289608001709,
      "learning_rate": 1.8228228228228228e-06,
      "loss": 0.4081,
      "step": 27230
    },
    {
      "epoch": 81.8,
      "grad_norm": 4.370038986206055,
      "learning_rate": 1.81981981981982e-06,
      "loss": 0.3643,
      "step": 27240
    },
    {
      "epoch": 81.83,
      "grad_norm": 4.464176654815674,
      "learning_rate": 1.816816816816817e-06,
      "loss": 0.3596,
      "step": 27250
    },
    {
      "epoch": 81.86,
      "grad_norm": 4.067475318908691,
      "learning_rate": 1.813813813813814e-06,
      "loss": 0.3262,
      "step": 27260
    },
    {
      "epoch": 81.89,
      "grad_norm": 6.306347370147705,
      "learning_rate": 1.810810810810811e-06,
      "loss": 0.4227,
      "step": 27270
    },
    {
      "epoch": 81.92,
      "grad_norm": 6.332818984985352,
      "learning_rate": 1.8078078078078078e-06,
      "loss": 0.3674,
      "step": 27280
    },
    {
      "epoch": 81.95,
      "grad_norm": 6.539882659912109,
      "learning_rate": 1.804804804804805e-06,
      "loss": 0.4095,
      "step": 27290
    },
    {
      "epoch": 81.98,
      "grad_norm": 4.516655445098877,
      "learning_rate": 1.801801801801802e-06,
      "loss": 0.3387,
      "step": 27300
    },
    {
      "epoch": 82.0,
      "eval_accuracy": 0.9199,
      "eval_loss": 0.3118031322956085,
      "eval_runtime": 26.094,
      "eval_samples_per_second": 383.23,
      "eval_steps_per_second": 1.533,
      "step": 27306
    },
    {
      "epoch": 82.01,
      "grad_norm": 4.243729114532471,
      "learning_rate": 1.798798798798799e-06,
      "loss": 0.3855,
      "step": 27310
    },
    {
      "epoch": 82.04,
      "grad_norm": 4.209694862365723,
      "learning_rate": 1.7957957957957958e-06,
      "loss": 0.3169,
      "step": 27320
    },
    {
      "epoch": 82.07,
      "grad_norm": 4.351260662078857,
      "learning_rate": 1.7927927927927929e-06,
      "loss": 0.3487,
      "step": 27330
    },
    {
      "epoch": 82.1,
      "grad_norm": 3.6804800033569336,
      "learning_rate": 1.7897897897897902e-06,
      "loss": 0.3887,
      "step": 27340
    },
    {
      "epoch": 82.13,
      "grad_norm": 4.920446395874023,
      "learning_rate": 1.786786786786787e-06,
      "loss": 0.3504,
      "step": 27350
    },
    {
      "epoch": 82.16,
      "grad_norm": 2.7672877311706543,
      "learning_rate": 1.783783783783784e-06,
      "loss": 0.3379,
      "step": 27360
    },
    {
      "epoch": 82.19,
      "grad_norm": 3.759749412536621,
      "learning_rate": 1.780780780780781e-06,
      "loss": 0.3534,
      "step": 27370
    },
    {
      "epoch": 82.22,
      "grad_norm": 5.417953014373779,
      "learning_rate": 1.777777777777778e-06,
      "loss": 0.362,
      "step": 27380
    },
    {
      "epoch": 82.25,
      "grad_norm": 4.77583122253418,
      "learning_rate": 1.7747747747747748e-06,
      "loss": 0.3211,
      "step": 27390
    },
    {
      "epoch": 82.28,
      "grad_norm": 4.040578842163086,
      "learning_rate": 1.7717717717717719e-06,
      "loss": 0.3729,
      "step": 27400
    },
    {
      "epoch": 82.31,
      "grad_norm": 6.474228382110596,
      "learning_rate": 1.7687687687687687e-06,
      "loss": 0.3436,
      "step": 27410
    },
    {
      "epoch": 82.34,
      "grad_norm": 5.873781681060791,
      "learning_rate": 1.765765765765766e-06,
      "loss": 0.3658,
      "step": 27420
    },
    {
      "epoch": 82.37,
      "grad_norm": 4.116581916809082,
      "learning_rate": 1.762762762762763e-06,
      "loss": 0.3546,
      "step": 27430
    },
    {
      "epoch": 82.4,
      "grad_norm": 4.436230659484863,
      "learning_rate": 1.7597597597597599e-06,
      "loss": 0.3849,
      "step": 27440
    },
    {
      "epoch": 82.43,
      "grad_norm": 5.607438564300537,
      "learning_rate": 1.756756756756757e-06,
      "loss": 0.364,
      "step": 27450
    },
    {
      "epoch": 82.46,
      "grad_norm": 5.597632884979248,
      "learning_rate": 1.7537537537537538e-06,
      "loss": 0.3084,
      "step": 27460
    },
    {
      "epoch": 82.49,
      "grad_norm": 3.4757421016693115,
      "learning_rate": 1.750750750750751e-06,
      "loss": 0.2759,
      "step": 27470
    },
    {
      "epoch": 82.52,
      "grad_norm": 5.107426166534424,
      "learning_rate": 1.7477477477477479e-06,
      "loss": 0.3418,
      "step": 27480
    },
    {
      "epoch": 82.55,
      "grad_norm": 4.416812419891357,
      "learning_rate": 1.744744744744745e-06,
      "loss": 0.3719,
      "step": 27490
    },
    {
      "epoch": 82.58,
      "grad_norm": 3.0949652194976807,
      "learning_rate": 1.7417417417417418e-06,
      "loss": 0.3791,
      "step": 27500
    },
    {
      "epoch": 82.61,
      "grad_norm": 3.1069371700286865,
      "learning_rate": 1.7387387387387388e-06,
      "loss": 0.3665,
      "step": 27510
    },
    {
      "epoch": 82.64,
      "grad_norm": 4.998491287231445,
      "learning_rate": 1.735735735735736e-06,
      "loss": 0.392,
      "step": 27520
    },
    {
      "epoch": 82.67,
      "grad_norm": 3.9115214347839355,
      "learning_rate": 1.732732732732733e-06,
      "loss": 0.3466,
      "step": 27530
    },
    {
      "epoch": 82.7,
      "grad_norm": 7.158092021942139,
      "learning_rate": 1.72972972972973e-06,
      "loss": 0.3945,
      "step": 27540
    },
    {
      "epoch": 82.73,
      "grad_norm": 3.646942138671875,
      "learning_rate": 1.7267267267267268e-06,
      "loss": 0.3675,
      "step": 27550
    },
    {
      "epoch": 82.76,
      "grad_norm": 2.7994065284729004,
      "learning_rate": 1.7237237237237239e-06,
      "loss": 0.3628,
      "step": 27560
    },
    {
      "epoch": 82.79,
      "grad_norm": 4.319077491760254,
      "learning_rate": 1.7207207207207207e-06,
      "loss": 0.4216,
      "step": 27570
    },
    {
      "epoch": 82.82,
      "grad_norm": 4.438078880310059,
      "learning_rate": 1.717717717717718e-06,
      "loss": 0.3594,
      "step": 27580
    },
    {
      "epoch": 82.85,
      "grad_norm": 5.1794610023498535,
      "learning_rate": 1.7147147147147146e-06,
      "loss": 0.3755,
      "step": 27590
    },
    {
      "epoch": 82.88,
      "grad_norm": 4.366214752197266,
      "learning_rate": 1.711711711711712e-06,
      "loss": 0.3263,
      "step": 27600
    },
    {
      "epoch": 82.91,
      "grad_norm": 6.042720794677734,
      "learning_rate": 1.708708708708709e-06,
      "loss": 0.3386,
      "step": 27610
    },
    {
      "epoch": 82.94,
      "grad_norm": 3.7862770557403564,
      "learning_rate": 1.7057057057057058e-06,
      "loss": 0.3235,
      "step": 27620
    },
    {
      "epoch": 82.97,
      "grad_norm": 6.545648574829102,
      "learning_rate": 1.7027027027027028e-06,
      "loss": 0.441,
      "step": 27630
    },
    {
      "epoch": 83.0,
      "eval_accuracy": 0.92,
      "eval_loss": 0.31168514490127563,
      "eval_runtime": 26.1131,
      "eval_samples_per_second": 382.95,
      "eval_steps_per_second": 1.532,
      "step": 27639
    },
    {
      "epoch": 83.0,
      "grad_norm": 4.174609661102295,
      "learning_rate": 1.6996996996996997e-06,
      "loss": 0.2663,
      "step": 27640
    },
    {
      "epoch": 83.03,
      "grad_norm": 4.8444061279296875,
      "learning_rate": 1.696696696696697e-06,
      "loss": 0.3656,
      "step": 27650
    },
    {
      "epoch": 83.06,
      "grad_norm": 3.6160123348236084,
      "learning_rate": 1.6936936936936938e-06,
      "loss": 0.2981,
      "step": 27660
    },
    {
      "epoch": 83.09,
      "grad_norm": 5.024894714355469,
      "learning_rate": 1.6906906906906909e-06,
      "loss": 0.3336,
      "step": 27670
    },
    {
      "epoch": 83.12,
      "grad_norm": 5.81716775894165,
      "learning_rate": 1.6876876876876877e-06,
      "loss": 0.3422,
      "step": 27680
    },
    {
      "epoch": 83.15,
      "grad_norm": 4.106879711151123,
      "learning_rate": 1.6846846846846848e-06,
      "loss": 0.3513,
      "step": 27690
    },
    {
      "epoch": 83.18,
      "grad_norm": 3.3921761512756348,
      "learning_rate": 1.681681681681682e-06,
      "loss": 0.3405,
      "step": 27700
    },
    {
      "epoch": 83.21,
      "grad_norm": 4.262903213500977,
      "learning_rate": 1.6786786786786789e-06,
      "loss": 0.3847,
      "step": 27710
    },
    {
      "epoch": 83.24,
      "grad_norm": 3.755544662475586,
      "learning_rate": 1.675675675675676e-06,
      "loss": 0.3272,
      "step": 27720
    },
    {
      "epoch": 83.27,
      "grad_norm": 5.057288646697998,
      "learning_rate": 1.6726726726726728e-06,
      "loss": 0.3626,
      "step": 27730
    },
    {
      "epoch": 83.3,
      "grad_norm": 4.16048526763916,
      "learning_rate": 1.6696696696696698e-06,
      "loss": 0.3726,
      "step": 27740
    },
    {
      "epoch": 83.33,
      "grad_norm": 3.6107938289642334,
      "learning_rate": 1.6666666666666667e-06,
      "loss": 0.3743,
      "step": 27750
    },
    {
      "epoch": 83.36,
      "grad_norm": 2.6226437091827393,
      "learning_rate": 1.663663663663664e-06,
      "loss": 0.3362,
      "step": 27760
    },
    {
      "epoch": 83.39,
      "grad_norm": 3.6641383171081543,
      "learning_rate": 1.6606606606606608e-06,
      "loss": 0.3794,
      "step": 27770
    },
    {
      "epoch": 83.42,
      "grad_norm": 3.930861234664917,
      "learning_rate": 1.6576576576576578e-06,
      "loss": 0.3394,
      "step": 27780
    },
    {
      "epoch": 83.45,
      "grad_norm": 8.251886367797852,
      "learning_rate": 1.6546546546546549e-06,
      "loss": 0.3275,
      "step": 27790
    },
    {
      "epoch": 83.48,
      "grad_norm": 5.807711124420166,
      "learning_rate": 1.6516516516516517e-06,
      "loss": 0.3511,
      "step": 27800
    },
    {
      "epoch": 83.51,
      "grad_norm": 3.4882266521453857,
      "learning_rate": 1.6486486486486488e-06,
      "loss": 0.3338,
      "step": 27810
    },
    {
      "epoch": 83.54,
      "grad_norm": 8.302092552185059,
      "learning_rate": 1.6456456456456456e-06,
      "loss": 0.344,
      "step": 27820
    },
    {
      "epoch": 83.57,
      "grad_norm": 4.071527481079102,
      "learning_rate": 1.6426426426426429e-06,
      "loss": 0.3865,
      "step": 27830
    },
    {
      "epoch": 83.6,
      "grad_norm": 3.8215830326080322,
      "learning_rate": 1.6396396396396397e-06,
      "loss": 0.3371,
      "step": 27840
    },
    {
      "epoch": 83.63,
      "grad_norm": 3.685642957687378,
      "learning_rate": 1.6366366366366368e-06,
      "loss": 0.3121,
      "step": 27850
    },
    {
      "epoch": 83.66,
      "grad_norm": 3.257276773452759,
      "learning_rate": 1.6336336336336336e-06,
      "loss": 0.3758,
      "step": 27860
    },
    {
      "epoch": 83.69,
      "grad_norm": 11.562396049499512,
      "learning_rate": 1.6306306306306307e-06,
      "loss": 0.3026,
      "step": 27870
    },
    {
      "epoch": 83.72,
      "grad_norm": 6.109564781188965,
      "learning_rate": 1.627627627627628e-06,
      "loss": 0.3394,
      "step": 27880
    },
    {
      "epoch": 83.75,
      "grad_norm": 3.819429874420166,
      "learning_rate": 1.6246246246246248e-06,
      "loss": 0.3727,
      "step": 27890
    },
    {
      "epoch": 83.78,
      "grad_norm": 3.833479404449463,
      "learning_rate": 1.6216216216216219e-06,
      "loss": 0.3356,
      "step": 27900
    },
    {
      "epoch": 83.81,
      "grad_norm": 3.8872108459472656,
      "learning_rate": 1.6186186186186187e-06,
      "loss": 0.3743,
      "step": 27910
    },
    {
      "epoch": 83.84,
      "grad_norm": 4.757948398590088,
      "learning_rate": 1.6156156156156157e-06,
      "loss": 0.3791,
      "step": 27920
    },
    {
      "epoch": 83.87,
      "grad_norm": 4.845849514007568,
      "learning_rate": 1.6126126126126126e-06,
      "loss": 0.318,
      "step": 27930
    },
    {
      "epoch": 83.9,
      "grad_norm": 3.234450578689575,
      "learning_rate": 1.6096096096096099e-06,
      "loss": 0.3461,
      "step": 27940
    },
    {
      "epoch": 83.93,
      "grad_norm": 4.019230365753174,
      "learning_rate": 1.6066066066066067e-06,
      "loss": 0.3549,
      "step": 27950
    },
    {
      "epoch": 83.96,
      "grad_norm": 4.257839202880859,
      "learning_rate": 1.6036036036036038e-06,
      "loss": 0.3258,
      "step": 27960
    },
    {
      "epoch": 83.99,
      "grad_norm": 4.626522541046143,
      "learning_rate": 1.6006006006006008e-06,
      "loss": 0.359,
      "step": 27970
    },
    {
      "epoch": 84.0,
      "eval_accuracy": 0.9195,
      "eval_loss": 0.3132447600364685,
      "eval_runtime": 26.239,
      "eval_samples_per_second": 381.112,
      "eval_steps_per_second": 1.524,
      "step": 27972
    },
    {
      "epoch": 84.02,
      "grad_norm": 3.7168285846710205,
      "learning_rate": 1.5975975975975977e-06,
      "loss": 0.2869,
      "step": 27980
    },
    {
      "epoch": 84.05,
      "grad_norm": 4.093753337860107,
      "learning_rate": 1.5945945945945947e-06,
      "loss": 0.3178,
      "step": 27990
    },
    {
      "epoch": 84.08,
      "grad_norm": 4.9695539474487305,
      "learning_rate": 1.5915915915915916e-06,
      "loss": 0.39,
      "step": 28000
    },
    {
      "epoch": 84.11,
      "grad_norm": 4.485487461090088,
      "learning_rate": 1.5885885885885888e-06,
      "loss": 0.3863,
      "step": 28010
    },
    {
      "epoch": 84.14,
      "grad_norm": 4.53755521774292,
      "learning_rate": 1.5855855855855857e-06,
      "loss": 0.3645,
      "step": 28020
    },
    {
      "epoch": 84.17,
      "grad_norm": 4.958364486694336,
      "learning_rate": 1.5825825825825827e-06,
      "loss": 0.3243,
      "step": 28030
    },
    {
      "epoch": 84.2,
      "grad_norm": 5.579946994781494,
      "learning_rate": 1.5795795795795796e-06,
      "loss": 0.3421,
      "step": 28040
    },
    {
      "epoch": 84.23,
      "grad_norm": 5.822466850280762,
      "learning_rate": 1.5765765765765766e-06,
      "loss": 0.3344,
      "step": 28050
    },
    {
      "epoch": 84.26,
      "grad_norm": 5.217677116394043,
      "learning_rate": 1.5735735735735739e-06,
      "loss": 0.353,
      "step": 28060
    },
    {
      "epoch": 84.29,
      "grad_norm": 4.095274925231934,
      "learning_rate": 1.5705705705705707e-06,
      "loss": 0.3462,
      "step": 28070
    },
    {
      "epoch": 84.32,
      "grad_norm": 5.008785247802734,
      "learning_rate": 1.5675675675675678e-06,
      "loss": 0.3739,
      "step": 28080
    },
    {
      "epoch": 84.35,
      "grad_norm": 4.568315029144287,
      "learning_rate": 1.5645645645645646e-06,
      "loss": 0.3519,
      "step": 28090
    },
    {
      "epoch": 84.38,
      "grad_norm": 4.050673961639404,
      "learning_rate": 1.5615615615615617e-06,
      "loss": 0.3452,
      "step": 28100
    },
    {
      "epoch": 84.41,
      "grad_norm": 9.300089836120605,
      "learning_rate": 1.5585585585585585e-06,
      "loss": 0.3172,
      "step": 28110
    },
    {
      "epoch": 84.44,
      "grad_norm": 5.783573150634766,
      "learning_rate": 1.5555555555555558e-06,
      "loss": 0.353,
      "step": 28120
    },
    {
      "epoch": 84.47,
      "grad_norm": 5.739729404449463,
      "learning_rate": 1.5525525525525526e-06,
      "loss": 0.3231,
      "step": 28130
    },
    {
      "epoch": 84.5,
      "grad_norm": 4.953646659851074,
      "learning_rate": 1.5495495495495497e-06,
      "loss": 0.3468,
      "step": 28140
    },
    {
      "epoch": 84.53,
      "grad_norm": 3.520857095718384,
      "learning_rate": 1.5465465465465467e-06,
      "loss": 0.3549,
      "step": 28150
    },
    {
      "epoch": 84.56,
      "grad_norm": 3.2164952754974365,
      "learning_rate": 1.5435435435435436e-06,
      "loss": 0.3355,
      "step": 28160
    },
    {
      "epoch": 84.59,
      "grad_norm": 3.103375196456909,
      "learning_rate": 1.5405405405405409e-06,
      "loss": 0.3678,
      "step": 28170
    },
    {
      "epoch": 84.62,
      "grad_norm": 5.175012588500977,
      "learning_rate": 1.5375375375375377e-06,
      "loss": 0.3786,
      "step": 28180
    },
    {
      "epoch": 84.65,
      "grad_norm": 3.8485748767852783,
      "learning_rate": 1.5345345345345348e-06,
      "loss": 0.3552,
      "step": 28190
    },
    {
      "epoch": 84.68,
      "grad_norm": 4.263490676879883,
      "learning_rate": 1.5315315315315316e-06,
      "loss": 0.3782,
      "step": 28200
    },
    {
      "epoch": 84.71,
      "grad_norm": 4.399095058441162,
      "learning_rate": 1.5285285285285287e-06,
      "loss": 0.3599,
      "step": 28210
    },
    {
      "epoch": 84.74,
      "grad_norm": 5.025137424468994,
      "learning_rate": 1.5255255255255255e-06,
      "loss": 0.3655,
      "step": 28220
    },
    {
      "epoch": 84.77,
      "grad_norm": 4.902331829071045,
      "learning_rate": 1.5225225225225225e-06,
      "loss": 0.4037,
      "step": 28230
    },
    {
      "epoch": 84.8,
      "grad_norm": 4.457612991333008,
      "learning_rate": 1.5195195195195198e-06,
      "loss": 0.3524,
      "step": 28240
    },
    {
      "epoch": 84.83,
      "grad_norm": 3.5555665493011475,
      "learning_rate": 1.5165165165165167e-06,
      "loss": 0.3907,
      "step": 28250
    },
    {
      "epoch": 84.86,
      "grad_norm": 3.3783371448516846,
      "learning_rate": 1.5135135135135137e-06,
      "loss": 0.3443,
      "step": 28260
    },
    {
      "epoch": 84.89,
      "grad_norm": 4.076903820037842,
      "learning_rate": 1.5105105105105106e-06,
      "loss": 0.281,
      "step": 28270
    },
    {
      "epoch": 84.92,
      "grad_norm": 4.224771022796631,
      "learning_rate": 1.5075075075075076e-06,
      "loss": 0.3562,
      "step": 28280
    },
    {
      "epoch": 84.95,
      "grad_norm": 4.400293827056885,
      "learning_rate": 1.5045045045045045e-06,
      "loss": 0.3039,
      "step": 28290
    },
    {
      "epoch": 84.98,
      "grad_norm": 4.277791976928711,
      "learning_rate": 1.5015015015015017e-06,
      "loss": 0.3106,
      "step": 28300
    },
    {
      "epoch": 85.0,
      "eval_accuracy": 0.9204,
      "eval_loss": 0.31311744451522827,
      "eval_runtime": 26.4274,
      "eval_samples_per_second": 378.395,
      "eval_steps_per_second": 1.514,
      "step": 28305
    },
    {
      "epoch": 85.02,
      "grad_norm": 6.217394828796387,
      "learning_rate": 1.4984984984984986e-06,
      "loss": 0.3836,
      "step": 28310
    },
    {
      "epoch": 85.05,
      "grad_norm": 5.343406677246094,
      "learning_rate": 1.4954954954954956e-06,
      "loss": 0.3762,
      "step": 28320
    },
    {
      "epoch": 85.08,
      "grad_norm": 4.6541643142700195,
      "learning_rate": 1.4924924924924927e-06,
      "loss": 0.3498,
      "step": 28330
    },
    {
      "epoch": 85.11,
      "grad_norm": 4.976070880889893,
      "learning_rate": 1.4894894894894895e-06,
      "loss": 0.3771,
      "step": 28340
    },
    {
      "epoch": 85.14,
      "grad_norm": 5.701066017150879,
      "learning_rate": 1.4864864864864868e-06,
      "loss": 0.391,
      "step": 28350
    },
    {
      "epoch": 85.17,
      "grad_norm": 4.469409465789795,
      "learning_rate": 1.4834834834834836e-06,
      "loss": 0.3437,
      "step": 28360
    },
    {
      "epoch": 85.2,
      "grad_norm": 4.064690589904785,
      "learning_rate": 1.4804804804804807e-06,
      "loss": 0.3214,
      "step": 28370
    },
    {
      "epoch": 85.23,
      "grad_norm": 6.355717182159424,
      "learning_rate": 1.4774774774774775e-06,
      "loss": 0.3571,
      "step": 28380
    },
    {
      "epoch": 85.26,
      "grad_norm": 3.9010183811187744,
      "learning_rate": 1.4744744744744746e-06,
      "loss": 0.3826,
      "step": 28390
    },
    {
      "epoch": 85.29,
      "grad_norm": 3.8871779441833496,
      "learning_rate": 1.4714714714714714e-06,
      "loss": 0.3368,
      "step": 28400
    },
    {
      "epoch": 85.32,
      "grad_norm": 3.0758235454559326,
      "learning_rate": 1.4684684684684685e-06,
      "loss": 0.3852,
      "step": 28410
    },
    {
      "epoch": 85.35,
      "grad_norm": 7.367844104766846,
      "learning_rate": 1.4654654654654657e-06,
      "loss": 0.3843,
      "step": 28420
    },
    {
      "epoch": 85.38,
      "grad_norm": 5.271114826202393,
      "learning_rate": 1.4624624624624626e-06,
      "loss": 0.3789,
      "step": 28430
    },
    {
      "epoch": 85.41,
      "grad_norm": 4.646884441375732,
      "learning_rate": 1.4594594594594596e-06,
      "loss": 0.2839,
      "step": 28440
    },
    {
      "epoch": 85.44,
      "grad_norm": 3.69492244720459,
      "learning_rate": 1.4564564564564565e-06,
      "loss": 0.419,
      "step": 28450
    },
    {
      "epoch": 85.47,
      "grad_norm": 4.678412437438965,
      "learning_rate": 1.4534534534534535e-06,
      "loss": 0.3551,
      "step": 28460
    },
    {
      "epoch": 85.5,
      "grad_norm": 7.434144020080566,
      "learning_rate": 1.4504504504504504e-06,
      "loss": 0.3454,
      "step": 28470
    },
    {
      "epoch": 85.53,
      "grad_norm": 3.7736716270446777,
      "learning_rate": 1.4474474474474477e-06,
      "loss": 0.3517,
      "step": 28480
    },
    {
      "epoch": 85.56,
      "grad_norm": 4.585010528564453,
      "learning_rate": 1.4444444444444445e-06,
      "loss": 0.3951,
      "step": 28490
    },
    {
      "epoch": 85.59,
      "grad_norm": 5.308398246765137,
      "learning_rate": 1.4414414414414416e-06,
      "loss": 0.3599,
      "step": 28500
    },
    {
      "epoch": 85.62,
      "grad_norm": 5.781804084777832,
      "learning_rate": 1.4384384384384386e-06,
      "loss": 0.3379,
      "step": 28510
    },
    {
      "epoch": 85.65,
      "grad_norm": 4.5535359382629395,
      "learning_rate": 1.4354354354354354e-06,
      "loss": 0.3562,
      "step": 28520
    },
    {
      "epoch": 85.68,
      "grad_norm": 4.323468208312988,
      "learning_rate": 1.4324324324324327e-06,
      "loss": 0.3539,
      "step": 28530
    },
    {
      "epoch": 85.71,
      "grad_norm": 3.2731239795684814,
      "learning_rate": 1.4294294294294296e-06,
      "loss": 0.3152,
      "step": 28540
    },
    {
      "epoch": 85.74,
      "grad_norm": 3.9777438640594482,
      "learning_rate": 1.4264264264264266e-06,
      "loss": 0.3599,
      "step": 28550
    },
    {
      "epoch": 85.77,
      "grad_norm": 3.968613862991333,
      "learning_rate": 1.4234234234234235e-06,
      "loss": 0.3435,
      "step": 28560
    },
    {
      "epoch": 85.8,
      "grad_norm": 3.4560418128967285,
      "learning_rate": 1.4204204204204205e-06,
      "loss": 0.4009,
      "step": 28570
    },
    {
      "epoch": 85.83,
      "grad_norm": 6.11572265625,
      "learning_rate": 1.4174174174174174e-06,
      "loss": 0.3233,
      "step": 28580
    },
    {
      "epoch": 85.86,
      "grad_norm": 6.127475261688232,
      "learning_rate": 1.4144144144144144e-06,
      "loss": 0.3681,
      "step": 28590
    },
    {
      "epoch": 85.89,
      "grad_norm": 3.571716547012329,
      "learning_rate": 1.4114114114114117e-06,
      "loss": 0.3432,
      "step": 28600
    },
    {
      "epoch": 85.92,
      "grad_norm": 4.2647528648376465,
      "learning_rate": 1.4084084084084085e-06,
      "loss": 0.3512,
      "step": 28610
    },
    {
      "epoch": 85.95,
      "grad_norm": 3.867506265640259,
      "learning_rate": 1.4054054054054056e-06,
      "loss": 0.3219,
      "step": 28620
    },
    {
      "epoch": 85.98,
      "grad_norm": 4.499976634979248,
      "learning_rate": 1.4024024024024024e-06,
      "loss": 0.3191,
      "step": 28630
    },
    {
      "epoch": 86.0,
      "eval_accuracy": 0.9201,
      "eval_loss": 0.31295347213745117,
      "eval_runtime": 26.6018,
      "eval_samples_per_second": 375.914,
      "eval_steps_per_second": 1.504,
      "step": 28638
    },
    {
      "epoch": 86.01,
      "grad_norm": 5.146522521972656,
      "learning_rate": 1.3993993993993995e-06,
      "loss": 0.3677,
      "step": 28640
    },
    {
      "epoch": 86.04,
      "grad_norm": 3.925790548324585,
      "learning_rate": 1.3963963963963963e-06,
      "loss": 0.3256,
      "step": 28650
    },
    {
      "epoch": 86.07,
      "grad_norm": 5.661437034606934,
      "learning_rate": 1.3933933933933936e-06,
      "loss": 0.3453,
      "step": 28660
    },
    {
      "epoch": 86.1,
      "grad_norm": 2.622417449951172,
      "learning_rate": 1.3903903903903904e-06,
      "loss": 0.3071,
      "step": 28670
    },
    {
      "epoch": 86.13,
      "grad_norm": 6.432394981384277,
      "learning_rate": 1.3873873873873875e-06,
      "loss": 0.3381,
      "step": 28680
    },
    {
      "epoch": 86.16,
      "grad_norm": 3.8260176181793213,
      "learning_rate": 1.3843843843843845e-06,
      "loss": 0.3447,
      "step": 28690
    },
    {
      "epoch": 86.19,
      "grad_norm": 4.630548000335693,
      "learning_rate": 1.3813813813813814e-06,
      "loss": 0.3053,
      "step": 28700
    },
    {
      "epoch": 86.22,
      "grad_norm": 4.513822078704834,
      "learning_rate": 1.3783783783783786e-06,
      "loss": 0.3923,
      "step": 28710
    },
    {
      "epoch": 86.25,
      "grad_norm": 3.1275057792663574,
      "learning_rate": 1.3753753753753755e-06,
      "loss": 0.3772,
      "step": 28720
    },
    {
      "epoch": 86.28,
      "grad_norm": 6.163814067840576,
      "learning_rate": 1.3723723723723725e-06,
      "loss": 0.3252,
      "step": 28730
    },
    {
      "epoch": 86.31,
      "grad_norm": 5.354124069213867,
      "learning_rate": 1.3693693693693694e-06,
      "loss": 0.351,
      "step": 28740
    },
    {
      "epoch": 86.34,
      "grad_norm": 4.566099643707275,
      "learning_rate": 1.3663663663663664e-06,
      "loss": 0.3701,
      "step": 28750
    },
    {
      "epoch": 86.37,
      "grad_norm": 3.471346139907837,
      "learning_rate": 1.3633633633633637e-06,
      "loss": 0.3199,
      "step": 28760
    },
    {
      "epoch": 86.4,
      "grad_norm": 4.914207458496094,
      "learning_rate": 1.3603603603603606e-06,
      "loss": 0.3755,
      "step": 28770
    },
    {
      "epoch": 86.43,
      "grad_norm": 3.921909809112549,
      "learning_rate": 1.3573573573573576e-06,
      "loss": 0.368,
      "step": 28780
    },
    {
      "epoch": 86.46,
      "grad_norm": 3.637758493423462,
      "learning_rate": 1.3543543543543545e-06,
      "loss": 0.3666,
      "step": 28790
    },
    {
      "epoch": 86.49,
      "grad_norm": 4.792076587677002,
      "learning_rate": 1.3513513513513515e-06,
      "loss": 0.3497,
      "step": 28800
    },
    {
      "epoch": 86.52,
      "grad_norm": 5.629250526428223,
      "learning_rate": 1.3483483483483484e-06,
      "loss": 0.3494,
      "step": 28810
    },
    {
      "epoch": 86.55,
      "grad_norm": 5.732989311218262,
      "learning_rate": 1.3453453453453454e-06,
      "loss": 0.3762,
      "step": 28820
    },
    {
      "epoch": 86.58,
      "grad_norm": 8.068380355834961,
      "learning_rate": 1.3423423423423422e-06,
      "loss": 0.3712,
      "step": 28830
    },
    {
      "epoch": 86.61,
      "grad_norm": 6.579488754272461,
      "learning_rate": 1.3393393393393395e-06,
      "loss": 0.3897,
      "step": 28840
    },
    {
      "epoch": 86.64,
      "grad_norm": 5.224200248718262,
      "learning_rate": 1.3363363363363366e-06,
      "loss": 0.3771,
      "step": 28850
    },
    {
      "epoch": 86.67,
      "grad_norm": 3.9246394634246826,
      "learning_rate": 1.3333333333333334e-06,
      "loss": 0.3353,
      "step": 28860
    },
    {
      "epoch": 86.7,
      "grad_norm": 6.095767498016357,
      "learning_rate": 1.3303303303303305e-06,
      "loss": 0.3718,
      "step": 28870
    },
    {
      "epoch": 86.73,
      "grad_norm": 4.3293776512146,
      "learning_rate": 1.3273273273273273e-06,
      "loss": 0.3086,
      "step": 28880
    },
    {
      "epoch": 86.76,
      "grad_norm": 3.3233001232147217,
      "learning_rate": 1.3243243243243246e-06,
      "loss": 0.387,
      "step": 28890
    },
    {
      "epoch": 86.79,
      "grad_norm": 4.108912944793701,
      "learning_rate": 1.3213213213213214e-06,
      "loss": 0.3706,
      "step": 28900
    },
    {
      "epoch": 86.82,
      "grad_norm": 5.592580318450928,
      "learning_rate": 1.3183183183183185e-06,
      "loss": 0.3254,
      "step": 28910
    },
    {
      "epoch": 86.85,
      "grad_norm": 3.880584716796875,
      "learning_rate": 1.3153153153153153e-06,
      "loss": 0.293,
      "step": 28920
    },
    {
      "epoch": 86.88,
      "grad_norm": 4.027377605438232,
      "learning_rate": 1.3123123123123124e-06,
      "loss": 0.3652,
      "step": 28930
    },
    {
      "epoch": 86.91,
      "grad_norm": 4.788450717926025,
      "learning_rate": 1.3093093093093096e-06,
      "loss": 0.3706,
      "step": 28940
    },
    {
      "epoch": 86.94,
      "grad_norm": 6.277264595031738,
      "learning_rate": 1.3063063063063065e-06,
      "loss": 0.3529,
      "step": 28950
    },
    {
      "epoch": 86.97,
      "grad_norm": 4.262212753295898,
      "learning_rate": 1.3033033033033035e-06,
      "loss": 0.2969,
      "step": 28960
    },
    {
      "epoch": 87.0,
      "grad_norm": 6.489366054534912,
      "learning_rate": 1.3003003003003004e-06,
      "loss": 0.3987,
      "step": 28970
    },
    {
      "epoch": 87.0,
      "eval_accuracy": 0.9202,
      "eval_loss": 0.3141060769557953,
      "eval_runtime": 26.0453,
      "eval_samples_per_second": 383.947,
      "eval_steps_per_second": 1.536,
      "step": 28971
    },
    {
      "epoch": 87.03,
      "grad_norm": 13.887879371643066,
      "learning_rate": 1.2972972972972974e-06,
      "loss": 0.4085,
      "step": 28980
    },
    {
      "epoch": 87.06,
      "grad_norm": 5.408322811126709,
      "learning_rate": 1.2942942942942943e-06,
      "loss": 0.3801,
      "step": 28990
    },
    {
      "epoch": 87.09,
      "grad_norm": 3.528775691986084,
      "learning_rate": 1.2912912912912913e-06,
      "loss": 0.3011,
      "step": 29000
    },
    {
      "epoch": 87.12,
      "grad_norm": 5.333277225494385,
      "learning_rate": 1.2882882882882882e-06,
      "loss": 0.3318,
      "step": 29010
    },
    {
      "epoch": 87.15,
      "grad_norm": 4.863485336303711,
      "learning_rate": 1.2852852852852854e-06,
      "loss": 0.391,
      "step": 29020
    },
    {
      "epoch": 87.18,
      "grad_norm": 4.352230072021484,
      "learning_rate": 1.2822822822822825e-06,
      "loss": 0.3088,
      "step": 29030
    },
    {
      "epoch": 87.21,
      "grad_norm": 5.921619892120361,
      "learning_rate": 1.2792792792792793e-06,
      "loss": 0.3323,
      "step": 29040
    },
    {
      "epoch": 87.24,
      "grad_norm": 3.968787431716919,
      "learning_rate": 1.2762762762762764e-06,
      "loss": 0.3412,
      "step": 29050
    },
    {
      "epoch": 87.27,
      "grad_norm": 5.20448637008667,
      "learning_rate": 1.2732732732732732e-06,
      "loss": 0.3408,
      "step": 29060
    },
    {
      "epoch": 87.3,
      "grad_norm": 4.800923824310303,
      "learning_rate": 1.2702702702702705e-06,
      "loss": 0.3507,
      "step": 29070
    },
    {
      "epoch": 87.33,
      "grad_norm": 5.60935640335083,
      "learning_rate": 1.2672672672672674e-06,
      "loss": 0.3395,
      "step": 29080
    },
    {
      "epoch": 87.36,
      "grad_norm": 4.371265888214111,
      "learning_rate": 1.2642642642642644e-06,
      "loss": 0.3354,
      "step": 29090
    },
    {
      "epoch": 87.39,
      "grad_norm": 4.773227691650391,
      "learning_rate": 1.2612612612612613e-06,
      "loss": 0.3183,
      "step": 29100
    },
    {
      "epoch": 87.42,
      "grad_norm": 3.942945718765259,
      "learning_rate": 1.2582582582582583e-06,
      "loss": 0.3217,
      "step": 29110
    },
    {
      "epoch": 87.45,
      "grad_norm": 3.5037994384765625,
      "learning_rate": 1.2552552552552556e-06,
      "loss": 0.4045,
      "step": 29120
    },
    {
      "epoch": 87.48,
      "grad_norm": 3.3998568058013916,
      "learning_rate": 1.2522522522522524e-06,
      "loss": 0.3525,
      "step": 29130
    },
    {
      "epoch": 87.51,
      "grad_norm": 4.385222911834717,
      "learning_rate": 1.2492492492492493e-06,
      "loss": 0.3798,
      "step": 29140
    },
    {
      "epoch": 87.54,
      "grad_norm": 3.4853687286376953,
      "learning_rate": 1.2462462462462463e-06,
      "loss": 0.3678,
      "step": 29150
    },
    {
      "epoch": 87.57,
      "grad_norm": 4.248010635375977,
      "learning_rate": 1.2432432432432434e-06,
      "loss": 0.3285,
      "step": 29160
    },
    {
      "epoch": 87.6,
      "grad_norm": 3.9796533584594727,
      "learning_rate": 1.2402402402402404e-06,
      "loss": 0.3474,
      "step": 29170
    },
    {
      "epoch": 87.63,
      "grad_norm": 3.916168689727783,
      "learning_rate": 1.2372372372372375e-06,
      "loss": 0.3478,
      "step": 29180
    },
    {
      "epoch": 87.66,
      "grad_norm": 5.578619956970215,
      "learning_rate": 1.2342342342342343e-06,
      "loss": 0.3995,
      "step": 29190
    },
    {
      "epoch": 87.69,
      "grad_norm": 4.324888706207275,
      "learning_rate": 1.2312312312312314e-06,
      "loss": 0.3168,
      "step": 29200
    },
    {
      "epoch": 87.72,
      "grad_norm": 6.634174346923828,
      "learning_rate": 1.2282282282282282e-06,
      "loss": 0.3369,
      "step": 29210
    },
    {
      "epoch": 87.75,
      "grad_norm": 4.07280158996582,
      "learning_rate": 1.2252252252252253e-06,
      "loss": 0.3602,
      "step": 29220
    },
    {
      "epoch": 87.78,
      "grad_norm": 4.786525249481201,
      "learning_rate": 1.2222222222222223e-06,
      "loss": 0.3611,
      "step": 29230
    },
    {
      "epoch": 87.81,
      "grad_norm": 9.206954956054688,
      "learning_rate": 1.2192192192192192e-06,
      "loss": 0.3904,
      "step": 29240
    },
    {
      "epoch": 87.84,
      "grad_norm": 4.202898979187012,
      "learning_rate": 1.2162162162162164e-06,
      "loss": 0.3612,
      "step": 29250
    },
    {
      "epoch": 87.87,
      "grad_norm": 11.730929374694824,
      "learning_rate": 1.2132132132132133e-06,
      "loss": 0.3367,
      "step": 29260
    },
    {
      "epoch": 87.9,
      "grad_norm": 4.191562652587891,
      "learning_rate": 1.2102102102102103e-06,
      "loss": 0.3062,
      "step": 29270
    },
    {
      "epoch": 87.93,
      "grad_norm": 4.959072113037109,
      "learning_rate": 1.2072072072072074e-06,
      "loss": 0.3251,
      "step": 29280
    },
    {
      "epoch": 87.96,
      "grad_norm": 4.947381973266602,
      "learning_rate": 1.2042042042042042e-06,
      "loss": 0.4022,
      "step": 29290
    },
    {
      "epoch": 87.99,
      "grad_norm": 4.479344367980957,
      "learning_rate": 1.2012012012012013e-06,
      "loss": 0.3327,
      "step": 29300
    },
    {
      "epoch": 88.0,
      "eval_accuracy": 0.9194,
      "eval_loss": 0.3137800395488739,
      "eval_runtime": 26.1193,
      "eval_samples_per_second": 382.859,
      "eval_steps_per_second": 1.531,
      "step": 29304
    },
    {
      "epoch": 88.02,
      "grad_norm": 6.0303826332092285,
      "learning_rate": 1.1981981981981983e-06,
      "loss": 0.3329,
      "step": 29310
    },
    {
      "epoch": 88.05,
      "grad_norm": 4.474847793579102,
      "learning_rate": 1.1951951951951952e-06,
      "loss": 0.3779,
      "step": 29320
    },
    {
      "epoch": 88.08,
      "grad_norm": 4.928411960601807,
      "learning_rate": 1.1921921921921922e-06,
      "loss": 0.3574,
      "step": 29330
    },
    {
      "epoch": 88.11,
      "grad_norm": 9.692339897155762,
      "learning_rate": 1.1891891891891893e-06,
      "loss": 0.3493,
      "step": 29340
    },
    {
      "epoch": 88.14,
      "grad_norm": 3.353285074234009,
      "learning_rate": 1.1861861861861864e-06,
      "loss": 0.3106,
      "step": 29350
    },
    {
      "epoch": 88.17,
      "grad_norm": 2.915330410003662,
      "learning_rate": 1.1831831831831834e-06,
      "loss": 0.3109,
      "step": 29360
    },
    {
      "epoch": 88.2,
      "grad_norm": 3.968358039855957,
      "learning_rate": 1.1801801801801803e-06,
      "loss": 0.3418,
      "step": 29370
    },
    {
      "epoch": 88.23,
      "grad_norm": 4.716326713562012,
      "learning_rate": 1.1771771771771773e-06,
      "loss": 0.4068,
      "step": 29380
    },
    {
      "epoch": 88.26,
      "grad_norm": 3.893982172012329,
      "learning_rate": 1.1741741741741742e-06,
      "loss": 0.3411,
      "step": 29390
    },
    {
      "epoch": 88.29,
      "grad_norm": 2.622589588165283,
      "learning_rate": 1.1711711711711712e-06,
      "loss": 0.2821,
      "step": 29400
    },
    {
      "epoch": 88.32,
      "grad_norm": 5.168035507202148,
      "learning_rate": 1.1681681681681683e-06,
      "loss": 0.312,
      "step": 29410
    },
    {
      "epoch": 88.35,
      "grad_norm": 4.591063499450684,
      "learning_rate": 1.1651651651651651e-06,
      "loss": 0.3603,
      "step": 29420
    },
    {
      "epoch": 88.38,
      "grad_norm": 5.766468048095703,
      "learning_rate": 1.1621621621621624e-06,
      "loss": 0.2897,
      "step": 29430
    },
    {
      "epoch": 88.41,
      "grad_norm": 4.102199077606201,
      "learning_rate": 1.1591591591591592e-06,
      "loss": 0.3404,
      "step": 29440
    },
    {
      "epoch": 88.44,
      "grad_norm": 3.0915396213531494,
      "learning_rate": 1.1561561561561563e-06,
      "loss": 0.3478,
      "step": 29450
    },
    {
      "epoch": 88.47,
      "grad_norm": 5.2455363273620605,
      "learning_rate": 1.1531531531531533e-06,
      "loss": 0.3793,
      "step": 29460
    },
    {
      "epoch": 88.5,
      "grad_norm": 4.164064884185791,
      "learning_rate": 1.1501501501501502e-06,
      "loss": 0.3344,
      "step": 29470
    },
    {
      "epoch": 88.53,
      "grad_norm": 3.9158265590667725,
      "learning_rate": 1.1471471471471472e-06,
      "loss": 0.3435,
      "step": 29480
    },
    {
      "epoch": 88.56,
      "grad_norm": 5.330009460449219,
      "learning_rate": 1.1441441441441443e-06,
      "loss": 0.3521,
      "step": 29490
    },
    {
      "epoch": 88.59,
      "grad_norm": 3.8261780738830566,
      "learning_rate": 1.1411411411411411e-06,
      "loss": 0.3143,
      "step": 29500
    },
    {
      "epoch": 88.62,
      "grad_norm": 4.802837371826172,
      "learning_rate": 1.1381381381381382e-06,
      "loss": 0.369,
      "step": 29510
    },
    {
      "epoch": 88.65,
      "grad_norm": 6.1245527267456055,
      "learning_rate": 1.1351351351351352e-06,
      "loss": 0.3668,
      "step": 29520
    },
    {
      "epoch": 88.68,
      "grad_norm": 6.6061296463012695,
      "learning_rate": 1.1321321321321323e-06,
      "loss": 0.3625,
      "step": 29530
    },
    {
      "epoch": 88.71,
      "grad_norm": 5.635161399841309,
      "learning_rate": 1.1291291291291293e-06,
      "loss": 0.3752,
      "step": 29540
    },
    {
      "epoch": 88.74,
      "grad_norm": 4.247094631195068,
      "learning_rate": 1.1261261261261262e-06,
      "loss": 0.3716,
      "step": 29550
    },
    {
      "epoch": 88.77,
      "grad_norm": 5.640427589416504,
      "learning_rate": 1.1231231231231232e-06,
      "loss": 0.3431,
      "step": 29560
    },
    {
      "epoch": 88.8,
      "grad_norm": 4.967761039733887,
      "learning_rate": 1.1201201201201203e-06,
      "loss": 0.3315,
      "step": 29570
    },
    {
      "epoch": 88.83,
      "grad_norm": 5.282632350921631,
      "learning_rate": 1.1171171171171171e-06,
      "loss": 0.3382,
      "step": 29580
    },
    {
      "epoch": 88.86,
      "grad_norm": 4.708310604095459,
      "learning_rate": 1.1141141141141142e-06,
      "loss": 0.3792,
      "step": 29590
    },
    {
      "epoch": 88.89,
      "grad_norm": 4.220743179321289,
      "learning_rate": 1.111111111111111e-06,
      "loss": 0.3738,
      "step": 29600
    },
    {
      "epoch": 88.92,
      "grad_norm": 2.9908206462860107,
      "learning_rate": 1.1081081081081083e-06,
      "loss": 0.3861,
      "step": 29610
    },
    {
      "epoch": 88.95,
      "grad_norm": 4.38171911239624,
      "learning_rate": 1.1051051051051051e-06,
      "loss": 0.3271,
      "step": 29620
    },
    {
      "epoch": 88.98,
      "grad_norm": 4.509523868560791,
      "learning_rate": 1.1021021021021022e-06,
      "loss": 0.3464,
      "step": 29630
    },
    {
      "epoch": 89.0,
      "eval_accuracy": 0.9207,
      "eval_loss": 0.31416377425193787,
      "eval_runtime": 26.3227,
      "eval_samples_per_second": 379.901,
      "eval_steps_per_second": 1.52,
      "step": 29637
    },
    {
      "epoch": 89.01,
      "grad_norm": 4.576425552368164,
      "learning_rate": 1.0990990990990993e-06,
      "loss": 0.2634,
      "step": 29640
    },
    {
      "epoch": 89.04,
      "grad_norm": 5.361783981323242,
      "learning_rate": 1.096096096096096e-06,
      "loss": 0.3808,
      "step": 29650
    },
    {
      "epoch": 89.07,
      "grad_norm": 3.7039408683776855,
      "learning_rate": 1.0930930930930932e-06,
      "loss": 0.3201,
      "step": 29660
    },
    {
      "epoch": 89.1,
      "grad_norm": 2.8414804935455322,
      "learning_rate": 1.0900900900900902e-06,
      "loss": 0.3634,
      "step": 29670
    },
    {
      "epoch": 89.13,
      "grad_norm": 6.98696756362915,
      "learning_rate": 1.087087087087087e-06,
      "loss": 0.3846,
      "step": 29680
    },
    {
      "epoch": 89.16,
      "grad_norm": 2.858682632446289,
      "learning_rate": 1.0840840840840841e-06,
      "loss": 0.3495,
      "step": 29690
    },
    {
      "epoch": 89.19,
      "grad_norm": 4.683915138244629,
      "learning_rate": 1.0810810810810812e-06,
      "loss": 0.3582,
      "step": 29700
    },
    {
      "epoch": 89.22,
      "grad_norm": 4.9959893226623535,
      "learning_rate": 1.0780780780780782e-06,
      "loss": 0.3459,
      "step": 29710
    },
    {
      "epoch": 89.25,
      "grad_norm": 3.1686313152313232,
      "learning_rate": 1.0750750750750753e-06,
      "loss": 0.3503,
      "step": 29720
    },
    {
      "epoch": 89.28,
      "grad_norm": 6.092320442199707,
      "learning_rate": 1.0720720720720721e-06,
      "loss": 0.3547,
      "step": 29730
    },
    {
      "epoch": 89.31,
      "grad_norm": 5.362958908081055,
      "learning_rate": 1.0690690690690692e-06,
      "loss": 0.3461,
      "step": 29740
    },
    {
      "epoch": 89.34,
      "grad_norm": 3.640472650527954,
      "learning_rate": 1.0660660660660662e-06,
      "loss": 0.3494,
      "step": 29750
    },
    {
      "epoch": 89.37,
      "grad_norm": 4.0758819580078125,
      "learning_rate": 1.063063063063063e-06,
      "loss": 0.259,
      "step": 29760
    },
    {
      "epoch": 89.4,
      "grad_norm": 4.850218296051025,
      "learning_rate": 1.0600600600600601e-06,
      "loss": 0.3851,
      "step": 29770
    },
    {
      "epoch": 89.43,
      "grad_norm": 5.28075647354126,
      "learning_rate": 1.0570570570570572e-06,
      "loss": 0.3388,
      "step": 29780
    },
    {
      "epoch": 89.46,
      "grad_norm": 4.323512077331543,
      "learning_rate": 1.0540540540540542e-06,
      "loss": 0.3485,
      "step": 29790
    },
    {
      "epoch": 89.49,
      "grad_norm": 4.59760856628418,
      "learning_rate": 1.051051051051051e-06,
      "loss": 0.3657,
      "step": 29800
    },
    {
      "epoch": 89.52,
      "grad_norm": 4.18367338180542,
      "learning_rate": 1.0480480480480481e-06,
      "loss": 0.3657,
      "step": 29810
    },
    {
      "epoch": 89.55,
      "grad_norm": 5.913191318511963,
      "learning_rate": 1.0450450450450452e-06,
      "loss": 0.3689,
      "step": 29820
    },
    {
      "epoch": 89.58,
      "grad_norm": 5.122714996337891,
      "learning_rate": 1.042042042042042e-06,
      "loss": 0.3297,
      "step": 29830
    },
    {
      "epoch": 89.61,
      "grad_norm": 5.049954891204834,
      "learning_rate": 1.039039039039039e-06,
      "loss": 0.3057,
      "step": 29840
    },
    {
      "epoch": 89.64,
      "grad_norm": 4.258790969848633,
      "learning_rate": 1.0360360360360361e-06,
      "loss": 0.3057,
      "step": 29850
    },
    {
      "epoch": 89.67,
      "grad_norm": 5.7420830726623535,
      "learning_rate": 1.033033033033033e-06,
      "loss": 0.3425,
      "step": 29860
    },
    {
      "epoch": 89.7,
      "grad_norm": 4.723175048828125,
      "learning_rate": 1.03003003003003e-06,
      "loss": 0.3643,
      "step": 29870
    },
    {
      "epoch": 89.73,
      "grad_norm": 5.273463249206543,
      "learning_rate": 1.027027027027027e-06,
      "loss": 0.4133,
      "step": 29880
    },
    {
      "epoch": 89.76,
      "grad_norm": 4.2874298095703125,
      "learning_rate": 1.0240240240240242e-06,
      "loss": 0.3049,
      "step": 29890
    },
    {
      "epoch": 89.79,
      "grad_norm": 4.636884689331055,
      "learning_rate": 1.0210210210210212e-06,
      "loss": 0.362,
      "step": 29900
    },
    {
      "epoch": 89.82,
      "grad_norm": 4.748985767364502,
      "learning_rate": 1.018018018018018e-06,
      "loss": 0.3403,
      "step": 29910
    },
    {
      "epoch": 89.85,
      "grad_norm": 3.7866580486297607,
      "learning_rate": 1.015015015015015e-06,
      "loss": 0.3132,
      "step": 29920
    },
    {
      "epoch": 89.88,
      "grad_norm": 5.007419586181641,
      "learning_rate": 1.0120120120120122e-06,
      "loss": 0.3786,
      "step": 29930
    },
    {
      "epoch": 89.91,
      "grad_norm": 4.309643268585205,
      "learning_rate": 1.009009009009009e-06,
      "loss": 0.3146,
      "step": 29940
    },
    {
      "epoch": 89.94,
      "grad_norm": 4.714195251464844,
      "learning_rate": 1.006006006006006e-06,
      "loss": 0.3255,
      "step": 29950
    },
    {
      "epoch": 89.97,
      "grad_norm": 5.178619861602783,
      "learning_rate": 1.0030030030030031e-06,
      "loss": 0.3441,
      "step": 29960
    },
    {
      "epoch": 90.0,
      "grad_norm": 16.105178833007812,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 0.3634,
      "step": 29970
    },
    {
      "epoch": 90.0,
      "eval_accuracy": 0.9207,
      "eval_loss": 0.31448256969451904,
      "eval_runtime": 26.3521,
      "eval_samples_per_second": 379.476,
      "eval_steps_per_second": 1.518,
      "step": 29970
    },
    {
      "epoch": 90.03,
      "grad_norm": 6.5296220779418945,
      "learning_rate": 9.969969969969972e-07,
      "loss": 0.3286,
      "step": 29980
    },
    {
      "epoch": 90.06,
      "grad_norm": 3.683039903640747,
      "learning_rate": 9.93993993993994e-07,
      "loss": 0.3542,
      "step": 29990
    },
    {
      "epoch": 90.09,
      "grad_norm": 5.204434871673584,
      "learning_rate": 9.909909909909911e-07,
      "loss": 0.4288,
      "step": 30000
    },
    {
      "epoch": 90.12,
      "grad_norm": 3.1721761226654053,
      "learning_rate": 9.87987987987988e-07,
      "loss": 0.3543,
      "step": 30010
    },
    {
      "epoch": 90.15,
      "grad_norm": 4.88484001159668,
      "learning_rate": 9.84984984984985e-07,
      "loss": 0.3097,
      "step": 30020
    },
    {
      "epoch": 90.18,
      "grad_norm": 5.034382343292236,
      "learning_rate": 9.81981981981982e-07,
      "loss": 0.3633,
      "step": 30030
    },
    {
      "epoch": 90.21,
      "grad_norm": 4.7634382247924805,
      "learning_rate": 9.78978978978979e-07,
      "loss": 0.3832,
      "step": 30040
    },
    {
      "epoch": 90.24,
      "grad_norm": 5.501009464263916,
      "learning_rate": 9.75975975975976e-07,
      "loss": 0.2972,
      "step": 30050
    },
    {
      "epoch": 90.27,
      "grad_norm": 3.3434524536132812,
      "learning_rate": 9.72972972972973e-07,
      "loss": 0.3391,
      "step": 30060
    },
    {
      "epoch": 90.3,
      "grad_norm": 7.160501480102539,
      "learning_rate": 9.6996996996997e-07,
      "loss": 0.3437,
      "step": 30070
    },
    {
      "epoch": 90.33,
      "grad_norm": 2.021047592163086,
      "learning_rate": 9.669669669669671e-07,
      "loss": 0.3584,
      "step": 30080
    },
    {
      "epoch": 90.36,
      "grad_norm": 4.661921977996826,
      "learning_rate": 9.63963963963964e-07,
      "loss": 0.3878,
      "step": 30090
    },
    {
      "epoch": 90.39,
      "grad_norm": 10.766969680786133,
      "learning_rate": 9.60960960960961e-07,
      "loss": 0.3046,
      "step": 30100
    },
    {
      "epoch": 90.42,
      "grad_norm": 4.374320983886719,
      "learning_rate": 9.57957957957958e-07,
      "loss": 0.3478,
      "step": 30110
    },
    {
      "epoch": 90.45,
      "grad_norm": 3.112799644470215,
      "learning_rate": 9.54954954954955e-07,
      "loss": 0.3234,
      "step": 30120
    },
    {
      "epoch": 90.48,
      "grad_norm": 3.8301897048950195,
      "learning_rate": 9.51951951951952e-07,
      "loss": 0.3249,
      "step": 30130
    },
    {
      "epoch": 90.51,
      "grad_norm": 5.908462047576904,
      "learning_rate": 9.48948948948949e-07,
      "loss": 0.3613,
      "step": 30140
    },
    {
      "epoch": 90.54,
      "grad_norm": 4.736770153045654,
      "learning_rate": 9.459459459459461e-07,
      "loss": 0.3848,
      "step": 30150
    },
    {
      "epoch": 90.57,
      "grad_norm": 4.781890392303467,
      "learning_rate": 9.42942942942943e-07,
      "loss": 0.3306,
      "step": 30160
    },
    {
      "epoch": 90.6,
      "grad_norm": 2.939910650253296,
      "learning_rate": 9.3993993993994e-07,
      "loss": 0.3135,
      "step": 30170
    },
    {
      "epoch": 90.63,
      "grad_norm": 4.981436252593994,
      "learning_rate": 9.36936936936937e-07,
      "loss": 0.346,
      "step": 30180
    },
    {
      "epoch": 90.66,
      "grad_norm": 5.331618785858154,
      "learning_rate": 9.33933933933934e-07,
      "loss": 0.3743,
      "step": 30190
    },
    {
      "epoch": 90.69,
      "grad_norm": 4.346016883850098,
      "learning_rate": 9.30930930930931e-07,
      "loss": 0.3656,
      "step": 30200
    },
    {
      "epoch": 90.72,
      "grad_norm": 4.141422271728516,
      "learning_rate": 9.27927927927928e-07,
      "loss": 0.3329,
      "step": 30210
    },
    {
      "epoch": 90.75,
      "grad_norm": 3.1103806495666504,
      "learning_rate": 9.24924924924925e-07,
      "loss": 0.3763,
      "step": 30220
    },
    {
      "epoch": 90.78,
      "grad_norm": 4.134851932525635,
      "learning_rate": 9.219219219219221e-07,
      "loss": 0.3978,
      "step": 30230
    },
    {
      "epoch": 90.81,
      "grad_norm": 6.171693325042725,
      "learning_rate": 9.189189189189191e-07,
      "loss": 0.3793,
      "step": 30240
    },
    {
      "epoch": 90.84,
      "grad_norm": 4.899822235107422,
      "learning_rate": 9.15915915915916e-07,
      "loss": 0.3399,
      "step": 30250
    },
    {
      "epoch": 90.87,
      "grad_norm": 4.265652179718018,
      "learning_rate": 9.12912912912913e-07,
      "loss": 0.3002,
      "step": 30260
    },
    {
      "epoch": 90.9,
      "grad_norm": 6.970982074737549,
      "learning_rate": 9.0990990990991e-07,
      "loss": 0.3748,
      "step": 30270
    },
    {
      "epoch": 90.93,
      "grad_norm": 4.818155288696289,
      "learning_rate": 9.06906906906907e-07,
      "loss": 0.3689,
      "step": 30280
    },
    {
      "epoch": 90.96,
      "grad_norm": 5.782536029815674,
      "learning_rate": 9.039039039039039e-07,
      "loss": 0.3694,
      "step": 30290
    },
    {
      "epoch": 90.99,
      "grad_norm": 3.527733087539673,
      "learning_rate": 9.00900900900901e-07,
      "loss": 0.3123,
      "step": 30300
    },
    {
      "epoch": 91.0,
      "eval_accuracy": 0.9197,
      "eval_loss": 0.3132767379283905,
      "eval_runtime": 26.2954,
      "eval_samples_per_second": 380.295,
      "eval_steps_per_second": 1.521,
      "step": 30303
    },
    {
      "epoch": 91.02,
      "grad_norm": 4.539586544036865,
      "learning_rate": 8.978978978978979e-07,
      "loss": 0.3454,
      "step": 30310
    },
    {
      "epoch": 91.05,
      "grad_norm": 5.961507320404053,
      "learning_rate": 8.948948948948951e-07,
      "loss": 0.4019,
      "step": 30320
    },
    {
      "epoch": 91.08,
      "grad_norm": 4.384204864501953,
      "learning_rate": 8.91891891891892e-07,
      "loss": 0.3051,
      "step": 30330
    },
    {
      "epoch": 91.11,
      "grad_norm": 3.877408504486084,
      "learning_rate": 8.88888888888889e-07,
      "loss": 0.325,
      "step": 30340
    },
    {
      "epoch": 91.14,
      "grad_norm": 3.735556125640869,
      "learning_rate": 8.858858858858859e-07,
      "loss": 0.3338,
      "step": 30350
    },
    {
      "epoch": 91.17,
      "grad_norm": 3.6671531200408936,
      "learning_rate": 8.82882882882883e-07,
      "loss": 0.3344,
      "step": 30360
    },
    {
      "epoch": 91.2,
      "grad_norm": 3.481395959854126,
      "learning_rate": 8.798798798798799e-07,
      "loss": 0.3287,
      "step": 30370
    },
    {
      "epoch": 91.23,
      "grad_norm": 5.657441139221191,
      "learning_rate": 8.768768768768769e-07,
      "loss": 0.3212,
      "step": 30380
    },
    {
      "epoch": 91.26,
      "grad_norm": 3.508744955062866,
      "learning_rate": 8.738738738738739e-07,
      "loss": 0.3448,
      "step": 30390
    },
    {
      "epoch": 91.29,
      "grad_norm": 4.640291690826416,
      "learning_rate": 8.708708708708709e-07,
      "loss": 0.3743,
      "step": 30400
    },
    {
      "epoch": 91.32,
      "grad_norm": 4.97070837020874,
      "learning_rate": 8.67867867867868e-07,
      "loss": 0.3411,
      "step": 30410
    },
    {
      "epoch": 91.35,
      "grad_norm": 3.490299940109253,
      "learning_rate": 8.64864864864865e-07,
      "loss": 0.4105,
      "step": 30420
    },
    {
      "epoch": 91.38,
      "grad_norm": 3.402111768722534,
      "learning_rate": 8.618618618618619e-07,
      "loss": 0.3758,
      "step": 30430
    },
    {
      "epoch": 91.41,
      "grad_norm": 4.324429035186768,
      "learning_rate": 8.58858858858859e-07,
      "loss": 0.3526,
      "step": 30440
    },
    {
      "epoch": 91.44,
      "grad_norm": 4.069542407989502,
      "learning_rate": 8.55855855855856e-07,
      "loss": 0.303,
      "step": 30450
    },
    {
      "epoch": 91.47,
      "grad_norm": 4.107100009918213,
      "learning_rate": 8.528528528528529e-07,
      "loss": 0.3915,
      "step": 30460
    },
    {
      "epoch": 91.5,
      "grad_norm": 5.373152256011963,
      "learning_rate": 8.498498498498498e-07,
      "loss": 0.3678,
      "step": 30470
    },
    {
      "epoch": 91.53,
      "grad_norm": 3.144451379776001,
      "learning_rate": 8.468468468468469e-07,
      "loss": 0.3303,
      "step": 30480
    },
    {
      "epoch": 91.56,
      "grad_norm": 3.854186534881592,
      "learning_rate": 8.438438438438439e-07,
      "loss": 0.3971,
      "step": 30490
    },
    {
      "epoch": 91.59,
      "grad_norm": 4.241054058074951,
      "learning_rate": 8.40840840840841e-07,
      "loss": 0.3687,
      "step": 30500
    },
    {
      "epoch": 91.62,
      "grad_norm": 5.219269275665283,
      "learning_rate": 8.37837837837838e-07,
      "loss": 0.3367,
      "step": 30510
    },
    {
      "epoch": 91.65,
      "grad_norm": 4.694491863250732,
      "learning_rate": 8.348348348348349e-07,
      "loss": 0.3929,
      "step": 30520
    },
    {
      "epoch": 91.68,
      "grad_norm": 3.9892783164978027,
      "learning_rate": 8.31831831831832e-07,
      "loss": 0.3733,
      "step": 30530
    },
    {
      "epoch": 91.71,
      "grad_norm": 4.450688362121582,
      "learning_rate": 8.288288288288289e-07,
      "loss": 0.2947,
      "step": 30540
    },
    {
      "epoch": 91.74,
      "grad_norm": 3.672638177871704,
      "learning_rate": 8.258258258258259e-07,
      "loss": 0.3789,
      "step": 30550
    },
    {
      "epoch": 91.77,
      "grad_norm": 4.482837200164795,
      "learning_rate": 8.228228228228228e-07,
      "loss": 0.3346,
      "step": 30560
    },
    {
      "epoch": 91.8,
      "grad_norm": 6.066772937774658,
      "learning_rate": 8.198198198198199e-07,
      "loss": 0.3526,
      "step": 30570
    },
    {
      "epoch": 91.83,
      "grad_norm": 5.765991687774658,
      "learning_rate": 8.168168168168168e-07,
      "loss": 0.3528,
      "step": 30580
    },
    {
      "epoch": 91.86,
      "grad_norm": 3.7455556392669678,
      "learning_rate": 8.13813813813814e-07,
      "loss": 0.3708,
      "step": 30590
    },
    {
      "epoch": 91.89,
      "grad_norm": 3.7361276149749756,
      "learning_rate": 8.108108108108109e-07,
      "loss": 0.3176,
      "step": 30600
    },
    {
      "epoch": 91.92,
      "grad_norm": 3.8778076171875,
      "learning_rate": 8.078078078078079e-07,
      "loss": 0.3292,
      "step": 30610
    },
    {
      "epoch": 91.95,
      "grad_norm": 2.9516711235046387,
      "learning_rate": 8.048048048048049e-07,
      "loss": 0.2913,
      "step": 30620
    },
    {
      "epoch": 91.98,
      "grad_norm": 4.212029933929443,
      "learning_rate": 8.018018018018019e-07,
      "loss": 0.3029,
      "step": 30630
    },
    {
      "epoch": 92.0,
      "eval_accuracy": 0.92,
      "eval_loss": 0.31380370259284973,
      "eval_runtime": 25.7552,
      "eval_samples_per_second": 388.272,
      "eval_steps_per_second": 1.553,
      "step": 30636
    },
    {
      "epoch": 92.01,
      "grad_norm": 4.563537120819092,
      "learning_rate": 7.987987987987988e-07,
      "loss": 0.2843,
      "step": 30640
    },
    {
      "epoch": 92.04,
      "grad_norm": 5.380393028259277,
      "learning_rate": 7.957957957957958e-07,
      "loss": 0.3538,
      "step": 30650
    },
    {
      "epoch": 92.07,
      "grad_norm": 5.2313714027404785,
      "learning_rate": 7.927927927927928e-07,
      "loss": 0.36,
      "step": 30660
    },
    {
      "epoch": 92.1,
      "grad_norm": 4.818750858306885,
      "learning_rate": 7.897897897897898e-07,
      "loss": 0.3096,
      "step": 30670
    },
    {
      "epoch": 92.13,
      "grad_norm": 3.9785690307617188,
      "learning_rate": 7.867867867867869e-07,
      "loss": 0.2951,
      "step": 30680
    },
    {
      "epoch": 92.16,
      "grad_norm": 4.332957744598389,
      "learning_rate": 7.837837837837839e-07,
      "loss": 0.4125,
      "step": 30690
    },
    {
      "epoch": 92.19,
      "grad_norm": 4.675868034362793,
      "learning_rate": 7.807807807807808e-07,
      "loss": 0.3306,
      "step": 30700
    },
    {
      "epoch": 92.22,
      "grad_norm": 5.006810188293457,
      "learning_rate": 7.777777777777779e-07,
      "loss": 0.3898,
      "step": 30710
    },
    {
      "epoch": 92.25,
      "grad_norm": 5.126926898956299,
      "learning_rate": 7.747747747747748e-07,
      "loss": 0.355,
      "step": 30720
    },
    {
      "epoch": 92.28,
      "grad_norm": 5.087444305419922,
      "learning_rate": 7.717717717717718e-07,
      "loss": 0.3424,
      "step": 30730
    },
    {
      "epoch": 92.31,
      "grad_norm": 5.345471382141113,
      "learning_rate": 7.687687687687688e-07,
      "loss": 0.3396,
      "step": 30740
    },
    {
      "epoch": 92.34,
      "grad_norm": 6.194934844970703,
      "learning_rate": 7.657657657657658e-07,
      "loss": 0.3909,
      "step": 30750
    },
    {
      "epoch": 92.37,
      "grad_norm": 4.657262802124023,
      "learning_rate": 7.627627627627627e-07,
      "loss": 0.3342,
      "step": 30760
    },
    {
      "epoch": 92.4,
      "grad_norm": 18.64208984375,
      "learning_rate": 7.597597597597599e-07,
      "loss": 0.3375,
      "step": 30770
    },
    {
      "epoch": 92.43,
      "grad_norm": 5.266354084014893,
      "learning_rate": 7.567567567567569e-07,
      "loss": 0.3513,
      "step": 30780
    },
    {
      "epoch": 92.46,
      "grad_norm": 5.636119365692139,
      "learning_rate": 7.537537537537538e-07,
      "loss": 0.3579,
      "step": 30790
    },
    {
      "epoch": 92.49,
      "grad_norm": 3.2974066734313965,
      "learning_rate": 7.507507507507509e-07,
      "loss": 0.3296,
      "step": 30800
    },
    {
      "epoch": 92.52,
      "grad_norm": 4.60905647277832,
      "learning_rate": 7.477477477477478e-07,
      "loss": 0.2948,
      "step": 30810
    },
    {
      "epoch": 92.55,
      "grad_norm": 3.9108526706695557,
      "learning_rate": 7.447447447447448e-07,
      "loss": 0.3275,
      "step": 30820
    },
    {
      "epoch": 92.58,
      "grad_norm": 3.9147696495056152,
      "learning_rate": 7.417417417417418e-07,
      "loss": 0.3305,
      "step": 30830
    },
    {
      "epoch": 92.61,
      "grad_norm": 4.329918384552002,
      "learning_rate": 7.387387387387388e-07,
      "loss": 0.3481,
      "step": 30840
    },
    {
      "epoch": 92.64,
      "grad_norm": 4.343500137329102,
      "learning_rate": 7.357357357357357e-07,
      "loss": 0.3215,
      "step": 30850
    },
    {
      "epoch": 92.67,
      "grad_norm": 3.3542425632476807,
      "learning_rate": 7.327327327327329e-07,
      "loss": 0.3339,
      "step": 30860
    },
    {
      "epoch": 92.7,
      "grad_norm": 6.222061634063721,
      "learning_rate": 7.297297297297298e-07,
      "loss": 0.3503,
      "step": 30870
    },
    {
      "epoch": 92.73,
      "grad_norm": 3.2602195739746094,
      "learning_rate": 7.267267267267268e-07,
      "loss": 0.3426,
      "step": 30880
    },
    {
      "epoch": 92.76,
      "grad_norm": 3.431351661682129,
      "learning_rate": 7.237237237237238e-07,
      "loss": 0.3249,
      "step": 30890
    },
    {
      "epoch": 92.79,
      "grad_norm": 3.704606771469116,
      "learning_rate": 7.207207207207208e-07,
      "loss": 0.3544,
      "step": 30900
    },
    {
      "epoch": 92.82,
      "grad_norm": 5.403119087219238,
      "learning_rate": 7.177177177177177e-07,
      "loss": 0.3488,
      "step": 30910
    },
    {
      "epoch": 92.85,
      "grad_norm": 5.552067756652832,
      "learning_rate": 7.147147147147148e-07,
      "loss": 0.36,
      "step": 30920
    },
    {
      "epoch": 92.88,
      "grad_norm": 3.5900611877441406,
      "learning_rate": 7.117117117117117e-07,
      "loss": 0.3496,
      "step": 30930
    },
    {
      "epoch": 92.91,
      "grad_norm": 3.687861680984497,
      "learning_rate": 7.087087087087087e-07,
      "loss": 0.346,
      "step": 30940
    },
    {
      "epoch": 92.94,
      "grad_norm": 7.328057289123535,
      "learning_rate": 7.057057057057058e-07,
      "loss": 0.3628,
      "step": 30950
    },
    {
      "epoch": 92.97,
      "grad_norm": 4.4473137855529785,
      "learning_rate": 7.027027027027028e-07,
      "loss": 0.3814,
      "step": 30960
    },
    {
      "epoch": 93.0,
      "eval_accuracy": 0.9192,
      "eval_loss": 0.31237778067588806,
      "eval_runtime": 25.9576,
      "eval_samples_per_second": 385.244,
      "eval_steps_per_second": 1.541,
      "step": 30969
    },
    {
      "epoch": 93.0,
      "grad_norm": 4.461737632751465,
      "learning_rate": 6.996996996996997e-07,
      "loss": 0.29,
      "step": 30970
    },
    {
      "epoch": 93.03,
      "grad_norm": 3.2840986251831055,
      "learning_rate": 6.966966966966968e-07,
      "loss": 0.2707,
      "step": 30980
    },
    {
      "epoch": 93.06,
      "grad_norm": 6.7376203536987305,
      "learning_rate": 6.936936936936937e-07,
      "loss": 0.3111,
      "step": 30990
    },
    {
      "epoch": 93.09,
      "grad_norm": 6.965914249420166,
      "learning_rate": 6.906906906906907e-07,
      "loss": 0.3857,
      "step": 31000
    },
    {
      "epoch": 93.12,
      "grad_norm": 2.9306857585906982,
      "learning_rate": 6.876876876876877e-07,
      "loss": 0.3607,
      "step": 31010
    },
    {
      "epoch": 93.15,
      "grad_norm": 3.026684045791626,
      "learning_rate": 6.846846846846847e-07,
      "loss": 0.3142,
      "step": 31020
    },
    {
      "epoch": 93.18,
      "grad_norm": 4.302062034606934,
      "learning_rate": 6.816816816816819e-07,
      "loss": 0.3442,
      "step": 31030
    },
    {
      "epoch": 93.21,
      "grad_norm": 4.217340469360352,
      "learning_rate": 6.786786786786788e-07,
      "loss": 0.3407,
      "step": 31040
    },
    {
      "epoch": 93.24,
      "grad_norm": 3.9915504455566406,
      "learning_rate": 6.756756756756758e-07,
      "loss": 0.3386,
      "step": 31050
    },
    {
      "epoch": 93.27,
      "grad_norm": 8.882112503051758,
      "learning_rate": 6.726726726726727e-07,
      "loss": 0.3418,
      "step": 31060
    },
    {
      "epoch": 93.3,
      "grad_norm": 5.864893913269043,
      "learning_rate": 6.696696696696698e-07,
      "loss": 0.3082,
      "step": 31070
    },
    {
      "epoch": 93.33,
      "grad_norm": 8.329072952270508,
      "learning_rate": 6.666666666666667e-07,
      "loss": 0.3414,
      "step": 31080
    },
    {
      "epoch": 93.36,
      "grad_norm": 6.774460792541504,
      "learning_rate": 6.636636636636637e-07,
      "loss": 0.3326,
      "step": 31090
    },
    {
      "epoch": 93.39,
      "grad_norm": 4.039747714996338,
      "learning_rate": 6.606606606606607e-07,
      "loss": 0.3431,
      "step": 31100
    },
    {
      "epoch": 93.42,
      "grad_norm": 4.581603527069092,
      "learning_rate": 6.576576576576577e-07,
      "loss": 0.3867,
      "step": 31110
    },
    {
      "epoch": 93.45,
      "grad_norm": 3.2808804512023926,
      "learning_rate": 6.546546546546548e-07,
      "loss": 0.3419,
      "step": 31120
    },
    {
      "epoch": 93.48,
      "grad_norm": 6.899603843688965,
      "learning_rate": 6.516516516516518e-07,
      "loss": 0.3569,
      "step": 31130
    },
    {
      "epoch": 93.51,
      "grad_norm": 4.4388885498046875,
      "learning_rate": 6.486486486486487e-07,
      "loss": 0.3402,
      "step": 31140
    },
    {
      "epoch": 93.54,
      "grad_norm": 4.982517242431641,
      "learning_rate": 6.456456456456457e-07,
      "loss": 0.384,
      "step": 31150
    },
    {
      "epoch": 93.57,
      "grad_norm": 4.4137654304504395,
      "learning_rate": 6.426426426426427e-07,
      "loss": 0.4024,
      "step": 31160
    },
    {
      "epoch": 93.6,
      "grad_norm": 7.69840669631958,
      "learning_rate": 6.396396396396397e-07,
      "loss": 0.3313,
      "step": 31170
    },
    {
      "epoch": 93.63,
      "grad_norm": 14.611039161682129,
      "learning_rate": 6.366366366366366e-07,
      "loss": 0.3637,
      "step": 31180
    },
    {
      "epoch": 93.66,
      "grad_norm": 4.412993907928467,
      "learning_rate": 6.336336336336337e-07,
      "loss": 0.379,
      "step": 31190
    },
    {
      "epoch": 93.69,
      "grad_norm": 4.273768901824951,
      "learning_rate": 6.306306306306306e-07,
      "loss": 0.3573,
      "step": 31200
    },
    {
      "epoch": 93.72,
      "grad_norm": 4.036172866821289,
      "learning_rate": 6.276276276276278e-07,
      "loss": 0.315,
      "step": 31210
    },
    {
      "epoch": 93.75,
      "grad_norm": 3.7316455841064453,
      "learning_rate": 6.246246246246246e-07,
      "loss": 0.3862,
      "step": 31220
    },
    {
      "epoch": 93.78,
      "grad_norm": 4.846069812774658,
      "learning_rate": 6.216216216216217e-07,
      "loss": 0.3672,
      "step": 31230
    },
    {
      "epoch": 93.81,
      "grad_norm": 5.222692966461182,
      "learning_rate": 6.186186186186187e-07,
      "loss": 0.3353,
      "step": 31240
    },
    {
      "epoch": 93.84,
      "grad_norm": 3.3685638904571533,
      "learning_rate": 6.156156156156157e-07,
      "loss": 0.3875,
      "step": 31250
    },
    {
      "epoch": 93.87,
      "grad_norm": 3.1866581439971924,
      "learning_rate": 6.126126126126126e-07,
      "loss": 0.3081,
      "step": 31260
    },
    {
      "epoch": 93.9,
      "grad_norm": 5.554765224456787,
      "learning_rate": 6.096096096096096e-07,
      "loss": 0.3279,
      "step": 31270
    },
    {
      "epoch": 93.93,
      "grad_norm": 5.637399673461914,
      "learning_rate": 6.066066066066066e-07,
      "loss": 0.3523,
      "step": 31280
    },
    {
      "epoch": 93.96,
      "grad_norm": 4.683924674987793,
      "learning_rate": 6.036036036036037e-07,
      "loss": 0.346,
      "step": 31290
    },
    {
      "epoch": 93.99,
      "grad_norm": 3.922468423843384,
      "learning_rate": 6.006006006006006e-07,
      "loss": 0.2953,
      "step": 31300
    },
    {
      "epoch": 94.0,
      "eval_accuracy": 0.9203,
      "eval_loss": 0.31257715821266174,
      "eval_runtime": 26.0542,
      "eval_samples_per_second": 383.815,
      "eval_steps_per_second": 1.535,
      "step": 31302
    },
    {
      "epoch": 94.02,
      "grad_norm": 4.449802398681641,
      "learning_rate": 5.975975975975976e-07,
      "loss": 0.3186,
      "step": 31310
    },
    {
      "epoch": 94.05,
      "grad_norm": 6.00807523727417,
      "learning_rate": 5.945945945945947e-07,
      "loss": 0.362,
      "step": 31320
    },
    {
      "epoch": 94.08,
      "grad_norm": 5.355147838592529,
      "learning_rate": 5.915915915915917e-07,
      "loss": 0.3827,
      "step": 31330
    },
    {
      "epoch": 94.11,
      "grad_norm": 4.773733139038086,
      "learning_rate": 5.885885885885887e-07,
      "loss": 0.3939,
      "step": 31340
    },
    {
      "epoch": 94.14,
      "grad_norm": 3.9060473442077637,
      "learning_rate": 5.855855855855856e-07,
      "loss": 0.3306,
      "step": 31350
    },
    {
      "epoch": 94.17,
      "grad_norm": 5.612941265106201,
      "learning_rate": 5.825825825825826e-07,
      "loss": 0.3102,
      "step": 31360
    },
    {
      "epoch": 94.2,
      "grad_norm": 4.536509037017822,
      "learning_rate": 5.795795795795796e-07,
      "loss": 0.3467,
      "step": 31370
    },
    {
      "epoch": 94.23,
      "grad_norm": 5.103543758392334,
      "learning_rate": 5.765765765765767e-07,
      "loss": 0.3405,
      "step": 31380
    },
    {
      "epoch": 94.26,
      "grad_norm": 3.99548602104187,
      "learning_rate": 5.735735735735736e-07,
      "loss": 0.3631,
      "step": 31390
    },
    {
      "epoch": 94.29,
      "grad_norm": 5.5115647315979,
      "learning_rate": 5.705705705705706e-07,
      "loss": 0.377,
      "step": 31400
    },
    {
      "epoch": 94.32,
      "grad_norm": 3.943998336791992,
      "learning_rate": 5.675675675675676e-07,
      "loss": 0.3507,
      "step": 31410
    },
    {
      "epoch": 94.35,
      "grad_norm": 5.907485008239746,
      "learning_rate": 5.645645645645647e-07,
      "loss": 0.3489,
      "step": 31420
    },
    {
      "epoch": 94.38,
      "grad_norm": 7.040783405303955,
      "learning_rate": 5.615615615615616e-07,
      "loss": 0.3312,
      "step": 31430
    },
    {
      "epoch": 94.41,
      "grad_norm": 4.399087905883789,
      "learning_rate": 5.585585585585586e-07,
      "loss": 0.3311,
      "step": 31440
    },
    {
      "epoch": 94.44,
      "grad_norm": 3.58413028717041,
      "learning_rate": 5.555555555555555e-07,
      "loss": 0.3988,
      "step": 31450
    },
    {
      "epoch": 94.47,
      "grad_norm": 15.95844841003418,
      "learning_rate": 5.525525525525526e-07,
      "loss": 0.3048,
      "step": 31460
    },
    {
      "epoch": 94.5,
      "grad_norm": 4.285422325134277,
      "learning_rate": 5.495495495495496e-07,
      "loss": 0.3826,
      "step": 31470
    },
    {
      "epoch": 94.53,
      "grad_norm": 3.7653510570526123,
      "learning_rate": 5.465465465465466e-07,
      "loss": 0.3802,
      "step": 31480
    },
    {
      "epoch": 94.56,
      "grad_norm": 3.588212728500366,
      "learning_rate": 5.435435435435435e-07,
      "loss": 0.3244,
      "step": 31490
    },
    {
      "epoch": 94.59,
      "grad_norm": 3.9439704418182373,
      "learning_rate": 5.405405405405406e-07,
      "loss": 0.3804,
      "step": 31500
    },
    {
      "epoch": 94.62,
      "grad_norm": 4.174720764160156,
      "learning_rate": 5.375375375375376e-07,
      "loss": 0.3476,
      "step": 31510
    },
    {
      "epoch": 94.65,
      "grad_norm": 3.3163976669311523,
      "learning_rate": 5.345345345345346e-07,
      "loss": 0.3458,
      "step": 31520
    },
    {
      "epoch": 94.68,
      "grad_norm": 3.437070608139038,
      "learning_rate": 5.315315315315315e-07,
      "loss": 0.2951,
      "step": 31530
    },
    {
      "epoch": 94.71,
      "grad_norm": 4.96079158782959,
      "learning_rate": 5.285285285285286e-07,
      "loss": 0.3308,
      "step": 31540
    },
    {
      "epoch": 94.74,
      "grad_norm": 3.980663776397705,
      "learning_rate": 5.255255255255255e-07,
      "loss": 0.3687,
      "step": 31550
    },
    {
      "epoch": 94.77,
      "grad_norm": 2.85925555229187,
      "learning_rate": 5.225225225225226e-07,
      "loss": 0.268,
      "step": 31560
    },
    {
      "epoch": 94.8,
      "grad_norm": 3.6773390769958496,
      "learning_rate": 5.195195195195195e-07,
      "loss": 0.322,
      "step": 31570
    },
    {
      "epoch": 94.83,
      "grad_norm": 3.816819429397583,
      "learning_rate": 5.165165165165165e-07,
      "loss": 0.3393,
      "step": 31580
    },
    {
      "epoch": 94.86,
      "grad_norm": 4.823004245758057,
      "learning_rate": 5.135135135135135e-07,
      "loss": 0.3802,
      "step": 31590
    },
    {
      "epoch": 94.89,
      "grad_norm": 4.860716819763184,
      "learning_rate": 5.105105105105106e-07,
      "loss": 0.3741,
      "step": 31600
    },
    {
      "epoch": 94.92,
      "grad_norm": 4.874477863311768,
      "learning_rate": 5.075075075075076e-07,
      "loss": 0.3741,
      "step": 31610
    },
    {
      "epoch": 94.95,
      "grad_norm": 3.4423089027404785,
      "learning_rate": 5.045045045045045e-07,
      "loss": 0.3578,
      "step": 31620
    },
    {
      "epoch": 94.98,
      "grad_norm": 5.386654376983643,
      "learning_rate": 5.015015015015016e-07,
      "loss": 0.3475,
      "step": 31630
    },
    {
      "epoch": 95.0,
      "eval_accuracy": 0.9206,
      "eval_loss": 0.3140873908996582,
      "eval_runtime": 25.8325,
      "eval_samples_per_second": 387.11,
      "eval_steps_per_second": 1.548,
      "step": 31635
    },
    {
      "epoch": 95.02,
      "grad_norm": 5.397531032562256,
      "learning_rate": 4.984984984984986e-07,
      "loss": 0.3081,
      "step": 31640
    },
    {
      "epoch": 95.05,
      "grad_norm": 4.5049824714660645,
      "learning_rate": 4.954954954954956e-07,
      "loss": 0.2944,
      "step": 31650
    },
    {
      "epoch": 95.08,
      "grad_norm": 5.925257205963135,
      "learning_rate": 4.924924924924925e-07,
      "loss": 0.3757,
      "step": 31660
    },
    {
      "epoch": 95.11,
      "grad_norm": 6.100909233093262,
      "learning_rate": 4.894894894894895e-07,
      "loss": 0.3221,
      "step": 31670
    },
    {
      "epoch": 95.14,
      "grad_norm": 6.157273769378662,
      "learning_rate": 4.864864864864865e-07,
      "loss": 0.3474,
      "step": 31680
    },
    {
      "epoch": 95.17,
      "grad_norm": 2.942246437072754,
      "learning_rate": 4.834834834834836e-07,
      "loss": 0.356,
      "step": 31690
    },
    {
      "epoch": 95.2,
      "grad_norm": 4.350193023681641,
      "learning_rate": 4.804804804804805e-07,
      "loss": 0.375,
      "step": 31700
    },
    {
      "epoch": 95.23,
      "grad_norm": 5.4187703132629395,
      "learning_rate": 4.774774774774775e-07,
      "loss": 0.3991,
      "step": 31710
    },
    {
      "epoch": 95.26,
      "grad_norm": 3.469822406768799,
      "learning_rate": 4.744744744744745e-07,
      "loss": 0.324,
      "step": 31720
    },
    {
      "epoch": 95.29,
      "grad_norm": 14.595053672790527,
      "learning_rate": 4.714714714714715e-07,
      "loss": 0.4067,
      "step": 31730
    },
    {
      "epoch": 95.32,
      "grad_norm": 4.813581466674805,
      "learning_rate": 4.684684684684685e-07,
      "loss": 0.3692,
      "step": 31740
    },
    {
      "epoch": 95.35,
      "grad_norm": 3.934116840362549,
      "learning_rate": 4.654654654654655e-07,
      "loss": 0.3081,
      "step": 31750
    },
    {
      "epoch": 95.38,
      "grad_norm": 3.947645902633667,
      "learning_rate": 4.624624624624625e-07,
      "loss": 0.3679,
      "step": 31760
    },
    {
      "epoch": 95.41,
      "grad_norm": 3.431896924972534,
      "learning_rate": 4.5945945945945953e-07,
      "loss": 0.3021,
      "step": 31770
    },
    {
      "epoch": 95.44,
      "grad_norm": 3.829011917114258,
      "learning_rate": 4.564564564564565e-07,
      "loss": 0.3665,
      "step": 31780
    },
    {
      "epoch": 95.47,
      "grad_norm": 5.203980922698975,
      "learning_rate": 4.534534534534535e-07,
      "loss": 0.3062,
      "step": 31790
    },
    {
      "epoch": 95.5,
      "grad_norm": 5.787779808044434,
      "learning_rate": 4.504504504504505e-07,
      "loss": 0.3318,
      "step": 31800
    },
    {
      "epoch": 95.53,
      "grad_norm": 4.036657333374023,
      "learning_rate": 4.4744744744744754e-07,
      "loss": 0.3812,
      "step": 31810
    },
    {
      "epoch": 95.56,
      "grad_norm": 3.6541788578033447,
      "learning_rate": 4.444444444444445e-07,
      "loss": 0.329,
      "step": 31820
    },
    {
      "epoch": 95.59,
      "grad_norm": 4.102051734924316,
      "learning_rate": 4.414414414414415e-07,
      "loss": 0.3426,
      "step": 31830
    },
    {
      "epoch": 95.62,
      "grad_norm": 5.160118579864502,
      "learning_rate": 4.3843843843843844e-07,
      "loss": 0.3566,
      "step": 31840
    },
    {
      "epoch": 95.65,
      "grad_norm": 3.3825466632843018,
      "learning_rate": 4.3543543543543544e-07,
      "loss": 0.311,
      "step": 31850
    },
    {
      "epoch": 95.68,
      "grad_norm": 4.075983047485352,
      "learning_rate": 4.324324324324325e-07,
      "loss": 0.3626,
      "step": 31860
    },
    {
      "epoch": 95.71,
      "grad_norm": 4.549976348876953,
      "learning_rate": 4.294294294294295e-07,
      "loss": 0.4073,
      "step": 31870
    },
    {
      "epoch": 95.74,
      "grad_norm": 4.406117916107178,
      "learning_rate": 4.2642642642642645e-07,
      "loss": 0.3075,
      "step": 31880
    },
    {
      "epoch": 95.77,
      "grad_norm": 3.083282709121704,
      "learning_rate": 4.2342342342342345e-07,
      "loss": 0.2793,
      "step": 31890
    },
    {
      "epoch": 95.8,
      "grad_norm": 4.250041961669922,
      "learning_rate": 4.204204204204205e-07,
      "loss": 0.3139,
      "step": 31900
    },
    {
      "epoch": 95.83,
      "grad_norm": 4.605595588684082,
      "learning_rate": 4.1741741741741745e-07,
      "loss": 0.3263,
      "step": 31910
    },
    {
      "epoch": 95.86,
      "grad_norm": 3.9657533168792725,
      "learning_rate": 4.1441441441441446e-07,
      "loss": 0.3575,
      "step": 31920
    },
    {
      "epoch": 95.89,
      "grad_norm": 3.7416341304779053,
      "learning_rate": 4.114114114114114e-07,
      "loss": 0.4011,
      "step": 31930
    },
    {
      "epoch": 95.92,
      "grad_norm": 3.5741846561431885,
      "learning_rate": 4.084084084084084e-07,
      "loss": 0.3258,
      "step": 31940
    },
    {
      "epoch": 95.95,
      "grad_norm": 4.933104038238525,
      "learning_rate": 4.0540540540540546e-07,
      "loss": 0.3196,
      "step": 31950
    },
    {
      "epoch": 95.98,
      "grad_norm": 3.783827781677246,
      "learning_rate": 4.0240240240240246e-07,
      "loss": 0.3406,
      "step": 31960
    },
    {
      "epoch": 96.0,
      "eval_accuracy": 0.9197,
      "eval_loss": 0.31409189105033875,
      "eval_runtime": 25.7986,
      "eval_samples_per_second": 387.617,
      "eval_steps_per_second": 1.55,
      "step": 31968
    },
    {
      "epoch": 96.01,
      "grad_norm": 3.8258960247039795,
      "learning_rate": 3.993993993993994e-07,
      "loss": 0.2989,
      "step": 31970
    },
    {
      "epoch": 96.04,
      "grad_norm": 8.178354263305664,
      "learning_rate": 3.963963963963964e-07,
      "loss": 0.3041,
      "step": 31980
    },
    {
      "epoch": 96.07,
      "grad_norm": 5.1330885887146,
      "learning_rate": 3.9339339339339347e-07,
      "loss": 0.3302,
      "step": 31990
    },
    {
      "epoch": 96.1,
      "grad_norm": 4.554864883422852,
      "learning_rate": 3.903903903903904e-07,
      "loss": 0.3139,
      "step": 32000
    },
    {
      "epoch": 96.13,
      "grad_norm": 5.574049472808838,
      "learning_rate": 3.873873873873874e-07,
      "loss": 0.3069,
      "step": 32010
    },
    {
      "epoch": 96.16,
      "grad_norm": 5.128723621368408,
      "learning_rate": 3.843843843843844e-07,
      "loss": 0.3264,
      "step": 32020
    },
    {
      "epoch": 96.19,
      "grad_norm": 4.984865188598633,
      "learning_rate": 3.8138138138138137e-07,
      "loss": 0.3636,
      "step": 32030
    },
    {
      "epoch": 96.22,
      "grad_norm": 7.759979724884033,
      "learning_rate": 3.7837837837837843e-07,
      "loss": 0.378,
      "step": 32040
    },
    {
      "epoch": 96.25,
      "grad_norm": 7.088198661804199,
      "learning_rate": 3.7537537537537543e-07,
      "loss": 0.3416,
      "step": 32050
    },
    {
      "epoch": 96.28,
      "grad_norm": 3.981818437576294,
      "learning_rate": 3.723723723723724e-07,
      "loss": 0.3049,
      "step": 32060
    },
    {
      "epoch": 96.31,
      "grad_norm": 3.994428873062134,
      "learning_rate": 3.693693693693694e-07,
      "loss": 0.3521,
      "step": 32070
    },
    {
      "epoch": 96.34,
      "grad_norm": 5.144979953765869,
      "learning_rate": 3.6636636636636644e-07,
      "loss": 0.3877,
      "step": 32080
    },
    {
      "epoch": 96.37,
      "grad_norm": 5.185286998748779,
      "learning_rate": 3.633633633633634e-07,
      "loss": 0.3951,
      "step": 32090
    },
    {
      "epoch": 96.4,
      "grad_norm": 3.5956671237945557,
      "learning_rate": 3.603603603603604e-07,
      "loss": 0.3168,
      "step": 32100
    },
    {
      "epoch": 96.43,
      "grad_norm": 3.5111818313598633,
      "learning_rate": 3.573573573573574e-07,
      "loss": 0.3339,
      "step": 32110
    },
    {
      "epoch": 96.46,
      "grad_norm": 4.944345474243164,
      "learning_rate": 3.5435435435435434e-07,
      "loss": 0.3712,
      "step": 32120
    },
    {
      "epoch": 96.49,
      "grad_norm": 3.8497438430786133,
      "learning_rate": 3.513513513513514e-07,
      "loss": 0.3425,
      "step": 32130
    },
    {
      "epoch": 96.52,
      "grad_norm": 4.525459289550781,
      "learning_rate": 3.483483483483484e-07,
      "loss": 0.3504,
      "step": 32140
    },
    {
      "epoch": 96.55,
      "grad_norm": 4.143781661987305,
      "learning_rate": 3.4534534534534535e-07,
      "loss": 0.3236,
      "step": 32150
    },
    {
      "epoch": 96.58,
      "grad_norm": 7.533895015716553,
      "learning_rate": 3.4234234234234235e-07,
      "loss": 0.3122,
      "step": 32160
    },
    {
      "epoch": 96.61,
      "grad_norm": 5.955483436584473,
      "learning_rate": 3.393393393393394e-07,
      "loss": 0.3859,
      "step": 32170
    },
    {
      "epoch": 96.64,
      "grad_norm": 4.334192752838135,
      "learning_rate": 3.3633633633633635e-07,
      "loss": 0.3599,
      "step": 32180
    },
    {
      "epoch": 96.67,
      "grad_norm": 4.097786903381348,
      "learning_rate": 3.3333333333333335e-07,
      "loss": 0.3478,
      "step": 32190
    },
    {
      "epoch": 96.7,
      "grad_norm": 4.746017932891846,
      "learning_rate": 3.3033033033033036e-07,
      "loss": 0.3343,
      "step": 32200
    },
    {
      "epoch": 96.73,
      "grad_norm": 3.965172052383423,
      "learning_rate": 3.273273273273274e-07,
      "loss": 0.3919,
      "step": 32210
    },
    {
      "epoch": 96.76,
      "grad_norm": 3.737344980239868,
      "learning_rate": 3.2432432432432436e-07,
      "loss": 0.3371,
      "step": 32220
    },
    {
      "epoch": 96.79,
      "grad_norm": 4.010631084442139,
      "learning_rate": 3.2132132132132136e-07,
      "loss": 0.3452,
      "step": 32230
    },
    {
      "epoch": 96.82,
      "grad_norm": 5.108850955963135,
      "learning_rate": 3.183183183183183e-07,
      "loss": 0.3384,
      "step": 32240
    },
    {
      "epoch": 96.85,
      "grad_norm": 4.691813945770264,
      "learning_rate": 3.153153153153153e-07,
      "loss": 0.3504,
      "step": 32250
    },
    {
      "epoch": 96.88,
      "grad_norm": 5.714572906494141,
      "learning_rate": 3.123123123123123e-07,
      "loss": 0.3388,
      "step": 32260
    },
    {
      "epoch": 96.91,
      "grad_norm": 3.284459114074707,
      "learning_rate": 3.0930930930930937e-07,
      "loss": 0.2918,
      "step": 32270
    },
    {
      "epoch": 96.94,
      "grad_norm": 3.6938838958740234,
      "learning_rate": 3.063063063063063e-07,
      "loss": 0.3261,
      "step": 32280
    },
    {
      "epoch": 96.97,
      "grad_norm": 5.403067111968994,
      "learning_rate": 3.033033033033033e-07,
      "loss": 0.3713,
      "step": 32290
    },
    {
      "epoch": 97.0,
      "grad_norm": 4.652668476104736,
      "learning_rate": 3.003003003003003e-07,
      "loss": 0.3448,
      "step": 32300
    },
    {
      "epoch": 97.0,
      "eval_accuracy": 0.9198,
      "eval_loss": 0.314133882522583,
      "eval_runtime": 26.1736,
      "eval_samples_per_second": 382.064,
      "eval_steps_per_second": 1.528,
      "step": 32301
    },
    {
      "epoch": 97.03,
      "grad_norm": 5.354833126068115,
      "learning_rate": 2.972972972972973e-07,
      "loss": 0.3472,
      "step": 32310
    },
    {
      "epoch": 97.06,
      "grad_norm": 4.751397609710693,
      "learning_rate": 2.9429429429429433e-07,
      "loss": 0.3253,
      "step": 32320
    },
    {
      "epoch": 97.09,
      "grad_norm": 6.813636779785156,
      "learning_rate": 2.912912912912913e-07,
      "loss": 0.356,
      "step": 32330
    },
    {
      "epoch": 97.12,
      "grad_norm": 4.35203218460083,
      "learning_rate": 2.8828828828828833e-07,
      "loss": 0.358,
      "step": 32340
    },
    {
      "epoch": 97.15,
      "grad_norm": 4.264711856842041,
      "learning_rate": 2.852852852852853e-07,
      "loss": 0.2937,
      "step": 32350
    },
    {
      "epoch": 97.18,
      "grad_norm": 5.273285388946533,
      "learning_rate": 2.8228228228228234e-07,
      "loss": 0.357,
      "step": 32360
    },
    {
      "epoch": 97.21,
      "grad_norm": 5.5505242347717285,
      "learning_rate": 2.792792792792793e-07,
      "loss": 0.3125,
      "step": 32370
    },
    {
      "epoch": 97.24,
      "grad_norm": 5.1180524826049805,
      "learning_rate": 2.762762762762763e-07,
      "loss": 0.3858,
      "step": 32380
    },
    {
      "epoch": 97.27,
      "grad_norm": 5.2147626876831055,
      "learning_rate": 2.732732732732733e-07,
      "loss": 0.347,
      "step": 32390
    },
    {
      "epoch": 97.3,
      "grad_norm": 5.263698101043701,
      "learning_rate": 2.702702702702703e-07,
      "loss": 0.3257,
      "step": 32400
    },
    {
      "epoch": 97.33,
      "grad_norm": 2.5007359981536865,
      "learning_rate": 2.672672672672673e-07,
      "loss": 0.3869,
      "step": 32410
    },
    {
      "epoch": 97.36,
      "grad_norm": 5.014791011810303,
      "learning_rate": 2.642642642642643e-07,
      "loss": 0.3397,
      "step": 32420
    },
    {
      "epoch": 97.39,
      "grad_norm": 3.0805418491363525,
      "learning_rate": 2.612612612612613e-07,
      "loss": 0.3024,
      "step": 32430
    },
    {
      "epoch": 97.42,
      "grad_norm": 4.988929271697998,
      "learning_rate": 2.5825825825825825e-07,
      "loss": 0.3576,
      "step": 32440
    },
    {
      "epoch": 97.45,
      "grad_norm": 3.6334080696105957,
      "learning_rate": 2.552552552552553e-07,
      "loss": 0.3885,
      "step": 32450
    },
    {
      "epoch": 97.48,
      "grad_norm": 3.4304449558258057,
      "learning_rate": 2.5225225225225225e-07,
      "loss": 0.3318,
      "step": 32460
    },
    {
      "epoch": 97.51,
      "grad_norm": 3.6762731075286865,
      "learning_rate": 2.492492492492493e-07,
      "loss": 0.3161,
      "step": 32470
    },
    {
      "epoch": 97.54,
      "grad_norm": 4.178286075592041,
      "learning_rate": 2.4624624624624625e-07,
      "loss": 0.3273,
      "step": 32480
    },
    {
      "epoch": 97.57,
      "grad_norm": 4.572354793548584,
      "learning_rate": 2.4324324324324326e-07,
      "loss": 0.3249,
      "step": 32490
    },
    {
      "epoch": 97.6,
      "grad_norm": 4.336079120635986,
      "learning_rate": 2.4024024024024026e-07,
      "loss": 0.3273,
      "step": 32500
    },
    {
      "epoch": 97.63,
      "grad_norm": 5.075711250305176,
      "learning_rate": 2.3723723723723726e-07,
      "loss": 0.3533,
      "step": 32510
    },
    {
      "epoch": 97.66,
      "grad_norm": 4.588191509246826,
      "learning_rate": 2.3423423423423426e-07,
      "loss": 0.3055,
      "step": 32520
    },
    {
      "epoch": 97.69,
      "grad_norm": 5.121078014373779,
      "learning_rate": 2.3123123123123124e-07,
      "loss": 0.3095,
      "step": 32530
    },
    {
      "epoch": 97.72,
      "grad_norm": 5.116728782653809,
      "learning_rate": 2.2822822822822824e-07,
      "loss": 0.3319,
      "step": 32540
    },
    {
      "epoch": 97.75,
      "grad_norm": 4.854680061340332,
      "learning_rate": 2.2522522522522524e-07,
      "loss": 0.3073,
      "step": 32550
    },
    {
      "epoch": 97.78,
      "grad_norm": 3.0708327293395996,
      "learning_rate": 2.2222222222222224e-07,
      "loss": 0.3133,
      "step": 32560
    },
    {
      "epoch": 97.81,
      "grad_norm": 4.557015895843506,
      "learning_rate": 2.1921921921921922e-07,
      "loss": 0.3733,
      "step": 32570
    },
    {
      "epoch": 97.84,
      "grad_norm": 5.618918418884277,
      "learning_rate": 2.1621621621621625e-07,
      "loss": 0.2956,
      "step": 32580
    },
    {
      "epoch": 97.87,
      "grad_norm": 4.315530776977539,
      "learning_rate": 2.1321321321321322e-07,
      "loss": 0.3163,
      "step": 32590
    },
    {
      "epoch": 97.9,
      "grad_norm": 3.887568473815918,
      "learning_rate": 2.1021021021021025e-07,
      "loss": 0.3162,
      "step": 32600
    },
    {
      "epoch": 97.93,
      "grad_norm": 5.347872257232666,
      "learning_rate": 2.0720720720720723e-07,
      "loss": 0.3691,
      "step": 32610
    },
    {
      "epoch": 97.96,
      "grad_norm": 5.586434841156006,
      "learning_rate": 2.042042042042042e-07,
      "loss": 0.365,
      "step": 32620
    },
    {
      "epoch": 97.99,
      "grad_norm": 3.6295406818389893,
      "learning_rate": 2.0120120120120123e-07,
      "loss": 0.3687,
      "step": 32630
    },
    {
      "epoch": 98.0,
      "eval_accuracy": 0.9205,
      "eval_loss": 0.31366539001464844,
      "eval_runtime": 26.9429,
      "eval_samples_per_second": 371.155,
      "eval_steps_per_second": 1.485,
      "step": 32634
    },
    {
      "epoch": 98.02,
      "grad_norm": 3.871833324432373,
      "learning_rate": 1.981981981981982e-07,
      "loss": 0.4683,
      "step": 32640
    },
    {
      "epoch": 98.05,
      "grad_norm": 3.0988991260528564,
      "learning_rate": 1.951951951951952e-07,
      "loss": 0.3901,
      "step": 32650
    },
    {
      "epoch": 98.08,
      "grad_norm": 4.379641056060791,
      "learning_rate": 1.921921921921922e-07,
      "loss": 0.3258,
      "step": 32660
    },
    {
      "epoch": 98.11,
      "grad_norm": 3.5822346210479736,
      "learning_rate": 1.8918918918918921e-07,
      "loss": 0.3026,
      "step": 32670
    },
    {
      "epoch": 98.14,
      "grad_norm": 6.377861022949219,
      "learning_rate": 1.861861861861862e-07,
      "loss": 0.4123,
      "step": 32680
    },
    {
      "epoch": 98.17,
      "grad_norm": 3.2342004776000977,
      "learning_rate": 1.8318318318318322e-07,
      "loss": 0.3574,
      "step": 32690
    },
    {
      "epoch": 98.2,
      "grad_norm": 4.315357208251953,
      "learning_rate": 1.801801801801802e-07,
      "loss": 0.322,
      "step": 32700
    },
    {
      "epoch": 98.23,
      "grad_norm": 3.566986560821533,
      "learning_rate": 1.7717717717717717e-07,
      "loss": 0.3391,
      "step": 32710
    },
    {
      "epoch": 98.26,
      "grad_norm": 4.169888019561768,
      "learning_rate": 1.741741741741742e-07,
      "loss": 0.3826,
      "step": 32720
    },
    {
      "epoch": 98.29,
      "grad_norm": 4.574073314666748,
      "learning_rate": 1.7117117117117117e-07,
      "loss": 0.3454,
      "step": 32730
    },
    {
      "epoch": 98.32,
      "grad_norm": 4.495599746704102,
      "learning_rate": 1.6816816816816818e-07,
      "loss": 0.3298,
      "step": 32740
    },
    {
      "epoch": 98.35,
      "grad_norm": 4.073917865753174,
      "learning_rate": 1.6516516516516518e-07,
      "loss": 0.3162,
      "step": 32750
    },
    {
      "epoch": 98.38,
      "grad_norm": 5.600374221801758,
      "learning_rate": 1.6216216216216218e-07,
      "loss": 0.3543,
      "step": 32760
    },
    {
      "epoch": 98.41,
      "grad_norm": 5.009756088256836,
      "learning_rate": 1.5915915915915916e-07,
      "loss": 0.3374,
      "step": 32770
    },
    {
      "epoch": 98.44,
      "grad_norm": 5.322586536407471,
      "learning_rate": 1.5615615615615616e-07,
      "loss": 0.3099,
      "step": 32780
    },
    {
      "epoch": 98.47,
      "grad_norm": 3.9506351947784424,
      "learning_rate": 1.5315315315315316e-07,
      "loss": 0.395,
      "step": 32790
    },
    {
      "epoch": 98.5,
      "grad_norm": 4.347846984863281,
      "learning_rate": 1.5015015015015016e-07,
      "loss": 0.3326,
      "step": 32800
    },
    {
      "epoch": 98.53,
      "grad_norm": 4.670065402984619,
      "learning_rate": 1.4714714714714716e-07,
      "loss": 0.3697,
      "step": 32810
    },
    {
      "epoch": 98.56,
      "grad_norm": 3.7971980571746826,
      "learning_rate": 1.4414414414414417e-07,
      "loss": 0.3176,
      "step": 32820
    },
    {
      "epoch": 98.59,
      "grad_norm": 2.8337886333465576,
      "learning_rate": 1.4114114114114117e-07,
      "loss": 0.3366,
      "step": 32830
    },
    {
      "epoch": 98.62,
      "grad_norm": 4.8089423179626465,
      "learning_rate": 1.3813813813813814e-07,
      "loss": 0.3562,
      "step": 32840
    },
    {
      "epoch": 98.65,
      "grad_norm": 3.430757999420166,
      "learning_rate": 1.3513513513513515e-07,
      "loss": 0.3087,
      "step": 32850
    },
    {
      "epoch": 98.68,
      "grad_norm": 4.315730094909668,
      "learning_rate": 1.3213213213213215e-07,
      "loss": 0.3713,
      "step": 32860
    },
    {
      "epoch": 98.71,
      "grad_norm": 6.535925388336182,
      "learning_rate": 1.2912912912912912e-07,
      "loss": 0.3145,
      "step": 32870
    },
    {
      "epoch": 98.74,
      "grad_norm": 4.162943363189697,
      "learning_rate": 1.2612612612612613e-07,
      "loss": 0.315,
      "step": 32880
    },
    {
      "epoch": 98.77,
      "grad_norm": 7.173022270202637,
      "learning_rate": 1.2312312312312313e-07,
      "loss": 0.3676,
      "step": 32890
    },
    {
      "epoch": 98.8,
      "grad_norm": 4.965452194213867,
      "learning_rate": 1.2012012012012013e-07,
      "loss": 0.3381,
      "step": 32900
    },
    {
      "epoch": 98.83,
      "grad_norm": 3.7585604190826416,
      "learning_rate": 1.1711711711711713e-07,
      "loss": 0.3771,
      "step": 32910
    },
    {
      "epoch": 98.86,
      "grad_norm": 3.651920795440674,
      "learning_rate": 1.1411411411411412e-07,
      "loss": 0.4056,
      "step": 32920
    },
    {
      "epoch": 98.89,
      "grad_norm": 4.883760452270508,
      "learning_rate": 1.1111111111111112e-07,
      "loss": 0.3325,
      "step": 32930
    },
    {
      "epoch": 98.92,
      "grad_norm": 3.536863327026367,
      "learning_rate": 1.0810810810810812e-07,
      "loss": 0.3775,
      "step": 32940
    },
    {
      "epoch": 98.95,
      "grad_norm": 4.0887837409973145,
      "learning_rate": 1.0510510510510513e-07,
      "loss": 0.3043,
      "step": 32950
    },
    {
      "epoch": 98.98,
      "grad_norm": 3.786797523498535,
      "learning_rate": 1.021021021021021e-07,
      "loss": 0.345,
      "step": 32960
    },
    {
      "epoch": 99.0,
      "eval_accuracy": 0.92,
      "eval_loss": 0.31443700194358826,
      "eval_runtime": 25.8743,
      "eval_samples_per_second": 386.483,
      "eval_steps_per_second": 1.546,
      "step": 32967
    },
    {
      "epoch": 99.01,
      "grad_norm": 3.2166411876678467,
      "learning_rate": 9.90990990990991e-08,
      "loss": 0.4414,
      "step": 32970
    },
    {
      "epoch": 99.04,
      "grad_norm": 3.6225943565368652,
      "learning_rate": 9.60960960960961e-08,
      "loss": 0.3675,
      "step": 32980
    },
    {
      "epoch": 99.07,
      "grad_norm": 8.521991729736328,
      "learning_rate": 9.30930930930931e-08,
      "loss": 0.3332,
      "step": 32990
    },
    {
      "epoch": 99.1,
      "grad_norm": 3.9980201721191406,
      "learning_rate": 9.00900900900901e-08,
      "loss": 0.2749,
      "step": 33000
    },
    {
      "epoch": 99.13,
      "grad_norm": 5.557321071624756,
      "learning_rate": 8.70870870870871e-08,
      "loss": 0.3677,
      "step": 33010
    },
    {
      "epoch": 99.16,
      "grad_norm": 3.6367788314819336,
      "learning_rate": 8.408408408408409e-08,
      "loss": 0.3265,
      "step": 33020
    },
    {
      "epoch": 99.19,
      "grad_norm": 3.7935874462127686,
      "learning_rate": 8.108108108108109e-08,
      "loss": 0.3023,
      "step": 33030
    },
    {
      "epoch": 99.22,
      "grad_norm": 11.532512664794922,
      "learning_rate": 7.807807807807808e-08,
      "loss": 0.3354,
      "step": 33040
    },
    {
      "epoch": 99.25,
      "grad_norm": 4.633002758026123,
      "learning_rate": 7.507507507507508e-08,
      "loss": 0.3079,
      "step": 33050
    },
    {
      "epoch": 99.28,
      "grad_norm": 6.079354286193848,
      "learning_rate": 7.207207207207208e-08,
      "loss": 0.3659,
      "step": 33060
    },
    {
      "epoch": 99.31,
      "grad_norm": 6.3850789070129395,
      "learning_rate": 6.906906906906907e-08,
      "loss": 0.356,
      "step": 33070
    },
    {
      "epoch": 99.34,
      "grad_norm": 3.221412420272827,
      "learning_rate": 6.606606606606607e-08,
      "loss": 0.3129,
      "step": 33080
    },
    {
      "epoch": 99.37,
      "grad_norm": 6.328848361968994,
      "learning_rate": 6.306306306306306e-08,
      "loss": 0.3521,
      "step": 33090
    },
    {
      "epoch": 99.4,
      "grad_norm": 7.482673645019531,
      "learning_rate": 6.006006006006006e-08,
      "loss": 0.3746,
      "step": 33100
    },
    {
      "epoch": 99.43,
      "grad_norm": 2.8720099925994873,
      "learning_rate": 5.705705705705706e-08,
      "loss": 0.2891,
      "step": 33110
    },
    {
      "epoch": 99.46,
      "grad_norm": 6.796731948852539,
      "learning_rate": 5.405405405405406e-08,
      "loss": 0.3427,
      "step": 33120
    },
    {
      "epoch": 99.49,
      "grad_norm": 3.5392496585845947,
      "learning_rate": 5.105105105105105e-08,
      "loss": 0.2814,
      "step": 33130
    },
    {
      "epoch": 99.52,
      "grad_norm": 4.383900165557861,
      "learning_rate": 4.804804804804805e-08,
      "loss": 0.3421,
      "step": 33140
    },
    {
      "epoch": 99.55,
      "grad_norm": 4.4157938957214355,
      "learning_rate": 4.504504504504505e-08,
      "loss": 0.341,
      "step": 33150
    },
    {
      "epoch": 99.58,
      "grad_norm": 6.63731050491333,
      "learning_rate": 4.2042042042042044e-08,
      "loss": 0.3427,
      "step": 33160
    },
    {
      "epoch": 99.61,
      "grad_norm": 4.230950355529785,
      "learning_rate": 3.903903903903904e-08,
      "loss": 0.329,
      "step": 33170
    },
    {
      "epoch": 99.64,
      "grad_norm": 4.978404521942139,
      "learning_rate": 3.603603603603604e-08,
      "loss": 0.3743,
      "step": 33180
    },
    {
      "epoch": 99.67,
      "grad_norm": 4.710839748382568,
      "learning_rate": 3.303303303303304e-08,
      "loss": 0.3405,
      "step": 33190
    },
    {
      "epoch": 99.7,
      "grad_norm": 4.96866512298584,
      "learning_rate": 3.003003003003003e-08,
      "loss": 0.32,
      "step": 33200
    },
    {
      "epoch": 99.73,
      "grad_norm": 4.506679534912109,
      "learning_rate": 2.702702702702703e-08,
      "loss": 0.3407,
      "step": 33210
    },
    {
      "epoch": 99.76,
      "grad_norm": 3.569267749786377,
      "learning_rate": 2.4024024024024027e-08,
      "loss": 0.2896,
      "step": 33220
    },
    {
      "epoch": 99.79,
      "grad_norm": 5.6849894523620605,
      "learning_rate": 2.1021021021021022e-08,
      "loss": 0.3784,
      "step": 33230
    },
    {
      "epoch": 99.82,
      "grad_norm": 4.4942708015441895,
      "learning_rate": 1.801801801801802e-08,
      "loss": 0.3281,
      "step": 33240
    },
    {
      "epoch": 99.85,
      "grad_norm": 3.4201173782348633,
      "learning_rate": 1.5015015015015016e-08,
      "loss": 0.3133,
      "step": 33250
    },
    {
      "epoch": 99.88,
      "grad_norm": 3.024629592895508,
      "learning_rate": 1.2012012012012013e-08,
      "loss": 0.3468,
      "step": 33260
    },
    {
      "epoch": 99.91,
      "grad_norm": 3.358059883117676,
      "learning_rate": 9.00900900900901e-09,
      "loss": 0.3192,
      "step": 33270
    },
    {
      "epoch": 99.94,
      "grad_norm": 4.0538411140441895,
      "learning_rate": 6.006006006006007e-09,
      "loss": 0.3465,
      "step": 33280
    },
    {
      "epoch": 99.97,
      "grad_norm": 3.809311866760254,
      "learning_rate": 3.0030030030030033e-09,
      "loss": 0.3481,
      "step": 33290
    },
    {
      "epoch": 100.0,
      "grad_norm": 56.771636962890625,
      "learning_rate": 0.0,
      "loss": 0.3582,
      "step": 33300
    },
    {
      "epoch": 100.0,
      "eval_accuracy": 0.9197,
      "eval_loss": 0.31422850489616394,
      "eval_runtime": 26.4136,
      "eval_samples_per_second": 378.593,
      "eval_steps_per_second": 1.514,
      "step": 33300
    },
    {
      "epoch": 100.0,
      "step": 33300,
      "total_flos": 3.296302301859836e+20,
      "train_loss": 0.7083145791847069,
      "train_runtime": 34418.8887,
      "train_samples_per_second": 123.479,
      "train_steps_per_second": 0.967
    }
  ],
  "logging_steps": 10,
  "max_steps": 33300,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 100,
  "save_steps": 500,
  "total_flos": 3.296302301859836e+20,
  "train_batch_size": 128,
  "trial_name": null,
  "trial_params": null
}